Hi guys,
I have a crazy curiosity at my hands. The STM32G474 is a "category 3" device, therefore it has 128kB of RAM with the upper 32kB being CCM-SRAM. It is really nice that it is all addressable in one continuous block.
When i use the upper 16kB of the CCM-SRAM my debugger (using OpenOCD) breaks. The program works, I can stop the programm, but i cannot resume after a debug break. I inevitably jump to the NMI-Error handler. So something must have gone wrong.
Strangely, I can circumvent this annoying debugging problem by adjusting the Linker-Script to:
MEMORY /* WORKING CONFIG */
{
/* used for ISR Vector table */
CCMRAM (xrw) : ORIGIN = 0x20000000+96k+16k, LENGTH = 4K
/* used for Stack (will start in CCM-Ram) and heap */
RAM (xrw) : ORIGIN = 0x20000000, LENGTH = 96k+16k
FLASH (rx) : ORIGIN = 0x8000000, LENGTH = 256K
}
(using just 16kb of the CCM-Ram)
However if the stack resides in the upper 16k of CCM-SCRAM by using, for example, 18kb of CCM-ram:
MEMORY /* BROKEN CONFIG */
{
/* used for ISR Vector table */
CCMRAM (xrw) : ORIGIN = 0x20000000+96k+18k, LENGTH = 4K
/* used for Stack (will start in CCM-Ram) and heap */
RAM (xrw) : ORIGIN = 0x20000000, LENGTH = 96k+18k
FLASH (rx) : ORIGIN = 0x8000000, LENGTH = 256K
}
I have the debugging problem.
I mean - i can confirm that the CCM-RAM is there.. after all the program is working and explicitly testing some addresses in the upper 16k of the CCM-ram does work after all:
However the debugger seems to have a hard time accessing it.
This might be a problem with OpenOCD or a hardware problem inside the STM32G4.
I am using STLink-V2 via SWD. I think an important part of the problem is that the IDE tries to load the contents of the stack variables. So CCM-ram is accessed through the debugger.
Recommendations on how one should tackle the problem are welcome!
