So with rt_timer_set_periodic(125000) on and my sound task running, I
still trip the watchdog. /proc/xenomai/stat is showing 9.5% usage by the
sound task and a whopping 16.5% by IRQ0. (This amount seems considerably
lower when watchdog support is not compiled in.) The weird thing is that
the watchdog keeps killing the thread even though it isn't at all
locking up the system.  (ROOT using 82%.)

 

I changed my code to use this:

 

if(rt_task_wait_period(0)) rt_task_wait_period(0);

 

This was intended to catch the -ETIMEDOUT return value in case it did
time out. I figure this maybe was causing the hard lockup I was seeing
before.

 

Gilles's patch did work for me regarding the rt_timer_set_mode()
problem.

 

It seems that for now, I'll have to disable the watchdog as I see this
behavior with periodic mode or aperiodic mode. (My watchdog timeout is
set for 1 second.)

 

Steven

 

_______________________________________________
Xenomai-help mailing list
[email protected]
https://mail.gna.org/listinfo/xenomai-help

Reply via email to