From c374eaacbc0d16983bbd1112fefa74dbe88e1e53 Mon Sep 17 00:00:00 2001
From: Wojtek Kosior <kwojtus@protonmail.com>
Date: Sat, 28 Dec 2019 15:55:29 +0100
Subject: use ldm instruction instead of libkernel for entering user mode; get
 rid of libkernel (no longer needed)

---
 memory.h | 13 ++-----------
 1 file changed, 2 insertions(+), 11 deletions(-)

(limited to 'memory.h')

diff --git a/memory.h b/memory.h
index 1c9ae35..d369694 100644
--- a/memory.h
+++ b/memory.h
@@ -24,24 +24,15 @@ extern const char __start;
 #define TRANSLATION_TABLE_END				\
   (TRANSLATION_TABLE_BASE + (uint32_t) (4096 * 4))
 
-#define LIBKERNEL_SECTION_START					\
-  (((TRANSLATION_TABLE_END - (uint32_t) 1) & ~((uint32_t) 0xfffff))	\
+#define PRIVILEGED_MEMORY_END					     \
+  (((TRANSLATION_TABLE_END - (uint32_t) 1) & ~((uint32_t) 0xfffff))  \
    + SECTION_SIZE)
 
-#define LIBKERNEL_SECTION_END			\
-  (LIBKERNEL_SECTION_START + SECTION_SIZE)
-
-// section for libkernel is flat-mapped
-#define LIBKERNEL_SECTION_NUMBER (LIBKERNEL_SECTION_START >> 20)
-
-#define PRIVILEGED_MEMORY_END LIBKERNEL_SECTION_END
-
 #define UNPRIVILEGED_MEMORY_START PRIVILEGED_MEMORY_END
 #define UNPRIVILEGED_MEMORY_END			\
   (UNPRIVILEGED_MEMORY_START + SECTION_SIZE)
 
 #define PL0_SECTION_NUMBER     ((uint32_t) 0b101010101010)
-#define PL0_SECTION_NUMBER_STR "0b101010101010"
 
 #define VIRTUAL_PL0_MEMORY_START (PL0_SECTION_NUMBER << 20)
 
-- 
cgit v1.2.3


From dea4cc9a00cd98f2749e9b658589a03addc6143b Mon Sep 17 00:00:00 2001
From: Wojtek Kosior <kwojtus@protonmail.com>
Date: Sat, 28 Dec 2019 16:46:08 +0100
Subject: clean-up memory.h

---
 memory.h | 22 ++++++++++++----------
 1 file changed, 12 insertions(+), 10 deletions(-)

(limited to 'memory.h')

diff --git a/memory.h b/memory.h
index d369694..e4493e2 100644
--- a/memory.h
+++ b/memory.h
@@ -1,9 +1,14 @@
 #ifndef MEMORY_H
 #define MEMORY_H
 
-#include "paging.h"
+#define POWER_OF_2(EXP) (((uint32_t) 1) << EXP)
 
-#define SECTION_SIZE (((uint32_t) 1) << 20)
+#define ALIGN_POWER_OF_2(ADDR, EXP)				\
+  (((ADDR - 1) & ~(POWER_OF_2(EXP) - 1)) + POWER_OF_2(EXP))
+
+#define SECTION_SIZE POWER_OF_2(20)
+
+#define ALIGN_SECTION(ADDR) ALIGN_POWER_OF_2(ADDR, 20)
 
 #define INTERRUPT_VECTOR_TABLE_START ((uint32_t) 0x0)
 
@@ -13,26 +18,23 @@
 extern const char __end;
 extern const char __start;
 
-#define KERNEL_START ((uint32_t) &__start)
+#define KERNEL_START ((uint32_t) &__start) // this is 0x8000
 #define KERNEL_END   ((uint32_t) &__end)
 
 // first 2^14 aligned address after the kernel
-#define TRANSLATION_TABLE_BASE						\
-  (((KERNEL_END - (uint32_t) 1) & ~((uint32_t) 0x3fff))			\
-   + (uint32_t) 0x4000)
+#define TRANSLATION_TABLE_BASE ALIGN_POWER_OF_2(KERNEL_END, 14)
 
 #define TRANSLATION_TABLE_END				\
   (TRANSLATION_TABLE_BASE + (uint32_t) (4096 * 4))
 
-#define PRIVILEGED_MEMORY_END					     \
-  (((TRANSLATION_TABLE_END - (uint32_t) 1) & ~((uint32_t) 0xfffff))  \
-   + SECTION_SIZE)
+#define PRIVILEGED_MEMORY_END ALIGN_SECTION(TRANSLATION_TABLE_END)
 
 #define UNPRIVILEGED_MEMORY_START PRIVILEGED_MEMORY_END
+
 #define UNPRIVILEGED_MEMORY_END			\
   (UNPRIVILEGED_MEMORY_START + SECTION_SIZE)
 
-#define PL0_SECTION_NUMBER     ((uint32_t) 0b101010101010)
+#define PL0_SECTION_NUMBER ((uint32_t) 0b101010101010)
 
 #define VIRTUAL_PL0_MEMORY_START (PL0_SECTION_NUMBER << 20)
 
-- 
cgit v1.2.3


From 700f4c412d42c9b9811269045c0e363a0331bba9 Mon Sep 17 00:00:00 2001
From: Wojtek Kosior <kwojtus@protonmail.com>
Date: Sat, 28 Dec 2019 21:54:42 +0100
Subject: split kernel into 2 stages; second stage gets copied to 0x0 and runs
 from there

---
 Makefile             | 29 ++++++++++++++++++-----
 boot.S               | 27 ---------------------
 demo_functionality.c | 26 ++------------------
 interrupt_vector.S   | 53 +++++++++++++++++++++++++++++------------
 interrupts.c         | 55 +++++++++++++++++++++++++++++++++++-------
 kernel.c             | 44 ----------------------------------
 kernel_stage1.S      | 67 ++++++++++++++++++++++++++++++++++++++++++++++++++++
 kernel_stage1.ld     | 27 +++++++++++++++++++++
 kernel_stage2.ld     | 52 ++++++++++++++++++++++++++++++++++++++++
 linker.ld            | 47 ------------------------------------
 memory.h             | 43 ++++++++++++++++++++++-----------
 setup.c              | 41 ++++++++++++++++++++++++++++++++
 12 files changed, 325 insertions(+), 186 deletions(-)
 delete mode 100644 boot.S
 delete mode 100644 kernel.c
 create mode 100644 kernel_stage1.S
 create mode 100644 kernel_stage1.ld
 create mode 100644 kernel_stage2.ld
 delete mode 100644 linker.ld
 create mode 100644 setup.c

(limited to 'memory.h')

diff --git a/Makefile b/Makefile
index 29efa86..68368a5 100644
--- a/Makefile
+++ b/Makefile
@@ -1,8 +1,10 @@
-CFLAGS=-mcpu=cortex-a7 -ffreestanding -std=gnu11 -Wall -Wextra -O2 -fPIC -I.
+CFLAGS=-mcpu=cortex-a7 -ffreestanding -std=gnu11 -Wall -Wextra -O2 -I.
 ELFFLAGS=-nostdlib -lgcc
 
 ARM_OBJECTS=kernel.o paging.o demo_functionality.o PL0_test.o uart.o loader_stage1.o loader_stage2.o
 
+KERNEL_STAGE2_OBJECTS=setup.o interrupt_vector.o interrupts.o uart.o demo_functionality.o paging.o ramfs_embeddable.o ramfs.o
+
 RAMFS_FILES=PL_0_test.img
 
 all : kernel.img
@@ -10,20 +12,32 @@ all : kernel.img
 %.o : %.c
 	arm-none-eabi-gcc $(CFLAGS) -c $^ -o $@
 
-%.o : %.S
-	arm-none-eabi-as -mcpu=cortex-a7 $^ -o $@
-
 %.img : %.elf
 	arm-none-eabi-objcopy $^ -O binary $@
 
+%.o : %.S
+	arm-none-eabi-as -mcpu=cortex-a7 $^ -o $@
+
 %_embeddable.o : %.img
 	arm-none-eabi-objcopy -I binary -O elf32-littlearm -B arm --rename-section .data=.rodata $^ $@
 
 PL_0_test.elf : PL0_test.o uart.o
 	arm-none-eabi-gcc -T PL0_test.ld -o $@ $(ELFFLAGS) $^
 
