Skip to contentSkip to conversation

ALSA: hda: tas2781: Fix missing setup at runtime PM

PATCH In review Takashi Iwai 2 participants 6 messages Oct 8, 2024 – Oct 9, 2024 Sound

lore
Share & tools
lore
#0 lore raw

PATCH 1/2 ALSA: hda: tas2781: Fix missing setup at runtime PM

tas2781_runtime_suspend() clears the playback_started flag, hence it leads to inconsistent state after runtime suspend is triggered, as if the device hasn't been opened yet. Also, the counterpart, alc2781_runtime_resume() doesn't call tasdevice_tuning_switch(), and this also causes the inconsistency.

This patch corrects the superfluous flag clearance and the missing call.

Fixes 5be27f1e3ec9 ("ALSA: hda/tas2781: Add tas2781 HDA driver")
---
 sound/pci/hda/tas2781_hda_i2c.c | 7 ++++---
 1 file changed, 4 insertions(+), 3 deletions(-)
sound/pci/hda/tas2781_hda_i2c.c +4 -3
index 370d847517f9..6d173b721fd0 100644
@@ -864,10 +864,8 @@ static int tas2781_runtime_suspend(struct device *dev)
 	/* The driver powers up the amplifiers at module load time.
 	 * Stop the playback if it's unused.
 	 */
-	if (tas_hda->priv->playback_started) {
+	if (tas_hda->priv->playback_started)
 		tasdevice_tuning_switch(tas_hda->priv, 1);
-		tas_hda->priv->playback_started = false;
-	}
 
 	mutex_unlock(&tas_hda->priv->codec_lock);
 
@@ -889,6 +887,9 @@ static int tas2781_runtime_resume(struct device *dev)
 	 */
 	tasdevice_apply_calibration(tas_hda->priv);
 
+	if (tas_hda->priv->playback_started)
+		tasdevice_tuning_switch(tas_hda->priv, 0);
+
 	mutex_unlock(&tas_hda->priv->codec_lock);
 
 	return 0;
Signature · 1 line
2.43.0
#1 lore raw

PATCH 2/2 ALSA: hda: tas2781: Simplify system PM with runtime PM

Now most of the calls in both system and runtime PMs are identical, use pm_runtime_force_suspend() and *_resume() for simplifying the code.

---
 sound/pci/hda/tas2781_hda_i2c.c | 27 ++-------------------------
 1 file changed, 2 insertions(+), 25 deletions(-)
sound/pci/hda/tas2781_hda_i2c.c +2 -25
index 6d173b721fd0..c35c676202b7 100644
@@ -901,19 +901,7 @@ static int tas2781_system_suspend(struct device *dev)
 
 	dev_dbg(tas_hda->priv->dev, "System Suspend\n");
 
-	mutex_lock(&tas_hda->priv->codec_lock);
-
-	/* Shutdown chip before system suspend */
-	if (tas_hda->priv->playback_started)
-		tasdevice_tuning_switch(tas_hda->priv, 1);
-
-	mutex_unlock(&tas_hda->priv->codec_lock);
-
-	/*
-	 * Reset GPIO may be shared, so cannot reset here.
-	 * However beyond this point, amps may be powered down.
-	 */
-	return 0;
+	return pm_runtime_force_suspend(dev);
 }
 
 static int tas2781_system_resume(struct device *dev)
@@ -924,26 +912,15 @@ static int tas2781_system_resume(struct device *dev)
 	dev_dbg(tas_hda->priv->dev, "System Resume\n");
 
 	mutex_lock(&tas_hda->priv->codec_lock);
-
 	for (i = 0; i < tas_hda->priv->ndev; i++) {
 		tas_hda->priv->tasdevice[i].cur_book = -1;
 		tas_hda->priv->tasdevice[i].cur_prog = -1;
 		tas_hda->priv->tasdevice[i].cur_conf = -1;
 	}
 	tasdevice_reset(tas_hda->priv);
-	tasdevice_prmg_load(tas_hda->priv, tas_hda->priv->cur_prog);
-
-	/* If calibrated data occurs error, dsp will still work with default
-	 * calibrated data inside algo.
-	 */
-	tasdevice_apply_calibration(tas_hda->priv);
-
-	if (tas_hda->priv->playback_started)
-		tasdevice_tuning_switch(tas_hda->priv, 0);
-
 	mutex_unlock(&tas_hda->priv->codec_lock);
 
