Flash loader rework

flashloaders/Makefile

@@ -1,38 +1,38 @@
-# Note that according to the original GPLed code, compiling is noted to be
-# as simple as gcc -c, this fails with my tests where this will lead to a wrong 
-# address read by the program.
-# This makefile will save your time from dealing with compile errors
-# Adjust CC if needed
+# The flash loader code cannot be compiled by the system gcc. This 
+# makefile use arm-none-eabi-gcc for this purpose
 
-CC = /opt/local/gcc-arm-none-eabi-8-2018-q4-major/bin/arm-none-eabi-gcc
+CROSS_COMPILE ?= arm-none-eabi-
 
-CFLAGS_thumb1 = -mcpu=Cortex-M0  -Tlinker.ld -ffreestanding -nostdlib
-CFLAGS_thumb2 = -mcpu=Cortex-M3  -Tlinker.ld -ffreestanding -nostdlib
+CC = $(CROSS_COMPILE)gcc
+OBJCOPY = $(CROSS_COMPILE)objcopy
 
-all: stm32vl.o stm32f0.o stm32l.o stm32f4.o stm32f4_lv.o stm32l4.o stm32f7.o stm32f7_lv.o
+XXD = xxd
+XXDFLAGS = -i -c 4 
 
-stm32vl.o: stm32f0.s
-	$(CC) stm32f0.s $(CFLAGS_thumb2)  -o stm32vl.o
-stm32f0.o: stm32f0.s
-	$(CC) stm32f0.s $(CFLAGS_thumb1)  -o stm32f0.o
-stm32l.o: stm32lx.s
-	$(CC) stm32lx.s $(CFLAGS_thumb2)  -o stm32l.o
-stm32f4.o: stm32f4.s
-	$(CC) stm32f4.s $(CFLAGS_thumb2)  -o stm32f4.o
-stm32f4_lv.o: stm32f4lv.s
-	$(CC) stm32f4lv.s $(CFLAGS_thumb2)  -o stm32f4_lv.o
-stm32l4.o: stm32l4.s
-	$(CC) stm32l4.s $(CFLAGS_thumb2)  -o stm32l4.o
-stm32f7.o: stm32f7.s
-	$(CC) stm32f7.s $(CFLAGS_thumb2)  -o stm32f7.o
-stm32f7_lv.o: stm32f7lv.s
-	$(CC) stm32f7lv.s $(CFLAGS_thumb2)  -o stm32f7_lv.o
+CFLAGS_ARMV6_M = -mcpu=Cortex-M0 -Tlinker.ld -ffreestanding -nostdlib
+CFLAGS_ARMV7_M = -mcpu=Cortex-M3 -Tlinker.ld -ffreestanding -nostdlib
 
-clean:
-	rm *.o
+all: stm32vl.h stm32f0.h stm32lx.h stm32f4.h stm32f4lv.h stm32l4.h stm32f7.h stm32f7lv.h
+
 
+%.h: %.bin
+	$(XXD) $(XXDFLAGS) $< $@
 
+%.bin: %.o
+	$(OBJCOPY) -O binary $< $@
+	rm $<
 
+# separate rule for STM32F0
+stm32f0.o: stm32f0.s
+	$(CC) stm32f0.s $(CFLAGS_ARMV6_M) -o stm32f0.o
 
+# separate rule for STM32F1/F3
+stm32vl.o: stm32f0.s
+	$(CC) stm32f0.s $(CFLAGS_ARMV7_M) -o stm32vl.o
 
+# generic rule for all other ARMv7-M
+%.o: *.s
+	$(CC) $< $(CFLAGS_ARMV7_M) -o $@
 
+clean:
+	rm -f *.h

flashloaders/cleanroom.md

@@ -1,6 +1,4 @@
-Original Chinese version can be found below.
-
-# Clean Room Documentation English Version
+# Flash Loader Documentation
 
 Code is situated in section `.text`
 
@@ -12,20 +10,19 @@ All parameters would be passed over registers
 
 `r0`: the base address of the copy source
 `r1`: the base address of the copy destination
-`r2`: the total word (4 bytes) count to be copied (with expeptions)
+`r2`: the count of byte to be copied
+`r3`: flash register offset (used to support two banks)
 
 **What the program is expected to do**:
 
-Copy data from source to destination, after which trigger a breakpint to exit. Before exit, `r2` must be cleared to zero to indicate that the copy is done.
+Copy data from source to destination, after which trigger a breakpint to exit. Before exit, `r2` must be less or equal to zero to indicate that the copy is done.
 
 **Limitation**: No stack operations are permitted. Registers ranging from `r3` to `r12` are free to use. Note that `r13` is `sp`(stack pointer), `r14` is `lr`(commonly used to store jump address), `r15` is `pc`(program counter).
 