-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) $^
+kernel_stage1.o : kernel_stage1.S kernel_stage2.img
+	arm-none-eabi-as -mcpu=cortex-a7 $< -o $@
+
+kernel.elf : kernel_stage1.ld kernel_stage1.o
+	arm-none-eabi-gcc -T $< -o $@ $(ELFFLAGS) kernel_stage1.o
+
+kernel.img : kernel.elf
+	arm-none-eabi-objcopy $^ -O binary $@
+
+kernel_stage2.elf : kernel_stage2.ld $(KERNEL_STAGE2_OBJECTS)
+	arm-none-eabi-gcc -T $< -o $@ $(ELFFLAGS) $(KERNEL_STAGE2_OBJECTS)
+
+#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_stage2.elf : loader_stage2.o uart.o
 	arm-none-eabi-gcc -T loader_stage2_linker.ld -o $@ $(ELFFLAGS) $^
@@ -42,6 +56,9 @@ loader.img : loader.elf
 qemu-elf : kernel.elf
 	qemu-system-arm -m 256 -M raspi2 -serial stdio -kernel $^
 
+qemu-img : kernel.img
+	qemu-system-arm -m 256 -M raspi2 -serial stdio -kernel $^
+
 qemu-bin : loader.img kernel.img pipe_image
 	./pipe_image --stdout | qemu-system-arm -m 256 -M raspi2 -serial stdio -kernel $<
 
diff --git a/boot.S b/boot.S
deleted file mode 100644
index 593ed11..0000000
--- a/boot.S
+++ /dev/null
@@ -1,27 +0,0 @@
-// 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_main
-	
-.global _boot // make entry point label global
-_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:
-        // Initialize the stack (_stack_top is defined in linker.ld)
-        ldr sp, =_stack_top
- 
-        // Call kernel_main
-        ldr r3, =kernel_main
-        bx r3
diff --git a/demo_functionality.c b/demo_functionality.c
index 4b002d6..420639b 100644
--- a/demo_functionality.c
+++ b/demo_functionality.c
@@ -53,7 +53,7 @@ void demo_current_mode(void)
   uart_puts(mode_name);
 }
 
-#define TRANSLATION_TABLE						\
+#define TRANSLATION_TABLE					     \
   ((short_section_descriptor_t volatile*) TRANSLATION_TABLE_BASE)
 
 extern char
@@ -144,29 +144,7 @@ void demo_go_unprivileged(void)
   write_SPSR(new_SPSR);
 
   uart_puts("All ready, jumping to PL0 code\n\r");
-  
+
   asm volatile("ldm %0, {r0 - r15} ^" ::
 	       "r" (PL0_regs));
 }
-
-extern char
-  __interrupts_start,
-  __interrupts_end,
-  __interrupts_size;
-
-extern void (*volatile system_reentry_point)(void);
-
-void system_reentry(void)
-{
-  uart_puts("re-entered system");
-  while(1);
-}
-
-void demo_setup_interrupts(void)
-{
-  system_reentry_point = system_reentry;
-
-  for (size_t i = 0; i < (size_t) &__interrupts_size; i++)
-    ((volatile char*) 0)[i] =
-      (&__interrupts_start)[i];
-}
diff --git a/interrupt_vector.S b/interrupt_vector.S
index d20bf6d..6037b7c 100644
--- a/interrupt_vector.S
+++ b/interrupt_vector.S
@@ -1,22 +1,45 @@
-.section ".interrupts.vector"
-
-.global abort_handler
-.local generic_handler	
-.global _interrupt_vectors
 _interrupt_vectors:
-	b generic_handler
-	b generic_handler
-	b generic_handler
+	b reset_handler_caller
+	b undef_handler_caller
+	b svc_handler_caller
 	b abort_handler_caller
 	b abort_handler_caller
-	b generic_handler
-	b generic_handler
-
-.section ".interrupts.text"
+	b generic_handler_caller
+	b irq_handler_caller
+	b fiq_handler_caller
+	
+reset_handler_caller:
+	ldr sp, =_stack_top
+	ldr r5, =reset_handler
+	bx r5
+	
+undef_handler_caller:
+	ldr sp, =_stack_top
+	ldr r5, =undefined_instruction_vector
+	bx r5
 
-generic_handler:
-	b generic_handler
+svc_handler_caller:
+	ldr sp, =_stack_top
+	ldr r5, =supervisor_call_handler
+	bx r5
+	
 abort_handler_caller:
-	mov sp, #0x8000
+	ldr sp, =_stack_top
 	ldr r5, =abort_handler
 	bx r5
+
+generic_handler_caller:
+	ldr sp, =_stack_top
+	ldr r5, =generic_handler
+	bx r5
+
+irq_handler_caller:
+	ldr sp, =_stack_top
+	ldr r5, =irq_handler
+	bx r5
+
+fiq_handler_caller:
+	ldr sp, =_stack_top
+	ldr r5, =fiq_handler
+	bx r5
+	
diff --git a/interrupts.c b/interrupts.c
index 6952f89..1b0590a 100644
--- a/interrupts.c
+++ b/interrupts.c
@@ -1,10 +1,20 @@
 #include "uart.h"
-/**
-    @brief The undefined instruction interrupt handler
 
-    If an undefined instruction is encountered, the CPU will start
-    executing this function. Just trap here as a debug solution.
-*/
+void setup(void);
+
+void reset_handler(void)
+{
+  static _Bool setup_done;
+  
+  if (!setup_done)
+    setup();
+
+  setup_done = 1;
+
+  // TODO do something here
+  while(1);
+}
+
 void
 __attribute__((interrupt("UNDEF")))
 __attribute__((section(".interrupts.text")))
@@ -17,13 +27,40 @@ undefined_instruction_vector(void)
     }
 }
 
-void __attribute__((section(".interrupts.data")))
-(*system_reentry_point) (void);
+void supervisor_call_handler(void)
+{
+  uart_puts("something svc happened\n\r");
+
+  while(1);
+}
 
