Skip to contentSkip to conversation

Realtek model parser problem with the new jack detection

DISCUSSION David Henningsson 2 participants 5 messages Jan 18, 2012 – Jan 19, 2012 Sound

lore
Share & tools
lore
#0 lore raw

Hi Takashi,

I'm troubleshooting a problem with the Realtek quirk/model parsers. The problem is basically that add_jack_kctls call snd_hda_jack_detect_enable, which in turn overwrites the current unsol tag.

E g, first the model parser sets a pin to AC_USRSP_EN | ALC_HP_EVENT, then comes snd_hda_jack_detect_enable and does AC_USRSP_EN | jack->tag.

I'm not sure of the best way to resolve this; either keep the current tag (and mark all jacks as dirty on an unsol event), or read back the current tag and set it as jack->action (and then always use the jack table). What do you think?

Example alsa-info for this is in
http://www.alsa-project.org/db/?f=5cc88967e9da547975924a06c948ece00f1f0263

Signature · 2 lines
David Henningsson, Canonical Ltd.
http://launchpad.net/~diwic
#1 lore raw

At Wed, 18 Jan 2012 17:28:32 +0100,

Quoted 7 lines · David Henningsson

Hi Takashi,

I'm troubleshooting a problem with the Realtek quirk/model parsers. The problem is basically that add_jack_kctls call snd_hda_jack_detect_enable, which in turn overwrites the current unsol tag.

Quoted 7 lines

E g, first the model parser sets a pin to AC_USRSP_EN | ALC_HP_EVENT, then comes snd_hda_jack_detect_enable and does AC_USRSP_EN | jack->tag.

I'm not sure of the best way to resolve this; either keep the current tag (and mark all jacks as dirty on an unsol event), or read back the current tag and set it as jack->action (and then always use the jack table). What do you think?

I think a safer way is to avoid calling snd_hda_jack_add_kctls() for non-auto-parser, and let quirks use its own unsol_event handler instead of alc_sku_unsol_event(). snd_hda_jack_detect() itself works both with or without jack-kctls.

I'll fix up the upstream tree.

thanks,

Takashi

#2 lore raw

At Thu, 19 Jan 2012 11:35:23 +0100,

Quoted 18 lines · Takashi Iwai

At Wed, 18 Jan 2012 17:28:32 +0100,

Quoted 7 lines · David Henningsson

Hi Takashi,

I'm troubleshooting a problem with the Realtek quirk/model parsers. The problem is basically that add_jack_kctls call snd_hda_jack_detect_enable, which in turn overwrites the current unsol tag.

Quoted 7 lines

E g, first the model parser sets a pin to AC_USRSP_EN | ALC_HP_EVENT, then comes snd_hda_jack_detect_enable and does AC_USRSP_EN | jack->tag.

I'm not sure of the best way to resolve this; either keep the current tag (and mark all jacks as dirty on an unsol event), or read back the current tag and set it as jack->action (and then always use the jack table). What do you think?

Quoted 6 lines

I think a safer way is to avoid calling snd_hda_jack_add_kctls() for non-auto-parser, and let quirks use its own unsol_event handler instead of alc_sku_unsol_event(). snd_hda_jack_detect() itself works both with or without jack-kctls.

I'll fix up the upstream tree.

The patch below.

Takashi

---
From: Takashi Iwai <[email protected]>
Subject: [PATCH] ALSA: hda/realtek - Avoid conflict of unsol-events with
static quirks

The recently added jack-kctl support sets the unsol event tags dynamically, while static quirks usually set the fixed tags in the init_verbs array. Due to this conflict, the own unsol event handler can't retrieve the tag and handle it properly any more.

For fixing this, avoid calling snd_hda_jack_add_kctls() for static quirks, and always let them use own handlers instead of the standard one for the auto-pareser.

