Re: [Xenomai] rtdm_nrtsig_pend FAQ

[Johann Obermayr <johann.obermayr@sigmatek.at>]

Am 12.11.2014 um 17:44 schrieb Gilles Chanteperdrix:
> On Wed, Nov 12, 2014 at 09:47:45AM +0100, Johann Obermayr wrote:
>> Hello,
>>
>> we use rtdm_nrtsig_pend follow.
>>
>> static rtdm_nrtsig_t wait_domain_nrt;
>> static DECLARE_WAIT_QUEUE_HEAD(wait_domain_wq);
>>
>> void period_update()
>> {
>>          wait_domain_pend_count++;
>>          rtdm_nrtsig_pend(&wait_domain_nrt);
>> }
>>
>> static void wait_domain_nrt_handler(rtdm_nrtsig_t nrt_sig, void *arg)
>> {
>>      wait_domain_call_count++;
>>      //wakeup wait_domain
>>      wait_domain_wakeup = 1;
>>      wake_up_interruptible(&wait_domain_wq);
>> }
>>
>>
>> period_update is called every millisecond.
>> this is on imx6 hardware with kernel 3.0.x and xenomai 2.6.2.1.
> >From what Thierry said, I suspect the patch for this kernel has a
> missing call to irq_exit. So, I would ask you to do the same as
> Thierry, please enable the I-pipe tracer, set a sufficient number of
> back trace points such as 10000, and trigger a trace when you detect
> that you have "lost" a notification (wait_dommain_pend_count > wait_domain_call_count + 1)
>
> Then send the trace to the mailing list.
>
Hi,

witch function i must use to make a i-pipe trace ?
here i am in a xenomai irq handle

void period_update()
{
     ......
     else if (wait_domain_pend_count > 100000 && wait_domain_pend_count 
!= wait_domain_call_count) {
             // dotrace

      }
}

Thanks for help
  Johann