[alsa-devel] HDA regression on Fujitsu S7020 laptop (ALC260 codec)

[Takashi Iwai]

At Sun, 17 Feb 2013 18:00:21 +0100,
Takashi Iwai wrote:
> 
> At Sun, 17 Feb 2013 17:59:29 +0100,
> Takashi Iwai wrote:
> > 
> > At Sun, 17 Feb 2013 23:23:16 +1030,
> > Jonathan Woithe wrote:
> > > 
> > > On Sun, Feb 17, 2013 at 09:32:38AM +0100, Takashi Iwai wrote:
> > > > > I've been able to test this (snapshot as of two days ago, reports version
> > > > > number 3.8.0-rc5+ via uname).  As expected there have been some
> > > > > improvements.
> > > > 
> > > > Thanks for testing.
> > > > 
> > > > > Things which work:
> > > > > 
> > > > >  * "Headphone Jack Mode" allows switching between "Headphone" and "Line
> > > > >    out".  Both of these mode settings appear to work.
> > > > > 
> > > > >  * There are separate "Headphone" and "Speaker" volume controls.
> > > > > 
> > > > >  * The "Headphone" volume control influences the headphone jack volume.
> > > > > 
> > > > >  * An "Auto-Mute Mode" control appears in alsamixer (it was missing on
> > > > >    earlier versions).
> > > > > 
> > > > >  * Alsamixer recognises the presence of two ADCs, giving "Capture" and
> > > > >    "Capture 1" controls, with corresponding "Input Source" controls.
> > > > > 
> > > > > Things which don't work or are missing:
> > > > > 
> > > > >  * "Headphone Jack Mode" lacks options to enable various "mic" or "line in"
> > > > >    modes (all of which are possible on this system)
> > > > 
> > > > Yes, it is still unimplemented.  So, it's expected not to be there
> > > > yet.  If you can hack it by yourself, let me know.
> > > 
> > > I may be able to give this a go, but from what I recall there was some
> > > debate as to how this should be implemented.  Has that been resolved yet?
> > 
> > No, we need to try to implement something and discuss later.
> > I think simply implementing two things would work: providing the vref
> > setup in Headphone Jack Mode enum, and add a pin to the Input Source.
> > User would require two setups, but it's less conflicting with giving
> > implicit dependency with each other, IMO.
> > 
> > > Is there any code I could use as a template or example?
> > 
> > No template, since the parser code has been completely rewritten
> > recently.  Just take a look at hda_generic.c.
> > 
> > 
> > > > >  * The "Input source" controls don't offer the "headphone jack" as a
> > > > >    possible source.
> > > > 
> > > > Ditto.
> > > 
> > > Ok.
> > > 
> > > > >  * The built-in speaker is completely silent under all situations I have
> > > > >    come up with.  The setting of "Auto-Mute Mode" has no effect, neither
> > > > >    does a manual unmute of the "Speaker" volume control.  While I don't
> > > > >    personally care much about this I suspect others might.
> > > > 
> > > > Hm, this is really mysterious.  I have no idea which init code causes
> > > > the problem, and it's hard to debug without a hardware.  Could you try
> > > > to figure out which setup is missing?
> > > 
> > > I spent a quite some time on this using hda-verb tonight with no success.
> > > I tried many combinations of mute settings and in the end unmuted every node
> > > I can see which lies between the DAC (nid 0x2) and the speaker in this
> > > laptop (nid 0x10).  That is:
> > > 
> > >  * Unmute the inputs on node 0x9
> > >  * Maximise the volume on node 0x9
> > >  * Unmute the input to node 0x10
> > > 
> > > Even with this the speakers remained mute.  I then tried using the relevant
> > > parts of the S702X initialisation sequence from 3.3.5 but even this didn't
> > > permit the speaker to make any sound.  Whatever's going on it's subtle.
> > > 
> > > I confirmed that the alsamixer "Speaker Volume" control manipulated the
> > > output volume control of node 0x9 as it should.
> > > 
> > > A quick comparison of /proc/asound/card0/codec#0 from 3.3.5 (where the
> > > speaker works) and 3.8.0-rc5+ (where it doesn't) failed to show anything
> > > obvious.  A more detailed comparison may be required.
> > 
> > Hm, then maybe the default initialization verbs for ALC260 does
> > something harmful for this machine?  A quick (untested) patch is
> > below.
> 
> Oops, forgot got attach.

Scratch that.  It doesn't work as expected.
Try the following one instead.


Takashi

---
diff --git a/sound/pci/hda/patch_realtek.c b/sound/pci/hda/patch_realtek.c
index 48c9d10..e0e0a5a 100644
--- a/sound/pci/hda/patch_realtek.c
+++ b/sound/pci/hda/patch_realtek.c
@@ -1516,8 +1516,14 @@ static void alc260_fixup_fsc_s7020(struct hda_codec *codec,
 {
 	struct alc_spec *spec = codec->spec;
 
-	if (action == HDA_FIXUP_ACT_PRE_PROBE)
+	switch (action) {
+	case HDA_FIXUP_ACT_PRE_PROBE:
 		spec->gen.add_out_jack_modes = 1;
+		break;
+	case HDA_FIXUP_ACT_PROBE:
+		spec->init_amp = ALC_INIT_NONE;
+		break;
+	}
 }
 
 static const struct hda_fixup alc260_fixups[] = {