---
 sound/pci/hda/alc880_quirks.c |   17 +++++++++++----
 sound/pci/hda/alc882_quirks.c |   15 +++++++++----
 sound/pci/hda/patch_realtek.c |   45 ++++++++++++++++++++++++++++------------
 3 files changed, 53 insertions(+), 24 deletions(-)
sound/pci/hda/alc880_quirks.c +12 -5
index 5b68435..501501e 100644
@@ -762,16 +762,22 @@ static void alc880_uniwill_unsol_event(struct hda_codec *codec,
 	/* Looks like the unsol event is incompatible with the standard
 	 * definition.  4bit tag is placed at 28 bit!
 	 */
-	switch (res >> 28) {
+	res >>= 28;
+	switch (res) {
 	case ALC_MIC_EVENT:
 		alc88x_simple_mic_automute(codec);
 		break;
 	default:
-		alc_sku_unsol_event(codec, res);
+		alc_exec_unsol_event(codec, res);
 		break;
 	}
 }
 
+static void alc880_unsol_event(struct hda_codec *codec, unsigned int res)
+{
+	alc_exec_unsol_event(codec, res >> 28);
+}
+
 static void alc880_uniwill_p53_setup(struct hda_codec *codec)
 {
 	struct alc_spec *spec = codec->spec;
@@ -800,10 +806,11 @@ static void alc880_uniwill_p53_unsol_event(struct hda_codec *codec,
 	/* Looks like the unsol event is incompatible with the standard
 	 * definition.  4bit tag is placed at 28 bit!
 	 */
-	if ((res >> 28) == ALC_DCVOL_EVENT)
+	res >>= 28;
+	if (res == ALC_DCVOL_EVENT)
 		alc880_uniwill_p53_dcvol_automute(codec);
 	else
-		alc_sku_unsol_event(codec, res);
+		alc_exec_unsol_event(codec, res);
 }
 
 /*
@@ -1677,7 +1684,7 @@ static const struct alc_config_preset alc880_presets[] = {
 		.channel_mode = alc880_lg_ch_modes,
 		.need_dac_fix = 1,
 		.input_mux = &alc880_lg_capture_source,
-		.unsol_event = alc_sku_unsol_event,
+		.unsol_event = alc880_unsol_event,
 		.setup = alc880_lg_setup,
 		.init_hook = alc_hp_automute,
 #ifdef CONFIG_SND_HDA_POWER_SAVE
sound/pci/hda/alc882_quirks.c +10 -5
index bdf0ed4..bb364a5 100644
@@ -730,6 +730,11 @@ static void alc889A_mb31_unsol_event(struct hda_codec *codec, unsigned int res)
 		alc889A_mb31_automute(codec);
 }
 
+static void alc882_unsol_event(struct hda_codec *codec, unsigned int res)
+{
+	alc_exec_unsol_event(codec, res >> 26);
+}
+
 /*
  * configuration and preset
  */
@@ -775,7 +780,7 @@ static const struct alc_config_preset alc882_presets[] = {
 			.channel_mode = alc885_mba21_ch_modes,
 			.num_channel_mode = ARRAY_SIZE(alc885_mba21_ch_modes),
 			.input_mux = &alc882_capture_source,
-			.unsol_event = alc_sku_unsol_event,
+			.unsol_event = alc882_unsol_event,
 			.setup = alc885_mba21_setup,
 			.init_hook = alc_hp_automute,
        },
@@ -791,7 +796,7 @@ static const struct alc_config_preset alc882_presets[] = {
 		.input_mux = &alc882_capture_source,
 		.dig_out_nid = ALC882_DIGOUT_NID,
 		.dig_in_nid = ALC882_DIGIN_NID,
-		.unsol_event = alc_sku_unsol_event,
+		.unsol_event = alc882_unsol_event,
 		.setup = alc885_mbp3_setup,
 		.init_hook = alc_hp_automute,
 	},
@@ -806,7 +811,7 @@ static const struct alc_config_preset alc882_presets[] = {
 		.input_mux = &mb5_capture_source,
 		.dig_out_nid = ALC882_DIGOUT_NID,
 		.dig_in_nid = ALC882_DIGIN_NID,
-		.unsol_event = alc_sku_unsol_event,
+		.unsol_event = alc882_unsol_event,
 		.setup = alc885_mb5_setup,
 		.init_hook = alc_hp_automute,
 	},
@@ -821,7 +826,7 @@ static const struct alc_config_preset alc882_presets[] = {
 		.input_mux = &macmini3_capture_source,
 		.dig_out_nid = ALC882_DIGOUT_NID,
 		.dig_in_nid = ALC882_DIGIN_NID,
-		.unsol_event = alc_sku_unsol_event,
+		.unsol_event = alc882_unsol_event,
 		.setup = alc885_macmini3_setup,
 		.init_hook = alc_hp_automute,
 	},
@@ -836,7 +841,7 @@ static const struct alc_config_preset alc882_presets[] = {
 		.input_mux = &alc889A_imac91_capture_source,
 		.dig_out_nid = ALC882_DIGOUT_NID,
 		.dig_in_nid = ALC882_DIGIN_NID,
-		.unsol_event = alc_sku_unsol_event,
+		.unsol_event = alc882_unsol_event,
 		.setup = alc885_imac91_setup,
 		.init_hook = alc_hp_automute,
 	},
sound/pci/hda/patch_realtek.c +31 -14

