* [PATCH] PM-runtime: fix deadlock when canceling hrtimer
@ 2019-02-21 7:59 Vincent Guittot
2019-02-21 9:35 ` Rafael J. Wysocki
0 siblings, 1 reply; 3+ messages in thread
From: Vincent Guittot @ 2019-02-21 7:59 UTC (permalink / raw)
To: linux-pm, linux-kernel, rjw, ulf.hansson; +Cc: sunzhaosheng, Vincent Guittot
When rpm_resume() desactivates the autosuspend timer, it should only try
to cancel hrtimer but not wait for the handler to finish because both
rpm_resume() and pm_suspend_timer_fn() are taking the power.lock.
We can have the deadlock sequence:
CPU0 CPU1
rpm_resume()
spin_lock_irqsave
pm_suspend_timer_fn()
spin_lock_irqsave
pm_runtime_deactivate_timer()
hrtimer_cancel()
hrtimer_try_to_cancel() is enough because dev->power.timer_expires is also
set to 0.
Fixes: 8234f6734c5d ("PM-runtime: Switch autosuspend over to using hrtimers")
Reported-by: Sunzhaosheng Sun(Zhaosheng) <sunzhaosheng@hisilicon.com>
Signed-off-by: Vincent Guittot <vincent.guittot@linaro.org>
---
drivers/base/power/runtime.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/drivers/base/power/runtime.c b/drivers/base/power/runtime.c
index 04407d9..78937c4 100644
--- a/drivers/base/power/runtime.c
+++ b/drivers/base/power/runtime.c
@@ -121,7 +121,7 @@ EXPORT_SYMBOL_GPL(pm_runtime_suspended_time);
static void pm_runtime_deactivate_timer(struct device *dev)
{
if (dev->power.timer_expires > 0) {
- hrtimer_cancel(&dev->power.suspend_timer);
+ hrtimer_try_to_cancel(&dev->power.suspend_timer);
dev->power.timer_expires = 0;
}
}
--
2.7.4
^ permalink raw reply related [flat|nested] 3+ messages in thread* Re: [PATCH] PM-runtime: fix deadlock when canceling hrtimer
2019-02-21 7:59 [PATCH] PM-runtime: fix deadlock when canceling hrtimer Vincent Guittot
@ 2019-02-21 9:35 ` Rafael J. Wysocki
2019-02-21 10:02 ` Vincent Guittot
0 siblings, 1 reply; 3+ messages in thread
From: Rafael J. Wysocki @ 2019-02-21 9:35 UTC (permalink / raw)
To: Vincent Guittot
Cc: Linux PM, Linux Kernel Mailing List, Rafael J. Wysocki,
Ulf Hansson, Zhaosheng Sun
On Thu, Feb 21, 2019 at 8:59 AM Vincent Guittot
<vincent.guittot@linaro.org> wrote:
>
> When rpm_resume() desactivates the autosuspend timer, it should only try
> to cancel hrtimer but not wait for the handler to finish because both
> rpm_resume() and pm_suspend_timer_fn() are taking the power.lock.
> We can have the deadlock sequence:
> CPU0 CPU1
> rpm_resume()
> spin_lock_irqsave
> pm_suspend_timer_fn()
> spin_lock_irqsave
> pm_runtime_deactivate_timer()
> hrtimer_cancel()
Everybody has missed this, not good.
> hrtimer_try_to_cancel() is enough because dev->power.timer_expires is also
> set to 0.
