RTOS/CC1310: Hanging in TI RTOS functions "Clock_walkQueueDynamic()"

OK. here is (hopefully)

the CCS Screenshot:

ROV Info:

OK, ones more:

Ok. Hopefully someone can help me now.
BR Egon

Hi M-W,

yes this is true. May this be a Problem?

Here you can see the construction of the clocks:

  /* Construct clock for App- and TT-Interval */

    Clock_Params clockParams;
    Clock_Params_init(&clockParams);
    clockParams.arg = (UArg)WAKEUP_SOURCE_SLEEPTIMER_APPL;
//    clockParams.period = 10000000/Clock_tickPeriod;   // 10s (10Mio us)   --> kann später mit Clock_setPeriod(handle, intervall[us]/Clock_tickPeriod); noch geändert werden
    Clock_construct(&AppClock, wakeupCb, 0, &clockParams);
    hAppClock = Clock_handle(&AppClock);

    clockParams.arg = (UArg)WAKEUP_SOURCE_SLEEPTIMER_TT;
//    clockParams.period = 10000000/Clock_tickPeriod;   // 10s (10Mio us)   --> kann später mit Clock_setPeriod(handle, intervall[us]/Clock_tickPeriod); noch geändert werden
    Clock_construct(&TTClock, wakeupCb, 0, &clockParams);
    hTTClock = Clock_handle(&TTClock);

           Clock_Params clockParams;
           Clock_Params_init(&clockParams);
           clockParams.arg = (UArg)WAKEUP_SOURCE_SLEEPTIMER_PROTOCOLL;
           Clock_construct(&HopClock, wakeupCb, (uint32)ui16SleepProtocoll*1000/Clock_tickPeriod, &clockParams);
           hHopClock = Clock_handle(&HopClock);
           Clock_start(hHopClock);

And the general Timer, I think the clocks derive from a GPT.

/*
 *  ============================ GPTimer begin =================================
 *  Remove unused entries to reduce flash usage both in Board.c and Board.h
 */
/* Place into subsections to allow the TI linker to remove items properly */
#if defined(__TI_COMPILER_VERSION__)
#pragma DATA_SECTION(GPTimerCC26XX_config, ".const:GPTimerCC26XX_config")
#pragma DATA_SECTION(gptimerCC26xxHWAttrs, ".const:gptimerCC26xxHWAttrs")
#endif

/* GPTimer hardware attributes, one per timer part (Timer 0A, 0B, 1A, 1B..) */
const GPTimerCC26XX_HWAttrs gptimerCC26xxHWAttrs[HWDEF_GPTIMERPARTSCOUNT] = {
    { .baseAddr = GPT0_BASE, .intNum = INT_GPT0A, .intPriority = (~0), .powerMngrId = PowerCC26XX_PERIPH_GPT0, .pinMux = GPT_PIN_0A, },
    { .baseAddr = GPT0_BASE, .intNum = INT_GPT0B, .intPriority = (~0), .powerMngrId = PowerCC26XX_PERIPH_GPT0, .pinMux = GPT_PIN_0B, },
    { .baseAddr = GPT1_BASE, .intNum = INT_GPT1A, .intPriority = (~0), .powerMngrId = PowerCC26XX_PERIPH_GPT1, .pinMux = GPT_PIN_1A, },
    { .baseAddr = GPT1_BASE, .intNum = INT_GPT1B, .intPriority = (~0), .powerMngrId = PowerCC26XX_PERIPH_GPT1, .pinMux = GPT_PIN_1B, },
    { .baseAddr = GPT2_BASE, .intNum = INT_GPT2A, .intPriority = (~0), .powerMngrId = PowerCC26XX_PERIPH_GPT2, .pinMux = GPT_PIN_2A, },
    { .baseAddr = GPT2_BASE, .intNum = INT_GPT2B, .intPriority = (~0), .powerMngrId = PowerCC26XX_PERIPH_GPT2, .pinMux = GPT_PIN_2B, },
    { .baseAddr = GPT3_BASE, .intNum = INT_GPT3A, .intPriority = (~0), .powerMngrId = PowerCC26XX_PERIPH_GPT3, .pinMux = GPT_PIN_3A, },
    { .baseAddr = GPT3_BASE, .intNum = INT_GPT3B, .intPriority = (~0), .powerMngrId = PowerCC26XX_PERIPH_GPT3, .pinMux = GPT_PIN_3B, },
};

