Peer manager fails with NRF_ERROR_STORAGE_FULL

[markus-k]

I'm having problems getting the peer manager to work in a project forked from the ble_app_template on my nRF52-DK (using pca10040e, S112), complaining it can't initialize it's flash storage. But it doesn't even work using one of the examples (unmodified), ble_app_cscs for example. It was working at some point when I first tried, but it stopped eventually. I also tried erasing the chip using --eraseall or --chiperase as suggested in other threads, without success. Programming the chip also fails (times out) sometimes, maybe that's related.

The log is spammed with these messages:

<error> peer_manager_pds: Could not initialize flash storage. fds_init() returned 0xB.
<error> peer_manager: pm_init failed because pds_init() returned NRF_ERROR_STORAGE_FULL.
<error> app: Fatal error
<warning> app: System reset
<error> peer_manager_pds: Could not initialize flash storage. fds_init() returned 0xB.
<error> peer_manager: pm_init failed because pds_init() returned NRF_ERROR_STORAGE_FULL.
<error> app: Fatal error
<warning> app: System reset

Debugging shows the following:

• pm_init() calls pds_init()
• pds_init() calls fds_register() successfully, then fds_init()
• fds_init() calls pages_init()
• in pages_init() page_identify() is called for all pages, which returns FDS_PAGE_UNDEFINED
• page_is_erased() returns false as *p_page_addr = 0x5f46524e (or anything else but 0xffffffff)
• total_pages_available is 0 eventually and pages_init() returns NO_PAGES
• fds_init() fails with FDS_ERR_NO_PAGES and pds_init() return NRF_ERROR_STORAGE_FULL

Using pca10040 ans S132 on the ble_app_cscs example doesn't lead to this error though, only pca10040e/S112.

I'm using nRF5_SDK_15.3.0_59ac345 with arm-none-eabi-gcc 8.3.0 and nrfjprog version 9.8.1/JLinkARM.dll version 6.44a on Linux (Arch).

Anything I'm missing? I'm just getting started with nRF52 development.

Edit 1: Since it might be related, attached is the log from nrfjprog when flashing fails.

log.log

[run_ar]

Seem to me you do not have enough flash space to run the application you are trying to run on the nRF52810 (pca100040e project). So you need to use chip with a larger flash size (regular pca100040 project for nRF52832)

[markus-k]

The example has worked before, so that'd suprise me. That'd also mean the nRF52810 is simply not compatible with the peer manager, since even the plain ble_app_template doesn't work?

This is the size of my output binary (with a good 100kB margin to the chip size):

   text	   data	    bss	    dec	    hex	filename
  93240	    324	   4332	  97896	  17e68	_build/nrf52810_xxaa.out

[run_ar]

Could you dump the content of UICR and post it here? something like this: nrfjprog --readuicr uicr.txt?

ble_app_template works for me after performing nrfjprog -e.

[markus-k]

Okay, after a lot more testing, it seems A) my own template fork was getting too big when compiled with -O0 instead of -Og for debugging, and B) there was some really weird erasing issue with the other examples I don't really understand. But I think I got it sorted now.

[run_ar]

Thank you for the update. And yes, the flash size is rather limited on the nRF52810, so only minor changes could cause the application to grow to big.

I also had some erase issues, which I have not had time to recreate. Most likely caused by sector erase when going from a large application to a small one (or some garbage left in UICR as I have been using that as well), so some date was stuck in flash. I was always able to make it work with a complete erase.