-	return 0;
+	return pm_runtime_force_resume(dev);
 }
 
 static const struct dev_pm_ops tas2781_hda_pm_ops = {
Signature · 1 line
2.43.0
#2 lore raw
In reply to Takashi Iwai

Hi Takashi,

Quoted 21 lines · Takashi Iwai

tas2781_runtime_suspend() clears the playback_started flag, hence it leads to inconsistent state after runtime suspend is triggered, as if the device hasn't been opened yet. Also, the counterpart, alc2781_runtime_resume() doesn't call tasdevice_tuning_switch(), and this also causes the inconsistency.

This patch corrects the superfluous flag clearance and the missing call.

Fixes 5be27f1e3ec9 ("ALSA: hda/tas2781: Add tas2781 HDA driver")
---
 sound/pci/hda/tas2781_hda_i2c.c | 7 ++++---
 1 file changed, 4 insertions(+), 3 deletions(-)
sound/pci/hda/tas2781_hda_i2c.c +0 -0
index 370d847517f9..6d173b721fd0 100644

sound/pci/hda/tas2781_hda_i2c.c static int tas2781_runtime_suspend(struct device *dev) · line 869

Quoted 9 lines
sound/pci/hda/tas2781_hda_i2c.c +1 -3
@@ -864,10 +864,8 @@ static int tas2781_runtime_suspend(struct device *dev)
 	/* The driver powers up the amplifiers at module load time.
 	 * Stop the playback if it's unused.
 	 */
-	if (tas_hda->priv->playback_started) {
+	if (tas_hda->priv->playback_started)
 		tasdevice_tuning_switch(tas_hda->priv, 1);
-		tas_hda->priv->playback_started = false;
-	}
 

This if-branch were added because I wanted to power up the amplifier at module load time in tasdev_fw_ready() with:

tasdevice_tuning_switch(tas_hda->priv, 0);
tas_hda->priv->playback_started = true;

but I didn't want it to stay powered forever, if there aren't any tas2781_hda_playback_hook() call with HDA_GEN_PCM_ACT_CLOSE action.

This seemed like a good idea at the time because the music didn't have to be restarted (for HDA_GEN_PCM_ACT_OPEN action) when the module was reloaded. But this is a rare use case, today I don't think it's important to have it in there.

In general cases, tas2781_hda_playback_hook() turns the amplifier on and off, and sets the playback_started flag.
If I understand correctly, the playback_started will be off at the time of runtime_suspend()/runtime_resume() calls.

So I think the mentioned two lines from the tasdev_fw_ready() function and the mentioned if-branch can be deleted.

Thanks,
Gergo Koteles

#3 lore raw

aOn Wed, 09 Oct 2024 16:32:52 +0200,

Quoted 49 lines · Gergo Koteles

Hi Takashi,

Quoted 21 lines · Takashi Iwai

tas2781_runtime_suspend() clears the playback_started flag, hence it leads to inconsistent state after runtime suspend is triggered, as if the device hasn't been opened yet. Also, the counterpart, alc2781_runtime_resume() doesn't call tasdevice_tuning_switch(), and this also causes the inconsistency.

This patch corrects the superfluous flag clearance and the missing call.

Fixes 5be27f1e3ec9 ("ALSA: hda/tas2781: Add tas2781 HDA driver")
---
 sound/pci/hda/tas2781_hda_i2c.c | 7 ++++---
 1 file changed, 4 insertions(+), 3 deletions(-)
sound/pci/hda/tas2781_hda_i2c.c +0 -0
index 370d847517f9..6d173b721fd0 100644

sound/pci/hda/tas2781_hda_i2c.c static int tas2781_runtime_suspend(struct device *dev) · line 869

Quoted 9 lines
sound/pci/hda/tas2781_hda_i2c.c +1 -3
@@ -864,10 +864,8 @@ static int tas2781_runtime_suspend(struct device *dev)
 	/* The driver powers up the amplifiers at module load time.
 	 * Stop the playback if it's unused.
 	 */
-	if (tas_hda->priv->playback_started) {
+	if (tas_hda->priv->playback_started)
 		tasdevice_tuning_switch(tas_hda->priv, 1);
-		tas_hda->priv->playback_started = false;
-	}
 

This if-branch were added because I wanted to power up the amplifier at module load time in tasdev_fw_ready() with:

tasdevice_tuning_switch(tas_hda->priv, 0);
tas_hda->priv->playback_started = true;

but I didn't want it to stay powered forever, if there aren't any tas2781_hda_playback_hook() call with HDA_GEN_PCM_ACT_CLOSE action.

