From aa4d426b4d3527d7e166df1a05058c9a4a0f6683 Mon Sep 17 00:00:00 2001 From: Wojtek Kosior Date: Fri, 30 Apr 2021 00:33:56 +0200 Subject: initial/final commit --- openssl-1.1.0h/crypto/aes/aes_cbc.c | 24 + openssl-1.1.0h/crypto/aes/aes_cfb.c | 43 + openssl-1.1.0h/crypto/aes/aes_core.c | 1367 ++++++ openssl-1.1.0h/crypto/aes/aes_ecb.c | 26 + openssl-1.1.0h/crypto/aes/aes_ige.c | 284 ++ openssl-1.1.0h/crypto/aes/aes_locl.h | 42 + openssl-1.1.0h/crypto/aes/aes_misc.c | 21 + openssl-1.1.0h/crypto/aes/aes_ofb.c | 19 + openssl-1.1.0h/crypto/aes/aes_wrap.c | 27 + openssl-1.1.0h/crypto/aes/aes_x86core.c | 1075 +++++ openssl-1.1.0h/crypto/aes/asm/aes-586.pl | 3000 ++++++++++++ openssl-1.1.0h/crypto/aes/asm/aes-armv4.pl | 1245 +++++ openssl-1.1.0h/crypto/aes/asm/aes-c64xplus.pl | 1382 ++++++ openssl-1.1.0h/crypto/aes/asm/aes-ia64.S | 1130 +++++ openssl-1.1.0h/crypto/aes/asm/aes-mips.pl | 2131 +++++++++ openssl-1.1.0h/crypto/aes/asm/aes-parisc.pl | 1029 ++++ openssl-1.1.0h/crypto/aes/asm/aes-ppc.pl | 1459 ++++++ openssl-1.1.0h/crypto/aes/asm/aes-s390x.pl | 2226 +++++++++ openssl-1.1.0h/crypto/aes/asm/aes-sparcv9.pl | 1192 +++++ openssl-1.1.0h/crypto/aes/asm/aes-x86_64.pl | 2820 +++++++++++ openssl-1.1.0h/crypto/aes/asm/aesfx-sparcv9.pl | 1270 +++++ openssl-1.1.0h/crypto/aes/asm/aesni-mb-x86_64.pl | 1402 ++++++ openssl-1.1.0h/crypto/aes/asm/aesni-sha1-x86_64.pl | 2072 ++++++++ .../crypto/aes/asm/aesni-sha256-x86_64.pl | 1717 +++++++ openssl-1.1.0h/crypto/aes/asm/aesni-x86.pl | 3413 +++++++++++++ openssl-1.1.0h/crypto/aes/asm/aesni-x86_64.pl | 5060 ++++++++++++++++++++ openssl-1.1.0h/crypto/aes/asm/aesp8-ppc.pl | 3805 +++++++++++++++ openssl-1.1.0h/crypto/aes/asm/aest4-sparcv9.pl | 929 ++++ openssl-1.1.0h/crypto/aes/asm/aesv8-armx.pl | 1008 ++++ openssl-1.1.0h/crypto/aes/asm/bsaes-armv7.pl | 2495 ++++++++++ openssl-1.1.0h/crypto/aes/asm/bsaes-x86_64.pl | 3111 ++++++++++++ openssl-1.1.0h/crypto/aes/asm/vpaes-armv8.pl | 1259 +++++ openssl-1.1.0h/crypto/aes/asm/vpaes-ppc.pl | 1594 ++++++ openssl-1.1.0h/crypto/aes/asm/vpaes-x86.pl | 916 ++++ openssl-1.1.0h/crypto/aes/asm/vpaes-x86_64.pl | 1215 +++++ openssl-1.1.0h/crypto/aes/build.info | 57 + 36 files changed, 51865 insertions(+) create mode 100644 openssl-1.1.0h/crypto/aes/aes_cbc.c create mode 100644 openssl-1.1.0h/crypto/aes/aes_cfb.c create mode 100644 openssl-1.1.0h/crypto/aes/aes_core.c create mode 100644 openssl-1.1.0h/crypto/aes/aes_ecb.c create mode 100644 openssl-1.1.0h/crypto/aes/aes_ige.c create mode 100644 openssl-1.1.0h/crypto/aes/aes_locl.h create mode 100644 openssl-1.1.0h/crypto/aes/aes_misc.c create mode 100644 openssl-1.1.0h/crypto/aes/aes_ofb.c create mode 100644 openssl-1.1.0h/crypto/aes/aes_wrap.c create mode 100644 openssl-1.1.0h/crypto/aes/aes_x86core.c create mode 100755 openssl-1.1.0h/crypto/aes/asm/aes-586.pl create mode 100644 openssl-1.1.0h/crypto/aes/asm/aes-armv4.pl create mode 100644 openssl-1.1.0h/crypto/aes/asm/aes-c64xplus.pl create mode 100644 openssl-1.1.0h/crypto/aes/asm/aes-ia64.S create mode 100644 openssl-1.1.0h/crypto/aes/asm/aes-mips.pl create mode 100644 openssl-1.1.0h/crypto/aes/asm/aes-parisc.pl create mode 100644 openssl-1.1.0h/crypto/aes/asm/aes-ppc.pl create mode 100644 openssl-1.1.0h/crypto/aes/asm/aes-s390x.pl create mode 100755 openssl-1.1.0h/crypto/aes/asm/aes-sparcv9.pl create mode 100755 openssl-1.1.0h/crypto/aes/asm/aes-x86_64.pl create mode 100644 openssl-1.1.0h/crypto/aes/asm/aesfx-sparcv9.pl create mode 100644 openssl-1.1.0h/crypto/aes/asm/aesni-mb-x86_64.pl create mode 100644 openssl-1.1.0h/crypto/aes/asm/aesni-sha1-x86_64.pl create mode 100644 openssl-1.1.0h/crypto/aes/asm/aesni-sha256-x86_64.pl create mode 100644 openssl-1.1.0h/crypto/aes/asm/aesni-x86.pl create mode 100644 openssl-1.1.0h/crypto/aes/asm/aesni-x86_64.pl create mode 100755 openssl-1.1.0h/crypto/aes/asm/aesp8-ppc.pl create mode 100644 openssl-1.1.0h/crypto/aes/asm/aest4-sparcv9.pl create mode 100755 openssl-1.1.0h/crypto/aes/asm/aesv8-armx.pl create mode 100644 openssl-1.1.0h/crypto/aes/asm/bsaes-armv7.pl create mode 100644 openssl-1.1.0h/crypto/aes/asm/bsaes-x86_64.pl create mode 100755 openssl-1.1.0h/crypto/aes/asm/vpaes-armv8.pl create mode 100644 openssl-1.1.0h/crypto/aes/asm/vpaes-ppc.pl create mode 100644 openssl-1.1.0h/crypto/aes/asm/vpaes-x86.pl create mode 100644 openssl-1.1.0h/crypto/aes/asm/vpaes-x86_64.pl create mode 100644 openssl-1.1.0h/crypto/aes/build.info (limited to 'openssl-1.1.0h/crypto/aes') diff --git a/openssl-1.1.0h/crypto/aes/aes_cbc.c b/openssl-1.1.0h/crypto/aes/aes_cbc.c new file mode 100644 index 0000000..342841f --- /dev/null +++ b/openssl-1.1.0h/crypto/aes/aes_cbc.c @@ -0,0 +1,24 @@ +/* + * Copyright 2002-2016 The OpenSSL Project Authors. All Rights Reserved. + * + * Licensed under the OpenSSL license (the "License"). You may not use + * this file except in compliance with the License. You can obtain a copy + * in the file LICENSE in the source distribution or at + * https://www.openssl.org/source/license.html + */ + +#include +#include + +void AES_cbc_encrypt(const unsigned char *in, unsigned char *out, + size_t len, const AES_KEY *key, + unsigned char *ivec, const int enc) +{ + + if (enc) + CRYPTO_cbc128_encrypt(in, out, len, key, ivec, + (block128_f) AES_encrypt); + else + CRYPTO_cbc128_decrypt(in, out, len, key, ivec, + (block128_f) AES_decrypt); +} diff --git a/openssl-1.1.0h/crypto/aes/aes_cfb.c b/openssl-1.1.0h/crypto/aes/aes_cfb.c new file mode 100644 index 0000000..f010e3c --- /dev/null +++ b/openssl-1.1.0h/crypto/aes/aes_cfb.c @@ -0,0 +1,43 @@ +/* + * Copyright 2002-2016 The OpenSSL Project Authors. All Rights Reserved. + * + * Licensed under the OpenSSL license (the "License"). You may not use + * this file except in compliance with the License. You can obtain a copy + * in the file LICENSE in the source distribution or at + * https://www.openssl.org/source/license.html + */ + +#include +#include + +/* + * The input and output encrypted as though 128bit cfb mode is being used. + * The extra state information to record how much of the 128bit block we have + * used is contained in *num; + */ + +void AES_cfb128_encrypt(const unsigned char *in, unsigned char *out, + size_t length, const AES_KEY *key, + unsigned char *ivec, int *num, const int enc) +{ + + CRYPTO_cfb128_encrypt(in, out, length, key, ivec, num, enc, + (block128_f) AES_encrypt); +} + +/* N.B. This expects the input to be packed, MS bit first */ +void AES_cfb1_encrypt(const unsigned char *in, unsigned char *out, + size_t length, const AES_KEY *key, + unsigned char *ivec, int *num, const int enc) +{ + CRYPTO_cfb128_1_encrypt(in, out, length, key, ivec, num, enc, + (block128_f) AES_encrypt); +} + +void AES_cfb8_encrypt(const unsigned char *in, unsigned char *out, + size_t length, const AES_KEY *key, + unsigned char *ivec, int *num, const int enc) +{ + CRYPTO_cfb128_8_encrypt(in, out, length, key, ivec, num, enc, + (block128_f) AES_encrypt); +} diff --git a/openssl-1.1.0h/crypto/aes/aes_core.c b/openssl-1.1.0h/crypto/aes/aes_core.c new file mode 100644 index 0000000..bd5c779 --- /dev/null +++ b/openssl-1.1.0h/crypto/aes/aes_core.c @@ -0,0 +1,1367 @@ +/* + * Copyright 2002-2016 The OpenSSL Project Authors. All Rights Reserved. + * + * Licensed under the OpenSSL license (the "License"). You may not use + * this file except in compliance with the License. You can obtain a copy + * in the file LICENSE in the source distribution or at + * https://www.openssl.org/source/license.html + */ + +/** + * rijndael-alg-fst.c + * + * @version 3.0 (December 2000) + * + * Optimised ANSI C code for the Rijndael cipher (now AES) + * + * @author Vincent Rijmen + * @author Antoon Bosselaers + * @author Paulo Barreto + * + * This code is hereby placed in the public domain. + * + * THIS SOFTWARE IS PROVIDED BY THE AUTHORS ''AS IS'' AND ANY EXPRESS + * OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED + * WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE + * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHORS OR CONTRIBUTORS BE + * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR + * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF + * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR + * BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, + * WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE + * OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, + * EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. + */ + +/* Note: rewritten a little bit to provide error control and an OpenSSL- + compatible API */ + +#include + +#include +#include +#include +#include "aes_locl.h" + +#ifndef AES_ASM +/*- +Te0[x] = S [x].[02, 01, 01, 03]; +Te1[x] = S [x].[03, 02, 01, 01]; +Te2[x] = S [x].[01, 03, 02, 01]; +Te3[x] = S [x].[01, 01, 03, 02]; + +Td0[x] = Si[x].[0e, 09, 0d, 0b]; +Td1[x] = Si[x].[0b, 0e, 09, 0d]; +Td2[x] = Si[x].[0d, 0b, 0e, 09]; +Td3[x] = Si[x].[09, 0d, 0b, 0e]; +Td4[x] = Si[x].[01]; +*/ + +static const u32 Te0[256] = { + 0xc66363a5U, 0xf87c7c84U, 0xee777799U, 0xf67b7b8dU, + 0xfff2f20dU, 0xd66b6bbdU, 0xde6f6fb1U, 0x91c5c554U, + 0x60303050U, 0x02010103U, 0xce6767a9U, 0x562b2b7dU, + 0xe7fefe19U, 0xb5d7d762U, 0x4dababe6U, 0xec76769aU, + 0x8fcaca45U, 0x1f82829dU, 0x89c9c940U, 0xfa7d7d87U, + 0xeffafa15U, 0xb25959ebU, 0x8e4747c9U, 0xfbf0f00bU, + 0x41adadecU, 0xb3d4d467U, 0x5fa2a2fdU, 0x45afafeaU, + 0x239c9cbfU, 0x53a4a4f7U, 0xe4727296U, 0x9bc0c05bU, + 0x75b7b7c2U, 0xe1fdfd1cU, 0x3d9393aeU, 0x4c26266aU, + 0x6c36365aU, 0x7e3f3f41U, 0xf5f7f702U, 0x83cccc4fU, + 0x6834345cU, 0x51a5a5f4U, 0xd1e5e534U, 0xf9f1f108U, + 0xe2717193U, 0xabd8d873U, 0x62313153U, 0x2a15153fU, + 0x0804040cU, 0x95c7c752U, 0x46232365U, 0x9dc3c35eU, + 0x30181828U, 0x379696a1U, 0x0a05050fU, 0x2f9a9ab5U, + 0x0e070709U, 0x24121236U, 0x1b80809bU, 0xdfe2e23dU, + 0xcdebeb26U, 0x4e272769U, 0x7fb2b2cdU, 0xea75759fU, + 0x1209091bU, 0x1d83839eU, 0x582c2c74U, 0x341a1a2eU, + 0x361b1b2dU, 0xdc6e6eb2U, 0xb45a5aeeU, 0x5ba0a0fbU, + 0xa45252f6U, 0x763b3b4dU, 0xb7d6d661U, 0x7db3b3ceU, + 0x5229297bU, 0xdde3e33eU, 0x5e2f2f71U, 0x13848497U, + 0xa65353f5U, 0xb9d1d168U, 0x00000000U, 0xc1eded2cU, + 0x40202060U, 0xe3fcfc1fU, 0x79b1b1c8U, 0xb65b5bedU, + 0xd46a6abeU, 0x8dcbcb46U, 0x67bebed9U, 0x7239394bU, + 0x944a4adeU, 0x984c4cd4U, 0xb05858e8U, 0x85cfcf4aU, + 0xbbd0d06bU, 0xc5efef2aU, 0x4faaaae5U, 0xedfbfb16U, + 0x864343c5U, 0x9a4d4dd7U, 0x66333355U, 0x11858594U, + 0x8a4545cfU, 0xe9f9f910U, 0x04020206U, 0xfe7f7f81U, + 0xa05050f0U, 0x783c3c44U, 0x259f9fbaU, 0x4ba8a8e3U, + 0xa25151f3U, 0x5da3a3feU, 0x804040c0U, 0x058f8f8aU, + 0x3f9292adU, 0x219d9dbcU, 0x70383848U, 0xf1f5f504U, + 0x63bcbcdfU, 0x77b6b6c1U, 0xafdada75U, 0x42212163U, + 0x20101030U, 0xe5ffff1aU, 0xfdf3f30eU, 0xbfd2d26dU, + 0x81cdcd4cU, 0x180c0c14U, 0x26131335U, 0xc3ecec2fU, + 0xbe5f5fe1U, 0x359797a2U, 0x884444ccU, 0x2e171739U, + 0x93c4c457U, 0x55a7a7f2U, 0xfc7e7e82U, 0x7a3d3d47U, + 0xc86464acU, 0xba5d5de7U, 0x3219192bU, 0xe6737395U, + 0xc06060a0U, 0x19818198U, 0x9e4f4fd1U, 0xa3dcdc7fU, + 0x44222266U, 0x542a2a7eU, 0x3b9090abU, 0x0b888883U, + 0x8c4646caU, 0xc7eeee29U, 0x6bb8b8d3U, 0x2814143cU, + 0xa7dede79U, 0xbc5e5ee2U, 0x160b0b1dU, 0xaddbdb76U, + 0xdbe0e03bU, 0x64323256U, 0x743a3a4eU, 0x140a0a1eU, + 0x924949dbU, 0x0c06060aU, 0x4824246cU, 0xb85c5ce4U, + 0x9fc2c25dU, 0xbdd3d36eU, 0x43acacefU, 0xc46262a6U, + 0x399191a8U, 0x319595a4U, 0xd3e4e437U, 0xf279798bU, + 0xd5e7e732U, 0x8bc8c843U, 0x6e373759U, 0xda6d6db7U, + 0x018d8d8cU, 0xb1d5d564U, 0x9c4e4ed2U, 0x49a9a9e0U, + 0xd86c6cb4U, 0xac5656faU, 0xf3f4f407U, 0xcfeaea25U, + 0xca6565afU, 0xf47a7a8eU, 0x47aeaee9U, 0x10080818U, + 0x6fbabad5U, 0xf0787888U, 0x4a25256fU, 0x5c2e2e72U, + 0x381c1c24U, 0x57a6a6f1U, 0x73b4b4c7U, 0x97c6c651U, + 0xcbe8e823U, 0xa1dddd7cU, 0xe874749cU, 0x3e1f1f21U, + 0x964b4bddU, 0x61bdbddcU, 0x0d8b8b86U, 0x0f8a8a85U, + 0xe0707090U, 0x7c3e3e42U, 0x71b5b5c4U, 0xcc6666aaU, + 0x904848d8U, 0x06030305U, 0xf7f6f601U, 0x1c0e0e12U, + 0xc26161a3U, 0x6a35355fU, 0xae5757f9U, 0x69b9b9d0U, + 0x17868691U, 0x99c1c158U, 0x3a1d1d27U, 0x279e9eb9U, + 0xd9e1e138U, 0xebf8f813U, 0x2b9898b3U, 0x22111133U, + 0xd26969bbU, 0xa9d9d970U, 0x078e8e89U, 0x339494a7U, + 0x2d9b9bb6U, 0x3c1e1e22U, 0x15878792U, 0xc9e9e920U, + 0x87cece49U, 0xaa5555ffU, 0x50282878U, 0xa5dfdf7aU, + 0x038c8c8fU, 0x59a1a1f8U, 0x09898980U, 0x1a0d0d17U, + 0x65bfbfdaU, 0xd7e6e631U, 0x844242c6U, 0xd06868b8U, + 0x824141c3U, 0x299999b0U, 0x5a2d2d77U, 0x1e0f0f11U, + 0x7bb0b0cbU, 0xa85454fcU, 0x6dbbbbd6U, 0x2c16163aU, +}; +static const u32 Te1[256] = { + 0xa5c66363U, 0x84f87c7cU, 0x99ee7777U, 0x8df67b7bU, + 0x0dfff2f2U, 0xbdd66b6bU, 0xb1de6f6fU, 0x5491c5c5U, + 0x50603030U, 0x03020101U, 0xa9ce6767U, 0x7d562b2bU, + 0x19e7fefeU, 0x62b5d7d7U, 0xe64dababU, 0x9aec7676U, + 0x458fcacaU, 0x9d1f8282U, 0x4089c9c9U, 0x87fa7d7dU, + 0x15effafaU, 0xebb25959U, 0xc98e4747U, 0x0bfbf0f0U, + 0xec41adadU, 0x67b3d4d4U, 0xfd5fa2a2U, 0xea45afafU, + 0xbf239c9cU, 0xf753a4a4U, 0x96e47272U, 0x5b9bc0c0U, + 0xc275b7b7U, 0x1ce1fdfdU, 0xae3d9393U, 0x6a4c2626U, + 0x5a6c3636U, 0x417e3f3fU, 0x02f5f7f7U, 0x4f83ccccU, + 0x5c683434U, 0xf451a5a5U, 0x34d1e5e5U, 0x08f9f1f1U, + 0x93e27171U, 0x73abd8d8U, 0x53623131U, 0x3f2a1515U, + 0x0c080404U, 0x5295c7c7U, 0x65462323U, 0x5e9dc3c3U, + 0x28301818U, 0xa1379696U, 0x0f0a0505U, 0xb52f9a9aU, + 0x090e0707U, 0x36241212U, 0x9b1b8080U, 0x3ddfe2e2U, + 0x26cdebebU, 0x694e2727U, 0xcd7fb2b2U, 0x9fea7575U, + 0x1b120909U, 0x9e1d8383U, 0x74582c2cU, 0x2e341a1aU, + 0x2d361b1bU, 0xb2dc6e6eU, 0xeeb45a5aU, 0xfb5ba0a0U, + 0xf6a45252U, 0x4d763b3bU, 0x61b7d6d6U, 0xce7db3b3U, + 0x7b522929U, 0x3edde3e3U, 0x715e2f2fU, 0x97138484U, + 0xf5a65353U, 0x68b9d1d1U, 0x00000000U, 0x2cc1ededU, + 0x60402020U, 0x1fe3fcfcU, 0xc879b1b1U, 0xedb65b5bU, + 0xbed46a6aU, 0x468dcbcbU, 0xd967bebeU, 0x4b723939U, + 0xde944a4aU, 0xd4984c4cU, 0xe8b05858U, 0x4a85cfcfU, + 0x6bbbd0d0U, 0x2ac5efefU, 0xe54faaaaU, 0x16edfbfbU, + 0xc5864343U, 0xd79a4d4dU, 0x55663333U, 0x94118585U, + 0xcf8a4545U, 0x10e9f9f9U, 0x06040202U, 0x81fe7f7fU, + 0xf0a05050U, 0x44783c3cU, 0xba259f9fU, 0xe34ba8a8U, + 0xf3a25151U, 0xfe5da3a3U, 0xc0804040U, 0x8a058f8fU, + 0xad3f9292U, 0xbc219d9dU, 0x48703838U, 0x04f1f5f5U, + 0xdf63bcbcU, 0xc177b6b6U, 0x75afdadaU, 0x63422121U, + 0x30201010U, 0x1ae5ffffU, 0x0efdf3f3U, 0x6dbfd2d2U, + 0x4c81cdcdU, 0x14180c0cU, 0x35261313U, 0x2fc3ececU, + 0xe1be5f5fU, 0xa2359797U, 0xcc884444U, 0x392e1717U, + 0x5793c4c4U, 0xf255a7a7U, 0x82fc7e7eU, 0x477a3d3dU, + 0xacc86464U, 0xe7ba5d5dU, 0x2b321919U, 0x95e67373U, + 0xa0c06060U, 0x98198181U, 0xd19e4f4fU, 0x7fa3dcdcU, + 0x66442222U, 0x7e542a2aU, 0xab3b9090U, 0x830b8888U, + 0xca8c4646U, 0x29c7eeeeU, 0xd36bb8b8U, 0x3c281414U, + 0x79a7dedeU, 0xe2bc5e5eU, 0x1d160b0bU, 0x76addbdbU, + 0x3bdbe0e0U, 0x56643232U, 0x4e743a3aU, 0x1e140a0aU, + 0xdb924949U, 0x0a0c0606U, 0x6c482424U, 0xe4b85c5cU, + 0x5d9fc2c2U, 0x6ebdd3d3U, 0xef43acacU, 0xa6c46262U, + 0xa8399191U, 0xa4319595U, 0x37d3e4e4U, 0x8bf27979U, + 0x32d5e7e7U, 0x438bc8c8U, 0x596e3737U, 0xb7da6d6dU, + 0x8c018d8dU, 0x64b1d5d5U, 0xd29c4e4eU, 0xe049a9a9U, + 0xb4d86c6cU, 0xfaac5656U, 0x07f3f4f4U, 0x25cfeaeaU, + 0xafca6565U, 0x8ef47a7aU, 0xe947aeaeU, 0x18100808U, + 0xd56fbabaU, 0x88f07878U, 0x6f4a2525U, 0x725c2e2eU, + 0x24381c1cU, 0xf157a6a6U, 0xc773b4b4U, 0x5197c6c6U, + 0x23cbe8e8U, 0x7ca1ddddU, 0x9ce87474U, 0x213e1f1fU, + 0xdd964b4bU, 0xdc61bdbdU, 0x860d8b8bU, 0x850f8a8aU, + 0x90e07070U, 0x427c3e3eU, 0xc471b5b5U, 0xaacc6666U, + 0xd8904848U, 0x05060303U, 0x01f7f6f6U, 0x121c0e0eU, + 0xa3c26161U, 0x5f6a3535U, 0xf9ae5757U, 0xd069b9b9U, + 0x91178686U, 0x5899c1c1U, 0x273a1d1dU, 0xb9279e9eU, + 0x38d9e1e1U, 0x13ebf8f8U, 0xb32b9898U, 0x33221111U, + 0xbbd26969U, 0x70a9d9d9U, 0x89078e8eU, 0xa7339494U, + 0xb62d9b9bU, 0x223c1e1eU, 0x92158787U, 0x20c9e9e9U, + 0x4987ceceU, 0xffaa5555U, 0x78502828U, 0x7aa5dfdfU, + 0x8f038c8cU, 0xf859a1a1U, 0x80098989U, 0x171a0d0dU, + 0xda65bfbfU, 0x31d7e6e6U, 0xc6844242U, 0xb8d06868U, + 0xc3824141U, 0xb0299999U, 0x775a2d2dU, 0x111e0f0fU, + 0xcb7bb0b0U, 0xfca85454U, 0xd66dbbbbU, 0x3a2c1616U, +}; +static const u32 Te2[256] = { + 0x63a5c663U, 0x7c84f87cU, 0x7799ee77U, 0x7b8df67bU, + 0xf20dfff2U, 0x6bbdd66bU, 0x6fb1de6fU, 0xc55491c5U, + 0x30506030U, 0x01030201U, 0x67a9ce67U, 0x2b7d562bU, + 0xfe19e7feU, 0xd762b5d7U, 0xabe64dabU, 0x769aec76U, + 0xca458fcaU, 0x829d1f82U, 0xc94089c9U, 0x7d87fa7dU, + 0xfa15effaU, 0x59ebb259U, 0x47c98e47U, 0xf00bfbf0U, + 0xadec41adU, 0xd467b3d4U, 0xa2fd5fa2U, 0xafea45afU, + 0x9cbf239cU, 0xa4f753a4U, 0x7296e472U, 0xc05b9bc0U, + 0xb7c275b7U, 0xfd1ce1fdU, 0x93ae3d93U, 0x266a4c26U, + 0x365a6c36U, 0x3f417e3fU, 0xf702f5f7U, 0xcc4f83ccU, + 0x345c6834U, 0xa5f451a5U, 0xe534d1e5U, 0xf108f9f1U, + 0x7193e271U, 0xd873abd8U, 0x31536231U, 0x153f2a15U, + 0x040c0804U, 0xc75295c7U, 0x23654623U, 0xc35e9dc3U, + 0x18283018U, 0x96a13796U, 0x050f0a05U, 0x9ab52f9aU, + 0x07090e07U, 0x12362412U, 0x809b1b80U, 0xe23ddfe2U, + 0xeb26cdebU, 0x27694e27U, 0xb2cd7fb2U, 0x759fea75U, + 0x091b1209U, 0x839e1d83U, 0x2c74582cU, 0x1a2e341aU, + 0x1b2d361bU, 0x6eb2dc6eU, 0x5aeeb45aU, 0xa0fb5ba0U, + 0x52f6a452U, 0x3b4d763bU, 0xd661b7d6U, 0xb3ce7db3U, + 0x297b5229U, 0xe33edde3U, 0x2f715e2fU, 0x84971384U, + 0x53f5a653U, 0xd168b9d1U, 0x00000000U, 0xed2cc1edU, + 0x20604020U, 0xfc1fe3fcU, 0xb1c879b1U, 0x5bedb65bU, + 0x6abed46aU, 0xcb468dcbU, 0xbed967beU, 0x394b7239U, + 0x4ade944aU, 0x4cd4984cU, 0x58e8b058U, 0xcf4a85cfU, + 0xd06bbbd0U, 0xef2ac5efU, 0xaae54faaU, 0xfb16edfbU, + 0x43c58643U, 0x4dd79a4dU, 0x33556633U, 0x85941185U, + 0x45cf8a45U, 0xf910e9f9U, 0x02060402U, 0x7f81fe7fU, + 0x50f0a050U, 0x3c44783cU, 0x9fba259fU, 0xa8e34ba8U, + 0x51f3a251U, 0xa3fe5da3U, 0x40c08040U, 0x8f8a058fU, + 0x92ad3f92U, 0x9dbc219dU, 0x38487038U, 0xf504f1f5U, + 0xbcdf63bcU, 0xb6c177b6U, 0xda75afdaU, 0x21634221U, + 0x10302010U, 0xff1ae5ffU, 0xf30efdf3U, 0xd26dbfd2U, + 0xcd4c81cdU, 0x0c14180cU, 0x13352613U, 0xec2fc3ecU, + 0x5fe1be5fU, 0x97a23597U, 0x44cc8844U, 0x17392e17U, + 0xc45793c4U, 0xa7f255a7U, 0x7e82fc7eU, 0x3d477a3dU, + 0x64acc864U, 0x5de7ba5dU, 0x192b3219U, 0x7395e673U, + 0x60a0c060U, 0x81981981U, 0x4fd19e4fU, 0xdc7fa3dcU, + 0x22664422U, 0x2a7e542aU, 0x90ab3b90U, 0x88830b88U, + 0x46ca8c46U, 0xee29c7eeU, 0xb8d36bb8U, 0x143c2814U, + 0xde79a7deU, 0x5ee2bc5eU, 0x0b1d160bU, 0xdb76addbU, + 0xe03bdbe0U, 0x32566432U, 0x3a4e743aU, 0x0a1e140aU, + 0x49db9249U, 0x060a0c06U, 0x246c4824U, 0x5ce4b85cU, + 0xc25d9fc2U, 0xd36ebdd3U, 0xacef43acU, 0x62a6c462U, + 0x91a83991U, 0x95a43195U, 0xe437d3e4U, 0x798bf279U, + 0xe732d5e7U, 0xc8438bc8U, 0x37596e37U, 0x6db7da6dU, + 0x8d8c018dU, 0xd564b1d5U, 0x4ed29c4eU, 0xa9e049a9U, + 0x6cb4d86cU, 0x56faac56U, 0xf407f3f4U, 0xea25cfeaU, + 0x65afca65U, 0x7a8ef47aU, 0xaee947aeU, 0x08181008U, + 0xbad56fbaU, 0x7888f078U, 0x256f4a25U, 0x2e725c2eU, + 0x1c24381cU, 0xa6f157a6U, 0xb4c773b4U, 0xc65197c6U, + 0xe823cbe8U, 0xdd7ca1ddU, 0x749ce874U, 0x1f213e1fU, + 0x4bdd964bU, 0xbddc61bdU, 0x8b860d8bU, 0x8a850f8aU, + 0x7090e070U, 0x3e427c3eU, 0xb5c471b5U, 0x66aacc66U, + 0x48d89048U, 0x03050603U, 0xf601f7f6U, 0x0e121c0eU, + 0x61a3c261U, 0x355f6a35U, 0x57f9ae57U, 0xb9d069b9U, + 0x86911786U, 0xc15899c1U, 0x1d273a1dU, 0x9eb9279eU, + 0xe138d9e1U, 0xf813ebf8U, 0x98b32b98U, 0x11332211U, + 0x69bbd269U, 0xd970a9d9U, 0x8e89078eU, 0x94a73394U, + 0x9bb62d9bU, 0x1e223c1eU, 0x87921587U, 0xe920c9e9U, + 0xce4987ceU, 0x55ffaa55U, 0x28785028U, 0xdf7aa5dfU, + 0x8c8f038cU, 0xa1f859a1U, 0x89800989U, 0x0d171a0dU, + 0xbfda65bfU, 0xe631d7e6U, 0x42c68442U, 0x68b8d068U, + 0x41c38241U, 0x99b02999U, 0x2d775a2dU, 0x0f111e0fU, + 0xb0cb7bb0U, 0x54fca854U, 0xbbd66dbbU, 0x163a2c16U, +}; +static const u32 Te3[256] = { + 0x6363a5c6U, 0x7c7c84f8U, 0x777799eeU, 0x7b7b8df6U, + 0xf2f20dffU, 0x6b6bbdd6U, 0x6f6fb1deU, 0xc5c55491U, + 0x30305060U, 0x01010302U, 0x6767a9ceU, 0x2b2b7d56U, + 0xfefe19e7U, 0xd7d762b5U, 0xababe64dU, 0x76769aecU, + 0xcaca458fU, 0x82829d1fU, 0xc9c94089U, 0x7d7d87faU, + 0xfafa15efU, 0x5959ebb2U, 0x4747c98eU, 0xf0f00bfbU, + 0xadadec41U, 0xd4d467b3U, 0xa2a2fd5fU, 0xafafea45U, + 0x9c9cbf23U, 0xa4a4f753U, 0x727296e4U, 0xc0c05b9bU, + 0xb7b7c275U, 0xfdfd1ce1U, 0x9393ae3dU, 0x26266a4cU, + 0x36365a6cU, 0x3f3f417eU, 0xf7f702f5U, 0xcccc4f83U, + 0x34345c68U, 0xa5a5f451U, 0xe5e534d1U, 0xf1f108f9U, + 0x717193e2U, 0xd8d873abU, 0x31315362U, 0x15153f2aU, + 0x04040c08U, 0xc7c75295U, 0x23236546U, 0xc3c35e9dU, + 0x18182830U, 0x9696a137U, 0x05050f0aU, 0x9a9ab52fU, + 0x0707090eU, 0x12123624U, 0x80809b1bU, 0xe2e23ddfU, + 0xebeb26cdU, 0x2727694eU, 0xb2b2cd7fU, 0x75759feaU, + 0x09091b12U, 0x83839e1dU, 0x2c2c7458U, 0x1a1a2e34U, + 0x1b1b2d36U, 0x6e6eb2dcU, 0x5a5aeeb4U, 0xa0a0fb5bU, + 0x5252f6a4U, 0x3b3b4d76U, 0xd6d661b7U, 0xb3b3ce7dU, + 0x29297b52U, 0xe3e33eddU, 0x2f2f715eU, 0x84849713U, + 0x5353f5a6U, 0xd1d168b9U, 0x00000000U, 0xeded2cc1U, + 0x20206040U, 0xfcfc1fe3U, 0xb1b1c879U, 0x5b5bedb6U, + 0x6a6abed4U, 0xcbcb468dU, 0xbebed967U, 0x39394b72U, + 0x4a4ade94U, 0x4c4cd498U, 0x5858e8b0U, 0xcfcf4a85U, + 0xd0d06bbbU, 0xefef2ac5U, 0xaaaae54fU, 0xfbfb16edU, + 0x4343c586U, 0x4d4dd79aU, 0x33335566U, 0x85859411U, + 0x4545cf8aU, 0xf9f910e9U, 0x02020604U, 0x7f7f81feU, + 0x5050f0a0U, 0x3c3c4478U, 0x9f9fba25U, 0xa8a8e34bU, + 0x5151f3a2U, 0xa3a3fe5dU, 0x4040c080U, 0x8f8f8a05U, + 0x9292ad3fU, 0x9d9dbc21U, 0x38384870U, 0xf5f504f1U, + 0xbcbcdf63U, 0xb6b6c177U, 0xdada75afU, 0x21216342U, + 0x10103020U, 0xffff1ae5U, 0xf3f30efdU, 0xd2d26dbfU, + 0xcdcd4c81U, 0x0c0c1418U, 0x13133526U, 0xecec2fc3U, + 0x5f5fe1beU, 0x9797a235U, 0x4444cc88U, 0x1717392eU, + 0xc4c45793U, 0xa7a7f255U, 0x7e7e82fcU, 0x3d3d477aU, + 0x6464acc8U, 0x5d5de7baU, 0x19192b32U, 0x737395e6U, + 0x6060a0c0U, 0x81819819U, 0x4f4fd19eU, 0xdcdc7fa3U, + 0x22226644U, 0x2a2a7e54U, 0x9090ab3bU, 0x8888830bU, + 0x4646ca8cU, 0xeeee29c7U, 0xb8b8d36bU, 0x14143c28U, + 0xdede79a7U, 0x5e5ee2bcU, 0x0b0b1d16U, 0xdbdb76adU, + 0xe0e03bdbU, 0x32325664U, 0x3a3a4e74U, 0x0a0a1e14U, + 0x4949db92U, 0x06060a0cU, 0x24246c48U, 0x5c5ce4b8U, + 0xc2c25d9fU, 0xd3d36ebdU, 0xacacef43U, 0x6262a6c4U, + 0x9191a839U, 0x9595a431U, 0xe4e437d3U, 0x79798bf2U, + 0xe7e732d5U, 0xc8c8438bU, 0x3737596eU, 0x6d6db7daU, + 0x8d8d8c01U, 0xd5d564b1U, 0x4e4ed29cU, 0xa9a9e049U, + 0x6c6cb4d8U, 0x5656faacU, 0xf4f407f3U, 0xeaea25cfU, + 0x6565afcaU, 0x7a7a8ef4U, 0xaeaee947U, 0x08081810U, + 0xbabad56fU, 0x787888f0U, 0x25256f4aU, 0x2e2e725cU, + 0x1c1c2438U, 0xa6a6f157U, 0xb4b4c773U, 0xc6c65197U, + 0xe8e823cbU, 0xdddd7ca1U, 0x74749ce8U, 0x1f1f213eU, + 0x4b4bdd96U, 0xbdbddc61U, 0x8b8b860dU, 0x8a8a850fU, + 0x707090e0U, 0x3e3e427cU, 0xb5b5c471U, 0x6666aaccU, + 0x4848d890U, 0x03030506U, 0xf6f601f7U, 0x0e0e121cU, + 0x6161a3c2U, 0x35355f6aU, 0x5757f9aeU, 0xb9b9d069U, + 0x86869117U, 0xc1c15899U, 0x1d1d273aU, 0x9e9eb927U, + 0xe1e138d9U, 0xf8f813ebU, 0x9898b32bU, 0x11113322U, + 0x6969bbd2U, 0xd9d970a9U, 0x8e8e8907U, 0x9494a733U, + 0x9b9bb62dU, 0x1e1e223cU, 0x87879215U, 0xe9e920c9U, + 0xcece4987U, 0x5555ffaaU, 0x28287850U, 0xdfdf7aa5U, + 0x8c8c8f03U, 0xa1a1f859U, 0x89898009U, 0x0d0d171aU, + 0xbfbfda65U, 0xe6e631d7U, 0x4242c684U, 0x6868b8d0U, + 0x4141c382U, 0x9999b029U, 0x2d2d775aU, 0x0f0f111eU, + 0xb0b0cb7bU, 0x5454fca8U, 0xbbbbd66dU, 0x16163a2cU, +}; + +static const u32 Td0[256] = { + 0x51f4a750U, 0x7e416553U, 0x1a17a4c3U, 0x3a275e96U, + 0x3bab6bcbU, 0x1f9d45f1U, 0xacfa58abU, 0x4be30393U, + 0x2030fa55U, 0xad766df6U, 0x88cc7691U, 0xf5024c25U, + 0x4fe5d7fcU, 0xc52acbd7U, 0x26354480U, 0xb562a38fU, + 0xdeb15a49U, 0x25ba1b67U, 0x45ea0e98U, 0x5dfec0e1U, + 0xc32f7502U, 0x814cf012U, 0x8d4697a3U, 0x6bd3f9c6U, + 0x038f5fe7U, 0x15929c95U, 0xbf6d7aebU, 0x955259daU, + 0xd4be832dU, 0x587421d3U, 0x49e06929U, 0x8ec9c844U, + 0x75c2896aU, 0xf48e7978U, 0x99583e6bU, 0x27b971ddU, + 0xbee14fb6U, 0xf088ad17U, 0xc920ac66U, 0x7dce3ab4U, + 0x63df4a18U, 0xe51a3182U, 0x97513360U, 0x62537f45U, + 0xb16477e0U, 0xbb6bae84U, 0xfe81a01cU, 0xf9082b94U, + 0x70486858U, 0x8f45fd19U, 0x94de6c87U, 0x527bf8b7U, + 0xab73d323U, 0x724b02e2U, 0xe31f8f57U, 0x6655ab2aU, + 0xb2eb2807U, 0x2fb5c203U, 0x86c57b9aU, 0xd33708a5U, + 0x302887f2U, 0x23bfa5b2U, 0x02036abaU, 0xed16825cU, + 0x8acf1c2bU, 0xa779b492U, 0xf307f2f0U, 0x4e69e2a1U, + 0x65daf4cdU, 0x0605bed5U, 0xd134621fU, 0xc4a6fe8aU, + 0x342e539dU, 0xa2f355a0U, 0x058ae132U, 0xa4f6eb75U, + 0x0b83ec39U, 0x4060efaaU, 0x5e719f06U, 0xbd6e1051U, + 0x3e218af9U, 0x96dd063dU, 0xdd3e05aeU, 0x4de6bd46U, + 0x91548db5U, 0x71c45d05U, 0x0406d46fU, 0x605015ffU, + 0x1998fb24U, 0xd6bde997U, 0x894043ccU, 0x67d99e77U, + 0xb0e842bdU, 0x07898b88U, 0xe7195b38U, 0x79c8eedbU, + 0xa17c0a47U, 0x7c420fe9U, 0xf8841ec9U, 0x00000000U, + 0x09808683U, 0x322bed48U, 0x1e1170acU, 0x6c5a724eU, + 0xfd0efffbU, 0x0f853856U, 0x3daed51eU, 0x362d3927U, + 0x0a0fd964U, 0x685ca621U, 0x9b5b54d1U, 0x24362e3aU, + 0x0c0a67b1U, 0x9357e70fU, 0xb4ee96d2U, 0x1b9b919eU, + 0x80c0c54fU, 0x61dc20a2U, 0x5a774b69U, 0x1c121a16U, + 0xe293ba0aU, 0xc0a02ae5U, 0x3c22e043U, 0x121b171dU, + 0x0e090d0bU, 0xf28bc7adU, 0x2db6a8b9U, 0x141ea9c8U, + 0x57f11985U, 0xaf75074cU, 0xee99ddbbU, 0xa37f60fdU, + 0xf701269fU, 0x5c72f5bcU, 0x44663bc5U, 0x5bfb7e34U, + 0x8b432976U, 0xcb23c6dcU, 0xb6edfc68U, 0xb8e4f163U, + 0xd731dccaU, 0x42638510U, 0x13972240U, 0x84c61120U, + 0x854a247dU, 0xd2bb3df8U, 0xaef93211U, 0xc729a16dU, + 0x1d9e2f4bU, 0xdcb230f3U, 0x0d8652ecU, 0x77c1e3d0U, + 0x2bb3166cU, 0xa970b999U, 0x119448faU, 0x47e96422U, + 0xa8fc8cc4U, 0xa0f03f1aU, 0x567d2cd8U, 0x223390efU, + 0x87494ec7U, 0xd938d1c1U, 0x8ccaa2feU, 0x98d40b36U, + 0xa6f581cfU, 0xa57ade28U, 0xdab78e26U, 0x3fadbfa4U, + 0x2c3a9de4U, 0x5078920dU, 0x6a5fcc9bU, 0x547e4662U, + 0xf68d13c2U, 0x90d8b8e8U, 0x2e39f75eU, 0x82c3aff5U, + 0x9f5d80beU, 0x69d0937cU, 0x6fd52da9U, 0xcf2512b3U, + 0xc8ac993bU, 0x10187da7U, 0xe89c636eU, 0xdb3bbb7bU, + 0xcd267809U, 0x6e5918f4U, 0xec9ab701U, 0x834f9aa8U, + 0xe6956e65U, 0xaaffe67eU, 0x21bccf08U, 0xef15e8e6U, + 0xbae79bd9U, 0x4a6f36ceU, 0xea9f09d4U, 0x29b07cd6U, + 0x31a4b2afU, 0x2a3f2331U, 0xc6a59430U, 0x35a266c0U, + 0x744ebc37U, 0xfc82caa6U, 0xe090d0b0U, 0x33a7d815U, + 0xf104984aU, 0x41ecdaf7U, 0x7fcd500eU, 0x1791f62fU, + 0x764dd68dU, 0x43efb04dU, 0xccaa4d54U, 0xe49604dfU, + 0x9ed1b5e3U, 0x4c6a881bU, 0xc12c1fb8U, 0x4665517fU, + 0x9d5eea04U, 0x018c355dU, 0xfa877473U, 0xfb0b412eU, + 0xb3671d5aU, 0x92dbd252U, 0xe9105633U, 0x6dd64713U, + 0x9ad7618cU, 0x37a10c7aU, 0x59f8148eU, 0xeb133c89U, + 0xcea927eeU, 0xb761c935U, 0xe11ce5edU, 0x7a47b13cU, + 0x9cd2df59U, 0x55f2733fU, 0x1814ce79U, 0x73c737bfU, + 0x53f7cdeaU, 0x5ffdaa5bU, 0xdf3d6f14U, 0x7844db86U, + 0xcaaff381U, 0xb968c43eU, 0x3824342cU, 0xc2a3405fU, + 0x161dc372U, 0xbce2250cU, 0x283c498bU, 0xff0d9541U, + 0x39a80171U, 0x080cb3deU, 0xd8b4e49cU, 0x6456c190U, + 0x7bcb8461U, 0xd532b670U, 0x486c5c74U, 0xd0b85742U, +}; +static const u32 Td1[256] = { + 0x5051f4a7U, 0x537e4165U, 0xc31a17a4U, 0x963a275eU, + 0xcb3bab6bU, 0xf11f9d45U, 0xabacfa58U, 0x934be303U, + 0x552030faU, 0xf6ad766dU, 0x9188cc76U, 0x25f5024cU, + 0xfc4fe5d7U, 0xd7c52acbU, 0x80263544U, 0x8fb562a3U, + 0x49deb15aU, 0x6725ba1bU, 0x9845ea0eU, 0xe15dfec0U, + 0x02c32f75U, 0x12814cf0U, 0xa38d4697U, 0xc66bd3f9U, + 0xe7038f5fU, 0x9515929cU, 0xebbf6d7aU, 0xda955259U, + 0x2dd4be83U, 0xd3587421U, 0x2949e069U, 0x448ec9c8U, + 0x6a75c289U, 0x78f48e79U, 0x6b99583eU, 0xdd27b971U, + 0xb6bee14fU, 0x17f088adU, 0x66c920acU, 0xb47dce3aU, + 0x1863df4aU, 0x82e51a31U, 0x60975133U, 0x4562537fU, + 0xe0b16477U, 0x84bb6baeU, 0x1cfe81a0U, 0x94f9082bU, + 0x58704868U, 0x198f45fdU, 0x8794de6cU, 0xb7527bf8U, + 0x23ab73d3U, 0xe2724b02U, 0x57e31f8fU, 0x2a6655abU, + 0x07b2eb28U, 0x032fb5c2U, 0x9a86c57bU, 0xa5d33708U, + 0xf2302887U, 0xb223bfa5U, 0xba02036aU, 0x5ced1682U, + 0x2b8acf1cU, 0x92a779b4U, 0xf0f307f2U, 0xa14e69e2U, + 0xcd65daf4U, 0xd50605beU, 0x1fd13462U, 0x8ac4a6feU, + 0x9d342e53U, 0xa0a2f355U, 0x32058ae1U, 0x75a4f6ebU, + 0x390b83ecU, 0xaa4060efU, 0x065e719fU, 0x51bd6e10U, + 0xf93e218aU, 0x3d96dd06U, 0xaedd3e05U, 0x464de6bdU, + 0xb591548dU, 0x0571c45dU, 0x6f0406d4U, 0xff605015U, + 0x241998fbU, 0x97d6bde9U, 0xcc894043U, 0x7767d99eU, + 0xbdb0e842U, 0x8807898bU, 0x38e7195bU, 0xdb79c8eeU, + 0x47a17c0aU, 0xe97c420fU, 0xc9f8841eU, 0x00000000U, + 0x83098086U, 0x48322bedU, 0xac1e1170U, 0x4e6c5a72U, + 0xfbfd0effU, 0x560f8538U, 0x1e3daed5U, 0x27362d39U, + 0x640a0fd9U, 0x21685ca6U, 0xd19b5b54U, 0x3a24362eU, + 0xb10c0a67U, 0x0f9357e7U, 0xd2b4ee96U, 0x9e1b9b91U, + 0x4f80c0c5U, 0xa261dc20U, 0x695a774bU, 0x161c121aU, + 0x0ae293baU, 0xe5c0a02aU, 0x433c22e0U, 0x1d121b17U, + 0x0b0e090dU, 0xadf28bc7U, 0xb92db6a8U, 0xc8141ea9U, + 0x8557f119U, 0x4caf7507U, 0xbbee99ddU, 0xfda37f60U, + 0x9ff70126U, 0xbc5c72f5U, 0xc544663bU, 0x345bfb7eU, + 0x768b4329U, 0xdccb23c6U, 0x68b6edfcU, 0x63b8e4f1U, + 0xcad731dcU, 0x10426385U, 0x40139722U, 0x2084c611U, + 0x7d854a24U, 0xf8d2bb3dU, 0x11aef932U, 0x6dc729a1U, + 0x4b1d9e2fU, 0xf3dcb230U, 0xec0d8652U, 0xd077c1e3U, + 0x6c2bb316U, 0x99a970b9U, 0xfa119448U, 0x2247e964U, + 0xc4a8fc8cU, 0x1aa0f03fU, 0xd8567d2cU, 0xef223390U, + 0xc787494eU, 0xc1d938d1U, 0xfe8ccaa2U, 0x3698d40bU, + 0xcfa6f581U, 0x28a57adeU, 0x26dab78eU, 0xa43fadbfU, + 0xe42c3a9dU, 0x0d507892U, 0x9b6a5fccU, 0x62547e46U, + 0xc2f68d13U, 0xe890d8b8U, 0x5e2e39f7U, 0xf582c3afU, + 0xbe9f5d80U, 0x7c69d093U, 0xa96fd52dU, 0xb3cf2512U, + 0x3bc8ac99U, 0xa710187dU, 0x6ee89c63U, 0x7bdb3bbbU, + 0x09cd2678U, 0xf46e5918U, 0x01ec9ab7U, 0xa8834f9aU, + 0x65e6956eU, 0x7eaaffe6U, 0x0821bccfU, 0xe6ef15e8U, + 0xd9bae79bU, 0xce4a6f36U, 0xd4ea9f09U, 0xd629b07cU, + 0xaf31a4b2U, 0x312a3f23U, 0x30c6a594U, 0xc035a266U, + 0x37744ebcU, 0xa6fc82caU, 0xb0e090d0U, 0x1533a7d8U, + 0x4af10498U, 0xf741ecdaU, 0x0e7fcd50U, 0x2f1791f6U, + 0x8d764dd6U, 0x4d43efb0U, 0x54ccaa4dU, 0xdfe49604U, + 0xe39ed1b5U, 0x1b4c6a88U, 0xb8c12c1fU, 0x7f466551U, + 0x049d5eeaU, 0x5d018c35U, 0x73fa8774U, 0x2efb0b41U, + 0x5ab3671dU, 0x5292dbd2U, 0x33e91056U, 0x136dd647U, + 0x8c9ad761U, 0x7a37a10cU, 0x8e59f814U, 0x89eb133cU, + 0xeecea927U, 0x35b761c9U, 0xede11ce5U, 0x3c7a47b1U, + 0x599cd2dfU, 0x3f55f273U, 0x791814ceU, 0xbf73c737U, + 0xea53f7cdU, 0x5b5ffdaaU, 0x14df3d6fU, 0x867844dbU, + 0x81caaff3U, 0x3eb968c4U, 0x2c382434U, 0x5fc2a340U, + 0x72161dc3U, 0x0cbce225U, 0x8b283c49U, 0x41ff0d95U, + 0x7139a801U, 0xde080cb3U, 0x9cd8b4e4U, 0x906456c1U, + 0x617bcb84U, 0x70d532b6U, 0x74486c5cU, 0x42d0b857U, +}; +static const u32 Td2[256] = { + 0xa75051f4U, 0x65537e41U, 0xa4c31a17U, 0x5e963a27U, + 0x6bcb3babU, 0x45f11f9dU, 0x58abacfaU, 0x03934be3U, + 0xfa552030U, 0x6df6ad76U, 0x769188ccU, 0x4c25f502U, + 0xd7fc4fe5U, 0xcbd7c52aU, 0x44802635U, 0xa38fb562U, + 0x5a49deb1U, 0x1b6725baU, 0x0e9845eaU, 0xc0e15dfeU, + 0x7502c32fU, 0xf012814cU, 0x97a38d46U, 0xf9c66bd3U, + 0x5fe7038fU, 0x9c951592U, 0x7aebbf6dU, 0x59da9552U, + 0x832dd4beU, 0x21d35874U, 0x692949e0U, 0xc8448ec9U, + 0x896a75c2U, 0x7978f48eU, 0x3e6b9958U, 0x71dd27b9U, + 0x4fb6bee1U, 0xad17f088U, 0xac66c920U, 0x3ab47dceU, + 0x4a1863dfU, 0x3182e51aU, 0x33609751U, 0x7f456253U, + 0x77e0b164U, 0xae84bb6bU, 0xa01cfe81U, 0x2b94f908U, + 0x68587048U, 0xfd198f45U, 0x6c8794deU, 0xf8b7527bU, + 0xd323ab73U, 0x02e2724bU, 0x8f57e31fU, 0xab2a6655U, + 0x2807b2ebU, 0xc2032fb5U, 0x7b9a86c5U, 0x08a5d337U, + 0x87f23028U, 0xa5b223bfU, 0x6aba0203U, 0x825ced16U, + 0x1c2b8acfU, 0xb492a779U, 0xf2f0f307U, 0xe2a14e69U, + 0xf4cd65daU, 0xbed50605U, 0x621fd134U, 0xfe8ac4a6U, + 0x539d342eU, 0x55a0a2f3U, 0xe132058aU, 0xeb75a4f6U, + 0xec390b83U, 0xefaa4060U, 0x9f065e71U, 0x1051bd6eU, + 0x8af93e21U, 0x063d96ddU, 0x05aedd3eU, 0xbd464de6U, + 0x8db59154U, 0x5d0571c4U, 0xd46f0406U, 0x15ff6050U, + 0xfb241998U, 0xe997d6bdU, 0x43cc8940U, 0x9e7767d9U, + 0x42bdb0e8U, 0x8b880789U, 0x5b38e719U, 0xeedb79c8U, + 0x0a47a17cU, 0x0fe97c42U, 0x1ec9f884U, 0x00000000U, + 0x86830980U, 0xed48322bU, 0x70ac1e11U, 0x724e6c5aU, + 0xfffbfd0eU, 0x38560f85U, 0xd51e3daeU, 0x3927362dU, + 0xd9640a0fU, 0xa621685cU, 0x54d19b5bU, 0x2e3a2436U, + 0x67b10c0aU, 0xe70f9357U, 0x96d2b4eeU, 0x919e1b9bU, + 0xc54f80c0U, 0x20a261dcU, 0x4b695a77U, 0x1a161c12U, + 0xba0ae293U, 0x2ae5c0a0U, 0xe0433c22U, 0x171d121bU, + 0x0d0b0e09U, 0xc7adf28bU, 0xa8b92db6U, 0xa9c8141eU, + 0x198557f1U, 0x074caf75U, 0xddbbee99U, 0x60fda37fU, + 0x269ff701U, 0xf5bc5c72U, 0x3bc54466U, 0x7e345bfbU, + 0x29768b43U, 0xc6dccb23U, 0xfc68b6edU, 0xf163b8e4U, + 0xdccad731U, 0x85104263U, 0x22401397U, 0x112084c6U, + 0x247d854aU, 0x3df8d2bbU, 0x3211aef9U, 0xa16dc729U, + 0x2f4b1d9eU, 0x30f3dcb2U, 0x52ec0d86U, 0xe3d077c1U, + 0x166c2bb3U, 0xb999a970U, 0x48fa1194U, 0x642247e9U, + 0x8cc4a8fcU, 0x3f1aa0f0U, 0x2cd8567dU, 0x90ef2233U, + 0x4ec78749U, 0xd1c1d938U, 0xa2fe8ccaU, 0x0b3698d4U, + 0x81cfa6f5U, 0xde28a57aU, 0x8e26dab7U, 0xbfa43fadU, + 0x9de42c3aU, 0x920d5078U, 0xcc9b6a5fU, 0x4662547eU, + 0x13c2f68dU, 0xb8e890d8U, 0xf75e2e39U, 0xaff582c3U, + 0x80be9f5dU, 0x937c69d0U, 0x2da96fd5U, 0x12b3cf25U, + 0x993bc8acU, 0x7da71018U, 0x636ee89cU, 0xbb7bdb3bU, + 0x7809cd26U, 0x18f46e59U, 0xb701ec9aU, 0x9aa8834fU, + 0x6e65e695U, 0xe67eaaffU, 0xcf0821bcU, 0xe8e6ef15U, + 0x9bd9bae7U, 0x36ce4a6fU, 0x09d4ea9fU, 0x7cd629b0U, + 0xb2af31a4U, 0x23312a3fU, 0x9430c6a5U, 0x66c035a2U, + 0xbc37744eU, 0xcaa6fc82U, 0xd0b0e090U, 0xd81533a7U, + 0x984af104U, 0xdaf741ecU, 0x500e7fcdU, 0xf62f1791U, + 0xd68d764dU, 0xb04d43efU, 0x4d54ccaaU, 0x04dfe496U, + 0xb5e39ed1U, 0x881b4c6aU, 0x1fb8c12cU, 0x517f4665U, + 0xea049d5eU, 0x355d018cU, 0x7473fa87U, 0x412efb0bU, + 0x1d5ab367U, 0xd25292dbU, 0x5633e910U, 0x47136dd6U, + 0x618c9ad7U, 0x0c7a37a1U, 0x148e59f8U, 0x3c89eb13U, + 0x27eecea9U, 0xc935b761U, 0xe5ede11cU, 0xb13c7a47U, + 0xdf599cd2U, 0x733f55f2U, 0xce791814U, 0x37bf73c7U, + 0xcdea53f7U, 0xaa5b5ffdU, 0x6f14df3dU, 0xdb867844U, + 0xf381caafU, 0xc43eb968U, 0x342c3824U, 0x405fc2a3U, + 0xc372161dU, 0x250cbce2U, 0x498b283cU, 0x9541ff0dU, + 0x017139a8U, 0xb3de080cU, 0xe49cd8b4U, 0xc1906456U, + 0x84617bcbU, 0xb670d532U, 0x5c74486cU, 0x5742d0b8U, +}; +static const u32 Td3[256] = { + 0xf4a75051U, 0x4165537eU, 0x17a4c31aU, 0x275e963aU, + 0xab6bcb3bU, 0x9d45f11fU, 0xfa58abacU, 0xe303934bU, + 0x30fa5520U, 0x766df6adU, 0xcc769188U, 0x024c25f5U, + 0xe5d7fc4fU, 0x2acbd7c5U, 0x35448026U, 0x62a38fb5U, + 0xb15a49deU, 0xba1b6725U, 0xea0e9845U, 0xfec0e15dU, + 0x2f7502c3U, 0x4cf01281U, 0x4697a38dU, 0xd3f9c66bU, + 0x8f5fe703U, 0x929c9515U, 0x6d7aebbfU, 0x5259da95U, + 0xbe832dd4U, 0x7421d358U, 0xe0692949U, 0xc9c8448eU, + 0xc2896a75U, 0x8e7978f4U, 0x583e6b99U, 0xb971dd27U, + 0xe14fb6beU, 0x88ad17f0U, 0x20ac66c9U, 0xce3ab47dU, + 0xdf4a1863U, 0x1a3182e5U, 0x51336097U, 0x537f4562U, + 0x6477e0b1U, 0x6bae84bbU, 0x81a01cfeU, 0x082b94f9U, + 0x48685870U, 0x45fd198fU, 0xde6c8794U, 0x7bf8b752U, + 0x73d323abU, 0x4b02e272U, 0x1f8f57e3U, 0x55ab2a66U, + 0xeb2807b2U, 0xb5c2032fU, 0xc57b9a86U, 0x3708a5d3U, + 0x2887f230U, 0xbfa5b223U, 0x036aba02U, 0x16825cedU, + 0xcf1c2b8aU, 0x79b492a7U, 0x07f2f0f3U, 0x69e2a14eU, + 0xdaf4cd65U, 0x05bed506U, 0x34621fd1U, 0xa6fe8ac4U, + 0x2e539d34U, 0xf355a0a2U, 0x8ae13205U, 0xf6eb75a4U, + 0x83ec390bU, 0x60efaa40U, 0x719f065eU, 0x6e1051bdU, + 0x218af93eU, 0xdd063d96U, 0x3e05aeddU, 0xe6bd464dU, + 0x548db591U, 0xc45d0571U, 0x06d46f04U, 0x5015ff60U, + 0x98fb2419U, 0xbde997d6U, 0x4043cc89U, 0xd99e7767U, + 0xe842bdb0U, 0x898b8807U, 0x195b38e7U, 0xc8eedb79U, + 0x7c0a47a1U, 0x420fe97cU, 0x841ec9f8U, 0x00000000U, + 0x80868309U, 0x2bed4832U, 0x1170ac1eU, 0x5a724e6cU, + 0x0efffbfdU, 0x8538560fU, 0xaed51e3dU, 0x2d392736U, + 0x0fd9640aU, 0x5ca62168U, 0x5b54d19bU, 0x362e3a24U, + 0x0a67b10cU, 0x57e70f93U, 0xee96d2b4U, 0x9b919e1bU, + 0xc0c54f80U, 0xdc20a261U, 0x774b695aU, 0x121a161cU, + 0x93ba0ae2U, 0xa02ae5c0U, 0x22e0433cU, 0x1b171d12U, + 0x090d0b0eU, 0x8bc7adf2U, 0xb6a8b92dU, 0x1ea9c814U, + 0xf1198557U, 0x75074cafU, 0x99ddbbeeU, 0x7f60fda3U, + 0x01269ff7U, 0x72f5bc5cU, 0x663bc544U, 0xfb7e345bU, + 0x4329768bU, 0x23c6dccbU, 0xedfc68b6U, 0xe4f163b8U, + 0x31dccad7U, 0x63851042U, 0x97224013U, 0xc6112084U, + 0x4a247d85U, 0xbb3df8d2U, 0xf93211aeU, 0x29a16dc7U, + 0x9e2f4b1dU, 0xb230f3dcU, 0x8652ec0dU, 0xc1e3d077U, + 0xb3166c2bU, 0x70b999a9U, 0x9448fa11U, 0xe9642247U, + 0xfc8cc4a8U, 0xf03f1aa0U, 0x7d2cd856U, 0x3390ef22U, + 0x494ec787U, 0x38d1c1d9U, 0xcaa2fe8cU, 0xd40b3698U, + 0xf581cfa6U, 0x7ade28a5U, 0xb78e26daU, 0xadbfa43fU, + 0x3a9de42cU, 0x78920d50U, 0x5fcc9b6aU, 0x7e466254U, + 0x8d13c2f6U, 0xd8b8e890U, 0x39f75e2eU, 0xc3aff582U, + 0x5d80be9fU, 0xd0937c69U, 0xd52da96fU, 0x2512b3cfU, + 0xac993bc8U, 0x187da710U, 0x9c636ee8U, 0x3bbb7bdbU, + 0x267809cdU, 0x5918f46eU, 0x9ab701ecU, 0x4f9aa883U, + 0x956e65e6U, 0xffe67eaaU, 0xbccf0821U, 0x15e8e6efU, + 0xe79bd9baU, 0x6f36ce4aU, 0x9f09d4eaU, 0xb07cd629U, + 0xa4b2af31U, 0x3f23312aU, 0xa59430c6U, 0xa266c035U, + 0x4ebc3774U, 0x82caa6fcU, 0x90d0b0e0U, 0xa7d81533U, + 0x04984af1U, 0xecdaf741U, 0xcd500e7fU, 0x91f62f17U, + 0x4dd68d76U, 0xefb04d43U, 0xaa4d54ccU, 0x9604dfe4U, + 0xd1b5e39eU, 0x6a881b4cU, 0x2c1fb8c1U, 0x65517f46U, + 0x5eea049dU, 0x8c355d01U, 0x877473faU, 0x0b412efbU, + 0x671d5ab3U, 0xdbd25292U, 0x105633e9U, 0xd647136dU, + 0xd7618c9aU, 0xa10c7a37U, 0xf8148e59U, 0x133c89ebU, + 0xa927eeceU, 0x61c935b7U, 0x1ce5ede1U, 0x47b13c7aU, + 0xd2df599cU, 0xf2733f55U, 0x14ce7918U, 0xc737bf73U, + 0xf7cdea53U, 0xfdaa5b5fU, 0x3d6f14dfU, 0x44db8678U, + 0xaff381caU, 0x68c43eb9U, 0x24342c38U, 0xa3405fc2U, + 0x1dc37216U, 0xe2250cbcU, 0x3c498b28U, 0x0d9541ffU, + 0xa8017139U, 0x0cb3de08U, 0xb4e49cd8U, 0x56c19064U, + 0xcb84617bU, 0x32b670d5U, 0x6c5c7448U, 0xb85742d0U, +}; +static const u8 Td4[256] = { + 0x52U, 0x09U, 0x6aU, 0xd5U, 0x30U, 0x36U, 0xa5U, 0x38U, + 0xbfU, 0x40U, 0xa3U, 0x9eU, 0x81U, 0xf3U, 0xd7U, 0xfbU, + 0x7cU, 0xe3U, 0x39U, 0x82U, 0x9bU, 0x2fU, 0xffU, 0x87U, + 0x34U, 0x8eU, 0x43U, 0x44U, 0xc4U, 0xdeU, 0xe9U, 0xcbU, + 0x54U, 0x7bU, 0x94U, 0x32U, 0xa6U, 0xc2U, 0x23U, 0x3dU, + 0xeeU, 0x4cU, 0x95U, 0x0bU, 0x42U, 0xfaU, 0xc3U, 0x4eU, + 0x08U, 0x2eU, 0xa1U, 0x66U, 0x28U, 0xd9U, 0x24U, 0xb2U, + 0x76U, 0x5bU, 0xa2U, 0x49U, 0x6dU, 0x8bU, 0xd1U, 0x25U, + 0x72U, 0xf8U, 0xf6U, 0x64U, 0x86U, 0x68U, 0x98U, 0x16U, + 0xd4U, 0xa4U, 0x5cU, 0xccU, 0x5dU, 0x65U, 0xb6U, 0x92U, + 0x6cU, 0x70U, 0x48U, 0x50U, 0xfdU, 0xedU, 0xb9U, 0xdaU, + 0x5eU, 0x15U, 0x46U, 0x57U, 0xa7U, 0x8dU, 0x9dU, 0x84U, + 0x90U, 0xd8U, 0xabU, 0x00U, 0x8cU, 0xbcU, 0xd3U, 0x0aU, + 0xf7U, 0xe4U, 0x58U, 0x05U, 0xb8U, 0xb3U, 0x45U, 0x06U, + 0xd0U, 0x2cU, 0x1eU, 0x8fU, 0xcaU, 0x3fU, 0x0fU, 0x02U, + 0xc1U, 0xafU, 0xbdU, 0x03U, 0x01U, 0x13U, 0x8aU, 0x6bU, + 0x3aU, 0x91U, 0x11U, 0x41U, 0x4fU, 0x67U, 0xdcU, 0xeaU, + 0x97U, 0xf2U, 0xcfU, 0xceU, 0xf0U, 0xb4U, 0xe6U, 0x73U, + 0x96U, 0xacU, 0x74U, 0x22U, 0xe7U, 0xadU, 0x35U, 0x85U, + 0xe2U, 0xf9U, 0x37U, 0xe8U, 0x1cU, 0x75U, 0xdfU, 0x6eU, + 0x47U, 0xf1U, 0x1aU, 0x71U, 0x1dU, 0x29U, 0xc5U, 0x89U, + 0x6fU, 0xb7U, 0x62U, 0x0eU, 0xaaU, 0x18U, 0xbeU, 0x1bU, + 0xfcU, 0x56U, 0x3eU, 0x4bU, 0xc6U, 0xd2U, 0x79U, 0x20U, + 0x9aU, 0xdbU, 0xc0U, 0xfeU, 0x78U, 0xcdU, 0x5aU, 0xf4U, + 0x1fU, 0xddU, 0xa8U, 0x33U, 0x88U, 0x07U, 0xc7U, 0x31U, + 0xb1U, 0x12U, 0x10U, 0x59U, 0x27U, 0x80U, 0xecU, 0x5fU, + 0x60U, 0x51U, 0x7fU, 0xa9U, 0x19U, 0xb5U, 0x4aU, 0x0dU, + 0x2dU, 0xe5U, 0x7aU, 0x9fU, 0x93U, 0xc9U, 0x9cU, 0xefU, + 0xa0U, 0xe0U, 0x3bU, 0x4dU, 0xaeU, 0x2aU, 0xf5U, 0xb0U, + 0xc8U, 0xebU, 0xbbU, 0x3cU, 0x83U, 0x53U, 0x99U, 0x61U, + 0x17U, 0x2bU, 0x04U, 0x7eU, 0xbaU, 0x77U, 0xd6U, 0x26U, + 0xe1U, 0x69U, 0x14U, 0x63U, 0x55U, 0x21U, 0x0cU, 0x7dU, +}; +static const u32 rcon[] = { + 0x01000000, 0x02000000, 0x04000000, 0x08000000, + 0x10000000, 0x20000000, 0x40000000, 0x80000000, + 0x1B000000, 0x36000000, /* for 128-bit blocks, Rijndael never uses more than 10 rcon values */ +}; + +/** + * Expand the cipher key into the encryption key schedule. + */ +int AES_set_encrypt_key(const unsigned char *userKey, const int bits, + AES_KEY *key) +{ + + u32 *rk; + int i = 0; + u32 temp; + + if (!userKey || !key) + return -1; + if (bits != 128 && bits != 192 && bits != 256) + return -2; + + rk = key->rd_key; + + if (bits == 128) + key->rounds = 10; + else if (bits == 192) + key->rounds = 12; + else + key->rounds = 14; + + rk[0] = GETU32(userKey ); + rk[1] = GETU32(userKey + 4); + rk[2] = GETU32(userKey + 8); + rk[3] = GETU32(userKey + 12); + if (bits == 128) { + while (1) { + temp = rk[3]; + rk[4] = rk[0] ^ + (Te2[(temp >> 16) & 0xff] & 0xff000000) ^ + (Te3[(temp >> 8) & 0xff] & 0x00ff0000) ^ + (Te0[(temp ) & 0xff] & 0x0000ff00) ^ + (Te1[(temp >> 24) ] & 0x000000ff) ^ + rcon[i]; + rk[5] = rk[1] ^ rk[4]; + rk[6] = rk[2] ^ rk[5]; + rk[7] = rk[3] ^ rk[6]; + if (++i == 10) { + return 0; + } + rk += 4; + } + } + rk[4] = GETU32(userKey + 16); + rk[5] = GETU32(userKey + 20); + if (bits == 192) { + while (1) { + temp = rk[ 5]; + rk[ 6] = rk[ 0] ^ + (Te2[(temp >> 16) & 0xff] & 0xff000000) ^ + (Te3[(temp >> 8) & 0xff] & 0x00ff0000) ^ + (Te0[(temp ) & 0xff] & 0x0000ff00) ^ + (Te1[(temp >> 24) ] & 0x000000ff) ^ + rcon[i]; + rk[ 7] = rk[ 1] ^ rk[ 6]; + rk[ 8] = rk[ 2] ^ rk[ 7]; + rk[ 9] = rk[ 3] ^ rk[ 8]; + if (++i == 8) { + return 0; + } + rk[10] = rk[ 4] ^ rk[ 9]; + rk[11] = rk[ 5] ^ rk[10]; + rk += 6; + } + } + rk[6] = GETU32(userKey + 24); + rk[7] = GETU32(userKey + 28); + if (bits == 256) { + while (1) { + temp = rk[ 7]; + rk[ 8] = rk[ 0] ^ + (Te2[(temp >> 16) & 0xff] & 0xff000000) ^ + (Te3[(temp >> 8) & 0xff] & 0x00ff0000) ^ + (Te0[(temp ) & 0xff] & 0x0000ff00) ^ + (Te1[(temp >> 24) ] & 0x000000ff) ^ + rcon[i]; + rk[ 9] = rk[ 1] ^ rk[ 8]; + rk[10] = rk[ 2] ^ rk[ 9]; + rk[11] = rk[ 3] ^ rk[10]; + if (++i == 7) { + return 0; + } + temp = rk[11]; + rk[12] = rk[ 4] ^ + (Te2[(temp >> 24) ] & 0xff000000) ^ + (Te3[(temp >> 16) & 0xff] & 0x00ff0000) ^ + (Te0[(temp >> 8) & 0xff] & 0x0000ff00) ^ + (Te1[(temp ) & 0xff] & 0x000000ff); + rk[13] = rk[ 5] ^ rk[12]; + rk[14] = rk[ 6] ^ rk[13]; + rk[15] = rk[ 7] ^ rk[14]; + + rk += 8; + } + } + return 0; +} + +/** + * Expand the cipher key into the decryption key schedule. + */ +int AES_set_decrypt_key(const unsigned char *userKey, const int bits, + AES_KEY *key) +{ + + u32 *rk; + int i, j, status; + u32 temp; + + /* first, start with an encryption schedule */ + status = AES_set_encrypt_key(userKey, bits, key); + if (status < 0) + return status; + + rk = key->rd_key; + + /* invert the order of the round keys: */ + for (i = 0, j = 4*(key->rounds); i < j; i += 4, j -= 4) { + temp = rk[i ]; rk[i ] = rk[j ]; rk[j ] = temp; + temp = rk[i + 1]; rk[i + 1] = rk[j + 1]; rk[j + 1] = temp; + temp = rk[i + 2]; rk[i + 2] = rk[j + 2]; rk[j + 2] = temp; + temp = rk[i + 3]; rk[i + 3] = rk[j + 3]; rk[j + 3] = temp; + } + /* apply the inverse MixColumn transform to all round keys but the first and the last: */ + for (i = 1; i < (key->rounds); i++) { + rk += 4; + rk[0] = + Td0[Te1[(rk[0] >> 24) ] & 0xff] ^ + Td1[Te1[(rk[0] >> 16) & 0xff] & 0xff] ^ + Td2[Te1[(rk[0] >> 8) & 0xff] & 0xff] ^ + Td3[Te1[(rk[0] ) & 0xff] & 0xff]; + rk[1] = + Td0[Te1[(rk[1] >> 24) ] & 0xff] ^ + Td1[Te1[(rk[1] >> 16) & 0xff] & 0xff] ^ + Td2[Te1[(rk[1] >> 8) & 0xff] & 0xff] ^ + Td3[Te1[(rk[1] ) & 0xff] & 0xff]; + rk[2] = + Td0[Te1[(rk[2] >> 24) ] & 0xff] ^ + Td1[Te1[(rk[2] >> 16) & 0xff] & 0xff] ^ + Td2[Te1[(rk[2] >> 8) & 0xff] & 0xff] ^ + Td3[Te1[(rk[2] ) & 0xff] & 0xff]; + rk[3] = + Td0[Te1[(rk[3] >> 24) ] & 0xff] ^ + Td1[Te1[(rk[3] >> 16) & 0xff] & 0xff] ^ + Td2[Te1[(rk[3] >> 8) & 0xff] & 0xff] ^ + Td3[Te1[(rk[3] ) & 0xff] & 0xff]; + } + return 0; +} + +/* + * Encrypt a single block + * in and out can overlap + */ +void AES_encrypt(const unsigned char *in, unsigned char *out, + const AES_KEY *key) { + + const u32 *rk; + u32 s0, s1, s2, s3, t0, t1, t2, t3; +#ifndef FULL_UNROLL + int r; +#endif /* ?FULL_UNROLL */ + + assert(in && out && key); + rk = key->rd_key; + + /* + * map byte array block to cipher state + * and add initial round key: + */ + s0 = GETU32(in ) ^ rk[0]; + s1 = GETU32(in + 4) ^ rk[1]; + s2 = GETU32(in + 8) ^ rk[2]; + s3 = GETU32(in + 12) ^ rk[3]; +#ifdef FULL_UNROLL + /* round 1: */ + t0 = Te0[s0 >> 24] ^ Te1[(s1 >> 16) & 0xff] ^ Te2[(s2 >> 8) & 0xff] ^ Te3[s3 & 0xff] ^ rk[ 4]; + t1 = Te0[s1 >> 24] ^ Te1[(s2 >> 16) & 0xff] ^ Te2[(s3 >> 8) & 0xff] ^ Te3[s0 & 0xff] ^ rk[ 5]; + t2 = Te0[s2 >> 24] ^ Te1[(s3 >> 16) & 0xff] ^ Te2[(s0 >> 8) & 0xff] ^ Te3[s1 & 0xff] ^ rk[ 6]; + t3 = Te0[s3 >> 24] ^ Te1[(s0 >> 16) & 0xff] ^ Te2[(s1 >> 8) & 0xff] ^ Te3[s2 & 0xff] ^ rk[ 7]; + /* round 2: */ + s0 = Te0[t0 >> 24] ^ Te1[(t1 >> 16) & 0xff] ^ Te2[(t2 >> 8) & 0xff] ^ Te3[t3 & 0xff] ^ rk[ 8]; + s1 = Te0[t1 >> 24] ^ Te1[(t2 >> 16) & 0xff] ^ Te2[(t3 >> 8) & 0xff] ^ Te3[t0 & 0xff] ^ rk[ 9]; + s2 = Te0[t2 >> 24] ^ Te1[(t3 >> 16) & 0xff] ^ Te2[(t0 >> 8) & 0xff] ^ Te3[t1 & 0xff] ^ rk[10]; + s3 = Te0[t3 >> 24] ^ Te1[(t0 >> 16) & 0xff] ^ Te2[(t1 >> 8) & 0xff] ^ Te3[t2 & 0xff] ^ rk[11]; + /* round 3: */ + t0 = Te0[s0 >> 24] ^ Te1[(s1 >> 16) & 0xff] ^ Te2[(s2 >> 8) & 0xff] ^ Te3[s3 & 0xff] ^ rk[12]; + t1 = Te0[s1 >> 24] ^ Te1[(s2 >> 16) & 0xff] ^ Te2[(s3 >> 8) & 0xff] ^ Te3[s0 & 0xff] ^ rk[13]; + t2 = Te0[s2 >> 24] ^ Te1[(s3 >> 16) & 0xff] ^ Te2[(s0 >> 8) & 0xff] ^ Te3[s1 & 0xff] ^ rk[14]; + t3 = Te0[s3 >> 24] ^ Te1[(s0 >> 16) & 0xff] ^ Te2[(s1 >> 8) & 0xff] ^ Te3[s2 & 0xff] ^ rk[15]; + /* round 4: */ + s0 = Te0[t0 >> 24] ^ Te1[(t1 >> 16) & 0xff] ^ Te2[(t2 >> 8) & 0xff] ^ Te3[t3 & 0xff] ^ rk[16]; + s1 = Te0[t1 >> 24] ^ Te1[(t2 >> 16) & 0xff] ^ Te2[(t3 >> 8) & 0xff] ^ Te3[t0 & 0xff] ^ rk[17]; + s2 = Te0[t2 >> 24] ^ Te1[(t3 >> 16) & 0xff] ^ Te2[(t0 >> 8) & 0xff] ^ Te3[t1 & 0xff] ^ rk[18]; + s3 = Te0[t3 >> 24] ^ Te1[(t0 >> 16) & 0xff] ^ Te2[(t1 >> 8) & 0xff] ^ Te3[t2 & 0xff] ^ rk[19]; + /* round 5: */ + t0 = Te0[s0 >> 24] ^ Te1[(s1 >> 16) & 0xff] ^ Te2[(s2 >> 8) & 0xff] ^ Te3[s3 & 0xff] ^ rk[20]; + t1 = Te0[s1 >> 24] ^ Te1[(s2 >> 16) & 0xff] ^ Te2[(s3 >> 8) & 0xff] ^ Te3[s0 & 0xff] ^ rk[21]; + t2 = Te0[s2 >> 24] ^ Te1[(s3 >> 16) & 0xff] ^ Te2[(s0 >> 8) & 0xff] ^ Te3[s1 & 0xff] ^ rk[22]; + t3 = Te0[s3 >> 24] ^ Te1[(s0 >> 16) & 0xff] ^ Te2[(s1 >> 8) & 0xff] ^ Te3[s2 & 0xff] ^ rk[23]; + /* round 6: */ + s0 = Te0[t0 >> 24] ^ Te1[(t1 >> 16) & 0xff] ^ Te2[(t2 >> 8) & 0xff] ^ Te3[t3 & 0xff] ^ rk[24]; + s1 = Te0[t1 >> 24] ^ Te1[(t2 >> 16) & 0xff] ^ Te2[(t3 >> 8) & 0xff] ^ Te3[t0 & 0xff] ^ rk[25]; + s2 = Te0[t2 >> 24] ^ Te1[(t3 >> 16) & 0xff] ^ Te2[(t0 >> 8) & 0xff] ^ Te3[t1 & 0xff] ^ rk[26]; + s3 = Te0[t3 >> 24] ^ Te1[(t0 >> 16) & 0xff] ^ Te2[(t1 >> 8) & 0xff] ^ Te3[t2 & 0xff] ^ rk[27]; + /* round 7: */ + t0 = Te0[s0 >> 24] ^ Te1[(s1 >> 16) & 0xff] ^ Te2[(s2 >> 8) & 0xff] ^ Te3[s3 & 0xff] ^ rk[28]; + t1 = Te0[s1 >> 24] ^ Te1[(s2 >> 16) & 0xff] ^ Te2[(s3 >> 8) & 0xff] ^ Te3[s0 & 0xff] ^ rk[29]; + t2 = Te0[s2 >> 24] ^ Te1[(s3 >> 16) & 0xff] ^ Te2[(s0 >> 8) & 0xff] ^ Te3[s1 & 0xff] ^ rk[30]; + t3 = Te0[s3 >> 24] ^ Te1[(s0 >> 16) & 0xff] ^ Te2[(s1 >> 8) & 0xff] ^ Te3[s2 & 0xff] ^ rk[31]; + /* round 8: */ + s0 = Te0[t0 >> 24] ^ Te1[(t1 >> 16) & 0xff] ^ Te2[(t2 >> 8) & 0xff] ^ Te3[t3 & 0xff] ^ rk[32]; + s1 = Te0[t1 >> 24] ^ Te1[(t2 >> 16) & 0xff] ^ Te2[(t3 >> 8) & 0xff] ^ Te3[t0 & 0xff] ^ rk[33]; + s2 = Te0[t2 >> 24] ^ Te1[(t3 >> 16) & 0xff] ^ Te2[(t0 >> 8) & 0xff] ^ Te3[t1 & 0xff] ^ rk[34]; + s3 = Te0[t3 >> 24] ^ Te1[(t0 >> 16) & 0xff] ^ Te2[(t1 >> 8) & 0xff] ^ Te3[t2 & 0xff] ^ rk[35]; + /* round 9: */ + t0 = Te0[s0 >> 24] ^ Te1[(s1 >> 16) & 0xff] ^ Te2[(s2 >> 8) & 0xff] ^ Te3[s3 & 0xff] ^ rk[36]; + t1 = Te0[s1 >> 24] ^ Te1[(s2 >> 16) & 0xff] ^ Te2[(s3 >> 8) & 0xff] ^ Te3[s0 & 0xff] ^ rk[37]; + t2 = Te0[s2 >> 24] ^ Te1[(s3 >> 16) & 0xff] ^ Te2[(s0 >> 8) & 0xff] ^ Te3[s1 & 0xff] ^ rk[38]; + t3 = Te0[s3 >> 24] ^ Te1[(s0 >> 16) & 0xff] ^ Te2[(s1 >> 8) & 0xff] ^ Te3[s2 & 0xff] ^ rk[39]; + if (key->rounds > 10) { + /* round 10: */ + s0 = Te0[t0 >> 24] ^ Te1[(t1 >> 16) & 0xff] ^ Te2[(t2 >> 8) & 0xff] ^ Te3[t3 & 0xff] ^ rk[40]; + s1 = Te0[t1 >> 24] ^ Te1[(t2 >> 16) & 0xff] ^ Te2[(t3 >> 8) & 0xff] ^ Te3[t0 & 0xff] ^ rk[41]; + s2 = Te0[t2 >> 24] ^ Te1[(t3 >> 16) & 0xff] ^ Te2[(t0 >> 8) & 0xff] ^ Te3[t1 & 0xff] ^ rk[42]; + s3 = Te0[t3 >> 24] ^ Te1[(t0 >> 16) & 0xff] ^ Te2[(t1 >> 8) & 0xff] ^ Te3[t2 & 0xff] ^ rk[43]; + /* round 11: */ + t0 = Te0[s0 >> 24] ^ Te1[(s1 >> 16) & 0xff] ^ Te2[(s2 >> 8) & 0xff] ^ Te3[s3 & 0xff] ^ rk[44]; + t1 = Te0[s1 >> 24] ^ Te1[(s2 >> 16) & 0xff] ^ Te2[(s3 >> 8) & 0xff] ^ Te3[s0 & 0xff] ^ rk[45]; + t2 = Te0[s2 >> 24] ^ Te1[(s3 >> 16) & 0xff] ^ Te2[(s0 >> 8) & 0xff] ^ Te3[s1 & 0xff] ^ rk[46]; + t3 = Te0[s3 >> 24] ^ Te1[(s0 >> 16) & 0xff] ^ Te2[(s1 >> 8) & 0xff] ^ Te3[s2 & 0xff] ^ rk[47]; + if (key->rounds > 12) { + /* round 12: */ + s0 = Te0[t0 >> 24] ^ Te1[(t1 >> 16) & 0xff] ^ Te2[(t2 >> 8) & 0xff] ^ Te3[t3 & 0xff] ^ rk[48]; + s1 = Te0[t1 >> 24] ^ Te1[(t2 >> 16) & 0xff] ^ Te2[(t3 >> 8) & 0xff] ^ Te3[t0 & 0xff] ^ rk[49]; + s2 = Te0[t2 >> 24] ^ Te1[(t3 >> 16) & 0xff] ^ Te2[(t0 >> 8) & 0xff] ^ Te3[t1 & 0xff] ^ rk[50]; + s3 = Te0[t3 >> 24] ^ Te1[(t0 >> 16) & 0xff] ^ Te2[(t1 >> 8) & 0xff] ^ Te3[t2 & 0xff] ^ rk[51]; + /* round 13: */ + t0 = Te0[s0 >> 24] ^ Te1[(s1 >> 16) & 0xff] ^ Te2[(s2 >> 8) & 0xff] ^ Te3[s3 & 0xff] ^ rk[52]; + t1 = Te0[s1 >> 24] ^ Te1[(s2 >> 16) & 0xff] ^ Te2[(s3 >> 8) & 0xff] ^ Te3[s0 & 0xff] ^ rk[53]; + t2 = Te0[s2 >> 24] ^ Te1[(s3 >> 16) & 0xff] ^ Te2[(s0 >> 8) & 0xff] ^ Te3[s1 & 0xff] ^ rk[54]; + t3 = Te0[s3 >> 24] ^ Te1[(s0 >> 16) & 0xff] ^ Te2[(s1 >> 8) & 0xff] ^ Te3[s2 & 0xff] ^ rk[55]; + } + } + rk += key->rounds << 2; +#else /* !FULL_UNROLL */ + /* + * Nr - 1 full rounds: + */ + r = key->rounds >> 1; + for (;;) { + t0 = + Te0[(s0 >> 24) ] ^ + Te1[(s1 >> 16) & 0xff] ^ + Te2[(s2 >> 8) & 0xff] ^ + Te3[(s3 ) & 0xff] ^ + rk[4]; + t1 = + Te0[(s1 >> 24) ] ^ + Te1[(s2 >> 16) & 0xff] ^ + Te2[(s3 >> 8) & 0xff] ^ + Te3[(s0 ) & 0xff] ^ + rk[5]; + t2 = + Te0[(s2 >> 24) ] ^ + Te1[(s3 >> 16) & 0xff] ^ + Te2[(s0 >> 8) & 0xff] ^ + Te3[(s1 ) & 0xff] ^ + rk[6]; + t3 = + Te0[(s3 >> 24) ] ^ + Te1[(s0 >> 16) & 0xff] ^ + Te2[(s1 >> 8) & 0xff] ^ + Te3[(s2 ) & 0xff] ^ + rk[7]; + + rk += 8; + if (--r == 0) { + break; + } + + s0 = + Te0[(t0 >> 24) ] ^ + Te1[(t1 >> 16) & 0xff] ^ + Te2[(t2 >> 8) & 0xff] ^ + Te3[(t3 ) & 0xff] ^ + rk[0]; + s1 = + Te0[(t1 >> 24) ] ^ + Te1[(t2 >> 16) & 0xff] ^ + Te2[(t3 >> 8) & 0xff] ^ + Te3[(t0 ) & 0xff] ^ + rk[1]; + s2 = + Te0[(t2 >> 24) ] ^ + Te1[(t3 >> 16) & 0xff] ^ + Te2[(t0 >> 8) & 0xff] ^ + Te3[(t1 ) & 0xff] ^ + rk[2]; + s3 = + Te0[(t3 >> 24) ] ^ + Te1[(t0 >> 16) & 0xff] ^ + Te2[(t1 >> 8) & 0xff] ^ + Te3[(t2 ) & 0xff] ^ + rk[3]; + } +#endif /* ?FULL_UNROLL */ + /* + * apply last round and + * map cipher state to byte array block: + */ + s0 = + (Te2[(t0 >> 24) ] & 0xff000000) ^ + (Te3[(t1 >> 16) & 0xff] & 0x00ff0000) ^ + (Te0[(t2 >> 8) & 0xff] & 0x0000ff00) ^ + (Te1[(t3 ) & 0xff] & 0x000000ff) ^ + rk[0]; + PUTU32(out , s0); + s1 = + (Te2[(t1 >> 24) ] & 0xff000000) ^ + (Te3[(t2 >> 16) & 0xff] & 0x00ff0000) ^ + (Te0[(t3 >> 8) & 0xff] & 0x0000ff00) ^ + (Te1[(t0 ) & 0xff] & 0x000000ff) ^ + rk[1]; + PUTU32(out + 4, s1); + s2 = + (Te2[(t2 >> 24) ] & 0xff000000) ^ + (Te3[(t3 >> 16) & 0xff] & 0x00ff0000) ^ + (Te0[(t0 >> 8) & 0xff] & 0x0000ff00) ^ + (Te1[(t1 ) & 0xff] & 0x000000ff) ^ + rk[2]; + PUTU32(out + 8, s2); + s3 = + (Te2[(t3 >> 24) ] & 0xff000000) ^ + (Te3[(t0 >> 16) & 0xff] & 0x00ff0000) ^ + (Te0[(t1 >> 8) & 0xff] & 0x0000ff00) ^ + (Te1[(t2 ) & 0xff] & 0x000000ff) ^ + rk[3]; + PUTU32(out + 12, s3); +} + +/* + * Decrypt a single block + * in and out can overlap + */ +void AES_decrypt(const unsigned char *in, unsigned char *out, + const AES_KEY *key) +{ + + const u32 *rk; + u32 s0, s1, s2, s3, t0, t1, t2, t3; +#ifndef FULL_UNROLL + int r; +#endif /* ?FULL_UNROLL */ + + assert(in && out && key); + rk = key->rd_key; + + /* + * map byte array block to cipher state + * and add initial round key: + */ + s0 = GETU32(in ) ^ rk[0]; + s1 = GETU32(in + 4) ^ rk[1]; + s2 = GETU32(in + 8) ^ rk[2]; + s3 = GETU32(in + 12) ^ rk[3]; +#ifdef FULL_UNROLL + /* round 1: */ + t0 = Td0[s0 >> 24] ^ Td1[(s3 >> 16) & 0xff] ^ Td2[(s2 >> 8) & 0xff] ^ Td3[s1 & 0xff] ^ rk[ 4]; + t1 = Td0[s1 >> 24] ^ Td1[(s0 >> 16) & 0xff] ^ Td2[(s3 >> 8) & 0xff] ^ Td3[s2 & 0xff] ^ rk[ 5]; + t2 = Td0[s2 >> 24] ^ Td1[(s1 >> 16) & 0xff] ^ Td2[(s0 >> 8) & 0xff] ^ Td3[s3 & 0xff] ^ rk[ 6]; + t3 = Td0[s3 >> 24] ^ Td1[(s2 >> 16) & 0xff] ^ Td2[(s1 >> 8) & 0xff] ^ Td3[s0 & 0xff] ^ rk[ 7]; + /* round 2: */ + s0 = Td0[t0 >> 24] ^ Td1[(t3 >> 16) & 0xff] ^ Td2[(t2 >> 8) & 0xff] ^ Td3[t1 & 0xff] ^ rk[ 8]; + s1 = Td0[t1 >> 24] ^ Td1[(t0 >> 16) & 0xff] ^ Td2[(t3 >> 8) & 0xff] ^ Td3[t2 & 0xff] ^ rk[ 9]; + s2 = Td0[t2 >> 24] ^ Td1[(t1 >> 16) & 0xff] ^ Td2[(t0 >> 8) & 0xff] ^ Td3[t3 & 0xff] ^ rk[10]; + s3 = Td0[t3 >> 24] ^ Td1[(t2 >> 16) & 0xff] ^ Td2[(t1 >> 8) & 0xff] ^ Td3[t0 & 0xff] ^ rk[11]; + /* round 3: */ + t0 = Td0[s0 >> 24] ^ Td1[(s3 >> 16) & 0xff] ^ Td2[(s2 >> 8) & 0xff] ^ Td3[s1 & 0xff] ^ rk[12]; + t1 = Td0[s1 >> 24] ^ Td1[(s0 >> 16) & 0xff] ^ Td2[(s3 >> 8) & 0xff] ^ Td3[s2 & 0xff] ^ rk[13]; + t2 = Td0[s2 >> 24] ^ Td1[(s1 >> 16) & 0xff] ^ Td2[(s0 >> 8) & 0xff] ^ Td3[s3 & 0xff] ^ rk[14]; + t3 = Td0[s3 >> 24] ^ Td1[(s2 >> 16) & 0xff] ^ Td2[(s1 >> 8) & 0xff] ^ Td3[s0 & 0xff] ^ rk[15]; + /* round 4: */ + s0 = Td0[t0 >> 24] ^ Td1[(t3 >> 16) & 0xff] ^ Td2[(t2 >> 8) & 0xff] ^ Td3[t1 & 0xff] ^ rk[16]; + s1 = Td0[t1 >> 24] ^ Td1[(t0 >> 16) & 0xff] ^ Td2[(t3 >> 8) & 0xff] ^ Td3[t2 & 0xff] ^ rk[17]; + s2 = Td0[t2 >> 24] ^ Td1[(t1 >> 16) & 0xff] ^ Td2[(t0 >> 8) & 0xff] ^ Td3[t3 & 0xff] ^ rk[18]; + s3 = Td0[t3 >> 24] ^ Td1[(t2 >> 16) & 0xff] ^ Td2[(t1 >> 8) & 0xff] ^ Td3[t0 & 0xff] ^ rk[19]; + /* round 5: */ + t0 = Td0[s0 >> 24] ^ Td1[(s3 >> 16) & 0xff] ^ Td2[(s2 >> 8) & 0xff] ^ Td3[s1 & 0xff] ^ rk[20]; + t1 = Td0[s1 >> 24] ^ Td1[(s0 >> 16) & 0xff] ^ Td2[(s3 >> 8) & 0xff] ^ Td3[s2 & 0xff] ^ rk[21]; + t2 = Td0[s2 >> 24] ^ Td1[(s1 >> 16) & 0xff] ^ Td2[(s0 >> 8) & 0xff] ^ Td3[s3 & 0xff] ^ rk[22]; + t3 = Td0[s3 >> 24] ^ Td1[(s2 >> 16) & 0xff] ^ Td2[(s1 >> 8) & 0xff] ^ Td3[s0 & 0xff] ^ rk[23]; + /* round 6: */ + s0 = Td0[t0 >> 24] ^ Td1[(t3 >> 16) & 0xff] ^ Td2[(t2 >> 8) & 0xff] ^ Td3[t1 & 0xff] ^ rk[24]; + s1 = Td0[t1 >> 24] ^ Td1[(t0 >> 16) & 0xff] ^ Td2[(t3 >> 8) & 0xff] ^ Td3[t2 & 0xff] ^ rk[25]; + s2 = Td0[t2 >> 24] ^ Td1[(t1 >> 16) & 0xff] ^ Td2[(t0 >> 8) & 0xff] ^ Td3[t3 & 0xff] ^ rk[26]; + s3 = Td0[t3 >> 24] ^ Td1[(t2 >> 16) & 0xff] ^ Td2[(t1 >> 8) & 0xff] ^ Td3[t0 & 0xff] ^ rk[27]; + /* round 7: */ + t0 = Td0[s0 >> 24] ^ Td1[(s3 >> 16) & 0xff] ^ Td2[(s2 >> 8) & 0xff] ^ Td3[s1 & 0xff] ^ rk[28]; + t1 = Td0[s1 >> 24] ^ Td1[(s0 >> 16) & 0xff] ^ Td2[(s3 >> 8) & 0xff] ^ Td3[s2 & 0xff] ^ rk[29]; + t2 = Td0[s2 >> 24] ^ Td1[(s1 >> 16) & 0xff] ^ Td2[(s0 >> 8) & 0xff] ^ Td3[s3 & 0xff] ^ rk[30]; + t3 = Td0[s3 >> 24] ^ Td1[(s2 >> 16) & 0xff] ^ Td2[(s1 >> 8) & 0xff] ^ Td3[s0 & 0xff] ^ rk[31]; + /* round 8: */ + s0 = Td0[t0 >> 24] ^ Td1[(t3 >> 16) & 0xff] ^ Td2[(t2 >> 8) & 0xff] ^ Td3[t1 & 0xff] ^ rk[32]; + s1 = Td0[t1 >> 24] ^ Td1[(t0 >> 16) & 0xff] ^ Td2[(t3 >> 8) & 0xff] ^ Td3[t2 & 0xff] ^ rk[33]; + s2 = Td0[t2 >> 24] ^ Td1[(t1 >> 16) & 0xff] ^ Td2[(t0 >> 8) & 0xff] ^ Td3[t3 & 0xff] ^ rk[34]; + s3 = Td0[t3 >> 24] ^ Td1[(t2 >> 16) & 0xff] ^ Td2[(t1 >> 8) & 0xff] ^ Td3[t0 & 0xff] ^ rk[35]; + /* round 9: */ + t0 = Td0[s0 >> 24] ^ Td1[(s3 >> 16) & 0xff] ^ Td2[(s2 >> 8) & 0xff] ^ Td3[s1 & 0xff] ^ rk[36]; + t1 = Td0[s1 >> 24] ^ Td1[(s0 >> 16) & 0xff] ^ Td2[(s3 >> 8) & 0xff] ^ Td3[s2 & 0xff] ^ rk[37]; + t2 = Td0[s2 >> 24] ^ Td1[(s1 >> 16) & 0xff] ^ Td2[(s0 >> 8) & 0xff] ^ Td3[s3 & 0xff] ^ rk[38]; + t3 = Td0[s3 >> 24] ^ Td1[(s2 >> 16) & 0xff] ^ Td2[(s1 >> 8) & 0xff] ^ Td3[s0 & 0xff] ^ rk[39]; + if (key->rounds > 10) { + /* round 10: */ + s0 = Td0[t0 >> 24] ^ Td1[(t3 >> 16) & 0xff] ^ Td2[(t2 >> 8) & 0xff] ^ Td3[t1 & 0xff] ^ rk[40]; + s1 = Td0[t1 >> 24] ^ Td1[(t0 >> 16) & 0xff] ^ Td2[(t3 >> 8) & 0xff] ^ Td3[t2 & 0xff] ^ rk[41]; + s2 = Td0[t2 >> 24] ^ Td1[(t1 >> 16) & 0xff] ^ Td2[(t0 >> 8) & 0xff] ^ Td3[t3 & 0xff] ^ rk[42]; + s3 = Td0[t3 >> 24] ^ Td1[(t2 >> 16) & 0xff] ^ Td2[(t1 >> 8) & 0xff] ^ Td3[t0 & 0xff] ^ rk[43]; + /* round 11: */ + t0 = Td0[s0 >> 24] ^ Td1[(s3 >> 16) & 0xff] ^ Td2[(s2 >> 8) & 0xff] ^ Td3[s1 & 0xff] ^ rk[44]; + t1 = Td0[s1 >> 24] ^ Td1[(s0 >> 16) & 0xff] ^ Td2[(s3 >> 8) & 0xff] ^ Td3[s2 & 0xff] ^ rk[45]; + t2 = Td0[s2 >> 24] ^ Td1[(s1 >> 16) & 0xff] ^ Td2[(s0 >> 8) & 0xff] ^ Td3[s3 & 0xff] ^ rk[46]; + t3 = Td0[s3 >> 24] ^ Td1[(s2 >> 16) & 0xff] ^ Td2[(s1 >> 8) & 0xff] ^ Td3[s0 & 0xff] ^ rk[47]; + if (key->rounds > 12) { + /* round 12: */ + s0 = Td0[t0 >> 24] ^ Td1[(t3 >> 16) & 0xff] ^ Td2[(t2 >> 8) & 0xff] ^ Td3[t1 & 0xff] ^ rk[48]; + s1 = Td0[t1 >> 24] ^ Td1[(t0 >> 16) & 0xff] ^ Td2[(t3 >> 8) & 0xff] ^ Td3[t2 & 0xff] ^ rk[49]; + s2 = Td0[t2 >> 24] ^ Td1[(t1 >> 16) & 0xff] ^ Td2[(t0 >> 8) & 0xff] ^ Td3[t3 & 0xff] ^ rk[50]; + s3 = Td0[t3 >> 24] ^ Td1[(t2 >> 16) & 0xff] ^ Td2[(t1 >> 8) & 0xff] ^ Td3[t0 & 0xff] ^ rk[51]; + /* round 13: */ + t0 = Td0[s0 >> 24] ^ Td1[(s3 >> 16) & 0xff] ^ Td2[(s2 >> 8) & 0xff] ^ Td3[s1 & 0xff] ^ rk[52]; + t1 = Td0[s1 >> 24] ^ Td1[(s0 >> 16) & 0xff] ^ Td2[(s3 >> 8) & 0xff] ^ Td3[s2 & 0xff] ^ rk[53]; + t2 = Td0[s2 >> 24] ^ Td1[(s1 >> 16) & 0xff] ^ Td2[(s0 >> 8) & 0xff] ^ Td3[s3 & 0xff] ^ rk[54]; + t3 = Td0[s3 >> 24] ^ Td1[(s2 >> 16) & 0xff] ^ Td2[(s1 >> 8) & 0xff] ^ Td3[s0 & 0xff] ^ rk[55]; + } + } + rk += key->rounds << 2; +#else /* !FULL_UNROLL */ + /* + * Nr - 1 full rounds: + */ + r = key->rounds >> 1; + for (;;) { + t0 = + Td0[(s0 >> 24) ] ^ + Td1[(s3 >> 16) & 0xff] ^ + Td2[(s2 >> 8) & 0xff] ^ + Td3[(s1 ) & 0xff] ^ + rk[4]; + t1 = + Td0[(s1 >> 24) ] ^ + Td1[(s0 >> 16) & 0xff] ^ + Td2[(s3 >> 8) & 0xff] ^ + Td3[(s2 ) & 0xff] ^ + rk[5]; + t2 = + Td0[(s2 >> 24) ] ^ + Td1[(s1 >> 16) & 0xff] ^ + Td2[(s0 >> 8) & 0xff] ^ + Td3[(s3 ) & 0xff] ^ + rk[6]; + t3 = + Td0[(s3 >> 24) ] ^ + Td1[(s2 >> 16) & 0xff] ^ + Td2[(s1 >> 8) & 0xff] ^ + Td3[(s0 ) & 0xff] ^ + rk[7]; + + rk += 8; + if (--r == 0) { + break; + } + + s0 = + Td0[(t0 >> 24) ] ^ + Td1[(t3 >> 16) & 0xff] ^ + Td2[(t2 >> 8) & 0xff] ^ + Td3[(t1 ) & 0xff] ^ + rk[0]; + s1 = + Td0[(t1 >> 24) ] ^ + Td1[(t0 >> 16) & 0xff] ^ + Td2[(t3 >> 8) & 0xff] ^ + Td3[(t2 ) & 0xff] ^ + rk[1]; + s2 = + Td0[(t2 >> 24) ] ^ + Td1[(t1 >> 16) & 0xff] ^ + Td2[(t0 >> 8) & 0xff] ^ + Td3[(t3 ) & 0xff] ^ + rk[2]; + s3 = + Td0[(t3 >> 24) ] ^ + Td1[(t2 >> 16) & 0xff] ^ + Td2[(t1 >> 8) & 0xff] ^ + Td3[(t0 ) & 0xff] ^ + rk[3]; + } +#endif /* ?FULL_UNROLL */ + /* + * apply last round and + * map cipher state to byte array block: + */ + s0 = + ((u32)Td4[(t0 >> 24) ] << 24) ^ + ((u32)Td4[(t3 >> 16) & 0xff] << 16) ^ + ((u32)Td4[(t2 >> 8) & 0xff] << 8) ^ + ((u32)Td4[(t1 ) & 0xff]) ^ + rk[0]; + PUTU32(out , s0); + s1 = + ((u32)Td4[(t1 >> 24) ] << 24) ^ + ((u32)Td4[(t0 >> 16) & 0xff] << 16) ^ + ((u32)Td4[(t3 >> 8) & 0xff] << 8) ^ + ((u32)Td4[(t2 ) & 0xff]) ^ + rk[1]; + PUTU32(out + 4, s1); + s2 = + ((u32)Td4[(t2 >> 24) ] << 24) ^ + ((u32)Td4[(t1 >> 16) & 0xff] << 16) ^ + ((u32)Td4[(t0 >> 8) & 0xff] << 8) ^ + ((u32)Td4[(t3 ) & 0xff]) ^ + rk[2]; + PUTU32(out + 8, s2); + s3 = + ((u32)Td4[(t3 >> 24) ] << 24) ^ + ((u32)Td4[(t2 >> 16) & 0xff] << 16) ^ + ((u32)Td4[(t1 >> 8) & 0xff] << 8) ^ + ((u32)Td4[(t0 ) & 0xff]) ^ + rk[3]; + PUTU32(out + 12, s3); +} + +#else /* AES_ASM */ + +static const u8 Te4[256] = { + 0x63U, 0x7cU, 0x77U, 0x7bU, 0xf2U, 0x6bU, 0x6fU, 0xc5U, + 0x30U, 0x01U, 0x67U, 0x2bU, 0xfeU, 0xd7U, 0xabU, 0x76U, + 0xcaU, 0x82U, 0xc9U, 0x7dU, 0xfaU, 0x59U, 0x47U, 0xf0U, + 0xadU, 0xd4U, 0xa2U, 0xafU, 0x9cU, 0xa4U, 0x72U, 0xc0U, + 0xb7U, 0xfdU, 0x93U, 0x26U, 0x36U, 0x3fU, 0xf7U, 0xccU, + 0x34U, 0xa5U, 0xe5U, 0xf1U, 0x71U, 0xd8U, 0x31U, 0x15U, + 0x04U, 0xc7U, 0x23U, 0xc3U, 0x18U, 0x96U, 0x05U, 0x9aU, + 0x07U, 0x12U, 0x80U, 0xe2U, 0xebU, 0x27U, 0xb2U, 0x75U, + 0x09U, 0x83U, 0x2cU, 0x1aU, 0x1bU, 0x6eU, 0x5aU, 0xa0U, + 0x52U, 0x3bU, 0xd6U, 0xb3U, 0x29U, 0xe3U, 0x2fU, 0x84U, + 0x53U, 0xd1U, 0x00U, 0xedU, 0x20U, 0xfcU, 0xb1U, 0x5bU, + 0x6aU, 0xcbU, 0xbeU, 0x39U, 0x4aU, 0x4cU, 0x58U, 0xcfU, + 0xd0U, 0xefU, 0xaaU, 0xfbU, 0x43U, 0x4dU, 0x33U, 0x85U, + 0x45U, 0xf9U, 0x02U, 0x7fU, 0x50U, 0x3cU, 0x9fU, 0xa8U, + 0x51U, 0xa3U, 0x40U, 0x8fU, 0x92U, 0x9dU, 0x38U, 0xf5U, + 0xbcU, 0xb6U, 0xdaU, 0x21U, 0x10U, 0xffU, 0xf3U, 0xd2U, + 0xcdU, 0x0cU, 0x13U, 0xecU, 0x5fU, 0x97U, 0x44U, 0x17U, + 0xc4U, 0xa7U, 0x7eU, 0x3dU, 0x64U, 0x5dU, 0x19U, 0x73U, + 0x60U, 0x81U, 0x4fU, 0xdcU, 0x22U, 0x2aU, 0x90U, 0x88U, + 0x46U, 0xeeU, 0xb8U, 0x14U, 0xdeU, 0x5eU, 0x0bU, 0xdbU, + 0xe0U, 0x32U, 0x3aU, 0x0aU, 0x49U, 0x06U, 0x24U, 0x5cU, + 0xc2U, 0xd3U, 0xacU, 0x62U, 0x91U, 0x95U, 0xe4U, 0x79U, + 0xe7U, 0xc8U, 0x37U, 0x6dU, 0x8dU, 0xd5U, 0x4eU, 0xa9U, + 0x6cU, 0x56U, 0xf4U, 0xeaU, 0x65U, 0x7aU, 0xaeU, 0x08U, + 0xbaU, 0x78U, 0x25U, 0x2eU, 0x1cU, 0xa6U, 0xb4U, 0xc6U, + 0xe8U, 0xddU, 0x74U, 0x1fU, 0x4bU, 0xbdU, 0x8bU, 0x8aU, + 0x70U, 0x3eU, 0xb5U, 0x66U, 0x48U, 0x03U, 0xf6U, 0x0eU, + 0x61U, 0x35U, 0x57U, 0xb9U, 0x86U, 0xc1U, 0x1dU, 0x9eU, + 0xe1U, 0xf8U, 0x98U, 0x11U, 0x69U, 0xd9U, 0x8eU, 0x94U, + 0x9bU, 0x1eU, 0x87U, 0xe9U, 0xceU, 0x55U, 0x28U, 0xdfU, + 0x8cU, 0xa1U, 0x89U, 0x0dU, 0xbfU, 0xe6U, 0x42U, 0x68U, + 0x41U, 0x99U, 0x2dU, 0x0fU, 0xb0U, 0x54U, 0xbbU, 0x16U +}; +static const u32 rcon[] = { + 0x01000000, 0x02000000, 0x04000000, 0x08000000, + 0x10000000, 0x20000000, 0x40000000, 0x80000000, + 0x1B000000, 0x36000000, /* for 128-bit blocks, Rijndael never uses more than 10 rcon values */ +}; + +/** + * Expand the cipher key into the encryption key schedule. + */ +int AES_set_encrypt_key(const unsigned char *userKey, const int bits, + AES_KEY *key) +{ + u32 *rk; + int i = 0; + u32 temp; + + if (!userKey || !key) + return -1; + if (bits != 128 && bits != 192 && bits != 256) + return -2; + + rk = key->rd_key; + + if (bits == 128) + key->rounds = 10; + else if (bits == 192) + key->rounds = 12; + else + key->rounds = 14; + + rk[0] = GETU32(userKey ); + rk[1] = GETU32(userKey + 4); + rk[2] = GETU32(userKey + 8); + rk[3] = GETU32(userKey + 12); + if (bits == 128) { + while (1) { + temp = rk[3]; + rk[4] = rk[0] ^ + ((u32)Te4[(temp >> 16) & 0xff] << 24) ^ + ((u32)Te4[(temp >> 8) & 0xff] << 16) ^ + ((u32)Te4[(temp ) & 0xff] << 8) ^ + ((u32)Te4[(temp >> 24) ]) ^ + rcon[i]; + rk[5] = rk[1] ^ rk[4]; + rk[6] = rk[2] ^ rk[5]; + rk[7] = rk[3] ^ rk[6]; + if (++i == 10) { + return 0; + } + rk += 4; + } + } + rk[4] = GETU32(userKey + 16); + rk[5] = GETU32(userKey + 20); + if (bits == 192) { + while (1) { + temp = rk[ 5]; + rk[ 6] = rk[ 0] ^ + ((u32)Te4[(temp >> 16) & 0xff] << 24) ^ + ((u32)Te4[(temp >> 8) & 0xff] << 16) ^ + ((u32)Te4[(temp ) & 0xff] << 8) ^ + ((u32)Te4[(temp >> 24) ]) ^ + rcon[i]; + rk[ 7] = rk[ 1] ^ rk[ 6]; + rk[ 8] = rk[ 2] ^ rk[ 7]; + rk[ 9] = rk[ 3] ^ rk[ 8]; + if (++i == 8) { + return 0; + } + rk[10] = rk[ 4] ^ rk[ 9]; + rk[11] = rk[ 5] ^ rk[10]; + rk += 6; + } + } + rk[6] = GETU32(userKey + 24); + rk[7] = GETU32(userKey + 28); + if (bits == 256) { + while (1) { + temp = rk[ 7]; + rk[ 8] = rk[ 0] ^ + ((u32)Te4[(temp >> 16) & 0xff] << 24) ^ + ((u32)Te4[(temp >> 8) & 0xff] << 16) ^ + ((u32)Te4[(temp ) & 0xff] << 8) ^ + ((u32)Te4[(temp >> 24) ]) ^ + rcon[i]; + rk[ 9] = rk[ 1] ^ rk[ 8]; + rk[10] = rk[ 2] ^ rk[ 9]; + rk[11] = rk[ 3] ^ rk[10]; + if (++i == 7) { + return 0; + } + temp = rk[11]; + rk[12] = rk[ 4] ^ + ((u32)Te4[(temp >> 24) ] << 24) ^ + ((u32)Te4[(temp >> 16) & 0xff] << 16) ^ + ((u32)Te4[(temp >> 8) & 0xff] << 8) ^ + ((u32)Te4[(temp ) & 0xff]); + rk[13] = rk[ 5] ^ rk[12]; + rk[14] = rk[ 6] ^ rk[13]; + rk[15] = rk[ 7] ^ rk[14]; + + rk += 8; + } + } + return 0; +} + +/** + * Expand the cipher key into the decryption key schedule. + */ +int AES_set_decrypt_key(const unsigned char *userKey, const int bits, + AES_KEY *key) +{ + + u32 *rk; + int i, j, status; + u32 temp; + + /* first, start with an encryption schedule */ + status = AES_set_encrypt_key(userKey, bits, key); + if (status < 0) + return status; + + rk = key->rd_key; + + /* invert the order of the round keys: */ + for (i = 0, j = 4*(key->rounds); i < j; i += 4, j -= 4) { + temp = rk[i ]; rk[i ] = rk[j ]; rk[j ] = temp; + temp = rk[i + 1]; rk[i + 1] = rk[j + 1]; rk[j + 1] = temp; + temp = rk[i + 2]; rk[i + 2] = rk[j + 2]; rk[j + 2] = temp; + temp = rk[i + 3]; rk[i + 3] = rk[j + 3]; rk[j + 3] = temp; + } + /* apply the inverse MixColumn transform to all round keys but the first and the last: */ + for (i = 1; i < (key->rounds); i++) { + rk += 4; + for (j = 0; j < 4; j++) { + u32 tp1, tp2, tp4, tp8, tp9, tpb, tpd, tpe, m; + + tp1 = rk[j]; + m = tp1 & 0x80808080; + tp2 = ((tp1 & 0x7f7f7f7f) << 1) ^ + ((m - (m >> 7)) & 0x1b1b1b1b); + m = tp2 & 0x80808080; + tp4 = ((tp2 & 0x7f7f7f7f) << 1) ^ + ((m - (m >> 7)) & 0x1b1b1b1b); + m = tp4 & 0x80808080; + tp8 = ((tp4 & 0x7f7f7f7f) << 1) ^ + ((m - (m >> 7)) & 0x1b1b1b1b); + tp9 = tp8 ^ tp1; + tpb = tp9 ^ tp2; + tpd = tp9 ^ tp4; + tpe = tp8 ^ tp4 ^ tp2; +#if defined(ROTATE) + rk[j] = tpe ^ ROTATE(tpd,16) ^ + ROTATE(tp9,24) ^ ROTATE(tpb,8); +#else + rk[j] = tpe ^ (tpd >> 16) ^ (tpd << 16) ^ + (tp9 >> 8) ^ (tp9 << 24) ^ + (tpb >> 24) ^ (tpb << 8); +#endif + } + } + return 0; +} + +#endif /* AES_ASM */ diff --git a/openssl-1.1.0h/crypto/aes/aes_ecb.c b/openssl-1.1.0h/crypto/aes/aes_ecb.c new file mode 100644 index 0000000..29bfc1a --- /dev/null +++ b/openssl-1.1.0h/crypto/aes/aes_ecb.c @@ -0,0 +1,26 @@ +/* + * Copyright 2002-2016 The OpenSSL Project Authors. All Rights Reserved. + * + * Licensed under the OpenSSL license (the "License"). You may not use + * this file except in compliance with the License. You can obtain a copy + * in the file LICENSE in the source distribution or at + * https://www.openssl.org/source/license.html + */ + +#include + +#include +#include "aes_locl.h" + +void AES_ecb_encrypt(const unsigned char *in, unsigned char *out, + const AES_KEY *key, const int enc) +{ + + assert(in && out && key); + assert((AES_ENCRYPT == enc) || (AES_DECRYPT == enc)); + + if (AES_ENCRYPT == enc) + AES_encrypt(in, out, key); + else + AES_decrypt(in, out, key); +} diff --git a/openssl-1.1.0h/crypto/aes/aes_ige.c b/openssl-1.1.0h/crypto/aes/aes_ige.c new file mode 100644 index 0000000..75f796c --- /dev/null +++ b/openssl-1.1.0h/crypto/aes/aes_ige.c @@ -0,0 +1,284 @@ +/* + * Copyright 2006-2016 The OpenSSL Project Authors. All Rights Reserved. + * + * Licensed under the OpenSSL license (the "License"). You may not use + * this file except in compliance with the License. You can obtain a copy + * in the file LICENSE in the source distribution or at + * https://www.openssl.org/source/license.html + */ + +#include "internal/cryptlib.h" + +#include +#include "aes_locl.h" + +#define N_WORDS (AES_BLOCK_SIZE / sizeof(unsigned long)) +typedef struct { + unsigned long data[N_WORDS]; +} aes_block_t; + +/* XXX: probably some better way to do this */ +#if defined(__i386__) || defined(__x86_64__) +# define UNALIGNED_MEMOPS_ARE_FAST 1 +#else +# define UNALIGNED_MEMOPS_ARE_FAST 0 +#endif + +#if UNALIGNED_MEMOPS_ARE_FAST +# define load_block(d, s) (d) = *(const aes_block_t *)(s) +# define store_block(d, s) *(aes_block_t *)(d) = (s) +#else +# define load_block(d, s) memcpy((d).data, (s), AES_BLOCK_SIZE) +# define store_block(d, s) memcpy((d), (s).data, AES_BLOCK_SIZE) +#endif + +/* N.B. The IV for this mode is _twice_ the block size */ + +void AES_ige_encrypt(const unsigned char *in, unsigned char *out, + size_t length, const AES_KEY *key, + unsigned char *ivec, const int enc) +{ + size_t n; + size_t len = length; + + if (length == 0) + return; + + OPENSSL_assert(in && out && key && ivec); + OPENSSL_assert((AES_ENCRYPT == enc) || (AES_DECRYPT == enc)); + OPENSSL_assert((length % AES_BLOCK_SIZE) == 0); + + len = length / AES_BLOCK_SIZE; + + if (AES_ENCRYPT == enc) { + if (in != out && + (UNALIGNED_MEMOPS_ARE_FAST + || ((size_t)in | (size_t)out | (size_t)ivec) % sizeof(long) == + 0)) { + aes_block_t *ivp = (aes_block_t *) ivec; + aes_block_t *iv2p = (aes_block_t *) (ivec + AES_BLOCK_SIZE); + + while (len) { + aes_block_t *inp = (aes_block_t *) in; + aes_block_t *outp = (aes_block_t *) out; + + for (n = 0; n < N_WORDS; ++n) + outp->data[n] = inp->data[n] ^ ivp->data[n]; + AES_encrypt((unsigned char *)outp->data, + (unsigned char *)outp->data, key); + for (n = 0; n < N_WORDS; ++n) + outp->data[n] ^= iv2p->data[n]; + ivp = outp; + iv2p = inp; + --len; + in += AES_BLOCK_SIZE; + out += AES_BLOCK_SIZE; + } + memcpy(ivec, ivp->data, AES_BLOCK_SIZE); + memcpy(ivec + AES_BLOCK_SIZE, iv2p->data, AES_BLOCK_SIZE); + } else { + aes_block_t tmp, tmp2; + aes_block_t iv; + aes_block_t iv2; + + load_block(iv, ivec); + load_block(iv2, ivec + AES_BLOCK_SIZE); + + while (len) { + load_block(tmp, in); + for (n = 0; n < N_WORDS; ++n) + tmp2.data[n] = tmp.data[n] ^ iv.data[n]; + AES_encrypt((unsigned char *)tmp2.data, + (unsigned char *)tmp2.data, key); + for (n = 0; n < N_WORDS; ++n) + tmp2.data[n] ^= iv2.data[n]; + store_block(out, tmp2); + iv = tmp2; + iv2 = tmp; + --len; + in += AES_BLOCK_SIZE; + out += AES_BLOCK_SIZE; + } + memcpy(ivec, iv.data, AES_BLOCK_SIZE); + memcpy(ivec + AES_BLOCK_SIZE, iv2.data, AES_BLOCK_SIZE); + } + } else { + if (in != out && + (UNALIGNED_MEMOPS_ARE_FAST + || ((size_t)in | (size_t)out | (size_t)ivec) % sizeof(long) == + 0)) { + aes_block_t *ivp = (aes_block_t *) ivec; + aes_block_t *iv2p = (aes_block_t *) (ivec + AES_BLOCK_SIZE); + + while (len) { + aes_block_t tmp; + aes_block_t *inp = (aes_block_t *) in; + aes_block_t *outp = (aes_block_t *) out; + + for (n = 0; n < N_WORDS; ++n) + tmp.data[n] = inp->data[n] ^ iv2p->data[n]; + AES_decrypt((unsigned char *)tmp.data, + (unsigned char *)outp->data, key); + for (n = 0; n < N_WORDS; ++n) + outp->data[n] ^= ivp->data[n]; + ivp = inp; + iv2p = outp; + --len; + in += AES_BLOCK_SIZE; + out += AES_BLOCK_SIZE; + } + memcpy(ivec, ivp->data, AES_BLOCK_SIZE); + memcpy(ivec + AES_BLOCK_SIZE, iv2p->data, AES_BLOCK_SIZE); + } else { + aes_block_t tmp, tmp2; + aes_block_t iv; + aes_block_t iv2; + + load_block(iv, ivec); + load_block(iv2, ivec + AES_BLOCK_SIZE); + + while (len) { + load_block(tmp, in); + tmp2 = tmp; + for (n = 0; n < N_WORDS; ++n) + tmp.data[n] ^= iv2.data[n]; + AES_decrypt((unsigned char *)tmp.data, + (unsigned char *)tmp.data, key); + for (n = 0; n < N_WORDS; ++n) + tmp.data[n] ^= iv.data[n]; + store_block(out, tmp); + iv = tmp2; + iv2 = tmp; + --len; + in += AES_BLOCK_SIZE; + out += AES_BLOCK_SIZE; + } + memcpy(ivec, iv.data, AES_BLOCK_SIZE); + memcpy(ivec + AES_BLOCK_SIZE, iv2.data, AES_BLOCK_SIZE); + } + } +} + +/* + * Note that its effectively impossible to do biIGE in anything other + * than a single pass, so no provision is made for chaining. + */ + +/* N.B. The IV for this mode is _four times_ the block size */ + +void AES_bi_ige_encrypt(const unsigned char *in, unsigned char *out, + size_t length, const AES_KEY *key, + const AES_KEY *key2, const unsigned char *ivec, + const int enc) +{ + size_t n; + size_t len = length; + unsigned char tmp[AES_BLOCK_SIZE]; + unsigned char tmp2[AES_BLOCK_SIZE]; + unsigned char tmp3[AES_BLOCK_SIZE]; + unsigned char prev[AES_BLOCK_SIZE]; + const unsigned char *iv; + const unsigned char *iv2; + + OPENSSL_assert(in && out && key && ivec); + OPENSSL_assert((AES_ENCRYPT == enc) || (AES_DECRYPT == enc)); + OPENSSL_assert((length % AES_BLOCK_SIZE) == 0); + + if (AES_ENCRYPT == enc) { + /* + * XXX: Do a separate case for when in != out (strictly should check + * for overlap, too) + */ + + /* First the forward pass */ + iv = ivec; + iv2 = ivec + AES_BLOCK_SIZE; + while (len >= AES_BLOCK_SIZE) { + for (n = 0; n < AES_BLOCK_SIZE; ++n) + out[n] = in[n] ^ iv[n]; + AES_encrypt(out, out, key); + for (n = 0; n < AES_BLOCK_SIZE; ++n) + out[n] ^= iv2[n]; + iv = out; + memcpy(prev, in, AES_BLOCK_SIZE); + iv2 = prev; + len -= AES_BLOCK_SIZE; + in += AES_BLOCK_SIZE; + out += AES_BLOCK_SIZE; + } + + /* And now backwards */ + iv = ivec + AES_BLOCK_SIZE * 2; + iv2 = ivec + AES_BLOCK_SIZE * 3; + len = length; + while (len >= AES_BLOCK_SIZE) { + out -= AES_BLOCK_SIZE; + /* + * XXX: reduce copies by alternating between buffers + */ + memcpy(tmp, out, AES_BLOCK_SIZE); + for (n = 0; n < AES_BLOCK_SIZE; ++n) + out[n] ^= iv[n]; + /* + * hexdump(stdout, "out ^ iv", out, AES_BLOCK_SIZE); + */ + AES_encrypt(out, out, key); + /* + * hexdump(stdout,"enc", out, AES_BLOCK_SIZE); + */ + /* + * hexdump(stdout,"iv2", iv2, AES_BLOCK_SIZE); + */ + for (n = 0; n < AES_BLOCK_SIZE; ++n) + out[n] ^= iv2[n]; + /* + * hexdump(stdout,"out", out, AES_BLOCK_SIZE); + */ + iv = out; + memcpy(prev, tmp, AES_BLOCK_SIZE); + iv2 = prev; + len -= AES_BLOCK_SIZE; + } + } else { + /* First backwards */ + iv = ivec + AES_BLOCK_SIZE * 2; + iv2 = ivec + AES_BLOCK_SIZE * 3; + in += length; + out += length; + while (len >= AES_BLOCK_SIZE) { + in -= AES_BLOCK_SIZE; + out -= AES_BLOCK_SIZE; + memcpy(tmp, in, AES_BLOCK_SIZE); + memcpy(tmp2, in, AES_BLOCK_SIZE); + for (n = 0; n < AES_BLOCK_SIZE; ++n) + tmp[n] ^= iv2[n]; + AES_decrypt(tmp, out, key); + for (n = 0; n < AES_BLOCK_SIZE; ++n) + out[n] ^= iv[n]; + memcpy(tmp3, tmp2, AES_BLOCK_SIZE); + iv = tmp3; + iv2 = out; + len -= AES_BLOCK_SIZE; + } + + /* And now forwards */ + iv = ivec; + iv2 = ivec + AES_BLOCK_SIZE; + len = length; + while (len >= AES_BLOCK_SIZE) { + memcpy(tmp, out, AES_BLOCK_SIZE); + memcpy(tmp2, out, AES_BLOCK_SIZE); + for (n = 0; n < AES_BLOCK_SIZE; ++n) + tmp[n] ^= iv2[n]; + AES_decrypt(tmp, out, key); + for (n = 0; n < AES_BLOCK_SIZE; ++n) + out[n] ^= iv[n]; + memcpy(tmp3, tmp2, AES_BLOCK_SIZE); + iv = tmp3; + iv2 = out; + len -= AES_BLOCK_SIZE; + in += AES_BLOCK_SIZE; + out += AES_BLOCK_SIZE; + } + } +} diff --git a/openssl-1.1.0h/crypto/aes/aes_locl.h b/openssl-1.1.0h/crypto/aes/aes_locl.h new file mode 100644 index 0000000..adee29d --- /dev/null +++ b/openssl-1.1.0h/crypto/aes/aes_locl.h @@ -0,0 +1,42 @@ +/* + * Copyright 2002-2016 The OpenSSL Project Authors. All Rights Reserved. + * + * Licensed under the OpenSSL license (the "License"). You may not use + * this file except in compliance with the License. You can obtain a copy + * in the file LICENSE in the source distribution or at + * https://www.openssl.org/source/license.html + */ + +#ifndef HEADER_AES_LOCL_H +# define HEADER_AES_LOCL_H + +# include +# include +# include +# include + +# if defined(_MSC_VER) && (defined(_M_IX86) || defined(_M_AMD64) || defined(_M_X64)) +# define SWAP(x) (_lrotl(x, 8) & 0x00ff00ff | _lrotr(x, 8) & 0xff00ff00) +# define GETU32(p) SWAP(*((u32 *)(p))) +# define PUTU32(ct, st) { *((u32 *)(ct)) = SWAP((st)); } +# else +# define GETU32(pt) (((u32)(pt)[0] << 24) ^ ((u32)(pt)[1] << 16) ^ ((u32)(pt)[2] << 8) ^ ((u32)(pt)[3])) +# define PUTU32(ct, st) { (ct)[0] = (u8)((st) >> 24); (ct)[1] = (u8)((st) >> 16); (ct)[2] = (u8)((st) >> 8); (ct)[3] = (u8)(st); } +# endif + +# ifdef AES_LONG +typedef unsigned long u32; +# else +typedef unsigned int u32; +# endif +typedef unsigned short u16; +typedef unsigned char u8; + +# define MAXKC (256/32) +# define MAXKB (256/8) +# define MAXNR 14 + +/* This controls loop-unrolling in aes_core.c */ +# undef FULL_UNROLL + +#endif /* !HEADER_AES_LOCL_H */ diff --git a/openssl-1.1.0h/crypto/aes/aes_misc.c b/openssl-1.1.0h/crypto/aes/aes_misc.c new file mode 100644 index 0000000..7403c84 --- /dev/null +++ b/openssl-1.1.0h/crypto/aes/aes_misc.c @@ -0,0 +1,21 @@ +/* + * Copyright 2002-2016 The OpenSSL Project Authors. All Rights Reserved. + * + * Licensed under the OpenSSL license (the "License"). You may not use + * this file except in compliance with the License. You can obtain a copy + * in the file LICENSE in the source distribution or at + * https://www.openssl.org/source/license.html + */ + +#include +#include +#include "aes_locl.h" + +const char *AES_options(void) +{ +#ifdef FULL_UNROLL + return "aes(full)"; +#else + return "aes(partial)"; +#endif +} diff --git a/openssl-1.1.0h/crypto/aes/aes_ofb.c b/openssl-1.1.0h/crypto/aes/aes_ofb.c new file mode 100644 index 0000000..215b538 --- /dev/null +++ b/openssl-1.1.0h/crypto/aes/aes_ofb.c @@ -0,0 +1,19 @@ +/* + * Copyright 2002-2016 The OpenSSL Project Authors. All Rights Reserved. + * + * Licensed under the OpenSSL license (the "License"). You may not use + * this file except in compliance with the License. You can obtain a copy + * in the file LICENSE in the source distribution or at + * https://www.openssl.org/source/license.html + */ + +#include +#include + +void AES_ofb128_encrypt(const unsigned char *in, unsigned char *out, + size_t length, const AES_KEY *key, + unsigned char *ivec, int *num) +{ + CRYPTO_ofb128_encrypt(in, out, length, key, ivec, num, + (block128_f) AES_encrypt); +} diff --git a/openssl-1.1.0h/crypto/aes/aes_wrap.c b/openssl-1.1.0h/crypto/aes/aes_wrap.c new file mode 100644 index 0000000..cae0b21 --- /dev/null +++ b/openssl-1.1.0h/crypto/aes/aes_wrap.c @@ -0,0 +1,27 @@ +/* + * Copyright 2008-2016 The OpenSSL Project Authors. All Rights Reserved. + * + * Licensed under the OpenSSL license (the "License"). You may not use + * this file except in compliance with the License. You can obtain a copy + * in the file LICENSE in the source distribution or at + * https://www.openssl.org/source/license.html + */ + +#include "internal/cryptlib.h" +#include +#include + +int AES_wrap_key(AES_KEY *key, const unsigned char *iv, + unsigned char *out, + const unsigned char *in, unsigned int inlen) +{ + return CRYPTO_128_wrap(key, iv, out, in, inlen, (block128_f) AES_encrypt); +} + +int AES_unwrap_key(AES_KEY *key, const unsigned char *iv, + unsigned char *out, + const unsigned char *in, unsigned int inlen) +{ + return CRYPTO_128_unwrap(key, iv, out, in, inlen, + (block128_f) AES_decrypt); +} diff --git a/openssl-1.1.0h/crypto/aes/aes_x86core.c b/openssl-1.1.0h/crypto/aes/aes_x86core.c new file mode 100644 index 0000000..95b49bb --- /dev/null +++ b/openssl-1.1.0h/crypto/aes/aes_x86core.c @@ -0,0 +1,1075 @@ +/* + * Copyright 2006-2016 The OpenSSL Project Authors. All Rights Reserved. + * + * Licensed under the OpenSSL license (the "License"). You may not use + * this file except in compliance with the License. You can obtain a copy + * in the file LICENSE in the source distribution or at + * https://www.openssl.org/source/license.html + */ + +/** + * rijndael-alg-fst.c + * + * @version 3.0 (December 2000) + * + * Optimised ANSI C code for the Rijndael cipher (now AES) + * + * @author Vincent Rijmen + * @author Antoon Bosselaers + * @author Paulo Barreto + * + * This code is hereby placed in the public domain. + * + * THIS SOFTWARE IS PROVIDED BY THE AUTHORS ''AS IS'' AND ANY EXPRESS + * OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED + * WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE + * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHORS OR CONTRIBUTORS BE + * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR + * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF + * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR + * BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, + * WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE + * OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, + * EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. + */ + +/* + * This is experimental x86[_64] derivative. It assumes little-endian + * byte order and expects CPU to sustain unaligned memory references. + * It is used as playground for cache-time attack mitigations and + * serves as reference C implementation for x86[_64] assembler. + * + * + */ + + +#include + +#include +#include +#include "aes_locl.h" + +/* + * These two parameters control which table, 256-byte or 2KB, is + * referenced in outer and respectively inner rounds. + */ +#define AES_COMPACT_IN_OUTER_ROUNDS +#ifdef AES_COMPACT_IN_OUTER_ROUNDS +/* AES_COMPACT_IN_OUTER_ROUNDS costs ~30% in performance, while + * adding AES_COMPACT_IN_INNER_ROUNDS reduces benchmark *further* + * by factor of ~2. */ +# undef AES_COMPACT_IN_INNER_ROUNDS +#endif + +#if 1 +static void prefetch256(const void *table) +{ + volatile unsigned long *t=(void *)table,ret; + unsigned long sum; + int i; + + /* 32 is common least cache-line size */ + for (sum=0,i=0;i<256/sizeof(t[0]);i+=32/sizeof(t[0])) sum ^= t[i]; + + ret = sum; +} +#else +# define prefetch256(t) +#endif + +#undef GETU32 +#define GETU32(p) (*((u32*)(p))) + +#if (defined(_WIN32) || defined(_WIN64)) && !defined(__MINGW32__) +typedef unsigned __int64 u64; +#define U64(C) C##UI64 +#elif defined(__arch64__) +typedef unsigned long u64; +#define U64(C) C##UL +#else +typedef unsigned long long u64; +#define U64(C) C##ULL +#endif + +#undef ROTATE +#if defined(_MSC_VER) +# define ROTATE(a,n) _lrotl(a,n) +#elif defined(__ICC) +# define ROTATE(a,n) _rotl(a,n) +#elif defined(__GNUC__) && __GNUC__>=2 +# if defined(__i386) || defined(__i386__) || defined(__x86_64) || defined(__x86_64__) +# define ROTATE(a,n) ({ register unsigned int ret; \ + asm ( \ + "roll %1,%0" \ + : "=r"(ret) \ + : "I"(n), "0"(a) \ + : "cc"); \ + ret; \ + }) +# endif +#endif +/*- +Te [x] = S [x].[02, 01, 01, 03, 02, 01, 01, 03]; +Te0[x] = S [x].[02, 01, 01, 03]; +Te1[x] = S [x].[03, 02, 01, 01]; +Te2[x] = S [x].[01, 03, 02, 01]; +Te3[x] = S [x].[01, 01, 03, 02]; +*/ +#define Te0 (u32)((u64*)((u8*)Te+0)) +#define Te1 (u32)((u64*)((u8*)Te+3)) +#define Te2 (u32)((u64*)((u8*)Te+2)) +#define Te3 (u32)((u64*)((u8*)Te+1)) +/*- +Td [x] = Si[x].[0e, 09, 0d, 0b, 0e, 09, 0d, 0b]; +Td0[x] = Si[x].[0e, 09, 0d, 0b]; +Td1[x] = Si[x].[0b, 0e, 09, 0d]; +Td2[x] = Si[x].[0d, 0b, 0e, 09]; +Td3[x] = Si[x].[09, 0d, 0b, 0e]; +Td4[x] = Si[x].[01]; +*/ +#define Td0 (u32)((u64*)((u8*)Td+0)) +#define Td1 (u32)((u64*)((u8*)Td+3)) +#define Td2 (u32)((u64*)((u8*)Td+2)) +#define Td3 (u32)((u64*)((u8*)Td+1)) + +static const u64 Te[256] = { + U64(0xa56363c6a56363c6), U64(0x847c7cf8847c7cf8), + U64(0x997777ee997777ee), U64(0x8d7b7bf68d7b7bf6), + U64(0x0df2f2ff0df2f2ff), U64(0xbd6b6bd6bd6b6bd6), + U64(0xb16f6fdeb16f6fde), U64(0x54c5c59154c5c591), + U64(0x5030306050303060), U64(0x0301010203010102), + U64(0xa96767cea96767ce), U64(0x7d2b2b567d2b2b56), + U64(0x19fefee719fefee7), U64(0x62d7d7b562d7d7b5), + U64(0xe6abab4de6abab4d), U64(0x9a7676ec9a7676ec), + U64(0x45caca8f45caca8f), U64(0x9d82821f9d82821f), + U64(0x40c9c98940c9c989), U64(0x877d7dfa877d7dfa), + U64(0x15fafaef15fafaef), U64(0xeb5959b2eb5959b2), + U64(0xc947478ec947478e), U64(0x0bf0f0fb0bf0f0fb), + U64(0xecadad41ecadad41), U64(0x67d4d4b367d4d4b3), + U64(0xfda2a25ffda2a25f), U64(0xeaafaf45eaafaf45), + U64(0xbf9c9c23bf9c9c23), U64(0xf7a4a453f7a4a453), + U64(0x967272e4967272e4), U64(0x5bc0c09b5bc0c09b), + U64(0xc2b7b775c2b7b775), U64(0x1cfdfde11cfdfde1), + U64(0xae93933dae93933d), U64(0x6a26264c6a26264c), + U64(0x5a36366c5a36366c), U64(0x413f3f7e413f3f7e), + U64(0x02f7f7f502f7f7f5), U64(0x4fcccc834fcccc83), + U64(0x5c3434685c343468), U64(0xf4a5a551f4a5a551), + U64(0x34e5e5d134e5e5d1), U64(0x08f1f1f908f1f1f9), + U64(0x937171e2937171e2), U64(0x73d8d8ab73d8d8ab), + U64(0x5331316253313162), U64(0x3f15152a3f15152a), + U64(0x0c0404080c040408), U64(0x52c7c79552c7c795), + U64(0x6523234665232346), U64(0x5ec3c39d5ec3c39d), + U64(0x2818183028181830), U64(0xa1969637a1969637), + U64(0x0f05050a0f05050a), U64(0xb59a9a2fb59a9a2f), + U64(0x0907070e0907070e), U64(0x3612122436121224), + U64(0x9b80801b9b80801b), U64(0x3de2e2df3de2e2df), + U64(0x26ebebcd26ebebcd), U64(0x6927274e6927274e), + U64(0xcdb2b27fcdb2b27f), U64(0x9f7575ea9f7575ea), + U64(0x1b0909121b090912), U64(0x9e83831d9e83831d), + U64(0x742c2c58742c2c58), U64(0x2e1a1a342e1a1a34), + U64(0x2d1b1b362d1b1b36), U64(0xb26e6edcb26e6edc), + U64(0xee5a5ab4ee5a5ab4), U64(0xfba0a05bfba0a05b), + U64(0xf65252a4f65252a4), U64(0x4d3b3b764d3b3b76), + U64(0x61d6d6b761d6d6b7), U64(0xceb3b37dceb3b37d), + U64(0x7b2929527b292952), U64(0x3ee3e3dd3ee3e3dd), + U64(0x712f2f5e712f2f5e), U64(0x9784841397848413), + U64(0xf55353a6f55353a6), U64(0x68d1d1b968d1d1b9), + U64(0x0000000000000000), U64(0x2cededc12cededc1), + U64(0x6020204060202040), U64(0x1ffcfce31ffcfce3), + U64(0xc8b1b179c8b1b179), U64(0xed5b5bb6ed5b5bb6), + U64(0xbe6a6ad4be6a6ad4), U64(0x46cbcb8d46cbcb8d), + U64(0xd9bebe67d9bebe67), U64(0x4b3939724b393972), + U64(0xde4a4a94de4a4a94), U64(0xd44c4c98d44c4c98), + U64(0xe85858b0e85858b0), U64(0x4acfcf854acfcf85), + U64(0x6bd0d0bb6bd0d0bb), U64(0x2aefefc52aefefc5), + U64(0xe5aaaa4fe5aaaa4f), U64(0x16fbfbed16fbfbed), + U64(0xc5434386c5434386), U64(0xd74d4d9ad74d4d9a), + U64(0x5533336655333366), U64(0x9485851194858511), + U64(0xcf45458acf45458a), U64(0x10f9f9e910f9f9e9), + U64(0x0602020406020204), U64(0x817f7ffe817f7ffe), + U64(0xf05050a0f05050a0), U64(0x443c3c78443c3c78), + U64(0xba9f9f25ba9f9f25), U64(0xe3a8a84be3a8a84b), + U64(0xf35151a2f35151a2), U64(0xfea3a35dfea3a35d), + U64(0xc0404080c0404080), U64(0x8a8f8f058a8f8f05), + U64(0xad92923fad92923f), U64(0xbc9d9d21bc9d9d21), + U64(0x4838387048383870), U64(0x04f5f5f104f5f5f1), + U64(0xdfbcbc63dfbcbc63), U64(0xc1b6b677c1b6b677), + U64(0x75dadaaf75dadaaf), U64(0x6321214263212142), + U64(0x3010102030101020), U64(0x1affffe51affffe5), + U64(0x0ef3f3fd0ef3f3fd), U64(0x6dd2d2bf6dd2d2bf), + U64(0x4ccdcd814ccdcd81), U64(0x140c0c18140c0c18), + U64(0x3513132635131326), U64(0x2fececc32fececc3), + U64(0xe15f5fbee15f5fbe), U64(0xa2979735a2979735), + U64(0xcc444488cc444488), U64(0x3917172e3917172e), + U64(0x57c4c49357c4c493), U64(0xf2a7a755f2a7a755), + U64(0x827e7efc827e7efc), U64(0x473d3d7a473d3d7a), + U64(0xac6464c8ac6464c8), U64(0xe75d5dbae75d5dba), + U64(0x2b1919322b191932), U64(0x957373e6957373e6), + U64(0xa06060c0a06060c0), U64(0x9881811998818119), + U64(0xd14f4f9ed14f4f9e), U64(0x7fdcdca37fdcdca3), + U64(0x6622224466222244), U64(0x7e2a2a547e2a2a54), + U64(0xab90903bab90903b), U64(0x8388880b8388880b), + U64(0xca46468cca46468c), U64(0x29eeeec729eeeec7), + U64(0xd3b8b86bd3b8b86b), U64(0x3c1414283c141428), + U64(0x79dedea779dedea7), U64(0xe25e5ebce25e5ebc), + U64(0x1d0b0b161d0b0b16), U64(0x76dbdbad76dbdbad), + U64(0x3be0e0db3be0e0db), U64(0x5632326456323264), + U64(0x4e3a3a744e3a3a74), U64(0x1e0a0a141e0a0a14), + U64(0xdb494992db494992), U64(0x0a06060c0a06060c), + U64(0x6c2424486c242448), U64(0xe45c5cb8e45c5cb8), + U64(0x5dc2c29f5dc2c29f), U64(0x6ed3d3bd6ed3d3bd), + U64(0xefacac43efacac43), U64(0xa66262c4a66262c4), + U64(0xa8919139a8919139), U64(0xa4959531a4959531), + U64(0x37e4e4d337e4e4d3), U64(0x8b7979f28b7979f2), + U64(0x32e7e7d532e7e7d5), U64(0x43c8c88b43c8c88b), + U64(0x5937376e5937376e), U64(0xb76d6ddab76d6dda), + U64(0x8c8d8d018c8d8d01), U64(0x64d5d5b164d5d5b1), + U64(0xd24e4e9cd24e4e9c), U64(0xe0a9a949e0a9a949), + U64(0xb46c6cd8b46c6cd8), U64(0xfa5656acfa5656ac), + U64(0x07f4f4f307f4f4f3), U64(0x25eaeacf25eaeacf), + U64(0xaf6565caaf6565ca), U64(0x8e7a7af48e7a7af4), + U64(0xe9aeae47e9aeae47), U64(0x1808081018080810), + U64(0xd5baba6fd5baba6f), U64(0x887878f0887878f0), + U64(0x6f25254a6f25254a), U64(0x722e2e5c722e2e5c), + U64(0x241c1c38241c1c38), U64(0xf1a6a657f1a6a657), + U64(0xc7b4b473c7b4b473), U64(0x51c6c69751c6c697), + U64(0x23e8e8cb23e8e8cb), U64(0x7cdddda17cdddda1), + U64(0x9c7474e89c7474e8), U64(0x211f1f3e211f1f3e), + U64(0xdd4b4b96dd4b4b96), U64(0xdcbdbd61dcbdbd61), + U64(0x868b8b0d868b8b0d), U64(0x858a8a0f858a8a0f), + U64(0x907070e0907070e0), U64(0x423e3e7c423e3e7c), + U64(0xc4b5b571c4b5b571), U64(0xaa6666ccaa6666cc), + U64(0xd8484890d8484890), U64(0x0503030605030306), + U64(0x01f6f6f701f6f6f7), U64(0x120e0e1c120e0e1c), + U64(0xa36161c2a36161c2), U64(0x5f35356a5f35356a), + U64(0xf95757aef95757ae), U64(0xd0b9b969d0b9b969), + U64(0x9186861791868617), U64(0x58c1c19958c1c199), + U64(0x271d1d3a271d1d3a), U64(0xb99e9e27b99e9e27), + U64(0x38e1e1d938e1e1d9), U64(0x13f8f8eb13f8f8eb), + U64(0xb398982bb398982b), U64(0x3311112233111122), + U64(0xbb6969d2bb6969d2), U64(0x70d9d9a970d9d9a9), + U64(0x898e8e07898e8e07), U64(0xa7949433a7949433), + U64(0xb69b9b2db69b9b2d), U64(0x221e1e3c221e1e3c), + U64(0x9287871592878715), U64(0x20e9e9c920e9e9c9), + U64(0x49cece8749cece87), U64(0xff5555aaff5555aa), + U64(0x7828285078282850), U64(0x7adfdfa57adfdfa5), + U64(0x8f8c8c038f8c8c03), U64(0xf8a1a159f8a1a159), + U64(0x8089890980898909), U64(0x170d0d1a170d0d1a), + U64(0xdabfbf65dabfbf65), U64(0x31e6e6d731e6e6d7), + U64(0xc6424284c6424284), U64(0xb86868d0b86868d0), + U64(0xc3414182c3414182), U64(0xb0999929b0999929), + U64(0x772d2d5a772d2d5a), U64(0x110f0f1e110f0f1e), + U64(0xcbb0b07bcbb0b07b), U64(0xfc5454a8fc5454a8), + U64(0xd6bbbb6dd6bbbb6d), U64(0x3a16162c3a16162c) +}; + +static const u8 Te4[256] = { + 0x63U, 0x7cU, 0x77U, 0x7bU, 0xf2U, 0x6bU, 0x6fU, 0xc5U, + 0x30U, 0x01U, 0x67U, 0x2bU, 0xfeU, 0xd7U, 0xabU, 0x76U, + 0xcaU, 0x82U, 0xc9U, 0x7dU, 0xfaU, 0x59U, 0x47U, 0xf0U, + 0xadU, 0xd4U, 0xa2U, 0xafU, 0x9cU, 0xa4U, 0x72U, 0xc0U, + 0xb7U, 0xfdU, 0x93U, 0x26U, 0x36U, 0x3fU, 0xf7U, 0xccU, + 0x34U, 0xa5U, 0xe5U, 0xf1U, 0x71U, 0xd8U, 0x31U, 0x15U, + 0x04U, 0xc7U, 0x23U, 0xc3U, 0x18U, 0x96U, 0x05U, 0x9aU, + 0x07U, 0x12U, 0x80U, 0xe2U, 0xebU, 0x27U, 0xb2U, 0x75U, + 0x09U, 0x83U, 0x2cU, 0x1aU, 0x1bU, 0x6eU, 0x5aU, 0xa0U, + 0x52U, 0x3bU, 0xd6U, 0xb3U, 0x29U, 0xe3U, 0x2fU, 0x84U, + 0x53U, 0xd1U, 0x00U, 0xedU, 0x20U, 0xfcU, 0xb1U, 0x5bU, + 0x6aU, 0xcbU, 0xbeU, 0x39U, 0x4aU, 0x4cU, 0x58U, 0xcfU, + 0xd0U, 0xefU, 0xaaU, 0xfbU, 0x43U, 0x4dU, 0x33U, 0x85U, + 0x45U, 0xf9U, 0x02U, 0x7fU, 0x50U, 0x3cU, 0x9fU, 0xa8U, + 0x51U, 0xa3U, 0x40U, 0x8fU, 0x92U, 0x9dU, 0x38U, 0xf5U, + 0xbcU, 0xb6U, 0xdaU, 0x21U, 0x10U, 0xffU, 0xf3U, 0xd2U, + 0xcdU, 0x0cU, 0x13U, 0xecU, 0x5fU, 0x97U, 0x44U, 0x17U, + 0xc4U, 0xa7U, 0x7eU, 0x3dU, 0x64U, 0x5dU, 0x19U, 0x73U, + 0x60U, 0x81U, 0x4fU, 0xdcU, 0x22U, 0x2aU, 0x90U, 0x88U, + 0x46U, 0xeeU, 0xb8U, 0x14U, 0xdeU, 0x5eU, 0x0bU, 0xdbU, + 0xe0U, 0x32U, 0x3aU, 0x0aU, 0x49U, 0x06U, 0x24U, 0x5cU, + 0xc2U, 0xd3U, 0xacU, 0x62U, 0x91U, 0x95U, 0xe4U, 0x79U, + 0xe7U, 0xc8U, 0x37U, 0x6dU, 0x8dU, 0xd5U, 0x4eU, 0xa9U, + 0x6cU, 0x56U, 0xf4U, 0xeaU, 0x65U, 0x7aU, 0xaeU, 0x08U, + 0xbaU, 0x78U, 0x25U, 0x2eU, 0x1cU, 0xa6U, 0xb4U, 0xc6U, + 0xe8U, 0xddU, 0x74U, 0x1fU, 0x4bU, 0xbdU, 0x8bU, 0x8aU, + 0x70U, 0x3eU, 0xb5U, 0x66U, 0x48U, 0x03U, 0xf6U, 0x0eU, + 0x61U, 0x35U, 0x57U, 0xb9U, 0x86U, 0xc1U, 0x1dU, 0x9eU, + 0xe1U, 0xf8U, 0x98U, 0x11U, 0x69U, 0xd9U, 0x8eU, 0x94U, + 0x9bU, 0x1eU, 0x87U, 0xe9U, 0xceU, 0x55U, 0x28U, 0xdfU, + 0x8cU, 0xa1U, 0x89U, 0x0dU, 0xbfU, 0xe6U, 0x42U, 0x68U, + 0x41U, 0x99U, 0x2dU, 0x0fU, 0xb0U, 0x54U, 0xbbU, 0x16U +}; + +static const u64 Td[256] = { + U64(0x50a7f45150a7f451), U64(0x5365417e5365417e), + U64(0xc3a4171ac3a4171a), U64(0x965e273a965e273a), + U64(0xcb6bab3bcb6bab3b), U64(0xf1459d1ff1459d1f), + U64(0xab58faacab58faac), U64(0x9303e34b9303e34b), + U64(0x55fa302055fa3020), U64(0xf66d76adf66d76ad), + U64(0x9176cc889176cc88), U64(0x254c02f5254c02f5), + U64(0xfcd7e54ffcd7e54f), U64(0xd7cb2ac5d7cb2ac5), + U64(0x8044352680443526), U64(0x8fa362b58fa362b5), + U64(0x495ab1de495ab1de), U64(0x671bba25671bba25), + U64(0x980eea45980eea45), U64(0xe1c0fe5de1c0fe5d), + U64(0x02752fc302752fc3), U64(0x12f04c8112f04c81), + U64(0xa397468da397468d), U64(0xc6f9d36bc6f9d36b), + U64(0xe75f8f03e75f8f03), U64(0x959c9215959c9215), + U64(0xeb7a6dbfeb7a6dbf), U64(0xda595295da595295), + U64(0x2d83bed42d83bed4), U64(0xd3217458d3217458), + U64(0x2969e0492969e049), U64(0x44c8c98e44c8c98e), + U64(0x6a89c2756a89c275), U64(0x78798ef478798ef4), + U64(0x6b3e58996b3e5899), U64(0xdd71b927dd71b927), + U64(0xb64fe1beb64fe1be), U64(0x17ad88f017ad88f0), + U64(0x66ac20c966ac20c9), U64(0xb43ace7db43ace7d), + U64(0x184adf63184adf63), U64(0x82311ae582311ae5), + U64(0x6033519760335197), U64(0x457f5362457f5362), + U64(0xe07764b1e07764b1), U64(0x84ae6bbb84ae6bbb), + U64(0x1ca081fe1ca081fe), U64(0x942b08f9942b08f9), + U64(0x5868487058684870), U64(0x19fd458f19fd458f), + U64(0x876cde94876cde94), U64(0xb7f87b52b7f87b52), + U64(0x23d373ab23d373ab), U64(0xe2024b72e2024b72), + U64(0x578f1fe3578f1fe3), U64(0x2aab55662aab5566), + U64(0x0728ebb20728ebb2), U64(0x03c2b52f03c2b52f), + U64(0x9a7bc5869a7bc586), U64(0xa50837d3a50837d3), + U64(0xf2872830f2872830), U64(0xb2a5bf23b2a5bf23), + U64(0xba6a0302ba6a0302), U64(0x5c8216ed5c8216ed), + U64(0x2b1ccf8a2b1ccf8a), U64(0x92b479a792b479a7), + U64(0xf0f207f3f0f207f3), U64(0xa1e2694ea1e2694e), + U64(0xcdf4da65cdf4da65), U64(0xd5be0506d5be0506), + U64(0x1f6234d11f6234d1), U64(0x8afea6c48afea6c4), + U64(0x9d532e349d532e34), U64(0xa055f3a2a055f3a2), + U64(0x32e18a0532e18a05), U64(0x75ebf6a475ebf6a4), + U64(0x39ec830b39ec830b), U64(0xaaef6040aaef6040), + U64(0x069f715e069f715e), U64(0x51106ebd51106ebd), + U64(0xf98a213ef98a213e), U64(0x3d06dd963d06dd96), + U64(0xae053eddae053edd), U64(0x46bde64d46bde64d), + U64(0xb58d5491b58d5491), U64(0x055dc471055dc471), + U64(0x6fd406046fd40604), U64(0xff155060ff155060), + U64(0x24fb981924fb9819), U64(0x97e9bdd697e9bdd6), + U64(0xcc434089cc434089), U64(0x779ed967779ed967), + U64(0xbd42e8b0bd42e8b0), U64(0x888b8907888b8907), + U64(0x385b19e7385b19e7), U64(0xdbeec879dbeec879), + U64(0x470a7ca1470a7ca1), U64(0xe90f427ce90f427c), + U64(0xc91e84f8c91e84f8), U64(0x0000000000000000), + U64(0x8386800983868009), U64(0x48ed2b3248ed2b32), + U64(0xac70111eac70111e), U64(0x4e725a6c4e725a6c), + U64(0xfbff0efdfbff0efd), U64(0x5638850f5638850f), + U64(0x1ed5ae3d1ed5ae3d), U64(0x27392d3627392d36), + U64(0x64d90f0a64d90f0a), U64(0x21a65c6821a65c68), + U64(0xd1545b9bd1545b9b), U64(0x3a2e36243a2e3624), + U64(0xb1670a0cb1670a0c), U64(0x0fe757930fe75793), + U64(0xd296eeb4d296eeb4), U64(0x9e919b1b9e919b1b), + U64(0x4fc5c0804fc5c080), U64(0xa220dc61a220dc61), + U64(0x694b775a694b775a), U64(0x161a121c161a121c), + U64(0x0aba93e20aba93e2), U64(0xe52aa0c0e52aa0c0), + U64(0x43e0223c43e0223c), U64(0x1d171b121d171b12), + U64(0x0b0d090e0b0d090e), U64(0xadc78bf2adc78bf2), + U64(0xb9a8b62db9a8b62d), U64(0xc8a91e14c8a91e14), + U64(0x8519f1578519f157), U64(0x4c0775af4c0775af), + U64(0xbbdd99eebbdd99ee), U64(0xfd607fa3fd607fa3), + U64(0x9f2601f79f2601f7), U64(0xbcf5725cbcf5725c), + U64(0xc53b6644c53b6644), U64(0x347efb5b347efb5b), + U64(0x7629438b7629438b), U64(0xdcc623cbdcc623cb), + U64(0x68fcedb668fcedb6), U64(0x63f1e4b863f1e4b8), + U64(0xcadc31d7cadc31d7), U64(0x1085634210856342), + U64(0x4022971340229713), U64(0x2011c6842011c684), + U64(0x7d244a857d244a85), U64(0xf83dbbd2f83dbbd2), + U64(0x1132f9ae1132f9ae), U64(0x6da129c76da129c7), + U64(0x4b2f9e1d4b2f9e1d), U64(0xf330b2dcf330b2dc), + U64(0xec52860dec52860d), U64(0xd0e3c177d0e3c177), + U64(0x6c16b32b6c16b32b), U64(0x99b970a999b970a9), + U64(0xfa489411fa489411), U64(0x2264e9472264e947), + U64(0xc48cfca8c48cfca8), U64(0x1a3ff0a01a3ff0a0), + U64(0xd82c7d56d82c7d56), U64(0xef903322ef903322), + U64(0xc74e4987c74e4987), U64(0xc1d138d9c1d138d9), + U64(0xfea2ca8cfea2ca8c), U64(0x360bd498360bd498), + U64(0xcf81f5a6cf81f5a6), U64(0x28de7aa528de7aa5), + U64(0x268eb7da268eb7da), U64(0xa4bfad3fa4bfad3f), + U64(0xe49d3a2ce49d3a2c), U64(0x0d9278500d927850), + U64(0x9bcc5f6a9bcc5f6a), U64(0x62467e5462467e54), + U64(0xc2138df6c2138df6), U64(0xe8b8d890e8b8d890), + U64(0x5ef7392e5ef7392e), U64(0xf5afc382f5afc382), + U64(0xbe805d9fbe805d9f), U64(0x7c93d0697c93d069), + U64(0xa92dd56fa92dd56f), U64(0xb31225cfb31225cf), + U64(0x3b99acc83b99acc8), U64(0xa77d1810a77d1810), + U64(0x6e639ce86e639ce8), U64(0x7bbb3bdb7bbb3bdb), + U64(0x097826cd097826cd), U64(0xf418596ef418596e), + U64(0x01b79aec01b79aec), U64(0xa89a4f83a89a4f83), + U64(0x656e95e6656e95e6), U64(0x7ee6ffaa7ee6ffaa), + U64(0x08cfbc2108cfbc21), U64(0xe6e815efe6e815ef), + U64(0xd99be7bad99be7ba), U64(0xce366f4ace366f4a), + U64(0xd4099fead4099fea), U64(0xd67cb029d67cb029), + U64(0xafb2a431afb2a431), U64(0x31233f2a31233f2a), + U64(0x3094a5c63094a5c6), U64(0xc066a235c066a235), + U64(0x37bc4e7437bc4e74), U64(0xa6ca82fca6ca82fc), + U64(0xb0d090e0b0d090e0), U64(0x15d8a73315d8a733), + U64(0x4a9804f14a9804f1), U64(0xf7daec41f7daec41), + U64(0x0e50cd7f0e50cd7f), U64(0x2ff691172ff69117), + U64(0x8dd64d768dd64d76), U64(0x4db0ef434db0ef43), + U64(0x544daacc544daacc), U64(0xdf0496e4df0496e4), + U64(0xe3b5d19ee3b5d19e), U64(0x1b886a4c1b886a4c), + U64(0xb81f2cc1b81f2cc1), U64(0x7f5165467f516546), + U64(0x04ea5e9d04ea5e9d), U64(0x5d358c015d358c01), + U64(0x737487fa737487fa), U64(0x2e410bfb2e410bfb), + U64(0x5a1d67b35a1d67b3), U64(0x52d2db9252d2db92), + U64(0x335610e9335610e9), U64(0x1347d66d1347d66d), + U64(0x8c61d79a8c61d79a), U64(0x7a0ca1377a0ca137), + U64(0x8e14f8598e14f859), U64(0x893c13eb893c13eb), + U64(0xee27a9ceee27a9ce), U64(0x35c961b735c961b7), + U64(0xede51ce1ede51ce1), U64(0x3cb1477a3cb1477a), + U64(0x59dfd29c59dfd29c), U64(0x3f73f2553f73f255), + U64(0x79ce141879ce1418), U64(0xbf37c773bf37c773), + U64(0xeacdf753eacdf753), U64(0x5baafd5f5baafd5f), + U64(0x146f3ddf146f3ddf), U64(0x86db447886db4478), + U64(0x81f3afca81f3afca), U64(0x3ec468b93ec468b9), + U64(0x2c3424382c342438), U64(0x5f40a3c25f40a3c2), + U64(0x72c31d1672c31d16), U64(0x0c25e2bc0c25e2bc), + U64(0x8b493c288b493c28), U64(0x41950dff41950dff), + U64(0x7101a8397101a839), U64(0xdeb30c08deb30c08), + U64(0x9ce4b4d89ce4b4d8), U64(0x90c1566490c15664), + U64(0x6184cb7b6184cb7b), U64(0x70b632d570b632d5), + U64(0x745c6c48745c6c48), U64(0x4257b8d04257b8d0) +}; +static const u8 Td4[256] = { + 0x52U, 0x09U, 0x6aU, 0xd5U, 0x30U, 0x36U, 0xa5U, 0x38U, + 0xbfU, 0x40U, 0xa3U, 0x9eU, 0x81U, 0xf3U, 0xd7U, 0xfbU, + 0x7cU, 0xe3U, 0x39U, 0x82U, 0x9bU, 0x2fU, 0xffU, 0x87U, + 0x34U, 0x8eU, 0x43U, 0x44U, 0xc4U, 0xdeU, 0xe9U, 0xcbU, + 0x54U, 0x7bU, 0x94U, 0x32U, 0xa6U, 0xc2U, 0x23U, 0x3dU, + 0xeeU, 0x4cU, 0x95U, 0x0bU, 0x42U, 0xfaU, 0xc3U, 0x4eU, + 0x08U, 0x2eU, 0xa1U, 0x66U, 0x28U, 0xd9U, 0x24U, 0xb2U, + 0x76U, 0x5bU, 0xa2U, 0x49U, 0x6dU, 0x8bU, 0xd1U, 0x25U, + 0x72U, 0xf8U, 0xf6U, 0x64U, 0x86U, 0x68U, 0x98U, 0x16U, + 0xd4U, 0xa4U, 0x5cU, 0xccU, 0x5dU, 0x65U, 0xb6U, 0x92U, + 0x6cU, 0x70U, 0x48U, 0x50U, 0xfdU, 0xedU, 0xb9U, 0xdaU, + 0x5eU, 0x15U, 0x46U, 0x57U, 0xa7U, 0x8dU, 0x9dU, 0x84U, + 0x90U, 0xd8U, 0xabU, 0x00U, 0x8cU, 0xbcU, 0xd3U, 0x0aU, + 0xf7U, 0xe4U, 0x58U, 0x05U, 0xb8U, 0xb3U, 0x45U, 0x06U, + 0xd0U, 0x2cU, 0x1eU, 0x8fU, 0xcaU, 0x3fU, 0x0fU, 0x02U, + 0xc1U, 0xafU, 0xbdU, 0x03U, 0x01U, 0x13U, 0x8aU, 0x6bU, + 0x3aU, 0x91U, 0x11U, 0x41U, 0x4fU, 0x67U, 0xdcU, 0xeaU, + 0x97U, 0xf2U, 0xcfU, 0xceU, 0xf0U, 0xb4U, 0xe6U, 0x73U, + 0x96U, 0xacU, 0x74U, 0x22U, 0xe7U, 0xadU, 0x35U, 0x85U, + 0xe2U, 0xf9U, 0x37U, 0xe8U, 0x1cU, 0x75U, 0xdfU, 0x6eU, + 0x47U, 0xf1U, 0x1aU, 0x71U, 0x1dU, 0x29U, 0xc5U, 0x89U, + 0x6fU, 0xb7U, 0x62U, 0x0eU, 0xaaU, 0x18U, 0xbeU, 0x1bU, + 0xfcU, 0x56U, 0x3eU, 0x4bU, 0xc6U, 0xd2U, 0x79U, 0x20U, + 0x9aU, 0xdbU, 0xc0U, 0xfeU, 0x78U, 0xcdU, 0x5aU, 0xf4U, + 0x1fU, 0xddU, 0xa8U, 0x33U, 0x88U, 0x07U, 0xc7U, 0x31U, + 0xb1U, 0x12U, 0x10U, 0x59U, 0x27U, 0x80U, 0xecU, 0x5fU, + 0x60U, 0x51U, 0x7fU, 0xa9U, 0x19U, 0xb5U, 0x4aU, 0x0dU, + 0x2dU, 0xe5U, 0x7aU, 0x9fU, 0x93U, 0xc9U, 0x9cU, 0xefU, + 0xa0U, 0xe0U, 0x3bU, 0x4dU, 0xaeU, 0x2aU, 0xf5U, 0xb0U, + 0xc8U, 0xebU, 0xbbU, 0x3cU, 0x83U, 0x53U, 0x99U, 0x61U, + 0x17U, 0x2bU, 0x04U, 0x7eU, 0xbaU, 0x77U, 0xd6U, 0x26U, + 0xe1U, 0x69U, 0x14U, 0x63U, 0x55U, 0x21U, 0x0cU, 0x7dU +}; + +static const u32 rcon[] = { + 0x00000001U, 0x00000002U, 0x00000004U, 0x00000008U, + 0x00000010U, 0x00000020U, 0x00000040U, 0x00000080U, + 0x0000001bU, 0x00000036U, /* for 128-bit blocks, Rijndael never uses more than 10 rcon values */ +}; + +/** + * Expand the cipher key into the encryption key schedule. + */ +int AES_set_encrypt_key(const unsigned char *userKey, const int bits, + AES_KEY *key) +{ + + u32 *rk; + int i = 0; + u32 temp; + + if (!userKey || !key) + return -1; + if (bits != 128 && bits != 192 && bits != 256) + return -2; + + rk = key->rd_key; + + if (bits==128) + key->rounds = 10; + else if (bits==192) + key->rounds = 12; + else + key->rounds = 14; + + rk[0] = GETU32(userKey ); + rk[1] = GETU32(userKey + 4); + rk[2] = GETU32(userKey + 8); + rk[3] = GETU32(userKey + 12); + if (bits == 128) { + while (1) { + temp = rk[3]; + rk[4] = rk[0] ^ + ((u32)Te4[(temp >> 8) & 0xff] ) ^ + ((u32)Te4[(temp >> 16) & 0xff] << 8) ^ + ((u32)Te4[(temp >> 24) ] << 16) ^ + ((u32)Te4[(temp ) & 0xff] << 24) ^ + rcon[i]; + rk[5] = rk[1] ^ rk[4]; + rk[6] = rk[2] ^ rk[5]; + rk[7] = rk[3] ^ rk[6]; + if (++i == 10) { + return 0; + } + rk += 4; + } + } + rk[4] = GETU32(userKey + 16); + rk[5] = GETU32(userKey + 20); + if (bits == 192) { + while (1) { + temp = rk[ 5]; + rk[ 6] = rk[ 0] ^ + ((u32)Te4[(temp >> 8) & 0xff] ) ^ + ((u32)Te4[(temp >> 16) & 0xff] << 8) ^ + ((u32)Te4[(temp >> 24) ] << 16) ^ + ((u32)Te4[(temp ) & 0xff] << 24) ^ + rcon[i]; + rk[ 7] = rk[ 1] ^ rk[ 6]; + rk[ 8] = rk[ 2] ^ rk[ 7]; + rk[ 9] = rk[ 3] ^ rk[ 8]; + if (++i == 8) { + return 0; + } + rk[10] = rk[ 4] ^ rk[ 9]; + rk[11] = rk[ 5] ^ rk[10]; + rk += 6; + } + } + rk[6] = GETU32(userKey + 24); + rk[7] = GETU32(userKey + 28); + if (bits == 256) { + while (1) { + temp = rk[ 7]; + rk[ 8] = rk[ 0] ^ + ((u32)Te4[(temp >> 8) & 0xff] ) ^ + ((u32)Te4[(temp >> 16) & 0xff] << 8) ^ + ((u32)Te4[(temp >> 24) ] << 16) ^ + ((u32)Te4[(temp ) & 0xff] << 24) ^ + rcon[i]; + rk[ 9] = rk[ 1] ^ rk[ 8]; + rk[10] = rk[ 2] ^ rk[ 9]; + rk[11] = rk[ 3] ^ rk[10]; + if (++i == 7) { + return 0; + } + temp = rk[11]; + rk[12] = rk[ 4] ^ + ((u32)Te4[(temp ) & 0xff] ) ^ + ((u32)Te4[(temp >> 8) & 0xff] << 8) ^ + ((u32)Te4[(temp >> 16) & 0xff] << 16) ^ + ((u32)Te4[(temp >> 24) ] << 24); + rk[13] = rk[ 5] ^ rk[12]; + rk[14] = rk[ 6] ^ rk[13]; + rk[15] = rk[ 7] ^ rk[14]; + + rk += 8; + } + } + return 0; +} + +/** + * Expand the cipher key into the decryption key schedule. + */ +int AES_set_decrypt_key(const unsigned char *userKey, const int bits, + AES_KEY *key) +{ + + u32 *rk; + int i, j, status; + u32 temp; + + /* first, start with an encryption schedule */ + status = AES_set_encrypt_key(userKey, bits, key); + if (status < 0) + return status; + + rk = key->rd_key; + + /* invert the order of the round keys: */ + for (i = 0, j = 4*(key->rounds); i < j; i += 4, j -= 4) { + temp = rk[i ]; rk[i ] = rk[j ]; rk[j ] = temp; + temp = rk[i + 1]; rk[i + 1] = rk[j + 1]; rk[j + 1] = temp; + temp = rk[i + 2]; rk[i + 2] = rk[j + 2]; rk[j + 2] = temp; + temp = rk[i + 3]; rk[i + 3] = rk[j + 3]; rk[j + 3] = temp; + } + /* apply the inverse MixColumn transform to all round keys but the first and the last: */ + for (i = 1; i < (key->rounds); i++) { + rk += 4; +#if 1 + for (j = 0; j < 4; j++) { + u32 tp1, tp2, tp4, tp8, tp9, tpb, tpd, tpe, m; + + tp1 = rk[j]; + m = tp1 & 0x80808080; + tp2 = ((tp1 & 0x7f7f7f7f) << 1) ^ + ((m - (m >> 7)) & 0x1b1b1b1b); + m = tp2 & 0x80808080; + tp4 = ((tp2 & 0x7f7f7f7f) << 1) ^ + ((m - (m >> 7)) & 0x1b1b1b1b); + m = tp4 & 0x80808080; + tp8 = ((tp4 & 0x7f7f7f7f) << 1) ^ + ((m - (m >> 7)) & 0x1b1b1b1b); + tp9 = tp8 ^ tp1; + tpb = tp9 ^ tp2; + tpd = tp9 ^ tp4; + tpe = tp8 ^ tp4 ^ tp2; +#if defined(ROTATE) + rk[j] = tpe ^ ROTATE(tpd,16) ^ + ROTATE(tp9,8) ^ ROTATE(tpb,24); +#else + rk[j] = tpe ^ (tpd >> 16) ^ (tpd << 16) ^ + (tp9 >> 24) ^ (tp9 << 8) ^ + (tpb >> 8) ^ (tpb << 24); +#endif + } +#else + rk[0] = + Td0[Te2[(rk[0] ) & 0xff] & 0xff] ^ + Td1[Te2[(rk[0] >> 8) & 0xff] & 0xff] ^ + Td2[Te2[(rk[0] >> 16) & 0xff] & 0xff] ^ + Td3[Te2[(rk[0] >> 24) ] & 0xff]; + rk[1] = + Td0[Te2[(rk[1] ) & 0xff] & 0xff] ^ + Td1[Te2[(rk[1] >> 8) & 0xff] & 0xff] ^ + Td2[Te2[(rk[1] >> 16) & 0xff] & 0xff] ^ + Td3[Te2[(rk[1] >> 24) ] & 0xff]; + rk[2] = + Td0[Te2[(rk[2] ) & 0xff] & 0xff] ^ + Td1[Te2[(rk[2] >> 8) & 0xff] & 0xff] ^ + Td2[Te2[(rk[2] >> 16) & 0xff] & 0xff] ^ + Td3[Te2[(rk[2] >> 24) ] & 0xff]; + rk[3] = + Td0[Te2[(rk[3] ) & 0xff] & 0xff] ^ + Td1[Te2[(rk[3] >> 8) & 0xff] & 0xff] ^ + Td2[Te2[(rk[3] >> 16) & 0xff] & 0xff] ^ + Td3[Te2[(rk[3] >> 24) ] & 0xff]; +#endif + } + return 0; +} + +/* + * Encrypt a single block + * in and out can overlap + */ +void AES_encrypt(const unsigned char *in, unsigned char *out, + const AES_KEY *key) +{ + + const u32 *rk; + u32 s0, s1, s2, s3, t[4]; + int r; + + assert(in && out && key); + rk = key->rd_key; + + /* + * map byte array block to cipher state + * and add initial round key: + */ + s0 = GETU32(in ) ^ rk[0]; + s1 = GETU32(in + 4) ^ rk[1]; + s2 = GETU32(in + 8) ^ rk[2]; + s3 = GETU32(in + 12) ^ rk[3]; + +#if defined(AES_COMPACT_IN_OUTER_ROUNDS) + prefetch256(Te4); + + t[0] = (u32)Te4[(s0 ) & 0xff] ^ + (u32)Te4[(s1 >> 8) & 0xff] << 8 ^ + (u32)Te4[(s2 >> 16) & 0xff] << 16 ^ + (u32)Te4[(s3 >> 24) ] << 24; + t[1] = (u32)Te4[(s1 ) & 0xff] ^ + (u32)Te4[(s2 >> 8) & 0xff] << 8 ^ + (u32)Te4[(s3 >> 16) & 0xff] << 16 ^ + (u32)Te4[(s0 >> 24) ] << 24; + t[2] = (u32)Te4[(s2 ) & 0xff] ^ + (u32)Te4[(s3 >> 8) & 0xff] << 8 ^ + (u32)Te4[(s0 >> 16) & 0xff] << 16 ^ + (u32)Te4[(s1 >> 24) ] << 24; + t[3] = (u32)Te4[(s3 ) & 0xff] ^ + (u32)Te4[(s0 >> 8) & 0xff] << 8 ^ + (u32)Te4[(s1 >> 16) & 0xff] << 16 ^ + (u32)Te4[(s2 >> 24) ] << 24; + + /* now do the linear transform using words */ + { int i; + u32 r0, r1, r2; + + for (i = 0; i < 4; i++) { + r0 = t[i]; + r1 = r0 & 0x80808080; + r2 = ((r0 & 0x7f7f7f7f) << 1) ^ + ((r1 - (r1 >> 7)) & 0x1b1b1b1b); +#if defined(ROTATE) + t[i] = r2 ^ ROTATE(r2,24) ^ ROTATE(r0,24) ^ + ROTATE(r0,16) ^ ROTATE(r0,8); +#else + t[i] = r2 ^ ((r2 ^ r0) << 24) ^ ((r2 ^ r0) >> 8) ^ + (r0 << 16) ^ (r0 >> 16) ^ + (r0 << 8) ^ (r0 >> 24); +#endif + t[i] ^= rk[4+i]; + } + } +#else + t[0] = Te0[(s0 ) & 0xff] ^ + Te1[(s1 >> 8) & 0xff] ^ + Te2[(s2 >> 16) & 0xff] ^ + Te3[(s3 >> 24) ] ^ + rk[4]; + t[1] = Te0[(s1 ) & 0xff] ^ + Te1[(s2 >> 8) & 0xff] ^ + Te2[(s3 >> 16) & 0xff] ^ + Te3[(s0 >> 24) ] ^ + rk[5]; + t[2] = Te0[(s2 ) & 0xff] ^ + Te1[(s3 >> 8) & 0xff] ^ + Te2[(s0 >> 16) & 0xff] ^ + Te3[(s1 >> 24) ] ^ + rk[6]; + t[3] = Te0[(s3 ) & 0xff] ^ + Te1[(s0 >> 8) & 0xff] ^ + Te2[(s1 >> 16) & 0xff] ^ + Te3[(s2 >> 24) ] ^ + rk[7]; +#endif + s0 = t[0]; s1 = t[1]; s2 = t[2]; s3 = t[3]; + + /* + * Nr - 2 full rounds: + */ + for (rk+=8,r=key->rounds-2; r>0; rk+=4,r--) { +#if defined(AES_COMPACT_IN_INNER_ROUNDS) + t[0] = (u32)Te4[(s0 ) & 0xff] ^ + (u32)Te4[(s1 >> 8) & 0xff] << 8 ^ + (u32)Te4[(s2 >> 16) & 0xff] << 16 ^ + (u32)Te4[(s3 >> 24) ] << 24; + t[1] = (u32)Te4[(s1 ) & 0xff] ^ + (u32)Te4[(s2 >> 8) & 0xff] << 8 ^ + (u32)Te4[(s3 >> 16) & 0xff] << 16 ^ + (u32)Te4[(s0 >> 24) ] << 24; + t[2] = (u32)Te4[(s2 ) & 0xff] ^ + (u32)Te4[(s3 >> 8) & 0xff] << 8 ^ + (u32)Te4[(s0 >> 16) & 0xff] << 16 ^ + (u32)Te4[(s1 >> 24) ] << 24; + t[3] = (u32)Te4[(s3 ) & 0xff] ^ + (u32)Te4[(s0 >> 8) & 0xff] << 8 ^ + (u32)Te4[(s1 >> 16) & 0xff] << 16 ^ + (u32)Te4[(s2 >> 24) ] << 24; + + /* now do the linear transform using words */ + { + int i; + u32 r0, r1, r2; + + for (i = 0; i < 4; i++) { + r0 = t[i]; + r1 = r0 & 0x80808080; + r2 = ((r0 & 0x7f7f7f7f) << 1) ^ + ((r1 - (r1 >> 7)) & 0x1b1b1b1b); +#if defined(ROTATE) + t[i] = r2 ^ ROTATE(r2,24) ^ ROTATE(r0,24) ^ + ROTATE(r0,16) ^ ROTATE(r0,8); +#else + t[i] = r2 ^ ((r2 ^ r0) << 24) ^ ((r2 ^ r0) >> 8) ^ + (r0 << 16) ^ (r0 >> 16) ^ + (r0 << 8) ^ (r0 >> 24); +#endif + t[i] ^= rk[i]; + } + } +#else + t[0] = Te0[(s0 ) & 0xff] ^ + Te1[(s1 >> 8) & 0xff] ^ + Te2[(s2 >> 16) & 0xff] ^ + Te3[(s3 >> 24) ] ^ + rk[0]; + t[1] = Te0[(s1 ) & 0xff] ^ + Te1[(s2 >> 8) & 0xff] ^ + Te2[(s3 >> 16) & 0xff] ^ + Te3[(s0 >> 24) ] ^ + rk[1]; + t[2] = Te0[(s2 ) & 0xff] ^ + Te1[(s3 >> 8) & 0xff] ^ + Te2[(s0 >> 16) & 0xff] ^ + Te3[(s1 >> 24) ] ^ + rk[2]; + t[3] = Te0[(s3 ) & 0xff] ^ + Te1[(s0 >> 8) & 0xff] ^ + Te2[(s1 >> 16) & 0xff] ^ + Te3[(s2 >> 24) ] ^ + rk[3]; +#endif + s0 = t[0]; s1 = t[1]; s2 = t[2]; s3 = t[3]; + } + /* + * apply last round and + * map cipher state to byte array block: + */ +#if defined(AES_COMPACT_IN_OUTER_ROUNDS) + prefetch256(Te4); + + *(u32*)(out+0) = + (u32)Te4[(s0 ) & 0xff] ^ + (u32)Te4[(s1 >> 8) & 0xff] << 8 ^ + (u32)Te4[(s2 >> 16) & 0xff] << 16 ^ + (u32)Te4[(s3 >> 24) ] << 24 ^ + rk[0]; + *(u32*)(out+4) = + (u32)Te4[(s1 ) & 0xff] ^ + (u32)Te4[(s2 >> 8) & 0xff] << 8 ^ + (u32)Te4[(s3 >> 16) & 0xff] << 16 ^ + (u32)Te4[(s0 >> 24) ] << 24 ^ + rk[1]; + *(u32*)(out+8) = + (u32)Te4[(s2 ) & 0xff] ^ + (u32)Te4[(s3 >> 8) & 0xff] << 8 ^ + (u32)Te4[(s0 >> 16) & 0xff] << 16 ^ + (u32)Te4[(s1 >> 24) ] << 24 ^ + rk[2]; + *(u32*)(out+12) = + (u32)Te4[(s3 ) & 0xff] ^ + (u32)Te4[(s0 >> 8) & 0xff] << 8 ^ + (u32)Te4[(s1 >> 16) & 0xff] << 16 ^ + (u32)Te4[(s2 >> 24) ] << 24 ^ + rk[3]; +#else + *(u32*)(out+0) = + (Te2[(s0 ) & 0xff] & 0x000000ffU) ^ + (Te3[(s1 >> 8) & 0xff] & 0x0000ff00U) ^ + (Te0[(s2 >> 16) & 0xff] & 0x00ff0000U) ^ + (Te1[(s3 >> 24) ] & 0xff000000U) ^ + rk[0]; + *(u32*)(out+4) = + (Te2[(s1 ) & 0xff] & 0x000000ffU) ^ + (Te3[(s2 >> 8) & 0xff] & 0x0000ff00U) ^ + (Te0[(s3 >> 16) & 0xff] & 0x00ff0000U) ^ + (Te1[(s0 >> 24) ] & 0xff000000U) ^ + rk[1]; + *(u32*)(out+8) = + (Te2[(s2 ) & 0xff] & 0x000000ffU) ^ + (Te3[(s3 >> 8) & 0xff] & 0x0000ff00U) ^ + (Te0[(s0 >> 16) & 0xff] & 0x00ff0000U) ^ + (Te1[(s1 >> 24) ] & 0xff000000U) ^ + rk[2]; + *(u32*)(out+12) = + (Te2[(s3 ) & 0xff] & 0x000000ffU) ^ + (Te3[(s0 >> 8) & 0xff] & 0x0000ff00U) ^ + (Te0[(s1 >> 16) & 0xff] & 0x00ff0000U) ^ + (Te1[(s2 >> 24) ] & 0xff000000U) ^ + rk[3]; +#endif +} + +/* + * Decrypt a single block + * in and out can overlap + */ +void AES_decrypt(const unsigned char *in, unsigned char *out, + const AES_KEY *key) +{ + + const u32 *rk; + u32 s0, s1, s2, s3, t[4]; + int r; + + assert(in && out && key); + rk = key->rd_key; + + /* + * map byte array block to cipher state + * and add initial round key: + */ + s0 = GETU32(in ) ^ rk[0]; + s1 = GETU32(in + 4) ^ rk[1]; + s2 = GETU32(in + 8) ^ rk[2]; + s3 = GETU32(in + 12) ^ rk[3]; + +#if defined(AES_COMPACT_IN_OUTER_ROUNDS) + prefetch256(Td4); + + t[0] = (u32)Td4[(s0 ) & 0xff] ^ + (u32)Td4[(s3 >> 8) & 0xff] << 8 ^ + (u32)Td4[(s2 >> 16) & 0xff] << 16 ^ + (u32)Td4[(s1 >> 24) ] << 24; + t[1] = (u32)Td4[(s1 ) & 0xff] ^ + (u32)Td4[(s0 >> 8) & 0xff] << 8 ^ + (u32)Td4[(s3 >> 16) & 0xff] << 16 ^ + (u32)Td4[(s2 >> 24) ] << 24; + t[2] = (u32)Td4[(s2 ) & 0xff] ^ + (u32)Td4[(s1 >> 8) & 0xff] << 8 ^ + (u32)Td4[(s0 >> 16) & 0xff] << 16 ^ + (u32)Td4[(s3 >> 24) ] << 24; + t[3] = (u32)Td4[(s3 ) & 0xff] ^ + (u32)Td4[(s2 >> 8) & 0xff] << 8 ^ + (u32)Td4[(s1 >> 16) & 0xff] << 16 ^ + (u32)Td4[(s0 >> 24) ] << 24; + + /* now do the linear transform using words */ + { + int i; + u32 tp1, tp2, tp4, tp8, tp9, tpb, tpd, tpe, m; + + for (i = 0; i < 4; i++) { + tp1 = t[i]; + m = tp1 & 0x80808080; + tp2 = ((tp1 & 0x7f7f7f7f) << 1) ^ + ((m - (m >> 7)) & 0x1b1b1b1b); + m = tp2 & 0x80808080; + tp4 = ((tp2 & 0x7f7f7f7f) << 1) ^ + ((m - (m >> 7)) & 0x1b1b1b1b); + m = tp4 & 0x80808080; + tp8 = ((tp4 & 0x7f7f7f7f) << 1) ^ + ((m - (m >> 7)) & 0x1b1b1b1b); + tp9 = tp8 ^ tp1; + tpb = tp9 ^ tp2; + tpd = tp9 ^ tp4; + tpe = tp8 ^ tp4 ^ tp2; +#if defined(ROTATE) + t[i] = tpe ^ ROTATE(tpd,16) ^ + ROTATE(tp9,8) ^ ROTATE(tpb,24); +#else + t[i] = tpe ^ (tpd >> 16) ^ (tpd << 16) ^ + (tp9 >> 24) ^ (tp9 << 8) ^ + (tpb >> 8) ^ (tpb << 24); +#endif + t[i] ^= rk[4+i]; + } + } +#else + t[0] = Td0[(s0 ) & 0xff] ^ + Td1[(s3 >> 8) & 0xff] ^ + Td2[(s2 >> 16) & 0xff] ^ + Td3[(s1 >> 24) ] ^ + rk[4]; + t[1] = Td0[(s1 ) & 0xff] ^ + Td1[(s0 >> 8) & 0xff] ^ + Td2[(s3 >> 16) & 0xff] ^ + Td3[(s2 >> 24) ] ^ + rk[5]; + t[2] = Td0[(s2 ) & 0xff] ^ + Td1[(s1 >> 8) & 0xff] ^ + Td2[(s0 >> 16) & 0xff] ^ + Td3[(s3 >> 24) ] ^ + rk[6]; + t[3] = Td0[(s3 ) & 0xff] ^ + Td1[(s2 >> 8) & 0xff] ^ + Td2[(s1 >> 16) & 0xff] ^ + Td3[(s0 >> 24) ] ^ + rk[7]; +#endif + s0 = t[0]; s1 = t[1]; s2 = t[2]; s3 = t[3]; + + /* + * Nr - 2 full rounds: + */ + for (rk+=8,r=key->rounds-2; r>0; rk+=4,r--) { +#if defined(AES_COMPACT_IN_INNER_ROUNDS) + t[0] = (u32)Td4[(s0 ) & 0xff] ^ + (u32)Td4[(s3 >> 8) & 0xff] << 8 ^ + (u32)Td4[(s2 >> 16) & 0xff] << 16 ^ + (u32)Td4[(s1 >> 24) ] << 24; + t[1] = (u32)Td4[(s1 ) & 0xff] ^ + (u32)Td4[(s0 >> 8) & 0xff] << 8 ^ + (u32)Td4[(s3 >> 16) & 0xff] << 16 ^ + (u32)Td4[(s2 >> 24) ] << 24; + t[2] = (u32)Td4[(s2 ) & 0xff] ^ + (u32)Td4[(s1 >> 8) & 0xff] << 8 ^ + (u32)Td4[(s0 >> 16) & 0xff] << 16 ^ + (u32)Td4[(s3 >> 24) ] << 24; + t[3] = (u32)Td4[(s3 ) & 0xff] ^ + (u32)Td4[(s2 >> 8) & 0xff] << 8 ^ + (u32)Td4[(s1 >> 16) & 0xff] << 16 ^ + (u32)Td4[(s0 >> 24) ] << 24; + + /* now do the linear transform using words */ + { + int i; + u32 tp1, tp2, tp4, tp8, tp9, tpb, tpd, tpe, m; + + for (i = 0; i < 4; i++) { + tp1 = t[i]; + m = tp1 & 0x80808080; + tp2 = ((tp1 & 0x7f7f7f7f) << 1) ^ + ((m - (m >> 7)) & 0x1b1b1b1b); + m = tp2 & 0x80808080; + tp4 = ((tp2 & 0x7f7f7f7f) << 1) ^ + ((m - (m >> 7)) & 0x1b1b1b1b); + m = tp4 & 0x80808080; + tp8 = ((tp4 & 0x7f7f7f7f) << 1) ^ + ((m - (m >> 7)) & 0x1b1b1b1b); + tp9 = tp8 ^ tp1; + tpb = tp9 ^ tp2; + tpd = tp9 ^ tp4; + tpe = tp8 ^ tp4 ^ tp2; +#if defined(ROTATE) + t[i] = tpe ^ ROTATE(tpd,16) ^ + ROTATE(tp9,8) ^ ROTATE(tpb,24); +#else + t[i] = tpe ^ (tpd >> 16) ^ (tpd << 16) ^ + (tp9 >> 24) ^ (tp9 << 8) ^ + (tpb >> 8) ^ (tpb << 24); +#endif + t[i] ^= rk[i]; + } + } +#else + t[0] = Td0[(s0 ) & 0xff] ^ + Td1[(s3 >> 8) & 0xff] ^ + Td2[(s2 >> 16) & 0xff] ^ + Td3[(s1 >> 24) ] ^ + rk[0]; + t[1] = Td0[(s1 ) & 0xff] ^ + Td1[(s0 >> 8) & 0xff] ^ + Td2[(s3 >> 16) & 0xff] ^ + Td3[(s2 >> 24) ] ^ + rk[1]; + t[2] = Td0[(s2 ) & 0xff] ^ + Td1[(s1 >> 8) & 0xff] ^ + Td2[(s0 >> 16) & 0xff] ^ + Td3[(s3 >> 24) ] ^ + rk[2]; + t[3] = Td0[(s3 ) & 0xff] ^ + Td1[(s2 >> 8) & 0xff] ^ + Td2[(s1 >> 16) & 0xff] ^ + Td3[(s0 >> 24) ] ^ + rk[3]; +#endif + s0 = t[0]; s1 = t[1]; s2 = t[2]; s3 = t[3]; + } + /* + * apply last round and + * map cipher state to byte array block: + */ + prefetch256(Td4); + + *(u32*)(out+0) = + ((u32)Td4[(s0 ) & 0xff]) ^ + ((u32)Td4[(s3 >> 8) & 0xff] << 8) ^ + ((u32)Td4[(s2 >> 16) & 0xff] << 16) ^ + ((u32)Td4[(s1 >> 24) ] << 24) ^ + rk[0]; + *(u32*)(out+4) = + ((u32)Td4[(s1 ) & 0xff]) ^ + ((u32)Td4[(s0 >> 8) & 0xff] << 8) ^ + ((u32)Td4[(s3 >> 16) & 0xff] << 16) ^ + ((u32)Td4[(s2 >> 24) ] << 24) ^ + rk[1]; + *(u32*)(out+8) = + ((u32)Td4[(s2 ) & 0xff]) ^ + ((u32)Td4[(s1 >> 8) & 0xff] << 8) ^ + ((u32)Td4[(s0 >> 16) & 0xff] << 16) ^ + ((u32)Td4[(s3 >> 24) ] << 24) ^ + rk[2]; + *(u32*)(out+12) = + ((u32)Td4[(s3 ) & 0xff]) ^ + ((u32)Td4[(s2 >> 8) & 0xff] << 8) ^ + ((u32)Td4[(s1 >> 16) & 0xff] << 16) ^ + ((u32)Td4[(s0 >> 24) ] << 24) ^ + rk[3]; +} diff --git a/openssl-1.1.0h/crypto/aes/asm/aes-586.pl b/openssl-1.1.0h/crypto/aes/asm/aes-586.pl new file mode 100755 index 0000000..1ba3565 --- /dev/null +++ b/openssl-1.1.0h/crypto/aes/asm/aes-586.pl @@ -0,0 +1,3000 @@ +#! /usr/bin/env perl +# Copyright 2004-2016 The OpenSSL Project Authors. All Rights Reserved. +# +# Licensed under the OpenSSL license (the "License"). You may not use +# this file except in compliance with the License. You can obtain a copy +# in the file LICENSE in the source distribution or at +# https://www.openssl.org/source/license.html + +# +# ==================================================================== +# Written by Andy Polyakov for the OpenSSL +# project. The module is, however, dual licensed under OpenSSL and +# CRYPTOGAMS licenses depending on where you obtain it. For further +# details see http://www.openssl.org/~appro/cryptogams/. +# ==================================================================== +# +# Version 4.3. +# +# You might fail to appreciate this module performance from the first +# try. If compared to "vanilla" linux-ia32-icc target, i.e. considered +# to be *the* best Intel C compiler without -KPIC, performance appears +# to be virtually identical... But try to re-configure with shared +# library support... Aha! Intel compiler "suddenly" lags behind by 30% +# [on P4, more on others]:-) And if compared to position-independent +# code generated by GNU C, this code performs *more* than *twice* as +# fast! Yes, all this buzz about PIC means that unlike other hand- +# coded implementations, this one was explicitly designed to be safe +# to use even in shared library context... This also means that this +# code isn't necessarily absolutely fastest "ever," because in order +# to achieve position independence an extra register has to be +# off-loaded to stack, which affects the benchmark result. +# +# Special note about instruction choice. Do you recall RC4_INT code +# performing poorly on P4? It might be the time to figure out why. +# RC4_INT code implies effective address calculations in base+offset*4 +# form. Trouble is that it seems that offset scaling turned to be +# critical path... At least eliminating scaling resulted in 2.8x RC4 +# performance improvement [as you might recall]. As AES code is hungry +# for scaling too, I [try to] avoid the latter by favoring off-by-2 +# shifts and masking the result with 0xFF<<2 instead of "boring" 0xFF. +# +# As was shown by Dean Gaudet , the above note turned +# void. Performance improvement with off-by-2 shifts was observed on +# intermediate implementation, which was spilling yet another register +# to stack... Final offset*4 code below runs just a tad faster on P4, +# but exhibits up to 10% improvement on other cores. +# +# Second version is "monolithic" replacement for aes_core.c, which in +# addition to AES_[de|en]crypt implements AES_set_[de|en]cryption_key. +# This made it possible to implement little-endian variant of the +# algorithm without modifying the base C code. Motivating factor for +# the undertaken effort was that it appeared that in tight IA-32 +# register window little-endian flavor could achieve slightly higher +# Instruction Level Parallelism, and it indeed resulted in up to 15% +# better performance on most recent µ-archs... +# +# Third version adds AES_cbc_encrypt implementation, which resulted in +# up to 40% performance imrovement of CBC benchmark results. 40% was +# observed on P4 core, where "overall" imrovement coefficient, i.e. if +# compared to PIC generated by GCC and in CBC mode, was observed to be +# as large as 4x:-) CBC performance is virtually identical to ECB now +# and on some platforms even better, e.g. 17.6 "small" cycles/byte on +# Opteron, because certain function prologues and epilogues are +# effectively taken out of the loop... +# +# Version 3.2 implements compressed tables and prefetch of these tables +# in CBC[!] mode. Former means that 3/4 of table references are now +# misaligned, which unfortunately has negative impact on elder IA-32 +# implementations, Pentium suffered 30% penalty, PIII - 10%. +# +# Version 3.3 avoids L1 cache aliasing between stack frame and +# S-boxes, and 3.4 - L1 cache aliasing even between key schedule. The +# latter is achieved by copying the key schedule to controlled place in +# stack. This unfortunately has rather strong impact on small block CBC +# performance, ~2x deterioration on 16-byte block if compared to 3.3. +# +# Version 3.5 checks if there is L1 cache aliasing between user-supplied +# key schedule and S-boxes and abstains from copying the former if +# there is no. This allows end-user to consciously retain small block +# performance by aligning key schedule in specific manner. +# +# Version 3.6 compresses Td4 to 256 bytes and prefetches it in ECB. +# +# Current ECB performance numbers for 128-bit key in CPU cycles per +# processed byte [measure commonly used by AES benchmarkers] are: +# +# small footprint fully unrolled +# P4 24 22 +# AMD K8 20 19 +# PIII 25 23 +# Pentium 81 78 +# +# Version 3.7 reimplements outer rounds as "compact." Meaning that +# first and last rounds reference compact 256 bytes S-box. This means +# that first round consumes a lot more CPU cycles and that encrypt +# and decrypt performance becomes asymmetric. Encrypt performance +# drops by 10-12%, while decrypt - by 20-25%:-( 256 bytes S-box is +# aggressively pre-fetched. +# +# Version 4.0 effectively rolls back to 3.6 and instead implements +# additional set of functions, _[x86|sse]_AES_[en|de]crypt_compact, +# which use exclusively 256 byte S-box. These functions are to be +# called in modes not concealing plain text, such as ECB, or when +# we're asked to process smaller amount of data [or unconditionally +# on hyper-threading CPU]. Currently it's called unconditionally from +# AES_[en|de]crypt, which affects all modes, but CBC. CBC routine +# still needs to be modified to switch between slower and faster +# mode when appropriate... But in either case benchmark landscape +# changes dramatically and below numbers are CPU cycles per processed +# byte for 128-bit key. +# +# ECB encrypt ECB decrypt CBC large chunk +# P4 52[54] 83[95] 23 +# AMD K8 46[41] 66[70] 18 +# PIII 41[50] 60[77] 24 +# Core 2 31[36] 45[64] 18.5 +# Atom 76[100] 96[138] 60 +# Pentium 115 150 77 +# +# Version 4.1 switches to compact S-box even in key schedule setup. +# +# Version 4.2 prefetches compact S-box in every SSE round or in other +# words every cache-line is *guaranteed* to be accessed within ~50 +# cycles window. Why just SSE? Because it's needed on hyper-threading +# CPU! Which is also why it's prefetched with 64 byte stride. Best +# part is that it has no negative effect on performance:-) +# +# Version 4.3 implements switch between compact and non-compact block +# functions in AES_cbc_encrypt depending on how much data was asked +# to be processed in one stroke. +# +###################################################################### +# Timing attacks are classified in two classes: synchronous when +# attacker consciously initiates cryptographic operation and collects +# timing data of various character afterwards, and asynchronous when +# malicious code is executed on same CPU simultaneously with AES, +# instruments itself and performs statistical analysis of this data. +# +# As far as synchronous attacks go the root to the AES timing +# vulnerability is twofold. Firstly, of 256 S-box elements at most 160 +# are referred to in single 128-bit block operation. Well, in C +# implementation with 4 distinct tables it's actually as little as 40 +# references per 256 elements table, but anyway... Secondly, even +# though S-box elements are clustered into smaller amount of cache- +# lines, smaller than 160 and even 40, it turned out that for certain +# plain-text pattern[s] or simply put chosen plain-text and given key +# few cache-lines remain unaccessed during block operation. Now, if +# attacker can figure out this access pattern, he can deduct the key +# [or at least part of it]. The natural way to mitigate this kind of +# attacks is to minimize the amount of cache-lines in S-box and/or +# prefetch them to ensure that every one is accessed for more uniform +# timing. But note that *if* plain-text was concealed in such way that +# input to block function is distributed *uniformly*, then attack +# wouldn't apply. Now note that some encryption modes, most notably +# CBC, do mask the plain-text in this exact way [secure cipher output +# is distributed uniformly]. Yes, one still might find input that +# would reveal the information about given key, but if amount of +# candidate inputs to be tried is larger than amount of possible key +# combinations then attack becomes infeasible. This is why revised +# AES_cbc_encrypt "dares" to switch to larger S-box when larger chunk +# of data is to be processed in one stroke. The current size limit of +# 512 bytes is chosen to provide same [diminishigly low] probability +# for cache-line to remain untouched in large chunk operation with +# large S-box as for single block operation with compact S-box and +# surely needs more careful consideration... +# +# As for asynchronous attacks. There are two flavours: attacker code +# being interleaved with AES on hyper-threading CPU at *instruction* +# level, and two processes time sharing single core. As for latter. +# Two vectors. 1. Given that attacker process has higher priority, +# yield execution to process performing AES just before timer fires +# off the scheduler, immediately regain control of CPU and analyze the +# cache state. For this attack to be efficient attacker would have to +# effectively slow down the operation by several *orders* of magnitute, +# by ratio of time slice to duration of handful of AES rounds, which +# unlikely to remain unnoticed. Not to mention that this also means +# that he would spend correspondigly more time to collect enough +# statistical data to mount the attack. It's probably appropriate to +# say that if adeversary reckons that this attack is beneficial and +# risks to be noticed, you probably have larger problems having him +# mere opportunity. In other words suggested code design expects you +# to preclude/mitigate this attack by overall system security design. +# 2. Attacker manages to make his code interrupt driven. In order for +# this kind of attack to be feasible, interrupt rate has to be high +# enough, again comparable to duration of handful of AES rounds. But +# is there interrupt source of such rate? Hardly, not even 1Gbps NIC +# generates interrupts at such raging rate... +# +# And now back to the former, hyper-threading CPU or more specifically +# Intel P4. Recall that asynchronous attack implies that malicious +# code instruments itself. And naturally instrumentation granularity +# has be noticeably lower than duration of codepath accessing S-box. +# Given that all cache-lines are accessed during that time that is. +# Current implementation accesses *all* cache-lines within ~50 cycles +# window, which is actually *less* than RDTSC latency on Intel P4! + +$0 =~ m/(.*[\/\\])[^\/\\]+$/; $dir=$1; +push(@INC,"${dir}","${dir}../../perlasm"); +require "x86asm.pl"; + +$output = pop; +open OUT,">$output"; +*STDOUT=*OUT; + +&asm_init($ARGV[0],"aes-586.pl",$x86only = $ARGV[$#ARGV] eq "386"); +&static_label("AES_Te"); +&static_label("AES_Td"); + +$s0="eax"; +$s1="ebx"; +$s2="ecx"; +$s3="edx"; +$key="edi"; +$acc="esi"; +$tbl="ebp"; + +# stack frame layout in _[x86|sse]_AES_* routines, frame is allocated +# by caller +$__ra=&DWP(0,"esp"); # return address +$__s0=&DWP(4,"esp"); # s0 backing store +$__s1=&DWP(8,"esp"); # s1 backing store +$__s2=&DWP(12,"esp"); # s2 backing store +$__s3=&DWP(16,"esp"); # s3 backing store +$__key=&DWP(20,"esp"); # pointer to key schedule +$__end=&DWP(24,"esp"); # pointer to end of key schedule +$__tbl=&DWP(28,"esp"); # %ebp backing store + +# stack frame layout in AES_[en|crypt] routines, which differs from +# above by 4 and overlaps by %ebp backing store +$_tbl=&DWP(24,"esp"); +$_esp=&DWP(28,"esp"); + +sub _data_word() { my $i; while(defined($i=shift)) { &data_word($i,$i); } } + +$speed_limit=512; # chunks smaller than $speed_limit are + # processed with compact routine in CBC mode +$small_footprint=1; # $small_footprint=1 code is ~5% slower [on + # recent µ-archs], but ~5 times smaller! + # I favor compact code to minimize cache + # contention and in hope to "collect" 5% back + # in real-life applications... + +$vertical_spin=0; # shift "verticaly" defaults to 0, because of + # its proof-of-concept status... +# Note that there is no decvert(), as well as last encryption round is +# performed with "horizontal" shifts. This is because this "vertical" +# implementation [one which groups shifts on a given $s[i] to form a +# "column," unlike "horizontal" one, which groups shifts on different +# $s[i] to form a "row"] is work in progress. It was observed to run +# few percents faster on Intel cores, but not AMD. On AMD K8 core it's +# whole 12% slower:-( So we face a trade-off... Shall it be resolved +# some day? Till then the code is considered experimental and by +# default remains dormant... + +sub encvert() +{ my ($te,@s) = @_; + my ($v0,$v1) = ($acc,$key); + + &mov ($v0,$s[3]); # copy s3 + &mov (&DWP(4,"esp"),$s[2]); # save s2 + &mov ($v1,$s[0]); # copy s0 + &mov (&DWP(8,"esp"),$s[1]); # save s1 + + &movz ($s[2],&HB($s[0])); + &and ($s[0],0xFF); + &mov ($s[0],&DWP(0,$te,$s[0],8)); # s0>>0 + &shr ($v1,16); + &mov ($s[3],&DWP(3,$te,$s[2],8)); # s0>>8 + &movz ($s[1],&HB($v1)); + &and ($v1,0xFF); + &mov ($s[2],&DWP(2,$te,$v1,8)); # s0>>16 + &mov ($v1,$v0); + &mov ($s[1],&DWP(1,$te,$s[1],8)); # s0>>24 + + &and ($v0,0xFF); + &xor ($s[3],&DWP(0,$te,$v0,8)); # s3>>0 + &movz ($v0,&HB($v1)); + &shr ($v1,16); + &xor ($s[2],&DWP(3,$te,$v0,8)); # s3>>8 + &movz ($v0,&HB($v1)); + &and ($v1,0xFF); + &xor ($s[1],&DWP(2,$te,$v1,8)); # s3>>16 + &mov ($v1,&DWP(4,"esp")); # restore s2 + &xor ($s[0],&DWP(1,$te,$v0,8)); # s3>>24 + + &mov ($v0,$v1); + &and ($v1,0xFF); + &xor ($s[2],&DWP(0,$te,$v1,8)); # s2>>0 + &movz ($v1,&HB($v0)); + &shr ($v0,16); + &xor ($s[1],&DWP(3,$te,$v1,8)); # s2>>8 + &movz ($v1,&HB($v0)); + &and ($v0,0xFF); + &xor ($s[0],&DWP(2,$te,$v0,8)); # s2>>16 + &mov ($v0,&DWP(8,"esp")); # restore s1 + &xor ($s[3],&DWP(1,$te,$v1,8)); # s2>>24 + + &mov ($v1,$v0); + &and ($v0,0xFF); + &xor ($s[1],&DWP(0,$te,$v0,8)); # s1>>0 + &movz ($v0,&HB($v1)); + &shr ($v1,16); + &xor ($s[0],&DWP(3,$te,$v0,8)); # s1>>8 + &movz ($v0,&HB($v1)); + &and ($v1,0xFF); + &xor ($s[3],&DWP(2,$te,$v1,8)); # s1>>16 + &mov ($key,$__key); # reincarnate v1 as key + &xor ($s[2],&DWP(1,$te,$v0,8)); # s1>>24 +} + +# Another experimental routine, which features "horizontal spin," but +# eliminates one reference to stack. Strangely enough runs slower... +sub enchoriz() +{ my ($v0,$v1) = ($key,$acc); + + &movz ($v0,&LB($s0)); # 3, 2, 1, 0* + &rotr ($s2,8); # 8,11,10, 9 + &mov ($v1,&DWP(0,$te,$v0,8)); # 0 + &movz ($v0,&HB($s1)); # 7, 6, 5*, 4 + &rotr ($s3,16); # 13,12,15,14 + &xor ($v1,&DWP(3,$te,$v0,8)); # 5 + &movz ($v0,&HB($s2)); # 8,11,10*, 9 + &rotr ($s0,16); # 1, 0, 3, 2 + &xor ($v1,&DWP(2,$te,$v0,8)); # 10 + &movz ($v0,&HB($s3)); # 13,12,15*,14 + &xor ($v1,&DWP(1,$te,$v0,8)); # 15, t[0] collected + &mov ($__s0,$v1); # t[0] saved + + &movz ($v0,&LB($s1)); # 7, 6, 5, 4* + &shr ($s1,16); # -, -, 7, 6 + &mov ($v1,&DWP(0,$te,$v0,8)); # 4 + &movz ($v0,&LB($s3)); # 13,12,15,14* + &xor ($v1,&DWP(2,$te,$v0,8)); # 14 + &movz ($v0,&HB($s0)); # 1, 0, 3*, 2 + &and ($s3,0xffff0000); # 13,12, -, - + &xor ($v1,&DWP(1,$te,$v0,8)); # 3 + &movz ($v0,&LB($s2)); # 8,11,10, 9* + &or ($s3,$s1); # 13,12, 7, 6 + &xor ($v1,&DWP(3,$te,$v0,8)); # 9, t[1] collected + &mov ($s1,$v1); # s[1]=t[1] + + &movz ($v0,&LB($s0)); # 1, 0, 3, 2* + &shr ($s2,16); # -, -, 8,11 + &mov ($v1,&DWP(2,$te,$v0,8)); # 2 + &movz ($v0,&HB($s3)); # 13,12, 7*, 6 + &xor ($v1,&DWP(1,$te,$v0,8)); # 7 + &movz ($v0,&HB($s2)); # -, -, 8*,11 + &xor ($v1,&DWP(0,$te,$v0,8)); # 8 + &mov ($v0,$s3); + &shr ($v0,24); # 13 + &xor ($v1,&DWP(3,$te,$v0,8)); # 13, t[2] collected + + &movz ($v0,&LB($s2)); # -, -, 8,11* + &shr ($s0,24); # 1* + &mov ($s2,&DWP(1,$te,$v0,8)); # 11 + &xor ($s2,&DWP(3,$te,$s0,8)); # 1 + &mov ($s0,$__s0); # s[0]=t[0] + &movz ($v0,&LB($s3)); # 13,12, 7, 6* + &shr ($s3,16); # , ,13,12 + &xor ($s2,&DWP(2,$te,$v0,8)); # 6 + &mov ($key,$__key); # reincarnate v0 as key + &and ($s3,0xff); # , ,13,12* + &mov ($s3,&DWP(0,$te,$s3,8)); # 12 + &xor ($s3,$s2); # s[2]=t[3] collected + &mov ($s2,$v1); # s[2]=t[2] +} + +# More experimental code... SSE one... Even though this one eliminates +# *all* references to stack, it's not faster... +sub sse_encbody() +{ + &movz ($acc,&LB("eax")); # 0 + &mov ("ecx",&DWP(0,$tbl,$acc,8)); # 0 + &pshufw ("mm2","mm0",0x0d); # 7, 6, 3, 2 + &movz ("edx",&HB("eax")); # 1 + &mov ("edx",&DWP(3,$tbl,"edx",8)); # 1 + &shr ("eax",16); # 5, 4 + + &movz ($acc,&LB("ebx")); # 10 + &xor ("ecx",&DWP(2,$tbl,$acc,8)); # 10 + &pshufw ("mm6","mm4",0x08); # 13,12, 9, 8 + &movz ($acc,&HB("ebx")); # 11 + &xor ("edx",&DWP(1,$tbl,$acc,8)); # 11 + &shr ("ebx",16); # 15,14 + + &movz ($acc,&HB("eax")); # 5 + &xor ("ecx",&DWP(3,$tbl,$acc,8)); # 5 + &movq ("mm3",QWP(16,$key)); + &movz ($acc,&HB("ebx")); # 15 + &xor ("ecx",&DWP(1,$tbl,$acc,8)); # 15 + &movd ("mm0","ecx"); # t[0] collected + + &movz ($acc,&LB("eax")); # 4 + &mov ("ecx",&DWP(0,$tbl,$acc,8)); # 4 + &movd ("eax","mm2"); # 7, 6, 3, 2 + &movz ($acc,&LB("ebx")); # 14 + &xor ("ecx",&DWP(2,$tbl,$acc,8)); # 14 + &movd ("ebx","mm6"); # 13,12, 9, 8 + + &movz ($acc,&HB("eax")); # 3 + &xor ("ecx",&DWP(1,$tbl,$acc,8)); # 3 + &movz ($acc,&HB("ebx")); # 9 + &xor ("ecx",&DWP(3,$tbl,$acc,8)); # 9 + &movd ("mm1","ecx"); # t[1] collected + + &movz ($acc,&LB("eax")); # 2 + &mov ("ecx",&DWP(2,$tbl,$acc,8)); # 2 + &shr ("eax",16); # 7, 6 + &punpckldq ("mm0","mm1"); # t[0,1] collected + &movz ($acc,&LB("ebx")); # 8 + &xor ("ecx",&DWP(0,$tbl,$acc,8)); # 8 + &shr ("ebx",16); # 13,12 + + &movz ($acc,&HB("eax")); # 7 + &xor ("ecx",&DWP(1,$tbl,$acc,8)); # 7 + &pxor ("mm0","mm3"); + &movz ("eax",&LB("eax")); # 6 + &xor ("edx",&DWP(2,$tbl,"eax",8)); # 6 + &pshufw ("mm1","mm0",0x08); # 5, 4, 1, 0 + &movz ($acc,&HB("ebx")); # 13 + &xor ("ecx",&DWP(3,$tbl,$acc,8)); # 13 + &xor ("ecx",&DWP(24,$key)); # t[2] + &movd ("mm4","ecx"); # t[2] collected + &movz ("ebx",&LB("ebx")); # 12 + &xor ("edx",&DWP(0,$tbl,"ebx",8)); # 12 + &shr ("ecx",16); + &movd ("eax","mm1"); # 5, 4, 1, 0 + &mov ("ebx",&DWP(28,$key)); # t[3] + &xor ("ebx","edx"); + &movd ("mm5","ebx"); # t[3] collected + &and ("ebx",0xffff0000); + &or ("ebx","ecx"); + + &punpckldq ("mm4","mm5"); # t[2,3] collected +} + +###################################################################### +# "Compact" block function +###################################################################### + +sub enccompact() +{ my $Fn = \&mov; + while ($#_>5) { pop(@_); $Fn=sub{}; } + my ($i,$te,@s)=@_; + my $tmp = $key; + my $out = $i==3?$s[0]:$acc; + + # $Fn is used in first compact round and its purpose is to + # void restoration of some values from stack, so that after + # 4xenccompact with extra argument $key value is left there... + if ($i==3) { &$Fn ($key,$__key); }##%edx + else { &mov ($out,$s[0]); } + &and ($out,0xFF); + if ($i==1) { &shr ($s[0],16); }#%ebx[1] + if ($i==2) { &shr ($s[0],24); }#%ecx[2] + &movz ($out,&BP(-128,$te,$out,1)); + + if ($i==3) { $tmp=$s[1]; }##%eax + &movz ($tmp,&HB($s[1])); + &movz ($tmp,&BP(-128,$te,$tmp,1)); + &shl ($tmp,8); + &xor ($out,$tmp); + + if ($i==3) { $tmp=$s[2]; &mov ($s[1],$__s0); }##%ebx + else { &mov ($tmp,$s[2]); + &shr ($tmp,16); } + if ($i==2) { &and ($s[1],0xFF); }#%edx[2] + &and ($tmp,0xFF); + &movz ($tmp,&BP(-128,$te,$tmp,1)); + &shl ($tmp,16); + &xor ($out,$tmp); + + if ($i==3) { $tmp=$s[3]; &mov ($s[2],$__s1); }##%ecx + elsif($i==2){ &movz ($tmp,&HB($s[3])); }#%ebx[2] + else { &mov ($tmp,$s[3]); + &shr ($tmp,24); } + &movz ($tmp,&BP(-128,$te,$tmp,1)); + &shl ($tmp,24); + &xor ($out,$tmp); + if ($i<2) { &mov (&DWP(4+4*$i,"esp"),$out); } + if ($i==3) { &mov ($s[3],$acc); } + &comment(); +} + +sub enctransform() +{ my @s = ($s0,$s1,$s2,$s3); + my $i = shift; + my $tmp = $tbl; + my $r2 = $key ; + + &and ($tmp,$s[$i]); + &lea ($r2,&DWP(0,$s[$i],$s[$i])); + &mov ($acc,$tmp); + &shr ($tmp,7); + &and ($r2,0xfefefefe); + &sub ($acc,$tmp); + &mov ($tmp,$s[$i]); + &and ($acc,0x1b1b1b1b); + &rotr ($tmp,16); + &xor ($acc,$r2); # r2 + &mov ($r2,$s[$i]); + + &xor ($s[$i],$acc); # r0 ^ r2 + &rotr ($r2,16+8); + &xor ($acc,$tmp); + &rotl ($s[$i],24); + &xor ($acc,$r2); + &mov ($tmp,0x80808080) if ($i!=1); + &xor ($s[$i],$acc); # ROTATE(r2^r0,24) ^ r2 +} + +&function_begin_B("_x86_AES_encrypt_compact"); + # note that caller is expected to allocate stack frame for me! + &mov ($__key,$key); # save key + + &xor ($s0,&DWP(0,$key)); # xor with key + &xor ($s1,&DWP(4,$key)); + &xor ($s2,&DWP(8,$key)); + &xor ($s3,&DWP(12,$key)); + + &mov ($acc,&DWP(240,$key)); # load key->rounds + &lea ($acc,&DWP(-2,$acc,$acc)); + &lea ($acc,&DWP(0,$key,$acc,8)); + &mov ($__end,$acc); # end of key schedule + + # prefetch Te4 + &mov ($key,&DWP(0-128,$tbl)); + &mov ($acc,&DWP(32-128,$tbl)); + &mov ($key,&DWP(64-128,$tbl)); + &mov ($acc,&DWP(96-128,$tbl)); + &mov ($key,&DWP(128-128,$tbl)); + &mov ($acc,&DWP(160-128,$tbl)); + &mov ($key,&DWP(192-128,$tbl)); + &mov ($acc,&DWP(224-128,$tbl)); + + &set_label("loop",16); + + &enccompact(0,$tbl,$s0,$s1,$s2,$s3,1); + &enccompact(1,$tbl,$s1,$s2,$s3,$s0,1); + &enccompact(2,$tbl,$s2,$s3,$s0,$s1,1); + &enccompact(3,$tbl,$s3,$s0,$s1,$s2,1); + &mov ($tbl,0x80808080); + &enctransform(2); + &enctransform(3); + &enctransform(0); + &enctransform(1); + &mov ($key,$__key); + &mov ($tbl,$__tbl); + &add ($key,16); # advance rd_key + &xor ($s0,&DWP(0,$key)); + &xor ($s1,&DWP(4,$key)); + &xor ($s2,&DWP(8,$key)); + &xor ($s3,&DWP(12,$key)); + + &cmp ($key,$__end); + &mov ($__key,$key); + &jb (&label("loop")); + + &enccompact(0,$tbl,$s0,$s1,$s2,$s3); + &enccompact(1,$tbl,$s1,$s2,$s3,$s0); + &enccompact(2,$tbl,$s2,$s3,$s0,$s1); + &enccompact(3,$tbl,$s3,$s0,$s1,$s2); + + &xor ($s0,&DWP(16,$key)); + &xor ($s1,&DWP(20,$key)); + &xor ($s2,&DWP(24,$key)); + &xor ($s3,&DWP(28,$key)); + + &ret (); +&function_end_B("_x86_AES_encrypt_compact"); + +###################################################################### +# "Compact" SSE block function. +###################################################################### +# +# Performance is not actually extraordinary in comparison to pure +# x86 code. In particular encrypt performance is virtually the same. +# Decrypt performance on the other hand is 15-20% better on newer +# µ-archs [but we're thankful for *any* improvement here], and ~50% +# better on PIII:-) And additionally on the pros side this code +# eliminates redundant references to stack and thus relieves/ +# minimizes the pressure on the memory bus. +# +# MMX register layout lsb +# +--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+ +# | mm4 | mm0 | +# +--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+ +# | s3 | s2 | s1 | s0 | +# +--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+ +# |15|14|13|12|11|10| 9| 8| 7| 6| 5| 4| 3| 2| 1| 0| +# +--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+ +# +# Indexes translate as s[N/4]>>(8*(N%4)), e.g. 5 means s1>>8. +# In this terms encryption and decryption "compact" permutation +# matrices can be depicted as following: +# +# encryption lsb # decryption lsb +# +----++----+----+----+----+ # +----++----+----+----+----+ +# | t0 || 15 | 10 | 5 | 0 | # | t0 || 7 | 10 | 13 | 0 | +# +----++----+----+----+----+ # +----++----+----+----+----+ +# | t1 || 3 | 14 | 9 | 4 | # | t1 || 11 | 14 | 1 | 4 | +# +----++----+----+----+----+ # +----++----+----+----+----+ +# | t2 || 7 | 2 | 13 | 8 | # | t2 || 15 | 2 | 5 | 8 | +# +----++----+----+----+----+ # +----++----+----+----+----+ +# | t3 || 11 | 6 | 1 | 12 | # | t3 || 3 | 6 | 9 | 12 | +# +----++----+----+----+----+ # +----++----+----+----+----+ +# +###################################################################### +# Why not xmm registers? Short answer. It was actually tested and +# was not any faster, but *contrary*, most notably on Intel CPUs. +# Longer answer. Main advantage of using mm registers is that movd +# latency is lower, especially on Intel P4. While arithmetic +# instructions are twice as many, they can be scheduled every cycle +# and not every second one when they are operating on xmm register, +# so that "arithmetic throughput" remains virtually the same. And +# finally the code can be executed even on elder SSE-only CPUs:-) + +sub sse_enccompact() +{ + &pshufw ("mm1","mm0",0x08); # 5, 4, 1, 0 + &pshufw ("mm5","mm4",0x0d); # 15,14,11,10 + &movd ("eax","mm1"); # 5, 4, 1, 0 + &movd ("ebx","mm5"); # 15,14,11,10 + &mov ($__key,$key); + + &movz ($acc,&LB("eax")); # 0 + &movz ("edx",&HB("eax")); # 1 + &pshufw ("mm2","mm0",0x0d); # 7, 6, 3, 2 + &movz ("ecx",&BP(-128,$tbl,$acc,1)); # 0 + &movz ($key,&LB("ebx")); # 10 + &movz ("edx",&BP(-128,$tbl,"edx",1)); # 1 + &shr ("eax",16); # 5, 4 + &shl ("edx",8); # 1 + + &movz ($acc,&BP(-128,$tbl,$key,1)); # 10 + &movz ($key,&HB("ebx")); # 11 + &shl ($acc,16); # 10 + &pshufw ("mm6","mm4",0x08); # 13,12, 9, 8 + &or ("ecx",$acc); # 10 + &movz ($acc,&BP(-128,$tbl,$key,1)); # 11 + &movz ($key,&HB("eax")); # 5 + &shl ($acc,24); # 11 + &shr ("ebx",16); # 15,14 + &or ("edx",$acc); # 11 + + &movz ($acc,&BP(-128,$tbl,$key,1)); # 5 + &movz ($key,&HB("ebx")); # 15 + &shl ($acc,8); # 5 + &or ("ecx",$acc); # 5 + &movz ($acc,&BP(-128,$tbl,$key,1)); # 15 + &movz ($key,&LB("eax")); # 4 + &shl ($acc,24); # 15 + &or ("ecx",$acc); # 15 + + &movz ($acc,&BP(-128,$tbl,$key,1)); # 4 + &movz ($key,&LB("ebx")); # 14 + &movd ("eax","mm2"); # 7, 6, 3, 2 + &movd ("mm0","ecx"); # t[0] collected + &movz ("ecx",&BP(-128,$tbl,$key,1)); # 14 + &movz ($key,&HB("eax")); # 3 + &shl ("ecx",16); # 14 + &movd ("ebx","mm6"); # 13,12, 9, 8 + &or ("ecx",$acc); # 14 + + &movz ($acc,&BP(-128,$tbl,$key,1)); # 3 + &movz ($key,&HB("ebx")); # 9 + &shl ($acc,24); # 3 + &or ("ecx",$acc); # 3 + &movz ($acc,&BP(-128,$tbl,$key,1)); # 9 + &movz ($key,&LB("ebx")); # 8 + &shl ($acc,8); # 9 + &shr ("ebx",16); # 13,12 + &or ("ecx",$acc); # 9 + + &movz ($acc,&BP(-128,$tbl,$key,1)); # 8 + &movz ($key,&LB("eax")); # 2 + &shr ("eax",16); # 7, 6 + &movd ("mm1","ecx"); # t[1] collected + &movz ("ecx",&BP(-128,$tbl,$key,1)); # 2 + &movz ($key,&HB("eax")); # 7 + &shl ("ecx",16); # 2 + &and ("eax",0xff); # 6 + &or ("ecx",$acc); # 2 + + &punpckldq ("mm0","mm1"); # t[0,1] collected + + &movz ($acc,&BP(-128,$tbl,$key,1)); # 7 + &movz ($key,&HB("ebx")); # 13 + &shl ($acc,24); # 7 + &and ("ebx",0xff); # 12 + &movz ("eax",&BP(-128,$tbl,"eax",1)); # 6 + &or ("ecx",$acc); # 7 + &shl ("eax",16); # 6 + &movz ($acc,&BP(-128,$tbl,$key,1)); # 13 + &or ("edx","eax"); # 6 + &shl ($acc,8); # 13 + &movz ("ebx",&BP(-128,$tbl,"ebx",1)); # 12 + &or ("ecx",$acc); # 13 + &or ("edx","ebx"); # 12 + &mov ($key,$__key); + &movd ("mm4","ecx"); # t[2] collected + &movd ("mm5","edx"); # t[3] collected + + &punpckldq ("mm4","mm5"); # t[2,3] collected +} + + if (!$x86only) { +&function_begin_B("_sse_AES_encrypt_compact"); + &pxor ("mm0",&QWP(0,$key)); # 7, 6, 5, 4, 3, 2, 1, 0 + &pxor ("mm4",&QWP(8,$key)); # 15,14,13,12,11,10, 9, 8 + + # note that caller is expected to allocate stack frame for me! + &mov ($acc,&DWP(240,$key)); # load key->rounds + &lea ($acc,&DWP(-2,$acc,$acc)); + &lea ($acc,&DWP(0,$key,$acc,8)); + &mov ($__end,$acc); # end of key schedule + + &mov ($s0,0x1b1b1b1b); # magic constant + &mov (&DWP(8,"esp"),$s0); + &mov (&DWP(12,"esp"),$s0); + + # prefetch Te4 + &mov ($s0,&DWP(0-128,$tbl)); + &mov ($s1,&DWP(32-128,$tbl)); + &mov ($s2,&DWP(64-128,$tbl)); + &mov ($s3,&DWP(96-128,$tbl)); + &mov ($s0,&DWP(128-128,$tbl)); + &mov ($s1,&DWP(160-128,$tbl)); + &mov ($s2,&DWP(192-128,$tbl)); + &mov ($s3,&DWP(224-128,$tbl)); + + &set_label("loop",16); + &sse_enccompact(); + &add ($key,16); + &cmp ($key,$__end); + &ja (&label("out")); + + &movq ("mm2",&QWP(8,"esp")); + &pxor ("mm3","mm3"); &pxor ("mm7","mm7"); + &movq ("mm1","mm0"); &movq ("mm5","mm4"); # r0 + &pcmpgtb("mm3","mm0"); &pcmpgtb("mm7","mm4"); + &pand ("mm3","mm2"); &pand ("mm7","mm2"); + &pshufw ("mm2","mm0",0xb1); &pshufw ("mm6","mm4",0xb1);# ROTATE(r0,16) + &paddb ("mm0","mm0"); &paddb ("mm4","mm4"); + &pxor ("mm0","mm3"); &pxor ("mm4","mm7"); # = r2 + &pshufw ("mm3","mm2",0xb1); &pshufw ("mm7","mm6",0xb1);# r0 + &pxor ("mm1","mm0"); &pxor ("mm5","mm4"); # r0^r2 + &pxor ("mm0","mm2"); &pxor ("mm4","mm6"); # ^= ROTATE(r0,16) + + &movq ("mm2","mm3"); &movq ("mm6","mm7"); + &pslld ("mm3",8); &pslld ("mm7",8); + &psrld ("mm2",24); &psrld ("mm6",24); + &pxor ("mm0","mm3"); &pxor ("mm4","mm7"); # ^= r0<<8 + &pxor ("mm0","mm2"); &pxor ("mm4","mm6"); # ^= r0>>24 + + &movq ("mm3","mm1"); &movq ("mm7","mm5"); + &movq ("mm2",&QWP(0,$key)); &movq ("mm6",&QWP(8,$key)); + &psrld ("mm1",8); &psrld ("mm5",8); + &mov ($s0,&DWP(0-128,$tbl)); + &pslld ("mm3",24); &pslld ("mm7",24); + &mov ($s1,&DWP(64-128,$tbl)); + &pxor ("mm0","mm1"); &pxor ("mm4","mm5"); # ^= (r2^r0)<<8 + &mov ($s2,&DWP(128-128,$tbl)); + &pxor ("mm0","mm3"); &pxor ("mm4","mm7"); # ^= (r2^r0)>>24 + &mov ($s3,&DWP(192-128,$tbl)); + + &pxor ("mm0","mm2"); &pxor ("mm4","mm6"); + &jmp (&label("loop")); + + &set_label("out",16); + &pxor ("mm0",&QWP(0,$key)); + &pxor ("mm4",&QWP(8,$key)); + + &ret (); +&function_end_B("_sse_AES_encrypt_compact"); + } + +###################################################################### +# Vanilla block function. +###################################################################### + +sub encstep() +{ my ($i,$te,@s) = @_; + my $tmp = $key; + my $out = $i==3?$s[0]:$acc; + + # lines marked with #%e?x[i] denote "reordered" instructions... + if ($i==3) { &mov ($key,$__key); }##%edx + else { &mov ($out,$s[0]); + &and ($out,0xFF); } + if ($i==1) { &shr ($s[0],16); }#%ebx[1] + if ($i==2) { &shr ($s[0],24); }#%ecx[2] + &mov ($out,&DWP(0,$te,$out,8)); + + if ($i==3) { $tmp=$s[1]; }##%eax + &movz ($tmp,&HB($s[1])); + &xor ($out,&DWP(3,$te,$tmp,8)); + + if ($i==3) { $tmp=$s[2]; &mov ($s[1],$__s0); }##%ebx + else { &mov ($tmp,$s[2]); + &shr ($tmp,16); } + if ($i==2) { &and ($s[1],0xFF); }#%edx[2] + &and ($tmp,0xFF); + &xor ($out,&DWP(2,$te,$tmp,8)); + + if ($i==3) { $tmp=$s[3]; &mov ($s[2],$__s1); }##%ecx + elsif($i==2){ &movz ($tmp,&HB($s[3])); }#%ebx[2] + else { &mov ($tmp,$s[3]); + &shr ($tmp,24) } + &xor ($out,&DWP(1,$te,$tmp,8)); + if ($i<2) { &mov (&DWP(4+4*$i,"esp"),$out); } + if ($i==3) { &mov ($s[3],$acc); } + &comment(); +} + +sub enclast() +{ my ($i,$te,@s)=@_; + my $tmp = $key; + my $out = $i==3?$s[0]:$acc; + + if ($i==3) { &mov ($key,$__key); }##%edx + else { &mov ($out,$s[0]); } + &and ($out,0xFF); + if ($i==1) { &shr ($s[0],16); }#%ebx[1] + if ($i==2) { &shr ($s[0],24); }#%ecx[2] + &mov ($out,&DWP(2,$te,$out,8)); + &and ($out,0x000000ff); + + if ($i==3) { $tmp=$s[1]; }##%eax + &movz ($tmp,&HB($s[1])); + &mov ($tmp,&DWP(0,$te,$tmp,8)); + &and ($tmp,0x0000ff00); + &xor ($out,$tmp); + + if ($i==3) { $tmp=$s[2]; &mov ($s[1],$__s0); }##%ebx + else { &mov ($tmp,$s[2]); + &shr ($tmp,16); } + if ($i==2) { &and ($s[1],0xFF); }#%edx[2] + &and ($tmp,0xFF); + &mov ($tmp,&DWP(0,$te,$tmp,8)); + &and ($tmp,0x00ff0000); + &xor ($out,$tmp); + + if ($i==3) { $tmp=$s[3]; &mov ($s[2],$__s1); }##%ecx + elsif($i==2){ &movz ($tmp,&HB($s[3])); }#%ebx[2] + else { &mov ($tmp,$s[3]); + &shr ($tmp,24); } + &mov ($tmp,&DWP(2,$te,$tmp,8)); + &and ($tmp,0xff000000); + &xor ($out,$tmp); + if ($i<2) { &mov (&DWP(4+4*$i,"esp"),$out); } + if ($i==3) { &mov ($s[3],$acc); } +} + +&function_begin_B("_x86_AES_encrypt"); + if ($vertical_spin) { + # I need high parts of volatile registers to be accessible... + &exch ($s1="edi",$key="ebx"); + &mov ($s2="esi",$acc="ecx"); + } + + # note that caller is expected to allocate stack frame for me! + &mov ($__key,$key); # save key + + &xor ($s0,&DWP(0,$key)); # xor with key + &xor ($s1,&DWP(4,$key)); + &xor ($s2,&DWP(8,$key)); + &xor ($s3,&DWP(12,$key)); + + &mov ($acc,&DWP(240,$key)); # load key->rounds + + if ($small_footprint) { + &lea ($acc,&DWP(-2,$acc,$acc)); + &lea ($acc,&DWP(0,$key,$acc,8)); + &mov ($__end,$acc); # end of key schedule + + &set_label("loop",16); + if ($vertical_spin) { + &encvert($tbl,$s0,$s1,$s2,$s3); + } else { + &encstep(0,$tbl,$s0,$s1,$s2,$s3); + &encstep(1,$tbl,$s1,$s2,$s3,$s0); + &encstep(2,$tbl,$s2,$s3,$s0,$s1); + &encstep(3,$tbl,$s3,$s0,$s1,$s2); + } + &add ($key,16); # advance rd_key + &xor ($s0,&DWP(0,$key)); + &xor ($s1,&DWP(4,$key)); + &xor ($s2,&DWP(8,$key)); + &xor ($s3,&DWP(12,$key)); + &cmp ($key,$__end); + &mov ($__key,$key); + &jb (&label("loop")); + } + else { + &cmp ($acc,10); + &jle (&label("10rounds")); + &cmp ($acc,12); + &jle (&label("12rounds")); + + &set_label("14rounds",4); + for ($i=1;$i<3;$i++) { + if ($vertical_spin) { + &encvert($tbl,$s0,$s1,$s2,$s3); + } else { + &encstep(0,$tbl,$s0,$s1,$s2,$s3); + &encstep(1,$tbl,$s1,$s2,$s3,$s0); + &encstep(2,$tbl,$s2,$s3,$s0,$s1); + &encstep(3,$tbl,$s3,$s0,$s1,$s2); + } + &xor ($s0,&DWP(16*$i+0,$key)); + &xor ($s1,&DWP(16*$i+4,$key)); + &xor ($s2,&DWP(16*$i+8,$key)); + &xor ($s3,&DWP(16*$i+12,$key)); + } + &add ($key,32); + &mov ($__key,$key); # advance rd_key + &set_label("12rounds",4); + for ($i=1;$i<3;$i++) { + if ($vertical_spin) { + &encvert($tbl,$s0,$s1,$s2,$s3); + } else { + &encstep(0,$tbl,$s0,$s1,$s2,$s3); + &encstep(1,$tbl,$s1,$s2,$s3,$s0); + &encstep(2,$tbl,$s2,$s3,$s0,$s1); + &encstep(3,$tbl,$s3,$s0,$s1,$s2); + } + &xor ($s0,&DWP(16*$i+0,$key)); + &xor ($s1,&DWP(16*$i+4,$key)); + &xor ($s2,&DWP(16*$i+8,$key)); + &xor ($s3,&DWP(16*$i+12,$key)); + } + &add ($key,32); + &mov ($__key,$key); # advance rd_key + &set_label("10rounds",4); + for ($i=1;$i<10;$i++) { + if ($vertical_spin) { + &encvert($tbl,$s0,$s1,$s2,$s3); + } else { + &encstep(0,$tbl,$s0,$s1,$s2,$s3); + &encstep(1,$tbl,$s1,$s2,$s3,$s0); + &encstep(2,$tbl,$s2,$s3,$s0,$s1); + &encstep(3,$tbl,$s3,$s0,$s1,$s2); + } + &xor ($s0,&DWP(16*$i+0,$key)); + &xor ($s1,&DWP(16*$i+4,$key)); + &xor ($s2,&DWP(16*$i+8,$key)); + &xor ($s3,&DWP(16*$i+12,$key)); + } + } + + if ($vertical_spin) { + # "reincarnate" some registers for "horizontal" spin... + &mov ($s1="ebx",$key="edi"); + &mov ($s2="ecx",$acc="esi"); + } + &enclast(0,$tbl,$s0,$s1,$s2,$s3); + &enclast(1,$tbl,$s1,$s2,$s3,$s0); + &enclast(2,$tbl,$s2,$s3,$s0,$s1); + &enclast(3,$tbl,$s3,$s0,$s1,$s2); + + &add ($key,$small_footprint?16:160); + &xor ($s0,&DWP(0,$key)); + &xor ($s1,&DWP(4,$key)); + &xor ($s2,&DWP(8,$key)); + &xor ($s3,&DWP(12,$key)); + + &ret (); + +&set_label("AES_Te",64); # Yes! I keep it in the code segment! + &_data_word(0xa56363c6, 0x847c7cf8, 0x997777ee, 0x8d7b7bf6); + &_data_word(0x0df2f2ff, 0xbd6b6bd6, 0xb16f6fde, 0x54c5c591); + &_data_word(0x50303060, 0x03010102, 0xa96767ce, 0x7d2b2b56); + &_data_word(0x19fefee7, 0x62d7d7b5, 0xe6abab4d, 0x9a7676ec); + &_data_word(0x45caca8f, 0x9d82821f, 0x40c9c989, 0x877d7dfa); + &_data_word(0x15fafaef, 0xeb5959b2, 0xc947478e, 0x0bf0f0fb); + &_data_word(0xecadad41, 0x67d4d4b3, 0xfda2a25f, 0xeaafaf45); + &_data_word(0xbf9c9c23, 0xf7a4a453, 0x967272e4, 0x5bc0c09b); + &_data_word(0xc2b7b775, 0x1cfdfde1, 0xae93933d, 0x6a26264c); + &_data_word(0x5a36366c, 0x413f3f7e, 0x02f7f7f5, 0x4fcccc83); + &_data_word(0x5c343468, 0xf4a5a551, 0x34e5e5d1, 0x08f1f1f9); + &_data_word(0x937171e2, 0x73d8d8ab, 0x53313162, 0x3f15152a); + &_data_word(0x0c040408, 0x52c7c795, 0x65232346, 0x5ec3c39d); + &_data_word(0x28181830, 0xa1969637, 0x0f05050a, 0xb59a9a2f); + &_data_word(0x0907070e, 0x36121224, 0x9b80801b, 0x3de2e2df); + &_data_word(0x26ebebcd, 0x6927274e, 0xcdb2b27f, 0x9f7575ea); + &_data_word(0x1b090912, 0x9e83831d, 0x742c2c58, 0x2e1a1a34); + &_data_word(0x2d1b1b36, 0xb26e6edc, 0xee5a5ab4, 0xfba0a05b); + &_data_word(0xf65252a4, 0x4d3b3b76, 0x61d6d6b7, 0xceb3b37d); + &_data_word(0x7b292952, 0x3ee3e3dd, 0x712f2f5e, 0x97848413); + &_data_word(0xf55353a6, 0x68d1d1b9, 0x00000000, 0x2cededc1); + &_data_word(0x60202040, 0x1ffcfce3, 0xc8b1b179, 0xed5b5bb6); + &_data_word(0xbe6a6ad4, 0x46cbcb8d, 0xd9bebe67, 0x4b393972); + &_data_word(0xde4a4a94, 0xd44c4c98, 0xe85858b0, 0x4acfcf85); + &_data_word(0x6bd0d0bb, 0x2aefefc5, 0xe5aaaa4f, 0x16fbfbed); + &_data_word(0xc5434386, 0xd74d4d9a, 0x55333366, 0x94858511); + &_data_word(0xcf45458a, 0x10f9f9e9, 0x06020204, 0x817f7ffe); + &_data_word(0xf05050a0, 0x443c3c78, 0xba9f9f25, 0xe3a8a84b); + &_data_word(0xf35151a2, 0xfea3a35d, 0xc0404080, 0x8a8f8f05); + &_data_word(0xad92923f, 0xbc9d9d21, 0x48383870, 0x04f5f5f1); + &_data_word(0xdfbcbc63, 0xc1b6b677, 0x75dadaaf, 0x63212142); + &_data_word(0x30101020, 0x1affffe5, 0x0ef3f3fd, 0x6dd2d2bf); + &_data_word(0x4ccdcd81, 0x140c0c18, 0x35131326, 0x2fececc3); + &_data_word(0xe15f5fbe, 0xa2979735, 0xcc444488, 0x3917172e); + &_data_word(0x57c4c493, 0xf2a7a755, 0x827e7efc, 0x473d3d7a); + &_data_word(0xac6464c8, 0xe75d5dba, 0x2b191932, 0x957373e6); + &_data_word(0xa06060c0, 0x98818119, 0xd14f4f9e, 0x7fdcdca3); + &_data_word(0x66222244, 0x7e2a2a54, 0xab90903b, 0x8388880b); + &_data_word(0xca46468c, 0x29eeeec7, 0xd3b8b86b, 0x3c141428); + &_data_word(0x79dedea7, 0xe25e5ebc, 0x1d0b0b16, 0x76dbdbad); + &_data_word(0x3be0e0db, 0x56323264, 0x4e3a3a74, 0x1e0a0a14); + &_data_word(0xdb494992, 0x0a06060c, 0x6c242448, 0xe45c5cb8); + &_data_word(0x5dc2c29f, 0x6ed3d3bd, 0xefacac43, 0xa66262c4); + &_data_word(0xa8919139, 0xa4959531, 0x37e4e4d3, 0x8b7979f2); + &_data_word(0x32e7e7d5, 0x43c8c88b, 0x5937376e, 0xb76d6dda); + &_data_word(0x8c8d8d01, 0x64d5d5b1, 0xd24e4e9c, 0xe0a9a949); + &_data_word(0xb46c6cd8, 0xfa5656ac, 0x07f4f4f3, 0x25eaeacf); + &_data_word(0xaf6565ca, 0x8e7a7af4, 0xe9aeae47, 0x18080810); + &_data_word(0xd5baba6f, 0x887878f0, 0x6f25254a, 0x722e2e5c); + &_data_word(0x241c1c38, 0xf1a6a657, 0xc7b4b473, 0x51c6c697); + &_data_word(0x23e8e8cb, 0x7cdddda1, 0x9c7474e8, 0x211f1f3e); + &_data_word(0xdd4b4b96, 0xdcbdbd61, 0x868b8b0d, 0x858a8a0f); + &_data_word(0x907070e0, 0x423e3e7c, 0xc4b5b571, 0xaa6666cc); + &_data_word(0xd8484890, 0x05030306, 0x01f6f6f7, 0x120e0e1c); + &_data_word(0xa36161c2, 0x5f35356a, 0xf95757ae, 0xd0b9b969); + &_data_word(0x91868617, 0x58c1c199, 0x271d1d3a, 0xb99e9e27); + &_data_word(0x38e1e1d9, 0x13f8f8eb, 0xb398982b, 0x33111122); + &_data_word(0xbb6969d2, 0x70d9d9a9, 0x898e8e07, 0xa7949433); + &_data_word(0xb69b9b2d, 0x221e1e3c, 0x92878715, 0x20e9e9c9); + &_data_word(0x49cece87, 0xff5555aa, 0x78282850, 0x7adfdfa5); + &_data_word(0x8f8c8c03, 0xf8a1a159, 0x80898909, 0x170d0d1a); + &_data_word(0xdabfbf65, 0x31e6e6d7, 0xc6424284, 0xb86868d0); + &_data_word(0xc3414182, 0xb0999929, 0x772d2d5a, 0x110f0f1e); + &_data_word(0xcbb0b07b, 0xfc5454a8, 0xd6bbbb6d, 0x3a16162c); + +#Te4 # four copies of Te4 to choose from to avoid L1 aliasing + &data_byte(0x63, 0x7c, 0x77, 0x7b, 0xf2, 0x6b, 0x6f, 0xc5); + &data_byte(0x30, 0x01, 0x67, 0x2b, 0xfe, 0xd7, 0xab, 0x76); + &data_byte(0xca, 0x82, 0xc9, 0x7d, 0xfa, 0x59, 0x47, 0xf0); + &data_byte(0xad, 0xd4, 0xa2, 0xaf, 0x9c, 0xa4, 0x72, 0xc0); + &data_byte(0xb7, 0xfd, 0x93, 0x26, 0x36, 0x3f, 0xf7, 0xcc); + &data_byte(0x34, 0xa5, 0xe5, 0xf1, 0x71, 0xd8, 0x31, 0x15); + &data_byte(0x04, 0xc7, 0x23, 0xc3, 0x18, 0x96, 0x05, 0x9a); + &data_byte(0x07, 0x12, 0x80, 0xe2, 0xeb, 0x27, 0xb2, 0x75); + &data_byte(0x09, 0x83, 0x2c, 0x1a, 0x1b, 0x6e, 0x5a, 0xa0); + &data_byte(0x52, 0x3b, 0xd6, 0xb3, 0x29, 0xe3, 0x2f, 0x84); + &data_byte(0x53, 0xd1, 0x00, 0xed, 0x20, 0xfc, 0xb1, 0x5b); + &data_byte(0x6a, 0xcb, 0xbe, 0x39, 0x4a, 0x4c, 0x58, 0xcf); + &data_byte(0xd0, 0xef, 0xaa, 0xfb, 0x43, 0x4d, 0x33, 0x85); + &data_byte(0x45, 0xf9, 0x02, 0x7f, 0x50, 0x3c, 0x9f, 0xa8); + &data_byte(0x51, 0xa3, 0x40, 0x8f, 0x92, 0x9d, 0x38, 0xf5); + &data_byte(0xbc, 0xb6, 0xda, 0x21, 0x10, 0xff, 0xf3, 0xd2); + &data_byte(0xcd, 0x0c, 0x13, 0xec, 0x5f, 0x97, 0x44, 0x17); + &data_byte(0xc4, 0xa7, 0x7e, 0x3d, 0x64, 0x5d, 0x19, 0x73); + &data_byte(0x60, 0x81, 0x4f, 0xdc, 0x22, 0x2a, 0x90, 0x88); + &data_byte(0x46, 0xee, 0xb8, 0x14, 0xde, 0x5e, 0x0b, 0xdb); + &data_byte(0xe0, 0x32, 0x3a, 0x0a, 0x49, 0x06, 0x24, 0x5c); + &data_byte(0xc2, 0xd3, 0xac, 0x62, 0x91, 0x95, 0xe4, 0x79); + &data_byte(0xe7, 0xc8, 0x37, 0x6d, 0x8d, 0xd5, 0x4e, 0xa9); + &data_byte(0x6c, 0x56, 0xf4, 0xea, 0x65, 0x7a, 0xae, 0x08); + &data_byte(0xba, 0x78, 0x25, 0x2e, 0x1c, 0xa6, 0xb4, 0xc6); + &data_byte(0xe8, 0xdd, 0x74, 0x1f, 0x4b, 0xbd, 0x8b, 0x8a); + &data_byte(0x70, 0x3e, 0xb5, 0x66, 0x48, 0x03, 0xf6, 0x0e); + &data_byte(0x61, 0x35, 0x57, 0xb9, 0x86, 0xc1, 0x1d, 0x9e); + &data_byte(0xe1, 0xf8, 0x98, 0x11, 0x69, 0xd9, 0x8e, 0x94); + &data_byte(0x9b, 0x1e, 0x87, 0xe9, 0xce, 0x55, 0x28, 0xdf); + &data_byte(0x8c, 0xa1, 0x89, 0x0d, 0xbf, 0xe6, 0x42, 0x68); + &data_byte(0x41, 0x99, 0x2d, 0x0f, 0xb0, 0x54, 0xbb, 0x16); + + &data_byte(0x63, 0x7c, 0x77, 0x7b, 0xf2, 0x6b, 0x6f, 0xc5); + &data_byte(0x30, 0x01, 0x67, 0x2b, 0xfe, 0xd7, 0xab, 0x76); + &data_byte(0xca, 0x82, 0xc9, 0x7d, 0xfa, 0x59, 0x47, 0xf0); + &data_byte(0xad, 0xd4, 0xa2, 0xaf, 0x9c, 0xa4, 0x72, 0xc0); + &data_byte(0xb7, 0xfd, 0x93, 0x26, 0x36, 0x3f, 0xf7, 0xcc); + &data_byte(0x34, 0xa5, 0xe5, 0xf1, 0x71, 0xd8, 0x31, 0x15); + &data_byte(0x04, 0xc7, 0x23, 0xc3, 0x18, 0x96, 0x05, 0x9a); + &data_byte(0x07, 0x12, 0x80, 0xe2, 0xeb, 0x27, 0xb2, 0x75); + &data_byte(0x09, 0x83, 0x2c, 0x1a, 0x1b, 0x6e, 0x5a, 0xa0); + &data_byte(0x52, 0x3b, 0xd6, 0xb3, 0x29, 0xe3, 0x2f, 0x84); + &data_byte(0x53, 0xd1, 0x00, 0xed, 0x20, 0xfc, 0xb1, 0x5b); + &data_byte(0x6a, 0xcb, 0xbe, 0x39, 0x4a, 0x4c, 0x58, 0xcf); + &data_byte(0xd0, 0xef, 0xaa, 0xfb, 0x43, 0x4d, 0x33, 0x85); + &data_byte(0x45, 0xf9, 0x02, 0x7f, 0x50, 0x3c, 0x9f, 0xa8); + &data_byte(0x51, 0xa3, 0x40, 0x8f, 0x92, 0x9d, 0x38, 0xf5); + &data_byte(0xbc, 0xb6, 0xda, 0x21, 0x10, 0xff, 0xf3, 0xd2); + &data_byte(0xcd, 0x0c, 0x13, 0xec, 0x5f, 0x97, 0x44, 0x17); + &data_byte(0xc4, 0xa7, 0x7e, 0x3d, 0x64, 0x5d, 0x19, 0x73); + &data_byte(0x60, 0x81, 0x4f, 0xdc, 0x22, 0x2a, 0x90, 0x88); + &data_byte(0x46, 0xee, 0xb8, 0x14, 0xde, 0x5e, 0x0b, 0xdb); + &data_byte(0xe0, 0x32, 0x3a, 0x0a, 0x49, 0x06, 0x24, 0x5c); + &data_byte(0xc2, 0xd3, 0xac, 0x62, 0x91, 0x95, 0xe4, 0x79); + &data_byte(0xe7, 0xc8, 0x37, 0x6d, 0x8d, 0xd5, 0x4e, 0xa9); + &data_byte(0x6c, 0x56, 0xf4, 0xea, 0x65, 0x7a, 0xae, 0x08); + &data_byte(0xba, 0x78, 0x25, 0x2e, 0x1c, 0xa6, 0xb4, 0xc6); + &data_byte(0xe8, 0xdd, 0x74, 0x1f, 0x4b, 0xbd, 0x8b, 0x8a); + &data_byte(0x70, 0x3e, 0xb5, 0x66, 0x48, 0x03, 0xf6, 0x0e); + &data_byte(0x61, 0x35, 0x57, 0xb9, 0x86, 0xc1, 0x1d, 0x9e); + &data_byte(0xe1, 0xf8, 0x98, 0x11, 0x69, 0xd9, 0x8e, 0x94); + &data_byte(0x9b, 0x1e, 0x87, 0xe9, 0xce, 0x55, 0x28, 0xdf); + &data_byte(0x8c, 0xa1, 0x89, 0x0d, 0xbf, 0xe6, 0x42, 0x68); + &data_byte(0x41, 0x99, 0x2d, 0x0f, 0xb0, 0x54, 0xbb, 0x16); + + &data_byte(0x63, 0x7c, 0x77, 0x7b, 0xf2, 0x6b, 0x6f, 0xc5); + &data_byte(0x30, 0x01, 0x67, 0x2b, 0xfe, 0xd7, 0xab, 0x76); + &data_byte(0xca, 0x82, 0xc9, 0x7d, 0xfa, 0x59, 0x47, 0xf0); + &data_byte(0xad, 0xd4, 0xa2, 0xaf, 0x9c, 0xa4, 0x72, 0xc0); + &data_byte(0xb7, 0xfd, 0x93, 0x26, 0x36, 0x3f, 0xf7, 0xcc); + &data_byte(0x34, 0xa5, 0xe5, 0xf1, 0x71, 0xd8, 0x31, 0x15); + &data_byte(0x04, 0xc7, 0x23, 0xc3, 0x18, 0x96, 0x05, 0x9a); + &data_byte(0x07, 0x12, 0x80, 0xe2, 0xeb, 0x27, 0xb2, 0x75); + &data_byte(0x09, 0x83, 0x2c, 0x1a, 0x1b, 0x6e, 0x5a, 0xa0); + &data_byte(0x52, 0x3b, 0xd6, 0xb3, 0x29, 0xe3, 0x2f, 0x84); + &data_byte(0x53, 0xd1, 0x00, 0xed, 0x20, 0xfc, 0xb1, 0x5b); + &data_byte(0x6a, 0xcb, 0xbe, 0x39, 0x4a, 0x4c, 0x58, 0xcf); + &data_byte(0xd0, 0xef, 0xaa, 0xfb, 0x43, 0x4d, 0x33, 0x85); + &data_byte(0x45, 0xf9, 0x02, 0x7f, 0x50, 0x3c, 0x9f, 0xa8); + &data_byte(0x51, 0xa3, 0x40, 0x8f, 0x92, 0x9d, 0x38, 0xf5); + &data_byte(0xbc, 0xb6, 0xda, 0x21, 0x10, 0xff, 0xf3, 0xd2); + &data_byte(0xcd, 0x0c, 0x13, 0xec, 0x5f, 0x97, 0x44, 0x17); + &data_byte(0xc4, 0xa7, 0x7e, 0x3d, 0x64, 0x5d, 0x19, 0x73); + &data_byte(0x60, 0x81, 0x4f, 0xdc, 0x22, 0x2a, 0x90, 0x88); + &data_byte(0x46, 0xee, 0xb8, 0x14, 0xde, 0x5e, 0x0b, 0xdb); + &data_byte(0xe0, 0x32, 0x3a, 0x0a, 0x49, 0x06, 0x24, 0x5c); + &data_byte(0xc2, 0xd3, 0xac, 0x62, 0x91, 0x95, 0xe4, 0x79); + &data_byte(0xe7, 0xc8, 0x37, 0x6d, 0x8d, 0xd5, 0x4e, 0xa9); + &data_byte(0x6c, 0x56, 0xf4, 0xea, 0x65, 0x7a, 0xae, 0x08); + &data_byte(0xba, 0x78, 0x25, 0x2e, 0x1c, 0xa6, 0xb4, 0xc6); + &data_byte(0xe8, 0xdd, 0x74, 0x1f, 0x4b, 0xbd, 0x8b, 0x8a); + &data_byte(0x70, 0x3e, 0xb5, 0x66, 0x48, 0x03, 0xf6, 0x0e); + &data_byte(0x61, 0x35, 0x57, 0xb9, 0x86, 0xc1, 0x1d, 0x9e); + &data_byte(0xe1, 0xf8, 0x98, 0x11, 0x69, 0xd9, 0x8e, 0x94); + &data_byte(0x9b, 0x1e, 0x87, 0xe9, 0xce, 0x55, 0x28, 0xdf); + &data_byte(0x8c, 0xa1, 0x89, 0x0d, 0xbf, 0xe6, 0x42, 0x68); + &data_byte(0x41, 0x99, 0x2d, 0x0f, 0xb0, 0x54, 0xbb, 0x16); + + &data_byte(0x63, 0x7c, 0x77, 0x7b, 0xf2, 0x6b, 0x6f, 0xc5); + &data_byte(0x30, 0x01, 0x67, 0x2b, 0xfe, 0xd7, 0xab, 0x76); + &data_byte(0xca, 0x82, 0xc9, 0x7d, 0xfa, 0x59, 0x47, 0xf0); + &data_byte(0xad, 0xd4, 0xa2, 0xaf, 0x9c, 0xa4, 0x72, 0xc0); + &data_byte(0xb7, 0xfd, 0x93, 0x26, 0x36, 0x3f, 0xf7, 0xcc); + &data_byte(0x34, 0xa5, 0xe5, 0xf1, 0x71, 0xd8, 0x31, 0x15); + &data_byte(0x04, 0xc7, 0x23, 0xc3, 0x18, 0x96, 0x05, 0x9a); + &data_byte(0x07, 0x12, 0x80, 0xe2, 0xeb, 0x27, 0xb2, 0x75); + &data_byte(0x09, 0x83, 0x2c, 0x1a, 0x1b, 0x6e, 0x5a, 0xa0); + &data_byte(0x52, 0x3b, 0xd6, 0xb3, 0x29, 0xe3, 0x2f, 0x84); + &data_byte(0x53, 0xd1, 0x00, 0xed, 0x20, 0xfc, 0xb1, 0x5b); + &data_byte(0x6a, 0xcb, 0xbe, 0x39, 0x4a, 0x4c, 0x58, 0xcf); + &data_byte(0xd0, 0xef, 0xaa, 0xfb, 0x43, 0x4d, 0x33, 0x85); + &data_byte(0x45, 0xf9, 0x02, 0x7f, 0x50, 0x3c, 0x9f, 0xa8); + &data_byte(0x51, 0xa3, 0x40, 0x8f, 0x92, 0x9d, 0x38, 0xf5); + &data_byte(0xbc, 0xb6, 0xda, 0x21, 0x10, 0xff, 0xf3, 0xd2); + &data_byte(0xcd, 0x0c, 0x13, 0xec, 0x5f, 0x97, 0x44, 0x17); + &data_byte(0xc4, 0xa7, 0x7e, 0x3d, 0x64, 0x5d, 0x19, 0x73); + &data_byte(0x60, 0x81, 0x4f, 0xdc, 0x22, 0x2a, 0x90, 0x88); + &data_byte(0x46, 0xee, 0xb8, 0x14, 0xde, 0x5e, 0x0b, 0xdb); + &data_byte(0xe0, 0x32, 0x3a, 0x0a, 0x49, 0x06, 0x24, 0x5c); + &data_byte(0xc2, 0xd3, 0xac, 0x62, 0x91, 0x95, 0xe4, 0x79); + &data_byte(0xe7, 0xc8, 0x37, 0x6d, 0x8d, 0xd5, 0x4e, 0xa9); + &data_byte(0x6c, 0x56, 0xf4, 0xea, 0x65, 0x7a, 0xae, 0x08); + &data_byte(0xba, 0x78, 0x25, 0x2e, 0x1c, 0xa6, 0xb4, 0xc6); + &data_byte(0xe8, 0xdd, 0x74, 0x1f, 0x4b, 0xbd, 0x8b, 0x8a); + &data_byte(0x70, 0x3e, 0xb5, 0x66, 0x48, 0x03, 0xf6, 0x0e); + &data_byte(0x61, 0x35, 0x57, 0xb9, 0x86, 0xc1, 0x1d, 0x9e); + &data_byte(0xe1, 0xf8, 0x98, 0x11, 0x69, 0xd9, 0x8e, 0x94); + &data_byte(0x9b, 0x1e, 0x87, 0xe9, 0xce, 0x55, 0x28, 0xdf); + &data_byte(0x8c, 0xa1, 0x89, 0x0d, 0xbf, 0xe6, 0x42, 0x68); + &data_byte(0x41, 0x99, 0x2d, 0x0f, 0xb0, 0x54, 0xbb, 0x16); +#rcon: + &data_word(0x00000001, 0x00000002, 0x00000004, 0x00000008); + &data_word(0x00000010, 0x00000020, 0x00000040, 0x00000080); + &data_word(0x0000001b, 0x00000036, 0x00000000, 0x00000000); + &data_word(0x00000000, 0x00000000, 0x00000000, 0x00000000); +&function_end_B("_x86_AES_encrypt"); + +# void AES_encrypt (const void *inp,void *out,const AES_KEY *key); +&function_begin("AES_encrypt"); + &mov ($acc,&wparam(0)); # load inp + &mov ($key,&wparam(2)); # load key + + &mov ($s0,"esp"); + &sub ("esp",36); + &and ("esp",-64); # align to cache-line + + # place stack frame just "above" the key schedule + &lea ($s1,&DWP(-64-63,$key)); + &sub ($s1,"esp"); + &neg ($s1); + &and ($s1,0x3C0); # modulo 1024, but aligned to cache-line + &sub ("esp",$s1); + &add ("esp",4); # 4 is reserved for caller's return address + &mov ($_esp,$s0); # save stack pointer + + &call (&label("pic_point")); # make it PIC! + &set_label("pic_point"); + &blindpop($tbl); + &picmeup($s0,"OPENSSL_ia32cap_P",$tbl,&label("pic_point")) if (!$x86only); + &lea ($tbl,&DWP(&label("AES_Te")."-".&label("pic_point"),$tbl)); + + # pick Te4 copy which can't "overlap" with stack frame or key schedule + &lea ($s1,&DWP(768-4,"esp")); + &sub ($s1,$tbl); + &and ($s1,0x300); + &lea ($tbl,&DWP(2048+128,$tbl,$s1)); + + if (!$x86only) { + &bt (&DWP(0,$s0),25); # check for SSE bit + &jnc (&label("x86")); + + &movq ("mm0",&QWP(0,$acc)); + &movq ("mm4",&QWP(8,$acc)); + &call ("_sse_AES_encrypt_compact"); + &mov ("esp",$_esp); # restore stack pointer + &mov ($acc,&wparam(1)); # load out + &movq (&QWP(0,$acc),"mm0"); # write output data + &movq (&QWP(8,$acc),"mm4"); + &emms (); + &function_end_A(); + } + &set_label("x86",16); + &mov ($_tbl,$tbl); + &mov ($s0,&DWP(0,$acc)); # load input data + &mov ($s1,&DWP(4,$acc)); + &mov ($s2,&DWP(8,$acc)); + &mov ($s3,&DWP(12,$acc)); + &call ("_x86_AES_encrypt_compact"); + &mov ("esp",$_esp); # restore stack pointer + &mov ($acc,&wparam(1)); # load out + &mov (&DWP(0,$acc),$s0); # write output data + &mov (&DWP(4,$acc),$s1); + &mov (&DWP(8,$acc),$s2); + &mov (&DWP(12,$acc),$s3); +&function_end("AES_encrypt"); + +#--------------------------------------------------------------------# + +###################################################################### +# "Compact" block function +###################################################################### + +sub deccompact() +{ my $Fn = \&mov; + while ($#_>5) { pop(@_); $Fn=sub{}; } + my ($i,$td,@s)=@_; + my $tmp = $key; + my $out = $i==3?$s[0]:$acc; + + # $Fn is used in first compact round and its purpose is to + # void restoration of some values from stack, so that after + # 4xdeccompact with extra argument $key, $s0 and $s1 values + # are left there... + if($i==3) { &$Fn ($key,$__key); } + else { &mov ($out,$s[0]); } + &and ($out,0xFF); + &movz ($out,&BP(-128,$td,$out,1)); + + if ($i==3) { $tmp=$s[1]; } + &movz ($tmp,&HB($s[1])); + &movz ($tmp,&BP(-128,$td,$tmp,1)); + &shl ($tmp,8); + &xor ($out,$tmp); + + if ($i==3) { $tmp=$s[2]; &mov ($s[1],$acc); } + else { mov ($tmp,$s[2]); } + &shr ($tmp,16); + &and ($tmp,0xFF); + &movz ($tmp,&BP(-128,$td,$tmp,1)); + &shl ($tmp,16); + &xor ($out,$tmp); + + if ($i==3) { $tmp=$s[3]; &$Fn ($s[2],$__s1); } + else { &mov ($tmp,$s[3]); } + &shr ($tmp,24); + &movz ($tmp,&BP(-128,$td,$tmp,1)); + &shl ($tmp,24); + &xor ($out,$tmp); + if ($i<2) { &mov (&DWP(4+4*$i,"esp"),$out); } + if ($i==3) { &$Fn ($s[3],$__s0); } +} + +# must be called with 2,3,0,1 as argument sequence!!! +sub dectransform() +{ my @s = ($s0,$s1,$s2,$s3); + my $i = shift; + my $tmp = $key; + my $tp2 = @s[($i+2)%4]; $tp2 = @s[2] if ($i==1); + my $tp4 = @s[($i+3)%4]; $tp4 = @s[3] if ($i==1); + my $tp8 = $tbl; + + &mov ($tmp,0x80808080); + &and ($tmp,$s[$i]); + &mov ($acc,$tmp); + &shr ($tmp,7); + &lea ($tp2,&DWP(0,$s[$i],$s[$i])); + &sub ($acc,$tmp); + &and ($tp2,0xfefefefe); + &and ($acc,0x1b1b1b1b); + &xor ($tp2,$acc); + &mov ($tmp,0x80808080); + + &and ($tmp,$tp2); + &mov ($acc,$tmp); + &shr ($tmp,7); + &lea ($tp4,&DWP(0,$tp2,$tp2)); + &sub ($acc,$tmp); + &and ($tp4,0xfefefefe); + &and ($acc,0x1b1b1b1b); + &xor ($tp2,$s[$i]); # tp2^tp1 + &xor ($tp4,$acc); + &mov ($tmp,0x80808080); + + &and ($tmp,$tp4); + &mov ($acc,$tmp); + &shr ($tmp,7); + &lea ($tp8,&DWP(0,$tp4,$tp4)); + &sub ($acc,$tmp); + &and ($tp8,0xfefefefe); + &and ($acc,0x1b1b1b1b); + &xor ($tp4,$s[$i]); # tp4^tp1 + &rotl ($s[$i],8); # = ROTATE(tp1,8) + &xor ($tp8,$acc); + + &xor ($s[$i],$tp2); + &xor ($tp2,$tp8); + &xor ($s[$i],$tp4); + &xor ($tp4,$tp8); + &rotl ($tp2,24); + &xor ($s[$i],$tp8); # ^= tp8^(tp4^tp1)^(tp2^tp1) + &rotl ($tp4,16); + &xor ($s[$i],$tp2); # ^= ROTATE(tp8^tp2^tp1,24) + &rotl ($tp8,8); + &xor ($s[$i],$tp4); # ^= ROTATE(tp8^tp4^tp1,16) + &mov ($s[0],$__s0) if($i==2); #prefetch $s0 + &mov ($s[1],$__s1) if($i==3); #prefetch $s1 + &mov ($s[2],$__s2) if($i==1); + &xor ($s[$i],$tp8); # ^= ROTATE(tp8,8) + + &mov ($s[3],$__s3) if($i==1); + &mov (&DWP(4+4*$i,"esp"),$s[$i]) if($i>=2); +} + +&function_begin_B("_x86_AES_decrypt_compact"); + # note that caller is expected to allocate stack frame for me! + &mov ($__key,$key); # save key + + &xor ($s0,&DWP(0,$key)); # xor with key + &xor ($s1,&DWP(4,$key)); + &xor ($s2,&DWP(8,$key)); + &xor ($s3,&DWP(12,$key)); + + &mov ($acc,&DWP(240,$key)); # load key->rounds + + &lea ($acc,&DWP(-2,$acc,$acc)); + &lea ($acc,&DWP(0,$key,$acc,8)); + &mov ($__end,$acc); # end of key schedule + + # prefetch Td4 + &mov ($key,&DWP(0-128,$tbl)); + &mov ($acc,&DWP(32-128,$tbl)); + &mov ($key,&DWP(64-128,$tbl)); + &mov ($acc,&DWP(96-128,$tbl)); + &mov ($key,&DWP(128-128,$tbl)); + &mov ($acc,&DWP(160-128,$tbl)); + &mov ($key,&DWP(192-128,$tbl)); + &mov ($acc,&DWP(224-128,$tbl)); + + &set_label("loop",16); + + &deccompact(0,$tbl,$s0,$s3,$s2,$s1,1); + &deccompact(1,$tbl,$s1,$s0,$s3,$s2,1); + &deccompact(2,$tbl,$s2,$s1,$s0,$s3,1); + &deccompact(3,$tbl,$s3,$s2,$s1,$s0,1); + &dectransform(2); + &dectransform(3); + &dectransform(0); + &dectransform(1); + &mov ($key,$__key); + &mov ($tbl,$__tbl); + &add ($key,16); # advance rd_key + &xor ($s0,&DWP(0,$key)); + &xor ($s1,&DWP(4,$key)); + &xor ($s2,&DWP(8,$key)); + &xor ($s3,&DWP(12,$key)); + + &cmp ($key,$__end); + &mov ($__key,$key); + &jb (&label("loop")); + + &deccompact(0,$tbl,$s0,$s3,$s2,$s1); + &deccompact(1,$tbl,$s1,$s0,$s3,$s2); + &deccompact(2,$tbl,$s2,$s1,$s0,$s3); + &deccompact(3,$tbl,$s3,$s2,$s1,$s0); + + &xor ($s0,&DWP(16,$key)); + &xor ($s1,&DWP(20,$key)); + &xor ($s2,&DWP(24,$key)); + &xor ($s3,&DWP(28,$key)); + + &ret (); +&function_end_B("_x86_AES_decrypt_compact"); + +###################################################################### +# "Compact" SSE block function. +###################################################################### + +sub sse_deccompact() +{ + &pshufw ("mm1","mm0",0x0c); # 7, 6, 1, 0 + &pshufw ("mm5","mm4",0x09); # 13,12,11,10 + &movd ("eax","mm1"); # 7, 6, 1, 0 + &movd ("ebx","mm5"); # 13,12,11,10 + &mov ($__key,$key); + + &movz ($acc,&LB("eax")); # 0 + &movz ("edx",&HB("eax")); # 1 + &pshufw ("mm2","mm0",0x06); # 3, 2, 5, 4 + &movz ("ecx",&BP(-128,$tbl,$acc,1)); # 0 + &movz ($key,&LB("ebx")); # 10 + &movz ("edx",&BP(-128,$tbl,"edx",1)); # 1 + &shr ("eax",16); # 7, 6 + &shl ("edx",8); # 1 + + &movz ($acc,&BP(-128,$tbl,$key,1)); # 10 + &movz ($key,&HB("ebx")); # 11 + &shl ($acc,16); # 10 + &pshufw ("mm6","mm4",0x03); # 9, 8,15,14 + &or ("ecx",$acc); # 10 + &movz ($acc,&BP(-128,$tbl,$key,1)); # 11 + &movz ($key,&HB("eax")); # 7 + &shl ($acc,24); # 11 + &shr ("ebx",16); # 13,12 + &or ("edx",$acc); # 11 + + &movz ($acc,&BP(-128,$tbl,$key,1)); # 7 + &movz ($key,&HB("ebx")); # 13 + &shl ($acc,24); # 7 + &or ("ecx",$acc); # 7 + &movz ($acc,&BP(-128,$tbl,$key,1)); # 13 + &movz ($key,&LB("eax")); # 6 + &shl ($acc,8); # 13 + &movd ("eax","mm2"); # 3, 2, 5, 4 + &or ("ecx",$acc); # 13 + + &movz ($acc,&BP(-128,$tbl,$key,1)); # 6 + &movz ($key,&LB("ebx")); # 12 + &shl ($acc,16); # 6 + &movd ("ebx","mm6"); # 9, 8,15,14 + &movd ("mm0","ecx"); # t[0] collected + &movz ("ecx",&BP(-128,$tbl,$key,1)); # 12 + &movz ($key,&LB("eax")); # 4 + &or ("ecx",$acc); # 12 + + &movz ($acc,&BP(-128,$tbl,$key,1)); # 4 + &movz ($key,&LB("ebx")); # 14 + &or ("edx",$acc); # 4 + &movz ($acc,&BP(-128,$tbl,$key,1)); # 14 + &movz ($key,&HB("eax")); # 5 + &shl ($acc,16); # 14 + &shr ("eax",16); # 3, 2 + &or ("edx",$acc); # 14 + + &movz ($acc,&BP(-128,$tbl,$key,1)); # 5 + &movz ($key,&HB("ebx")); # 15 + &shr ("ebx",16); # 9, 8 + &shl ($acc,8); # 5 + &movd ("mm1","edx"); # t[1] collected + &movz ("edx",&BP(-128,$tbl,$key,1)); # 15 + &movz ($key,&HB("ebx")); # 9 + &shl ("edx",24); # 15 + &and ("ebx",0xff); # 8 + &or ("edx",$acc); # 15 + + &punpckldq ("mm0","mm1"); # t[0,1] collected + + &movz ($acc,&BP(-128,$tbl,$key,1)); # 9 + &movz ($key,&LB("eax")); # 2 + &shl ($acc,8); # 9 + &movz ("eax",&HB("eax")); # 3 + &movz ("ebx",&BP(-128,$tbl,"ebx",1)); # 8 + &or ("ecx",$acc); # 9 + &movz ($acc,&BP(-128,$tbl,$key,1)); # 2 + &or ("edx","ebx"); # 8 + &shl ($acc,16); # 2 + &movz ("eax",&BP(-128,$tbl,"eax",1)); # 3 + &or ("edx",$acc); # 2 + &shl ("eax",24); # 3 + &or ("ecx","eax"); # 3 + &mov ($key,$__key); + &movd ("mm4","edx"); # t[2] collected + &movd ("mm5","ecx"); # t[3] collected + + &punpckldq ("mm4","mm5"); # t[2,3] collected +} + + if (!$x86only) { +&function_begin_B("_sse_AES_decrypt_compact"); + &pxor ("mm0",&QWP(0,$key)); # 7, 6, 5, 4, 3, 2, 1, 0 + &pxor ("mm4",&QWP(8,$key)); # 15,14,13,12,11,10, 9, 8 + + # note that caller is expected to allocate stack frame for me! + &mov ($acc,&DWP(240,$key)); # load key->rounds + &lea ($acc,&DWP(-2,$acc,$acc)); + &lea ($acc,&DWP(0,$key,$acc,8)); + &mov ($__end,$acc); # end of key schedule + + &mov ($s0,0x1b1b1b1b); # magic constant + &mov (&DWP(8,"esp"),$s0); + &mov (&DWP(12,"esp"),$s0); + + # prefetch Td4 + &mov ($s0,&DWP(0-128,$tbl)); + &mov ($s1,&DWP(32-128,$tbl)); + &mov ($s2,&DWP(64-128,$tbl)); + &mov ($s3,&DWP(96-128,$tbl)); + &mov ($s0,&DWP(128-128,$tbl)); + &mov ($s1,&DWP(160-128,$tbl)); + &mov ($s2,&DWP(192-128,$tbl)); + &mov ($s3,&DWP(224-128,$tbl)); + + &set_label("loop",16); + &sse_deccompact(); + &add ($key,16); + &cmp ($key,$__end); + &ja (&label("out")); + + # ROTATE(x^y,N) == ROTATE(x,N)^ROTATE(y,N) + &movq ("mm3","mm0"); &movq ("mm7","mm4"); + &movq ("mm2","mm0",1); &movq ("mm6","mm4",1); + &movq ("mm1","mm0"); &movq ("mm5","mm4"); + &pshufw ("mm0","mm0",0xb1); &pshufw ("mm4","mm4",0xb1);# = ROTATE(tp0,16) + &pslld ("mm2",8); &pslld ("mm6",8); + &psrld ("mm3",8); &psrld ("mm7",8); + &pxor ("mm0","mm2"); &pxor ("mm4","mm6"); # ^= tp0<<8 + &pxor ("mm0","mm3"); &pxor ("mm4","mm7"); # ^= tp0>>8 + &pslld ("mm2",16); &pslld ("mm6",16); + &psrld ("mm3",16); &psrld ("mm7",16); + &pxor ("mm0","mm2"); &pxor ("mm4","mm6"); # ^= tp0<<24 + &pxor ("mm0","mm3"); &pxor ("mm4","mm7"); # ^= tp0>>24 + + &movq ("mm3",&QWP(8,"esp")); + &pxor ("mm2","mm2"); &pxor ("mm6","mm6"); + &pcmpgtb("mm2","mm1"); &pcmpgtb("mm6","mm5"); + &pand ("mm2","mm3"); &pand ("mm6","mm3"); + &paddb ("mm1","mm1"); &paddb ("mm5","mm5"); + &pxor ("mm1","mm2"); &pxor ("mm5","mm6"); # tp2 + &movq ("mm3","mm1"); &movq ("mm7","mm5"); + &movq ("mm2","mm1"); &movq ("mm6","mm5"); + &pxor ("mm0","mm1"); &pxor ("mm4","mm5"); # ^= tp2 + &pslld ("mm3",24); &pslld ("mm7",24); + &psrld ("mm2",8); &psrld ("mm6",8); + &pxor ("mm0","mm3"); &pxor ("mm4","mm7"); # ^= tp2<<24 + &pxor ("mm0","mm2"); &pxor ("mm4","mm6"); # ^= tp2>>8 + + &movq ("mm2",&QWP(8,"esp")); + &pxor ("mm3","mm3"); &pxor ("mm7","mm7"); + &pcmpgtb("mm3","mm1"); &pcmpgtb("mm7","mm5"); + &pand ("mm3","mm2"); &pand ("mm7","mm2"); + &paddb ("mm1","mm1"); &paddb ("mm5","mm5"); + &pxor ("mm1","mm3"); &pxor ("mm5","mm7"); # tp4 + &pshufw ("mm3","mm1",0xb1); &pshufw ("mm7","mm5",0xb1); + &pxor ("mm0","mm1"); &pxor ("mm4","mm5"); # ^= tp4 + &pxor ("mm0","mm3"); &pxor ("mm4","mm7"); # ^= ROTATE(tp4,16) + + &pxor ("mm3","mm3"); &pxor ("mm7","mm7"); + &pcmpgtb("mm3","mm1"); &pcmpgtb("mm7","mm5"); + &pand ("mm3","mm2"); &pand ("mm7","mm2"); + &paddb ("mm1","mm1"); &paddb ("mm5","mm5"); + &pxor ("mm1","mm3"); &pxor ("mm5","mm7"); # tp8 + &pxor ("mm0","mm1"); &pxor ("mm4","mm5"); # ^= tp8 + &movq ("mm3","mm1"); &movq ("mm7","mm5"); + &pshufw ("mm2","mm1",0xb1); &pshufw ("mm6","mm5",0xb1); + &pxor ("mm0","mm2"); &pxor ("mm4","mm6"); # ^= ROTATE(tp8,16) + &pslld ("mm1",8); &pslld ("mm5",8); + &psrld ("mm3",8); &psrld ("mm7",8); + &movq ("mm2",&QWP(0,$key)); &movq ("mm6",&QWP(8,$key)); + &pxor ("mm0","mm1"); &pxor ("mm4","mm5"); # ^= tp8<<8 + &pxor ("mm0","mm3"); &pxor ("mm4","mm7"); # ^= tp8>>8 + &mov ($s0,&DWP(0-128,$tbl)); + &pslld ("mm1",16); &pslld ("mm5",16); + &mov ($s1,&DWP(64-128,$tbl)); + &psrld ("mm3",16); &psrld ("mm7",16); + &mov ($s2,&DWP(128-128,$tbl)); + &pxor ("mm0","mm1"); &pxor ("mm4","mm5"); # ^= tp8<<24 + &mov ($s3,&DWP(192-128,$tbl)); + &pxor ("mm0","mm3"); &pxor ("mm4","mm7"); # ^= tp8>>24 + + &pxor ("mm0","mm2"); &pxor ("mm4","mm6"); + &jmp (&label("loop")); + + &set_label("out",16); + &pxor ("mm0",&QWP(0,$key)); + &pxor ("mm4",&QWP(8,$key)); + + &ret (); +&function_end_B("_sse_AES_decrypt_compact"); + } + +###################################################################### +# Vanilla block function. +###################################################################### + +sub decstep() +{ my ($i,$td,@s) = @_; + my $tmp = $key; + my $out = $i==3?$s[0]:$acc; + + # no instructions are reordered, as performance appears + # optimal... or rather that all attempts to reorder didn't + # result in better performance [which by the way is not a + # bit lower than ecryption]. + if($i==3) { &mov ($key,$__key); } + else { &mov ($out,$s[0]); } + &and ($out,0xFF); + &mov ($out,&DWP(0,$td,$out,8)); + + if ($i==3) { $tmp=$s[1]; } + &movz ($tmp,&HB($s[1])); + &xor ($out,&DWP(3,$td,$tmp,8)); + + if ($i==3) { $tmp=$s[2]; &mov ($s[1],$acc); } + else { &mov ($tmp,$s[2]); } + &shr ($tmp,16); + &and ($tmp,0xFF); + &xor ($out,&DWP(2,$td,$tmp,8)); + + if ($i==3) { $tmp=$s[3]; &mov ($s[2],$__s1); } + else { &mov ($tmp,$s[3]); } + &shr ($tmp,24); + &xor ($out,&DWP(1,$td,$tmp,8)); + if ($i<2) { &mov (&DWP(4+4*$i,"esp"),$out); } + if ($i==3) { &mov ($s[3],$__s0); } + &comment(); +} + +sub declast() +{ my ($i,$td,@s)=@_; + my $tmp = $key; + my $out = $i==3?$s[0]:$acc; + + if($i==0) { &lea ($td,&DWP(2048+128,$td)); + &mov ($tmp,&DWP(0-128,$td)); + &mov ($acc,&DWP(32-128,$td)); + &mov ($tmp,&DWP(64-128,$td)); + &mov ($acc,&DWP(96-128,$td)); + &mov ($tmp,&DWP(128-128,$td)); + &mov ($acc,&DWP(160-128,$td)); + &mov ($tmp,&DWP(192-128,$td)); + &mov ($acc,&DWP(224-128,$td)); + &lea ($td,&DWP(-128,$td)); } + if($i==3) { &mov ($key,$__key); } + else { &mov ($out,$s[0]); } + &and ($out,0xFF); + &movz ($out,&BP(0,$td,$out,1)); + + if ($i==3) { $tmp=$s[1]; } + &movz ($tmp,&HB($s[1])); + &movz ($tmp,&BP(0,$td,$tmp,1)); + &shl ($tmp,8); + &xor ($out,$tmp); + + if ($i==3) { $tmp=$s[2]; &mov ($s[1],$acc); } + else { mov ($tmp,$s[2]); } + &shr ($tmp,16); + &and ($tmp,0xFF); + &movz ($tmp,&BP(0,$td,$tmp,1)); + &shl ($tmp,16); + &xor ($out,$tmp); + + if ($i==3) { $tmp=$s[3]; &mov ($s[2],$__s1); } + else { &mov ($tmp,$s[3]); } + &shr ($tmp,24); + &movz ($tmp,&BP(0,$td,$tmp,1)); + &shl ($tmp,24); + &xor ($out,$tmp); + if ($i<2) { &mov (&DWP(4+4*$i,"esp"),$out); } + if ($i==3) { &mov ($s[3],$__s0); + &lea ($td,&DWP(-2048,$td)); } +} + +&function_begin_B("_x86_AES_decrypt"); + # note that caller is expected to allocate stack frame for me! + &mov ($__key,$key); # save key + + &xor ($s0,&DWP(0,$key)); # xor with key + &xor ($s1,&DWP(4,$key)); + &xor ($s2,&DWP(8,$key)); + &xor ($s3,&DWP(12,$key)); + + &mov ($acc,&DWP(240,$key)); # load key->rounds + + if ($small_footprint) { + &lea ($acc,&DWP(-2,$acc,$acc)); + &lea ($acc,&DWP(0,$key,$acc,8)); + &mov ($__end,$acc); # end of key schedule + &set_label("loop",16); + &decstep(0,$tbl,$s0,$s3,$s2,$s1); + &decstep(1,$tbl,$s1,$s0,$s3,$s2); + &decstep(2,$tbl,$s2,$s1,$s0,$s3); + &decstep(3,$tbl,$s3,$s2,$s1,$s0); + &add ($key,16); # advance rd_key + &xor ($s0,&DWP(0,$key)); + &xor ($s1,&DWP(4,$key)); + &xor ($s2,&DWP(8,$key)); + &xor ($s3,&DWP(12,$key)); + &cmp ($key,$__end); + &mov ($__key,$key); + &jb (&label("loop")); + } + else { + &cmp ($acc,10); + &jle (&label("10rounds")); + &cmp ($acc,12); + &jle (&label("12rounds")); + + &set_label("14rounds",4); + for ($i=1;$i<3;$i++) { + &decstep(0,$tbl,$s0,$s3,$s2,$s1); + &decstep(1,$tbl,$s1,$s0,$s3,$s2); + &decstep(2,$tbl,$s2,$s1,$s0,$s3); + &decstep(3,$tbl,$s3,$s2,$s1,$s0); + &xor ($s0,&DWP(16*$i+0,$key)); + &xor ($s1,&DWP(16*$i+4,$key)); + &xor ($s2,&DWP(16*$i+8,$key)); + &xor ($s3,&DWP(16*$i+12,$key)); + } + &add ($key,32); + &mov ($__key,$key); # advance rd_key + &set_label("12rounds",4); + for ($i=1;$i<3;$i++) { + &decstep(0,$tbl,$s0,$s3,$s2,$s1); + &decstep(1,$tbl,$s1,$s0,$s3,$s2); + &decstep(2,$tbl,$s2,$s1,$s0,$s3); + &decstep(3,$tbl,$s3,$s2,$s1,$s0); + &xor ($s0,&DWP(16*$i+0,$key)); + &xor ($s1,&DWP(16*$i+4,$key)); + &xor ($s2,&DWP(16*$i+8,$key)); + &xor ($s3,&DWP(16*$i+12,$key)); + } + &add ($key,32); + &mov ($__key,$key); # advance rd_key + &set_label("10rounds",4); + for ($i=1;$i<10;$i++) { + &decstep(0,$tbl,$s0,$s3,$s2,$s1); + &decstep(1,$tbl,$s1,$s0,$s3,$s2); + &decstep(2,$tbl,$s2,$s1,$s0,$s3); + &decstep(3,$tbl,$s3,$s2,$s1,$s0); + &xor ($s0,&DWP(16*$i+0,$key)); + &xor ($s1,&DWP(16*$i+4,$key)); + &xor ($s2,&DWP(16*$i+8,$key)); + &xor ($s3,&DWP(16*$i+12,$key)); + } + } + + &declast(0,$tbl,$s0,$s3,$s2,$s1); + &declast(1,$tbl,$s1,$s0,$s3,$s2); + &declast(2,$tbl,$s2,$s1,$s0,$s3); + &declast(3,$tbl,$s3,$s2,$s1,$s0); + + &add ($key,$small_footprint?16:160); + &xor ($s0,&DWP(0,$key)); + &xor ($s1,&DWP(4,$key)); + &xor ($s2,&DWP(8,$key)); + &xor ($s3,&DWP(12,$key)); + + &ret (); + +&set_label("AES_Td",64); # Yes! I keep it in the code segment! + &_data_word(0x50a7f451, 0x5365417e, 0xc3a4171a, 0x965e273a); + &_data_word(0xcb6bab3b, 0xf1459d1f, 0xab58faac, 0x9303e34b); + &_data_word(0x55fa3020, 0xf66d76ad, 0x9176cc88, 0x254c02f5); + &_data_word(0xfcd7e54f, 0xd7cb2ac5, 0x80443526, 0x8fa362b5); + &_data_word(0x495ab1de, 0x671bba25, 0x980eea45, 0xe1c0fe5d); + &_data_word(0x02752fc3, 0x12f04c81, 0xa397468d, 0xc6f9d36b); + &_data_word(0xe75f8f03, 0x959c9215, 0xeb7a6dbf, 0xda595295); + &_data_word(0x2d83bed4, 0xd3217458, 0x2969e049, 0x44c8c98e); + &_data_word(0x6a89c275, 0x78798ef4, 0x6b3e5899, 0xdd71b927); + &_data_word(0xb64fe1be, 0x17ad88f0, 0x66ac20c9, 0xb43ace7d); + &_data_word(0x184adf63, 0x82311ae5, 0x60335197, 0x457f5362); + &_data_word(0xe07764b1, 0x84ae6bbb, 0x1ca081fe, 0x942b08f9); + &_data_word(0x58684870, 0x19fd458f, 0x876cde94, 0xb7f87b52); + &_data_word(0x23d373ab, 0xe2024b72, 0x578f1fe3, 0x2aab5566); + &_data_word(0x0728ebb2, 0x03c2b52f, 0x9a7bc586, 0xa50837d3); + &_data_word(0xf2872830, 0xb2a5bf23, 0xba6a0302, 0x5c8216ed); + &_data_word(0x2b1ccf8a, 0x92b479a7, 0xf0f207f3, 0xa1e2694e); + &_data_word(0xcdf4da65, 0xd5be0506, 0x1f6234d1, 0x8afea6c4); + &_data_word(0x9d532e34, 0xa055f3a2, 0x32e18a05, 0x75ebf6a4); + &_data_word(0x39ec830b, 0xaaef6040, 0x069f715e, 0x51106ebd); + &_data_word(0xf98a213e, 0x3d06dd96, 0xae053edd, 0x46bde64d); + &_data_word(0xb58d5491, 0x055dc471, 0x6fd40604, 0xff155060); + &_data_word(0x24fb9819, 0x97e9bdd6, 0xcc434089, 0x779ed967); + &_data_word(0xbd42e8b0, 0x888b8907, 0x385b19e7, 0xdbeec879); + &_data_word(0x470a7ca1, 0xe90f427c, 0xc91e84f8, 0x00000000); + &_data_word(0x83868009, 0x48ed2b32, 0xac70111e, 0x4e725a6c); + &_data_word(0xfbff0efd, 0x5638850f, 0x1ed5ae3d, 0x27392d36); + &_data_word(0x64d90f0a, 0x21a65c68, 0xd1545b9b, 0x3a2e3624); + &_data_word(0xb1670a0c, 0x0fe75793, 0xd296eeb4, 0x9e919b1b); + &_data_word(0x4fc5c080, 0xa220dc61, 0x694b775a, 0x161a121c); + &_data_word(0x0aba93e2, 0xe52aa0c0, 0x43e0223c, 0x1d171b12); + &_data_word(0x0b0d090e, 0xadc78bf2, 0xb9a8b62d, 0xc8a91e14); + &_data_word(0x8519f157, 0x4c0775af, 0xbbdd99ee, 0xfd607fa3); + &_data_word(0x9f2601f7, 0xbcf5725c, 0xc53b6644, 0x347efb5b); + &_data_word(0x7629438b, 0xdcc623cb, 0x68fcedb6, 0x63f1e4b8); + &_data_word(0xcadc31d7, 0x10856342, 0x40229713, 0x2011c684); + &_data_word(0x7d244a85, 0xf83dbbd2, 0x1132f9ae, 0x6da129c7); + &_data_word(0x4b2f9e1d, 0xf330b2dc, 0xec52860d, 0xd0e3c177); + &_data_word(0x6c16b32b, 0x99b970a9, 0xfa489411, 0x2264e947); + &_data_word(0xc48cfca8, 0x1a3ff0a0, 0xd82c7d56, 0xef903322); + &_data_word(0xc74e4987, 0xc1d138d9, 0xfea2ca8c, 0x360bd498); + &_data_word(0xcf81f5a6, 0x28de7aa5, 0x268eb7da, 0xa4bfad3f); + &_data_word(0xe49d3a2c, 0x0d927850, 0x9bcc5f6a, 0x62467e54); + &_data_word(0xc2138df6, 0xe8b8d890, 0x5ef7392e, 0xf5afc382); + &_data_word(0xbe805d9f, 0x7c93d069, 0xa92dd56f, 0xb31225cf); + &_data_word(0x3b99acc8, 0xa77d1810, 0x6e639ce8, 0x7bbb3bdb); + &_data_word(0x097826cd, 0xf418596e, 0x01b79aec, 0xa89a4f83); + &_data_word(0x656e95e6, 0x7ee6ffaa, 0x08cfbc21, 0xe6e815ef); + &_data_word(0xd99be7ba, 0xce366f4a, 0xd4099fea, 0xd67cb029); + &_data_word(0xafb2a431, 0x31233f2a, 0x3094a5c6, 0xc066a235); + &_data_word(0x37bc4e74, 0xa6ca82fc, 0xb0d090e0, 0x15d8a733); + &_data_word(0x4a9804f1, 0xf7daec41, 0x0e50cd7f, 0x2ff69117); + &_data_word(0x8dd64d76, 0x4db0ef43, 0x544daacc, 0xdf0496e4); + &_data_word(0xe3b5d19e, 0x1b886a4c, 0xb81f2cc1, 0x7f516546); + &_data_word(0x04ea5e9d, 0x5d358c01, 0x737487fa, 0x2e410bfb); + &_data_word(0x5a1d67b3, 0x52d2db92, 0x335610e9, 0x1347d66d); + &_data_word(0x8c61d79a, 0x7a0ca137, 0x8e14f859, 0x893c13eb); + &_data_word(0xee27a9ce, 0x35c961b7, 0xede51ce1, 0x3cb1477a); + &_data_word(0x59dfd29c, 0x3f73f255, 0x79ce1418, 0xbf37c773); + &_data_word(0xeacdf753, 0x5baafd5f, 0x146f3ddf, 0x86db4478); + &_data_word(0x81f3afca, 0x3ec468b9, 0x2c342438, 0x5f40a3c2); + &_data_word(0x72c31d16, 0x0c25e2bc, 0x8b493c28, 0x41950dff); + &_data_word(0x7101a839, 0xdeb30c08, 0x9ce4b4d8, 0x90c15664); + &_data_word(0x6184cb7b, 0x70b632d5, 0x745c6c48, 0x4257b8d0); + +#Td4: # four copies of Td4 to choose from to avoid L1 aliasing + &data_byte(0x52, 0x09, 0x6a, 0xd5, 0x30, 0x36, 0xa5, 0x38); + &data_byte(0xbf, 0x40, 0xa3, 0x9e, 0x81, 0xf3, 0xd7, 0xfb); + &data_byte(0x7c, 0xe3, 0x39, 0x82, 0x9b, 0x2f, 0xff, 0x87); + &data_byte(0x34, 0x8e, 0x43, 0x44, 0xc4, 0xde, 0xe9, 0xcb); + &data_byte(0x54, 0x7b, 0x94, 0x32, 0xa6, 0xc2, 0x23, 0x3d); + &data_byte(0xee, 0x4c, 0x95, 0x0b, 0x42, 0xfa, 0xc3, 0x4e); + &data_byte(0x08, 0x2e, 0xa1, 0x66, 0x28, 0xd9, 0x24, 0xb2); + &data_byte(0x76, 0x5b, 0xa2, 0x49, 0x6d, 0x8b, 0xd1, 0x25); + &data_byte(0x72, 0xf8, 0xf6, 0x64, 0x86, 0x68, 0x98, 0x16); + &data_byte(0xd4, 0xa4, 0x5c, 0xcc, 0x5d, 0x65, 0xb6, 0x92); + &data_byte(0x6c, 0x70, 0x48, 0x50, 0xfd, 0xed, 0xb9, 0xda); + &data_byte(0x5e, 0x15, 0x46, 0x57, 0xa7, 0x8d, 0x9d, 0x84); + &data_byte(0x90, 0xd8, 0xab, 0x00, 0x8c, 0xbc, 0xd3, 0x0a); + &data_byte(0xf7, 0xe4, 0x58, 0x05, 0xb8, 0xb3, 0x45, 0x06); + &data_byte(0xd0, 0x2c, 0x1e, 0x8f, 0xca, 0x3f, 0x0f, 0x02); + &data_byte(0xc1, 0xaf, 0xbd, 0x03, 0x01, 0x13, 0x8a, 0x6b); + &data_byte(0x3a, 0x91, 0x11, 0x41, 0x4f, 0x67, 0xdc, 0xea); + &data_byte(0x97, 0xf2, 0xcf, 0xce, 0xf0, 0xb4, 0xe6, 0x73); + &data_byte(0x96, 0xac, 0x74, 0x22, 0xe7, 0xad, 0x35, 0x85); + &data_byte(0xe2, 0xf9, 0x37, 0xe8, 0x1c, 0x75, 0xdf, 0x6e); + &data_byte(0x47, 0xf1, 0x1a, 0x71, 0x1d, 0x29, 0xc5, 0x89); + &data_byte(0x6f, 0xb7, 0x62, 0x0e, 0xaa, 0x18, 0xbe, 0x1b); + &data_byte(0xfc, 0x56, 0x3e, 0x4b, 0xc6, 0xd2, 0x79, 0x20); + &data_byte(0x9a, 0xdb, 0xc0, 0xfe, 0x78, 0xcd, 0x5a, 0xf4); + &data_byte(0x1f, 0xdd, 0xa8, 0x33, 0x88, 0x07, 0xc7, 0x31); + &data_byte(0xb1, 0x12, 0x10, 0x59, 0x27, 0x80, 0xec, 0x5f); + &data_byte(0x60, 0x51, 0x7f, 0xa9, 0x19, 0xb5, 0x4a, 0x0d); + &data_byte(0x2d, 0xe5, 0x7a, 0x9f, 0x93, 0xc9, 0x9c, 0xef); + &data_byte(0xa0, 0xe0, 0x3b, 0x4d, 0xae, 0x2a, 0xf5, 0xb0); + &data_byte(0xc8, 0xeb, 0xbb, 0x3c, 0x83, 0x53, 0x99, 0x61); + &data_byte(0x17, 0x2b, 0x04, 0x7e, 0xba, 0x77, 0xd6, 0x26); + &data_byte(0xe1, 0x69, 0x14, 0x63, 0x55, 0x21, 0x0c, 0x7d); + + &data_byte(0x52, 0x09, 0x6a, 0xd5, 0x30, 0x36, 0xa5, 0x38); + &data_byte(0xbf, 0x40, 0xa3, 0x9e, 0x81, 0xf3, 0xd7, 0xfb); + &data_byte(0x7c, 0xe3, 0x39, 0x82, 0x9b, 0x2f, 0xff, 0x87); + &data_byte(0x34, 0x8e, 0x43, 0x44, 0xc4, 0xde, 0xe9, 0xcb); + &data_byte(0x54, 0x7b, 0x94, 0x32, 0xa6, 0xc2, 0x23, 0x3d); + &data_byte(0xee, 0x4c, 0x95, 0x0b, 0x42, 0xfa, 0xc3, 0x4e); + &data_byte(0x08, 0x2e, 0xa1, 0x66, 0x28, 0xd9, 0x24, 0xb2); + &data_byte(0x76, 0x5b, 0xa2, 0x49, 0x6d, 0x8b, 0xd1, 0x25); + &data_byte(0x72, 0xf8, 0xf6, 0x64, 0x86, 0x68, 0x98, 0x16); + &data_byte(0xd4, 0xa4, 0x5c, 0xcc, 0x5d, 0x65, 0xb6, 0x92); + &data_byte(0x6c, 0x70, 0x48, 0x50, 0xfd, 0xed, 0xb9, 0xda); + &data_byte(0x5e, 0x15, 0x46, 0x57, 0xa7, 0x8d, 0x9d, 0x84); + &data_byte(0x90, 0xd8, 0xab, 0x00, 0x8c, 0xbc, 0xd3, 0x0a); + &data_byte(0xf7, 0xe4, 0x58, 0x05, 0xb8, 0xb3, 0x45, 0x06); + &data_byte(0xd0, 0x2c, 0x1e, 0x8f, 0xca, 0x3f, 0x0f, 0x02); + &data_byte(0xc1, 0xaf, 0xbd, 0x03, 0x01, 0x13, 0x8a, 0x6b); + &data_byte(0x3a, 0x91, 0x11, 0x41, 0x4f, 0x67, 0xdc, 0xea); + &data_byte(0x97, 0xf2, 0xcf, 0xce, 0xf0, 0xb4, 0xe6, 0x73); + &data_byte(0x96, 0xac, 0x74, 0x22, 0xe7, 0xad, 0x35, 0x85); + &data_byte(0xe2, 0xf9, 0x37, 0xe8, 0x1c, 0x75, 0xdf, 0x6e); + &data_byte(0x47, 0xf1, 0x1a, 0x71, 0x1d, 0x29, 0xc5, 0x89); + &data_byte(0x6f, 0xb7, 0x62, 0x0e, 0xaa, 0x18, 0xbe, 0x1b); + &data_byte(0xfc, 0x56, 0x3e, 0x4b, 0xc6, 0xd2, 0x79, 0x20); + &data_byte(0x9a, 0xdb, 0xc0, 0xfe, 0x78, 0xcd, 0x5a, 0xf4); + &data_byte(0x1f, 0xdd, 0xa8, 0x33, 0x88, 0x07, 0xc7, 0x31); + &data_byte(0xb1, 0x12, 0x10, 0x59, 0x27, 0x80, 0xec, 0x5f); + &data_byte(0x60, 0x51, 0x7f, 0xa9, 0x19, 0xb5, 0x4a, 0x0d); + &data_byte(0x2d, 0xe5, 0x7a, 0x9f, 0x93, 0xc9, 0x9c, 0xef); + &data_byte(0xa0, 0xe0, 0x3b, 0x4d, 0xae, 0x2a, 0xf5, 0xb0); + &data_byte(0xc8, 0xeb, 0xbb, 0x3c, 0x83, 0x53, 0x99, 0x61); + &data_byte(0x17, 0x2b, 0x04, 0x7e, 0xba, 0x77, 0xd6, 0x26); + &data_byte(0xe1, 0x69, 0x14, 0x63, 0x55, 0x21, 0x0c, 0x7d); + + &data_byte(0x52, 0x09, 0x6a, 0xd5, 0x30, 0x36, 0xa5, 0x38); + &data_byte(0xbf, 0x40, 0xa3, 0x9e, 0x81, 0xf3, 0xd7, 0xfb); + &data_byte(0x7c, 0xe3, 0x39, 0x82, 0x9b, 0x2f, 0xff, 0x87); + &data_byte(0x34, 0x8e, 0x43, 0x44, 0xc4, 0xde, 0xe9, 0xcb); + &data_byte(0x54, 0x7b, 0x94, 0x32, 0xa6, 0xc2, 0x23, 0x3d); + &data_byte(0xee, 0x4c, 0x95, 0x0b, 0x42, 0xfa, 0xc3, 0x4e); + &data_byte(0x08, 0x2e, 0xa1, 0x66, 0x28, 0xd9, 0x24, 0xb2); + &data_byte(0x76, 0x5b, 0xa2, 0x49, 0x6d, 0x8b, 0xd1, 0x25); + &data_byte(0x72, 0xf8, 0xf6, 0x64, 0x86, 0x68, 0x98, 0x16); + &data_byte(0xd4, 0xa4, 0x5c, 0xcc, 0x5d, 0x65, 0xb6, 0x92); + &data_byte(0x6c, 0x70, 0x48, 0x50, 0xfd, 0xed, 0xb9, 0xda); + &data_byte(0x5e, 0x15, 0x46, 0x57, 0xa7, 0x8d, 0x9d, 0x84); + &data_byte(0x90, 0xd8, 0xab, 0x00, 0x8c, 0xbc, 0xd3, 0x0a); + &data_byte(0xf7, 0xe4, 0x58, 0x05, 0xb8, 0xb3, 0x45, 0x06); + &data_byte(0xd0, 0x2c, 0x1e, 0x8f, 0xca, 0x3f, 0x0f, 0x02); + &data_byte(0xc1, 0xaf, 0xbd, 0x03, 0x01, 0x13, 0x8a, 0x6b); + &data_byte(0x3a, 0x91, 0x11, 0x41, 0x4f, 0x67, 0xdc, 0xea); + &data_byte(0x97, 0xf2, 0xcf, 0xce, 0xf0, 0xb4, 0xe6, 0x73); + &data_byte(0x96, 0xac, 0x74, 0x22, 0xe7, 0xad, 0x35, 0x85); + &data_byte(0xe2, 0xf9, 0x37, 0xe8, 0x1c, 0x75, 0xdf, 0x6e); + &data_byte(0x47, 0xf1, 0x1a, 0x71, 0x1d, 0x29, 0xc5, 0x89); + &data_byte(0x6f, 0xb7, 0x62, 0x0e, 0xaa, 0x18, 0xbe, 0x1b); + &data_byte(0xfc, 0x56, 0x3e, 0x4b, 0xc6, 0xd2, 0x79, 0x20); + &data_byte(0x9a, 0xdb, 0xc0, 0xfe, 0x78, 0xcd, 0x5a, 0xf4); + &data_byte(0x1f, 0xdd, 0xa8, 0x33, 0x88, 0x07, 0xc7, 0x31); + &data_byte(0xb1, 0x12, 0x10, 0x59, 0x27, 0x80, 0xec, 0x5f); + &data_byte(0x60, 0x51, 0x7f, 0xa9, 0x19, 0xb5, 0x4a, 0x0d); + &data_byte(0x2d, 0xe5, 0x7a, 0x9f, 0x93, 0xc9, 0x9c, 0xef); + &data_byte(0xa0, 0xe0, 0x3b, 0x4d, 0xae, 0x2a, 0xf5, 0xb0); + &data_byte(0xc8, 0xeb, 0xbb, 0x3c, 0x83, 0x53, 0x99, 0x61); + &data_byte(0x17, 0x2b, 0x04, 0x7e, 0xba, 0x77, 0xd6, 0x26); + &data_byte(0xe1, 0x69, 0x14, 0x63, 0x55, 0x21, 0x0c, 0x7d); + + &data_byte(0x52, 0x09, 0x6a, 0xd5, 0x30, 0x36, 0xa5, 0x38); + &data_byte(0xbf, 0x40, 0xa3, 0x9e, 0x81, 0xf3, 0xd7, 0xfb); + &data_byte(0x7c, 0xe3, 0x39, 0x82, 0x9b, 0x2f, 0xff, 0x87); + &data_byte(0x34, 0x8e, 0x43, 0x44, 0xc4, 0xde, 0xe9, 0xcb); + &data_byte(0x54, 0x7b, 0x94, 0x32, 0xa6, 0xc2, 0x23, 0x3d); + &data_byte(0xee, 0x4c, 0x95, 0x0b, 0x42, 0xfa, 0xc3, 0x4e); + &data_byte(0x08, 0x2e, 0xa1, 0x66, 0x28, 0xd9, 0x24, 0xb2); + &data_byte(0x76, 0x5b, 0xa2, 0x49, 0x6d, 0x8b, 0xd1, 0x25); + &data_byte(0x72, 0xf8, 0xf6, 0x64, 0x86, 0x68, 0x98, 0x16); + &data_byte(0xd4, 0xa4, 0x5c, 0xcc, 0x5d, 0x65, 0xb6, 0x92); + &data_byte(0x6c, 0x70, 0x48, 0x50, 0xfd, 0xed, 0xb9, 0xda); + &data_byte(0x5e, 0x15, 0x46, 0x57, 0xa7, 0x8d, 0x9d, 0x84); + &data_byte(0x90, 0xd8, 0xab, 0x00, 0x8c, 0xbc, 0xd3, 0x0a); + &data_byte(0xf7, 0xe4, 0x58, 0x05, 0xb8, 0xb3, 0x45, 0x06); + &data_byte(0xd0, 0x2c, 0x1e, 0x8f, 0xca, 0x3f, 0x0f, 0x02); + &data_byte(0xc1, 0xaf, 0xbd, 0x03, 0x01, 0x13, 0x8a, 0x6b); + &data_byte(0x3a, 0x91, 0x11, 0x41, 0x4f, 0x67, 0xdc, 0xea); + &data_byte(0x97, 0xf2, 0xcf, 0xce, 0xf0, 0xb4, 0xe6, 0x73); + &data_byte(0x96, 0xac, 0x74, 0x22, 0xe7, 0xad, 0x35, 0x85); + &data_byte(0xe2, 0xf9, 0x37, 0xe8, 0x1c, 0x75, 0xdf, 0x6e); + &data_byte(0x47, 0xf1, 0x1a, 0x71, 0x1d, 0x29, 0xc5, 0x89); + &data_byte(0x6f, 0xb7, 0x62, 0x0e, 0xaa, 0x18, 0xbe, 0x1b); + &data_byte(0xfc, 0x56, 0x3e, 0x4b, 0xc6, 0xd2, 0x79, 0x20); + &data_byte(0x9a, 0xdb, 0xc0, 0xfe, 0x78, 0xcd, 0x5a, 0xf4); + &data_byte(0x1f, 0xdd, 0xa8, 0x33, 0x88, 0x07, 0xc7, 0x31); + &data_byte(0xb1, 0x12, 0x10, 0x59, 0x27, 0x80, 0xec, 0x5f); + &data_byte(0x60, 0x51, 0x7f, 0xa9, 0x19, 0xb5, 0x4a, 0x0d); + &data_byte(0x2d, 0xe5, 0x7a, 0x9f, 0x93, 0xc9, 0x9c, 0xef); + &data_byte(0xa0, 0xe0, 0x3b, 0x4d, 0xae, 0x2a, 0xf5, 0xb0); + &data_byte(0xc8, 0xeb, 0xbb, 0x3c, 0x83, 0x53, 0x99, 0x61); + &data_byte(0x17, 0x2b, 0x04, 0x7e, 0xba, 0x77, 0xd6, 0x26); + &data_byte(0xe1, 0x69, 0x14, 0x63, 0x55, 0x21, 0x0c, 0x7d); +&function_end_B("_x86_AES_decrypt"); + +# void AES_decrypt (const void *inp,void *out,const AES_KEY *key); +&function_begin("AES_decrypt"); + &mov ($acc,&wparam(0)); # load inp + &mov ($key,&wparam(2)); # load key + + &mov ($s0,"esp"); + &sub ("esp",36); + &and ("esp",-64); # align to cache-line + + # place stack frame just "above" the key schedule + &lea ($s1,&DWP(-64-63,$key)); + &sub ($s1,"esp"); + &neg ($s1); + &and ($s1,0x3C0); # modulo 1024, but aligned to cache-line + &sub ("esp",$s1); + &add ("esp",4); # 4 is reserved for caller's return address + &mov ($_esp,$s0); # save stack pointer + + &call (&label("pic_point")); # make it PIC! + &set_label("pic_point"); + &blindpop($tbl); + &picmeup($s0,"OPENSSL_ia32cap_P",$tbl,&label("pic_point")) if(!