 **Requirement**: After every single copy, wait until the flash finishes. The detailed single copy length and the way to check can be found below. Address of `flash_base` shall be two-bytes aligned.
 
 ## stm32f0.s
 
-**Exception**: `r2` stores the total half word (2 bytes) count to be copied
-
 `flash_base`: 0x40022000
 
 `FLASH_CR`: offset from `flash_base` is 16
@@ -37,11 +34,11 @@ Copy data from source to destination, after which trigger a breakpint to exit. B
 
 **Special requirements**:
 
-Before every copy, read a word from FLASH_CR, set the lowest bit to 1 and write back. Copy one half word each time.
+Before every copy, read a word from FLASH_CR, set the PG bit to 1 and write back. Copy one half word each time.
 
-How to wait for the write process: read a word from FLASH_SR, loop until the content is not 1. After that, check FLASH_SR, proceed if the content is 4, otherwise exit.
+How to wait for the write process: read a word from FLASH_SR, loop until the busy bit is reset. After that, FLASH_SR is check. The process is interrupted if the error bit (0x04) is set.
 
-Exit: after the copying process and before triggering the breakpoint, clear the lowest bit in FLASH_CR.
+Exit: after the copying process and before triggering the breakpoint, clear the PG bit in FLASH_CR.
 
 ## stm32f4.s
 
@@ -56,7 +53,8 @@ Exit: after the copying process and before triggering the breakpoint, clear the
 **Special requirements**:
 
 Copy one word each time.
-How to wait for the write process: read a half word from FLASH_SR, loop until the content is not 1.
+
+How to wait for the write process: read a word from FLASH_SR, loop until the busy bit is reset.
 
 ## stm32f4lv.s
 
@@ -71,7 +69,7 @@ How to wait for the write process: read a half word from FLASH_SR, loop until th
 
 Copy one byte each time.
 
-How to wait from the write process: read a half word from FLASH_SR, loop until the content is not 1.
+How to wait from the write process: read a half word from FLASH_SR, loop until the busy bit is reset.
 
 ## stm32f7.s
 
@@ -89,16 +87,14 @@ Mostly same with `stm32f4.s`. Require establishing a memory barrier after every
 
 Mostly same with `stm32f7.s`. Copy one byte each time.
 
-## stm32l0x.s
+## stm32lx.s
 
 **Special Requirements**:
 
 Copy one word each time. No wait for write.
 
 ## stm32l4.s
 
-**Exception**: r2 stores the double word count to be copied.
-
 `flash_base`: 0x40022000
 `FLASH_BSY`: offset from `flash_base` is 0x12
 
@@ -109,14 +105,10 @@ Copy one word each time. No wait for write.
 
 Copy one double word each time (More than one registers are allowed).
 
-How to wait for the write process: read a half word from `FLASH_BSY`, loop until the lowest bit turns non-1.
-
-## stm32lx.s
+How to wait for the write process: read a half word from `FLASH_BSY`, loop until the busy bit is reset.
 
-Same with stm32l0x.s.
 
-
-# 净室工程文档-原始中文版
+# 净室工程文档-原始中文版 (out of date)
 
 代码位于的section：`.text`
 编译制导添加`.syntax unified`
@@ -139,8 +131,6 @@ Same with stm32l0x.s.
 
 ## stm32f0.s
 
-例外：`r2`:拷贝half word（2字节）数
-
 特殊地址定义：`flash_base`:定义为0x40022000
 
 `FLASH_CR`: 相对`flash_base`的offset为16
@@ -230,4 +220,4 @@ Same with stm32l0x.s.
 
 ## stm32lx.s
 
-要求与stm32l0x.s相同
+要求与stm32l0x.s相同

flashloaders/stm32f0.s

@@ -1,6 +1,14 @@
     .syntax unified
     .text
 
+    /*
+     * Arguments:
+     *   r0 - source memory ptr
+     *   r1 - target memory ptr
+     *   r2 - count of bytes
+     *   r3 - flash register offset
+     */
+
     .global copy
 copy:
     /*
@@ -17,54 +25,56 @@ copy:
      */
     nop
     nop
-    ldr r7, =flash_base
-    ldr r4, [r7]
-    ldr r7, =flash_off_cr
-    ldr r6, [r7]
-    adds r6, r6, r4
-    ldr r7, =flash_off_sr
-    ldr r5, [r7]
-    adds r5, r5, r4
 
