Skip to contentSkip to conversation

ASoC: soc-pcm: BE dai needs prepare when pause release after resume

PATCH Superseded [email protected] 2 participants 4 messages May 5, 2019 – May 7, 2019

Subsystems (2)

v1 of 2 Superseded Latest is v2 →

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 ASoC: soc-pcm: BE dai needs prepare when pause release after resume

From: Libin Yang <[email protected]>

If playback/capture is paused and system enters S3, after system returns from suspend, BE dai needs to call prepare() callback when playback/capture is released from pause.

---
 sound/soc/soc-pcm.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)
sound/soc/soc-pcm.c +2 -1
index d2aa560..0888995 100644
@@ -2471,7 +2471,8 @@ int dpcm_be_dai_prepare(struct snd_soc_pcm_runtime *fe, int stream)
 
 		if ((be->dpcm[stream].state != SND_SOC_DPCM_STATE_HW_PARAMS) &&
 		    (be->dpcm[stream].state != SND_SOC_DPCM_STATE_STOP) &&
-		    (be->dpcm[stream].state != SND_SOC_DPCM_STATE_SUSPEND))
+		    (be->dpcm[stream].state != SND_SOC_DPCM_STATE_SUSPEND) &&
+		    (be->dpcm[stream].state != SND_SOC_DPCM_STATE_PAUSED))
 			continue;
 
 		dev_dbg(be->dev, "ASoC: prepare BE %s\n",
Signature · 1 line
2.7.4
#1 lore raw
Quoted 5 lines · On 5/5/19 1:37 AM, [email protected] wrote:

From: Libin Yang <[email protected]>

If playback/capture is paused and system enters S3, after system returns from suspend, BE dai needs to call prepare() callback when playback/capture is released from pause.

Libin, this patch was discussed at length on GitHub [1] and the commit message does not convey any of the information we looked into. It's not very helpful to send such patches to a larger audience without explaining context and goals. I personally still have no idea of the state machine and if all solutions need this or if this is only needed in the case where the RESUME_INFO flag is not set.

[1] https://github.com/thesofproject/linux/pull/868

Quoted 21 lines
---
  sound/soc/soc-pcm.c | 3 ++-
  1 file changed, 2 insertions(+), 1 deletion(-)
sound/soc/soc-pcm.c +2 -1
index d2aa560..0888995 100644
@@ -2471,7 +2471,8 @@ int dpcm_be_dai_prepare(struct snd_soc_pcm_runtime *fe, int stream)
 
  		if ((be->dpcm[stream].state != SND_SOC_DPCM_STATE_HW_PARAMS) &&
  		    (be->dpcm[stream].state != SND_SOC_DPCM_STATE_STOP) &&
-		    (be->dpcm[stream].state != SND_SOC_DPCM_STATE_SUSPEND))
+		    (be->dpcm[stream].state != SND_SOC_DPCM_STATE_SUSPEND) &&
+		    (be->dpcm[stream].state != SND_SOC_DPCM_STATE_PAUSED))
  			continue;
 
  		dev_dbg(be->dev, "ASoC: prepare BE %s\n",
#2 lore raw

Hi Pierre,

Quoted 13 lines

-----Original Message-----
From: Pierre-Louis Bossart [mailto:[email protected]]
Sent: Monday, May 6, 2019 11:08 PM
To: Yang, Libin <[email protected]>; [email protected]; [email protected]
Cc: Sridharan, Ranjani <[email protected]>; Wang, Rander <[email protected]>
Subject: Re: [alsa-devel] [PATCH] ASoC: soc-pcm: BE dai needs prepare when pause release after resume

Quoted 2 lines · On 5/5/19 1:37 AM, [email protected] wrote:

From: Libin Yang <[email protected]>

Quoted 7 lines
Quoted 3 lines

If playback/capture is paused and system enters S3, after system returns from suspend, BE dai needs to call prepare() callback when playback/capture is released from pause.

Libin, this patch was discussed at length on GitHub [1] and the commit message does not convey any of the information we looked into. It's not very helpful to send such patches to a larger audience without explaining context

Thanks for pointing it out. I will add more context in the patch description.

