I'm creating a bootloader for STM32F429 with the gnu toolchain (9.2.1) and trying to reserve some flash memory for user data, shared by bootloader and application. I want to reserve the second of its first four 16K flash sectors for this, since they're nice and small: all other sectors are 128K. The memory layout should look look like this:
ISR vector table : 0x08000000 (428 bytes)
padding : 0x080001ac (15956 bytes)
-----------------------------
user data : 0x08004000 (16K)
-----------------------------
bootloader code : 0x08008000 (max 224K, for total of max. 256K)
I have modified ST's linker script to look like this:
/* Specify the memory areas */
MEMORY
{
FLASH (rx) : ORIGIN = 0x08000000, LENGTH = 256K /* max. bootloader binary size */
CCMRAM (rw) : ORIGIN = 0x10000000, LENGTH = 64K
RAM (xrw) : ORIGIN = 0x20000000, LENGTH = 191K /* deduct 1K for NOINIT data */
NOINIT (rwx) : ORIGIN = 0x2002FC00, LENGTH = 1K /* NOINIT data will survive reset */
}
/* Define output sections */
SECTIONS
{
/* The startup code goes first into FLASH */
.isr_vector :
{
. = ALIGN(4);
KEEP(*(.isr_vector)) /* Startup code */
. = ALIGN(4);
} >FLASH
/* Then comes user flash sector, make sure to get one of the first 4 that are 16K in size
.user_flash :
{
. = ALIGN(0x4000)
KEEP(*(.user_flash))
. = ALIGN(0x4000)
} >FLASH
/* ... more sections below not shown here */
}
In my bootloader.cpp, I declare the user flash as follows:
__attribute__((__section__(".user_flash"))) const uint32_t user_flash[0x4000 / sizeof(uint32_t)] = {0};
However, in the debugger, the address of user_flash is not 0x08004000 as expected, but 0x0801b4e8.
Removing the __attribute__((__section__(".user_flash"))) from the declaration yields a different adress, proving the attribute has at least some effect.
Running arm-non-eabi-objdump -h on the .elf file confirms the (incorrect) adress of the user_flash section.
I have also tried declaring the section with . = 0x08004000, to no effect.
Another approach I tried is to place .isr_vector in its own 16K section, followed by a 16K section for the user flash and finally a (256-16-16) = 224K FLASH section. This produced a bogus binary that would hardfault.
What am I doing wrong here?
Edit:
I've tried inserting a .padding section after .isr_vector to try to align .user_flash that way, but no matter what I do, .text and .rodata sections seem to always come immediately after .padding, even though they are declared after the .user_flash section.
