sd_softdevice_disable in hardfault.

[victor]

Hi, I want to logg faults on my system so in the app_error_handler I want to store some data to flash. This location in flash is erased earlier when no error was present so I only want to do a write here. I want to call sd_softdevice_disable and then nrf_nvmc_write_bytes but sd_softdevice_disable hangs the system. Is there any other way to shut down the softdevice that I can do while in error state?

[Susheel Nuguru]

Can you tell us more about what you mean by sd_softdevice_disable hangs the system? does it call the softdevice_assert_handler? As fasr as I know, disabling the softdevice should work if the error happened in application and not the softdevice. Putting some code here might help speed up our analysis.

[victor]

This would be the stack trace when it locks:

sd_softdevice_disable (Locks here) settings_set_hardfault_reason (It is called hardfault but could be any fault) app_error_handler sd_assert_handler

So sd_softdevice_disable is called from the sd_assert_handler. I assume that this may be why it locks.

Some code:

void sd_assert_handler(uint32_t pc, uint16_t line_num, const uint8_t* p_file_name)
{
	app_error_handler(pc,line_num,p_file_name);
}
void app_error_handler(uint32_t error_code, uint32_t line_num, const uint8_t * p_file_name)
{
	settings_set_hardfault_reason(error_code,line_num,(char*)p_file_name);
	char str[256];
	sprintf(str, "C:%d, L:%d, F:%s\n",
			(int)error_code,
			(int)line_num,
			p_file_name);

	error_loop(str);
}

void settings_set_hardfault_reason(uint32_t code, uint16_t line, char *file)
{
	if(internal_lp_settings[0].hardfault_reason.line==0xFFFF)
	{
		char* filename=strrchr(file, '/') ? strrchr(file, '/') + 1 : file;
		strncpy((char*)&internal_lp_settings[0].hardfault_reason.file,(char*) &filename[0], MIN(strlen(filename)+1,sizeof(internal_lp_settings[0].hardfault_reason.file)));
		internal_lp_settings[0].hardfault_reason.code=code;
		internal_lp_settings[0].hardfault_reason.line=line;
		volatile pstorage_handle_t block_handle;
		pstorage_block_identifier_get(&pstorage_lp_flash_handle,0,(pstorage_handle_t*)&block_handle);
		sd_softdevice_disable();//Disable softdevice to ensure that no callback calls are made
		uint32_t hard_fault_reason_offset=sizeof(internal_lp_settings[0])-sizeof(internal_lp_settings[0].hardfault_reason);
		nrf_nvmc_write_bytes(block_handle.block_id+hard_fault_reason_offset,(void*)&internal_lp_settings[0].hardfault_reason,sizeof(internal_lp_settings[0].hardfault_reason));
	}
}

[Susheel Nuguru]

You cannot call softdevice calls in softdevice assert handler. Softdevice assert handler is running in the hard_fault context. That explains the system lock up. I am more concerned about the softdevice_assert that was triggered. That should not have happened in normal conditions. Can you tell me the file_name and line number values in sd_assert_handler? how and when was it triggered?

[victor]

It is line 189 in timeslotmanager. This happens when I pause the program when dubbging(Which is how I force a fault to test the fault logging function) but it also happens some times when i run the program. I think it may have something to do with flash writes that locks the chip for some time which results in my timeslot deadline not beeing met.

[Susheel Nuguru]

if you want to write to flash along with your timeslots, then you NEED to use sd_flash_xx functions. Otherwise the flash operations are happening without the knowledge of softdevice and hence the hardfault.