Acer aspire 3830TG and Conexant 506c/20588

From 69bb710e13ae842d77d584a75088e008834a1722 Mon Sep 17 00:00:00 2001

From 9505d48fb556f977c5247c9564481301de39b258 Mon Sep 17 00:00:00 2001

The new Conexant 5066 auto-parser is barely finished and already we might need a quirk system for it...

Anyway, in this bug, the internal speaker is not working (and was not working before the auto-parser either). I'm attaching two patches that are quite simple, and at least the first one (add ID 506c) I think should be applied right away.

The problem: The internal speaker is at node 0x1f and you need to turn on EAPD on node 0x1b for the speaker to sound. Now I'm not sure how to fix this in the best way. An "init verb" quirk, or a "use EAPD from this pin" quirk? Try to copy-paste realtek's quirk system, and if so, refactor it into hda_codec.c, or write something new?

What are your ideas?

Alsa-info reference: https://launchpadlibrarian.net/73966502/alsa-info
Bug reference:
https://bugs.launchpad.net/ubuntu/+source/alsa-driver/+bug/783582

Signature · 3 lines

David Henningsson
http://launchpad.net/~diwic
[2 0001-ALSA-HDA-Add-a-new-Conexant-codec-ID-506c.patch <text/x-patch (7bit)>]

Quoted 1 line

From 69bb710e13ae842d77d584a75088e008834a1722 Mon Sep 17 00:00:00 2001

From: David Henningsson <[email protected]>
Date: Tue, 21 Jun 2011 20:51:34 +0200
Subject: [PATCH 1/2] ALSA: HDA: Add a new Conexant codec ID (506c)

Conexant ID 506c was found on Acer Aspire 3830TG. As users report no playback, sending to stable should be safe.

Cc [email protected]

BugLink https://bugs.launchpad.net/bugs/783582

Reported-by andROOM

Signed-off-by David Henningsson <[email protected]>

---
 sound/pci/hda/patch_conexant.c |    3 +++
 1 files changed, 3 insertions(+), 0 deletions(-)

sound/pci/hda/patch_conexant.c +3 -0

index 6e86427..6c9b742 100644
@@ -4390,6 +4390,8 @@ static const struct hda_codec_preset snd_hda_preset_conexant[] = {
 	  .patch = patch_cxt5066 },
 	{ .id = 0x14f15069, .name = "CX20585",
 	  .patch = patch_cxt5066 },
+	{ .id = 0x14f1506c, .name = "CX20588",
+	  .patch = patch_cxt5066 },
 	{ .id = 0x14f1506e, .name = "CX20590",
 	  .patch = patch_cxt5066 },
 	{ .id = 0x14f15097, .name = "CX20631",

@@ -4418,6 +4420,7 @@ MODULE_ALIAS("snd-hda-codec-id:14f15066");
 MODULE_ALIAS("snd-hda-codec-id:14f15067");
 MODULE_ALIAS("snd-hda-codec-id:14f15068");
 MODULE_ALIAS("snd-hda-codec-id:14f15069");
+MODULE_ALIAS("snd-hda-codec-id:14f1506c");
 MODULE_ALIAS("snd-hda-codec-id:14f1506e");
 MODULE_ALIAS("snd-hda-codec-id:14f15097");
 MODULE_ALIAS("snd-hda-codec-id:14f15098");

Signature · 3 lines

1.7.4.1

[3 0002-ALSA-HDA-Add-model-auto-quirk-for-Acer-Aspire-3830TG.patch <text/x-patch (7bit)>]

Quoted 1 line

From 9505d48fb556f977c5247c9564481301de39b258 Mon Sep 17 00:00:00 2001

From: David Henningsson <[email protected]>
Date: Tue, 21 Jun 2011 21:01:52 +0200
Subject: [PATCH 2/2] ALSA: HDA: Add model=auto quirk for Acer Aspire 3830TG

Since we're not using the new auto parser as a fallback yet, add it manually as a quirk.

Signed-off-by David Henningsson <[email protected]>

---
 sound/pci/hda/patch_conexant.c |    1 +
 1 files changed, 1 insertions(+), 0 deletions(-)

sound/pci/hda/patch_conexant.c +1 -0

index 6c9b742..91c8c0c 100644
@@ -3075,6 +3075,7 @@ static const char * const cxt5066_models[CXT5066_MODELS] = {
 };
 
 static const struct snd_pci_quirk cxt5066_cfg_tbl[] = {
+	SND_PCI_QUIRK(0x1025, 0x054c, "Acer Aspire 3830TG", CXT5066_AUTO),
 	SND_PCI_QUIRK_MASK(0x1025, 0xff00, 0x0400, "Acer", CXT5066_IDEAPAD),
 	SND_PCI_QUIRK(0x1028, 0x02d8, "Dell Vostro", CXT5066_DELL_VOSTRO),
 	SND_PCI_QUIRK(0x1028, 0x02f5, "Dell Vostro 320", CXT5066_IDEAPAD),

Signature · 1 line

1.7.4.1

At Tue, 28 Jun 2011 12:44:19 +0100,

Quoted 8 lines · David Henningsson

The new Conexant 5066 auto-parser is barely finished and already we might need a quirk system for it...

Anyway, in this bug, the internal speaker is not working (and was not working before the auto-parser either). I'm attaching two patches that are quite simple, and at least the first one (add ID 506c) I think should be applied right away.

Thanks, applied to fix/hda for 3.0-rc. Meanwhile, I'll change the branch for 3.1 to enable model=auto as default for Conexant (Thanks for remind :)

Quoted 6 lines

The problem: The internal speaker is at node 0x1f and you need to turn on EAPD on node 0x1b for the speaker to sound. Now I'm not sure how to fix this in the best way. An "init verb" quirk, or a "use EAPD from this pin" quirk? Try to copy-paste realtek's quirk system, and if so, refactor it into hda_codec.c, or write something new?

Quoted 1 line

What are your ideas?

We can simply turn on all EAPDs for 5066 and older chips. These codecs have just one or two EAPD pins, and we have no tightly coupled EAPD control (yet), so it'd be safe to turn all on.

