sometime fds_gc() does not call the user handler.

[masataka hasegawa]

Hi,

I use the FDS on our project firmware.
The firmware saves some configurations to FDS at shutdown and the factory reset.
At this time, sometime fds_gc() does not call the user handler. (freq = 3/100)

Are there some known bugs?
or
Do you know any informations?

environment:
- SDK15
- nRF52840
- use Softdevice (use peer manager)
- FDS_VIRTUAL_PAGES 5
- FDS_VIRTUAL_PAGE_SIZE 1024
- FDS_BACKEND 2
- FDS_OP_QUEUE_SIZE 8
- FDS_CRC_CHECK_ON_READ 0
- FDS_CRC_CHECK_ON_WRITE 0
- FDS_MAX_USERS 8
- NRF_FSTORAGE_PARAM_CHECK_DISABLED 0
- NRF_FSTORAGE_SD_QUEUE_SIZE 4
- NRF_FSTORAGE_SD_MAX_RETRIES 8
- NRF_FSTORAGE_SD_MAX_WRITE_SIZE 4096

[masataka hasegawa]

Hi,

I added debug code (detect timeout and retry sd_flash_write.)
sd_flash_write returns NRF_ERROR_BUSY.

Thanks,

[masataka hasegawa]

Hi,

I want to reset flash module in softdevice.
Are there the reset API in softdevice?

Thanks,

[Kenneth]

Hello,

* @retval ::NRF_ERROR_BUSY           The previous command has not yet completed.

This means that there is already an on-going flash operation, you will have to wait until NRF_EVT_FLASH_OPERATION_SUCCESS or NRF_EVT_FLASH_OPERATION_ERROR events and try again.

There is no way to reset the flash module in the softdevice no.

Best regards,
Kenneth

[masataka hasegawa]

Hi,

Yes, I know the meaning of NRF_ERROR_BUSY.
My experiment is below.

1. our firmware calls fds_gc.
2. fds_gc calls some fstorage APIs.
                         :
3. fds_gc calls sd_flash_write.
4. no event and no call user handler.
5. check timeout.
6. if timeout is detected, f/w calls event handler itself.
7. fds_gc goes to next step.
8. fds_gc calls some sd APIs (sd_flash_write...)
9. sd_flash_write return NRF_ERROR_BUSY.

1 ~ 4, 7 ~ 9 = original SDK code
5, 6, = added my experiment code

This result means that sd_flash_write of No 3 have not complete.

Thanks,

[Kenneth]

Hello,

From my understand you have also been in contact with my colleague Zou-san. We are not able to recreate your issue at the moment. 

We will need a (simplified test) project from you that we can run on a nRF52840-DK to recreate your issue here, can you please arrange that for us?

Our FreeRTOS expert is back on Monday, he can look at your project then. Also please send us your system_nrf52.c file in your project so we can check that you have a workaround for errata_108().

Best regards,
Kenneth

[masataka hasegawa]

Hi,

I checked modules/nrfx/mdk/system_nrf52.c,
and I found the workaround code of errata 108.

if (errata_108()){
    *(volatile uint32_t *)0x40000EE4 = *(volatile uint32_t *)0x10000258 & 0x0000004F;
}

And we are discussing Zou-san.
We are trying to reproduce it at simple projects.

Thanks,