On Thu 2016-02-25 13:59:32, Peter Zijlstra wrote:
> On Wed, Feb 24, 2016 at 05:18:05PM +0100, Petr Mladek wrote:
> > @@ -770,7 +782,22 @@ void delayed_kthread_work_timer_fn(unsigned long
> > __data)
> > if (WARN_ON_ONCE(!worker))
> > return;
> >
> > - spin_lock(&worker->lock);
> > + /*
> > + * We might be unable to take the lock if someone is trying to
> > + * cancel this work and calls del_timer_sync() when this callback
> > + * has already been removed from the timer list.
> > + */
> > + while (!spin_trylock(&worker->lock)) {
> > + /*
> > + * Busy wait with spin_is_locked() to avoid cache bouncing.
> > + * Break when canceling is set to avoid a deadlock.
> > + */
> > + do {
> > + if (work->canceling)
> > + return;
> > + cpu_relax();
> > + } while (spin_is_locked(&worker->lock));
> > + }
> > /* Work must not be used with more workers, see queue_kthread_work(). */
> > WARN_ON_ONCE(work->worker != worker);
> >
>
> This is pretty vile; why can't you drop the lock over del_timer_sync() ?
We would need to take the lock later and check if nobody has set the timer
again in the meantime.
Now, timer_active() check is not reliable. It does not check if the
timer handler is running at the moment. I tried to implement
a safe timer_active()[1] but nobody was keen to take it.
Even if we have that timer_active() check, we would need to add
some relock/check/try_again stuff around the del_timer_sync().
So, it would just move the complexity somewhere else.
I think that the current solution is quite elegant after all.
Thanks a lot Tejun for the idea.
Reference:
[1] http://thread.gmane.org/gmane.linux.kernel.mm/144964/focus=144965
Thanks a lot for review,
Petr