$x86only); + &lea ($tbl,&DWP(&label("AES_Td")."-".&label("pic_point"),$tbl)); + + # pick Td4 copy which can't "overlap" with stack frame or key schedule + &lea ($s1,&DWP(768-4,"esp")); + &sub ($s1,$tbl); + &and ($s1,0x300); + &lea ($tbl,&DWP(2048+128,$tbl,$s1)); + + if (!$x86only) { + &bt (&DWP(0,$s0),25); # check for SSE bit + &jnc (&label("x86")); + + &movq ("mm0",&QWP(0,$acc)); + &movq ("mm4",&QWP(8,$acc)); + &call ("_sse_AES_decrypt_compact"); + &mov ("esp",$_esp); # restore stack pointer + &mov ($acc,&wparam(1)); # load out + &movq (&QWP(0,$acc),"mm0"); # write output data + &movq (&QWP(8,$acc),"mm4"); + &emms (); + &function_end_A(); + } + &set_label("x86",16); + &mov ($_tbl,$tbl); + &mov ($s0,&DWP(0,$acc)); # load input data + &mov ($s1,&DWP(4,$acc)); + &mov ($s2,&DWP(8,$acc)); + &mov ($s3,&DWP(12,$acc)); + &call ("_x86_AES_decrypt_compact"); + &mov ("esp",$_esp); # restore stack pointer + &mov ($acc,&wparam(1)); # load out + &mov (&DWP(0,$acc),$s0); # write output data + &mov (&DWP(4,$acc),$s1); + &mov (&DWP(8,$acc),$s2); + &mov (&DWP(12,$acc),$s3); +&function_end("AES_decrypt"); + +# void AES_cbc_encrypt (const void char *inp, unsigned char *out, +# size_t length, const AES_KEY *key, +# unsigned char *ivp,const int enc); +{ +# stack frame layout +# -4(%esp) # return address 0(%esp) +# 0(%esp) # s0 backing store 4(%esp) +# 4(%esp) # s1 backing store 8(%esp) +# 8(%esp) # s2 backing store 12(%esp) +# 12(%esp) # s3 backing store 16(%esp) +# 16(%esp) # key backup 20(%esp) +# 20(%esp) # end of key schedule 24(%esp) +# 24(%esp) # %ebp backup 28(%esp) +# 28(%esp) # %esp backup +my $_inp=&DWP(32,"esp"); # copy of wparam(0) +my $_out=&DWP(36,"esp"); # copy of wparam(1) +my $_len=&DWP(40,"esp"); # copy of wparam(2) +my $_key=&DWP(44,"esp"); # copy of wparam(3) +my $_ivp=&DWP(48,"esp"); # copy of wparam(4) +my $_tmp=&DWP(52,"esp"); # volatile variable +# +my $ivec=&DWP(60,"esp"); # ivec[16] +my $aes_key=&DWP(76,"esp"); # copy of aes_key +my $mark=&DWP(76+240,"esp"); # copy of aes_key->rounds + +&function_begin("AES_cbc_encrypt"); + &mov ($s2 eq "ecx"? $s2 : "",&wparam(2)); # load len + &cmp ($s2,0); + &je (&label("drop_out")); + + &call (&label("pic_point")); # make it PIC! + &set_label("pic_point"); + &blindpop($tbl); + &picmeup($s0,"OPENSSL_ia32cap_P",$tbl,&label("pic_point")) if(!$x86only); + + &cmp (&wparam(5),0); + &lea ($tbl,&DWP(&label("AES_Te")."-".&label("pic_point"),$tbl)); + &jne (&label("picked_te")); + &lea ($tbl,&DWP(&label("AES_Td")."-".&label("AES_Te"),$tbl)); + &set_label("picked_te"); + + # one can argue if this is required + &pushf (); + &cld (); + + &cmp ($s2,$speed_limit); + &jb (&label("slow_way")); + &test ($s2,15); + &jnz (&label("slow_way")); + if (!$x86only) { + &bt (&DWP(0,$s0),28); # check for hyper-threading bit + &jc (&label("slow_way")); + } + # pre-allocate aligned stack frame... + &lea ($acc,&DWP(-80-244,"esp")); + &and ($acc,-64); + + # ... and make sure it doesn't alias with $tbl modulo 4096 + &mov ($s0,$tbl); + &lea ($s1,&DWP(2048+256,$tbl)); + &mov ($s3,$acc); + &and ($s0,0xfff); # s = %ebp&0xfff + &and ($s1,0xfff); # e = (%ebp+2048+256)&0xfff + &and ($s3,0xfff); # p = %esp&0xfff + + &cmp ($s3,$s1); # if (p>=e) %esp =- (p-e); + &jb (&label("tbl_break_out")); + &sub ($s3,$s1); + &sub ($acc,$s3); + &jmp (&label("tbl_ok")); + &set_label("tbl_break_out",4); # else %esp -= (p-s)&0xfff + framesz; + &sub ($s3,$s0); + &and ($s3,0xfff); + &add ($s3,384); + &sub ($acc,$s3); + &set_label("tbl_ok",4); + + &lea ($s3,&wparam(0)); # obtain pointer to parameter block + &exch ("esp",$acc); # allocate stack frame + &add ("esp",4); # reserve for return address! + &mov ($_tbl,$tbl); # save %ebp + &mov ($_esp,$acc); # save %esp + + &mov ($s0,&DWP(0,$s3)); # load inp + &mov ($s1,&DWP(4,$s3)); # load out + #&mov ($s2,&DWP(8,$s3)); # load len + &mov ($key,&DWP(12,$s3)); # load key + &mov ($acc,&DWP(16,$s3)); # load ivp + &mov ($s3,&DWP(20,$s3)); # load enc flag + + &mov ($_inp,$s0); # save copy of inp + &mov ($_out,$s1); # save copy of out + &mov ($_len,$s2); # save copy of len + &mov ($_key,$key); # save copy of key + &mov ($_ivp,$acc); # save copy of ivp + + &mov ($mark,0); # copy of aes_key->rounds = 0; + # do we copy key schedule to stack? + &mov ($s1 eq "ebx" ? $s1 : "",$key); + &mov ($s2 eq "ecx" ? $s2 : "",244/4); + &sub ($s1,$tbl); + &mov ("esi",$key); + &and ($s1,0xfff); + &lea ("edi",$aes_key); + &cmp ($s1,2048+256); + &jb (&label("do_copy")); + &cmp ($s1,4096-244); + &jb (&label("skip_copy")); + &set_label("do_copy",4); + &mov ($_key,"edi"); + &data_word(0xA5F3F689); # rep movsd + &set_label("skip_copy"); + + &mov ($key,16); + &set_label("prefetch_tbl",4); + &mov ($s0,&DWP(0,$tbl)); + &mov ($s1,&DWP(32,$tbl)); + &mov ($s2,&DWP(64,$tbl)); + &mov ($acc,&DWP(96,$tbl)); + &lea ($tbl,&DWP(128,$tbl)); + &sub ($key,1); + &jnz (&label("prefetch_tbl")); + &sub ($tbl,2048); + + &mov ($acc,$_inp); + &mov ($key,$_ivp); + + &cmp ($s3,0); + &je (&label("fast_decrypt")); + +#----------------------------- ENCRYPT -----------------------------# + &mov ($s0,&DWP(0,$key)); # load iv + &mov ($s1,&DWP(4,$key)); + + &set_label("fast_enc_loop",16); + &mov ($s2,&DWP(8,$key)); + &mov ($s3,&DWP(12,$key)); + + &xor ($s0,&DWP(0,$acc)); # xor input data + &xor ($s1,&DWP(4,$acc)); + &xor ($s2,&DWP(8,$acc)); + &xor ($s3,&DWP(12,$acc)); + + &mov ($key,$_key); # load key + &call ("_x86_AES_encrypt"); + + &mov ($acc,$_inp); # load inp + &mov ($key,$_out); # load out + + &mov (&DWP(0,$key),$s0); # save output data + &mov (&DWP(4,$key),$s1); + &mov (&DWP(8,$key),$s2); + &mov (&DWP(12,$key),$s3); + + &lea ($acc,&DWP(16,$acc)); # advance inp + &mov ($s2,$_len); # load len + &mov ($_inp,$acc); # save inp + &lea ($s3,&DWP(16,$key)); # advance out + &mov ($_out,$s3); # save out + &sub ($s2,16); # decrease len + &mov ($_len,$s2); # save len + &jnz (&label("fast_enc_loop")); + &mov ($acc,$_ivp); # load ivp + &mov ($s2,&DWP(8,$key)); # restore last 2 dwords + &mov ($s3,&DWP(12,$key)); + &mov (&DWP(0,$acc),$s0); # save ivec + &mov (&DWP(4,$acc),$s1); + &mov (&DWP(8,$acc),$s2); + &mov (&DWP(12,$acc),$s3); + + &cmp ($mark,0); # was the key schedule copied? + &mov ("edi",$_key); + &je (&label("skip_ezero")); + # zero copy of key schedule + &mov ("ecx",240/4); + &xor ("eax","eax"); + &align (4); + &data_word(0xABF3F689); # rep stosd + &set_label("skip_ezero"); + &mov ("esp",$_esp); + &popf (); + &set_label("drop_out"); + &function_end_A(); + &pushf (); # kludge, never executed + +#----------------------------- DECRYPT -----------------------------# +&set_label("fast_decrypt",16); + + &cmp ($acc,$_out); + &je (&label("fast_dec_in_place")); # in-place processing... + + &mov ($_tmp,$key); + + &align (4); + &set_label("fast_dec_loop",16); + &mov ($s0,&DWP(0,$acc)); # read input + &mov ($s1,&DWP(4,$acc)); + &mov ($s2,&DWP(8,$acc)); + &mov ($s3,&DWP(12,$acc)); + + &mov ($key,$_key); # load key + &call ("_x86_AES_decrypt"); + + &mov ($key,$_tmp); # load ivp + &mov ($acc,$_len); # load len + &xor ($s0,&DWP(0,$key)); # xor iv + &xor ($s1,&DWP(4,$key)); + &xor ($s2,&DWP(8,$key)); + &xor ($s3,&DWP(12,$key)); + + &mov ($key,$_out); # load out + &mov ($acc,$_inp); # load inp + + &mov (&DWP(0,$key),$s0); # write output + &mov (&DWP(4,$key),$s1); + &mov (&DWP(8,$key),$s2); + &mov (&DWP(12,$key),$s3); + + &mov ($s2,$_len); # load len + &mov ($_tmp,$acc); # save ivp + &lea ($acc,&DWP(16,$acc)); # advance inp + &mov ($_inp,$acc); # save inp + &lea ($key,&DWP(16,$key)); # advance out + &mov ($_out,$key); # save out + &sub ($s2,16); # decrease len + &mov ($_len,$s2); # save len + &jnz (&label("fast_dec_loop")); + &mov ($key,$_tmp); # load temp ivp + &mov ($acc,$_ivp); # load user ivp + &mov ($s0,&DWP(0,$key)); # load iv + &mov ($s1,&DWP(4,$key)); + &mov ($s2,&DWP(8,$key)); + &mov ($s3,&DWP(12,$key)); + &mov (&DWP(0,$acc),$s0); # copy back to user + &mov (&DWP(4,$acc),$s1); + &mov (&DWP(8,$acc),$s2); + &mov (&DWP(12,$acc),$s3); + &jmp (&label("fast_dec_out")); + + &set_label("fast_dec_in_place",16); + &set_label("fast_dec_in_place_loop"); + &mov ($s0,&DWP(0,$acc)); # read input + &mov ($s1,&DWP(4,$acc)); + &mov ($s2,&DWP(8,$acc)); + &mov ($s3,&DWP(12,$acc)); + + &lea ($key,$ivec); + &mov (&DWP(0,$key),$s0); # copy to temp + &mov (&DWP(4,$key),$s1); + &mov (&DWP(8,$key),$s2); + &mov (&DWP(12,$key),$s3); + + &mov ($key,$_key); # load key + &call ("_x86_AES_decrypt"); + + &mov ($key,$_ivp); # load ivp + &mov ($acc,$_out); # load out + &xor ($s0,&DWP(0,$key)); # xor iv + &xor ($s1,&DWP(4,$key)); + &xor ($s2,&DWP(8,$key)); + &xor ($s3,&DWP(12,$key)); + + &mov (&DWP(0,$acc),$s0); # write output + &mov (&DWP(4,$acc),$s1); + &mov (&DWP(8,$acc),$s2); + &mov (&DWP(12,$acc),$s3); + + &lea ($acc,&DWP(16,$acc)); # advance out + &mov ($_out,$acc); # save out + + &lea ($acc,$ivec); + &mov ($s0,&DWP(0,$acc)); # read temp + &mov ($s1,&DWP(4,$acc)); + &mov ($s2,&DWP(8,$acc)); + &mov ($s3,&DWP(12,$acc)); + + &mov (&DWP(0,$key),$s0); # copy iv + &mov (&DWP(4,$key),$s1); + &mov (&DWP(8,$key),$s2); + &mov (&DWP(12,$key),$s3); + + &mov ($acc,$_inp); # load inp + &mov ($s2,$_len); # load len + &lea ($acc,&DWP(16,$acc)); # advance inp + &mov ($_inp,$acc); # save inp + &sub ($s2,16); # decrease len + &mov ($_len,$s2); # save len + &jnz (&label("fast_dec_in_place_loop")); + + &set_label("fast_dec_out",4); + &cmp ($mark,0); # was the key schedule copied? + &mov ("edi",$_key); + &je (&label("skip_dzero")); + # zero copy of key schedule + &mov ("ecx",240/4); + &xor ("eax","eax"); + &align (4); + &data_word(0xABF3F689); # rep stosd + &set_label("skip_dzero"); + &mov ("esp",$_esp); + &popf (); + &function_end_A(); + &pushf (); # kludge, never executed + +#--------------------------- SLOW ROUTINE ---------------------------# +&set_label("slow_way",16); + + &mov ($s0,&DWP(0,$s0)) if (!$x86only);# load OPENSSL_ia32cap + &mov ($key,&wparam(3)); # load key + + # pre-allocate aligned stack frame... + &lea ($acc,&DWP(-80,"esp")); + &and ($acc,-64); + + # ... and make sure it doesn't alias with $key modulo 1024 + &lea ($s1,&DWP(-80-63,$key)); + &sub ($s1,$acc); + &neg ($s1); + &and ($s1,0x3C0); # modulo 1024, but aligned to cache-line + &sub ($acc,$s1); + + # pick S-box copy which can't overlap with stack frame or $key + &lea ($s1,&DWP(768,$acc)); + &sub ($s1,$tbl); + &and ($s1,0x300); + &lea ($tbl,&DWP(2048+128,$tbl,$s1)); + + &lea ($s3,&wparam(0)); # pointer to parameter block + + &exch ("esp",$acc); + &add ("esp",4); # reserve for return address! + &mov ($_tbl,$tbl); # save %ebp + &mov ($_esp,$acc); # save %esp + &mov ($_tmp,$s0); # save OPENSSL_ia32cap + + &mov ($s0,&DWP(0,$s3)); # load inp + &mov ($s1,&DWP(4,$s3)); # load out + #&mov ($s2,&DWP(8,$s3)); # load len + #&mov ($key,&DWP(12,$s3)); # load key + &mov ($acc,&DWP(16,$s3)); # load ivp + &mov ($s3,&DWP(20,$s3)); # load enc flag + + &mov ($_inp,$s0); # save copy of inp + &mov ($_out,$s1); # save copy of out + &mov ($_len,$s2); # save copy of len + &mov ($_key,$key); # save copy of key + &mov ($_ivp,$acc); # save copy of ivp + + &mov ($key,$acc); + &mov ($acc,$s0); + + &cmp ($s3,0); + &je (&label("slow_decrypt")); + +#--------------------------- SLOW ENCRYPT ---------------------------# + &cmp ($s2,16); + &mov ($s3,$s1); + &jb (&label("slow_enc_tail")); + + if (!$x86only) { + &bt ($_tmp,25); # check for SSE bit + &jnc (&label("slow_enc_x86")); + + &movq ("mm0",&QWP(0,$key)); # load iv + &movq ("mm4",&QWP(8,$key)); + + &set_label("slow_enc_loop_sse",16); + &pxor ("mm0",&QWP(0,$acc)); # xor input data + &pxor ("mm4",&QWP(8,$acc)); + + &mov ($key,$_key); + &call ("_sse_AES_encrypt_compact"); + + &mov ($acc,$_inp); # load inp + &mov ($key,$_out); # load out + &mov ($s2,$_len); # load len + + &movq (&QWP(0,$key),"mm0"); # save output data + &movq (&QWP(8,$key),"mm4"); + + &lea ($acc,&DWP(16,$acc)); # advance inp + &mov ($_inp,$acc); # save inp + &lea ($s3,&DWP(16,$key)); # advance out + &mov ($_out,$s3); # save out + &sub ($s2,16); # decrease len + &cmp ($s2,16); + &mov ($_len,$s2); # save len + &jae (&label("slow_enc_loop_sse")); + &test ($s2,15); + &jnz (&label("slow_enc_tail")); + &mov ($acc,$_ivp); # load ivp + &movq (&QWP(0,$acc),"mm0"); # save ivec + &movq (&QWP(8,$acc),"mm4"); + &emms (); + &mov ("esp",$_esp); + &popf (); + &function_end_A(); + &pushf (); # kludge, never executed + } + &set_label("slow_enc_x86",16); + &mov ($s0,&DWP(0,$key)); # load iv + &mov ($s1,&DWP(4,$key)); + + &set_label("slow_enc_loop_x86",4); + &mov ($s2,&DWP(8,$key)); + &mov ($s3,&DWP(12,$key)); + + &xor ($s0,&DWP(0,$acc)); # xor input data + &xor ($s1,&DWP(4,$acc)); + &xor ($s2,&DWP(8,$acc)); + &xor ($s3,&DWP(12,$acc)); + + &mov ($key,$_key); # load key + &call ("_x86_AES_encrypt_compact"); + + &mov ($acc,$_inp); # load inp + &mov ($key,$_out); # load out + + &mov (&DWP(0,$key),$s0); # save output data + &mov (&DWP(4,$key),$s1); + &mov (&DWP(8,$key),$s2); + &mov (&DWP(12,$key),$s3); + + &mov ($s2,$_len); # load len + &lea ($acc,&DWP(16,$acc)); # advance inp + &mov ($_inp,$acc); # save inp + &lea ($s3,&DWP(16,$key)); # advance out + &mov ($_out,$s3); # save out + &sub ($s2,16); # decrease len + &cmp ($s2,16); + &mov ($_len,$s2); # save len + &jae (&label("slow_enc_loop_x86")); + &test ($s2,15); + &jnz (&label("slow_enc_tail")); + &mov ($acc,$_ivp); # load ivp + &mov ($s2,&DWP(8,$key)); # restore last dwords + &mov ($s3,&DWP(12,$key)); + &mov (&DWP(0,$acc),$s0); # save ivec + &mov (&DWP(4,$acc),$s1); + &mov (&DWP(8,$acc),$s2); + &mov (&DWP(12,$acc),$s3); + + &mov ("esp",$_esp); + &popf (); + &function_end_A(); + &pushf (); # kludge, never executed + + &set_label("slow_enc_tail",16); + &emms () if (!$x86only); + &mov ($key eq "edi"? $key:"",$s3); # load out to edi + &mov ($s1,16); + &sub ($s1,$s2); + &cmp ($key,$acc eq "esi"? $acc:""); # compare with inp + &je (&label("enc_in_place")); + &align (4); + &data_word(0xA4F3F689); # rep movsb # copy input + &jmp (&label("enc_skip_in_place")); + &set_label("enc_in_place"); + &lea ($key,&DWP(0,$key,$s2)); + &set_label("enc_skip_in_place"); + &mov ($s2,$s1); + &xor ($s0,$s0); + &align (4); + &data_word(0xAAF3F689); # rep stosb # zero tail + + &mov ($key,$_ivp); # restore ivp + &mov ($acc,$s3); # output as input + &mov ($s0,&DWP(0,$key)); + &mov ($s1,&DWP(4,$key)); + &mov ($_len,16); # len=16 + &jmp (&label("slow_enc_loop_x86")); # one more spin... + +#--------------------------- SLOW DECRYPT ---------------------------# +&set_label("slow_decrypt",16); + if (!$x86only) { + &bt ($_tmp,25); # check for SSE bit + &jnc (&label("slow_dec_loop_x86")); + + &set_label("slow_dec_loop_sse",4); + &movq ("mm0",&QWP(0,$acc)); # read input + &movq ("mm4",&QWP(8,$acc)); + + &mov ($key,$_key); + &call ("_sse_AES_decrypt_compact"); + + &mov ($acc,$_inp); # load inp + &lea ($s0,$ivec); + &mov ($s1,$_out); # load out + &mov ($s2,$_len); # load len + &mov ($key,$_ivp); # load ivp + + &movq ("mm1",&QWP(0,$acc)); # re-read input + &movq ("mm5",&QWP(8,$acc)); + + &pxor ("mm0",&QWP(0,$key)); # xor iv + &pxor ("mm4",&QWP(8,$key)); + + &movq (&QWP(0,$key),"mm1"); # copy input to iv + &movq (&QWP(8,$key),"mm5"); + + &sub ($s2,16); # decrease len + &jc (&label("slow_dec_partial_sse")); + + &movq (&QWP(0,$s1),"mm0"); # write output + &movq (&QWP(8,$s1),"mm4"); + + &lea ($s1,&DWP(16,$s1)); # advance out + &mov ($_out,$s1); # save out + &lea ($acc,&DWP(16,$acc)); # advance inp + &mov ($_inp,$acc); # save inp + &mov ($_len,$s2); # save len + &jnz (&label("slow_dec_loop_sse")); + &emms (); + &mov ("esp",$_esp); + &popf (); + &function_end_A(); + &pushf (); # kludge, never executed + + &set_label("slow_dec_partial_sse",16); + &movq (&QWP(0,$s0),"mm0"); # save output to temp + &movq (&QWP(8,$s0),"mm4"); + &emms (); + + &add ($s2 eq "ecx" ? "ecx":"",16); + &mov ("edi",$s1); # out + &mov ("esi",$s0); # temp + &align (4); + &data_word(0xA4F3F689); # rep movsb # copy partial output + + &mov ("esp",$_esp); + &popf (); + &function_end_A(); + &pushf (); # kludge, never executed + } + &set_label("slow_dec_loop_x86",16); + &mov ($s0,&DWP(0,$acc)); # read input + &mov ($s1,&DWP(4,$acc)); + &mov ($s2,&DWP(8,$acc)); + &mov ($s3,&DWP(12,$acc)); + + &lea ($key,$ivec); + &mov (&DWP(0,$key),$s0); # copy to temp + &mov (&DWP(4,$key),$s1); + &mov (&DWP(8,$key),$s2); + &mov (&DWP(12,$key),$s3); + + &mov ($key,$_key); # load key + &call ("_x86_AES_decrypt_compact"); + + &mov ($key,$_ivp); # load ivp + &mov ($acc,$_len); # load len + &xor ($s0,&DWP(0,$key)); # xor iv + &xor ($s1,&DWP(4,$key)); + &xor ($s2,&DWP(8,$key)); + &xor ($s3,&DWP(12,$key)); + + &sub ($acc,16); + &jc (&label("slow_dec_partial_x86")); + + &mov ($_len,$acc); # save len + &mov ($acc,$_out); # load out + + &mov (&DWP(0,$acc),$s0); # write output + &mov (&DWP(4,$acc),$s1); + &mov (&DWP(8,$acc),$s2); + &mov (&DWP(12,$acc),$s3); + + &lea ($acc,&DWP(16,$acc)); # advance out + &mov ($_out,$acc); # save out + + &lea ($acc,$ivec); + &mov ($s0,&DWP(0,$acc)); # read temp + &mov ($s1,&DWP(4,$acc)); + &mov ($s2,&DWP(8,$acc)); + &mov ($s3,&DWP(12,$acc)); + + &mov (&DWP(0,$key),$s0); # copy it to iv + &mov (&DWP(4,$key),$s1); + &mov (&DWP(8,$key),$s2); + &mov (&DWP(12,$key),$s3); + + &mov ($acc,$_inp); # load inp + &lea ($acc,&DWP(16,$acc)); # advance inp + &mov ($_inp,$acc); # save inp + &jnz (&label("slow_dec_loop_x86")); + &mov ("esp",$_esp); + &popf (); + &function_end_A(); + &pushf (); # kludge, never executed + + &set_label("slow_dec_partial_x86",16); + &lea ($acc,$ivec); + &mov (&DWP(0,$acc),$s0); # save output to temp + &mov (&DWP(4,$acc),$s1); + &mov (&DWP(8,$acc),$s2); + &mov (&DWP(12,$acc),$s3); + + &mov ($acc,$_inp); + &mov ($s0,&DWP(0,$acc)); # re-read input + &mov ($s1,&DWP(4,$acc)); + &mov ($s2,&DWP(8,$acc)); + &mov ($s3,&DWP(12,$acc)); + + &mov (&DWP(0,$key),$s0); # copy it to iv + &mov (&DWP(4,$key),$s1); + &mov (&DWP(8,$key),$s2); + &mov (&DWP(12,$key),$s3); + + &mov ("ecx",$_len); + &mov ("edi",$_out); + &lea ("esi",$ivec); + &align (4); + &data_word(0xA4F3F689); # rep movsb # copy partial output + + &mov ("esp",$_esp); + &popf (); +&function_end("AES_cbc_encrypt"); +} + +#------------------------------------------------------------------# + +sub enckey() +{ + &movz ("esi",&LB("edx")); # rk[i]>>0 + &movz ("ebx",&BP(-128,$tbl,"esi",1)); + &movz ("esi",&HB("edx")); # rk[i]>>8 + &shl ("ebx",24); + &xor ("eax","ebx"); + + &movz ("ebx",&BP(-128,$tbl,"esi",1)); + &shr ("edx",16); + &movz ("esi",&LB("edx")); # rk[i]>>16 + &xor ("eax","ebx"); + + &movz ("ebx",&BP(-128,$tbl,"esi",1)); + &movz ("esi",&HB("edx")); # rk[i]>>24 + &shl ("ebx",8); + &xor ("eax","ebx"); + + &movz ("ebx",&BP(-128,$tbl,"esi",1)); + &shl ("ebx",16); + &xor ("eax","ebx"); + + &xor ("eax",&DWP(1024-128,$tbl,"ecx",4)); # rcon +} + +&function_begin("_x86_AES_set_encrypt_key"); + &mov ("esi",&wparam(1)); # user supplied key + &mov ("edi",&wparam(3)); # private key schedule + + &test ("esi",-1); + &jz (&label("badpointer")); + &test ("edi",-1); + &jz (&label("badpointer")); + + &call (&label("pic_point")); + &set_label("pic_point"); + &blindpop($tbl); + &lea ($tbl,&DWP(&label("AES_Te")."-".&label("pic_point"),$tbl)); + &lea ($tbl,&DWP(2048+128,$tbl)); + + # prefetch Te4 + &mov ("eax",&DWP(0-128,$tbl)); + &mov ("ebx",&DWP(32-128,$tbl)); + &mov ("ecx",&DWP(64-128,$tbl)); + &mov ("edx",&DWP(96-128,$tbl)); + &mov ("eax",&DWP(128-128,$tbl)); + &mov ("ebx",&DWP(160-128,$tbl)); + &mov ("ecx",&DWP(192-128,$tbl)); + &mov ("edx",&DWP(224-128,$tbl)); + + &mov ("ecx",&wparam(2)); # number of bits in key + &cmp ("ecx",128); + &je (&label("10rounds")); + &cmp ("ecx",192); + &je (&label("12rounds")); + &cmp ("ecx",256); + &je (&label("14rounds")); + &mov ("eax",-2); # invalid number of bits + &jmp (&label("exit")); + + &set_label("10rounds"); + &mov ("eax",&DWP(0,"esi")); # copy first 4 dwords + &mov ("ebx",&DWP(4,"esi")); + &mov ("ecx",&DWP(8,"esi")); + &mov ("edx",&DWP(12,"esi")); + &mov (&DWP(0,"edi"),"eax"); + &mov (&DWP(4,"edi"),"ebx"); + &mov (&DWP(8,"edi"),"ecx"); + &mov (&DWP(12,"edi"),"edx"); + + &xor ("ecx","ecx"); + &jmp (&label("10shortcut")); + + &align (4); + &set_label("10loop"); + &mov ("eax",&DWP(0,"edi")); # rk[0] + &mov ("edx",&DWP(12,"edi")); # rk[3] + &set_label("10shortcut"); + &enckey (); + + &mov (&DWP(16,"edi"),"eax"); # rk[4] + &xor ("eax",&DWP(4,"edi")); + &mov (&DWP(20,"edi"),"eax"); # rk[5] + &xor ("eax",&DWP(8,"edi")); + &mov (&DWP(24,"edi"),"eax"); # rk[6] + &xor ("eax",&DWP(12,"edi")); + &mov (&DWP(28,"edi"),"eax"); # rk[7] + &inc ("ecx"); + &add ("edi",16); + &cmp ("ecx",10); + &jl (&label("10loop")); + + &mov (&DWP(80,"edi"),10); # setup number of rounds + &xor ("eax","eax"); + &jmp (&label("exit")); + + &set_label("12rounds"); + &mov ("eax",&DWP(0,"esi")); # copy first 6 dwords + &mov ("ebx",&DWP(4,"esi")); + &mov ("ecx",&DWP(8,"esi")); + &mov ("edx",&DWP(12,"esi")); + &mov (&DWP(0,"edi"),"eax"); + &mov (&DWP(4,"edi"),"ebx"); + &mov (&DWP(8,"edi"),"ecx"); + &mov (&DWP(12,"edi"),"edx"); + &mov ("ecx",&DWP(16,"esi")); + &mov ("edx",&DWP(20,"esi")); + &mov (&DWP(16,"edi"),"ecx"); + &mov (&DWP(20,"edi"),"edx"); + + &xor ("ecx","ecx"); + &jmp (&label("12shortcut")); + + &align (4); + &set_label("12loop"); + &mov ("eax",&DWP(0,"edi")); # rk[0] + &mov ("edx",&DWP(20,"edi")); # rk[5] + &set_label("12shortcut"); + &enckey (); + + &mov (&DWP(24,"edi"),"eax"); # rk[6] + &xor ("eax",&DWP(4,"edi")); + &mov (&DWP(28,"edi"),"eax"); # rk[7] + &xor ("eax",&DWP(8,"edi")); + &mov (&DWP(32,"edi"),"eax"); # rk[8] + &xor ("eax",&DWP(12,"edi")); + &mov (&DWP(36,"edi"),"eax"); # rk[9] + + &cmp ("ecx",7); + &je (&label("12break")); + &inc ("ecx"); + + &xor ("eax",&DWP(16,"edi")); + &mov (&DWP(40,"edi"),"eax"); # rk[10] + &xor ("eax",&DWP(20,"edi")); + &mov (&DWP(44,"edi"),"eax"); # rk[11] + + &add ("edi",24); + &jmp (&label("12loop")); + + &set_label("12break"); + &mov (&DWP(72,"edi"),12); # setup number of rounds + &xor ("eax","eax"); + &jmp (&label("exit")); + + &set_label("14rounds"); + &mov ("eax",&DWP(0,"esi")); # copy first 8 dwords + &mov ("ebx",&DWP(4,"esi")); + &mov ("ecx",&DWP(8,"esi")); + &mov ("edx",&DWP(12,"esi")); + &mov (&DWP(0,"edi"),"eax"); + &mov (&DWP(4,"edi"),"ebx"); + &mov (&DWP(8,"edi"),"ecx"); + &mov (&DWP(12,"edi"),"edx"); + &mov ("eax",&DWP(16,"esi")); + &mov ("ebx",&DWP(20,"esi")); + &mov ("ecx",&DWP(24,"esi")); + &mov ("edx",&DWP(28,"esi")); + &mov (&DWP(16,"edi"),"eax"); + &mov (&DWP(20,"edi"),"ebx"); + &mov (&DWP(24,"edi"),"ecx"); + &mov (&DWP(28,"edi"),"edx"); + + &xor ("ecx","ecx"); + &jmp (&label("14shortcut")); + + &align (4); + &set_label("14loop"); + &mov ("edx",&DWP(28,"edi")); # rk[7] + &set_label("14shortcut"); + &mov ("eax",&DWP(0,"edi")); # rk[0] + + &enckey (); + + &mov (&DWP(32,"edi"),"eax"); # rk[8] + &xor ("eax",&DWP(4,"edi")); + &mov (&DWP(36,"edi"),"eax"); # rk[9] + &xor ("eax",&DWP(8,"edi")); + &mov (&DWP(40,"edi"),"eax"); # rk[10] + &xor ("eax",&DWP(12,"edi")); + &mov (&DWP(44,"edi"),"eax"); # rk[11] + + &cmp ("ecx",6); + &je (&label("14break")); + &inc ("ecx"); + + &mov ("edx","eax"); + &mov ("eax",&DWP(16,"edi")); # rk[4] + &movz ("esi",&LB("edx")); # rk[11]>>0 + &movz ("ebx",&BP(-128,$tbl,"esi",1)); + &movz ("esi",&HB("edx")); # rk[11]>>8 + &xor ("eax","ebx"); + + &movz ("ebx",&BP(-128,$tbl,"esi",1)); + &shr ("edx",16); + &shl ("ebx",8); + &movz ("esi",&LB("edx")); # rk[11]>>16 + &xor ("eax","ebx"); + + &movz ("ebx",&BP(-128,$tbl,"esi",1)); + &movz ("esi",&HB("edx")); # rk[11]>>24 + &shl ("ebx",16); + &xor ("eax","ebx"); + + &movz ("ebx",&BP(-128,$tbl,"esi",1)); + &shl ("ebx",24); + &xor ("eax","ebx"); + + &mov (&DWP(48,"edi"),"eax"); # rk[12] + &xor ("eax",&DWP(20,"edi")); + &mov (&DWP(52,"edi"),"eax"); # rk[13] + &xor ("eax",&DWP(24,"edi")); + &mov (&DWP(56,"edi"),"eax"); # rk[14] + &xor ("eax",&DWP(28,"edi")); + &mov (&DWP(60,"edi"),"eax"); # rk[15] + + &add ("edi",32); + &jmp (&label("14loop")); + + &set_label("14break"); + &mov (&DWP(48,"edi"),14); # setup number of rounds + &xor ("eax","eax"); + &jmp (&label("exit")); + + &set_label("badpointer"); + &mov ("eax",-1); + &set_label("exit"); +&function_end("_x86_AES_set_encrypt_key"); + +# int AES_set_encrypt_key(const unsigned char *userKey, const int bits, +# AES_KEY *key) +&function_begin_B("AES_set_encrypt_key"); + &call ("_x86_AES_set_encrypt_key"); + &ret (); +&function_end_B("AES_set_encrypt_key"); + +sub deckey() +{ my ($i,$key,$tp1,$tp2,$tp4,$tp8) = @_; + my $tmp = $tbl; + + &mov ($tmp,0x80808080); + &and ($tmp,$tp1); + &lea ($tp2,&DWP(0,$tp1,$tp1)); + &mov ($acc,$tmp); + &shr ($tmp,7); + &sub ($acc,$tmp); + &and ($tp2,0xfefefefe); + &and ($acc,0x1b1b1b1b); + &xor ($tp2,$acc); + &mov ($tmp,0x80808080); + + &and ($tmp,$tp2); + &lea ($tp4,&DWP(0,$tp2,$tp2)); + &mov ($acc,$tmp); + &shr ($tmp,7); + &sub ($acc,$tmp); + &and ($tp4,0xfefefefe); + &and ($acc,0x1b1b1b1b); + &xor ($tp2,$tp1); # tp2^tp1 + &xor ($tp4,$acc); + &mov ($tmp,0x80808080); + + &and ($tmp,$tp4); + &lea ($tp8,&DWP(0,$tp4,$tp4)); + &mov ($acc,$tmp); + &shr ($tmp,7); + &xor ($tp4,$tp1); # tp4^tp1 + &sub ($acc,$tmp); + &and ($tp8,0xfefefefe); + &and ($acc,0x1b1b1b1b); + &rotl ($tp1,8); # = ROTATE(tp1,8) + &xor ($tp8,$acc); + + &mov ($tmp,&DWP(4*($i+1),$key)); # modulo-scheduled load + + &xor ($tp1,$tp2); + &xor ($tp2,$tp8); + &xor ($tp1,$tp4); + &rotl ($tp2,24); + &xor ($tp4,$tp8); + &xor ($tp1,$tp8); # ^= tp8^(tp4^tp1)^(tp2^tp1) + &rotl ($tp4,16); + &xor ($tp1,$tp2); # ^= ROTATE(tp8^tp2^tp1,24) + &rotl ($tp8,8); + &xor ($tp1,$tp4); # ^= ROTATE(tp8^tp4^tp1,16) + &mov ($tp2,$tmp); + &xor ($tp1,$tp8); # ^= ROTATE(tp8,8) + + &mov (&DWP(4*$i,$key),$tp1); +} + +# int AES_set_decrypt_key(const unsigned char *userKey, const int bits, +# AES_KEY *key) +&function_begin_B("AES_set_decrypt_key"); + &call ("_x86_AES_set_encrypt_key"); + &cmp ("eax",0); + &je (&label("proceed")); + &ret (); + + &set_label("proceed"); + &push ("ebp"); + &push ("ebx"); + &push ("esi"); + &push ("edi"); + + &mov ("esi",&wparam(2)); + &mov ("ecx",&DWP(240,"esi")); # pull number of rounds + &lea ("ecx",&DWP(0,"","ecx",4)); + &lea ("edi",&DWP(0,"esi","ecx",4)); # pointer to last chunk + + &set_label("invert",4); # invert order of chunks + &mov ("eax",&DWP(0,"esi")); + &mov ("ebx",&DWP(4,"esi")); + &mov ("ecx",&DWP(0,"edi")); + &mov ("edx",&DWP(4,"edi")); + &mov (&DWP(0,"edi"),"eax"); + &mov (&DWP(4,"edi"),"ebx"); + &mov (&DWP(0,"esi"),"ecx"); + &mov (&DWP(4,"esi"),"edx"); + &mov ("eax",&DWP(8,"esi")); + &mov ("ebx",&DWP(12,"esi")); + &mov ("ecx",&DWP(8,"edi")); + &mov ("edx",&DWP(12,"edi")); + &mov (&DWP(8,"edi"),"eax"); + &mov (&DWP(12,"edi"),"ebx"); + &mov (&DWP(8,"esi"),"ecx"); + &mov (&DWP(12,"esi"),"edx"); + &add ("esi",16); + &sub ("edi",16); + &cmp ("esi","edi"); + &jne (&label("invert")); + + &mov ($key,&wparam(2)); + &mov ($acc,&DWP(240,$key)); # pull number of rounds + &lea ($acc,&DWP(-2,$acc,$acc)); + &lea ($acc,&DWP(0,$key,$acc,8)); + &mov (&wparam(2),$acc); + + &mov ($s0,&DWP(16,$key)); # modulo-scheduled load + &set_label("permute",4); # permute the key schedule + &add ($key,16); + &deckey (0,$key,$s0,$s1,$s2,$s3); + &deckey (1,$key,$s1,$s2,$s3,$s0); + &deckey (2,$key,$s2,$s3,$s0,$s1); + &deckey (3,$key,$s3,$s0,$s1,$s2); + &cmp ($key,&wparam(2)); + &jb (&label("permute")); + + &xor ("eax","eax"); # return success +&function_end("AES_set_decrypt_key"); +&asciz("AES for x86, CRYPTOGAMS by "); + +&asm_finish(); + +close STDOUT; diff --git a/openssl-1.1.0h/crypto/aes/asm/aes-armv4.pl b/openssl-1.1.0h/crypto/aes/asm/aes-armv4.pl new file mode 100644 index 0000000..9981589 --- /dev/null +++ b/openssl-1.1.0h/crypto/aes/asm/aes-armv4.pl @@ -0,0 +1,1245 @@ +#! /usr/bin/env perl +# Copyright 2007-2018 The OpenSSL Project Authors. All Rights Reserved. +# +# Licensed under the OpenSSL license (the "License"). You may not use +# this file except in compliance with the License. You can obtain a copy +# in the file LICENSE in the source distribution or at +# https://www.openssl.org/source/license.html + + +# ==================================================================== +# Written by Andy Polyakov for the OpenSSL +# project. The module is, however, dual licensed under OpenSSL and +# CRYPTOGAMS licenses depending on where you obtain it. For further +# details see http://www.openssl.org/~appro/cryptogams/. +# ==================================================================== + +# AES for ARMv4 + +# January 2007. +# +# Code uses single 1K S-box and is >2 times faster than code generated +# by gcc-3.4.1. This is thanks to unique feature of ARMv4 ISA, which +# allows to merge logical or arithmetic operation with shift or rotate +# in one instruction and emit combined result every cycle. The module +# is endian-neutral. The performance is ~42 cycles/byte for 128-bit +# key [on single-issue Xscale PXA250 core]. + +# May 2007. +# +# AES_set_[en|de]crypt_key is added. + +# July 2010. +# +# Rescheduling for dual-issue pipeline resulted in 12% improvement on +# Cortex A8 core and ~25 cycles per byte processed with 128-bit key. + +# February 2011. +# +# Profiler-assisted and platform-specific optimization resulted in 16% +# improvement on Cortex A8 core and ~21.5 cycles per byte. + +$flavour = shift; +if ($flavour=~/\w[\w\-]*\.\w+$/) { $output=$flavour; undef $flavour; } +else { while (($output=shift) && ($output!~/\w[\w\-]*\.\w+$/)) {} } + +if ($flavour && $flavour ne "void") { + $0 =~ m/(.*[\/\\])[^\/\\]+$/; $dir=$1; + ( $xlate="${dir}arm-xlate.pl" and -f $xlate ) or + ( $xlate="${dir}../../perlasm/arm-xlate.pl" and -f $xlate) or + die "can't locate arm-xlate.pl"; + + open STDOUT,"| \"$^X\" $xlate $flavour $output"; +} else { + open STDOUT,">$output"; +} + +$s0="r0"; +$s1="r1"; +$s2="r2"; +$s3="r3"; +$t1="r4"; +$t2="r5"; +$t3="r6"; +$i1="r7"; +$i2="r8"; +$i3="r9"; + +$tbl="r10"; +$key="r11"; +$rounds="r12"; + +$code=<<___; +#ifndef __KERNEL__ +# include "arm_arch.h" +#else +# define __ARM_ARCH__ __LINUX_ARM_ARCH__ +#endif + +.text +#if defined(__thumb2__) && !defined(__APPLE__) +.syntax unified +.thumb +#else +.code 32 +#undef __thumb2__ +#endif + +.type AES_Te,%object +.align 5 +AES_Te: +.word 0xc66363a5, 0xf87c7c84, 0xee777799, 0xf67b7b8d +.word 0xfff2f20d, 0xd66b6bbd, 0xde6f6fb1, 0x91c5c554 +.word 0x60303050, 0x02010103, 0xce6767a9, 0x562b2b7d +.word 0xe7fefe19, 0xb5d7d762, 0x4dababe6, 0xec76769a +.word 0x8fcaca45, 0x1f82829d, 0x89c9c940, 0xfa7d7d87 +.word 0xeffafa15, 0xb25959eb, 0x8e4747c9, 0xfbf0f00b +.word 0x41adadec, 0xb3d4d467, 0x5fa2a2fd, 0x45afafea +.word 0x239c9cbf, 0x53a4a4f7, 0xe4727296, 0x9bc0c05b +.word 0x75b7b7c2, 0xe1fdfd1c, 0x3d9393ae, 0x4c26266a +.word 0x6c36365a, 0x7e3f3f41, 0xf5f7f702, 0x83cccc4f +.word 0x6834345c, 0x51a5a5f4, 0xd1e5e534, 0xf9f1f108 +.word 0xe2717193, 0xabd8d873, 0x62313153, 0x2a15153f +.word 0x0804040c, 0x95c7c752, 0x46232365, 0x9dc3c35e +.word 0x30181828, 0x379696a1, 0x0a05050f, 0x2f9a9ab5 +.word 0x0e070709, 0x24121236, 0x1b80809b, 0xdfe2e23d +.word 0xcdebeb26, 0x4e272769, 0x7fb2b2cd, 0xea75759f +.word 0x1209091b, 0x1d83839e, 0x582c2c74, 0x341a1a2e +.word 0x361b1b2d, 0xdc6e6eb2, 0xb45a5aee, 0x5ba0a0fb +.word 0xa45252f6, 0x763b3b4d, 0xb7d6d661, 0x7db3b3ce +.word 0x5229297b, 0xdde3e33e, 0x5e2f2f71, 0x13848497 +.word 0xa65353f5, 0xb9d1d168, 0x00000000, 0xc1eded2c +.word 0x40202060, 0xe3fcfc1f, 0x79b1b1c8, 0xb65b5bed +.word 0xd46a6abe, 0x8dcbcb46, 0x67bebed9, 0x7239394b +.word 0x944a4ade, 0x984c4cd4, 0xb05858e8, 0x85cfcf4a +.word 0xbbd0d06b, 0xc5efef2a, 0x4faaaae5, 0xedfbfb16 +.word 0x864343c5, 0x9a4d4dd7, 0x66333355, 0x11858594 +.word 0x8a4545cf, 0xe9f9f910, 0x04020206, 0xfe7f7f81 +.word 0xa05050f0, 0x783c3c44, 0x259f9fba, 0x4ba8a8e3 +.word 0xa25151f3, 0x5da3a3fe, 0x804040c0, 0x058f8f8a +.word 0x3f9292ad, 0x219d9dbc, 0x70383848, 0xf1f5f504 +.word 0x63bcbcdf, 0x77b6b6c1, 0xafdada75, 0x42212163 +.word 0x20101030, 0xe5ffff1a, 0xfdf3f30e, 0xbfd2d26d +.word 0x81cdcd4c, 0x180c0c14, 0x26131335, 0xc3ecec2f +.word 0xbe5f5fe1, 0x359797a2, 0x884444cc, 0x2e171739 +.word 0x93c4c457, 0x55a7a7f2, 0xfc7e7e82, 0x7a3d3d47 +.word 0xc86464ac, 0xba5d5de7, 0x3219192b, 0xe6737395 +.word 0xc06060a0, 0x19818198, 0x9e4f4fd1, 0xa3dcdc7f +.word 0x44222266, 0x542a2a7e, 0x3b9090ab, 0x0b888883 +.word 0x8c4646ca, 0xc7eeee29, 0x6bb8b8d3, 0x2814143c +.word 0xa7dede79, 0xbc5e5ee2, 0x160b0b1d, 0xaddbdb76 +.word 0xdbe0e03b, 0x64323256, 0x743a3a4e, 0x140a0a1e +.word 0x924949db, 0x0c06060a, 0x4824246c, 0xb85c5ce4 +.word 0x9fc2c25d, 0xbdd3d36e, 0x43acacef, 0xc46262a6 +.word 0x399191a8, 0x319595a4, 0xd3e4e437, 0xf279798b +.word 0xd5e7e732, 0x8bc8c843, 0x6e373759, 0xda6d6db7 +.word 0x018d8d8c, 0xb1d5d564, 0x9c4e4ed2, 0x49a9a9e0 +.word 0xd86c6cb4, 0xac5656fa, 0xf3f4f407, 0xcfeaea25 +.word 0xca6565af, 0xf47a7a8e, 0x47aeaee9, 0x10080818 +.word 0x6fbabad5, 0xf0787888, 0x4a25256f, 0x5c2e2e72 +.word 0x381c1c24, 0x57a6a6f1, 0x73b4b4c7, 0x97c6c651 +.word 0xcbe8e823, 0xa1dddd7c, 0xe874749c, 0x3e1f1f21 +.word 0x964b4bdd, 0x61bdbddc, 0x0d8b8b86, 0x0f8a8a85 +.word 0xe0707090, 0x7c3e3e42, 0x71b5b5c4, 0xcc6666aa +.word 0x904848d8, 0x06030305, 0xf7f6f601, 0x1c0e0e12 +.word 0xc26161a3, 0x6a35355f, 0xae5757f9, 0x69b9b9d0 +.word 0x17868691, 0x99c1c158, 0x3a1d1d27, 0x279e9eb9 +.word 0xd9e1e138, 0xebf8f813, 0x2b9898b3, 0x22111133 +.word 0xd26969bb, 0xa9d9d970, 0x078e8e89, 0x339494a7 +.word 0x2d9b9bb6, 0x3c1e1e22, 0x15878792, 0xc9e9e920 +.word 0x87cece49, 0xaa5555ff, 0x50282878, 0xa5dfdf7a +.word 0x038c8c8f, 0x59a1a1f8, 0x09898980, 0x1a0d0d17 +.word 0x65bfbfda, 0xd7e6e631, 0x844242c6, 0xd06868b8 +.word 0x824141c3, 0x299999b0, 0x5a2d2d77, 0x1e0f0f11 +.word 0x7bb0b0cb, 0xa85454fc, 0x6dbbbbd6, 0x2c16163a +@ Te4[256] +.byte 0x63, 0x7c, 0x77, 0x7b, 0xf2, 0x6b, 0x6f, 0xc5 +.byte 0x30, 0x01, 0x67, 0x2b, 0xfe, 0xd7, 0xab, 0x76 +.byte 0xca, 0x82, 0xc9, 0x7d, 0xfa, 0x59, 0x47, 0xf0 +.byte 0xad, 0xd4, 0xa2, 0xaf, 0x9c, 0xa4, 0x72, 0xc0 +.byte 0xb7, 0xfd, 0x93, 0x26, 0x36, 0x3f, 0xf7, 0xcc +.byte 0x34, 0xa5, 0xe5, 0xf1, 0x71, 0xd8, 0x31, 0x15 +.byte 0x04, 0xc7, 0x23, 0xc3, 0x18, 0x96, 0x05, 0x9a +.byte 0x07, 0x12, 0x80, 0xe2, 0xeb, 0x27, 0xb2, 0x75 +.byte 0x09, 0x83, 0x2c, 0x1a, 0x1b, 0x6e, 0x5a, 0xa0 +.byte 0x52, 0x3b, 0xd6, 0xb3, 0x29, 0xe3, 0x2f, 0x84 +.byte 0x53, 0xd1, 0x00, 0xed, 0x20, 0xfc, 0xb1, 0x5b +.byte 0x6a, 0xcb, 0xbe, 0x39, 0x4a, 0x4c, 0x58, 0xcf +.byte 0xd0, 0xef, 0xaa, 0xfb, 0x43, 0x4d, 0x33, 0x85 +.byte 0x45, 0xf9, 0x02, 0x7f, 0x50, 0x3c, 0x9f, 0xa8 +.byte 0x51, 0xa3, 0x40, 0x8f, 0x92, 0x9d, 0x38, 0xf5 +.byte 0xbc, 0xb6, 0xda, 0x21, 0x10, 0xff, 0xf3, 0xd2 +.byte 0xcd, 0x0c, 0x13, 0xec, 0x5f, 0x97, 0x44, 0x17 +.byte 0xc4, 0xa7, 0x7e, 0x3d, 0x64, 0x5d, 0x19, 0x73 +.byte 0x60, 0x81, 0x4f, 0xdc, 0x22, 0x2a, 0x90, 0x88 +.byte 0x46, 0xee, 0xb8, 0x14, 0xde, 0x5e, 0x0b, 0xdb +.byte 0xe0, 0x32, 0x3a, 0x0a, 0x49, 0x06, 0x24, 0x5c +.byte 0xc2, 0xd3, 0xac, 0x62, 0x91, 0x95, 0xe4, 0x79 +.byte 0xe7, 0xc8, 0x37, 0x6d, 0x8d, 0xd5, 0x4e, 0xa9 +.byte 0x6c, 0x56, 0xf4, 0xea, 0x65, 0x7a, 0xae, 0x08 +.byte 0xba, 0x78, 0x25, 0x2e, 0x1c, 0xa6, 0xb4, 0xc6 +.byte 0xe8, 0xdd, 0x74, 0x1f, 0x4b, 0xbd, 0x8b, 0x8a +.byte 0x70, 0x3e, 0xb5, 0x66, 0x48, 0x03, 0xf6, 0x0e +.byte 0x61, 0x35, 0x57, 0xb9, 0x86, 0xc1, 0x1d, 0x9e +.byte 0xe1, 0xf8, 0x98, 0x11, 0x69, 0xd9, 0x8e, 0x94 +.byte 0x9b, 0x1e, 0x87, 0xe9, 0xce, 0x55, 0x28, 0xdf +.byte 0x8c, 0xa1, 0x89, 0x0d, 0xbf, 0xe6, 0x42, 0x68 +.byte 0x41, 0x99, 0x2d, 0x0f, 0xb0, 0x54, 0xbb, 0x16 +@ rcon[] +.word 0x01000000, 0x02000000, 0x04000000, 0x08000000 +.word 0x10000000, 0x20000000, 0x40000000, 0x80000000 +.word 0x1B000000, 0x36000000, 0, 0, 0, 0, 0, 0 +.size AES_Te,.-AES_Te + +@ void AES_encrypt(const unsigned char *in, unsigned char *out, +@ const AES_KEY *key) { +.global AES_encrypt +.type AES_encrypt,%function +.align 5 +AES_encrypt: +#ifndef __thumb2__ + sub r3,pc,#8 @ AES_encrypt +#else + adr r3,. +#endif + stmdb sp!,{r1,r4-r12,lr} +#if defined(__thumb2__) || defined(__APPLE__) + adr $tbl,AES_Te +#else + sub $tbl,r3,#AES_encrypt-AES_Te @ Te +#endif + mov $rounds,r0 @ inp + mov $key,r2 +#if __ARM_ARCH__<7 + ldrb $s0,[$rounds,#3] @ load input data in endian-neutral + ldrb $t1,[$rounds,#2] @ manner... + ldrb $t2,[$rounds,#1] + ldrb $t3,[$rounds,#0] + orr $s0,$s0,$t1,lsl#8 + ldrb $s1,[$rounds,#7] + orr $s0,$s0,$t2,lsl#16 + ldrb $t1,[$rounds,#6] + orr $s0,$s0,$t3,lsl#24 + ldrb $t2,[$rounds,#5] + ldrb $t3,[$rounds,#4] + orr $s1,$s1,$t1,lsl#8 + ldrb $s2,[$rounds,#11] + orr $s1,$s1,$t2,lsl#16 + ldrb $t1,[$rounds,#10] + orr $s1,$s1,$t3,lsl#24 + ldrb $t2,[$rounds,#9] + ldrb $t3,[$rounds,#8] + orr $s2,$s2,$t1,lsl#8 + ldrb $s3,[$rounds,#15] + orr $s2,$s2,$t2,lsl#16 + ldrb $t1,[$rounds,#14] + orr $s2,$s2,$t3,lsl#24 + ldrb $t2,[$rounds,#13] + ldrb $t3,[$rounds,#12] + orr $s3,$s3,$t1,lsl#8 + orr $s3,$s3,$t2,lsl#16 + orr $s3,$s3,$t3,lsl#24 +#else + ldr $s0,[$rounds,#0] + ldr $s1,[$rounds,#4] + ldr $s2,[$rounds,#8] + ldr $s3,[$rounds,#12] +#ifdef __ARMEL__ + rev $s0,$s0 + rev $s1,$s1 + rev $s2,$s2 + rev $s3,$s3 +#endif +#endif + bl _armv4_AES_encrypt + + ldr $rounds,[sp],#4 @ pop out +#if __ARM_ARCH__>=7 +#ifdef __ARMEL__ + rev $s0,$s0 + rev $s1,$s1 + rev $s2,$s2 + rev $s3,$s3 +#endif + str $s0,[$rounds,#0] + str $s1,[$rounds,#4] + str $s2,[$rounds,#8] + str $s3,[$rounds,#12] +#else + mov $t1,$s0,lsr#24 @ write output in endian-neutral + mov $t2,$s0,lsr#16 @ manner... + mov $t3,$s0,lsr#8 + strb $t1,[$rounds,#0] + strb $t2,[$rounds,#1] + mov $t1,$s1,lsr#24 + strb $t3,[$rounds,#2] + mov $t2,$s1,lsr#16 + strb $s0,[$rounds,#3] + mov $t3,$s1,lsr#8 + strb $t1,[$rounds,#4] + strb $t2,[$rounds,#5] + mov $t1,$s2,lsr#24 + strb $t3,[$rounds,#6] + mov $t2,$s2,lsr#16 + strb $s1,[$rounds,#7] + mov $t3,$s2,lsr#8 + strb $t1,[$rounds,#8] + strb $t2,[$rounds,#9] + mov $t1,$s3,lsr#24 + strb $t3,[$rounds,#10] + mov $t2,$s3,lsr#16 + strb $s2,[$rounds,#11] + mov $t3,$s3,lsr#8 + strb $t1,[$rounds,#12] + strb $t2,[$rounds,#13] + strb $t3,[$rounds,#14] + strb $s3,[$rounds,#15] +#endif +#if __ARM_ARCH__>=5 + ldmia sp!,{r4-r12,pc} +#else + ldmia sp!,{r4-r12,lr} + tst lr,#1 + moveq pc,lr @ be binary compatible with V4, yet + bx lr @ interoperable with Thumb ISA:-) +#endif +.size AES_encrypt,.-AES_encrypt + +.type _armv4_AES_encrypt,%function +.align 2 +_armv4_AES_encrypt: + str lr,[sp,#-4]! @ push lr + ldmia $key!,{$t1-$i1} + eor $s0,$s0,$t1 + ldr $rounds,[$key,#240-16] + eor $s1,$s1,$t2 + eor $s2,$s2,$t3 + eor $s3,$s3,$i1 + sub $rounds,$rounds,#1 + mov lr,#255 + + and $i1,lr,$s0 + and $i2,lr,$s0,lsr#8 + and $i3,lr,$s0,lsr#16 + mov $s0,$s0,lsr#24 +.Lenc_loop: + ldr $t1,[$tbl,$i1,lsl#2] @ Te3[s0>>0] + and $i1,lr,$s1,lsr#16 @ i0 + ldr $t2,[$tbl,$i2,lsl#2] @ Te2[s0>>8] + and $i2,lr,$s1 + ldr $t3,[$tbl,$i3,lsl#2] @ Te1[s0>>16] + and $i3,lr,$s1,lsr#8 + ldr $s0,[$tbl,$s0,lsl#2] @ Te0[s0>>24] + mov $s1,$s1,lsr#24 + + ldr $i1,[$tbl,$i1,lsl#2] @ Te1[s1>>16] + ldr $i2,[$tbl,$i2,lsl#2] @ Te3[s1>>0] + ldr $i3,[$tbl,$i3,lsl#2] @ Te2[s1>>8] + eor $s0,$s0,$i1,ror#8 + ldr $s1,[$tbl,$s1,lsl#2] @ Te0[s1>>24] + and $i1,lr,$s2,lsr#8 @ i0 + eor $t2,$t2,$i2,ror#8 + and $i2,lr,$s2,lsr#16 @ i1 + eor $t3,$t3,$i3,ror#8 + and $i3,lr,$s2 + ldr $i1,[$tbl,$i1,lsl#2] @ Te2[s2>>8] + eor $s1,$s1,$t1,ror#24 + ldr $i2,[$tbl,$i2,lsl#2] @ Te1[s2>>16] + mov $s2,$s2,lsr#24 + + ldr $i3,[$tbl,$i3,lsl#2] @ Te3[s2>>0] + eor $s0,$s0,$i1,ror#16 + ldr $s2,[$tbl,$s2,lsl#2] @ Te0[s2>>24] + and $i1,lr,$s3 @ i0 + eor $s1,$s1,$i2,ror#8 + and $i2,lr,$s3,lsr#8 @ i1 + eor $t3,$t3,$i3,ror#16 + and $i3,lr,$s3,lsr#16 @ i2 + ldr $i1,[$tbl,$i1,lsl#2] @ Te3[s3>>0] + eor $s2,$s2,$t2,ror#16 + ldr $i2,[$tbl,$i2,lsl#2] @ Te2[s3>>8] + mov $s3,$s3,lsr#24 + + ldr $i3,[$tbl,$i3,lsl#2] @ Te1[s3>>16] + eor $s0,$s0,$i1,ror#24 + ldr $i1,[$key],#16 + eor $s1,$s1,$i2,ror#16 + ldr $s3,[$tbl,$s3,lsl#2] @ Te0[s3>>24] + eor $s2,$s2,$i3,ror#8 + ldr $t1,[$key,#-12] + eor $s3,$s3,$t3,ror#8 + + ldr $t2,[$key,#-8] + eor $s0,$s0,$i1 + ldr $t3,[$key,#-4] + and $i1,lr,$s0 + eor $s1,$s1,$t1 + and $i2,lr,$s0,lsr#8 + eor $s2,$s2,$t2 + and $i3,lr,$s0,lsr#16 + eor $s3,$s3,$t3 + mov $s0,$s0,lsr#24 + + subs $rounds,$rounds,#1 + bne .Lenc_loop + + add $tbl,$tbl,#2 + + ldrb $t1,[$tbl,$i1,lsl#2] @ Te4[s0>>0] + and $i1,lr,$s1,lsr#16 @ i0 + ldrb $t2,[$tbl,$i2,lsl#2] @ Te4[s0>>8] + and $i2,lr,$s1 + ldrb $t3,[$tbl,$i3,lsl#2] @ Te4[s0>>16] + and $i3,lr,$s1,lsr#8 + ldrb $s0,[$tbl,$s0,lsl#2] @ Te4[s0>>24] + mov $s1,$s1,lsr#24 + + ldrb $i1,[$tbl,$i1,lsl#2] @ Te4[s1>>16] + ldrb $i2,[$tbl,$i2,lsl#2] @ Te4[s1>>0] + ldrb $i3,[$tbl,$i3,lsl#2] @ Te4[s1>>8] + eor $s0,$i1,$s0,lsl#8 + ldrb $s1,[$tbl,$s1,lsl#2] @ Te4[s1>>24] + and $i1,lr,$s2,lsr#8 @ i0 + eor $t2,$i2,$t2,lsl#8 + and $i2,lr,$s2,lsr#16 @ i1 + eor $t3,$i3,$t3,lsl#8 + and $i3,lr,$s2 + ldrb $i1,[$tbl,$i1,lsl#2] @ Te4[s2>>8] + eor $s1,$t1,$s1,lsl#24 + ldrb $i2,[$tbl,$i2,lsl#2] @ Te4[s2>>16] + mov $s2,$s2,lsr#24 + + ldrb $i3,[$tbl,$i3,lsl#2] @ Te4[s2>>0] + eor $s0,$i1,$s0,lsl#8 + ldrb $s2,[$tbl,$s2,lsl#2] @ Te4[s2>>24] + and $i1,lr,$s3 @ i0 + eor $s1,$s1,$i2,lsl#16 + and $i2,lr,$s3,lsr#8 @ i1 + eor $t3,$i3,$t3,lsl#8 + and $i3,lr,$s3,lsr#16 @ i2 + ldrb $i1,[$tbl,$i1,lsl#2] @ Te4[s3>>0] + eor $s2,$t2,$s2,lsl#24 + ldrb $i2,[$tbl,$i2,lsl#2] @ Te4[s3>>8] + mov $s3,$s3,lsr#24 + + ldrb $i3,[$tbl,$i3,lsl#2] @ Te4[s3>>16] + eor $s0,$i1,$s0,lsl#8 + ldr $i1,[$key,#0] + ldrb $s3,[$tbl,$s3,lsl#2] @ Te4[s3>>24] + eor $s1,$s1,$i2,lsl#8 + ldr $t1,[$key,#4] + eor $s2,$s2,$i3,lsl#16 + ldr $t2,[$key,#8] + eor $s3,$t3,$s3,lsl#24 + ldr $t3,[$key,#12] + + eor $s0,$s0,$i1 + eor $s1,$s1,$t1 + eor $s2,$s2,$t2 + eor $s3,$s3,$t3 + + sub $tbl,$tbl,#2 + ldr pc,[sp],#4 @ pop and return +.size _armv4_AES_encrypt,.-_armv4_AES_encrypt + +.global AES_set_encrypt_key +.type AES_set_encrypt_key,%function +.align 5 +AES_set_encrypt_key: +_armv4_AES_set_encrypt_key: +#ifndef __thumb2__ + sub r3,pc,#8 @ AES_set_encrypt_key +#else + adr r3,. +#endif + teq r0,#0 +#ifdef __thumb2__ + itt eq @ Thumb2 thing, sanity check in ARM +#endif + moveq r0,#-1 + beq .Labrt + teq r2,#0 +#ifdef __thumb2__ + itt eq @ Thumb2 thing, sanity check in ARM +#endif + moveq r0,#-1 + beq .Labrt + + teq r1,#128 + beq .Lok + teq r1,#192 + beq .Lok + teq r1,#256 +#ifdef __thumb2__ + itt ne @ Thumb2 thing, sanity check in ARM +#endif + movne r0,#-1 + bne .Labrt + +.Lok: stmdb sp!,{r4-r12,lr} + mov $rounds,r0 @ inp + mov lr,r1 @ bits + mov $key,r2 @ key + +#if defined(__thumb2__) || defined(__APPLE__) + adr $tbl,AES_Te+1024 @ Te4 +#else + sub $tbl,r3,#_armv4_AES_set_encrypt_key-AES_Te-1024 @ Te4 +#endif + +#if __ARM_ARCH__<7 + ldrb $s0,[$rounds,#3] @ load input data in endian-neutral + ldrb $t1,[$rounds,#2] @ manner... + ldrb $t2,[$rounds,#1] + ldrb $t3,[$rounds,#0] + orr $s0,$s0,$t1,lsl#8 + ldrb $s1,[$rounds,#7] + orr $s0,$s0,$t2,lsl#16 + ldrb $t1,[$rounds,#6] + orr $s0,$s0,$t3,lsl#24 + ldrb $t2,[$rounds,#5] + ldrb $t3,[$rounds,#4] + orr $s1,$s1,$t1,lsl#8 + ldrb $s2,[$rounds,#11] + orr $s1,$s1,$t2,lsl#16 + ldrb $t1,[$rounds,#10] + orr $s1,$s1,$t3,lsl#24 + ldrb $t2,[$rounds,#9] + ldrb $t3,[$rounds,#8] + orr $s2,$s2,$t1,lsl#8 + ldrb $s3,[$rounds,#15] + orr $s2,$s2,$t2,lsl#16 + ldrb $t1,[$rounds,#14] + orr $s2,$s2,$t3,lsl#24 + ldrb $t2,[$rounds,#13] + ldrb $t3,[$rounds,#12] + orr $s3,$s3,$t1,lsl#8 + str $s0,[$key],#16 + orr $s3,$s3,$t2,lsl#16 + str $s1,[$key,#-12] + orr $s3,$s3,$t3,lsl#24 + str $s2,[$key,#-8] + str $s3,[$key,#-4] +#else + ldr $s0,[$rounds,#0] + ldr $s1,[$rounds,#4] + ldr $s2,[$rounds,#8] + ldr $s3,[$rounds,#12] +#ifdef __ARMEL__ + rev $s0,$s0 + rev $s1,$s1 + rev $s2,$s2 + rev $s3,$s3 +#endif + str $s0,[$key],#16 + str $s1,[$key,#-12] + str $s2,[$key,#-8] + str $s3,[$key,#-4] +#endif + + teq lr,#128 + bne .Lnot128 + mov $rounds,#10 + str $rounds,[$key,#240-16] + add $t3,$tbl,#256 @ rcon + mov lr,#255 + +.L128_loop: + and $t2,lr,$s3,lsr#24 + and $i1,lr,$s3,lsr#16 + ldrb $t2,[$tbl,$t2] + and $i2,lr,$s3,lsr#8 + ldrb $i1,[$tbl,$i1] + and $i3,lr,$s3 + ldrb $i2,[$tbl,$i2] + orr $t2,$t2,$i1,lsl#24 + ldrb $i3,[$tbl,$i3] + orr $t2,$t2,$i2,lsl#16 + ldr $t1,[$t3],#4 @ rcon[i++] + orr $t2,$t2,$i3,lsl#8 + eor $t2,$t2,$t1 + eor $s0,$s0,$t2 @ rk[4]=rk[0]^... + eor $s1,$s1,$s0 @ rk[5]=rk[1]^rk[4] + str $s0,[$key],#16 + eor $s2,$s2,$s1 @ rk[6]=rk[2]^rk[5] + str $s1,[$key,#-12] + eor $s3,$s3,$s2 @ rk[7]=rk[3]^rk[6] + str $s2,[$key,#-8] + subs $rounds,$rounds,#1 + str $s3,[$key,#-4] + bne .L128_loop + sub r2,$key,#176 + b .Ldone + +.Lnot128: +#if __ARM_ARCH__<7 + ldrb $i2,[$rounds,#19] + ldrb $t1,[$rounds,#18] + ldrb $t2,[$rounds,#17] + ldrb $t3,[$rounds,#16] + orr $i2,$i2,$t1,lsl#8 + ldrb $i3,[$rounds,#23] + orr $i2,$i2,$t2,lsl#16 + ldrb $t1,[$rounds,#22] + orr $i2,$i2,$t3,lsl#24 + ldrb $t2,[$rounds,#21] + ldrb $t3,[$rounds,#20] + orr $i3,$i3,$t1,lsl#8 + orr $i3,$i3,$t2,lsl#16 + str $i2,[$key],#8 + orr $i3,$i3,$t3,lsl#24 + str $i3,[$key,#-4] +#else + ldr $i2,[$rounds,#16] + ldr $i3,[$rounds,#20] +#ifdef __ARMEL__ + rev $i2,$i2 + rev $i3,$i3 +#endif + str $i2,[$key],#8 + str $i3,[$key,#-4] +#endif + + teq lr,#192 + bne .Lnot192 + mov $rounds,#12 + str $rounds,[$key,#240-24] + add $t3,$tbl,#256 @ rcon + mov lr,#255 + mov $rounds,#8 + +.L192_loop: + and $t2,lr,$i3,lsr#24 + and $i1,lr,$i3,lsr#16 + ldrb $t2,[$tbl,$t2] + and $i2,lr,$i3,lsr#8 + ldrb $i1,[$tbl,$i1] + and $i3,lr,$i3 + ldrb $i2,[$tbl,$i2] + orr $t2,$t2,$i1,lsl#24 + ldrb $i3,[$tbl,$i3] + orr $t2,$t2,$i2,lsl#16 + ldr $t1,[$t3],#4 @ rcon[i++] + orr $t2,$t2,$i3,lsl#8 + eor $i3,$t2,$t1 + eor $s0,$s0,$i3 @ rk[6]=rk[0]^... + eor $s1,$s1,$s0 @ rk[7]=rk[1]^rk[6] + str $s0,[$key],#24 + eor $s2,$s2,$s1 @ rk[8]=rk[2]^rk[7] + str $s1,[$key,#-20] + eor $s3,$s3,$s2 @ rk[9]=rk[3]^rk[8] + str $s2,[$key,#-16] + subs $rounds,$rounds,#1 + str $s3,[$key,#-12] +#ifdef __thumb2__ + itt eq @ Thumb2 thing, sanity check in ARM +#endif + subeq r2,$key,#216 + beq .Ldone + + ldr $i1,[$key,#-32] + ldr $i2,[$key,#-28] + eor $i1,$i1,$s3 @ rk[10]=rk[4]^rk[9] + eor $i3,$i2,$i1 @ rk[11]=rk[5]^rk[10] + str $i1,[$key,#-8] + str $i3,[$key,#-4] + b .L192_loop + +.Lnot192: +#if __ARM_ARCH__<7 + ldrb $i2,[$rounds,#27] + ldrb $t1,[$rounds,#26] + ldrb $t2,[$rounds,#25] + ldrb $t3,[$rounds,#24] + orr $i2,$i2,$t1,lsl#8 + ldrb $i3,[$rounds,#31] + orr $i2,$i2,$t2,lsl#16 + ldrb $t1,[$rounds,#30] + orr $i2,$i2,$t3,lsl#24 + ldrb $t2,[$rounds,#29] + ldrb $t3,[$rounds,#28] + orr $i3,$i3,$t1,lsl#8 + orr $i3,$i3,$t2,lsl#16 + str $i2,[$key],#8 + orr $i3,$i3,$t3,lsl#24 + str $i3,[$key,#-4] +#else + ldr $i2,[$rounds,#24] + ldr $i3,[$rounds,#28] +#ifdef __ARMEL__ + rev $i2,$i2 + rev $i3,$i3 +#endif + str $i2,[$key],#8 + str $i3,[$key,#-4] +#endif + + mov $rounds,#14 + str $rounds,[$key,#240-32] + add $t3,$tbl,#256 @ rcon + mov lr,#255 + mov $rounds,#7 + +.L256_loop: + and $t2,lr,$i3,lsr#24 + and $i1,lr,$i3,lsr#16 + ldrb $t2,[$tbl,$t2] + and $i2,lr,$i3,lsr#8 + ldrb $i1,[$tbl,$i1] + and $i3,lr,$i3 + ldrb $i2,[$tbl,$i2] + orr $t2,$t2,$i1,lsl#24 + ldrb $i3,[$tbl,$i3] + orr $t2,$t2,$i2,lsl#16 + ldr $t1,[$t3],#4 @ rcon[i++] + orr $t2,$t2,$i3,lsl#8 + eor $i3,$t2,$t1 + eor $s0,$s0,$i3 @ rk[8]=rk[0]^... + eor $s1,$s1,$s0 @ rk[9]=rk[1]^rk[8] + str $s0,[$key],#32 + eor $s2,$s2,$s1 @ rk[10]=rk[2]^rk[9] + str $s1,[$key,#-28] + eor $s3,$s3,$s2 @ rk[11]=rk[3]^rk[10] + str $s2,[$key,#-24] + subs $rounds,$rounds,#1 + str $s3,[$key,#-20] +#ifdef __thumb2__ + itt eq @ Thumb2 thing, sanity check in ARM +#endif + subeq r2,$key,#256 + beq .Ldone + + and $t2,lr,$s3 + and $i1,lr,$s3,lsr#8 + ldrb $t2,[$tbl,$t2] + and $i2,lr,$s3,lsr#16 + ldrb $i1,[$tbl,$i1] + and $i3,lr,$s3,lsr#24 + ldrb $i2,[$tbl,$i2] + orr $t2,$t2,$i1,lsl#8 + ldrb $i3,[$tbl,$i3] + orr $t2,$t2,$i2,lsl#16 + ldr $t1,[$key,#-48] + orr $t2,$t2,$i3,lsl#24 + + ldr $i1,[$key,#-44] + ldr $i2,[$key,#-40] + eor $t1,$t1,$t2 @ rk[12]=rk[4]^... + ldr $i3,[$key,#-36] + eor $i1,$i1,$t1 @ rk[13]=rk[5]^rk[12] + str $t1,[$key,#-16] + eor $i2,$i2,$i1 @ rk[14]=rk[6]^rk[13] + str $i1,[$key,#-12] + eor $i3,$i3,$i2 @ rk[15]=rk[7]^rk[14] + str $i2,[$key,#-8] + str $i3,[$key,#-4] + b .L256_loop + +.align 2 +.Ldone: mov r0,#0 + ldmia sp!,{r4-r12,lr} +.Labrt: +#if __ARM_ARCH__>=5 + ret @ bx lr +#else + tst lr,#1 + moveq pc,lr @ be binary compatible with V4, yet + bx lr @ interoperable with Thumb ISA:-) +#endif +.size AES_set_encrypt_key,.-AES_set_encrypt_key + +.global AES_set_decrypt_key +.type AES_set_decrypt_key,%function +.align 5 +AES_set_decrypt_key: + str lr,[sp,#-4]! @ push lr + bl _armv4_AES_set_encrypt_key + teq r0,#0 + ldr lr,[sp],#4 @ pop lr + bne .Labrt + + mov r0,r2 @ AES_set_encrypt_key preserves r2, + mov r1,r2 @ which is AES_KEY *key + b _armv4_AES_set_enc2dec_key +.size AES_set_decrypt_key,.-AES_set_decrypt_key + +@ void AES_set_enc2dec_key(const AES_KEY *inp,AES_KEY *out) +.global AES_set_enc2dec_key +.type AES_set_enc2dec_key,%function +.align 5 +AES_set_enc2dec_key: +_armv4_AES_set_enc2dec_key: + stmdb sp!,{r4-r12,lr} + + ldr $rounds,[r0,#240] + mov $i1,r0 @ input + add $i2,r0,$rounds,lsl#4 + mov $key,r1 @ output + add $tbl,r1,$rounds,lsl#4 + str $rounds,[r1,#240] + +.Linv: ldr $s0,[$i1],#16 + ldr $s1,[$i1,#-12] + ldr $s2,[$i1,#-8] + ldr $s3,[$i1,#-4] + ldr $t1,[$i2],#-16 + ldr $t2,[$i2,#16+4] + ldr $t3,[$i2,#16+8] + ldr $i3,[$i2,#16+12] + str $s0,[$tbl],#-16 + str $s1,[$tbl,#16+4] + str $s2,[$tbl,#16+8] + str $s3,[$tbl,#16+12] + str $t1,[$key],#16 + str $t2,[$key,#-12] + str $t3,[$key,#-8] + str $i3,[$key,#-4] + teq $i1,$i2 + bne .Linv + + ldr $s0,[$i1] + ldr $s1,[$i1,#4] + ldr $s2,[$i1,#8] + ldr $s3,[$i1,#12] + str $s0,[$key] + str $s1,[$key,#4] + str $s2,[$key,#8] + str $s3,[$key,#12] + sub $key,$key,$rounds,lsl#3 +___ +$mask80=$i1; +$mask1b=$i2; +$mask7f=$i3; +$code.=<<___; + ldr $s0,[$key,#16]! @ prefetch tp1 + mov $mask80,#0x80 + mov $mask1b,#0x1b + orr $mask80,$mask80,#0x8000 + orr $mask1b,$mask1b,#0x1b00 + orr $mask80,$mask80,$mask80,lsl#16 + orr $mask1b,$mask1b,$mask1b,lsl#16 + sub $rounds,$rounds,#1 + mvn $mask7f,$mask80 + mov $rounds,$rounds,lsl#2 @ (rounds-1)*4 + +.Lmix: and $t1,$s0,$mask80 + and $s1,$s0,$mask7f + sub $t1,$t1,$t1,lsr#7 + and $t1,$t1,$mask1b + eor $s1,$t1,$s1,lsl#1 @ tp2 + + and $t1,$s1,$mask80 + and $s2,$s1,$mask7f + sub $t1,$t1,$t1,lsr#7 + and $t1,$t1,$mask1b + eor $s2,$t1,$s2,lsl#1 @ tp4 + + and $t1,$s2,$mask80 + and $s3,$s2,$mask7f + sub $t1,$t1,$t1,lsr#7 + and $t1,$t1,$mask1b + eor $s3,$t1,$s3,lsl#1 @ tp8 + + eor $t1,$s1,$s2 + eor $t2,$s0,$s3 @ tp9 + eor $t1,$t1,$s3 @ tpe + eor $t1,$t1,$s1,ror#24 + eor $t1,$t1,$t2,ror#24 @ ^= ROTATE(tpb=tp9^tp2,8) + eor $t1,$t1,$s2,ror#16 + eor $t1,$t1,$t2,ror#16 @ ^= ROTATE(tpd=tp9^tp4,16) + eor $t1,$t1,$t2,ror#8 @ ^= ROTATE(tp9,24) + + ldr $s0,[$key,#4] @ prefetch tp1 + str $t1,[$key],#4 + subs $rounds,$rounds,#1 + bne .Lmix + + mov r0,#0 +#if __ARM_ARCH__>=5 + ldmia sp!,{r4-r12,pc} +#else + ldmia sp!,{r4-r12,lr} + tst lr,#1 + moveq pc,lr @ be binary compatible with V4, yet + bx lr @ interoperable with Thumb ISA:-) +#endif +.size AES_set_enc2dec_key,.-AES_set_enc2dec_key + +.type AES_Td,%object +.align 5 +AES_Td: +.word 0x51f4a750, 0x7e416553, 0x1a17a4c3, 0x3a275e96 +.word 0x3bab6bcb, 0x1f9d45f1, 0xacfa58ab, 0x4be30393 +.word 0x2030fa55, 0xad766df6, 0x88cc7691, 0xf5024c25 +.word 0x4fe5d7fc, 0xc52acbd7, 0x26354480, 0xb562a38f +.word 0xdeb15a49, 0x25ba1b67, 0x45ea0e98, 0x5dfec0e1 +.word 0xc32f7502, 0x814cf012, 0x8d4697a3, 0x6bd3f9c6 +.word 0x038f5fe7, 0x15929c95, 0xbf6d7aeb, 0x955259da +.word 0xd4be832d, 0x587421d3, 0x49e06929, 0x8ec9c844 +.word 0x75c2896a, 0xf48e7978, 0x99583e6b, 0x27b971dd +.word 0xbee14fb6, 0xf088ad17, 0xc920ac66, 0x7dce3ab4 +.word 0x63df4a18, 0xe51a3182, 0x97513360, 0x62537f45 +.word 0xb16477e0, 0xbb6bae84, 0xfe81a01c, 0xf9082b94 +.word 0x70486858, 0x8f45fd19, 0x94de6c87, 0x527bf8b7 +.word 0xab73d323, 0x724b02e2, 0xe31f8f57, 0x6655ab2a +.word 0xb2eb2807, 0x2fb5c203, 0x86c57b9a, 0xd33708a5 +.word 0x302887f2, 0x23bfa5b2, 0x02036aba, 0xed16825c +.word 0x8acf1c2b, 0xa779b492, 0xf307f2f0, 0x4e69e2a1 +.word 0x65daf4cd, 0x0605bed5, 0xd134621f, 0xc4a6fe8a +.word 0x342e539d, 0xa2f355a0, 0x058ae132, 0xa4f6eb75 +.word 0x0b83ec39, 0x4060efaa, 0x5e719f06, 0xbd6e1051 +.word 0x3e218af9, 0x96dd063d, 0xdd3e05ae, 0x4de6bd46 +.word 0x91548db5, 0x71c45d05, 0x0406d46f, 0x605015ff +.word 0x1998fb24, 0xd6bde997, 0x894043cc, 0x67d99e77 +.word 0xb0e842bd, 0x07898b88, 0xe7195b38, 0x79c8eedb +.word 0xa17c0a47, 0x7c420fe9, 0xf8841ec9, 0x00000000 +.word 0x09808683, 0x322bed48, 0x1e1170ac, 0x6c5a724e +.word 0xfd0efffb, 0x0f853856, 0x3daed51e, 0x362d3927 +.word 0x0a0fd964, 0x685ca621, 0x9b5b54d1, 0x24362e3a +.word 0x0c0a67b1, 0x9357e70f, 0xb4ee96d2, 0x1b9b919e +.word 0x80c0c54f, 0x61dc20a2, 0x5a774b69, 0x1c121a16 +.word 0xe293ba0a, 0xc0a02ae5, 0x3c22e043, 0x121b171d +.word 0x0e090d0b, 0xf28bc7ad, 0x2db6a8b9, 0x141ea9c8 +.word 0x57f11985, 0xaf75074c, 0xee99ddbb, 0xa37f60fd +.word 0xf701269f, 0x5c72f5bc, 0x44663bc5, 0x5bfb7e34 +.word 0x8b432976, 0xcb23c6dc, 0xb6edfc68, 0xb8e4f163 +.word 0xd731dcca, 0x42638510, 0x13972240, 0x84c61120 +.word 0x854a247d, 0xd2bb3df8, 0xaef93211, 0xc729a16d +.word 0x1d9e2f4b, 0xdcb230f3, 0x0d8652ec, 0x77c1e3d0 +.word 0x2bb3166c, 0xa970b999, 0x119448fa, 0x47e96422 +.word 0xa8fc8cc4, 0xa0f03f1a, 0x567d2cd8, 0x223390ef +.word 0x87494ec7, 0xd938d1c1, 0x8ccaa2fe, 0x98d40b36 +.word 0xa6f581cf, 0xa57ade28, 0xdab78e26, 0x3fadbfa4 +.word 0x2c3a9de4, 0x5078920d, 0x6a5fcc9b, 0x547e4662 +.word 0xf68d13c2, 0x90d8b8e8, 0x2e39f75e, 0x82c3aff5 +.word 0x9f5d80be, 0x69d0937c, 0x6fd52da9, 0xcf2512b3 +.word 0xc8ac993b, 0x10187da7, 0xe89c636e, 0xdb3bbb7b +.word 0xcd267809, 0x6e5918f4, 0xec9ab701, 0x834f9aa8 +.word 0xe6956e65, 0xaaffe67e, 0x21bccf08, 0xef15e8e6 +.word 0xbae79bd9, 0x4a6f36ce, 0xea9f09d4, 0x29b07cd6 +.word 0x31a4b2af, 0x2a3f2331, 0xc6a59430, 0x35a266c0 +.word 0x744ebc37, 0xfc82caa6, 0xe090d0b0, 0x33a7d815 +.word 0xf104984a, 0x41ecdaf7, 0x7fcd500e, 0x1791f62f +.word 0x764dd68d, 0x43efb04d, 0xccaa4d54, 0xe49604df +.word 0x9ed1b5e3, 0x4c6a881b, 0xc12c1fb8, 0x4665517f +.word 0x9d5eea04, 0x018c355d, 0xfa877473, 0xfb0b412e +.word 0xb3671d5a, 0x92dbd252, 0xe9105633, 0x6dd64713 +.word 0x9ad7618c, 0x37a10c7a, 0x59f8148e, 0xeb133c89 +.word 0xcea927ee, 0xb761c935, 0xe11ce5ed, 0x7a47b13c +.word 0x9cd2df59, 0x55f2733f, 0x1814ce79, 0x73c737bf +.word 0x53f7cdea, 0x5ffdaa5b, 0xdf3d6f14, 0x7844db86 +.word 0xcaaff381, 0xb968c43e, 0x3824342c, 0xc2a3405f +.word 0x161dc372, 0xbce2250c, 0x283c498b, 0xff0d9541 +.word 0x39a80171, 0x080cb3de, 0xd8b4e49c, 0x6456c190 +.word 0x7bcb8461, 0xd532b670, 0x486c5c74, 0xd0b85742 +@ Td4[256] +.byte 0x52, 0x09, 0x6a, 0xd5, 0x30, 0x36, 0xa5, 0x38 +.byte 0xbf, 0x40, 0xa3, 0x9e, 0x81, 0xf3, 0xd7, 0xfb +.byte 0x7c, 0xe3, 0x39, 0x82, 0x9b, 0x2f, 0xff, 0x87 +.byte 0x34, 0x8e, 0x43, 0x44, 0xc4, 0xde, 0xe9, 0xcb +.byte 0x54, 0x7b, 0x94, 0x32, 0xa6, 0xc2, 0x23, 0x3d +.byte 0xee, 0x4c, 0x95, 0x0b, 0x42, 0xfa, 0xc3, 0x4e +.byte 0x08, 0x2e, 0xa1, 0x66, 0x28, 0xd9, 0x24, 0xb2 +.byte 0x76, 0x5b, 0xa2, 0x49, 0x6d, 0x8b, 0xd1, 0x25 +.byte 0x72, 0xf8, 0xf6, 0x64, 0x86, 0x68, 0x98, 0x16 +.byte 0xd4, 0xa4, 0x5c, 0xcc, 0x5d, 0x65, 0xb6, 0x92 +.byte 0x6c, 0x70, 0x48, 0x50, 0xfd, 0xed, 0xb9, 0xda +.byte 0x5e, 0x15, 0x46, 0x57, 0xa7, 0x8d, 0x9d, 0x84 +.byte 0x90, 0xd8, 0xab, 0x00, 0x8c, 0xbc, 0xd3, 0x0a +.byte 0xf7, 0xe4, 0x58, 0x05, 0xb8, 0xb3, 0x45, 0x06 +.byte 0xd0, 0x2c, 0x1e, 0x8f, 0xca, 0x3f, 0x0f, 0x02 +.byte 0xc1, 0xaf, 0xbd, 0x03, 0x01, 0x13, 0x8a, 0x6b +.byte 0x3a, 0x91, 0x11, 0x41, 0x4f, 0x67, 0xdc, 0xea +.byte 0x97, 0xf2, 0xcf, 0xce, 0xf0, 0xb4, 0xe6, 0x73 +.byte 0x96, 0xac, 0x74, 0x22, 0xe7, 0xad, 0x35, 0x85 +.byte 0xe2, 0xf9, 0x37, 0xe8, 0x1c, 0x75, 0xdf, 0x6e +.byte 0x47, 0xf1, 0x1a, 0x71, 0x1d, 0x29, 0xc5, 0x89 +.byte 0x6f, 0xb7, 0x62, 0x0e, 0xaa, 0x18, 0xbe, 0x1b +.byte 0xfc, 0x56, 0x3e, 0x4b, 0xc6, 0xd2, 0x79, 0x20 +.byte 0x9a, 0xdb, 0xc0, 0xfe, 0x78, 0xcd, 0x5a, 0xf4 +.byte 0x1f, 0xdd, 0xa8, 0x33, 0x88, 0x07, 0xc7, 0x31 +.byte 0xb1, 0x12, 0x10, 0x59, 0x27, 0x80, 0xec, 0x5f +.byte 0x60, 0x51, 0x7f, 0xa9, 0x19, 0xb5, 0x4a, 0x0d +.byte 0x2d, 0xe5, 0x7a, 0x9f, 0x93, 0xc9, 0x9c, 0xef +.byte 0xa0, 0xe0, 0x3b, 0x4d, 0xae, 0x2a, 0xf5, 0xb0 +.byte 0xc8, 0xeb, 0xbb, 0x3c, 0x83, 0x53, 0x99, 0x61 +.byte 0x17, 0x2b, 0x04, 0x7e, 0xba, 0x77, 0xd6, 0x26 +.byte 0xe1, 0x69, 0x14, 0x63, 0x55, 0x21, 0x0c, 0x7d +.size AES_Td,.-AES_Td + +@ void AES_decrypt(const unsigned char *in, unsigned char *out, +@ const AES_KEY *key) { +.global AES_decrypt +.type AES_decrypt,%function +.align 5 +AES_decrypt: +#ifndef __thumb2__ + sub r3,pc,#8 @ AES_decrypt +#else + adr r3,. +#endif + stmdb sp!,{r1,r4-r12,lr} +#if defined(__thumb2__) || defined(__APPLE__) + adr $tbl,AES_Td +#else + sub $tbl,r3,#AES_decrypt-AES_Td @ Td +#endif + mov $rounds,r0 @ inp + mov $key,r2 +#if __ARM_ARCH__<7 + ldrb $s0,[$rounds,#3] @ load input data in endian-neutral + ldrb $t1,[$rounds,#2] @ manner... + ldrb $t2,[$rounds,#1] + ldrb $t3,[$rounds,#0] + orr $s0,$s0,$t1,lsl#8 + ldrb $s1,[$rounds,#7] + orr $s0,$s0,$t2,lsl#16 + ldrb $t1,[$rounds,#6] + orr $s0,$s0,$t3,lsl#24 + ldrb $t2,[$rounds,#5] + ldrb $t3,[$rounds,#4] + orr $s1,$s1,$t1,lsl#8 + ldrb $s2,[$rounds,#11] + orr $s1,$s1,$t2,lsl#16 + ldrb $t1,[$rounds,#10] + orr $s1,$s1,$t3,lsl#24 + ldrb $t2,[$rounds,#9] + ldrb $t3,[$rounds,#8] + orr $s2,$s2,$t1,lsl#8 + ldrb $s3,[$rounds,#15] + orr $s2,$s2,$t2,lsl#16 + ldrb $t1,[$rounds,#14] + orr $s2,$s2,$t3,lsl#24 + ldrb $t2,[$rounds,#13] + ldrb $t3,[$rounds,#12] + orr $s3,$s3,$t1,lsl#8 + orr $s3,$s3,$t2,lsl#16 + orr $s3,$s3,$t3,lsl#24 +#else + ldr $s0,[$rounds,#0] + ldr $s1,[$rounds,#4] + ldr $s2,[$rounds,#8] + ldr $s3,[$rounds,#12] +#ifdef __ARMEL__ + rev $s0,$s0 + rev $s1,$s1 + rev $s2,$s2 + rev $s3,$s3 +#endif +#endif + bl _armv4_AES_decrypt + + ldr $rounds,[sp],#4 @ pop out +#if __ARM_ARCH__>=7 +#ifdef __ARMEL__ + rev $s0,$s0 + rev $s1,$s1 + rev $s2,$s2 + rev $s3,$s3 +#endif + str $s0,[$rounds,#0] + str $s1,[$rounds,#4] + str $s2,[$rounds,#8] + str $s3,[$rounds,#12] +#else + mov $t1,$s0,lsr#24 @ write output in endian-neutral + mov $t2,$s0,lsr#16 @ manner... + mov $t3,$s0,lsr#8 + strb $t1,[$rounds,#0] + strb $t2,[$rounds,#1] + mov $t1,$s1,lsr#24 + strb $t3,[$rounds,#2] + mov $t2,$s1,lsr#16 + strb $s0,[$rounds,#3] + mov $t3,$s1,lsr#8 + strb $t1,[$rounds,#4] + strb $t2,[$rounds,#5] + mov $t1,$s2,lsr#24 + strb $t3,[$rounds,#6] + mov $t2,$s2,lsr#16 + strb $s1,[$rounds,#7] + mov $t3,$s2,lsr#8 + strb $t1,[$rounds,#8] + strb $t2,[$rounds,#9] + mov $t1,$s3,lsr#24 + strb $t3,[$rounds,#10] + mov $t2,$s3,lsr#16 + strb $s2,[$rounds,#11] + mov $t3,$s3,lsr#8 + strb $t1,[$rounds,#12] + strb $t2,[$rounds,#13] + strb $t3,[$rounds,#14] + strb $s3,[$rounds,#15] +#endif +#if __ARM_ARCH__>=5 + ldmia sp!,{r4-r12,pc} +#else + ldmia sp!,{r4-r12,lr} + tst lr,#1 + moveq pc,lr @ be binary compatible with V4, yet + bx lr @ interoperable with Thumb ISA:-) +#endif +.size AES_decrypt,.-AES_decrypt + +.type _armv4_AES_decrypt,%function +.align 2 +_armv4_AES_decrypt: + str lr,[sp,#-4]! @ push lr + ldmia $key!,{$t1-$i1} + eor $s0,$s0,$t1 + ldr $rounds,[$key,#240-16] + eor $s1,$s1,$t2 + eor $s2,$s2,$t3 + eor $s3,$s3,$i1 + sub $rounds,$rounds,#1 + mov lr,#255 + + and $i1,lr,$s0,lsr#16 + and $i2,lr,$s0,lsr#8 + and $i3,lr,$s0 + mov $s0,$s0,lsr#24 +.Ldec_loop: + ldr $t1,[$tbl,$i1,lsl#2] @ Td1[s0>>16] + and $i1,lr,$s1 @ i0 + ldr $t2,[$tbl,$i2,lsl#2] @ Td2[s0>>8] + and $i2,lr,$s1,lsr#16 + ldr $t3,[$tbl,$i3,lsl#2] @ Td3[s0>>0] + and $i3,lr,$s1,lsr#8 + ldr $s0,[$tbl,$s0,lsl#2] @ Td0[s0>>24] + mov $s1,$s1,lsr#24 + + ldr $i1,[$tbl,$i1,lsl#2] @ Td3[s1>>0] + ldr $i2,[$tbl,$i2,lsl#2] @ Td1[s1>>16] + ldr $i3,[$tbl,$i3,lsl#2] @ Td2[s1>>8] + eor $s0,$s0,$i1,ror#24 + ldr $s1,[$tbl,$s1,lsl#2] @ Td0[s1>>24] + and $i1,lr,$s2,lsr#8 @ i0 + eor $t2,$i2,$t2,ror#8 + and $i2,lr,$s2 @ i1 + eor $t3,$i3,$t3,ror#8 + and $i3,lr,$s2,lsr#16 + ldr $i1,[$tbl,$i1,lsl#2] @ Td2[s2>>8] + eor $s1,$s1,$t1,ror#8 + ldr $i2,[$tbl,$i2,lsl#2] @ Td3[s2>>0] + mov $s2,$s2,lsr#24 + + ldr $i3,[$tbl,$i3,lsl#2] @ Td1[s2>>16] + eor $s0,$s0,$i1,ror#16 + ldr $s2,[$tbl,$s2,lsl#2] @ Td0[s2>>24] + and $i1,lr,$s3,lsr#16 @ i0 + eor $s1,$s1,$i2,ror#24 + and $i2,lr,$s3,lsr#8 @ i1 + eor $t3,$i3,$t3,ror#8 + and $i3,lr,$s3 @ i2 + ldr $i1,[$tbl,$i1,lsl#2] @ Td1[s3>>16] + eor $s2,$s2,$t2,ror#8 + ldr $i2,[$tbl,$i2,lsl#2] @ Td2[s3>>8] + mov $s3,$s3,lsr#24 + + ldr $i3,[$tbl,$i3,lsl#2] @ Td3[s3>>0] + eor $s0,$s0,$i1,ror#8 + ldr $i1,[$key],#16 + eor $s1,$s1,$i2,ror#16 + ldr $s3,[$tbl,$s3,lsl#2] @ Td0[s3>>24] + eor $s2,$s2,$i3,ror#24 + + ldr $t1,[$key,#-12] + eor $s0,$s0,$i1 + ldr $t2,[$key,#-8] + eor $s3,$s3,$t3,ror#8 + ldr $t3,[$key,#-4] + and $i1,lr,$s0,lsr#16 + eor $s1,$s1,$t1 + and $i2,lr,$s0,lsr#8 + eor $s2,$s2,$t2 + and $i3,lr,$s0 + eor $s3,$s3,$t3 + mov $s0,$s0,lsr#24 + + subs $rounds,$rounds,#1 + bne .Ldec_loop + + add $tbl,$tbl,#1024 + + ldr $t2,[$tbl,#0] @ prefetch Td4 + ldr $t3,[$tbl,#32] + ldr $t1,[$tbl,#64] + ldr $t2,[$tbl,#96] + ldr $t3,[$tbl,#128] + ldr $t1,[$tbl,#160] + ldr $t2,[$tbl,#192] + ldr $t3,[$tbl,#224] + + ldrb $s0,[$tbl,$s0] @ Td4[s0>>24] + ldrb $t1,[$tbl,$i1] @ Td4[s0>>16] + and $i1,lr,$s1 @ i0 + ldrb $t2,[$tbl,$i2] @ Td4[s0>>8] + and $i2,lr,$s1,lsr#16 + ldrb $t3,[$tbl,$i3] @ Td4[s0>>0] + and $i3,lr,$s1,lsr#8 + + add $s1,$tbl,$s1,lsr#24 + ldrb $i1,[$tbl,$i1] @ Td4[s1>>0] + ldrb $s1,[$s1] @ Td4[s1>>24] + ldrb $i2,[$tbl,$i2] @ Td4[s1>>16] + eor $s0,$i1,$s0,lsl#24 + ldrb $i3,[$tbl,$i3] @ Td4[s1>>8] + eor $s1,$t1,$s1,lsl#8 + and $i1,lr,$s2,lsr#8 @ i0 + eor $t2,$t2,$i2,lsl#8 + and $i2,lr,$s2 @ i1 + ldrb $i1,[$tbl,$i1] @ Td4[s2>>8] + eor $t3,$t3,$i3,lsl#8 + ldrb $i2,[$tbl,$i2] @ Td4[s2>>0] + and $i3,lr,$s2,lsr#16 + + add $s2,$tbl,$s2,lsr#24 + ldrb $s2,[$s2] @ Td4[s2>>24] + eor $s0,$s0,$i1,lsl#8 + ldrb $i3,[$tbl,$i3] @ Td4[s2>>16] + eor $s1,$i2,$s1,lsl#16 + and $i1,lr,$s3,lsr#16 @ i0 + eor $s2,$t2,$s2,lsl#16 + and $i2,lr,$s3,lsr#8 @ i1 + ldrb $i1,[$tbl,$i1] @ Td4[s3>>16] + eor $t3,$t3,$i3,lsl#16 + ldrb $i2,[$tbl,$i2] @ Td4[s3>>8] + and $i3,lr,$s3 @ i2 + + add $s3,$tbl,$s3,lsr#24 + ldrb $i3,[$tbl,$i3] @ Td4[s3>>0] + ldrb $s3,[$s3] @ Td4[s3>>24] + eor $s0,$s0,$i1,lsl#16 + ldr $i1,[$key,#0] + eor $s1,$s1,$i2,lsl#8 + ldr $t1,[$key,#4] + eor $s2,$i3,$s2,lsl#8 + ldr $t2,[$key,#8] + eor $s3,$t3,$s3,lsl#24 + ldr $t3,[$key,#12] + + eor $s0,$s0,$i1 + eor $s1,$s1,$t1 + eor $s2,$s2,$t2 + eor $s3,$s3,$t3 + + sub $tbl,$tbl,#1024 + ldr pc,[sp],#4 @ pop and return +.size _armv4_AES_decrypt,.-_armv4_AES_decrypt +.asciz "AES for ARMv4, CRYPTOGAMS by " +.align 2 +___ + +$code =~ s/\bbx\s+lr\b/.word\t0xe12fff1e/gm; # make it possible to compile with -march=armv4 +$code =~ s/\bret\b/bx\tlr/gm; + +open SELF,$0; +while() { + next if (/^#!/); + last if (!s/^#/@/ and !/^$/); + print; +} +close SELF; + +print $code; +close STDOUT; # enforce flush diff --git a/openssl-1.1.0h/crypto/aes/asm/aes-c64xplus.pl b/openssl-1.1.0h/crypto/aes/asm/aes-c64xplus.pl new file mode 100644 index 0000000..19d2cc1 --- /dev/null +++ b/openssl-1.1.0h/crypto/aes/asm/aes-c64xplus.pl @@ -0,0 +1,1382 @@ +#! /usr/bin/env perl +# Copyright 2012-2016 The OpenSSL Project Authors. All Rights Reserved. +# +# Licensed under the OpenSSL license (the "License"). You may not use +# this file except in compliance with the License. You can obtain a copy +# in the file LICENSE in the source distribution or at +# https://www.openssl.org/source/license.html + +# +# ==================================================================== +# Written by Andy Polyakov for the OpenSSL +# project. The module is, however, dual licensed under OpenSSL and +# CRYPTOGAMS licenses depending on where you obtain it. For further +# details see http://www.openssl.org/~appro/cryptogams/. +# ==================================================================== +# +# [Endian-neutral] AES for C64x+. +# +# Even though SPLOOPs are scheduled for 13 cycles, and thus expected +# performance is ~8.5 cycles per byte processed with 128-bit key, +# measured performance turned to be ~10 cycles per byte. Discrepancy +# must be caused by limitations of L1D memory banking(*), see SPRU871 +# TI publication for further details. If any consolation it's still +# ~20% faster than TI's linear assembly module anyway... Compared to +# aes_core.c compiled with cl6x 6.0 with -mv6400+ -o2 options this +# code is 3.75x faster and almost 3x smaller (tables included). +# +# (*) This means that there might be subtle correlation between data +# and timing and one can wonder if it can be ... attacked:-( +# On the other hand this also means that *if* one chooses to +# implement *4* T-tables variant [instead of 1 T-table as in +# this implementation, or in addition to], then one ought to +# *interleave* them. Even though it complicates addressing, +# references to interleaved tables would be guaranteed not to +# clash. I reckon that it should be possible to break 8 cycles +# per byte "barrier," i.e. improve by ~20%, naturally at the +# cost of 8x increased pressure on L1D. 8x because you'd have +# to interleave both Te and Td tables... + +while (($output=shift) && ($output!~/\w[\w\-]*\.\w+$/)) {} +open STDOUT,">$output"; + +($TEA,$TEB)=("A5","B5"); +($KPA,$KPB)=("A3","B1"); +@K=("A6","B6","A7","B7"); +@s=("A8","B8","A9","B9"); +@Te0=@Td0=("A16","B16","A17","B17"); +@Te1=@Td1=("A18","B18","A19","B19"); +@Te2=@Td2=("A20","B20","A21","B21"); +@Te3=@Td3=("A22","B22","A23","B23"); + +$code=<<___; + .text + + .if .ASSEMBLER_VERSION<7000000 + .asg 0,__TI_EABI__ + .endif + .if __TI_EABI__ + .nocmp + .asg AES_encrypt,_AES_encrypt + .asg AES_decrypt,_AES_decrypt + .asg AES_set_encrypt_key,_AES_set_encrypt_key + .asg AES_set_decrypt_key,_AES_set_decrypt_key + .asg AES_ctr32_encrypt,_AES_ctr32_encrypt + .endif + + .asg B3,RA + .asg A4,INP + .asg B4,OUT + .asg A6,KEY + .asg A4,RET + .asg B15,SP + + .eval 24,EXT0 + .eval 16,EXT1 + .eval 8,EXT2 + .eval 0,EXT3 + .eval 8,TBL1 + .eval 16,TBL2 + .eval 24,TBL3 + + .if .BIG_ENDIAN + .eval 24-EXT0,EXT0 + .eval 24-EXT1,EXT1 + .eval 24-EXT2,EXT2 + .eval 24-EXT3,EXT3 + .eval 32-TBL1,TBL1 + .eval 32-TBL2,TBL2 + .eval 32-TBL3,TBL3 + .endif + + .global _AES_encrypt +_AES_encrypt: + .asmfunc + MVK 1,B2 +__encrypt: + .if __TI_EABI__ + [B2] LDNDW *INP++,A9:A8 ; load input +|| MVKL \$PCR_OFFSET(AES_Te,__encrypt),$TEA +|| ADDKPC __encrypt,B0 + [B2] LDNDW *INP++,B9:B8 +|| MVKH \$PCR_OFFSET(AES_Te,__encrypt),$TEA +|| ADD 0,KEY,$KPA +|| ADD 4,KEY,$KPB + .else + [B2] LDNDW *INP++,A9:A8 ; load input +|| MVKL (AES_Te-__encrypt),$TEA +|| ADDKPC __encrypt,B0 + [B2] LDNDW *INP++,B9:B8 +|| MVKH (AES_Te-__encrypt),$TEA +|| ADD 0,KEY,$KPA +|| ADD 4,KEY,$KPB + .endif + LDW *$KPA++[2],$Te0[0] ; zero round key +|| LDW *$KPB++[2],$Te0[1] +|| MVK 60,A0 +|| ADD B0,$TEA,$TEA ; AES_Te + LDW *KEY[A0],B0 ; rounds +|| MVK 1024,A0 ; sizeof(AES_Te) + LDW *$KPA++[2],$Te0[2] +|| LDW *$KPB++[2],$Te0[3] +|| MV $TEA,$TEB + NOP + .if .BIG_ENDIAN + MV A9,$s[0] +|| MV A8,$s[1] +|| MV B9,$s[2] +|| MV B8,$s[3] + .else + MV A8,$s[0] +|| MV A9,$s[1] +|| MV B8,$s[2] +|| MV B9,$s[3] + .endif + XOR $Te0[0],$s[0],$s[0] +|| XOR $Te0[1],$s[1],$s[1] +|| LDW *$KPA++[2],$K[0] ; 1st round key +|| LDW *$KPB++[2],$K[1] + SUB B0,2,B0 + + SPLOOPD 13 +|| MVC B0,ILC +|| LDW *$KPA++[2],$K[2] +|| LDW *$KPB++[2],$K[3] +;;==================================================================== + EXTU $s[1],EXT1,24,$Te1[1] +|| EXTU $s[0],EXT3,24,$Te3[0] + LDW *${TEB}[$Te1[1]],$Te1[1] ; Te1[s1>>8], t0 +|| LDW *${TEA}[$Te3[0]],$Te3[0] ; Te3[s0>>24], t1 +|| XOR $s[2],$Te0[2],$s[2] ; modulo-scheduled +|| XOR $s[3],$Te0[3],$s[3] ; modulo-scheduled +|| EXTU $s[1],EXT3,24,$Te3[1] +|| EXTU $s[0],EXT1,24,$Te1[0] + LDW *${TEB}[$Te3[1]],$Te3[1] ; Te3[s1>>24], t2 +|| LDW *${TEA}[$Te1[0]],$Te1[0] ; Te1[s0>>8], t3 +|| EXTU $s[2],EXT2,24,$Te2[2] +|| EXTU $s[3],EXT2,24,$Te2[3] + LDW *${TEA}[$Te2[2]],$Te2[2] ; Te2[s2>>16], t0 +|| LDW *${TEB}[$Te2[3]],$Te2[3] ; Te2[s3>>16], t1 +|| EXTU $s[3],EXT3,24,$Te3[3] +|| EXTU $s[2],EXT1,24,$Te1[2] + LDW *${TEB}[$Te3[3]],$Te3[3] ; Te3[s3>>24], t0 +|| LDW *${TEA}[$Te1[2]],$Te1[2] ; Te1[s2>>8], t1 +|| EXTU $s[0],EXT2,24,$Te2[0] +|| EXTU $s[1],EXT2,24,$Te2[1] + LDW *${TEA}[$Te2[0]],$Te2[0] ; Te2[s0>>16], t2 +|| LDW *${TEB}[$Te2[1]],$Te2[1] ; Te2[s1>>16], t3 +|| EXTU $s[3],EXT1,24,$Te1[3] +|| EXTU $s[2],EXT3,24,$Te3[2] + LDW *${TEB}[$Te1[3]],$Te1[3] ; Te1[s3>>8], t2 +|| LDW *${TEA}[$Te3[2]],$Te3[2] ; Te3[s2>>24], t3 +|| ROTL $Te1[1],TBL1,$Te3[0] ; t0 +|| ROTL $Te3[0],TBL3,$Te1[1] ; t1 +|| EXTU $s[0],EXT0,24,$Te0[0] +|| EXTU $s[1],EXT0,24,$Te0[1] + LDW *${TEA}[$Te0[0]],$Te0[0] ; Te0[s0], t0 +|| LDW *${TEB}[$Te0[1]],$Te0[1] ; Te0[s1], t1 +|| ROTL $Te3[1],TBL3,$Te1[0] ; t2 +|| ROTL $Te1[0],TBL1,$Te3[1] ; t3 +|| EXTU $s[2],EXT0,24,$Te0[2] +|| EXTU $s[3],EXT0,24,$Te0[3] + LDW *${TEA}[$Te0[2]],$Te0[2] ; Te0[s2], t2 +|| LDW *${TEB}[$Te0[3]],$Te0[3] ; Te0[s3], t3 +|| ROTL $Te2[2],TBL2,$Te2[2] ; t0 +|| ROTL $Te2[3],TBL2,$Te2[3] ; t1 +|| XOR $K[0],$Te3[0],$s[0] +|| XOR $K[1],$Te1[1],$s[1] + ROTL $Te3[3],TBL3,$Te1[2] ; t0 +|| ROTL $Te1[2],TBL1,$Te3[3] ; t1 +|| XOR $K[2],$Te1[0],$s[2] +|| XOR $K[3],$Te3[1],$s[3] +|| LDW *$KPA++[2],$K[0] ; next round key +|| LDW *$KPB++[2],$K[1] + ROTL $Te2[0],TBL2,$Te2[0] ; t2 +|| ROTL $Te2[1],TBL2,$Te2[1] ; t3 +|| XOR $s[0],$Te2[2],$s[0] +|| XOR $s[1],$Te2[3],$s[1] +|| LDW *$KPA++[2],$K[2] +|| LDW *$KPB++[2],$K[3] + ROTL $Te1[3],TBL1,$Te3[2] ; t2 +|| ROTL $Te3[2],TBL3,$Te1[3] ; t3 +|| XOR $s[0],$Te1[2],$s[0] +|| XOR $s[1],$Te3[3],$s[1] + XOR $s[2],$Te2[0],$s[2] +|| XOR $s[3],$Te2[1],$s[3] +|| XOR $s[0],$Te0[0],$s[0] +|| XOR $s[1],$Te0[1],$s[1] + SPKERNEL +|| XOR.L $s[2],$Te3[2],$s[2] +|| XOR.L $s[3],$Te1[3],$s[3] +;;==================================================================== + ADD.D ${TEA},A0,${TEA} ; point to Te4 +|| ADD.D ${TEB},A0,${TEB} +|| EXTU $s[1],EXT1,24,$Te1[1] +|| EXTU $s[0],EXT3,24,$Te3[0] + LDBU *${TEB}[$Te1[1]],$Te1[1] ; Te1[s1>>8], t0 +|| LDBU *${TEA}[$Te3[0]],$Te3[0] ; Te3[s0>>24], t1 +|| XOR $s[2],$Te0[2],$s[2] ; modulo-scheduled +|| XOR $s[3],$Te0[3],$s[3] ; modulo-scheduled +|| EXTU $s[0],EXT0,24,$Te0[0] +|| EXTU $s[1],EXT0,24,$Te0[1] + LDBU *${TEA}[$Te0[0]],$Te0[0] ; Te0[s0], t0 +|| LDBU *${TEB}[$Te0[1]],$Te0[1] ; Te0[s1], t1 +|| EXTU $s[3],EXT3,24,$Te3[3] +|| EXTU $s[2],EXT1,24,$Te1[2] + LDBU *${TEB}[$Te3[3]],$Te3[3] ; Te3[s3>>24], t0 +|| LDBU *${TEA}[$Te1[2]],$Te1[2] ; Te1[s2>>8], t1 +|| EXTU $s[2],EXT2,24,$Te2[2] +|| EXTU $s[3],EXT2,24,$Te2[3] + LDBU *${TEA}[$Te2[2]],$Te2[2] ; Te2[s2>>16], t0 +|| LDBU *${TEB}[$Te2[3]],$Te2[3] ; Te2[s3>>16], t1 +|| EXTU $s[1],EXT3,24,$Te3[1] +|| EXTU $s[0],EXT1,24,$Te1[0] + LDBU *${TEB}[$Te3[1]],$Te3[1] ; Te3[s1>>24], t2 +|| LDBU *${TEA}[$Te1[0]],$Te1[0] ; Te1[s0>>8], t3 +|| EXTU $s[3],EXT1,24,$Te1[3] +|| EXTU $s[2],EXT3,24,$Te3[2] + LDBU *${TEB}[$Te1[3]],$Te1[3] ; Te1[s3>>8], t2 +|| LDBU *${TEA}[$Te3[2]],$Te3[2] ; Te3[s2>>24], t3 +|| EXTU $s[2],EXT0,24,$Te0[2] +|| EXTU $s[3],EXT0,24,$Te0[3] + LDBU *${TEA}[$Te0[2]],$Te0[2] ; Te0[s2], t2 +|| LDBU *${TEB}[$Te0[3]],$Te0[3] ; Te0[s3], t3 +|| EXTU $s[0],EXT2,24,$Te2[0] +|| EXTU $s[1],EXT2,24,$Te2[1] + LDBU *${TEA}[$Te2[0]],$Te2[0] ; Te2[s0>>16], t2 +|| LDBU *${TEB}[$Te2[1]],$Te2[1] ; Te2[s1>>16], t3 + + .if .BIG_ENDIAN + PACK2 $Te0[0],$Te1[1],$Te0[0] +|| PACK2 $Te0[1],$Te1[2],$Te0[1] + PACK2 $Te2[2],$Te3[3],$Te2[2] +|| PACK2 $Te2[3],$Te3[0],$Te2[3] + PACKL4 $Te0[0],$Te2[2],$Te0[0] +|| PACKL4 $Te0[1],$Te2[3],$Te0[1] + XOR $K[0],$Te0[0],$Te0[0] ; s[0] +|| XOR $K[1],$Te0[1],$Te0[1] ; s[1] + + PACK2 $Te0[2],$Te1[3],$Te0[2] +|| PACK2 $Te0[3],$Te1[0],$Te0[3] + PACK2 $Te2[0],$Te3[1],$Te2[0] +|| PACK2 $Te2[1],$Te3[2],$Te2[1] +|| BNOP RA + PACKL4 $Te0[2],$Te2[0],$Te0[2] +|| PACKL4 $Te0[3],$Te2[1],$Te0[3] + XOR $K[2],$Te0[2],$Te0[2] ; s[2] +|| XOR $K[3],$Te0[3],$Te0[3] ; s[3] + + MV $Te0[0],A9 +|| MV $Te0[1],A8 + MV $Te0[2],B9 +|| MV $Te0[3],B8 +|| [B2] STNDW A9:A8,*OUT++ + [B2] STNDW B9:B8,*OUT++ + .else + PACK2 $Te1[1],$Te0[0],$Te1[1] +|| PACK2 $Te1[2],$Te0[1],$Te1[2] + PACK2 $Te3[3],$Te2[2],$Te3[3] +|| PACK2 $Te3[0],$Te2[3],$Te3[0] + PACKL4 $Te3[3],$Te1[1],$Te1[1] +|| PACKL4 $Te3[0],$Te1[2],$Te1[2] + XOR $K[0],$Te1[1],$Te1[1] ; s[0] +|| XOR $K[1],$Te1[2],$Te1[2] ; s[1] + + PACK2 $Te1[3],$Te0[2],$Te1[3] +|| PACK2 $Te1[0],$Te0[3],$Te1[0] + PACK2 $Te3[1],$Te2[0],$Te3[1] +|| PACK2 $Te3[2],$Te2[1],$Te3[2] +|| BNOP RA + PACKL4 $Te3[1],$Te1[3],$Te1[3] +|| PACKL4 $Te3[2],$Te1[0],$Te1[0] + XOR $K[2],$Te1[3],$Te1[3] ; s[2] +|| XOR $K[3],$Te1[0],$Te1[0] ; s[3] + + MV $Te1[1],A8 +|| MV $Te1[2],A9 + MV $Te1[3],B8 +|| MV $Te1[0],B9 +|| [B2] STNDW A9:A8,*OUT++ + [B2] STNDW B9:B8,*OUT++ + .endif + .endasmfunc + + .global _AES_decrypt +_AES_decrypt: + .asmfunc + MVK 1,B2 +__decrypt: + .if __TI_EABI__ + [B2] LDNDW *INP++,A9:A8 ; load input +|| MVKL \$PCR_OFFSET(AES_Td,__decrypt),$TEA +|| ADDKPC __decrypt,B0 + [B2] LDNDW *INP++,B9:B8 +|| MVKH \$PCR_OFFSET(AES_Td,__decrypt),$TEA +|| ADD 0,KEY,$KPA +|| ADD 4,KEY,$KPB + .else + [B2] LDNDW *INP++,A9:A8 ; load input +|| MVKL (AES_Td-__decrypt),$TEA +|| ADDKPC __decrypt,B0 + [B2] LDNDW *INP++,B9:B8 +|| MVKH (AES_Td-__decrypt),$TEA +|| ADD 0,KEY,$KPA +|| ADD 4,KEY,$KPB + .endif + LDW *$KPA++[2],$Td0[0] ; zero round key +|| LDW *$KPB++[2],$Td0[1] +|| MVK 60,A0 +|| ADD B0,$TEA,$TEA ; AES_Td + LDW *KEY[A0],B0 ; rounds +|| MVK 1024,A0 ; sizeof(AES_Td) + LDW *$KPA++[2],$Td0[2] +|| LDW *$KPB++[2],$Td0[3] +|| MV $TEA,$TEB + NOP + .if .BIG_ENDIAN + MV A9,$s[0] +|| MV A8,$s[1] +|| MV B9,$s[2] +|| MV B8,$s[3] + .else + MV A8,$s[0] +|| MV A9,$s[1] +|| MV B8,$s[2] +|| MV B9,$s[3] + .endif + XOR $Td0[0],$s[0],$s[0] +|| XOR $Td0[1],$s[1],$s[1] +|| LDW *$KPA++[2],$K[0] ; 1st round key +|| LDW *$KPB++[2],$K[1] + SUB B0,2,B0 + + SPLOOPD 13 +|| MVC B0,ILC +|| LDW *$KPA++[2],$K[2] +|| LDW *$KPB++[2],$K[3] +;;==================================================================== + EXTU $s[1],EXT3,24,$Td3[1] +|| EXTU $s[0],EXT1,24,$Td1[0] + LDW *${TEB}[$Td3[1]],$Td3[1] ; Td3[s1>>24], t0 +|| LDW *${TEA}[$Td1[0]],$Td1[0] ; Td1[s0>>8], t1 +|| XOR $s[2],$Td0[2],$s[2] ; modulo-scheduled +|| XOR $s[3],$Td0[3],$s[3] ; modulo-scheduled +|| EXTU $s[1],EXT1,24,$Td1[1] +|| EXTU $s[0],EXT3,24,$Td3[0] + LDW *${TEB}[$Td1[1]],$Td1[1] ; Td1[s1>>8], t2 +|| LDW *${TEA}[$Td3[0]],$Td3[0] ; Td3[s0>>24], t3 +|| EXTU $s[2],EXT2,24,$Td2[2] +|| EXTU $s[3],EXT2,24,$Td2[3] + LDW *${TEA}[$Td2[2]],$Td2[2] ; Td2[s2>>16], t0 +|| LDW *${TEB}[$Td2[3]],$Td2[3] ; Td2[s3>>16], t1 +|| EXTU $s[3],EXT1,24,$Td1[3] +|| EXTU $s[2],EXT3,24,$Td3[2] + LDW *${TEB}[$Td1[3]],$Td1[3] ; Td1[s3>>8], t0 +|| LDW *${TEA}[$Td3[2]],$Td3[2] ; Td3[s2>>24], t1 +|| EXTU $s[0],EXT2,24,$Td2[0] +|| EXTU $s[1],EXT2,24,$Td2[1] + LDW *${TEA}[$Td2[0]],$Td2[0] ; Td2[s0>>16], t2 +|| LDW *${TEB}[$Td2[1]],$Td2[1] ; Td2[s1>>16], t3 +|| EXTU $s[3],EXT3,24,$Td3[3] +|| EXTU $s[2],EXT1,24,$Td1[2] + LDW *${TEB}[$Td3[3]],$Td3[3] ; Td3[s3>>24], t2 +|| LDW *${TEA}[$Td1[2]],$Td1[2] ; Td1[s2>>8], t3 +|| ROTL $Td3[1],TBL3,$Td1[0] ; t0 +|| ROTL $Td1[0],TBL1,$Td3[1] ; t1 +|| EXTU $s[0],EXT0,24,$Td0[0] +|| EXTU $s[1],EXT0,24,$Td0[1] + LDW *${TEA}[$Td0[0]],$Td0[0] ; Td0[s0], t0 +|| LDW *${TEB}[$Td0[1]],$Td0[1] ; Td0[s1], t1 +|| ROTL $Td1[1],TBL1,$Td3[0] ; t2 +|| ROTL $Td3[0],TBL3,$Td1[1] ; t3 +|| EXTU $s[2],EXT0,24,$Td0[2] +|| EXTU $s[3],EXT0,24,$Td0[3] + LDW *${TEA}[$Td0[2]],$Td0[2] ; Td0[s2], t2 +|| LDW *${TEB}[$Td0[3]],$Td0[3] ; Td0[s3], t3 +|| ROTL $Td2[2],TBL2,$Td2[2] ; t0 +|| ROTL $Td2[3],TBL2,$Td2[3] ; t1 +|| XOR $K[0],$Td1[0],$s[0] +|| XOR $K[1],$Td3[1],$s[1] + ROTL $Td1[3],TBL1,$Td3[2] ; t0 +|| ROTL $Td3[2],TBL3,$Td1[3] ; t1 +|| XOR $K[2],$Td3[0],$s[2] +|| XOR $K[3],$Td1[1],$s[3] +|| LDW *$KPA++[2],$K[0] ; next round key +|| LDW *$KPB++[2],$K[1] + ROTL $Td2[0],TBL2,$Td2[0] ; t2 +|| ROTL $Td2[1],TBL2,$Td2[1] ; t3 +|| XOR $s[0],$Td2[2],$s[0] +|| XOR $s[1],$Td2[3],$s[1] +|| LDW *$KPA++[2],$K[2] +|| LDW *$KPB++[2],$K[3] + ROTL $Td3[3],TBL3,$Td1[2] ; t2 +|| ROTL $Td1[2],TBL1,$Td3[3] ; t3 +|| XOR $s[0],$Td3[2],$s[0] +|| XOR $s[1],$Td1[3],$s[1] + XOR $s[2],$Td2[0],$s[2] +|| XOR $s[3],$Td2[1],$s[3] +|| XOR $s[0],$Td0[0],$s[0] +|| XOR $s[1],$Td0[1],$s[1] + SPKERNEL +|| XOR.L $s[2],$Td1[2],$s[2] +|| XOR.L $s[3],$Td3[3],$s[3] +;;==================================================================== + ADD.D ${TEA},A0,${TEA} ; point to Td4 +|| ADD.D ${TEB},A0,${TEB} +|| EXTU $s[1],EXT3,24,$Td3[1] +|| EXTU $s[0],EXT1,24,$Td1[0] + LDBU *${TEB}[$Td3[1]],$Td3[1] ; Td3[s1>>24], t0 +|| LDBU *${TEA}[$Td1[0]],$Td1[0] ; Td1[s0>>8], t1 +|| XOR $s[2],$Td0[2],$s[2] ; modulo-scheduled +|| XOR $s[3],$Td0[3],$s[3] ; modulo-scheduled +|| EXTU $s[0],EXT0,24,$Td0[0] +|| EXTU $s[1],EXT0,24,$Td0[1] + LDBU *${TEA}[$Td0[0]],$Td0[0] ; Td0[s0], t0 +|| LDBU *${TEB}[$Td0[1]],$Td0[1] ; Td0[s1], t1 +|| EXTU $s[2],EXT2,24,$Td2[2] +|| EXTU $s[3],EXT2,24,$Td2[3] + LDBU *${TEA}[$Td2[2]],$Td2[2] ; Td2[s2>>16], t0 +|| LDBU *${TEB}[$Td2[3]],$Td2[3] ; Td2[s3>>16], t1 +|| EXTU $s[3],EXT1,24,$Td1[3] +|| EXTU $s[2],EXT3,24,$Td3[2] + LDBU *${TEB}[$Td1[3]],$Td1[3] ; Td1[s3>>8], t0 +|| LDBU *${TEA}[$Td3[2]],$Td3[2] ; Td3[s2>>24], t1 +|| EXTU $s[1],EXT1,24,$Td1[1] +|| EXTU $s[0],EXT3,24,$Td3[0] + LDBU *${TEB}[$Td1[1]],$Td1[1] ; Td1[s1>>8], t2 +|| LDBU *${TEA}[$Td3[0]],$Td3[0] ; Td3[s0>>24], t3 +|| EXTU $s[0],EXT2,24,$Td2[0] +|| EXTU $s[1],EXT2,24,$Td2[1] + LDBU *${TEA}[$Td2[0]],$Td2[0] ; Td2[s0>>16], t2 +|| LDBU *${TEB}[$Td2[1]],$Td2[1] ; Td2[s1>>16], t3 +|| EXTU $s[3],EXT3,24,$Td3[3] +|| EXTU $s[2],EXT1,24,$Td1[2] + LDBU *${TEB}[$Td3[3]],$Td3[3] ; Td3[s3>>24], t2 +|| LDBU *${TEA}[$Td1[2]],$Td1[2] ; Td1[s2>>8], t3 +|| EXTU $s[2],EXT0,24,$Td0[2] +|| EXTU $s[3],EXT0,24,$Td0[3] + LDBU *${TEA}[$Td0[2]],$Td0[2] ; Td0[s2], t2 +|| LDBU *${TEB}[$Td0[3]],$Td0[3] ; Td0[s3], t3 + + .if .BIG_ENDIAN + PACK2 $Td0[0],$Td1[3],$Td0[0] +|| PACK2 $Td0[1],$Td1[0],$Td0[1] + PACK2 $Td2[2],$Td3[1],$Td2[2] +|| PACK2 $Td2[3],$Td3[2],$Td2[3] + PACKL4 $Td0[0],$Td2[2],$Td0[0] +|| PACKL4 $Td0[1],$Td2[3],$Td0[1] + XOR $K[0],$Td0[0],$Td0[0] ; s[0] +|| XOR $K[1],$Td0[1],$Td0[1] ; s[1] + + PACK2 $Td0[2],$Td1[1],$Td0[2] +|| PACK2 $Td0[3],$Td1[2],$Td0[3] + PACK2 $Td2[0],$Td3[3],$Td2[0] +|| PACK2 $Td2[1],$Td3[0],$Td2[1] +|| BNOP RA + PACKL4 $Td0[2],$Td2[0],$Td0[2] +|| PACKL4 $Td0[3],$Td2[1],$Td0[3] + XOR $K[2],$Td0[2],$Td0[2] ; s[2] +|| XOR $K[3],$Td0[3],$Td0[3] ; s[3] + + MV $Td0[0],A9 +|| MV $Td0[1],A8 + MV $Td0[2],B9 +|| MV $Td0[3],B8 +|| [B2] STNDW A9:A8,*OUT++ + [B2] STNDW B9:B8,*OUT++ + .else + PACK2 $Td1[3],$Td0[0],$Td1[3] +|| PACK2 $Td1[0],$Td0[1],$Td1[0] + PACK2 $Td3[1],$Td2[2],$Td3[1] +|| PACK2 $Td3[2],$Td2[3],$Td3[2] + PACKL4 $Td3[1],$Td1[3],$Td1[3] +|| PACKL4 $Td3[2],$Td1[0],$Td1[0] + XOR $K[0],$Td1[3],$Td1[3] ; s[0] +|| XOR $K[1],$Td1[0],$Td1[0] ; s[1] + + PACK2 $Td1[1],$Td0[2],$Td1[1] +|| PACK2 $Td1[2],$Td0[3],$Td1[2] + PACK2 $Td3[3],$Td2[0],$Td3[3] +|| PACK2 $Td3[0],$Td2[1],$Td3[0] +|| BNOP RA + PACKL4 $Td3[3],$Td1[1],$Td1[1] +|| PACKL4 $Td3[0],$Td1[2],$Td1[2] + XOR $K[2],$Td1[1],$Td1[1] ; s[2] +|| XOR $K[3],$Td1[2],$Td1[2] ; s[3] + + MV $Td1[3],A8 +|| MV $Td1[0],A9 + MV $Td1[1],B8 +|| MV $Td1[2],B9 +|| [B2] STNDW A9:A8,*OUT++ + [B2] STNDW B9:B8,*OUT++ + .endif + .endasmfunc +___ +{ +my @K=(@K,@s); # extended key +my @Te4=map("B$_",(16..19)); + +my @Kx9=@Te0; # used in AES_set_decrypt_key +my @KxB=@Te1; +my @KxD=@Te2; +my @KxE=@Te3; + +$code.=<<___; + .asg OUT,BITS + + .global _AES_set_encrypt_key +_AES_set_encrypt_key: +__set_encrypt_key: + .asmfunc + MV INP,A0 +|| SHRU BITS,5,BITS ; 128-192-256 -> 4-6-8 +|| MV KEY,A1 + [!A0] B RA +||[!A0] MVK -1,RET +||[!A0] MVK 1,A1 ; only one B RA + [!A1] B RA +||[!A1] MVK -1,RET +||[!A1] MVK 0,A0 +|| MVK 0,B0 +|| MVK 0,A1 + [A0] LDNDW *INP++,A9:A8 +|| [A0] CMPEQ 4,BITS,B0 +|| [A0] CMPLT 3,BITS,A1 + [B0] B key128? +|| [A1] LDNDW *INP++,B9:B8 +|| [A0] CMPEQ 6,BITS,B0 +|| [A0] CMPLT 5,BITS,A1 + [B0] B key192? +|| [A1] LDNDW *INP++,B17:B16 +|| [A0] CMPEQ 8,BITS,B0 +|| [A0] CMPLT 7,BITS,A1 + [B0] B key256? +|| [A1] LDNDW *INP++,B19:B18 + + .if __TI_EABI__ + [A0] ADD 0,KEY,$KPA +|| [A0] ADD 4,KEY,$KPB +|| [A0] MVKL \$PCR_OFFSET(AES_Te4,__set_encrypt_key),$TEA +|| [A0] ADDKPC __set_encrypt_key,B6 + [A0] MVKH \$PCR_OFFSET(AES_Te4,__set_encrypt_key),$TEA + [A0] ADD B6,$TEA,$TEA ; AES_Te4 + .else + [A0] ADD 0,KEY,$KPA +|| [A0] ADD 4,KEY,$KPB +|| [A0] MVKL (AES_Te4-__set_encrypt_key),$TEA +|| [A0] ADDKPC __set_encrypt_key,B6 + [A0] MVKH (AES_Te4-__set_encrypt_key),$TEA + [A0] ADD B6,$TEA,$TEA ; AES_Te4 + .endif + NOP + NOP + + BNOP RA,5 +|| MVK -2,RET ; unknown bit length +|| MVK 0,B0 ; redundant +;;==================================================================== +;;==================================================================== +key128?: + .if .BIG_ENDIAN + MV A9,$K[0] +|| MV A8,$K[1] +|| MV B9,$Te4[2] +|| MV B8,$K[3] + .else + MV A8,$K[0] +|| MV A9,$K[1] +|| MV B8,$Te4[2] +|| MV B9,$K[3] + .endif + + MVK 256,A0 +|| MVK 9,B0 + + SPLOOPD 14 +|| MVC B0,ILC +|| MV $TEA,$TEB +|| ADD $TEA,A0,A30 ; rcon +;;==================================================================== + LDW *A30++[1],A31 ; rcon[i] +|| MV $Te4[2],$K[2] +|| EXTU $K[3],EXT1,24,$Te4[0] + LDBU *${TEB}[$Te4[0]],$Te4[0] +|| MV $K[3],A0 +|| EXTU $K[3],EXT2,24,$Te4[1] + LDBU *${TEB}[$Te4[1]],$Te4[1] +|| EXTU A0,EXT3,24,A0 +|| EXTU $K[3],EXT0,24,$Te4[3] + .if .BIG_ENDIAN + LDBU *${TEA}[A0],$Te4[3] +|| LDBU *${TEB}[$Te4[3]],A0 + .else + LDBU *${TEA}[A0],A0 +|| LDBU *${TEB}[$Te4[3]],$Te4[3] + .endif + + STW $K[0],*$KPA++[2] +|| STW $K[1],*$KPB++[2] + STW $K[2],*$KPA++[2] +|| STW $K[3],*$KPB++[2] + + XOR A31,$K[0],$K[0] ; ^=rcon[i] + .if .BIG_ENDIAN + PACK2 $Te4[0],$Te4[1],$Te4[1] + PACK2 $Te4[3],A0,$Te4[3] + PACKL4 $Te4[1],$Te4[3],$Te4[3] + .else + PACK2 $Te4[1],$Te4[0],$Te4[1] + PACK2 $Te4[3],A0,$Te4[3] + PACKL4 $Te4[3],$Te4[1],$Te4[3] + .endif + XOR $Te4[3],$K[0],$Te4[0] ; K[0] + XOR $Te4[0],$K[1],$K[1] ; K[1] + MV $Te4[0],$K[0] +|| XOR $K[1],$K[2],$Te4[2] ; K[2] + XOR $Te4[2],$K[3],$K[3] ; K[3] + SPKERNEL +;;==================================================================== + BNOP RA + MV $Te4[2],$K[2] +|| STW $K[0],*$KPA++[2] +|| STW $K[1],*$KPB++[2] + STW $K[2],*$KPA++[2] +|| STW $K[3],*$KPB++[2] + MVK 10,B0 ; rounds + STW B0,*++${KPB}[15] + MVK 0,RET +;;==================================================================== +;;==================================================================== +key192?: + .if .BIG_ENDIAN + MV A9,$K[0] +|| MV A8,$K[1] +|| MV B9,$K[2] +|| MV B8,$K[3] + MV B17,$Te4[2] +|| MV B16,$K[5] + .else + MV A8,$K[0] +|| MV A9,$K[1] +|| MV B8,$K[2] +|| MV B9,$K[3] + MV B16,$Te4[2] +|| MV B17,$K[5] + .endif + + MVK 256,A0 +|| MVK 6,B0 + MV $TEA,$TEB +|| ADD $TEA,A0,A30 ; rcon +;;==================================================================== +loop192?: + LDW *A30++[1],A31 ; rcon[i] +|| MV $Te4[2],$K[4] +|| EXTU $K[5],EXT1,24,$Te4[0] + LDBU *${TEB}[$Te4[0]],$Te4[0] +|| MV $K[5],A0 +|| EXTU $K[5],EXT2,24,$Te4[1] + LDBU *${TEB}[$Te4[1]],$Te4[1] +|| EXTU A0,EXT3,24,A0 +|| EXTU $K[5],EXT0,24,$Te4[3] + .if .BIG_ENDIAN + LDBU *${TEA}[A0],$Te4[3] +|| LDBU *${TEB}[$Te4[3]],A0 + .else + LDBU *${TEA}[A0],A0 +|| LDBU *${TEB}[$Te4[3]],$Te4[3] + .endif + + STW $K[0],*$KPA++[2] +|| STW $K[1],*$KPB++[2] + STW $K[2],*$KPA++[2] +|| STW $K[3],*$KPB++[2] + STW $K[4],*$KPA++[2] +|| STW $K[5],*$KPB++[2] + + XOR A31,$K[0],$K[0] ; ^=rcon[i] + .if .BIG_ENDIAN + PACK2 $Te4[0],$Te4[1],$Te4[1] +|| PACK2 $Te4[3],A0,$Te4[3] + PACKL4 $Te4[1],$Te4[3],$Te4[3] + .else + PACK2 $Te4[1],$Te4[0],$Te4[1] +|| PACK2 $Te4[3],A0,$Te4[3] + PACKL4 $Te4[3],$Te4[1],$Te4[3] + .endif + BDEC loop192?,B0 +|| XOR $Te4[3],$K[0],$Te4[0] ; K[0] + XOR $Te4[0],$K[1],$K[1] ; K[1] + MV $Te4[0],$K[0] +|| XOR $K[1],$K[2],$Te4[2] ; K[2] + XOR $Te4[2],$K[3],$K[3] ; K[3] + MV $Te4[2],$K[2] +|| XOR $K[3],$K[4],$Te4[2] ; K[4] + XOR $Te4[2],$K[5],$K[5] ; K[5] +;;==================================================================== + BNOP RA + STW $K[0],*$KPA++[2] +|| STW $K[1],*$KPB++[2] + STW $K[2],*$KPA++[2] +|| STW $K[3],*$KPB++[2] + MVK 12,B0 ; rounds + STW B0,*++${KPB}[7] + MVK 0,RET +;;==================================================================== +;;==================================================================== +key256?: + .if .BIG_ENDIAN + MV A9,$K[0] +|| MV A8,$K[1] +|| MV B9,$K[2] +|| MV B8,$K[3] + MV B17,$K[4] +|| MV B16,$K[5] +|| MV B19,$Te4[2] +|| MV B18,$K[7] + .else + MV A8,$K[0] +|| MV A9,$K[1] +|| MV B8,$K[2] +|| MV B9,$K[3] + MV B16,$K[4] +|| MV B17,$K[5] +|| MV B18,$Te4[2] +|| MV B19,$K[7] + .endif + + MVK 256,A0 +|| MVK 6,B0 + MV $TEA,$TEB +|| ADD $TEA,A0,A30 ; rcon +;;==================================================================== +loop256?: + LDW *A30++[1],A31 ; rcon[i] +|| MV $Te4[2],$K[6] +|| EXTU $K[7],EXT1,24,$Te4[0] + LDBU *${TEB}[$Te4[0]],$Te4[0] +|| MV $K[7],A0 +|| EXTU $K[7],EXT2,24,$Te4[1] + LDBU *${TEB}[$Te4[1]],$Te4[1] +|| EXTU A0,EXT3,24,A0 +|| EXTU $K[7],EXT0,24,$Te4[3] + .if .BIG_ENDIAN + LDBU *${TEA}[A0],$Te4[3] +|| LDBU *${TEB}[$Te4[3]],A0 + .else + LDBU *${TEA}[A0],A0 +|| LDBU *${TEB}[$Te4[3]],$Te4[3] + .endif + + STW $K[0],*$KPA++[2] +|| STW $K[1],*$KPB++[2] + STW $K[2],*$KPA++[2] +|| STW $K[3],*$KPB++[2] + STW $K[4],*$KPA++[2] +|| STW $K[5],*$KPB++[2] + STW $K[6],*$KPA++[2] +|| STW $K[7],*$KPB++[2] +|| XOR A31,$K[0],$K[0] ; ^=rcon[i] + .if .BIG_ENDIAN + PACK2 $Te4[0],$Te4[1],$Te4[1] +|| PACK2 $Te4[3],A0,$Te4[3] + PACKL4 $Te4[1],$Te4[3],$Te4[3] +||[!B0] B done256? + .else + PACK2 $Te4[1],$Te4[0],$Te4[1] +|| PACK2 $Te4[3],A0,$Te4[3] + PACKL4 $Te4[3],$Te4[1],$Te4[3] +||[!B0] B done256? + .endif + XOR $Te4[3],$K[0],$Te4[0] ; K[0] + XOR $Te4[0],$K[1],$K[1] ; K[1] + MV $Te4[0],$K[0] +|| XOR $K[1],$K[2],$Te4[2] ; K[2] + XOR $Te4[2],$K[3],$K[3] ; K[3] + + MV $Te4[2],$K[2] +|| [B0] EXTU $K[3],EXT0,24,$Te4[0] +|| [B0] SUB B0,1,B0 + LDBU *${TEB}[$Te4[0]],$Te4[0] +|| MV $K[3],A0 +|| EXTU $K[3],EXT1,24,$Te4[1] + LDBU *${TEB}[$Te4[1]],$Te4[1] +|| EXTU A0,EXT2,24,A0 +|| EXTU $K[3],EXT3,24,$Te4[3] + + .if .BIG_ENDIAN + LDBU *${TEA}[A0],$Te4[3] +|| LDBU *${TEB}[$Te4[3]],A0 + NOP 3 + PACK2 $Te4[0],$Te4[1],$Te4[1] + PACK2 $Te4[3],A0,$Te4[3] +|| B loop256? + PACKL4 $Te4[1],$Te4[3],$Te4[3] + .else + LDBU *${TEA}[A0],A0 +|| LDBU *${TEB}[$Te4[3]],$Te4[3] + NOP 3 + PACK2 $Te4[1],$Te4[0],$Te4[1] + PACK2 $Te4[3],A0,$Te4[3] +|| B loop256? + PACKL4 $Te4[3],$Te4[1],$Te4[3] + .endif + + XOR $Te4[3],$K[4],$Te4[0] ; K[4] + XOR $Te4[0],$K[5],$K[5] ; K[5] + MV $Te4[0],$K[4] +|| XOR $K[5],$K[6],$Te4[2] ; K[6] + XOR $Te4[2],$K[7],$K[7] ; K[7] +;;==================================================================== +done256?: + BNOP RA + STW $K[0],*$KPA++[2] +|| STW $K[1],*$KPB++[2] + STW $K[2],*$KPA++[2] +|| STW $K[3],*$KPB++[2] + MVK 14,B0 ; rounds + STW B0,*--${KPB}[1] + MVK 0,RET + .endasmfunc + + .global _AES_set_decrypt_key +_AES_set_decrypt_key: + .asmfunc + B __set_encrypt_key ; guarantee local call + MV KEY,B30 ; B30 is not modified + MV RA, B31 ; B31 is not modified + ADDKPC ret?,RA,2 +ret?: ; B0 holds rounds or zero + [!B0] BNOP B31 ; return if zero + [B0] SHL B0,4,A0 ; offset to last round key + [B0] SHRU B0,1,B1 + [B0] SUB B1,1,B1 + [B0] MVK 0x0000001B,B3 ; AES polynomial + [B0] MVKH 0x07000000,B3 + + SPLOOPD 9 ; flip round keys +|| MVC B1,ILC +|| MV B30,$KPA +|| ADD B30,A0,$KPB +|| MVK 16,A0 ; sizeof(round key) +;;==================================================================== + LDW *${KPA}[0],A16 +|| LDW *${KPB}[0],B16 + LDW *${KPA}[1],A17 +|| LDW *${KPB}[1],B17 + LDW *${KPA}[2],A18 +|| LDW *${KPB}[2],B18 + LDW *${KPA}[3],A19 +|| ADD $KPA,A0,$KPA +|| LDW *${KPB}[3],B19 +|| SUB $KPB,A0,$KPB + NOP + STW B16,*${KPA}[-4] +|| STW A16,*${KPB}[4] + STW B17,*${KPA}[-3] +|| STW A17,*${KPB}[5] + STW B18,*${KPA}[-2] +|| STW A18,*${KPB}[6] + STW B19,*${KPA}[-1] +|| STW A19,*${KPB}[7] + SPKERNEL +;;==================================================================== + SUB B0,1,B0 ; skip last round +|| ADD B30,A0,$KPA ; skip first round +|| ADD B30,A0,$KPB +|| MVC GFPGFR,B30 ; save GFPGFR + LDW *${KPA}[0],$K[0] +|| LDW *${KPB}[1],$K[1] +|| MVC B3,GFPGFR + LDW *${KPA}[2],$K[2] +|| LDW *${KPB}[3],$K[3] + MVK 0x00000909,A24 +|| MVK 0x00000B0B,B24 + MVKH 0x09090000,A24 +|| MVKH 0x0B0B0000,B24 + MVC B0,ILC +|| SUB B0,1,B0 + + GMPY4 $K[0],A24,$Kx9[0] ; ·0x09 +|| GMPY4 $K[1],A24,$Kx9[1] +|| MVK 0x00000D0D,A25 +|| MVK 0x00000E0E,B25 + GMPY4 $K[2],A24,$Kx9[2] +|| GMPY4 $K[3],A24,$Kx9[3] +|| MVKH 0x0D0D0000,A25 +|| MVKH 0x0E0E0000,B25 + + GMPY4 $K[0],B24,$KxB[0] ; ·0x0B +|| GMPY4 $K[1],B24,$KxB[1] + GMPY4 $K[2],B24,$KxB[2] +|| GMPY4 $K[3],B24,$KxB[3] + + SPLOOP 11 ; InvMixColumns +;;==================================================================== + GMPY4 $K[0],A25,$KxD[0] ; ·0x0D +|| GMPY4 $K[1],A25,$KxD[1] +|| SWAP2 $Kx9[0],$Kx9[0] ; rotate by 16 +|| SWAP2 $Kx9[1],$Kx9[1] +|| MV $K[0],$s[0] ; this or DINT +|| MV $K[1],$s[1] +|| [B0] LDW *${KPA}[4],$K[0] +|| [B0] LDW *${KPB}[5],$K[1] + GMPY4 $K[2],A25,$KxD[2] +|| GMPY4 $K[3],A25,$KxD[3] +|| SWAP2 $Kx9[2],$Kx9[2] +|| SWAP2 $Kx9[3],$Kx9[3] +|| MV $K[2],$s[2] +|| MV $K[3],$s[3] +|| [B0] LDW *${KPA}[6],$K[2] +|| [B0] LDW *${KPB}[7],$K[3] + + GMPY4 $s[0],B25,$KxE[0] ; ·0x0E +|| GMPY4 $s[1],B25,$KxE[1] +|| XOR $Kx9[0],$KxB[0],$KxB[0] +|| XOR $Kx9[1],$KxB[1],$KxB[1] + GMPY4 $s[2],B25,$KxE[2] +|| GMPY4 $s[3],B25,$KxE[3] +|| XOR $Kx9[2],$KxB[2],$KxB[2] +|| XOR $Kx9[3],$KxB[3],$KxB[3] + + ROTL $KxB[0],TBL3,$KxB[0] +|| ROTL $KxB[1],TBL3,$KxB[1] +|| SWAP2 $KxD[0],$KxD[0] ; rotate by 16 +|| SWAP2 $KxD[1],$KxD[1] + ROTL $KxB[2],TBL3,$KxB[2] +|| ROTL $KxB[3],TBL3,$KxB[3] +|| SWAP2 $KxD[2],$KxD[2] +|| SWAP2 $KxD[3],$KxD[3] + + XOR $KxE[0],$KxD[0],$KxE[0] +|| XOR $KxE[1],$KxD[1],$KxE[1] +|| [B0] GMPY4 $K[0],A24,$Kx9[0] ; ·0x09 +|| [B0] GMPY4 $K[1],A24,$Kx9[1] +|| ADDAW $KPA,4,$KPA + XOR $KxE[2],$KxD[2],$KxE[2] +|| XOR $KxE[3],$KxD[3],$KxE[3] +|| [B0] GMPY4 $K[2],A24,$Kx9[2] +|| [B0] GMPY4 $K[3],A24,$Kx9[3] +|| ADDAW $KPB,4,$KPB + + XOR $KxB[0],$KxE[0],$KxE[0] +|| XOR $KxB[1],$KxE[1],$KxE[1] +|| [B0] GMPY4 $K[0],B24,$KxB[0] ; ·0x0B +|| [B0] GMPY4 $K[1],B24,$KxB[1] + XOR $KxB[2],$KxE[2],$KxE[2] +|| XOR $KxB[3],$KxE[3],$KxE[3] +|| [B0] GMPY4 $K[2],B24,$KxB[2] +|| [B0] GMPY4 $K[3],B24,$KxB[3] +|| STW $KxE[0],*${KPA}[-4] +|| STW $KxE[1],*${KPB}[-3] + STW $KxE[2],*${KPA}[-2] +|| STW $KxE[3],*${KPB}[-1] +|| [B0] SUB B0,1,B0 + SPKERNEL +;;==================================================================== + BNOP B31,3 + MVC B30,GFPGFR ; restore GFPGFR(*) + MVK 0,RET + .endasmfunc +___ +# (*) Even though ABI doesn't specify GFPGFR as non-volatile, there +# are code samples out there that *assume* its default value. +} +{ +my ($inp,$out,$blocks,$key,$ivp)=("A4","B4","A6","B6","A8"); +$code.=<<___; + .global _AES_ctr32_encrypt +_AES_ctr32_encrypt: + .asmfunc + LDNDW *${ivp}[0],A31:A30 ; load counter value +|| MV $blocks,A2 ; reassign $blocks +|| DMV RA,$key,B27:B26 ; reassign RA and $key + LDNDW *${ivp}[1],B31:B30 +|| MVK 0,B2 ; don't let __encrypt load input +|| MVK 0,A1 ; and postpone writing output + .if .BIG_ENDIAN + NOP + .else + NOP 4 + SWAP2 B31,B31 ; keep least significant 32 bits + SWAP4 B31,B31 ; in host byte order + .endif +ctr32_loop?: + [A2] BNOP __encrypt +|| [A1] XOR A29,A9,A9 ; input^Ek(counter) +|| [A1] XOR A28,A8,A8 +|| [A2] LDNDW *INP++,A29:A28 ; load input + [!A2] BNOP B27 ; return +|| [A1] XOR B29,B9,B9 +|| [A1] XOR B28,B8,B8 +|| [A2] LDNDW *INP++,B29:B28 + .if .BIG_ENDIAN + [A1] STNDW A9:A8,*OUT++ ; save output +|| [A2] DMV A31,A30,A9:A8 ; pass counter value to __encrypt + [A1] STNDW B9:B8,*OUT++ +|| [A2] DMV B31,B30,B9:B8 +|| [A2] ADD B30,1,B30 ; counter++ + .else + [A1] STNDW A9:A8,*OUT++ ; save output +|| [A2] DMV A31,A30,A9:A8 +|| [A2] SWAP2 B31,B0 +|| [A2] ADD B31,1,B31 ; counter++ + [A1] STNDW B9:B8,*OUT++ +|| [A2] MV B30,B8 +|| [A2] SWAP4 B0,B9 + .endif + [A2] ADDKPC ctr32_loop?,RA ; return to ctr32_loop? +|| [A2] MV B26,KEY ; pass $key +|| [A2] SUB A2,1,A2 ; $blocks-- +||[!A1] MVK 1,A1 + NOP + NOP + .endasmfunc +___ +} +# Tables are kept in endian-neutral manner +$code.=<<___; + .if __TI_EABI__ + .sect ".text:aes_asm.const" + .else + .sect ".const:aes_asm" + .endif + .align 128 +AES_Te: + .byte 0xc6,0x63,0x63,0xa5, 0xf8,0x7c,0x7c,0x84 + .byte 0xee,0x77,0x77,0x99, 0xf6,0x7b,0x7b,0x8d + .byte 0xff,0xf2,0xf2,0x0d, 0xd6,0x6b,0x6b,0xbd + .byte 0xde,0x6f,0x6f,0xb1, 0x91,0xc5,0xc5,0x54 + .byte 0x60,0x30,0x30,0x50, 0x02,0x01,0x01,0x03 + .byte 0xce,0x67,0x67,0xa9, 0x56,0x2b,0x2b,0x7d + .byte 0xe7,0xfe,0xfe,0x19, 0xb5,0xd7,0xd7,0x62 + .byte 0x4d,0xab,0xab,0xe6, 0xec,0x76,0x76,0x9a + .byte 0x8f,0xca,0xca,0x45, 0x1f,0x82,0x82,0x9d + .byte 0x89,0xc9,0xc9,0x40, 0xfa,0x7d,0x7d,0x87 + .byte 0xef,0xfa,0xfa,0x15, 0xb2,0x59,0x59,0xeb + .byte 0x8e,0x47,0x47,0xc9, 0xfb,0xf0,0xf0,0x0b + .byte 0x41,0xad,0xad,0xec, 0xb3,0xd4,0xd4,0x67 + .byte 0x5f,0xa2,0xa2,0xfd, 0x45,0xaf,0xaf,0xea + .byte 0x23,0x9c,0x9c,0xbf, 0x53,0xa4,0xa4,0xf7 + .byte 0xe4,0x72,0x72,0x96, 0x9b,0xc0,0xc0,0x5b + .byte 0x75,0xb7,0xb7,0xc2, 0xe1,0xfd,0xfd,0x1c + .byte 0x3d,0x93,0x93,0xae, 0x4c,0x26,0x26,0x6a + .byte 0x6c,0x36,0x36,0x5a, 0x7e,0x3f,0x3f,0x41 + .byte 0xf5,0xf7,0xf7,0x02, 0x83,0xcc,0xcc,0x4f + .byte 0x68,0x34,0x34,0x5c, 0x51,0xa5,0xa5,0xf4 + .byte 0xd1,0xe5,0xe5,0x34, 0xf9,0xf1,0xf1,0x08 + .byte 0xe2,0x71,0x71,0x93, 0xab,0xd8,0xd8,0x73 + .byte 0x62,0x31,0x31,0x53, 0x2a,0x15,0x15,0x3f + .byte 0x08,0x04,0x04,0x0c, 0x95,0xc7,0xc7,0x52 + .byte 0x46,0x23,0x23,0x65, 0x9d,0xc3,0xc3,0x5e + .byte 0x30,0x18,0x18,0x28, 0x37,0x96,0x96,0xa1 + .byte 0x0a,0x05,0x05,0x0f, 0x2f,0x9a,0x9a,0xb5 + .byte 0x0e,0x07,0x07,0x09, 0x24,0x12,0x12,0x36 + .byte 0x1b,0x80,0x80,0x9b, 0xdf,0xe2,0xe2,0x3d + .byte 0xcd,0xeb,0xeb,0x26, 0x4e,0x27,0x27,0x69 + .byte 0x7f,0xb2,0xb2,0xcd, 0xea,0x75,0x75,0x9f + .byte 0x12,0x09,0x09,0x1b, 0x1d,0x83,0x83,0x9e + .byte 0x58,0x2c,0x2c,0x74, 0x34,0x1a,0x1a,0x2e + .byte 0x36,0x1b,0x1b,0x2d, 0xdc,0x6e,0x6e,0xb2 + .byte 0xb4,0x5a,0x5a,0xee, 0x5b,0xa0,0xa0,0xfb + .byte 0xa4,0x52,0x52,0xf6, 0x76,0x3b,0x3b,0x4d + .byte 0xb7,0xd6,0xd6,0x61, 0x7d,0xb3,0xb3,0xce + .byte 0x52,0x29,0x29,0x7b, 0xdd,0xe3,0xe3,0x3e + .byte 0x5e,0x2f,0x2f,0x71, 0x13,0x84,0x84,0x97 + .byte 0xa6,0x53,0x53,0xf5, 0xb9,0xd1,0xd1,0x68 + .byte 0x00,0x00,0x00,0x00, 0xc1,0xed,0xed,0x2c + .byte 0x40,0x20,0x20,0x60, 0xe3,0xfc,0xfc,0x1f + .byte 0x79,0xb1,0xb1,0xc8, 0xb6,0x5b,0x5b,0xed + .byte 0xd4,0x6a,0x6a,0xbe, 0x8d,0xcb,0xcb,0x46 + .byte 0x67,0xbe,0xbe,0xd9, 0x72,0x39,0x39,0x4b + .byte 0x94,0x4a,0x4a,0xde, 0x98,0x4c,0x4c,0xd4 + .byte 0xb0,0x58,0x58,0xe8, 0x85,0xcf,0xcf,0x4a + .byte 0xbb,0xd0,0xd0,0x6b, 0xc5,0xef,0xef,0x2a + .byte 0x4f,0xaa,0xaa,0xe5, 0xed,0xfb,0xfb,0x16 + .byte 0x86,0x43,0x43,0xc5, 0x9a,0x4d,0x4d,0xd7 + .byte 0x66,0x33,0x33,0x55, 0x11,0x85,0x85,0x94 + .byte 0x8a,0x45,0x45,0xcf, 0xe9,0xf9,0xf9,0x10 + .byte 0x04,0x02,0x02,0x06, 0xfe,0x7f,0x7f,0x81 + .byte 0xa0,0x50,0x50,0xf0, 0x78,0x3c,0x3c,0x44 + .byte 0x25,0x9f,0x9f,0xba, 0x4b,0xa8,0xa8,0xe3 + .byte 0xa2,0x51,0x51,0xf3, 0x5d,0xa3,0xa3,0xfe + .byte 0x80,0x40,0x40,0xc0, 0x05,0x8f,0x8f,0x8a + .byte 0x3f,0x92,0x92,0xad, 0x21,0x9d,0x9d,0xbc + .byte 0x70,0x38,0x38,0x48, 0xf1,0xf5,0xf5,0x04 + .byte 0x63,0xbc,0xbc,0xdf, 0x77,0xb6,0xb6,0xc1 + .byte 0xaf,0xda,0xda,0x75, 0x42,0x21,0x21,0x63 + .byte 0x20,0x10,0x10,0x30, 0xe5,0xff,0xff,0x1a + .byte 0xfd,0xf3,0xf3,0x0e, 0xbf,0xd2,0xd2,0x6d + .byte 0x81,0xcd,0xcd,0x4c, 0x18,0x0c,0x0c,0x14 + .byte 0x26,0x13,0x13,0x35, 0xc3,0xec,0xec,0x2f + .byte 0xbe,0x5f,0x5f,0xe1, 0x35,0x97,0x97,0xa2 + .byte 0x88,0x44,0x44,0xcc, 0x2e,0x17,0x17,0x39 + .byte 0x93,0xc4,0xc4,0x57, 0x55,0xa7,0xa7,0xf2 + .byte 0xfc,0x7e,0x7e,0x82, 0x7a,0x3d,0x3d,0x47 + .byte 0xc8,0x64,0x64,0xac, 0xba,0x5d,0x5d,0xe7 + .byte 0x32,0x19,0x19,0x2b, 0xe6,0x73,0x73,0x95 + .byte 0xc0,0x60,0x60,0xa0, 0x19,0x81,0x81,0x98 + .byte 0x9e,0x4f,0x4f,0xd1, 0xa3,0xdc,0xdc,0x7f + .byte 0x44,0x22,0x22,0x66, 0x54,0x2a,0x2a,0x7e + .byte 0x3b,0x90,0x90,0xab, 0x0b,0x88,0x88,0x83 + .byte 0x8c,0x46,0x46,0xca, 0xc7,0xee,0xee,0x29 + .byte 0x6b,0xb8,0xb8,0xd3, 0x28,0x14,0x14,0x3c + .byte 0xa7,0xde,0xde,0x79, 0xbc,0x5e,0x5e,0xe2 + .byte 0x16,0x0b,0x0b,0x1d, 0xad,0xdb,0xdb,0x76 + .byte 0xdb,0xe0,0xe0,0x3b, 0x64,0x32,0x32,0x56 + .byte 0x74,0x3a,0x3a,0x4e, 0x14,0x0a,0x0a,0x1e + .byte 0x92,0x49,0x49,0xdb, 0x0c,0x06,0x06,0x0a + .byte 0x48,0x24,0x24,0x6c, 0xb8,0x5c,0x5c,0xe4 + .byte 0x9f,0xc2,0xc2,0x5d, 0xbd,0xd3,0xd3,0x6e + .byte 0x43,0xac,0xac,0xef, 0xc4,0x62,0x62,0xa6 + .byte 0x39,0x91,0x91,0xa8, 0x31,0x95,0x95,0xa4 + .byte 0xd3,0xe4,0xe4,0x37, 0xf2,0x79,0x79,0x8b + .byte 0xd5,0xe7,0xe7,0x32, 0x8b,0xc8,0xc8,0x43 + .byte 0x6e,0x37,0x37,0x59, 0xda,0x6d,0x6d,0xb7 + .byte 0x01,0x8d,0x8d,0x8c, 0xb1,0xd5,0xd5,0x64 + .byte 0x9c,0x4e,0x4e,0xd2, 0x49,0xa9,0xa9,0xe0 + .byte 0xd8,0x6c,0x6c,0xb4, 0xac,0x56,0x56,0xfa + .byte 0xf3,0xf4,0xf4,0x07, 0xcf,0xea,0xea,0x25 + .byte 0xca,0x65,0x65,0xaf, 0xf4,0x7a,0x7a,0x8e + .byte 0x47,0xae,0xae,0xe9, 0x10,0x08,0x08,0x18 + .byte 0x6f,0xba,0xba,0xd5, 0xf0,0x78,0x78,0x88 + .byte 0x4a,0x25,0x25,0x6f, 0x5c,0x2e,0x2e,0x72 + .byte 0x38,0x1c,0x1c,0x24, 0x57,0xa6,0xa6,0xf1 + .byte 0x73,0xb4,0xb4,0xc7, 0x97,0xc6,0xc6,0x51 + .byte 0xcb,0xe8,0xe8,0x23, 0xa1,0xdd,0xdd,0x7c + .byte 0xe8,0x74,0x74,0x9c, 0x3e,0x1f,0x1f,0x21 + .byte 0x96,0x4b,0x4b,0xdd, 0x61,0xbd,0xbd,0xdc + .byte 0x0d,0x8b,0x8b,0x86, 0x0f,0x8a,0x8a,0x85 + .byte 0xe0,0x70,0x70,0x90, 0x7c,0x3e,0x3e,0x42 + .byte 0x71,0xb5,0xb5,0xc4, 0xcc,0x66,0x66,0xaa + .byte 0x90,0x48,0x48,0xd8, 0x06,0x03,0x03,0x05 + .byte 0xf7,0xf6,0xf6,0x01, 0x1c,0x0e,0x0e,0x12 + .byte 0xc2,0x61,0x61,0xa3, 0x6a,0x35,0x35,0x5f + .byte 0xae,0x57,0x57,0xf9, 0x69,0xb9,0xb9,0xd0 + .byte 0x17,0x86,0x86,0x91, 0x99,0xc1,0xc1,0x58 + .byte 0x3a,0x1d,0x1d,0x27, 0x27,0x9e,0x9e,0xb9 + .byte 0xd9,0xe1,0xe1,0x38, 0xeb,0xf8,0xf8,0x13 + .byte 0x2b,0x98,0x98,0xb3, 0x22,0x11,0x11,0x33 + .byte 0xd2,0x69,0x69,0xbb, 0xa9,0xd9,0xd9,0x70 + .byte 0x07,0x8e,0x8e,0x89, 0x33,0x94,0x94,0xa7 + .byte 0x2d,0x9b,0x9b,0xb6, 0x3c,0x1e,0x1e,0x22 + .byte 0x15,0x87,0x87,0x92, 0xc9,0xe9,0xe9,0x20 + .byte 0x87,0xce,0xce,0x49, 0xaa,0x55,0x55,0xff + .byte 0x50,0x28,0x28,0x78, 0xa5,0xdf,0xdf,0x7a + .byte 0x03,0x8c,0x8c,0x8f, 0x59,0xa1,0xa1,0xf8 + .byte 0x09,0x89,0x89,0x80, 0x1a,0x0d,0x0d,0x17 + .byte 0x65,0xbf,0xbf,0xda, 0xd7,0xe6,0xe6,0x31 + .byte 0x84,0x42,0x42,0xc6, 0xd0,0x68,0x68,0xb8 + .byte 0x82,0x41,0x41,0xc3, 0x29,0x99,0x99,0xb0 + .byte 0x5a,0x2d,0x2d,0x77, 0x1e,0x0f,0x0f,0x11 + .byte 0x7b,0xb0,0xb0,0xcb, 0xa8,0x54,0x54,0xfc + .byte 0x6d,0xbb,0xbb,0xd6, 0x2c,0x16,0x16,0x3a +AES_Te4: + .byte 0x63, 0x7c, 0x77, 0x7b, 0xf2, 0x6b, 0x6f, 0xc5 + .byte 0x30, 0x01, 0x67, 0x2b, 0xfe, 0xd7, 0xab, 0x76 + .byte 0xca, 0x82, 0xc9, 0x7d, 0xfa, 0x59, 0x47, 0xf0 + .byte 0xad, 0xd4, 0xa2, 0xaf, 0x9c, 0xa4, 0x72, 0xc0 + .byte 0xb7, 0xfd, 0x93, 0x26, 0x36, 0x3f, 0xf7, 0xcc + .byte 0x34, 0xa5, 0xe5, 0xf1, 0x71, 0xd8, 0x31, 0x15 + .byte 0x04, 0xc7, 0x23, 0xc3, 0x18, 0x96, 0x05, 0x9a + .byte 0x07, 0x12, 0x80, 0xe2, 0xeb, 0x27, 0xb2, 0x75 + .byte 0x09, 0x83, 0x2c, 0x1a, 0x1b, 0x6e, 0x5a, 0xa0 + .byte 0x52, 0x3b, 0xd6, 0xb3, 0x29, 0xe3, 0x2f, 0x84 + .byte 0x53, 0xd1, 0x00, 0xed, 0x20, 0xfc, 0xb1, 0x5b + .byte 0x6a, 0xcb, 0xbe, 0x39, 0x4a, 0x4c, 0x58, 0xcf + .byte 0xd0, 0xef, 0xaa, 0xfb, 0x43, 0x4d, 0x33, 0x85 + .byte 0x45, 0xf9, 0x02, 0x7f, 0x50, 0x3c, 0x9f, 0xa8 + .byte 0x51, 0xa3, 0x40, 0x8f, 0x92, 0x9d, 0x38, 0xf5 + .byte 0xbc, 0xb6, 0xda, 0x21, 0x10, 0xff, 0xf3, 0xd2 + .byte 0xcd, 0x0c, 0x13, 0xec, 0x5f, 0x97, 0x44, 0x17 + .byte 0xc4, 0xa7, 0x7e, 0x3d, 0x64, 0x5d, 0x19, 0x73 + .byte 0x60, 0x81, 0x4f, 0xdc, 0x22, 0x2a, 0x90, 0x88 + .byte 0x46, 0xee, 0xb8, 0x14, 0xde, 0x5e, 0x0b, 0xdb + .byte 0xe0, 0x32, 0x3a, 0x0a, 0x49, 0x06, 0x24, 0x5c + .byte 0xc2, 0xd3, 0xac, 0x62, 0x91, 0x95, 0xe4, 0x79 + .byte 0xe7, 0xc8, 0x37, 0x6d, 0x8d, 0xd5, 0x4e, 0xa9 + .byte 0x6c, 0x56, 0xf4, 0xea, 0x65, 0x7a, 0xae, 0x08 + .byte 0xba, 0x78, 0x25, 0x2e, 0x1c, 0xa6, 0xb4, 0xc6 + .byte 0xe8, 0xdd, 0x74, 0x1f, 0x4b, 0xbd, 0x8b, 0x8a + .byte 0x70, 0x3e, 0xb5, 0x66, 0x48, 0x03, 0xf6, 0x0e + .byte 0x61, 0x35, 0x57, 0xb9, 0x86, 0xc1, 0x1d, 0x9e + .byte 0xe1, 0xf8, 0x98, 0x11, 0x69, 0xd9, 0x8e, 0x94 + .byte 0x9b, 0x1e, 0x87, 0xe9, 0xce, 0x55, 0x28, 0xdf + .byte 0x8c, 0xa1, 0x89, 0x0d, 0xbf, 0xe6, 0x42, 0x68 + .byte 0x41, 0x99, 0x2d, 0x0f, 0xb0, 0x54, 0xbb, 0x16 +rcon: + .byte 0x01,0x00,0x00,0x00, 0x02,0x00,0x00,0x00 + .byte 0x04,0x00,0x00,0x00, 0x08,0x00,0x00,0x00 + .byte 0x10,0x00,0x00,0x00, 0x20,0x00,0x00,0x00 + .byte 0x40,0x00,0x00,0x00, 0x80,0x00,0x00,0x00 + .byte 0x1B,0x00,0x00,0x00, 0x36,0x00,0x00,0x00 + .align 128 +AES_Td: + .byte 0x51,0xf4,0xa7,0x50, 0x7e,0x41,0x65,0x53 + .byte 0x1a,0x17,0xa4,0xc3, 0x3a,0x27,0x5e,0x96 + .byte 0x3b,0xab,0x6b,0xcb, 0x1f,0x9d,0x45,0xf1 + .byte 0xac,0xfa,0x58,0xab, 0x4b,0xe3,0x03,0x93 + .byte 0x20,0x30,0xfa,0x55, 0xad,0x76,0x6d,0xf6 + .byte 0x88,0xcc,0x76,0x91, 0xf5,0x02,0x4c,0x25 + .byte 0x4f,0xe5,0xd7,0xfc, 0xc5,0x2a,0xcb,0xd7 + .byte 0x26,0x35,0x44,0x80, 0xb5,0x62,0xa3,0x8f + .byte 0xde,0xb1,0x5a,0x49, 0x25,0xba,0x1b,0x67 + .byte 0x45,0xea,0x0e,0x98, 0x5d,0xfe,0xc0,0xe1 + .byte 0xc3,0x2f,0x75,0x02, 0x81,0x4c,0xf0,0x12 + .byte 0x8d,0x46,0x97,0xa3, 0x6b,0xd3,0xf9,0xc6 + .byte 0x03,0x8f,0x5f,0xe7, 0x15,0x92,0x9c,0x95 + .byte 0xbf,0x6d,0x7a,0xeb, 0x95,0x52,0x59,0xda + .byte 0xd4,0xbe,0x83,0x2d, 0x58,0x74,0x21,0xd3 + .byte 0x49,0xe0,0x69,0x29, 0x8e,0xc9,0xc8,0x44 + .byte 0x75,0xc2,0x89,0x6a, 0xf4,0x8e,0x79,0x78 + .byte 0x99,0x58,0x3e,0x6b, 0x27,0xb9,0x71,0xdd + .byte 0xbe,0xe1,0x4f,0xb6, 0xf0,0x88,0xad,0x17 + .byte 0xc9,0x20,0xac,0x66, 0x7d,0xce,0x3a,0xb4 + .byte 0x63,0xdf,0x4a,0x18, 0xe5,0x1a,0x31,0x82 + .byte 0x97,0x51,0x33,0x60, 0x62,0x53,0x7f,0x45 + .byte 0xb1,0x64,0x77,0xe0, 0xbb,0x6b,0xae,0x84 + .byte 0xfe,0x81,0xa0,0x1c, 0xf9,0x08,0x2b,0x94 + .byte 0x70,0x48,0x68,0x58, 0x8f,0x45,0xfd,0x19 + .byte 0x94,0xde,0x6c,0x87, 0x52,0x7b,0xf8,0xb7 + .byte 0xab,0x73,0xd3,0x23, 0x72,0x4b,0x02,0xe2 + .byte 0xe3,0x1f,0x8f,0x57, 0x66,0x55,0xab,0x2a + .byte 0xb2,0xeb,0x28,0x07, 0x2f,0xb5,0xc2,0x03 + .byte 0x86,0xc5,0x7b,0x9a, 0xd3,0x37,0x08,0xa5 + .byte 0x30,0x28,0x87,0xf2, 0x23,0xbf,0xa5,0xb2 + .byte 0x02,0x03,0x6a,0xba, 0xed,0x16,0x82,0x5c + .byte 0x8a,0xcf,0x1c,0x2b, 0xa7,0x79,0xb4,0x92 + .byte 0xf3,0x07,0xf2,0xf0, 0x4e,0x69,0xe2,0xa1 + .byte 0x65,0xda,0xf4,0xcd, 0x06,0x05,0xbe,0xd5 + .byte 0xd1,0x34,0x62,0x1f, 0xc4,0xa6,0xfe,0x8a + .byte 0x34,0x2e,0x53,0x9d, 0xa2,0xf3,0x55,0xa0 + .byte 0x05,0x8a,0xe1,0x32, 0xa4,0xf6,0xeb,0x75 + .byte 0x0b,0x83,0xec,0x39, 0x40,0x60,0xef,0xaa + .byte 0x5e,0x71,0x9f,0x06, 0xbd,0x6e,0x10,0x51 + .byte 0x3e,0x21,0x8a,0xf9, 0x96,0xdd,0x06,0x3d + .byte 0xdd,0x3e,0x05,0xae, 0x4d,0xe6,0xbd,0x46 + .byte 0x91,0x54,0x8d,0xb5, 0x71,0xc4,0x5d,0x05 + .byte 0x04,0x06,0xd4,0x6f, 0x60,0x50,0x15,0xff + .byte 0x19,0x98,0xfb,0x24, 0xd6,0xbd,0xe9,0x97 + .byte 0x89,0x40,0x43,0xcc, 0x67,0xd9,0x9e,0x77 + .byte 0xb0,0xe8,0x42,0xbd, 0x07,0x89,0x8b,0x88 + .byte 0xe7,0x19,0x5b,0x38, 0x79,0xc8,0xee,0xdb + .byte 0xa1,0x7c,0x0a,0x47, 0x7c,0x42,0x0f,0xe9 + .byte 0xf8,0x84,0x1e,0xc9, 0x00,0x00,0x00,0x00 + .byte 0x09,0x80,0x86,0x83, 0x32,0x2b,0xed,0x48 + .byte 0x1e,0x11,0x70,0xac, 0x6c,0x5a,0x72,0x4e + .byte 0xfd,0x0e,0xff,0xfb, 0x0f,0x85,0x38,0x56 + .byte 0x3d,0xae,0xd5,0x1e, 0x36,0x2d,0x39,0x27 + .byte 0x0a,0x0f,0xd9,0x64, 0x68,0x5c,0xa6,0x21 + .byte 0x9b,0x5b,0x54,0xd1, 0x24,0x36,0x2e,0x3a + .byte 0x0c,0x0a,0x67,0xb1, 0x93,0x57,0xe7,0x0f + .byte 0xb4,0xee,0x96,0xd2, 0x1b,0x9b,0x91,0x9e + .byte 0x80,0xc0,0xc5,0x4f, 0x61,0xdc,0x20,0xa2 + .byte 0x5a,0x77,0x4b,0x69, 0x1c,0x12,0x1a,0x16 + .byte 0xe2,0x93,0xba,0x0a, 0xc0,0xa0,0x2a,0xe5 + .byte 0x3c,0x22,0xe0,0x43, 0x12,0x1b,0x17,0x1d + .byte 0x0e,0x09,0x0d,0x0b, 0xf2,0x8b,0xc7,0xad + .byte 0x2d,0xb6,0xa8,0xb9, 0x14,0x1e,0xa9,0xc8 + .byte 0x57,0xf1,0x19,0x85, 0xaf,0x75,0x07,0x4c + .byte 0xee,0x99,0xdd,0xbb, 0xa3,0x7f,0x60,0xfd + .byte 0xf7,0x01,0x26,0x9f, 0x5c,0x72,0xf5,0xbc + .byte 0x44,0x66,0x3b,0xc5, 0x5b,0xfb,0x7e,0x34 + .byte 0x8b,0x43,0x29,0x76, 0xcb,0x23,0xc6,0xdc + .byte 0xb6,0xed,0xfc,0x68, 0xb8,0xe4,0xf1,0x63 + .byte 0xd7,0x31,0xdc,0xca, 0x42,0x63,0x85,0x10 + .byte 0x13,0x97,0x22,0x40, 0x84,0xc6,0x11,0x20 + .byte 0x85,0x4a,0x24,0x7d, 0xd2,0xbb,0x3d,0xf8 + .byte 0xae,0xf9,0x32,0x11, 0xc7,0x29,0xa1,0x6d + .byte 0x1d,0x9e,0x2f,0x4b, 0xdc,0xb2,0x30,0xf3 + .byte 0x0d,0x86,0x52,0xec, 0x77,0xc1,0xe3,0xd0 + .byte 0x2b,0xb3,0x16,0x6c, 0xa9,0x70,0xb9,0x99 + .byte 0x11,0x94,0x48,0xfa, 0x47,0xe9,0x64,0x22 + .byte 0xa8,0xfc,0x8c,0xc4, 0xa0,0xf0,0x3f,0x1a + .byte 0x56,0x7d,0x2c,0xd8, 0x22,0x33,0x90,0xef + .byte 0x87,0x49,0x4e,0xc7, 0xd9,0x38,0xd1,0xc1 + .byte 0x8c,0xca,0xa2,0xfe, 0x98,0xd4,0x0b,0x36 + .byte 0xa6,0xf5,0x81,0xcf, 0xa5,0x7a,0xde,0x28 + .byte 0xda,0xb7,0x8e,0x26, 0x3f,0xad,0xbf,0xa4 + .byte 0x2c,0x3a,0x9d,0xe4, 0x50,0x78,0x92,0x0d + .byte 0x6a,0x5f,0xcc,0x9b, 0x54,0x7e,0x46,0x62 + .byte 0xf6,0x8d,0x13,0xc2, 0x90,0xd8,0xb8,0xe8 + .byte 0x2e,0x39,0xf7,0x5e, 0x82,0xc3,0xaf,0xf5 + .byte 0x9f,0x5d,0x80,0xbe, 0x69,0xd0,0x93,0x7c + .byte 0x6f,0xd5,0x2d,0xa9, 0xcf,0x25,0x12,0xb3 + .byte 0xc8,0xac,0x99,0x3b, 0x10,0x18,0x7d,0xa7 + .byte 0xe8,0x9c,0x63,0x6e, 0xdb,0x3b,0xbb,0x7b + .byte 0xcd,0x26,0x78,0x09, 0x6e,0x59,0x18,0xf4 + .byte 0xec,0x9a,0xb7,0x01, 0x83,0x4f,0x9a,0xa8 + .byte 0xe6,0x95,0x6e,0x65, 0xaa,0xff,0xe6,0x7e + .byte 0x21,0xbc,0xcf,0x08, 0xef,0x15,0xe8,0xe6 + .byte 0xba,0xe7,0x9b,0xd9, 0x4a,0x6f,0x36,0xce + .byte 0xea,0x9f,0x09,0xd4, 0x29,0xb0,0x7c,0xd6 + .byte 0x31,0xa4,0xb2,0xaf, 0x2a,0x3f,0x23,0x31 + .byte 0xc6,0xa5,0x94,0x30, 0x35,0xa2,0x66,0xc0 + .byte 0x74,0x4e,0xbc,0x37, 0xfc,0x82,0xca,0xa6 + .byte 0xe0,0x90,0xd0,0xb0, 0x33,0xa7,0xd8,0x15 + .byte 0xf1,0x04,0x98,0x4a, 0x41,0xec,0xda,0xf7 + .byte 0x7f,0xcd,0x50,0x0e, 0x17,0x91,0xf6,0x2f + .byte 0x76,0x4d,0xd6,0x8d, 0x43,0xef,0xb0,0x4d + .byte 0xcc,0xaa,0x4d,0x54, 0xe4,0x96,0x04,0xdf + .byte 0x9e,0xd1,0xb5,0xe3, 0x4c,0x6a,0x88,0x1b + .byte 0xc1,0x2c,0x1f,0xb8, 0x46,0x65,0x51,0x7f + .byte 0x9d,0x5e,0xea,0x04, 0x01,0x8c,0x35,0x5d + .byte 0xfa,0x87,0x74,0x73, 0xfb,0x0b,0x41,0x2e + .byte 0xb3,0x67,0x1d,0x5a, 0x92,0xdb,0xd2,0x52 + .byte 0xe9,0x10,0x56,0x33, 0x6d,0xd6,0x47,0x13 + .byte 0x9a,0xd7,0x61,0x8c, 0x37,0xa1,0x0c,0x7a + .byte 0x59,0xf8,0x14,0x8e, 0xeb,0x13,0x3c,0x89 + .byte 0xce,0xa9,0x27,0xee, 0xb7,0x61,0xc9,0x35 + .byte 0xe1,0x1c,0xe5,0xed, 0x7a,0x47,0xb1,0x3c + .byte 0x9c,0xd2,0xdf,0x59, 0x55,0xf2,0x73,0x3f + .byte 0x18,0x14,0xce,0x79, 0x73,0xc7,0x37,0xbf + .byte 0x53,0xf7,0xcd,0xea, 0x5f,0xfd,0xaa,0x5b + .byte 0xdf,0x3d,0x6f,0x14, 0x78,0x44,0xdb,0x86 + .byte 0xca,0xaf,0xf3,0x81, 0xb9,0x68,0xc4,0x3e + .byte 0x38,0x24,0x34,0x2c, 0xc2,0xa3,0x40,0x5f + .byte 0x16,0x1d,0xc3,0x72, 0xbc,0xe2,0x25,0x0c + .byte 0x28,0x3c,0x49,0x8b, 0xff,0x0d,0x95,0x41 + .byte 0x39,0xa8,0x01,0x71, 0x08,0x0c,0xb3,0xde + .byte 0xd8,0xb4,0xe4,0x9c, 0x64,0x56,0xc1,0x90 + .byte 0x7b,0xcb,0x84,0x61, 0xd5,0x32,0xb6,0x70 + .byte 0x48,0x6c,0x5c,0x74, 0xd0,0xb8,0x57,0x42 +AES_Td4: + .byte 0x52, 0x09, 0x6a, 0xd5, 0x30, 0x36, 0xa5, 0x38 + .byte 0xbf, 0x40, 0xa3, 0x9e, 0x81, 0xf3, 0xd7, 0xfb + .byte 0x7c, 0xe3, 0x39, 0x82, 0x9b, 0x2f, 0xff, 0x87 + .byte 0x34, 0x8e, 0x43, 0x44, 0xc4, 0xde, 0xe9, 0xcb + .byte 0x54, 0x7b, 0x94, 0x32, 0xa6, 0xc2, 0x23, 0x3d + .byte 0xee, 0x4c, 0x95, 0x0b, 0x42, 0xfa, 0xc3, 0x4e + .byte 0x08, 0x2e, 0xa1, 0x66, 0x28, 0xd9, 0x24, 0xb2 + .byte 0x76, 0x5b, 0xa2, 0x49, 0x6d, 0x8b, 0xd1, 0x25 + .byte 0x72, 0xf8, 0xf6, 0x64, 0x86, 0x68, 0x98, 0x16 + .byte 0xd4, 0xa4, 0x5c, 0xcc, 0x5d, 0x65, 0xb6, 0x92 + .byte 0x6c, 0x70, 0x48, 0x50, 0xfd, 0xed, 0xb9, 0xda + .byte 0x5e, 0x15, 0x46, 0x57, 0xa7, 0x8d, 0x9d, 0x84 + .byte 0x90, 0xd8, 0xab, 0x00, 0x8c, 0xbc, 0xd3, 0x0a + .byte 0xf7, 0xe4, 0x58, 0x05, 0xb8, 0xb3, 0x45, 0x06 + .byte 0xd0, 0x2c, 0x1e, 0x8f, 0xca, 0x3f, 0x0f, 0x02 + .byte 0xc1, 0xaf, 0xbd, 0x03, 0x01, 0x13, 0x8a, 0x6b + .byte 0x3a, 0x91, 0x11, 0x41, 0x4f, 0x67, 0xdc, 0xea + .byte 0x97, 0xf2, 0xcf, 0xce, 0xf0, 0xb4, 0xe6, 0x73 + .byte 0x96, 0xac, 0x74, 0x22, 0xe7, 0xad, 0x35, 0x85 + .byte 0xe2, 0xf9, 0x37, 0xe8, 0x1c, 0x75, 0xdf, 0x6e + .byte 0x47, 0xf1, 0x1a, 0x71, 0x1d, 0x29, 0xc5, 0x89 + .byte 0x6f, 0xb7, 0x62, 0x0e, 0xaa, 0x18, 0xbe, 0x1b + .byte 0xfc, 0x56, 0x3e, 0x4b, 0xc6, 0xd2, 0x79, 0x20 + .byte 0x9a, 0xdb, 0xc0, 0xfe, 0x78, 0xcd, 0x5a, 0xf4 + .byte 0x1f, 0xdd, 0xa8, 0x33, 0x88, 0x07, 0xc7, 0x31 + .byte 0xb1, 0x12, 0x10, 0x59, 0x27, 0x80, 0xec, 0x5f + .byte 0x60, 0x51, 0x7f, 0xa9, 0x19, 0xb5, 0x4a, 0x0d + .byte 0x2d, 0xe5, 0x7a, 0x9f, 0x93, 0xc9, 0x9c, 0xef + .byte 0xa0, 0xe0, 0x3b, 0x4d, 0xae, 0x2a, 0xf5, 0xb0 + .byte 0xc8, 0xeb, 0xbb, 0x3c, 0x83, 0x53, 0x99, 0x61 + .byte 0x17, 0x2b, 0x04, 0x7e, 0xba, 0x77, 0xd6, 0x26 + .byte 0xe1, 0x69, 0x14, 0x63, 0x55, 0x21, 0x0c, 0x7d + .cstring "AES for C64x+, CRYPTOGAMS by " + .align 4 +___ + +print $code; +close STDOUT; diff --git a/openssl-1.1.0h/crypto/aes/asm/aes-ia64.S b/openssl-1.1.0h/crypto/aes/asm/aes-ia64.S new file mode 100644 index 0000000..f7f1f63 --- /dev/null +++ b/openssl-1.1.0h/crypto/aes/asm/aes-ia64.S @@ -0,0 +1,1130 @@ +// Copyright 2004-2016 The OpenSSL Project Authors. All Rights Reserved. +// +// Licensed under the OpenSSL license (the "License"). You may not use +// this file except in compliance with the License. You can obtain a copy +// in the file LICENSE in the source distribution or at +// https://www.openssl.org/source/license.html +// +// ==================================================================== +// Written by Andy Polyakov for the OpenSSL +// project. Rights for redistribution and usage in source and binary +// forms are granted according to the OpenSSL license. +// ==================================================================== +// +// What's wrong with compiler generated code? Compiler never uses +// variable 'shr' which is pairable with 'extr'/'dep' instructions. +// Then it uses 'zxt' which is an I-type, but can be replaced with +// 'and' which in turn can be assigned to M-port [there're double as +// much M-ports as there're I-ports on Itanium 2]. By sacrificing few +// registers for small constants (255, 24 and 16) to be used with +// 'shr' and 'and' instructions I can achieve better ILP, Instruction +// Level Parallelism, and performance. This code outperforms GCC 3.3 +// generated code by over factor of 2 (two), GCC 3.4 - by 70% and +// HP C - by 40%. Measured best-case scenario, i.e. aligned +// big-endian input, ECB timing on Itanium 2 is (18 + 13*rounds) +// ticks per block, or 9.25 CPU cycles per byte for 128 bit key. + +// Version 1.2 mitigates the hazard of cache-timing attacks by +// a) compressing S-boxes from 8KB to 2KB+256B, b) scheduling +// references to S-boxes for L2 cache latency, c) prefetching T[ed]4 +// prior last round. As result performance dropped to (26 + 15*rounds) +// ticks per block or 11 cycles per byte processed with 128-bit key. +// This is ~16% deterioration. For reference Itanium 2 L1 cache has +// 64 bytes line size and L2 - 128 bytes... + +.ident "aes-ia64.S, version 1.2" +.ident "IA-64 ISA artwork by Andy Polyakov " +.explicit +.text + +rk0=r8; rk1=r9; + +pfssave=r2; +lcsave=r10; +prsave=r3; +maskff=r11; +twenty4=r14; +sixteen=r15; + +te00=r16; te11=r17; te22=r18; te33=r19; +te01=r20; te12=r21; te23=r22; te30=r23; +te02=r24; te13=r25; te20=r26; te31=r27; +te03=r28; te10=r29; te21=r30; te32=r31; + +// these are rotating... +t0=r32; s0=r33; +t1=r34; s1=r35; +t2=r36; s2=r37; +t3=r38; s3=r39; + +te0=r40; te1=r41; te2=r42; te3=r43; + +#if defined(_HPUX_SOURCE) && !defined(_LP64) +# define ADDP addp4 +#else +# define ADDP add +#endif + +// Offsets from Te0 +#define TE0 0 +#define TE2 2 +#if defined(_HPUX_SOURCE) || defined(B_ENDIAN) +#define TE1 3 +#define TE3 1 +#else +#define TE1 1 +#define TE3 3 +#endif + +// This implies that AES_KEY comprises 32-bit key schedule elements +// even on LP64 platforms. +#ifndef KSZ +# define KSZ 4 +# define LDKEY ld4 +#endif + +.proc _ia64_AES_encrypt# +// Input: rk0-rk1 +// te0 +// te3 as AES_KEY->rounds!!! +// s0-s3 +// maskff,twenty4,sixteen +// Output: r16,r20,r24,r28 as s0-s3 +// Clobber: r16-r31,rk0-rk1,r32-r43 +.align 32 +_ia64_AES_encrypt: + .prologue + .altrp b6 + .body +{ .mmi; alloc r16=ar.pfs,12,0,0,8 + LDKEY t0=[rk0],2*KSZ + mov pr.rot=1<<16 } +{ .mmi; LDKEY t1=[rk1],2*KSZ + add te1=TE1,te0 + add te3=-3,te3 };; +{ .mib; LDKEY t2=[rk0],2*KSZ + mov ar.ec=2 } +{ .mib; LDKEY t3=[rk1],2*KSZ + add te2=TE2,te0 + brp.loop.imp .Le_top,.Le_end-16 };; + +{ .mmi; xor s0=s0,t0 + xor s1=s1,t1 + mov ar.lc=te3 } +{ .mmi; xor s2=s2,t2 + xor s3=s3,t3 + add te3=TE3,te0 };; + +.align 32 +.Le_top: +{ .mmi; (p0) LDKEY t0=[rk0],2*KSZ // 0/0:rk[0] + (p0) and te33=s3,maskff // 0/0:s3&0xff + (p0) extr.u te22=s2,8,8 } // 0/0:s2>>8&0xff +{ .mmi; (p0) LDKEY t1=[rk1],2*KSZ // 0/1:rk[1] + (p0) and te30=s0,maskff // 0/1:s0&0xff + (p0) shr.u te00=s0,twenty4 };; // 0/0:s0>>24 +{ .mmi; (p0) LDKEY t2=[rk0],2*KSZ // 1/2:rk[2] + (p0) shladd te33=te33,3,te3 // 1/0:te0+s0>>24 + (p0) extr.u te23=s3,8,8 } // 1/1:s3>>8&0xff +{ .mmi; (p0) LDKEY t3=[rk1],2*KSZ // 1/3:rk[3] + (p0) shladd te30=te30,3,te3 // 1/1:te3+s0 + (p0) shr.u te01=s1,twenty4 };; // 1/1:s1>>24 +{ .mmi; (p0) ld4 te33=[te33] // 2/0:te3[s3&0xff] + (p0) shladd te22=te22,3,te2 // 2/0:te2+s2>>8&0xff + (p0) extr.u te20=s0,8,8 } // 2/2:s0>>8&0xff +{ .mmi; (p0) ld4 te30=[te30] // 2/1:te3[s0] + (p0) shladd te23=te23,3,te2 // 2/1:te2+s3>>8 + (p0) shr.u te02=s2,twenty4 };; // 2/2:s2>>24 +{ .mmi; (p0) ld4 te22=[te22] // 3/0:te2[s2>>8] + (p0) shladd te20=te20,3,te2 // 3/2:te2+s0>>8 + (p0) extr.u te21=s1,8,8 } // 3/3:s1>>8&0xff +{ .mmi; (p0) ld4 te23=[te23] // 3/1:te2[s3>>8] + (p0) shladd te00=te00,3,te0 // 3/0:te0+s0>>24 + (p0) shr.u te03=s3,twenty4 };; // 3/3:s3>>24 +{ .mmi; (p0) ld4 te20=[te20] // 4/2:te2[s0>>8] + (p0) shladd te21=te21,3,te2 // 4/3:te3+s2 + (p0) extr.u te11=s1,16,8 } // 4/0:s1>>16&0xff +{ .mmi; (p0) ld4 te00=[te00] // 4/0:te0[s0>>24] + (p0) shladd te01=te01,3,te0 // 4/1:te0+s1>>24 + (p0) shr.u te13=s3,sixteen };; // 4/2:s3>>16 +{ .mmi; (p0) ld4 te21=[te21] // 5/3:te2[s1>>8] + (p0) shladd te11=te11,3,te1 // 5/0:te1+s1>>16 + (p0) extr.u te12=s2,16,8 } // 5/1:s2>>16&0xff +{ .mmi; (p0) ld4 te01=[te01] // 5/1:te0[s1>>24] + (p0) shladd te02=te02,3,te0 // 5/2:te0+s2>>24 + (p0) and te31=s1,maskff };; // 5/2:s1&0xff +{ .mmi; (p0) ld4 te11=[te11] // 6/0:te1[s1>>16] + (p0) shladd te12=te12,3,te1 // 6/1:te1+s2>>16 + (p0) extr.u te10=s0,16,8 } // 6/3:s0>>16&0xff +{ .mmi; (p0) ld4 te02=[te02] // 6/2:te0[s2>>24] + (p0) shladd te03=te03,3,te0 // 6/3:te1+s0>>16 + (p0) and te32=s2,maskff };; // 6/3:s2&0xff + +{ .mmi; (p0) ld4 te12=[te12] // 7/1:te1[s2>>16] + (p0) shladd te31=te31,3,te3 // 7/2:te3+s1&0xff + (p0) and te13=te13,maskff} // 7/2:s3>>16&0xff +{ .mmi; (p0) ld4 te03=[te03] // 7/3:te0[s3>>24] + (p0) shladd te32=te32,3,te3 // 7/3:te3+s2 + (p0) xor t0=t0,te33 };; // 7/0: +{ .mmi; (p0) ld4 te31=[te31] // 8/2:te3[s1] + (p0) shladd te13=te13,3,te1 // 8/2:te1+s3>>16 + (p0) xor t0=t0,te22 } // 8/0: +{ .mmi; (p0) ld4 te32=[te32] // 8/3:te3[s2] + (p0) shladd te10=te10,3,te1 // 8/3:te1+s0>>16 + (p0) xor t1=t1,te30 };; // 8/1: +{ .mmi; (p0) ld4 te13=[te13] // 9/2:te1[s3>>16] + (p0) ld4 te10=[te10] // 9/3:te1[s0>>16] + (p0) xor t0=t0,te00 };; // 9/0: !L2 scheduling +{ .mmi; (p0) xor t1=t1,te23 // 10[9]/1: + (p0) xor t2=t2,te20 // 10[9]/2: + (p0) xor t3=t3,te21 };; // 10[9]/3: +{ .mmi; (p0) xor t0=t0,te11 // 11[10]/0:done! + (p0) xor t1=t1,te01 // 11[10]/1: + (p0) xor t2=t2,te02 };; // 11[10]/2: !L2 scheduling +{ .mmi; (p0) xor t3=t3,te03 // 12[10]/3: + (p16) cmp.eq p0,p17=r0,r0 };; // 12[10]/clear (p17) +{ .mmi; (p0) xor t1=t1,te12 // 13[11]/1:done! + (p0) xor t2=t2,te31 // 13[11]/2: + (p0) xor t3=t3,te32 } // 13[11]/3: +{ .mmi; (p17) add te0=2048,te0 // 13[11]/ + (p17) add te1=2048+64-TE1,te1};; // 13[11]/ +{ .mib; (p0) xor t2=t2,te13 // 14[12]/2:done! + (p17) add te2=2048+128-TE2,te2} // 14[12]/ +{ .mib; (p0) xor t3=t3,te10 // 14[12]/3:done! + (p17) add te3=2048+192-TE3,te3 // 14[12]/ + br.ctop.sptk .Le_top };; +.Le_end: + + +{ .mmi; ld8 te12=[te0] // prefetch Te4 + ld8 te31=[te1] } +{ .mmi; ld8 te10=[te2] + ld8 te32=[te3] } + +{ .mmi; LDKEY t0=[rk0],2*KSZ // 0/0:rk[0] + and te33=s3,maskff // 0/0:s3&0xff + extr.u te22=s2,8,8 } // 0/0:s2>>8&0xff +{ .mmi; LDKEY t1=[rk1],2*KSZ // 0/1:rk[1] + and te30=s0,maskff // 0/1:s0&0xff + shr.u te00=s0,twenty4 };; // 0/0:s0>>24 +{ .mmi; LDKEY t2=[rk0],2*KSZ // 1/2:rk[2] + add te33=te33,te0 // 1/0:te0+s0>>24 + extr.u te23=s3,8,8 } // 1/1:s3>>8&0xff +{ .mmi; LDKEY t3=[rk1],2*KSZ // 1/3:rk[3] + add te30=te30,te0 // 1/1:te0+s0 + shr.u te01=s1,twenty4 };; // 1/1:s1>>24 +{ .mmi; ld1 te33=[te33] // 2/0:te0[s3&0xff] + add te22=te22,te0 // 2/0:te0+s2>>8&0xff + extr.u te20=s0,8,8 } // 2/2:s0>>8&0xff +{ .mmi; ld1 te30=[te30] // 2/1:te0[s0] + add te23=te23,te0 // 2/1:te0+s3>>8 + shr.u te02=s2,twenty4 };; // 2/2:s2>>24 +{ .mmi; ld1 te22=[te22] // 3/0:te0[s2>>8] + add te20=te20,te0 // 3/2:te0+s0>>8 + extr.u te21=s1,8,8 } // 3/3:s1>>8&0xff +{ .mmi; ld1 te23=[te23] // 3/1:te0[s3>>8] + add te00=te00,te0 // 3/0:te0+s0>>24 + shr.u te03=s3,twenty4 };; // 3/3:s3>>24 +{ .mmi; ld1 te20=[te20] // 4/2:te0[s0>>8] + add te21=te21,te0 // 4/3:te0+s2 + extr.u te11=s1,16,8 } // 4/0:s1>>16&0xff +{ .mmi; ld1 te00=[te00] // 4/0:te0[s0>>24] + add te01=te01,te0 // 4/1:te0+s1>>24 + shr.u te13=s3,sixteen };; // 4/2:s3>>16 +{ .mmi; ld1 te21=[te21] // 5/3:te0[s1>>8] + add te11=te11,te0 // 5/0:te0+s1>>16 + extr.u te12=s2,16,8 } // 5/1:s2>>16&0xff +{ .mmi; ld1 te01=[te01] // 5/1:te0[s1>>24] + add te02=te02,te0 // 5/2:te0+s2>>24 + and te31=s1,maskff };; // 5/2:s1&0xff +{ .mmi; ld1 te11=[te11] // 6/0:te0[s1>>16] + add te12=te12,te0 // 6/1:te0+s2>>16 + extr.u te10=s0,16,8 } // 6/3:s0>>16&0xff +{ .mmi; ld1 te02=[te02] // 6/2:te0[s2>>24] + add te03=te03,te0 // 6/3:te0+s0>>16 + and te32=s2,maskff };; // 6/3:s2&0xff + +{ .mmi; ld1 te12=[te12] // 7/1:te0[s2>>16] + add te31=te31,te0 // 7/2:te0+s1&0xff + dep te33=te22,te33,8,8} // 7/0: +{ .mmi; ld1 te03=[te03] // 7/3:te0[s3>>24] + add te32=te32,te0 // 7/3:te0+s2 + and te13=te13,maskff};; // 7/2:s3>>16&0xff +{ .mmi; ld1 te31=[te31] // 8/2:te0[s1] + add te13=te13,te0 // 8/2:te0+s3>>16 + dep te30=te23,te30,8,8} // 8/1: +{ .mmi; ld1 te32=[te32] // 8/3:te0[s2] + add te10=te10,te0 // 8/3:te0+s0>>16 + shl te00=te00,twenty4};; // 8/0: +{ .mii; ld1 te13=[te13] // 9/2:te0[s3>>16] + dep te33=te11,te33,16,8 // 9/0: + shl te01=te01,twenty4};; // 9/1: +{ .mii; ld1 te10=[te10] // 10/3:te0[s0>>16] + dep te31=te20,te31,8,8 // 10/2: + shl te02=te02,twenty4};; // 10/2: +{ .mii; xor t0=t0,te33 // 11/0: + dep te32=te21,te32,8,8 // 11/3: + shl te12=te12,sixteen};; // 11/1: +{ .mii; xor r16=t0,te00 // 12/0:done! + dep te31=te13,te31,16,8 // 12/2: + shl te03=te03,twenty4};; // 12/3: +{ .mmi; xor t1=t1,te01 // 13/1: + xor t2=t2,te02 // 13/2: + dep te32=te10,te32,16,8};; // 13/3: +{ .mmi; xor t1=t1,te30 // 14/1: + xor r24=t2,te31 // 14/2:done! + xor t3=t3,te32 };; // 14/3: +{ .mib; xor r20=t1,te12 // 15/1:done! + xor r28=t3,te03 // 15/3:done! + br.ret.sptk b6 };; +.endp _ia64_AES_encrypt# + +// void AES_encrypt (const void *in,void *out,const AES_KEY *key); +.global AES_encrypt# +.proc AES_encrypt# +.align 32 +AES_encrypt: + .prologue + .save ar.pfs,pfssave +{ .mmi; alloc pfssave=ar.pfs,3,1,12,0 + and out0=3,in0 + mov r3=ip } +{ .mmi; ADDP in0=0,in0 + mov loc0=psr.um + ADDP out11=KSZ*60,in2 };; // &AES_KEY->rounds + +{ .mmi; ld4 out11=[out11] // AES_KEY->rounds + add out8=(AES_Te#-AES_encrypt#),r3 // Te0 + .save pr,prsave + mov prsave=pr } +{ .mmi; rum 1<<3 // clear um.ac + .save ar.lc,lcsave + mov lcsave=ar.lc };; + + .body +#if defined(_HPUX_SOURCE) // HPUX is big-endian, cut 15+15 cycles... +{ .mib; cmp.ne p6,p0=out0,r0 + add out0=4,in0 +(p6) br.dpnt.many .Le_i_unaligned };; + +{ .mmi; ld4 out1=[in0],8 // s0 + and out9=3,in1 + mov twenty4=24 } +{ .mmi; ld4 out3=[out0],8 // s1 + ADDP rk0=0,in2 + mov sixteen=16 };; +{ .mmi; ld4 out5=[in0] // s2 + cmp.ne p6,p0=out9,r0 + mov maskff=0xff } +{ .mmb; ld4 out7=[out0] // s3 + ADDP rk1=KSZ,in2 + br.call.sptk.many b6=_ia64_AES_encrypt };; + +{ .mib; ADDP in0=4,in1 + ADDP in1=0,in1 +(p6) br.spnt .Le_o_unaligned };; + +{ .mii; mov psr.um=loc0 + mov ar.pfs=pfssave + mov ar.lc=lcsave };; +{ .mmi; st4 [in1]=r16,8 // s0 + st4 [in0]=r20,8 // s1 + mov pr=prsave,0x1ffff };; +{ .mmb; st4 [in1]=r24 // s2 + st4 [in0]=r28 // s3 + br.ret.sptk.many b0 };; +#endif + +.align 32 +.Le_i_unaligned: +{ .mmi; add out0=1,in0 + add out2=2,in0 + add out4=3,in0 };; +{ .mmi; ld1 r16=[in0],4 + ld1 r17=[out0],4 }//;; +{ .mmi; ld1 r18=[out2],4 + ld1 out1=[out4],4 };; // s0 +{ .mmi; ld1 r20=[in0],4 + ld1 r21=[out0],4 }//;; +{ .mmi; ld1 r22=[out2],4 + ld1 out3=[out4],4 };; // s1 +{ .mmi; ld1 r24=[in0],4 + ld1 r25=[out0],4 }//;; +{ .mmi; ld1 r26=[out2],4 + ld1 out5=[out4],4 };; // s2 +{ .mmi; ld1 r28=[in0] + ld1 r29=[out0] }//;; +{ .mmi; ld1 r30=[out2] + ld1 out7=[out4] };; // s3 + +{ .mii; + dep out1=r16,out1,24,8 //;; + dep out3=r20,out3,24,8 }//;; +{ .mii; ADDP rk0=0,in2 + dep out5=r24,out5,24,8 //;; + dep out7=r28,out7,24,8 };; +{ .mii; ADDP rk1=KSZ,in2 + dep out1=r17,out1,16,8 //;; + dep out3=r21,out3,16,8 }//;; +{ .mii; mov twenty4=24 + dep out5=r25,out5,16,8 //;; + dep out7=r29,out7,16,8 };; +{ .mii; mov sixteen=16 + dep out1=r18,out1,8,8 //;; + dep out3=r22,out3,8,8 }//;; +{ .mii; mov maskff=0xff + dep out5=r26,out5,8,8 //;; + dep out7=r30,out7,8,8 };; + +{ .mib; br.call.sptk.many b6=_ia64_AES_encrypt };; + +.Le_o_unaligned: +{ .mii; ADDP out0=0,in1 + extr.u r17=r16,8,8 // s0 + shr.u r19=r16,twenty4 }//;; +{ .mii; ADDP out1=1,in1 + extr.u r18=r16,16,8 + shr.u r23=r20,twenty4 }//;; // s1 +{ .mii; ADDP out2=2,in1 + extr.u r21=r20,8,8 + shr.u r22=r20,sixteen }//;; +{ .mii; ADDP out3=3,in1 + extr.u r25=r24,8,8 // s2 + shr.u r27=r24,twenty4 };; +{ .mii; st1 [out3]=r16,4 + extr.u r26=r24,16,8 + shr.u r31=r28,twenty4 }//;; // s3 +{ .mii; st1 [out2]=r17,4 + extr.u r29=r28,8,8 + shr.u r30=r28,sixteen }//;; + +{ .mmi; st1 [out1]=r18,4 + st1 [out0]=r19,4 };; +{ .mmi; st1 [out3]=r20,4 + st1 [out2]=r21,4 }//;; +{ .mmi; st1 [out1]=r22,4 + st1 [out0]=r23,4 };; +{ .mmi; st1 [out3]=r24,4 + st1 [out2]=r25,4 + mov pr=prsave,0x1ffff }//;; +{ .mmi; st1 [out1]=r26,4 + st1 [out0]=r27,4 + mov ar.pfs=pfssave };; +{ .mmi; st1 [out3]=r28 + st1 [out2]=r29 + mov ar.lc=lcsave }//;; +{ .mmi; st1 [out1]=r30 + st1 [out0]=r31 } +{ .mfb; mov psr.um=loc0 // restore user mask + br.ret.sptk.many b0 };; +.endp AES_encrypt# + +// *AES_decrypt are autogenerated by the following script: +#if 0 +#!/usr/bin/env perl +print "// *AES_decrypt are autogenerated by the following script:\n#if 0\n"; +open(PROG,'<'.$0); while() { print; } close(PROG); +print "#endif\n"; +while(<>) { + $process=1 if (/\.proc\s+_ia64_AES_encrypt/); + next if (!$process); + + #s/te00=s0/td00=s0/; s/te00/td00/g; + s/te11=s1/td13=s3/; s/te11/td13/g; + #s/te22=s2/td22=s2/; s/te22/td22/g; + s/te33=s3/td31=s1/; s/te33/td31/g; + + #s/te01=s1/td01=s1/; s/te01/td01/g; + s/te12=s2/td10=s0/; s/te12/td10/g; + #s/te23=s3/td23=s3/; s/te23/td23/g; + s/te30=s0/td32=s2/; s/te30/td32/g; + + #s/te02=s2/td02=s2/; s/te02/td02/g; + s/te13=s3/td11=s1/; s/te13/td11/g; + #s/te20=s0/td20=s0/; s/te20/td20/g; + s/te31=s1/td33=s3/; s/te31/td33/g; + + #s/te03=s3/td03=s3/; s/te03/td03/g; + s/te10=s0/td12=s2/; s/te10/td12/g; + #s/te21=s1/td21=s1/; s/te21/td21/g; + s/te32=s2/td30=s0/; s/te32/td30/g; + + s/td/te/g; + + s/AES_encrypt/AES_decrypt/g; + s/\.Le_/.Ld_/g; + s/AES_Te#/AES_Td#/g; + + print; + + exit if (/\.endp\s+AES_decrypt/); +} +#endif +.proc _ia64_AES_decrypt# +// Input: rk0-rk1 +// te0 +// te3 as AES_KEY->rounds!!! +// s0-s3 +// maskff,twenty4,sixteen +// Output: r16,r20,r24,r28 as s0-s3 +// Clobber: r16-r31,rk0-rk1,r32-r43 +.align 32 +_ia64_AES_decrypt: + .prologue + .altrp b6 + .body +{ .mmi; alloc r16=ar.pfs,12,0,0,8 + LDKEY t0=[rk0],2*KSZ + mov pr.rot=1<<16 } +{ .mmi; LDKEY t1=[rk1],2*KSZ + add te1=TE1,te0 + add te3=-3,te3 };; +{ .mib; LDKEY t2=[rk0],2*KSZ + mov ar.ec=2 } +{ .mib; LDKEY t3=[rk1],2*KSZ + add te2=TE2,te0 + brp.loop.imp .Ld_top,.Ld_end-16 };; + +{ .mmi; xor s0=s0,t0 + xor s1=s1,t1 + mov ar.lc=te3 } +{ .mmi; xor s2=s2,t2 + xor s3=s3,t3 + add te3=TE3,te0 };; + +.align 32 +.Ld_top: +{ .mmi; (p0) LDKEY t0=[rk0],2*KSZ // 0/0:rk[0] + (p0) and te31=s1,maskff // 0/0:s3&0xff + (p0) extr.u te22=s2,8,8 } // 0/0:s2>>8&0xff +{ .mmi; (p0) LDKEY t1=[rk1],2*KSZ // 0/1:rk[1] + (p0) and te32=s2,maskff // 0/1:s0&0xff + (p0) shr.u te00=s0,twenty4 };; // 0/0:s0>>24 +{ .mmi; (p0) LDKEY t2=[rk0],2*KSZ // 1/2:rk[2] + (p0) shladd te31=te31,3,te3 // 1/0:te0+s0>>24 + (p0) extr.u te23=s3,8,8 } // 1/1:s3>>8&0xff +{ .mmi; (p0) LDKEY t3=[rk1],2*KSZ // 1/3:rk[3] + (p0) shladd te32=te32,3,te3 // 1/1:te3+s0 + (p0) shr.u te01=s1,twenty4 };; // 1/1:s1>>24 +{ .mmi; (p0) ld4 te31=[te31] // 2/0:te3[s3&0xff] + (p0) shladd te22=te22,3,te2 // 2/0:te2+s2>>8&0xff + (p0) extr.u te20=s0,8,8 } // 2/2:s0>>8&0xff +{ .mmi; (p0) ld4 te32=[te32] // 2/1:te3[s0] + (p0) shladd te23=te23,3,te2 // 2/1:te2+s3>>8 + (p0) shr.u te02=s2,twenty4 };; // 2/2:s2>>24 +{ .mmi; (p0) ld4 te22=[te22] // 3/0:te2[s2>>8] + (p0) shladd te20=te20,3,te2 // 3/2:te2+s0>>8 + (p0) extr.u te21=s1,8,8 } // 3/3:s1>>8&0xff +{ .mmi; (p0) ld4 te23=[te23] // 3/1:te2[s3>>8] + (p0) shladd te00=te00,3,te0 // 3/0:te0+s0>>24 + (p0) shr.u te03=s3,twenty4 };; // 3/3:s3>>24 +{ .mmi; (p0) ld4 te20=[te20] // 4/2:te2[s0>>8] + (p0) shladd te21=te21,3,te2 // 4/3:te3+s2 + (p0) extr.u te13=s3,16,8 } // 4/0:s1>>16&0xff +{ .mmi; (p0) ld4 te00=[te00] // 4/0:te0[s0>>24] + (p0) shladd te01=te01,3,te0 // 4/1:te0+s1>>24 + (p0) shr.u te11=s1,sixteen };; // 4/2:s3>>16 +{ .mmi; (p0) ld4 te21=[te21] // 5/3:te2[s1>>8] + (p0) shladd te13=te13,3,te1 // 5/0:te1+s1>>16 + (p0) extr.u te10=s0,16,8 } // 5/1:s2>>16&0xff +{ .mmi; (p0) ld4 te01=[te01] // 5/1:te0[s1>>24] + (p0) shladd te02=te02,3,te0 // 5/2:te0+s2>>24 + (p0) and te33=s3,maskff };; // 5/2:s1&0xff +{ .mmi; (p0) ld4 te13=[te13] // 6/0:te1[s1>>16] + (p0) shladd te10=te10,3,te1 // 6/1:te1+s2>>16 + (p0) extr.u te12=s2,16,8 } // 6/3:s0>>16&0xff +{ .mmi; (p0) ld4 te02=[te02] // 6/2:te0[s2>>24] + (p0) shladd te03=te03,3,te0 // 6/3:te1+s0>>16 + (p0) and te30=s0,maskff };; // 6/3:s2&0xff + +{ .mmi; (p0) ld4 te10=[te10] // 7/1:te1[s2>>16] + (p0) shladd te33=te33,3,te3 // 7/2:te3+s1&0xff + (p0) and te11=te11,maskff} // 7/2:s3>>16&0xff +{ .mmi; (p0) ld4 te03=[te03] // 7/3:te0[s3>>24] + (p0) shladd te30=te30,3,te3 // 7/3:te3+s2 + (p0) xor t0=t0,te31 };; // 7/0: +{ .mmi; (p0) ld4 te33=[te33] // 8/2:te3[s1] + (p0) shladd te11=te11,3,te1 // 8/2:te1+s3>>16 + (p0) xor t0=t0,te22 } // 8/0: +{ .mmi; (p0) ld4 te30=[te30] // 8/3:te3[s2] + (p0) shladd te12=te12,3,te1 // 8/3:te1+s0>>16 + (p0) xor t1=t1,te32 };; // 8/1: +{ .mmi; (p0) ld4 te11=[te11] // 9/2:te1[s3>>16] + (p0) ld4 te12=[te12] // 9/3:te1[s0>>16] + (p0) xor t0=t0,te00 };; // 9/0: !L2 scheduling +{ .mmi; (p0) xor t1=t1,te23 // 10[9]/1: + (p0) xor t2=t2,te20 // 10[9]/2: + (p0) xor t3=t3,te21 };; // 10[9]/3: +{ .mmi; (p0) xor t0=t0,te13 // 11[10]/0:done! + (p0) xor t1=t1,te01 // 11[10]/1: + (p0) xor t2=t2,te02 };; // 11[10]/2: !L2 scheduling +{ .mmi; (p0) xor t3=t3,te03 // 12[10]/3: + (p16) cmp.eq p0,p17=r0,r0 };; // 12[10]/clear (p17) +{ .mmi; (p0) xor t1=t1,te10 // 13[11]/1:done! + (p0) xor t2=t2,te33 // 13[11]/2: + (p0) xor t3=t3,te30 } // 13[11]/3: +{ .mmi; (p17) add te0=2048,te0 // 13[11]/ + (p17) add te1=2048+64-TE1,te1};; // 13[11]/ +{ .mib; (p0) xor t2=t2,te11 // 14[12]/2:done! + (p17) add te2=2048+128-TE2,te2} // 14[12]/ +{ .mib; (p0) xor t3=t3,te12 // 14[12]/3:done! + (p17) add te3=2048+192-TE3,te3 // 14[12]/ + br.ctop.sptk .Ld_top };; +.Ld_end: + + +{ .mmi; ld8 te10=[te0] // prefetch Td4 + ld8 te33=[te1] } +{ .mmi; ld8 te12=[te2] + ld8 te30=[te3] } + +{ .mmi; LDKEY t0=[rk0],2*KSZ // 0/0:rk[0] + and te31=s1,maskff // 0/0:s3&0xff + extr.u te22=s2,8,8 } // 0/0:s2>>8&0xff +{ .mmi; LDKEY t1=[rk1],2*KSZ // 0/1:rk[1] + and te32=s2,maskff // 0/1:s0&0xff + shr.u te00=s0,twenty4 };; // 0/0:s0>>24 +{ .mmi; LDKEY t2=[rk0],2*KSZ // 1/2:rk[2] + add te31=te31,te0 // 1/0:te0+s0>>24 + extr.u te23=s3,8,8 } // 1/1:s3>>8&0xff +{ .mmi; LDKEY t3=[rk1],2*KSZ // 1/3:rk[3] + add te32=te32,te0 // 1/1:te0+s0 + shr.u te01=s1,twenty4 };; // 1/1:s1>>24 +{ .mmi; ld1 te31=[te31] // 2/0:te0[s3&0xff] + add te22=te22,te0 // 2/0:te0+s2>>8&0xff + extr.u te20=s0,8,8 } // 2/2:s0>>8&0xff +{ .mmi; ld1 te32=[te32] // 2/1:te0[s0] + add te23=te23,te0 // 2/1:te0+s3>>8 + shr.u te02=s2,twenty4 };; // 2/2:s2>>24 +{ .mmi; ld1 te22=[te22] // 3/0:te0[s2>>8] + add te20=te20,te0 // 3/2:te0+s0>>8 + extr.u te21=s1,8,8 } // 3/3:s1>>8&0xff +{ .mmi; ld1 te23=[te23] // 3/1:te0[s3>>8] + add te00=te00,te0 // 3/0:te0+s0>>24 + shr.u te03=s3,twenty4 };; // 3/3:s3>>24 +{ .mmi; ld1 te20=[te20] // 4/2:te0[s0>>8] + add te21=te21,te0 // 4/3:te0+s2 + extr.u te13=s3,16,8 } // 4/0:s1>>16&0xff +{ .mmi; ld1 te00=[te00] // 4/0:te0[s0>>24] + add te01=te01,te0 // 4/1:te0+s1>>24 + shr.u te11=s1,sixteen };; // 4/2:s3>>16 +{ .mmi; ld1 te21=[te21] // 5/3:te0[s1>>8] + add te13=te13,te0 // 5/0:te0+s1>>16 + extr.u te10=s0,16,8 } // 5/1:s2>>16&0xff +{ .mmi; ld1 te01=[te01] // 5/1:te0[s1>>24] + add te02=te02,te0 // 5/2:te0+s2>>24 + and te33=s3,maskff };; // 5/2:s1&0xff +{ .mmi; ld1 te13=[te13] // 6/0:te0[s1>>16] + add te10=te10,te0 // 6/1:te0+s2>>16 + extr.u te12=s2,16,8 } // 6/3:s0>>16&0xff +{ .mmi; ld1 te02=[te02] // 6/2:te0[s2>>24] + add te03=te03,te0 // 6/3:te0+s0>>16 + and te30=s0,maskff };; // 6/3:s2&0xff + +{ .mmi; ld1 te10=[te10] // 7/1:te0[s2>>16] + add te33=te33,te0 // 7/2:te0+s1&0xff + dep te31=te22,te31,8,8} // 7/0: +{ .mmi; ld1 te03=[te03] // 7/3:te0[s3>>24] + add te30=te30,te0 // 7/3:te0+s2 + and te11=te11,maskff};; // 7/2:s3>>16&0xff +{ .mmi; ld1 te33=[te33] // 8/2:te0[s1] + add te11=te11,te0 // 8/2:te0+s3>>16 + dep te32=te23,te32,8,8} // 8/1: +{ .mmi; ld1 te30=[te30] // 8/3:te0[s2] + add te12=te12,te0 // 8/3:te0+s0>>16 + shl te00=te00,twenty4};; // 8/0: +{ .mii; ld1 te11=[te11] // 9/2:te0[s3>>16] + dep te31=te13,te31,16,8 // 9/0: + shl te01=te01,twenty4};; // 9/1: +{ .mii; ld1 te12=[te12] // 10/3:te0[s0>>16] + dep te33=te20,te33,8,8 // 10/2: + shl te02=te02,twenty4};; // 10/2: +{ .mii; xor t0=t0,te31 // 11/0: + dep te30=te21,te30,8,8 // 11/3: + shl te10=te10,sixteen};; // 11/1: +{ .mii; xor r16=t0,te00 // 12/0:done! + dep te33=te11,te33,16,8 // 12/2: + shl te03=te03,twenty4};; // 12/3: +{ .mmi; xor t1=t1,te01 // 13/1: + xor t2=t2,te02 // 13/2: + dep te30=te12,te30,16,8};; // 13/3: +{ .mmi; xor t1=t1,te32 // 14/1: + xor r24=t2,te33 // 14/2:done! + xor t3=t3,te30 };; // 14/3: +{ .mib; xor r20=t1,te10 // 15/1:done! + xor r28=t3,te03 // 15/3:done! + br.ret.sptk b6 };; +.endp _ia64_AES_decrypt# + +// void AES_decrypt (const void *in,void *out,const AES_KEY *key); +.global AES_decrypt# +.proc AES_decrypt# +.align 32 +AES_decrypt: + .prologue + .save ar.pfs,pfssave +{ .mmi; alloc pfssave=ar.pfs,3,1,12,0 + and out0=3,in0 + mov r3=ip } +{ .mmi; ADDP in0=0,in0 + mov loc0=psr.um + ADDP out11=KSZ*60,in2 };; // &AES_KEY->rounds + +{ .mmi; ld4 out11=[out11] // AES_KEY->rounds + add out8=(AES_Td#-AES_decrypt#),r3 // Te0 + .save pr,prsave + mov prsave=pr } +{ .mmi; rum 1<<3 // clear um.ac + .save ar.lc,lcsave + mov lcsave=ar.lc };; + + .body +#if defined(_HPUX_SOURCE) // HPUX is big-endian, cut 15+15 cycles... +{ .mib; cmp.ne p6,p0=out0,r0 + add out0=4,in0 +(p6) br.dpnt.many .Ld_i_unaligned };; + +{ .mmi; ld4 out1=[in0],8 // s0 + and out9=3,in1 + mov twenty4=24 } +{ .mmi; ld4 out3=[out0],8 // s1 + ADDP rk0=0,in2 + mov sixteen=16 };; +{ .mmi; ld4 out5=[in0] // s2 + cmp.ne p6,p0=out9,r0 + mov maskff=0xff } +{ .mmb; ld4 out7=[out0] // s3 + ADDP rk1=KSZ,in2 + br.call.sptk.many b6=_ia64_AES_decrypt };; + +{ .mib; ADDP in0=4,in1 + ADDP in1=0,in1 +(p6) br.spnt .Ld_o_unaligned };; + +{ .mii; mov psr.um=loc0 + mov ar.pfs=pfssave + mov ar.lc=lcsave };; +{ .mmi; st4 [in1]=r16,8 // s0 + st4 [in0]=r20,8 // s1 + mov pr=prsave,0x1ffff };; +{ .mmb; st4 [in1]=r24 // s2 + st4 [in0]=r28 // s3 + br.ret.sptk.many b0 };; +#endif + +.align 32 +.Ld_i_unaligned: +{ .mmi; add out0=1,in0 + add out2=2,in0 + add out4=3,in0 };; +{ .mmi; ld1 r16=[in0],4 + ld1 r17=[out0],4 }//;; +{ .mmi; ld1 r18=[out2],4 + ld1 out1=[out4],4 };; // s0 +{ .mmi; ld1 r20=[in0],4 + ld1 r21=[out0],4 }//;; +{ .mmi; ld1 r22=[out2],4 + ld1 out3=[out4],4 };; // s1 +{ .mmi; ld1 r24=[in0],4 + ld1 r25=[out0],4 }//;; +{ .mmi; ld1 r26=[out2],4 + ld1 out5=[out4],4 };; // s2 +{ .mmi; ld1 r28=[in0] + ld1 r29=[out0] }//;; +{ .mmi; ld1 r30=[out2] + ld1 out7=[out4] };; // s3 + +{ .mii; + dep out1=r16,out1,24,8 //;; + dep out3=r20,out3,24,8 }//;; +{ .mii; ADDP rk0=0,in2 + dep out5=r24,out5,24,8 //;; + dep out7=r28,out7,24,8 };; +{ .mii; ADDP rk1=KSZ,in2 + dep out1=r17,out1,16,8 //;; + dep out3=r21,out3,16,8 }//;; +{ .mii; mov twenty4=24 + dep out5=r25,out5,16,8 //;; + dep out7=r29,out7,16,8 };; +{ .mii; mov sixteen=16 + dep out1=r18,out1,8,8 //;; + dep out3=r22,out3,8,8 }//;; +{ .mii; mov maskff=0xff + dep out5=r26,out5,8,8 //;; + dep out7=r30,out7,8,8 };; + +{ .mib; br.call.sptk.many b6=_ia64_AES_decrypt };; + +.Ld_o_unaligned: +{ .mii; ADDP out0=0,in1 + extr.u r17=r16,8,8 // s0 + shr.u r19=r16,twenty4 }//;; +{ .mii; ADDP out1=1,in1 + extr.u r18=r16,16,8 + shr.u r23=r20,twenty4 }//;; // s1 +{ .mii; ADDP out2=2,in1 + extr.u r21=r20,8,8 + shr.u r22=r20,sixteen }//;; +{ .mii; ADDP out3=3,in1 + extr.u r25=r24,8,8 // s2 + shr.u r27=r24,twenty4 };; +{ .mii; st1 [out3]=r16,4 + extr.u r26=r24,16,8 + shr.u r31=r28,twenty4 }//;; // s3 +{ .mii; st1 [out2]=r17,4 + extr.u r29=r28,8,8 + shr.u r30=r28,sixteen }//;; + +{ .mmi; st1 [out1]=r18,4 + st1 [out0]=r19,4 };; +{ .mmi; st1 [out3]=r20,4 + st1 [out2]=r21,4 }//;; +{ .mmi; st1 [out1]=r22,4 + st1 [out0]=r23,4 };; +{ .mmi; st1 [out3]=r24,4 + st1 [out2]=r25,4 + mov pr=prsave,0x1ffff }//;; +{ .mmi; st1 [out1]=r26,4 + st1 [out0]=r27,4 + mov ar.pfs=pfssave };; +{ .mmi; st1 [out3]=r28 + st1 [out2]=r29 + mov ar.lc=lcsave }//;; +{ .mmi; st1 [out1]=r30 + st1 [out0]=r31 } +{ .mfb; mov psr.um=loc0 // restore user mask + br.ret.sptk.many b0 };; +.endp AES_decrypt# + +// leave it in .text segment... +.align 64 +.global AES_Te# +.type AES_Te#,@object +AES_Te: data4 0xc66363a5,0xc66363a5, 0xf87c7c84,0xf87c7c84 + data4 0xee777799,0xee777799, 0xf67b7b8d,0xf67b7b8d + data4 0xfff2f20d,0xfff2f20d, 0xd66b6bbd,0xd66b6bbd + data4 0xde6f6fb1,0xde6f6fb1, 0x91c5c554,0x91c5c554 + data4 0x60303050,0x60303050, 0x02010103,0x02010103 + data4 0xce6767a9,0xce6767a9, 0x562b2b7d,0x562b2b7d + data4 0xe7fefe19,0xe7fefe19, 0xb5d7d762,0xb5d7d762 + data4 0x4dababe6,0x4dababe6, 0xec76769a,0xec76769a + data4 0x8fcaca45,0x8fcaca45, 0x1f82829d,0x1f82829d + data4 0x89c9c940,0x89c9c940, 0xfa7d7d87,0xfa7d7d87 + data4 0xeffafa15,0xeffafa15, 0xb25959eb,0xb25959eb + data4 0x8e4747c9,0x8e4747c9, 0xfbf0f00b,0xfbf0f00b + data4 0x41adadec,0x41adadec, 0xb3d4d467,0xb3d4d467 + data4 0x5fa2a2fd,0x5fa2a2fd, 0x45afafea,0x45afafea + data4 0x239c9cbf,0x239c9cbf, 0x53a4a4f7,0x53a4a4f7 + data4 0xe4727296,0xe4727296, 0x9bc0c05b,0x9bc0c05b + data4 0x75b7b7c2,0x75b7b7c2, 0xe1fdfd1c,0xe1fdfd1c + data4 0x3d9393ae,0x3d9393ae, 0x4c26266a,0x4c26266a + data4 0x6c36365a,0x6c36365a, 0x7e3f3f41,0x7e3f3f41 + data4 0xf5f7f702,0xf5f7f702, 0x83cccc4f,0x83cccc4f + data4 0x6834345c,0x6834345c, 0x51a5a5f4,0x51a5a5f4 + data4 0xd1e5e534,0xd1e5e534, 0xf9f1f108,0xf9f1f108 + data4 0xe2717193,0xe2717193, 0xabd8d873,0xabd8d873 + data4 0x62313153,0x62313153, 0x2a15153f,0x2a15153f + data4 0x0804040c,0x0804040c, 0x95c7c752,0x95c7c752 + data4 0x46232365,0x46232365, 0x9dc3c35e,0x9dc3c35e + data4 0x30181828,0x30181828, 0x379696a1,0x379696a1 + data4 0x0a05050f,0x0a05050f, 0x2f9a9ab5,0x2f9a9ab5 + data4 0x0e070709,0x0e070709, 0x24121236,0x24121236 + data4 0x1b80809b,0x1b80809b, 0xdfe2e23d,0xdfe2e23d + data4 0xcdebeb26,0xcdebeb26, 0x4e272769,0x4e272769 + data4 0x7fb2b2cd,0x7fb2b2cd, 0xea75759f,0xea75759f + data4 0x1209091b,0x1209091b, 0x1d83839e,0x1d83839e + data4 0x582c2c74,0x582c2c74, 0x341a1a2e,0x341a1a2e + data4 0x361b1b2d,0x361b1b2d, 0xdc6e6eb2,0xdc6e6eb2 + data4 0xb45a5aee,0xb45a5aee, 0x5ba0a0fb,0x5ba0a0fb + data4 0xa45252f6,0xa45252f6, 0x763b3b4d,0x763b3b4d + data4 0xb7d6d661,0xb7d6d661, 0x7db3b3ce,0x7db3b3ce + data4 0x5229297b,0x5229297b, 0xdde3e33e,0xdde3e33e + data4 0x5e2f2f71,0x5e2f2f71, 0x13848497,0x13848497 + data4 0xa65353f5,0xa65353f5, 0xb9d1d168,0xb9d1d168 + data4 0x00000000,0x00000000, 0xc1eded2c,0xc1eded2c + data4 0x40202060,0x40202060, 0xe3fcfc1f,0xe3fcfc1f + data4 0x79b1b1c8,0x79b1b1c8, 0xb65b5bed,0xb65b5bed + data4 0xd46a6abe,0xd46a6abe, 0x8dcbcb46,0x8dcbcb46 + data4 0x67bebed9,0x67bebed9, 0x7239394b,0x7239394b + data4 0x944a4ade,0x944a4ade, 0x984c4cd4,0x984c4cd4 + data4 0xb05858e8,0xb05858e8, 0x85cfcf4a,0x85cfcf4a + data4 0xbbd0d06b,0xbbd0d06b, 0xc5efef2a,0xc5efef2a + data4 0x4faaaae5,0x4faaaae5, 0xedfbfb16,0xedfbfb16 + data4 0x864343c5,0x864343c5, 0x9a4d4dd7,0x9a4d4dd7 + data4 0x66333355,0x66333355, 0x11858594,0x11858594 + data4 0x8a4545cf,0x8a4545cf, 0xe9f9f910,0xe9f9f910 + data4 0x04020206,0x04020206, 0xfe7f7f81,0xfe7f7f81 + data4 0xa05050f0,0xa05050f0, 0x783c3c44,0x783c3c44 + data4 0x259f9fba,0x259f9fba, 0x4ba8a8e3,0x4ba8a8e3 + data4 0xa25151f3,0xa25151f3, 0x5da3a3fe,0x5da3a3fe + data4 0x804040c0,0x804040c0, 0x058f8f8a,0x058f8f8a + data4 0x3f9292ad,0x3f9292ad, 0x219d9dbc,0x219d9dbc + data4 0x70383848,0x70383848, 0xf1f5f504,0xf1f5f504 + data4 0x63bcbcdf,0x63bcbcdf, 0x77b6b6c1,0x77b6b6c1 + data4 0xafdada75,0xafdada75, 0x42212163,0x42212163 + data4 0x20101030,0x20101030, 0xe5ffff1a,0xe5ffff1a + data4 0xfdf3f30e,0xfdf3f30e, 0xbfd2d26d,0xbfd2d26d + data4 0x81cdcd4c,0x81cdcd4c, 0x180c0c14,0x180c0c14 + data4 0x26131335,0x26131335, 0xc3ecec2f,0xc3ecec2f + data4 0xbe5f5fe1,0xbe5f5fe1, 0x359797a2,0x359797a2 + data4 0x884444cc,0x884444cc, 0x2e171739,0x2e171739 + data4 0x93c4c457,0x93c4c457, 0x55a7a7f2,0x55a7a7f2 + data4 0xfc7e7e82,0xfc7e7e82, 0x7a3d3d47,0x7a3d3d47 + data4 0xc86464ac,0xc86464ac, 0xba5d5de7,0xba5d5de7 + data4 0x3219192b,0x3219192b, 0xe6737395,0xe6737395 + data4 0xc06060a0,0xc06060a0, 0x19818198,0x19818198 + data4 0x9e4f4fd1,0x9e4f4fd1, 0xa3dcdc7f,0xa3dcdc7f + data4 0x44222266,0x44222266, 0x542a2a7e,0x542a2a7e + data4 0x3b9090ab,0x3b9090ab, 0x0b888883,0x0b888883 + data4 0x8c4646ca,0x8c4646ca, 0xc7eeee29,0xc7eeee29 + data4 0x6bb8b8d3,0x6bb8b8d3, 0x2814143c,0x2814143c + data4 0xa7dede79,0xa7dede79, 0xbc5e5ee2,0xbc5e5ee2 + data4 0x160b0b1d,0x160b0b1d, 0xaddbdb76,0xaddbdb76 + data4 0xdbe0e03b,0xdbe0e03b, 0x64323256,0x64323256 + data4 0x743a3a4e,0x743a3a4e, 0x140a0a1e,0x140a0a1e + data4 0x924949db,0x924949db, 0x0c06060a,0x0c06060a + data4 0x4824246c,0x4824246c, 0xb85c5ce4,0xb85c5ce4 + data4 0x9fc2c25d,0x9fc2c25d, 0xbdd3d36e,0xbdd3d36e + data4 0x43acacef,0x43acacef, 0xc46262a6,0xc46262a6 + data4 0x399191a8,0x399191a8, 0x319595a4,0x319595a4 + data4 0xd3e4e437,0xd3e4e437, 0xf279798b,0xf279798b + data4 0xd5e7e732,0xd5e7e732, 0x8bc8c843,0x8bc8c843 + data4 0x6e373759,0x6e373759, 0xda6d6db7,0xda6d6db7 + data4 0x018d8d8c,0x018d8d8c, 0xb1d5d564,0xb1d5d564 + data4 0x9c4e4ed2,0x9c4e4ed2, 0x49a9a9e0,0x49a9a9e0 + data4 0xd86c6cb4,0xd86c6cb4, 0xac5656fa,0xac5656fa + data4 0xf3f4f407,0xf3f4f407, 0xcfeaea25,0xcfeaea25 + data4 0xca6565af,0xca6565af, 0xf47a7a8e,0xf47a7a8e + data4 0x47aeaee9,0x47aeaee9, 0x10080818,0x10080818 + data4 0x6fbabad5,0x6fbabad5, 0xf0787888,0xf0787888 + data4 0x4a25256f,0x4a25256f, 0x5c2e2e72,0x5c2e2e72 + data4 0x381c1c24,0x381c1c24, 0x57a6a6f1,0x57a6a6f1 + data4 0x73b4b4c7,0x73b4b4c7, 0x97c6c651,0x97c6c651 + data4 0xcbe8e823,0xcbe8e823, 0xa1dddd7c,0xa1dddd7c + data4 0xe874749c,0xe874749c, 0x3e1f1f21,0x3e1f1f21 + data4 0x964b4bdd,0x964b4bdd, 0x61bdbddc,0x61bdbddc + data4 0x0d8b8b86,0x0d8b8b86, 0x0f8a8a85,0x0f8a8a85 + data4 0xe0707090,0xe0707090, 0x7c3e3e42,0x7c3e3e42 + data4 0x71b5b5c4,0x71b5b5c4, 0xcc6666aa,0xcc6666aa + data4 0x904848d8,0x904848d8, 0x06030305,0x06030305 + data4 0xf7f6f601,0xf7f6f601, 0x1c0e0e12,0x1c0e0e12 + data4 0xc26161a3,0xc26161a3, 0x6a35355f,0x6a35355f + data4 0xae5757f9,0xae5757f9, 0x69b9b9d0,0x69b9b9d0 + data4 0x17868691,0x17868691, 0x99c1c158,0x99c1c158 + data4 0x3a1d1d27,0x3a1d1d27, 0x279e9eb9,0x279e9eb9 + data4 0xd9e1e138,0xd9e1e138, 0xebf8f813,0xebf8f813 + data4 0x2b9898b3,0x2b9898b3, 0x22111133,0x22111133 + data4 0xd26969bb,0xd26969bb, 0xa9d9d970,0xa9d9d970 + data4 0x078e8e89,0x078e8e89, 0x339494a7,0x339494a7 + data4 0x2d9b9bb6,0x2d9b9bb6, 0x3c1e1e22,0x3c1e1e22 + data4 0x15878792,0x15878792, 0xc9e9e920,0xc9e9e920 + data4 0x87cece49,0x87cece49, 0xaa5555ff,0xaa5555ff + data4 0x50282878,0x50282878, 0xa5dfdf7a,0xa5dfdf7a + data4 0x038c8c8f,0x038c8c8f, 0x59a1a1f8,0x59a1a1f8 + data4 0x09898980,0x09898980, 0x1a0d0d17,0x1a0d0d17 + data4 0x65bfbfda,0x65bfbfda, 0xd7e6e631,0xd7e6e631 + data4 0x844242c6,0x844242c6, 0xd06868b8,0xd06868b8 + data4 0x824141c3,0x824141c3, 0x299999b0,0x299999b0 + data4 0x5a2d2d77,0x5a2d2d77, 0x1e0f0f11,0x1e0f0f11 + data4 0x7bb0b0cb,0x7bb0b0cb, 0xa85454fc,0xa85454fc + data4 0x6dbbbbd6,0x6dbbbbd6, 0x2c16163a,0x2c16163a +// Te4: + data1 0x63, 0x7c, 0x77, 0x7b, 0xf2, 0x6b, 0x6f, 0xc5 + data1 0x30, 0x01, 0x67, 0x2b, 0xfe, 0xd7, 0xab, 0x76 + data1 0xca, 0x82, 0xc9, 0x7d, 0xfa, 0x59, 0x47, 0xf0 + data1 0xad, 0xd4, 0xa2, 0xaf, 0x9c, 0xa4, 0x72, 0xc0 + data1 0xb7, 0xfd, 0x93, 0x26, 0x36, 0x3f, 0xf7, 0xcc + data1 0x34, 0xa5, 0xe5, 0xf1, 0x71, 0xd8, 0x31, 0x15 + data1 0x04, 0xc7, 0x23, 0xc3, 0x18, 0x96, 0x05, 0x9a + data1 0x07, 0x12, 0x80, 0xe2, 0xeb, 0x27, 0xb2, 0x75 + data1 0x09, 0x83, 0x2c, 0x1a, 0x1b, 0x6e, 0x5a, 0xa0 + data1 0x52, 0x3b, 0xd6, 0xb3, 0x29, 0xe3, 0x2f, 0x84 + data1 0x53, 0xd1, 0x00, 0xed, 0x20, 0xfc, 0xb1, 0x5b + data1 0x6a, 0xcb, 0xbe, 0x39, 0x4a, 0x4c, 0x58, 0xcf + data1 0xd0, 0xef, 0xaa, 0xfb, 0x43, 0x4d, 0x33, 0x85 + data1 0x45, 0xf9, 0x02, 0x7f, 0x50, 0x3c, 0x9f, 0xa8 + data1 0x51, 0xa3, 0x40, 0x8f, 0x92, 0x9d, 0x38, 0xf5 + data1 0xbc, 0xb6, 0xda, 0x21, 0x10, 0xff, 0xf3, 0xd2 + data1 0xcd, 0x0c, 0x13, 0xec, 0x5f, 0x97, 0x44, 0x17 + data1 0xc4, 0xa7, 0x7e, 0x3d, 0x64, 0x5d, 0x19, 0x73 + data1 0x60, 0x81, 0x4f, 0xdc, 0x22, 0x2a, 0x90, 0x88 + data1 0x46, 0xee, 0xb8, 0x14, 0xde, 0x5e, 0x0b, 0xdb + data1 0xe0, 0x32, 0x3a, 0x0a, 0x49, 0x06, 0x24, 0x5c + data1 0xc2, 0xd3, 0xac, 0x62, 0x91, 0x95, 0xe4, 0x79 + data1 0xe7, 0xc8, 0x37, 0x6d, 0x8d, 0xd5, 0x4e, 0xa9 + data1 0x6c, 0x56, 0xf4, 0xea, 0x65, 0x7a, 0xae, 0x08 + data1 0xba, 0x78, 0x25, 0x2e, 0x1c, 0xa6, 0xb4, 0xc6 + data1 0xe8, 0xdd, 0x74, 0x1f, 0x4b, 0xbd, 0x8b, 0x8a + data1 0x70, 0x3e, 0xb5, 0x66, 0x48, 0x03, 0xf6, 0x0e + data1 0x61, 0x35, 0x57, 0xb9, 0x86, 0xc1, 0x1d, 0x9e + data1 0xe1, 0xf8, 0x98, 0x11, 0x69, 0xd9, 0x8e, 0x94 + data1 0x9b, 0x1e, 0x87, 0xe9, 0xce, 0x55, 0x28, 0xdf + data1 0x8c, 0xa1, 0x89, 0x0d, 0xbf, 0xe6, 0x42, 0x68 + data1 0x41, 0x99, 0x2d, 0x0f, 0xb0, 0x54, 0xbb, 0x16 +.size AES_Te#,2048+256 // HP-UX assembler fails to ".-AES_Te#" + +.align 64 +.global AES_Td# +.type AES_Td#,@object +AES_Td: data4 0x51f4a750,0x51f4a750, 0x7e416553,0x7e416553 + data4 0x1a17a4c3,0x1a17a4c3, 0x3a275e96,0x3a275e96 + data4 0x3bab6bcb,0x3bab6bcb, 0x1f9d45f1,0x1f9d45f1 + data4 0xacfa58ab,0xacfa58ab, 0x4be30393,0x4be30393 + data4 0x2030fa55,0x2030fa55, 0xad766df6,0xad766df6 + data4 0x88cc7691,0x88cc7691, 0xf5024c25,0xf5024c25 + data4 0x4fe5d7fc,0x4fe5d7fc, 0xc52acbd7,0xc52acbd7 + data4 0x26354480,0x26354480, 0xb562a38f,0xb562a38f + data4 0xdeb15a49,0xdeb15a49, 0x25ba1b67,0x25ba1b67 + data4 0x45ea0e98,0x45ea0e98, 0x5dfec0e1,0x5dfec0e1 + data4 0xc32f7502,0xc32f7502, 0x814cf012,0x814cf012 + data4 0x8d4697a3,0x8d4697a3, 0x6bd3f9c6,0x6bd3f9c6 + data4 0x038f5fe7,0x038f5fe7, 0x15929c95,0x15929c95 + data4 0xbf6d7aeb,0xbf6d7aeb, 0x955259da,0x955259da + data4 0xd4be832d,0xd4be832d, 0x587421d3,0x587421d3 + data4 0x49e06929,0x49e06929, 0x8ec9c844,0x8ec9c844 + data4 0x75c2896a,0x75c2896a, 0xf48e7978,0xf48e7978 + data4 0x99583e6b,0x99583e6b, 0x27b971dd,0x27b971dd + data4 0xbee14fb6,0xbee14fb6, 0xf088ad17,0xf088ad17 + data4 0xc920ac66,0xc920ac66, 0x7dce3ab4,0x7dce3ab4 + data4 0x63df4a18,0x63df4a18, 0xe51a3182,0xe51a3182 + data4 0x97513360,0x97513360, 0x62537f45,0x62537f45 + data4 0xb16477e0,0xb16477e0, 0xbb6bae84,0xbb6bae84 + data4 0xfe81a01c,0xfe81a01c, 0xf9082b94,0xf9082b94 + data4 0x70486858,0x70486858, 0x8f45fd19,0x8f45fd19 + data4 0x94de6c87,0x94de6c87, 0x527bf8b7,0x527bf8b7 + data4 0xab73d323,0xab73d323, 0x724b02e2,0x724b02e2 + data4 0xe31f8f57,0xe31f8f57, 0x6655ab2a,0x6655ab2a + data4 0xb2eb2807,0xb2eb2807, 0x2fb5c203,0x2fb5c203 + data4 0x86c57b9a,0x86c57b9a, 0xd33708a5,0xd33708a5 + data4 0x302887f2,0x302887f2, 0x23bfa5b2,0x23bfa5b2 + data4 0x02036aba,0x02036aba, 0xed16825c,0xed16825c + data4 0x8acf1c2b,0x8acf1c2b, 0xa779b492,0xa779b492 + data4 0xf307f2f0,0xf307f2f0, 0x4e69e2a1,0x4e69e2a1 + data4 0x65daf4cd,0x65daf4cd, 0x0605bed5,0x0605bed5 + data4 0xd134621f,0xd134621f, 0xc4a6fe8a,0xc4a6fe8a + data4 0x342e539d,0x342e539d, 0xa2f355a0,0xa2f355a0 + data4 0x058ae132,0x058ae132, 0xa4f6eb75,0xa4f6eb75 + data4 0x0b83ec39,0x0b83ec39, 0x4060efaa,0x4060efaa + data4 0x5e719f06,0x5e719f06, 0xbd6e1051,0xbd6e1051 + data4 0x3e218af9,0x3e218af9, 0x96dd063d,0x96dd063d + data4 0xdd3e05ae,0xdd3e05ae, 0x4de6bd46,0x4de6bd46 + data4 0x91548db5,0x91548db5, 0x71c45d05,0x71c45d05 + data4 0x0406d46f,0x0406d46f, 0x605015ff,0x605015ff + data4 0x1998fb24,0x1998fb24, 0xd6bde997,0xd6bde997 + data4 0x894043cc,0x894043cc, 0x67d99e77,0x67d99e77 + data4 0xb0e842bd,0xb0e842bd, 0x07898b88,0x07898b88 + data4 0xe7195b38,0xe7195b38, 0x79c8eedb,0x79c8eedb + data4 0xa17c0a47,0xa17c0a47, 0x7c420fe9,0x7c420fe9 + data4 0xf8841ec9,0xf8841ec9, 0x00000000,0x00000000 + data4 0x09808683,0x09808683, 0x322bed48,0x322bed48 + data4 0x1e1170ac,0x1e1170ac, 0x6c5a724e,0x6c5a724e + data4 0xfd0efffb,0xfd0efffb, 0x0f853856,0x0f853856 + data4 0x3daed51e,0x3daed51e, 0x362d3927,0x362d3927 + data4 0x0a0fd964,0x0a0fd964, 0x685ca621,0x685ca621 + data4 0x9b5b54d1,0x9b5b54d1, 0x24362e3a,0x24362e3a + data4 0x0c0a67b1,0x0c0a67b1, 0x9357e70f,0x9357e70f + data4 0xb4ee96d2,0xb4ee96d2, 0x1b9b919e,0x1b9b919e + data4 0x80c0c54f,0x80c0c54f, 0x61dc20a2,0x61dc20a2 + data4 0x5a774b69,0x5a774b69, 0x1c121a16,0x1c121a16 + data4 0xe293ba0a,0xe293ba0a, 0xc0a02ae5,0xc0a02ae5 + data4 0x3c22e043,0x3c22e043, 0x121b171d,0x121b171d + data4 0x0e090d0b,0x0e090d0b, 0xf28bc7ad,0xf28bc7ad + data4 0x2db6a8b9,0x2db6a8b9, 0x141ea9c8,0x141ea9c8 + data4 0x57f11985,0x57f11985, 0xaf75074c,0xaf75074c + data4 0xee99ddbb,0xee99ddbb, 0xa37f60fd,0xa37f60fd + data4 0xf701269f,0xf701269f, 0x5c72f5bc,0x5c72f5bc + data4 0x44663bc5,0x44663bc5, 0x5bfb7e34,0x5bfb7e34 + data4 0x8b432976,0x8b432976, 0xcb23c6dc,0xcb23c6dc + data4 0xb6edfc68,0xb6edfc68, 0xb8e4f163,0xb8e4f163 + data4 0xd731dcca,0xd731dcca, 0x42638510,0x42638510 + data4 0x13972240,0x13972240, 0x84c61120,0x84c61120 + data4 0x854a247d,0x854a247d, 0xd2bb3df8,0xd2bb3df8 + data4 0xaef93211,0xaef93211, 0xc729a16d,0xc729a16d + data4 0x1d9e2f4b,0x1d9e2f4b, 0xdcb230f3,0xdcb230f3 + data4 0x0d8652ec,0x0d8652ec, 0x77c1e3d0,0x77c1e3d0 + data4 0x2bb3166c,0x2bb3166c, 0xa970b999,0xa970b999 + data4 0x119448fa,0x119448fa, 0x47e96422,0x47e96422 + data4 0xa8fc8cc4,0xa8fc8cc4, 0xa0f03f1a,0xa0f03f1a + data4 0x567d2cd8,0x567d2cd8, 0x223390ef,0x223390ef + data4 0x87494ec7,0x87494ec7, 0xd938d1c1,0xd938d1c1 + data4 0x8ccaa2fe,0x8ccaa2fe, 0x98d40b36,0x98d40b36 + data4 0xa6f581cf,0xa6f581cf, 0xa57ade28,0xa57ade28 + data4 0xdab78e26,0xdab78e26, 0x3fadbfa4,0x3fadbfa4 + data4 0x2c3a9de4,0x2c3a9de4, 0x5078920d,0x5078920d + data4 0x6a5fcc9b,0x6a5fcc9b, 0x547e4662,0x547e4662 + data4 0xf68d13c2,0xf68d13c2, 0x90d8b8e8,0x90d8b8e8 + data4 0x2e39f75e,0x2e39f75e, 0x82c3aff5,0x82c3aff5 + data4 0x9f5d80be,0x9f5d80be, 0x69d0937c,0x69d0937c + data4 0x6fd52da9,0x6fd52da9, 0xcf2512b3,0xcf2512b3 + data4 0xc8ac993b,0xc8ac993b, 0x10187da7,0x10187da7 + data4 0xe89c636e,0xe89c636e, 0xdb3bbb7b,0xdb3bbb7b + data4 0xcd267809,0xcd267809, 0x6e5918f4,0x6e5918f4 + data4 0xec9ab701,0xec9ab701, 0x834f9aa8,0x834f9aa8 + data4 0xe6956e65,0xe6956e65, 0xaaffe67e,0xaaffe67e + data4 0x21bccf08,0x21bccf08, 0xef15e8e6,0xef15e8e6 + data4 0xbae79bd9,0xbae79bd9, 0x4a6f36ce,0x4a6f36ce + data4 0xea9f09d4,0xea9f09d4, 0x29b07cd6,0x29b07cd6 + data4 0x31a4b2af,0x31a4b2af, 0x2a3f2331,0x2a3f2331 + data4 0xc6a59430,0xc6a59430, 0x35a266c0,0x35a266c0 + data4 0x744ebc37,0x744ebc37, 0xfc82caa6,0xfc82caa6 + data4 0xe090d0b0,0xe090d0b0, 0x33a7d815,0x33a7d815 + data4 0xf104984a,0xf104984a, 0x41ecdaf7,0x41ecdaf7 + data4 0x7fcd500e,0x7fcd500e, 0x1791f62f,0x1791f62f + data4 0x764dd68d,0x764dd68d, 0x43efb04d,0x43efb04d + data4 0xccaa4d54,0xccaa4d54, 0xe49604df,0xe49604df + data4 0x9ed1b5e3,0x9ed1b5e3, 0x4c6a881b,0x4c6a881b + data4 0xc12c1fb8,0xc12c1fb8, 0x4665517f,0x4665517f + data4 0x9d5eea04,0x9d5eea04, 0x018c355d,0x018c355d + data4 0xfa877473,0xfa877473, 0xfb0b412e,0xfb0b412e + data4 0xb3671d5a,0xb3671d5a, 0x92dbd252,0x92dbd252 + data4 0xe9105633,0xe9105633, 0x6dd64713,0x6dd64713 + data4 0x9ad7618c,0x9ad7618c, 0x37a10c7a,0x37a10c7a + data4 0x59f8148e,0x59f8148e, 0xeb133c89,0xeb133c89 + data4 0xcea927ee,0xcea927ee, 0xb761c935,0xb761c935 + data4 0xe11ce5ed,0xe11ce5ed, 0x7a47b13c,0x7a47b13c + data4 0x9cd2df59,0x9cd2df59, 0x55f2733f,0x55f2733f + data4 0x1814ce79,0x1814ce79, 0x73c737bf,0x73c737bf + data4 0x53f7cdea,0x53f7cdea, 0x5ffdaa5b,0x5ffdaa5b + data4 0xdf3d6f14,0xdf3d6f14, 0x7844db86,0x7844db86 + data4 0xcaaff381,0xcaaff381, 0xb968c43e,0xb968c43e + data4 0x3824342c,0x3824342c, 0xc2a3405f,0xc2a3405f + data4 0x161dc372,0x161dc372, 0xbce2250c,0xbce2250c + data4 0x283c498b,0x283c498b, 0xff0d9541,0xff0d9541 + data4 0x39a80171,0x39a80171, 0x080cb3de,0x080cb3de + data4 0xd8b4e49c,0xd8b4e49c, 0x6456c190,0x6456c190 + data4 0x7bcb8461,0x7bcb8461, 0xd532b670,0xd532b670 + data4 0x486c5c74,0x486c5c74, 0xd0b85742,0xd0b85742 +// Td4: + data1 0x52, 0x09, 0x6a, 0xd5, 0x30, 0x36, 0xa5, 0x38 + data1 0xbf, 0x40, 0xa3, 0x9e, 0x81, 0xf3, 0xd7, 0xfb + data1 0x7c, 0xe3, 0x39, 0x82, 0x9b, 0x2f, 0xff, 0x87 + data1 0x34, 0x8e, 0x43, 0x44, 0xc4, 0xde, 0xe9, 0xcb + data1 0x54, 0x7b, 0x94, 0x32, 0xa6, 0xc2, 0x23, 0x3d + data1 0xee, 0x4c, 0x95, 0x0b, 0x42, 0xfa, 0xc3, 0x4e + data1 0x08, 0x2e, 0xa1, 0x66, 0x28, 0xd9, 0x24, 0xb2 + data1 0x76, 0x5b, 0xa2, 0x49, 0x6d, 0x8b, 0xd1, 0x25 + data1 0x72, 0xf8, 0xf6, 0x64, 0x86, 0x68, 0x98, 0x16 + data1 0xd4, 0xa4, 0x5c, 0xcc, 0x5d, 0x65, 0xb6, 0x92 + data1 0x6c, 0x70, 0x48, 0x50, 0xfd, 0xed, 0xb9, 0xda + data1 0x5e, 0x15, 0x46, 0x57, 0xa7, 0x8d, 0x9d, 0x84 + data1 0x90, 0xd8, 0xab, 0x00, 0x8c, 0xbc, 0xd3, 0x0a + data1 0xf7, 0xe4, 0x58, 0x05, 0xb8, 0xb3, 0x45, 0x06 + data1 0xd0, 0x2c, 0x1e, 0x8f, 0xca, 0x3f, 0x0f, 0x02 + data1 0xc1, 0xaf, 0xbd, 0x03, 0x01, 0x13, 0x8a, 0x6b + data1 0x3a, 0x91, 0x11, 0x41, 0x4f, 0x67, 0xdc, 0xea + data1 0x97, 0xf2, 0xcf, 0xce, 0xf0, 0xb4, 0xe6, 0x73 + data1 0x96, 0xac, 0x74, 0x22, 0xe7, 0xad, 0x35, 0x85 + data1 0xe2, 0xf9, 0x37, 0xe8, 0x1c, 0x75, 0xdf, 0x6e + data1 0x47, 0xf1, 0x1a, 0x71, 0x1d, 0x29, 0xc5, 0x89 + data1 0x6f, 0xb7, 0x62, 0x0e, 0xaa, 0x18, 0xbe, 0x1b + data1 0xfc, 0x56, 0x3e, 0x4b, 0xc6, 0xd2, 0x79, 0x20 + data1 0x9a, 0xdb, 0xc0, 0xfe, 0x78, 0xcd, 0x5a, 0xf4 + data1 0x1f, 0xdd, 0xa8, 0x33, 0x88, 0x07, 0xc7, 0x31 + data1 0xb1, 0x12, 0x10, 0x59, 0x27, 0x80, 0xec, 0x5f + data1 0x60, 0x51, 0x7f, 0xa9, 0x19, 0xb5, 0x4a, 0x0d + data1 0x2d, 0xe5, 0x7a, 0x9f, 0x93, 0xc9, 0x9c, 0xef + data1 0xa0, 0xe0, 0x3b, 0x4d, 0xae, 0x2a, 0xf5, 0xb0 + data1 0xc8, 0xeb, 0xbb, 0x3c, 0x83, 0x53, 0x99, 0x61 + data1 0x17, 0x2b, 0x04, 0x7e, 0xba, 0x77, 0xd6, 0x26 + data1 0xe1, 0x69, 0x14, 0x63, 0x55, 0x21, 0x0c, 0x7d +.size AES_Td#,2048+256 // HP-UX assembler fails to ".-AES_Td#" diff --git a/openssl-1.1.0h/crypto/aes/asm/aes-mips.pl b/openssl-1.1.0h/crypto/aes/asm/aes-mips.pl new file mode 100644 index 0000000..439578d --- /dev/null +++ b/openssl-1.1.0h/crypto/aes/asm/aes-mips.pl @@ -0,0 +1,2131 @@ +#! /usr/bin/env perl +# Copyright 2010-2016 The OpenSSL Project Authors. All Rights Reserved. +# +# Licensed under the OpenSSL license (the "License"). You may not use +# this file except in compliance with the License. You can obtain a copy +# in the file LICENSE in the source distribution or at +# https://www.openssl.org/source/license.html + + +# ==================================================================== +# Written by Andy Polyakov for the OpenSSL +# project. The module is, however, dual licensed under OpenSSL and +# CRYPTOGAMS licenses depending on where you obtain it. For further +# details see http://www.openssl.org/~appro/cryptogams/. +# ==================================================================== + +# AES for MIPS + +# October 2010 +# +# Code uses 1K[+256B] S-box and on single-issue core [such as R5000] +# spends ~68 cycles per byte processed with 128-bit key. This is ~16% +# faster than gcc-generated code, which is not very impressive. But +# recall that compressed S-box requires extra processing, namely +# additional rotations. Rotations are implemented with lwl/lwr pairs, +# which is normally used for loading unaligned data. Another cool +# thing about this module is its endian neutrality, which means that +# it processes data without ever changing byte order... + +# September 2012 +# +# Add MIPS32R2 (~10% less instructions) and SmartMIPS ASE (further +# ~25% less instructions) code. Note that there is no run-time switch, +# instead, code path is chosen upon pre-process time, pass -mips32r2 +# or/and -msmartmips. + +###################################################################### +# There is a number of MIPS ABI in use, O32 and N32/64 are most +# widely used. Then there is a new contender: NUBI. It appears that if +# one picks the latter, it's possible to arrange code in ABI neutral +# manner. Therefore let's stick to NUBI register layout: +# +($zero,$at,$t0,$t1,$t2)=map("\$$_",(0..2,24,25)); +($a0,$a1,$a2,$a3,$a4,$a5,$a6,$a7)=map("\$$_",(4..11)); +($s0,$s1,$s2,$s3,$s4,$s5,$s6,$s7,$s8,$s9,$s10,$s11)=map("\$$_",(12..23)); +($gp,$tp,$sp,$fp,$ra)=map("\$$_",(3,28..31)); +# +# The return value is placed in $a0. Following coding rules facilitate +# interoperability: +# +# - never ever touch $tp, "thread pointer", former $gp; +# - copy return value to $t0, former $v0 [or to $a0 if you're adapting +# old code]; +# - on O32 populate $a4-$a7 with 'lw $aN,4*N($sp)' if necessary; +# +# For reference here is register layout for N32/64 MIPS ABIs: +# +# ($zero,$at,$v0,$v1)=map("\$$_",(0..3)); +# ($a0,$a1,$a2,$a3,$a4,$a5,$a6,$a7)=map("\$$_",(4..11)); +# ($t0,$t1,$t2,$t3,$t8,$t9)=map("\$$_",(12..15,24,25)); +# ($s0,$s1,$s2,$s3,$s4,$s5,$s6,$s7)=map("\$$_",(16..23)); +# ($gp,$sp,$fp,$ra)=map("\$$_",(28..31)); +# +$flavour = shift || "o32"; # supported flavours are o32,n32,64,nubi32,nubi64 + +if ($flavour =~ /64|n32/i) { + $PTR_LA="dla"; + $PTR_ADD="dadd"; # incidentally works even on n32 + $PTR_SUB="dsub"; # incidentally works even on n32 + $PTR_INS="dins"; + $REG_S="sd"; + $REG_L="ld"; + $PTR_SLL="dsll"; # incidentally works even on n32 + $SZREG=8; +} else { + $PTR_LA="la"; + $PTR_ADD="add"; + $PTR_SUB="sub"; + $PTR_INS="ins"; + $REG_S="sw"; + $REG_L="lw"; + $PTR_SLL="sll"; + $SZREG=4; +} +$pf = ($flavour =~ /nubi/i) ? $t0 : $t2; +# +# +# +###################################################################### + +$big_endian=(`echo MIPSEL | $ENV{CC} -E -`=~/MIPSEL/)?1:0 if ($ENV{CC}); + +for (@ARGV) { $output=$_ if (/\w[\w\-]*\.\w+$/); } +open STDOUT,">$output"; + +if (!defined($big_endian)) +{ $big_endian=(unpack('L',pack('N',1))==1); } + +while (($output=shift) && ($output!~/\w[\w\-]*\.\w+$/)) {} +open STDOUT,">$output"; + +my ($MSB,$LSB)=(0,3); # automatically converted to little-endian + +$code.=<<___; +.text +#ifdef OPENSSL_FIPSCANISTER +# include +#endif + +#if defined(__mips_smartmips) && !defined(_MIPS_ARCH_MIPS32R2) +#define _MIPS_ARCH_MIPS32R2 +#endif + +#if !defined(__mips_eabi) && (!defined(__vxworks) || defined(__pic__)) +.option pic2 +#endif +.set noat +___ + +{{{ +my $FRAMESIZE=16*$SZREG; +my $SAVED_REGS_MASK = ($flavour =~ /nubi/i) ? "0xc0fff008" : "0xc0ff0000"; + +my ($inp,$out,$key,$Tbl,$s0,$s1,$s2,$s3)=($a0,$a1,$a2,$a3,$a4,$a5,$a6,$a7); +my ($i0,$i1,$i2,$i3)=($at,$t0,$t1,$t2); +my ($t0,$t1,$t2,$t3,$t4,$t5,$t6,$t7,$t8,$t9,$t10,$t11) = map("\$$_",(12..23)); +my ($key0,$cnt)=($gp,$fp); + +# instuction ordering is "stolen" from output from MIPSpro assembler +# invoked with -mips3 -O3 arguments... +$code.=<<___; +.align 5 +.ent _mips_AES_encrypt +_mips_AES_encrypt: + .frame $sp,0,$ra + .set reorder + lw $t0,0($key) + lw $t1,4($key) + lw $t2,8($key) + lw $t3,12($key) + lw $cnt,240($key) + $PTR_ADD $key0,$key,16 + + xor $s0,$t0 + xor $s1,$t1 + xor $s2,$t2 + xor $s3,$t3 + + sub $cnt,1 +#if defined(__mips_smartmips) + ext $i0,$s1,16,8 +.Loop_enc: + ext $i1,$s2,16,8 + ext $i2,$s3,16,8 + ext $i3,$s0,16,8 + lwxs $t0,$i0($Tbl) # Te1[s1>>16] + ext $i0,$s2,8,8 + lwxs $t1,$i1($Tbl) # Te1[s2>>16] + ext $i1,$s3,8,8 + lwxs $t2,$i2($Tbl) # Te1[s3>>16] + ext $i2,$s0,8,8 + lwxs $t3,$i3($Tbl) # Te1[s0>>16] + ext $i3,$s1,8,8 + + lwxs $t4,$i0($Tbl) # Te2[s2>>8] + ext $i0,$s3,0,8 + lwxs $t5,$i1($Tbl) # Te2[s3>>8] + ext $i1,$s0,0,8 + lwxs $t6,$i2($Tbl) # Te2[s0>>8] + ext $i2,$s1,0,8 + lwxs $t7,$i3($Tbl) # Te2[s1>>8] + ext $i3,$s2,0,8 + + lwxs $t8,$i0($Tbl) # Te3[s3] + ext $i0,$s0,24,8 + lwxs $t9,$i1($Tbl) # Te3[s0] + ext $i1,$s1,24,8 + lwxs $t10,$i2($Tbl) # Te3[s1] + ext $i2,$s2,24,8 + lwxs $t11,$i3($Tbl) # Te3[s2] + ext $i3,$s3,24,8 + + rotr $t0,$t0,8 + rotr $t1,$t1,8 + rotr $t2,$t2,8 + rotr $t3,$t3,8 + + rotr $t4,$t4,16 + rotr $t5,$t5,16 + rotr $t6,$t6,16 + rotr $t7,$t7,16 + + xor $t0,$t4 + lwxs $t4,$i0($Tbl) # Te0[s0>>24] + xor $t1,$t5 + lwxs $t5,$i1($Tbl) # Te0[s1>>24] + xor $t2,$t6 + lwxs $t6,$i2($Tbl) # Te0[s2>>24] + xor $t3,$t7 + lwxs $t7,$i3($Tbl) # Te0[s3>>24] + + rotr $t8,$t8,24 + lw $s0,0($key0) + rotr $t9,$t9,24 + lw $s1,4($key0) + rotr $t10,$t10,24 + lw $s2,8($key0) + rotr $t11,$t11,24 + lw $s3,12($key0) + + xor $t0,$t8 + xor $t1,$t9 + xor $t2,$t10 + xor $t3,$t11 + + xor $t0,$t4 + xor $t1,$t5 + xor $t2,$t6 + xor $t3,$t7 + + sub $cnt,1 + $PTR_ADD $key0,16 + xor $s0,$t0 + xor $s1,$t1 + xor $s2,$t2 + xor $s3,$t3 + .set noreorder + bnez $cnt,.Loop_enc + ext $i0,$s1,16,8 + + _xtr $i0,$s1,16-2 +#else + _xtr $i0,$s1,16-2 +.Loop_enc: + _xtr $i1,$s2,16-2 + _xtr $i2,$s3,16-2 + _xtr $i3,$s0,16-2 + and $i0,0x3fc + and $i1,0x3fc + and $i2,0x3fc + and $i3,0x3fc + $PTR_ADD $i0,$Tbl + $PTR_ADD $i1,$Tbl + $PTR_ADD $i2,$Tbl + $PTR_ADD $i3,$Tbl +#if defined(_MIPS_ARCH_MIPS32R2) || defined(_MIPS_ARCH_MIPS64R2) + lw $t0,0($i0) # Te1[s1>>16] + _xtr $i0,$s2,8-2 + lw $t1,0($i1) # Te1[s2>>16] + _xtr $i1,$s3,8-2 + lw $t2,0($i2) # Te1[s3>>16] + _xtr $i2,$s0,8-2 + lw $t3,0($i3) # Te1[s0>>16] + _xtr $i3,$s1,8-2 +#else + lwl $t0,3($i0) # Te1[s1>>16] + lwl $t1,3($i1) # Te1[s2>>16] + lwl $t2,3($i2) # Te1[s3>>16] + lwl $t3,3($i3) # Te1[s0>>16] + lwr $t0,2($i0) # Te1[s1>>16] + _xtr $i0,$s2,8-2 + lwr $t1,2($i1) # Te1[s2>>16] + _xtr $i1,$s3,8-2 + lwr $t2,2($i2) # Te1[s3>>16] + _xtr $i2,$s0,8-2 + lwr $t3,2($i3) # Te1[s0>>16] + _xtr $i3,$s1,8-2 +#endif + and $i0,0x3fc + and $i1,0x3fc + and $i2,0x3fc + and $i3,0x3fc + $PTR_ADD $i0,$Tbl + $PTR_ADD $i1,$Tbl + $PTR_ADD $i2,$Tbl + $PTR_ADD $i3,$Tbl +#if defined(_MIPS_ARCH_MIPS32R2) || defined(_MIPS_ARCH_MIPS64R2) + rotr $t0,$t0,8 + rotr $t1,$t1,8 + rotr $t2,$t2,8 + rotr $t3,$t3,8 +# if defined(_MIPSEL) + lw $t4,0($i0) # Te2[s2>>8] + _xtr $i0,$s3,0-2 + lw $t5,0($i1) # Te2[s3>>8] + _xtr $i1,$s0,0-2 + lw $t6,0($i2) # Te2[s0>>8] + _xtr $i2,$s1,0-2 + lw $t7,0($i3) # Te2[s1>>8] + _xtr $i3,$s2,0-2 + + and $i0,0x3fc + and $i1,0x3fc + and $i2,0x3fc + and $i3,0x3fc + $PTR_ADD $i0,$Tbl + $PTR_ADD $i1,$Tbl + $PTR_ADD $i2,$Tbl + $PTR_ADD $i3,$Tbl + lw $t8,0($i0) # Te3[s3] + $PTR_INS $i0,$s0,2,8 + lw $t9,0($i1) # Te3[s0] + $PTR_INS $i1,$s1,2,8 + lw $t10,0($i2) # Te3[s1] + $PTR_INS $i2,$s2,2,8 + lw $t11,0($i3) # Te3[s2] + $PTR_INS $i3,$s3,2,8 +# else + lw $t4,0($i0) # Te2[s2>>8] + $PTR_INS $i0,$s3,2,8 + lw $t5,0($i1) # Te2[s3>>8] + $PTR_INS $i1,$s0,2,8 + lw $t6,0($i2) # Te2[s0>>8] + $PTR_INS $i2,$s1,2,8 + lw $t7,0($i3) # Te2[s1>>8] + $PTR_INS $i3,$s2,2,8 + + lw $t8,0($i0) # Te3[s3] + _xtr $i0,$s0,24-2 + lw $t9,0($i1) # Te3[s0] + _xtr $i1,$s1,24-2 + lw $t10,0($i2) # Te3[s1] + _xtr $i2,$s2,24-2 + lw $t11,0($i3) # Te3[s2] + _xtr $i3,$s3,24-2 + + and $i0,0x3fc + and $i1,0x3fc + and $i2,0x3fc + and $i3,0x3fc + $PTR_ADD $i0,$Tbl + $PTR_ADD $i1,$Tbl + $PTR_ADD $i2,$Tbl + $PTR_ADD $i3,$Tbl +# endif + rotr $t4,$t4,16 + rotr $t5,$t5,16 + rotr $t6,$t6,16 + rotr $t7,$t7,16 + + rotr $t8,$t8,24 + rotr $t9,$t9,24 + rotr $t10,$t10,24 + rotr $t11,$t11,24 +#else + lwl $t4,2($i0) # Te2[s2>>8] + lwl $t5,2($i1) # Te2[s3>>8] + lwl $t6,2($i2) # Te2[s0>>8] + lwl $t7,2($i3) # Te2[s1>>8] + lwr $t4,1($i0) # Te2[s2>>8] + _xtr $i0,$s3,0-2 + lwr $t5,1($i1) # Te2[s3>>8] + _xtr $i1,$s0,0-2 + lwr $t6,1($i2) # Te2[s0>>8] + _xtr $i2,$s1,0-2 + lwr $t7,1($i3) # Te2[s1>>8] + _xtr $i3,$s2,0-2 + + and $i0,0x3fc + and $i1,0x3fc + and $i2,0x3fc + and $i3,0x3fc + $PTR_ADD $i0,$Tbl + $PTR_ADD $i1,$Tbl + $PTR_ADD $i2,$Tbl + $PTR_ADD $i3,$Tbl + lwl $t8,1($i0) # Te3[s3] + lwl $t9,1($i1) # Te3[s0] + lwl $t10,1($i2) # Te3[s1] + lwl $t11,1($i3) # Te3[s2] + lwr $t8,0($i0) # Te3[s3] + _xtr $i0,$s0,24-2 + lwr $t9,0($i1) # Te3[s0] + _xtr $i1,$s1,24-2 + lwr $t10,0($i2) # Te3[s1] + _xtr $i2,$s2,24-2 + lwr $t11,0($i3) # Te3[s2] + _xtr $i3,$s3,24-2 + + and $i0,0x3fc + and $i1,0x3fc + and $i2,0x3fc + and $i3,0x3fc + $PTR_ADD $i0,$Tbl + $PTR_ADD $i1,$Tbl + $PTR_ADD $i2,$Tbl + $PTR_ADD $i3,$Tbl +#endif + xor $t0,$t4 + lw $t4,0($i0) # Te0[s0>>24] + xor $t1,$t5 + lw $t5,0($i1) # Te0[s1>>24] + xor $t2,$t6 + lw $t6,0($i2) # Te0[s2>>24] + xor $t3,$t7 + lw $t7,0($i3) # Te0[s3>>24] + + xor $t0,$t8 + lw $s0,0($key0) + xor $t1,$t9 + lw $s1,4($key0) + xor $t2,$t10 + lw $s2,8($key0) + xor $t3,$t11 + lw $s3,12($key0) + + xor $t0,$t4 + xor $t1,$t5 + xor $t2,$t6 + xor $t3,$t7 + + sub $cnt,1 + $PTR_ADD $key0,16 + xor $s0,$t0 + xor $s1,$t1 + xor $s2,$t2 + xor $s3,$t3 + .set noreorder + bnez $cnt,.Loop_enc + _xtr $i0,$s1,16-2 +#endif + + .set reorder + _xtr $i1,$s2,16-2 + _xtr $i2,$s3,16-2 + _xtr $i3,$s0,16-2 + and $i0,0x3fc + and $i1,0x3fc + and $i2,0x3fc + and $i3,0x3fc + $PTR_ADD $i0,$Tbl + $PTR_ADD $i1,$Tbl + $PTR_ADD $i2,$Tbl + $PTR_ADD $i3,$Tbl + lbu $t0,2($i0) # Te4[s1>>16] + _xtr $i0,$s2,8-2 + lbu $t1,2($i1) # Te4[s2>>16] + _xtr $i1,$s3,8-2 + lbu $t2,2($i2) # Te4[s3>>16] + _xtr $i2,$s0,8-2 + lbu $t3,2($i3) # Te4[s0>>16] + _xtr $i3,$s1,8-2 + + and $i0,0x3fc + and $i1,0x3fc + and $i2,0x3fc + and $i3,0x3fc + $PTR_ADD $i0,$Tbl + $PTR_ADD $i1,$Tbl + $PTR_ADD $i2,$Tbl + $PTR_ADD $i3,$Tbl +#if defined(_MIPS_ARCH_MIPS32R2) || defined(_MIPS_ARCH_MIPS64R2) +# if defined(_MIPSEL) + lbu $t4,2($i0) # Te4[s2>>8] + $PTR_INS $i0,$s0,2,8 + lbu $t5,2($i1) # Te4[s3>>8] + $PTR_INS $i1,$s1,2,8 + lbu $t6,2($i2) # Te4[s0>>8] + $PTR_INS $i2,$s2,2,8 + lbu $t7,2($i3) # Te4[s1>>8] + $PTR_INS $i3,$s3,2,8 + + lbu $t8,2($i0) # Te4[s0>>24] + _xtr $i0,$s3,0-2 + lbu $t9,2($i1) # Te4[s1>>24] + _xtr $i1,$s0,0-2 + lbu $t10,2($i2) # Te4[s2>>24] + _xtr $i2,$s1,0-2 + lbu $t11,2($i3) # Te4[s3>>24] + _xtr $i3,$s2,0-2 + + and $i0,0x3fc + and $i1,0x3fc + and $i2,0x3fc + and $i3,0x3fc + $PTR_ADD $i0,$Tbl + $PTR_ADD $i1,$Tbl + $PTR_ADD $i2,$Tbl + $PTR_ADD $i3,$Tbl +# else + lbu $t4,2($i0) # Te4[s2>>8] + _xtr $i0,$s0,24-2 + lbu $t5,2($i1) # Te4[s3>>8] + _xtr $i1,$s1,24-2 + lbu $t6,2($i2) # Te4[s0>>8] + _xtr $i2,$s2,24-2 + lbu $t7,2($i3) # Te4[s1>>8] + _xtr $i3,$s3,24-2 + + and $i0,0x3fc + and $i1,0x3fc + and $i2,0x3fc + and $i3,0x3fc + $PTR_ADD $i0,$Tbl + $PTR_ADD $i1,$Tbl + $PTR_ADD $i2,$Tbl + $PTR_ADD $i3,$Tbl + lbu $t8,2($i0) # Te4[s0>>24] + $PTR_INS $i0,$s3,2,8 + lbu $t9,2($i1) # Te4[s1>>24] + $PTR_INS $i1,$s0,2,8 + lbu $t10,2($i2) # Te4[s2>>24] + $PTR_INS $i2,$s1,2,8 + lbu $t11,2($i3) # Te4[s3>>24] + $PTR_INS $i3,$s2,2,8 +# endif + _ins $t0,16 + _ins $t1,16 + _ins $t2,16 + _ins $t3,16 + + _ins2 $t0,$t4,8 + lbu $t4,2($i0) # Te4[s3] + _ins2 $t1,$t5,8 + lbu $t5,2($i1) # Te4[s0] + _ins2 $t2,$t6,8 + lbu $t6,2($i2) # Te4[s1] + _ins2 $t3,$t7,8 + lbu $t7,2($i3) # Te4[s2] + + _ins2 $t0,$t8,24 + lw $s0,0($key0) + _ins2 $t1,$t9,24 + lw $s1,4($key0) + _ins2 $t2,$t10,24 + lw $s2,8($key0) + _ins2 $t3,$t11,24 + lw $s3,12($key0) + + _ins2 $t0,$t4,0 + _ins2 $t1,$t5,0 + _ins2 $t2,$t6,0 + _ins2 $t3,$t7,0 +#else + lbu $t4,2($i0) # Te4[s2>>8] + _xtr $i0,$s0,24-2 + lbu $t5,2($i1) # Te4[s3>>8] + _xtr $i1,$s1,24-2 + lbu $t6,2($i2) # Te4[s0>>8] + _xtr $i2,$s2,24-2 + lbu $t7,2($i3) # Te4[s1>>8] + _xtr $i3,$s3,24-2 + + and $i0,0x3fc + and $i1,0x3fc + and $i2,0x3fc + and $i3,0x3fc + $PTR_ADD $i0,$Tbl + $PTR_ADD $i1,$Tbl + $PTR_ADD $i2,$Tbl + $PTR_ADD $i3,$Tbl + lbu $t8,2($i0) # Te4[s0>>24] + _xtr $i0,$s3,0-2 + lbu $t9,2($i1) # Te4[s1>>24] + _xtr $i1,$s0,0-2 + lbu $t10,2($i2) # Te4[s2>>24] + _xtr $i2,$s1,0-2 + lbu $t11,2($i3) # Te4[s3>>24] + _xtr $i3,$s2,0-2 + + and $i0,0x3fc + and $i1,0x3fc + and $i2,0x3fc + and $i3,0x3fc + $PTR_ADD $i0,$Tbl + $PTR_ADD $i1,$Tbl + $PTR_ADD $i2,$Tbl + $PTR_ADD $i3,$Tbl + + _ins $t0,16 + _ins $t1,16 + _ins $t2,16 + _ins $t3,16 + + _ins $t4,8 + _ins $t5,8 + _ins $t6,8 + _ins $t7,8 + + xor $t0,$t4 + lbu $t4,2($i0) # Te4[s3] + xor $t1,$t5 + lbu $t5,2($i1) # Te4[s0] + xor $t2,$t6 + lbu $t6,2($i2) # Te4[s1] + xor $t3,$t7 + lbu $t7,2($i3) # Te4[s2] + + _ins $t8,24 + lw $s0,0($key0) + _ins $t9,24 + lw $s1,4($key0) + _ins $t10,24 + lw $s2,8($key0) + _ins $t11,24 + lw $s3,12($key0) + + xor $t0,$t8 + xor $t1,$t9 + xor $t2,$t10 + xor $t3,$t11 + + _ins $t4,0 + _ins $t5,0 + _ins $t6,0 + _ins $t7,0 + + xor $t0,$t4 + xor $t1,$t5 + xor $t2,$t6 + xor $t3,$t7 +#endif + xor $s0,$t0 + xor $s1,$t1 + xor $s2,$t2 + xor $s3,$t3 + + jr $ra +.end _mips_AES_encrypt + +.align 5 +.globl AES_encrypt +.ent AES_encrypt +AES_encrypt: + .frame $sp,$FRAMESIZE,$ra + .mask $SAVED_REGS_MASK,-$SZREG + .set noreorder +___ +$code.=<<___ if ($flavour =~ /o32/i); # o32 PIC-ification + .cpload $pf +___ +$code.=<<___; + $PTR_SUB $sp,$FRAMESIZE + $REG_S $ra,$FRAMESIZE-1*$SZREG($sp) + $REG_S $fp,$FRAMESIZE-2*$SZREG($sp) + $REG_S $s11,$FRAMESIZE-3*$SZREG($sp) + $REG_S $s10,$FRAMESIZE-4*$SZREG($sp) + $REG_S $s9,$FRAMESIZE-5*$SZREG($sp) + $REG_S $s8,$FRAMESIZE-6*$SZREG($sp) + $REG_S $s7,$FRAMESIZE-7*$SZREG($sp) + $REG_S $s6,$FRAMESIZE-8*$SZREG($sp) + $REG_S $s5,$FRAMESIZE-9*$SZREG($sp) + $REG_S $s4,$FRAMESIZE-10*$SZREG($sp) +___ +$code.=<<___ if ($flavour =~ /nubi/i); # optimize non-nubi prologue + $REG_S \$15,$FRAMESIZE-11*$SZREG($sp) + $REG_S \$14,$FRAMESIZE-12*$SZREG($sp) + $REG_S \$13,$FRAMESIZE-13*$SZREG($sp) + $REG_S \$12,$FRAMESIZE-14*$SZREG($sp) + $REG_S $gp,$FRAMESIZE-15*$SZREG($sp) +___ +$code.=<<___ if ($flavour !~ /o32/i); # non-o32 PIC-ification + .cplocal $Tbl + .cpsetup $pf,$zero,AES_encrypt +___ +$code.=<<___; + .set reorder + $PTR_LA $Tbl,AES_Te # PIC-ified 'load address' + + lwl $s0,0+$MSB($inp) + lwl $s1,4+$MSB($inp) + lwl $s2,8+$MSB($inp) + lwl $s3,12+$MSB($inp) + lwr $s0,0+$LSB($inp) + lwr $s1,4+$LSB($inp) + lwr $s2,8+$LSB($inp) + lwr $s3,12+$LSB($inp) + + bal _mips_AES_encrypt + + swr $s0,0+$LSB($out) + swr $s1,4+$LSB($out) + swr $s2,8+$LSB($out) + swr $s3,12+$LSB($out) + swl $s0,0+$MSB($out) + swl $s1,4+$MSB($out) + swl $s2,8+$MSB($out) + swl $s3,12+$MSB($out) + + .set noreorder + $REG_L $ra,$FRAMESIZE-1*$SZREG($sp) + $REG_L $fp,$FRAMESIZE-2*$SZREG($sp) + $REG_L $s11,$FRAMESIZE-3*$SZREG($sp) + $REG_L $s10,$FRAMESIZE-4*$SZREG($sp) + $REG_L $s9,$FRAMESIZE-5*$SZREG($sp) + $REG_L $s8,$FRAMESIZE-6*$SZREG($sp) + $REG_L $s7,$FRAMESIZE-7*$SZREG($sp) + $REG_L $s6,$FRAMESIZE-8*$SZREG($sp) + $REG_L $s5,$FRAMESIZE-9*$SZREG($sp) + $REG_L $s4,$FRAMESIZE-10*$SZREG($sp) +___ +$code.=<<___ if ($flavour =~ /nubi/i); + $REG_L \$15,$FRAMESIZE-11*$SZREG($sp) + $REG_L \$14,$FRAMESIZE-12*$SZREG($sp) + $REG_L \$13,$FRAMESIZE-13*$SZREG($sp) + $REG_L \$12,$FRAMESIZE-14*$SZREG($sp) + $REG_L $gp,$FRAMESIZE-15*$SZREG($sp) +___ +$code.=<<___; + jr $ra + $PTR_ADD $sp,$FRAMESIZE +.end AES_encrypt +___ + +$code.=<<___; +.align 5 +.ent _mips_AES_decrypt +_mips_AES_decrypt: + .frame $sp,0,$ra + .set reorder + lw $t0,0($key) + lw $t1,4($key) + lw $t2,8($key) + lw $t3,12($key) + lw $cnt,240($key) + $PTR_ADD $key0,$key,16 + + xor $s0,$t0 + xor $s1,$t1 + xor $s2,$t2 + xor $s3,$t3 + + sub $cnt,1 +#if defined(__mips_smartmips) + ext $i0,$s3,16,8 +.Loop_dec: + ext $i1,$s0,16,8 + ext $i2,$s1,16,8 + ext $i3,$s2,16,8 + lwxs $t0,$i0($Tbl) # Td1[s3>>16] + ext $i0,$s2,8,8 + lwxs $t1,$i1($Tbl) # Td1[s0>>16] + ext $i1,$s3,8,8 + lwxs $t2,$i2($Tbl) # Td1[s1>>16] + ext $i2,$s0,8,8 + lwxs $t3,$i3($Tbl) # Td1[s2>>16] + ext $i3,$s1,8,8 + + lwxs $t4,$i0($Tbl) # Td2[s2>>8] + ext $i0,$s1,0,8 + lwxs $t5,$i1($Tbl) # Td2[s3>>8] + ext $i1,$s2,0,8 + lwxs $t6,$i2($Tbl) # Td2[s0>>8] + ext $i2,$s3,0,8 + lwxs $t7,$i3($Tbl) # Td2[s1>>8] + ext $i3,$s0,0,8 + + lwxs $t8,$i0($Tbl) # Td3[s1] + ext $i0,$s0,24,8 + lwxs $t9,$i1($Tbl) # Td3[s2] + ext $i1,$s1,24,8 + lwxs $t10,$i2($Tbl) # Td3[s3] + ext $i2,$s2,24,8 + lwxs $t11,$i3($Tbl) # Td3[s0] + ext $i3,$s3,24,8 + + rotr $t0,$t0,8 + rotr $t1,$t1,8 + rotr $t2,$t2,8 + rotr $t3,$t3,8 + + rotr $t4,$t4,16 + rotr $t5,$t5,16 + rotr $t6,$t6,16 + rotr $t7,$t7,16 + + xor $t0,$t4 + lwxs $t4,$i0($Tbl) # Td0[s0>>24] + xor $t1,$t5 + lwxs $t5,$i1($Tbl) # Td0[s1>>24] + xor $t2,$t6 + lwxs $t6,$i2($Tbl) # Td0[s2>>24] + xor $t3,$t7 + lwxs $t7,$i3($Tbl) # Td0[s3>>24] + + rotr $t8,$t8,24 + lw $s0,0($key0) + rotr $t9,$t9,24 + lw $s1,4($key0) + rotr $t10,$t10,24 + lw $s2,8($key0) + rotr $t11,$t11,24 + lw $s3,12($key0) + + xor $t0,$t8 + xor $t1,$t9 + xor $t2,$t10 + xor $t3,$t11 + + xor $t0,$t4 + xor $t1,$t5 + xor $t2,$t6 + xor $t3,$t7 + + sub $cnt,1 + $PTR_ADD $key0,16 + xor $s0,$t0 + xor $s1,$t1 + xor $s2,$t2 + xor $s3,$t3 + .set noreorder + bnez $cnt,.Loop_dec + ext $i0,$s3,16,8 + + _xtr $i0,$s3,16-2 +#else + _xtr $i0,$s3,16-2 +.Loop_dec: + _xtr $i1,$s0,16-2 + _xtr $i2,$s1,16-2 + _xtr $i3,$s2,16-2 + and $i0,0x3fc + and $i1,0x3fc + and $i2,0x3fc + and $i3,0x3fc + $PTR_ADD $i0,$Tbl + $PTR_ADD $i1,$Tbl + $PTR_ADD $i2,$Tbl + $PTR_ADD $i3,$Tbl +#if defined(_MIPS_ARCH_MIPS32R2) || defined(_MIPS_ARCH_MIPS64R2) + lw $t0,0($i0) # Td1[s3>>16] + _xtr $i0,$s2,8-2 + lw $t1,0($i1) # Td1[s0>>16] + _xtr $i1,$s3,8-2 + lw $t2,0($i2) # Td1[s1>>16] + _xtr $i2,$s0,8-2 + lw $t3,0($i3) # Td1[s2>>16] + _xtr $i3,$s1,8-2 +#else + lwl $t0,3($i0) # Td1[s3>>16] + lwl $t1,3($i1) # Td1[s0>>16] + lwl $t2,3($i2) # Td1[s1>>16] + lwl $t3,3($i3) # Td1[s2>>16] + lwr $t0,2($i0) # Td1[s3>>16] + _xtr $i0,$s2,8-2 + lwr $t1,2($i1) # Td1[s0>>16] + _xtr $i1,$s3,8-2 + lwr $t2,2($i2) # Td1[s1>>16] + _xtr $i2,$s0,8-2 + lwr $t3,2($i3) # Td1[s2>>16] + _xtr $i3,$s1,8-2 +#endif + + and $i0,0x3fc + and $i1,0x3fc + and $i2,0x3fc + and $i3,0x3fc + $PTR_ADD $i0,$Tbl + $PTR_ADD $i1,$Tbl + $PTR_ADD $i2,$Tbl + $PTR_ADD $i3,$Tbl +#if defined(_MIPS_ARCH_MIPS32R2) || defined(_MIPS_ARCH_MIPS64R2) + rotr $t0,$t0,8 + rotr $t1,$t1,8 + rotr $t2,$t2,8 + rotr $t3,$t3,8 +# if defined(_MIPSEL) + lw $t4,0($i0) # Td2[s2>>8] + _xtr $i0,$s1,0-2 + lw $t5,0($i1) # Td2[s3>>8] + _xtr $i1,$s2,0-2 + lw $t6,0($i2) # Td2[s0>>8] + _xtr $i2,$s3,0-2 + lw $t7,0($i3) # Td2[s1>>8] + _xtr $i3,$s0,0-2 + + and $i0,0x3fc + and $i1,0x3fc + and $i2,0x3fc + and $i3,0x3fc + $PTR_ADD $i0,$Tbl + $PTR_ADD $i1,$Tbl + $PTR_ADD $i2,$Tbl + $PTR_ADD $i3,$Tbl + lw $t8,0($i0) # Td3[s1] + $PTR_INS $i0,$s0,2,8 + lw $t9,0($i1) # Td3[s2] + $PTR_INS $i1,$s1,2,8 + lw $t10,0($i2) # Td3[s3] + $PTR_INS $i2,$s2,2,8 + lw $t11,0($i3) # Td3[s0] + $PTR_INS $i3,$s3,2,8 +#else + lw $t4,0($i0) # Td2[s2>>8] + $PTR_INS $i0,$s1,2,8 + lw $t5,0($i1) # Td2[s3>>8] + $PTR_INS $i1,$s2,2,8 + lw $t6,0($i2) # Td2[s0>>8] + $PTR_INS $i2,$s3,2,8 + lw $t7,0($i3) # Td2[s1>>8] + $PTR_INS $i3,$s0,2,8 + + lw $t8,0($i0) # Td3[s1] + _xtr $i0,$s0,24-2 + lw $t9,0($i1) # Td3[s2] + _xtr $i1,$s1,24-2 + lw $t10,0($i2) # Td3[s3] + _xtr $i2,$s2,24-2 + lw $t11,0($i3) # Td3[s0] + _xtr $i3,$s3,24-2 + + and $i0,0x3fc + and $i1,0x3fc + and $i2,0x3fc + and $i3,0x3fc + $PTR_ADD $i0,$Tbl + $PTR_ADD $i1,$Tbl + $PTR_ADD $i2,$Tbl + $PTR_ADD $i3,$Tbl +#endif + rotr $t4,$t4,16 + rotr $t5,$t5,16 + rotr $t6,$t6,16 + rotr $t7,$t7,16 + + rotr $t8,$t8,24 + rotr $t9,$t9,24 + rotr $t10,$t10,24 + rotr $t11,$t11,24 +#else + lwl $t4,2($i0) # Td2[s2>>8] + lwl $t5,2($i1) # Td2[s3>>8] + lwl $t6,2($i2) # Td2[s0>>8] + lwl $t7,2($i3) # Td2[s1>>8] + lwr $t4,1($i0) # Td2[s2>>8] + _xtr $i0,$s1,0-2 + lwr $t5,1($i1) # Td2[s3>>8] + _xtr $i1,$s2,0-2 + lwr $t6,1($i2) # Td2[s0>>8] + _xtr $i2,$s3,0-2 + lwr $t7,1($i3) # Td2[s1>>8] + _xtr $i3,$s0,0-2 + + and $i0,0x3fc + and $i1,0x3fc + and $i2,0x3fc + and $i3,0x3fc + $PTR_ADD $i0,$Tbl + $PTR_ADD $i1,$Tbl + $PTR_ADD $i2,$Tbl + $PTR_ADD $i3,$Tbl + lwl $t8,1($i0) # Td3[s1] + lwl $t9,1($i1) # Td3[s2] + lwl $t10,1($i2) # Td3[s3] + lwl $t11,1($i3) # Td3[s0] + lwr $t8,0($i0) # Td3[s1] + _xtr $i0,$s0,24-2 + lwr $t9,0($i1) # Td3[s2] + _xtr $i1,$s1,24-2 + lwr $t10,0($i2) # Td3[s3] + _xtr $i2,$s2,24-2 + lwr $t11,0($i3) # Td3[s0] + _xtr $i3,$s3,24-2 + + and $i0,0x3fc + and $i1,0x3fc + and $i2,0x3fc + and $i3,0x3fc + $PTR_ADD $i0,$Tbl + $PTR_ADD $i1,$Tbl + $PTR_ADD $i2,$Tbl + $PTR_ADD $i3,$Tbl +#endif + + xor $t0,$t4 + lw $t4,0($i0) # Td0[s0>>24] + xor $t1,$t5 + lw $t5,0($i1) # Td0[s1>>24] + xor $t2,$t6 + lw $t6,0($i2) # Td0[s2>>24] + xor $t3,$t7 + lw $t7,0($i3) # Td0[s3>>24] + + xor $t0,$t8 + lw $s0,0($key0) + xor $t1,$t9 + lw $s1,4($key0) + xor $t2,$t10 + lw $s2,8($key0) + xor $t3,$t11 + lw $s3,12($key0) + + xor $t0,$t4 + xor $t1,$t5 + xor $t2,$t6 + xor $t3,$t7 + + sub $cnt,1 + $PTR_ADD $key0,16 + xor $s0,$t0 + xor $s1,$t1 + xor $s2,$t2 + xor $s3,$t3 + .set noreorder + bnez $cnt,.Loop_dec + _xtr $i0,$s3,16-2 +#endif + + .set reorder + lw $t4,1024($Tbl) # prefetch Td4 + _xtr $i0,$s3,16 + lw $t5,1024+32($Tbl) + _xtr $i1,$s0,16 + lw $t6,1024+64($Tbl) + _xtr $i2,$s1,16 + lw $t7,1024+96($Tbl) + _xtr $i3,$s2,16 + lw $t8,1024+128($Tbl) + and $i0,0xff + lw $t9,1024+160($Tbl) + and $i1,0xff + lw $t10,1024+192($Tbl) + and $i2,0xff + lw $t11,1024+224($Tbl) + and $i3,0xff + + $PTR_ADD $i0,$Tbl + $PTR_ADD $i1,$Tbl + $PTR_ADD $i2,$Tbl + $PTR_ADD $i3,$Tbl + lbu $t0,1024($i0) # Td4[s3>>16] + _xtr $i0,$s2,8 + lbu $t1,1024($i1) # Td4[s0>>16] + _xtr $i1,$s3,8 + lbu $t2,1024($i2) # Td4[s1>>16] + _xtr $i2,$s0,8 + lbu $t3,1024($i3) # Td4[s2>>16] + _xtr $i3,$s1,8 + + and $i0,0xff + and $i1,0xff + and $i2,0xff + and $i3,0xff + $PTR_ADD $i0,$Tbl + $PTR_ADD $i1,$Tbl + $PTR_ADD $i2,$Tbl + $PTR_ADD $i3,$Tbl +#if defined(_MIPS_ARCH_MIPS32R2) || defined(_MIPS_ARCH_MIPS64R2) +# if defined(_MIPSEL) + lbu $t4,1024($i0) # Td4[s2>>8] + $PTR_INS $i0,$s0,0,8 + lbu $t5,1024($i1) # Td4[s3>>8] + $PTR_INS $i1,$s1,0,8 + lbu $t6,1024($i2) # Td4[s0>>8] + $PTR_INS $i2,$s2,0,8 + lbu $t7,1024($i3) # Td4[s1>>8] + $PTR_INS $i3,$s3,0,8 + + lbu $t8,1024($i0) # Td4[s0>>24] + _xtr $i0,$s1,0 + lbu $t9,1024($i1) # Td4[s1>>24] + _xtr $i1,$s2,0 + lbu $t10,1024($i2) # Td4[s2>>24] + _xtr $i2,$s3,0 + lbu $t11,1024($i3) # Td4[s3>>24] + _xtr $i3,$s0,0 + + $PTR_ADD $i0,$Tbl + $PTR_ADD $i1,$Tbl + $PTR_ADD $i2,$Tbl + $PTR_ADD $i3,$Tbl +# else + lbu $t4,1024($i0) # Td4[s2>>8] + _xtr $i0,$s0,24 + lbu $t5,1024($i1) # Td4[s3>>8] + _xtr $i1,$s1,24 + lbu $t6,1024($i2) # Td4[s0>>8] + _xtr $i2,$s2,24 + lbu $t7,1024($i3) # Td4[s1>>8] + _xtr $i3,$s3,24 + + $PTR_ADD $i0,$Tbl + $PTR_ADD $i1,$Tbl + $PTR_ADD $i2,$Tbl + $PTR_ADD $i3,$Tbl + lbu $t8,1024($i0) # Td4[s0>>24] + $PTR_INS $i0,$s1,0,8 + lbu $t9,1024($i1) # Td4[s1>>24] + $PTR_INS $i1,$s2,0,8 + lbu $t10,1024($i2) # Td4[s2>>24] + $PTR_INS $i2,$s3,0,8 + lbu $t11,1024($i3) # Td4[s3>>24] + $PTR_INS $i3,$s0,0,8 +# endif + _ins $t0,16 + _ins $t1,16 + _ins $t2,16 + _ins $t3,16 + + _ins2 $t0,$t4,8 + lbu $t4,1024($i0) # Td4[s1] + _ins2 $t1,$t5,8 + lbu $t5,1024($i1) # Td4[s2] + _ins2 $t2,$t6,8 + lbu $t6,1024($i2) # Td4[s3] + _ins2 $t3,$t7,8 + lbu $t7,1024($i3) # Td4[s0] + + _ins2 $t0,$t8,24 + lw $s0,0($key0) + _ins2 $t1,$t9,24 + lw $s1,4($key0) + _ins2 $t2,$t10,24 + lw $s2,8($key0) + _ins2 $t3,$t11,24 + lw $s3,12($key0) + + _ins2 $t0,$t4,0 + _ins2 $t1,$t5,0 + _ins2 $t2,$t6,0 + _ins2 $t3,$t7,0 +#else + lbu $t4,1024($i0) # Td4[s2>>8] + _xtr $i0,$s0,24 + lbu $t5,1024($i1) # Td4[s3>>8] + _xtr $i1,$s1,24 + lbu $t6,1024($i2) # Td4[s0>>8] + _xtr $i2,$s2,24 + lbu $t7,1024($i3) # Td4[s1>>8] + _xtr $i3,$s3,24 + + $PTR_ADD $i0,$Tbl + $PTR_ADD $i1,$Tbl + $PTR_ADD $i2,$Tbl + $PTR_ADD $i3,$Tbl + lbu $t8,1024($i0) # Td4[s0>>24] + _xtr $i0,$s1,0 + lbu $t9,1024($i1) # Td4[s1>>24] + _xtr $i1,$s2,0 + lbu $t10,1024($i2) # Td4[s2>>24] + _xtr $i2,$s3,0 + lbu $t11,1024($i3) # Td4[s3>>24] + _xtr $i3,$s0,0 + + $PTR_ADD $i0,$Tbl + $PTR_ADD $i1,$Tbl + $PTR_ADD $i2,$Tbl + $PTR_ADD $i3,$Tbl + + _ins $t0,16 + _ins $t1,16 + _ins $t2,16 + _ins $t3,16 + + _ins $t4,8 + _ins $t5,8 + _ins $t6,8 + _ins $t7,8 + + xor $t0,$t4 + lbu $t4,1024($i0) # Td4[s1] + xor $t1,$t5 + lbu $t5,1024($i1) # Td4[s2] + xor $t2,$t6 + lbu $t6,1024($i2) # Td4[s3] + xor $t3,$t7 + lbu $t7,1024($i3) # Td4[s0] + + _ins $t8,24 + lw $s0,0($key0) + _ins $t9,24 + lw $s1,4($key0) + _ins $t10,24 + lw $s2,8($key0) + _ins $t11,24 + lw $s3,12($key0) + + xor $t0,$t8 + xor $t1,$t9 + xor $t2,$t10 + xor $t3,$t11 + + _ins $t4,0 + _ins $t5,0 + _ins $t6,0 + _ins $t7,0 + + xor $t0,$t4 + xor $t1,$t5 + xor $t2,$t6 + xor $t3,$t7 +#endif + + xor $s0,$t0 + xor $s1,$t1 + xor $s2,$t2 + xor $s3,$t3 + + jr $ra +.end _mips_AES_decrypt + +.align 5 +.globl AES_decrypt +.ent AES_decrypt +AES_decrypt: + .frame $sp,$FRAMESIZE,$ra + .mask $SAVED_REGS_MASK,-$SZREG + .set noreorder +___ +$code.