This seemed like a good idea at the time because the music didn't have to be restarted (for HDA_GEN_PCM_ACT_OPEN action) when the module was reloaded. But this is a rare use case, today I don't think it's important to have it in there.

Quoted 7 lines

In general cases, tas2781_hda_playback_hook() turns the amplifier on and off, and sets the playback_started flag.
If I understand correctly, the playback_started will be off at the time of runtime_suspend()/runtime_resume() calls.

So I think the mentioned two lines from the tasdev_fw_ready() function and the mentioned if-branch can be deleted.

Ah, right, there is a tweak of runtime PM there, which makes harder to track the state change...
Can we just drop the amp in tasdev_fw_ready()? Then the whole runtime suspend can be dropped. And, the conditional amp off/on can be moved again back to the system suspend/resume. That makes less confusing.

In anyway, it seems that my patches aren't really correct. Let's discard them at first.

thanks,

Takashi

#4 lore raw
Quoted 56 lines · Takashi Iwai

aOn Wed, 09 Oct 2024 16:32:52 +0200,

Quoted 54 lines · Gergo Koteles

Hi Takashi,

Quoted 21 lines · Takashi Iwai

tas2781_runtime_suspend() clears the playback_started flag, hence it leads to inconsistent state after runtime suspend is triggered, as if the device hasn't been opened yet. Also, the counterpart, alc2781_runtime_resume() doesn't call tasdevice_tuning_switch(), and this also causes the inconsistency.

This patch corrects the superfluous flag clearance and the missing call.

Fixes 5be27f1e3ec9 ("ALSA: hda/tas2781: Add tas2781 HDA driver")
---
 sound/pci/hda/tas2781_hda_i2c.c | 7 ++++---
 1 file changed, 4 insertions(+), 3 deletions(-)
sound/pci/hda/tas2781_hda_i2c.c +0 -0
index 370d847517f9..6d173b721fd0 100644

sound/pci/hda/tas2781_hda_i2c.c static int tas2781_runtime_suspend(struct device *dev) · line 869

Quoted 9 lines
sound/pci/hda/tas2781_hda_i2c.c +1 -3
@@ -864,10 +864,8 @@ static int tas2781_runtime_suspend(struct device *dev)
 	/* The driver powers up the amplifiers at module load time.
 	 * Stop the playback if it's unused.
 	 */
-	if (tas_hda->priv->playback_started) {
+	if (tas_hda->priv->playback_started)
 		tasdevice_tuning_switch(tas_hda->priv, 1);
-		tas_hda->priv->playback_started = false;
-	}
 

This if-branch were added because I wanted to power up the amplifier at module load time in tasdev_fw_ready() with:

tasdevice_tuning_switch(tas_hda->priv, 0);
tas_hda->priv->playback_started = true;

but I didn't want it to stay powered forever, if there aren't any tas2781_hda_playback_hook() call with HDA_GEN_PCM_ACT_CLOSE action.

This seemed like a good idea at the time because the music didn't have to be restarted (for HDA_GEN_PCM_ACT_OPEN action) when the module was reloaded. But this is a rare use case, today I don't think it's important to have it in there.

In general cases, tas2781_hda_playback_hook() turns the amplifier on and off, and sets the playback_started flag.
If I understand correctly, the playback_started will be off at the time of runtime_suspend()/runtime_resume() calls.

Quoted 7 lines
Quoted 2 lines

So I think the mentioned two lines from the tasdev_fw_ready() function and the mentioned if-branch can be deleted.

Ah, right, there is a tweak of runtime PM there, which makes harder to track the state change...
Can we just drop the amp in tasdev_fw_ready()?

