Re: [PATCH] PM / sleep: fix unbalanced pm runtime disable in __device_suspend_late()

["Rafael J. Wysocki" <rjw@rjwysocki.net>]

On Friday, May 20, 2016 07:21:03 PM Grygorii Strashko wrote:
> On 05/20/2016 03:18 PM, Rafael J. Wysocki wrote:
> > On Thursday, May 19, 2016 08:11:34 PM Grygorii Strashko wrote:
> >> On 05/19/2016 04:38 PM, Rafael J. Wysocki wrote:
> >>> On Fri, May 13, 2016 at 8:03 PM, Grygorii Strashko
> >>> <grygorii.strashko@ti.com> wrote:
> >>>> The PM runtime will be left disabled for the device if its .suspend_late()
> >>>> callback fails and async suspend is not allowed for this device. In
> >>>> this case device will not be added in dpm_late_early_list and
> >>>> dpm_resume_early() will ignore this device, as result PM runtime will
> >>>> be disabled for it forever (side effect: after 8 subsequent failures
> >>>> for the same device the PM runtime will be reenabled due to
> >>>> disable_depth overflow).
> >>>>
> >>>> Hence, re-enable PM runtime in __device_suspend_late() if
> >>>> .suspend_late() callback fails and async suspend is not allowed for
> >>>> this device.
> >>>>
> >>>> Signed-off-by: Grygorii Strashko <grygorii.strashko@ti.com>
> >>>> ---
> >>>>    drivers/base/power/main.c | 7 +++++--
> >>>>    1 file changed, 5 insertions(+), 2 deletions(-)
> >>>>
> >>>> diff --git a/drivers/base/power/main.c b/drivers/base/power/main.c
> >>>> index 6e7c3cc..9b266e5 100644
> >>>> --- a/drivers/base/power/main.c
> >>>> +++ b/drivers/base/power/main.c
> >>>> @@ -1207,10 +1207,13 @@ static int __device_suspend_late(struct device *dev, pm_message_t state, bool as
> >>>>           }
> >>>>
> >>>>           error = dpm_run_callback(callback, dev, state, info);
> >>>> -       if (!error)
> >>>> +       if (!error) {
> >>>>                   dev->power.is_late_suspended = true;
> >>>> -       else
> >>>> +       } else {
> >> 		Point [1]
> >>>>                   async_error = error;
> >>>> +               if (!is_async(dev))
> >>>
> >>> Why is the is_async() check necessary here?
> >> 			
> >> A: deviceX is suspended *async* and reached point [1], in this case:
> >> - deviceX has been added in dpm_late_early_list already
> >> - dpm_suspend_late() will detect async_error and call dpm_resume_early()
> >> - dpm_resume_early() will call device_resume_early() for deviceX
> >> - device_resume_early() will re-enable PM runtime
> >> {
> >> ...
> >> 	if (!dev->power.is_late_suspended)
> >> 		goto Out;
> >>
> >> 	...
> >>   Out:
> >> 	TRACE_RESUME(error);
> >>
> >> 	pm_runtime_enable(dev);
> >> ^^^^^^^^^^^^
> >> 	complete_all(&dev->power.completion);
> >> 	return error;
> >> }	
> >> 	
> >>
> >> B: deviceX is suspended *sync* and reached point [1], in this case:
> >> - deviceX has not been added in dpm_late_early_list yet
> >> - dpm_suspend_late() will detect sync_error and call dpm_resume_early()
> >> - dpm_resume_early() will ignore deviceX
> >>
> >> if i'll not check for !is_async(dev) then pm_runtime_enable(dev)
> >> will be called twice for deviceX with this patch.
> > 
> > OK, thanks!
> > 
> > So to me, the problem is that we handle failures in that code inconsistently
> > depending on whether or not async suspend/resume is enabled for the device.
> > 
> > I'd rather make it consistent than add extra checks to it, so the patch below
> > is how I would fix this.
> > 
> > ---
> > From: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
> > Subject: [PATCH] PM / sleep: Handle failures in device_suspend_late() consistently
> > 
> > Grygorii Strashko reports:
> > 
> >   The PM runtime will be left disabled for the device if its
> >   .suspend_late() callback fails and async suspend is not allowed
> >   for this device. In this case device will not be added in
> >   dpm_late_early_list and dpm_resume_early() will ignore this
> >   device, as result PM runtime will be disabled for it forever
> >   (side effect: after 8 subsequent failures for the same device
> >   the PM runtime will be reenabled due to disable_depth overflow).
> > 
> > To fix this problem, add devices to dpm_late_early_list regardless
> > of whether or not device_suspend_late() returns errors for them.
> > 
> > That will ensure failures in there to be handled consistently for
> > all devices regardless of their async suspend/resume status.
> > 
> > Reported-by: Grygorii Strashko <grygorii.strashko@ti.com>
> > Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
> > ---
> >   drivers/base/power/main.c |    5 +++--
> >   1 file changed, 3 insertions(+), 2 deletions(-)
> > 
> > Index: linux-pm/drivers/base/power/main.c
> > ===================================================================
> > --- linux-pm.orig/drivers/base/power/main.c
> > +++ linux-pm/drivers/base/power/main.c
> > @@ -1267,14 +1267,15 @@ int dpm_suspend_late(pm_message_t state)
> >   		error = device_suspend_late(dev);
> >   
> >   		mutex_lock(&dpm_list_mtx);
> > +		if (!list_empty(&dev->power.entry))
> > +			list_move(&dev->power.entry, &dpm_late_early_list);
> > +
> >   		if (error) {
> >   			pm_dev_err(dev, state, " late", error);
> >   			dpm_save_failed_dev(dev_name(dev));
> >   			put_device(dev);
> >   			break;
> >   		}
> > -		if (!list_empty(&dev->power.entry))
> > -			list_move(&dev->power.entry, &dpm_late_early_list);
> >   		put_device(dev);
> >   
> >   		if (async_error)
> > 
> 
> Yep, it works too.
> Tested-by: Grygorii Strashko <grygorii.strashko@ti.com>

OK, thanks!

Applied.

> By the way, there is third option:)

Well, that at least would require a comment explaining what's going on.

> +++ b/drivers/base/power/main.c
> @@ -1211,8 +1211,7 @@ static int __device_suspend_late(struct device *dev, pm_message_t state, bool as
>                 dev->power.is_late_suspended = true;
>         } else {
>                 async_error = error;
> -               if (!is_async(dev))
> -                       pm_runtime_enable(dev);
> +               error = 0;
>         }

Thanks,
Rafael