=<<___ if ($flavour =~ /o32/i); # o32 PIC-ification + .cpload $pf +___ +$code.=<<___; + $PTR_SUB $sp,$FRAMESIZE + $REG_S $ra,$FRAMESIZE-1*$SZREG($sp) + $REG_S $fp,$FRAMESIZE-2*$SZREG($sp) + $REG_S $s11,$FRAMESIZE-3*$SZREG($sp) + $REG_S $s10,$FRAMESIZE-4*$SZREG($sp) + $REG_S $s9,$FRAMESIZE-5*$SZREG($sp) + $REG_S $s8,$FRAMESIZE-6*$SZREG($sp) + $REG_S $s7,$FRAMESIZE-7*$SZREG($sp) + $REG_S $s6,$FRAMESIZE-8*$SZREG($sp) + $REG_S $s5,$FRAMESIZE-9*$SZREG($sp) + $REG_S $s4,$FRAMESIZE-10*$SZREG($sp) +___ +$code.=<<___ if ($flavour =~ /nubi/i); # optimize non-nubi prologue + $REG_S \$15,$FRAMESIZE-11*$SZREG($sp) + $REG_S \$14,$FRAMESIZE-12*$SZREG($sp) + $REG_S \$13,$FRAMESIZE-13*$SZREG($sp) + $REG_S \$12,$FRAMESIZE-14*$SZREG($sp) + $REG_S $gp,$FRAMESIZE-15*$SZREG($sp) +___ +$code.=<<___ if ($flavour !~ /o32/i); # non-o32 PIC-ification + .cplocal $Tbl + .cpsetup $pf,$zero,AES_decrypt +___ +$code.=<<___; + .set reorder + $PTR_LA $Tbl,AES_Td # PIC-ified 'load address' + + lwl $s0,0+$MSB($inp) + lwl $s1,4+$MSB($inp) + lwl $s2,8+$MSB($inp) + lwl $s3,12+$MSB($inp) + lwr $s0,0+$LSB($inp) + lwr $s1,4+$LSB($inp) + lwr $s2,8+$LSB($inp) + lwr $s3,12+$LSB($inp) + + bal _mips_AES_decrypt + + swr $s0,0+$LSB($out) + swr $s1,4+$LSB($out) + swr $s2,8+$LSB($out) + swr $s3,12+$LSB($out) + swl $s0,0+$MSB($out) + swl $s1,4+$MSB($out) + swl $s2,8+$MSB($out) + swl $s3,12+$MSB($out) + + .set noreorder + $REG_L $ra,$FRAMESIZE-1*$SZREG($sp) + $REG_L $fp,$FRAMESIZE-2*$SZREG($sp) + $REG_L $s11,$FRAMESIZE-3*$SZREG($sp) + $REG_L $s10,$FRAMESIZE-4*$SZREG($sp) + $REG_L $s9,$FRAMESIZE-5*$SZREG($sp) + $REG_L $s8,$FRAMESIZE-6*$SZREG($sp) + $REG_L $s7,$FRAMESIZE-7*$SZREG($sp) + $REG_L $s6,$FRAMESIZE-8*$SZREG($sp) + $REG_L $s5,$FRAMESIZE-9*$SZREG($sp) + $REG_L $s4,$FRAMESIZE-10*$SZREG($sp) +___ +$code.=<<___ if ($flavour =~ /nubi/i); + $REG_L \$15,$FRAMESIZE-11*$SZREG($sp) + $REG_L \$14,$FRAMESIZE-12*$SZREG($sp) + $REG_L \$13,$FRAMESIZE-13*$SZREG($sp) + $REG_L \$12,$FRAMESIZE-14*$SZREG($sp) + $REG_L $gp,$FRAMESIZE-15*$SZREG($sp) +___ +$code.=<<___; + jr $ra + $PTR_ADD $sp,$FRAMESIZE +.end AES_decrypt +___ +}}} + +{{{ +my $FRAMESIZE=8*$SZREG; +my $SAVED_REGS_MASK = ($flavour =~ /nubi/i) ? "0xc000f008" : "0xc0000000"; + +my ($inp,$bits,$key,$Tbl)=($a0,$a1,$a2,$a3); +my ($rk0,$rk1,$rk2,$rk3,$rk4,$rk5,$rk6,$rk7)=($a4,$a5,$a6,$a7,$s0,$s1,$s2,$s3); +my ($i0,$i1,$i2,$i3)=($at,$t0,$t1,$t2); +my ($rcon,$cnt)=($gp,$fp); + +$code.=<<___; +.align 5 +.ent _mips_AES_set_encrypt_key +_mips_AES_set_encrypt_key: + .frame $sp,0,$ra + .set noreorder + beqz $inp,.Lekey_done + li $t0,-1 + beqz $key,.Lekey_done + $PTR_ADD $rcon,$Tbl,256 + + .set reorder + lwl $rk0,0+$MSB($inp) # load 128 bits + lwl $rk1,4+$MSB($inp) + lwl $rk2,8+$MSB($inp) + lwl $rk3,12+$MSB($inp) + li $at,128 + lwr $rk0,0+$LSB($inp) + lwr $rk1,4+$LSB($inp) + lwr $rk2,8+$LSB($inp) + lwr $rk3,12+$LSB($inp) + .set noreorder + beq $bits,$at,.L128bits + li $cnt,10 + + .set reorder + lwl $rk4,16+$MSB($inp) # load 192 bits + lwl $rk5,20+$MSB($inp) + li $at,192 + lwr $rk4,16+$LSB($inp) + lwr $rk5,20+$LSB($inp) + .set noreorder + beq $bits,$at,.L192bits + li $cnt,8 + + .set reorder + lwl $rk6,24+$MSB($inp) # load 256 bits + lwl $rk7,28+$MSB($inp) + li $at,256 + lwr $rk6,24+$LSB($inp) + lwr $rk7,28+$LSB($inp) + .set noreorder + beq $bits,$at,.L256bits + li $cnt,7 + + b .Lekey_done + li $t0,-2 + +.align 4 +.L128bits: + .set reorder + srl $i0,$rk3,16 + srl $i1,$rk3,8 + and $i0,0xff + and $i1,0xff + and $i2,$rk3,0xff + srl $i3,$rk3,24 + $PTR_ADD $i0,$Tbl + $PTR_ADD $i1,$Tbl + $PTR_ADD $i2,$Tbl + $PTR_ADD $i3,$Tbl + lbu $i0,0($i0) + lbu $i1,0($i1) + lbu $i2,0($i2) + lbu $i3,0($i3) + + sw $rk0,0($key) + sw $rk1,4($key) + sw $rk2,8($key) + sw $rk3,12($key) + sub $cnt,1 + $PTR_ADD $key,16 + + _bias $i0,24 + _bias $i1,16 + _bias $i2,8 + _bias $i3,0 + + xor $rk0,$i0 + lw $i0,0($rcon) + xor $rk0,$i1 + xor $rk0,$i2 + xor $rk0,$i3 + xor $rk0,$i0 + + xor $rk1,$rk0 + xor $rk2,$rk1 + xor $rk3,$rk2 + + .set noreorder + bnez $cnt,.L128bits + $PTR_ADD $rcon,4 + + sw $rk0,0($key) + sw $rk1,4($key) + sw $rk2,8($key) + li $cnt,10 + sw $rk3,12($key) + li $t0,0 + sw $cnt,80($key) + b .Lekey_done + $PTR_SUB $key,10*16 + +.align 4 +.L192bits: + .set reorder + srl $i0,$rk5,16 + srl $i1,$rk5,8 + and $i0,0xff + and $i1,0xff + and $i2,$rk5,0xff + srl $i3,$rk5,24 + $PTR_ADD $i0,$Tbl + $PTR_ADD $i1,$Tbl + $PTR_ADD $i2,$Tbl + $PTR_ADD $i3,$Tbl + lbu $i0,0($i0) + lbu $i1,0($i1) + lbu $i2,0($i2) + lbu $i3,0($i3) + + sw $rk0,0($key) + sw $rk1,4($key) + sw $rk2,8($key) + sw $rk3,12($key) + sw $rk4,16($key) + sw $rk5,20($key) + sub $cnt,1 + $PTR_ADD $key,24 + + _bias $i0,24 + _bias $i1,16 + _bias $i2,8 + _bias $i3,0 + + xor $rk0,$i0 + lw $i0,0($rcon) + xor $rk0,$i1 + xor $rk0,$i2 + xor $rk0,$i3 + xor $rk0,$i0 + + xor $rk1,$rk0 + xor $rk2,$rk1 + xor $rk3,$rk2 + xor $rk4,$rk3 + xor $rk5,$rk4 + + .set noreorder + bnez $cnt,.L192bits + $PTR_ADD $rcon,4 + + sw $rk0,0($key) + sw $rk1,4($key) + sw $rk2,8($key) + li $cnt,12 + sw $rk3,12($key) + li $t0,0 + sw $cnt,48($key) + b .Lekey_done + $PTR_SUB $key,12*16 + +.align 4 +.L256bits: + .set reorder + srl $i0,$rk7,16 + srl $i1,$rk7,8 + and $i0,0xff + and $i1,0xff + and $i2,$rk7,0xff + srl $i3,$rk7,24 + $PTR_ADD $i0,$Tbl + $PTR_ADD $i1,$Tbl + $PTR_ADD $i2,$Tbl + $PTR_ADD $i3,$Tbl + lbu $i0,0($i0) + lbu $i1,0($i1) + lbu $i2,0($i2) + lbu $i3,0($i3) + + sw $rk0,0($key) + sw $rk1,4($key) + sw $rk2,8($key) + sw $rk3,12($key) + sw $rk4,16($key) + sw $rk5,20($key) + sw $rk6,24($key) + sw $rk7,28($key) + sub $cnt,1 + + _bias $i0,24 + _bias $i1,16 + _bias $i2,8 + _bias $i3,0 + + xor $rk0,$i0 + lw $i0,0($rcon) + xor $rk0,$i1 + xor $rk0,$i2 + xor $rk0,$i3 + xor $rk0,$i0 + + xor $rk1,$rk0 + xor $rk2,$rk1 + xor $rk3,$rk2 + beqz $cnt,.L256bits_done + + srl $i0,$rk3,24 + srl $i1,$rk3,16 + srl $i2,$rk3,8 + and $i3,$rk3,0xff + and $i1,0xff + and $i2,0xff + $PTR_ADD $i0,$Tbl + $PTR_ADD $i1,$Tbl + $PTR_ADD $i2,$Tbl + $PTR_ADD $i3,$Tbl + lbu $i0,0($i0) + lbu $i1,0($i1) + lbu $i2,0($i2) + lbu $i3,0($i3) + sll $i0,24 + sll $i1,16 + sll $i2,8 + + xor $rk4,$i0 + xor $rk4,$i1 + xor $rk4,$i2 + xor $rk4,$i3 + + xor $rk5,$rk4 + xor $rk6,$rk5 + xor $rk7,$rk6 + + $PTR_ADD $key,32 + .set noreorder + b .L256bits + $PTR_ADD $rcon,4 + +.L256bits_done: + sw $rk0,32($key) + sw $rk1,36($key) + sw $rk2,40($key) + li $cnt,14 + sw $rk3,44($key) + li $t0,0 + sw $cnt,48($key) + $PTR_SUB $key,12*16 + +.Lekey_done: + jr $ra + nop +.end _mips_AES_set_encrypt_key + +.globl AES_set_encrypt_key +.ent AES_set_encrypt_key +AES_set_encrypt_key: + .frame $sp,$FRAMESIZE,$ra + .mask $SAVED_REGS_MASK,-$SZREG + .set noreorder +___ +$code.=<<___ if ($flavour =~ /o32/i); # o32 PIC-ification + .cpload $pf +___ +$code.=<<___; + $PTR_SUB $sp,$FRAMESIZE + $REG_S $ra,$FRAMESIZE-1*$SZREG($sp) + $REG_S $fp,$FRAMESIZE-2*$SZREG($sp) +___ +$code.=<<___ if ($flavour =~ /nubi/i); # optimize non-nubi prologue + $REG_S $s3,$FRAMESIZE-3*$SZREG($sp) + $REG_S $s2,$FRAMESIZE-4*$SZREG($sp) + $REG_S $s1,$FRAMESIZE-5*$SZREG($sp) + $REG_S $s0,$FRAMESIZE-6*$SZREG($sp) + $REG_S $gp,$FRAMESIZE-7*$SZREG($sp) +___ +$code.=<<___ if ($flavour !~ /o32/i); # non-o32 PIC-ification + .cplocal $Tbl + .cpsetup $pf,$zero,AES_set_encrypt_key +___ +$code.=<<___; + .set reorder + $PTR_LA $Tbl,AES_Te4 # PIC-ified 'load address' + + bal _mips_AES_set_encrypt_key + + .set noreorder + move $a0,$t0 + $REG_L $ra,$FRAMESIZE-1*$SZREG($sp) + $REG_L $fp,$FRAMESIZE-2*$SZREG($sp) +___ +$code.=<<___ if ($flavour =~ /nubi/i); + $REG_L $s3,$FRAMESIZE-11*$SZREG($sp) + $REG_L $s2,$FRAMESIZE-12*$SZREG($sp) + $REG_L $s1,$FRAMESIZE-13*$SZREG($sp) + $REG_L $s0,$FRAMESIZE-14*$SZREG($sp) + $REG_L $gp,$FRAMESIZE-15*$SZREG($sp) +___ +$code.=<<___; + jr $ra + $PTR_ADD $sp,$FRAMESIZE +.end AES_set_encrypt_key +___ + +my ($head,$tail)=($inp,$bits); +my ($tp1,$tp2,$tp4,$tp8,$tp9,$tpb,$tpd,$tpe)=($a4,$a5,$a6,$a7,$s0,$s1,$s2,$s3); +my ($m,$x80808080,$x7f7f7f7f,$x1b1b1b1b)=($at,$t0,$t1,$t2); +$code.=<<___; +.align 5 +.globl AES_set_decrypt_key +.ent AES_set_decrypt_key +AES_set_decrypt_key: + .frame $sp,$FRAMESIZE,$ra + .mask $SAVED_REGS_MASK,-$SZREG + .set noreorder +___ +$code.=<<___ if ($flavour =~ /o32/i); # o32 PIC-ification + .cpload $pf +___ +$code.=<<___; + $PTR_SUB $sp,$FRAMESIZE + $REG_S $ra,$FRAMESIZE-1*$SZREG($sp) + $REG_S $fp,$FRAMESIZE-2*$SZREG($sp) +___ +$code.=<<___ if ($flavour =~ /nubi/i); # optimize non-nubi prologue + $REG_S $s3,$FRAMESIZE-3*$SZREG($sp) + $REG_S $s2,$FRAMESIZE-4*$SZREG($sp) + $REG_S $s1,$FRAMESIZE-5*$SZREG($sp) + $REG_S $s0,$FRAMESIZE-6*$SZREG($sp) + $REG_S $gp,$FRAMESIZE-7*$SZREG($sp) +___ +$code.=<<___ if ($flavour !~ /o32/i); # non-o32 PIC-ification + .cplocal $Tbl + .cpsetup $pf,$zero,AES_set_decrypt_key +___ +$code.=<<___; + .set reorder + $PTR_LA $Tbl,AES_Te4 # PIC-ified 'load address' + + bal _mips_AES_set_encrypt_key + + bltz $t0,.Ldkey_done + + sll $at,$cnt,4 + $PTR_ADD $head,$key,0 + $PTR_ADD $tail,$key,$at +.align 4 +.Lswap: + lw $rk0,0($head) + lw $rk1,4($head) + lw $rk2,8($head) + lw $rk3,12($head) + lw $rk4,0($tail) + lw $rk5,4($tail) + lw $rk6,8($tail) + lw $rk7,12($tail) + sw $rk0,0($tail) + sw $rk1,4($tail) + sw $rk2,8($tail) + sw $rk3,12($tail) + $PTR_ADD $head,16 + $PTR_SUB $tail,16 + sw $rk4,-16($head) + sw $rk5,-12($head) + sw $rk6,-8($head) + sw $rk7,-4($head) + bne $head,$tail,.Lswap + + lw $tp1,16($key) # modulo-scheduled + lui $x80808080,0x8080 + sub $cnt,1 + or $x80808080,0x8080 + sll $cnt,2 + $PTR_ADD $key,16 + lui $x1b1b1b1b,0x1b1b + nor $x7f7f7f7f,$zero,$x80808080 + or $x1b1b1b1b,0x1b1b +.align 4 +.Lmix: + and $m,$tp1,$x80808080 + and $tp2,$tp1,$x7f7f7f7f + srl $tp4,$m,7 + addu $tp2,$tp2 # tp2<<1 + subu $m,$tp4 + and $m,$x1b1b1b1b + xor $tp2,$m + + and $m,$tp2,$x80808080 + and $tp4,$tp2,$x7f7f7f7f + srl $tp8,$m,7 + addu $tp4,$tp4 # tp4<<1 + subu $m,$tp8 + and $m,$x1b1b1b1b + xor $tp4,$m + + and $m,$tp4,$x80808080 + and $tp8,$tp4,$x7f7f7f7f + srl $tp9,$m,7 + addu $tp8,$tp8 # tp8<<1 + subu $m,$tp9 + and $m,$x1b1b1b1b + xor $tp8,$m + + xor $tp9,$tp8,$tp1 + xor $tpe,$tp8,$tp4 + xor $tpb,$tp9,$tp2 + xor $tpd,$tp9,$tp4 + +#if defined(_MIPS_ARCH_MIPS32R2) || defined(_MIPS_ARCH_MIPS64R2) + rotr $tp1,$tpd,16 + xor $tpe,$tp2 + rotr $tp2,$tp9,8 + xor $tpe,$tp1 + rotr $tp4,$tpb,24 + xor $tpe,$tp2 + lw $tp1,4($key) # modulo-scheduled + xor $tpe,$tp4 +#else + _ror $tp1,$tpd,16 + xor $tpe,$tp2 + _ror $tp2,$tpd,-16 + xor $tpe,$tp1 + _ror $tp1,$tp9,8 + xor $tpe,$tp2 + _ror $tp2,$tp9,-24 + xor $tpe,$tp1 + _ror $tp1,$tpb,24 + xor $tpe,$tp2 + _ror $tp2,$tpb,-8 + xor $tpe,$tp1 + lw $tp1,4($key) # modulo-scheduled + xor $tpe,$tp2 +#endif + sub $cnt,1 + sw $tpe,0($key) + $PTR_ADD $key,4 + bnez $cnt,.Lmix + + li $t0,0 +.Ldkey_done: + .set noreorder + move $a0,$t0 + $REG_L $ra,$FRAMESIZE-1*$SZREG($sp) + $REG_L $fp,$FRAMESIZE-2*$SZREG($sp) +___ +$code.=<<___ if ($flavour =~ /nubi/i); + $REG_L $s3,$FRAMESIZE-11*$SZREG($sp) + $REG_L $s2,$FRAMESIZE-12*$SZREG($sp) + $REG_L $s1,$FRAMESIZE-13*$SZREG($sp) + $REG_L $s0,$FRAMESIZE-14*$SZREG($sp) + $REG_L $gp,$FRAMESIZE-15*$SZREG($sp) +___ +$code.=<<___; + jr $ra + $PTR_ADD $sp,$FRAMESIZE +.end AES_set_decrypt_key +___ +}}} + +###################################################################### +# Tables are kept in endian-neutral manner +$code.=<<___; +.rdata +.align 10 +AES_Te: +.byte 0xc6,0x63,0x63,0xa5, 0xf8,0x7c,0x7c,0x84 # Te0 +.byte 0xee,0x77,0x77,0x99, 0xf6,0x7b,0x7b,0x8d +.byte 0xff,0xf2,0xf2,0x0d, 0xd6,0x6b,0x6b,0xbd +.byte 0xde,0x6f,0x6f,0xb1, 0x91,0xc5,0xc5,0x54 +.byte 0x60,0x30,0x30,0x50, 0x02,0x01,0x01,0x03 +.byte 0xce,0x67,0x67,0xa9, 0x56,0x2b,0x2b,0x7d +.byte 0xe7,0xfe,0xfe,0x19, 0xb5,0xd7,0xd7,0x62 +.byte 0x4d,0xab,0xab,0xe6, 0xec,0x76,0x76,0x9a +.byte 0x8f,0xca,0xca,0x45, 0x1f,0x82,0x82,0x9d +.byte 0x89,0xc9,0xc9,0x40, 0xfa,0x7d,0x7d,0x87 +.byte 0xef,0xfa,0xfa,0x15, 0xb2,0x59,0x59,0xeb +.byte 0x8e,0x47,0x47,0xc9, 0xfb,0xf0,0xf0,0x0b +.byte 0x41,0xad,0xad,0xec, 0xb3,0xd4,0xd4,0x67 +.byte 0x5f,0xa2,0xa2,0xfd, 0x45,0xaf,0xaf,0xea +.byte 0x23,0x9c,0x9c,0xbf, 0x53,0xa4,0xa4,0xf7 +.byte 0xe4,0x72,0x72,0x96, 0x9b,0xc0,0xc0,0x5b +.byte 0x75,0xb7,0xb7,0xc2, 0xe1,0xfd,0xfd,0x1c +.byte 0x3d,0x93,0x93,0xae, 0x4c,0x26,0x26,0x6a +.byte 0x6c,0x36,0x36,0x5a, 0x7e,0x3f,0x3f,0x41 +.byte 0xf5,0xf7,0xf7,0x02, 0x83,0xcc,0xcc,0x4f +.byte 0x68,0x34,0x34,0x5c, 0x51,0xa5,0xa5,0xf4 +.byte 0xd1,0xe5,0xe5,0x34, 0xf9,0xf1,0xf1,0x08 +.byte 0xe2,0x71,0x71,0x93, 0xab,0xd8,0xd8,0x73 +.byte 0x62,0x31,0x31,0x53, 0x2a,0x15,0x15,0x3f +.byte 0x08,0x04,0x04,0x0c, 0x95,0xc7,0xc7,0x52 +.byte 0x46,0x23,0x23,0x65, 0x9d,0xc3,0xc3,0x5e +.byte 0x30,0x18,0x18,0x28, 0x37,0x96,0x96,0xa1 +.byte 0x0a,0x05,0x05,0x0f, 0x2f,0x9a,0x9a,0xb5 +.byte 0x0e,0x07,0x07,0x09, 0x24,0x12,0x12,0x36 +.byte 0x1b,0x80,0x80,0x9b, 0xdf,0xe2,0xe2,0x3d +.byte 0xcd,0xeb,0xeb,0x26, 0x4e,0x27,0x27,0x69 +.byte 0x7f,0xb2,0xb2,0xcd, 0xea,0x75,0x75,0x9f +.byte 0x12,0x09,0x09,0x1b, 0x1d,0x83,0x83,0x9e +.byte 0x58,0x2c,0x2c,0x74, 0x34,0x1a,0x1a,0x2e +.byte 0x36,0x1b,0x1b,0x2d, 0xdc,0x6e,0x6e,0xb2 +.byte 0xb4,0x5a,0x5a,0xee, 0x5b,0xa0,0xa0,0xfb +.byte 0xa4,0x52,0x52,0xf6, 0x76,0x3b,0x3b,0x4d +.byte 0xb7,0xd6,0xd6,0x61, 0x7d,0xb3,0xb3,0xce +.byte 0x52,0x29,0x29,0x7b, 0xdd,0xe3,0xe3,0x3e +.byte 0x5e,0x2f,0x2f,0x71, 0x13,0x84,0x84,0x97 +.byte 0xa6,0x53,0x53,0xf5, 0xb9,0xd1,0xd1,0x68 +.byte 0x00,0x00,0x00,0x00, 0xc1,0xed,0xed,0x2c +.byte 0x40,0x20,0x20,0x60, 0xe3,0xfc,0xfc,0x1f +.byte 0x79,0xb1,0xb1,0xc8, 0xb6,0x5b,0x5b,0xed +.byte 0xd4,0x6a,0x6a,0xbe, 0x8d,0xcb,0xcb,0x46 +.byte 0x67,0xbe,0xbe,0xd9, 0x72,0x39,0x39,0x4b +.byte 0x94,0x4a,0x4a,0xde, 0x98,0x4c,0x4c,0xd4 +.byte 0xb0,0x58,0x58,0xe8, 0x85,0xcf,0xcf,0x4a +.byte 0xbb,0xd0,0xd0,0x6b, 0xc5,0xef,0xef,0x2a +.byte 0x4f,0xaa,0xaa,0xe5, 0xed,0xfb,0xfb,0x16 +.byte 0x86,0x43,0x43,0xc5, 0x9a,0x4d,0x4d,0xd7 +.byte 0x66,0x33,0x33,0x55, 0x11,0x85,0x85,0x94 +.byte 0x8a,0x45,0x45,0xcf, 0xe9,0xf9,0xf9,0x10 +.byte 0x04,0x02,0x02,0x06, 0xfe,0x7f,0x7f,0x81 +.byte 0xa0,0x50,0x50,0xf0, 0x78,0x3c,0x3c,0x44 +.byte 0x25,0x9f,0x9f,0xba, 0x4b,0xa8,0xa8,0xe3 +.byte 0xa2,0x51,0x51,0xf3, 0x5d,0xa3,0xa3,0xfe +.byte 0x80,0x40,0x40,0xc0, 0x05,0x8f,0x8f,0x8a +.byte 0x3f,0x92,0x92,0xad, 0x21,0x9d,0x9d,0xbc +.byte 0x70,0x38,0x38,0x48, 0xf1,0xf5,0xf5,0x04 +.byte 0x63,0xbc,0xbc,0xdf, 0x77,0xb6,0xb6,0xc1 +.byte 0xaf,0xda,0xda,0x75, 0x42,0x21,0x21,0x63 +.byte 0x20,0x10,0x10,0x30, 0xe5,0xff,0xff,0x1a +.byte 0xfd,0xf3,0xf3,0x0e, 0xbf,0xd2,0xd2,0x6d +.byte 0x81,0xcd,0xcd,0x4c, 0x18,0x0c,0x0c,0x14 +.byte 0x26,0x13,0x13,0x35, 0xc3,0xec,0xec,0x2f +.byte 0xbe,0x5f,0x5f,0xe1, 0x35,0x97,0x97,0xa2 +.byte 0x88,0x44,0x44,0xcc, 0x2e,0x17,0x17,0x39 +.byte 0x93,0xc4,0xc4,0x57, 0x55,0xa7,0xa7,0xf2 +.byte 0xfc,0x7e,0x7e,0x82, 0x7a,0x3d,0x3d,0x47 +.byte 0xc8,0x64,0x64,0xac, 0xba,0x5d,0x5d,0xe7 +.byte 0x32,0x19,0x19,0x2b, 0xe6,0x73,0x73,0x95 +.byte 0xc0,0x60,0x60,0xa0, 0x19,0x81,0x81,0x98 +.byte 0x9e,0x4f,0x4f,0xd1, 0xa3,0xdc,0xdc,0x7f +.byte 0x44,0x22,0x22,0x66, 0x54,0x2a,0x2a,0x7e +.byte 0x3b,0x90,0x90,0xab, 0x0b,0x88,0x88,0x83 +.byte 0x8c,0x46,0x46,0xca, 0xc7,0xee,0xee,0x29 +.byte 0x6b,0xb8,0xb8,0xd3, 0x28,0x14,0x14,0x3c +.byte 0xa7,0xde,0xde,0x79, 0xbc,0x5e,0x5e,0xe2 +.byte 0x16,0x0b,0x0b,0x1d, 0xad,0xdb,0xdb,0x76 +.byte 0xdb,0xe0,0xe0,0x3b, 0x64,0x32,0x32,0x56 +.byte 0x74,0x3a,0x3a,0x4e, 0x14,0x0a,0x0a,0x1e +.byte 0x92,0x49,0x49,0xdb, 0x0c,0x06,0x06,0x0a +.byte 0x48,0x24,0x24,0x6c, 0xb8,0x5c,0x5c,0xe4 +.byte 0x9f,0xc2,0xc2,0x5d, 0xbd,0xd3,0xd3,0x6e +.byte 0x43,0xac,0xac,0xef, 0xc4,0x62,0x62,0xa6 +.byte 0x39,0x91,0x91,0xa8, 0x31,0x95,0x95,0xa4 +.byte 0xd3,0xe4,0xe4,0x37, 0xf2,0x79,0x79,0x8b +.byte 0xd5,0xe7,0xe7,0x32, 0x8b,0xc8,0xc8,0x43 +.byte 0x6e,0x37,0x37,0x59, 0xda,0x6d,0x6d,0xb7 +.byte 0x01,0x8d,0x8d,0x8c, 0xb1,0xd5,0xd5,0x64 +.byte 0x9c,0x4e,0x4e,0xd2, 0x49,0xa9,0xa9,0xe0 +.byte 0xd8,0x6c,0x6c,0xb4, 0xac,0x56,0x56,0xfa +.byte 0xf3,0xf4,0xf4,0x07, 0xcf,0xea,0xea,0x25 +.byte 0xca,0x65,0x65,0xaf, 0xf4,0x7a,0x7a,0x8e +.byte 0x47,0xae,0xae,0xe9, 0x10,0x08,0x08,0x18 +.byte 0x6f,0xba,0xba,0xd5, 0xf0,0x78,0x78,0x88 +.byte 0x4a,0x25,0x25,0x6f, 0x5c,0x2e,0x2e,0x72 +.byte 0x38,0x1c,0x1c,0x24, 0x57,0xa6,0xa6,0xf1 +.byte 0x73,0xb4,0xb4,0xc7, 0x97,0xc6,0xc6,0x51 +.byte 0xcb,0xe8,0xe8,0x23, 0xa1,0xdd,0xdd,0x7c +.byte 0xe8,0x74,0x74,0x9c, 0x3e,0x1f,0x1f,0x21 +.byte 0x96,0x4b,0x4b,0xdd, 0x61,0xbd,0xbd,0xdc +.byte 0x0d,0x8b,0x8b,0x86, 0x0f,0x8a,0x8a,0x85 +.byte 0xe0,0x70,0x70,0x90, 0x7c,0x3e,0x3e,0x42 +.byte 0x71,0xb5,0xb5,0xc4, 0xcc,0x66,0x66,0xaa +.byte 0x90,0x48,0x48,0xd8, 0x06,0x03,0x03,0x05 +.byte 0xf7,0xf6,0xf6,0x01, 0x1c,0x0e,0x0e,0x12 +.byte 0xc2,0x61,0x61,0xa3, 0x6a,0x35,0x35,0x5f +.byte 0xae,0x57,0x57,0xf9, 0x69,0xb9,0xb9,0xd0 +.byte 0x17,0x86,0x86,0x91, 0x99,0xc1,0xc1,0x58 +.byte 0x3a,0x1d,0x1d,0x27, 0x27,0x9e,0x9e,0xb9 +.byte 0xd9,0xe1,0xe1,0x38, 0xeb,0xf8,0xf8,0x13 +.byte 0x2b,0x98,0x98,0xb3, 0x22,0x11,0x11,0x33 +.byte 0xd2,0x69,0x69,0xbb, 0xa9,0xd9,0xd9,0x70 +.byte 0x07,0x8e,0x8e,0x89, 0x33,0x94,0x94,0xa7 +.byte 0x2d,0x9b,0x9b,0xb6, 0x3c,0x1e,0x1e,0x22 +.byte 0x15,0x87,0x87,0x92, 0xc9,0xe9,0xe9,0x20 +.byte 0x87,0xce,0xce,0x49, 0xaa,0x55,0x55,0xff +.byte 0x50,0x28,0x28,0x78, 0xa5,0xdf,0xdf,0x7a +.byte 0x03,0x8c,0x8c,0x8f, 0x59,0xa1,0xa1,0xf8 +.byte 0x09,0x89,0x89,0x80, 0x1a,0x0d,0x0d,0x17 +.byte 0x65,0xbf,0xbf,0xda, 0xd7,0xe6,0xe6,0x31 +.byte 0x84,0x42,0x42,0xc6, 0xd0,0x68,0x68,0xb8 +.byte 0x82,0x41,0x41,0xc3, 0x29,0x99,0x99,0xb0 +.byte 0x5a,0x2d,0x2d,0x77, 0x1e,0x0f,0x0f,0x11 +.byte 0x7b,0xb0,0xb0,0xcb, 0xa8,0x54,0x54,0xfc +.byte 0x6d,0xbb,0xbb,0xd6, 0x2c,0x16,0x16,0x3a + +AES_Td: +.byte 0x51,0xf4,0xa7,0x50, 0x7e,0x41,0x65,0x53 # Td0 +.byte 0x1a,0x17,0xa4,0xc3, 0x3a,0x27,0x5e,0x96 +.byte 0x3b,0xab,0x6b,0xcb, 0x1f,0x9d,0x45,0xf1 +.byte 0xac,0xfa,0x58,0xab, 0x4b,0xe3,0x03,0x93 +.byte 0x20,0x30,0xfa,0x55, 0xad,0x76,0x6d,0xf6 +.byte 0x88,0xcc,0x76,0x91, 0xf5,0x02,0x4c,0x25 +.byte 0x4f,0xe5,0xd7,0xfc, 0xc5,0x2a,0xcb,0xd7 +.byte 0x26,0x35,0x44,0x80, 0xb5,0x62,0xa3,0x8f +.byte 0xde,0xb1,0x5a,0x49, 0x25,0xba,0x1b,0x67 +.byte 0x45,0xea,0x0e,0x98, 0x5d,0xfe,0xc0,0xe1 +.byte 0xc3,0x2f,0x75,0x02, 0x81,0x4c,0xf0,0x12 +.byte 0x8d,0x46,0x97,0xa3, 0x6b,0xd3,0xf9,0xc6 +.byte 0x03,0x8f,0x5f,0xe7, 0x15,0x92,0x9c,0x95 +.byte 0xbf,0x6d,0x7a,0xeb, 0x95,0x52,0x59,0xda +.byte 0xd4,0xbe,0x83,0x2d, 0x58,0x74,0x21,0xd3 +.byte 0x49,0xe0,0x69,0x29, 0x8e,0xc9,0xc8,0x44 +.byte 0x75,0xc2,0x89,0x6a, 0xf4,0x8e,0x79,0x78 +.byte 0x99,0x58,0x3e,0x6b, 0x27,0xb9,0x71,0xdd +.byte 0xbe,0xe1,0x4f,0xb6, 0xf0,0x88,0xad,0x17 +.byte 0xc9,0x20,0xac,0x66, 0x7d,0xce,0x3a,0xb4 +.byte 0x63,0xdf,0x4a,0x18, 0xe5,0x1a,0x31,0x82 +.byte 0x97,0x51,0x33,0x60, 0x62,0x53,0x7f,0x45 +.byte 0xb1,0x64,0x77,0xe0, 0xbb,0x6b,0xae,0x84 +.byte 0xfe,0x81,0xa0,0x1c, 0xf9,0x08,0x2b,0x94 +.byte 0x70,0x48,0x68,0x58, 0x8f,0x45,0xfd,0x19 +.byte 0x94,0xde,0x6c,0x87, 0x52,0x7b,0xf8,0xb7 +.byte 0xab,0x73,0xd3,0x23, 0x72,0x4b,0x02,0xe2 +.byte 0xe3,0x1f,0x8f,0x57, 0x66,0x55,0xab,0x2a +.byte 0xb2,0xeb,0x28,0x07, 0x2f,0xb5,0xc2,0x03 +.byte 0x86,0xc5,0x7b,0x9a, 0xd3,0x37,0x08,0xa5 +.byte 0x30,0x28,0x87,0xf2, 0x23,0xbf,0xa5,0xb2 +.byte 0x02,0x03,0x6a,0xba, 0xed,0x16,0x82,0x5c +.byte 0x8a,0xcf,0x1c,0x2b, 0xa7,0x79,0xb4,0x92 +.byte 0xf3,0x07,0xf2,0xf0, 0x4e,0x69,0xe2,0xa1 +.byte 0x65,0xda,0xf4,0xcd, 0x06,0x05,0xbe,0xd5 +.byte 0xd1,0x34,0x62,0x1f, 0xc4,0xa6,0xfe,0x8a +.byte 0x34,0x2e,0x53,0x9d, 0xa2,0xf3,0x55,0xa0 +.byte 0x05,0x8a,0xe1,0x32, 0xa4,0xf6,0xeb,0x75 +.byte 0x0b,0x83,0xec,0x39, 0x40,0x60,0xef,0xaa +.byte 0x5e,0x71,0x9f,0x06, 0xbd,0x6e,0x10,0x51 +.byte 0x3e,0x21,0x8a,0xf9, 0x96,0xdd,0x06,0x3d +.byte 0xdd,0x3e,0x05,0xae, 0x4d,0xe6,0xbd,0x46 +.byte 0x91,0x54,0x8d,0xb5, 0x71,0xc4,0x5d,0x05 +.byte 0x04,0x06,0xd4,0x6f, 0x60,0x50,0x15,0xff +.byte 0x19,0x98,0xfb,0x24, 0xd6,0xbd,0xe9,0x97 +.byte 0x89,0x40,0x43,0xcc, 0x67,0xd9,0x9e,0x77 +.byte 0xb0,0xe8,0x42,0xbd, 0x07,0x89,0x8b,0x88 +.byte 0xe7,0x19,0x5b,0x38, 0x79,0xc8,0xee,0xdb +.byte 0xa1,0x7c,0x0a,0x47, 0x7c,0x42,0x0f,0xe9 +.byte 0xf8,0x84,0x1e,0xc9, 0x00,0x00,0x00,0x00 +.byte 0x09,0x80,0x86,0x83, 0x32,0x2b,0xed,0x48 +.byte 0x1e,0x11,0x70,0xac, 0x6c,0x5a,0x72,0x4e +.byte 0xfd,0x0e,0xff,0xfb, 0x0f,0x85,0x38,0x56 +.byte 0x3d,0xae,0xd5,0x1e, 0x36,0x2d,0x39,0x27 +.byte 0x0a,0x0f,0xd9,0x64, 0x68,0x5c,0xa6,0x21 +.byte 0x9b,0x5b,0x54,0xd1, 0x24,0x36,0x2e,0x3a +.byte 0x0c,0x0a,0x67,0xb1, 0x93,0x57,0xe7,0x0f +.byte 0xb4,0xee,0x96,0xd2, 0x1b,0x9b,0x91,0x9e +.byte 0x80,0xc0,0xc5,0x4f, 0x61,0xdc,0x20,0xa2 +.byte 0x5a,0x77,0x4b,0x69, 0x1c,0x12,0x1a,0x16 +.byte 0xe2,0x93,0xba,0x0a, 0xc0,0xa0,0x2a,0xe5 +.byte 0x3c,0x22,0xe0,0x43, 0x12,0x1b,0x17,0x1d +.byte 0x0e,0x09,0x0d,0x0b, 0xf2,0x8b,0xc7,0xad +.byte 0x2d,0xb6,0xa8,0xb9, 0x14,0x1e,0xa9,0xc8 +.byte 0x57,0xf1,0x19,0x85, 0xaf,0x75,0x07,0x4c +.byte 0xee,0x99,0xdd,0xbb, 0xa3,0x7f,0x60,0xfd +.byte 0xf7,0x01,0x26,0x9f, 0x5c,0x72,0xf5,0xbc +.byte 0x44,0x66,0x3b,0xc5, 0x5b,0xfb,0x7e,0x34 +.byte 0x8b,0x43,0x29,0x76, 0xcb,0x23,0xc6,0xdc +.byte 0xb6,0xed,0xfc,0x68, 0xb8,0xe4,0xf1,0x63 +.byte 0xd7,0x31,0xdc,0xca, 0x42,0x63,0x85,0x10 +.byte 0x13,0x97,0x22,0x40, 0x84,0xc6,0x11,0x20 +.byte 0x85,0x4a,0x24,0x7d, 0xd2,0xbb,0x3d,0xf8 +.byte 0xae,0xf9,0x32,0x11, 0xc7,0x29,0xa1,0x6d +.byte 0x1d,0x9e,0x2f,0x4b, 0xdc,0xb2,0x30,0xf3 +.byte 0x0d,0x86,0x52,0xec, 0x77,0xc1,0xe3,0xd0 +.byte 0x2b,0xb3,0x16,0x6c, 0xa9,0x70,0xb9,0x99 +.byte 0x11,0x94,0x48,0xfa, 0x47,0xe9,0x64,0x22 +.byte 0xa8,0xfc,0x8c,0xc4, 0xa0,0xf0,0x3f,0x1a +.byte 0x56,0x7d,0x2c,0xd8, 0x22,0x33,0x90,0xef +.byte 0x87,0x49,0x4e,0xc7, 0xd9,0x38,0xd1,0xc1 +.byte 0x8c,0xca,0xa2,0xfe, 0x98,0xd4,0x0b,0x36 +.byte 0xa6,0xf5,0x81,0xcf, 0xa5,0x7a,0xde,0x28 +.byte 0xda,0xb7,0x8e,0x26, 0x3f,0xad,0xbf,0xa4 +.byte 0x2c,0x3a,0x9d,0xe4, 0x50,0x78,0x92,0x0d +.byte 0x6a,0x5f,0xcc,0x9b, 0x54,0x7e,0x46,0x62 +.byte 0xf6,0x8d,0x13,0xc2, 0x90,0xd8,0xb8,0xe8 +.byte 0x2e,0x39,0xf7,0x5e, 0x82,0xc3,0xaf,0xf5 +.byte 0x9f,0x5d,0x80,0xbe, 0x69,0xd0,0x93,0x7c +.byte 0x6f,0xd5,0x2d,0xa9, 0xcf,0x25,0x12,0xb3 +.byte 0xc8,0xac,0x99,0x3b, 0x10,0x18,0x7d,0xa7 +.byte 0xe8,0x9c,0x63,0x6e, 0xdb,0x3b,0xbb,0x7b +.byte 0xcd,0x26,0x78,0x09, 0x6e,0x59,0x18,0xf4 +.byte 0xec,0x9a,0xb7,0x01, 0x83,0x4f,0x9a,0xa8 +.byte 0xe6,0x95,0x6e,0x65, 0xaa,0xff,0xe6,0x7e +.byte 0x21,0xbc,0xcf,0x08, 0xef,0x15,0xe8,0xe6 +.byte 0xba,0xe7,0x9b,0xd9, 0x4a,0x6f,0x36,0xce +.byte 0xea,0x9f,0x09,0xd4, 0x29,0xb0,0x7c,0xd6 +.byte 0x31,0xa4,0xb2,0xaf, 0x2a,0x3f,0x23,0x31 +.byte 0xc6,0xa5,0x94,0x30, 0x35,0xa2,0x66,0xc0 +.byte 0x74,0x4e,0xbc,0x37, 0xfc,0x82,0xca,0xa6 +.byte 0xe0,0x90,0xd0,0xb0, 0x33,0xa7,0xd8,0x15 +.byte 0xf1,0x04,0x98,0x4a, 0x41,0xec,0xda,0xf7 +.byte 0x7f,0xcd,0x50,0x0e, 0x17,0x91,0xf6,0x2f +.byte 0x76,0x4d,0xd6,0x8d, 0x43,0xef,0xb0,0x4d +.byte 0xcc,0xaa,0x4d,0x54, 0xe4,0x96,0x04,0xdf +.byte 0x9e,0xd1,0xb5,0xe3, 0x4c,0x6a,0x88,0x1b +.byte 0xc1,0x2c,0x1f,0xb8, 0x46,0x65,0x51,0x7f +.byte 0x9d,0x5e,0xea,0x04, 0x01,0x8c,0x35,0x5d +.byte 0xfa,0x87,0x74,0x73, 0xfb,0x0b,0x41,0x2e +.byte 0xb3,0x67,0x1d,0x5a, 0x92,0xdb,0xd2,0x52 +.byte 0xe9,0x10,0x56,0x33, 0x6d,0xd6,0x47,0x13 +.byte 0x9a,0xd7,0x61,0x8c, 0x37,0xa1,0x0c,0x7a +.byte 0x59,0xf8,0x14,0x8e, 0xeb,0x13,0x3c,0x89 +.byte 0xce,0xa9,0x27,0xee, 0xb7,0x61,0xc9,0x35 +.byte 0xe1,0x1c,0xe5,0xed, 0x7a,0x47,0xb1,0x3c +.byte 0x9c,0xd2,0xdf,0x59, 0x55,0xf2,0x73,0x3f +.byte 0x18,0x14,0xce,0x79, 0x73,0xc7,0x37,0xbf +.byte 0x53,0xf7,0xcd,0xea, 0x5f,0xfd,0xaa,0x5b +.byte 0xdf,0x3d,0x6f,0x14, 0x78,0x44,0xdb,0x86 +.byte 0xca,0xaf,0xf3,0x81, 0xb9,0x68,0xc4,0x3e +.byte 0x38,0x24,0x34,0x2c, 0xc2,0xa3,0x40,0x5f +.byte 0x16,0x1d,0xc3,0x72, 0xbc,0xe2,0x25,0x0c +.byte 0x28,0x3c,0x49,0x8b, 0xff,0x0d,0x95,0x41 +.byte 0x39,0xa8,0x01,0x71, 0x08,0x0c,0xb3,0xde +.byte 0xd8,0xb4,0xe4,0x9c, 0x64,0x56,0xc1,0x90 +.byte 0x7b,0xcb,0x84,0x61, 0xd5,0x32,0xb6,0x70 +.byte 0x48,0x6c,0x5c,0x74, 0xd0,0xb8,0x57,0x42 + +.byte 0x52, 0x09, 0x6a, 0xd5, 0x30, 0x36, 0xa5, 0x38 # Td4 +.byte 0xbf, 0x40, 0xa3, 0x9e, 0x81, 0xf3, 0xd7, 0xfb +.byte 0x7c, 0xe3, 0x39, 0x82, 0x9b, 0x2f, 0xff, 0x87 +.byte 0x34, 0x8e, 0x43, 0x44, 0xc4, 0xde, 0xe9, 0xcb +.byte 0x54, 0x7b, 0x94, 0x32, 0xa6, 0xc2, 0x23, 0x3d +.byte 0xee, 0x4c, 0x95, 0x0b, 0x42, 0xfa, 0xc3, 0x4e +.byte 0x08, 0x2e, 0xa1, 0x66, 0x28, 0xd9, 0x24, 0xb2 +.byte 0x76, 0x5b, 0xa2, 0x49, 0x6d, 0x8b, 0xd1, 0x25 +.byte 0x72, 0xf8, 0xf6, 0x64, 0x86, 0x68, 0x98, 0x16 +.byte 0xd4, 0xa4, 0x5c, 0xcc, 0x5d, 0x65, 0xb6, 0x92 +.byte 0x6c, 0x70, 0x48, 0x50, 0xfd, 0xed, 0xb9, 0xda +.byte 0x5e, 0x15, 0x46, 0x57, 0xa7, 0x8d, 0x9d, 0x84 +.byte 0x90, 0xd8, 0xab, 0x00, 0x8c, 0xbc, 0xd3, 0x0a +.byte 0xf7, 0xe4, 0x58, 0x05, 0xb8, 0xb3, 0x45, 0x06 +.byte 0xd0, 0x2c, 0x1e, 0x8f, 0xca, 0x3f, 0x0f, 0x02 +.byte 0xc1, 0xaf, 0xbd, 0x03, 0x01, 0x13, 0x8a, 0x6b +.byte 0x3a, 0x91, 0x11, 0x41, 0x4f, 0x67, 0xdc, 0xea +.byte 0x97, 0xf2, 0xcf, 0xce, 0xf0, 0xb4, 0xe6, 0x73 +.byte 0x96, 0xac, 0x74, 0x22, 0xe7, 0xad, 0x35, 0x85 +.byte 0xe2, 0xf9, 0x37, 0xe8, 0x1c, 0x75, 0xdf, 0x6e +.byte 0x47, 0xf1, 0x1a, 0x71, 0x1d, 0x29, 0xc5, 0x89 +.byte 0x6f, 0xb7, 0x62, 0x0e, 0xaa, 0x18, 0xbe, 0x1b +.byte 0xfc, 0x56, 0x3e, 0x4b, 0xc6, 0xd2, 0x79, 0x20 +.byte 0x9a, 0xdb, 0xc0, 0xfe, 0x78, 0xcd, 0x5a, 0xf4 +.byte 0x1f, 0xdd, 0xa8, 0x33, 0x88, 0x07, 0xc7, 0x31 +.byte 0xb1, 0x12, 0x10, 0x59, 0x27, 0x80, 0xec, 0x5f +.byte 0x60, 0x51, 0x7f, 0xa9, 0x19, 0xb5, 0x4a, 0x0d +.byte 0x2d, 0xe5, 0x7a, 0x9f, 0x93, 0xc9, 0x9c, 0xef +.byte 0xa0, 0xe0, 0x3b, 0x4d, 0xae, 0x2a, 0xf5, 0xb0 +.byte 0xc8, 0xeb, 0xbb, 0x3c, 0x83, 0x53, 0x99, 0x61 +.byte 0x17, 0x2b, 0x04, 0x7e, 0xba, 0x77, 0xd6, 0x26 +.byte 0xe1, 0x69, 0x14, 0x63, 0x55, 0x21, 0x0c, 0x7d + +AES_Te4: +.byte 0x63, 0x7c, 0x77, 0x7b, 0xf2, 0x6b, 0x6f, 0xc5 # Te4 +.byte 0x30, 0x01, 0x67, 0x2b, 0xfe, 0xd7, 0xab, 0x76 +.byte 0xca, 0x82, 0xc9, 0x7d, 0xfa, 0x59, 0x47, 0xf0 +.byte 0xad, 0xd4, 0xa2, 0xaf, 0x9c, 0xa4, 0x72, 0xc0 +.byte 0xb7, 0xfd, 0x93, 0x26, 0x36, 0x3f, 0xf7, 0xcc +.byte 0x34, 0xa5, 0xe5, 0xf1, 0x71, 0xd8, 0x31, 0x15 +.byte 0x04, 0xc7, 0x23, 0xc3, 0x18, 0x96, 0x05, 0x9a +.byte 0x07, 0x12, 0x80, 0xe2, 0xeb, 0x27, 0xb2, 0x75 +.byte 0x09, 0x83, 0x2c, 0x1a, 0x1b, 0x6e, 0x5a, 0xa0 +.byte 0x52, 0x3b, 0xd6, 0xb3, 0x29, 0xe3, 0x2f, 0x84 +.byte 0x53, 0xd1, 0x00, 0xed, 0x20, 0xfc, 0xb1, 0x5b +.byte 0x6a, 0xcb, 0xbe, 0x39, 0x4a, 0x4c, 0x58, 0xcf +.byte 0xd0, 0xef, 0xaa, 0xfb, 0x43, 0x4d, 0x33, 0x85 +.byte 0x45, 0xf9, 0x02, 0x7f, 0x50, 0x3c, 0x9f, 0xa8 +.byte 0x51, 0xa3, 0x40, 0x8f, 0x92, 0x9d, 0x38, 0xf5 +.byte 0xbc, 0xb6, 0xda, 0x21, 0x10, 0xff, 0xf3, 0xd2 +.byte 0xcd, 0x0c, 0x13, 0xec, 0x5f, 0x97, 0x44, 0x17 +.byte 0xc4, 0xa7, 0x7e, 0x3d, 0x64, 0x5d, 0x19, 0x73 +.byte 0x60, 0x81, 0x4f, 0xdc, 0x22, 0x2a, 0x90, 0x88 +.byte 0x46, 0xee, 0xb8, 0x14, 0xde, 0x5e, 0x0b, 0xdb +.byte 0xe0, 0x32, 0x3a, 0x0a, 0x49, 0x06, 0x24, 0x5c +.byte 0xc2, 0xd3, 0xac, 0x62, 0x91, 0x95, 0xe4, 0x79 +.byte 0xe7, 0xc8, 0x37, 0x6d, 0x8d, 0xd5, 0x4e, 0xa9 +.byte 0x6c, 0x56, 0xf4, 0xea, 0x65, 0x7a, 0xae, 0x08 +.byte 0xba, 0x78, 0x25, 0x2e, 0x1c, 0xa6, 0xb4, 0xc6 +.byte 0xe8, 0xdd, 0x74, 0x1f, 0x4b, 0xbd, 0x8b, 0x8a +.byte 0x70, 0x3e, 0xb5, 0x66, 0x48, 0x03, 0xf6, 0x0e +.byte 0x61, 0x35, 0x57, 0xb9, 0x86, 0xc1, 0x1d, 0x9e +.byte 0xe1, 0xf8, 0x98, 0x11, 0x69, 0xd9, 0x8e, 0x94 +.byte 0x9b, 0x1e, 0x87, 0xe9, 0xce, 0x55, 0x28, 0xdf +.byte 0x8c, 0xa1, 0x89, 0x0d, 0xbf, 0xe6, 0x42, 0x68 +.byte 0x41, 0x99, 0x2d, 0x0f, 0xb0, 0x54, 0xbb, 0x16 + +.byte 0x01,0x00,0x00,0x00, 0x02,0x00,0x00,0x00 # rcon +.byte 0x04,0x00,0x00,0x00, 0x08,0x00,0x00,0x00 +.byte 0x10,0x00,0x00,0x00, 0x20,0x00,0x00,0x00 +.byte 0x40,0x00,0x00,0x00, 0x80,0x00,0x00,0x00 +.byte 0x1B,0x00,0x00,0x00, 0x36,0x00,0x00,0x00 +___ + +foreach (split("\n",$code)) { + s/\`([^\`]*)\`/eval $1/ge; + + # made-up _instructions, _xtr, _ins, _ror and _bias, cope + # with byte order dependencies... + if (/^\s+_/) { + s/(_[a-z]+\s+)(\$[0-9]+),([^,]+)(#.*)*$/$1$2,$2,$3/; + + s/_xtr\s+(\$[0-9]+),(\$[0-9]+),([0-9]+(\-2)*)/ + sprintf("srl\t$1,$2,%d",$big_endian ? eval($3) + : eval("24-$3"))/e or + s/_ins\s+(\$[0-9]+),(\$[0-9]+),([0-9]+)/ + sprintf("sll\t$1,$2,%d",$big_endian ? eval($3) + : eval("24-$3"))/e or + s/_ins2\s+(\$[0-9]+),(\$[0-9]+),([0-9]+)/ + sprintf("ins\t$1,$2,%d,8",$big_endian ? eval($3) + : eval("24-$3"))/e or + s/_ror\s+(\$[0-9]+),(\$[0-9]+),(\-?[0-9]+)/ + sprintf("srl\t$1,$2,%d",$big_endian ? eval($3) + : eval("$3*-1"))/e or + s/_bias\s+(\$[0-9]+),(\$[0-9]+),([0-9]+)/ + sprintf("sll\t$1,$2,%d",$big_endian ? eval($3) + : eval("($3-16)&31"))/e; + + s/srl\s+(\$[0-9]+),(\$[0-9]+),\-([0-9]+)/ + sprintf("sll\t$1,$2,$3")/e or + s/srl\s+(\$[0-9]+),(\$[0-9]+),0/ + sprintf("and\t$1,$2,0xff")/e or + s/(sll\s+\$[0-9]+,\$[0-9]+,0)/#$1/; + } + + # convert lwl/lwr and swr/swl to little-endian order + if (!$big_endian && /^\s+[sl]w[lr]\s+/) { + s/([sl]wl.*)([0-9]+)\((\$[0-9]+)\)/ + sprintf("$1%d($3)",eval("$2-$2%4+($2%4-1)&3"))/e or + s/([sl]wr.*)([0-9]+)\((\$[0-9]+)\)/ + sprintf("$1%d($3)",eval("$2-$2%4+($2%4+1)&3"))/e; + } + + if (!$big_endian) { + s/(rotr\s+\$[0-9]+,\$[0-9]+),([0-9]+)/sprintf("$1,%d",32-$2)/e; + s/(ext\s+\$[0-9]+,\$[0-9]+),([0-9]+),8/sprintf("$1,%d,8",24-$2)/e; + } + + print $_,"\n"; +} + +close STDOUT; diff --git a/openssl-1.1.0h/crypto/aes/asm/aes-parisc.pl b/openssl-1.1.0h/crypto/aes/asm/aes-parisc.pl new file mode 100644 index 0000000..2c785bc --- /dev/null +++ b/openssl-1.1.0h/crypto/aes/asm/aes-parisc.pl @@ -0,0 +1,1029 @@ +#! /usr/bin/env perl +# Copyright 2009-2016 The OpenSSL Project Authors. All Rights Reserved. +# +# Licensed under the OpenSSL license (the "License"). You may not use +# this file except in compliance with the License. You can obtain a copy +# in the file LICENSE in the source distribution or at +# https://www.openssl.org/source/license.html + + +# ==================================================================== +# Written by Andy Polyakov for the OpenSSL +# project. The module is, however, dual licensed under OpenSSL and +# CRYPTOGAMS licenses depending on where you obtain it. For further +# details see http://www.openssl.org/~appro/cryptogams/. +# ==================================================================== + +# AES for PA-RISC. +# +# June 2009. +# +# The module is mechanical transliteration of aes-sparcv9.pl, but with +# a twist: S-boxes are compressed even further down to 1K+256B. On +# PA-7100LC performance is ~40% better than gcc 3.2 generated code and +# is about 33 cycles per byte processed with 128-bit key. Newer CPUs +# perform at 16 cycles per byte. It's not faster than code generated +# by vendor compiler, but recall that it has compressed S-boxes, which +# requires extra processing. +# +# Special thanks to polarhome.com for providing HP-UX account. + +$flavour = shift; +$output = shift; +open STDOUT,">$output"; + +if ($flavour =~ /64/) { + $LEVEL ="2.0W"; + $SIZE_T =8; + $FRAME_MARKER =80; + $SAVED_RP =16; + $PUSH ="std"; + $PUSHMA ="std,ma"; + $POP ="ldd"; + $POPMB ="ldd,mb"; +} else { + $LEVEL ="1.0"; + $SIZE_T =4; + $FRAME_MARKER =48; + $SAVED_RP =20; + $PUSH ="stw"; + $PUSHMA ="stwm"; + $POP ="ldw"; + $POPMB ="ldwm"; +} + +$FRAME=16*$SIZE_T+$FRAME_MARKER;# 16 saved regs + frame marker + # [+ argument transfer] +$inp="%r26"; # arg0 +$out="%r25"; # arg1 +$key="%r24"; # arg2 + +($s0,$s1,$s2,$s3) = ("%r1","%r2","%r3","%r4"); +($t0,$t1,$t2,$t3) = ("%r5","%r6","%r7","%r8"); + +($acc0, $acc1, $acc2, $acc3, $acc4, $acc5, $acc6, $acc7, + $acc8, $acc9,$acc10,$acc11,$acc12,$acc13,$acc14,$acc15) = +("%r9","%r10","%r11","%r12","%r13","%r14","%r15","%r16", +"%r17","%r18","%r19","%r20","%r21","%r22","%r23","%r26"); + +$tbl="%r28"; +$rounds="%r29"; + +$code=<<___; + .LEVEL $LEVEL + .SPACE \$TEXT\$ + .SUBSPA \$CODE\$,QUAD=0,ALIGN=8,ACCESS=0x2C,CODE_ONLY + + .EXPORT AES_encrypt,ENTRY,ARGW0=GR,ARGW1=GR,ARGW2=GR + .ALIGN 64 +AES_encrypt + .PROC + .CALLINFO FRAME=`$FRAME-16*$SIZE_T`,NO_CALLS,SAVE_RP,ENTRY_GR=18 + .ENTRY + $PUSH %r2,-$SAVED_RP(%sp) ; standard prologue + $PUSHMA %r3,$FRAME(%sp) + $PUSH %r4,`-$FRAME+1*$SIZE_T`(%sp) + $PUSH %r5,`-$FRAME+2*$SIZE_T`(%sp) + $PUSH %r6,`-$FRAME+3*$SIZE_T`(%sp) + $PUSH %r7,`-$FRAME+4*$SIZE_T`(%sp) + $PUSH %r8,`-$FRAME+5*$SIZE_T`(%sp) + $PUSH %r9,`-$FRAME+6*$SIZE_T`(%sp) + $PUSH %r10,`-$FRAME+7*$SIZE_T`(%sp) + $PUSH %r11,`-$FRAME+8*$SIZE_T`(%sp) + $PUSH %r12,`-$FRAME+9*$SIZE_T`(%sp) + $PUSH %r13,`-$FRAME+10*$SIZE_T`(%sp) + $PUSH %r14,`-$FRAME+11*$SIZE_T`(%sp) + $PUSH %r15,`-$FRAME+12*$SIZE_T`(%sp) + $PUSH %r16,`-$FRAME+13*$SIZE_T`(%sp) + $PUSH %r17,`-$FRAME+14*$SIZE_T`(%sp) + $PUSH %r18,`-$FRAME+15*$SIZE_T`(%sp) + + blr %r0,$tbl + ldi 3,$t0 +L\$enc_pic + andcm $tbl,$t0,$tbl + ldo L\$AES_Te-L\$enc_pic($tbl),$tbl + + and $inp,$t0,$t0 + sub $inp,$t0,$inp + ldw 0($inp),$s0 + ldw 4($inp),$s1 + ldw 8($inp),$s2 + comib,= 0,$t0,L\$enc_inp_aligned + ldw 12($inp),$s3 + + sh3addl $t0,%r0,$t0 + subi 32,$t0,$t0 + mtctl $t0,%cr11 + ldw 16($inp),$t1 + vshd $s0,$s1,$s0 + vshd $s1,$s2,$s1 + vshd $s2,$s3,$s2 + vshd $s3,$t1,$s3 + +L\$enc_inp_aligned + bl _parisc_AES_encrypt,%r31 + nop + + extru,<> $out,31,2,%r0 + b L\$enc_out_aligned + nop + + _srm $s0,24,$acc0 + _srm $s0,16,$acc1 + stb $acc0,0($out) + _srm $s0,8,$acc2 + stb $acc1,1($out) + _srm $s1,24,$acc4 + stb $acc2,2($out) + _srm $s1,16,$acc5 + stb $s0,3($out) + _srm $s1,8,$acc6 + stb $acc4,4($out) + _srm $s2,24,$acc0 + stb $acc5,5($out) + _srm $s2,16,$acc1 + stb $acc6,6($out) + _srm $s2,8,$acc2 + stb $s1,7($out) + _srm $s3,24,$acc4 + stb $acc0,8($out) + _srm $s3,16,$acc5 + stb $acc1,9($out) + _srm $s3,8,$acc6 + stb $acc2,10($out) + stb $s2,11($out) + stb $acc4,12($out) + stb $acc5,13($out) + stb $acc6,14($out) + b L\$enc_done + stb $s3,15($out) + +L\$enc_out_aligned + stw $s0,0($out) + stw $s1,4($out) + stw $s2,8($out) + stw $s3,12($out) + +L\$enc_done + $POP `-$FRAME-$SAVED_RP`(%sp),%r2 ; standard epilogue + $POP `-$FRAME+1*$SIZE_T`(%sp),%r4 + $POP `-$FRAME+2*$SIZE_T`(%sp),%r5 + $POP `-$FRAME+3*$SIZE_T`(%sp),%r6 + $POP `-$FRAME+4*$SIZE_T`(%sp),%r7 + $POP `-$FRAME+5*$SIZE_T`(%sp),%r8 + $POP `-$FRAME+6*$SIZE_T`(%sp),%r9 + $POP `-$FRAME+7*$SIZE_T`(%sp),%r10 + $POP `-$FRAME+8*$SIZE_T`(%sp),%r11 + $POP `-$FRAME+9*$SIZE_T`(%sp),%r12 + $POP `-$FRAME+10*$SIZE_T`(%sp),%r13 + $POP `-$FRAME+11*$SIZE_T`(%sp),%r14 + $POP `-$FRAME+12*$SIZE_T`(%sp),%r15 + $POP `-$FRAME+13*$SIZE_T`(%sp),%r16 + $POP `-$FRAME+14*$SIZE_T`(%sp),%r17 + $POP `-$FRAME+15*$SIZE_T`(%sp),%r18 + bv (%r2) + .EXIT + $POPMB -$FRAME(%sp),%r3 + .PROCEND + + .ALIGN 16 +_parisc_AES_encrypt + .PROC + .CALLINFO MILLICODE + .ENTRY + ldw 240($key),$rounds + ldw 0($key),$t0 + ldw 4($key),$t1 + ldw 8($key),$t2 + _srm $rounds,1,$rounds + xor $t0,$s0,$s0 + ldw 12($key),$t3 + _srm $s0,24,$acc0 + xor $t1,$s1,$s1 + ldw 16($key),$t0 + _srm $s1,16,$acc1 + xor $t2,$s2,$s2 + ldw 20($key),$t1 + xor $t3,$s3,$s3 + ldw 24($key),$t2 + ldw 28($key),$t3 +L\$enc_loop + _srm $s2,8,$acc2 + ldwx,s $acc0($tbl),$acc0 + _srm $s3,0,$acc3 + ldwx,s $acc1($tbl),$acc1 + _srm $s1,24,$acc4 + ldwx,s $acc2($tbl),$acc2 + _srm $s2,16,$acc5 + ldwx,s $acc3($tbl),$acc3 + _srm $s3,8,$acc6 + ldwx,s $acc4($tbl),$acc4 + _srm $s0,0,$acc7 + ldwx,s $acc5($tbl),$acc5 + _srm $s2,24,$acc8 + ldwx,s $acc6($tbl),$acc6 + _srm $s3,16,$acc9 + ldwx,s $acc7($tbl),$acc7 + _srm $s0,8,$acc10 + ldwx,s $acc8($tbl),$acc8 + _srm $s1,0,$acc11 + ldwx,s $acc9($tbl),$acc9 + _srm $s3,24,$acc12 + ldwx,s $acc10($tbl),$acc10 + _srm $s0,16,$acc13 + ldwx,s $acc11($tbl),$acc11 + _srm $s1,8,$acc14 + ldwx,s $acc12($tbl),$acc12 + _srm $s2,0,$acc15 + ldwx,s $acc13($tbl),$acc13 + ldwx,s $acc14($tbl),$acc14 + ldwx,s $acc15($tbl),$acc15 + addib,= -1,$rounds,L\$enc_last + ldo 32($key),$key + + _ror $acc1,8,$acc1 + xor $acc0,$t0,$t0 + ldw 0($key),$s0 + _ror $acc2,16,$acc2 + xor $acc1,$t0,$t0 + ldw 4($key),$s1 + _ror $acc3,24,$acc3 + xor $acc2,$t0,$t0 + ldw 8($key),$s2 + _ror $acc5,8,$acc5 + xor $acc3,$t0,$t0 + ldw 12($key),$s3 + _ror $acc6,16,$acc6 + xor $acc4,$t1,$t1 + _ror $acc7,24,$acc7 + xor $acc5,$t1,$t1 + _ror $acc9,8,$acc9 + xor $acc6,$t1,$t1 + _ror $acc10,16,$acc10 + xor $acc7,$t1,$t1 + _ror $acc11,24,$acc11 + xor $acc8,$t2,$t2 + _ror $acc13,8,$acc13 + xor $acc9,$t2,$t2 + _ror $acc14,16,$acc14 + xor $acc10,$t2,$t2 + _ror $acc15,24,$acc15 + xor $acc11,$t2,$t2 + xor $acc12,$acc14,$acc14 + xor $acc13,$t3,$t3 + _srm $t0,24,$acc0 + xor $acc14,$t3,$t3 + _srm $t1,16,$acc1 + xor $acc15,$t3,$t3 + + _srm $t2,8,$acc2 + ldwx,s $acc0($tbl),$acc0 + _srm $t3,0,$acc3 + ldwx,s $acc1($tbl),$acc1 + _srm $t1,24,$acc4 + ldwx,s $acc2($tbl),$acc2 + _srm $t2,16,$acc5 + ldwx,s $acc3($tbl),$acc3 + _srm $t3,8,$acc6 + ldwx,s $acc4($tbl),$acc4 + _srm $t0,0,$acc7 + ldwx,s $acc5($tbl),$acc5 + _srm $t2,24,$acc8 + ldwx,s $acc6($tbl),$acc6 + _srm $t3,16,$acc9 + ldwx,s $acc7($tbl),$acc7 + _srm $t0,8,$acc10 + ldwx,s $acc8($tbl),$acc8 + _srm $t1,0,$acc11 + ldwx,s $acc9($tbl),$acc9 + _srm $t3,24,$acc12 + ldwx,s $acc10($tbl),$acc10 + _srm $t0,16,$acc13 + ldwx,s $acc11($tbl),$acc11 + _srm $t1,8,$acc14 + ldwx,s $acc12($tbl),$acc12 + _srm $t2,0,$acc15 + ldwx,s $acc13($tbl),$acc13 + _ror $acc1,8,$acc1 + ldwx,s $acc14($tbl),$acc14 + + _ror $acc2,16,$acc2 + xor $acc0,$s0,$s0 + ldwx,s $acc15($tbl),$acc15 + _ror $acc3,24,$acc3 + xor $acc1,$s0,$s0 + ldw 16($key),$t0 + _ror $acc5,8,$acc5 + xor $acc2,$s0,$s0 + ldw 20($key),$t1 + _ror $acc6,16,$acc6 + xor $acc3,$s0,$s0 + ldw 24($key),$t2 + _ror $acc7,24,$acc7 + xor $acc4,$s1,$s1 + ldw 28($key),$t3 + _ror $acc9,8,$acc9 + xor $acc5,$s1,$s1 + ldw 1024+0($tbl),%r0 ; prefetch te4 + _ror $acc10,16,$acc10 + xor $acc6,$s1,$s1 + ldw 1024+32($tbl),%r0 ; prefetch te4 + _ror $acc11,24,$acc11 + xor $acc7,$s1,$s1 + ldw 1024+64($tbl),%r0 ; prefetch te4 + _ror $acc13,8,$acc13 + xor $acc8,$s2,$s2 + ldw 1024+96($tbl),%r0 ; prefetch te4 + _ror $acc14,16,$acc14 + xor $acc9,$s2,$s2 + ldw 1024+128($tbl),%r0 ; prefetch te4 + _ror $acc15,24,$acc15 + xor $acc10,$s2,$s2 + ldw 1024+160($tbl),%r0 ; prefetch te4 + _srm $s0,24,$acc0 + xor $acc11,$s2,$s2 + ldw 1024+192($tbl),%r0 ; prefetch te4 + xor $acc12,$acc14,$acc14 + xor $acc13,$s3,$s3 + ldw 1024+224($tbl),%r0 ; prefetch te4 + _srm $s1,16,$acc1 + xor $acc14,$s3,$s3 + b L\$enc_loop + xor $acc15,$s3,$s3 + + .ALIGN 16 +L\$enc_last + ldo 1024($tbl),$rounds + _ror $acc1,8,$acc1 + xor $acc0,$t0,$t0 + ldw 0($key),$s0 + _ror $acc2,16,$acc2 + xor $acc1,$t0,$t0 + ldw 4($key),$s1 + _ror $acc3,24,$acc3 + xor $acc2,$t0,$t0 + ldw 8($key),$s2 + _ror $acc5,8,$acc5 + xor $acc3,$t0,$t0 + ldw 12($key),$s3 + _ror $acc6,16,$acc6 + xor $acc4,$t1,$t1 + _ror $acc7,24,$acc7 + xor $acc5,$t1,$t1 + _ror $acc9,8,$acc9 + xor $acc6,$t1,$t1 + _ror $acc10,16,$acc10 + xor $acc7,$t1,$t1 + _ror $acc11,24,$acc11 + xor $acc8,$t2,$t2 + _ror $acc13,8,$acc13 + xor $acc9,$t2,$t2 + _ror $acc14,16,$acc14 + xor $acc10,$t2,$t2 + _ror $acc15,24,$acc15 + xor $acc11,$t2,$t2 + xor $acc12,$acc14,$acc14 + xor $acc13,$t3,$t3 + _srm $t0,24,$acc0 + xor $acc14,$t3,$t3 + _srm $t1,16,$acc1 + xor $acc15,$t3,$t3 + + _srm $t2,8,$acc2 + ldbx $acc0($rounds),$acc0 + _srm $t1,24,$acc4 + ldbx $acc1($rounds),$acc1 + _srm $t2,16,$acc5 + _srm $t3,0,$acc3 + ldbx $acc2($rounds),$acc2 + ldbx $acc3($rounds),$acc3 + _srm $t3,8,$acc6 + ldbx $acc4($rounds),$acc4 + _srm $t2,24,$acc8 + ldbx $acc5($rounds),$acc5 + _srm $t3,16,$acc9 + _srm $t0,0,$acc7 + ldbx $acc6($rounds),$acc6 + ldbx $acc7($rounds),$acc7 + _srm $t0,8,$acc10 + ldbx $acc8($rounds),$acc8 + _srm $t3,24,$acc12 + ldbx $acc9($rounds),$acc9 + _srm $t0,16,$acc13 + _srm $t1,0,$acc11 + ldbx $acc10($rounds),$acc10 + _srm $t1,8,$acc14 + ldbx $acc11($rounds),$acc11 + ldbx $acc12($rounds),$acc12 + ldbx $acc13($rounds),$acc13 + _srm $t2,0,$acc15 + ldbx $acc14($rounds),$acc14 + + dep $acc0,7,8,$acc3 + ldbx $acc15($rounds),$acc15 + dep $acc4,7,8,$acc7 + dep $acc1,15,8,$acc3 + dep $acc5,15,8,$acc7 + dep $acc2,23,8,$acc3 + dep $acc6,23,8,$acc7 + xor $acc3,$s0,$s0 + xor $acc7,$s1,$s1 + dep $acc8,7,8,$acc11 + dep $acc12,7,8,$acc15 + dep $acc9,15,8,$acc11 + dep $acc13,15,8,$acc15 + dep $acc10,23,8,$acc11 + dep $acc14,23,8,$acc15 + xor $acc11,$s2,$s2 + + bv (%r31) + .EXIT + xor $acc15,$s3,$s3 + .PROCEND + + .ALIGN 64 +L\$AES_Te + .WORD 0xc66363a5, 0xf87c7c84, 0xee777799, 0xf67b7b8d + .WORD 0xfff2f20d, 0xd66b6bbd, 0xde6f6fb1, 0x91c5c554 + .WORD 0x60303050, 0x02010103, 0xce6767a9, 0x562b2b7d + .WORD 0xe7fefe19, 0xb5d7d762, 0x4dababe6, 0xec76769a + .WORD 0x8fcaca45, 0x1f82829d, 0x89c9c940, 0xfa7d7d87 + .WORD 0xeffafa15, 0xb25959eb, 0x8e4747c9, 0xfbf0f00b + .WORD 0x41adadec, 0xb3d4d467, 0x5fa2a2fd, 0x45afafea + .WORD 0x239c9cbf, 0x53a4a4f7, 0xe4727296, 0x9bc0c05b + .WORD 0x75b7b7c2, 0xe1fdfd1c, 0x3d9393ae, 0x4c26266a + .WORD 0x6c36365a, 0x7e3f3f41, 0xf5f7f702, 0x83cccc4f + .WORD 0x6834345c, 0x51a5a5f4, 0xd1e5e534, 0xf9f1f108 + .WORD 0xe2717193, 0xabd8d873, 0x62313153, 0x2a15153f + .WORD 0x0804040c, 0x95c7c752, 0x46232365, 0x9dc3c35e + .WORD 0x30181828, 0x379696a1, 0x0a05050f, 0x2f9a9ab5 + .WORD 0x0e070709, 0x24121236, 0x1b80809b, 0xdfe2e23d + .WORD 0xcdebeb26, 0x4e272769, 0x7fb2b2cd, 0xea75759f + .WORD 0x1209091b, 0x1d83839e, 0x582c2c74, 0x341a1a2e + .WORD 0x361b1b2d, 0xdc6e6eb2, 0xb45a5aee, 0x5ba0a0fb + .WORD 0xa45252f6, 0x763b3b4d, 0xb7d6d661, 0x7db3b3ce + .WORD 0x5229297b, 0xdde3e33e, 0x5e2f2f71, 0x13848497 + .WORD 0xa65353f5, 0xb9d1d168, 0x00000000, 0xc1eded2c + .WORD 0x40202060, 0xe3fcfc1f, 0x79b1b1c8, 0xb65b5bed + .WORD 0xd46a6abe, 0x8dcbcb46, 0x67bebed9, 0x7239394b + .WORD 0x944a4ade, 0x984c4cd4, 0xb05858e8, 0x85cfcf4a + .WORD 0xbbd0d06b, 0xc5efef2a, 0x4faaaae5, 0xedfbfb16 + .WORD 0x864343c5, 0x9a4d4dd7, 0x66333355, 0x11858594 + .WORD 0x8a4545cf, 0xe9f9f910, 0x04020206, 0xfe7f7f81 + .WORD 0xa05050f0, 0x783c3c44, 0x259f9fba, 0x4ba8a8e3 + .WORD 0xa25151f3, 0x5da3a3fe, 0x804040c0, 0x058f8f8a + .WORD 0x3f9292ad, 0x219d9dbc, 0x70383848, 0xf1f5f504 + .WORD 0x63bcbcdf, 0x77b6b6c1, 0xafdada75, 0x42212163 + .WORD 0x20101030, 0xe5ffff1a, 0xfdf3f30e, 0xbfd2d26d + .WORD 0x81cdcd4c, 0x180c0c14, 0x26131335, 0xc3ecec2f + .WORD 0xbe5f5fe1, 0x359797a2, 0x884444cc, 0x2e171739 + .WORD 0x93c4c457, 0x55a7a7f2, 0xfc7e7e82, 0x7a3d3d47 + .WORD 0xc86464ac, 0xba5d5de7, 0x3219192b, 0xe6737395 + .WORD 0xc06060a0, 0x19818198, 0x9e4f4fd1, 0xa3dcdc7f + .WORD 0x44222266, 0x542a2a7e, 0x3b9090ab, 0x0b888883 + .WORD 0x8c4646ca, 0xc7eeee29, 0x6bb8b8d3, 0x2814143c + .WORD 0xa7dede79, 0xbc5e5ee2, 0x160b0b1d, 0xaddbdb76 + .WORD 0xdbe0e03b, 0x64323256, 0x743a3a4e, 0x140a0a1e + .WORD 0x924949db, 0x0c06060a, 0x4824246c, 0xb85c5ce4 + .WORD 0x9fc2c25d, 0xbdd3d36e, 0x43acacef, 0xc46262a6 + .WORD 0x399191a8, 0x319595a4, 0xd3e4e437, 0xf279798b + .WORD 0xd5e7e732, 0x8bc8c843, 0x6e373759, 0xda6d6db7 + .WORD 0x018d8d8c, 0xb1d5d564, 0x9c4e4ed2, 0x49a9a9e0 + .WORD 0xd86c6cb4, 0xac5656fa, 0xf3f4f407, 0xcfeaea25 + .WORD 0xca6565af, 0xf47a7a8e, 0x47aeaee9, 0x10080818 + .WORD 0x6fbabad5, 0xf0787888, 0x4a25256f, 0x5c2e2e72 + .WORD 0x381c1c24, 0x57a6a6f1, 0x73b4b4c7, 0x97c6c651 + .WORD 0xcbe8e823, 0xa1dddd7c, 0xe874749c, 0x3e1f1f21 + .WORD 0x964b4bdd, 0x61bdbddc, 0x0d8b8b86, 0x0f8a8a85 + .WORD 0xe0707090, 0x7c3e3e42, 0x71b5b5c4, 0xcc6666aa + .WORD 0x904848d8, 0x06030305, 0xf7f6f601, 0x1c0e0e12 + .WORD 0xc26161a3, 0x6a35355f, 0xae5757f9, 0x69b9b9d0 + .WORD 0x17868691, 0x99c1c158, 0x3a1d1d27, 0x279e9eb9 + .WORD 0xd9e1e138, 0xebf8f813, 0x2b9898b3, 0x22111133 + .WORD 0xd26969bb, 0xa9d9d970, 0x078e8e89, 0x339494a7 + .WORD 0x2d9b9bb6, 0x3c1e1e22, 0x15878792, 0xc9e9e920 + .WORD 0x87cece49, 0xaa5555ff, 0x50282878, 0xa5dfdf7a + .WORD 0x038c8c8f, 0x59a1a1f8, 0x09898980, 0x1a0d0d17 + .WORD 0x65bfbfda, 0xd7e6e631, 0x844242c6, 0xd06868b8 + .WORD 0x824141c3, 0x299999b0, 0x5a2d2d77, 0x1e0f0f11 + .WORD 0x7bb0b0cb, 0xa85454fc, 0x6dbbbbd6, 0x2c16163a + .BYTE 0x63, 0x7c, 0x77, 0x7b, 0xf2, 0x6b, 0x6f, 0xc5 + .BYTE 0x30, 0x01, 0x67, 0x2b, 0xfe, 0xd7, 0xab, 0x76 + .BYTE 0xca, 0x82, 0xc9, 0x7d, 0xfa, 0x59, 0x47, 0xf0 + .BYTE 0xad, 0xd4, 0xa2, 0xaf, 0x9c, 0xa4, 0x72, 0xc0 + .BYTE 0xb7, 0xfd, 0x93, 0x26, 0x36, 0x3f, 0xf7, 0xcc + .BYTE 0x34, 0xa5, 0xe5, 0xf1, 0x71, 0xd8, 0x31, 0x15 + .BYTE 0x04, 0xc7, 0x23, 0xc3, 0x18, 0x96, 0x05, 0x9a + .BYTE 0x07, 0x12, 0x80, 0xe2, 0xeb, 0x27, 0xb2, 0x75 + .BYTE 0x09, 0x83, 0x2c, 0x1a, 0x1b, 0x6e, 0x5a, 0xa0 + .BYTE 0x52, 0x3b, 0xd6, 0xb3, 0x29, 0xe3, 0x2f, 0x84 + .BYTE 0x53, 0xd1, 0x00, 0xed, 0x20, 0xfc, 0xb1, 0x5b + .BYTE 0x6a, 0xcb, 0xbe, 0x39, 0x4a, 0x4c, 0x58, 0xcf + .BYTE 0xd0, 0xef, 0xaa, 0xfb, 0x43, 0x4d, 0x33, 0x85 + .BYTE 0x45, 0xf9, 0x02, 0x7f, 0x50, 0x3c, 0x9f, 0xa8 + .BYTE 0x51, 0xa3, 0x40, 0x8f, 0x92, 0x9d, 0x38, 0xf5 + .BYTE 0xbc, 0xb6, 0xda, 0x21, 0x10, 0xff, 0xf3, 0xd2 + .BYTE 0xcd, 0x0c, 0x13, 0xec, 0x5f, 0x97, 0x44, 0x17 + .BYTE 0xc4, 0xa7, 0x7e, 0x3d, 0x64, 0x5d, 0x19, 0x73 + .BYTE 0x60, 0x81, 0x4f, 0xdc, 0x22, 0x2a, 0x90, 0x88 + .BYTE 0x46, 0xee, 0xb8, 0x14, 0xde, 0x5e, 0x0b, 0xdb + .BYTE 0xe0, 0x32, 0x3a, 0x0a, 0x49, 0x06, 0x24, 0x5c + .BYTE 0xc2, 0xd3, 0xac, 0x62, 0x91, 0x95, 0xe4, 0x79 + .BYTE 0xe7, 0xc8, 0x37, 0x6d, 0x8d, 0xd5, 0x4e, 0xa9 + .BYTE 0x6c, 0x56, 0xf4, 0xea, 0x65, 0x7a, 0xae, 0x08 + .BYTE 0xba, 0x78, 0x25, 0x2e, 0x1c, 0xa6, 0xb4, 0xc6 + .BYTE 0xe8, 0xdd, 0x74, 0x1f, 0x4b, 0xbd, 0x8b, 0x8a + .BYTE 0x70, 0x3e, 0xb5, 0x66, 0x48, 0x03, 0xf6, 0x0e + .BYTE 0x61, 0x35, 0x57, 0xb9, 0x86, 0xc1, 0x1d, 0x9e + .BYTE 0xe1, 0xf8, 0x98, 0x11, 0x69, 0xd9, 0x8e, 0x94 + .BYTE 0x9b, 0x1e, 0x87, 0xe9, 0xce, 0x55, 0x28, 0xdf + .BYTE 0x8c, 0xa1, 0x89, 0x0d, 0xbf, 0xe6, 0x42, 0x68 + .BYTE 0x41, 0x99, 0x2d, 0x0f, 0xb0, 0x54, 0xbb, 0x16 +___ + +$code.=<<___; + .EXPORT AES_decrypt,ENTRY,ARGW0=GR,ARGW1=GR,ARGW2=GR + .ALIGN 16 +AES_decrypt + .PROC + .CALLINFO FRAME=`$FRAME-16*$SIZE_T`,NO_CALLS,SAVE_RP,ENTRY_GR=18 + .ENTRY + $PUSH %r2,-$SAVED_RP(%sp) ; standard prologue + $PUSHMA %r3,$FRAME(%sp) + $PUSH %r4,`-$FRAME+1*$SIZE_T`(%sp) + $PUSH %r5,`-$FRAME+2*$SIZE_T`(%sp) + $PUSH %r6,`-$FRAME+3*$SIZE_T`(%sp) + $PUSH %r7,`-$FRAME+4*$SIZE_T`(%sp) + $PUSH %r8,`-$FRAME+5*$SIZE_T`(%sp) + $PUSH %r9,`-$FRAME+6*$SIZE_T`(%sp) + $PUSH %r10,`-$FRAME+7*$SIZE_T`(%sp) + $PUSH %r11,`-$FRAME+8*$SIZE_T`(%sp) + $PUSH %r12,`-$FRAME+9*$SIZE_T`(%sp) + $PUSH %r13,`-$FRAME+10*$SIZE_T`(%sp) + $PUSH %r14,`-$FRAME+11*$SIZE_T`(%sp) + $PUSH %r15,`-$FRAME+12*$SIZE_T`(%sp) + $PUSH %r16,`-$FRAME+13*$SIZE_T`(%sp) + $PUSH %r17,`-$FRAME+14*$SIZE_T`(%sp) + $PUSH %r18,`-$FRAME+15*$SIZE_T`(%sp) + + blr %r0,$tbl + ldi 3,$t0 +L\$dec_pic + andcm $tbl,$t0,$tbl + ldo L\$AES_Td-L\$dec_pic($tbl),$tbl + + and $inp,$t0,$t0 + sub $inp,$t0,$inp + ldw 0($inp),$s0 + ldw 4($inp),$s1 + ldw 8($inp),$s2 + comib,= 0,$t0,L\$dec_inp_aligned + ldw 12($inp),$s3 + + sh3addl $t0,%r0,$t0 + subi 32,$t0,$t0 + mtctl $t0,%cr11 + ldw 16($inp),$t1 + vshd $s0,$s1,$s0 + vshd $s1,$s2,$s1 + vshd $s2,$s3,$s2 + vshd $s3,$t1,$s3 + +L\$dec_inp_aligned + bl _parisc_AES_decrypt,%r31 + nop + + extru,<> $out,31,2,%r0 + b L\$dec_out_aligned + nop + + _srm $s0,24,$acc0 + _srm $s0,16,$acc1 + stb $acc0,0($out) + _srm $s0,8,$acc2 + stb $acc1,1($out) + _srm $s1,24,$acc4 + stb $acc2,2($out) + _srm $s1,16,$acc5 + stb $s0,3($out) + _srm $s1,8,$acc6 + stb $acc4,4($out) + _srm $s2,24,$acc0 + stb $acc5,5($out) + _srm $s2,16,$acc1 + stb $acc6,6($out) + _srm $s2,8,$acc2 + stb $s1,7($out) + _srm $s3,24,$acc4 + stb $acc0,8($out) + _srm $s3,16,$acc5 + stb $acc1,9($out) + _srm $s3,8,$acc6 + stb $acc2,10($out) + stb $s2,11($out) + stb $acc4,12($out) + stb $acc5,13($out) + stb $acc6,14($out) + b L\$dec_done + stb $s3,15($out) + +L\$dec_out_aligned + stw $s0,0($out) + stw $s1,4($out) + stw $s2,8($out) + stw $s3,12($out) + +L\$dec_done + $POP `-$FRAME-$SAVED_RP`(%sp),%r2 ; standard epilogue + $POP `-$FRAME+1*$SIZE_T`(%sp),%r4 + $POP `-$FRAME+2*$SIZE_T`(%sp),%r5 + $POP `-$FRAME+3*$SIZE_T`(%sp),%r6 + $POP `-$FRAME+4*$SIZE_T`(%sp),%r7 + $POP `-$FRAME+5*$SIZE_T`(%sp),%r8 + $POP `-$FRAME+6*$SIZE_T`(%sp),%r9 + $POP `-$FRAME+7*$SIZE_T`(%sp),%r10 + $POP `-$FRAME+8*$SIZE_T`(%sp),%r11 + $POP `-$FRAME+9*$SIZE_T`(%sp),%r12 + $POP `-$FRAME+10*$SIZE_T`(%sp),%r13 + $POP `-$FRAME+11*$SIZE_T`(%sp),%r14 + $POP `-$FRAME+12*$SIZE_T`(%sp),%r15 + $POP `-$FRAME+13*$SIZE_T`(%sp),%r16 + $POP `-$FRAME+14*$SIZE_T`(%sp),%r17 + $POP `-$FRAME+15*$SIZE_T`(%sp),%r18 + bv (%r2) + .EXIT + $POPMB -$FRAME(%sp),%r3 + .PROCEND + + .ALIGN 16 +_parisc_AES_decrypt + .PROC + .CALLINFO MILLICODE + .ENTRY + ldw 240($key),$rounds + ldw 0($key),$t0 + ldw 4($key),$t1 + ldw 8($key),$t2 + ldw 12($key),$t3 + _srm $rounds,1,$rounds + xor $t0,$s0,$s0 + ldw 16($key),$t0 + xor $t1,$s1,$s1 + ldw 20($key),$t1 + _srm $s0,24,$acc0 + xor $t2,$s2,$s2 + ldw 24($key),$t2 + xor $t3,$s3,$s3 + ldw 28($key),$t3 + _srm $s3,16,$acc1 +L\$dec_loop + _srm $s2,8,$acc2 + ldwx,s $acc0($tbl),$acc0 + _srm $s1,0,$acc3 + ldwx,s $acc1($tbl),$acc1 + _srm $s1,24,$acc4 + ldwx,s $acc2($tbl),$acc2 + _srm $s0,16,$acc5 + ldwx,s $acc3($tbl),$acc3 + _srm $s3,8,$acc6 + ldwx,s $acc4($tbl),$acc4 + _srm $s2,0,$acc7 + ldwx,s $acc5($tbl),$acc5 + _srm $s2,24,$acc8 + ldwx,s $acc6($tbl),$acc6 + _srm $s1,16,$acc9 + ldwx,s $acc7($tbl),$acc7 + _srm $s0,8,$acc10 + ldwx,s $acc8($tbl),$acc8 + _srm $s3,0,$acc11 + ldwx,s $acc9($tbl),$acc9 + _srm $s3,24,$acc12 + ldwx,s $acc10($tbl),$acc10 + _srm $s2,16,$acc13 + ldwx,s $acc11($tbl),$acc11 + _srm $s1,8,$acc14 + ldwx,s $acc12($tbl),$acc12 + _srm $s0,0,$acc15 + ldwx,s $acc13($tbl),$acc13 + ldwx,s $acc14($tbl),$acc14 + ldwx,s $acc15($tbl),$acc15 + addib,= -1,$rounds,L\$dec_last + ldo 32($key),$key + + _ror $acc1,8,$acc1 + xor $acc0,$t0,$t0 + ldw 0($key),$s0 + _ror $acc2,16,$acc2 + xor $acc1,$t0,$t0 + ldw 4($key),$s1 + _ror $acc3,24,$acc3 + xor $acc2,$t0,$t0 + ldw 8($key),$s2 + _ror $acc5,8,$acc5 + xor $acc3,$t0,$t0 + ldw 12($key),$s3 + _ror $acc6,16,$acc6 + xor $acc4,$t1,$t1 + _ror $acc7,24,$acc7 + xor $acc5,$t1,$t1 + _ror $acc9,8,$acc9 + xor $acc6,$t1,$t1 + _ror $acc10,16,$acc10 + xor $acc7,$t1,$t1 + _ror $acc11,24,$acc11 + xor $acc8,$t2,$t2 + _ror $acc13,8,$acc13 + xor $acc9,$t2,$t2 + _ror $acc14,16,$acc14 + xor $acc10,$t2,$t2 + _ror $acc15,24,$acc15 + xor $acc11,$t2,$t2 + xor $acc12,$acc14,$acc14 + xor $acc13,$t3,$t3 + _srm $t0,24,$acc0 + xor $acc14,$t3,$t3 + xor $acc15,$t3,$t3 + _srm $t3,16,$acc1 + + _srm $t2,8,$acc2 + ldwx,s $acc0($tbl),$acc0 + _srm $t1,0,$acc3 + ldwx,s $acc1($tbl),$acc1 + _srm $t1,24,$acc4 + ldwx,s $acc2($tbl),$acc2 + _srm $t0,16,$acc5 + ldwx,s $acc3($tbl),$acc3 + _srm $t3,8,$acc6 + ldwx,s $acc4($tbl),$acc4 + _srm $t2,0,$acc7 + ldwx,s $acc5($tbl),$acc5 + _srm $t2,24,$acc8 + ldwx,s $acc6($tbl),$acc6 + _srm $t1,16,$acc9 + ldwx,s $acc7($tbl),$acc7 + _srm $t0,8,$acc10 + ldwx,s $acc8($tbl),$acc8 + _srm $t3,0,$acc11 + ldwx,s $acc9($tbl),$acc9 + _srm $t3,24,$acc12 + ldwx,s $acc10($tbl),$acc10 + _srm $t2,16,$acc13 + ldwx,s $acc11($tbl),$acc11 + _srm $t1,8,$acc14 + ldwx,s $acc12($tbl),$acc12 + _srm $t0,0,$acc15 + ldwx,s $acc13($tbl),$acc13 + _ror $acc1,8,$acc1 + ldwx,s $acc14($tbl),$acc14 + + _ror $acc2,16,$acc2 + xor $acc0,$s0,$s0 + ldwx,s $acc15($tbl),$acc15 + _ror $acc3,24,$acc3 + xor $acc1,$s0,$s0 + ldw 16($key),$t0 + _ror $acc5,8,$acc5 + xor $acc2,$s0,$s0 + ldw 20($key),$t1 + _ror $acc6,16,$acc6 + xor $acc3,$s0,$s0 + ldw 24($key),$t2 + _ror $acc7,24,$acc7 + xor $acc4,$s1,$s1 + ldw 28($key),$t3 + _ror $acc9,8,$acc9 + xor $acc5,$s1,$s1 + ldw 1024+0($tbl),%r0 ; prefetch td4 + _ror $acc10,16,$acc10 + xor $acc6,$s1,$s1 + ldw 1024+32($tbl),%r0 ; prefetch td4 + _ror $acc11,24,$acc11 + xor $acc7,$s1,$s1 + ldw 1024+64($tbl),%r0 ; prefetch td4 + _ror $acc13,8,$acc13 + xor $acc8,$s2,$s2 + ldw 1024+96($tbl),%r0 ; prefetch td4 + _ror $acc14,16,$acc14 + xor $acc9,$s2,$s2 + ldw 1024+128($tbl),%r0 ; prefetch td4 + _ror $acc15,24,$acc15 + xor $acc10,$s2,$s2 + ldw 1024+160($tbl),%r0 ; prefetch td4 + _srm $s0,24,$acc0 + xor $acc11,$s2,$s2 + ldw 1024+192($tbl),%r0 ; prefetch td4 + xor $acc12,$acc14,$acc14 + xor $acc13,$s3,$s3 + ldw 1024+224($tbl),%r0 ; prefetch td4 + xor $acc14,$s3,$s3 + xor $acc15,$s3,$s3 + b L\$dec_loop + _srm $s3,16,$acc1 + + .ALIGN 16 +L\$dec_last + ldo 1024($tbl),$rounds + _ror $acc1,8,$acc1 + xor $acc0,$t0,$t0 + ldw 0($key),$s0 + _ror $acc2,16,$acc2 + xor $acc1,$t0,$t0 + ldw 4($key),$s1 + _ror $acc3,24,$acc3 + xor $acc2,$t0,$t0 + ldw 8($key),$s2 + _ror $acc5,8,$acc5 + xor $acc3,$t0,$t0 + ldw 12($key),$s3 + _ror $acc6,16,$acc6 + xor $acc4,$t1,$t1 + _ror $acc7,24,$acc7 + xor $acc5,$t1,$t1 + _ror $acc9,8,$acc9 + xor $acc6,$t1,$t1 + _ror $acc10,16,$acc10 + xor $acc7,$t1,$t1 + _ror $acc11,24,$acc11 + xor $acc8,$t2,$t2 + _ror $acc13,8,$acc13 + xor $acc9,$t2,$t2 + _ror $acc14,16,$acc14 + xor $acc10,$t2,$t2 + _ror $acc15,24,$acc15 + xor $acc11,$t2,$t2 + xor $acc12,$acc14,$acc14 + xor $acc13,$t3,$t3 + _srm $t0,24,$acc0 + xor $acc14,$t3,$t3 + xor $acc15,$t3,$t3 + _srm $t3,16,$acc1 + + _srm $t2,8,$acc2 + ldbx $acc0($rounds),$acc0 + _srm $t1,24,$acc4 + ldbx $acc1($rounds),$acc1 + _srm $t0,16,$acc5 + _srm $t1,0,$acc3 + ldbx $acc2($rounds),$acc2 + ldbx $acc3($rounds),$acc3 + _srm $t3,8,$acc6 + ldbx $acc4($rounds),$acc4 + _srm $t2,24,$acc8 + ldbx $acc5($rounds),$acc5 + _srm $t1,16,$acc9 + _srm $t2,0,$acc7 + ldbx $acc6($rounds),$acc6 + ldbx $acc7($rounds),$acc7 + _srm $t0,8,$acc10 + ldbx $acc8($rounds),$acc8 + _srm $t3,24,$acc12 + ldbx $acc9($rounds),$acc9 + _srm $t2,16,$acc13 + _srm $t3,0,$acc11 + ldbx $acc10($rounds),$acc10 + _srm $t1,8,$acc14 + ldbx $acc11($rounds),$acc11 + ldbx $acc12($rounds),$acc12 + ldbx $acc13($rounds),$acc13 + _srm $t0,0,$acc15 + ldbx $acc14($rounds),$acc14 + + dep $acc0,7,8,$acc3 + ldbx $acc15($rounds),$acc15 + dep $acc4,7,8,$acc7 + dep $acc1,15,8,$acc3 + dep $acc5,15,8,$acc7 + dep $acc2,23,8,$acc3 + dep $acc6,23,8,$acc7 + xor $acc3,$s0,$s0 + xor $acc7,$s1,$s1 + dep $acc8,7,8,$acc11 + dep $acc12,7,8,$acc15 + dep $acc9,15,8,$acc11 + dep $acc13,15,8,$acc15 + dep $acc10,23,8,$acc11 + dep $acc14,23,8,$acc15 + xor $acc11,$s2,$s2 + + bv (%r31) + .EXIT + xor $acc15,$s3,$s3 + .PROCEND + + .ALIGN 64 +L\$AES_Td + .WORD 0x51f4a750, 0x7e416553, 0x1a17a4c3, 0x3a275e96 + .WORD 0x3bab6bcb, 0x1f9d45f1, 0xacfa58ab, 0x4be30393 + .WORD 0x2030fa55, 0xad766df6, 0x88cc7691, 0xf5024c25 + .WORD 0x4fe5d7fc, 0xc52acbd7, 0x26354480, 0xb562a38f + .WORD 0xdeb15a49, 0x25ba1b67, 0x45ea0e98, 0x5dfec0e1 + .WORD 0xc32f7502, 0x814cf012, 0x8d4697a3, 0x6bd3f9c6 + .WORD 0x038f5fe7, 0x15929c95, 0xbf6d7aeb, 0x955259da + .WORD 0xd4be832d, 0x587421d3, 0x49e06929, 0x8ec9c844 + .WORD 0x75c2896a, 0xf48e7978, 0x99583e6b, 0x27b971dd + .WORD 0xbee14fb6, 0xf088ad17, 0xc920ac66, 0x7dce3ab4 + .WORD 0x63df4a18, 0xe51a3182, 0x97513360, 0x62537f45 + .WORD 0xb16477e0, 0xbb6bae84, 0xfe81a01c, 0xf9082b94 + .WORD 0x70486858, 0x8f45fd19, 0x94de6c87, 0x527bf8b7 + .WORD 0xab73d323, 0x724b02e2, 0xe31f8f57, 0x6655ab2a + .WORD 0xb2eb2807, 0x2fb5c203, 0x86c57b9a, 0xd33708a5 + .WORD 0x302887f2, 0x23bfa5b2, 0x02036aba, 0xed16825c + .WORD 0x8acf1c2b, 0xa779b492, 0xf307f2f0, 0x4e69e2a1 + .WORD 0x65daf4cd, 0x0605bed5, 0xd134621f, 0xc4a6fe8a + .WORD 0x342e539d, 0xa2f355a0, 0x058ae132, 0xa4f6eb75 + .WORD 0x0b83ec39, 0x4060efaa, 0x5e719f06, 0xbd6e1051 + .WORD 0x3e218af9, 0x96dd063d, 0xdd3e05ae, 0x4de6bd46 + .WORD 0x91548db5, 0x71c45d05, 0x0406d46f, 0x605015ff + .WORD 0x1998fb24, 0xd6bde997, 0x894043cc, 0x67d99e77 + .WORD 0xb0e842bd, 0x07898b88, 0xe7195b38, 0x79c8eedb + .WORD 0xa17c0a47, 0x7c420fe9, 0xf8841ec9, 0x00000000 + .WORD 0x09808683, 0x322bed48, 0x1e1170ac, 0x6c5a724e + .WORD 0xfd0efffb, 0x0f853856, 0x3daed51e, 0x362d3927 + .WORD 0x0a0fd964, 0x685ca621, 0x9b5b54d1, 0x24362e3a + .WORD 0x0c0a67b1, 0x9357e70f, 0xb4ee96d2, 0x1b9b919e + .WORD 0x80c0c54f, 0x61dc20a2, 0x5a774b69, 0x1c121a16 + .WORD 0xe293ba0a, 0xc0a02ae5, 0x3c22e043, 0x121b171d + .WORD 0x0e090d0b, 0xf28bc7ad, 0x2db6a8b9, 0x141ea9c8 + .WORD 0x57f11985, 0xaf75074c, 0xee99ddbb, 0xa37f60fd + .WORD 0xf701269f, 0x5c72f5bc, 0x44663bc5, 0x5bfb7e34 + .WORD 0x8b432976, 0xcb23c6dc, 0xb6edfc68, 0xb8e4f163 + .WORD 0xd731dcca, 0x42638510, 0x13972240, 0x84c61120 + .WORD 0x854a247d, 0xd2bb3df8, 0xaef93211, 0xc729a16d + .WORD 0x1d9e2f4b, 0xdcb230f3, 0x0d8652ec, 0x77c1e3d0 + .WORD 0x2bb3166c, 0xa970b999, 0x119448fa, 0x47e96422 + .WORD 0xa8fc8cc4, 0xa0f03f1a, 0x567d2cd8, 0x223390ef + .WORD 0x87494ec7, 0xd938d1c1, 0x8ccaa2fe, 0x98d40b36 + .WORD 0xa6f581cf, 0xa57ade28, 0xdab78e26, 0x3fadbfa4 + .WORD 0x2c3a9de4, 0x5078920d, 0x6a5fcc9b, 0x547e4662 + .WORD 0xf68d13c2, 0x90d8b8e8, 0x2e39f75e, 0x82c3aff5 + .WORD 0x9f5d80be, 0x69d0937c, 0x6fd52da9, 0xcf2512b3 + .WORD 0xc8ac993b, 0x10187da7, 0xe89c636e, 0xdb3bbb7b + .WORD 0xcd267809, 0x6e5918f4, 0xec9ab701, 0x834f9aa8 + .WORD 0xe6956e65, 0xaaffe67e, 0x21bccf08, 0xef15e8e6 + .WORD 0xbae79bd9, 0x4a6f36ce, 0xea9f09d4, 0x29b07cd6 + .WORD 0x31a4b2af, 0x2a3f2331, 0xc6a59430, 0x35a266c0 + .WORD 0x744ebc37, 0xfc82caa6, 0xe090d0b0, 0x33a7d815 + .WORD 0xf104984a, 0x41ecdaf7, 0x7fcd500e, 0x1791f62f + .WORD 0x764dd68d, 0x43efb04d, 0xccaa4d54, 0xe49604df + .WORD 0x9ed1b5e3, 0x4c6a881b, 0xc12c1fb8, 0x4665517f + .WORD 0x9d5eea04, 0x018c355d, 0xfa877473, 0xfb0b412e + .WORD 0xb3671d5a, 0x92dbd252, 0xe9105633, 0x6dd64713 + .WORD 0x9ad7618c, 0x37a10c7a, 0x59f8148e, 0xeb133c89 + .WORD 0xcea927ee, 0xb761c935, 0xe11ce5ed, 0x7a47b13c + .WORD 0x9cd2df59, 0x55f2733f, 0x1814ce79, 0x73c737bf + .WORD 0x53f7cdea, 0x5ffdaa5b, 0xdf3d6f14, 0x7844db86 + .WORD 0xcaaff381, 0xb968c43e, 0x3824342c, 0xc2a3405f + .WORD 0x161dc372, 0xbce2250c, 0x283c498b, 0xff0d9541 + .WORD 0x39a80171, 0x080cb3de, 0xd8b4e49c, 0x6456c190 + .WORD 0x7bcb8461, 0xd532b670, 0x486c5c74, 0xd0b85742 + .BYTE 0x52, 0x09, 0x6a, 0xd5, 0x30, 0x36, 0xa5, 0x38 + .BYTE 0xbf, 0x40, 0xa3, 0x9e, 0x81, 0xf3, 0xd7, 0xfb + .BYTE 0x7c, 0xe3, 0x39, 0x82, 0x9b, 0x2f, 0xff, 0x87 + .BYTE 0x34, 0x8e, 0x43, 0x44, 0xc4, 0xde, 0xe9, 0xcb + .BYTE 0x54, 0x7b, 0x94, 0x32, 0xa6, 0xc2, 0x23, 0x3d + .BYTE 0xee, 0x4c, 0x95, 0x0b, 0x42, 0xfa, 0xc3, 0x4e + .BYTE 0x08, 0x2e, 0xa1, 0x66, 0x28, 0xd9, 0x24, 0xb2 + .BYTE 0x76, 0x5b, 0xa2, 0x49, 0x6d, 0x8b, 0xd1, 0x25 + .BYTE 0x72, 0xf8, 0xf6, 0x64, 0x86, 0x68, 0x98, 0x16 + .BYTE 0xd4, 0xa4, 0x5c, 0xcc, 0x5d, 0x65, 0xb6, 0x92 + .BYTE 0x6c, 0x70, 0x48, 0x50, 0xfd, 0xed, 0xb9, 0xda + .BYTE 0x5e, 0x15, 0x46, 0x57, 0xa7, 0x8d, 0x9d, 0x84 + .BYTE 0x90, 0xd8, 0xab, 0x00, 0x8c, 0xbc, 0xd3, 0x0a + .BYTE 0xf7, 0xe4, 0x58, 0x05, 0xb8, 0xb3, 0x45, 0x06 + .BYTE 0xd0, 0x2c, 0x1e, 0x8f, 0xca, 0x3f, 0x0f, 0x02 + .BYTE 0xc1, 0xaf, 0xbd, 0x03, 0x01, 0x13, 0x8a, 0x6b + .BYTE 0x3a, 0x91, 0x11, 0x41, 0x4f, 0x67, 0xdc, 0xea + .BYTE 0x97, 0xf2, 0xcf, 0xce, 0xf0, 0xb4, 0xe6, 0x73 + .BYTE 0x96, 0xac, 0x74, 0x22, 0xe7, 0xad, 0x35, 0x85 + .BYTE 0xe2, 0xf9, 0x37, 0xe8, 0x1c, 0x75, 0xdf, 0x6e + .BYTE 0x47, 0xf1, 0x1a, 0x71, 0x1d, 0x29, 0xc5, 0x89 + .BYTE 0x6f, 0xb7, 0x62, 0x0e, 0xaa, 0x18, 0xbe, 0x1b + .BYTE 0xfc, 0x56, 0x3e, 0x4b, 0xc6, 0xd2, 0x79, 0x20 + .BYTE 0x9a, 0xdb, 0xc0, 0xfe, 0x78, 0xcd, 0x5a, 0xf4 + .BYTE 0x1f, 0xdd, 0xa8, 0x33, 0x88, 0x07, 0xc7, 0x31 + .BYTE 0xb1, 0x12, 0x10, 0x59, 0x27, 0x80, 0xec, 0x5f + .BYTE 0x60, 0x51, 0x7f, 0xa9, 0x19, 0xb5, 0x4a, 0x0d + .BYTE 0x2d, 0xe5, 0x7a, 0x9f, 0x93, 0xc9, 0x9c, 0xef + .BYTE 0xa0, 0xe0, 0x3b, 0x4d, 0xae, 0x2a, 0xf5, 0xb0 + .BYTE 0xc8, 0xeb, 0xbb, 0x3c, 0x83, 0x53, 0x99, 0x61 + .BYTE 0x17, 0x2b, 0x04, 0x7e, 0xba, 0x77, 0xd6, 0x26 + .BYTE 0xe1, 0x69, 0x14, 0x63, 0x55, 0x21, 0x0c, 0x7d + .STRINGZ "AES for PA-RISC, CRYPTOGAMS by " +___ + +foreach (split("\n",$code)) { + s/\`([^\`]*)\`/eval $1/ge; + + # translate made up instructons: _ror, _srm + s/_ror(\s+)(%r[0-9]+),/shd$1$2,$2,/ or + + s/_srm(\s+%r[0-9]+),([0-9]+),/ + $SIZE_T==4 ? sprintf("extru%s,%d,8,",$1,31-$2) + : sprintf("extrd,u%s,%d,8,",$1,63-$2)/e; + + s/,\*/,/ if ($SIZE_T==4); + s/\bbv\b(.*\(%r2\))/bve$1/ if ($SIZE_T==8); + print $_,"\n"; +} +close STDOUT; diff --git a/openssl-1.1.0h/crypto/aes/asm/aes-ppc.pl b/openssl-1.1.0h/crypto/aes/asm/aes-ppc.pl new file mode 100644 index 0000000..1558d8e --- /dev/null +++ b/openssl-1.1.0h/crypto/aes/asm/aes-ppc.pl @@ -0,0 +1,1459 @@ +#! /usr/bin/env perl +# Copyright 2007-2016 The OpenSSL Project Authors. All Rights Reserved. +# +# Licensed under the OpenSSL license (the "License"). You may not use +# this file except in compliance with the License. You can obtain a copy +# in the file LICENSE in the source distribution or at +# https://www.openssl.org/source/license.html + + +# ==================================================================== +# Written by Andy Polyakov for the OpenSSL +# project. The module is, however, dual licensed under OpenSSL and +# CRYPTOGAMS licenses depending on where you obtain it. For further +# details see http://www.openssl.org/~appro/cryptogams/. +# ==================================================================== + +# Needs more work: key setup, CBC routine... +# +# ppc_AES_[en|de]crypt perform at 18 cycles per byte processed with +# 128-bit key, which is ~40% better than 64-bit code generated by gcc +# 4.0. But these are not the ones currently used! Their "compact" +# counterparts are, for security reason. ppc_AES_encrypt_compact runs +# at 1/2 of ppc_AES_encrypt speed, while ppc_AES_decrypt_compact - +# at 1/3 of ppc_AES_decrypt. + +# February 2010 +# +# Rescheduling instructions to favour Power6 pipeline gave 10% +# performance improvement on the platform in question (and marginal +# improvement even on others). It should be noted that Power6 fails +# to process byte in 18 cycles, only in 23, because it fails to issue +# 4 load instructions in two cycles, only in 3. As result non-compact +# block subroutines are 25% slower than one would expect. Compact +# functions scale better, because they have pure computational part, +# which scales perfectly with clock frequency. To be specific +# ppc_AES_encrypt_compact operates at 42 cycles per byte, while +# ppc_AES_decrypt_compact - at 55 (in 64-bit build). + +$flavour = shift; + +if ($flavour =~ /64/) { + $SIZE_T =8; + $LRSAVE =2*$SIZE_T; + $STU ="stdu"; + $POP ="ld"; + $PUSH ="std"; +} elsif ($flavour =~ /32/) { + $SIZE_T =4; + $LRSAVE =$SIZE_T; + $STU ="stwu"; + $POP ="lwz"; + $PUSH ="stw"; +} else { die "nonsense $flavour"; } + +$LITTLE_ENDIAN = ($flavour=~/le$/) ? $SIZE_T : 0; + +$0 =~ m/(.*[\/\\])[^\/\\]+$/; $dir=$1; +( $xlate="${dir}ppc-xlate.pl" and -f $xlate ) or +( $xlate="${dir}../../perlasm/ppc-xlate.pl" and -f $xlate) or +die "can't locate ppc-xlate.pl"; + +open STDOUT,"| $^X $xlate $flavour ".shift || die "can't call $xlate: $!"; + +$FRAME=32*$SIZE_T; + +sub _data_word() +{ my $i; + while(defined($i=shift)) { $code.=sprintf"\t.long\t0x%08x,0x%08x\n",$i,$i; } +} + +$sp="r1"; +$toc="r2"; +$inp="r3"; +$out="r4"; +$key="r5"; + +$Tbl0="r3"; +$Tbl1="r6"; +$Tbl2="r7"; +$Tbl3=$out; # stay away from "r2"; $out is offloaded to stack + +$s0="r8"; +$s1="r9"; +$s2="r10"; +$s3="r11"; + +$t0="r12"; +$t1="r0"; # stay away from "r13"; +$t2="r14"; +$t3="r15"; + +$acc00="r16"; +$acc01="r17"; +$acc02="r18"; +$acc03="r19"; + +$acc04="r20"; +$acc05="r21"; +$acc06="r22"; +$acc07="r23"; + +$acc08="r24"; +$acc09="r25"; +$acc10="r26"; +$acc11="r27"; + +$acc12="r28"; +$acc13="r29"; +$acc14="r30"; +$acc15="r31"; + +$mask80=$Tbl2; +$mask1b=$Tbl3; + +$code.=<<___; +.machine "any" +.text + +.align 7 +LAES_Te: + mflr r0 + bcl 20,31,\$+4 + mflr $Tbl0 ; vvvvv "distance" between . and 1st data entry + addi $Tbl0,$Tbl0,`128-8` + mtlr r0 + blr + .long 0 + .byte 0,12,0x14,0,0,0,0,0 + .space `64-9*4` +LAES_Td: + mflr r0 + bcl 20,31,\$+4 + mflr $Tbl0 ; vvvvvvvv "distance" between . and 1st data entry + addi $Tbl0,$Tbl0,`128-64-8+2048+256` + mtlr r0 + blr + .long 0 + .byte 0,12,0x14,0,0,0,0,0 + .space `128-64-9*4` +___ +&_data_word( + 0xc66363a5, 0xf87c7c84, 0xee777799, 0xf67b7b8d, + 0xfff2f20d, 0xd66b6bbd, 0xde6f6fb1, 0x91c5c554, + 0x60303050, 0x02010103, 0xce6767a9, 0x562b2b7d, + 0xe7fefe19, 0xb5d7d762, 0x4dababe6, 0xec76769a, + 0x8fcaca45, 0x1f82829d, 0x89c9c940, 0xfa7d7d87, + 0xeffafa15, 0xb25959eb, 0x8e4747c9, 0xfbf0f00b, + 0x41adadec, 0xb3d4d467, 0x5fa2a2fd, 0x45afafea, + 0x239c9cbf, 0x53a4a4f7, 0xe4727296, 0x9bc0c05b, + 0x75b7b7c2, 0xe1fdfd1c, 0x3d9393ae, 0x4c26266a, + 0x6c36365a, 0x7e3f3f41, 0xf5f7f702, 0x83cccc4f, + 0x6834345c, 0x51a5a5f4, 0xd1e5e534, 0xf9f1f108, + 0xe2717193, 0xabd8d873, 0x62313153, 0x2a15153f, + 0x0804040c, 0x95c7c752, 0x46232365, 0x9dc3c35e, + 0x30181828, 0x379696a1, 0x0a05050f, 0x2f9a9ab5, + 0x0e070709, 0x24121236, 0x1b80809b, 0xdfe2e23d, + 0xcdebeb26, 0x4e272769, 0x7fb2b2cd, 0xea75759f, + 0x1209091b, 0x1d83839e, 0x582c2c74, 0x341a1a2e, + 0x361b1b2d, 0xdc6e6eb2, 0xb45a5aee, 0x5ba0a0fb, + 0xa45252f6, 0x763b3b4d, 0xb7d6d661, 0x7db3b3ce, + 0x5229297b, 0xdde3e33e, 0x5e2f2f71, 0x13848497, + 0xa65353f5, 0xb9d1d168, 0x00000000, 0xc1eded2c, + 0x40202060, 0xe3fcfc1f, 0x79b1b1c8, 0xb65b5bed, + 0xd46a6abe, 0x8dcbcb46, 0x67bebed9, 0x7239394b, + 0x944a4ade, 0x984c4cd4, 0xb05858e8, 0x85cfcf4a, + 0xbbd0d06b, 0xc5efef2a, 0x4faaaae5, 0xedfbfb16, + 0x864343c5, 0x9a4d4dd7, 0x66333355, 0x11858594, + 0x8a4545cf, 0xe9f9f910, 0x04020206, 0xfe7f7f81, + 0xa05050f0, 0x783c3c44, 0x259f9fba, 0x4ba8a8e3, + 0xa25151f3, 0x5da3a3fe, 0x804040c0, 0x058f8f8a, + 0x3f9292ad, 0x219d9dbc, 0x70383848, 0xf1f5f504, + 0x63bcbcdf, 0x77b6b6c1, 0xafdada75, 0x42212163, + 0x20101030, 0xe5ffff1a, 0xfdf3f30e, 0xbfd2d26d, + 0x81cdcd4c, 0x180c0c14, 0x26131335, 0xc3ecec2f, + 0xbe5f5fe1, 0x359797a2, 0x884444cc, 0x2e171739, + 0x93c4c457, 0x55a7a7f2, 0xfc7e7e82, 0x7a3d3d47, + 0xc86464ac, 0xba5d5de7, 0x3219192b, 0xe6737395, + 0xc06060a0, 0x19818198, 0x9e4f4fd1, 0xa3dcdc7f, + 0x44222266, 0x542a2a7e, 0x3b9090ab, 0x0b888883, + 0x8c4646ca, 0xc7eeee29, 0x6bb8b8d3, 0x2814143c, + 0xa7dede79, 0xbc5e5ee2, 0x160b0b1d, 0xaddbdb76, + 0xdbe0e03b, 0x64323256, 0x743a3a4e, 0x140a0a1e, + 0x924949db, 0x0c06060a, 0x4824246c, 0xb85c5ce4, + 0x9fc2c25d, 0xbdd3d36e, 0x43acacef, 0xc46262a6, + 0x399191a8, 0x319595a4, 0xd3e4e437, 0xf279798b, + 0xd5e7e732, 0x8bc8c843, 0x6e373759, 0xda6d6db7, + 0x018d8d8c, 0xb1d5d564, 0x9c4e4ed2, 0x49a9a9e0, + 0xd86c6cb4, 0xac5656fa, 0xf3f4f407, 0xcfeaea25, + 0xca6565af, 0xf47a7a8e, 0x47aeaee9, 0x10080818, + 0x6fbabad5, 0xf0787888, 0x4a25256f, 0x5c2e2e72, + 0x381c1c24, 0x57a6a6f1, 0x73b4b4c7, 0x97c6c651, + 0xcbe8e823, 0xa1dddd7c, 0xe874749c, 0x3e1f1f21, + 0x964b4bdd, 0x61bdbddc, 0x0d8b8b86, 0x0f8a8a85, + 0xe0707090, 0x7c3e3e42, 0x71b5b5c4, 0xcc6666aa, + 0x904848d8, 0x06030305, 0xf7f6f601, 0x1c0e0e12, + 0xc26161a3, 0x6a35355f, 0xae5757f9, 0x69b9b9d0, + 0x17868691, 0x99c1c158, 0x3a1d1d27, 0x279e9eb9, + 0xd9e1e138, 0xebf8f813, 0x2b9898b3, 0x22111133, + 0xd26969bb, 0xa9d9d970, 0x078e8e89, 0x339494a7, + 0x2d9b9bb6, 0x3c1e1e22, 0x15878792, 0xc9e9e920, + 0x87cece49, 0xaa5555ff, 0x50282878, 0xa5dfdf7a, + 0x038c8c8f, 0x59a1a1f8, 0x09898980, 0x1a0d0d17, + 0x65bfbfda, 0xd7e6e631, 0x844242c6, 0xd06868b8, + 0x824141c3, 0x299999b0, 0x5a2d2d77, 0x1e0f0f11, + 0x7bb0b0cb, 0xa85454fc, 0x6dbbbbd6, 0x2c16163a); +$code.=<<___; +.byte 0x63, 0x7c, 0x77, 0x7b, 0xf2, 0x6b, 0x6f, 0xc5 +.byte 0x30, 0x01, 0x67, 0x2b, 0xfe, 0xd7, 0xab, 0x76 +.byte 0xca, 0x82, 0xc9, 0x7d, 0xfa, 0x59, 0x47, 0xf0 +.byte 0xad, 0xd4, 0xa2, 0xaf, 0x9c, 0xa4, 0x72, 0xc0 +.byte 0xb7, 0xfd, 0x93, 0x26, 0x36, 0x3f, 0xf7, 0xcc +.byte 0x34, 0xa5, 0xe5, 0xf1, 0x71, 0xd8, 0x31, 0x15 +.byte 0x04, 0xc7, 0x23, 0xc3, 0x18, 0x96, 0x05, 0x9a +.byte 0x07, 0x12, 0x80, 0xe2, 0xeb, 0x27, 0xb2, 0x75 +.byte 0x09, 0x83, 0x2c, 0x1a, 0x1b, 0x6e, 0x5a, 0xa0 +.byte 0x52, 0x3b, 0xd6, 0xb3, 0x29, 0xe3, 0x2f, 0x84 +.byte 0x53, 0xd1, 0x00, 0xed, 0x20, 0xfc, 0xb1, 0x5b +.byte 0x6a, 0xcb, 0xbe, 0x39, 0x4a, 0x4c, 0x58, 0xcf +.byte 0xd0, 0xef, 0xaa, 0xfb, 0x43, 0x4d, 0x33, 0x85 +.byte 0x45, 0xf9, 0x02, 0x7f, 0x50, 0x3c, 0x9f, 0xa8 +.byte 0x51, 0xa3, 0x40, 0x8f, 0x92, 0x9d, 0x38, 0xf5 +.byte 0xbc, 0xb6, 0xda, 0x21, 0x10, 0xff, 0xf3, 0xd2 +.byte 0xcd, 0x0c, 0x13, 0xec, 0x5f, 0x97, 0x44, 0x17 +.byte 0xc4, 0xa7, 0x7e, 0x3d, 0x64, 0x5d, 0x19, 0x73 +.byte 0x60, 0x81, 0x4f, 0xdc, 0x22, 0x2a, 0x90, 0x88 +.byte 0x46, 0xee, 0xb8, 0x14, 0xde, 0x5e, 0x0b, 0xdb +.byte 0xe0, 0x32, 0x3a, 0x0a, 0x49, 0x06, 0x24, 0x5c +.byte 0xc2, 0xd3, 0xac, 0x62, 0x91, 0x95, 0xe4, 0x79 +.byte 0xe7, 0xc8, 0x37, 0x6d, 0x8d, 0xd5, 0x4e, 0xa9 +.byte 0x6c, 0x56, 0xf4, 0xea, 0x65, 0x7a, 0xae, 0x08 +.byte 0xba, 0x78, 0x25, 0x2e, 0x1c, 0xa6, 0xb4, 0xc6 +.byte 0xe8, 0xdd, 0x74, 0x1f, 0x4b, 0xbd, 0x8b, 0x8a +.byte 0x70, 0x3e, 0xb5, 0x66, 0x48, 0x03, 0xf6, 0x0e +.byte 0x61, 0x35, 0x57, 0xb9, 0x86, 0xc1, 0x1d, 0x9e +.byte 0xe1, 0xf8, 0x98, 0x11, 0x69, 0xd9, 0x8e, 0x94 +.byte 0x9b, 0x1e, 0x87, 0xe9, 0xce, 0x55, 0x28, 0xdf +.byte 0x8c, 0xa1, 0x89, 0x0d, 0xbf, 0xe6, 0x42, 0x68 +.byte 0x41, 0x99, 0x2d, 0x0f, 0xb0, 0x54, 0xbb, 0x16 +___ +&_data_word( + 0x51f4a750, 0x7e416553, 0x1a17a4c3, 0x3a275e96, + 0x3bab6bcb, 0x1f9d45f1, 0xacfa58ab, 0x4be30393, + 0x2030fa55, 0xad766df6, 0x88cc7691, 0xf5024c25, + 0x4fe5d7fc, 0xc52acbd7, 0x26354480, 0xb562a38f, + 0xdeb15a49, 0x25ba1b67, 0x45ea0e98, 0x5dfec0e1, + 0xc32f7502, 0x814cf012, 0x8d4697a3, 0x6bd3f9c6, + 0x038f5fe7, 0x15929c95, 0xbf6d7aeb, 0x955259da, + 0xd4be832d, 0x587421d3, 0x49e06929, 0x8ec9c844, + 0x75c2896a, 0xf48e7978, 0x99583e6b, 0x27b971dd, + 0xbee14fb6, 0xf088ad17, 0xc920ac66, 0x7dce3ab4, + 0x63df4a18, 0xe51a3182, 0x97513360, 0x62537f45, + 0xb16477e0, 0xbb6bae84, 0xfe81a01c, 0xf9082b94, + 0x70486858, 0x8f45fd19, 0x94de6c87, 0x527bf8b7, + 0xab73d323, 0x724b02e2, 0xe31f8f57, 0x6655ab2a, + 0xb2eb2807, 0x2fb5c203, 0x86c57b9a, 0xd33708a5, + 0x302887f2, 0x23bfa5b2, 0x02036aba, 0xed16825c, + 0x8acf1c2b, 0xa779b492, 0xf307f2f0, 0x4e69e2a1, + 0x65daf4cd, 0x0605bed5, 0xd134621f, 0xc4a6fe8a, + 0x342e539d, 0xa2f355a0, 0x058ae132, 0xa4f6eb75, + 0x0b83ec39, 0x4060efaa, 0x5e719f06, 0xbd6e1051, + 0x3e218af9, 0x96dd063d, 0xdd3e05ae, 0x4de6bd46, + 0x91548db5, 0x71c45d05, 0x0406d46f, 0x605015ff, + 0x1998fb24, 0xd6bde997, 0x894043cc, 0x67d99e77, + 0xb0e842bd, 0x07898b88, 0xe7195b38, 0x79c8eedb, + 0xa17c0a47, 0x7c420fe9, 0xf8841ec9, 0x00000000, + 0x09808683, 0x322bed48, 0x1e1170ac, 0x6c5a724e, + 0xfd0efffb, 0x0f853856, 0x3daed51e, 0x362d3927, + 0x0a0fd964, 0x685ca621, 0x9b5b54d1, 0x24362e3a, + 0x0c0a67b1, 0x9357e70f, 0xb4ee96d2, 0x1b9b919e, + 0x80c0c54f, 0x61dc20a2, 0x5a774b69, 0x1c121a16, + 0xe293ba0a, 0xc0a02ae5, 0x3c22e043, 0x121b171d, + 0x0e090d0b, 0xf28bc7ad, 0x2db6a8b9, 0x141ea9c8, + 0x57f11985, 0xaf75074c, 0xee99ddbb, 0xa37f60fd, + 0xf701269f, 0x5c72f5bc, 0x44663bc5, 0x5bfb7e34, + 0x8b432976, 0xcb23c6dc, 0xb6edfc68, 0xb8e4f163, + 0xd731dcca, 0x42638510, 0x13972240, 0x84c61120, + 0x854a247d, 0xd2bb3df8, 0xaef93211, 0xc729a16d, + 0x1d9e2f4b, 0xdcb230f3, 0x0d8652ec, 0x77c1e3d0, + 0x2bb3166c, 0xa970b999, 0x119448fa, 0x47e96422, + 0xa8fc8cc4, 0xa0f03f1a, 0x567d2cd8, 0x223390ef, + 0x87494ec7, 0xd938d1c1, 0x8ccaa2fe, 0x98d40b36, + 0xa6f581cf, 0xa57ade28, 0xdab78e26, 0x3fadbfa4, + 0x2c3a9de4, 0x5078920d, 0x6a5fcc9b, 0x547e4662, + 0xf68d13c2, 0x90d8b8e8, 0x2e39f75e, 0x82c3aff5, + 0x9f5d80be, 0x69d0937c, 0x6fd52da9, 0xcf2512b3, + 0xc8ac993b, 0x10187da7, 0xe89c636e, 0xdb3bbb7b, + 0xcd267809, 0x6e5918f4, 0xec9ab701, 0x834f9aa8, + 0xe6956e65, 0xaaffe67e, 0x21bccf08, 0xef15e8e6, + 0xbae79bd9, 0x4a6f36ce, 0xea9f09d4, 0x29b07cd6, + 0x31a4b2af, 0x2a3f2331, 0xc6a59430, 0x35a266c0, + 0x744ebc37, 0xfc82caa6, 0xe090d0b0, 0x33a7d815, + 0xf104984a, 0x41ecdaf7, 0x7fcd500e, 0x1791f62f, + 0x764dd68d, 0x43efb04d, 0xccaa4d54, 0xe49604df, + 0x9ed1b5e3, 0x4c6a881b, 0xc12c1fb8, 0x4665517f, + 0x9d5eea04, 0x018c355d, 0xfa877473, 0xfb0b412e, + 0xb3671d5a, 0x92dbd252, 0xe9105633, 0x6dd64713, + 0x9ad7618c, 0x37a10c7a, 0x59f8148e, 0xeb133c89, + 0xcea927ee, 0xb761c935, 0xe11ce5ed, 0x7a47b13c, + 0x9cd2df59, 0x55f2733f, 0x1814ce79, 0x73c737bf, + 0x53f7cdea, 0x5ffdaa5b, 0xdf3d6f14, 0x7844db86, + 0xcaaff381, 0xb968c43e, 0x3824342c, 0xc2a3405f, + 0x161dc372, 0xbce2250c, 0x283c498b, 0xff0d9541, + 0x39a80171, 0x080cb3de, 0xd8b4e49c, 0x6456c190, + 0x7bcb8461, 0xd532b670, 0x486c5c74, 0xd0b85742); +$code.=<<___; +.byte 0x52, 0x09, 0x6a, 0xd5, 0x30, 0x36, 0xa5, 0x38 +.byte 0xbf, 0x40, 0xa3, 0x9e, 0x81, 0xf3, 0xd7, 0xfb +.byte 0x7c, 0xe3, 0x39, 0x82, 0x9b, 0x2f, 0xff, 0x87 +.byte 0x34, 0x8e, 0x43, 0x44, 0xc4, 0xde, 0xe9, 0xcb +.byte 0x54, 0x7b, 0x94, 0x32, 0xa6, 0xc2, 0x23, 0x3d +.byte 0xee, 0x4c, 0x95, 0x0b, 0x42, 0xfa, 0xc3, 0x4e +.byte 0x08, 0x2e, 0xa1, 0x66, 0x28, 0xd9, 0x24, 0xb2 +.byte 0x76, 0x5b, 0xa2, 0x49, 0x6d, 0x8b, 0xd1, 0x25 +.byte 0x72, 0xf8, 0xf6, 0x64, 0x86, 0x68, 0x98, 0x16 +.byte 0xd4, 0xa4, 0x5c, 0xcc, 0x5d, 0x65, 0xb6, 0x92 +.byte 0x6c, 0x70, 0x48, 0x50, 0xfd, 0xed, 0xb9, 0xda +.byte 0x5e, 0x15, 0x46, 0x57, 0xa7, 0x8d, 0x9d, 0x84 +.byte 0x90, 0xd8, 0xab, 0x00, 0x8c, 0xbc, 0xd3, 0x0a +.byte 0xf7, 0xe4, 0x58, 0x05, 0xb8, 0xb3, 0x45, 0x06 +.byte 0xd0, 0x2c, 0x1e, 0x8f, 0xca, 0x3f, 0x0f, 0x02 +.byte 0xc1, 0xaf, 0xbd, 0x03, 0x01, 0x13, 0x8a, 0x6b +.byte 0x3a, 0x91, 0x11, 0x41, 0x4f, 0x67, 0xdc, 0xea +.byte 0x97, 0xf2, 0xcf, 0xce, 0xf0, 0xb4, 0xe6, 0x73 +.byte 0x96, 0xac, 0x74, 0x22, 0xe7, 0xad, 0x35, 0x85 +.byte 0xe2, 0xf9, 0x37, 0xe8, 0x1c, 0x75, 0xdf, 0x6e +.byte 0x47, 0xf1, 0x1a, 0x71, 0x1d, 0x29, 0xc5, 0x89 +.byte 0x6f, 0xb7, 0x62, 0x0e, 0xaa, 0x18, 0xbe, 0x1b +.byte 0xfc, 0x56, 0x3e, 0x4b, 0xc6, 0xd2, 0x79, 0x20 +.byte 0x9a, 0xdb, 0xc0, 0xfe, 0x78, 0xcd, 0x5a, 0xf4 +.byte 0x1f, 0xdd, 0xa8, 0x33, 0x88, 0x07, 0xc7, 0x31 +.byte 0xb1, 0x12, 0x10, 0x59, 0x27, 0x80, 0xec, 0x5f +.byte 0x60, 0x51, 0x7f, 0xa9, 0x19, 0xb5, 0x4a, 0x0d +.byte 0x2d, 0xe5, 0x7a, 0x9f, 0x93, 0xc9, 0x9c, 0xef +.byte 0xa0, 0xe0, 0x3b, 0x4d, 0xae, 0x2a, 0xf5, 0xb0 +.byte 0xc8, 0xeb, 0xbb, 0x3c, 0x83, 0x53, 0x99, 0x61 +.byte 0x17, 0x2b, 0x04, 0x7e, 0xba, 0x77, 0xd6, 0x26 +.byte 0xe1, 0x69, 0x14, 0x63, 0x55, 0x21, 0x0c, 0x7d + + +.globl .AES_encrypt +.align 7 +.AES_encrypt: + $STU $sp,-$FRAME($sp) + mflr r0 + + $PUSH $out,`$FRAME-$SIZE_T*19`($sp) + $PUSH r14,`$FRAME-$SIZE_T*18`($sp) + $PUSH r15,`$FRAME-$SIZE_T*17`($sp) + $PUSH r16,`$FRAME-$SIZE_T*16`($sp) + $PUSH r17,`$FRAME-$SIZE_T*15`($sp) + $PUSH r18,`$FRAME-$SIZE_T*14`($sp) + $PUSH r19,`$FRAME-$SIZE_T*13`($sp) + $PUSH r20,`$FRAME-$SIZE_T*12`($sp) + $PUSH r21,`$FRAME-$SIZE_T*11`($sp) + $PUSH r22,`$FRAME-$SIZE_T*10`($sp) + $PUSH r23,`$FRAME-$SIZE_T*9`($sp) + $PUSH r24,`$FRAME-$SIZE_T*8`($sp) + $PUSH r25,`$FRAME-$SIZE_T*7`($sp) + $PUSH r26,`$FRAME-$SIZE_T*6`($sp) + $PUSH r27,`$FRAME-$SIZE_T*5`($sp) + $PUSH r28,`$FRAME-$SIZE_T*4`($sp) + $PUSH r29,`$FRAME-$SIZE_T*3`($sp) + $PUSH r30,`$FRAME-$SIZE_T*2`($sp) + $PUSH r31,`$FRAME-$SIZE_T*1`($sp) + $PUSH r0,`$FRAME+$LRSAVE`($sp) + + andi. $t0,$inp,3 + andi. $t1,$out,3 + or. $t0,$t0,$t1 + bne Lenc_unaligned + +Lenc_unaligned_ok: +___ +$code.=<<___ if (!$LITTLE_ENDIAN); + lwz $s0,0($inp) + lwz $s1,4($inp) + lwz $s2,8($inp) + lwz $s3,12($inp) +___ +$code.=<<___ if ($LITTLE_ENDIAN); + lwz $t0,0($inp) + lwz $t1,4($inp) + lwz $t2,8($inp) + lwz $t3,12($inp) + rotlwi $s0,$t0,8 + rotlwi $s1,$t1,8 + rotlwi $s2,$t2,8 + rotlwi $s3,$t3,8 + rlwimi $s0,$t0,24,0,7 + rlwimi $s1,$t1,24,0,7 + rlwimi $s2,$t2,24,0,7 + rlwimi $s3,$t3,24,0,7 + rlwimi $s0,$t0,24,16,23 + rlwimi $s1,$t1,24,16,23 + rlwimi $s2,$t2,24,16,23 + rlwimi $s3,$t3,24,16,23 +___ +$code.=<<___; + bl LAES_Te + bl Lppc_AES_encrypt_compact + $POP $out,`$FRAME-$SIZE_T*19`($sp) +___ +$code.=<<___ if ($LITTLE_ENDIAN); + rotlwi $t0,$s0,8 + rotlwi $t1,$s1,8 + rotlwi $t2,$s2,8 + rotlwi $t3,$s3,8 + rlwimi $t0,$s0,24,0,7 + rlwimi $t1,$s1,24,0,7 + rlwimi $t2,$s2,24,0,7 + rlwimi $t3,$s3,24,0,7 + rlwimi $t0,$s0,24,16,23 + rlwimi $t1,$s1,24,16,23 + rlwimi $t2,$s2,24,16,23 + rlwimi $t3,$s3,24,16,23 + stw $t0,0($out) + stw $t1,4($out) + stw $t2,8($out) + stw $t3,12($out) +___ +$code.=<<___ if (!$LITTLE_ENDIAN); + stw $s0,0($out) + stw $s1,4($out) + stw $s2,8($out) + stw $s3,12($out) +___ +$code.=<<___; + b Lenc_done + +Lenc_unaligned: + subfic $t0,$inp,4096 + subfic $t1,$out,4096 + andi. $t0,$t0,4096-16 + beq Lenc_xpage + andi. $t1,$t1,4096-16 + bne Lenc_unaligned_ok + +Lenc_xpage: + lbz $acc00,0($inp) + lbz $acc01,1($inp) + lbz $acc02,2($inp) + lbz $s0,3($inp) + lbz $acc04,4($inp) + lbz $acc05,5($inp) + lbz $acc06,6($inp) + lbz $s1,7($inp) + lbz $acc08,8($inp) + lbz $acc09,9($inp) + lbz $acc10,10($inp) + insrwi $s0,$acc00,8,0 + lbz $s2,11($inp) + insrwi $s1,$acc04,8,0 + lbz $acc12,12($inp) + insrwi $s0,$acc01,8,8 + lbz $acc13,13($inp) + insrwi $s1,$acc05,8,8 + lbz $acc14,14($inp) + insrwi $s0,$acc02,8,16 + lbz $s3,15($inp) + insrwi $s1,$acc06,8,16 + insrwi $s2,$acc08,8,0 + insrwi $s3,$acc12,8,0 + insrwi $s2,$acc09,8,8 + insrwi $s3,$acc13,8,8 + insrwi $s2,$acc10,8,16 + insrwi $s3,$acc14,8,16 + + bl LAES_Te + bl Lppc_AES_encrypt_compact + $POP $out,`$FRAME-$SIZE_T*19`($sp) + + extrwi $acc00,$s0,8,0 + extrwi $acc01,$s0,8,8 + stb $acc00,0($out) + extrwi $acc02,$s0,8,16 + stb $acc01,1($out) + stb $acc02,2($out) + extrwi $acc04,$s1,8,0 + stb $s0,3($out) + extrwi $acc05,$s1,8,8 + stb $acc04,4($out) + extrwi $acc06,$s1,8,16 + stb $acc05,5($out) + stb $acc06,6($out) + extrwi $acc08,$s2,8,0 + stb $s1,7($out) + extrwi $acc09,$s2,8,8 + stb $acc08,8($out) + extrwi $acc10,$s2,8,16 + stb $acc09,9($out) + stb $acc10,10($out) + extrwi $acc12,$s3,8,0 + stb $s2,11($out) + extrwi $acc13,$s3,8,8 + stb $acc12,12($out) + extrwi $acc14,$s3,8,16 + stb $acc13,13($out) + stb $acc14,14($out) + stb $s3,15($out) + +Lenc_done: + $POP r0,`$FRAME+$LRSAVE`($sp) + $POP r14,`$FRAME-$SIZE_T*18`($sp) + $POP r15,`$FRAME-$SIZE_T*17`($sp) + $POP r16,`$FRAME-$SIZE_T*16`($sp) + $POP r17,`$FRAME-$SIZE_T*15`($sp) + $POP r18,`$FRAME-$SIZE_T*14`($sp) + $POP r19,`$FRAME-$SIZE_T*13`($sp) + $POP r20,`$FRAME-$SIZE_T*12`($sp) + $POP r21,`$FRAME-$SIZE_T*11`($sp) + $POP r22,`$FRAME-$SIZE_T*10`($sp) + $POP r23,`$FRAME-$SIZE_T*9`($sp) + $POP r24,`$FRAME-$SIZE_T*8`($sp) + $POP r25,`$FRAME-$SIZE_T*7`($sp) + $POP r26,`$FRAME-$SIZE_T*6`($sp) + $POP r27,`$FRAME-$SIZE_T*5`($sp) + $POP r28,`$FRAME-$SIZE_T*4`($sp) + $POP r29,`$FRAME-$SIZE_T*3`($sp) + $POP r30,`$FRAME-$SIZE_T*2`($sp) + $POP r31,`$FRAME-$SIZE_T*1`($sp) + mtlr r0 + addi $sp,$sp,$FRAME + blr + .long 0 + .byte 0,12,4,1,0x80,18,3,0 + .long 0 + +.align 5 +Lppc_AES_encrypt: + lwz $acc00,240($key) + addi $Tbl1,$Tbl0,3 + lwz $t0,0($key) + addi $Tbl2,$Tbl0,2 + lwz $t1,4($key) + addi $Tbl3,$Tbl0,1 + lwz $t2,8($key) + addi $acc00,$acc00,-1 + lwz $t3,12($key) + addi $key,$key,16 + xor $s0,$s0,$t0 + xor $s1,$s1,$t1 + xor $s2,$s2,$t2 + xor $s3,$s3,$t3 + mtctr $acc00 +.align 4 +Lenc_loop: + rlwinm $acc00,$s0,`32-24+3`,21,28 + rlwinm $acc01,$s1,`32-24+3`,21,28 + rlwinm $acc02,$s2,`32-24+3`,21,28 + rlwinm $acc03,$s3,`32-24+3`,21,28 + lwz $t0,0($key) + rlwinm $acc04,$s1,`32-16+3`,21,28 + lwz $t1,4($key) + rlwinm $acc05,$s2,`32-16+3`,21,28 + lwz $t2,8($key) + rlwinm $acc06,$s3,`32-16+3`,21,28 + lwz $t3,12($key) + rlwinm $acc07,$s0,`32-16+3`,21,28 + lwzx $acc00,$Tbl0,$acc00 + rlwinm $acc08,$s2,`32-8+3`,21,28 + lwzx $acc01,$Tbl0,$acc01 + rlwinm $acc09,$s3,`32-8+3`,21,28 + lwzx $acc02,$Tbl0,$acc02 + rlwinm $acc10,$s0,`32-8+3`,21,28 + lwzx $acc03,$Tbl0,$acc03 + rlwinm $acc11,$s1,`32-8+3`,21,28 + lwzx $acc04,$Tbl1,$acc04 + rlwinm $acc12,$s3,`0+3`,21,28 + lwzx $acc05,$Tbl1,$acc05 + rlwinm $acc13,$s0,`0+3`,21,28 + lwzx $acc06,$Tbl1,$acc06 + rlwinm $acc14,$s1,`0+3`,21,28 + lwzx $acc07,$Tbl1,$acc07 + rlwinm $acc15,$s2,`0+3`,21,28 + lwzx $acc08,$Tbl2,$acc08 + xor $t0,$t0,$acc00 + lwzx $acc09,$Tbl2,$acc09 + xor $t1,$t1,$acc01 + lwzx $acc10,$Tbl2,$acc10 + xor $t2,$t2,$acc02 + lwzx $acc11,$Tbl2,$acc11 + xor $t3,$t3,$acc03 + lwzx $acc12,$Tbl3,$acc12 + xor $t0,$t0,$acc04 + lwzx $acc13,$Tbl3,$acc13 + xor $t1,$t1,$acc05 + lwzx $acc14,$Tbl3,$acc14 + xor $t2,$t2,$acc06 + lwzx $acc15,$Tbl3,$acc15 + xor $t3,$t3,$acc07 + xor $t0,$t0,$acc08 + xor $t1,$t1,$acc09 + xor $t2,$t2,$acc10 + xor $t3,$t3,$acc11 + xor $s0,$t0,$acc12 + xor $s1,$t1,$acc13 + xor $s2,$t2,$acc14 + xor $s3,$t3,$acc15 + addi $key,$key,16 + bdnz Lenc_loop + + addi $Tbl2,$Tbl0,2048 + nop + lwz $t0,0($key) + rlwinm $acc00,$s0,`32-24`,24,31 + lwz $t1,4($key) + rlwinm $acc01,$s1,`32-24`,24,31 + lwz $t2,8($key) + rlwinm $acc02,$s2,`32-24`,24,31 + lwz $t3,12($key) + rlwinm $acc03,$s3,`32-24`,24,31 + lwz $acc08,`2048+0`($Tbl0) ! prefetch Te4 + rlwinm $acc04,$s1,`32-16`,24,31 + lwz $acc09,`2048+32`($Tbl0) + rlwinm $acc05,$s2,`32-16`,24,31 + lwz $acc10,`2048+64`($Tbl0) + rlwinm $acc06,$s3,`32-16`,24,31 + lwz $acc11,`2048+96`($Tbl0) + rlwinm $acc07,$s0,`32-16`,24,31 + lwz $acc12,`2048+128`($Tbl0) + rlwinm $acc08,$s2,`32-8`,24,31 + lwz $acc13,`2048+160`($Tbl0) + rlwinm $acc09,$s3,`32-8`,24,31 + lwz $acc14,`2048+192`($Tbl0) + rlwinm $acc10,$s0,`32-8`,24,31 + lwz $acc15,`2048+224`($Tbl0) + rlwinm $acc11,$s1,`32-8`,24,31 + lbzx $acc00,$Tbl2,$acc00 + rlwinm $acc12,$s3,`0`,24,31 + lbzx $acc01,$Tbl2,$acc01 + rlwinm $acc13,$s0,`0`,24,31 + lbzx $acc02,$Tbl2,$acc02 + rlwinm $acc14,$s1,`0`,24,31 + lbzx $acc03,$Tbl2,$acc03 + rlwinm $acc15,$s2,`0`,24,31 + lbzx $acc04,$Tbl2,$acc04 + rlwinm $s0,$acc00,24,0,7 + lbzx $acc05,$Tbl2,$acc05 + rlwinm $s1,$acc01,24,0,7 + lbzx $acc06,$Tbl2,$acc06 + rlwinm $s2,$acc02,24,0,7 + lbzx $acc07,$Tbl2,$acc07 + rlwinm $s3,$acc03,24,0,7 + lbzx $acc08,$Tbl2,$acc08 + rlwimi $s0,$acc04,16,8,15 + lbzx $acc09,$Tbl2,$acc09 + rlwimi $s1,$acc05,16,8,15 + lbzx $acc10,$Tbl2,$acc10 + rlwimi $s2,$acc06,16,8,15 + lbzx $acc11,$Tbl2,$acc11 + rlwimi $s3,$acc07,16,8,15 + lbzx $acc12,$Tbl2,$acc12 + rlwimi $s0,$acc08,8,16,23 + lbzx $acc13,$Tbl2,$acc13 + rlwimi $s1,$acc09,8,16,23 + lbzx $acc14,$Tbl2,$acc14 + rlwimi $s2,$acc10,8,16,23 + lbzx $acc15,$Tbl2,$acc15 + rlwimi $s3,$acc11,8,16,23 + or $s0,$s0,$acc12 + or $s1,$s1,$acc13 + or $s2,$s2,$acc14 + or $s3,$s3,$acc15 + xor $s0,$s0,$t0 + xor $s1,$s1,$t1 + xor $s2,$s2,$t2 + xor $s3,$s3,$t3 + blr + .long 0 + .byte 0,12,0x14,0,0,0,0,0 + +.align 4 +Lppc_AES_encrypt_compact: + lwz $acc00,240($key) + addi $Tbl1,$Tbl0,2048 + lwz $t0,0($key) + lis $mask80,0x8080 + lwz $t1,4($key) + lis $mask1b,0x1b1b + lwz $t2,8($key) + ori $mask80,$mask80,0x8080 + lwz $t3,12($key) + ori $mask1b,$mask1b,0x1b1b + addi $key,$key,16 + mtctr $acc00 +.align 4 +Lenc_compact_loop: + xor $s0,$s0,$t0 + xor $s1,$s1,$t1 + rlwinm $acc00,$s0,`32-24`,24,31 + xor $s2,$s2,$t2 + rlwinm $acc01,$s1,`32-24`,24,31 + xor $s3,$s3,$t3 + rlwinm $acc02,$s2,`32-24`,24,31 + rlwinm $acc03,$s3,`32-24`,24,31 + rlwinm $acc04,$s1,`32-16`,24,31 + rlwinm $acc05,$s2,`32-16`,24,31 + rlwinm $acc06,$s3,`32-16`,24,31 + rlwinm $acc07,$s0,`32-16`,24,31 + lbzx $acc00,$Tbl1,$acc00 + rlwinm $acc08,$s2,`32-8`,24,31 + lbzx $acc01,$Tbl1,$acc01 + rlwinm $acc09,$s3,`32-8`,24,31 + lbzx $acc02,$Tbl1,$acc02 + rlwinm $acc10,$s0,`32-8`,24,31 + lbzx $acc03,$Tbl1,$acc03 + rlwinm $acc11,$s1,`32-8`,24,31 + lbzx $acc04,$Tbl1,$acc04 + rlwinm $acc12,$s3,`0`,24,31 + lbzx $acc05,$Tbl1,$acc05 + rlwinm $acc13,$s0,`0`,24,31 + lbzx $acc06,$Tbl1,$acc06 + rlwinm $acc14,$s1,`0`,24,31 + lbzx $acc07,$Tbl1,$acc07 + rlwinm $acc15,$s2,`0`,24,31 + lbzx $acc08,$Tbl1,$acc08 + rlwinm $s0,$acc00,24,0,7 + lbzx $acc09,$Tbl1,$acc09 + rlwinm $s1,$acc01,24,0,7 + lbzx $acc10,$Tbl1,$acc10 + rlwinm $s2,$acc02,24,0,7 + lbzx $acc11,$Tbl1,$acc11 + rlwinm $s3,$acc03,24,0,7 + lbzx $acc12,$Tbl1,$acc12 + rlwimi $s0,$acc04,16,8,15 + lbzx $acc13,$Tbl1,$acc13 + rlwimi $s1,$acc05,16,8,15 + lbzx $acc14,$Tbl1,$acc14 + rlwimi $s2,$acc06,16,8,15 + lbzx $acc15,$Tbl1,$acc15 + rlwimi $s3,$acc07,16,8,15 + rlwimi $s0,$acc08,8,16,23 + rlwimi $s1,$acc09,8,16,23 + rlwimi $s2,$acc10,8,16,23 + rlwimi $s3,$acc11,8,16,23 + lwz $t0,0($key) + or $s0,$s0,$acc12 + lwz $t1,4($key) + or $s1,$s1,$acc13 + lwz $t2,8($key) + or $s2,$s2,$acc14 + lwz $t3,12($key) + or $s3,$s3,$acc15 + + addi $key,$key,16 + bdz Lenc_compact_done + + and $acc00,$s0,$mask80 # r1=r0&0x80808080 + and $acc01,$s1,$mask80 + and $acc02,$s2,$mask80 + and $acc03,$s3,$mask80 + srwi $acc04,$acc00,7 # r1>>7 + andc $acc08,$s0,$mask80 # r0&0x7f7f7f7f + srwi $acc05,$acc01,7 + andc $acc09,$s1,$mask80 + srwi $acc06,$acc02,7 + andc $acc10,$s2,$mask80 + srwi $acc07,$acc03,7 + andc $acc11,$s3,$mask80 + sub $acc00,$acc00,$acc04 # r1-(r1>>7) + sub $acc01,$acc01,$acc05 + sub $acc02,$acc02,$acc06 + sub $acc03,$acc03,$acc07 + add $acc08,$acc08,$acc08 # (r0&0x7f7f7f7f)<<1 + add $acc09,$acc09,$acc09 + add $acc10,$acc10,$acc10 + add $acc11,$acc11,$acc11 + and $acc00,$acc00,$mask1b # (r1-(r1>>7))&0x1b1b1b1b + and $acc01,$acc01,$mask1b + and $acc02,$acc02,$mask1b + and $acc03,$acc03,$mask1b + xor $acc00,$acc00,$acc08 # r2 + xor $acc01,$acc01,$acc09 + rotlwi $acc12,$s0,16 # ROTATE(r0,16) + xor $acc02,$acc02,$acc10 + rotlwi $acc13,$s1,16 + xor $acc03,$acc03,$acc11 + rotlwi $acc14,$s2,16 + + xor $s0,$s0,$acc00 # r0^r2 + rotlwi $acc15,$s3,16 + xor $s1,$s1,$acc01 + rotrwi $s0,$s0,24 # ROTATE(r2^r0,24) + xor $s2,$s2,$acc02 + rotrwi $s1,$s1,24 + xor $s3,$s3,$acc03 + rotrwi $s2,$s2,24 + xor $s0,$s0,$acc00 # ROTATE(r2^r0,24)^r2 + rotrwi $s3,$s3,24 + xor $s1,$s1,$acc01 + xor $s2,$s2,$acc02 + xor $s3,$s3,$acc03 + rotlwi $acc08,$acc12,8 # ROTATE(r0,24) + xor $s0,$s0,$acc12 # + rotlwi $acc09,$acc13,8 + xor $s1,$s1,$acc13 + rotlwi $acc10,$acc14,8 + xor $s2,$s2,$acc14 + rotlwi $acc11,$acc15,8 + xor $s3,$s3,$acc15 + xor $s0,$s0,$acc08 # + xor $s1,$s1,$acc09 + xor $s2,$s2,$acc10 + xor $s3,$s3,$acc11 + + b Lenc_compact_loop +.align 4 +Lenc_compact_done: + xor $s0,$s0,$t0 + xor $s1,$s1,$t1 + xor $s2,$s2,$t2 + xor $s3,$s3,$t3 + blr + .long 0 + .byte 0,12,0x14,0,0,0,0,0 +.size .AES_encrypt,.-.AES_encrypt + +.globl .AES_decrypt +.align 7 +.AES_decrypt: + $STU $sp,-$FRAME($sp) + mflr r0 + + $PUSH $out,`$FRAME-$SIZE_T*19`($sp) + $PUSH r14,`$FRAME-$SIZE_T*18`($sp) + $PUSH r15,`$FRAME-$SIZE_T*17`($sp) + $PUSH r16,`$FRAME-$SIZE_T*16`($sp) + $PUSH r17,`$FRAME-$SIZE_T*15`($sp) + $PUSH r18,`$FRAME-$SIZE_T*14`($sp) + $PUSH r19,`$FRAME-$SIZE_T*13`($sp) + $PUSH r20,`$FRAME-$SIZE_T*12`($sp) + $PUSH r21,`$FRAME-$SIZE_T*11`($sp) + $PUSH r22,`$FRAME-$SIZE_T*10`($sp) + $PUSH r23,`$FRAME-$SIZE_T*9`($sp) + $PUSH r24,`$FRAME-$SIZE_T*8`($sp) + $PUSH r25,`$FRAME-$SIZE_T*7`($sp) + $PUSH r26,`$FRAME-$SIZE_T*6`($sp) + $PUSH r27,`$FRAME-$SIZE_T*5`($sp) + $PUSH r28,`$FRAME-$SIZE_T*4`($sp) + $PUSH r29,`$FRAME-$SIZE_T*3`($sp) + $PUSH r30,`$FRAME-$SIZE_T*2`($sp) + $PUSH r31,`$FRAME-$SIZE_T*1`($sp) + $PUSH r0,`$FRAME+$LRSAVE`($sp) + + andi. $t0,$inp,3 + andi. $t1,$out,3 + or. $t0,$t0,$t1 + bne Ldec_unaligned + +Ldec_unaligned_ok: +___ +$code.=<<___ if (!$LITTLE_ENDIAN); + lwz $s0,0($inp) + lwz $s1,4($inp) + lwz $s2,8($inp) + lwz $s3,12($inp) +___ +$code.=<<___ if ($LITTLE_ENDIAN); + lwz $t0,0($inp) + lwz $t1,4($inp) + lwz $t2,8($inp) + lwz $t3,12($inp) + rotlwi $s0,$t0,8 + rotlwi $s1,$t1,8 + rotlwi $s2,$t2,8 + rotlwi $s3,$t3,8 + rlwimi $s0,$t0,24,0,7 + rlwimi $s1,$t1,24,0,7 + rlwimi $s2,$t2,24,0,7 + rlwimi $s3,$t3,24,0,7 + rlwimi $s0,$t0,24,16,23 + rlwimi $s1,$t1,24,16,23 + rlwimi $s2,$t2,24,16,23 + rlwimi $s3,$t3,24,16,23 +___ +$code.=<<___; + bl LAES_Td + bl Lppc_AES_decrypt_compact + $POP $out,`$FRAME-$SIZE_T*19`($sp) +___ +$code.=<<___ if ($LITTLE_ENDIAN); + rotlwi $t0,$s0,8 + rotlwi $t1,$s1,8 + rotlwi $t2,$s2,8 + rotlwi $t3,$s3,8 + rlwimi $t0,$s0,24,0,7 + rlwimi $t1,$s1,24,0,7 + rlwimi $t2,$s2,24,0,7 + rlwimi $t3,$s3,24,0,7 + rlwimi $t0,$s0,24,16,23 + rlwimi $t1,$s1,24,16,23 + rlwimi $t2,$s2,24,16,23 + rlwimi $t3,$s3,24,16,23 + stw $t0,0($out) + stw $t1,4($out) + stw $t2,8($out) + stw $t3,12($out) +___ +$code.=<<___ if (!$LITTLE_ENDIAN); + stw $s0,0($out) + stw $s1,4($out) + stw $s2,8($out) + stw $s3,12($out) +___ +$code.=<<___; + b Ldec_done + +Ldec_unaligned: + subfic $t0,$inp,4096 + subfic $t1,$out,4096 + andi. $t0,$t0,4096-16 + beq Ldec_xpage + andi. $t1,$t1,4096-16 + bne Ldec_unaligned_ok + +Ldec_xpage: + lbz $acc00,0($inp) + lbz $acc01,1($inp) + lbz $acc02,2($inp) + lbz $s0,3($inp) + lbz $acc04,4($inp) + lbz $acc05,5($inp) + lbz $acc06,6($inp) + lbz $s1,7($inp) + lbz $acc08,8($inp) + lbz $acc09,9($inp) + lbz $acc10,10($inp) + insrwi $s0,$acc00,8,0 + lbz $s2,11($inp) + insrwi $s1,$acc04,8,0 + lbz $acc12,12($inp) + insrwi $s0,$acc01,8,8 + lbz $acc13,13($inp) + insrwi $s1,$acc05,8,8 + lbz $acc14,14($inp) + insrwi $s0,$acc02,8,16 + lbz $s3,15($inp) + insrwi $s1,$acc06,8,16 + insrwi $s2,$acc08,8,0 + insrwi $s3,$acc12,8,0 + insrwi $s2,$acc09,8,8 + insrwi $s3,$acc13,8,8 + insrwi $s2,$acc10,8,16 + insrwi $s3,$acc14,8,16 + + bl LAES_Td + bl Lppc_AES_decrypt_compact + $POP $out,`$FRAME-$SIZE_T*19`($sp) + + extrwi $acc00,$s0,8,0 + extrwi $acc01,$s0,8,8 + stb $acc00,0($out) + extrwi $acc02,$s0,8,16 + stb $acc01,1($out) + stb $acc02,2($out) + extrwi $acc04,$s1,8,0 + stb $s0,3($out) + extrwi $acc05,$s1,8,8 + stb $acc04,4($out) + extrwi $acc06,$s1,8,16 + stb $acc05,5($out) + stb $acc06,6($out) + extrwi $acc08,$s2,8,0 + stb $s1,7($out) + extrwi $acc09,$s2,8,8 + stb $acc08,8($out) + extrwi $acc10,$s2,8,16 + stb $acc09,9($out) + stb $acc10,10($out) + extrwi $acc12,$s3,8,0 + stb $s2,11($out) + extrwi $acc13,$s3,8,8 + stb $acc12,12($out) + extrwi $acc14,$s3,8,16 + stb $acc13,13($out) + stb $acc14,14($out) + stb $s3,15($out) + +Ldec_done: + $POP r0,`$FRAME+$LRSAVE`($sp) + $POP r14,`$FRAME-$SIZE_T*18`($sp) + $POP r15,`$FRAME-$SIZE_T*17`($sp) + $POP r16,`$FRAME-$SIZE_T*16`($sp) + $POP r17,`$FRAME-$SIZE_T*15`($sp) + $POP r18,`$FRAME-$SIZE_T*14`($sp) + $POP r19,`$FRAME-$SIZE_T*13`($sp) + $POP r20,`$FRAME-$SIZE_T*12`($sp) + $POP r21,`$FRAME-$SIZE_T*11`($sp) + $POP r22,`$FRAME-$SIZE_T*10`($sp) + $POP r23,`$FRAME-$SIZE_T*9`($sp) + $POP r24,`$FRAME-$SIZE_T*8`($sp) + $POP r25,`$FRAME-$SIZE_T*7`($sp) + $POP r26,`$FRAME-$SIZE_T*6`($sp) + $POP r27,`$FRAME-$SIZE_T*5`($sp) + $POP r28,`$FRAME-$SIZE_T*4`($sp) + $POP r29,`$FRAME-$SIZE_T*3`($sp) + $POP r30,`$FRAME-$SIZE_T*2`($sp) + $POP r31,`$FRAME-$SIZE_T*1`($sp) + mtlr r0 + addi $sp,$sp,$FRAME + blr + .long 0 + .byte 0,12,4,1,0x80,18,3,0 + .long 0 + +.align 5 +Lppc_AES_decrypt: + lwz $acc00,240($key) + addi $Tbl1,$Tbl0,3 + lwz $t0,0($key) + addi $Tbl2,$Tbl0,2 + lwz $t1,4($key) + addi $Tbl3,$Tbl0,1 + lwz $t2,8($key) + addi $acc00,$acc00,-1 + lwz $t3,12($key) + addi $key,$key,16 + xor $s0,$s0,$t0 + xor $s1,$s1,$t1 + xor $s2,$s2,$t2 + xor $s3,$s3,$t3 + mtctr $acc00 +.align 4 +Ldec_loop: + rlwinm $acc00,$s0,`32-24+3`,21,28 + rlwinm $acc01,$s1,`32-24+3`,21,28 + rlwinm $acc02,$s2,`32-24+3`,21,28 + rlwinm $acc03,$s3,`32-24+3`,21,28 + lwz $t0,0($key) + rlwinm $acc04,$s3,`32-16+3`,21,28 + lwz $t1,4($key) + rlwinm $acc05,$s0,`32-16+3`,21,28 + lwz $t2,8($key) + rlwinm $acc06,$s1,`32-16+3`,21,28 + lwz $t3,12($key) + rlwinm $acc07,$s2,`32-16+3`,21,28 + lwzx $acc00,$Tbl0,$acc00 + rlwinm $acc08,$s2,`32-8+3`,21,28 + lwzx $acc01,$Tbl0,$acc01 + rlwinm $acc09,$s3,`32-8+3`,21,28 + lwzx $acc02,$Tbl0,$acc02 + rlwinm $acc10,$s0,`32-8+3`,21,28 + lwzx $acc03,$Tbl0,$acc03 + rlwinm $acc11,$s1,`32-8+3`,21,28 + lwzx $acc04,$Tbl1,$acc04 + rlwinm $acc12,$s1,`0+3`,21,28 + lwzx $acc05,$Tbl1,$acc05 + rlwinm $acc13,$s2,`0+3`,21,28 + lwzx $acc06,$Tbl1,$acc06 + rlwinm $acc14,$s3,`0+3`,21,28 + lwzx $acc07,$Tbl1,$acc07 + rlwinm $acc15,$s0,`0+3`,21,28 + lwzx $acc08,$Tbl2,$acc08 + xor $t0,$t0,$acc00 + lwzx $acc09,$Tbl2,$acc09 + xor $t1,$t1,$acc01 + lwzx $acc10,$Tbl2,$acc10 + xor $t2,$t2,$acc02 + lwzx $acc11,$Tbl2,$acc11 + xor $t3,$t3,$acc03 + lwzx $acc12,$Tbl3,$acc12 + xor $t0,$t0,$acc04 + lwzx $acc13,$Tbl3,$acc13 + xor $t1,$t1,$acc05 + lwzx $acc14,$Tbl3,$acc14 + xor $t2,$t2,$acc06 + lwzx $acc15,$Tbl3,$acc15 + xor $t3,$t3,$acc07 + xor $t0,$t0,$acc08 + xor $t1,$t1,$acc09 + xor $t2,$t2,$acc10 + xor $t3,$t3,$acc11 + xor $s0,$t0,$acc12 + xor $s1,$t1,$acc13 + xor $s2,$t2,$acc14 + xor $s3,$t3,$acc15 + addi $key,$key,16 + bdnz Ldec_loop + + addi $Tbl2,$Tbl0,2048 + nop + lwz $t0,0($key) + rlwinm $acc00,$s0,`32-24`,24,31 + lwz $t1,4($key) + rlwinm $acc01,$s1,`32-24`,24,31 + lwz $t2,8($key) + rlwinm $acc02,$s2,`32-24`,24,31 + lwz $t3,12($key) + rlwinm $acc03,$s3,`32-24`,24,31 + lwz $acc08,`2048+0`($Tbl0) ! prefetch Td4 + rlwinm $acc04,$s3,`32-16`,24,31 + lwz $acc09,`2048+32`($Tbl0) + rlwinm $acc05,$s0,`32-16`,24,31 + lwz $acc10,`2048+64`($Tbl0) + lbzx $acc00,$Tbl2,$acc00 + lwz $acc11,`2048+96`($Tbl0) + lbzx $acc01,$Tbl2,$acc01 + lwz $acc12,`2048+128`($Tbl0) + rlwinm $acc06,$s1,`32-16`,24,31 + lwz $acc13,`2048+160`($Tbl0) + rlwinm $acc07,$s2,`32-16`,24,31 + lwz $acc14,`2048+192`($Tbl0) + rlwinm $acc08,$s2,`32-8`,24,31 + lwz $acc15,`2048+224`($Tbl0) + rlwinm $acc09,$s3,`32-8`,24,31 + lbzx $acc02,$Tbl2,$acc02 + rlwinm $acc10,$s0,`32-8`,24,31 + lbzx $acc03,$Tbl2,$acc03 + rlwinm $acc11,$s1,`32-8`,24,31 + lbzx $acc04,$Tbl2,$acc04 + rlwinm $acc12,$s1,`0`,24,31 + lbzx $acc05,$Tbl2,$acc05 + rlwinm $acc13,$s2,`0`,24,31 + lbzx $acc06,$Tbl2,$acc06 + rlwinm $acc14,$s3,`0`,24,31 + lbzx $acc07,$Tbl2,$acc07 + rlwinm $acc15,$s0,`0`,24,31 + lbzx $acc08,$Tbl2,$acc08 + rlwinm $s0,$acc00,24,0,7 + lbzx $acc09,$Tbl2,$acc09 + rlwinm $s1,$acc01,24,0,7 + lbzx $acc10,$Tbl2,$acc10 + rlwinm $s2,$acc02,24,0,7 + lbzx $acc11,$Tbl2,$acc11 + rlwinm $s3,$acc03,24,0,7 + lbzx $acc12,$Tbl2,$acc12 + rlwimi $s0,$acc04,16,8,15 + lbzx $acc13,$Tbl2,$acc13 + rlwimi $s1,$acc05,16,8,15 + lbzx $acc14,$Tbl2,$acc14 + rlwimi $s2,$acc06,16,8,15 + lbzx $acc15,$Tbl2,$acc15 + rlwimi $s3,$acc07,16,8,15 + rlwimi $s0,$acc08,8,16,23 + rlwimi $s1,$acc09,8,16,23 + rlwimi $s2,$acc10,8,16,23 + rlwimi $s3,$acc11,8,16,23 + or $s0,$s0,$acc12 + or $s1,$s1,$acc13 + or $s2,$s2,$acc14 + or $s3,$s3,$acc15 + xor $s0,$s0,$t0 + xor $s1,$s1,$t1 + xor $s2,$s2,$t2 + xor $s3,$s3,$t3 + blr + .long 0 + .byte 0,12,0x14,0,0,0,0,0 + +.align 4 +Lppc_AES_decrypt_compact: + lwz $acc00,240($key) + addi $Tbl1,$Tbl0,2048 + lwz $t0,0($key) + lis $mask80,0x8080 + lwz $t1,4($key) + lis $mask1b,0x1b1b + lwz $t2,8($key) + ori $mask80,$mask80,0x8080 + lwz $t3,12($key) + ori $mask1b,$mask1b,0x1b1b + addi $key,$key,16 +___ +$code.=<<___ if ($SIZE_T==8); + insrdi $mask80,$mask80,32,0 + insrdi $mask1b,$mask1b,32,0 +___ +$code.=<<___; + mtctr $acc00 +.align 4 +Ldec_compact_loop: + xor $s0,$s0,$t0 + xor $s1,$s1,$t1 + rlwinm $acc00,$s0,`32-24`,24,31 + xor $s2,$s2,$t2 + rlwinm $acc01,$s1,`32-24`,24,31 + xor $s3,$s3,$t3 + rlwinm $acc02,$s2,`32-24`,24,31 + rlwinm $acc03,$s3,`32-24`,24,31 + rlwinm $acc04,$s3,`32-16`,24,31 + rlwinm $acc05,$s0,`32-16`,24,31 + rlwinm $acc06,$s1,`32-16`,24,31 + rlwinm $acc07,$s2,`32-16`,24,31 + lbzx $acc00,$Tbl1,$acc00 + rlwinm $acc08,$s2,`32-8`,24,31 + lbzx $acc01,$Tbl1,$acc01 + rlwinm $acc09,$s3,`32-8`,24,31 + lbzx $acc02,$Tbl1,$acc02 + rlwinm $acc10,$s0,`32-8`,24,31 + lbzx $acc03,$Tbl1,$acc03 + rlwinm $acc11,$s1,`32-8`,24,31 + lbzx $acc04,$Tbl1,$acc04 + rlwinm $acc12,$s1,`0`,24,31 + lbzx $acc05,$Tbl1,$acc05 + rlwinm $acc13,$s2,`0`,24,31 + lbzx $acc06,$Tbl1,$acc06 + rlwinm $acc14,$s3,`0`,24,31 + lbzx $acc07,$Tbl1,$acc07 + rlwinm $acc15,$s0,`0`,24,31 + lbzx $acc08,$Tbl1,$acc08 + rlwinm $s0,$acc00,24,0,7 + lbzx $acc09,$Tbl1,$acc09 + rlwinm $s1,$acc01,24,0,7 + lbzx $acc10,$Tbl1,$acc10 + rlwinm $s2,$acc02,24,0,7 + lbzx $acc11,$Tbl1,$acc11 + rlwinm $s3,$acc03,24,0,7 + lbzx $acc12,$Tbl1,$acc12 + rlwimi $s0,$acc04,16,8,15 + lbzx $acc13,$Tbl1,$acc13 + rlwimi $s1,$acc05,16,8,15 + lbzx $acc14,$Tbl1,$acc14 + rlwimi $s2,$acc06,16,8,15 + lbzx $acc15,$Tbl1,$acc15 + rlwimi $s3,$acc07,16,8,15 + rlwimi $s0,$acc08,8,16,23 + rlwimi $s1,$acc09,8,16,23 + rlwimi $s2,$acc10,8,16,23 + rlwimi $s3,$acc11,8,16,23 + lwz $t0,0($key) + or $s0,$s0,$acc12 + lwz $t1,4($key) + or $s1,$s1,$acc13 + lwz $t2,8($key) + or $s2,$s2,$acc14 + lwz $t3,12($key) + or $s3,$s3,$acc15 + + addi $key,$key,16 + bdz Ldec_compact_done +___ +$code.=<<___ if ($SIZE_T==8); + # vectorized permutation improves decrypt performance by 10% + insrdi $s0,$s1,32,0 + insrdi $s2,$s3,32,0 + + and $acc00,$s0,$mask80 # r1=r0&0x80808080 + and $acc02,$s2,$mask80 + srdi $acc04,$acc00,7 # r1>>7 + srdi $acc06,$acc02,7 + andc $acc08,$s0,$mask80 # r0&0x7f7f7f7f + andc $acc10,$s2,$mask80 + sub $acc00,$acc00,$acc04 # r1-(r1>>7) + sub $acc02,$acc02,$acc06 + add $acc08,$acc08,$acc08 # (r0&0x7f7f7f7f)<<1 + add $acc10,$acc10,$acc10 + and $acc00,$acc00,$mask1b # (r1-(r1>>7))&0x1b1b1b1b + and $acc02,$acc02,$mask1b + xor $acc00,$acc00,$acc08 # r2 + xor $acc02,$acc02,$acc10 + + and $acc04,$acc00,$mask80 # r1=r2&0x80808080 + and $acc06,$acc02,$mask80 + srdi $acc08,$acc04,7 # r1>>7 + srdi $acc10,$acc06,7 + andc $acc12,$acc00,$mask80 # r2&0x7f7f7f7f + andc $acc14,$acc02,$mask80 + sub $acc04,$acc04,$acc08 # r1-(r1>>7) + sub $acc06,$acc06,$acc10 + add $acc12,$acc12,$acc12 # (r2&0x7f7f7f7f)<<1 + add $acc14,$acc14,$acc14 + and $acc04,$acc04,$mask1b # (r1-(r1>>7))&0x1b1b1b1b + and $acc06,$acc06,$mask1b + xor $acc04,$acc04,$acc12 # r4 + xor $acc06,$acc06,$acc14 + + and $acc08,$acc04,$mask80 # r1=r4&0x80808080 + and $acc10,$acc06,$mask80 + srdi $acc12,$acc08,7 # r1>>7 + srdi $acc14,$acc10,7 + sub $acc08,$acc08,$acc12 # r1-(r1>>7) + sub $acc10,$acc10,$acc14 + andc $acc12,$acc04,$mask80 # r4&0x7f7f7f7f + andc $acc14,$acc06,$mask80 + add $acc12,$acc12,$acc12 # (r4&0x7f7f7f7f)<<1 + add $acc14,$acc14,$acc14 + and $acc08,$acc08,$mask1b # (r1-(r1>>7))&0x1b1b1b1b + and $acc10,$acc10,$mask1b + xor $acc08,$acc08,$acc12 # r8 + xor $acc10,$acc10,$acc14 + + xor $acc00,$acc00,$s0 # r2^r0 + xor $acc02,$acc02,$s2 + xor $acc04,$acc04,$s0 # r4^r0 + xor $acc06,$acc06,$s2 + + extrdi $acc01,$acc00,32,0 + extrdi $acc03,$acc02,32,0 + extrdi $acc05,$acc04,32,0 + extrdi $acc07,$acc06,32,0 + extrdi $acc09,$acc08,32,0 + extrdi $acc11,$acc10,32,0 +___ +$code.=<<___ if ($SIZE_T==4); + and $acc00,$s0,$mask80 # r1=r0&0x80808080 + and $acc01,$s1,$mask80 + and $acc02,$s2,$mask80 + and $acc03,$s3,$mask80 + srwi $acc04,$acc00,7 # r1>>7 + andc $acc08,$s0,$mask80 # r0&0x7f7f7f7f + srwi $acc05,$acc01,7 + andc $acc09,$s1,$mask80 + srwi $acc06,$acc02,7 + andc $acc10,$s2,$mask80 + srwi $acc07,$acc03,7 + andc $acc11,$s3,$mask80 + sub $acc00,$acc00,$acc04 # r1-(r1>>7) + sub $acc01,$acc01,$acc05 + sub $acc02,$acc02,$acc06 + sub $acc03,$acc03,$acc07 + add $acc08,$acc08,$acc08 # (r0&0x7f7f7f7f)<<1 + add $acc09,$acc09,$acc09 + add $acc10,$acc10,$acc10 + add $acc11,$acc11,$acc11 + and $acc00,$acc00,$mask1b # (r1-(r1>>7))&0x1b1b1b1b + and $acc01,$acc01,$mask1b + and $acc02,$acc02,$mask1b + and $acc03,$acc03,$mask1b + xor $acc00,$acc00,$acc08 # r2 + xor $acc01,$acc01,$acc09 + xor $acc02,$acc02,$acc10 + xor $acc03,$acc03,$acc11 + + and $acc04,$acc00,$mask80 # r1=r2&0x80808080 + and $acc05,$acc01,$mask80 + and $acc06,$acc02,$mask80 + and $acc07,$acc03,$mask80 + srwi $acc08,$acc04,7 # r1>>7 + andc $acc12,$acc00,$mask80 # r2&0x7f7f7f7f + srwi $acc09,$acc05,7 + andc $acc13,$acc01,$mask80 + srwi $acc10,$acc06,7 + andc $acc14,$acc02,$mask80 + srwi $acc11,$acc07,7 + andc $acc15,$acc03,$mask80 + sub $acc04,$acc04,$acc08 # r1-(r1>>7) + sub $acc05,$acc05,$acc09 + sub $acc06,$acc06,$acc10 + sub $acc07,$acc07,$acc11 + add $acc12,$acc12,$acc12 # (r2&0x7f7f7f7f)<<1 + add $acc13,$acc13,$acc13 + add $acc14,$acc14,$acc14 + add $acc15,$acc15,$acc15 + and $acc04,$acc04,$mask1b # (r1-(r1>>7))&0x1b1b1b1b + and $acc05,$acc05,$mask1b + and $acc06,$acc06,$mask1b + and $acc07,$acc07,$mask1b + xor $acc04,$acc04,$acc12 # r4 + xor $acc05,$acc05,$acc13 + xor $acc06,$acc06,$acc14 + xor $acc07,$acc07,$acc15 + + and $acc08,$acc04,$mask80 # r1=r4&0x80808080 + and $acc09,$acc05,$mask80 + srwi $acc12,$acc08,7 # r1>>7 + and $acc10,$acc06,$mask80 + srwi $acc13,$acc09,7 + and $acc11,$acc07,$mask80 + srwi $acc14,$acc10,7 + sub $acc08,$acc08,$acc12 # r1-(r1>>7) + srwi $acc15,$acc11,7 + sub $acc09,$acc09,$acc13 + sub $acc10,$acc10,$acc14 + sub $acc11,$acc11,$acc15 + andc $acc12,$acc04,$mask80 # r4&0x7f7f7f7f + andc $acc13,$acc05,$mask80 + andc $acc14,$acc06,$mask80 + andc $acc15,$acc07,$mask80 + add $acc12,$acc12,$acc12 # (r4&0x7f7f7f7f)<<1 + add $acc13,$acc13,$acc13 + add $acc14,$acc14,$acc14 + add $acc15,$acc15,$acc15 + and $acc08,$acc08,$mask1b # (r1-(r1>>7))&0x1b1b1b1b + and $acc09,$acc09,$mask1b + and $acc10,$acc10,$mask1b + and $acc11,$acc11,$mask1b + xor $acc08,$acc08,$acc12 # r8 + xor $acc09,$acc09,$acc13 + xor $acc10,$acc10,$acc14 + xor $acc11,$acc11,$acc15 + + xor $acc00,$acc00,$s0 # r2^r0 + xor $acc01,$acc01,$s1 + xor $acc02,$acc02,$s2 + xor $acc03,$acc03,$s3 + xor $acc04,$acc04,$s0 # r4^r0 + xor $acc05,$acc05,$s1 + xor $acc06,$acc06,$s2 + xor $acc07,$acc07,$s3 +___ +$code.=<<___; + rotrwi $s0,$s0,8 # = ROTATE(r0,8) + rotrwi $s1,$s1,8 + xor $s0,$s0,$acc00 # ^= r2^r0 + rotrwi $s2,$s2,8 + xor $s1,$s1,$acc01 + rotrwi $s3,$s3,8 + xor $s2,$s2,$acc02 + xor $s3,$s3,$acc03 + xor $acc00,$acc00,$acc08 + xor $acc01,$acc01,$acc09 + xor $acc02,$acc02,$acc10 + xor $acc03,$acc03,$acc11 + xor $s0,$s0,$acc04 # ^= r4^r0 + rotrwi $acc00,$acc00,24 + xor $s1,$s1,$acc05 + rotrwi $acc01,$acc01,24 + xor $s2,$s2,$acc06 + rotrwi $acc02,$acc02,24 + xor $s3,$s3,$acc07 + rotrwi $acc03,$acc03,24 + xor $acc04,$acc04,$acc08 + xor $acc05,$acc05,$acc09 + xor $acc06,$acc06,$acc10 + xor $acc07,$acc07,$acc11 + xor $s0,$s0,$acc08 # ^= r8 [^((r4^r0)^(r2^r0)=r4^r2)] + rotrwi $acc04,$acc04,16 + xor $s1,$s1,$acc09 + rotrwi $acc05,$acc05,16 + xor $s2,$s2,$acc10 + rotrwi $acc06,$acc06,16 + xor $s3,$s3,$acc11 + rotrwi $acc07,$acc07,16 + xor $s0,$s0,$acc00 # ^= ROTATE(r8^r2^r0,24) + rotrwi $acc08,$acc08,8 + xor $s1,$s1,$acc01 + rotrwi $acc09,$acc09,8 + xor $s2,$s2,$acc02 + rotrwi $acc10,$acc10,8 + xor $s3,$s3,$acc03 + rotrwi $acc11,$acc11,8 + xor $s0,$s0,$acc04 # ^= ROTATE(r8^r4^r0,16) + xor $s1,$s1,$acc05 + xor $s2,$s2,$acc06 + xor $s3,$s3,$acc07 + xor $s0,$s0,$acc08 # ^= ROTATE(r8,8) + xor $s1,$s1,$acc09 + xor $s2,$s2,$acc10 + xor $s3,$s3,$acc11 + + b Ldec_compact_loop +.align 4 +Ldec_compact_done: + xor $s0,$s0,$t0 + xor $s1,$s1,$t1 + xor $s2,$s2,$t2 + xor $s3,$s3,$t3 + blr + .long 0 + .byte 0,12,0x14,0,0,0,0,0 +.size .AES_decrypt,.-.AES_decrypt + +.asciz "AES for PPC, CRYPTOGAMS by " +.align 7 +___ + +$code =~ s/\`([^\`]*)\`/eval $1/gem; +print $code; +close STDOUT; diff --git a/openssl-1.1.0h/crypto/aes/asm/aes-s390x.pl b/openssl-1.1.0h/crypto/aes/asm/aes-s390x.pl new file mode 100644 index 0000000..fd8a737 --- /dev/null +++ b/openssl-1.1.0h/crypto/aes/asm/aes-s390x.pl @@ -0,0 +1,2226 @@ +#! /usr/bin/env perl +# Copyright 2007-2016 The OpenSSL Project Authors. All Rights Reserved. +# +# Licensed under the OpenSSL license (the "License"). You may not use +# this file except in compliance with the License. You can obtain a copy +# in the file LICENSE in the source distribution or at +# https://www.openssl.org/source/license.html + + +# ==================================================================== +# Written by Andy Polyakov for the OpenSSL +# project. The module is, however, dual licensed under OpenSSL and +# CRYPTOGAMS licenses depending on where you obtain it. For further +# details see http://www.openssl.org/~appro/cryptogams/. +# ==================================================================== + +# AES for s390x. + +# April 2007. +# +# Software performance improvement over gcc-generated code is ~70% and +# in absolute terms is ~73 cycles per byte processed with 128-bit key. +# You're likely to exclaim "why so slow?" Keep in mind that z-CPUs are +# *strictly* in-order execution and issued instruction [in this case +# load value from memory is critical] has to complete before execution +# flow proceeds. S-boxes are compressed to 2KB[+256B]. +# +# As for hardware acceleration support. It's basically a "teaser," as +# it can and should be improved in several ways. Most notably support +# for CBC is not utilized, nor multiple blocks are ever processed. +# Then software key schedule can be postponed till hardware support +# detection... Performance improvement over assembler is reportedly +# ~2.5x, but can reach >8x [naturally on larger chunks] if proper +# support is implemented. + +# May 2007. +# +# Implement AES_set_[en|de]crypt_key. Key schedule setup is avoided +# for 128-bit keys, if hardware support is detected. + +# Januray 2009. +# +# Add support for hardware AES192/256 and reschedule instructions to +# minimize/avoid Address Generation Interlock hazard and to favour +# dual-issue z10 pipeline. This gave ~25% improvement on z10 and +# almost 50% on z9. The gain is smaller on z10, because being dual- +# issue z10 makes it improssible to eliminate the interlock condition: +# critial path is not long enough. Yet it spends ~24 cycles per byte +# processed with 128-bit key. +# +# Unlike previous version hardware support detection takes place only +# at the moment of key schedule setup, which is denoted in key->rounds. +# This is done, because deferred key setup can't be made MT-safe, not +# for keys longer than 128 bits. +# +# Add AES_cbc_encrypt, which gives incredible performance improvement, +# it was measured to be ~6.6x. It's less than previously mentioned 8x, +# because software implementation was optimized. + +# May 2010. +# +# Add AES_ctr32_encrypt. If hardware-assisted, it provides up to 4.3x +# performance improvement over "generic" counter mode routine relying +# on single-block, also hardware-assisted, AES_encrypt. "Up to" refers +# to the fact that exact throughput value depends on current stack +# frame alignment within 4KB page. In worst case you get ~75% of the +# maximum, but *on average* it would be as much as ~98%. Meaning that +# worst case is unlike, it's like hitting ravine on plateau. + +# November 2010. +# +# Adapt for -m31 build. If kernel supports what's called "highgprs" +# feature on Linux [see /proc/cpuinfo], it's possible to use 64-bit +# instructions and achieve "64-bit" performance even in 31-bit legacy +# application context. The feature is not specific to any particular +# processor, as long as it's "z-CPU". Latter implies that the code +# remains z/Architecture specific. On z990 it was measured to perform +# 2x better than code generated by gcc 4.3. + +# December 2010. +# +# Add support for z196 "cipher message with counter" instruction. +# Note however that it's disengaged, because it was measured to +# perform ~12% worse than vanilla km-based code... + +# February 2011. +# +# Add AES_xts_[en|de]crypt. This includes support for z196 km-xts-aes +# instructions, which deliver ~70% improvement at 8KB block size over +# vanilla km-based code, 37% - at most like 512-bytes block size. + +$flavour = shift; + +if ($flavour =~ /3[12]/) { + $SIZE_T=4; + $g=""; +} else { + $SIZE_T=8; + $g="g"; +} + +while (($output=shift) && ($output!~/\w[\w\-]*\.\w+$/)) {} +open STDOUT,">$output"; + +$softonly=0; # allow hardware support + +$t0="%r0"; $mask="%r0"; +$t1="%r1"; +$t2="%r2"; $inp="%r2"; +$t3="%r3"; $out="%r3"; $bits="%r3"; +$key="%r4"; +$i1="%r5"; +$i2="%r6"; +$i3="%r7"; +$s0="%r8"; +$s1="%r9"; +$s2="%r10"; +$s3="%r11"; +$tbl="%r12"; +$rounds="%r13"; +$ra="%r14"; +$sp="%r15"; + +$stdframe=16*$SIZE_T+4*8; + +sub _data_word() +{ my $i; + while(defined($i=shift)) { $code.=sprintf".long\t0x%08x,0x%08x\n",$i,$i; } +} + +$code=<<___; +.text + +.type AES_Te,\@object +.align 256 +AES_Te: +___ +&_data_word( + 0xc66363a5, 0xf87c7c84, 0xee777799, 0xf67b7b8d, + 0xfff2f20d, 0xd66b6bbd, 0xde6f6fb1, 0x91c5c554, + 0x60303050, 0x02010103, 0xce6767a9, 0x562b2b7d, + 0xe7fefe19, 0xb5d7d762, 0x4dababe6, 0xec76769a, + 0x8fcaca45, 0x1f82829d, 0x89c9c940, 0xfa7d7d87, + 0xeffafa15, 0xb25959eb, 0x8e4747c9, 0xfbf0f00b, + 0x41adadec, 0xb3d4d467, 0x5fa2a2fd, 0x45afafea, + 0x239c9cbf, 0x53a4a4f7, 0xe4727296, 0x9bc0c05b, + 0x75b7b7c2, 0xe1fdfd1c, 0x3d9393ae, 0x4c26266a, + 0x6c36365a, 0x7e3f3f41, 0xf5f7f702, 0x83cccc4f, + 0x6834345c, 0x51a5a5f4, 0xd1e5e534, 0xf9f1f108, + 0xe2717193, 0xabd8d873, 0x62313153, 0x2a15153f, + 0x0804040c, 0x95c7c752, 0x46232365, 0x9dc3c35e, + 0x30181828, 0x379696a1, 0x0a05050f, 0x2f9a9ab5, + 0x0e070709, 0x24121236, 0x1b80809b, 0xdfe2e23d, + 0xcdebeb26, 0x4e272769, 0x7fb2b2cd, 0xea75759f, + 0x1209091b, 0x1d83839e, 0x582c2c74, 0x341a1a2e, + 0x361b1b2d, 0xdc6e6eb2, 0xb45a5aee, 0x5ba0a0fb, + 0xa45252f6, 0x763b3b4d, 0xb7d6d661, 0x7db3b3ce, + 0x5229297b, 0xdde3e33e, 0x5e2f2f71, 0x13848497, + 0xa65353f5, 0xb9d1d168, 0x00000000, 0xc1eded2c, + 0x40202060, 0xe3fcfc1f, 0x79b1b1c8, 0xb65b5bed, + 0xd46a6abe, 0x8dcbcb46, 0x67bebed9, 0x7239394b, + 0x944a4ade, 0x984c4cd4, 0xb05858e8, 0x85cfcf4a, + 0xbbd0d06b, 0xc5efef2a, 0x4faaaae5, 0xedfbfb16, + 0x864343c5, 0x9a4d4dd7, 0x66333355, 0x11858594, + 0x8a4545cf, 0xe9f9f910, 0x04020206, 0xfe7f7f81, + 0xa05050f0, 0x783c3c44, 0x259f9fba, 0x4ba8a8e3, + 0xa25151f3, 0x5da3a3fe, 0x804040c0, 0x058f8f8a, + 0x3f9292ad, 0x219d9dbc, 0x70383848, 0xf1f5f504, + 0x63bcbcdf, 0x77b6b6c1, 0xafdada75, 0x42212163, + 0x20101030, 0xe5ffff1a, 0xfdf3f30e, 0xbfd2d26d, + 0x81cdcd4c, 0x180c0c14, 0x26131335, 0xc3ecec2f, + 0xbe5f5fe1, 0x359797a2, 0x884444cc, 0x2e171739, + 0x93c4c457, 0x55a7a7f2, 0xfc7e7e82, 0x7a3d3d47, + 0xc86464ac, 0xba5d5de7, 0x3219192b, 0xe6737395, + 0xc06060a0, 0x19818198, 0x9e4f4fd1, 0xa3dcdc7f, + 0x44222266, 0x542a2a7e, 0x3b9090ab, 0x0b888883, + 0x8c4646ca, 0xc7eeee29, 0x6bb8b8d3, 0x2814143c, + 0xa7dede79, 0xbc5e5ee2, 0x160b0b1d, 0xaddbdb76, + 0xdbe0e03b, 0x64323256, 0x743a3a4e, 0x140a0a1e, + 0x924949db, 0x0c06060a, 0x4824246c, 0xb85c5ce4, + 0x9fc2c25d, 0xbdd3d36e, 0x43acacef, 0xc46262a6, + 0x399191a8, 0x319595a4, 0xd3e4e437, 0xf279798b, + 0xd5e7e732, 0x8bc8c843, 0x6e373759, 0xda6d6db7, + 0x018d8d8c, 0xb1d5d564, 0x9c4e4ed2, 0x49a9a9e0, + 0xd86c6cb4, 0xac5656fa, 0xf3f4f407, 0xcfeaea25, + 0xca6565af, 0xf47a7a8e, 0x47aeaee9, 0x10080818, + 0x6fbabad5, 0xf0787888, 0x4a25256f, 0x5c2e2e72, + 0x381c1c24, 0x57a6a6f1, 0x73b4b4c7, 0x97c6c651, + 0xcbe8e823, 0xa1dddd7c, 0xe874749c, 0x3e1f1f21, + 0x964b4bdd, 0x61bdbddc, 0x0d8b8b86, 0x0f8a8a85, + 0xe0707090, 0x7c3e3e42, 0x71b5b5c4, 0xcc6666aa, + 0x904848d8, 0x06030305, 0xf7f6f601, 0x1c0e0e12, + 0xc26161a3, 0x6a35355f, 0xae5757f9, 0x69b9b9d0, + 0x17868691, 0x99c1c158, 0x3a1d1d27, 0x279e9eb9, + 0xd9e1e138, 0xebf8f813, 0x2b9898b3, 0x22111133, + 0xd26969bb, 0xa9d9d970, 0x078e8e89, 0x339494a7, + 0x2d9b9bb6, 0x3c1e1e22, 0x15878792, 0xc9e9e920, + 0x87cece49, 0xaa5555ff, 0x50282878, 0xa5dfdf7a, + 0x038c8c8f, 0x59a1a1f8, 0x09898980, 0x1a0d0d17, + 0x65bfbfda, 0xd7e6e631, 0x844242c6, 0xd06868b8, + 0x824141c3, 0x299999b0, 0x5a2d2d77, 0x1e0f0f11, + 0x7bb0b0cb, 0xa85454fc, 0x6dbbbbd6, 0x2c16163a); +$code.=<<___; +# Te4[256] +.byte 0x63, 0x7c, 0x77, 0x7b, 0xf2, 0x6b, 0x6f, 0xc5 +.byte 0x30, 0x01, 0x67, 0x2b, 0xfe, 0xd7, 0xab, 0x76 +.byte 0xca, 0x82, 0xc9, 0x7d, 0xfa, 0x59, 0x47, 0xf0 +.byte 0xad, 0xd4, 0xa2, 0xaf, 0x9c, 0xa4, 0x72, 0xc0 +.byte 0xb7, 0xfd, 0x93, 0x26, 0x36, 0x3f, 0xf7, 0xcc +.byte 0x34, 0xa5, 0xe5, 0xf1, 0x71, 0xd8, 0x31, 0x15 +.byte 0x04, 0xc7, 0x23, 0xc3, 0x18, 0x96, 0x05, 0x9a +.byte 0x07, 0x12, 0x80, 0xe2, 0xeb, 0x27, 0xb2, 0x75 +.byte 0x09, 0x83, 0x2c, 0x1a, 0x1b, 0x6e, 0x5a, 0xa0 +.byte 0x52, 0x3b, 0xd6, 0xb3, 0x29, 0xe3, 0x2f, 0x84 +.byte 0x53, 0xd1, 0x00, 0xed, 0x20, 0xfc, 0xb1, 0x5b +.byte 0x6a, 0xcb, 0xbe, 0x39, 0x4a, 0x4c, 0x58, 0xcf +.byte 0xd0, 0xef, 0xaa, 0xfb, 0x43, 0x4d, 0x33, 0x85 +.byte 0x45, 0xf9, 0x02, 0x7f, 0x50, 0x3c, 0x9f, 0xa8 +.byte 0x51, 0xa3, 0x40, 0x8f, 0x92, 0x9d, 0x38, 0xf5 +.byte 0xbc, 0xb6, 0xda, 0x21, 0x10, 0xff, 0xf3, 0xd2 +.byte 0xcd, 0x0c, 0x13, 0xec, 0x5f, 0x97, 0x44, 0x17 +.byte 0xc4, 0xa7, 0x7e, 0x3d, 0x64, 0x5d, 0x19, 0x73 +.byte 0x60, 0x81, 0x4f, 0xdc, 0x22, 0x2a, 0x90, 0x88 +.byte 0x46, 0xee, 0xb8, 0x14, 0xde, 0x5e, 0x0b, 0xdb +.byte 0xe0, 0x32, 0x3a, 0x0a, 0x49, 0x06, 0x24, 0x5c +.byte 0xc2, 0xd3, 0xac, 0x62, 0x91, 0x95, 0xe4, 0x79 +.byte 0xe7, 0xc8, 0x37, 0x6d, 0x8d, 0xd5, 0x4e, 0xa9 +.byte 0x6c, 0x56, 0xf4, 0xea, 0x65, 0x7a, 0xae, 0x08 +.byte 0xba, 0x78, 0x25, 0x2e, 0x1c, 0xa6, 0xb4, 0xc6 +.byte 0xe8, 0xdd, 0x74, 0x1f, 0x4b, 0xbd, 0x8b, 0x8a +.byte 0x70, 0x3e, 0xb5, 0x66, 0x48, 0x03, 0xf6, 0x0e +.byte 0x61, 0x35, 0x57, 0xb9, 0x86, 0xc1, 0x1d, 0x9e +.byte 0xe1, 0xf8, 0x98, 0x11, 0x69, 0xd9, 0x8e, 0x94 +.byte 0x9b, 0x1e, 0x87, 0xe9, 0xce, 0x55, 0x28, 0xdf +.byte 0x8c, 0xa1, 0x89, 0x0d, 0xbf, 0xe6, 0x42, 0x68 +.byte 0x41, 0x99, 0x2d, 0x0f, 0xb0, 0x54, 0xbb, 0x16 +# rcon[] +.long 0x01000000, 0x02000000, 0x04000000, 0x08000000 +.long 0x10000000, 0x20000000, 0x40000000, 0x80000000 +.long 0x1B000000, 0x36000000, 0, 0, 0, 0, 0, 0 +.align 256 +.size AES_Te,.-AES_Te + +# void AES_encrypt(const unsigned char *inp, unsigned char *out, +# const AES_KEY *key) { +.globl AES_encrypt +.type AES_encrypt,\@function +AES_encrypt: +___ +$code.=<<___ if (!$softonly); + l %r0,240($key) + lhi %r1,16 + clr %r0,%r1 + jl .Lesoft + + la %r1,0($key) + #la %r2,0($inp) + la %r4,0($out) + lghi %r3,16 # single block length + .long 0xb92e0042 # km %r4,%r2 + brc 1,.-4 # can this happen? + br %r14 +.align 64 +.Lesoft: +___ +$code.=<<___; + stm${g} %r3,$ra,3*$SIZE_T($sp) + + llgf $s0,0($inp) + llgf $s1,4($inp) + llgf $s2,8($inp) + llgf $s3,12($inp) + + larl $tbl,AES_Te + bras $ra,_s390x_AES_encrypt + + l${g} $out,3*$SIZE_T($sp) + st $s0,0($out) + st $s1,4($out) + st $s2,8($out) + st $s3,12($out) + + lm${g} %r6,$ra,6*$SIZE_T($sp) + br $ra +.size AES_encrypt,.-AES_encrypt + +.type _s390x_AES_encrypt,\@function +.align 16 +_s390x_AES_encrypt: + st${g} $ra,15*$SIZE_T($sp) + x $s0,0($key) + x $s1,4($key) + x $s2,8($key) + x $s3,12($key) + l $rounds,240($key) + llill $mask,`0xff<<3` + aghi $rounds,-1 + j .Lenc_loop +.align 16 +.Lenc_loop: + sllg $t1,$s0,`0+3` + srlg $t2,$s0,`8-3` + srlg $t3,$s0,`16-3` + srl $s0,`24-3` + nr $s0,$mask + ngr $t1,$mask + nr $t2,$mask + nr $t3,$mask + + srlg $i1,$s1,`16-3` # i0 + sllg $i2,$s1,`0+3` + srlg $i3,$s1,`8-3` + srl $s1,`24-3` + nr $i1,$mask + nr $s1,$mask + ngr $i2,$mask + nr $i3,$mask + + l $s0,0($s0,$tbl) # Te0[s0>>24] + l $t1,1($t1,$tbl) # Te3[s0>>0] + l $t2,2($t2,$tbl) # Te2[s0>>8] + l $t3,3($t3,$tbl) # Te1[s0>>16] + + x $s0,3($i1,$tbl) # Te1[s1>>16] + l $s1,0($s1,$tbl) # Te0[s1>>24] + x $t2,1($i2,$tbl) # Te3[s1>>0] + x $t3,2($i3,$tbl) # Te2[s1>>8] + + srlg $i1,$s2,`8-3` # i0 + srlg $i2,$s2,`16-3` # i1 + nr $i1,$mask + nr $i2,$mask + sllg $i3,$s2,`0+3` + srl $s2,`24-3` + nr $s2,$mask + ngr $i3,$mask + + xr $s1,$t1 + srlg $ra,$s3,`8-3` # i1 + sllg $t1,$s3,`0+3` # i0 + nr $ra,$mask + la $key,16($key) + ngr $t1,$mask + + x $s0,2($i1,$tbl) # Te2[s2>>8] + x $s1,3($i2,$tbl) # Te1[s2>>16] + l $s2,0($s2,$tbl) # Te0[s2>>24] + x $t3,1($i3,$tbl) # Te3[s2>>0] + + srlg $i3,$s3,`16-3` # i2 + xr $s2,$t2 + srl $s3,`24-3` + nr $i3,$mask + nr $s3,$mask + + x $s0,0($key) + x $s1,4($key) + x $s2,8($key) + x $t3,12($key) + + x $s0,1($t1,$tbl) # Te3[s3>>0] + x $s1,2($ra,$tbl) # Te2[s3>>8] + x $s2,3($i3,$tbl) # Te1[s3>>16] + l $s3,0($s3,$tbl) # Te0[s3>>24] + xr $s3,$t3 + + brct $rounds,.Lenc_loop + .align 16 + + sllg $t1,$s0,`0+3` + srlg $t2,$s0,`8-3` + ngr $t1,$mask + srlg $t3,$s0,`16-3` + srl $s0,`24-3` + nr $s0,$mask + nr $t2,$mask + nr $t3,$mask + + srlg $i1,$s1,`16-3` # i0 + sllg $i2,$s1,`0+3` + ngr $i2,$mask + srlg $i3,$s1,`8-3` + srl $s1,`24-3` + nr $i1,$mask + nr $s1,$mask + nr $i3,$mask + + llgc $s0,2($s0,$tbl) # Te4[s0>>24] + llgc $t1,2($t1,$tbl) # Te4[s0>>0] + sll $s0,24 + llgc $t2,2($t2,$tbl) # Te4[s0>>8] + llgc $t3,2($t3,$tbl) # Te4[s0>>16] + sll $t2,8 + sll $t3,16 + + llgc $i1,2($i1,$tbl) # Te4[s1>>16] + llgc $s1,2($s1,$tbl) # Te4[s1>>24] + llgc $i2,2($i2,$tbl) # Te4[s1>>0] + llgc $i3,2($i3,$tbl) # Te4[s1>>8] + sll $i1,16 + sll $s1,24 + sll $i3,8 + or $s0,$i1 + or $s1,$t1 + or $t2,$i2 + or $t3,$i3 + + srlg $i1,$s2,`8-3` # i0 + srlg $i2,$s2,`16-3` # i1 + nr $i1,$mask + nr $i2,$mask + sllg $i3,$s2,`0+3` + srl $s2,`24-3` + ngr $i3,$mask + nr $s2,$mask + + sllg $t1,$s3,`0+3` # i0 + srlg $ra,$s3,`8-3` # i1 + ngr $t1,$mask + + llgc $i1,2($i1,$tbl) # Te4[s2>>8] + llgc $i2,2($i2,$tbl) # Te4[s2>>16] + sll $i1,8 + llgc $s2,2($s2,$tbl) # Te4[s2>>24] + llgc $i3,2($i3,$tbl) # Te4[s2>>0] + sll $i2,16 + nr $ra,$mask + sll $s2,24 + or $s0,$i1 + or $s1,$i2 + or $s2,$t2 + or $t3,$i3 + + srlg $i3,$s3,`16-3` # i2 + srl $s3,`24-3` + nr $i3,$mask + nr $s3,$mask + + l $t0,16($key) + l $t2,20($key) + + llgc $i1,2($t1,$tbl) # Te4[s3>>0] + llgc $i2,2($ra,$tbl) # Te4[s3>>8] + llgc $i3,2($i3,$tbl) # Te4[s3>>16] + llgc $s3,2($s3,$tbl) # Te4[s3>>24] + sll $i2,8 + sll $i3,16 + sll $s3,24 + or $s0,$i1 + or $s1,$i2 + or $s2,$i3 + or $s3,$t3 + + l${g} $ra,15*$SIZE_T($sp) + xr $s0,$t0 + xr $s1,$t2 + x $s2,24($key) + x $s3,28($key) + + br $ra +.size _s390x_AES_encrypt,.-_s390x_AES_encrypt +___ + +$code.=<<___; +.type AES_Td,\@object +.align 256 +AES_Td: +___ +&_data_word( + 0x51f4a750, 0x7e416553, 0x1a17a4c3, 0x3a275e96, + 0x3bab6bcb, 0x1f9d45f1, 0xacfa58ab, 0x4be30393, + 0x2030fa55, 0xad766df6, 0x88cc7691, 0xf5024c25, + 0x4fe5d7fc, 0xc52acbd7, 0x26354480, 0xb562a38f, + 0xdeb15a49, 0x25ba1b67, 0x45ea0e98, 0x5dfec0e1, + 0xc32f7502, 0x814cf012, 0x8d4697a3, 0x6bd3f9c6, + 0x038f5fe7, 0x15929c95, 0xbf6d7aeb, 0x955259da, + 0xd4be832d, 0x587421d3, 0x49e06929, 0x8ec9c844, + 0x75c2896a, 0xf48e7978, 0x99583e6b, 0x27b971dd, + 0xbee14fb6, 0xf088ad17, 0xc920ac66, 0x7dce3ab4, + 0x63df4a18, 0xe51a3182, 0x97513360, 0x62537f45, + 0xb16477e0, 0xbb6bae84, 0xfe81a01c, 0xf9082b94, + 0x70486858, 0x8f45fd19, 0x94de6c87, 0x527bf8b7, + 0xab73d323, 0x724b02e2, 0xe31f8f57, 0x6655ab2a, + 0xb2eb2807, 0x2fb5c203, 0x86c57b9a, 0xd33708a5, + 0x302887f2, 0x23bfa5b2, 0x02036aba, 0xed16825c, + 0x8acf1c2b, 0xa779b492, 0xf307f2f0, 0x4e69e2a1, + 0x65daf4cd, 0x0605bed5, 0xd134621f, 0xc4a6fe8a, + 0x342e539d, 0xa2f355a0, 0x058ae132, 0xa4f6eb75, + 0x0b83ec39, 0x4060efaa, 0x5e719f06, 0xbd6e1051, + 0x3e218af9, 0x96dd063d, 0xdd3e05ae, 0x4de6bd46, + 0x91548db5, 0x71c45d05, 0x0406d46f, 0x605015ff, + 0x1998fb24, 0xd6bde997, 0x894043cc, 0x67d99e77, + 0xb0e842bd, 0x07898b88, 0xe7195b38, 0x79c8eedb, + 0xa17c0a47, 0x7c420fe9, 0xf8841ec9, 0x00000000, + 0x09808683, 0x322bed48, 0x1e1170ac, 0x6c5a724e, + 0xfd0efffb, 0x0f853856, 0x3daed51e, 0x362d3927, + 0x0a0fd964, 0x685ca621, 0x9b5b54d1, 0x24362e3a, + 0x0c0a67b1, 0x9357e70f, 0xb4ee96d2, 0x1b9b919e, + 0x80c0c54f, 0x61dc20a2, 0x5a774b69, 0x1c121a16, + 0xe293ba0a, 0xc0a02ae5, 0x3c22e043, 0x121b171d, + 0x0e090d0b, 0xf28bc7ad, 0x2db6a8b9, 0x141ea9c8, + 0x57f11985, 0xaf75074c, 0xee99ddbb, 0xa37f60fd, + 0xf701269f, 0x5c72f5bc, 0x44663bc5, 0x5bfb7e34, + 0x8b432976, 0xcb23c6dc, 0xb6edfc68, 0xb8e4f163, + 0xd731dcca, 0x42638510, 0x13972240, 0x84c61120, + 0x854a247d, 0xd2bb3df8, 0xaef93211, 0xc729a16d, + 0x1d9e2f4b, 0xdcb230f3, 0x0d8652ec, 0x77c1e3d0, + 0x2bb3166c, 0xa970b999, 0x119448fa, 0x47e96422, + 0xa8fc8cc4, 0xa0f03f1a, 0x567d2cd8, 0x223390ef, + 0x87494ec7, 0xd938d1c1, 0x8ccaa2fe, 0x98d40b36, + 0xa6f581cf, 0xa57ade28, 0xdab78e26, 0x3fadbfa4, + 0x2c3a9de4, 0x5078920d, 0x6a5fcc9b, 0x547e4662, + 0xf68d13c2, 0x90d8b8e8, 0x2e39f75e, 0x82c3aff5, + 0x9f5d80be, 0x69d0937c, 0x6fd52da9, 0xcf2512b3, + 0xc8ac993b, 0x10187da7, 0xe89c636e, 0xdb3bbb7b, + 0xcd267809, 0x6e5918f4, 0xec9ab701, 0x834f9aa8, + 0xe6956e65, 0xaaffe67e, 0x21bccf08, 0xef15e8e6, + 0xbae79bd9, 0x4a6f36ce, 0xea9f09d4, 0x29b07cd6, + 0x31a4b2af, 0x2a3f2331, 0xc6a59430, 0x35a266c0, + 0x744ebc37, 0xfc82caa6, 0xe090d0b0, 0x33a7d815, + 0xf104984a, 0x41ecdaf7, 0x7fcd500e, 0x1791f62f, + 0x764dd68d, 0x43efb04d, 0xccaa4d54, 0xe49604df, + 0x9ed1b5e3, 0x4c6a881b, 0xc12c1fb8, 0x4665517f, + 0x9d5eea04, 0x018c355d, 0xfa877473, 0xfb0b412e, + 0xb3671d5a, 0x92dbd252, 0xe9105633, 0x6dd64713, + 0x9ad7618c, 0x37a10c7a, 0x59f8148e, 0xeb133c89, + 0xcea927ee, 0xb761c935, 0xe11ce5ed, 0x7a47b13c, + 0x9cd2df59, 0x55f2733f, 0x1814ce79, 0x73c737bf, + 0x53f7cdea, 0x5ffdaa5b, 0xdf3d6f14, 0x7844db86, + 0xcaaff381, 0xb968c43e, 0x3824342c, 0xc2a3405f, + 0x161dc372, 0xbce2250c, 0x283c498b, 0xff0d9541, + 0x39a80171, 0x080cb3de, 0xd8b4e49c, 0x6456c190, + 0x7bcb8461, 0xd532b670, 0x486c5c74, 0xd0b85742); +$code.=<<___; +# Td4[256] +.byte 0x52, 0x09, 0x6a, 0xd5, 0x30, 0x36, 0xa5, 0x38 +.byte 0xbf, 0x40, 0xa3, 0x9e, 0x81, 0xf3, 0xd7, 0xfb +.byte 0x7c, 0xe3, 0x39, 0x82, 0x9b, 0x2f, 0xff, 0x87 +.byte 0x34, 0x8e, 0x43, 0x44, 0xc4, 0xde, 0xe9, 0xcb +.byte 0x54, 0x7b, 0x94, 0x32, 0xa6, 0xc2, 0x23, 0x3d +.byte 0xee, 0x4c, 0x95, 0x0b, 0x42, 0xfa, 0xc3, 0x4e +.byte 0x08, 0x2e, 0xa1, 0x66, 0x28, 0xd9, 0x24, 0xb2 +.byte 0x76, 0x5b, 0xa2, 0x49, 0x6d, 0x8b, 0xd1, 0x25 +.byte 0x72, 0xf8, 0xf6, 0x64, 0x86, 0x68, 0x98, 0x16 +.byte 0xd4, 0xa4, 0x5c, 0xcc, 0x5d, 0x65, 0xb6, 0x92 +.byte 0x6c, 0x70, 0x48, 0x50, 0xfd, 0xed, 0xb9, 0xda +.byte 0x5e, 0x15, 0x46, 0x57, 0xa7, 0x8d, 0x9d, 0x84 +.byte 0x90, 0xd8, 0xab, 0x00, 0x8c, 0xbc, 0xd3, 0x0a +.byte 0xf7, 0xe4, 0x58, 0x05, 0xb8, 0xb3, 0x45, 0x06 +.byte 0xd0, 0x2c, 0x1e, 0x8f, 0xca, 0x3f, 0x0f, 0x02 +.byte 0xc1, 0xaf, 0xbd, 0x03, 0x01, 0x13, 0x8a, 0x6b +.byte 0x3a, 0x91, 0x11, 0x41, 0x4f, 0x67, 0xdc, 0xea +.byte 0x97, 0xf2, 0xcf, 0xce, 0xf0, 0xb4, 0xe6, 0x73 +.byte 0x96, 0xac, 0x74, 0x22, 0xe7, 0xad, 0x35, 0x85 +.byte 0xe2, 0xf9, 0x37, 0xe8, 0x1c, 0x75, 0xdf, 0x6e +.byte 0x47, 0xf1, 0x1a, 0x71, 0x1d, 0x29, 0xc5, 0x89 +.byte 0x6f, 0xb7, 0x62, 0x0e, 0xaa, 0x18, 0xbe, 0x1b +.byte 0xfc, 0x56, 0x3e, 0x4b, 0xc6, 0xd2, 0x79, 0x20 +.byte 0x9a, 0xdb, 0xc0, 0xfe, 0x78, 0xcd, 0x5a, 0xf4 +.byte 0x1f, 0xdd, 0xa8, 0x33, 0x88, 0x07, 0xc7, 0x31 +.byte 0xb1, 0x12, 0x10, 0x59, 0x27, 0x80, 0xec, 0x5f +.byte 0x60, 0x51, 0x7f, 0xa9, 0x19, 0xb5, 0x4a, 0x0d +.byte 0x2d, 0xe5, 0x7a, 0x9f, 0x93, 0xc9, 0x9c, 0xef +.byte 0xa0, 0xe0, 0x3b, 0x4d, 0xae, 0x2a, 0xf5, 0xb0 +.byte 0xc8, 0xeb, 0xbb, 0x3c, 0x83, 0x53, 0x99, 0x61 +.byte 0x17, 0x2b, 0x04, 0x7e, 0xba, 0x77, 0xd6, 0x26 +.byte 0xe1, 0x69, 0x14, 0x63, 0x55, 0x21, 0x0c, 0x7d +.size AES_Td,.-AES_Td + +# void AES_decrypt(const unsigned char *inp, unsigned char *out, +# const AES_KEY *key) { +.globl AES_decrypt +.type AES_decrypt,\@function +AES_decrypt: +___ +$code.=<<___ if (!$softonly); + l %r0,240($key) + lhi %r1,16 + clr %r0,%r1 + jl .Ldsoft + + la %r1,0($key) + #la %r2,0($inp) + la %r4,0($out) + lghi %r3,16 # single block length + .long 0xb92e0042 # km %r4,%r2 + brc 1,.-4 # can this happen? + br %r14 +.align 64 +.Ldsoft: +___ +$code.=<<___; + stm${g} %r3,$ra,3*$SIZE_T($sp) + + llgf $s0,0($inp) + llgf $s1,4($inp) + llgf $s2,8($inp) + llgf $s3,12($inp) + + larl $tbl,AES_Td + bras $ra,_s390x_AES_decrypt + + l${g} $out,3*$SIZE_T($sp) + st $s0,0($out) + st $s1,4($out) + st $s2,8($out) + st $s3,12($out) + + lm${g} %r6,$ra,6*$SIZE_T($sp) + br $ra +.size AES_decrypt,.-AES_decrypt + +.type _s390x_AES_decrypt,\@function +.align 16 +_s390x_AES_decrypt: + st${g} $ra,15*$SIZE_T($sp) + x $s0,0($key) + x $s1,4($key) + x $s2,8($key) + x $s3,12($key) + l $rounds,240($key) + llill $mask,`0xff<<3` + aghi $rounds,-1 + j .Ldec_loop +.align 16 +.Ldec_loop: + srlg $t1,$s0,`16-3` + srlg $t2,$s0,`8-3` + sllg $t3,$s0,`0+3` + srl $s0,`24-3` + nr $s0,$mask + nr $t1,$mask + nr $t2,$mask + ngr $t3,$mask + + sllg $i1,$s1,`0+3` # i0 + srlg $i2,$s1,`16-3` + srlg $i3,$s1,`8-3` + srl $s1,`24-3` + ngr $i1,$mask + nr $s1,$mask + nr $i2,$mask + nr $i3,$mask + + l $s0,0($s0,$tbl) # Td0[s0>>24] + l $t1,3($t1,$tbl) # Td1[s0>>16] + l $t2,2($t2,$tbl) # Td2[s0>>8] + l $t3,1($t3,$tbl) # Td3[s0>>0] + + x $s0,1($i1,$tbl) # Td3[s1>>0] + l $s1,0($s1,$tbl) # Td0[s1>>24] + x $t2,3($i2,$tbl) # Td1[s1>>16] + x $t3,2($i3,$tbl) # Td2[s1>>8] + + srlg $i1,$s2,`8-3` # i0 + sllg $i2,$s2,`0+3` # i1 + srlg $i3,$s2,`16-3` + srl $s2,`24-3` + nr $i1,$mask + ngr $i2,$mask + nr $s2,$mask + nr $i3,$mask + + xr $s1,$t1 + srlg $ra,$s3,`8-3` # i1 + srlg $t1,$s3,`16-3` # i0 + nr $ra,$mask + la $key,16($key) + nr $t1,$mask + + x $s0,2($i1,$tbl) # Td2[s2>>8] + x $s1,1($i2,$tbl) # Td3[s2>>0] + l $s2,0($s2,$tbl) # Td0[s2>>24] + x $t3,3($i3,$tbl) # Td1[s2>>16] + + sllg $i3,$s3,`0+3` # i2 + srl $s3,`24-3` + ngr $i3,$mask + nr $s3,$mask + + xr $s2,$t2 + x $s0,0($key) + x $s1,4($key) + x $s2,8($key) + x $t3,12($key) + + x $s0,3($t1,$tbl) # Td1[s3>>16] + x $s1,2($ra,$tbl) # Td2[s3>>8] + x $s2,1($i3,$tbl) # Td3[s3>>0] + l $s3,0($s3,$tbl) # Td0[s3>>24] + xr $s3,$t3 + + brct $rounds,.Ldec_loop + .align 16 + + l $t1,`2048+0`($tbl) # prefetch Td4 + l $t2,`2048+64`($tbl) + l $t3,`2048+128`($tbl) + l $i1,`2048+192`($tbl) + llill $mask,0xff + + srlg $i3,$s0,24 # i0 + srlg $t1,$s0,16 + srlg $t2,$s0,8 + nr $s0,$mask # i3 + nr $t1,$mask + + srlg $i1,$s1,24 + nr $t2,$mask + srlg $i2,$s1,16 + srlg $ra,$s1,8 + nr $s1,$mask # i0 + nr $i2,$mask + nr $ra,$mask + + llgc $i3,2048($i3,$tbl) # Td4[s0>>24] + llgc $t1,2048($t1,$tbl) # Td4[s0>>16] + llgc $t2,2048($t2,$tbl) # Td4[s0>>8] + sll $t1,16 + llgc $t3,2048($s0,$tbl) # Td4[s0>>0] + sllg $s0,$i3,24 + sll $t2,8 + + llgc $s1,2048($s1,$tbl) # Td4[s1>>0] + llgc $i1,2048($i1,$tbl) # Td4[s1>>24] + llgc $i2,2048($i2,$tbl) # Td4[s1>>16] + sll $i1,24 + llgc $i3,2048($ra,$tbl) # Td4[s1>>8] + sll $i2,16 + sll $i3,8 + or $s0,$s1 + or $t1,$i1 + or $t2,$i2 + or $t3,$i3 + + srlg $i1,$s2,8 # i0 + srlg $i2,$s2,24 + srlg $i3,$s2,16 + nr $s2,$mask # i1 + nr $i1,$mask + nr $i3,$mask + llgc $i1,2048($i1,$tbl) # Td4[s2>>8] + llgc $s1,2048($s2,$tbl) # Td4[s2>>0] + llgc $i2,2048($i2,$tbl) # Td4[s2>>24] + llgc $i3,2048($i3,$tbl) # Td4[s2>>16] + sll $i1,8 + sll $i2,24 + or $s0,$i1 + sll $i3,16 + or $t2,$i2 + or $t3,$i3 + + srlg $i1,$s3,16 # i0 + srlg $i2,$s3,8 # i1 + srlg $i3,$s3,24 + nr $s3,$mask # i2 + nr $i1,$mask + nr $i2,$mask + + l${g} $ra,15*$SIZE_T($sp) + or $s1,$t1 + l $t0,16($key) + l $t1,20($key) + + llgc $i1,2048($i1,$tbl) # Td4[s3>>16] + llgc $i2,2048($i2,$tbl) # Td4[s3>>8] + sll $i1,16 + llgc $s2,2048($s3,$tbl) # Td4[s3>>0] + llgc $s3,2048($i3,$tbl) # Td4[s3>>24] + sll $i2,8 + sll $s3,24 + or $s0,$i1 + or $s1,$i2 + or $s2,$t2 + or $s3,$t3 + + xr $s0,$t0 + xr $s1,$t1 + x $s2,24($key) + x $s3,28($key) + + br $ra +.size _s390x_AES_decrypt,.-_s390x_AES_decrypt +___ + +$code.=<<___; +# void AES_set_encrypt_key(const unsigned char *in, int bits, +# AES_KEY *key) { +.globl AES_set_encrypt_key +.type AES_set_encrypt_key,\@function +.align 16 +AES_set_encrypt_key: +_s390x_AES_set_encrypt_key: + lghi $t0,0 + cl${g}r $inp,$t0 + je .Lminus1 + cl${g}r $key,$t0 + je .Lminus1 + + lghi $t0,128 + clr $bits,$t0 + je .Lproceed + lghi $t0,192 + clr $bits,$t0 + je .Lproceed + lghi $t0,256 + clr $bits,$t0 + je .Lproceed + lghi %r2,-2 + br %r14 + +.align 16 +.Lproceed: +___ +$code.=<<___ if (!$softonly); + # convert bits to km(c) code, [128,192,256]->[18,19,20] + lhi %r5,-128 + lhi %r0,18 + ar %r5,$bits + srl %r5,6 + ar %r5,%r0 + + larl %r1,OPENSSL_s390xcap_P + llihh %r0,0x8000 + srlg %r0,%r0,0(%r5) + ng %r0,32(%r1) # check availability of both km... + ng %r0,48(%r1) # ...and kmc support for given key length + jz .Lekey_internal + + lmg %r0,%r1,0($inp) # just copy 128 bits... + stmg %r0,%r1,0($key) + lhi %r0,192 + cr $bits,%r0 + jl 1f + lg %r1,16($inp) + stg %r1,16($key) + je 1f + lg %r1,24($inp) + stg %r1,24($key) +1: st $bits,236($key) # save bits [for debugging purposes] + lgr $t0,%r5 + st %r5,240($key) # save km(c) code + lghi %r2,0 + br %r14 +___ +$code.=<<___; +.align 16 +.Lekey_internal: + stm${g} %r4,%r13,4*$SIZE_T($sp) # all non-volatile regs and $key + + larl $tbl,AES_Te+2048 + + llgf $s0,0($inp) + llgf $s1,4($inp) + llgf $s2,8($inp) + llgf $s3,12($inp) + st $s0,0($key) + st $s1,4($key) + st $s2,8($key) + st $s3,12($key) + lghi $t0,128 + cr $bits,$t0 + jne .Lnot128 + + llill $mask,0xff + lghi $t3,0 # i=0 + lghi $rounds,10 + st $rounds,240($key) + + llgfr $t2,$s3 # temp=rk[3] + srlg $i1,$s3,8 + srlg $i2,$s3,16 + srlg $i3,$s3,24 + nr $t2,$mask + nr $i1,$mask + nr $i2,$mask + +.align 16 +.L128_loop: + la $t2,0($t2,$tbl) + la $i1,0($i1,$tbl) + la $i2,0($i2,$tbl) + la $i3,0($i3,$tbl) + icm $t2,2,0($t2) # Te4[rk[3]>>0]<<8 + icm $t2,4,0($i1) # Te4[rk[3]>>8]<<16 + icm $t2,8,0($i2) # Te4[rk[3]>>16]<<24 + icm $t2,1,0($i3) # Te4[rk[3]>>24] + x $t2,256($t3,$tbl) # rcon[i] + xr $s0,$t2 # rk[4]=rk[0]^... + xr $s1,$s0 # rk[5]=rk[1]^rk[4] + xr $s2,$s1 # rk[6]=rk[2]^rk[5] + xr $s3,$s2 # rk[7]=rk[3]^rk[6] + + llgfr $t2,$s3 # temp=rk[3] + srlg $i1,$s3,8 + srlg $i2,$s3,16 + nr $t2,$mask + nr $i1,$mask + srlg $i3,$s3,24 + nr $i2,$mask + + st $s0,16($key) + st $s1,20($key) + st $s2,24($key) + st $s3,28($key) + la $key,16($key) # key+=4 + la $t3,4($t3) # i++ + brct $rounds,.L128_loop + lghi $t0,10 + lghi %r2,0 + lm${g} %r4,%r13,4*$SIZE_T($sp) + br $ra + +.align 16 +.Lnot128: + llgf $t0,16($inp) + llgf $t1,20($inp) + st $t0,16($key) + st $t1,20($key) + lghi $t0,192 + cr $bits,$t0 + jne .Lnot192 + + llill $mask,0xff + lghi $t3,0 # i=0 + lghi $rounds,12 + st $rounds,240($key) + lghi $rounds,8 + + srlg $i1,$t1,8 + srlg $i2,$t1,16 + srlg $i3,$t1,24 + nr $t1,$mask + nr $i1,$mask + nr $i2,$mask + +.align 16 +.L192_loop: + la $t1,0($t1,$tbl) + la $i1,0($i1,$tbl) + la $i2,0($i2,$tbl) + la $i3,0($i3,$tbl) + icm $t1,2,0($t1) # Te4[rk[5]>>0]<<8 + icm $t1,4,0($i1) # Te4[rk[5]>>8]<<16 + icm $t1,8,0($i2) # Te4[rk[5]>>16]<<24 + icm $t1,1,0($i3) # Te4[rk[5]>>24] + x $t1,256($t3,$tbl) # rcon[i] + xr $s0,$t1 # rk[6]=rk[0]^... + xr $s1,$s0 # rk[7]=rk[1]^rk[6] + xr $s2,$s1 # rk[8]=rk[2]^rk[7] + xr $s3,$s2 # rk[9]=rk[3]^rk[8] + + st $s0,24($key) + st $s1,28($key) + st $s2,32($key) + st $s3,36($key) + brct $rounds,.L192_continue + lghi $t0,12 + lghi %r2,0 + lm${g} %r4,%r13,4*$SIZE_T($sp) + br $ra + +.align 16 +.L192_continue: + lgr $t1,$s3 + x $t1,16($key) # rk[10]=rk[4]^rk[9] + st $t1,40($key) + x $t1,20($key) # rk[11]=rk[5]^rk[10] + st $t1,44($key) + + srlg $i1,$t1,8 + srlg $i2,$t1,16 + srlg $i3,$t1,24 + nr $t1,$mask + nr $i1,$mask + nr $i2,$mask + + la $key,24($key) # key+=6 + la $t3,4($t3) # i++ + j .L192_loop + +.align 16 +.Lnot192: + llgf $t0,24($inp) + llgf $t1,28($inp) + st $t0,24($key) + st $t1,28($key) + llill $mask,0xff + lghi $t3,0 # i=0 + lghi $rounds,14 + st $rounds,240($key) + lghi $rounds,7 + + srlg $i1,$t1,8 + srlg $i2,$t1,16 + srlg $i3,$t1,24 + nr $t1,$mask + nr $i1,$mask + nr $i2,$mask + +.align 16 +.L256_loop: + la $t1,0($t1,$tbl) + la $i1,0($i1,$tbl) + la $i2,0($i2,$tbl) + la $i3,0($i3,$tbl) + icm $t1,2,0($t1) # Te4[rk[7]>>0]<<8 + icm $t1,4,0($i1) # Te4[rk[7]>>8]<<16 + icm $t1,8,0($i2) # Te4[rk[7]>>16]<<24 + icm $t1,1,0($i3) # Te4[rk[7]>>24] + x $t1,256($t3,$tbl) # rcon[i] + xr $s0,$t1 # rk[8]=rk[0]^... + xr $s1,$s0 # rk[9]=rk[1]^rk[8] + xr $s2,$s1 # rk[10]=rk[2]^rk[9] + xr $s3,$s2 # rk[11]=rk[3]^rk[10] + st $s0,32($key) + st $s1,36($key) + st $s2,40($key) + st $s3,44($key) + brct $rounds,.L256_continue + lghi $t0,14 + lghi %r2,0 + lm${g} %r4,%r13,4*$SIZE_T($sp) + br $ra + +.align 16 +.L256_continue: + lgr $t1,$s3 # temp=rk[11] + srlg $i1,$s3,8 + srlg $i2,$s3,16 + srlg $i3,$s3,24 + nr $t1,$mask + nr $i1,$mask + nr $i2,$mask + la $t1,0($t1,$tbl) + la $i1,0($i1,$tbl) + la $i2,0($i2,$tbl) + la $i3,0($i3,$tbl) + llgc $t1,0($t1) # Te4[rk[11]>>0] + icm $t1,2,0($i1) # Te4[rk[11]>>8]<<8 + icm $t1,4,0($i2) # Te4[rk[11]>>16]<<16 + icm $t1,8,0($i3) # Te4[rk[11]>>24]<<24 + x $t1,16($key) # rk[12]=rk[4]^... + st $t1,48($key) + x $t1,20($key) # rk[13]=rk[5]^rk[12] + st $t1,52($key) + x $t1,24($key) # rk[14]=rk[6]^rk[13] + st $t1,56($key) + x $t1,28($key) # rk[15]=rk[7]^rk[14] + st $t1,60($key) + + srlg $i1,$t1,8 + srlg $i2,$t1,16 + srlg $i3,$t1,24 + nr $t1,$mask + nr $i1,$mask + nr $i2,$mask + + la $key,32($key) # key+=8 + la $t3,4($t3) # i++ + j .L256_loop + +.Lminus1: + lghi %r2,-1 + br $ra +.size AES_set_encrypt_key,.-AES_set_encrypt_key + +# void AES_set_decrypt_key(const unsigned char *in, int bits, +# AES_KEY *key) { +.globl AES_set_decrypt_key +.type AES_set_decrypt_key,\@function +.align 16 +AES_set_decrypt_key: + #st${g} $key,4*$SIZE_T($sp) # I rely on AES_set_encrypt_key to + st${g} $ra,14*$SIZE_T($sp) # save non-volatile registers and $key! + bras $ra,_s390x_AES_set_encrypt_key + #l${g} $key,4*$SIZE_T($sp) + l${g} $ra,14*$SIZE_T($sp) + ltgr %r2,%r2 + bnzr $ra +___ +$code.=<<___ if (!$softonly); + #l $t0,240($key) + lhi $t1,16 + cr $t0,$t1 + jl .Lgo + oill $t0,0x80 # set "decrypt" bit + st $t0,240($key) + br $ra +___ +$code.=<<___; +.align 16 +.Lgo: lgr $rounds,$t0 #llgf $rounds,240($key) + la $i1,0($key) + sllg $i2,$rounds,4 + la $i2,0($i2,$key) + srl $rounds,1 + lghi $t1,-16 + +.align 16 +.Linv: lmg $s0,$s1,0($i1) + lmg $s2,$s3,0($i2) + stmg $s0,$s1,0($i2) + stmg $s2,$s3,0($i1) + la $i1,16($i1) + la $i2,0($t1,$i2) + brct $rounds,.Linv +___ +$mask80=$i1; +$mask1b=$i2; +$maskfe=$i3; +$code.=<<___; + llgf $rounds,240($key) + aghi $rounds,-1 + sll $rounds,2 # (rounds-1)*4 + llilh $mask80,0x8080 + llilh $mask1b,0x1b1b + llilh $maskfe,0xfefe + oill $mask80,0x8080 + oill $mask1b,0x1b1b + oill $maskfe,0xfefe + +.align 16 +.Lmix: l $s0,16($key) # tp1 + lr $s1,$s0 + ngr $s1,$mask80 + srlg $t1,$s1,7 + slr $s1,$t1 + nr $s1,$mask1b + sllg $t1,$s0,1 + nr $t1,$maskfe + xr $s1,$t1 # tp2 + + lr $s2,$s1 + ngr $s2,$mask80 + srlg $t1,$s2,7 + slr $s2,$t1 + nr $s2,$mask1b + sllg $t1,$s1,1 + nr $t1,$maskfe + xr $s2,$t1 # tp4 + + lr $s3,$s2 + ngr $s3,$mask80 + srlg $t1,$s3,7 + slr $s3,$t1 + nr $s3,$mask1b + sllg $t1,$s2,1 + nr $t1,$maskfe + xr $s3,$t1 # tp8 + + xr $s1,$s0 # tp2^tp1 + xr $s2,$s0 # tp4^tp1 + rll $s0,$s0,24 # = ROTATE(tp1,8) + xr $s2,$s3 # ^=tp8 + xr $s0,$s1 # ^=tp2^tp1 + xr $s1,$s3 # tp2^tp1^tp8 + xr $s0,$s2 # ^=tp4^tp1^tp8 + rll $s1,$s1,8 + rll $s2,$s2,16 + xr $s0,$s1 # ^= ROTATE(tp8^tp2^tp1,24) + rll $s3,$s3,24 + xr $s0,$s2 # ^= ROTATE(tp8^tp4^tp1,16) + xr $s0,$s3 # ^= ROTATE(tp8,8) + + st $s0,16($key) + la $key,4($key) + brct $rounds,.Lmix + + lm${g} %r6,%r13,6*$SIZE_T($sp)# as was saved by AES_set_encrypt_key! + lghi %r2,0 + br $ra +.size AES_set_decrypt_key,.-AES_set_decrypt_key +___ + +######################################################################## +# void AES_cbc_encrypt(const unsigned char *in, unsigned char *out, +# size_t length, const AES_KEY *key, +# unsigned char *ivec, const int enc) +{ +my $inp="%r2"; +my $out="%r4"; # length and out are swapped +my $len="%r3"; +my $key="%r5"; +my $ivp="%r6"; + +$code.=<<___; +.globl AES_cbc_encrypt +.type AES_cbc_encrypt,\@function +.align 16 +AES_cbc_encrypt: + xgr %r3,%r4 # flip %r3 and %r4, out and len + xgr %r4,%r3 + xgr %r3,%r4 +___ +$code.=<<___ if (!$softonly); + lhi %r0,16 + cl %r0,240($key) + jh .Lcbc_software + + lg %r0,0($ivp) # copy ivec + lg %r1,8($ivp) + stmg %r0,%r1,16($sp) + lmg %r0,%r1,0($key) # copy key, cover 256 bit + stmg %r0,%r1,32($sp) + lmg %r0,%r1,16($key) + stmg %r0,%r1,48($sp) + l %r0,240($key) # load kmc code + lghi $key,15 # res=len%16, len-=res; + ngr $key,$len + sl${g}r $len,$key + la %r1,16($sp) # parameter block - ivec || key + jz .Lkmc_truncated + .long 0xb92f0042 # kmc %r4,%r2 + brc 1,.-4 # pay attention to "partial completion" + ltr $key,$key + jnz .Lkmc_truncated +.Lkmc_done: + lmg %r0,%r1,16($sp) # copy ivec to caller + stg %r0,0($ivp) + stg %r1,8($ivp) + br $ra +.align 16 +.Lkmc_truncated: + ahi $key,-1 # it's the way it's encoded in mvc + tmll %r0,0x80 + jnz .Lkmc_truncated_dec + lghi %r1,0 + stg %r1,16*$SIZE_T($sp) + stg %r1,16*$SIZE_T+8($sp) + bras %r1,1f + mvc 16*$SIZE_T(1,$sp),0($inp) +1: ex $key,0(%r1) + la %r1,16($sp) # restore parameter block + la $inp,16*$SIZE_T($sp) + lghi $len,16 + .long 0xb92f0042 # kmc %r4,%r2 + j .Lkmc_done +.align 16 +.Lkmc_truncated_dec: + st${g} $out,4*$SIZE_T($sp) + la $out,16*$SIZE_T($sp) + lghi $len,16 + .long 0xb92f0042 # kmc %r4,%r2 + l${g} $out,4*$SIZE_T($sp) + bras %r1,2f + mvc 0(1,$out),16*$SIZE_T($sp) +2: ex $key,0(%r1) + j .Lkmc_done +.align 16 +.Lcbc_software: +___ +$code.=<<___; + stm${g} $key,$ra,5*$SIZE_T($sp) + lhi %r0,0 + cl %r0,`$stdframe+$SIZE_T-4`($sp) + je .Lcbc_decrypt + + larl $tbl,AES_Te + + llgf $s0,0($ivp) + llgf $s1,4($ivp) + llgf $s2,8($ivp) + llgf $s3,12($ivp) + + lghi $t0,16 + sl${g}r $len,$t0 + brc 4,.Lcbc_enc_tail # if borrow +.Lcbc_enc_loop: + stm${g} $inp,$out,2*$SIZE_T($sp) + x $s0,0($inp) + x $s1,4($inp) + x $s2,8($inp) + x $s3,12($inp) + lgr %r4,$key + + bras $ra,_s390x_AES_encrypt + + lm${g} $inp,$key,2*$SIZE_T($sp) + st $s0,0($out) + st $s1,4($out) + st $s2,8($out) + st $s3,12($out) + + la $inp,16($inp) + la $out,16($out) + lghi $t0,16 + lt${g}r $len,$len + jz .Lcbc_enc_done + sl${g}r $len,$t0 + brc 4,.Lcbc_enc_tail # if borrow + j .Lcbc_enc_loop +.align 16 +.Lcbc_enc_done: + l${g} $ivp,6*$SIZE_T($sp) + st $s0,0($ivp) + st $s1,4($ivp) + st $s2,8($ivp) + st $s3,12($ivp) + + lm${g} %r7,$ra,7*$SIZE_T($sp) + br $ra + +.align 16 +.Lcbc_enc_tail: + aghi $len,15 + lghi $t0,0 + stg $t0,16*$SIZE_T($sp) + stg $t0,16*$SIZE_T+8($sp) + bras $t1,3f + mvc 16*$SIZE_T(1,$sp),0($inp) +3: ex $len,0($t1) + lghi $len,0 + la $inp,16*$SIZE_T($sp) + j .Lcbc_enc_loop + +.align 16 +.Lcbc_decrypt: + larl $tbl,AES_Td + + lg $t0,0($ivp) + lg $t1,8($ivp) + stmg $t0,$t1,16*$SIZE_T($sp) + +.Lcbc_dec_loop: + stm${g} $inp,$out,2*$SIZE_T($sp) + llgf $s0,0($inp) + llgf $s1,4($inp) + llgf $s2,8($inp) + llgf $s3,12($inp) + lgr %r4,$key + + bras $ra,_s390x_AES_decrypt + + lm${g} $inp,$key,2*$SIZE_T($sp) + sllg $s0,$s0,32 + sllg $s2,$s2,32 + lr $s0,$s1 + lr $s2,$s3 + + lg $t0,0($inp) + lg $t1,8($inp) + xg $s0,16*$SIZE_T($sp) + xg $s2,16*$SIZE_T+8($sp) + lghi $s1,16 + sl${g}r $len,$s1 + brc 4,.Lcbc_dec_tail # if borrow + brc 2,.Lcbc_dec_done # if zero + stg $s0,0($out) + stg $s2,8($out) + stmg $t0,$t1,16*$SIZE_T($sp) + + la $inp,16($inp) + la $out,16($out) + j .Lcbc_dec_loop + +.Lcbc_dec_done: + stg $s0,0($out) + stg $s2,8($out) +.Lcbc_dec_exit: + lm${g} %r6,$ra,6*$SIZE_T($sp) + stmg $t0,$t1,0($ivp) + + br $ra + +.align 16 +.Lcbc_dec_tail: + aghi $len,15 + stg $s0,16*$SIZE_T($sp) + stg $s2,16*$SIZE_T+8($sp) + bras $s1,4f + mvc 0(1,$out),16*$SIZE_T($sp) +4: ex $len,0($s1) + j .Lcbc_dec_exit +.size AES_cbc_encrypt,.-AES_cbc_encrypt +___ +} +######################################################################## +# void AES_ctr32_encrypt(const unsigned char *in, unsigned char *out, +# size_t blocks, const AES_KEY *key, +# const unsigned char *ivec) +{ +my $inp="%r2"; +my $out="%r4"; # blocks and out are swapped +my $len="%r3"; +my $key="%r5"; my $iv0="%r5"; +my $ivp="%r6"; +my $fp ="%r7"; + +$code.=<<___; +.globl AES_ctr32_encrypt +.type AES_ctr32_encrypt,\@function +.align 16 +AES_ctr32_encrypt: + xgr %r3,%r4 # flip %r3 and %r4, $out and $len + xgr %r4,%r3 + xgr %r3,%r4 + llgfr $len,$len # safe in ctr32 subroutine even in 64-bit case +___ +$code.=<<___ if (!$softonly); + l %r0,240($key) + lhi %r1,16 + clr %r0,%r1 + jl .Lctr32_software + + stm${g} %r6,$s3,6*$SIZE_T($sp) + + slgr $out,$inp + la %r1,0($key) # %r1 is permanent copy of $key + lg $iv0,0($ivp) # load ivec + lg $ivp,8($ivp) + + # prepare and allocate stack frame at the top of 4K page + # with 1K reserved for eventual signal handling + lghi $s0,-1024-256-16# guarantee at least 256-bytes buffer + lghi $s1,-4096 + algr $s0,$sp + lgr $fp,$sp + ngr $s0,$s1 # align at page boundary + slgr $fp,$s0 # total buffer size + lgr $s2,$sp + lghi $s1,1024+16 # sl[g]fi is extended-immediate facility + slgr $fp,$s1 # deduct reservation to get usable buffer size + # buffer size is at lest 256 and at most 3072+256-16 + + la $sp,1024($s0) # alloca + srlg $fp,$fp,4 # convert bytes to blocks, minimum 16 + st${g} $s2,0($sp) # back-chain + st${g} $fp,$SIZE_T($sp) + + slgr $len,$fp + brc 1,.Lctr32_hw_switch # not zero, no borrow + algr $fp,$len # input is shorter than allocated buffer + lghi $len,0 + st${g} $fp,$SIZE_T($sp) + +.Lctr32_hw_switch: +___ +$code.=<<___ if (!$softonly && 0);# kmctr code was measured to be ~12% slower + llgfr $s0,%r0 + lgr $s1,%r1 + larl %r1,OPENSSL_s390xcap_P + llihh %r0,0x8000 # check if kmctr supports the function code + srlg %r0,%r0,0($s0) + ng %r0,64(%r1) # check kmctr capability vector + lgr %r0,$s0 + lgr %r1,$s1 + jz .Lctr32_km_loop + +####### kmctr code + algr $out,$inp # restore $out + lgr $s1,$len # $s1 undertakes $len + j .Lctr32_kmctr_loop +.align 16 +.Lctr32_kmctr_loop: + la $s2,16($sp) + lgr $s3,$fp +.Lctr32_kmctr_prepare: + stg $iv0,0($s2) + stg $ivp,8($s2) + la $s2,16($s2) + ahi $ivp,1 # 32-bit increment, preserves upper half + brct $s3,.Lctr32_kmctr_prepare + + #la $inp,0($inp) # inp + sllg $len,$fp,4 # len + #la $out,0($out) # out + la $s2,16($sp) # iv + .long 0xb92da042 # kmctr $out,$s2,$inp + brc 1,.-4 # pay attention to "partial completion" + + slgr $s1,$fp + brc 1,.Lctr32_kmctr_loop # not zero, no borrow + algr $fp,$s1 + lghi $s1,0 + brc 4+1,.Lctr32_kmctr_loop # not zero + + l${g} $sp,0($sp) + lm${g} %r6,$s3,6*$SIZE_T($sp) + br $ra +.align 16 +___ +$code.=<<___ if (!$softonly); +.Lctr32_km_loop: + la $s2,16($sp) + lgr $s3,$fp +.Lctr32_km_prepare: + stg $iv0,0($s2) + stg $ivp,8($s2) + la $s2,16($s2) + ahi $ivp,1 # 32-bit increment, preserves upper half + brct $s3,.Lctr32_km_prepare + + la $s0,16($sp) # inp + sllg $s1,$fp,4 # len + la $s2,16($sp) # out + .long 0xb92e00a8 # km %r10,%r8 + brc 1,.-4 # pay attention to "partial completion" + + la $s2,16($sp) + lgr $s3,$fp + slgr $s2,$inp +.Lctr32_km_xor: + lg $s0,0($inp) + lg $s1,8($inp) + xg $s0,0($s2,$inp) + xg $s1,8($s2,$inp) + stg $s0,0($out,$inp) + stg $s1,8($out,$inp) + la $inp,16($inp) + brct $s3,.Lctr32_km_xor + + slgr $len,$fp + brc 1,.Lctr32_km_loop # not zero, no borrow + algr $fp,$len + lghi $len,0 + brc 4+1,.Lctr32_km_loop # not zero + + l${g} $s0,0($sp) + l${g} $s1,$SIZE_T($sp) + la $s2,16($sp) +.Lctr32_km_zap: + stg $s0,0($s2) + stg $s0,8($s2) + la $s2,16($s2) + brct $s1,.Lctr32_km_zap + + la $sp,0($s0) + lm${g} %r6,$s3,6*$SIZE_T($sp) + br $ra +.align 16 +.Lctr32_software: +___ +$code.=<<___; + stm${g} $key,$ra,5*$SIZE_T($sp) + sl${g}r $inp,$out + larl $tbl,AES_Te + llgf $t1,12($ivp) + +.Lctr32_loop: + stm${g} $inp,$out,2*$SIZE_T($sp) + llgf $s0,0($ivp) + llgf $s1,4($ivp) + llgf $s2,8($ivp) + lgr $s3,$t1 + st $t1,16*$SIZE_T($sp) + lgr %r4,$key + + bras $ra,_s390x_AES_encrypt + + lm${g} $inp,$ivp,2*$SIZE_T($sp) + llgf $t1,16*$SIZE_T($sp) + x $s0,0($inp,$out) + x $s1,4($inp,$out) + x $s2,8($inp,$out) + x $s3,12($inp,$out) + stm $s0,$s3,0($out) + + la $out,16($out) + ahi $t1,1 # 32-bit increment + brct $len,.Lctr32_loop + + lm${g} %r6,$ra,6*$SIZE_T($sp) + br $ra +.size AES_ctr32_encrypt,.-AES_ctr32_encrypt +___ +} + +######################################################################## +# void AES_xts_encrypt(const char *inp,char *out,size_t len, +# const AES_KEY *key1, const AES_KEY *key2, +# const unsigned char iv[16]); +# +{ +my $inp="%r2"; +my $out="%r4"; # len and out are swapped +my $len="%r3"; +my $key1="%r5"; # $i1 +my $key2="%r6"; # $i2 +my $fp="%r7"; # $i3 +my $tweak=16*$SIZE_T+16; # or $stdframe-16, bottom of the frame... + +$code.=<<___; +.type _s390x_xts_km,\@function +.align 16 +_s390x_xts_km: +___ +$code.=<<___ if(1); + llgfr $s0,%r0 # put aside the function code + lghi $s1,0x7f + nr $s1,%r0 + larl %r1,OPENSSL_s390xcap_P + llihh %r0,0x8000 + srlg %r0,%r0,32($s1) # check for 32+function code + ng %r0,32(%r1) # check km capability vector + lgr %r0,$s0 # restore the function code + la %r1,0($key1) # restore $key1 + jz .Lxts_km_vanilla + + lmg $i2,$i3,$tweak($sp) # put aside the tweak value + algr $out,$inp + + oill %r0,32 # switch to xts function code + aghi $s1,-18 # + sllg $s1,$s1,3 # (function code - 18)*8, 0 or 16 + la %r1,$tweak-16($sp) + slgr %r1,$s1 # parameter block position + lmg $s0,$s3,0($key1) # load 256 bits of key material, + stmg $s0,$s3,0(%r1) # and copy it to parameter block. + # yes, it contains junk and overlaps + # with the tweak in 128-bit case. + # it's done to avoid conditional + # branch. + stmg $i2,$i3,$tweak($sp) # "re-seat" the tweak value + + .long 0xb92e0042 # km %r4,%r2 + brc 1,.-4 # pay attention to "partial completion" + + lrvg $s0,$tweak+0($sp) # load the last tweak + lrvg $s1,$tweak+8($sp) + stmg %r0,%r3,$tweak-32($sp) # wipe copy of the key + + nill %r0,0xffdf # switch back to original function code + la %r1,0($key1) # restore pointer to $key1 + slgr $out,$inp + + llgc $len,2*$SIZE_T-1($sp) + nill $len,0x0f # $len%=16 + br $ra + +.align 16 +.Lxts_km_vanilla: +___ +$code.=<<___; + # prepare and allocate stack frame at the top of 4K page + # with 1K reserved for eventual signal handling + lghi $s0,-1024-256-16# guarantee at least 256-bytes buffer + lghi $s1,-4096 + algr $s0,$sp + lgr $fp,$sp + ngr $s0,$s1 # align at page boundary + slgr $fp,$s0 # total buffer size + lgr $s2,$sp + lghi $s1,1024+16 # sl[g]fi is extended-immediate facility + slgr $fp,$s1 # deduct reservation to get usable buffer size + # buffer size is at lest 256 and at most 3072+256-16 + + la $sp,1024($s0) # alloca + nill $fp,0xfff0 # round to 16*n + st${g} $s2,0($sp) # back-chain + nill $len,0xfff0 # redundant + st${g} $fp,$SIZE_T($sp) + + slgr $len,$fp + brc 1,.Lxts_km_go # not zero, no borrow + algr $fp,$len # input is shorter than allocated buffer + lghi $len,0 + st${g} $fp,$SIZE_T($sp) + +.Lxts_km_go: + lrvg $s0,$tweak+0($s2) # load the tweak value in little-endian + lrvg $s1,$tweak+8($s2) + + la $s2,16($sp) # vector of ascending tweak values + slgr $s2,$inp + srlg $s3,$fp,4 + j .Lxts_km_start + +.Lxts_km_loop: + la $s2,16($sp) + slgr $s2,$inp + srlg $s3,$fp,4 +.Lxts_km_prepare: + lghi $i1,0x87 + srag $i2,$s1,63 # broadcast upper bit + ngr $i1,$i2 # rem + algr $s0,$s0 + alcgr $s1,$s1 + xgr $s0,$i1 +.Lxts_km_start: + lrvgr $i1,$s0 # flip byte order + lrvgr $i2,$s1 + stg $i1,0($s2,$inp) + stg $i2,8($s2,$inp) + xg $i1,0($inp) + xg $i2,8($inp) + stg $i1,0($out,$inp) + stg $i2,8($out,$inp) + la $inp,16($inp) + brct $s3,.Lxts_km_prepare + + slgr $inp,$fp # rewind $inp + la $s2,0($out,$inp) + lgr $s3,$fp + .long 0xb92e00aa # km $s2,$s2 + brc 1,.-4 # pay attention to "partial completion" + + la $s2,16($sp) + slgr $s2,$inp + srlg $s3,$fp,4 +.Lxts_km_xor: + lg $i1,0($out,$inp) + lg $i2,8($out,$inp) + xg $i1,0($s2,$inp) + xg $i2,8($s2,$inp) + stg $i1,0($out,$inp) + stg $i2,8($out,$inp) + la $inp,16($inp) + brct $s3,.Lxts_km_xor + + slgr $len,$fp + brc 1,.Lxts_km_loop # not zero, no borrow + algr $fp,$len + lghi $len,0 + brc 4+1,.Lxts_km_loop # not zero + + l${g} $i1,0($sp) # back-chain + llgf $fp,`2*$SIZE_T-4`($sp) # bytes used + la $i2,16($sp) + srlg $fp,$fp,4 +.Lxts_km_zap: + stg $i1,0($i2) + stg $i1,8($i2) + la $i2,16($i2) + brct $fp,.Lxts_km_zap + + la $sp,0($i1) + llgc $len,2*$SIZE_T-1($i1) + nill $len,0x0f # $len%=16 + bzr $ra + + # generate one more tweak... + lghi $i1,0x87 + srag $i2,$s1,63 # broadcast upper bit + ngr $i1,$i2 # rem + algr $s0,$s0 + alcgr $s1,$s1 + xgr $s0,$i1 + + ltr $len,$len # clear zero flag + br $ra +.size _s390x_xts_km,.-_s390x_xts_km + +.globl AES_xts_encrypt +.type AES_xts_encrypt,\@function +.align 16 +AES_xts_encrypt: + xgr %r3,%r4 # flip %r3 and %r4, $out and $len + xgr %r4,%r3 + xgr %r3,%r4 +___ +$code.=<<___ if ($SIZE_T==4); + llgfr $len,$len +___ +$code.=<<___; + st${g} $len,1*$SIZE_T($sp) # save copy of $len + srag $len,$len,4 # formally wrong, because it expands + # sign byte, but who can afford asking + # to process more than 2^63-1 bytes? + # I use it, because it sets condition + # code... + bcr 8,$ra # abort if zero (i.e. less than 16) +___ +$code.=<<___ if (!$softonly); + llgf %r0,240($key2) + lhi %r1,16 + clr %r0,%r1 + jl .Lxts_enc_software + + st${g} $ra,5*$SIZE_T($sp) + stm${g} %r6,$s3,6*$SIZE_T($sp) + + sllg $len,$len,4 # $len&=~15 + slgr $out,$inp + + # generate the tweak value + l${g} $s3,$stdframe($sp) # pointer to iv + la $s2,$tweak($sp) + lmg $s0,$s1,0($s3) + lghi $s3,16 + stmg $s0,$s1,0($s2) + la %r1,0($key2) # $key2 is not needed anymore + .long 0xb92e00aa # km $s2,$s2, generate the tweak + brc 1,.-4 # can this happen? + + l %r0,240($key1) + la %r1,0($key1) # $key1 is not needed anymore + bras $ra,_s390x_xts_km + jz .Lxts_enc_km_done + + aghi $inp,-16 # take one step back + la $i3,0($out,$inp) # put aside real $out +.Lxts_enc_km_steal: + llgc $i1,16($inp) + llgc $i2,0($out,$inp) + stc $i1,0($out,$inp) + stc $i2,16($out,$inp) + la $inp,1($inp) + brct $len,.Lxts_enc_km_steal + + la $s2,0($i3) + lghi $s3,16 + lrvgr $i1,$s0 # flip byte order + lrvgr $i2,$s1 + xg $i1,0($s2) + xg $i2,8($s2) + stg $i1,0($s2) + stg $i2,8($s2) + .long 0xb92e00aa # km $s2,$s2 + brc 1,.-4 # can this happen? + lrvgr $i1,$s0 # flip byte order + lrvgr $i2,$s1 + xg $i1,0($i3) + xg $i2,8($i3) + stg $i1,0($i3) + stg $i2,8($i3) + +.Lxts_enc_km_done: + stg $sp,$tweak+0($sp) # wipe tweak + stg $sp,$tweak+8($sp) + l${g} $ra,5*$SIZE_T($sp) + lm${g} %r6,$s3,6*$SIZE_T($sp) + br $ra +.align 16 +.Lxts_enc_software: +___ +$code.=<<___; + stm${g} %r6,$ra,6*$SIZE_T($sp) + + slgr $out,$inp + + l${g} $s3,$stdframe($sp) # ivp + llgf $s0,0($s3) # load iv + llgf $s1,4($s3) + llgf $s2,8($s3) + llgf $s3,12($s3) + stm${g} %r2,%r5,2*$SIZE_T($sp) + la $key,0($key2) + larl $tbl,AES_Te + bras $ra,_s390x_AES_encrypt # generate the tweak + lm${g} %r2,%r5,2*$SIZE_T($sp) + stm $s0,$s3,$tweak($sp) # save the tweak + j .Lxts_enc_enter + +.align 16 +.Lxts_enc_loop: + lrvg $s1,$tweak+0($sp) # load the tweak in little-endian + lrvg $s3,$tweak+8($sp) + lghi %r1,0x87 + srag %r0,$s3,63 # broadcast upper bit + ngr %r1,%r0 # rem + algr $s1,$s1 + alcgr $s3,$s3 + xgr $s1,%r1 + lrvgr $s1,$s1 # flip byte order + lrvgr $s3,$s3 + srlg $s0,$s1,32 # smash the tweak to 4x32-bits + stg $s1,$tweak+0($sp) # save the tweak + llgfr $s1,$s1 + srlg $s2,$s3,32 + stg $s3,$tweak+8($sp) + llgfr $s3,$s3 + la $inp,16($inp) # $inp+=16 +.Lxts_enc_enter: + x $s0,0($inp) # ^=*($inp) + x $s1,4($inp) + x $s2,8($inp) + x $s3,12($inp) + stm${g} %r2,%r3,2*$SIZE_T($sp) # only two registers are changing + la $key,0($key1) + bras $ra,_s390x_AES_encrypt + lm${g} %r2,%r5,2*$SIZE_T($sp) + x $s0,$tweak+0($sp) # ^=tweak + x $s1,$tweak+4($sp) + x $s2,$tweak+8($sp) + x $s3,$tweak+12($sp) + st $s0,0($out,$inp) + st $s1,4($out,$inp) + st $s2,8($out,$inp) + st $s3,12($out,$inp) + brct${g} $len,.Lxts_enc_loop + + llgc $len,`2*$SIZE_T-1`($sp) + nill $len,0x0f # $len%16 + jz .Lxts_enc_done + + la $i3,0($inp,$out) # put aside real $out +.Lxts_enc_steal: + llgc %r0,16($inp) + llgc %r1,0($out,$inp) + stc %r0,0($out,$inp) + stc %r1,16($out,$inp) + la $inp,1($inp) + brct $len,.Lxts_enc_steal + la $out,0($i3) # restore real $out + + # generate last tweak... + lrvg $s1,$tweak+0($sp) # load the tweak in little-endian + lrvg $s3,$tweak+8($sp) + lghi %r1,0x87 + srag %r0,$s3,63 # broadcast upper bit + ngr %r1,%r0 # rem + algr $s1,$s1 + alcgr $s3,$s3 + xgr $s1,%r1 + lrvgr $s1,$s1 # flip byte order + lrvgr $s3,$s3 + srlg $s0,$s1,32 # smash the tweak to 4x32-bits + stg $s1,$tweak+0($sp) # save the tweak + llgfr $s1,$s1 + srlg $s2,$s3,32 + stg $s3,$tweak+8($sp) + llgfr $s3,$s3 + + x $s0,0($out) # ^=*(inp)|stolen cipther-text + x $s1,4($out) + x $s2,8($out) + x $s3,12($out) + st${g} $out,4*$SIZE_T($sp) + la $key,0($key1) + bras $ra,_s390x_AES_encrypt + l${g} $out,4*$SIZE_T($sp) + x $s0,`$tweak+0`($sp) # ^=tweak + x $s1,`$tweak+4`($sp) + x $s2,`$tweak+8`($sp) + x $s3,`$tweak+12`($sp) + st $s0,0($out) + st $s1,4($out) + st $s2,8($out) + st $s3,12($out) + +.Lxts_enc_done: + stg $sp,$tweak+0($sp) # wipe tweak + stg $sp,$twesk+8($sp) + lm${g} %r6,$ra,6*$SIZE_T($sp) + br $ra +.size AES_xts_encrypt,.-AES_xts_encrypt +___ +# void AES_xts_decrypt(const char *inp,char *out,size_t len, +# const AES_KEY *key1, const AES_KEY *key2, +# const unsigned char iv[16]); +# +$code.=<<___; +.globl AES_xts_decrypt +.type AES_xts_decrypt,\@function +.align 16 +AES_xts_decrypt: + xgr %r3,%r4 # flip %r3 and %r4, $out and $len + xgr %r4,%r3 + xgr %r3,%r4 +___ +$code.=<<___ if ($SIZE_T==4); + llgfr $len,$len +___ +$code.=<<___; + st${g} $len,1*$SIZE_T($sp) # save copy of $len + aghi $len,-16 + bcr 4,$ra # abort if less than zero. formally + # wrong, because $len is unsigned, + # but who can afford asking to + # process more than 2^63-1 bytes? + tmll $len,0x0f + jnz .Lxts_dec_proceed + aghi $len,16 +.Lxts_dec_proceed: +___ +$code.=<<___ if (!$softonly); + llgf %r0,240($key2) + lhi %r1,16 + clr %r0,%r1 + jl .Lxts_dec_software + + st${g} $ra,5*$SIZE_T($sp) + stm${g} %r6,$s3,6*$SIZE_T($sp) + + nill $len,0xfff0 # $len&=~15 + slgr $out,$inp + + # generate the tweak value + l${g} $s3,$stdframe($sp) # pointer to iv + la $s2,$tweak($sp) + lmg $s0,$s1,0($s3) + lghi $s3,16 + stmg $s0,$s1,0($s2) + la %r1,0($key2) # $key2 is not needed past this point + .long 0xb92e00aa # km $s2,$s2, generate the tweak + brc 1,.-4 # can this happen? + + l %r0,240($key1) + la %r1,0($key1) # $key1 is not needed anymore + + ltgr $len,$len + jz .Lxts_dec_km_short + bras $ra,_s390x_xts_km + jz .Lxts_dec_km_done + + lrvgr $s2,$s0 # make copy in reverse byte order + lrvgr $s3,$s1 + j .Lxts_dec_km_2ndtweak + +.Lxts_dec_km_short: + llgc $len,`2*$SIZE_T-1`($sp) + nill $len,0x0f # $len%=16 + lrvg $s0,$tweak+0($sp) # load the tweak + lrvg $s1,$tweak+8($sp) + lrvgr $s2,$s0 # make copy in reverse byte order + lrvgr $s3,$s1 + +.Lxts_dec_km_2ndtweak: + lghi $i1,0x87 + srag $i2,$s1,63 # broadcast upper bit + ngr $i1,$i2 # rem + algr $s0,$s0 + alcgr $s1,$s1 + xgr $s0,$i1 + lrvgr $i1,$s0 # flip byte order + lrvgr $i2,$s1 + + xg $i1,0($inp) + xg $i2,8($inp) + stg $i1,0($out,$inp) + stg $i2,8($out,$inp) + la $i2,0($out,$inp) + lghi $i3,16 + .long 0xb92e0066 # km $i2,$i2 + brc 1,.-4 # can this happen? + lrvgr $i1,$s0 + lrvgr $i2,$s1 + xg $i1,0($out,$inp) + xg $i2,8($out,$inp) + stg $i1,0($out,$inp) + stg $i2,8($out,$inp) + + la $i3,0($out,$inp) # put aside real $out +.Lxts_dec_km_steal: + llgc $i1,16($inp) + llgc $i2,0($out,$inp) + stc $i1,0($out,$inp) + stc $i2,16($out,$inp) + la $inp,1($inp) + brct $len,.Lxts_dec_km_steal + + lgr $s0,$s2 + lgr $s1,$s3 + xg $s0,0($i3) + xg $s1,8($i3) + stg $s0,0($i3) + stg $s1,8($i3) + la $s0,0($i3) + lghi $s1,16 + .long 0xb92e0088 # km $s0,$s0 + brc 1,.-4 # can this happen? + xg $s2,0($i3) + xg $s3,8($i3) + stg $s2,0($i3) + stg $s3,8($i3) +.Lxts_dec_km_done: + stg $sp,$tweak+0($sp) # wipe tweak + stg $sp,$tweak+8($sp) + l${g} $ra,5*$SIZE_T($sp) + lm${g} %r6,$s3,6*$SIZE_T($sp) + br $ra +.align 16 +.Lxts_dec_software: +___ +$code.=<<___; + stm${g} %r6,$ra,6*$SIZE_T($sp) + + srlg $len,$len,4 + slgr $out,$inp + + l${g} $s3,$stdframe($sp) # ivp + llgf $s0,0($s3) # load iv + llgf $s1,4($s3) + llgf $s2,8($s3) + llgf $s3,12($s3) + stm${g} %r2,%r5,2*$SIZE_T($sp) + la $key,0($key2) + larl $tbl,AES_Te + bras $ra,_s390x_AES_encrypt # generate the tweak + lm${g} %r2,%r5,2*$SIZE_T($sp) + larl $tbl,AES_Td + lt${g}r $len,$len + stm $s0,$s3,$tweak($sp) # save the tweak + jz .Lxts_dec_short + j .Lxts_dec_enter + +.align 16 +.Lxts_dec_loop: + lrvg $s1,$tweak+0($sp) # load the tweak in little-endian + lrvg $s3,$tweak+8($sp) + lghi %r1,0x87 + srag %r0,$s3,63 # broadcast upper bit + ngr %r1,%r0 # rem + algr $s1,$s1 + alcgr $s3,$s3 + xgr $s1,%r1 + lrvgr $s1,$s1 # flip byte order + lrvgr $s3,$s3 + srlg $s0,$s1,32 # smash the tweak to 4x32-bits + stg $s1,$tweak+0($sp) # save the tweak + llgfr $s1,$s1 + srlg $s2,$s3,32 + stg $s3,$tweak+8($sp) + llgfr $s3,$s3 +.Lxts_dec_enter: + x $s0,0($inp) # tweak^=*(inp) + x $s1,4($inp) + x $s2,8($inp) + x $s3,12($inp) + stm${g} %r2,%r3,2*$SIZE_T($sp) # only two registers are changing + la $key,0($key1) + bras $ra,_s390x_AES_decrypt + lm${g} %r2,%r5,2*$SIZE_T($sp) + x $s0,$tweak+0($sp) # ^=tweak + x $s1,$tweak+4($sp) + x $s2,$tweak+8($sp) + x $s3,$tweak+12($sp) + st $s0,0($out,$inp) + st $s1,4($out,$inp) + st $s2,8($out,$inp) + st $s3,12($out,$inp) + la $inp,16($inp) + brct${g} $len,.Lxts_dec_loop + + llgc $len,`2*$SIZE_T-1`($sp) + nill $len,0x0f # $len%16 + jz .Lxts_dec_done + + # generate pair of tweaks... + lrvg $s1,$tweak+0($sp) # load the tweak in little-endian + lrvg $s3,$tweak+8($sp) + lghi %r1,0x87 + srag %r0,$s3,63 # broadcast upper bit + ngr %r1,%r0 # rem + algr $s1,$s1 + alcgr $s3,$s3 + xgr $s1,%r1 + lrvgr $i2,$s1 # flip byte order + lrvgr $i3,$s3 + stmg $i2,$i3,$tweak($sp) # save the 1st tweak + j .Lxts_dec_2ndtweak + +.align 16 +.Lxts_dec_short: + llgc $len,`2*$SIZE_T-1`($sp) + nill $len,0x0f # $len%16 + lrvg $s1,$tweak+0($sp) # load the tweak in little-endian + lrvg $s3,$tweak+8($sp) +.Lxts_dec_2ndtweak: + lghi %r1,0x87 + srag %r0,$s3,63 # broadcast upper bit + ngr %r1,%r0 # rem + algr $s1,$s1 + alcgr $s3,$s3 + xgr $s1,%r1 + lrvgr $s1,$s1 # flip byte order + lrvgr $s3,$s3 + srlg $s0,$s1,32 # smash the tweak to 4x32-bits + stg $s1,$tweak-16+0($sp) # save the 2nd tweak + llgfr $s1,$s1 + srlg $s2,$s3,32 + stg $s3,$tweak-16+8($sp) + llgfr $s3,$s3 + + x $s0,0($inp) # tweak_the_2nd^=*(inp) + x $s1,4($inp) + x $s2,8($inp) + x $s3,12($inp) + stm${g} %r2,%r3,2*$SIZE_T($sp) + la $key,0($key1) + bras $ra,_s390x_AES_decrypt + lm${g} %r2,%r5,2*$SIZE_T($sp) + x $s0,$tweak-16+0($sp) # ^=tweak_the_2nd + x $s1,$tweak-16+4($sp) + x $s2,$tweak-16+8($sp) + x $s3,$tweak-16+12($sp) + st $s0,0($out,$inp) + st $s1,4($out,$inp) + st $s2,8($out,$inp) + st $s3,12($out,$inp) + + la $i3,0($out,$inp) # put aside real $out +.Lxts_dec_steal: + llgc %r0,16($inp) + llgc %r1,0($out,$inp) + stc %r0,0($out,$inp) + stc %r1,16($out,$inp) + la $inp,1($inp) + brct $len,.Lxts_dec_steal + la $out,0($i3) # restore real $out + + lm $s0,$s3,$tweak($sp) # load the 1st tweak + x $s0,0($out) # tweak^=*(inp)|stolen cipher-text + x $s1,4($out) + x $s2,8($out) + x $s3,12($out) + st${g} $out,4*$SIZE_T($sp) + la $key,0($key1) + bras $ra,_s390x_AES_decrypt + l${g} $out,4*$SIZE_T($sp) + x $s0,$tweak+0($sp) # ^=tweak + x $s1,$tweak+4($sp) + x $s2,$tweak+8($sp) + x $s3,$tweak+12($sp) + st $s0,0($out) + st $s1,4($out) + st $s2,8($out) + st $s3,12($out) + stg $sp,$tweak-16+0($sp) # wipe 2nd tweak + stg $sp,$tweak-16+8($sp) +.Lxts_dec_done: + stg $sp,$tweak+0($sp) # wipe tweak + stg $sp,$twesk+8($sp) + lm${g} %r6,$ra,6*$SIZE_T($sp) + br $ra +.size AES_xts_decrypt,.-AES_xts_decrypt +___ +} +$code.=<<___; +.string "AES for s390x, CRYPTOGAMS by " +___ + +$code =~ s/\`([^\`]*)\`/eval $1/gem; +print $code; +close STDOUT; # force flush diff --git a/openssl-1.1.0h/crypto/aes/asm/aes-sparcv9.pl b/openssl-1.1.0h/crypto/aes/asm/aes-sparcv9.pl new file mode 100755 index 0000000..883fae8 --- /dev/null +++ b/openssl-1.1.0h/crypto/aes/asm/aes-sparcv9.pl @@ -0,0 +1,1192 @@ +#! /usr/bin/env perl +# Copyright 2005-2016 The OpenSSL Project Authors. All Rights Reserved. +# +# Licensed under the OpenSSL license (the "License"). You may not use +# this file except in compliance with the License. You can obtain a copy +# in the file LICENSE in the source distribution or at +# https://www.openssl.org/source/license.html + +# +# ==================================================================== +# Written by Andy Polyakov for the OpenSSL +# project. Rights for redistribution and usage in source and binary +# forms are granted according to the OpenSSL license. +# ==================================================================== +# +# Version 1.1 +# +# The major reason for undertaken effort was to mitigate the hazard of +# cache-timing attack. This is [currently and initially!] addressed in +# two ways. 1. S-boxes are compressed from 5KB to 2KB+256B size each. +# 2. References to them are scheduled for L2 cache latency, meaning +# that the tables don't have to reside in L1 cache. Once again, this +# is an initial draft and one should expect more countermeasures to +# be implemented... +# +# Version 1.1 prefetches T[ed]4 in order to mitigate attack on last +# round. +# +# Even though performance was not the primary goal [on the contrary, +# extra shifts "induced" by compressed S-box and longer loop epilogue +# "induced" by scheduling for L2 have negative effect on performance], +# the code turned out to run in ~23 cycles per processed byte en-/ +# decrypted with 128-bit key. This is pretty good result for code +# with mentioned qualities and UltraSPARC core. Compared to Sun C +# generated code my encrypt procedure runs just few percents faster, +# while decrypt one - whole 50% faster [yes, Sun C failed to generate +# optimal decrypt procedure]. Compared to GNU C generated code both +# procedures are more than 60% faster:-) + +$output = pop; +open STDOUT,">$output"; + +$frame="STACK_FRAME"; +$bias="STACK_BIAS"; +$locals=16; + +$acc0="%l0"; +$acc1="%o0"; +$acc2="%o1"; +$acc3="%o2"; + +$acc4="%l1"; +$acc5="%o3"; +$acc6="%o4"; +$acc7="%o5"; + +$acc8="%l2"; +$acc9="%o7"; +$acc10="%g1"; +$acc11="%g2"; + +$acc12="%l3"; +$acc13="%g3"; +$acc14="%g4"; +$acc15="%g5"; + +$t0="%l4"; +$t1="%l5"; +$t2="%l6"; +$t3="%l7"; + +$s0="%i0"; +$s1="%i1"; +$s2="%i2"; +$s3="%i3"; +$tbl="%i4"; +$key="%i5"; +$rounds="%i7"; # aliases with return address, which is off-loaded to stack + +sub _data_word() +{ my $i; + while(defined($i=shift)) { $code.=sprintf"\t.long\t0x%08x,0x%08x\n",$i,$i; } +} + +$code.=<<___; +#include "sparc_arch.h" + +#ifdef __arch64__ +.register %g2,#scratch +.register %g3,#scratch +#endif +.section ".text",#alloc,#execinstr + +.align 256 +AES_Te: +___ +&_data_word( + 0xc66363a5, 0xf87c7c84, 0xee777799, 0xf67b7b8d, + 0xfff2f20d, 0xd66b6bbd, 0xde6f6fb1, 0x91c5c554, + 0x60303050, 0x02010103, 0xce6767a9, 0x562b2b7d, + 0xe7fefe19, 0xb5d7d762, 0x4dababe6, 0xec76769a, + 0x8fcaca45, 0x1f82829d, 0x89c9c940, 0xfa7d7d87, + 0xeffafa15, 0xb25959eb, 0x8e4747c9, 0xfbf0f00b, + 0x41adadec, 0xb3d4d467, 0x5fa2a2fd, 0x45afafea, + 0x239c9cbf, 0x53a4a4f7, 0xe4727296, 0x9bc0c05b, + 0x75b7b7c2, 0xe1fdfd1c, 0x3d9393ae, 0x4c26266a, + 0x6c36365a, 0x7e3f3f41, 0xf5f7f702, 0x83cccc4f, + 0x6834345c, 0x51a5a5f4, 0xd1e5e534, 0xf9f1f108, + 0xe2717193, 0xabd8d873, 0x62313153, 0x2a15153f, + 0x0804040c, 0x95c7c752, 0x46232365, 0x9dc3c35e, + 0x30181828, 0x379696a1, 0x0a05050f, 0x2f9a9ab5, + 0x0e070709, 0x24121236, 0x1b80809b, 0xdfe2e23d, + 0xcdebeb26, 0x4e272769, 0x7fb2b2cd, 0xea75759f, + 0x1209091b, 0x1d83839e, 0x582c2c74, 0x341a1a2e, + 0x361b1b2d, 0xdc6e6eb2, 0xb45a5aee, 0x5ba0a0fb, + 0xa45252f6, 0x763b3b4d, 0xb7d6d661, 0x7db3b3ce, + 0x5229297b, 0xdde3e33e, 0x5e2f2f71, 0x13848497, + 0xa65353f5, 0xb9d1d168, 0x00000000, 0xc1eded2c, + 0x40202060, 0xe3fcfc1f, 0x79b1b1c8, 0xb65b5bed, + 0xd46a6abe, 0x8dcbcb46, 0x67bebed9, 0x7239394b, + 0x944a4ade, 0x984c4cd4, 0xb05858e8, 0x85cfcf4a, + 0xbbd0d06b, 0xc5efef2a, 0x4faaaae5, 0xedfbfb16, + 0x864343c5, 0x9a4d4dd7, 0x66333355, 0x11858594, + 0x8a4545cf, 0xe9f9f910, 0x04020206, 0xfe7f7f81, + 0xa05050f0, 0x783c3c44, 0x259f9fba, 0x4ba8a8e3, + 0xa25151f3, 0x5da3a3fe, 0x804040c0, 0x058f8f8a, + 0x3f9292ad, 0x219d9dbc, 0x70383848, 0xf1f5f504, + 0x63bcbcdf, 0x77b6b6c1, 0xafdada75, 0x42212163, + 0x20101030, 0xe5ffff1a, 0xfdf3f30e, 0xbfd2d26d, + 0x81cdcd4c, 0x180c0c14, 0x26131335, 0xc3ecec2f, + 0xbe5f5fe1, 0x359797a2, 0x884444cc, 0x2e171739, + 0x93c4c457, 0x55a7a7f2, 0xfc7e7e82, 0x7a3d3d47, + 0xc86464ac, 0xba5d5de7, 0x3219192b, 0xe6737395, + 0xc06060a0, 0x19818198, 0x9e4f4fd1, 0xa3dcdc7f, + 0x44222266, 0x542a2a7e, 0x3b9090ab, 0x0b888883, + 0x8c4646ca, 0xc7eeee29, 0x6bb8b8d3, 0x2814143c, + 0xa7dede79, 0xbc5e5ee2, 0x160b0b1d, 0xaddbdb76, + 0xdbe0e03b, 0x64323256, 0x743a3a4e, 0x140a0a1e, + 0x924949db, 0x0c06060a, 0x4824246c, 0xb85c5ce4, + 0x9fc2c25d, 0xbdd3d36e, 0x43acacef, 0xc46262a6, + 0x399191a8, 0x319595a4, 0xd3e4e437, 0xf279798b, + 0xd5e7e732, 0x8bc8c843, 0x6e373759, 0xda6d6db7, + 0x018d8d8c, 0xb1d5d564, 0x9c4e4ed2, 0x49a9a9e0, + 0xd86c6cb4, 0xac5656fa, 0xf3f4f407, 0xcfeaea25, + 0xca6565af, 0xf47a7a8e, 0x47aeaee9, 0x10080818, + 0x6fbabad5, 0xf0787888, 0x4a25256f, 0x5c2e2e72, + 0x381c1c24, 0x57a6a6f1, 0x73b4b4c7, 0x97c6c651, + 0xcbe8e823, 0xa1dddd7c, 0xe874749c, 0x3e1f1f21, + 0x964b4bdd, 0x61bdbddc, 0x0d8b8b86, 0x0f8a8a85, + 0xe0707090, 0x7c3e3e42, 0x71b5b5c4, 0xcc6666aa, + 0x904848d8, 0x06030305, 0xf7f6f601, 0x1c0e0e12, + 0xc26161a3, 0x6a35355f, 0xae5757f9, 0x69b9b9d0, + 0x17868691, 0x99c1c158, 0x3a1d1d27, 0x279e9eb9, + 0xd9e1e138, 0xebf8f813, 0x2b9898b3, 0x22111133, + 0xd26969bb, 0xa9d9d970, 0x078e8e89, 0x339494a7, + 0x2d9b9bb6, 0x3c1e1e22, 0x15878792, 0xc9e9e920, + 0x87cece49, 0xaa5555ff, 0x50282878, 0xa5dfdf7a, + 0x038c8c8f, 0x59a1a1f8, 0x09898980, 0x1a0d0d17, + 0x65bfbfda, 0xd7e6e631, 0x844242c6, 0xd06868b8, + 0x824141c3, 0x299999b0, 0x5a2d2d77, 0x1e0f0f11, + 0x7bb0b0cb, 0xa85454fc, 0x6dbbbbd6, 0x2c16163a); +$code.=<<___; + .byte 0x63, 0x7c, 0x77, 0x7b, 0xf2, 0x6b, 0x6f, 0xc5 + .byte 0x30, 0x01, 0x67, 0x2b, 0xfe, 0xd7, 0xab, 0x76 + .byte 0xca, 0x82, 0xc9, 0x7d, 0xfa, 0x59, 0x47, 0xf0 + .byte 0xad, 0xd4, 0xa2, 0xaf, 0x9c, 0xa4, 0x72, 0xc0 + .byte 0xb7, 0xfd, 0x93, 0x26, 0x36, 0x3f, 0xf7, 0xcc + .byte 0x34, 0xa5, 0xe5, 0xf1, 0x71, 0xd8, 0x31, 0x15 + .byte 0x04, 0xc7, 0x23, 0xc3, 0x18, 0x96, 0x05, 0x9a + .byte 0x07, 0x12, 0x80, 0xe2, 0xeb, 0x27, 0xb2, 0x75 + .byte 0x09, 0x83, 0x2c, 0x1a, 0x1b, 0x6e, 0x5a, 0xa0 + .byte 0x52, 0x3b, 0xd6, 0xb3, 0x29, 0xe3, 0x2f, 0x84 + .byte 0x53, 0xd1, 0x00, 0xed, 0x20, 0xfc, 0xb1, 0x5b + .byte 0x6a, 0xcb, 0xbe, 0x39, 0x4a, 0x4c, 0x58, 0xcf + .byte 0xd0, 0xef, 0xaa, 0xfb, 0x43, 0x4d, 0x33, 0x85 + .byte 0x45, 0xf9, 0x02, 0x7f, 0x50, 0x3c, 0x9f, 0xa8 + .byte 0x51, 0xa3, 0x40, 0x8f, 0x92, 0x9d, 0x38, 0xf5 + .byte 0xbc, 0xb6, 0xda, 0x21, 0x10, 0xff, 0xf3, 0xd2 + .byte 0xcd, 0x0c, 0x13, 0xec, 0x5f, 0x97, 0x44, 0x17 + .byte 0xc4, 0xa7, 0x7e, 0x3d, 0x64, 0x5d, 0x19, 0x73 + .byte 0x60, 0x81, 0x4f, 0xdc, 0x22, 0x2a, 0x90, 0x88 + .byte 0x46, 0xee, 0xb8, 0x14, 0xde, 0x5e, 0x0b, 0xdb + .byte 0xe0, 0x32, 0x3a, 0x0a, 0x49, 0x06, 0x24, 0x5c + .byte 0xc2, 0xd3, 0xac, 0x62, 0x91, 0x95, 0xe4, 0x79 + .byte 0xe7, 0xc8, 0x37, 0x6d, 0x8d, 0xd5, 0x4e, 0xa9 + .byte 0x6c, 0x56, 0xf4, 0xea, 0x65, 0x7a, 0xae, 0x08 + .byte 0xba, 0x78, 0x25, 0x2e, 0x1c, 0xa6, 0xb4, 0xc6 + .byte 0xe8, 0xdd, 0x74, 0x1f, 0x4b, 0xbd, 0x8b, 0x8a + .byte 0x70, 0x3e, 0xb5, 0x66, 0x48, 0x03, 0xf6, 0x0e + .byte 0x61, 0x35, 0x57, 0xb9, 0x86, 0xc1, 0x1d, 0x9e + .byte 0xe1, 0xf8, 0x98, 0x11, 0x69, 0xd9, 0x8e, 0x94 + .byte 0x9b, 0x1e, 0x87, 0xe9, 0xce, 0x55, 0x28, 0xdf + .byte 0x8c, 0xa1, 0x89, 0x0d, 0xbf, 0xe6, 0x42, 0x68 + .byte 0x41, 0x99, 0x2d, 0x0f, 0xb0, 0x54, 0xbb, 0x16 +.type AES_Te,#object +.size AES_Te,(.-AES_Te) + +.align 64 +.skip 16 +_sparcv9_AES_encrypt: + save %sp,-$frame-$locals,%sp + stx %i7,[%sp+$bias+$frame+0] ! off-load return address + ld [$key+240],$rounds + ld [$key+0],$t0 + ld [$key+4],$t1 ! + ld [$key+8],$t2 + srl $rounds,1,$rounds + xor $t0,$s0,$s0 + ld [$key+12],$t3 + srl $s0,21,$acc0 + xor $t1,$s1,$s1 + ld [$key+16],$t0 + srl $s1,13,$acc1 ! + xor $t2,$s2,$s2 + ld [$key+20],$t1 + xor $t3,$s3,$s3 + ld [$key+24],$t2 + and $acc0,2040,$acc0 + ld [$key+28],$t3 + nop +.Lenc_loop: + srl $s2,5,$acc2 ! + and $acc1,2040,$acc1 + ldx [$tbl+$acc0],$acc0 + sll $s3,3,$acc3 + and $acc2,2040,$acc2 + ldx [$tbl+$acc1],$acc1 + srl $s1,21,$acc4 + and $acc3,2040,$acc3 + ldx [$tbl+$acc2],$acc2 ! + srl $s2,13,$acc5 + and $acc4,2040,$acc4 + ldx [$tbl+$acc3],$acc3 + srl $s3,5,$acc6 + and $acc5,2040,$acc5 + ldx [$tbl+$acc4],$acc4 + fmovs %f0,%f0 + sll $s0,3,$acc7 ! + and $acc6,2040,$acc6 + ldx [$tbl+$acc5],$acc5 + srl $s2,21,$acc8 + and $acc7,2040,$acc7 + ldx [$tbl+$acc6],$acc6 + srl $s3,13,$acc9 + and $acc8,2040,$acc8 + ldx [$tbl+$acc7],$acc7 ! + srl $s0,5,$acc10 + and $acc9,2040,$acc9 + ldx [$tbl+$acc8],$acc8 + sll $s1,3,$acc11 + and $acc10,2040,$acc10 + ldx [$tbl+$acc9],$acc9 + fmovs %f0,%f0 + srl $s3,21,$acc12 ! + and $acc11,2040,$acc11 + ldx [$tbl+$acc10],$acc10 + srl $s0,13,$acc13 + and $acc12,2040,$acc12 + ldx [$tbl+$acc11],$acc11 + srl $s1,5,$acc14 + and $acc13,2040,$acc13 + ldx [$tbl+$acc12],$acc12 ! + sll $s2,3,$acc15 + and $acc14,2040,$acc14 + ldx [$tbl+$acc13],$acc13 + and $acc15,2040,$acc15 + add $key,32,$key + ldx [$tbl+$acc14],$acc14 + fmovs %f0,%f0 + subcc $rounds,1,$rounds ! + ldx [$tbl+$acc15],$acc15 + bz,a,pn %icc,.Lenc_last + add $tbl,2048,$rounds + + srlx $acc1,8,$acc1 + xor $acc0,$t0,$t0 + ld [$key+0],$s0 + fmovs %f0,%f0 + srlx $acc2,16,$acc2 ! + xor $acc1,$t0,$t0 + ld [$key+4],$s1 + srlx $acc3,24,$acc3 + xor $acc2,$t0,$t0 + ld [$key+8],$s2 + srlx $acc5,8,$acc5 + xor $acc3,$t0,$t0 + ld [$key+12],$s3 ! + srlx $acc6,16,$acc6 + xor $acc4,$t1,$t1 + fmovs %f0,%f0 + srlx $acc7,24,$acc7 + xor $acc5,$t1,$t1 + srlx $acc9,8,$acc9 + xor $acc6,$t1,$t1 + srlx $acc10,16,$acc10 ! + xor $acc7,$t1,$t1 + srlx $acc11,24,$acc11 + xor $acc8,$t2,$t2 + srlx $acc13,8,$acc13 + xor $acc9,$t2,$t2 + srlx $acc14,16,$acc14 + xor $acc10,$t2,$t2 + srlx $acc15,24,$acc15 ! + xor $acc11,$t2,$t2 + xor $acc12,$acc14,$acc14 + xor $acc13,$t3,$t3 + srl $t0,21,$acc0 + xor $acc14,$t3,$t3 + srl $t1,13,$acc1 + xor $acc15,$t3,$t3 + + and $acc0,2040,$acc0 ! + srl $t2,5,$acc2 + and $acc1,2040,$acc1 + ldx [$tbl+$acc0],$acc0 + sll $t3,3,$acc3 + and $acc2,2040,$acc2 + ldx [$tbl+$acc1],$acc1 + fmovs %f0,%f0 + srl $t1,21,$acc4 ! + and $acc3,2040,$acc3 + ldx [$tbl+$acc2],$acc2 + srl $t2,13,$acc5 + and $acc4,2040,$acc4 + ldx [$tbl+$acc3],$acc3 + srl $t3,5,$acc6 + and $acc5,2040,$acc5 + ldx [$tbl+$acc4],$acc4 ! + sll $t0,3,$acc7 + and $acc6,2040,$acc6 + ldx [$tbl+$acc5],$acc5 + srl $t2,21,$acc8 + and $acc7,2040,$acc7 + ldx [$tbl+$acc6],$acc6 + fmovs %f0,%f0 + srl $t3,13,$acc9 ! + and $acc8,2040,$acc8 + ldx [$tbl+$acc7],$acc7 + srl $t0,5,$acc10 + and $acc9,2040,$acc9 + ldx [$tbl+$acc8],$acc8 + sll $t1,3,$acc11 + and $acc10,2040,$acc10 + ldx [$tbl+$acc9],$acc9 ! + srl $t3,21,$acc12 + and $acc11,2040,$acc11 + ldx [$tbl+$acc10],$acc10 + srl $t0,13,$acc13 + and $acc12,2040,$acc12 + ldx [$tbl+$acc11],$acc11 + fmovs %f0,%f0 + srl $t1,5,$acc14 ! + and $acc13,2040,$acc13 + ldx [$tbl+$acc12],$acc12 + sll $t2,3,$acc15 + and $acc14,2040,$acc14 + ldx [$tbl+$acc13],$acc13 + srlx $acc1,8,$acc1 + and $acc15,2040,$acc15 + ldx [$tbl+$acc14],$acc14 ! + + srlx $acc2,16,$acc2 + xor $acc0,$s0,$s0 + ldx [$tbl+$acc15],$acc15 + srlx $acc3,24,$acc3 + xor $acc1,$s0,$s0 + ld [$key+16],$t0 + fmovs %f0,%f0 + srlx $acc5,8,$acc5 ! + xor $acc2,$s0,$s0 + ld [$key+20],$t1 + srlx $acc6,16,$acc6 + xor $acc3,$s0,$s0 + ld [$key+24],$t2 + srlx $acc7,24,$acc7 + xor $acc4,$s1,$s1 + ld [$key+28],$t3 ! + srlx $acc9,8,$acc9 + xor $acc5,$s1,$s1 + ldx [$tbl+2048+0],%g0 ! prefetch te4 + srlx $acc10,16,$acc10 + xor $acc6,$s1,$s1 + ldx [$tbl+2048+32],%g0 ! prefetch te4 + srlx $acc11,24,$acc11 + xor $acc7,$s1,$s1 + ldx [$tbl+2048+64],%g0 ! prefetch te4 + srlx $acc13,8,$acc13 + xor $acc8,$s2,$s2 + ldx [$tbl+2048+96],%g0 ! prefetch te4 + srlx $acc14,16,$acc14 ! + xor $acc9,$s2,$s2 + ldx [$tbl+2048+128],%g0 ! prefetch te4 + srlx $acc15,24,$acc15 + xor $acc10,$s2,$s2 + ldx [$tbl+2048+160],%g0 ! prefetch te4 + srl $s0,21,$acc0 + xor $acc11,$s2,$s2 + ldx [$tbl+2048+192],%g0 ! prefetch te4 + xor $acc12,$acc14,$acc14 + xor $acc13,$s3,$s3 + ldx [$tbl+2048+224],%g0 ! prefetch te4 + srl $s1,13,$acc1 ! + xor $acc14,$s3,$s3 + xor $acc15,$s3,$s3 + ba .Lenc_loop + and $acc0,2040,$acc0 + +.align 32 +.Lenc_last: + srlx $acc1,8,$acc1 ! + xor $acc0,$t0,$t0 + ld [$key+0],$s0 + srlx $acc2,16,$acc2 + xor $acc1,$t0,$t0 + ld [$key+4],$s1 + srlx $acc3,24,$acc3 + xor $acc2,$t0,$t0 + ld [$key+8],$s2 ! + srlx $acc5,8,$acc5 + xor $acc3,$t0,$t0 + ld [$key+12],$s3 + srlx $acc6,16,$acc6 + xor $acc4,$t1,$t1 + srlx $acc7,24,$acc7 + xor $acc5,$t1,$t1 + srlx $acc9,8,$acc9 ! + xor $acc6,$t1,$t1 + srlx $acc10,16,$acc10 + xor $acc7,$t1,$t1 + srlx $acc11,24,$acc11 + xor $acc8,$t2,$t2 + srlx $acc13,8,$acc13 + xor $acc9,$t2,$t2 + srlx $acc14,16,$acc14 ! + xor $acc10,$t2,$t2 + srlx $acc15,24,$acc15 + xor $acc11,$t2,$t2 + xor $acc12,$acc14,$acc14 + xor $acc13,$t3,$t3 + srl $t0,24,$acc0 + xor $acc14,$t3,$t3 + srl $t1,16,$acc1 ! + xor $acc15,$t3,$t3 + + srl $t2,8,$acc2 + and $acc1,255,$acc1 + ldub [$rounds+$acc0],$acc0 + srl $t1,24,$acc4 + and $acc2,255,$acc2 + ldub [$rounds+$acc1],$acc1 + srl $t2,16,$acc5 ! + and $t3,255,$acc3 + ldub [$rounds+$acc2],$acc2 + ldub [$rounds+$acc3],$acc3 + srl $t3,8,$acc6 + and $acc5,255,$acc5 + ldub [$rounds+$acc4],$acc4 + fmovs %f0,%f0 + srl $t2,24,$acc8 ! + and $acc6,255,$acc6 + ldub [$rounds+$acc5],$acc5 + srl $t3,16,$acc9 + and $t0,255,$acc7 + ldub [$rounds+$acc6],$acc6 + ldub [$rounds+$acc7],$acc7 + fmovs %f0,%f0 + srl $t0,8,$acc10 ! + and $acc9,255,$acc9 + ldub [$rounds+$acc8],$acc8 + srl $t3,24,$acc12 + and $acc10,255,$acc10 + ldub [$rounds+$acc9],$acc9 + srl $t0,16,$acc13 + and $t1,255,$acc11 + ldub [$rounds+$acc10],$acc10 ! + srl $t1,8,$acc14 + and $acc13,255,$acc13 + ldub [$rounds+$acc11],$acc11 + ldub [$rounds+$acc12],$acc12 + and $acc14,255,$acc14 + ldub [$rounds+$acc13],$acc13 + and $t2,255,$acc15 + ldub [$rounds+$acc14],$acc14 ! + + sll $acc0,24,$acc0 + xor $acc3,$s0,$s0 + ldub [$rounds+$acc15],$acc15 + sll $acc1,16,$acc1 + xor $acc0,$s0,$s0 + ldx [%sp+$bias+$frame+0],%i7 ! restore return address + fmovs %f0,%f0 + sll $acc2,8,$acc2 ! + xor $acc1,$s0,$s0 + sll $acc4,24,$acc4 + xor $acc2,$s0,$s0 + sll $acc5,16,$acc5 + xor $acc7,$s1,$s1 + sll $acc6,8,$acc6 + xor $acc4,$s1,$s1 + sll $acc8,24,$acc8 ! + xor $acc5,$s1,$s1 + sll $acc9,16,$acc9 + xor $acc11,$s2,$s2 + sll $acc10,8,$acc10 + xor $acc6,$s1,$s1 + sll $acc12,24,$acc12 + xor $acc8,$s2,$s2 + sll $acc13,16,$acc13 ! + xor $acc9,$s2,$s2 + sll $acc14,8,$acc14 + xor $acc10,$s2,$s2 + xor $acc12,$acc14,$acc14 + xor $acc13,$s3,$s3 + xor $acc14,$s3,$s3 + xor $acc15,$s3,$s3 + + ret + restore +.type _sparcv9_AES_encrypt,#function +.size _sparcv9_AES_encrypt,(.-_sparcv9_AES_encrypt) + +.align 32 +.globl AES_encrypt +AES_encrypt: + or %o0,%o1,%g1 + andcc %g1,3,%g0 + bnz,pn %xcc,.Lunaligned_enc + save %sp,-$frame,%sp + + ld [%i0+0],%o0 + ld [%i0+4],%o1 + ld [%i0+8],%o2 + ld [%i0+12],%o3 + +1: call .+8 + add %o7,AES_Te-1b,%o4 + call _sparcv9_AES_encrypt + mov %i2,%o5 + + st %o0,[%i1+0] + st %o1,[%i1+4] + st %o2,[%i1+8] + st %o3,[%i1+12] + + ret + restore + +.align 32 +.Lunaligned_enc: + ldub [%i0+0],%l0 + ldub [%i0+1],%l1 + ldub [%i0+2],%l2 + + sll %l0,24,%l0 + ldub [%i0+3],%l3 + sll %l1,16,%l1 + ldub [%i0+4],%l4 + sll %l2,8,%l2 + or %l1,%l0,%l0 + ldub [%i0+5],%l5 + sll %l4,24,%l4 + or %l3,%l2,%l2 + ldub [%i0+6],%l6 + sll %l5,16,%l5 + or %l0,%l2,%o0 + ldub [%i0+7],%l7 + + sll %l6,8,%l6 + or %l5,%l4,%l4 + ldub [%i0+8],%l0 + or %l7,%l6,%l6 + ldub [%i0+9],%l1 + or %l4,%l6,%o1 + ldub [%i0+10],%l2 + + sll %l0,24,%l0 + ldub [%i0+11],%l3 + sll %l1,16,%l1 + ldub [%i0+12],%l4 + sll %l2,8,%l2 + or %l1,%l0,%l0 + ldub [%i0+13],%l5 + sll %l4,24,%l4 + or %l3,%l2,%l2 + ldub [%i0+14],%l6 + sll %l5,16,%l5 + or %l0,%l2,%o2 + ldub [%i0+15],%l7 + + sll %l6,8,%l6 + or %l5,%l4,%l4 + or %l7,%l6,%l6 + or %l4,%l6,%o3 + +1: call .+8 + add %o7,AES_Te-1b,%o4 + call _sparcv9_AES_encrypt + mov %i2,%o5 + + srl %o0,24,%l0 + srl %o0,16,%l1 + stb %l0,[%i1+0] + srl %o0,8,%l2 + stb %l1,[%i1+1] + stb %l2,[%i1+2] + srl %o1,24,%l4 + stb %o0,[%i1+3] + + srl %o1,16,%l5 + stb %l4,[%i1+4] + srl %o1,8,%l6 + stb %l5,[%i1+5] + stb %l6,[%i1+6] + srl %o2,24,%l0 + stb %o1,[%i1+7] + + srl %o2,16,%l1 + stb %l0,[%i1+8] + srl %o2,8,%l2 + stb %l1,[%i1+9] + stb %l2,[%i1+10] + srl %o3,24,%l4 + stb %o2,[%i1+11] + + srl %o3,16,%l5 + stb %l4,[%i1+12] + srl %o3,8,%l6 + stb %l5,[%i1+13] + stb %l6,[%i1+14] + stb %o3,[%i1+15] + + ret + restore +.type AES_encrypt,#function +.size AES_encrypt,(.-AES_encrypt) + +___ + +$code.=<<___; +.align 256 +AES_Td: +___ +&_data_word( + 0x51f4a750, 0x7e416553, 0x1a17a4c3, 0x3a275e96, + 0x3bab6bcb, 0x1f9d45f1, 0xacfa58ab, 0x4be30393, + 0x2030fa55, 0xad766df6, 0x88cc7691, 0xf5024c25, + 0x4fe5d7fc, 0xc52acbd7, 0x26354480, 0xb562a38f, + 0xdeb15a49, 0x25ba1b67, 0x45ea0e98, 0x5dfec0e1, + 0xc32f7502, 0x814cf012, 0x8d4697a3, 0x6bd3f9c6, + 0x038f5fe7, 0x15929c95, 0xbf6d7aeb, 0x955259da, + 0xd4be832d, 0x587421d3, 0x49e06929, 0x8ec9c844, + 0x75c2896a, 0xf48e7978, 0x99583e6b, 0x27b971dd, + 0xbee14fb6, 0xf088ad17, 0xc920ac66, 0x7dce3ab4, + 0x63df4a18, 0xe51a3182, 0x97513360, 0x62537f45, + 0xb16477e0, 0xbb6bae84, 0xfe81a01c, 0xf9082b94, + 0x70486858, 0x8f45fd19, 0x94de6c87, 0x527bf8b7, + 0xab73d323, 0x724b02e2, 0xe31f8f57, 0x6655ab2a, + 0xb2eb2807, 0x2fb5c203, 0x86c57b9a, 0xd33708a5, + 0x302887f2, 0x23bfa5b2, 0x02036aba, 0xed16825c, + 0x8acf1c2b, 0xa779b492, 0xf307f2f0, 0x4e69e2a1, + 0x65daf4cd, 0x0605bed5, 0xd134621f, 0xc4a6fe8a, + 0x342e539d, 0xa2f355a0, 0x058ae132, 0xa4f6eb75, + 0x0b83ec39, 0x4060efaa, 0x5e719f06, 0xbd6e1051, + 0x3e218af9, 0x96dd063d, 0xdd3e05ae, 0x4de6bd46, + 0x91548db5, 0x71c45d05, 0x0406d46f, 0x605015ff, + 0x1998fb24, 0xd6bde997, 0x894043cc, 0x67d99e77, + 0xb0e842bd, 0x07898b88, 0xe7195b38, 0x79c8eedb, + 0xa17c0a47, 0x7c420fe9, 0xf8841ec9, 0x00000000, + 0x09808683, 0x322bed48, 0x1e1170ac, 0x6c5a724e, + 0xfd0efffb, 0x0f853856, 0x3daed51e, 0x362d3927, + 0x0a0fd964, 0x685ca621, 0x9b5b54d1, 0x24362e3a, + 0x0c0a67b1, 0x9357e70f, 0xb4ee96d2, 0x1b9b919e, + 0x80c0c54f, 0x61dc20a2, 0x5a774b69, 0x1c121a16, + 0xe293ba0a, 0xc0a02ae5, 0x3c22e043, 0x121b171d, + 0x0e090d0b, 0xf28bc7ad, 0x2db6a8b9, 0x141ea9c8, + 0x57f11985, 0xaf75074c, 0xee99ddbb, 0xa37f60fd, + 0xf701269f, 0x5c72f5bc, 0x44663bc5, 0x5bfb7e34, + 0x8b432976, 0xcb23c6dc, 0xb6edfc68, 0xb8e4f163, + 0xd731dcca, 0x42638510, 0x13972240, 0x84c61120, + 0x854a247d, 0xd2bb3df8, 0xaef93211, 0xc729a16d, + 0x1d9e2f4b, 0xdcb230f3, 0x0d8652ec, 0x77c1e3d0, + 0x2bb3166c, 0xa970b999, 0x119448fa, 0x47e96422, + 0xa8fc8cc4, 0xa0f03f1a, 0x567d2cd8, 0x223390ef, + 0x87494ec7, 0xd938d1c1, 0x8ccaa2fe, 0x98d40b36, + 0xa6f581cf, 0xa57ade28, 0xdab78e26, 0x3fadbfa4, + 0x2c3a9de4, 0x5078920d, 0x6a5fcc9b, 0x547e4662, + 0xf68d13c2, 0x90d8b8e8, 0x2e39f75e, 0x82c3aff5, + 0x9f5d80be, 0x69d0937c, 0x6fd52da9, 0xcf2512b3, + 0xc8ac993b, 0x10187da7, 0xe89c636e, 0xdb3bbb7b, + 0xcd267809, 0x6e5918f4, 0xec9ab701, 0x834f9aa8, + 0xe6956e65, 0xaaffe67e, 0x21bccf08, 0xef15e8e6, + 0xbae79bd9, 0x4a6f36ce, 0xea9f09d4, 0x29b07cd6, + 0x31a4b2af, 0x2a3f2331, 0xc6a59430, 0x35a266c0, + 0x744ebc37, 0xfc82caa6, 0xe090d0b0, 0x33a7d815, + 0xf104984a, 0x41ecdaf7, 0x7fcd500e, 0x1791f62f, + 0x764dd68d, 0x43efb04d, 0xccaa4d54, 0xe49604df, + 0x9ed1b5e3, 0x4c6a881b, 0xc12c1fb8, 0x4665517f, + 0x9d5eea04, 0x018c355d, 0xfa877473, 0xfb0b412e, + 0xb3671d5a, 0x92dbd252, 0xe9105633, 0x6dd64713, + 0x9ad7618c, 0x37a10c7a, 0x59f8148e, 0xeb133c89, + 0xcea927ee, 0xb761c935, 0xe11ce5ed, 0x7a47b13c, + 0x9cd2df59, 0x55f2733f, 0x1814ce79, 0x73c737bf, + 0x53f7cdea, 0x5ffdaa5b, 0xdf3d6f14, 0x7844db86, + 0xcaaff381, 0xb968c43e, 0x3824342c, 0xc2a3405f, + 0x161dc372, 0xbce2250c, 0x283c498b, 0xff0d9541, + 0x39a80171, 0x080cb3de, 0xd8b4e49c, 0x6456c190, + 0x7bcb8461, 0xd532b670, 0x486c5c74, 0xd0b85742); +$code.=<<___; + .byte 0x52, 0x09, 0x6a, 0xd5, 0x30, 0x36, 0xa5, 0x38 + .byte 0xbf, 0x40, 0xa3, 0x9e, 0x81, 0xf3, 0xd7, 0xfb + .byte 0x7c, 0xe3, 0x39, 0x82, 0x9b, 0x2f, 0xff, 0x87 + .byte 0x34, 0x8e, 0x43, 0x44, 0xc4, 0xde, 0xe9, 0xcb + .byte 0x54, 0x7b, 0x94, 0x32, 0xa6, 0xc2, 0x23, 0x3d + .byte 0xee, 0x4c, 0x95, 0x0b, 0x42, 0xfa, 0xc3, 0x4e + .byte 0x08, 0x2e, 0xa1, 0x66, 0x28, 0xd9, 0x24, 0xb2 + .byte 0x76, 0x5b, 0xa2, 0x49, 0x6d, 0x8b, 0xd1, 0x25 + .byte 0x72, 0xf8, 0xf6, 0x64, 0x86, 0x68, 0x98, 0x16 + .byte 0xd4, 0xa4, 0x5c, 0xcc, 0x5d, 0x65, 0xb6, 0x92 + .byte 0x6c, 0x70, 0x48, 0x50, 0xfd, 0xed, 0xb9, 0xda + .byte 0x5e, 0x15, 0x46, 0x57, 0xa7, 0x8d, 0x9d, 0x84 + .byte 0x90, 0xd8, 0xab, 0x00, 0x8c, 0xbc, 0xd3, 0x0a + .byte 0xf7, 0xe4, 0x58, 0x05, 0xb8, 0xb3, 0x45, 0x06 + .byte 0xd0, 0x2c, 0x1e, 0x8f, 0xca, 0x3f, 0x0f, 0x02 + .byte 0xc1, 0xaf, 0xbd, 0x03, 0x01, 0x13, 0x8a, 0x6b + .byte 0x3a, 0x91, 0x11, 0x41, 0x4f, 0x67, 0xdc, 0xea + .byte 0x97, 0xf2, 0xcf, 0xce, 0xf0, 0xb4, 0xe6, 0x73 + .byte 0x96, 0xac, 0x74, 0x22, 0xe7, 0xad, 0x35, 0x85 + .byte 0xe2, 0xf9, 0x37, 0xe8, 0x1c, 0x75, 0xdf, 0x6e + .byte 0x47, 0xf1, 0x1a, 0x71, 0x1d, 0x29, 0xc5, 0x89 + .byte 0x6f, 0xb7, 0x62, 0x0e, 0xaa, 0x18, 0xbe, 0x1b + .byte 0xfc, 0x56, 0x3e, 0x4b, 0xc6, 0xd2, 0x79, 0x20 + .byte 0x9a, 0xdb, 0xc0, 0xfe, 0x78, 0xcd, 0x5a, 0xf4 + .byte 0x1f, 0xdd, 0xa8, 0x33, 0x88, 0x07, 0xc7, 0x31 + .byte 0xb1, 0x12, 0x10, 0x59, 0x27, 0x80, 0xec, 0x5f + .byte 0x60, 0x51, 0x7f, 0xa9, 0x19, 0xb5, 0x4a, 0x0d + .byte 0x2d, 0xe5, 0x7a, 0x9f, 0x93, 0xc9, 0x9c, 0xef + .byte 0xa0, 0xe0, 0x3b, 0x4d, 0xae, 0x2a, 0xf5, 0xb0 + .byte 0xc8, 0xeb, 0xbb, 0x3c, 0x83, 0x53, 0x99, 0x61 + .byte 0x17, 0x2b, 0x04, 0x7e, 0xba, 0x77, 0xd6, 0x26 + .byte 0xe1, 0x69, 0x14, 0x63, 0x55, 0x21, 0x0c, 0x7d +.type AES_Td,#object +.size AES_Td,(.-AES_Td) + +.align 64 +.skip 16 +_sparcv9_AES_decrypt: + save %sp,-$frame-$locals,%sp + stx %i7,[%sp+$bias+$frame+0] ! off-load return address + ld [$key+240],$rounds + ld [$key+0],$t0 + ld [$key+4],$t1 ! + ld [$key+8],$t2 + ld [$key+12],$t3 + srl $rounds,1,$rounds + xor $t0,$s0,$s0 + ld [$key+16],$t0 + xor $t1,$s1,$s1 + ld [$key+20],$t1 + srl $s0,21,$acc0 ! + xor $t2,$s2,$s2 + ld [$key+24],$t2 + xor $t3,$s3,$s3 + and $acc0,2040,$acc0 + ld [$key+28],$t3 + srl $s3,13,$acc1 + nop +.Ldec_loop: + srl $s2,5,$acc2 ! + and $acc1,2040,$acc1 + ldx [$tbl+$acc0],$acc0 + sll $s1,3,$acc3 + and $acc2,2040,$acc2 + ldx [$tbl+$acc1],$acc1 + srl $s1,21,$acc4 + and $acc3,2040,$acc3 + ldx [$tbl+$acc2],$acc2 ! + srl $s0,13,$acc5 + and $acc4,2040,$acc4 + ldx [$tbl+$acc3],$acc3 + srl $s3,5,$acc6 + and $acc5,2040,$acc5 + ldx [$tbl+$acc4],$acc4 + fmovs %f0,%f0 + sll $s2,3,$acc7 ! + and $acc6,2040,$acc6 + ldx [$tbl+$acc5],$acc5 + srl $s2,21,$acc8 + and $acc7,2040,$acc7 + ldx [$tbl+$acc6],$acc6 + srl $s1,13,$acc9 + and $acc8,2040,$acc8 + ldx [$tbl+$acc7],$acc7 ! + srl $s0,5,$acc10 + and $acc9,2040,$acc9 + ldx [$tbl+$acc8],$acc8 + sll $s3,3,$acc11 + and $acc10,2040,$acc10 + ldx [$tbl+$acc9],$acc9 + fmovs %f0,%f0 + srl $s3,21,$acc12 ! + and $acc11,2040,$acc11 + ldx [$tbl+$acc10],$acc10 + srl $s2,13,$acc13 + and $acc12,2040,$acc12 + ldx [$tbl+$acc11],$acc11 + srl $s1,5,$acc14 + and $acc13,2040,$acc13 + ldx [$tbl+$acc12],$acc12 ! + sll $s0,3,$acc15 + and $acc14,2040,$acc14 + ldx [$tbl+$acc13],$acc13 + and $acc15,2040,$acc15 + add $key,32,$key + ldx [$tbl+$acc14],$acc14 + fmovs %f0,%f0 + subcc $rounds,1,$rounds ! + ldx [$tbl+$acc15],$acc15 + bz,a,pn %icc,.Ldec_last + add $tbl,2048,$rounds + + srlx $acc1,8,$acc1 + xor $acc0,$t0,$t0 + ld [$key+0],$s0 + fmovs %f0,%f0 + srlx $acc2,16,$acc2 ! + xor $acc1,$t0,$t0 + ld [$key+4],$s1 + srlx $acc3,24,$acc3 + xor $acc2,$t0,$t0 + ld [$key+8],$s2 + srlx $acc5,8,$acc5 + xor $acc3,$t0,$t0 + ld [$key+12],$s3 ! + srlx $acc6,16,$acc6 + xor $acc4,$t1,$t1 + fmovs %f0,%f0 + srlx $acc7,24,$acc7 + xor $acc5,$t1,$t1 + srlx $acc9,8,$acc9 + xor $acc6,$t1,$t1 + srlx $acc10,16,$acc10 ! + xor $acc7,$t1,$t1 + srlx $acc11,24,$acc11 + xor $acc8,$t2,$t2 + srlx $acc13,8,$acc13 + xor $acc9,$t2,$t2 + srlx $acc14,16,$acc14 + xor $acc10,$t2,$t2 + srlx $acc15,24,$acc15 ! + xor $acc11,$t2,$t2 + xor $acc12,$acc14,$acc14 + xor $acc13,$t3,$t3 + srl $t0,21,$acc0 + xor $acc14,$t3,$t3 + xor $acc15,$t3,$t3 + srl $t3,13,$acc1 + + and $acc0,2040,$acc0 ! + srl $t2,5,$acc2 + and $acc1,2040,$acc1 + ldx [$tbl+$acc0],$acc0 + sll $t1,3,$acc3 + and $acc2,2040,$acc2 + ldx [$tbl+$acc1],$acc1 + fmovs %f0,%f0 + srl $t1,21,$acc4 ! + and $acc3,2040,$acc3 + ldx [$tbl+$acc2],$acc2 + srl $t0,13,$acc5 + and $acc4,2040,$acc4 + ldx [$tbl+$acc3],$acc3 + srl $t3,5,$acc6 + and $acc5,2040,$acc5 + ldx [$tbl+$acc4],$acc4 ! + sll $t2,3,$acc7 + and $acc6,2040,$acc6 + ldx [$tbl+$acc5],$acc5 + srl $t2,21,$acc8 + and $acc7,2040,$acc7 + ldx [$tbl+$acc6],$acc6 + fmovs %f0,%f0 + srl $t1,13,$acc9 ! + and $acc8,2040,$acc8 + ldx [$tbl+$acc7],$acc7 + srl $t0,5,$acc10 + and $acc9,2040,$acc9 + ldx [$tbl+$acc8],$acc8 + sll $t3,3,$acc11 + and $acc10,2040,$acc10 + ldx [$tbl+$acc9],$acc9 ! + srl $t3,21,$acc12 + and $acc11,2040,$acc11 + ldx [$tbl+$acc10],$acc10 + srl $t2,13,$acc13 + and $acc12,2040,$acc12 + ldx [$tbl+$acc11],$acc11 + fmovs %f0,%f0 + srl $t1,5,$acc14 ! + and $acc13,2040,$acc13 + ldx [$tbl+$acc12],$acc12 + sll $t0,3,$acc15 + and $acc14,2040,$acc14 + ldx [$tbl+$acc13],$acc13 + srlx $acc1,8,$acc1 + and $acc15,2040,$acc15 + ldx [$tbl+$acc14],$acc14 ! + + srlx $acc2,16,$acc2 + xor $acc0,$s0,$s0 + ldx [$tbl+$acc15],$acc15 + srlx $acc3,24,$acc3 + xor $acc1,$s0,$s0 + ld [$key+16],$t0 + fmovs %f0,%f0 + srlx $acc5,8,$acc5 ! + xor $acc2,$s0,$s0 + ld [$key+20],$t1 + srlx $acc6,16,$acc6 + xor $acc3,$s0,$s0 + ld [$key+24],$t2 + srlx $acc7,24,$acc7 + xor $acc4,$s1,$s1 + ld [$key+28],$t3 ! + srlx $acc9,8,$acc9 + xor $acc5,$s1,$s1 + ldx [$tbl+2048+0],%g0 ! prefetch td4 + srlx $acc10,16,$acc10 + xor $acc6,$s1,$s1 + ldx [$tbl+2048+32],%g0 ! prefetch td4 + srlx $acc11,24,$acc11 + xor $acc7,$s1,$s1 + ldx [$tbl+2048+64],%g0 ! prefetch td4 + srlx $acc13,8,$acc13 + xor $acc8,$s2,$s2 + ldx [$tbl+2048+96],%g0 ! prefetch td4 + srlx $acc14,16,$acc14 ! + xor $acc9,$s2,$s2 + ldx [$tbl+2048+128],%g0 ! prefetch td4 + srlx $acc15,24,$acc15 + xor $acc10,$s2,$s2 + ldx [$tbl+2048+160],%g0 ! prefetch td4 + srl $s0,21,$acc0 + xor $acc11,$s2,$s2 + ldx [$tbl+2048+192],%g0 ! prefetch td4 + xor $acc12,$acc14,$acc14 + xor $acc13,$s3,$s3 + ldx [$tbl+2048+224],%g0 ! prefetch td4 + and $acc0,2040,$acc0 ! + xor $acc14,$s3,$s3 + xor $acc15,$s3,$s3 + ba .Ldec_loop + srl $s3,13,$acc1 + +.align 32 +.Ldec_last: + srlx $acc1,8,$acc1 ! + xor $acc0,$t0,$t0 + ld [$key+0],$s0 + srlx $acc2,16,$acc2 + xor $acc1,$t0,$t0 + ld [$key+4],$s1 + srlx $acc3,24,$acc3 + xor $acc2,$t0,$t0 + ld [$key+8],$s2 ! + srlx $acc5,8,$acc5 + xor $acc3,$t0,$t0 + ld [$key+12],$s3 + srlx $acc6,16,$acc6 + xor $acc4,$t1,$t1 + srlx $acc7,24,$acc7 + xor $acc5,$t1,$t1 + srlx $acc9,8,$acc9 ! + xor $acc6,$t1,$t1 + srlx $acc10,16,$acc10 + xor $acc7,$t1,$t1 + srlx $acc11,24,$acc11 + xor $acc8,$t2,$t2 + srlx $acc13,8,$acc13 + xor $acc9,$t2,$t2 + srlx $acc14,16,$acc14 ! + xor $acc10,$t2,$t2 + srlx $acc15,24,$acc15 + xor $acc11,$t2,$t2 + xor $acc12,$acc14,$acc14 + xor $acc13,$t3,$t3 + srl $t0,24,$acc0 + xor $acc14,$t3,$t3 + xor $acc15,$t3,$t3 ! + srl $t3,16,$acc1 + + srl $t2,8,$acc2 + and $acc1,255,$acc1 + ldub [$rounds+$acc0],$acc0 + srl $t1,24,$acc4 + and $acc2,255,$acc2 + ldub [$rounds+$acc1],$acc1 + srl $t0,16,$acc5 ! + and $t1,255,$acc3 + ldub [$rounds+$acc2],$acc2 + ldub [$rounds+$acc3],$acc3 + srl $t3,8,$acc6 + and $acc5,255,$acc5 + ldub [$rounds+$acc4],$acc4 + fmovs %f0,%f0 + srl $t2,24,$acc8 ! + and $acc6,255,$acc6 + ldub [$rounds+$acc5],$acc5 + srl $t1,16,$acc9 + and $t2,255,$acc7 + ldub [$rounds+$acc6],$acc6 + ldub [$rounds+$acc7],$acc7 + fmovs %f0,%f0 + srl $t0,8,$acc10 ! + and $acc9,255,$acc9 + ldub [$rounds+$acc8],$acc8 + srl $t3,24,$acc12 + and $acc10,255,$acc10 + ldub [$rounds+$acc9],$acc9 + srl $t2,16,$acc13 + and $t3,255,$acc11 + ldub [$rounds+$acc10],$acc10 ! + srl $t1,8,$acc14 + and $acc13,255,$acc13 + ldub [$rounds+$acc11],$acc11 + ldub [$rounds+$acc12],$acc12 + and $acc14,255,$acc14 + ldub [$rounds+$acc13],$acc13 + and $t0,255,$acc15 + ldub [$rounds+$acc14],$acc14 ! + + sll $acc0,24,$acc0 + xor $acc3,$s0,$s0 + ldub [$rounds+$acc15],$acc15 + sll $acc1,16,$acc1 + xor $acc0,$s0,$s0 + ldx [%sp+$bias+$frame+0],%i7 ! restore return address + fmovs %f0,%f0 + sll $acc2,8,$acc2 ! + xor $acc1,$s0,$s0 + sll $acc4,24,$acc4 + xor $acc2,$s0,$s0 + sll $acc5,16,$acc5 + xor $acc7,$s1,$s1 + sll $acc6,8,$acc6 + xor $acc4,$s1,$s1 + sll $acc8,24,$acc8 ! + xor $acc5,$s1,$s1 + sll $acc9,16,$acc9 + xor $acc11,$s2,$s2 + sll $acc10,8,$acc10 + xor $acc6,$s1,$s1 + sll $acc12,24,$acc12 + xor $acc8,$s2,$s2 + sll $acc13,16,$acc13 ! + xor $acc9,$s2,$s2 + sll $acc14,8,$acc14 + xor $acc10,$s2,$s2 + xor $acc12,$acc14,$acc14 + xor $acc13,$s3,$s3 + xor $acc14,$s3,$s3 + xor $acc15,$s3,$s3 + + ret + restore +.type _sparcv9_AES_decrypt,#function +.size _sparcv9_AES_decrypt,(.