Mesh seqnum reached maximum

RE: Mesh seqnum reached maximum

huybk213 — Thu, 02 Jul 2020 07:22:18 GMT

I have the same problem, but it seem really hard to fix
https://devzone.nordicsemi.com/f/nordic-q-a/63168/nrf-mesh-publish-error-15-forbidden

RE: Mesh seqnum reached maximum

Marciano-PL — Fri, 07 Feb 2020 23:45:59 GMT

Other issues opened with the same problem. I think the Mesh SDK fails to systematically reset the sequence number when the IV index gets incremented.

20d ago: https://devzone.nordicsemi.com/f/nordic-q-a/56741/nrf52832-nrf-sdk-for-mesh-light_switch_client-getting-nrf_error_forbidden-error/230923#230923

1mo ago: https://devzone.nordicsemi.com/f/nordic-q-a/55988/mesh-and-uart-nrf_error_forbidden

3mo ago: https://devzone.nordicsemi.com/f/nordic-q-a/54029/access_model_publish-returns-nrf_error_forbidden-error

RE: Mesh seqnum reached maximum

Marciano-PL — Fri, 07 Feb 2020 23:38:27 GMT

I'm finding that this issue isn't truly solved. I left my DK on for several weeks and when I returned to debug, I found that although the IV index had been incremented, the sequence number remained at the maximum. Each time my application called for sending a message, m_net_state.seqnum < m_net_state.seqnum_max_available evaluated as false, indicating that it never rolled over.

Setup

nRF52840-DK
nRF5 SDK v16.0.0
Mesh SDK v4.0.0
S140 v7.0.0

RE: Mesh seqnum reached maximum

Hung Bui — Mon, 22 Jan 2018 11:34:41 GMT

The sequence number is explained at section 2.3.9.4 in Mesh Profile v1.0

According to the spec, IV index need to be updated to avoid the sequence number wrap around. We have implemented that in net_state_seqnum_alloc() in net_state.c. IV index is supposed to be updated.

How long did it take you to see the issue ? To reach the maximum sequence number, it takes dozens of day to 1, 2 months. Do you see the problem if you disable flash (PERSISTENT_STORAGE = 0) ?

I would suggest to change the NETWORK_SEQNUM_MAX to something smaller and run the test again and check if iv_update_trigger_if_pending() is called.

RE: Mesh seqnum reached maximum

Will Dimmit — Thu, 18 Jan 2018 20:08:00 GMT

Looking at the code, I believe that net_state_seqnum_alloc() is trying to request a new block of sequence numbers with the call seqnum_block_allocate() That should require a write to flash, which may be lengthy. I'm guessing this function is supposed to block TX while the new block is being allocated, but walking back up the chain of function calls to access_model_publish(), I don't see a retry mechanism.

I haven't seen this issue yet in my testing, but I worry that I may. I'm looking forward to what Nordic has to say about this.

Cheers,

Will