At Tue, 28 Jun 2011 14:15:40 +0200,

Quoted 22 lines · Takashi Iwai

At Tue, 28 Jun 2011 12:44:19 +0100,

Quoted 8 lines · David Henningsson

The new Conexant 5066 auto-parser is barely finished and already we might need a quirk system for it...

Anyway, in this bug, the internal speaker is not working (and was not working before the auto-parser either). I'm attaching two patches that are quite simple, and at least the first one (add ID 506c) I think should be applied right away.

Thanks, applied to fix/hda for 3.0-rc. Meanwhile, I'll change the branch for 3.1 to enable model=auto as default for Conexant (Thanks for remind :)

Quoted 6 lines

The problem: The internal speaker is at node 0x1f and you need to turn on EAPD on node 0x1b for the speaker to sound. Now I'm not sure how to fix this in the best way. An "init verb" quirk, or a "use EAPD from this pin" quirk? Try to copy-paste realtek's quirk system, and if so, refactor it into hda_codec.c, or write something new?

Quoted 5 lines

Quoted 1 line

What are your ideas?

We can simply turn on all EAPDs for 5066 and older chips. These codecs have just one or two EAPD pins, and we have no tightly coupled EAPD control (yet), so it'd be safe to turn all on.

Or, maybe something like below would be cleverer... (Totally untested, of course.)

From: Takashi Iwai<[email protected]>
Subject: [PATCH] ALSA: hda - Handle extra EAPD in Conexant auto-parser

There seem some machines referring to EAPD bit of the unassigned pin for the speaker EAPD of a different pin, thus we need to control the EAPD of unused pins as well.

Signed-off-by Takashi Iwai<[email protected]>

---
  sound/pci/hda/patch_conexant.c |   42 +++++++++++++++++++++++++++++++++++++++-
  1 files changed, 41 insertions(+), 1 deletions(-)

sound/pci/hda/patch_conexant.c +8 -1

index 4ca880b..dbea3d0 100644
@@ -155,6 +155,10 @@ struct conexant_spec {
  	unsigned int mic_boost; /* offset into cxt5066_analog_mic_boost */
 
  	unsigned int beep_amp;
+
+	/* extra EAPD pins */
+	unsigned int num_eapds;
+	hda_nid_t eapds[4];
  };
 
  static int conexant_playback_pcm_open(struct hda_pcm_stream *hinfo,

@@ -3485,7 +3489,7 @@ static void cx_auto_update_speakers(struct hda_codec *codec)
  	/* if LO is a copy of either HP or Speaker, don't need to handle it */
  	if (cfg->line_out_pins[0] == cfg->hp_pins[0] ||
  	    cfg->line_out_pins[0] == cfg->speaker_pins[0])
-		return;
+		goto turn_extra_eapd;
  	if (spec->auto_mute) {
  		/* mute LO in auto-mode when HP jack is present */
  		if (cfg->line_out_type == AUTO_PIN_SPEAKER_OUT ||

@@ -3495,6 +3499,9 @@ static void cx_auto_update_speakers(struct hda_codec *codec)
  			on = 1;
  	}
  	do_automute(codec, cfg->line_outs, cfg->line_out_pins, on);
+ turn_extra_eapd:
+	/* turn on/off extra EAPDs, too */
+	cx_auto_turn_eapd(codec, spec->num_eapds, spec->eapds, on);
  }
 
  static void cx_auto_hp_automute(struct hda_codec *codec)

sound/pci/hda/patch_conexant.c +31 -0

+	int i;
+	for (i = 0; i<  nums; i++)
+		if (list[i] == nid)
+			return true;
+	return false;
+}
+
+/* parse extra-EAPD that aren't assigned to any pins */
+static void cx_auto_parse_eapd(struct hda_codec *codec)
+{
+	struct conexant_spec *spec = codec->spec;
+	struct auto_pin_cfg *cfg =&spec->autocfg;
+	hda_nid_t nid, end_nid;
+
+	end_nid = codec->start_nid + codec->num_nodes;
+	for (nid = codec->start_nid; nid<  end_nid; nid++) {
+		if (get_wcaps_type(get_wcaps(codec, nid)) != AC_WID_PIN)
+			continue;
+		if (!(snd_hda_query_pin_caps(codec, nid)&  AC_PINCAP_EAPD))
+			continue;
+		if (found_in_nid_list(nid, cfg->line_out_pins, cfg->line_outs) ||
+		    found_in_nid_list(nid, cfg->hp_pins, cfg->hp_outs) ||
+		    found_in_nid_list(nid, cfg->speaker_pins, cfg->speaker_outs))
+			continue;
+		spec->eapds[spec->num_eapds++] = nid;
+		if (spec->num_eapds>= ARRAY_SIZE(spec->eapds))
+			break;
+	}
+}
+
  static int cx_auto_parse_auto_config(struct hda_codec *codec)
  {
  	struct conexant_spec *spec = codec->spec;

@@ -3914,6 +3953,7 @@ static int cx_auto_parse_auto_config(struct hda_codec *codec)
  	cx_auto_parse_input(codec);
  	cx_auto_parse_digital(codec);
  	cx_auto_parse_beep(codec);
+	cx_auto_parse_eapd(codec);
  	return 0;
  }
 

Quoted 30 lines · Takashi Iwai

At Tue, 28 Jun 2011 14:15:40 +0200,

Quoted 27 lines · Takashi Iwai

At Tue, 28 Jun 2011 12:44:19 +0100,

Quoted 8 lines · David Henningsson

The new Conexant 5066 auto-parser is barely finished and already we might need a quirk system for it...

Anyway, in this bug, the internal speaker is not working (and was not working before the auto-parser either). I'm attaching two patches that are quite simple, and at least the first one (add ID 506c) I think should be applied right away.

Thanks, applied to fix/hda for 3.0-rc. Meanwhile, I'll change the branch for 3.1 to enable model=auto as default for Conexant (Thanks for remind :)

Quoted 7 lines

The problem: The internal speaker is at node 0x1f and you need to turn on EAPD on node 0x1b for the speaker to sound. Now I'm not sure how to fix this in the best way. An "init verb" quirk, or a "use EAPD from this pin" quirk? Try to copy-paste realtek's quirk system, and if so, refactor it into hda_codec.c, or write something new?

