Problem with fstorage changes in SDK v14

[Richard D.]

I'm migrating an existing application from SDK v13 to SDK v15 and have run into a minor issue with the fstorage changes introduced in SDK v14.

My application uses both fds and fstorage libraries. In v13, this worked because fstorage figured out for itself where to locate FDS and my own flash areas. In v14, one is required to specify flash start/end addresses in nrf_fstorage_t which would logically be just below the area used by FDS, but there is no way to determine which memory is used by fds.

The obvious solution is an FDS api which returns the starting address of flash used by FDS (m_fs.start_addr). I can implement this myself by hacking fds, but want to bring this to the attention of the Nordic developers in the context of the Mesh SDK which is completely missing the functionality previously provided by fstorage (this thread).

Succinctly, concurrent use of fds and fstorage is a common use case for which we need appropriate api's.

Thanks,
Richard

[Jørgen Holmefjord]

Hi,

The FDS area is quite simple to figure out from the application. The parameters to fstorage is set in fds_init, using the following functions:

static uint32_t flash_end_addr()
{
    uint32_t const bootloader_addr = NRF_UICR->NRFFW[0];
    uint32_t const page_sz         = NRF_FICR->CODEPAGESIZE;
    uint32_t const code_sz         = NRF_FICR->CODESIZE;

    return (bootloader_addr != 0xFFFFFFFF) ? bootloader_addr : (code_sz * page_sz);
}


static void flash_bounds_set(void)
{
    uint32_t flash_size  = (FDS_PHY_PAGES * FDS_PHY_PAGE_SIZE * sizeof(uint32_t));
    m_fs.end_addr   = flash_end_addr();
    m_fs.start_addr = m_fs.end_addr - flash_size;
}

Best regards,
Jørgen

[Richard D.]

Hi Jørgen,

I don't understand your comment.  Are suggesting that my application duplicate the code in the FDS static (private) function?

Richard

[Jørgen Holmefjord]

Yes, I'm saying that the FDS area should be static and known to the application, from the configuration of number of pages in sdk_config.h, and the size of the flash that is stored in FICR.

[Richard D.]

While I agree that applications can compute the location used by FDS, it would be far better if FDS exposed an api something like

uint32_t fds_get_flash_start_address(void);

With such an API, applications are decoupled from the FDS and bootloader implementation details.

[Jørgen Holmefjord]

I have forwarded your request to our SDK team.