boardloader, bootloader: jump_to memory clearing and simplify code to asm

SConscript.bootloader

@@ -76,6 +76,7 @@ SOURCE_STMHAL = [
 ]
 
 SOURCE_BOOTLOADER = [
+    'embed/common/util.s',
     'embed/bootloader/startup.S',
     'embed/bootloader/header.S',
     'embed/bootloader/main.c',
@@ -141,6 +142,7 @@ env.Replace(
         'MCU_SERIES_F4',
         'PB_FIELD_16BIT',
     ] + CPPDEFINES_MOD,
+    ASFLAGS='-mthumb -mcpu=cortex-m4 -mfloat-abi=hard -mfpu=fpv4-sp-d16',
     ASPPFLAGS='$CFLAGS $CCFLAGS', )
 
 env.Replace(

SConscript.firmware

@@ -344,6 +344,7 @@ env.Replace(
         'STM32F405xx',
         ('STM32_HAL_H', '"<stm32f4xx_hal.h>"'),
     ] + CPPDEFINES_MOD,
+    ASFLAGS='-mthumb -mcpu=cortex-m4 -mfloat-abi=hard -mfpu=fpv4-sp-d16',
     ASPPFLAGS='$CFLAGS $CCFLAGS', )
 
 env.Replace(

embed/bootloader/memory.ld

@@ -97,3 +97,10 @@ _ram_start = ORIGIN(RAM);
 _ram_end = ORIGIN(RAM) + LENGTH(RAM);
 _heap_start = _ebss; /* heap starts just after statically allocated memory */
 _heap_end = 0x2001c000; /* tunable */
+/* used by the startup code to wipe memory */
+ccmram_start = ORIGIN(CCMRAM);
+ccmram_end = ORIGIN(CCMRAM) + LENGTH(CCMRAM);
+
+/* used by the startup code to wipe memory */
+sram_start = ORIGIN(RAM);
+sram_end = ORIGIN(RAM) + LENGTH(RAM);

embed/common/util.s

@@ -16,4 +16,36 @@ memset_reg:
   bne .L_loop_begin
   bx lr
 
+  .set SCB_VTOR, 0xE000ED08 // reference "Cortex-M4 Devices Generic User Guide" section 4.3
+
+  .global jump_to
+  .type jump_to, STT_FUNC
+jump_to:
+  mov r4, r0            // save input argument r0
+  // todo: this subroutine re-points the exception handlers before the C code
+  //       that comprises them have been given a good environment to run.
+  //       so, the this needs to disable interrupts before the VTOR
+  //       switch and then the reset_handler of the next stage needs to re-enable interrupts.
+  // todo: CPSID f
+  // wipe memory at the end of the current stage of code
+  ldr r0, =ccmram_start // r0 - point to beginning of CCMRAM
+  ldr r1, =ccmram_end   // r1 - point to byte after the end of CCMRAM
+  ldr r2, =0            // r2 - the word-sized value to be written
+  bl memset_reg
+  ldr r0, =sram_start   // r0 - point to beginning of SRAM
+  ldr r1, =sram_end     // r1 - point to byte after the end of SRAM
+  ldr r2, =0            // r2 - the word-sized value to be written
+  bl memset_reg
+  // todo: need to think through exception handler races for the VTOR and MSP change below
+  //       there are probably corner cases still.
+  // use the next stage's exception handlers
+  ldr r0, =SCB_VTOR
+  str r4, [r0]
+  // give the next stage a fresh main stack pointer
+  ldr r0, [r4]
+  msr msp, r0
+  // go on to the next stage
+  ldr r0, [r4, 4]
+  bx r0
+
   .end

embed/trezorhal/common.c

@@ -54,13 +54,6 @@ void periph_init(void) {
     DWT->CTRL |= DWT_CTRL_CYCCNTENA_Msk;  // Enable Cycle Count Register
 }
 
-void jump_to(uint32_t start)
-{
-    SCB->VTOR = start;
-    __asm__ volatile("msr msp, %0"::"g" (*(volatile uint32_t *)start));
-    (*(void (**)())(start + 4))();
-}
-
 void hal_delay(uint32_t ms)
 {
     HAL_Delay(ms);