rejoin drain rewiev

[Team-MT]

Title: Critical bugs in rejoin logic causing unnecessary battery drain and non-optimal hub search

Description:

While reviewing the rejoin and steering logic in src/zb_appCb.c, I found two serious bugs and a few minor discrepancies that together cause excessive radio activity and unpredictability in network reconnection. These issues directly increase battery consumption and make the device less reliable in finding/keeping the coordinator.

1. Inverted timer cancel condition in zb_bdbInitCb (major)

File: src/zb_appCb.c
Block: #if REJOIN_FAILURE_TIMER inside zb_bdbInitCb

if(joinedNetwork) {
    zb_rejoinReqWithBackOff(zb_apsChannelMaskGet(), g_bdbAttrs.scanDuration);
    if(!g_sensorAppCtx.timerRejoinBackoffEvt){   // ❌ wrong: should be without "!"
        TL_ZB_TIMER_CANCEL(&g_sensorAppCtx.timerRejoinBackoffEvt);
    }
}

Problem:
The condition if(!evt) cancels the timer only when the timer does NOT exist. When the device is already on a network but receives BDB_INIT_STATUS_FAILURE, the existing rejoin-backoff timer is never cancelled. As a result, periodic rejoin attempts keep firing even while connected, wasting power and generating unnecessary radio traffic.

fix:
Remove the negation:

if(g_sensorAppCtx.timerRejoinBackoffEvt) {
    TL_ZB_TIMER_CANCEL(&g_sensorAppCtx.timerRejoinBackoffEvt);
}

Missing return in sensorDevice_rejoinBackoff (major)
File: src/zb_appCb.c
Function: sensorDevice_rejoinBackoff – #else branch (when REJOIN_FAILURE_TIMER is not active)

#else
    static bool rejoinMode = REJOIN_SECURITY;
    ...
    rejoinMode = !rejoinMode;
    // ❌ missing return 0;
#endif
}

Problem:
The function is declared as static s32 sensorDevice_rejoinBackoff(void *arg), but the #else path lacks a final return statement. This is undefined behavior. The return value of a timer callback usually controls whether the timer is re-scheduled (0 = restart, -1 = stop). A garbage return value can either stop the rejoin backoff entirely (device never recovers) or restart it with an unexpected interval, leading to erratic reconnection attempts and increased power drain.

Fix:
Add return 0; before the #endif:

    rejoinMode = !rejoinMode;
    return 0;
#endif

Impact summary

• Battery drain: The device performs unnecessary rejoin attempts even when already connected.
• Unreliable reconnection: The broken return value can break the backoff mechanism.
• Network noise: Extra Zigbee traffic may affect coordinator performance.

[pvvx]

Thanks for finding the typo.

1. But this does not lead to the described actions.
   In this branch, the RejoinBackoffEvt timer must be assigned or rejoinBackoff() must be executed.
   In the end, this condition is an unnecessary piece of code.

2. Code without the REJOIN_FAILURE_TIMER option is not used.
   This is because the network search sequence defined in the SDK will then run with the default pauses defined in the standard.
   The SDK (and standard) require the network search function to run almost continuously, which drains the battery within a couple of hours. When you disable the REJOIN_FAILURE_TIMER option, the compiler will automatically notify you about the error of missing “return n”.