Peer manager returns invalid peer id

[Moritz]

In our testing we encounter some devices (Centrals with nrf52832), where the peer manager event PM_EVT_BONDED_PEER_CONNECTED contains a invalid peer id 0x4000 (allowed values are between 0..255). This seems to be caused by a wrong FDS-record. The function "pds_peer_data_iterate" scans for all fds records with the record key PM_PEER_DATA_ID_BONDING (0xC007) and hereby finds a record that looks like this in flash:

What's strange about this record is that it matches the gap-address of the connecting peer, but the file id (marked in yellow) is 0x0000. Which will be reported as peer id 0x4000 (see function "file_id_to_peer_id" in peer_data_storage.c).

Looking into the peer_manager code, I haven't found a way how it could happen that it writes a file id of 0x0000. What do you think? Is this a software bug or a hardware issue. Maybe a reset happens while the file id is written by FDS in the function "record_header_write_finalize". But this also seems quite unlikely and we have seen this issue with multiple devices.

This is on SDK12.2

[Petter Myhre]

I'm not sure why this is happening. Could you explain a bit more how you do your testing? Any suggestions on how we can try to reproduce it here?

[Moritz]

In this case the testing was just pairing 2 nrf52-devices (Just works with LE Secure Connections) and sometimes it fails with the above described behaviour. It's the same code base where we found the issue regarding fstorage that Adrian reported. I think it's quite likely that the described race condition in fstorage is also the root of this issue.