I think, yes. I tested it with TAS2563 (Lenovo Yoga 7 14ARB7 Laptop).

Quoted 4 lines

Then the whole runtime
suspend can be dropped. And, the conditional amp off/on can be moved again back to the system suspend/resume. That makes less confusing.

The runtime_resume() checks whether "Speaker Program Id" kcontrol (as cur_prog variable) has changed with tasdevice_prmg_load() function. If so, updates it, so I don't think it can be dropped easily without moving this functionality somewhere else.

I'm not sure if the calibration data needs to be updated before each power-up or only after a program change.

Thanks,
Gergo Koteles

#5 lore raw
Quoted 73 lines · Gergo Koteles
Quoted 56 lines · Takashi Iwai

aOn Wed, 09 Oct 2024 16:32:52 +0200,

Quoted 54 lines · Gergo Koteles

Hi Takashi,

Quoted 21 lines · Takashi Iwai

tas2781_runtime_suspend() clears the playback_started flag, hence it leads to inconsistent state after runtime suspend is triggered, as if the device hasn't been opened yet. Also, the counterpart, alc2781_runtime_resume() doesn't call tasdevice_tuning_switch(), and this also causes the inconsistency.

This patch corrects the superfluous flag clearance and the missing call.

Fixes 5be27f1e3ec9 ("ALSA: hda/tas2781: Add tas2781 HDA driver")
---
 sound/pci/hda/tas2781_hda_i2c.c | 7 ++++---
 1 file changed, 4 insertions(+), 3 deletions(-)
sound/pci/hda/tas2781_hda_i2c.c +0 -0
index 370d847517f9..6d173b721fd0 100644

sound/pci/hda/tas2781_hda_i2c.c static int tas2781_runtime_suspend(struct device *dev) · line 869

Quoted 9 lines
sound/pci/hda/tas2781_hda_i2c.c +1 -3
@@ -864,10 +864,8 @@ static int tas2781_runtime_suspend(struct device *dev)
 	/* The driver powers up the amplifiers at module load time.
 	 * Stop the playback if it's unused.
 	 */
-	if (tas_hda->priv->playback_started) {
+	if (tas_hda->priv->playback_started)
 		tasdevice_tuning_switch(tas_hda->priv, 1);
-		tas_hda->priv->playback_started = false;
-	}
 

This if-branch were added because I wanted to power up the amplifier at module load time in tasdev_fw_ready() with:

tasdevice_tuning_switch(tas_hda->priv, 0);
tas_hda->priv->playback_started = true;

but I didn't want it to stay powered forever, if there aren't any tas2781_hda_playback_hook() call with HDA_GEN_PCM_ACT_CLOSE action.

This seemed like a good idea at the time because the music didn't have to be restarted (for HDA_GEN_PCM_ACT_OPEN action) when the module was reloaded. But this is a rare use case, today I don't think it's important to have it in there.

In general cases, tas2781_hda_playback_hook() turns the amplifier on and off, and sets the playback_started flag.
If I understand correctly, the playback_started will be off at the time of runtime_suspend()/runtime_resume() calls.

Quoted 7 lines
Quoted 2 lines

So I think the mentioned two lines from the tasdev_fw_ready() function and the mentioned if-branch can be deleted.

Ah, right, there is a tweak of runtime PM there, which makes harder to track the state change...
Can we just drop the amp in tasdev_fw_ready()?

I think, yes. I tested it with TAS2563 (Lenovo Yoga 7 14ARB7 Laptop).

Quoted 4 lines

Then the whole runtime
suspend can be dropped. And, the conditional amp off/on can be moved again back to the system suspend/resume. That makes less confusing.

Quoted 4 lines

The runtime_resume() checks whether "Speaker Program Id" kcontrol (as cur_prog variable) has changed with tasdevice_prmg_load() function. If so, updates it, so I don't think it can be dropped easily without moving this functionality somewhere else.

Yes, I meant to drop only runtime suspend, too.

Quoted 2 lines

I'm not sure if the calibration data needs to be updated before each power-up or only after a program change.

That leads to another issue: it seems that the speaker program id change doesn't influence on the behavior immediately, and it appears only after the runtime resume. It's a bit tricky, and better to be documented somewhere.

Takashi