 void
 __attribute__((interrupt("ABORT")))
-__attribute__((section(".interrupts.text")))
 abort_handler(void)
 {
-  system_reentry_point();
+  uart_puts("re-entered system\n\r");
+
+  while(1);
+}
+
+void generic_handler(void)
+{
+  uart_puts("something weird happened\n\r");
+
+  while(1);
+}
+
+void irq_handler(void)
+{
+  uart_puts("irq happened\n\r");
+
+  while(1);
+}
+
+void fiq_handler(void)
+{
+  uart_puts("fiq happened\n\r");
+
+  while(1);
 }
+
diff --git a/kernel.c b/kernel.c
deleted file mode 100644
index 719ceff..0000000
--- a/kernel.c
+++ /dev/null
@@ -1,44 +0,0 @@
-#include "uart.h"
-#include "demo_functionality.h"
-#include "paging.h"
-
-void kernel_main(void)
-{
-  uart_init();
-
-  // When we attach screen session after loading kernel with socat
-  // we miss kernel's greeting... So we'll make the kernel wait for
-  // one char we're going to send from within screen
-  uart_getc();
-  
-  uart_puts("Hello, kernel World!\r\n");
-
-  // prints some info
-  demo_paging_support();
-
-  // prints some info
-  demo_current_mode();
-
-  // prints some info and sets upp translation table, turns on MMU
-  setup_flat_map();
-
-  demo_setup_PL0();
-
-  demo_setup_interrupts();
-  
-  // prints some info and sets up a section for PL0 code, loads a blob
-  // there and jumps to it... never, ever, ever returns
-  demo_go_unprivileged();
-  
-  while(1);
-  
-  while (1)
-    {
-      char c = uart_getc();
-
-      uart_putc(c);
-      
-      if (c == '\r')
-	uart_putc('\n');
-    }
-}
diff --git a/kernel_stage1.S b/kernel_stage1.S
new file mode 100644
index 0000000..1e0f614
--- /dev/null
+++ b/kernel_stage1.S
@@ -0,0 +1,67 @@
+/* arm mode, cortex-a7 compatibility
+ *
+ * _boot is entry point for the kernel.
+ *
+ * Kernel copies it's embedded stage 2 to address 0x0 and jumps to
+ * it (to the reset handler). Registers r0 - r2 are arguments for
+ * the kernel, but we're not using them for now.
+ *
+ * This file is based on (and almost identical with) loader_stage1.S
+ */
+	
+.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 kernel to address 0x0
+
+	// 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, #0
+
+	// load blob size to r5
+	// The size might get too big for an immediate value, so
+	// we load it from memory.
+	adr r5, blob_size
+	ldr r5, [r5]
+
+	// r6 is the counter - counts the bytes copied
+	mov r6, #0
+
+	// This initial piece of code might get overwritten when we
+	// copy stage2, so the actual copying loop shall be after
+	// stage2 blob. We want this asm code to be PIC, so we're
+	// computing address of stage2_end into r7.
+	add r7, r3, r5
+	bx r7
+	
+blob_size:
+	.word stage2_end - stage2_start
+
+.align 4
+stage2_start:
+	.incbin "kernel_stage2.img"
+stage2_end:
+
+	// 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
+	
+        // Call stage2 of the kernel (branch to 0x0,
+	// which is the reset handler).
+        bx r4
diff --git a/kernel_stage1.ld b/kernel_stage1.ld
new file mode 100644
index 0000000..3130634
--- /dev/null
+++ b/kernel_stage1.ld
@@ -0,0 +1,27 @@
+ENTRY(_boot) /* defined in boot.S; qemu needs it to run elf file */
+
+/* Code starts at  0x8000 - that's where RPis in 32-bit mode load
+ * kernel at. My experiments do, however, show, that qemu emulating
+ * RPi2 loads the kernel at 0x10000! (took some pain to find out).
+ * rpi-open-firmware, on the other hand, loads kernel at 0x2000000!
+ * This is not really a problem, since:
+ *   1. We can use our bootloader to load the kernel at 0x8000
+ *   2. We've rewritten stage 1 of both bootloader and kernel in
+ *      careful assembly, so that they should work regardless of
+ *      where they are loaded.
+ *   3. In qemu, we can load kernel.elf instead of raw binary
+ *      (qemu will do the right thing then)
+ */
+
+SECTIONS
+{
+
+	. = 0x8000;
+
+	__start = .;
+	.kernel_stage1 :
+	{
+		KEEP(kernel_stage1.o)
+	}
+	__end = .;
+}
diff --git a/kernel_stage2.ld b/kernel_stage2.ld
new file mode 100644
index 0000000..d3a23bf
--- /dev/null
+++ b/kernel_stage2.ld
@@ -0,0 +1,52 @@
+/* This sesond stage of the kernel is run from address 0x0 */
+
+TRANSLATION_TABLE_SIZE = 4096 * 4;
+MMU_SECTION_SIZE = 1 << 20;
+
+SECTIONS
+{
+
+	. = 0x0;
+
+	__start = .;
+	.kernel_stage2 :
+	{
+		KEEP(interrupt_vector.o)
+		. = ALIGN(4);
+		ramfs_embeddable.o
+		(*)
+	}
+	__end = .;
+
+	. = ALIGN(1 << 14);
+	
+	.translation_table (NOLOAD) :
+	{
+		_translation_table_start = .;
+		
+		. = . + TRANSLATION_TABLE_SIZE;
+
+		_translation_table_end = .;
+	}
+
+	. = ALIGN(1 << 20);
+	. = . + MMU_SECTION_SIZE;
+	
+	.stack (NOLOAD) :
+	{
+		_stack_start = .;
+		
+		. = . + MMU_SECTION_SIZE;
+
+		_stack_top = .;
+	}
+
+	.unprivileged_memory (NOLOAD) :
+	{
+		_unprivileged_memory_start = .;
+		
+		. = . + MMU_SECTION_SIZE;
+
+		_unprivileged_memory_end = .;
+	}
+}
diff --git a/linker.ld b/linker.ld
deleted file mode 100644
index 444bbf6..0000000
--- a/linker.ld
+++ /dev/null
@@ -1,47 +0,0 @@
-ENTRY(_boot) /* defined in boot.S; qemu needs it to run elf file */
- 
-SECTIONS
-{
-    /* Starts at  0x8000 - that's where RPis in 32-bit mode load */
-    /* kernel at. My experiments do, however, show, that qemu */
-    /* emulating RPi2 loads the kernel at 0x10000! (took some pain */
-    /* to find out). rpi-open-firmware, on the other hand, loads */
-    /* kernel at 0x2000000! */
-    /* This is not really a problem, since: */
-    /*   1. We can use our bootloader to load the kernel at 0x8000 */
-    /*   2. Stage 1 of the bootloader is written in careful */
-    /*      assembly, so that the loader itself should work */
-    /*      regardless of where it is loaded. */
-    /*   3. In qemu, we can load kernel.elf instead of raw binary */
-    /*      (qemu will do the right thing then) */
-
-    . = 0x8000;
-
-    /* RPi in 64-bit mode uses address 0x80000 instead */
-    
-    __start = .;
-    .kernel :
-    {
-        __kernel_start = .;
-	KEEP(boot.o)
-	. = ALIGN(4);
-	ramfs_embeddable.o
-	*(EXCLUDE_FILE (libkernel.o interrupt_vector.o interrupts.o) *)
-	__kernel_end = .;
-    }
-    __kernel_size = __kernel_end - __kernel_start;
-
-    .interrupts :
-    {
-	__interrupts_start = .;
-	KEEP(*(.interrupts.vector))
-	interrupt_vector.o
-	interrupts.o
-    	__interrupts_end = .;
-    }
-    __interrupts_size = __interrupts_end - __interrupts_start;
-    
-    __end = .;
-
-    _stack_top = __start;
-}
diff --git a/memory.h b/memory.h
index e4493e2..adc3bc0 100644
--- a/memory.h
+++ b/memory.h
@@ -1,7 +1,10 @@
 #ifndef MEMORY_H
 #define MEMORY_H
 
-#define POWER_OF_2(EXP) (((uint32_t) 1) << EXP)
+// These macros were heavily used b4 I moved all the address
+// computation to the linker script. Now I'm just keeping them
+// in case they're needed for something else :)
+#define POWER_OF_2(EXP) (((size_t) 1) << EXP)
 
 #define ALIGN_POWER_OF_2(ADDR, EXP)				\
   (((ADDR - 1) & ~(POWER_OF_2(EXP) - 1)) + POWER_OF_2(EXP))
@@ -10,33 +13,45 @@
 
 #define ALIGN_SECTION(ADDR) ALIGN_POWER_OF_2(ADDR, 20)
 
-#define INTERRUPT_VECTOR_TABLE_START ((uint32_t) 0x0)
 
-#define STACK_START ((uint32_t) 0x4000)
-#define STACK_END   ((uint32_t) 0x8000)
+// memory layout
+
+#define INTERRUPT_VECTOR_TABLE_START ((uint32_t) 0x0)
 
+// all those symbols are defined in the linker script
 extern const char __end;
 extern const char __start;
+extern const char _translation_table_start;
+extern const char _translation_table_end;
+extern const char _stack_start;
+extern const char _stack_top;
+extern const char _unprivileged_memory_start;
+extern const char _unprivileged_memory_end;
 
-#define KERNEL_START ((uint32_t) &__start) // this is 0x8000
-#define KERNEL_END   ((uint32_t) &__end)
+#define KERNEL_START ((size_t) &__start) // this is 0x0
+#define KERNEL_END   ((size_t) &__end)
 
 // first 2^14 aligned address after the kernel
-#define TRANSLATION_TABLE_BASE ALIGN_POWER_OF_2(KERNEL_END, 14)
-
-#define TRANSLATION_TABLE_END				\
-  (TRANSLATION_TABLE_BASE + (uint32_t) (4096 * 4))
+#define TRANSLATION_TABLE_BASE ((size_t) &_translation_table_start)
+#define TRANSLATION_TABLE_END  ((size_t) &_translation_table_end)
 
-#define PRIVILEGED_MEMORY_END ALIGN_SECTION(TRANSLATION_TABLE_END)
+// first section after the translation table is left unused;
+// the next section is used as the stack
+#define STACK_START ((size_t) &_stack_start)
+#define STACK_END   ((size_t) &_stack_top)
 
-#define UNPRIVILEGED_MEMORY_START PRIVILEGED_MEMORY_END
+#define PRIVILEGED_MEMORY_END STACK_END
 
+#define UNPRIVILEGED_MEMORY_START				\
+  ((size_t) &_unprivileged_memory_start)  // equal to STACK_END
 #define UNPRIVILEGED_MEMORY_END			\
-  (UNPRIVILEGED_MEMORY_START + SECTION_SIZE)
+  ((size_t) &_unprivileged_memory_end)
 
-#define PL0_SECTION_NUMBER ((uint32_t) 0b101010101010)
+#define PL0_SECTION_NUMBER ((size_t) 0xaaa)
 
 #define VIRTUAL_PL0_MEMORY_START (PL0_SECTION_NUMBER << 20)
+#define VIRTUAL_PL0_MEMORY_END			\
+  (VIRTUAL_PL0_MEMORY_START + SECTION_SIZE)
 
