RTOS/CC2640R2F: Why sometimes my watchdog does not execute watchdog's callback and reset directly?

Hi Joakim,

The following is my code.

Please reference it.

static void SimpleBLEPeripheral_init(void)
{
  ...

  Board_initWatchdog();
  
  Watchdog_Params_init(&params);
  params.resetMode = Watchdog_RESET_OFF;
  params.callbackFxn = SimpleBLEPeripheral_watchDog_cb;
  hWdt = Watchdog_open(Board_WATCHDOG0, &params);
  if (hWdt == NULL) {
    // Error opening watchdog
    for (;;){}
  }
  // set timeout period to 1000 ms 
  wdtTicks = Watchdog_convertMsToTicks(hWdt, 1000);
  Watchdog_setReload(hWdt, wdtTicks);

  ...

}


static void SimpleBLEPeripheral_processAppMsg(sbpEvt_t *pMsg)
{
  ...

    case SBC_SS_RST_EVT:
      {
        RecoveryStatusStruct mStruct;
        
        if(osal_snv_read(0x83,sizeof(RecoveryStatusStruct),&mStruct)!=SUCCESS)
          mStruct.rstCounter = 0; // init reset counter
        
        // Store tag
        mStruct.rstTag = 0xA5;
        
        // Store oled is on or off
        if(BSP_OLED_Status()) // isOledOn
          mStruct.isOledOff = 0;
        else
          mStruct.isOledOff = 1;
        
        // Store timeticks
        mStruct.utcTimes = (uint32_t)UTC_getClock();
        
        // Store reset counter
        if(mStruct.rstCounter<127)
          mStruct.rstCounter++;
        else
          mStruct.rstCounter = 0;

#ifdef BK_USE_UART
        char txBuf[30] = {0};
        tiny_sprintf(txBuf, "Pre(hex):%02X,%08X,%02X,%01X\r\n", mStruct.rstTag, mStruct.utcTimes, mStruct.rstCounter, mStruct.isOledOff);
        logPrinter_send((uint8_t*)txBuf, strlen(txBuf));
#endif // BK_USE_UART
        
        if(osal_snv_write(0x83,sizeof(RecoveryStatusStruct),&mStruct)==SUCCESS)
        {
#ifdef BK_USE_UART
          char txBuf0[30] = {0};
          tiny_sprintf(txBuf0, "Pre(hex):Success to write.\r\n");
          logPrinter_send((uint8_t*)txBuf0, strlen(txBuf0));
#endif // BK_USE_UART
          BSP_SYSTEM_RESET();
        }
        else
        {
#ifdef BK_USE_UART
          char txBuf1[30] = {0};
          tiny_sprintf(txBuf1, "Pre(hex):Fail to write.\r\n");
          logPrinter_send((uint8_t*)txBuf1, strlen(txBuf1));
#endif // BK_USE_UART
        }

        Watchdog_clear(hWdt);
        SimpleBLEPeripheral_enqueueMsg(SBC_SS_RST_EVT, 0x00);
      }
      break;

    default:
      // Do nothing.
      break;
  }
}


void SimpleBLEPeripheral_watchDog_cb(uintptr_t handle)
{
  Watchdog_clear(hWdt);
  SimpleBLEPeripheral_enqueueMsg(SBC_SS_RST_EVT, 0x00);
}

Thanks.

Hi,

I'm not sure that you can call any RTOS or stack APIs in the watchdog callback function. Also since the watchdog interrupt is a non-maskable interrupt (NMI) it could occur at times that the regular code thinks it's protected and atomic.

If I were to guess, I'd say that something in the WDT NMI is causing a cpu fault, which would (per the arm docs) lead to a cpu lockup, and (according to the TI reference swcu117) a reset.

Luckily, you can read some registers on boot to detect what caused the reset to happen. It's a bit of a trick, basically we must do two things

* Disable Warm-reset-to-pin-reset, so that WDT/LOCKUP are tracked as reset sources
* Read out the Warmreset status bits before the regular reset functions clear the result

#include <ti/devices/cc26x0r2/inc/hw_prcm.h>

volatile uint32_t warmreset_stat __attribute__((noinit));
void extra_reset_fxn()
{
    warmreset_stat = HWREG(PRCM_BASE + PRCM_O_WARMRESET);
}

void *mainThread(void *arg0)
{
    Watchdog_Handle watchdogHandle;
    Watchdog_Params params;
    uint32_t        reloadValue;

    HWREG(PRCM_BASE + PRCM_O_WARMRESET) &= ~PRCM_WARMRESET_WR_TO_PINRESET_M;

Also in the RTOS config (.cfg) file, at the very top, this reset function must be added as the first to execute:

var Reset = xdc.useModule('xdc.runtime.Reset');
Reset.fxns[Reset.fxns.length++] = '&extra_reset_fxn';

It would be interesting to see, after a reset, what the value of warmreset_stat is. 0x01 would indicate WDT, 0x02 LOCKUP.

Best regards,
Aslak

Hi Aslak,

I have added the code you mentioned in my workspace like below.

@simple_peripheral.c

static void SimpleBLEPeripheral_init(void)
{
  ...
  HWREG(PRCM_BASE + PRCM_O_WARMRESET) &= ~PRCM_WARMRESET_WR_TO_PINRESET_M;
  ...
}

#include <ti/devices/cc26x0r2/inc/hw_prcm.h>
volatile uint32_t warmreset_stat __attribute__((noinit));
void extra_reset_fxn()
{
  warmreset_stat = HWREG(PRCM_BASE + PRCM_O_WARMRESET);
}

@source/ti/blestack/common/cc26xx/kernel/cc2640/config/cc2640_r2_csdk.cfg

/* Debug */
var Reset = xdc.useModule('xdc.runtime.Reset');
Reset.fxns[Reset.fxns.length++] = '&extra_reset_fxn';
...
// var Reset = xdc.useModule('xdc.runtime.Reset');

But the board cannot reset like before and I need to re-power on to reset the board, it like stuck in the reset process.

