All of lore.kernel.org
 help / color / mirror / Atom feed
From: Robert Lee <lerobert@google.com>
To: Takashi Iwai <tiwai@suse.de>
Cc: vkoul@kernel.org, perex@perex.cz, tiwai@suse.com,
	alsa-devel@alsa-project.org, linux-kernel@vger.kernel.org,
	carterhsu@google.com, zxinhui@google.com, bubblefang@google.com
Subject: Re: [Patch v2] ALSA: compress: allow to leave draining state when pausing in draining
Date: Thu, 8 Jul 2021 21:47:24 +0800	[thread overview]
Message-ID: <CAOM6g_Cv6rsLOAb0+Lr_YkjHpKfw+zvWXH0X5LKR=Z4dtXGRng@mail.gmail.com> (raw)
In-Reply-To: <s5ho8bd59q4.wl-tiwai@suse.de>

Hi Takashi,

It is a little complex to describe the design in detail, but try to
explain simply
what issue we meet.

If w/o the change,  after user resumes from the pause, our system would call
snd_compr_drain() or snd_compr_partial_drain() again after it returns from
previous drain (when EOF reaches). Then it will block in this drain and no one
wake it up because EOF has already reached. I add this change to return from
the previous drain.

And yes, after user resumes it, it will change state to RUNNING. Then it will
call snd_compr_drain() or snd_compr_partial_drain() very soon and change
state to DRAINING again.

Actually, I am wondering how the pause-during-drain can keep the state in
DRAINING. It should have a different design. :)

I also checked the snd_compr_open() comment, and it doesn't mention that
we cannot pause in DRAINING state. Looks like it needs to be updated according
to these changes. Maybe it can be updated in another commit?

* SNDRV_PCM_STATE_DRAINING: When stream is draining current data. This is done
 *      by calling SNDRV_COMPRESS_DRAIN.
 * SNDRV_PCM_STATE_PAUSED: When stream is paused. This is done by calling
 *      SNDRV_COMPRESS_PAUSE. It can be stopped or resumed by calling
 *      SNDRV_COMPRESS_STOP or SNDRV_COMPRESS_RESUME respectively.

thanks,
Robert.

Takashi Iwai <tiwai@suse.de> 於 2021年7月8日 週四 下午7:24寫道:


Takashi Iwai <tiwai@suse.de> 於 2021年7月8日 週四 下午7:24寫道:
>
> On Thu, 08 Jul 2021 04:08:15 +0200,
> Robert Lee wrote:
> >
> > When compress offload pauses in draining state, not all platforms
> > need to keep in draining state. Some platforms may call drain or
> > partial drain again when resume from pause in draining, so it needs
> > to wake up from snd_compress_wait_for_drain() in this case.
> >
> > Call API snd_compr_leave_draining_in_pause(), if the platform
> > doesn't need to keep in draining state when pause in draining
> > state.
> >
> > Signed-off-by: Robert Lee <lerobert@google.com>
> > ---
> >  include/sound/compress_driver.h | 14 ++++++++++++++
> >  sound/core/compress_offload.c   |  8 +++++++-
> >  2 files changed, 21 insertions(+), 1 deletion(-)
> >
> > diff --git a/include/sound/compress_driver.h b/include/sound/compress_driver.h
> > index 277087f635f3..e16524a93a14 100644
> > --- a/include/sound/compress_driver.h
> > +++ b/include/sound/compress_driver.h
> > @@ -145,6 +145,7 @@ struct snd_compr_ops {
> >   * @lock: device lock
> >   * @device: device id
> >   * @use_pause_in_draining: allow pause in draining, true when set
> > + * @leave_draining_in_pause: leave draining state when pausing in draining
> >   */
> >  struct snd_compr {
> >       const char *name;
> > @@ -156,6 +157,7 @@ struct snd_compr {
> >       struct mutex lock;
> >       int device;
> >       bool use_pause_in_draining;
> > +     bool leave_draining_in_pause;
> >  #ifdef CONFIG_SND_VERBOSE_PROCFS
> >       /* private: */
> >       char id[64];
> > @@ -182,6 +184,18 @@ static inline void snd_compr_use_pause_in_draining(struct snd_compr_stream *subs
> >       substream->device->use_pause_in_draining = true;
> >  }
> >
> > +/**
> > + * snd_compr_leave_draining_in_pause - Leave draining state when pause in draining
> > + * @substream: compress substream to set
> > + *
> > + * In some platform, we need to leave draining state when we use pause in draining.
> > + * Add API to allow leave draining state.
> > + */
> > +static inline void snd_compr_leave_draining_in_pause(struct snd_compr_stream *substream)
> > +{
> > +     substream->device->leave_draining_in_pause = true;
> > +}
> > +
> >  /* dsp driver callback apis
> >   * For playback: driver should call snd_compress_fragment_elapsed() to let the
> >   * framework know that a fragment has been consumed from the ring buffer
> > diff --git a/sound/core/compress_offload.c b/sound/core/compress_offload.c
> > index 21ce4c056a92..c6e5c8f072d7 100644
> > --- a/sound/core/compress_offload.c
> > +++ b/sound/core/compress_offload.c
> > @@ -719,8 +719,14 @@ static int snd_compr_pause(struct snd_compr_stream *stream)
> >               if (!stream->device->use_pause_in_draining)
> >                       return -EPERM;
> >               retval = stream->ops->trigger(stream, SNDRV_PCM_TRIGGER_PAUSE_PUSH);
> > -             if (!retval)
> > +             if (!retval) {
> > +                     if (stream->device->leave_draining_in_pause) {
> > +                             stream->runtime->state = SNDRV_PCM_STATE_PAUSED;
> > +                             wake_up(&stream->runtime->sleep);
> > +                             break;
> > +                     }
> >                       stream->pause_in_draining = true;
> > +             }
>
> Hrm, what actually happens with this new flag?  It changes the state
> to PAUSED even if it's done during the draining.  Then user resumes
> the pause via snd_compr_resume(), and now the state changes to
> RUNNING.  OTOH, if the draining runs normally, it'll end up with
> SETUP.
>
> Even if the above is even designed behavior, it must be described
> properly somewhere.  The state change is described in snd_compr_open()
> comment, and the new behavior should be mentioned there as well.
> (Admittedly, the previous hack for the pause-during-drain is also
> missing and should have been mentioned there; but an excuse is that
> the pause-during-drain doesn't change the state itself :)
>
>
> thanks,
>
> Takashi

