On Fri, 01 Sep 2017 13:09:36 +0200,
Andy Shevchenko wrote:
>
> On Thu, 2017-08-31 at 11:33 -0700, Dmitry Torokhov wrote:
>
> > > +
> > > + ret = enable_irq_wake(irq);
> > > + if (ret)
> > > + dev_warn(dev, "Can't enable IRQ as wake source:
> > > %d\n", ret);
> >
> > We do not normally enable wake IRQs in probe, but instead do:
> >
> > device_init_wakeup(&pdev->dev, true);
> >
> > in probe() and then check it in suspend/resume:
> >
> > if (device_may_wakeup(dev)) {
> > err = enable_irq_wake(XXX->irq);
> > if (!err)
> > XXX->irq_wake_enabled = true;
> > }
> >
> > ...
> >
> > if (XXX->irq_wake_enabled)
>
> No need to duplicate a flag which IRQ core already has.
>
> See, for example, commit
> aef3ad103a68 ("serial: core: remove unneeded irq_wake flag")
I couldn't find the commit, but checked the web search.
So everything is set up and done by dev_pm_set_wake_irq(), no need to
create the own PM ops just for that, right? That stuff was already in
v3 patch.
thanks,
Takashi