90 lines not on this page · raw message

Signature · 1 line
1.7.8.3
#3 lore raw
Quoted 20 lines · Takashi Iwai

At Thu, 19 Jan 2012 11:35:23 +0100,

Quoted 18 lines · Takashi Iwai

At Wed, 18 Jan 2012 17:28:32 +0100,

Quoted 14 lines · David Henningsson

Hi Takashi,

I'm troubleshooting a problem with the Realtek quirk/model parsers. The problem is basically that add_jack_kctls call snd_hda_jack_detect_enable, which in turn overwrites the current unsol tag.

E g, first the model parser sets a pin to AC_USRSP_EN | ALC_HP_EVENT, then comes snd_hda_jack_detect_enable and does AC_USRSP_EN | jack->tag.

I'm not sure of the best way to resolve this; either keep the current tag (and mark all jacks as dirty on an unsol event), or read back the current tag and set it as jack->action (and then always use the jack table). What do you think?

Quoted 8 lines
Quoted 6 lines

I think a safer way is to avoid calling snd_hda_jack_add_kctls() for non-auto-parser, and let quirks use its own unsol_event handler instead of alc_sku_unsol_event(). snd_hda_jack_detect() itself works both with or without jack-kctls.

I'll fix up the upstream tree.

The patch below.

Thanks for quick handling. This will mean no jack kctls for model parsers, which is a bit sad, but hopefully its an incentive for people to move to auto-parsers.

One review comment below:

Quoted 168 lines

Takashi

---
From: Takashi Iwai<[email protected]>
Subject: [PATCH] ALSA: hda/realtek - Avoid conflict of unsol-events with
static quirks

The recently added jack-kctl support sets the unsol event tags dynamically, while static quirks usually set the fixed tags in the init_verbs array. Due to this conflict, the own unsol event handler can't retrieve the tag and handle it properly any more.

For fixing this, avoid calling snd_hda_jack_add_kctls() for static quirks, and always let them use own handlers instead of the standard one for the auto-pareser.

---
  sound/pci/hda/alc880_quirks.c |   17 +++++++++++----
  sound/pci/hda/alc882_quirks.c |   15 +++++++++----
  sound/pci/hda/patch_realtek.c |   45 ++++++++++++++++++++++++++++------------
  3 files changed, 53 insertions(+), 24 deletions(-)
sound/pci/hda/alc880_quirks.c +12 -5
index 5b68435..501501e 100644
@@ -762,16 +762,22 @@ static void alc880_uniwill_unsol_event(struct hda_codec *codec,
  	/* Looks like the unsol event is incompatible with the standard
  	 * definition.  4bit tag is placed at 28 bit!
  	 */
-	switch (res>>  28) {
+	res>>= 28;
+	switch (res) {
  	case ALC_MIC_EVENT:
  		alc88x_simple_mic_automute(codec);
  		break;
  	default:
-		alc_sku_unsol_event(codec, res);
+		alc_exec_unsol_event(codec, res);
  		break;
  	}
  }
 
