Don't understand page numbering for sd_flash_write and sd_flash_erase

[David Smoot]

I want to use the last page of codespace for storing a single 32 bit word of non volatile data. I am failing so far and I think I just don't understand the page numbering scheme the soft device API is using.

So the page number I think I want to use for app data is: NRF_FICR->CODESIZE -1.

I have a simple function where I want to store three bytes in flash by stuffing them in a 32 bit word:

uint32_t write_cal_data(uint8_t I680, uint8_t I850, uint8_t I970)
{
	uint32_t retval;
	retval = sd_flash_page_erase(get_app_data_pagenum());
	if (retval != NRF_SUCCESS)
		return retval;
	uint32_t * p_dest = (uint32_t *)get_app_data_base_addr();
	uint32_t combined_currents = (I680 << 16) | (I850 << 8) | I970;
	retval = sd_flash_write(p_dest, &combined_currents, 1);
	return retval;
}

That code completes with all the return values equal to NRF_SUCCESS.

But when I try to access the stored data, I dereference address (NRF_FICR->CODEPAGESIZE * (NRF_FICR->CODESIZE - 1));, I do not get the data I previously stored.

If I write a 32 bit word to page number "NRF_FICR->CODESIZE -1", am I correct to then access that word at the address above? What am I missing? Thanks, David

[Håkon Alseth]

Hi David,

The softdevice will schedule the write-process in between connection intervals and give you an event back when the write has finished (NRF_EVT_FLASH_OPERATION_SUCCESS or NRF_EVT_FLASH_OPERATION_ERROR) in the callback set in function "softdevice_sys_evt_handler_set(my_handler);".

Note: if you're using pstorage module, it uses the last two pages by-default, and you should then use calls to pstorage instead of directly accessing the sd_*-API.

Best regards Håkon

[David Smoot]

Thanks, I will try the call back method and the pstorage API if the call back does not work.

Can you confirm that my addressing is correct? If I use the sd_flash_write function to write to page number NRF_FICR->CODESIZE -1, can I then access the first address on that page by dereferencing (NRF_FICR->CODEPAGESIZE * (NRF_FICR->CODESIZE - 1))?

Thanks, David

[Håkon Alseth]

Hi David,

That should be valid, unless there's something else in your application using that address area.

Best regards Håkon

[David Smoot]

Some progress due to a very smart coworker. A reason the code above does not work as I posted it is that I am trying to schedule a write operation immediately after scheduling the erase operation. The erase operation has not completed yet when I call sd_flash_write. I am still working on my code, will try to update this thread if / when I get it correct.

[David Smoot]

Ok, I don't know if this will help anyone else but here goes in case:

My original code did not work for two reasons:

1. If you use the sd_flash_* API only allows one pending operation at a time. Learned this from a reply on another thread flash_write.
2. But more importantly is the scope of the variable being written. If you use a local function variable as an argument to sd_flash_write, it will most likely be out of scope and not holding the data you want when the soft device gets around to actually performing the flash write.