What are your ideas?

We can simply turn on all EAPDs for 5066 and older chips. These codecs have just one or two EAPD pins, and we have no tightly coupled EAPD control (yet), so it'd be safe to turn all on.

Quoted 2 lines

Or, maybe something like below would be cleverer... (Totally untested, of course.)

Yes, very nice, I was thinking of something like that. That will be best from a power consumption perspective. My only worry is that if turning EAPD on and off causes pops (I don't know if they do), it will be better to leave it on all the time.

Quoted 47 lines

From: Takashi Iwai<[email protected]>
Subject: [PATCH] ALSA: hda - Handle extra EAPD in Conexant auto-parser

There seem some machines referring to EAPD bit of the unassigned pin for the speaker EAPD of a different pin, thus we need to control the EAPD of unused pins as well.

Signed-off-by Takashi Iwai<[email protected]>

---
  sound/pci/hda/patch_conexant.c |   42 +++++++++++++++++++++++++++++++++++++++-
  1 files changed, 41 insertions(+), 1 deletions(-)

sound/pci/hda/patch_conexant.c +8 -1

index 4ca880b..dbea3d0 100644
@@ -155,6 +155,10 @@ struct conexant_spec {
  	unsigned int mic_boost; /* offset into cxt5066_analog_mic_boost */
 
  	unsigned int beep_amp;
+
+	/* extra EAPD pins */
+	unsigned int num_eapds;
+	hda_nid_t eapds[4];
  };
 
  static int conexant_playback_pcm_open(struct hda_pcm_stream *hinfo,

@@ -3485,7 +3489,7 @@ static void cx_auto_update_speakers(struct hda_codec *codec)
  	/* if LO is a copy of either HP or Speaker, don't need to handle it */
  	if (cfg->line_out_pins[0] == cfg->hp_pins[0] ||
  	    cfg->line_out_pins[0] == cfg->speaker_pins[0])
-		return;
+		goto turn_extra_eapd;
  	if (spec->auto_mute) {
  		/* mute LO in auto-mode when HP jack is present */
  		if (cfg->line_out_type == AUTO_PIN_SPEAKER_OUT ||

@@ -3495,6 +3499,9 @@ static void cx_auto_update_speakers(struct hda_codec *codec)
  			on = 1;
  	}
  	do_automute(codec, cfg->line_outs, cfg->line_out_pins, on);
+ turn_extra_eapd:
+	/* turn on/off extra EAPDs, too */
+	cx_auto_turn_eapd(codec, spec->num_eapds, spec->eapds, on);
  }
 
  static void cx_auto_hp_automute(struct hda_codec *codec)

sound/pci/hda/patch_conexant.c static void cx_auto_parse_beep(struct hda_codec *codec) · line 3912

Quoted 5 lines

sound/pci/hda/patch_conexant.c +2 -0

@@ -3901,6 +3908,38 @@ static void cx_auto_parse_beep(struct hda_codec *codec)
  #define cx_auto_parse_beep(codec)
  #endif
 
+static bool found_in_nid_list(hda_nid_t nid, const hda_nid_t *list, int nums)
+{

Hmm, I've seen variations of this function at more than one place already. Maybe a function like this would be good to have in hda_codec.c?

int snd_hda_find_nid(hda_nid_t nid, const hda_nid_t *list, int list_length)

{
	int i;
	for (i = 0; i < list_length; i++)
		if (list[i] == nid)
			return i;
	return -1;
}

#define bool snd_hda_nid_exists(nid, list, list_length) (snd_hda_find_nid(nid, list, list_length) != -1)

At Wed, 29 Jun 2011 10:16:51 +0100,

Quoted 32 lines · David Henningsson

Quoted 30 lines · Takashi Iwai

At Tue, 28 Jun 2011 14:15:40 +0200,

Quoted 27 lines · Takashi Iwai

At Tue, 28 Jun 2011 12:44:19 +0100,

Quoted 8 lines · David Henningsson

The new Conexant 5066 auto-parser is barely finished and already we might need a quirk system for it...

Anyway, in this bug, the internal speaker is not working (and was not working before the auto-parser either). I'm attaching two patches that are quite simple, and at least the first one (add ID 506c) I think should be applied right away.

Thanks, applied to fix/hda for 3.0-rc. Meanwhile, I'll change the branch for 3.1 to enable model=auto as default for Conexant (Thanks for remind :)

Quoted 7 lines

The problem: The internal speaker is at node 0x1f and you need to turn on EAPD on node 0x1b for the speaker to sound. Now I'm not sure how to fix this in the best way. An "init verb" quirk, or a "use EAPD from this pin" quirk? Try to copy-paste realtek's quirk system, and if so, refactor it into hda_codec.c, or write something new?

What are your ideas?

We can simply turn on all EAPDs for 5066 and older chips. These codecs have just one or two EAPD pins, and we have no tightly coupled EAPD control (yet), so it'd be safe to turn all on.

Quoted 7 lines

Quoted 2 lines

Or, maybe something like below would be cleverer... (Totally untested, of course.)

Yes, very nice, I was thinking of something like that. That will be best from a power consumption perspective. My only worry is that if turning EAPD on and off causes pops (I don't know if they do), it will be better to leave it on all the time.

Turning all EAPD on would be even much easier.
OTOH, we already control EAPD dynamically for the active pins, so this behavior could be determined better by some switch.

Quoted 65 lines

Quoted 47 lines

From: Takashi Iwai<[email protected]>
Subject: [PATCH] ALSA: hda - Handle extra EAPD in Conexant auto-parser

There seem some machines referring to EAPD bit of the unassigned pin for the speaker EAPD of a different pin, thus we need to control the EAPD of unused pins as well.