/*  GPTimer objects, one per full-width timer (A+B) (Timer 0, Timer 1..) */
GPTimerCC26XX_Object gptimerCC26XXObjects[HWDEF_GPTIMERCOUNT];

/* GPTimer configuration (used as GPTimer_Handle by driver and application) */
const GPTimerCC26XX_Config GPTimerCC26XX_config[HWDEF_GPTIMERPARTSCOUNT] = {
    { &gptimerCC26XXObjects[0], &gptimerCC26xxHWAttrs[0], GPT_A },
    { &gptimerCC26XXObjects[0], &gptimerCC26xxHWAttrs[1], GPT_B },
    { &gptimerCC26XXObjects[1], &gptimerCC26xxHWAttrs[2], GPT_A },
    { &gptimerCC26XXObjects[1], &gptimerCC26xxHWAttrs[3], GPT_B },
    { &gptimerCC26XXObjects[2], &gptimerCC26xxHWAttrs[4], GPT_A },
    { &gptimerCC26XXObjects[2], &gptimerCC26xxHWAttrs[5], GPT_B },
    { &gptimerCC26XXObjects[3], &gptimerCC26xxHWAttrs[6], GPT_A },
    { &gptimerCC26XXObjects[3], &gptimerCC26xxHWAttrs[7], GPT_B },
};

/*
 *  ============================ GPTimer end ===================================
 */

BR,
Noge

Hi M-W,

what I can say to the swi, th function UInt32 Clock_walkQueueDynamic(Bool service, UInt32 thisTick) is not left by return, the firmware always run into the function. But I do not know which module / swi it is, because I newer come out. I had a look at our board file, but I do not have a module with a swi priority of 5.

To the clocks, yes period and timeouts are set at other different places on the application. and also starting the clocks.

Do you know, if it is a SWI of the knl which runs on priority 5?

BR,
Egon

Hi M-W,

the "obj->fxn(obj->arg);" is not called. The condition if (obj->active == TRUE) { ...} always stays false.
What you said about the Void Clock_workFuncDynamic(UArg arg0, UArg arg1) function, I cant follow you, because the firmware dont come back to it.
Here there is a screenshot of it

BR,
Noge

Hi M-W,

there is no callback function called from  obj->fxn(obj->arg);. The condition  if (obj->active == TRUE) {...} in the UInt32 Clock_walkQueueDynamic(Bool service, UInt32 thisTick) is always false and it does not return from this function. How to check/link  the  ServiceTick, I canot follow you.

here is a new creenshot from the Void Clock_workFuncDynamic(UArg arg0, UArg arg1) 

What I could see, the trouble is caused from my last clock instance I created. and it is done like this:

           Clock_Params clockParams;
           Clock_Params_init(&clockParams);
           clockParams.arg = (UArg)WAKEUP_SOURCE_SLEEPTIMER_PROTOCOLL;
           Clock_construct(&HopClock, wakeupCb, (uint32)ui16SleepProtocoll*1000/Clock_tickPeriod, &clockParams);
           hHopClock = Clock_handle(&HopClock);
           Clock_start(hHopClock);

Is there anythin wrong?

BR,

Noge

Hi M-W,

Sharing the project will be difficult, because it is confidential information. Part of it may work. But before I will try to have a look at the stack. I used the Stack usage tap of the CCS, but it remaisn empty, Why?

Her please see my screenshot:

BR Noge

Hi M-W,

See my screenshot:

What do you think, it seems the stack is empty, but once was filled up to the address 0x20001094. Is the stack mok, or has there been an overflow?
BR Noge

Hi M-W,

I was on a sick leave, so some time passed. But now I'm back on work.  So let' focus on the clock module.

here you can see the actual ROV:

Actual I do not know what is messed up in  the clock module.

