Headphone not working on a desktop machine

[David Henningsson]

On a desktop machine we're currently enabling, the front headphone output does not work (no or very weak output).

Please don't triage, this is for tracking purposes only.

[David Henningsson (diwic)]

If there are no internal speakers, we should not turn the eapd switch
off, because it might be necessary to keep high for Headphone.

BugLink: https://bugs.launchpad.net/bugs/1155016
Signed-off-by: David Henningsson <email address hidden>
---
 sound/pci/hda/patch_sigmatel.c | 25 +++++++++++++++++++++++++
 1 file changed, 25 insertions(+)

Hi Takashi,

I encountered the bug when working with a pre-release machine (so no alsa-info, unfortunately).

Feel free to commit if you think this is a good idea. Or can you think of a use case
where this would cause a regression?

diff --git a/sound/pci/hda/patch_sigmatel.c b/sound/pci/hda/patch_sigmatel.c
index d57c81e..66200ea8 100644
--- a/sound/pci/hda/patch_sigmatel.c
+++ b/sound/pci/hda/patch_sigmatel.c
@@ -3524,6 +3524,31 @@ static int stac_parse_auto_config(struct hda_codec *codec)
  if (err < 0)
   return err;

+ /* Don't GPIO-mute speakers if there are no internal speakers, because
+ the GPIO might be necessary for Headphone */
+ if (spec->eapd_switch) {
+ hda_nid_t *nid_pin;
+ int nids, i;
+ bool found = false;
+ if (spec->gen.autocfg.line_out_type == AUTO_PIN_SPEAKER_OUT) {
+ nid_pin = spec->gen.autocfg.line_out_pins;
+ nids = spec->gen.autocfg.line_outs;
+ } else {
+ nid_pin = spec->gen.autocfg.speaker_pins;
+ nids = spec->gen.autocfg.speaker_outs;
+ }
+ for (i = 0; i < nids; i++) {
+ unsigned int def_conf = snd_hda_codec_get_pincfg(codec, nid_pin[i]);
+ unsigned int attr = snd_hda_get_input_pin_attr(def_conf);
+ if (attr == INPUT_PIN_ATTR_INT) {
+ found = true;
+ break;
+ }
+ }
+ if (!found)
+ spec->eapd_switch = 0;
+ }
+
  /* minimum value is actually mute */
  spec->gen.vmaster_tlv[3] |= TLV_DB_SCALE_MUTE;

--
1.7.9.5

[Takashi Iwai (tiwai)]

At Thu, 14 Mar 2013 10:42:00 +0100,
David Henningsson wrote:
>
> If there are no internal speakers, we should not turn the eapd switch
> off, because it might be necessary to keep high for Headphone.
>
> BugLink: https://bugs.launchpad.net/bugs/1155016
> Signed-off-by: David Henningsson <email address hidden>
> ---
> sound/pci/hda/patch_sigmatel.c | 25 +++++++++++++++++++++++++
> 1 file changed, 25 insertions(+)
>
> Hi Takashi,
>
> I encountered the bug when working with a pre-release machine (so no alsa-info, unfortunately).
>
> Feel free to commit if you think this is a good idea. Or can you think of a use case
> where this would cause a regression?

I'm inclined for avoiding applying this at that place for all IDT/STAC
codecs, but only to specific codec. patch_stac92hd73xx() is the only
IDT codecs setting spec->eapd_switch, and other places are only for
legacy STAC codecs, which I don't want to touch too much any longer.

spec->eapd_switch can be set on/off well even after calling
stac_parse_auto_config(), I guess, so it can be done in patch_*() as
well.

That is, I mean a patch like below.

thanks,

Takashi

---
diff --git a/sound/pci/hda/patch_sigmatel.c b/sound/pci/hda/patch_sigmatel.c
index d57c81e..b82cbc5 100644
--- a/sound/pci/hda/patch_sigmatel.c
+++ b/sound/pci/hda/patch_sigmatel.c
@@ -815,6 +815,29 @@ static int find_mute_led_cfg(struct hda_codec *codec, int default_polarity)
  return 0;
 }

+/* check whether a built-in speaker is included in parsed pins */
+static bool has_builtin_speaker(struct hda_codec *codec)
+{
+ struct sigmatel_spec *spec = codec->spec;
+ hda_nid_t *nid_pin;
+ int nids, i;
+
+ if (spec->gen.autocfg.line_out_type == AUTO_PIN_SPEAKER_OUT) {
+ nid_pin = spec->gen.autocfg.line_out_pins;
+ nids = spec->gen.autocfg.line_outs;
+ } else {
+ nid_pin = spec->gen.autocfg.speaker_pins;
+ nids = spec->gen.autocfg.speaker_outs;
+ }
+
+ for (i = 0; i < nids; i++) {
+ unsigned int def_conf = snd_hda_codec_get_pincfg(codec, nid_pin[i]);
+ if (snd_hda_get_input_pin_attr(def_conf) == INPUT_PIN_ATTR_INT)
+ return true;
+ }
+ return false;
+}
+
 /*
  * PC beep controls
  */
@@ -3891,6 +3914,12 @@ static int patch_stac92hd73xx(struct hda_codec *codec)
   return err;
  }

+ /* Don't GPIO-mute speakers if there are no internal speakers, because
+ * the GPIO might be necessary for Headphone
+ */
+ if (spec->eapd_switch && !has_builtin_speaker(codec))
+ spec->eapd_switch = false;
+
  codec->proc_widget_hook = stac92hd7x_proc_hook;

  snd_hda_apply_fixup(codec, HDA_FIXUP_ACT_PROBE);

[David Henningsson (diwic)]

Patch committed upstream (after some modification by upstream) as http://git.kernel.org/cgit/linux/kernel/git/tiwai/sound.git/commit/?id=303985f81019571db0b3a6f01fc7f03eb350657e

Will be in tomorrow's daily dkms package.

[Launchpad Janitor (janitor)]

This bug was fixed in the package linux - 3.9.0-0.2

---------------
linux (3.9.0-0.2) saucy; urgency=low

  [ Tim Gardner ]

  * Enable extras packaging for amd64/i386.
    Fixes build depenencies with brittany and linux-meta.
 -- Tim Gardner <email address hidden> Mon, 29 Apr 2013 05:37:01 -0600