Signed-off-by Takashi Iwai<[email protected]>

---
   sound/pci/hda/patch_conexant.c |   42 +++++++++++++++++++++++++++++++++++++++-
   1 files changed, 41 insertions(+), 1 deletions(-)

sound/pci/hda/patch_conexant.c +8 -1

index 4ca880b..dbea3d0 100644
@@ -155,6 +155,10 @@ struct conexant_spec {
   	unsigned int mic_boost; /* offset into cxt5066_analog_mic_boost */
 
   	unsigned int beep_amp;
+
+	/* extra EAPD pins */
+	unsigned int num_eapds;
+	hda_nid_t eapds[4];
   };
 
   static int conexant_playback_pcm_open(struct hda_pcm_stream *hinfo,

@@ -3485,7 +3489,7 @@ static void cx_auto_update_speakers(struct hda_codec *codec)
   	/* if LO is a copy of either HP or Speaker, don't need to handle it */
   	if (cfg->line_out_pins[0] == cfg->hp_pins[0] ||
   	    cfg->line_out_pins[0] == cfg->speaker_pins[0])
-		return;
+		goto turn_extra_eapd;
   	if (spec->auto_mute) {
   		/* mute LO in auto-mode when HP jack is present */
   		if (cfg->line_out_type == AUTO_PIN_SPEAKER_OUT ||

@@ -3495,6 +3499,9 @@ static void cx_auto_update_speakers(struct hda_codec *codec)
   			on = 1;
   	}
   	do_automute(codec, cfg->line_outs, cfg->line_out_pins, on);
+ turn_extra_eapd:
+	/* turn on/off extra EAPDs, too */
+	cx_auto_turn_eapd(codec, spec->num_eapds, spec->eapds, on);
   }
 
   static void cx_auto_hp_automute(struct hda_codec *codec)

sound/pci/hda/patch_conexant.c static void cx_auto_parse_beep(struct hda_codec *codec) · line 3912

Quoted 5 lines

sound/pci/hda/patch_conexant.c +2 -0

@@ -3901,6 +3908,38 @@ static void cx_auto_parse_beep(struct hda_codec *codec)
   #define cx_auto_parse_beep(codec)
   #endif
 
+static bool found_in_nid_list(hda_nid_t nid, const hda_nid_t *list, int nums)
+{

Hmm, I've seen variations of this function at more than one place already. Maybe a function like this would be good to have in hda_codec.c?

int snd_hda_find_nid(hda_nid_t nid, const hda_nid_t *list, int list_length)

{
	int i;
	for (i = 0; i<  list_length; i++)
		if (list[i] == nid)
			return i;
	return -1;
}

Quoted 2 lines

#define bool snd_hda_nid_exists(nid, list, list_length) (snd_hda_find_nid(nid, list, list_length) != -1)

Yes, I was thinking of it (and actually wrote some code yesterday), but for now I implemented it as runnable on 3.0 as is ;)

Quoted 111 lines · Takashi Iwai

At Wed, 29 Jun 2011 10:16:51 +0100,

Quoted 32 lines · David Henningsson

Quoted 30 lines · Takashi Iwai

At Tue, 28 Jun 2011 14:15:40 +0200,

Quoted 27 lines · Takashi Iwai

At Tue, 28 Jun 2011 12:44:19 +0100,

Quoted 8 lines · David Henningsson

The new Conexant 5066 auto-parser is barely finished and already we might need a quirk system for it...

Anyway, in this bug, the internal speaker is not working (and was not working before the auto-parser either). I'm attaching two patches that are quite simple, and at least the first one (add ID 506c) I think should be applied right away.

Thanks, applied to fix/hda for 3.0-rc. Meanwhile, I'll change the branch for 3.1 to enable model=auto as default for Conexant (Thanks for remind :)

Quoted 7 lines

The problem: The internal speaker is at node 0x1f and you need to turn on EAPD on node 0x1b for the speaker to sound. Now I'm not sure how to fix this in the best way. An "init verb" quirk, or a "use EAPD from this pin" quirk? Try to copy-paste realtek's quirk system, and if so, refactor it into hda_codec.c, or write something new?

What are your ideas?

We can simply turn on all EAPDs for 5066 and older chips. These codecs have just one or two EAPD pins, and we have no tightly coupled EAPD control (yet), so it'd be safe to turn all on.

Quoted 7 lines

Quoted 2 lines

Or, maybe something like below would be cleverer... (Totally untested, of course.)

Yes, very nice, I was thinking of something like that. That will be best from a power consumption perspective. My only worry is that if turning EAPD on and off causes pops (I don't know if they do), it will be better to leave it on all the time.

Turning all EAPD on would be even much easier.
OTOH, we already control EAPD dynamically for the active pins, so this behavior could be determined better by some switch.

Quoted 65 lines

Quoted 47 lines

From: Takashi Iwai<[email protected]>
Subject: [PATCH] ALSA: hda - Handle extra EAPD in Conexant auto-parser

There seem some machines referring to EAPD bit of the unassigned pin for the speaker EAPD of a different pin, thus we need to control the EAPD of unused pins as well.

Signed-off-by Takashi Iwai<[email protected]>

---
   sound/pci/hda/patch_conexant.c |   42 +++++++++++++++++++++++++++++++++++++++-
   1 files changed, 41 insertions(+), 1 deletions(-)

sound/pci/hda/patch_conexant.c +8 -1

index 4ca880b..dbea3d0 100644
@@ -155,6 +155,10 @@ struct conexant_spec {
   	unsigned int mic_boost; /* offset into cxt5066_analog_mic_boost */
 
   	unsigned int beep_amp;
+
+	/* extra EAPD pins */
+	unsigned int num_eapds;
+	hda_nid_t eapds[4];
   };
 
   static int conexant_playback_pcm_open(struct hda_pcm_stream *hinfo,

@@ -3485,7 +3489,7 @@ static void cx_auto_update_speakers(struct hda_codec *codec)
   	/* if LO is a copy of either HP or Speaker, don't need to handle it */
   	if (cfg->line_out_pins[0] == cfg->hp_pins[0] ||
   	    cfg->line_out_pins[0] == cfg->speaker_pins[0])
-		return;
+		goto turn_extra_eapd;
   	if (spec->auto_mute) {
   		/* mute LO in auto-mode when HP jack is present */
   		if (cfg->line_out_type == AUTO_PIN_SPEAKER_OUT ||