 #endif // MEMORY_H
 
diff --git a/setup.c b/setup.c
new file mode 100644
index 0000000..48df825
--- /dev/null
+++ b/setup.c
@@ -0,0 +1,41 @@
+#include "uart.h"
+#include "demo_functionality.h"
+#include "paging.h"
+
+void setup(void)
+{
+  uart_init();
+
+  // When we attach screen session after loading kernel with socat
+  // we miss kernel's greeting... So we'll make the kernel wait for
+  // one char we're going to send from within screen
+  uart_getc();
+  
+  uart_puts("Hello, kernel World!\r\n");
+
+  // prints some info
+  demo_paging_support();
+
+  // prints some info
+  demo_current_mode();
+
+  // prints some info and sets upp translation table, turns on MMU
+  setup_flat_map();
+
+  // prints some info and sets up a section for PL0 code,
+  // loads a blob there 
+  demo_setup_PL0();
+
+  // jumps to unprivileged code... never, ever, ever returns
+  demo_go_unprivileged();
+  
+  while (1)
+    {
+      char c = uart_getc();
+
+      uart_putc(c);
+      
+      if (c == '\r')
+	uart_putc('\n');
+    }
+}
-- 
cgit v1.2.3


From c9e045dc2170a99c9f32386e3e53aee9e01a8e7c Mon Sep 17 00:00:00 2001
From: Wojtek Kosior <kwojtus@protonmail.com>
Date: Mon, 30 Dec 2019 17:34:23 +0100
Subject: io api rework

---
 Makefile             | 10 ++++---
 PL0_test.c           |  6 ++---
 PL0_utils.c          | 14 +++-------
 PL0_utils.h          |  6 +----
 demo_functionality.c | 52 ++++++++++++++++++-------------------
 interrupts.c         | 20 +++++++-------
 io.c                 | 64 +++++++++++++++++++++++++++++++++++++++++++++
 io.h                 | 26 +++++++++++++++++++
 loader_stage2.c      | 11 ++++----
 memory.h             |  2 ++
 paging.c             | 27 ++++++++-----------
 setup.c              | 13 +++++-----
 strings.c            | 73 ++++++++++++++++++++++++++++++++++++++++++++++++++++
 strings.h            | 22 +++++++++++-----
 uart.c               | 10 ++-----
 uart.h               |  5 ++--
 16 files changed, 258 insertions(+), 103 deletions(-)
 create mode 100644 io.c
 create mode 100644 io.h
 create mode 100644 strings.c

(limited to 'memory.h')

diff --git a/Makefile b/Makefile
index c3fc2f1..1e8f8fe 100644
--- a/Makefile
+++ b/Makefile
@@ -3,9 +3,11 @@ ELFFLAGS=-nostdlib -lgcc
 
 ARM_OBJECTS=kernel.o paging.o demo_functionality.o PL0_test.o uart.o loader_stage1.o loader_stage2.o
 
-KERNEL_STAGE2_OBJECTS=setup.o interrupt_vector.o interrupts.o uart.o demo_functionality.o paging.o ramfs_embeddable.o ramfs.o
+KERNEL_STAGE2_OBJECTS=setup.o interrupt_vector.o interrupts.o uart.o demo_functionality.o paging.o ramfs_embeddable.o ramfs.o strings.o io.o
 
-PL_0_TEST_OBJECTS=PL0_utils.o svc.o PL0_test.o
+PL_0_TEST_OBJECTS=PL0_utils.o svc.o PL0_test.o strings.o io.o
+
+LOADER_STAGE2_OBJECTS=uart.o strings.o io.o loader_stage2.o
 
 RAMFS_FILES=PL_0_test.img
 
@@ -41,8 +43,8 @@ kernel_stage2.elf : kernel_stage2.ld $(KERNEL_STAGE2_OBJECTS)
 #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_stage2.elf : loader_stage2.o uart.o
-	arm-none-eabi-gcc -T loader_stage2_linker.ld -o $@ $(ELFFLAGS) $^
+loader_stage2.elf : loader_stage2_linker.ld $(LOADER_STAGE2_OBJECTS)
+	arm-none-eabi-gcc -T $< -o $@ $(ELFFLAGS) $(LOADER_STAGE2_OBJECTS)
 
 loader_stage1.o : loader_stage1.S  loader_stage2.img
 	arm-none-eabi-as -mcpu=cortex-a7 $< -o $@
diff --git a/PL0_test.c b/PL0_test.c
index f476eb7..ac74ad4 100644
--- a/PL0_test.c
+++ b/PL0_test.c
@@ -15,15 +15,15 @@ void PL0_main(void)
   first_kernel_byte[0] = *(char*) 0x0;
   first_kernel_byte[1] = '\0';
 
-  uart_puts(first_kernel_byte);
+  puts(first_kernel_byte);
     
   while (1)
     {
       char c = getchar();
-
-      putchar(c);
       
       if (c == '\r')
 	putchar('\n');
+
+      putchar(c);
     }
 }
diff --git a/PL0_utils.c b/PL0_utils.c
index 2cede90..d83edb9 100644
--- a/PL0_utils.c
+++ b/PL0_utils.c
@@ -2,26 +2,18 @@
 #include <stdint.h>
 
 #include "svc_interface.h"
+#include "PL0_utils.h"
 
 // most generic definition possible
 // the actual function defined in svc.S
 uint32_t svc(enum svc_type, ...);
 
-void putchar(int character)
+void putchar(char character)
 {
   svc(UART_PUTCHAR, character);
 }
 
-int getchar(void)
+char getchar(void)
 {
   return svc(UART_GETCHAR);
 }
-
-void puts(char *string)
-{
-  for (size_t i = 0; string[i]; i++)
-    putchar(string[i]);
-
-  putchar('\n');
-  putchar('\r');
-}
diff --git a/PL0_utils.h b/PL0_utils.h
index ba72fd3..c26a100 100644
--- a/PL0_utils.h
+++ b/PL0_utils.h
@@ -1,10 +1,6 @@
 #ifndef PL0_UTILS_H
 #define PL0_UTILS_H
 
-void putchar(int character);
-
-int getchar(void);
-
-void puts(char *string);
+#include "io.h"
 
 #endif // PL0_UTILS_H
diff --git a/demo_functionality.c b/demo_functionality.c
index d8fbdb5..2db40f8 100644
--- a/demo_functionality.c
+++ b/demo_functionality.c
@@ -1,4 +1,4 @@
-#include "uart.h"
+#include "io.h"
 #include "psr.h"
 #include "memory.h"
 #include "translation_table_descriptors.h"
@@ -14,16 +14,16 @@ void demo_paging_support(void)
   
   switch(ID_MMFR0 & 0xf) /* lowest 4 bits indicate VMSA support */
     {
-    case 0 : paging = "no paging\n\r"; break;
-    case 1 : paging = "implementation defined paging\n\r"; break;
-    case 2 : paging = "VMSAv6, with cache and TLB type registers\n\r"; break;
-    case 3 : paging = "VMSAv7, with support for remapping and access flag\n\r"; break;
-    case 4 : paging = "VMSAv7 with PXN bit supported\n\r"; break;
-    case 5 : paging = "VMSAv7, PXN and long format descriptors. EPAE is supported.\n\r"; break;
-    default : paging = "?_? unknown paging ?_?\n\r";
+    case 0 : paging = "no paging"; break;
+    case 1 : paging = "implementation defined paging"; break;
+    case 2 : paging = "VMSAv6, with cache and TLB type registers"; break;
+    case 3 : paging = "VMSAv7, with support for remapping and access flag"; break;
+    case 4 : paging = "VMSAv7 with PXN bit supported"; break;
+    case 5 : paging = "VMSAv7, PXN and long format descriptors. EPAE is supported."; break;
+    default : paging = "?_? unknown paging ?_?";
     }
   
-  uart_puts(paging);
+  puts(paging);
 }
 
 void demo_current_mode(void)
@@ -37,20 +37,20 @@ void demo_current_mode(void)
   
   switch(CPSR.fields.PSR_MODE_4_0)
     {
-    case MODE_USER       : mode_name = "User (PL0)\r\n"; break;
-    case MODE_FIQ        : mode_name = "FIQ (PL1)\r\n"; break;
-    case MODE_IRQ        : mode_name = "IRQ (PL1)\r\n"; break;
-    case MODE_SUPERVISOR : mode_name = "Supervisor (PL1)\r\n"; break;
-    case MODE_MONITOR    : mode_name = "Monitor (PL1)\r\n"; break;
-    case MODE_ABORT      : mode_name = "Abort (PL1)\r\n"; break;
-    case MODE_HYPERVISOR : mode_name = "Hyp (PL2)\r\n"; break;
-    case MODE_UNDEFINED  : mode_name = "Undefined (PL1)\r\n"; break;
-    case MODE_SYSTEM     : mode_name = "System (PL1)\r\n"; break;
-    default : mode_name = "Unknown mode\r\n"; break;
+    case MODE_USER       : mode_name = "User (PL0)"; break;
+    case MODE_FIQ        : mode_name = "FIQ (PL1)"; break;
+    case MODE_IRQ        : mode_name = "IRQ (PL1)"; break;
+    case MODE_SUPERVISOR : mode_name = "Supervisor (PL1)"; break;
+    case MODE_MONITOR    : mode_name = "Monitor (PL1)"; break;
+    case MODE_ABORT      : mode_name = "Abort (PL1)"; break;
+    case MODE_HYPERVISOR : mode_name = "Hyp (PL2)"; break;
+    case MODE_UNDEFINED  : mode_name = "Undefined (PL1)"; break;
+    case MODE_SYSTEM     : mode_name = "System (PL1)"; break;
+    default : mode_name = "Unknown mode"; break;
     }
 