+static void alc880_unsol_event(struct hda_codec *codec, unsigned int res)
+{
+	alc_exec_unsol_event(codec, res>>  28);
+}
+
  static void alc880_uniwill_p53_setup(struct hda_codec *codec)
  {
  	struct alc_spec *spec = codec->spec;
@@ -800,10 +806,11 @@ static void alc880_uniwill_p53_unsol_event(struct hda_codec *codec,
  	/* Looks like the unsol event is incompatible with the standard
  	 * definition.  4bit tag is placed at 28 bit!
  	 */
-	if ((res>>  28) == ALC_DCVOL_EVENT)
+	res>>= 28;
+	if (res == ALC_DCVOL_EVENT)
  		alc880_uniwill_p53_dcvol_automute(codec);
  	else
-		alc_sku_unsol_event(codec, res);
+		alc_exec_unsol_event(codec, res);
  }
 
  /*
@@ -1677,7 +1684,7 @@ static const struct alc_config_preset alc880_presets[] = {
  		.channel_mode = alc880_lg_ch_modes,
  		.need_dac_fix = 1,
  		.input_mux =&alc880_lg_capture_source,
-		.unsol_event = alc_sku_unsol_event,
+		.unsol_event = alc880_unsol_event,
  		.setup = alc880_lg_setup,
  		.init_hook = alc_hp_automute,
  #ifdef CONFIG_SND_HDA_POWER_SAVE
sound/pci/hda/alc882_quirks.c +10 -5
index bdf0ed4..bb364a5 100644
@@ -730,6 +730,11 @@ static void alc889A_mb31_unsol_event(struct hda_codec *codec, unsigned int res)
  		alc889A_mb31_automute(codec);
  }
 
+static void alc882_unsol_event(struct hda_codec *codec, unsigned int res)
+{
+	alc_exec_unsol_event(codec, res>>  26);
+}
+
  /*
   * configuration and preset
   */
@@ -775,7 +780,7 @@ static const struct alc_config_preset alc882_presets[] = {
  			.channel_mode = alc885_mba21_ch_modes,
  			.num_channel_mode = ARRAY_SIZE(alc885_mba21_ch_modes),
  			.input_mux =&alc882_capture_source,
-			.unsol_event = alc_sku_unsol_event,
+			.unsol_event = alc882_unsol_event,
  			.setup = alc885_mba21_setup,
  			.init_hook = alc_hp_automute,
         },
@@ -791,7 +796,7 @@ static const struct alc_config_preset alc882_presets[] = {
  		.input_mux =&alc882_capture_source,
  		.dig_out_nid = ALC882_DIGOUT_NID,
  		.dig_in_nid = ALC882_DIGIN_NID,
-		.unsol_event = alc_sku_unsol_event,
+		.unsol_event = alc882_unsol_event,
  		.setup = alc885_mbp3_setup,
  		.init_hook = alc_hp_automute,
  	},
@@ -806,7 +811,7 @@ static const struct alc_config_preset alc882_presets[] = {
  		.input_mux =&mb5_capture_source,
  		.dig_out_nid = ALC882_DIGOUT_NID,
  		.dig_in_nid = ALC882_DIGIN_NID,
-		.unsol_event = alc_sku_unsol_event,
+		.unsol_event = alc882_unsol_event,
  		.setup = alc885_mb5_setup,
  		.init_hook = alc_hp_automute,
  	},
@@ -821,7 +826,7 @@ static const struct alc_config_preset alc882_presets[] = {
  		.input_mux =&macmini3_capture_source,
  		.dig_out_nid = ALC882_DIGOUT_NID,
  		.dig_in_nid = ALC882_DIGIN_NID,
-		.unsol_event = alc_sku_unsol_event,
+		.unsol_event = alc882_unsol_event,
  		.setup = alc885_macmini3_setup,
  		.init_hook = alc_hp_automute,
  	},
@@ -836,7 +841,7 @@ static const struct alc_config_preset alc882_presets[] = {
  		.input_mux =&alc889A_imac91_capture_source,
  		.dig_out_nid = ALC882_DIGOUT_NID,
  		.dig_in_nid = ALC882_DIGIN_NID,
-		.unsol_event = alc_sku_unsol_event,
+		.unsol_event = alc882_unsol_event,
  		.setup = alc885_imac91_setup,
  		.init_hook = alc_hp_automute,
  	},
sound/pci/hda/patch_realtek.c +3 -10
index 61ccbe8..2326bf3 100644
@@ -621,17 +621,10 @@ static void alc_mic_automute(struct hda_codec *codec)
  		alc_mux_select(codec, 0, spec->int_mic_idx, false);
  }
 