>
> Fixes: 8234f6734c5d ("PM-runtime: Switch autosuspend over to using hrtimers")
> Reported-by: Sunzhaosheng Sun(Zhaosheng) <sunzhaosheng@hisilicon.com>
> Signed-off-by: Vincent Guittot <vincent.guittot@linaro.org>
> ---
> drivers/base/power/runtime.c | 2 +-
> 1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/drivers/base/power/runtime.c b/drivers/base/power/runtime.c
> index 04407d9..78937c4 100644
> --- a/drivers/base/power/runtime.c
> +++ b/drivers/base/power/runtime.c
> @@ -121,7 +121,7 @@ EXPORT_SYMBOL_GPL(pm_runtime_suspended_time);
> static void pm_runtime_deactivate_timer(struct device *dev)
> {
> if (dev->power.timer_expires > 0) {
> - hrtimer_cancel(&dev->power.suspend_timer);
> + hrtimer_try_to_cancel(&dev->power.suspend_timer);
> dev->power.timer_expires = 0;
> }
> }
> --
Applied with some changelog modifications, thanks!
^ permalink raw reply [flat|nested] 3+ messages in thread* Re: [PATCH] PM-runtime: fix deadlock when canceling hrtimer
2019-02-21 9:35 ` Rafael J. Wysocki
@ 2019-02-21 10:02 ` Vincent Guittot
0 siblings, 0 replies; 3+ messages in thread
From: Vincent Guittot @ 2019-02-21 10:02 UTC (permalink / raw)
To: Rafael J. Wysocki
Cc: Linux PM, Linux Kernel Mailing List, Rafael J. Wysocki,
Ulf Hansson, Zhaosheng Sun
On Thu, 21 Feb 2019 at 10:35, Rafael J. Wysocki <rafael@kernel.org> wrote:
>
> On Thu, Feb 21, 2019 at 8:59 AM Vincent Guittot
> <vincent.guittot@linaro.org> wrote:
> >
> > When rpm_resume() desactivates the autosuspend timer, it should only try
> > to cancel hrtimer but not wait for the handler to finish because both
> > rpm_resume() and pm_suspend_timer_fn() are taking the power.lock.
> > We can have the deadlock sequence:
> > CPU0 CPU1
> > rpm_resume()
> > spin_lock_irqsave
> > pm_suspend_timer_fn()
> > spin_lock_irqsave
> > pm_runtime_deactivate_timer()
> > hrtimer_cancel()
>
> Everybody has missed this, not good.
yes. default behavior if different between timer and hrtimer
del_timer() is equivalent to hrtimer_try_to_cancel()
and hrtimer_cancel() to del_timer_sync()
>
> > hrtimer_try_to_cancel() is enough because dev->power.timer_expires is also
> > set to 0.
> >
> > Fixes: 8234f6734c5d ("PM-runtime: Switch autosuspend over to using hrtimers")
> > Reported-by: Sunzhaosheng Sun(Zhaosheng) <sunzhaosheng@hisilicon.com>
> > Signed-off-by: Vincent Guittot <vincent.guittot@linaro.org>
> > ---
> > drivers/base/power/runtime.c | 2 +-
> > 1 file changed, 1 insertion(+), 1 deletion(-)
> >
> > diff --git a/drivers/base/power/runtime.c b/drivers/base/power/runtime.c
> > index 04407d9..78937c4 100644
> > --- a/drivers/base/power/runtime.c
> > +++ b/drivers/base/power/runtime.c
> > @@ -121,7 +121,7 @@ EXPORT_SYMBOL_GPL(pm_runtime_suspended_time);
> > static void pm_runtime_deactivate_timer(struct device *dev)
> > {
> > if (dev->power.timer_expires > 0) {
> > - hrtimer_cancel(&dev->power.suspend_timer);
> > + hrtimer_try_to_cancel(&dev->power.suspend_timer);
> > dev->power.timer_expires = 0;
> > }
> > }
> > --
>
> Applied with some changelog modifications, thanks!
Thanks
^ permalink raw reply [flat|nested] 3+ messages in thread
end of thread, other threads:[~2019-02-21 10:02 UTC | newest]
Thread overview: 3+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2019-02-21 7:59 [PATCH] PM-runtime: fix deadlock when canceling hrtimer Vincent Guittot
2019-02-21 9:35 ` Rafael J. Wysocki
2019-02-21 10:02 ` Vincent Guittot
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).