-  uart_puts("current mode: ");
-  uart_puts(mode_name);
+  prints("current mode: ");
+  puts(mode_name);
 }
 
 #define TRANSLATION_TABLE					     \
@@ -69,7 +69,7 @@ void demo_setup_PL0(void)
   if (find_file(&_binary_ramfs_img_start, "PL_0_test.img",
 		&PL_0_test_img))
     {
-      uart_puts("PL_0_test.img not found :(\r\n");
+      puts("PL_0_test.img not found :(");
       asm volatile ("wfi");
     }
 
@@ -96,20 +96,20 @@ void demo_setup_PL0(void)
   
   // check that translation works... by copying a string using one
   // mapping and reading it using other :D
-  char message[] = "mapped sections for PL0 code\n\r";
+  char message[] = "mapped sections for PL0 code";
   
   unsigned int i;
   for (i = 0; i < sizeof(message); i++)
     ((volatile char*) UNPRIVILEGED_MEMORY_START)[i] = message[i];
   
-  uart_puts((char*) VIRTUAL_PL0_MEMORY_START);
+  puts((char*) VIRTUAL_PL0_MEMORY_START);
 
   // now paste a userspace program to that section
   for (uint32_t i = 0; i < PL_0_test_img.file_size; i++)
     ((volatile char*) VIRTUAL_PL0_MEMORY_START)[i] =
       PL_0_test_img.file_contents[i];
 
-  uart_puts("copied PL0 code to it's section\n\r");
+  puts("copied PL0 code to it's section");
 }
 
 // needed for array initialization in demo_go_unprivileged()
@@ -132,7 +132,7 @@ void demo_go_unprivileged(void)
   new_SPSR.fields.PSR_MODE_4_0 = MODE_USER;
   write_SPSR(new_SPSR);
 
-  uart_puts("All ready, jumping to PL0 code\n\r");
+  puts("All ready, jumping to PL0 code");
 
   asm volatile("cps %[sysmode]\n\r"
 	       "mov sp, %[stackaddr]\n\r"
diff --git a/interrupts.c b/interrupts.c
index f47bc1d..3102761 100644
--- a/interrupts.c
+++ b/interrupts.c
@@ -1,4 +1,4 @@
-#include "uart.h"
+#include "io.h"
 #include "svc_interface.h"
 
 void setup(void);
@@ -18,7 +18,7 @@ void reset_handler(void)
 
 void undefined_instruction_vector(void)
 {
-    uart_puts("Undefined instruction occured");
+    puts("Undefined instruction occured");
     while( 1 )
     {
         /* Do Nothing! */
@@ -32,16 +32,16 @@ uint32_t supervisor_call_handler(enum svc_type request, uint32_t arg1,
   
   switch(request) {
   case UART_PUTCHAR:
-    uart_putc(arg1);
+    putchar(arg1);
     break;
   case UART_GETCHAR:
-    return uart_getc();
+    return getchar();
   case UART_WRITE:
-    uart_puts("UART_WRITE not implemented!!!!!\n\r");
+    puts("UART_WRITE not implemented!!!!!");
     break;
   default:
     // perhaps we should kill the process now?
-    uart_puts("unknown supervisor call type!!!!!\n\r");
+    puts("unknown supervisor call type!!!!!");
   }
 
   return 0; // a dummy value
@@ -49,28 +49,28 @@ uint32_t supervisor_call_handler(enum svc_type request, uint32_t arg1,
 
 void abort_handler(void)
 {
-  uart_puts("re-entered system\n\r");
+  puts("re-entered system");
 
   while(1);
 }
 
 void generic_handler(void)
 {
-  uart_puts("something weird happened\n\r");
+  puts("something weird happened");
 
   while(1);
 }
 
 void irq_handler(void)
 {
-  uart_puts("irq happened\n\r");
+  puts("irq happened");
 
   while(1);
 }
 
 void fiq_handler(void)
 {
-  uart_puts("fiq happened\n\r");
+  puts("fiq happened");
 
   while(1);
 }
diff --git a/io.c b/io.c
new file mode 100644
index 0000000..f35fda9
--- /dev/null
+++ b/io.c
@@ -0,0 +1,64 @@
+#include <stddef.h>
+
+#include "io.h"
+#include "strings.h"
+
+void puts(char string[])
+{
+  prints(string);
+
+  putchar('\n');
+  putchar('\r');
+}
+
+void prints(char string[])
+{
+  for (size_t i = 0; string[i]; i++)
+    putchar(string[i]);
+}
+
+void printdec(uint32_t number)
+{
+  char buf[11];
+
+  uint32_to_decstring(number, buf);
+
+  prints(buf);
+}
+
+void printhex(uint32_t number)
+{
+  char buf[9];
+
+  uint32_to_hexstring(number, buf);
+
+  prints(buf);
+}
+
+void printbin(uint32_t number)
+{
+  char buf[33];
+
+  uint32_to_binstring(number, buf);
+
+  prints(buf);
+}
+
+void printdect(uint32_t number)
+{
+  char buf[11];
+
+  uint32_to_decstringt(number, buf);
+
+  prints(buf);
+}
+
+void printhext(uint32_t number)
+{
+  char buf[9];
+
+  uint32_to_hexstringt(number, buf);
+
+  prints(buf);
+}
+
diff --git a/io.h b/io.h
new file mode 100644
index 0000000..161728a
--- /dev/null
+++ b/io.h
@@ -0,0 +1,26 @@
+#ifndef IO_H
+#define IO_H
+
+#include <stdint.h>
+
+// putchar() and getchar() are not part of io.c, but it's useful to
+// have those symbols declared here
+void putchar(char c);
+
+char getchar(void);
+
+void puts(char string[]);
+
+void prints(char string[]);
+
+void printdec(uint32_t number);
+
+void printhex(uint32_t number);
+
+void printbin(uint32_t number);
+
+void printdect(uint32_t number);
+
+void printhext(uint32_t number);
+
+#endif // IO_H
diff --git a/loader_stage2.c b/loader_stage2.c
index f49d94c..e05de51 100644
--- a/loader_stage2.c
+++ b/loader_stage2.c
@@ -1,6 +1,7 @@
 #include <stddef.h>
 #include <stdint.h>
 #include <uart.h>
+#include <io.h>
 #include <global.h>
 
 void *const kernel_load_addr = ((void*) 0x8000);
@@ -17,10 +18,10 @@ void _stage2_main(uint32_t r0, uint32_t r1, uint32_t atags)
   // get kernel size via uart (little endian)
   uint32_t b0, b1, b2, b3;
   
-  b0 = uart_getc();
-  b1 = uart_getc();
-  b2 = uart_getc();
-  b3 = uart_getc();
+  b0 = getchar();
+  b1 = getchar();
+  b2 = getchar();
+  b3 = getchar();
   
   uint32_t kernel_size = b0 | (b1 << 8) | (b2 << 16) | (b3 << 24);
 
@@ -28,7 +29,7 @@ void _stage2_main(uint32_t r0, uint32_t r1, uint32_t atags)
   char *dst = kernel_load_addr, *end = dst + kernel_size;
 
   while (dst < end)
-    *(dst++) = uart_getc();
+    *(dst++) = getchar();
 
   // jump to kernel
   ((void(*)(uint32_t, uint32_t, uint32_t)) kernel_load_addr)
diff --git a/memory.h b/memory.h
index adc3bc0..f6ece42 100644
--- a/memory.h
+++ b/memory.h
@@ -1,6 +1,8 @@
 #ifndef MEMORY_H
 #define MEMORY_H
 