-_sparcv9_AES_decrypt) + +.align 32 +.globl AES_decrypt +AES_decrypt: + or %o0,%o1,%g1 + andcc %g1,3,%g0 + bnz,pn %xcc,.Lunaligned_dec + save %sp,-$frame,%sp + + ld [%i0+0],%o0 + ld [%i0+4],%o1 + ld [%i0+8],%o2 + ld [%i0+12],%o3 + +1: call .+8 + add %o7,AES_Td-1b,%o4 + call _sparcv9_AES_decrypt + mov %i2,%o5 + + st %o0,[%i1+0] + st %o1,[%i1+4] + st %o2,[%i1+8] + st %o3,[%i1+12] + + ret + restore + +.align 32 +.Lunaligned_dec: + ldub [%i0+0],%l0 + ldub [%i0+1],%l1 + ldub [%i0+2],%l2 + + sll %l0,24,%l0 + ldub [%i0+3],%l3 + sll %l1,16,%l1 + ldub [%i0+4],%l4 + sll %l2,8,%l2 + or %l1,%l0,%l0 + ldub [%i0+5],%l5 + sll %l4,24,%l4 + or %l3,%l2,%l2 + ldub [%i0+6],%l6 + sll %l5,16,%l5 + or %l0,%l2,%o0 + ldub [%i0+7],%l7 + + sll %l6,8,%l6 + or %l5,%l4,%l4 + ldub [%i0+8],%l0 + or %l7,%l6,%l6 + ldub [%i0+9],%l1 + or %l4,%l6,%o1 + ldub [%i0+10],%l2 + + sll %l0,24,%l0 + ldub [%i0+11],%l3 + sll %l1,16,%l1 + ldub [%i0+12],%l4 + sll %l2,8,%l2 + or %l1,%l0,%l0 + ldub [%i0+13],%l5 + sll %l4,24,%l4 + or %l3,%l2,%l2 + ldub [%i0+14],%l6 + sll %l5,16,%l5 + or %l0,%l2,%o2 + ldub [%i0+15],%l7 + + sll %l6,8,%l6 + or %l5,%l4,%l4 + or %l7,%l6,%l6 + or %l4,%l6,%o3 + +1: call .+8 + add %o7,AES_Td-1b,%o4 + call _sparcv9_AES_decrypt + mov %i2,%o5 + + srl %o0,24,%l0 + srl %o0,16,%l1 + stb %l0,[%i1+0] + srl %o0,8,%l2 + stb %l1,[%i1+1] + stb %l2,[%i1+2] + srl %o1,24,%l4 + stb %o0,[%i1+3] + + srl %o1,16,%l5 + stb %l4,[%i1+4] + srl %o1,8,%l6 + stb %l5,[%i1+5] + stb %l6,[%i1+6] + srl %o2,24,%l0 + stb %o1,[%i1+7] + + srl %o2,16,%l1 + stb %l0,[%i1+8] + srl %o2,8,%l2 + stb %l1,[%i1+9] + stb %l2,[%i1+10] + srl %o3,24,%l4 + stb %o2,[%i1+11] + + srl %o3,16,%l5 + stb %l4,[%i1+12] + srl %o3,8,%l6 + stb %l5,[%i1+13] + stb %l6,[%i1+14] + stb %o3,[%i1+15] + + ret + restore +.type AES_decrypt,#function +.size AES_decrypt,(.-AES_decrypt) +___ + +# fmovs instructions substituting for FP nops were originally added +# to meet specific instruction alignment requirements to maximize ILP. +# As UltraSPARC T1, a.k.a. Niagara, has shared FPU, FP nops can have +# undesired effect, so just omit them and sacrifice some portion of +# percent in performance... +$code =~ s/fmovs.*$//gm; + +print $code; +close STDOUT; # ensure flush diff --git a/openssl-1.1.0h/crypto/aes/asm/aes-x86_64.pl b/openssl-1.1.0h/crypto/aes/asm/aes-x86_64.pl new file mode 100755 index 0000000..ce4ca30 --- /dev/null +++ b/openssl-1.1.0h/crypto/aes/asm/aes-x86_64.pl @@ -0,0 +1,2820 @@ +#! /usr/bin/env perl +# Copyright 2005-2016 The OpenSSL Project Authors. All Rights Reserved. +# +# Licensed under the OpenSSL license (the "License"). You may not use +# this file except in compliance with the License. You can obtain a copy +# in the file LICENSE in the source distribution or at +# https://www.openssl.org/source/license.html + +# +# ==================================================================== +# Written by Andy Polyakov for the OpenSSL +# project. The module is, however, dual licensed under OpenSSL and +# CRYPTOGAMS licenses depending on where you obtain it. For further +# details see http://www.openssl.org/~appro/cryptogams/. +# ==================================================================== +# +# Version 2.1. +# +# aes-*-cbc benchmarks are improved by >70% [compared to gcc 3.3.2 on +# Opteron 240 CPU] plus all the bells-n-whistles from 32-bit version +# [you'll notice a lot of resemblance], such as compressed S-boxes +# in little-endian byte order, prefetch of these tables in CBC mode, +# as well as avoiding L1 cache aliasing between stack frame and key +# schedule and already mentioned tables, compressed Td4... +# +# Performance in number of cycles per processed byte for 128-bit key: +# +# ECB encrypt ECB decrypt CBC large chunk +# AMD64 33 43 13.0 +# EM64T 38 56 18.6(*) +# Core 2 30 42 14.5(*) +# Atom 65 86 32.1(*) +# +# (*) with hyper-threading off + +$flavour = shift; +$output = shift; +if ($flavour =~ /\./) { $output = $flavour; undef $flavour; } + +$win64=0; $win64=1 if ($flavour =~ /[nm]asm|mingw64/ || $output =~ /\.asm$/); + +$0 =~ m/(.*[\/\\])[^\/\\]+$/; $dir=$1; +( $xlate="${dir}x86_64-xlate.pl" and -f $xlate ) or +( $xlate="${dir}../../perlasm/x86_64-xlate.pl" and -f $xlate) or +die "can't locate x86_64-xlate.pl"; + +open OUT,"| \"$^X\" \"$xlate\" $flavour \"$output\""; +*STDOUT=*OUT; + +$verticalspin=1; # unlike 32-bit version $verticalspin performs + # ~15% better on both AMD and Intel cores +$speed_limit=512; # see aes-586.pl for details + +$code=".text\n"; + +$s0="%eax"; +$s1="%ebx"; +$s2="%ecx"; +$s3="%edx"; +$acc0="%esi"; $mask80="%rsi"; +$acc1="%edi"; $maskfe="%rdi"; +$acc2="%ebp"; $mask1b="%rbp"; +$inp="%r8"; +$out="%r9"; +$t0="%r10d"; +$t1="%r11d"; +$t2="%r12d"; +$rnds="%r13d"; +$sbox="%r14"; +$key="%r15"; + +sub hi() { my $r=shift; $r =~ s/%[er]([a-d])x/%\1h/; $r; } +sub lo() { my $r=shift; $r =~ s/%[er]([a-d])x/%\1l/; + $r =~ s/%[er]([sd]i)/%\1l/; + $r =~ s/%(r[0-9]+)[d]?/%\1b/; $r; } +sub LO() { my $r=shift; $r =~ s/%r([a-z]+)/%e\1/; + $r =~ s/%r([0-9]+)/%r\1d/; $r; } +sub _data_word() +{ my $i; + while(defined($i=shift)) { $code.=sprintf".long\t0x%08x,0x%08x\n",$i,$i; } +} +sub data_word() +{ my $i; + my $last=pop(@_); + $code.=".long\t"; + while(defined($i=shift)) { $code.=sprintf"0x%08x,",$i; } + $code.=sprintf"0x%08x\n",$last; +} + +sub data_byte() +{ my $i; + my $last=pop(@_); + $code.=".byte\t"; + while(defined($i=shift)) { $code.=sprintf"0x%02x,",$i&0xff; } + $code.=sprintf"0x%02x\n",$last&0xff; +} + +sub encvert() +{ my $t3="%r8d"; # zaps $inp! + +$code.=<<___; + # favor 3-way issue Opteron pipeline... + movzb `&lo("$s0")`,$acc0 + movzb `&lo("$s1")`,$acc1 + movzb `&lo("$s2")`,$acc2 + mov 0($sbox,$acc0,8),$t0 + mov 0($sbox,$acc1,8),$t1 + mov 0($sbox,$acc2,8),$t2 + + movzb `&hi("$s1")`,$acc0 + movzb `&hi("$s2")`,$acc1 + movzb `&lo("$s3")`,$acc2 + xor 3($sbox,$acc0,8),$t0 + xor 3($sbox,$acc1,8),$t1 + mov 0($sbox,$acc2,8),$t3 + + movzb `&hi("$s3")`,$acc0 + shr \$16,$s2 + movzb `&hi("$s0")`,$acc2 + xor 3($sbox,$acc0,8),$t2 + shr \$16,$s3 + xor 3($sbox,$acc2,8),$t3 + + shr \$16,$s1 + lea 16($key),$key + shr \$16,$s0 + + movzb `&lo("$s2")`,$acc0 + movzb `&lo("$s3")`,$acc1 + movzb `&lo("$s0")`,$acc2 + xor 2($sbox,$acc0,8),$t0 + xor 2($sbox,$acc1,8),$t1 + xor 2($sbox,$acc2,8),$t2 + + movzb `&hi("$s3")`,$acc0 + movzb `&hi("$s0")`,$acc1 + movzb `&lo("$s1")`,$acc2 + xor 1($sbox,$acc0,8),$t0 + xor 1($sbox,$acc1,8),$t1 + xor 2($sbox,$acc2,8),$t3 + + mov 12($key),$s3 + movzb `&hi("$s1")`,$acc1 + movzb `&hi("$s2")`,$acc2 + mov 0($key),$s0 + xor 1($sbox,$acc1,8),$t2 + xor 1($sbox,$acc2,8),$t3 + + mov 4($key),$s1 + mov 8($key),$s2 + xor $t0,$s0 + xor $t1,$s1 + xor $t2,$s2 + xor $t3,$s3 +___ +} + +sub enclastvert() +{ my $t3="%r8d"; # zaps $inp! + +$code.=<<___; + movzb `&lo("$s0")`,$acc0 + movzb `&lo("$s1")`,$acc1 + movzb `&lo("$s2")`,$acc2 + movzb 2($sbox,$acc0,8),$t0 + movzb 2($sbox,$acc1,8),$t1 + movzb 2($sbox,$acc2,8),$t2 + + movzb `&lo("$s3")`,$acc0 + movzb `&hi("$s1")`,$acc1 + movzb `&hi("$s2")`,$acc2 + movzb 2($sbox,$acc0,8),$t3 + mov 0($sbox,$acc1,8),$acc1 #$t0 + mov 0($sbox,$acc2,8),$acc2 #$t1 + + and \$0x0000ff00,$acc1 + and \$0x0000ff00,$acc2 + + xor $acc1,$t0 + xor $acc2,$t1 + shr \$16,$s2 + + movzb `&hi("$s3")`,$acc0 + movzb `&hi("$s0")`,$acc1 + shr \$16,$s3 + mov 0($sbox,$acc0,8),$acc0 #$t2 + mov 0($sbox,$acc1,8),$acc1 #$t3 + + and \$0x0000ff00,$acc0 + and \$0x0000ff00,$acc1 + shr \$16,$s1 + xor $acc0,$t2 + xor $acc1,$t3 + shr \$16,$s0 + + movzb `&lo("$s2")`,$acc0 + movzb `&lo("$s3")`,$acc1 + movzb `&lo("$s0")`,$acc2 + mov 0($sbox,$acc0,8),$acc0 #$t0 + mov 0($sbox,$acc1,8),$acc1 #$t1 + mov 0($sbox,$acc2,8),$acc2 #$t2 + + and \$0x00ff0000,$acc0 + and \$0x00ff0000,$acc1 + and \$0x00ff0000,$acc2 + + xor $acc0,$t0 + xor $acc1,$t1 + xor $acc2,$t2 + + movzb `&lo("$s1")`,$acc0 + movzb `&hi("$s3")`,$acc1 + movzb `&hi("$s0")`,$acc2 + mov 0($sbox,$acc0,8),$acc0 #$t3 + mov 2($sbox,$acc1,8),$acc1 #$t0 + mov 2($sbox,$acc2,8),$acc2 #$t1 + + and \$0x00ff0000,$acc0 + and \$0xff000000,$acc1 + and \$0xff000000,$acc2 + + xor $acc0,$t3 + xor $acc1,$t0 + xor $acc2,$t1 + + movzb `&hi("$s1")`,$acc0 + movzb `&hi("$s2")`,$acc1 + mov 16+12($key),$s3 + mov 2($sbox,$acc0,8),$acc0 #$t2 + mov 2($sbox,$acc1,8),$acc1 #$t3 + mov 16+0($key),$s0 + + and \$0xff000000,$acc0 + and \$0xff000000,$acc1 + + xor $acc0,$t2 + xor $acc1,$t3 + + mov 16+4($key),$s1 + mov 16+8($key),$s2 + xor $t0,$s0 + xor $t1,$s1 + xor $t2,$s2 + xor $t3,$s3 +___ +} + +sub encstep() +{ my ($i,@s) = @_; + my $tmp0=$acc0; + my $tmp1=$acc1; + my $tmp2=$acc2; + my $out=($t0,$t1,$t2,$s[0])[$i]; + + if ($i==3) { + $tmp0=$s[1]; + $tmp1=$s[2]; + $tmp2=$s[3]; + } + $code.=" movzb ".&lo($s[0]).",$out\n"; + $code.=" mov $s[2],$tmp1\n" if ($i!=3); + $code.=" lea 16($key),$key\n" if ($i==0); + + $code.=" movzb ".&hi($s[1]).",$tmp0\n"; + $code.=" mov 0($sbox,$out,8),$out\n"; + + $code.=" shr \$16,$tmp1\n"; + $code.=" mov $s[3],$tmp2\n" if ($i!=3); + $code.=" xor 3($sbox,$tmp0,8),$out\n"; + + $code.=" movzb ".&lo($tmp1).",$tmp1\n"; + $code.=" shr \$24,$tmp2\n"; + $code.=" xor 4*$i($key),$out\n"; + + $code.=" xor 2($sbox,$tmp1,8),$out\n"; + $code.=" xor 1($sbox,$tmp2,8),$out\n"; + + $code.=" mov $t0,$s[1]\n" if ($i==3); + $code.=" mov $t1,$s[2]\n" if ($i==3); + $code.=" mov $t2,$s[3]\n" if ($i==3); + $code.="\n"; +} + +sub enclast() +{ my ($i,@s)=@_; + my $tmp0=$acc0; + my $tmp1=$acc1; + my $tmp2=$acc2; + my $out=($t0,$t1,$t2,$s[0])[$i]; + + if ($i==3) { + $tmp0=$s[1]; + $tmp1=$s[2]; + $tmp2=$s[3]; + } + $code.=" movzb ".&lo($s[0]).",$out\n"; + $code.=" mov $s[2],$tmp1\n" if ($i!=3); + + $code.=" mov 2($sbox,$out,8),$out\n"; + $code.=" shr \$16,$tmp1\n"; + $code.=" mov $s[3],$tmp2\n" if ($i!=3); + + $code.=" and \$0x000000ff,$out\n"; + $code.=" movzb ".&hi($s[1]).",$tmp0\n"; + $code.=" movzb ".&lo($tmp1).",$tmp1\n"; + $code.=" shr \$24,$tmp2\n"; + + $code.=" mov 0($sbox,$tmp0,8),$tmp0\n"; + $code.=" mov 0($sbox,$tmp1,8),$tmp1\n"; + $code.=" mov 2($sbox,$tmp2,8),$tmp2\n"; + + $code.=" and \$0x0000ff00,$tmp0\n"; + $code.=" and \$0x00ff0000,$tmp1\n"; + $code.=" and \$0xff000000,$tmp2\n"; + + $code.=" xor $tmp0,$out\n"; + $code.=" mov $t0,$s[1]\n" if ($i==3); + $code.=" xor $tmp1,$out\n"; + $code.=" mov $t1,$s[2]\n" if ($i==3); + $code.=" xor $tmp2,$out\n"; + $code.=" mov $t2,$s[3]\n" if ($i==3); + $code.="\n"; +} + +$code.=<<___; +.type _x86_64_AES_encrypt,\@abi-omnipotent +.align 16 +_x86_64_AES_encrypt: + xor 0($key),$s0 # xor with key + xor 4($key),$s1 + xor 8($key),$s2 + xor 12($key),$s3 + + mov 240($key),$rnds # load key->rounds + sub \$1,$rnds + jmp .Lenc_loop +.align 16 +.Lenc_loop: +___ + if ($verticalspin) { &encvert(); } + else { &encstep(0,$s0,$s1,$s2,$s3); + &encstep(1,$s1,$s2,$s3,$s0); + &encstep(2,$s2,$s3,$s0,$s1); + &encstep(3,$s3,$s0,$s1,$s2); + } +$code.=<<___; + sub \$1,$rnds + jnz .Lenc_loop +___ + if ($verticalspin) { &enclastvert(); } + else { &enclast(0,$s0,$s1,$s2,$s3); + &enclast(1,$s1,$s2,$s3,$s0); + &enclast(2,$s2,$s3,$s0,$s1); + &enclast(3,$s3,$s0,$s1,$s2); + $code.=<<___; + xor 16+0($key),$s0 # xor with key + xor 16+4($key),$s1 + xor 16+8($key),$s2 + xor 16+12($key),$s3 +___ + } +$code.=<<___; + .byte 0xf3,0xc3 # rep ret +.size _x86_64_AES_encrypt,.-_x86_64_AES_encrypt +___ + +# it's possible to implement this by shifting tN by 8, filling least +# significant byte with byte load and finally bswap-ing at the end, +# but such partial register load kills Core 2... +sub enccompactvert() +{ my ($t3,$t4,$t5)=("%r8d","%r9d","%r13d"); + +$code.=<<___; + movzb `&lo("$s0")`,$t0 + movzb `&lo("$s1")`,$t1 + movzb `&lo("$s2")`,$t2 + movzb `&lo("$s3")`,$t3 + movzb `&hi("$s1")`,$acc0 + movzb `&hi("$s2")`,$acc1 + shr \$16,$s2 + movzb `&hi("$s3")`,$acc2 + movzb ($sbox,$t0,1),$t0 + movzb ($sbox,$t1,1),$t1 + movzb ($sbox,$t2,1),$t2 + movzb ($sbox,$t3,1),$t3 + + movzb ($sbox,$acc0,1),$t4 #$t0 + movzb `&hi("$s0")`,$acc0 + movzb ($sbox,$acc1,1),$t5 #$t1 + movzb `&lo("$s2")`,$acc1 + movzb ($sbox,$acc2,1),$acc2 #$t2 + movzb ($sbox,$acc0,1),$acc0 #$t3 + + shl \$8,$t4 + shr \$16,$s3 + shl \$8,$t5 + xor $t4,$t0 + shr \$16,$s0 + movzb `&lo("$s3")`,$t4 + shr \$16,$s1 + xor $t5,$t1 + shl \$8,$acc2 + movzb `&lo("$s0")`,$t5 + movzb ($sbox,$acc1,1),$acc1 #$t0 + xor $acc2,$t2 + + shl \$8,$acc0 + movzb `&lo("$s1")`,$acc2 + shl \$16,$acc1 + xor $acc0,$t3 + movzb ($sbox,$t4,1),$t4 #$t1 + movzb `&hi("$s3")`,$acc0 + movzb ($sbox,$t5,1),$t5 #$t2 + xor $acc1,$t0 + + shr \$8,$s2 + movzb `&hi("$s0")`,$acc1 + shl \$16,$t4 + shr \$8,$s1 + shl \$16,$t5 + xor $t4,$t1 + movzb ($sbox,$acc2,1),$acc2 #$t3 + movzb ($sbox,$acc0,1),$acc0 #$t0 + movzb ($sbox,$acc1,1),$acc1 #$t1 + movzb ($sbox,$s2,1),$s3 #$t3 + movzb ($sbox,$s1,1),$s2 #$t2 + + shl \$16,$acc2 + xor $t5,$t2 + shl \$24,$acc0 + xor $acc2,$t3 + shl \$24,$acc1 + xor $acc0,$t0 + shl \$24,$s3 + xor $acc1,$t1 + shl \$24,$s2 + mov $t0,$s0 + mov $t1,$s1 + xor $t2,$s2 + xor $t3,$s3 +___ +} + +sub enctransform_ref() +{ my $sn = shift; + my ($acc,$r2,$tmp)=("%r8d","%r9d","%r13d"); + +$code.=<<___; + mov $sn,$acc + and \$0x80808080,$acc + mov $acc,$tmp + shr \$7,$tmp + lea ($sn,$sn),$r2 + sub $tmp,$acc + and \$0xfefefefe,$r2 + and \$0x1b1b1b1b,$acc + mov $sn,$tmp + xor $acc,$r2 + + xor $r2,$sn + rol \$24,$sn + xor $r2,$sn + ror \$16,$tmp + xor $tmp,$sn + ror \$8,$tmp + xor $tmp,$sn +___ +} + +# unlike decrypt case it does not pay off to parallelize enctransform +sub enctransform() +{ my ($t3,$r20,$r21)=($acc2,"%r8d","%r9d"); + +$code.=<<___; + mov \$0x80808080,$t0 + mov \$0x80808080,$t1 + and $s0,$t0 + and $s1,$t1 + mov $t0,$acc0 + mov $t1,$acc1 + shr \$7,$t0 + lea ($s0,$s0),$r20 + shr \$7,$t1 + lea ($s1,$s1),$r21 + sub $t0,$acc0 + sub $t1,$acc1 + and \$0xfefefefe,$r20 + and \$0xfefefefe,$r21 + and \$0x1b1b1b1b,$acc0 + and \$0x1b1b1b1b,$acc1 + mov $s0,$t0 + mov $s1,$t1 + xor $acc0,$r20 + xor $acc1,$r21 + + xor $r20,$s0 + xor $r21,$s1 + mov \$0x80808080,$t2 + rol \$24,$s0 + mov \$0x80808080,$t3 + rol \$24,$s1 + and $s2,$t2 + and $s3,$t3 + xor $r20,$s0 + xor $r21,$s1 + mov $t2,$acc0 + ror \$16,$t0 + mov $t3,$acc1 + ror \$16,$t1 + lea ($s2,$s2),$r20 + shr \$7,$t2 + xor $t0,$s0 + shr \$7,$t3 + xor $t1,$s1 + ror \$8,$t0 + lea ($s3,$s3),$r21 + ror \$8,$t1 + sub $t2,$acc0 + sub $t3,$acc1 + xor $t0,$s0 + xor $t1,$s1 + + and \$0xfefefefe,$r20 + and \$0xfefefefe,$r21 + and \$0x1b1b1b1b,$acc0 + and \$0x1b1b1b1b,$acc1 + mov $s2,$t2 + mov $s3,$t3 + xor $acc0,$r20 + xor $acc1,$r21 + + ror \$16,$t2 + xor $r20,$s2 + ror \$16,$t3 + xor $r21,$s3 + rol \$24,$s2 + mov 0($sbox),$acc0 # prefetch Te4 + rol \$24,$s3 + xor $r20,$s2 + mov 64($sbox),$acc1 + xor $r21,$s3 + mov 128($sbox),$r20 + xor $t2,$s2 + ror \$8,$t2 + xor $t3,$s3 + ror \$8,$t3 + xor $t2,$s2 + mov 192($sbox),$r21 + xor $t3,$s3 +___ +} + +$code.=<<___; +.type _x86_64_AES_encrypt_compact,\@abi-omnipotent +.align 16 +_x86_64_AES_encrypt_compact: + lea 128($sbox),$inp # size optimization + mov 0-128($inp),$acc1 # prefetch Te4 + mov 32-128($inp),$acc2 + mov 64-128($inp),$t0 + mov 96-128($inp),$t1 + mov 128-128($inp),$acc1 + mov 160-128($inp),$acc2 + mov 192-128($inp),$t0 + mov 224-128($inp),$t1 + jmp .Lenc_loop_compact +.align 16 +.Lenc_loop_compact: + xor 0($key),$s0 # xor with key + xor 4($key),$s1 + xor 8($key),$s2 + xor 12($key),$s3 + lea 16($key),$key +___ + &enccompactvert(); +$code.=<<___; + cmp 16(%rsp),$key + je .Lenc_compact_done +___ + &enctransform(); +$code.=<<___; + jmp .Lenc_loop_compact +.align 16 +.Lenc_compact_done: + xor 0($key),$s0 + xor 4($key),$s1 + xor 8($key),$s2 + xor 12($key),$s3 + .byte 0xf3,0xc3 # rep ret +.size _x86_64_AES_encrypt_compact,.-_x86_64_AES_encrypt_compact +___ + +# void AES_encrypt (const void *inp,void *out,const AES_KEY *key); +$code.=<<___; +.globl AES_encrypt +.type AES_encrypt,\@function,3 +.align 16 +.globl asm_AES_encrypt +.hidden asm_AES_encrypt +asm_AES_encrypt: +AES_encrypt: + push %rbx + push %rbp + push %r12 + push %r13 + push %r14 + push %r15 + + # allocate frame "above" key schedule + mov %rsp,%r10 + lea -63(%rdx),%rcx # %rdx is key argument + and \$-64,%rsp + sub %rsp,%rcx + neg %rcx + and \$0x3c0,%rcx + sub %rcx,%rsp + sub \$32,%rsp + + mov %rsi,16(%rsp) # save out + mov %r10,24(%rsp) # save real stack pointer +.Lenc_prologue: + + mov %rdx,$key + mov 240($key),$rnds # load rounds + + mov 0(%rdi),$s0 # load input vector + mov 4(%rdi),$s1 + mov 8(%rdi),$s2 + mov 12(%rdi),$s3 + + shl \$4,$rnds + lea ($key,$rnds),%rbp + mov $key,(%rsp) # key schedule + mov %rbp,8(%rsp) # end of key schedule + + # pick Te4 copy which can't "overlap" with stack frame or key schedule + lea .LAES_Te+2048(%rip),$sbox + lea 768(%rsp),%rbp + sub $sbox,%rbp + and \$0x300,%rbp + lea ($sbox,%rbp),$sbox + + call _x86_64_AES_encrypt_compact + + mov 16(%rsp),$out # restore out + mov 24(%rsp),%rsi # restore saved stack pointer + mov $s0,0($out) # write output vector + mov $s1,4($out) + mov $s2,8($out) + mov $s3,12($out) + + mov (%rsi),%r15 + mov 8(%rsi),%r14 + mov 16(%rsi),%r13 + mov 24(%rsi),%r12 + mov 32(%rsi),%rbp + mov 40(%rsi),%rbx + lea 48(%rsi),%rsp +.Lenc_epilogue: + ret +.size AES_encrypt,.-AES_encrypt +___ + +#------------------------------------------------------------------# + +sub decvert() +{ my $t3="%r8d"; # zaps $inp! + +$code.=<<___; + # favor 3-way issue Opteron pipeline... + movzb `&lo("$s0")`,$acc0 + movzb `&lo("$s1")`,$acc1 + movzb `&lo("$s2")`,$acc2 + mov 0($sbox,$acc0,8),$t0 + mov 0($sbox,$acc1,8),$t1 + mov 0($sbox,$acc2,8),$t2 + + movzb `&hi("$s3")`,$acc0 + movzb `&hi("$s0")`,$acc1 + movzb `&lo("$s3")`,$acc2 + xor 3($sbox,$acc0,8),$t0 + xor 3($sbox,$acc1,8),$t1 + mov 0($sbox,$acc2,8),$t3 + + movzb `&hi("$s1")`,$acc0 + shr \$16,$s0 + movzb `&hi("$s2")`,$acc2 + xor 3($sbox,$acc0,8),$t2 + shr \$16,$s3 + xor 3($sbox,$acc2,8),$t3 + + shr \$16,$s1 + lea 16($key),$key + shr \$16,$s2 + + movzb `&lo("$s2")`,$acc0 + movzb `&lo("$s3")`,$acc1 + movzb `&lo("$s0")`,$acc2 + xor 2($sbox,$acc0,8),$t0 + xor 2($sbox,$acc1,8),$t1 + xor 2($sbox,$acc2,8),$t2 + + movzb `&hi("$s1")`,$acc0 + movzb `&hi("$s2")`,$acc1 + movzb `&lo("$s1")`,$acc2 + xor 1($sbox,$acc0,8),$t0 + xor 1($sbox,$acc1,8),$t1 + xor 2($sbox,$acc2,8),$t3 + + movzb `&hi("$s3")`,$acc0 + mov 12($key),$s3 + movzb `&hi("$s0")`,$acc2 + xor 1($sbox,$acc0,8),$t2 + mov 0($key),$s0 + xor 1($sbox,$acc2,8),$t3 + + xor $t0,$s0 + mov 4($key),$s1 + mov 8($key),$s2 + xor $t2,$s2 + xor $t1,$s1 + xor $t3,$s3 +___ +} + +sub declastvert() +{ my $t3="%r8d"; # zaps $inp! + +$code.=<<___; + lea 2048($sbox),$sbox # size optimization + movzb `&lo("$s0")`,$acc0 + movzb `&lo("$s1")`,$acc1 + movzb `&lo("$s2")`,$acc2 + movzb ($sbox,$acc0,1),$t0 + movzb ($sbox,$acc1,1),$t1 + movzb ($sbox,$acc2,1),$t2 + + movzb `&lo("$s3")`,$acc0 + movzb `&hi("$s3")`,$acc1 + movzb `&hi("$s0")`,$acc2 + movzb ($sbox,$acc0,1),$t3 + movzb ($sbox,$acc1,1),$acc1 #$t0 + movzb ($sbox,$acc2,1),$acc2 #$t1 + + shl \$8,$acc1 + shl \$8,$acc2 + + xor $acc1,$t0 + xor $acc2,$t1 + shr \$16,$s3 + + movzb `&hi("$s1")`,$acc0 + movzb `&hi("$s2")`,$acc1 + shr \$16,$s0 + movzb ($sbox,$acc0,1),$acc0 #$t2 + movzb ($sbox,$acc1,1),$acc1 #$t3 + + shl \$8,$acc0 + shl \$8,$acc1 + shr \$16,$s1 + xor $acc0,$t2 + xor $acc1,$t3 + shr \$16,$s2 + + movzb `&lo("$s2")`,$acc0 + movzb `&lo("$s3")`,$acc1 + movzb `&lo("$s0")`,$acc2 + movzb ($sbox,$acc0,1),$acc0 #$t0 + movzb ($sbox,$acc1,1),$acc1 #$t1 + movzb ($sbox,$acc2,1),$acc2 #$t2 + + shl \$16,$acc0 + shl \$16,$acc1 + shl \$16,$acc2 + + xor $acc0,$t0 + xor $acc1,$t1 + xor $acc2,$t2 + + movzb `&lo("$s1")`,$acc0 + movzb `&hi("$s1")`,$acc1 + movzb `&hi("$s2")`,$acc2 + movzb ($sbox,$acc0,1),$acc0 #$t3 + movzb ($sbox,$acc1,1),$acc1 #$t0 + movzb ($sbox,$acc2,1),$acc2 #$t1 + + shl \$16,$acc0 + shl \$24,$acc1 + shl \$24,$acc2 + + xor $acc0,$t3 + xor $acc1,$t0 + xor $acc2,$t1 + + movzb `&hi("$s3")`,$acc0 + movzb `&hi("$s0")`,$acc1 + mov 16+12($key),$s3 + movzb ($sbox,$acc0,1),$acc0 #$t2 + movzb ($sbox,$acc1,1),$acc1 #$t3 + mov 16+0($key),$s0 + + shl \$24,$acc0 + shl \$24,$acc1 + + xor $acc0,$t2 + xor $acc1,$t3 + + mov 16+4($key),$s1 + mov 16+8($key),$s2 + lea -2048($sbox),$sbox + xor $t0,$s0 + xor $t1,$s1 + xor $t2,$s2 + xor $t3,$s3 +___ +} + +sub decstep() +{ my ($i,@s) = @_; + my $tmp0=$acc0; + my $tmp1=$acc1; + my $tmp2=$acc2; + my $out=($t0,$t1,$t2,$s[0])[$i]; + + $code.=" mov $s[0],$out\n" if ($i!=3); + $tmp1=$s[2] if ($i==3); + $code.=" mov $s[2],$tmp1\n" if ($i!=3); + $code.=" and \$0xFF,$out\n"; + + $code.=" mov 0($sbox,$out,8),$out\n"; + $code.=" shr \$16,$tmp1\n"; + $tmp2=$s[3] if ($i==3); + $code.=" mov $s[3],$tmp2\n" if ($i!=3); + + $tmp0=$s[1] if ($i==3); + $code.=" movzb ".&hi($s[1]).",$tmp0\n"; + $code.=" and \$0xFF,$tmp1\n"; + $code.=" shr \$24,$tmp2\n"; + + $code.=" xor 3($sbox,$tmp0,8),$out\n"; + $code.=" xor 2($sbox,$tmp1,8),$out\n"; + $code.=" xor 1($sbox,$tmp2,8),$out\n"; + + $code.=" mov $t2,$s[1]\n" if ($i==3); + $code.=" mov $t1,$s[2]\n" if ($i==3); + $code.=" mov $t0,$s[3]\n" if ($i==3); + $code.="\n"; +} + +sub declast() +{ my ($i,@s)=@_; + my $tmp0=$acc0; + my $tmp1=$acc1; + my $tmp2=$acc2; + my $out=($t0,$t1,$t2,$s[0])[$i]; + + $code.=" mov $s[0],$out\n" if ($i!=3); + $tmp1=$s[2] if ($i==3); + $code.=" mov $s[2],$tmp1\n" if ($i!=3); + $code.=" and \$0xFF,$out\n"; + + $code.=" movzb 2048($sbox,$out,1),$out\n"; + $code.=" shr \$16,$tmp1\n"; + $tmp2=$s[3] if ($i==3); + $code.=" mov $s[3],$tmp2\n" if ($i!=3); + + $tmp0=$s[1] if ($i==3); + $code.=" movzb ".&hi($s[1]).",$tmp0\n"; + $code.=" and \$0xFF,$tmp1\n"; + $code.=" shr \$24,$tmp2\n"; + + $code.=" movzb 2048($sbox,$tmp0,1),$tmp0\n"; + $code.=" movzb 2048($sbox,$tmp1,1),$tmp1\n"; + $code.=" movzb 2048($sbox,$tmp2,1),$tmp2\n"; + + $code.=" shl \$8,$tmp0\n"; + $code.=" shl \$16,$tmp1\n"; + $code.=" shl \$24,$tmp2\n"; + + $code.=" xor $tmp0,$out\n"; + $code.=" mov $t2,$s[1]\n" if ($i==3); + $code.=" xor $tmp1,$out\n"; + $code.=" mov $t1,$s[2]\n" if ($i==3); + $code.=" xor $tmp2,$out\n"; + $code.=" mov $t0,$s[3]\n" if ($i==3); + $code.="\n"; +} + +$code.=<<___; +.type _x86_64_AES_decrypt,\@abi-omnipotent +.align 16 +_x86_64_AES_decrypt: + xor 0($key),$s0 # xor with key + xor 4($key),$s1 + xor 8($key),$s2 + xor 12($key),$s3 + + mov 240($key),$rnds # load key->rounds + sub \$1,$rnds + jmp .Ldec_loop +.align 16 +.Ldec_loop: +___ + if ($verticalspin) { &decvert(); } + else { &decstep(0,$s0,$s3,$s2,$s1); + &decstep(1,$s1,$s0,$s3,$s2); + &decstep(2,$s2,$s1,$s0,$s3); + &decstep(3,$s3,$s2,$s1,$s0); + $code.=<<___; + lea 16($key),$key + xor 0($key),$s0 # xor with key + xor 4($key),$s1 + xor 8($key),$s2 + xor 12($key),$s3 +___ + } +$code.=<<___; + sub \$1,$rnds + jnz .Ldec_loop +___ + if ($verticalspin) { &declastvert(); } + else { &declast(0,$s0,$s3,$s2,$s1); + &declast(1,$s1,$s0,$s3,$s2); + &declast(2,$s2,$s1,$s0,$s3); + &declast(3,$s3,$s2,$s1,$s0); + $code.=<<___; + xor 16+0($key),$s0 # xor with key + xor 16+4($key),$s1 + xor 16+8($key),$s2 + xor 16+12($key),$s3 +___ + } +$code.=<<___; + .byte 0xf3,0xc3 # rep ret +.size _x86_64_AES_decrypt,.-_x86_64_AES_decrypt +___ + +sub deccompactvert() +{ my ($t3,$t4,$t5)=("%r8d","%r9d","%r13d"); + +$code.=<<___; + movzb `&lo("$s0")`,$t0 + movzb `&lo("$s1")`,$t1 + movzb `&lo("$s2")`,$t2 + movzb `&lo("$s3")`,$t3 + movzb `&hi("$s3")`,$acc0 + movzb `&hi("$s0")`,$acc1 + shr \$16,$s3 + movzb `&hi("$s1")`,$acc2 + movzb ($sbox,$t0,1),$t0 + movzb ($sbox,$t1,1),$t1 + movzb ($sbox,$t2,1),$t2 + movzb ($sbox,$t3,1),$t3 + + movzb ($sbox,$acc0,1),$t4 #$t0 + movzb `&hi("$s2")`,$acc0 + movzb ($sbox,$acc1,1),$t5 #$t1 + movzb ($sbox,$acc2,1),$acc2 #$t2 + movzb ($sbox,$acc0,1),$acc0 #$t3 + + shr \$16,$s2 + shl \$8,$t5 + shl \$8,$t4 + movzb `&lo("$s2")`,$acc1 + shr \$16,$s0 + xor $t4,$t0 + shr \$16,$s1 + movzb `&lo("$s3")`,$t4 + + shl \$8,$acc2 + xor $t5,$t1 + shl \$8,$acc0 + movzb `&lo("$s0")`,$t5 + movzb ($sbox,$acc1,1),$acc1 #$t0 + xor $acc2,$t2 + movzb `&lo("$s1")`,$acc2 + + shl \$16,$acc1 + xor $acc0,$t3 + movzb ($sbox,$t4,1),$t4 #$t1 + movzb `&hi("$s1")`,$acc0 + movzb ($sbox,$acc2,1),$acc2 #$t3 + xor $acc1,$t0 + movzb ($sbox,$t5,1),$t5 #$t2 + movzb `&hi("$s2")`,$acc1 + + shl \$16,$acc2 + shl \$16,$t4 + shl \$16,$t5 + xor $acc2,$t3 + movzb `&hi("$s3")`,$acc2 + xor $t4,$t1 + shr \$8,$s0 + xor $t5,$t2 + + movzb ($sbox,$acc0,1),$acc0 #$t0 + movzb ($sbox,$acc1,1),$s1 #$t1 + movzb ($sbox,$acc2,1),$s2 #$t2 + movzb ($sbox,$s0,1),$s3 #$t3 + + mov $t0,$s0 + shl \$24,$acc0 + shl \$24,$s1 + shl \$24,$s2 + xor $acc0,$s0 + shl \$24,$s3 + xor $t1,$s1 + xor $t2,$s2 + xor $t3,$s3 +___ +} + +# parallelized version! input is pair of 64-bit values: %rax=s1.s0 +# and %rcx=s3.s2, output is four 32-bit values in %eax=s0, %ebx=s1, +# %ecx=s2 and %edx=s3. +sub dectransform() +{ my ($tp10,$tp20,$tp40,$tp80,$acc0)=("%rax","%r8", "%r9", "%r10","%rbx"); + my ($tp18,$tp28,$tp48,$tp88,$acc8)=("%rcx","%r11","%r12","%r13","%rdx"); + my $prefetch = shift; + +$code.=<<___; + mov $mask80,$tp40 + mov $mask80,$tp48 + and $tp10,$tp40 + and $tp18,$tp48 + mov $tp40,$acc0 + mov $tp48,$acc8 + shr \$7,$tp40 + lea ($tp10,$tp10),$tp20 + shr \$7,$tp48 + lea ($tp18,$tp18),$tp28 + sub $tp40,$acc0 + sub $tp48,$acc8 + and $maskfe,$tp20 + and $maskfe,$tp28 + and $mask1b,$acc0 + and $mask1b,$acc8 + xor $acc0,$tp20 + xor $acc8,$tp28 + mov $mask80,$tp80 + mov $mask80,$tp88 + + and $tp20,$tp80 + and $tp28,$tp88 + mov $tp80,$acc0 + mov $tp88,$acc8 + shr \$7,$tp80 + lea ($tp20,$tp20),$tp40 + shr \$7,$tp88 + lea ($tp28,$tp28),$tp48 + sub $tp80,$acc0 + sub $tp88,$acc8 + and $maskfe,$tp40 + and $maskfe,$tp48 + and $mask1b,$acc0 + and $mask1b,$acc8 + xor $acc0,$tp40 + xor $acc8,$tp48 + mov $mask80,$tp80 + mov $mask80,$tp88 + + and $tp40,$tp80 + and $tp48,$tp88 + mov $tp80,$acc0 + mov $tp88,$acc8 + shr \$7,$tp80 + xor $tp10,$tp20 # tp2^=tp1 + shr \$7,$tp88 + xor $tp18,$tp28 # tp2^=tp1 + sub $tp80,$acc0 + sub $tp88,$acc8 + lea ($tp40,$tp40),$tp80 + lea ($tp48,$tp48),$tp88 + xor $tp10,$tp40 # tp4^=tp1 + xor $tp18,$tp48 # tp4^=tp1 + and $maskfe,$tp80 + and $maskfe,$tp88 + and $mask1b,$acc0 + and $mask1b,$acc8 + xor $acc0,$tp80 + xor $acc8,$tp88 + + xor $tp80,$tp10 # tp1^=tp8 + xor $tp88,$tp18 # tp1^=tp8 + xor $tp80,$tp20 # tp2^tp1^=tp8 + xor $tp88,$tp28 # tp2^tp1^=tp8 + mov $tp10,$acc0 + mov $tp18,$acc8 + xor $tp80,$tp40 # tp4^tp1^=tp8 + shr \$32,$acc0 + xor $tp88,$tp48 # tp4^tp1^=tp8 + shr \$32,$acc8 + xor $tp20,$tp80 # tp8^=tp8^tp2^tp1=tp2^tp1 + rol \$8,`&LO("$tp10")` # ROTATE(tp1^tp8,8) + xor $tp28,$tp88 # tp8^=tp8^tp2^tp1=tp2^tp1 + rol \$8,`&LO("$tp18")` # ROTATE(tp1^tp8,8) + xor $tp40,$tp80 # tp2^tp1^=tp8^tp4^tp1=tp8^tp4^tp2 + rol \$8,`&LO("$acc0")` # ROTATE(tp1^tp8,8) + xor $tp48,$tp88 # tp2^tp1^=tp8^tp4^tp1=tp8^tp4^tp2 + + rol \$8,`&LO("$acc8")` # ROTATE(tp1^tp8,8) + xor `&LO("$tp80")`,`&LO("$tp10")` + shr \$32,$tp80 + xor `&LO("$tp88")`,`&LO("$tp18")` + shr \$32,$tp88 + xor `&LO("$tp80")`,`&LO("$acc0")` + xor `&LO("$tp88")`,`&LO("$acc8")` + + mov $tp20,$tp80 + rol \$24,`&LO("$tp20")` # ROTATE(tp2^tp1^tp8,24) + mov $tp28,$tp88 + rol \$24,`&LO("$tp28")` # ROTATE(tp2^tp1^tp8,24) + shr \$32,$tp80 + xor `&LO("$tp20")`,`&LO("$tp10")` + shr \$32,$tp88 + xor `&LO("$tp28")`,`&LO("$tp18")` + rol \$24,`&LO("$tp80")` # ROTATE(tp2^tp1^tp8,24) + mov $tp40,$tp20 + rol \$24,`&LO("$tp88")` # ROTATE(tp2^tp1^tp8,24) + mov $tp48,$tp28 + shr \$32,$tp20 + xor `&LO("$tp80")`,`&LO("$acc0")` + shr \$32,$tp28 + xor `&LO("$tp88")`,`&LO("$acc8")` + + `"mov 0($sbox),$mask80" if ($prefetch)` + rol \$16,`&LO("$tp40")` # ROTATE(tp4^tp1^tp8,16) + `"mov 64($sbox),$maskfe" if ($prefetch)` + rol \$16,`&LO("$tp48")` # ROTATE(tp4^tp1^tp8,16) + `"mov 128($sbox),$mask1b" if ($prefetch)` + rol \$16,`&LO("$tp20")` # ROTATE(tp4^tp1^tp8,16) + `"mov 192($sbox),$tp80" if ($prefetch)` + xor `&LO("$tp40")`,`&LO("$tp10")` + rol \$16,`&LO("$tp28")` # ROTATE(tp4^tp1^tp8,16) + xor `&LO("$tp48")`,`&LO("$tp18")` + `"mov 256($sbox),$tp88" if ($prefetch)` + xor `&LO("$tp20")`,`&LO("$acc0")` + xor `&LO("$tp28")`,`&LO("$acc8")` +___ +} + +$code.=<<___; +.type _x86_64_AES_decrypt_compact,\@abi-omnipotent +.align 16 +_x86_64_AES_decrypt_compact: + lea 128($sbox),$inp # size optimization + mov 0-128($inp),$acc1 # prefetch Td4 + mov 32-128($inp),$acc2 + mov 64-128($inp),$t0 + mov 96-128($inp),$t1 + mov 128-128($inp),$acc1 + mov 160-128($inp),$acc2 + mov 192-128($inp),$t0 + mov 224-128($inp),$t1 + jmp .Ldec_loop_compact + +.align 16 +.Ldec_loop_compact: + xor 0($key),$s0 # xor with key + xor 4($key),$s1 + xor 8($key),$s2 + xor 12($key),$s3 + lea 16($key),$key +___ + &deccompactvert(); +$code.=<<___; + cmp 16(%rsp),$key + je .Ldec_compact_done + + mov 256+0($sbox),$mask80 + shl \$32,%rbx + shl \$32,%rdx + mov 256+8($sbox),$maskfe + or %rbx,%rax + or %rdx,%rcx + mov 256+16($sbox),$mask1b +___ + &dectransform(1); +$code.=<<___; + jmp .Ldec_loop_compact +.align 16 +.Ldec_compact_done: + xor 0($key),$s0 + xor 4($key),$s1 + xor 8($key),$s2 + xor 12($key),$s3 + .byte 0xf3,0xc3 # rep ret +.size _x86_64_AES_decrypt_compact,.-_x86_64_AES_decrypt_compact +___ + +# void AES_decrypt (const void *inp,void *out,const AES_KEY *key); +$code.=<<___; +.globl AES_decrypt +.type AES_decrypt,\@function,3 +.align 16 +.globl asm_AES_decrypt +.hidden asm_AES_decrypt +asm_AES_decrypt: +AES_decrypt: + push %rbx + push %rbp + push %r12 + push %r13 + push %r14 + push %r15 + + # allocate frame "above" key schedule + mov %rsp,%r10 + lea -63(%rdx),%rcx # %rdx is key argument + and \$-64,%rsp + sub %rsp,%rcx + neg %rcx + and \$0x3c0,%rcx + sub %rcx,%rsp + sub \$32,%rsp + + mov %rsi,16(%rsp) # save out + mov %r10,24(%rsp) # save real stack pointer +.Ldec_prologue: + + mov %rdx,$key + mov 240($key),$rnds # load rounds + + mov 0(%rdi),$s0 # load input vector + mov 4(%rdi),$s1 + mov 8(%rdi),$s2 + mov 12(%rdi),$s3 + + shl \$4,$rnds + lea ($key,$rnds),%rbp + mov $key,(%rsp) # key schedule + mov %rbp,8(%rsp) # end of key schedule + + # pick Td4 copy which can't "overlap" with stack frame or key schedule + lea .LAES_Td+2048(%rip),$sbox + lea 768(%rsp),%rbp + sub $sbox,%rbp + and \$0x300,%rbp + lea ($sbox,%rbp),$sbox + shr \$3,%rbp # recall "magic" constants! + add %rbp,$sbox + + call _x86_64_AES_decrypt_compact + + mov 16(%rsp),$out # restore out + mov 24(%rsp),%rsi # restore saved stack pointer + mov $s0,0($out) # write output vector + mov $s1,4($out) + mov $s2,8($out) + mov $s3,12($out) + + mov (%rsi),%r15 + mov 8(%rsi),%r14 + mov 16(%rsi),%r13 + mov 24(%rsi),%r12 + mov 32(%rsi),%rbp + mov 40(%rsi),%rbx + lea 48(%rsi),%rsp +.Ldec_epilogue: + ret +.size AES_decrypt,.-AES_decrypt +___ +#------------------------------------------------------------------# + +sub enckey() +{ +$code.=<<___; + movz %dl,%esi # rk[i]>>0 + movzb -128(%rbp,%rsi),%ebx + movz %dh,%esi # rk[i]>>8 + shl \$24,%ebx + xor %ebx,%eax + + movzb -128(%rbp,%rsi),%ebx + shr \$16,%edx + movz %dl,%esi # rk[i]>>16 + xor %ebx,%eax + + movzb -128(%rbp,%rsi),%ebx + movz %dh,%esi # rk[i]>>24 + shl \$8,%ebx + xor %ebx,%eax + + movzb -128(%rbp,%rsi),%ebx + shl \$16,%ebx + xor %ebx,%eax + + xor 1024-128(%rbp,%rcx,4),%eax # rcon +___ +} + +# int AES_set_encrypt_key(const unsigned char *userKey, const int bits, +# AES_KEY *key) +$code.=<<___; +.globl AES_set_encrypt_key +.type AES_set_encrypt_key,\@function,3 +.align 16 +AES_set_encrypt_key: + push %rbx + push %rbp + push %r12 # redundant, but allows to share + push %r13 # exception handler... + push %r14 + push %r15 + sub \$8,%rsp +.Lenc_key_prologue: + + call _x86_64_AES_set_encrypt_key + + mov 40(%rsp),%rbp + mov 48(%rsp),%rbx + add \$56,%rsp +.Lenc_key_epilogue: + ret +.size AES_set_encrypt_key,.-AES_set_encrypt_key + +.type _x86_64_AES_set_encrypt_key,\@abi-omnipotent +.align 16 +_x86_64_AES_set_encrypt_key: + mov %esi,%ecx # %ecx=bits + mov %rdi,%rsi # %rsi=userKey + mov %rdx,%rdi # %rdi=key + + test \$-1,%rsi + jz .Lbadpointer + test \$-1,%rdi + jz .Lbadpointer + + lea .LAES_Te(%rip),%rbp + lea 2048+128(%rbp),%rbp + + # prefetch Te4 + mov 0-128(%rbp),%eax + mov 32-128(%rbp),%ebx + mov 64-128(%rbp),%r8d + mov 96-128(%rbp),%edx + mov 128-128(%rbp),%eax + mov 160-128(%rbp),%ebx + mov 192-128(%rbp),%r8d + mov 224-128(%rbp),%edx + + cmp \$128,%ecx + je .L10rounds + cmp \$192,%ecx + je .L12rounds + cmp \$256,%ecx + je .L14rounds + mov \$-2,%rax # invalid number of bits + jmp .Lexit + +.L10rounds: + mov 0(%rsi),%rax # copy first 4 dwords + mov 8(%rsi),%rdx + mov %rax,0(%rdi) + mov %rdx,8(%rdi) + + shr \$32,%rdx + xor %ecx,%ecx + jmp .L10shortcut +.align 4 +.L10loop: + mov 0(%rdi),%eax # rk[0] + mov 12(%rdi),%edx # rk[3] +.L10shortcut: +___ + &enckey (); +$code.=<<___; + mov %eax,16(%rdi) # rk[4] + xor 4(%rdi),%eax + mov %eax,20(%rdi) # rk[5] + xor 8(%rdi),%eax + mov %eax,24(%rdi) # rk[6] + xor 12(%rdi),%eax + mov %eax,28(%rdi) # rk[7] + add \$1,%ecx + lea 16(%rdi),%rdi + cmp \$10,%ecx + jl .L10loop + + movl \$10,80(%rdi) # setup number of rounds + xor %rax,%rax + jmp .Lexit + +.L12rounds: + mov 0(%rsi),%rax # copy first 6 dwords + mov 8(%rsi),%rbx + mov 16(%rsi),%rdx + mov %rax,0(%rdi) + mov %rbx,8(%rdi) + mov %rdx,16(%rdi) + + shr \$32,%rdx + xor %ecx,%ecx + jmp .L12shortcut +.align 4 +.L12loop: + mov 0(%rdi),%eax # rk[0] + mov 20(%rdi),%edx # rk[5] +.L12shortcut: +___ + &enckey (); +$code.=<<___; + mov %eax,24(%rdi) # rk[6] + xor 4(%rdi),%eax + mov %eax,28(%rdi) # rk[7] + xor 8(%rdi),%eax + mov %eax,32(%rdi) # rk[8] + xor 12(%rdi),%eax + mov %eax,36(%rdi) # rk[9] + + cmp \$7,%ecx + je .L12break + add \$1,%ecx + + xor 16(%rdi),%eax + mov %eax,40(%rdi) # rk[10] + xor 20(%rdi),%eax + mov %eax,44(%rdi) # rk[11] + + lea 24(%rdi),%rdi + jmp .L12loop +.L12break: + movl \$12,72(%rdi) # setup number of rounds + xor %rax,%rax + jmp .Lexit + +.L14rounds: + mov 0(%rsi),%rax # copy first 8 dwords + mov 8(%rsi),%rbx + mov 16(%rsi),%rcx + mov 24(%rsi),%rdx + mov %rax,0(%rdi) + mov %rbx,8(%rdi) + mov %rcx,16(%rdi) + mov %rdx,24(%rdi) + + shr \$32,%rdx + xor %ecx,%ecx + jmp .L14shortcut +.align 4 +.L14loop: + mov 0(%rdi),%eax # rk[0] + mov 28(%rdi),%edx # rk[4] +.L14shortcut: +___ + &enckey (); +$code.=<<___; + mov %eax,32(%rdi) # rk[8] + xor 4(%rdi),%eax + mov %eax,36(%rdi) # rk[9] + xor 8(%rdi),%eax + mov %eax,40(%rdi) # rk[10] + xor 12(%rdi),%eax + mov %eax,44(%rdi) # rk[11] + + cmp \$6,%ecx + je .L14break + add \$1,%ecx + + mov %eax,%edx + mov 16(%rdi),%eax # rk[4] + movz %dl,%esi # rk[11]>>0 + movzb -128(%rbp,%rsi),%ebx + movz %dh,%esi # rk[11]>>8 + xor %ebx,%eax + + movzb -128(%rbp,%rsi),%ebx + shr \$16,%edx + shl \$8,%ebx + movz %dl,%esi # rk[11]>>16 + xor %ebx,%eax + + movzb -128(%rbp,%rsi),%ebx + movz %dh,%esi # rk[11]>>24 + shl \$16,%ebx + xor %ebx,%eax + + movzb -128(%rbp,%rsi),%ebx + shl \$24,%ebx + xor %ebx,%eax + + mov %eax,48(%rdi) # rk[12] + xor 20(%rdi),%eax + mov %eax,52(%rdi) # rk[13] + xor 24(%rdi),%eax + mov %eax,56(%rdi) # rk[14] + xor 28(%rdi),%eax + mov %eax,60(%rdi) # rk[15] + + lea 32(%rdi),%rdi + jmp .L14loop +.L14break: + movl \$14,48(%rdi) # setup number of rounds + xor %rax,%rax + jmp .Lexit + +.Lbadpointer: + mov \$-1,%rax +.Lexit: + .byte 0xf3,0xc3 # rep ret +.size _x86_64_AES_set_encrypt_key,.-_x86_64_AES_set_encrypt_key +___ + +sub deckey_ref() +{ my ($i,$ptr,$te,$td) = @_; + my ($tp1,$tp2,$tp4,$tp8,$acc)=("%eax","%ebx","%edi","%edx","%r8d"); +$code.=<<___; + mov $i($ptr),$tp1 + mov $tp1,$acc + and \$0x80808080,$acc + mov $acc,$tp4 + shr \$7,$tp4 + lea 0($tp1,$tp1),$tp2 + sub $tp4,$acc + and \$0xfefefefe,$tp2 + and \$0x1b1b1b1b,$acc + xor $tp2,$acc + mov $acc,$tp2 + + and \$0x80808080,$acc + mov $acc,$tp8 + shr \$7,$tp8 + lea 0($tp2,$tp2),$tp4 + sub $tp8,$acc + and \$0xfefefefe,$tp4 + and \$0x1b1b1b1b,$acc + xor $tp1,$tp2 # tp2^tp1 + xor $tp4,$acc + mov $acc,$tp4 + + and \$0x80808080,$acc + mov $acc,$tp8 + shr \$7,$tp8 + sub $tp8,$acc + lea 0($tp4,$tp4),$tp8 + xor $tp1,$tp4 # tp4^tp1 + and \$0xfefefefe,$tp8 + and \$0x1b1b1b1b,$acc + xor $acc,$tp8 + + xor $tp8,$tp1 # tp1^tp8 + rol \$8,$tp1 # ROTATE(tp1^tp8,8) + xor $tp8,$tp2 # tp2^tp1^tp8 + xor $tp8,$tp4 # tp4^tp1^tp8 + xor $tp2,$tp8 + xor $tp4,$tp8 # tp8^(tp8^tp4^tp1)^(tp8^tp2^tp1)=tp8^tp4^tp2 + + xor $tp8,$tp1 + rol \$24,$tp2 # ROTATE(tp2^tp1^tp8,24) + xor $tp2,$tp1 + rol \$16,$tp4 # ROTATE(tp4^tp1^tp8,16) + xor $tp4,$tp1 + + mov $tp1,$i($ptr) +___ +} + +# int AES_set_decrypt_key(const unsigned char *userKey, const int bits, +# AES_KEY *key) +$code.=<<___; +.globl AES_set_decrypt_key +.type AES_set_decrypt_key,\@function,3 +.align 16 +AES_set_decrypt_key: + push %rbx + push %rbp + push %r12 + push %r13 + push %r14 + push %r15 + push %rdx # save key schedule +.Ldec_key_prologue: + + call _x86_64_AES_set_encrypt_key + mov (%rsp),%r8 # restore key schedule + cmp \$0,%eax + jne .Labort + + mov 240(%r8),%r14d # pull number of rounds + xor %rdi,%rdi + lea (%rdi,%r14d,4),%rcx + mov %r8,%rsi + lea (%r8,%rcx,4),%rdi # pointer to last chunk +.align 4 +.Linvert: + mov 0(%rsi),%rax + mov 8(%rsi),%rbx + mov 0(%rdi),%rcx + mov 8(%rdi),%rdx + mov %rax,0(%rdi) + mov %rbx,8(%rdi) + mov %rcx,0(%rsi) + mov %rdx,8(%rsi) + lea 16(%rsi),%rsi + lea -16(%rdi),%rdi + cmp %rsi,%rdi + jne .Linvert + + lea .LAES_Te+2048+1024(%rip),%rax # rcon + + mov 40(%rax),$mask80 + mov 48(%rax),$maskfe + mov 56(%rax),$mask1b + + mov %r8,$key + sub \$1,%r14d +.align 4 +.Lpermute: + lea 16($key),$key + mov 0($key),%rax + mov 8($key),%rcx +___ + &dectransform (); +$code.=<<___; + mov %eax,0($key) + mov %ebx,4($key) + mov %ecx,8($key) + mov %edx,12($key) + sub \$1,%r14d + jnz .Lpermute + + xor %rax,%rax +.Labort: + mov 8(%rsp),%r15 + mov 16(%rsp),%r14 + mov 24(%rsp),%r13 + mov 32(%rsp),%r12 + mov 40(%rsp),%rbp + mov 48(%rsp),%rbx + add \$56,%rsp +.Ldec_key_epilogue: + ret +.size AES_set_decrypt_key,.-AES_set_decrypt_key +___ + +# void AES_cbc_encrypt (const void char *inp, unsigned char *out, +# size_t length, const AES_KEY *key, +# unsigned char *ivp,const int enc); +{ +# stack frame layout +# -8(%rsp) return address +my $keyp="0(%rsp)"; # one to pass as $key +my $keyend="8(%rsp)"; # &(keyp->rd_key[4*keyp->rounds]) +my $_rsp="16(%rsp)"; # saved %rsp +my $_inp="24(%rsp)"; # copy of 1st parameter, inp +my $_out="32(%rsp)"; # copy of 2nd parameter, out +my $_len="40(%rsp)"; # copy of 3rd parameter, length +my $_key="48(%rsp)"; # copy of 4th parameter, key +my $_ivp="56(%rsp)"; # copy of 5th parameter, ivp +my $ivec="64(%rsp)"; # ivec[16] +my $aes_key="80(%rsp)"; # copy of aes_key +my $mark="80+240(%rsp)"; # copy of aes_key->rounds + +$code.=<<___; +.globl AES_cbc_encrypt +.type AES_cbc_encrypt,\@function,6 +.align 16 +.extern OPENSSL_ia32cap_P +.globl asm_AES_cbc_encrypt +.hidden asm_AES_cbc_encrypt +asm_AES_cbc_encrypt: +AES_cbc_encrypt: + cmp \$0,%rdx # check length + je .Lcbc_epilogue + pushfq + push %rbx + push %rbp + push %r12 + push %r13 + push %r14 + push %r15 +.Lcbc_prologue: + + cld + mov %r9d,%r9d # clear upper half of enc + + lea .LAES_Te(%rip),$sbox + cmp \$0,%r9 + jne .Lcbc_picked_te + lea .LAES_Td(%rip),$sbox +.Lcbc_picked_te: + + mov OPENSSL_ia32cap_P(%rip),%r10d + cmp \$$speed_limit,%rdx + jb .Lcbc_slow_prologue + test \$15,%rdx + jnz .Lcbc_slow_prologue + bt \$28,%r10d + jc .Lcbc_slow_prologue + + # allocate aligned stack frame... + lea -88-248(%rsp),$key + and \$-64,$key + + # ... and make sure it doesn't alias with AES_T[ed] modulo 4096 + mov $sbox,%r10 + lea 2304($sbox),%r11 + mov $key,%r12 + and \$0xFFF,%r10 # s = $sbox&0xfff + and \$0xFFF,%r11 # e = ($sbox+2048)&0xfff + and \$0xFFF,%r12 # p = %rsp&0xfff + + cmp %r11,%r12 # if (p=>e) %rsp =- (p-e); + jb .Lcbc_te_break_out + sub %r11,%r12 + sub %r12,$key + jmp .Lcbc_te_ok +.Lcbc_te_break_out: # else %rsp -= (p-s)&0xfff + framesz + sub %r10,%r12 + and \$0xFFF,%r12 + add \$320,%r12 + sub %r12,$key +.align 4 +.Lcbc_te_ok: + + xchg %rsp,$key + #add \$8,%rsp # reserve for return address! + mov $key,$_rsp # save %rsp +.Lcbc_fast_body: + mov %rdi,$_inp # save copy of inp + mov %rsi,$_out # save copy of out + mov %rdx,$_len # save copy of len + mov %rcx,$_key # save copy of key + mov %r8,$_ivp # save copy of ivp + movl \$0,$mark # copy of aes_key->rounds = 0; + mov %r8,%rbp # rearrange input arguments + mov %r9,%rbx + mov %rsi,$out + mov %rdi,$inp + mov %rcx,$key + + mov 240($key),%eax # key->rounds + # do we copy key schedule to stack? + mov $key,%r10 + sub $sbox,%r10 + and \$0xfff,%r10 + cmp \$2304,%r10 + jb .Lcbc_do_ecopy + cmp \$4096-248,%r10 + jb .Lcbc_skip_ecopy +.align 4 +.Lcbc_do_ecopy: + mov $key,%rsi + lea $aes_key,%rdi + lea $aes_key,$key + mov \$240/8,%ecx + .long 0x90A548F3 # rep movsq + mov %eax,(%rdi) # copy aes_key->rounds +.Lcbc_skip_ecopy: + mov $key,$keyp # save key pointer + + mov \$18,%ecx +.align 4 +.Lcbc_prefetch_te: + mov 0($sbox),%r10 + mov 32($sbox),%r11 + mov 64($sbox),%r12 + mov 96($sbox),%r13 + lea 128($sbox),$sbox + sub \$1,%ecx + jnz .Lcbc_prefetch_te + lea -2304($sbox),$sbox + + cmp \$0,%rbx + je .LFAST_DECRYPT + +#----------------------------- ENCRYPT -----------------------------# + mov 0(%rbp),$s0 # load iv + mov 4(%rbp),$s1 + mov 8(%rbp),$s2 + mov 12(%rbp),$s3 + +.align 4 +.Lcbc_fast_enc_loop: + xor 0($inp),$s0 + xor 4($inp),$s1 + xor 8($inp),$s2 + xor 12($inp),$s3 + mov $keyp,$key # restore key + mov $inp,$_inp # if ($verticalspin) save inp + + call _x86_64_AES_encrypt + + mov $_inp,$inp # if ($verticalspin) restore inp + mov $_len,%r10 + mov $s0,0($out) + mov $s1,4($out) + mov $s2,8($out) + mov $s3,12($out) + + lea 16($inp),$inp + lea 16($out),$out + sub \$16,%r10 + test \$-16,%r10 + mov %r10,$_len + jnz .Lcbc_fast_enc_loop + mov $_ivp,%rbp # restore ivp + mov $s0,0(%rbp) # save ivec + mov $s1,4(%rbp) + mov $s2,8(%rbp) + mov $s3,12(%rbp) + + jmp .Lcbc_fast_cleanup + +#----------------------------- DECRYPT -----------------------------# +.align 16 +.LFAST_DECRYPT: + cmp $inp,$out + je .Lcbc_fast_dec_in_place + + mov %rbp,$ivec +.align 4 +.Lcbc_fast_dec_loop: + mov 0($inp),$s0 # read input + mov 4($inp),$s1 + mov 8($inp),$s2 + mov 12($inp),$s3 + mov $keyp,$key # restore key + mov $inp,$_inp # if ($verticalspin) save inp + + call _x86_64_AES_decrypt + + mov $ivec,%rbp # load ivp + mov $_inp,$inp # if ($verticalspin) restore inp + mov $_len,%r10 # load len + xor 0(%rbp),$s0 # xor iv + xor 4(%rbp),$s1 + xor 8(%rbp),$s2 + xor 12(%rbp),$s3 + mov $inp,%rbp # current input, next iv + + sub \$16,%r10 + mov %r10,$_len # update len + mov %rbp,$ivec # update ivp + + mov $s0,0($out) # write output + mov $s1,4($out) + mov $s2,8($out) + mov $s3,12($out) + + lea 16($inp),$inp + lea 16($out),$out + jnz .Lcbc_fast_dec_loop + mov $_ivp,%r12 # load user ivp + mov 0(%rbp),%r10 # load iv + mov 8(%rbp),%r11 + mov %r10,0(%r12) # copy back to user + mov %r11,8(%r12) + jmp .Lcbc_fast_cleanup + +.align 16 +.Lcbc_fast_dec_in_place: + mov 0(%rbp),%r10 # copy iv to stack + mov 8(%rbp),%r11 + mov %r10,0+$ivec + mov %r11,8+$ivec +.align 4 +.Lcbc_fast_dec_in_place_loop: + mov 0($inp),$s0 # load input + mov 4($inp),$s1 + mov 8($inp),$s2 + mov 12($inp),$s3 + mov $keyp,$key # restore key + mov $inp,$_inp # if ($verticalspin) save inp + + call _x86_64_AES_decrypt + + mov $_inp,$inp # if ($verticalspin) restore inp + mov $_len,%r10 + xor 0+$ivec,$s0 + xor 4+$ivec,$s1 + xor 8+$ivec,$s2 + xor 12+$ivec,$s3 + + mov 0($inp),%r11 # load input + mov 8($inp),%r12 + sub \$16,%r10 + jz .Lcbc_fast_dec_in_place_done + + mov %r11,0+$ivec # copy input to iv + mov %r12,8+$ivec + + mov $s0,0($out) # save output [zaps input] + mov $s1,4($out) + mov $s2,8($out) + mov $s3,12($out) + + lea 16($inp),$inp + lea 16($out),$out + mov %r10,$_len + jmp .Lcbc_fast_dec_in_place_loop +.Lcbc_fast_dec_in_place_done: + mov $_ivp,%rdi + mov %r11,0(%rdi) # copy iv back to user + mov %r12,8(%rdi) + + mov $s0,0($out) # save output [zaps input] + mov $s1,4($out) + mov $s2,8($out) + mov $s3,12($out) + +.align 4 +.Lcbc_fast_cleanup: + cmpl \$0,$mark # was the key schedule copied? + lea $aes_key,%rdi + je .Lcbc_exit + mov \$240/8,%ecx + xor %rax,%rax + .long 0x90AB48F3 # rep stosq + + jmp .Lcbc_exit + +#--------------------------- SLOW ROUTINE ---------------------------# +.align 16 +.Lcbc_slow_prologue: + # allocate aligned stack frame... + lea -88(%rsp),%rbp + and \$-64,%rbp + # ... just "above" key schedule + lea -88-63(%rcx),%r10 + sub %rbp,%r10 + neg %r10 + and \$0x3c0,%r10 + sub %r10,%rbp + + xchg %rsp,%rbp + #add \$8,%rsp # reserve for return address! + mov %rbp,$_rsp # save %rsp +.Lcbc_slow_body: + #mov %rdi,$_inp # save copy of inp + #mov %rsi,$_out # save copy of out + #mov %rdx,$_len # save copy of len + #mov %rcx,$_key # save copy of key + mov %r8,$_ivp # save copy of ivp + mov %r8,%rbp # rearrange input arguments + mov %r9,%rbx + mov %rsi,$out + mov %rdi,$inp + mov %rcx,$key + mov %rdx,%r10 + + mov 240($key),%eax + mov $key,$keyp # save key pointer + shl \$4,%eax + lea ($key,%rax),%rax + mov %rax,$keyend + + # pick Te4 copy which can't "overlap" with stack frame or key scdedule + lea 2048($sbox),$sbox + lea 768-8(%rsp),%rax + sub $sbox,%rax + and \$0x300,%rax + lea ($sbox,%rax),$sbox + + cmp \$0,%rbx + je .LSLOW_DECRYPT + +#--------------------------- SLOW ENCRYPT ---------------------------# + test \$-16,%r10 # check upon length + mov 0(%rbp),$s0 # load iv + mov 4(%rbp),$s1 + mov 8(%rbp),$s2 + mov 12(%rbp),$s3 + jz .Lcbc_slow_enc_tail # short input... + +.align 4 +.Lcbc_slow_enc_loop: + xor 0($inp),$s0 + xor 4($inp),$s1 + xor 8($inp),$s2 + xor 12($inp),$s3 + mov $keyp,$key # restore key + mov $inp,$_inp # save inp + mov $out,$_out # save out + mov %r10,$_len # save len + + call _x86_64_AES_encrypt_compact + + mov $_inp,$inp # restore inp + mov $_out,$out # restore out + mov $_len,%r10 # restore len + mov $s0,0($out) + mov $s1,4($out) + mov $s2,8($out) + mov $s3,12($out) + + lea 16($inp),$inp + lea 16($out),$out + sub \$16,%r10 + test \$-16,%r10 + jnz .Lcbc_slow_enc_loop + test \$15,%r10 + jnz .Lcbc_slow_enc_tail + mov $_ivp,%rbp # restore ivp + mov $s0,0(%rbp) # save ivec + mov $s1,4(%rbp) + mov $s2,8(%rbp) + mov $s3,12(%rbp) + + jmp .Lcbc_exit + +.align 4 +.Lcbc_slow_enc_tail: + mov %rax,%r11 + mov %rcx,%r12 + mov %r10,%rcx + mov $inp,%rsi + mov $out,%rdi + .long 0x9066A4F3 # rep movsb + mov \$16,%rcx # zero tail + sub %r10,%rcx + xor %rax,%rax + .long 0x9066AAF3 # rep stosb + mov $out,$inp # this is not a mistake! + mov \$16,%r10 # len=16 + mov %r11,%rax + mov %r12,%rcx + jmp .Lcbc_slow_enc_loop # one more spin... +#--------------------------- SLOW DECRYPT ---------------------------# +.align 16 +.LSLOW_DECRYPT: + shr \$3,%rax + add %rax,$sbox # recall "magic" constants! + + mov 0(%rbp),%r11 # copy iv to stack + mov 8(%rbp),%r12 + mov %r11,0+$ivec + mov %r12,8+$ivec + +.align 4 +.Lcbc_slow_dec_loop: + mov 0($inp),$s0 # load input + mov 4($inp),$s1 + mov 8($inp),$s2 + mov 12($inp),$s3 + mov $keyp,$key # restore key + mov $inp,$_inp # save inp + mov $out,$_out # save out + mov %r10,$_len # save len + + call _x86_64_AES_decrypt_compact + + mov $_inp,$inp # restore inp + mov $_out,$out # restore out + mov $_len,%r10 + xor 0+$ivec,$s0 + xor 4+$ivec,$s1 + xor 8+$ivec,$s2 + xor 12+$ivec,$s3 + + mov 0($inp),%r11 # load input + mov 8($inp),%r12 + sub \$16,%r10 + jc .Lcbc_slow_dec_partial + jz .Lcbc_slow_dec_done + + mov %r11,0+$ivec # copy input to iv + mov %r12,8+$ivec + + mov $s0,0($out) # save output [can zap input] + mov $s1,4($out) + mov $s2,8($out) + mov $s3,12($out) + + lea 16($inp),$inp + lea 16($out),$out + jmp .Lcbc_slow_dec_loop +.Lcbc_slow_dec_done: + mov $_ivp,%rdi + mov %r11,0(%rdi) # copy iv back to user + mov %r12,8(%rdi) + + mov $s0,0($out) # save output [can zap input] + mov $s1,4($out) + mov $s2,8($out) + mov $s3,12($out) + + jmp .Lcbc_exit + +.align 4 +.Lcbc_slow_dec_partial: + mov $_ivp,%rdi + mov %r11,0(%rdi) # copy iv back to user + mov %r12,8(%rdi) + + mov $s0,0+$ivec # save output to stack + mov $s1,4+$ivec + mov $s2,8+$ivec + mov $s3,12+$ivec + + mov $out,%rdi + lea $ivec,%rsi + lea 16(%r10),%rcx + .long 0x9066A4F3 # rep movsb + jmp .Lcbc_exit + +.align 16 +.Lcbc_exit: + mov $_rsp,%rsi + mov (%rsi),%r15 + mov 8(%rsi),%r14 + mov 16(%rsi),%r13 + mov 24(%rsi),%r12 + mov 32(%rsi),%rbp + mov 40(%rsi),%rbx + lea 48(%rsi),%rsp +.Lcbc_popfq: + popfq +.Lcbc_epilogue: + ret +.size AES_cbc_encrypt,.-AES_cbc_encrypt +___ +} + +$code.=<<___; +.align 64 +.LAES_Te: +___ + &_data_word(0xa56363c6, 0x847c7cf8, 0x997777ee, 0x8d7b7bf6); + &_data_word(0x0df2f2ff, 0xbd6b6bd6, 0xb16f6fde, 0x54c5c591); + &_data_word(0x50303060, 0x03010102, 0xa96767ce, 0x7d2b2b56); + &_data_word(0x19fefee7, 0x62d7d7b5, 0xe6abab4d, 0x9a7676ec); + &_data_word(0x45caca8f, 0x9d82821f, 0x40c9c989, 0x877d7dfa); + &_data_word(0x15fafaef, 0xeb5959b2, 0xc947478e, 0x0bf0f0fb); + &_data_word(0xecadad41, 0x67d4d4b3, 0xfda2a25f, 0xeaafaf45); + &_data_word(0xbf9c9c23, 0xf7a4a453, 0x967272e4, 0x5bc0c09b); + &_data_word(0xc2b7b775, 0x1cfdfde1, 0xae93933d, 0x6a26264c); + &_data_word(0x5a36366c, 0x413f3f7e, 0x02f7f7f5, 0x4fcccc83); + &_data_word(0x5c343468, 0xf4a5a551, 0x34e5e5d1, 0x08f1f1f9); + &_data_word(0x937171e2, 0x73d8d8ab, 0x53313162, 0x3f15152a); + &_data_word(0x0c040408, 0x52c7c795, 0x65232346, 0x5ec3c39d); + &_data_word(0x28181830, 0xa1969637, 0x0f05050a, 0xb59a9a2f); + &_data_word(0x0907070e, 0x36121224, 0x9b80801b, 0x3de2e2df); + &_data_word(0x26ebebcd, 0x6927274e, 0xcdb2b27f, 0x9f7575ea); + &_data_word(0x1b090912, 0x9e83831d, 0x742c2c58, 0x2e1a1a34); + &_data_word(0x2d1b1b36, 0xb26e6edc, 0xee5a5ab4, 0xfba0a05b); + &_data_word(0xf65252a4, 0x4d3b3b76, 0x61d6d6b7, 0xceb3b37d); + &_data_word(0x7b292952, 0x3ee3e3dd, 0x712f2f5e, 0x97848413); + &_data_word(0xf55353a6, 0x68d1d1b9, 0x00000000, 0x2cededc1); + &_data_word(0x60202040, 0x1ffcfce3, 0xc8b1b179, 0xed5b5bb6); + &_data_word(0xbe6a6ad4, 0x46cbcb8d, 0xd9bebe67, 0x4b393972); + &_data_word(0xde4a4a94, 0xd44c4c98, 0xe85858b0, 0x4acfcf85); + &_data_word(0x6bd0d0bb, 0x2aefefc5, 0xe5aaaa4f, 0x16fbfbed); + &_data_word(0xc5434386, 0xd74d4d9a, 0x55333366, 0x94858511); + &_data_word(0xcf45458a, 0x10f9f9e9, 0x06020204, 0x817f7ffe); + &_data_word(0xf05050a0, 0x443c3c78, 0xba9f9f25, 0xe3a8a84b); + &_data_word(0xf35151a2, 0xfea3a35d, 0xc0404080, 0x8a8f8f05); + &_data_word(0xad92923f, 0xbc9d9d21, 0x48383870, 0x04f5f5f1); + &_data_word(0xdfbcbc63, 0xc1b6b677, 0x75dadaaf, 0x63212142); + &_data_word(0x30101020, 0x1affffe5, 0x0ef3f3fd, 0x6dd2d2bf); + &_data_word(0x4ccdcd81, 0x140c0c18, 0x35131326, 0x2fececc3); + &_data_word(0xe15f5fbe, 0xa2979735, 0xcc444488, 0x3917172e); + &_data_word(0x57c4c493, 0xf2a7a755, 0x827e7efc, 0x473d3d7a); + &_data_word(0xac6464c8, 0xe75d5dba, 0x2b191932, 0x957373e6); + &_data_word(0xa06060c0, 0x98818119, 0xd14f4f9e, 0x7fdcdca3); + &_data_word(0x66222244, 0x7e2a2a54, 0xab90903b, 0x8388880b); + &_data_word(0xca46468c, 0x29eeeec7, 0xd3b8b86b, 0x3c141428); + &_data_word(0x79dedea7, 0xe25e5ebc, 0x1d0b0b16, 0x76dbdbad); + &_data_word(0x3be0e0db, 0x56323264, 0x4e3a3a74, 0x1e0a0a14); + &_data_word(0xdb494992, 0x0a06060c, 0x6c242448, 0xe45c5cb8); + &_data_word(0x5dc2c29f, 0x6ed3d3bd, 0xefacac43, 0xa66262c4); + &_data_word(0xa8919139, 0xa4959531, 0x37e4e4d3, 0x8b7979f2); + &_data_word(0x32e7e7d5, 0x43c8c88b, 0x5937376e, 0xb76d6dda); + &_data_word(0x8c8d8d01, 0x64d5d5b1, 0xd24e4e9c, 0xe0a9a949); + &_data_word(0xb46c6cd8, 0xfa5656ac, 0x07f4f4f3, 0x25eaeacf); + &_data_word(0xaf6565ca, 0x8e7a7af4, 0xe9aeae47, 0x18080810); + &_data_word(0xd5baba6f, 0x887878f0, 0x6f25254a, 0x722e2e5c); + &_data_word(0x241c1c38, 0xf1a6a657, 0xc7b4b473, 0x51c6c697); + &_data_word(0x23e8e8cb, 0x7cdddda1, 0x9c7474e8, 0x211f1f3e); + &_data_word(0xdd4b4b96, 0xdcbdbd61, 0x868b8b0d, 0x858a8a0f); + &_data_word(0x907070e0, 0x423e3e7c, 0xc4b5b571, 0xaa6666cc); + &_data_word(0xd8484890, 0x05030306, 0x01f6f6f7, 0x120e0e1c); + &_data_word(0xa36161c2, 0x5f35356a, 0xf95757ae, 0xd0b9b969); + &_data_word(0x91868617, 0x58c1c199, 0x271d1d3a, 0xb99e9e27); + &_data_word(0x38e1e1d9, 0x13f8f8eb, 0xb398982b, 0x33111122); + &_data_word(0xbb6969d2, 0x70d9d9a9, 0x898e8e07, 0xa7949433); + &_data_word(0xb69b9b2d, 0x221e1e3c, 0x92878715, 0x20e9e9c9); + &_data_word(0x49cece87, 0xff5555aa, 0x78282850, 0x7adfdfa5); + &_data_word(0x8f8c8c03, 0xf8a1a159, 0x80898909, 0x170d0d1a); + &_data_word(0xdabfbf65, 0x31e6e6d7, 0xc6424284, 0xb86868d0); + &_data_word(0xc3414182, 0xb0999929, 0x772d2d5a, 0x110f0f1e); + &_data_word(0xcbb0b07b, 0xfc5454a8, 0xd6bbbb6d, 0x3a16162c); + +#Te4 # four copies of Te4 to choose from to avoid L1 aliasing + &data_byte(0x63, 0x7c, 0x77, 0x7b, 0xf2, 0x6b, 0x6f, 0xc5); + &data_byte(0x30, 0x01, 0x67, 0x2b, 0xfe, 0xd7, 0xab, 0x76); + &data_byte(0xca, 0x82, 0xc9, 0x7d, 0xfa, 0x59, 0x47, 0xf0); + &data_byte(0xad, 0xd4, 0xa2, 0xaf, 0x9c, 0xa4, 0x72, 0xc0); + &data_byte(0xb7, 0xfd, 0x93, 0x26, 0x36, 0x3f, 0xf7, 0xcc); + &data_byte(0x34, 0xa5, 0xe5, 0xf1, 0x71, 0xd8, 0x31, 0x15); + &data_byte(0x04, 0xc7, 0x23, 0xc3, 0x18, 0x96, 0x05, 0x9a); + &data_byte(0x07, 0x12, 0x80, 0xe2, 0xeb, 0x27, 0xb2, 0x75); + &data_byte(0x09, 0x83, 0x2c, 0x1a, 0x1b, 0x6e, 0x5a, 0xa0); + &data_byte(0x52, 0x3b, 0xd6, 0xb3, 0x29, 0xe3, 0x2f, 0x84); + &data_byte(0x53, 0xd1, 0x00, 0xed, 0x20, 0xfc, 0xb1, 0x5b); + &data_byte(0x6a, 0xcb, 0xbe, 0x39, 0x4a, 0x4c, 0x58, 0xcf); + &data_byte(0xd0, 0xef, 0xaa, 0xfb, 0x43, 0x4d, 0x33, 0x85); + &data_byte(0x45, 0xf9, 0x02, 0x7f, 0x50, 0x3c, 0x9f, 0xa8); + &data_byte(0x51, 0xa3, 0x40, 0x8f, 0x92, 0x9d, 0x38, 0xf5); + &data_byte(0xbc, 0xb6, 0xda, 0x21, 0x10, 0xff, 0xf3, 0xd2); + &data_byte(0xcd, 0x0c, 0x13, 0xec, 0x5f, 0x97, 0x44, 0x17); + &data_byte(0xc4, 0xa7, 0x7e, 0x3d, 0x64, 0x5d, 0x19, 0x73); + &data_byte(0x60, 0x81, 0x4f, 0xdc, 0x22, 0x2a, 0x90, 0x88); + &data_byte(0x46, 0xee, 0xb8, 0x14, 0xde, 0x5e, 0x0b, 0xdb); + &data_byte(0xe0, 0x32, 0x3a, 0x0a, 0x49, 0x06, 0x24, 0x5c); + &data_byte(0xc2, 0xd3, 0xac, 0x62, 0x91, 0x95, 0xe4, 0x79); + &data_byte(0xe7, 0xc8, 0x37, 0x6d, 0x8d, 0xd5, 0x4e, 0xa9); + &data_byte(0x6c, 0x56, 0xf4, 0xea, 0x65, 0x7a, 0xae, 0x08); + &data_byte(0xba, 0x78, 0x25, 0x2e, 0x1c, 0xa6, 0xb4, 0xc6); + &data_byte(0xe8, 0xdd, 0x74, 0x1f, 0x4b, 0xbd, 0x8b, 0x8a); + &data_byte(0x70, 0x3e, 0xb5, 0x66, 0x48, 0x03, 0xf6, 0x0e); + &data_byte(0x61, 0x35, 0x57, 0xb9, 0x86, 0xc1, 0x1d, 0x9e); + &data_byte(0xe1, 0xf8, 0x98, 0x11, 0x69, 0xd9, 0x8e, 0x94); + &data_byte(0x9b, 0x1e, 0x87, 0xe9, 0xce, 0x55, 0x28, 0xdf); + &data_byte(0x8c, 0xa1, 0x89, 0x0d, 0xbf, 0xe6, 0x42, 0x68); + &data_byte(0x41, 0x99, 0x2d, 0x0f, 0xb0, 0x54, 0xbb, 0x16); + + &data_byte(0x63, 0x7c, 0x77, 0x7b, 0xf2, 0x6b, 0x6f, 0xc5); + &data_byte(0x30, 0x01, 0x67, 0x2b, 0xfe, 0xd7, 0xab, 0x76); + &data_byte(0xca, 0x82, 0xc9, 0x7d, 0xfa, 0x59, 0x47, 0xf0); + &data_byte(0xad, 0xd4, 0xa2, 0xaf, 0x9c, 0xa4, 0x72, 0xc0); + &data_byte(0xb7, 0xfd, 0x93, 0x26, 0x36, 0x3f, 0xf7, 0xcc); + &data_byte(0x34, 0xa5, 0xe5, 0xf1, 0x71, 0xd8, 0x31, 0x15); + &data_byte(0x04, 0xc7, 0x23, 0xc3, 0x18, 0x96, 0x05, 0x9a); + &data_byte(0x07, 0x12, 0x80, 0xe2, 0xeb, 0x27, 0xb2, 0x75); + &data_byte(0x09, 0x83, 0x2c, 0x1a, 0x1b, 0x6e, 0x5a, 0xa0); + &data_byte(0x52, 0x3b, 0xd6, 0xb3, 0x29, 0xe3, 0x2f, 0x84); + &data_byte(0x53, 0xd1, 0x00, 0xed, 0x20, 0xfc, 0xb1, 0x5b); + &data_byte(0x6a, 0xcb, 0xbe, 0x39, 0x4a, 0x4c, 0x58, 0xcf); + &data_byte(0xd0, 0xef, 0xaa, 0xfb, 0x43, 0x4d, 0x33, 0x85); + &data_byte(0x45, 0xf9, 0x02, 0x7f, 0x50, 0x3c, 0x9f, 0xa8); + &data_byte(0x51, 0xa3, 0x40, 0x8f, 0x92, 0x9d, 0x38, 0xf5); + &data_byte(0xbc, 0xb6, 0xda, 0x21, 0x10, 0xff, 0xf3, 0xd2); + &data_byte(0xcd, 0x0c, 0x13, 0xec, 0x5f, 0x97, 0x44, 0x17); + &data_byte(0xc4, 0xa7, 0x7e, 0x3d, 0x64, 0x5d, 0x19, 0x73); + &data_byte(0x60, 0x81, 0x4f, 0xdc, 0x22, 0x2a, 0x90, 0x88); + &data_byte(0x46, 0xee, 0xb8, 0x14, 0xde, 0x5e, 0x0b, 0xdb); + &data_byte(0xe0, 0x32, 0x3a, 0x0a, 0x49, 0x06, 0x24, 0x5c); + &data_byte(0xc2, 0xd3, 0xac, 0x62, 0x91, 0x95, 0xe4, 0x79); + &data_byte(0xe7, 0xc8, 0x37, 0x6d, 0x8d, 0xd5, 0x4e, 0xa9); + &data_byte(0x6c, 0x56, 0xf4, 0xea, 0x65, 0x7a, 0xae, 0x08); + &data_byte(0xba, 0x78, 0x25, 0x2e, 0x1c, 0xa6, 0xb4, 0xc6); + &data_byte(0xe8, 0xdd, 0x74, 0x1f, 0x4b, 0xbd, 0x8b, 0x8a); + &data_byte(0x70, 0x3e, 0xb5, 0x66, 0x48, 0x03, 0xf6, 0x0e); + &data_byte(0x61, 0x35, 0x57, 0xb9, 0x86, 0xc1, 0x1d, 0x9e); + &data_byte(0xe1, 0xf8, 0x98, 0x11, 0x69, 0xd9, 0x8e, 0x94); + &data_byte(0x9b, 0x1e, 0x87, 0xe9, 0xce, 0x55, 0x28, 0xdf); + &data_byte(0x8c, 0xa1, 0x89, 0x0d, 0xbf, 0xe6, 0x42, 0x68); + &data_byte(0x41, 0x99, 0x2d, 0x0f, 0xb0, 0x54, 0xbb, 0x16); + + &data_byte(0x63, 0x7c, 0x77, 0x7b, 0xf2, 0x6b, 0x6f, 0xc5); + &data_byte(0x30, 0x01, 0x67, 0x2b, 0xfe, 0xd7, 0xab, 0x76); + &data_byte(0xca, 0x82, 0xc9, 0x7d, 0xfa, 0x59, 0x47, 0xf0); + &data_byte(0xad, 0xd4, 0xa2, 0xaf, 0x9c, 0xa4, 0x72, 0xc0); + &data_byte(0xb7, 0xfd, 0x93, 0x26, 0x36, 0x3f, 0xf7, 0xcc); + &data_byte(0x34, 0xa5, 0xe5, 0xf1, 0x71, 0xd8, 0x31, 0x15); + &data_byte(0x04, 0xc7, 0x23, 0xc3, 0x18, 0x96, 0x05, 0x9a); + &data_byte(0x07, 0x12, 0x80, 0xe2, 0xeb, 0x27, 0xb2, 0x75); + &data_byte(0x09, 0x83, 0x2c, 0x1a, 0x1b, 0x6e, 0x5a, 0xa0); + &data_byte(0x52, 0x3b, 0xd6, 0xb3, 0x29, 0xe3, 0x2f, 0x84); + &data_byte(0x53, 0xd1, 0x00, 0xed, 0x20, 0xfc, 0xb1, 0x5b); + &data_byte(0x6a, 0xcb, 0xbe, 0x39, 0x4a, 0x4c, 0x58, 0xcf); + &data_byte(0xd0, 0xef, 0xaa, 0xfb, 0x43, 0x4d, 0x33, 0x85); + &data_byte(0x45, 0xf9, 0x02, 0x7f, 0x50, 0x3c, 0x9f, 0xa8); + &data_byte(0x51, 0xa3, 0x40, 0x8f, 0x92, 0x9d, 0x38, 0xf5); + &data_byte(0xbc, 0xb6, 0xda, 0x21, 0x10, 0xff, 0xf3, 0xd2); + &data_byte(0xcd, 0x0c, 0x13, 0xec, 0x5f, 0x97, 0x44, 0x17); + &data_byte(0xc4, 0xa7, 0x7e, 0x3d, 0x64, 0x5d, 0x19, 0x73); + &data_byte(0x60, 0x81, 0x4f, 0xdc, 0x22, 0x2a, 0x90, 0x88); + &data_byte(0x46, 0xee, 0xb8, 0x14, 0xde, 0x5e, 0x0b, 0xdb); + &data_byte(0xe0, 0x32, 0x3a, 0x0a, 0x49, 0x06, 0x24, 0x5c); + &data_byte(0xc2, 0xd3, 0xac, 0x62, 0x91, 0x95, 0xe4, 0x79); + &data_byte(0xe7, 0xc8, 0x37, 0x6d, 0x8d, 0xd5, 0x4e, 0xa9); + &data_byte(0x6c, 0x56, 0xf4, 0xea, 0x65, 0x7a, 0xae, 0x08); + &data_byte(0xba, 0x78, 0x25, 0x2e, 0x1c, 0xa6, 0xb4, 0xc6); + &data_byte(0xe8, 0xdd, 0x74, 0x1f, 0x4b, 0xbd, 0x8b, 0x8a); + &data_byte(0x70, 0x3e, 0xb5, 0x66, 0x48, 0x03, 0xf6, 0x0e); + &data_byte(0x61, 0x35, 0x57, 0xb9, 0x86, 0xc1, 0x1d, 0x9e); + &data_byte(0xe1, 0xf8, 0x98, 0x11, 0x69, 0xd9, 0x8e, 0x94); + &data_byte(0x9b, 0x1e, 0x87, 0xe9, 0xce, 0x55, 0x28, 0xdf); + &data_byte(0x8c, 0xa1, 0x89, 0x0d, 0xbf, 0xe6, 0x42, 0x68); + &data_byte(0x41, 0x99, 0x2d, 0x0f, 0xb0, 0x54, 0xbb, 0x16); + + &data_byte(0x63, 0x7c, 0x77, 0x7b, 0xf2, 0x6b, 0x6f, 0xc5); + &data_byte(0x30, 0x01, 0x67, 0x2b, 0xfe, 0xd7, 0xab, 0x76); + &data_byte(0xca, 0x82, 0xc9, 0x7d, 0xfa, 0x59, 0x47, 0xf0); + &data_byte(0xad, 0xd4, 0xa2, 0xaf, 0x9c, 0xa4, 0x72, 0xc0); + &data_byte(0xb7, 0xfd, 0x93, 0x26, 0x36, 0x3f, 0xf7, 0xcc); + &data_byte(0x34, 0xa5, 0xe5, 0xf1, 0x71, 0xd8, 0x31, 0x15); + &data_byte(0x04, 0xc7, 0x23, 0xc3, 0x18, 0x96, 0x05, 0x9a); + &data_byte(0x07, 0x12, 0x80, 0xe2, 0xeb, 0x27, 0xb2, 0x75); + &data_byte(0x09, 0x83, 0x2c, 0x1a, 0x1b, 0x6e, 0x5a, 0xa0); + &data_byte(0x52, 0x3b, 0xd6, 0xb3, 0x29, 0xe3, 0x2f, 0x84); + &data_byte(0x53, 0xd1, 0x00, 0xed, 0x20, 0xfc, 0xb1, 0x5b); + &data_byte(0x6a, 0xcb, 0xbe, 0x39, 0x4a, 0x4c, 0x58, 0xcf); + &data_byte(0xd0, 0xef, 0xaa, 0xfb, 0x43, 0x4d, 0x33, 0x85); + &data_byte(0x45, 0xf9, 0x02, 0x7f, 0x50, 0x3c, 0x9f, 0xa8); + &data_byte(0x51, 0xa3, 0x40, 0x8f, 0x92, 0x9d, 0x38, 0xf5); + &data_byte(0xbc, 0xb6, 0xda, 0x21, 0x10, 0xff, 0xf3, 0xd2); + &data_byte(0xcd, 0x0c, 0x13, 0xec, 0x5f, 0x97, 0x44, 0x17); + &data_byte(0xc4, 0xa7, 0x7e, 0x3d, 0x64, 0x5d, 0x19, 0x73); + &data_byte(0x60, 0x81, 0x4f, 0xdc, 0x22, 0x2a, 0x90, 0x88); + &data_byte(0x46, 0xee, 0xb8, 0x14, 0xde, 0x5e, 0x0b, 0xdb); + &data_byte(0xe0, 0x32, 0x3a, 0x0a, 0x49, 0x06, 0x24, 0x5c); + &data_byte(0xc2, 0xd3, 0xac, 0x62, 0x91, 0x95, 0xe4, 0x79); + &data_byte(0xe7, 0xc8, 0x37, 0x6d, 0x8d, 0xd5, 0x4e, 0xa9); + &data_byte(0x6c, 0x56, 0xf4, 0xea, 0x65, 0x7a, 0xae, 0x08); + &data_byte(0xba, 0x78, 0x25, 0x2e, 0x1c, 0xa6, 0xb4, 0xc6); + &data_byte(0xe8, 0xdd, 0x74, 0x1f, 0x4b, 0xbd, 0x8b, 0x8a); + &data_byte(0x70, 0x3e, 0xb5, 0x66, 0x48, 0x03, 0xf6, 0x0e); + &data_byte(0x61, 0x35, 0x57, 0xb9, 0x86, 0xc1, 0x1d, 0x9e); + &data_byte(0xe1, 0xf8, 0x98, 0x11, 0x69, 0xd9, 0x8e, 0x94); + &data_byte(0x9b, 0x1e, 0x87, 0xe9, 0xce, 0x55, 0x28, 0xdf); + &data_byte(0x8c, 0xa1, 0x89, 0x0d, 0xbf, 0xe6, 0x42, 0x68); + &data_byte(0x41, 0x99, 0x2d, 0x0f, 0xb0, 0x54, 0xbb, 0x16); +#rcon: +$code.=<<___; + .long 0x00000001, 0x00000002, 0x00000004, 0x00000008 + .long 0x00000010, 0x00000020, 0x00000040, 0x00000080 + .long 0x0000001b, 0x00000036, 0x80808080, 0x80808080 + .long 0xfefefefe, 0xfefefefe, 0x1b1b1b1b, 0x1b1b1b1b +___ +$code.=<<___; +.align 64 +.LAES_Td: +___ + &_data_word(0x50a7f451, 0x5365417e, 0xc3a4171a, 0x965e273a); + &_data_word(0xcb6bab3b, 0xf1459d1f, 0xab58faac, 0x9303e34b); + &_data_word(0x55fa3020, 0xf66d76ad, 0x9176cc88, 0x254c02f5); + &_data_word(0xfcd7e54f, 0xd7cb2ac5, 0x80443526, 0x8fa362b5); + &_data_word(0x495ab1de, 0x671bba25, 0x980eea45, 0xe1c0fe5d); + &_data_word(0x02752fc3, 0x12f04c81, 0xa397468d, 0xc6f9d36b); + &_data_word(0xe75f8f03, 0x959c9215, 0xeb7a6dbf, 0xda595295); + &_data_word(0x2d83bed4, 0xd3217458, 0x2969e049, 0x44c8c98e); + &_data_word(0x6a89c275, 0x78798ef4, 0x6b3e5899, 0xdd71b927); + &_data_word(0xb64fe1be, 0x17ad88f0, 0x66ac20c9, 0xb43ace7d); + &_data_word(0x184adf63, 0x82311ae5, 0x60335197, 0x457f5362); + &_data_word(0xe07764b1, 0x84ae6bbb, 0x1ca081fe, 0x942b08f9); + &_data_word(0x58684870, 0x19fd458f, 0x876cde94, 0xb7f87b52); + &_data_word(0x23d373ab, 0xe2024b72, 0x578f1fe3, 0x2aab5566); + &_data_word(0x0728ebb2, 0x03c2b52f, 0x9a7bc586, 0xa50837d3); + &_data_word(0xf2872830, 0xb2a5bf23, 0xba6a0302, 0x5c8216ed); + &_data_word(0x2b1ccf8a, 0x92b479a7, 0xf0f207f3, 0xa1e2694e); + &_data_word(0xcdf4da65, 0xd5be0506, 0x1f6234d1, 0x8afea6c4); + &_data_word(0x9d532e34, 0xa055f3a2, 0x32e18a05, 0x75ebf6a4); + &_data_word(0x39ec830b, 0xaaef6040, 0x069f715e, 0x51106ebd); + &_data_word(0xf98a213e, 0x3d06dd96, 0xae053edd, 0x46bde64d); + &_data_word(0xb58d5491, 0x055dc471, 0x6fd40604, 0xff155060); + &_data_word(0x24fb9819, 0x97e9bdd6, 0xcc434089, 0x779ed967); + &_data_word(0xbd42e8b0, 0x888b8907, 0x385b19e7, 0xdbeec879); + &_data_word(0x470a7ca1, 0xe90f427c, 0xc91e84f8, 0x00000000); + &_data_word(0x83868009, 0x48ed2b32, 0xac70111e, 0x4e725a6c); + &_data_word(0xfbff0efd, 0x5638850f, 0x1ed5ae3d, 0x27392d36); + &_data_word(0x64d90f0a, 0x21a65c68, 0xd1545b9b, 0x3a2e3624); + &_data_word(0xb1670a0c, 0x0fe75793, 0xd296eeb4, 0x9e919b1b); + &_data_word(0x4fc5c080, 0xa220dc61, 0x694b775a, 0x161a121c); + &_data_word(0x0aba93e2, 0xe52aa0c0, 0x43e0223c, 0x1d171b12); + &_data_word(0x0b0d090e, 0xadc78bf2, 0xb9a8b62d, 0xc8a91e14); + &_data_word(0x8519f157, 0x4c0775af, 0xbbdd99ee, 0xfd607fa3); + &_data_word(0x9f2601f7, 0xbcf5725c, 0xc53b6644, 0x347efb5b); + &_data_word(0x7629438b, 0xdcc623cb, 0x68fcedb6, 0x63f1e4b8); + &_data_word(0xcadc31d7, 0x10856342, 0x40229713, 0x2011c684); + &_data_word(0x7d244a85, 0xf83dbbd2, 0x1132f9ae, 0x6da129c7); + &_data_word(0x4b2f9e1d, 0xf330b2dc, 0xec52860d, 0xd0e3c177); + &_data_word(0x6c16b32b, 0x99b970a9, 0xfa489411, 0x2264e947); + &_data_word(0xc48cfca8, 0x1a3ff0a0, 0xd82c7d56, 0xef903322); + &_data_word(0xc74e4987, 0xc1d138d9, 0xfea2ca8c, 0x360bd498); + &_data_word(0xcf81f5a6, 0x28de7aa5, 0x268eb7da, 0xa4bfad3f); + &_data_word(0xe49d3a2c, 0x0d927850, 0x9bcc5f6a, 0x62467e54); + &_data_word(0xc2138df6, 0xe8b8d890, 0x5ef7392e, 0xf5afc382); + &_data_word(0xbe805d9f, 0x7c93d069, 0xa92dd56f, 0xb31225cf); + &_data_word(0x3b99acc8, 0xa77d1810, 0x6e639ce8, 0x7bbb3bdb); + &_data_word(0x097826cd, 0xf418596e, 0x01b79aec, 0xa89a4f83); + &_data_word(0x656e95e6, 0x7ee6ffaa, 0x08cfbc21, 0xe6e815ef); + &_data_word(0xd99be7ba, 0xce366f4a, 0xd4099fea, 0xd67cb029); + &_data_word(0xafb2a431, 0x31233f2a, 0x3094a5c6, 0xc066a235); + &_data_word(0x37bc4e74, 0xa6ca82fc, 0xb0d090e0, 0x15d8a733); + &_data_word(0x4a9804f1, 0xf7daec41, 0x0e50cd7f, 0x2ff69117); + &_data_word(0x8dd64d76, 0x4db0ef43, 0x544daacc, 0xdf0496e4); + &_data_word(0xe3b5d19e, 0x1b886a4c, 0xb81f2cc1, 0x7f516546); + &_data_word(0x04ea5e9d, 0x5d358c01, 0x737487fa, 0x2e410bfb); + &_data_word(0x5a1d67b3, 0x52d2db92, 0x335610e9, 0x1347d66d); + &_data_word(0x8c61d79a, 0x7a0ca137, 0x8e14f859, 0x893c13eb); + &_data_word(0xee27a9ce, 0x35c961b7, 0xede51ce1, 0x3cb1477a); + &_data_word(0x59dfd29c, 0x3f73f255, 0x79ce1418, 0xbf37c773); + &_data_word(0xeacdf753, 0x5baafd5f, 0x146f3ddf, 0x86db4478); + &_data_word(0x81f3afca, 0x3ec468b9, 0x2c342438, 0x5f40a3c2); + &_data_word(0x72c31d16, 0x0c25e2bc, 0x8b493c28, 0x41950dff); + &_data_word(0x7101a839, 0xdeb30c08, 0x9ce4b4d8, 0x90c15664); + &_data_word(0x6184cb7b, 0x70b632d5, 0x745c6c48, 0x4257b8d0); + +#Td4: # four copies of Td4 to choose from to avoid L1 aliasing + &data_byte(0x52, 0x09, 0x6a, 0xd5, 0x30, 0x36, 0xa5, 0x38); + &data_byte(0xbf, 0x40, 0xa3, 0x9e, 0x81, 0xf3, 0xd7, 0xfb); + &data_byte(0x7c, 0xe3, 0x39, 0x82, 0x9b, 0x2f, 0xff, 0x87); + &data_byte(0x34, 0x8e, 0x43, 0x44, 0xc4, 0xde, 0xe9, 0xcb); + &data_byte(0x54, 0x7b, 0x94, 0x32, 0xa6, 0xc2, 0x23, 0x3d); + &data_byte(0xee, 0x4c, 0x95, 0x0b, 0x42, 0xfa, 0xc3, 0x4e); + &data_byte(0x08, 0x2e, 0xa1, 0x66, 0x28, 0xd9, 0x24, 0xb2); + &data_byte(0x76, 0x5b, 0xa2, 0x49, 0x6d, 0x8b, 0xd1, 0x25); + &data_byte(0x72, 0xf8, 0xf6, 0x64, 0x86, 0x68, 0x98, 0x16); + &data_byte(0xd4, 0xa4, 0x5c, 0xcc, 0x5d, 0x65, 0xb6, 0x92); + &data_byte(0x6c, 0x70, 0x48, 0x50, 0xfd, 0xed, 0xb9, 0xda); + &data_byte(0x5e, 0x15, 0x46, 0x57, 0xa7, 0x8d, 0x9d, 0x84); + &data_byte(0x90, 0xd8, 0xab, 0x00, 0x8c, 0xbc, 0xd3, 0x0a); + &data_byte(0xf7, 0xe4, 0x58, 0x05, 0xb8, 0xb3, 0x45, 0x06); + &data_byte(0xd0, 0x2c, 0x1e, 0x8f, 0xca, 0x3f, 0x0f, 0x02); + &data_byte(0xc1, 0xaf, 0xbd, 0x03, 0x01, 0x13, 0x8a, 0x6b); + &data_byte(0x3a, 0x91, 0x11, 0x41, 0x4f, 0x67, 0xdc, 0xea); + &data_byte(0x97, 0xf2, 0xcf, 0xce, 0xf0, 0xb4, 0xe6, 0x73); + &data_byte(0x96, 0xac, 0x74, 0x22, 0xe7, 0xad, 0x35, 0x85); + &data_byte(0xe2, 0xf9, 0x37, 0xe8, 0x1c, 0x75, 0xdf, 0x6e); + &data_byte(0x47, 0xf1, 0x1a, 0x71, 0x1d, 0x29, 0xc5, 0x89); + &data_byte(0x6f, 0xb7, 0x62, 0x0e, 0xaa, 0x18, 0xbe, 0x1b); + &data_byte(0xfc, 0x56, 0x3e, 0x4b, 0xc6, 0xd2, 0x79, 0x20); + &data_byte(0x9a, 0xdb, 0xc0, 0xfe, 0x78, 0xcd, 0x5a, 0xf4); + &data_byte(0x1f, 0xdd, 0xa8, 0x33, 0x88, 0x07, 0xc7, 0x31); + &data_byte(0xb1, 0x12, 0x10, 0x59, 0x27, 0x80, 0xec, 0x5f); + &data_byte(0x60, 0x51, 0x7f, 0xa9, 0x19, 0xb5, 0x4a, 0x0d); + &data_byte(0x2d, 0xe5, 0x7a, 0x9f, 0x93, 0xc9, 0x9c, 0xef); + &data_byte(0xa0, 0xe0, 0x3b, 0x4d, 0xae, 0x2a, 0xf5, 0xb0); + &data_byte(0xc8, 0xeb, 0xbb, 0x3c, 0x83, 0x53, 0x99, 0x61); + &data_byte(0x17, 0x2b, 0x04, 0x7e, 0xba, 0x77, 0xd6, 0x26); + &data_byte(0xe1, 0x69, 0x14, 0x63, 0x55, 0x21, 0x0c, 0x7d); +$code.=<<___; + .long 0x80808080, 0x80808080, 0xfefefefe, 0xfefefefe + .long 0x1b1b1b1b, 0x1b1b1b1b, 0, 0 +___ + &data_byte(0x52, 0x09, 0x6a, 0xd5, 0x30, 0x36, 0xa5, 0x38); + &data_byte(0xbf, 0x40, 0xa3, 0x9e, 0x81, 0xf3, 0xd7, 0xfb); + &data_byte(0x7c, 0xe3, 0x39, 0x82, 0x9b, 0x2f, 0xff, 0x87); + &data_byte(0x34, 0x8e, 0x43, 0x44, 0xc4, 0xde, 0xe9, 0xcb); + &data_byte(0x54, 0x7b, 0x94, 0x32, 0xa6, 0xc2, 0x23, 0x3d); + &data_byte(0xee, 0x4c, 0x95, 0x0b, 0x42, 0xfa, 0xc3, 0x4e); + &data_byte(0x08, 0x2e, 0xa1, 0x66, 0x28, 0xd9, 0x24, 0xb2); + &data_byte(0x76, 0x5b, 0xa2, 0x49, 0x6d, 0x8b, 0xd1, 0x25); + &data_byte(0x72, 0xf8, 0xf6, 0x64, 0x86, 0x68, 0x98, 0x16); + &data_byte(0xd4, 0xa4, 0x5c, 0xcc, 0x5d, 0x65, 0xb6, 0x92); + &data_byte(0x6c, 0x70, 0x48, 0x50, 0xfd, 0xed, 0xb9, 0xda); + &data_byte(0x5e, 0x15, 0x46, 0x57, 0xa7, 0x8d, 0x9d, 0x84); + &data_byte(0x90, 0xd8, 0xab, 0x00, 0x8c, 0xbc, 0xd3, 0x0a); + &data_byte(0xf7, 0xe4, 0x58, 0x05, 0xb8, 0xb3, 0x45, 0x06); + &data_byte(0xd0, 0x2c, 0x1e, 0x8f, 0xca, 0x3f, 0x0f, 0x02); + &data_byte(0xc1, 0xaf, 0xbd, 0x03, 0x01, 0x13, 0x8a, 0x6b); + &data_byte(0x3a, 0x91, 0x11, 0x41, 0x4f, 0x67, 0xdc, 0xea); + &data_byte(0x97, 0xf2, 0xcf, 0xce, 0xf0, 0xb4, 0xe6, 0x73); + &data_byte(0x96, 0xac, 0x74, 0x22, 0xe7, 0xad, 0x35, 0x85); + &data_byte(0xe2, 0xf9, 0x37, 0xe8, 0x1c, 0x75, 0xdf, 0x6e); + &data_byte(0x47, 0xf1, 0x1a, 0x71, 0x1d, 0x29, 0xc5, 0x89); + &data_byte(0x6f, 0xb7, 0x62, 0x0e, 0xaa, 0x18, 0xbe, 0x1b); + &data_byte(0xfc, 0x56, 0x3e, 0x4b, 0xc6, 0xd2, 0x79, 0x20); + &data_byte(0x9a, 0xdb, 0xc0, 0xfe, 0x78, 0xcd, 0x5a, 0xf4); + &data_byte(0x1f, 0xdd, 0xa8, 0x33, 0x88, 0x07, 0xc7, 0x31); + &data_byte(0xb1, 0x12, 0x10, 0x59, 0x27, 0x80, 0xec, 0x5f); + &data_byte(0x60, 0x51, 0x7f, 0xa9, 0x19, 0xb5, 0x4a, 0x0d); + &data_byte(0x2d, 0xe5, 0x7a, 0x9f, 0x93, 0xc9, 0x9c, 0xef); + &data_byte(0xa0, 0xe0, 0x3b, 0x4d, 0xae, 0x2a, 0xf5, 0xb0); + &data_byte(0xc8, 0xeb, 0xbb, 0x3c, 0x83, 0x53, 0x99, 0x61); + &data_byte(0x17, 0x2b, 0x04, 0x7e, 0xba, 0x77, 0xd6, 0x26); + &data_byte(0xe1, 0x69, 0x14, 0x63, 0x55, 0x21, 0x0c, 0x7d); +$code.=<<___; + .long 0x80808080, 0x80808080, 0xfefefefe, 0xfefefefe + .long 0x1b1b1b1b, 0x1b1b1b1b, 0, 0 +___ + &data_byte(0x52, 0x09, 0x6a, 0xd5, 0x30, 0x36, 0xa5, 0x38); + &data_byte(0xbf, 0x40, 0xa3, 0x9e, 0x81, 0xf3, 0xd7, 0xfb); + &data_byte(0x7c, 0xe3, 0x39, 0x82, 0x9b, 0x2f, 0xff, 0x87); + &data_byte(0x34, 0x8e, 0x43, 0x44, 0xc4, 0xde, 0xe9, 0xcb); + &data_byte(0x54, 0x7b, 0x94, 0x32, 0xa6, 0xc2, 0x23, 0x3d); + &data_byte(0xee, 0x4c, 0x95, 0x0b, 0x42, 0xfa, 0xc3, 0x4e); + &data_byte(0x08, 0x2e, 0xa1, 0x66, 0x28, 0xd9, 0x24, 0xb2); + &data_byte(0x76, 0x5b, 0xa2, 0x49, 0x6d, 0x8b, 0xd1, 0x25); + &data_byte(0x72, 0xf8, 0xf6, 0x64, 0x86, 0x68, 0x98, 0x16); + &data_byte(0xd4, 0xa4, 0x5c, 0xcc, 0x5d, 0x65, 0xb6, 0x92); + &data_byte(0x6c, 0x70, 0x48, 0x50, 0xfd, 0xed, 0xb9, 0xda); + &data_byte(0x5e, 0x15, 0x46, 0x57, 0xa7, 0x8d, 0x9d, 0x84); + &data_byte(0x90, 0xd8, 0xab, 0x00, 0x8c, 0xbc, 0xd3, 0x0a); + &data_byte(0xf7, 0xe4, 0x58, 0x05, 0xb8, 0xb3, 0x45, 0x06); + &data_byte(0xd0, 0x2c, 0x1e, 0x8f, 0xca, 0x3f, 0x0f, 0x02); + &data_byte(0xc1, 0xaf, 0xbd, 0x03, 0x01, 0x13, 0x8a, 0x6b); + &data_byte(0x3a, 0x91, 0x11, 0x41, 0x4f, 0x67, 0xdc, 0xea); + &data_byte(0x97, 0xf2, 0xcf, 0xce, 0xf0, 0xb4, 0xe6, 0x73); + &data_byte(0x96, 0xac, 0x74, 0x22, 0xe7, 0xad, 0x35, 0x85); + &data_byte(0xe2, 0xf9, 0x37, 0xe8, 0x1c, 0x75, 0xdf, 0x6e); + &data_byte(0x47, 0xf1, 0x1a, 0x71, 0x1d, 0x29, 0xc5, 0x89); + &data_byte(0x6f, 0xb7, 0x62, 0x0e, 0xaa, 0x18, 0xbe, 0x1b); + &data_byte(0xfc, 0x56, 0x3e, 0x4b, 0xc6, 0xd2, 0x79, 0x20); + &data_byte(0x9a, 0xdb, 0xc0, 0xfe, 0x78, 0xcd, 0x5a, 0xf4); + &data_byte(0x1f, 0xdd, 0xa8, 0x33, 0x88, 0x07, 0xc7, 0x31); + &data_byte(0xb1, 0x12, 0x10, 0x59, 0x27, 0x80, 0xec, 0x5f); + &data_byte(0x60, 0x51, 0x7f, 0xa9, 0x19, 0xb5, 0x4a, 0x0d); + &data_byte(0x2d, 0xe5, 0x7a, 0x9f, 0x93, 0xc9, 0x9c, 0xef); + &data_byte(0xa0, 0xe0, 0x3b, 0x4d, 0xae, 0x2a, 0xf5, 0xb0); + &data_byte(0xc8, 0xeb, 0xbb, 0x3c, 0x83, 0x53, 0x99, 0x61); + &data_byte(0x17, 0x2b, 0x04, 0x7e, 0xba, 0x77, 0xd6, 0x26); + &data_byte(0xe1, 0x69, 0x14, 0x63, 0x55, 0x21, 0x0c, 0x7d); +$code.=<<___; + .long 0x80808080, 0x80808080, 0xfefefefe, 0xfefefefe + .long 0x1b1b1b1b, 0x1b1b1b1b, 0, 0 +___ + &data_byte(0x52, 0x09, 0x6a, 0xd5, 0x30, 0x36, 0xa5, 0x38); + &data_byte(0xbf, 0x40, 0xa3, 0x9e, 0x81, 0xf3, 0xd7, 0xfb); + &data_byte(0x7c, 0xe3, 0x39, 0x82, 0x9b, 0x2f, 0xff, 0x87); + &data_byte(0x34, 0x8e, 0x43, 0x44, 0xc4, 0xde, 0xe9, 0xcb); + &data_byte(0x54, 0x7b, 0x94, 0x32, 0xa6, 0xc2, 0x23, 0x3d); + &data_byte(0xee, 0x4c, 0x95, 0x0b, 0x42, 0xfa, 0xc3, 0x4e); + &data_byte(0x08, 0x2e, 0xa1, 0x66, 0x28, 0xd9, 0x24, 0xb2); + &data_byte(0x76, 0x5b, 0xa2, 0x49, 0x6d, 0x8b, 0xd1, 0x25); + &data_byte(0x72, 0xf8, 0xf6, 0x64, 0x86, 0x68, 0x98, 0x16); + &data_byte(0xd4, 0xa4, 0x5c, 0xcc, 0x5d, 0x65, 0xb6, 0x92); + &data_byte(0x6c, 0x70, 0x48, 0x50, 0xfd, 0xed, 0xb9, 0xda); + &data_byte(0x5e, 0x15, 0x46, 0x57, 0xa7, 0x8d, 0x9d, 0x84); + &data_byte(0x90, 0xd8, 0xab, 0x00, 0x8c, 0xbc, 0xd3, 0x0a); + &data_byte(0xf7, 0xe4, 0x58, 0x05, 0xb8, 0xb3, 0x45, 0x06); + &data_byte(0xd0, 0x2c, 0x1e, 0x8f, 0xca, 0x3f, 0x0f, 0x02); + &data_byte(0xc1, 0xaf, 0xbd, 0x03, 0x01, 0x13, 0x8a, 0x6b); + &data_byte(0x3a, 0x91, 0x11, 0x41, 0x4f, 0x67, 0xdc, 0xea); + &data_byte(0x97, 0xf2, 0xcf, 0xce, 0xf0, 0xb4, 0xe6, 0x73); + &data_byte(0x96, 0xac, 0x74, 0x22, 0xe7, 0xad, 0x35, 0x85); + &data_byte(0xe2, 0xf9, 0x37, 0xe8, 0x1c, 0x75, 0xdf, 0x6e); + &data_byte(0x47, 0xf1, 0x1a, 0x71, 0x1d, 0x29, 0xc5, 0x89); + &data_byte(0x6f, 0xb7, 0x62, 0x0e, 0xaa, 0x18, 0xbe, 0x1b); + &data_byte(0xfc, 0x56, 0x3e, 0x4b, 0xc6, 0xd2, 0x79, 0x20); + &data_byte(0x9a, 0xdb, 0xc0, 0xfe, 0x78, 0xcd, 0x5a, 0xf4); + &data_byte(0x1f, 0xdd, 0xa8, 0x33, 0x88, 0x07, 0xc7, 0x31); + &data_byte(0xb1, 0x12, 0x10, 0x59, 0x27, 0x80, 0xec, 0x5f); + &data_byte(0x60, 0x51, 0x7f, 0xa9, 0x19, 0xb5, 0x4a, 0x0d); + &data_byte(0x2d, 0xe5, 0x7a, 0x9f, 0x93, 0xc9, 0x9c, 0xef); + &data_byte(0xa0, 0xe0, 0x3b, 0x4d, 0xae, 0x2a, 0xf5, 0xb0); + &data_byte(0xc8, 0xeb, 0xbb, 0x3c, 0x83, 0x53, 0x99, 0x61); + &data_byte(0x17, 0x2b, 0x04, 0x7e, 0xba, 0x77, 0xd6, 0x26); + &data_byte(0xe1, 0x69, 0x14, 0x63, 0x55, 0x21, 0x0c, 0x7d); +$code.=<<___; + .long 0x80808080, 0x80808080, 0xfefefefe, 0xfefefefe + .long 0x1b1b1b1b, 0x1b1b1b1b, 0, 0 +.asciz "AES for x86_64, CRYPTOGAMS by " +.align 64 +___ + +# EXCEPTION_DISPOSITION handler (EXCEPTION_RECORD *rec,ULONG64 frame, +# CONTEXT *context,DISPATCHER_CONTEXT *disp) +if ($win64) { +$rec="%rcx"; +$frame="%rdx"; +$context="%r8"; +$disp="%r9"; + +$code.=<<___; +.extern __imp_RtlVirtualUnwind +.type block_se_handler,\@abi-omnipotent +.align 16 +block_se_handler: + push %rsi + push %rdi + push %rbx + push %rbp + push %r12 + push %r13 + push %r14 + push %r15 + pushfq + sub \$64,%rsp + + mov 120($context),%rax # pull context->Rax + mov 248($context),%rbx # pull context->Rip + + mov 8($disp),%rsi # disp->ImageBase + mov 56($disp),%r11 # disp->HandlerData + + mov 0(%r11),%r10d # HandlerData[0] + lea (%rsi,%r10),%r10 # prologue label + cmp %r10,%rbx # context->RipRsp + + mov 4(%r11),%r10d # HandlerData[1] + lea (%rsi,%r10),%r10 # epilogue label + cmp %r10,%rbx # context->Rip>=epilogue label + jae .Lin_block_prologue + + mov 24(%rax),%rax # pull saved real stack pointer + lea 48(%rax),%rax # adjust... + + mov -8(%rax),%rbx + mov -16(%rax),%rbp + mov -24(%rax),%r12 + mov -32(%rax),%r13 + mov -40(%rax),%r14 + mov -48(%rax),%r15 + mov %rbx,144($context) # restore context->Rbx + mov %rbp,160($context) # restore context->Rbp + mov %r12,216($context) # restore context->R12 + mov %r13,224($context) # restore context->R13 + mov %r14,232($context) # restore context->R14 + mov %r15,240($context) # restore context->R15 + +.Lin_block_prologue: + mov 8(%rax),%rdi + mov 16(%rax),%rsi + mov %rax,152($context) # restore context->Rsp + mov %rsi,168($context) # restore context->Rsi + mov %rdi,176($context) # restore context->Rdi + + jmp .Lcommon_seh_exit +.size block_se_handler,.-block_se_handler + +.type key_se_handler,\@abi-omnipotent +.align 16 +key_se_handler: + push %rsi + push %rdi + push %rbx + push %rbp + push %r12 + push %r13 + push %r14 + push %r15 + pushfq + sub \$64,%rsp + + mov 120($context),%rax # pull context->Rax + mov 248($context),%rbx # pull context->Rip + + mov 8($disp),%rsi # disp->ImageBase + mov 56($disp),%r11 # disp->HandlerData + + mov 0(%r11),%r10d # HandlerData[0] + lea (%rsi,%r10),%r10 # prologue label + cmp %r10,%rbx # context->RipRsp + + mov 4(%r11),%r10d # HandlerData[1] + lea (%rsi,%r10),%r10 # epilogue label + cmp %r10,%rbx # context->Rip>=epilogue label + jae .Lin_key_prologue + + lea 56(%rax),%rax + + mov -8(%rax),%rbx + mov -16(%rax),%rbp + mov -24(%rax),%r12 + mov -32(%rax),%r13 + mov -40(%rax),%r14 + mov -48(%rax),%r15 + mov %rbx,144($context) # restore context->Rbx + mov %rbp,160($context) # restore context->Rbp + mov %r12,216($context) # restore context->R12 + mov %r13,224($context) # restore context->R13 + mov %r14,232($context) # restore context->R14 + mov %r15,240($context) # restore context->R15 + +.Lin_key_prologue: + mov 8(%rax),%rdi + mov 16(%rax),%rsi + mov %rax,152($context) # restore context->Rsp + mov %rsi,168($context) # restore context->Rsi + mov %rdi,176($context) # restore context->Rdi + + jmp .Lcommon_seh_exit +.size key_se_handler,.-key_se_handler + +.type cbc_se_handler,\@abi-omnipotent +.align 16 +cbc_se_handler: + push %rsi + push %rdi + push %rbx + push %rbp + push %r12 + push %r13 + push %r14 + push %r15 + pushfq + sub \$64,%rsp + + mov 120($context),%rax # pull context->Rax + mov 248($context),%rbx # pull context->Rip + + lea .Lcbc_prologue(%rip),%r10 + cmp %r10,%rbx # context->Rip<.Lcbc_prologue + jb .Lin_cbc_prologue + + lea .Lcbc_fast_body(%rip),%r10 + cmp %r10,%rbx # context->Rip<.Lcbc_fast_body + jb .Lin_cbc_frame_setup + + lea .Lcbc_slow_prologue(%rip),%r10 + cmp %r10,%rbx # context->Rip<.Lcbc_slow_prologue + jb .Lin_cbc_body + + lea .Lcbc_slow_body(%rip),%r10 + cmp %r10,%rbx # context->Rip<.Lcbc_slow_body + jb .Lin_cbc_frame_setup + +.Lin_cbc_body: + mov 152($context),%rax # pull context->Rsp + + lea .Lcbc_epilogue(%rip),%r10 + cmp %r10,%rbx # context->Rip>=.Lcbc_epilogue + jae .Lin_cbc_prologue + + lea 8(%rax),%rax + + lea .Lcbc_popfq(%rip),%r10 + cmp %r10,%rbx # context->Rip>=.Lcbc_popfq + jae .Lin_cbc_prologue + + mov `16-8`(%rax),%rax # biased $_rsp + lea 56(%rax),%rax + +.Lin_cbc_frame_setup: + mov -16(%rax),%rbx + mov -24(%rax),%rbp + mov -32(%rax),%r12 + mov -40(%rax),%r13 + mov -48(%rax),%r14 + mov -56(%rax),%r15 + mov %rbx,144($context) # restore context->Rbx + mov %rbp,160($context) # restore context->Rbp + mov %r12,216($context) # restore context->R12 + mov %r13,224($context) # restore context->R13 + mov %r14,232($context) # restore context->R14 + mov %r15,240($context) # restore context->R15 + +.Lin_cbc_prologue: + mov 8(%rax),%rdi + mov 16(%rax),%rsi + mov %rax,152($context) # restore context->Rsp + mov %rsi,168($context) # restore context->Rsi + mov %rdi,176($context) # restore context->Rdi + +.Lcommon_seh_exit: + + mov 40($disp),%rdi # disp->ContextRecord + mov $context,%rsi # context + mov \$`1232/8`,%ecx # sizeof(CONTEXT) + .long 0xa548f3fc # cld; rep movsq + + mov $disp,%rsi + xor %rcx,%rcx # arg1, UNW_FLAG_NHANDLER + mov 8(%rsi),%rdx # arg2, disp->ImageBase + mov 0(%rsi),%r8 # arg3, disp->ControlPc + mov 16(%rsi),%r9 # arg4, disp->FunctionEntry + mov 40(%rsi),%r10 # disp->ContextRecord + lea 56(%rsi),%r11 # &disp->HandlerData + lea 24(%rsi),%r12 # &disp->EstablisherFrame + mov %r10,32(%rsp) # arg5 + mov %r11,40(%rsp) # arg6 + mov %r12,48(%rsp) # arg7 + mov %rcx,56(%rsp) # arg8, (NULL) + call *__imp_RtlVirtualUnwind(%rip) + + mov \$1,%eax # ExceptionContinueSearch + add \$64,%rsp + popfq + pop %r15 + pop %r14 + pop %r13 + pop %r12 + pop %rbp + pop %rbx + pop %rdi + pop %rsi + ret +.size cbc_se_handler,.-cbc_se_handler + +.section .pdata +.align 4 + .rva .LSEH_begin_AES_encrypt + .rva .LSEH_end_AES_encrypt + .rva .LSEH_info_AES_encrypt + + .rva .LSEH_begin_AES_decrypt + .rva .LSEH_end_AES_decrypt + .rva .LSEH_info_AES_decrypt + + .rva .LSEH_begin_AES_set_encrypt_key + .rva .LSEH_end_AES_set_encrypt_key + .rva .LSEH_info_AES_set_encrypt_key + + .rva .LSEH_begin_AES_set_decrypt_key + .rva .LSEH_end_AES_set_decrypt_key + .rva .LSEH_info_AES_set_decrypt_key + + .rva .LSEH_begin_AES_cbc_encrypt + .rva .LSEH_end_AES_cbc_encrypt + .rva .LSEH_info_AES_cbc_encrypt + +.section .xdata +.align 8 +.LSEH_info_AES_encrypt: + .byte 9,0,0,0 + .rva block_se_handler + .rva .Lenc_prologue,.Lenc_epilogue # HandlerData[] +.LSEH_info_AES_decrypt: + .byte 9,0,0,0 + .rva block_se_handler + .rva .Ldec_prologue,.Ldec_epilogue # HandlerData[] +.LSEH_info_AES_set_encrypt_key: + .byte 9,0,0,0 + .rva key_se_handler + .rva .Lenc_key_prologue,.Lenc_key_epilogue # HandlerData[] +.LSEH_info_AES_set_decrypt_key: + .byte 9,0,0,0 + .rva key_se_handler + .rva .Ldec_key_prologue,.Ldec_key_epilogue # HandlerData[] +.LSEH_info_AES_cbc_encrypt: + .byte 9,0,0,0 + .rva cbc_se_handler +___ +} + +$code =~ s/\`([^\`]*)\`/eval($1)/gem; + +print $code; + +close STDOUT; diff --git a/openssl-1.1.0h/crypto/aes/asm/aesfx-sparcv9.pl b/openssl-1.1.0h/crypto/aes/asm/aesfx-sparcv9.pl new file mode 100644 index 0000000..04b3cf7 --- /dev/null +++ b/openssl-1.1.0h/crypto/aes/asm/aesfx-sparcv9.pl @@ -0,0 +1,1270 @@ +#! /usr/bin/env perl +# Copyright 2016 The OpenSSL Project Authors. All Rights Reserved. +# +# Licensed under the OpenSSL license (the "License"). You may not use +# this file except in compliance with the License. You can obtain a copy +# in the file LICENSE in the source distribution or at +# https://www.openssl.org/source/license.html + +# +# ==================================================================== +# Written by Andy Polyakov for the OpenSSL +# project. The module is, however, dual licensed under OpenSSL and +# CRYPTOGAMS licenses depending on where you obtain it. For further +# details see http://www.openssl.org/~appro/cryptogams/. +# ==================================================================== + +# March 2016 +# +# Initial support for Fujitsu SPARC64 X/X+ comprises minimally +# required key setup and single-block procedures. +# +# April 2016 +# +# Add "teaser" CBC and CTR mode-specific subroutines. "Teaser" means +# that parallelizeable nature of CBC decrypt and CTR is not utilized +# yet. CBC encrypt on the other hand is as good as it can possibly +# get processing one byte in 4.1 cycles with 128-bit key on SPARC64 X. +# This is ~6x faster than pure software implementation... +# +# July 2016 +# +# Switch from faligndata to fshiftorx, which allows to omit alignaddr +# instructions and improve single-block and short-input performance +# with misaligned data. + +$output = pop; +open STDOUT,">$output"; + +{ +my ($inp,$out,$key,$rounds,$tmp,$mask) = map("%o$_",(0..5)); + +$code.=<<___; +#include "sparc_arch.h" + +#define LOCALS (STACK_BIAS+STACK_FRAME) + +.text + +.globl aes_fx_encrypt +.align 32 +aes_fx_encrypt: + and $inp, 7, $tmp ! is input aligned? + andn $inp, 7, $inp + ldd [$key + 0], %f6 ! round[0] + ldd [$key + 8], %f8 + mov %o7, %g1 + ld [$key + 240], $rounds + +1: call .+8 + add %o7, .Linp_align-1b, %o7 + + sll $tmp, 3, $tmp + ldd [$inp + 0], %f0 ! load input + brz,pt $tmp, .Lenc_inp_aligned + ldd [$inp + 8], %f2 + + ldd [%o7 + $tmp], %f14 ! shift left params + ldd [$inp + 16], %f4 + fshiftorx %f0, %f2, %f14, %f0 + fshiftorx %f2, %f4, %f14, %f2 + +.Lenc_inp_aligned: + ldd [$key + 16], %f10 ! round[1] + ldd [$key + 24], %f12 + + fxor %f0, %f6, %f0 ! ^=round[0] + fxor %f2, %f8, %f2 + ldd [$key + 32], %f6 ! round[2] + ldd [$key + 40], %f8 + add $key, 32, $key + sub $rounds, 4, $rounds + +.Loop_enc: + fmovd %f0, %f4 + faesencx %f2, %f10, %f0 + faesencx %f4, %f12, %f2 + ldd [$key + 16], %f10 + ldd [$key + 24], %f12 + add $key, 32, $key + + fmovd %f0, %f4 + faesencx %f2, %f6, %f0 + faesencx %f4, %f8, %f2 + ldd [$key + 0], %f6 + ldd [$key + 8], %f8 + + brnz,a $rounds, .Loop_enc + sub $rounds, 2, $rounds + + andcc $out, 7, $tmp ! is output aligned? + andn $out, 7, $out + mov 0xff, $mask + srl $mask, $tmp, $mask + add %o7, 64, %o7 + sll $tmp, 3, $tmp + + fmovd %f0, %f4 + faesencx %f2, %f10, %f0 + faesencx %f4, %f12, %f2 + ldd [%o7 + $tmp], %f14 ! shift right params + + fmovd %f0, %f4 + faesenclx %f2, %f6, %f0 + faesenclx %f4, %f8, %f2 + + bnz,pn %icc, .Lenc_out_unaligned + mov %g1, %o7 + + std %f0, [$out + 0] + retl + std %f2, [$out + 8] + +.align 16 +.Lenc_out_unaligned: + add $out, 16, $inp + orn %g0, $mask, $tmp + fshiftorx %f0, %f0, %f14, %f4 + fshiftorx %f0, %f2, %f14, %f6 + fshiftorx %f2, %f2, %f14, %f8 + + stda %f4, [$out + $mask]0xc0 ! partial store + std %f6, [$out + 8] + stda %f8, [$inp + $tmp]0xc0 ! partial store + retl + nop +.type aes_fx_encrypt,#function +.size aes_fx_encrypt,.-aes_fx_encrypt + +.globl aes_fx_decrypt +.align 32 +aes_fx_decrypt: + and $inp, 7, $tmp ! is input aligned? + andn $inp, 7, $inp + ldd [$key + 0], %f6 ! round[0] + ldd [$key + 8], %f8 + mov %o7, %g1 + ld [$key + 240], $rounds + +1: call .+8 + add %o7, .Linp_align-1b, %o7 + + sll $tmp, 3, $tmp + ldd [$inp + 0], %f0 ! load input + brz,pt $tmp, .Ldec_inp_aligned + ldd [$inp + 8], %f2 + + ldd [%o7 + $tmp], %f14 ! shift left params + ldd [$inp + 16], %f4 + fshiftorx %f0, %f2, %f14, %f0 + fshiftorx %f2, %f4, %f14, %f2 + +.Ldec_inp_aligned: + ldd [$key + 16], %f10 ! round[1] + ldd [$key + 24], %f12 + + fxor %f0, %f6, %f0 ! ^=round[0] + fxor %f2, %f8, %f2 + ldd [$key + 32], %f6 ! round[2] + ldd [$key + 40], %f8 + add $key, 32, $key + sub $rounds, 4, $rounds + +.Loop_dec: + fmovd %f0, %f4 + faesdecx %f2, %f10, %f0 + faesdecx %f4, %f12, %f2 + ldd [$key + 16], %f10 + ldd [$key + 24], %f12 + add $key, 32, $key + + fmovd %f0, %f4 + faesdecx %f2, %f6, %f0 + faesdecx %f4, %f8, %f2 + ldd [$key + 0], %f6 + ldd [$key + 8], %f8 + + brnz,a $rounds, .Loop_dec + sub $rounds, 2, $rounds + + andcc $out, 7, $tmp ! is output aligned? + andn $out, 7, $out + mov 0xff, $mask + srl $mask, $tmp, $mask + add %o7, 64, %o7 + sll $tmp, 3, $tmp + + fmovd %f0, %f4 + faesdecx %f2, %f10, %f0 + faesdecx %f4, %f12, %f2 + ldd [%o7 + $tmp], %f14 ! shift right params + + fmovd %f0, %f4 + faesdeclx %f2, %f6, %f0 + faesdeclx %f4, %f8, %f2 + + bnz,pn %icc, .Ldec_out_unaligned + mov %g1, %o7 + + std %f0, [$out + 0] + retl + std %f2, [$out + 8] + +.align 16 +.Ldec_out_unaligned: + add $out, 16, $inp + orn %g0, $mask, $tmp + fshiftorx %f0, %f0, %f14, %f4 + fshiftorx %f0, %f2, %f14, %f6 + fshiftorx %f2, %f2, %f14, %f8 + + stda %f4, [$out + $mask]0xc0 ! partial store + std %f6, [$out + 8] + stda %f8, [$inp + $tmp]0xc0 ! partial store + retl + nop +.type aes_fx_decrypt,#function +.size aes_fx_decrypt,.-aes_fx_decrypt +___ +} +{ +my ($inp,$bits,$out,$tmp,$inc) = map("%o$_",(0..5)); +$code.=<<___; +.globl aes_fx_set_decrypt_key +.align 32 +aes_fx_set_decrypt_key: + b .Lset_encrypt_key + mov -1, $inc + retl + nop +.type aes_fx_set_decrypt_key,#function +.size aes_fx_set_decrypt_key,.-aes_fx_set_decrypt_key + +.globl aes_fx_set_encrypt_key +.align 32 +aes_fx_set_encrypt_key: + mov 1, $inc + nop +.Lset_encrypt_key: + and $inp, 7, $tmp + andn $inp, 7, $inp + sll $tmp, 3, $tmp + mov %o7, %g1 + +1: call .+8 + add %o7, .Linp_align-1b, %o7 + + ldd [%o7 + $tmp], %f10 ! shift left params + mov %g1, %o7 + + cmp $bits, 192 + ldd [$inp + 0], %f0 + bl,pt %icc, .L128 + ldd [$inp + 8], %f2 + + be,pt %icc, .L192 + ldd [$inp + 16], %f4 + brz,pt $tmp, .L256aligned + ldd [$inp + 24], %f6 + + ldd [$inp + 32], %f8 + fshiftorx %f0, %f2, %f10, %f0 + fshiftorx %f2, %f4, %f10, %f2 + fshiftorx %f4, %f6, %f10, %f4 + fshiftorx %f6, %f8, %f10, %f6 + +.L256aligned: + mov 14, $bits + and $inc, `14*16`, $tmp + st $bits, [$out + 240] ! store rounds + add $out, $tmp, $out ! start or end of key schedule + sllx $inc, 4, $inc ! 16 or -16 +___ +for ($i=0; $i<6; $i++) { + $code.=<<___; + std %f0, [$out + 0] + faeskeyx %f6, `0x10+$i`, %f0 + std %f2, [$out + 8] + add $out, $inc, $out + faeskeyx %f0, 0x00, %f2 + std %f4, [$out + 0] + faeskeyx %f2, 0x01, %f4 + std %f6, [$out + 8] + add $out, $inc, $out + faeskeyx %f4, 0x00, %f6 +___ +} +$code.=<<___; + std %f0, [$out + 0] + faeskeyx %f6, `0x10+$i`, %f0 + std %f2, [$out + 8] + add $out, $inc, $out + faeskeyx %f0, 0x00, %f2 + std %f4,[$out + 0] + std %f6,[$out + 8] + add $out, $inc, $out + std %f0,[$out + 0] + std %f2,[$out + 8] + retl + xor %o0, %o0, %o0 ! return 0 + +.align 16 +.L192: + brz,pt $tmp, .L192aligned + nop + + ldd [$inp + 24], %f6 + fshiftorx %f0, %f2, %f10, %f0 + fshiftorx %f2, %f4, %f10, %f2 + fshiftorx %f4, %f6, %f10, %f4 + +.L192aligned: + mov 12, $bits + and $inc, `12*16`, $tmp + st $bits, [$out + 240] ! store rounds + add $out, $tmp, $out ! start or end of key schedule + sllx $inc, 4, $inc ! 16 or -16 +___ +for ($i=0; $i<8; $i+=2) { + $code.=<<___; + std %f0, [$out + 0] + faeskeyx %f4, `0x10+$i`, %f0 + std %f2, [$out + 8] + add $out, $inc, $out + faeskeyx %f0, 0x00, %f2 + std %f4, [$out + 0] + faeskeyx %f2, 0x00, %f4 + std %f0, [$out + 8] + add $out, $inc, $out + faeskeyx %f4, `0x10+$i+1`, %f0 + std %f2, [$out + 0] + faeskeyx %f0, 0x00, %f2 + std %f4, [$out + 8] + add $out, $inc, $out +___ +$code.=<<___ if ($i<6); + faeskeyx %f2, 0x00, %f4 +___ +} +$code.=<<___; + std %f0, [$out + 0] + std %f2, [$out + 8] + retl + xor %o0, %o0, %o0 ! return 0 + +.align 16 +.L128: + brz,pt $tmp, .L128aligned + nop + + ldd [$inp + 16], %f4 + fshiftorx %f0, %f2, %f10, %f0 + fshiftorx %f2, %f4, %f10, %f2 + +.L128aligned: + mov 10, $bits + and $inc, `10*16`, $tmp + st $bits, [$out + 240] ! store rounds + add $out, $tmp, $out ! start or end of key schedule + sllx $inc, 4, $inc ! 16 or -16 +___ +for ($i=0; $i<10; $i++) { + $code.=<<___; + std %f0, [$out + 0] + faeskeyx %f2, `0x10+$i`, %f0 + std %f2, [$out + 8] + add $out, $inc, $out + faeskeyx %f0, 0x00, %f2 +___ +} +$code.=<<___; + std %f0, [$out + 0] + std %f2, [$out + 8] + retl + xor %o0, %o0, %o0 ! return 0 +.type aes_fx_set_encrypt_key,#function +.size aes_fx_set_encrypt_key,.-aes_fx_set_encrypt_key +___ +} +{ +my ($inp,$out,$len,$key,$ivp,$dir) = map("%i$_",(0..5)); +my ($rounds,$inner,$end,$inc,$ialign,$oalign,$mask) = map("%l$_",(0..7)); +my ($iv0,$iv1,$r0hi,$r0lo,$rlhi,$rllo,$in0,$in1,$intail,$outhead,$fshift) + = map("%f$_",grep { !($_ & 1) } (16 .. 62)); +my ($ileft,$iright) = ($ialign,$oalign); + +$code.=<<___; +.globl aes_fx_cbc_encrypt +.align 32 +aes_fx_cbc_encrypt: + save %sp, -STACK_FRAME-16, %sp + srln $len, 4, $len + and $inp, 7, $ialign + andn $inp, 7, $inp + brz,pn $len, .Lcbc_no_data + sll $ialign, 3, $ileft + +1: call .+8 + add %o7, .Linp_align-1b, %o7 + + ld [$key + 240], $rounds + and $out, 7, $oalign + ld [$ivp + 0], %f0 ! load ivec + andn $out, 7, $out + ld [$ivp + 4], %f1 + sll $oalign, 3, $mask + ld [$ivp + 8], %f2 + ld [$ivp + 12], %f3 + + sll $rounds, 4, $rounds + add $rounds, $key, $end + ldd [$key + 0], $r0hi ! round[0] + ldd [$key + 8], $r0lo + + add $inp, 16, $inp + sub $len, 1, $len + ldd [$end + 0], $rlhi ! round[last] + ldd [$end + 8], $rllo + + mov 16, $inc + movrz $len, 0, $inc + ldd [$key + 16], %f10 ! round[1] + ldd [$key + 24], %f12 + + ldd [%o7 + $ileft], $fshift ! shift left params + add %o7, 64, %o7 + ldd [$inp - 16], $in0 ! load input + ldd [$inp - 8], $in1 + ldda [$inp]0x82, $intail ! non-faulting load + brz $dir, .Lcbc_decrypt + add $inp, $inc, $inp ! inp+=16 + + fxor $r0hi, %f0, %f0 ! ivec^=round[0] + fxor $r0lo, %f2, %f2 + fshiftorx $in0, $in1, $fshift, $in0 + fshiftorx $in1, $intail, $fshift, $in1 + nop + +.Loop_cbc_enc: + fxor $in0, %f0, %f0 ! inp^ivec^round[0] + fxor $in1, %f2, %f2 + ldd [$key + 32], %f6 ! round[2] + ldd [$key + 40], %f8 + add $key, 32, $end + sub $rounds, 16*6, $inner + +.Lcbc_enc: + fmovd %f0, %f4 + faesencx %f2, %f10, %f0 + faesencx %f4, %f12, %f2 + ldd [$end + 16], %f10 + ldd [$end + 24], %f12 + add $end, 32, $end + + fmovd %f0, %f4 + faesencx %f2, %f6, %f0 + faesencx %f4, %f8, %f2 + ldd [$end + 0], %f6 + ldd [$end + 8], %f8 + + brnz,a $inner, .Lcbc_enc + sub $inner, 16*2, $inner + + fmovd %f0, %f4 + faesencx %f2, %f10, %f0 + faesencx %f4, %f12, %f2 + ldd [$end + 16], %f10 ! round[last-1] + ldd [$end + 24], %f12 + + movrz $len, 0, $inc + fmovd $intail, $in0 + ldd [$inp - 8], $in1 ! load next input block + ldda [$inp]0x82, $intail ! non-faulting load + add $inp, $inc, $inp ! inp+=16 + + fmovd %f0, %f4 + faesencx %f2, %f6, %f0 + faesencx %f4, %f8, %f2 + + fshiftorx $in0, $in1, $fshift, $in0 + fshiftorx $in1, $intail, $fshift, $in1 + + fmovd %f0, %f4 + faesencx %f2, %f10, %f0 + faesencx %f4, %f12, %f2 + ldd [$key + 16], %f10 ! round[1] + ldd [$key + 24], %f12 + + fxor $r0hi, $in0, $in0 ! inp^=round[0] + fxor $r0lo, $in1, $in1 + + fmovd %f0, %f4 + faesenclx %f2, $rlhi, %f0 + faesenclx %f4, $rllo, %f2 + + brnz,pn $oalign, .Lcbc_enc_unaligned_out + nop + + std %f0, [$out + 0] + std %f2, [$out + 8] + add $out, 16, $out + + brnz,a $len, .Loop_cbc_enc + sub $len, 1, $len + + st %f0, [$ivp + 0] ! output ivec + st %f1, [$ivp + 4] + st %f2, [$ivp + 8] + st %f3, [$ivp + 12] + +.Lcbc_no_data: + ret + restore + +.align 32 +.Lcbc_enc_unaligned_out: + ldd [%o7 + $mask], $fshift ! shift right params + mov 0xff, $mask + srl $mask, $oalign, $mask + sub %g0, $ileft, $iright + + fshiftorx %f0, %f0, $fshift, %f6 + fshiftorx %f0, %f2, $fshift, %f8 + + stda %f6, [$out + $mask]0xc0 ! partial store + orn %g0, $mask, $mask + std %f8, [$out + 8] + add $out, 16, $out + brz $len, .Lcbc_enc_unaligned_out_done + sub $len, 1, $len + b .Loop_cbc_enc_unaligned_out + nop + +.align 32 +.Loop_cbc_enc_unaligned_out: + fmovd %f2, $outhead + fxor $in0, %f0, %f0 ! inp^ivec^round[0] + fxor $in1, %f2, %f2 + ldd [$key + 32], %f6 ! round[2] + ldd [$key + 40], %f8 + + fmovd %f0, %f4 + faesencx %f2, %f10, %f0 + faesencx %f4, %f12, %f2 + ldd [$key + 48], %f10 ! round[3] + ldd [$key + 56], %f12 + + ldx [$inp - 16], %o0 + ldx [$inp - 8], %o1 + brz $ileft, .Lcbc_enc_aligned_inp + movrz $len, 0, $inc + + ldx [$inp], %o2 + sllx %o0, $ileft, %o0 + srlx %o1, $iright, %g1 + sllx %o1, $ileft, %o1 + or %g1, %o0, %o0 + srlx %o2, $iright, %o2 + or %o2, %o1, %o1 + +.Lcbc_enc_aligned_inp: + fmovd %f0, %f4 + faesencx %f2, %f6, %f0 + faesencx %f4, %f8, %f2 + ldd [$key + 64], %f6 ! round[4] + ldd [$key + 72], %f8 + add $key, 64, $end + sub $rounds, 16*8, $inner + + stx %o0, [%sp + LOCALS + 0] + stx %o1, [%sp + LOCALS + 8] + add $inp, $inc, $inp ! inp+=16 + nop + +.Lcbc_enc_unaligned: + fmovd %f0, %f4 + faesencx %f2, %f10, %f0 + faesencx %f4, %f12, %f2 + ldd [$end + 16], %f10 + ldd [$end + 24], %f12 + add $end, 32, $end + + fmovd %f0, %f4 + faesencx %f2, %f6, %f0 + faesencx %f4, %f8, %f2 + ldd [$end + 0], %f6 + ldd [$end + 8], %f8 + + brnz,a $inner, .Lcbc_enc_unaligned + sub $inner, 16*2, $inner + + fmovd %f0, %f4 + faesencx %f2, %f10, %f0 + faesencx %f4, %f12, %f2 + ldd [$end + 16], %f10 ! round[last-1] + ldd [$end + 24], %f12 + + fmovd %f0, %f4 + faesencx %f2, %f6, %f0 + faesencx %f4, %f8, %f2 + + ldd [%sp + LOCALS + 0], $in0 + ldd [%sp + LOCALS + 8], $in1 + + fmovd %f0, %f4 + faesencx %f2, %f10, %f0 + faesencx %f4, %f12, %f2 + ldd [$key + 16], %f10 ! round[1] + ldd [$key + 24], %f12 + + fxor $r0hi, $in0, $in0 ! inp^=round[0] + fxor $r0lo, $in1, $in1 + + fmovd %f0, %f4 + faesenclx %f2, $rlhi, %f0 + faesenclx %f4, $rllo, %f2 + + fshiftorx $outhead, %f0, $fshift, %f6 + fshiftorx %f0, %f2, $fshift, %f8 + std %f6, [$out + 0] + std %f8, [$out + 8] + add $out, 16, $out + + brnz,a $len, .Loop_cbc_enc_unaligned_out + sub $len, 1, $len + +.Lcbc_enc_unaligned_out_done: + fshiftorx %f2, %f2, $fshift, %f8 + stda %f8, [$out + $mask]0xc0 ! partial store + + st %f0, [$ivp + 0] ! output ivec + st %f1, [$ivp + 4] + st %f2, [$ivp + 8] + st %f3, [$ivp + 12] + + ret + restore + +.align 32 +.Lcbc_decrypt: + fshiftorx $in0, $in1, $fshift, $in0 + fshiftorx $in1, $intail, $fshift, $in1 + fmovd %f0, $iv0 + fmovd %f2, $iv1 + +.Loop_cbc_dec: + fxor $in0, $r0hi, %f0 ! inp^round[0] + fxor $in1, $r0lo, %f2 + ldd [$key + 32], %f6 ! round[2] + ldd [$key + 40], %f8 + add $key, 32, $end + sub $rounds, 16*6, $inner + +.Lcbc_dec: + fmovd %f0, %f4 + faesdecx %f2, %f10, %f0 + faesdecx %f4, %f12, %f2 + ldd [$end + 16], %f10 + ldd [$end + 24], %f12 + add $end, 32, $end + + fmovd %f0, %f4 + faesdecx %f2, %f6, %f0 + faesdecx %f4, %f8, %f2 + ldd [$end + 0], %f6 + ldd [$end + 8], %f8 + + brnz,a $inner, .Lcbc_dec + sub $inner, 16*2, $inner + + fmovd %f0, %f4 + faesdecx %f2, %f10, %f0 + faesdecx %f4, %f12, %f2 + ldd [$end + 16], %f10 ! round[last-1] + ldd [$end + 24], %f12 + + fmovd %f0, %f4 + faesdecx %f2, %f6, %f0 + faesdecx %f4, %f8, %f2 + fxor $iv0, $rlhi, %f6 ! ivec^round[last] + fxor $iv1, $rllo, %f8 + fmovd $in0, $iv0 + fmovd $in1, $iv1 + + movrz $len, 0, $inc + fmovd $intail, $in0 + ldd [$inp - 8], $in1 ! load next input block + ldda [$inp]0x82, $intail ! non-faulting load + add $inp, $inc, $inp ! inp+=16 + + fmovd %f0, %f4 + faesdecx %f2, %f10, %f0 + faesdecx %f4, %f12, %f2 + ldd [$key + 16], %f10 ! round[1] + ldd [$key + 24], %f12 + + fshiftorx $in0, $in1, $fshift, $in0 + fshiftorx $in1, $intail, $fshift, $in1 + + fmovd %f0, %f4 + faesdeclx %f2, %f6, %f0 + faesdeclx %f4, %f8, %f2 + + brnz,pn $oalign, .Lcbc_dec_unaligned_out + nop + + std %f0, [$out + 0] + std %f2, [$out + 8] + add $out, 16, $out + + brnz,a $len, .Loop_cbc_dec + sub $len, 1, $len + + st $iv0, [$ivp + 0] ! output ivec + st $iv0#lo, [$ivp + 4] + st $iv1, [$ivp + 8] + st $iv1#lo, [$ivp + 12] + + ret + restore + +.align 32 +.Lcbc_dec_unaligned_out: + ldd [%o7 + $mask], $fshift ! shift right params + mov 0xff, $mask + srl $mask, $oalign, $mask + sub %g0, $ileft, $iright + + fshiftorx %f0, %f0, $fshift, %f6 + fshiftorx %f0, %f2, $fshift, %f8 + + stda %f6, [$out + $mask]0xc0 ! partial store + orn %g0, $mask, $mask + std %f8, [$out + 8] + add $out, 16, $out + brz $len, .Lcbc_dec_unaligned_out_done + sub $len, 1, $len + b .Loop_cbc_dec_unaligned_out + nop + +.align 32 +.Loop_cbc_dec_unaligned_out: + fmovd %f2, $outhead + fxor $in0, $r0hi, %f0 ! inp^round[0] + fxor $in1, $r0lo, %f2 + ldd [$key + 32], %f6 ! round[2] + ldd [$key + 40], %f8 + + fmovd %f0, %f4 + faesdecx %f2, %f10, %f0 + faesdecx %f4, %f12, %f2 + ldd [$key + 48], %f10 ! round[3] + ldd [$key + 56], %f12 + + ldx [$inp - 16], %o0 + ldx [$inp - 8], %o1 + brz $ileft, .Lcbc_dec_aligned_inp + movrz $len, 0, $inc + + ldx [$inp], %o2 + sllx %o0, $ileft, %o0 + srlx %o1, $iright, %g1 + sllx %o1, $ileft, %o1 + or %g1, %o0, %o0 + srlx %o2, $iright, %o2 + or %o2, %o1, %o1 + +.Lcbc_dec_aligned_inp: + fmovd %f0, %f4 + faesdecx %f2, %f6, %f0 + faesdecx %f4, %f8, %f2 + ldd [$key + 64], %f6 ! round[4] + ldd [$key + 72], %f8 + add $key, 64, $end + sub $rounds, 16*8, $inner + + stx %o0, [%sp + LOCALS + 0] + stx %o1, [%sp + LOCALS + 8] + add $inp, $inc, $inp ! inp+=16 + nop + +.Lcbc_dec_unaligned: + fmovd %f0, %f4 + faesdecx %f2, %f10, %f0 + faesdecx %f4, %f12, %f2 + ldd [$end + 16], %f10 + ldd [$end + 24], %f12 + add $end, 32, $end + + fmovd %f0, %f4 + faesdecx %f2, %f6, %f0 + faesdecx %f4, %f8, %f2 + ldd [$end + 0], %f6 + ldd [$end + 8], %f8 + + brnz,a $inner, .Lcbc_dec_unaligned + sub $inner, 16*2, $inner + + fmovd %f0, %f4 + faesdecx %f2, %f10, %f0 + faesdecx %f4, %f12, %f2 + ldd [$end + 16], %f10 ! round[last-1] + ldd [$end + 24], %f12 + + fmovd %f0, %f4 + faesdecx %f2, %f6, %f0 + faesdecx %f4, %f8, %f2 + + fxor $iv0, $rlhi, %f6 ! ivec^round[last] + fxor $iv1, $rllo, %f8 + fmovd $in0, $iv0 + fmovd $in1, $iv1 + ldd [%sp + LOCALS + 0], $in0 + ldd [%sp + LOCALS + 8], $in1 + + fmovd %f0, %f4 + faesdecx %f2, %f10, %f0 + faesdecx %f4, %f12, %f2 + ldd [$key + 16], %f10 ! round[1] + ldd [$key + 24], %f12 + + fmovd %f0, %f4 + faesdeclx %f2, %f6, %f0 + faesdeclx %f4, %f8, %f2 + + fshiftorx $outhead, %f0, $fshift, %f6 + fshiftorx %f0, %f2, $fshift, %f8 + std %f6, [$out + 0] + std %f8, [$out + 8] + add $out, 16, $out + + brnz,a $len, .Loop_cbc_dec_unaligned_out + sub $len, 1, $len + +.Lcbc_dec_unaligned_out_done: + fshiftorx %f2, %f2, $fshift, %f8 + stda %f8, [$out + $mask]0xc0 ! partial store + + st $iv0, [$ivp + 0] ! output ivec + st $iv0#lo, [$ivp + 4] + st $iv1, [$ivp + 8] + st $iv1#lo, [$ivp + 12] + + ret + restore +.type aes_fx_cbc_encrypt,#function +.size aes_fx_cbc_encrypt,.-aes_fx_cbc_encrypt +___ +} +{ +my ($inp,$out,$len,$key,$ivp) = map("%i$_",(0..5)); +my ($rounds,$inner,$end,$inc,$ialign,$oalign,$mask) = map("%l$_",(0..7)); +my ($ctr0,$ctr1,$r0hi,$r0lo,$rlhi,$rllo,$in0,$in1,$intail,$outhead,$fshift) + = map("%f$_",grep { !($_ & 1) } (16 .. 62)); +my ($ileft,$iright) = ($ialign, $oalign); +my $one = "%f14"; + +$code.=<<___; +.globl aes_fx_ctr32_encrypt_blocks +.align 32 +aes_fx_ctr32_encrypt_blocks: + save %sp, -STACK_FRAME-16, %sp + srln $len, 0, $len + and $inp, 7, $ialign + andn $inp, 7, $inp + brz,pn $len, .Lctr32_no_data + sll $ialign, 3, $ileft + +.Lpic: call .+8 + add %o7, .Linp_align - .Lpic, %o7 + + ld [$key + 240], $rounds + and $out, 7, $oalign + ld [$ivp + 0], $ctr0 ! load counter + andn $out, 7, $out + ld [$ivp + 4], $ctr0#lo + sll $oalign, 3, $mask + ld [$ivp + 8], $ctr1 + ld [$ivp + 12], $ctr1#lo + ldd [%o7 + 128], $one + + sll $rounds, 4, $rounds + add $rounds, $key, $end + ldd [$key + 0], $r0hi ! round[0] + ldd [$key + 8], $r0lo + + add $inp, 16, $inp + sub $len, 1, $len + ldd [$key + 16], %f10 ! round[1] + ldd [$key + 24], %f12 + + mov 16, $inc + movrz $len, 0, $inc + ldd [$end + 0], $rlhi ! round[last] + ldd [$end + 8], $rllo + + ldd [%o7 + $ileft], $fshift ! shiftleft params + add %o7, 64, %o7 + ldd [$inp - 16], $in0 ! load input + ldd [$inp - 8], $in1 + ldda [$inp]0x82, $intail ! non-faulting load + add $inp, $inc, $inp ! inp+=16 + + fshiftorx $in0, $in1, $fshift, $in0 + fshiftorx $in1, $intail, $fshift, $in1 + +.Loop_ctr32: + fxor $ctr0, $r0hi, %f0 ! counter^round[0] + fxor $ctr1, $r0lo, %f2 + ldd [$key + 32], %f6 ! round[2] + ldd [$key + 40], %f8 + add $key, 32, $end + sub $rounds, 16*6, $inner + +.Lctr32_enc: + fmovd %f0, %f4 + faesencx %f2, %f10, %f0 + faesencx %f4, %f12, %f2 + ldd [$end + 16], %f10 + ldd [$end + 24], %f12 + add $end, 32, $end + + fmovd %f0, %f4 + faesencx %f2, %f6, %f0 + faesencx %f4, %f8, %f2 + ldd [$end + 0], %f6 + ldd [$end + 8], %f8 + + brnz,a $inner, .Lctr32_enc + sub $inner, 16*2, $inner + + fmovd %f0, %f4 + faesencx %f2, %f10, %f0 + faesencx %f4, %f12, %f2 + ldd [$end + 16], %f10 ! round[last-1] + ldd [$end + 24], %f12 + + fmovd %f0, %f4 + faesencx %f2, %f6, %f0 + faesencx %f4, %f8, %f2 + fxor $in0, $rlhi, %f6 ! inp^round[last] + fxor $in1, $rllo, %f8 + + movrz $len, 0, $inc + fmovd $intail, $in0 + ldd [$inp - 8], $in1 ! load next input block + ldda [$inp]0x82, $intail ! non-faulting load + add $inp, $inc, $inp ! inp+=16 + + fmovd %f0, %f4 + faesencx %f2, %f10, %f0 + faesencx %f4, %f12, %f2 + ldd [$key + 16], %f10 ! round[1] + ldd [$key + 24], %f12 + + fshiftorx $in0, $in1, $fshift, $in0 + fshiftorx $in1, $intail, $fshift, $in1 + fpadd32 $ctr1, $one, $ctr1 ! increment counter + + fmovd %f0, %f4 + faesenclx %f2, %f6, %f0 + faesenclx %f4, %f8, %f2 + + brnz,pn $oalign, .Lctr32_unaligned_out + nop + + std %f0, [$out + 0] + std %f2, [$out + 8] + add $out, 16, $out + + brnz,a $len, .Loop_ctr32 + sub $len, 1, $len + +.Lctr32_no_data: + ret + restore + +.align 32 +.Lctr32_unaligned_out: + ldd [%o7 + $mask], $fshift ! shift right params + mov 0xff, $mask + srl $mask, $oalign, $mask + sub %g0, $ileft, $iright + + fshiftorx %f0, %f0, $fshift, %f6 + fshiftorx %f0, %f2, $fshift, %f8 + + stda %f6, [$out + $mask]0xc0 ! partial store + orn %g0, $mask, $mask + std %f8, [$out + 8] + add $out, 16, $out + brz $len, .Lctr32_unaligned_out_done + sub $len, 1, $len + b .Loop_ctr32_unaligned_out + nop + +.align 32 +.Loop_ctr32_unaligned_out: + fmovd %f2, $outhead + fxor $ctr0, $r0hi, %f0 ! counter^round[0] + fxor $ctr1, $r0lo, %f2 + ldd [$key + 32], %f6 ! round[2] + ldd [$key + 40], %f8 + + fmovd %f0, %f4 + faesencx %f2, %f10, %f0 + faesencx %f4, %f12, %f2 + ldd [$key + 48], %f10 ! round[3] + ldd [$key + 56], %f12 + + ldx [$inp - 16], %o0 + ldx [$inp - 8], %o1 + brz $ileft, .Lctr32_aligned_inp + movrz $len, 0, $inc + + ldx [$inp], %o2 + sllx %o0, $ileft, %o0 + srlx %o1, $iright, %g1 + sllx %o1, $ileft, %o1 + or %g1, %o0, %o0 + srlx %o2, $iright, %o2 + or %o2, %o1, %o1 + +.Lctr32_aligned_inp: + fmovd %f0, %f4 + faesencx %f2, %f6, %f0 + faesencx %f4, %f8, %f2 + ldd [$key + 64], %f6 ! round[4] + ldd [$key + 72], %f8 + add $key, 64, $end + sub $rounds, 16*8, $inner + + stx %o0, [%sp + LOCALS + 0] + stx %o1, [%sp + LOCALS + 8] + add $inp, $inc, $inp ! inp+=16 + nop + +.Lctr32_enc_unaligned: + fmovd %f0, %f4 + faesencx %f2, %f10, %f0 + faesencx %f4, %f12, %f2 + ldd [$end + 16], %f10 + ldd [$end + 24], %f12 + add $end, 32, $end + + fmovd %f0, %f4 + faesencx %f2, %f6, %f0 + faesencx %f4, %f8, %f2 + ldd [$end + 0], %f6 + ldd [$end + 8], %f8 + + brnz,a $inner, .Lctr32_enc_unaligned + sub $inner, 16*2, $inner + + fmovd %f0, %f4 + faesencx %f2, %f10, %f0 + faesencx %f4, %f12, %f2 + ldd [$end + 16], %f10 ! round[last-1] + ldd [$end + 24], %f12 + fpadd32 $ctr1, $one, $ctr1 ! increment counter + + fmovd %f0, %f4 + faesencx %f2, %f6, %f0 + faesencx %f4, %f8, %f2 + fxor $in0, $rlhi, %f6 ! inp^round[last] + fxor $in1, $rllo, %f8 + ldd [%sp + LOCALS + 0], $in0 + ldd [%sp + LOCALS + 8], $in1 + + fmovd %f0, %f4 + faesencx %f2, %f10, %f0 + faesencx %f4, %f12, %f2 + ldd [$key + 16], %f10 ! round[1] + ldd [$key + 24], %f12 + + fmovd %f0, %f4 + faesenclx %f2, %f6, %f0 + faesenclx %f4, %f8, %f2 + + fshiftorx $outhead, %f0, $fshift, %f6 + fshiftorx %f0, %f2, $fshift, %f8 + std %f6, [$out + 0] + std %f8, [$out + 8] + add $out, 16, $out + + brnz,a $len, .Loop_ctr32_unaligned_out + sub $len, 1, $len + +.Lctr32_unaligned_out_done: + fshiftorx %f2, %f2, $fshift, %f8 + stda %f8, [$out + $mask]0xc0 ! partial store + + ret + restore +.type aes_fx_ctr32_encrypt_blocks,#function +.size aes_fx_ctr32_encrypt_blocks,.-aes_fx_ctr32_encrypt_blocks + +.align 32 +.Linp_align: ! fshiftorx parameters for left shift toward %rs1 + .byte 0, 0, 64, 0, 0, 64, 0, -64 + .byte 0, 0, 56, 8, 0, 56, 8, -56 + .byte 0, 0, 48, 16, 0, 48, 16, -48 + .byte 0, 0, 40, 24, 0, 40, 24, -40 + .byte 0, 0, 32, 32, 0, 32, 32, -32 + .byte 0, 0, 24, 40, 0, 24, 40, -24 + .byte 0, 0, 16, 48, 0, 16, 48, -16 + .byte 0, 0, 8, 56, 0, 8, 56, -8 +.Lout_align: ! fshiftorx parameters for right shift toward %rs2 + .byte 0, 0, 0, 64, 0, 0, 64, 0 + .byte 0, 0, 8, 56, 0, 8, 56, -8 + .byte 0, 0, 16, 48, 0, 16, 48, -16 + .byte 0, 0, 24, 40, 0, 24, 40, -24 + .byte 0, 0, 32, 32, 0, 32, 32, -32 + .byte 0, 0, 40, 24, 0, 40, 24, -40 + .byte 0, 0, 48, 16, 0, 48, 16, -48 + .byte 0, 0, 56, 8, 0, 56, 8, -56 +.Lone: + .word 0, 1 +.asciz "AES for Fujitsu SPARC64 X, CRYPTOGAMS by " +.align 4 +___ +} +# Purpose of these subroutines is to explicitly encode VIS instructions, +# so that one can compile the module without having to specify VIS +# extensions on compiler command line, e.g. -xarch=v9 vs. -xarch=v9a. +# Idea is to reserve for option to produce "universal" binary and let +# programmer detect if current CPU is VIS capable at run-time. +sub unvis { +my ($mnemonic,$rs1,$rs2,$rd)=@_; +my ($ref,$opf); +my %visopf = ( "faligndata" => 0x048, + "bshuffle" => 0x04c, + "fpadd32" => 0x052, + "fxor" => 0x06c, + "fsrc2" => 0x078 ); + + $ref = "$mnemonic\t$rs1,$rs2,$rd"; + + if ($opf=$visopf{$mnemonic}) { + foreach ($rs1,$rs2,$rd) { + return $ref if (!/%f([0-9]{1,2})/); + $_=$1; + if ($1>=32) { + return $ref if ($1&1); + # re-encode for upper double register addressing + $_=($1|$1>>5)&31; + } + } + + return sprintf ".word\t0x%08x !%s", + 0x81b00000|$rd<<25|$rs1<<14|$opf<<5|$rs2, + $ref; + } else { + return $ref; + } +} + +sub unvis3 { +my ($mnemonic,$rs1,$rs2,$rd)=@_; +my %bias = ( "g" => 0, "o" => 8, "l" => 16, "i" => 24 ); +my ($ref,$opf); +my %visopf = ( "alignaddr" => 0x018, + "bmask" => 0x019, + "alignaddrl" => 0x01a ); + + $ref = "$mnemonic\t$rs1,$rs2,$rd"; + + if ($opf=$visopf{$mnemonic}) { + foreach ($rs1,$rs2,$rd) { + return $ref if (!/%([goli])([0-9])/); + $_=$bias{$1}+$2; + } + + return sprintf ".word\t0x%08x !%s", + 0x81b00000|$rd<<25|$rs1<<14|$opf<<5|$rs2, + $ref; + } else { + return $ref; + } +} + +sub unfx { +my ($mnemonic,$rs1,$rs2,$rd)=@_; +my ($ref,$opf); +my %aesopf = ( "faesencx" => 0x90, + "faesdecx" => 0x91, + "faesenclx" => 0x92, + "faesdeclx" => 0x93, + "faeskeyx" => 0x94 ); + + $ref = "$mnemonic\t$rs1,$rs2,$rd"; + + if (defined($opf=$aesopf{$mnemonic})) { + $rs2 = ($rs2 =~ /%f([0-6]*[02468])/) ? (($1|$1>>5)&31) : $rs2; + $rs2 = oct($rs2) if ($rs2 =~ /^0/); + + foreach ($rs1,$rd) { + return $ref if (!/%f([0-9]{1,2})/); + $_=$1; + if ($1>=32) { + return $ref if ($1&1); + # re-encode for upper double register addressing + $_=($1|$1>>5)&31; + } + } + + return sprintf ".word\t0x%08x !%s", + 2<<30|$rd<<25|0x36<<19|$rs1<<14|$opf<<5|$rs2, + $ref; + } else { + return $ref; + } +} + +sub unfx3src { +my ($mnemonic,$rs1,$rs2,$rs3,$rd)=@_; +my ($ref,$opf); +my %aesopf = ( "fshiftorx" => 0x0b ); + + $ref = "$mnemonic\t$rs1,$rs2,$rs3,$rd"; + + if (defined($opf=$aesopf{$mnemonic})) { + foreach ($rs1,$rs2,$rs3,$rd) { + return $ref if (!/%f([0-9]{1,2})/); + $_=$1; + if ($1>=32) { + return $ref if ($1&1); + # re-encode for upper double register addressing + $_=($1|$1>>5)&31; + } + } + + return sprintf ".word\t0x%08x !%s", + 2<<30|$rd<<25|0x37<<19|$rs1<<14|$rs3<<9|$opf<<5|$rs2, + $ref; + } else { + return $ref; + } +} + +foreach (split("\n",$code)) { + s/\`([^\`]*)\`/eval $1/ge; + + s/%f([0-9]+)#lo/sprintf "%%f%d",$1+1/ge; + + s/\b(faes[^x]{3,4}x)\s+(%f[0-9]{1,2}),\s*([%fx0-9]+),\s*(%f[0-9]{1,2})/ + &unfx($1,$2,$3,$4) + /ge or + s/\b([f][^\s]*)\s+(%f[0-9]{1,2}),\s*(%f[0-9]{1,2}),\s*(%f[0-9]{1,2}),\s*(%f[0-9]{1,2})/ + &unfx3src($1,$2,$3,$4,$5) + /ge or + s/\b([fb][^\s]*)\s+(%f[0-9]{1,2}),\s*(%f[0-9]{1,2}),\s*(%f[0-9]{1,2})/ + &unvis($1,$2,$3,$4) + /ge or + s/\b(alignaddr[l]*)\s+(%[goli][0-7]),\s*(%[goli][0-7]),\s*(%[goli][0-7])/ + &unvis3($1,$2,$3,$4) + /ge; + print $_,"\n"; +} + +close STDOUT; diff --git a/openssl-1.1.0h/crypto/aes/asm/aesni-mb-x86_64.pl b/openssl-1.1.0h/crypto/aes/asm/aesni-mb-x86_64.pl new file mode 100644 index 0000000..aa2735e --- /dev/null +++ b/openssl-1.1.0h/crypto/aes/asm/aesni-mb-x86_64.pl @@ -0,0 +1,1402 @@ +#! /usr/bin/env perl +# Copyright 2013-2016 The OpenSSL Project Authors. All Rights Reserved. +# +# Licensed under the OpenSSL license (the "License"). You may not use +# this file except in compliance with the License. You can obtain a copy +# in the file LICENSE in the source distribution or at +# https://www.openssl.org/source/license.html + + +# ==================================================================== +# Written by Andy Polyakov for the OpenSSL +# project. The module is, however, dual licensed under OpenSSL and +# CRYPTOGAMS licenses depending on where you obtain it. For further +# details see http://www.openssl.org/~appro/cryptogams/. +# ==================================================================== + +# Multi-buffer AES-NI procedures process several independent buffers +# in parallel by interleaving independent instructions. +# +# Cycles per byte for interleave factor 4: +# +# asymptotic measured +# --------------------------- +# Westmere 5.00/4=1.25 5.13/4=1.28 +# Atom 15.0/4=3.75 ?15.7/4=3.93 +# Sandy Bridge 5.06/4=1.27 5.18/4=1.29 +# Ivy Bridge 5.06/4=1.27 5.14/4=1.29 +# Haswell 4.44/4=1.11 4.44/4=1.11 +# Bulldozer 5.75/4=1.44 5.76/4=1.44 +# +# Cycles per byte for interleave factor 8 (not implemented for +# pre-AVX processors, where higher interleave factor incidentally +# doesn't result in improvement): +# +# asymptotic measured +# --------------------------- +# Sandy Bridge 5.06/8=0.64 7.10/8=0.89(*) +# Ivy Bridge 5.06/8=0.64 7.14/8=0.89(*) +# Haswell 5.00/8=0.63 5.00/8=0.63 +# Bulldozer 5.75/8=0.72 5.77/8=0.72 +# +# (*) Sandy/Ivy Bridge are known to handle high interleave factors +# suboptimally; + +$flavour = shift; +$output = shift; +if ($flavour =~ /\./) { $output = $flavour; undef $flavour; } + +$win64=0; $win64=1 if ($flavour =~ /[nm]asm|mingw64/ || $output =~ /\.asm$/); + +$0 =~ m/(.*[\/\\])[^\/\\]+$/; $dir=$1; +( $xlate="${dir}x86_64-xlate.pl" and -f $xlate ) or +( $xlate="${dir}../../perlasm/x86_64-xlate.pl" and -f $xlate) or +die "can't locate x86_64-xlate.pl"; + +$avx=0; + +if (`$ENV{CC} -Wa,-v -c -o /dev/null -x assembler /dev/null 2>&1` + =~ /GNU assembler version ([2-9]\.[0-9]+)/) { + $avx = ($1>=2.19) + ($1>=2.22); +} + +if (!$avx && $win64 && ($flavour =~ /nasm/ || $ENV{ASM} =~ /nasm/) && + `nasm -v 2>&1` =~ /NASM version ([2-9]\.[0-9]+)/) { + $avx = ($1>=2.09) + ($1>=2.10); +} + +if (!$avx && $win64 && ($flavour =~ /masm/ || $ENV{ASM} =~ /ml64/) && + `ml64 2>&1` =~ /Version ([0-9]+)\./) { + $avx = ($1>=10) + ($1>=11); +} + +if (!$avx && `$ENV{CC} -v 2>&1` =~ /((?:^clang|LLVM) version|.*based on LLVM) ([3-9]\.[0-9]+)/) { + $avx = ($2>=3.0) + ($2>3.0); +} + +open OUT,"| \"$^X\" \"$xlate\" $flavour \"$output\""; +*STDOUT=*OUT; + +# void aesni_multi_cbc_encrypt ( +# struct { void *inp,*out; int blocks; double iv[2]; } inp[8]; +# const AES_KEY *key, +# int num); /* 1 or 2 */ +# +$inp="%rdi"; # 1st arg +$key="%rsi"; # 2nd arg +$num="%edx"; + +@inptr=map("%r$_",(8..11)); +@outptr=map("%r$_",(12..15)); + +($rndkey0,$rndkey1)=("%xmm0","%xmm1"); +@out=map("%xmm$_",(2..5)); +@inp=map("%xmm$_",(6..9)); +($counters,$mask,$zero)=map("%xmm$_",(10..12)); + +($rounds,$one,$sink,$offset)=("%eax","%ecx","%rbp","%rbx"); + +$code.=<<___; +.text + +.extern OPENSSL_ia32cap_P + +.globl aesni_multi_cbc_encrypt +.type aesni_multi_cbc_encrypt,\@function,3 +.align 32 +aesni_multi_cbc_encrypt: +___ +$code.=<<___ if ($avx); + cmp \$2,$num + jb .Lenc_non_avx + mov OPENSSL_ia32cap_P+4(%rip),%ecx + test \$`1<<28`,%ecx # AVX bit + jnz _avx_cbc_enc_shortcut + jmp .Lenc_non_avx +.align 16 +.Lenc_non_avx: +___ +$code.=<<___; + mov %rsp,%rax + push %rbx + push %rbp + push %r12 + push %r13 + push %r14 + push %r15 +___ +$code.=<<___ if ($win64); + lea -0xa8(%rsp),%rsp + movaps %xmm6,(%rsp) + movaps %xmm7,0x10(%rsp) + movaps %xmm8,0x20(%rsp) + movaps %xmm9,0x30(%rsp) + movaps %xmm10,0x40(%rsp) + movaps %xmm11,0x50(%rsp) + movaps %xmm12,0x60(%rsp) + movaps %xmm13,-0x68(%rax) # not used, saved to share se_handler + movaps %xmm14,-0x58(%rax) + movaps %xmm15,-0x48(%rax) +___ +$code.=<<___; + # stack layout + # + # +0 output sink + # +16 input sink [original %rsp and $num] + # +32 counters + + sub \$48,%rsp + and \$-64,%rsp + mov %rax,16(%rsp) # original %rsp + +.Lenc4x_body: + movdqu ($key),$zero # 0-round key + lea 0x78($key),$key # size optimization + lea 40*2($inp),$inp + +.Lenc4x_loop_grande: + mov $num,24(%rsp) # original $num + xor $num,$num +___ +for($i=0;$i<4;$i++) { + $code.=<<___; + mov `40*$i+16-40*2`($inp),$one # borrow $one for number of blocks + mov `40*$i+0-40*2`($inp),@inptr[$i] + cmp $num,$one + mov `40*$i+8-40*2`($inp),@outptr[$i] + cmovg $one,$num # find maximum + test $one,$one + movdqu `40*$i+24-40*2`($inp),@out[$i] # load IV + mov $one,`32+4*$i`(%rsp) # initialize counters + cmovle %rsp,@inptr[$i] # cancel input +___ +} +$code.=<<___; + test $num,$num + jz .Lenc4x_done + + movups 0x10-0x78($key),$rndkey1 + pxor $zero,@out[0] + movups 0x20-0x78($key),$rndkey0 + pxor $zero,@out[1] + mov 0xf0-0x78($key),$rounds + pxor $zero,@out[2] + movdqu (@inptr[0]),@inp[0] # load inputs + pxor $zero,@out[3] + movdqu (@inptr[1]),@inp[1] + pxor @inp[0],@out[0] + movdqu (@inptr[2]),@inp[2] + pxor @inp[1],@out[1] + movdqu (@inptr[3]),@inp[3] + pxor @inp[2],@out[2] + pxor @inp[3],@out[3] + movdqa 32(%rsp),$counters # load counters + xor $offset,$offset + jmp .Loop_enc4x + +.align 32 +.Loop_enc4x: + add \$16,$offset + lea 16(%rsp),$sink # sink pointer + mov \$1,$one # constant of 1 + sub $offset,$sink + + aesenc $rndkey1,@out[0] + prefetcht0 31(@inptr[0],$offset) # prefetch input + prefetcht0 31(@inptr[1],$offset) + aesenc $rndkey1,@out[1] + prefetcht0 31(@inptr[2],$offset) + prefetcht0 31(@inptr[2],$offset) + aesenc $rndkey1,@out[2] + aesenc $rndkey1,@out[3] + movups 0x30-0x78($key),$rndkey1 +___ +for($i=0;$i<4;$i++) { +my $rndkey = ($i&1) ? $rndkey1 : $rndkey0; +$code.=<<___; + cmp `32+4*$i`(%rsp),$one + aesenc $rndkey,@out[0] + aesenc $rndkey,@out[1] + aesenc $rndkey,@out[2] + cmovge $sink,@inptr[$i] # cancel input + cmovg $sink,@outptr[$i] # sink output + aesenc $rndkey,@out[3] + movups `0x40+16*$i-0x78`($key),$rndkey +___ +} +$code.=<<___; + movdqa $counters,$mask + aesenc $rndkey0,@out[0] + prefetcht0 15(@outptr[0],$offset) # prefetch output + prefetcht0 15(@outptr[1],$offset) + aesenc $rndkey0,@out[1] + prefetcht0 15(@outptr[2],$offset) + prefetcht0 15(@outptr[3],$offset) + aesenc $rndkey0,@out[2] + aesenc $rndkey0,@out[3] + movups 0x80-0x78($key),$rndkey0 + pxor $zero,$zero + + aesenc $rndkey1,@out[0] + pcmpgtd $zero,$mask + movdqu -0x78($key),$zero # reload 0-round key + aesenc $rndkey1,@out[1] + paddd $mask,$counters # decrement counters + movdqa $counters,32(%rsp) # update counters + aesenc $rndkey1,@out[2] + aesenc $rndkey1,@out[3] + movups 0x90-0x78($key),$rndkey1 + + cmp \$11,$rounds + + aesenc $rndkey0,@out[0] + aesenc $rndkey0,@out[1] + aesenc $rndkey0,@out[2] + aesenc $rndkey0,@out[3] + movups 0xa0-0x78($key),$rndkey0 + + jb .Lenc4x_tail + + aesenc $rndkey1,@out[0] + aesenc $rndkey1,@out[1] + aesenc $rndkey1,@out[2] + aesenc $rndkey1,@out[3] + movups 0xb0-0x78($key),$rndkey1 + + aesenc $rndkey0,@out[0] + aesenc $rndkey0,@out[1] + aesenc $rndkey0,@out[2] + aesenc $rndkey0,@out[3] + movups 0xc0-0x78($key),$rndkey0 + + je .Lenc4x_tail + + aesenc $rndkey1,@out[0] + aesenc $rndkey1,@out[1] + aesenc $rndkey1,@out[2] + aesenc $rndkey1,@out[3] + movups 0xd0-0x78($key),$rndkey1 + + aesenc $rndkey0,@out[0] + aesenc $rndkey0,@out[1] + aesenc $rndkey0,@out[2] + aesenc $rndkey0,@out[3] + movups 0xe0-0x78($key),$rndkey0 + jmp .Lenc4x_tail + +.align 32 +.Lenc4x_tail: + aesenc $rndkey1,@out[0] + aesenc $rndkey1,@out[1] + aesenc $rndkey1,@out[2] + aesenc $rndkey1,@out[3] + movdqu (@inptr[0],$offset),@inp[0] + movdqu 0x10-0x78($key),$rndkey1 + + aesenclast $rndkey0,@out[0] + movdqu (@inptr[1],$offset),@inp[1] + pxor $zero,@inp[0] + aesenclast $rndkey0,@out[1] + movdqu (@inptr[2],$offset),@inp[2] + pxor $zero,@inp[1] + aesenclast $rndkey0,@out[2] + movdqu (@inptr[3],$offset),@inp[3] + pxor $zero,@inp[2] + aesenclast $rndkey0,@out[3] + movdqu 0x20-0x78($key),$rndkey0 + pxor $zero,@inp[3] + + movups @out[0],-16(@outptr[0],$offset) + pxor @inp[0],@out[0] + movups @out[1],-16(@outptr[1],$offset) + pxor @inp[1],@out[1] + movups @out[2],-16(@outptr[2],$offset) + pxor @inp[2],@out[2] + movups @out[3],-16(@outptr[3],$offset) + pxor @inp[3],@out[3] + + dec $num + jnz .Loop_enc4x + + mov 16(%rsp),%rax # original %rsp + mov 24(%rsp),$num + + #pxor @inp[0],@out[0] + #pxor @inp[1],@out[1] + #movdqu @out[0],`40*0+24-40*2`($inp) # output iv FIX ME! + #pxor @inp[2],@out[2] + #movdqu @out[1],`40*1+24-40*2`($inp) + #pxor @inp[3],@out[3] + #movdqu @out[2],`40*2+24-40*2`($inp) # won't fix, let caller + #movdqu @out[3],`40*3+24-40*2`($inp) # figure this out... + + lea `40*4`($inp),$inp + dec $num + jnz .Lenc4x_loop_grande + +.Lenc4x_done: +___ +$code.=<<___ if ($win64); + movaps -0xd8(%rax),%xmm6 + movaps -0xc8(%rax),%xmm7 + movaps -0xb8(%rax),%xmm8 + movaps -0xa8(%rax),%xmm9 + movaps -0x98(%rax),%xmm10 + movaps -0x88(%rax),%xmm11 + movaps -0x78(%rax),%xmm12 + #movaps -0x68(%rax),%xmm13 + #movaps -0x58(%rax),%xmm14 + #movaps -0x48(%rax),%xmm15 +___ +$code.=<<___; + mov -48(%rax),%r15 + mov -40(%rax),%r14 + mov -32(%rax),%r13 + mov -24(%rax),%r12 + mov -16(%rax),%rbp + mov -8(%rax),%rbx + lea (%rax),%rsp +.Lenc4x_epilogue: + ret +.size aesni_multi_cbc_encrypt,.-aesni_multi_cbc_encrypt + +.globl aesni_multi_cbc_decrypt +.type aesni_multi_cbc_decrypt,\@function,3 +.align 32 +aesni_multi_cbc_decrypt: +___ +$code.=<<___ if ($avx); + cmp \$2,$num + jb .Ldec_non_avx + mov OPENSSL_ia32cap_P+4(%rip),%ecx + test \$`1<<28`,%ecx # AVX bit + jnz _avx_cbc_dec_shortcut + jmp .Ldec_non_avx +.align 16 +.Ldec_non_avx: +___ +$code.=<<___; + mov %rsp,%rax + push %rbx + push %rbp + push %r12 + push %r13 + push %r14 + push %r15 +___ +$code.=<<___ if ($win64); + lea -0xa8(%rsp),%rsp + movaps %xmm6,(%rsp) + movaps %xmm7,0x10(%rsp) + movaps %xmm8,0x20(%rsp) + movaps %xmm9,0x30(%rsp) + movaps %xmm10,0x40(%rsp) + movaps %xmm11,0x50(%rsp) + movaps %xmm12,0x60(%rsp) + movaps %xmm13,-0x68(%rax) # not used, saved to share se_handler + movaps %xmm14,-0x58(%rax) + movaps %xmm15,-0x48(%rax) +___ +$code.=<<___; + # stack layout + # + # +0 output sink + # +16 input sink [original %rsp and $num] + # +32 counters + + sub \$48,%rsp + and \$-64,%rsp + mov %rax,16(%rsp) # original %rsp + +.Ldec4x_body: + movdqu ($key),$zero # 0-round key + lea 0x78($key),$key # size optimization + lea 40*2($inp),$inp + +.Ldec4x_loop_grande: + mov $num,24(%rsp) # original $num + xor $num,$num +___ +for($i=0;$i<4;$i++) { + $code.=<<___; + mov `40*$i+16-40*2`($inp),$one # borrow $one for number of blocks + mov `40*$i+0-40*2`($inp),@inptr[$i] + cmp $num,$one + mov `40*$i+8-40*2`($inp),@outptr[$i] + cmovg $one,$num # find maximum + test $one,$one + movdqu `40*$i+24-40*2`($inp),@inp[$i] # load IV + mov $one,`32+4*$i`(%rsp) # initialize counters + cmovle %rsp,@inptr[$i] # cancel input +___ +} +$code.=<<___; + test $num,$num + jz .Ldec4x_done + + movups 0x10-0x78($key),$rndkey1 + movups 0x20-0x78($key),$rndkey0 + mov 0xf0-0x78($key),$rounds + movdqu (@inptr[0]),@out[0] # load inputs + movdqu (@inptr[1]),@out[1] + pxor $zero,@out[0] + movdqu (@inptr[2]),@out[2] + pxor $zero,@out[1] + movdqu (@inptr[3]),@out[3] + pxor $zero,@out[2] + pxor $zero,@out[3] + movdqa 32(%rsp),$counters # load counters + xor $offset,$offset + jmp .Loop_dec4x + +.align 32 +.Loop_dec4x: + add \$16,$offset + lea 16(%rsp),$sink # sink pointer + mov \$1,$one # constant of 1 + sub $offset,$sink + + aesdec $rndkey1,@out[0] + prefetcht0 31(@inptr[0],$offset) # prefetch input + prefetcht0 31(@inptr[1],$offset) + aesdec $rndkey1,@out[1] + prefetcht0 31(@inptr[2],$offset) + prefetcht0 31(@inptr[3],$offset) + aesdec $rndkey1,@out[2] + aesdec $rndkey1,@out[3] + movups 0x30-0x78($key),$rndkey1 +___ +for($i=0;$i<4;$i++) { +my $rndkey = ($i&1) ? $rndkey1 : $rndkey0; +$code.=<<___; + cmp `32+4*$i`(%rsp),$one + aesdec $rndkey,@out[0] + aesdec $rndkey,@out[1] + aesdec $rndkey,@out[2] + cmovge $sink,@inptr[$i] # cancel input + cmovg $sink,@outptr[$i] # sink output + aesdec $rndkey,@out[3] + movups `0x40+16*$i-0x78`($key),$rndkey +___ +} +$code.=<<___; + movdqa $counters,$mask + aesdec $rndkey0,@out[0] + prefetcht0 15(@outptr[0],$offset) # prefetch output + prefetcht0 15(@outptr[1],$offset) + aesdec $rndkey0,@out[1] + prefetcht0 15(@outptr[2],$offset) + prefetcht0 15(@outptr[3],$offset) + aesdec $rndkey0,@out[2] + aesdec $rndkey0,@out[3] + movups 0x80-0x78($key),$rndkey0 + pxor $zero,$zero + + aesdec $rndkey1,@out[0] + pcmpgtd $zero,$mask + movdqu -0x78($key),$zero # reload 0-round key + aesdec $rndkey1,@out[1] + paddd $mask,$counters # decrement counters + movdqa $counters,32(%rsp) # update counters + aesdec $rndkey1,@out[2] + aesdec $rndkey1,@out[3] + movups 0x90-0x78($key),$rndkey1 + + cmp \$11,$rounds + + aesdec $rndkey0,@out[0] + aesdec $rndkey0,@out[1] + aesdec $rndkey0,@out[2] + aesdec $rndkey0,@out[3] + movups 0xa0-0x78($key),$rndkey0 + + jb .Ldec4x_tail + + aesdec $rndkey1,@out[0] + aesdec $rndkey1,@out[1] + aesdec $rndkey1,@out[2] + aesdec $rndkey1,@out[3] + movups 0xb0-0x78($key),$rndkey1 + + aesdec $rndkey0,@out[0] + aesdec $rndkey0,@out[1] + aesdec $rndkey0,@out[2] + aesdec $rndkey0,@out[3] + movups 0xc0-0x78($key),$rndkey0 + + je .Ldec4x_tail + + aesdec $rndkey1,@out[0] + aesdec $rndkey1,@out[1] + aesdec $rndkey1,@out[2] + aesdec $rndkey1,@out[3] + movups 0xd0-0x78($key),$rndkey1 + + aesdec $rndkey0,@out[0] + aesdec $rndkey0,@out[1] + aesdec $rndkey0,@out[2] + aesdec $rndkey0,@out[3] + movups 0xe0-0x78($key),$rndkey0 + jmp .Ldec4x_tail + +.align 32 +.Ldec4x_tail: + aesdec $rndkey1,@out[0] + aesdec $rndkey1,@out[1] + aesdec $rndkey1,@out[2] + pxor $rndkey0,@inp[0] + pxor $rndkey0,@inp[1] + aesdec $rndkey1,@out[3] + movdqu 0x10-0x78($key),$rndkey1 + pxor $rndkey0,@inp[2] + pxor $rndkey0,@inp[3] + movdqu 0x20-0x78($key),$rndkey0 + + aesdeclast @inp[0],@out[0] + aesdeclast @inp[1],@out[1] + movdqu -16(@inptr[0],$offset),@inp[0] # load next IV + movdqu -16(@inptr[1],$offset),@inp[1] + aesdeclast @inp[2],@out[2] + aesdeclast @inp[3],@out[3] + movdqu -16(@inptr[2],$offset),@inp[2] + movdqu -16(@inptr[3],$offset),@inp[3] + + movups @out[0],-16(@outptr[0],$offset) + movdqu (@inptr[0],$offset),@out[0] + movups @out[1],-16(@outptr[1],$offset) + movdqu (@inptr[1],$offset),@out[1] + pxor $zero,@out[0] + movups @out[2],-16(@outptr[2],$offset) + movdqu (@inptr[2],$offset),@out[2] + pxor $zero,@out[1] + movups @out[3],-16(@outptr[3],$offset) + movdqu (@inptr[3],$offset),@out[3] + pxor $zero,@out[2] + pxor $zero,@out[3] + + dec $num + jnz .Loop_dec4x + + mov 16(%rsp),%rax # original %rsp + mov 24(%rsp),$num + + lea `40*4`($inp),$inp + dec $num + jnz .Ldec4x_loop_grande + +.Ldec4x_done: +___ +$code.=<<___ if ($win64); + movaps -0xd8(%rax),%xmm6 + movaps -0xc8(%rax),%xmm7 + movaps -0xb8(%rax),%xmm8 + movaps -0xa8(%rax),%xmm9 + movaps -0x98(%rax),%xmm10 + movaps -0x88(%rax),%xmm11 + movaps -0x78(%rax),%xmm12 + #movaps -0x68(%rax),%xmm13 + #movaps -0x58(%rax),%xmm14 + #movaps -0x48(%rax),%xmm15 +___ +$code.=<<___; + mov -48(%rax),%r15 + mov -40(%rax),%r14 + mov -32(%rax),%r13 + mov -24(%rax),%r12 + mov -16(%rax),%rbp + mov -8(%rax),%rbx + lea (%rax),%rsp +.Ldec4x_epilogue: + ret +.size aesni_multi_cbc_decrypt,.-aesni_multi_cbc_decrypt +___ + + if ($avx) {{{ +my @ptr=map("%r$_",(8..15)); +my $offload=$sink; + +my @out=map("%xmm$_",(2..9)); +my @inp=map("%xmm$_",(10..13)); +my ($counters,$zero)=("%xmm14","%xmm15"); + +$code.=<<___; +.type aesni_multi_cbc_encrypt_avx,\@function,3 +.align 32 +aesni_multi_cbc_encrypt_avx: +_avx_cbc_enc_shortcut: + mov %rsp,%rax + push %rbx + push %rbp + push %r12 + push %r13 + push %r14 + push %r15 +___ +$code.=<<___ if ($win64); + lea -0xa8(%rsp),%rsp + movaps %xmm6,(%rsp) + movaps %xmm7,0x10(%rsp) + movaps %xmm8,0x20(%rsp) + movaps %xmm9,0x30(%rsp) + movaps %xmm10,0x40(%rsp) + movaps %xmm11,0x50(%rsp) + movaps %xmm12,-0x78(%rax) + movaps %xmm13,-0x68(%rax) + movaps %xmm14,-0x58(%rax) + movaps %xmm15,-0x48(%rax) +___ +$code.=<<___; + # stack layout + # + # +0 output sink + # +16 input sink [original %rsp and $num] + # +32 counters + # +64 distances between inputs and outputs + # +128 off-load area for @inp[0..3] + + sub \$192,%rsp + and \$-128,%rsp + mov %rax,16(%rsp) # original %rsp + +.Lenc8x_body: + vzeroupper + vmovdqu ($key),$zero # 0-round key + lea 0x78($key),$key # size optimization + lea 40*4($inp),$inp + shr \$1,$num + +.Lenc8x_loop_grande: + #mov $num,24(%rsp) # original $num + xor $num,$num +___ +for($i=0;$i<8;$i++) { + my $temp = $i ? $offload : $offset; + $code.=<<___; + mov `40*$i+16-40*4`($inp),$one # borrow $one for number of blocks + mov `40*$i+0-40*4`($inp),@ptr[$i] # input pointer + cmp $num,$one + mov `40*$i+8-40*4`($inp),$temp # output pointer + cmovg $one,$num # find maximum + test $one,$one + vmovdqu `40*$i+24-40*4`($inp),@out[$i] # load IV + mov $one,`32+4*$i`(%rsp) # initialize counters + cmovle %rsp,@ptr[$i] # cancel input + sub @ptr[$i],$temp # distance between input and output + mov $temp,`64+8*$i`(%rsp) # initialize distances +___ +} +$code.=<<___; + test $num,$num + jz .Lenc8x_done + + vmovups 0x10-0x78($key),$rndkey1 + vmovups 0x20-0x78($key),$rndkey0 + mov 0xf0-0x78($key),$rounds + + vpxor (@ptr[0]),$zero,@inp[0] # load inputs and xor with 0-round + lea 128(%rsp),$offload # offload area + vpxor (@ptr[1]),$zero,@inp[1] + vpxor (@ptr[2]),$zero,@inp[2] + vpxor (@ptr[3]),$zero,@inp[3] + vpxor @inp[0],@out[0],@out[0] + vpxor (@ptr[4]),$zero,@inp[0] + vpxor @inp[1],@out[1],@out[1] + vpxor (@ptr[5]),$zero,@inp[1] + vpxor @inp[2],@out[2],@out[2] + vpxor (@ptr[6]),$zero,@inp[2] + vpxor @inp[3],@out[3],@out[3] + vpxor (@ptr[7]),$zero,@inp[3] + vpxor @inp[0],@out[4],@out[4] + mov \$1,$one # constant of 1 + vpxor @inp[1],@out[5],@out[5] + vpxor @inp[2],@out[6],@out[6] + vpxor @inp[3],@out[7],@out[7] + jmp .Loop_enc8x + +.align 32 +.Loop_enc8x: +___ +for($i=0;$i<8;$i++) { +my $rndkey=($i&1)?$rndkey0:$rndkey1; +$code.=<<___; + vaesenc $rndkey,@out[0],@out[0] + cmp 32+4*$i(%rsp),$one +___ +$code.=<<___ if ($i); + mov 64+8*$i(%rsp),$offset +___ +$code.=<<___; + vaesenc $rndkey,@out[1],@out[1] + prefetcht0 31(@ptr[$i]) # prefetch input + vaesenc $rndkey,@out[2],@out[2] +___ +$code.=<<___ if ($i>1); + prefetcht0 15(@ptr[$i-2]) # prefetch output +___ +$code.=<<___; + vaesenc $rndkey,@out[3],@out[3] + lea (@ptr[$i],$offset),$offset + cmovge %rsp,@ptr[$i] # cancel input + vaesenc $rndkey,@out[4],@out[4] + cmovg %rsp,$offset # sink output + vaesenc $rndkey,@out[5],@out[5] + sub @ptr[$i],$offset + vaesenc $rndkey,@out[6],@out[6] + vpxor 16(@ptr[$i]),$zero,@inp[$i%4] # load input and xor with 0-round + mov $offset,64+8*$i(%rsp) + vaesenc $rndkey,@out[7],@out[7] + vmovups `16*(3+$i)-0x78`($key),$rndkey + lea 16(@ptr[$i],$offset),@ptr[$i] # switch to output +___ +$code.=<<___ if ($i<4) + vmovdqu @inp[$i%4],`16*$i`($offload) # off-load +___ +} +$code.=<<___; + vmovdqu 32(%rsp),$counters + prefetcht0 15(@ptr[$i-2]) # prefetch output + prefetcht0 15(@ptr[$i-1]) + cmp \$11,$rounds + jb .Lenc8x_tail + + vaesenc $rndkey1,@out[0],@out[0] + vaesenc $rndkey1,@out[1],@out[1] + vaesenc $rndkey1,@out[2],@out[2] + vaesenc $rndkey1,@out[3],@out[3] + vaesenc $rndkey1,@out[4],@out[4] + vaesenc $rndkey1,@out[5],@out[5] + vaesenc $rndkey1,@out[6],@out[6] + vaesenc $rndkey1,@out[7],@out[7] + vmovups 0xb0-0x78($key),$rndkey1 + + vaesenc $rndkey0,@out[0],@out[0] + vaesenc $rndkey0,@out[1],@out[1] + vaesenc $rndkey0,@out[2],@out[2] + vaesenc $rndkey0,@out[3],@out[3] + vaesenc $rndkey0,@out[4],@out[4] + vaesenc $rndkey0,@out[5],@out[5] + vaesenc $rndkey0,@out[6],@out[6] + vaesenc $rndkey0,@out[7],@out[7] + vmovups 0xc0-0x78($key),$rndkey0 + je .Lenc8x_tail + + vaesenc $rndkey1,@out[0],@out[0] + vaesenc $rndkey1,@out[1],@out[1] + vaesenc $rndkey1,@out[2],@out[2] + vaesenc $rndkey1,@out[3],@out[3] + vaesenc $rndkey1,@out[4],@out[4] + vaesenc $rndkey1,@out[5],@out[5] + vaesenc $rndkey1,@out[6],@out[6] + vaesenc $rndkey1,@out[7],@out[7] + vmovups 0xd0-0x78($key),$rndkey1 + + vaesenc $rndkey0,@out[0],@out[0] + vaesenc $rndkey0,@out[1],@out[1] + vaesenc $rndkey0,@out[2],@out[2] + vaesenc $rndkey0,@out[3],@out[3] + vaesenc $rndkey0,@out[4],@out[4] + vaesenc $rndkey0,@out[5],@out[5] + vaesenc $rndkey0,@out[6],@out[6] + vaesenc $rndkey0,@out[7],@out[7] + vmovups 0xe0-0x78($key),$rndkey0 + +.Lenc8x_tail: + vaesenc $rndkey1,@out[0],@out[0] + vpxor $zero,$zero,$zero + vaesenc $rndkey1,@out[1],@out[1] + vaesenc $rndkey1,@out[2],@out[2] + vpcmpgtd $zero,$counters,$zero + vaesenc $rndkey1,@out[3],@out[3] + vaesenc $rndkey1,@out[4],@out[4] + vpaddd $counters,$zero,$zero # decrement counters + vmovdqu 48(%rsp),$counters + vaesenc $rndkey1,@out[5],@out[5] + mov 64(%rsp),$offset # pre-load 1st offset + vaesenc $rndkey1,@out[6],@out[6] + vaesenc $rndkey1,@out[7],@out[7] + vmovups 0x10-0x78($key),$rndkey1 + + vaesenclast $rndkey0,@out[0],@out[0] + vmovdqa $zero,32(%rsp) # update counters + vpxor $zero,$zero,$zero + vaesenclast $rndkey0,@out[1],@out[1] + vaesenclast $rndkey0,@out[2],@out[2] + vpcmpgtd $zero,$counters,$zero + vaesenclast $rndkey0,@out[3],@out[3] + vaesenclast $rndkey0,@out[4],@out[4] + vpaddd $zero,$counters,$counters # decrement counters + vmovdqu -0x78($key),$zero # 0-round + vaesenclast $rndkey0,@out[5],@out[5] + vaesenclast $rndkey0,@out[6],@out[6] + vmovdqa $counters,48(%rsp) # update counters + vaesenclast $rndkey0,@out[7],@out[7] + vmovups 0x20-0x78($key),$rndkey0 + + vmovups @out[0],-16(@ptr[0]) # write output + sub $offset,@ptr[0] # switch to input + vpxor 0x00($offload),@out[0],@out[0] + vmovups @out[1],-16(@ptr[1]) + sub `64+1*8`(%rsp),@ptr[1] + vpxor 0x10($offload),@out[1],@out[1] + vmovups @out[2],-16(@ptr[2]) + sub `64+2*8`(%rsp),@ptr[2] + vpxor 0x20($offload),@out[2],@out[2] + vmovups @out[3],-16(@ptr[3]) + sub `64+3*8`(%rsp),@ptr[3] + vpxor 0x30($offload),@out[3],@out[3] + vmovups @out[4],-16(@ptr[4]) + sub `64+4*8`(%rsp),@ptr[4] + vpxor @inp[0],@out[4],@out[4] + vmovups @out[5],-16(@ptr[5]) + sub `64+5*8`(%rsp),@ptr[5] + vpxor @inp[1],@out[5],@out[5] + vmovups @out[6],-16(@ptr[6]) + sub `64+6*8`(%rsp),@ptr[6] + vpxor @inp[2],@out[6],@out[6] + vmovups @out[7],-16(@ptr[7]) + sub `64+7*8`(%rsp),@ptr[7] + vpxor @inp[3],@out[7],@out[7] + + dec $num + jnz .Loop_enc8x + + mov 16(%rsp),%rax # original %rsp + #mov 24(%rsp),$num + #lea `40*8`($inp),$inp + #dec $num + #jnz .Lenc8x_loop_grande + +.Lenc8x_done: + vzeroupper +___ +$code.=<<___ if ($win64); + movaps -0xd8(%rax),%xmm6 + movaps -0xc8(%rax),%xmm7 + movaps -0xb8(%rax),%xmm8 + movaps -0xa8(%rax),%xmm9 + movaps -0x98(%rax),%xmm10 + movaps -0x88(%rax),%xmm11 + movaps -0x78(%rax),%xmm12 + movaps -0x68(%rax),%xmm13 + movaps -0x58(%rax),%xmm14 + movaps -0x48(%rax),%xmm15 +___ +$code.=<<___; + mov -48(%rax),%r15 + mov -40(%rax),%r14 + mov -32(%rax),%r13 + mov -24(%rax),%r12 + mov -16(%rax),%rbp + mov -8(%rax),%rbx + lea (%rax),%rsp +.Lenc8x_epilogue: + ret +.size aesni_multi_cbc_encrypt_avx,.-aesni_multi_cbc_encrypt_avx + +.type aesni_multi_cbc_decrypt_avx,\@function,3 +.align 32 +aesni_multi_cbc_decrypt_avx: +_avx_cbc_dec_shortcut: + mov %rsp,%rax + push %rbx + push %rbp + push %r12 + push %r13 + push %r14 + push %r15 +___ +$code.=<<___ if ($win64); + lea -0xa8(%rsp),%rsp + movaps %xmm6,(%rsp) + movaps %xmm7,0x10(%rsp) + movaps %xmm8,0x20(%rsp) + movaps %xmm9,0x30(%rsp) + movaps %xmm10,0x40(%rsp) + movaps %xmm11,0x50(%rsp) + movaps %xmm12,-0x78(%rax) + movaps %xmm13,-0x68(%rax) + movaps %xmm14,-0x58(%rax) + movaps %xmm15,-0x48(%rax) +___ +$code.=<<___; + # stack layout + # + # +0 output sink + # +16 input sink [original %rsp and $num] + # +32 counters + # +64 distances between inputs and outputs + # +128 off-load area for @inp[0..3] + # +192 IV/input offload + + sub \$256,%rsp + and \$-256,%rsp + sub \$192,%rsp + mov %rax,16(%rsp) # original %rsp + +.Ldec8x_body: + vzeroupper + vmovdqu ($key),$zero # 0-round key + lea 0x78($key),$key # size optimization + lea 40*4($inp),$inp + shr \$1,$num + +.Ldec8x_loop_grande: + #mov $num,24(%rsp) # original $num + xor $num,$num +___ +for($i=0;$i<8;$i++) { + my $temp = $i ? $offload : $offset; + $code.=<<___; + mov `40*$i+16-40*4`($inp),$one # borrow $one for number of blocks + mov `40*$i+0-40*4`($inp),@ptr[$i] # input pointer + cmp $num,$one + mov `40*$i+8-40*4`($inp),$temp # output pointer + cmovg $one,$num # find maximum + test $one,$one + vmovdqu `40*$i+24-40*4`($inp),@out[$i] # load IV + mov $one,`32+4*$i`(%rsp) # initialize counters + cmovle %rsp,@ptr[$i] # cancel input + sub @ptr[$i],$temp # distance between input and output + mov $temp,`64+8*$i`(%rsp) # initialize distances + vmovdqu @out[$i],`192+16*$i`(%rsp) # offload IV +___ +} +$code.=<<___; + test $num,$num + jz .Ldec8x_done + + vmovups 0x10-0x78($key),$rndkey1 + vmovups 0x20-0x78($key),$rndkey0 + mov 0xf0-0x78($key),$rounds + lea 192+128(%rsp),$offload # offload area + + vmovdqu (@ptr[0]),@out[0] # load inputs + vmovdqu (@ptr[1]),@out[1] + vmovdqu (@ptr[2]),@out[2] + vmovdqu (@ptr[3]),@out[3] + vmovdqu (@ptr[4]),@out[4] + vmovdqu (@ptr[5]),@out[5] + vmovdqu (@ptr[6]),@out[6] + vmovdqu (@ptr[7]),@out[7] + vmovdqu @out[0],0x00($offload) # offload inputs + vpxor $zero,@out[0],@out[0] # xor inputs with 0-round + vmovdqu @out[1],0x10($offload) + vpxor $zero,@out[1],@out[1] + vmovdqu @out[2],0x20($offload) + vpxor $zero,@out[2],@out[2] + vmovdqu @out[3],0x30($offload) + vpxor $zero,@out[3],@out[3] + vmovdqu @out[4],0x40($offload) + vpxor $zero,@out[4],@out[4] + vmovdqu @out[5],0x50($offload) + vpxor $zero,@out[5],@out[5] + vmovdqu @out[6],0x60($offload) + vpxor $zero,@out[6],@out[6] + vmovdqu @out[7],0x70($offload) + vpxor $zero,@out[7],@out[7] + xor \$0x80,$offload + mov \$1,$one # constant of 1 + jmp .Loop_dec8x + +.align 32 +.Loop_dec8x: +___ +for($i=0;$i<8;$i++) { +my $rndkey=($i&1)?$rndkey0:$rndkey1; +$code.=<<___; + vaesdec $rndkey,@out[0],@out[0] + cmp 32+4*$i(%rsp),$one +___ +$code.=<<___ if ($i); + mov 64+8*$i(%rsp),$offset +___ +$code.=<<___; + vaesdec $rndkey,@out[1],@out[1] + prefetcht0 31(@ptr[$i]) # prefetch input + vaesdec $rndkey,@out[2],@out[2] +___ +$code.=<<___ if ($i>1); + prefetcht0 15(@ptr[$i-2]) # prefetch output +___ +$code.=<<___; + vaesdec $rndkey,@out[3],@out[3] + lea (@ptr[$i],$offset),$offset + cmovge %rsp,@ptr[$i] # cancel input + vaesdec $rndkey,@out[4],@out[4] + cmovg %rsp,$offset # sink output + vaesdec $rndkey,@out[5],@out[5] + sub @ptr[$i],$offset + vaesdec $rndkey,@out[6],@out[6] + vmovdqu 16(@ptr[$i]),@inp[$i%4] # load input + mov $offset,64+8*$i(%rsp) + vaesdec $rndkey,@out[7],@out[7] + vmovups `16*(3+$i)-0x78`($key),$rndkey + lea 16(@ptr[$i],$offset),@ptr[$i] # switch to output +___ +$code.=<<___ if ($i<4); + vmovdqu @inp[$i%4],`128+16*$i`(%rsp) # off-load +___ +} +$code.=<<___; + vmovdqu 32(%rsp),$counters + prefetcht0 15(@ptr[$i-2]) # prefetch output + prefetcht0 15(@ptr[$i-1]) + cmp \$11,$rounds + jb .Ldec8x_tail + + vaesdec $rndkey1,@out[0],@out[0] + vaesdec $rndkey1,@out[1],@out[1] + vaesdec $rndkey1,@out[2],@out[2] + vaesdec $rndkey1,@out[3],@out[3] + vaesdec $rndkey1,@out[4],@out[4] + vaesdec $rndkey1,@out[5],@out[5] + vaesdec $rndkey1,@out[6],@out[6] + vaesdec $rndkey1,@out[7],@out[7] + vmovups 0xb0-0x78($key),$rndkey1 + + vaesdec $rndkey0,@out[0],@out[0] + vaesdec $rndkey0,@out[1],@out[1] + vaesdec $rndkey0,@out[2],@out[2] + vaesdec $rndkey0,@out[3],@out[3] + vaesdec $rndkey0,@out[4],@out[4] + vaesdec $rndkey0,@out[5],@out[5] + vaesdec $rndkey0,@out[6],@out[6] + vaesdec $rndkey0,@out[7],@out[7] + vmovups 0xc0-0x78($key),$rndkey0 + je .Ldec8x_tail + + vaesdec $rndkey1,@out[0],@out[0] + vaesdec $rndkey1,@out[1],@out[1] + vaesdec $rndkey1,@out[2],@out[2] + vaesdec $rndkey1,@out[3],@out[3] + vaesdec $rndkey1,@out[4],@out[4] + vaesdec $rndkey1,@out[5],@out[5] + vaesdec $rndkey1,@out[6],@out[6] + vaesdec $rndkey1,@out[7],@out[7] + vmovups 0xd0-0x78($key),$rndkey1 + + vaesdec $rndkey0,@out[0],@out[0] + vaesdec $rndkey0,@out[1],@out[1] + vaesdec $rndkey0,@out[2],@out[2] + vaesdec $rndkey0,@out[3],@out[3] + vaesdec $rndkey0,@out[4],@out[4] + vaesdec $rndkey0,@out[5],@out[5] + vaesdec $rndkey0,@out[6],@out[6] + vaesdec $rndkey0,@out[7],@out[7] + vmovups 0xe0-0x78($key),$rndkey0 + +.Ldec8x_tail: + vaesdec $rndkey1,@out[0],@out[0] + vpxor $zero,$zero,$zero + vaesdec $rndkey1,@out[1],@out[1] + vaesdec $rndkey1,@out[2],@out[2] + vpcmpgtd $zero,$counters,$zero + vaesdec $rndkey1,@out[3],@out[3] + vaesdec $rndkey1,@out[4],@out[4] + vpaddd $counters,$zero,$zero # decrement counters + vmovdqu 48(%rsp),$counters + vaesdec $rndkey1,@out[5],@out[5] + mov 64(%rsp),$offset # pre-load 1st offset + vaesdec $rndkey1,@out[6],@out[6] + vaesdec $rndkey1,@out[7],@out[7] + vmovups 0x10-0x78($key),$rndkey1 + + vaesdeclast $rndkey0,@out[0],@out[0] + vmovdqa $zero,32(%rsp) # update counters + vpxor $zero,$zero,$zero + vaesdeclast $rndkey0,@out[1],@out[1] + vpxor 0x00($offload),@out[0],@out[0] # xor with IV + vaesdeclast $rndkey0,@out[2],@out[2] + vpxor 0x10($offload),@out[1],@out[1] + vpcmpgtd $zero,$counters,$zero + vaesdeclast $rndkey0,@out[3],@out[3] + vpxor 0x20($offload),@out[2],@out[2] + vaesdeclast $rndkey0,@out[4],@out[4] + vpxor 0x30($offload),@out[3],@out[3] + vpaddd $zero,$counters,$counters # decrement counters + vmovdqu -0x78($key),$zero # 0-round + vaesdeclast $rndkey0,@out[5],@out[5] + vpxor 0x40($offload),@out[4],@out[4] + vaesdeclast $rndkey0,@out[6],@out[6] + vpxor 0x50($offload),@out[5],@out[5] + vmovdqa $counters,48(%rsp) # update counters + vaesdeclast $rndkey0,@out[7],@out[7] + vpxor 0x60($offload),@out[6],@out[6] + vmovups 0x20-0x78($key),$rndkey0 + + vmovups @out[0],-16(@ptr[0]) # write output + sub $offset,@ptr[0] # switch to input + vmovdqu 128+0(%rsp),@out[0] + vpxor 0x70($offload),@out[7],@out[7] + vmovups @out[1],-16(@ptr[1]) + sub `64+1*8`(%rsp),@ptr[1] + vmovdqu @out[0],0x00($offload) + vpxor $zero,@out[0],@out[0] + vmovdqu 128+16(%rsp),@out[1] + vmovups @out[2],-16(@ptr[2]) + sub `64+2*8`(%rsp),@ptr[2] + vmovdqu @out[1],0x10($offload) + vpxor $zero,@out[1],@out[1] + vmovdqu 128+32(%rsp),@out[2] + vmovups @out[3],-16(@ptr[3]) + sub `64+3*8`(%rsp),@ptr[3] + vmovdqu @out[2],0x20($offload) + vpxor $zero,@out[2],@out[2] + vmovdqu 128+48(%rsp),@out[3] + vmovups @out[4],-16(@ptr[4]) + sub `64+4*8`(%rsp),@ptr[4] + vmovdqu @out[3],0x30($offload) + vpxor $zero,@out[3],@out[3] + vmovdqu @inp[0],0x40($offload) + vpxor @inp[0],$zero,@out[4] + vmovups @out[5],-16(@ptr[5]) + sub `64+5*8`(%rsp),@ptr[5] + vmovdqu @inp[1],0x50($offload) + vpxor @inp[1],$zero,@out[5] + vmovups @out[6],-16(@ptr[6]) + sub `64+6*8`(%rsp),@ptr[6] + vmovdqu @inp[2],0x60($offload) + vpxor @inp[2],$zero,@out[6] + vmovups @out[7],-16(@ptr[7]) + sub `64+7*8`(%rsp),@ptr[7] + vmovdqu @inp[3],0x70($offload) + vpxor @inp[3],$zero,@out[7] + + xor \$128,$offload + dec $num + jnz .Loop_dec8x + + mov 16(%rsp),%rax # original %rsp + #mov 24(%rsp),$num + #lea `40*8`($inp),$inp + #dec $num + #jnz .Ldec8x_loop_grande + +.Ldec8x_done: + vzeroupper +___ +$code.=<<___ if ($win64); + movaps -0xd8(%rax),%xmm6 + movaps -0xc8(%rax),%xmm7 + movaps -0xb8(%rax),%xmm8 + movaps -0xa8(%rax),%xmm9 + movaps -0x98(%rax),%xmm10 + movaps -0x88(%rax),%xmm11 + movaps -0x78(%rax),%xmm12 + movaps -0x68(%rax),%xmm13 + movaps -0x58(%rax),%xmm14 + movaps -0x48(%rax),%xmm15 +___ +$code.=<<___; + mov -48(%rax),%r15 + mov -40(%rax),%r14 + mov -32(%rax),%r13 + mov -24(%rax),%r12 + mov -16(%rax),%rbp + mov -8(%rax),%rbx + lea (%rax),%rsp +.Ldec8x_epilogue: + ret +.size aesni_multi_cbc_decrypt_avx,.-aesni_multi_cbc_decrypt_avx +___ + }}} + +if ($win64) { +# EXCEPTION_DISPOSITION handler (EXCEPTION_RECORD *rec,ULONG64 frame, +# CONTEXT *context,DISPATCHER_CONTEXT *disp) +$rec="%rcx"; +$frame="%rdx"; +$context="%r8"; +$disp="%r9"; + +$code.=<<___; +.extern __imp_RtlVirtualUnwind +.type se_handler,\@abi-omnipotent +.align 16 +se_handler: + push %rsi + push %rdi + push %rbx + push %rbp + push %r12 + push %r13 + push %r14 + push %r15 + pushfq + sub \$64,%rsp + + mov 120($context),%rax # pull context->Rax + mov 248($context),%rbx # pull context->Rip + + mov 8($disp),%rsi # disp->ImageBase + mov 56($disp),%r11 # disp->HandlerData + + mov 0(%r11),%r10d # HandlerData[0] + lea (%rsi,%r10),%r10 # prologue label + cmp %r10,%rbx # context->Rip<.Lprologue + jb .Lin_prologue + + mov 152($context),%rax # pull context->Rsp + + mov 4(%r11),%r10d # HandlerData[1] + lea (%rsi,%r10),%r10 # epilogue label + cmp %r10,%rbx # context->Rip>=.Lepilogue + jae .Lin_prologue + + mov 16(%rax),%rax # pull saved stack pointer + + mov -8(%rax),%rbx + mov -16(%rax),%rbp + mov -24(%rax),%r12 + mov -32(%rax),%r13 + mov -40(%rax),%r14 + mov -48(%rax),%r15 + mov %rbx,144($context) # restore context->Rbx + mov %rbp,160($context) # restore context->Rbp + mov %r12,216($context) # restore cotnext->R12 + mov %r13,224($context) # restore cotnext->R13 + mov %r14,232($context) # restore cotnext->R14 + mov %r15,240($context) # restore cotnext->R15 + + lea -56-10*16(%rax),%rsi + lea 512($context),%rdi # &context.Xmm6 + mov \$20,%ecx + .long 0xa548f3fc # cld; rep movsq + +.Lin_prologue: + mov 8(%rax),%rdi + mov 16(%rax),%rsi + mov %rax,152($context) # restore context->Rsp + mov %rsi,168($context) # restore context->Rsi + mov %rdi,176($context) # restore context->Rdi + + mov 40($disp),%rdi # disp->ContextRecord + mov $context,%rsi # context + mov \$154,%ecx # sizeof(CONTEXT) + .long 0xa548f3fc # cld; rep movsq + + mov $disp,%rsi + xor %rcx,%rcx # arg1, UNW_FLAG_NHANDLER + mov 8(%rsi),%rdx # arg2, disp->ImageBase + mov 0(%rsi),%r8 # arg3, disp->ControlPc + mov 16(%rsi),%r9 # arg4, disp->FunctionEntry + mov 40(%rsi),%r10 # disp->ContextRecord + lea 56(%rsi),%r11 # &disp->HandlerData + lea 24(%rsi),%r12 # &disp->EstablisherFrame + mov %r10,32(%rsp) # arg5 + mov %r11,40(%rsp) # arg6 + mov %r12,48(%rsp) # arg7 + mov %rcx,56(%rsp) # arg8, (NULL) + call *__imp_RtlVirtualUnwind(%rip) + + mov \$1,%eax # ExceptionContinueSearch + add \$64,%rsp + popfq + pop %r15 + pop %r14 + pop %r13 + pop %r12 + pop %rbp + pop %rbx + pop %rdi + pop %rsi + ret +.size se_handler,.-se_handler + +.section .pdata +.align 4 + .rva .LSEH_begin_aesni_multi_cbc_encrypt + .rva .LSEH_end_aesni_multi_cbc_encrypt + .rva .LSEH_info_aesni_multi_cbc_encrypt + .rva .LSEH_begin_aesni_multi_cbc_decrypt + .rva .LSEH_end_aesni_multi_cbc_decrypt + .rva .LSEH_info_aesni_multi_cbc_decrypt +___ +$code.=<<___ if ($avx); + .rva .LSEH_begin_aesni_multi_cbc_encrypt_avx + .rva .LSEH_end_aesni_multi_cbc_encrypt_avx + .rva .LSEH_info_aesni_multi_cbc_encrypt_avx + .rva .LSEH_begin_aesni_multi_cbc_decrypt_avx + .rva .LSEH_end_aesni_multi_cbc_decrypt_avx + .rva .LSEH_info_aesni_multi_cbc_decrypt_avx +___ +$code.=<<___; +.section .xdata +.align 8 +.LSEH_info_aesni_multi_cbc_encrypt: + .byte 9,0,0,0 + .rva se_handler + .rva .Lenc4x_body,.Lenc4x_epilogue # HandlerData[] +.LSEH_info_aesni_multi_cbc_decrypt: + .byte 9,0,0,0 + .rva se_handler + .rva .Ldec4x_body,.Ldec4x_epilogue # HandlerData[] +___ +$code.=<<___ if ($avx); +.LSEH_info_aesni_multi_cbc_encrypt_avx: + .byte 9,0,0,0 + .rva se_handler + .rva .Lenc8x_body,.Lenc8x_epilogue # HandlerData[] +.LSEH_info_aesni_multi_cbc_decrypt_avx: + .byte 9,0,0,0 + .rva se_handler + .rva .Ldec8x_body,.Ldec8x_epilogue # HandlerData[] +___ +} +#################################################################### + +sub rex { + local *opcode=shift; + my ($dst,$src)=@_; + my $rex=0; + + $rex|=0x04 if($dst>=8); + $rex|=0x01 if($src>=8); + push @opcode,$rex|0x40 if($rex); +} + +sub aesni { + my $line=shift; + my @opcode=(0x66); + + if ($line=~/(aeskeygenassist)\s+\$([x0-9a-f]+),\s*%xmm([0-9]+),\s*%xmm([0-9]+)/) { + rex(\@opcode,$4,$3); + push @opcode,0x0f,0x3a,0xdf; + push @opcode,0xc0|($3&7)|(($4&7)<<3); # ModR/M + my $c=$2; + push @opcode,$c=~/^0/?oct($c):$c; + return ".byte\t".join(',',@opcode); + } + elsif ($line=~/(aes[a-z]+)\s+%xmm([0-9]+),\s*%xmm([0-9]+)/) { + my %opcodelet = ( + "aesimc" => 0xdb, + "aesenc" => 0xdc, "aesenclast" => 0xdd, + "aesdec" => 0xde, "aesdeclast" => 0xdf + ); + return undef if (!defined($opcodelet{$1})); + rex(\@opcode,$3,$2); + push @opcode,0x0f,0x38,$opcodelet{$1}; + push @opcode,0xc0|($2&7)|(($3&7)<<3); # ModR/M + return ".byte\t".join(',',@opcode); + } + elsif ($line=~/(aes[a-z]+)\s+([0x1-9a-fA-F]*)\(%rsp\),\s*%xmm([0-9]+)/) { + my %opcodelet = ( + "aesenc" => 0xdc, "aesenclast" => 0xdd, + "aesdec" => 0xde, "aesdeclast" => 0xdf + ); + return undef if (!defined($opcodelet{$1})); + my $off = $2; + push @opcode,0x44 if ($3>=8); + push @opcode,0x0f,0x38,$opcodelet{$1}; + push @opcode,0x44|(($3&7)<<3),0x24; # ModR/M + push @opcode,($off=~/^0/?oct($off):$off)&0xff; + return ".byte\t".join(',',@opcode); + } + return $line; +} + +$code =~ s/\`([^\`]*)\`/eval($1)/gem; +$code =~ s/\b(aes.*%xmm[0-9]+).*$/aesni($1)/gem; + +print $code; +close STDOUT; diff --git a/openssl-1.1.0h/crypto/aes/asm/aesni-sha1-x86_64.pl b/openssl-1.1.0h/crypto/aes/asm/aesni-sha1-x86_64.pl new file mode 100644 index 0000000..33a7f0c --- /dev/null +++ b/openssl-1.1.0h/crypto/aes/asm/aesni-sha1-x86_64.pl @@ -0,0 +1,2072 @@ +#! /usr/bin/env perl +# Copyright 2011-2016 The OpenSSL Project Authors. All Rights Reserved. +# +# Licensed under the OpenSSL license (the "License"). You may not use +# this file except in compliance with the License. You can obtain a copy +# in the file LICENSE in the source distribution or at +# https://www.openssl.org/source/license.html + +# +# ==================================================================== +# Written by Andy Polyakov for the OpenSSL +# project. The module is, however, dual licensed under OpenSSL and +# CRYPTOGAMS licenses depending on where you obtain it. For further +# details see http://www.openssl.org/~appro/cryptogams/. +# ==================================================================== +# +# June 2011 +# +# This is AESNI-CBC+SHA1 "stitch" implementation. The idea, as spelled +# in http://download.intel.com/design/intarch/papers/323686.pdf, is +# that since AESNI-CBC encrypt exhibit *very* low instruction-level +# parallelism, interleaving it with another algorithm would allow to +# utilize processor resources better and achieve better performance. +# SHA1 instruction sequences(*) are taken from sha1-x86_64.pl and +# AESNI code is weaved into it. Below are performance numbers in +# cycles per processed byte, less is better, for standalone AESNI-CBC +# encrypt, sum of the latter and standalone SHA1, and "stitched" +# subroutine: +# +# AES-128-CBC +SHA1 stitch gain +# Westmere 3.77[+5.3] 9.07 6.55 +38% +# Sandy Bridge 5.05[+5.0(6.1)] 10.06(11.15) 5.98(7.05) +68%(+58%) +# Ivy Bridge 5.05[+4.6] 9.65 5.54 +74% +# Haswell 4.43[+3.6(4.2)] 8.00(8.58) 4.55(5.21) +75%(+65%) +# Skylake 2.63[+3.5(4.1)] 6.17(6.69) 4.23(4.44) +46%(+51%) +# Bulldozer 5.77[+6.0] 11.72 6.37 +84% +# Ryzen(**) 2.71[+1.93] 4.64 2.74 +69% +# Goldmont(**) 3.82[+1.70] 5.52 4.20 +31% +# +# AES-192-CBC +# Westmere 4.51 9.81 6.80 +44% +# Sandy Bridge 6.05 11.06(12.15) 6.11(7.19) +81%(+69%) +# Ivy Bridge 6.05 10.65 6.07 +75% +# Haswell 5.29 8.86(9.44) 5.32(5.32) +67%(+77%) +# Bulldozer 6.89 12.84 6.96 +84% +# +# AES-256-CBC +# Westmere 5.25 10.55 7.21 +46% +# Sandy Bridge 7.05 12.06(13.15) 7.12(7.72) +69%(+70%) +# Ivy Bridge 7.05 11.65 7.12 +64% +# Haswell 6.19 9.76(10.34) 6.21(6.25) +57%(+65%) +# Skylake 3.62 7.16(7.68) 4.56(4.76) +57%(+61%) +# Bulldozer 8.00 13.95 8.25 +69% +# Ryzen(**) 3.71 5.64 3.72 +52% +# Goldmont(**) 5.35 7.05 5.76 +22% +# +# (*) There are two code paths: SSSE3 and AVX. See sha1-568.pl for +# background information. Above numbers in parentheses are SSSE3 +# results collected on AVX-capable CPU, i.e. apply on OSes that +# don't support AVX. +# (**) SHAEXT results. +# +# Needless to mention that it makes no sense to implement "stitched" +# *decrypt* subroutine. Because *both* AESNI-CBC decrypt and SHA1 +# fully utilize parallelism, so stitching would not give any gain +# anyway. Well, there might be some, e.g. because of better cache +# locality... For reference, here are performance results for +# standalone AESNI-CBC decrypt: +# +# AES-128-CBC AES-192-CBC AES-256-CBC +# Westmere 1.25 1.50 1.75 +# Sandy Bridge 0.74 0.91 1.09 +# Ivy Bridge 0.74 0.90 1.11 +# Haswell 0.63 0.76 0.88 +# Bulldozer 0.70 0.85 0.99 + +# And indeed: +# +# AES-256-CBC +SHA1 stitch gain +# Westmere 1.75 7.20 6.68 +7.8% +# Sandy Bridge 1.09 6.09(7.22) 5.82(6.95) +4.6%(+3.9%) +# Ivy Bridge 1.11 5.70 5.45 +4.6% +# Haswell 0.88 4.45(5.00) 4.39(4.69) +1.4%(*)(+6.6%) +# Bulldozer 0.99 6.95 5.95 +17%(**) +# +# (*) Tiny improvement coefficient on Haswell is because we compare +# AVX1 stitch to sum with AVX2 SHA1. +# (**) Execution is fully dominated by integer code sequence and +# SIMD still hardly shows [in single-process benchmark;-] + +$flavour = shift; +$output = shift; +if ($flavour =~ /\./) { $output = $flavour; undef $flavour; } + +$win64=0; $win64=1 if ($flavour =~ /[nm]asm|mingw64/ || $output =~ /\.asm$/); + +$0 =~ m/(.*[\/\\])[^\/\\]+$/; $dir=$1; +( $xlate="${dir}x86_64-xlate.pl" and -f $xlate ) or +( $xlate="${dir}../../perlasm/x86_64-xlate.pl" and -f $xlate) or +die "can't locate x86_64-xlate.pl"; + +$avx=1 if (`$ENV{CC} -Wa,-v -c -o /dev/null -x assembler /dev/null 2>&1` + =~ /GNU assembler version ([2-9]\.[0-9]+)/ && + $1>=2.19); +$avx=1 if (!$avx && $win64 && ($flavour =~ /nasm/ || $ENV{ASM} =~ /nasm/) && + `nasm -v 2>&1` =~ /NASM version ([2-9]\.[0-9]+)/ && + $1>=2.09); +$avx=1 if (!$avx && $win64 && ($flavour =~ /masm/ || $ENV{ASM} =~ /ml64/) && + `ml64 2>&1` =~ /Version ([0-9]+)\./ && + $1>=10); +$avx=1 if (!$avx && `$ENV{CC} -v 2>&1` =~ /((?:^clang|LLVM) version|.*based on LLVM) ([3-9]\.[0-9]+)/ && $2>=3.0); + +$shaext=1; ### set to zero if compiling for 1.0.1 + +$stitched_decrypt=0; + +open OUT,"| \"$^X\" \"$xlate\" $flavour \"$output\""; +*STDOUT=*OUT; + +# void aesni_cbc_sha1_enc(const void *inp, +# void *out, +# size_t length, +# const AES_KEY *key, +# unsigned char *iv, +# SHA_CTX *ctx, +# const void *in0); + +$code.=<<___; +.text +.extern OPENSSL_ia32cap_P + +.globl aesni_cbc_sha1_enc +.type aesni_cbc_sha1_enc,\@abi-omnipotent +.align 32 +aesni_cbc_sha1_enc: + # caller should check for SSSE3 and AES-NI bits + mov OPENSSL_ia32cap_P+0(%rip),%r10d + mov OPENSSL_ia32cap_P+4(%rip),%r11 +___ +$code.=<<___ if ($shaext); + bt \$61,%r11 # check SHA bit + jc aesni_cbc_sha1_enc_shaext +___ +$code.=<<___ if ($avx); + and \$`1<<28`,%r11d # mask AVX bit + and \$`1<<30`,%r10d # mask "Intel CPU" bit + or %r11d,%r10d + cmp \$`1<<28|1<<30`,%r10d + je aesni_cbc_sha1_enc_avx +___ +$code.=<<___; + jmp aesni_cbc_sha1_enc_ssse3 + ret +.size aesni_cbc_sha1_enc,.-aesni_cbc_sha1_enc +___ + +my ($in0,$out,$len,$key,$ivp,$ctx,$inp)=("%rdi","%rsi","%rdx","%rcx","%r8","%r9","%r10"); + +my $Xi=4; +my @X=map("%xmm$_",(4..7,0..3)); +my @Tx=map("%xmm$_",(8..10)); +my @V=($A,$B,$C,$D,$E)=("%eax","%ebx","%ecx","%edx","%ebp"); # size optimization +my @T=("%esi","%edi"); +my $j=0; my $jj=0; my $r=0; my $sn=0; my $rx=0; +my $K_XX_XX="%r11"; +my ($rndkey0,$iv,$in)=map("%xmm$_",(11..13)); # for enc +my @rndkey=("%xmm14","%xmm15"); # for enc +my ($inout0,$inout1,$inout2,$inout3)=map("%xmm$_",(12..15)); # for dec + +if (1) { # reassign for Atom Silvermont + # The goal is to minimize amount of instructions with more than + # 3 prefix bytes. Or in more practical terms to keep AES-NI *and* + # SSSE3 instructions to upper half of the register bank. + @X=map("%xmm$_",(8..11,4..7)); + @Tx=map("%xmm$_",(12,13,3)); + ($iv,$in,$rndkey0)=map("%xmm$_",(2,14,15)); + @rndkey=("%xmm0","%xmm1"); +} + +sub AUTOLOAD() # thunk [simplified] 32-bit style perlasm +{ my $opcode = $AUTOLOAD; $opcode =~ s/.*:://; + my $arg = pop; + $arg = "\$$arg" if ($arg*1 eq $arg); + $code .= "\t$opcode\t".join(',',$arg,reverse @_)."\n"; +} + +my $_rol=sub { &rol(@_) }; +my $_ror=sub { &ror(@_) }; + +$code.=<<___; +.type aesni_cbc_sha1_enc_ssse3,\@function,6 +.align 32 +aesni_cbc_sha1_enc_ssse3: + mov `($win64?56:8)`(%rsp),$inp # load 7th argument + #shr \$6,$len # debugging artefact + #jz .Lepilogue_ssse3 # debugging artefact + push %rbx + push %rbp + push %r12 + push %r13 + push %r14 + push %r15 + lea `-104-($win64?10*16:0)`(%rsp),%rsp + #mov $in0,$inp # debugging artefact + #lea 64(%rsp),$ctx # debugging artefact +___ +$code.=<<___ if ($win64); + movaps %xmm6,96+0(%rsp) + movaps %xmm7,96+16(%rsp) + movaps %xmm8,96+32(%rsp) + movaps %xmm9,96+48(%rsp) + movaps %xmm10,96+64(%rsp) + movaps %xmm11,96+80(%rsp) + movaps %xmm12,96+96(%rsp) + movaps %xmm13,96+112(%rsp) + movaps %xmm14,96+128(%rsp) + movaps %xmm15,96+144(%rsp) +.Lprologue_ssse3: +___ +$code.=<<___; + mov $in0,%r12 # reassign arguments + mov $out,%r13 + mov $len,%r14 + lea 112($key),%r15 # size optimization + movdqu ($ivp),$iv # load IV + mov $ivp,88(%rsp) # save $ivp +___ +($in0,$out,$len,$key)=map("%r$_",(12..15)); # reassign arguments +my $rounds="${ivp}d"; +$code.=<<___; + shl \$6,$len + sub $in0,$out + mov 240-112($key),$rounds + add $inp,$len # end of input + + lea K_XX_XX(%rip),$K_XX_XX + mov 0($ctx),$A # load context + mov 4($ctx),$B + mov 8($ctx),$C + mov 12($ctx),$D + mov $B,@T[0] # magic seed + mov 16($ctx),$E + mov $C,@T[1] + xor $D,@T[1] + and @T[1],@T[0] + + movdqa 64($K_XX_XX),@Tx[2] # pbswap mask + movdqa 0($K_XX_XX),@Tx[1] # K_00_19 + movdqu 0($inp),@X[-4&7] # load input to %xmm[0-3] + movdqu 16($inp),@X[-3&7] + movdqu 32($inp),@X[-2&7] + movdqu 48($inp),@X[-1&7] + pshufb @Tx[2],@X[-4&7] # byte swap + pshufb @Tx[2],@X[-3&7] + pshufb @Tx[2],@X[-2&7] + add \$64,$inp + paddd @Tx[1],@X[-4&7] # add K_00_19 + pshufb @Tx[2],@X[-1&7] + paddd @Tx[1],@X[-3&7] + paddd @Tx[1],@X[-2&7] + movdqa @X[-4&7],0(%rsp) # X[]+K xfer to IALU + psubd @Tx[1],@X[-4&7] # restore X[] + movdqa @X[-3&7],16(%rsp) + psubd @Tx[1],@X[-3&7] + movdqa @X[-2&7],32(%rsp) + psubd @Tx[1],@X[-2&7] + movups -112($key),$rndkey0 # $key[0] + movups 16-112($key),$rndkey[0] # forward reference + jmp .Loop_ssse3 +___ + +my $aesenc=sub { + use integer; + my ($n,$k)=($r/10,$r%10); + if ($k==0) { + $code.=<<___; + movups `16*$n`($in0),$in # load input + xorps $rndkey0,$in +___ + $code.=<<___ if ($n); + movups $iv,`16*($n-1)`($out,$in0) # write output +___ + $code.=<<___; + xorps $in,$iv + movups `32+16*$k-112`($key),$rndkey[1] + aesenc $rndkey[0],$iv +___ + } elsif ($k==9) { + $sn++; + $code.=<<___; + cmp \$11,$rounds + jb .Laesenclast$sn + movups `32+16*($k+0)-112`($key),$rndkey[1] + aesenc $rndkey[0],$iv + movups `32+16*($k+1)-112`($key),$rndkey[0] + aesenc $rndkey[1],$iv + je .Laesenclast$sn + movups `32+16*($k+2)-112`($key),$rndkey[1] + aesenc $rndkey[0],$iv + movups `32+16*($k+3)-112`($key),$rndkey[0] + aesenc $rndkey[1],$iv +.Laesenclast$sn: + aesenclast $rndkey[0],$iv + movups 16-112($key),$rndkey[1] # forward reference +___ + } else { + $code.=<<___; + movups `32+16*$k-112`($key),$rndkey[1] + aesenc $rndkey[0],$iv +___ + } + $r++; unshift(@rndkey,pop(@rndkey)); +}; + +sub Xupdate_ssse3_16_31() # recall that $Xi starts with 4 +{ use integer; + my $body = shift; + my @insns = (&$body,&$body,&$body,&$body); # 40 instructions + my ($a,$b,$c,$d,$e); + + eval(shift(@insns)); # ror + &pshufd (@X[0],@X[-4&7],0xee); # was &movdqa (@X[0],@X[-3&7]); + eval(shift(@insns)); + &movdqa (@Tx[0],@X[-1&7]); + &paddd (@Tx[1],@X[-1&7]); + eval(shift(@insns)); + eval(shift(@insns)); + + &punpcklqdq(@X[0],@X[-3&7]); # compose "X[-14]" in "X[0]", was &palignr(@X[0],@X[-4&7],8); + eval(shift(@insns)); + eval(shift(@insns)); # rol + eval(shift(@insns)); + &psrldq (@Tx[0],4); # "X[-3]", 3 dwords + eval(shift(@insns)); + eval(shift(@insns)); + + &pxor (@X[0],@X[-4&7]); # "X[0]"^="X[-16]" + eval(shift(@insns)); + eval(shift(@insns)); # ror + &pxor (@Tx[0],@X[-2&7]); # "X[-3]"^"X[-8]" + eval(shift(@insns)); + eval(shift(@insns)); + eval(shift(@insns)); + + &pxor (@X[0],@Tx[0]); # "X[0]"^="X[-3]"^"X[-8]" + eval(shift(@insns)); + eval(shift(@insns)); # rol + &movdqa (eval(16*(($Xi-1)&3))."(%rsp)",@Tx[1]); # X[]+K xfer to IALU + eval(shift(@insns)); + eval(shift(@insns)); + + &movdqa (@Tx[2],@X[0]); + eval(shift(@insns)); + eval(shift(@insns)); + eval(shift(@insns)); # ror + &movdqa (@Tx[0],@X[0]); + eval(shift(@insns)); + + &pslldq (@Tx[2],12); # "X[0]"<<96, extract one dword + &paddd (@X[0],@X[0]); + eval(shift(@insns)); + eval(shift(@insns)); + + &psrld (@Tx[0],31); + eval(shift(@insns)); + eval(shift(@insns)); # rol + eval(shift(@insns)); + &movdqa (@Tx[1],@Tx[2]); + eval(shift(@insns)); + eval(shift(@insns)); + + &psrld (@Tx[2],30); + eval(shift(@insns)); + eval(shift(@insns)); # ror + &por (@X[0],@Tx[0]); # "X[0]"<<<=1 + eval(shift(@insns)); + eval(shift(@insns)); + eval(shift(@insns)); + + &pslld (@Tx[1],2); + &pxor (@X[0],@Tx[2]); + eval(shift(@insns)); + &movdqa (@Tx[2],eval(16*(($Xi)/5))."($K_XX_XX)"); # K_XX_XX + eval(shift(@insns)); # rol + eval(shift(@insns)); + eval(shift(@insns)); + + &pxor (@X[0],@Tx[1]); # "X[0]"^=("X[0]">>96)<<<2 + &pshufd (@Tx[1],@X[-1&7],0xee) if ($Xi==7); # was &movdqa (@Tx[0],@X[-1&7]) in Xupdate_ssse3_32_79 + + foreach (@insns) { eval; } # remaining instructions [if any] + + $Xi++; push(@X,shift(@X)); # "rotate" X[] + push(@Tx,shift(@Tx)); +} + +sub Xupdate_ssse3_32_79() +{ use integer; + my $body = shift; + my @insns = (&$body,&$body,&$body,&$body); # 32 to 44 instructions + my ($a,$b,$c,$d,$e); + + eval(shift(@insns)) if ($Xi==8); + &pxor (@X[0],@X[-4&7]); # "X[0]"="X[-32]"^"X[-16]" + eval(shift(@insns)) if ($Xi==8); + eval(shift(@insns)); # body_20_39 + eval(shift(@insns)); + eval(shift(@insns)) if (@insns[1] =~ /_ror/); + eval(shift(@insns)) if (@insns[0] =~ /_ror/); + &punpcklqdq(@Tx[0],@X[-1&7]); # compose "X[-6]", was &palignr(@Tx[0],@X[-2&7],8); + eval(shift(@insns)); + eval(shift(@insns)); # rol + + &pxor (@X[0],@X[-7&7]); # "X[0]"^="X[-28]" + eval(shift(@insns)); + eval(shift(@insns)); + if ($Xi%5) { + &movdqa (@Tx[2],@Tx[1]);# "perpetuate" K_XX_XX... + } else { # ... or load next one + &movdqa (@Tx[2],eval(16*($Xi/5))."($K_XX_XX)"); + } + eval(shift(@insns)); # ror + &paddd (@Tx[1],@X[-1&7]); + eval(shift(@insns)); + + &pxor (@X[0],@Tx[0]); # "X[0]"^="X[-6]" + eval(shift(@insns)); # body_20_39 + eval(shift(@insns)); + eval(shift(@insns)); + eval(shift(@insns)); # rol + eval(shift(@insns)) if (@insns[0] =~ /_ror/); + + &movdqa (@Tx[0],@X[0]); + eval(shift(@insns)); + eval(shift(@insns)); + &movdqa (eval(16*(($Xi-1)&3))."(%rsp)",@Tx[1]); # X[]+K xfer to IALU + eval(shift(@insns)); # ror + eval(shift(@insns)); + eval(shift(@insns)); # body_20_39 + + &pslld (@X[0],2); + eval(shift(@insns)); + eval(shift(@insns)); + &psrld (@Tx[0],30); + eval(shift(@insns)) if (@insns[0] =~ /_rol/);# rol + eval(shift(@insns)); + eval(shift(@insns)); + eval(shift(@insns)); # ror + + &por (@X[0],@Tx[0]); # "X[0]"<<<=2 + eval(shift(@insns)); + eval(shift(@insns)); # body_20_39 + eval(shift(@insns)) if (@insns[1] =~ /_rol/); + eval(shift(@insns)) if (@insns[0] =~ /_rol/); + &pshufd(@Tx[1],@X[-1&7],0xee) if ($Xi<19); # was &movdqa (@Tx[1],@X[0]) + eval(shift(@insns)); + eval(shift(@insns)); # rol + eval(shift(@insns)); + eval(shift(@insns)); + eval(shift(@insns)); # rol + eval(shift(@insns)); + + foreach (@insns) { eval; } # remaining instructions + + $Xi++; push(@X,shift(@X)); # "rotate" X[] + push(@Tx,shift(@Tx)); +} + +sub Xuplast_ssse3_80() +{ use integer; + my $body = shift; + my @insns = (&$body,&$body,&$body,&$body); # 32 instructions + my ($a,$b,$c,$d,$e); + + eval(shift(@insns)); + eval(shift(@insns)); + eval(shift(@insns)); + eval(shift(@insns)); + &paddd (@Tx[1],@X[-1&7]); + eval(shift(@insns)); + eval(shift(@insns)); + + &movdqa (eval(16*(($Xi-1)&3))."(%rsp)",@Tx[1]); # X[]+K xfer IALU + + foreach (@insns) { eval; } # remaining instructions + + &cmp ($inp,$len); + &je (shift); + + unshift(@Tx,pop(@Tx)); + + &movdqa (@Tx[2],"64($K_XX_XX)"); # pbswap mask + &movdqa (@Tx[1],"0($K_XX_XX)"); # K_00_19 + &movdqu (@X[-4&7],"0($inp)"); # load input + &movdqu (@X[-3&7],"16($inp)"); + &movdqu (@X[-2&7],"32($inp)"); + &movdqu (@X[-1&7],"48($inp)"); + &pshufb (@X[-4&7],@Tx[2]); # byte swap + &add ($inp,64); + + $Xi=0; +} + +sub Xloop_ssse3() +{ use integer; + my $body = shift; + my @insns = (&$body,&$body,&$body,&$body); # 32 instructions + my ($a,$b,$c,$d,$e); + + eval(shift(@insns)); + eval(shift(@insns)); + eval(shift(@insns)); + &pshufb (@X[($Xi-3)&7],@Tx[2]); + eval(shift(@insns)); + eval(shift(@insns)); + eval(shift(@insns)); + eval(shift(@insns)); + &paddd (@X[($Xi-4)&7],@Tx[1]); + eval(shift(@insns)); + eval(shift(@insns)); + eval(shift(@insns)); + eval(shift(@insns)); + &movdqa (eval(16*$Xi)."(%rsp)",@X[($Xi-4)&7]); # X[]+K xfer to IALU + eval(shift(@insns)); + eval(shift(@insns)); + eval(shift(@insns)); + eval(shift(@insns)); + &psubd (@X[($Xi-4)&7],@Tx[1]); + + foreach (@insns) { eval; } + $Xi++; +} + +sub Xtail_ssse3() +{ use integer; + my $body = shift; + my @insns = (&$body,&$body,&$body,&$body); # 32 instructions + my ($a,$b,$c,$d,$e); + + foreach (@insns) { eval; } +} + +my @body_00_19 = ( + '($a,$b,$c,$d,$e)=@V;'. + '&$_ror ($b,$j?7:2);', # $b>>>2 + '&xor (@T[0],$d);', + '&mov (@T[1],$a);', # $b for next round + + '&add ($e,eval(4*($j&15))."(%rsp)");',# X[]+K xfer + '&xor ($b,$c);', # $c^$d for next round + + '&$_rol ($a,5);', + '&add ($e,@T[0]);', + '&and (@T[1],$b);', # ($b&($c^$d)) for next round + + '&xor ($b,$c);', # restore $b + '&add ($e,$a);' .'$j++; unshift(@V,pop(@V)); unshift(@T,pop(@T));' + ); + +sub body_00_19 () { # ((c^d)&b)^d + # on start @T[0]=(c^d)&b + return &body_20_39() if ($rx==19); $rx++; + + use integer; + my ($k,$n); + my @r=@body_00_19; + + $n = scalar(@r); + $k = (($jj+1)*12/20)*20*$n/12; # 12 aesencs per these 20 rounds + @r[$k%$n].='&$aesenc();' if ($jj==$k/$n); + $jj++; + + return @r; +} + +my @body_20_39 = ( + '($a,$b,$c,$d,$e)=@V;'. + '&add ($e,eval(4*($j&15))."(%rsp)");',# X[]+K xfer + '&xor (@T[0],$d) if($j==19);'. + '&xor (@T[0],$c) if($j> 19);', # ($b^$d^$c) + '&mov (@T[1],$a);', # $b for next round + + '&$_rol ($a,5);', + '&add ($e,@T[0]);', + '&xor (@T[1],$c) if ($j< 79);', # $b^$d for next round + + '&$_ror ($b,7);', # $b>>>2 + '&add ($e,$a);' .'$j++; unshift(@V,pop(@V)); unshift(@T,pop(@T));' + ); + +sub body_20_39 () { # b^d^c + # on entry @T[0]=b^d + return &body_40_59() if ($rx==39); $rx++; + + use integer; + my ($k,$n); + my @r=@body_20_39; + + $n = scalar(@r); + $k = (($jj+1)*8/20)*20*$n/8; # 8 aesencs per these 20 rounds + @r[$k%$n].='&$aesenc();' if ($jj==$k/$n && $rx!=20); + $jj++; + + return @r; +} + +my @body_40_59 = ( + '($a,$b,$c,$d,$e)=@V;'. + '&add ($e,eval(4*($j&15))."(%rsp)");',# X[]+K xfer + '&and (@T[0],$c) if ($j>=40);', # (b^c)&(c^d) + '&xor ($c,$d) if ($j>=40);', # restore $c + + '&$_ror ($b,7);', # $b>>>2 + '&mov (@T[1],$a);', # $b for next round + '&xor (@T[0],$c);', + + '&$_rol ($a,5);', + '&add ($e,@T[0]);', + '&xor (@T[1],$c) if ($j==59);'. + '&xor (@T[1],$b) if ($j< 59);', # b^c for next round + + '&xor ($b,$c) if ($j< 59);', # c^d for next round + '&add ($e,$a);' .'$j++; unshift(@V,pop(@V)); unshift(@T,pop(@T));' + ); + +sub body_40_59 () { # ((b^c)&(c^d))^c + # on entry @T[0]=(b^c), (c^=d) + $rx++; + + use integer; + my ($k,$n); + my @r=@body_40_59; + + $n = scalar(@r); + $k=(($jj+1)*12/20)*20*$n/12; # 12 aesencs per these 20 rounds + @r[$k%$n].='&$aesenc();' if ($jj==$k/$n && $rx!=40); + $jj++; + + return @r; +} +$code.=<<___; +.align 32 +.Loop_ssse3: +___ + &Xupdate_ssse3_16_31(\&body_00_19); + &Xupdate_ssse3_16_31(\&body_00_19); + &Xupdate_ssse3_16_31(\&body_00_19); + &Xupdate_ssse3_16_31(\&body_00_19); + &Xupdate_ssse3_32_79(\&body_00_19); + &Xupdate_ssse3_32_79(\&body_20_39); + &Xupdate_ssse3_32_79(\&body_20_39); + &Xupdate_ssse3_32_79(\&body_20_39); + &Xupdate_ssse3_32_79(\&body_20_39); + &Xupdate_ssse3_32_79(\&body_20_39); + &Xupdate_ssse3_32_79(\&body_40_59); + &Xupdate_ssse3_32_79(\&body_40_59); + &Xupdate_ssse3_32_79(\&body_40_59); + &Xupdate_ssse3_32_79(\&body_40_59); + &Xupdate_ssse3_32_79(\&body_40_59); + &Xupdate_ssse3_32_79(\&body_20_39); + &Xuplast_ssse3_80(\&body_20_39,".Ldone_ssse3"); # can jump to "done" + + $saved_j=$j; @saved_V=@V; + $saved_r=$r; @saved_rndkey=@rndkey; + + &Xloop_ssse3(\&body_20_39); + &Xloop_ssse3(\&body_20_39); + &Xloop_ssse3(\&body_20_39); + +$code.=<<___; + movups $iv,48($out,$in0) # write output + lea 64($in0),$in0 + + add 0($ctx),$A # update context + add 4($ctx),@T[0] + add 8($ctx),$C + add 12($ctx),$D + mov $A,0($ctx) + add 16($ctx),$E + mov @T[0],4($ctx) + mov @T[0],$B # magic seed + mov $C,8($ctx) + mov $C,@T[1] + mov $D,12($ctx) + xor $D,@T[1] + mov $E,16($ctx) + and @T[1],@T[0] + jmp .Loop_ssse3 + +.Ldone_ssse3: +___ + $jj=$j=$saved_j; @V=@saved_V; + $r=$saved_r; @rndkey=@saved_rndkey; + + &Xtail_ssse3(\&body_20_39); + &Xtail_ssse3(\&body_20_39); + &Xtail_ssse3(\&body_20_39); + +$code.=<<___; + movups $iv,48($out,$in0) # write output + mov 88(%rsp),$ivp # restore $ivp + + add 0($ctx),$A # update context + add 4($ctx),@T[0] + add 8($ctx),$C + mov $A,0($ctx) + add 12($ctx),$D + mov @T[0],4($ctx) + add 16($ctx),$E + mov $C,8($ctx) + mov $D,12($ctx) + mov $E,16($ctx) + movups $iv,($ivp) # write IV +___ +$code.=<<___ if ($win64); + movaps 96+0(%rsp),%xmm6 + movaps 96+16(%rsp),%xmm7 + movaps 96+32(%rsp),%xmm8 + movaps 96+48(%rsp),%xmm9 + movaps 96+64(%rsp),%xmm10 + movaps 96+80(%rsp),%xmm11 + movaps 96+96(%rsp),%xmm12 + movaps 96+112(%rsp),%xmm13 + movaps 96+128(%rsp),%xmm14 + movaps 96+144(%rsp),%xmm15 +___ +$code.=<<___; + lea `104+($win64?10*16:0)`(%rsp),%rsi + mov 0(%rsi),%r15 + mov 8(%rsi),%r14 + mov 16(%rsi),%r13 + mov 24(%rsi),%r12 + mov 32(%rsi),%rbp + mov 40(%rsi),%rbx + lea 48(%rsi),%rsp +.Lepilogue_ssse3: + ret +.size aesni_cbc_sha1_enc_ssse3,.-aesni_cbc_sha1_enc_ssse3 +___ + + if ($stitched_decrypt) {{{ +# reset +($in0,$out,$len,$key,$ivp,$ctx,$inp)=("%rdi","%rsi","%rdx","%rcx","%r8","%r9","%r10"); +$j=$jj=$r=$rx=0; +$Xi=4; + +# reassign for Atom Silvermont (see above) +($inout0,$inout1,$inout2,$inout3,$rndkey0)=map("%xmm$_",(0..4)); +@X=map("%xmm$_",(8..13,6,7)); +@Tx=map("%xmm$_",(14,15,5)); + +my @aes256_dec = ( + '&movdqu($inout0,"0x00($in0)");', + '&movdqu($inout1,"0x10($in0)"); &pxor ($inout0,$rndkey0);', + '&movdqu($inout2,"0x20($in0)"); &pxor ($inout1,$rndkey0);', + '&movdqu($inout3,"0x30($in0)"); &pxor ($inout2,$rndkey0);', + + '&pxor ($inout3,$rndkey0); &movups ($rndkey0,"16-112($key)");', + '&movaps("64(%rsp)",@X[2]);', # save IV, originally @X[3] + undef,undef + ); +for ($i=0;$i<13;$i++) { + push (@aes256_dec,( + '&aesdec ($inout0,$rndkey0);', + '&aesdec ($inout1,$rndkey0);', + '&aesdec ($inout2,$rndkey0);', + '&aesdec ($inout3,$rndkey0); &movups($rndkey0,"'.(16*($i+2)-112).'($key)");' + )); + push (@aes256_dec,(undef,undef)) if (($i>=3 && $i<=5) || $i>=11); + push (@aes256_dec,(undef,undef)) if ($i==5); +} +push(@aes256_dec,( + '&aesdeclast ($inout0,$rndkey0); &movups (@X[0],"0x00($in0)");', + '&aesdeclast ($inout1,$rndkey0); &movups (@X[1],"0x10($in0)");', + '&aesdeclast ($inout2,$rndkey0); &movups (@X[2],"0x20($in0)");', + '&aesdeclast ($inout3,$rndkey0); &movups (@X[3],"0x30($in0)");', + + '&xorps ($inout0,"64(%rsp)"); &movdqu ($rndkey0,"-112($key)");', + '&xorps ($inout1,@X[0]); &movups ("0x00($out,$in0)",$inout0);', + '&xorps ($inout2,@X[1]); &movups ("0x10($out,$in0)",$inout1);', + '&xorps ($inout3,@X[2]); &movups ("0x20($out,$in0)",$inout2);', + + '&movups ("0x30($out,$in0)",$inout3);' + )); + +sub body_00_19_dec () { # ((c^d)&b)^d + # on start @T[0]=(c^d)&b + return &body_20_39_dec() if ($rx==19); + + my @r=@body_00_19; + + unshift (@r,@aes256_dec[$rx]) if (@aes256_dec[$rx]); + $rx++; + + return @r; +} + +sub body_20_39_dec () { # b^d^c + # on entry @T[0]=b^d + return &body_40_59_dec() if ($rx==39); + + my @r=@body_20_39; + + unshift (@r,@aes256_dec[$rx]) if (@aes256_dec[$rx]); + $rx++; + + return @r; +} + +sub body_40_59_dec () { # ((b^c)&(c^d))^c + # on entry @T[0]=(b^c), (c^=d) + + my @r=@body_40_59; + + unshift (@r,@aes256_dec[$rx]) if (@aes256_dec[$rx]); + $rx++; + + return @r; +} + +$code.=<<___; +.globl aesni256_cbc_sha1_dec +.type aesni256_cbc_sha1_dec,\@abi-omnipotent +.align 32 +aesni256_cbc_sha1_dec: + # caller should check for SSSE3 and AES-NI bits + mov OPENSSL_ia32cap_P+0(%rip),%r10d + mov OPENSSL_ia32cap_P+4(%rip),%r11d +___ +$code.=<<___ if ($avx); + and \$`1<<28`,%r11d # mask AVX bit + and \$`1<<30`,%r10d # mask "Intel CPU" bit + or %r11d,%r10d + cmp \$`1<<28|1<<30`,%r10d + je aesni256_cbc_sha1_dec_avx +___ +$code.=<<___; + jmp aesni256_cbc_sha1_dec_ssse3 + ret +.size aesni256_cbc_sha1_dec,.-aesni256_cbc_sha1_dec + +.type aesni256_cbc_sha1_dec_ssse3,\@function,6 +.align 32 +aesni256_cbc_sha1_dec_ssse3: + mov `($win64?56:8)`(%rsp),$inp # load 7th argument + push %rbx + push %rbp + push %r12 + push %r13 + push %r14 + push %r15 + lea `-104-($win64?10*16:0)`(%rsp),%rsp +___ +$code.=<<___ if ($win64); + movaps %xmm6,96+0(%rsp) + movaps %xmm7,96+16(%rsp) + movaps %xmm8,96+32(%rsp) + movaps %xmm9,96+48(%rsp) + movaps %xmm10,96+64(%rsp) + movaps %xmm11,96+80(%rsp) + movaps %xmm12,96+96(%rsp) + movaps %xmm13,96+112(%rsp) + movaps %xmm14,96+128(%rsp) + movaps %xmm15,96+144(%rsp) +.Lprologue_dec_ssse3: +___ +$code.=<<___; + mov $in0,%r12 # reassign arguments + mov $out,%r13 + mov $len,%r14 + lea 112($key),%r15 # size optimization + movdqu ($ivp),@X[3] # load IV + #mov $ivp,88(%rsp) # save $ivp +___ +($in0,$out,$len,$key)=map("%r$_",(12..15)); # reassign arguments +$code.=<<___; + shl \$6,$len + sub $in0,$out + add $inp,$len # end of input + + lea K_XX_XX(%rip),$K_XX_XX + mov 0($ctx),$A # load context + mov 4($ctx),$B + mov 8($ctx),$C + mov 12($ctx),$D + mov $B,@T[0] # magic seed + mov 16($ctx),$E + mov $C,@T[1] + xor $D,@T[1] + and @T[1],@T[0] + + movdqa 64($K_XX_XX),@Tx[2] # pbswap mask + movdqa 0($K_XX_XX),@Tx[1] # K_00_19 + movdqu 0($inp),@X[-4&7] # load input to %xmm[0-3] + movdqu 16($inp),@X[-3&7] + movdqu 32($inp),@X[-2&7] + movdqu 48($inp),@X[-1&7] + pshufb @Tx[2],@X[-4&7] # byte swap + add \$64,$inp + pshufb @Tx[2],@X[-3&7] + pshufb @Tx[2],@X[-2&7] + pshufb @Tx[2],@X[-1&7] + paddd @Tx[1],@X[-4&7] # add K_00_19 + paddd @Tx[1],@X[-3&7] + paddd @Tx[1],@X[-2&7] + movdqa @X[-4&7],0(%rsp) # X[]+K xfer to IALU + psubd @Tx[1],@X[-4&7] # restore X[] + movdqa @X[-3&7],16(%rsp) + psubd @Tx[1],@X[-3&7] + movdqa @X[-2&7],32(%rsp) + psubd @Tx[1],@X[-2&7] + movdqu -112($key),$rndkey0 # $key[0] + jmp .Loop_dec_ssse3 + +.align 32 +.Loop_dec_ssse3: +___ + &Xupdate_ssse3_16_31(\&body_00_19_dec); + &Xupdate_ssse3_16_31(\&body_00_19_dec); + &Xupdate_ssse3_16_31(\&body_00_19_dec); + &Xupdate_ssse3_16_31(\&body_00_19_dec); + &Xupdate_ssse3_32_79(\&body_00_19_dec); + &Xupdate_ssse3_32_79(\&body_20_39_dec); + &Xupdate_ssse3_32_79(\&body_20_39_dec); + &Xupdate_ssse3_32_79(\&body_20_39_dec); + &Xupdate_ssse3_32_79(\&body_20_39_dec); + &Xupdate_ssse3_32_79(\&body_20_39_dec); + &Xupdate_ssse3_32_79(\&body_40_59_dec); + &Xupdate_ssse3_32_79(\&body_40_59_dec); + &Xupdate_ssse3_32_79(\&body_40_59_dec); + &Xupdate_ssse3_32_79(\&body_40_59_dec); + &Xupdate_ssse3_32_79(\&body_40_59_dec); + &Xupdate_ssse3_32_79(\&body_20_39_dec); + &Xuplast_ssse3_80(\&body_20_39_dec,".Ldone_dec_ssse3"); # can jump to "done" + + $saved_j=$j; @saved_V=@V; + $saved_rx=$rx; + + &Xloop_ssse3(\&body_20_39_dec); + &Xloop_ssse3(\&body_20_39_dec); + &Xloop_ssse3(\&body_20_39_dec); + + eval(@aes256_dec[-1]); # last store +$code.=<<___; + lea 64($in0),$in0 + + add 0($ctx),$A # update context + add 4($ctx),@T[0] + add 8($ctx),$C + add 12($ctx),$D + mov $A,0($ctx) + add 16($ctx),$E + mov @T[0],4($ctx) + mov @T[0],$B # magic seed + mov $C,8($ctx) + mov $C,@T[1] + mov $D,12($ctx) + xor $D,@T[1] + mov $E,16($ctx) + and @T[1],@T[0] + jmp .Loop_dec_ssse3 + +.Ldone_dec_ssse3: +___ + $jj=$j=$saved_j; @V=@saved_V; + $rx=$saved_rx; + + &Xtail_ssse3(\&body_20_39_dec); + &Xtail_ssse3(\&body_20_39_dec); + &Xtail_ssse3(\&body_20_39_dec); + + eval(@aes256_dec[-1]); # last store +$code.=<<___; + add 0($ctx),$A # update context + add 4($ctx),@T[0] + add 8($ctx),$C + mov $A,0($ctx) + add 12($ctx),$D + mov @T[0],4($ctx) + add 16($ctx),$E + mov $C,8($ctx) + mov $D,12($ctx) + mov $E,16($ctx) + movups @X[3],($ivp) # write IV +___ +$code.=<<___ if ($win64); + movaps 96+0(%rsp),%xmm6 + movaps 96+16(%rsp),%xmm7 + movaps 96+32(%rsp),%xmm8 + movaps 96+48(%rsp),%xmm9 + movaps 96+64(%rsp),%xmm10 + movaps 96+80(%rsp),%xmm11 + movaps 96+96(%rsp),%xmm12 + movaps 96+112(%rsp),%xmm13 + movaps 96+128(%rsp),%xmm14 + movaps 96+144(%rsp),%xmm15 +___ +$code.=<<___; + lea `104+($win64?10*16:0)`(%rsp),%rsi + mov 0(%rsi),%r15 + mov 8(%rsi),%r14 + mov 16(%rsi),%r13 + mov 24(%rsi),%r12 + mov 32(%rsi),%rbp + mov 40(%rsi),%rbx + lea 48(%rsi),%rsp +.Lepilogue_dec_ssse3: + ret +.size aesni256_cbc_sha1_dec_ssse3,.-aesni256_cbc_sha1_dec_ssse3 +___ + }}} +$j=$jj=$r=$rx=0; + +if ($avx) { +my ($in0,$out,$len,$key,$ivp,$ctx,$inp)=("%rdi","%rsi","%rdx","%rcx","%r8","%r9","%r10"); + +my $Xi=4; +my @X=map("%xmm$_",(4..7,0..3)); +my @Tx=map("%xmm$_",(8..10)); +my @V=($A,$B,$C,$D,$E)=("%eax","%ebx","%ecx","%edx","%ebp"); # size optimization +my @T=("%esi","%edi"); +my ($rndkey0,$iv,$in)=map("%xmm$_",(11..13)); +my @rndkey=("%xmm14","%xmm15"); +my ($inout0,$inout1,$inout2,$inout3)=map("%xmm$_",(12..15)); # for dec +my $Kx=@Tx[2]; + +my $_rol=sub { &shld(@_[0],@_) }; +my $_ror=sub { &shrd(@_[0],@_) }; + +$code.=<<___; +.type aesni_cbc_sha1_enc_avx,\@function,6 +.align 32 +aesni_cbc_sha1_enc_avx: + mov `($win64?56:8)`(%rsp),$inp # load 7th argument + #shr \$6,$len # debugging artefact + #jz .Lepilogue_avx # debugging artefact + push %rbx + push %rbp + push %r12 + push %r13 + push %r14 + push %r15 + lea `-104-($win64?10*16:0)`(%rsp),%rsp + #mov $in0,$inp # debugging artefact + #lea 64(%rsp),$ctx # debugging artefact +___ +$code.=<<___ if ($win64); + movaps %xmm6,96+0(%rsp) + movaps %xmm7,96+16(%rsp) + movaps %xmm8,96+32(%rsp) + movaps %xmm9,96+48(%rsp) + movaps %xmm10,96+64(%rsp) + movaps %xmm11,96+80(%rsp) + movaps %xmm12,96+96(%rsp) + movaps %xmm13,96+112(%rsp) + movaps %xmm14,96+128(%rsp) + movaps %xmm15,96+144(%rsp) +.Lprologue_avx: +___ +$code.=<<___; + vzeroall + mov $in0,%r12 # reassign arguments + mov $out,%r13 + mov $len,%r14 + lea 112($key),%r15 # size optimization + vmovdqu ($ivp),$iv # load IV + mov $ivp,88(%rsp) # save $ivp +___ +($in0,$out,$len,$key)=map("%r$_",(12..15)); # reassign arguments +my $rounds="${ivp}d"; +$code.=<<___; + shl \$6,$len + sub $in0,$out + mov 240-112($key),$rounds + add $inp,$len # end of input + + lea K_XX_XX(%rip),$K_XX_XX + mov 0($ctx),$A # load context + mov 4($ctx),$B + mov 8($ctx),$C + mov 12($ctx),$D + mov $B,@T[0] # magic seed + mov 16($ctx),$E + mov $C,@T[1] + xor $D,@T[1] + and @T[1],@T[0] + + vmovdqa 64($K_XX_XX),@X[2] # pbswap mask + vmovdqa 0($K_XX_XX),$Kx # K_00_19 + vmovdqu 0($inp),@X[-4&7] # load input to %xmm[0-3] + vmovdqu 16($inp),@X[-3&7] + vmovdqu 32($inp),@X[-2&7] + vmovdqu 48($inp),@X[-1&7] + vpshufb @X[2],@X[-4&7],@X[-4&7] # byte swap + add \$64,$inp + vpshufb @X[2],@X[-3&7],@X[-3&7] + vpshufb @X[2],@X[-2&7],@X[-2&7] + vpshufb @X[2],@X[-1&7],@X[-1&7] + vpaddd $Kx,@X[-4&7],@X[0] # add K_00_19 + vpaddd $Kx,@X[-3&7],@X[1] + vpaddd $Kx,@X[-2&7],@X[2] + vmovdqa @X[0],0(%rsp) # X[]+K xfer to IALU + vmovdqa @X[1],16(%rsp) + vmovdqa @X[2],32(%rsp) + vmovups -112($key),$rndkey[1] # $key[0] + vmovups 16-112($key),$rndkey[0] # forward reference + jmp .Loop_avx +___ + +my $aesenc=sub { + use integer; + my ($n,$k)=($r/10,$r%10); + if ($k==0) { + $code.=<<___; + vmovdqu `16*$n`($in0),$in # load input + vpxor $rndkey[1],$in,$in +___ + $code.=<<___ if ($n); + vmovups $iv,`16*($n-1)`($out,$in0) # write output +___ + $code.=<<___; + vpxor $in,$iv,$iv + vaesenc $rndkey[0],$iv,$iv + vmovups `32+16*$k-112`($key),$rndkey[1] +___ + } elsif ($k==9) { + $sn++; + $code.=<<___; + cmp \$11,$rounds + jb .Lvaesenclast$sn + vaesenc $rndkey[0],$iv,$iv + vmovups `32+16*($k+0)-112`($key),$rndkey[1] + vaesenc $rndkey[1],$iv,$iv + vmovups `32+16*($k+1)-112`($key),$rndkey[0] + je .Lvaesenclast$sn + vaesenc $rndkey[0],$iv,$iv + vmovups `32+16*($k+2)-112`($key),$rndkey[1] + vaesenc $rndkey[1],$iv,$iv + vmovups `32+16*($k+3)-112`($key),$rndkey[0] +.Lvaesenclast$sn: + vaesenclast $rndkey[0],$iv,$iv + vmovups -112($key),$rndkey[0] + vmovups 16-112($key),$rndkey[1] # forward reference +___ + } else { + $code.=<<___; + vaesenc $rndkey[0],$iv,$iv + vmovups `32+16*$k-112`($key),$rndkey[1] +___ + } + $r++; unshift(@rndkey,pop(@rndkey)); +}; + +sub Xupdate_avx_16_31() # recall that $Xi starts with 4 +{ use integer; + my $body = shift; + my @insns = (&$body,&$body,&$body,&$body); # 40 instructions + my ($a,$b,$c,$d,$e); + + eval(shift(@insns)); + eval(shift(@insns)); + &vpalignr(@X[0],@X[-3&7],@X[-4&7],8); # compose "X[-14]" in "X[0]" + eval(shift(@insns)); + eval(shift(@insns)); + + &vpaddd (@Tx[1],$Kx,@X[-1&7]); + eval(shift(@insns)); + eval(shift(@insns)); + &vpsrldq(@Tx[0],@X[-1&7],4); # "X[-3]", 3 dwords + eval(shift(@insns)); + eval(shift(@insns)); + &vpxor (@X[0],@X[0],@X[-4&7]); # "X[0]"^="X[-16]" + eval(shift(@insns)); + eval(shift(@insns)); + + &vpxor (@Tx[0],@Tx[0],@X[-2&7]); # "X[-3]"^"X[-8]" + eval(shift(@insns)); + eval(shift(@insns)); + eval(shift(@insns)); + eval(shift(@insns)); + + &vpxor (@X[0],@X[0],@Tx[0]); # "X[0]"^="X[-3]"^"X[-8]" + eval(shift(@insns)); + eval(shift(@insns)); + &vmovdqa (eval(16*(($Xi-1)&3))."(%rsp)",@Tx[1]); # X[]+K xfer to IALU + eval(shift(@insns)); + eval(shift(@insns)); + + &vpsrld (@Tx[0],@X[0],31); + eval(shift(@insns)); + eval(shift(@insns)); + eval(shift(@insns)); + eval(shift(@insns)); + + &vpslldq(@Tx[1],@X[0],12); # "X[0]"<<96, extract one dword + &vpaddd (@X[0],@X[0],@X[0]); + eval(shift(@insns)); + eval(shift(@insns)); + eval(shift(@insns)); + eval(shift(@insns)); + + &vpor (@X[0],@X[0],@Tx[0]); # "X[0]"<<<=1 + &vpsrld (@Tx[0],@Tx[1],30); + eval(shift(@insns)); + eval(shift(@insns)); + eval(shift(@insns)); + eval(shift(@insns)); + + &vpslld (@Tx[1],@Tx[1],2); + &vpxor (@X[0],@X[0],@Tx[0]); + eval(shift(@insns)); + eval(shift(@insns)); + eval(shift(@insns)); + eval(shift(@insns)); + + &vpxor (@X[0],@X[0],@Tx[1]); # "X[0]"^=("X[0]">>96)<<<2 + eval(shift(@insns)); + eval(shift(@insns)); + &vmovdqa ($Kx,eval(16*(($Xi)/5))."($K_XX_XX)") if ($Xi%5==0); # K_XX_XX + eval(shift(@insns)); + eval(shift(@insns)); + + + foreach (@insns) { eval; } # remaining instructions [if any] + + $Xi++; push(@X,shift(@X)); # "rotate" X[] +} + +sub Xupdate_avx_32_79() +{ use integer; + my $body = shift; + my @insns = (&$body,&$body,&$body,&$body); # 32 to 48 instructions + my ($a,$b,$c,$d,$e); + + &vpalignr(@Tx[0],@X[-1&7],@X[-2&7],8); # compose "X[-6]" + &vpxor (@X[0],@X[0],@X[-4&7]); # "X[0]"="X[-32]"^"X[-16]" + eval(shift(@insns)); # body_20_39 + eval(shift(@insns)); + eval(shift(@insns)); + eval(shift(@insns)); # rol + + &vpxor (@X[0],@X[0],@X[-7&7]); # "X[0]"^="X[-28]" + eval(shift(@insns)); + eval(shift(@insns)) if (@insns[0] !~ /&ro[rl]/); + &vpaddd (@Tx[1],$Kx,@X[-1&7]); + &vmovdqa ($Kx,eval(16*($Xi/5))."