Here I can give you the Callback functions:

void wakeupCb(UArg arg)
{
    unsigned short wakeupsource = (unsigned short) arg;

//    if( wakeupsource == WAKEUP_SOURCE_PORT )  // Wakeup by Port       //##pw 2017_07_03 anstatt generell für alle Ports die WS, wird nun einzeln unterschieden (je nach Key)
    if( wakeupsource == WAKEUP_SOURCE_SLEEPTIMER_APPL )    // Wakeup by Appl-Timer
    {
        // spezieller Code im Falle Wakeup durch App-Intervall
        // zuerst findet die Sonderbehandlung für ein sehr langes Intervall statt (>~11,9h), damit wir danach sichergehen können, dass das Intervall auch tatsächlich vollständig abgelaufen ist...
        UInt32 ticksFor1s = 1000000/Clock_tickPeriod;   // Zeit für 1s
        UInt16 maxSPeriod = ((UInt32)0xFFFFFFFF)/ticksFor1s;    // max. mögliche Anzahl von Sekunden die in einer Periode Platz finden (Achtung: max. liegt bei 42.949 Sekunden (~11,9h), theoretisch können aber 65.535 eingegeben werden...)

        if( SleepLoopMaxAppl > maxSPeriod ) // das App-Intervall ist länger als die max. mögl. Sekunden die in der Periode Platz finden (32bit Wertebereich, aber 1 Tick pro 10us!)
        {
            Clock_stop(hAppClock);
            if (SleepLoopMaxAppl >= SleepLoopCounterAppl)
            {
                if( (SleepLoopMaxAppl - SleepLoopCounterAppl) <= maxSPeriod )   // das war der letzte Durchlauf
                {
                    SleepLoopCounterAppl = 0;
                    Clock_setTimeout(hAppClock, maxSPeriod * ticksFor1s);
                }
                else    // es sind weitere Durchläufe nötig
                {
                    wakeupsource = 0;   // da gesamtes Intervall noch nicht abgelaufen ist, auch noch kein Wakeup-Handling
                    SleepLoopCounterAppl += maxSPeriod;
                    if( (SleepLoopMaxAppl - SleepLoopCounterAppl) < maxSPeriod )    // es bleibt nur mehr ein Rest (letzter Durchlauf bis Wakeup)
                    {
                        Clock_setTimeout(hAppClock, (SleepLoopMaxAppl - SleepLoopCounterAppl) * ticksFor1s);
                    }
                    else    // oder noch mind. eine ganze maxPeriode
                    {
                        Clock_setTimeout(hAppClock, maxSPeriod * ticksFor1s);
                    }
                }
                Clock_start(hAppClock);
            }
            else
            {
                System_printf("Callback overflow%x \n",WakeUpSource);
            }
        }
        if( wakeupsource )  // Zeug für Intervall nur machen, wenn es auch wirklich abgelaufen ist...
        {
            // Wichtig: hier darf nichts blockierendes gemacht werden (wir sind im IRQ-Handler!)
        }
    }
    else if( wakeupsource ==  WAKEUP_SOURCE_ULP)  // Wakeup by TT-Timer
    {
    }
    else if( wakeupsource ==  WAKEUP_SOURCE_SLEEPTIMER_PROTOCOLL)
    {
        ui8RF_State = RF_WAKEUP;
    }
    if( wakeupsource )  // durch Sonderbehandlung kann ev. noch verhindert werden, dass wirklich das Wakeup durchkommt
    {
        WakeUpSource |= wakeupsource;
#ifndef DEBUG_SLEEP
        //Task_getMode();
        Semaphore_post(hWakeupSem); // Tasks die auf diese wakeupSem warten, werden aufgeweckt
        //
#endif
#ifdef SYSTEM_PRT
        System_printf("Wakeup CB %x \n",WakeUpSource);
#endif
    }
}

void clk_CallBk(UArg arg)
{
    ui8RF_State = RF_WAKEUP;
    WakeUpSource = arg;
    Semaphore_post(hWakeupSem);
}

Any Ideas how to gon on?
BR Noge

Hi Noge, welcome back.

Unfortunately, it is very hard to say what could be wrong in your case. If you would be able to share a project that re-produces the issue so that I can debug it, I could possibly help you figuring this out a lot sooner (either here or in private).

As for the Raw clock view, what is the value of the "Clock module" variables (the once at the top, for example "nextScheduledTick") when you have the hiccup inside the walk function?

Also, as the callback is rather big, have you made sure that both the System stack and tasks stacks are intact? You can find info on the "System stack" if you look under "Hwi" in ROV and select "Raw" view.