+#include <stddef.h>
+
 // These macros were heavily used b4 I moved all the address
 // computation to the linker script. Now I'm just keeping them
 // in case they're needed for something else :)
diff --git a/paging.c b/paging.c
index 7c2a9de..2985e7e 100644
--- a/paging.c
+++ b/paging.c
@@ -1,24 +1,21 @@
 #include "cp_regs.h"
-#include "uart.h"
 #include "strings.h"
 #include "memory.h"
 #include "translation_table_descriptors.h"
+#include "io.h"
 
 void setup_flat_map(void)
 {
-  char bits[33]; // for printing uint32_t bit values
-  
   // compute translation table base address
   // translation table shall start at first 2^14-bytes aligned
   // address after the kernel image
 
-  uint32_to_bits(TRANSLATION_TABLE_BASE, bits);
-  uart_puts("binary representation of chosen"
-	    " lvl1 translation table address: ");
-  uart_puts(bits); uart_puts("\n\r");
+  prints("chosen lvl1 translation table address: 0x");
+  printhex(TRANSLATION_TABLE_BASE);
+  puts("");
  
   // flat map all memory
-  uart_puts("preparing translation table\n\r");
+  puts("preparing translation table");
   short_descriptor_lvl1_t volatile *translation_table =
     (short_descriptor_lvl1_t*) TRANSLATION_TABLE_BASE;
   
@@ -35,8 +32,7 @@ void setup_flat_map(void)
       };
 
   // meddle with domain settings
-  uart_puts("setting domain0 to client access"
-	    " and blocking other domains\n\r");
+  puts("setting domain0 to client access and blocking other domains");
 
   DACR_t DACR = 0;
   DACR = set_domain_permissions(DACR, 0, DOMAIN_CLIENT_ACCESS);
@@ -55,7 +51,7 @@ void setup_flat_map(void)
   // we also disable data and instruction caches and the MMU
   
   // some of this is redundant (i.e. MMU should already be disabled)
-  uart_puts("setting C, I, AFE and TRE to 0 in SCTLR\n\r");
+  puts("setting C, I, AFE and TRE to 0 in SCTLR");
 
   SCTLR_t SCTLR;
   asm("mrc p15, 0, %0, c1, c0, 0" : "=r" (SCTLR.raw));
@@ -70,8 +66,8 @@ void setup_flat_map(void)
 
   // TODO: move invalidation instructions to some header as inlines
   
-  uart_puts("invalidating instruction cache, branch prediction,"
-	    " and entire main TLB\n\r");
+  puts("invalidating instruction cache, branch prediction,"
+       " and entire main TLB");
   
   // invalidate instruction cache
   asm("mcr p15, 0, r0, c7, c5, 0\n\r" // r0 gets ignored
@@ -86,8 +82,7 @@ void setup_flat_map(void)
       "isb" :: "r" (0) : "memory");
 
   // now set TTBCR to use TTBR0 exclusively
-  uart_puts("Setting TTBCR.N to 0, so that"
-	    " TTBR0 is used everywhere\n\r");
+  puts("Setting TTBCR.N to 0, so that TTBR0 is used everywhere");
   
   uint32_t TTBCR = 0;
   asm("mcr p15, 0, %0, c2, c0, 2" :: "r" (TTBCR));
@@ -102,7 +97,7 @@ void setup_flat_map(void)
   asm("mcr p15, 0, %0, c2, c0, 0" :: "r" (TTBR0.raw));
 
   // enable MMU
-  uart_puts("enabling the MMU\n\r");
+  puts("enabling the MMU");
 
   // redundant - we already have SCTLR contents in the variable
   // asm("mrc p15, 0, %0, c1, c0, 0" : "=r" (SCTLR.raw));
diff --git a/setup.c b/setup.c
index 48df825..3542043 100644
--- a/setup.c
+++ b/setup.c
@@ -1,4 +1,5 @@
 #include "uart.h"
+#include "io.h"
 #include "demo_functionality.h"
 #include "paging.h"
 
@@ -9,9 +10,9 @@ void setup(void)
   // When we attach screen session after loading kernel with socat
   // we miss kernel's greeting... So we'll make the kernel wait for
   // one char we're going to send from within screen
-  uart_getc();
+  getchar();
   
-  uart_puts("Hello, kernel World!\r\n");
+  puts("Hello, kernel World!");
 
   // prints some info
   demo_paging_support();
@@ -31,11 +32,11 @@ void setup(void)
   
   while (1)
     {
-      char c = uart_getc();
+      char c = getchar();
 
-      uart_putc(c);
-      
       if (c == '\r')
-	uart_putc('\n');
+	putchar('\n');
+
+      putchar(c);
     }
 }
diff --git a/strings.c b/strings.c
new file mode 100644
index 0000000..55aff97
--- /dev/null
+++ b/strings.c
@@ -0,0 +1,73 @@
+#include <stddef.h>
+
+#include "strings.h"
+
+void uint32_to_dec(uint32_t number, char buf[10])
+{
+  for (int i = 0; i < 10; i++)
+    {
+      buf[10 - 1 - i] = number % 10;
+      number /= 10;
+    }
+}
+
+void uint32_to_hex(uint32_t number, char buf[8])
+{
+  for (int i = 0; i < 8; i++)
+    {
+      unsigned char quadbit = (number >> ((8 - i - 1) * 4)) & 0xf;
+      buf[i] =  quadbit > 9 ? quadbit - 10 + 'a' : quadbit + '0';
+    }
+}
+
+void uint32_to_bin(uint32_t number, char buf[32])
+{
+  for (int i = 0; i < 32; i++)
+    buf[i] = ((number >> (32 - i - 1)) & 1) ? '1' : '0';
+}
+
+void uint32_to_decstring(uint32_t number, char buf[11])
+{
+  uint32_to_dec(number, buf);
+  buf[10] = '\0';
+}
+
+void uint32_to_hexstring(uint32_t number, char buf[9])
+{
+  uint32_to_hex(number, buf);
+  buf[9] = '\0';
+}
+
+void uint32_to_binstring(uint32_t number, char buf[33])
+{
+  uint32_to_bin(number, buf);
+  buf[32] = '\0';
+}
+
+void trim_0s(char string[])
+{
+  size_t i;
+  for (i = 0; string[i] != '\0' && string[i] != '0'; i++);
+
+  size_t j = 0;
+  
+  if (string[i] == '\0')
+    string[j++] = string[i--];
+
+  do
+    string[j] = string[i + j];
+  while (string[j++]);
+}
+
+void uint32_to_decstringt(uint32_t number, char buf[11])
+{
+  uint32_to_decstring(number, buf);
+  trim_0s(buf);
+}
+
+void uint32_to_hexstringt(uint32_t number, char buf[9])
+{
+  uint32_to_hexstring(number, buf);
+  trim_0s(buf);
+}
+
diff --git a/strings.h b/strings.h
index 8d4b511..bfe6fd0 100644
--- a/strings.h
+++ b/strings.h
@@ -3,12 +3,22 @@
 
 #include <stdint.h>
 
-static inline void uint32_to_bits(uint32_t number, char *buf)
-{
-  for (int i = 0; i < 32; i++)
-    buf[i] = ((number >> (32 - i - 1)) & 1) ? '1' : '0';
+void uint32_to_dec(uint32_t number, char buf[10]);
 
-  buf[32] = '\0';
-}
+void uint32_to_hex(uint32_t number, char buf[8]);
+
+void uint32_to_bin(uint32_t number, char buf[32]);
+
+void uint32_to_decstring(uint32_t number, char buf[11]);
+
+void uint32_to_hexstring(uint32_t number, char buf[9]);
+
+void uint32_to_binstring(uint32_t number, char buf[33]);
+
+void trim_0s(char string[]);
+
+void uint32_to_decstringt(uint32_t number, char buf[11]);
+
+void uint32_to_hexstringt(uint32_t number, char buf[9]);
 
 #endif // STRINGS_H
diff --git a/uart.c b/uart.c
index 139cfe0..2030538 100644
--- a/uart.c
+++ b/uart.c
@@ -63,22 +63,16 @@ void uart_init()
   mmio_write(UART0_CR, (1 << 0) | (1 << 8) | (1 << 9));
 }
 
-void uart_putc(unsigned char c)
+void putchar(char c)
 {
   // Wait for UART to become ready to transmit.
   while ( mmio_read(UART0_FR) & (1 << 5) ) { }
   mmio_write(UART0_DR, c);
 }
 
-unsigned char uart_getc()
+char getchar(void)
 {
   // Wait for UART to have received something.
   while ( mmio_read(UART0_FR) & (1 << 4) ) { }
   return mmio_read(UART0_DR);
 }
-
-void uart_puts(const char* str)
-{
-  for (size_t i = 0; str[i] != '\0'; i ++)
-    uart_putc((unsigned char)str[i]);
-}
diff --git a/uart.h b/uart.h
index ec29d47..d5e931b 100644
--- a/uart.h
+++ b/uart.h
@@ -39,8 +39,7 @@ enum
   };
 
 void uart_init();
-void uart_putc(unsigned char c);
-unsigned char uart_getc();
-void uart_puts(const char* str);
+void putchar(char c);
+char getchar(void);
 
