unexpected BLE_GATTC_EVT_TIMEOUT

[DmitrySol]

Hi.

We use nRF5_SDK_17.0.2_d674dde as central

And phone as peripheral.

Usually system works correctly for a long time.

But sometimes it enter in strange state.

Connections are broken every 30 seconds after successful connection.

I have think previously, it is because of no activity on NUS service.

So sending NUS message after BLE_NUS_C_EVT_DISCOVERY_COMPLET was added.

But it doesn't help.

Please teach a nature of this behavior and advice steps to prevent it.

Thanks. BR.

[Kenneth]

Hi,

30 seconds is typically a timeout for a procedure that is not handled. An on-air sniffer log would be very useful:

https://www.nordicsemi.com/Software-and-tools/Development-Tools/nRF-Sniffer-for-Bluetooth-LE 
https://infocenter.nordicsemi.com/pdf/nRF_Sniffer_BLE_UG_v3.2.pdf (there might be newer depending on when you read this)

Kenneth

[DmitrySol]

Thanks for answer.

But I can't catch log because of rarely problem.

We have logs from uart. But it is truncated because of size of buffers of host controller.

So I hesitate that it can be useful.

[DmitrySol]

Hi

I was try to insert LTK in PASS Key / OOB place.
Encryption was not successful.

LTK: AD FE E8 94 50 18 16 DD D8 B1 F0 75 9E 74 6D 7F

I have attempted with:
0xADFEE894501816DDD8B1F0759E746D7F
0x7F6D749E75F0B1D8DD16185094E8FEAD

[DmitrySol]

Hi.

What can be cause of unsuccessful decryption with using LTK?

[Kenneth]

Sorry for the late answer, it seems you are right, I am not able to make this work either at the moment.

I think you will have to sniff the initial connection that include bonding, then as long as you don't restart wireshark, you should be able to sniff and decrypt the the problematic re-connection later when you start receiving the BLE_GATTC_EVT_TIMEOUT.

Kenneth

[DmitrySol]

Thanks for reply.

But it is impossible. Issue is appearing very rarely.

I  can't support wireshark  to be connected so long time.

Perhaps any other ideas?

[Kenneth]

Some of the more expensive proffesional sniffers such as Ellisys and Fronline can insert LTK, but those are typically rather expensive.

An alternative would be to dump all the events ID's when the issue occur, e.g. in ble_evt_handler() log the 'p_ble_evt->header.evt_id's. Some of them are frequent, like BLE_GATTS_EVT_HVN_TX_COMPLETE, so they can be ignored, I am more interested if you find an event that occurs when failing vs. when working.

Kenneth

[DmitrySol]

Hi! Kenneth.

I attach logs of normal and abnormal work.

If you need some more specific log.
Please say what types of logs I need to add.

Logs can be truncated a bit because of size of log buffers.

BR. Dmitry.

Log of abnormal work

Log of normal work

[Kenneth]

I don't know what may be the problem here, sorry. You will just have to experiment, e.g. with delays or try to look deeper into the sequence of softdevice api calls and events. In worst case you will have to take a few steps back, e.g. go back to the default nRF5 SDK examples and work your way until it start failing. An sniffer log would be very helpful here, without it it's almost impossible to know the reason of timeout without just experimenting and trying things.

I can find differences, for instance there are differences seemingly both in security and gatt discovery in your log. Considering you are experiencing a timeout, it is due do:

https://infocenter.nordicsemi.com/topic/com.nordic.infocenter.s132.api.v7.2.0/group___b_l_e___g_a_t_t_c___t_i_m_e_o_u_t___m_s_c.html

However, which of these procedures are causing the timeout is is difficult to know:

https://infocenter.nordicsemi.com/topic/com.nordic.infocenter.s132.api.v7.2.0/group___b_l_e___g_a_t_t_c___m_s_c.html

Kenneth