Quoted 3 lines

and goals. I personally still have no idea of the state machine and if all solutions need this or if this is only needed in the case where the RESUME_INFO flag is not set.

Let me describe the bug for the reviewers who doesn't know the details.

The patch is trying to fix the below bug:
Playback -> pause -> suspend -> resume -> pause release.

When pause release, playback should continue to play. However we found playback will quit abnormally.

This is because we should restore the registers after S3. Currently RESUME_INFO flag is not set, so alsa will call prepare() callback. But as you can see, dpcm_be_dai_prepare() will block calling prepare() because it lacks of this patch. This means prepare() of BE will not be called, which is wrong.

So this patch adds the judgement to let the driver call the prepare() if the state is SND_SOC_DPCM_STATE_PAUSED. And the state is SND_SOC_DPCM_STATE_PAUSED because it is paused. Although S3 happens before pause release, but the state is still in SND_SOC_DPCM_STATE_PAUSED based on the code of dpcm_be_dai_trigger().

If RESUME_INFO flag is not set, prepare() will not be called by alsa after S3. So the patch is not necessary.

Regards,
Libin

Quoted 31 lines

[1] https://github.com/thesofproject/linux/pull/868

Quoted 11 lines
---
  sound/soc/soc-pcm.c | 3 ++-
  1 file changed, 2 insertions(+), 1 deletion(-)
sound/soc/soc-pcm.c index renamed from sound/soc/soc-pcm.c +0 -0

d2aa560..0888995 100644

sound/soc/soc-pcm.c int dpcm_be_dai_prepare(struct · line 2473

Quoted 3 lines
sound/soc/soc-pcm.c +0 -0
@@ -2471,7 +2471,8 @@ int dpcm_be_dai_prepare(struct
 snd_soc_pcm_runtime *fe, int stream)
 
  		if ((be->dpcm[stream].state !=

SND_SOC_DPCM_STATE_HW_PARAMS) &&

sound/soc/soc-pcm.c int dpcm_be_dai_prepare(struct · line 2474

Quoted 1 line
sound/soc/soc-pcm.c +0 -0
  		    (be->dpcm[stream].state != SND_SOC_DPCM_STATE_STOP)

&&

sound/soc/soc-pcm.c int dpcm_be_dai_prepare(struct · line 2475

Quoted 1 line
sound/soc/soc-pcm.c +0 -1
-		    (be->dpcm[stream].state !=

SND_SOC_DPCM_STATE_SUSPEND))

sound/soc/soc-pcm.c int dpcm_be_dai_prepare(struct · line 2475

Quoted 1 line
sound/soc/soc-pcm.c +1 -0
+		    (be->dpcm[stream].state !=

SND_SOC_DPCM_STATE_SUSPEND) &&

sound/soc/soc-pcm.c int dpcm_be_dai_prepare(struct · line 2476

Quoted 1 line
sound/soc/soc-pcm.c +1 -0
+		    (be->dpcm[stream].state !=

SND_SOC_DPCM_STATE_PAUSED))

Quoted 4 lines
sound/soc/soc-pcm.c +0 -0
  			continue;
 

dev_dbg(be->dev, "ASoC: prepare BE %s\n",

#3 lore raw
In reply to Pierre-Louis Bossart
Quoted 56 lines

-----Original Message-----
From: Yang, Libin
Sent: Tuesday, May 7, 2019 4:15 PM
To: 'Pierre-Louis Bossart' <[email protected]>; alsa- [email protected]; [email protected]
Cc: Sridharan, Ranjani <[email protected]>; Wang, Rander <[email protected]>
Subject: RE: [alsa-devel] [PATCH] ASoC: soc-pcm: BE dai needs prepare when pause release after resume

Hi Pierre,

Quoted 14 lines

-----Original Message-----
From: Pierre-Louis Bossart [mailto:[email protected]]
Sent: Monday, May 6, 2019 11:08 PM
To: Yang, Libin <[email protected]>; [email protected]; [email protected]
Cc: Sridharan, Ranjani <[email protected]>; Wang, Rander <[email protected]>
Subject: Re: [alsa-devel] [PATCH] ASoC: soc-pcm: BE dai needs prepare when pause release after resume

Quoted 2 lines · On 5/5/19 1:37 AM, [email protected] wrote:

From: Libin Yang <[email protected]>

Quoted 8 lines
Quoted 3 lines

If playback/capture is paused and system enters S3, after system returns from suspend, BE dai needs to call prepare() callback when playback/capture is released from pause.

