Skip to contentSkip to conversation

ALSA: Realtek: Early Forbid of runtime PM

PATCH Superseded Harsha Priya 2 participants 3 messages Aug 27, 2020 Sound

v1 of 2 Superseded Latest is v2 →

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 ALSA: Realtek: Early Forbid of runtime PM

For Realtek codecs, pm_runtime_forbid() is called in the probe function after the build_controls(). In a stress test, occasionally the runtime PM calls are invoked before controls are built. This causes the codec to be runtime suspended before probe completes. Because of this, not all controls are enumerated correctly and audio does not work until system is rebooted.

This patch calls pm_runtime_forbid() early to fix the issue. Multiple stress tests of 2000+ cycles has been done to test the fix.

---
 sound/pci/hda/patch_realtek.c | 3 +++
 1 file changed, 3 insertions(+)
sound/pci/hda/patch_realtek.c +3 -0
index 98789691a479..30ea07fd6735 100644
@@ -8418,6 +8418,7 @@ static int patch_alc269(struct hda_codec *codec)
 {
 	struct alc_spec *spec;
 	int err;
+	struct device *dev = hda_codec_dev(codec);
 
 	err = alc_alloc_spec(codec, 0x0b);
 	if (err < 0)
@@ -8430,6 +8431,8 @@ static int patch_alc269(struct hda_codec *codec)
 #ifdef CONFIG_PM
 	codec->patch_ops.suspend = alc269_suspend;
 	codec->patch_ops.resume = alc269_resume;
+	pm_runtime_dont_use_autosuspend(dev);
+	pm_runtime_forbid(dev);
 #endif
 	spec->shutup = alc_default_shutup;
 	spec->init_hook = alc_default_init;
Signature · 1 line
2.17.1
#1 lore raw
Quoted 8 lines · Harsha Priya

For Realtek codecs, pm_runtime_forbid() is called in the probe function after the build_controls(). In a stress test, occasionally the runtime PM calls are invoked before controls are built. This causes the codec to be runtime suspended before probe completes. Because of this, not all controls are enumerated correctly and audio does not work until system is rebooted.

Quoted 6 lines

This patch calls pm_runtime_forbid() early to fix the issue. Multiple stress tests of 2000+ cycles has been done to test the fix.

The behavior shouldn't be specific to that model, also not to codec vendors, but it's rather a generic problem, so it's no right place to correct, I suppose.

Can we simply call pm_runtime_forbid() at creating a codec object like below?

thanks,

Takashi

sound/pci/hda/hda_codec.c +3 -0
@@ -1000,6 +1000,9 @@ int snd_hda_codec_device_new(struct hda_bus *bus, struct snd_card *card,
 	if (err < 0)
 		goto error;
 
+	/* PM runtime needs to be enabled later after binding codec */
+	pm_runtime_forbid(&codec->core.dev);
+
 	return 0;
 
  error:
#2 lore raw
Quoted 11 lines
Quoted 11 lines

For Realtek codecs, pm_runtime_forbid() is called in the probe function after the build_controls(). In a stress test, occasionally the runtime PM calls are invoked before controls are built. This causes the codec to be runtime suspended before probe completes. Because of this, not all controls are enumerated correctly and audio does not work until system is rebooted.

This patch calls pm_runtime_forbid() early to fix the issue. Multiple stress tests of 2000+ cycles has been done to test the fix.

Quoted 8 lines
Quoted 3 lines

The behavior shouldn't be specific to that model, also not to codec vendors, but it's rather a generic problem, so it's no right place to correct, I suppose.

Can we simply call pm_runtime_forbid() at creating a codec object like below?

We tried this change and its working as expected. I believe this will address similar issues across all codecs. Will send a patch with this fix.

Quoted 19 lines

thanks,

Takashi

sound/pci/hda/hda_codec.c +3 -0
@@ -1000,6 +1000,9 @@ int snd_hda_codec_device_new(struct hda_bus *bus,
 struct snd_card *card,
 	if (err < 0)
 		goto error;
 
+	/* PM runtime needs to be enabled later after binding codec */
+	pm_runtime_forbid(&codec->core.dev);
+
 	return 0;
 

error: