Re: [linux-pm] pm_runtime_suspended() and non-pm_runtime-using (i2c) drivers

["Rafael J. Wysocki" <rjw-KKrjLPT3xs0@public.gmane.org>]

On Wednesday, December 15, 2010, Rafael J. Wysocki wrote:
> On Wednesday, December 15, 2010, Rafael J. Wysocki wrote:
> > On Tuesday, December 14, 2010, Alan Stern wrote:
> > > On Tue, 14 Dec 2010, Rabin Vincent wrote:
> > > 
> > > > Hello,
> > > > 
> > > > If an i2c driver uses dev_pm_ops and pm_runtime_suspended() returns true
> > > > for the device,  the i2c core will not call the driver's pm->suspend()
> > > > routine.  Similar behaviour (except for the if dev_pm_ops check) is
> > > > present in the generic PM ops provided in
> > > > drivers/base/power/generic_ops.c.
> > > > 
> > > > Since pm_runtime_suspended() returns true if the relevant driver did not
> > > > call any pm_runtime functions, this means that any driver which does not
> > > > use pm_runtime APIs will not get its pm->suspend() callback called
> > > > during system sleep, if CONFIG_PM_RUNTIME is enabled.
> > > > 
> > > > For the i2c case, there are several such drivers (in drivers/input/*,
> > > > etc) lacking these calls.  How is this to be handled?  Do all of these
> > > > drivers need to be patched to use the pm_runtime API if they are to be
> > > > used on a kernel with PM_RUNTIME enabled?
> > > 
> > > I'm not familiar with the details of how the i2c subsystem works.  But
> > > in general, the subsystem code should call pm_runtime_set_active()  
> > > for every device before registering it.  Then if a driver doesn't use
> > > any runtime-PM functions, pm_runtime_suspended() will return false.
> > 
> > I rather think that our current definition of pm_runtime_suspended() is
> > not really adequate.
> > 
> > Namely, it shouldn't really return true if runtime PM is not enabled
> > (ie. power.disable_depth > 1).
> 
> That should have been power.disable_depth > 0, of course.  Sorry.
> 
> > That change would fix the issue at hand,  wouldn't it?

Below is a patch.  I don't think it needs to be done under the lock at
the moment if the callers of pm_runtime_enable() always remember to set
the appropriate device status before calling it.

Thanks,
Rafael

---
From: Rafael J. Wysocki <rjw-KKrjLPT3xs0@public.gmane.org>
Subject: PM / Runtime: Fix pm_runtime_suspended()

pm_runtime_suspended() shouldn't return true if the runtime PM of the
given device is disabled.

Signed-off-by: Rafael J. Wysocki <rjw-KKrjLPT3xs0@public.gmane.org>
---
 include/linux/pm_runtime.h |    3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

Index: linux-2.6/include/linux/pm_runtime.h
===================================================================
--- linux-2.6.orig/include/linux/pm_runtime.h
+++ linux-2.6/include/linux/pm_runtime.h
@@ -78,7 +78,8 @@ static inline void device_set_run_wake(s
 
 static inline bool pm_runtime_suspended(struct device *dev)
 {
-	return dev->power.runtime_status == RPM_SUSPENDED;
+	return dev->power.runtime_status == RPM_SUSPENDED
+		&& !dev->power.disable_depth;
 }
 
 static inline void pm_runtime_mark_last_busy(struct device *dev)