@@ -3495,6 +3499,9 @@ static void cx_auto_update_speakers(struct hda_codec *codec)
   			on = 1;
   	}
   	do_automute(codec, cfg->line_outs, cfg->line_out_pins, on);
+ turn_extra_eapd:
+	/* turn on/off extra EAPDs, too */
+	cx_auto_turn_eapd(codec, spec->num_eapds, spec->eapds, on);
   }
 
   static void cx_auto_hp_automute(struct hda_codec *codec)

sound/pci/hda/patch_conexant.c static void cx_auto_parse_beep(struct hda_codec *codec) · line 3912

Quoted 5 lines

sound/pci/hda/patch_conexant.c +2 -0

@@ -3901,6 +3908,38 @@ static void cx_auto_parse_beep(struct hda_codec *codec)
   #define cx_auto_parse_beep(codec)
   #endif
 
+static bool found_in_nid_list(hda_nid_t nid, const hda_nid_t *list, int nums)
+{

Hmm, I've seen variations of this function at more than one place already. Maybe a function like this would be good to have in hda_codec.c?

int snd_hda_find_nid(hda_nid_t nid, const hda_nid_t *list, int list_length)

{
	int i;
	for (i = 0; i<  list_length; i++)
		if (list[i] == nid)
			return i;
	return -1;
}

Quoted 5 lines

Quoted 2 lines

#define bool snd_hda_nid_exists(nid, list, list_length) (snd_hda_find_nid(nid, list, list_length) != -1)

Yes, I was thinking of it (and actually wrote some code yesterday), but for now I implemented it as runnable on 3.0 as is ;)

Sorry, did you actually commit this code (I didn't see it in your tree), or were you expecting me to ask downstream to test it first?

At Thu, 30 Jun 2011 07:55:58 +0100,

Quoted 119 lines · David Henningsson

Quoted 116 lines · Takashi Iwai

At Wed, 29 Jun 2011 10:16:51 +0100,

Quoted 32 lines · David Henningsson

Quoted 30 lines · Takashi Iwai

At Tue, 28 Jun 2011 14:15:40 +0200,

Quoted 27 lines · Takashi Iwai

At Tue, 28 Jun 2011 12:44:19 +0100,

Quoted 8 lines · David Henningsson

The new Conexant 5066 auto-parser is barely finished and already we might need a quirk system for it...

Anyway, in this bug, the internal speaker is not working (and was not working before the auto-parser either). I'm attaching two patches that are quite simple, and at least the first one (add ID 506c) I think should be applied right away.

Thanks, applied to fix/hda for 3.0-rc. Meanwhile, I'll change the branch for 3.1 to enable model=auto as default for Conexant (Thanks for remind :)

Quoted 7 lines

The problem: The internal speaker is at node 0x1f and you need to turn on EAPD on node 0x1b for the speaker to sound. Now I'm not sure how to fix this in the best way. An "init verb" quirk, or a "use EAPD from this pin" quirk? Try to copy-paste realtek's quirk system, and if so, refactor it into hda_codec.c, or write something new?

What are your ideas?

We can simply turn on all EAPDs for 5066 and older chips. These codecs have just one or two EAPD pins, and we have no tightly coupled EAPD control (yet), so it'd be safe to turn all on.

Quoted 7 lines

Quoted 2 lines

Or, maybe something like below would be cleverer... (Totally untested, of course.)

Yes, very nice, I was thinking of something like that. That will be best from a power consumption perspective. My only worry is that if turning EAPD on and off causes pops (I don't know if they do), it will be better to leave it on all the time.

Turning all EAPD on would be even much easier.
OTOH, we already control EAPD dynamically for the active pins, so this behavior could be determined better by some switch.

Quoted 65 lines

Quoted 47 lines

From: Takashi Iwai<[email protected]>
Subject: [PATCH] ALSA: hda - Handle extra EAPD in Conexant auto-parser

There seem some machines referring to EAPD bit of the unassigned pin for the speaker EAPD of a different pin, thus we need to control the EAPD of unused pins as well.

Signed-off-by Takashi Iwai<[email protected]>

---
    sound/pci/hda/patch_conexant.c |   42 +++++++++++++++++++++++++++++++++++++++-
    1 files changed, 41 insertions(+), 1 deletions(-)

sound/pci/hda/patch_conexant.c +8 -1

index 4ca880b..dbea3d0 100644
@@ -155,6 +155,10 @@ struct conexant_spec {
    	unsigned int mic_boost; /* offset into cxt5066_analog_mic_boost */
 
    	unsigned int beep_amp;
+
+	/* extra EAPD pins */
+	unsigned int num_eapds;
+	hda_nid_t eapds[4];
    };
 
    static int conexant_playback_pcm_open(struct hda_pcm_stream *hinfo,

@@ -3485,7 +3489,7 @@ static void cx_auto_update_speakers(struct hda_codec *codec)
    	/* if LO is a copy of either HP or Speaker, don't need to handle it */
    	if (cfg->line_out_pins[0] == cfg->hp_pins[0] ||
    	    cfg->line_out_pins[0] == cfg->speaker_pins[0])
-		return;
+		goto turn_extra_eapd;
    	if (spec->auto_mute) {
    		/* mute LO in auto-mode when HP jack is present */
    		if (cfg->line_out_type == AUTO_PIN_SPEAKER_OUT ||

@@ -3495,6 +3499,9 @@ static void cx_auto_update_speakers(struct hda_codec *codec)
    			on = 1;
    	}
    	do_automute(codec, cfg->line_outs, cfg->line_out_pins, on);
+ turn_extra_eapd:
+	/* turn on/off extra EAPDs, too */
+	cx_auto_turn_eapd(codec, spec->num_eapds, spec->eapds, on);
    }
 
    static void cx_auto_hp_automute(struct hda_codec *codec)

sound/pci/hda/patch_conexant.c static void cx_auto_parse_beep(struct hda_codec *codec) · line 3912

Quoted 5 lines

sound/pci/hda/patch_conexant.c +2 -0

@@ -3901,6 +3908,38 @@ static void cx_auto_parse_beep(struct hda_codec *codec)
    #define cx_auto_parse_beep(codec)
    #endif
 
+static bool found_in_nid_list(hda_nid_t nid, const hda_nid_t *list, int nums)
+{

Hmm, I've seen variations of this function at more than one place already. Maybe a function like this would be good to have in hda_codec.c?

int snd_hda_find_nid(hda_nid_t nid, const hda_nid_t *list, int list_length)

{
	int i;
	for (i = 0; i<   list_length; i++)
		if (list[i] == nid)
			return i;
	return -1;
}

Quoted 2 lines

#define bool snd_hda_nid_exists(nid, list, list_length) (snd_hda_find_nid(nid, list, list_length) != -1)

Yes, I was thinking of it (and actually wrote some code yesterday), but for now I implemented it as runnable on 3.0 as is ;)

Quoted 2 lines

Sorry, did you actually commit this code (I didn't see it in your tree), or were you expecting me to ask downstream to test it first?

