How to identify the reason for a WDOG event

[tbiberdorf]

I'm working with the Kinetis K60 and I'm experiencing WDOG events happening and I cannot determine why its happening.  I cannot believe they are happening due to the WDOG not being refreshed within it's timeout period.

I have the WDOG IRQ enabled and I'm taking a snap shot of the WDOG timeout periods and they appear to be well below the timeout limits.

I my case I've extended the timeout period to an extreme level to verify.

My timeout period is set to 10 seconds with a 1 second task refreshing the WDOG.

My IRQ posting is giving the following values:

WDOG_TMROUTH = 0x92

WDOG_TMROUTL = 0x7ce4

Here is my WDOG init method:

void wdtInit()

{

    set_irq_priority (INT_Watchdog-16, 0);

    enable_irq(INT_Watchdog-16) ; 

    wdtUnlock();

    WDOG_TOVALH = 0x1c9c;

    WDOG_TOVALL = 0x3800; /* 10 second period */

    WDOG_WINH = 100;

    WDOG_WINL = 100;

    /* WDOG_STCTRLH: DISTESTWDOG=0,BYTESEL=0,TESTSEL=0,TESTWDOG=0,

     * WAITEN=1,STOPEN=1,DBGEN=1,ALLOWUPDATE=1,WINEN=0

     * ,IRQRSTEN=1,CLKSRC=0,WDOGEN=1 */

    WDOG_STCTRLH = WDOG_STCTRLH_BYTESEL(0x00) |

            WDOG_STCTRLH_WAITEN_MASK |

            WDOG_STCTRLH_STOPEN_MASK |

            WDOG_STCTRLH_DBGEN_MASK  |

            WDOG_STCTRLH_ALLOWUPDATE_MASK |

            WDOG_STCTRLH_IRQRSTEN_MASK |

            WDOG_STCTRLH_CLKSRC_MASK |

            WDOG_STCTRLH_WDOGEN_MASK |

            0x0100U;  

}

After the WDOG event triggers, I do examine the reset registers:

RCM_SRS0 = 0x20

RCM_SRS1 = 0x0

and these indicate that the cause was a watchdog timeout, but with my TMROUTH and TMROUTL values do not indicate that refresh time has expired.

Can there be another reason or hidden setting that I'm missing?

Thanks

Terry

[tbiberdorf · Accepted solution]

I believe the cause of the WDOG events turns out to be how I managed the WDOG_REFRESH operations.

I failed to read the fine print details that the two write operations must be completed within 20 bus cycles.

I have many IRQs running, which could easily exceed the 20 bus cycle delay.

By adding a Disable/Enable interrupts operation around my WDOG_REFRESH calls, the WDOG resets stopped happening.