-/* unsolicited event for HP jack sensing */
-static void alc_sku_unsol_event(struct hda_codec *codec, unsigned int res)
+/* handle the specified unsol action (ALC_XXX_EVENT) */
+static void alc_exec_unsol_event(struct hda_codec *codec, int action)
  {
-	struct alc_spec *spec = codec->spec;
-	if (codec->vendor_id == 0x10ec0880)
-		res>>= 28;
-	else
-		res>>= 26;
-	if (spec->use_jack_tbl)
-		res = snd_hda_jack_get_action(codec, res);
-	switch (res) {
+	switch (action) {
  	case ALC_HP_EVENT:
  		alc_hp_automute(codec);
  		break;
@@ -645,6 +638,19 @@ static void alc_sku_unsol_event(struct hda_codec *codec, unsigned int res)
  	snd_hda_jack_report_sync(codec);
  }
 

sound/pci/hda/patch_realtek.c static void alc_sku_unsol_event(struct hda_codec *codec, unsigned int res) · line 650

Quoted 10 lines
sound/pci/hda/patch_realtek.c +10 -0
+/* unsolicited event for HP jack sensing */
+static void alc_sku_unsol_event(struct hda_codec *codec, unsigned int res)
+{
+	struct alc_spec *spec = codec->spec;
+	if (codec->vendor_id == 0x10ec0880)
+		res>>= 28;
+	else
+		res>>= 26;
+	if (spec->use_jack_tbl)
+		res = snd_hda_jack_get_action(codec, res);

Can the check for spec->use_jack_tbl can be removed here?

Quoted 23 lines
sound/pci/hda/patch_realtek.c +7 -2
+	alc_exec_unsol_event(codec, res);
+}
+
  /* call init functions of standard auto-mute helpers */
  static void alc_inithook(struct hda_codec *codec)
  {
@@ -1883,7 +1889,7 @@ static const struct snd_kcontrol_new alc_beep_mixer[] = {
  };
  #endif
 
-static int alc_build_controls(struct hda_codec *codec)
+static int __alc_build_controls(struct hda_codec *codec)
  {
  	struct alc_spec *spec = codec->spec;
  	struct snd_kcontrol *kctl = NULL;
@@ -2029,11 +2035,16 @@ static int alc_build_controls(struct hda_codec *codec)
 
  	alc_free_kctls(codec); /* no longer needed */
 
-	err = snd_hda_jack_add_kctls(codec,&spec->autocfg);
+	return 0;
+}
+

sound/pci/hda/patch_realtek.c static int alc_build_controls(struct hda_codec *codec) · line 2048

Quoted 10 lines
sound/pci/hda/patch_realtek.c +5 -2
+static int alc_build_controls(struct hda_codec *codec)
+{
+	struct alc_spec *spec = codec->spec;
+	int err = __alc_build_controls(codec);
  	if (err<  0)
  		return err;
-
-	return 0;
+	return snd_hda_jack_add_kctls(codec,&spec->autocfg);
  }

I think you can simplify this by

	if (spec->use_jack_tbl)
		err = snd_hda_jack_add_kctls(codec,&spec->autocfg);

If so, only one alc_build_controls is necessary.

