Re: [patch 5/8] hrtimer remove state field

[Thomas Gleixner]

On Sun, 2006-03-12 at 16:17 +0100, Roman Zippel wrote:
> > 
> > CPU 0
> > spin_unlock_irq(base->lock)	
> > 			CPU1
> > 			signal is dequeued
> > 				timer is requeued
> > 			user space code is executed
> > 			user space code sets SIG_IGN
> > restart = fn();
> > 
> > Now fn() calls send_sigqeue() which returns 1, resulting in ret =
> > HRTIMER_RESTART which leads to requeueing of an enqueued timer.
> 
> I'm not quite sure I follow, when the timer is running no signal should be 
> queued, so nothing can be dequeued and no new timer can be requeued.
> If that somehow is possible (although I don't see how), you'd found a bug 
> in the signal/posix timer code, which should not be worked around in the 
> hrtimer run queue.

How do you want to prevent that a signal is dequeued on one CPU while
the softirq expires the timer on another CPU ? This can not be
prevented.

Of course we can check hrtimer_active() in posix_timer_fn(), but I have
to check if there is a problem the other way round. I cook up a patch.

	tglx



-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/