Libin, this patch was discussed at length on GitHub [1] and the commit message does not convey any of the information we looked into. It's not very helpful to send such patches to a larger audience without explaining context

Thanks for pointing it out. I will add more context in the patch description.

Quoted 3 lines

and goals. I personally still have no idea of the state machine and if all solutions need this or if this is only needed in the case where the RESUME_INFO flag is not set.

Let me describe the bug for the reviewers who doesn't know the details.

The patch is trying to fix the below bug:
Playback -> pause -> suspend -> resume -> pause release.

When pause release, playback should continue to play. However we found playback will quit abnormally.

This is because we should restore the registers after S3.
Currently RESUME_INFO flag is not set, so alsa will call prepare() callback. But as you can see, dpcm_be_dai_prepare() will block calling prepare() because it lacks of this patch. This means prepare() of BE will not be called, which is wrong.

Quoted 8 lines

So this patch adds the judgement to let the driver call the prepare() if the state is SND_SOC_DPCM_STATE_PAUSED. And the state is SND_SOC_DPCM_STATE_PAUSED because it is paused. Although
S3 happens before pause release, but the state is still in SND_SOC_DPCM_STATE_PAUSED based on the code of dpcm_be_dai_trigger().

If RESUME_INFO flag is not set, prepare() will not be called by alsa after S3. So the patch is not necessary.

Typo: " If RESUME_INFO flag is not set " should be "If RESUME_INFO flag is set"

Regards,
Libin

Quoted 35 lines

Regards,
Libin

Quoted 31 lines

[1] https://github.com/thesofproject/linux/pull/868

Quoted 11 lines
---
  sound/soc/soc-pcm.c | 3 ++-
  1 file changed, 2 insertions(+), 1 deletion(-)
sound/soc/soc-pcm.c index renamed from sound/soc/soc-pcm.c +0 -0

d2aa560..0888995 100644

sound/soc/soc-pcm.c int dpcm_be_dai_prepare(struct · line 2473

Quoted 3 lines
sound/soc/soc-pcm.c +0 -0
@@ -2471,7 +2471,8 @@ int dpcm_be_dai_prepare(struct
 snd_soc_pcm_runtime *fe, int stream)
 
  		if ((be->dpcm[stream].state !=

SND_SOC_DPCM_STATE_HW_PARAMS) &&

sound/soc/soc-pcm.c int dpcm_be_dai_prepare(struct · line 2474

Quoted 1 line
sound/soc/soc-pcm.c +0 -0
  		    (be->dpcm[stream].state != SND_SOC_DPCM_STATE_STOP)

&&

sound/soc/soc-pcm.c int dpcm_be_dai_prepare(struct · line 2475

Quoted 1 line
sound/soc/soc-pcm.c +0 -1
-		    (be->dpcm[stream].state !=

SND_SOC_DPCM_STATE_SUSPEND))

sound/soc/soc-pcm.c int dpcm_be_dai_prepare(struct · line 2475

Quoted 1 line
sound/soc/soc-pcm.c +1 -0
+		    (be->dpcm[stream].state !=

SND_SOC_DPCM_STATE_SUSPEND) &&

sound/soc/soc-pcm.c int dpcm_be_dai_prepare(struct · line 2476

Quoted 1 line
sound/soc/soc-pcm.c +1 -0
+		    (be->dpcm[stream].state !=

SND_SOC_DPCM_STATE_PAUSED))

Quoted 4 lines
sound/soc/soc-pcm.c +0 -0
  			continue;
 

dev_dbg(be->dev, "ASoC: prepare BE %s\n",