Quoted 29 lines
sound/pci/hda/patch_realtek.c +6 -0
@@ -4168,6 +4179,8 @@ static int patch_alc880(struct hda_codec *codec)
  	codec->patch_ops = alc_patch_ops;
  	if (board_config == ALC_MODEL_AUTO)
  		spec->init_hook = alc_auto_init_std;
+	else
+		codec->patch_ops.build_controls = __alc_build_controls;
  #ifdef CONFIG_SND_HDA_POWER_SAVE
  	if (!spec->loopback.amplist)
  		spec->loopback.amplist = alc880_loopbacks;
@@ -4297,6 +4310,8 @@ static int patch_alc260(struct hda_codec *codec)
  	codec->patch_ops = alc_patch_ops;
  	if (board_config == ALC_MODEL_AUTO)
  		spec->init_hook = alc_auto_init_std;
+	else
+		codec->patch_ops.build_controls = __alc_build_controls;
  	spec->shutup = alc_eapd_shutup;
  #ifdef CONFIG_SND_HDA_POWER_SAVE
  	if (!spec->loopback.amplist)
@@ -4691,6 +4706,8 @@ static int patch_alc882(struct hda_codec *codec)
  	codec->patch_ops = alc_patch_ops;
  	if (board_config == ALC_MODEL_AUTO)
  		spec->init_hook = alc_auto_init_std;
+	else
+		codec->patch_ops.build_controls = __alc_build_controls;
 
  #ifdef CONFIG_SND_HDA_POWER_SAVE
  	if (!spec->loopback.amplist)
Signature · 2 lines
David Henningsson, Canonical Ltd.
http://launchpad.net/~diwic
#4 lore raw

At Thu, 19 Jan 2012 12:39:28 +0100,

Quoted 41 lines · David Henningsson
Quoted 20 lines · Takashi Iwai

At Thu, 19 Jan 2012 11:35:23 +0100,

Quoted 18 lines · Takashi Iwai

At Wed, 18 Jan 2012 17:28:32 +0100,

Quoted 14 lines · David Henningsson

Hi Takashi,

I'm troubleshooting a problem with the Realtek quirk/model parsers. The problem is basically that add_jack_kctls call snd_hda_jack_detect_enable, which in turn overwrites the current unsol tag.

E g, first the model parser sets a pin to AC_USRSP_EN | ALC_HP_EVENT, then comes snd_hda_jack_detect_enable and does AC_USRSP_EN | jack->tag.

I'm not sure of the best way to resolve this; either keep the current tag (and mark all jacks as dirty on an unsol event), or read back the current tag and set it as jack->action (and then always use the jack table). What do you think?

Quoted 8 lines
Quoted 6 lines

I think a safer way is to avoid calling snd_hda_jack_add_kctls() for non-auto-parser, and let quirks use its own unsol_event handler instead of alc_sku_unsol_event(). snd_hda_jack_detect() itself works both with or without jack-kctls.

I'll fix up the upstream tree.

The patch below.

Thanks for quick handling. This will mean no jack kctls for model parsers, which is a bit sad, but hopefully its an incentive for people to move to auto-parsers.

One review comment below:

Quoted 4 lines
sound/pci/hda/patch_realtek.c +0 -0
@@ -645,6 +638,19 @@ static void alc_sku_unsol_event(struct hda_codec *codec, unsigned int res)
  	snd_hda_jack_report_sync(codec);
  }
 
Quoted 12 lines

sound/pci/hda/patch_realtek.c static void alc_sku_unsol_event(struct hda_codec *codec, unsigned int res) · line 650

Quoted 10 lines
sound/pci/hda/patch_realtek.c +10 -0
+/* unsolicited event for HP jack sensing */
+static void alc_sku_unsol_event(struct hda_codec *codec, unsigned int res)
+{
+	struct alc_spec *spec = codec->spec;
+	if (codec->vendor_id == 0x10ec0880)
+		res>>= 28;
+	else
+		res>>= 26;
+	if (spec->use_jack_tbl)
+		res = snd_hda_jack_get_action(codec, res);

Can the check for spec->use_jack_tbl can be removed here?

Hm, good point. Actually, the whole changes for creating own unsol_event() weren't necessary.

OTOH, separating the handlers makes cleaner, i.e. eventually we can drop spec->use_jack_tbl field completely. Let's do it in another patch.

thanks,

Takashi