WARNING: multiple messages have this Message-ID (diff)
From: Robert Lee <lerobert@google.com>
To: Takashi Iwai <tiwai@suse.de>
Cc: alsa-devel@alsa-project.org, zxinhui@google.com,
	carterhsu@google.com, linux-kernel@vger.kernel.org,
	tiwai@suse.com, vkoul@kernel.org, bubblefang@google.com
Subject: Re: [Patch v2] ALSA: compress: allow to leave draining state when pausing in draining
Date: Thu, 8 Jul 2021 21:47:24 +0800	[thread overview]
Message-ID: <CAOM6g_Cv6rsLOAb0+Lr_YkjHpKfw+zvWXH0X5LKR=Z4dtXGRng@mail.gmail.com> (raw)
In-Reply-To: <s5ho8bd59q4.wl-tiwai@suse.de>

Hi Takashi,

It is a little complex to describe the design in detail, but try to
explain simply
what issue we meet.

If w/o the change,  after user resumes from the pause, our system would call
snd_compr_drain() or snd_compr_partial_drain() again after it returns from
previous drain (when EOF reaches). Then it will block in this drain and no one
wake it up because EOF has already reached. I add this change to return from
the previous drain.

And yes, after user resumes it, it will change state to RUNNING. Then it will
call snd_compr_drain() or snd_compr_partial_drain() very soon and change
state to DRAINING again.

Actually, I am wondering how the pause-during-drain can keep the state in
DRAINING. It should have a different design. :)

I also checked the snd_compr_open() comment, and it doesn't mention that
we cannot pause in DRAINING state. Looks like it needs to be updated according
to these changes. Maybe it can be updated in another commit?

* SNDRV_PCM_STATE_DRAINING: When stream is draining current data. This is done
 *      by calling SNDRV_COMPRESS_DRAIN.
 * SNDRV_PCM_STATE_PAUSED: When stream is paused. This is done by calling
 *      SNDRV_COMPRESS_PAUSE. It can be stopped or resumed by calling
 *      SNDRV_COMPRESS_STOP or SNDRV_COMPRESS_RESUME respectively.

thanks,
Robert.

Takashi Iwai <tiwai@suse.de> 於 2021年7月8日 週四 下午7:24寫道:


Takashi Iwai <tiwai@suse.de> 於 2021年7月8日 週四 下午7:24寫道:
>
> On Thu, 08 Jul 2021 04:08:15 +0200,
> Robert Lee wrote:
> >
> > When compress offload pauses in draining state, not all platforms
> > need to keep in draining state. Some platforms may call drain or
> > partial drain again when resume from pause in draining, so it needs
> > to wake up from snd_compress_wait_for_drain() in this case.
> >
> > Call API snd_compr_leave_draining_in_pause(), if the platform
> > doesn't need to keep in draining state when pause in draining
> > state.
> >
> > Signed-off-by: Robert Lee <lerobert@google.com>
> > ---
> >  include/sound/compress_driver.h | 14 ++++++++++++++
> >  sound/core/compress_offload.c   |  8 +++++++-
> >  2 files changed, 21 insertions(+), 1 deletion(-)
> >
> > diff --git a/include/sound/compress_driver.h b/include/sound/compress_driver.h
> > index 277087f635f3..e16524a93a14 100644
> > --- a/include/sound/compress_driver.h
> > +++ b/include/sound/compress_driver.h
> > @@ -145,6 +145,7 @@ struct snd_compr_ops {
> >   * @lock: device lock
> >   * @device: device id
> >   * @use_pause_in_draining: allow pause in draining, true when set
> > + * @leave_draining_in_pause: leave draining state when pausing in draining
> >   */
> >  struct snd_compr {
> >       const char *name;
> > @@ -156,6 +157,7 @@ struct snd_compr {
> >       struct mutex lock;
> >       int device;
> >       bool use_pause_in_draining;
> > +     bool leave_draining_in_pause;
> >  #ifdef CONFIG_SND_VERBOSE_PROCFS
> >       /* private: */
> >       char id[64];
> > @@ -182,6 +184,18 @@ static inline void snd_compr_use_pause_in_draining(struct snd_compr_stream *subs
> >       substream->device->use_pause_in_draining = true;
> >  }
> >
> > +/**
> > + * snd_compr_leave_draining_in_pause - Leave draining state when pause in draining
> > + * @substream: compress substream to set
> > + *
> > + * In some platform, we need to leave draining state when we use pause in draining.
> > + * Add API to allow leave draining state.
> > + */
> > +static inline void snd_compr_leave_draining_in_pause(struct snd_compr_stream *substream)
> > +{
> > +     substream->device->leave_draining_in_pause = true;
> > +}
> > +
> >  /* dsp driver callback apis
> >   * For playback: driver should call snd_compress_fragment_elapsed() to let the
> >   * framework know that a fragment has been consumed from the ring buffer
> > diff --git a/sound/core/compress_offload.c b/sound/core/compress_offload.c
> > index 21ce4c056a92..c6e5c8f072d7 100644
> > --- a/sound/core/compress_offload.c
> > +++ b/sound/core/compress_offload.c
> > @@ -719,8 +719,14 @@ static int snd_compr_pause(struct snd_compr_stream *stream)
> >               if (!stream->device->use_pause_in_draining)
> >                       return -EPERM;
> >               retval = stream->ops->trigger(stream, SNDRV_PCM_TRIGGER_PAUSE_PUSH);
> > -             if (!retval)
> > +             if (!retval) {
> > +                     if (stream->device->leave_draining_in_pause) {
> > +                             stream->runtime->state = SNDRV_PCM_STATE_PAUSED;
> > +                             wake_up(&stream->runtime->sleep);
> > +                             break;
> > +                     }
> >                       stream->pause_in_draining = true;
> > +             }
>
> Hrm, what actually happens with this new flag?  It changes the state
> to PAUSED even if it's done during the draining.  Then user resumes
> the pause via snd_compr_resume(), and now the state changes to
> RUNNING.  OTOH, if the draining runs normally, it'll end up with
> SETUP.
>
> Even if the above is even designed behavior, it must be described
> properly somewhere.  The state change is described in snd_compr_open()
> comment, and the new behavior should be mentioned there as well.
> (Admittedly, the previous hack for the pause-during-drain is also
> missing and should have been mentioned there; but an excuse is that
> the pause-during-drain doesn't change the state itself :)
>
>
> thanks,
>
> Takashi

  reply	other threads:[~2021-07-08 13:47 UTC|newest]

Thread overview: 12+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2021-07-08  2:08 [Patch v2] ALSA: compress: allow to leave draining state when pausing in draining Robert Lee
2021-07-08  2:08 ` Robert Lee
2021-07-08 11:24 ` Takashi Iwai
2021-07-08 11:24   ` Takashi Iwai
2021-07-08 13:47   ` Robert Lee [this message]
2021-07-08 13:47     ` Robert Lee
2021-07-08 14:53     ` Jaroslav Kysela
2021-07-08 14:53       ` Jaroslav Kysela
2021-07-09  2:08       ` Robert Lee
2021-07-09  2:08         ` Robert Lee
2021-07-09  7:47         ` Takashi Iwai
2021-07-09  7:47           ` Takashi Iwai

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to='CAOM6g_Cv6rsLOAb0+Lr_YkjHpKfw+zvWXH0X5LKR=Z4dtXGRng@mail.gmail.com' \
    --to=lerobert@google.com \
    --cc=alsa-devel@alsa-project.org \
    --cc=bubblefang@google.com \
    --cc=carterhsu@google.com \
    --cc=linux-kernel@vger.kernel.org \
    --cc=perex@perex.cz \
    --cc=tiwai@suse.com \
    --cc=tiwai@suse.de \
    --cc=vkoul@kernel.org \
    --cc=zxinhui@google.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.