($K_XX_XX)") if ($Xi%5==0); + eval(shift(@insns)); # ror + eval(shift(@insns)); + + &vpxor (@X[0],@X[0],@Tx[0]); # "X[0]"^="X[-6]" + eval(shift(@insns)); # body_20_39 + eval(shift(@insns)); + eval(shift(@insns)); + eval(shift(@insns)); # rol + + &vpsrld (@Tx[0],@X[0],30); + &vmovdqa (eval(16*(($Xi-1)&3))."(%rsp)",@Tx[1]); # X[]+K xfer to IALU + eval(shift(@insns)); + eval(shift(@insns)); + eval(shift(@insns)); # ror + eval(shift(@insns)); + + &vpslld (@X[0],@X[0],2); + eval(shift(@insns)); # body_20_39 + eval(shift(@insns)); + eval(shift(@insns)); + eval(shift(@insns)); # rol + eval(shift(@insns)); + eval(shift(@insns)); + eval(shift(@insns)); # ror + eval(shift(@insns)); + + &vpor (@X[0],@X[0],@Tx[0]); # "X[0]"<<<=2 + eval(shift(@insns)); # body_20_39 + eval(shift(@insns)); + eval(shift(@insns)); + eval(shift(@insns)); # rol + eval(shift(@insns)); + eval(shift(@insns)); + eval(shift(@insns)); # rol + eval(shift(@insns)); + + foreach (@insns) { eval; } # remaining instructions + + $Xi++; push(@X,shift(@X)); # "rotate" X[] +} + +sub Xuplast_avx_80() +{ use integer; + my $body = shift; + my @insns = (&$body,&$body,&$body,&$body); # 32 instructions + my ($a,$b,$c,$d,$e); + + eval(shift(@insns)); + &vpaddd (@Tx[1],$Kx,@X[-1&7]); + eval(shift(@insns)); + eval(shift(@insns)); + eval(shift(@insns)); + eval(shift(@insns)); + + &vmovdqa (eval(16*(($Xi-1)&3))."(%rsp)",@Tx[1]); # X[]+K xfer IALU + + foreach (@insns) { eval; } # remaining instructions + + &cmp ($inp,$len); + &je (shift); + + &vmovdqa(@Tx[1],"64($K_XX_XX)"); # pbswap mask + &vmovdqa($Kx,"0($K_XX_XX)"); # K_00_19 + &vmovdqu(@X[-4&7],"0($inp)"); # load input + &vmovdqu(@X[-3&7],"16($inp)"); + &vmovdqu(@X[-2&7],"32($inp)"); + &vmovdqu(@X[-1&7],"48($inp)"); + &vpshufb(@X[-4&7],@X[-4&7],@Tx[1]); # byte swap + &add ($inp,64); + + $Xi=0; +} + +sub Xloop_avx() +{ use integer; + my $body = shift; + my @insns = (&$body,&$body,&$body,&$body); # 32 instructions + my ($a,$b,$c,$d,$e); + + eval(shift(@insns)); + eval(shift(@insns)); + &vpshufb(@X[($Xi-3)&7],@X[($Xi-3)&7],@Tx[1]); + eval(shift(@insns)); + eval(shift(@insns)); + &vpaddd (@Tx[0],@X[($Xi-4)&7],$Kx); + eval(shift(@insns)); + eval(shift(@insns)); + eval(shift(@insns)); + eval(shift(@insns)); + &vmovdqa(eval(16*$Xi)."(%rsp)",@Tx[0]); # X[]+K xfer to IALU + eval(shift(@insns)); + eval(shift(@insns)); + + foreach (@insns) { eval; } + $Xi++; +} + +sub Xtail_avx() +{ use integer; + my $body = shift; + my @insns = (&$body,&$body,&$body,&$body); # 32 instructions + my ($a,$b,$c,$d,$e); + + foreach (@insns) { eval; } +} + +$code.=<<___; +.align 32 +.Loop_avx: +___ + &Xupdate_avx_16_31(\&body_00_19); + &Xupdate_avx_16_31(\&body_00_19); + &Xupdate_avx_16_31(\&body_00_19); + &Xupdate_avx_16_31(\&body_00_19); + &Xupdate_avx_32_79(\&body_00_19); + &Xupdate_avx_32_79(\&body_20_39); + &Xupdate_avx_32_79(\&body_20_39); + &Xupdate_avx_32_79(\&body_20_39); + &Xupdate_avx_32_79(\&body_20_39); + &Xupdate_avx_32_79(\&body_20_39); + &Xupdate_avx_32_79(\&body_40_59); + &Xupdate_avx_32_79(\&body_40_59); + &Xupdate_avx_32_79(\&body_40_59); + &Xupdate_avx_32_79(\&body_40_59); + &Xupdate_avx_32_79(\&body_40_59); + &Xupdate_avx_32_79(\&body_20_39); + &Xuplast_avx_80(\&body_20_39,".Ldone_avx"); # can jump to "done" + + $saved_j=$j; @saved_V=@V; + $saved_r=$r; @saved_rndkey=@rndkey; + + &Xloop_avx(\&body_20_39); + &Xloop_avx(\&body_20_39); + &Xloop_avx(\&body_20_39); + +$code.=<<___; + vmovups $iv,48($out,$in0) # write output + lea 64($in0),$in0 + + add 0($ctx),$A # update context + add 4($ctx),@T[0] + add 8($ctx),$C + add 12($ctx),$D + mov $A,0($ctx) + add 16($ctx),$E + mov @T[0],4($ctx) + mov @T[0],$B # magic seed + mov $C,8($ctx) + mov $C,@T[1] + mov $D,12($ctx) + xor $D,@T[1] + mov $E,16($ctx) + and @T[1],@T[0] + jmp .Loop_avx + +.Ldone_avx: +___ + $jj=$j=$saved_j; @V=@saved_V; + $r=$saved_r; @rndkey=@saved_rndkey; + + &Xtail_avx(\&body_20_39); + &Xtail_avx(\&body_20_39); + &Xtail_avx(\&body_20_39); + +$code.=<<___; + vmovups $iv,48($out,$in0) # write output + mov 88(%rsp),$ivp # restore $ivp + + add 0($ctx),$A # update context + add 4($ctx),@T[0] + add 8($ctx),$C + mov $A,0($ctx) + add 12($ctx),$D + mov @T[0],4($ctx) + add 16($ctx),$E + mov $C,8($ctx) + mov $D,12($ctx) + mov $E,16($ctx) + vmovups $iv,($ivp) # write IV + vzeroall +___ +$code.=<<___ if ($win64); + movaps 96+0(%rsp),%xmm6 + movaps 96+16(%rsp),%xmm7 + movaps 96+32(%rsp),%xmm8 + movaps 96+48(%rsp),%xmm9 + movaps 96+64(%rsp),%xmm10 + movaps 96+80(%rsp),%xmm11 + movaps 96+96(%rsp),%xmm12 + movaps 96+112(%rsp),%xmm13 + movaps 96+128(%rsp),%xmm14 + movaps 96+144(%rsp),%xmm15 +___ +$code.=<<___; + lea `104+($win64?10*16:0)`(%rsp),%rsi + mov 0(%rsi),%r15 + mov 8(%rsi),%r14 + mov 16(%rsi),%r13 + mov 24(%rsi),%r12 + mov 32(%rsi),%rbp + mov 40(%rsi),%rbx + lea 48(%rsi),%rsp +.Lepilogue_avx: + ret +.size aesni_cbc_sha1_enc_avx,.-aesni_cbc_sha1_enc_avx +___ + + if ($stitched_decrypt) {{{ +# reset +($in0,$out,$len,$key,$ivp,$ctx,$inp)=("%rdi","%rsi","%rdx","%rcx","%r8","%r9","%r10"); + +$j=$jj=$r=$rx=0; +$Xi=4; + +@aes256_dec = ( + '&vpxor ($inout0,$rndkey0,"0x00($in0)");', + '&vpxor ($inout1,$rndkey0,"0x10($in0)");', + '&vpxor ($inout2,$rndkey0,"0x20($in0)");', + '&vpxor ($inout3,$rndkey0,"0x30($in0)");', + + '&vmovups($rndkey0,"16-112($key)");', + '&vmovups("64(%rsp)",@X[2]);', # save IV, originally @X[3] + undef,undef + ); +for ($i=0;$i<13;$i++) { + push (@aes256_dec,( + '&vaesdec ($inout0,$inout0,$rndkey0);', + '&vaesdec ($inout1,$inout1,$rndkey0);', + '&vaesdec ($inout2,$inout2,$rndkey0);', + '&vaesdec ($inout3,$inout3,$rndkey0); &vmovups($rndkey0,"'.(16*($i+2)-112).'($key)");' + )); + push (@aes256_dec,(undef,undef)) if (($i>=3 && $i<=5) || $i>=11); + push (@aes256_dec,(undef,undef)) if ($i==5); +} +push(@aes256_dec,( + '&vaesdeclast ($inout0,$inout0,$rndkey0); &vmovups(@X[0],"0x00($in0)");', + '&vaesdeclast ($inout1,$inout1,$rndkey0); &vmovups(@X[1],"0x10($in0)");', + '&vaesdeclast ($inout2,$inout2,$rndkey0); &vmovups(@X[2],"0x20($in0)");', + '&vaesdeclast ($inout3,$inout3,$rndkey0); &vmovups(@X[3],"0x30($in0)");', + + '&vxorps ($inout0,$inout0,"64(%rsp)"); &vmovdqu($rndkey0,"-112($key)");', + '&vxorps ($inout1,$inout1,@X[0]); &vmovups("0x00($out,$in0)",$inout0);', + '&vxorps ($inout2,$inout2,@X[1]); &vmovups("0x10($out,$in0)",$inout1);', + '&vxorps ($inout3,$inout3,@X[2]); &vmovups("0x20($out,$in0)",$inout2);', + + '&vmovups ("0x30($out,$in0)",$inout3);' + )); + +$code.=<<___; +.type aesni256_cbc_sha1_dec_avx,\@function,6 +.align 32 +aesni256_cbc_sha1_dec_avx: + mov `($win64?56:8)`(%rsp),$inp # load 7th argument + push %rbx + push %rbp + push %r12 + push %r13 + push %r14 + push %r15 + lea `-104-($win64?10*16:0)`(%rsp),%rsp +___ +$code.=<<___ if ($win64); + movaps %xmm6,96+0(%rsp) + movaps %xmm7,96+16(%rsp) + movaps %xmm8,96+32(%rsp) + movaps %xmm9,96+48(%rsp) + movaps %xmm10,96+64(%rsp) + movaps %xmm11,96+80(%rsp) + movaps %xmm12,96+96(%rsp) + movaps %xmm13,96+112(%rsp) + movaps %xmm14,96+128(%rsp) + movaps %xmm15,96+144(%rsp) +.Lprologue_dec_avx: +___ +$code.=<<___; + vzeroall + mov $in0,%r12 # reassign arguments + mov $out,%r13 + mov $len,%r14 + lea 112($key),%r15 # size optimization + vmovdqu ($ivp),@X[3] # load IV +___ +($in0,$out,$len,$key)=map("%r$_",(12..15)); # reassign arguments +$code.=<<___; + shl \$6,$len + sub $in0,$out + add $inp,$len # end of input + + lea K_XX_XX(%rip),$K_XX_XX + mov 0($ctx),$A # load context + mov 4($ctx),$B + mov 8($ctx),$C + mov 12($ctx),$D + mov $B,@T[0] # magic seed + mov 16($ctx),$E + mov $C,@T[1] + xor $D,@T[1] + and @T[1],@T[0] + + vmovdqa 64($K_XX_XX),@X[2] # pbswap mask + vmovdqa 0($K_XX_XX),$Kx # K_00_19 + vmovdqu 0($inp),@X[-4&7] # load input to %xmm[0-3] + vmovdqu 16($inp),@X[-3&7] + vmovdqu 32($inp),@X[-2&7] + vmovdqu 48($inp),@X[-1&7] + vpshufb @X[2],@X[-4&7],@X[-4&7] # byte swap + add \$64,$inp + vpshufb @X[2],@X[-3&7],@X[-3&7] + vpshufb @X[2],@X[-2&7],@X[-2&7] + vpshufb @X[2],@X[-1&7],@X[-1&7] + vpaddd $Kx,@X[-4&7],@X[0] # add K_00_19 + vpaddd $Kx,@X[-3&7],@X[1] + vpaddd $Kx,@X[-2&7],@X[2] + vmovdqa @X[0],0(%rsp) # X[]+K xfer to IALU + vmovdqa @X[1],16(%rsp) + vmovdqa @X[2],32(%rsp) + vmovups -112($key),$rndkey0 # $key[0] + jmp .Loop_dec_avx + +.align 32 +.Loop_dec_avx: +___ + &Xupdate_avx_16_31(\&body_00_19_dec); + &Xupdate_avx_16_31(\&body_00_19_dec); + &Xupdate_avx_16_31(\&body_00_19_dec); + &Xupdate_avx_16_31(\&body_00_19_dec); + &Xupdate_avx_32_79(\&body_00_19_dec); + &Xupdate_avx_32_79(\&body_20_39_dec); + &Xupdate_avx_32_79(\&body_20_39_dec); + &Xupdate_avx_32_79(\&body_20_39_dec); + &Xupdate_avx_32_79(\&body_20_39_dec); + &Xupdate_avx_32_79(\&body_20_39_dec); + &Xupdate_avx_32_79(\&body_40_59_dec); + &Xupdate_avx_32_79(\&body_40_59_dec); + &Xupdate_avx_32_79(\&body_40_59_dec); + &Xupdate_avx_32_79(\&body_40_59_dec); + &Xupdate_avx_32_79(\&body_40_59_dec); + &Xupdate_avx_32_79(\&body_20_39_dec); + &Xuplast_avx_80(\&body_20_39_dec,".Ldone_dec_avx"); # can jump to "done" + + $saved_j=$j; @saved_V=@V; + $saved_rx=$rx; + + &Xloop_avx(\&body_20_39_dec); + &Xloop_avx(\&body_20_39_dec); + &Xloop_avx(\&body_20_39_dec); + + eval(@aes256_dec[-1]); # last store +$code.=<<___; + lea 64($in0),$in0 + + add 0($ctx),$A # update context + add 4($ctx),@T[0] + add 8($ctx),$C + add 12($ctx),$D + mov $A,0($ctx) + add 16($ctx),$E + mov @T[0],4($ctx) + mov @T[0],$B # magic seed + mov $C,8($ctx) + mov $C,@T[1] + mov $D,12($ctx) + xor $D,@T[1] + mov $E,16($ctx) + and @T[1],@T[0] + jmp .Loop_dec_avx + +.Ldone_dec_avx: +___ + $jj=$j=$saved_j; @V=@saved_V; + $rx=$saved_rx; + + &Xtail_avx(\&body_20_39_dec); + &Xtail_avx(\&body_20_39_dec); + &Xtail_avx(\&body_20_39_dec); + + eval(@aes256_dec[-1]); # last store +$code.=<<___; + + add 0($ctx),$A # update context + add 4($ctx),@T[0] + add 8($ctx),$C + mov $A,0($ctx) + add 12($ctx),$D + mov @T[0],4($ctx) + add 16($ctx),$E + mov $C,8($ctx) + mov $D,12($ctx) + mov $E,16($ctx) + vmovups @X[3],($ivp) # write IV + vzeroall +___ +$code.=<<___ if ($win64); + movaps 96+0(%rsp),%xmm6 + movaps 96+16(%rsp),%xmm7 + movaps 96+32(%rsp),%xmm8 + movaps 96+48(%rsp),%xmm9 + movaps 96+64(%rsp),%xmm10 + movaps 96+80(%rsp),%xmm11 + movaps 96+96(%rsp),%xmm12 + movaps 96+112(%rsp),%xmm13 + movaps 96+128(%rsp),%xmm14 + movaps 96+144(%rsp),%xmm15 +___ +$code.=<<___; + lea `104+($win64?10*16:0)`(%rsp),%rsi + mov 0(%rsi),%r15 + mov 8(%rsi),%r14 + mov 16(%rsi),%r13 + mov 24(%rsi),%r12 + mov 32(%rsi),%rbp + mov 40(%rsi),%rbx + lea 48(%rsi),%rsp +.Lepilogue_dec_avx: + ret +.size aesni256_cbc_sha1_dec_avx,.-aesni256_cbc_sha1_dec_avx +___ + }}} +} +$code.=<<___; +.align 64 +K_XX_XX: +.long 0x5a827999,0x5a827999,0x5a827999,0x5a827999 # K_00_19 +.long 0x6ed9eba1,0x6ed9eba1,0x6ed9eba1,0x6ed9eba1 # K_20_39 +.long 0x8f1bbcdc,0x8f1bbcdc,0x8f1bbcdc,0x8f1bbcdc # K_40_59 +.long 0xca62c1d6,0xca62c1d6,0xca62c1d6,0xca62c1d6 # K_60_79 +.long 0x00010203,0x04050607,0x08090a0b,0x0c0d0e0f # pbswap mask +.byte 0xf,0xe,0xd,0xc,0xb,0xa,0x9,0x8,0x7,0x6,0x5,0x4,0x3,0x2,0x1,0x0 + +.asciz "AESNI-CBC+SHA1 stitch for x86_64, CRYPTOGAMS by " +.align 64 +___ + if ($shaext) {{{ +($in0,$out,$len,$key,$ivp,$ctx,$inp)=("%rdi","%rsi","%rdx","%rcx","%r8","%r9","%r10"); + +$rounds="%r11d"; + +($iv,$in,$rndkey0)=map("%xmm$_",(2,14,15)); +@rndkey=("%xmm0","%xmm1"); +$r=0; + +my ($BSWAP,$ABCD,$E,$E_,$ABCD_SAVE,$E_SAVE)=map("%xmm$_",(7..12)); +my @MSG=map("%xmm$_",(3..6)); + +$code.=<<___; +.type aesni_cbc_sha1_enc_shaext,\@function,6 +.align 32 +aesni_cbc_sha1_enc_shaext: + mov `($win64?56:8)`(%rsp),$inp # load 7th argument +___ +$code.=<<___ if ($win64); + lea `-8-10*16`(%rsp),%rsp + movaps %xmm6,-8-10*16(%rax) + movaps %xmm7,-8-9*16(%rax) + movaps %xmm8,-8-8*16(%rax) + movaps %xmm9,-8-7*16(%rax) + movaps %xmm10,-8-6*16(%rax) + movaps %xmm11,-8-5*16(%rax) + movaps %xmm12,-8-4*16(%rax) + movaps %xmm13,-8-3*16(%rax) + movaps %xmm14,-8-2*16(%rax) + movaps %xmm15,-8-1*16(%rax) +.Lprologue_shaext: +___ +$code.=<<___; + movdqu ($ctx),$ABCD + movd 16($ctx),$E + movdqa K_XX_XX+0x50(%rip),$BSWAP # byte-n-word swap + + mov 240($key),$rounds + sub $in0,$out + movups ($key),$rndkey0 # $key[0] + movups ($ivp),$iv # load IV + movups 16($key),$rndkey[0] # forward reference + lea 112($key),$key # size optimization + + pshufd \$0b00011011,$ABCD,$ABCD # flip word order + pshufd \$0b00011011,$E,$E # flip word order + jmp .Loop_shaext + +.align 16 +.Loop_shaext: +___ + &$aesenc(); +$code.=<<___; + movdqu ($inp),@MSG[0] + movdqa $E,$E_SAVE # offload $E + pshufb $BSWAP,@MSG[0] + movdqu 0x10($inp),@MSG[1] + movdqa $ABCD,$ABCD_SAVE # offload $ABCD +___ + &$aesenc(); +$code.=<<___; + pshufb $BSWAP,@MSG[1] + + paddd @MSG[0],$E + movdqu 0x20($inp),@MSG[2] + lea 0x40($inp),$inp + pxor $E_SAVE,@MSG[0] # black magic +___ + &$aesenc(); +$code.=<<___; + pxor $E_SAVE,@MSG[0] # black magic + movdqa $ABCD,$E_ + pshufb $BSWAP,@MSG[2] + sha1rnds4 \$0,$E,$ABCD # 0-3 + sha1nexte @MSG[1],$E_ +___ + &$aesenc(); +$code.=<<___; + sha1msg1 @MSG[1],@MSG[0] + movdqu -0x10($inp),@MSG[3] + movdqa $ABCD,$E + pshufb $BSWAP,@MSG[3] +___ + &$aesenc(); +$code.=<<___; + sha1rnds4 \$0,$E_,$ABCD # 4-7 + sha1nexte @MSG[2],$E + pxor @MSG[2],@MSG[0] + sha1msg1 @MSG[2],@MSG[1] +___ + &$aesenc(); + +for($i=2;$i<20-4;$i++) { +$code.=<<___; + movdqa $ABCD,$E_ + sha1rnds4 \$`int($i/5)`,$E,$ABCD # 8-11 + sha1nexte @MSG[3],$E_ +___ + &$aesenc(); +$code.=<<___; + sha1msg2 @MSG[3],@MSG[0] + pxor @MSG[3],@MSG[1] + sha1msg1 @MSG[3],@MSG[2] +___ + ($E,$E_)=($E_,$E); + push(@MSG,shift(@MSG)); + + &$aesenc(); +} +$code.=<<___; + movdqa $ABCD,$E_ + sha1rnds4 \$3,$E,$ABCD # 64-67 + sha1nexte @MSG[3],$E_ + sha1msg2 @MSG[3],@MSG[0] + pxor @MSG[3],@MSG[1] +___ + &$aesenc(); +$code.=<<___; + movdqa $ABCD,$E + sha1rnds4 \$3,$E_,$ABCD # 68-71 + sha1nexte @MSG[0],$E + sha1msg2 @MSG[0],@MSG[1] +___ + &$aesenc(); +$code.=<<___; + movdqa $E_SAVE,@MSG[0] + movdqa $ABCD,$E_ + sha1rnds4 \$3,$E,$ABCD # 72-75 + sha1nexte @MSG[1],$E_ +___ + &$aesenc(); +$code.=<<___; + movdqa $ABCD,$E + sha1rnds4 \$3,$E_,$ABCD # 76-79 + sha1nexte $MSG[0],$E +___ + while($r<40) { &$aesenc(); } # remaining aesenc's +$code.=<<___; + dec $len + + paddd $ABCD_SAVE,$ABCD + movups $iv,48($out,$in0) # write output + lea 64($in0),$in0 + jnz .Loop_shaext + + pshufd \$0b00011011,$ABCD,$ABCD + pshufd \$0b00011011,$E,$E + movups $iv,($ivp) # write IV + movdqu $ABCD,($ctx) + movd $E,16($ctx) +___ +$code.=<<___ if ($win64); + movaps -8-10*16(%rax),%xmm6 + movaps -8-9*16(%rax),%xmm7 + movaps -8-8*16(%rax),%xmm8 + movaps -8-7*16(%rax),%xmm9 + movaps -8-6*16(%rax),%xmm10 + movaps -8-5*16(%rax),%xmm11 + movaps -8-4*16(%rax),%xmm12 + movaps -8-3*16(%rax),%xmm13 + movaps -8-2*16(%rax),%xmm14 + movaps -8-1*16(%rax),%xmm15 + mov %rax,%rsp +.Lepilogue_shaext: +___ +$code.=<<___; + ret +.size aesni_cbc_sha1_enc_shaext,.-aesni_cbc_sha1_enc_shaext +___ + }}} +# EXCEPTION_DISPOSITION handler (EXCEPTION_RECORD *rec,ULONG64 frame, +# CONTEXT *context,DISPATCHER_CONTEXT *disp) +if ($win64) { +$rec="%rcx"; +$frame="%rdx"; +$context="%r8"; +$disp="%r9"; + +$code.=<<___; +.extern __imp_RtlVirtualUnwind +.type ssse3_handler,\@abi-omnipotent +.align 16 +ssse3_handler: + push %rsi + push %rdi + push %rbx + push %rbp + push %r12 + push %r13 + push %r14 + push %r15 + pushfq + sub \$64,%rsp + + mov 120($context),%rax # pull context->Rax + mov 248($context),%rbx # pull context->Rip + + mov 8($disp),%rsi # disp->ImageBase + mov 56($disp),%r11 # disp->HandlerData + + mov 0(%r11),%r10d # HandlerData[0] + lea (%rsi,%r10),%r10 # prologue label + cmp %r10,%rbx # context->RipRsp + + mov 4(%r11),%r10d # HandlerData[1] + lea (%rsi,%r10),%r10 # epilogue label + cmp %r10,%rbx # context->Rip>=epilogue label + jae .Lcommon_seh_tail +___ +$code.=<<___ if ($shaext); + lea aesni_cbc_sha1_enc_shaext(%rip),%r10 + cmp %r10,%rbx + jb .Lseh_no_shaext + + lea (%rax),%rsi + lea 512($context),%rdi # &context.Xmm6 + mov \$20,%ecx + .long 0xa548f3fc # cld; rep movsq + lea 168(%rax),%rax # adjust stack pointer + jmp .Lcommon_seh_tail +.Lseh_no_shaext: +___ +$code.=<<___; + lea 96(%rax),%rsi + lea 512($context),%rdi # &context.Xmm6 + mov \$20,%ecx + .long 0xa548f3fc # cld; rep movsq + lea `104+10*16`(%rax),%rax # adjust stack pointer + + mov 0(%rax),%r15 + mov 8(%rax),%r14 + mov 16(%rax),%r13 + mov 24(%rax),%r12 + mov 32(%rax),%rbp + mov 40(%rax),%rbx + lea 48(%rax),%rax + mov %rbx,144($context) # restore context->Rbx + mov %rbp,160($context) # restore context->Rbp + mov %r12,216($context) # restore context->R12 + mov %r13,224($context) # restore context->R13 + mov %r14,232($context) # restore context->R14 + mov %r15,240($context) # restore context->R15 + +.Lcommon_seh_tail: + mov 8(%rax),%rdi + mov 16(%rax),%rsi + mov %rax,152($context) # restore context->Rsp + mov %rsi,168($context) # restore context->Rsi + mov %rdi,176($context) # restore context->Rdi + + mov 40($disp),%rdi # disp->ContextRecord + mov $context,%rsi # context + mov \$154,%ecx # sizeof(CONTEXT) + .long 0xa548f3fc # cld; rep movsq + + mov $disp,%rsi + xor %rcx,%rcx # arg1, UNW_FLAG_NHANDLER + mov 8(%rsi),%rdx # arg2, disp->ImageBase + mov 0(%rsi),%r8 # arg3, disp->ControlPc + mov 16(%rsi),%r9 # arg4, disp->FunctionEntry + mov 40(%rsi),%r10 # disp->ContextRecord + lea 56(%rsi),%r11 # &disp->HandlerData + lea 24(%rsi),%r12 # &disp->EstablisherFrame + mov %r10,32(%rsp) # arg5 + mov %r11,40(%rsp) # arg6 + mov %r12,48(%rsp) # arg7 + mov %rcx,56(%rsp) # arg8, (NULL) + call *__imp_RtlVirtualUnwind(%rip) + + mov \$1,%eax # ExceptionContinueSearch + add \$64,%rsp + popfq + pop %r15 + pop %r14 + pop %r13 + pop %r12 + pop %rbp + pop %rbx + pop %rdi + pop %rsi + ret +.size ssse3_handler,.-ssse3_handler + +.section .pdata +.align 4 + .rva .LSEH_begin_aesni_cbc_sha1_enc_ssse3 + .rva .LSEH_end_aesni_cbc_sha1_enc_ssse3 + .rva .LSEH_info_aesni_cbc_sha1_enc_ssse3 +___ +$code.=<<___ if ($avx); + .rva .LSEH_begin_aesni_cbc_sha1_enc_avx + .rva .LSEH_end_aesni_cbc_sha1_enc_avx + .rva .LSEH_info_aesni_cbc_sha1_enc_avx +___ +$code.=<<___ if ($shaext); + .rva .LSEH_begin_aesni_cbc_sha1_enc_shaext + .rva .LSEH_end_aesni_cbc_sha1_enc_shaext + .rva .LSEH_info_aesni_cbc_sha1_enc_shaext +___ +$code.=<<___; +.section .xdata +.align 8 +.LSEH_info_aesni_cbc_sha1_enc_ssse3: + .byte 9,0,0,0 + .rva ssse3_handler + .rva .Lprologue_ssse3,.Lepilogue_ssse3 # HandlerData[] +___ +$code.=<<___ if ($avx); +.LSEH_info_aesni_cbc_sha1_enc_avx: + .byte 9,0,0,0 + .rva ssse3_handler + .rva .Lprologue_avx,.Lepilogue_avx # HandlerData[] +___ +$code.=<<___ if ($shaext); +.LSEH_info_aesni_cbc_sha1_enc_shaext: + .byte 9,0,0,0 + .rva ssse3_handler + .rva .Lprologue_shaext,.Lepilogue_shaext # HandlerData[] +___ +} + +#################################################################### +sub rex { + local *opcode=shift; + my ($dst,$src)=@_; + my $rex=0; + + $rex|=0x04 if($dst>=8); + $rex|=0x01 if($src>=8); + unshift @opcode,$rex|0x40 if($rex); +} + +sub sha1rnds4 { + if (@_[0] =~ /\$([x0-9a-f]+),\s*%xmm([0-9]+),\s*%xmm([0-9]+)/) { + my @opcode=(0x0f,0x3a,0xcc); + rex(\@opcode,$3,$2); + push @opcode,0xc0|($2&7)|(($3&7)<<3); # ModR/M + my $c=$1; + push @opcode,$c=~/^0/?oct($c):$c; + return ".byte\t".join(',',@opcode); + } else { + return "sha1rnds4\t".@_[0]; + } +} + +sub sha1op38 { + my $instr = shift; + my %opcodelet = ( + "sha1nexte" => 0xc8, + "sha1msg1" => 0xc9, + "sha1msg2" => 0xca ); + + if (defined($opcodelet{$instr}) && @_[0] =~ /%xmm([0-9]+),\s*%xmm([0-9]+)/) { + my @opcode=(0x0f,0x38); + rex(\@opcode,$2,$1); + push @opcode,$opcodelet{$instr}; + push @opcode,0xc0|($1&7)|(($2&7)<<3); # ModR/M + return ".byte\t".join(',',@opcode); + } else { + return $instr."\t".@_[0]; + } +} + +sub aesni { + my $line=shift; + my @opcode=(0x0f,0x38); + + if ($line=~/(aes[a-z]+)\s+%xmm([0-9]+),\s*%xmm([0-9]+)/) { + my %opcodelet = ( + "aesenc" => 0xdc, "aesenclast" => 0xdd, + "aesdec" => 0xde, "aesdeclast" => 0xdf + ); + return undef if (!defined($opcodelet{$1})); + rex(\@opcode,$3,$2); + push @opcode,$opcodelet{$1},0xc0|($2&7)|(($3&7)<<3); # ModR/M + unshift @opcode,0x66; + return ".byte\t".join(',',@opcode); + } + return $line; +} + +foreach (split("\n",$code)) { + s/\`([^\`]*)\`/eval $1/geo; + + s/\b(sha1rnds4)\s+(.*)/sha1rnds4($2)/geo or + s/\b(sha1[^\s]*)\s+(.*)/sha1op38($1,$2)/geo or + s/\b(aes.*%xmm[0-9]+).*$/aesni($1)/geo; + + print $_,"\n"; +} +close STDOUT; diff --git a/openssl-1.1.0h/crypto/aes/asm/aesni-sha256-x86_64.pl b/openssl-1.1.0h/crypto/aes/asm/aesni-sha256-x86_64.pl new file mode 100644 index 0000000..0e49f26 --- /dev/null +++ b/openssl-1.1.0h/crypto/aes/asm/aesni-sha256-x86_64.pl @@ -0,0 +1,1717 @@ +#! /usr/bin/env perl +# Copyright 2013-2016 The OpenSSL Project Authors. All Rights Reserved. +# +# Licensed under the OpenSSL license (the "License"). You may not use +# this file except in compliance with the License. You can obtain a copy +# in the file LICENSE in the source distribution or at +# https://www.openssl.org/source/license.html + +# +# ==================================================================== +# Written by Andy Polyakov for the OpenSSL +# project. The module is, however, dual licensed under OpenSSL and +# CRYPTOGAMS licenses depending on where you obtain it. For further +# details see http://www.openssl.org/~appro/cryptogams/. +# ==================================================================== +# +# January 2013 +# +# This is AESNI-CBC+SHA256 stitch implementation. The idea, as spelled +# in http://download.intel.com/design/intarch/papers/323686.pdf, is +# that since AESNI-CBC encrypt exhibit *very* low instruction-level +# parallelism, interleaving it with another algorithm would allow to +# utilize processor resources better and achieve better performance. +# SHA256 instruction sequences(*) are taken from sha512-x86_64.pl and +# AESNI code is weaved into it. As SHA256 dominates execution time, +# stitch performance does not depend on AES key length. Below are +# performance numbers in cycles per processed byte, less is better, +# for standalone AESNI-CBC encrypt, standalone SHA256, and stitched +# subroutine: +# +# AES-128/-192/-256+SHA256 this(**) gain +# Sandy Bridge 5.05/6.05/7.05+11.6 13.0 +28%/36%/43% +# Ivy Bridge 5.05/6.05/7.05+10.3 11.6 +32%/41%/50% +# Haswell 4.43/5.29/6.19+7.80 8.79 +39%/49%/59% +# Skylake 2.62/3.14/3.62+7.70 8.10 +27%/34%/40% +# Bulldozer 5.77/6.89/8.00+13.7 13.7 +42%/50%/58% +# Ryzen(***) 2.71/-/3.71+2.05 2.74/-/3.73 +74%/-/54% +# Goldmont(***) 3.82/-/5.35+4.16 4.73/-/5.94 +69%/-/60% +# +# (*) there are XOP, AVX1 and AVX2 code paths, meaning that +# Westmere is omitted from loop, this is because gain was not +# estimated high enough to justify the effort; +# (**) these are EVP-free results, results obtained with 'speed +# -evp aes-256-cbc-hmac-sha256' will vary by percent or two; +# (***) these are SHAEXT results; + +$flavour = shift; +$output = shift; +if ($flavour =~ /\./) { $output = $flavour; undef $flavour; } + +$win64=0; $win64=1 if ($flavour =~ /[nm]asm|mingw64/ || $output =~ /\.asm$/); + +$0 =~ m/(.*[\/\\])[^\/\\]+$/; $dir=$1; +( $xlate="${dir}x86_64-xlate.pl" and -f $xlate ) or +( $xlate="${dir}../../perlasm/x86_64-xlate.pl" and -f $xlate) or +die "can't locate x86_64-xlate.pl"; + +if (`$ENV{CC} -Wa,-v -c -o /dev/null -x assembler /dev/null 2>&1` + =~ /GNU assembler version ([2-9]\.[0-9]+)/) { + $avx = ($1>=2.19) + ($1>=2.22); +} + +if (!$avx && $win64 && ($flavour =~ /nasm/ || $ENV{ASM} =~ /nasm/) && + `nasm -v 2>&1` =~ /NASM version ([2-9]\.[0-9]+)/) { + $avx = ($1>=2.09) + ($1>=2.10); +} + +if (!$avx && $win64 && ($flavour =~ /masm/ || $ENV{ASM} =~ /ml64/) && + `ml64 2>&1` =~ /Version ([0-9]+)\./) { + $avx = ($1>=10) + ($1>=12); +} + +if (!$avx && `$ENV{CC} -v 2>&1` =~ /((?:^clang|LLVM) version|.*based on LLVM) ([3-9]\.[0-9]+)/) { + $avx = ($2>=3.0) + ($2>3.0); +} + +$shaext=$avx; ### set to zero if compiling for 1.0.1 +$avx=1 if (!$shaext && $avx); + +open OUT,"| \"$^X\" \"$xlate\" $flavour \"$output\""; +*STDOUT=*OUT; + +$func="aesni_cbc_sha256_enc"; +$TABLE="K256"; +$SZ=4; +@ROT=($A,$B,$C,$D,$E,$F,$G,$H)=("%eax","%ebx","%ecx","%edx", + "%r8d","%r9d","%r10d","%r11d"); +($T1,$a0,$a1,$a2,$a3)=("%r12d","%r13d","%r14d","%r15d","%esi"); +@Sigma0=( 2,13,22); +@Sigma1=( 6,11,25); +@sigma0=( 7,18, 3); +@sigma1=(17,19,10); +$rounds=64; + +######################################################################## +# void aesni_cbc_sha256_enc(const void *inp, +# void *out, +# size_t length, +# const AES_KEY *key, +# unsigned char *iv, +# SHA256_CTX *ctx, +# const void *in0); +($inp, $out, $len, $key, $ivp, $ctx, $in0) = +("%rdi","%rsi","%rdx","%rcx","%r8","%r9","%r10"); + +$Tbl="%rbp"; + +$_inp="16*$SZ+0*8(%rsp)"; +$_out="16*$SZ+1*8(%rsp)"; +$_end="16*$SZ+2*8(%rsp)"; +$_key="16*$SZ+3*8(%rsp)"; +$_ivp="16*$SZ+4*8(%rsp)"; +$_ctx="16*$SZ+5*8(%rsp)"; +$_in0="16*$SZ+6*8(%rsp)"; +$_rsp="16*$SZ+7*8(%rsp)"; +$framesz=16*$SZ+8*8; + +$code=<<___; +.text + +.extern OPENSSL_ia32cap_P +.globl $func +.type $func,\@abi-omnipotent +.align 16 +$func: +___ + if ($avx) { +$code.=<<___; + lea OPENSSL_ia32cap_P(%rip),%r11 + mov \$1,%eax + cmp \$0,`$win64?"%rcx":"%rdi"` + je .Lprobe + mov 0(%r11),%eax + mov 4(%r11),%r10 +___ +$code.=<<___ if ($shaext); + bt \$61,%r10 # check for SHA + jc ${func}_shaext +___ +$code.=<<___; + mov %r10,%r11 + shr \$32,%r11 + + test \$`1<<11`,%r10d # check for XOP + jnz ${func}_xop +___ +$code.=<<___ if ($avx>1); + and \$`1<<8|1<<5|1<<3`,%r11d # check for BMI2+AVX2+BMI1 + cmp \$`1<<8|1<<5|1<<3`,%r11d + je ${func}_avx2 +___ +$code.=<<___; + and \$`1<<28`,%r10d # check for AVX + jnz ${func}_avx + ud2 +___ + } +$code.=<<___; + xor %eax,%eax + cmp \$0,`$win64?"%rcx":"%rdi"` + je .Lprobe + ud2 +.Lprobe: + ret +.size $func,.-$func + +.align 64 +.type $TABLE,\@object +$TABLE: + .long 0x428a2f98,0x71374491,0xb5c0fbcf,0xe9b5dba5 + .long 0x428a2f98,0x71374491,0xb5c0fbcf,0xe9b5dba5 + .long 0x3956c25b,0x59f111f1,0x923f82a4,0xab1c5ed5 + .long 0x3956c25b,0x59f111f1,0x923f82a4,0xab1c5ed5 + .long 0xd807aa98,0x12835b01,0x243185be,0x550c7dc3 + .long 0xd807aa98,0x12835b01,0x243185be,0x550c7dc3 + .long 0x72be5d74,0x80deb1fe,0x9bdc06a7,0xc19bf174 + .long 0x72be5d74,0x80deb1fe,0x9bdc06a7,0xc19bf174 + .long 0xe49b69c1,0xefbe4786,0x0fc19dc6,0x240ca1cc + .long 0xe49b69c1,0xefbe4786,0x0fc19dc6,0x240ca1cc + .long 0x2de92c6f,0x4a7484aa,0x5cb0a9dc,0x76f988da + .long 0x2de92c6f,0x4a7484aa,0x5cb0a9dc,0x76f988da + .long 0x983e5152,0xa831c66d,0xb00327c8,0xbf597fc7 + .long 0x983e5152,0xa831c66d,0xb00327c8,0xbf597fc7 + .long 0xc6e00bf3,0xd5a79147,0x06ca6351,0x14292967 + .long 0xc6e00bf3,0xd5a79147,0x06ca6351,0x14292967 + .long 0x27b70a85,0x2e1b2138,0x4d2c6dfc,0x53380d13 + .long 0x27b70a85,0x2e1b2138,0x4d2c6dfc,0x53380d13 + .long 0x650a7354,0x766a0abb,0x81c2c92e,0x92722c85 + .long 0x650a7354,0x766a0abb,0x81c2c92e,0x92722c85 + .long 0xa2bfe8a1,0xa81a664b,0xc24b8b70,0xc76c51a3 + .long 0xa2bfe8a1,0xa81a664b,0xc24b8b70,0xc76c51a3 + .long 0xd192e819,0xd6990624,0xf40e3585,0x106aa070 + .long 0xd192e819,0xd6990624,0xf40e3585,0x106aa070 + .long 0x19a4c116,0x1e376c08,0x2748774c,0x34b0bcb5 + .long 0x19a4c116,0x1e376c08,0x2748774c,0x34b0bcb5 + .long 0x391c0cb3,0x4ed8aa4a,0x5b9cca4f,0x682e6ff3 + .long 0x391c0cb3,0x4ed8aa4a,0x5b9cca4f,0x682e6ff3 + .long 0x748f82ee,0x78a5636f,0x84c87814,0x8cc70208 + .long 0x748f82ee,0x78a5636f,0x84c87814,0x8cc70208 + .long 0x90befffa,0xa4506ceb,0xbef9a3f7,0xc67178f2 + .long 0x90befffa,0xa4506ceb,0xbef9a3f7,0xc67178f2 + + .long 0x00010203,0x04050607,0x08090a0b,0x0c0d0e0f + .long 0x00010203,0x04050607,0x08090a0b,0x0c0d0e0f + .long 0,0,0,0, 0,0,0,0, -1,-1,-1,-1 + .long 0,0,0,0, 0,0,0,0 + .asciz "AESNI-CBC+SHA256 stitch for x86_64, CRYPTOGAMS by " +.align 64 +___ + +###################################################################### +# SIMD code paths +# +{{{ +($iv,$inout,$roundkey,$temp, + $mask10,$mask12,$mask14,$offload)=map("%xmm$_",(8..15)); + +$aesni_cbc_idx=0; +@aesni_cbc_block = ( +## &vmovdqu ($roundkey,"0x00-0x80($inp)");' +## &vmovdqu ($inout,($inp)); +## &mov ($_inp,$inp); + + '&vpxor ($inout,$inout,$roundkey);'. + ' &vmovdqu ($roundkey,"0x10-0x80($inp)");', + + '&vpxor ($inout,$inout,$iv);', + + '&vaesenc ($inout,$inout,$roundkey);'. + ' &vmovdqu ($roundkey,"0x20-0x80($inp)");', + + '&vaesenc ($inout,$inout,$roundkey);'. + ' &vmovdqu ($roundkey,"0x30-0x80($inp)");', + + '&vaesenc ($inout,$inout,$roundkey);'. + ' &vmovdqu ($roundkey,"0x40-0x80($inp)");', + + '&vaesenc ($inout,$inout,$roundkey);'. + ' &vmovdqu ($roundkey,"0x50-0x80($inp)");', + + '&vaesenc ($inout,$inout,$roundkey);'. + ' &vmovdqu ($roundkey,"0x60-0x80($inp)");', + + '&vaesenc ($inout,$inout,$roundkey);'. + ' &vmovdqu ($roundkey,"0x70-0x80($inp)");', + + '&vaesenc ($inout,$inout,$roundkey);'. + ' &vmovdqu ($roundkey,"0x80-0x80($inp)");', + + '&vaesenc ($inout,$inout,$roundkey);'. + ' &vmovdqu ($roundkey,"0x90-0x80($inp)");', + + '&vaesenc ($inout,$inout,$roundkey);'. + ' &vmovdqu ($roundkey,"0xa0-0x80($inp)");', + + '&vaesenclast ($temp,$inout,$roundkey);'. + ' &vaesenc ($inout,$inout,$roundkey);'. + ' &vmovdqu ($roundkey,"0xb0-0x80($inp)");', + + '&vpand ($iv,$temp,$mask10);'. + ' &vaesenc ($inout,$inout,$roundkey);'. + ' &vmovdqu ($roundkey,"0xc0-0x80($inp)");', + + '&vaesenclast ($temp,$inout,$roundkey);'. + ' &vaesenc ($inout,$inout,$roundkey);'. + ' &vmovdqu ($roundkey,"0xd0-0x80($inp)");', + + '&vpand ($temp,$temp,$mask12);'. + ' &vaesenc ($inout,$inout,$roundkey);'. + '&vmovdqu ($roundkey,"0xe0-0x80($inp)");', + + '&vpor ($iv,$iv,$temp);'. + ' &vaesenclast ($temp,$inout,$roundkey);'. + ' &vmovdqu ($roundkey,"0x00-0x80($inp)");' + +## &mov ($inp,$_inp); +## &mov ($out,$_out); +## &vpand ($temp,$temp,$mask14); +## &vpor ($iv,$iv,$temp); +## &vmovdqu ($iv,($out,$inp); +## &lea (inp,16($inp)); +); + +my $a4=$T1; +my ($a,$b,$c,$d,$e,$f,$g,$h); + +sub AUTOLOAD() # thunk [simplified] 32-bit style perlasm +{ my $opcode = $AUTOLOAD; $opcode =~ s/.*:://; + my $arg = pop; + $arg = "\$$arg" if ($arg*1 eq $arg); + $code .= "\t$opcode\t".join(',',$arg,reverse @_)."\n"; +} + +sub body_00_15 () { + ( + '($a,$b,$c,$d,$e,$f,$g,$h)=@ROT;'. + + '&ror ($a0,$Sigma1[2]-$Sigma1[1])', + '&mov ($a,$a1)', + '&mov ($a4,$f)', + + '&xor ($a0,$e)', + '&ror ($a1,$Sigma0[2]-$Sigma0[1])', + '&xor ($a4,$g)', # f^g + + '&ror ($a0,$Sigma1[1]-$Sigma1[0])', + '&xor ($a1,$a)', + '&and ($a4,$e)', # (f^g)&e + + @aesni_cbc_block[$aesni_cbc_idx++]. + '&xor ($a0,$e)', + '&add ($h,$SZ*($i&15)."(%rsp)")', # h+=X[i]+K[i] + '&mov ($a2,$a)', + + '&ror ($a1,$Sigma0[1]-$Sigma0[0])', + '&xor ($a4,$g)', # Ch(e,f,g)=((f^g)&e)^g + '&xor ($a2,$b)', # a^b, b^c in next round + + '&ror ($a0,$Sigma1[0])', # Sigma1(e) + '&add ($h,$a4)', # h+=Ch(e,f,g) + '&and ($a3,$a2)', # (b^c)&(a^b) + + '&xor ($a1,$a)', + '&add ($h,$a0)', # h+=Sigma1(e) + '&xor ($a3,$b)', # Maj(a,b,c)=Ch(a^b,c,b) + + '&add ($d,$h)', # d+=h + '&ror ($a1,$Sigma0[0])', # Sigma0(a) + '&add ($h,$a3)', # h+=Maj(a,b,c) + + '&mov ($a0,$d)', + '&add ($a1,$h);'. # h+=Sigma0(a) + '($a2,$a3) = ($a3,$a2); unshift(@ROT,pop(@ROT)); $i++;' + ); +} + +if ($avx) {{ +###################################################################### +# XOP code path +# +$code.=<<___; +.type ${func}_xop,\@function,6 +.align 64 +${func}_xop: +.Lxop_shortcut: + mov `($win64?56:8)`(%rsp),$in0 # load 7th parameter + push %rbx + push %rbp + push %r12 + push %r13 + push %r14 + push %r15 + mov %rsp,%r11 # copy %rsp + sub \$`$framesz+$win64*16*10`,%rsp + and \$-64,%rsp # align stack frame + + shl \$6,$len + sub $inp,$out # re-bias + sub $inp,$in0 + add $inp,$len # end of input + + #mov $inp,$_inp # saved later + mov $out,$_out + mov $len,$_end + #mov $key,$_key # remains resident in $inp register + mov $ivp,$_ivp + mov $ctx,$_ctx + mov $in0,$_in0 + mov %r11,$_rsp +___ +$code.=<<___ if ($win64); + movaps %xmm6,`$framesz+16*0`(%rsp) + movaps %xmm7,`$framesz+16*1`(%rsp) + movaps %xmm8,`$framesz+16*2`(%rsp) + movaps %xmm9,`$framesz+16*3`(%rsp) + movaps %xmm10,`$framesz+16*4`(%rsp) + movaps %xmm11,`$framesz+16*5`(%rsp) + movaps %xmm12,`$framesz+16*6`(%rsp) + movaps %xmm13,`$framesz+16*7`(%rsp) + movaps %xmm14,`$framesz+16*8`(%rsp) + movaps %xmm15,`$framesz+16*9`(%rsp) +___ +$code.=<<___; +.Lprologue_xop: + vzeroall + + mov $inp,%r12 # borrow $a4 + lea 0x80($key),$inp # size optimization, reassign + lea $TABLE+`$SZ*2*$rounds+32`(%rip),%r13 # borrow $a0 + mov 0xf0-0x80($inp),%r14d # rounds, borrow $a1 + mov $ctx,%r15 # borrow $a2 + mov $in0,%rsi # borrow $a3 + vmovdqu ($ivp),$iv # load IV + sub \$9,%r14 + + mov $SZ*0(%r15),$A + mov $SZ*1(%r15),$B + mov $SZ*2(%r15),$C + mov $SZ*3(%r15),$D + mov $SZ*4(%r15),$E + mov $SZ*5(%r15),$F + mov $SZ*6(%r15),$G + mov $SZ*7(%r15),$H + + vmovdqa 0x00(%r13,%r14,8),$mask14 + vmovdqa 0x10(%r13,%r14,8),$mask12 + vmovdqa 0x20(%r13,%r14,8),$mask10 + vmovdqu 0x00-0x80($inp),$roundkey + jmp .Lloop_xop +___ + if ($SZ==4) { # SHA256 + my @X = map("%xmm$_",(0..3)); + my ($t0,$t1,$t2,$t3) = map("%xmm$_",(4..7)); + +$code.=<<___; +.align 16 +.Lloop_xop: + vmovdqa $TABLE+`$SZ*2*$rounds`(%rip),$t3 + vmovdqu 0x00(%rsi,%r12),@X[0] + vmovdqu 0x10(%rsi,%r12),@X[1] + vmovdqu 0x20(%rsi,%r12),@X[2] + vmovdqu 0x30(%rsi,%r12),@X[3] + vpshufb $t3,@X[0],@X[0] + lea $TABLE(%rip),$Tbl + vpshufb $t3,@X[1],@X[1] + vpshufb $t3,@X[2],@X[2] + vpaddd 0x00($Tbl),@X[0],$t0 + vpshufb $t3,@X[3],@X[3] + vpaddd 0x20($Tbl),@X[1],$t1 + vpaddd 0x40($Tbl),@X[2],$t2 + vpaddd 0x60($Tbl),@X[3],$t3 + vmovdqa $t0,0x00(%rsp) + mov $A,$a1 + vmovdqa $t1,0x10(%rsp) + mov $B,$a3 + vmovdqa $t2,0x20(%rsp) + xor $C,$a3 # magic + vmovdqa $t3,0x30(%rsp) + mov $E,$a0 + jmp .Lxop_00_47 + +.align 16 +.Lxop_00_47: + sub \$-16*2*$SZ,$Tbl # size optimization + vmovdqu (%r12),$inout # $a4 + mov %r12,$_inp # $a4 +___ +sub XOP_256_00_47 () { +my $j = shift; +my $body = shift; +my @X = @_; +my @insns = (&$body,&$body,&$body,&$body); # 104 instructions + + &vpalignr ($t0,@X[1],@X[0],$SZ); # X[1..4] + eval(shift(@insns)); + eval(shift(@insns)); + &vpalignr ($t3,@X[3],@X[2],$SZ); # X[9..12] + eval(shift(@insns)); + eval(shift(@insns)); + &vprotd ($t1,$t0,8*$SZ-$sigma0[1]); + eval(shift(@insns)); + eval(shift(@insns)); + &vpsrld ($t0,$t0,$sigma0[2]); + eval(shift(@insns)); + eval(shift(@insns)); + &vpaddd (@X[0],@X[0],$t3); # X[0..3] += X[9..12] + eval(shift(@insns)); + eval(shift(@insns)); + eval(shift(@insns)); + eval(shift(@insns)); + &vprotd ($t2,$t1,$sigma0[1]-$sigma0[0]); + eval(shift(@insns)); + eval(shift(@insns)); + &vpxor ($t0,$t0,$t1); + eval(shift(@insns)); + eval(shift(@insns)); + eval(shift(@insns)); + eval(shift(@insns)); + &vprotd ($t3,@X[3],8*$SZ-$sigma1[1]); + eval(shift(@insns)); + eval(shift(@insns)); + &vpxor ($t0,$t0,$t2); # sigma0(X[1..4]) + eval(shift(@insns)); + eval(shift(@insns)); + &vpsrld ($t2,@X[3],$sigma1[2]); + eval(shift(@insns)); + eval(shift(@insns)); + &vpaddd (@X[0],@X[0],$t0); # X[0..3] += sigma0(X[1..4]) + eval(shift(@insns)); + eval(shift(@insns)); + &vprotd ($t1,$t3,$sigma1[1]-$sigma1[0]); + eval(shift(@insns)); + eval(shift(@insns)); + &vpxor ($t3,$t3,$t2); + eval(shift(@insns)); + eval(shift(@insns)); + eval(shift(@insns)); + eval(shift(@insns)); + &vpxor ($t3,$t3,$t1); # sigma1(X[14..15]) + eval(shift(@insns)); + eval(shift(@insns)); + eval(shift(@insns)); + eval(shift(@insns)); + &vpsrldq ($t3,$t3,8); + eval(shift(@insns)); + eval(shift(@insns)); + eval(shift(@insns)); + eval(shift(@insns)); + &vpaddd (@X[0],@X[0],$t3); # X[0..1] += sigma1(X[14..15]) + eval(shift(@insns)); + eval(shift(@insns)); + eval(shift(@insns)); + eval(shift(@insns)); + &vprotd ($t3,@X[0],8*$SZ-$sigma1[1]); + eval(shift(@insns)); + eval(shift(@insns)); + &vpsrld ($t2,@X[0],$sigma1[2]); + eval(shift(@insns)); + eval(shift(@insns)); + &vprotd ($t1,$t3,$sigma1[1]-$sigma1[0]); + eval(shift(@insns)); + eval(shift(@insns)); + &vpxor ($t3,$t3,$t2); + eval(shift(@insns)); + eval(shift(@insns)); + eval(shift(@insns)); + eval(shift(@insns)); + &vpxor ($t3,$t3,$t1); # sigma1(X[16..17]) + eval(shift(@insns)); + eval(shift(@insns)); + eval(shift(@insns)); + eval(shift(@insns)); + &vpslldq ($t3,$t3,8); # 22 instructions + eval(shift(@insns)); + eval(shift(@insns)); + eval(shift(@insns)); + eval(shift(@insns)); + &vpaddd (@X[0],@X[0],$t3); # X[2..3] += sigma1(X[16..17]) + eval(shift(@insns)); + eval(shift(@insns)); + eval(shift(@insns)); + eval(shift(@insns)); + &vpaddd ($t2,@X[0],16*2*$j."($Tbl)"); + foreach (@insns) { eval; } # remaining instructions + &vmovdqa (16*$j."(%rsp)",$t2); +} + + $aesni_cbc_idx=0; + for ($i=0,$j=0; $j<4; $j++) { + &XOP_256_00_47($j,\&body_00_15,@X); + push(@X,shift(@X)); # rotate(@X) + } + &mov ("%r12",$_inp); # borrow $a4 + &vpand ($temp,$temp,$mask14); + &mov ("%r15",$_out); # borrow $a2 + &vpor ($iv,$iv,$temp); + &vmovdqu ("(%r15,%r12)",$iv); # write output + &lea ("%r12","16(%r12)"); # inp++ + + &cmpb ($SZ-1+16*2*$SZ."($Tbl)",0); + &jne (".Lxop_00_47"); + + &vmovdqu ($inout,"(%r12)"); + &mov ($_inp,"%r12"); + + $aesni_cbc_idx=0; + for ($i=0; $i<16; ) { + foreach(body_00_15()) { eval; } + } + } +$code.=<<___; + mov $_inp,%r12 # borrow $a4 + mov $_out,%r13 # borrow $a0 + mov $_ctx,%r15 # borrow $a2 + mov $_in0,%rsi # borrow $a3 + + vpand $mask14,$temp,$temp + mov $a1,$A + vpor $temp,$iv,$iv + vmovdqu $iv,(%r13,%r12) # write output + lea 16(%r12),%r12 # inp++ + + add $SZ*0(%r15),$A + add $SZ*1(%r15),$B + add $SZ*2(%r15),$C + add $SZ*3(%r15),$D + add $SZ*4(%r15),$E + add $SZ*5(%r15),$F + add $SZ*6(%r15),$G + add $SZ*7(%r15),$H + + cmp $_end,%r12 + + mov $A,$SZ*0(%r15) + mov $B,$SZ*1(%r15) + mov $C,$SZ*2(%r15) + mov $D,$SZ*3(%r15) + mov $E,$SZ*4(%r15) + mov $F,$SZ*5(%r15) + mov $G,$SZ*6(%r15) + mov $H,$SZ*7(%r15) + + jb .Lloop_xop + + mov $_ivp,$ivp + mov $_rsp,%rsi + vmovdqu $iv,($ivp) # output IV + vzeroall +___ +$code.=<<___ if ($win64); + movaps `$framesz+16*0`(%rsp),%xmm6 + movaps `$framesz+16*1`(%rsp),%xmm7 + movaps `$framesz+16*2`(%rsp),%xmm8 + movaps `$framesz+16*3`(%rsp),%xmm9 + movaps `$framesz+16*4`(%rsp),%xmm10 + movaps `$framesz+16*5`(%rsp),%xmm11 + movaps `$framesz+16*6`(%rsp),%xmm12 + movaps `$framesz+16*7`(%rsp),%xmm13 + movaps `$framesz+16*8`(%rsp),%xmm14 + movaps `$framesz+16*9`(%rsp),%xmm15 +___ +$code.=<<___; + mov (%rsi),%r15 + mov 8(%rsi),%r14 + mov 16(%rsi),%r13 + mov 24(%rsi),%r12 + mov 32(%rsi),%rbp + mov 40(%rsi),%rbx + lea 48(%rsi),%rsp +.Lepilogue_xop: + ret +.size ${func}_xop,.-${func}_xop +___ +###################################################################### +# AVX+shrd code path +# +local *ror = sub { &shrd(@_[0],@_) }; + +$code.=<<___; +.type ${func}_avx,\@function,6 +.align 64 +${func}_avx: +.Lavx_shortcut: + mov `($win64?56:8)`(%rsp),$in0 # load 7th parameter + push %rbx + push %rbp + push %r12 + push %r13 + push %r14 + push %r15 + mov %rsp,%r11 # copy %rsp + sub \$`$framesz+$win64*16*10`,%rsp + and \$-64,%rsp # align stack frame + + shl \$6,$len + sub $inp,$out # re-bias + sub $inp,$in0 + add $inp,$len # end of input + + #mov $inp,$_inp # saved later + mov $out,$_out + mov $len,$_end + #mov $key,$_key # remains resident in $inp register + mov $ivp,$_ivp + mov $ctx,$_ctx + mov $in0,$_in0 + mov %r11,$_rsp +___ +$code.=<<___ if ($win64); + movaps %xmm6,`$framesz+16*0`(%rsp) + movaps %xmm7,`$framesz+16*1`(%rsp) + movaps %xmm8,`$framesz+16*2`(%rsp) + movaps %xmm9,`$framesz+16*3`(%rsp) + movaps %xmm10,`$framesz+16*4`(%rsp) + movaps %xmm11,`$framesz+16*5`(%rsp) + movaps %xmm12,`$framesz+16*6`(%rsp) + movaps %xmm13,`$framesz+16*7`(%rsp) + movaps %xmm14,`$framesz+16*8`(%rsp) + movaps %xmm15,`$framesz+16*9`(%rsp) +___ +$code.=<<___; +.Lprologue_avx: + vzeroall + + mov $inp,%r12 # borrow $a4 + lea 0x80($key),$inp # size optimization, reassign + lea $TABLE+`$SZ*2*$rounds+32`(%rip),%r13 # borrow $a0 + mov 0xf0-0x80($inp),%r14d # rounds, borrow $a1 + mov $ctx,%r15 # borrow $a2 + mov $in0,%rsi # borrow $a3 + vmovdqu ($ivp),$iv # load IV + sub \$9,%r14 + + mov $SZ*0(%r15),$A + mov $SZ*1(%r15),$B + mov $SZ*2(%r15),$C + mov $SZ*3(%r15),$D + mov $SZ*4(%r15),$E + mov $SZ*5(%r15),$F + mov $SZ*6(%r15),$G + mov $SZ*7(%r15),$H + + vmovdqa 0x00(%r13,%r14,8),$mask14 + vmovdqa 0x10(%r13,%r14,8),$mask12 + vmovdqa 0x20(%r13,%r14,8),$mask10 + vmovdqu 0x00-0x80($inp),$roundkey +___ + if ($SZ==4) { # SHA256 + my @X = map("%xmm$_",(0..3)); + my ($t0,$t1,$t2,$t3) = map("%xmm$_",(4..7)); + +$code.=<<___; + jmp .Lloop_avx +.align 16 +.Lloop_avx: + vmovdqa $TABLE+`$SZ*2*$rounds`(%rip),$t3 + vmovdqu 0x00(%rsi,%r12),@X[0] + vmovdqu 0x10(%rsi,%r12),@X[1] + vmovdqu 0x20(%rsi,%r12),@X[2] + vmovdqu 0x30(%rsi,%r12),@X[3] + vpshufb $t3,@X[0],@X[0] + lea $TABLE(%rip),$Tbl + vpshufb $t3,@X[1],@X[1] + vpshufb $t3,@X[2],@X[2] + vpaddd 0x00($Tbl),@X[0],$t0 + vpshufb $t3,@X[3],@X[3] + vpaddd 0x20($Tbl),@X[1],$t1 + vpaddd 0x40($Tbl),@X[2],$t2 + vpaddd 0x60($Tbl),@X[3],$t3 + vmovdqa $t0,0x00(%rsp) + mov $A,$a1 + vmovdqa $t1,0x10(%rsp) + mov $B,$a3 + vmovdqa $t2,0x20(%rsp) + xor $C,$a3 # magic + vmovdqa $t3,0x30(%rsp) + mov $E,$a0 + jmp .Lavx_00_47 + +.align 16 +.Lavx_00_47: + sub \$-16*2*$SZ,$Tbl # size optimization + vmovdqu (%r12),$inout # $a4 + mov %r12,$_inp # $a4 +___ +sub Xupdate_256_AVX () { + ( + '&vpalignr ($t0,@X[1],@X[0],$SZ)', # X[1..4] + '&vpalignr ($t3,@X[3],@X[2],$SZ)', # X[9..12] + '&vpsrld ($t2,$t0,$sigma0[0]);', + '&vpaddd (@X[0],@X[0],$t3)', # X[0..3] += X[9..12] + '&vpsrld ($t3,$t0,$sigma0[2])', + '&vpslld ($t1,$t0,8*$SZ-$sigma0[1]);', + '&vpxor ($t0,$t3,$t2)', + '&vpshufd ($t3,@X[3],0b11111010)',# X[14..15] + '&vpsrld ($t2,$t2,$sigma0[1]-$sigma0[0]);', + '&vpxor ($t0,$t0,$t1)', + '&vpslld ($t1,$t1,$sigma0[1]-$sigma0[0]);', + '&vpxor ($t0,$t0,$t2)', + '&vpsrld ($t2,$t3,$sigma1[2]);', + '&vpxor ($t0,$t0,$t1)', # sigma0(X[1..4]) + '&vpsrlq ($t3,$t3,$sigma1[0]);', + '&vpaddd (@X[0],@X[0],$t0)', # X[0..3] += sigma0(X[1..4]) + '&vpxor ($t2,$t2,$t3);', + '&vpsrlq ($t3,$t3,$sigma1[1]-$sigma1[0])', + '&vpxor ($t2,$t2,$t3)', # sigma1(X[14..15]) + '&vpshufd ($t2,$t2,0b10000100)', + '&vpsrldq ($t2,$t2,8)', + '&vpaddd (@X[0],@X[0],$t2)', # X[0..1] += sigma1(X[14..15]) + '&vpshufd ($t3,@X[0],0b01010000)',# X[16..17] + '&vpsrld ($t2,$t3,$sigma1[2])', + '&vpsrlq ($t3,$t3,$sigma1[0])', + '&vpxor ($t2,$t2,$t3);', + '&vpsrlq ($t3,$t3,$sigma1[1]-$sigma1[0])', + '&vpxor ($t2,$t2,$t3)', + '&vpshufd ($t2,$t2,0b11101000)', + '&vpslldq ($t2,$t2,8)', + '&vpaddd (@X[0],@X[0],$t2)' # X[2..3] += sigma1(X[16..17]) + ); +} + +sub AVX_256_00_47 () { +my $j = shift; +my $body = shift; +my @X = @_; +my @insns = (&$body,&$body,&$body,&$body); # 104 instructions + + foreach (Xupdate_256_AVX()) { # 29 instructions + eval; + eval(shift(@insns)); + eval(shift(@insns)); + eval(shift(@insns)); + } + &vpaddd ($t2,@X[0],16*2*$j."($Tbl)"); + foreach (@insns) { eval; } # remaining instructions + &vmovdqa (16*$j."(%rsp)",$t2); +} + + $aesni_cbc_idx=0; + for ($i=0,$j=0; $j<4; $j++) { + &AVX_256_00_47($j,\&body_00_15,@X); + push(@X,shift(@X)); # rotate(@X) + } + &mov ("%r12",$_inp); # borrow $a4 + &vpand ($temp,$temp,$mask14); + &mov ("%r15",$_out); # borrow $a2 + &vpor ($iv,$iv,$temp); + &vmovdqu ("(%r15,%r12)",$iv); # write output + &lea ("%r12","16(%r12)"); # inp++ + + &cmpb ($SZ-1+16*2*$SZ."($Tbl)",0); + &jne (".Lavx_00_47"); + + &vmovdqu ($inout,"(%r12)"); + &mov ($_inp,"%r12"); + + $aesni_cbc_idx=0; + for ($i=0; $i<16; ) { + foreach(body_00_15()) { eval; } + } + + } +$code.=<<___; + mov $_inp,%r12 # borrow $a4 + mov $_out,%r13 # borrow $a0 + mov $_ctx,%r15 # borrow $a2 + mov $_in0,%rsi # borrow $a3 + + vpand $mask14,$temp,$temp + mov $a1,$A + vpor $temp,$iv,$iv + vmovdqu $iv,(%r13,%r12) # write output + lea 16(%r12),%r12 # inp++ + + add $SZ*0(%r15),$A + add $SZ*1(%r15),$B + add $SZ*2(%r15),$C + add $SZ*3(%r15),$D + add $SZ*4(%r15),$E + add $SZ*5(%r15),$F + add $SZ*6(%r15),$G + add $SZ*7(%r15),$H + + cmp $_end,%r12 + + mov $A,$SZ*0(%r15) + mov $B,$SZ*1(%r15) + mov $C,$SZ*2(%r15) + mov $D,$SZ*3(%r15) + mov $E,$SZ*4(%r15) + mov $F,$SZ*5(%r15) + mov $G,$SZ*6(%r15) + mov $H,$SZ*7(%r15) + jb .Lloop_avx + + mov $_ivp,$ivp + mov $_rsp,%rsi + vmovdqu $iv,($ivp) # output IV + vzeroall +___ +$code.=<<___ if ($win64); + movaps `$framesz+16*0`(%rsp),%xmm6 + movaps `$framesz+16*1`(%rsp),%xmm7 + movaps `$framesz+16*2`(%rsp),%xmm8 + movaps `$framesz+16*3`(%rsp),%xmm9 + movaps `$framesz+16*4`(%rsp),%xmm10 + movaps `$framesz+16*5`(%rsp),%xmm11 + movaps `$framesz+16*6`(%rsp),%xmm12 + movaps `$framesz+16*7`(%rsp),%xmm13 + movaps `$framesz+16*8`(%rsp),%xmm14 + movaps `$framesz+16*9`(%rsp),%xmm15 +___ +$code.=<<___; + mov (%rsi),%r15 + mov 8(%rsi),%r14 + mov 16(%rsi),%r13 + mov 24(%rsi),%r12 + mov 32(%rsi),%rbp + mov 40(%rsi),%rbx + lea 48(%rsi),%rsp +.Lepilogue_avx: + ret +.size ${func}_avx,.-${func}_avx +___ + +if ($avx>1) {{ +###################################################################### +# AVX2+BMI code path +# +my $a5=$SZ==4?"%esi":"%rsi"; # zap $inp +my $PUSH8=8*2*$SZ; +use integer; + +sub bodyx_00_15 () { + # at start $a1 should be zero, $a3 - $b^$c and $a4 copy of $f + ( + '($a,$b,$c,$d,$e,$f,$g,$h)=@ROT;'. + + '&add ($h,(32*($i/(16/$SZ))+$SZ*($i%(16/$SZ)))%$PUSH8.$base)', # h+=X[i]+K[i] + '&and ($a4,$e)', # f&e + '&rorx ($a0,$e,$Sigma1[2])', + '&rorx ($a2,$e,$Sigma1[1])', + + '&lea ($a,"($a,$a1)")', # h+=Sigma0(a) from the past + '&lea ($h,"($h,$a4)")', + '&andn ($a4,$e,$g)', # ~e&g + '&xor ($a0,$a2)', + + '&rorx ($a1,$e,$Sigma1[0])', + '&lea ($h,"($h,$a4)")', # h+=Ch(e,f,g)=(e&f)+(~e&g) + '&xor ($a0,$a1)', # Sigma1(e) + '&mov ($a2,$a)', + + '&rorx ($a4,$a,$Sigma0[2])', + '&lea ($h,"($h,$a0)")', # h+=Sigma1(e) + '&xor ($a2,$b)', # a^b, b^c in next round + '&rorx ($a1,$a,$Sigma0[1])', + + '&rorx ($a0,$a,$Sigma0[0])', + '&lea ($d,"($d,$h)")', # d+=h + '&and ($a3,$a2)', # (b^c)&(a^b) + @aesni_cbc_block[$aesni_cbc_idx++]. + '&xor ($a1,$a4)', + + '&xor ($a3,$b)', # Maj(a,b,c)=Ch(a^b,c,b) + '&xor ($a1,$a0)', # Sigma0(a) + '&lea ($h,"($h,$a3)");'. # h+=Maj(a,b,c) + '&mov ($a4,$e)', # copy of f in future + + '($a2,$a3) = ($a3,$a2); unshift(@ROT,pop(@ROT)); $i++;' + ); + # and at the finish one has to $a+=$a1 +} + +$code.=<<___; +.type ${func}_avx2,\@function,6 +.align 64 +${func}_avx2: +.Lavx2_shortcut: + mov `($win64?56:8)`(%rsp),$in0 # load 7th parameter + push %rbx + push %rbp + push %r12 + push %r13 + push %r14 + push %r15 + mov %rsp,%r11 # copy %rsp + sub \$`2*$SZ*$rounds+8*8+$win64*16*10`,%rsp + and \$-256*$SZ,%rsp # align stack frame + add \$`2*$SZ*($rounds-8)`,%rsp + + shl \$6,$len + sub $inp,$out # re-bias + sub $inp,$in0 + add $inp,$len # end of input + + #mov $inp,$_inp # saved later + #mov $out,$_out # kept in $offload + mov $len,$_end + #mov $key,$_key # remains resident in $inp register + mov $ivp,$_ivp + mov $ctx,$_ctx + mov $in0,$_in0 + mov %r11,$_rsp +___ +$code.=<<___ if ($win64); + movaps %xmm6,`$framesz+16*0`(%rsp) + movaps %xmm7,`$framesz+16*1`(%rsp) + movaps %xmm8,`$framesz+16*2`(%rsp) + movaps %xmm9,`$framesz+16*3`(%rsp) + movaps %xmm10,`$framesz+16*4`(%rsp) + movaps %xmm11,`$framesz+16*5`(%rsp) + movaps %xmm12,`$framesz+16*6`(%rsp) + movaps %xmm13,`$framesz+16*7`(%rsp) + movaps %xmm14,`$framesz+16*8`(%rsp) + movaps %xmm15,`$framesz+16*9`(%rsp) +___ +$code.=<<___; +.Lprologue_avx2: + vzeroall + + mov $inp,%r13 # borrow $a0 + vpinsrq \$1,$out,$offload,$offload + lea 0x80($key),$inp # size optimization, reassign + lea $TABLE+`$SZ*2*$rounds+32`(%rip),%r12 # borrow $a4 + mov 0xf0-0x80($inp),%r14d # rounds, borrow $a1 + mov $ctx,%r15 # borrow $a2 + mov $in0,%rsi # borrow $a3 + vmovdqu ($ivp),$iv # load IV + lea -9(%r14),%r14 + + vmovdqa 0x00(%r12,%r14,8),$mask14 + vmovdqa 0x10(%r12,%r14,8),$mask12 + vmovdqa 0x20(%r12,%r14,8),$mask10 + + sub \$-16*$SZ,%r13 # inp++, size optimization + mov $SZ*0(%r15),$A + lea (%rsi,%r13),%r12 # borrow $a0 + mov $SZ*1(%r15),$B + cmp $len,%r13 # $_end + mov $SZ*2(%r15),$C + cmove %rsp,%r12 # next block or random data + mov $SZ*3(%r15),$D + mov $SZ*4(%r15),$E + mov $SZ*5(%r15),$F + mov $SZ*6(%r15),$G + mov $SZ*7(%r15),$H + vmovdqu 0x00-0x80($inp),$roundkey +___ + if ($SZ==4) { # SHA256 + my @X = map("%ymm$_",(0..3)); + my ($t0,$t1,$t2,$t3) = map("%ymm$_",(4..7)); + +$code.=<<___; + jmp .Loop_avx2 +.align 16 +.Loop_avx2: + vmovdqa $TABLE+`$SZ*2*$rounds`(%rip),$t3 + vmovdqu -16*$SZ+0(%rsi,%r13),%xmm0 + vmovdqu -16*$SZ+16(%rsi,%r13),%xmm1 + vmovdqu -16*$SZ+32(%rsi,%r13),%xmm2 + vmovdqu -16*$SZ+48(%rsi,%r13),%xmm3 + + vinserti128 \$1,(%r12),@X[0],@X[0] + vinserti128 \$1,16(%r12),@X[1],@X[1] + vpshufb $t3,@X[0],@X[0] + vinserti128 \$1,32(%r12),@X[2],@X[2] + vpshufb $t3,@X[1],@X[1] + vinserti128 \$1,48(%r12),@X[3],@X[3] + + lea $TABLE(%rip),$Tbl + vpshufb $t3,@X[2],@X[2] + lea -16*$SZ(%r13),%r13 + vpaddd 0x00($Tbl),@X[0],$t0 + vpshufb $t3,@X[3],@X[3] + vpaddd 0x20($Tbl),@X[1],$t1 + vpaddd 0x40($Tbl),@X[2],$t2 + vpaddd 0x60($Tbl),@X[3],$t3 + vmovdqa $t0,0x00(%rsp) + xor $a1,$a1 + vmovdqa $t1,0x20(%rsp) + lea -$PUSH8(%rsp),%rsp + mov $B,$a3 + vmovdqa $t2,0x00(%rsp) + xor $C,$a3 # magic + vmovdqa $t3,0x20(%rsp) + mov $F,$a4 + sub \$-16*2*$SZ,$Tbl # size optimization + jmp .Lavx2_00_47 + +.align 16 +.Lavx2_00_47: + vmovdqu (%r13),$inout + vpinsrq \$0,%r13,$offload,$offload +___ + +sub AVX2_256_00_47 () { +my $j = shift; +my $body = shift; +my @X = @_; +my @insns = (&$body,&$body,&$body,&$body); # 96 instructions +my $base = "+2*$PUSH8(%rsp)"; + + &lea ("%rsp","-$PUSH8(%rsp)") if (($j%2)==0); + foreach (Xupdate_256_AVX()) { # 29 instructions + eval; + eval(shift(@insns)); + eval(shift(@insns)); + eval(shift(@insns)); + } + &vpaddd ($t2,@X[0],16*2*$j."($Tbl)"); + foreach (@insns) { eval; } # remaining instructions + &vmovdqa ((32*$j)%$PUSH8."(%rsp)",$t2); +} + $aesni_cbc_idx=0; + for ($i=0,$j=0; $j<4; $j++) { + &AVX2_256_00_47($j,\&bodyx_00_15,@X); + push(@X,shift(@X)); # rotate(@X) + } + &vmovq ("%r13",$offload); # borrow $a0 + &vpextrq ("%r15",$offload,1); # borrow $a2 + &vpand ($temp,$temp,$mask14); + &vpor ($iv,$iv,$temp); + &vmovdqu ("(%r15,%r13)",$iv); # write output + &lea ("%r13","16(%r13)"); # inp++ + + &lea ($Tbl,16*2*$SZ."($Tbl)"); + &cmpb (($SZ-1)."($Tbl)",0); + &jne (".Lavx2_00_47"); + + &vmovdqu ($inout,"(%r13)"); + &vpinsrq ($offload,$offload,"%r13",0); + + $aesni_cbc_idx=0; + for ($i=0; $i<16; ) { + my $base=$i<8?"+$PUSH8(%rsp)":"(%rsp)"; + foreach(bodyx_00_15()) { eval; } + } + } +$code.=<<___; + vpextrq \$1,$offload,%r12 # $_out, borrow $a4 + vmovq $offload,%r13 # $_inp, borrow $a0 + mov `2*$SZ*$rounds+5*8`(%rsp),%r15 # $_ctx, borrow $a2 + add $a1,$A + lea `2*$SZ*($rounds-8)`(%rsp),$Tbl + + vpand $mask14,$temp,$temp + vpor $temp,$iv,$iv + vmovdqu $iv,(%r12,%r13) # write output + lea 16(%r13),%r13 + + add $SZ*0(%r15),$A + add $SZ*1(%r15),$B + add $SZ*2(%r15),$C + add $SZ*3(%r15),$D + add $SZ*4(%r15),$E + add $SZ*5(%r15),$F + add $SZ*6(%r15),$G + add $SZ*7(%r15),$H + + mov $A,$SZ*0(%r15) + mov $B,$SZ*1(%r15) + mov $C,$SZ*2(%r15) + mov $D,$SZ*3(%r15) + mov $E,$SZ*4(%r15) + mov $F,$SZ*5(%r15) + mov $G,$SZ*6(%r15) + mov $H,$SZ*7(%r15) + + cmp `$PUSH8+2*8`($Tbl),%r13 # $_end + je .Ldone_avx2 + + xor $a1,$a1 + mov $B,$a3 + mov $F,$a4 + xor $C,$a3 # magic + jmp .Lower_avx2 +.align 16 +.Lower_avx2: + vmovdqu (%r13),$inout + vpinsrq \$0,%r13,$offload,$offload +___ + $aesni_cbc_idx=0; + for ($i=0; $i<16; ) { + my $base="+16($Tbl)"; + foreach(bodyx_00_15()) { eval; } + &lea ($Tbl,"-$PUSH8($Tbl)") if ($i==8); + } +$code.=<<___; + vmovq $offload,%r13 # borrow $a0 + vpextrq \$1,$offload,%r15 # borrow $a2 + vpand $mask14,$temp,$temp + vpor $temp,$iv,$iv + lea -$PUSH8($Tbl),$Tbl + vmovdqu $iv,(%r15,%r13) # write output + lea 16(%r13),%r13 # inp++ + cmp %rsp,$Tbl + jae .Lower_avx2 + + mov `2*$SZ*$rounds+5*8`(%rsp),%r15 # $_ctx, borrow $a2 + lea 16*$SZ(%r13),%r13 + mov `2*$SZ*$rounds+6*8`(%rsp),%rsi # $_in0, borrow $a3 + add $a1,$A + lea `2*$SZ*($rounds-8)`(%rsp),%rsp + + add $SZ*0(%r15),$A + add $SZ*1(%r15),$B + add $SZ*2(%r15),$C + add $SZ*3(%r15),$D + add $SZ*4(%r15),$E + add $SZ*5(%r15),$F + add $SZ*6(%r15),$G + lea (%rsi,%r13),%r12 + add $SZ*7(%r15),$H + + cmp $_end,%r13 + + mov $A,$SZ*0(%r15) + cmove %rsp,%r12 # next block or stale data + mov $B,$SZ*1(%r15) + mov $C,$SZ*2(%r15) + mov $D,$SZ*3(%r15) + mov $E,$SZ*4(%r15) + mov $F,$SZ*5(%r15) + mov $G,$SZ*6(%r15) + mov $H,$SZ*7(%r15) + + jbe .Loop_avx2 + lea (%rsp),$Tbl + +.Ldone_avx2: + lea ($Tbl),%rsp + mov $_ivp,$ivp + mov $_rsp,%rsi + vmovdqu $iv,($ivp) # output IV + vzeroall +___ +$code.=<<___ if ($win64); + movaps `$framesz+16*0`(%rsp),%xmm6 + movaps `$framesz+16*1`(%rsp),%xmm7 + movaps `$framesz+16*2`(%rsp),%xmm8 + movaps `$framesz+16*3`(%rsp),%xmm9 + movaps `$framesz+16*4`(%rsp),%xmm10 + movaps `$framesz+16*5`(%rsp),%xmm11 + movaps `$framesz+16*6`(%rsp),%xmm12 + movaps `$framesz+16*7`(%rsp),%xmm13 + movaps `$framesz+16*8`(%rsp),%xmm14 + movaps `$framesz+16*9`(%rsp),%xmm15 +___ +$code.=<<___; + mov (%rsi),%r15 + mov 8(%rsi),%r14 + mov 16(%rsi),%r13 + mov 24(%rsi),%r12 + mov 32(%rsi),%rbp + mov 40(%rsi),%rbx + lea 48(%rsi),%rsp +.Lepilogue_avx2: + ret +.size ${func}_avx2,.-${func}_avx2 +___ +}} +}} +{{ +my ($in0,$out,$len,$key,$ivp,$ctx,$inp)=("%rdi","%rsi","%rdx","%rcx","%r8","%r9","%r10"); + +my ($rounds,$Tbl)=("%r11d","%rbx"); + +my ($iv,$in,$rndkey0)=map("%xmm$_",(6,14,15)); +my @rndkey=("%xmm4","%xmm5"); +my $r=0; +my $sn=0; + +my ($Wi,$ABEF,$CDGH,$TMP,$BSWAP,$ABEF_SAVE,$CDGH_SAVE)=map("%xmm$_",(0..3,7..9)); +my @MSG=map("%xmm$_",(10..13)); + +my $aesenc=sub { + use integer; + my ($n,$k)=($r/10,$r%10); + if ($k==0) { + $code.=<<___; + movups `16*$n`($in0),$in # load input + xorps $rndkey0,$in +___ + $code.=<<___ if ($n); + movups $iv,`16*($n-1)`($out,$in0) # write output +___ + $code.=<<___; + xorps $in,$iv + movups `32+16*$k-112`($key),$rndkey[1] + aesenc $rndkey[0],$iv +___ + } elsif ($k==9) { + $sn++; + $code.=<<___; + cmp \$11,$rounds + jb .Laesenclast$sn + movups `32+16*($k+0)-112`($key),$rndkey[1] + aesenc $rndkey[0],$iv + movups `32+16*($k+1)-112`($key),$rndkey[0] + aesenc $rndkey[1],$iv + je .Laesenclast$sn + movups `32+16*($k+2)-112`($key),$rndkey[1] + aesenc $rndkey[0],$iv + movups `32+16*($k+3)-112`($key),$rndkey[0] + aesenc $rndkey[1],$iv +.Laesenclast$sn: + aesenclast $rndkey[0],$iv + movups 16-112($key),$rndkey[1] # forward reference + nop +___ + } else { + $code.=<<___; + movups `32+16*$k-112`($key),$rndkey[1] + aesenc $rndkey[0],$iv +___ + } + $r++; unshift(@rndkey,pop(@rndkey)); +}; + +if ($shaext) { +my $Tbl="%rax"; + +$code.=<<___; +.type ${func}_shaext,\@function,6 +.align 32 +${func}_shaext: + mov `($win64?56:8)`(%rsp),$inp # load 7th argument +___ +$code.=<<___ if ($win64); + lea `-8-10*16`(%rsp),%rsp + movaps %xmm6,-8-10*16(%rax) + movaps %xmm7,-8-9*16(%rax) + movaps %xmm8,-8-8*16(%rax) + movaps %xmm9,-8-7*16(%rax) + movaps %xmm10,-8-6*16(%rax) + movaps %xmm11,-8-5*16(%rax) + movaps %xmm12,-8-4*16(%rax) + movaps %xmm13,-8-3*16(%rax) + movaps %xmm14,-8-2*16(%rax) + movaps %xmm15,-8-1*16(%rax) +.Lprologue_shaext: +___ +$code.=<<___; + lea K256+0x80(%rip),$Tbl + movdqu ($ctx),$ABEF # DCBA + movdqu 16($ctx),$CDGH # HGFE + movdqa 0x200-0x80($Tbl),$TMP # byte swap mask + + mov 240($key),$rounds + sub $in0,$out + movups ($key),$rndkey0 # $key[0] + movups ($ivp),$iv # load IV + movups 16($key),$rndkey[0] # forward reference + lea 112($key),$key # size optimization + + pshufd \$0x1b,$ABEF,$Wi # ABCD + pshufd \$0xb1,$ABEF,$ABEF # CDAB + pshufd \$0x1b,$CDGH,$CDGH # EFGH + movdqa $TMP,$BSWAP # offload + palignr \$8,$CDGH,$ABEF # ABEF + punpcklqdq $Wi,$CDGH # CDGH + + jmp .Loop_shaext + +.align 16 +.Loop_shaext: + movdqu ($inp),@MSG[0] + movdqu 0x10($inp),@MSG[1] + movdqu 0x20($inp),@MSG[2] + pshufb $TMP,@MSG[0] + movdqu 0x30($inp),@MSG[3] + + movdqa 0*32-0x80($Tbl),$Wi + paddd @MSG[0],$Wi + pshufb $TMP,@MSG[1] + movdqa $CDGH,$CDGH_SAVE # offload + movdqa $ABEF,$ABEF_SAVE # offload +___ + &$aesenc(); +$code.=<<___; + sha256rnds2 $ABEF,$CDGH # 0-3 + pshufd \$0x0e,$Wi,$Wi +___ + &$aesenc(); +$code.=<<___; + sha256rnds2 $CDGH,$ABEF + + movdqa 1*32-0x80($Tbl),$Wi + paddd @MSG[1],$Wi + pshufb $TMP,@MSG[2] + lea 0x40($inp),$inp +___ + &$aesenc(); +$code.=<<___; + sha256rnds2 $ABEF,$CDGH # 4-7 + pshufd \$0x0e,$Wi,$Wi +___ + &$aesenc(); +$code.=<<___; + sha256rnds2 $CDGH,$ABEF + + movdqa 2*32-0x80($Tbl),$Wi + paddd @MSG[2],$Wi + pshufb $TMP,@MSG[3] + sha256msg1 @MSG[1],@MSG[0] +___ + &$aesenc(); +$code.=<<___; + sha256rnds2 $ABEF,$CDGH # 8-11 + pshufd \$0x0e,$Wi,$Wi + movdqa @MSG[3],$TMP + palignr \$4,@MSG[2],$TMP + paddd $TMP,@MSG[0] +___ + &$aesenc(); +$code.=<<___; + sha256rnds2 $CDGH,$ABEF + + movdqa 3*32-0x80($Tbl),$Wi + paddd @MSG[3],$Wi + sha256msg2 @MSG[3],@MSG[0] + sha256msg1 @MSG[2],@MSG[1] +___ + &$aesenc(); +$code.=<<___; + sha256rnds2 $ABEF,$CDGH # 12-15 + pshufd \$0x0e,$Wi,$Wi +___ + &$aesenc(); +$code.=<<___; + movdqa @MSG[0],$TMP + palignr \$4,@MSG[3],$TMP + paddd $TMP,@MSG[1] + sha256rnds2 $CDGH,$ABEF +___ +for($i=4;$i<16-3;$i++) { + &$aesenc() if (($r%10)==0); +$code.=<<___; + movdqa $i*32-0x80($Tbl),$Wi + paddd @MSG[0],$Wi + sha256msg2 @MSG[0],@MSG[1] + sha256msg1 @MSG[3],@MSG[2] +___ + &$aesenc(); +$code.=<<___; + sha256rnds2 $ABEF,$CDGH # 16-19... + pshufd \$0x0e,$Wi,$Wi + movdqa @MSG[1],$TMP + palignr \$4,@MSG[0],$TMP + paddd $TMP,@MSG[2] +___ + &$aesenc(); + &$aesenc() if ($r==19); +$code.=<<___; + sha256rnds2 $CDGH,$ABEF +___ + push(@MSG,shift(@MSG)); +} +$code.=<<___; + movdqa 13*32-0x80($Tbl),$Wi + paddd @MSG[0],$Wi + sha256msg2 @MSG[0],@MSG[1] + sha256msg1 @MSG[3],@MSG[2] +___ + &$aesenc(); +$code.=<<___; + sha256rnds2 $ABEF,$CDGH # 52-55 + pshufd \$0x0e,$Wi,$Wi + movdqa @MSG[1],$TMP + palignr \$4,@MSG[0],$TMP + paddd $TMP,@MSG[2] +___ + &$aesenc(); + &$aesenc(); +$code.=<<___; + sha256rnds2 $CDGH,$ABEF + + movdqa 14*32-0x80($Tbl),$Wi + paddd @MSG[1],$Wi + sha256msg2 @MSG[1],@MSG[2] + movdqa $BSWAP,$TMP +___ + &$aesenc(); +$code.=<<___; + sha256rnds2 $ABEF,$CDGH # 56-59 + pshufd \$0x0e,$Wi,$Wi +___ + &$aesenc(); +$code.=<<___; + sha256rnds2 $CDGH,$ABEF + + movdqa 15*32-0x80($Tbl),$Wi + paddd @MSG[2],$Wi +___ + &$aesenc(); + &$aesenc(); +$code.=<<___; + sha256rnds2 $ABEF,$CDGH # 60-63 + pshufd \$0x0e,$Wi,$Wi +___ + &$aesenc(); +$code.=<<___; + sha256rnds2 $CDGH,$ABEF + #pxor $CDGH,$rndkey0 # black magic +___ + while ($r<40) { &$aesenc(); } # remaining aesenc's +$code.=<<___; + #xorps $CDGH,$rndkey0 # black magic + paddd $CDGH_SAVE,$CDGH + paddd $ABEF_SAVE,$ABEF + + dec $len + movups $iv,48($out,$in0) # write output + lea 64($in0),$in0 + jnz .Loop_shaext + + pshufd \$0xb1,$CDGH,$CDGH # DCHG + pshufd \$0x1b,$ABEF,$TMP # FEBA + pshufd \$0xb1,$ABEF,$ABEF # BAFE + punpckhqdq $CDGH,$ABEF # DCBA + palignr \$8,$TMP,$CDGH # HGFE + + movups $iv,($ivp) # write IV + movdqu $ABEF,($ctx) + movdqu $CDGH,16($ctx) +___ +$code.=<<___ if ($win64); + movaps 0*16(%rsp),%xmm6 + movaps 1*16(%rsp),%xmm7 + movaps 2*16(%rsp),%xmm8 + movaps 3*16(%rsp),%xmm9 + movaps 4*16(%rsp),%xmm10 + movaps 5*16(%rsp),%xmm11 + movaps 6*16(%rsp),%xmm12 + movaps 7*16(%rsp),%xmm13 + movaps 8*16(%rsp),%xmm14 + movaps 9*16(%rsp),%xmm15 + lea 8+10*16(%rsp),%rsp +.Lepilogue_shaext: +___ +$code.=<<___; + ret +.size ${func}_shaext,.-${func}_shaext +___ +} +}}}}} + +# EXCEPTION_DISPOSITION handler (EXCEPTION_RECORD *rec,ULONG64 frame, +# CONTEXT *context,DISPATCHER_CONTEXT *disp) +if ($win64 && $avx) { +$rec="%rcx"; +$frame="%rdx"; +$context="%r8"; +$disp="%r9"; + +$code.=<<___; +.extern __imp_RtlVirtualUnwind +.type se_handler,\@abi-omnipotent +.align 16 +se_handler: + push %rsi + push %rdi + push %rbx + push %rbp + push %r12 + push %r13 + push %r14 + push %r15 + pushfq + sub \$64,%rsp + + mov 120($context),%rax # pull context->Rax + mov 248($context),%rbx # pull context->Rip + + mov 8($disp),%rsi # disp->ImageBase + mov 56($disp),%r11 # disp->HanderlData + + mov 0(%r11),%r10d # HandlerData[0] + lea (%rsi,%r10),%r10 # prologue label + cmp %r10,%rbx # context->RipRsp + + mov 4(%r11),%r10d # HandlerData[1] + lea (%rsi,%r10),%r10 # epilogue label + cmp %r10,%rbx # context->Rip>=epilogue label + jae .Lin_prologue +___ +$code.=<<___ if ($shaext); + lea aesni_cbc_sha256_enc_shaext(%rip),%r10 + cmp %r10,%rbx + jb .Lnot_in_shaext + + lea (%rax),%rsi + lea 512($context),%rdi # &context.Xmm6 + mov \$20,%ecx + .long 0xa548f3fc # cld; rep movsq + lea 168(%rax),%rax # adjust stack pointer + jmp .Lin_prologue +.Lnot_in_shaext: +___ +$code.=<<___ if ($avx>1); + lea .Lavx2_shortcut(%rip),%r10 + cmp %r10,%rbx # context->RipRbx + mov %rbp,160($context) # restore context->Rbp + mov %r12,216($context) # restore context->R12 + mov %r13,224($context) # restore context->R13 + mov %r14,232($context) # restore context->R14 + mov %r15,240($context) # restore context->R15 + + lea 16*$SZ+8*8(%rsi),%rsi # Xmm6- save area + lea 512($context),%rdi # &context.Xmm6 + mov \$20,%ecx + .long 0xa548f3fc # cld; rep movsq + +.Lin_prologue: + mov 8(%rax),%rdi + mov 16(%rax),%rsi + mov %rax,152($context) # restore context->Rsp + mov %rsi,168($context) # restore context->Rsi + mov %rdi,176($context) # restore context->Rdi + + mov 40($disp),%rdi # disp->ContextRecord + mov $context,%rsi # context + mov \$154,%ecx # sizeof(CONTEXT) + .long 0xa548f3fc # cld; rep movsq + + mov $disp,%rsi + xor %rcx,%rcx # arg1, UNW_FLAG_NHANDLER + mov 8(%rsi),%rdx # arg2, disp->ImageBase + mov 0(%rsi),%r8 # arg3, disp->ControlPc + mov 16(%rsi),%r9 # arg4, disp->FunctionEntry + mov 40(%rsi),%r10 # disp->ContextRecord + lea 56(%rsi),%r11 # &disp->HandlerData + lea 24(%rsi),%r12 # &disp->EstablisherFrame + mov %r10,32(%rsp) # arg5 + mov %r11,40(%rsp) # arg6 + mov %r12,48(%rsp) # arg7 + mov %rcx,56(%rsp) # arg8, (NULL) + call *__imp_RtlVirtualUnwind(%rip) + + mov \$1,%eax # ExceptionContinueSearch + add \$64,%rsp + popfq + pop %r15 + pop %r14 + pop %r13 + pop %r12 + pop %rbp + pop %rbx + pop %rdi + pop %rsi + ret +.size se_handler,.-se_handler + +.section .pdata + .rva .LSEH_begin_${func}_xop + .rva .LSEH_end_${func}_xop + .rva .LSEH_info_${func}_xop + + .rva .LSEH_begin_${func}_avx + .rva .LSEH_end_${func}_avx + .rva .LSEH_info_${func}_avx +___ +$code.=<<___ if ($avx>1); + .rva .LSEH_begin_${func}_avx2 + .rva .LSEH_end_${func}_avx2 + .rva .LSEH_info_${func}_avx2 +___ +$code.=<<___ if ($shaext); + .rva .LSEH_begin_${func}_shaext + .rva .LSEH_end_${func}_shaext + .rva .LSEH_info_${func}_shaext +___ +$code.=<<___; +.section .xdata +.align 8 +.LSEH_info_${func}_xop: + .byte 9,0,0,0 + .rva se_handler + .rva .Lprologue_xop,.Lepilogue_xop # HandlerData[] + +.LSEH_info_${func}_avx: + .byte 9,0,0,0 + .rva se_handler + .rva .Lprologue_avx,.Lepilogue_avx # HandlerData[] +___ +$code.=<<___ if ($avx>1); +.LSEH_info_${func}_avx2: + .byte 9,0,0,0 + .rva se_handler + .rva .Lprologue_avx2,.Lepilogue_avx2 # HandlerData[] +___ +$code.=<<___ if ($shaext); +.LSEH_info_${func}_shaext: + .byte 9,0,0,0 + .rva se_handler + .rva .Lprologue_shaext,.Lepilogue_shaext # HandlerData[] +___ +} + +#################################################################### +sub rex { + local *opcode=shift; + my ($dst,$src)=@_; + my $rex=0; + + $rex|=0x04 if($dst>=8); + $rex|=0x01 if($src>=8); + unshift @opcode,$rex|0x40 if($rex); +} + +{ + my %opcodelet = ( + "sha256rnds2" => 0xcb, + "sha256msg1" => 0xcc, + "sha256msg2" => 0xcd ); + + sub sha256op38 { + my $instr = shift; + + if (defined($opcodelet{$instr}) && @_[0] =~ /%xmm([0-9]+),\s*%xmm([0-9]+)/) { + my @opcode=(0x0f,0x38); + rex(\@opcode,$2,$1); + push @opcode,$opcodelet{$instr}; + push @opcode,0xc0|($1&7)|(($2&7)<<3); # ModR/M + return ".byte\t".join(',',@opcode); + } else { + return $instr."\t".@_[0]; + } + } +} + +$code =~ s/\`([^\`]*)\`/eval $1/gem; +$code =~ s/\b(sha256[^\s]*)\s+(.*)/sha256op38($1,$2)/gem; +print $code; +close STDOUT; diff --git a/openssl-1.1.0h/crypto/aes/asm/aesni-x86.pl b/openssl-1.1.0h/crypto/aes/asm/aesni-x86.pl new file mode 100644 index 0000000..ed1a47c --- /dev/null +++ b/openssl-1.1.0h/crypto/aes/asm/aesni-x86.pl @@ -0,0 +1,3413 @@ +#! /usr/bin/env perl +# Copyright 2009-2016 The OpenSSL Project Authors. All Rights Reserved. +# +# Licensed under the OpenSSL license (the "License"). You may not use +# this file except in compliance with the License. You can obtain a copy +# in the file LICENSE in the source distribution or at +# https://www.openssl.org/source/license.html + + +# ==================================================================== +# Written by Andy Polyakov for the OpenSSL +# project. The module is, however, dual licensed under OpenSSL and +# CRYPTOGAMS licenses depending on where you obtain it. For further +# details see http://www.openssl.org/~appro/cryptogams/. +# ==================================================================== +# +# This module implements support for Intel AES-NI extension. In +# OpenSSL context it's used with Intel engine, but can also be used as +# drop-in replacement for crypto/aes/asm/aes-586.pl [see below for +# details]. +# +# Performance. +# +# To start with see corresponding paragraph in aesni-x86_64.pl... +# Instead of filling table similar to one found there I've chosen to +# summarize *comparison* results for raw ECB, CTR and CBC benchmarks. +# The simplified table below represents 32-bit performance relative +# to 64-bit one in every given point. Ratios vary for different +# encryption modes, therefore interval values. +# +# 16-byte 64-byte 256-byte 1-KB 8-KB +# 53-67% 67-84% 91-94% 95-98% 97-99.5% +# +# Lower ratios for smaller block sizes are perfectly understandable, +# because function call overhead is higher in 32-bit mode. Largest +# 8-KB block performance is virtually same: 32-bit code is less than +# 1% slower for ECB, CBC and CCM, and ~3% slower otherwise. + +# January 2011 +# +# See aesni-x86_64.pl for details. Unlike x86_64 version this module +# interleaves at most 6 aes[enc|dec] instructions, because there are +# not enough registers for 8x interleave [which should be optimal for +# Sandy Bridge]. Actually, performance results for 6x interleave +# factor presented in aesni-x86_64.pl (except for CTR) are for this +# module. + +# April 2011 +# +# Add aesni_xts_[en|de]crypt. Westmere spends 1.50 cycles processing +# one byte out of 8KB with 128-bit key, Sandy Bridge - 1.09. + +# November 2015 +# +# Add aesni_ocb_[en|de]crypt. + +###################################################################### +# Current large-block performance in cycles per byte processed with +# 128-bit key (less is better). +# +# CBC en-/decrypt CTR XTS ECB OCB +# Westmere 3.77/1.37 1.37 1.52 1.27 +# * Bridge 5.07/0.98 0.99 1.09 0.91 1.10 +# Haswell 4.44/0.80 0.97 1.03 0.72 0.76 +# Silvermont 5.77/3.56 3.67 4.03 3.46 4.03 +# Bulldozer 5.80/0.98 1.05 1.24 0.93 1.23 + +$PREFIX="aesni"; # if $PREFIX is set to "AES", the script + # generates drop-in replacement for + # crypto/aes/asm/aes-586.pl:-) +$inline=1; # inline _aesni_[en|de]crypt + +$0 =~ m/(.*[\/\\])[^\/\\]+$/; $dir=$1; +push(@INC,"${dir}","${dir}../../perlasm"); +require "x86asm.pl"; + +$output = pop; +open OUT,">$output"; +*STDOUT=*OUT; + +&asm_init($ARGV[0],$0); + +&external_label("OPENSSL_ia32cap_P"); +&static_label("key_const"); + +if ($PREFIX eq "aesni") { $movekey=\&movups; } +else { $movekey=\&movups; } + +$len="eax"; +$rounds="ecx"; +$key="edx"; +$inp="esi"; +$out="edi"; +$rounds_="ebx"; # backup copy for $rounds +$key_="ebp"; # backup copy for $key + +$rndkey0="xmm0"; +$rndkey1="xmm1"; +$inout0="xmm2"; +$inout1="xmm3"; +$inout2="xmm4"; +$inout3="xmm5"; $in1="xmm5"; +$inout4="xmm6"; $in0="xmm6"; +$inout5="xmm7"; $ivec="xmm7"; + +# AESNI extension +sub aeskeygenassist +{ my($dst,$src,$imm)=@_; + if ("$dst:$src" =~ /xmm([0-7]):xmm([0-7])/) + { &data_byte(0x66,0x0f,0x3a,0xdf,0xc0|($1<<3)|$2,$imm); } +} +sub aescommon +{ my($opcodelet,$dst,$src)=@_; + if ("$dst:$src" =~ /xmm([0-7]):xmm([0-7])/) + { &data_byte(0x66,0x0f,0x38,$opcodelet,0xc0|($1<<3)|$2);} +} +sub aesimc { aescommon(0xdb,@_); } +sub aesenc { aescommon(0xdc,@_); } +sub aesenclast { aescommon(0xdd,@_); } +sub aesdec { aescommon(0xde,@_); } +sub aesdeclast { aescommon(0xdf,@_); } + +# Inline version of internal aesni_[en|de]crypt1 +{ my $sn; +sub aesni_inline_generate1 +{ my ($p,$inout,$ivec)=@_; $inout=$inout0 if (!defined($inout)); + $sn++; + + &$movekey ($rndkey0,&QWP(0,$key)); + &$movekey ($rndkey1,&QWP(16,$key)); + &xorps ($ivec,$rndkey0) if (defined($ivec)); + &lea ($key,&DWP(32,$key)); + &xorps ($inout,$ivec) if (defined($ivec)); + &xorps ($inout,$rndkey0) if (!defined($ivec)); + &set_label("${p}1_loop_$sn"); + eval"&aes${p} ($inout,$rndkey1)"; + &dec ($rounds); + &$movekey ($rndkey1,&QWP(0,$key)); + &lea ($key,&DWP(16,$key)); + &jnz (&label("${p}1_loop_$sn")); + eval"&aes${p}last ($inout,$rndkey1)"; +}} + +sub aesni_generate1 # fully unrolled loop +{ my ($p,$inout)=@_; $inout=$inout0 if (!defined($inout)); + + &function_begin_B("_aesni_${p}rypt1"); + &movups ($rndkey0,&QWP(0,$key)); + &$movekey ($rndkey1,&QWP(0x10,$key)); + &xorps ($inout,$rndkey0); + &$movekey ($rndkey0,&QWP(0x20,$key)); + &lea ($key,&DWP(0x30,$key)); + &cmp ($rounds,11); + &jb (&label("${p}128")); + &lea ($key,&DWP(0x20,$key)); + &je (&label("${p}192")); + &lea ($key,&DWP(0x20,$key)); + eval"&aes${p} ($inout,$rndkey1)"; + &$movekey ($rndkey1,&QWP(-0x40,$key)); + eval"&aes${p} ($inout,$rndkey0)"; + &$movekey ($rndkey0,&QWP(-0x30,$key)); + &set_label("${p}192"); + eval"&aes${p} ($inout,$rndkey1)"; + &$movekey ($rndkey1,&QWP(-0x20,$key)); + eval"&aes${p} ($inout,$rndkey0)"; + &$movekey ($rndkey0,&QWP(-0x10,$key)); + &set_label("${p}128"); + eval"&aes${p} ($inout,$rndkey1)"; + &$movekey ($rndkey1,&QWP(0,$key)); + eval"&aes${p} ($inout,$rndkey0)"; + &$movekey ($rndkey0,&QWP(0x10,$key)); + eval"&aes${p} ($inout,$rndkey1)"; + &$movekey ($rndkey1,&QWP(0x20,$key)); + eval"&aes${p} ($inout,$rndkey0)"; + &$movekey ($rndkey0,&QWP(0x30,$key)); + eval"&aes${p} ($inout,$rndkey1)"; + &$movekey ($rndkey1,&QWP(0x40,$key)); + eval"&aes${p} ($inout,$rndkey0)"; + &$movekey ($rndkey0,&QWP(0x50,$key)); + eval"&aes${p} ($inout,$rndkey1)"; + &$movekey ($rndkey1,&QWP(0x60,$key)); + eval"&aes${p} ($inout,$rndkey0)"; + &$movekey ($rndkey0,&QWP(0x70,$key)); + eval"&aes${p} ($inout,$rndkey1)"; + eval"&aes${p}last ($inout,$rndkey0)"; + &ret(); + &function_end_B("_aesni_${p}rypt1"); +} + +# void $PREFIX_encrypt (const void *inp,void *out,const AES_KEY *key); +&aesni_generate1("enc") if (!$inline); +&function_begin_B("${PREFIX}_encrypt"); + &mov ("eax",&wparam(0)); + &mov ($key,&wparam(2)); + &movups ($inout0,&QWP(0,"eax")); + &mov ($rounds,&DWP(240,$key)); + &mov ("eax",&wparam(1)); + if ($inline) + { &aesni_inline_generate1("enc"); } + else + { &call ("_aesni_encrypt1"); } + &pxor ($rndkey0,$rndkey0); # clear register bank + &pxor ($rndkey1,$rndkey1); + &movups (&QWP(0,"eax"),$inout0); + &pxor ($inout0,$inout0); + &ret (); +&function_end_B("${PREFIX}_encrypt"); + +# void $PREFIX_decrypt (const void *inp,void *out,const AES_KEY *key); +&aesni_generate1("dec") if(!$inline); +&function_begin_B("${PREFIX}_decrypt"); + &mov ("eax",&wparam(0)); + &mov ($key,&wparam(2)); + &movups ($inout0,&QWP(0,"eax")); + &mov ($rounds,&DWP(240,$key)); + &mov ("eax",&wparam(1)); + if ($inline) + { &aesni_inline_generate1("dec"); } + else + { &call ("_aesni_decrypt1"); } + &pxor ($rndkey0,$rndkey0); # clear register bank + &pxor ($rndkey1,$rndkey1); + &movups (&QWP(0,"eax"),$inout0); + &pxor ($inout0,$inout0); + &ret (); +&function_end_B("${PREFIX}_decrypt"); + +# _aesni_[en|de]cryptN are private interfaces, N denotes interleave +# factor. Why 3x subroutine were originally used in loops? Even though +# aes[enc|dec] latency was originally 6, it could be scheduled only +# every *2nd* cycle. Thus 3x interleave was the one providing optimal +# utilization, i.e. when subroutine's throughput is virtually same as +# of non-interleaved subroutine [for number of input blocks up to 3]. +# This is why it originally made no sense to implement 2x subroutine. +# But times change and it became appropriate to spend extra 192 bytes +# on 2x subroutine on Atom Silvermont account. For processors that +# can schedule aes[enc|dec] every cycle optimal interleave factor +# equals to corresponding instructions latency. 8x is optimal for +# * Bridge, but it's unfeasible to accommodate such implementation +# in XMM registers addreassable in 32-bit mode and therefore maximum +# of 6x is used instead... + +sub aesni_generate2 +{ my $p=shift; + + &function_begin_B("_aesni_${p}rypt2"); + &$movekey ($rndkey0,&QWP(0,$key)); + &shl ($rounds,4); + &$movekey ($rndkey1,&QWP(16,$key)); + &xorps ($inout0,$rndkey0); + &pxor ($inout1,$rndkey0); + &$movekey ($rndkey0,&QWP(32,$key)); + &lea ($key,&DWP(32,$key,$rounds)); + &neg ($rounds); + &add ($rounds,16); + + &set_label("${p}2_loop"); + eval"&aes${p} ($inout0,$rndkey1)"; + eval"&aes${p} ($inout1,$rndkey1)"; + &$movekey ($rndkey1,&QWP(0,$key,$rounds)); + &add ($rounds,32); + eval"&aes${p} ($inout0,$rndkey0)"; + eval"&aes${p} ($inout1,$rndkey0)"; + &$movekey ($rndkey0,&QWP(-16,$key,$rounds)); + &jnz (&label("${p}2_loop")); + eval"&aes${p} ($inout0,$rndkey1)"; + eval"&aes${p} ($inout1,$rndkey1)"; + eval"&aes${p}last ($inout0,$rndkey0)"; + eval"&aes${p}last ($inout1,$rndkey0)"; + &ret(); + &function_end_B("_aesni_${p}rypt2"); +} + +sub aesni_generate3 +{ my $p=shift; + + &function_begin_B("_aesni_${p}rypt3"); + &$movekey ($rndkey0,&QWP(0,$key)); + &shl ($rounds,4); + &$movekey ($rndkey1,&QWP(16,$key)); + &xorps ($inout0,$rndkey0); + &pxor ($inout1,$rndkey0); + &pxor ($inout2,$rndkey0); + &$movekey ($rndkey0,&QWP(32,$key)); + &lea ($key,&DWP(32,$key,$rounds)); + &neg ($rounds); + &add ($rounds,16); + + &set_label("${p}3_loop"); + eval"&aes${p} ($inout0,$rndkey1)"; + eval"&aes${p} ($inout1,$rndkey1)"; + eval"&aes${p} ($inout2,$rndkey1)"; + &$movekey ($rndkey1,&QWP(0,$key,$rounds)); + &add ($rounds,32); + eval"&aes${p} ($inout0,$rndkey0)"; + eval"&aes${p} ($inout1,$rndkey0)"; + eval"&aes${p} ($inout2,$rndkey0)"; + &$movekey ($rndkey0,&QWP(-16,$key,$rounds)); + &jnz (&label("${p}3_loop")); + eval"&aes${p} ($inout0,$rndkey1)"; + eval"&aes${p} ($inout1,$rndkey1)"; + eval"&aes${p} ($inout2,$rndkey1)"; + eval"&aes${p}last ($inout0,$rndkey0)"; + eval"&aes${p}last ($inout1,$rndkey0)"; + eval"&aes${p}last ($inout2,$rndkey0)"; + &ret(); + &function_end_B("_aesni_${p}rypt3"); +} + +# 4x interleave is implemented to improve small block performance, +# most notably [and naturally] 4 block by ~30%. One can argue that one +# should have implemented 5x as well, but improvement would be <20%, +# so it's not worth it... +sub aesni_generate4 +{ my $p=shift; + + &function_begin_B("_aesni_${p}rypt4"); + &$movekey ($rndkey0,&QWP(0,$key)); + &$movekey ($rndkey1,&QWP(16,$key)); + &shl ($rounds,4); + &xorps ($inout0,$rndkey0); + &pxor ($inout1,$rndkey0); + &pxor ($inout2,$rndkey0); + &pxor ($inout3,$rndkey0); + &$movekey ($rndkey0,&QWP(32,$key)); + &lea ($key,&DWP(32,$key,$rounds)); + &neg ($rounds); + &data_byte (0x0f,0x1f,0x40,0x00); + &add ($rounds,16); + + &set_label("${p}4_loop"); + eval"&aes${p} ($inout0,$rndkey1)"; + eval"&aes${p} ($inout1,$rndkey1)"; + eval"&aes${p} ($inout2,$rndkey1)"; + eval"&aes${p} ($inout3,$rndkey1)"; + &$movekey ($rndkey1,&QWP(0,$key,$rounds)); + &add ($rounds,32); + eval"&aes${p} ($inout0,$rndkey0)"; + eval"&aes${p} ($inout1,$rndkey0)"; + eval"&aes${p} ($inout2,$rndkey0)"; + eval"&aes${p} ($inout3,$rndkey0)"; + &$movekey ($rndkey0,&QWP(-16,$key,$rounds)); + &jnz (&label("${p}4_loop")); + + eval"&aes${p} ($inout0,$rndkey1)"; + eval"&aes${p} ($inout1,$rndkey1)"; + eval"&aes${p} ($inout2,$rndkey1)"; + eval"&aes${p} ($inout3,$rndkey1)"; + eval"&aes${p}last ($inout0,$rndkey0)"; + eval"&aes${p}last ($inout1,$rndkey0)"; + eval"&aes${p}last ($inout2,$rndkey0)"; + eval"&aes${p}last ($inout3,$rndkey0)"; + &ret(); + &function_end_B("_aesni_${p}rypt4"); +} + +sub aesni_generate6 +{ my $p=shift; + + &function_begin_B("_aesni_${p}rypt6"); + &static_label("_aesni_${p}rypt6_enter"); + &$movekey ($rndkey0,&QWP(0,$key)); + &shl ($rounds,4); + &$movekey ($rndkey1,&QWP(16,$key)); + &xorps ($inout0,$rndkey0); + &pxor ($inout1,$rndkey0); # pxor does better here + &pxor ($inout2,$rndkey0); + eval"&aes${p} ($inout0,$rndkey1)"; + &pxor ($inout3,$rndkey0); + &pxor ($inout4,$rndkey0); + eval"&aes${p} ($inout1,$rndkey1)"; + &lea ($key,&DWP(32,$key,$rounds)); + &neg ($rounds); + eval"&aes${p} ($inout2,$rndkey1)"; + &pxor ($inout5,$rndkey0); + &$movekey ($rndkey0,&QWP(0,$key,$rounds)); + &add ($rounds,16); + &jmp (&label("_aesni_${p}rypt6_inner")); + + &set_label("${p}6_loop",16); + eval"&aes${p} ($inout0,$rndkey1)"; + eval"&aes${p} ($inout1,$rndkey1)"; + eval"&aes${p} ($inout2,$rndkey1)"; + &set_label("_aesni_${p}rypt6_inner"); + eval"&aes${p} ($inout3,$rndkey1)"; + eval"&aes${p} ($inout4,$rndkey1)"; + eval"&aes${p} ($inout5,$rndkey1)"; + &set_label("_aesni_${p}rypt6_enter"); + &$movekey ($rndkey1,&QWP(0,$key,$rounds)); + &add ($rounds,32); + eval"&aes${p} ($inout0,$rndkey0)"; + eval"&aes${p} ($inout1,$rndkey0)"; + eval"&aes${p} ($inout2,$rndkey0)"; + eval"&aes${p} ($inout3,$rndkey0)"; + eval"&aes${p} ($inout4,$rndkey0)"; + eval"&aes${p} ($inout5,$rndkey0)"; + &$movekey ($rndkey0,&QWP(-16,$key,$rounds)); + &jnz (&label("${p}6_loop")); + + eval"&aes${p} ($inout0,$rndkey1)"; + eval"&aes${p} ($inout1,$rndkey1)"; + eval"&aes${p} ($inout2,$rndkey1)"; + eval"&aes${p} ($inout3,$rndkey1)"; + eval"&aes${p} ($inout4,$rndkey1)"; + eval"&aes${p} ($inout5,$rndkey1)"; + eval"&aes${p}last ($inout0,$rndkey0)"; + eval"&aes${p}last ($inout1,$rndkey0)"; + eval"&aes${p}last ($inout2,$rndkey0)"; + eval"&aes${p}last ($inout3,$rndkey0)"; + eval"&aes${p}last ($inout4,$rndkey0)"; + eval"&aes${p}last ($inout5,$rndkey0)"; + &ret(); + &function_end_B("_aesni_${p}rypt6"); +} +&aesni_generate2("enc") if ($PREFIX eq "aesni"); +&aesni_generate2("dec"); +&aesni_generate3("enc") if ($PREFIX eq "aesni"); +&aesni_generate3("dec"); +&aesni_generate4("enc") if ($PREFIX eq "aesni"); +&aesni_generate4("dec"); +&aesni_generate6("enc") if ($PREFIX eq "aesni"); +&aesni_generate6("dec"); + +if ($PREFIX eq "aesni") { +###################################################################### +# void aesni_ecb_encrypt (const void *in, void *out, +# size_t length, const AES_KEY *key, +# int enc); +&function_begin("aesni_ecb_encrypt"); + &mov ($inp,&wparam(0)); + &mov ($out,&wparam(1)); + &mov ($len,&wparam(2)); + &mov ($key,&wparam(3)); + &mov ($rounds_,&wparam(4)); + &and ($len,-16); + &jz (&label("ecb_ret")); + &mov ($rounds,&DWP(240,$key)); + &test ($rounds_,$rounds_); + &jz (&label("ecb_decrypt")); + + &mov ($key_,$key); # backup $key + &mov ($rounds_,$rounds); # backup $rounds + &cmp ($len,0x60); + &jb (&label("ecb_enc_tail")); + + &movdqu ($inout0,&QWP(0,$inp)); + &movdqu ($inout1,&QWP(0x10,$inp)); + &movdqu ($inout2,&QWP(0x20,$inp)); + &movdqu ($inout3,&QWP(0x30,$inp)); + &movdqu ($inout4,&QWP(0x40,$inp)); + &movdqu ($inout5,&QWP(0x50,$inp)); + &lea ($inp,&DWP(0x60,$inp)); + &sub ($len,0x60); + &jmp (&label("ecb_enc_loop6_enter")); + +&set_label("ecb_enc_loop6",16); + &movups (&QWP(0,$out),$inout0); + &movdqu ($inout0,&QWP(0,$inp)); + &movups (&QWP(0x10,$out),$inout1); + &movdqu ($inout1,&QWP(0x10,$inp)); + &movups (&QWP(0x20,$out),$inout2); + &movdqu ($inout2,&QWP(0x20,$inp)); + &movups (&QWP(0x30,$out),$inout3); + &movdqu ($inout3,&QWP(0x30,$inp)); + &movups (&QWP(0x40,$out),$inout4); + &movdqu ($inout4,&QWP(0x40,$inp)); + &movups (&QWP(0x50,$out),$inout5); + &lea ($out,&DWP(0x60,$out)); + &movdqu ($inout5,&QWP(0x50,$inp)); + &lea ($inp,&DWP(0x60,$inp)); +&set_label("ecb_enc_loop6_enter"); + + &call ("_aesni_encrypt6"); + + &mov ($key,$key_); # restore $key + &mov ($rounds,$rounds_); # restore $rounds + &sub ($len,0x60); + &jnc (&label("ecb_enc_loop6")); + + &movups (&QWP(0,$out),$inout0); + &movups (&QWP(0x10,$out),$inout1); + &movups (&QWP(0x20,$out),$inout2); + &movups (&QWP(0x30,$out),$inout3); + &movups (&QWP(0x40,$out),$inout4); + &movups (&QWP(0x50,$out),$inout5); + &lea ($out,&DWP(0x60,$out)); + &add ($len,0x60); + &jz (&label("ecb_ret")); + +&set_label("ecb_enc_tail"); + &movups ($inout0,&QWP(0,$inp)); + &cmp ($len,0x20); + &jb (&label("ecb_enc_one")); + &movups ($inout1,&QWP(0x10,$inp)); + &je (&label("ecb_enc_two")); + &movups ($inout2,&QWP(0x20,$inp)); + &cmp ($len,0x40); + &jb (&label("ecb_enc_three")); + &movups ($inout3,&QWP(0x30,$inp)); + &je (&label("ecb_enc_four")); + &movups ($inout4,&QWP(0x40,$inp)); + &xorps ($inout5,$inout5); + &call ("_aesni_encrypt6"); + &movups (&QWP(0,$out),$inout0); + &movups (&QWP(0x10,$out),$inout1); + &movups (&QWP(0x20,$out),$inout2); + &movups (&QWP(0x30,$out),$inout3); + &movups (&QWP(0x40,$out),$inout4); + jmp (&label("ecb_ret")); + +&set_label("ecb_enc_one",16); + if ($inline) + { &aesni_inline_generate1("enc"); } + else + { &call ("_aesni_encrypt1"); } + &movups (&QWP(0,$out),$inout0); + &jmp (&label("ecb_ret")); + +&set_label("ecb_enc_two",16); + &call ("_aesni_encrypt2"); + &movups (&QWP(0,$out),$inout0); + &movups (&QWP(0x10,$out),$inout1); + &jmp (&label("ecb_ret")); + +&set_label("ecb_enc_three",16); + &call ("_aesni_encrypt3"); + &movups (&QWP(0,$out),$inout0); + &movups (&QWP(0x10,$out),$inout1); + &movups (&QWP(0x20,$out),$inout2); + &jmp (&label("ecb_ret")); + +&set_label("ecb_enc_four",16); + &call ("_aesni_encrypt4"); + &movups (&QWP(0,$out),$inout0); + &movups (&QWP(0x10,$out),$inout1); + &movups (&QWP(0x20,$out),$inout2); + &movups (&QWP(0x30,$out),$inout3); + &jmp (&label("ecb_ret")); +###################################################################### +&set_label("ecb_decrypt",16); + &mov ($key_,$key); # backup $key + &mov ($rounds_,$rounds); # backup $rounds + &cmp ($len,0x60); + &jb (&label("ecb_dec_tail")); + + &movdqu ($inout0,&QWP(0,$inp)); + &movdqu ($inout1,&QWP(0x10,$inp)); + &movdqu ($inout2,&QWP(0x20,$inp)); + &movdqu ($inout3,&QWP(0x30,$inp)); + &movdqu ($inout4,&QWP(0x40,$inp)); + &movdqu ($inout5,&QWP(0x50,$inp)); + &lea ($inp,&DWP(0x60,$inp)); + &sub ($len,0x60); + &jmp (&label("ecb_dec_loop6_enter")); + +&set_label("ecb_dec_loop6",16); + &movups (&QWP(0,$out),$inout0); + &movdqu ($inout0,&QWP(0,$inp)); + &movups (&QWP(0x10,$out),$inout1); + &movdqu ($inout1,&QWP(0x10,$inp)); + &movups (&QWP(0x20,$out),$inout2); + &movdqu ($inout2,&QWP(0x20,$inp)); + &movups (&QWP(0x30,$out),$inout3); + &movdqu ($inout3,&QWP(0x30,$inp)); + &movups (&QWP(0x40,$out),$inout4); + &movdqu ($inout4,&QWP(0x40,$inp)); + &movups (&QWP(0x50,$out),$inout5); + &lea ($out,&DWP(0x60,$out)); + &movdqu ($inout5,&QWP(0x50,$inp)); + &lea ($inp,&DWP(0x60,$inp)); +&set_label("ecb_dec_loop6_enter"); + + &call ("_aesni_decrypt6"); + + &mov ($key,$key_); # restore $key + &mov ($rounds,$rounds_); # restore $rounds + &sub ($len,0x60); + &jnc (&label("ecb_dec_loop6")); + + &movups (&QWP(0,$out),$inout0); + &movups (&QWP(0x10,$out),$inout1); + &movups (&QWP(0x20,$out),$inout2); + &movups (&QWP(0x30,$out),$inout3); + &movups (&QWP(0x40,$out),$inout4); + &movups (&QWP(0x50,$out),$inout5); + &lea ($out,&DWP(0x60,$out)); + &add ($len,0x60); + &jz (&label("ecb_ret")); + +&set_label("ecb_dec_tail"); + &movups ($inout0,&QWP(0,$inp)); + &cmp ($len,0x20); + &jb (&label("ecb_dec_one")); + &movups ($inout1,&QWP(0x10,$inp)); + &je (&label("ecb_dec_two")); + &movups ($inout2,&QWP(0x20,$inp)); + &cmp ($len,0x40); + &jb (&label("ecb_dec_three")); + &movups ($inout3,&QWP(0x30,$inp)); + &je (&label("ecb_dec_four")); + &movups ($inout4,&QWP(0x40,$inp)); + &xorps ($inout5,$inout5); + &call ("_aesni_decrypt6"); + &movups (&QWP(0,$out),$inout0); + &movups (&QWP(0x10,$out),$inout1); + &movups (&QWP(0x20,$out),$inout2); + &movups (&QWP(0x30,$out),$inout3); + &movups (&QWP(0x40,$out),$inout4); + &jmp (&label("ecb_ret")); + +&set_label("ecb_dec_one",16); + if ($inline) + { &aesni_inline_generate1("dec"); } + else + { &call ("_aesni_decrypt1"); } + &movups (&QWP(0,$out),$inout0); + &jmp (&label("ecb_ret")); + +&set_label("ecb_dec_two",16); + &call ("_aesni_decrypt2"); + &movups (&QWP(0,$out),$inout0); + &movups (&QWP(0x10,$out),$inout1); + &jmp (&label("ecb_ret")); + +&set_label("ecb_dec_three",16); + &call ("_aesni_decrypt3"); + &movups (&QWP(0,$out),$inout0); + &movups (&QWP(0x10,$out),$inout1); + &movups (&QWP(0x20,$out),$inout2); + &jmp (&label("ecb_ret")); + +&set_label("ecb_dec_four",16); + &call ("_aesni_decrypt4"); + &movups (&QWP(0,$out),$inout0); + &movups (&QWP(0x10,$out),$inout1); + &movups (&QWP(0x20,$out),$inout2); + &movups (&QWP(0x30,$out),$inout3); + +&set_label("ecb_ret"); + &pxor ("xmm0","xmm0"); # clear register bank + &pxor ("xmm1","xmm1"); + &pxor ("xmm2","xmm2"); + &pxor ("xmm3","xmm3"); + &pxor ("xmm4","xmm4"); + &pxor ("xmm5","xmm5"); + &pxor ("xmm6","xmm6"); + &pxor ("xmm7","xmm7"); +&function_end("aesni_ecb_encrypt"); + +###################################################################### +# void aesni_ccm64_[en|de]crypt_blocks (const void *in, void *out, +# size_t blocks, const AES_KEY *key, +# const char *ivec,char *cmac); +# +# Handles only complete blocks, operates on 64-bit counter and +# does not update *ivec! Nor does it finalize CMAC value +# (see engine/eng_aesni.c for details) +# +{ my $cmac=$inout1; +&function_begin("aesni_ccm64_encrypt_blocks"); + &mov ($inp,&wparam(0)); + &mov ($out,&wparam(1)); + &mov ($len,&wparam(2)); + &mov ($key,&wparam(3)); + &mov ($rounds_,&wparam(4)); + &mov ($rounds,&wparam(5)); + &mov ($key_,"esp"); + &sub ("esp",60); + &and ("esp",-16); # align stack + &mov (&DWP(48,"esp"),$key_); + + &movdqu ($ivec,&QWP(0,$rounds_)); # load ivec + &movdqu ($cmac,&QWP(0,$rounds)); # load cmac + &mov ($rounds,&DWP(240,$key)); + + # compose byte-swap control mask for pshufb on stack + &mov (&DWP(0,"esp"),0x0c0d0e0f); + &mov (&DWP(4,"esp"),0x08090a0b); + &mov (&DWP(8,"esp"),0x04050607); + &mov (&DWP(12,"esp"),0x00010203); + + # compose counter increment vector on stack + &mov ($rounds_,1); + &xor ($key_,$key_); + &mov (&DWP(16,"esp"),$rounds_); + &mov (&DWP(20,"esp"),$key_); + &mov (&DWP(24,"esp"),$key_); + &mov (&DWP(28,"esp"),$key_); + + &shl ($rounds,4); + &mov ($rounds_,16); + &lea ($key_,&DWP(0,$key)); + &movdqa ($inout3,&QWP(0,"esp")); + &movdqa ($inout0,$ivec); + &lea ($key,&DWP(32,$key,$rounds)); + &sub ($rounds_,$rounds); + &pshufb ($ivec,$inout3); + +&set_label("ccm64_enc_outer"); + &$movekey ($rndkey0,&QWP(0,$key_)); + &mov ($rounds,$rounds_); + &movups ($in0,&QWP(0,$inp)); + + &xorps ($inout0,$rndkey0); + &$movekey ($rndkey1,&QWP(16,$key_)); + &xorps ($rndkey0,$in0); + &xorps ($cmac,$rndkey0); # cmac^=inp + &$movekey ($rndkey0,&QWP(32,$key_)); + +&set_label("ccm64_enc2_loop"); + &aesenc ($inout0,$rndkey1); + &aesenc ($cmac,$rndkey1); + &$movekey ($rndkey1,&QWP(0,$key,$rounds)); + &add ($rounds,32); + &aesenc ($inout0,$rndkey0); + &aesenc ($cmac,$rndkey0); + &$movekey ($rndkey0,&QWP(-16,$key,$rounds)); + &jnz (&label("ccm64_enc2_loop")); + &aesenc ($inout0,$rndkey1); + &aesenc ($cmac,$rndkey1); + &paddq ($ivec,&QWP(16,"esp")); + &dec ($len); + &aesenclast ($inout0,$rndkey0); + &aesenclast ($cmac,$rndkey0); + + &lea ($inp,&DWP(16,$inp)); + &xorps ($in0,$inout0); # inp^=E(ivec) + &movdqa ($inout0,$ivec); + &movups (&QWP(0,$out),$in0); # save output + &pshufb ($inout0,$inout3); + &lea ($out,&DWP(16,$out)); + &jnz (&label("ccm64_enc_outer")); + + &mov ("esp",&DWP(48,"esp")); + &mov ($out,&wparam(5)); + &movups (&QWP(0,$out),$cmac); + + &pxor ("xmm0","xmm0"); # clear register bank + &pxor ("xmm1","xmm1"); + &pxor ("xmm2","xmm2"); + &pxor ("xmm3","xmm3"); + &pxor ("xmm4","xmm4"); + &pxor ("xmm5","xmm5"); + &pxor ("xmm6","xmm6"); + &pxor ("xmm7","xmm7"); +&function_end("aesni_ccm64_encrypt_blocks"); + +&function_begin("aesni_ccm64_decrypt_blocks"); + &mov ($inp,&wparam(0)); + &mov ($out,&wparam(1)); + &mov ($len,&wparam(2)); + &mov ($key,&wparam(3)); + &mov ($rounds_,&wparam(4)); + &mov ($rounds,&wparam(5)); + &mov ($key_,"esp"); + &sub ("esp",60); + &and ("esp",-16); # align stack + &mov (&DWP(48,"esp"),$key_); + + &movdqu ($ivec,&QWP(0,$rounds_)); # load ivec + &movdqu ($cmac,&QWP(0,$rounds)); # load cmac + &mov ($rounds,&DWP(240,$key)); + + # compose byte-swap control mask for pshufb on stack + &mov (&DWP(0,"esp"),0x0c0d0e0f); + &mov (&DWP(4,"esp"),0x08090a0b); + &mov (&DWP(8,"esp"),0x04050607); + &mov (&DWP(12,"esp"),0x00010203); + + # compose counter increment vector on stack + &mov ($rounds_,1); + &xor ($key_,$key_); + &mov (&DWP(16,"esp"),$rounds_); + &mov (&DWP(20,"esp"),$key_); + &mov (&DWP(24,"esp"),$key_); + &mov (&DWP(28,"esp"),$key_); + + &movdqa ($inout3,&QWP(0,"esp")); # bswap mask + &movdqa ($inout0,$ivec); + + &mov ($key_,$key); + &mov ($rounds_,$rounds); + + &pshufb ($ivec,$inout3); + if ($inline) + { &aesni_inline_generate1("enc"); } + else + { &call ("_aesni_encrypt1"); } + &shl ($rounds_,4); + &mov ($rounds,16); + &movups ($in0,&QWP(0,$inp)); # load inp + &paddq ($ivec,&QWP(16,"esp")); + &lea ($inp,&QWP(16,$inp)); + &sub ($rounds,$rounds_); + &lea ($key,&DWP(32,$key_,$rounds_)); + &mov ($rounds_,$rounds); + &jmp (&label("ccm64_dec_outer")); + +&set_label("ccm64_dec_outer",16); + &xorps ($in0,$inout0); # inp ^= E(ivec) + &movdqa ($inout0,$ivec); + &movups (&QWP(0,$out),$in0); # save output + &lea ($out,&DWP(16,$out)); + &pshufb ($inout0,$inout3); + + &sub ($len,1); + &jz (&label("ccm64_dec_break")); + + &$movekey ($rndkey0,&QWP(0,$key_)); + &mov ($rounds,$rounds_); + &$movekey ($rndkey1,&QWP(16,$key_)); + &xorps ($in0,$rndkey0); + &xorps ($inout0,$rndkey0); + &xorps ($cmac,$in0); # cmac^=out + &$movekey ($rndkey0,&QWP(32,$key_)); + +&set_label("ccm64_dec2_loop"); + &aesenc ($inout0,$rndkey1); + &aesenc ($cmac,$rndkey1); + &$movekey ($rndkey1,&QWP(0,$key,$rounds)); + &add ($rounds,32); + &aesenc ($inout0,$rndkey0); + &aesenc ($cmac,$rndkey0); + &$movekey ($rndkey0,&QWP(-16,$key,$rounds)); + &jnz (&label("ccm64_dec2_loop")); + &movups ($in0,&QWP(0,$inp)); # load inp + &paddq ($ivec,&QWP(16,"esp")); + &aesenc ($inout0,$rndkey1); + &aesenc ($cmac,$rndkey1); + &aesenclast ($inout0,$rndkey0); + &aesenclast ($cmac,$rndkey0); + &lea ($inp,&QWP(16,$inp)); + &jmp (&label("ccm64_dec_outer")); + +&set_label("ccm64_dec_break",16); + &mov ($rounds,&DWP(240,$key_)); + &mov ($key,$key_); + if ($inline) + { &aesni_inline_generate1("enc",$cmac,$in0); } + else + { &call ("_aesni_encrypt1",$cmac); } + + &mov ("esp",&DWP(48,"esp")); + &mov ($out,&wparam(5)); + &movups (&QWP(0,$out),$cmac); + + &pxor ("xmm0","xmm0"); # clear register bank + &pxor ("xmm1","xmm1"); + &pxor ("xmm2","xmm2"); + &pxor ("xmm3","xmm3"); + &pxor ("xmm4","xmm4"); + &pxor ("xmm5","xmm5"); + &pxor ("xmm6","xmm6"); + &pxor ("xmm7","xmm7"); +&function_end("aesni_ccm64_decrypt_blocks"); +} + +###################################################################### +# void aesni_ctr32_encrypt_blocks (const void *in, void *out, +# size_t blocks, const AES_KEY *key, +# const char *ivec); +# +# Handles only complete blocks, operates on 32-bit counter and +# does not update *ivec! (see crypto/modes/ctr128.c for details) +# +# stack layout: +# 0 pshufb mask +# 16 vector addend: 0,6,6,6 +# 32 counter-less ivec +# 48 1st triplet of counter vector +# 64 2nd triplet of counter vector +# 80 saved %esp + +&function_begin("aesni_ctr32_encrypt_blocks"); + &mov ($inp,&wparam(0)); + &mov ($out,&wparam(1)); + &mov ($len,&wparam(2)); + &mov ($key,&wparam(3)); + &mov ($rounds_,&wparam(4)); + &mov ($key_,"esp"); + &sub ("esp",88); + &and ("esp",-16); # align stack + &mov (&DWP(80,"esp"),$key_); + + &cmp ($len,1); + &je (&label("ctr32_one_shortcut")); + + &movdqu ($inout5,&QWP(0,$rounds_)); # load ivec + + # compose byte-swap control mask for pshufb on stack + &mov (&DWP(0,"esp"),0x0c0d0e0f); + &mov (&DWP(4,"esp"),0x08090a0b); + &mov (&DWP(8,"esp"),0x04050607); + &mov (&DWP(12,"esp"),0x00010203); + + # compose counter increment vector on stack + &mov ($rounds,6); + &xor ($key_,$key_); + &mov (&DWP(16,"esp"),$rounds); + &mov (&DWP(20,"esp"),$rounds); + &mov (&DWP(24,"esp"),$rounds); + &mov (&DWP(28,"esp"),$key_); + + &pextrd ($rounds_,$inout5,3); # pull 32-bit counter + &pinsrd ($inout5,$key_,3); # wipe 32-bit counter + + &mov ($rounds,&DWP(240,$key)); # key->rounds + + # compose 2 vectors of 3x32-bit counters + &bswap ($rounds_); + &pxor ($rndkey0,$rndkey0); + &pxor ($rndkey1,$rndkey1); + &movdqa ($inout0,&QWP(0,"esp")); # load byte-swap mask + &pinsrd ($rndkey0,$rounds_,0); + &lea ($key_,&DWP(3,$rounds_)); + &pinsrd ($rndkey1,$key_,0); + &inc ($rounds_); + &pinsrd ($rndkey0,$rounds_,1); + &inc ($key_); + &pinsrd ($rndkey1,$key_,1); + &inc ($rounds_); + &pinsrd ($rndkey0,$rounds_,2); + &inc ($key_); + &pinsrd ($rndkey1,$key_,2); + &movdqa (&QWP(48,"esp"),$rndkey0); # save 1st triplet + &pshufb ($rndkey0,$inout0); # byte swap + &movdqu ($inout4,&QWP(0,$key)); # key[0] + &movdqa (&QWP(64,"esp"),$rndkey1); # save 2nd triplet + &pshufb ($rndkey1,$inout0); # byte swap + + &pshufd ($inout0,$rndkey0,3<<6); # place counter to upper dword + &pshufd ($inout1,$rndkey0,2<<6); + &cmp ($len,6); + &jb (&label("ctr32_tail")); + &pxor ($inout5,$inout4); # counter-less ivec^key[0] + &shl ($rounds,4); + &mov ($rounds_,16); + &movdqa (&QWP(32,"esp"),$inout5); # save counter-less ivec^key[0] + &mov ($key_,$key); # backup $key + &sub ($rounds_,$rounds); # backup twisted $rounds + &lea ($key,&DWP(32,$key,$rounds)); + &sub ($len,6); + &jmp (&label("ctr32_loop6")); + +&set_label("ctr32_loop6",16); + # inlining _aesni_encrypt6's prologue gives ~6% improvement... + &pshufd ($inout2,$rndkey0,1<<6); + &movdqa ($rndkey0,&QWP(32,"esp")); # pull counter-less ivec + &pshufd ($inout3,$rndkey1,3<<6); + &pxor ($inout0,$rndkey0); # merge counter-less ivec + &pshufd ($inout4,$rndkey1,2<<6); + &pxor ($inout1,$rndkey0); + &pshufd ($inout5,$rndkey1,1<<6); + &$movekey ($rndkey1,&QWP(16,$key_)); + &pxor ($inout2,$rndkey0); + &pxor ($inout3,$rndkey0); + &aesenc ($inout0,$rndkey1); + &pxor ($inout4,$rndkey0); + &pxor ($inout5,$rndkey0); + &aesenc ($inout1,$rndkey1); + &$movekey ($rndkey0,&QWP(32,$key_)); + &mov ($rounds,$rounds_); + &aesenc ($inout2,$rndkey1); + &aesenc ($inout3,$rndkey1); + &aesenc ($inout4,$rndkey1); + &aesenc ($inout5,$rndkey1); + + &call (&label("_aesni_encrypt6_enter")); + + &movups ($rndkey1,&QWP(0,$inp)); + &movups ($rndkey0,&QWP(0x10,$inp)); + &xorps ($inout0,$rndkey1); + &movups ($rndkey1,&QWP(0x20,$inp)); + &xorps ($inout1,$rndkey0); + &movups (&QWP(0,$out),$inout0); + &movdqa ($rndkey0,&QWP(16,"esp")); # load increment + &xorps ($inout2,$rndkey1); + &movdqa ($rndkey1,&QWP(64,"esp")); # load 2nd triplet + &movups (&QWP(0x10,$out),$inout1); + &movups (&QWP(0x20,$out),$inout2); + + &paddd ($rndkey1,$rndkey0); # 2nd triplet increment + &paddd ($rndkey0,&QWP(48,"esp")); # 1st triplet increment + &movdqa ($inout0,&QWP(0,"esp")); # load byte swap mask + + &movups ($inout1,&QWP(0x30,$inp)); + &movups ($inout2,&QWP(0x40,$inp)); + &xorps ($inout3,$inout1); + &movups ($inout1,&QWP(0x50,$inp)); + &lea ($inp,&DWP(0x60,$inp)); + &movdqa (&QWP(48,"esp"),$rndkey0); # save 1st triplet + &pshufb ($rndkey0,$inout0); # byte swap + &xorps ($inout4,$inout2); + &movups (&QWP(0x30,$out),$inout3); + &xorps ($inout5,$inout1); + &movdqa (&QWP(64,"esp"),$rndkey1); # save 2nd triplet + &pshufb ($rndkey1,$inout0); # byte swap + &movups (&QWP(0x40,$out),$inout4); + &pshufd ($inout0,$rndkey0,3<<6); + &movups (&QWP(0x50,$out),$inout5); + &lea ($out,&DWP(0x60,$out)); + + &pshufd ($inout1,$rndkey0,2<<6); + &sub ($len,6); + &jnc (&label("ctr32_loop6")); + + &add ($len,6); + &jz (&label("ctr32_ret")); + &movdqu ($inout5,&QWP(0,$key_)); + &mov ($key,$key_); + &pxor ($inout5,&QWP(32,"esp")); # restore count-less ivec + &mov ($rounds,&DWP(240,$key_)); # restore $rounds + +&set_label("ctr32_tail"); + &por ($inout0,$inout5); + &cmp ($len,2); + &jb (&label("ctr32_one")); + + &pshufd ($inout2,$rndkey0,1<<6); + &por ($inout1,$inout5); + &je (&label("ctr32_two")); + + &pshufd ($inout3,$rndkey1,3<<6); + &por ($inout2,$inout5); + &cmp ($len,4); + &jb (&label("ctr32_three")); + + &pshufd ($inout4,$rndkey1,2<<6); + &por ($inout3,$inout5); + &je (&label("ctr32_four")); + + &por ($inout4,$inout5); + &call ("_aesni_encrypt6"); + &movups ($rndkey1,&QWP(0,$inp)); + &movups ($rndkey0,&QWP(0x10,$inp)); + &xorps ($inout0,$rndkey1); + &movups ($rndkey1,&QWP(0x20,$inp)); + &xorps ($inout1,$rndkey0); + &movups ($rndkey0,&QWP(0x30,$inp)); + &xorps ($inout2,$rndkey1); + &movups ($rndkey1,&QWP(0x40,$inp)); + &xorps ($inout3,$rndkey0); + &movups (&QWP(0,$out),$inout0); + &xorps ($inout4,$rndkey1); + &movups (&QWP(0x10,$out),$inout1); + &movups (&QWP(0x20,$out),$inout2); + &movups (&QWP(0x30,$out),$inout3); + &movups (&QWP(0x40,$out),$inout4); + &jmp (&label("ctr32_ret")); + +&set_label("ctr32_one_shortcut",16); + &movups ($inout0,&QWP(0,$rounds_)); # load ivec + &mov ($rounds,&DWP(240,$key)); + +&set_label("ctr32_one"); + if ($inline) + { &aesni_inline_generate1("enc"); } + else + { &call ("_aesni_encrypt1"); } + &movups ($in0,&QWP(0,$inp)); + &xorps ($in0,$inout0); + &movups (&QWP(0,$out),$in0); + &jmp (&label("ctr32_ret")); + +&set_label("ctr32_two",16); + &call ("_aesni_encrypt2"); + &movups ($inout3,&QWP(0,$inp)); + &movups ($inout4,&QWP(0x10,$inp)); + &xorps ($inout0,$inout3); + &xorps ($inout1,$inout4); + &movups (&QWP(0,$out),$inout0); + &movups (&QWP(0x10,$out),$inout1); + &jmp (&label("ctr32_ret")); + +&set_label("ctr32_three",16); + &call ("_aesni_encrypt3"); + &movups ($inout3,&QWP(0,$inp)); + &movups ($inout4,&QWP(0x10,$inp)); + &xorps ($inout0,$inout3); + &movups ($inout5,&QWP(0x20,$inp)); + &xorps ($inout1,$inout4); + &movups (&QWP(0,$out),$inout0); + &xorps ($inout2,$inout5); + &movups (&QWP(0x10,$out),$inout1); + &movups (&QWP(0x20,$out),$inout2); + &jmp (&label("ctr32_ret")); + +&set_label("ctr32_four",16); + &call ("_aesni_encrypt4"); + &movups ($inout4,&QWP(0,$inp)); + &movups ($inout5,&QWP(0x10,$inp)); + &movups ($rndkey1,&QWP(0x20,$inp)); + &xorps ($inout0,$inout4); + &movups ($rndkey0,&QWP(0x30,$inp)); + &xorps ($inout1,$inout5); + &movups (&QWP(0,$out),$inout0); + &xorps ($inout2,$rndkey1); + &movups (&QWP(0x10,$out),$inout1); + &xorps ($inout3,$rndkey0); + &movups (&QWP(0x20,$out),$inout2); + &movups (&QWP(0x30,$out),$inout3); + +&set_label("ctr32_ret"); + &pxor ("xmm0","xmm0"); # clear register bank + &pxor ("xmm1","xmm1"); + &pxor ("xmm2","xmm2"); + &pxor ("xmm3","xmm3"); + &pxor ("xmm4","xmm4"); + &movdqa (&QWP(32,"esp"),"xmm0"); # clear stack + &pxor ("xmm5","xmm5"); + &movdqa (&QWP(48,"esp"),"xmm0"); + &pxor ("xmm6","xmm6"); + &movdqa (&QWP(64,"esp"),"xmm0"); + &pxor ("xmm7","xmm7"); + &mov ("esp",&DWP(80,"esp")); +&function_end("aesni_ctr32_encrypt_blocks"); + +###################################################################### +# void aesni_xts_[en|de]crypt(const char *inp,char *out,size_t len, +# const AES_KEY *key1, const AES_KEY *key2 +# const unsigned char iv[16]); +# +{ my ($tweak,$twtmp,$twres,$twmask)=($rndkey1,$rndkey0,$inout0,$inout1); + +&function_begin("aesni_xts_encrypt"); + &mov ($key,&wparam(4)); # key2 + &mov ($inp,&wparam(5)); # clear-text tweak + + &mov ($rounds,&DWP(240,$key)); # key2->rounds + &movups ($inout0,&QWP(0,$inp)); + if ($inline) + { &aesni_inline_generate1("enc"); } + else + { &call ("_aesni_encrypt1"); } + + &mov ($inp,&wparam(0)); + &mov ($out,&wparam(1)); + &mov ($len,&wparam(2)); + &mov ($key,&wparam(3)); # key1 + + &mov ($key_,"esp"); + &sub ("esp",16*7+8); + &mov ($rounds,&DWP(240,$key)); # key1->rounds + &and ("esp",-16); # align stack + + &mov (&DWP(16*6+0,"esp"),0x87); # compose the magic constant + &mov (&DWP(16*6+4,"esp"),0); + &mov (&DWP(16*6+8,"esp"),1); + &mov (&DWP(16*6+12,"esp"),0); + &mov (&DWP(16*7+0,"esp"),$len); # save original $len + &mov (&DWP(16*7+4,"esp"),$key_); # save original %esp + + &movdqa ($tweak,$inout0); + &pxor ($twtmp,$twtmp); + &movdqa ($twmask,&QWP(6*16,"esp")); # 0x0...010...87 + &pcmpgtd($twtmp,$tweak); # broadcast upper bits + + &and ($len,-16); + &mov ($key_,$key); # backup $key + &mov ($rounds_,$rounds); # backup $rounds + &sub ($len,16*6); + &jc (&label("xts_enc_short")); + + &shl ($rounds,4); + &mov ($rounds_,16); + &sub ($rounds_,$rounds); + &lea ($key,&DWP(32,$key,$rounds)); + &jmp (&label("xts_enc_loop6")); + +&set_label("xts_enc_loop6",16); + for ($i=0;$i<4;$i++) { + &pshufd ($twres,$twtmp,0x13); + &pxor ($twtmp,$twtmp); + &movdqa (&QWP(16*$i,"esp"),$tweak); + &paddq ($tweak,$tweak); # &psllq($tweak,1); + &pand ($twres,$twmask); # isolate carry and residue + &pcmpgtd ($twtmp,$tweak); # broadcast upper bits + &pxor ($tweak,$twres); + } + &pshufd ($inout5,$twtmp,0x13); + &movdqa (&QWP(16*$i++,"esp"),$tweak); + &paddq ($tweak,$tweak); # &psllq($tweak,1); + &$movekey ($rndkey0,&QWP(0,$key_)); + &pand ($inout5,$twmask); # isolate carry and residue + &movups ($inout0,&QWP(0,$inp)); # load input + &pxor ($inout5,$tweak); + + # inline _aesni_encrypt6 prologue and flip xor with tweak and key[0] + &mov ($rounds,$rounds_); # restore $rounds + &movdqu ($inout1,&QWP(16*1,$inp)); + &xorps ($inout0,$rndkey0); # input^=rndkey[0] + &movdqu ($inout2,&QWP(16*2,$inp)); + &pxor ($inout1,$rndkey0); + &movdqu ($inout3,&QWP(16*3,$inp)); + &pxor ($inout2,$rndkey0); + &movdqu ($inout4,&QWP(16*4,$inp)); + &pxor ($inout3,$rndkey0); + &movdqu ($rndkey1,&QWP(16*5,$inp)); + &pxor ($inout4,$rndkey0); + &lea ($inp,&DWP(16*6,$inp)); + &pxor ($inout0,&QWP(16*0,"esp")); # input^=tweak + &movdqa (&QWP(16*$i,"esp"),$inout5); # save last tweak + &pxor ($inout5,$rndkey1); + + &$movekey ($rndkey1,&QWP(16,$key_)); + &pxor ($inout1,&QWP(16*1,"esp")); + &pxor ($inout2,&QWP(16*2,"esp")); + &aesenc ($inout0,$rndkey1); + &pxor ($inout3,&QWP(16*3,"esp")); + &pxor ($inout4,&QWP(16*4,"esp")); + &aesenc ($inout1,$rndkey1); + &pxor ($inout5,$rndkey0); + &$movekey ($rndkey0,&QWP(32,$key_)); + &aesenc ($inout2,$rndkey1); + &aesenc ($inout3,$rndkey1); + &aesenc ($inout4,$rndkey1); + &aesenc ($inout5,$rndkey1); + &call (&label("_aesni_encrypt6_enter")); + + &movdqa ($tweak,&QWP(16*5,"esp")); # last tweak + &pxor ($twtmp,$twtmp); + &xorps ($inout0,&QWP(16*0,"esp")); # output^=tweak + &pcmpgtd ($twtmp,$tweak); # broadcast upper bits + &xorps ($inout1,&QWP(16*1,"esp")); + &movups (&QWP(16*0,$out),$inout0); # write output + &xorps ($inout2,&QWP(16*2,"esp")); + &movups (&QWP(16*1,$out),$inout1); + &xorps ($inout3,&QWP(16*3,"esp")); + &movups (&QWP(16*2,$out),$inout2); + &xorps ($inout4,&QWP(16*4,"esp")); + &movups (&QWP(16*3,$out),$inout3); + &xorps ($inout5,$tweak); + &movups (&QWP(16*4,$out),$inout4); + &pshufd ($twres,$twtmp,0x13); + &movups (&QWP(16*5,$out),$inout5); + &lea ($out,&DWP(16*6,$out)); + &movdqa ($twmask,&QWP(16*6,"esp")); # 0x0...010...87 + + &pxor ($twtmp,$twtmp); + &paddq ($tweak,$tweak); # &psllq($tweak,1); + &pand ($twres,$twmask); # isolate carry and residue + &pcmpgtd($twtmp,$tweak); # broadcast upper bits + &pxor ($tweak,$twres); + + &sub ($len,16*6); + &jnc (&label("xts_enc_loop6")); + + &mov ($rounds,&DWP(240,$key_)); # restore $rounds + &mov ($key,$key_); # restore $key + &mov ($rounds_,$rounds); + +&set_label("xts_enc_short"); + &add ($len,16*6); + &jz (&label("xts_enc_done6x")); + + &movdqa ($inout3,$tweak); # put aside previous tweak + &cmp ($len,0x20); + &jb (&label("xts_enc_one")); + + &pshufd ($twres,$twtmp,0x13); + &pxor ($twtmp,$twtmp); + &paddq ($tweak,$tweak); # &psllq($tweak,1); + &pand ($twres,$twmask); # isolate carry and residue + &pcmpgtd($twtmp,$tweak); # broadcast upper bits + &pxor ($tweak,$twres); + &je (&label("xts_enc_two")); + + &pshufd ($twres,$twtmp,0x13); + &pxor ($twtmp,$twtmp); + &movdqa ($inout4,$tweak); # put aside previous tweak + &paddq ($tweak,$tweak); # &psllq($tweak,1); + &pand ($twres,$twmask); # isolate carry and residue + &pcmpgtd($twtmp,$tweak); # broadcast upper bits + &pxor ($tweak,$twres); + &cmp ($len,0x40); + &jb (&label("xts_enc_three")); + + &pshufd ($twres,$twtmp,0x13); + &pxor ($twtmp,$twtmp); + &movdqa ($inout5,$tweak); # put aside previous tweak + &paddq ($tweak,$tweak); # &psllq($tweak,1); + &pand ($twres,$twmask); # isolate carry and residue + &pcmpgtd($twtmp,$tweak); # broadcast upper bits + &pxor ($tweak,$twres); + &movdqa (&QWP(16*0,"esp"),$inout3); + &movdqa (&QWP(16*1,"esp"),$inout4); + &je (&label("xts_enc_four")); + + &movdqa (&QWP(16*2,"esp"),$inout5); + &pshufd ($inout5,$twtmp,0x13); + &movdqa (&QWP(16*3,"esp"),$tweak); + &paddq ($tweak,$tweak); # &psllq($inout0,1); + &pand ($inout5,$twmask); # isolate carry and residue + &pxor ($inout5,$tweak); + + &movdqu ($inout0,&QWP(16*0,$inp)); # load input + &movdqu ($inout1,&QWP(16*1,$inp)); + &movdqu ($inout2,&QWP(16*2,$inp)); + &pxor ($inout0,&QWP(16*0,"esp")); # input^=tweak + &movdqu ($inout3,&QWP(16*3,$inp)); + &pxor ($inout1,&QWP(16*1,"esp")); + &movdqu ($inout4,&QWP(16*4,$inp)); + &pxor ($inout2,&QWP(16*2,"esp")); + &lea ($inp,&DWP(16*5,$inp)); + &pxor ($inout3,&QWP(16*3,"esp")); + &movdqa (&QWP(16*4,"esp"),$inout5); # save last tweak + &pxor ($inout4,$inout5); + + &call ("_aesni_encrypt6"); + + &movaps ($tweak,&QWP(16*4,"esp")); # last tweak + &xorps ($inout0,&QWP(16*0,"esp")); # output^=tweak + &xorps ($inout1,&QWP(16*1,"esp")); + &xorps ($inout2,&QWP(16*2,"esp")); + &movups (&QWP(16*0,$out),$inout0); # write output + &xorps ($inout3,&QWP(16*3,"esp")); + &movups (&QWP(16*1,$out),$inout1); + &xorps ($inout4,$tweak); + &movups (&QWP(16*2,$out),$inout2); + &movups (&QWP(16*3,$out),$inout3); + &movups (&QWP(16*4,$out),$inout4); + &lea ($out,&DWP(16*5,$out)); + &jmp (&label("xts_enc_done")); + +&set_label("xts_enc_one",16); + &movups ($inout0,&QWP(16*0,$inp)); # load input + &lea ($inp,&DWP(16*1,$inp)); + &xorps ($inout0,$inout3); # input^=tweak + if ($inline) + { &aesni_inline_generate1("enc"); } + else + { &call ("_aesni_encrypt1"); } + &xorps ($inout0,$inout3); # output^=tweak + &movups (&QWP(16*0,$out),$inout0); # write output + &lea ($out,&DWP(16*1,$out)); + + &movdqa ($tweak,$inout3); # last tweak + &jmp (&label("xts_enc_done")); + +&set_label("xts_enc_two",16); + &movaps ($inout4,$tweak); # put aside last tweak + + &movups ($inout0,&QWP(16*0,$inp)); # load input + &movups ($inout1,&QWP(16*1,$inp)); + &lea ($inp,&DWP(16*2,$inp)); + &xorps ($inout0,$inout3); # input^=tweak + &xorps ($inout1,$inout4); + + &call ("_aesni_encrypt2"); + + &xorps ($inout0,$inout3); # output^=tweak + &xorps ($inout1,$inout4); + &movups (&QWP(16*0,$out),$inout0); # write output + &movups (&QWP(16*1,$out),$inout1); + &lea ($out,&DWP(16*2,$out)); + + &movdqa ($tweak,$inout4); # last tweak + &jmp (&label("xts_enc_done")); + +&set_label("xts_enc_three",16); + &movaps ($inout5,$tweak); # put aside last tweak + &movups ($inout0,&QWP(16*0,$inp)); # load input + &movups ($inout1,&QWP(16*1,$inp)); + &movups ($inout2,&QWP(16*2,$inp)); + &lea ($inp,&DWP(16*3,$inp)); + &xorps ($inout0,$inout3); # input^=tweak + &xorps ($inout1,$inout4); + &xorps ($inout2,$inout5); + + &call ("_aesni_encrypt3"); + + &xorps ($inout0,$inout3); # output^=tweak + &xorps ($inout1,$inout4); + &xorps ($inout2,$inout5); + &movups (&QWP(16*0,$out),$inout0); # write output + &movups (&QWP(16*1,$out),$inout1); + &movups (&QWP(16*2,$out),$inout2); + &lea ($out,&DWP(16*3,$out)); + + &movdqa ($tweak,$inout5); # last tweak + &jmp (&label("xts_enc_done")); + +&set_label("xts_enc_four",16); + &movaps ($inout4,$tweak); # put aside last tweak + + &movups ($inout0,&QWP(16*0,$inp)); # load input + &movups ($inout1,&QWP(16*1,$inp)); + &movups ($inout2,&QWP(16*2,$inp)); + &xorps ($inout0,&QWP(16*0,"esp")); # input^=tweak + &movups ($inout3,&QWP(16*3,$inp)); + &lea ($inp,&DWP(16*4,$inp)); + &xorps ($inout1,&QWP(16*1,"esp")); + &xorps ($inout2,$inout5); + &xorps ($inout3,$inout4); + + &call ("_aesni_encrypt4"); + + &xorps ($inout0,&QWP(16*0,"esp")); # output^=tweak + &xorps ($inout1,&QWP(16*1,"esp")); + &xorps ($inout2,$inout5); + &movups (&QWP(16*0,$out),$inout0); # write output + &xorps ($inout3,$inout4); + &movups (&QWP(16*1,$out),$inout1); + &movups (&QWP(16*2,$out),$inout2); + &movups (&QWP(16*3,$out),$inout3); + &lea ($out,&DWP(16*4,$out)); + + &movdqa ($tweak,$inout4); # last tweak + &jmp (&label("xts_enc_done")); + +&set_label("xts_enc_done6x",16); # $tweak is pre-calculated + &mov ($len,&DWP(16*7+0,"esp")); # restore original $len + &and ($len,15); + &jz (&label("xts_enc_ret")); + &movdqa ($inout3,$tweak); + &mov (&DWP(16*7+0,"esp"),$len); # save $len%16 + &jmp (&label("xts_enc_steal")); + +&set_label("xts_enc_done",16); + &mov ($len,&DWP(16*7+0,"esp")); # restore original $len + &pxor ($twtmp,$twtmp); + &and ($len,15); + &jz (&label("xts_enc_ret")); + + &pcmpgtd($twtmp,$tweak); # broadcast upper bits + &mov (&DWP(16*7+0,"esp"),$len); # save $len%16 + &pshufd ($inout3,$twtmp,0x13); + &paddq ($tweak,$tweak); # &psllq($tweak,1); + &pand ($inout3,&QWP(16*6,"esp")); # isolate carry and residue + &pxor ($inout3,$tweak); + +&set_label("xts_enc_steal"); + &movz ($rounds,&BP(0,$inp)); + &movz ($key,&BP(-16,$out)); + &lea ($inp,&DWP(1,$inp)); + &mov (&BP(-16,$out),&LB($rounds)); + &mov (&BP(0,$out),&LB($key)); + &lea ($out,&DWP(1,$out)); + &sub ($len,1); + &jnz (&label("xts_enc_steal")); + + &sub ($out,&DWP(16*7+0,"esp")); # rewind $out + &mov ($key,$key_); # restore $key + &mov ($rounds,$rounds_); # restore $rounds + + &movups ($inout0,&QWP(-16,$out)); # load input + &xorps ($inout0,$inout3); # input^=tweak + if ($inline) + { &aesni_inline_generate1("enc"); } + else + { &call ("_aesni_encrypt1"); } + &xorps ($inout0,$inout3); # output^=tweak + &movups (&QWP(-16,$out),$inout0); # write output + +&set_label("xts_enc_ret"); + &pxor ("xmm0","xmm0"); # clear register bank + &pxor ("xmm1","xmm1"); + &pxor ("xmm2","xmm2"); + &movdqa (&QWP(16*0,"esp"),"xmm0"); # clear stack + &pxor ("xmm3","xmm3"); + &movdqa (&QWP(16*1,"esp"),"xmm0"); + &pxor ("xmm4","xmm4"); + &movdqa (&QWP(16*2,"esp"),"xmm0"); + &pxor ("xmm5","xmm5"); + &movdqa (&QWP(16*3,"esp"),"xmm0"); + &pxor ("xmm6","xmm6"); + &movdqa (&QWP(16*4,"esp"),"xmm0"); + &pxor ("xmm7","xmm7"); + &movdqa (&QWP(16*5,"esp"),"xmm0"); + &mov ("esp",&DWP(16*7+4,"esp")); # restore %esp +&function_end("aesni_xts_encrypt"); + +&function_begin("aesni_xts_decrypt"); + &mov ($key,&wparam(4)); # key2 + &mov ($inp,&wparam(5)); # clear-text tweak + + &mov ($rounds,&DWP(240,$key)); # key2->rounds + &movups ($inout0,&QWP(0,$inp)); + if ($inline) + { &aesni_inline_generate1("enc"); } + else + { &call ("_aesni_encrypt1"); } + + &mov ($inp,&wparam(0)); + &mov ($out,&wparam(1)); + &mov ($len,&wparam(2)); + &mov ($key,&wparam(3)); # key1 + + &mov ($key_,"esp"); + &sub ("esp",16*7+8); + &and ("esp",-16); # align stack + + &xor ($rounds_,$rounds_); # if(len%16) len-=16; + &test ($len,15); + &setnz (&LB($rounds_)); + &shl ($rounds_,4); + &sub ($len,$rounds_); + + &mov (&DWP(16*6+0,"esp"),0x87); # compose the magic constant + &mov (&DWP(16*6+4,"esp"),0); + &mov (&DWP(16*6+8,"esp"),1); + &mov (&DWP(16*6+12,"esp"),0); + &mov (&DWP(16*7+0,"esp"),$len); # save original $len + &mov (&DWP(16*7+4,"esp"),$key_); # save original %esp + + &mov ($rounds,&DWP(240,$key)); # key1->rounds + &mov ($key_,$key); # backup $key + &mov ($rounds_,$rounds); # backup $rounds + + &movdqa ($tweak,$inout0); + &pxor ($twtmp,$twtmp); + &movdqa ($twmask,&QWP(6*16,"esp")); # 0x0...010...87 + &pcmpgtd($twtmp,$tweak); # broadcast upper bits + + &and ($len,-16); + &sub ($len,16*6); + &jc (&label("xts_dec_short")); + + &shl ($rounds,4); + &mov ($rounds_,16); + &sub ($rounds_,$rounds); + &lea ($key,&DWP(32,$key,$rounds)); + &jmp (&label("xts_dec_loop6")); + +&set_label("xts_dec_loop6",16); + for ($i=0;$i<4;$i++) { + &pshufd ($twres,$twtmp,0x13); + &pxor ($twtmp,$twtmp); + &movdqa (&QWP(16*$i,"esp"),$tweak); + &paddq ($tweak,$tweak); # &psllq($tweak,1); + &pand ($twres,$twmask); # isolate carry and residue + &pcmpgtd ($twtmp,$tweak); # broadcast upper bits + &pxor ($tweak,$twres); + } + &pshufd ($inout5,$twtmp,0x13); + &movdqa (&QWP(16*$i++,"esp"),$tweak); + &paddq ($tweak,$tweak); # &psllq($tweak,1); + &$movekey ($rndkey0,&QWP(0,$key_)); + &pand ($inout5,$twmask); # isolate carry and residue + &movups ($inout0,&QWP(0,$inp)); # load input + &pxor ($inout5,$tweak); + + # inline _aesni_encrypt6 prologue and flip xor with tweak and key[0] + &mov ($rounds,$rounds_); + &movdqu ($inout1,&QWP(16*1,$inp)); + &xorps ($inout0,$rndkey0); # input^=rndkey[0] + &movdqu ($inout2,&QWP(16*2,$inp)); + &pxor ($inout1,$rndkey0); + &movdqu ($inout3,&QWP(16*3,$inp)); + &pxor ($inout2,$rndkey0); + &movdqu ($inout4,&QWP(16*4,$inp)); + &pxor ($inout3,$rndkey0); + &movdqu ($rndkey1,&QWP(16*5,$inp)); + &pxor ($inout4,$rndkey0); + &lea ($inp,&DWP(16*6,$inp)); + &pxor ($inout0,&QWP(16*0,"esp")); # input^=tweak + &movdqa (&QWP(16*$i,"esp"),$inout5); # save last tweak + &pxor ($inout5,$rndkey1); + + &$movekey ($rndkey1,&QWP(16,$key_)); + &pxor ($inout1,&QWP(16*1,"esp")); + &pxor ($inout2,&QWP(16*2,"esp")); + &aesdec ($inout0,$rndkey1); + &pxor ($inout3,&QWP(16*3,"esp")); + &pxor ($inout4,&QWP(16*4,"esp")); + &aesdec ($inout1,$rndkey1); + &pxor ($inout5,$rndkey0); + &$movekey ($rndkey0,&QWP(32,$key_)); + &aesdec ($inout2,$rndkey1); + &aesdec ($inout3,$rndkey1); + &aesdec ($inout4,$rndkey1); + &aesdec ($inout5,$rndkey1); + &call (&label("_aesni_decrypt6_enter")); + + &movdqa ($tweak,&QWP(16*5,"esp")); # last tweak + &pxor ($twtmp,$twtmp); + &xorps ($inout0,&QWP(16*0,"esp")); # output^=tweak + &pcmpgtd ($twtmp,$tweak); # broadcast upper bits + &xorps ($inout1,&QWP(16*1,"esp")); + &movups (&QWP(16*0,$out),$inout0); # write output + &xorps ($inout2,&QWP(16*2,"esp")); + &movups (&QWP(16*1,$out),$inout1); + &xorps ($inout3,&QWP(16*3,"esp")); + &movups (&QWP(16*2,$out),$inout2); + &xorps ($inout4,&QWP(16*4,"esp")); + &movups (&QWP(16*3,$out),$inout3); + &xorps ($inout5,$tweak); + &movups (&QWP(16*4,$out),$inout4); + &pshufd ($twres,$twtmp,0x13); + &movups (&QWP(16*5,$out),$inout5); + &lea ($out,&DWP(16*6,$out)); + &movdqa ($twmask,&QWP(16*6,"esp")); # 0x0...010...87 + + &pxor ($twtmp,$twtmp); + &paddq ($tweak,$tweak); # &psllq($tweak,1); + &pand ($twres,$twmask); # isolate carry and residue + &pcmpgtd($twtmp,$tweak); # broadcast upper bits + &pxor ($tweak,$twres); + + &sub ($len,16*6); + &jnc (&label("xts_dec_loop6")); + + &mov ($rounds,&DWP(240,$key_)); # restore $rounds + &mov ($key,$key_); # restore $key + &mov ($rounds_,$rounds); + +&set_label("xts_dec_short"); + &add ($len,16*6); + &jz (&label("xts_dec_done6x")); + + &movdqa ($inout3,$tweak); # put aside previous tweak + &cmp ($len,0x20); + &jb (&label("xts_dec_one")); + + &pshufd ($twres,$twtmp,0x13); + &pxor ($twtmp,$twtmp); + &paddq ($tweak,$tweak); # &psllq($tweak,1); + &pand ($twres,$twmask); # isolate carry and residue + &pcmpgtd($twtmp,$tweak); # broadcast upper bits + &pxor ($tweak,$twres); + &je (&label("xts_dec_two")); + + &pshufd ($twres,$twtmp,0x13); + &pxor ($twtmp,$twtmp); + &movdqa ($inout4,$tweak); # put aside previous tweak + &paddq ($tweak,$tweak); # &psllq($tweak,1); + &pand ($twres,$twmask); # isolate carry and residue + &pcmpgtd($twtmp,$tweak); # broadcast upper bits + &pxor ($tweak,$twres); + &cmp ($len,0x40); + &jb (&label("xts_dec_three")); + + &pshufd ($twres,$twtmp,0x13); + &pxor ($twtmp,$twtmp); + &movdqa ($inout5,$tweak); # put aside previous tweak + &paddq ($tweak,$tweak); # &psllq($tweak,1); + &pand ($twres,$twmask); # isolate carry and residue + &pcmpgtd($twtmp,$tweak); # broadcast upper bits + &pxor ($tweak,$twres); + &movdqa (&QWP(16*0,"esp"),$inout3); + &movdqa (&QWP(16*1,"esp"),$inout4); + &je (&label("xts_dec_four")); + + &movdqa (&QWP(16*2,"esp"),$inout5); + &pshufd ($inout5,$twtmp,0x13); + &movdqa (&QWP(16*3,"esp"),$tweak); + &paddq ($tweak,$tweak); # &psllq($inout0,1); + &pand ($inout5,$twmask); # isolate carry and residue + &pxor ($inout5,$tweak); + + &movdqu ($inout0,&QWP(16*0,$inp)); # load input + &movdqu ($inout1,&QWP(16*1,$inp)); + &movdqu ($inout2,&QWP(16*2,$inp)); + &pxor ($inout0,&QWP(16*0,"esp")); # input^=tweak + &movdqu ($inout3,&QWP(16*3,$inp)); + &pxor ($inout1,&QWP(16*1,"esp")); + &movdqu ($inout4,&QWP(16*4,$inp)); + &pxor ($inout2,&QWP(16*2,"esp")); + &lea ($inp,&DWP(16*5,$inp)); + &pxor ($inout3,&QWP(16*3,"esp")); + &movdqa (&QWP(16*4,"esp"),$inout5); # save last tweak + &pxor ($inout4,$inout5); + + &call ("_aesni_decrypt6"); + + &movaps ($tweak,&QWP(16*4,"esp")); # last tweak + &xorps ($inout0,&QWP(16*0,"esp")); # output^=tweak + &xorps ($inout1,&QWP(16*1,"esp")); + &xorps ($inout2,&QWP(16*2,"esp")); + &movups (&QWP(16*0,$out),$inout0); # write output + &xorps ($inout3,&QWP(16*3,"esp")); + &movups (&QWP(16*1,$out),$inout1); + &xorps ($inout4,$tweak); + &movups (&QWP(16*2,$out),$inout2); + &movups (&QWP(16*3,$out),$inout3); + &movups (&QWP(16*4,$out),$inout4); + &lea ($out,&DWP(16*5,$out)); + &jmp (&label("xts_dec_done")); + +&set_label("xts_dec_one",16); + &movups ($inout0,&QWP(16*0,$inp)); # load input + &lea ($inp,&DWP(16*1,$inp)); + &xorps ($inout0,$inout3); # input^=tweak + if ($inline) + { &aesni_inline_generate1("dec"); } + else + { &call ("_aesni_decrypt1"); } + &xorps ($inout0,$inout3); # output^=tweak + &movups (&QWP(16*0,$out),$inout0); # write output + &lea ($out,&DWP(16*1,$out)); + + &movdqa ($tweak,$inout3); # last tweak + &jmp (&label("xts_dec_done")); + +&set_label("xts_dec_two",16); + &movaps ($inout4,$tweak); # put aside last tweak + + &movups ($inout0,&QWP(16*0,$inp)); # load input + &movups ($inout1,&QWP(16*1,$inp)); + &lea ($inp,&DWP(16*2,$inp)); + &xorps ($inout0,$inout3); # input^=tweak + &xorps ($inout1,$inout4); + + &call ("_aesni_decrypt2"); + + &xorps ($inout0,$inout3); # output^=tweak + &xorps ($inout1,$inout4); + &movups (&QWP(16*0,$out),$inout0); # write output + &movups (&QWP(16*1,$out),$inout1); + &lea ($out,&DWP(16*2,$out)); + + &movdqa ($tweak,$inout4); # last tweak + &jmp (&label("xts_dec_done")); + +&set_label("xts_dec_three",16); + &movaps ($inout5,$tweak); # put aside last tweak + &movups ($inout0,&QWP(16*0,$inp)); # load input + &movups ($inout1,&QWP(16*1,$inp)); + &movups ($inout2,&QWP(16*2,$inp)); + &lea ($inp,&DWP(16*3,$inp)); + &xorps ($inout0,$inout3); # input^=tweak + &xorps ($inout1,$inout4); + &xorps ($inout2,$inout5); + + &call ("_aesni_decrypt3"); + + &xorps ($inout0,$inout3); # output^=tweak + &xorps ($inout1,$inout4); + &xorps ($inout2,$inout5); + &movups (&QWP(16*0,$out),$inout0); # write output + &movups (&QWP(16*1,$out),$inout1); + &movups (&QWP(16*2,$out),$inout2); + &lea ($out,&DWP(16*3,$out)); + + &movdqa ($tweak,$inout5); # last tweak + &jmp (&label("xts_dec_done")); + +&set_label("xts_dec_four",16); + &movaps ($inout4,$tweak); # put aside last tweak + + &movups ($inout0,&QWP(16*0,$inp)); # load input + &movups ($inout1,&QWP(16*1,$inp)); + &movups ($inout2,&QWP(16*2,$inp)); + &xorps ($inout0,&QWP(16*0,"esp")); # input^=tweak + &movups ($inout3,&QWP(16*3,$inp)); + &lea ($inp,&DWP(16*4,$inp)); + &xorps ($inout1,&QWP(16*1,"esp")); + &xorps ($inout2,$inout5); + &xorps ($inout3,$inout4); + + &call ("_aesni_decrypt4"); + + &xorps ($inout0,&QWP(16*0,"esp")); # output^=tweak + &xorps ($inout1,&QWP(16*1,"esp")); + &xorps ($inout2,$inout5); + &movups (&QWP(16*0,$out),$inout0); # write output + &xorps ($inout3,$inout4); + &movups (&QWP(16*1,$out),$inout1); + &movups (&QWP(16*2,$out),$inout2); + &movups (&QWP(16*3,$out),$inout3); + &lea ($out,&DWP(16*4,$out)); + + &movdqa ($tweak,$inout4); # last tweak + &jmp (&label("xts_dec_done")); + +&set_label("xts_dec_done6x",16); # $tweak is pre-calculated + &mov ($len,&DWP(16*7+0,"esp")); # restore original $len + &and ($len,15); + &jz (&label("xts_dec_ret")); + &mov (&DWP(16*7+0,"esp"),$len); # save $len%16 + &jmp (&label("xts_dec_only_one_more")); + +&set_label("xts_dec_done",16); + &mov ($len,&DWP(16*7+0,"esp")); # restore original $len + &pxor ($twtmp,$twtmp); + &and ($len,15); + &jz (&label("xts_dec_ret")); + + &pcmpgtd($twtmp,$tweak); # broadcast upper bits + &mov (&DWP(16*7+0,"esp"),$len); # save $len%16 + &pshufd ($twres,$twtmp,0x13); + &pxor ($twtmp,$twtmp); + &movdqa ($twmask,&QWP(16*6,"esp")); + &paddq ($tweak,$tweak); # &psllq($tweak,1); + &pand ($twres,$twmask); # isolate carry and residue + &pcmpgtd($twtmp,$tweak); # broadcast upper bits + &pxor ($tweak,$twres); + +&set_label("xts_dec_only_one_more"); + &pshufd ($inout3,$twtmp,0x13); + &movdqa ($inout4,$tweak); # put aside previous tweak + &paddq ($tweak,$tweak); # &psllq($tweak,1); + &pand ($inout3,$twmask); # isolate carry and residue + &pxor ($inout3,$tweak); + + &mov ($key,$key_); # restore $key + &mov ($rounds,$rounds_); # restore $rounds + + &movups ($inout0,&QWP(0,$inp)); # load input + &xorps ($inout0,$inout3); # input^=tweak + if ($inline) + { &aesni_inline_generate1("dec"); } + else + { &call ("_aesni_decrypt1"); } + &xorps ($inout0,$inout3); # output^=tweak + &movups (&QWP(0,$out),$inout0); # write output + +&set_label("xts_dec_steal"); + &movz ($rounds,&BP(16,$inp)); + &movz ($key,&BP(0,$out)); + &lea ($inp,&DWP(1,$inp)); + &mov (&BP(0,$out),&LB($rounds)); + &mov (&BP(16,$out),&LB($key)); + &lea ($out,&DWP(1,$out)); + &sub ($len,1); + &jnz (&label("xts_dec_steal")); + + &sub ($out,&DWP(16*7+0,"esp")); # rewind $out + &mov ($key,$key_); # restore $key + &mov ($rounds,$rounds_); # restore $rounds + + &movups ($inout0,&QWP(0,$out)); # load input + &xorps ($inout0,$inout4); # input^=tweak + if ($inline) + { &aesni_inline_generate1("dec"); } + else + { &call ("_aesni_decrypt1"); } + &xorps ($inout0,$inout4); # output^=tweak + &movups (&QWP(0,$out),$inout0); # write output + +&set_label("xts_dec_ret"); + &pxor ("xmm0","xmm0"); # clear register bank + &pxor ("xmm1","xmm1"); + &pxor ("xmm2","xmm2"); + &movdqa (&QWP(16*0,"esp"),"xmm0"); # clear stack + &pxor ("xmm3","xmm3"); + &movdqa (&QWP(16*1,"esp"),"xmm0"); + &pxor ("xmm4","xmm4"); + &movdqa (&QWP(16*2,"esp"),"xmm0"); + &pxor ("xmm5","xmm5"); + &movdqa (&QWP(16*3,"esp"),"xmm0"); + &pxor ("xmm6","xmm6"); + &movdqa (&QWP(16*4,"esp"),"xmm0"); + &pxor ("xmm7","xmm7"); + &movdqa (&QWP(16*5,"esp"),"xmm0"); + &mov ("esp",&DWP(16*7+4,"esp")); # restore %esp +&function_end("aesni_xts_decrypt"); +} + +###################################################################### +# void aesni_ocb_[en|de]crypt(const char *inp, char *out, size_t blocks, +# const AES_KEY *key, unsigned int start_block_num, +# unsigned char offset_i[16], const unsigned char L_[][16], +# unsigned char checksum[16]); +# +{ +# offsets within stack frame +my $checksum = 16*6; +my ($key_off,$rounds_off,$out_off,$end_off,$esp_off)=map(16*7+4*$_,(0..4)); + +# reassigned registers +my ($l_,$block,$i1,$i3,$i5) = ($rounds_,$key_,$rounds,$len,$out); +# $l_, $blocks, $inp, $key are permanently allocated in registers; +# remaining non-volatile ones are offloaded to stack, which even +# stay invariant after written to stack. + +&function_begin("aesni_ocb_encrypt"); + &mov ($rounds,&wparam(5)); # &offset_i + &mov ($rounds_,&wparam(7)); # &checksum + + &mov ($inp,&wparam(0)); + &mov ($out,&wparam(1)); + &mov ($len,&wparam(2)); + &mov ($key,&wparam(3)); + &movdqu ($rndkey0,&QWP(0,$rounds)); # load offset_i + &mov ($block,&wparam(4)); # start_block_num + &movdqu ($rndkey1,&QWP(0,$rounds_)); # load checksum + &mov ($l_,&wparam(6)); # L_ + + &mov ($rounds,"esp"); + &sub ("esp",$esp_off+4); # alloca + &and ("esp",-16); # align stack + + &sub ($out,$inp); + &shl ($len,4); + &lea ($len,&DWP(-16*6,$inp,$len)); # end of input - 16*6 + &mov (&DWP($out_off,"esp"),$out); + &mov (&DWP($end_off,"esp"),$len); + &mov (&DWP($esp_off,"esp"),$rounds); + + &mov ($rounds,&DWP(240,$key)); + + &test ($block,1); + &jnz (&label("odd")); + + &bsf ($i3,$block); + &add ($block,1); + &shl ($i3,4); + &movdqu ($inout5,&QWP(0,$l_,$i3)); + &mov ($i3,$key); # put aside key + + &movdqu ($inout0,&QWP(16*0,$inp)); # load input + &lea ($inp,&DWP(16,$inp)); + + &pxor ($inout5,$rndkey0); # ^ last offset_i + &pxor ($rndkey1,$inout0); # checksum + &pxor ($inout0,$inout5); # ^ offset_i + + &movdqa ($inout4,$rndkey1); + if ($inline) + { &aesni_inline_generate1("enc"); } + else + { &call ("_aesni_encrypt1"); } + + &xorps ($inout0,$inout5); # ^ offset_i + &movdqa ($rndkey0,$inout5); # pass last offset_i + &movdqa ($rndkey1,$inout4); # pass the checksum + + &movups (&QWP(-16,$out,$inp),$inout0); # store output + + &mov ($rounds,&DWP(240,$i3)); + &mov ($key,$i3); # restore key + &mov ($len,&DWP($end_off,"esp")); + +&set_label("odd"); + &shl ($rounds,4); + &mov ($out,16); + &sub ($out,$rounds); # twisted rounds + &mov (&DWP($key_off,"esp"),$key); + &lea ($key,&DWP(32,$key,$rounds)); # end of key schedule + &mov (&DWP($rounds_off,"esp"),$out); + + &cmp ($inp,$len); + &ja (&label("short")); + &jmp (&label("grandloop")); + +&set_label("grandloop",32); + &lea ($i1,&DWP(1,$block)); + &lea ($i3,&DWP(3,$block)); + &lea ($i5,&DWP(5,$block)); + &add ($block,6); + &bsf ($i1,$i1); + &bsf ($i3,$i3); + &bsf ($i5,$i5); + &shl ($i1,4); + &shl ($i3,4); + &shl ($i5,4); + &movdqu ($inout0,&QWP(0,$l_)); + &movdqu ($inout1,&QWP(0,$l_,$i1)); + &mov ($rounds,&DWP($rounds_off,"esp")); + &movdqa ($inout2,$inout0); + &movdqu ($inout3,&QWP(0,$l_,$i3)); + &movdqa ($inout4,$inout0); + &movdqu ($inout5,&QWP(0,$l_,$i5)); + + &pxor ($inout0,$rndkey0); # ^ last offset_i + &pxor ($inout1,$inout0); + &movdqa (&QWP(16*0,"esp"),$inout0); + &pxor ($inout2,$inout1); + &movdqa (&QWP(16*1,"esp"),$inout1); + &pxor ($inout3,$inout2); + &movdqa (&QWP(16*2,"esp"),$inout2); + &pxor ($inout4,$inout3); + &movdqa (&QWP(16*3,"esp"),$inout3); + &pxor ($inout5,$inout4); + &movdqa (&QWP(16*4,"esp"),$inout4); + &movdqa (&QWP(16*5,"esp"),$inout5); + + &$movekey ($rndkey0,&QWP(-48,$key,$rounds)); + &movdqu ($inout0,&QWP(16*0,$inp)); # load input + &movdqu ($inout1,&QWP(16*1,$inp)); + &movdqu ($inout2,&QWP(16*2,$inp)); + &movdqu ($inout3,&QWP(16*3,$inp)); + &movdqu ($inout4,&QWP(16*4,$inp)); + &movdqu ($inout5,&QWP(16*5,$inp)); + &lea ($inp,&DWP(16*6,$inp)); + + &pxor ($rndkey1,$inout0); # checksum + &pxor ($inout0,$rndkey0); # ^ roundkey[0] + &pxor ($rndkey1,$inout1); + &pxor ($inout1,$rndkey0); + &pxor ($rndkey1,$inout2); + &pxor ($inout2,$rndkey0); + &pxor ($rndkey1,$inout3); + &pxor ($inout3,$rndkey0); + &pxor ($rndkey1,$inout4); + &pxor ($inout4,$rndkey0); + &pxor ($rndkey1,$inout5); + &pxor ($inout5,$rndkey0); + &movdqa (&QWP($checksum,"esp"),$rndkey1); + + &$movekey ($rndkey1,&QWP(-32,$key,$rounds)); + &pxor ($inout0,&QWP(16*0,"esp")); # ^ offset_i + &pxor ($inout1,&QWP(16*1,"esp")); + &pxor ($inout2,&QWP(16*2,"esp")); + &pxor ($inout3,&QWP(16*3,"esp")); + &pxor ($inout4,&QWP(16*4,"esp")); + &pxor ($inout5,&QWP(16*5,"esp")); + + &$movekey ($rndkey0,&QWP(-16,$key,$rounds)); + &aesenc ($inout0,$rndkey1); + &aesenc ($inout1,$rndkey1); + &aesenc ($inout2,$rndkey1); + &aesenc ($inout3,$rndkey1); + &aesenc ($inout4,$rndkey1); + &aesenc ($inout5,$rndkey1); + + &mov ($out,&DWP($out_off,"esp")); + &mov ($len,&DWP($end_off,"esp")); + &call ("_aesni_encrypt6_enter"); + + &movdqa ($rndkey0,&QWP(16*5,"esp")); # pass last offset_i + &pxor ($inout0,&QWP(16*0,"esp")); # ^ offset_i + &pxor ($inout1,&QWP(16*1,"esp")); + &pxor ($inout2,&QWP(16*2,"esp")); + &pxor ($inout3,&QWP(16*3,"esp")); + &pxor ($inout4,&QWP(16*4,"esp")); + &pxor ($inout5,$rndkey0); + &movdqa ($rndkey1,&QWP($checksum,"esp"));# pass the checksum + + &movdqu (&QWP(-16*6,$out,$inp),$inout0);# store output + &movdqu (&QWP(-16*5,$out,$inp),$inout1); + &movdqu (&QWP(-16*4,$out,$inp),$inout2); + &movdqu (&QWP(-16*3,$out,$inp),$inout3); + &movdqu (&QWP(-16*2,$out,$inp),$inout4); + &movdqu (&QWP(-16*1,$out,$inp),$inout5); + &cmp ($inp,$len); # done yet? + &jb (&label("grandloop")); + +&set_label("short"); + &add ($len,16*6); + &sub ($len,$inp); + &jz (&label("done")); + + &cmp ($len,16*2); + &jb (&label("one")); + &je (&label("two")); + + &cmp ($len,16*4); + &jb (&label("three")); + &je (&label("four")); + + &lea ($i1,&DWP(1,$block)); + &lea ($i3,&DWP(3,$block)); + &bsf ($i1,$i1); + &bsf ($i3,$i3); + &shl ($i1,4); + &shl ($i3,4); + &movdqu ($inout0,&QWP(0,$l_)); + &movdqu ($inout1,&QWP(0,$l_,$i1)); + &mov ($rounds,&DWP($rounds_off,"esp")); + &movdqa ($inout2,$inout0); + &movdqu ($inout3,&QWP(0,$l_,$i3)); + &movdqa ($inout4,$inout0); + + &pxor ($inout0,$rndkey0); # ^ last offset_i + &pxor ($inout1,$inout0); + &movdqa (&QWP(16*0,"esp"),$inout0); + &pxor ($inout2,$inout1); + &movdqa (&QWP(16*1,"esp"),$inout1); + &pxor ($inout3,$inout2); + &movdqa (&QWP(16*2,"esp"),$inout2); + &pxor ($inout4,$inout3); + &movdqa (&QWP(16*3,"esp"),$inout3); + &pxor ($inout5,$inout4); + &movdqa (&QWP(16*4,"esp"),$inout4); + + &$movekey ($rndkey0,&QWP(-48,$key,$rounds)); + &movdqu ($inout0,&QWP(16*0,$inp)); # load input + &movdqu ($inout1,&QWP(16*1,$inp)); + &movdqu ($inout2,&QWP(16*2,$inp)); + &movdqu ($inout3,&QWP(16*3,$inp)); + &movdqu ($inout4,&QWP(16*4,$inp)); + &pxor ($inout5,$inout5); + + &pxor ($rndkey1,$inout0); # checksum + &pxor ($inout0,$rndkey0); # ^ roundkey[0] + &pxor ($rndkey1,$inout1); + &pxor ($inout1,$rndkey0); + &pxor ($rndkey1,$inout2); + &pxor ($inout2,$rndkey0); + &pxor ($rndkey1,$inout3); + &pxor ($inout3,$rndkey0); + &pxor ($rndkey1,$inout4); + &pxor ($inout4,$rndkey0); + &movdqa (&QWP($checksum,"esp"),$rndkey1); + + &$movekey ($rndkey1,&QWP(-32,$key,$rounds)); + &pxor ($inout0,&QWP(16*0,"esp")); # ^ offset_i + &pxor ($inout1,&QWP(16*1,"esp")); + &pxor ($inout2,&QWP(16*2,"esp")); + &pxor ($inout3,&QWP(16*3,"esp")); + &pxor ($inout4,&QWP(16*4,"esp")); + + &$movekey ($rndkey0,&QWP(-16,$key,$rounds)); + &aesenc ($inout0,$rndkey1); + &aesenc ($inout1,$rndkey1); + &aesenc ($inout2,$rndkey1); + &aesenc ($inout3,$rndkey1); + &aesenc ($inout4,$rndkey1); + &aesenc ($inout5,$rndkey1); + + &mov ($out,&DWP($out_off,"esp")); + &call ("_aesni_encrypt6_enter"); + + &movdqa ($rndkey0,&QWP(16*4,"esp")); # pass last offset_i + &pxor ($inout0,&QWP(16*0,"esp")); # ^ offset_i + &pxor ($inout1,&QWP(16*1,"esp")); + &pxor ($inout2,&QWP(16*2,"esp")); + &pxor ($inout3,&QWP(16*3,"esp")); + &pxor ($inout4,$rndkey0); + &movdqa ($rndkey1,&QWP($checksum,"esp"));# pass the checksum + + &movdqu (&QWP(16*0,$out,$inp),$inout0); # store output + &movdqu (&QWP(16*1,$out,$inp),$inout1); + &movdqu (&QWP(16*2,$out,$inp),$inout2); + &movdqu (&QWP(16*3,$out,$inp),$inout3); + &movdqu (&QWP(16*4,$out,$inp),$inout4); + + &jmp (&label("done")); + +&set_label("one",16); + &movdqu ($inout5,&QWP(0,$l_)); + &mov ($key,&DWP($key_off,"esp")); # restore key + + &movdqu ($inout0,&QWP(16*0,$inp)); # load input + &mov ($rounds,&DWP(240,$key)); + + &pxor ($inout5,$rndkey0); # ^ last offset_i + &pxor ($rndkey1,$inout0); # checksum + &pxor ($inout0,$inout5); # ^ offset_i + + &movdqa ($inout4,$rndkey1); + &mov ($out,&DWP($out_off,"esp")); + if ($inline) + { &aesni_inline_generate1("enc"); } + else + { &call ("_aesni_encrypt1"); } + + &xorps ($inout0,$inout5); # ^ offset_i + &movdqa ($rndkey0,$inout5); # pass last offset_i + &movdqa ($rndkey1,$inout4); # pass the checksum + &movups (&QWP(0,$out,$inp),$inout0); + + &jmp (&label("done")); + +&set_label("two",16); + &lea ($i1,&DWP(1,$block)); + &mov ($key,&DWP($key_off,"esp")); # restore key + &bsf ($i1,$i1); + &shl ($i1,4); + &movdqu ($inout4,&QWP(0,$l_)); + &movdqu ($inout5,&QWP(0,$l_,$i1)); + + &movdqu ($inout0,&QWP(16*0,$inp)); # load input + &movdqu ($inout1,&QWP(16*1,$inp)); + &mov ($rounds,&DWP(240,$key)); + + &pxor ($inout4,$rndkey0); # ^ last offset_i + &pxor ($inout5,$inout4); + + &pxor ($rndkey1,$inout0); # checksum + &pxor ($inout0,$inout4); # ^ offset_i + &pxor ($rndkey1,$inout1); + &pxor ($inout1,$inout5); + + &movdqa ($inout3,$rndkey1) + &mov ($out,&DWP($out_off,"esp")); + &call ("_aesni_encrypt2"); + + &xorps ($inout0,$inout4); # ^ offset_i + &xorps ($inout1,$inout5); + &movdqa ($rndkey0,$inout5); # pass last offset_i + &movdqa ($rndkey1,$inout3); # pass the checksum + &movups (&QWP(16*0,$out,$inp),$inout0); # store output + &movups (&QWP(16*1,$out,$inp),$inout1); + + &jmp (&label("done")); + +&set_label("three",16); + &lea ($i1,&DWP(1,$block)); + &mov ($key,&DWP($key_off,"esp")); # restore key + &bsf ($i1,$i1); + &shl ($i1,4); + &movdqu ($inout3,&QWP(0,$l_)); + &movdqu ($inout4,&QWP(0,$l_,$i1)); + &movdqa ($inout5,$inout3); + + &movdqu ($inout0,&QWP(16*0,$inp)); # load input + &movdqu ($inout1,&QWP(16*1,$inp)); + &movdqu ($inout2,&QWP(16*2,$inp)); + &mov ($rounds,&DWP(240,$key)); + + &pxor ($inout3,$rndkey0); # ^ last offset_i + &pxor ($inout4,$inout3); + &pxor ($inout5,$inout4); + + &pxor ($rndkey1,$inout0); # checksum + &pxor ($inout0,$inout3); # ^ offset_i + &pxor ($rndkey1,$inout1); + &pxor ($inout1,$inout4); + &pxor ($rndkey1,$inout2); + &pxor ($inout2,$inout5); + + &movdqa (&QWP($checksum,"esp"),$rndkey1); + &mov ($out,&DWP($out_off,"esp")); + &call ("_aesni_encrypt3"); + + &xorps ($inout0,$inout3); # ^ offset_i + &xorps ($inout1,$inout4); + &xorps ($inout2,$inout5); + &movdqa ($rndkey0,$inout5); # pass last offset_i + &movdqa ($rndkey1,&QWP($checksum,"esp"));# pass the checksum + &movups (&QWP(16*0,$out,$inp),$inout0); # store output + &movups (&QWP(16*1,$out,$inp),$inout1); + &movups (&QWP(16*2,$out,$inp),$inout2); + + &jmp (&label("done")); + +&set_label("four",16); + &lea ($i1,&DWP(1,$block)); + &lea ($i3,&DWP(3,$block)); + &bsf ($i1,$i1); + &bsf ($i3,$i3); + &mov ($key,&DWP($key_off,"esp")); # restore key + &shl ($i1,4); + &shl ($i3,4); + &movdqu ($inout2,&QWP(0,$l_)); + &movdqu ($inout3,&QWP(0,$l_,$i1)); + &movdqa ($inout4,$inout2); + &movdqu ($inout5,&QWP(0,$l_,$i3)); + + &pxor ($inout2,$rndkey0); # ^ last offset_i + &movdqu ($inout0,&QWP(16*0,$inp)); # load input + &pxor ($inout3,$inout2); + &movdqu ($inout1,&QWP(16*1,$inp)); + &pxor ($inout4,$inout3); + &movdqa (&QWP(16*0,"esp"),$inout2); + &pxor ($inout5,$inout4); + &movdqa (&QWP(16*1,"esp"),$inout3); + &movdqu ($inout2,&QWP(16*2,$inp)); + &movdqu ($inout3,&QWP(16*3,$inp)); + &mov ($rounds,&DWP(240,$key)); + + &pxor ($rndkey1,$inout0); # checksum + &pxor ($inout0,&QWP(16*0,"esp")); # ^ offset_i + &pxor ($rndkey1,$inout1); + &pxor ($inout1,&QWP(16*1,"esp")); + &pxor ($rndkey1,$inout2); + &pxor ($inout2,$inout4); + &pxor ($rndkey1,$inout3); + &pxor ($inout3,$inout5); + + &movdqa (&QWP($checksum,"esp"),$rndkey1) + &mov ($out,&DWP($out_off,"esp")); + &call ("_aesni_encrypt4"); + + &xorps ($inout0,&QWP(16*0,"esp")); # ^ offset_i + &xorps ($inout1,&QWP(16*1,"esp")); + &xorps ($inout2,$inout4); + &movups (&QWP(16*0,$out,$inp),$inout0); # store output + &xorps ($inout3,$inout5); + &movups (&QWP(16*1,$out,$inp),$inout1); + &movdqa ($rndkey0,$inout5); # pass last offset_i + &movups (&QWP(16*2,$out,$inp),$inout2); + &movdqa ($rndkey1,&QWP($checksum,"esp"));# pass the checksum + &movups (&QWP(16*3,$out,$inp),$inout3); + +&set_label("done"); + &mov ($key,&DWP($esp_off,"esp")); + &pxor ($inout0,$inout0); # clear register bank + &pxor ($inout1,$inout1); + &movdqa (&QWP(16*0,"esp"),$inout0); # clear stack + &pxor ($inout2,$inout2); + &movdqa (&QWP(16*1,"esp"),$inout0); + &pxor ($inout3,$inout3); + &movdqa (&QWP(16*2,"esp"),$inout0); + &pxor ($inout4,$inout4); + &movdqa (&QWP(16*3,"esp"),$inout0); + &pxor ($inout5,$inout5); + &movdqa (&QWP(16*4,"esp"),$inout0); + &movdqa (&QWP(16*5,"esp"),$inout0); + &movdqa (&QWP(16*6,"esp"),$inout0); + + &lea ("esp",&DWP(0,$key)); + &mov ($rounds,&wparam(5)); # &offset_i + &mov ($rounds_,&wparam(7)); # &checksum + &movdqu (&QWP(0,$rounds),$rndkey0); + &pxor ($rndkey0,$rndkey0); + &movdqu (&QWP(0,$rounds_),$rndkey1); + &pxor ($rndkey1,$rndkey1); +&function_end("aesni_ocb_encrypt"); + +&function_begin("aesni_ocb_decrypt"); + &mov ($rounds,&wparam(5)); # &offset_i + &mov ($rounds_,&wparam(7)); # &checksum + + &mov ($inp,&wparam(0)); + &mov ($out,&wparam(1)); + &mov ($len,&wparam(2)); + &mov ($key,&wparam(3)); + &movdqu ($rndkey0,&QWP(0,$rounds)); # load offset_i + &mov ($block,&wparam(4)); # start_block_num + &movdqu ($rndkey1,&QWP(0,$rounds_)); # load checksum + &mov ($l_,&wparam(6)); # L_ + + &mov ($rounds,"esp"); + &sub ("esp",$esp_off+4); # alloca + &and ("esp",-16); # align stack + + &sub ($out,$inp); + &shl ($len,4); + &lea ($len,&DWP(-16*6,$inp,$len)); # end of input - 16*6 + &mov (&DWP($out_off,"esp"),$out); + &mov (&DWP($end_off,"esp"),$len); + &mov (&DWP($esp_off,"esp"),$rounds); + + &mov ($rounds,&DWP(240,$key)); + + &test ($block,1); + &jnz (&label("odd")); + + &bsf ($i3,$block); + &add ($block,1); + &shl ($i3,4); + &movdqu ($inout5,&QWP(0,$l_,$i3)); + &mov ($i3,$key); # put aside key + + &movdqu ($inout0,&QWP(16*0,$inp)); # load input + &lea ($inp,&DWP(16,$inp)); + + &pxor ($inout5,$rndkey0); # ^ last offset_i + &pxor ($inout0,$inout5); # ^ offset_i + + &movdqa ($inout4,$rndkey1); + if ($inline) + { &aesni_inline_generate1("dec"); } + else + { &call ("_aesni_decrypt1"); } + + &xorps ($inout0,$inout5); # ^ offset_i + &movaps ($rndkey1,$inout4); # pass the checksum + &movdqa ($rndkey0,$inout5); # pass last offset_i + &xorps ($rndkey1,$inout0); # checksum + &movups (&QWP(-16,$out,$inp),$inout0); # store output + + &mov ($rounds,&DWP(240,$i3)); + &mov ($key,$i3); # restore key + &mov ($len,&DWP($end_off,"esp")); + +&set_label("odd"); + &shl ($rounds,4); + &mov ($out,16); + &sub ($out,$rounds); # twisted rounds + &mov (&DWP($key_off,"esp"),$key); + &lea ($key,&DWP(32,$key,$rounds)); # end of key schedule + &mov (&DWP($rounds_off,"esp"),$out); + + &cmp ($inp,$len); + &ja (&label("short")); + &jmp (&label("grandloop")); + +&set_label("grandloop",32); + &lea ($i1,&DWP(1,$block)); + &lea ($i3,&DWP(3,$block)); + &lea ($i5,&DWP(5,$block)); + &add ($block,6); + &bsf ($i1,$i1); + &bsf ($i3,$i3); + &bsf ($i5,$i5); + &shl ($i1,4); + &shl ($i3,4); + &shl ($i5,4); + &movdqu ($inout0,&QWP(0,$l_)); + &movdqu ($inout1,&QWP(0,$l_,$i1)); + &mov ($rounds,&DWP($rounds_off,"esp")); + &movdqa ($inout2,$inout0); + &movdqu ($inout3,&QWP(0,$l_,$i3)); + &movdqa ($inout4,$inout0); + &movdqu ($inout5,&QWP(0,$l_,$i5)); + + &pxor ($inout0,$rndkey0); # ^ last offset_i + &pxor ($inout1,$inout0); + &movdqa (&QWP(16*0,"esp"),$inout0); + &pxor ($inout2,$inout1); + &movdqa (&QWP(16*1,"esp"),$inout1); + &pxor ($inout3,$inout2); + &movdqa (&QWP(16*2,"esp"),$inout2); + &pxor ($inout4,$inout3); + &movdqa (&QWP(16*3,"esp"),$inout3); + &pxor ($inout5,$inout4); + &movdqa (&QWP(16*4,"esp"),$inout4); + &movdqa (&QWP(16*5,"esp"),$inout5); + + &$movekey ($rndkey0,&QWP(-48,$key,$rounds)); + &movdqu ($inout0,&QWP(16*0,$inp)); # load input + &movdqu ($inout1,&QWP(16*1,$inp)); + &movdqu ($inout2,&QWP(16*2,$inp)); + &movdqu ($inout3,&QWP(16*3,$inp)); + &movdqu ($inout4,&QWP(16*4,$inp)); + &movdqu ($inout5,&QWP(16*5,$inp)); + &lea ($inp,&DWP(16*6,$inp)); + + &movdqa (&QWP($checksum,"esp"),$rndkey1); + &pxor ($inout0,$rndkey0); # ^ roundkey[0] + &pxor ($inout1,$rndkey0); + &pxor ($inout2,$rndkey0); + &pxor ($inout3,$rndkey0); + &pxor ($inout4,$rndkey0); + &pxor ($inout5,$rndkey0); + + &$movekey ($rndkey1,&QWP(-32,$key,$rounds)); + &pxor ($inout0,&QWP(16*0,"esp")); # ^ offset_i + &pxor ($inout1,&QWP(16*1,"esp")); + &pxor ($inout2,&QWP(16*2,"esp")); + &pxor ($inout3,&QWP(16*3,"esp")); + &pxor ($inout4,&QWP(16*4,"esp")); + &pxor ($inout5,&QWP(16*5,"esp")); + + &$movekey ($rndkey0,&QWP(-16,$key,$rounds)); + &aesdec ($inout0,$rndkey1); + &aesdec ($inout1,$rndkey1); + &aesdec ($inout2,$rndkey1); + &aesdec ($inout3,$rndkey1); + &aesdec ($inout4,$rndkey1); + &aesdec ($inout5,$rndkey1); + + &mov ($out,&DWP($out_off,"esp")); + &mov ($len,&DWP($end_off,"esp")); + &call ("_aesni_decrypt6_enter"); + + &movdqa ($rndkey0,&QWP(16*5,"esp")); # pass last offset_i + &pxor ($inout0,&QWP(16*0,"esp")); # ^ offset_i + &movdqa ($rndkey1,&QWP($checksum,"esp")); + &pxor ($inout1,&QWP(16*1,"esp")); + &pxor ($inout2,&QWP(16*2,"esp")); + &pxor ($inout3,&QWP(16*3,"esp")); + &pxor ($inout4,&QWP(16*4,"esp")); + &pxor ($inout5,$rndkey0); + + &pxor ($rndkey1,$inout0); # checksum + &movdqu (&QWP(-16*6,$out,$inp),$inout0);# store output + &pxor ($rndkey1,$inout1); + &movdqu (&QWP(-16*5,$out,$inp),$inout1); + &pxor ($rndkey1,$inout2); + &movdqu (&QWP(-16*4,$out,$inp),$inout2); + &pxor ($rndkey1,$inout3); + &movdqu (&QWP(-16*3,$out,$inp),$inout3); + &pxor ($rndkey1,$inout4); + &movdqu (&QWP(-16*2,$out,$inp),$inout4); + &pxor ($rndkey1,$inout5); + &movdqu (&QWP(-16*1,$out,$inp),$inout5); + &cmp ($inp,$len); # done yet? + &jb (&label("grandloop")); + +&set_label("short"); + &add ($len,16*6); + &sub ($len,$inp); + &jz (&label("done")); + + &cmp ($len,16*2); + &jb (&label("one")); + &je (&label("two")); + + &cmp ($len,16*4); + &jb (&label("three")); + &je (&label("four")); + + &lea ($i1,&DWP(1,$block)); + &lea ($i3,&DWP(3,$block)); + &bsf ($i1,$i1); + &bsf ($i3,$i3); + &shl ($i1,4); + &shl ($i3,4); + &movdqu ($inout0,&QWP(0,$l_)); + &movdqu ($inout1,&QWP(0,$l_,$i1)); + &mov ($rounds,&DWP($rounds_off,"esp")); + &movdqa ($inout2,$inout0); + &movdqu ($inout3,&QWP(0,$l_,$i3)); + &movdqa ($inout4,$inout0); + + &pxor ($inout0,$rndkey0); # ^ last offset_i + &pxor ($inout1,$inout0); + &movdqa (&QWP(16*0,"esp"),$inout0); + &pxor ($inout2,$inout1); + &movdqa (&QWP(16*1,"esp"),$inout1); + &pxor ($inout3,$inout2); + &movdqa (&QWP(16*2,"esp"),$inout2); + &pxor ($inout4,$inout3); + &movdqa (&QWP(16*3,"esp"),$inout3); + &pxor ($inout5,$inout4); + &movdqa (&QWP(16*4,"esp"),$inout4); + + &$movekey ($rndkey0,&QWP(-48,$key,$rounds)); + &movdqu ($inout0,&QWP(16*0,$inp)); # load input + &movdqu ($inout1,&QWP(16*1,$inp)); + &movdqu ($inout2,&QWP(16*2,$inp)); + &movdqu ($inout3,&QWP(16*3,$inp)); + &movdqu ($inout4,&QWP(16*4,$inp)); + &pxor ($inout5,$inout5); + + &movdqa (&QWP($checksum,"esp"),$rndkey1); + &pxor ($inout0,$rndkey0); # ^ roundkey[0] + &pxor ($inout1,$rndkey0); + &pxor ($inout2,$rndkey0); + &pxor ($inout3,$rndkey0); + &pxor ($inout4,$rndkey0); + + &$movekey ($rndkey1,&QWP(-32,$key,$rounds)); + &pxor ($inout0,&QWP(16*0,"esp")); # ^ offset_i + &pxor ($inout1,&QWP(16*1,"esp")); + &pxor ($inout2,&QWP(16*2,"esp")); + &pxor ($inout3,&QWP(16*3,"esp")); + &pxor ($inout4,&QWP(16*4,"esp")); + + &$movekey ($rndkey0,&QWP(-16,$key,$rounds)); + &aesdec ($inout0,$rndkey1); + &aesdec ($inout1,$rndkey1); + &aesdec ($inout2,$rndkey1); + &aesdec ($inout3,$rndkey1); + &aesdec ($inout4,$rndkey1); + &aesdec ($inout5,$rndkey1); + + &mov ($out,&DWP($out_off,"esp")); + &call ("_aesni_decrypt6_enter"); + + &movdqa ($rndkey0,&QWP(16*4,"esp")); # pass last offset_i + &pxor ($inout0,&QWP(16*0,"esp")); # ^ offset_i + &movdqa ($rndkey1,&QWP($checksum,"esp")); + &pxor ($inout1,&QWP(16*1,"esp")); + &pxor ($inout2,&QWP(16*2,"esp")); + &pxor ($inout3,&QWP(16*3,"esp")); + &pxor ($inout4,$rndkey0); + + &pxor ($rndkey1,$inout0); # checksum + &movdqu (&QWP(16*0,$out,$inp),$inout0); # store output + &pxor ($rndkey1,$inout1); + &movdqu (&QWP(16*1,$out,$inp),$inout1); + &pxor ($rndkey1,$inout2); + &movdqu (&QWP(16*2,$out,$inp),$inout2); + &pxor ($rndkey1,$inout3); + &movdqu (&QWP(16*3,$out,$inp),$inout3); + &pxor ($rndkey1,$inout4); + &movdqu (&QWP(16*4,$out,$inp),$inout4); + + &jmp (&label("done")); + +&set_label("one",16); + &movdqu ($inout5,&QWP(0,$l_)); + &mov ($key,&DWP($key_off,"esp")); # restore key + + &movdqu ($inout0,&QWP(16*0,$inp)); # load input + &mov ($rounds,&DWP(240,$key)); + + &pxor ($inout5,$rndkey0); # ^ last offset_i + &pxor ($inout0,$inout5); # ^ offset_i + + &movdqa ($inout4,$rndkey1); + &mov ($out,&DWP($out_off,"esp")); + if ($inline) + { &aesni_inline_generate1("dec"); } + else + { &call ("_aesni_decrypt1"); } + + &xorps ($inout0,$inout5); # ^ offset_i + &movaps ($rndkey1,$inout4); # pass the checksum + &movdqa ($rndkey0,$inout5); # pass last offset_i + &xorps ($rndkey1,$inout0); # checksum + &movups (&QWP(0,$out,$inp),$inout0); + + &jmp (&label("done")); + +&set_label("two",16); + &lea ($i1,&DWP(1,$block)); + &mov ($key,&DWP($key_off,"esp")); # restore key + &bsf ($i1,$i1); + &shl ($i1,4); + &movdqu ($inout4,&QWP(0,$l_)); + &movdqu ($inout5,&QWP(0,$l_,$i1)); + + &movdqu ($inout0,&QWP(16*0,$inp)); # load input + &movdqu ($inout1,&QWP(16*1,$inp)); + &mov ($rounds,&DWP(240,$key)); + + &movdqa ($inout3,$rndkey1); + &pxor ($inout4,$rndkey0); # ^ last offset_i + &pxor ($inout5,$inout4); + + &pxor ($inout0,$inout4); # ^ offset_i + &pxor ($inout1,$inout5); + + &mov ($out,&DWP($out_off,"esp")); + &call ("_aesni_decrypt2"); + + &xorps ($inout0,$inout4); # ^ offset_i + &xorps ($inout1,$inout5); + &movdqa ($rndkey0,$inout5); # pass last offset_i + &xorps ($inout3,$inout0); # checksum + &movups (&QWP(16*0,$out,$inp),$inout0); # store output + &xorps ($inout3,$inout1); + &movups (&QWP(16*1,$out,$inp),$inout1); + &movaps ($rndkey1,$inout3); # pass the checksum + + &jmp (&label("done")); + +&set_label("three",16); + &lea ($i1,&DWP(1,$block)); + &mov ($key,&DWP($key_off,"esp")); # restore key + &bsf ($i1,$i1); + &shl ($i1,4); + &movdqu ($inout3,&QWP(0,$l_)); + &movdqu ($inout4,&QWP(0,$l_,$i1)); + &movdqa ($inout5,$inout3); + + &movdqu ($inout0,&QWP(16*0,$inp)); # load input + &movdqu ($inout1,&QWP(16*1,$inp)); + &movdqu ($inout2,&QWP(16*2,$inp)); + &mov ($rounds,&DWP(240,$key)); + + &movdqa (&QWP($checksum,"esp"),$rndkey1); + &pxor ($inout3,$rndkey0); # ^ last offset_i + &pxor ($inout4,$inout3); + &pxor ($inout5,$inout4); + + &pxor ($inout0,$inout3); # ^ offset_i + &pxor ($inout1,$inout4); + &pxor ($inout2,$inout5); + + &mov ($out,&DWP($out_off,"esp")); + &call ("_aesni_decrypt3"); + + &movdqa ($rndkey1,&QWP($checksum,"esp"));# pass the checksum + &xorps ($inout0,$inout3); # ^ offset_i + &xorps ($inout1,$inout4); + &xorps ($inout2,$inout5); + &movups (&QWP(16*0,$out,$inp),$inout0); # store output + &pxor ($rndkey1,$inout0); # checksum + &movdqa ($rndkey0,$inout5); # pass last offset_i + &movups (&QWP(16*1,$out,$inp),$inout1); + &pxor ($rndkey1,$inout1); + &movups (&QWP(16*2,$out,$inp),$inout2); + &pxor ($rndkey1,$inout2); + + &jmp (&label("done")); + +&set_label("four",16); + &lea ($i1,&DWP(1,$block)); + &lea ($i3,&DWP(3,$block)); + &bsf ($i1,$i1); + &bsf ($i3,$i3); + &mov ($key,&DWP($key_off,"esp")); # restore key + &shl ($i1,4); + &shl ($i3,4); + &movdqu ($inout2,&QWP(0,$l_)); + &movdqu ($inout3,&QWP(0,$l_,$i1)); + &movdqa ($inout4,$inout2); + &movdqu ($inout5,&QWP(0,$l_,$i3)); + + &pxor ($inout2,$rndkey0); # ^ last offset_i + &movdqu ($inout0,&QWP(16*0,$inp)); # load input + &pxor ($inout3,$inout2); + &movdqu ($inout1,&QWP(16*1,$inp)); + &pxor ($inout4,$inout3); + &movdqa (&QWP(16*0,"esp"),$inout2); + &pxor ($inout5,$inout4); + &movdqa (&QWP(16*1,"esp"),$inout3); + &movdqu ($inout2,&QWP(16*2,$inp)); + &movdqu ($inout3,&QWP(16*3,$inp)); + &mov ($rounds,&DWP(240,$key)); + + &movdqa (&QWP($checksum,"esp"),$rndkey1); + &pxor ($inout0,&QWP(16*0,"esp")); # ^ offset_i + &pxor ($inout1,&QWP(16*1,"esp")); + &pxor ($inout2,$inout4); + &pxor ($inout3,$inout5); + + &mov ($out,&DWP($out_off,"esp")); + &call ("_aesni_decrypt4"); + + &movdqa ($rndkey1,&QWP($checksum,"esp"));# pass the checksum + &xorps ($inout0,&QWP(16*0,"esp")); # ^ offset_i + &xorps ($inout1,&QWP(16*1,"esp")); + &xorps ($inout2,$inout4); + &movups (&QWP(16*0,$out,$inp),$inout0); # store output + &pxor ($rndkey1,$inout0); # checksum + &xorps ($inout3,$inout5); + &movups (&QWP(16*1,$out,$inp),$inout1); + &pxor ($rndkey1,$inout1); + &movdqa ($rndkey0,$inout5); # pass last offset_i + &movups (&QWP(16*2,$out,$inp),$inout2); + &pxor ($rndkey1,$inout2); + &movups (&QWP(16*3,$out,$inp),$inout3); + &pxor ($rndkey1,$inout3); + +&set_label("done"); + &mov ($key,&DWP($esp_off,"esp")); + &pxor ($inout0,$inout0); # clear register bank + &pxor ($inout1,$inout1); + &movdqa (&QWP(16*0,"esp"),$inout0); # clear stack + &pxor ($inout2,$inout2); + &movdqa (&QWP(16*1,"esp"),$inout0); + &pxor ($inout3,$inout3); + &movdqa (&QWP(16*2,"esp"),$inout0); + &pxor ($inout4,$inout4); + &movdqa (&QWP(16*3,"esp"),$inout0); + &pxor ($inout5,$inout5); + &movdqa (&QWP(16*4,"esp"),$inout0); + &movdqa (&QWP(16*5,"esp"),$inout0); + &movdqa (&QWP(16*6,"esp"),$inout0); + + &lea ("esp",&DWP(0,$key)); + &mov ($rounds,&wparam(5)); # &offset_i + &mov ($rounds_,&wparam(7)); # &checksum + &movdqu (&QWP(0,$rounds),$rndkey0); + &pxor ($rndkey0,$rndkey0); + &movdqu (&QWP(0,$rounds_),$rndkey1); + &pxor ($rndkey1,$rndkey1); +&function_end("aesni_ocb_decrypt"); +} +} + +###################################################################### +# void $PREFIX_cbc_encrypt (const void *inp, void *out, +# size_t length, const AES_KEY *key, +# unsigned char *ivp,const int enc); +&function_begin("${PREFIX}_cbc_encrypt"); + &mov ($inp,&wparam(0)); + &mov ($rounds_,"esp"); + &mov ($out,&wparam(1)); + &sub ($rounds_,24); + &mov ($len,&wparam(2)); + &and ($rounds_,-16); + &mov ($key,&wparam(3)); + &mov ($key_,&wparam(4)); + &test ($len,$len); + &jz (&label("cbc_abort")); + + &cmp (&wparam(5),0); + &xchg ($rounds_,"esp"); # alloca + &movups ($ivec,&QWP(0,$key_)); # load IV + &mov ($rounds,&DWP(240,$key)); + &mov ($key_,$key); # backup $key + &mov (&DWP(16,"esp"),$rounds_); # save original %esp + &mov ($rounds_,$rounds); # backup $rounds + &je (&label("cbc_decrypt")); + + &movaps ($inout0,$ivec); + &cmp ($len,16); + &jb (&label("cbc_enc_tail")); + &sub ($len,16); + &jmp (&label("cbc_enc_loop")); + +&set_label("cbc_enc_loop",16); + &movups ($ivec,&QWP(0,$inp)); # input actually + &lea ($inp,&DWP(16,$inp)); + if ($inline) + { &aesni_inline_generate1("enc",$inout0,$ivec); } + else + { &xorps($inout0,$ivec); &call("_aesni_encrypt1"); } + &mov ($rounds,$rounds_); # restore $rounds + &mov ($key,$key_); # restore $key + &movups (&QWP(0,$out),$inout0); # store output + &lea ($out,&DWP(16,$out)); + &sub ($len,16); + &jnc (&label("cbc_enc_loop")); + &add ($len,16); + &jnz (&label("cbc_enc_tail")); + &movaps ($ivec,$inout0); + &pxor ($inout0,$inout0); + &jmp (&label("cbc_ret")); + +&set_label("cbc_enc_tail"); + &mov ("ecx",$len); # zaps $rounds + &data_word(0xA4F3F689); # rep movsb + &mov ("ecx",16); # zero tail + &sub ("ecx",$len); + &xor ("eax","eax"); # zaps $len + &data_word(0xAAF3F689); # rep stosb + &lea ($out,&DWP(-16,$out)); # rewind $out by 1 block + &mov ($rounds,$rounds_); # restore $rounds + &mov ($inp,$out); # $inp and $out are the same + &mov ($key,$key_); # restore $key + &jmp (&label("cbc_enc_loop")); +###################################################################### +&set_label("cbc_decrypt",16); + &cmp ($len,0x50); + &jbe (&label("cbc_dec_tail")); + &movaps (&QWP(0,"esp"),$ivec); # save IV + &sub ($len,0x50); + &jmp (&label("cbc_dec_loop6_enter")); + +&set_label("cbc_dec_loop6",16); + &movaps (&QWP(0,"esp"),$rndkey0); # save IV + &movups (&QWP(0,$out),$inout5); + &lea ($out,&DWP(0x10,$out)); +&set_label("cbc_dec_loop6_enter"); + &movdqu ($inout0,&QWP(0,$inp)); + &movdqu ($inout1,&QWP(0x10,$inp)); + &movdqu ($inout2,&QWP(0x20,$inp)); + &movdqu ($inout3,&QWP(0x30,$inp)); + &movdqu ($inout4,&QWP(0x40,$inp)); + &movdqu ($inout5,&QWP(0x50,$inp)); + + &call ("_aesni_decrypt6"); + + &movups ($rndkey1,&QWP(0,$inp)); + &movups ($rndkey0,&QWP(0x10,$inp)); + &xorps ($inout0,&QWP(0,"esp")); # ^=IV + &xorps ($inout1,$rndkey1); + &movups ($rndkey1,&QWP(0x20,$inp)); + &xorps ($inout2,$rndkey0); + &movups ($rndkey0,&QWP(0x30,$inp)); + &xorps ($inout3,$rndkey1); + &movups ($rndkey1,&QWP(0x40,$inp)); + &xorps ($inout4,$rndkey0); + &movups ($rndkey0,&QWP(0x50,$inp)); # IV + &xorps ($inout5,$rndkey1); + &movups (&QWP(0,$out),$inout0); + &movups (&QWP(0x10,$out),$inout1); + &lea ($inp,&DWP(0x60,$inp)); + &movups (&QWP(0x20,$out),$inout2); + &mov ($rounds,$rounds_); # restore $rounds + &movups (&QWP(0x30,$out),$inout3); + &mov ($key,$key_); # restore $key + &movups (&QWP(0x40,$out),$inout4); + &lea ($out,&DWP(0x50,$out)); + &sub ($len,0x60); + &ja (&label("cbc_dec_loop6")); + + &movaps ($inout0,$inout5); + &movaps ($ivec,$rndkey0); + &add ($len,0x50); + &jle (&label("cbc_dec_clear_tail_collected")); + &movups (&QWP(0,$out),$inout0); + &lea ($out,&DWP(0x10,$out)); +&set_label("cbc_dec_tail"); + &movups ($inout0,&QWP(0,$inp)); + &movaps ($in0,$inout0); + &cmp ($len,0x10); + &jbe (&label("cbc_dec_one")); + + &movups ($inout1,&QWP(0x10,$inp)); + &movaps ($in1,$inout1); + &cmp ($len,0x20); + &jbe (&label("cbc_dec_two")); + + &movups ($inout2,&QWP(0x20,$inp)); + &cmp ($len,0x30); + &jbe (&label("cbc_dec_three")); + + &movups ($inout3,&QWP(0x30,$inp)); + &cmp ($len,0x40); + &jbe (&label("cbc_dec_four")); + + &movups ($inout4,&QWP(0x40,$inp)); + &movaps (&QWP(0,"esp"),$ivec); # save IV + &movups ($inout0,&QWP(0,$inp)); + &xorps ($inout5,$inout5); + &call ("_aesni_decrypt6"); + &movups ($rndkey1,&QWP(0,$inp)); + &movups ($rndkey0,&QWP(0x10,$inp)); + &xorps ($inout0,&QWP(0,"esp")); # ^= IV + &xorps ($inout1,$rndkey1); + &movups ($rndkey1,&QWP(0x20,$inp)); + &xorps ($inout2,$rndkey0); + &movups ($rndkey0,&QWP(0x30,$inp)); + &xorps ($inout3,$rndkey1); + &movups ($ivec,&QWP(0x40,$inp)); # IV + &xorps ($inout4,$rndkey0); + &movups (&QWP(0,$out),$inout0); + &movups (&QWP(0x10,$out),$inout1); + &pxor ($inout1,$inout1); + &movups (&QWP(0x20,$out),$inout2); + &pxor ($inout2,$inout2); + &movups (&QWP(0x30,$out),$inout3); + &pxor ($inout3,$inout3); + &lea ($out,&DWP(0x40,$out)); + &movaps ($inout0,$inout4); + &pxor ($inout4,$inout4); + &sub ($len,0x50); + &jmp (&label("cbc_dec_tail_collected")); + +&set_label("cbc_dec_one",16); + if ($inline) + { &aesni_inline_generate1("dec"); } + else + { &call ("_aesni_decrypt1"); } + &xorps ($inout0,$ivec); + &movaps ($ivec,$in0); + &sub ($len,0x10); + &jmp (&label("cbc_dec_tail_collected")); + +&set_label("cbc_dec_two",16); + &call ("_aesni_decrypt2"); + &xorps ($inout0,$ivec); + &xorps ($inout1,$in0); + &movups (&QWP(0,$out),$inout0); + &movaps ($inout0,$inout1); + &pxor ($inout1,$inout1); + &lea ($out,&DWP(0x10,$out)); + &movaps ($ivec,$in1); + &sub ($len,0x20); + &jmp (&label("cbc_dec_tail_collected")); + +&set_label("cbc_dec_three",16); + &call ("_aesni_decrypt3"); + &xorps ($inout0,$ivec); + &xorps ($inout1,$in0); + &xorps ($inout2,$in1); + &movups (&QWP(0,$out),$inout0); + &movaps ($inout0,$inout2); + &pxor ($inout2,$inout2); + &movups (&QWP(0x10,$out),$inout1); + &pxor ($inout1,$inout1); + &lea ($out,&DWP(0x20,$out)); + &movups ($ivec,&QWP(0x20,$inp)); + &sub ($len,0x30); + &jmp (&label("cbc_dec_tail_collected")); + +&set_label("cbc_dec_four",16); + &call ("_aesni_decrypt4"); + &movups ($rndkey1,&QWP(0x10,$inp)); + &movups ($rndkey0,&QWP(0x20,$inp)); + &xorps ($inout0,$ivec); + &movups ($ivec,&QWP(0x30,$inp)); + &xorps ($inout1,$in0); + &movups (&QWP(0,$out),$inout0); + &xorps ($inout2,$rndkey1); + &movups (&QWP(0x10,$out),$inout1); + &pxor ($inout1,$inout1); + &xorps ($inout3,$rndkey0); + &movups (&QWP(0x20,$out),$inout2); + &pxor ($inout2,$inout2); + &lea ($out,&DWP(0x30,$out)); + &movaps ($inout0,$inout3); + &pxor ($inout3,$inout3); + &sub ($len,0x40); + &jmp (&label("cbc_dec_tail_collected")); + +&set_label("cbc_dec_clear_tail_collected",16); + &pxor ($inout1,$inout1); + &pxor ($inout2,$inout2); + &pxor ($inout3,$inout3); + &pxor ($inout4,$inout4); +&set_label("cbc_dec_tail_collected"); + &and ($len,15); + &jnz (&label("cbc_dec_tail_partial")); + &movups (&QWP(0,$out),$inout0); + &pxor ($rndkey0,$rndkey0); + &jmp (&label("cbc_ret")); + +&set_label("cbc_dec_tail_partial",16); + &movaps (&QWP(0,"esp"),$inout0); + &pxor ($rndkey0,$rndkey0); + &mov ("ecx",16); + &mov ($inp,"esp"); + &sub ("ecx",$len); + &data_word(0xA4F3F689); # rep movsb + &movdqa (&QWP(0,"esp"),$inout0); + +&set_label("cbc_ret"); + &mov ("esp",&DWP(16,"esp")); # pull original %esp + &mov ($key_,&wparam(4)); + &pxor ($inout0,$inout0); + &pxor ($rndkey1,$rndkey1); + &movups (&QWP(0,$key_),$ivec); # output IV + &pxor ($ivec,$ivec); +&set_label("cbc_abort"); +&function_end("${PREFIX}_cbc_encrypt"); + +###################################################################### +# Mechanical port from aesni-x86_64.pl. +# +# _aesni_set_encrypt_key is private interface, +# input: +# "eax" const unsigned char *userKey +# $rounds int bits +# $key AES_KEY *key +# output: +# "eax" return code +# $round rounds + +&function_begin_B("_aesni_set_encrypt_key"); + &push ("ebp"); + &push ("ebx"); + &test ("eax","eax"); + &jz (&label("bad_pointer")); + &test ($key,$key); + &jz (&label("bad_pointer")); + + &call (&label("pic")); +&set_label("pic"); + &blindpop("ebx"); + &lea ("ebx",&DWP(&label("key_const")."-".&label("pic"),"ebx")); + + &picmeup("ebp","OPENSSL_ia32cap_P","ebx",&label("key_const")); + &movups ("xmm0",&QWP(0,"eax")); # pull first 128 bits of *userKey + &xorps ("xmm4","xmm4"); # low dword of xmm4 is assumed 0 + &mov ("ebp",&DWP(4,"ebp")); + &lea ($key,&DWP(16,$key)); + &and ("ebp",1<<28|1<<11); # AVX and XOP bits + &cmp ($rounds,256); + &je (&label("14rounds")); + &cmp ($rounds,192); + &je (&label("12rounds")); + &cmp ($rounds,128); + &jne (&label("bad_keybits")); + +&set_label("10rounds",16); + &cmp ("ebp",1<<28); + &je (&label("10rounds_alt")); + + &mov ($rounds,9); + &$movekey (&QWP(-16,$key),"xmm0"); # round 0 + &aeskeygenassist("xmm1","xmm0",0x01); # round 1 + &call (&label("key_128_cold")); + &aeskeygenassist("xmm1","xmm0",0x2); # round 2 + &call (&label("key_128")); + &aeskeygenassist("xmm1","xmm0",0x04); # round 3 + &call (&label("key_128")); + &aeskeygenassist("xmm1","xmm0",0x08); # round 4 + &call (&label("key_128")); + &aeskeygenassist("xmm1","xmm0",0x10); # round 5 + &call (&label("key_128")); + &aeskeygenassist("xmm1","xmm0",0x20); # round 6 + &call (&label("key_128")); + &aeskeygenassist("xmm1","xmm0",0x40); # round 7 + &call (&label("key_128")); + &aeskeygenassist("xmm1","xmm0",0x80); # round 8 + &call (&label("key_128")); + &aeskeygenassist("xmm1","xmm0",0x1b); # round 9 + &call (&label("key_128")); + &aeskeygenassist("xmm1","xmm0",0x36); # round 10 + &call (&label("key_128")); + &$movekey (&QWP(0,$key),"xmm0"); + &mov (&DWP(80,$key),$rounds); + + &jmp (&label("good_key")); + +&set_label("key_128",16); + &$movekey (&QWP(0,$key),"xmm0"); + &lea ($key,&DWP(16,$key)); +&set_label("key_128_cold"); + &shufps ("xmm4","xmm0",0b00010000); + &xorps ("xmm0","xmm4"); + &shufps ("xmm4","xmm0",0b10001100); + &xorps ("xmm0","xmm4"); + &shufps ("xmm1","xmm1",0b11111111); # critical path + &xorps ("xmm0","xmm1"); + &ret(); + +&set_label("10rounds_alt",16); + &movdqa ("xmm5",&QWP(0x00,"ebx")); + &mov ($rounds,8); + &movdqa ("xmm4",&QWP(0x20,"ebx")); + &movdqa ("xmm2","xmm0"); + &movdqu (&QWP(-16,$key),"xmm0"); + +&set_label("loop_key128"); + &pshufb ("xmm0","xmm5"); + &aesenclast ("xmm0","xmm4"); + &pslld ("xmm4",1); + &lea ($key,&DWP(16,$key)); + + &movdqa ("xmm3","xmm2"); + &pslldq ("xmm2",4); + &pxor ("xmm3","xmm2"); + &pslldq ("xmm2",4); + &pxor ("xmm3","xmm2"); + &pslldq ("xmm2",4); + &pxor ("xmm2","xmm3"); + + &pxor ("xmm0","xmm2"); + &movdqu (&QWP(-16,$key),"xmm0"); + &movdqa ("xmm2","xmm0"); + + &dec ($rounds); + &jnz (&label("loop_key128")); + + &movdqa ("xmm4",&QWP(0x30,"ebx")); + + &pshufb ("xmm0","xmm5"); + &aesenclast ("xmm0","xmm4"); + &pslld ("xmm4",1); + + &movdqa ("xmm3","xmm2"); + &pslldq ("xmm2",4); + &pxor ("xmm3","xmm2"); + &pslldq ("xmm2",4); + &pxor ("xmm3","xmm2"); + &pslldq ("xmm2",4); + &pxor ("xmm2","xmm3"); + + &pxor ("xmm0","xmm2"); + &movdqu (&QWP(0,$key),"xmm0"); + + &movdqa ("xmm2","xmm0"); + &pshufb ("xmm0","xmm5"); + &aesenclast ("xmm0","xmm4"); + + &movdqa ("xmm3","xmm2"); + &pslldq ("xmm2",4); + &pxor ("xmm3","xmm2"); + &pslldq ("xmm2",4); + &pxor ("xmm3","xmm2"); + &pslldq ("xmm2",4); + &pxor ("xmm2","xmm3"); + + &pxor ("xmm0","xmm2"); + &movdqu (&QWP(16,$key),"xmm0"); + + &mov ($rounds,9); + &mov (&DWP(96,$key),$rounds); + + &jmp (&label("good_key")); + +&set_label("12rounds",16); + &movq ("xmm2",&QWP(16,"eax")); # remaining 1/3 of *userKey + &cmp ("ebp",1<<28); + &je (&label("12rounds_alt")); + + &mov ($rounds,11); + &$movekey (&QWP(-16,$key),"xmm0"); # round 0 + &aeskeygenassist("xmm1","xmm2",0x01); # round 1,2 + &call (&label("key_192a_cold")); + &aeskeygenassist("xmm1","xmm2",0x02); # round 2,3 + &call (&label("key_192b")); + &aeskeygenassist("xmm1","xmm2",0x04); # round 4,5 + &call (&label("key_192a")); + &aeskeygenassist("xmm1","xmm2",0x08); # round 5,6 + &call (&label("key_192b")); + &aeskeygenassist("xmm1","xmm2",0x10); # round 7,8 + &call (&label("key_192a")); + &aeskeygenassist("xmm1","xmm2",0x20); # round 8,9 + &call (&label("key_192b")); + &aeskeygenassist("xmm1","xmm2",0x40); # round 10,11 + &call (&label("key_192a")); + &aeskeygenassist("xmm1","xmm2",0x80); # round 11,12 + &call (&label("key_192b")); + &$movekey (&QWP(0,$key),"xmm0"); + &mov (&DWP(48,$key),$rounds); + + &jmp (&label("good_key")); + +&set_label("key_192a",16); + &$movekey (&QWP(0,$key),"xmm0"); + &lea ($key,&DWP(16,$key)); +&set_label("key_192a_cold",16); + &movaps ("xmm5","xmm2"); +&set_label("key_192b_warm"); + &shufps ("xmm4","xmm0",0b00010000); + &movdqa ("xmm3","xmm2"); + &xorps ("xmm0","xmm4"); + &shufps ("xmm4","xmm0",0b10001100); + &pslldq ("xmm3",4); + &xorps ("xmm0","xmm4"); + &pshufd ("xmm1","xmm1",0b01010101); # critical path + &pxor ("xmm2","xmm3"); + &pxor ("xmm0","xmm1"); + &pshufd ("xmm3","xmm0",0b11111111); + &pxor ("xmm2","xmm3"); + &ret(); + +&set_label("key_192b",16); + &movaps ("xmm3","xmm0"); + &shufps ("xmm5","xmm0",0b01000100); + &$movekey (&QWP(0,$key),"xmm5"); + &shufps ("xmm3","xmm2",0b01001110); + &$movekey (&QWP(16,$key),"xmm3"); + &lea ($key,&DWP(32,$key)); + &jmp (&label("key_192b_warm")); + +&set_label("12rounds_alt",16); + &movdqa ("xmm5",&QWP(0x10,"ebx")); + &movdqa ("xmm4",&QWP(0x20,"ebx")); + &mov ($rounds,8); + &movdqu (&QWP(-16,$key),"xmm0"); + +&set_label("loop_key192"); + &movq (&QWP(0,$key),"xmm2"); + &movdqa ("xmm1","xmm2"); + &pshufb ("xmm2","xmm5"); + &aesenclast ("xmm2","xmm4"); + &pslld ("xmm4",1); + &lea ($key,&DWP(24,$key)); + + &movdqa ("xmm3","xmm0"); + &pslldq ("xmm0",4); + &pxor ("xmm3","xmm0"); + &pslldq ("xmm0",4); + &pxor ("xmm3","xmm0"); + &pslldq ("xmm0",4); + &pxor ("xmm0","xmm3"); + + &pshufd ("xmm3","xmm0",0xff); + &pxor ("xmm3","xmm1"); + &pslldq ("xmm1",4); + &pxor ("xmm3","xmm1"); + + &pxor ("xmm0","xmm2"); + &pxor ("xmm2","xmm3"); + &movdqu (&QWP(-16,$key),"xmm0"); + + &dec ($rounds); + &jnz (&label("loop_key192")); + + &mov ($rounds,11); + &mov (&DWP(32,$key),$rounds); + + &jmp (&label("good_key")); + +&set_label("14rounds",16); + &movups ("xmm2",&QWP(16,"eax")); # remaining half of *userKey + &lea ($key,&DWP(16,$key)); + &cmp ("ebp",1<<28); + &je (&label("14rounds_alt")); + + &mov ($rounds,13); + &$movekey (&QWP(-32,$key),"xmm0"); # round 0 + &$movekey (&QWP(-16,$key),"xmm2"); # round 1 + &aeskeygenassist("xmm1","xmm2",0x01); # round 2 + &call (&label("key_256a_cold")); + &aeskeygenassist("xmm1","xmm0",0x01); # round 3 + &call (&label("key_256b")); + &aeskeygenassist("xmm1","xmm2",0x02); # round 4 + &call (&label("key_256a")); + &aeskeygenassist("xmm1","xmm0",0x02); # round 5 + &call (&label("key_256b")); + &aeskeygenassist("xmm1","xmm2",0x04); # round 6 + &call (&label("key_256a")); + &aeskeygenassist("xmm1","xmm0",0x04); # round 7 + &call (&label("key_256b")); + &aeskeygenassist("xmm1","xmm2",0x08); # round 8 + &call (&label("key_256a")); + &aeskeygenassist("xmm1","xmm0",0x08); # round 9 + &call (&label("key_256b")); + &aeskeygenassist("xmm1","xmm2",0x10); # round 10 + &call (&label("key_256a")); + &aeskeygenassist("xmm1","xmm0",0x10); # round 11 + &call (&label("key_256b")); + &aeskeygenassist("xmm1","xmm2",0x20); # round 12 + &call (&label("key_256a")); + &aeskeygenassist("xmm1","xmm0",0x20); # round 13 + &call (&label("key_256b")); + &aeskeygenassist("xmm1","xmm2",0x40); # round 14 + &call (&label("key_256a")); + &$movekey (&QWP(0,$key),"xmm0"); + &mov (&DWP(16,$key),$rounds); + &xor ("eax","eax"); + + &jmp (&label("good_key")); + +&set_label("key_256a",16); + &$movekey (&QWP(0,$key),"xmm2"); + &lea ($key,&DWP(16,$key)); +&set_label("key_256a_cold"); + &shufps ("xmm4","xmm0",0b00010000); + &xorps ("xmm0","xmm4"); + &shufps ("xmm4","xmm0",0b10001100); + &xorps ("xmm0","xmm4"); + &shufps ("xmm1","xmm1",0b11111111); # critical path + &xorps ("xmm0","xmm1"); + &ret(); + +&set_label("key_256b",16); + &$movekey (&QWP(0,$key),"xmm0"); + &lea ($key,&DWP(16,$key)); + + &shufps ("xmm4","xmm2",0b00010000); + &xorps ("xmm2","xmm4"); + &shufps ("xmm4","xmm2",0b10001100); + &xorps ("xmm2","xmm4"); + &shufps ("xmm1","xmm1",0b10101010); # critical path + &xorps ("xmm2","xmm1"); + &ret(); + +&set_label("14rounds_alt",16); + &movdqa ("xmm5",&QWP(0x00,"ebx")); + &movdqa ("xmm4",&QWP(0x20,"ebx")); + &mov ($rounds,7); + &movdqu (&QWP(-32,$key),"xmm0"); + &movdqa ("xmm1","xmm2"); + &movdqu (&QWP(-16,$key),"xmm2"); + +&set_label("loop_key256"); + &pshufb ("xmm2","xmm5"); + &aesenclast ("xmm2","xmm4"); + + &movdqa ("xmm3","xmm0"); + &pslldq ("xmm0",4); + &pxor ("xmm3","xmm0"); + &pslldq ("xmm0",4); + &pxor ("xmm3","xmm0"); + &pslldq ("xmm0",4); + &pxor ("xmm0","xmm3"); + &pslld ("xmm4",1); + + &pxor ("xmm0","xmm2"); + &movdqu (&QWP(0,$key),"xmm0"); + + &dec ($rounds); + &jz (&label("done_key256")); + + &pshufd ("xmm2","xmm0",0xff); + &pxor ("xmm3","xmm3"); + &aesenclast ("xmm2","xmm3"); + + &movdqa ("xmm3","xmm1"); + &pslldq ("xmm1",4); + &pxor ("xmm3","xmm1"); + &pslldq ("xmm1",4); + &pxor ("xmm3","xmm1"); + &pslldq ("xmm1",4); + &pxor ("xmm1","xmm3"); + + &pxor ("xmm2","xmm1"); + &movdqu (&QWP(16,$key),"xmm2"); + &lea ($key,&DWP(32,$key)); + &movdqa ("xmm1","xmm2"); + &jmp (&label("loop_key256")); + +&set_label("done_key256"); + &mov ($rounds,13); + &mov (&DWP(16,$key),$rounds); + +&set_label("good_key"); + &pxor ("xmm0","xmm0"); + &pxor ("xmm1","xmm1"); + &pxor ("xmm2","xmm2"); + &pxor ("xmm3","xmm3"); + &pxor ("xmm4","xmm4"); + &pxor ("xmm5","xmm5"); + &xor ("eax","eax"); + &pop ("ebx"); + &pop ("ebp"); + &ret (); + +&set_label("bad_pointer",4); + &mov ("eax",-1); + &pop ("ebx"); + &pop ("ebp"); + &ret (); +&set_label("bad_keybits",4); + &pxor ("xmm0","xmm0"); + &mov ("eax",-2); + &pop ("ebx"); + &pop ("ebp"); + &ret (); +&function_end_B("_aesni_set_encrypt_key"); + +# int $PREFIX_set_encrypt_key (const unsigned char *userKey, int bits, +# AES_KEY *key) +&function_begin_B("${PREFIX}_set_encrypt_key"); + &mov ("eax",&wparam(0)); + &mov ($rounds,&wparam(1)); + &mov ($key,&wparam(2)); + &call ("_aesni_set_encrypt_key"); + &ret (); +&function_end_B("${PREFIX}_set_encrypt_key"); + +# int $PREFIX_set_decrypt_key (const unsigned char *userKey, int bits, +# AES_KEY *key) +&function_begin_B("${PREFIX}_set_decrypt_key"); + &mov ("eax",&wparam(0)); + &mov ($rounds,&wparam(1)); + &mov ($key,&wparam(2)); + &call ("_aesni_set_encrypt_key"); + &mov ($key,&wparam(2)); + &shl ($rounds,4); # rounds-1 after _aesni_set_encrypt_key + &test ("eax","eax"); + &jnz (&label("dec_key_ret")); + &lea ("eax",&DWP(16,$key,$rounds)); # end of key schedule + + &$movekey ("xmm0",&QWP(0,$key)); # just swap + &$movekey ("xmm1",&QWP(0,"eax")); + &$movekey (&QWP(0,"eax"),"xmm0"); + &$movekey (&QWP(0,$key),"xmm1"); + &lea ($key,&DWP(16,$key)); + &lea ("eax",&DWP(-16,"eax")); + +&set_label("dec_key_inverse"); + &$movekey ("xmm0",&QWP(0,$key)); # swap and inverse + &$movekey ("xmm1",&QWP(0,"eax")); + &aesimc ("xmm0","xmm0"); + &aesimc ("xmm1","xmm1"); + &lea ($key,&DWP(16,$key)); + &lea ("eax",&DWP(-16,"eax")); + &$movekey (&QWP(16,"eax"),"xmm0"); + &$movekey (&QWP(-16,$key),"xmm1"); + &cmp ("eax",$key); + &ja (&label("dec_key_inverse")); + + &$movekey ("xmm0",&QWP(0,$key)); # inverse middle + &aesimc ("xmm0","xmm0"); + &$movekey (&QWP(0,$key),"xmm0"); + + &pxor ("xmm0","xmm0"); + &pxor ("xmm1","xmm1"); + &xor ("eax","eax"); # return success +&set_label("dec_key_ret"); + &ret (); +&function_end_B("${PREFIX}_set_decrypt_key"); + +&set_label("key_const",64); +&data_word(0x0c0f0e0d,0x0c0f0e0d,0x0c0f0e0d,0x0c0f0e0d); +&data_word(0x04070605,0x04070605,0x04070605,0x04070605); +&data_word(1,1,1,1); +&data_word(0x1b,0x1b,0x1b,0x1b); +&asciz("AES for Intel AES-NI, CRYPTOGAMS by "); + +&asm_finish(); + +close STDOUT; diff --git a/openssl-1.1.0h/crypto/aes/asm/aesni-x86_64.pl b/openssl-1.1.0h/crypto/aes/asm/aesni-x86_64.pl new file mode 100644 index 0000000..98ca179 --- /dev/null +++ b/openssl-1.1.0h/crypto/aes/asm/aesni-x86_64.pl @@ -0,0 +1,5060 @@ +#! /usr/bin/env perl +# Copyright 2009-2016 The OpenSSL Project Authors. All Rights Reserved. +# +# Licensed under the OpenSSL license (the "License"). You may not use +# this file except in compliance with the License. You can obtain a copy +# in the file LICENSE in the source distribution or at +# https://www.openssl.org/source/license.html + +# +# ==================================================================== +# Written by Andy Polyakov for the OpenSSL +# project. The module is, however, dual licensed under OpenSSL and +# CRYPTOGAMS licenses depending on where you obtain it. For further +# details see http://www.openssl.org/~appro/cryptogams/. +# ==================================================================== +# +# This module implements support for Intel AES-NI extension. In +# OpenSSL context it's used with Intel engine, but can also be used as +# drop-in replacement for crypto/aes/asm/aes-x86_64.pl [see below for +# details]. +# +# Performance. +# +# Given aes(enc|dec) instructions' latency asymptotic performance for +# non-parallelizable modes such as CBC encrypt is 3.75 cycles per byte +# processed with 128-bit key. And given their throughput asymptotic +# performance for parallelizable modes is 1.25 cycles per byte. Being +# asymptotic limit it's not something you commonly achieve in reality, +# but how close does one get? Below are results collected for +# different modes and block sized. Pairs of numbers are for en-/ +# decryption. +# +# 16-byte 64-byte 256-byte 1-KB 8-KB +# ECB 4.25/4.25 1.38/1.38 1.28/1.28 1.26/1.26 1.26/1.26 +# CTR 5.42/5.42 1.92/1.92 1.44/1.44 1.28/1.28 1.26/1.26 +# CBC 4.38/4.43 4.15/1.43 4.07/1.32 4.07/1.29 4.06/1.28 +# CCM 5.66/9.42 4.42/5.41 4.16/4.40 4.09/4.15 4.06/4.07 +# OFB 5.42/5.42 4.64/4.64 4.44/4.44 4.39/4.39 4.38/4.38 +# CFB 5.73/5.85 5.56/5.62 5.48/5.56 5.47/5.55 5.47/5.55 +# +# ECB, CTR, CBC and CCM results are free from EVP overhead. This means +# that otherwise used 'openssl speed -evp aes-128-??? -engine aesni +# [-decrypt]' will exhibit 10-15% worse results for smaller blocks. +# The results were collected with specially crafted speed.c benchmark +# in order to compare them with results reported in "Intel Advanced +# Encryption Standard (AES) New Instruction Set" White Paper Revision +# 3.0 dated May 2010. All above results are consistently better. This +# module also provides better performance for block sizes smaller than +# 128 bytes in points *not* represented in the above table. +# +# Looking at the results for 8-KB buffer. +# +# CFB and OFB results are far from the limit, because implementation +# uses "generic" CRYPTO_[c|o]fb128_encrypt interfaces relying on +# single-block aesni_encrypt, which is not the most optimal way to go. +# CBC encrypt result is unexpectedly high and there is no documented +# explanation for it. Seemingly there is a small penalty for feeding +# the result back to AES unit the way it's done in CBC mode. There is +# nothing one can do and the result appears optimal. CCM result is +# identical to CBC, because CBC-MAC is essentially CBC encrypt without +# saving output. CCM CTR "stays invisible," because it's neatly +# interleaved wih CBC-MAC. This provides ~30% improvement over +# "straghtforward" CCM implementation with CTR and CBC-MAC performed +# disjointly. Parallelizable modes practically achieve the theoretical +# limit. +# +# Looking at how results vary with buffer size. +# +# Curves are practically saturated at 1-KB buffer size. In most cases +# "256-byte" performance is >95%, and "64-byte" is ~90% of "8-KB" one. +# CTR curve doesn't follow this pattern and is "slowest" changing one +# with "256-byte" result being 87% of "8-KB." This is because overhead +# in CTR mode is most computationally intensive. Small-block CCM +# decrypt is slower than encrypt, because first CTR and last CBC-MAC +# iterations can't be interleaved. +# +# Results for 192- and 256-bit keys. +# +# EVP-free results were observed to scale perfectly with number of +# rounds for larger block sizes, i.e. 192-bit result being 10/12 times +# lower and 256-bit one - 10/14. Well, in CBC encrypt case differences +# are a tad smaller, because the above mentioned penalty biases all +# results by same constant value. In similar way function call +# overhead affects small-block performance, as well as OFB and CFB +# results. Differences are not large, most common coefficients are +# 10/11.7 and 10/13.4 (as opposite to 10/12.0 and 10/14.0), but one +# observe even 10/11.2 and 10/12.4 (CTR, OFB, CFB)... + +# January 2011 +# +# While Westmere processor features 6 cycles latency for aes[enc|dec] +# instructions, which can be scheduled every second cycle, Sandy +# Bridge spends 8 cycles per instruction, but it can schedule them +# every cycle. This means that code targeting Westmere would perform +# suboptimally on Sandy Bridge. Therefore this update. +# +# In addition, non-parallelizable CBC encrypt (as well as CCM) is +# optimized. Relative improvement might appear modest, 8% on Westmere, +# but in absolute terms it's 3.77 cycles per byte encrypted with +# 128-bit key on Westmere, and 5.07 - on Sandy Bridge. These numbers +# should be compared to asymptotic limits of 3.75 for Westmere and +# 5.00 for Sandy Bridge. Actually, the fact that they get this close +# to asymptotic limits is quite amazing. Indeed, the limit is +# calculated as latency times number of rounds, 10 for 128-bit key, +# and divided by 16, the number of bytes in block, or in other words +# it accounts *solely* for aesenc instructions. But there are extra +# instructions, and numbers so close to the asymptotic limits mean +# that it's as if it takes as little as *one* additional cycle to +# execute all of them. How is it possible? It is possible thanks to +# out-of-order execution logic, which manages to overlap post- +# processing of previous block, things like saving the output, with +# actual encryption of current block, as well as pre-processing of +# current block, things like fetching input and xor-ing it with +# 0-round element of the key schedule, with actual encryption of +# previous block. Keep this in mind... +# +# For parallelizable modes, such as ECB, CBC decrypt, CTR, higher +# performance is achieved by interleaving instructions working on +# independent blocks. In which case asymptotic limit for such modes +# can be obtained by dividing above mentioned numbers by AES +# instructions' interleave factor. Westmere can execute at most 3 +# instructions at a time, meaning that optimal interleave factor is 3, +# and that's where the "magic" number of 1.25 come from. "Optimal +# interleave factor" means that increase of interleave factor does +# not improve performance. The formula has proven to reflect reality +# pretty well on Westmere... Sandy Bridge on the other hand can +# execute up to 8 AES instructions at a time, so how does varying +# interleave factor affect the performance? Here is table for ECB +# (numbers are cycles per byte processed with 128-bit key): +# +# instruction interleave factor 3x 6x 8x +# theoretical asymptotic limit 1.67 0.83 0.625 +# measured performance for 8KB block 1.05 0.86 0.84 +# +# "as if" interleave factor 4.7x 5.8x 6.0x +# +# Further data for other parallelizable modes: +# +# CBC decrypt 1.16 0.93 0.74 +# CTR 1.14 0.91 0.74 +# +# Well, given 3x column it's probably inappropriate to call the limit +# asymptotic, if it can be surpassed, isn't it? What happens there? +# Rewind to CBC paragraph for the answer. Yes, out-of-order execution +# magic is responsible for this. Processor overlaps not only the +# additional instructions with AES ones, but even AES instuctions +# processing adjacent triplets of independent blocks. In the 6x case +# additional instructions still claim disproportionally small amount +# of additional cycles, but in 8x case number of instructions must be +# a tad too high for out-of-order logic to cope with, and AES unit +# remains underutilized... As you can see 8x interleave is hardly +# justifiable, so there no need to feel bad that 32-bit aesni-x86.pl +# utilizies 6x interleave because of limited register bank capacity. +# +# Higher interleave factors do have negative impact on Westmere +# performance. While for ECB mode it's negligible ~1.5%, other +# parallelizables perform ~5% worse, which is outweighed by ~25% +# improvement on Sandy Bridge. To balance regression on Westmere +# CTR mode was implemented with 6x aesenc interleave factor. + +# April 2011 +# +# Add aesni_xts_[en|de]crypt. Westmere spends 1.25 cycles processing +# one byte out of 8KB with 128-bit key, Sandy Bridge - 0.90. Just like +# in CTR mode AES instruction interleave factor was chosen to be 6x. + +# November 2015 +# +# Add aesni_ocb_[en|de]crypt. AES instruction interleave factor was +# chosen to be 6x. + +###################################################################### +# Current large-block performance in cycles per byte processed with +# 128-bit key (less is better). +# +# CBC en-/decrypt CTR XTS ECB OCB +# Westmere 3.77/1.25 1.25 1.25 1.26 +# * Bridge 5.07/0.74 0.75 0.90 0.85 0.98 +# Haswell 4.44/0.63 0.63 0.73 0.63 0.70 +# Skylake 2.62/0.63 0.63 0.63 0.63 +# Silvermont 5.75/3.54 3.56 4.12 3.87(*) 4.11 +# Goldmont 3.82/1.26 1.26 1.29 1.29 1.50 +# Bulldozer 5.77/0.70 0.72 0.90 0.70 0.95 +# +# (*) Atom Silvermont ECB result is suboptimal because of penalties +# incurred by operations on %xmm8-15. As ECB is not considered +# critical, nothing was done to mitigate the problem. + +$PREFIX="aesni"; # if $PREFIX is set to "AES", the script + # generates drop-in replacement for + # crypto/aes/asm/aes-x86_64.pl:-) + +$flavour = shift; +$output = shift; +if ($flavour =~ /\./) { $output = $flavour; undef $flavour; } + +$win64=0; $win64=1 if ($flavour =~ /[nm]asm|mingw64/ || $output =~ /\.asm$/); + +$0 =~ m/(.*[\/\\])[^\/\\]+$/; $dir=$1; +( $xlate="${dir}x86_64-xlate.pl" and -f $xlate ) or +( $xlate="${dir}../../perlasm/x86_64-xlate.pl" and -f $xlate) or +die "can't locate x86_64-xlate.pl"; + +open OUT,"| \"$^X\" \"$xlate\" $flavour \"$output\""; +*STDOUT=*OUT; + +$movkey = $PREFIX eq "aesni" ? "movups" : "movups"; +@_4args=$win64? ("%rcx","%rdx","%r8", "%r9") : # Win64 order + ("%rdi","%rsi","%rdx","%rcx"); # Unix order + +$code=".text\n"; +$code.=".extern OPENSSL_ia32cap_P\n"; + +$rounds="%eax"; # input to and changed by aesni_[en|de]cryptN !!! +# this is natural Unix argument order for public $PREFIX_[ecb|cbc]_encrypt ... +$inp="%rdi"; +$out="%rsi"; +$len="%rdx"; +$key="%rcx"; # input to and changed by aesni_[en|de]cryptN !!! +$ivp="%r8"; # cbc, ctr, ... + +$rnds_="%r10d"; # backup copy for $rounds +$key_="%r11"; # backup copy for $key + +# %xmm register layout +$rndkey0="%xmm0"; $rndkey1="%xmm1"; +$inout0="%xmm2"; $inout1="%xmm3"; +$inout2="%xmm4"; $inout3="%xmm5"; +$inout4="%xmm6"; $inout5="%xmm7"; +$inout6="%xmm8"; $inout7="%xmm9"; + +$in2="%xmm6"; $in1="%xmm7"; # used in CBC decrypt, CTR, ... +$in0="%xmm8"; $iv="%xmm9"; + +# Inline version of internal aesni_[en|de]crypt1. +# +# Why folded loop? Because aes[enc|dec] is slow enough to accommodate +# cycles which take care of loop variables... +{ my $sn; +sub aesni_generate1 { +my ($p,$key,$rounds,$inout,$ivec)=@_; $inout=$inout0 if (!defined($inout)); +++$sn; +$code.=<<___; + $movkey ($key),$rndkey0 + $movkey 16($key),$rndkey1 +___ +$code.=<<___ if (defined($ivec)); + xorps $rndkey0,$ivec + lea 32($key),$key + xorps $ivec,$inout +___ +$code.=<<___ if (!defined($ivec)); + lea 32($key),$key + xorps $rndkey0,$inout +___ +$code.=<<___; +.Loop_${p}1_$sn: + aes${p} $rndkey1,$inout + dec $rounds + $movkey ($key),$rndkey1 + lea 16($key),$key + jnz .Loop_${p}1_$sn # loop body is 16 bytes + aes${p}last $rndkey1,$inout +___ +}} +# void $PREFIX_[en|de]crypt (const void *inp,void *out,const AES_KEY *key); +# +{ my ($inp,$out,$key) = @_4args; + +$code.=<<___; +.globl ${PREFIX}_encrypt +.type ${PREFIX}_encrypt,\@abi-omnipotent +.align 16 +${PREFIX}_encrypt: + movups ($inp),$inout0 # load input + mov 240($key),$rounds # key->rounds +___ + &aesni_generate1("enc",$key,$rounds); +$code.=<<___; + pxor $rndkey0,$rndkey0 # clear register bank + pxor $rndkey1,$rndkey1 + movups $inout0,($out) # output + pxor $inout0,$inout0 + ret +.size ${PREFIX}_encrypt,.-${PREFIX}_encrypt + +.globl ${PREFIX}_decrypt +.type ${PREFIX}_decrypt,\@abi-omnipotent +.align 16 +${PREFIX}_decrypt: + movups ($inp),$inout0 # load input + mov 240($key),$rounds # key->rounds +___ + &aesni_generate1("dec",$key,$rounds); +$code.=<<___; + pxor $rndkey0,$rndkey0 # clear register bank + pxor $rndkey1,$rndkey1 + movups $inout0,($out) # output + pxor $inout0,$inout0 + ret +.size ${PREFIX}_decrypt, .-${PREFIX}_decrypt +___ +} + +# _aesni_[en|de]cryptN are private interfaces, N denotes interleave +# factor. Why 3x subroutine were originally used in loops? Even though +# aes[enc|dec] latency was originally 6, it could be scheduled only +# every *2nd* cycle. Thus 3x interleave was the one providing optimal +# utilization, i.e. when subroutine's throughput is virtually same as +# of non-interleaved subroutine [for number of input blocks up to 3]. +# This is why it originally made no sense to implement 2x subroutine. +# But times change and it became appropriate to spend extra 192 bytes +# on 2x subroutine on Atom Silvermont account. For processors that +# can schedule aes[enc|dec] every cycle optimal interleave factor +# equals to corresponding instructions latency. 8x is optimal for +# * Bridge and "super-optimal" for other Intel CPUs... + +sub aesni_generate2 { +my $dir=shift; +# As already mentioned it takes in $key and $rounds, which are *not* +# preserved. $inout[0-1] is cipher/clear text... +$code.=<<___; +.type _aesni_${dir}rypt2,\@abi-omnipotent +.align 16 +_aesni_${dir}rypt2: + $movkey ($key),$rndkey0 + shl \$4,$rounds + $movkey 16($key),$rndkey1 + xorps $rndkey0,$inout0 + xorps $rndkey0,$inout1 + $movkey 32($key),$rndkey0 + lea 32($key,$rounds),$key + neg %rax # $rounds + add \$16,%rax + +.L${dir}_loop2: + aes${dir} $rndkey1,$inout0 + aes${dir} $rndkey1,$inout1 + $movkey ($key,%rax),$rndkey1 + add \$32,%rax + aes${dir} $rndkey0,$inout0 + aes${dir} $rndkey0,$inout1 + $movkey -16($key,%rax),$rndkey0 + jnz .L${dir}_loop2 + + aes${dir} $rndkey1,$inout0 + aes${dir} $rndkey1,$inout1 + aes${dir}last $rndkey0,$inout0 + aes${dir}last $rndkey0,$inout1 + ret +.size _aesni_${dir}rypt2,.-_aesni_${dir}rypt2 +___ +} +sub aesni_generate3 { +my $dir=shift; +# As already mentioned it takes in $key and $rounds, which are *not* +# preserved. $inout[0-2] is cipher/clear text... +$code.=<<___; +.type _aesni_${dir}rypt3,\@abi-omnipotent +.align 16 +_aesni_${dir}rypt3: + $movkey ($key),$rndkey0 + shl \$4,$rounds + $movkey 16($key),$rndkey1 + xorps $rndkey0,$inout0 + xorps $rndkey0,$inout1 + xorps $rndkey0,$inout2 + $movkey 32($key),$rndkey0 + lea 32($key,$rounds),$key + neg %rax # $rounds + add \$16,%rax + +.L${dir}_loop3: + aes${dir} $rndkey1,$inout0 + aes${dir} $rndkey1,$inout1 + aes${dir} $rndkey1,$inout2 + $movkey ($key,%rax),$rndkey1 + add \$32,%rax + aes${dir} $rndkey0,$inout0 + aes${dir} $rndkey0,$inout1 + aes${dir} $rndkey0,$inout2 + $movkey -16($key,%rax),$rndkey0 + jnz .L${dir}_loop3 + + aes${dir} $rndkey1,$inout0 + aes${dir} $rndkey1,$inout1 + aes${dir} $rndkey1,$inout2 + aes${dir}last $rndkey0,$inout0 + aes${dir}last $rndkey0,$inout1 + aes${dir}last $rndkey0,$inout2 + ret +.size _aesni_${dir}rypt3,.