rewrite the stage 1 of bootloader in a more compact and (mostly) load-addr-independent way

7 files changed, 97 insertions, 103 deletions

diff --git a/Makefile b/Makefile
index 92d200f..29efa86 100644
--- a/Makefile
+++ b/Makefile
@@ -25,8 +25,14 @@ PL_0_test.elf : PL0_test.o uart.o
 kernel.elf : boot.o kernel.o uart.o demo_functionality.o paging.o interrupt_vector.o interrupts.o ramfs_embeddable.o ramfs.o
 	arm-none-eabi-gcc -T linker.ld -o $@ $(ELFFLAGS) $^
 
-loader.elf : boot.o loader_stage1.o loader_stage2.o uart.o
-	arm-none-eabi-gcc -T loader_linker.ld -o $@ $(ELFFLAGS) $^
+loader_stage2.elf : loader_stage2.o uart.o
+	arm-none-eabi-gcc -T loader_stage2_linker.ld -o $@ $(ELFFLAGS) $^
+
+loader_stage1.o : loader_stage1.S  loader_stage2.img
+	arm-none-eabi-as -mcpu=cortex-a7 $< -o $@
+
+loader.elf : loader_stage1.o
+	arm-none-eabi-gcc -T loader_stage1_linker.ld -o $@ $(ELFFLAGS) $^
 
 loader.img : loader.elf
 	arm-none-eabi-objcopy $^ -O binary $@
@@ -36,7 +42,7 @@ loader.img : loader.elf
 qemu-elf : kernel.elf
 	qemu-system-arm -m 256 -M raspi2 -serial stdio -kernel $^
 
-qemu-bin : loader.elf kernel.img pipe_image
+qemu-bin : loader.img kernel.img pipe_image
 	./pipe_image --stdout | qemu-system-arm -m 256 -M raspi2 -serial stdio -kernel $<
 