Not committed yet. At least for 3.0, I'd commit only the fix that was tested. If this can be tested quickly and the result is positive, I'll queue it up for 3.0. Otherwise put it to the branch for 3.1.

Quoted 124 lines · Takashi Iwai

At Thu, 30 Jun 2011 07:55:58 +0100,

Quoted 121 lines · David Henningsson

Quoted 111 lines · Takashi Iwai

At Wed, 29 Jun 2011 10:16:51 +0100,

Quoted 32 lines · David Henningsson

Quoted 30 lines · Takashi Iwai

At Tue, 28 Jun 2011 14:15:40 +0200,

Quoted 27 lines · Takashi Iwai

At Tue, 28 Jun 2011 12:44:19 +0100,

Quoted 8 lines · David Henningsson

The new Conexant 5066 auto-parser is barely finished and already we might need a quirk system for it...

Anyway, in this bug, the internal speaker is not working (and was not working before the auto-parser either). I'm attaching two patches that are quite simple, and at least the first one (add ID 506c) I think should be applied right away.

Thanks, applied to fix/hda for 3.0-rc. Meanwhile, I'll change the branch for 3.1 to enable model=auto as default for Conexant (Thanks for remind :)

Quoted 7 lines

The problem: The internal speaker is at node 0x1f and you need to turn on EAPD on node 0x1b for the speaker to sound. Now I'm not sure how to fix this in the best way. An "init verb" quirk, or a "use EAPD from this pin" quirk? Try to copy-paste realtek's quirk system, and if so, refactor it into hda_codec.c, or write something new?

What are your ideas?

We can simply turn on all EAPDs for 5066 and older chips. These codecs have just one or two EAPD pins, and we have no tightly coupled EAPD control (yet), so it'd be safe to turn all on.

Quoted 7 lines

Quoted 2 lines

Or, maybe something like below would be cleverer... (Totally untested, of course.)

Yes, very nice, I was thinking of something like that. That will be best from a power consumption perspective. My only worry is that if turning EAPD on and off causes pops (I don't know if they do), it will be better to leave it on all the time.

Turning all EAPD on would be even much easier.
OTOH, we already control EAPD dynamically for the active pins, so this behavior could be determined better by some switch.

Quoted 65 lines

Quoted 47 lines

From: Takashi Iwai<[email protected]>
Subject: [PATCH] ALSA: hda - Handle extra EAPD in Conexant auto-parser

There seem some machines referring to EAPD bit of the unassigned pin for the speaker EAPD of a different pin, thus we need to control the EAPD of unused pins as well.

Signed-off-by Takashi Iwai<[email protected]>

---
    sound/pci/hda/patch_conexant.c |   42 +++++++++++++++++++++++++++++++++++++++-
    1 files changed, 41 insertions(+), 1 deletions(-)

sound/pci/hda/patch_conexant.c +8 -1

index 4ca880b..dbea3d0 100644
@@ -155,6 +155,10 @@ struct conexant_spec {
    	unsigned int mic_boost; /* offset into cxt5066_analog_mic_boost */
 
    	unsigned int beep_amp;
+
+	/* extra EAPD pins */
+	unsigned int num_eapds;
+	hda_nid_t eapds[4];
    };
 
    static int conexant_playback_pcm_open(struct hda_pcm_stream *hinfo,

@@ -3485,7 +3489,7 @@ static void cx_auto_update_speakers(struct hda_codec *codec)
    	/* if LO is a copy of either HP or Speaker, don't need to handle it */
    	if (cfg->line_out_pins[0] == cfg->hp_pins[0] ||
    	    cfg->line_out_pins[0] == cfg->speaker_pins[0])
-		return;
+		goto turn_extra_eapd;
    	if (spec->auto_mute) {
    		/* mute LO in auto-mode when HP jack is present */
    		if (cfg->line_out_type == AUTO_PIN_SPEAKER_OUT ||

@@ -3495,6 +3499,9 @@ static void cx_auto_update_speakers(struct hda_codec *codec)
    			on = 1;
    	}
    	do_automute(codec, cfg->line_outs, cfg->line_out_pins, on);
+ turn_extra_eapd:
+	/* turn on/off extra EAPDs, too */
+	cx_auto_turn_eapd(codec, spec->num_eapds, spec->eapds, on);
    }
 
    static void cx_auto_hp_automute(struct hda_codec *codec)

sound/pci/hda/patch_conexant.c static void cx_auto_parse_beep(struct hda_codec *codec) · line 3912

Quoted 5 lines

sound/pci/hda/patch_conexant.c +2 -0

@@ -3901,6 +3908,38 @@ static void cx_auto_parse_beep(struct hda_codec *codec)
    #define cx_auto_parse_beep(codec)
    #endif
 
+static bool found_in_nid_list(hda_nid_t nid, const hda_nid_t *list, int nums)
+{

Hmm, I've seen variations of this function at more than one place already. Maybe a function like this would be good to have in hda_codec.c?

int snd_hda_find_nid(hda_nid_t nid, const hda_nid_t *list, int list_length)

{
	int i;
	for (i = 0; i<   list_length; i++)
		if (list[i] == nid)
			return i;
	return -1;
}

Quoted 5 lines

Quoted 2 lines

#define bool snd_hda_nid_exists(nid, list, list_length) (snd_hda_find_nid(nid, list, list_length) != -1)

Yes, I was thinking of it (and actually wrote some code yesterday), but for now I implemented it as runnable on 3.0 as is ;)

Sorry, did you actually commit this code (I didn't see it in your tree), or were you expecting me to ask downstream to test it first?