-_aesni_${dir}rypt3 +___ +} +# 4x interleave is implemented to improve small block performance, +# most notably [and naturally] 4 block by ~30%. One can argue that one +# should have implemented 5x as well, but improvement would be <20%, +# so it's not worth it... +sub aesni_generate4 { +my $dir=shift; +# As already mentioned it takes in $key and $rounds, which are *not* +# preserved. $inout[0-3] is cipher/clear text... +$code.=<<___; +.type _aesni_${dir}rypt4,\@abi-omnipotent +.align 16 +_aesni_${dir}rypt4: + $movkey ($key),$rndkey0 + shl \$4,$rounds + $movkey 16($key),$rndkey1 + xorps $rndkey0,$inout0 + xorps $rndkey0,$inout1 + xorps $rndkey0,$inout2 + xorps $rndkey0,$inout3 + $movkey 32($key),$rndkey0 + lea 32($key,$rounds),$key + neg %rax # $rounds + .byte 0x0f,0x1f,0x00 + add \$16,%rax + +.L${dir}_loop4: + aes${dir} $rndkey1,$inout0 + aes${dir} $rndkey1,$inout1 + aes${dir} $rndkey1,$inout2 + aes${dir} $rndkey1,$inout3 + $movkey ($key,%rax),$rndkey1 + add \$32,%rax + aes${dir} $rndkey0,$inout0 + aes${dir} $rndkey0,$inout1 + aes${dir} $rndkey0,$inout2 + aes${dir} $rndkey0,$inout3 + $movkey -16($key,%rax),$rndkey0 + jnz .L${dir}_loop4 + + aes${dir} $rndkey1,$inout0 + aes${dir} $rndkey1,$inout1 + aes${dir} $rndkey1,$inout2 + aes${dir} $rndkey1,$inout3 + aes${dir}last $rndkey0,$inout0 + aes${dir}last $rndkey0,$inout1 + aes${dir}last $rndkey0,$inout2 + aes${dir}last $rndkey0,$inout3 + ret +.size _aesni_${dir}rypt4,.-_aesni_${dir}rypt4 +___ +} +sub aesni_generate6 { +my $dir=shift; +# As already mentioned it takes in $key and $rounds, which are *not* +# preserved. $inout[0-5] is cipher/clear text... +$code.=<<___; +.type _aesni_${dir}rypt6,\@abi-omnipotent +.align 16 +_aesni_${dir}rypt6: + $movkey ($key),$rndkey0 + shl \$4,$rounds + $movkey 16($key),$rndkey1 + xorps $rndkey0,$inout0 + pxor $rndkey0,$inout1 + pxor $rndkey0,$inout2 + aes${dir} $rndkey1,$inout0 + lea 32($key,$rounds),$key + neg %rax # $rounds + aes${dir} $rndkey1,$inout1 + pxor $rndkey0,$inout3 + pxor $rndkey0,$inout4 + aes${dir} $rndkey1,$inout2 + pxor $rndkey0,$inout5 + $movkey ($key,%rax),$rndkey0 + add \$16,%rax + jmp .L${dir}_loop6_enter +.align 16 +.L${dir}_loop6: + aes${dir} $rndkey1,$inout0 + aes${dir} $rndkey1,$inout1 + aes${dir} $rndkey1,$inout2 +.L${dir}_loop6_enter: + aes${dir} $rndkey1,$inout3 + aes${dir} $rndkey1,$inout4 + aes${dir} $rndkey1,$inout5 + $movkey ($key,%rax),$rndkey1 + add \$32,%rax + aes${dir} $rndkey0,$inout0 + aes${dir} $rndkey0,$inout1 + aes${dir} $rndkey0,$inout2 + aes${dir} $rndkey0,$inout3 + aes${dir} $rndkey0,$inout4 + aes${dir} $rndkey0,$inout5 + $movkey -16($key,%rax),$rndkey0 + jnz .L${dir}_loop6 + + aes${dir} $rndkey1,$inout0 + aes${dir} $rndkey1,$inout1 + aes${dir} $rndkey1,$inout2 + aes${dir} $rndkey1,$inout3 + aes${dir} $rndkey1,$inout4 + aes${dir} $rndkey1,$inout5 + aes${dir}last $rndkey0,$inout0 + aes${dir}last $rndkey0,$inout1 + aes${dir}last $rndkey0,$inout2 + aes${dir}last $rndkey0,$inout3 + aes${dir}last $rndkey0,$inout4 + aes${dir}last $rndkey0,$inout5 + ret +.size _aesni_${dir}rypt6,.-_aesni_${dir}rypt6 +___ +} +sub aesni_generate8 { +my $dir=shift; +# As already mentioned it takes in $key and $rounds, which are *not* +# preserved. $inout[0-7] is cipher/clear text... +$code.=<<___; +.type _aesni_${dir}rypt8,\@abi-omnipotent +.align 16 +_aesni_${dir}rypt8: + $movkey ($key),$rndkey0 + shl \$4,$rounds + $movkey 16($key),$rndkey1 + xorps $rndkey0,$inout0 + xorps $rndkey0,$inout1 + pxor $rndkey0,$inout2 + pxor $rndkey0,$inout3 + pxor $rndkey0,$inout4 + lea 32($key,$rounds),$key + neg %rax # $rounds + aes${dir} $rndkey1,$inout0 + pxor $rndkey0,$inout5 + pxor $rndkey0,$inout6 + aes${dir} $rndkey1,$inout1 + pxor $rndkey0,$inout7 + $movkey ($key,%rax),$rndkey0 + add \$16,%rax + jmp .L${dir}_loop8_inner +.align 16 +.L${dir}_loop8: + aes${dir} $rndkey1,$inout0 + aes${dir} $rndkey1,$inout1 +.L${dir}_loop8_inner: + aes${dir} $rndkey1,$inout2 + aes${dir} $rndkey1,$inout3 + aes${dir} $rndkey1,$inout4 + aes${dir} $rndkey1,$inout5 + aes${dir} $rndkey1,$inout6 + aes${dir} $rndkey1,$inout7 +.L${dir}_loop8_enter: + $movkey ($key,%rax),$rndkey1 + add \$32,%rax + aes${dir} $rndkey0,$inout0 + aes${dir} $rndkey0,$inout1 + aes${dir} $rndkey0,$inout2 + aes${dir} $rndkey0,$inout3 + aes${dir} $rndkey0,$inout4 + aes${dir} $rndkey0,$inout5 + aes${dir} $rndkey0,$inout6 + aes${dir} $rndkey0,$inout7 + $movkey -16($key,%rax),$rndkey0 + jnz .L${dir}_loop8 + + aes${dir} $rndkey1,$inout0 + aes${dir} $rndkey1,$inout1 + aes${dir} $rndkey1,$inout2 + aes${dir} $rndkey1,$inout3 + aes${dir} $rndkey1,$inout4 + aes${dir} $rndkey1,$inout5 + aes${dir} $rndkey1,$inout6 + aes${dir} $rndkey1,$inout7 + aes${dir}last $rndkey0,$inout0 + aes${dir}last $rndkey0,$inout1 + aes${dir}last $rndkey0,$inout2 + aes${dir}last $rndkey0,$inout3 + aes${dir}last $rndkey0,$inout4 + aes${dir}last $rndkey0,$inout5 + aes${dir}last $rndkey0,$inout6 + aes${dir}last $rndkey0,$inout7 + ret +.size _aesni_${dir}rypt8,.-_aesni_${dir}rypt8 +___ +} +&aesni_generate2("enc") if ($PREFIX eq "aesni"); +&aesni_generate2("dec"); +&aesni_generate3("enc") if ($PREFIX eq "aesni"); +&aesni_generate3("dec"); +&aesni_generate4("enc") if ($PREFIX eq "aesni"); +&aesni_generate4("dec"); +&aesni_generate6("enc") if ($PREFIX eq "aesni"); +&aesni_generate6("dec"); +&aesni_generate8("enc") if ($PREFIX eq "aesni"); +&aesni_generate8("dec"); + +if ($PREFIX eq "aesni") { +######################################################################## +# void aesni_ecb_encrypt (const void *in, void *out, +# size_t length, const AES_KEY *key, +# int enc); +$code.=<<___; +.globl aesni_ecb_encrypt +.type aesni_ecb_encrypt,\@function,5 +.align 16 +aesni_ecb_encrypt: +___ +$code.=<<___ if ($win64); + lea -0x58(%rsp),%rsp + movaps %xmm6,(%rsp) # offload $inout4..7 + movaps %xmm7,0x10(%rsp) + movaps %xmm8,0x20(%rsp) + movaps %xmm9,0x30(%rsp) +.Lecb_enc_body: +___ +$code.=<<___; + and \$-16,$len # if ($len<16) + jz .Lecb_ret # return + + mov 240($key),$rounds # key->rounds + $movkey ($key),$rndkey0 + mov $key,$key_ # backup $key + mov $rounds,$rnds_ # backup $rounds + test %r8d,%r8d # 5th argument + jz .Lecb_decrypt +#--------------------------- ECB ENCRYPT ------------------------------# + cmp \$0x80,$len # if ($len<8*16) + jb .Lecb_enc_tail # short input + + movdqu ($inp),$inout0 # load 8 input blocks + movdqu 0x10($inp),$inout1 + movdqu 0x20($inp),$inout2 + movdqu 0x30($inp),$inout3 + movdqu 0x40($inp),$inout4 + movdqu 0x50($inp),$inout5 + movdqu 0x60($inp),$inout6 + movdqu 0x70($inp),$inout7 + lea 0x80($inp),$inp # $inp+=8*16 + sub \$0x80,$len # $len-=8*16 (can be zero) + jmp .Lecb_enc_loop8_enter +.align 16 +.Lecb_enc_loop8: + movups $inout0,($out) # store 8 output blocks + mov $key_,$key # restore $key + movdqu ($inp),$inout0 # load 8 input blocks + mov $rnds_,$rounds # restore $rounds + movups $inout1,0x10($out) + movdqu 0x10($inp),$inout1 + movups $inout2,0x20($out) + movdqu 0x20($inp),$inout2 + movups $inout3,0x30($out) + movdqu 0x30($inp),$inout3 + movups $inout4,0x40($out) + movdqu 0x40($inp),$inout4 + movups $inout5,0x50($out) + movdqu 0x50($inp),$inout5 + movups $inout6,0x60($out) + movdqu 0x60($inp),$inout6 + movups $inout7,0x70($out) + lea 0x80($out),$out # $out+=8*16 + movdqu 0x70($inp),$inout7 + lea 0x80($inp),$inp # $inp+=8*16 +.Lecb_enc_loop8_enter: + + call _aesni_encrypt8 + + sub \$0x80,$len + jnc .Lecb_enc_loop8 # loop if $len-=8*16 didn't borrow + + movups $inout0,($out) # store 8 output blocks + mov $key_,$key # restore $key + movups $inout1,0x10($out) + mov $rnds_,$rounds # restore $rounds + movups $inout2,0x20($out) + movups $inout3,0x30($out) + movups $inout4,0x40($out) + movups $inout5,0x50($out) + movups $inout6,0x60($out) + movups $inout7,0x70($out) + lea 0x80($out),$out # $out+=8*16 + add \$0x80,$len # restore real remaining $len + jz .Lecb_ret # done if ($len==0) + +.Lecb_enc_tail: # $len is less than 8*16 + movups ($inp),$inout0 + cmp \$0x20,$len + jb .Lecb_enc_one + movups 0x10($inp),$inout1 + je .Lecb_enc_two + movups 0x20($inp),$inout2 + cmp \$0x40,$len + jb .Lecb_enc_three + movups 0x30($inp),$inout3 + je .Lecb_enc_four + movups 0x40($inp),$inout4 + cmp \$0x60,$len + jb .Lecb_enc_five + movups 0x50($inp),$inout5 + je .Lecb_enc_six + movdqu 0x60($inp),$inout6 + xorps $inout7,$inout7 + call _aesni_encrypt8 + movups $inout0,($out) # store 7 output blocks + movups $inout1,0x10($out) + movups $inout2,0x20($out) + movups $inout3,0x30($out) + movups $inout4,0x40($out) + movups $inout5,0x50($out) + movups $inout6,0x60($out) + jmp .Lecb_ret +.align 16 +.Lecb_enc_one: +___ + &aesni_generate1("enc",$key,$rounds); +$code.=<<___; + movups $inout0,($out) # store one output block + jmp .Lecb_ret +.align 16 +.Lecb_enc_two: + call _aesni_encrypt2 + movups $inout0,($out) # store 2 output blocks + movups $inout1,0x10($out) + jmp .Lecb_ret +.align 16 +.Lecb_enc_three: + call _aesni_encrypt3 + movups $inout0,($out) # store 3 output blocks + movups $inout1,0x10($out) + movups $inout2,0x20($out) + jmp .Lecb_ret +.align 16 +.Lecb_enc_four: + call _aesni_encrypt4 + movups $inout0,($out) # store 4 output blocks + movups $inout1,0x10($out) + movups $inout2,0x20($out) + movups $inout3,0x30($out) + jmp .Lecb_ret +.align 16 +.Lecb_enc_five: + xorps $inout5,$inout5 + call _aesni_encrypt6 + movups $inout0,($out) # store 5 output blocks + movups $inout1,0x10($out) + movups $inout2,0x20($out) + movups $inout3,0x30($out) + movups $inout4,0x40($out) + jmp .Lecb_ret +.align 16 +.Lecb_enc_six: + call _aesni_encrypt6 + movups $inout0,($out) # store 6 output blocks + movups $inout1,0x10($out) + movups $inout2,0x20($out) + movups $inout3,0x30($out) + movups $inout4,0x40($out) + movups $inout5,0x50($out) + jmp .Lecb_ret + #--------------------------- ECB DECRYPT ------------------------------# +.align 16 +.Lecb_decrypt: + cmp \$0x80,$len # if ($len<8*16) + jb .Lecb_dec_tail # short input + + movdqu ($inp),$inout0 # load 8 input blocks + movdqu 0x10($inp),$inout1 + movdqu 0x20($inp),$inout2 + movdqu 0x30($inp),$inout3 + movdqu 0x40($inp),$inout4 + movdqu 0x50($inp),$inout5 + movdqu 0x60($inp),$inout6 + movdqu 0x70($inp),$inout7 + lea 0x80($inp),$inp # $inp+=8*16 + sub \$0x80,$len # $len-=8*16 (can be zero) + jmp .Lecb_dec_loop8_enter +.align 16 +.Lecb_dec_loop8: + movups $inout0,($out) # store 8 output blocks + mov $key_,$key # restore $key + movdqu ($inp),$inout0 # load 8 input blocks + mov $rnds_,$rounds # restore $rounds + movups $inout1,0x10($out) + movdqu 0x10($inp),$inout1 + movups $inout2,0x20($out) + movdqu 0x20($inp),$inout2 + movups $inout3,0x30($out) + movdqu 0x30($inp),$inout3 + movups $inout4,0x40($out) + movdqu 0x40($inp),$inout4 + movups $inout5,0x50($out) + movdqu 0x50($inp),$inout5 + movups $inout6,0x60($out) + movdqu 0x60($inp),$inout6 + movups $inout7,0x70($out) + lea 0x80($out),$out # $out+=8*16 + movdqu 0x70($inp),$inout7 + lea 0x80($inp),$inp # $inp+=8*16 +.Lecb_dec_loop8_enter: + + call _aesni_decrypt8 + + $movkey ($key_),$rndkey0 + sub \$0x80,$len + jnc .Lecb_dec_loop8 # loop if $len-=8*16 didn't borrow + + movups $inout0,($out) # store 8 output blocks + pxor $inout0,$inout0 # clear register bank + mov $key_,$key # restore $key + movups $inout1,0x10($out) + pxor $inout1,$inout1 + mov $rnds_,$rounds # restore $rounds + movups $inout2,0x20($out) + pxor $inout2,$inout2 + movups $inout3,0x30($out) + pxor $inout3,$inout3 + movups $inout4,0x40($out) + pxor $inout4,$inout4 + movups $inout5,0x50($out) + pxor $inout5,$inout5 + movups $inout6,0x60($out) + pxor $inout6,$inout6 + movups $inout7,0x70($out) + pxor $inout7,$inout7 + lea 0x80($out),$out # $out+=8*16 + add \$0x80,$len # restore real remaining $len + jz .Lecb_ret # done if ($len==0) + +.Lecb_dec_tail: + movups ($inp),$inout0 + cmp \$0x20,$len + jb .Lecb_dec_one + movups 0x10($inp),$inout1 + je .Lecb_dec_two + movups 0x20($inp),$inout2 + cmp \$0x40,$len + jb .Lecb_dec_three + movups 0x30($inp),$inout3 + je .Lecb_dec_four + movups 0x40($inp),$inout4 + cmp \$0x60,$len + jb .Lecb_dec_five + movups 0x50($inp),$inout5 + je .Lecb_dec_six + movups 0x60($inp),$inout6 + $movkey ($key),$rndkey0 + xorps $inout7,$inout7 + call _aesni_decrypt8 + movups $inout0,($out) # store 7 output blocks + pxor $inout0,$inout0 # clear register bank + movups $inout1,0x10($out) + pxor $inout1,$inout1 + movups $inout2,0x20($out) + pxor $inout2,$inout2 + movups $inout3,0x30($out) + pxor $inout3,$inout3 + movups $inout4,0x40($out) + pxor $inout4,$inout4 + movups $inout5,0x50($out) + pxor $inout5,$inout5 + movups $inout6,0x60($out) + pxor $inout6,$inout6 + pxor $inout7,$inout7 + jmp .Lecb_ret +.align 16 +.Lecb_dec_one: +___ + &aesni_generate1("dec",$key,$rounds); +$code.=<<___; + movups $inout0,($out) # store one output block + pxor $inout0,$inout0 # clear register bank + jmp .Lecb_ret +.align 16 +.Lecb_dec_two: + call _aesni_decrypt2 + movups $inout0,($out) # store 2 output blocks + pxor $inout0,$inout0 # clear register bank + movups $inout1,0x10($out) + pxor $inout1,$inout1 + jmp .Lecb_ret +.align 16 +.Lecb_dec_three: + call _aesni_decrypt3 + movups $inout0,($out) # store 3 output blocks + pxor $inout0,$inout0 # clear register bank + movups $inout1,0x10($out) + pxor $inout1,$inout1 + movups $inout2,0x20($out) + pxor $inout2,$inout2 + jmp .Lecb_ret +.align 16 +.Lecb_dec_four: + call _aesni_decrypt4 + movups $inout0,($out) # store 4 output blocks + pxor $inout0,$inout0 # clear register bank + movups $inout1,0x10($out) + pxor $inout1,$inout1 + movups $inout2,0x20($out) + pxor $inout2,$inout2 + movups $inout3,0x30($out) + pxor $inout3,$inout3 + jmp .Lecb_ret +.align 16 +.Lecb_dec_five: + xorps $inout5,$inout5 + call _aesni_decrypt6 + movups $inout0,($out) # store 5 output blocks + pxor $inout0,$inout0 # clear register bank + movups $inout1,0x10($out) + pxor $inout1,$inout1 + movups $inout2,0x20($out) + pxor $inout2,$inout2 + movups $inout3,0x30($out) + pxor $inout3,$inout3 + movups $inout4,0x40($out) + pxor $inout4,$inout4 + pxor $inout5,$inout5 + jmp .Lecb_ret +.align 16 +.Lecb_dec_six: + call _aesni_decrypt6 + movups $inout0,($out) # store 6 output blocks + pxor $inout0,$inout0 # clear register bank + movups $inout1,0x10($out) + pxor $inout1,$inout1 + movups $inout2,0x20($out) + pxor $inout2,$inout2 + movups $inout3,0x30($out) + pxor $inout3,$inout3 + movups $inout4,0x40($out) + pxor $inout4,$inout4 + movups $inout5,0x50($out) + pxor $inout5,$inout5 + +.Lecb_ret: + xorps $rndkey0,$rndkey0 # %xmm0 + pxor $rndkey1,$rndkey1 +___ +$code.=<<___ if ($win64); + movaps (%rsp),%xmm6 + movaps %xmm0,(%rsp) # clear stack + movaps 0x10(%rsp),%xmm7 + movaps %xmm0,0x10(%rsp) + movaps 0x20(%rsp),%xmm8 + movaps %xmm0,0x20(%rsp) + movaps 0x30(%rsp),%xmm9 + movaps %xmm0,0x30(%rsp) + lea 0x58(%rsp),%rsp +.Lecb_enc_ret: +___ +$code.=<<___; + ret +.size aesni_ecb_encrypt,.-aesni_ecb_encrypt +___ + +{ +###################################################################### +# void aesni_ccm64_[en|de]crypt_blocks (const void *in, void *out, +# size_t blocks, const AES_KEY *key, +# const char *ivec,char *cmac); +# +# Handles only complete blocks, operates on 64-bit counter and +# does not update *ivec! Nor does it finalize CMAC value +# (see engine/eng_aesni.c for details) +# +{ +my $cmac="%r9"; # 6th argument + +my $increment="%xmm9"; +my $iv="%xmm6"; +my $bswap_mask="%xmm7"; + +$code.=<<___; +.globl aesni_ccm64_encrypt_blocks +.type aesni_ccm64_encrypt_blocks,\@function,6 +.align 16 +aesni_ccm64_encrypt_blocks: +___ +$code.=<<___ if ($win64); + lea -0x58(%rsp),%rsp + movaps %xmm6,(%rsp) # $iv + movaps %xmm7,0x10(%rsp) # $bswap_mask + movaps %xmm8,0x20(%rsp) # $in0 + movaps %xmm9,0x30(%rsp) # $increment +.Lccm64_enc_body: +___ +$code.=<<___; + mov 240($key),$rounds # key->rounds + movdqu ($ivp),$iv + movdqa .Lincrement64(%rip),$increment + movdqa .Lbswap_mask(%rip),$bswap_mask + + shl \$4,$rounds + mov \$16,$rnds_ + lea 0($key),$key_ + movdqu ($cmac),$inout1 + movdqa $iv,$inout0 + lea 32($key,$rounds),$key # end of key schedule + pshufb $bswap_mask,$iv + sub %rax,%r10 # twisted $rounds + jmp .Lccm64_enc_outer +.align 16 +.Lccm64_enc_outer: + $movkey ($key_),$rndkey0 + mov %r10,%rax + movups ($inp),$in0 # load inp + + xorps $rndkey0,$inout0 # counter + $movkey 16($key_),$rndkey1 + xorps $in0,$rndkey0 + xorps $rndkey0,$inout1 # cmac^=inp + $movkey 32($key_),$rndkey0 + +.Lccm64_enc2_loop: + aesenc $rndkey1,$inout0 + aesenc $rndkey1,$inout1 + $movkey ($key,%rax),$rndkey1 + add \$32,%rax + aesenc $rndkey0,$inout0 + aesenc $rndkey0,$inout1 + $movkey -16($key,%rax),$rndkey0 + jnz .Lccm64_enc2_loop + aesenc $rndkey1,$inout0 + aesenc $rndkey1,$inout1 + paddq $increment,$iv + dec $len # $len-- ($len is in blocks) + aesenclast $rndkey0,$inout0 + aesenclast $rndkey0,$inout1 + + lea 16($inp),$inp + xorps $inout0,$in0 # inp ^= E(iv) + movdqa $iv,$inout0 + movups $in0,($out) # save output + pshufb $bswap_mask,$inout0 + lea 16($out),$out # $out+=16 + jnz .Lccm64_enc_outer # loop if ($len!=0) + + pxor $rndkey0,$rndkey0 # clear register bank + pxor $rndkey1,$rndkey1 + pxor $inout0,$inout0 + movups $inout1,($cmac) # store resulting mac + pxor $inout1,$inout1 + pxor $in0,$in0 + pxor $iv,$iv +___ +$code.=<<___ if ($win64); + movaps (%rsp),%xmm6 + movaps %xmm0,(%rsp) # clear stack + movaps 0x10(%rsp),%xmm7 + movaps %xmm0,0x10(%rsp) + movaps 0x20(%rsp),%xmm8 + movaps %xmm0,0x20(%rsp) + movaps 0x30(%rsp),%xmm9 + movaps %xmm0,0x30(%rsp) + lea 0x58(%rsp),%rsp +.Lccm64_enc_ret: +___ +$code.=<<___; + ret +.size aesni_ccm64_encrypt_blocks,.-aesni_ccm64_encrypt_blocks +___ +###################################################################### +$code.=<<___; +.globl aesni_ccm64_decrypt_blocks +.type aesni_ccm64_decrypt_blocks,\@function,6 +.align 16 +aesni_ccm64_decrypt_blocks: +___ +$code.=<<___ if ($win64); + lea -0x58(%rsp),%rsp + movaps %xmm6,(%rsp) # $iv + movaps %xmm7,0x10(%rsp) # $bswap_mask + movaps %xmm8,0x20(%rsp) # $in8 + movaps %xmm9,0x30(%rsp) # $increment +.Lccm64_dec_body: +___ +$code.=<<___; + mov 240($key),$rounds # key->rounds + movups ($ivp),$iv + movdqu ($cmac),$inout1 + movdqa .Lincrement64(%rip),$increment + movdqa .Lbswap_mask(%rip),$bswap_mask + + movaps $iv,$inout0 + mov $rounds,$rnds_ + mov $key,$key_ + pshufb $bswap_mask,$iv +___ + &aesni_generate1("enc",$key,$rounds); +$code.=<<___; + shl \$4,$rnds_ + mov \$16,$rounds + movups ($inp),$in0 # load inp + paddq $increment,$iv + lea 16($inp),$inp # $inp+=16 + sub %r10,%rax # twisted $rounds + lea 32($key_,$rnds_),$key # end of key schedule + mov %rax,%r10 + jmp .Lccm64_dec_outer +.align 16 +.Lccm64_dec_outer: + xorps $inout0,$in0 # inp ^= E(iv) + movdqa $iv,$inout0 + movups $in0,($out) # save output + lea 16($out),$out # $out+=16 + pshufb $bswap_mask,$inout0 + + sub \$1,$len # $len-- ($len is in blocks) + jz .Lccm64_dec_break # if ($len==0) break + + $movkey ($key_),$rndkey0 + mov %r10,%rax + $movkey 16($key_),$rndkey1 + xorps $rndkey0,$in0 + xorps $rndkey0,$inout0 + xorps $in0,$inout1 # cmac^=out + $movkey 32($key_),$rndkey0 + jmp .Lccm64_dec2_loop +.align 16 +.Lccm64_dec2_loop: + aesenc $rndkey1,$inout0 + aesenc $rndkey1,$inout1 + $movkey ($key,%rax),$rndkey1 + add \$32,%rax + aesenc $rndkey0,$inout0 + aesenc $rndkey0,$inout1 + $movkey -16($key,%rax),$rndkey0 + jnz .Lccm64_dec2_loop + movups ($inp),$in0 # load input + paddq $increment,$iv + aesenc $rndkey1,$inout0 + aesenc $rndkey1,$inout1 + aesenclast $rndkey0,$inout0 + aesenclast $rndkey0,$inout1 + lea 16($inp),$inp # $inp+=16 + jmp .Lccm64_dec_outer + +.align 16 +.Lccm64_dec_break: + #xorps $in0,$inout1 # cmac^=out + mov 240($key_),$rounds +___ + &aesni_generate1("enc",$key_,$rounds,$inout1,$in0); +$code.=<<___; + pxor $rndkey0,$rndkey0 # clear register bank + pxor $rndkey1,$rndkey1 + pxor $inout0,$inout0 + movups $inout1,($cmac) # store resulting mac + pxor $inout1,$inout1 + pxor $in0,$in0 + pxor $iv,$iv +___ +$code.=<<___ if ($win64); + movaps (%rsp),%xmm6 + movaps %xmm0,(%rsp) # clear stack + movaps 0x10(%rsp),%xmm7 + movaps %xmm0,0x10(%rsp) + movaps 0x20(%rsp),%xmm8 + movaps %xmm0,0x20(%rsp) + movaps 0x30(%rsp),%xmm9 + movaps %xmm0,0x30(%rsp) + lea 0x58(%rsp),%rsp +.Lccm64_dec_ret: +___ +$code.=<<___; + ret +.size aesni_ccm64_decrypt_blocks,.-aesni_ccm64_decrypt_blocks +___ +} +###################################################################### +# void aesni_ctr32_encrypt_blocks (const void *in, void *out, +# size_t blocks, const AES_KEY *key, +# const char *ivec); +# +# Handles only complete blocks, operates on 32-bit counter and +# does not update *ivec! (see crypto/modes/ctr128.c for details) +# +# Overhaul based on suggestions from Shay Gueron and Vlad Krasnov, +# http://rt.openssl.org/Ticket/Display.html?id=3021&user=guest&pass=guest. +# Keywords are full unroll and modulo-schedule counter calculations +# with zero-round key xor. +{ +my ($in0,$in1,$in2,$in3,$in4,$in5)=map("%xmm$_",(10..15)); +my ($key0,$ctr)=("${key_}d","${ivp}d"); +my $frame_size = 0x80 + ($win64?160:0); + +$code.=<<___; +.globl aesni_ctr32_encrypt_blocks +.type aesni_ctr32_encrypt_blocks,\@function,5 +.align 16 +aesni_ctr32_encrypt_blocks: + cmp \$1,$len + jne .Lctr32_bulk + + # handle single block without allocating stack frame, + # useful when handling edges + movups ($ivp),$inout0 + movups ($inp),$inout1 + mov 240($key),%edx # key->rounds +___ + &aesni_generate1("enc",$key,"%edx"); +$code.=<<___; + pxor $rndkey0,$rndkey0 # clear register bank + pxor $rndkey1,$rndkey1 + xorps $inout1,$inout0 + pxor $inout1,$inout1 + movups $inout0,($out) + xorps $inout0,$inout0 + jmp .Lctr32_epilogue + +.align 16 +.Lctr32_bulk: + lea (%rsp),%rax + push %rbp + sub \$$frame_size,%rsp + and \$-16,%rsp # Linux kernel stack can be incorrectly seeded +___ +$code.=<<___ if ($win64); + movaps %xmm6,-0xa8(%rax) # offload everything + movaps %xmm7,-0x98(%rax) + movaps %xmm8,-0x88(%rax) + movaps %xmm9,-0x78(%rax) + movaps %xmm10,-0x68(%rax) + movaps %xmm11,-0x58(%rax) + movaps %xmm12,-0x48(%rax) + movaps %xmm13,-0x38(%rax) + movaps %xmm14,-0x28(%rax) + movaps %xmm15,-0x18(%rax) +.Lctr32_body: +___ +$code.=<<___; + lea -8(%rax),%rbp + + # 8 16-byte words on top of stack are counter values + # xor-ed with zero-round key + + movdqu ($ivp),$inout0 + movdqu ($key),$rndkey0 + mov 12($ivp),$ctr # counter LSB + pxor $rndkey0,$inout0 + mov 12($key),$key0 # 0-round key LSB + movdqa $inout0,0x00(%rsp) # populate counter block + bswap $ctr + movdqa $inout0,$inout1 + movdqa $inout0,$inout2 + movdqa $inout0,$inout3 + movdqa $inout0,0x40(%rsp) + movdqa $inout0,0x50(%rsp) + movdqa $inout0,0x60(%rsp) + mov %rdx,%r10 # about to borrow %rdx + movdqa $inout0,0x70(%rsp) + + lea 1($ctr),%rax + lea 2($ctr),%rdx + bswap %eax + bswap %edx + xor $key0,%eax + xor $key0,%edx + pinsrd \$3,%eax,$inout1 + lea 3($ctr),%rax + movdqa $inout1,0x10(%rsp) + pinsrd \$3,%edx,$inout2 + bswap %eax + mov %r10,%rdx # restore %rdx + lea 4($ctr),%r10 + movdqa $inout2,0x20(%rsp) + xor $key0,%eax + bswap %r10d + pinsrd \$3,%eax,$inout3 + xor $key0,%r10d + movdqa $inout3,0x30(%rsp) + lea 5($ctr),%r9 + mov %r10d,0x40+12(%rsp) + bswap %r9d + lea 6($ctr),%r10 + mov 240($key),$rounds # key->rounds + xor $key0,%r9d + bswap %r10d + mov %r9d,0x50+12(%rsp) + xor $key0,%r10d + lea 7($ctr),%r9 + mov %r10d,0x60+12(%rsp) + bswap %r9d + mov OPENSSL_ia32cap_P+4(%rip),%r10d + xor $key0,%r9d + and \$`1<<26|1<<22`,%r10d # isolate XSAVE+MOVBE + mov %r9d,0x70+12(%rsp) + + $movkey 0x10($key),$rndkey1 + + movdqa 0x40(%rsp),$inout4 + movdqa 0x50(%rsp),$inout5 + + cmp \$8,$len # $len is in blocks + jb .Lctr32_tail # short input if ($len<8) + + sub \$6,$len # $len is biased by -6 + cmp \$`1<<22`,%r10d # check for MOVBE without XSAVE + je .Lctr32_6x # [which denotes Atom Silvermont] + + lea 0x80($key),$key # size optimization + sub \$2,$len # $len is biased by -8 + jmp .Lctr32_loop8 + +.align 16 +.Lctr32_6x: + shl \$4,$rounds + mov \$48,$rnds_ + bswap $key0 + lea 32($key,$rounds),$key # end of key schedule + sub %rax,%r10 # twisted $rounds + jmp .Lctr32_loop6 + +.align 16 +.Lctr32_loop6: + add \$6,$ctr # next counter value + $movkey -48($key,$rnds_),$rndkey0 + aesenc $rndkey1,$inout0 + mov $ctr,%eax + xor $key0,%eax + aesenc $rndkey1,$inout1 + movbe %eax,`0x00+12`(%rsp) # store next counter value + lea 1($ctr),%eax + aesenc $rndkey1,$inout2 + xor $key0,%eax + movbe %eax,`0x10+12`(%rsp) + aesenc $rndkey1,$inout3 + lea 2($ctr),%eax + xor $key0,%eax + aesenc $rndkey1,$inout4 + movbe %eax,`0x20+12`(%rsp) + lea 3($ctr),%eax + aesenc $rndkey1,$inout5 + $movkey -32($key,$rnds_),$rndkey1 + xor $key0,%eax + + aesenc $rndkey0,$inout0 + movbe %eax,`0x30+12`(%rsp) + lea 4($ctr),%eax + aesenc $rndkey0,$inout1 + xor $key0,%eax + movbe %eax,`0x40+12`(%rsp) + aesenc $rndkey0,$inout2 + lea 5($ctr),%eax + xor $key0,%eax + aesenc $rndkey0,$inout3 + movbe %eax,`0x50+12`(%rsp) + mov %r10,%rax # mov $rnds_,$rounds + aesenc $rndkey0,$inout4 + aesenc $rndkey0,$inout5 + $movkey -16($key,$rnds_),$rndkey0 + + call .Lenc_loop6 + + movdqu ($inp),$inout6 # load 6 input blocks + movdqu 0x10($inp),$inout7 + movdqu 0x20($inp),$in0 + movdqu 0x30($inp),$in1 + movdqu 0x40($inp),$in2 + movdqu 0x50($inp),$in3 + lea 0x60($inp),$inp # $inp+=6*16 + $movkey -64($key,$rnds_),$rndkey1 + pxor $inout0,$inout6 # inp^=E(ctr) + movaps 0x00(%rsp),$inout0 # load next counter [xor-ed with 0 round] + pxor $inout1,$inout7 + movaps 0x10(%rsp),$inout1 + pxor $inout2,$in0 + movaps 0x20(%rsp),$inout2 + pxor $inout3,$in1 + movaps 0x30(%rsp),$inout3 + pxor $inout4,$in2 + movaps 0x40(%rsp),$inout4 + pxor $inout5,$in3 + movaps 0x50(%rsp),$inout5 + movdqu $inout6,($out) # store 6 output blocks + movdqu $inout7,0x10($out) + movdqu $in0,0x20($out) + movdqu $in1,0x30($out) + movdqu $in2,0x40($out) + movdqu $in3,0x50($out) + lea 0x60($out),$out # $out+=6*16 + + sub \$6,$len + jnc .Lctr32_loop6 # loop if $len-=6 didn't borrow + + add \$6,$len # restore real remaining $len + jz .Lctr32_done # done if ($len==0) + + lea -48($rnds_),$rounds + lea -80($key,$rnds_),$key # restore $key + neg $rounds + shr \$4,$rounds # restore $rounds + jmp .Lctr32_tail + +.align 32 +.Lctr32_loop8: + add \$8,$ctr # next counter value + movdqa 0x60(%rsp),$inout6 + aesenc $rndkey1,$inout0 + mov $ctr,%r9d + movdqa 0x70(%rsp),$inout7 + aesenc $rndkey1,$inout1 + bswap %r9d + $movkey 0x20-0x80($key),$rndkey0 + aesenc $rndkey1,$inout2 + xor $key0,%r9d + nop + aesenc $rndkey1,$inout3 + mov %r9d,0x00+12(%rsp) # store next counter value + lea 1($ctr),%r9 + aesenc $rndkey1,$inout4 + aesenc $rndkey1,$inout5 + aesenc $rndkey1,$inout6 + aesenc $rndkey1,$inout7 + $movkey 0x30-0x80($key),$rndkey1 +___ +for($i=2;$i<8;$i++) { +my $rndkeyx = ($i&1)?$rndkey1:$rndkey0; +$code.=<<___; + bswap %r9d + aesenc $rndkeyx,$inout0 + aesenc $rndkeyx,$inout1 + xor $key0,%r9d + .byte 0x66,0x90 + aesenc $rndkeyx,$inout2 + aesenc $rndkeyx,$inout3 + mov %r9d,`0x10*($i-1)`+12(%rsp) + lea $i($ctr),%r9 + aesenc $rndkeyx,$inout4 + aesenc $rndkeyx,$inout5 + aesenc $rndkeyx,$inout6 + aesenc $rndkeyx,$inout7 + $movkey `0x20+0x10*$i`-0x80($key),$rndkeyx +___ +} +$code.=<<___; + bswap %r9d + aesenc $rndkey0,$inout0 + aesenc $rndkey0,$inout1 + aesenc $rndkey0,$inout2 + xor $key0,%r9d + movdqu 0x00($inp),$in0 # start loading input + aesenc $rndkey0,$inout3 + mov %r9d,0x70+12(%rsp) + cmp \$11,$rounds + aesenc $rndkey0,$inout4 + aesenc $rndkey0,$inout5 + aesenc $rndkey0,$inout6 + aesenc $rndkey0,$inout7 + $movkey 0xa0-0x80($key),$rndkey0 + + jb .Lctr32_enc_done + + aesenc $rndkey1,$inout0 + aesenc $rndkey1,$inout1 + aesenc $rndkey1,$inout2 + aesenc $rndkey1,$inout3 + aesenc $rndkey1,$inout4 + aesenc $rndkey1,$inout5 + aesenc $rndkey1,$inout6 + aesenc $rndkey1,$inout7 + $movkey 0xb0-0x80($key),$rndkey1 + + aesenc $rndkey0,$inout0 + aesenc $rndkey0,$inout1 + aesenc $rndkey0,$inout2 + aesenc $rndkey0,$inout3 + aesenc $rndkey0,$inout4 + aesenc $rndkey0,$inout5 + aesenc $rndkey0,$inout6 + aesenc $rndkey0,$inout7 + $movkey 0xc0-0x80($key),$rndkey0 + je .Lctr32_enc_done + + aesenc $rndkey1,$inout0 + aesenc $rndkey1,$inout1 + aesenc $rndkey1,$inout2 + aesenc $rndkey1,$inout3 + aesenc $rndkey1,$inout4 + aesenc $rndkey1,$inout5 + aesenc $rndkey1,$inout6 + aesenc $rndkey1,$inout7 + $movkey 0xd0-0x80($key),$rndkey1 + + aesenc $rndkey0,$inout0 + aesenc $rndkey0,$inout1 + aesenc $rndkey0,$inout2 + aesenc $rndkey0,$inout3 + aesenc $rndkey0,$inout4 + aesenc $rndkey0,$inout5 + aesenc $rndkey0,$inout6 + aesenc $rndkey0,$inout7 + $movkey 0xe0-0x80($key),$rndkey0 + jmp .Lctr32_enc_done + +.align 16 +.Lctr32_enc_done: + movdqu 0x10($inp),$in1 + pxor $rndkey0,$in0 # input^=round[last] + movdqu 0x20($inp),$in2 + pxor $rndkey0,$in1 + movdqu 0x30($inp),$in3 + pxor $rndkey0,$in2 + movdqu 0x40($inp),$in4 + pxor $rndkey0,$in3 + movdqu 0x50($inp),$in5 + pxor $rndkey0,$in4 + pxor $rndkey0,$in5 + aesenc $rndkey1,$inout0 + aesenc $rndkey1,$inout1 + aesenc $rndkey1,$inout2 + aesenc $rndkey1,$inout3 + aesenc $rndkey1,$inout4 + aesenc $rndkey1,$inout5 + aesenc $rndkey1,$inout6 + aesenc $rndkey1,$inout7 + movdqu 0x60($inp),$rndkey1 # borrow $rndkey1 for inp[6] + lea 0x80($inp),$inp # $inp+=8*16 + + aesenclast $in0,$inout0 # $inN is inp[N]^round[last] + pxor $rndkey0,$rndkey1 # borrowed $rndkey + movdqu 0x70-0x80($inp),$in0 + aesenclast $in1,$inout1 + pxor $rndkey0,$in0 + movdqa 0x00(%rsp),$in1 # load next counter block + aesenclast $in2,$inout2 + aesenclast $in3,$inout3 + movdqa 0x10(%rsp),$in2 + movdqa 0x20(%rsp),$in3 + aesenclast $in4,$inout4 + aesenclast $in5,$inout5 + movdqa 0x30(%rsp),$in4 + movdqa 0x40(%rsp),$in5 + aesenclast $rndkey1,$inout6 + movdqa 0x50(%rsp),$rndkey0 + $movkey 0x10-0x80($key),$rndkey1#real 1st-round key + aesenclast $in0,$inout7 + + movups $inout0,($out) # store 8 output blocks + movdqa $in1,$inout0 + movups $inout1,0x10($out) + movdqa $in2,$inout1 + movups $inout2,0x20($out) + movdqa $in3,$inout2 + movups $inout3,0x30($out) + movdqa $in4,$inout3 + movups $inout4,0x40($out) + movdqa $in5,$inout4 + movups $inout5,0x50($out) + movdqa $rndkey0,$inout5 + movups $inout6,0x60($out) + movups $inout7,0x70($out) + lea 0x80($out),$out # $out+=8*16 + + sub \$8,$len + jnc .Lctr32_loop8 # loop if $len-=8 didn't borrow + + add \$8,$len # restore real remainig $len + jz .Lctr32_done # done if ($len==0) + lea -0x80($key),$key + +.Lctr32_tail: + # note that at this point $inout0..5 are populated with + # counter values xor-ed with 0-round key + lea 16($key),$key + cmp \$4,$len + jb .Lctr32_loop3 + je .Lctr32_loop4 + + # if ($len>4) compute 7 E(counter) + shl \$4,$rounds + movdqa 0x60(%rsp),$inout6 + pxor $inout7,$inout7 + + $movkey 16($key),$rndkey0 + aesenc $rndkey1,$inout0 + aesenc $rndkey1,$inout1 + lea 32-16($key,$rounds),$key# prepare for .Lenc_loop8_enter + neg %rax + aesenc $rndkey1,$inout2 + add \$16,%rax # prepare for .Lenc_loop8_enter + movups ($inp),$in0 + aesenc $rndkey1,$inout3 + aesenc $rndkey1,$inout4 + movups 0x10($inp),$in1 # pre-load input + movups 0x20($inp),$in2 + aesenc $rndkey1,$inout5 + aesenc $rndkey1,$inout6 + + call .Lenc_loop8_enter + + movdqu 0x30($inp),$in3 + pxor $in0,$inout0 + movdqu 0x40($inp),$in0 + pxor $in1,$inout1 + movdqu $inout0,($out) # store output + pxor $in2,$inout2 + movdqu $inout1,0x10($out) + pxor $in3,$inout3 + movdqu $inout2,0x20($out) + pxor $in0,$inout4 + movdqu $inout3,0x30($out) + movdqu $inout4,0x40($out) + cmp \$6,$len + jb .Lctr32_done # $len was 5, stop store + + movups 0x50($inp),$in1 + xorps $in1,$inout5 + movups $inout5,0x50($out) + je .Lctr32_done # $len was 6, stop store + + movups 0x60($inp),$in2 + xorps $in2,$inout6 + movups $inout6,0x60($out) + jmp .Lctr32_done # $len was 7, stop store + +.align 32 +.Lctr32_loop4: + aesenc $rndkey1,$inout0 + lea 16($key),$key + dec $rounds + aesenc $rndkey1,$inout1 + aesenc $rndkey1,$inout2 + aesenc $rndkey1,$inout3 + $movkey ($key),$rndkey1 + jnz .Lctr32_loop4 + aesenclast $rndkey1,$inout0 + aesenclast $rndkey1,$inout1 + movups ($inp),$in0 # load input + movups 0x10($inp),$in1 + aesenclast $rndkey1,$inout2 + aesenclast $rndkey1,$inout3 + movups 0x20($inp),$in2 + movups 0x30($inp),$in3 + + xorps $in0,$inout0 + movups $inout0,($out) # store output + xorps $in1,$inout1 + movups $inout1,0x10($out) + pxor $in2,$inout2 + movdqu $inout2,0x20($out) + pxor $in3,$inout3 + movdqu $inout3,0x30($out) + jmp .Lctr32_done # $len was 4, stop store + +.align 32 +.Lctr32_loop3: + aesenc $rndkey1,$inout0 + lea 16($key),$key + dec $rounds + aesenc $rndkey1,$inout1 + aesenc $rndkey1,$inout2 + $movkey ($key),$rndkey1 + jnz .Lctr32_loop3 + aesenclast $rndkey1,$inout0 + aesenclast $rndkey1,$inout1 + aesenclast $rndkey1,$inout2 + + movups ($inp),$in0 # load input + xorps $in0,$inout0 + movups $inout0,($out) # store output + cmp \$2,$len + jb .Lctr32_done # $len was 1, stop store + + movups 0x10($inp),$in1 + xorps $in1,$inout1 + movups $inout1,0x10($out) + je .Lctr32_done # $len was 2, stop store + + movups 0x20($inp),$in2 + xorps $in2,$inout2 + movups $inout2,0x20($out) # $len was 3, stop store + +.Lctr32_done: + xorps %xmm0,%xmm0 # clear regiser bank + xor $key0,$key0 + pxor %xmm1,%xmm1 + pxor %xmm2,%xmm2 + pxor %xmm3,%xmm3 + pxor %xmm4,%xmm4 + pxor %xmm5,%xmm5 +___ +$code.=<<___ if (!$win64); + pxor %xmm6,%xmm6 + pxor %xmm7,%xmm7 + movaps %xmm0,0x00(%rsp) # clear stack + pxor %xmm8,%xmm8 + movaps %xmm0,0x10(%rsp) + pxor %xmm9,%xmm9 + movaps %xmm0,0x20(%rsp) + pxor %xmm10,%xmm10 + movaps %xmm0,0x30(%rsp) + pxor %xmm11,%xmm11 + movaps %xmm0,0x40(%rsp) + pxor %xmm12,%xmm12 + movaps %xmm0,0x50(%rsp) + pxor %xmm13,%xmm13 + movaps %xmm0,0x60(%rsp) + pxor %xmm14,%xmm14 + movaps %xmm0,0x70(%rsp) + pxor %xmm15,%xmm15 +___ +$code.=<<___ if ($win64); + movaps -0xa0(%rbp),%xmm6 + movaps %xmm0,-0xa0(%rbp) # clear stack + movaps -0x90(%rbp),%xmm7 + movaps %xmm0,-0x90(%rbp) + movaps -0x80(%rbp),%xmm8 + movaps %xmm0,-0x80(%rbp) + movaps -0x70(%rbp),%xmm9 + movaps %xmm0,-0x70(%rbp) + movaps -0x60(%rbp),%xmm10 + movaps %xmm0,-0x60(%rbp) + movaps -0x50(%rbp),%xmm11 + movaps %xmm0,-0x50(%rbp) + movaps -0x40(%rbp),%xmm12 + movaps %xmm0,-0x40(%rbp) + movaps -0x30(%rbp),%xmm13 + movaps %xmm0,-0x30(%rbp) + movaps -0x20(%rbp),%xmm14 + movaps %xmm0,-0x20(%rbp) + movaps -0x10(%rbp),%xmm15 + movaps %xmm0,-0x10(%rbp) + movaps %xmm0,0x00(%rsp) + movaps %xmm0,0x10(%rsp) + movaps %xmm0,0x20(%rsp) + movaps %xmm0,0x30(%rsp) + movaps %xmm0,0x40(%rsp) + movaps %xmm0,0x50(%rsp) + movaps %xmm0,0x60(%rsp) + movaps %xmm0,0x70(%rsp) +___ +$code.=<<___; + lea (%rbp),%rsp + pop %rbp +.Lctr32_epilogue: + ret +.size aesni_ctr32_encrypt_blocks,.-aesni_ctr32_encrypt_blocks +___ +} + +###################################################################### +# void aesni_xts_[en|de]crypt(const char *inp,char *out,size_t len, +# const AES_KEY *key1, const AES_KEY *key2 +# const unsigned char iv[16]); +# +{ +my @tweak=map("%xmm$_",(10..15)); +my ($twmask,$twres,$twtmp)=("%xmm8","%xmm9",@tweak[4]); +my ($key2,$ivp,$len_)=("%r8","%r9","%r9"); +my $frame_size = 0x70 + ($win64?160:0); + +$code.=<<___; +.globl aesni_xts_encrypt +.type aesni_xts_encrypt,\@function,6 +.align 16 +aesni_xts_encrypt: + lea (%rsp),%rax + push %rbp + sub \$$frame_size,%rsp + and \$-16,%rsp # Linux kernel stack can be incorrectly seeded +___ +$code.=<<___ if ($win64); + movaps %xmm6,-0xa8(%rax) # offload everything + movaps %xmm7,-0x98(%rax) + movaps %xmm8,-0x88(%rax) + movaps %xmm9,-0x78(%rax) + movaps %xmm10,-0x68(%rax) + movaps %xmm11,-0x58(%rax) + movaps %xmm12,-0x48(%rax) + movaps %xmm13,-0x38(%rax) + movaps %xmm14,-0x28(%rax) + movaps %xmm15,-0x18(%rax) +.Lxts_enc_body: +___ +$code.=<<___; + lea -8(%rax),%rbp + movups ($ivp),$inout0 # load clear-text tweak + mov 240(%r8),$rounds # key2->rounds + mov 240($key),$rnds_ # key1->rounds +___ + # generate the tweak + &aesni_generate1("enc",$key2,$rounds,$inout0); +$code.=<<___; + $movkey ($key),$rndkey0 # zero round key + mov $key,$key_ # backup $key + mov $rnds_,$rounds # backup $rounds + shl \$4,$rnds_ + mov $len,$len_ # backup $len + and \$-16,$len + + $movkey 16($key,$rnds_),$rndkey1 # last round key + + movdqa .Lxts_magic(%rip),$twmask + movdqa $inout0,@tweak[5] + pshufd \$0x5f,$inout0,$twres + pxor $rndkey0,$rndkey1 +___ + # alternative tweak calculation algorithm is based on suggestions + # by Shay Gueron. psrad doesn't conflict with AES-NI instructions + # and should help in the future... + for ($i=0;$i<4;$i++) { + $code.=<<___; + movdqa $twres,$twtmp + paddd $twres,$twres + movdqa @tweak[5],@tweak[$i] + psrad \$31,$twtmp # broadcast upper bits + paddq @tweak[5],@tweak[5] + pand $twmask,$twtmp + pxor $rndkey0,@tweak[$i] + pxor $twtmp,@tweak[5] +___ + } +$code.=<<___; + movdqa @tweak[5],@tweak[4] + psrad \$31,$twres + paddq @tweak[5],@tweak[5] + pand $twmask,$twres + pxor $rndkey0,@tweak[4] + pxor $twres,@tweak[5] + movaps $rndkey1,0x60(%rsp) # save round[0]^round[last] + + sub \$16*6,$len + jc .Lxts_enc_short # if $len-=6*16 borrowed + + mov \$16+96,$rounds + lea 32($key_,$rnds_),$key # end of key schedule + sub %r10,%rax # twisted $rounds + $movkey 16($key_),$rndkey1 + mov %rax,%r10 # backup twisted $rounds + lea .Lxts_magic(%rip),%r8 + jmp .Lxts_enc_grandloop + +.align 32 +.Lxts_enc_grandloop: + movdqu `16*0`($inp),$inout0 # load input + movdqa $rndkey0,$twmask + movdqu `16*1`($inp),$inout1 + pxor @tweak[0],$inout0 # input^=tweak^round[0] + movdqu `16*2`($inp),$inout2 + pxor @tweak[1],$inout1 + aesenc $rndkey1,$inout0 + movdqu `16*3`($inp),$inout3 + pxor @tweak[2],$inout2 + aesenc $rndkey1,$inout1 + movdqu `16*4`($inp),$inout4 + pxor @tweak[3],$inout3 + aesenc $rndkey1,$inout2 + movdqu `16*5`($inp),$inout5 + pxor @tweak[5],$twmask # round[0]^=tweak[5] + movdqa 0x60(%rsp),$twres # load round[0]^round[last] + pxor @tweak[4],$inout4 + aesenc $rndkey1,$inout3 + $movkey 32($key_),$rndkey0 + lea `16*6`($inp),$inp + pxor $twmask,$inout5 + + pxor $twres,@tweak[0] # calclulate tweaks^round[last] + aesenc $rndkey1,$inout4 + pxor $twres,@tweak[1] + movdqa @tweak[0],`16*0`(%rsp) # put aside tweaks^round[last] + aesenc $rndkey1,$inout5 + $movkey 48($key_),$rndkey1 + pxor $twres,@tweak[2] + + aesenc $rndkey0,$inout0 + pxor $twres,@tweak[3] + movdqa @tweak[1],`16*1`(%rsp) + aesenc $rndkey0,$inout1 + pxor $twres,@tweak[4] + movdqa @tweak[2],`16*2`(%rsp) + aesenc $rndkey0,$inout2 + aesenc $rndkey0,$inout3 + pxor $twres,$twmask + movdqa @tweak[4],`16*4`(%rsp) + aesenc $rndkey0,$inout4 + aesenc $rndkey0,$inout5 + $movkey 64($key_),$rndkey0 + movdqa $twmask,`16*5`(%rsp) + pshufd \$0x5f,@tweak[5],$twres + jmp .Lxts_enc_loop6 +.align 32 +.Lxts_enc_loop6: + aesenc $rndkey1,$inout0 + aesenc $rndkey1,$inout1 + aesenc $rndkey1,$inout2 + aesenc $rndkey1,$inout3 + aesenc $rndkey1,$inout4 + aesenc $rndkey1,$inout5 + $movkey -64($key,%rax),$rndkey1 + add \$32,%rax + + aesenc $rndkey0,$inout0 + aesenc $rndkey0,$inout1 + aesenc $rndkey0,$inout2 + aesenc $rndkey0,$inout3 + aesenc $rndkey0,$inout4 + aesenc $rndkey0,$inout5 + $movkey -80($key,%rax),$rndkey0 + jnz .Lxts_enc_loop6 + + movdqa (%r8),$twmask # start calculating next tweak + movdqa $twres,$twtmp + paddd $twres,$twres + aesenc $rndkey1,$inout0 + paddq @tweak[5],@tweak[5] + psrad \$31,$twtmp + aesenc $rndkey1,$inout1 + pand $twmask,$twtmp + $movkey ($key_),@tweak[0] # load round[0] + aesenc $rndkey1,$inout2 + aesenc $rndkey1,$inout3 + aesenc $rndkey1,$inout4 + pxor $twtmp,@tweak[5] + movaps @tweak[0],@tweak[1] # copy round[0] + aesenc $rndkey1,$inout5 + $movkey -64($key),$rndkey1 + + movdqa $twres,$twtmp + aesenc $rndkey0,$inout0 + paddd $twres,$twres + pxor @tweak[5],@tweak[0] + aesenc $rndkey0,$inout1 + psrad \$31,$twtmp + paddq @tweak[5],@tweak[5] + aesenc $rndkey0,$inout2 + aesenc $rndkey0,$inout3 + pand $twmask,$twtmp + movaps @tweak[1],@tweak[2] + aesenc $rndkey0,$inout4 + pxor $twtmp,@tweak[5] + movdqa $twres,$twtmp + aesenc $rndkey0,$inout5 + $movkey -48($key),$rndkey0 + + paddd $twres,$twres + aesenc $rndkey1,$inout0 + pxor @tweak[5],@tweak[1] + psrad \$31,$twtmp + aesenc $rndkey1,$inout1 + paddq @tweak[5],@tweak[5] + pand $twmask,$twtmp + aesenc $rndkey1,$inout2 + aesenc $rndkey1,$inout3 + movdqa @tweak[3],`16*3`(%rsp) + pxor $twtmp,@tweak[5] + aesenc $rndkey1,$inout4 + movaps @tweak[2],@tweak[3] + movdqa $twres,$twtmp + aesenc $rndkey1,$inout5 + $movkey -32($key),$rndkey1 + + paddd $twres,$twres + aesenc $rndkey0,$inout0 + pxor @tweak[5],@tweak[2] + psrad \$31,$twtmp + aesenc $rndkey0,$inout1 + paddq @tweak[5],@tweak[5] + pand $twmask,$twtmp + aesenc $rndkey0,$inout2 + aesenc $rndkey0,$inout3 + aesenc $rndkey0,$inout4 + pxor $twtmp,@tweak[5] + movaps @tweak[3],@tweak[4] + aesenc $rndkey0,$inout5 + + movdqa $twres,$rndkey0 + paddd $twres,$twres + aesenc $rndkey1,$inout0 + pxor @tweak[5],@tweak[3] + psrad \$31,$rndkey0 + aesenc $rndkey1,$inout1 + paddq @tweak[5],@tweak[5] + pand $twmask,$rndkey0 + aesenc $rndkey1,$inout2 + aesenc $rndkey1,$inout3 + pxor $rndkey0,@tweak[5] + $movkey ($key_),$rndkey0 + aesenc $rndkey1,$inout4 + aesenc $rndkey1,$inout5 + $movkey 16($key_),$rndkey1 + + pxor @tweak[5],@tweak[4] + aesenclast `16*0`(%rsp),$inout0 + psrad \$31,$twres + paddq @tweak[5],@tweak[5] + aesenclast `16*1`(%rsp),$inout1 + aesenclast `16*2`(%rsp),$inout2 + pand $twmask,$twres + mov %r10,%rax # restore $rounds + aesenclast `16*3`(%rsp),$inout3 + aesenclast `16*4`(%rsp),$inout4 + aesenclast `16*5`(%rsp),$inout5 + pxor $twres,@tweak[5] + + lea `16*6`($out),$out # $out+=6*16 + movups $inout0,`-16*6`($out) # store 6 output blocks + movups $inout1,`-16*5`($out) + movups $inout2,`-16*4`($out) + movups $inout3,`-16*3`($out) + movups $inout4,`-16*2`($out) + movups $inout5,`-16*1`($out) + sub \$16*6,$len + jnc .Lxts_enc_grandloop # loop if $len-=6*16 didn't borrow + + mov \$16+96,$rounds + sub $rnds_,$rounds + mov $key_,$key # restore $key + shr \$4,$rounds # restore original value + +.Lxts_enc_short: + # at the point @tweak[0..5] are populated with tweak values + mov $rounds,$rnds_ # backup $rounds + pxor $rndkey0,@tweak[0] + add \$16*6,$len # restore real remaining $len + jz .Lxts_enc_done # done if ($len==0) + + pxor $rndkey0,@tweak[1] + cmp \$0x20,$len + jb .Lxts_enc_one # $len is 1*16 + pxor $rndkey0,@tweak[2] + je .Lxts_enc_two # $len is 2*16 + + pxor $rndkey0,@tweak[3] + cmp \$0x40,$len + jb .Lxts_enc_three # $len is 3*16 + pxor $rndkey0,@tweak[4] + je .Lxts_enc_four # $len is 4*16 + + movdqu ($inp),$inout0 # $len is 5*16 + movdqu 16*1($inp),$inout1 + movdqu 16*2($inp),$inout2 + pxor @tweak[0],$inout0 + movdqu 16*3($inp),$inout3 + pxor @tweak[1],$inout1 + movdqu 16*4($inp),$inout4 + lea 16*5($inp),$inp # $inp+=5*16 + pxor @tweak[2],$inout2 + pxor @tweak[3],$inout3 + pxor @tweak[4],$inout4 + pxor $inout5,$inout5 + + call _aesni_encrypt6 + + xorps @tweak[0],$inout0 + movdqa @tweak[5],@tweak[0] + xorps @tweak[1],$inout1 + xorps @tweak[2],$inout2 + movdqu $inout0,($out) # store 5 output blocks + xorps @tweak[3],$inout3 + movdqu $inout1,16*1($out) + xorps @tweak[4],$inout4 + movdqu $inout2,16*2($out) + movdqu $inout3,16*3($out) + movdqu $inout4,16*4($out) + lea 16*5($out),$out # $out+=5*16 + jmp .Lxts_enc_done + +.align 16 +.Lxts_enc_one: + movups ($inp),$inout0 + lea 16*1($inp),$inp # inp+=1*16 + xorps @tweak[0],$inout0 +___ + &aesni_generate1("enc",$key,$rounds); +$code.=<<___; + xorps @tweak[0],$inout0 + movdqa @tweak[1],@tweak[0] + movups $inout0,($out) # store one output block + lea 16*1($out),$out # $out+=1*16 + jmp .Lxts_enc_done + +.align 16 +.Lxts_enc_two: + movups ($inp),$inout0 + movups 16($inp),$inout1 + lea 32($inp),$inp # $inp+=2*16 + xorps @tweak[0],$inout0 + xorps @tweak[1],$inout1 + + call _aesni_encrypt2 + + xorps @tweak[0],$inout0 + movdqa @tweak[2],@tweak[0] + xorps @tweak[1],$inout1 + movups $inout0,($out) # store 2 output blocks + movups $inout1,16*1($out) + lea 16*2($out),$out # $out+=2*16 + jmp .Lxts_enc_done + +.align 16 +.Lxts_enc_three: + movups ($inp),$inout0 + movups 16*1($inp),$inout1 + movups 16*2($inp),$inout2 + lea 16*3($inp),$inp # $inp+=3*16 + xorps @tweak[0],$inout0 + xorps @tweak[1],$inout1 + xorps @tweak[2],$inout2 + + call _aesni_encrypt3 + + xorps @tweak[0],$inout0 + movdqa @tweak[3],@tweak[0] + xorps @tweak[1],$inout1 + xorps @tweak[2],$inout2 + movups $inout0,($out) # store 3 output blocks + movups $inout1,16*1($out) + movups $inout2,16*2($out) + lea 16*3($out),$out # $out+=3*16 + jmp .Lxts_enc_done + +.align 16 +.Lxts_enc_four: + movups ($inp),$inout0 + movups 16*1($inp),$inout1 + movups 16*2($inp),$inout2 + xorps @tweak[0],$inout0 + movups 16*3($inp),$inout3 + lea 16*4($inp),$inp # $inp+=4*16 + xorps @tweak[1],$inout1 + xorps @tweak[2],$inout2 + xorps @tweak[3],$inout3 + + call _aesni_encrypt4 + + pxor @tweak[0],$inout0 + movdqa @tweak[4],@tweak[0] + pxor @tweak[1],$inout1 + pxor @tweak[2],$inout2 + movdqu $inout0,($out) # store 4 output blocks + pxor @tweak[3],$inout3 + movdqu $inout1,16*1($out) + movdqu $inout2,16*2($out) + movdqu $inout3,16*3($out) + lea 16*4($out),$out # $out+=4*16 + jmp .Lxts_enc_done + +.align 16 +.Lxts_enc_done: + and \$15,$len_ # see if $len%16 is 0 + jz .Lxts_enc_ret + mov $len_,$len + +.Lxts_enc_steal: + movzb ($inp),%eax # borrow $rounds ... + movzb -16($out),%ecx # ... and $key + lea 1($inp),$inp + mov %al,-16($out) + mov %cl,0($out) + lea 1($out),$out + sub \$1,$len + jnz .Lxts_enc_steal + + sub $len_,$out # rewind $out + mov $key_,$key # restore $key + mov $rnds_,$rounds # restore $rounds + + movups -16($out),$inout0 + xorps @tweak[0],$inout0 +___ + &aesni_generate1("enc",$key,$rounds); +$code.=<<___; + xorps @tweak[0],$inout0 + movups $inout0,-16($out) + +.Lxts_enc_ret: + xorps %xmm0,%xmm0 # clear register bank + pxor %xmm1,%xmm1 + pxor %xmm2,%xmm2 + pxor %xmm3,%xmm3 + pxor %xmm4,%xmm4 + pxor %xmm5,%xmm5 +___ +$code.=<<___ if (!$win64); + pxor %xmm6,%xmm6 + pxor %xmm7,%xmm7 + movaps %xmm0,0x00(%rsp) # clear stack + pxor %xmm8,%xmm8 + movaps %xmm0,0x10(%rsp) + pxor %xmm9,%xmm9 + movaps %xmm0,0x20(%rsp) + pxor %xmm10,%xmm10 + movaps %xmm0,0x30(%rsp) + pxor %xmm11,%xmm11 + movaps %xmm0,0x40(%rsp) + pxor %xmm12,%xmm12 + movaps %xmm0,0x50(%rsp) + pxor %xmm13,%xmm13 + movaps %xmm0,0x60(%rsp) + pxor %xmm14,%xmm14 + pxor %xmm15,%xmm15 +___ +$code.=<<___ if ($win64); + movaps -0xa0(%rbp),%xmm6 + movaps %xmm0,-0xa0(%rbp) # clear stack + movaps -0x90(%rbp),%xmm7 + movaps %xmm0,-0x90(%rbp) + movaps -0x80(%rbp),%xmm8 + movaps %xmm0,-0x80(%rbp) + movaps -0x70(%rbp),%xmm9 + movaps %xmm0,-0x70(%rbp) + movaps -0x60(%rbp),%xmm10 + movaps %xmm0,-0x60(%rbp) + movaps -0x50(%rbp),%xmm11 + movaps %xmm0,-0x50(%rbp) + movaps -0x40(%rbp),%xmm12 + movaps %xmm0,-0x40(%rbp) + movaps -0x30(%rbp),%xmm13 + movaps %xmm0,-0x30(%rbp) + movaps -0x20(%rbp),%xmm14 + movaps %xmm0,-0x20(%rbp) + movaps -0x10(%rbp),%xmm15 + movaps %xmm0,-0x10(%rbp) + movaps %xmm0,0x00(%rsp) + movaps %xmm0,0x10(%rsp) + movaps %xmm0,0x20(%rsp) + movaps %xmm0,0x30(%rsp) + movaps %xmm0,0x40(%rsp) + movaps %xmm0,0x50(%rsp) + movaps %xmm0,0x60(%rsp) +___ +$code.=<<___; + lea (%rbp),%rsp + pop %rbp +.Lxts_enc_epilogue: + ret +.size aesni_xts_encrypt,.-aesni_xts_encrypt +___ + +$code.=<<___; +.globl aesni_xts_decrypt +.type aesni_xts_decrypt,\@function,6 +.align 16 +aesni_xts_decrypt: + lea (%rsp),%rax + push %rbp + sub \$$frame_size,%rsp + and \$-16,%rsp # Linux kernel stack can be incorrectly seeded +___ +$code.=<<___ if ($win64); + movaps %xmm6,-0xa8(%rax) # offload everything + movaps %xmm7,-0x98(%rax) + movaps %xmm8,-0x88(%rax) + movaps %xmm9,-0x78(%rax) + movaps %xmm10,-0x68(%rax) + movaps %xmm11,-0x58(%rax) + movaps %xmm12,-0x48(%rax) + movaps %xmm13,-0x38(%rax) + movaps %xmm14,-0x28(%rax) + movaps %xmm15,-0x18(%rax) +.Lxts_dec_body: +___ +$code.=<<___; + lea -8(%rax),%rbp + movups ($ivp),$inout0 # load clear-text tweak + mov 240($key2),$rounds # key2->rounds + mov 240($key),$rnds_ # key1->rounds +___ + # generate the tweak + &aesni_generate1("enc",$key2,$rounds,$inout0); +$code.=<<___; + xor %eax,%eax # if ($len%16) len-=16; + test \$15,$len + setnz %al + shl \$4,%rax + sub %rax,$len + + $movkey ($key),$rndkey0 # zero round key + mov $key,$key_ # backup $key + mov $rnds_,$rounds # backup $rounds + shl \$4,$rnds_ + mov $len,$len_ # backup $len + and \$-16,$len + + $movkey 16($key,$rnds_),$rndkey1 # last round key + + movdqa .Lxts_magic(%rip),$twmask + movdqa $inout0,@tweak[5] + pshufd \$0x5f,$inout0,$twres + pxor $rndkey0,$rndkey1 +___ + for ($i=0;$i<4;$i++) { + $code.=<<___; + movdqa $twres,$twtmp + paddd $twres,$twres + movdqa @tweak[5],@tweak[$i] + psrad \$31,$twtmp # broadcast upper bits + paddq @tweak[5],@tweak[5] + pand $twmask,$twtmp + pxor $rndkey0,@tweak[$i] + pxor $twtmp,@tweak[5] +___ + } +$code.=<<___; + movdqa @tweak[5],@tweak[4] + psrad \$31,$twres + paddq @tweak[5],@tweak[5] + pand $twmask,$twres + pxor $rndkey0,@tweak[4] + pxor $twres,@tweak[5] + movaps $rndkey1,0x60(%rsp) # save round[0]^round[last] + + sub \$16*6,$len + jc .Lxts_dec_short # if $len-=6*16 borrowed + + mov \$16+96,$rounds + lea 32($key_,$rnds_),$key # end of key schedule + sub %r10,%rax # twisted $rounds + $movkey 16($key_),$rndkey1 + mov %rax,%r10 # backup twisted $rounds + lea .Lxts_magic(%rip),%r8 + jmp .Lxts_dec_grandloop + +.align 32 +.Lxts_dec_grandloop: + movdqu `16*0`($inp),$inout0 # load input + movdqa $rndkey0,$twmask + movdqu `16*1`($inp),$inout1 + pxor @tweak[0],$inout0 # intput^=tweak^round[0] + movdqu `16*2`($inp),$inout2 + pxor @tweak[1],$inout1 + aesdec $rndkey1,$inout0 + movdqu `16*3`($inp),$inout3 + pxor @tweak[2],$inout2 + aesdec $rndkey1,$inout1 + movdqu `16*4`($inp),$inout4 + pxor @tweak[3],$inout3 + aesdec $rndkey1,$inout2 + movdqu `16*5`($inp),$inout5 + pxor @tweak[5],$twmask # round[0]^=tweak[5] + movdqa 0x60(%rsp),$twres # load round[0]^round[last] + pxor @tweak[4],$inout4 + aesdec $rndkey1,$inout3 + $movkey 32($key_),$rndkey0 + lea `16*6`($inp),$inp + pxor $twmask,$inout5 + + pxor $twres,@tweak[0] # calclulate tweaks^round[last] + aesdec $rndkey1,$inout4 + pxor $twres,@tweak[1] + movdqa @tweak[0],`16*0`(%rsp) # put aside tweaks^last round key + aesdec $rndkey1,$inout5 + $movkey 48($key_),$rndkey1 + pxor $twres,@tweak[2] + + aesdec $rndkey0,$inout0 + pxor $twres,@tweak[3] + movdqa @tweak[1],`16*1`(%rsp) + aesdec $rndkey0,$inout1 + pxor $twres,@tweak[4] + movdqa @tweak[2],`16*2`(%rsp) + aesdec $rndkey0,$inout2 + aesdec $rndkey0,$inout3 + pxor $twres,$twmask + movdqa @tweak[4],`16*4`(%rsp) + aesdec $rndkey0,$inout4 + aesdec $rndkey0,$inout5 + $movkey 64($key_),$rndkey0 + movdqa $twmask,`16*5`(%rsp) + pshufd \$0x5f,@tweak[5],$twres + jmp .Lxts_dec_loop6 +.align 32 +.Lxts_dec_loop6: + aesdec $rndkey1,$inout0 + aesdec $rndkey1,$inout1 + aesdec $rndkey1,$inout2 + aesdec $rndkey1,$inout3 + aesdec $rndkey1,$inout4 + aesdec $rndkey1,$inout5 + $movkey -64($key,%rax),$rndkey1 + add \$32,%rax + + aesdec $rndkey0,$inout0 + aesdec $rndkey0,$inout1 + aesdec $rndkey0,$inout2 + aesdec $rndkey0,$inout3 + aesdec $rndkey0,$inout4 + aesdec $rndkey0,$inout5 + $movkey -80($key,%rax),$rndkey0 + jnz .Lxts_dec_loop6 + + movdqa (%r8),$twmask # start calculating next tweak + movdqa $twres,$twtmp + paddd $twres,$twres + aesdec $rndkey1,$inout0 + paddq @tweak[5],@tweak[5] + psrad \$31,$twtmp + aesdec $rndkey1,$inout1 + pand $twmask,$twtmp + $movkey ($key_),@tweak[0] # load round[0] + aesdec $rndkey1,$inout2 + aesdec $rndkey1,$inout3 + aesdec $rndkey1,$inout4 + pxor $twtmp,@tweak[5] + movaps @tweak[0],@tweak[1] # copy round[0] + aesdec $rndkey1,$inout5 + $movkey -64($key),$rndkey1 + + movdqa $twres,$twtmp + aesdec $rndkey0,$inout0 + paddd $twres,$twres + pxor @tweak[5],@tweak[0] + aesdec $rndkey0,$inout1 + psrad \$31,$twtmp + paddq @tweak[5],@tweak[5] + aesdec $rndkey0,$inout2 + aesdec $rndkey0,$inout3 + pand $twmask,$twtmp + movaps @tweak[1],@tweak[2] + aesdec $rndkey0,$inout4 + pxor $twtmp,@tweak[5] + movdqa $twres,$twtmp + aesdec $rndkey0,$inout5 + $movkey -48($key),$rndkey0 + + paddd $twres,$twres + aesdec $rndkey1,$inout0 + pxor @tweak[5],@tweak[1] + psrad \$31,$twtmp + aesdec $rndkey1,$inout1 + paddq @tweak[5],@tweak[5] + pand $twmask,$twtmp + aesdec $rndkey1,$inout2 + aesdec $rndkey1,$inout3 + movdqa @tweak[3],`16*3`(%rsp) + pxor $twtmp,@tweak[5] + aesdec $rndkey1,$inout4 + movaps @tweak[2],@tweak[3] + movdqa $twres,$twtmp + aesdec $rndkey1,$inout5 + $movkey -32($key),$rndkey1 + + paddd $twres,$twres + aesdec $rndkey0,$inout0 + pxor @tweak[5],@tweak[2] + psrad \$31,$twtmp + aesdec $rndkey0,$inout1 + paddq @tweak[5],@tweak[5] + pand $twmask,$twtmp + aesdec $rndkey0,$inout2 + aesdec $rndkey0,$inout3 + aesdec $rndkey0,$inout4 + pxor $twtmp,@tweak[5] + movaps @tweak[3],@tweak[4] + aesdec $rndkey0,$inout5 + + movdqa $twres,$rndkey0 + paddd $twres,$twres + aesdec $rndkey1,$inout0 + pxor @tweak[5],@tweak[3] + psrad \$31,$rndkey0 + aesdec $rndkey1,$inout1 + paddq @tweak[5],@tweak[5] + pand $twmask,$rndkey0 + aesdec $rndkey1,$inout2 + aesdec $rndkey1,$inout3 + pxor $rndkey0,@tweak[5] + $movkey ($key_),$rndkey0 + aesdec $rndkey1,$inout4 + aesdec $rndkey1,$inout5 + $movkey 16($key_),$rndkey1 + + pxor @tweak[5],@tweak[4] + aesdeclast `16*0`(%rsp),$inout0 + psrad \$31,$twres + paddq @tweak[5],@tweak[5] + aesdeclast `16*1`(%rsp),$inout1 + aesdeclast `16*2`(%rsp),$inout2 + pand $twmask,$twres + mov %r10,%rax # restore $rounds + aesdeclast `16*3`(%rsp),$inout3 + aesdeclast `16*4`(%rsp),$inout4 + aesdeclast `16*5`(%rsp),$inout5 + pxor $twres,@tweak[5] + + lea `16*6`($out),$out # $out+=6*16 + movups $inout0,`-16*6`($out) # store 6 output blocks + movups $inout1,`-16*5`($out) + movups $inout2,`-16*4`($out) + movups $inout3,`-16*3`($out) + movups $inout4,`-16*2`($out) + movups $inout5,`-16*1`($out) + sub \$16*6,$len + jnc .Lxts_dec_grandloop # loop if $len-=6*16 didn't borrow + + mov \$16+96,$rounds + sub $rnds_,$rounds + mov $key_,$key # restore $key + shr \$4,$rounds # restore original value + +.Lxts_dec_short: + # at the point @tweak[0..5] are populated with tweak values + mov $rounds,$rnds_ # backup $rounds + pxor $rndkey0,@tweak[0] + pxor $rndkey0,@tweak[1] + add \$16*6,$len # restore real remaining $len + jz .Lxts_dec_done # done if ($len==0) + + pxor $rndkey0,@tweak[2] + cmp \$0x20,$len + jb .Lxts_dec_one # $len is 1*16 + pxor $rndkey0,@tweak[3] + je .Lxts_dec_two # $len is 2*16 + + pxor $rndkey0,@tweak[4] + cmp \$0x40,$len + jb .Lxts_dec_three # $len is 3*16 + je .Lxts_dec_four # $len is 4*16 + + movdqu ($inp),$inout0 # $len is 5*16 + movdqu 16*1($inp),$inout1 + movdqu 16*2($inp),$inout2 + pxor @tweak[0],$inout0 + movdqu 16*3($inp),$inout3 + pxor @tweak[1],$inout1 + movdqu 16*4($inp),$inout4 + lea 16*5($inp),$inp # $inp+=5*16 + pxor @tweak[2],$inout2 + pxor @tweak[3],$inout3 + pxor @tweak[4],$inout4 + + call _aesni_decrypt6 + + xorps @tweak[0],$inout0 + xorps @tweak[1],$inout1 + xorps @tweak[2],$inout2 + movdqu $inout0,($out) # store 5 output blocks + xorps @tweak[3],$inout3 + movdqu $inout1,16*1($out) + xorps @tweak[4],$inout4 + movdqu $inout2,16*2($out) + pxor $twtmp,$twtmp + movdqu $inout3,16*3($out) + pcmpgtd @tweak[5],$twtmp + movdqu $inout4,16*4($out) + lea 16*5($out),$out # $out+=5*16 + pshufd \$0x13,$twtmp,@tweak[1] # $twres + and \$15,$len_ + jz .Lxts_dec_ret + + movdqa @tweak[5],@tweak[0] + paddq @tweak[5],@tweak[5] # psllq 1,$tweak + pand $twmask,@tweak[1] # isolate carry and residue + pxor @tweak[5],@tweak[1] + jmp .Lxts_dec_done2 + +.align 16 +.Lxts_dec_one: + movups ($inp),$inout0 + lea 16*1($inp),$inp # $inp+=1*16 + xorps @tweak[0],$inout0 +___ + &aesni_generate1("dec",$key,$rounds); +$code.=<<___; + xorps @tweak[0],$inout0 + movdqa @tweak[1],@tweak[0] + movups $inout0,($out) # store one output block + movdqa @tweak[2],@tweak[1] + lea 16*1($out),$out # $out+=1*16 + jmp .Lxts_dec_done + +.align 16 +.Lxts_dec_two: + movups ($inp),$inout0 + movups 16($inp),$inout1 + lea 32($inp),$inp # $inp+=2*16 + xorps @tweak[0],$inout0 + xorps @tweak[1],$inout1 + + call _aesni_decrypt2 + + xorps @tweak[0],$inout0 + movdqa @tweak[2],@tweak[0] + xorps @tweak[1],$inout1 + movdqa @tweak[3],@tweak[1] + movups $inout0,($out) # store 2 output blocks + movups $inout1,16*1($out) + lea 16*2($out),$out # $out+=2*16 + jmp .Lxts_dec_done + +.align 16 +.Lxts_dec_three: + movups ($inp),$inout0 + movups 16*1($inp),$inout1 + movups 16*2($inp),$inout2 + lea 16*3($inp),$inp # $inp+=3*16 + xorps @tweak[0],$inout0 + xorps @tweak[1],$inout1 + xorps @tweak[2],$inout2 + + call _aesni_decrypt3 + + xorps @tweak[0],$inout0 + movdqa @tweak[3],@tweak[0] + xorps @tweak[1],$inout1 + movdqa @tweak[4],@tweak[1] + xorps @tweak[2],$inout2 + movups $inout0,($out) # store 3 output blocks + movups $inout1,16*1($out) + movups $inout2,16*2($out) + lea 16*3($out),$out # $out+=3*16 + jmp .Lxts_dec_done + +.align 16 +.Lxts_dec_four: + movups ($inp),$inout0 + movups 16*1($inp),$inout1 + movups 16*2($inp),$inout2 + xorps @tweak[0],$inout0 + movups 16*3($inp),$inout3 + lea 16*4($inp),$inp # $inp+=4*16 + xorps @tweak[1],$inout1 + xorps @tweak[2],$inout2 + xorps @tweak[3],$inout3 + + call _aesni_decrypt4 + + pxor @tweak[0],$inout0 + movdqa @tweak[4],@tweak[0] + pxor @tweak[1],$inout1 + movdqa @tweak[5],@tweak[1] + pxor @tweak[2],$inout2 + movdqu $inout0,($out) # store 4 output blocks + pxor @tweak[3],$inout3 + movdqu $inout1,16*1($out) + movdqu $inout2,16*2($out) + movdqu $inout3,16*3($out) + lea 16*4($out),$out # $out+=4*16 + jmp .Lxts_dec_done + +.align 16 +.Lxts_dec_done: + and \$15,$len_ # see if $len%16 is 0 + jz .Lxts_dec_ret +.Lxts_dec_done2: + mov $len_,$len + mov $key_,$key # restore $key + mov $rnds_,$rounds # restore $rounds + + movups ($inp),$inout0 + xorps @tweak[1],$inout0 +___ + &aesni_generate1("dec",$key,$rounds); +$code.=<<___; + xorps @tweak[1],$inout0 + movups $inout0,($out) + +.Lxts_dec_steal: + movzb 16($inp),%eax # borrow $rounds ... + movzb ($out),%ecx # ... and $key + lea 1($inp),$inp + mov %al,($out) + mov %cl,16($out) + lea 1($out),$out + sub \$1,$len + jnz .Lxts_dec_steal + + sub $len_,$out # rewind $out + mov $key_,$key # restore $key + mov $rnds_,$rounds # restore $rounds + + movups ($out),$inout0 + xorps @tweak[0],$inout0 +___ + &aesni_generate1("dec",$key,$rounds); +$code.=<<___; + xorps @tweak[0],$inout0 + movups $inout0,($out) + +.Lxts_dec_ret: + xorps %xmm0,%xmm0 # clear register bank + pxor %xmm1,%xmm1 + pxor %xmm2,%xmm2 + pxor %xmm3,%xmm3 + pxor %xmm4,%xmm4 + pxor %xmm5,%xmm5 +___ +$code.=<<___ if (!$win64); + pxor %xmm6,%xmm6 + pxor %xmm7,%xmm7 + movaps %xmm0,0x00(%rsp) # clear stack + pxor %xmm8,%xmm8 + movaps %xmm0,0x10(%rsp) + pxor %xmm9,%xmm9 + movaps %xmm0,0x20(%rsp) + pxor %xmm10,%xmm10 + movaps %xmm0,0x30(%rsp) + pxor %xmm11,%xmm11 + movaps %xmm0,0x40(%rsp) + pxor %xmm12,%xmm12 + movaps %xmm0,0x50(%rsp) + pxor %xmm13,%xmm13 + movaps %xmm0,0x60(%rsp) + pxor %xmm14,%xmm14 + pxor %xmm15,%xmm15 +___ +$code.=<<___ if ($win64); + movaps -0xa0(%rbp),%xmm6 + movaps %xmm0,-0xa0(%rbp) # clear stack + movaps -0x90(%rbp),%xmm7 + movaps %xmm0,-0x90(%rbp) + movaps -0x80(%rbp),%xmm8 + movaps %xmm0,-0x80(%rbp) + movaps -0x70(%rbp),%xmm9 + movaps %xmm0,-0x70(%rbp) + movaps -0x60(%rbp),%xmm10 + movaps %xmm0,-0x60(%rbp) + movaps -0x50(%rbp),%xmm11 + movaps %xmm0,-0x50(%rbp) + movaps -0x40(%rbp),%xmm12 + movaps %xmm0,-0x40(%rbp) + movaps -0x30(%rbp),%xmm13 + movaps %xmm0,-0x30(%rbp) + movaps -0x20(%rbp),%xmm14 + movaps %xmm0,-0x20(%rbp) + movaps -0x10(%rbp),%xmm15 + movaps %xmm0,-0x10(%rbp) + movaps %xmm0,0x00(%rsp) + movaps %xmm0,0x10(%rsp) + movaps %xmm0,0x20(%rsp) + movaps %xmm0,0x30(%rsp) + movaps %xmm0,0x40(%rsp) + movaps %xmm0,0x50(%rsp) + movaps %xmm0,0x60(%rsp) +___ +$code.=<<___; + lea (%rbp),%rsp + pop %rbp +.Lxts_dec_epilogue: + ret +.size aesni_xts_decrypt,.-aesni_xts_decrypt +___ +} + +###################################################################### +# void aesni_ocb_[en|de]crypt(const char *inp, char *out, size_t blocks, +# const AES_KEY *key, unsigned int start_block_num, +# unsigned char offset_i[16], const unsigned char L_[][16], +# unsigned char checksum[16]); +# +{ +my @offset=map("%xmm$_",(10..15)); +my ($checksum,$rndkey0l)=("%xmm8","%xmm9"); +my ($block_num,$offset_p)=("%r8","%r9"); # 5th and 6th arguments +my ($L_p,$checksum_p) = ("%rbx","%rbp"); +my ($i1,$i3,$i5) = ("%r12","%r13","%r14"); +my $seventh_arg = $win64 ? 56 : 8; +my $blocks = $len; + +$code.=<<___; +.globl aesni_ocb_encrypt +.type aesni_ocb_encrypt,\@function,6 +.align 32 +aesni_ocb_encrypt: + lea (%rsp),%rax + push %rbx + push %rbp + push %r12 + push %r13 + push %r14 +___ +$code.=<<___ if ($win64); + lea -0xa0(%rsp),%rsp + movaps %xmm6,0x00(%rsp) # offload everything + movaps %xmm7,0x10(%rsp) + movaps %xmm8,0x20(%rsp) + movaps %xmm9,0x30(%rsp) + movaps %xmm10,0x40(%rsp) + movaps %xmm11,0x50(%rsp) + movaps %xmm12,0x60(%rsp) + movaps %xmm13,0x70(%rsp) + movaps %xmm14,0x80(%rsp) + movaps %xmm15,0x90(%rsp) +.Locb_enc_body: +___ +$code.=<<___; + mov $seventh_arg(%rax),$L_p # 7th argument + mov $seventh_arg+8(%rax),$checksum_p# 8th argument + + mov 240($key),$rnds_ + mov $key,$key_ + shl \$4,$rnds_ + $movkey ($key),$rndkey0l # round[0] + $movkey 16($key,$rnds_),$rndkey1 # round[last] + + movdqu ($offset_p),@offset[5] # load last offset_i + pxor $rndkey1,$rndkey0l # round[0] ^ round[last] + pxor $rndkey1,@offset[5] # offset_i ^ round[last] + + mov \$16+32,$rounds + lea 32($key_,$rnds_),$key + $movkey 16($key_),$rndkey1 # round[1] + sub %r10,%rax # twisted $rounds + mov %rax,%r10 # backup twisted $rounds + + movdqu ($L_p),@offset[0] # L_0 for all odd-numbered blocks + movdqu ($checksum_p),$checksum # load checksum + + test \$1,$block_num # is first block number odd? + jnz .Locb_enc_odd + + bsf $block_num,$i1 + add \$1,$block_num + shl \$4,$i1 + movdqu ($L_p,$i1),$inout5 # borrow + movdqu ($inp),$inout0 + lea 16($inp),$inp + + call __ocb_encrypt1 + + movdqa $inout5,@offset[5] + movups $inout0,($out) + lea 16($out),$out + sub \$1,$blocks + jz .Locb_enc_done + +.Locb_enc_odd: + lea 1($block_num),$i1 # even-numbered blocks + lea 3($block_num),$i3 + lea 5($block_num),$i5 + lea 6($block_num),$block_num + bsf $i1,$i1 # ntz(block) + bsf $i3,$i3 + bsf $i5,$i5 + shl \$4,$i1 # ntz(block) -> table offset + shl \$4,$i3 + shl \$4,$i5 + + sub \$6,$blocks + jc .Locb_enc_short + jmp .Locb_enc_grandloop + +.align 32 +.Locb_enc_grandloop: + movdqu `16*0`($inp),$inout0 # load input + movdqu `16*1`($inp),$inout1 + movdqu `16*2`($inp),$inout2 + movdqu `16*3`($inp),$inout3 + movdqu `16*4`($inp),$inout4 + movdqu `16*5`($inp),$inout5 + lea `16*6`($inp),$inp + + call __ocb_encrypt6 + + movups $inout0,`16*0`($out) # store output + movups $inout1,`16*1`($out) + movups $inout2,`16*2`($out) + movups $inout3,`16*3`($out) + movups $inout4,`16*4`($out) + movups $inout5,`16*5`($out) + lea `16*6`($out),$out + sub \$6,$blocks + jnc .Locb_enc_grandloop + +.Locb_enc_short: + add \$6,$blocks + jz .Locb_enc_done + + movdqu `16*0`($inp),$inout0 + cmp \$2,$blocks + jb .Locb_enc_one + movdqu `16*1`($inp),$inout1 + je .Locb_enc_two + + movdqu `16*2`($inp),$inout2 + cmp \$4,$blocks + jb .Locb_enc_three + movdqu `16*3`($inp),$inout3 + je .Locb_enc_four + + movdqu `16*4`($inp),$inout4 + pxor $inout5,$inout5 + + call __ocb_encrypt6 + + movdqa @offset[4],@offset[5] + movups $inout0,`16*0`($out) + movups $inout1,`16*1`($out) + movups $inout2,`16*2`($out) + movups $inout3,`16*3`($out) + movups $inout4,`16*4`($out) + + jmp .Locb_enc_done + +.align 16 +.Locb_enc_one: + movdqa @offset[0],$inout5 # borrow + + call __ocb_encrypt1 + + movdqa $inout5,@offset[5] + movups $inout0,`16*0`($out) + jmp .Locb_enc_done + +.align 16 +.Locb_enc_two: + pxor $inout2,$inout2 + pxor $inout3,$inout3 + + call __ocb_encrypt4 + + movdqa @offset[1],@offset[5] + movups $inout0,`16*0`($out) + movups $inout1,`16*1`($out) + + jmp .Locb_enc_done + +.align 16 +.Locb_enc_three: + pxor $inout3,$inout3 + + call __ocb_encrypt4 + + movdqa @offset[2],@offset[5] + movups $inout0,`16*0`($out) + movups $inout1,`16*1`($out) + movups $inout2,`16*2`($out) + + jmp .Locb_enc_done + +.align 16 +.Locb_enc_four: + call __ocb_encrypt4 + + movdqa @offset[3],@offset[5] + movups $inout0,`16*0`($out) + movups $inout1,`16*1`($out) + movups $inout2,`16*2`($out) + movups $inout3,`16*3`($out) + +.Locb_enc_done: + pxor $rndkey0,@offset[5] # "remove" round[last] + movdqu $checksum,($checksum_p) # store checksum + movdqu @offset[5],($offset_p) # store last offset_i + + xorps %xmm0,%xmm0 # clear register bank + pxor %xmm1,%xmm1 + pxor %xmm2,%xmm2 + pxor %xmm3,%xmm3 + pxor %xmm4,%xmm4 + pxor %xmm5,%xmm5 +___ +$code.=<<___ if (!$win64); + pxor %xmm6,%xmm6 + pxor %xmm7,%xmm7 + pxor %xmm8,%xmm8 + pxor %xmm9,%xmm9 + pxor %xmm10,%xmm10 + pxor %xmm11,%xmm11 + pxor %xmm12,%xmm12 + pxor %xmm13,%xmm13 + pxor %xmm14,%xmm14 + pxor %xmm15,%xmm15 +___ +$code.=<<___ if ($win64); + movaps 0x00(%rsp),%xmm6 + movaps %xmm0,0x00(%rsp) # clear stack + movaps 0x10(%rsp),%xmm7 + movaps %xmm0,0x10(%rsp) + movaps 0x20(%rsp),%xmm8 + movaps %xmm0,0x20(%rsp) + movaps 0x30(%rsp),%xmm9 + movaps %xmm0,0x30(%rsp) + movaps 0x40(%rsp),%xmm10 + movaps %xmm0,0x40(%rsp) + movaps 0x50(%rsp),%xmm11 + movaps %xmm0,0x50(%rsp) + movaps 0x60(%rsp),%xmm12 + movaps %xmm0,0x60(%rsp) + movaps 0x70(%rsp),%xmm13 + movaps %xmm0,0x70(%rsp) + movaps 0x80(%rsp),%xmm14 + movaps %xmm0,0x80(%rsp) + movaps 0x90(%rsp),%xmm15 + movaps %xmm0,0x90(%rsp) + lea 0xa0+0x28(%rsp),%rax +.Locb_enc_pop: + lea 0xa0(%rsp),%rsp +___ +$code.=<<___; + pop %r14 + pop %r13 + pop %r12 + pop %rbp + pop %rbx +.Locb_enc_epilogue: + ret +.size aesni_ocb_encrypt,.-aesni_ocb_encrypt + +.type __ocb_encrypt6,\@abi-omnipotent +.align 32 +__ocb_encrypt6: + pxor $rndkey0l,@offset[5] # offset_i ^ round[0] + movdqu ($L_p,$i1),@offset[1] + movdqa @offset[0],@offset[2] + movdqu ($L_p,$i3),@offset[3] + movdqa @offset[0],@offset[4] + pxor @offset[5],@offset[0] + movdqu ($L_p,$i5),@offset[5] + pxor @offset[0],@offset[1] + pxor $inout0,$checksum # accumulate checksum + pxor @offset[0],$inout0 # input ^ round[0] ^ offset_i + pxor @offset[1],@offset[2] + pxor $inout1,$checksum + pxor @offset[1],$inout1 + pxor @offset[2],@offset[3] + pxor $inout2,$checksum + pxor @offset[2],$inout2 + pxor @offset[3],@offset[4] + pxor $inout3,$checksum + pxor @offset[3],$inout3 + pxor @offset[4],@offset[5] + pxor $inout4,$checksum + pxor @offset[4],$inout4 + pxor $inout5,$checksum + pxor @offset[5],$inout5 + $movkey 32($key_),$rndkey0 + + lea 1($block_num),$i1 # even-numbered blocks + lea 3($block_num),$i3 + lea 5($block_num),$i5 + add \$6,$block_num + pxor $rndkey0l,@offset[0] # offset_i ^ round[last] + bsf $i1,$i1 # ntz(block) + bsf $i3,$i3 + bsf $i5,$i5 + + aesenc $rndkey1,$inout0 + aesenc $rndkey1,$inout1 + aesenc $rndkey1,$inout2 + aesenc $rndkey1,$inout3 + pxor $rndkey0l,@offset[1] + pxor $rndkey0l,@offset[2] + aesenc $rndkey1,$inout4 + pxor $rndkey0l,@offset[3] + pxor $rndkey0l,@offset[4] + aesenc $rndkey1,$inout5 + $movkey 48($key_),$rndkey1 + pxor $rndkey0l,@offset[5] + + aesenc $rndkey0,$inout0 + aesenc $rndkey0,$inout1 + aesenc $rndkey0,$inout2 + aesenc $rndkey0,$inout3 + aesenc $rndkey0,$inout4 + aesenc $rndkey0,$inout5 + $movkey 64($key_),$rndkey0 + shl \$4,$i1 # ntz(block) -> table offset + shl \$4,$i3 + jmp .Locb_enc_loop6 + +.align 32 +.Locb_enc_loop6: + aesenc $rndkey1,$inout0 + aesenc $rndkey1,$inout1 + aesenc $rndkey1,$inout2 + aesenc $rndkey1,$inout3 + aesenc $rndkey1,$inout4 + aesenc $rndkey1,$inout5 + $movkey ($key,%rax),$rndkey1 + add \$32,%rax + + aesenc $rndkey0,$inout0 + aesenc $rndkey0,$inout1 + aesenc $rndkey0,$inout2 + aesenc $rndkey0,$inout3 + aesenc $rndkey0,$inout4 + aesenc $rndkey0,$inout5 + $movkey -16($key,%rax),$rndkey0 + jnz .Locb_enc_loop6 + + aesenc $rndkey1,$inout0 + aesenc $rndkey1,$inout1 + aesenc $rndkey1,$inout2 + aesenc $rndkey1,$inout3 + aesenc $rndkey1,$inout4 + aesenc $rndkey1,$inout5 + $movkey 16($key_),$rndkey1 + shl \$4,$i5 + + aesenclast @offset[0],$inout0 + movdqu ($L_p),@offset[0] # L_0 for all odd-numbered blocks + mov %r10,%rax # restore twisted rounds + aesenclast @offset[1],$inout1 + aesenclast @offset[2],$inout2 + aesenclast @offset[3],$inout3 + aesenclast @offset[4],$inout4 + aesenclast @offset[5],$inout5 + ret +.size __ocb_encrypt6,.-__ocb_encrypt6 + +.type __ocb_encrypt4,\@abi-omnipotent +.align 32 +__ocb_encrypt4: + pxor $rndkey0l,@offset[5] # offset_i ^ round[0] + movdqu ($L_p,$i1),@offset[1] + movdqa @offset[0],@offset[2] + movdqu ($L_p,$i3),@offset[3] + pxor @offset[5],@offset[0] + pxor @offset[0],@offset[1] + pxor $inout0,$checksum # accumulate checksum + pxor @offset[0],$inout0 # input ^ round[0] ^ offset_i + pxor @offset[1],@offset[2] + pxor $inout1,$checksum + pxor @offset[1],$inout1 + pxor @offset[2],@offset[3] + pxor $inout2,$checksum + pxor @offset[2],$inout2 + pxor $inout3,$checksum + pxor @offset[3],$inout3 + $movkey 32($key_),$rndkey0 + + pxor $rndkey0l,@offset[0] # offset_i ^ round[last] + pxor $rndkey0l,@offset[1] + pxor $rndkey0l,@offset[2] + pxor $rndkey0l,@offset[3] + + aesenc $rndkey1,$inout0 + aesenc $rndkey1,$inout1 + aesenc $rndkey1,$inout2 + aesenc $rndkey1,$inout3 + $movkey 48($key_),$rndkey1 + + aesenc $rndkey0,$inout0 + aesenc $rndkey0,$inout1 + aesenc $rndkey0,$inout2 + aesenc $rndkey0,$inout3 + $movkey 64($key_),$rndkey0 + jmp .Locb_enc_loop4 + +.align 32 +.Locb_enc_loop4: + aesenc $rndkey1,$inout0 + aesenc $rndkey1,$inout1 + aesenc $rndkey1,$inout2 + aesenc $rndkey1,$inout3 + $movkey ($key,%rax),$rndkey1 + add \$32,%rax + + aesenc $rndkey0,$inout0 + aesenc $rndkey0,$inout1 + aesenc $rndkey0,$inout2 + aesenc $rndkey0,$inout3 + $movkey -16($key,%rax),$rndkey0 + jnz .Locb_enc_loop4 + + aesenc $rndkey1,$inout0 + aesenc $rndkey1,$inout1 + aesenc $rndkey1,$inout2 + aesenc $rndkey1,$inout3 + $movkey 16($key_),$rndkey1 + mov %r10,%rax # restore twisted rounds + + aesenclast @offset[0],$inout0 + aesenclast @offset[1],$inout1 + aesenclast @offset[2],$inout2 + aesenclast @offset[3],$inout3 + ret +.size __ocb_encrypt4,.-__ocb_encrypt4 + +.type __ocb_encrypt1,\@abi-omnipotent +.align 32 +__ocb_encrypt1: + pxor @offset[5],$inout5 # offset_i + pxor $rndkey0l,$inout5 # offset_i ^ round[0] + pxor $inout0,$checksum # accumulate checksum + pxor $inout5,$inout0 # input ^ round[0] ^ offset_i + $movkey 32($key_),$rndkey0 + + aesenc $rndkey1,$inout0 + $movkey 48($key_),$rndkey1 + pxor $rndkey0l,$inout5 # offset_i ^ round[last] + + aesenc $rndkey0,$inout0 + $movkey 64($key_),$rndkey0 + jmp .Locb_enc_loop1 + +.align 32 +.Locb_enc_loop1: + aesenc $rndkey1,$inout0 + $movkey ($key,%rax),$rndkey1 + add \$32,%rax + + aesenc $rndkey0,$inout0 + $movkey -16($key,%rax),$rndkey0 + jnz .Locb_enc_loop1 + + aesenc $rndkey1,$inout0 + $movkey 16($key_),$rndkey1 # redundant in tail + mov %r10,%rax # restore twisted rounds + + aesenclast $inout5,$inout0 + ret +.size __ocb_encrypt1,.-__ocb_encrypt1 + +.globl aesni_ocb_decrypt +.type aesni_ocb_decrypt,\@function,6 +.align 32 +aesni_ocb_decrypt: + lea (%rsp),%rax + push %rbx + push %rbp + push %r12 + push %r13 + push %r14 +___ +$code.=<<___ if ($win64); + lea -0xa0(%rsp),%rsp + movaps %xmm6,0x00(%rsp) # offload everything + movaps %xmm7,0x10(%rsp) + movaps %xmm8,0x20(%rsp) + movaps %xmm9,0x30(%rsp) + movaps %xmm10,0x40(%rsp) + movaps %xmm11,0x50(%rsp) + movaps %xmm12,0x60(%rsp) + movaps %xmm13,0x70(%rsp) + movaps %xmm14,0x80(%rsp) + movaps %xmm15,0x90(%rsp) +.Locb_dec_body: +___ +$code.=<<___; + mov $seventh_arg(%rax),$L_p # 7th argument + mov $seventh_arg+8(%rax),$checksum_p# 8th argument + + mov 240($key),$rnds_ + mov $key,$key_ + shl \$4,$rnds_ + $movkey ($key),$rndkey0l # round[0] + $movkey 16($key,$rnds_),$rndkey1 # round[last] + + movdqu ($offset_p),@offset[5] # load last offset_i + pxor $rndkey1,$rndkey0l # round[0] ^ round[last] + pxor $rndkey1,@offset[5] # offset_i ^ round[last] + + mov \$16+32,$rounds + lea 32($key_,$rnds_),$key + $movkey 16($key_),$rndkey1 # round[1] + sub %r10,%rax # twisted $rounds + mov %rax,%r10 # backup twisted $rounds + + movdqu ($L_p),@offset[0] # L_0 for all odd-numbered blocks + movdqu ($checksum_p),$checksum # load checksum + + test \$1,$block_num # is first block number odd? + jnz .Locb_dec_odd + + bsf $block_num,$i1 + add \$1,$block_num + shl \$4,$i1 + movdqu ($L_p,$i1),$inout5 # borrow + movdqu ($inp),$inout0 + lea 16($inp),$inp + + call __ocb_decrypt1 + + movdqa $inout5,@offset[5] + movups $inout0,($out) + xorps $inout0,$checksum # accumulate checksum + lea 16($out),$out + sub \$1,$blocks + jz .Locb_dec_done + +.Locb_dec_odd: + lea 1($block_num),$i1 # even-numbered blocks + lea 3($block_num),$i3 + lea 5($block_num),$i5 + lea 6($block_num),$block_num + bsf $i1,$i1 # ntz(block) + bsf $i3,$i3 + bsf $i5,$i5 + shl \$4,$i1 # ntz(block) -> table offset + shl \$4,$i3 + shl \$4,$i5 + + sub \$6,$blocks + jc .Locb_dec_short + jmp .Locb_dec_grandloop + +.align 32 +.Locb_dec_grandloop: + movdqu `16*0`($inp),$inout0 # load input + movdqu `16*1`($inp),$inout1 + movdqu `16*2`($inp),$inout2 + movdqu `16*3`($inp),$inout3 + movdqu `16*4`($inp),$inout4 + movdqu `16*5`($inp),$inout5 + lea `16*6`($inp),$inp + + call __ocb_decrypt6 + + movups $inout0,`16*0`($out) # store output + pxor $inout0,$checksum # accumulate checksum + movups $inout1,`16*1`($out) + pxor $inout1,$checksum + movups $inout2,`16*2`($out) + pxor $inout2,$checksum + movups $inout3,`16*3`($out) + pxor $inout3,$checksum + movups $inout4,`16*4`($out) + pxor $inout4,$checksum + movups $inout5,`16*5`($out) + pxor $inout5,$checksum + lea `16*6`($out),$out + sub \$6,$blocks + jnc .Locb_dec_grandloop + +.Locb_dec_short: + add \$6,$blocks + jz .Locb_dec_done + + movdqu `16*0`($inp),$inout0 + cmp \$2,$blocks + jb .Locb_dec_one + movdqu `16*1`($inp),$inout1 + je .Locb_dec_two + + movdqu `16*2`($inp),$inout2 + cmp \$4,$blocks + jb .Locb_dec_three + movdqu `16*3`($inp),$inout3 + je .Locb_dec_four + + movdqu `16*4`($inp),$inout4 + pxor $inout5,$inout5 + + call __ocb_decrypt6 + + movdqa @offset[4],@offset[5] + movups $inout0,`16*0`($out) # store output + pxor $inout0,$checksum # accumulate checksum + movups $inout1,`16*1`($out) + pxor $inout1,$checksum + movups $inout2,`16*2`($out) + pxor $inout2,$checksum + movups $inout3,`16*3`($out) + pxor $inout3,$checksum + movups $inout4,`16*4`($out) + pxor $inout4,$checksum + + jmp .Locb_dec_done + +.align 16 +.Locb_dec_one: + movdqa @offset[0],$inout5 # borrow + + call __ocb_decrypt1 + + movdqa $inout5,@offset[5] + movups $inout0,`16*0`($out) # store output + xorps $inout0,$checksum # accumulate checksum + jmp .Locb_dec_done + +.align 16 +.Locb_dec_two: + pxor $inout2,$inout2 + pxor $inout3,$inout3 + + call __ocb_decrypt4 + + movdqa @offset[1],@offset[5] + movups $inout0,`16*0`($out) # store output + xorps $inout0,$checksum # accumulate checksum + movups $inout1,`16*1`($out) + xorps $inout1,$checksum + + jmp .Locb_dec_done + +.align 16 +.Locb_dec_three: + pxor $inout3,$inout3 + + call __ocb_decrypt4 + + movdqa @offset[2],@offset[5] + movups $inout0,`16*0`($out) # store output + xorps $inout0,$checksum # accumulate checksum + movups $inout1,`16*1`($out) + xorps $inout1,$checksum + movups $inout2,`16*2`($out) + xorps $inout2,$checksum + + jmp .Locb_dec_done + +.align 16 +.Locb_dec_four: + call __ocb_decrypt4 + + movdqa @offset[3],@offset[5] + movups $inout0,`16*0`($out) # store output + pxor $inout0,$checksum # accumulate checksum + movups $inout1,`16*1`($out) + pxor $inout1,$checksum + movups $inout2,`16*2`($out) + pxor $inout2,$checksum + movups $inout3,`16*3`($out) + pxor $inout3,$checksum + +.Locb_dec_done: + pxor $rndkey0,@offset[5] # "remove" round[last] + movdqu $checksum,($checksum_p) # store checksum + movdqu @offset[5],($offset_p) # store last offset_i + + xorps %xmm0,%xmm0 # clear register bank + pxor %xmm1,%xmm1 + pxor %xmm2,%xmm2 + pxor %xmm3,%xmm3 + pxor %xmm4,%xmm4 + pxor %xmm5,%xmm5 +___ +$code.=<<___ if (!$win64); + pxor %xmm6,%xmm6 + pxor %xmm7,%xmm7 + pxor %xmm8,%xmm8 + pxor %xmm9,%xmm9 + pxor %xmm10,%xmm10 + pxor %xmm11,%xmm11 + pxor %xmm12,%xmm12 + pxor %xmm13,%xmm13 + pxor %xmm14,%xmm14 + pxor %xmm15,%xmm15 +___ +$code.=<<___ if ($win64); + movaps 0x00(%rsp),%xmm6 + movaps %xmm0,0x00(%rsp) # clear stack + movaps 0x10(%rsp),%xmm7 + movaps %xmm0,0x10(%rsp) + movaps 0x20(%rsp),%xmm8 + movaps %xmm0,0x20(%rsp) + movaps 0x30(%rsp),%xmm9 + movaps %xmm0,0x30(%rsp) + movaps 0x40(%rsp),%xmm10 + movaps %xmm0,0x40(%rsp) + movaps 0x50(%rsp),%xmm11 + movaps %xmm0,0x50(%rsp) + movaps 0x60(%rsp),%xmm12 + movaps %xmm0,0x60(%rsp) + movaps 0x70(%rsp),%xmm13 + movaps %xmm0,0x70(%rsp) + movaps 0x80(%rsp),%xmm14 + movaps %xmm0,0x80(%rsp) + movaps 0x90(%rsp),%xmm15 + movaps %xmm0,0x90(%rsp) + lea 0xa0+0x28(%rsp),%rax +.Locb_dec_pop: + lea 0xa0(%rsp),%rsp +___ +$code.=<<___; + pop %r14 + pop %r13 + pop %r12 + pop %rbp + pop %rbx +.Locb_dec_epilogue: + ret +.size aesni_ocb_decrypt,.-aesni_ocb_decrypt + +.type __ocb_decrypt6,\@abi-omnipotent +.align 32 +__ocb_decrypt6: + pxor $rndkey0l,@offset[5] # offset_i ^ round[0] + movdqu ($L_p,$i1),@offset[1] + movdqa @offset[0],@offset[2] + movdqu ($L_p,$i3),@offset[3] + movdqa @offset[0],@offset[4] + pxor @offset[5],@offset[0] + movdqu ($L_p,$i5),@offset[5] + pxor @offset[0],@offset[1] + pxor @offset[0],$inout0 # input ^ round[0] ^ offset_i + pxor @offset[1],@offset[2] + pxor @offset[1],$inout1 + pxor @offset[2],@offset[3] + pxor @offset[2],$inout2 + pxor @offset[3],@offset[4] + pxor @offset[3],$inout3 + pxor @offset[4],@offset[5] + pxor @offset[4],$inout4 + pxor @offset[5],$inout5 + $movkey 32($key_),$rndkey0 + + lea 1($block_num),$i1 # even-numbered blocks + lea 3($block_num),$i3 + lea 5($block_num),$i5 + add \$6,$block_num + pxor $rndkey0l,@offset[0] # offset_i ^ round[last] + bsf $i1,$i1 # ntz(block) + bsf $i3,$i3 + bsf $i5,$i5 + + aesdec $rndkey1,$inout0 + aesdec $rndkey1,$inout1 + aesdec $rndkey1,$inout2 + aesdec $rndkey1,$inout3 + pxor $rndkey0l,@offset[1] + pxor $rndkey0l,@offset[2] + aesdec $rndkey1,$inout4 + pxor $rndkey0l,@offset[3] + pxor $rndkey0l,@offset[4] + aesdec $rndkey1,$inout5 + $movkey 48($key_),$rndkey1 + pxor $rndkey0l,@offset[5] + + aesdec $rndkey0,$inout0 + aesdec $rndkey0,$inout1 + aesdec $rndkey0,$inout2 + aesdec $rndkey0,$inout3 + aesdec $rndkey0,$inout4 + aesdec $rndkey0,$inout5 + $movkey 64($key_),$rndkey0 + shl \$4,$i1 # ntz(block) -> table offset + shl \$4,$i3 + jmp .Locb_dec_loop6 + +.align 32 +.Locb_dec_loop6: + aesdec $rndkey1,$inout0 + aesdec $rndkey1,$inout1 + aesdec $rndkey1,$inout2 + aesdec $rndkey1,$inout3 + aesdec $rndkey1,$inout4 + aesdec $rndkey1,$inout5 + $movkey ($key,%rax),$rndkey1 + add \$32,%rax + + aesdec $rndkey0,$inout0 + aesdec $rndkey0,$inout1 + aesdec $rndkey0,$inout2 + aesdec $rndkey0,$inout3 + aesdec $rndkey0,$inout4 + aesdec $rndkey0,$inout5 + $movkey -16($key,%rax),$rndkey0 + jnz .Locb_dec_loop6 + + aesdec $rndkey1,$inout0 + aesdec $rndkey1,$inout1 + aesdec $rndkey1,$inout2 + aesdec $rndkey1,$inout3 + aesdec $rndkey1,$inout4 + aesdec $rndkey1,$inout5 + $movkey 16($key_),$rndkey1 + shl \$4,$i5 + + aesdeclast @offset[0],$inout0 + movdqu ($L_p),@offset[0] # L_0 for all odd-numbered blocks + mov %r10,%rax # restore twisted rounds + aesdeclast @offset[1],$inout1 + aesdeclast @offset[2],$inout2 + aesdeclast @offset[3],$inout3 + aesdeclast @offset[4],$inout4 + aesdeclast @offset[5],$inout5 + ret +.size __ocb_decrypt6,.-__ocb_decrypt6 + +.type __ocb_decrypt4,\@abi-omnipotent +.align 32 +__ocb_decrypt4: + pxor $rndkey0l,@offset[5] # offset_i ^ round[0] + movdqu ($L_p,$i1),@offset[1] + movdqa @offset[0],@offset[2] + movdqu ($L_p,$i3),@offset[3] + pxor @offset[5],@offset[0] + pxor @offset[0],@offset[1] + pxor @offset[0],$inout0 # input ^ round[0] ^ offset_i + pxor @offset[1],@offset[2] + pxor @offset[1],$inout1 + pxor @offset[2],@offset[3] + pxor @offset[2],$inout2 + pxor @offset[3],$inout3 + $movkey 32($key_),$rndkey0 + + pxor $rndkey0l,@offset[0] # offset_i ^ round[last] + pxor $rndkey0l,@offset[1] + pxor $rndkey0l,@offset[2] + pxor $rndkey0l,@offset[3] + + aesdec $rndkey1,$inout0 + aesdec $rndkey1,$inout1 + aesdec $rndkey1,$inout2 + aesdec $rndkey1,$inout3 + $movkey 48($key_),$rndkey1 + + aesdec $rndkey0,$inout0 + aesdec $rndkey0,$inout1 + aesdec $rndkey0,$inout2 + aesdec $rndkey0,$inout3 + $movkey 64($key_),$rndkey0 + jmp .Locb_dec_loop4 + +.align 32 +.Locb_dec_loop4: + aesdec $rndkey1,$inout0 + aesdec $rndkey1,$inout1 + aesdec $rndkey1,$inout2 + aesdec $rndkey1,$inout3 + $movkey ($key,%rax),$rndkey1 + add \$32,%rax + + aesdec $rndkey0,$inout0 + aesdec $rndkey0,$inout1 + aesdec $rndkey0,$inout2 + aesdec $rndkey0,$inout3 + $movkey -16($key,%rax),$rndkey0 + jnz .Locb_dec_loop4 + + aesdec $rndkey1,$inout0 + aesdec $rndkey1,$inout1 + aesdec $rndkey1,$inout2 + aesdec $rndkey1,$inout3 + $movkey 16($key_),$rndkey1 + mov %r10,%rax # restore twisted rounds + + aesdeclast @offset[0],$inout0 + aesdeclast @offset[1],$inout1 + aesdeclast @offset[2],$inout2 + aesdeclast @offset[3],$inout3 + ret +.size __ocb_decrypt4,.-__ocb_decrypt4 + +.type __ocb_decrypt1,\@abi-omnipotent +.align 32 +__ocb_decrypt1: + pxor @offset[5],$inout5 # offset_i + pxor $rndkey0l,$inout5 # offset_i ^ round[0] + pxor $inout5,$inout0 # input ^ round[0] ^ offset_i + $movkey 32($key_),$rndkey0 + + aesdec $rndkey1,$inout0 + $movkey 48($key_),$rndkey1 + pxor $rndkey0l,$inout5 # offset_i ^ round[last] + + aesdec $rndkey0,$inout0 + $movkey 64($key_),$rndkey0 + jmp .Locb_dec_loop1 + +.align 32 +.Locb_dec_loop1: + aesdec $rndkey1,$inout0 + $movkey ($key,%rax),$rndkey1 + add \$32,%rax + + aesdec $rndkey0,$inout0 + $movkey -16($key,%rax),$rndkey0 + jnz .Locb_dec_loop1 + + aesdec $rndkey1,$inout0 + $movkey 16($key_),$rndkey1 # redundant in tail + mov %r10,%rax # restore twisted rounds + + aesdeclast $inout5,$inout0 + ret +.size __ocb_decrypt1,.-__ocb_decrypt1 +___ +} }} + +######################################################################## +# void $PREFIX_cbc_encrypt (const void *inp, void *out, +# size_t length, const AES_KEY *key, +# unsigned char *ivp,const int enc); +{ +my $frame_size = 0x10 + ($win64?0xa0:0); # used in decrypt +my ($iv,$in0,$in1,$in2,$in3,$in4)=map("%xmm$_",(10..15)); +my $inp_=$key_; + +$code.=<<___; +.globl ${PREFIX}_cbc_encrypt +.type ${PREFIX}_cbc_encrypt,\@function,6 +.align 16 +${PREFIX}_cbc_encrypt: + test $len,$len # check length + jz .Lcbc_ret + + mov 240($key),$rnds_ # key->rounds + mov $key,$key_ # backup $key + test %r9d,%r9d # 6th argument + jz .Lcbc_decrypt +#--------------------------- CBC ENCRYPT ------------------------------# + movups ($ivp),$inout0 # load iv as initial state + mov $rnds_,$rounds + cmp \$16,$len + jb .Lcbc_enc_tail + sub \$16,$len + jmp .Lcbc_enc_loop +.align 16 +.Lcbc_enc_loop: + movups ($inp),$inout1 # load input + lea 16($inp),$inp + #xorps $inout1,$inout0 +___ + &aesni_generate1("enc",$key,$rounds,$inout0,$inout1); +$code.=<<___; + mov $rnds_,$rounds # restore $rounds + mov $key_,$key # restore $key + movups $inout0,0($out) # store output + lea 16($out),$out + sub \$16,$len + jnc .Lcbc_enc_loop + add \$16,$len + jnz .Lcbc_enc_tail + pxor $rndkey0,$rndkey0 # clear register bank + pxor $rndkey1,$rndkey1 + movups $inout0,($ivp) + pxor $inout0,$inout0 + pxor $inout1,$inout1 + jmp .Lcbc_ret + +.Lcbc_enc_tail: + mov $len,%rcx # zaps $key + xchg $inp,$out # $inp is %rsi and $out is %rdi now + .long 0x9066A4F3 # rep movsb + mov \$16,%ecx # zero tail + sub $len,%rcx + xor %eax,%eax + .long 0x9066AAF3 # rep stosb + lea -16(%rdi),%rdi # rewind $out by 1 block + mov $rnds_,$rounds # restore $rounds + mov %rdi,%rsi # $inp and $out are the same + mov $key_,$key # restore $key + xor $len,$len # len=16 + jmp .Lcbc_enc_loop # one more spin + #--------------------------- CBC DECRYPT ------------------------------# +.align 16 +.Lcbc_decrypt: + cmp \$16,$len + jne .Lcbc_decrypt_bulk + + # handle single block without allocating stack frame, + # useful in ciphertext stealing mode + movdqu ($inp),$inout0 # load input + movdqu ($ivp),$inout1 # load iv + movdqa $inout0,$inout2 # future iv +___ + &aesni_generate1("dec",$key,$rnds_); +$code.=<<___; + pxor $rndkey0,$rndkey0 # clear register bank + pxor $rndkey1,$rndkey1 + movdqu $inout2,($ivp) # store iv + xorps $inout1,$inout0 # ^=iv + pxor $inout1,$inout1 + movups $inout0,($out) # store output + pxor $inout0,$inout0 + jmp .Lcbc_ret +.align 16 +.Lcbc_decrypt_bulk: + lea (%rsp),%rax + push %rbp + sub \$$frame_size,%rsp + and \$-16,%rsp # Linux kernel stack can be incorrectly seeded +___ +$code.=<<___ if ($win64); + movaps %xmm6,0x10(%rsp) + movaps %xmm7,0x20(%rsp) + movaps %xmm8,0x30(%rsp) + movaps %xmm9,0x40(%rsp) + movaps %xmm10,0x50(%rsp) + movaps %xmm11,0x60(%rsp) + movaps %xmm12,0x70(%rsp) + movaps %xmm13,0x80(%rsp) + movaps %xmm14,0x90(%rsp) + movaps %xmm15,0xa0(%rsp) +.Lcbc_decrypt_body: +___ +$code.=<<___; + lea -8(%rax),%rbp + movups ($ivp),$iv + mov $rnds_,$rounds + cmp \$0x50,$len + jbe .Lcbc_dec_tail + + $movkey ($key),$rndkey0 + movdqu 0x00($inp),$inout0 # load input + movdqu 0x10($inp),$inout1 + movdqa $inout0,$in0 + movdqu 0x20($inp),$inout2 + movdqa $inout1,$in1 + movdqu 0x30($inp),$inout3 + movdqa $inout2,$in2 + movdqu 0x40($inp),$inout4 + movdqa $inout3,$in3 + movdqu 0x50($inp),$inout5 + movdqa $inout4,$in4 + mov OPENSSL_ia32cap_P+4(%rip),%r9d + cmp \$0x70,$len + jbe .Lcbc_dec_six_or_seven + + and \$`1<<26|1<<22`,%r9d # isolate XSAVE+MOVBE + sub \$0x50,$len # $len is biased by -5*16 + cmp \$`1<<22`,%r9d # check for MOVBE without XSAVE + je .Lcbc_dec_loop6_enter # [which denotes Atom Silvermont] + sub \$0x20,$len # $len is biased by -7*16 + lea 0x70($key),$key # size optimization + jmp .Lcbc_dec_loop8_enter +.align 16 +.Lcbc_dec_loop8: + movups $inout7,($out) + lea 0x10($out),$out +.Lcbc_dec_loop8_enter: + movdqu 0x60($inp),$inout6 + pxor $rndkey0,$inout0 + movdqu 0x70($inp),$inout7 + pxor $rndkey0,$inout1 + $movkey 0x10-0x70($key),$rndkey1 + pxor $rndkey0,$inout2 + xor $inp_,$inp_ + cmp \$0x70,$len # is there at least 0x60 bytes ahead? + pxor $rndkey0,$inout3 + pxor $rndkey0,$inout4 + pxor $rndkey0,$inout5 + pxor $rndkey0,$inout6 + + aesdec $rndkey1,$inout0 + pxor $rndkey0,$inout7 + $movkey 0x20-0x70($key),$rndkey0 + aesdec $rndkey1,$inout1 + aesdec $rndkey1,$inout2 + aesdec $rndkey1,$inout3 + aesdec $rndkey1,$inout4 + aesdec $rndkey1,$inout5 + aesdec $rndkey1,$inout6 + setnc ${inp_}b + shl \$7,$inp_ + aesdec $rndkey1,$inout7 + add $inp,$inp_ + $movkey 0x30-0x70($key),$rndkey1 +___ +for($i=1;$i<12;$i++) { +my $rndkeyx = ($i&1)?$rndkey0:$rndkey1; +$code.=<<___ if ($i==7); + cmp \$11,$rounds +___ +$code.=<<___; + aesdec $rndkeyx,$inout0 + aesdec $rndkeyx,$inout1 + aesdec $rndkeyx,$inout2 + aesdec $rndkeyx,$inout3 + aesdec $rndkeyx,$inout4 + aesdec $rndkeyx,$inout5 + aesdec $rndkeyx,$inout6 + aesdec $rndkeyx,$inout7 + $movkey `0x30+0x10*$i`-0x70($key),$rndkeyx +___ +$code.=<<___ if ($i<6 || (!($i&1) && $i>7)); + nop +___ +$code.=<<___ if ($i==7); + jb .Lcbc_dec_done +___ +$code.=<<___ if ($i==9); + je .Lcbc_dec_done +___ +$code.=<<___ if ($i==11); + jmp .Lcbc_dec_done +___ +} +$code.=<<___; +.align 16 +.Lcbc_dec_done: + aesdec $rndkey1,$inout0 + aesdec $rndkey1,$inout1 + pxor $rndkey0,$iv + pxor $rndkey0,$in0 + aesdec $rndkey1,$inout2 + aesdec $rndkey1,$inout3 + pxor $rndkey0,$in1 + pxor $rndkey0,$in2 + aesdec $rndkey1,$inout4 + aesdec $rndkey1,$inout5 + pxor $rndkey0,$in3 + pxor $rndkey0,$in4 + aesdec $rndkey1,$inout6 + aesdec $rndkey1,$inout7 + movdqu 0x50($inp),$rndkey1 + + aesdeclast $iv,$inout0 + movdqu 0x60($inp),$iv # borrow $iv + pxor $rndkey0,$rndkey1 + aesdeclast $in0,$inout1 + pxor $rndkey0,$iv + movdqu 0x70($inp),$rndkey0 # next IV + aesdeclast $in1,$inout2 + lea 0x80($inp),$inp + movdqu 0x00($inp_),$in0 + aesdeclast $in2,$inout3 + aesdeclast $in3,$inout4 + movdqu 0x10($inp_),$in1 + movdqu 0x20($inp_),$in2 + aesdeclast $in4,$inout5 + aesdeclast $rndkey1,$inout6 + movdqu 0x30($inp_),$in3 + movdqu 0x40($inp_),$in4 + aesdeclast $iv,$inout7 + movdqa $rndkey0,$iv # return $iv + movdqu 0x50($inp_),$rndkey1 + $movkey -0x70($key),$rndkey0 + + movups $inout0,($out) # store output + movdqa $in0,$inout0 + movups $inout1,0x10($out) + movdqa $in1,$inout1 + movups $inout2,0x20($out) + movdqa $in2,$inout2 + movups $inout3,0x30($out) + movdqa $in3,$inout3 + movups $inout4,0x40($out) + movdqa $in4,$inout4 + movups $inout5,0x50($out) + movdqa $rndkey1,$inout5 + movups $inout6,0x60($out) + lea 0x70($out),$out + + sub \$0x80,$len + ja .Lcbc_dec_loop8 + + movaps $inout7,$inout0 + lea -0x70($key),$key + add \$0x70,$len + jle .Lcbc_dec_clear_tail_collected + movups $inout7,($out) + lea 0x10($out),$out + cmp \$0x50,$len + jbe .Lcbc_dec_tail + + movaps $in0,$inout0 +.Lcbc_dec_six_or_seven: + cmp \$0x60,$len + ja .Lcbc_dec_seven + + movaps $inout5,$inout6 + call _aesni_decrypt6 + pxor $iv,$inout0 # ^= IV + movaps $inout6,$iv + pxor $in0,$inout1 + movdqu $inout0,($out) + pxor $in1,$inout2 + movdqu $inout1,0x10($out) + pxor $inout1,$inout1 # clear register bank + pxor $in2,$inout3 + movdqu $inout2,0x20($out) + pxor $inout2,$inout2 + pxor $in3,$inout4 + movdqu $inout3,0x30($out) + pxor $inout3,$inout3 + pxor $in4,$inout5 + movdqu $inout4,0x40($out) + pxor $inout4,$inout4 + lea 0x50($out),$out + movdqa $inout5,$inout0 + pxor $inout5,$inout5 + jmp .Lcbc_dec_tail_collected + +.align 16 +.Lcbc_dec_seven: + movups 0x60($inp),$inout6 + xorps $inout7,$inout7 + call _aesni_decrypt8 + movups 0x50($inp),$inout7 + pxor $iv,$inout0 # ^= IV + movups 0x60($inp),$iv + pxor $in0,$inout1 + movdqu $inout0,($out) + pxor $in1,$inout2 + movdqu $inout1,0x10($out) + pxor $inout1,$inout1 # clear register bank + pxor $in2,$inout3 + movdqu $inout2,0x20($out) + pxor $inout2,$inout2 + pxor $in3,$inout4 + movdqu $inout3,0x30($out) + pxor $inout3,$inout3 + pxor $in4,$inout5 + movdqu $inout4,0x40($out) + pxor $inout4,$inout4 + pxor $inout7,$inout6 + movdqu $inout5,0x50($out) + pxor $inout5,$inout5 + lea 0x60($out),$out + movdqa $inout6,$inout0 + pxor $inout6,$inout6 + pxor $inout7,$inout7 + jmp .Lcbc_dec_tail_collected + +.align 16 +.Lcbc_dec_loop6: + movups $inout5,($out) + lea 0x10($out),$out + movdqu 0x00($inp),$inout0 # load input + movdqu 0x10($inp),$inout1 + movdqa $inout0,$in0 + movdqu 0x20($inp),$inout2 + movdqa $inout1,$in1 + movdqu 0x30($inp),$inout3 + movdqa $inout2,$in2 + movdqu 0x40($inp),$inout4 + movdqa $inout3,$in3 + movdqu 0x50($inp),$inout5 + movdqa $inout4,$in4 +.Lcbc_dec_loop6_enter: + lea 0x60($inp),$inp + movdqa $inout5,$inout6 + + call _aesni_decrypt6 + + pxor $iv,$inout0 # ^= IV + movdqa $inout6,$iv + pxor $in0,$inout1 + movdqu $inout0,($out) + pxor $in1,$inout2 + movdqu $inout1,0x10($out) + pxor $in2,$inout3 + movdqu $inout2,0x20($out) + pxor $in3,$inout4 + mov $key_,$key + movdqu $inout3,0x30($out) + pxor $in4,$inout5 + mov $rnds_,$rounds + movdqu $inout4,0x40($out) + lea 0x50($out),$out + sub \$0x60,$len + ja .Lcbc_dec_loop6 + + movdqa $inout5,$inout0 + add \$0x50,$len + jle .Lcbc_dec_clear_tail_collected + movups $inout5,($out) + lea 0x10($out),$out + +.Lcbc_dec_tail: + movups ($inp),$inout0 + sub \$0x10,$len + jbe .Lcbc_dec_one # $len is 1*16 or less + + movups 0x10($inp),$inout1 + movaps $inout0,$in0 + sub \$0x10,$len + jbe .Lcbc_dec_two # $len is 2*16 or less + + movups 0x20($inp),$inout2 + movaps $inout1,$in1 + sub \$0x10,$len + jbe .Lcbc_dec_three # $len is 3*16 or less + + movups 0x30($inp),$inout3 + movaps $inout2,$in2 + sub \$0x10,$len + jbe .Lcbc_dec_four # $len is 4*16 or less + + movups 0x40($inp),$inout4 # $len is 5*16 or less + movaps $inout3,$in3 + movaps $inout4,$in4 + xorps $inout5,$inout5 + call _aesni_decrypt6 + pxor $iv,$inout0 + movaps $in4,$iv + pxor $in0,$inout1 + movdqu $inout0,($out) + pxor $in1,$inout2 + movdqu $inout1,0x10($out) + pxor $inout1,$inout1 # clear register bank + pxor $in2,$inout3 + movdqu $inout2,0x20($out) + pxor $inout2,$inout2 + pxor $in3,$inout4 + movdqu $inout3,0x30($out) + pxor $inout3,$inout3 + lea 0x40($out),$out + movdqa $inout4,$inout0 + pxor $inout4,$inout4 + pxor $inout5,$inout5 + sub \$0x10,$len + jmp .Lcbc_dec_tail_collected + +.align 16 +.Lcbc_dec_one: + movaps $inout0,$in0 +___ + &aesni_generate1("dec",$key,$rounds); +$code.=<<___; + xorps $iv,$inout0 + movaps $in0,$iv + jmp .Lcbc_dec_tail_collected +.align 16 +.Lcbc_dec_two: + movaps $inout1,$in1 + call _aesni_decrypt2 + pxor $iv,$inout0 + movaps $in1,$iv + pxor $in0,$inout1 + movdqu $inout0,($out) + movdqa $inout1,$inout0 + pxor $inout1,$inout1 # clear register bank + lea 0x10($out),$out + jmp .Lcbc_dec_tail_collected +.align 16 +.Lcbc_dec_three: + movaps $inout2,$in2 + call _aesni_decrypt3 + pxor $iv,$inout0 + movaps $in2,$iv + pxor $in0,$inout1 + movdqu $inout0,($out) + pxor $in1,$inout2 + movdqu $inout1,0x10($out) + pxor $inout1,$inout1 # clear register bank + movdqa $inout2,$inout0 + pxor $inout2,$inout2 + lea 0x20($out),$out + jmp .Lcbc_dec_tail_collected +.align 16 +.Lcbc_dec_four: + movaps $inout3,$in3 + call _aesni_decrypt4 + pxor $iv,$inout0 + movaps $in3,$iv + pxor $in0,$inout1 + movdqu $inout0,($out) + pxor $in1,$inout2 + movdqu $inout1,0x10($out) + pxor $inout1,$inout1 # clear register bank + pxor $in2,$inout3 + movdqu $inout2,0x20($out) + pxor $inout2,$inout2 + movdqa $inout3,$inout0 + pxor $inout3,$inout3 + lea 0x30($out),$out + jmp .Lcbc_dec_tail_collected + +.align 16 +.Lcbc_dec_clear_tail_collected: + pxor $inout1,$inout1 # clear register bank + pxor $inout2,$inout2 + pxor $inout3,$inout3 +___ +$code.=<<___ if (!$win64); + pxor $inout4,$inout4 # %xmm6..9 + pxor $inout5,$inout5 + pxor $inout6,$inout6 + pxor $inout7,$inout7 +___ +$code.=<<___; +.Lcbc_dec_tail_collected: + movups $iv,($ivp) + and \$15,$len + jnz .Lcbc_dec_tail_partial + movups $inout0,($out) + pxor $inout0,$inout0 + jmp .Lcbc_dec_ret +.align 16 +.Lcbc_dec_tail_partial: + movaps $inout0,(%rsp) + pxor $inout0,$inout0 + mov \$16,%rcx + mov $out,%rdi + sub $len,%rcx + lea (%rsp),%rsi + .long 0x9066A4F3 # rep movsb + movdqa $inout0,(%rsp) + +.Lcbc_dec_ret: + xorps $rndkey0,$rndkey0 # %xmm0 + pxor $rndkey1,$rndkey1 +___ +$code.=<<___ if ($win64); + movaps 0x10(%rsp),%xmm6 + movaps %xmm0,0x10(%rsp) # clear stack + movaps 0x20(%rsp),%xmm7 + movaps %xmm0,0x20(%rsp) + movaps 0x30(%rsp),%xmm8 + movaps %xmm0,0x30(%rsp) + movaps 0x40(%rsp),%xmm9 + movaps %xmm0,0x40(%rsp) + movaps 0x50(%rsp),%xmm10 + movaps %xmm0,0x50(%rsp) + movaps 0x60(%rsp),%xmm11 + movaps %xmm0,0x60(%rsp) + movaps 0x70(%rsp),%xmm12 + movaps %xmm0,0x70(%rsp) + movaps 0x80(%rsp),%xmm13 + movaps %xmm0,0x80(%rsp) + movaps 0x90(%rsp),%xmm14 + movaps %xmm0,0x90(%rsp) + movaps 0xa0(%rsp),%xmm15 + movaps %xmm0,0xa0(%rsp) +___ +$code.=<<___; + lea (%rbp),%rsp + pop %rbp +.Lcbc_ret: + ret +.size ${PREFIX}_cbc_encrypt,.-${PREFIX}_cbc_encrypt +___ +} +# int ${PREFIX}_set_decrypt_key(const unsigned char *inp, +# int bits, AES_KEY *key) +# +# input: $inp user-supplied key +# $bits $inp length in bits +# $key pointer to key schedule +# output: %eax 0 denoting success, -1 or -2 - failure (see C) +# *$key key schedule +# +{ my ($inp,$bits,$key) = @_4args; + $bits =~ s/%r/%e/; + +$code.=<<___; +.globl ${PREFIX}_set_decrypt_key +.type ${PREFIX}_set_decrypt_key,\@abi-omnipotent +.align 16 +${PREFIX}_set_decrypt_key: + .byte 0x48,0x83,0xEC,0x08 # sub rsp,8 + call __aesni_set_encrypt_key + shl \$4,$bits # rounds-1 after _aesni_set_encrypt_key + test %eax,%eax + jnz .Ldec_key_ret + lea 16($key,$bits),$inp # points at the end of key schedule + + $movkey ($key),%xmm0 # just swap + $movkey ($inp),%xmm1 + $movkey %xmm0,($inp) + $movkey %xmm1,($key) + lea 16($key),$key + lea -16($inp),$inp + +.Ldec_key_inverse: + $movkey ($key),%xmm0 # swap and inverse + $movkey ($inp),%xmm1 + aesimc %xmm0,%xmm0 + aesimc %xmm1,%xmm1 + lea 16($key),$key + lea -16($inp),$inp + $movkey %xmm0,16($inp) + $movkey %xmm1,-16($key) + cmp $key,$inp + ja .Ldec_key_inverse + + $movkey ($key),%xmm0 # inverse middle + aesimc %xmm0,%xmm0 + pxor %xmm1,%xmm1 + $movkey %xmm0,($inp) + pxor %xmm0,%xmm0 +.Ldec_key_ret: + add \$8,%rsp + ret +.LSEH_end_set_decrypt_key: +.size ${PREFIX}_set_decrypt_key,.-${PREFIX}_set_decrypt_key +___ + +# This is based on submission by +# +# Huang Ying +# Vinodh Gopal +# Kahraman Akdemir +# +# Aggressively optimized in respect to aeskeygenassist's critical path +# and is contained in %xmm0-5 to meet Win64 ABI requirement. +# +# int ${PREFIX}_set_encrypt_key(const unsigned char *inp, +# int bits, AES_KEY * const key); +# +# input: $inp user-supplied key +# $bits $inp length in bits +# $key pointer to key schedule +# output: %eax 0 denoting success, -1 or -2 - failure (see C) +# $bits rounds-1 (used in aesni_set_decrypt_key) +# *$key key schedule +# $key pointer to key schedule (used in +# aesni_set_decrypt_key) +# +# Subroutine is frame-less, which means that only volatile registers +# are used. Note that it's declared "abi-omnipotent", which means that +# amount of volatile registers is smaller on Windows. +# +$code.=<<___; +.globl ${PREFIX}_set_encrypt_key +.type ${PREFIX}_set_encrypt_key,\@abi-omnipotent +.align 16 +${PREFIX}_set_encrypt_key: +__aesni_set_encrypt_key: + .byte 0x48,0x83,0xEC,0x08 # sub rsp,8 + mov \$-1,%rax + test $inp,$inp + jz .Lenc_key_ret + test $key,$key + jz .Lenc_key_ret + + mov \$`1<<28|1<<11`,%r10d # AVX and XOP bits + movups ($inp),%xmm0 # pull first 128 bits of *userKey + xorps %xmm4,%xmm4 # low dword of xmm4 is assumed 0 + and OPENSSL_ia32cap_P+4(%rip),%r10d + lea 16($key),%rax # %rax is used as modifiable copy of $key + cmp \$256,$bits + je .L14rounds + cmp \$192,$bits + je .L12rounds + cmp \$128,$bits + jne .Lbad_keybits + +.L10rounds: + mov \$9,$bits # 10 rounds for 128-bit key + cmp \$`1<<28`,%r10d # AVX, bit no XOP + je .L10rounds_alt + + $movkey %xmm0,($key) # round 0 + aeskeygenassist \$0x1,%xmm0,%xmm1 # round 1 + call .Lkey_expansion_128_cold + aeskeygenassist \$0x2,%xmm0,%xmm1 # round 2 + call .Lkey_expansion_128 + aeskeygenassist \$0x4,%xmm0,%xmm1 # round 3 + call .Lkey_expansion_128 + aeskeygenassist \$0x8,%xmm0,%xmm1 # round 4 + call .Lkey_expansion_128 + aeskeygenassist \$0x10,%xmm0,%xmm1 # round 5 + call .Lkey_expansion_128 + aeskeygenassist \$0x20,%xmm0,%xmm1 # round 6 + call .Lkey_expansion_128 + aeskeygenassist \$0x40,%xmm0,%xmm1 # round 7 + call .Lkey_expansion_128 + aeskeygenassist \$0x80,%xmm0,%xmm1 # round 8 + call .Lkey_expansion_128 + aeskeygenassist \$0x1b,%xmm0,%xmm1 # round 9 + call .Lkey_expansion_128 + aeskeygenassist \$0x36,%xmm0,%xmm1 # round 10 + call .Lkey_expansion_128 + $movkey %xmm0,(%rax) + mov $bits,80(%rax) # 240(%rdx) + xor %eax,%eax + jmp .Lenc_key_ret + +.align 16 +.L10rounds_alt: + movdqa .Lkey_rotate(%rip),%xmm5 + mov \$8,%r10d + movdqa .Lkey_rcon1(%rip),%xmm4 + movdqa %xmm0,%xmm2 + movdqu %xmm0,($key) + jmp .Loop_key128 + +.align 16 +.Loop_key128: + pshufb %xmm5,%xmm0 + aesenclast %xmm4,%xmm0 + pslld \$1,%xmm4 + lea 16(%rax),%rax + + movdqa %xmm2,%xmm3 + pslldq \$4,%xmm2 + pxor %xmm2,%xmm3 + pslldq \$4,%xmm2 + pxor %xmm2,%xmm3 + pslldq \$4,%xmm2 + pxor %xmm3,%xmm2 + + pxor %xmm2,%xmm0 + movdqu %xmm0,-16(%rax) + movdqa %xmm0,%xmm2 + + dec %r10d + jnz .Loop_key128 + + movdqa .Lkey_rcon1b(%rip),%xmm4 + + pshufb %xmm5,%xmm0 + aesenclast %xmm4,%xmm0 + pslld \$1,%xmm4 + + movdqa %xmm2,%xmm3 + pslldq \$4,%xmm2 + pxor %xmm2,%xmm3 + pslldq \$4,%xmm2 + pxor %xmm2,%xmm3 + pslldq \$4,%xmm2 + pxor %xmm3,%xmm2 + + pxor %xmm2,%xmm0 + movdqu %xmm0,(%rax) + + movdqa %xmm0,%xmm2 + pshufb %xmm5,%xmm0 + aesenclast %xmm4,%xmm0 + + movdqa %xmm2,%xmm3 + pslldq \$4,%xmm2 + pxor %xmm2,%xmm3 + pslldq \$4,%xmm2 + pxor %xmm2,%xmm3 + pslldq \$4,%xmm2 + pxor %xmm3,%xmm2 + + pxor %xmm2,%xmm0 + movdqu %xmm0,16(%rax) + + mov $bits,96(%rax) # 240($key) + xor %eax,%eax + jmp .Lenc_key_ret + +.align 16 +.L12rounds: + movq 16($inp),%xmm2 # remaining 1/3 of *userKey + mov \$11,$bits # 12 rounds for 192 + cmp \$`1<<28`,%r10d # AVX, but no XOP + je .L12rounds_alt + + $movkey %xmm0,($key) # round 0 + aeskeygenassist \$0x1,%xmm2,%xmm1 # round 1,2 + call .Lkey_expansion_192a_cold + aeskeygenassist \$0x2,%xmm2,%xmm1 # round 2,3 + call .Lkey_expansion_192b + aeskeygenassist \$0x4,%xmm2,%xmm1 # round 4,5 + call .Lkey_expansion_192a + aeskeygenassist \$0x8,%xmm2,%xmm1 # round 5,6 + call .Lkey_expansion_192b + aeskeygenassist \$0x10,%xmm2,%xmm1 # round 7,8 + call .Lkey_expansion_192a + aeskeygenassist \$0x20,%xmm2,%xmm1 # round 8,9 + call .Lkey_expansion_192b + aeskeygenassist \$0x40,%xmm2,%xmm1 # round 10,11 + call .Lkey_expansion_192a + aeskeygenassist \$0x80,%xmm2,%xmm1 # round 11,12 + call .Lkey_expansion_192b + $movkey %xmm0,(%rax) + mov $bits,48(%rax) # 240(%rdx) + xor %rax, %rax + jmp .Lenc_key_ret + +.align 16 +.L12rounds_alt: + movdqa .Lkey_rotate192(%rip),%xmm5 + movdqa .Lkey_rcon1(%rip),%xmm4 + mov \$8,%r10d + movdqu %xmm0,($key) + jmp .Loop_key192 + +.align 16 +.Loop_key192: + movq %xmm2,0(%rax) + movdqa %xmm2,%xmm1 + pshufb %xmm5,%xmm2 + aesenclast %xmm4,%xmm2 + pslld \$1, %xmm4 + lea 24(%rax),%rax + + movdqa %xmm0,%xmm3 + pslldq \$4,%xmm0 + pxor %xmm0,%xmm3 + pslldq \$4,%xmm0 + pxor %xmm0,%xmm3 + pslldq \$4,%xmm0 + pxor %xmm3,%xmm0 + + pshufd \$0xff,%xmm0,%xmm3 + pxor %xmm1,%xmm3 + pslldq \$4,%xmm1 + pxor %xmm1,%xmm3 + + pxor %xmm2,%xmm0 + pxor %xmm3,%xmm2 + movdqu %xmm0,-16(%rax) + + dec %r10d + jnz .Loop_key192 + + mov $bits,32(%rax) # 240($key) + xor %eax,%eax + jmp .Lenc_key_ret + +.align 16 +.L14rounds: + movups 16($inp),%xmm2 # remaning half of *userKey + mov \$13,$bits # 14 rounds for 256 + lea 16(%rax),%rax + cmp \$`1<<28`,%r10d # AVX, but no XOP + je .L14rounds_alt + + $movkey %xmm0,($key) # round 0 + $movkey %xmm2,16($key) # round 1 + aeskeygenassist \$0x1,%xmm2,%xmm1 # round 2 + call .Lkey_expansion_256a_cold + aeskeygenassist \$0x1,%xmm0,%xmm1 # round 3 + call .Lkey_expansion_256b + aeskeygenassist \$0x2,%xmm2,%xmm1 # round 4 + call .Lkey_expansion_256a + aeskeygenassist \$0x2,%xmm0,%xmm1 # round 5 + call .Lkey_expansion_256b + aeskeygenassist \$0x4,%xmm2,%xmm1 # round 6 + call .Lkey_expansion_256a + aeskeygenassist \$0x4,%xmm0,%xmm1 # round 7 + call .Lkey_expansion_256b + aeskeygenassist \$0x8,%xmm2,%xmm1 # round 8 + call .Lkey_expansion_256a + aeskeygenassist \$0x8,%xmm0,%xmm1 # round 9 + call .Lkey_expansion_256b + aeskeygenassist \$0x10,%xmm2,%xmm1 # round 10 + call .Lkey_expansion_256a + aeskeygenassist \$0x10,%xmm0,%xmm1 # round 11 + call .Lkey_expansion_256b + aeskeygenassist \$0x20,%xmm2,%xmm1 # round 12 + call .Lkey_expansion_256a + aeskeygenassist \$0x20,%xmm0,%xmm1 # round 13 + call .Lkey_expansion_256b + aeskeygenassist \$0x40,%xmm2,%xmm1 # round 14 + call .Lkey_expansion_256a + $movkey %xmm0,(%rax) + mov $bits,16(%rax) # 240(%rdx) + xor %rax,%rax + jmp .Lenc_key_ret + +.align 16 +.L14rounds_alt: + movdqa .Lkey_rotate(%rip),%xmm5 + movdqa .Lkey_rcon1(%rip),%xmm4 + mov \$7,%r10d + movdqu %xmm0,0($key) + movdqa %xmm2,%xmm1 + movdqu %xmm2,16($key) + jmp .Loop_key256 + +.align 16 +.Loop_key256: + pshufb %xmm5,%xmm2 + aesenclast %xmm4,%xmm2 + + movdqa %xmm0,%xmm3 + pslldq \$4,%xmm0 + pxor %xmm0,%xmm3 + pslldq \$4,%xmm0 + pxor %xmm0,%xmm3 + pslldq \$4,%xmm0 + pxor %xmm3,%xmm0 + pslld \$1,%xmm4 + + pxor %xmm2,%xmm0 + movdqu %xmm0,(%rax) + + dec %r10d + jz .Ldone_key256 + + pshufd \$0xff,%xmm0,%xmm2 + pxor %xmm3,%xmm3 + aesenclast %xmm3,%xmm2 + + movdqa %xmm1,%xmm3 + pslldq \$4,%xmm1 + pxor %xmm1,%xmm3 + pslldq \$4,%xmm1 + pxor %xmm1,%xmm3 + pslldq \$4,%xmm1 + pxor %xmm3,%xmm1 + + pxor %xmm1,%xmm2 + movdqu %xmm2,16(%rax) + lea 32(%rax),%rax + movdqa %xmm2,%xmm1 + + jmp .Loop_key256 + +.Ldone_key256: + mov $bits,16(%rax) # 240($key) + xor %eax,%eax + jmp .Lenc_key_ret + +.align 16 +.Lbad_keybits: + mov \$-2,%rax +.Lenc_key_ret: + pxor %xmm0,%xmm0 + pxor %xmm1,%xmm1 + pxor %xmm2,%xmm2 + pxor %xmm3,%xmm3 + pxor %xmm4,%xmm4 + pxor %xmm5,%xmm5 + add \$8,%rsp + ret +.LSEH_end_set_encrypt_key: + +.align 16 +.Lkey_expansion_128: + $movkey %xmm0,(%rax) + lea 16(%rax),%rax +.Lkey_expansion_128_cold: + shufps \$0b00010000,%xmm0,%xmm4 + xorps %xmm4, %xmm0 + shufps \$0b10001100,%xmm0,%xmm4 + xorps %xmm4, %xmm0 + shufps \$0b11111111,%xmm1,%xmm1 # critical path + xorps %xmm1,%xmm0 + ret + +.align 16 +.Lkey_expansion_192a: + $movkey %xmm0,(%rax) + lea 16(%rax),%rax +.Lkey_expansion_192a_cold: + movaps %xmm2, %xmm5 +.Lkey_expansion_192b_warm: + shufps \$0b00010000,%xmm0,%xmm4 + movdqa %xmm2,%xmm3 + xorps %xmm4,%xmm0 + shufps \$0b10001100,%xmm0,%xmm4 + pslldq \$4,%xmm3 + xorps %xmm4,%xmm0 + pshufd \$0b01010101,%xmm1,%xmm1 # critical path + pxor %xmm3,%xmm2 + pxor %xmm1,%xmm0 + pshufd \$0b11111111,%xmm0,%xmm3 + pxor %xmm3,%xmm2 + ret + +.align 16 +.Lkey_expansion_192b: + movaps %xmm0,%xmm3 + shufps \$0b01000100,%xmm0,%xmm5 + $movkey %xmm5,(%rax) + shufps \$0b01001110,%xmm2,%xmm3 + $movkey %xmm3,16(%rax) + lea 32(%rax),%rax + jmp .Lkey_expansion_192b_warm + +.align 16 +.Lkey_expansion_256a: + $movkey %xmm2,(%rax) + lea 16(%rax),%rax +.Lkey_expansion_256a_cold: + shufps \$0b00010000,%xmm0,%xmm4 + xorps %xmm4,%xmm0 + shufps \$0b10001100,%xmm0,%xmm4 + xorps %xmm4,%xmm0 + shufps \$0b11111111,%xmm1,%xmm1 # critical path + xorps %xmm1,%xmm0 + ret + +.align 16 +.Lkey_expansion_256b: + $movkey %xmm0,(%rax) + lea 16(%rax),%rax + + shufps \$0b00010000,%xmm2,%xmm4 + xorps %xmm4,%xmm2 + shufps \$0b10001100,%xmm2,%xmm4 + xorps %xmm4,%xmm2 + shufps \$0b10101010,%xmm1,%xmm1 # critical path + xorps %xmm1,%xmm2 + ret +.size ${PREFIX}_set_encrypt_key,.-${PREFIX}_set_encrypt_key +.size __aesni_set_encrypt_key,.-__aesni_set_encrypt_key +___ +} + +$code.=<<___; +.align 64 +.Lbswap_mask: + .byte 15,14,13,12,11,10,9,8,7,6,5,4,3,2,1,0 +.Lincrement32: + .long 6,6,6,0 +.Lincrement64: + .long 1,0,0,0 +.Lxts_magic: + .long 0x87,0,1,0 +.Lincrement1: + .byte 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,1 +.Lkey_rotate: + .long 0x0c0f0e0d,0x0c0f0e0d,0x0c0f0e0d,0x0c0f0e0d +.Lkey_rotate192: + .long 0x04070605,0x04070605,0x04070605,0x04070605 +.Lkey_rcon1: + .long 1,1,1,1 +.Lkey_rcon1b: + .long 0x1b,0x1b,0x1b,0x1b + +.asciz "AES for Intel AES-NI, CRYPTOGAMS by " +.align 64 +___ + +# EXCEPTION_DISPOSITION handler (EXCEPTION_RECORD *rec,ULONG64 frame, +# CONTEXT *context,DISPATCHER_CONTEXT *disp) +if ($win64) { +$rec="%rcx"; +$frame="%rdx"; +$context="%r8"; +$disp="%r9"; + +$code.=<<___; +.extern __imp_RtlVirtualUnwind +___ +$code.=<<___ if ($PREFIX eq "aesni"); +.type ecb_ccm64_se_handler,\@abi-omnipotent +.align 16 +ecb_ccm64_se_handler: + push %rsi + push %rdi + push %rbx + push %rbp + push %r12 + push %r13 + push %r14 + push %r15 + pushfq + sub \$64,%rsp + + mov 120($context),%rax # pull context->Rax + mov 248($context),%rbx # pull context->Rip + + mov 8($disp),%rsi # disp->ImageBase + mov 56($disp),%r11 # disp->HandlerData + + mov 0(%r11),%r10d # HandlerData[0] + lea (%rsi,%r10),%r10 # prologue label + cmp %r10,%rbx # context->RipRsp + + mov 4(%r11),%r10d # HandlerData[1] + lea (%rsi,%r10),%r10 # epilogue label + cmp %r10,%rbx # context->Rip>=epilogue label + jae .Lcommon_seh_tail + + lea 0(%rax),%rsi # %xmm save area + lea 512($context),%rdi # &context.Xmm6 + mov \$8,%ecx # 4*sizeof(%xmm0)/sizeof(%rax) + .long 0xa548f3fc # cld; rep movsq + lea 0x58(%rax),%rax # adjust stack pointer + + jmp .Lcommon_seh_tail +.size ecb_ccm64_se_handler,.-ecb_ccm64_se_handler + +.type ctr_xts_se_handler,\@abi-omnipotent +.align 16 +ctr_xts_se_handler: + push %rsi + push %rdi + push %rbx + push %rbp + push %r12 + push %r13 + push %r14 + push %r15 + pushfq + sub \$64,%rsp + + mov 120($context),%rax # pull context->Rax + mov 248($context),%rbx # pull context->Rip + + mov 8($disp),%rsi # disp->ImageBase + mov 56($disp),%r11 # disp->HandlerData + + mov 0(%r11),%r10d # HandlerData[0] + lea (%rsi,%r10),%r10 # prologue lable + cmp %r10,%rbx # context->RipRsp + + mov 4(%r11),%r10d # HandlerData[1] + lea (%rsi,%r10),%r10 # epilogue label + cmp %r10,%rbx # context->Rip>=epilogue label + jae .Lcommon_seh_tail + + mov 160($context),%rax # pull context->Rbp + lea -0xa0(%rax),%rsi # %xmm save area + lea 512($context),%rdi # & context.Xmm6 + mov \$20,%ecx # 10*sizeof(%xmm0)/sizeof(%rax) + .long 0xa548f3fc # cld; rep movsq + + jmp .Lcommon_rbp_tail +.size ctr_xts_se_handler,.-ctr_xts_se_handler + +.type ocb_se_handler,\@abi-omnipotent +.align 16 +ocb_se_handler: + push %rsi + push %rdi + push %rbx + push %rbp + push %r12 + push %r13 + push %r14 + push %r15 + pushfq + sub \$64,%rsp + + mov 120($context),%rax # pull context->Rax + mov 248($context),%rbx # pull context->Rip + + mov 8($disp),%rsi # disp->ImageBase + mov 56($disp),%r11 # disp->HandlerData + + mov 0(%r11),%r10d # HandlerData[0] + lea (%rsi,%r10),%r10 # prologue lable + cmp %r10,%rbx # context->RipRip>=epilogue label + jae .Lcommon_seh_tail + + mov 8(%r11),%r10d # HandlerData[2] + lea (%rsi,%r10),%r10 + cmp %r10,%rbx # context->Rip>=pop label + jae .Locb_no_xmm + + mov 152($context),%rax # pull context->Rsp + + lea (%rax),%rsi # %xmm save area + lea 512($context),%rdi # & context.Xmm6 + mov \$20,%ecx # 10*sizeof(%xmm0)/sizeof(%rax) + .long 0xa548f3fc # cld; rep movsq + lea 0xa0+0x28(%rax),%rax + +.Locb_no_xmm: + mov -8(%rax),%rbx + mov -16(%rax),%rbp + mov -24(%rax),%r12 + mov -32(%rax),%r13 + mov -40(%rax),%r14 + + mov %rbx,144($context) # restore context->Rbx + mov %rbp,160($context) # restore context->Rbp + mov %r12,216($context) # restore context->R12 + mov %r13,224($context) # restore context->R13 + mov %r14,232($context) # restore context->R14 + + jmp .Lcommon_seh_tail +.size ocb_se_handler,.-ocb_se_handler +___ +$code.=<<___; +.type cbc_se_handler,\@abi-omnipotent +.align 16 +cbc_se_handler: + push %rsi + push %rdi + push %rbx + push %rbp + push %r12 + push %r13 + push %r14 + push %r15 + pushfq + sub \$64,%rsp + + mov 152($context),%rax # pull context->Rsp + mov 248($context),%rbx # pull context->Rip + + lea .Lcbc_decrypt_bulk(%rip),%r10 + cmp %r10,%rbx # context->Rip<"prologue" label + jb .Lcommon_seh_tail + + lea .Lcbc_decrypt_body(%rip),%r10 + cmp %r10,%rbx # context->RipRip>="epilogue" label + jae .Lcommon_seh_tail + + lea 16(%rax),%rsi # %xmm save area + lea 512($context),%rdi # &context.Xmm6 + mov \$20,%ecx # 10*sizeof(%xmm0)/sizeof(%rax) + .long 0xa548f3fc # cld; rep movsq + +.Lcommon_rbp_tail: + mov 160($context),%rax # pull context->Rbp + mov (%rax),%rbp # restore saved %rbp + lea 8(%rax),%rax # adjust stack pointer + mov %rbp,160($context) # restore context->Rbp + jmp .Lcommon_seh_tail + +.Lrestore_cbc_rax: + mov 120($context),%rax + +.Lcommon_seh_tail: + mov 8(%rax),%rdi + mov 16(%rax),%rsi + mov %rax,152($context) # restore context->Rsp + mov %rsi,168($context) # restore context->Rsi + mov %rdi,176($context) # restore context->Rdi + + mov 40($disp),%rdi # disp->ContextRecord + mov $context,%rsi # context + mov \$154,%ecx # sizeof(CONTEXT) + .long 0xa548f3fc # cld; rep movsq + + mov $disp,%rsi + xor %rcx,%rcx # arg1, UNW_FLAG_NHANDLER + mov 8(%rsi),%rdx # arg2, disp->ImageBase + mov 0(%rsi),%r8 # arg3, disp->ControlPc + mov 16(%rsi),%r9 # arg4, disp->FunctionEntry + mov 40(%rsi),%r10 # disp->ContextRecord + lea 56(%rsi),%r11 # &disp->HandlerData + lea 24(%rsi),%r12 # &disp->EstablisherFrame + mov %r10,32(%rsp) # arg5 + mov %r11,40(%rsp) # arg6 + mov %r12,48(%rsp) # arg7 + mov %rcx,56(%rsp) # arg8, (NULL) + call *__imp_RtlVirtualUnwind(%rip) + + mov \$1,%eax # ExceptionContinueSearch + add \$64,%rsp + popfq + pop %r15 + pop %r14 + pop %r13 + pop %r12 + pop %rbp + pop %rbx + pop %rdi + pop %rsi + ret +.size cbc_se_handler,.-cbc_se_handler + +.section .pdata +.align 4 +___ +$code.=<<___ if ($PREFIX eq "aesni"); + .rva .LSEH_begin_aesni_ecb_encrypt + .rva .LSEH_end_aesni_ecb_encrypt + .rva .LSEH_info_ecb + + .rva .LSEH_begin_aesni_ccm64_encrypt_blocks + .rva .LSEH_end_aesni_ccm64_encrypt_blocks + .rva .LSEH_info_ccm64_enc + + .rva .LSEH_begin_aesni_ccm64_decrypt_blocks + .rva .LSEH_end_aesni_ccm64_decrypt_blocks + .rva .LSEH_info_ccm64_dec + + .rva .LSEH_begin_aesni_ctr32_encrypt_blocks + .rva .LSEH_end_aesni_ctr32_encrypt_blocks + .rva .LSEH_info_ctr32 + + .rva .LSEH_begin_aesni_xts_encrypt + .rva .LSEH_end_aesni_xts_encrypt + .rva .LSEH_info_xts_enc + + .rva .LSEH_begin_aesni_xts_decrypt + .rva .LSEH_end_aesni_xts_decrypt + .rva .LSEH_info_xts_dec + + .rva .LSEH_begin_aesni_ocb_encrypt + .rva .LSEH_end_aesni_ocb_encrypt + .rva .LSEH_info_ocb_enc + + .rva .LSEH_begin_aesni_ocb_decrypt + .rva .LSEH_end_aesni_ocb_decrypt + .rva .LSEH_info_ocb_dec +___ +$code.=<<___; + .rva .LSEH_begin_${PREFIX}_cbc_encrypt + .rva .LSEH_end_${PREFIX}_cbc_encrypt + .rva .LSEH_info_cbc + + .rva ${PREFIX}_set_decrypt_key + .rva .LSEH_end_set_decrypt_key + .rva .LSEH_info_key + + .rva ${PREFIX}_set_encrypt_key + .rva .LSEH_end_set_encrypt_key + .rva .LSEH_info_key +.section .xdata +.align 8 +___ +$code.=<<___ if ($PREFIX eq "aesni"); +.LSEH_info_ecb: + .byte 9,0,0,0 + .rva ecb_ccm64_se_handler + .rva .Lecb_enc_body,.Lecb_enc_ret # HandlerData[] +.LSEH_info_ccm64_enc: + .byte 9,0,0,0 + .rva ecb_ccm64_se_handler + .rva .Lccm64_enc_body,.Lccm64_enc_ret # HandlerData[] +.LSEH_info_ccm64_dec: + .byte 9,0,0,0 + .rva ecb_ccm64_se_handler + .rva .Lccm64_dec_body,.Lccm64_dec_ret # HandlerData[] +.LSEH_info_ctr32: + .byte 9,0,0,0 + .rva ctr_xts_se_handler + .rva .Lctr32_body,.Lctr32_epilogue # HandlerData[] +.LSEH_info_xts_enc: + .byte 9,0,0,0 + .rva ctr_xts_se_handler + .rva .Lxts_enc_body,.Lxts_enc_epilogue # HandlerData[] +.LSEH_info_xts_dec: + .byte 9,0,0,0 + .rva ctr_xts_se_handler + .rva .Lxts_dec_body,.Lxts_dec_epilogue # HandlerData[] +.LSEH_info_ocb_enc: + .byte 9,0,0,0 + .rva ocb_se_handler + .rva .Locb_enc_body,.Locb_enc_epilogue # HandlerData[] + .rva .Locb_enc_pop + .long 0 +.LSEH_info_ocb_dec: + .byte 9,0,0,0 + .rva ocb_se_handler + .rva .Locb_dec_body,.Locb_dec_epilogue # HandlerData[] + .rva .Locb_dec_pop + .long 0 +___ +$code.=<<___; +.LSEH_info_cbc: + .byte 9,0,0,0 + .rva cbc_se_handler +.LSEH_info_key: + .byte 0x01,0x04,0x01,0x00 + .byte 0x04,0x02,0x00,0x00 # sub rsp,8 +___ +} + +sub rex { + local *opcode=shift; + my ($dst,$src)=@_; + my $rex=0; + + $rex|=0x04 if($dst>=8); + $rex|=0x01 if($src>=8); + push @opcode,$rex|0x40 if($rex); +} + +sub aesni { + my $line=shift; + my @opcode=(0x66); + + if ($line=~/(aeskeygenassist)\s+\$([x0-9a-f]+),\s*%xmm([0-9]+),\s*%xmm([0-9]+)/) { + rex(\@opcode,$4,$3); + push @opcode,0x0f,0x3a,0xdf; + push @opcode,0xc0|($3&7)|(($4&7)<<3); # ModR/M + my $c=$2; + push @opcode,$c=~/^0/?oct($c):$c; + return ".byte\t".join(',',@opcode); + } + elsif ($line=~/(aes[a-z]+)\s+%xmm([0-9]+),\s*%xmm([0-9]+)/) { + my %opcodelet = ( + "aesimc" => 0xdb, + "aesenc" => 0xdc, "aesenclast" => 0xdd, + "aesdec" => 0xde, "aesdeclast" => 0xdf + ); + return undef if (!defined($opcodelet{$1})); + rex(\@opcode,$3,$2); + push @opcode,0x0f,0x38,$opcodelet{$1}; + push @opcode,0xc0|($2&7)|(($3&7)<<3); # ModR/M + return ".byte\t".join(',',@opcode); + } + elsif ($line=~/(aes[a-z]+)\s+([0x1-9a-fA-F]*)\(%rsp\),\s*%xmm([0-9]+)/) { + my %opcodelet = ( + "aesenc" => 0xdc, "aesenclast" => 0xdd, + "aesdec" => 0xde, "aesdeclast" => 0xdf + ); + return undef if (!defined($opcodelet{$1})); + my $off = $2; + push @opcode,0x44 if ($3>=8); + push @opcode,0x0f,0x38,$opcodelet{$1}; + push @opcode,0x44|(($3&7)<<3),0x24; # ModR/M + push @opcode,($off=~/^0/?oct($off):$off)&0xff; + return ".byte\t".join(',',@opcode); + } + return $line; +} + +sub movbe { + ".byte 0x0f,0x38,0xf1,0x44,0x24,".shift; +} + +$code =~ s/\`([^\`]*)\`/eval($1)/gem; +$code =~ s/\b(aes.*%xmm[0-9]+).*$/aesni($1)/gem; +#$code =~ s/\bmovbe\s+%eax/bswap %eax; mov %eax/gm; # debugging artefact +$code =~ s/\bmovbe\s+%eax,\s*([0-9]+)\(%rsp\)/movbe($1)/gem; + +print $code; + +close STDOUT; diff --git a/openssl-1.1.0h/crypto/aes/asm/aesp8-ppc.pl b/openssl-1.1.0h/crypto/aes/asm/aesp8-ppc.pl new file mode 100755 index 0000000..b7e92f6 --- /dev/null +++ b/openssl-1.1.0h/crypto/aes/asm/aesp8-ppc.pl @@ -0,0 +1,3805 @@ +#! /usr/bin/env perl +# Copyright 2014-2016 The OpenSSL Project Authors. All Rights Reserved. +# +# Licensed under the OpenSSL license (the "License"). You may not use +# this file except in compliance with the License. You can obtain a copy +# in the file LICENSE in the source distribution or at +# https://www.openssl.org/source/license.html + +# +# ==================================================================== +# Written by Andy Polyakov for the OpenSSL +# project. The module is, however, dual licensed under OpenSSL and +# CRYPTOGAMS licenses depending on where you obtain it. For further +# details see http://www.openssl.org/~appro/cryptogams/. +# ==================================================================== +# +# This module implements support for AES instructions as per PowerISA +# specification version 2.07, first implemented by POWER8 processor. +# The module is endian-agnostic in sense that it supports both big- +# and little-endian cases. Data alignment in parallelizable modes is +# handled with VSX loads and stores, which implies MSR.VSX flag being +# set. It should also be noted that ISA specification doesn't prohibit +# alignment exceptions for these instructions on page boundaries. +# Initially alignment was handled in pure AltiVec/VMX way [when data +# is aligned programmatically, which in turn guarantees exception- +# free execution], but it turned to hamper performance when vcipher +# instructions are interleaved. It's reckoned that eventual +# misalignment penalties at page boundaries are in average lower +# than additional overhead in pure AltiVec approach. +# +# May 2016 +# +# Add XTS subroutine, 9x on little- and 12x improvement on big-endian +# systems were measured. +# +###################################################################### +# Current large-block performance in cycles per byte processed with +# 128-bit key (less is better). +# +# CBC en-/decrypt CTR XTS +# POWER8[le] 3.96/0.72 0.74 1.1 +# POWER8[be] 3.75/0.65 0.66 1.0 + +$flavour = shift; + +if ($flavour =~ /64/) { + $SIZE_T =8; + $LRSAVE =2*$SIZE_T; + $STU ="stdu"; + $POP ="ld"; + $PUSH ="std"; + $UCMP ="cmpld"; + $SHL ="sldi"; +} elsif ($flavour =~ /32/) { + $SIZE_T =4; + $LRSAVE =$SIZE_T; + $STU ="stwu"; + $POP ="lwz"; + $PUSH ="stw"; + $UCMP ="cmplw"; + $SHL ="slwi"; +} else { die "nonsense $flavour"; } + +$LITTLE_ENDIAN = ($flavour=~/le$/) ? $SIZE_T : 0; + +$0 =~ m/(.*[\/\\])[^\/\\]+$/; $dir=$1; +( $xlate="${dir}ppc-xlate.pl" and -f $xlate ) or +( $xlate="${dir}../../perlasm/ppc-xlate.pl" and -f $xlate) or +die "can't locate ppc-xlate.pl"; + +open STDOUT,"| $^X $xlate $flavour ".shift || die "can't call $xlate: $!"; + +$FRAME=8*$SIZE_T; +$prefix="aes_p8"; + +$sp="r1"; +$vrsave="r12"; + +######################################################################### +{{{ # Key setup procedures # +my ($inp,$bits,$out,$ptr,$cnt,$rounds)=map("r$_",(3..8)); +my ($zero,$in0,$in1,$key,$rcon,$mask,$tmp)=map("v$_",(0..6)); +my ($stage,$outperm,$outmask,$outhead,$outtail)=map("v$_",(7..11)); + +$code.=<<___; +.machine "any" + +.text + +.align 7 +rcon: +.long 0x01000000, 0x01000000, 0x01000000, 0x01000000 ?rev +.long 0x1b000000, 0x1b000000, 0x1b000000, 0x1b000000 ?rev +.long 0x0d0e0f0c, 0x0d0e0f0c, 0x0d0e0f0c, 0x0d0e0f0c ?rev +.long 0,0,0,0 ?asis +Lconsts: + mflr r0 + bcl 20,31,\$+4 + mflr $ptr #vvvvv "distance between . and rcon + addi $ptr,$ptr,-0x48 + mtlr r0 + blr + .long 0 + .byte 0,12,0x14,0,0,0,0,0 +.asciz "AES for PowerISA 2.07, CRYPTOGAMS by " + +.globl .${prefix}_set_encrypt_key +.align 5 +.${prefix}_set_encrypt_key: +Lset_encrypt_key: + mflr r11 + $PUSH r11,$LRSAVE($sp) + + li $ptr,-1 + ${UCMP}i $inp,0 + beq- Lenc_key_abort # if ($inp==0) return -1; + ${UCMP}i $out,0 + beq- Lenc_key_abort # if ($out==0) return -1; + li $ptr,-2 + cmpwi $bits,128 + blt- Lenc_key_abort + cmpwi $bits,256 + bgt- Lenc_key_abort + andi. r0,$bits,0x3f + bne- Lenc_key_abort + + lis r0,0xfff0 + mfspr $vrsave,256 + mtspr 256,r0 + + bl Lconsts + mtlr r11 + + neg r9,$inp + lvx $in0,0,$inp + addi $inp,$inp,15 # 15 is not typo + lvsr $key,0,r9 # borrow $key + li r8,0x20 + cmpwi $bits,192 + lvx $in1,0,$inp + le?vspltisb $mask,0x0f # borrow $mask + lvx $rcon,0,$ptr + le?vxor $key,$key,$mask # adjust for byte swap + lvx $mask,r8,$ptr + addi $ptr,$ptr,0x10 + vperm $in0,$in0,$in1,$key # align [and byte swap in LE] + li $cnt,8 + vxor $zero,$zero,$zero + mtctr $cnt + + ?lvsr $outperm,0,$out + vspltisb $outmask,-1 + lvx $outhead,0,$out + ?vperm $outmask,$zero,$outmask,$outperm + + blt Loop128 + addi $inp,$inp,8 + beq L192 + addi $inp,$inp,8 + b L256 + +.align 4 +Loop128: + vperm $key,$in0,$in0,$mask # rotate-n-splat + vsldoi $tmp,$zero,$in0,12 # >>32 + vperm $outtail,$in0,$in0,$outperm # rotate + vsel $stage,$outhead,$outtail,$outmask + vmr $outhead,$outtail + vcipherlast $key,$key,$rcon + stvx $stage,0,$out + addi $out,$out,16 + + vxor $in0,$in0,$tmp + vsldoi $tmp,$zero,$tmp,12 # >>32 + vxor $in0,$in0,$tmp + vsldoi $tmp,$zero,$tmp,12 # >>32 + vxor $in0,$in0,$tmp + vadduwm $rcon,$rcon,$rcon + vxor $in0,$in0,$key + bdnz Loop128 + + lvx $rcon,0,$ptr # last two round keys + + vperm $key,$in0,$in0,$mask # rotate-n-splat + vsldoi $tmp,$zero,$in0,12 # >>32 + vperm $outtail,$in0,$in0,$outperm # rotate + vsel $stage,$outhead,$outtail,$outmask + vmr $outhead,$outtail + vcipherlast $key,$key,$rcon + stvx $stage,0,$out + addi $out,$out,16 + + vxor $in0,$in0,$tmp + vsldoi $tmp,$zero,$tmp,12 # >>32 + vxor $in0,$in0,$tmp + vsldoi $tmp,$zero,$tmp,12 # >>32 + vxor $in0,$in0,$tmp + vadduwm $rcon,$rcon,$rcon + vxor $in0,$in0,$key + + vperm $key,$in0,$in0,$mask # rotate-n-splat + vsldoi $tmp,$zero,$in0,12 # >>32 + vperm $outtail,$in0,$in0,$outperm # rotate + vsel $stage,$outhead,$outtail,$outmask + vmr $outhead,$outtail + vcipherlast $key,$key,$rcon + stvx $stage,0,$out + addi $out,$out,16 + + vxor $in0,$in0,$tmp + vsldoi $tmp,$zero,$tmp,12 # >>32 + vxor $in0,$in0,$tmp + vsldoi $tmp,$zero,$tmp,12 # >>32 + vxor $in0,$in0,$tmp + vxor $in0,$in0,$key + vperm $outtail,$in0,$in0,$outperm # rotate + vsel $stage,$outhead,$outtail,$outmask + vmr $outhead,$outtail + stvx $stage,0,$out + + addi $inp,$out,15 # 15 is not typo + addi $out,$out,0x50 + + li $rounds,10 + b Ldone + +.align 4 +L192: + lvx $tmp,0,$inp + li $cnt,4 + vperm $outtail,$in0,$in0,$outperm # rotate + vsel $stage,$outhead,$outtail,$outmask + vmr $outhead,$outtail + stvx $stage,0,$out + addi $out,$out,16 + vperm $in1,$in1,$tmp,$key # align [and byte swap in LE] + vspltisb $key,8 # borrow $key + mtctr $cnt + vsububm $mask,$mask,$key # adjust the mask + +Loop192: + vperm $key,$in1,$in1,$mask # roate-n-splat + vsldoi $tmp,$zero,$in0,12 # >>32 + vcipherlast $key,$key,$rcon + + vxor $in0,$in0,$tmp + vsldoi $tmp,$zero,$tmp,12 # >>32 + vxor $in0,$in0,$tmp + vsldoi $tmp,$zero,$tmp,12 # >>32 + vxor $in0,$in0,$tmp + + vsldoi $stage,$zero,$in1,8 + vspltw $tmp,$in0,3 + vxor $tmp,$tmp,$in1 + vsldoi $in1,$zero,$in1,12 # >>32 + vadduwm $rcon,$rcon,$rcon + vxor $in1,$in1,$tmp + vxor $in0,$in0,$key + vxor $in1,$in1,$key + vsldoi $stage,$stage,$in0,8 + + vperm $key,$in1,$in1,$mask # rotate-n-splat + vsldoi $tmp,$zero,$in0,12 # >>32 + vperm $outtail,$stage,$stage,$outperm # rotate + vsel $stage,$outhead,$outtail,$outmask + vmr $outhead,$outtail + vcipherlast $key,$key,$rcon + stvx $stage,0,$out + addi $out,$out,16 + + vsldoi $stage,$in0,$in1,8 + vxor $in0,$in0,$tmp + vsldoi $tmp,$zero,$tmp,12 # >>32 + vperm $outtail,$stage,$stage,$outperm # rotate + vsel $stage,$outhead,$outtail,$outmask + vmr $outhead,$outtail + vxor $in0,$in0,$tmp + vsldoi $tmp,$zero,$tmp,12 # >>32 + vxor $in0,$in0,$tmp + stvx $stage,0,$out + addi $out,$out,16 + + vspltw $tmp,$in0,3 + vxor $tmp,$tmp,$in1 + vsldoi $in1,$zero,$in1,12 # >>32 + vadduwm $rcon,$rcon,$rcon + vxor $in1,$in1,$tmp + vxor $in0,$in0,$key + vxor $in1,$in1,$key + vperm $outtail,$in0,$in0,$outperm # rotate + vsel $stage,$outhead,$outtail,$outmask + vmr $outhead,$outtail + stvx $stage,0,$out + addi $inp,$out,15 # 15 is not typo + addi $out,$out,16 + bdnz Loop192 + + li $rounds,12 + addi $out,$out,0x20 + b Ldone + +.align 4 +L256: + lvx $tmp,0,$inp + li $cnt,7 + li $rounds,14 + vperm $outtail,$in0,$in0,$outperm # rotate + vsel $stage,$outhead,$outtail,$outmask + vmr $outhead,$outtail + stvx $stage,0,$out + addi $out,$out,16 + vperm $in1,$in1,$tmp,$key # align [and byte swap in LE] + mtctr $cnt + +Loop256: + vperm $key,$in1,$in1,$mask # rotate-n-splat + vsldoi $tmp,$zero,$in0,12 # >>32 + vperm $outtail,$in1,$in1,$outperm # rotate + vsel $stage,$outhead,$outtail,$outmask + vmr $outhead,$outtail + vcipherlast $key,$key,$rcon + stvx $stage,0,$out + addi $out,$out,16 + + vxor $in0,$in0,$tmp + vsldoi $tmp,$zero,$tmp,12 # >>32 + vxor $in0,$in0,$tmp + vsldoi $tmp,$zero,$tmp,12 # >>32 + vxor $in0,$in0,$tmp + vadduwm $rcon,$rcon,$rcon + vxor $in0,$in0,$key + vperm $outtail,$in0,$in0,$outperm # rotate + vsel $stage,$outhead,$outtail,$outmask + vmr $outhead,$outtail + stvx $stage,0,$out + addi $inp,$out,15 # 15 is not typo + addi $out,$out,16 + bdz Ldone + + vspltw $key,$in0,3 # just splat + vsldoi $tmp,$zero,$in1,12 # >>32 + vsbox $key,$key + + vxor $in1,$in1,$tmp + vsldoi $tmp,$zero,$tmp,12 # >>32 + vxor $in1,$in1,$tmp + vsldoi $tmp,$zero,$tmp,12 # >>32 + vxor $in1,$in1,$tmp + + vxor $in1,$in1,$key + b Loop256 + +.align 4 +Ldone: + lvx $in1,0,$inp # redundant in aligned case + vsel $in1,$outhead,$in1,$outmask + stvx $in1,0,$inp + li $ptr,0 + mtspr 256,$vrsave + stw $rounds,0($out) + +Lenc_key_abort: + mr r3,$ptr + blr + .long 0 + .byte 0,12,0x14,1,0,0,3,0 + .long 0 +.size .${prefix}_set_encrypt_key,.-.${prefix}_set_encrypt_key + +.globl .${prefix}_set_decrypt_key +.align 5 +.${prefix}_set_decrypt_key: + $STU $sp,-$FRAME($sp) + mflr r10 + $PUSH r10,$FRAME+$LRSAVE($sp) + bl Lset_encrypt_key + mtlr r10 + + cmpwi r3,0 + bne- Ldec_key_abort + + slwi $cnt,$rounds,4 + subi $inp,$out,240 # first round key + srwi $rounds,$rounds,1 + add $out,$inp,$cnt # last round key + mtctr $rounds + +Ldeckey: + lwz r0, 0($inp) + lwz r6, 4($inp) + lwz r7, 8($inp) + lwz r8, 12($inp) + addi $inp,$inp,16 + lwz r9, 0($out) + lwz r10,4($out) + lwz r11,8($out) + lwz r12,12($out) + stw r0, 0($out) + stw r6, 4($out) + stw r7, 8($out) + stw r8, 12($out) + subi $out,$out,16 + stw r9, -16($inp) + stw r10,-12($inp) + stw r11,-8($inp) + stw r12,-4($inp) + bdnz Ldeckey + + xor r3,r3,r3 # return value +Ldec_key_abort: + addi $sp,$sp,$FRAME + blr + .long 0 + .byte 0,12,4,1,0x80,0,3,0 + .long 0 +.size .${prefix}_set_decrypt_key,.-.${prefix}_set_decrypt_key +___ +}}} +######################################################################### +{{{ # Single block en- and decrypt procedures # +sub gen_block () { +my $dir = shift; +my $n = $dir eq "de" ? "n" : ""; +my ($inp,$out,$key,$rounds,$idx)=map("r$_",(3..7)); + +$code.=<<___; +.globl .${prefix}_${dir}crypt +.align 5 +.${prefix}_${dir}crypt: + lwz $rounds,240($key) + lis r0,0xfc00 + mfspr $vrsave,256 + li $idx,15 # 15 is not typo + mtspr 256,r0 + + lvx v0,0,$inp + neg r11,$out + lvx v1,$idx,$inp + lvsl v2,0,$inp # inpperm + le?vspltisb v4,0x0f + ?lvsl v3,0,r11 # outperm + le?vxor v2,v2,v4 + li $idx,16 + vperm v0,v0,v1,v2 # align [and byte swap in LE] + lvx v1,0,$key + ?lvsl v5,0,$key # keyperm + srwi $rounds,$rounds,1 + lvx v2,$idx,$key + addi $idx,$idx,16 + subi $rounds,$rounds,1 + ?vperm v1,v1,v2,v5 # align round key + + vxor v0,v0,v1 + lvx v1,$idx,$key + addi $idx,$idx,16 + mtctr $rounds + +Loop_${dir}c: + ?vperm v2,v2,v1,v5 + v${n}cipher v0,v0,v2 + lvx v2,$idx,$key + addi $idx,$idx,16 + ?vperm v1,v1,v2,v5 + v${n}cipher v0,v0,v1 + lvx v1,$idx,$key + addi $idx,$idx,16 + bdnz Loop_${dir}c + + ?vperm v2,v2,v1,v5 + v${n}cipher v0,v0,v2 + lvx v2,$idx,$key + ?vperm v1,v1,v2,v5 + v${n}cipherlast v0,v0,v1 + + vspltisb v2,-1 + vxor v1,v1,v1 + li $idx,15 # 15 is not typo + ?vperm v2,v1,v2,v3 # outmask + le?vxor v3,v3,v4 + lvx v1,0,$out # outhead + vperm v0,v0,v0,v3 # rotate [and byte swap in LE] + vsel v1,v1,v0,v2 + lvx v4,$idx,$out + stvx v1,0,$out + vsel v0,v0,v4,v2 + stvx v0,$idx,$out + + mtspr 256,$vrsave + blr + .long 0 + .byte 0,12,0x14,0,0,0,3,0 + .long 0 +.size .${prefix}_${dir}crypt,.-.${prefix}_${dir}crypt +___ +} +&gen_block("en"); +&gen_block("de"); +}}} +######################################################################### +{{{ # CBC en- and decrypt procedures # +my ($inp,$out,$len,$key,$ivp,$enc,$rounds,$idx)=map("r$_",(3..10)); +my ($rndkey0,$rndkey1,$inout,$tmp)= map("v$_",(0..3)); +my ($ivec,$inptail,$inpperm,$outhead,$outperm,$outmask,$keyperm)= + map("v$_",(4..10)); +$code.=<<___; +.globl .${prefix}_cbc_encrypt +.align 5 +.${prefix}_cbc_encrypt: + ${UCMP}i $len,16 + bltlr- + + cmpwi $enc,0 # test direction + lis r0,0xffe0 + mfspr $vrsave,256 + mtspr 256,r0 + + li $idx,15 + vxor $rndkey0,$rndkey0,$rndkey0 + le?vspltisb $tmp,0x0f + + lvx $ivec,0,$ivp # load [unaligned] iv + lvsl $inpperm,0,$ivp + lvx $inptail,$idx,$ivp + le?vxor $inpperm,$inpperm,$tmp + vperm $ivec,$ivec,$inptail,$inpperm + + neg r11,$inp + ?lvsl $keyperm,0,$key # prepare for unaligned key + lwz $rounds,240($key) + + lvsr $inpperm,0,r11 # prepare for unaligned load + lvx $inptail,0,$inp + addi $inp,$inp,15 # 15 is not typo + le?vxor $inpperm,$inpperm,$tmp + + ?lvsr $outperm,0,$out # prepare for unaligned store + vspltisb $outmask,-1 + lvx $outhead,0,$out + ?vperm $outmask,$rndkey0,$outmask,$outperm + le?vxor $outperm,$outperm,$tmp + + srwi $rounds,$rounds,1 + li $idx,16 + subi $rounds,$rounds,1 + beq Lcbc_dec + +Lcbc_enc: + vmr $inout,$inptail + lvx $inptail,0,$inp + addi $inp,$inp,16 + mtctr $rounds + subi $len,$len,16 # len-=16 + + lvx $rndkey0,0,$key + vperm $inout,$inout,$inptail,$inpperm + lvx $rndkey1,$idx,$key + addi $idx,$idx,16 + ?vperm $rndkey0,$rndkey0,$rndkey1,$keyperm + vxor $inout,$inout,$rndkey0 + lvx $rndkey0,$idx,$key + addi $idx,$idx,16 + vxor $inout,$inout,$ivec + +Loop_cbc_enc: + ?vperm $rndkey1,$rndkey1,$rndkey0,$keyperm + vcipher $inout,$inout,$rndkey1 + lvx $rndkey1,$idx,$key + addi $idx,$idx,16 + ?vperm $rndkey0,$rndkey0,$rndkey1,$keyperm + vcipher $inout,$inout,$rndkey0 + lvx $rndkey0,$idx,$key + addi $idx,$idx,16 + bdnz Loop_cbc_enc + + ?vperm $rndkey1,$rndkey1,$rndkey0,$keyperm + vcipher $inout,$inout,$rndkey1 + lvx $rndkey1,$idx,$key + li $idx,16 + ?vperm $rndkey0,$rndkey0,$rndkey1,$keyperm + vcipherlast $ivec,$inout,$rndkey0 + ${UCMP}i $len,16 + + vperm $tmp,$ivec,$ivec,$outperm + vsel $inout,$outhead,$tmp,$outmask + vmr $outhead,$tmp + stvx $inout,0,$out + addi $out,$out,16 + bge Lcbc_enc + + b Lcbc_done + +.align 4 +Lcbc_dec: + ${UCMP}i $len,128 + bge _aesp8_cbc_decrypt8x + vmr $tmp,$inptail + lvx $inptail,0,$inp + addi $inp,$inp,16 + mtctr $rounds + subi $len,$len,16 # len-=16 + + lvx $rndkey0,0,$key + vperm $tmp,$tmp,$inptail,$inpperm + lvx $rndkey1,$idx,$key + addi $idx,$idx,16 + ?vperm $rndkey0,$rndkey0,$rndkey1,$keyperm + vxor $inout,$tmp,$rndkey0 + lvx $rndkey0,$idx,$key + addi $idx,$idx,16 + +Loop_cbc_dec: + ?vperm $rndkey1,$rndkey1,$rndkey0,$keyperm + vncipher $inout,$inout,$rndkey1 + lvx $rndkey1,$idx,$key + addi $idx,$idx,16 + ?vperm $rndkey0,$rndkey0,$rndkey1,$keyperm + vncipher $inout,$inout,$rndkey0 + lvx $rndkey0,$idx,$key + addi $idx,$idx,16 + bdnz Loop_cbc_dec + + ?vperm $rndkey1,$rndkey1,$rndkey0,$keyperm + vncipher $inout,$inout,$rndkey1 + lvx $rndkey1,$idx,$key + li $idx,16 + ?vperm $rndkey0,$rndkey0,$rndkey1,$keyperm + vncipherlast $inout,$inout,$rndkey0 + ${UCMP}i $len,16 + + vxor $inout,$inout,$ivec + vmr $ivec,$tmp + vperm $tmp,$inout,$inout,$outperm + vsel $inout,$outhead,$tmp,$outmask + vmr $outhead,$tmp + stvx $inout,0,$out + addi $out,$out,16 + bge Lcbc_dec + +Lcbc_done: + addi $out,$out,-1 + lvx $inout,0,$out # redundant in aligned case + vsel $inout,$outhead,$inout,$outmask + stvx $inout,0,$out + + neg $enc,$ivp # write [unaligned] iv + li $idx,15 # 15 is not typo + vxor $rndkey0,$rndkey0,$rndkey0 + vspltisb $outmask,-1 + le?vspltisb $tmp,0x0f + ?lvsl $outperm,0,$enc + ?vperm $outmask,$rndkey0,$outmask,$outperm + le?vxor $outperm,$outperm,$tmp + lvx $outhead,0,$ivp + vperm $ivec,$ivec,$ivec,$outperm + vsel $inout,$outhead,$ivec,$outmask + lvx $inptail,$idx,$ivp + stvx $inout,0,$ivp + vsel $inout,$ivec,$inptail,$outmask + stvx $inout,$idx,$ivp + + mtspr 256,$vrsave + blr + .long 0 + .byte 0,12,0x14,0,0,0,6,0 + .long 0 +___ +######################################################################### +{{ # Optimized CBC decrypt procedure # +my $key_="r11"; +my ($x00,$x10,$x20,$x30,$x40,$x50,$x60,$x70)=map("r$_",(0,8,26..31)); + $x00=0 if ($flavour =~ /osx/); +my ($in0, $in1, $in2, $in3, $in4, $in5, $in6, $in7 )=map("v$_",(0..3,10..13)); +my ($out0,$out1,$out2,$out3,$out4,$out5,$out6,$out7)=map("v$_",(14..21)); +my $rndkey0="v23"; # v24-v25 rotating buffer for first found keys + # v26-v31 last 6 round keys +my ($tmp,$keyperm)=($in3,$in4); # aliases with "caller", redundant assignment + +$code.=<<___; +.align 5 +_aesp8_cbc_decrypt8x: + $STU $sp,-`($FRAME+21*16+6*$SIZE_T)`($sp) + li r10,`$FRAME+8*16+15` + li r11,`$FRAME+8*16+31` + stvx v20,r10,$sp # ABI says so + addi r10,r10,32 + stvx v21,r11,$sp + addi r11,r11,32 + stvx v22,r10,$sp + addi r10,r10,32 + stvx v23,r11,$sp + addi r11,r11,32 + stvx v24,r10,$sp + addi r10,r10,32 + stvx v25,r11,$sp + addi r11,r11,32 + stvx v26,r10,$sp + addi r10,r10,32 + stvx v27,r11,$sp + addi r11,r11,32 + stvx v28,r10,$sp + addi r10,r10,32 + stvx v29,r11,$sp + addi r11,r11,32 + stvx v30,r10,$sp + stvx v31,r11,$sp + li r0,-1 + stw $vrsave,`$FRAME+21*16-4`($sp) # save vrsave + li $x10,0x10 + $PUSH r26,`$FRAME+21*16+0*$SIZE_T`($sp) + li $x20,0x20 + $PUSH r27,`$FRAME+21*16+1*$SIZE_T`($sp) + li $x30,0x30 + $PUSH r28,`$FRAME+21*16+2*$SIZE_T`($sp) + li $x40,0x40 + $PUSH r29,`$FRAME+21*16+3*$SIZE_T`($sp) + li $x50,0x50 + $PUSH r30,`$FRAME+21*16+4*$SIZE_T`($sp) + li $x60,0x60 + $PUSH r31,`$FRAME+21*16+5*$SIZE_T`($sp) + li $x70,0x70 + mtspr 256,r0 + + subi $rounds,$rounds,3 # -4 in total + subi $len,$len,128 # bias + + lvx $rndkey0,$x00,$key # load key schedule + lvx v30,$x10,$key + addi $key,$key,0x20 + lvx v31,$x00,$key + ?vperm $rndkey0,$rndkey0,v30,$keyperm + addi $key_,$sp,$FRAME+15 + mtctr $rounds + +Load_cbc_dec_key: + ?vperm v24,v30,v31,$keyperm + lvx v30,$x10,$key + addi $key,$key,0x20 + stvx v24,$x00,$key_ # off-load round[1] + ?vperm v25,v31,v30,$keyperm + lvx v31,$x00,$key + stvx v25,$x10,$key_ # off-load round[2] + addi $key_,$key_,0x20 + bdnz Load_cbc_dec_key + + lvx v26,$x10,$key + ?vperm v24,v30,v31,$keyperm + lvx v27,$x20,$key + stvx v24,$x00,$key_ # off-load round[3] + ?vperm v25,v31,v26,$keyperm + lvx v28,$x30,$key + stvx v25,$x10,$key_ # off-load round[4] + addi $key_,$sp,$FRAME+15 # rewind $key_ + ?vperm v26,v26,v27,$keyperm + lvx v29,$x40,$key + ?vperm v27,v27,v28,$keyperm + lvx v30,$x50,$key + ?vperm v28,v28,v29,$keyperm + lvx v31,$x60,$key + ?vperm v29,v29,v30,$keyperm + lvx $out0,$x70,$key # borrow $out0 + ?vperm v30,v30,v31,$keyperm + lvx v24,$x00,$key_ # pre-load round[1] + ?vperm v31,v31,$out0,$keyperm + lvx v25,$x10,$key_ # pre-load round[2] + + #lvx $inptail,0,$inp # "caller" already did this + #addi $inp,$inp,15 # 15 is not typo + subi $inp,$inp,15 # undo "caller" + + le?li $idx,8 + lvx_u $in0,$x00,$inp # load first 8 "words" + le?lvsl $inpperm,0,$idx + le?vspltisb $tmp,0x0f + lvx_u $in1,$x10,$inp + le?vxor $inpperm,$inpperm,$tmp # transform for lvx_u/stvx_u + lvx_u $in2,$x20,$inp + le?vperm $in0,$in0,$in0,$inpperm + lvx_u $in3,$x30,$inp + le?vperm $in1,$in1,$in1,$inpperm + lvx_u $in4,$x40,$inp + le?vperm $in2,$in2,$in2,$inpperm + vxor $out0,$in0,$rndkey0 + lvx_u $in5,$x50,$inp + le?vperm $in3,$in3,$in3,$inpperm + vxor $out1,$in1,$rndkey0 + lvx_u $in6,$x60,$inp + le?vperm $in4,$in4,$in4,$inpperm + vxor $out2,$in2,$rndkey0 + lvx_u $in7,$x70,$inp + addi $inp,$inp,0x80 + le?vperm $in5,$in5,$in5,$inpperm + vxor $out3,$in3,$rndkey0 + le?vperm $in6,$in6,$in6,$inpperm + vxor $out4,$in4,$rndkey0 + le?vperm $in7,$in7,$in7,$inpperm + vxor $out5,$in5,$rndkey0 + vxor $out6,$in6,$rndkey0 + vxor $out7,$in7,$rndkey0 + + mtctr $rounds + b Loop_cbc_dec8x +.align 5 +Loop_cbc_dec8x: + vncipher $out0,$out0,v24 + vncipher $out1,$out1,v24 + vncipher $out2,$out2,v24 + vncipher $out3,$out3,v24 + vncipher $out4,$out4,v24 + vncipher $out5,$out5,v24 + vncipher $out6,$out6,v24 + vncipher $out7,$out7,v24 + lvx v24,$x20,$key_ # round[3] + addi $key_,$key_,0x20 + + vncipher $out0,$out0,v25 + vncipher $out1,$out1,v25 + vncipher $out2,$out2,v25 + vncipher $out3,$out3,v25 + vncipher $out4,$out4,v25 + vncipher $out5,$out5,v25 + vncipher $out6,$out6,v25 + vncipher $out7,$out7,v25 + lvx v25,$x10,$key_ # round[4] + bdnz Loop_cbc_dec8x + + subic $len,$len,128 # $len-=128 + vncipher $out0,$out0,v24 + vncipher $out1,$out1,v24 + vncipher $out2,$out2,v24 + vncipher $out3,$out3,v24 + vncipher $out4,$out4,v24 + vncipher $out5,$out5,v24 + vncipher $out6,$out6,v24 + vncipher $out7,$out7,v24 + + subfe. r0,r0,r0 # borrow?-1:0 + vncipher $out0,$out0,v25 + vncipher $out1,$out1,v25 + vncipher $out2,$out2,v25 + vncipher $out3,$out3,v25 + vncipher $out4,$out4,v25 + vncipher $out5,$out5,v25 + vncipher $out6,$out6,v25 + vncipher $out7,$out7,v25 + + and r0,r0,$len + vncipher $out0,$out0,v26 + vncipher $out1,$out1,v26 + vncipher $out2,$out2,v26 + vncipher $out3,$out3,v26 + vncipher $out4,$out4,v26 + vncipher $out5,$out5,v26 + vncipher $out6,$out6,v26 + vncipher $out7,$out7,v26 + + add $inp,$inp,r0 # $inp is adjusted in such + # way that at exit from the + # loop inX-in7 are loaded + # with last "words" + vncipher $out0,$out0,v27 + vncipher $out1,$out1,v27 + vncipher $out2,$out2,v27 + vncipher $out3,$out3,v27 + vncipher $out4,$out4,v27 + vncipher $out5,$out5,v27 + vncipher $out6,$out6,v27 + vncipher $out7,$out7,v27 + + addi $key_,$sp,$FRAME+15 # rewind $key_ + vncipher $out0,$out0,v28 + vncipher $out1,$out1,v28 + vncipher $out2,$out2,v28 + vncipher $out3,$out3,v28 + vncipher $out4,$out4,v28 + vncipher $out5,$out5,v28 + vncipher $out6,$out6,v28 + vncipher $out7,$out7,v28 + lvx v24,$x00,$key_ # re-pre-load round[1] + + vncipher $out0,$out0,v29 + vncipher $out1,$out1,v29 + vncipher $out2,$out2,v29 + vncipher $out3,$out3,v29 + vncipher $out4,$out4,v29 + vncipher $out5,$out5,v29 + vncipher $out6,$out6,v29 + vncipher $out7,$out7,v29 + lvx v25,$x10,$key_ # re-pre-load round[2] + + vncipher $out0,$out0,v30 + vxor $ivec,$ivec,v31 # xor with last round key + vncipher $out1,$out1,v30 + vxor $in0,$in0,v31 + vncipher $out2,$out2,v30 + vxor $in1,$in1,v31 + vncipher $out3,$out3,v30 + vxor $in2,$in2,v31 + vncipher $out4,$out4,v30 + vxor $in3,$in3,v31 + vncipher $out5,$out5,v30 + vxor $in4,$in4,v31 + vncipher $out6,$out6,v30 + vxor $in5,$in5,v31 + vncipher $out7,$out7,v30 + vxor $in6,$in6,v31 + + vncipherlast $out0,$out0,$ivec + vncipherlast $out1,$out1,$in0 + lvx_u $in0,$x00,$inp # load next input block + vncipherlast $out2,$out2,$in1 + lvx_u $in1,$x10,$inp + vncipherlast $out3,$out3,$in2 + le?vperm $in0,$in0,$in0,$inpperm + lvx_u $in2,$x20,$inp + vncipherlast $out4,$out4,$in3 + le?vperm $in1,$in1,$in1,$inpperm + lvx_u $in3,$x30,$inp + vncipherlast $out5,$out5,$in4 + le?vperm $in2,$in2,$in2,$inpperm + lvx_u $in4,$x40,$inp + vncipherlast $out6,$out6,$in5 + le?vperm $in3,$in3,$in3,$inpperm + lvx_u $in5,$x50,$inp + vncipherlast $out7,$out7,$in6 + le?vperm $in4,$in4,$in4,$inpperm + lvx_u $in6,$x60,$inp + vmr $ivec,$in7 + le?vperm $in5,$in5,$in5,$inpperm + lvx_u $in7,$x70,$inp + addi $inp,$inp,0x80 + + le?vperm $out0,$out0,$out0,$inpperm + le?vperm $out1,$out1,$out1,$inpperm + stvx_u $out0,$x00,$out + le?vperm $in6,$in6,$in6,$inpperm + vxor $out0,$in0,$rndkey0 + le?vperm $out2,$out2,$out2,$inpperm + stvx_u $out1,$x10,$out + le?vperm $in7,$in7,$in7,$inpperm + vxor $out1,$in1,$rndkey0 + le?vperm $out3,$out3,$out3,$inpperm + stvx_u $out2,$x20,$out + vxor $out2,$in2,$rndkey0 + le?vperm $out4,$out4,$out4,$inpperm + stvx_u $out3,$x30,$out + vxor $out3,$in3,$rndkey0 + le?vperm $out5,$out5,$out5,$inpperm + stvx_u $out4,$x40,$out + vxor $out4,$in4,$rndkey0 + le?vperm $out6,$out6,$out6,$inpperm + stvx_u $out5,$x50,$out + vxor $out5,$in5,$rndkey0 + le?vperm $out7,$out7,$out7,$inpperm + stvx_u $out6,$x60,$out + vxor $out6,$in6,$rndkey0 + stvx_u $out7,$x70,$out + addi $out,$out,0x80 + vxor $out7,$in7,$rndkey0 + + mtctr $rounds + beq Loop_cbc_dec8x # did $len-=128 borrow? + + addic. $len,$len,128 + beq Lcbc_dec8x_done + nop + nop + +Loop_cbc_dec8x_tail: # up to 7 "words" tail... + vncipher $out1,$out1,v24 + vncipher $out2,$out2,v24 + vncipher $out3,$out3,v24 + vncipher $out4,$out4,v24 + vncipher $out5,$out5,v24 + vncipher $out6,$out6,v24 + vncipher $out7,$out7,v24 + lvx v24,$x20,$key_ # round[3] + addi $key_,$key_,0x20 + + vncipher $out1,$out1,v25 + vncipher $out2,$out2,v25 + vncipher $out3,$out3,v25 + vncipher $out4,$out4,v25 + vncipher $out5,$out5,v25 + vncipher $out6,$out6,v25 + vncipher $out7,$out7,v25 + lvx v25,$x10,$key_ # round[4] + bdnz Loop_cbc_dec8x_tail + + vncipher $out1,$out1,v24 + vncipher $out2,$out2,v24 + vncipher $out3,$out3,v24 + vncipher $out4,$out4,v24 + vncipher $out5,$out5,v24 + vncipher $out6,$out6,v24 + vncipher $out7,$out7,v24 + + vncipher $out1,$out1,v25 + vncipher $out2,$out2,v25 + vncipher $out3,$out3,v25 + vncipher $out4,$out4,v25 + vncipher $out5,$out5,v25 + vncipher $out6,$out6,v25 + vncipher $out7,$out7,v25 + + vncipher $out1,$out1,v26 + vncipher $out2,$out2,v26 + vncipher $out3,$out3,v26 + vncipher $out4,$out4,v26 + vncipher $out5,$out5,v26 + vncipher $out6,$out6,v26 + vncipher $out7,$out7,v26 + + vncipher $out1,$out1,v27 + vncipher $out2,$out2,v27 + vncipher $out3,$out3,v27 + vncipher $out4,$out4,v27 + vncipher $out5,$out5,v27 + vncipher $out6,$out6,v27 + vncipher $out7,$out7,v27 + + vncipher $out1,$out1,v28 + vncipher $out2,$out2,v28 + vncipher $out3,$out3,v28 + vncipher $out4,$out4,v28 + vncipher $out5,$out5,v28 + vncipher $out6,$out6,v28 + vncipher $out7,$out7,v28 + + vncipher $out1,$out1,v29 + vncipher $out2,$out2,v29 + vncipher $out3,$out3,v29 + vncipher $out4,$out4,v29 + vncipher $out5,$out5,v29 + vncipher $out6,$out6,v29 + vncipher $out7,$out7,v29 + + vncipher $out1,$out1,v30 + vxor $ivec,$ivec,v31 # last round key + vncipher $out2,$out2,v30 + vxor $in1,$in1,v31 + vncipher $out3,$out3,v30 + vxor $in2,$in2,v31 + vncipher $out4,$out4,v30 + vxor $in3,$in3,v31 + vncipher $out5,$out5,v30 + vxor $in4,$in4,v31 + vncipher $out6,$out6,v30 + vxor $in5,$in5,v31 + vncipher $out7,$out7,v30 + vxor $in6,$in6,v31 + + cmplwi $len,32 # switch($len) + blt Lcbc_dec8x_one + nop + beq Lcbc_dec8x_two + cmplwi $len,64 + blt Lcbc_dec8x_three + nop + beq Lcbc_dec8x_four + cmplwi $len,96 + blt Lcbc_dec8x_five + nop + beq Lcbc_dec8x_six + +Lcbc_dec8x_seven: + vncipherlast $out1,$out1,$ivec + vncipherlast $out2,$out2,$in1 + vncipherlast $out3,$out3,$in2 + vncipherlast $out4,$out4,$in3 + vncipherlast $out5,$out5,$in4 + vncipherlast $out6,$out6,$in5 + vncipherlast $out7,$out7,$in6 + vmr $ivec,$in7 + + le?vperm $out1,$out1,$out1,$inpperm + le?vperm $out2,$out2,$out2,$inpperm + stvx_u $out1,$x00,$out + le?vperm $out3,$out3,$out3,$inpperm + stvx_u $out2,$x10,$out + le?vperm $out4,$out4,$out4,$inpperm + stvx_u $out3,$x20,$out + le?vperm $out5,$out5,$out5,$inpperm + stvx_u $out4,$x30,$out + le?vperm $out6,$out6,$out6,$inpperm + stvx_u $out5,$x40,$out + le?vperm $out7,$out7,$out7,$inpperm + stvx_u $out6,$x50,$out + stvx_u $out7,$x60,$out + addi $out,$out,0x70 + b Lcbc_dec8x_done + +.align 5 +Lcbc_dec8x_six: + vncipherlast $out2,$out2,$ivec + vncipherlast $out3,$out3,$in2 + vncipherlast $out4,$out4,$in3 + vncipherlast $out5,$out5,$in4 + vncipherlast $out6,$out6,$in5 + vncipherlast $out7,$out7,$in6 + vmr $ivec,$in7 + + le?vperm $out2,$out2,$out2,$inpperm + le?vperm $out3,$out3,$out3,$inpperm + stvx_u $out2,$x00,$out + le?vperm $out4,$out4,$out4,$inpperm + stvx_u $out3,$x10,$out + le?vperm $out5,$out5,$out5,$inpperm + stvx_u $out4,$x20,$out + le?vperm $out6,$out6,$out6,$inpperm + stvx_u $out5,$x30,$out + le?vperm $out7,$out7,$out7,$inpperm + stvx_u $out6,$x40,$out + stvx_u $out7,$x50,$out + addi $out,$out,0x60 + b Lcbc_dec8x_done + +.align 5 +Lcbc_dec8x_five: + vncipherlast $out3,$out3,$ivec + vncipherlast $out4,$out4,$in3 + vncipherlast $out5,$out5,$in4 + vncipherlast $out6,$out6,$in5 + vncipherlast $out7,$out7,$in6 + vmr $ivec,$in7 + + le?vperm $out3,$out3,$out3,$inpperm + le?vperm $out4,$out4,$out4,$inpperm + stvx_u $out3,$x00,$out + le?vperm $out5,$out5,$out5,$inpperm + stvx_u $out4,$x10,$out + le?vperm $out6,$out6,$out6,$inpperm + stvx_u $out5,$x20,$out + le?vperm $out7,$out7,$out7,$inpperm + stvx_u $out6,$x30,$out + stvx_u $out7,$x40,$out + addi $out,$out,0x50 + b Lcbc_dec8x_done + +.align 5 +Lcbc_dec8x_four: + vncipherlast $out4,$out4,$ivec + vncipherlast $out5,$out5,$in4 + vncipherlast $out6,$out6,$in5 + vncipherlast $out7,$out7,$in6 + vmr $ivec,$in7 + + le?vperm $out4,$out4,$out4,$inpperm + le?vperm $out5,$out5,$out5,$inpperm + stvx_u $out4,$x00,$out + le?vperm $out6,$out6,$out6,$inpperm + stvx_u $out5,$x10,$out + le?vperm $out7,$out7,$out7,$inpperm + stvx_u $out6,$x20,$out + stvx_u $out7,$x30,$out + addi $out,$out,0x40 + b Lcbc_dec8x_done + +.align 5 +Lcbc_dec8x_three: + vncipherlast $out5,$out5,$ivec + vncipherlast $out6,$out6,$in5 + vncipherlast $out7,$out7,$in6 + vmr $ivec,$in7 + + le?vperm $out5,$out5,$out5,$inpperm + le?vperm $out6,$out6,$out6,$inpperm + stvx_u $out5,$x00,$out + le?vperm $out7,$out7,$out7,$inpperm + stvx_u $out6,$x10,$out + stvx_u $out7,$x20,$out + addi $out,$out,0x30 + b Lcbc_dec8x_done + +.align 5 +Lcbc_dec8x_two: + vncipherlast $out6,$out6,$ivec + vncipherlast $out7,$out7,$in6 + vmr $ivec,$in7 + + le?vperm $out6,$out6,$out6,$inpperm + le?vperm $out7,$out7,$out7,$inpperm + stvx_u $out6,$x00,$out + stvx_u $out7,$x10,$out + addi $out,$out,0x20 + b Lcbc_dec8x_done + +.align 5 +Lcbc_dec8x_one: + vncipherlast $out7,$out7,$ivec + vmr $ivec,$in7 + + le?vperm $out7,$out7,$out7,$inpperm + stvx_u $out7,0,$out + addi $out,$out,0x10 + +Lcbc_dec8x_done: + le?vperm $ivec,$ivec,$ivec,$inpperm + stvx_u $ivec,0,$ivp # write [unaligned] iv + + li r10,`$FRAME+15` + li r11,`$FRAME+31` + stvx $inpperm,r10,$sp # wipe copies of round keys + addi r10,r10,32 + stvx $inpperm,r11,$sp + addi r11,r11,32 + stvx $inpperm,r10,$sp + addi r10,r10,32 + stvx $inpperm,r11,$sp + addi r11,r11,32 + stvx $inpperm,r10,$sp + addi r10,r10,32 + stvx $inpperm,r11,$sp + addi r11,r11,32 + stvx $inpperm,r10,$sp + addi r10,r10,32 + stvx $inpperm,r11,$sp + addi r11,r11,32 + + mtspr 256,$vrsave + lvx v20,r10,$sp # ABI says so + addi r10,r10,32 + lvx v21,r11,$sp + addi r11,r11,32 + lvx v22,r10,$sp + addi r10,r10,32 + lvx v23,r11,$sp + addi r11,r11,32 + lvx v24,r10,$sp + addi r10,r10,32 + lvx v25,r11,$sp + addi r11,r11,32 + lvx v26,r10,$sp + addi r10,r10,32 + lvx v27,r11,$sp + addi r11,r11,32 + lvx v28,r10,$sp + addi r10,r10,32 + lvx v29,r11,$sp + addi r11,r11,32 + lvx v30,r10,$sp + lvx v31,r11,$sp + $POP r26,`$FRAME+21*16+0*$SIZE_T`($sp) + $POP r27,`$FRAME+21*16+1*$SIZE_T`($sp) + $POP r28,`$FRAME+21*16+2*$SIZE_T`($sp) + $POP r29,`$FRAME+21*16+3*$SIZE_T`($sp) + $POP r30,`$FRAME+21*16+4*$SIZE_T`($sp) + $POP r31,`$FRAME+21*16+5*$SIZE_T`($sp) + addi $sp,$sp,`$FRAME+21*16+6*$SIZE_T` + blr + .long 0 + .byte 0,12,0x04,0,0x80,6,6,0 + .long 0 +.size .${prefix}_cbc_encrypt,.-.${prefix}_cbc_encrypt +___ +}} }}} + +######################################################################### +{{{ # CTR procedure[s] # +my ($inp,$out,$len,$key,$ivp,$x10,$rounds,$idx)=map("r$_",(3..10)); +my ($rndkey0,$rndkey1,$inout,$tmp)= map("v$_",(0..3)); +my ($ivec,$inptail,$inpperm,$outhead,$outperm,$outmask,$keyperm,$one)= + map("v$_",(4..11)); +my $dat=$tmp; + +$code.=<<___; +.globl .${prefix}_ctr32_encrypt_blocks +.align 5 +.${prefix}_ctr32_encrypt_blocks: + ${UCMP}i $len,1 + bltlr- + + lis r0,0xfff0 + mfspr $vrsave,256 + mtspr 256,r0 + + li $idx,15 + vxor $rndkey0,$rndkey0,$rndkey0 + le?vspltisb $tmp,0x0f + + lvx $ivec,0,$ivp # load [unaligned] iv + lvsl $inpperm,0,$ivp + lvx $inptail,$idx,$ivp + vspltisb $one,1 + le?vxor $inpperm,$inpperm,$tmp + vperm $ivec,$ivec,$inptail,$inpperm + vsldoi $one,$rndkey0,$one,1 + + neg r11,$inp + ?lvsl $keyperm,0,$key # prepare for unaligned key + lwz $rounds,240($key) + + lvsr $inpperm,0,r11 # prepare for unaligned load + lvx $inptail,0,$inp + addi $inp,$inp,15 # 15 is not typo + le?vxor $inpperm,$inpperm,$tmp + + srwi $rounds,$rounds,1 + li $idx,16 + subi $rounds,$rounds,1 + + ${UCMP}i $len,8 + bge _aesp8_ctr32_encrypt8x + + ?lvsr $outperm,0,$out # prepare for unaligned store + vspltisb $outmask,-1 + lvx $outhead,0,$out + ?vperm $outmask,$rndkey0,$outmask,$outperm + le?vxor $outperm,$outperm,$tmp + + lvx $rndkey0,0,$key + mtctr $rounds + lvx $rndkey1,$idx,$key + addi $idx,$idx,16 + ?vperm $rndkey0,$rndkey0,$rndkey1,$keyperm + vxor $inout,$ivec,$rndkey0 + lvx $rndkey0,$idx,$key + addi $idx,$idx,16 + b Loop_ctr32_enc + +.align 5 +Loop_ctr32_enc: + ?vperm $rndkey1,$rndkey1,$rndkey0,$keyperm + vcipher $inout,$inout,$rndkey1 + lvx $rndkey1,$idx,$key + addi $idx,$idx,16 + ?vperm $rndkey0,$rndkey0,$rndkey1,$keyperm + vcipher $inout,$inout,$rndkey0 + lvx $rndkey0,$idx,$key + addi $idx,$idx,16 + bdnz Loop_ctr32_enc + + vadduwm $ivec,$ivec,$one + vmr $dat,$inptail + lvx $inptail,0,$inp + addi $inp,$inp,16 + subic. $len,$len,1 # blocks-- + + ?vperm $rndkey1,$rndkey1,$rndkey0,$keyperm + vcipher $inout,$inout,$rndkey1 + lvx $rndkey1,$idx,$key + vperm $dat,$dat,$inptail,$inpperm + li $idx,16 + ?vperm $rndkey1,$rndkey0,$rndkey1,$keyperm + lvx $rndkey0,0,$key + vxor $dat,$dat,$rndkey1 # last round key + vcipherlast $inout,$inout,$dat + + lvx $rndkey1,$idx,$key + addi $idx,$idx,16 + vperm $inout,$inout,$inout,$outperm + vsel $dat,$outhead,$inout,$outmask + mtctr $rounds + ?vperm $rndkey0,$rndkey0,$rndkey1,$keyperm + vmr $outhead,$inout + vxor $inout,$ivec,$rndkey0 + lvx $rndkey0,$idx,$key + addi $idx,$idx,16 + stvx $dat,0,$out + addi $out,$out,16 + bne Loop_ctr32_enc + + addi $out,$out,-1 + lvx $inout,0,$out # redundant in aligned case + vsel $inout,$outhead,$inout,$outmask + stvx $inout,0,$out + + mtspr 256,$vrsave + blr + .long 0 + .byte 0,12,0x14,0,0,0,6,0 + .long 0 +___ +######################################################################### +{{ # Optimized CTR procedure # +my $key_="r11"; +my ($x00,$x10,$x20,$x30,$x40,$x50,$x60,$x70)=map("r$_",(0,8,26..31)); + $x00=0 if ($flavour =~ /osx/); +my ($in0, $in1, $in2, $in3, $in4, $in5, $in6, $in7 )=map("v$_",(0..3,10,12..14)); +my ($out0,$out1,$out2,$out3,$out4,$out5,$out6,$out7)=map("v$_",(15..22)); +my $rndkey0="v23"; # v24-v25 rotating buffer for first found keys + # v26-v31 last 6 round keys +my ($tmp,$keyperm)=($in3,$in4); # aliases with "caller", redundant assignment +my ($two,$three,$four)=($outhead,$outperm,$outmask); + +$code.=<<___; +.align 5 +_aesp8_ctr32_encrypt8x: + $STU $sp,-`($FRAME+21*16+6*$SIZE_T)`($sp) + li r10,`$FRAME+8*16+15` + li r11,`$FRAME+8*16+31` + stvx v20,r10,$sp # ABI says so + addi r10,r10,32 + stvx v21,r11,$sp + addi r11,r11,32 + stvx v22,r10,$sp + addi r10,r10,32 + stvx v23,r11,$sp + addi r11,r11,32 + stvx v24,r10,$sp + addi r10,r10,32 + stvx v25,r11,$sp + addi r11,r11,32 + stvx v26,r10,$sp + addi r10,r10,32 + stvx v27,r11,$sp + addi r11,r11,32 + stvx v28,r10,$sp + addi r10,r10,32 + stvx v29,r11,$sp + addi r11,r11,32 + stvx v30,r10,$sp + stvx v31,r11,$sp + li r0,-1 + stw $vrsave,`$FRAME+21*16-4`($sp) # save vrsave + li $x10,0x10 + $PUSH r26,`$FRAME+21*16+0*$SIZE_T`($sp) + li $x20,0x20 + $PUSH r27,`$FRAME+21*16+1*$SIZE_T`($sp) + li $x30,0x30 + $PUSH r28,`$FRAME+21*16+2*$SIZE_T`($sp) + li $x40,0x40 + $PUSH r29,`$FRAME+21*16+3*$SIZE_T`($sp) + li $x50,0x50 + $PUSH r30,`$FRAME+21*16+4*$SIZE_T`($sp) + li $x60,0x60 + $PUSH r31,`$FRAME+21*16+5*$SIZE_T`($sp) + li $x70,0x70 + mtspr 256,r0 + + subi $rounds,$rounds,3 # -4 in total + + lvx $rndkey0,$x00,$key # load key schedule + lvx v30,$x10,$key + addi $key,$key,0x20 + lvx v31,$x00,$key + ?vperm $rndkey0,$rndkey0,v30,$keyperm + addi $key_,$sp,$FRAME+15 + mtctr $rounds + +Load_ctr32_enc_key: + ?vperm v24,v30,v31,$keyperm + lvx v30,$x10,$key + addi $key,$key,0x20 + stvx v24,$x00,$key_ # off-load round[1] + ?vperm v25,v31,v30,$keyperm + lvx v31,$x00,$key + stvx v25,$x10,$key_ # off-load round[2] + addi $key_,$key_,0x20 + bdnz Load_ctr32_enc_key + + lvx v26,$x10,$key + ?vperm v24,v30,v31,$keyperm + lvx v27,$x20,$key + stvx v24,$x00,$key_ # off-load round[3] + ?vperm v25,v31,v26,$keyperm + lvx v28,$x30,$key + stvx v25,$x10,$key_ # off-load round[4] + addi $key_,$sp,$FRAME+15 # rewind $key_ + ?vperm v26,v26,v27,$keyperm + lvx v29,$x40,$key + ?vperm v27,v27,v28,$keyperm + lvx v30,$x50,$key + ?vperm v28,v28,v29,$keyperm + lvx v31,$x60,$key + ?vperm v29,v29,v30,$keyperm + lvx $out0,$x70,$key # borrow $out0 + ?vperm v30,v30,v31,$keyperm + lvx v24,$x00,$key_ # pre-load round[1] + ?vperm v31,v31,$out0,$keyperm + lvx v25,$x10,$key_ # pre-load round[2] + + vadduwm $two,$one,$one + subi $inp,$inp,15 # undo "caller" + $SHL $len,$len,4 + + vadduwm $out1,$ivec,$one # counter values ... + vadduwm $out2,$ivec,$two + vxor $out0,$ivec,$rndkey0 # ... xored with rndkey[0] + le?li $idx,8 + vadduwm $out3,$out1,$two + vxor $out1,$out1,$rndkey0 + le?lvsl $inpperm,0,$idx + vadduwm $out4,$out2,$two + vxor $out2,$out2,$rndkey0 + le?vspltisb $tmp,0x0f + vadduwm $out5,$out3,$two + vxor $out3,$out3,$rndkey0 + le?vxor $inpperm,$inpperm,$tmp # transform for lvx_u/stvx_u + vadduwm $out6,$out4,$two + vxor $out4,$out4,$rndkey0 + vadduwm $out7,$out5,$two + vxor $out5,$out5,$rndkey0 + vadduwm $ivec,$out6,$two # next counter value + vxor $out6,$out6,$rndkey0 + vxor $out7,$out7,$rndkey0 + + mtctr $rounds + b Loop_ctr32_enc8x +.align 5 +Loop_ctr32_enc8x: + vcipher $out0,$out0,v24 + vcipher $out1,$out1,v24 + vcipher $out2,$out2,v24 + vcipher $out3,$out3,v24 + vcipher $out4,$out4,v24 + vcipher $out5,$out5,v24 + vcipher $out6,$out6,v24 + vcipher $out7,$out7,v24 +Loop_ctr32_enc8x_middle: + lvx v24,$x20,$key_ # round[3] + addi $key_,$key_,0x20 + + vcipher $out0,$out0,v25 + vcipher $out1,$out1,v25 + vcipher $out2,$out2,v25 + vcipher $out3,$out3,v25 + vcipher $out4,$out4,v25 + vcipher $out5,$out5,v25 + vcipher $out6,$out6,v25 + vcipher $out7,$out7,v25 + lvx v25,$x10,$key_ # round[4] + bdnz Loop_ctr32_enc8x + + subic r11,$len,256 # $len-256, borrow $key_ + vcipher $out0,$out0,v24 + vcipher $out1,$out1,v24 + vcipher $out2,$out2,v24 + vcipher $out3,$out3,v24 + vcipher $out4,$out4,v24 + vcipher $out5,$out5,v24 + vcipher $out6,$out6,v24 + vcipher $out7,$out7,v24 + + subfe r0,r0,r0 # borrow?-1:0 + vcipher $out0,$out0,v25 + vcipher $out1,$out1,v25 + vcipher $out2,$out2,v25 + vcipher $out3,$out3,v25 + vcipher $out4,$out4,v25 + vcipher $out5,$out5,v25 + vcipher $out6,$out6,v25 + vcipher $out7,$out7,v25 + + and r0,r0,r11 + addi $key_,$sp,$FRAME+15 # rewind $key_ + vcipher $out0,$out0,v26 + vcipher $out1,$out1,v26 + vcipher $out2,$out2,v26 + vcipher $out3,$out3,v26 + vcipher $out4,$out4,v26 + vcipher $out5,$out5,v26 + vcipher $out6,$out6,v26 + vcipher $out7,$out7,v26 + lvx v24,$x00,$key_ # re-pre-load round[1] + + subic $len,$len,129 # $len-=129 + vcipher $out0,$out0,v27 + addi $len,$len,1 # $len-=128 really + vcipher $out1,$out1,v27 + vcipher $out2,$out2,v27 + vcipher $out3,$out3,v27 + vcipher $out4,$out4,v27 + vcipher $out5,$out5,v27 + vcipher $out6,$out6,v27 + vcipher $out7,$out7,v27 + lvx v25,$x10,$key_ # re-pre-load round[2] + + vcipher $out0,$out0,v28 + lvx_u $in0,$x00,$inp # load input + vcipher $out1,$out1,v28 + lvx_u $in1,$x10,$inp + vcipher $out2,$out2,v28 + lvx_u $in2,$x20,$inp + vcipher $out3,$out3,v28 + lvx_u $in3,$x30,$inp + vcipher $out4,$out4,v28 + lvx_u $in4,$x40,$inp + vcipher $out5,$out5,v28 + lvx_u $in5,$x50,$inp + vcipher $out6,$out6,v28 + lvx_u $in6,$x60,$inp + vcipher $out7,$out7,v28 + lvx_u $in7,$x70,$inp + addi $inp,$inp,0x80 + + vcipher $out0,$out0,v29 + le?vperm $in0,$in0,$in0,$inpperm + vcipher $out1,$out1,v29 + le?vperm $in1,$in1,$in1,$inpperm + vcipher $out2,$out2,v29 + le?vperm $in2,$in2,$in2,$inpperm + vcipher $out3,$out3,v29 + le?vperm $in3,$in3,$in3,$inpperm + vcipher $out4,$out4,v29 + le?vperm $in4,$in4,$in4,$inpperm + vcipher $out5,$out5,v29 + le?vperm $in5,$in5,$in5,$inpperm + vcipher $out6,$out6,v29 + le?vperm $in6,$in6,$in6,$inpperm + vcipher $out7,$out7,v29 + le?vperm $in7,$in7,$in7,$inpperm + + add $inp,$inp,r0 # $inp is adjusted in such + # way that at exit from the + # loop inX-in7 are loaded + # with last "words" + subfe. r0,r0,r0 # borrow?-1:0 + vcipher $out0,$out0,v30 + vxor $in0,$in0,v31 # xor with last round key + vcipher $out1,$out1,v30 + vxor $in1,$in1,v31 + vcipher $out2,$out2,v30 + vxor $in2,$in2,v31 + vcipher $out3,$out3,v30 + vxor $in3,$in3,v31 + vcipher $out4,$out4,v30 + vxor $in4,$in4,v31 + vcipher $out5,$out5,v30 + vxor $in5,$in5,v31 + vcipher $out6,$out6,v30 + vxor $in6,$in6,v31 + vcipher $out7,$out7,v30 + vxor $in7,$in7,v31 + + bne Lctr32_enc8x_break # did $len-129 borrow? + + vcipherlast $in0,$out0,$in0 + vcipherlast $in1,$out1,$in1 + vadduwm $out1,$ivec,$one # counter values ... + vcipherlast $in2,$out2,$in2 + vadduwm $out2,$ivec,$two + vxor $out0,$ivec,$rndkey0 # ... xored with rndkey[0] + vcipherlast $in3,$out3,$in3 + vadduwm $out3,$out1,$two + vxor $out1,$out1,$rndkey0 + vcipherlast $in4,$out4,$in4 + vadduwm $out4,$out2,$two + vxor $out2,$out2,$rndkey0 + vcipherlast $in5,$out5,$in5 + vadduwm $out5,$out3,$two + vxor $out3,$out3,$rndkey0 + vcipherlast $in6,$out6,$in6 + vadduwm $out6,$out4,$two + vxor $out4,$out4,$rndkey0 + vcipherlast $in7,$out7,$in7 + vadduwm $out7,$out5,$two + vxor $out5,$out5,$rndkey0 + le?vperm $in0,$in0,$in0,$inpperm + vadduwm $ivec,$out6,$two # next counter value + vxor $out6,$out6,$rndkey0 + le?vperm $in1,$in1,$in1,$inpperm + vxor $out7,$out7,$rndkey0 + mtctr $rounds + + vcipher $out0,$out0,v24 + stvx_u $in0,$x00,$out + le?vperm $in2,$in2,$in2,$inpperm + vcipher $out1,$out1,v24 + stvx_u $in1,$x10,$out + le?vperm $in3,$in3,$in3,$inpperm + vcipher $out2,$out2,v24 + stvx_u $in2,$x20,$out + le?vperm $in4,$in4,$in4,$inpperm + vcipher $out3,$out3,v24 + stvx_u $in3,$x30,$out + le?vperm $in5,$in5,$in5,$inpperm + vcipher $out4,$out4,v24 + stvx_u $in4,$x40,$out + le?vperm $in6,$in6,$in6,$inpperm + vcipher $out5,$out5,v24 + stvx_u $in5,$x50,$out + le?vperm $in7,$in7,$in7,$inpperm + vcipher $out6,$out6,v24 + stvx_u $in6,$x60,$out + vcipher $out7,$out7,v24 + stvx_u $in7,$x70,$out + addi $out,$out,0x80 + + b Loop_ctr32_enc8x_middle + +.align 5 +Lctr32_enc8x_break: + cmpwi $len,-0x60 + blt Lctr32_enc8x_one + nop + beq Lctr32_enc8x_two + cmpwi $len,-0x40 + blt Lctr32_enc8x_three + nop + beq Lctr32_enc8x_four + cmpwi $len,-0x20 + blt Lctr32_enc8x_five + nop + beq Lctr32_enc8x_six + cmpwi $len,0x00 + blt Lctr32_enc8x_seven + +Lctr32_enc8x_eight: + vcipherlast $out0,$out0,$in0 + vcipherlast $out1,$out1,$in1 + vcipherlast $out2,$out2,$in2 + vcipherlast $out3,$out3,$in3 + vcipherlast $out4,$out4,$in4 + vcipherlast $out5,$out5,$in5 + vcipherlast $out6,$out6,$in6 + vcipherlast $out7,$out7,$in7 + + le?vperm $out0,$out0,$out0,$inpperm + le?vperm $out1,$out1,$out1,$inpperm + stvx_u $out0,$x00,$out + le?vperm $out2,$out2,$out2,$inpperm + stvx_u $out1,$x10,$out + le?vperm $out3,$out3,$out3,$inpperm + stvx_u $out2,$x20,$out + le?vperm $out4,$out4,$out4,$inpperm + stvx_u $out3,$x30,$out + le?vperm $out5,$out5,$out5,$inpperm + stvx_u $out4,$x40,$out + le?vperm $out6,$out6,$out6,$inpperm + stvx_u $out5,$x50,$out + le?vperm $out7,$out7,$out7,$inpperm + stvx_u $out6,$x60,$out + stvx_u $out7,$x70,$out + addi $out,$out,0x80 + b Lctr32_enc8x_done + +.align 5 +Lctr32_enc8x_seven: + vcipherlast $out0,$out0,$in1 + vcipherlast $out1,$out1,$in2 + vcipherlast $out2,$out2,$in3 + vcipherlast $out3,$out3,$in4 + vcipherlast $out4,$out4,$in5 + vcipherlast $out5,$out5,$in6 + vcipherlast $out6,$out6,$in7 + + le?vperm $out0,$out0,$out0,$inpperm + le?vperm $out1,$out1,$out1,$inpperm + stvx_u $out0,$x00,$out + le?vperm $out2,$out2,$out2,$inpperm + stvx_u $out1,$x10,$out + le?vperm $out3,$out3,$out3,$inpperm + stvx_u $out2,$x20,$out + le?vperm $out4,$out4,$out4,$inpperm + stvx_u $out3,$x30,$out + le?vperm $out5,$out5,$out5,$inpperm + stvx_u $out4,$x40,$out + le?vperm $out6,$out6,$out6,$inpperm + stvx_u $out5,$x50,$out + stvx_u $out6,$x60,$out + addi $out,$out,0x70 + b Lctr32_enc8x_done + +.align 5 +Lctr32_enc8x_six: + vcipherlast $out0,$out0,$in2 + vcipherlast $out1,$out1,$in3 + vcipherlast $out2,$out2,$in4 + vcipherlast $out3,$out3,$in5 + vcipherlast $out4,$out4,$in6 + vcipherlast $out5,$out5,$in7 + + le?vperm $out0,$out0,$out0,$inpperm + le?vperm $out1,$out1,$out1,$inpperm + stvx_u $out0,$x00,$out + le?vperm $out2,$out2,$out2,$inpperm + stvx_u $out1,$x10,$out + le?vperm $out3,$out3,$out3,$inpperm + stvx_u $out2,$x20,$out + le?vperm $out4,$out4,$out4,$inpperm + stvx_u $out3,$x30,$out + le?vperm $out5,$out5,$out5,$inpperm + stvx_u $out4,$x40,$out + stvx_u $out5,$x50,$out + addi $out,$out,0x60 + b Lctr32_enc8x_done + +.align 5 +Lctr32_enc8x_five: + vcipherlast $out0,$out0,$in3 + vcipherlast $out1,$out1,$in4 + vcipherlast $out2,$out2,$in5 + vcipherlast $out3,$out3,$in6 + vcipherlast $out4,$out4,$in7 + + le?vperm $out0,$out0,$out0,$inpperm + le?vperm $out1,$out1,$out1,$inpperm + stvx_u $out0,$x00,$out + le?vperm $out2,$out2,$out2,$inpperm + stvx_u $out1,$x10,$out + le?vperm $out3,$out3,$out3,$inpperm + stvx_u $out2,$x20,$out + le?vperm $out4,$out4,$out4,$inpperm + stvx_u $out3,$x30,$out + stvx_u $out4,$x40,$out + addi $out,$out,0x50 + b Lctr32_enc8x_done + +.align 5 +Lctr32_enc8x_four: + vcipherlast $out0,$out0,$in4 + vcipherlast $out1,$out1,$in5 + vcipherlast $out2,$out2,$in6 + vcipherlast $out3,$out3,$in7 + + le?vperm $out0,$out0,$out0,$inpperm + le?vperm $out1,$out1,$out1,$inpperm + stvx_u $out0,$x00,$out + le?vperm $out2,$out2,$out2,$inpperm + stvx_u $out1,$x10,$out + le?vperm $out3,$out3,$out3,$inpperm + stvx_u $out2,$x20,$out + stvx_u $out3,$x30,$out + addi $out,$out,0x40 + b Lctr32_enc8x_done + +.align 5 +Lctr32_enc8x_three: + vcipherlast $out0,$out0,$in5 + vcipherlast $out1,$out1,$in6 + vcipherlast $out2,$out2,$in7 + + le?vperm $out0,$out0,$out0,$inpperm + le?vperm $out1,$out1,$out1,$inpperm + stvx_u $out0,$x00,$out + le?vperm $out2,$out2,$out2,$inpperm + stvx_u $out1,$x10,$out + stvx_u $out2,$x20,$out + addi $out,$out,0x30 + b Lcbc_dec8x_done + +.align 5 +Lctr32_enc8x_two: + vcipherlast $out0,$out0,$in6 + vcipherlast $out1,$out1,$in7 + + le?vperm $out0,$out0,$out0,$inpperm + le?vperm $out1,$out1,$out1,$inpperm + stvx_u $out0,$x00,$out + stvx_u $out1,$x10,$out + addi $out,$out,0x20 + b Lcbc_dec8x_done + +.align 5 +Lctr32_enc8x_one: + vcipherlast $out0,$out0,$in7 + + le?vperm $out0,$out0,$out0,$inpperm + stvx_u $out0,0,$out + addi $out,$out,0x10 + +Lctr32_enc8x_done: + li r10,`$FRAME+15` + li r11,`$FRAME+31` + stvx $inpperm,r10,$sp # wipe copies of round keys + addi r10,r10,32 + stvx $inpperm,r11,$sp + addi r11,r11,32 + stvx $inpperm,r10,$sp + addi r10,r10,32 + stvx $inpperm,r11,$sp + addi r11,r11,32 + stvx $inpperm,r10,$sp + addi r10,r10,32 + stvx $inpperm,r11,$sp + addi r11,r11,32 + stvx $inpperm,r10,$sp + addi r10,r10,32 + stvx $inpperm,r11,$sp + addi r11,r11,32 + + mtspr 256,$vrsave + lvx v20,r10,$sp # ABI says so + addi r10,r10,32 + lvx v21,r11,$sp + addi r11,r11,32 + lvx v22,r10,$sp + addi r10,r10,32 + lvx v23,r11,$sp + addi r11,r11,32 + lvx v24,r10,$sp + addi r10,r10,32 + lvx v25,r11,$sp + addi r11,r11,32 + lvx v26,r10,$sp + addi r10,r10,32 + lvx v27,r11,$sp + addi r11,r11,32 + lvx v28,r10,$sp + addi r10,r10,32 + lvx v29,r11,$sp + addi r11,r11,32 + lvx v30,r10,$sp + lvx v31,r11,$sp + $POP r26,`$FRAME+21*16+0*$SIZE_T`($sp) + $POP r27,`$FRAME+21*16+1*$SIZE_T`($sp) + $POP r28,`$FRAME+21*16+2*$SIZE_T`($sp) + $POP r29,`$FRAME+21*16+3*$SIZE_T`($sp) + $POP r30,`$FRAME+21*16+4*$SIZE_T`($sp) + $POP r31,`$FRAME+21*16+5*$SIZE_T`($sp) + addi $sp,$sp,`$FRAME+21*16+6*$SIZE_T` + blr + .long 0 + .byte 0,12,0x04,0,0x80,6,6,0 + .long 0 +.size .${prefix}_ctr32_encrypt_blocks,.-.${prefix}_ctr32_encrypt_blocks +___ +}} }}} + +######################################################################### +{{{ # XTS procedures # +# int aes_p8_xts_[en|de]crypt(const char *inp, char *out, size_t len, # +# const AES_KEY *key1, const AES_KEY *key2, # +# [const] unsigned char iv[16]); # +# If $key2 is NULL, then a "tweak chaining" mode is engaged, in which # +# input tweak value is assumed to be encrypted already, and last tweak # +# value, one suitable for consecutive call on same chunk of data, is # +# written back to original buffer. In addition, in "tweak chaining" # +# mode only complete input blocks are processed. # + +my ($inp,$out,$len,$key1,$key2,$ivp,$rounds,$idx) = map("r$_",(3..10)); +my ($rndkey0,$rndkey1,$inout) = map("v$_",(0..2)); +my ($output,$inptail,$inpperm,$leperm,$keyperm) = map("v$_",(3..7)); +my ($tweak,$seven,$eighty7,$tmp,$tweak1) = map("v$_",(8..12)); +my $taillen = $key2; + + ($inp,$idx) = ($idx,$inp); # reassign + +$code.=<<___; +.globl .${prefix}_xts_encrypt +.align 5 +.${prefix}_xts_encrypt: + mr $inp,r3 # reassign + li r3,-1 + ${UCMP}i $len,16 + bltlr- + + lis r0,0xfff0 + mfspr r12,256 # save vrsave + li r11,0 + mtspr 256,r0 + + vspltisb $seven,0x07 # 0x070707..07 + le?lvsl $leperm,r11,r11 + le?vspltisb $tmp,0x0f + le?vxor $leperm,$leperm,$seven + + li $idx,15 + lvx $tweak,0,$ivp # load [unaligned] iv + lvsl $inpperm,0,$ivp + lvx $inptail,$idx,$ivp + le?vxor $inpperm,$inpperm,$tmp + vperm $tweak,$tweak,$inptail,$inpperm + + neg r11,$inp + lvsr $inpperm,0,r11 # prepare for unaligned load + lvx $inout,0,$inp + addi $inp,$inp,15 # 15 is not typo + le?vxor $inpperm,$inpperm,$tmp + + ${UCMP}i $key2,0 # key2==NULL? + beq Lxts_enc_no_key2 + + ?lvsl $keyperm,0,$key2 # prepare for unaligned key + lwz $rounds,240($key2) + srwi $rounds,$rounds,1 + subi $rounds,$rounds,1 + li $idx,16 + + lvx $rndkey0,0,$key2 + lvx $rndkey1,$idx,$key2 + addi $idx,$idx,16 + ?vperm $rndkey0,$rndkey0,$rndkey1,$keyperm + vxor $tweak,$tweak,$rndkey0 + lvx $rndkey0,$idx,$key2 + addi $idx,$idx,16 + mtctr $rounds + +Ltweak_xts_enc: + ?vperm $rndkey1,$rndkey1,$rndkey0,$keyperm + vcipher $tweak,$tweak,$rndkey1 + lvx $rndkey1,$idx,$key2 + addi $idx,$idx,16 + ?vperm $rndkey0,$rndkey0,$rndkey1,$keyperm + vcipher $tweak,$tweak,$rndkey0 + lvx $rndkey0,$idx,$key2 + addi $idx,$idx,16 + bdnz Ltweak_xts_enc + + ?vperm $rndkey1,$rndkey1,$rndkey0,$keyperm + vcipher $tweak,$tweak,$rndkey1 + lvx $rndkey1,$idx,$key2 + ?vperm $rndkey0,$rndkey0,$rndkey1,$keyperm + vcipherlast $tweak,$tweak,$rndkey0 + + li $ivp,0 # don't chain the tweak + b Lxts_enc + +Lxts_enc_no_key2: + li $idx,-16 + and $len,$len,$idx # in "tweak chaining" + # mode only complete + # blocks are processed +Lxts_enc: + lvx $inptail,0,$inp + addi $inp,$inp,16 + + ?lvsl $keyperm,0,$key1 # prepare for unaligned key + lwz $rounds,240($key1) + srwi $rounds,$rounds,1 + subi $rounds,$rounds,1 + li $idx,16 + + vslb $eighty7,$seven,$seven # 0x808080..80 + vor $eighty7,$eighty7,$seven # 0x878787..87 + vspltisb $tmp,1 # 0x010101..01 + vsldoi $eighty7,$eighty7,$tmp,15 # 0x870101..01 + + ${UCMP}i $len,96 + bge _aesp8_xts_encrypt6x + + andi. $taillen,$len,15 + subic r0,$len,32 + subi $taillen,$taillen,16 + subfe r0,r0,r0 + and r0,r0,$taillen + add $inp,$inp,r0 + + lvx $rndkey0,0,$key1 + lvx $rndkey1,$idx,$key1 + addi $idx,$idx,16 + vperm $inout,$inout,$inptail,$inpperm + ?vperm $rndkey0,$rndkey0,$rndkey1,$keyperm + vxor $inout,$inout,$tweak + vxor $inout,$inout,$rndkey0 + lvx $rndkey0,$idx,$key1 + addi $idx,$idx,16 + mtctr $rounds + b Loop_xts_enc + +.align 5 +Loop_xts_enc: + ?vperm $rndkey1,$rndkey1,$rndkey0,$keyperm + vcipher $inout,$inout,$rndkey1 + lvx $rndkey1,$idx,$key1 + addi $idx,$idx,16 + ?vperm $rndkey0,$rndkey0,$rndkey1,$keyperm + vcipher $inout,$inout,$rndkey0 + lvx $rndkey0,$idx,$key1 + addi $idx,$idx,16 + bdnz Loop_xts_enc + + ?vperm $rndkey1,$rndkey1,$rndkey0,$keyperm + vcipher $inout,$inout,$rndkey1 + lvx $rndkey1,$idx,$key1 + li $idx,16 + ?vperm $rndkey0,$rndkey0,$rndkey1,$keyperm + vxor $rndkey0,$rndkey0,$tweak + vcipherlast $output,$inout,$rndkey0 + + le?vperm $tmp,$output,$output,$leperm + be?nop + le?stvx_u $tmp,0,$out + be?stvx_u $output,0,$out + addi $out,$out,16 + + subic. $len,$len,16 + beq Lxts_enc_done + + vmr $inout,$inptail + lvx $inptail,0,$inp + addi $inp,$inp,16 + lvx $rndkey0,0,$key1 + lvx $rndkey1,$idx,$key1 + addi $idx,$idx,16 + + subic r0,$len,32 + subfe r0,r0,r0 + and r0,r0,$taillen + add $inp,$inp,r0 + + vsrab $tmp,$tweak,$seven # next tweak value + vaddubm $tweak,$tweak,$tweak + vsldoi $tmp,$tmp,$tmp,15 + vand $tmp,$tmp,$eighty7 + vxor $tweak,$tweak,$tmp + + vperm $inout,$inout,$inptail,$inpperm + ?vperm $rndkey0,$rndkey0,$rndkey1,$keyperm + vxor $inout,$inout,$tweak + vxor $output,$output,$rndkey0 # just in case $len<16 + vxor $inout,$inout,$rndkey0 + lvx $rndkey0,$idx,$key1 + addi $idx,$idx,16 + + mtctr $rounds + ${UCMP}i $len,16 + bge Loop_xts_enc + + vxor $output,$output,$tweak + lvsr $inpperm,0,$len # $inpperm is no longer needed + vxor $inptail,$inptail,$inptail # $inptail is no longer needed + vspltisb $tmp,-1 + vperm $inptail,$inptail,$tmp,$inpperm + vsel $inout,$inout,$output,$inptail + + subi r11,$out,17 + subi $out,$out,16 + mtctr $len + li $len,16 +Loop_xts_enc_steal: + lbzu r0,1(r11) + stb r0,16(r11) + bdnz Loop_xts_enc_steal + + mtctr $rounds + b Loop_xts_enc # one more time... + +Lxts_enc_done: + ${UCMP}i $ivp,0 + beq Lxts_enc_ret + + vsrab $tmp,$tweak,$seven # next tweak value + vaddubm $tweak,$tweak,$tweak + vsldoi $tmp,$tmp,$tmp,15 + vand $tmp,$tmp,$eighty7 + vxor $tweak,$tweak,$tmp + + le?vperm $tweak,$tweak,$tweak,$leperm + stvx_u $tweak,0,$ivp + +Lxts_enc_ret: + mtspr 256,r12 # restore vrsave + li r3,0 + blr + .long 0 + .byte 0,12,0x04,0,0x80,6,6,0 + .long 0 +.size .${prefix}_xts_encrypt,.-.${prefix}_xts_encrypt + +.globl .${prefix}_xts_decrypt +.align 5 +.${prefix}_xts_decrypt: + mr $inp,r3 # reassign + li r3,-1 + ${UCMP}i $len,16 + bltlr- + + lis r0,0xfff8 + mfspr r12,256 # save vrsave + li r11,0 + mtspr 256,r0 + + andi. r0,$len,15 + neg r0,r0 + andi. r0,r0,16 + sub $len,$len,r0 + + vspltisb $seven,0x07 # 0x070707..07 + le?lvsl $leperm,r11,r11 + le?vspltisb $tmp,0x0f + le?vxor $leperm,$leperm,$seven + + li $idx,15 + lvx $tweak,0,$ivp # load [unaligned] iv + lvsl $inpperm,0,$ivp + lvx $inptail,$idx,$ivp + le?vxor $inpperm,$inpperm,$tmp + vperm $tweak,$tweak,$inptail,$inpperm + + neg r11,$inp + lvsr $inpperm,0,r11 # prepare for unaligned load + lvx $inout,0,$inp + addi $inp,$inp,15 # 15 is not typo + le?vxor $inpperm,$inpperm,$tmp + + ${UCMP}i $key2,0 # key2==NULL? + beq Lxts_dec_no_key2 + + ?lvsl $keyperm,0,$key2 # prepare for unaligned key + lwz $rounds,240($key2) + srwi $rounds,$rounds,1 + subi $rounds,$rounds,1 + li $idx,16 + + lvx $rndkey0,0,$key2 + lvx $rndkey1,$idx,$key2 + addi $idx,$idx,16 + ?vperm $rndkey0,$rndkey0,$rndkey1,$keyperm + vxor $tweak,$tweak,$rndkey0 + lvx $rndkey0,$idx,$key2 + addi $idx,$idx,16 + mtctr $rounds + +Ltweak_xts_dec: + ?vperm $rndkey1,$rndkey1,$rndkey0,$keyperm + vcipher $tweak,$tweak,$rndkey1 + lvx $rndkey1,$idx,$key2 + addi $idx,$idx,16 + ?vperm $rndkey0,$rndkey0,$rndkey1,$keyperm + vcipher $tweak,$tweak,$rndkey0 + lvx $rndkey0,$idx,$key2 + addi $idx,$idx,16 + bdnz Ltweak_xts_dec + + ?vperm $rndkey1,$rndkey1,$rndkey0,$keyperm + vcipher $tweak,$tweak,$rndkey1 + lvx $rndkey1,$idx,$key2 + ?vperm $rndkey0,$rndkey0,$rndkey1,$keyperm + vcipherlast $tweak,$tweak,$rndkey0 + + li $ivp,0 # don't chain the tweak + b Lxts_dec + +Lxts_dec_no_key2: + neg $idx,$len + andi. $idx,$idx,15 + add $len,$len,$idx # in "tweak chaining" + # mode only complete + # blocks are processed +Lxts_dec: + lvx $inptail,0,$inp + addi $inp,$inp,16 + + ?lvsl $keyperm,0,$key1 # prepare for unaligned key + lwz $rounds,240($key1) + srwi $rounds,$rounds,1 + subi $rounds,$rounds,1 + li $idx,16 + + vslb $eighty7,$seven,$seven # 0x808080..80 + vor $eighty7,$eighty7,$seven # 0x878787..87 + vspltisb $tmp,1 # 0x010101..01 + vsldoi $eighty7,$eighty7,$tmp,15 # 0x870101..01 + + ${UCMP}i $len,96 + bge _aesp8_xts_decrypt6x + + lvx $rndkey0,0,$key1 + lvx $rndkey1,$idx,$key1 + addi $idx,$idx,16 + vperm $inout,$inout,$inptail,$inpperm + ?vperm $rndkey0,$rndkey0,$rndkey1,$keyperm + vxor $inout,$inout,$tweak + vxor $inout,$inout,$rndkey0 + lvx $rndkey0,$idx,$key1 + addi $idx,$idx,16 + mtctr $rounds + + ${UCMP}i $len,16 + blt Ltail_xts_dec + be?b Loop_xts_dec + +.align 5 +Loop_xts_dec: + ?vperm $rndkey1,$rndkey1,$rndkey0,$keyperm + vncipher $inout,$inout,$rndkey1 + lvx $rndkey1,$idx,$key1 + addi $idx,$idx,16 + ?vperm $rndkey0,$rndkey0,$rndkey1,$keyperm + vncipher $inout,$inout,$rndkey0 + lvx $rndkey0,$idx,$key1 + addi $idx,$idx,16 + bdnz Loop_xts_dec + + ?vperm $rndkey1,$rndkey1,$rndkey0,$keyperm + vncipher $inout,$inout,$rndkey1 + lvx $rndkey1,$idx,$key1 + li $idx,16 + ?vperm $rndkey0,$rndkey0,$rndkey1,$keyperm + vxor $rndkey0,$rndkey0,$tweak + vncipherlast $output,$inout,$rndkey0 + + le?vperm $tmp,$output,$output,$leperm + be?nop + le?stvx_u $tmp,0,$out + be?stvx_u $output,0,$out + addi $out,$out,16 + + subic. $len,$len,16 + beq Lxts_dec_done + + vmr $inout,$inptail + lvx $inptail,0,$inp + addi $inp,$inp,16 + lvx $rndkey0,0,$key1 + lvx $rndkey1,$idx,$key1 + addi $idx,$idx,16 + + vsrab $tmp,$tweak,$seven # next tweak value + vaddubm $tweak,$tweak,$tweak + vsldoi $tmp,$tmp,$tmp,15 + vand $tmp,$tmp,$eighty7 + vxor $tweak,$tweak,$tmp + + vperm $inout,$inout,$inptail,$inpperm + ?vperm $rndkey0,$rndkey0,$rndkey1,$keyperm + vxor $inout,$inout,$tweak + vxor $inout,$inout,$rndkey0 + lvx $rndkey0,$idx,$key1 + addi $idx,$idx,16 + + mtctr $rounds + ${UCMP}i $len,16 + bge Loop_xts_dec + +Ltail_xts_dec: + vsrab $tmp,$tweak,$seven # next tweak value + vaddubm $tweak1,$tweak,$tweak + vsldoi $tmp,$tmp,$tmp,15 + vand $tmp,$tmp,$eighty7 + vxor $tweak1,$tweak1,$tmp + + subi $inp,$inp,16 + add $inp,$inp,$len + + vxor $inout,$inout,$tweak # :-( + vxor $inout,$inout,$tweak1 # :-) + +Loop_xts_dec_short: + ?vperm $rndkey1,$rndkey1,$rndkey0,$keyperm + vncipher $inout,$inout,$rndkey1 + lvx $rndkey1,$idx,$key1 + addi $idx,$idx,16 + ?vperm $rndkey0,$rndkey0,$rndkey1,$keyperm + vncipher $inout,$inout,$rndkey0 + lvx $rndkey0,$idx,$key1 + addi $idx,$idx,16 + bdnz Loop_xts_dec_short + + ?vperm $rndkey1,$rndkey1,$rndkey0,$keyperm + vncipher $inout,$inout,$rndkey1 + lvx $rndkey1,$idx,$key1 + li $idx,16 + ?vperm $rndkey0,$rndkey0,$rndkey1,$keyperm + vxor $rndkey0,$rndkey0,$tweak1 + vncipherlast $output,$inout,$rndkey0 + + le?vperm $tmp,$output,$output,$leperm + be?nop + le?stvx_u $tmp,0,$out + be?stvx_u $output,0,$out + + vmr $inout,$inptail + lvx $inptail,0,$inp + #addi $inp,$inp,16 + lvx $rndkey0,0,$key1 + lvx $rndkey1,$idx,$key1 + addi $idx,$idx,16 + vperm $inout,$inout,$inptail,$inpperm + ?vperm $rndkey0,$rndkey0,$rndkey1,$keyperm + + lvsr $inpperm,0,$len # $inpperm is no longer needed + vxor $inptail,$inptail,$inptail # $inptail is no longer needed + vspltisb $tmp,-1 + vperm $inptail,$inptail,$tmp,$inpperm + vsel $inout,$inout,$output,$inptail + + vxor $rndkey0,$rndkey0,$tweak + vxor $inout,$inout,$rndkey0 + lvx $rndkey0,$idx,$key1 + addi $idx,$idx,16 + + subi r11,$out,1 + mtctr $len + li $len,16 +Loop_xts_dec_steal: + lbzu r0,1(r11) + stb r0,16(r11) + bdnz Loop_xts_dec_steal + + mtctr $rounds + b Loop_xts_dec # one more time... + +Lxts_dec_done: + ${UCMP}i $ivp,0 + beq Lxts_dec_ret + + vsrab $tmp,$tweak,$seven # next tweak value + vaddubm $tweak,$tweak,$tweak + vsldoi $tmp,$tmp,$tmp,15 + vand $tmp,$tmp,$eighty7 + vxor $tweak,$tweak,$tmp + + le?vperm $tweak,$tweak,$tweak,$leperm + stvx_u $tweak,0,$ivp + +Lxts_dec_ret: + mtspr 256,r12 # restore vrsave + li r3,0 + blr + .long 0 + .byte 0,12,0x04,0,0x80,6,6,0 + .long 0 +.size .${prefix}_xts_decrypt,.-.${prefix}_xts_decrypt +___ +######################################################################### +{{ # Optimized XTS procedures # +my $key_=$key2; +my ($x00,$x10,$x20,$x30,$x40,$x50,$x60,$x70)=map("r$_",(0,3,26..31)); + $x00=0 if ($flavour =~ /osx/); +my ($in0, $in1, $in2, $in3, $in4, $in5 )=map("v$_",(0..5)); +my ($out0, $out1, $out2, $out3, $out4, $out5)=map("v$_",(7,12..16)); +my ($twk0, $twk1, $twk2, $twk3, $twk4, $twk5)=map("v$_",(17..22)); +my $rndkey0="v23"; # v24-v25 rotating buffer for first found keys + # v26-v31 last 6 round keys +my ($keyperm)=($out0); # aliases with "caller", redundant assignment +my $taillen=$x70; + +$code.=<<___; +.align 5 +_aesp8_xts_encrypt6x: + $STU $sp,-`($FRAME+21*16+6*$SIZE_T)`($sp) + mflr r11 + li r7,`$FRAME+8*16+15` + li r3,`$FRAME+8*16+31` + $PUSH r11,`$FRAME+21*16+6*$SIZE_T+$LRSAVE`($sp) + stvx v20,r7,$sp # ABI says so + addi r7,r7,32 + stvx v21,r3,$sp + addi r3,r3,32 + stvx v22,r7,$sp + addi r7,r7,32 + stvx v23,r3,$sp + addi r3,r3,32 + stvx v24,r7,$sp + addi r7,r7,32 + stvx v25,r3,$sp + addi r3,r3,32 + stvx v26,r7,$sp + addi r7,r7,32 + stvx v27,r3,$sp + addi r3,r3,32 + stvx v28,r7,$sp + addi r7,r7,32 + stvx v29,r3,$sp + addi r3,r3,32 + stvx v30,r7,$sp + stvx v31,r3,$sp + li r0,-1 + stw $vrsave,`$FRAME+21*16-4`($sp) # save vrsave + li $x10,0x10 + $PUSH r26,`$FRAME+21*16+0*$SIZE_T`($sp) + li $x20,0x20 + $PUSH r27,`$FRAME+21*16+1*$SIZE_T`($sp) + li $x30,0x30 + $PUSH r28,`$FRAME+21*16+2*$SIZE_T`($sp) + li $x40,0x40 + $PUSH r29,`$FRAME+21*16+3*$SIZE_T`($sp) + li $x50,0x50 + $PUSH r30,`$FRAME+21*16+4*$SIZE_T`($sp) + li $x60,0x60 + $PUSH r31,`$FRAME+21*16+5*$SIZE_T`($sp) + li $x70,0x70 + mtspr 256,r0 + + subi $rounds,$rounds,3 # -4 in total + + lvx $rndkey0,$x00,$key1 # load key schedule + lvx v30,$x10,$key1 + addi $key1,$key1,0x20 + lvx v31,$x00,$key1 + ?vperm $rndkey0,$rndkey0,v30,$keyperm + addi $key_,$sp,$FRAME+15 + mtctr $rounds + +Load_xts_enc_key: + ?vperm v24,v30,v31,$keyperm + lvx v30,$x10,$key1 + addi $key1,$key1,0x20 + stvx v24,$x00,$key_ # off-load round[1] + ?vperm v25,v31,v30,$keyperm + lvx v31,$x00,$key1 + stvx v25,$x10,$key_ # off-load round[2] + addi $key_,$key_,0x20 + bdnz Load_xts_enc_key + + lvx v26,$x10,$key1 + ?vperm v24,v30,v31,$keyperm + lvx v27,$x20,$key1 + stvx v24,$x00,$key_ # off-load round[3] + ?vperm v25,v31,v26,$keyperm + lvx v28,$x30,$key1 + stvx v25,$x10,$key_ # off-load round[4] + addi $key_,$sp,$FRAME+15 # rewind $key_ + ?vperm v26,v26,v27,$keyperm + lvx v29,$x40,$key1 + ?vperm v27,v27,v28,$keyperm + lvx v30,$x50,$key1 + ?vperm v28,v28,v29,$keyperm + lvx v31,$x60,$key1 + ?vperm v29,v29,v30,$keyperm + lvx $twk5,$x70,$key1 # borrow $twk5 + ?vperm v30,v30,v31,$keyperm + lvx v24,$x00,$key_ # pre-load round[1] + ?vperm v31,v31,$twk5,$keyperm + lvx v25,$x10,$key_ # pre-load round[2] + + vperm $in0,$inout,$inptail,$inpperm + subi $inp,$inp,31 # undo "caller" + vxor $twk0,$tweak,$rndkey0 + vsrab $tmp,$tweak,$seven # next tweak value + vaddubm $tweak,$tweak,$tweak + vsldoi $tmp,$tmp,$tmp,15 + vand $tmp,$tmp,$eighty7 + vxor $out0,$in0,$twk0 + vxor $tweak,$tweak,$tmp + + lvx_u $in1,$x10,$inp + vxor $twk1,$tweak,$rndkey0 + vsrab $tmp,$tweak,$seven # next tweak value + vaddubm $tweak,$tweak,$tweak + vsldoi $tmp,$tmp,$tmp,15 + le?vperm $in1,$in1,$in1,$leperm + vand $tmp,$tmp,$eighty7 + vxor $out1,$in1,$twk1 + vxor $tweak,$tweak,$tmp + + lvx_u $in2,$x20,$inp + andi. $taillen,$len,15 + vxor $twk2,$tweak,$rndkey0 + vsrab $tmp,$tweak,$seven # next tweak value + vaddubm $tweak,$tweak,$tweak + vsldoi $tmp,$tmp,$tmp,15 + le?vperm $in2,$in2,$in2,$leperm + vand $tmp,$tmp,$eighty7 + vxor $out2,$in2,$twk2 + vxor $tweak,$tweak,$tmp + + lvx_u $in3,$x30,$inp + sub $len,$len,$taillen + vxor $twk3,$tweak,$rndkey0 + vsrab $tmp,$tweak,$seven # next tweak value + vaddubm $tweak,$tweak,$tweak + vsldoi $tmp,$tmp,$tmp,15 + le?vperm $in3,$in3,$in3,$leperm + vand $tmp,$tmp,$eighty7 + vxor $out3,$in3,$twk3 + vxor $tweak,$tweak,$tmp + + lvx_u $in4,$x40,$inp + subi $len,$len,0x60 + vxor $twk4,$tweak,$rndkey0 + vsrab $tmp,$tweak,$seven # next tweak value + vaddubm $tweak,$tweak,$tweak + vsldoi $tmp,$tmp,$tmp,15 + le?vperm $in4,$in4,$in4,$leperm + vand $tmp,$tmp,$eighty7 + vxor $out4,$in4,$twk4 + vxor $tweak,$tweak,$tmp + + lvx_u $in5,$x50,$inp + addi $inp,$inp,0x60 + vxor $twk5,$tweak,$rndkey0 + vsrab $tmp,$tweak,$seven # next tweak value + vaddubm $tweak,$tweak,$tweak + vsldoi $tmp,$tmp,$tmp,15 + le?vperm $in5,$in5,$in5,$leperm + vand $tmp,$tmp,$eighty7 + vxor $out5,$in5,$twk5 + vxor $tweak,$tweak,$tmp + + vxor v31,v31,$rndkey0 + mtctr $rounds + b Loop_xts_enc6x + +.align 5 +Loop_xts_enc6x: + vcipher $out0,$out0,v24 + vcipher $out1,$out1,v24 + vcipher $out2,$out2,v24 + vcipher $out3,$out3,v24 + vcipher $out4,$out4,v24 + vcipher $out5,$out5,v24 + lvx v24,$x20,$key_ # round[3] + addi $key_,$key_,0x20 + + vcipher $out0,$out0,v25 + vcipher $out1,$out1,v25 + vcipher $out2,$out2,v25 + vcipher $out3,$out3,v25 + vcipher $out4,$out4,v25 + vcipher $out5,$out5,v25 + lvx v25,$x10,$key_ # round[4] + bdnz Loop_xts_enc6x + + subic $len,$len,96 # $len-=96 + vxor $in0,$twk0,v31 # xor with last round key + vcipher $out0,$out0,v24 + vcipher $out1,$out1,v24 + vsrab $tmp,$tweak,$seven # next tweak value + vxor $twk0,$tweak,$rndkey0 + vaddubm $tweak,$tweak,$tweak + vcipher $out2,$out2,v24 + vcipher $out3,$out3,v24 + vsldoi $tmp,$tmp,$tmp,15 + vcipher $out4,$out4,v24 + vcipher $out5,$out5,v24 + + subfe. r0,r0,r0 # borrow?