Re: [PATCH] PM: sleep: core: Resume suspended device if direct-complete is disabled

From: Rafael J. Wysocki
Date: Thu Jan 28 2021 - 05:25:00 EST


On Thu, Jan 28, 2021 at 9:11 AM Takashi Iwai <tiwai@xxxxxxx> wrote:
>
> On Thu, 31 Dec 2020 07:03:19 +0100,
> Kai-Heng Feng wrote:
> >
> > HDA controller can't be runtime-suspended after commit 215a22ed31a1
> > ("ALSA: hda: Refactor codjc PM to use direct-complete optimization"),
> > which enables direct-complete for HDA codec.
> >
> > The HDA codec driver doesn't expect direct-complete will be disabled
> > after it returns a positive value from prepare() callback. So freeze()
> > is called directly when it's runtime-suspended, breaks the balance of
> > its internal codec_powered counting.
> >
> > So if a device is prepared for direct-complete but PM core breaks the
> > assumption, resume the device to keep PM operations balanced.
> >
> > Fixes: 215a22ed31a1 ("ALSA: hda: Refactor codec PM to use direct-complete optimization")
> > Signed-off-by: Kai-Heng Feng <kai.heng.feng@xxxxxxxxxxxxx>
>
> Kai-Heng, is this fix still needed for 5.11?
>
> The description mentions about HD-audio controller, while the recent
> revert was the HD-audio codec, so I suppose it's still affected?

Regardless, this is not the right fix, because it doesn't take driver
flags into account AFAICS.

> > ---
> > drivers/base/power/main.c | 4 ++++
> > 1 file changed, 4 insertions(+)
> >
> > diff --git a/drivers/base/power/main.c b/drivers/base/power/main.c
> > index 46793276598d..9c0e25a92ad0 100644
> > --- a/drivers/base/power/main.c
> > +++ b/drivers/base/power/main.c
> > @@ -1849,6 +1849,10 @@ static int device_prepare(struct device *dev, pm_message_t state)
> > (ret > 0 || dev->power.no_pm_callbacks) &&
> > !dev_pm_test_driver_flags(dev, DPM_FLAG_NO_DIRECT_COMPLETE);
> > spin_unlock_irq(&dev->power.lock);
> > +
> > + if (ret > 0 && !dev->power.direct_complete)
> > + pm_runtime_resume(dev);
> > +
> > return 0;
> > }
> >
> > --
> > 2.29.2
> >