-loop:
-    # FLASH_CR ^= 1
+    # load flash control register address
+    # add r3 to flash_base for support dual bank (see flash_loader.c)
+    ldr r7, flash_base
+    add r7, r7, r3
+    ldr r6, flash_off_cr
+    add r6, r6, r7
+    ldr r5, flash_off_sr
+    add r5, r5, r7
+
+    # FLASH_CR |= 0x01 (set PG)
     ldr r7, =0x1
-    ldr r3, [r6]
-    orrs r3, r3, r7
-    str r3, [r6]
+    ldr r4, [r6]
+    orrs r4, r4, r7
+    str r4, [r6]
 
+loop:
     # copy 2 bytes
-    ldrh r3, [r0]
-    strh r3, [r1]
+    ldrh r4, [r0]
+    strh r4, [r1]
 
-    ldr r7, =2
-    adds r0, r0, r7
-    adds r1, r1, r7
+    # increment address
+    adds r0, r0, #0x2
+    adds r1, r1, #0x2
 
-    # wait if FLASH_SR == 1
+    # BUSY flag
+    ldr r7, =0x01
 wait:
-    ldr r7, =0x1
-    ldr r3, [r5]
-    tst r3, r7
-    beq wait
+    # get FLASH_SR
+    ldr r4, [r5]
 
-    # exit if FLASH_SR != 4
-    ldr r7, =0x4
-    tst r3, r7
+    # wait until BUSY flag is reset
+    tst r4, r7
+    bne wait
+
+    # test PGERR or WRPRTERR flag is reset
+    ldr r7, =0x14
+    tst r4, r7
     bne exit
 
-    # loop if r2 != 0
-    ldr r7, =0x1
-    subs r2, r2, r7
-    cmp r2, #0
-    bne loop
+    # loop if count > 0
+    subs r2, r2, #0x2
+    bgt loop
 
 exit:
     # FLASH_CR &= ~1
     ldr r7, =0x1
-    ldr r3, [r6]
-    bics r3, r3, r7
-    str r3, [r6]
+    ldr r4, [r6]
+    bics r4, r4, r7
+    str r4, [r6]
 
     bkpt
 

flashloaders/stm32f4.s

@@ -1,6 +1,14 @@
     .syntax unified
     .text
 
+    /*
+     * Arguments:
+     *   r0 - source memory ptr
+     *   r1 - target memory ptr
+     *   r2 - count of bytes
+     *   r3 - flash register offset
+     */
+
     .global copy
 copy:
     ldr r12, flash_base
@@ -9,22 +17,24 @@ copy:
 
 loop:
     # copy 4 bytes
-    ldr r3, [r0]
-    str r3, [r1]
+    ldr r4, [r0]
+    str r4, [r1]
 
+    # increment address
     add r0, r0, #4
     add r1, r1, #4
 
-    # wait if FLASH_SR == 1
 wait:
-    ldrh r3, [r10]
-    tst r3, #0x1
-    beq wait
-
-    # loop if r2 != 0
-    sub r2, r2, #1
-    cmp r2, #0
-    bne loop
+    # get FLASH_SR
+    ldrh r4, [r10]
+
+    # wait until BUSY flag is reset
+    tst r4, #0x1
+    bne wait
+
+    # loop if count > 0
+    subs r2, r2, #4
+    bgt loop
 
 exit:
     bkpt

flashloaders/stm32f4lv.s

@@ -1,36 +1,40 @@
     .syntax unified
     .text
 
+    /*
+     * Arguments:
+     *   r0 - source memory ptr
+     *   r1 - target memory ptr
+     *   r2 - count of bytes
+     *   r3 - flash register offset
+     */
+
     .global copy
 copy:
     ldr r12, flash_base
     ldr r10, flash_off_sr
     add r10, r10, r12
 
-    # tip 1: original r2 indicates the count of 4 bytes need to copy,
-    #   but we can only copy one byte each time.
-    #   as we have no flash larger than 1GB, we do a little trick here.
-    # tip 2: r2 is always a power of 2
-    mov r2, r2, lsl#2
-
 loop:
     # copy 1 byte
-    ldrb r3, [r0]
-    strb r3, [r1]
+    ldrb r4, [r0]
+    strb r4, [r1]
 
+    # increment address
     add r0, r0, #1
     add r1, r1, #1
 
-    # wait if FLASH_SR == 1
 wait:
-    ldrh r3, [r10]
-    tst r3, #0x1
-    beq wait
-
-    # loop if r2 != 0
-    sub r2, r2, #1
-    cmp r2, #0
-    bne loop
+    # get FLASH_SR
+    ldrh r4, [r10]
+
+    # wait until BUSY flag is reset
+    tst r4, #0x1
+    bne wait
+
+    # loop if count > 0
+    subs r2, r2, #1
+    bgt loop
 
 exit:
     bkpt