 run-on-rpi : kernel.img pipe_image
diff --git a/loader_linker.ld b/loader_linker.ld
deleted file mode 100644
index 45c88ff..0000000
--- a/loader_linker.ld
+++ /dev/null
@@ -1,72 +0,0 @@
-ENTRY(_boot)
- 
-SECTIONS
-{
-    /* Starts at LOADER_ADDR. */
-    /* Warning! Internet says RPis in 32-bit mode load binary at 0x8000! */
-    /* My experiments do, however, show, that qemu emulating RPi2 */
-    /* loads it at 0x10000! (took some pain to find out) */
-    
-    /* . = 0x10000; */
-    
-    /* rpi-open-firmware, on the other hand, loads it at 0x2000000 */
-    /* (and this should be not-so-hard to change by modifying the */
-    /* firmware) */
-    
-    . = 0x2000000;
-
-    /* For AArch64, use . = 0x80000; Unless this too is wrong */
-    
-    __start = .;
-    __text_stage1_start = .;
-    .text_stage1 :
-    {
-        KEEP(*(.text.boot))
-        *(.text.boot)
-        *(.text.stage1main)
-    }
-    . = ALIGN(4096); /* align to page size */
-    __text_stage1_end = .;
-
-    __stage2_start = .;
-    __text_start = .;
-    .text :
-    {
-        KEEP(*(.text.stage2main))
-        *(.text.boot)
-    }
-    . = ALIGN(4096); /* align to page size */
-    __text_end = .;
-    
-    __rodata_start = .;
-    .rodata :
-    {
-        *(.rodata)
-    }
-    . = ALIGN(4096); /* align to page size */
-    __rodata_end = .;
- 
-    __data_start = .;
-    .data :
-    {
-        *(.data)
-    }
-    . = ALIGN(4096); /* align to page size */
-    __data_end = .;
- 
-    __bss_start = .;
-    .bss :
-    {
-        bss = .;
-        *(.bss)
-    }
-    . = ALIGN(4096); /* align to page size */
-    __bss_end = .;
-    __bss_size = __bss_end - __bss_start;
-
-    __stage2_end = .;
-    __stage2_size = __stage2_end - __stage2_start;
-    __end = .;
-
-    _stack_top = 0x4000;
-}
diff --git a/loader_stage1.S b/loader_stage1.S
new file mode 100644
index 0000000..adf2e12
--- /dev/null
+++ b/loader_stage1.S
@@ -0,0 +1,55 @@
+// armv7 mode
+ 
+// Entry point for the kernel.
+// r15 -> should begin execution at 0x8000.
+// r0 -> 0x00000000
+// r1 -> 0x00000C42
+// r2 -> 0x00000100 - start of ATAGS
+// preserve these registers as argument for kernel
+
+.global _boot
+_boot:
+        // Only let the first core execute
+        mrc p15, 0, r3, c0, c0, 5
+        and r3, r3, #3
+        cmp r3, #0
+        beq proceed
+	// this is a kind of blef - races can theoretically still occur
+	// when the main core overwrites this part of memory
+	wfe
+
+proceed:
+	// copy stage2 of the loader to address 0x4000
+
+	// first, load address of stage2_start to r3 (a PIC way)
+	adr r3, stage2_start
+
+	// load destination address for stage2 code to r4
+	mov r4, #0x4000
+
+	// load blob size to r5
+	mov r5, #(stage2_end - stage2_start)
+
+	// r6 is the counter - counts the bytes copied
+	mov r6, #0
+
+	// each word of the blob is loaded to r7 and stored
+	// from r7 to it's destination in a loop
+loop:
+	ldr r7, [r3, r6]
+	str r7, [r4, r6]
+	add r6, r6, #4
+	cmp r6, r5
+	blo loop
+	
+        // Initialize the stack
+	// _stack_top is defined in loader_stage1_linker.ld
+	ldr sp, =_stack_top
+	
+        // Call stage2 of the loader (branch to 0x4000)
+        bx r4
+
+.align 4
+stage2_start:	
+	.incbin "loader_stage2.img"
+stage2_end:
diff --git a/loader_stage1.c b/loader_stage1.c
deleted file mode 100644
index aca439c..0000000
--- a/loader_stage1.c
+++ /dev/null
@@ -1,26 +0,0 @@
-#include <stddef.h>
-#include <stdint.h>
-#include <global.h>
-
-#define STAGE2_ADDR ((volatile char*) 0x4000)
-
-// There's one tricky thing about managing executable's own code.
-// Mainly, symbols are visible to c code as extern chars, but the actual
-// values are their adresses... see the code below
-extern char
-  __stage2_start,
-  __stage2_end,
-  __stage2_size;
-
-__attribute__((section(".text.stage1main")))
-void kernel_main(uint32_t r0, uint32_t r1, uint32_t atags)
-{
-  // stage2 of the bootloader is part of executable;
-  // copy it over to it's destination place
-  // TODO implement a memcpy() somewhere and use it instead of loops
-  for (size_t i = 0; i < (size_t) &__stage2_size; i++)
-    STAGE2_ADDR[i] = (&__stage2_start)[i];
-
-  // jump to stage2
-  ((void(*)(uint32_t, uint32_t, uint32_t))STAGE2_ADDR)(r0, r1, atags);
-}
diff --git a/loader_stage1_linker.ld b/loader_stage1_linker.ld
new file mode 100644
index 0000000..711fcbf
--- /dev/null
+++ b/loader_stage1_linker.ld
@@ -0,0 +1,16 @@
+ENTRY(_boot)
+ 
+SECTIONS
+{
+	/* see linker.ld for details */
+	. = 0x2000000;
+
+	__start = .;
+	loader_stage1 :
+	{
+		KEEP(loader_stage1.o)
+	}
+	__end = .;
+
+    _stack_top = 0x8000;
+}
diff --git a/loader_stage2.c b/loader_stage2.c
index 7b5087a..65dda0a 100644
--- a/loader_stage2.c
+++ b/loader_stage2.c
@@ -5,8 +5,7 @@
 
 void *const kernel_load_addr = ((void*) 0x8000);
 
-void __attribute__((section(".text.stage2main")))
-stage2(uint32_t r0, uint32_t r1, uint32_t atags)
+void _stage2_main(uint32_t r0, uint32_t r1, uint32_t atags)
 {
   // Declare as unused
   (void) r0;
diff --git a/loader_stage2_linker.ld b/loader_stage2_linker.ld
new file mode 100644
index 0000000..33e79e9
--- /dev/null
+++ b/loader_stage2_linker.ld
@@ -0,0 +1,16 @@
+ENTRY(_stage2_main)
+ 
+SECTIONS
+{
+	/* see loader_stage1.S for details */
+	. = 0x4000;
+
+	__start = .;
+	loader_stage2 :
+	{
+		KEEP(loader_stage2.o(.text))
+		loader_stage2.o
+		uart.o
+	}
+	__end = .;
+}