Then I try to mark the code of "Disable Warm-reset-to-pin-reset" and I check another register in the self-defined reset function(extra_reset_fxn()), the code like below:

@simple_peripheral.c

#include <ti/devices/cc26x0r2/inc/hw_prcm.h>
#include <ti/devices/cc26x0r2/inc/hw_aon_sysctl.h>

__no_init volatile uint32_t warmreset_stat;
__no_init volatile uint32_t reset_ctrl;
void extra_reset_fxn()
{
  warmreset_stat = HWREG(PRCM_BASE + PRCM_O_WARMRESET);
  reset_ctrl     = HWREG(AON_SYSCTL_BASE + AON_SYSCTL_O_RESETCTL);
}

static void SimpleBLEPeripheral_init(void)
{
  ...
//HWREG(PRCM_BASE + PRCM_O_WARMRESET) &= ~PRCM_WARMRESET_WR_TO_PINRESET_M;
  ...
}

I get the following log message.

---------------------Power On--------------------
17:29:35, [[Device Initial]]
17:29:35, Reg     : WARMRESET ,RESETCTL
17:29:35, Reg(hex): 0x00000004,0x00003AF2 --> RESET_SRC:1h = Reset pin
17:29:35, snv_read Fail!!
17:30:16, Pre:Success to snv_write.
---------------------Reboot--------------------
17:30:16, [[Device Initial]]
17:30:16, Reg(hex):0x00000004,0x00003AFC --> RESET_SRC:6h = Software reset via SYSRESET register
17:30:16, Rst:Success to snv_read.
17:31:03, Pre:Success to snv_write.
---------------------Reboot--------------------
...
---------------------Reboot--------------------
17:38:21, [[Device Initial]]
17:38:21, Reg(hex):0x00000004,0x00003AFC --> RESET_SRC:6h = Software reset via SYSRESET register
17:38:21, Rst:Success to snv_read.
                                         --> [Error] Didn't execute the code of storing the current status
---------------------Reboot--------------------
17:39:04, [[Device Initial]]
17:39:04, Reg(hex):0x00000004,0x00003AFE --> RESET_SRC:7h = Software reset via PRCM warm reset request
17:39:04, Rst:Success to snv_read.
17:39:43, Pre:Success to snv_write.
---------------------Reboot--------------------
17:39:43, [[Device Initial]]
17:39:43, Reg(hex):0x00000004,0x00003AFC --> RESET_SRC:6h = Software reset via SYSRESET register
17:39:43, Rst:Success to snv_read.
17:40:30, Pre:Success to snv_write.
---------------------Reboot--------------------
...

Why the data of WARMRESET register didn't been changed after reset by my watchdog?
Does it cause by that the resetMode of my watchdog is Watchdog_RESET_ON?

According to the above information, should we say that something in the WDT NMI is causing a CPU fault and a reset?
If yes, how can I do to prevent this error?

Hi Aslak,

In order to check that I can get the correct data of RESTCTL register,

I change the reset mode of the watchdog into "Watchdog_RESET_ON" and disable "Warm-reset-to-pin-reset".

Let the watchdog timeout. Then my device will halt in some process.

I get the data of RESTCTL register is 0x04 after power-cycle.

Why the data of RESTCTL register I read is 0x04 after power-cycle? (I think it should be 0 by default. (ref:swcu117.pdf, section 6.8.2.4.37))

Thanks.

Hi Aslak,

Thank you so much for your help.

I'm so sorry that I write the wrong register name in the last post.
And I actually want to say is the WARMRESET register, not RESTCTL register.

So my question is the following, 

In order to check that I can get the correct data of WARMRESET register,

I change the reset mode of the watchdog into "Watchdog_RESET_ON" and disable "Warm-reset-to-pin-reset".

Let the watchdog timeout. Then my device will halt in some process.

I get the data of WARMRESET register is 0x04 after power-cycle.

Why the data of WARMRESET register I read is 0x04 after power-cycle? (I think it should be 0 by default. (ref:swcu117.pdf, section 6.8.2.4.37))

Thanks.

Hi Aslak,

Yes, I have added the following code at the top of the cfg file,cc2640_r2_csdk.cfg.

 ...
 ******************************************************************************
 Release Name: simplelink_cc2640r2_sdk_1_30_00_25
 Release Date: 2017-03-02 20:08:35
 *****************************************************************************/

var Reset = xdc.useModule('xdc.runtime.Reset');
Reset.fxns[Reset.fxns.length++] = '&extra_reset_fxn';

/* ================ ROM configuration ================ */
/*
 * To use BIOS in flash, comment out the code block below.
 */
if (typeof NO_ROM == 'undefined' || (typeof NO_ROM != 'undefined' && NO_ROM == 0))
{
...

Is it the correct place to add codes?

Hi Aslak,

Because the project has not had enough time, the cause of this problem may not continue in a short time, I must find other ways to solve the problem.

I put the results of the experiment on my Evernote(link). I hope it could be useful for somebody.

Assuming that this abnormal restart is caused by CPU LOCKUP, is there any way to solve or avoid this problem?

Or is there any other way to keep some important parameters before the watchdog reboots the system?