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 <[email protected]>
---
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
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 <[email protected]>
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?
thanks,
Takashi
> ---
> 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
>
On Thu, Jan 28, 2021 at 4:09 PM Takashi Iwai <[email protected]> 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 <[email protected]>
>
> Kai-Heng, is this fix still needed for 5.11?
No it's not needed anymore because "ALSA: hda: Balance runtime/system
PM if direct-complete is disabled" is in place.
>
> The description mentions about HD-audio controller, while the recent
> revert was the HD-audio codec, so I suppose it's still affected?
Not affected anymore if above mentioned patch is applied.
Kai-Heng
>
>
> thanks,
>
> Takashi
>
> > ---
> > 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
> >
On Thu, 28 Jan 2021 09:15:10 +0100,
Kai-Heng Feng wrote:
>
> On Thu, Jan 28, 2021 at 4:09 PM Takashi Iwai <[email protected]> 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 <[email protected]>
> >
> > Kai-Heng, is this fix still needed for 5.11?
>
> No it's not needed anymore because "ALSA: hda: Balance runtime/system
> PM if direct-complete is disabled" is in place.
>
> >
> > The description mentions about HD-audio controller, while the recent
> > revert was the HD-audio codec, so I suppose it's still affected?
>
> Not affected anymore if above mentioned patch is applied.
OK, thanks for clarification!
Takashi
On Thu, Jan 28, 2021 at 9:11 AM Takashi Iwai <[email protected]> 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 <[email protected]>
>
> 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
> >