 #endif // UART_H
-- 
cgit v1.2.3


From 3387b288a53422131f22f1fc96168a128edc9902 Mon Sep 17 00:00:00 2001
From: Wojtek Kosior <kwojtus@protonmail.com>
Date: Tue, 31 Dec 2019 12:04:14 +0100
Subject: remove misleading const

---
 memory.h | 16 ++++++++--------
 1 file changed, 8 insertions(+), 8 deletions(-)

(limited to 'memory.h')

diff --git a/memory.h b/memory.h
index f6ece42..7de556d 100644
--- a/memory.h
+++ b/memory.h
@@ -21,14 +21,14 @@
 #define INTERRUPT_VECTOR_TABLE_START ((uint32_t) 0x0)
 
 // all those symbols are defined in the linker script
-extern const char __end;
-extern const char __start;
-extern const char _translation_table_start;
-extern const char _translation_table_end;
-extern const char _stack_start;
-extern const char _stack_top;
-extern const char _unprivileged_memory_start;
-extern const char _unprivileged_memory_end;
+extern char __end;
+extern char __start;
+extern char _translation_table_start;
+extern char _translation_table_end;
+extern char _stack_start;
+extern char _stack_top;
+extern char _unprivileged_memory_start;
+extern char _unprivileged_memory_end;
 
 #define KERNEL_START ((size_t) &__start) // this is 0x0
 #define KERNEL_END   ((size_t) &__end)
-- 
cgit v1.2.3


From 7dcea5fdafe66d8bcf1eeacbaf3f3f3b1c258dfc Mon Sep 17 00:00:00 2001
From: Wojtek Kosior <kwojtus@protonmail.com>
Date: Tue, 31 Dec 2019 13:50:14 +0100
Subject: implement basic memory section allocation for processes

---
 kernel_stage2.ld |  12 ++++-
 memory.h         |   6 +++
 paging.c         | 132 +++++++++++++++++++++++++++++++++++++++++++++++++++++++
 paging.h         |   8 ++++
 4 files changed, 157 insertions(+), 1 deletion(-)

(limited to 'memory.h')

diff --git a/kernel_stage2.ld b/kernel_stage2.ld
index dce2d1f..a30105e 100644
--- a/kernel_stage2.ld
+++ b/kernel_stage2.ld
@@ -1,6 +1,7 @@
 /* This sesond stage of the kernel is run from address 0x0 */
 
 TRANSLATION_TABLE_SIZE = 4096 * 4;
+SECTIONS_LIST_SIZE = 4096 * 8;
 MMU_SECTION_SIZE = 1 << 20;
 
 SECTIONS
@@ -39,9 +40,18 @@ SECTIONS
 		_translation_table_end = .;
 	}
 
+	.sections_list (NOLOAD) :
+	{
+		_sections_list_start = .;
+		
+		. = . + SECTIONS_LIST_SIZE;
+
+		_sections_list_end = .;
+	}
+
 	. = ALIGN(1 << 20);
 	. = . + MMU_SECTION_SIZE;
-	
+
 	.stack (NOLOAD) :
 	{
 		_stack_start = .;
diff --git a/memory.h b/memory.h
index 7de556d..2ab5485 100644
--- a/memory.h
+++ b/memory.h
@@ -25,6 +25,8 @@ extern char __end;
 extern char __start;
 extern char _translation_table_start;
 extern char _translation_table_end;
+extern char _sections_list_start;
+extern char _sections_list_end;
 extern char _stack_start;
 extern char _stack_top;
 extern char _unprivileged_memory_start;
@@ -37,6 +39,10 @@ extern char _unprivileged_memory_end;
 #define TRANSLATION_TABLE_BASE ((size_t) &_translation_table_start)
 #define TRANSLATION_TABLE_END  ((size_t) &_translation_table_end)
 
+// another 32KB after the translation table are used for sections list
+#define SECTIONS_LIST_START ((size_t) &_sections_list_start)
+#define SECTIONS_LIST_END  ((size_t) &_sections_list_end)
+
 // first section after the translation table is left unused;
 // the next section is used as the stack
 #define STACK_START ((size_t) &_stack_start)
diff --git a/paging.c b/paging.c
index 2985e7e..0de63b8 100644
--- a/paging.c
+++ b/paging.c
@@ -4,6 +4,8 @@
 #include "translation_table_descriptors.h"
 #include "io.h"
 
+#include "paging.h"
+
 void setup_flat_map(void)
 {
   // compute translation table base address
@@ -107,3 +109,133 @@ void setup_flat_map(void)
   asm("mcr p15, 0, %0, c1, c0, 0\n\r"
       "isb" :: "r" (SCTLR.raw) : "memory");
 }
+
+#define OWNER_FREE ((void*) 0)
+#define OWNER_KERNEL ((void*) 1)
+#define OWNER_SPLIT ((void*) 2)
+
+// we want to maintain a list of free and used physical sections
+struct section_node
+{
+  // we're going to add processes, process management and
+  // struct process. Then, owner will be struct process*.
+  void *owner; // 0 if free, 1 if used by kernel, 2 if split to pages
+
+  // it's actually a 2-directional lists;
+  // end of list is marked by reference to itself;
+  // we use offsets into sections_list array instead of pointers;
+  uint16_t prev, next;
+};
+
+static struct section_node volatile *sections_list;
+
+static uint16_t
+  all_sections_count, kernel_sections_count,
+  split_sections_count, free_sections_count;
+
+// those are undefined when the corresponding count is 0;
+static uint16_t
+  first_free_section, first_kernel_section, first_split_section;
+
+void setup_pager_structures(uint32_t available_mem)
+{
+  all_sections_count = available_mem / SECTION_SIZE;
+  kernel_sections_count = PRIVILEGED_MEMORY_END / SECTION_SIZE;
+  free_sections_count = all_sections_count - kernel_sections_count;
+  split_sections_count = 0;
+  
+  sections_list = (struct section_node*) SECTIONS_LIST_START;
+
+  for (uint16_t i = 0; i < kernel_sections_count; i++)
+    sections_list[i] = (struct section_node) {
+      .owner = OWNER_KERNEL,
+      .prev = i == 0 ? i : i - 1,
+      .next = i == kernel_sections_count - 1 ? i : i + 1
+    };
+
+  first_kernel_section = 0;
+  
+  for (uint16_t i = kernel_sections_count;
+       i < all_sections_count; i++)
+    sections_list[i] = (struct section_node) {
+      .owner = OWNER_FREE,
+      .prev = i == kernel_sections_count ? i : i - 1,
+      .next = i == all_sections_count - 1 ? i : i + 1
+    };
+
+  first_free_section = kernel_sections_count;
+}
+
+// return section number or CLAIM_FAILURE in case of failure
+static uint16_t claim_section(void *owner)
+{
+  if (!free_sections_count)
+    return CLAIM_FAILURE; // failure
+
+  uint16_t section = first_free_section;
+  
+  if (--free_sections_count)
+    {
+      uint16_t next;
+      
+      next = sections_list[section].next;
+      sections_list[next].prev = next;
+
+      first_free_section = next;
+    }
+
+  if (owner == OWNER_KERNEL)
+    {
+      sections_list[first_kernel_section].prev = section;
+
+      sections_list[section] = (struct section_node) {
+	.owner = owner,
+	.prev = section,
+	.next = first_kernel_section
+      };
+
+      kernel_sections_count++;
+
+      first_kernel_section = section;
+    }
+  else
+    sections_list[section] = (struct section_node) {
+      .owner = owner,
+      .prev = section,
+      .next = section
+    };
+
+  return section;
+}
+
+// return values like claim_section()
+uint16_t claim_and_map_section
+(void *owner, uint16_t where_to_map, uint8_t access_permissions)
+{
+  uint16_t section = claim_section(owner);
+  
+  if (section == CLAIM_FAILURE)
+    return section;
+  
+  short_section_descriptor_t volatile *section_entry =
+    &((short_section_descriptor_t*)
+      TRANSLATION_TABLE_BASE)[where_to_map];
+
+  short_section_descriptor_t descriptor = *section_entry;
+
+  // set up address of section
+  descriptor.SECTION_BASE_ADDRESS_31_20 = section;
+
+  // set requested permissions on section
+  descriptor.ACCESS_PERMISSIONS_2 = access_permissions >> 2;
+  descriptor.ACCESS_PERMISSIONS_1_0 = access_permissions & 0b011;
+
+  // write modified descriptor to the table
+  *section_entry = descriptor;
+  
+  // invalidate main Translation Lookup Buffer
+  asm("mcr p15, 0, r1, c8, c7, 0\n\r"
+      "isb" ::: "memory");
+
+  return section;
+}
diff --git a/paging.h b/paging.h
index 1218fb8..1715a6e 100644
--- a/paging.h
+++ b/paging.h
@@ -3,4 +3,12 @@
 
 void setup_flat_map(void);
 
