[BUG RT] WARNING: at kernel/sched.c:5071 2.6.23-rc1-rt7

[BUG RT] WARNING: at kernel/sched.c:5071 2.6.23-rc1-rt7

[Steven Rostedt]

Ingo,

The below ifndef, shouldn't that be ifndef CONFIG_PREEMPT_SOFTIRQS ?
I hit that warning while I was running !PREEMPT_RT but with both hard
and softiqs as threads.


int __sched cond_resched_softirq(void)
{
#ifndef CONFIG_PREEMPT_RT
	WARN_ON_ONCE(!in_softirq());
#endif
	if (need_resched() && system_state == SYSTEM_RUNNING) {
		local_bh_enable();
		__cond_resched();
		local_bh_disable();
		return 1;
	}
	return 0;
}


-- Steve

P.S. I really found out that the system becomes VERY non-responsive when
you run with both hard and softirqs as threads, but with
PREEMPT_NONE ;-)

Re: [BUG RT] WARNING: at kernel/sched.c:5071 2.6.23-rc1-rt7

[Ingo Molnar]

* Steven Rostedt <rostedt@goodmis.org> wrote:

> Ingo,
> 
> The below ifndef, shouldn't that be ifndef CONFIG_PREEMPT_SOFTIRQS ? I 
> hit that warning while I was running !PREEMPT_RT but with both hard 
> and softiqs as threads.

yeah, indeed - fixed.

> P.S. I really found out that the system becomes VERY non-responsive 
> when you run with both hard and softirqs as threads, but with 
> PREEMPT_NONE ;-)

hm. That's not supposed to happen. Could there be some wakeup or softirq 
processing problem?

	Ingo

Re: [BUG RT] WARNING: at kernel/sched.c:5071 2.6.23-rc1-rt7

[Steven Rostedt]

On Sun, 2007-08-05 at 08:56 +0200, Ingo Molnar wrote:
> * Steven Rostedt <rostedt@goodmis.org> wrote:

> 
> > P.S. I really found out that the system becomes VERY non-responsive 
> > when you run with both hard and softirqs as threads, but with 
> > PREEMPT_NONE ;-)
> 
> hm. That's not supposed to happen. Could there be some wakeup or softirq 
> processing problem?
> 
> 	Ingo

Well, it happened right after I installed the ath driver (from
http://svn.madwifi.org/trunk ).  But I've recompiled my system with
PREEMPT_RT and it works fine now.  I can investigate it later, but I
currently need this system working with a PREEMPT_RT kernel.

-- Steve