Quoted 3 lines

Not committed yet. At least for 3.0, I'd commit only the fix that was tested. If this can be tested quickly and the result is positive, I'll queue it up for 3.0. Otherwise put it to the branch for 3.1.

Hmm, that came back with mixed reviews. First tester said it didn't help (still had to manually turn on EAPD on 0x1b), second tester said it did work, but that headphones stopped working instead...

At Mon, 04 Jul 2011 13:02:17 +0200,

Quoted 130 lines · David Henningsson

Quoted 127 lines · Takashi Iwai

At Thu, 30 Jun 2011 07:55:58 +0100,

Quoted 113 lines · David Henningsson

Quoted 111 lines · Takashi Iwai

At Wed, 29 Jun 2011 10:16:51 +0100,

Quoted 32 lines · David Henningsson

Quoted 30 lines · Takashi Iwai

At Tue, 28 Jun 2011 14:15:40 +0200,

Quoted 27 lines · Takashi Iwai

At Tue, 28 Jun 2011 12:44:19 +0100,

Quoted 8 lines · David Henningsson

The new Conexant 5066 auto-parser is barely finished and already we might need a quirk system for it...

Anyway, in this bug, the internal speaker is not working (and was not working before the auto-parser either). I'm attaching two patches that are quite simple, and at least the first one (add ID 506c) I think should be applied right away.

Thanks, applied to fix/hda for 3.0-rc. Meanwhile, I'll change the branch for 3.1 to enable model=auto as default for Conexant (Thanks for remind :)

Quoted 7 lines

The problem: The internal speaker is at node 0x1f and you need to turn on EAPD on node 0x1b for the speaker to sound. Now I'm not sure how to fix this in the best way. An "init verb" quirk, or a "use EAPD from this pin" quirk? Try to copy-paste realtek's quirk system, and if so, refactor it into hda_codec.c, or write something new?

What are your ideas?

We can simply turn on all EAPDs for 5066 and older chips. These codecs have just one or two EAPD pins, and we have no tightly coupled EAPD control (yet), so it'd be safe to turn all on.

Quoted 7 lines

Quoted 2 lines

Or, maybe something like below would be cleverer... (Totally untested, of course.)

Yes, very nice, I was thinking of something like that. That will be best from a power consumption perspective. My only worry is that if turning EAPD on and off causes pops (I don't know if they do), it will be better to leave it on all the time.

Turning all EAPD on would be even much easier.
OTOH, we already control EAPD dynamically for the active pins, so this behavior could be determined better by some switch.

Quoted 65 lines

Quoted 47 lines

From: Takashi Iwai<[email protected]>
Subject: [PATCH] ALSA: hda - Handle extra EAPD in Conexant auto-parser

There seem some machines referring to EAPD bit of the unassigned pin for the speaker EAPD of a different pin, thus we need to control the EAPD of unused pins as well.

Signed-off-by Takashi Iwai<[email protected]>

---
    sound/pci/hda/patch_conexant.c |   42 +++++++++++++++++++++++++++++++++++++++-
    1 files changed, 41 insertions(+), 1 deletions(-)

sound/pci/hda/patch_conexant.c +8 -1

index 4ca880b..dbea3d0 100644
@@ -155,6 +155,10 @@ struct conexant_spec {
    	unsigned int mic_boost; /* offset into cxt5066_analog_mic_boost */
 
    	unsigned int beep_amp;
+
+	/* extra EAPD pins */
+	unsigned int num_eapds;
+	hda_nid_t eapds[4];
    };
 
    static int conexant_playback_pcm_open(struct hda_pcm_stream *hinfo,

@@ -3485,7 +3489,7 @@ static void cx_auto_update_speakers(struct hda_codec *codec)
    	/* if LO is a copy of either HP or Speaker, don't need to handle it */
    	if (cfg->line_out_pins[0] == cfg->hp_pins[0] ||
    	    cfg->line_out_pins[0] == cfg->speaker_pins[0])
-		return;
+		goto turn_extra_eapd;
    	if (spec->auto_mute) {
    		/* mute LO in auto-mode when HP jack is present */
    		if (cfg->line_out_type == AUTO_PIN_SPEAKER_OUT ||

@@ -3495,6 +3499,9 @@ static void cx_auto_update_speakers(struct hda_codec *codec)
    			on = 1;
    	}
    	do_automute(codec, cfg->line_outs, cfg->line_out_pins, on);
+ turn_extra_eapd:
+	/* turn on/off extra EAPDs, too */
+	cx_auto_turn_eapd(codec, spec->num_eapds, spec->eapds, on);
    }
 
    static void cx_auto_hp_automute(struct hda_codec *codec)

sound/pci/hda/patch_conexant.c static void cx_auto_parse_beep(struct hda_codec *codec) · line 3912

Quoted 5 lines

sound/pci/hda/patch_conexant.c +2 -0

@@ -3901,6 +3908,38 @@ static void cx_auto_parse_beep(struct hda_codec *codec)
    #define cx_auto_parse_beep(codec)
    #endif
 
+static bool found_in_nid_list(hda_nid_t nid, const hda_nid_t *list, int nums)
+{

Hmm, I've seen variations of this function at more than one place already. Maybe a function like this would be good to have in hda_codec.c?

int snd_hda_find_nid(hda_nid_t nid, const hda_nid_t *list, int list_length)

{
	int i;
	for (i = 0; i<   list_length; i++)
		if (list[i] == nid)
			return i;
	return -1;
}

Quoted 8 lines

Quoted 5 lines

Quoted 2 lines

#define bool snd_hda_nid_exists(nid, list, list_length) (snd_hda_find_nid(nid, list, list_length) != -1)

Yes, I was thinking of it (and actually wrote some code yesterday), but for now I implemented it as runnable on 3.0 as is ;)

Sorry, did you actually commit this code (I didn't see it in your tree), or were you expecting me to ask downstream to test it first?

Not committed yet. At least for 3.0, I'd commit only the fix that was tested. If this can be tested quickly and the result is positive, I'll queue it up for 3.0. Otherwise put it to the branch for 3.1.

Quoted 3 lines

Hmm, that came back with mixed reviews. First tester said it didn't help (still had to manually turn on EAPD on 0x1b), second tester said it did work, but that headphones stopped working instead...