-1:0 + vand $tmp,$tmp,$eighty7 + vcipher $out0,$out0,v25 + vcipher $out1,$out1,v25 + vxor $tweak,$tweak,$tmp + vcipher $out2,$out2,v25 + vcipher $out3,$out3,v25 + vxor $in1,$twk1,v31 + vsrab $tmp,$tweak,$seven # next tweak value + vxor $twk1,$tweak,$rndkey0 + vcipher $out4,$out4,v25 + vcipher $out5,$out5,v25 + + and r0,r0,$len + vaddubm $tweak,$tweak,$tweak + vsldoi $tmp,$tmp,$tmp,15 + vcipher $out0,$out0,v26 + vcipher $out1,$out1,v26 + vand $tmp,$tmp,$eighty7 + vcipher $out2,$out2,v26 + vcipher $out3,$out3,v26 + vxor $tweak,$tweak,$tmp + vcipher $out4,$out4,v26 + vcipher $out5,$out5,v26 + + add $inp,$inp,r0 # $inp is adjusted in such + # way that at exit from the + # loop inX-in5 are loaded + # with last "words" + vxor $in2,$twk2,v31 + vsrab $tmp,$tweak,$seven # next tweak value + vxor $twk2,$tweak,$rndkey0 + vaddubm $tweak,$tweak,$tweak + vcipher $out0,$out0,v27 + vcipher $out1,$out1,v27 + vsldoi $tmp,$tmp,$tmp,15 + vcipher $out2,$out2,v27 + vcipher $out3,$out3,v27 + vand $tmp,$tmp,$eighty7 + vcipher $out4,$out4,v27 + vcipher $out5,$out5,v27 + + addi $key_,$sp,$FRAME+15 # rewind $key_ + vxor $tweak,$tweak,$tmp + vcipher $out0,$out0,v28 + vcipher $out1,$out1,v28 + vxor $in3,$twk3,v31 + vsrab $tmp,$tweak,$seven # next tweak value + vxor $twk3,$tweak,$rndkey0 + vcipher $out2,$out2,v28 + vcipher $out3,$out3,v28 + vaddubm $tweak,$tweak,$tweak + vsldoi $tmp,$tmp,$tmp,15 + vcipher $out4,$out4,v28 + vcipher $out5,$out5,v28 + lvx v24,$x00,$key_ # re-pre-load round[1] + vand $tmp,$tmp,$eighty7 + + vcipher $out0,$out0,v29 + vcipher $out1,$out1,v29 + vxor $tweak,$tweak,$tmp + vcipher $out2,$out2,v29 + vcipher $out3,$out3,v29 + vxor $in4,$twk4,v31 + vsrab $tmp,$tweak,$seven # next tweak value + vxor $twk4,$tweak,$rndkey0 + vcipher $out4,$out4,v29 + vcipher $out5,$out5,v29 + lvx v25,$x10,$key_ # re-pre-load round[2] + vaddubm $tweak,$tweak,$tweak + vsldoi $tmp,$tmp,$tmp,15 + + vcipher $out0,$out0,v30 + vcipher $out1,$out1,v30 + vand $tmp,$tmp,$eighty7 + vcipher $out2,$out2,v30 + vcipher $out3,$out3,v30 + vxor $tweak,$tweak,$tmp + vcipher $out4,$out4,v30 + vcipher $out5,$out5,v30 + vxor $in5,$twk5,v31 + vsrab $tmp,$tweak,$seven # next tweak value + vxor $twk5,$tweak,$rndkey0 + + vcipherlast $out0,$out0,$in0 + lvx_u $in0,$x00,$inp # load next input block + vaddubm $tweak,$tweak,$tweak + vsldoi $tmp,$tmp,$tmp,15 + vcipherlast $out1,$out1,$in1 + lvx_u $in1,$x10,$inp + vcipherlast $out2,$out2,$in2 + le?vperm $in0,$in0,$in0,$leperm + lvx_u $in2,$x20,$inp + vand $tmp,$tmp,$eighty7 + vcipherlast $out3,$out3,$in3 + le?vperm $in1,$in1,$in1,$leperm + lvx_u $in3,$x30,$inp + vcipherlast $out4,$out4,$in4 + le?vperm $in2,$in2,$in2,$leperm + lvx_u $in4,$x40,$inp + vxor $tweak,$tweak,$tmp + vcipherlast $tmp,$out5,$in5 # last block might be needed + # in stealing mode + le?vperm $in3,$in3,$in3,$leperm + lvx_u $in5,$x50,$inp + addi $inp,$inp,0x60 + le?vperm $in4,$in4,$in4,$leperm + le?vperm $in5,$in5,$in5,$leperm + + le?vperm $out0,$out0,$out0,$leperm + le?vperm $out1,$out1,$out1,$leperm + stvx_u $out0,$x00,$out # store output + vxor $out0,$in0,$twk0 + le?vperm $out2,$out2,$out2,$leperm + stvx_u $out1,$x10,$out + vxor $out1,$in1,$twk1 + le?vperm $out3,$out3,$out3,$leperm + stvx_u $out2,$x20,$out + vxor $out2,$in2,$twk2 + le?vperm $out4,$out4,$out4,$leperm + stvx_u $out3,$x30,$out + vxor $out3,$in3,$twk3 + le?vperm $out5,$tmp,$tmp,$leperm + stvx_u $out4,$x40,$out + vxor $out4,$in4,$twk4 + le?stvx_u $out5,$x50,$out + be?stvx_u $tmp, $x50,$out + vxor $out5,$in5,$twk5 + addi $out,$out,0x60 + + mtctr $rounds + beq Loop_xts_enc6x # did $len-=96 borrow? + + addic. $len,$len,0x60 + beq Lxts_enc6x_zero + cmpwi $len,0x20 + blt Lxts_enc6x_one + nop + beq Lxts_enc6x_two + cmpwi $len,0x40 + blt Lxts_enc6x_three + nop + beq Lxts_enc6x_four + +Lxts_enc6x_five: + vxor $out0,$in1,$twk0 + vxor $out1,$in2,$twk1 + vxor $out2,$in3,$twk2 + vxor $out3,$in4,$twk3 + vxor $out4,$in5,$twk4 + + bl _aesp8_xts_enc5x + + le?vperm $out0,$out0,$out0,$leperm + vmr $twk0,$twk5 # unused tweak + le?vperm $out1,$out1,$out1,$leperm + stvx_u $out0,$x00,$out # store output + le?vperm $out2,$out2,$out2,$leperm + stvx_u $out1,$x10,$out + le?vperm $out3,$out3,$out3,$leperm + stvx_u $out2,$x20,$out + vxor $tmp,$out4,$twk5 # last block prep for stealing + le?vperm $out4,$out4,$out4,$leperm + stvx_u $out3,$x30,$out + stvx_u $out4,$x40,$out + addi $out,$out,0x50 + bne Lxts_enc6x_steal + b Lxts_enc6x_done + +.align 4 +Lxts_enc6x_four: + vxor $out0,$in2,$twk0 + vxor $out1,$in3,$twk1 + vxor $out2,$in4,$twk2 + vxor $out3,$in5,$twk3 + vxor $out4,$out4,$out4 + + bl _aesp8_xts_enc5x + + le?vperm $out0,$out0,$out0,$leperm + vmr $twk0,$twk4 # unused tweak + le?vperm $out1,$out1,$out1,$leperm + stvx_u $out0,$x00,$out # store output + le?vperm $out2,$out2,$out2,$leperm + stvx_u $out1,$x10,$out + vxor $tmp,$out3,$twk4 # last block prep for stealing + le?vperm $out3,$out3,$out3,$leperm + stvx_u $out2,$x20,$out + stvx_u $out3,$x30,$out + addi $out,$out,0x40 + bne Lxts_enc6x_steal + b Lxts_enc6x_done + +.align 4 +Lxts_enc6x_three: + vxor $out0,$in3,$twk0 + vxor $out1,$in4,$twk1 + vxor $out2,$in5,$twk2 + vxor $out3,$out3,$out3 + vxor $out4,$out4,$out4 + + bl _aesp8_xts_enc5x + + le?vperm $out0,$out0,$out0,$leperm + vmr $twk0,$twk3 # unused tweak + le?vperm $out1,$out1,$out1,$leperm + stvx_u $out0,$x00,$out # store output + vxor $tmp,$out2,$twk3 # last block prep for stealing + le?vperm $out2,$out2,$out2,$leperm + stvx_u $out1,$x10,$out + stvx_u $out2,$x20,$out + addi $out,$out,0x30 + bne Lxts_enc6x_steal + b Lxts_enc6x_done + +.align 4 +Lxts_enc6x_two: + vxor $out0,$in4,$twk0 + vxor $out1,$in5,$twk1 + vxor $out2,$out2,$out2 + vxor $out3,$out3,$out3 + vxor $out4,$out4,$out4 + + bl _aesp8_xts_enc5x + + le?vperm $out0,$out0,$out0,$leperm + vmr $twk0,$twk2 # unused tweak + vxor $tmp,$out1,$twk2 # last block prep for stealing + le?vperm $out1,$out1,$out1,$leperm + stvx_u $out0,$x00,$out # store output + stvx_u $out1,$x10,$out + addi $out,$out,0x20 + bne Lxts_enc6x_steal + b Lxts_enc6x_done + +.align 4 +Lxts_enc6x_one: + vxor $out0,$in5,$twk0 + nop +Loop_xts_enc1x: + vcipher $out0,$out0,v24 + lvx v24,$x20,$key_ # round[3] + addi $key_,$key_,0x20 + + vcipher $out0,$out0,v25 + lvx v25,$x10,$key_ # round[4] + bdnz Loop_xts_enc1x + + add $inp,$inp,$taillen + cmpwi $taillen,0 + vcipher $out0,$out0,v24 + + subi $inp,$inp,16 + vcipher $out0,$out0,v25 + + lvsr $inpperm,0,$taillen + vcipher $out0,$out0,v26 + + lvx_u $in0,0,$inp + vcipher $out0,$out0,v27 + + addi $key_,$sp,$FRAME+15 # rewind $key_ + vcipher $out0,$out0,v28 + lvx v24,$x00,$key_ # re-pre-load round[1] + + vcipher $out0,$out0,v29 + lvx v25,$x10,$key_ # re-pre-load round[2] + vxor $twk0,$twk0,v31 + + le?vperm $in0,$in0,$in0,$leperm + vcipher $out0,$out0,v30 + + vperm $in0,$in0,$in0,$inpperm + vcipherlast $out0,$out0,$twk0 + + vmr $twk0,$twk1 # unused tweak + vxor $tmp,$out0,$twk1 # last block prep for stealing + le?vperm $out0,$out0,$out0,$leperm + stvx_u $out0,$x00,$out # store output + addi $out,$out,0x10 + bne Lxts_enc6x_steal + b Lxts_enc6x_done + +.align 4 +Lxts_enc6x_zero: + cmpwi $taillen,0 + beq Lxts_enc6x_done + + add $inp,$inp,$taillen + subi $inp,$inp,16 + lvx_u $in0,0,$inp + lvsr $inpperm,0,$taillen # $in5 is no more + le?vperm $in0,$in0,$in0,$leperm + vperm $in0,$in0,$in0,$inpperm + vxor $tmp,$tmp,$twk0 +Lxts_enc6x_steal: + vxor $in0,$in0,$twk0 + vxor $out0,$out0,$out0 + vspltisb $out1,-1 + vperm $out0,$out0,$out1,$inpperm + vsel $out0,$in0,$tmp,$out0 # $tmp is last block, remember? + + subi r30,$out,17 + subi $out,$out,16 + mtctr $taillen +Loop_xts_enc6x_steal: + lbzu r0,1(r30) + stb r0,16(r30) + bdnz Loop_xts_enc6x_steal + + li $taillen,0 + mtctr $rounds + b Loop_xts_enc1x # one more time... + +.align 4 +Lxts_enc6x_done: + ${UCMP}i $ivp,0 + beq Lxts_enc6x_ret + + vxor $tweak,$twk0,$rndkey0 + le?vperm $tweak,$tweak,$tweak,$leperm + stvx_u $tweak,0,$ivp + +Lxts_enc6x_ret: + mtlr r11 + li r10,`$FRAME+15` + li r11,`$FRAME+31` + stvx $seven,r10,$sp # wipe copies of round keys + addi r10,r10,32 + stvx $seven,r11,$sp + addi r11,r11,32 + stvx $seven,r10,$sp + addi r10,r10,32 + stvx $seven,r11,$sp + addi r11,r11,32 + stvx $seven,r10,$sp + addi r10,r10,32 + stvx $seven,r11,$sp + addi r11,r11,32 + stvx $seven,r10,$sp + addi r10,r10,32 + stvx $seven,r11,$sp + addi r11,r11,32 + + mtspr 256,$vrsave + lvx v20,r10,$sp # ABI says so + addi r10,r10,32 + lvx v21,r11,$sp + addi r11,r11,32 + lvx v22,r10,$sp + addi r10,r10,32 + lvx v23,r11,$sp + addi r11,r11,32 + lvx v24,r10,$sp + addi r10,r10,32 + lvx v25,r11,$sp + addi r11,r11,32 + lvx v26,r10,$sp + addi r10,r10,32 + lvx v27,r11,$sp + addi r11,r11,32 + lvx v28,r10,$sp + addi r10,r10,32 + lvx v29,r11,$sp + addi r11,r11,32 + lvx v30,r10,$sp + lvx v31,r11,$sp + $POP r26,`$FRAME+21*16+0*$SIZE_T`($sp) + $POP r27,`$FRAME+21*16+1*$SIZE_T`($sp) + $POP r28,`$FRAME+21*16+2*$SIZE_T`($sp) + $POP r29,`$FRAME+21*16+3*$SIZE_T`($sp) + $POP r30,`$FRAME+21*16+4*$SIZE_T`($sp) + $POP r31,`$FRAME+21*16+5*$SIZE_T`($sp) + addi $sp,$sp,`$FRAME+21*16+6*$SIZE_T` + blr + .long 0 + .byte 0,12,0x04,1,0x80,6,6,0 + .long 0 + +.align 5 +_aesp8_xts_enc5x: + vcipher $out0,$out0,v24 + vcipher $out1,$out1,v24 + vcipher $out2,$out2,v24 + vcipher $out3,$out3,v24 + vcipher $out4,$out4,v24 + lvx v24,$x20,$key_ # round[3] + addi $key_,$key_,0x20 + + vcipher $out0,$out0,v25 + vcipher $out1,$out1,v25 + vcipher $out2,$out2,v25 + vcipher $out3,$out3,v25 + vcipher $out4,$out4,v25 + lvx v25,$x10,$key_ # round[4] + bdnz _aesp8_xts_enc5x + + add $inp,$inp,$taillen + cmpwi $taillen,0 + vcipher $out0,$out0,v24 + vcipher $out1,$out1,v24 + vcipher $out2,$out2,v24 + vcipher $out3,$out3,v24 + vcipher $out4,$out4,v24 + + subi $inp,$inp,16 + vcipher $out0,$out0,v25 + vcipher $out1,$out1,v25 + vcipher $out2,$out2,v25 + vcipher $out3,$out3,v25 + vcipher $out4,$out4,v25 + vxor $twk0,$twk0,v31 + + vcipher $out0,$out0,v26 + lvsr $inpperm,0,$taillen # $in5 is no more + vcipher $out1,$out1,v26 + vcipher $out2,$out2,v26 + vcipher $out3,$out3,v26 + vcipher $out4,$out4,v26 + vxor $in1,$twk1,v31 + + vcipher $out0,$out0,v27 + lvx_u $in0,0,$inp + vcipher $out1,$out1,v27 + vcipher $out2,$out2,v27 + vcipher $out3,$out3,v27 + vcipher $out4,$out4,v27 + vxor $in2,$twk2,v31 + + addi $key_,$sp,$FRAME+15 # rewind $key_ + vcipher $out0,$out0,v28 + vcipher $out1,$out1,v28 + vcipher $out2,$out2,v28 + vcipher $out3,$out3,v28 + vcipher $out4,$out4,v28 + lvx v24,$x00,$key_ # re-pre-load round[1] + vxor $in3,$twk3,v31 + + vcipher $out0,$out0,v29 + le?vperm $in0,$in0,$in0,$leperm + vcipher $out1,$out1,v29 + vcipher $out2,$out2,v29 + vcipher $out3,$out3,v29 + vcipher $out4,$out4,v29 + lvx v25,$x10,$key_ # re-pre-load round[2] + vxor $in4,$twk4,v31 + + vcipher $out0,$out0,v30 + vperm $in0,$in0,$in0,$inpperm + vcipher $out1,$out1,v30 + vcipher $out2,$out2,v30 + vcipher $out3,$out3,v30 + vcipher $out4,$out4,v30 + + vcipherlast $out0,$out0,$twk0 + vcipherlast $out1,$out1,$in1 + vcipherlast $out2,$out2,$in2 + vcipherlast $out3,$out3,$in3 + vcipherlast $out4,$out4,$in4 + blr + .long 0 + .byte 0,12,0x14,0,0,0,0,0 + +.align 5 +_aesp8_xts_decrypt6x: + $STU $sp,-`($FRAME+21*16+6*$SIZE_T)`($sp) + mflr r11 + li r7,`$FRAME+8*16+15` + li r3,`$FRAME+8*16+31` + $PUSH r11,`$FRAME+21*16+6*$SIZE_T+$LRSAVE`($sp) + stvx v20,r7,$sp # ABI says so + addi r7,r7,32 + stvx v21,r3,$sp + addi r3,r3,32 + stvx v22,r7,$sp + addi r7,r7,32 + stvx v23,r3,$sp + addi r3,r3,32 + stvx v24,r7,$sp + addi r7,r7,32 + stvx v25,r3,$sp + addi r3,r3,32 + stvx v26,r7,$sp + addi r7,r7,32 + stvx v27,r3,$sp + addi r3,r3,32 + stvx v28,r7,$sp + addi r7,r7,32 + stvx v29,r3,$sp + addi r3,r3,32 + stvx v30,r7,$sp + stvx v31,r3,$sp + li r0,-1 + stw $vrsave,`$FRAME+21*16-4`($sp) # save vrsave + li $x10,0x10 + $PUSH r26,`$FRAME+21*16+0*$SIZE_T`($sp) + li $x20,0x20 + $PUSH r27,`$FRAME+21*16+1*$SIZE_T`($sp) + li $x30,0x30 + $PUSH r28,`$FRAME+21*16+2*$SIZE_T`($sp) + li $x40,0x40 + $PUSH r29,`$FRAME+21*16+3*$SIZE_T`($sp) + li $x50,0x50 + $PUSH r30,`$FRAME+21*16+4*$SIZE_T`($sp) + li $x60,0x60 + $PUSH r31,`$FRAME+21*16+5*$SIZE_T`($sp) + li $x70,0x70 + mtspr 256,r0 + + subi $rounds,$rounds,3 # -4 in total + + lvx $rndkey0,$x00,$key1 # load key schedule + lvx v30,$x10,$key1 + addi $key1,$key1,0x20 + lvx v31,$x00,$key1 + ?vperm $rndkey0,$rndkey0,v30,$keyperm + addi $key_,$sp,$FRAME+15 + mtctr $rounds + +Load_xts_dec_key: + ?vperm v24,v30,v31,$keyperm + lvx v30,$x10,$key1 + addi $key1,$key1,0x20 + stvx v24,$x00,$key_ # off-load round[1] + ?vperm v25,v31,v30,$keyperm + lvx v31,$x00,$key1 + stvx v25,$x10,$key_ # off-load round[2] + addi $key_,$key_,0x20 + bdnz Load_xts_dec_key + + lvx v26,$x10,$key1 + ?vperm v24,v30,v31,$keyperm + lvx v27,$x20,$key1 + stvx v24,$x00,$key_ # off-load round[3] + ?vperm v25,v31,v26,$keyperm + lvx v28,$x30,$key1 + stvx v25,$x10,$key_ # off-load round[4] + addi $key_,$sp,$FRAME+15 # rewind $key_ + ?vperm v26,v26,v27,$keyperm + lvx v29,$x40,$key1 + ?vperm v27,v27,v28,$keyperm + lvx v30,$x50,$key1 + ?vperm v28,v28,v29,$keyperm + lvx v31,$x60,$key1 + ?vperm v29,v29,v30,$keyperm + lvx $twk5,$x70,$key1 # borrow $twk5 + ?vperm v30,v30,v31,$keyperm + lvx v24,$x00,$key_ # pre-load round[1] + ?vperm v31,v31,$twk5,$keyperm + lvx v25,$x10,$key_ # pre-load round[2] + + vperm $in0,$inout,$inptail,$inpperm + subi $inp,$inp,31 # undo "caller" + vxor $twk0,$tweak,$rndkey0 + vsrab $tmp,$tweak,$seven # next tweak value + vaddubm $tweak,$tweak,$tweak + vsldoi $tmp,$tmp,$tmp,15 + vand $tmp,$tmp,$eighty7 + vxor $out0,$in0,$twk0 + vxor $tweak,$tweak,$tmp + + lvx_u $in1,$x10,$inp + vxor $twk1,$tweak,$rndkey0 + vsrab $tmp,$tweak,$seven # next tweak value + vaddubm $tweak,$tweak,$tweak + vsldoi $tmp,$tmp,$tmp,15 + le?vperm $in1,$in1,$in1,$leperm + vand $tmp,$tmp,$eighty7 + vxor $out1,$in1,$twk1 + vxor $tweak,$tweak,$tmp + + lvx_u $in2,$x20,$inp + andi. $taillen,$len,15 + vxor $twk2,$tweak,$rndkey0 + vsrab $tmp,$tweak,$seven # next tweak value + vaddubm $tweak,$tweak,$tweak + vsldoi $tmp,$tmp,$tmp,15 + le?vperm $in2,$in2,$in2,$leperm + vand $tmp,$tmp,$eighty7 + vxor $out2,$in2,$twk2 + vxor $tweak,$tweak,$tmp + + lvx_u $in3,$x30,$inp + sub $len,$len,$taillen + vxor $twk3,$tweak,$rndkey0 + vsrab $tmp,$tweak,$seven # next tweak value + vaddubm $tweak,$tweak,$tweak + vsldoi $tmp,$tmp,$tmp,15 + le?vperm $in3,$in3,$in3,$leperm + vand $tmp,$tmp,$eighty7 + vxor $out3,$in3,$twk3 + vxor $tweak,$tweak,$tmp + + lvx_u $in4,$x40,$inp + subi $len,$len,0x60 + vxor $twk4,$tweak,$rndkey0 + vsrab $tmp,$tweak,$seven # next tweak value + vaddubm $tweak,$tweak,$tweak + vsldoi $tmp,$tmp,$tmp,15 + le?vperm $in4,$in4,$in4,$leperm + vand $tmp,$tmp,$eighty7 + vxor $out4,$in4,$twk4 + vxor $tweak,$tweak,$tmp + + lvx_u $in5,$x50,$inp + addi $inp,$inp,0x60 + vxor $twk5,$tweak,$rndkey0 + vsrab $tmp,$tweak,$seven # next tweak value + vaddubm $tweak,$tweak,$tweak + vsldoi $tmp,$tmp,$tmp,15 + le?vperm $in5,$in5,$in5,$leperm + vand $tmp,$tmp,$eighty7 + vxor $out5,$in5,$twk5 + vxor $tweak,$tweak,$tmp + + vxor v31,v31,$rndkey0 + mtctr $rounds + b Loop_xts_dec6x + +.align 5 +Loop_xts_dec6x: + vncipher $out0,$out0,v24 + vncipher $out1,$out1,v24 + vncipher $out2,$out2,v24 + vncipher $out3,$out3,v24 + vncipher $out4,$out4,v24 + vncipher $out5,$out5,v24 + lvx v24,$x20,$key_ # round[3] + addi $key_,$key_,0x20 + + vncipher $out0,$out0,v25 + vncipher $out1,$out1,v25 + vncipher $out2,$out2,v25 + vncipher $out3,$out3,v25 + vncipher $out4,$out4,v25 + vncipher $out5,$out5,v25 + lvx v25,$x10,$key_ # round[4] + bdnz Loop_xts_dec6x + + subic $len,$len,96 # $len-=96 + vxor $in0,$twk0,v31 # xor with last round key + vncipher $out0,$out0,v24 + vncipher $out1,$out1,v24 + vsrab $tmp,$tweak,$seven # next tweak value + vxor $twk0,$tweak,$rndkey0 + vaddubm $tweak,$tweak,$tweak + vncipher $out2,$out2,v24 + vncipher $out3,$out3,v24 + vsldoi $tmp,$tmp,$tmp,15 + vncipher $out4,$out4,v24 + vncipher $out5,$out5,v24 + + subfe. r0,r0,r0 # borrow?-1:0 + vand $tmp,$tmp,$eighty7 + vncipher $out0,$out0,v25 + vncipher $out1,$out1,v25 + vxor $tweak,$tweak,$tmp + vncipher $out2,$out2,v25 + vncipher $out3,$out3,v25 + vxor $in1,$twk1,v31 + vsrab $tmp,$tweak,$seven # next tweak value + vxor $twk1,$tweak,$rndkey0 + vncipher $out4,$out4,v25 + vncipher $out5,$out5,v25 + + and r0,r0,$len + vaddubm $tweak,$tweak,$tweak + vsldoi $tmp,$tmp,$tmp,15 + vncipher $out0,$out0,v26 + vncipher $out1,$out1,v26 + vand $tmp,$tmp,$eighty7 + vncipher $out2,$out2,v26 + vncipher $out3,$out3,v26 + vxor $tweak,$tweak,$tmp + vncipher $out4,$out4,v26 + vncipher $out5,$out5,v26 + + add $inp,$inp,r0 # $inp is adjusted in such + # way that at exit from the + # loop inX-in5 are loaded + # with last "words" + vxor $in2,$twk2,v31 + vsrab $tmp,$tweak,$seven # next tweak value + vxor $twk2,$tweak,$rndkey0 + vaddubm $tweak,$tweak,$tweak + vncipher $out0,$out0,v27 + vncipher $out1,$out1,v27 + vsldoi $tmp,$tmp,$tmp,15 + vncipher $out2,$out2,v27 + vncipher $out3,$out3,v27 + vand $tmp,$tmp,$eighty7 + vncipher $out4,$out4,v27 + vncipher $out5,$out5,v27 + + addi $key_,$sp,$FRAME+15 # rewind $key_ + vxor $tweak,$tweak,$tmp + vncipher $out0,$out0,v28 + vncipher $out1,$out1,v28 + vxor $in3,$twk3,v31 + vsrab $tmp,$tweak,$seven # next tweak value + vxor $twk3,$tweak,$rndkey0 + vncipher $out2,$out2,v28 + vncipher $out3,$out3,v28 + vaddubm $tweak,$tweak,$tweak + vsldoi $tmp,$tmp,$tmp,15 + vncipher $out4,$out4,v28 + vncipher $out5,$out5,v28 + lvx v24,$x00,$key_ # re-pre-load round[1] + vand $tmp,$tmp,$eighty7 + + vncipher $out0,$out0,v29 + vncipher $out1,$out1,v29 + vxor $tweak,$tweak,$tmp + vncipher $out2,$out2,v29 + vncipher $out3,$out3,v29 + vxor $in4,$twk4,v31 + vsrab $tmp,$tweak,$seven # next tweak value + vxor $twk4,$tweak,$rndkey0 + vncipher $out4,$out4,v29 + vncipher $out5,$out5,v29 + lvx v25,$x10,$key_ # re-pre-load round[2] + vaddubm $tweak,$tweak,$tweak + vsldoi $tmp,$tmp,$tmp,15 + + vncipher $out0,$out0,v30 + vncipher $out1,$out1,v30 + vand $tmp,$tmp,$eighty7 + vncipher $out2,$out2,v30 + vncipher $out3,$out3,v30 + vxor $tweak,$tweak,$tmp + vncipher $out4,$out4,v30 + vncipher $out5,$out5,v30 + vxor $in5,$twk5,v31 + vsrab $tmp,$tweak,$seven # next tweak value + vxor $twk5,$tweak,$rndkey0 + + vncipherlast $out0,$out0,$in0 + lvx_u $in0,$x00,$inp # load next input block + vaddubm $tweak,$tweak,$tweak + vsldoi $tmp,$tmp,$tmp,15 + vncipherlast $out1,$out1,$in1 + lvx_u $in1,$x10,$inp + vncipherlast $out2,$out2,$in2 + le?vperm $in0,$in0,$in0,$leperm + lvx_u $in2,$x20,$inp + vand $tmp,$tmp,$eighty7 + vncipherlast $out3,$out3,$in3 + le?vperm $in1,$in1,$in1,$leperm + lvx_u $in3,$x30,$inp + vncipherlast $out4,$out4,$in4 + le?vperm $in2,$in2,$in2,$leperm + lvx_u $in4,$x40,$inp + vxor $tweak,$tweak,$tmp + vncipherlast $out5,$out5,$in5 + le?vperm $in3,$in3,$in3,$leperm + lvx_u $in5,$x50,$inp + addi $inp,$inp,0x60 + le?vperm $in4,$in4,$in4,$leperm + le?vperm $in5,$in5,$in5,$leperm + + le?vperm $out0,$out0,$out0,$leperm + le?vperm $out1,$out1,$out1,$leperm + stvx_u $out0,$x00,$out # store output + vxor $out0,$in0,$twk0 + le?vperm $out2,$out2,$out2,$leperm + stvx_u $out1,$x10,$out + vxor $out1,$in1,$twk1 + le?vperm $out3,$out3,$out3,$leperm + stvx_u $out2,$x20,$out + vxor $out2,$in2,$twk2 + le?vperm $out4,$out4,$out4,$leperm + stvx_u $out3,$x30,$out + vxor $out3,$in3,$twk3 + le?vperm $out5,$out5,$out5,$leperm + stvx_u $out4,$x40,$out + vxor $out4,$in4,$twk4 + stvx_u $out5,$x50,$out + vxor $out5,$in5,$twk5 + addi $out,$out,0x60 + + mtctr $rounds + beq Loop_xts_dec6x # did $len-=96 borrow? + + addic. $len,$len,0x60 + beq Lxts_dec6x_zero + cmpwi $len,0x20 + blt Lxts_dec6x_one + nop + beq Lxts_dec6x_two + cmpwi $len,0x40 + blt Lxts_dec6x_three + nop + beq Lxts_dec6x_four + +Lxts_dec6x_five: + vxor $out0,$in1,$twk0 + vxor $out1,$in2,$twk1 + vxor $out2,$in3,$twk2 + vxor $out3,$in4,$twk3 + vxor $out4,$in5,$twk4 + + bl _aesp8_xts_dec5x + + le?vperm $out0,$out0,$out0,$leperm + vmr $twk0,$twk5 # unused tweak + vxor $twk1,$tweak,$rndkey0 + le?vperm $out1,$out1,$out1,$leperm + stvx_u $out0,$x00,$out # store output + vxor $out0,$in0,$twk1 + le?vperm $out2,$out2,$out2,$leperm + stvx_u $out1,$x10,$out + le?vperm $out3,$out3,$out3,$leperm + stvx_u $out2,$x20,$out + le?vperm $out4,$out4,$out4,$leperm + stvx_u $out3,$x30,$out + stvx_u $out4,$x40,$out + addi $out,$out,0x50 + bne Lxts_dec6x_steal + b Lxts_dec6x_done + +.align 4 +Lxts_dec6x_four: + vxor $out0,$in2,$twk0 + vxor $out1,$in3,$twk1 + vxor $out2,$in4,$twk2 + vxor $out3,$in5,$twk3 + vxor $out4,$out4,$out4 + + bl _aesp8_xts_dec5x + + le?vperm $out0,$out0,$out0,$leperm + vmr $twk0,$twk4 # unused tweak + vmr $twk1,$twk5 + le?vperm $out1,$out1,$out1,$leperm + stvx_u $out0,$x00,$out # store output + vxor $out0,$in0,$twk5 + le?vperm $out2,$out2,$out2,$leperm + stvx_u $out1,$x10,$out + le?vperm $out3,$out3,$out3,$leperm + stvx_u $out2,$x20,$out + stvx_u $out3,$x30,$out + addi $out,$out,0x40 + bne Lxts_dec6x_steal + b Lxts_dec6x_done + +.align 4 +Lxts_dec6x_three: + vxor $out0,$in3,$twk0 + vxor $out1,$in4,$twk1 + vxor $out2,$in5,$twk2 + vxor $out3,$out3,$out3 + vxor $out4,$out4,$out4 + + bl _aesp8_xts_dec5x + + le?vperm $out0,$out0,$out0,$leperm + vmr $twk0,$twk3 # unused tweak + vmr $twk1,$twk4 + le?vperm $out1,$out1,$out1,$leperm + stvx_u $out0,$x00,$out # store output + vxor $out0,$in0,$twk4 + le?vperm $out2,$out2,$out2,$leperm + stvx_u $out1,$x10,$out + stvx_u $out2,$x20,$out + addi $out,$out,0x30 + bne Lxts_dec6x_steal + b Lxts_dec6x_done + +.align 4 +Lxts_dec6x_two: + vxor $out0,$in4,$twk0 + vxor $out1,$in5,$twk1 + vxor $out2,$out2,$out2 + vxor $out3,$out3,$out3 + vxor $out4,$out4,$out4 + + bl _aesp8_xts_dec5x + + le?vperm $out0,$out0,$out0,$leperm + vmr $twk0,$twk2 # unused tweak + vmr $twk1,$twk3 + le?vperm $out1,$out1,$out1,$leperm + stvx_u $out0,$x00,$out # store output + vxor $out0,$in0,$twk3 + stvx_u $out1,$x10,$out + addi $out,$out,0x20 + bne Lxts_dec6x_steal + b Lxts_dec6x_done + +.align 4 +Lxts_dec6x_one: + vxor $out0,$in5,$twk0 + nop +Loop_xts_dec1x: + vncipher $out0,$out0,v24 + lvx v24,$x20,$key_ # round[3] + addi $key_,$key_,0x20 + + vncipher $out0,$out0,v25 + lvx v25,$x10,$key_ # round[4] + bdnz Loop_xts_dec1x + + subi r0,$taillen,1 + vncipher $out0,$out0,v24 + + andi. r0,r0,16 + cmpwi $taillen,0 + vncipher $out0,$out0,v25 + + sub $inp,$inp,r0 + vncipher $out0,$out0,v26 + + lvx_u $in0,0,$inp + vncipher $out0,$out0,v27 + + addi $key_,$sp,$FRAME+15 # rewind $key_ + vncipher $out0,$out0,v28 + lvx v24,$x00,$key_ # re-pre-load round[1] + + vncipher $out0,$out0,v29 + lvx v25,$x10,$key_ # re-pre-load round[2] + vxor $twk0,$twk0,v31 + + le?vperm $in0,$in0,$in0,$leperm + vncipher $out0,$out0,v30 + + mtctr $rounds + vncipherlast $out0,$out0,$twk0 + + vmr $twk0,$twk1 # unused tweak + vmr $twk1,$twk2 + le?vperm $out0,$out0,$out0,$leperm + stvx_u $out0,$x00,$out # store output + addi $out,$out,0x10 + vxor $out0,$in0,$twk2 + bne Lxts_dec6x_steal + b Lxts_dec6x_done + +.align 4 +Lxts_dec6x_zero: + cmpwi $taillen,0 + beq Lxts_dec6x_done + + lvx_u $in0,0,$inp + le?vperm $in0,$in0,$in0,$leperm + vxor $out0,$in0,$twk1 +Lxts_dec6x_steal: + vncipher $out0,$out0,v24 + lvx v24,$x20,$key_ # round[3] + addi $key_,$key_,0x20 + + vncipher $out0,$out0,v25 + lvx v25,$x10,$key_ # round[4] + bdnz Lxts_dec6x_steal + + add $inp,$inp,$taillen + vncipher $out0,$out0,v24 + + cmpwi $taillen,0 + vncipher $out0,$out0,v25 + + lvx_u $in0,0,$inp + vncipher $out0,$out0,v26 + + lvsr $inpperm,0,$taillen # $in5 is no more + vncipher $out0,$out0,v27 + + addi $key_,$sp,$FRAME+15 # rewind $key_ + vncipher $out0,$out0,v28 + lvx v24,$x00,$key_ # re-pre-load round[1] + + vncipher $out0,$out0,v29 + lvx v25,$x10,$key_ # re-pre-load round[2] + vxor $twk1,$twk1,v31 + + le?vperm $in0,$in0,$in0,$leperm + vncipher $out0,$out0,v30 + + vperm $in0,$in0,$in0,$inpperm + vncipherlast $tmp,$out0,$twk1 + + le?vperm $out0,$tmp,$tmp,$leperm + le?stvx_u $out0,0,$out + be?stvx_u $tmp,0,$out + + vxor $out0,$out0,$out0 + vspltisb $out1,-1 + vperm $out0,$out0,$out1,$inpperm + vsel $out0,$in0,$tmp,$out0 + vxor $out0,$out0,$twk0 + + subi r30,$out,1 + mtctr $taillen +Loop_xts_dec6x_steal: + lbzu r0,1(r30) + stb r0,16(r30) + bdnz Loop_xts_dec6x_steal + + li $taillen,0 + mtctr $rounds + b Loop_xts_dec1x # one more time... + +.align 4 +Lxts_dec6x_done: + ${UCMP}i $ivp,0 + beq Lxts_dec6x_ret + + vxor $tweak,$twk0,$rndkey0 + le?vperm $tweak,$tweak,$tweak,$leperm + stvx_u $tweak,0,$ivp + +Lxts_dec6x_ret: + mtlr r11 + li r10,`$FRAME+15` + li r11,`$FRAME+31` + stvx $seven,r10,$sp # wipe copies of round keys + addi r10,r10,32 + stvx $seven,r11,$sp + addi r11,r11,32 + stvx $seven,r10,$sp + addi r10,r10,32 + stvx $seven,r11,$sp + addi r11,r11,32 + stvx $seven,r10,$sp + addi r10,r10,32 + stvx $seven,r11,$sp + addi r11,r11,32 + stvx $seven,r10,$sp + addi r10,r10,32 + stvx $seven,r11,$sp + addi r11,r11,32 + + mtspr 256,$vrsave + lvx v20,r10,$sp # ABI says so + addi r10,r10,32 + lvx v21,r11,$sp + addi r11,r11,32 + lvx v22,r10,$sp + addi r10,r10,32 + lvx v23,r11,$sp + addi r11,r11,32 + lvx v24,r10,$sp + addi r10,r10,32 + lvx v25,r11,$sp + addi r11,r11,32 + lvx v26,r10,$sp + addi r10,r10,32 + lvx v27,r11,$sp + addi r11,r11,32 + lvx v28,r10,$sp + addi r10,r10,32 + lvx v29,r11,$sp + addi r11,r11,32 + lvx v30,r10,$sp + lvx v31,r11,$sp + $POP r26,`$FRAME+21*16+0*$SIZE_T`($sp) + $POP r27,`$FRAME+21*16+1*$SIZE_T`($sp) + $POP r28,`$FRAME+21*16+2*$SIZE_T`($sp) + $POP r29,`$FRAME+21*16+3*$SIZE_T`($sp) + $POP r30,`$FRAME+21*16+4*$SIZE_T`($sp) + $POP r31,`$FRAME+21*16+5*$SIZE_T`($sp) + addi $sp,$sp,`$FRAME+21*16+6*$SIZE_T` + blr + .long 0 + .byte 0,12,0x04,1,0x80,6,6,0 + .long 0 + +.align 5 +_aesp8_xts_dec5x: + vncipher $out0,$out0,v24 + vncipher $out1,$out1,v24 + vncipher $out2,$out2,v24 + vncipher $out3,$out3,v24 + vncipher $out4,$out4,v24 + lvx v24,$x20,$key_ # round[3] + addi $key_,$key_,0x20 + + vncipher $out0,$out0,v25 + vncipher $out1,$out1,v25 + vncipher $out2,$out2,v25 + vncipher $out3,$out3,v25 + vncipher $out4,$out4,v25 + lvx v25,$x10,$key_ # round[4] + bdnz _aesp8_xts_dec5x + + subi r0,$taillen,1 + vncipher $out0,$out0,v24 + vncipher $out1,$out1,v24 + vncipher $out2,$out2,v24 + vncipher $out3,$out3,v24 + vncipher $out4,$out4,v24 + + andi. r0,r0,16 + cmpwi $taillen,0 + vncipher $out0,$out0,v25 + vncipher $out1,$out1,v25 + vncipher $out2,$out2,v25 + vncipher $out3,$out3,v25 + vncipher $out4,$out4,v25 + vxor $twk0,$twk0,v31 + + sub $inp,$inp,r0 + vncipher $out0,$out0,v26 + vncipher $out1,$out1,v26 + vncipher $out2,$out2,v26 + vncipher $out3,$out3,v26 + vncipher $out4,$out4,v26 + vxor $in1,$twk1,v31 + + vncipher $out0,$out0,v27 + lvx_u $in0,0,$inp + vncipher $out1,$out1,v27 + vncipher $out2,$out2,v27 + vncipher $out3,$out3,v27 + vncipher $out4,$out4,v27 + vxor $in2,$twk2,v31 + + addi $key_,$sp,$FRAME+15 # rewind $key_ + vncipher $out0,$out0,v28 + vncipher $out1,$out1,v28 + vncipher $out2,$out2,v28 + vncipher $out3,$out3,v28 + vncipher $out4,$out4,v28 + lvx v24,$x00,$key_ # re-pre-load round[1] + vxor $in3,$twk3,v31 + + vncipher $out0,$out0,v29 + le?vperm $in0,$in0,$in0,$leperm + vncipher $out1,$out1,v29 + vncipher $out2,$out2,v29 + vncipher $out3,$out3,v29 + vncipher $out4,$out4,v29 + lvx v25,$x10,$key_ # re-pre-load round[2] + vxor $in4,$twk4,v31 + + vncipher $out0,$out0,v30 + vncipher $out1,$out1,v30 + vncipher $out2,$out2,v30 + vncipher $out3,$out3,v30 + vncipher $out4,$out4,v30 + + vncipherlast $out0,$out0,$twk0 + vncipherlast $out1,$out1,$in1 + vncipherlast $out2,$out2,$in2 + vncipherlast $out3,$out3,$in3 + vncipherlast $out4,$out4,$in4 + mtctr $rounds + blr + .long 0 + .byte 0,12,0x14,0,0,0,0,0 +___ +}} }}} + +my $consts=1; +foreach(split("\n",$code)) { + s/\`([^\`]*)\`/eval($1)/geo; + + # constants table endian-specific conversion + if ($consts && m/\.(long|byte)\s+(.+)\s+(\?[a-z]*)$/o) { + my $conv=$3; + my @bytes=(); + + # convert to endian-agnostic format + if ($1 eq "long") { + foreach (split(/,\s*/,$2)) { + my $l = /^0/?oct:int; + push @bytes,($l>>24)&0xff,($l>>16)&0xff,($l>>8)&0xff,$l&0xff; + } + } else { + @bytes = map(/^0/?oct:int,split(/,\s*/,$2)); + } + + # little-endian conversion + if ($flavour =~ /le$/o) { + SWITCH: for($conv) { + /\?inv/ && do { @bytes=map($_^0xf,@bytes); last; }; + /\?rev/ && do { @bytes=reverse(@bytes); last; }; + } + } + + #emit + print ".byte\t",join(',',map (sprintf("0x%02x",$_),@bytes)),"\n"; + next; + } + $consts=0 if (m/Lconsts:/o); # end of table + + # instructions prefixed with '?' are endian-specific and need + # to be adjusted accordingly... + if ($flavour =~ /le$/o) { # little-endian + s/le\?//o or + s/be\?/#be#/o or + s/\?lvsr/lvsl/o or + s/\?lvsl/lvsr/o or + s/\?(vperm\s+v[0-9]+,\s*)(v[0-9]+,\s*)(v[0-9]+,\s*)(v[0-9]+)/$1$3$2$4/o or + s/\?(vsldoi\s+v[0-9]+,\s*)(v[0-9]+,)\s*(v[0-9]+,\s*)([0-9]+)/$1$3$2 16-$4/o or + s/\?(vspltw\s+v[0-9]+,\s*)(v[0-9]+,)\s*([0-9])/$1$2 3-$3/o; + } else { # big-endian + s/le\?/#le#/o or + s/be\?//o or + s/\?([a-z]+)/$1/o; + } + + print $_,"\n"; +} + +close STDOUT; diff --git a/openssl-1.1.0h/crypto/aes/asm/aest4-sparcv9.pl b/openssl-1.1.0h/crypto/aes/asm/aest4-sparcv9.pl new file mode 100644 index 0000000..bf479c6 --- /dev/null +++ b/openssl-1.1.0h/crypto/aes/asm/aest4-sparcv9.pl @@ -0,0 +1,929 @@ +#! /usr/bin/env perl +# Copyright 2012-2016 The OpenSSL Project Authors. All Rights Reserved. +# +# Licensed under the OpenSSL license (the "License"). You may not use +# this file except in compliance with the License. You can obtain a copy +# in the file LICENSE in the source distribution or at +# https://www.openssl.org/source/license.html + + +# ==================================================================== +# Written by David S. Miller and Andy Polyakov +# . The module is licensed under 2-clause BSD +# license. October 2012. All rights reserved. +# ==================================================================== + +###################################################################### +# AES for SPARC T4. +# +# AES round instructions complete in 3 cycles and can be issued every +# cycle. It means that round calculations should take 4*rounds cycles, +# because any given round instruction depends on result of *both* +# previous instructions: +# +# |0 |1 |2 |3 |4 +# |01|01|01| +# |23|23|23| +# |01|01|... +# |23|... +# +# Provided that fxor [with IV] takes 3 cycles to complete, critical +# path length for CBC encrypt would be 3+4*rounds, or in other words +# it should process one byte in at least (3+4*rounds)/16 cycles. This +# estimate doesn't account for "collateral" instructions, such as +# fetching input from memory, xor-ing it with zero-round key and +# storing the result. Yet, *measured* performance [for data aligned +# at 64-bit boundary!] deviates from this equation by less than 0.5%: +# +# 128-bit key 192- 256- +# CBC encrypt 2.70/2.90(*) 3.20/3.40 3.70/3.90 +# (*) numbers after slash are for +# misaligned data. +# +# Out-of-order execution logic managed to fully overlap "collateral" +# instructions with those on critical path. Amazing! +# +# As with Intel AES-NI, question is if it's possible to improve +# performance of parallelizeable modes by interleaving round +# instructions. Provided round instruction latency and throughput +# optimal interleave factor is 2. But can we expect 2x performance +# improvement? Well, as round instructions can be issued one per +# cycle, they don't saturate the 2-way issue pipeline and therefore +# there is room for "collateral" calculations... Yet, 2x speed-up +# over CBC encrypt remains unattaintable: +# +# 128-bit key 192- 256- +# CBC decrypt 1.64/2.11 1.89/2.37 2.23/2.61 +# CTR 1.64/2.08(*) 1.89/2.33 2.23/2.61 +# (*) numbers after slash are for +# misaligned data. +# +# Estimates based on amount of instructions under assumption that +# round instructions are not pairable with any other instruction +# suggest that latter is the actual case and pipeline runs +# underutilized. It should be noted that T4 out-of-order execution +# logic is so capable that performance gain from 2x interleave is +# not even impressive, ~7-13% over non-interleaved code, largest +# for 256-bit keys. + +# To anchor to something else, software implementation processes +# one byte in 29 cycles with 128-bit key on same processor. Intel +# Sandy Bridge encrypts byte in 5.07 cycles in CBC mode and decrypts +# in 0.93, naturally with AES-NI. + +$0 =~ m/(.*[\/\\])[^\/\\]+$/; $dir=$1; +push(@INC,"${dir}","${dir}../../perlasm"); +require "sparcv9_modes.pl"; + +$output = pop; +open STDOUT,">$output"; + +$::evp=1; # if $evp is set to 0, script generates module with +# AES_[en|de]crypt, AES_set_[en|de]crypt_key and AES_cbc_encrypt entry +# points. These however are not fully compatible with openssl/aes.h, +# because they expect AES_KEY to be aligned at 64-bit boundary. When +# used through EVP, alignment is arranged at EVP layer. Second thing +# that is arranged by EVP is at least 32-bit alignment of IV. + +###################################################################### +# single-round subroutines +# +{ +my ($inp,$out,$key,$rounds,$tmp,$mask)=map("%o$_",(0..5)); + +$code.=<<___; +#include "sparc_arch.h" + +#ifdef __arch64__ +.register %g2,#scratch +.register %g3,#scratch +#endif + +.text + +.globl aes_t4_encrypt +.align 32 +aes_t4_encrypt: + andcc $inp, 7, %g1 ! is input aligned? + andn $inp, 7, $inp + + ldx [$key + 0], %g4 + ldx [$key + 8], %g5 + + ldx [$inp + 0], %o4 + bz,pt %icc, 1f + ldx [$inp + 8], %o5 + ldx [$inp + 16], $inp + sll %g1, 3, %g1 + sub %g0, %g1, %o3 + sllx %o4, %g1, %o4 + sllx %o5, %g1, %g1 + srlx %o5, %o3, %o5 + srlx $inp, %o3, %o3 + or %o5, %o4, %o4 + or %o3, %g1, %o5 +1: + ld [$key + 240], $rounds + ldd [$key + 16], %f12 + ldd [$key + 24], %f14 + xor %g4, %o4, %o4 + xor %g5, %o5, %o5 + movxtod %o4, %f0 + movxtod %o5, %f2 + srl $rounds, 1, $rounds + ldd [$key + 32], %f16 + sub $rounds, 1, $rounds + ldd [$key + 40], %f18 + add $key, 48, $key + +.Lenc: + aes_eround01 %f12, %f0, %f2, %f4 + aes_eround23 %f14, %f0, %f2, %f2 + ldd [$key + 0], %f12 + ldd [$key + 8], %f14 + sub $rounds,1,$rounds + aes_eround01 %f16, %f4, %f2, %f0 + aes_eround23 %f18, %f4, %f2, %f2 + ldd [$key + 16], %f16 + ldd [$key + 24], %f18 + brnz,pt $rounds, .Lenc + add $key, 32, $key + + andcc $out, 7, $tmp ! is output aligned? + aes_eround01 %f12, %f0, %f2, %f4 + aes_eround23 %f14, %f0, %f2, %f2 + aes_eround01_l %f16, %f4, %f2, %f0 + aes_eround23_l %f18, %f4, %f2, %f2 + + bnz,pn %icc, 2f + nop + + std %f0, [$out + 0] + retl + std %f2, [$out + 8] + +2: alignaddrl $out, %g0, $out + mov 0xff, $mask + srl $mask, $tmp, $mask + + faligndata %f0, %f0, %f4 + faligndata %f0, %f2, %f6 + faligndata %f2, %f2, %f8 + + stda %f4, [$out + $mask]0xc0 ! partial store + std %f6, [$out + 8] + add $out, 16, $out + orn %g0, $mask, $mask + retl + stda %f8, [$out + $mask]0xc0 ! partial store +.type aes_t4_encrypt,#function +.size aes_t4_encrypt,.-aes_t4_encrypt + +.globl aes_t4_decrypt +.align 32 +aes_t4_decrypt: + andcc $inp, 7, %g1 ! is input aligned? + andn $inp, 7, $inp + + ldx [$key + 0], %g4 + ldx [$key + 8], %g5 + + ldx [$inp + 0], %o4 + bz,pt %icc, 1f + ldx [$inp + 8], %o5 + ldx [$inp + 16], $inp + sll %g1, 3, %g1 + sub %g0, %g1, %o3 + sllx %o4, %g1, %o4 + sllx %o5, %g1, %g1 + srlx %o5, %o3, %o5 + srlx $inp, %o3, %o3 + or %o5, %o4, %o4 + or %o3, %g1, %o5 +1: + ld [$key + 240], $rounds + ldd [$key + 16], %f12 + ldd [$key + 24], %f14 + xor %g4, %o4, %o4 + xor %g5, %o5, %o5 + movxtod %o4, %f0 + movxtod %o5, %f2 + srl $rounds, 1, $rounds + ldd [$key + 32], %f16 + sub $rounds, 1, $rounds + ldd [$key + 40], %f18 + add $key, 48, $key + +.Ldec: + aes_dround01 %f12, %f0, %f2, %f4 + aes_dround23 %f14, %f0, %f2, %f2 + ldd [$key + 0], %f12 + ldd [$key + 8], %f14 + sub $rounds,1,$rounds + aes_dround01 %f16, %f4, %f2, %f0 + aes_dround23 %f18, %f4, %f2, %f2 + ldd [$key + 16], %f16 + ldd [$key + 24], %f18 + brnz,pt $rounds, .Ldec + add $key, 32, $key + + andcc $out, 7, $tmp ! is output aligned? + aes_dround01 %f12, %f0, %f2, %f4 + aes_dround23 %f14, %f0, %f2, %f2 + aes_dround01_l %f16, %f4, %f2, %f0 + aes_dround23_l %f18, %f4, %f2, %f2 + + bnz,pn %icc, 2f + nop + + std %f0, [$out + 0] + retl + std %f2, [$out + 8] + +2: alignaddrl $out, %g0, $out + mov 0xff, $mask + srl $mask, $tmp, $mask + + faligndata %f0, %f0, %f4 + faligndata %f0, %f2, %f6 + faligndata %f2, %f2, %f8 + + stda %f4, [$out + $mask]0xc0 ! partial store + std %f6, [$out + 8] + add $out, 16, $out + orn %g0, $mask, $mask + retl + stda %f8, [$out + $mask]0xc0 ! partial store +.type aes_t4_decrypt,#function +.size aes_t4_decrypt,.-aes_t4_decrypt +___ +} + +###################################################################### +# key setup subroutines +# +{ +my ($inp,$bits,$out,$tmp)=map("%o$_",(0..5)); +$code.=<<___; +.globl aes_t4_set_encrypt_key +.align 32 +aes_t4_set_encrypt_key: +.Lset_encrypt_key: + and $inp, 7, $tmp + alignaddr $inp, %g0, $inp + cmp $bits, 192 + ldd [$inp + 0], %f0 + bl,pt %icc,.L128 + ldd [$inp + 8], %f2 + + be,pt %icc,.L192 + ldd [$inp + 16], %f4 + brz,pt $tmp, .L256aligned + ldd [$inp + 24], %f6 + + ldd [$inp + 32], %f8 + faligndata %f0, %f2, %f0 + faligndata %f2, %f4, %f2 + faligndata %f4, %f6, %f4 + faligndata %f6, %f8, %f6 +.L256aligned: +___ +for ($i=0; $i<6; $i++) { + $code.=<<___; + std %f0, [$out + `32*$i+0`] + aes_kexpand1 %f0, %f6, $i, %f0 + std %f2, [$out + `32*$i+8`] + aes_kexpand2 %f2, %f0, %f2 + std %f4, [$out + `32*$i+16`] + aes_kexpand0 %f4, %f2, %f4 + std %f6, [$out + `32*$i+24`] + aes_kexpand2 %f6, %f4, %f6 +___ +} +$code.=<<___; + std %f0, [$out + `32*$i+0`] + aes_kexpand1 %f0, %f6, $i, %f0 + std %f2, [$out + `32*$i+8`] + aes_kexpand2 %f2, %f0, %f2 + std %f4, [$out + `32*$i+16`] + std %f6, [$out + `32*$i+24`] + std %f0, [$out + `32*$i+32`] + std %f2, [$out + `32*$i+40`] + + mov 14, $tmp + st $tmp, [$out + 240] + retl + xor %o0, %o0, %o0 + +.align 16 +.L192: + brz,pt $tmp, .L192aligned + nop + + ldd [$inp + 24], %f6 + faligndata %f0, %f2, %f0 + faligndata %f2, %f4, %f2 + faligndata %f4, %f6, %f4 +.L192aligned: +___ +for ($i=0; $i<7; $i++) { + $code.=<<___; + std %f0, [$out + `24*$i+0`] + aes_kexpand1 %f0, %f4, $i, %f0 + std %f2, [$out + `24*$i+8`] + aes_kexpand2 %f2, %f0, %f2 + std %f4, [$out + `24*$i+16`] + aes_kexpand2 %f4, %f2, %f4 +___ +} +$code.=<<___; + std %f0, [$out + `24*$i+0`] + aes_kexpand1 %f0, %f4, $i, %f0 + std %f2, [$out + `24*$i+8`] + aes_kexpand2 %f2, %f0, %f2 + std %f4, [$out + `24*$i+16`] + std %f0, [$out + `24*$i+24`] + std %f2, [$out + `24*$i+32`] + + mov 12, $tmp + st $tmp, [$out + 240] + retl + xor %o0, %o0, %o0 + +.align 16 +.L128: + brz,pt $tmp, .L128aligned + nop + + ldd [$inp + 16], %f4 + faligndata %f0, %f2, %f0 + faligndata %f2, %f4, %f2 +.L128aligned: +___ +for ($i=0; $i<10; $i++) { + $code.=<<___; + std %f0, [$out + `16*$i+0`] + aes_kexpand1 %f0, %f2, $i, %f0 + std %f2, [$out + `16*$i+8`] + aes_kexpand2 %f2, %f0, %f2 +___ +} +$code.=<<___; + std %f0, [$out + `16*$i+0`] + std %f2, [$out + `16*$i+8`] + + mov 10, $tmp + st $tmp, [$out + 240] + retl + xor %o0, %o0, %o0 +.type aes_t4_set_encrypt_key,#function +.size aes_t4_set_encrypt_key,.-aes_t4_set_encrypt_key + +.globl aes_t4_set_decrypt_key +.align 32 +aes_t4_set_decrypt_key: + mov %o7, %o5 + call .Lset_encrypt_key + nop + + mov %o5, %o7 + sll $tmp, 4, $inp ! $tmp is number of rounds + add $tmp, 2, $tmp + add $out, $inp, $inp ! $inp=$out+16*rounds + srl $tmp, 2, $tmp ! $tmp=(rounds+2)/4 + +.Lkey_flip: + ldd [$out + 0], %f0 + ldd [$out + 8], %f2 + ldd [$out + 16], %f4 + ldd [$out + 24], %f6 + ldd [$inp + 0], %f8 + ldd [$inp + 8], %f10 + ldd [$inp - 16], %f12 + ldd [$inp - 8], %f14 + sub $tmp, 1, $tmp + std %f0, [$inp + 0] + std %f2, [$inp + 8] + std %f4, [$inp - 16] + std %f6, [$inp - 8] + std %f8, [$out + 0] + std %f10, [$out + 8] + std %f12, [$out + 16] + std %f14, [$out + 24] + add $out, 32, $out + brnz $tmp, .Lkey_flip + sub $inp, 32, $inp + + retl + xor %o0, %o0, %o0 +.type aes_t4_set_decrypt_key,#function +.size aes_t4_set_decrypt_key,.-aes_t4_set_decrypt_key +___ +} + +{{{ +my ($inp,$out,$len,$key,$ivec,$enc)=map("%i$_",(0..5)); +my ($ileft,$iright,$ooff,$omask,$ivoff)=map("%l$_",(1..7)); + +$code.=<<___; +.align 32 +_aes128_encrypt_1x: +___ +for ($i=0; $i<4; $i++) { + $code.=<<___; + aes_eround01 %f`16+8*$i+0`, %f0, %f2, %f4 + aes_eround23 %f`16+8*$i+2`, %f0, %f2, %f2 + aes_eround01 %f`16+8*$i+4`, %f4, %f2, %f0 + aes_eround23 %f`16+8*$i+6`, %f4, %f2, %f2 +___ +} +$code.=<<___; + aes_eround01 %f48, %f0, %f2, %f4 + aes_eround23 %f50, %f0, %f2, %f2 + aes_eround01_l %f52, %f4, %f2, %f0 + retl + aes_eround23_l %f54, %f4, %f2, %f2 +.type _aes128_encrypt_1x,#function +.size _aes128_encrypt_1x,.-_aes128_encrypt_1x + +.align 32 +_aes128_encrypt_2x: +___ +for ($i=0; $i<4; $i++) { + $code.=<<___; + aes_eround01 %f`16+8*$i+0`, %f0, %f2, %f8 + aes_eround23 %f`16+8*$i+2`, %f0, %f2, %f2 + aes_eround01 %f`16+8*$i+0`, %f4, %f6, %f10 + aes_eround23 %f`16+8*$i+2`, %f4, %f6, %f6 + aes_eround01 %f`16+8*$i+4`, %f8, %f2, %f0 + aes_eround23 %f`16+8*$i+6`, %f8, %f2, %f2 + aes_eround01 %f`16+8*$i+4`, %f10, %f6, %f4 + aes_eround23 %f`16+8*$i+6`, %f10, %f6, %f6 +___ +} +$code.=<<___; + aes_eround01 %f48, %f0, %f2, %f8 + aes_eround23 %f50, %f0, %f2, %f2 + aes_eround01 %f48, %f4, %f6, %f10 + aes_eround23 %f50, %f4, %f6, %f6 + aes_eround01_l %f52, %f8, %f2, %f0 + aes_eround23_l %f54, %f8, %f2, %f2 + aes_eround01_l %f52, %f10, %f6, %f4 + retl + aes_eround23_l %f54, %f10, %f6, %f6 +.type _aes128_encrypt_2x,#function +.size _aes128_encrypt_2x,.-_aes128_encrypt_2x + +.align 32 +_aes128_loadkey: + ldx [$key + 0], %g4 + ldx [$key + 8], %g5 +___ +for ($i=2; $i<22;$i++) { # load key schedule + $code.=<<___; + ldd [$key + `8*$i`], %f`12+2*$i` +___ +} +$code.=<<___; + retl + nop +.type _aes128_loadkey,#function +.size _aes128_loadkey,.-_aes128_loadkey +_aes128_load_enckey=_aes128_loadkey +_aes128_load_deckey=_aes128_loadkey + +___ + +&alg_cbc_encrypt_implement("aes",128); +if ($::evp) { + &alg_ctr32_implement("aes",128); + &alg_xts_implement("aes",128,"en"); + &alg_xts_implement("aes",128,"de"); +} +&alg_cbc_decrypt_implement("aes",128); + +$code.=<<___; +.align 32 +_aes128_decrypt_1x: +___ +for ($i=0; $i<4; $i++) { + $code.=<<___; + aes_dround01 %f`16+8*$i+0`, %f0, %f2, %f4 + aes_dround23 %f`16+8*$i+2`, %f0, %f2, %f2 + aes_dround01 %f`16+8*$i+4`, %f4, %f2, %f0 + aes_dround23 %f`16+8*$i+6`, %f4, %f2, %f2 +___ +} +$code.=<<___; + aes_dround01 %f48, %f0, %f2, %f4 + aes_dround23 %f50, %f0, %f2, %f2 + aes_dround01_l %f52, %f4, %f2, %f0 + retl + aes_dround23_l %f54, %f4, %f2, %f2 +.type _aes128_decrypt_1x,#function +.size _aes128_decrypt_1x,.-_aes128_decrypt_1x + +.align 32 +_aes128_decrypt_2x: +___ +for ($i=0; $i<4; $i++) { + $code.=<<___; + aes_dround01 %f`16+8*$i+0`, %f0, %f2, %f8 + aes_dround23 %f`16+8*$i+2`, %f0, %f2, %f2 + aes_dround01 %f`16+8*$i+0`, %f4, %f6, %f10 + aes_dround23 %f`16+8*$i+2`, %f4, %f6, %f6 + aes_dround01 %f`16+8*$i+4`, %f8, %f2, %f0 + aes_dround23 %f`16+8*$i+6`, %f8, %f2, %f2 + aes_dround01 %f`16+8*$i+4`, %f10, %f6, %f4 + aes_dround23 %f`16+8*$i+6`, %f10, %f6, %f6 +___ +} +$code.=<<___; + aes_dround01 %f48, %f0, %f2, %f8 + aes_dround23 %f50, %f0, %f2, %f2 + aes_dround01 %f48, %f4, %f6, %f10 + aes_dround23 %f50, %f4, %f6, %f6 + aes_dround01_l %f52, %f8, %f2, %f0 + aes_dround23_l %f54, %f8, %f2, %f2 + aes_dround01_l %f52, %f10, %f6, %f4 + retl + aes_dround23_l %f54, %f10, %f6, %f6 +.type _aes128_decrypt_2x,#function +.size _aes128_decrypt_2x,.-_aes128_decrypt_2x +___ + +$code.=<<___; +.align 32 +_aes192_encrypt_1x: +___ +for ($i=0; $i<5; $i++) { + $code.=<<___; + aes_eround01 %f`16+8*$i+0`, %f0, %f2, %f4 + aes_eround23 %f`16+8*$i+2`, %f0, %f2, %f2 + aes_eround01 %f`16+8*$i+4`, %f4, %f2, %f0 + aes_eround23 %f`16+8*$i+6`, %f4, %f2, %f2 +___ +} +$code.=<<___; + aes_eround01 %f56, %f0, %f2, %f4 + aes_eround23 %f58, %f0, %f2, %f2 + aes_eround01_l %f60, %f4, %f2, %f0 + retl + aes_eround23_l %f62, %f4, %f2, %f2 +.type _aes192_encrypt_1x,#function +.size _aes192_encrypt_1x,.-_aes192_encrypt_1x + +.align 32 +_aes192_encrypt_2x: +___ +for ($i=0; $i<5; $i++) { + $code.=<<___; + aes_eround01 %f`16+8*$i+0`, %f0, %f2, %f8 + aes_eround23 %f`16+8*$i+2`, %f0, %f2, %f2 + aes_eround01 %f`16+8*$i+0`, %f4, %f6, %f10 + aes_eround23 %f`16+8*$i+2`, %f4, %f6, %f6 + aes_eround01 %f`16+8*$i+4`, %f8, %f2, %f0 + aes_eround23 %f`16+8*$i+6`, %f8, %f2, %f2 + aes_eround01 %f`16+8*$i+4`, %f10, %f6, %f4 + aes_eround23 %f`16+8*$i+6`, %f10, %f6, %f6 +___ +} +$code.=<<___; + aes_eround01 %f56, %f0, %f2, %f8 + aes_eround23 %f58, %f0, %f2, %f2 + aes_eround01 %f56, %f4, %f6, %f10 + aes_eround23 %f58, %f4, %f6, %f6 + aes_eround01_l %f60, %f8, %f2, %f0 + aes_eround23_l %f62, %f8, %f2, %f2 + aes_eround01_l %f60, %f10, %f6, %f4 + retl + aes_eround23_l %f62, %f10, %f6, %f6 +.type _aes192_encrypt_2x,#function +.size _aes192_encrypt_2x,.-_aes192_encrypt_2x + +.align 32 +_aes256_encrypt_1x: + aes_eround01 %f16, %f0, %f2, %f4 + aes_eround23 %f18, %f0, %f2, %f2 + ldd [$key + 208], %f16 + ldd [$key + 216], %f18 + aes_eround01 %f20, %f4, %f2, %f0 + aes_eround23 %f22, %f4, %f2, %f2 + ldd [$key + 224], %f20 + ldd [$key + 232], %f22 +___ +for ($i=1; $i<6; $i++) { + $code.=<<___; + aes_eround01 %f`16+8*$i+0`, %f0, %f2, %f4 + aes_eround23 %f`16+8*$i+2`, %f0, %f2, %f2 + aes_eround01 %f`16+8*$i+4`, %f4, %f2, %f0 + aes_eround23 %f`16+8*$i+6`, %f4, %f2, %f2 +___ +} +$code.=<<___; + aes_eround01 %f16, %f0, %f2, %f4 + aes_eround23 %f18, %f0, %f2, %f2 + ldd [$key + 16], %f16 + ldd [$key + 24], %f18 + aes_eround01_l %f20, %f4, %f2, %f0 + aes_eround23_l %f22, %f4, %f2, %f2 + ldd [$key + 32], %f20 + retl + ldd [$key + 40], %f22 +.type _aes256_encrypt_1x,#function +.size _aes256_encrypt_1x,.-_aes256_encrypt_1x + +.align 32 +_aes256_encrypt_2x: + aes_eround01 %f16, %f0, %f2, %f8 + aes_eround23 %f18, %f0, %f2, %f2 + aes_eround01 %f16, %f4, %f6, %f10 + aes_eround23 %f18, %f4, %f6, %f6 + ldd [$key + 208], %f16 + ldd [$key + 216], %f18 + aes_eround01 %f20, %f8, %f2, %f0 + aes_eround23 %f22, %f8, %f2, %f2 + aes_eround01 %f20, %f10, %f6, %f4 + aes_eround23 %f22, %f10, %f6, %f6 + ldd [$key + 224], %f20 + ldd [$key + 232], %f22 +___ +for ($i=1; $i<6; $i++) { + $code.=<<___; + aes_eround01 %f`16+8*$i+0`, %f0, %f2, %f8 + aes_eround23 %f`16+8*$i+2`, %f0, %f2, %f2 + aes_eround01 %f`16+8*$i+0`, %f4, %f6, %f10 + aes_eround23 %f`16+8*$i+2`, %f4, %f6, %f6 + aes_eround01 %f`16+8*$i+4`, %f8, %f2, %f0 + aes_eround23 %f`16+8*$i+6`, %f8, %f2, %f2 + aes_eround01 %f`16+8*$i+4`, %f10, %f6, %f4 + aes_eround23 %f`16+8*$i+6`, %f10, %f6, %f6 +___ +} +$code.=<<___; + aes_eround01 %f16, %f0, %f2, %f8 + aes_eround23 %f18, %f0, %f2, %f2 + aes_eround01 %f16, %f4, %f6, %f10 + aes_eround23 %f18, %f4, %f6, %f6 + ldd [$key + 16], %f16 + ldd [$key + 24], %f18 + aes_eround01_l %f20, %f8, %f2, %f0 + aes_eround23_l %f22, %f8, %f2, %f2 + aes_eround01_l %f20, %f10, %f6, %f4 + aes_eround23_l %f22, %f10, %f6, %f6 + ldd [$key + 32], %f20 + retl + ldd [$key + 40], %f22 +.type _aes256_encrypt_2x,#function +.size _aes256_encrypt_2x,.-_aes256_encrypt_2x + +.align 32 +_aes192_loadkey: + ldx [$key + 0], %g4 + ldx [$key + 8], %g5 +___ +for ($i=2; $i<26;$i++) { # load key schedule + $code.=<<___; + ldd [$key + `8*$i`], %f`12+2*$i` +___ +} +$code.=<<___; + retl + nop +.type _aes192_loadkey,#function +.size _aes192_loadkey,.-_aes192_loadkey +_aes256_loadkey=_aes192_loadkey +_aes192_load_enckey=_aes192_loadkey +_aes192_load_deckey=_aes192_loadkey +_aes256_load_enckey=_aes192_loadkey +_aes256_load_deckey=_aes192_loadkey +___ + +&alg_cbc_encrypt_implement("aes",256); +&alg_cbc_encrypt_implement("aes",192); +if ($::evp) { + &alg_ctr32_implement("aes",256); + &alg_xts_implement("aes",256,"en"); + &alg_xts_implement("aes",256,"de"); + &alg_ctr32_implement("aes",192); +} +&alg_cbc_decrypt_implement("aes",192); +&alg_cbc_decrypt_implement("aes",256); + +$code.=<<___; +.align 32 +_aes256_decrypt_1x: + aes_dround01 %f16, %f0, %f2, %f4 + aes_dround23 %f18, %f0, %f2, %f2 + ldd [$key + 208], %f16 + ldd [$key + 216], %f18 + aes_dround01 %f20, %f4, %f2, %f0 + aes_dround23 %f22, %f4, %f2, %f2 + ldd [$key + 224], %f20 + ldd [$key + 232], %f22 +___ +for ($i=1; $i<6; $i++) { + $code.=<<___; + aes_dround01 %f`16+8*$i+0`, %f0, %f2, %f4 + aes_dround23 %f`16+8*$i+2`, %f0, %f2, %f2 + aes_dround01 %f`16+8*$i+4`, %f4, %f2, %f0 + aes_dround23 %f`16+8*$i+6`, %f4, %f2, %f2 +___ +} +$code.=<<___; + aes_dround01 %f16, %f0, %f2, %f4 + aes_dround23 %f18, %f0, %f2, %f2 + ldd [$key + 16], %f16 + ldd [$key + 24], %f18 + aes_dround01_l %f20, %f4, %f2, %f0 + aes_dround23_l %f22, %f4, %f2, %f2 + ldd [$key + 32], %f20 + retl + ldd [$key + 40], %f22 +.type _aes256_decrypt_1x,#function +.size _aes256_decrypt_1x,.-_aes256_decrypt_1x + +.align 32 +_aes256_decrypt_2x: + aes_dround01 %f16, %f0, %f2, %f8 + aes_dround23 %f18, %f0, %f2, %f2 + aes_dround01 %f16, %f4, %f6, %f10 + aes_dround23 %f18, %f4, %f6, %f6 + ldd [$key + 208], %f16 + ldd [$key + 216], %f18 + aes_dround01 %f20, %f8, %f2, %f0 + aes_dround23 %f22, %f8, %f2, %f2 + aes_dround01 %f20, %f10, %f6, %f4 + aes_dround23 %f22, %f10, %f6, %f6 + ldd [$key + 224], %f20 + ldd [$key + 232], %f22 +___ +for ($i=1; $i<6; $i++) { + $code.=<<___; + aes_dround01 %f`16+8*$i+0`, %f0, %f2, %f8 + aes_dround23 %f`16+8*$i+2`, %f0, %f2, %f2 + aes_dround01 %f`16+8*$i+0`, %f4, %f6, %f10 + aes_dround23 %f`16+8*$i+2`, %f4, %f6, %f6 + aes_dround01 %f`16+8*$i+4`, %f8, %f2, %f0 + aes_dround23 %f`16+8*$i+6`, %f8, %f2, %f2 + aes_dround01 %f`16+8*$i+4`, %f10, %f6, %f4 + aes_dround23 %f`16+8*$i+6`, %f10, %f6, %f6 +___ +} +$code.=<<___; + aes_dround01 %f16, %f0, %f2, %f8 + aes_dround23 %f18, %f0, %f2, %f2 + aes_dround01 %f16, %f4, %f6, %f10 + aes_dround23 %f18, %f4, %f6, %f6 + ldd [$key + 16], %f16 + ldd [$key + 24], %f18 + aes_dround01_l %f20, %f8, %f2, %f0 + aes_dround23_l %f22, %f8, %f2, %f2 + aes_dround01_l %f20, %f10, %f6, %f4 + aes_dround23_l %f22, %f10, %f6, %f6 + ldd [$key + 32], %f20 + retl + ldd [$key + 40], %f22 +.type _aes256_decrypt_2x,#function +.size _aes256_decrypt_2x,.-_aes256_decrypt_2x + +.align 32 +_aes192_decrypt_1x: +___ +for ($i=0; $i<5; $i++) { + $code.=<<___; + aes_dround01 %f`16+8*$i+0`, %f0, %f2, %f4 + aes_dround23 %f`16+8*$i+2`, %f0, %f2, %f2 + aes_dround01 %f`16+8*$i+4`, %f4, %f2, %f0 + aes_dround23 %f`16+8*$i+6`, %f4, %f2, %f2 +___ +} +$code.=<<___; + aes_dround01 %f56, %f0, %f2, %f4 + aes_dround23 %f58, %f0, %f2, %f2 + aes_dround01_l %f60, %f4, %f2, %f0 + retl + aes_dround23_l %f62, %f4, %f2, %f2 +.type _aes192_decrypt_1x,#function +.size _aes192_decrypt_1x,.-_aes192_decrypt_1x + +.align 32 +_aes192_decrypt_2x: +___ +for ($i=0; $i<5; $i++) { + $code.=<<___; + aes_dround01 %f`16+8*$i+0`, %f0, %f2, %f8 + aes_dround23 %f`16+8*$i+2`, %f0, %f2, %f2 + aes_dround01 %f`16+8*$i+0`, %f4, %f6, %f10 + aes_dround23 %f`16+8*$i+2`, %f4, %f6, %f6 + aes_dround01 %f`16+8*$i+4`, %f8, %f2, %f0 + aes_dround23 %f`16+8*$i+6`, %f8, %f2, %f2 + aes_dround01 %f`16+8*$i+4`, %f10, %f6, %f4 + aes_dround23 %f`16+8*$i+6`, %f10, %f6, %f6 +___ +} +$code.=<<___; + aes_dround01 %f56, %f0, %f2, %f8 + aes_dround23 %f58, %f0, %f2, %f2 + aes_dround01 %f56, %f4, %f6, %f10 + aes_dround23 %f58, %f4, %f6, %f6 + aes_dround01_l %f60, %f8, %f2, %f0 + aes_dround23_l %f62, %f8, %f2, %f2 + aes_dround01_l %f60, %f10, %f6, %f4 + retl + aes_dround23_l %f62, %f10, %f6, %f6 +.type _aes192_decrypt_2x,#function +.size _aes192_decrypt_2x,.-_aes192_decrypt_2x +___ +}}} + +if (!$::evp) { +$code.=<<___; +.global AES_encrypt +AES_encrypt=aes_t4_encrypt +.global AES_decrypt +AES_decrypt=aes_t4_decrypt +.global AES_set_encrypt_key +.align 32 +AES_set_encrypt_key: + andcc %o2, 7, %g0 ! check alignment + bnz,a,pn %icc, 1f + mov -1, %o0 + brz,a,pn %o0, 1f + mov -1, %o0 + brz,a,pn %o2, 1f + mov -1, %o0 + andncc %o1, 0x1c0, %g0 + bnz,a,pn %icc, 1f + mov -2, %o0 + cmp %o1, 128 + bl,a,pn %icc, 1f + mov -2, %o0 + b aes_t4_set_encrypt_key + nop +1: retl + nop +.type AES_set_encrypt_key,#function +.size AES_set_encrypt_key,.-AES_set_encrypt_key + +.global AES_set_decrypt_key +.align 32 +AES_set_decrypt_key: + andcc %o2, 7, %g0 ! check alignment + bnz,a,pn %icc, 1f + mov -1, %o0 + brz,a,pn %o0, 1f + mov -1, %o0 + brz,a,pn %o2, 1f + mov -1, %o0 + andncc %o1, 0x1c0, %g0 + bnz,a,pn %icc, 1f + mov -2, %o0 + cmp %o1, 128 + bl,a,pn %icc, 1f + mov -2, %o0 + b aes_t4_set_decrypt_key + nop +1: retl + nop +.type AES_set_decrypt_key,#function +.size AES_set_decrypt_key,.-AES_set_decrypt_key +___ + +my ($inp,$out,$len,$key,$ivec,$enc)=map("%o$_",(0..5)); + +$code.=<<___; +.globl AES_cbc_encrypt +.align 32 +AES_cbc_encrypt: + ld [$key + 240], %g1 + nop + brz $enc, .Lcbc_decrypt + cmp %g1, 12 + + bl,pt %icc, aes128_t4_cbc_encrypt + nop + be,pn %icc, aes192_t4_cbc_encrypt + nop + ba aes256_t4_cbc_encrypt + nop + +.Lcbc_decrypt: + bl,pt %icc, aes128_t4_cbc_decrypt + nop + be,pn %icc, aes192_t4_cbc_decrypt + nop + ba aes256_t4_cbc_decrypt + nop +.type AES_cbc_encrypt,#function +.size AES_cbc_encrypt,.-AES_cbc_encrypt +___ +} +$code.=<<___; +.asciz "AES for SPARC T4, David S. Miller, Andy Polyakov" +.align 4 +___ + +&emit_assembler(); + +close STDOUT; diff --git a/openssl-1.1.0h/crypto/aes/asm/aesv8-armx.pl b/openssl-1.1.0h/crypto/aes/asm/aesv8-armx.pl new file mode 100755 index 0000000..1782d5b --- /dev/null +++ b/openssl-1.1.0h/crypto/aes/asm/aesv8-armx.pl @@ -0,0 +1,1008 @@ +#! /usr/bin/env perl +# Copyright 2014-2016 The OpenSSL Project Authors. All Rights Reserved. +# +# Licensed under the OpenSSL license (the "License"). You may not use +# this file except in compliance with the License. You can obtain a copy +# in the file LICENSE in the source distribution or at +# https://www.openssl.org/source/license.html + +# +# ==================================================================== +# Written by Andy Polyakov for the OpenSSL +# project. The module is, however, dual licensed under OpenSSL and +# CRYPTOGAMS licenses depending on where you obtain it. For further +# details see http://www.openssl.org/~appro/cryptogams/. +# ==================================================================== +# +# This module implements support for ARMv8 AES instructions. The +# module is endian-agnostic in sense that it supports both big- and +# little-endian cases. As does it support both 32- and 64-bit modes +# of operation. Latter is achieved by limiting amount of utilized +# registers to 16, which implies additional NEON load and integer +# instructions. This has no effect on mighty Apple A7, where results +# are literally equal to the theoretical estimates based on AES +# instruction latencies and issue rates. On Cortex-A53, an in-order +# execution core, this costs up to 10-15%, which is partially +# compensated by implementing dedicated code path for 128-bit +# CBC encrypt case. On Cortex-A57 parallelizable mode performance +# seems to be limited by sheer amount of NEON instructions... +# +# Performance in cycles per byte processed with 128-bit key: +# +# CBC enc CBC dec CTR +# Apple A7 2.39 1.20 1.20 +# Cortex-A53 1.32 1.29 1.46 +# Cortex-A57(*) 1.95 0.85 0.93 +# Denver 1.96 0.86 0.80 +# Mongoose 1.33 1.20 1.20 +# +# (*) original 3.64/1.34/1.32 results were for r0p0 revision +# and are still same even for updated module; + +$flavour = shift; +$output = shift; + +$0 =~ m/(.*[\/\\])[^\/\\]+$/; $dir=$1; +( $xlate="${dir}arm-xlate.pl" and -f $xlate ) or +( $xlate="${dir}../../perlasm/arm-xlate.pl" and -f $xlate) or +die "can't locate arm-xlate.pl"; + +open OUT,"| \"$^X\" $xlate $flavour $output"; +*STDOUT=*OUT; + +$prefix="aes_v8"; + +$code=<<___; +#include "arm_arch.h" + +#if __ARM_MAX_ARCH__>=7 +.text +___ +$code.=".arch armv8-a+crypto\n" if ($flavour =~ /64/); +$code.=<<___ if ($flavour !~ /64/); +.arch armv7-a // don't confuse not-so-latest binutils with argv8 :-) +.fpu neon +.code 32 +#undef __thumb2__ +___ + +# Assembler mnemonics are an eclectic mix of 32- and 64-bit syntax, +# NEON is mostly 32-bit mnemonics, integer - mostly 64. Goal is to +# maintain both 32- and 64-bit codes within single module and +# transliterate common code to either flavour with regex vodoo. +# +{{{ +my ($inp,$bits,$out,$ptr,$rounds)=("x0","w1","x2","x3","w12"); +my ($zero,$rcon,$mask,$in0,$in1,$tmp,$key)= + $flavour=~/64/? map("q$_",(0..6)) : map("q$_",(0..3,8..10)); + + +$code.=<<___; +.align 5 +.Lrcon: +.long 0x01,0x01,0x01,0x01 +.long 0x0c0f0e0d,0x0c0f0e0d,0x0c0f0e0d,0x0c0f0e0d // rotate-n-splat +.long 0x1b,0x1b,0x1b,0x1b + +.globl ${prefix}_set_encrypt_key +.type ${prefix}_set_encrypt_key,%function +.align 5 +${prefix}_set_encrypt_key: +.Lenc_key: +___ +$code.=<<___ if ($flavour =~ /64/); + stp x29,x30,[sp,#-16]! + add x29,sp,#0 +___ +$code.=<<___; + mov $ptr,#-1 + cmp $inp,#0 + b.eq .Lenc_key_abort + cmp $out,#0 + b.eq .Lenc_key_abort + mov $ptr,#-2 + cmp $bits,#128 + b.lt .Lenc_key_abort + cmp $bits,#256 + b.gt .Lenc_key_abort + tst $bits,#0x3f + b.ne .Lenc_key_abort + + adr $ptr,.Lrcon + cmp $bits,#192 + + veor $zero,$zero,$zero + vld1.8 {$in0},[$inp],#16 + mov $bits,#8 // reuse $bits + vld1.32 {$rcon,$mask},[$ptr],#32 + + b.lt .Loop128 + b.eq .L192 + b .L256 + +.align 4 +.Loop128: + vtbl.8 $key,{$in0},$mask + vext.8 $tmp,$zero,$in0,#12 + vst1.32 {$in0},[$out],#16 + aese $key,$zero + subs $bits,$bits,#1 + + veor $in0,$in0,$tmp + vext.8 $tmp,$zero,$tmp,#12 + veor $in0,$in0,$tmp + vext.8 $tmp,$zero,$tmp,#12 + veor $key,$key,$rcon + veor $in0,$in0,$tmp + vshl.u8 $rcon,$rcon,#1 + veor $in0,$in0,$key + b.ne .Loop128 + + vld1.32 {$rcon},[$ptr] + + vtbl.8 $key,{$in0},$mask + vext.8 $tmp,$zero,$in0,#12 + vst1.32 {$in0},[$out],#16 + aese $key,$zero + + veor $in0,$in0,$tmp + vext.8 $tmp,$zero,$tmp,#12 + veor $in0,$in0,$tmp + vext.8 $tmp,$zero,$tmp,#12 + veor $key,$key,$rcon + veor $in0,$in0,$tmp + vshl.u8 $rcon,$rcon,#1 + veor $in0,$in0,$key + + vtbl.8 $key,{$in0},$mask + vext.8 $tmp,$zero,$in0,#12 + vst1.32 {$in0},[$out],#16 + aese $key,$zero + + veor $in0,$in0,$tmp + vext.8 $tmp,$zero,$tmp,#12 + veor $in0,$in0,$tmp + vext.8 $tmp,$zero,$tmp,#12 + veor $key,$key,$rcon + veor $in0,$in0,$tmp + veor $in0,$in0,$key + vst1.32 {$in0},[$out] + add $out,$out,#0x50 + + mov $rounds,#10 + b .Ldone + +.align 4 +.L192: + vld1.8 {$in1},[$inp],#8 + vmov.i8 $key,#8 // borrow $key + vst1.32 {$in0},[$out],#16 + vsub.i8 $mask,$mask,$key // adjust the mask + +.Loop192: + vtbl.8 $key,{$in1},$mask + vext.8 $tmp,$zero,$in0,#12 + vst1.32 {$in1},[$out],#8 + aese $key,$zero + subs $bits,$bits,#1 + + veor $in0,$in0,$tmp + vext.8 $tmp,$zero,$tmp,#12 + veor $in0,$in0,$tmp + vext.8 $tmp,$zero,$tmp,#12 + veor $in0,$in0,$tmp + + vdup.32 $tmp,${in0}[3] + veor $tmp,$tmp,$in1 + veor $key,$key,$rcon + vext.8 $in1,$zero,$in1,#12 + vshl.u8 $rcon,$rcon,#1 + veor $in1,$in1,$tmp + veor $in0,$in0,$key + veor $in1,$in1,$key + vst1.32 {$in0},[$out],#16 + b.ne .Loop192 + + mov $rounds,#12 + add $out,$out,#0x20 + b .Ldone + +.align 4 +.L256: + vld1.8 {$in1},[$inp] + mov $bits,#7 + mov $rounds,#14 + vst1.32 {$in0},[$out],#16 + +.Loop256: + vtbl.8 $key,{$in1},$mask + vext.8 $tmp,$zero,$in0,#12 + vst1.32 {$in1},[$out],#16 + aese $key,$zero + subs $bits,$bits,#1 + + veor $in0,$in0,$tmp + vext.8 $tmp,$zero,$tmp,#12 + veor $in0,$in0,$tmp + vext.8 $tmp,$zero,$tmp,#12 + veor $key,$key,$rcon + veor $in0,$in0,$tmp + vshl.u8 $rcon,$rcon,#1 + veor $in0,$in0,$key + vst1.32 {$in0},[$out],#16 + b.eq .Ldone + + vdup.32 $key,${in0}[3] // just splat + vext.8 $tmp,$zero,$in1,#12 + aese $key,$zero + + veor $in1,$in1,$tmp + vext.8 $tmp,$zero,$tmp,#12 + veor $in1,$in1,$tmp + vext.8 $tmp,$zero,$tmp,#12 + veor $in1,$in1,$tmp + + veor $in1,$in1,$key + b .Loop256 + +.Ldone: + str $rounds,[$out] + mov $ptr,#0 + +.Lenc_key_abort: + mov x0,$ptr // return value + `"ldr x29,[sp],#16" if ($flavour =~ /64/)` + ret +.size ${prefix}_set_encrypt_key,.-${prefix}_set_encrypt_key + +.globl ${prefix}_set_decrypt_key +.type ${prefix}_set_decrypt_key,%function +.align 5 +${prefix}_set_decrypt_key: +___ +$code.=<<___ if ($flavour =~ /64/); + stp x29,x30,[sp,#-16]! + add x29,sp,#0 +___ +$code.=<<___ if ($flavour !~ /64/); + stmdb sp!,{r4,lr} +___ +$code.=<<___; + bl .Lenc_key + + cmp x0,#0 + b.ne .Ldec_key_abort + + sub $out,$out,#240 // restore original $out + mov x4,#-16 + add $inp,$out,x12,lsl#4 // end of key schedule + + vld1.32 {v0.16b},[$out] + vld1.32 {v1.16b},[$inp] + vst1.32 {v0.16b},[$inp],x4 + vst1.32 {v1.16b},[$out],#16 + +.Loop_imc: + vld1.32 {v0.16b},[$out] + vld1.32 {v1.16b},[$inp] + aesimc v0.16b,v0.16b + aesimc v1.16b,v1.16b + vst1.32 {v0.16b},[$inp],x4 + vst1.32 {v1.16b},[$out],#16 + cmp $inp,$out + b.hi .Loop_imc + + vld1.32 {v0.16b},[$out] + aesimc v0.16b,v0.16b + vst1.32 {v0.16b},[$inp] + + eor x0,x0,x0 // return value +.Ldec_key_abort: +___ +$code.=<<___ if ($flavour !~ /64/); + ldmia sp!,{r4,pc} +___ +$code.=<<___ if ($flavour =~ /64/); + ldp x29,x30,[sp],#16 + ret +___ +$code.=<<___; +.size ${prefix}_set_decrypt_key,.-${prefix}_set_decrypt_key +___ +}}} +{{{ +sub gen_block () { +my $dir = shift; +my ($e,$mc) = $dir eq "en" ? ("e","mc") : ("d","imc"); +my ($inp,$out,$key)=map("x$_",(0..2)); +my $rounds="w3"; +my ($rndkey0,$rndkey1,$inout)=map("q$_",(0..3)); + +$code.=<<___; +.globl ${prefix}_${dir}crypt +.type ${prefix}_${dir}crypt,%function +.align 5 +${prefix}_${dir}crypt: + ldr $rounds,[$key,#240] + vld1.32 {$rndkey0},[$key],#16 + vld1.8 {$inout},[$inp] + sub $rounds,$rounds,#2 + vld1.32 {$rndkey1},[$key],#16 + +.Loop_${dir}c: + aes$e $inout,$rndkey0 + aes$mc $inout,$inout + vld1.32 {$rndkey0},[$key],#16 + subs $rounds,$rounds,#2 + aes$e $inout,$rndkey1 + aes$mc $inout,$inout + vld1.32 {$rndkey1},[$key],#16 + b.gt .Loop_${dir}c + + aes$e $inout,$rndkey0 + aes$mc $inout,$inout + vld1.32 {$rndkey0},[$key] + aes$e $inout,$rndkey1 + veor $inout,$inout,$rndkey0 + + vst1.8 {$inout},[$out] + ret +.size ${prefix}_${dir}crypt,.-${prefix}_${dir}crypt +___ +} +&gen_block("en"); +&gen_block("de"); +}}} +{{{ +my ($inp,$out,$len,$key,$ivp)=map("x$_",(0..4)); my $enc="w5"; +my ($rounds,$cnt,$key_,$step,$step1)=($enc,"w6","x7","x8","x12"); +my ($dat0,$dat1,$in0,$in1,$tmp0,$tmp1,$ivec,$rndlast)=map("q$_",(0..7)); + +my ($dat,$tmp,$rndzero_n_last)=($dat0,$tmp0,$tmp1); +my ($key4,$key5,$key6,$key7)=("x6","x12","x14",$key); + +### q8-q15 preloaded key schedule + +$code.=<<___; +.globl ${prefix}_cbc_encrypt +.type ${prefix}_cbc_encrypt,%function +.align 5 +${prefix}_cbc_encrypt: +___ +$code.=<<___ if ($flavour =~ /64/); + stp x29,x30,[sp,#-16]! + add x29,sp,#0 +___ +$code.=<<___ if ($flavour !~ /64/); + mov ip,sp + stmdb sp!,{r4-r8,lr} + vstmdb sp!,{d8-d15} @ ABI specification says so + ldmia ip,{r4-r5} @ load remaining args +___ +$code.=<<___; + subs $len,$len,#16 + mov $step,#16 + b.lo .Lcbc_abort + cclr $step,eq + + cmp $enc,#0 // en- or decrypting? + ldr $rounds,[$key,#240] + and $len,$len,#-16 + vld1.8 {$ivec},[$ivp] + vld1.8 {$dat},[$inp],$step + + vld1.32 {q8-q9},[$key] // load key schedule... + sub $rounds,$rounds,#6 + add $key_,$key,x5,lsl#4 // pointer to last 7 round keys + sub $rounds,$rounds,#2 + vld1.32 {q10-q11},[$key_],#32 + vld1.32 {q12-q13},[$key_],#32 + vld1.32 {q14-q15},[$key_],#32 + vld1.32 {$rndlast},[$key_] + + add $key_,$key,#32 + mov $cnt,$rounds + b.eq .Lcbc_dec + + cmp $rounds,#2 + veor $dat,$dat,$ivec + veor $rndzero_n_last,q8,$rndlast + b.eq .Lcbc_enc128 + + vld1.32 {$in0-$in1},[$key_] + add $key_,$key,#16 + add $key4,$key,#16*4 + add $key5,$key,#16*5 + aese $dat,q8 + aesmc $dat,$dat + add $key6,$key,#16*6 + add $key7,$key,#16*7 + b .Lenter_cbc_enc + +.align 4 +.Loop_cbc_enc: + aese $dat,q8 + aesmc $dat,$dat + vst1.8 {$ivec},[$out],#16 +.Lenter_cbc_enc: + aese $dat,q9 + aesmc $dat,$dat + aese $dat,$in0 + aesmc $dat,$dat + vld1.32 {q8},[$key4] + cmp $rounds,#4 + aese $dat,$in1 + aesmc $dat,$dat + vld1.32 {q9},[$key5] + b.eq .Lcbc_enc192 + + aese $dat,q8 + aesmc $dat,$dat + vld1.32 {q8},[$key6] + aese $dat,q9 + aesmc $dat,$dat + vld1.32 {q9},[$key7] + nop + +.Lcbc_enc192: + aese $dat,q8 + aesmc $dat,$dat + subs $len,$len,#16 + aese $dat,q9 + aesmc $dat,$dat + cclr $step,eq + aese $dat,q10 + aesmc $dat,$dat + aese $dat,q11 + aesmc $dat,$dat + vld1.8 {q8},[$inp],$step + aese $dat,q12 + aesmc $dat,$dat + veor q8,q8,$rndzero_n_last + aese $dat,q13 + aesmc $dat,$dat + vld1.32 {q9},[$key_] // re-pre-load rndkey[1] + aese $dat,q14 + aesmc $dat,$dat + aese $dat,q15 + veor $ivec,$dat,$rndlast + b.hs .Loop_cbc_enc + + vst1.8 {$ivec},[$out],#16 + b .Lcbc_done + +.align 5 +.Lcbc_enc128: + vld1.32 {$in0-$in1},[$key_] + aese $dat,q8 + aesmc $dat,$dat + b .Lenter_cbc_enc128 +.Loop_cbc_enc128: + aese $dat,q8 + aesmc $dat,$dat + vst1.8 {$ivec},[$out],#16 +.Lenter_cbc_enc128: + aese $dat,q9 + aesmc $dat,$dat + subs $len,$len,#16 + aese $dat,$in0 + aesmc $dat,$dat + cclr $step,eq + aese $dat,$in1 + aesmc $dat,$dat + aese $dat,q10 + aesmc $dat,$dat + aese $dat,q11 + aesmc $dat,$dat + vld1.8 {q8},[$inp],$step + aese $dat,q12 + aesmc $dat,$dat + aese $dat,q13 + aesmc $dat,$dat + aese $dat,q14 + aesmc $dat,$dat + veor q8,q8,$rndzero_n_last + aese $dat,q15 + veor $ivec,$dat,$rndlast + b.hs .Loop_cbc_enc128 + + vst1.8 {$ivec},[$out],#16 + b .Lcbc_done +___ +{ +my ($dat2,$in2,$tmp2)=map("q$_",(10,11,9)); +$code.=<<___; +.align 5 +.Lcbc_dec: + vld1.8 {$dat2},[$inp],#16 + subs $len,$len,#32 // bias + add $cnt,$rounds,#2 + vorr $in1,$dat,$dat + vorr $dat1,$dat,$dat + vorr $in2,$dat2,$dat2 + b.lo .Lcbc_dec_tail + + vorr $dat1,$dat2,$dat2 + vld1.8 {$dat2},[$inp],#16 + vorr $in0,$dat,$dat + vorr $in1,$dat1,$dat1 + vorr $in2,$dat2,$dat2 + +.Loop3x_cbc_dec: + aesd $dat0,q8 + aesimc $dat0,$dat0 + aesd $dat1,q8 + aesimc $dat1,$dat1 + aesd $dat2,q8 + aesimc $dat2,$dat2 + vld1.32 {q8},[$key_],#16 + subs $cnt,$cnt,#2 + aesd $dat0,q9 + aesimc $dat0,$dat0 + aesd $dat1,q9 + aesimc $dat1,$dat1 + aesd $dat2,q9 + aesimc $dat2,$dat2 + vld1.32 {q9},[$key_],#16 + b.gt .Loop3x_cbc_dec + + aesd $dat0,q8 + aesimc $dat0,$dat0 + aesd $dat1,q8 + aesimc $dat1,$dat1 + aesd $dat2,q8 + aesimc $dat2,$dat2 + veor $tmp0,$ivec,$rndlast + subs $len,$len,#0x30 + veor $tmp1,$in0,$rndlast + mov.lo x6,$len // x6, $cnt, is zero at this point + aesd $dat0,q9 + aesimc $dat0,$dat0 + aesd $dat1,q9 + aesimc $dat1,$dat1 + aesd $dat2,q9 + aesimc $dat2,$dat2 + veor $tmp2,$in1,$rndlast + add $inp,$inp,x6 // $inp is adjusted in such way that + // at exit from the loop $dat1-$dat2 + // are loaded with last "words" + vorr $ivec,$in2,$in2 + mov $key_,$key + aesd $dat0,q12 + aesimc $dat0,$dat0 + aesd $dat1,q12 + aesimc $dat1,$dat1 + aesd $dat2,q12 + aesimc $dat2,$dat2 + vld1.8 {$in0},[$inp],#16 + aesd $dat0,q13 + aesimc $dat0,$dat0 + aesd $dat1,q13 + aesimc $dat1,$dat1 + aesd $dat2,q13 + aesimc $dat2,$dat2 + vld1.8 {$in1},[$inp],#16 + aesd $dat0,q14 + aesimc $dat0,$dat0 + aesd $dat1,q14 + aesimc $dat1,$dat1 + aesd $dat2,q14 + aesimc $dat2,$dat2 + vld1.8 {$in2},[$inp],#16 + aesd $dat0,q15 + aesd $dat1,q15 + aesd $dat2,q15 + vld1.32 {q8},[$key_],#16 // re-pre-load rndkey[0] + add $cnt,$rounds,#2 + veor $tmp0,$tmp0,$dat0 + veor $tmp1,$tmp1,$dat1 + veor $dat2,$dat2,$tmp2 + vld1.32 {q9},[$key_],#16 // re-pre-load rndkey[1] + vst1.8 {$tmp0},[$out],#16 + vorr $dat0,$in0,$in0 + vst1.8 {$tmp1},[$out],#16 + vorr $dat1,$in1,$in1 + vst1.8 {$dat2},[$out],#16 + vorr $dat2,$in2,$in2 + b.hs .Loop3x_cbc_dec + + cmn $len,#0x30 + b.eq .Lcbc_done + nop + +.Lcbc_dec_tail: + aesd $dat1,q8 + aesimc $dat1,$dat1 + aesd $dat2,q8 + aesimc $dat2,$dat2 + vld1.32 {q8},[$key_],#16 + subs $cnt,$cnt,#2 + aesd $dat1,q9 + aesimc $dat1,$dat1 + aesd $dat2,q9 + aesimc $dat2,$dat2 + vld1.32 {q9},[$key_],#16 + b.gt .Lcbc_dec_tail + + aesd $dat1,q8 + aesimc $dat1,$dat1 + aesd $dat2,q8 + aesimc $dat2,$dat2 + aesd $dat1,q9 + aesimc $dat1,$dat1 + aesd $dat2,q9 + aesimc $dat2,$dat2 + aesd $dat1,q12 + aesimc $dat1,$dat1 + aesd $dat2,q12 + aesimc $dat2,$dat2 + cmn $len,#0x20 + aesd $dat1,q13 + aesimc $dat1,$dat1 + aesd $dat2,q13 + aesimc $dat2,$dat2 + veor $tmp1,$ivec,$rndlast + aesd $dat1,q14 + aesimc $dat1,$dat1 + aesd $dat2,q14 + aesimc $dat2,$dat2 + veor $tmp2,$in1,$rndlast + aesd $dat1,q15 + aesd $dat2,q15 + b.eq .Lcbc_dec_one + veor $tmp1,$tmp1,$dat1 + veor $tmp2,$tmp2,$dat2 + vorr $ivec,$in2,$in2 + vst1.8 {$tmp1},[$out],#16 + vst1.8 {$tmp2},[$out],#16 + b .Lcbc_done + +.Lcbc_dec_one: + veor $tmp1,$tmp1,$dat2 + vorr $ivec,$in2,$in2 + vst1.8 {$tmp1},[$out],#16 + +.Lcbc_done: + vst1.8 {$ivec},[$ivp] +.Lcbc_abort: +___ +} +$code.=<<___ if ($flavour !~ /64/); + vldmia sp!,{d8-d15} + ldmia sp!,{r4-r8,pc} +___ +$code.=<<___ if ($flavour =~ /64/); + ldr x29,[sp],#16 + ret +___ +$code.=<<___; +.size ${prefix}_cbc_encrypt,.-${prefix}_cbc_encrypt +___ +}}} +{{{ +my ($inp,$out,$len,$key,$ivp)=map("x$_",(0..4)); +my ($rounds,$cnt,$key_)=("w5","w6","x7"); +my ($ctr,$tctr0,$tctr1,$tctr2)=map("w$_",(8..10,12)); +my $step="x12"; # aliases with $tctr2 + +my ($dat0,$dat1,$in0,$in1,$tmp0,$tmp1,$ivec,$rndlast)=map("q$_",(0..7)); +my ($dat2,$in2,$tmp2)=map("q$_",(10,11,9)); + +my ($dat,$tmp)=($dat0,$tmp0); + +### q8-q15 preloaded key schedule + +$code.=<<___; +.globl ${prefix}_ctr32_encrypt_blocks +.type ${prefix}_ctr32_encrypt_blocks,%function +.align 5 +${prefix}_ctr32_encrypt_blocks: +___ +$code.=<<___ if ($flavour =~ /64/); + stp x29,x30,[sp,#-16]! + add x29,sp,#0 +___ +$code.=<<___ if ($flavour !~ /64/); + mov ip,sp + stmdb sp!,{r4-r10,lr} + vstmdb sp!,{d8-d15} @ ABI specification says so + ldr r4, [ip] @ load remaining arg +___ +$code.=<<___; + ldr $rounds,[$key,#240] + + ldr $ctr, [$ivp, #12] + vld1.32 {$dat0},[$ivp] + + vld1.32 {q8-q9},[$key] // load key schedule... + sub $rounds,$rounds,#4 + mov $step,#16 + cmp $len,#2 + add $key_,$key,x5,lsl#4 // pointer to last 5 round keys + sub $rounds,$rounds,#2 + vld1.32 {q12-q13},[$key_],#32 + vld1.32 {q14-q15},[$key_],#32 + vld1.32 {$rndlast},[$key_] + add $key_,$key,#32 + mov $cnt,$rounds + cclr $step,lo +#ifndef __ARMEB__ + rev $ctr, $ctr +#endif + vorr $dat1,$dat0,$dat0 + add $tctr1, $ctr, #1 + vorr $dat2,$dat0,$dat0 + add $ctr, $ctr, #2 + vorr $ivec,$dat0,$dat0 + rev $tctr1, $tctr1 + vmov.32 ${dat1}[3],$tctr1 + b.ls .Lctr32_tail + rev $tctr2, $ctr + sub $len,$len,#3 // bias + vmov.32 ${dat2}[3],$tctr2 + b .Loop3x_ctr32 + +.align 4 +.Loop3x_ctr32: + aese $dat0,q8 + aesmc $dat0,$dat0 + aese $dat1,q8 + aesmc $dat1,$dat1 + aese $dat2,q8 + aesmc $dat2,$dat2 + vld1.32 {q8},[$key_],#16 + subs $cnt,$cnt,#2 + aese $dat0,q9 + aesmc $dat0,$dat0 + aese $dat1,q9 + aesmc $dat1,$dat1 + aese $dat2,q9 + aesmc $dat2,$dat2 + vld1.32 {q9},[$key_],#16 + b.gt .Loop3x_ctr32 + + aese $dat0,q8 + aesmc $tmp0,$dat0 + aese $dat1,q8 + aesmc $tmp1,$dat1 + vld1.8 {$in0},[$inp],#16 + vorr $dat0,$ivec,$ivec + aese $dat2,q8 + aesmc $dat2,$dat2 + vld1.8 {$in1},[$inp],#16 + vorr $dat1,$ivec,$ivec + aese $tmp0,q9 + aesmc $tmp0,$tmp0 + aese $tmp1,q9 + aesmc $tmp1,$tmp1 + vld1.8 {$in2},[$inp],#16 + mov $key_,$key + aese $dat2,q9 + aesmc $tmp2,$dat2 + vorr $dat2,$ivec,$ivec + add $tctr0,$ctr,#1 + aese $tmp0,q12 + aesmc $tmp0,$tmp0 + aese $tmp1,q12 + aesmc $tmp1,$tmp1 + veor $in0,$in0,$rndlast + add $tctr1,$ctr,#2 + aese $tmp2,q12 + aesmc $tmp2,$tmp2 + veor $in1,$in1,$rndlast + add $ctr,$ctr,#3 + aese $tmp0,q13 + aesmc $tmp0,$tmp0 + aese $tmp1,q13 + aesmc $tmp1,$tmp1 + veor $in2,$in2,$rndlast + rev $tctr0,$tctr0 + aese $tmp2,q13 + aesmc $tmp2,$tmp2 + vmov.32 ${dat0}[3], $tctr0 + rev $tctr1,$tctr1 + aese $tmp0,q14 + aesmc $tmp0,$tmp0 + aese $tmp1,q14 + aesmc $tmp1,$tmp1 + vmov.32 ${dat1}[3], $tctr1 + rev $tctr2,$ctr + aese $tmp2,q14 + aesmc $tmp2,$tmp2 + vmov.32 ${dat2}[3], $tctr2 + subs $len,$len,#3 + aese $tmp0,q15 + aese $tmp1,q15 + aese $tmp2,q15 + + veor $in0,$in0,$tmp0 + vld1.32 {q8},[$key_],#16 // re-pre-load rndkey[0] + vst1.8 {$in0},[$out],#16 + veor $in1,$in1,$tmp1 + mov $cnt,$rounds + vst1.8 {$in1},[$out],#16 + veor $in2,$in2,$tmp2 + vld1.32 {q9},[$key_],#16 // re-pre-load rndkey[1] + vst1.8 {$in2},[$out],#16 + b.hs .Loop3x_ctr32 + + adds $len,$len,#3 + b.eq .Lctr32_done + cmp $len,#1 + mov $step,#16 + cclr $step,eq + +.Lctr32_tail: + aese $dat0,q8 + aesmc $dat0,$dat0 + aese $dat1,q8 + aesmc $dat1,$dat1 + vld1.32 {q8},[$key_],#16 + subs $cnt,$cnt,#2 + aese $dat0,q9 + aesmc $dat0,$dat0 + aese $dat1,q9 + aesmc $dat1,$dat1 + vld1.32 {q9},[$key_],#16 + b.gt .Lctr32_tail + + aese $dat0,q8 + aesmc $dat0,$dat0 + aese $dat1,q8 + aesmc $dat1,$dat1 + aese $dat0,q9 + aesmc $dat0,$dat0 + aese $dat1,q9 + aesmc $dat1,$dat1 + vld1.8 {$in0},[$inp],$step + aese $dat0,q12 + aesmc $dat0,$dat0 + aese $dat1,q12 + aesmc $dat1,$dat1 + vld1.8 {$in1},[$inp] + aese $dat0,q13 + aesmc $dat0,$dat0 + aese $dat1,q13 + aesmc $dat1,$dat1 + veor $in0,$in0,$rndlast + aese $dat0,q14 + aesmc $dat0,$dat0 + aese $dat1,q14 + aesmc $dat1,$dat1 + veor $in1,$in1,$rndlast + aese $dat0,q15 + aese $dat1,q15 + + cmp $len,#1 + veor $in0,$in0,$dat0 + veor $in1,$in1,$dat1 + vst1.8 {$in0},[$out],#16 + b.eq .Lctr32_done + vst1.8 {$in1},[$out] + +.Lctr32_done: +___ +$code.=<<___ if ($flavour !~ /64/); + vldmia sp!,{d8-d15} + ldmia sp!,{r4-r10,pc} +___ +$code.=<<___ if ($flavour =~ /64/); + ldr x29,[sp],#16 + ret +___ +$code.=<<___; +.size ${prefix}_ctr32_encrypt_blocks,.-${prefix}_ctr32_encrypt_blocks +___ +}}} +$code.=<<___; +#endif +___ +######################################## +if ($flavour =~ /64/) { ######## 64-bit code + my %opcode = ( + "aesd" => 0x4e285800, "aese" => 0x4e284800, + "aesimc"=> 0x4e287800, "aesmc" => 0x4e286800 ); + + local *unaes = sub { + my ($mnemonic,$arg)=@_; + + $arg =~ m/[qv]([0-9]+)[^,]*,\s*[qv]([0-9]+)/o && + sprintf ".inst\t0x%08x\t//%s %s", + $opcode{$mnemonic}|$1|($2<<5), + $mnemonic,$arg; + }; + + foreach(split("\n",$code)) { + s/\`([^\`]*)\`/eval($1)/geo; + + s/\bq([0-9]+)\b/"v".($1<8?$1:$1+8).".16b"/geo; # old->new registers + s/@\s/\/\//o; # old->new style commentary + + #s/[v]?(aes\w+)\s+([qv].*)/unaes($1,$2)/geo or + s/cclr\s+([wx])([^,]+),\s*([a-z]+)/csel $1$2,$1zr,$1$2,$3/o or + s/mov\.([a-z]+)\s+([wx][0-9]+),\s*([wx][0-9]+)/csel $2,$3,$2,$1/o or + s/vmov\.i8/movi/o or # fix up legacy mnemonics + s/vext\.8/ext/o or + s/vrev32\.8/rev32/o or + s/vtst\.8/cmtst/o or + s/vshr/ushr/o or + s/^(\s+)v/$1/o or # strip off v prefix + s/\bbx\s+lr\b/ret/o; + + # fix up remainig legacy suffixes + s/\.[ui]?8//o; + m/\],#8/o and s/\.16b/\.8b/go; + s/\.[ui]?32//o and s/\.16b/\.4s/go; + s/\.[ui]?64//o and s/\.16b/\.2d/go; + s/\.[42]([sd])\[([0-3])\]/\.$1\[$2\]/o; + + print $_,"\n"; + } +} else { ######## 32-bit code + my %opcode = ( + "aesd" => 0xf3b00340, "aese" => 0xf3b00300, + "aesimc"=> 0xf3b003c0, "aesmc" => 0xf3b00380 ); + + local *unaes = sub { + my ($mnemonic,$arg)=@_; + + if ($arg =~ m/[qv]([0-9]+)[^,]*,\s*[qv]([0-9]+)/o) { + my $word = $opcode{$mnemonic}|(($1&7)<<13)|(($1&8)<<19) + |(($2&7)<<1) |(($2&8)<<2); + # since ARMv7 instructions are always encoded little-endian. + # correct solution is to use .inst directive, but older + # assemblers don't implement it:-( + sprintf ".byte\t0x%02x,0x%02x,0x%02x,0x%02x\t@ %s %s", + $word&0xff,($word>>8)&0xff, + ($word>>16)&0xff,($word>>24)&0xff, + $mnemonic,$arg; + } + }; + + sub unvtbl { + my $arg=shift; + + $arg =~ m/q([0-9]+),\s*\{q([0-9]+)\},\s*q([0-9]+)/o && + sprintf "vtbl.8 d%d,{q%d},d%d\n\t". + "vtbl.8 d%d,{q%d},d%d", 2*$1,$2,2*$3, 2*$1+1,$2,2*$3+1; + } + + sub unvdup32 { + my $arg=shift; + + $arg =~ m/q([0-9]+),\s*q([0-9]+)\[([0-3])\]/o && + sprintf "vdup.32 q%d,d%d[%d]",$1,2*$2+($3>>1),$3&1; + } + + sub unvmov32 { + my $arg=shift; + + $arg =~ m/q([0-9]+)\[([0-3])\],(.*)/o && + sprintf "vmov.32 d%d[%d],%s",2*$1+($2>>1),$2&1,$3; + } + + foreach(split("\n",$code)) { + s/\`([^\`]*)\`/eval($1)/geo; + + s/\b[wx]([0-9]+)\b/r$1/go; # new->old registers + s/\bv([0-9])\.[12468]+[bsd]\b/q$1/go; # new->old registers + s/\/\/\s?/@ /o; # new->old style commentary + + # fix up remainig new-style suffixes + s/\{q([0-9]+)\},\s*\[(.+)\],#8/sprintf "{d%d},[$2]!",2*$1/eo or + s/\],#[0-9]+/]!/o; + + s/[v]?(aes\w+)\s+([qv].*)/unaes($1,$2)/geo or + s/cclr\s+([^,]+),\s*([a-z]+)/mov$2 $1,#0/o or + s/vtbl\.8\s+(.*)/unvtbl($1)/geo or + s/vdup\.32\s+(.*)/unvdup32($1)/geo or + s/vmov\.32\s+(.*)/unvmov32($1)/geo or + s/^(\s+)b\./$1b/o or + s/^(\s+)mov\./$1mov/o or + s/^(\s+)ret/$1bx\tlr/o; + + print $_,"\n"; + } +} + +close STDOUT; diff --git a/openssl-1.1.0h/crypto/aes/asm/bsaes-armv7.pl b/openssl-1.1.0h/crypto/aes/asm/bsaes-armv7.pl new file mode 100644 index 0000000..7af38af --- /dev/null +++ b/openssl-1.1.0h/crypto/aes/asm/bsaes-armv7.pl @@ -0,0 +1,2495 @@ +#! /usr/bin/env perl +# Copyright 2012-2018 The OpenSSL Project Authors. All Rights Reserved. +# +# Licensed under the OpenSSL license (the "License"). You may not use +# this file except in compliance with the License. You can obtain a copy +# in the file LICENSE in the source distribution or at +# https://www.openssl.org/source/license.html + + +# ==================================================================== +# Written by Andy Polyakov for the OpenSSL +# project. The module is, however, dual licensed under OpenSSL and +# CRYPTOGAMS licenses depending on where you obtain it. For further +# details see http://www.openssl.org/~appro/cryptogams/. +# +# Specific modes and adaptation for Linux kernel by Ard Biesheuvel +# . Permission to use under GPL terms is +# granted. +# ==================================================================== + +# Bit-sliced AES for ARM NEON +# +# February 2012. +# +# This implementation is direct adaptation of bsaes-x86_64 module for +# ARM NEON. Except that this module is endian-neutral [in sense that +# it can be compiled for either endianness] by courtesy of vld1.8's +# neutrality. Initial version doesn't implement interface to OpenSSL, +# only low-level primitives and unsupported entry points, just enough +# to collect performance results, which for Cortex-A8 core are: +# +# encrypt 19.5 cycles per byte processed with 128-bit key +# decrypt 22.1 cycles per byte processed with 128-bit key +# key conv. 440 cycles per 128-bit key/0.18 of 8x block +# +# Snapdragon S4 encrypts byte in 17.6 cycles and decrypts in 19.7, +# which is [much] worse than anticipated (for further details see +# http://www.openssl.org/~appro/Snapdragon-S4.html). +# +# Cortex-A15 manages in 14.2/16.1 cycles [when integer-only code +# manages in 20.0 cycles]. +# +# When comparing to x86_64 results keep in mind that NEON unit is +# [mostly] single-issue and thus can't [fully] benefit from +# instruction-level parallelism. And when comparing to aes-armv4 +# results keep in mind key schedule conversion overhead (see +# bsaes-x86_64.pl for further details)... +# +# + +# April-August 2013 +# +# Add CBC, CTR and XTS subroutines, adapt for kernel use. +# +# + +$flavour = shift; +if ($flavour=~/\w[\w\-]*\.\w+$/) { $output=$flavour; undef $flavour; } +else { while (($output=shift) && ($output!~/\w[\w\-]*\.\w+$/)) {} } + +if ($flavour && $flavour ne "void") { + $0 =~ m/(.*[\/\\])[^\/\\]+$/; $dir=$1; + ( $xlate="${dir}arm-xlate.pl" and -f $xlate ) or + ( $xlate="${dir}../../perlasm/arm-xlate.pl" and -f $xlate) or + die "can't locate arm-xlate.pl"; + + open STDOUT,"| \"$^X\" $xlate $flavour $output"; +} else { + open STDOUT,">$output"; +} + +my ($inp,$out,$len,$key)=("r0","r1","r2","r3"); +my @XMM=map("q$_",(0..15)); + +{ +my ($key,$rounds,$const)=("r4","r5","r6"); + +sub Dlo() { shift=~m|q([1]?[0-9])|?"d".($1*2):""; } +sub Dhi() { shift=~m|q([1]?[0-9])|?"d".($1*2+1):""; } + +sub Sbox { +# input in lsb > [b0, b1, b2, b3, b4, b5, b6, b7] < msb +# output in lsb > [b0, b1, b4, b6, b3, b7, b2, b5] < msb +my @b=@_[0..7]; +my @t=@_[8..11]; +my @s=@_[12..15]; + &InBasisChange (@b); + &Inv_GF256 (@b[6,5,0,3,7,1,4,2],@t,@s); + &OutBasisChange (@b[7,1,4,2,6,5,0,3]); +} + +sub InBasisChange { +# input in lsb > [b0, b1, b2, b3, b4, b5, b6, b7] < msb +# output in lsb > [b6, b5, b0, b3, b7, b1, b4, b2] < msb +my @b=@_[0..7]; +$code.=<<___; + veor @b[2], @b[2], @b[1] + veor @b[5], @b[5], @b[6] + veor @b[3], @b[3], @b[0] + veor @b[6], @b[6], @b[2] + veor @b[5], @b[5], @b[0] + + veor @b[6], @b[6], @b[3] + veor @b[3], @b[3], @b[7] + veor @b[7], @b[7], @b[5] + veor @b[3], @b[3], @b[4] + veor @b[4], @b[4], @b[5] + + veor @b[2], @b[2], @b[7] + veor @b[3], @b[3], @b[1] + veor @b[1], @b[1], @b[5] +___ +} + +sub OutBasisChange { +# input in lsb > [b0, b1, b2, b3, b4, b5, b6, b7] < msb +# output in lsb > [b6, b1, b2, b4, b7, b0, b3, b5] < msb +my @b=@_[0..7]; +$code.=<<___; + veor @b[0], @b[0], @b[6] + veor @b[1], @b[1], @b[4] + veor @b[4], @b[4], @b[6] + veor @b[2], @b[2], @b[0] + veor @b[6], @b[6], @b[1] + + veor @b[1], @b[1], @b[5] + veor @b[5], @b[5], @b[3] + veor @b[3], @b[3], @b[7] + veor @b[7], @b[7], @b[5] + veor @b[2], @b[2], @b[5] + + veor @b[4], @b[4], @b[7] +___ +} + +sub InvSbox { +# input in lsb > [b0, b1, b2, b3, b4, b5, b6, b7] < msb +# output in lsb > [b0, b1, b6, b4, b2, b7, b3, b5] < msb +my @b=@_[0..7]; +my @t=@_[8..11]; +my @s=@_[12..15]; + &InvInBasisChange (@b); + &Inv_GF256 (@b[5,1,2,6,3,7,0,4],@t,@s); + &InvOutBasisChange (@b[3,7,0,4,5,1,2,6]); +} + +sub InvInBasisChange { # OutBasisChange in reverse (with twist) +my @b=@_[5,1,2,6,3,7,0,4]; +$code.=<<___ + veor @b[1], @b[1], @b[7] + veor @b[4], @b[4], @b[7] + + veor @b[7], @b[7], @b[5] + veor @b[1], @b[1], @b[3] + veor @b[2], @b[2], @b[5] + veor @b[3], @b[3], @b[7] + + veor @b[6], @b[6], @b[1] + veor @b[2], @b[2], @b[0] + veor @b[5], @b[5], @b[3] + veor @b[4], @b[4], @b[6] + veor @b[0], @b[0], @b[6] + veor @b[1], @b[1], @b[4] +___ +} + +sub InvOutBasisChange { # InBasisChange in reverse +my @b=@_[2,5,7,3,6,1,0,4]; +$code.=<<___; + veor @b[1], @b[1], @b[5] + veor @b[2], @b[2], @b[7] + + veor @b[3], @b[3], @b[1] + veor @b[4], @b[4], @b[5] + veor @b[7], @b[7], @b[5] + veor @b[3], @b[3], @b[4] + veor @b[5], @b[5], @b[0] + veor @b[3], @b[3], @b[7] + veor @b[6], @b[6], @b[2] + veor @b[2], @b[2], @b[1] + veor @b[6], @b[6], @b[3] + + veor @b[3], @b[3], @b[0] + veor @b[5], @b[5], @b[6] +___ +} + +sub Mul_GF4 { +#;************************************************************* +#;* Mul_GF4: Input x0-x1,y0-y1 Output x0-x1 Temp t0 (8) * +#;************************************************************* +my ($x0,$x1,$y0,$y1,$t0,$t1)=@_; +$code.=<<___; + veor $t0, $y0, $y1 + vand $t0, $t0, $x0 + veor $x0, $x0, $x1 + vand $t1, $x1, $y0 + vand $x0, $x0, $y1 + veor $x1, $t1, $t0 + veor $x0, $x0, $t1 +___ +} + +sub Mul_GF4_N { # not used, see next subroutine +# multiply and scale by N +my ($x0,$x1,$y0,$y1,$t0)=@_; +$code.=<<___; + veor $t0, $y0, $y1 + vand $t0, $t0, $x0 + veor $x0, $x0, $x1 + vand $x1, $x1, $y0 + vand $x0, $x0, $y1 + veor $x1, $x1, $x0 + veor $x0, $x0, $t0 +___ +} + +sub Mul_GF4_N_GF4 { +# interleaved Mul_GF4_N and Mul_GF4 +my ($x0,$x1,$y0,$y1,$t0, + $x2,$x3,$y2,$y3,$t1)=@_; +$code.=<<___; + veor $t0, $y0, $y1 + veor $t1, $y2, $y3 + vand $t0, $t0, $x0 + vand $t1, $t1, $x2 + veor $x0, $x0, $x1 + veor $x2, $x2, $x3 + vand $x1, $x1, $y0 + vand $x3, $x3, $y2 + vand $x0, $x0, $y1 + vand $x2, $x2, $y3 + veor $x1, $x1, $x0 + veor $x2, $x2, $x3 + veor $x0, $x0, $t0 + veor $x3, $x3, $t1 +___ +} +sub Mul_GF16_2 { +my @x=@_[0..7]; +my @y=@_[8..11]; +my @t=@_[12..15]; +$code.=<<___; + veor @t[0], @x[0], @x[2] + veor @t[1], @x[1], @x[3] +___ + &Mul_GF4 (@x[0], @x[1], @y[0], @y[1], @t[2..3]); +$code.=<<___; + veor @y[0], @y[0], @y[2] + veor @y[1], @y[1], @y[3] +___ + Mul_GF4_N_GF4 (@t[0], @t[1], @y[0], @y[1], @t[3], + @x[2], @x[3], @y[2], @y[3], @t[2]); +$code.=<<___; + veor @x[0], @x[0], @t[0] + veor @x[2], @x[2], @t[0] + veor @x[1], @x[1], @t[1] + veor @x[3], @x[3], @t[1] + + veor @t[0], @x[4], @x[6] + veor @t[1], @x[5], @x[7] +___ + &Mul_GF4_N_GF4 (@t[0], @t[1], @y[0], @y[1], @t[3], + @x[6], @x[7], @y[2], @y[3], @t[2]); +$code.=<<___; + veor @y[0], @y[0], @y[2] + veor @y[1], @y[1], @y[3] +___ + &Mul_GF4 (@x[4], @x[5], @y[0], @y[1], @t[2..3]); +$code.=<<___; + veor @x[4], @x[4], @t[0] + veor @x[6], @x[6], @t[0] + veor @x[5], @x[5], @t[1] + veor @x[7], @x[7], @t[1] +___ +} +sub Inv_GF256 { +#;******************************************************************** +#;* Inv_GF256: Input x0-x7 Output x0-x7 Temp t0-t3,s0-s3 (144) * +#;******************************************************************** +my @x=@_[0..7]; +my @t=@_[8..11]; +my @s=@_[12..15]; +# direct optimizations from hardware +$code.=<<___; + veor @t[3], @x[4], @x[6] + veor @t[2], @x[5], @x[7] + veor @t[1], @x[1], @x[3] + veor @s[1], @x[7], @x[6] + vmov @t[0], @t[2] + veor @s[0], @x[0], @x[2] + + vorr @t[2], @t[2], @t[1] + veor @s[3], @t[3], @t[0] + vand @s[2], @t[3], @s[0] + vorr @t[3], @t[3], @s[0] + veor @s[0], @s[0], @t[1] + vand @t[0], @t[0], @t[1] + veor @t[1], @x[3], @x[2] + vand @s[3], @s[3], @s[0] + vand @s[1], @s[1], @t[1] + veor @t[1], @x[4], @x[5] + veor @s[0], @x[1], @x[0] + veor @t[3], @t[3], @s[1] + veor @t[2], @t[2], @s[1] + vand @s[1], @t[1], @s[0] + vorr @t[1], @t[1], @s[0] + veor @t[3], @t[3], @s[3] + veor @t[0], @t[0], @s[1] + veor @t[2], @t[2], @s[2] + veor @t[1], @t[1], @s[3] + veor @t[0], @t[0], @s[2] + vand @s[0], @x[7], @x[3] + veor @t[1], @t[1], @s[2] + vand @s[1], @x[6], @x[2] + vand @s[2], @x[5], @x[1] + vorr @s[3], @x[4], @x[0] + veor @t[3], @t[3], @s[0] + veor @t[1], @t[1], @s[2] + veor @t[0], @t[0], @s[3] + veor @t[2], @t[2], @s[1] + + @ Inv_GF16 \t0, \t1, \t2, \t3, \s0, \s1, \s2, \s3 + + @ new smaller inversion + + vand @s[2], @t[3], @t[1] + vmov @s[0], @t[0] + + veor @s[1], @t[2], @s[2] + veor @s[3], @t[0], @s[2] + veor @s[2], @t[0], @s[2] @ @s[2]=@s[3] + + vbsl @s[1], @t[1], @t[0] + vbsl @s[3], @t[3], @t[2] + veor @t[3], @t[3], @t[2] + + vbsl @s[0], @s[1], @s[2] + vbsl @t[0], @s[2], @s[1] + + vand @s[2], @s[0], @s[3] + veor @t[1], @t[1], @t[0] + + veor @s[2], @s[2], @t[3] +___ +# output in s3, s2, s1, t1 + +# Mul_GF16_2 \x0, \x1, \x2, \x3, \x4, \x5, \x6, \x7, \t2, \t3, \t0, \t1, \s0, \s1, \s2, \s3 + +# Mul_GF16_2 \x0, \x1, \x2, \x3, \x4, \x5, \x6, \x7, \s3, \s2, \s1, \t1, \s0, \t0, \t2, \t3 + &Mul_GF16_2(@x,@s[3,2,1],@t[1],@s[0],@t[0,2,3]); + +### output msb > [x3,x2,x1,x0,x7,x6,x5,x4] < lsb +} + +# AES linear components + +sub ShiftRows { +my @x=@_[0..7]; +my @t=@_[8..11]; +my $mask=pop; +$code.=<<___; + vldmia $key!, {@t[0]-@t[3]} + veor @t[0], @t[0], @x[0] + veor @t[1], @t[1], @x[1] + vtbl.8 `&Dlo(@x[0])`, {@t[0]}, `&Dlo($mask)` + vtbl.8 `&Dhi(@x[0])`, {@t[0]}, `&Dhi($mask)` + vldmia $key!, {@t[0]} + veor @t[2], @t[2], @x[2] + vtbl.8 `&Dlo(@x[1])`, {@t[1]}, `&Dlo($mask)` + vtbl.8 `&Dhi(@x[1])`, {@t[1]}, `&Dhi($mask)` + vldmia $key!, {@t[1]} + veor @t[3], @t[3], @x[3] + vtbl.8 `&Dlo(@x[2])`, {@t[2]}, `&Dlo($mask)` + vtbl.8 `&Dhi(@x[2])`, {@t[2]}, `&Dhi($mask)` + vldmia $key!, {@t[2]} + vtbl.8 `&Dlo(@x[3])`, {@t[3]}, `&Dlo($mask)` + vtbl.8 `&Dhi(@x[3])`, {@t[3]}, `&Dhi($mask)` + vldmia $key!, {@t[3]} + veor @t[0], @t[0], @x[4] + veor @t[1], @t[1], @x[5] + vtbl.8 `&Dlo(@x[4])`, {@t[0]}, `&Dlo($mask)` + vtbl.8 `&Dhi(@x[4])`, {@t[0]}, `&Dhi($mask)` + veor @t[2], @t[2], @x[6] + vtbl.8 `&Dlo(@x[5])`, {@t[1]}, `&Dlo($mask)` + vtbl.8 `&Dhi(@x[5])`, {@t[1]}, `&Dhi($mask)` + veor @t[3], @t[3], @x[7] + vtbl.8 `&Dlo(@x[6])`, {@t[2]}, `&Dlo($mask)` + vtbl.8 `&Dhi(@x[6])`, {@t[2]}, `&Dhi($mask)` + vtbl.8 `&Dlo(@x[7])`, {@t[3]}, `&Dlo($mask)` + vtbl.8 `&Dhi(@x[7])`, {@t[3]}, `&Dhi($mask)` +___ +} + +sub MixColumns { +# modified to emit output in order suitable for feeding back to aesenc[last] +my @x=@_[0..7]; +my @t=@_[8..15]; +my $inv=@_[16]; # optional +$code.=<<___; + vext.8 @t[0], @x[0], @x[0], #12 @ x0 <<< 32 + vext.8 @t[1], @x[1], @x[1], #12 + veor @x[0], @x[0], @t[0] @ x0 ^ (x0 <<< 32) + vext.8 @t[2], @x[2], @x[2], #12 + veor @x[1], @x[1], @t[1] + vext.8 @t[3], @x[3], @x[3], #12 + veor @x[2], @x[2], @t[2] + vext.8 @t[4], @x[4], @x[4], #12 + veor @x[3], @x[3], @t[3] + vext.8 @t[5], @x[5], @x[5], #12 + veor @x[4], @x[4], @t[4] + vext.8 @t[6], @x[6], @x[6], #12 + veor @x[5], @x[5], @t[5] + vext.8 @t[7], @x[7], @x[7], #12 + veor @x[6], @x[6], @t[6] + + veor @t[1], @t[1], @x[0] + veor @x[7], @x[7], @t[7] + vext.8 @x[0], @x[0], @x[0], #8 @ (x0 ^ (x0 <<< 32)) <<< 64) + veor @t[2], @t[2], @x[1] + veor @t[0], @t[0], @x[7] + veor @t[1], @t[1], @x[7] + vext.8 @x[1], @x[1], @x[1], #8 + veor @t[5], @t[5], @x[4] + veor @x[0], @x[0], @t[0] + veor @t[6], @t[6], @x[5] + veor @x[1], @x[1], @t[1] + vext.8 @t[0], @x[4], @x[4], #8 + veor @t[4], @t[4], @x[3] + vext.8 @t[1], @x[5], @x[5], #8 + veor @t[7], @t[7], @x[6] + vext.8 @x[4], @x[3], @x[3], #8 + veor @t[3], @t[3], @x[2] + vext.8 @x[5], @x[7], @x[7], #8 + veor @t[4], @t[4], @x[7] + vext.8 @x[3], @x[6], @x[6], #8 + veor @t[3], @t[3], @x[7] + vext.8 @x[6], @x[2], @x[2], #8 + veor @x[7], @t[1], @t[5] +___ +$code.=<<___ if (!$inv); + veor @x[2], @t[0], @t[4] + veor @x[4], @x[4], @t[3] + veor @x[5], @x[5], @t[7] + veor @x[3], @x[3], @t[6] + @ vmov @x[2], @t[0] + veor @x[6], @x[6], @t[2] + @ vmov @x[7], @t[1] +___ +$code.=<<___ if ($inv); + veor @t[3], @t[3], @x[4] + veor @x[5], @x[5], @t[7] + veor @x[2], @x[3], @t[6] + veor @x[3], @t[0], @t[4] + veor @x[4], @x[6], @t[2] + vmov @x[6], @t[3] + @ vmov @x[7], @t[1] +___ +} + +sub InvMixColumns_orig { +my @x=@_[0..7]; +my @t=@_[8..15]; + +$code.=<<___; + @ multiplication by 0x0e + vext.8 @t[7], @x[7], @x[7], #12 + vmov @t[2], @x[2] + veor @x[2], @x[2], @x[5] @ 2 5 + veor @x[7], @x[7], @x[5] @ 7 5 + vext.8 @t[0], @x[0], @x[0], #12 + vmov @t[5], @x[5] + veor @x[5], @x[5], @x[0] @ 5 0 [1] + veor @x[0], @x[0], @x[1] @ 0 1 + vext.8 @t[1], @x[1], @x[1], #12 + veor @x[1], @x[1], @x[2] @ 1 25 + veor @x[0], @x[0], @x[6] @ 01 6 [2] + vext.8 @t[3], @x[3], @x[3], #12 + veor @x[1], @x[1], @x[3] @ 125 3 [4] + veor @x[2], @x[2], @x[0] @ 25 016 [3] + veor @x[3], @x[3], @x[7] @ 3 75 + veor @x[7], @x[7], @x[6] @ 75 6 [0] + vext.8 @t[6], @x[6], @x[6], #12 + vmov @t[4], @x[4] + veor @x[6], @x[6], @x[4] @ 6 4 + veor @x[4], @x[4], @x[3] @ 4 375 [6] + veor @x[3], @x[3], @x[7] @ 375 756=36 + veor @x[6], @x[6], @t[5] @ 64 5 [7] + veor @x[3], @x[3], @t[2] @ 36 2 + vext.8 @t[5], @t[5], @t[5], #12 + veor @x[3], @x[3], @t[4] @ 362 4 [5] +___ + my @y = @x[7,5,0,2,1,3,4,6]; +$code.=<<___; + @ multiplication by 0x0b + veor @y[1], @y[1], @y[0] + veor @y[0], @y[0], @t[0] + vext.8 @t[2], @t[2], @t[2], #12 + veor @y[1], @y[1], @t[1] + veor @y[0], @y[0], @t[5] + vext.8 @t[4], @t[4], @t[4], #12 + veor @y[1], @y[1], @t[6] + veor @y[0], @y[0], @t[7] + veor @t[7], @t[7], @t[6] @ clobber t[7] + + veor @y[3], @y[3], @t[0] + veor @y[1], @y[1], @y[0] + vext.8 @t[0], @t[0], @t[0], #12 + veor @y[2], @y[2], @t[1] + veor @y[4], @y[4], @t[1] + vext.8 @t[1], @t[1], @t[1], #12 + veor @y[2], @y[2], @t[2] + veor @y[3], @y[3], @t[2] + veor @y[5], @y[5], @t[2] + veor @y[2], @y[2], @t[7] + vext.8 @t[2], @t[2], @t[2], #12 + veor @y[3], @y[3], @t[3] + veor @y[6], @y[6], @t[3] + veor @y[4], @y[4], @t[3] + veor @y[7], @y[7], @t[4] + vext.8 @t[3], @t[3], @t[3], #12 + veor @y[5], @y[5], @t[4] + veor @y[7], @y[7], @t[7] + veor @t[7], @t[7], @t[5] @ clobber t[7] even more + veor @y[3], @y[3], @t[5] + veor @y[4], @y[4], @t[4] + + veor @y[5], @y[5], @t[7] + vext.8 @t[4], @t[4], @t[4], #12 + veor @y[6], @y[6], @t[7] + veor @y[4], @y[4], @t[7] + + veor @t[7], @t[7], @t[5] + vext.8 @t[5], @t[5], @t[5], #12 + + @ multiplication by 0x0d + veor @y[4], @y[4], @y[7] + veor @t[7], @t[7], @t[6] @ restore t[7] + veor @y[7], @y[7], @t[4] + vext.8 @t[6], @t[6], @t[6], #12 + veor @y[2], @y[2], @t[0] + veor @y[7], @y[7], @t[5] + vext.8 @t[7], @t[7], @t[7], #12 + veor @y[2], @y[2], @t[2] + + veor @y[3], @y[3], @y[1] + veor @y[1], @y[1], @t[1] + veor @y[0], @y[0], @t[0] + veor @y[3], @y[3], @t[0] + veor @y[1], @y[1], @t[5] + veor @y[0], @y[0], @t[5] + vext.8 @t[0], @t[0], @t[0], #12 + veor @y[1], @y[1], @t[7] + veor @y[0], @y[0], @t[6] + veor @y[3], @y[3], @y[1] + veor @y[4], @y[4], @t[1] + vext.8 @t[1], @t[1], @t[1], #12 + + veor @y[7], @y[7], @t[7] + veor @y[4], @y[4], @t[2] + veor @y[5], @y[5], @t[2] + veor @y[2], @y[2], @t[6] + veor @t[6], @t[6], @t[3] @ clobber t[6] + vext.8 @t[2], @t[2], @t[2], #12 + veor @y[4], @y[4], @y[7] + veor @y[3], @y[3], @t[6] + + veor @y[6], @y[6], @t[6] + veor @y[5], @y[5], @t[5] + vext.8 @t[5], @t[5], @t[5], #12 + veor @y[6], @y[6], @t[4] + vext.8 @t[4], @t[4], @t[4], #12 + veor @y[5], @y[5], @t[6] + veor @y[6], @y[6], @t[7] + vext.8 @t[7], @t[7], @t[7], #12 + veor @t[6], @t[6], @t[3] @ restore t[6] + vext.8 @t[3], @t[3], @t[3], #12 + + @ multiplication by 0x09 + veor @y[4], @y[4], @y[1] + veor @t[1], @t[1], @y[1] @ t[1]=y[1] + veor @t[0], @t[0], @t[5] @ clobber t[0] + vext.8 @t[6], @t[6], @t[6], #12 + veor @t[1], @t[1], @t[5] + veor @y[3], @y[3], @t[0] + veor @t[0], @t[0], @y[0] @ t[0]=y[0] + veor @t[1], @t[1], @t[6] + veor @t[6], @t[6], @t[7] @ clobber t[6] + veor @y[4], @y[4], @t[1] + veor @y[7], @y[7], @t[4] + veor @y[6], @y[6], @t[3] + veor @y[5], @y[5], @t[2] + veor @t[4], @t[4], @y[4] @ t[4]=y[4] + veor @t[3], @t[3], @y[3] @ t[3]=y[3] + veor @t[5], @t[5], @y[5] @ t[5]=y[5] + veor @t[2], @t[2], @y[2] @ t[2]=y[2] + veor @t[3], @t[3], @t[7] + veor @XMM[5], @t[5], @t[6] + veor @XMM[6], @t[6], @y[6] @ t[6]=y[6] + veor @XMM[2], @t[2], @t[6] + veor @XMM[7], @t[7], @y[7] @ t[7]=y[7] + + vmov @XMM[0], @t[0] + vmov @XMM[1], @t[1] + @ vmov @XMM[2], @t[2] + vmov @XMM[3], @t[3] + vmov @XMM[4], @t[4] + @ vmov @XMM[5], @t[5] + @ vmov @XMM[6], @t[6] + @ vmov @XMM[7], @t[7] +___ +} + +sub InvMixColumns { +my @x=@_[0..7]; +my @t=@_[8..15]; + +# Thanks to Jussi Kivilinna for providing pointer to +# +# | 0e 0b 0d 09 | | 02 03 01 01 | | 05 00 04 00 | +# | 09 0e 0b 0d | = | 01 02 03 01 | x | 00 05 00 04 | +# | 0d 09 0e 0b | | 01 01 02 03 | | 04 00 05 00 | +# | 0b 0d 09 0e | | 03 01 01 02 | | 00 04 00 05 | + +$code.=<<___; + @ multiplication by 0x05-0x00-0x04-0x00 + vext.8 @t[0], @x[0], @x[0], #8 + vext.8 @t[6], @x[6], @x[6], #8 + vext.8 @t[7], @x[7], @x[7], #8 + veor @t[0], @t[0], @x[0] + vext.8 @t[1], @x[1], @x[1], #8 + veor @t[6], @t[6], @x[6] + vext.8 @t[2], @x[2], @x[2], #8 + veor @t[7], @t[7], @x[7] + vext.8 @t[3], @x[3], @x[3], #8 + veor @t[1], @t[1], @x[1] + vext.8 @t[4], @x[4], @x[4], #8 + veor @t[2], @t[2], @x[2] + vext.8 @t[5], @x[5], @x[5], #8 + veor @t[3], @t[3], @x[3] + veor @t[4], @t[4], @x[4] + veor @t[5], @t[5], @x[5] + + veor @x[0], @x[0], @t[6] + veor @x[1], @x[1], @t[6] + veor @x[2], @x[2], @t[0] + veor @x[4], @x[4], @t[2] + veor @x[3], @x[3], @t[1] + veor @x[1], @x[1], @t[7] + veor @x[2], @x[2], @t[7] + veor @x[4], @x[4], @t[6] + veor @x[5], @x[5], @t[3] + veor @x[3], @x[3], @t[6] + veor @x[6], @x[6], @t[4] + veor @x[4], @x[4], @t[7] + veor @x[5], @x[5], @t[7] + veor @x[7], @x[7], @t[5] +___ + &MixColumns (@x,@t,1); # flipped 2<->3 and 4<->6 +} + +sub swapmove { +my ($a,$b,$n,$mask,$t)=@_; +$code.=<<___; + vshr.u64 $t, $b, #$n + veor $t, $t, $a + vand $t, $t, $mask + veor $a, $a, $t + vshl.u64 $t, $t, #$n + veor $b, $b, $t +___ +} +sub swapmove2x { +my ($a0,$b0,$a1,$b1,$n,$mask,$t0,$t1)=@_; +$code.=<<___; + vshr.u64 $t0, $b0, #$n + vshr.u64 $t1, $b1, #$n + veor $t0, $t0, $a0 + veor $t1, $t1, $a1 + vand $t0, $t0, $mask + vand $t1, $t1, $mask + veor $a0, $a0, $t0 + vshl.u64 $t0, $t0, #$n + veor $a1, $a1, $t1 + vshl.u64 $t1, $t1, #$n + veor $b0, $b0, $t0 + veor $b1, $b1, $t1 +___ +} + +sub bitslice { +my @x=reverse(@_[0..7]); +my ($t0,$t1,$t2,$t3)=@_[8..11]; +$code.=<<___; + vmov.i8 $t0,#0x55 @ compose .LBS0 + vmov.i8 $t1,#0x33 @ compose .LBS1 +___ + &swapmove2x(@x[0,1,2,3],1,$t0,$t2,$t3); + &swapmove2x(@x[4,5,6,7],1,$t0,$t2,$t3); +$code.=<<___; + vmov.i8 $t0,#0x0f @ compose .LBS2 +___ + &swapmove2x(@x[0,2,1,3],2,$t1,$t2,$t3); + &swapmove2x(@x[4,6,5,7],2,$t1,$t2,$t3); + + &swapmove2x(@x[0,4,1,5],4,$t0,$t2,$t3); + &swapmove2x(@x[2,6,3,7],4,$t0,$t2,$t3); +} + +$code.=<<___; +#ifndef __KERNEL__ +# include "arm_arch.h" + +# define VFP_ABI_PUSH vstmdb sp!,{d8-d15} +# define VFP_ABI_POP vldmia sp!,{d8-d15} +# define VFP_ABI_FRAME 0x40 +#else +# define VFP_ABI_PUSH +# define VFP_ABI_POP +# define VFP_ABI_FRAME 0 +# define BSAES_ASM_EXTENDED_KEY +# define XTS_CHAIN_TWEAK +# define __ARM_ARCH__ __LINUX_ARM_ARCH__ +# define __ARM_MAX_ARCH__ 7 +#endif + +#ifdef __thumb__ +# define adrl adr +#endif + +#if __ARM_MAX_ARCH__>=7 +.arch armv7-a +.fpu neon + +.text +.syntax unified @ ARMv7-capable assembler is expected to handle this +#if defined(__thumb2__) && !defined(__APPLE__) +.thumb +#else +.code 32 +# undef __thumb2__ +#endif + +.type _bsaes_decrypt8,%function +.align 4 +_bsaes_decrypt8: + adr $const,. + vldmia $key!, {@XMM[9]} @ round 0 key +#if defined(__thumb2__) || defined(__APPLE__) + adr $const,.LM0ISR +#else + add $const,$const,#.LM0ISR-_bsaes_decrypt8 +#endif + + vldmia $const!, {@XMM[8]} @ .LM0ISR + veor @XMM[10], @XMM[0], @XMM[9] @ xor with round0 key + veor @XMM[11], @XMM[1], @XMM[9] + vtbl.8 `&Dlo(@XMM[0])`, {@XMM[10]}, `&Dlo(@XMM[8])` + vtbl.8 `&Dhi(@XMM[0])`, {@XMM[10]}, `&Dhi(@XMM[8])` + veor @XMM[12], @XMM[2], @XMM[9] + vtbl.8 `&Dlo(@XMM[1])`, {@XMM[11]}, `&Dlo(@XMM[8])` + vtbl.8 `&Dhi(@XMM[1])`, {@XMM[11]}, `&Dhi(@XMM[8])` + veor @XMM[13], @XMM[3], @XMM[9] + vtbl.8 `&Dlo(@XMM[2])`, {@XMM[12]}, `&Dlo(@XMM[8])` + vtbl.8 `&Dhi(@XMM[2])`, {@XMM[12]}, `&Dhi(@XMM[8])` + veor @XMM[14], @XMM[4], @XMM[9] + vtbl.8 `&Dlo(@XMM[3])`, {@XMM[13]}, `&Dlo(@XMM[8])` + vtbl.8 `&Dhi(@XMM[3])`, {@XMM[13]}, `&Dhi(@XMM[8])` + veor @XMM[15], @XMM[5], @XMM[9] + vtbl.8 `&Dlo(@XMM[4])`, {@XMM[14]}, `&Dlo(@XMM[8])` + vtbl.8 `&Dhi(@XMM[4])`, {@XMM[14]}, `&Dhi(@XMM[8])` + veor @XMM[10], @XMM[6], @XMM[9] + vtbl.8 `&Dlo(@XMM[5])`, {@XMM[15]}, `&Dlo(@XMM[8])` + vtbl.8 `&Dhi(@XMM[5])`, {@XMM[15]}, `&Dhi(@XMM[8])` + veor @XMM[11], @XMM[7], @XMM[9] + vtbl.8 `&Dlo(@XMM[6])`, {@XMM[10]}, `&Dlo(@XMM[8])` + vtbl.8 `&Dhi(@XMM[6])`, {@XMM[10]}, `&Dhi(@XMM[8])` + vtbl.8 `&Dlo(@XMM[7])`, {@XMM[11]}, `&Dlo(@XMM[8])` + vtbl.8 `&Dhi(@XMM[7])`, {@XMM[11]}, `&Dhi(@XMM[8])` +___ + &bitslice (@XMM[0..7, 8..11]); +$code.=<<___; + sub $rounds,$rounds,#1 + b .Ldec_sbox +.align 4 +.Ldec_loop: +___ + &ShiftRows (@XMM[0..7, 8..12]); +$code.=".Ldec_sbox:\n"; + &InvSbox (@XMM[0..7, 8..15]); +$code.=<<___; + subs $rounds,$rounds,#1 + bcc .Ldec_done +___ + &InvMixColumns (@XMM[0,1,6,4,2,7,3,5, 8..15]); +$code.=<<___; + vldmia $const, {@XMM[12]} @ .LISR + ite eq @ Thumb2 thing, sanity check in ARM + addeq $const,$const,#0x10 + bne .Ldec_loop + vldmia $const, {@XMM[12]} @ .LISRM0 + b .Ldec_loop +.align 4 +.Ldec_done: +___ + &bitslice (@XMM[0,1,6,4,2,7,3,5, 8..11]); +$code.=<<___; + vldmia $key, {@XMM[8]} @ last round key + veor @XMM[6], @XMM[6], @XMM[8] + veor @XMM[4], @XMM[4], @XMM[8] + veor @XMM[2], @XMM[2], @XMM[8] + veor @XMM[7], @XMM[7], @XMM[8] + veor @XMM[3], @XMM[3], @XMM[8] + veor @XMM[5], @XMM[5], @XMM[8] + veor @XMM[0], @XMM[0], @XMM[8] + veor @XMM[1], @XMM[1], @XMM[8] + bx lr +.size _bsaes_decrypt8,.-_bsaes_decrypt8 + +.type _bsaes_const,%object +.align 6 +_bsaes_const: +.LM0ISR: @ InvShiftRows constants + .quad 0x0a0e0206070b0f03, 0x0004080c0d010509 +.LISR: + .quad 0x0504070602010003, 0x0f0e0d0c080b0a09 +.LISRM0: + .quad 0x01040b0e0205080f, 0x0306090c00070a0d +.LM0SR: @ ShiftRows constants + .quad 0x0a0e02060f03070b, 0x0004080c05090d01 +.LSR: + .quad 0x0504070600030201, 0x0f0e0d0c0a09080b +.LSRM0: + .quad 0x0304090e00050a0f, 0x01060b0c0207080d +.LM0: + .quad 0x02060a0e03070b0f, 0x0004080c0105090d +.LREVM0SR: + .quad 0x090d01050c000408, 0x03070b0f060a0e02 +.asciz "Bit-sliced AES for NEON, CRYPTOGAMS by " +.align 6 +.size _bsaes_const,.-_bsaes_const + +.type _bsaes_encrypt8,%function +.align 4 +_bsaes_encrypt8: + adr $const,. + vldmia $key!, {@XMM[9]} @ round 0 key +#if defined(__thumb2__) || defined(__APPLE__) + adr $const,.LM0SR +#else + sub $const,$const,#_bsaes_encrypt8-.LM0SR +#endif + + vldmia $const!, {@XMM[8]} @ .LM0SR +_bsaes_encrypt8_alt: + veor @XMM[10], @XMM[0], @XMM[9] @ xor with round0 key + veor @XMM[11], @XMM[1], @XMM[9] + vtbl.8 `&Dlo(@XMM[0])`, {@XMM[10]}, `&Dlo(@XMM[8])` + vtbl.8 `&Dhi(@XMM[0])`, {@XMM[10]}, `&Dhi(@XMM[8])` + veor @XMM[12], @XMM[2], @XMM[9] + vtbl.8 `&Dlo(@XMM[1])`, {@XMM[11]}, `&Dlo(@XMM[8])` + vtbl.8 `&Dhi(@XMM[1])`, {@XMM[11]}, `&Dhi(@XMM[8])` + veor @XMM[13], @XMM[3], @XMM[9] + vtbl.8 `&Dlo(@XMM[2])`, {@XMM[12]}, `&Dlo(@XMM[8])` + vtbl.8 `&Dhi(@XMM[2])`, {@XMM[12]}, `&Dhi(@XMM[8])` + veor @XMM[14], @XMM[4], @XMM[9] + vtbl.8 `&Dlo(@XMM[3])`, {@XMM[13]}, `&Dlo(@XMM[8])` + vtbl.8 `&Dhi(@XMM[3])`, {@XMM[13]}, `&Dhi(@XMM[8])` + veor @XMM[15], @XMM[5], @XMM[9] + vtbl.8 `&Dlo(@XMM[4])`, {@XMM[14]}, `&Dlo(@XMM[8])` + vtbl.8 `&Dhi(@XMM[4])`, {@XMM[14]}, `&Dhi(@XMM[8])` + veor @XMM[10], @XMM[6], @XMM[9] + vtbl.8 `&Dlo(@XMM[5])`, {@XMM[15]}, `&Dlo(@XMM[8])` + vtbl.8 `&Dhi(@XMM[5])`, {@XMM[15]}, `&Dhi(@XMM[8])` + veor @XMM[11], @XMM[7], @XMM[9] + vtbl.8 `&Dlo(@XMM[6])`, {@XMM[10]}, `&Dlo(@XMM[8])` + vtbl.8 `&Dhi(@XMM[6])`, {@XMM[10]}, `&Dhi(@XMM[8])` + vtbl.8 `&Dlo(@XMM[7])`, {@XMM[11]}, `&Dlo(@XMM[8])` + vtbl.8 `&Dhi(@XMM[7])`, {@XMM[11]}, `&Dhi(@XMM[8])` +_bsaes_encrypt8_bitslice: +___ + &bitslice (@XMM[0..7, 8..11]); +$code.=<<___; + sub $rounds,$rounds,#1 + b .Lenc_sbox +.align 4 +.Lenc_loop: +___ + &ShiftRows (@XMM[0..7, 8..12]); +$code.=".Lenc_sbox:\n"; + &Sbox (@XMM[0..7, 8..15]); +$code.=<<___; + subs $rounds,$rounds,#1 + bcc .Lenc_done +___ + &MixColumns (@XMM[0,1,4,6,3,7,2,5, 8..15]); +$code.=<<___; + vldmia $const, {@XMM[12]} @ .LSR + ite eq @ Thumb2 thing, samity check in ARM + addeq $const,$const,#0x10 + bne .Lenc_loop + vldmia $const, {@XMM[12]} @ .LSRM0 + b .Lenc_loop +.align 4 +.Lenc_done: +___ + # output in lsb > [t0, t1, t4, t6, t3, t7, t2, t5] < msb + &bitslice (@XMM[0,1,4,6,3,7,2,5, 8..11]); +$code.=<<___; + vldmia $key, {@XMM[8]} @ last round key + veor @XMM[4], @XMM[4], @XMM[8] + veor @XMM[6], @XMM[6], @XMM[8] + veor @XMM[3], @XMM[3], @XMM[8] + veor @XMM[7], @XMM[7], @XMM[8] + veor @XMM[2], @XMM[2], @XMM[8] + veor @XMM[5], @XMM[5], @XMM[8] + veor @XMM[0], @XMM[0], @XMM[8] + veor @XMM[1], @XMM[1], @XMM[8] + bx lr +.size _bsaes_encrypt8,.-_bsaes_encrypt8 +___ +} +{ +my ($out,$inp,$rounds,$const)=("r12","r4","r5","r6"); + +sub bitslice_key { +my @x=reverse(@_[0..7]); +my ($bs0,$bs1,$bs2,$t2,$t3)=@_[8..12]; + + &swapmove (@x[0,1],1,$bs0,$t2,$t3); +$code.=<<___; + @ &swapmove(@x[2,3],1,$t0,$t2,$t3); + vmov @x[2], @x[0] + vmov @x[3], @x[1] +___ + #&swapmove2x(@x[4,5,6,7],1,$t0,$t2,$t3); + + &swapmove2x (@x[0,2,1,3],2,$bs1,$t2,$t3); +$code.=<<___; + @ &swapmove2x(@x[4,6,5,7],2,$t1,$t2,$t3); + vmov @x[4], @x[0] + vmov @x[6], @x[2] + vmov @x[5], @x[1] + vmov @x[7], @x[3] +___ + &swapmove2x (@x[0,4,1,5],4,$bs2,$t2,$t3); + &swapmove2x (@x[2,6,3,7],4,$bs2,$t2,$t3); +} + +$code.=<<___; +.type _bsaes_key_convert,%function +.align 4 +_bsaes_key_convert: + adr $const,. + vld1.8 {@XMM[7]}, [$inp]! @ load round 0 key +#if defined(__thumb2__) || defined(__APPLE__) + adr $const,.LM0 +#else + sub $const,$const,#_bsaes_key_convert-.LM0 +#endif + vld1.8 {@XMM[15]}, [$inp]! @ load round 1 key + + vmov.i8 @XMM[8], #0x01 @ bit masks + vmov.i8 @XMM[9], #0x02 + vmov.i8 @XMM[10], #0x04 + vmov.i8 @XMM[11], #0x08 + vmov.i8 @XMM[12], #0x10 + vmov.i8 @XMM[13], #0x20 + vldmia $const, {@XMM[14]} @ .LM0 + +#ifdef __ARMEL__ + vrev32.8 @XMM[7], @XMM[7] + vrev32.8 @XMM[15], @XMM[15] +#endif + sub $rounds,$rounds,#1 + vstmia $out!, {@XMM[7]} @ save round 0 key + b .Lkey_loop + +.align 4 +.Lkey_loop: + vtbl.8 `&Dlo(@XMM[7])`,{@XMM[15]},`&Dlo(@XMM[14])` + vtbl.8 `&Dhi(@XMM[7])`,{@XMM[15]},`&Dhi(@XMM[14])` + vmov.i8 @XMM[6], #0x40 + vmov.i8 @XMM[15], #0x80 + + vtst.8 @XMM[0], @XMM[7], @XMM[8] + vtst.8 @XMM[1], @XMM[7], @XMM[9] + vtst.8 @XMM[2], @XMM[7], @XMM[10] + vtst.8 @XMM[3], @XMM[7], @XMM[11] + vtst.8 @XMM[4], @XMM[7], @XMM[12] + vtst.8 @XMM[5], @XMM[7], @XMM[13] + vtst.8 @XMM[6], @XMM[7], @XMM[6] + vtst.8 @XMM[7], @XMM[7], @XMM[15] + vld1.8 {@XMM[15]}, [$inp]! @ load next round key + vmvn @XMM[0], @XMM[0] @ "pnot" + vmvn @XMM[1], @XMM[1] + vmvn @XMM[5], @XMM[5] + vmvn @XMM[6], @XMM[6] +#ifdef __ARMEL__ + vrev32.8 @XMM[15], @XMM[15] +#endif + subs $rounds,$rounds,#1 + vstmia $out!,{@XMM[0]-@XMM[7]} @ write bit-sliced round key + bne .Lkey_loop + + vmov.i8 @XMM[7],#0x63 @ compose .L63 + @ don't save last round key + bx lr +.size _bsaes_key_convert,.-_bsaes_key_convert +___ +} + +if (0) { # following four functions are unsupported interface + # used for benchmarking... +$code.=<<___; +.globl bsaes_enc_key_convert +.type bsaes_enc_key_convert,%function +.align 4 +bsaes_enc_key_convert: + stmdb sp!,{r4-r6,lr} + vstmdb sp!,{d8-d15} @ ABI specification says so + + ldr r5,[$inp,#240] @ pass rounds + mov r4,$inp @ pass key + mov r12,$out @ pass key schedule + bl _bsaes_key_convert + veor @XMM[7],@XMM[7],@XMM[15] @ fix up last round key + vstmia r12, {@XMM[7]} @ save last round key + + vldmia sp!,{d8-d15} + ldmia sp!,{r4-r6,pc} +.size bsaes_enc_key_convert,.-bsaes_enc_key_convert + +.globl bsaes_encrypt_128 +.type bsaes_encrypt_128,%function +.align 4 +bsaes_encrypt_128: + stmdb sp!,{r4-r6,lr} + vstmdb sp!,{d8-d15} @ ABI specification says so +.Lenc128_loop: + vld1.8 {@XMM[0]-@XMM[1]}, [$inp]! @ load input + vld1.8 {@XMM[2]-@XMM[3]}, [$inp]! + mov r4,$key @ pass the key + vld1.8 {@XMM[4]-@XMM[5]}, [$inp]! + mov r5,#10 @ pass rounds + vld1.8 {@XMM[6]-@XMM[7]}, [$inp]! + + bl _bsaes_encrypt8 + + vst1.8 {@XMM[0]-@XMM[1]}, [$out]! @ write output + vst1.8 {@XMM[4]}, [$out]! + vst1.8 {@XMM[6]}, [$out]! + vst1.8 {@XMM[3]}, [$out]! + vst1.8 {@XMM[7]}, [$out]! + vst1.8 {@XMM[2]}, [$out]! + subs $len,$len,#0x80 + vst1.8 {@XMM[5]}, [$out]! + bhi .Lenc128_loop + + vldmia sp!,{d8-d15} + ldmia sp!,{r4-r6,pc} +.size bsaes_encrypt_128,.-bsaes_encrypt_128 + +.globl bsaes_dec_key_convert +.type bsaes_dec_key_convert,%function +.align 4 +bsaes_dec_key_convert: + stmdb sp!,{r4-r6,lr} + vstmdb sp!,{d8-d15} @ ABI specification says so + + ldr r5,[$inp,#240] @ pass rounds + mov r4,$inp @ pass key + mov r12,$out @ pass key schedule + bl _bsaes_key_convert + vldmia $out, {@XMM[6]} + vstmia r12, {@XMM[15]} @ save last round key + veor @XMM[7], @XMM[7], @XMM[6] @ fix up round 0 key + vstmia $out, {@XMM[7]} + + vldmia sp!,{d8-d15} + ldmia sp!,{r4-r6,pc} +.size bsaes_dec_key_convert,.-bsaes_dec_key_convert + +.globl bsaes_decrypt_128 +.type bsaes_decrypt_128,%function +.align 4 +bsaes_decrypt_128: + stmdb sp!,{r4-r6,lr} + vstmdb sp!,{d8-d15} @ ABI specification says so +.Ldec128_loop: + vld1.8 {@XMM[0]-@XMM[1]}, [$inp]! @ load input + vld1.8 {@XMM[2]-@XMM[3]}, [$inp]! + mov r4,$key @ pass the key + vld1.8 {@XMM[4]-@XMM[5]}, [$inp]! + mov r5,#10 @ pass rounds + vld1.8 {@XMM[6]-@XMM[7]}, [$inp]! + + bl _bsaes_decrypt8 + + vst1.8 {@XMM[0]-@XMM[1]}, [$out]! @ write output + vst1.8 {@XMM[6]}, [$out]! + vst1.8 {@XMM[4]}, [$out]! + vst1.8 {@XMM[2]}, [$out]! + vst1.8 {@XMM[7]}, [$out]! + vst1.8 {@XMM[3]}, [$out]! + subs $len,$len,#0x80 + vst1.8 {@XMM[5]}, [$out]! + bhi .Ldec128_loop + + vldmia sp!,{d8-d15} + ldmia sp!,{r4-r6,pc} +.size bsaes_decrypt_128,.-bsaes_decrypt_128 +___ +} +{ +my ($inp,$out,$len,$key, $ivp,$fp,$rounds)=map("r$_",(0..3,8..10)); +my ($keysched)=("sp"); + +$code.=<<___; +.extern AES_cbc_encrypt +.extern AES_decrypt + +.global bsaes_cbc_encrypt +.type bsaes_cbc_encrypt,%function +.align 5 +bsaes_cbc_encrypt: +#ifndef __KERNEL__ + cmp $len, #128 +#ifndef __thumb__ + blo AES_cbc_encrypt +#else + bhs 1f + b AES_cbc_encrypt +1: +#endif +#endif + + @ it is up to the caller to make sure we are called with enc == 0 + + mov ip, sp + stmdb sp!, {r4-r10, lr} + VFP_ABI_PUSH + ldr $ivp, [ip] @ IV is 1st arg on the stack + mov $len, $len, lsr#4 @ len in 16 byte blocks + sub sp, #0x10 @ scratch space to carry over the IV + mov $fp, sp @ save sp + + ldr $rounds, [$key, #240] @ get # of rounds +#ifndef BSAES_ASM_EXTENDED_KEY + @ allocate the key schedule on the stack + sub r12, sp, $rounds, lsl#7 @ 128 bytes per inner round key + add r12, #`128-32` @ sifze of bit-slices key schedule + + @ populate the key schedule + mov r4, $key @ pass key + mov r5, $rounds @ pass # of rounds + mov sp, r12 @ sp is $keysched + bl _bsaes_key_convert + vldmia $keysched, {@XMM[6]} + vstmia r12, {@XMM[15]} @ save last round key + veor @XMM[7], @XMM[7], @XMM[6] @ fix up round 0 key + vstmia $keysched, {@XMM[7]} +#else + ldr r12, [$key, #244] + eors r12, #1 + beq 0f + + @ populate the key schedule + str r12, [$key, #244] + mov r4, $key @ pass key + mov r5, $rounds @ pass # of rounds + add r12, $key, #248 @ pass key schedule + bl _bsaes_key_convert + add r4, $key, #248 + vldmia r4, {@XMM[6]} + vstmia r12, {@XMM[15]} @ save last round key + veor @XMM[7], @XMM[7], @XMM[6] @ fix up round 0 key + vstmia r4, {@XMM[7]} + +.align 2 +0: +#endif + + vld1.8 {@XMM[15]}, [$ivp] @ load IV + b .Lcbc_dec_loop + +.align 4 +.Lcbc_dec_loop: + subs $len, $len, #0x8 + bmi .Lcbc_dec_loop_finish + + vld1.8 {@XMM[0]-@XMM[1]}, [$inp]! @ load input + vld1.8 {@XMM[2]-@XMM[3]}, [$inp]! +#ifndef BSAES_ASM_EXTENDED_KEY + mov r4, $keysched @ pass the key +#else + add r4, $key, #248 +#endif + vld1.8 {@XMM[4]-@XMM[5]}, [$inp]! + mov r5, $rounds + vld1.8 {@XMM[6]-@XMM[7]}, [$inp] + sub $inp, $inp, #0x60 + vstmia $fp, {@XMM[15]} @ put aside IV + + bl _bsaes_decrypt8 + + vldmia $fp, {@XMM[14]} @ reload IV + vld1.8 {@XMM[8]-@XMM[9]}, [$inp]! @ reload input + veor @XMM[0], @XMM[0], @XMM[14] @ ^= IV + vld1.8 {@XMM[10]-@XMM[11]}, [$inp]! + veor @XMM[1], @XMM[1], @XMM[8] + veor @XMM[6], @XMM[6], @XMM[9] + vld1.8 {@XMM[12]-@XMM[13]}, [$inp]! + veor @XMM[4], @XMM[4], @XMM[10] + veor @XMM[2], @XMM[2], @XMM[11] + vld1.8 {@XMM[14]-@XMM[15]}, [$inp]! + veor @XMM[7], @XMM[7], @XMM[12] + vst1.8 {@XMM[0]-@XMM[1]}, [$out]! @ write output + veor @XMM[3], @XMM[3], @XMM[13] + vst1.8 {@XMM[6]}, [$out]! + veor @XMM[5], @XMM[5], @XMM[14] + vst1.8 {@XMM[4]}, [$out]! + vst1.8 {@XMM[2]}, [$out]! + vst1.8 {@XMM[7]}, [$out]! + vst1.8 {@XMM[3]}, [$out]! + vst1.8 {@XMM[5]}, [$out]! + + b .Lcbc_dec_loop + +.Lcbc_dec_loop_finish: + adds $len, $len, #8 + beq .Lcbc_dec_done + + vld1.8 {@XMM[0]}, [$inp]! @ load input + cmp $len, #2 + blo .Lcbc_dec_one + vld1.8 {@XMM[1]}, [$inp]! +#ifndef BSAES_ASM_EXTENDED_KEY + mov r4, $keysched @ pass the key +#else + add r4, $key, #248 +#endif + mov r5, $rounds + vstmia $fp, {@XMM[15]} @ put aside IV + beq .Lcbc_dec_two + vld1.8 {@XMM[2]}, [$inp]! + cmp $len, #4 + blo .Lcbc_dec_three + vld1.8 {@XMM[3]}, [$inp]! + beq .Lcbc_dec_four + vld1.8 {@XMM[4]}, [$inp]! + cmp $len, #6 + blo .Lcbc_dec_five + vld1.8 {@XMM[5]}, [$inp]! + beq .Lcbc_dec_six + vld1.8 {@XMM[6]}, [$inp]! + sub $inp, $inp, #0x70 + + bl _bsaes_decrypt8 + + vldmia $fp, {@XMM[14]} @ reload IV + vld1.8 {@XMM[8]-@XMM[9]}, [$inp]! @ reload input + veor @XMM[0], @XMM[0], @XMM[14] @ ^= IV + vld1.8 {@XMM[10]-@XMM[11]}, [$inp]! + veor @XMM[1], @XMM[1], @XMM[8] + veor @XMM[6], @XMM[6], @XMM[9] + vld1.8 {@XMM[12]-@XMM[13]}, [$inp]! + veor @XMM[4], @XMM[4], @XMM[10] + veor @XMM[2], @XMM[2], @XMM[11] + vld1.8 {@XMM[15]}, [$inp]! + veor @XMM[7], @XMM[7], @XMM[12] + vst1.8 {@XMM[0]-@XMM[1]}, [$out]! @ write output + veor @XMM[3], @XMM[3], @XMM[13] + vst1.8 {@XMM[6]}, [$out]! + vst1.8 {@XMM[4]}, [$out]! + vst1.8 {@XMM[2]}, [$out]! + vst1.8 {@XMM[7]}, [$out]! + vst1.8 {@XMM[3]}, [$out]! + b .Lcbc_dec_done +.align 4 +.Lcbc_dec_six: + sub $inp, $inp, #0x60 + bl _bsaes_decrypt8 + vldmia $fp,{@XMM[14]} @ reload IV + vld1.8 {@XMM[8]-@XMM[9]}, [$inp]! @ reload input + veor @XMM[0], @XMM[0], @XMM[14] @ ^= IV + vld1.8 {@XMM[10]-@XMM[11]}, [$inp]! + veor @XMM[1], @XMM[1], @XMM[8] + veor @XMM[6], @XMM[6], @XMM[9] + vld1.8 {@XMM[12]}, [$inp]! + veor @XMM[4], @XMM[4], @XMM[10] + veor @XMM[2], @XMM[2], @XMM[11] + vld1.8 {@XMM[15]}, [$inp]! + veor @XMM[7], @XMM[7], @XMM[12] + vst1.8 {@XMM[0]-@XMM[1]}, [$out]! @ write output + vst1.8 {@XMM[6]}, [$out]! + vst1.8 {@XMM[4]}, [$out]! + vst1.8 {@XMM[2]}, [$out]! + vst1.8 {@XMM[7]}, [$out]! + b .Lcbc_dec_done +.align 4 +.Lcbc_dec_five: + sub $inp, $inp, #0x50 + bl _bsaes_decrypt8 + vldmia $fp, {@XMM[14]} @ reload IV + vld1.8 {@XMM[8]-@XMM[9]}, [$inp]! @ reload input + veor @XMM[0], @XMM[0], @XMM[14] @ ^= IV + vld1.8 {@XMM[10]-@XMM[11]}, [$inp]! + veor @XMM[1], @XMM[1], @XMM[8] + veor @XMM[6], @XMM[6], @XMM[9] + vld1.8 {@XMM[15]}, [$inp]! + veor @XMM[4], @XMM[4], @XMM[10] + vst1.8 {@XMM[0]-@XMM[1]}, [$out]! @ write output + veor @XMM[2], @XMM[2], @XMM[11] + vst1.8 {@XMM[6]}, [$out]! + vst1.8 {@XMM[4]}, [$out]! + vst1.8 {@XMM[2]}, [$out]! + b .Lcbc_dec_done +.align 4 +.Lcbc_dec_four: + sub $inp, $inp, #0x40 + bl _bsaes_decrypt8 + vldmia $fp, {@XMM[14]} @ reload IV + vld1.8 {@XMM[8]-@XMM[9]}, [$inp]! @ reload input + veor @XMM[0], @XMM[0], @XMM[14] @ ^= IV + vld1.8 {@XMM[10]}, [$inp]! + veor @XMM[1], @XMM[1], @XMM[8] + veor @XMM[6], @XMM[6], @XMM[9] + vld1.8 {@XMM[15]}, [$inp]! + veor @XMM[4], @XMM[4], @XMM[10] + vst1.8 {@XMM[0]-@XMM[1]}, [$out]! @ write output + vst1.8 {@XMM[6]}, [$out]! + vst1.8 {@XMM[4]}, [$out]! + b .Lcbc_dec_done +.align 4 +.Lcbc_dec_three: + sub $inp, $inp, #0x30 + bl _bsaes_decrypt8 + vldmia $fp, {@XMM[14]} @ reload IV + vld1.8 {@XMM[8]-@XMM[9]}, [$inp]! @ reload input + veor @XMM[0], @XMM[0], @XMM[14] @ ^= IV + vld1.8 {@XMM[15]}, [$inp]! + veor @XMM[1], @XMM[1], @XMM[8] + veor @XMM[6], @XMM[6], @XMM[9] + vst1.8 {@XMM[0]-@XMM[1]}, [$out]! @ write output + vst1.8 {@XMM[6]}, [$out]! + b .Lcbc_dec_done +.align 4 +.Lcbc_dec_two: + sub $inp, $inp, #0x20 + bl _bsaes_decrypt8 + vldmia $fp, {@XMM[14]} @ reload IV + vld1.8 {@XMM[8]}, [$inp]! @ reload input + veor @XMM[0], @XMM[0], @XMM[14] @ ^= IV + vld1.8 {@XMM[15]}, [$inp]! @ reload input + veor @XMM[1], @XMM[1], @XMM[8] + vst1.8 {@XMM[0]-@XMM[1]}, [$out]! @ write output + b .Lcbc_dec_done +.align 4 +.Lcbc_dec_one: + sub $inp, $inp, #0x10 + mov $rounds, $out @ save original out pointer + mov $out, $fp @ use the iv scratch space as out buffer + mov r2, $key + vmov @XMM[4],@XMM[15] @ just in case ensure that IV + vmov @XMM[5],@XMM[0] @ and input are preserved + bl AES_decrypt + vld1.8 {@XMM[0]}, [$fp] @ load result + veor @XMM[0], @XMM[0], @XMM[4] @ ^= IV + vmov @XMM[15], @XMM[5] @ @XMM[5] holds input + vst1.8 {@XMM[0]}, [$rounds] @ write output + +.Lcbc_dec_done: +#ifndef BSAES_ASM_EXTENDED_KEY + vmov.i32 q0, #0 + vmov.i32 q1, #0 +.Lcbc_dec_bzero: @ wipe key schedule [if any] + vstmia $keysched!, {q0-q1} + cmp $keysched, $fp + bne .Lcbc_dec_bzero +#endif + + mov sp, $fp + add sp, #0x10 @ add sp,$fp,#0x10 is no good for thumb + vst1.8 {@XMM[15]}, [$ivp] @ return IV + VFP_ABI_POP + ldmia sp!, {r4-r10, pc} +.size bsaes_cbc_encrypt,.-bsaes_cbc_encrypt +___ +} +{ +my ($inp,$out,$len,$key, $ctr,$fp,$rounds)=(map("r$_",(0..3,8..10))); +my $const = "r6"; # shared with _bsaes_encrypt8_alt +my $keysched = "sp"; + +$code.=<<___; +.extern AES_encrypt +.global bsaes_ctr32_encrypt_blocks +.type bsaes_ctr32_encrypt_blocks,%function +.align 5 +bsaes_ctr32_encrypt_blocks: + cmp $len, #8 @ use plain AES for + blo .Lctr_enc_short @ small sizes + + mov ip, sp + stmdb sp!, {r4-r10, lr} + VFP_ABI_PUSH + ldr $ctr, [ip] @ ctr is 1st arg on the stack + sub sp, sp, #0x10 @ scratch space to carry over the ctr + mov $fp, sp @ save sp + + ldr $rounds, [$key, #240] @ get # of rounds +#ifndef BSAES_ASM_EXTENDED_KEY + @ allocate the key schedule on the stack + sub r12, sp, $rounds, lsl#7 @ 128 bytes per inner round key + add r12, #`128-32` @ size of bit-sliced key schedule + + @ populate the key schedule + mov r4, $key @ pass key + mov r5, $rounds @ pass # of rounds + mov sp, r12 @ sp is $keysched + bl _bsaes_key_convert + veor @XMM[7],@XMM[7],@XMM[15] @ fix up last round key + vstmia r12, {@XMM[7]} @ save last round key + + vld1.8 {@XMM[0]}, [$ctr] @ load counter +#ifdef __APPLE__ + mov $ctr, #:lower16:(.LREVM0SR-.LM0) + add $ctr, $const, $ctr +#else + add $ctr, $const, #.LREVM0SR-.LM0 @ borrow $ctr +#endif + vldmia $keysched, {@XMM[4]} @ load round0 key +#else + ldr r12, [$key, #244] + eors r12, #1 + beq 0f + + @ populate the key schedule + str r12, [$key, #244] + mov r4, $key @ pass key + mov r5, $rounds @ pass # of rounds + add r12, $key, #248 @ pass key schedule + bl _bsaes_key_convert + veor @XMM[7],@XMM[7],@XMM[15] @ fix up last round key + vstmia r12, {@XMM[7]} @ save last round key + +.align 2 +0: add r12, $key, #248 + vld1.8 {@XMM[0]}, [$ctr] @ load counter + adrl $ctr, .LREVM0SR @ borrow $ctr + vldmia r12, {@XMM[4]} @ load round0 key + sub sp, #0x10 @ place for adjusted round0 key +#endif + + vmov.i32 @XMM[8],#1 @ compose 1<<96 + veor @XMM[9],@XMM[9],@XMM[9] + vrev32.8 @XMM[0],@XMM[0] + vext.8 @XMM[8],@XMM[9],@XMM[8],#4 + vrev32.8 @XMM[4],@XMM[4] + vadd.u32 @XMM[9],@XMM[8],@XMM[8] @ compose 2<<96 + vstmia $keysched, {@XMM[4]} @ save adjusted round0 key + b .Lctr_enc_loop + +.align 4 +.Lctr_enc_loop: + vadd.u32 @XMM[10], @XMM[8], @XMM[9] @ compose 3<<96 + vadd.u32 @XMM[1], @XMM[0], @XMM[8] @ +1 + vadd.u32 @XMM[2], @XMM[0], @XMM[9] @ +2 + vadd.u32 @XMM[3], @XMM[0], @XMM[10] @ +3 + vadd.u32 @XMM[4], @XMM[1], @XMM[10] + vadd.u32 @XMM[5], @XMM[2], @XMM[10] + vadd.u32 @XMM[6], @XMM[3], @XMM[10] + vadd.u32 @XMM[7], @XMM[4], @XMM[10] + vadd.u32 @XMM[10], @XMM[5], @XMM[10] @ next counter + + @ Borrow prologue from _bsaes_encrypt8 to use the opportunity + @ to flip byte order in 32-bit counter + + vldmia $keysched, {@XMM[9]} @ load round0 key +#ifndef BSAES_ASM_EXTENDED_KEY + add r4, $keysched, #0x10 @ pass next round key +#else + add r4, $key, #`248+16` +#endif + vldmia $ctr, {@XMM[8]} @ .LREVM0SR + mov r5, $rounds @ pass rounds + vstmia $fp, {@XMM[10]} @ save next counter +#ifdef __APPLE__ + mov $const, #:lower16:(.LREVM0SR-.LSR) + sub $const, $ctr, $const +#else + sub $const, $ctr, #.LREVM0SR-.LSR @ pass constants +#endif + + bl _bsaes_encrypt8_alt + + subs $len, $len, #8 + blo .Lctr_enc_loop_done + + vld1.8 {@XMM[8]-@XMM[9]}, [$inp]! @ load input + vld1.8 {@XMM[10]-@XMM[11]}, [$inp]! + veor @XMM[0], @XMM[8] + veor @XMM[1], @XMM[9] + vld1.8 {@XMM[12]-@XMM[13]}, [$inp]! + veor @XMM[4], @XMM[10] + veor @XMM[6], @XMM[11] + vld1.8 {@XMM[14]-@XMM[15]}, [$inp]! + veor @XMM[3], @XMM[12] + vst1.8 {@XMM[0]-@XMM[1]}, [$out]! @ write output + veor @XMM[7], @XMM[13] + veor @XMM[2], @XMM[14] + vst1.8 {@XMM[4]}, [$out]! + veor @XMM[5], @XMM[15] + vst1.8 {@XMM[6]}, [$out]! + vmov.i32 @XMM[8], #1 @ compose 1<<96 + vst1.8 {@XMM[3]}, [$out]! + veor @XMM[9], @XMM[9], @XMM[9] + vst1.8 {@XMM[7]}, [$out]! + vext.8 @XMM[8], @XMM[9], @XMM[8], #4 + vst1.8 {@XMM[2]}, [$out]! + vadd.u32 @XMM[9],@XMM[8],@XMM[8] @ compose 2<<96 + vst1.8 {@XMM[5]}, [$out]! + vldmia $fp, {@XMM[0]} @ load counter + + bne .Lctr_enc_loop + b .Lctr_enc_done + +.align 4 +.Lctr_enc_loop_done: + add $len, $len, #8 + vld1.8 {@XMM[8]}, [$inp]! @ load input + veor @XMM[0], @XMM[8] + vst1.8 {@XMM[0]}, [$out]! @ write output + cmp $len, #2 + blo .Lctr_enc_done + vld1.8 {@XMM[9]}, [$inp]! + veor @XMM[1], @XMM[9] + vst1.8 {@XMM[1]}, [$out]! + beq .Lctr_enc_done + vld1.8 {@XMM[10]}, [$inp]! + veor @XMM[4], @XMM[10] + vst1.8 {@XMM[4]}, [$out]! + cmp $len, #4 + blo .Lctr_enc_done + vld1.8 {@XMM[11]}, [$inp]! + veor @XMM[6], @XMM[11] + vst1.8 {@XMM[6]}, [$out]! + beq .Lctr_enc_done + vld1.8 {@XMM[12]}, [$inp]! + veor @XMM[3], @XMM[12] + vst1.8 {@XMM[3]}, [$out]! + cmp $len, #6 + blo .Lctr_enc_done + vld1.8 {@XMM[13]}, [$inp]! + veor @XMM[7], @XMM[13] + vst1.8 {@XMM[7]}, [$out]! + beq .Lctr_enc_done + vld1.8 {@XMM[14]}, [$inp] + veor @XMM[2], @XMM[14] + vst1.8 {@XMM[2]}, [$out]! + +.Lctr_enc_done: + vmov.i32 q0, #0 + vmov.i32 q1, #0 +#ifndef BSAES_ASM_EXTENDED_KEY +.Lctr_enc_bzero: @ wipe key schedule [if any] + vstmia $keysched!, {q0-q1} + cmp $keysched, $fp + bne .Lctr_enc_bzero +#else + vstmia $keysched, {q0-q1} +#endif + + mov sp, $fp + add sp, #0x10 @ add sp,$fp,#0x10 is no good for thumb + VFP_ABI_POP + ldmia sp!, {r4-r10, pc} @ return + +.align 4 +.Lctr_enc_short: + ldr ip, [sp] @ ctr pointer is passed on stack + stmdb sp!, {r4-r8, lr} + + mov r4, $inp @ copy arguments + mov r5, $out + mov r6, $len + mov r7, $key + ldr r8, [ip, #12] @ load counter LSW + vld1.8 {@XMM[1]}, [ip] @ load whole counter value +#ifdef __ARMEL__ + rev r8, r8 +#endif + sub sp, sp, #0x10 + vst1.8 {@XMM[1]}, [sp] @ copy counter value + sub sp, sp, #0x10 + +.Lctr_enc_short_loop: + add r0, sp, #0x10 @ input counter value + mov r1, sp @ output on the stack + mov r2, r7 @ key + + bl AES_encrypt + + vld1.8 {@XMM[0]}, [r4]! @ load input + vld1.8 {@XMM[1]}, [sp] @ load encrypted counter + add r8, r8, #1 +#ifdef __ARMEL__ + rev r0, r8 + str r0, [sp, #0x1c] @ next counter value +#else + str r8, [sp, #0x1c] @ next counter value +#endif + veor @XMM[0],@XMM[0],@XMM[1] + vst1.8 {@XMM[0]}, [r5]! @ store output + subs r6, r6, #1 + bne .Lctr_enc_short_loop + + vmov.i32 q0, #0 + vmov.i32 q1, #0 + vstmia sp!, {q0-q1} + + ldmia sp!, {r4-r8, pc} +.size bsaes_ctr32_encrypt_blocks,.-bsaes_ctr32_encrypt_blocks +___ +} +{ +###################################################################### +# void bsaes_xts_[en|de]crypt(const char *inp,char *out,size_t len, +# const AES_KEY *key1, const AES_KEY *key2, +# const unsigned char iv[16]); +# +my ($inp,$out,$len,$key,$rounds,$magic,$fp)=(map("r$_",(7..10,1..3))); +my $const="r6"; # returned by _bsaes_key_convert +my $twmask=@XMM[5]; +my @T=@XMM[6..7]; + +$code.=<<___; +.globl bsaes_xts_encrypt +.type bsaes_xts_encrypt,%function +.align 4 +bsaes_xts_encrypt: + mov ip, sp + stmdb sp!, {r4-r10, lr} @ 0x20 + VFP_ABI_PUSH + mov r6, sp @ future $fp + + mov $inp, r0 + mov $out, r1 + mov $len, r2 + mov $key, r3 + + sub r0, sp, #0x10 @ 0x10 + bic r0, #0xf @ align at 16 bytes + mov sp, r0 + +#ifdef XTS_CHAIN_TWEAK + ldr r0, [ip] @ pointer to input tweak +#else + @ generate initial tweak + ldr r0, [ip, #4] @ iv[] + mov r1, sp + ldr r2, [ip, #0] @ key2 + bl AES_encrypt + mov r0,sp @ pointer to initial tweak +#endif + + ldr $rounds, [$key, #240] @ get # of rounds + mov $fp, r6 +#ifndef BSAES_ASM_EXTENDED_KEY + @ allocate the key schedule on the stack + sub r12, sp, $rounds, lsl#7 @ 128 bytes per inner round key + @ add r12, #`128-32` @ size of bit-sliced key schedule + sub r12, #`32+16` @ place for tweak[9] + + @ populate the key schedule + mov r4, $key @ pass key + mov r5, $rounds @ pass # of rounds + mov sp, r12 + add r12, #0x90 @ pass key schedule + bl _bsaes_key_convert + veor @XMM[7], @XMM[7], @XMM[15] @ fix up last round key + vstmia r12, {@XMM[7]} @ save last round key +#else + ldr r12, [$key, #244] + eors r12, #1 + beq 0f + + str r12, [$key, #244] + mov r4, $key @ pass key + mov r5, $rounds @ pass # of rounds + add r12, $key, #248 @ pass key schedule + bl _bsaes_key_convert + veor @XMM[7], @XMM[7], @XMM[15] @ fix up last round key + vstmia r12, {@XMM[7]} + +.align 2 +0: sub sp, #0x90 @ place for tweak[9] +#endif + + vld1.8 {@XMM[8]}, [r0] @ initial tweak + adr $magic, .Lxts_magic + + subs $len, #0x80 + blo .Lxts_enc_short + b .Lxts_enc_loop + +.align 4 +.Lxts_enc_loop: + vldmia $magic, {$twmask} @ load XTS magic + vshr.s64 @T[0], @XMM[8], #63 + mov r0, sp + vand @T[0], @T[0], $twmask +___ +for($i=9;$i<16;$i++) { +$code.=<<___; + vadd.u64 @XMM[$i], @XMM[$i-1], @XMM[$i-1] + vst1.64 {@XMM[$i-1]}, [r0,:128]! + vswp `&Dhi("@T[0]")`,`&Dlo("@T[0]")` + vshr.s64 @T[1], @XMM[$i], #63 + veor @XMM[$i], @XMM[$i], @T[0] + vand @T[1], @T[1], $twmask +___ + @T=reverse(@T); + +$code.=<<___ if ($i>=10); + vld1.8 {@XMM[$i-10]}, [$inp]! +___ +$code.=<<___ if ($i>=11); + veor @XMM[$i-11], @XMM[$i-11], @XMM[$i-3] +___ +} +$code.=<<___; + vadd.u64 @XMM[8], @XMM[15], @XMM[15] + vst1.64 {@XMM[15]}, [r0,:128]! + vswp `&Dhi("@T[0]")`,`&Dlo("@T[0]")` + veor @XMM[8], @XMM[8], @T[0] + vst1.64 {@XMM[8]}, [r0,:128] @ next round tweak + + vld1.8 {@XMM[6]-@XMM[7]}, [$inp]! + veor @XMM[5], @XMM[5], @XMM[13] +#ifndef BSAES_ASM_EXTENDED_KEY + add r4, sp, #0x90 @ pass key schedule +#else + add r4, $key, #248 @ pass key schedule +#endif + veor @XMM[6], @XMM[6], @XMM[14] + mov r5, $rounds @ pass rounds + veor @XMM[7], @XMM[7], @XMM[15] + mov r0, sp + + bl _bsaes_encrypt8 + + vld1.64 {@XMM[ 8]-@XMM[ 9]}, [r0,:128]! + vld1.64 {@XMM[10]-@XMM[11]}, [r0,:128]! + veor @XMM[0], @XMM[0], @XMM[ 8] + vld1.64 {@XMM[12]-@XMM[13]}, [r0,:128]! + veor @XMM[1], @XMM[1], @XMM[ 9] + veor @XMM[8], @XMM[4], @XMM[10] + vst1.8 {@XMM[0]-@XMM[1]}, [$out]! + veor @XMM[9], @XMM[6], @XMM[11] + vld1.64 {@XMM[14]-@XMM[15]}, [r0,:128]! + veor @XMM[10], @XMM[3], @XMM[12] + vst1.8 {@XMM[8]-@XMM[9]}, [$out]! + veor @XMM[11], @XMM[7], @XMM[13] + veor @XMM[12], @XMM[2], @XMM[14] + vst1.8 {@XMM[10]-@XMM[11]}, [$out]! + veor @XMM[13], @XMM[5], @XMM[15] + vst1.8 {@XMM[12]-@XMM[13]}, [$out]! + + vld1.64 {@XMM[8]}, [r0,:128] @ next round tweak + + subs $len, #0x80 + bpl .Lxts_enc_loop + +.Lxts_enc_short: + adds $len, #0x70 + bmi .Lxts_enc_done + + vldmia $magic, {$twmask} @ load XTS magic + vshr.s64 @T[0], @XMM[8], #63 + mov r0, sp + vand @T[0], @T[0], $twmask +___ +for($i=9;$i<16;$i++) { +$code.=<<___; + vadd.u64 @XMM[$i], @XMM[$i-1], @XMM[$i-1] + vst1.64 {@XMM[$i-1]}, [r0,:128]! + vswp `&Dhi("@T[0]")`,`&Dlo("@T[0]")` + vshr.s64 @T[1], @XMM[$i], #63 + veor @XMM[$i], @XMM[$i], @T[0] + vand @T[1], @T[1], $twmask +___ + @T=reverse(@T); + +$code.=<<___ if ($i>=10); + vld1.8 {@XMM[$i-10]}, [$inp]! + subs $len, #0x10 + bmi .Lxts_enc_`$i-9` +___ +$code.=<<___ if ($i>=11); + veor @XMM[$i-11], @XMM[$i-11], @XMM[$i-3] +___ +} +$code.=<<___; + sub $len, #0x10 + vst1.64 {@XMM[15]}, [r0,:128] @ next round tweak + + vld1.8 {@XMM[6]}, [$inp]! + veor @XMM[5], @XMM[5], @XMM[13] +#ifndef BSAES_ASM_EXTENDED_KEY + add r4, sp, #0x90 @ pass key schedule +#else + add r4, $key, #248 @ pass key schedule +#endif + veor @XMM[6], @XMM[6], @XMM[14] + mov r5, $rounds @ pass rounds + mov r0, sp + + bl _bsaes_encrypt8 + + vld1.64 {@XMM[ 8]-@XMM[ 9]}, [r0,:128]! + vld1.64 {@XMM[10]-@XMM[11]}, [r0,:128]! + veor @XMM[0], @XMM[0], @XMM[ 8] + vld1.64 {@XMM[12]-@XMM[13]}, [r0,:128]! + veor @XMM[1], @XMM[1], @XMM[ 9] + veor @XMM[8], @XMM[4], @XMM[10] + vst1.8 {@XMM[0]-@XMM[1]}, [$out]! + veor @XMM[9], @XMM[6], @XMM[11] + vld1.64 {@XMM[14]}, [r0,:128]! + veor @XMM[10], @XMM[3], @XMM[12] + vst1.8 {@XMM[8]-@XMM[9]}, [$out]! + veor @XMM[11], @XMM[7], @XMM[13] + veor @XMM[12], @XMM[2], @XMM[14] + vst1.8 {@XMM[10]-@XMM[11]}, [$out]! + vst1.8 {@XMM[12]}, [$out]! + + vld1.64 {@XMM[8]}, [r0,:128] @ next round tweak + b .Lxts_enc_done +.align 4 +.Lxts_enc_6: + veor @XMM[4], @XMM[4], @XMM[12] +#ifndef BSAES_ASM_EXTENDED_KEY + add r4, sp, #0x90 @ pass key schedule +#else + add r4, $key, #248 @ pass key schedule +#endif + veor @XMM[5], @XMM[5], @XMM[13] + mov r5, $rounds @ pass rounds + mov r0, sp + + bl _bsaes_encrypt8 + + vld1.64 {@XMM[ 8]-@XMM[ 9]}, [r0,:128]! + vld1.64 {@XMM[10]-@XMM[11]}, [r0,:128]! + veor @XMM[0], @XMM[0], @XMM[ 8] + vld1.64 {@XMM[12]-@XMM[13]}, [r0,:128]! + veor @XMM[1], @XMM[1], @XMM[ 9] + veor @XMM[8], @XMM[4], @XMM[10] + vst1.8 {@XMM[0]-@XMM[1]}, [$out]! + veor @XMM[9], @XMM[6], @XMM[11] + veor @XMM[10], @XMM[3], @XMM[12] + vst1.8 {@XMM[8]-@XMM[9]}, [$out]! + veor @XMM[11], @XMM[7], @XMM[13] + vst1.8 {@XMM[10]-@XMM[11]}, [$out]! + + vld1.64 {@XMM[8]}, [r0,:128] @ next round tweak + b .Lxts_enc_done + +@ put this in range for both ARM and Thumb mode adr instructions +.align 5 +.Lxts_magic: + .quad 1, 0x87 + +.align 5 +.Lxts_enc_5: + veor @XMM[3], @XMM[3], @XMM[11] +#ifndef BSAES_ASM_EXTENDED_KEY + add r4, sp, #0x90 @ pass key schedule +#else + add r4, $key, #248 @ pass key schedule +#endif + veor @XMM[4], @XMM[4], @XMM[12] + mov r5, $rounds @ pass rounds + mov r0, sp + + bl _bsaes_encrypt8 + + vld1.64 {@XMM[ 8]-@XMM[ 9]}, [r0,:128]! + vld1.64 {@XMM[10]-@XMM[11]}, [r0,:128]! + veor @XMM[0], @XMM[0], @XMM[ 8] + vld1.64 {@XMM[12]}, [r0,:128]! + veor @XMM[1], @XMM[1], @XMM[ 9] + veor @XMM[8], @XMM[4], @XMM[10] + vst1.8 {@XMM[0]-@XMM[1]}, [$out]! + veor @XMM[9], @XMM[6], @XMM[11] + veor @XMM[10], @XMM[3], @XMM[12] + vst1.8 {@XMM[8]-@XMM[9]}, [$out]! + vst1.8 {@XMM[10]}, [$out]! + + vld1.64 {@XMM[8]}, [r0,:128] @ next round tweak + b .Lxts_enc_done +.align 4 +.Lxts_enc_4: + veor @XMM[2], @XMM[2], @XMM[10] +#ifndef BSAES_ASM_EXTENDED_KEY + add r4, sp, #0x90 @ pass key schedule +#else + add r4, $key, #248 @ pass key schedule +#endif + veor @XMM[3], @XMM[3], @XMM[11] + mov r5, $rounds @ pass rounds + mov r0, sp + + bl _bsaes_encrypt8 + + vld1.64 {@XMM[ 8]-@XMM[ 9]}, [r0,:128]! + vld1.64 {@XMM[10]-@XMM[11]}, [r0,:128]! + veor @XMM[0], @XMM[0], @XMM[ 8] + veor @XMM[1], @XMM[1], @XMM[ 9] + veor @XMM[8], @XMM[4], @XMM[10] + vst1.8 {@XMM[0]-@XMM[1]}, [$out]! + veor @XMM[9], @XMM[6], @XMM[11] + vst1.8 {@XMM[8]-@XMM[9]}, [$out]! + + vld1.64 {@XMM[8]}, [r0,:128] @ next round tweak + b .Lxts_enc_done +.align 4 +.Lxts_enc_3: + veor @XMM[1], @XMM[1], @XMM[9] +#ifndef BSAES_ASM_EXTENDED_KEY + add r4, sp, #0x90 @ pass key schedule +#else + add r4, $key, #248 @ pass key schedule +#endif + veor @XMM[2], @XMM[2], @XMM[10] + mov r5, $rounds @ pass rounds + mov r0, sp + + bl _bsaes_encrypt8 + + vld1.64 {@XMM[8]-@XMM[9]}, [r0,:128]! + vld1.64 {@XMM[10]}, [r0,:128]! + veor @XMM[0], @XMM[0], @XMM[ 8] + veor @XMM[1], @XMM[1], @XMM[ 9] + veor @XMM[8], @XMM[4], @XMM[10] + vst1.8 {@XMM[0]-@XMM[1]}, [$out]! + vst1.8 {@XMM[8]}, [$out]! + + vld1.64 {@XMM[8]}, [r0,:128] @ next round tweak + b .Lxts_enc_done +.align 4 +.Lxts_enc_2: + veor @XMM[0], @XMM[0], @XMM[8] +#ifndef BSAES_ASM_EXTENDED_KEY + add r4, sp, #0x90 @ pass key schedule +#else + add r4, $key, #248 @ pass key schedule +#endif + veor @XMM[1], @XMM[1], @XMM[9] + mov r5, $rounds @ pass rounds + mov r0, sp + + bl _bsaes_encrypt8 + + vld1.64 {@XMM[8]-@XMM[9]}, [r0,:128]! + veor @XMM[0], @XMM[0], @XMM[ 8] + veor @XMM[1], @XMM[1], @XMM[ 9] + vst1.8 {@XMM[0]-@XMM[1]}, [$out]! + + vld1.64 {@XMM[8]}, [r0,:128] @ next round tweak + b .Lxts_enc_done +.align 4 +.Lxts_enc_1: + mov r0, sp + veor @XMM[0], @XMM[0], @XMM[8] + mov r1, sp + vst1.8 {@XMM[0]}, [sp,:128] + mov r2, $key + mov r4, $fp @ preserve fp + + bl AES_encrypt + + vld1.8 {@XMM[0]}, [sp,:128] + veor @XMM[0], @XMM[0], @XMM[8] + vst1.8 {@XMM[0]}, [$out]! + mov $fp, r4 + + vmov @XMM[8], @XMM[9] @ next round tweak + +.Lxts_enc_done: +#ifndef XTS_CHAIN_TWEAK + adds $len, #0x10 + beq .Lxts_enc_ret + sub r6, $out, #0x10 + +.Lxts_enc_steal: + ldrb r0, [$inp], #1 + ldrb r1, [$out, #-0x10] + strb r0, [$out, #-0x10] + strb r1, [$out], #1 + + subs $len, #1 + bhi .Lxts_enc_steal + + vld1.8 {@XMM[0]}, [r6] + mov r0, sp + veor @XMM[0], @XMM[0], @XMM[8] + mov r1, sp + vst1.8 {@XMM[0]}, [sp,:128] + mov r2, $key + mov r4, $fp @ preserve fp + + bl AES_encrypt + + vld1.8 {@XMM[0]}, [sp,:128] + veor @XMM[0], @XMM[0], @XMM[8] + vst1.8 {@XMM[0]}, [r6] + mov $fp, r4 +#endif + +.Lxts_enc_ret: + bic r0, $fp, #0xf + vmov.i32 q0, #0 + vmov.i32 q1, #0 +#ifdef XTS_CHAIN_TWEAK + ldr r1, [$fp, #0x20+VFP_ABI_FRAME] @ chain tweak +#endif +.Lxts_enc_bzero: @ wipe key schedule [if any] + vstmia sp!, {q0-q1} + cmp sp, r0 + bne .Lxts_enc_bzero + + mov sp, $fp +#ifdef XTS_CHAIN_TWEAK + vst1.8 {@XMM[8]}, [r1] +#endif + VFP_ABI_POP + ldmia sp!, {r4-r10, pc} @ return + +.size bsaes_xts_encrypt,.-bsaes_xts_encrypt + +.globl bsaes_xts_decrypt +.type bsaes_xts_decrypt,%function +.align 4 +bsaes_xts_decrypt: + mov ip, sp + stmdb sp!, {r4-r10, lr} @ 0x20 + VFP_ABI_PUSH + mov r6, sp @ future $fp + + mov $inp, r0 + mov $out, r1 + mov $len, r2 + mov $key, r3 + + sub r0, sp, #0x10 @ 0x10 + bic r0, #0xf @ align at 16 bytes + mov sp, r0 + +#ifdef XTS_CHAIN_TWEAK + ldr r0, [ip] @ pointer to input tweak +#else + @ generate initial tweak + ldr r0, [ip, #4] @ iv[] + mov r1, sp + ldr r2, [ip, #0] @ key2 + bl AES_encrypt + mov r0, sp @ pointer to initial tweak +#endif + + ldr $rounds, [$key, #240] @ get # of rounds + mov $fp, r6 +#ifndef BSAES_ASM_EXTENDED_KEY + @ allocate the key schedule on the stack + sub r12, sp, $rounds, lsl#7 @ 128 bytes per inner round key + @ add r12, #`128-32` @ size of bit-sliced key schedule + sub r12, #`32+16` @ place for tweak[9] + + @ populate the key schedule + mov r4, $key @ pass key + mov r5, $rounds @ pass # of rounds + mov sp, r12 + add r12, #0x90 @ pass key schedule + bl _bsaes_key_convert + add r4, sp, #0x90 + vldmia r4, {@XMM[6]} + vstmia r12, {@XMM[15]} @ save last round key + veor @XMM[7], @XMM[7], @XMM[6] @ fix up round 0 key + vstmia r4, {@XMM[7]} +#else + ldr r12, [$key, #244] + eors r12, #1 + beq 0f + + str r12, [$key, #244] + mov r4, $key @ pass key + mov r5, $rounds @ pass # of rounds + add r12, $key, #248 @ pass key schedule + bl _bsaes_key_convert + add r4, $key, #248 + vldmia r4, {@XMM[6]} + vstmia r12, {@XMM[15]} @ save last round key + veor @XMM[7], @XMM[7], @XMM[6] @ fix up round 0 key + vstmia r4, {@XMM[7]} + +.align 2 +0: sub sp, #0x90 @ place for tweak[9] +#endif + vld1.8 {@XMM[8]}, [r0] @ initial tweak + adr $magic, .Lxts_magic + +#ifndef XTS_CHAIN_TWEAK + tst $len, #0xf @ if not multiple of 16 + it ne @ Thumb2 thing, sanity check in ARM + subne $len, #0x10 @ subtract another 16 bytes +#endif + subs $len, #0x80 + + blo .Lxts_dec_short + b .Lxts_dec_loop + +.align 4 +.Lxts_dec_loop: + vldmia $magic, {$twmask} @ load XTS magic + vshr.s64 @T[0], @XMM[8], #63 + mov r0, sp + vand @T[0], @T[0], $twmask +___ +for($i=9;$i<16;$i++) { +$code.=<<___; + vadd.u64 @XMM[$i], @XMM[$i-1], @XMM[$i-1] + vst1.64 {@XMM[$i-1]}, [r0,:128]! + vswp `&Dhi("@T[0]")`,`&Dlo("@T[0]")` + vshr.s64 @T[1], @XMM[$i], #63 + veor @XMM[$i], @XMM[$i], @T[0] + vand @T[1], @T[1], $twmask +___ + @T=reverse(@T); + +$code.=<<___ if ($i>=10); + vld1.8 {@XMM[$i-10]}, [$inp]! +___ +$code.=<<___ if ($i>=11); + veor @XMM[$i-11], @XMM[$i-11], @XMM[$i-3] +___ +} +$code.=<<___; + vadd.u64 @XMM[8], @XMM[15], @XMM[15] + vst1.64 {@XMM[15]}, [r0,:128]! + vswp `&Dhi("@T[0]")`,`&Dlo("@T[0]")` + veor @XMM[8], @XMM[8], @T[0] + vst1.64 {@XMM[8]}, [r0,:128] @ next round tweak + + vld1.8 {@XMM[6]-@XMM[7]}, [$inp]! + veor @XMM[5], @XMM[5], @XMM[13] +#ifndef BSAES_ASM_EXTENDED_KEY + add r4, sp, #0x90 @ pass key schedule +#else + add r4, $key, #248 @ pass key schedule +#endif + veor @XMM[6], @XMM[6], @XMM[14] + mov r5, $rounds @ pass rounds + veor @XMM[7], @XMM[7], @XMM[15] + mov r0, sp + + bl _bsaes_decrypt8 + + vld1.64 {@XMM[ 8]-@XMM[ 9]}, [r0,:128]! + vld1.64 {@XMM[10]-@XMM[11]}, [r0,:128]! + veor @XMM[0], @XMM[0], @XMM[ 8] + vld1.64 {@XMM[12]-@XMM[13]}, [r0,:128]! + veor @XMM[1], @XMM[1], @XMM[ 9] + veor @XMM[8], @XMM[6], @XMM[10] + vst1.8 {@XMM[0]-@XMM[1]}, [$out]! + veor @XMM[9], @XMM[4], @XMM[11] + vld1.64 {@XMM[14]-@XMM[15]}, [r0,:128]! + veor @XMM[10], @XMM[2], @XMM[12] + vst1.8 {@XMM[8]-@XMM[9]}, [$out]! + veor @XMM[11], @XMM[7], @XMM[13] + veor @XMM[12], @XMM[3], @XMM[14] + vst1.8 {@XMM[10]-@XMM[11]}, [$out]! + veor @XMM[13], @XMM[5], @XMM[15] + vst1.8 {@XMM[12]-@XMM[13]}, [$out]! + + vld1.64 {@XMM[8]}, [r0,:128] @ next round tweak + + subs $len, #0x80 + bpl .Lxts_dec_loop + +.Lxts_dec_short: + adds $len, #0x70 + bmi .Lxts_dec_done + + vldmia $magic, {$twmask} @ load XTS magic + vshr.s64 @T[0], @XMM[8], #63 + mov r0, sp + vand @T[0], @T[0], $twmask +___ +for($i=9;$i<16;$i++) { +$code.=<<___; + vadd.u64 @XMM[$i], @XMM[$i-1], @XMM[$i-1] + vst1.64 {@XMM[$i-1]}, [r0,:128]! + vswp `&Dhi("@T[0]")`,`&Dlo("@T[0]")` + vshr.s64 @T[1], @XMM[$i], #63 + veor @XMM[$i], @XMM[$i], @T[0] + vand @T[1], @T[1], $twmask +___ + @T=reverse(@T); + +$code.=<<___ if ($i>=10); + vld1.8 {@XMM[$i-10]}, [$inp]! + subs $len, #0x10 + bmi .Lxts_dec_`$i-9` +___ +$code.=<<___ if ($i>=11); + veor @XMM[$i-11], @XMM[$i-11], @XMM[$i-3] +___ +} +$code.=<<___; + sub $len, #0x10 + vst1.64 {@XMM[15]}, [r0,:128] @ next round tweak + + vld1.8 {@XMM[6]}, [$inp]! + veor @XMM[5], @XMM[5], @XMM[13] +#ifndef BSAES_ASM_EXTENDED_KEY + add r4, sp, #0x90 @ pass key schedule +#else + add r4, $key, #248 @ pass key schedule +#endif + veor @XMM[6], @XMM[6], @XMM[14] + mov r5, $rounds @ pass rounds + mov r0, sp + + bl _bsaes_decrypt8 + + vld1.64 {@XMM[ 8]-@XMM[ 9]}, [r0,:128]! + vld1.64 {@XMM[10]-@XMM[11]}, [r0,:128]! + veor @XMM[0], @XMM[0], @XMM[ 8] + vld1.64 {@XMM[12]-@XMM[13]}, [r0,:128]! + veor @XMM[1], @XMM[1], @XMM[ 9] + veor @XMM[8], @XMM[6], @XMM[10] + vst1.8 {@XMM[0]-@XMM[1]}, [$out]! + veor @XMM[9], @XMM[4], @XMM[11] + vld1.64 {@XMM[14]}, [r0,:128]! + veor @XMM[10], @XMM[2], @XMM[12] + vst1.8 {@XMM[8]-@XMM[9]}, [$out]! + veor @XMM[11], @XMM[7], @XMM[13] + veor @XMM[12], @XMM[3], @XMM[14] + vst1.8 {@XMM[10]-@XMM[11]}, [$out]! + vst1.8 {@XMM[12]}, [$out]! + + vld1.64 {@XMM[8]}, [r0,:128] @ next round tweak + b .Lxts_dec_done +.align 4 +.Lxts_dec_6: + vst1.64 {@XMM[14]}, [r0,:128] @ next round tweak + + veor @XMM[4], @XMM[4], @XMM[12] +#ifndef BSAES_ASM_EXTENDED_KEY + add r4, sp, #0x90 @ pass key schedule +#else + add r4, $key, #248 @ pass key schedule +#endif + veor @XMM[5], @XMM[5], @XMM[13] + mov r5, $rounds @ pass rounds + mov r0, sp + + bl _bsaes_decrypt8 + + vld1.64 {@XMM[ 8]-@XMM[ 9]}, [r0,:128]! + vld1.64 {@XMM[10]-@XMM[11]}, [r0,:128]! + veor @XMM[0], @XMM[0], @XMM[ 8] + vld1.64 {@XMM[12]-@XMM[13]}, [r0,:128]! + veor @XMM[1], @XMM[1], @XMM[ 9] + veor @XMM[8], @XMM[6], @XMM[10] + vst1.8 {@XMM[0]-@XMM[1]}, [$out]! + veor @XMM[9], @XMM[4], @XMM[11] + veor @XMM[10], @XMM[2], @XMM[12] + vst1.8 {@XMM[8]-@XMM[9]}, [$out]! + veor @XMM[11], @XMM[7], @XMM[13] + vst1.8 {@XMM[10]-@XMM[11]}, [$out]! + + vld1.64 {@XMM[8]}, [r0,:128] @ next round tweak + b .Lxts_dec_done +.align 4 +.Lxts_dec_5: + veor @XMM[3], @XMM[3], @XMM[11] +#ifndef BSAES_ASM_EXTENDED_KEY + add r4, sp, #0x90 @ pass key schedule +#else + add r4, $key, #248 @ pass key schedule +#endif + veor @XMM[4], @XMM[4], @XMM[12] + mov r5, $rounds @ pass rounds + mov r0, sp + + bl _bsaes_decrypt8 + + vld1.64 {@XMM[ 8]-@XMM[ 9]}, [r0,:128]! + vld1.64 {@XMM[10]-@XMM[11]}, [r0,:128]! + veor @XMM[0], @XMM[0], @XMM[ 8] + vld1.64 {@XMM[12]}, [r0,:128]! + veor @XMM[1], @XMM[1], @XMM[ 9] + veor @XMM[8], @XMM[6], @XMM[10] + vst1.8 {@XMM[0]-@XMM[1]}, [$out]! + veor @XMM[9], @XMM[4], @XMM[11] + veor @XMM[10], @XMM[2], @XMM[12] + vst1.8 {@XMM[8]-@XMM[9]}, [$out]! + vst1.8 {@XMM[10]}, [$out]! + + vld1.64 {@XMM[8]}, [r0,:128] @ next round tweak + b .Lxts_dec_done +.align 4 +.Lxts_dec_4: + veor @XMM[2], @XMM[2], @XMM[10] +#ifndef BSAES_ASM_EXTENDED_KEY + add r4, sp, #0x90 @ pass key schedule +#else + add r4, $key, #248 @ pass key schedule +#endif + veor @XMM[3], @XMM[3], @XMM[11] + mov r5, $rounds @ pass rounds + mov r0, sp + + bl _bsaes_decrypt8 + + vld1.64 {@XMM[ 8]-@XMM[ 9]}, [r0,:128]! + vld1.64 {@XMM[10]-@XMM[11]}, [r0,:128]! + veor @XMM[0], @XMM[0], @XMM[ 8] + veor @XMM[1], @XMM[1], @XMM[ 9] + veor @XMM[8], @XMM[6], @XMM[10] + vst1.8 {@XMM[0]-@XMM[1]}, [$out]! + veor @XMM[9], @XMM[4], @XMM[11] + vst1.8 {@XMM[8]-@XMM[9]}, [$out]! + + vld1.64 {@XMM[8]}, [r0,:128] @ next round tweak + b .Lxts_dec_done +.align 4 +.Lxts_dec_3: + veor @XMM[1], @XMM[1], @XMM[9] +#ifndef BSAES_ASM_EXTENDED_KEY + add r4, sp, #0x90 @ pass key schedule +#else + add r4, $key, #248 @ pass key schedule +#endif + veor @XMM[2], @XMM[2], @XMM[10] + mov r5, $rounds @ pass rounds + mov r0, sp + + bl _bsaes_decrypt8 + + vld1.64 {@XMM[8]-@XMM[9]}, [r0,:128]! + vld1.64 {@XMM[10]}, [r0,:128]! + veor @XMM[0], @XMM[0], @XMM[ 8] + veor @XMM[1], @XMM[1], @XMM[ 9] + veor @XMM[8], @XMM[6], @XMM[10] + vst1.8 {@XMM[0]-@XMM[1]}, [$out]! + vst1.8 {@XMM[8]}, [$out]! + + vld1.64 {@XMM[8]}, [r0,:128] @ next round tweak + b .Lxts_dec_done +.align 4 +.Lxts_dec_2: + veor @XMM[0], @XMM[0], @XMM[8] +#ifndef BSAES_ASM_EXTENDED_KEY + add r4, sp, #0x90 @ pass key schedule +#else + add r4, $key, #248 @ pass key schedule +#endif + veor @XMM[1], @XMM[1], @XMM[9] + mov r5, $rounds @ pass rounds + mov r0, sp + + bl _bsaes_decrypt8 + + vld1.64 {@XMM[8]-@XMM[9]}, [r0,:128]! + veor @XMM[0], @XMM[0], @XMM[ 8] + veor @XMM[1], @XMM[1], @XMM[ 9] + vst1.8 {@XMM[0]-@XMM[1]}, [$out]! + + vld1.64 {@XMM[8]}, [r0,:128] @ next round tweak + b .Lxts_dec_done +.align 4 +.Lxts_dec_1: + mov r0, sp + veor @XMM[0], @XMM[0], @XMM[8] + mov r1, sp + vst1.8 {@XMM[0]}, [sp,:128] + mov r5, $magic @ preserve magic + mov r2, $key + mov r4, $fp @ preserve fp + + bl AES_decrypt + + vld1.8 {@XMM[0]}, [sp,:128] + veor @XMM[0], @XMM[0], @XMM[8] + vst1.8 {@XMM[0]}, [$out]! + mov $fp, r4 + mov $magic, r5 + + vmov @XMM[8], @XMM[9] @ next round tweak + +.Lxts_dec_done: +#ifndef XTS_CHAIN_TWEAK + adds $len, #0x10 + beq .Lxts_dec_ret + + @ calculate one round of extra tweak for the stolen ciphertext + vldmia $magic, {$twmask} + vshr.s64 @XMM[6], @XMM[8], #63 + vand @XMM[6], @XMM[6], $twmask + vadd.u64 @XMM[9], @XMM[8], @XMM[8] + vswp `&Dhi("@XMM[6]")`,`&Dlo("@XMM[6]")` + veor @XMM[9], @XMM[9], @XMM[6] + + @ perform the final decryption with the last tweak value + vld1.8 {@XMM[0]}, [$inp]! + mov r0, sp + veor @XMM[0], @XMM[0], @XMM[9] + mov r1, sp + vst1.8 {@XMM[0]}, [sp,:128] + mov r2, $key + mov r4, $fp @ preserve fp + + bl AES_decrypt + + vld1.8 {@XMM[0]}, [sp,:128] + veor @XMM[0], @XMM[0], @XMM[9] + vst1.8 {@XMM[0]}, [$out] + + mov r6, $out +.Lxts_dec_steal: + ldrb r1, [$out] + ldrb r0, [$inp], #1 + strb r1, [$out, #0x10] + strb r0, [$out], #1 + + subs $len, #1 + bhi .Lxts_dec_steal + + vld1.8 {@XMM[0]}, [r6] + mov r0, sp + veor @XMM[0], @XMM[8] + mov r1, sp + vst1.8 {@XMM[0]}, [sp,:128] + mov r2, $key + + bl AES_decrypt + + vld1.8 {@XMM[0]}, [sp,:128] + veor @XMM[0], @XMM[0], @XMM[8] + vst1.8 {@XMM[0]}, [r6] + mov $fp, r4 +#endif + +.Lxts_dec_ret: + bic r0, $fp, #0xf + vmov.i32 q0, #0 + vmov.i32 q1, #0 +#ifdef XTS_CHAIN_TWEAK + ldr r1, [$fp, #0x20+VFP_ABI_FRAME] @ chain tweak +#endif +.Lxts_dec_bzero: @ wipe key schedule [if any] + vstmia sp!, {q0-q1} + cmp sp, r0 + bne .Lxts_dec_bzero + + mov sp, $fp +#ifdef XTS_CHAIN_TWEAK + vst1.8 {@XMM[8]}, [r1] +#endif + VFP_ABI_POP + ldmia sp!, {r4-r10, pc} @ return + +.size bsaes_xts_decrypt,.-bsaes_xts_decrypt +___ +} +$code.=<<___; +#endif +___ + +$code =~ s/\`([^\`]*)\`/eval($1)/gem; + +open SELF,$0; +while() { + next if (/^#!/); + last if (!s/^#/@/ and !/^$/); + print; +} +close SELF; + +print $code; + +close STDOUT; diff --git a/openssl-1.1.0h/crypto/aes/asm/bsaes-x86_64.pl b/openssl-1.1.0h/crypto/aes/asm/bsaes-x86_64.pl new file mode 100644 index 0000000..921d870 --- /dev/null +++ b/openssl-1.1.0h/crypto/aes/asm/bsaes-x86_64.pl @@ -0,0 +1,3111 @@ +#! /usr/bin/env perl +# Copyright 2011-2016 The OpenSSL Project Authors. All Rights Reserved. +# +# Licensed under the OpenSSL license (the "License"). You may not use +# this file except in compliance with the License. You can obtain a copy +# in the file LICENSE in the source distribution or at +# https://www.openssl.org/source/license.html + + +################################################################### +### AES-128 [originally in CTR mode] ### +### bitsliced implementation for Intel Core 2 processors ### +### requires support of SSE extensions up to SSSE3 ### +### Author: Emilia Käsper and Peter Schwabe ### +### Date: 2009-03-19 ### +### Public domain ### +### ### +### See http://homes.esat.kuleuven.be/~ekasper/#software for ### +### further information. ### +################################################################### +# +# September 2011. +# +# Started as transliteration to "perlasm" the original code has +# undergone following changes: +# +# - code was made position-independent; +# - rounds were folded into a loop resulting in >5x size reduction +# from 12.5KB to 2.2KB; +# - above was possibile thanks to mixcolumns() modification that +# allowed to feed its output back to aesenc[last], this was +# achieved at cost of two additional inter-registers moves; +# - some instruction reordering and interleaving; +# - this module doesn't implement key setup subroutine, instead it +# relies on conversion of "conventional" key schedule as returned +# by AES_set_encrypt_key (see discussion below); +# - first and last round keys are treated differently, which allowed +# to skip one shiftrows(), reduce bit-sliced key schedule and +# speed-up conversion by 22%; +# - support for 192- and 256-bit keys was added; +# +# Resulting performance in CPU cycles spent to encrypt one byte out +# of 4096-byte buffer with 128-bit key is: +# +# Emilia's this(*) difference +# +# Core 2 9.30 8.69 +7% +# Nehalem(**) 7.63 6.88 +11% +# Atom 17.1 16.4 +4% +# Silvermont - 12.9 +# Goldmont - 8.85 +# +# (*) Comparison is not completely fair, because "this" is ECB, +# i.e. no extra processing such as counter values calculation +# and xor-ing input as in Emilia's CTR implementation is +# performed. However, the CTR calculations stand for not more +# than 1% of total time, so comparison is *rather* fair. +# +# (**) Results were collected on Westmere, which is considered to +# be equivalent to Nehalem for this code. +# +# As for key schedule conversion subroutine. Interface to OpenSSL +# relies on per-invocation on-the-fly conversion. This naturally +# has impact on performance, especially for short inputs. Conversion +# time in CPU cycles and its ratio to CPU cycles spent in 8x block +# function is: +# +# conversion conversion/8x block +# Core 2 240 0.22 +# Nehalem 180 0.20 +# Atom 430 0.20 +# +# The ratio values mean that 128-byte blocks will be processed +# 16-18% slower, 256-byte blocks - 9-10%, 384-byte blocks - 6-7%, +# etc. Then keep in mind that input sizes not divisible by 128 are +# *effectively* slower, especially shortest ones, e.g. consecutive +# 144-byte blocks are processed 44% slower than one would expect, +# 272 - 29%, 400 - 22%, etc. Yet, despite all these "shortcomings" +# it's still faster than ["hyper-threading-safe" code path in] +# aes-x86_64.pl on all lengths above 64 bytes... +# +# October 2011. +# +# Add decryption procedure. Performance in CPU cycles spent to decrypt +# one byte out of 4096-byte buffer with 128-bit key is: +# +# Core 2 9.98 +# Nehalem 7.80 +# Atom 17.9 +# Silvermont 14.0 +# Goldmont 10.2 +# +# November 2011. +# +# Add bsaes_xts_[en|de]crypt. Less-than-80-bytes-block performance is +# suboptimal, but XTS is meant to be used with larger blocks... +# +# + +$flavour = shift; +$output = shift; +if ($flavour =~ /\./) { $output = $flavour; undef $flavour; } + +$win64=0; $win64=1 if ($flavour =~ /[nm]asm|mingw64/ || $output =~ /\.asm$/); + +$0 =~ m/(.*[\/\\])[^\/\\]+$/; $dir=$1; +( $xlate="${dir}x86_64-xlate.pl" and -f $xlate ) or +( $xlate="${dir}../../perlasm/x86_64-xlate.pl" and -f $xlate) or +die "can't locate x86_64-xlate.pl"; + +open OUT,"| \"$^X\" \"$xlate\" $flavour \"$output\""; +*STDOUT=*OUT; + +my ($inp,$out,$len,$key,$ivp)=("%rdi","%rsi","%rdx","%rcx"); +my @XMM=map("%xmm$_",(15,0..14)); # best on Atom, +10% over (0..15) +my $ecb=0; # suppress unreferenced ECB subroutines, spare some space... + +{ +my ($key,$rounds,$const)=("%rax","%r10d","%r11"); + +sub Sbox { +# input in lsb > [b0, b1, b2, b3, b4, b5, b6, b7] < msb +# output in lsb > [b0, b1, b4, b6, b3, b7, b2, b5] < msb +my @b=@_[0..7]; +my @t=@_[8..11]; +my @s=@_[12..15]; + &InBasisChange (@b); + &Inv_GF256 (@b[6,5,0,3,7,1,4,2],@t,@s); + &OutBasisChange (@b[7,1,4,2,6,5,0,3]); +} + +sub InBasisChange { +# input in lsb > [b0, b1, b2, b3, b4, b5, b6, b7] < msb +# output in lsb > [b6, b5, b0, b3, b7, b1, b4, b2] < msb +my @b=@_[0..7]; +$code.=<<___; + pxor @b[6], @b[5] + pxor @b[1], @b[2] + pxor @b[0], @b[3] + pxor @b[2], @b[6] + pxor @b[0], @b[5] + + pxor @b[3], @b[6] + pxor @b[7], @b[3] + pxor @b[5], @b[7] + pxor @b[4], @b[3] + pxor @b[5], @b[4] + pxor @b[1], @b[3] + + pxor @b[7], @b[2] + pxor @b[5], @b[1] +___ +} + +sub OutBasisChange { +# input in lsb > [b0, b1, b2, b3, b4, b5, b6, b7] < msb +# output in lsb > [b6, b1, b2, b4, b7, b0, b3, b5] < msb +my @b=@_[0..7]; +$code.=<<___; + pxor @b[6], @b[0] + pxor @b[4], @b[1] + pxor @b[0], @b[2] + pxor @b[6], @b[4] + pxor @b[1], @b[6] + + pxor @b[5], @b[1] + pxor @b[3], @b[5] + pxor @b[7], @b[3] + pxor @b[5], @b[7] + pxor @b[5], @b[2] + + pxor @b[7], @b[4] +___ +} + +sub InvSbox { +# input in lsb > [b0, b1, b2, b3, b4, b5, b6, b7] < msb +# output in lsb > [b0, b1, b6, b4, b2, b7, b3, b5] < msb +my @b=@_[0..7]; +my @t=@_[8..11]; +my @s=@_[12..15]; + &InvInBasisChange (@b); + &Inv_GF256 (@b[5,1,2,6,3,7,0,4],@t,@s); + &InvOutBasisChange (@b[3,7,0,4,5,1,2,6]); +} + +sub InvInBasisChange { # OutBasisChange in reverse +my @b=@_[5,1,2,6,3,7,0,4]; +$code.=<<___ + pxor @b[7], @b[4] + + pxor @b[5], @b[7] + pxor @b[5], @b[2] + pxor @b[7], @b[3] + pxor @b[3], @b[5] + pxor @b[5], @b[1] + + pxor @b[1], @b[6] + pxor @b[0], @b[2] + pxor @b[6], @b[4] + pxor @b[6], @b[0] + pxor @b[4], @b[1] +___ +} + +sub InvOutBasisChange { # InBasisChange in reverse +my @b=@_[2,5,7,3,6,1,0,4]; +$code.=<<___; + pxor @b[5], @b[1] + pxor @b[7], @b[2] + + pxor @b[1], @b[3] + pxor @b[5], @b[4] + pxor @b[5], @b[7] + pxor @b[4], @b[3] + pxor @b[0], @b[5] + pxor @b[7], @b[3] + pxor @b[2], @b[6] + pxor @b[1], @b[2] + pxor @b[3], @b[6] + + pxor @b[0], @b[3] + pxor @b[6], @b[5] +___ +} + +sub Mul_GF4 { +#;************************************************************* +#;* Mul_GF4: Input x0-x1,y0-y1 Output x0-x1 Temp t0 (8) * +#;************************************************************* +my ($x0,$x1,$y0,$y1,$t0)=@_; +$code.=<<___; + movdqa $y0, $t0 + pxor $y1, $t0 + pand $x0, $t0 + pxor $x1, $x0 + pand $y0, $x1 + pand $y1, $x0 + pxor $x1, $x0 + pxor $t0, $x1 +___ +} + +sub Mul_GF4_N { # not used, see next subroutine +# multiply and scale by N +my ($x0,$x1,$y0,$y1,$t0)=@_; +$code.=<<___; + movdqa $y0, $t0 + pxor $y1, $t0 + pand $x0, $t0 + pxor $x1, $x0 + pand $y0, $x1 + pand $y1, $x0 + pxor $x0, $x1 + pxor $t0, $x0 +___ +} + +sub Mul_GF4_N_GF4 { +# interleaved Mul_GF4_N and Mul_GF4 +my ($x0,$x1,$y0,$y1,$t0, + $x2,$x3,$y2,$y3,$t1)=@_; +$code.=<<___; + movdqa $y0, $t0 + movdqa $y2, $t1 + pxor $y1, $t0 + pxor $y3, $t1 + pand $x0, $t0 + pand $x2, $t1 + pxor $x1, $x0 + pxor $x3, $x2 + pand $y0, $x1 + pand $y2, $x3 + pand $y1, $x0 + pand $y3, $x2 + pxor $x0, $x1 + pxor $x3, $x2 + pxor $t0, $x0 + pxor $t1, $x3 +___ +} +sub Mul_GF16_2 { +my @x=@_[0..7]; +my @y=@_[8..11]; +my @t=@_[12..15]; +$code.=<<___; + movdqa @x[0], @t[0] + movdqa @x[1], @t[1] +___ + &Mul_GF4 (@x[0], @x[1], @y[0], @y[1], @t[2]); +$code.=<<___; + pxor @x[2], @t[0] + pxor @x[3], @t[1] + pxor @y[2], @y[0] + pxor @y[3], @y[1] +___ + Mul_GF4_N_GF4 (@t[0], @t[1], @y[0], @y[1], @t[3], + @x[2], @x[3], @y[2], @y[3], @t[2]); +$code.=<<___; + pxor @t[0], @x[0] + pxor @t[0], @x[2] + pxor @t[1], @x[1] + pxor @t[1], @x[3] + + movdqa @x[4], @t[0] + movdqa @x[5], @t[1] + pxor @x[6], @t[0] + pxor @x[7], @t[1] +___ + &Mul_GF4_N_GF4 (@t[0], @t[1], @y[0], @y[1], @t[3], + @x[6], @x[7], @y[2], @y[3], @t[2]); +$code.=<<___; + pxor @y[2], @y[0] + pxor @y[3], @y[1] +___ + &Mul_GF4 (@x[4], @x[5], @y[0], @y[1], @t[3]); +$code.=<<___; + pxor @t[0], @x[4] + pxor @t[0], @x[6] + pxor @t[1], @x[5] + pxor @t[1], @x[7] +___ +} +sub Inv_GF256 { +#;******************************************************************** +#;* Inv_GF256: Input x0-x7 Output x0-x7 Temp t0-t3,s0-s3 (144) * +#;******************************************************************** +my @x=@_[0..7]; +my @t=@_[8..11]; +my @s=@_[12..15]; +# direct optimizations from hardware +$code.=<<___; + movdqa @x[4], @t[3] + movdqa @x[5], @t[2] + movdqa @x[1], @t[1] + movdqa @x[7], @s[1] + movdqa @x[0], @s[0] + + pxor @x[6], @t[3] + pxor @x[7], @t[2] + pxor @x[3], @t[1] + movdqa @t[3], @s[2] + pxor @x[6], @s[1] + movdqa @t[2], @t[0] + pxor @x[2], @s[0] + movdqa @t[3], @s[3] + + por @t[1], @t[2] + por @s[0], @t[3] + pxor @t[0], @s[3] + pand @s[0], @s[2] + pxor @t[1], @s[0] + pand @t[1], @t[0] + pand @s[0], @s[3] + movdqa @x[3], @s[0] + pxor @x[2], @s[0] + pand @s[0], @s[1] + pxor @s[1], @t[3] + pxor @s[1], @t[2] + movdqa @x[4], @s[1] + movdqa @x[1], @s[0] + pxor @x[5], @s[1] + pxor @x[0], @s[0] + movdqa @s[1], @t[1] + pand @s[0], @s[1] + por @s[0], @t[1] + pxor @s[1], @t[0] + pxor @s[3], @t[3] + pxor @s[2], @t[2] + pxor @s[3], @t[1] + movdqa @x[7], @s[0] + pxor @s[2], @t[0] + movdqa @x[6], @s[1] + pxor @s[2], @t[1] + movdqa @x[5], @s[2] + pand @x[3], @s[0] + movdqa @x[4], @s[3] + pand @x[2], @s[1] + pand @x[1], @s[2] + por @x[0], @s[3] + pxor @s[0], @t[3] + pxor @s[1], @t[2] + pxor @s[2], @t[1] + pxor @s[3], @t[0] + + #Inv_GF16 \t0, \t1, \t2, \t3, \s0, \s1, \s2, \s3 + + # new smaller inversion + + movdqa @t[3], @s[0] + pand @t[1], @t[3] + pxor @t[2], @s[0] + + movdqa @t[0], @s[2] + movdqa @s[0], @s[3] + pxor @t[3], @s[2] + pand @s[2], @s[3] + + movdqa @t[1], @s[1] + pxor @t[2], @s[3] + pxor @t[0], @s[1] + + pxor @t[2], @t[3] + + pand @t[3], @s[1] + + movdqa @s[2], @t[2] + pxor @t[0], @s[1] + + pxor @s[1], @t[2] + pxor @s[1], @t[1] + + pand @t[0], @t[2] + + pxor @t[2], @s[2] + pxor @t[2], @t[1] + + pand @s[3], @s[2] + + pxor @s[0], @s[2] +___ +# output in s3, s2, s1, t1 + +# Mul_GF16_2 \x0, \x1, \x2, \x3, \x4, \x5, \x6, \x7, \t2, \t3, \t0, \t1, \s0, \s1, \s2, \s3 + +# Mul_GF16_2 \x0, \x1, \x2, \x3, \x4, \x5, \x6, \x7, \s3, \s2, \s1, \t1, \s0, \t0, \t2, \t3 + &Mul_GF16_2(@x,@s[3,2,1],@t[1],@s[0],@t[0,2,3]); + +### output msb > [x3,x2,x1,x0,x7,x6,x5,x4] < lsb +} + +# AES linear components + +sub ShiftRows { +my @x=@_[0..7]; +my $mask=pop; +$code.=<<___; + pxor 0x00($key),@x[0] + pxor 0x10($key),@x[1] + pxor 0x20($key),@x[2] + pxor 0x30($key),@x[3] + pshufb $mask,@x[0] + pshufb $mask,@x[1] + pxor 0x40($key),@x[4] + pxor 0x50($key),@x[5] + pshufb $mask,@x[2] + pshufb $mask,@x[3] + pxor 0x60($key),@x[6] + pxor 0x70($key),@x[7] + pshufb $mask,@x[4] + pshufb $mask,@x[5] + pshufb $mask,@x[6] + pshufb $mask,@x[7] + lea 0x80($key),$key +___ +} + +sub MixColumns { +# modified to emit output in order suitable for feeding back to aesenc[last] +my @x=@_[0..7]; +my @t=@_[8..15]; +my $inv=@_[16]; # optional +$code.=<<___; + pshufd \$0x93, @x[0], @t[0] # x0 <<< 32 + pshufd \$0x93, @x[1], @t[1] + pxor @t[0], @x[0] # x0 ^ (x0 <<< 32) + pshufd \$0x93, @x[2], @t[2] + pxor @t[1], @x[1] + pshufd \$0x93, @x[3], @t[3] + pxor @t[2], @x[2] + pshufd \$0x93, @x[4], @t[4] + pxor @t[3], @x[3] + pshufd \$0x93, @x[5], @t[5] + pxor @t[4], @x[4] + pshufd \$0x93, @x[6], @t[6] + pxor @t[5], @x[5] + pshufd \$0x93, @x[7], @t[7] + pxor @t[6], @x[6] + pxor @t[7], @x[7] + + pxor @x[0], @t[1] + pxor @x[7], @t[0] + pxor @x[7], @t[1] + pshufd \$0x4E, @x[0], @x[0] # (x0 ^ (x0 <<< 32)) <<< 64) + pxor @x[1], @t[2] + pshufd \$0x4E, @x[1], @x[1] + pxor @x[4], @t[5] + pxor @t[0], @x[0] + pxor @x[5], @t[6] + pxor @t[1], @x[1] + pxor @x[3], @t[4] + pshufd \$0x4E, @x[4], @t[0] + pxor @x[6], @t[7] + pshufd \$0x4E, @x[5], @t[1] + pxor @x[2], @t[3] + pshufd \$0x4E, @x[3], @x[4] + pxor @x[7], @t[3] + pshufd \$0x4E, @x[7], @x[5] + pxor @x[7], @t[4] + pshufd \$0x4E, @x[6], @x[3] + pxor @t[4], @t[0] + pshufd \$0x4E, @x[2], @x[6] + pxor @t[5], @t[1] +___ +$code.=<<___ if (!$inv); + pxor @t[3], @x[4] + pxor @t[7], @x[5] + pxor @t[6], @x[3] + movdqa @t[0], @x[2] + pxor @t[2], @x[6] + movdqa @t[1], @x[7] +___ +$code.=<<___ if ($inv); + pxor @x[4], @t[3] + pxor @t[7], @x[5] + pxor @x[3], @t[6] + movdqa @t[0], @x[3] + pxor @t[2], @x[6] + movdqa @t[6], @x[2] + movdqa @t[1], @x[7] + movdqa @x[6], @x[4] + movdqa @t[3], @x[6] +___ +} + +sub InvMixColumns_orig { +my @x=@_[0..7]; +my @t=@_[8..15]; + +$code.=<<___; + # multiplication by 0x0e + pshufd \$0x93, @x[7], @t[7] + movdqa @x[2], @t[2] + pxor @x[5], @x[7] # 7 5 + pxor @x[5], @x[2] # 2 5 + pshufd \$0x93, @x[0], @t[0] + movdqa @x[5], @t[5] + pxor @x[0], @x[5] # 5 0 [1] + pxor @x[1], @x[0] # 0 1 + pshufd \$0x93, @x[1], @t[1] + pxor @x[2], @x[1] # 1 25 + pxor @x[6], @x[0] # 01 6 [2] + pxor @x[3], @x[1] # 125 3 [4] + pshufd \$0x93, @x[3], @t[3] + pxor @x[0], @x[2] # 25 016 [3] + pxor @x[7], @x[3] # 3 75 + pxor @x[6], @x[7] # 75 6 [0] + pshufd \$0x93, @x[6], @t[6] + movdqa @x[4], @t[4] + pxor @x[4], @x[6] # 6 4 + pxor @x[3], @x[4] # 4 375 [6] + pxor @x[7], @x[3] # 375 756=36 + pxor @t[5], @x[6] # 64 5 [7] + pxor @t[2], @x[3] # 36 2 + pxor @t[4], @x[3] # 362 4 [5] + pshufd \$0x93, @t[5], @t[5] +___ + my @y = @x[7,5,0,2,1,3,4,6]; +$code.=<<___; + # multiplication by 0x0b + pxor @y[0], @y[1] + pxor @t[0], @y[0] + pxor @t[1], @y[1] + pshufd \$0x93, @t[2], @t[2] + pxor @t[5], @y[0] + pxor @t[6], @y[1] + pxor @t[7], @y[0] + pshufd \$0x93, @t[4], @t[4] + pxor @t[6], @t[7] # clobber t[7] + pxor @y[0], @y[1] + + pxor @t[0], @y[3] + pshufd \$0x93, @t[0], @t[0] + pxor @t[1], @y[2] + pxor @t[1], @y[4] + pxor @t[2], @y[2] + pshufd \$0x93, @t[1], @t[1] + pxor @t[2], @y[3] + pxor @t[2], @y[5] + pxor @t[7], @y[2] + pshufd \$0x93, @t[2], @t[2] + pxor @t[3], @y[3] + pxor @t[3], @y[6] + pxor @t[3], @y[4] + pshufd \$0x93, @t[3], @t[3] + pxor @t[4], @y[7] + pxor @t[4], @y[5] + pxor @t[7], @y[7] + pxor @t[5], @y[3] + pxor @t[4], @y[4] + pxor @t[5], @t[7] # clobber t[7] even more + + pxor @t[7], @y[5] + pshufd \$0x93, @t[4], @t[4] + pxor @t[7], @y[6] + pxor @t[7], @y[4] + + pxor @t[5], @t[7] + pshufd \$0x93, @t[5], @t[5] + pxor @t[6], @t[7] # restore t[7] + + # multiplication by 0x0d + pxor @y[7], @y[4] + pxor @t[4], @y[7] + pshufd \$0x93, @t[6], @t[6] + pxor @t[0], @y[2] + pxor @t[5], @y[7] + pxor @t[2], @y[2] + pshufd \$0x93, @t[7], @t[7] + + pxor @y[1], @y[3] + pxor @t[1], @y[1] + pxor @t[0], @y[0] + pxor @t[0], @y[3] + pxor @t[5], @y[1] + pxor @t[5], @y[0] + pxor @t[7], @y[1] + pshufd \$0x93, @t[0], @t[0] + pxor @t[6], @y[0] + pxor @y[1], @y[3] + pxor @t[1], @y[4] + pshufd \$0x93, @t[1], @t[1] + + pxor @t[7], @y[7] + pxor @t[2], @y[4] + pxor @t[2], @y[5] + pshufd \$0x93, @t[2], @t[2] + pxor @t[6], @y[2] + pxor @t[3], @t[6] # clobber t[6] + pxor @y[7], @y[4] + pxor @t[6], @y[3] + + pxor @t[6], @y[6] + pxor @t[5], @y[5] + pxor @t[4], @y[6] + pshufd \$0x93, @t[4], @t[4] + pxor @t[6], @y[5] + pxor @t[7], @y[6] + pxor @t[3], @t[6] # restore t[6] + + pshufd \$0x93, @t[5], @t[5] + pshufd \$0x93, @t[6], @t[6] + pshufd \$0x93, @t[7], @t[7] + pshufd \$0x93, @t[3], @t[3] + + # multiplication by 0x09 + pxor @y[1], @y[4] + pxor @y[1], @t[1] # t[1]=y[1] + pxor @t[5], @t[0] # clobber t[0] + pxor @t[5], @t[1] + pxor @t[0], @y[3] + pxor @y[0], @t[0] # t[0]=y[0] + pxor @t[6], @t[1] + pxor @t[7], @t[6] # clobber t[6] + pxor @t[1], @y[4] + pxor @t[4], @y[7] + pxor @y[4], @t[4] # t[4]=y[4] + pxor @t[3], @y[6] + pxor @y[3], @t[3] # t[3]=y[3] + pxor @t[2], @y[5] + pxor @y[2], @t[2] # t[2]=y[2] + pxor @t[7], @t[3] + pxor @y[5], @t[5] # t[5]=y[5] + pxor @t[6], @t[2] + pxor @t[6], @t[5] + pxor @y[6], @t[6] # t[6]=y[6] + pxor @y[7], @t[7] # t[7]=y[7] + + movdqa @t[0],@XMM[0] + movdqa @t[1],@XMM[1] + movdqa @t[2],@XMM[2] + movdqa @t[3],@XMM[3] + movdqa @t[4],@XMM[4] + movdqa @t[5],@XMM[5] + movdqa @t[6],@XMM[6] + movdqa @t[7],@XMM[7] +___ +} + +sub InvMixColumns { +my @x=@_[0..7]; +my @t=@_[8..15]; + +# Thanks to Jussi Kivilinna for providing pointer to +# +# | 0e 0b 0d 09 | | 02 03 01 01 | | 05 00 04 00 | +# | 09 0e 0b 0d | = | 01 02 03 01 | x | 00 05 00 04 | +# | 0d 09 0e 0b | | 01 01 02 03 | | 04 00 05 00 | +# | 0b 0d 09 0e | | 03 01 01 02 | | 00 04 00 05 | + +$code.=<<___; + # multiplication by 0x05-0x00-0x04-0x00 + pshufd \$0x4E, @x[0], @t[0] + pshufd \$0x4E, @x[6], @t[6] + pxor @x[0], @t[0] + pshufd \$0x4E, @x[7], @t[7] + pxor @x[6], @t[6] + pshufd \$0x4E, @x[1], @t[1] + pxor @x[7], @t[7] + pshufd \$0x4E, @x[2], @t[2] + pxor @x[1], @t[1] + pshufd \$0x4E, @x[3], @t[3] + pxor @x[2], @t[2] + pxor @t[6], @x[0] + pxor @t[6], @x[1] + pshufd \$0x4E, @x[4], @t[4] + pxor @x[3], @t[3] + pxor @t[0], @x[2] + pxor @t[1], @x[3] + pshufd \$0x4E, @x[5], @t[5] + pxor @x[4], @t[4] + pxor @t[7], @x[1] + pxor @t[2], @x[4] + pxor @x[5], @t[5] + + pxor @t[7], @x[2] + pxor @t[6], @x[3] + pxor @t[6], @x[4] + pxor @t[3], @x[5] + pxor @t[4], @x[6] + pxor @t[7], @x[4] + pxor @t[7], @x[5] + pxor @t[5], @x[7] +___ + &MixColumns (@x,@t,1); # flipped 2<->3 and 4<->6 +} + +sub aesenc { # not used +my @b=@_[0..7]; +my @t=@_[8..15]; +$code.=<<___; + movdqa 0x30($const),@t[0] # .LSR +___ + &ShiftRows (@b,@t[0]); + &Sbox (@b,@t); + &MixColumns (@b[0,1,4,6,3,7,2,5],@t); +} + +sub aesenclast { # not used +my @b=@_[0..7]; +my @t=@_[8..15]; +$code.=<<___; + movdqa 0x40($const),@t[0] # .LSRM0 +___ + &ShiftRows (@b,@t[0]); + &Sbox (@b,@t); +$code.=<<___ + pxor 0x00($key),@b[0] + pxor 0x10($key),@b[1] + pxor 0x20($key),@b[4] + pxor 0x30($key),@b[6] + pxor 0x40($key),@b[3] + pxor 0x50($key),@b[7] + pxor 0x60($key),@b[2] + pxor 0x70($key),@b[5] +___ +} + +sub swapmove { +my ($a,$b,$n,$mask,$t)=@_; +$code.=<<___; + movdqa $b,$t + psrlq \$$n,$b + pxor $a,$b + pand $mask,$b + pxor $b,$a + psllq \$$n,$b + pxor $t,$b +___ +} +sub swapmove2x { +my ($a0,$b0,$a1,$b1,$n,$mask,$t0,$t1)=@_; +$code.=<<___; + movdqa $b0,$t0 + psrlq \$$n,$b0 + movdqa $b1,$t1 + psrlq \$$n,$b1 + pxor $a0,$b0 + pxor $a1,$b1 + pand $mask,$b0 + pand $mask,$b1 + pxor $b0,$a0 + psllq \$$n,$b0 + pxor $b1,$a1 + psllq \$$n,$b1 + pxor $t0,$b0 + pxor $t1,$b1 +___ +} + +sub bitslice { +my @x=reverse(@_[0..7]); +my ($t0,$t1,$t2,$t3)=@_[8..11]; +$code.=<<___; + movdqa 0x00($const),$t0 # .LBS0 + movdqa 0x10($const),$t1 # .LBS1 +___ + &swapmove2x(@x[0,1,2,3],1,$t0,$t2,$t3); + &swapmove2x(@x[4,5,6,7],1,$t0,$t2,$t3); +$code.=<<___; + movdqa 0x20($const),$t0 # .LBS2 +___ + &swapmove2x(@x[0,2,1,3],2,$t1,$t2,$t3); + &swapmove2x(@x[4,6,5,7],2,$t1,$t2,$t3); + + &swapmove2x(@x[0,4,1,5],4,$t0,$t2,$t3); + &swapmove2x(@x[2,6,3,7],4,$t0,$t2,$t3); +} + +$code.=<<___; +.text + +.extern asm_AES_encrypt +.extern asm_AES_decrypt + +.type _bsaes_encrypt8,\@abi-omnipotent +.align 64 +_bsaes_encrypt8: + lea .LBS0(%rip), $const # constants table + + movdqa ($key), @XMM[9] # round 0 key + lea 0x10($key), $key + movdqa 0x50($const), @XMM[8] # .LM0SR + pxor @XMM[9], @XMM[0] # xor with round0 key + pxor @XMM[9], @XMM[1] + pxor @XMM[9], @XMM[2] + pxor @XMM[9], @XMM[3] + pshufb @XMM[8], @XMM[0] + pshufb @XMM[8], @XMM[1] + pxor @XMM[9], @XMM[4] + pxor @XMM[9], @XMM[5] + pshufb @XMM[8], @XMM[2] + pshufb @XMM[8], @XMM[3] + pxor @XMM[9], @XMM[6] + pxor @XMM[9], @XMM[7] + pshufb @XMM[8], @XMM[4] + pshufb @XMM[8], @XMM[5] + pshufb @XMM[8], @XMM[6] + pshufb @XMM[8], @XMM[7] +_bsaes_encrypt8_bitslice: +___ + &bitslice (@XMM[0..7, 8..11]); +$code.=<<___; + dec $rounds + jmp .Lenc_sbox +.align 16 +.Lenc_loop: +___ + &ShiftRows (@XMM[0..7, 8]); +$code.=".Lenc_sbox:\n"; + &Sbox (@XMM[0..7, 8..15]); +$code.=<<___; + dec $rounds + jl .Lenc_done +___ + &MixColumns (@XMM[0,1,4,6,3,7,2,5, 8..15]); +$code.=<<___; + movdqa 0x30($const), @XMM[8] # .LSR + jnz .Lenc_loop + movdqa 0x40($const), @XMM[8] # .LSRM0 + jmp .Lenc_loop +.align 16 +.Lenc_done: +___ + # output in lsb > [t0, t1, t4, t6, t3, t7, t2, t5] < msb + &bitslice (@XMM[0,1,4,6,3,7,2,5, 8..11]); +$code.=<<___; + movdqa ($key), @XMM[8] # last round key + pxor @XMM[8], @XMM[4] + pxor @XMM[8], @XMM[6] + pxor @XMM[8], @XMM[3] + pxor @XMM[8], @XMM[7] + pxor @XMM[8], @XMM[2] + pxor @XMM[8], @XMM[5] + pxor @XMM[8], @XMM[0] + pxor @XMM[8], @XMM[1] + ret +.size _bsaes_encrypt8,.-_bsaes_encrypt8 + +.type _bsaes_decrypt8,\@abi-omnipotent +.align 64 +_bsaes_decrypt8: + lea .LBS0(%rip), $const # constants table + + movdqa ($key), @XMM[9] # round 0 key + lea 0x10($key), $key + movdqa -0x30($const), @XMM[8] # .LM0ISR + pxor @XMM[9], @XMM[0] # xor with round0 key + pxor @XMM[9], @XMM[1] + pxor @XMM[9], @XMM[2] + pxor @XMM[9], @XMM[3] + pshufb @XMM[8], @XMM[0] + pshufb @XMM[8], @XMM[1] + pxor @XMM[9], @XMM[4] + pxor @XMM[9], @XMM[5] + pshufb @XMM[8], @XMM[2] + pshufb @XMM[8], @XMM[3] + pxor @XMM[9], @XMM[6] + pxor @XMM[9], @XMM[7] + pshufb @XMM[8], @XMM[4] + pshufb @XMM[8], @XMM[5] + pshufb @XMM[8], @XMM[6] + pshufb @XMM[8], @XMM[7] +___ + &bitslice (@XMM[0..7, 8..11]); +$code.=<<___; + dec $rounds + jmp .Ldec_sbox +.align 16 +.Ldec_loop: +___ + &ShiftRows (@XMM[0..7, 8]); +$code.=".Ldec_sbox:\n"; + &InvSbox (@XMM[0..7, 8..15]); +$code.=<<___; + dec $rounds + jl .Ldec_done +___ + &InvMixColumns (@XMM[0,1,6,4,2,7,3,5, 8..15]); +$code.=<<___; + movdqa -0x10($const), @XMM[8] # .LISR + jnz .Ldec_loop + movdqa -0x20($const), @XMM[8] # .LISRM0 + jmp .Ldec_loop +.align 16 +.Ldec_done: +___ + &bitslice (@XMM[0,1,6,4,2,7,3,5, 8..11]); +$code.=<<___; + movdqa ($key), @XMM[8] # last round key + pxor @XMM[8], @XMM[6] + pxor @XMM[8], @XMM[4] + pxor @XMM[8], @XMM[2] + pxor @XMM[8], @XMM[7] + pxor @XMM[8], @XMM[3] + pxor @XMM[8], @XMM[5] + pxor @XMM[8], @XMM[0] + pxor @XMM[8], @XMM[1] + ret +.size _bsaes_decrypt8,.-_bsaes_decrypt8 +___ +} +{ +my ($out,$inp,$rounds,$const)=("%rax","%rcx","%r10d","%r11"); + +sub bitslice_key { +my @x=reverse(@_[0..7]); +my ($bs0,$bs1,$bs2,$t2,$t3)=@_[8..12]; + + &swapmove (@x[0,1],1,$bs0,$t2,$t3); +$code.=<<___; + #&swapmove(@x[2,3],1,$t0,$t2,$t3); + movdqa @x[0], @x[2] + movdqa @x[1], @x[3] +___ + #&swapmove2x(@x[4,5,6,7],1,$t0,$t2,$t3); + + &swapmove2x (@x[0,2,1,3],2,$bs1,$t2,$t3); +$code.=<<___; + #&swapmove2x(@x[4,6,5,7],2,$t1,$t2,$t3); + movdqa @x[0], @x[4] + movdqa @x[2], @x[6] + movdqa @x[1], @x[5] + movdqa @x[3], @x[7] +___ + &swapmove2x (@x[0,4,1,5],4,$bs2,$t2,$t3); + &swapmove2x (@x[2,6,3,7],4,$bs2,$t2,$t3); +} + +$code.=<<___; +.type _bsaes_key_convert,\@abi-omnipotent +.align 16 +_bsaes_key_convert: + lea .Lmasks(%rip), $const + movdqu ($inp), %xmm7 # load round 0 key + lea 0x10($inp), $inp + movdqa 0x00($const), %xmm0 # 0x01... + movdqa 0x10($const), %xmm1 # 0x02... + movdqa 0x20($const), %xmm2 # 0x04... + movdqa 0x30($const), %xmm3 # 0x08... + movdqa 0x40($const), %xmm4 # .LM0 + pcmpeqd %xmm5, %xmm5 # .LNOT + + movdqu ($inp), %xmm6 # load round 1 key + movdqa %xmm7, ($out) # save round 0 key + lea 0x10($out), $out + dec $rounds + jmp .Lkey_loop +.align 16 +.Lkey_loop: + pshufb %xmm4, %xmm6 # .LM0 + + movdqa %xmm0, %xmm8 + movdqa %xmm1, %xmm9 + + pand %xmm6, %xmm8 + pand %xmm6, %xmm9 + movdqa %xmm2, %xmm10 + pcmpeqb %xmm0, %xmm8 + psllq \$4, %xmm0 # 0x10... + movdqa %xmm3, %xmm11 + pcmpeqb %xmm1, %xmm9 + psllq \$4, %xmm1 # 0x20... + + pand %xmm6, %xmm10 + pand %xmm6, %xmm11 + movdqa %xmm0, %xmm12 + pcmpeqb %xmm2, %xmm10 + psllq \$4, %xmm2 # 0x40... + movdqa %xmm1, %xmm13 + pcmpeqb %xmm3, %xmm11 + psllq \$4, %xmm3 # 0x80... + + movdqa %xmm2, %xmm14 + movdqa %xmm3, %xmm15 + pxor %xmm5, %xmm8 # "pnot" + pxor %xmm5, %xmm9 + + pand %xmm6, %xmm12 + pand %xmm6, %xmm13 + movdqa %xmm8, 0x00($out) # write bit-sliced round key + pcmpeqb %xmm0, %xmm12 + psrlq \$4, %xmm0 # 0x01... + movdqa %xmm9, 0x10($out) + pcmpeqb %xmm1, %xmm13 + psrlq \$4, %xmm1 # 0x02... + lea 0x10($inp), $inp + + pand %xmm6, %xmm14 + pand %xmm6, %xmm15 + movdqa %xmm10, 0x20($out) + pcmpeqb %xmm2, %xmm14 + psrlq \$4, %xmm2 # 0x04... + movdqa %xmm11, 0x30($out) + pcmpeqb %xmm3, %xmm15 + psrlq \$4, %xmm3 # 0x08... + movdqu ($inp), %xmm6 # load next round key + + pxor %xmm5, %xmm13 # "pnot" + pxor %xmm5, %xmm14 + movdqa %xmm12, 0x40($out) + movdqa %xmm13, 0x50($out) + movdqa %xmm14, 0x60($out) + movdqa %xmm15, 0x70($out) + lea 0x80($out),$out + dec $rounds + jnz .Lkey_loop + + movdqa 0x50($const), %xmm7 # .L63 + #movdqa %xmm6, ($out) # don't save last round key + ret +.size _bsaes_key_convert,.-_bsaes_key_convert +___ +} + +if (0 && !$win64) { # following four functions are unsupported interface + # used for benchmarking... +$code.=<<___; +.globl bsaes_enc_key_convert +.type bsaes_enc_key_convert,\@function,2 +.align 16 +bsaes_enc_key_convert: + mov 240($inp),%r10d # pass rounds + mov $inp,%rcx # pass key + mov $out,%rax # pass key schedule + call _bsaes_key_convert + pxor %xmm6,%xmm7 # fix up last round key + movdqa %xmm7,(%rax) # save last round key + ret +.size bsaes_enc_key_convert,.-bsaes_enc_key_convert + +.globl bsaes_encrypt_128 +.type bsaes_encrypt_128,\@function,4 +.align 16 +bsaes_encrypt_128: +.Lenc128_loop: + movdqu 0x00($inp), @XMM[0] # load input + movdqu 0x10($inp), @XMM[1] + movdqu 0x20($inp), @XMM[2] + movdqu 0x30($inp), @XMM[3] + movdqu 0x40($inp), @XMM[4] + movdqu 0x50($inp), @XMM[5] + movdqu 0x60($inp), @XMM[6] + movdqu 0x70($inp), @XMM[7] + mov $key, %rax # pass the $key + lea 0x80($inp), $inp + mov \$10,%r10d + + call _bsaes_encrypt8 + + movdqu @XMM[0], 0x00($out) # write output + movdqu @XMM[1], 0x10($out) + movdqu @XMM[4], 0x20($out) + movdqu @XMM[6], 0x30($out) + movdqu @XMM[3], 0x40($out) + movdqu @XMM[7], 0x50($out) + movdqu @XMM[2], 0x60($out) + movdqu @XMM[5], 0x70($out) + lea 0x80($out), $out + sub \$0x80,$len + ja .Lenc128_loop + ret +.size bsaes_encrypt_128,.-bsaes_encrypt_128 + +.globl bsaes_dec_key_convert +.type bsaes_dec_key_convert,\@function,2 +.align 16 +bsaes_dec_key_convert: + mov 240($inp),%r10d # pass rounds + mov $inp,%rcx # pass key + mov $out,%rax # pass key schedule + call _bsaes_key_convert + pxor ($out),%xmm7 # fix up round 0 key + movdqa %xmm6,(%rax) # save last round key + movdqa %xmm7,($out) + ret +.size bsaes_dec_key_convert,.-bsaes_dec_key_convert + +.globl bsaes_decrypt_128 +.type bsaes_decrypt_128,\@function,4 +.align 16 +bsaes_decrypt_128: +.Ldec128_loop: + movdqu 0x00($inp), @XMM[0] # load input + movdqu 0x10($inp), @XMM[1] + movdqu 0x20($inp), @XMM[2] + movdqu 0x30($inp), @XMM[3] + movdqu 0x40($inp), @XMM[4] + movdqu 0x50($inp), @XMM[5] + movdqu 0x60($inp), @XMM[6] + movdqu 0x70($inp), @XMM[7] + mov $key, %rax # pass the $key + lea 0x80($inp), $inp + mov \$10,%r10d + + call _bsaes_decrypt8 + + movdqu @XMM[0], 0x00($out) # write output + movdqu @XMM[1], 0x10($out) + movdqu @XMM[6], 0x20($out) + movdqu @XMM[4], 0x30($out) + movdqu @XMM[2], 0x40($out) + movdqu @XMM[7], 0x50($out) + movdqu @XMM[3], 0x60($out) + movdqu @XMM[5], 0x70($out) + lea 0x80($out), $out + sub \$0x80,$len + ja .Ldec128_loop + ret +.size bsaes_decrypt_128,.-bsaes_decrypt_128 +___ +} +{ +###################################################################### +# +# OpenSSL interface +# +my ($arg1,$arg2,$arg3,$arg4,$arg5,$arg6)=$win64 ? ("%rcx","%rdx","%r8","%r9","%r10","%r11d") + : ("%rdi","%rsi","%rdx","%rcx","%r8","%r9d"); +my ($inp,$out,$len,$key)=("%r12","%r13","%r14","%r15"); + +if ($ecb) { +$code.=<<___; +.globl bsaes_ecb_encrypt_blocks +.type bsaes_ecb_encrypt_blocks,\@abi-omnipotent +.align 16 +bsaes_ecb_encrypt_blocks: + mov %rsp, %rax +.Lecb_enc_prologue: + push %rbp + push %rbx + push %r12 + push %r13 + push %r14 + push %r15 + lea -0x48(%rsp),%rsp +___ +$code.=<<___ if ($win64); + lea -0xa0(%rsp), %rsp + movaps %xmm6, 0x40(%rsp) + movaps %xmm7, 0x50(%rsp) + movaps %xmm8, 0x60(%rsp) + movaps %xmm9, 0x70(%rsp) + movaps %xmm10, 0x80(%rsp) + movaps %xmm11, 0x90(%rsp) + movaps %xmm12, 0xa0(%rsp) + movaps %xmm13, 0xb0(%rsp) + movaps %xmm14, 0xc0(%rsp) + movaps %xmm15, 0xd0(%rsp) +.Lecb_enc_body: +___ +$code.=<<___; + mov %rsp,%rbp # backup %rsp + mov 240($arg4),%eax # rounds + mov $arg1,$inp # backup arguments + mov $arg2,$out + mov $arg3,$len + mov $arg4,$key + cmp \$8,$arg3 + jb .Lecb_enc_short + + mov %eax,%ebx # backup rounds + shl \$7,%rax # 128 bytes per inner round key + sub \$`128-32`,%rax # size of bit-sliced key schedule + sub %rax,%rsp + mov %rsp,%rax # pass key schedule + mov $key,%rcx # pass key + mov %ebx,%r10d # pass rounds + call _bsaes_key_convert + pxor %xmm6,%xmm7 # fix up last round key + movdqa %xmm7,(%rax) # save last round key + + sub \$8,$len +.Lecb_enc_loop: + movdqu 0x00($inp), @XMM[0] # load input + movdqu 0x10($inp), @XMM[1] + movdqu 0x20($inp), @XMM[2] + movdqu 0x30($inp), @XMM[3] + movdqu 0x40($inp), @XMM[4] + movdqu 0x50($inp), @XMM[5] + mov %rsp, %rax # pass key schedule + movdqu 0x60($inp), @XMM[6] + mov %ebx,%r10d # pass rounds + movdqu 0x70($inp), @XMM[7] + lea 0x80($inp), $inp + + call _bsaes_encrypt8 + + movdqu @XMM[0], 0x00($out) # write output + movdqu @XMM[1], 0x10($out) + movdqu @XMM[4], 0x20($out) + movdqu @XMM[6], 0x30($out) + movdqu @XMM[3], 0x40($out) + movdqu @XMM[7], 0x50($out) + movdqu @XMM[2], 0x60($out) + movdqu @XMM[5], 0x70($out) + lea 0x80($out), $out + sub \$8,$len + jnc .Lecb_enc_loop + + add \$8,$len + jz .Lecb_enc_done + + movdqu 0x00($inp), @XMM[0] # load input + mov %rsp, %rax # pass key schedule + mov %ebx,%r10d # pass rounds + cmp \$2,$len + jb .Lecb_enc_one + movdqu 0x10($inp), @XMM[1] + je .Lecb_enc_two + movdqu 0x20($inp), @XMM[2] + cmp \$4,$len + jb .Lecb_enc_three + movdqu 0x30($inp), @XMM[3] + je .Lecb_enc_four + movdqu 0x40($inp), @XMM[4] + cmp \$6,$len + jb .Lecb_enc_five + movdqu 0x50($inp), @XMM[5] + je .Lecb_enc_six + movdqu 0x60($inp), @XMM[6] + call _bsaes_encrypt8 + movdqu @XMM[0], 0x00($out) # write output + movdqu @XMM[1], 0x10($out) + movdqu @XMM[4], 0x20($out) + movdqu @XMM[6], 0x30($out) + movdqu @XMM[3], 0x40($out) + movdqu @XMM[7], 0x50($out) + movdqu @XMM[2], 0x60($out) + jmp .Lecb_enc_done +.align 16 +.Lecb_enc_six: + call _bsaes_encrypt8 + movdqu @XMM[0], 0x00($out) # write output + movdqu @XMM[1], 0x10($out) + movdqu @XMM[4], 0x20($out) + movdqu @XMM[6], 0x30($out) + movdqu @XMM[3], 0x40($out) + movdqu @XMM[7], 0x50($out) + jmp .Lecb_enc_done +.align 16 +.Lecb_enc_five: + call _bsaes_encrypt8 + movdqu @XMM[0], 0x00($out) # write output + movdqu @XMM[1], 0x10($out) + movdqu @XMM[4], 0x20($out) + movdqu @XMM[6], 0x30($out) + movdqu @XMM[3], 0x40($out) + jmp .Lecb_enc_done +.align 16 +.Lecb_enc_four: + call _bsaes_encrypt8 + movdqu @XMM[0], 0x00($out) # write output + movdqu @XMM[1], 0x10($out) + movdqu @XMM[4], 0x20($out) + movdqu @XMM[6], 0x30($out) + jmp .Lecb_enc_done +.align 16 +.Lecb_enc_three: + call _bsaes_encrypt8 + movdqu @XMM[0], 0x00($out) # write output + movdqu @XMM[1], 0x10($out) + movdqu @XMM[4], 0x20($out) + jmp .Lecb_enc_done +.align 16 +.Lecb_enc_two: + call _bsaes_encrypt8 + movdqu @XMM[0], 0x00($out) # write output + movdqu @XMM[1], 0x10($out) + jmp .Lecb_enc_done +.align 16 +.Lecb_enc_one: + call _bsaes_encrypt8 + movdqu @XMM[0], 0x00($out) # write output + jmp .Lecb_enc_done +.align 16 +.Lecb_enc_short: + lea ($inp), $arg1 + lea ($out), $arg2 + lea ($key), $arg3 + call asm_AES_encrypt + lea 16($inp), $inp + lea 16($out), $out + dec $len + jnz .Lecb_enc_short + +.Lecb_enc_done: + lea (%rsp),%rax + pxor %xmm0, %xmm0 +.Lecb_enc_bzero: # wipe key schedule [if any] + movdqa %xmm0, 0x00(%rax) + movdqa %xmm0, 0x10(%rax) + lea 0x20(%rax), %rax + cmp %rax, %rbp + jb .Lecb_enc_bzero + + lea (%rbp),%rsp # restore %rsp +___ +$code.=<<___ if ($win64); + movaps 0x40(%rbp), %xmm6 + movaps 0x50(%rbp), %xmm7 + movaps 0x60(%rbp), %xmm8 + movaps 0x70(%rbp), %xmm9 + movaps 0x80(%rbp), %xmm10 + movaps 0x90(%rbp), %xmm11 + movaps 0xa0(%rbp), %xmm12 + movaps 0xb0(%rbp), %xmm13 + movaps 0xc0(%rbp), %xmm14 + movaps 0xd0(%rbp), %xmm15 + lea 0xa0(%rbp), %rsp +___ +$code.=<<___; + mov 0x48(%rsp), %r15 + mov 0x50(%rsp), %r14 + mov 0x58(%rsp), %r13 + mov 0x60(%rsp), %r12 + mov 0x68(%rsp), %rbx + mov 0x70(%rsp), %rax + lea 0x78(%rsp), %rsp + mov %rax, %rbp +.Lecb_enc_epilogue: + ret +.size bsaes_ecb_encrypt_blocks,.-bsaes_ecb_encrypt_blocks + +.globl bsaes_ecb_decrypt_blocks +.type bsaes_ecb_decrypt_blocks,\@abi-omnipotent +.align 16 +bsaes_ecb_decrypt_blocks: + mov %rsp, %rax +.Lecb_dec_prologue: + push %rbp + push %rbx + push %r12 + push %r13 + push %r14 + push %r15 + lea -0x48(%rsp),%rsp +___ +$code.=<<___ if ($win64); + lea -0xa0(%rsp), %rsp + movaps %xmm6, 0x40(%rsp) + movaps %xmm7, 0x50(%rsp) + movaps %xmm8, 0x60(%rsp) + movaps %xmm9, 0x70(%rsp) + movaps %xmm10, 0x80(%rsp) + movaps %xmm11, 0x90(%rsp) + movaps %xmm12, 0xa0(%rsp) + movaps %xmm13, 0xb0(%rsp) + movaps %xmm14, 0xc0(%rsp) + movaps %xmm15, 0xd0(%rsp) +.Lecb_dec_body: +___ +$code.=<<___; + mov %rsp,%rbp # backup %rsp + mov 240($arg4),%eax # rounds + mov $arg1,$inp # backup arguments + mov $arg2,$out + mov $arg3,$len + mov $arg4,$key + cmp \$8,$arg3 + jb .Lecb_dec_short + + mov %eax,%ebx # backup rounds + shl \$7,%rax # 128 bytes per inner round key + sub \$`128-32`,%rax # size of bit-sliced key schedule + sub %rax,%rsp + mov %rsp,%rax # pass key schedule + mov $key,%rcx # pass key + mov %ebx,%r10d # pass rounds + call _bsaes_key_convert + pxor (%rsp),%xmm7 # fix up 0 round key + movdqa %xmm6,(%rax) # save last round key + movdqa %xmm7,(%rsp) + + sub \$8,$len +.Lecb_dec_loop: + movdqu 0x00($inp), @XMM[0] # load input + movdqu 0x10($inp), @XMM[1] + movdqu 0x20($inp), @XMM[2] + movdqu 0x30($inp), @XMM[3] + movdqu 0x40($inp), @XMM[4] + movdqu 0x50($inp), @XMM[5] + mov %rsp, %rax # pass key schedule + movdqu 0x60($inp), @XMM[6] + mov %ebx,%r10d # pass rounds + movdqu 0x70($inp), @XMM[7] + lea 0x80($inp), $inp + + call _bsaes_decrypt8 + + movdqu @XMM[0], 0x00($out) # write output + movdqu @XMM[1], 0x10($out) + movdqu @XMM[6], 0x20($out) + movdqu @XMM[4], 0x30($out) + movdqu @XMM[2], 0x40($out) + movdqu @XMM[7], 0x50($out) + movdqu @XMM[3], 0x60($out) + movdqu @XMM[5], 0x70($out) + lea 0x80($out), $out + sub \$8,$len + jnc .Lecb_dec_loop + + add \$8,$len + jz .Lecb_dec_done + + movdqu 0x00($inp), @XMM[0] # load input + mov %rsp, %rax # pass key schedule + mov %ebx,%r10d # pass rounds + cmp \$2,$len + jb .Lecb_dec_one + movdqu 0x10($inp), @XMM[1] + je .Lecb_dec_two + movdqu 0x20($inp), @XMM[2] + cmp \$4,$len + jb .Lecb_dec_three + movdqu 0x30($inp), @XMM[3] + je .Lecb_dec_four + movdqu 0x40($inp), @XMM[4] + cmp \$6,$len + jb .Lecb_dec_five + movdqu 0x50($inp), @XMM[5] + je .Lecb_dec_six + movdqu 0x60($inp), @XMM[6] + call _bsaes_decrypt8 + movdqu @XMM[0], 0x00($out) # write output + movdqu @XMM[1], 0x10($out) + movdqu @XMM[6], 0x20($out) + movdqu @XMM[4], 0x30($out) + movdqu @XMM[2], 0x40($out) + movdqu @XMM[7], 0x50($out) + movdqu @XMM[3], 0x60($out) + jmp .Lecb_dec_done +.align 16 +.Lecb_dec_six: + call _bsaes_decrypt8 + movdqu @XMM[0], 0x00($out) # write output + movdqu @XMM[1], 0x10($out) + movdqu @XMM[6], 0x20($out) + movdqu @XMM[4], 0x30($out) + movdqu @XMM[2], 0x40($out) + movdqu @XMM[7], 0x50($out) + jmp .Lecb_dec_done +.align 16 +.Lecb_dec_five: + call _bsaes_decrypt8 + movdqu @XMM[0], 0x00($out) # write output + movdqu @XMM[1], 0x10($out) + movdqu @XMM[6], 0x20($out) + movdqu @XMM[4], 0x30($out) + movdqu @XMM[2], 0x40($out) + jmp .Lecb_dec_done +.align 16 +.Lecb_dec_four: + call _bsaes_decrypt8 + movdqu @XMM[0], 0x00($out) # write output + movdqu @XMM[1], 0x10($out) + movdqu @XMM[6], 0x20($out) + movdqu @XMM[4], 0x30($out) + jmp .Lecb_dec_done +.align 16 +.Lecb_dec_three: + call _bsaes_decrypt8 + movdqu @XMM[0], 0x00($out) # write output + movdqu @XMM[1], 0x10($out) + movdqu @XMM[6], 0x20($out) + jmp .Lecb_dec_done +.align 16 +.Lecb_dec_two: + call _bsaes_decrypt8 + movdqu @XMM[0], 0x00($out) # write output + movdqu @XMM[1], 0x10($out) + jmp .Lecb_dec_done +.align 16 +.Lecb_dec_one: + call _bsaes_decrypt8 + movdqu @XMM[0], 0x00($out) # write output + jmp .Lecb_dec_done +.align 16 +.Lecb_dec_short: + lea ($inp), $arg1 + lea ($out), $arg2 + lea ($key), $arg3 + call asm_AES_decrypt + lea 16($inp), $inp + lea 16($out), $out + dec $len + jnz .Lecb_dec_short + +.Lecb_dec_done: + lea (%rsp),%rax + pxor %xmm0, %xmm0 +.Lecb_dec_bzero: # wipe key schedule [if any] + movdqa %xmm0, 0x00(%rax) + movdqa %xmm0, 0x10(%rax) + lea 0x20(%rax), %rax + cmp %rax, %rbp + jb .Lecb_dec_bzero + + lea (%rbp),%rsp # restore %rsp +___ +$code.=<<___ if ($win64); + movaps 0x40(%rbp), %xmm6 + movaps 0x50(%rbp), %xmm7 + movaps 0x60(%rbp), %xmm8 + movaps 0x70(%rbp), %xmm9 + movaps 0x80(%rbp), %xmm10 + movaps 0x90(%rbp), %xmm11 + movaps 0xa0(%rbp), %xmm12 + movaps 0xb0(%rbp), %xmm13 + movaps 0xc0(%rbp), %xmm14 + movaps 0xd0(%rbp), %xmm15 + lea 0xa0(%rbp), %rsp +___ +$code.=<<___; + mov 0x48(%rsp), %r15 + mov 0x50(%rsp), %r14 + mov 0x58(%rsp), %r13 + mov 0x60(%rsp), %r12 + mov 0x68(%rsp), %rbx + mov 0x70(%rsp), %rax + lea 0x78(%rsp), %rsp + mov %rax, %rbp +.Lecb_dec_epilogue: + ret +.size bsaes_ecb_decrypt_blocks,.-bsaes_ecb_decrypt_blocks +___ +} +$code.=<<___; +.extern asm_AES_cbc_encrypt +.globl bsaes_cbc_encrypt +.type bsaes_cbc_encrypt,\@abi-omnipotent +.align 16 +bsaes_cbc_encrypt: +___ +$code.=<<___ if ($win64); + mov 48(%rsp),$arg6 # pull direction flag +___ +$code.=<<___; + cmp \$0,$arg6 + jne asm_AES_cbc_encrypt + cmp \$128,$arg3 + jb asm_AES_cbc_encrypt + + mov %rsp, %rax +.Lcbc_dec_prologue: + push %rbp + push %rbx + push %r12 + push %r13 + push %r14 + push %r15 + lea -0x48(%rsp), %rsp +___ +$code.=<<___ if ($win64); + mov 0xa0(%rsp),$arg5 # pull ivp + lea -0xa0(%rsp), %rsp + movaps %xmm6, 0x40(%rsp) + movaps %xmm7, 0x50(%rsp) + movaps %xmm8, 0x60(%rsp) + movaps %xmm9, 0x70(%rsp) + movaps %xmm10, 0x80(%rsp) + movaps %xmm11, 0x90(%rsp) + movaps %xmm12, 0xa0(%rsp) + movaps %xmm13, 0xb0(%rsp) + movaps %xmm14, 0xc0(%rsp) + movaps %xmm15, 0xd0(%rsp) +.Lcbc_dec_body: +___ +$code.=<<___; + mov %rsp, %rbp # backup %rsp + mov 240($arg4), %eax # rounds + mov $arg1, $inp # backup arguments + mov $arg2, $out + mov $arg3, $len + mov $arg4, $key + mov $arg5, %rbx + shr \$4, $len # bytes to blocks + + mov %eax, %edx # rounds + shl \$7, %rax # 128 bytes per inner round key + sub \$`128-32`, %rax # size of bit-sliced key schedule + sub %rax, %rsp + + mov %rsp, %rax # pass key schedule + mov $key, %rcx # pass key + mov %edx, %r10d # pass rounds + call _bsaes_key_convert + pxor (%rsp),%xmm7 # fix up 0 round key + movdqa %xmm6,(%rax) # save last round key + movdqa %xmm7,(%rsp) + + movdqu (%rbx), @XMM[15] # load IV + sub \$8,$len +.Lcbc_dec_loop: + movdqu 0x00($inp), @XMM[0] # load input + movdqu 0x10($inp), @XMM[1] + movdqu 0x20($inp), @XMM[2] + movdqu 0x30($inp), @XMM[3] + movdqu 0x40($inp), @XMM[4] + movdqu 0x50($inp), @XMM[5] + mov %rsp, %rax # pass key schedule + movdqu 0x60($inp), @XMM[6] + mov %edx,%r10d # pass rounds + movdqu 0x70($inp), @XMM[7] + movdqa @XMM[15], 0x20(%rbp) # put aside IV + + call _bsaes_decrypt8 + + pxor 0x20(%rbp), @XMM[0] # ^= IV + movdqu 0x00($inp), @XMM[8] # re-load input + movdqu 0x10($inp), @XMM[9] + pxor @XMM[8], @XMM[1] + movdqu 0x20($inp), @XMM[10] + pxor @XMM[9], @XMM[6] + movdqu 0x30($inp), @XMM[11] + pxor @XMM[10], @XMM[4] + movdqu 0x40($inp), @XMM[12] + pxor @XMM[11], @XMM[2] + movdqu 0x50($inp), @XMM[13] + pxor @XMM[12], @XMM[7] + movdqu 0x60($inp), @XMM[14] + pxor @XMM[13], @XMM[3] + movdqu 0x70($inp), @XMM[15] # IV + pxor @XMM[14], @XMM[5] + movdqu @XMM[0], 0x00($out) # write output + lea 0x80($inp), $inp + movdqu @XMM[1], 0x10($out) + movdqu @XMM[6], 0x20($out) + movdqu @XMM[4], 0x30($out) + movdqu @XMM[2], 0x40($out) + movdqu @XMM[7], 0x50($out) + movdqu @XMM[3], 0x60($out) + movdqu @XMM[5], 0x70($out) + lea 0x80($out), $out + sub \$8,$len + jnc .Lcbc_dec_loop + + add \$8,$len + jz .Lcbc_dec_done + + movdqu 0x00($inp), @XMM[0] # load input + mov %rsp, %rax # pass key schedule + mov %edx, %r10d # pass rounds + cmp \$2,$len + jb .Lcbc_dec_one + movdqu 0x10($inp), @XMM[1] + je .Lcbc_dec_two + movdqu 0x20($inp), @XMM[2] + cmp \$4,$len + jb .Lcbc_dec_three + movdqu 0x30($inp), @XMM[3] + je .Lcbc_dec_four + movdqu 0x40($inp), @XMM[4] + cmp \$6,$len + jb .Lcbc_dec_five + movdqu 0x50($inp), @XMM[5] + je .Lcbc_dec_six + movdqu 0x60($inp), @XMM[6] + movdqa @XMM[15], 0x20(%rbp) # put aside IV + call _bsaes_decrypt8 + pxor 0x20(%rbp), @XMM[0] # ^= IV + movdqu 0x00($inp), @XMM[8] # re-load input + movdqu 0x10($inp), @XMM[9] + pxor @XMM[8], @XMM[1] + movdqu 0x20($inp), @XMM[10] + pxor @XMM[9], @XMM[6] + movdqu 0x30($inp), @XMM[11] + pxor @XMM[10], @XMM[4] + movdqu 0x40($inp), @XMM[12] + pxor @XMM[11], @XMM[2] + movdqu 0x50($inp), @XMM[13] + pxor @XMM[12], @XMM[7] + movdqu 0x60($inp), @XMM[15] # IV + pxor @XMM[13], @XMM[3] + movdqu @XMM[0], 0x00($out) # write output + movdqu @XMM[1], 0x10($out) + movdqu @XMM[6], 0x20($out) + movdqu @XMM[4], 0x30($out) + movdqu @XMM[2], 0x40($out) + movdqu @XMM[7], 0x50($out) + movdqu @XMM[3], 0x60($out) + jmp .Lcbc_dec_done +.align 16 +.Lcbc_dec_six: + movdqa @XMM[15], 0x20(%rbp) # put aside IV + call _bsaes_decrypt8 + pxor 0x20(%rbp), @XMM[0] # ^= IV + movdqu 0x00($inp), @XMM[8] # re-load input + movdqu 0x10($inp), @XMM[9] + pxor @XMM[8], @XMM[1] + movdqu 0x20($inp), @XMM[10] + pxor @XMM[9], @XMM[6] + movdqu 0x30($inp), @XMM[11] + pxor @XMM[10], @XMM[4] + movdqu 0x40($inp), @XMM[12] + pxor @XMM[11], @XMM[2] + movdqu 0x50($inp), @XMM[15] # IV + pxor @XMM[12], @XMM[7] + movdqu @XMM[0], 0x00($out) # write output + movdqu @XMM[1], 0x10($out) + movdqu @XMM[6], 0x20($out) + movdqu @XMM[4], 0x30($out) + movdqu @XMM[2], 0x40($out) + movdqu @XMM[7], 0x50($out) + jmp .Lcbc_dec_done +.align 16 +.Lcbc_dec_five: + movdqa @XMM[15], 0x20(%rbp) # put aside IV + call _bsaes_decrypt8 + pxor 0x20(%rbp), @XMM[0] # ^= IV + movdqu 0x00($inp), @XMM[8] # re-load input + movdqu 0x10($inp), @XMM[9] + pxor @XMM[8], @XMM[1] + movdqu 0x20($inp), @XMM[10] + pxor @XMM[9], @XMM[6] + movdqu 0x30($inp), @XMM[11] + pxor @XMM[10], @XMM[4] + movdqu 0x40($inp), @XMM[15] # IV + pxor @XMM[11], @XMM[2] + movdqu @XMM[0], 0x00($out) # write output + movdqu @XMM[1], 0x10($out) + movdqu @XMM[6], 0x20($out) + movdqu @XMM[4], 0x30($out) + movdqu @XMM[2], 0x40($out) + jmp .Lcbc_dec_done +.align 16 +.Lcbc_dec_four: + movdqa @XMM[15], 0x20(%rbp) # put aside IV + call _bsaes_decrypt8 + pxor 0x20(%rbp), @XMM[0] # ^= IV + movdqu 0x00($inp), @XMM[8] # re-load input + movdqu 0x10($inp), @XMM[9] + pxor @XMM[8], @XMM[1] + movdqu 0x20($inp), @XMM[10] + pxor @XMM[9], @XMM[6] + movdqu 0x30($inp), @XMM[15] # IV + pxor @XMM[10], @XMM[4] + movdqu @XMM[0], 0x00($out) # write output + movdqu @XMM[1], 0x10($out) + movdqu @XMM[6], 0x20($out) + movdqu @XMM[4], 0x30($out) + jmp .Lcbc_dec_done +.align 16 +.Lcbc_dec_three: + movdqa @XMM[15], 0x20(%rbp) # put aside IV + call _bsaes_decrypt8 + pxor 0x20(%rbp), @XMM[0] # ^= IV + movdqu 0x00($inp), @XMM[8] # re-load input + movdqu 0x10($inp), @XMM[9] + pxor @XMM[8], @XMM[1] + movdqu 0x20($inp), @XMM[15] # IV + pxor @XMM[9], @XMM[6] + movdqu @XMM[0], 0x00($out) # write output + movdqu @XMM[1], 0x10($out) + movdqu @XMM[6], 0x20($out) + jmp .Lcbc_dec_done +.align 16 +.Lcbc_dec_two: + movdqa @XMM[15], 0x20(%rbp) # put aside IV + call _bsaes_decrypt8 + pxor 0x20(%rbp), @XMM[0] # ^= IV + movdqu 0x00($inp), @XMM[8] # re-load input + movdqu 0x10($inp), @XMM[15] # IV + pxor @XMM[8], @XMM[1] + movdqu @XMM[0], 0x00($out) # write output + movdqu @XMM[1], 0x10($out) + jmp .Lcbc_dec_done +.align 16 +.Lcbc_dec_one: + lea ($inp), $arg1 + lea 0x20(%rbp), $arg2 # buffer output + lea ($key), $arg3 + call asm_AES_decrypt # doesn't touch %xmm + pxor 0x20(%rbp), @XMM[15] # ^= IV + movdqu @XMM[15], ($out) # write output + movdqa @XMM[0], @XMM[15] # IV + +.Lcbc_dec_done: + movdqu @XMM[15], (%rbx) # return IV + lea (%rsp), %rax + pxor %xmm0, %xmm0 +.Lcbc_dec_bzero: # wipe key schedule [if any] + movdqa %xmm0, 0x00(%rax) + movdqa %xmm0, 0x10(%rax) + lea 0x20(%rax), %rax + cmp %rax, %rbp + ja .Lcbc_dec_bzero + + lea (%rbp),%rsp # restore %rsp +___ +$code.