_int_install_kernel_isr, some strange behavior

arnogir — Thu, 24 Aug 2017 13:08:08 GMT

Hello

I'm using a K60 on KDS3.2.0 with MQX 4.2.

To drive stepper motor, I use the Kernel ISR to instead of MQX ISR to have less latency as possible.

This work fine, but I have a strange behavior which lead me to ask some question.

First, this is my simplified code:

uint16_t   DV_Stepper_iFifoLevel[8];
uint16_t   DV_Stepper_iAllFifoLevelSum;

...

void MyInitFunction(void)
{
...
  _int_install_kernel_isr(INT_FTM0, DV_Stepper_iFTM0IRQ);
  _bsp_int_init(INT_FTM0, 1, 0, TRUE);

  _int_install_kernel_isr(INT_FTM3, DV_Stepper_iFTM3IRQ);
  _bsp_int_init(INT_FTM3, 1, 0, TRUE);

}

void MainTask(void)
{
....
  DV_Stepper_iFifoLevel[MotorID]++;
  DV_Stepper_iAllFifoLevelSum++;
....
}


/* Interrupt function ..... */

static void DV_Stepper_iFTM0IRQ(void)
{
  DV_Stepper_iFTMxIRQ(0);
}


static void DV_Stepper_iFTM3IRQ(void)
{
  DV_Stepper_iFTMxIRQ(3);
}



static inline void DV_Stepper_iFTMxIRQ(uint8_t FTMIndex)
{

  /* Chn+1 Interrupt */
  if ((DV_Stepper_FTMx_Cn1SC(FTMIndex, 0) & FTM_CnSC_CHF_MASK) != 0)
  {  
    DV_Stepper_iMotorStepControl(FTMIndex, 0);
    DV_Stepper_FTMx_Cn1SC(FTMIndex, 0) &= ~FTM_CnSC_CHF_MASK; /* clear flag */
  }
  /* Chn+1 Interrupt */
  if ((DV_Stepper_FTMx_Cn1SC(FTMIndex, 1) & FTM_CnSC_CHF_MASK) != 0)
  {
    DV_Stepper_iMotorStepControl(FTMIndex, 1);
    DV_Stepper_FTMx_Cn1SC(FTMIndex, 1) &= ~FTM_CnSC_CHF_MASK; /* clear flag */
  }
  /* Chn+1 Interrupt */
  if ((DV_Stepper_FTMx_Cn1SC(FTMIndex, 2) & FTM_CnSC_CHF_MASK) != 0)
  {
    DV_Stepper_iMotorStepControl(FTMIndex, 2);
    DV_Stepper_FTMx_Cn1SC(FTMIndex, 2) &= ~FTM_CnSC_CHF_MASK; /* clear flag */
  }
  /* Chn+1 Interrupt */
  if ((DV_Stepper_FTMx_Cn1SC(FTMIndex, 3) & FTM_CnSC_CHF_MASK) != 0)
  {
    DV_Stepper_iMotorStepControl(FTMIndex, 3);
    DV_Stepper_FTMx_Cn1SC(FTMIndex, 3) &= ~FTM_CnSC_CHF_MASK; /* clear flag */
  }  
}


static inline void DV_Stepper_iMotorStepControl(const uint8_t FTMIndex, const uint8_t ChannelPairIndex)
{
  const MotorAssocation[2][4] = 
  {
    /* FTM 0 */
    {0, 1, 2, 3},

    /* FTM 3 */
    {4, 5, 6, 7}
   };
  const uint8_t MotorID = MotorAssocation[FTMIndex][ChannelPairIndex];
  
  .....
  DV_Stepper_iFifoLevel[MotorID]--;
  DV_Stepper_iAllFifoLevelSum--;
  ......
}‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍

Like you can see, I have a table of counter DV_Stepper_iFifoLevel[8] and a global sum counter DV_Stepper_iAllFifoLevelSum.

Therorically, DV_Stepper_iFifoLevel[0] + DV_Stepper_iFifoLevel[1]+... DV_Stepper_iFifoLevel[7] is always equal to the DV_Stepper_iAllFifoLevelSum.

But sometime, at the end of my behavior, the sum is equal to 1 2 or 4 like each Fifo independant counter DV_Stepper_iFifoLevel[n] arr all to 0!

I can't explain this!.

-> Is a Kernel ISR can be interrupted itself?

-> do you see some case which could be explain this?

Thank

Re: _int_install_kernel_isr, some strange behavior

arnogir — Mon, 28 Aug 2017 11:30:29 GMT

In fact, if Main_Task is interrupted between counter increment, this nos cause problem.

But if interrupt occurs during increment (which take many assembler instruction) this cause problem.

So I disabled FTM interrupt before increment both counter, this resolve my problem.

Re: _int_install_kernel_isr, some strange behavior

danielchen — Tue, 29 Aug 2017 02:51:06 GMT

Thanks for sharing

MQX Software SolutionsのトピックRe: _int_install_kernel_isr, some strange behavior

_int_install_kernel_isr, some strange behavior

Re: _int_install_kernel_isr, some strange behavior

Re: _int_install_kernel_isr, some strange behavior