ZigBee Application crash - ERROR 29 [RET_OVERFLOW] at [...]zigbee_app_utils.c:849

[Rogger]

Application crashes with the following log

[06:45:22.727,691] <err> zigbee_app_utils: ERROR 29 [RET_OVERFLOW] at WEST_TOPDIR/nrf/subsys/zigbee/lib/zigbee_app_utils/zigbee_app_utils.c:849
[06:45:22.727,691] <err> zboss_osif: ZBOSS fatal error occurred

The relevant lines from zigbee_app_utils

 

#if defined CONFIG_ZIGBEE_ROLE_END_DEVICE
wait_for_user_input = false;
is_rejoin_start_scheduled = false;

zb_ret_t zb_err_code = ZB_SCHEDULE_APP_ALARM(
stop_network_rejoin,
ZB_TRUE,
ZB_MILLISECONDS_TO_BEACON_INTERVAL(
ZB_DEV_REJOIN_TIMEOUT_MS));
ZB_ERROR_CHECK(zb_err_code);
#endif

It seems that the Alarm cannot be scheduled due to RET_OVERFLOW.

- The implementation of the signal handler is here https://github.com/Rogger/schneggi-sensor/blob/64b2408f1053a15622c618249d887ca7694eb250/src/main.c#L704, in there is as well the call to app utils.

Context:

SDK & toolchain in use 2.5.0

Any help or guidance is highly appreciated, thank you in advance

Best Regards

Michael

[Elfving]

Hi,

Could you get me the network key for this trace?

Regards,

Elfving

[Rogger]

Hi,

Sure, the key: ffb12ed4228949b66f1a72143a263808 (normal byte order).

Thanks!

[Elfving]

Hello pwpot,

Have you created another DevZone case for this issue yourself?

Regards,

Elfving

[Pawel(embeddedsolutions.pl)]

Hi, Elfving

I am going to but at this moment I cannot provide any detailed info about the issue as device is already on client site.

I conclude based on some LED blinks and on my own experience.
Some time ago I made a mistake in code which resulted in ZBOSS Queue overflow. 
I was scheduling one of the callbacks multiple times without proper cancellation.
As a result of that ZBOSS scheduler stopped working and the ZBOSS Queue did not recover from such state.

In my case the problem is also related with rejoin and it happens less often then how it is in Rogger code, its days, sometimes weeks, not hours. 
I assume this is because Rogger is scheduling ZBOSS callback to use it as a main while loop (very often) where I am doing it mostly during rejoin.

I am not sure if ZBOSS deadlocks or sth and if can it be detected by classic watchdog. 

Is it possible to increase ZB_CONFIG_SCHEDULER_Q_SIZE and still be FOTA compatible ? 
Developing with ZBOSS guide says that: 
The memory configuration must not be changed between firmware versions if the firmware is going to be upgraded using the OTA DFU.

I think it is very generic info and maybe ZB_CONFIG_SCHEDULER_Q_SIZE is not the part of that.

[Rogger]

Hi Pwpot,

Thanks for the tip, unfortunately this did not help.

[Elfving]

Hello and sorry about the wait,

Have you had any development on this case from your side, or are you still facing the same issue?

Regards,

Elfving

[Rogger]

Unfortunately yes, I am still facing the same issue.

[Elfving]

I don't know exactly what the issue is. Though it could be like pwpot was suggesting here, that you are scheduling one of the callbacks multiple times without proper cancellation, or that you have recursive tasks. Could it be that you are also not freeing the buffers?

You might want to check all the places where you call ZB_SCHEDULE_APP_CALLBACK() or ZB_SCHEDULE_APP_ALARM().

Could you also try lowering down the Q_SIZE again, to see if maybe that maybe was the only thing that postponed the error?

Regards,

Elfving

[Rogger]

I agree, it looks like an issue related to not freeing buffers or not canceled scheduled callbacks.

But since I can reproduce the issue with the light bulb example (not modified, SDK 2.5.2) I thin it might be a bug in the ZigBee stack.

I will try to dig deeper to find more information.

[Pawel(embeddedsolutions.pl)]

Hi Rogger, 

I went through your code one more time and I found something. 
I think calling 
ZB_SCHEDULE_APP_ALARM(sensor_loop, ZB_ALARM_ANY_PARAM,
ZB_MILLISECONDS_TO_BEACON_INTERVAL(0));
in main is not necessary because your sensor_loop first call will be scheduled by the ZB_ZDO_SIGNAL_SKIP_STARTUP signal anyway. 

I suggest sth like this: 

 int main(void)
 {
    ...
    ...
 
    /* Start Zigbee default thread */
	zigbee_enable();
    while (!zigbee_is_stack_started()) {
        /* Sleep is mandatory (Main thread goes to sleep, ZBOSS thread starts working):
         * Allow the ZBOSS Thread to be created and started.
         */
        k_msleep(200);
    }
    
    return 0;
}