=<<___ if ($win64); + movaps 0x40(%rbp), %xmm6 + movaps 0x50(%rbp), %xmm7 + movaps 0x60(%rbp), %xmm8 + movaps 0x70(%rbp), %xmm9 + movaps 0x80(%rbp), %xmm10 + movaps 0x90(%rbp), %xmm11 + movaps 0xa0(%rbp), %xmm12 + movaps 0xb0(%rbp), %xmm13 + movaps 0xc0(%rbp), %xmm14 + movaps 0xd0(%rbp), %xmm15 + lea 0xa0(%rbp), %rsp +___ +$code.=<<___; + mov 0x48(%rsp), %r15 + mov 0x50(%rsp), %r14 + mov 0x58(%rsp), %r13 + mov 0x60(%rsp), %r12 + mov 0x68(%rsp), %rbx + mov 0x70(%rsp), %rax + lea 0x78(%rsp), %rsp + mov %rax, %rbp +.Lcbc_dec_epilogue: + ret +.size bsaes_cbc_encrypt,.-bsaes_cbc_encrypt + +.globl bsaes_ctr32_encrypt_blocks +.type bsaes_ctr32_encrypt_blocks,\@abi-omnipotent +.align 16 +bsaes_ctr32_encrypt_blocks: + mov %rsp, %rax +.Lctr_enc_prologue: + push %rbp + push %rbx + push %r12 + push %r13 + push %r14 + push %r15 + lea -0x48(%rsp), %rsp +___ +$code.=<<___ if ($win64); + mov 0xa0(%rsp),$arg5 # pull ivp + lea -0xa0(%rsp), %rsp + movaps %xmm6, 0x40(%rsp) + movaps %xmm7, 0x50(%rsp) + movaps %xmm8, 0x60(%rsp) + movaps %xmm9, 0x70(%rsp) + movaps %xmm10, 0x80(%rsp) + movaps %xmm11, 0x90(%rsp) + movaps %xmm12, 0xa0(%rsp) + movaps %xmm13, 0xb0(%rsp) + movaps %xmm14, 0xc0(%rsp) + movaps %xmm15, 0xd0(%rsp) +.Lctr_enc_body: +___ +$code.=<<___; + mov %rsp, %rbp # backup %rsp + movdqu ($arg5), %xmm0 # load counter + mov 240($arg4), %eax # rounds + mov $arg1, $inp # backup arguments + mov $arg2, $out + mov $arg3, $len + mov $arg4, $key + movdqa %xmm0, 0x20(%rbp) # copy counter + cmp \$8, $arg3 + jb .Lctr_enc_short + + mov %eax, %ebx # rounds + shl \$7, %rax # 128 bytes per inner round key + sub \$`128-32`, %rax # size of bit-sliced key schedule + sub %rax, %rsp + + mov %rsp, %rax # pass key schedule + mov $key, %rcx # pass key + mov %ebx, %r10d # pass rounds + call _bsaes_key_convert + pxor %xmm6,%xmm7 # fix up last round key + movdqa %xmm7,(%rax) # save last round key + + movdqa (%rsp), @XMM[9] # load round0 key + lea .LADD1(%rip), %r11 + movdqa 0x20(%rbp), @XMM[0] # counter copy + movdqa -0x20(%r11), @XMM[8] # .LSWPUP + pshufb @XMM[8], @XMM[9] # byte swap upper part + pshufb @XMM[8], @XMM[0] + movdqa @XMM[9], (%rsp) # save adjusted round0 key + jmp .Lctr_enc_loop +.align 16 +.Lctr_enc_loop: + movdqa @XMM[0], 0x20(%rbp) # save counter + movdqa @XMM[0], @XMM[1] # prepare 8 counter values + movdqa @XMM[0], @XMM[2] + paddd 0x00(%r11), @XMM[1] # .LADD1 + movdqa @XMM[0], @XMM[3] + paddd 0x10(%r11), @XMM[2] # .LADD2 + movdqa @XMM[0], @XMM[4] + paddd 0x20(%r11), @XMM[3] # .LADD3 + movdqa @XMM[0], @XMM[5] + paddd 0x30(%r11), @XMM[4] # .LADD4 + movdqa @XMM[0], @XMM[6] + paddd 0x40(%r11), @XMM[5] # .LADD5 + movdqa @XMM[0], @XMM[7] + paddd 0x50(%r11), @XMM[6] # .LADD6 + paddd 0x60(%r11), @XMM[7] # .LADD7 + + # Borrow prologue from _bsaes_encrypt8 to use the opportunity + # to flip byte order in 32-bit counter + movdqa (%rsp), @XMM[9] # round 0 key + lea 0x10(%rsp), %rax # pass key schedule + movdqa -0x10(%r11), @XMM[8] # .LSWPUPM0SR + pxor @XMM[9], @XMM[0] # xor with round0 key + pxor @XMM[9], @XMM[1] + pxor @XMM[9], @XMM[2] + pxor @XMM[9], @XMM[3] + pshufb @XMM[8], @XMM[0] + pshufb @XMM[8], @XMM[1] + pxor @XMM[9], @XMM[4] + pxor @XMM[9], @XMM[5] + pshufb @XMM[8], @XMM[2] + pshufb @XMM[8], @XMM[3] + pxor @XMM[9], @XMM[6] + pxor @XMM[9], @XMM[7] + pshufb @XMM[8], @XMM[4] + pshufb @XMM[8], @XMM[5] + pshufb @XMM[8], @XMM[6] + pshufb @XMM[8], @XMM[7] + lea .LBS0(%rip), %r11 # constants table + mov %ebx,%r10d # pass rounds + + call _bsaes_encrypt8_bitslice + + sub \$8,$len + jc .Lctr_enc_loop_done + + movdqu 0x00($inp), @XMM[8] # load input + movdqu 0x10($inp), @XMM[9] + movdqu 0x20($inp), @XMM[10] + movdqu 0x30($inp), @XMM[11] + movdqu 0x40($inp), @XMM[12] + movdqu 0x50($inp), @XMM[13] + movdqu 0x60($inp), @XMM[14] + movdqu 0x70($inp), @XMM[15] + lea 0x80($inp),$inp + pxor @XMM[0], @XMM[8] + movdqa 0x20(%rbp), @XMM[0] # load counter + pxor @XMM[9], @XMM[1] + movdqu @XMM[8], 0x00($out) # write output + pxor @XMM[10], @XMM[4] + movdqu @XMM[1], 0x10($out) + pxor @XMM[11], @XMM[6] + movdqu @XMM[4], 0x20($out) + pxor @XMM[12], @XMM[3] + movdqu @XMM[6], 0x30($out) + pxor @XMM[13], @XMM[7] + movdqu @XMM[3], 0x40($out) + pxor @XMM[14], @XMM[2] + movdqu @XMM[7], 0x50($out) + pxor @XMM[15], @XMM[5] + movdqu @XMM[2], 0x60($out) + lea .LADD1(%rip), %r11 + movdqu @XMM[5], 0x70($out) + lea 0x80($out), $out + paddd 0x70(%r11), @XMM[0] # .LADD8 + jnz .Lctr_enc_loop + + jmp .Lctr_enc_done +.align 16 +.Lctr_enc_loop_done: + add \$8, $len + movdqu 0x00($inp), @XMM[8] # load input + pxor @XMM[8], @XMM[0] + movdqu @XMM[0], 0x00($out) # write output + cmp \$2,$len + jb .Lctr_enc_done + movdqu 0x10($inp), @XMM[9] + pxor @XMM[9], @XMM[1] + movdqu @XMM[1], 0x10($out) + je .Lctr_enc_done + movdqu 0x20($inp), @XMM[10] + pxor @XMM[10], @XMM[4] + movdqu @XMM[4], 0x20($out) + cmp \$4,$len + jb .Lctr_enc_done + movdqu 0x30($inp), @XMM[11] + pxor @XMM[11], @XMM[6] + movdqu @XMM[6], 0x30($out) + je .Lctr_enc_done + movdqu 0x40($inp), @XMM[12] + pxor @XMM[12], @XMM[3] + movdqu @XMM[3], 0x40($out) + cmp \$6,$len + jb .Lctr_enc_done + movdqu 0x50($inp), @XMM[13] + pxor @XMM[13], @XMM[7] + movdqu @XMM[7], 0x50($out) + je .Lctr_enc_done + movdqu 0x60($inp), @XMM[14] + pxor @XMM[14], @XMM[2] + movdqu @XMM[2], 0x60($out) + jmp .Lctr_enc_done + +.align 16 +.Lctr_enc_short: + lea 0x20(%rbp), $arg1 + lea 0x30(%rbp), $arg2 + lea ($key), $arg3 + call asm_AES_encrypt + movdqu ($inp), @XMM[1] + lea 16($inp), $inp + mov 0x2c(%rbp), %eax # load 32-bit counter + bswap %eax + pxor 0x30(%rbp), @XMM[1] + inc %eax # increment + movdqu @XMM[1], ($out) + bswap %eax + lea 16($out), $out + mov %eax, 0x2c(%rsp) # save 32-bit counter + dec $len + jnz .Lctr_enc_short + +.Lctr_enc_done: + lea (%rsp), %rax + pxor %xmm0, %xmm0 +.Lctr_enc_bzero: # wipe key schedule [if any] + movdqa %xmm0, 0x00(%rax) + movdqa %xmm0, 0x10(%rax) + lea 0x20(%rax), %rax + cmp %rax, %rbp + ja .Lctr_enc_bzero + + lea (%rbp),%rsp # restore %rsp +___ +$code.=<<___ if ($win64); + movaps 0x40(%rbp), %xmm6 + movaps 0x50(%rbp), %xmm7 + movaps 0x60(%rbp), %xmm8 + movaps 0x70(%rbp), %xmm9 + movaps 0x80(%rbp), %xmm10 + movaps 0x90(%rbp), %xmm11 + movaps 0xa0(%rbp), %xmm12 + movaps 0xb0(%rbp), %xmm13 + movaps 0xc0(%rbp), %xmm14 + movaps 0xd0(%rbp), %xmm15 + lea 0xa0(%rbp), %rsp +___ +$code.=<<___; + mov 0x48(%rsp), %r15 + mov 0x50(%rsp), %r14 + mov 0x58(%rsp), %r13 + mov 0x60(%rsp), %r12 + mov 0x68(%rsp), %rbx + mov 0x70(%rsp), %rax + lea 0x78(%rsp), %rsp + mov %rax, %rbp +.Lctr_enc_epilogue: + ret +.size bsaes_ctr32_encrypt_blocks,.-bsaes_ctr32_encrypt_blocks +___ +###################################################################### +# void bsaes_xts_[en|de]crypt(const char *inp,char *out,size_t len, +# const AES_KEY *key1, const AES_KEY *key2, +# const unsigned char iv[16]); +# +my ($twmask,$twres,$twtmp)=@XMM[13..15]; +$arg6=~s/d$//; + +$code.=<<___; +.globl bsaes_xts_encrypt +.type bsaes_xts_encrypt,\@abi-omnipotent +.align 16 +bsaes_xts_encrypt: + mov %rsp, %rax +.Lxts_enc_prologue: + push %rbp + push %rbx + push %r12 + push %r13 + push %r14 + push %r15 + lea -0x48(%rsp), %rsp +___ +$code.=<<___ if ($win64); + mov 0xa0(%rsp),$arg5 # pull key2 + mov 0xa8(%rsp),$arg6 # pull ivp + lea -0xa0(%rsp), %rsp + movaps %xmm6, 0x40(%rsp) + movaps %xmm7, 0x50(%rsp) + movaps %xmm8, 0x60(%rsp) + movaps %xmm9, 0x70(%rsp) + movaps %xmm10, 0x80(%rsp) + movaps %xmm11, 0x90(%rsp) + movaps %xmm12, 0xa0(%rsp) + movaps %xmm13, 0xb0(%rsp) + movaps %xmm14, 0xc0(%rsp) + movaps %xmm15, 0xd0(%rsp) +.Lxts_enc_body: +___ +$code.=<<___; + mov %rsp, %rbp # backup %rsp + mov $arg1, $inp # backup arguments + mov $arg2, $out + mov $arg3, $len + mov $arg4, $key + + lea ($arg6), $arg1 + lea 0x20(%rbp), $arg2 + lea ($arg5), $arg3 + call asm_AES_encrypt # generate initial tweak + + mov 240($key), %eax # rounds + mov $len, %rbx # backup $len + + mov %eax, %edx # rounds + shl \$7, %rax # 128 bytes per inner round key + sub \$`128-32`, %rax # size of bit-sliced key schedule + sub %rax, %rsp + + mov %rsp, %rax # pass key schedule + mov $key, %rcx # pass key + mov %edx, %r10d # pass rounds + call _bsaes_key_convert + pxor %xmm6, %xmm7 # fix up last round key + movdqa %xmm7, (%rax) # save last round key + + and \$-16, $len + sub \$0x80, %rsp # place for tweak[8] + movdqa 0x20(%rbp), @XMM[7] # initial tweak + + pxor $twtmp, $twtmp + movdqa .Lxts_magic(%rip), $twmask + pcmpgtd @XMM[7], $twtmp # broadcast upper bits + + sub \$0x80, $len + jc .Lxts_enc_short + jmp .Lxts_enc_loop + +.align 16 +.Lxts_enc_loop: +___ + for ($i=0;$i<7;$i++) { + $code.=<<___; + pshufd \$0x13, $twtmp, $twres + pxor $twtmp, $twtmp + movdqa @XMM[7], @XMM[$i] + movdqa @XMM[7], `0x10*$i`(%rsp)# save tweak[$i] + paddq @XMM[7], @XMM[7] # psllq 1,$tweak + pand $twmask, $twres # isolate carry and residue + pcmpgtd @XMM[7], $twtmp # broadcast upper bits + pxor $twres, @XMM[7] +___ + $code.=<<___ if ($i>=1); + movdqu `0x10*($i-1)`($inp), @XMM[8+$i-1] +___ + $code.=<<___ if ($i>=2); + pxor @XMM[8+$i-2], @XMM[$i-2]# input[] ^ tweak[] +___ + } +$code.=<<___; + movdqu 0x60($inp), @XMM[8+6] + pxor @XMM[8+5], @XMM[5] + movdqu 0x70($inp), @XMM[8+7] + lea 0x80($inp), $inp + movdqa @XMM[7], 0x70(%rsp) + pxor @XMM[8+6], @XMM[6] + lea 0x80(%rsp), %rax # pass key schedule + pxor @XMM[8+7], @XMM[7] + mov %edx, %r10d # pass rounds + + call _bsaes_encrypt8 + + pxor 0x00(%rsp), @XMM[0] # ^= tweak[] + pxor 0x10(%rsp), @XMM[1] + movdqu @XMM[0], 0x00($out) # write output + pxor 0x20(%rsp), @XMM[4] + movdqu @XMM[1], 0x10($out) + pxor 0x30(%rsp), @XMM[6] + movdqu @XMM[4], 0x20($out) + pxor 0x40(%rsp), @XMM[3] + movdqu @XMM[6], 0x30($out) + pxor 0x50(%rsp), @XMM[7] + movdqu @XMM[3], 0x40($out) + pxor 0x60(%rsp), @XMM[2] + movdqu @XMM[7], 0x50($out) + pxor 0x70(%rsp), @XMM[5] + movdqu @XMM[2], 0x60($out) + movdqu @XMM[5], 0x70($out) + lea 0x80($out), $out + + movdqa 0x70(%rsp), @XMM[7] # prepare next iteration tweak + pxor $twtmp, $twtmp + movdqa .Lxts_magic(%rip), $twmask + pcmpgtd @XMM[7], $twtmp + pshufd \$0x13, $twtmp, $twres + pxor $twtmp, $twtmp + paddq @XMM[7], @XMM[7] # psllq 1,$tweak + pand $twmask, $twres # isolate carry and residue + pcmpgtd @XMM[7], $twtmp # broadcast upper bits + pxor $twres, @XMM[7] + + sub \$0x80,$len + jnc .Lxts_enc_loop + +.Lxts_enc_short: + add \$0x80, $len + jz .Lxts_enc_done +___ + for ($i=0;$i<7;$i++) { + $code.=<<___; + pshufd \$0x13, $twtmp, $twres + pxor $twtmp, $twtmp + movdqa @XMM[7], @XMM[$i] + movdqa @XMM[7], `0x10*$i`(%rsp)# save tweak[$i] + paddq @XMM[7], @XMM[7] # psllq 1,$tweak + pand $twmask, $twres # isolate carry and residue + pcmpgtd @XMM[7], $twtmp # broadcast upper bits + pxor $twres, @XMM[7] +___ + $code.=<<___ if ($i>=1); + movdqu `0x10*($i-1)`($inp), @XMM[8+$i-1] + cmp \$`0x10*$i`,$len + je .Lxts_enc_$i +___ + $code.=<<___ if ($i>=2); + pxor @XMM[8+$i-2], @XMM[$i-2]# input[] ^ tweak[] +___ + } +$code.=<<___; + movdqu 0x60($inp), @XMM[8+6] + pxor @XMM[8+5], @XMM[5] + movdqa @XMM[7], 0x70(%rsp) + lea 0x70($inp), $inp + pxor @XMM[8+6], @XMM[6] + lea 0x80(%rsp), %rax # pass key schedule + mov %edx, %r10d # pass rounds + + call _bsaes_encrypt8 + + pxor 0x00(%rsp), @XMM[0] # ^= tweak[] + pxor 0x10(%rsp), @XMM[1] + movdqu @XMM[0], 0x00($out) # write output + pxor 0x20(%rsp), @XMM[4] + movdqu @XMM[1], 0x10($out) + pxor 0x30(%rsp), @XMM[6] + movdqu @XMM[4], 0x20($out) + pxor 0x40(%rsp), @XMM[3] + movdqu @XMM[6], 0x30($out) + pxor 0x50(%rsp), @XMM[7] + movdqu @XMM[3], 0x40($out) + pxor 0x60(%rsp), @XMM[2] + movdqu @XMM[7], 0x50($out) + movdqu @XMM[2], 0x60($out) + lea 0x70($out), $out + + movdqa 0x70(%rsp), @XMM[7] # next iteration tweak + jmp .Lxts_enc_done +.align 16 +.Lxts_enc_6: + pxor @XMM[8+4], @XMM[4] + lea 0x60($inp), $inp + pxor @XMM[8+5], @XMM[5] + lea 0x80(%rsp), %rax # pass key schedule + mov %edx, %r10d # pass rounds + + call _bsaes_encrypt8 + + pxor 0x00(%rsp), @XMM[0] # ^= tweak[] + pxor 0x10(%rsp), @XMM[1] + movdqu @XMM[0], 0x00($out) # write output + pxor 0x20(%rsp), @XMM[4] + movdqu @XMM[1], 0x10($out) + pxor 0x30(%rsp), @XMM[6] + movdqu @XMM[4], 0x20($out) + pxor 0x40(%rsp), @XMM[3] + movdqu @XMM[6], 0x30($out) + pxor 0x50(%rsp), @XMM[7] + movdqu @XMM[3], 0x40($out) + movdqu @XMM[7], 0x50($out) + lea 0x60($out), $out + + movdqa 0x60(%rsp), @XMM[7] # next iteration tweak + jmp .Lxts_enc_done +.align 16 +.Lxts_enc_5: + pxor @XMM[8+3], @XMM[3] + lea 0x50($inp), $inp + pxor @XMM[8+4], @XMM[4] + lea 0x80(%rsp), %rax # pass key schedule + mov %edx, %r10d # pass rounds + + call _bsaes_encrypt8 + + pxor 0x00(%rsp), @XMM[0] # ^= tweak[] + pxor 0x10(%rsp), @XMM[1] + movdqu @XMM[0], 0x00($out) # write output + pxor 0x20(%rsp), @XMM[4] + movdqu @XMM[1], 0x10($out) + pxor 0x30(%rsp), @XMM[6] + movdqu @XMM[4], 0x20($out) + pxor 0x40(%rsp), @XMM[3] + movdqu @XMM[6], 0x30($out) + movdqu @XMM[3], 0x40($out) + lea 0x50($out), $out + + movdqa 0x50(%rsp), @XMM[7] # next iteration tweak + jmp .Lxts_enc_done +.align 16 +.Lxts_enc_4: + pxor @XMM[8+2], @XMM[2] + lea 0x40($inp), $inp + pxor @XMM[8+3], @XMM[3] + lea 0x80(%rsp), %rax # pass key schedule + mov %edx, %r10d # pass rounds + + call _bsaes_encrypt8 + + pxor 0x00(%rsp), @XMM[0] # ^= tweak[] + pxor 0x10(%rsp), @XMM[1] + movdqu @XMM[0], 0x00($out) # write output + pxor 0x20(%rsp), @XMM[4] + movdqu @XMM[1], 0x10($out) + pxor 0x30(%rsp), @XMM[6] + movdqu @XMM[4], 0x20($out) + movdqu @XMM[6], 0x30($out) + lea 0x40($out), $out + + movdqa 0x40(%rsp), @XMM[7] # next iteration tweak + jmp .Lxts_enc_done +.align 16 +.Lxts_enc_3: + pxor @XMM[8+1], @XMM[1] + lea 0x30($inp), $inp + pxor @XMM[8+2], @XMM[2] + lea 0x80(%rsp), %rax # pass key schedule + mov %edx, %r10d # pass rounds + + call _bsaes_encrypt8 + + pxor 0x00(%rsp), @XMM[0] # ^= tweak[] + pxor 0x10(%rsp), @XMM[1] + movdqu @XMM[0], 0x00($out) # write output + pxor 0x20(%rsp), @XMM[4] + movdqu @XMM[1], 0x10($out) + movdqu @XMM[4], 0x20($out) + lea 0x30($out), $out + + movdqa 0x30(%rsp), @XMM[7] # next iteration tweak + jmp .Lxts_enc_done +.align 16 +.Lxts_enc_2: + pxor @XMM[8+0], @XMM[0] + lea 0x20($inp), $inp + pxor @XMM[8+1], @XMM[1] + lea 0x80(%rsp), %rax # pass key schedule + mov %edx, %r10d # pass rounds + + call _bsaes_encrypt8 + + pxor 0x00(%rsp), @XMM[0] # ^= tweak[] + pxor 0x10(%rsp), @XMM[1] + movdqu @XMM[0], 0x00($out) # write output + movdqu @XMM[1], 0x10($out) + lea 0x20($out), $out + + movdqa 0x20(%rsp), @XMM[7] # next iteration tweak + jmp .Lxts_enc_done +.align 16 +.Lxts_enc_1: + pxor @XMM[0], @XMM[8] + lea 0x10($inp), $inp + movdqa @XMM[8], 0x20(%rbp) + lea 0x20(%rbp), $arg1 + lea 0x20(%rbp), $arg2 + lea ($key), $arg3 + call asm_AES_encrypt # doesn't touch %xmm + pxor 0x20(%rbp), @XMM[0] # ^= tweak[] + #pxor @XMM[8], @XMM[0] + #lea 0x80(%rsp), %rax # pass key schedule + #mov %edx, %r10d # pass rounds + #call _bsaes_encrypt8 + #pxor 0x00(%rsp), @XMM[0] # ^= tweak[] + movdqu @XMM[0], 0x00($out) # write output + lea 0x10($out), $out + + movdqa 0x10(%rsp), @XMM[7] # next iteration tweak + +.Lxts_enc_done: + and \$15, %ebx + jz .Lxts_enc_ret + mov $out, %rdx + +.Lxts_enc_steal: + movzb ($inp), %eax + movzb -16(%rdx), %ecx + lea 1($inp), $inp + mov %al, -16(%rdx) + mov %cl, 0(%rdx) + lea 1(%rdx), %rdx + sub \$1,%ebx + jnz .Lxts_enc_steal + + movdqu -16($out), @XMM[0] + lea 0x20(%rbp), $arg1 + pxor @XMM[7], @XMM[0] + lea 0x20(%rbp), $arg2 + movdqa @XMM[0], 0x20(%rbp) + lea ($key), $arg3 + call asm_AES_encrypt # doesn't touch %xmm + pxor 0x20(%rbp), @XMM[7] + movdqu @XMM[7], -16($out) + +.Lxts_enc_ret: + lea (%rsp), %rax + pxor %xmm0, %xmm0 +.Lxts_enc_bzero: # wipe key schedule [if any] + movdqa %xmm0, 0x00(%rax) + movdqa %xmm0, 0x10(%rax) + lea 0x20(%rax), %rax + cmp %rax, %rbp + ja .Lxts_enc_bzero + + lea (%rbp),%rsp # restore %rsp +___ +$code.=<<___ if ($win64); + movaps 0x40(%rbp), %xmm6 + movaps 0x50(%rbp), %xmm7 + movaps 0x60(%rbp), %xmm8 + movaps 0x70(%rbp), %xmm9 + movaps 0x80(%rbp), %xmm10 + movaps 0x90(%rbp), %xmm11 + movaps 0xa0(%rbp), %xmm12 + movaps 0xb0(%rbp), %xmm13 + movaps 0xc0(%rbp), %xmm14 + movaps 0xd0(%rbp), %xmm15 + lea 0xa0(%rbp), %rsp +___ +$code.=<<___; + mov 0x48(%rsp), %r15 + mov 0x50(%rsp), %r14 + mov 0x58(%rsp), %r13 + mov 0x60(%rsp), %r12 + mov 0x68(%rsp), %rbx + mov 0x70(%rsp), %rax + lea 0x78(%rsp), %rsp + mov %rax, %rbp +.Lxts_enc_epilogue: + ret +.size bsaes_xts_encrypt,.-bsaes_xts_encrypt + +.globl bsaes_xts_decrypt +.type bsaes_xts_decrypt,\@abi-omnipotent +.align 16 +bsaes_xts_decrypt: + mov %rsp, %rax +.Lxts_dec_prologue: + push %rbp + push %rbx + push %r12 + push %r13 + push %r14 + push %r15 + lea -0x48(%rsp), %rsp +___ +$code.=<<___ if ($win64); + mov 0xa0(%rsp),$arg5 # pull key2 + mov 0xa8(%rsp),$arg6 # pull ivp + lea -0xa0(%rsp), %rsp + movaps %xmm6, 0x40(%rsp) + movaps %xmm7, 0x50(%rsp) + movaps %xmm8, 0x60(%rsp) + movaps %xmm9, 0x70(%rsp) + movaps %xmm10, 0x80(%rsp) + movaps %xmm11, 0x90(%rsp) + movaps %xmm12, 0xa0(%rsp) + movaps %xmm13, 0xb0(%rsp) + movaps %xmm14, 0xc0(%rsp) + movaps %xmm15, 0xd0(%rsp) +.Lxts_dec_body: +___ +$code.=<<___; + mov %rsp, %rbp # backup %rsp + mov $arg1, $inp # backup arguments + mov $arg2, $out + mov $arg3, $len + mov $arg4, $key + + lea ($arg6), $arg1 + lea 0x20(%rbp), $arg2 + lea ($arg5), $arg3 + call asm_AES_encrypt # generate initial tweak + + mov 240($key), %eax # rounds + mov $len, %rbx # backup $len + + mov %eax, %edx # rounds + shl \$7, %rax # 128 bytes per inner round key + sub \$`128-32`, %rax # size of bit-sliced key schedule + sub %rax, %rsp + + mov %rsp, %rax # pass key schedule + mov $key, %rcx # pass key + mov %edx, %r10d # pass rounds + call _bsaes_key_convert + pxor (%rsp), %xmm7 # fix up round 0 key + movdqa %xmm6, (%rax) # save last round key + movdqa %xmm7, (%rsp) + + xor %eax, %eax # if ($len%16) len-=16; + and \$-16, $len + test \$15, %ebx + setnz %al + shl \$4, %rax + sub %rax, $len + + sub \$0x80, %rsp # place for tweak[8] + movdqa 0x20(%rbp), @XMM[7] # initial tweak + + pxor $twtmp, $twtmp + movdqa .Lxts_magic(%rip), $twmask + pcmpgtd @XMM[7], $twtmp # broadcast upper bits + + sub \$0x80, $len + jc .Lxts_dec_short + jmp .Lxts_dec_loop + +.align 16 +.Lxts_dec_loop: +___ + for ($i=0;$i<7;$i++) { + $code.=<<___; + pshufd \$0x13, $twtmp, $twres + pxor $twtmp, $twtmp + movdqa @XMM[7], @XMM[$i] + movdqa @XMM[7], `0x10*$i`(%rsp)# save tweak[$i] + paddq @XMM[7], @XMM[7] # psllq 1,$tweak + pand $twmask, $twres # isolate carry and residue + pcmpgtd @XMM[7], $twtmp # broadcast upper bits + pxor $twres, @XMM[7] +___ + $code.=<<___ if ($i>=1); + movdqu `0x10*($i-1)`($inp), @XMM[8+$i-1] +___ + $code.=<<___ if ($i>=2); + pxor @XMM[8+$i-2], @XMM[$i-2]# input[] ^ tweak[] +___ + } +$code.=<<___; + movdqu 0x60($inp), @XMM[8+6] + pxor @XMM[8+5], @XMM[5] + movdqu 0x70($inp), @XMM[8+7] + lea 0x80($inp), $inp + movdqa @XMM[7], 0x70(%rsp) + pxor @XMM[8+6], @XMM[6] + lea 0x80(%rsp), %rax # pass key schedule + pxor @XMM[8+7], @XMM[7] + mov %edx, %r10d # pass rounds + + call _bsaes_decrypt8 + + pxor 0x00(%rsp), @XMM[0] # ^= tweak[] + pxor 0x10(%rsp), @XMM[1] + movdqu @XMM[0], 0x00($out) # write output + pxor 0x20(%rsp), @XMM[6] + movdqu @XMM[1], 0x10($out) + pxor 0x30(%rsp), @XMM[4] + movdqu @XMM[6], 0x20($out) + pxor 0x40(%rsp), @XMM[2] + movdqu @XMM[4], 0x30($out) + pxor 0x50(%rsp), @XMM[7] + movdqu @XMM[2], 0x40($out) + pxor 0x60(%rsp), @XMM[3] + movdqu @XMM[7], 0x50($out) + pxor 0x70(%rsp), @XMM[5] + movdqu @XMM[3], 0x60($out) + movdqu @XMM[5], 0x70($out) + lea 0x80($out), $out + + movdqa 0x70(%rsp), @XMM[7] # prepare next iteration tweak + pxor $twtmp, $twtmp + movdqa .Lxts_magic(%rip), $twmask + pcmpgtd @XMM[7], $twtmp + pshufd \$0x13, $twtmp, $twres + pxor $twtmp, $twtmp + paddq @XMM[7], @XMM[7] # psllq 1,$tweak + pand $twmask, $twres # isolate carry and residue + pcmpgtd @XMM[7], $twtmp # broadcast upper bits + pxor $twres, @XMM[7] + + sub \$0x80,$len + jnc .Lxts_dec_loop + +.Lxts_dec_short: + add \$0x80, $len + jz .Lxts_dec_done +___ + for ($i=0;$i<7;$i++) { + $code.=<<___; + pshufd \$0x13, $twtmp, $twres + pxor $twtmp, $twtmp + movdqa @XMM[7], @XMM[$i] + movdqa @XMM[7], `0x10*$i`(%rsp)# save tweak[$i] + paddq @XMM[7], @XMM[7] # psllq 1,$tweak + pand $twmask, $twres # isolate carry and residue + pcmpgtd @XMM[7], $twtmp # broadcast upper bits + pxor $twres, @XMM[7] +___ + $code.=<<___ if ($i>=1); + movdqu `0x10*($i-1)`($inp), @XMM[8+$i-1] + cmp \$`0x10*$i`,$len + je .Lxts_dec_$i +___ + $code.=<<___ if ($i>=2); + pxor @XMM[8+$i-2], @XMM[$i-2]# input[] ^ tweak[] +___ + } +$code.=<<___; + movdqu 0x60($inp), @XMM[8+6] + pxor @XMM[8+5], @XMM[5] + movdqa @XMM[7], 0x70(%rsp) + lea 0x70($inp), $inp + pxor @XMM[8+6], @XMM[6] + lea 0x80(%rsp), %rax # pass key schedule + mov %edx, %r10d # pass rounds + + call _bsaes_decrypt8 + + pxor 0x00(%rsp), @XMM[0] # ^= tweak[] + pxor 0x10(%rsp), @XMM[1] + movdqu @XMM[0], 0x00($out) # write output + pxor 0x20(%rsp), @XMM[6] + movdqu @XMM[1], 0x10($out) + pxor 0x30(%rsp), @XMM[4] + movdqu @XMM[6], 0x20($out) + pxor 0x40(%rsp), @XMM[2] + movdqu @XMM[4], 0x30($out) + pxor 0x50(%rsp), @XMM[7] + movdqu @XMM[2], 0x40($out) + pxor 0x60(%rsp), @XMM[3] + movdqu @XMM[7], 0x50($out) + movdqu @XMM[3], 0x60($out) + lea 0x70($out), $out + + movdqa 0x70(%rsp), @XMM[7] # next iteration tweak + jmp .Lxts_dec_done +.align 16 +.Lxts_dec_6: + pxor @XMM[8+4], @XMM[4] + lea 0x60($inp), $inp + pxor @XMM[8+5], @XMM[5] + lea 0x80(%rsp), %rax # pass key schedule + mov %edx, %r10d # pass rounds + + call _bsaes_decrypt8 + + pxor 0x00(%rsp), @XMM[0] # ^= tweak[] + pxor 0x10(%rsp), @XMM[1] + movdqu @XMM[0], 0x00($out) # write output + pxor 0x20(%rsp), @XMM[6] + movdqu @XMM[1], 0x10($out) + pxor 0x30(%rsp), @XMM[4] + movdqu @XMM[6], 0x20($out) + pxor 0x40(%rsp), @XMM[2] + movdqu @XMM[4], 0x30($out) + pxor 0x50(%rsp), @XMM[7] + movdqu @XMM[2], 0x40($out) + movdqu @XMM[7], 0x50($out) + lea 0x60($out), $out + + movdqa 0x60(%rsp), @XMM[7] # next iteration tweak + jmp .Lxts_dec_done +.align 16 +.Lxts_dec_5: + pxor @XMM[8+3], @XMM[3] + lea 0x50($inp), $inp + pxor @XMM[8+4], @XMM[4] + lea 0x80(%rsp), %rax # pass key schedule + mov %edx, %r10d # pass rounds + + call _bsaes_decrypt8 + + pxor 0x00(%rsp), @XMM[0] # ^= tweak[] + pxor 0x10(%rsp), @XMM[1] + movdqu @XMM[0], 0x00($out) # write output + pxor 0x20(%rsp), @XMM[6] + movdqu @XMM[1], 0x10($out) + pxor 0x30(%rsp), @XMM[4] + movdqu @XMM[6], 0x20($out) + pxor 0x40(%rsp), @XMM[2] + movdqu @XMM[4], 0x30($out) + movdqu @XMM[2], 0x40($out) + lea 0x50($out), $out + + movdqa 0x50(%rsp), @XMM[7] # next iteration tweak + jmp .Lxts_dec_done +.align 16 +.Lxts_dec_4: + pxor @XMM[8+2], @XMM[2] + lea 0x40($inp), $inp + pxor @XMM[8+3], @XMM[3] + lea 0x80(%rsp), %rax # pass key schedule + mov %edx, %r10d # pass rounds + + call _bsaes_decrypt8 + + pxor 0x00(%rsp), @XMM[0] # ^= tweak[] + pxor 0x10(%rsp), @XMM[1] + movdqu @XMM[0], 0x00($out) # write output + pxor 0x20(%rsp), @XMM[6] + movdqu @XMM[1], 0x10($out) + pxor 0x30(%rsp), @XMM[4] + movdqu @XMM[6], 0x20($out) + movdqu @XMM[4], 0x30($out) + lea 0x40($out), $out + + movdqa 0x40(%rsp), @XMM[7] # next iteration tweak + jmp .Lxts_dec_done +.align 16 +.Lxts_dec_3: + pxor @XMM[8+1], @XMM[1] + lea 0x30($inp), $inp + pxor @XMM[8+2], @XMM[2] + lea 0x80(%rsp), %rax # pass key schedule + mov %edx, %r10d # pass rounds + + call _bsaes_decrypt8 + + pxor 0x00(%rsp), @XMM[0] # ^= tweak[] + pxor 0x10(%rsp), @XMM[1] + movdqu @XMM[0], 0x00($out) # write output + pxor 0x20(%rsp), @XMM[6] + movdqu @XMM[1], 0x10($out) + movdqu @XMM[6], 0x20($out) + lea 0x30($out), $out + + movdqa 0x30(%rsp), @XMM[7] # next iteration tweak + jmp .Lxts_dec_done +.align 16 +.Lxts_dec_2: + pxor @XMM[8+0], @XMM[0] + lea 0x20($inp), $inp + pxor @XMM[8+1], @XMM[1] + lea 0x80(%rsp), %rax # pass key schedule + mov %edx, %r10d # pass rounds + + call _bsaes_decrypt8 + + pxor 0x00(%rsp), @XMM[0] # ^= tweak[] + pxor 0x10(%rsp), @XMM[1] + movdqu @XMM[0], 0x00($out) # write output + movdqu @XMM[1], 0x10($out) + lea 0x20($out), $out + + movdqa 0x20(%rsp), @XMM[7] # next iteration tweak + jmp .Lxts_dec_done +.align 16 +.Lxts_dec_1: + pxor @XMM[0], @XMM[8] + lea 0x10($inp), $inp + movdqa @XMM[8], 0x20(%rbp) + lea 0x20(%rbp), $arg1 + lea 0x20(%rbp), $arg2 + lea ($key), $arg3 + call asm_AES_decrypt # doesn't touch %xmm + pxor 0x20(%rbp), @XMM[0] # ^= tweak[] + #pxor @XMM[8], @XMM[0] + #lea 0x80(%rsp), %rax # pass key schedule + #mov %edx, %r10d # pass rounds + #call _bsaes_decrypt8 + #pxor 0x00(%rsp), @XMM[0] # ^= tweak[] + movdqu @XMM[0], 0x00($out) # write output + lea 0x10($out), $out + + movdqa 0x10(%rsp), @XMM[7] # next iteration tweak + +.Lxts_dec_done: + and \$15, %ebx + jz .Lxts_dec_ret + + pxor $twtmp, $twtmp + movdqa .Lxts_magic(%rip), $twmask + pcmpgtd @XMM[7], $twtmp + pshufd \$0x13, $twtmp, $twres + movdqa @XMM[7], @XMM[6] + paddq @XMM[7], @XMM[7] # psllq 1,$tweak + pand $twmask, $twres # isolate carry and residue + movdqu ($inp), @XMM[0] + pxor $twres, @XMM[7] + + lea 0x20(%rbp), $arg1 + pxor @XMM[7], @XMM[0] + lea 0x20(%rbp), $arg2 + movdqa @XMM[0], 0x20(%rbp) + lea ($key), $arg3 + call asm_AES_decrypt # doesn't touch %xmm + pxor 0x20(%rbp), @XMM[7] + mov $out, %rdx + movdqu @XMM[7], ($out) + +.Lxts_dec_steal: + movzb 16($inp), %eax + movzb (%rdx), %ecx + lea 1($inp), $inp + mov %al, (%rdx) + mov %cl, 16(%rdx) + lea 1(%rdx), %rdx + sub \$1,%ebx + jnz .Lxts_dec_steal + + movdqu ($out), @XMM[0] + lea 0x20(%rbp), $arg1 + pxor @XMM[6], @XMM[0] + lea 0x20(%rbp), $arg2 + movdqa @XMM[0], 0x20(%rbp) + lea ($key), $arg3 + call asm_AES_decrypt # doesn't touch %xmm + pxor 0x20(%rbp), @XMM[6] + movdqu @XMM[6], ($out) + +.Lxts_dec_ret: + lea (%rsp), %rax + pxor %xmm0, %xmm0 +.Lxts_dec_bzero: # wipe key schedule [if any] + movdqa %xmm0, 0x00(%rax) + movdqa %xmm0, 0x10(%rax) + lea 0x20(%rax), %rax + cmp %rax, %rbp + ja .Lxts_dec_bzero + + lea (%rbp),%rsp # restore %rsp +___ +$code.=<<___ if ($win64); + movaps 0x40(%rbp), %xmm6 + movaps 0x50(%rbp), %xmm7 + movaps 0x60(%rbp), %xmm8 + movaps 0x70(%rbp), %xmm9 + movaps 0x80(%rbp), %xmm10 + movaps 0x90(%rbp), %xmm11 + movaps 0xa0(%rbp), %xmm12 + movaps 0xb0(%rbp), %xmm13 + movaps 0xc0(%rbp), %xmm14 + movaps 0xd0(%rbp), %xmm15 + lea 0xa0(%rbp), %rsp +___ +$code.=<<___; + mov 0x48(%rsp), %r15 + mov 0x50(%rsp), %r14 + mov 0x58(%rsp), %r13 + mov 0x60(%rsp), %r12 + mov 0x68(%rsp), %rbx + mov 0x70(%rsp), %rax + lea 0x78(%rsp), %rsp + mov %rax, %rbp +.Lxts_dec_epilogue: + ret +.size bsaes_xts_decrypt,.-bsaes_xts_decrypt +___ +} +$code.=<<___; +.type _bsaes_const,\@object +.align 64 +_bsaes_const: +.LM0ISR: # InvShiftRows constants + .quad 0x0a0e0206070b0f03, 0x0004080c0d010509 +.LISRM0: + .quad 0x01040b0e0205080f, 0x0306090c00070a0d +.LISR: + .quad 0x0504070602010003, 0x0f0e0d0c080b0a09 +.LBS0: # bit-slice constants + .quad 0x5555555555555555, 0x5555555555555555 +.LBS1: + .quad 0x3333333333333333, 0x3333333333333333 +.LBS2: + .quad 0x0f0f0f0f0f0f0f0f, 0x0f0f0f0f0f0f0f0f +.LSR: # shiftrows constants + .quad 0x0504070600030201, 0x0f0e0d0c0a09080b +.LSRM0: + .quad 0x0304090e00050a0f, 0x01060b0c0207080d +.LM0SR: + .quad 0x0a0e02060f03070b, 0x0004080c05090d01 +.LSWPUP: # byte-swap upper dword + .quad 0x0706050403020100, 0x0c0d0e0f0b0a0908 +.LSWPUPM0SR: + .quad 0x0a0d02060c03070b, 0x0004080f05090e01 +.LADD1: # counter increment constants + .quad 0x0000000000000000, 0x0000000100000000 +.LADD2: + .quad 0x0000000000000000, 0x0000000200000000 +.LADD3: + .quad 0x0000000000000000, 0x0000000300000000 +.LADD4: + .quad 0x0000000000000000, 0x0000000400000000 +.LADD5: + .quad 0x0000000000000000, 0x0000000500000000 +.LADD6: + .quad 0x0000000000000000, 0x0000000600000000 +.LADD7: + .quad 0x0000000000000000, 0x0000000700000000 +.LADD8: + .quad 0x0000000000000000, 0x0000000800000000 +.Lxts_magic: + .long 0x87,0,1,0 +.Lmasks: + .quad 0x0101010101010101, 0x0101010101010101 + .quad 0x0202020202020202, 0x0202020202020202 + .quad 0x0404040404040404, 0x0404040404040404 + .quad 0x0808080808080808, 0x0808080808080808 +.LM0: + .quad 0x02060a0e03070b0f, 0x0004080c0105090d +.L63: + .quad 0x6363636363636363, 0x6363636363636363 +.asciz "Bit-sliced AES for x86_64/SSSE3, Emilia Käsper, Peter Schwabe, Andy Polyakov" +.align 64 +.size _bsaes_const,.-_bsaes_const +___ + +# EXCEPTION_DISPOSITION handler (EXCEPTION_RECORD *rec,ULONG64 frame, +# CONTEXT *context,DISPATCHER_CONTEXT *disp) +if ($win64) { +$rec="%rcx"; +$frame="%rdx"; +$context="%r8"; +$disp="%r9"; + +$code.=<<___; +.extern __imp_RtlVirtualUnwind +.type se_handler,\@abi-omnipotent +.align 16 +se_handler: + push %rsi + push %rdi + push %rbx + push %rbp + push %r12 + push %r13 + push %r14 + push %r15 + pushfq + sub \$64,%rsp + + mov 120($context),%rax # pull context->Rax + mov 248($context),%rbx # pull context->Rip + + mov 8($disp),%rsi # disp->ImageBase + mov 56($disp),%r11 # disp->HandlerData + + mov 0(%r11),%r10d # HandlerData[0] + lea (%rsi,%r10),%r10 # prologue label + cmp %r10,%rbx # context->RipRsp + + mov 4(%r11),%r10d # HandlerData[1] + lea (%rsi,%r10),%r10 # epilogue label + cmp %r10,%rbx # context->Rip>=epilogue label + jae .Lin_prologue + + mov 160($context),%rax # pull context->Rbp + + lea 0x40(%rax),%rsi # %xmm save area + lea 512($context),%rdi # &context.Xmm6 + mov \$20,%ecx # 10*sizeof(%xmm0)/sizeof(%rax) + .long 0xa548f3fc # cld; rep movsq + lea 0xa0(%rax),%rax # adjust stack pointer + + mov 0x70(%rax),%rbp + mov 0x68(%rax),%rbx + mov 0x60(%rax),%r12 + mov 0x58(%rax),%r13 + mov 0x50(%rax),%r14 + mov 0x48(%rax),%r15 + lea 0x78(%rax),%rax # adjust stack pointer + mov %rbx,144($context) # restore context->Rbx + mov %rbp,160($context) # restore context->Rbp + mov %r12,216($context) # restore context->R12 + mov %r13,224($context) # restore context->R13 + mov %r14,232($context) # restore context->R14 + mov %r15,240($context) # restore context->R15 + +.Lin_prologue: + mov %rax,152($context) # restore context->Rsp + + mov 40($disp),%rdi # disp->ContextRecord + mov $context,%rsi # context + mov \$`1232/8`,%ecx # sizeof(CONTEXT) + .long 0xa548f3fc # cld; rep movsq + + mov $disp,%rsi + xor %rcx,%rcx # arg1, UNW_FLAG_NHANDLER + mov 8(%rsi),%rdx # arg2, disp->ImageBase + mov 0(%rsi),%r8 # arg3, disp->ControlPc + mov 16(%rsi),%r9 # arg4, disp->FunctionEntry + mov 40(%rsi),%r10 # disp->ContextRecord + lea 56(%rsi),%r11 # &disp->HandlerData + lea 24(%rsi),%r12 # &disp->EstablisherFrame + mov %r10,32(%rsp) # arg5 + mov %r11,40(%rsp) # arg6 + mov %r12,48(%rsp) # arg7 + mov %rcx,56(%rsp) # arg8, (NULL) + call *__imp_RtlVirtualUnwind(%rip) + + mov \$1,%eax # ExceptionContinueSearch + add \$64,%rsp + popfq + pop %r15 + pop %r14 + pop %r13 + pop %r12 + pop %rbp + pop %rbx + pop %rdi + pop %rsi + ret +.size se_handler,.-se_handler + +.section .pdata +.align 4 +___ +$code.=<<___ if ($ecb); + .rva .Lecb_enc_prologue + .rva .Lecb_enc_epilogue + .rva .Lecb_enc_info + + .rva .Lecb_dec_prologue + .rva .Lecb_dec_epilogue + .rva .Lecb_dec_info +___ +$code.=<<___; + .rva .Lcbc_dec_prologue + .rva .Lcbc_dec_epilogue + .rva .Lcbc_dec_info + + .rva .Lctr_enc_prologue + .rva .Lctr_enc_epilogue + .rva .Lctr_enc_info + + .rva .Lxts_enc_prologue + .rva .Lxts_enc_epilogue + .rva .Lxts_enc_info + + .rva .Lxts_dec_prologue + .rva .Lxts_dec_epilogue + .rva .Lxts_dec_info + +.section .xdata +.align 8 +___ +$code.=<<___ if ($ecb); +.Lecb_enc_info: + .byte 9,0,0,0 + .rva se_handler + .rva .Lecb_enc_body,.Lecb_enc_epilogue # HandlerData[] +.Lecb_dec_info: + .byte 9,0,0,0 + .rva se_handler + .rva .Lecb_dec_body,.Lecb_dec_epilogue # HandlerData[] +___ +$code.=<<___; +.Lcbc_dec_info: + .byte 9,0,0,0 + .rva se_handler + .rva .Lcbc_dec_body,.Lcbc_dec_epilogue # HandlerData[] +.Lctr_enc_info: + .byte 9,0,0,0 + .rva se_handler + .rva .Lctr_enc_body,.Lctr_enc_epilogue # HandlerData[] +.Lxts_enc_info: + .byte 9,0,0,0 + .rva se_handler + .rva .Lxts_enc_body,.Lxts_enc_epilogue # HandlerData[] +.Lxts_dec_info: + .byte 9,0,0,0 + .rva se_handler + .rva .Lxts_dec_body,.Lxts_dec_epilogue # HandlerData[] +___ +} + +$code =~ s/\`([^\`]*)\`/eval($1)/gem; + +print $code; + +close STDOUT; diff --git a/openssl-1.1.0h/crypto/aes/asm/vpaes-armv8.pl b/openssl-1.1.0h/crypto/aes/asm/vpaes-armv8.pl new file mode 100755 index 0000000..d6b5f56 --- /dev/null +++ b/openssl-1.1.0h/crypto/aes/asm/vpaes-armv8.pl @@ -0,0 +1,1259 @@ +#! /usr/bin/env perl +# Copyright 2015-2016 The OpenSSL Project Authors. All Rights Reserved. +# +# Licensed under the OpenSSL license (the "License"). You may not use +# this file except in compliance with the License. You can obtain a copy +# in the file LICENSE in the source distribution or at +# https://www.openssl.org/source/license.html + + +###################################################################### +## Constant-time SSSE3 AES core implementation. +## version 0.1 +## +## By Mike Hamburg (Stanford University), 2009 +## Public domain. +## +## For details see http://shiftleft.org/papers/vector_aes/ and +## http://crypto.stanford.edu/vpaes/. +## +###################################################################### +# ARMv8 NEON adaptation by +# +# Reason for undertaken effort is that there is at least one popular +# SoC based on Cortex-A53 that doesn't have crypto extensions. +# +# CBC enc ECB enc/dec(*) [bit-sliced enc/dec] +# Cortex-A53 21.5 18.1/20.6 [17.5/19.8 ] +# Cortex-A57 36.0(**) 20.4/24.9(**) [14.4/16.6 ] +# X-Gene 45.9(**) 45.8/57.7(**) [33.1/37.6(**) ] +# Denver(***) 16.6(**) 15.1/17.8(**) [8.80/9.93 ] +# Apple A7(***) 22.7(**) 10.9/14.3 [8.45/10.0 ] +# Mongoose(***) 26.3(**) 21.0/25.0(**) [13.3/16.8 ] +# +# (*) ECB denotes approximate result for parallelizeable modes +# such as CBC decrypt, CTR, etc.; +# (**) these results are worse than scalar compiler-generated +# code, but it's constant-time and therefore preferred; +# (***) presented for reference/comparison purposes; + +$flavour = shift; +while (($output=shift) && ($output!~/\w[\w\-]*\.\w+$/)) {} + +$0 =~ m/(.*[\/\\])[^\/\\]+$/; $dir=$1; +( $xlate="${dir}arm-xlate.pl" and -f $xlate ) or +( $xlate="${dir}../../perlasm/arm-xlate.pl" and -f $xlate) or +die "can't locate arm-xlate.pl"; + +open OUT,"| \"$^X\" $xlate $flavour $output"; +*STDOUT=*OUT; + +$code.=<<___; +.text + +.type _vpaes_consts,%object +.align 7 // totally strategic alignment +_vpaes_consts: +.Lk_mc_forward: // mc_forward + .quad 0x0407060500030201, 0x0C0F0E0D080B0A09 + .quad 0x080B0A0904070605, 0x000302010C0F0E0D + .quad 0x0C0F0E0D080B0A09, 0x0407060500030201 + .quad 0x000302010C0F0E0D, 0x080B0A0904070605 +.Lk_mc_backward:// mc_backward + .quad 0x0605040702010003, 0x0E0D0C0F0A09080B + .quad 0x020100030E0D0C0F, 0x0A09080B06050407 + .quad 0x0E0D0C0F0A09080B, 0x0605040702010003 + .quad 0x0A09080B06050407, 0x020100030E0D0C0F +.Lk_sr: // sr + .quad 0x0706050403020100, 0x0F0E0D0C0B0A0908 + .quad 0x030E09040F0A0500, 0x0B06010C07020D08 + .quad 0x0F060D040B020900, 0x070E050C030A0108 + .quad 0x0B0E0104070A0D00, 0x0306090C0F020508 + +// +// "Hot" constants +// +.Lk_inv: // inv, inva + .quad 0x0E05060F0D080180, 0x040703090A0B0C02 + .quad 0x01040A060F0B0780, 0x030D0E0C02050809 +.Lk_ipt: // input transform (lo, hi) + .quad 0xC2B2E8985A2A7000, 0xCABAE09052227808 + .quad 0x4C01307D317C4D00, 0xCD80B1FCB0FDCC81 +.Lk_sbo: // sbou, sbot + .quad 0xD0D26D176FBDC700, 0x15AABF7AC502A878 + .quad 0xCFE474A55FBB6A00, 0x8E1E90D1412B35FA +.Lk_sb1: // sb1u, sb1t + .quad 0x3618D415FAE22300, 0x3BF7CCC10D2ED9EF + .quad 0xB19BE18FCB503E00, 0xA5DF7A6E142AF544 +.Lk_sb2: // sb2u, sb2t + .quad 0x69EB88400AE12900, 0xC2A163C8AB82234A + .quad 0xE27A93C60B712400, 0x5EB7E955BC982FCD + +// +// Decryption stuff +// +.Lk_dipt: // decryption input transform + .quad 0x0F505B040B545F00, 0x154A411E114E451A + .quad 0x86E383E660056500, 0x12771772F491F194 +.Lk_dsbo: // decryption sbox final output + .quad 0x1387EA537EF94000, 0xC7AA6DB9D4943E2D + .quad 0x12D7560F93441D00, 0xCA4B8159D8C58E9C +.Lk_dsb9: // decryption sbox output *9*u, *9*t + .quad 0x851C03539A86D600, 0xCAD51F504F994CC9 + .quad 0xC03B1789ECD74900, 0x725E2C9EB2FBA565 +.Lk_dsbd: // decryption sbox output *D*u, *D*t + .quad 0x7D57CCDFE6B1A200, 0xF56E9B13882A4439 + .quad 0x3CE2FAF724C6CB00, 0x2931180D15DEEFD3 +.Lk_dsbb: // decryption sbox output *B*u, *B*t + .quad 0xD022649296B44200, 0x602646F6B0F2D404 + .quad 0xC19498A6CD596700, 0xF3FF0C3E3255AA6B +.Lk_dsbe: // decryption sbox output *E*u, *E*t + .quad 0x46F2929626D4D000, 0x2242600464B4F6B0 + .quad 0x0C55A6CDFFAAC100, 0x9467F36B98593E32 + +// +// Key schedule constants +// +.Lk_dksd: // decryption key schedule: invskew x*D + .quad 0xFEB91A5DA3E44700, 0x0740E3A45A1DBEF9 + .quad 0x41C277F4B5368300, 0x5FDC69EAAB289D1E +.Lk_dksb: // decryption key schedule: invskew x*B + .quad 0x9A4FCA1F8550D500, 0x03D653861CC94C99 + .quad 0x115BEDA7B6FC4A00, 0xD993256F7E3482C8 +.Lk_dkse: // decryption key schedule: invskew x*E + 0x63 + .quad 0xD5031CCA1FC9D600, 0x53859A4C994F5086 + .quad 0xA23196054FDC7BE8, 0xCD5EF96A20B31487 +.Lk_dks9: // decryption key schedule: invskew x*9 + .quad 0xB6116FC87ED9A700, 0x4AED933482255BFC + .quad 0x4576516227143300, 0x8BB89FACE9DAFDCE + +.Lk_rcon: // rcon + .quad 0x1F8391B9AF9DEEB6, 0x702A98084D7C7D81 + +.Lk_opt: // output transform + .quad 0xFF9F4929D6B66000, 0xF7974121DEBE6808 + .quad 0x01EDBD5150BCEC00, 0xE10D5DB1B05C0CE0 +.Lk_deskew: // deskew tables: inverts the sbox's "skew" + .quad 0x07E4A34047A4E300, 0x1DFEB95A5DBEF91A + .quad 0x5F36B5DC83EA6900, 0x2841C2ABF49D1E77 + +.asciz "Vector Permutaion AES for ARMv8, Mike Hamburg (Stanford University)" +.size _vpaes_consts,.-_vpaes_consts +.align 6 +___ + +{ +my ($inp,$out,$key) = map("x$_",(0..2)); + +my ($invlo,$invhi,$iptlo,$ipthi,$sbou,$sbot) = map("v$_.16b",(18..23)); +my ($sb1u,$sb1t,$sb2u,$sb2t) = map("v$_.16b",(24..27)); +my ($sb9u,$sb9t,$sbdu,$sbdt,$sbbu,$sbbt,$sbeu,$sbet)=map("v$_.16b",(24..31)); + +$code.=<<___; +## +## _aes_preheat +## +## Fills register %r10 -> .aes_consts (so you can -fPIC) +## and %xmm9-%xmm15 as specified below. +## +.type _vpaes_encrypt_preheat,%function +.align 4 +_vpaes_encrypt_preheat: + adr x10, .Lk_inv + movi v17.16b, #0x0f + ld1 {v18.2d-v19.2d}, [x10],#32 // .Lk_inv + ld1 {v20.2d-v23.2d}, [x10],#64 // .Lk_ipt, .Lk_sbo + ld1 {v24.2d-v27.2d}, [x10] // .Lk_sb1, .Lk_sb2 + ret +.size _vpaes_encrypt_preheat,.-_vpaes_encrypt_preheat + +## +## _aes_encrypt_core +## +## AES-encrypt %xmm0. +## +## Inputs: +## %xmm0 = input +## %xmm9-%xmm15 as in _vpaes_preheat +## (%rdx) = scheduled keys +## +## Output in %xmm0 +## Clobbers %xmm1-%xmm5, %r9, %r10, %r11, %rax +## Preserves %xmm6 - %xmm8 so you get some local vectors +## +## +.type _vpaes_encrypt_core,%function +.align 4 +_vpaes_encrypt_core: + mov x9, $key + ldr w8, [$key,#240] // pull rounds + adr x11, .Lk_mc_forward+16 + // vmovdqa .Lk_ipt(%rip), %xmm2 # iptlo + ld1 {v16.2d}, [x9], #16 // vmovdqu (%r9), %xmm5 # round0 key + and v1.16b, v7.16b, v17.16b // vpand %xmm9, %xmm0, %xmm1 + ushr v0.16b, v7.16b, #4 // vpsrlb \$4, %xmm0, %xmm0 + tbl v1.16b, {$iptlo}, v1.16b // vpshufb %xmm1, %xmm2, %xmm1 + // vmovdqa .Lk_ipt+16(%rip), %xmm3 # ipthi + tbl v2.16b, {$ipthi}, v0.16b // vpshufb %xmm0, %xmm3, %xmm2 + eor v0.16b, v1.16b, v16.16b // vpxor %xmm5, %xmm1, %xmm0 + eor v0.16b, v0.16b, v2.16b // vpxor %xmm2, %xmm0, %xmm0 + b .Lenc_entry + +.align 4 +.Lenc_loop: + // middle of middle round + add x10, x11, #0x40 + tbl v4.16b, {$sb1t}, v2.16b // vpshufb %xmm2, %xmm13, %xmm4 # 4 = sb1u + ld1 {v1.2d}, [x11], #16 // vmovdqa -0x40(%r11,%r10), %xmm1 # .Lk_mc_forward[] + tbl v0.16b, {$sb1u}, v3.16b // vpshufb %xmm3, %xmm12, %xmm0 # 0 = sb1t + eor v4.16b, v4.16b, v16.16b // vpxor %xmm5, %xmm4, %xmm4 # 4 = sb1u + k + tbl v5.16b, {$sb2t}, v2.16b // vpshufb %xmm2, %xmm15, %xmm5 # 4 = sb2u + eor v0.16b, v0.16b, v4.16b // vpxor %xmm4, %xmm0, %xmm0 # 0 = A + tbl v2.16b, {$sb2u}, v3.16b // vpshufb %xmm3, %xmm14, %xmm2 # 2 = sb2t + ld1 {v4.2d}, [x10] // vmovdqa (%r11,%r10), %xmm4 # .Lk_mc_backward[] + tbl v3.16b, {v0.16b}, v1.16b // vpshufb %xmm1, %xmm0, %xmm3 # 0 = B + eor v2.16b, v2.16b, v5.16b // vpxor %xmm5, %xmm2, %xmm2 # 2 = 2A + tbl v0.16b, {v0.16b}, v4.16b // vpshufb %xmm4, %xmm0, %xmm0 # 3 = D + eor v3.16b, v3.16b, v2.16b // vpxor %xmm2, %xmm3, %xmm3 # 0 = 2A+B + tbl v4.16b, {v3.16b}, v1.16b // vpshufb %xmm1, %xmm3, %xmm4 # 0 = 2B+C + eor v0.16b, v0.16b, v3.16b // vpxor %xmm3, %xmm0, %xmm0 # 3 = 2A+B+D + and x11, x11, #~(1<<6) // and \$0x30, %r11 # ... mod 4 + eor v0.16b, v0.16b, v4.16b // vpxor %xmm4, %xmm0, %xmm0 # 0 = 2A+3B+C+D + sub w8, w8, #1 // nr-- + +.Lenc_entry: + // top of round + and v1.16b, v0.16b, v17.16b // vpand %xmm0, %xmm9, %xmm1 # 0 = k + ushr v0.16b, v0.16b, #4 // vpsrlb \$4, %xmm0, %xmm0 # 1 = i + tbl v5.16b, {$invhi}, v1.16b // vpshufb %xmm1, %xmm11, %xmm5 # 2 = a/k + eor v1.16b, v1.16b, v0.16b // vpxor %xmm0, %xmm1, %xmm1 # 0 = j + tbl v3.16b, {$invlo}, v0.16b // vpshufb %xmm0, %xmm10, %xmm3 # 3 = 1/i + tbl v4.16b, {$invlo}, v1.16b // vpshufb %xmm1, %xmm10, %xmm4 # 4 = 1/j + eor v3.16b, v3.16b, v5.16b // vpxor %xmm5, %xmm3, %xmm3 # 3 = iak = 1/i + a/k + eor v4.16b, v4.16b, v5.16b // vpxor %xmm5, %xmm4, %xmm4 # 4 = jak = 1/j + a/k + tbl v2.16b, {$invlo}, v3.16b // vpshufb %xmm3, %xmm10, %xmm2 # 2 = 1/iak + tbl v3.16b, {$invlo}, v4.16b // vpshufb %xmm4, %xmm10, %xmm3 # 3 = 1/jak + eor v2.16b, v2.16b, v1.16b // vpxor %xmm1, %xmm2, %xmm2 # 2 = io + eor v3.16b, v3.16b, v0.16b // vpxor %xmm0, %xmm3, %xmm3 # 3 = jo + ld1 {v16.2d}, [x9],#16 // vmovdqu (%r9), %xmm5 + cbnz w8, .Lenc_loop + + // middle of last round + add x10, x11, #0x80 + // vmovdqa -0x60(%r10), %xmm4 # 3 : sbou .Lk_sbo + // vmovdqa -0x50(%r10), %xmm0 # 0 : sbot .Lk_sbo+16 + tbl v4.16b, {$sbou}, v2.16b // vpshufb %xmm2, %xmm4, %xmm4 # 4 = sbou + ld1 {v1.2d}, [x10] // vmovdqa 0x40(%r11,%r10), %xmm1 # .Lk_sr[] + tbl v0.16b, {$sbot}, v3.16b // vpshufb %xmm3, %xmm0, %xmm0 # 0 = sb1t + eor v4.16b, v4.16b, v16.16b // vpxor %xmm5, %xmm4, %xmm4 # 4 = sb1u + k + eor v0.16b, v0.16b, v4.16b // vpxor %xmm4, %xmm0, %xmm0 # 0 = A + tbl v0.16b, {v0.16b}, v1.16b // vpshufb %xmm1, %xmm0, %xmm0 + ret +.size _vpaes_encrypt_core,.-_vpaes_encrypt_core + +.globl vpaes_encrypt +.type vpaes_encrypt,%function +.align 4 +vpaes_encrypt: + stp x29,x30,[sp,#-16]! + add x29,sp,#0 + + ld1 {v7.16b}, [$inp] + bl _vpaes_encrypt_preheat + bl _vpaes_encrypt_core + st1 {v0.16b}, [$out] + + ldp x29,x30,[sp],#16 + ret +.size vpaes_encrypt,.-vpaes_encrypt + +.type _vpaes_encrypt_2x,%function +.align 4 +_vpaes_encrypt_2x: + mov x9, $key + ldr w8, [$key,#240] // pull rounds + adr x11, .Lk_mc_forward+16 + // vmovdqa .Lk_ipt(%rip), %xmm2 # iptlo + ld1 {v16.2d}, [x9], #16 // vmovdqu (%r9), %xmm5 # round0 key + and v1.16b, v14.16b, v17.16b // vpand %xmm9, %xmm0, %xmm1 + ushr v0.16b, v14.16b, #4 // vpsrlb \$4, %xmm0, %xmm0 + and v9.16b, v15.16b, v17.16b + ushr v8.16b, v15.16b, #4 + tbl v1.16b, {$iptlo}, v1.16b // vpshufb %xmm1, %xmm2, %xmm1 + tbl v9.16b, {$iptlo}, v9.16b + // vmovdqa .Lk_ipt+16(%rip), %xmm3 # ipthi + tbl v2.16b, {$ipthi}, v0.16b // vpshufb %xmm0, %xmm3, %xmm2 + tbl v10.16b, {$ipthi}, v8.16b + eor v0.16b, v1.16b, v16.16b // vpxor %xmm5, %xmm1, %xmm0 + eor v8.16b, v9.16b, v16.16b + eor v0.16b, v0.16b, v2.16b // vpxor %xmm2, %xmm0, %xmm0 + eor v8.16b, v8.16b, v10.16b + b .Lenc_2x_entry + +.align 4 +.Lenc_2x_loop: + // middle of middle round + add x10, x11, #0x40 + tbl v4.16b, {$sb1t}, v2.16b // vpshufb %xmm2, %xmm13, %xmm4 # 4 = sb1u + tbl v12.16b, {$sb1t}, v10.16b + ld1 {v1.2d}, [x11], #16 // vmovdqa -0x40(%r11,%r10), %xmm1 # .Lk_mc_forward[] + tbl v0.16b, {$sb1u}, v3.16b // vpshufb %xmm3, %xmm12, %xmm0 # 0 = sb1t + tbl v8.16b, {$sb1u}, v11.16b + eor v4.16b, v4.16b, v16.16b // vpxor %xmm5, %xmm4, %xmm4 # 4 = sb1u + k + eor v12.16b, v12.16b, v16.16b + tbl v5.16b, {$sb2t}, v2.16b // vpshufb %xmm2, %xmm15, %xmm5 # 4 = sb2u + tbl v13.16b, {$sb2t}, v10.16b + eor v0.16b, v0.16b, v4.16b // vpxor %xmm4, %xmm0, %xmm0 # 0 = A + eor v8.16b, v8.16b, v12.16b + tbl v2.16b, {$sb2u}, v3.16b // vpshufb %xmm3, %xmm14, %xmm2 # 2 = sb2t + tbl v10.16b, {$sb2u}, v11.16b + ld1 {v4.2d}, [x10] // vmovdqa (%r11,%r10), %xmm4 # .Lk_mc_backward[] + tbl v3.16b, {v0.16b}, v1.16b // vpshufb %xmm1, %xmm0, %xmm3 # 0 = B + tbl v11.16b, {v8.16b}, v1.16b + eor v2.16b, v2.16b, v5.16b // vpxor %xmm5, %xmm2, %xmm2 # 2 = 2A + eor v10.16b, v10.16b, v13.16b + tbl v0.16b, {v0.16b}, v4.16b // vpshufb %xmm4, %xmm0, %xmm0 # 3 = D + tbl v8.16b, {v8.16b}, v4.16b + eor v3.16b, v3.16b, v2.16b // vpxor %xmm2, %xmm3, %xmm3 # 0 = 2A+B + eor v11.16b, v11.16b, v10.16b + tbl v4.16b, {v3.16b}, v1.16b // vpshufb %xmm1, %xmm3, %xmm4 # 0 = 2B+C + tbl v12.16b, {v11.16b},v1.16b + eor v0.16b, v0.16b, v3.16b // vpxor %xmm3, %xmm0, %xmm0 # 3 = 2A+B+D + eor v8.16b, v8.16b, v11.16b + and x11, x11, #~(1<<6) // and \$0x30, %r11 # ... mod 4 + eor v0.16b, v0.16b, v4.16b // vpxor %xmm4, %xmm0, %xmm0 # 0 = 2A+3B+C+D + eor v8.16b, v8.16b, v12.16b + sub w8, w8, #1 // nr-- + +.Lenc_2x_entry: + // top of round + and v1.16b, v0.16b, v17.16b // vpand %xmm0, %xmm9, %xmm1 # 0 = k + ushr v0.16b, v0.16b, #4 // vpsrlb \$4, %xmm0, %xmm0 # 1 = i + and v9.16b, v8.16b, v17.16b + ushr v8.16b, v8.16b, #4 + tbl v5.16b, {$invhi},v1.16b // vpshufb %xmm1, %xmm11, %xmm5 # 2 = a/k + tbl v13.16b, {$invhi},v9.16b + eor v1.16b, v1.16b, v0.16b // vpxor %xmm0, %xmm1, %xmm1 # 0 = j + eor v9.16b, v9.16b, v8.16b + tbl v3.16b, {$invlo},v0.16b // vpshufb %xmm0, %xmm10, %xmm3 # 3 = 1/i + tbl v11.16b, {$invlo},v8.16b + tbl v4.16b, {$invlo},v1.16b // vpshufb %xmm1, %xmm10, %xmm4 # 4 = 1/j + tbl v12.16b, {$invlo},v9.16b + eor v3.16b, v3.16b, v5.16b // vpxor %xmm5, %xmm3, %xmm3 # 3 = iak = 1/i + a/k + eor v11.16b, v11.16b, v13.16b + eor v4.16b, v4.16b, v5.16b // vpxor %xmm5, %xmm4, %xmm4 # 4 = jak = 1/j + a/k + eor v12.16b, v12.16b, v13.16b + tbl v2.16b, {$invlo},v3.16b // vpshufb %xmm3, %xmm10, %xmm2 # 2 = 1/iak + tbl v10.16b, {$invlo},v11.16b + tbl v3.16b, {$invlo},v4.16b // vpshufb %xmm4, %xmm10, %xmm3 # 3 = 1/jak + tbl v11.16b, {$invlo},v12.16b + eor v2.16b, v2.16b, v1.16b // vpxor %xmm1, %xmm2, %xmm2 # 2 = io + eor v10.16b, v10.16b, v9.16b + eor v3.16b, v3.16b, v0.16b // vpxor %xmm0, %xmm3, %xmm3 # 3 = jo + eor v11.16b, v11.16b, v8.16b + ld1 {v16.2d}, [x9],#16 // vmovdqu (%r9), %xmm5 + cbnz w8, .Lenc_2x_loop + + // middle of last round + add x10, x11, #0x80 + // vmovdqa -0x60(%r10), %xmm4 # 3 : sbou .Lk_sbo + // vmovdqa -0x50(%r10), %xmm0 # 0 : sbot .Lk_sbo+16 + tbl v4.16b, {$sbou}, v2.16b // vpshufb %xmm2, %xmm4, %xmm4 # 4 = sbou + tbl v12.16b, {$sbou}, v10.16b + ld1 {v1.2d}, [x10] // vmovdqa 0x40(%r11,%r10), %xmm1 # .Lk_sr[] + tbl v0.16b, {$sbot}, v3.16b // vpshufb %xmm3, %xmm0, %xmm0 # 0 = sb1t + tbl v8.16b, {$sbot}, v11.16b + eor v4.16b, v4.16b, v16.16b // vpxor %xmm5, %xmm4, %xmm4 # 4 = sb1u + k + eor v12.16b, v12.16b, v16.16b + eor v0.16b, v0.16b, v4.16b // vpxor %xmm4, %xmm0, %xmm0 # 0 = A + eor v8.16b, v8.16b, v12.16b + tbl v0.16b, {v0.16b},v1.16b // vpshufb %xmm1, %xmm0, %xmm0 + tbl v1.16b, {v8.16b},v1.16b + ret +.size _vpaes_encrypt_2x,.-_vpaes_encrypt_2x + +.type _vpaes_decrypt_preheat,%function +.align 4 +_vpaes_decrypt_preheat: + adr x10, .Lk_inv + movi v17.16b, #0x0f + adr x11, .Lk_dipt + ld1 {v18.2d-v19.2d}, [x10],#32 // .Lk_inv + ld1 {v20.2d-v23.2d}, [x11],#64 // .Lk_dipt, .Lk_dsbo + ld1 {v24.2d-v27.2d}, [x11],#64 // .Lk_dsb9, .Lk_dsbd + ld1 {v28.2d-v31.2d}, [x11] // .Lk_dsbb, .Lk_dsbe + ret +.size _vpaes_decrypt_preheat,.-_vpaes_decrypt_preheat + +## +## Decryption core +## +## Same API as encryption core. +## +.type _vpaes_decrypt_core,%function +.align 4 +_vpaes_decrypt_core: + mov x9, $key + ldr w8, [$key,#240] // pull rounds + + // vmovdqa .Lk_dipt(%rip), %xmm2 # iptlo + lsl x11, x8, #4 // mov %rax, %r11; shl \$4, %r11 + eor x11, x11, #0x30 // xor \$0x30, %r11 + adr x10, .Lk_sr + and x11, x11, #0x30 // and \$0x30, %r11 + add x11, x11, x10 + adr x10, .Lk_mc_forward+48 + + ld1 {v16.2d}, [x9],#16 // vmovdqu (%r9), %xmm4 # round0 key + and v1.16b, v7.16b, v17.16b // vpand %xmm9, %xmm0, %xmm1 + ushr v0.16b, v7.16b, #4 // vpsrlb \$4, %xmm0, %xmm0 + tbl v2.16b, {$iptlo}, v1.16b // vpshufb %xmm1, %xmm2, %xmm2 + ld1 {v5.2d}, [x10] // vmovdqa .Lk_mc_forward+48(%rip), %xmm5 + // vmovdqa .Lk_dipt+16(%rip), %xmm1 # ipthi + tbl v0.16b, {$ipthi}, v0.16b // vpshufb %xmm0, %xmm1, %xmm0 + eor v2.16b, v2.16b, v16.16b // vpxor %xmm4, %xmm2, %xmm2 + eor v0.16b, v0.16b, v2.16b // vpxor %xmm2, %xmm0, %xmm0 + b .Ldec_entry + +.align 4 +.Ldec_loop: +// +// Inverse mix columns +// + // vmovdqa -0x20(%r10),%xmm4 # 4 : sb9u + // vmovdqa -0x10(%r10),%xmm1 # 0 : sb9t + tbl v4.16b, {$sb9u}, v2.16b // vpshufb %xmm2, %xmm4, %xmm4 # 4 = sb9u + tbl v1.16b, {$sb9t}, v3.16b // vpshufb %xmm3, %xmm1, %xmm1 # 0 = sb9t + eor v0.16b, v4.16b, v16.16b // vpxor %xmm4, %xmm0, %xmm0 + // vmovdqa 0x00(%r10),%xmm4 # 4 : sbdu + eor v0.16b, v0.16b, v1.16b // vpxor %xmm1, %xmm0, %xmm0 # 0 = ch + // vmovdqa 0x10(%r10),%xmm1 # 0 : sbdt + + tbl v4.16b, {$sbdu}, v2.16b // vpshufb %xmm2, %xmm4, %xmm4 # 4 = sbdu + tbl v0.16b, {v0.16b}, v5.16b // vpshufb %xmm5, %xmm0, %xmm0 # MC ch + tbl v1.16b, {$sbdt}, v3.16b // vpshufb %xmm3, %xmm1, %xmm1 # 0 = sbdt + eor v0.16b, v0.16b, v4.16b // vpxor %xmm4, %xmm0, %xmm0 # 4 = ch + // vmovdqa 0x20(%r10), %xmm4 # 4 : sbbu + eor v0.16b, v0.16b, v1.16b // vpxor %xmm1, %xmm0, %xmm0 # 0 = ch + // vmovdqa 0x30(%r10), %xmm1 # 0 : sbbt + + tbl v4.16b, {$sbbu}, v2.16b // vpshufb %xmm2, %xmm4, %xmm4 # 4 = sbbu + tbl v0.16b, {v0.16b}, v5.16b // vpshufb %xmm5, %xmm0, %xmm0 # MC ch + tbl v1.16b, {$sbbt}, v3.16b // vpshufb %xmm3, %xmm1, %xmm1 # 0 = sbbt + eor v0.16b, v0.16b, v4.16b // vpxor %xmm4, %xmm0, %xmm0 # 4 = ch + // vmovdqa 0x40(%r10), %xmm4 # 4 : sbeu + eor v0.16b, v0.16b, v1.16b // vpxor %xmm1, %xmm0, %xmm0 # 0 = ch + // vmovdqa 0x50(%r10), %xmm1 # 0 : sbet + + tbl v4.16b, {$sbeu}, v2.16b // vpshufb %xmm2, %xmm4, %xmm4 # 4 = sbeu + tbl v0.16b, {v0.16b}, v5.16b // vpshufb %xmm5, %xmm0, %xmm0 # MC ch + tbl v1.16b, {$sbet}, v3.16b // vpshufb %xmm3, %xmm1, %xmm1 # 0 = sbet + eor v0.16b, v0.16b, v4.16b // vpxor %xmm4, %xmm0, %xmm0 # 4 = ch + ext v5.16b, v5.16b, v5.16b, #12 // vpalignr \$12, %xmm5, %xmm5, %xmm5 + eor v0.16b, v0.16b, v1.16b // vpxor %xmm1, %xmm0, %xmm0 # 0 = ch + sub w8, w8, #1 // sub \$1,%rax # nr-- + +.Ldec_entry: + // top of round + and v1.16b, v0.16b, v17.16b // vpand %xmm9, %xmm0, %xmm1 # 0 = k + ushr v0.16b, v0.16b, #4 // vpsrlb \$4, %xmm0, %xmm0 # 1 = i + tbl v2.16b, {$invhi}, v1.16b // vpshufb %xmm1, %xmm11, %xmm2 # 2 = a/k + eor v1.16b, v1.16b, v0.16b // vpxor %xmm0, %xmm1, %xmm1 # 0 = j + tbl v3.16b, {$invlo}, v0.16b // vpshufb %xmm0, %xmm10, %xmm3 # 3 = 1/i + tbl v4.16b, {$invlo}, v1.16b // vpshufb %xmm1, %xmm10, %xmm4 # 4 = 1/j + eor v3.16b, v3.16b, v2.16b // vpxor %xmm2, %xmm3, %xmm3 # 3 = iak = 1/i + a/k + eor v4.16b, v4.16b, v2.16b // vpxor %xmm2, %xmm4, %xmm4 # 4 = jak = 1/j + a/k + tbl v2.16b, {$invlo}, v3.16b // vpshufb %xmm3, %xmm10, %xmm2 # 2 = 1/iak + tbl v3.16b, {$invlo}, v4.16b // vpshufb %xmm4, %xmm10, %xmm3 # 3 = 1/jak + eor v2.16b, v2.16b, v1.16b // vpxor %xmm1, %xmm2, %xmm2 # 2 = io + eor v3.16b, v3.16b, v0.16b // vpxor %xmm0, %xmm3, %xmm3 # 3 = jo + ld1 {v16.2d}, [x9],#16 // vmovdqu (%r9), %xmm0 + cbnz w8, .Ldec_loop + + // middle of last round + // vmovdqa 0x60(%r10), %xmm4 # 3 : sbou + tbl v4.16b, {$sbou}, v2.16b // vpshufb %xmm2, %xmm4, %xmm4 # 4 = sbou + // vmovdqa 0x70(%r10), %xmm1 # 0 : sbot + ld1 {v2.2d}, [x11] // vmovdqa -0x160(%r11), %xmm2 # .Lk_sr-.Lk_dsbd=-0x160 + tbl v1.16b, {$sbot}, v3.16b // vpshufb %xmm3, %xmm1, %xmm1 # 0 = sb1t + eor v4.16b, v4.16b, v16.16b // vpxor %xmm0, %xmm4, %xmm4 # 4 = sb1u + k + eor v0.16b, v1.16b, v4.16b // vpxor %xmm4, %xmm1, %xmm0 # 0 = A + tbl v0.16b, {v0.16b}, v2.16b // vpshufb %xmm2, %xmm0, %xmm0 + ret +.size _vpaes_decrypt_core,.-_vpaes_decrypt_core + +.globl vpaes_decrypt +.type vpaes_decrypt,%function +.align 4 +vpaes_decrypt: + stp x29,x30,[sp,#-16]! + add x29,sp,#0 + + ld1 {v7.16b}, [$inp] + bl _vpaes_decrypt_preheat + bl _vpaes_decrypt_core + st1 {v0.16b}, [$out] + + ldp x29,x30,[sp],#16 + ret +.size vpaes_decrypt,.-vpaes_decrypt + +// v14-v15 input, v0-v1 output +.type _vpaes_decrypt_2x,%function +.align 4 +_vpaes_decrypt_2x: + mov x9, $key + ldr w8, [$key,#240] // pull rounds + + // vmovdqa .Lk_dipt(%rip), %xmm2 # iptlo + lsl x11, x8, #4 // mov %rax, %r11; shl \$4, %r11 + eor x11, x11, #0x30 // xor \$0x30, %r11 + adr x10, .Lk_sr + and x11, x11, #0x30 // and \$0x30, %r11 + add x11, x11, x10 + adr x10, .Lk_mc_forward+48 + + ld1 {v16.2d}, [x9],#16 // vmovdqu (%r9), %xmm4 # round0 key + and v1.16b, v14.16b, v17.16b // vpand %xmm9, %xmm0, %xmm1 + ushr v0.16b, v14.16b, #4 // vpsrlb \$4, %xmm0, %xmm0 + and v9.16b, v15.16b, v17.16b + ushr v8.16b, v15.16b, #4 + tbl v2.16b, {$iptlo},v1.16b // vpshufb %xmm1, %xmm2, %xmm2 + tbl v10.16b, {$iptlo},v9.16b + ld1 {v5.2d}, [x10] // vmovdqa .Lk_mc_forward+48(%rip), %xmm5 + // vmovdqa .Lk_dipt+16(%rip), %xmm1 # ipthi + tbl v0.16b, {$ipthi},v0.16b // vpshufb %xmm0, %xmm1, %xmm0 + tbl v8.16b, {$ipthi},v8.16b + eor v2.16b, v2.16b, v16.16b // vpxor %xmm4, %xmm2, %xmm2 + eor v10.16b, v10.16b, v16.16b + eor v0.16b, v0.16b, v2.16b // vpxor %xmm2, %xmm0, %xmm0 + eor v8.16b, v8.16b, v10.16b + b .Ldec_2x_entry + +.align 4 +.Ldec_2x_loop: +// +// Inverse mix columns +// + // vmovdqa -0x20(%r10),%xmm4 # 4 : sb9u + // vmovdqa -0x10(%r10),%xmm1 # 0 : sb9t + tbl v4.16b, {$sb9u}, v2.16b // vpshufb %xmm2, %xmm4, %xmm4 # 4 = sb9u + tbl v12.16b, {$sb9u}, v10.16b + tbl v1.16b, {$sb9t}, v3.16b // vpshufb %xmm3, %xmm1, %xmm1 # 0 = sb9t + tbl v9.16b, {$sb9t}, v11.16b + eor v0.16b, v4.16b, v16.16b // vpxor %xmm4, %xmm0, %xmm0 + eor v8.16b, v12.16b, v16.16b + // vmovdqa 0x00(%r10),%xmm4 # 4 : sbdu + eor v0.16b, v0.16b, v1.16b // vpxor %xmm1, %xmm0, %xmm0 # 0 = ch + eor v8.16b, v8.16b, v9.16b // vpxor %xmm1, %xmm0, %xmm0 # 0 = ch + // vmovdqa 0x10(%r10),%xmm1 # 0 : sbdt + + tbl v4.16b, {$sbdu}, v2.16b // vpshufb %xmm2, %xmm4, %xmm4 # 4 = sbdu + tbl v12.16b, {$sbdu}, v10.16b + tbl v0.16b, {v0.16b},v5.16b // vpshufb %xmm5, %xmm0, %xmm0 # MC ch + tbl v8.16b, {v8.16b},v5.16b + tbl v1.16b, {$sbdt}, v3.16b // vpshufb %xmm3, %xmm1, %xmm1 # 0 = sbdt + tbl v9.16b, {$sbdt}, v11.16b + eor v0.16b, v0.16b, v4.16b // vpxor %xmm4, %xmm0, %xmm0 # 4 = ch + eor v8.16b, v8.16b, v12.16b + // vmovdqa 0x20(%r10), %xmm4 # 4 : sbbu + eor v0.16b, v0.16b, v1.16b // vpxor %xmm1, %xmm0, %xmm0 # 0 = ch + eor v8.16b, v8.16b, v9.16b + // vmovdqa 0x30(%r10), %xmm1 # 0 : sbbt + + tbl v4.16b, {$sbbu}, v2.16b // vpshufb %xmm2, %xmm4, %xmm4 # 4 = sbbu + tbl v12.16b, {$sbbu}, v10.16b + tbl v0.16b, {v0.16b},v5.16b // vpshufb %xmm5, %xmm0, %xmm0 # MC ch + tbl v8.16b, {v8.16b},v5.16b + tbl v1.16b, {$sbbt}, v3.16b // vpshufb %xmm3, %xmm1, %xmm1 # 0 = sbbt + tbl v9.16b, {$sbbt}, v11.16b + eor v0.16b, v0.16b, v4.16b // vpxor %xmm4, %xmm0, %xmm0 # 4 = ch + eor v8.16b, v8.16b, v12.16b + // vmovdqa 0x40(%r10), %xmm4 # 4 : sbeu + eor v0.16b, v0.16b, v1.16b // vpxor %xmm1, %xmm0, %xmm0 # 0 = ch + eor v8.16b, v8.16b, v9.16b + // vmovdqa 0x50(%r10), %xmm1 # 0 : sbet + + tbl v4.16b, {$sbeu}, v2.16b // vpshufb %xmm2, %xmm4, %xmm4 # 4 = sbeu + tbl v12.16b, {$sbeu}, v10.16b + tbl v0.16b, {v0.16b},v5.16b // vpshufb %xmm5, %xmm0, %xmm0 # MC ch + tbl v8.16b, {v8.16b},v5.16b + tbl v1.16b, {$sbet}, v3.16b // vpshufb %xmm3, %xmm1, %xmm1 # 0 = sbet + tbl v9.16b, {$sbet}, v11.16b + eor v0.16b, v0.16b, v4.16b // vpxor %xmm4, %xmm0, %xmm0 # 4 = ch + eor v8.16b, v8.16b, v12.16b + ext v5.16b, v5.16b, v5.16b, #12 // vpalignr \$12, %xmm5, %xmm5, %xmm5 + eor v0.16b, v0.16b, v1.16b // vpxor %xmm1, %xmm0, %xmm0 # 0 = ch + eor v8.16b, v8.16b, v9.16b + sub w8, w8, #1 // sub \$1,%rax # nr-- + +.Ldec_2x_entry: + // top of round + and v1.16b, v0.16b, v17.16b // vpand %xmm9, %xmm0, %xmm1 # 0 = k + ushr v0.16b, v0.16b, #4 // vpsrlb \$4, %xmm0, %xmm0 # 1 = i + and v9.16b, v8.16b, v17.16b + ushr v8.16b, v8.16b, #4 + tbl v2.16b, {$invhi},v1.16b // vpshufb %xmm1, %xmm11, %xmm2 # 2 = a/k + tbl v10.16b, {$invhi},v9.16b + eor v1.16b, v1.16b, v0.16b // vpxor %xmm0, %xmm1, %xmm1 # 0 = j + eor v9.16b, v9.16b, v8.16b + tbl v3.16b, {$invlo},v0.16b // vpshufb %xmm0, %xmm10, %xmm3 # 3 = 1/i + tbl v11.16b, {$invlo},v8.16b + tbl v4.16b, {$invlo},v1.16b // vpshufb %xmm1, %xmm10, %xmm4 # 4 = 1/j + tbl v12.16b, {$invlo},v9.16b + eor v3.16b, v3.16b, v2.16b // vpxor %xmm2, %xmm3, %xmm3 # 3 = iak = 1/i + a/k + eor v11.16b, v11.16b, v10.16b + eor v4.16b, v4.16b, v2.16b // vpxor %xmm2, %xmm4, %xmm4 # 4 = jak = 1/j + a/k + eor v12.16b, v12.16b, v10.16b + tbl v2.16b, {$invlo},v3.16b // vpshufb %xmm3, %xmm10, %xmm2 # 2 = 1/iak + tbl v10.16b, {$invlo},v11.16b + tbl v3.16b, {$invlo},v4.16b // vpshufb %xmm4, %xmm10, %xmm3 # 3 = 1/jak + tbl v11.16b, {$invlo},v12.16b + eor v2.16b, v2.16b, v1.16b // vpxor %xmm1, %xmm2, %xmm2 # 2 = io + eor v10.16b, v10.16b, v9.16b + eor v3.16b, v3.16b, v0.16b // vpxor %xmm0, %xmm3, %xmm3 # 3 = jo + eor v11.16b, v11.16b, v8.16b + ld1 {v16.2d}, [x9],#16 // vmovdqu (%r9), %xmm0 + cbnz w8, .Ldec_2x_loop + + // middle of last round + // vmovdqa 0x60(%r10), %xmm4 # 3 : sbou + tbl v4.16b, {$sbou}, v2.16b // vpshufb %xmm2, %xmm4, %xmm4 # 4 = sbou + tbl v12.16b, {$sbou}, v10.16b + // vmovdqa 0x70(%r10), %xmm1 # 0 : sbot + tbl v1.16b, {$sbot}, v3.16b // vpshufb %xmm3, %xmm1, %xmm1 # 0 = sb1t + tbl v9.16b, {$sbot}, v11.16b + ld1 {v2.2d}, [x11] // vmovdqa -0x160(%r11), %xmm2 # .Lk_sr-.Lk_dsbd=-0x160 + eor v4.16b, v4.16b, v16.16b // vpxor %xmm0, %xmm4, %xmm4 # 4 = sb1u + k + eor v12.16b, v12.16b, v16.16b + eor v0.16b, v1.16b, v4.16b // vpxor %xmm4, %xmm1, %xmm0 # 0 = A + eor v8.16b, v9.16b, v12.16b + tbl v0.16b, {v0.16b},v2.16b // vpshufb %xmm2, %xmm0, %xmm0 + tbl v1.16b, {v8.16b},v2.16b + ret +.size _vpaes_decrypt_2x,.-_vpaes_decrypt_2x +___ +} +{ +my ($inp,$bits,$out,$dir)=("x0","w1","x2","w3"); +my ($invlo,$invhi,$iptlo,$ipthi,$rcon) = map("v$_.16b",(18..21,8)); + +$code.=<<___; +######################################################## +## ## +## AES key schedule ## +## ## +######################################################## +.type _vpaes_key_preheat,%function +.align 4 +_vpaes_key_preheat: + adr x10, .Lk_inv + movi v16.16b, #0x5b // .Lk_s63 + adr x11, .Lk_sb1 + movi v17.16b, #0x0f // .Lk_s0F + ld1 {v18.2d-v21.2d}, [x10] // .Lk_inv, .Lk_ipt + adr x10, .Lk_dksd + ld1 {v22.2d-v23.2d}, [x11] // .Lk_sb1 + adr x11, .Lk_mc_forward + ld1 {v24.2d-v27.2d}, [x10],#64 // .Lk_dksd, .Lk_dksb + ld1 {v28.2d-v31.2d}, [x10],#64 // .Lk_dkse, .Lk_dks9 + ld1 {v8.2d}, [x10] // .Lk_rcon + ld1 {v9.2d}, [x11] // .Lk_mc_forward[0] + ret +.size _vpaes_key_preheat,.-_vpaes_key_preheat + +.type _vpaes_schedule_core,%function +.align 4 +_vpaes_schedule_core: + stp x29, x30, [sp,#-16]! + add x29,sp,#0 + + bl _vpaes_key_preheat // load the tables + + ld1 {v0.16b}, [$inp],#16 // vmovdqu (%rdi), %xmm0 # load key (unaligned) + + // input transform + mov v3.16b, v0.16b // vmovdqa %xmm0, %xmm3 + bl _vpaes_schedule_transform + mov v7.16b, v0.16b // vmovdqa %xmm0, %xmm7 + + adr x10, .Lk_sr // lea .Lk_sr(%rip),%r10 + add x8, x8, x10 + cbnz $dir, .Lschedule_am_decrypting + + // encrypting, output zeroth round key after transform + st1 {v0.2d}, [$out] // vmovdqu %xmm0, (%rdx) + b .Lschedule_go + +.Lschedule_am_decrypting: + // decrypting, output zeroth round key after shiftrows + ld1 {v1.2d}, [x8] // vmovdqa (%r8,%r10), %xmm1 + tbl v3.16b, {v3.16b}, v1.16b // vpshufb %xmm1, %xmm3, %xmm3 + st1 {v3.2d}, [$out] // vmovdqu %xmm3, (%rdx) + eor x8, x8, #0x30 // xor \$0x30, %r8 + +.Lschedule_go: + cmp $bits, #192 // cmp \$192, %esi + b.hi .Lschedule_256 + b.eq .Lschedule_192 + // 128: fall though + +## +## .schedule_128 +## +## 128-bit specific part of key schedule. +## +## This schedule is really simple, because all its parts +## are accomplished by the subroutines. +## +.Lschedule_128: + mov $inp, #10 // mov \$10, %esi + +.Loop_schedule_128: + sub $inp, $inp, #1 // dec %esi + bl _vpaes_schedule_round + cbz $inp, .Lschedule_mangle_last + bl _vpaes_schedule_mangle // write output + b .Loop_schedule_128 + +## +## .aes_schedule_192 +## +## 192-bit specific part of key schedule. +## +## The main body of this schedule is the same as the 128-bit +## schedule, but with more smearing. The long, high side is +## stored in %xmm7 as before, and the short, low side is in +## the high bits of %xmm6. +## +## This schedule is somewhat nastier, however, because each +## round produces 192 bits of key material, or 1.5 round keys. +## Therefore, on each cycle we do 2 rounds and produce 3 round +## keys. +## +.align 4 +.Lschedule_192: + sub $inp, $inp, #8 + ld1 {v0.16b}, [$inp] // vmovdqu 8(%rdi),%xmm0 # load key part 2 (very unaligned) + bl _vpaes_schedule_transform // input transform + mov v6.16b, v0.16b // vmovdqa %xmm0, %xmm6 # save short part + eor v4.16b, v4.16b, v4.16b // vpxor %xmm4, %xmm4, %xmm4 # clear 4 + ins v6.d[0], v4.d[0] // vmovhlps %xmm4, %xmm6, %xmm6 # clobber low side with zeros + mov $inp, #4 // mov \$4, %esi + +.Loop_schedule_192: + sub $inp, $inp, #1 // dec %esi + bl _vpaes_schedule_round + ext v0.16b, v6.16b, v0.16b, #8 // vpalignr \$8,%xmm6,%xmm0,%xmm0 + bl _vpaes_schedule_mangle // save key n + bl _vpaes_schedule_192_smear + bl _vpaes_schedule_mangle // save key n+1 + bl _vpaes_schedule_round + cbz $inp, .Lschedule_mangle_last + bl _vpaes_schedule_mangle // save key n+2 + bl _vpaes_schedule_192_smear + b .Loop_schedule_192 + +## +## .aes_schedule_256 +## +## 256-bit specific part of key schedule. +## +## The structure here is very similar to the 128-bit +## schedule, but with an additional "low side" in +## %xmm6. The low side's rounds are the same as the +## high side's, except no rcon and no rotation. +## +.align 4 +.Lschedule_256: + ld1 {v0.16b}, [$inp] // vmovdqu 16(%rdi),%xmm0 # load key part 2 (unaligned) + bl _vpaes_schedule_transform // input transform + mov $inp, #7 // mov \$7, %esi + +.Loop_schedule_256: + sub $inp, $inp, #1 // dec %esi + bl _vpaes_schedule_mangle // output low result + mov v6.16b, v0.16b // vmovdqa %xmm0, %xmm6 # save cur_lo in xmm6 + + // high round + bl _vpaes_schedule_round + cbz $inp, .Lschedule_mangle_last + bl _vpaes_schedule_mangle + + // low round. swap xmm7 and xmm6 + dup v0.4s, v0.s[3] // vpshufd \$0xFF, %xmm0, %xmm0 + movi v4.16b, #0 + mov v5.16b, v7.16b // vmovdqa %xmm7, %xmm5 + mov v7.16b, v6.16b // vmovdqa %xmm6, %xmm7 + bl _vpaes_schedule_low_round + mov v7.16b, v5.16b // vmovdqa %xmm5, %xmm7 + + b .Loop_schedule_256 + +## +## .aes_schedule_mangle_last +## +## Mangler for last round of key schedule +## Mangles %xmm0 +## when encrypting, outputs out(%xmm0) ^ 63 +## when decrypting, outputs unskew(%xmm0) +## +## Always called right before return... jumps to cleanup and exits +## +.align 4 +.Lschedule_mangle_last: + // schedule last round key from xmm0 + adr x11, .Lk_deskew // lea .Lk_deskew(%rip),%r11 # prepare to deskew + cbnz $dir, .Lschedule_mangle_last_dec + + // encrypting + ld1 {v1.2d}, [x8] // vmovdqa (%r8,%r10),%xmm1 + adr x11, .Lk_opt // lea .Lk_opt(%rip), %r11 # prepare to output transform + add $out, $out, #32 // add \$32, %rdx + tbl v0.16b, {v0.16b}, v1.16b // vpshufb %xmm1, %xmm0, %xmm0 # output permute + +.Lschedule_mangle_last_dec: + ld1 {v20.2d-v21.2d}, [x11] // reload constants + sub $out, $out, #16 // add \$-16, %rdx + eor v0.16b, v0.16b, v16.16b // vpxor .Lk_s63(%rip), %xmm0, %xmm0 + bl _vpaes_schedule_transform // output transform + st1 {v0.2d}, [$out] // vmovdqu %xmm0, (%rdx) # save last key + + // cleanup + eor v0.16b, v0.16b, v0.16b // vpxor %xmm0, %xmm0, %xmm0 + eor v1.16b, v1.16b, v1.16b // vpxor %xmm1, %xmm1, %xmm1 + eor v2.16b, v2.16b, v2.16b // vpxor %xmm2, %xmm2, %xmm2 + eor v3.16b, v3.16b, v3.16b // vpxor %xmm3, %xmm3, %xmm3 + eor v4.16b, v4.16b, v4.16b // vpxor %xmm4, %xmm4, %xmm4 + eor v5.16b, v5.16b, v5.16b // vpxor %xmm5, %xmm5, %xmm5 + eor v6.16b, v6.16b, v6.16b // vpxor %xmm6, %xmm6, %xmm6 + eor v7.16b, v7.16b, v7.16b // vpxor %xmm7, %xmm7, %xmm7 + ldp x29, x30, [sp],#16 + ret +.size _vpaes_schedule_core,.-_vpaes_schedule_core + +## +## .aes_schedule_192_smear +## +## Smear the short, low side in the 192-bit key schedule. +## +## Inputs: +## %xmm7: high side, b a x y +## %xmm6: low side, d c 0 0 +## %xmm13: 0 +## +## Outputs: +## %xmm6: b+c+d b+c 0 0 +## %xmm0: b+c+d b+c b a +## +.type _vpaes_schedule_192_smear,%function +.align 4 +_vpaes_schedule_192_smear: + movi v1.16b, #0 + dup v0.4s, v7.s[3] + ins v1.s[3], v6.s[2] // vpshufd \$0x80, %xmm6, %xmm1 # d c 0 0 -> c 0 0 0 + ins v0.s[0], v7.s[2] // vpshufd \$0xFE, %xmm7, %xmm0 # b a _ _ -> b b b a + eor v6.16b, v6.16b, v1.16b // vpxor %xmm1, %xmm6, %xmm6 # -> c+d c 0 0 + eor v1.16b, v1.16b, v1.16b // vpxor %xmm1, %xmm1, %xmm1 + eor v6.16b, v6.16b, v0.16b // vpxor %xmm0, %xmm6, %xmm6 # -> b+c+d b+c b a + mov v0.16b, v6.16b // vmovdqa %xmm6, %xmm0 + ins v6.d[0], v1.d[0] // vmovhlps %xmm1, %xmm6, %xmm6 # clobber low side with zeros + ret +.size _vpaes_schedule_192_smear,.-_vpaes_schedule_192_smear + +## +## .aes_schedule_round +## +## Runs one main round of the key schedule on %xmm0, %xmm7 +## +## Specifically, runs subbytes on the high dword of %xmm0 +## then rotates it by one byte and xors into the low dword of +## %xmm7. +## +## Adds rcon from low byte of %xmm8, then rotates %xmm8 for +## next rcon. +## +## Smears the dwords of %xmm7 by xoring the low into the +## second low, result into third, result into highest. +## +## Returns results in %xmm7 = %xmm0. +## Clobbers %xmm1-%xmm4, %r11. +## +.type _vpaes_schedule_round,%function +.align 4 +_vpaes_schedule_round: + // extract rcon from xmm8 + movi v4.16b, #0 // vpxor %xmm4, %xmm4, %xmm4 + ext v1.16b, $rcon, v4.16b, #15 // vpalignr \$15, %xmm8, %xmm4, %xmm1 + ext $rcon, $rcon, $rcon, #15 // vpalignr \$15, %xmm8, %xmm8, %xmm8 + eor v7.16b, v7.16b, v1.16b // vpxor %xmm1, %xmm7, %xmm7 + + // rotate + dup v0.4s, v0.s[3] // vpshufd \$0xFF, %xmm0, %xmm0 + ext v0.16b, v0.16b, v0.16b, #1 // vpalignr \$1, %xmm0, %xmm0, %xmm0 + + // fall through... + + // low round: same as high round, but no rotation and no rcon. +_vpaes_schedule_low_round: + // smear xmm7 + ext v1.16b, v4.16b, v7.16b, #12 // vpslldq \$4, %xmm7, %xmm1 + eor v7.16b, v7.16b, v1.16b // vpxor %xmm1, %xmm7, %xmm7 + ext v4.16b, v4.16b, v7.16b, #8 // vpslldq \$8, %xmm7, %xmm4 + + // subbytes + and v1.16b, v0.16b, v17.16b // vpand %xmm9, %xmm0, %xmm1 # 0 = k + ushr v0.16b, v0.16b, #4 // vpsrlb \$4, %xmm0, %xmm0 # 1 = i + eor v7.16b, v7.16b, v4.16b // vpxor %xmm4, %xmm7, %xmm7 + tbl v2.16b, {$invhi}, v1.16b // vpshufb %xmm1, %xmm11, %xmm2 # 2 = a/k + eor v1.16b, v1.16b, v0.16b // vpxor %xmm0, %xmm1, %xmm1 # 0 = j + tbl v3.16b, {$invlo}, v0.16b // vpshufb %xmm0, %xmm10, %xmm3 # 3 = 1/i + eor v3.16b, v3.16b, v2.16b // vpxor %xmm2, %xmm3, %xmm3 # 3 = iak = 1/i + a/k + tbl v4.16b, {$invlo}, v1.16b // vpshufb %xmm1, %xmm10, %xmm4 # 4 = 1/j + eor v7.16b, v7.16b, v16.16b // vpxor .Lk_s63(%rip), %xmm7, %xmm7 + tbl v3.16b, {$invlo}, v3.16b // vpshufb %xmm3, %xmm10, %xmm3 # 2 = 1/iak + eor v4.16b, v4.16b, v2.16b // vpxor %xmm2, %xmm4, %xmm4 # 4 = jak = 1/j + a/k + tbl v2.16b, {$invlo}, v4.16b // vpshufb %xmm4, %xmm10, %xmm2 # 3 = 1/jak + eor v3.16b, v3.16b, v1.16b // vpxor %xmm1, %xmm3, %xmm3 # 2 = io + eor v2.16b, v2.16b, v0.16b // vpxor %xmm0, %xmm2, %xmm2 # 3 = jo + tbl v4.16b, {v23.16b}, v3.16b // vpshufb %xmm3, %xmm13, %xmm4 # 4 = sbou + tbl v1.16b, {v22.16b}, v2.16b // vpshufb %xmm2, %xmm12, %xmm1 # 0 = sb1t + eor v1.16b, v1.16b, v4.16b // vpxor %xmm4, %xmm1, %xmm1 # 0 = sbox output + + // add in smeared stuff + eor v0.16b, v1.16b, v7.16b // vpxor %xmm7, %xmm1, %xmm0 + eor v7.16b, v1.16b, v7.16b // vmovdqa %xmm0, %xmm7 + ret +.size _vpaes_schedule_round,.-_vpaes_schedule_round + +## +## .aes_schedule_transform +## +## Linear-transform %xmm0 according to tables at (%r11) +## +## Requires that %xmm9 = 0x0F0F... as in preheat +## Output in %xmm0 +## Clobbers %xmm1, %xmm2 +## +.type _vpaes_schedule_transform,%function +.align 4 +_vpaes_schedule_transform: + and v1.16b, v0.16b, v17.16b // vpand %xmm9, %xmm0, %xmm1 + ushr v0.16b, v0.16b, #4 // vpsrlb \$4, %xmm0, %xmm0 + // vmovdqa (%r11), %xmm2 # lo + tbl v2.16b, {$iptlo}, v1.16b // vpshufb %xmm1, %xmm2, %xmm2 + // vmovdqa 16(%r11), %xmm1 # hi + tbl v0.16b, {$ipthi}, v0.16b // vpshufb %xmm0, %xmm1, %xmm0 + eor v0.16b, v0.16b, v2.16b // vpxor %xmm2, %xmm0, %xmm0 + ret +.size _vpaes_schedule_transform,.-_vpaes_schedule_transform + +## +## .aes_schedule_mangle +## +## Mangle xmm0 from (basis-transformed) standard version +## to our version. +## +## On encrypt, +## xor with 0x63 +## multiply by circulant 0,1,1,1 +## apply shiftrows transform +## +## On decrypt, +## xor with 0x63 +## multiply by "inverse mixcolumns" circulant E,B,D,9 +## deskew +## apply shiftrows transform +## +## +## Writes out to (%rdx), and increments or decrements it +## Keeps track of round number mod 4 in %r8 +## Preserves xmm0 +## Clobbers xmm1-xmm5 +## +.type _vpaes_schedule_mangle,%function +.align 4 +_vpaes_schedule_mangle: + mov v4.16b, v0.16b // vmovdqa %xmm0, %xmm4 # save xmm0 for later + // vmovdqa .Lk_mc_forward(%rip),%xmm5 + cbnz $dir, .Lschedule_mangle_dec + + // encrypting + eor v4.16b, v0.16b, v16.16b // vpxor .Lk_s63(%rip), %xmm0, %xmm4 + add $out, $out, #16 // add \$16, %rdx + tbl v4.16b, {v4.16b}, v9.16b // vpshufb %xmm5, %xmm4, %xmm4 + tbl v1.16b, {v4.16b}, v9.16b // vpshufb %xmm5, %xmm4, %xmm1 + tbl v3.16b, {v1.16b}, v9.16b // vpshufb %xmm5, %xmm1, %xmm3 + eor v4.16b, v4.16b, v1.16b // vpxor %xmm1, %xmm4, %xmm4 + ld1 {v1.2d}, [x8] // vmovdqa (%r8,%r10), %xmm1 + eor v3.16b, v3.16b, v4.16b // vpxor %xmm4, %xmm3, %xmm3 + + b .Lschedule_mangle_both +.align 4 +.Lschedule_mangle_dec: + // inverse mix columns + // lea .Lk_dksd(%rip),%r11 + ushr v1.16b, v4.16b, #4 // vpsrlb \$4, %xmm4, %xmm1 # 1 = hi + and v4.16b, v4.16b, v17.16b // vpand %xmm9, %xmm4, %xmm4 # 4 = lo + + // vmovdqa 0x00(%r11), %xmm2 + tbl v2.16b, {v24.16b}, v4.16b // vpshufb %xmm4, %xmm2, %xmm2 + // vmovdqa 0x10(%r11), %xmm3 + tbl v3.16b, {v25.16b}, v1.16b // vpshufb %xmm1, %xmm3, %xmm3 + eor v3.16b, v3.16b, v2.16b // vpxor %xmm2, %xmm3, %xmm3 + tbl v3.16b, {v3.16b}, v9.16b // vpshufb %xmm5, %xmm3, %xmm3 + + // vmovdqa 0x20(%r11), %xmm2 + tbl v2.16b, {v26.16b}, v4.16b // vpshufb %xmm4, %xmm2, %xmm2 + eor v2.16b, v2.16b, v3.16b // vpxor %xmm3, %xmm2, %xmm2 + // vmovdqa 0x30(%r11), %xmm3 + tbl v3.16b, {v27.16b}, v1.16b // vpshufb %xmm1, %xmm3, %xmm3 + eor v3.16b, v3.16b, v2.16b // vpxor %xmm2, %xmm3, %xmm3 + tbl v3.16b, {v3.16b}, v9.16b // vpshufb %xmm5, %xmm3, %xmm3 + + // vmovdqa 0x40(%r11), %xmm2 + tbl v2.16b, {v28.16b}, v4.16b // vpshufb %xmm4, %xmm2, %xmm2 + eor v2.16b, v2.16b, v3.16b // vpxor %xmm3, %xmm2, %xmm2 + // vmovdqa 0x50(%r11), %xmm3 + tbl v3.16b, {v29.16b}, v1.16b // vpshufb %xmm1, %xmm3, %xmm3 + eor v3.16b, v3.16b, v2.16b // vpxor %xmm2, %xmm3, %xmm3 + + // vmovdqa 0x60(%r11), %xmm2 + tbl v2.16b, {v30.16b}, v4.16b // vpshufb %xmm4, %xmm2, %xmm2 + tbl v3.16b, {v3.16b}, v9.16b // vpshufb %xmm5, %xmm3, %xmm3 + // vmovdqa 0x70(%r11), %xmm4 + tbl v4.16b, {v31.16b}, v1.16b // vpshufb %xmm1, %xmm4, %xmm4 + ld1 {v1.2d}, [x8] // vmovdqa (%r8,%r10), %xmm1 + eor v2.16b, v2.16b, v3.16b // vpxor %xmm3, %xmm2, %xmm2 + eor v3.16b, v4.16b, v2.16b // vpxor %xmm2, %xmm4, %xmm3 + + sub $out, $out, #16 // add \$-16, %rdx + +.Lschedule_mangle_both: + tbl v3.16b, {v3.16b}, v1.16b // vpshufb %xmm1, %xmm3, %xmm3 + add x8, x8, #64-16 // add \$-16, %r8 + and x8, x8, #~(1<<6) // and \$0x30, %r8 + st1 {v3.2d}, [$out] // vmovdqu %xmm3, (%rdx) + ret +.size _vpaes_schedule_mangle,.-_vpaes_schedule_mangle + +.globl vpaes_set_encrypt_key +.type vpaes_set_encrypt_key,%function +.align 4 +vpaes_set_encrypt_key: + stp x29,x30,[sp,#-16]! + add x29,sp,#0 + stp d8,d9,[sp,#-16]! // ABI spec says so + + lsr w9, $bits, #5 // shr \$5,%eax + add w9, w9, #5 // \$5,%eax + str w9, [$out,#240] // mov %eax,240(%rdx) # AES_KEY->rounds = nbits/32+5; + + mov $dir, #0 // mov \$0,%ecx + mov x8, #0x30 // mov \$0x30,%r8d + bl _vpaes_schedule_core + eor x0, x0, x0 + + ldp d8,d9,[sp],#16 + ldp x29,x30,[sp],#16 + ret +.size vpaes_set_encrypt_key,.-vpaes_set_encrypt_key + +.globl vpaes_set_decrypt_key +.type vpaes_set_decrypt_key,%function +.align 4 +vpaes_set_decrypt_key: + stp x29,x30,[sp,#-16]! + add x29,sp,#0 + stp d8,d9,[sp,#-16]! // ABI spec says so + + lsr w9, $bits, #5 // shr \$5,%eax + add w9, w9, #5 // \$5,%eax + str w9, [$out,#240] // mov %eax,240(%rdx) # AES_KEY->rounds = nbits/32+5; + lsl w9, w9, #4 // shl \$4,%eax + add $out, $out, #16 // lea 16(%rdx,%rax),%rdx + add $out, $out, x9 + + mov $dir, #1 // mov \$1,%ecx + lsr w8, $bits, #1 // shr \$1,%r8d + and x8, x8, #32 // and \$32,%r8d + eor x8, x8, #32 // xor \$32,%r8d # nbits==192?0:32 + bl _vpaes_schedule_core + + ldp d8,d9,[sp],#16 + ldp x29,x30,[sp],#16 + ret +.size vpaes_set_decrypt_key,.-vpaes_set_decrypt_key +___ +} +{ +my ($inp,$out,$len,$key,$ivec,$dir) = map("x$_",(0..5)); + +$code.=<<___; +.globl vpaes_cbc_encrypt +.type vpaes_cbc_encrypt,%function +.align 4 +vpaes_cbc_encrypt: + cbz $len, .Lcbc_abort + cmp w5, #0 // check direction + b.eq vpaes_cbc_decrypt + + stp x29,x30,[sp,#-16]! + add x29,sp,#0 + + mov x17, $len // reassign + mov x2, $key // reassign + + ld1 {v0.16b}, [$ivec] // load ivec + bl _vpaes_encrypt_preheat + b .Lcbc_enc_loop + +.align 4 +.Lcbc_enc_loop: + ld1 {v7.16b}, [$inp],#16 // load input + eor v7.16b, v7.16b, v0.16b // xor with ivec + bl _vpaes_encrypt_core + st1 {v0.16b}, [$out],#16 // save output + subs x17, x17, #16 + b.hi .Lcbc_enc_loop + + st1 {v0.16b}, [$ivec] // write ivec + + ldp x29,x30,[sp],#16 +.Lcbc_abort: + ret +.size vpaes_cbc_encrypt,.-vpaes_cbc_encrypt + +.type vpaes_cbc_decrypt,%function +.align 4 +vpaes_cbc_decrypt: + stp x29,x30,[sp,#-16]! + add x29,sp,#0 + stp d8,d9,[sp,#-16]! // ABI spec says so + stp d10,d11,[sp,#-16]! + stp d12,d13,[sp,#-16]! + stp d14,d15,[sp,#-16]! + + mov x17, $len // reassign + mov x2, $key // reassign + ld1 {v6.16b}, [$ivec] // load ivec + bl _vpaes_decrypt_preheat + tst x17, #16 + b.eq .Lcbc_dec_loop2x + + ld1 {v7.16b}, [$inp], #16 // load input + bl _vpaes_decrypt_core + eor v0.16b, v0.16b, v6.16b // xor with ivec + orr v6.16b, v7.16b, v7.16b // next ivec value + st1 {v0.16b}, [$out], #16 + subs x17, x17, #16 + b.ls .Lcbc_dec_done + +.align 4 +.Lcbc_dec_loop2x: + ld1 {v14.16b,v15.16b}, [$inp], #32 + bl _vpaes_decrypt_2x + eor v0.16b, v0.16b, v6.16b // xor with ivec + eor v1.16b, v1.16b, v14.16b + orr v6.16b, v15.16b, v15.16b + st1 {v0.16b,v1.16b}, [$out], #32 + subs x17, x17, #32 + b.hi .Lcbc_dec_loop2x + +.Lcbc_dec_done: + st1 {v6.16b}, [$ivec] + + ldp d14,d15,[sp],#16 + ldp d12,d13,[sp],#16 + ldp d10,d11,[sp],#16 + ldp d8,d9,[sp],#16 + ldp x29,x30,[sp],#16 + ret +.size vpaes_cbc_decrypt,.-vpaes_cbc_decrypt +___ +if (1) { +$code.=<<___; +.globl vpaes_ecb_encrypt +.type vpaes_ecb_encrypt,%function +.align 4 +vpaes_ecb_encrypt: + stp x29,x30,[sp,#-16]! + add x29,sp,#0 + stp d8,d9,[sp,#-16]! // ABI spec says so + stp d10,d11,[sp,#-16]! + stp d12,d13,[sp,#-16]! + stp d14,d15,[sp,#-16]! + + mov x17, $len + mov x2, $key + bl _vpaes_encrypt_preheat + tst x17, #16 + b.eq .Lecb_enc_loop + + ld1 {v7.16b}, [$inp],#16 + bl _vpaes_encrypt_core + st1 {v0.16b}, [$out],#16 + subs x17, x17, #16 + b.ls .Lecb_enc_done + +.align 4 +.Lecb_enc_loop: + ld1 {v14.16b,v15.16b}, [$inp], #32 + bl _vpaes_encrypt_2x + st1 {v0.16b,v1.16b}, [$out], #32 + subs x17, x17, #32 + b.hi .Lecb_enc_loop + +.Lecb_enc_done: + ldp d14,d15,[sp],#16 + ldp d12,d13,[sp],#16 + ldp d10,d11,[sp],#16 + ldp d8,d9,[sp],#16 + ldp x29,x30,[sp],#16 + ret +.size vpaes_ecb_encrypt,.-vpaes_ecb_encrypt + +.globl vpaes_ecb_decrypt +.type vpaes_ecb_decrypt,%function +.align 4 +vpaes_ecb_decrypt: + stp x29,x30,[sp,#-16]! + add x29,sp,#0 + stp d8,d9,[sp,#-16]! // ABI spec says so + stp d10,d11,[sp,#-16]! + stp d12,d13,[sp,#-16]! + stp d14,d15,[sp,#-16]! + + mov x17, $len + mov x2, $key + bl _vpaes_decrypt_preheat + tst x17, #16 + b.eq .Lecb_dec_loop + + ld1 {v7.16b}, [$inp],#16 + bl _vpaes_encrypt_core + st1 {v0.16b}, [$out],#16 + subs x17, x17, #16 + b.ls .Lecb_dec_done + +.align 4 +.Lecb_dec_loop: + ld1 {v14.16b,v15.16b}, [$inp], #32 + bl _vpaes_decrypt_2x + st1 {v0.16b,v1.16b}, [$out], #32 + subs x17, x17, #32 + b.hi .Lecb_dec_loop + +.Lecb_dec_done: + ldp d14,d15,[sp],#16 + ldp d12,d13,[sp],#16 + ldp d10,d11,[sp],#16 + ldp d8,d9,[sp],#16 + ldp x29,x30,[sp],#16 + ret +.size vpaes_ecb_decrypt,.-vpaes_ecb_decrypt +___ +} } +print $code; + +close STDOUT; diff --git a/openssl-1.1.0h/crypto/aes/asm/vpaes-ppc.pl b/openssl-1.1.0h/crypto/aes/asm/vpaes-ppc.pl new file mode 100644 index 0000000..bb38fbe --- /dev/null +++ b/openssl-1.1.0h/crypto/aes/asm/vpaes-ppc.pl @@ -0,0 +1,1594 @@ +#! /usr/bin/env perl +# Copyright 2013-2016 The OpenSSL Project Authors. All Rights Reserved. +# +# Licensed under the OpenSSL license (the "License"). You may not use +# this file except in compliance with the License. You can obtain a copy +# in the file LICENSE in the source distribution or at +# https://www.openssl.org/source/license.html + + +###################################################################### +## Constant-time SSSE3 AES core implementation. +## version 0.1 +## +## By Mike Hamburg (Stanford University), 2009 +## Public domain. +## +## For details see http://shiftleft.org/papers/vector_aes/ and +## http://crypto.stanford.edu/vpaes/. + +# CBC encrypt/decrypt performance in cycles per byte processed with +# 128-bit key. +# +# aes-ppc.pl this +# PPC74x0/G4e 35.5/52.1/(23.8) 11.9(*)/15.4 +# PPC970/G5 37.9/55.0/(28.5) 22.2/28.5 +# POWER6 42.7/54.3/(28.2) 63.0/92.8(**) +# POWER7 32.3/42.9/(18.4) 18.5/23.3 +# +# (*) This is ~10% worse than reported in paper. The reason is +# twofold. This module doesn't make any assumption about +# key schedule (or data for that matter) alignment and handles +# it in-line. Secondly it, being transliterated from +# vpaes-x86_64.pl, relies on "nested inversion" better suited +# for Intel CPUs. +# (**) Inadequate POWER6 performance is due to astronomic AltiVec +# latency, 9 cycles per simple logical operation. + +$flavour = shift; + +if ($flavour =~ /64/) { + $SIZE_T =8; + $LRSAVE =2*$SIZE_T; + $STU ="stdu"; + $POP ="ld"; + $PUSH ="std"; + $UCMP ="cmpld"; +} elsif ($flavour =~ /32/) { + $SIZE_T =4; + $LRSAVE =$SIZE_T; + $STU ="stwu"; + $POP ="lwz"; + $PUSH ="stw"; + $UCMP ="cmplw"; +} else { die "nonsense $flavour"; } + +$sp="r1"; +$FRAME=6*$SIZE_T+13*16; # 13*16 is for v20-v31 offload + +$0 =~ m/(.*[\/\\])[^\/\\]+$/; $dir=$1; +( $xlate="${dir}ppc-xlate.pl" and -f $xlate ) or +( $xlate="${dir}../../perlasm/ppc-xlate.pl" and -f $xlate) or +die "can't locate ppc-xlate.pl"; + +open STDOUT,"| $^X $xlate $flavour ".shift || die "can't call $xlate: $!"; + +$code.=<<___; +.machine "any" + +.text + +.align 7 # totally strategic alignment +_vpaes_consts: +Lk_mc_forward: # mc_forward + .long 0x01020300, 0x05060704, 0x090a0b08, 0x0d0e0f0c ?inv + .long 0x05060704, 0x090a0b08, 0x0d0e0f0c, 0x01020300 ?inv + .long 0x090a0b08, 0x0d0e0f0c, 0x01020300, 0x05060704 ?inv + .long 0x0d0e0f0c, 0x01020300, 0x05060704, 0x090a0b08 ?inv +Lk_mc_backward: # mc_backward + .long 0x03000102, 0x07040506, 0x0b08090a, 0x0f0c0d0e ?inv + .long 0x0f0c0d0e, 0x03000102, 0x07040506, 0x0b08090a ?inv + .long 0x0b08090a, 0x0f0c0d0e, 0x03000102, 0x07040506 ?inv + .long 0x07040506, 0x0b08090a, 0x0f0c0d0e, 0x03000102 ?inv +Lk_sr: # sr + .long 0x00010203, 0x04050607, 0x08090a0b, 0x0c0d0e0f ?inv + .long 0x00050a0f, 0x04090e03, 0x080d0207, 0x0c01060b ?inv + .long 0x0009020b, 0x040d060f, 0x08010a03, 0x0c050e07 ?inv + .long 0x000d0a07, 0x04010e0b, 0x0805020f, 0x0c090603 ?inv + +## +## "Hot" constants +## +Lk_inv: # inv, inva + .long 0xf001080d, 0x0f06050e, 0x020c0b0a, 0x09030704 ?rev + .long 0xf0070b0f, 0x060a0401, 0x09080502, 0x0c0e0d03 ?rev +Lk_ipt: # input transform (lo, hi) + .long 0x00702a5a, 0x98e8b2c2, 0x08782252, 0x90e0baca ?rev + .long 0x004d7c31, 0x7d30014c, 0x81ccfdb0, 0xfcb180cd ?rev +Lk_sbo: # sbou, sbot + .long 0x00c7bd6f, 0x176dd2d0, 0x78a802c5, 0x7abfaa15 ?rev + .long 0x006abb5f, 0xa574e4cf, 0xfa352b41, 0xd1901e8e ?rev +Lk_sb1: # sb1u, sb1t + .long 0x0023e2fa, 0x15d41836, 0xefd92e0d, 0xc1ccf73b ?rev + .long 0x003e50cb, 0x8fe19bb1, 0x44f52a14, 0x6e7adfa5 ?rev +Lk_sb2: # sb2u, sb2t + .long 0x0029e10a, 0x4088eb69, 0x4a2382ab, 0xc863a1c2 ?rev + .long 0x0024710b, 0xc6937ae2, 0xcd2f98bc, 0x55e9b75e ?rev + +## +## Decryption stuff +## +Lk_dipt: # decryption input transform + .long 0x005f540b, 0x045b500f, 0x1a454e11, 0x1e414a15 ?rev + .long 0x00650560, 0xe683e386, 0x94f191f4, 0x72177712 ?rev +Lk_dsbo: # decryption sbox final output + .long 0x0040f97e, 0x53ea8713, 0x2d3e94d4, 0xb96daac7 ?rev + .long 0x001d4493, 0x0f56d712, 0x9c8ec5d8, 0x59814bca ?rev +Lk_dsb9: # decryption sbox output *9*u, *9*t + .long 0x00d6869a, 0x53031c85, 0xc94c994f, 0x501fd5ca ?rev + .long 0x0049d7ec, 0x89173bc0, 0x65a5fbb2, 0x9e2c5e72 ?rev +Lk_dsbd: # decryption sbox output *D*u, *D*t + .long 0x00a2b1e6, 0xdfcc577d, 0x39442a88, 0x139b6ef5 ?rev + .long 0x00cbc624, 0xf7fae23c, 0xd3efde15, 0x0d183129 ?rev +Lk_dsbb: # decryption sbox output *B*u, *B*t + .long 0x0042b496, 0x926422d0, 0x04d4f2b0, 0xf6462660 ?rev + .long 0x006759cd, 0xa69894c1, 0x6baa5532, 0x3e0cfff3 ?rev +Lk_dsbe: # decryption sbox output *E*u, *E*t + .long 0x00d0d426, 0x9692f246, 0xb0f6b464, 0x04604222 ?rev + .long 0x00c1aaff, 0xcda6550c, 0x323e5998, 0x6bf36794 ?rev + +## +## Key schedule constants +## +Lk_dksd: # decryption key schedule: invskew x*D + .long 0x0047e4a3, 0x5d1ab9fe, 0xf9be1d5a, 0xa4e34007 ?rev + .long 0x008336b5, 0xf477c241, 0x1e9d28ab, 0xea69dc5f ?rev +Lk_dksb: # decryption key schedule: invskew x*B + .long 0x00d55085, 0x1fca4f9a, 0x994cc91c, 0x8653d603 ?rev + .long 0x004afcb6, 0xa7ed5b11, 0xc882347e, 0x6f2593d9 ?rev +Lk_dkse: # decryption key schedule: invskew x*E + 0x63 + .long 0x00d6c91f, 0xca1c03d5, 0x86504f99, 0x4c9a8553 ?rev + .long 0xe87bdc4f, 0x059631a2, 0x8714b320, 0x6af95ecd ?rev +Lk_dks9: # decryption key schedule: invskew x*9 + .long 0x00a7d97e, 0xc86f11b6, 0xfc5b2582, 0x3493ed4a ?rev + .long 0x00331427, 0x62517645, 0xcefddae9, 0xac9fb88b ?rev + +Lk_rcon: # rcon + .long 0xb6ee9daf, 0xb991831f, 0x817d7c4d, 0x08982a70 ?asis +Lk_s63: + .long 0x5b5b5b5b, 0x5b5b5b5b, 0x5b5b5b5b, 0x5b5b5b5b ?asis + +Lk_opt: # output transform + .long 0x0060b6d6, 0x29499fff, 0x0868bede, 0x214197f7 ?rev + .long 0x00ecbc50, 0x51bded01, 0xe00c5cb0, 0xb15d0de1 ?rev +Lk_deskew: # deskew tables: inverts the sbox's "skew" + .long 0x00e3a447, 0x40a3e407, 0x1af9be5d, 0x5ab9fe1d ?rev + .long 0x0069ea83, 0xdcb5365f, 0x771e9df4, 0xabc24128 ?rev +.align 5 +Lconsts: + mflr r0 + bcl 20,31,\$+4 + mflr r12 #vvvvv "distance between . and _vpaes_consts + addi r12,r12,-0x308 + mtlr r0 + blr + .long 0 + .byte 0,12,0x14,0,0,0,0,0 +.asciz "Vector Permutation AES for AltiVec, Mike Hamburg (Stanford University)" +.align 6 +___ + +my ($inptail,$inpperm,$outhead,$outperm,$outmask,$keyperm) = map("v$_",(26..31)); +{ +my ($inp,$out,$key) = map("r$_",(3..5)); + +my ($invlo,$invhi,$iptlo,$ipthi,$sbou,$sbot) = map("v$_",(10..15)); +my ($sb1u,$sb1t,$sb2u,$sb2t) = map("v$_",(16..19)); +my ($sb9u,$sb9t,$sbdu,$sbdt,$sbbu,$sbbt,$sbeu,$sbet)=map("v$_",(16..23)); + +$code.=<<___; +## +## _aes_preheat +## +## Fills register %r10 -> .aes_consts (so you can -fPIC) +## and %xmm9-%xmm15 as specified below. +## +.align 4 +_vpaes_encrypt_preheat: + mflr r8 + bl Lconsts + mtlr r8 + li r11, 0xc0 # Lk_inv + li r10, 0xd0 + li r9, 0xe0 # Lk_ipt + li r8, 0xf0 + vxor v7, v7, v7 # 0x00..00 + vspltisb v8,4 # 0x04..04 + vspltisb v9,0x0f # 0x0f..0f + lvx $invlo, r12, r11 + li r11, 0x100 + lvx $invhi, r12, r10 + li r10, 0x110 + lvx $iptlo, r12, r9 + li r9, 0x120 + lvx $ipthi, r12, r8 + li r8, 0x130 + lvx $sbou, r12, r11 + li r11, 0x140 + lvx $sbot, r12, r10 + li r10, 0x150 + lvx $sb1u, r12, r9 + lvx $sb1t, r12, r8 + lvx $sb2u, r12, r11 + lvx $sb2t, r12, r10 + blr + .long 0 + .byte 0,12,0x14,0,0,0,0,0 + +## +## _aes_encrypt_core +## +## AES-encrypt %xmm0. +## +## Inputs: +## %xmm0 = input +## %xmm9-%xmm15 as in _vpaes_preheat +## (%rdx) = scheduled keys +## +## Output in %xmm0 +## Clobbers %xmm1-%xmm6, %r9, %r10, %r11, %rax +## +## +.align 5 +_vpaes_encrypt_core: + lwz r8, 240($key) # pull rounds + li r9, 16 + lvx v5, 0, $key # vmovdqu (%r9), %xmm5 # round0 key + li r11, 0x10 + lvx v6, r9, $key + addi r9, r9, 16 + ?vperm v5, v5, v6, $keyperm # align round key + addi r10, r11, 0x40 + vsrb v1, v0, v8 # vpsrlb \$4, %xmm0, %xmm0 + vperm v0, $iptlo, $iptlo, v0 # vpshufb %xmm1, %xmm2, %xmm1 + vperm v1, $ipthi, $ipthi, v1 # vpshufb %xmm0, %xmm3, %xmm2 + vxor v0, v0, v5 # vpxor %xmm5, %xmm1, %xmm0 + vxor v0, v0, v1 # vpxor %xmm2, %xmm0, %xmm0 + mtctr r8 + b Lenc_entry + +.align 4 +Lenc_loop: + # middle of middle round + vperm v4, $sb1t, v7, v2 # vpshufb %xmm2, %xmm13, %xmm4 # 4 = sb1u + lvx v1, r12, r11 # vmovdqa -0x40(%r11,%r10), %xmm1 # .Lk_mc_forward[] + addi r11, r11, 16 + vperm v0, $sb1u, v7, v3 # vpshufb %xmm3, %xmm12, %xmm0 # 0 = sb1t + vxor v4, v4, v5 # vpxor %xmm5, %xmm4, %xmm4 # 4 = sb1u + k + andi. r11, r11, 0x30 # and \$0x30, %r11 # ... mod 4 + vperm v5, $sb2t, v7, v2 # vpshufb %xmm2, %xmm15, %xmm5 # 4 = sb2u + vxor v0, v0, v4 # vpxor %xmm4, %xmm0, %xmm0 # 0 = A + vperm v2, $sb2u, v7, v3 # vpshufb %xmm3, %xmm14, %xmm2 # 2 = sb2t + lvx v4, r12, r10 # vmovdqa (%r11,%r10), %xmm4 # .Lk_mc_backward[] + addi r10, r11, 0x40 + vperm v3, v0, v7, v1 # vpshufb %xmm1, %xmm0, %xmm3 # 0 = B + vxor v2, v2, v5 # vpxor %xmm5, %xmm2, %xmm2 # 2 = 2A + vperm v0, v0, v7, v4 # vpshufb %xmm4, %xmm0, %xmm0 # 3 = D + vxor v3, v3, v2 # vpxor %xmm2, %xmm3, %xmm3 # 0 = 2A+B + vperm v4, v3, v7, v1 # vpshufb %xmm1, %xmm3, %xmm4 # 0 = 2B+C + vxor v0, v0, v3 # vpxor %xmm3, %xmm0, %xmm0 # 3 = 2A+B+D + vxor v0, v0, v4 # vpxor %xmm4, %xmm0, %xmm0 # 0 = 2A+3B+C+D + +Lenc_entry: + # top of round + vsrb v1, v0, v8 # vpsrlb \$4, %xmm0, %xmm0 # 1 = i + vperm v5, $invhi, $invhi, v0 # vpshufb %xmm1, %xmm11, %xmm5 # 2 = a/k + vxor v0, v0, v1 # vpxor %xmm0, %xmm1, %xmm1 # 0 = j + vperm v3, $invlo, $invlo, v1 # vpshufb %xmm0, %xmm10, %xmm3 # 3 = 1/i + vperm v4, $invlo, $invlo, v0 # vpshufb %xmm1, %xmm10, %xmm4 # 4 = 1/j + vand v0, v0, v9 + vxor v3, v3, v5 # vpxor %xmm5, %xmm3, %xmm3 # 3 = iak = 1/i + a/k + vxor v4, v4, v5 # vpxor %xmm5, %xmm4, %xmm4 # 4 = jak = 1/j + a/k + vperm v2, $invlo, v7, v3 # vpshufb %xmm3, %xmm10, %xmm2 # 2 = 1/iak + vmr v5, v6 + lvx v6, r9, $key # vmovdqu (%r9), %xmm5 + vperm v3, $invlo, v7, v4 # vpshufb %xmm4, %xmm10, %xmm3 # 3 = 1/jak + addi r9, r9, 16 + vxor v2, v2, v0 # vpxor %xmm1, %xmm2, %xmm2 # 2 = io + ?vperm v5, v5, v6, $keyperm # align round key + vxor v3, v3, v1 # vpxor %xmm0, %xmm3, %xmm3 # 3 = jo + bdnz Lenc_loop + + # middle of last round + addi r10, r11, 0x80 + # vmovdqa -0x60(%r10), %xmm4 # 3 : sbou .Lk_sbo + # vmovdqa -0x50(%r10), %xmm0 # 0 : sbot .Lk_sbo+16 + vperm v4, $sbou, v7, v2 # vpshufb %xmm2, %xmm4, %xmm4 # 4 = sbou + lvx v1, r12, r10 # vmovdqa 0x40(%r11,%r10), %xmm1 # .Lk_sr[] + vperm v0, $sbot, v7, v3 # vpshufb %xmm3, %xmm0, %xmm0 # 0 = sb1t + vxor v4, v4, v5 # vpxor %xmm5, %xmm4, %xmm4 # 4 = sb1u + k + vxor v0, v0, v4 # vpxor %xmm4, %xmm0, %xmm0 # 0 = A + vperm v0, v0, v7, v1 # vpshufb %xmm1, %xmm0, %xmm0 + blr + .long 0 + .byte 0,12,0x14,0,0,0,0,0 + +.globl .vpaes_encrypt +.align 5 +.vpaes_encrypt: + $STU $sp,-$FRAME($sp) + li r10,`15+6*$SIZE_T` + li r11,`31+6*$SIZE_T` + mflr r6 + mfspr r7, 256 # save vrsave + stvx v20,r10,$sp + addi r10,r10,32 + stvx v21,r11,$sp + addi r11,r11,32 + stvx v22,r10,$sp + addi r10,r10,32 + stvx v23,r11,$sp + addi r11,r11,32 + stvx v24,r10,$sp + addi r10,r10,32 + stvx v25,r11,$sp + addi r11,r11,32 + stvx v26,r10,$sp + addi r10,r10,32 + stvx v27,r11,$sp + addi r11,r11,32 + stvx v28,r10,$sp + addi r10,r10,32 + stvx v29,r11,$sp + addi r11,r11,32 + stvx v30,r10,$sp + stvx v31,r11,$sp + stw r7,`$FRAME-4`($sp) # save vrsave + li r0, -1 + $PUSH r6,`$FRAME+$LRSAVE`($sp) + mtspr 256, r0 # preserve all AltiVec registers + + bl _vpaes_encrypt_preheat + + ?lvsl $inpperm, 0, $inp # prepare for unaligned access + lvx v0, 0, $inp + addi $inp, $inp, 15 # 15 is not a typo + ?lvsr $outperm, 0, $out + ?lvsl $keyperm, 0, $key # prepare for unaligned access + lvx $inptail, 0, $inp # redundant in aligned case + ?vperm v0, v0, $inptail, $inpperm + + bl _vpaes_encrypt_core + + andi. r8, $out, 15 + li r9, 16 + beq Lenc_out_aligned + + vperm v0, v0, v0, $outperm # rotate right/left + mtctr r9 +Lenc_out_unaligned: + stvebx v0, 0, $out + addi $out, $out, 1 + bdnz Lenc_out_unaligned + b Lenc_done + +.align 4 +Lenc_out_aligned: + stvx v0, 0, $out +Lenc_done: + + li r10,`15+6*$SIZE_T` + li r11,`31+6*$SIZE_T` + mtlr r6 + mtspr 256, r7 # restore vrsave + lvx v20,r10,$sp + addi r10,r10,32 + lvx v21,r11,$sp + addi r11,r11,32 + lvx v22,r10,$sp + addi r10,r10,32 + lvx v23,r11,$sp + addi r11,r11,32 + lvx v24,r10,$sp + addi r10,r10,32 + lvx v25,r11,$sp + addi r11,r11,32 + lvx v26,r10,$sp + addi r10,r10,32 + lvx v27,r11,$sp + addi r11,r11,32 + lvx v28,r10,$sp + addi r10,r10,32 + lvx v29,r11,$sp + addi r11,r11,32 + lvx v30,r10,$sp + lvx v31,r11,$sp + addi $sp,$sp,$FRAME + blr + .long 0 + .byte 0,12,0x04,1,0x80,0,3,0 + .long 0 +.size .vpaes_encrypt,.-.vpaes_encrypt + +.align 4 +_vpaes_decrypt_preheat: + mflr r8 + bl Lconsts + mtlr r8 + li r11, 0xc0 # Lk_inv + li r10, 0xd0 + li r9, 0x160 # Ldipt + li r8, 0x170 + vxor v7, v7, v7 # 0x00..00 + vspltisb v8,4 # 0x04..04 + vspltisb v9,0x0f # 0x0f..0f + lvx $invlo, r12, r11 + li r11, 0x180 + lvx $invhi, r12, r10 + li r10, 0x190 + lvx $iptlo, r12, r9 + li r9, 0x1a0 + lvx $ipthi, r12, r8 + li r8, 0x1b0 + lvx $sbou, r12, r11 + li r11, 0x1c0 + lvx $sbot, r12, r10 + li r10, 0x1d0 + lvx $sb9u, r12, r9 + li r9, 0x1e0 + lvx $sb9t, r12, r8 + li r8, 0x1f0 + lvx $sbdu, r12, r11 + li r11, 0x200 + lvx $sbdt, r12, r10 + li r10, 0x210 + lvx $sbbu, r12, r9 + lvx $sbbt, r12, r8 + lvx $sbeu, r12, r11 + lvx $sbet, r12, r10 + blr + .long 0 + .byte 0,12,0x14,0,0,0,0,0 + +## +## Decryption core +## +## Same API as encryption core. +## +.align 4 +_vpaes_decrypt_core: + lwz r8, 240($key) # pull rounds + li r9, 16 + lvx v5, 0, $key # vmovdqu (%r9), %xmm4 # round0 key + li r11, 0x30 + lvx v6, r9, $key + addi r9, r9, 16 + ?vperm v5, v5, v6, $keyperm # align round key + vsrb v1, v0, v8 # vpsrlb \$4, %xmm0, %xmm0 + vperm v0, $iptlo, $iptlo, v0 # vpshufb %xmm1, %xmm2, %xmm2 + vperm v1, $ipthi, $ipthi, v1 # vpshufb %xmm0, %xmm1, %xmm0 + vxor v0, v0, v5 # vpxor %xmm4, %xmm2, %xmm2 + vxor v0, v0, v1 # vpxor %xmm2, %xmm0, %xmm0 + mtctr r8 + b Ldec_entry + +.align 4 +Ldec_loop: +# +# Inverse mix columns +# + lvx v0, r12, r11 # v5 and v0 are flipped + # vmovdqa -0x20(%r10),%xmm4 # 4 : sb9u + # vmovdqa -0x10(%r10),%xmm1 # 0 : sb9t + vperm v4, $sb9u, v7, v2 # vpshufb %xmm2, %xmm4, %xmm4 # 4 = sb9u + subi r11, r11, 16 + vperm v1, $sb9t, v7, v3 # vpshufb %xmm3, %xmm1, %xmm1 # 0 = sb9t + andi. r11, r11, 0x30 + vxor v5, v5, v4 # vpxor %xmm4, %xmm0, %xmm0 + # vmovdqa 0x00(%r10),%xmm4 # 4 : sbdu + vxor v5, v5, v1 # vpxor %xmm1, %xmm0, %xmm0 # 0 = ch + # vmovdqa 0x10(%r10),%xmm1 # 0 : sbdt + + vperm v4, $sbdu, v7, v2 # vpshufb %xmm2, %xmm4, %xmm4 # 4 = sbdu + vperm v5, v5, v7, v0 # vpshufb %xmm5, %xmm0, %xmm0 # MC ch + vperm v1, $sbdt, v7, v3 # vpshufb %xmm3, %xmm1, %xmm1 # 0 = sbdt + vxor v5, v5, v4 # vpxor %xmm4, %xmm0, %xmm0 # 4 = ch + # vmovdqa 0x20(%r10), %xmm4 # 4 : sbbu + vxor v5, v5, v1 # vpxor %xmm1, %xmm0, %xmm0 # 0 = ch + # vmovdqa 0x30(%r10), %xmm1 # 0 : sbbt + + vperm v4, $sbbu, v7, v2 # vpshufb %xmm2, %xmm4, %xmm4 # 4 = sbbu + vperm v5, v5, v7, v0 # vpshufb %xmm5, %xmm0, %xmm0 # MC ch + vperm v1, $sbbt, v7, v3 # vpshufb %xmm3, %xmm1, %xmm1 # 0 = sbbt + vxor v5, v5, v4 # vpxor %xmm4, %xmm0, %xmm0 # 4 = ch + # vmovdqa 0x40(%r10), %xmm4 # 4 : sbeu + vxor v5, v5, v1 # vpxor %xmm1, %xmm0, %xmm0 # 0 = ch + # vmovdqa 0x50(%r10), %xmm1 # 0 : sbet + + vperm v4, $sbeu, v7, v2 # vpshufb %xmm2, %xmm4, %xmm4 # 4 = sbeu + vperm v5, v5, v7, v0 # vpshufb %xmm5, %xmm0, %xmm0 # MC ch + vperm v1, $sbet, v7, v3 # vpshufb %xmm3, %xmm1, %xmm1 # 0 = sbet + vxor v0, v5, v4 # vpxor %xmm4, %xmm0, %xmm0 # 4 = ch + vxor v0, v0, v1 # vpxor %xmm1, %xmm0, %xmm0 # 0 = ch + +Ldec_entry: + # top of round + vsrb v1, v0, v8 # vpsrlb \$4, %xmm0, %xmm0 # 1 = i + vperm v2, $invhi, $invhi, v0 # vpshufb %xmm1, %xmm11, %xmm2 # 2 = a/k + vxor v0, v0, v1 # vpxor %xmm0, %xmm1, %xmm1 # 0 = j + vperm v3, $invlo, $invlo, v1 # vpshufb %xmm0, %xmm10, %xmm3 # 3 = 1/i + vperm v4, $invlo, $invlo, v0 # vpshufb %xmm1, %xmm10, %xmm4 # 4 = 1/j + vand v0, v0, v9 + vxor v3, v3, v2 # vpxor %xmm2, %xmm3, %xmm3 # 3 = iak = 1/i + a/k + vxor v4, v4, v2 # vpxor %xmm2, %xmm4, %xmm4 # 4 = jak = 1/j + a/k + vperm v2, $invlo, v7, v3 # vpshufb %xmm3, %xmm10, %xmm2 # 2 = 1/iak + vmr v5, v6 + lvx v6, r9, $key # vmovdqu (%r9), %xmm0 + vperm v3, $invlo, v7, v4 # vpshufb %xmm4, %xmm10, %xmm3 # 3 = 1/jak + addi r9, r9, 16 + vxor v2, v2, v0 # vpxor %xmm1, %xmm2, %xmm2 # 2 = io + ?vperm v5, v5, v6, $keyperm # align round key + vxor v3, v3, v1 # vpxor %xmm0, %xmm3, %xmm3 # 3 = jo + bdnz Ldec_loop + + # middle of last round + addi r10, r11, 0x80 + # vmovdqa 0x60(%r10), %xmm4 # 3 : sbou + vperm v4, $sbou, v7, v2 # vpshufb %xmm2, %xmm4, %xmm4 # 4 = sbou + # vmovdqa 0x70(%r10), %xmm1 # 0 : sbot + lvx v2, r12, r10 # vmovdqa -0x160(%r11), %xmm2 # .Lk_sr-.Lk_dsbd=-0x160 + vperm v1, $sbot, v7, v3 # vpshufb %xmm3, %xmm1, %xmm1 # 0 = sb1t + vxor v4, v4, v5 # vpxor %xmm0, %xmm4, %xmm4 # 4 = sb1u + k + vxor v0, v1, v4 # vpxor %xmm4, %xmm1, %xmm0 # 0 = A + vperm v0, v0, v7, v2 # vpshufb %xmm2, %xmm0, %xmm0 + blr + .long 0 + .byte 0,12,0x14,0,0,0,0,0 + +.globl .vpaes_decrypt +.align 5 +.vpaes_decrypt: + $STU $sp,-$FRAME($sp) + li r10,`15+6*$SIZE_T` + li r11,`31+6*$SIZE_T` + mflr r6 + mfspr r7, 256 # save vrsave + stvx v20,r10,$sp + addi r10,r10,32 + stvx v21,r11,$sp + addi r11,r11,32 + stvx v22,r10,$sp + addi r10,r10,32 + stvx v23,r11,$sp + addi r11,r11,32 + stvx v24,r10,$sp + addi r10,r10,32 + stvx v25,r11,$sp + addi r11,r11,32 + stvx v26,r10,$sp + addi r10,r10,32 + stvx v27,r11,$sp + addi r11,r11,32 + stvx v28,r10,$sp + addi r10,r10,32 + stvx v29,r11,$sp + addi r11,r11,32 + stvx v30,r10,$sp + stvx v31,r11,$sp + stw r7,`$FRAME-4`($sp) # save vrsave + li r0, -1 + $PUSH r6,`$FRAME+$LRSAVE`($sp) + mtspr 256, r0 # preserve all AltiVec registers + + bl _vpaes_decrypt_preheat + + ?lvsl $inpperm, 0, $inp # prepare for unaligned access + lvx v0, 0, $inp + addi $inp, $inp, 15 # 15 is not a typo + ?lvsr $outperm, 0, $out + ?lvsl $keyperm, 0, $key + lvx $inptail, 0, $inp # redundant in aligned case + ?vperm v0, v0, $inptail, $inpperm + + bl _vpaes_decrypt_core + + andi. r8, $out, 15 + li r9, 16 + beq Ldec_out_aligned + + vperm v0, v0, v0, $outperm # rotate right/left + mtctr r9 +Ldec_out_unaligned: + stvebx v0, 0, $out + addi $out, $out, 1 + bdnz Ldec_out_unaligned + b Ldec_done + +.align 4 +Ldec_out_aligned: + stvx v0, 0, $out +Ldec_done: + + li r10,`15+6*$SIZE_T` + li r11,`31+6*$SIZE_T` + mtlr r6 + mtspr 256, r7 # restore vrsave + lvx v20,r10,$sp + addi r10,r10,32 + lvx v21,r11,$sp + addi r11,r11,32 + lvx v22,r10,$sp + addi r10,r10,32 + lvx v23,r11,$sp + addi r11,r11,32 + lvx v24,r10,$sp + addi r10,r10,32 + lvx v25,r11,$sp + addi r11,r11,32 + lvx v26,r10,$sp + addi r10,r10,32 + lvx v27,r11,$sp + addi r11,r11,32 + lvx v28,r10,$sp + addi r10,r10,32 + lvx v29,r11,$sp + addi r11,r11,32 + lvx v30,r10,$sp + lvx v31,r11,$sp + addi $sp,$sp,$FRAME + blr + .long 0 + .byte 0,12,0x04,1,0x80,0,3,0 + .long 0 +.size .vpaes_decrypt,.-.vpaes_decrypt + +.globl .vpaes_cbc_encrypt +.align 5 +.vpaes_cbc_encrypt: + ${UCMP}i r5,16 + bltlr- + + $STU $sp,-`($FRAME+2*$SIZE_T)`($sp) + mflr r0 + li r10,`15+6*$SIZE_T` + li r11,`31+6*$SIZE_T` + mfspr r12, 256 + stvx v20,r10,$sp + addi r10,r10,32 + stvx v21,r11,$sp + addi r11,r11,32 + stvx v22,r10,$sp + addi r10,r10,32 + stvx v23,r11,$sp + addi r11,r11,32 + stvx v24,r10,$sp + addi r10,r10,32 + stvx v25,r11,$sp + addi r11,r11,32 + stvx v26,r10,$sp + addi r10,r10,32 + stvx v27,r11,$sp + addi r11,r11,32 + stvx v28,r10,$sp + addi r10,r10,32 + stvx v29,r11,$sp + addi r11,r11,32 + stvx v30,r10,$sp + stvx v31,r11,$sp + stw r12,`$FRAME-4`($sp) # save vrsave + $PUSH r30,`$FRAME+$SIZE_T*0`($sp) + $PUSH r31,`$FRAME+$SIZE_T*1`($sp) + li r9, -16 + $PUSH r0, `$FRAME+$SIZE_T*2+$LRSAVE`($sp) + + and r30, r5, r9 # copy length&-16 + andi. r9, $out, 15 # is $out aligned? + mr r5, r6 # copy pointer to key + mr r31, r7 # copy pointer to iv + li r6, -1 + mcrf cr1, cr0 # put aside $out alignment flag + mr r7, r12 # copy vrsave + mtspr 256, r6 # preserve all AltiVec registers + + lvx v24, 0, r31 # load [potentially unaligned] iv + li r9, 15 + ?lvsl $inpperm, 0, r31 + lvx v25, r9, r31 + ?vperm v24, v24, v25, $inpperm + + cmpwi r8, 0 # test direction + neg r8, $inp # prepare for unaligned access + vxor v7, v7, v7 + ?lvsl $keyperm, 0, $key + ?lvsr $outperm, 0, $out + ?lvsr $inpperm, 0, r8 # -$inp + vnor $outmask, v7, v7 # 0xff..ff + lvx $inptail, 0, $inp + ?vperm $outmask, v7, $outmask, $outperm + addi $inp, $inp, 15 # 15 is not a typo + + beq Lcbc_decrypt + + bl _vpaes_encrypt_preheat + li r0, 16 + + beq cr1, Lcbc_enc_loop # $out is aligned + + vmr v0, $inptail + lvx $inptail, 0, $inp + addi $inp, $inp, 16 + ?vperm v0, v0, $inptail, $inpperm + vxor v0, v0, v24 # ^= iv + + bl _vpaes_encrypt_core + + andi. r8, $out, 15 + vmr v24, v0 # put aside iv + sub r9, $out, r8 + vperm $outhead, v0, v0, $outperm # rotate right/left + +Lcbc_enc_head: + stvebx $outhead, r8, r9 + cmpwi r8, 15 + addi r8, r8, 1 + bne Lcbc_enc_head + + sub. r30, r30, r0 # len -= 16 + addi $out, $out, 16 + beq Lcbc_unaligned_done + +Lcbc_enc_loop: + vmr v0, $inptail + lvx $inptail, 0, $inp + addi $inp, $inp, 16 + ?vperm v0, v0, $inptail, $inpperm + vxor v0, v0, v24 # ^= iv + + bl _vpaes_encrypt_core + + vmr v24, v0 # put aside iv + sub. r30, r30, r0 # len -= 16 + vperm v0, v0, v0, $outperm # rotate right/left + vsel v1, $outhead, v0, $outmask + vmr $outhead, v0 + stvx v1, 0, $out + addi $out, $out, 16 + bne Lcbc_enc_loop + + b Lcbc_done + +.align 5 +Lcbc_decrypt: + bl _vpaes_decrypt_preheat + li r0, 16 + + beq cr1, Lcbc_dec_loop # $out is aligned + + vmr v0, $inptail + lvx $inptail, 0, $inp + addi $inp, $inp, 16 + ?vperm v0, v0, $inptail, $inpperm + vmr v25, v0 # put aside input + + bl _vpaes_decrypt_core + + andi. r8, $out, 15 + vxor v0, v0, v24 # ^= iv + vmr v24, v25 + sub r9, $out, r8 + vperm $outhead, v0, v0, $outperm # rotate right/left + +Lcbc_dec_head: + stvebx $outhead, r8, r9 + cmpwi r8, 15 + addi r8, r8, 1 + bne Lcbc_dec_head + + sub. r30, r30, r0 # len -= 16 + addi $out, $out, 16 + beq Lcbc_unaligned_done + +Lcbc_dec_loop: + vmr v0, $inptail + lvx $inptail, 0, $inp + addi $inp, $inp, 16 + ?vperm v0, v0, $inptail, $inpperm + vmr v25, v0 # put aside input + + bl _vpaes_decrypt_core + + vxor v0, v0, v24 # ^= iv + vmr v24, v25 + sub. r30, r30, r0 # len -= 16 + vperm v0, v0, v0, $outperm # rotate right/left + vsel v1, $outhead, v0, $outmask + vmr $outhead, v0 + stvx v1, 0, $out + addi $out, $out, 16 + bne Lcbc_dec_loop + +Lcbc_done: + beq cr1, Lcbc_write_iv # $out is aligned + +Lcbc_unaligned_done: + andi. r8, $out, 15 + sub $out, $out, r8 + li r9, 0 +Lcbc_tail: + stvebx $outhead, r9, $out + addi r9, r9, 1 + cmpw r9, r8 + bne Lcbc_tail + +Lcbc_write_iv: + neg r8, r31 # write [potentially unaligned] iv + li r10, 4 + ?lvsl $outperm, 0, r8 + li r11, 8 + li r12, 12 + vperm v24, v24, v24, $outperm # rotate right/left + stvewx v24, 0, r31 # ivp is at least 32-bit aligned + stvewx v24, r10, r31 + stvewx v24, r11, r31 + stvewx v24, r12, r31 + + mtspr 256, r7 # restore vrsave + li r10,`15+6*$SIZE_T` + li r11,`31+6*$SIZE_T` + lvx v20,r10,$sp + addi r10,r10,32 + lvx v21,r11,$sp + addi r11,r11,32 + lvx v22,r10,$sp + addi r10,r10,32 + lvx v23,r11,$sp + addi r11,r11,32 + lvx v24,r10,$sp + addi r10,r10,32 + lvx v25,r11,$sp + addi r11,r11,32 + lvx v26,r10,$sp + addi r10,r10,32 + lvx v27,r11,$sp + addi r11,r11,32 + lvx v28,r10,$sp + addi r10,r10,32 + lvx v29,r11,$sp + addi r11,r11,32 + lvx v30,r10,$sp + lvx v31,r11,$sp +Lcbc_abort: + $POP r0, `$FRAME+$SIZE_T*2+$LRSAVE`($sp) + $POP r30,`$FRAME+$SIZE_T*0`($sp) + $POP r31,`$FRAME+$SIZE_T*1`($sp) + mtlr r0 + addi $sp,$sp,`$FRAME+$SIZE_T*2` + blr + .long 0 + .byte 0,12,0x04,1,0x80,2,6,0 + .long 0 +.size .vpaes_cbc_encrypt,.-.vpaes_cbc_encrypt +___ +} +{ +my ($inp,$bits,$out)=map("r$_",(3..5)); +my $dir="cr1"; +my ($invlo,$invhi,$iptlo,$ipthi,$rcon) = map("v$_",(10..13,24)); + +$code.=<<___; +######################################################## +## ## +## AES key schedule ## +## ## +######################################################## +.align 4 +_vpaes_key_preheat: + mflr r8 + bl Lconsts + mtlr r8 + li r11, 0xc0 # Lk_inv + li r10, 0xd0 + li r9, 0xe0 # L_ipt + li r8, 0xf0 + + vspltisb v8,4 # 0x04..04 + vxor v9,v9,v9 # 0x00..00 + lvx $invlo, r12, r11 # Lk_inv + li r11, 0x120 + lvx $invhi, r12, r10 + li r10, 0x130 + lvx $iptlo, r12, r9 # Lk_ipt + li r9, 0x220 + lvx $ipthi, r12, r8 + li r8, 0x230 + + lvx v14, r12, r11 # Lk_sb1 + li r11, 0x240 + lvx v15, r12, r10 + li r10, 0x250 + + lvx v16, r12, r9 # Lk_dksd + li r9, 0x260 + lvx v17, r12, r8 + li r8, 0x270 + lvx v18, r12, r11 # Lk_dksb + li r11, 0x280 + lvx v19, r12, r10 + li r10, 0x290 + lvx v20, r12, r9 # Lk_dkse + li r9, 0x2a0 + lvx v21, r12, r8 + li r8, 0x2b0 + lvx v22, r12, r11 # Lk_dks9 + lvx v23, r12, r10 + + lvx v24, r12, r9 # Lk_rcon + lvx v25, 0, r12 # Lk_mc_forward[0] + lvx v26, r12, r8 # Lks63 + blr + .long 0 + .byte 0,12,0x14,0,0,0,0,0 + +.align 4 +_vpaes_schedule_core: + mflr r7 + + bl _vpaes_key_preheat # load the tables + + #lvx v0, 0, $inp # vmovdqu (%rdi), %xmm0 # load key (unaligned) + neg r8, $inp # prepare for unaligned access + lvx v0, 0, $inp + addi $inp, $inp, 15 # 15 is not typo + ?lvsr $inpperm, 0, r8 # -$inp + lvx v6, 0, $inp # v6 serves as inptail + addi $inp, $inp, 8 + ?vperm v0, v0, v6, $inpperm + + # input transform + vmr v3, v0 # vmovdqa %xmm0, %xmm3 + bl _vpaes_schedule_transform + vmr v7, v0 # vmovdqa %xmm0, %xmm7 + + bne $dir, Lschedule_am_decrypting + + # encrypting, output zeroth round key after transform + li r8, 0x30 # mov \$0x30,%r8d + li r9, 4 + li r10, 8 + li r11, 12 + + ?lvsr $outperm, 0, $out # prepare for unaligned access + vnor $outmask, v9, v9 # 0xff..ff + ?vperm $outmask, v9, $outmask, $outperm + + #stvx v0, 0, $out # vmovdqu %xmm0, (%rdx) + vperm $outhead, v0, v0, $outperm # rotate right/left + stvewx $outhead, 0, $out # some are superfluous + stvewx $outhead, r9, $out + stvewx $outhead, r10, $out + addi r10, r12, 0x80 # lea .Lk_sr(%rip),%r10 + stvewx $outhead, r11, $out + b Lschedule_go + +Lschedule_am_decrypting: + srwi r8, $bits, 1 # shr \$1,%r8d + andi. r8, r8, 32 # and \$32,%r8d + xori r8, r8, 32 # xor \$32,%r8d # nbits==192?0:32 + addi r10, r12, 0x80 # lea .Lk_sr(%rip),%r10 + # decrypting, output zeroth round key after shiftrows + lvx v1, r8, r10 # vmovdqa (%r8,%r10), %xmm1 + li r9, 4 + li r10, 8 + li r11, 12 + vperm v4, v3, v3, v1 # vpshufb %xmm1, %xmm3, %xmm3 + + neg r0, $out # prepare for unaligned access + ?lvsl $outperm, 0, r0 + vnor $outmask, v9, v9 # 0xff..ff + ?vperm $outmask, $outmask, v9, $outperm + + #stvx v4, 0, $out # vmovdqu %xmm3, (%rdx) + vperm $outhead, v4, v4, $outperm # rotate right/left + stvewx $outhead, 0, $out # some are superfluous + stvewx $outhead, r9, $out + stvewx $outhead, r10, $out + addi r10, r12, 0x80 # lea .Lk_sr(%rip),%r10 + stvewx $outhead, r11, $out + addi $out, $out, 15 # 15 is not typo + xori r8, r8, 0x30 # xor \$0x30, %r8 + +Lschedule_go: + cmplwi $bits, 192 # cmp \$192, %esi + bgt Lschedule_256 + beq Lschedule_192 + # 128: fall though + +## +## .schedule_128 +## +## 128-bit specific part of key schedule. +## +## This schedule is really simple, because all its parts +## are accomplished by the subroutines. +## +Lschedule_128: + li r0, 10 # mov \$10, %esi + mtctr r0 + +Loop_schedule_128: + bl _vpaes_schedule_round + bdz Lschedule_mangle_last # dec %esi + bl _vpaes_schedule_mangle # write output + b Loop_schedule_128 + +## +## .aes_schedule_192 +## +## 192-bit specific part of key schedule. +## +## The main body of this schedule is the same as the 128-bit +## schedule, but with more smearing. The long, high side is +## stored in %xmm7 as before, and the short, low side is in +## the high bits of %xmm6. +## +## This schedule is somewhat nastier, however, because each +## round produces 192 bits of key material, or 1.5 round keys. +## Therefore, on each cycle we do 2 rounds and produce 3 round +## keys. +## +.align 4 +Lschedule_192: + li r0, 4 # mov \$4, %esi + lvx v0, 0, $inp + ?vperm v0, v6, v0, $inpperm + ?vsldoi v0, v3, v0, 8 # vmovdqu 8(%rdi),%xmm0 # load key part 2 (very unaligned) + bl _vpaes_schedule_transform # input transform + ?vsldoi v6, v0, v9, 8 + ?vsldoi v6, v9, v6, 8 # clobber "low" side with zeros + mtctr r0 + +Loop_schedule_192: + bl _vpaes_schedule_round + ?vsldoi v0, v6, v0, 8 # vpalignr \$8,%xmm6,%xmm0,%xmm0 + bl _vpaes_schedule_mangle # save key n + bl _vpaes_schedule_192_smear + bl _vpaes_schedule_mangle # save key n+1 + bl _vpaes_schedule_round + bdz Lschedule_mangle_last # dec %esi + bl _vpaes_schedule_mangle # save key n+2 + bl _vpaes_schedule_192_smear + b Loop_schedule_192 + +## +## .aes_schedule_256 +## +## 256-bit specific part of key schedule. +## +## The structure here is very similar to the 128-bit +## schedule, but with an additional "low side" in +## %xmm6. The low side's rounds are the same as the +## high side's, except no rcon and no rotation. +## +.align 4 +Lschedule_256: + li r0, 7 # mov \$7, %esi + addi $inp, $inp, 8 + lvx v0, 0, $inp # vmovdqu 16(%rdi),%xmm0 # load key part 2 (unaligned) + ?vperm v0, v6, v0, $inpperm + bl _vpaes_schedule_transform # input transform + mtctr r0 + +Loop_schedule_256: + bl _vpaes_schedule_mangle # output low result + vmr v6, v0 # vmovdqa %xmm0, %xmm6 # save cur_lo in xmm6 + + # high round + bl _vpaes_schedule_round + bdz Lschedule_mangle_last # dec %esi + bl _vpaes_schedule_mangle + + # low round. swap xmm7 and xmm6 + ?vspltw v0, v0, 3 # vpshufd \$0xFF, %xmm0, %xmm0 + vmr v5, v7 # vmovdqa %xmm7, %xmm5 + vmr v7, v6 # vmovdqa %xmm6, %xmm7 + bl _vpaes_schedule_low_round + vmr v7, v5 # vmovdqa %xmm5, %xmm7 + + b Loop_schedule_256 +## +## .aes_schedule_mangle_last +## +## Mangler for last round of key schedule +## Mangles %xmm0 +## when encrypting, outputs out(%xmm0) ^ 63 +## when decrypting, outputs unskew(%xmm0) +## +## Always called right before return... jumps to cleanup and exits +## +.align 4 +Lschedule_mangle_last: + # schedule last round key from xmm0 + li r11, 0x2e0 # lea .Lk_deskew(%rip),%r11 + li r9, 0x2f0 + bne $dir, Lschedule_mangle_last_dec + + # encrypting + lvx v1, r8, r10 # vmovdqa (%r8,%r10),%xmm1 + li r11, 0x2c0 # lea .Lk_opt(%rip), %r11 # prepare to output transform + li r9, 0x2d0 # prepare to output transform + vperm v0, v0, v0, v1 # vpshufb %xmm1, %xmm0, %xmm0 # output permute + + lvx $iptlo, r11, r12 # reload $ipt + lvx $ipthi, r9, r12 + addi $out, $out, 16 # add \$16, %rdx + vxor v0, v0, v26 # vpxor .Lk_s63(%rip), %xmm0, %xmm0 + bl _vpaes_schedule_transform # output transform + + #stvx v0, r0, $out # vmovdqu %xmm0, (%rdx) # save last key + vperm v0, v0, v0, $outperm # rotate right/left + li r10, 4 + vsel v2, $outhead, v0, $outmask + li r11, 8 + stvx v2, 0, $out + li r12, 12 + stvewx v0, 0, $out # some (or all) are redundant + stvewx v0, r10, $out + stvewx v0, r11, $out + stvewx v0, r12, $out + b Lschedule_mangle_done + +.align 4 +Lschedule_mangle_last_dec: + lvx $iptlo, r11, r12 # reload $ipt + lvx $ipthi, r9, r12 + addi $out, $out, -16 # add \$-16, %rdx + vxor v0, v0, v26 # vpxor .Lk_s63(%rip), %xmm0, %xmm0 + bl _vpaes_schedule_transform # output transform + + #stvx v0, r0, $out # vmovdqu %xmm0, (%rdx) # save last key + addi r9, $out, -15 # -15 is not typo + vperm v0, v0, v0, $outperm # rotate right/left + li r10, 4 + vsel v2, $outhead, v0, $outmask + li r11, 8 + stvx v2, 0, $out + li r12, 12 + stvewx v0, 0, r9 # some (or all) are redundant + stvewx v0, r10, r9 + stvewx v0, r11, r9 + stvewx v0, r12, r9 + + +Lschedule_mangle_done: + mtlr r7 + # cleanup + vxor v0, v0, v0 # vpxor %xmm0, %xmm0, %xmm0 + vxor v1, v1, v1 # vpxor %xmm1, %xmm1, %xmm1 + vxor v2, v2, v2 # vpxor %xmm2, %xmm2, %xmm2 + vxor v3, v3, v3 # vpxor %xmm3, %xmm3, %xmm3 + vxor v4, v4, v4 # vpxor %xmm4, %xmm4, %xmm4 + vxor v5, v5, v5 # vpxor %xmm5, %xmm5, %xmm5 + vxor v6, v6, v6 # vpxor %xmm6, %xmm6, %xmm6 + vxor v7, v7, v7 # vpxor %xmm7, %xmm7, %xmm7 + + blr + .long 0 + .byte 0,12,0x14,0,0,0,0,0 + +## +## .aes_schedule_192_smear +## +## Smear the short, low side in the 192-bit key schedule. +## +## Inputs: +## %xmm7: high side, b a x y +## %xmm6: low side, d c 0 0 +## %xmm13: 0 +## +## Outputs: +## %xmm6: b+c+d b+c 0 0 +## %xmm0: b+c+d b+c b a +## +.align 4 +_vpaes_schedule_192_smear: + ?vspltw v0, v7, 3 + ?vsldoi v1, v9, v6, 12 # vpshufd \$0x80, %xmm6, %xmm1 # d c 0 0 -> c 0 0 0 + ?vsldoi v0, v7, v0, 8 # vpshufd \$0xFE, %xmm7, %xmm0 # b a _ _ -> b b b a + vxor v6, v6, v1 # vpxor %xmm1, %xmm6, %xmm6 # -> c+d c 0 0 + vxor v6, v6, v0 # vpxor %xmm0, %xmm6, %xmm6 # -> b+c+d b+c b a + vmr v0, v6 + ?vsldoi v6, v6, v9, 8 + ?vsldoi v6, v9, v6, 8 # clobber low side with zeros + blr + .long 0 + .byte 0,12,0x14,0,0,0,0,0 + +## +## .aes_schedule_round +## +## Runs one main round of the key schedule on %xmm0, %xmm7 +## +## Specifically, runs subbytes on the high dword of %xmm0 +## then rotates it by one byte and xors into the low dword of +## %xmm7. +## +## Adds rcon from low byte of %xmm8, then rotates %xmm8 for +## next rcon. +## +## Smears the dwords of %xmm7 by xoring the low into the +## second low, result into third, result into highest. +## +## Returns results in %xmm7 = %xmm0. +## Clobbers %xmm1-%xmm4, %r11. +## +.align 4 +_vpaes_schedule_round: + # extract rcon from xmm8 + #vxor v4, v4, v4 # vpxor %xmm4, %xmm4, %xmm4 + ?vsldoi v1, $rcon, v9, 15 # vpalignr \$15, %xmm8, %xmm4, %xmm1 + ?vsldoi $rcon, $rcon, $rcon, 15 # vpalignr \$15, %xmm8, %xmm8, %xmm8 + vxor v7, v7, v1 # vpxor %xmm1, %xmm7, %xmm7 + + # rotate + ?vspltw v0, v0, 3 # vpshufd \$0xFF, %xmm0, %xmm0 + ?vsldoi v0, v0, v0, 1 # vpalignr \$1, %xmm0, %xmm0, %xmm0 + + # fall through... + + # low round: same as high round, but no rotation and no rcon. +_vpaes_schedule_low_round: + # smear xmm7 + ?vsldoi v1, v9, v7, 12 # vpslldq \$4, %xmm7, %xmm1 + vxor v7, v7, v1 # vpxor %xmm1, %xmm7, %xmm7 + vspltisb v1, 0x0f # 0x0f..0f + ?vsldoi v4, v9, v7, 8 # vpslldq \$8, %xmm7, %xmm4 + + # subbytes + vand v1, v1, v0 # vpand %xmm9, %xmm0, %xmm1 # 0 = k + vsrb v0, v0, v8 # vpsrlb \$4, %xmm0, %xmm0 # 1 = i + vxor v7, v7, v4 # vpxor %xmm4, %xmm7, %xmm7 + vperm v2, $invhi, v9, v1 # vpshufb %xmm1, %xmm11, %xmm2 # 2 = a/k + vxor v1, v1, v0 # vpxor %xmm0, %xmm1, %xmm1 # 0 = j + vperm v3, $invlo, v9, v0 # vpshufb %xmm0, %xmm10, %xmm3 # 3 = 1/i + vxor v3, v3, v2 # vpxor %xmm2, %xmm3, %xmm3 # 3 = iak = 1/i + a/k + vperm v4, $invlo, v9, v1 # vpshufb %xmm1, %xmm10, %xmm4 # 4 = 1/j + vxor v7, v7, v26 # vpxor .Lk_s63(%rip), %xmm7, %xmm7 + vperm v3, $invlo, v9, v3 # vpshufb %xmm3, %xmm10, %xmm3 # 2 = 1/iak + vxor v4, v4, v2 # vpxor %xmm2, %xmm4, %xmm4 # 4 = jak = 1/j + a/k + vperm v2, $invlo, v9, v4 # vpshufb %xmm4, %xmm10, %xmm2 # 3 = 1/jak + vxor v3, v3, v1 # vpxor %xmm1, %xmm3, %xmm3 # 2 = io + vxor v2, v2, v0 # vpxor %xmm0, %xmm2, %xmm2 # 3 = jo + vperm v4, v15, v9, v3 # vpshufb %xmm3, %xmm13, %xmm4 # 4 = sbou + vperm v1, v14, v9, v2 # vpshufb %xmm2, %xmm12, %xmm1 # 0 = sb1t + vxor v1, v1, v4 # vpxor %xmm4, %xmm1, %xmm1 # 0 = sbox output + + # add in smeared stuff + vxor v0, v1, v7 # vpxor %xmm7, %xmm1, %xmm0 + vxor v7, v1, v7 # vmovdqa %xmm0, %xmm7 + blr + .long 0 + .byte 0,12,0x14,0,0,0,0,0 + +## +## .aes_schedule_transform +## +## Linear-transform %xmm0 according to tables at (%r11) +## +## Requires that %xmm9 = 0x0F0F... as in preheat +## Output in %xmm0 +## Clobbers %xmm2 +## +.align 4 +_vpaes_schedule_transform: + #vand v1, v0, v9 # vpand %xmm9, %xmm0, %xmm1 + vsrb v2, v0, v8 # vpsrlb \$4, %xmm0, %xmm0 + # vmovdqa (%r11), %xmm2 # lo + vperm v0, $iptlo, $iptlo, v0 # vpshufb %xmm1, %xmm2, %xmm2 + # vmovdqa 16(%r11), %xmm1 # hi + vperm v2, $ipthi, $ipthi, v2 # vpshufb %xmm0, %xmm1, %xmm0 + vxor v0, v0, v2 # vpxor %xmm2, %xmm0, %xmm0 + blr + .long 0 + .byte 0,12,0x14,0,0,0,0,0 + +## +## .aes_schedule_mangle +## +## Mangle xmm0 from (basis-transformed) standard version +## to our version. +## +## On encrypt, +## xor with 0x63 +## multiply by circulant 0,1,1,1 +## apply shiftrows transform +## +## On decrypt, +## xor with 0x63 +## multiply by "inverse mixcolumns" circulant E,B,D,9 +## deskew +## apply shiftrows transform +## +## +## Writes out to (%rdx), and increments or decrements it +## Keeps track of round number mod 4 in %r8 +## Preserves xmm0 +## Clobbers xmm1-xmm5 +## +.align 4 +_vpaes_schedule_mangle: + #vmr v4, v0 # vmovdqa %xmm0, %xmm4 # save xmm0 for later + # vmovdqa .Lk_mc_forward(%rip),%xmm5 + bne $dir, Lschedule_mangle_dec + + # encrypting + vxor v4, v0, v26 # vpxor .Lk_s63(%rip), %xmm0, %xmm4 + addi $out, $out, 16 # add \$16, %rdx + vperm v4, v4, v4, v25 # vpshufb %xmm5, %xmm4, %xmm4 + vperm v1, v4, v4, v25 # vpshufb %xmm5, %xmm4, %xmm1 + vperm v3, v1, v1, v25 # vpshufb %xmm5, %xmm1, %xmm3 + vxor v4, v4, v1 # vpxor %xmm1, %xmm4, %xmm4 + lvx v1, r8, r10 # vmovdqa (%r8,%r10), %xmm1 + vxor v3, v3, v4 # vpxor %xmm4, %xmm3, %xmm3 + + vperm v3, v3, v3, v1 # vpshufb %xmm1, %xmm3, %xmm3 + addi r8, r8, -16 # add \$-16, %r8 + andi. r8, r8, 0x30 # and \$0x30, %r8 + + #stvx v3, 0, $out # vmovdqu %xmm3, (%rdx) + vperm v1, v3, v3, $outperm # rotate right/left + vsel v2, $outhead, v1, $outmask + vmr $outhead, v1 + stvx v2, 0, $out + blr + +.align 4 +Lschedule_mangle_dec: + # inverse mix columns + # lea .Lk_dksd(%rip),%r11 + vsrb v1, v0, v8 # vpsrlb \$4, %xmm4, %xmm1 # 1 = hi + #and v4, v0, v9 # vpand %xmm9, %xmm4, %xmm4 # 4 = lo + + # vmovdqa 0x00(%r11), %xmm2 + vperm v2, v16, v16, v0 # vpshufb %xmm4, %xmm2, %xmm2 + # vmovdqa 0x10(%r11), %xmm3 + vperm v3, v17, v17, v1 # vpshufb %xmm1, %xmm3, %xmm3 + vxor v3, v3, v2 # vpxor %xmm2, %xmm3, %xmm3 + vperm v3, v3, v9, v25 # vpshufb %xmm5, %xmm3, %xmm3 + + # vmovdqa 0x20(%r11), %xmm2 + vperm v2, v18, v18, v0 # vpshufb %xmm4, %xmm2, %xmm2 + vxor v2, v2, v3 # vpxor %xmm3, %xmm2, %xmm2 + # vmovdqa 0x30(%r11), %xmm3 + vperm v3, v19, v19, v1 # vpshufb %xmm1, %xmm3, %xmm3 + vxor v3, v3, v2 # vpxor %xmm2, %xmm3, %xmm3 + vperm v3, v3, v9, v25 # vpshufb %xmm5, %xmm3, %xmm3 + + # vmovdqa 0x40(%r11), %xmm2 + vperm v2, v20, v20, v0 # vpshufb %xmm4, %xmm2, %xmm2 + vxor v2, v2, v3 # vpxor %xmm3, %xmm2, %xmm2 + # vmovdqa 0x50(%r11), %xmm3 + vperm v3, v21, v21, v1 # vpshufb %xmm1, %xmm3, %xmm3 + vxor v3, v3, v2 # vpxor %xmm2, %xmm3, %xmm3 + + # vmovdqa 0x60(%r11), %xmm2 + vperm v2, v22, v22, v0 # vpshufb %xmm4, %xmm2, %xmm2 + vperm v3, v3, v9, v25 # vpshufb %xmm5, %xmm3, %xmm3 + # vmovdqa 0x70(%r11), %xmm4 + vperm v4, v23, v23, v1 # vpshufb %xmm1, %xmm4, %xmm4 + lvx v1, r8, r10 # vmovdqa (%r8,%r10), %xmm1 + vxor v2, v2, v3 # vpxor %xmm3, %xmm2, %xmm2 + vxor v3, v4, v2 # vpxor %xmm2, %xmm4, %xmm3 + + addi $out, $out, -16 # add \$-16, %rdx + + vperm v3, v3, v3, v1 # vpshufb %xmm1, %xmm3, %xmm3 + addi r8, r8, -16 # add \$-16, %r8 + andi. r8, r8, 0x30 # and \$0x30, %r8 + + #stvx v3, 0, $out # vmovdqu %xmm3, (%rdx) + vperm v1, v3, v3, $outperm # rotate right/left + vsel v2, $outhead, v1, $outmask + vmr $outhead, v1 + stvx v2, 0, $out + blr + .long 0 + .byte 0,12,0x14,0,0,0,0,0 + +.globl .vpaes_set_encrypt_key +.align 5 +.vpaes_set_encrypt_key: + $STU $sp,-$FRAME($sp) + li r10,`15+6*$SIZE_T` + li r11,`31+6*$SIZE_T` + mflr r0 + mfspr r6, 256 # save vrsave + stvx v20,r10,$sp + addi r10,r10,32 + stvx v21,r11,$sp + addi r11,r11,32 + stvx v22,r10,$sp + addi r10,r10,32 + stvx v23,r11,$sp + addi r11,r11,32 + stvx v24,r10,$sp + addi r10,r10,32 + stvx v25,r11,$sp + addi r11,r11,32 + stvx v26,r10,$sp + addi r10,r10,32 + stvx v27,r11,$sp + addi r11,r11,32 + stvx v28,r10,$sp + addi r10,r10,32 + stvx v29,r11,$sp + addi r11,r11,32 + stvx v30,r10,$sp + stvx v31,r11,$sp + stw r6,`$FRAME-4`($sp) # save vrsave + li r7, -1 + $PUSH r0, `$FRAME+$LRSAVE`($sp) + mtspr 256, r7 # preserve all AltiVec registers + + srwi r9, $bits, 5 # shr \$5,%eax + addi r9, r9, 6 # add \$5,%eax + stw r9, 240($out) # mov %eax,240(%rdx) # AES_KEY->rounds = nbits/32+5; + + cmplw $dir, $bits, $bits # set encrypt direction + li r8, 0x30 # mov \$0x30,%r8d + bl _vpaes_schedule_core + + $POP r0, `$FRAME+$LRSAVE`($sp) + li r10,`15+6*$SIZE_T` + li r11,`31+6*$SIZE_T` + mtspr 256, r6 # restore vrsave + mtlr r0 + xor r3, r3, r3 + lvx v20,r10,$sp + addi r10,r10,32 + lvx v21,r11,$sp + addi r11,r11,32 + lvx v22,r10,$sp + addi r10,r10,32 + lvx v23,r11,$sp + addi r11,r11,32 + lvx v24,r10,$sp + addi r10,r10,32 + lvx v25,r11,$sp + addi r11,r11,32 + lvx v26,r10,$sp + addi r10,r10,32 + lvx v27,r11,$sp + addi r11,r11,32 + lvx v28,r10,$sp + addi r10,r10,32 + lvx v29,r11,$sp + addi r11,r11,32 + lvx v30,r10,$sp + lvx v31,r11,$sp + addi $sp,$sp,$FRAME + blr + .long 0 + .byte 0,12,0x04,1,0x80,0,3,0 + .long 0 +.size .vpaes_set_encrypt_key,.-.vpaes_set_encrypt_key + +.globl .vpaes_set_decrypt_key +.align 4 +.vpaes_set_decrypt_key: + $STU $sp,-$FRAME($sp) + li r10,`15+6*$SIZE_T` + li r11,`31+6*$SIZE_T` + mflr r0 + mfspr r6, 256 # save vrsave + stvx v20,r10,$sp + addi r10,r10,32 + stvx v21,r11,$sp + addi r11,r11,32 + stvx v22,r10,$sp + addi r10,r10,32 + stvx v23,r11,$sp + addi r11,r11,32 + stvx v24,r10,$sp + addi r10,r10,32 + stvx v25,r11,$sp + addi r11,r11,32 + stvx v26,r10,$sp + addi r10,r10,32 + stvx v27,r11,$sp + addi r11,r11,32 + stvx v28,r10,$sp + addi r10,r10,32 + stvx v29,r11,$sp + addi r11,r11,32 + stvx v30,r10,$sp + stvx v31,r11,$sp + stw r6,`$FRAME-4`($sp) # save vrsave + li r7, -1 + $PUSH r0, `$FRAME+$LRSAVE`($sp) + mtspr 256, r7 # preserve all AltiVec registers + + srwi r9, $bits, 5 # shr \$5,%eax + addi r9, r9, 6 # add \$5,%eax + stw r9, 240($out) # mov %eax,240(%rdx) # AES_KEY->rounds = nbits/32+5; + + slwi r9, r9, 4 # shl \$4,%eax + add $out, $out, r9 # lea (%rdx,%rax),%rdx + + cmplwi $dir, $bits, 0 # set decrypt direction + srwi r8, $bits, 1 # shr \$1,%r8d + andi. r8, r8, 32 # and \$32,%r8d + xori r8, r8, 32 # xor \$32,%r8d # nbits==192?0:32 + bl _vpaes_schedule_core + + $POP r0, `$FRAME+$LRSAVE`($sp) + li r10,`15+6*$SIZE_T` + li r11,`31+6*$SIZE_T` + mtspr 256, r6 # restore vrsave + mtlr r0 + xor r3, r3, r3 + lvx v20,r10,$sp + addi r10,r10,32 + lvx v21,r11,$sp + addi r11,r11,32 + lvx v22,r10,$sp + addi r10,r10,32 + lvx v23,r11,$sp + addi r11,r11,32 + lvx v24,r10,$sp + addi r10,r10,32 + lvx v25,r11,$sp + addi r11,r11,32 + lvx v26,r10,$sp + addi r10,r10,32 + lvx v27,r11,$sp + addi r11,r11,32 + lvx v28,r10,$sp + addi r10,r10,32 + lvx v29,r11,$sp + addi r11,r11,32 + lvx v30,r10,$sp + lvx v31,r11,$sp + addi $sp,$sp,$FRAME + blr + .long 0 + .byte 0,12,0x04,1,0x80,0,3,0 + .long 0 +.size .vpaes_set_decrypt_key,.-.vpaes_set_decrypt_key +___ +} + +my $consts=1; +foreach (split("\n",$code)) { + s/\`([^\`]*)\`/eval $1/geo; + + # constants table endian-specific conversion + if ($consts && m/\.long\s+(.+)\s+(\?[a-z]*)$/o) { + my $conv=$2; + my @bytes=(); + + # convert to endian-agnostic format + foreach (split(/,\s+/,$1)) { + my $l = /^0/?oct:int; + push @bytes,($l>>24)&0xff,($l>>16)&0xff,($l>>8)&0xff,$l&0xff; + } + + # little-endian conversion + if ($flavour =~ /le$/o) { + SWITCH: for($conv) { + /\?inv/ && do { @bytes=map($_^0xf,@bytes); last; }; + /\?rev/ && do { @bytes=reverse(@bytes); last; }; + } + } + + #emit + print ".byte\t",join(',',map (sprintf("0x%02x",$_),@bytes)),"\n"; + next; + } + $consts=0 if (m/Lconsts:/o); # end of table + + # instructions prefixed with '?' are endian-specific and need + # to be adjusted accordingly... + if ($flavour =~ /le$/o) { # little-endian + s/\?lvsr/lvsl/o or + s/\?lvsl/lvsr/o or + s/\?(vperm\s+v[0-9]+,\s*)(v[0-9]+,\s*)(v[0-9]+,\s*)(v[0-9]+)/$1$3$2$4/o or + s/\?(vsldoi\s+v[0-9]+,\s*)(v[0-9]+,)\s*(v[0-9]+,\s*)([0-9]+)/$1$3$2 16-$4/o or + s/\?(vspltw\s+v[0-9]+,\s*)(v[0-9]+,)\s*([0-9])/$1$2 3-$3/o; + } else { # big-endian + s/\?([a-z]+)/$1/o; + } + + print $_,"\n"; +} + +close STDOUT; diff --git a/openssl-1.1.0h/crypto/aes/asm/vpaes-x86.pl b/openssl-1.1.0h/crypto/aes/asm/vpaes-x86.pl new file mode 100644 index 0000000..47615c0 --- /dev/null +++ b/openssl-1.1.0h/crypto/aes/asm/vpaes-x86.pl @@ -0,0 +1,916 @@ +#! /usr/bin/env perl +# Copyright 2011-2016 The OpenSSL Project Authors. All Rights Reserved. +# +# Licensed under the OpenSSL license (the "License"). You may not use +# this file except in compliance with the License. You can obtain a copy +# in the file LICENSE in the source distribution or at +# https://www.openssl.org/source/license.html + + +###################################################################### +## Constant-time SSSE3 AES core implementation. +## version 0.1 +## +## By Mike Hamburg (Stanford University), 2009 +## Public domain. +## +## For details see http://shiftleft.org/papers/vector_aes/ and +## http://crypto.stanford.edu/vpaes/. + +###################################################################### +# September 2011. +# +# Port vpaes-x86_64.pl as 32-bit "almost" drop-in replacement for +# aes-586.pl. "Almost" refers to the fact that AES_cbc_encrypt +# doesn't handle partial vectors (doesn't have to if called from +# EVP only). "Drop-in" implies that this module doesn't share key +# schedule structure with the original nor does it make assumption +# about its alignment... +# +# Performance summary. aes-586.pl column lists large-block CBC +# encrypt/decrypt/with-hyper-threading-off(*) results in cycles per +# byte processed with 128-bit key, and vpaes-x86.pl column - [also +# large-block CBC] encrypt/decrypt. +# +# aes-586.pl vpaes-x86.pl +# +# Core 2(**) 28.1/41.4/18.3 21.9/25.2(***) +# Nehalem 27.9/40.4/18.1 10.2/11.9 +# Atom 70.7/92.1/60.1 61.1/75.4(***) +# Silvermont 45.4/62.9/24.1 49.2/61.1(***) +# +# (*) "Hyper-threading" in the context refers rather to cache shared +# among multiple cores, than to specifically Intel HTT. As vast +# majority of contemporary cores share cache, slower code path +# is common place. In other words "with-hyper-threading-off" +# results are presented mostly for reference purposes. +# +# (**) "Core 2" refers to initial 65nm design, a.k.a. Conroe. +# +# (***) Less impressive improvement on Core 2 and Atom is due to slow +# pshufb, yet it's respectable +28%/64% improvement on Core 2 +# and +15% on Atom (as implied, over "hyper-threading-safe" +# code path). +# +# + +$0 =~ m/(.*[\/\\])[^\/\\]+$/; $dir=$1; +push(@INC,"${dir}","${dir}../../perlasm"); +require "x86asm.pl"; + +$output = pop; +open OUT,">$output"; +*STDOUT=*OUT; + +&asm_init($ARGV[0],"vpaes-x86.pl",$x86only = $ARGV[$#ARGV] eq "386"); + +$PREFIX="vpaes"; + +my ($round, $base, $magic, $key, $const, $inp, $out)= + ("eax", "ebx", "ecx", "edx","ebp", "esi","edi"); + +&static_label("_vpaes_consts"); +&static_label("_vpaes_schedule_low_round"); + +&set_label("_vpaes_consts",64); +$k_inv=-0x30; # inv, inva + &data_word(0x0D080180,0x0E05060F,0x0A0B0C02,0x04070309); + &data_word(0x0F0B0780,0x01040A06,0x02050809,0x030D0E0C); + +$k_s0F=-0x10; # s0F + &data_word(0x0F0F0F0F,0x0F0F0F0F,0x0F0F0F0F,0x0F0F0F0F); + +$k_ipt=0x00; # input transform (lo, hi) + &data_word(0x5A2A7000,0xC2B2E898,0x52227808,0xCABAE090); + &data_word(0x317C4D00,0x4C01307D,0xB0FDCC81,0xCD80B1FC); + +$k_sb1=0x20; # sb1u, sb1t + &data_word(0xCB503E00,0xB19BE18F,0x142AF544,0xA5DF7A6E); + &data_word(0xFAE22300,0x3618D415,0x0D2ED9EF,0x3BF7CCC1); +$k_sb2=0x40; # sb2u, sb2t + &data_word(0x0B712400,0xE27A93C6,0xBC982FCD,0x5EB7E955); + &data_word(0x0AE12900,0x69EB8840,0xAB82234A,0xC2A163C8); +$k_sbo=0x60; # sbou, sbot + &data_word(0x6FBDC700,0xD0D26D17,0xC502A878,0x15AABF7A); + &data_word(0x5FBB6A00,0xCFE474A5,0x412B35FA,0x8E1E90D1); + +$k_mc_forward=0x80; # mc_forward + &data_word(0x00030201,0x04070605,0x080B0A09,0x0C0F0E0D); + &data_word(0x04070605,0x080B0A09,0x0C0F0E0D,0x00030201); + &data_word(0x080B0A09,0x0C0F0E0D,0x00030201,0x04070605); + &data_word(0x0C0F0E0D,0x00030201,0x04070605,0x080B0A09); + +$k_mc_backward=0xc0; # mc_backward + &data_word(0x02010003,0x06050407,0x0A09080B,0x0E0D0C0F); + &data_word(0x0E0D0C0F,0x02010003,0x06050407,0x0A09080B); + &data_word(0x0A09080B,0x0E0D0C0F,0x02010003,0x06050407); + &data_word(0x06050407,0x0A09080B,0x0E0D0C0F,0x02010003); + +$k_sr=0x100; # sr + &data_word(0x03020100,0x07060504,0x0B0A0908,0x0F0E0D0C); + &data_word(0x0F0A0500,0x030E0904,0x07020D08,0x0B06010C); + &data_word(0x0B020900,0x0F060D04,0x030A0108,0x070E050C); + &data_word(0x070A0D00,0x0B0E0104,0x0F020508,0x0306090C); + +$k_rcon=0x140; # rcon + &data_word(0xAF9DEEB6,0x1F8391B9,0x4D7C7D81,0x702A9808); + +$k_s63=0x150; # s63: all equal to 0x63 transformed + &data_word(0x5B5B5B5B,0x5B5B5B5B,0x5B5B5B5B,0x5B5B5B5B); + +$k_opt=0x160; # output transform + &data_word(0xD6B66000,0xFF9F4929,0xDEBE6808,0xF7974121); + &data_word(0x50BCEC00,0x01EDBD51,0xB05C0CE0,0xE10D5DB1); + +$k_deskew=0x180; # deskew tables: inverts the sbox's "skew" + &data_word(0x47A4E300,0x07E4A340,0x5DBEF91A,0x1DFEB95A); + &data_word(0x83EA6900,0x5F36B5DC,0xF49D1E77,0x2841C2AB); +## +## Decryption stuff +## Key schedule constants +## +$k_dksd=0x1a0; # decryption key schedule: invskew x*D + &data_word(0xA3E44700,0xFEB91A5D,0x5A1DBEF9,0x0740E3A4); + &data_word(0xB5368300,0x41C277F4,0xAB289D1E,0x5FDC69EA); +$k_dksb=0x1c0; # decryption key schedule: invskew x*B + &data_word(0x8550D500,0x9A4FCA1F,0x1CC94C99,0x03D65386); + &data_word(0xB6FC4A00,0x115BEDA7,0x7E3482C8,0xD993256F); +$k_dkse=0x1e0; # decryption key schedule: invskew x*E + 0x63 + &data_word(0x1FC9D600,0xD5031CCA,0x994F5086,0x53859A4C); + &data_word(0x4FDC7BE8,0xA2319605,0x20B31487,0xCD5EF96A); +$k_dks9=0x200; # decryption key schedule: invskew x*9 + &data_word(0x7ED9A700,0xB6116FC8,0x82255BFC,0x4AED9334); + &data_word(0x27143300,0x45765162,0xE9DAFDCE,0x8BB89FAC); + +## +## Decryption stuff +## Round function constants +## +$k_dipt=0x220; # decryption input transform + &data_word(0x0B545F00,0x0F505B04,0x114E451A,0x154A411E); + &data_word(0x60056500,0x86E383E6,0xF491F194,0x12771772); + +$k_dsb9=0x240; # decryption sbox output *9*u, *9*t + &data_word(0x9A86D600,0x851C0353,0x4F994CC9,0xCAD51F50); + &data_word(0xECD74900,0xC03B1789,0xB2FBA565,0x725E2C9E); +$k_dsbd=0x260; # decryption sbox output *D*u, *D*t + &data_word(0xE6B1A200,0x7D57CCDF,0x882A4439,0xF56E9B13); + &data_word(0x24C6CB00,0x3CE2FAF7,0x15DEEFD3,0x2931180D); +$k_dsbb=0x280; # decryption sbox output *B*u, *B*t + &data_word(0x96B44200,0xD0226492,0xB0F2D404,0x602646F6); + &data_word(0xCD596700,0xC19498A6,0x3255AA6B,0xF3FF0C3E); +$k_dsbe=0x2a0; # decryption sbox output *E*u, *E*t + &data_word(0x26D4D000,0x46F29296,0x64B4F6B0,0x22426004); + &data_word(0xFFAAC100,0x0C55A6CD,0x98593E32,0x9467F36B); +$k_dsbo=0x2c0; # decryption sbox final output + &data_word(0x7EF94000,0x1387EA53,0xD4943E2D,0xC7AA6DB9); + &data_word(0x93441D00,0x12D7560F,0xD8C58E9C,0xCA4B8159); +&asciz ("Vector Permutation AES for x86/SSSE3, Mike Hamburg (Stanford University)"); +&align (64); + +&function_begin_B("_vpaes_preheat"); + &add ($const,&DWP(0,"esp")); + &movdqa ("xmm7",&QWP($k_inv,$const)); + &movdqa ("xmm6",&QWP($k_s0F,$const)); + &ret (); +&function_end_B("_vpaes_preheat"); + +## +## _aes_encrypt_core +## +## AES-encrypt %xmm0. +## +## Inputs: +## %xmm0 = input +## %xmm6-%xmm7 as in _vpaes_preheat +## (%edx) = scheduled keys +## +## Output in %xmm0 +## Clobbers %xmm1-%xmm5, %eax, %ebx, %ecx, %edx +## +## +&function_begin_B("_vpaes_encrypt_core"); + &mov ($magic,16); + &mov ($round,&DWP(240,$key)); + &movdqa ("xmm1","xmm6") + &movdqa ("xmm2",&QWP($k_ipt,$const)); + &pandn ("xmm1","xmm0"); + &pand ("xmm0","xmm6"); + &movdqu ("xmm5",&QWP(0,$key)); + &pshufb ("xmm2","xmm0"); + &movdqa ("xmm0",&QWP($k_ipt+16,$const)); + &pxor ("xmm2","xmm5"); + &psrld ("xmm1",4); + &add ($key,16); + &pshufb ("xmm0","xmm1"); + &lea ($base,&DWP($k_mc_backward,$const)); + &pxor ("xmm0","xmm2"); + &jmp (&label("enc_entry")); + + +&set_label("enc_loop",16); + # middle of middle round + &movdqa ("xmm4",&QWP($k_sb1,$const)); # 4 : sb1u + &movdqa ("xmm0",&QWP($k_sb1+16,$const));# 0 : sb1t + &pshufb ("xmm4","xmm2"); # 4 = sb1u + &pshufb ("xmm0","xmm3"); # 0 = sb1t + &pxor ("xmm4","xmm5"); # 4 = sb1u + k + &movdqa ("xmm5",&QWP($k_sb2,$const)); # 4 : sb2u + &pxor ("xmm0","xmm4"); # 0 = A + &movdqa ("xmm1",&QWP(-0x40,$base,$magic));# .Lk_mc_forward[] + &pshufb ("xmm5","xmm2"); # 4 = sb2u + &movdqa ("xmm2",&QWP($k_sb2+16,$const));# 2 : sb2t + &movdqa ("xmm4",&QWP(0,$base,$magic)); # .Lk_mc_backward[] + &pshufb ("xmm2","xmm3"); # 2 = sb2t + &movdqa ("xmm3","xmm0"); # 3 = A + &pxor ("xmm2","xmm5"); # 2 = 2A + &pshufb ("xmm0","xmm1"); # 0 = B + &add ($key,16); # next key + &pxor ("xmm0","xmm2"); # 0 = 2A+B + &pshufb ("xmm3","xmm4"); # 3 = D + &add ($magic,16); # next mc + &pxor ("xmm3","xmm0"); # 3 = 2A+B+D + &pshufb ("xmm0","xmm1"); # 0 = 2B+C + &and ($magic,0x30); # ... mod 4 + &sub ($round,1); # nr-- + &pxor ("xmm0","xmm3"); # 0 = 2A+3B+C+D + +&set_label("enc_entry"); + # top of round + &movdqa ("xmm1","xmm6"); # 1 : i + &movdqa ("xmm5",&QWP($k_inv+16,$const));# 2 : a/k + &pandn ("xmm1","xmm0"); # 1 = i<<4 + &psrld ("xmm1",4); # 1 = i + &pand ("xmm0","xmm6"); # 0 = k + &pshufb ("xmm5","xmm0"); # 2 = a/k + &movdqa ("xmm3","xmm7"); # 3 : 1/i + &pxor ("xmm0","xmm1"); # 0 = j + &pshufb ("xmm3","xmm1"); # 3 = 1/i + &movdqa ("xmm4","xmm7"); # 4 : 1/j + &pxor ("xmm3","xmm5"); # 3 = iak = 1/i + a/k + &pshufb ("xmm4","xmm0"); # 4 = 1/j + &movdqa ("xmm2","xmm7"); # 2 : 1/iak + &pxor ("xmm4","xmm5"); # 4 = jak = 1/j + a/k + &pshufb ("xmm2","xmm3"); # 2 = 1/iak + &movdqa ("xmm3","xmm7"); # 3 : 1/jak + &pxor ("xmm2","xmm0"); # 2 = io + &pshufb ("xmm3","xmm4"); # 3 = 1/jak + &movdqu ("xmm5",&QWP(0,$key)); + &pxor ("xmm3","xmm1"); # 3 = jo + &jnz (&label("enc_loop")); + + # middle of last round + &movdqa ("xmm4",&QWP($k_sbo,$const)); # 3 : sbou .Lk_sbo + &movdqa ("xmm0",&QWP($k_sbo+16,$const));# 3 : sbot .Lk_sbo+16 + &pshufb ("xmm4","xmm2"); # 4 = sbou + &pxor ("xmm4","xmm5"); # 4 = sb1u + k + &pshufb ("xmm0","xmm3"); # 0 = sb1t + &movdqa ("xmm1",&QWP(0x40,$base,$magic));# .Lk_sr[] + &pxor ("xmm0","xmm4"); # 0 = A + &pshufb ("xmm0","xmm1"); + &ret (); +&function_end_B("_vpaes_encrypt_core"); + +## +## Decryption core +## +## Same API as encryption core. +## +&function_begin_B("_vpaes_decrypt_core"); + &lea ($base,&DWP($k_dsbd,$const)); + &mov ($round,&DWP(240,$key)); + &movdqa ("xmm1","xmm6"); + &movdqa ("xmm2",&QWP($k_dipt-$k_dsbd,$base)); + &pandn ("xmm1","xmm0"); + &mov ($magic,$round); + &psrld ("xmm1",4) + &movdqu ("xmm5",&QWP(0,$key)); + &shl ($magic,4); + &pand ("xmm0","xmm6"); + &pshufb ("xmm2","xmm0"); + &movdqa ("xmm0",&QWP($k_dipt-$k_dsbd+16,$base)); + &xor ($magic,0x30); + &pshufb ("xmm0","xmm1"); + &and ($magic,0x30); + &pxor ("xmm2","xmm5"); + &movdqa ("xmm5",&QWP($k_mc_forward+48,$const)); + &pxor ("xmm0","xmm2"); + &add ($key,16); + &lea ($magic,&DWP($k_sr-$k_dsbd,$base,$magic)); + &jmp (&label("dec_entry")); + +&set_label("dec_loop",16); +## +## Inverse mix columns +## + &movdqa ("xmm4",&QWP(-0x20,$base)); # 4 : sb9u + &movdqa ("xmm1",&QWP(-0x10,$base)); # 0 : sb9t + &pshufb ("xmm4","xmm2"); # 4 = sb9u + &pshufb ("xmm1","xmm3"); # 0 = sb9t + &pxor ("xmm0","xmm4"); + &movdqa ("xmm4",&QWP(0,$base)); # 4 : sbdu + &pxor ("xmm0","xmm1"); # 0 = ch + &movdqa ("xmm1",&QWP(0x10,$base)); # 0 : sbdt + + &pshufb ("xmm4","xmm2"); # 4 = sbdu + &pshufb ("xmm0","xmm5"); # MC ch + &pshufb ("xmm1","xmm3"); # 0 = sbdt + &pxor ("xmm0","xmm4"); # 4 = ch + &movdqa ("xmm4",&QWP(0x20,$base)); # 4 : sbbu + &pxor ("xmm0","xmm1"); # 0 = ch + &movdqa ("xmm1",&QWP(0x30,$base)); # 0 : sbbt + + &pshufb ("xmm4","xmm2"); # 4 = sbbu + &pshufb ("xmm0","xmm5"); # MC ch + &pshufb ("xmm1","xmm3"); # 0 = sbbt + &pxor ("xmm0","xmm4"); # 4 = ch + &movdqa ("xmm4",&QWP(0x40,$base)); # 4 : sbeu + &pxor ("xmm0","xmm1"); # 0 = ch + &movdqa ("xmm1",&QWP(0x50,$base)); # 0 : sbet + + &pshufb ("xmm4","xmm2"); # 4 = sbeu + &pshufb ("xmm0","xmm5"); # MC ch + &pshufb ("xmm1","xmm3"); # 0 = sbet + &pxor ("xmm0","xmm4"); # 4 = ch + &add ($key,16); # next round key + &palignr("xmm5","xmm5",12); + &pxor ("xmm0","xmm1"); # 0 = ch + &sub ($round,1); # nr-- + +&set_label("dec_entry"); + # top of round + &movdqa ("xmm1","xmm6"); # 1 : i + &movdqa ("xmm2",&QWP($k_inv+16,$const));# 2 : a/k + &pandn ("xmm1","xmm0"); # 1 = i<<4 + &pand ("xmm0","xmm6"); # 0 = k + &psrld ("xmm1",4); # 1 = i + &pshufb ("xmm2","xmm0"); # 2 = a/k + &movdqa ("xmm3","xmm7"); # 3 : 1/i + &pxor ("xmm0","xmm1"); # 0 = j + &pshufb ("xmm3","xmm1"); # 3 = 1/i + &movdqa ("xmm4","xmm7"); # 4 : 1/j + &pxor ("xmm3","xmm2"); # 3 = iak = 1/i + a/k + &pshufb ("xmm4","xmm0"); # 4 = 1/j + &pxor ("xmm4","xmm2"); # 4 = jak = 1/j + a/k + &movdqa ("xmm2","xmm7"); # 2 : 1/iak + &pshufb ("xmm2","xmm3"); # 2 = 1/iak + &movdqa ("xmm3","xmm7"); # 3 : 1/jak + &pxor ("xmm2","xmm0"); # 2 = io + &pshufb ("xmm3","xmm4"); # 3 = 1/jak + &movdqu ("xmm0",&QWP(0,$key)); + &pxor ("xmm3","xmm1"); # 3 = jo + &jnz (&label("dec_loop")); + + # middle of last round + &movdqa ("xmm4",&QWP(0x60,$base)); # 3 : sbou + &pshufb ("xmm4","xmm2"); # 4 = sbou + &pxor ("xmm4","xmm0"); # 4 = sb1u + k + &movdqa ("xmm0",&QWP(0x70,$base)); # 0 : sbot + &movdqa ("xmm2",&QWP(0,$magic)); + &pshufb ("xmm0","xmm3"); # 0 = sb1t + &pxor ("xmm0","xmm4"); # 0 = A + &pshufb ("xmm0","xmm2"); + &ret (); +&function_end_B("_vpaes_decrypt_core"); + +######################################################## +## ## +## AES key schedule ## +## ## +######################################################## +&function_begin_B("_vpaes_schedule_core"); + &add ($const,&DWP(0,"esp")); + &movdqu ("xmm0",&QWP(0,$inp)); # load key (unaligned) + &movdqa ("xmm2",&QWP($k_rcon,$const)); # load rcon + + # input transform + &movdqa ("xmm3","xmm0"); + &lea ($base,&DWP($k_ipt,$const)); + &movdqa (&QWP(4,"esp"),"xmm2"); # xmm8 + &call ("_vpaes_schedule_transform"); + &movdqa ("xmm7","xmm0"); + + &test ($out,$out); + &jnz (&label("schedule_am_decrypting")); + + # encrypting, output zeroth round key after transform + &movdqu (&QWP(0,$key),"xmm0"); + &jmp (&label("schedule_go")); + +&set_label("schedule_am_decrypting"); + # decrypting, output zeroth round key after shiftrows + &movdqa ("xmm1",&QWP($k_sr,$const,$magic)); + &pshufb ("xmm3","xmm1"); + &movdqu (&QWP(0,$key),"xmm3"); + &xor ($magic,0x30); + +&set_label("schedule_go"); + &cmp ($round,192); + &ja (&label("schedule_256")); + &je (&label("schedule_192")); + # 128: fall though + +## +## .schedule_128 +## +## 128-bit specific part of key schedule. +## +## This schedule is really simple, because all its parts +## are accomplished by the subroutines. +## +&set_label("schedule_128"); + &mov ($round,10); + +&set_label("loop_schedule_128"); + &call ("_vpaes_schedule_round"); + &dec ($round); + &jz (&label("schedule_mangle_last")); + &call ("_vpaes_schedule_mangle"); # write output + &jmp (&label("loop_schedule_128")); + +## +## .aes_schedule_192 +## +## 192-bit specific part of key schedule. +## +## The main body of this schedule is the same as the 128-bit +## schedule, but with more smearing. The long, high side is +## stored in %xmm7 as before, and the short, low side is in +## the high bits of %xmm6. +## +## This schedule is somewhat nastier, however, because each +## round produces 192 bits of key material, or 1.5 round keys. +## Therefore, on each cycle we do 2 rounds and produce 3 round +## keys. +## +&set_label("schedule_192",16); + &movdqu ("xmm0",&QWP(8,$inp)); # load key part 2 (very unaligned) + &call ("_vpaes_schedule_transform"); # input transform + &movdqa ("xmm6","xmm0"); # save short part + &pxor ("xmm4","xmm4"); # clear 4 + &movhlps("xmm6","xmm4"); # clobber low side with zeros + &mov ($round,4); + +&set_label("loop_schedule_192"); + &call ("_vpaes_schedule_round"); + &palignr("xmm0","xmm6",8); + &call ("_vpaes_schedule_mangle"); # save key n + &call ("_vpaes_schedule_192_smear"); + &call ("_vpaes_schedule_mangle"); # save key n+1 + &call ("_vpaes_schedule_round"); + &dec ($round); + &jz (&label("schedule_mangle_last")); + &call ("_vpaes_schedule_mangle"); # save key n+2 + &call ("_vpaes_schedule_192_smear"); + &jmp (&label("loop_schedule_192")); + +## +## .aes_schedule_256 +## +## 256-bit specific part of key schedule. +## +## The structure here is very similar to the 128-bit +## schedule, but with an additional "low side" in +## %xmm6. The low side's rounds are the same as the +## high side's, except no rcon and no rotation. +## +&set_label("schedule_256",16); + &movdqu ("xmm0",&QWP(16,$inp)); # load key part 2 (unaligned) + &call ("_vpaes_schedule_transform"); # input transform + &mov ($round,7); + +&set_label("loop_schedule_256"); + &call ("_vpaes_schedule_mangle"); # output low result + &movdqa ("xmm6","xmm0"); # save cur_lo in xmm6 + + # high round + &call ("_vpaes_schedule_round"); + &dec ($round); + &jz (&label("schedule_mangle_last")); + &call ("_vpaes_schedule_mangle"); + + # low round. swap xmm7 and xmm6 + &pshufd ("xmm0","xmm0",0xFF); + &movdqa (&QWP(20,"esp"),"xmm7"); + &movdqa ("xmm7","xmm6"); + &call ("_vpaes_schedule_low_round"); + &movdqa ("xmm7",&QWP(20,"esp")); + + &jmp (&label("loop_schedule_256")); + +## +## .aes_schedule_mangle_last +## +## Mangler for last round of key schedule +## Mangles %xmm0 +## when encrypting, outputs out(%xmm0) ^ 63 +## when decrypting, outputs unskew(%xmm0) +## +## Always called right before return... jumps to cleanup and exits +## +&set_label("schedule_mangle_last",16); + # schedule last round key from xmm0 + &lea ($base,&DWP($k_deskew,$const)); + &test ($out,$out); + &jnz (&label("schedule_mangle_last_dec")); + + # encrypting + &movdqa ("xmm1",&QWP($k_sr,$const,$magic)); + &pshufb ("xmm0","xmm1"); # output permute + &lea ($base,&DWP($k_opt,$const)); # prepare to output transform + &add ($key,32); + +&set_label("schedule_mangle_last_dec"); + &add ($key,-16); + &pxor ("xmm0",&QWP($k_s63,$const)); + &call ("_vpaes_schedule_transform"); # output transform + &movdqu (&QWP(0,$key),"xmm0"); # save last key + + # cleanup + &pxor ("xmm0","xmm0"); + &pxor ("xmm1","xmm1"); + &pxor ("xmm2","xmm2"); + &pxor ("xmm3","xmm3"); + &pxor ("xmm4","xmm4"); + &pxor ("xmm5","xmm5"); + &pxor ("xmm6","xmm6"); + &pxor ("xmm7","xmm7"); + &ret (); +&function_end_B("_vpaes_schedule_core"); + +## +## .aes_schedule_192_smear +## +## Smear the short, low side in the 192-bit key schedule. +## +## Inputs: +## %xmm7: high side, b a x y +## %xmm6: low side, d c 0 0 +## %xmm13: 0 +## +## Outputs: +## %xmm6: b+c+d b+c 0 0 +## %xmm0: b+c+d b+c b a +## +&function_begin_B("_vpaes_schedule_192_smear"); + &pshufd ("xmm1","xmm6",0x80); # d c 0 0 -> c 0 0 0 + &pshufd ("xmm0","xmm7",0xFE); # b a _ _ -> b b b a + &pxor ("xmm6","xmm1"); # -> c+d c 0 0 + &pxor ("xmm1","xmm1"); + &pxor ("xmm6","xmm0"); # -> b+c+d b+c b a + &movdqa ("xmm0","xmm6"); + &movhlps("xmm6","xmm1"); # clobber low side with zeros + &ret (); +&function_end_B("_vpaes_schedule_192_smear"); + +## +## .aes_schedule_round +## +## Runs one main round of the key schedule on %xmm0, %xmm7 +## +## Specifically, runs subbytes on the high dword of %xmm0 +## then rotates it by one byte and xors into the low dword of +## %xmm7. +## +## Adds rcon from low byte of %xmm8, then rotates %xmm8 for +## next rcon. +## +## Smears the dwords of %xmm7 by xoring the low into the +## second low, result into third, result into highest. +## +## Returns results in %xmm7 = %xmm0. +## Clobbers %xmm1-%xmm5. +## +&function_begin_B("_vpaes_schedule_round"); + # extract rcon from xmm8 + &movdqa ("xmm2",&QWP(8,"esp")); # xmm8 + &pxor ("xmm1","xmm1"); + &palignr("xmm1","xmm2",15); + &palignr("xmm2","xmm2",15); + &pxor ("xmm7","xmm1"); + + # rotate + &pshufd ("xmm0","xmm0",0xFF); + &palignr("xmm0","xmm0",1); + + # fall through... + &movdqa (&QWP(8,"esp"),"xmm2"); # xmm8 + + # low round: same as high round, but no rotation and no rcon. +&set_label("_vpaes_schedule_low_round"); + # smear xmm7 + &movdqa ("xmm1","xmm7"); + &pslldq ("xmm7",4); + &pxor ("xmm7","xmm1"); + &movdqa ("xmm1","xmm7"); + &pslldq ("xmm7",8); + &pxor ("xmm7","xmm1"); + &pxor ("xmm7",&QWP($k_s63,$const)); + + # subbyte + &movdqa ("xmm4",&QWP($k_s0F,$const)); + &movdqa ("xmm5",&QWP($k_inv,$const)); # 4 : 1/j + &movdqa ("xmm1","xmm4"); + &pandn ("xmm1","xmm0"); + &psrld ("xmm1",4); # 1 = i + &pand ("xmm0","xmm4"); # 0 = k + &movdqa ("xmm2",&QWP($k_inv+16,$const));# 2 : a/k + &pshufb ("xmm2","xmm0"); # 2 = a/k + &pxor ("xmm0","xmm1"); # 0 = j + &movdqa ("xmm3","xmm5"); # 3 : 1/i + &pshufb ("xmm3","xmm1"); # 3 = 1/i + &pxor ("xmm3","xmm2"); # 3 = iak = 1/i + a/k + &movdqa ("xmm4","xmm5"); # 4 : 1/j + &pshufb ("xmm4","xmm0"); # 4 = 1/j + &pxor ("xmm4","xmm2"); # 4 = jak = 1/j + a/k + &movdqa ("xmm2","xmm5"); # 2 : 1/iak + &pshufb ("xmm2","xmm3"); # 2 = 1/iak + &pxor ("xmm2","xmm0"); # 2 = io + &movdqa ("xmm3","xmm5"); # 3 : 1/jak + &pshufb ("xmm3","xmm4"); # 3 = 1/jak + &pxor ("xmm3","xmm1"); # 3 = jo + &movdqa ("xmm4",&QWP($k_sb1,$const)); # 4 : sbou + &pshufb ("xmm4","xmm2"); # 4 = sbou + &movdqa ("xmm0",&QWP($k_sb1+16,$const));# 0 : sbot + &pshufb ("xmm0","xmm3"); # 0 = sb1t + &pxor ("xmm0","xmm4"); # 0 = sbox output + + # add in smeared stuff + &pxor ("xmm0","xmm7"); + &movdqa ("xmm7","xmm0"); + &ret (); +&function_end_B("_vpaes_schedule_round"); + +## +## .aes_schedule_transform +## +## Linear-transform %xmm0 according to tables at (%ebx) +## +## Output in %xmm0 +## Clobbers %xmm1, %xmm2 +## +&function_begin_B("_vpaes_schedule_transform"); + &movdqa ("xmm2",&QWP($k_s0F,$const)); + &movdqa ("xmm1","xmm2"); + &pandn ("xmm1","xmm0"); + &psrld ("xmm1",4); + &pand ("xmm0","xmm2"); + &movdqa ("xmm2",&QWP(0,$base)); + &pshufb ("xmm2","xmm0"); + &movdqa ("xmm0",&QWP(16,$base)); + &pshufb ("xmm0","xmm1"); + &pxor ("xmm0","xmm2"); + &ret (); +&function_end_B("_vpaes_schedule_transform"); + +## +## .aes_schedule_mangle +## +## Mangle xmm0 from (basis-transformed) standard version +## to our version. +## +## On encrypt, +## xor with 0x63 +## multiply by circulant 0,1,1,1 +## apply shiftrows transform +## +## On decrypt, +## xor with 0x63 +## multiply by "inverse mixcolumns" circulant E,B,D,9 +## deskew +## apply shiftrows transform +## +## +## Writes out to (%edx), and increments or decrements it +## Keeps track of round number mod 4 in %ecx +## Preserves xmm0 +## Clobbers xmm1-xmm5 +## +&function_begin_B("_vpaes_schedule_mangle"); + &movdqa ("xmm4","xmm0"); # save xmm0 for later + &movdqa ("xmm5",&QWP($k_mc_forward,$const)); + &test ($out,$out); + &jnz (&label("schedule_mangle_dec")); + + # encrypting + &add ($key,16); + &pxor ("xmm4",&QWP($k_s63,$const)); + &pshufb ("xmm4","xmm5"); + &movdqa ("xmm3","xmm4"); + &pshufb ("xmm4","xmm5"); + &pxor ("xmm3","xmm4"); + &pshufb ("xmm4","xmm5"); + &pxor ("xmm3","xmm4"); + + &jmp (&label("schedule_mangle_both")); + +&set_label("schedule_mangle_dec",16); + # inverse mix columns + &movdqa ("xmm2",&QWP($k_s0F,$const)); + &lea ($inp,&DWP($k_dksd,$const)); + &movdqa ("xmm1","xmm2"); + &pandn ("xmm1","xmm4"); + &psrld ("xmm1",4); # 1 = hi + &pand ("xmm4","xmm2"); # 4 = lo + + &movdqa ("xmm2",&QWP(0,$inp)); + &pshufb ("xmm2","xmm4"); + &movdqa ("xmm3",&QWP(0x10,$inp)); + &pshufb ("xmm3","xmm1"); + &pxor ("xmm3","xmm2"); + &pshufb ("xmm3","xmm5"); + + &movdqa ("xmm2",&QWP(0x20,$inp)); + &pshufb ("xmm2","xmm4"); + &pxor ("xmm2","xmm3"); + &movdqa ("xmm3",&QWP(0x30,$inp)); + &pshufb ("xmm3","xmm1"); + &pxor ("xmm3","xmm2"); + &pshufb ("xmm3","xmm5"); + + &movdqa ("xmm2",&QWP(0x40,$inp)); + &pshufb ("xmm2","xmm4"); + &pxor ("xmm2","xmm3"); + &movdqa ("xmm3",&QWP(0x50,$inp)); + &pshufb ("xmm3","xmm1"); + &pxor ("xmm3","xmm2"); + &pshufb ("xmm3","xmm5"); + + &movdqa ("xmm2",&QWP(0x60,$inp)); + &pshufb ("xmm2","xmm4"); + &pxor ("xmm2","xmm3"); + &movdqa ("xmm3",&QWP(0x70,$inp)); + &pshufb ("xmm3","xmm1"); + &pxor ("xmm3","xmm2"); + + &add ($key,-16); + +&set_label("schedule_mangle_both"); + &movdqa ("xmm1",&QWP($k_sr,$const,$magic)); + &pshufb ("xmm3","xmm1"); + &add ($magic,-16); + &and ($magic,0x30); + &movdqu (&QWP(0,$key),"xmm3"); + &ret (); +&function_end_B("_vpaes_schedule_mangle"); + +# +# Interface to OpenSSL +# +&function_begin("${PREFIX}_set_encrypt_key"); + &mov ($inp,&wparam(0)); # inp + &lea ($base,&DWP(-56,"esp")); + &mov ($round,&wparam(1)); # bits + &and ($base,-16); + &mov ($key,&wparam(2)); # key + &xchg ($base,"esp"); # alloca + &mov (&DWP(48,"esp"),$base); + + &mov ($base,$round); + &shr ($base,5); + &add ($base,5); + &mov (&DWP(240,$key),$base); # AES_KEY->rounds = nbits/32+5; + &mov ($magic,0x30); + &mov ($out,0); + + &lea ($const,&DWP(&label("_vpaes_consts")."+0x30-".&label("pic_point"))); + &call ("_vpaes_schedule_core"); +&set_label("pic_point"); + + &mov ("esp",&DWP(48,"esp")); + &xor ("eax","eax"); +&function_end("${PREFIX}_set_encrypt_key"); + +&function_begin("${PREFIX}_set_decrypt_key"); + &mov ($inp,&wparam(0)); # inp + &lea ($base,&DWP(-56,"esp")); + &mov ($round,&wparam(1)); # bits + &and ($base,-16); + &mov ($key,&wparam(2)); # key + &xchg ($base,"esp"); # alloca + &mov (&DWP(48,"esp"),$base); + + &mov ($base,$round); + &shr ($base,5); + &add ($base,5); + &mov (&DWP(240,$key),$base); # AES_KEY->rounds = nbits/32+5; + &shl ($base,4); + &lea ($key,&DWP(16,$key,$base)); + + &mov ($out,1); + &mov ($magic,$round); + &shr ($magic,1); + &and ($magic,32); + &xor ($magic,32); # nbist==192?0:32; + + &lea ($const,&DWP(&label("_vpaes_consts")."+0x30-".&label("pic_point"))); + &call ("_vpaes_schedule_core"); +&set_label("pic_point"); + + &mov ("esp",&DWP(48,"esp")); + &xor ("eax","eax"); +&function_end("${PREFIX}_set_decrypt_key"); + +&function_begin("${PREFIX}_encrypt"); + &lea ($const,&DWP(&label("_vpaes_consts")."+0x30-".&label("pic_point"))); + &call ("_vpaes_preheat"); +&set_label("pic_point"); + &mov ($inp,&wparam(0)); # inp + &lea ($base,&DWP(-56,"esp")); + &mov ($out,&wparam(1)); # out + &and ($base,-16); + &mov ($key,&wparam(2)); # key + &xchg ($base,"esp"); # alloca + &mov (&DWP(48,"esp"),$base); + + &movdqu ("xmm0",&QWP(0,$inp)); + &call ("_vpaes_encrypt_core"); + &movdqu (&QWP(0,$out),"xmm0"); + + &mov ("esp",&DWP(48,"esp")); +&function_end("${PREFIX}_encrypt"); + +&function_begin("${PREFIX}_decrypt"); + &lea ($const,&DWP(&label("_vpaes_consts")."+0x30-".&label("pic_point"))); + &call ("_vpaes_preheat"); +&set_label("pic_point"); + &mov ($inp,&wparam(0)); # inp + &lea ($base,&DWP(-56,"esp")); + &mov ($out,&wparam(1)); # out + &and ($base,-16); + &mov ($key,&wparam(2)); # key + &xchg ($base,"esp"); # alloca + &mov (&DWP(48,"esp"),$base); + + &movdqu ("xmm0",&QWP(0,$inp)); + &call ("_vpaes_decrypt_core"); + &movdqu (&QWP(0,$out),"xmm0"); + + &mov ("esp",&DWP(48,"esp")); +&function_end("${PREFIX}_decrypt"); + +&function_begin("${PREFIX}_cbc_encrypt"); + &mov ($inp,&wparam(0)); # inp + &mov ($out,&wparam(1)); # out + &mov ($round,&wparam(2)); # len + &mov ($key,&wparam(3)); # key + &sub ($round,16); + &jc (&label("cbc_abort")); + &lea ($base,&DWP(-56,"esp")); + &mov ($const,&wparam(4)); # ivp + &and ($base,-16); + &mov ($magic,&wparam(5)); # enc + &xchg ($base,"esp"); # alloca + &movdqu ("xmm1",&QWP(0,$const)); # load IV + &sub ($out,$inp); + &mov (&DWP(48,"esp"),$base); + + &mov (&DWP(0,"esp"),$out); # save out + &mov (&DWP(4,"esp"),$key) # save key + &mov (&DWP(8,"esp"),$const); # save ivp + &mov ($out,$round); # $out works as $len + + &lea ($const,&DWP(&label("_vpaes_consts")."+0x30-".&label("pic_point"))); + &call ("_vpaes_preheat"); +&set_label("pic_point"); + &cmp ($magic,0); + &je (&label("cbc_dec_loop")); + &jmp (&label("cbc_enc_loop")); + +&set_label("cbc_enc_loop",16); + &movdqu ("xmm0",&QWP(0,$inp)); # load input + &pxor ("xmm0","xmm1"); # inp^=iv + &call ("_vpaes_encrypt_core"); + &mov ($base,&DWP(0,"esp")); # restore out + &mov ($key,&DWP(4,"esp")); # restore key + &movdqa ("xmm1","xmm0"); + &movdqu (&QWP(0,$base,$inp),"xmm0"); # write output + &lea ($inp,&DWP(16,$inp)); + &sub ($out,16); + &jnc (&label("cbc_enc_loop")); + &jmp (&label("cbc_done")); + +&set_label("cbc_dec_loop",16); + &movdqu ("xmm0",&QWP(0,$inp)); # load input + &movdqa (&QWP(16,"esp"),"xmm1"); # save IV + &movdqa (&QWP(32,"esp"),"xmm0"); # save future IV + &call ("_vpaes_decrypt_core"); + &mov ($base,&DWP(0,"esp")); # restore out + &mov ($key,&DWP(4,"esp")); # restore key + &pxor ("xmm0",&QWP(16,"esp")); # out^=iv + &movdqa ("xmm1",&QWP(32,"esp")); # load next IV + &movdqu (&QWP(0,$base,$inp),"xmm0"); # write output + &lea ($inp,&DWP(16,$inp)); + &sub ($out,16); + &jnc (&label("cbc_dec_loop")); + +&set_label("cbc_done"); + &mov ($base,&DWP(8,"esp")); # restore ivp + &mov ("esp",&DWP(48,"esp")); + &movdqu (&QWP(0,$base),"xmm1"); # write IV +&set_label("cbc_abort"); +&function_end("${PREFIX}_cbc_encrypt"); + +&asm_finish(); + +close STDOUT; diff --git a/openssl-1.1.0h/crypto/aes/asm/vpaes-x86_64.pl b/openssl-1.1.0h/crypto/aes/asm/vpaes-x86_64.pl new file mode 100644 index 0000000..422e8ee --- /dev/null +++ b/openssl-1.1.0h/crypto/aes/asm/vpaes-x86_64.pl @@ -0,0 +1,1215 @@ +#! /usr/bin/env perl +# Copyright 2011-2016 The OpenSSL Project Authors. All Rights Reserved. +# +# Licensed under the OpenSSL license (the "License"). You may not use +# this file except in compliance with the License. You can obtain a copy +# in the file LICENSE in the source distribution or at +# https://www.openssl.org/source/license.html + + +###################################################################### +## Constant-time SSSE3 AES core implementation. +## version 0.1 +## +## By Mike Hamburg (Stanford University), 2009 +## Public domain. +## +## For details see http://shiftleft.org/papers/vector_aes/ and +## http://crypto.stanford.edu/vpaes/. + +###################################################################### +# September 2011. +# +# Interface to OpenSSL as "almost" drop-in replacement for +# aes-x86_64.pl. "Almost" refers to the fact that AES_cbc_encrypt +# doesn't handle partial vectors (doesn't have to if called from +# EVP only). "Drop-in" implies that this module doesn't share key +# schedule structure with the original nor does it make assumption +# about its alignment... +# +# Performance summary. aes-x86_64.pl column lists large-block CBC +# encrypt/decrypt/with-hyper-threading-off(*) results in cycles per +# byte processed with 128-bit key, and vpaes-x86_64.pl column - +# [also large-block CBC] encrypt/decrypt. +# +# aes-x86_64.pl vpaes-x86_64.pl +# +# Core 2(**) 29.6/41.1/14.3 21.9/25.2(***) +# Nehalem 29.6/40.3/14.6 10.0/11.8 +# Atom 57.3/74.2/32.1 60.9/77.2(***) +# Silvermont 52.7/64.0/19.5 48.8/60.8(***) +# Goldmont 38.9/49.0/17.8 10.6/12.6 +# +# (*) "Hyper-threading" in the context refers rather to cache shared +# among multiple cores, than to specifically Intel HTT. As vast +# majority of contemporary cores share cache, slower code path +# is common place. In other words "with-hyper-threading-off" +# results are presented mostly for reference purposes. +# +# (**) "Core 2" refers to initial 65nm design, a.k.a. Conroe. +# +# (***) Less impressive improvement on Core 2 and Atom is due to slow +# pshufb, yet it's respectable +36%/62% improvement on Core 2 +# (as implied, over "hyper-threading-safe" code path). +# +# + +$flavour = shift; +$output = shift; +if ($flavour =~ /\./) { $output = $flavour; undef $flavour; } + +$win64=0; $win64=1 if ($flavour =~ /[nm]asm|mingw64/ || $output =~ /\.asm$/); + +$0 =~ m/(.*[\/\\])[^\/\\]+$/; $dir=$1; +( $xlate="${dir}x86_64-xlate.pl" and -f $xlate ) or +( $xlate="${dir}../../perlasm/x86_64-xlate.pl" and -f $xlate) or +die "can't locate x86_64-xlate.pl"; + +open OUT,"| \"$^X\" \"$xlate\" $flavour \"$output\""; +*STDOUT=*OUT; + +$PREFIX="vpaes"; + +$code.=<<___; +.text + +## +## _aes_encrypt_core +## +## AES-encrypt %xmm0. +## +## Inputs: +## %xmm0 = input +## %xmm9-%xmm15 as in _vpaes_preheat +## (%rdx) = scheduled keys +## +## Output in %xmm0 +## Clobbers %xmm1-%xmm5, %r9, %r10, %r11, %rax +## Preserves %xmm6 - %xmm8 so you get some local vectors +## +## +.type _vpaes_encrypt_core,\@abi-omnipotent +.align 16 +_vpaes_encrypt_core: + mov %rdx, %r9 + mov \$16, %r11 + mov 240(%rdx),%eax + movdqa %xmm9, %xmm1 + movdqa .Lk_ipt(%rip), %xmm2 # iptlo + pandn %xmm0, %xmm1 + movdqu (%r9), %xmm5 # round0 key + psrld \$4, %xmm1 + pand %xmm9, %xmm0 + pshufb %xmm0, %xmm2 + movdqa .Lk_ipt+16(%rip), %xmm0 # ipthi + pshufb %xmm1, %xmm0 + pxor %xmm5, %xmm2 + add \$16, %r9 + pxor %xmm2, %xmm0 + lea .Lk_mc_backward(%rip),%r10 + jmp .Lenc_entry + +.align 16 +.Lenc_loop: + # middle of middle round + movdqa %xmm13, %xmm4 # 4 : sb1u + movdqa %xmm12, %xmm0 # 0 : sb1t + pshufb %xmm2, %xmm4 # 4 = sb1u + pshufb %xmm3, %xmm0 # 0 = sb1t + pxor %xmm5, %xmm4 # 4 = sb1u + k + movdqa %xmm15, %xmm5 # 4 : sb2u + pxor %xmm4, %xmm0 # 0 = A + movdqa -0x40(%r11,%r10), %xmm1 # .Lk_mc_forward[] + pshufb %xmm2, %xmm5 # 4 = sb2u + movdqa (%r11,%r10), %xmm4 # .Lk_mc_backward[] + movdqa %xmm14, %xmm2 # 2 : sb2t + pshufb %xmm3, %xmm2 # 2 = sb2t + movdqa %xmm0, %xmm3 # 3 = A + pxor %xmm5, %xmm2 # 2 = 2A + pshufb %xmm1, %xmm0 # 0 = B + add \$16, %r9 # next key + pxor %xmm2, %xmm0 # 0 = 2A+B + pshufb %xmm4, %xmm3 # 3 = D + add \$16, %r11 # next mc + pxor %xmm0, %xmm3 # 3 = 2A+B+D + pshufb %xmm1, %xmm0 # 0 = 2B+C + and \$0x30, %r11 # ... mod 4 + sub \$1,%rax # nr-- + pxor %xmm3, %xmm0 # 0 = 2A+3B+C+D + +.Lenc_entry: + # top of round + movdqa %xmm9, %xmm1 # 1 : i + movdqa %xmm11, %xmm5 # 2 : a/k + pandn %xmm0, %xmm1 # 1 = i<<4 + psrld \$4, %xmm1 # 1 = i + pand %xmm9, %xmm0 # 0 = k + pshufb %xmm0, %xmm5 # 2 = a/k + movdqa %xmm10, %xmm3 # 3 : 1/i + pxor %xmm1, %xmm0 # 0 = j + pshufb %xmm1, %xmm3 # 3 = 1/i + movdqa %xmm10, %xmm4 # 4 : 1/j + pxor %xmm5, %xmm3 # 3 = iak = 1/i + a/k + pshufb %xmm0, %xmm4 # 4 = 1/j + movdqa %xmm10, %xmm2 # 2 : 1/iak + pxor %xmm5, %xmm4 # 4 = jak = 1/j + a/k + pshufb %xmm3, %xmm2 # 2 = 1/iak + movdqa %xmm10, %xmm3 # 3 : 1/jak + pxor %xmm0, %xmm2 # 2 = io + pshufb %xmm4, %xmm3 # 3 = 1/jak + movdqu (%r9), %xmm5 + pxor %xmm1, %xmm3 # 3 = jo + jnz .Lenc_loop + + # middle of last round + movdqa -0x60(%r10), %xmm4 # 3 : sbou .Lk_sbo + movdqa -0x50(%r10), %xmm0 # 0 : sbot .Lk_sbo+16 + pshufb %xmm2, %xmm4 # 4 = sbou + pxor %xmm5, %xmm4 # 4 = sb1u + k + pshufb %xmm3, %xmm0 # 0 = sb1t + movdqa 0x40(%r11,%r10), %xmm1 # .Lk_sr[] + pxor %xmm4, %xmm0 # 0 = A + pshufb %xmm1, %xmm0 + ret +.size _vpaes_encrypt_core,.-_vpaes_encrypt_core + +## +## Decryption core +## +## Same API as encryption core. +## +.type _vpaes_decrypt_core,\@abi-omnipotent +.align 16 +_vpaes_decrypt_core: + mov %rdx, %r9 # load key + mov 240(%rdx),%eax + movdqa %xmm9, %xmm1 + movdqa .Lk_dipt(%rip), %xmm2 # iptlo + pandn %xmm0, %xmm1 + mov %rax, %r11 + psrld \$4, %xmm1 + movdqu (%r9), %xmm5 # round0 key + shl \$4, %r11 + pand %xmm9, %xmm0 + pshufb %xmm0, %xmm2 + movdqa .Lk_dipt+16(%rip), %xmm0 # ipthi + xor \$0x30, %r11 + lea .Lk_dsbd(%rip),%r10 + pshufb %xmm1, %xmm0 + and \$0x30, %r11 + pxor %xmm5, %xmm2 + movdqa .Lk_mc_forward+48(%rip), %xmm5 + pxor %xmm2, %xmm0 + add \$16, %r9 + add %r10, %r11 + jmp .Ldec_entry + +.align 16 +.Ldec_loop: +## +## Inverse mix columns +## + movdqa -0x20(%r10),%xmm4 # 4 : sb9u + movdqa -0x10(%r10),%xmm1 # 0 : sb9t + pshufb %xmm2, %xmm4 # 4 = sb9u + pshufb %xmm3, %xmm1 # 0 = sb9t + pxor %xmm4, %xmm0 + movdqa 0x00(%r10),%xmm4 # 4 : sbdu + pxor %xmm1, %xmm0 # 0 = ch + movdqa 0x10(%r10),%xmm1 # 0 : sbdt + + pshufb %xmm2, %xmm4 # 4 = sbdu + pshufb %xmm5, %xmm0 # MC ch + pshufb %xmm3, %xmm1 # 0 = sbdt + pxor %xmm4, %xmm0 # 4 = ch + movdqa 0x20(%r10),%xmm4 # 4 : sbbu + pxor %xmm1, %xmm0 # 0 = ch + movdqa 0x30(%r10),%xmm1 # 0 : sbbt + + pshufb %xmm2, %xmm4 # 4 = sbbu + pshufb %xmm5, %xmm0 # MC ch + pshufb %xmm3, %xmm1 # 0 = sbbt + pxor %xmm4, %xmm0 # 4 = ch + movdqa 0x40(%r10),%xmm4 # 4 : sbeu + pxor %xmm1, %xmm0 # 0 = ch + movdqa 0x50(%r10),%xmm1 # 0 : sbet + + pshufb %xmm2, %xmm4 # 4 = sbeu + pshufb %xmm5, %xmm0 # MC ch + pshufb %xmm3, %xmm1 # 0 = sbet + pxor %xmm4, %xmm0 # 4 = ch + add \$16, %r9 # next round key + palignr \$12, %xmm5, %xmm5 + pxor %xmm1, %xmm0 # 0 = ch + sub \$1,%rax # nr-- + +.Ldec_entry: + # top of round + movdqa %xmm9, %xmm1 # 1 : i + pandn %xmm0, %xmm1 # 1 = i<<4 + movdqa %xmm11, %xmm2 # 2 : a/k + psrld \$4, %xmm1 # 1 = i + pand %xmm9, %xmm0 # 0 = k + pshufb %xmm0, %xmm2 # 2 = a/k + movdqa %xmm10, %xmm3 # 3 : 1/i + pxor %xmm1, %xmm0 # 0 = j + pshufb %xmm1, %xmm3 # 3 = 1/i + movdqa %xmm10, %xmm4 # 4 : 1/j + pxor %xmm2, %xmm3 # 3 = iak = 1/i + a/k + pshufb %xmm0, %xmm4 # 4 = 1/j + pxor %xmm2, %xmm4 # 4 = jak = 1/j + a/k + movdqa %xmm10, %xmm2 # 2 : 1/iak + pshufb %xmm3, %xmm2 # 2 = 1/iak + movdqa %xmm10, %xmm3 # 3 : 1/jak + pxor %xmm0, %xmm2 # 2 = io + pshufb %xmm4, %xmm3 # 3 = 1/jak + movdqu (%r9), %xmm0 + pxor %xmm1, %xmm3 # 3 = jo + jnz .Ldec_loop + + # middle of last round + movdqa 0x60(%r10), %xmm4 # 3 : sbou + pshufb %xmm2, %xmm4 # 4 = sbou + pxor %xmm0, %xmm4 # 4 = sb1u + k + movdqa 0x70(%r10), %xmm0 # 0 : sbot + movdqa -0x160(%r11), %xmm2 # .Lk_sr-.Lk_dsbd=-0x160 + pshufb %xmm3, %xmm0 # 0 = sb1t + pxor %xmm4, %xmm0 # 0 = A + pshufb %xmm2, %xmm0 + ret +.size _vpaes_decrypt_core,.-_vpaes_decrypt_core + +######################################################## +## ## +## AES key schedule ## +## ## +######################################################## +.type _vpaes_schedule_core,\@abi-omnipotent +.align 16 +_vpaes_schedule_core: + # rdi = key + # rsi = size in bits + # rdx = buffer + # rcx = direction. 0=encrypt, 1=decrypt + + call _vpaes_preheat # load the tables + movdqa .Lk_rcon(%rip), %xmm8 # load rcon + movdqu (%rdi), %xmm0 # load key (unaligned) + + # input transform + movdqa %xmm0, %xmm3 + lea .Lk_ipt(%rip), %r11 + call _vpaes_schedule_transform + movdqa %xmm0, %xmm7 + + lea .Lk_sr(%rip),%r10 + test %rcx, %rcx + jnz .Lschedule_am_decrypting + + # encrypting, output zeroth round key after transform + movdqu %xmm0, (%rdx) + jmp .Lschedule_go + +.Lschedule_am_decrypting: + # decrypting, output zeroth round key after shiftrows + movdqa (%r8,%r10),%xmm1 + pshufb %xmm1, %xmm3 + movdqu %xmm3, (%rdx) + xor \$0x30, %r8 + +.Lschedule_go: + cmp \$192, %esi + ja .Lschedule_256 + je .Lschedule_192 + # 128: fall though + +## +## .schedule_128 +## +## 128-bit specific part of key schedule. +## +## This schedule is really simple, because all its parts +## are accomplished by the subroutines. +## +.Lschedule_128: + mov \$10, %esi + +.Loop_schedule_128: + call _vpaes_schedule_round + dec %rsi + jz .Lschedule_mangle_last + call _vpaes_schedule_mangle # write output + jmp .Loop_schedule_128 + +## +## .aes_schedule_192 +## +## 192-bit specific part of key schedule. +## +## The main body of this schedule is the same as the 128-bit +## schedule, but with more smearing. The long, high side is +## stored in %xmm7 as before, and the short, low side is in +## the high bits of %xmm6. +## +## This schedule is somewhat nastier, however, because each +## round produces 192 bits of key material, or 1.5 round keys. +## Therefore, on each cycle we do 2 rounds and produce 3 round +## keys. +## +.align 16 +.Lschedule_192: + movdqu 8(%rdi),%xmm0 # load key part 2 (very unaligned) + call _vpaes_schedule_transform # input transform + movdqa %xmm0, %xmm6 # save short part + pxor %xmm4, %xmm4 # clear 4 + movhlps %xmm4, %xmm6 # clobber low side with zeros + mov \$4, %esi + +.Loop_schedule_192: + call _vpaes_schedule_round + palignr \$8,%xmm6,%xmm0 + call _vpaes_schedule_mangle # save key n + call _vpaes_schedule_192_smear + call _vpaes_schedule_mangle # save key n+1 + call _vpaes_schedule_round + dec %rsi + jz .Lschedule_mangle_last + call _vpaes_schedule_mangle # save key n+2 + call _vpaes_schedule_192_smear + jmp .Loop_schedule_192 + +## +## .aes_schedule_256 +## +## 256-bit specific part of key schedule. +## +## The structure here is very similar to the 128-bit +## schedule, but with an additional "low side" in +## %xmm6. The low side's rounds are the same as the +## high side's, except no rcon and no rotation. +## +.align 16 +.Lschedule_256: + movdqu 16(%rdi),%xmm0 # load key part 2 (unaligned) + call _vpaes_schedule_transform # input transform + mov \$7, %esi + +.Loop_schedule_256: + call _vpaes_schedule_mangle # output low result + movdqa %xmm0, %xmm6 # save cur_lo in xmm6 + + # high round + call _vpaes_schedule_round + dec %rsi + jz .Lschedule_mangle_last + call _vpaes_schedule_mangle + + # low round. swap xmm7 and xmm6 + pshufd \$0xFF, %xmm0, %xmm0 + movdqa %xmm7, %xmm5 + movdqa %xmm6, %xmm7 + call _vpaes_schedule_low_round + movdqa %xmm5, %xmm7 + + jmp .Loop_schedule_256 + + +## +## .aes_schedule_mangle_last +## +## Mangler for last round of key schedule +## Mangles %xmm0 +## when encrypting, outputs out(%xmm0) ^ 63 +## when decrypting, outputs unskew(%xmm0) +## +## Always called right before return... jumps to cleanup and exits +## +.align 16 +.Lschedule_mangle_last: + # schedule last round key from xmm0 + lea .Lk_deskew(%rip),%r11 # prepare to deskew + test %rcx, %rcx + jnz .Lschedule_mangle_last_dec + + # encrypting + movdqa (%r8,%r10),%xmm1 + pshufb %xmm1, %xmm0 # output permute + lea .Lk_opt(%rip), %r11 # prepare to output transform + add \$32, %rdx + +.Lschedule_mangle_last_dec: + add \$-16, %rdx + pxor .Lk_s63(%rip), %xmm0 + call _vpaes_schedule_transform # output transform + movdqu %xmm0, (%rdx) # save last key + + # cleanup + pxor %xmm0, %xmm0 + pxor %xmm1, %xmm1 + pxor %xmm2, %xmm2 + pxor %xmm3, %xmm3 + pxor %xmm4, %xmm4 + pxor %xmm5, %xmm5 + pxor %xmm6, %xmm6 + pxor %xmm7, %xmm7 + ret +.size _vpaes_schedule_core,.-_vpaes_schedule_core + +## +## .aes_schedule_192_smear +## +## Smear the short, low side in the 192-bit key schedule. +## +## Inputs: +## %xmm7: high side, b a x y +## %xmm6: low side, d c 0 0 +## %xmm13: 0 +## +## Outputs: +## %xmm6: b+c+d b+c 0 0 +## %xmm0: b+c+d b+c b a +## +.type _vpaes_schedule_192_smear,\@abi-omnipotent +.align 16 +_vpaes_schedule_192_smear: + pshufd \$0x80, %xmm6, %xmm1 # d c 0 0 -> c 0 0 0 + pshufd \$0xFE, %xmm7, %xmm0 # b a _ _ -> b b b a + pxor %xmm1, %xmm6 # -> c+d c 0 0 + pxor %xmm1, %xmm1 + pxor %xmm0, %xmm6 # -> b+c+d b+c b a + movdqa %xmm6, %xmm0 + movhlps %xmm1, %xmm6 # clobber low side with zeros + ret +.size _vpaes_schedule_192_smear,.-_vpaes_schedule_192_smear + +## +## .aes_schedule_round +## +## Runs one main round of the key schedule on %xmm0, %xmm7 +## +## Specifically, runs subbytes on the high dword of %xmm0 +## then rotates it by one byte and xors into the low dword of +## %xmm7. +## +## Adds rcon from low byte of %xmm8, then rotates %xmm8 for +## next rcon. +## +## Smears the dwords of %xmm7 by xoring the low into the +## second low, result into third, result into highest. +## +## Returns results in %xmm7 = %xmm0. +## Clobbers %xmm1-%xmm4, %r11. +## +.type _vpaes_schedule_round,\@abi-omnipotent +.align 16 +_vpaes_schedule_round: + # extract rcon from xmm8 + pxor %xmm1, %xmm1 + palignr \$15, %xmm8, %xmm1 + palignr \$15, %xmm8, %xmm8 + pxor %xmm1, %xmm7 + + # rotate + pshufd \$0xFF, %xmm0, %xmm0 + palignr \$1, %xmm0, %xmm0 + + # fall through... + + # low round: same as high round, but no rotation and no rcon. +_vpaes_schedule_low_round: + # smear xmm7 + movdqa %xmm7, %xmm1 + pslldq \$4, %xmm7 + pxor %xmm1, %xmm7 + movdqa %xmm7, %xmm1 + pslldq \$8, %xmm7 + pxor %xmm1, %xmm7 + pxor .Lk_s63(%rip), %xmm7 + + # subbytes + movdqa %xmm9, %xmm1 + pandn %xmm0, %xmm1 + psrld \$4, %xmm1 # 1 = i + pand %xmm9, %xmm0 # 0 = k + movdqa %xmm11, %xmm2 # 2 : a/k + pshufb %xmm0, %xmm2 # 2 = a/k + pxor %xmm1, %xmm0 # 0 = j + movdqa %xmm10, %xmm3 # 3 : 1/i + pshufb %xmm1, %xmm3 # 3 = 1/i + pxor %xmm2, %xmm3 # 3 = iak = 1/i + a/k + movdqa %xmm10, %xmm4 # 4 : 1/j + pshufb %xmm0, %xmm4 # 4 = 1/j + pxor %xmm2, %xmm4 # 4 = jak = 1/j + a/k + movdqa %xmm10, %xmm2 # 2 : 1/iak + pshufb %xmm3, %xmm2 # 2 = 1/iak + pxor %xmm0, %xmm2 # 2 = io + movdqa %xmm10, %xmm3 # 3 : 1/jak + pshufb %xmm4, %xmm3 # 3 = 1/jak + pxor %xmm1, %xmm3 # 3 = jo + movdqa %xmm13, %xmm4 # 4 : sbou + pshufb %xmm2, %xmm4 # 4 = sbou + movdqa %xmm12, %xmm0 # 0 : sbot + pshufb %xmm3, %xmm0 # 0 = sb1t + pxor %xmm4, %xmm0 # 0 = sbox output + + # add in smeared stuff + pxor %xmm7, %xmm0 + movdqa %xmm0, %xmm7 + ret +.size _vpaes_schedule_round,.-_vpaes_schedule_round + +## +## .aes_schedule_transform +## +## Linear-transform %xmm0 according to tables at (%r11) +## +## Requires that %xmm9 = 0x0F0F... as in preheat +## Output in %xmm0 +## Clobbers %xmm1, %xmm2 +## +.type _vpaes_schedule_transform,\@abi-omnipotent +.align 16 +_vpaes_schedule_transform: + movdqa %xmm9, %xmm1 + pandn %xmm0, %xmm1 + psrld \$4, %xmm1 + pand %xmm9, %xmm0 + movdqa (%r11), %xmm2 # lo + pshufb %xmm0, %xmm2 + movdqa 16(%r11), %xmm0 # hi + pshufb %xmm1, %xmm0 + pxor %xmm2, %xmm0 + ret +.size _vpaes_schedule_transform,.-_vpaes_schedule_transform + +## +## .aes_schedule_mangle +## +## Mangle xmm0 from (basis-transformed) standard version +## to our version. +## +## On encrypt, +## xor with 0x63 +## multiply by circulant 0,1,1,1 +## apply shiftrows transform +## +## On decrypt, +## xor with 0x63 +## multiply by "inverse mixcolumns" circulant E,B,D,9 +## deskew +## apply shiftrows transform +## +## +## Writes out to (%rdx), and increments or decrements it +## Keeps track of round number mod 4 in %r8 +## Preserves xmm0 +## Clobbers xmm1-xmm5 +## +.type _vpaes_schedule_mangle,\@abi-omnipotent +.align 16 +_vpaes_schedule_mangle: + movdqa %xmm0, %xmm4 # save xmm0 for later + movdqa .Lk_mc_forward(%rip),%xmm5 + test %rcx, %rcx + jnz .Lschedule_mangle_dec + + # encrypting + add \$16, %rdx + pxor .Lk_s63(%rip),%xmm4 + pshufb %xmm5, %xmm4 + movdqa %xmm4, %xmm3 + pshufb %xmm5, %xmm4 + pxor %xmm4, %xmm3 + pshufb %xmm5, %xmm4 + pxor %xmm4, %xmm3 + + jmp .Lschedule_mangle_both +.align 16 +.Lschedule_mangle_dec: + # inverse mix columns + lea .Lk_dksd(%rip),%r11 + movdqa %xmm9, %xmm1 + pandn %xmm4, %xmm1 + psrld \$4, %xmm1 # 1 = hi + pand %xmm9, %xmm4 # 4 = lo + + movdqa 0x00(%r11), %xmm2 + pshufb %xmm4, %xmm2 + movdqa 0x10(%r11), %xmm3 + pshufb %xmm1, %xmm3 + pxor %xmm2, %xmm3 + pshufb %xmm5, %xmm3 + + movdqa 0x20(%r11), %xmm2 + pshufb %xmm4, %xmm2 + pxor %xmm3, %xmm2 + movdqa 0x30(%r11), %xmm3 + pshufb %xmm1, %xmm3 + pxor %xmm2, %xmm3 + pshufb %xmm5, %xmm3 + + movdqa 0x40(%r11), %xmm2 + pshufb %xmm4, %xmm2 + pxor %xmm3, %xmm2 + movdqa 0x50(%r11), %xmm3 + pshufb %xmm1, %xmm3 + pxor %xmm2, %xmm3 + pshufb %xmm5, %xmm3 + + movdqa 0x60(%r11), %xmm2 + pshufb %xmm4, %xmm2 + pxor %xmm3, %xmm2 + movdqa 0x70(%r11), %xmm3 + pshufb %xmm1, %xmm3 + pxor %xmm2, %xmm3 + + add \$-16, %rdx + +.Lschedule_mangle_both: + movdqa (%r8,%r10),%xmm1 + pshufb %xmm1,%xmm3 + add \$-16, %r8 + and \$0x30, %r8 + movdqu %xmm3, (%rdx) + ret +.size _vpaes_schedule_mangle,.-_vpaes_schedule_mangle + +# +# Interface to OpenSSL +# +.globl ${PREFIX}_set_encrypt_key +.type ${PREFIX}_set_encrypt_key,\@function,3 +.align 16 +${PREFIX}_set_encrypt_key: +___ +$code.=<<___ if ($win64); + lea -0xb8(%rsp),%rsp + movaps %xmm6,0x10(%rsp) + movaps %xmm7,0x20(%rsp) + movaps %xmm8,0x30(%rsp) + movaps %xmm9,0x40(%rsp) + movaps %xmm10,0x50(%rsp) + movaps %xmm11,0x60(%rsp) + movaps %xmm12,0x70(%rsp) + movaps %xmm13,0x80(%rsp) + movaps %xmm14,0x90(%rsp) + movaps %xmm15,0xa0(%rsp) +.Lenc_key_body: +___ +$code.=<<___; + mov %esi,%eax + shr \$5,%eax + add \$5,%eax + mov %eax,240(%rdx) # AES_KEY->rounds = nbits/32+5; + + mov \$0,%ecx + mov \$0x30,%r8d + call _vpaes_schedule_core +___ +$code.=<<___ if ($win64); + movaps 0x10(%rsp),%xmm6 + movaps 0x20(%rsp),%xmm7 + movaps 0x30(%rsp),%xmm8 + movaps 0x40(%rsp),%xmm9 + movaps 0x50(%rsp),%xmm10 + movaps 0x60(%rsp),%xmm11 + movaps 0x70(%rsp),%xmm12 + movaps 0x80(%rsp),%xmm13 + movaps 0x90(%rsp),%xmm14 + movaps 0xa0(%rsp),%xmm15 + lea 0xb8(%rsp),%rsp +.Lenc_key_epilogue: +___ +$code.=<<___; + xor %eax,%eax + ret +.size ${PREFIX}_set_encrypt_key,.-${PREFIX}_set_encrypt_key + +.globl ${PREFIX}_set_decrypt_key +.type ${PREFIX}_set_decrypt_key,\@function,3 +.align 16 +${PREFIX}_set_decrypt_key: +___ +$code.=<<___ if ($win64); + lea -0xb8(%rsp),%rsp + movaps %xmm6,0x10(%rsp) + movaps %xmm7,0x20(%rsp) + movaps %xmm8,0x30(%rsp) + movaps %xmm9,0x40(%rsp) + movaps %xmm10,0x50(%rsp) + movaps %xmm11,0x60(%rsp) + movaps %xmm12,0x70(%rsp) + movaps %xmm13,0x80(%rsp) + movaps %xmm14,0x90(%rsp) + movaps %xmm15,0xa0(%rsp) +.Ldec_key_body: +___ +$code.=<<___; + mov %esi,%eax + shr \$5,%eax + add \$5,%eax + mov %eax,240(%rdx) # AES_KEY->rounds = nbits/32+5; + shl \$4,%eax + lea 16(%rdx,%rax),%rdx + + mov \$1,%ecx + mov %esi,%r8d + shr \$1,%r8d + and \$32,%r8d + xor \$32,%r8d # nbits==192?0:32 + call _vpaes_schedule_core +___ +$code.=<<___ if ($win64); + movaps 0x10(%rsp),%xmm6 + movaps 0x20(%rsp),%xmm7 + movaps 0x30(%rsp),%xmm8 + movaps 0x40(%rsp),%xmm9 + movaps 0x50(%rsp),%xmm10 + movaps 0x60(%rsp),%xmm11 + movaps 0x70(%rsp),%xmm12 + movaps 0x80(%rsp),%xmm13 + movaps 0x90(%rsp),%xmm14 + movaps 0xa0(%rsp),%xmm15 + lea 0xb8(%rsp),%rsp +.Ldec_key_epilogue: +___ +$code.=<<___; + xor %eax,%eax + ret +.size ${PREFIX}_set_decrypt_key,.-${PREFIX}_set_decrypt_key + +.globl ${PREFIX}_encrypt +.type ${PREFIX}_encrypt,\@function,3 +.align 16 +${PREFIX}_encrypt: +___ +$code.=<<___ if ($win64); + lea -0xb8(%rsp),%rsp + movaps %xmm6,0x10(%rsp) + movaps %xmm7,0x20(%rsp) + movaps %xmm8,0x30(%rsp) + movaps %xmm9,0x40(%rsp) + movaps %xmm10,0x50(%rsp) + movaps %xmm11,0x60(%rsp) + movaps %xmm12,0x70(%rsp) + movaps %xmm13,0x80(%rsp) + movaps %xmm14,0x90(%rsp) + movaps %xmm15,0xa0(%rsp) +.Lenc_body: +___ +$code.=<<___; + movdqu (%rdi),%xmm0 + call _vpaes_preheat + call _vpaes_encrypt_core + movdqu %xmm0,(%rsi) +___ +$code.=<<___ if ($win64); + movaps 0x10(%rsp),%xmm6 + movaps 0x20(%rsp),%xmm7 + movaps 0x30(%rsp),%xmm8 + movaps 0x40(%rsp),%xmm9 + movaps 0x50(%rsp),%xmm10 + movaps 0x60(%rsp),%xmm11 + movaps 0x70(%rsp),%xmm12 + movaps 0x80(%rsp),%xmm13 + movaps 0x90(%rsp),%xmm14 + movaps 0xa0(%rsp),%xmm15 + lea 0xb8(%rsp),%rsp +.Lenc_epilogue: +___ +$code.=<<___; + ret +.size ${PREFIX}_encrypt,.-${PREFIX}_encrypt + +.globl ${PREFIX}_decrypt +.type ${PREFIX}_decrypt,\@function,3 +.align 16 +${PREFIX}_decrypt: +___ +$code.=<<___ if ($win64); + lea -0xb8(%rsp),%rsp + movaps %xmm6,0x10(%rsp) + movaps %xmm7,0x20(%rsp) + movaps %xmm8,0x30(%rsp) + movaps %xmm9,0x40(%rsp) + movaps %xmm10,0x50(%rsp) + movaps %xmm11,0x60(%rsp) + movaps %xmm12,0x70(%rsp) + movaps %xmm13,0x80(%rsp) + movaps %xmm14,0x90(%rsp) + movaps %xmm15,0xa0(%rsp) +.Ldec_body: +___ +$code.=<<___; + movdqu (%rdi),%xmm0 + call _vpaes_preheat + call _vpaes_decrypt_core + movdqu %xmm0,(%rsi) +___ +$code.=<<___ if ($win64); + movaps 0x10(%rsp),%xmm6 + movaps 0x20(%rsp),%xmm7 + movaps 0x30(%rsp),%xmm8 + movaps 0x40(%rsp),%xmm9 + movaps 0x50(%rsp),%xmm10 + movaps 0x60(%rsp),%xmm11 + movaps 0x70(%rsp),%xmm12 + movaps 0x80(%rsp),%xmm13 + movaps 0x90(%rsp),%xmm14 + movaps 0xa0(%rsp),%xmm15 + lea 0xb8(%rsp),%rsp +.Ldec_epilogue: +___ +$code.=<<___; + ret +.size ${PREFIX}_decrypt,.-${PREFIX}_decrypt +___ +{ +my ($inp,$out,$len,$key,$ivp,$enc)=("%rdi","%rsi","%rdx","%rcx","%r8","%r9"); +# void AES_cbc_encrypt (const void char *inp, unsigned char *out, +# size_t length, const AES_KEY *key, +# unsigned char *ivp,const int enc); +$code.=<<___; +.globl ${PREFIX}_cbc_encrypt +.type ${PREFIX}_cbc_encrypt,\@function,6 +.align 16 +${PREFIX}_cbc_encrypt: + xchg $key,$len +___ +($len,$key)=($key,$len); +$code.=<<___; + sub \$16,$len + jc .Lcbc_abort +___ +$code.=<<___ if ($win64); + lea -0xb8(%rsp),%rsp + movaps %xmm6,0x10(%rsp) + movaps %xmm7,0x20(%rsp) + movaps %xmm8,0x30(%rsp) + movaps %xmm9,0x40(%rsp) + movaps %xmm10,0x50(%rsp) + movaps %xmm11,0x60(%rsp) + movaps %xmm12,0x70(%rsp) + movaps %xmm13,0x80(%rsp) + movaps %xmm14,0x90(%rsp) + movaps %xmm15,0xa0(%rsp) +.Lcbc_body: +___ +$code.=<<___; + movdqu ($ivp),%xmm6 # load IV + sub $inp,$out + call _vpaes_preheat + cmp \$0,${enc}d + je .Lcbc_dec_loop + jmp .Lcbc_enc_loop +.align 16 +.Lcbc_enc_loop: + movdqu ($inp),%xmm0 + pxor %xmm6,%xmm0 + call _vpaes_encrypt_core + movdqa %xmm0,%xmm6 + movdqu %xmm0,($out,$inp) + lea 16($inp),$inp + sub \$16,$len + jnc .Lcbc_enc_loop + jmp .Lcbc_done +.align 16 +.Lcbc_dec_loop: + movdqu ($inp),%xmm0 + movdqa %xmm0,%xmm7 + call _vpaes_decrypt_core + pxor %xmm6,%xmm0 + movdqa %xmm7,%xmm6 + movdqu %xmm0,($out,$inp) + lea 16($inp),$inp + sub \$16,$len + jnc .Lcbc_dec_loop +.Lcbc_done: + movdqu %xmm6,($ivp) # save IV +___ +$code.=<<___ if ($win64); + movaps 0x10(%rsp),%xmm6 + movaps 0x20(%rsp),%xmm7 + movaps 0x30(%rsp),%xmm8 + movaps 0x40(%rsp),%xmm9 + movaps 0x50(%rsp),%xmm10 + movaps 0x60(%rsp),%xmm11 + movaps 0x70(%rsp),%xmm12 + movaps 0x80(%rsp),%xmm13 + movaps 0x90(%rsp),%xmm14 + movaps 0xa0(%rsp),%xmm15 + lea 0xb8(%rsp),%rsp +.Lcbc_epilogue: +___ +$code.=<<___; +.Lcbc_abort: + ret +.size ${PREFIX}_cbc_encrypt,.-${PREFIX}_cbc_encrypt +___ +} +$code.=<<___; +## +## _aes_preheat +## +## Fills register %r10 -> .aes_consts (so you can -fPIC) +## and %xmm9-%xmm15 as specified below. +## +.type _vpaes_preheat,\@abi-omnipotent +.align 16 +_vpaes_preheat: + lea .Lk_s0F(%rip), %r10 + movdqa -0x20(%r10), %xmm10 # .Lk_inv + movdqa -0x10(%r10), %xmm11 # .Lk_inv+16 + movdqa 0x00(%r10), %xmm9 # .Lk_s0F + movdqa 0x30(%r10), %xmm13 # .Lk_sb1 + movdqa 0x40(%r10), %xmm12 # .Lk_sb1+16 + movdqa 0x50(%r10), %xmm15 # .Lk_sb2 + movdqa 0x60(%r10), %xmm14 # .Lk_sb2+16 + ret +.size _vpaes_preheat,.-_vpaes_preheat +######################################################## +## ## +## Constants ## +## ## +######################################################## +.type _vpaes_consts,\@object +.align 64 +_vpaes_consts: +.Lk_inv: # inv, inva + .quad 0x0E05060F0D080180, 0x040703090A0B0C02 + .quad 0x01040A060F0B0780, 0x030D0E0C02050809 + +.Lk_s0F: # s0F + .quad 0x0F0F0F0F0F0F0F0F, 0x0F0F0F0F0F0F0F0F + +.Lk_ipt: # input transform (lo, hi) + .quad 0xC2B2E8985A2A7000, 0xCABAE09052227808 + .quad 0x4C01307D317C4D00, 0xCD80B1FCB0FDCC81 + +.Lk_sb1: # sb1u, sb1t + .quad 0xB19BE18FCB503E00, 0xA5DF7A6E142AF544 + .quad 0x3618D415FAE22300, 0x3BF7CCC10D2ED9EF +.Lk_sb2: # sb2u, sb2t + .quad 0xE27A93C60B712400, 0x5EB7E955BC982FCD + .quad 0x69EB88400AE12900, 0xC2A163C8AB82234A +.Lk_sbo: # sbou, sbot + .quad 0xD0D26D176FBDC700, 0x15AABF7AC502A878 + .quad 0xCFE474A55FBB6A00, 0x8E1E90D1412B35FA + +.Lk_mc_forward: # mc_forward + .quad 0x0407060500030201, 0x0C0F0E0D080B0A09 + .quad 0x080B0A0904070605, 0x000302010C0F0E0D + .quad 0x0C0F0E0D080B0A09, 0x0407060500030201 + .quad 0x000302010C0F0E0D, 0x080B0A0904070605 + +.Lk_mc_backward:# mc_backward + .quad 0x0605040702010003, 0x0E0D0C0F0A09080B + .quad 0x020100030E0D0C0F, 0x0A09080B06050407 + .quad 0x0E0D0C0F0A09080B, 0x0605040702010003 + .quad 0x0A09080B06050407, 0x020100030E0D0C0F + +.Lk_sr: # sr + .quad 0x0706050403020100, 0x0F0E0D0C0B0A0908 + .quad 0x030E09040F0A0500, 0x0B06010C07020D08 + .quad 0x0F060D040B020900, 0x070E050C030A0108 + .quad 0x0B0E0104070A0D00, 0x0306090C0F020508 + +.Lk_rcon: # rcon + .quad 0x1F8391B9AF9DEEB6, 0x702A98084D7C7D81 + +.Lk_s63: # s63: all equal to 0x63 transformed + .quad 0x5B5B5B5B5B5B5B5B, 0x5B5B5B5B5B5B5B5B + +.Lk_opt: # output transform + .quad 0xFF9F4929D6B66000, 0xF7974121DEBE6808 + .quad 0x01EDBD5150BCEC00, 0xE10D5DB1B05C0CE0 + +.Lk_deskew: # deskew tables: inverts the sbox's "skew" + .quad 0x07E4A34047A4E300, 0x1DFEB95A5DBEF91A + .quad 0x5F36B5DC83EA6900, 0x2841C2ABF49D1E77 + +## +## Decryption stuff +## Key schedule constants +## +.Lk_dksd: # decryption key schedule: invskew x*D + .quad 0xFEB91A5DA3E44700, 0x0740E3A45A1DBEF9 + .quad 0x41C277F4B5368300, 0x5FDC69EAAB289D1E +.Lk_dksb: # decryption key schedule: invskew x*B + .quad 0x9A4FCA1F8550D500, 0x03D653861CC94C99 + .quad 0x115BEDA7B6FC4A00, 0xD993256F7E3482C8 +.Lk_dkse: # decryption key schedule: invskew x*E + 0x63 + .quad 0xD5031CCA1FC9D600, 0x53859A4C994F5086 + .quad 0xA23196054FDC7BE8, 0xCD5EF96A20B31487 +.Lk_dks9: # decryption key schedule: invskew x*9 + .quad 0xB6116FC87ED9A700, 0x4AED933482255BFC + .quad 0x4576516227143300, 0x8BB89FACE9DAFDCE + +## +## Decryption stuff +## Round function constants +## +.Lk_dipt: # decryption input transform + .quad 0x0F505B040B545F00, 0x154A411E114E451A + .quad 0x86E383E660056500, 0x12771772F491F194 + +.Lk_dsb9: # decryption sbox output *9*u, *9*t + .quad 0x851C03539A86D600, 0xCAD51F504F994CC9 + .quad 0xC03B1789ECD74900, 0x725E2C9EB2FBA565 +.Lk_dsbd: # decryption sbox output *D*u, *D*t + .quad 0x7D57CCDFE6B1A200, 0xF56E9B13882A4439 + .quad 0x3CE2FAF724C6CB00, 0x2931180D15DEEFD3 +.Lk_dsbb: # decryption sbox output *B*u, *B*t + .quad 0xD022649296B44200, 0x602646F6B0F2D404 + .quad 0xC19498A6CD596700, 0xF3FF0C3E3255AA6B +.Lk_dsbe: # decryption sbox output *E*u, *E*t + .quad 0x46F2929626D4D000, 0x2242600464B4F6B0 + .quad 0x0C55A6CDFFAAC100, 0x9467F36B98593E32 +.Lk_dsbo: # decryption sbox final output + .quad 0x1387EA537EF94000, 0xC7AA6DB9D4943E2D + .quad 0x12D7560F93441D00, 0xCA4B8159D8C58E9C +.asciz "Vector Permutation AES for x86_64/SSSE3, Mike Hamburg (Stanford University)" +.align 64 +.size _vpaes_consts,.-_vpaes_consts +___ + +if ($win64) { +# EXCEPTION_DISPOSITION handler (EXCEPTION_RECORD *rec,ULONG64 frame, +# CONTEXT *context,DISPATCHER_CONTEXT *disp) +$rec="%rcx"; +$frame="%rdx"; +$context="%r8"; +$disp="%r9"; + +$code.=<<___; +.extern __imp_RtlVirtualUnwind +.type se_handler,\@abi-omnipotent +.align 16 +se_handler: + push %rsi + push %rdi + push %rbx + push %rbp + push %r12 + push %r13 + push %r14 + push %r15 + pushfq + sub \$64,%rsp + + mov 120($context),%rax # pull context->Rax + mov 248($context),%rbx # pull context->Rip + + mov 8($disp),%rsi # disp->ImageBase + mov 56($disp),%r11 # disp->HandlerData + + mov 0(%r11),%r10d # HandlerData[0] + lea (%rsi,%r10),%r10 # prologue label + cmp %r10,%rbx # context->RipRsp + + mov 4(%r11),%r10d # HandlerData[1] + lea (%rsi,%r10),%r10 # epilogue label + cmp %r10,%rbx # context->Rip>=epilogue label + jae .Lin_prologue + + lea 16(%rax),%rsi # %xmm save area + lea 512($context),%rdi # &context.Xmm6 + mov \$20,%ecx # 10*sizeof(%xmm0)/sizeof(%rax) + .long 0xa548f3fc # cld; rep movsq + lea 0xb8(%rax),%rax # adjust stack pointer + +.Lin_prologue: + mov 8(%rax),%rdi + mov 16(%rax),%rsi + mov %rax,152($context) # restore context->Rsp + mov %rsi,168($context) # restore context->Rsi + mov %rdi,176($context) # restore context->Rdi + + mov 40($disp),%rdi # disp->ContextRecord + mov $context,%rsi # context + mov \$`1232/8`,%ecx # sizeof(CONTEXT) + .long 0xa548f3fc # cld; rep movsq + + mov $disp,%rsi + xor %rcx,%rcx # arg1, UNW_FLAG_NHANDLER + mov 8(%rsi),%rdx # arg2, disp->ImageBase + mov 0(%rsi),%r8 # arg3, disp->ControlPc + mov 16(%rsi),%r9 # arg4, disp->FunctionEntry + mov 40(%rsi),%r10 # disp->ContextRecord + lea 56(%rsi),%r11 # &disp->HandlerData + lea 24(%rsi),%r12 # &disp->EstablisherFrame + mov %r10,32(%rsp) # arg5 + mov %r11,40(%rsp) # arg6 + mov %r12,48(%rsp) # arg7 + mov %rcx,56(%rsp) # arg8, (NULL) + call *__imp_RtlVirtualUnwind(%rip) + + mov \$1,%eax # ExceptionContinueSearch + add \$64,%rsp + popfq + pop %r15 + pop %r14 + pop %r13 + pop %r12 + pop %rbp + pop %rbx + pop %rdi + pop %rsi + ret +.size se_handler,.-se_handler + +.section .pdata +.align 4 + .rva .LSEH_begin_${PREFIX}_set_encrypt_key + .rva .LSEH_end_${PREFIX}_set_encrypt_key + .rva .LSEH_info_${PREFIX}_set_encrypt_key + + .rva .LSEH_begin_${PREFIX}_set_decrypt_key + .rva .LSEH_end_${PREFIX}_set_decrypt_key + .rva .LSEH_info_${PREFIX}_set_decrypt_key + + .rva .LSEH_begin_${PREFIX}_encrypt + .rva .LSEH_end_${PREFIX}_encrypt + .rva .LSEH_info_${PREFIX}_encrypt + + .rva .LSEH_begin_${PREFIX}_decrypt + .rva .LSEH_end_${PREFIX}_decrypt + .rva .LSEH_info_${PREFIX}_decrypt + + .rva .LSEH_begin_${PREFIX}_cbc_encrypt + .rva .LSEH_end_${PREFIX}_cbc_encrypt + .rva .LSEH_info_${PREFIX}_cbc_encrypt + +.section .xdata +.align 8 +.LSEH_info_${PREFIX}_set_encrypt_key: + .byte 9,0,0,0 + .rva se_handler + .rva .Lenc_key_body,.Lenc_key_epilogue # HandlerData[] +.LSEH_info_${PREFIX}_set_decrypt_key: + .byte 9,0,0,0 + .rva se_handler + .rva .Ldec_key_body,.Ldec_key_epilogue # HandlerData[] +.LSEH_info_${PREFIX}_encrypt: + .byte 9,0,0,0 + .rva se_handler + .rva .Lenc_body,.Lenc_epilogue # HandlerData[] +.LSEH_info_${PREFIX}_decrypt: + .byte 9,0,0,0 + .rva se_handler + .rva .Ldec_body,.Ldec_epilogue # HandlerData[] +.LSEH_info_${PREFIX}_cbc_encrypt: + .byte 9,0,0,0 + .rva se_handler + .rva .Lcbc_body,.Lcbc_epilogue # HandlerData[] +___ +} + +$code =~ s/\`([^\`]*)\`/eval($1)/gem; + +print $code; + +close STDOUT; diff --git a/openssl-1.1.0h/crypto/aes/build.info b/openssl-1.1.0h/crypto/aes/build.info new file mode 100644 index 0000000..cf6cb5e --- /dev/null +++ b/openssl-1.1.0h/crypto/aes/build.info @@ -0,0 +1,57 @@ +LIBS=../../libcrypto +SOURCE[../../libcrypto]=\ + aes_misc.c aes_ecb.c aes_cfb.c aes_ofb.c \ + aes_ige.c aes_wrap.c {- $target{aes_asm_src} -} + +GENERATE[aes-ia64.s]=asm/aes-ia64.S + +GENERATE[aes-586.s]=asm/aes-586.pl $(PERLASM_SCHEME) $(CFLAGS) $(LIB_CFLAGS) $(PROCESSOR) +DEPEND[aes-586.s]=../perlasm/x86asm.pl +GENERATE[vpaes-x86.s]=asm/vpaes-x86.pl $(PERLASM_SCHEME) $(CFLAGS) $(LIB_CFLAGS) $(PROCESSOR) +DEPEND[vpaes-586.s]=../perlasm/x86asm.pl +GENERATE[aesni-x86.s]=asm/aesni-x86.pl $(PERLASM_SCHEME) $(CFLAGS) $(LIB_CFLAGS) $(PROCESSOR) +DEPEND[aesni-586.s]=../perlasm/x86asm.pl + +GENERATE[aes-x86_64.s]=asm/aes-x86_64.pl $(PERLASM_SCHEME) +GENERATE[vpaes-x86_64.s]=asm/vpaes-x86_64.pl $(PERLASM_SCHEME) +GENERATE[bsaes-x86_64.s]=asm/bsaes-x86_64.pl $(PERLASM_SCHEME) +GENERATE[aesni-x86_64.s]=asm/aesni-x86_64.pl $(PERLASM_SCHEME) +GENERATE[aesni-sha1-x86_64.s]=asm/aesni-sha1-x86_64.pl $(PERLASM_SCHEME) +GENERATE[aesni-sha256-x86_64.s]=asm/aesni-sha256-x86_64.pl $(PERLASM_SCHEME) +GENERATE[aesni-mb-x86_64.s]=asm/aesni-mb-x86_64.pl $(PERLASM_SCHEME) + +GENERATE[aes-sparcv9.S]=asm/aes-sparcv9.pl $(PERLASM_SCHEME) +INCLUDE[aes-sparcv9.o]=.. +GENERATE[aest4-sparcv9.S]=asm/aest4-sparcv9.pl $(PERLASM_SCHEME) +INCLUDE[aest4-sparcv9.o]=.. +DEPEND[aest4-sparcv9.S]=../perlasm/sparcv9_modes.pl +GENERATE[aesfx-sparcv9.S]=asm/aesfx-sparcv9.pl $(PERLASM_SCHEME) +INCLUDE[aesfx-sparcv9.o]=.. + +GENERATE[aes-ppc.s]=asm/aes-ppc.pl $(PERLASM_SCHEME) +GENERATE[vpaes-ppc.s]=asm/vpaes-ppc.pl $(PERLASM_SCHEME) +GENERATE[aesp8-ppc.s]=asm/aesp8-ppc.pl $(PERLASM_SCHEME) + +GENERATE[aes-parisc.s]=asm/aes-parisc.pl $(PERLASM_SCHEME) + +GENERATE[aes-mips.S]=asm/aes-mips.pl $(PERLASM_SCHEME) + +GENERATE[aesv8-armx.S]=asm/aesv8-armx.pl $(PERLASM_SCHEME) +INCLUDE[aesv8-armx.o]=.. +GENERATE[vpaes-armv8.S]=asm/vpaes-armv8.pl $(PERLASM_SCHEME) + +GENERATE[aes-armv4.S]=asm/aes-armv4.pl $(PERLASM_SCHEME) +INCLUDE[aes-armv4.o]=.. +GENERATE[bsaes-armv7.S]=asm/bsaes-armv7.pl $(PERLASM_SCHEME) +INCLUDE[bsaes-armv7.o]=.. + +BEGINRAW[Makefile] +##### AES assembler implementations + +# GNU make "catch all" +{- $builddir -}/aes-%.S: {- $sourcedir -}/asm/aes-%.pl + CC="$(CC)" $(PERL) $< $(PERLASM_SCHEME) $@ +{- $builddir -}/bsaes-%.S: {- $sourcedir -}/asm/bsaes-%.pl + CC="$(CC)" $(PERL) $< $(PERLASM_SCHEME) $@ + +ENDRAW[Makefile] -- cgit v1.2.3