On Wed, Jan 16, 2013 at 12:14:01PM +0000, Philip, Avinash wrote: > On Mon, Jan 14, 2013 at 12:38:56, Thierry Reding wrote: > > On Thu, Jan 10, 2013 at 06:33:43PM +0530, Philip Avinash wrote: > > [...] > > > diff --git a/drivers/pwm/pwm-tiehrpwm.c b/drivers/pwm/pwm-tiehrpwm.c > > [...] > > > +static int ehrpwm_pwm_suspend(struct device *dev) > > > +{ > > > + struct ehrpwm_pwm_chip *pc = dev_get_drvdata(dev); > > > + > > > + ehrpwm_pwm_context_save(pc); > > > + pm_runtime_put_sync(dev); > > > + return 0; > > > +} > > > + > > > +static int ehrpwm_pwm_resume(struct device *dev) > > > +{ > > > + struct ehrpwm_pwm_chip *pc = dev_get_drvdata(dev); > > > + > > > + pm_runtime_get_sync(dev); > > > + ehrpwm_pwm_context_restore(pc); > > > + return 0; > > > +} > > > > According to Documentation/power/runtime_pm.txt, the PM core runs the > > pm_runtime_get_noresume() and pm_runtime_put_sync() before executing the > > .suspend() and .resume() callbacks. Are you sure you need to call them > > here explicitly? > > I understand the problem of calling pm_runtime_get_sync() from .resume(). > But this has to be called if pwm was running while going to suspend so that > pwm can continues to run after resume. Okay. I misread the documentation and/or your patch. The documentation says that the core calls pm_runtime_get_noresume() before executing the .suspend() callback so you're not in fact calling it twice. Sorry for the confusion. > So I will add check of test_bit(PWMF_ENABLED, &pwm->flags) before > pm_runtime_get/put_sync calls. Yes, that sounds reasonable. Thierry
Attachment:
pgplmLt103dCa.pgp
Description: PGP signature