avoid erase of user space flash when debugging

[bet]

Posted on July 14, 2016 at 16:49

Hi,

I use a STM32F205RG6 chip and splitted the flash into 2 parts, 512K for the program and 512K for user's data:

MEMORY
{
FLASH (rx) : ORIGIN = 0x8000000, LENGTH = 512K
DATA (rwx) : ORIGIN = 0x8080000, LENGTH = 512K
RAM (xrw) : ORIGIN = 0x20000000, LENGTH = 128K
}
#in SECTIONS
.user_data :
{
. = ALIGN(4);
*(.user_data)
. = ALIGN(4);
} > DATA

It works like a charm and I can use the .user_data part in my code without worrying about messing the other part. The only issue that I have is that when I start the application (in debug or normal mode) via ST-link /v2 (through the IDE System workbench here but it shouldn't make any difference), The full flash is always erased. I modified the openocd flashing script for stm32f2 family (stm32f2x.cfg) to specify a flash bank size of 512K

#flash bank $_FLASHNAME stm32f2x 0 0 0 0 $_TARGETNAME
flash bank $_FLASHNAME stm32f2x 0 0x80000 0 0 $_TARGETNAME

but than, the flashing algorithm complain about a checksum mismatch (indeed the second half still has my data, as I want) Does someone know how to cope with such a situation, is there something to do to completely ignore the DATA part when flashing or should I just create a helper method to reload/reconfigure the second part of the flash everytime I start/debug the application via the ST-link? Thanks and regards, Zoyhar. #flash-stm32-st-link-openocd

[Tesla DeLorean]

Posted on July 14, 2016 at 17:19

Most commercial tools have a check box to do an sector/block level erase as required, Keil certainly does, using the ST-LINK, and others.

[Radosław]

Posted on July 14, 2016 at 22:20

NOLOAD.

Search linker manual for this.

[bet]

Posted on July 15, 2016 at 09:04

Thanks dembek, it was exactly what I was looking for and even with no need to hack the flashing scripts 🙂

So for the ones who reach this post with the same needs, adding a NOLAD in the section that you want to keep untouched, will force the linker not to initialize that section and therefore keep it as is

.user_data (NOLOAD) :
{
. = ALIGN(4);
KEEP (*(.user_data))
. = ALIGN(4);
} > DATA

Thanks again dembek and thanks as well Clive1, I was thinking to get in touch with AC6 to know if such a functionnality exist in system workbench but whenever possible, I prefer to know how it works underneath. Many thanks, Zoyhar.

[Radosław]

Posted on July 15, 2016 at 09:24

This is GCC linker feature. Becose AC6 is based on GCC, everything will work exacly this same.