The latter comment suggests that EAPD must be always on, then?

At Mon, 04 Jul 2011 13:05:44 +0200,

Quoted 133 lines · Takashi Iwai

At Mon, 04 Jul 2011 13:02:17 +0200,

Quoted 130 lines · David Henningsson

Quoted 127 lines · Takashi Iwai

At Thu, 30 Jun 2011 07:55:58 +0100,

Quoted 113 lines · David Henningsson

Quoted 111 lines · Takashi Iwai

At Wed, 29 Jun 2011 10:16:51 +0100,

Quoted 32 lines · David Henningsson

Quoted 30 lines · Takashi Iwai

At Tue, 28 Jun 2011 14:15:40 +0200,

Quoted 27 lines · Takashi Iwai

At Tue, 28 Jun 2011 12:44:19 +0100,

Quoted 8 lines · David Henningsson

The new Conexant 5066 auto-parser is barely finished and already we might need a quirk system for it...

Anyway, in this bug, the internal speaker is not working (and was not working before the auto-parser either). I'm attaching two patches that are quite simple, and at least the first one (add ID 506c) I think should be applied right away.

Thanks, applied to fix/hda for 3.0-rc. Meanwhile, I'll change the branch for 3.1 to enable model=auto as default for Conexant (Thanks for remind :)

Quoted 7 lines

The problem: The internal speaker is at node 0x1f and you need to turn on EAPD on node 0x1b for the speaker to sound. Now I'm not sure how to fix this in the best way. An "init verb" quirk, or a "use EAPD from this pin" quirk? Try to copy-paste realtek's quirk system, and if so, refactor it into hda_codec.c, or write something new?

What are your ideas?

We can simply turn on all EAPDs for 5066 and older chips. These codecs have just one or two EAPD pins, and we have no tightly coupled EAPD control (yet), so it'd be safe to turn all on.

Quoted 7 lines

Quoted 2 lines

Or, maybe something like below would be cleverer... (Totally untested, of course.)

Yes, very nice, I was thinking of something like that. That will be best from a power consumption perspective. My only worry is that if turning EAPD on and off causes pops (I don't know if they do), it will be better to leave it on all the time.

Turning all EAPD on would be even much easier.
OTOH, we already control EAPD dynamically for the active pins, so this behavior could be determined better by some switch.

Quoted 65 lines

Quoted 47 lines

From: Takashi Iwai<[email protected]>
Subject: [PATCH] ALSA: hda - Handle extra EAPD in Conexant auto-parser

There seem some machines referring to EAPD bit of the unassigned pin for the speaker EAPD of a different pin, thus we need to control the EAPD of unused pins as well.

Signed-off-by Takashi Iwai<[email protected]>

---
    sound/pci/hda/patch_conexant.c |   42 +++++++++++++++++++++++++++++++++++++++-
    1 files changed, 41 insertions(+), 1 deletions(-)

sound/pci/hda/patch_conexant.c +8 -1

index 4ca880b..dbea3d0 100644
@@ -155,6 +155,10 @@ struct conexant_spec {
    	unsigned int mic_boost; /* offset into cxt5066_analog_mic_boost */
 
    	unsigned int beep_amp;
+
+	/* extra EAPD pins */
+	unsigned int num_eapds;
+	hda_nid_t eapds[4];
    };
 
    static int conexant_playback_pcm_open(struct hda_pcm_stream *hinfo,

@@ -3485,7 +3489,7 @@ static void cx_auto_update_speakers(struct hda_codec *codec)
    	/* if LO is a copy of either HP or Speaker, don't need to handle it */
    	if (cfg->line_out_pins[0] == cfg->hp_pins[0] ||
    	    cfg->line_out_pins[0] == cfg->speaker_pins[0])
-		return;
+		goto turn_extra_eapd;
    	if (spec->auto_mute) {
    		/* mute LO in auto-mode when HP jack is present */
    		if (cfg->line_out_type == AUTO_PIN_SPEAKER_OUT ||

@@ -3495,6 +3499,9 @@ static void cx_auto_update_speakers(struct hda_codec *codec)
    			on = 1;
    	}
    	do_automute(codec, cfg->line_outs, cfg->line_out_pins, on);
+ turn_extra_eapd:
+	/* turn on/off extra EAPDs, too */
+	cx_auto_turn_eapd(codec, spec->num_eapds, spec->eapds, on);
    }
 
    static void cx_auto_hp_automute(struct hda_codec *codec)

sound/pci/hda/patch_conexant.c static void cx_auto_parse_beep(struct hda_codec *codec) · line 3912

Quoted 5 lines

sound/pci/hda/patch_conexant.c +2 -0

@@ -3901,6 +3908,38 @@ static void cx_auto_parse_beep(struct hda_codec *codec)
    #define cx_auto_parse_beep(codec)
    #endif
 
+static bool found_in_nid_list(hda_nid_t nid, const hda_nid_t *list, int nums)
+{

Hmm, I've seen variations of this function at more than one place already. Maybe a function like this would be good to have in hda_codec.c?

int snd_hda_find_nid(hda_nid_t nid, const hda_nid_t *list, int list_length)

{
	int i;
	for (i = 0; i<   list_length; i++)
		if (list[i] == nid)
			return i;
	return -1;
}

Quoted 8 lines

Quoted 5 lines

Quoted 2 lines

#define bool snd_hda_nid_exists(nid, list, list_length) (snd_hda_find_nid(nid, list, list_length) != -1)

Yes, I was thinking of it (and actually wrote some code yesterday), but for now I implemented it as runnable on 3.0 as is ;)

Sorry, did you actually commit this code (I didn't see it in your tree), or were you expecting me to ask downstream to test it first?

Not committed yet. At least for 3.0, I'd commit only the fix that was tested. If this can be tested quickly and the result is positive, I'll queue it up for 3.0. Otherwise put it to the branch for 3.1.

Quoted 5 lines

Quoted 3 lines

Hmm, that came back with mixed reviews. First tester said it didn't help (still had to manually turn on EAPD on 0x1b), second tester said it did work, but that headphones stopped working instead...

The latter comment suggests that EAPD must be always on, then?

So, what about the patch below instead?