+void setup_pager_structures(uint32_t available_mem);
+
+#define CLAIM_FAILURE 0xffff
+
+// returns section number or CLAIM_FAILURE in case of failure
+uint16_t claim_and_map_section
+(void *owner, uint16_t where_to_map, uint8_t access_permissions);
+
 #endif // PAGING_H
-- 
cgit v1.2.3


From 030cc96003d4518463b40f20d4194e51b1745c86 Mon Sep 17 00:00:00 2001
From: Wojtek Kosior <kwojtus@protonmail.com>
Date: Tue, 31 Dec 2019 17:36:50 +0100
Subject: don't statically define the unprivileged section (we're going to have
 many dynamically-claimed unprivileged sections)

---
 kernel_stage2.ld | 9 ---------
 memory.h         | 7 +------
 setup.c          | 3 ++-
 3 files changed, 3 insertions(+), 16 deletions(-)

(limited to 'memory.h')

diff --git a/kernel_stage2.ld b/kernel_stage2.ld
index a30105e..a3df063 100644
--- a/kernel_stage2.ld
+++ b/kernel_stage2.ld
@@ -60,13 +60,4 @@ SECTIONS
 
 		_stack_top = .;
 	}
-
-	.unprivileged_memory (NOLOAD) :
-	{
-		_unprivileged_memory_start = .;
-		
-		. = . + MMU_SECTION_SIZE;
-
-		_unprivileged_memory_end = .;
-	}
 }
diff --git a/memory.h b/memory.h
index 2ab5485..c7952f0 100644
--- a/memory.h
+++ b/memory.h
@@ -29,8 +29,6 @@ extern char _sections_list_start;
 extern char _sections_list_end;
 extern char _stack_start;
 extern char _stack_top;
-extern char _unprivileged_memory_start;
-extern char _unprivileged_memory_end;
 
 #define KERNEL_START ((size_t) &__start) // this is 0x0
 #define KERNEL_END   ((size_t) &__end)
@@ -50,11 +48,8 @@ extern char _unprivileged_memory_end;
 
 #define PRIVILEGED_MEMORY_END STACK_END
 
-#define UNPRIVILEGED_MEMORY_START				\
-  ((size_t) &_unprivileged_memory_start)  // equal to STACK_END
-#define UNPRIVILEGED_MEMORY_END			\
-  ((size_t) &_unprivileged_memory_end)
 
+// the following describes the virtual section for our PL0 programs
 #define PL0_SECTION_NUMBER ((size_t) 0xaaa)
 
 #define VIRTUAL_PL0_MEMORY_START (PL0_SECTION_NUMBER << 20)
diff --git a/setup.c b/setup.c
index b7acf60..63ca054 100644
--- a/setup.c
+++ b/setup.c
@@ -75,7 +75,8 @@ void setup(uint32_t r0, uint32_t machine_type,
       memory_size = 192 * POWER_OF_2(20);
     }
 
-  if (memory_size < UNPRIVILEGED_MEMORY_END)
+  // assume we need at least one section for PL0
+  if (memory_size < PRIVILEGED_MEMORY_END + SECTION_SIZE)
     {
       puts("Not enough memory to continue");
       while (1);
-- 
cgit v1.2.3


From 0ec2016f80ffb5971362559b98b10744bb5bf893 Mon Sep 17 00:00:00 2001
From: Wojtek Kosior <kwojtus@protonmail.com>
Date: Tue, 31 Dec 2019 18:54:28 +0100
Subject: split stack into irq, fiq and supervisor stacks; add proper handling
 of those in interrupt handlers (asm code)

---
 interrupt_vector.S | 25 ++++++++++++++-----------
 kernel_stage2.ld   | 22 +++++++++++++++++++---
 memory.h           | 18 +++++++++++++++---
 3 files changed, 48 insertions(+), 17 deletions(-)

(limited to 'memory.h')

diff --git a/interrupt_vector.S b/interrupt_vector.S
index 88b32bf..af80eec 100644
--- a/interrupt_vector.S
+++ b/interrupt_vector.S
@@ -9,39 +9,42 @@ _interrupt_vectors:
 	b fiq_handler_caller
 	
 reset_handler_caller:
-	ldr sp, =_stack_top
+	ldr sp, =_supervisor_stack_top
 	ldr r5, =reset_handler
 	bx r5
 	
 undef_handler_caller:
-	ldr sp, =_stack_top
+	ldr sp, =_supervisor_stack_top
 	ldr r5, =undefined_instruction_vector
 	bx r5
 
 svc_handler_caller:
-	ldr sp, =_stack_top
+	ldr sp, =_supervisor_stack_top
+	push {r5, lr}
 	ldr r5, =supervisor_call_handler
-	push {lr}
 	blx r5
-	ldm sp!, {pc} ^
+	ldm sp!, {r5, pc} ^
 	
 abort_handler_caller:
-	ldr sp, =_stack_top
+	ldr sp, =_supervisor_stack_top
 	ldr r5, =abort_handler
 	bx r5
 
 generic_handler_caller:
-	ldr sp, =_stack_top
+	ldr sp, =_supervisor_stack_top
 	ldr r5, =generic_handler
 	bx r5
 
 irq_handler_caller:
-	ldr sp, =_stack_top
-	ldr r5, =irq_handler
-	bx r5
+	ldr sp, =_irq_stack_top
+	sub lr, #4
+	push {r0-r3, lr}
+	ldr r3, =irq_handler
+	blx r3
+	ldm sp!, {r0-r3, pc} ^
 
 fiq_handler_caller:
-	ldr sp, =_stack_top
+	ldr sp, =_fiq_stack_top
 	ldr r5, =fiq_handler
 	bx r5
 	
diff --git a/kernel_stage2.ld b/kernel_stage2.ld
index a3df063..15b61ec 100644
--- a/kernel_stage2.ld
+++ b/kernel_stage2.ld
@@ -56,8 +56,24 @@ SECTIONS
 	{
 		_stack_start = .;
 		
-		. = . + MMU_SECTION_SIZE;
-
-		_stack_top = .;
+		_fiq_stack_start = .;
+		
+		. = . + (1 << 18);
+		
+		_fiq_stack_top = .;
+		
+		_irq_stack_start = .;
+		
+		. = . + (1 << 18);
+		
+		_irq_stack_top = .;
+		
+		_supervisor_stack_start = .;
+		
+		. = . + (1 << 19);
+		
+		_supervisor_stack_top = .;
+		
+		_stack_end = .;
 	}
 }
diff --git a/memory.h b/memory.h
index c7952f0..bdeba52 100644
--- a/memory.h
+++ b/memory.h
@@ -28,7 +28,13 @@ extern char _translation_table_end;
 extern char _sections_list_start;
 extern char _sections_list_end;
 extern char _stack_start;
-extern char _stack_top;
+extern char _fiq_stack_start;
+extern char _fiq_stack_top;
+extern char _irq_stack_start;
+extern char _irq_stack_top;
+extern char _supervisor_stack_start;
+extern char _supervisor_stack_top;
+extern char _stack_end;
 
 #define KERNEL_START ((size_t) &__start) // this is 0x0
 #define KERNEL_END   ((size_t) &__end)
@@ -43,8 +49,14 @@ extern char _stack_top;
 
 // first section after the translation table is left unused;
 // the next section is used as the stack
-#define STACK_START ((size_t) &_stack_start)
-#define STACK_END   ((size_t) &_stack_top)
+#define STACK_START            ((size_t) &_stack_start)
+#define FIQ_STACK_START        ((size_t) &_fiq_stack_start)
+#define FIQ_STACK_END          ((size_t) &_fiq_stack_top)
+#define IRQ_STACK_START        ((size_t) &_irq_stack_start)
+#define IRQ_STACK_END          ((size_t) &_irq_stack_top)
+#define SUPERVISOR_STACK_START ((size_t) &_supervisor_stack_start)
+#define SUPERVISOR_STACK_END   ((size_t) &_supervisor_stack_top)
+#define STACK_END              ((size_t) &_stack_end)
 
 #define PRIVILEGED_MEMORY_END STACK_END
 
-- 
cgit v1.2.3