[alsa-devel] [Fwd: Re: [Regression] With the new 2.6.33 when I plug headphones, the speaker doen't get off anymore]

[giggzounet]

Takashi Iwai a écrit :
> At Mon, 01 Mar 2010 14:46:10 +0100,
> Guillaume De Nayer wrote:
>> Takashi Iwai a écrit :
>>> At Sun, 28 Feb 2010 15:29:25 +0100,
>>> giggz wrote:
>>>> giggz a écrit :
>>>>> giggzounet a écrit :
>>>>>> Hi,
>>>>>>
>>>>>> I have installed debian stable lenny + backports on an eeepc 1201n. With
>>>>>> the 2.6.30 or 2.6.32 from lenny-backport I don't have problem with sound
>>>>>> : when I plug the headphones, speakers get off.
>>>>>>
>>>>>> I have tested with 2.6.33 and when I plug headphones, speakers don't get
>>>>>> off anymore. so I have sound in headphones and speakers.
>>>>>>
>>>>>> 22:42 giggz at baal ~ % cat /proc/asound/card0/codec#* | grep Codec
>>>>>> Codec: Realtek ALC269
>>>>>> Codec: Nvidia MCP7A HDMI
>>>>>>
>>>>>> I attach the output of alsa-info with the 2.6.32.9 and the 2.6.33 (I'm
>>>>>> in a Uni Campus and it's quite difficult to have internet on a external
>>>>>> laptop...so I attach them at this mail and I don't upload them on pastbin).
>>>>>>
>>>>>> I have opened a bug on the kernel bugzilla (bug 15399). Should I provide
>>>>>> anything more ?
>>>>>>
>>>>> Following the advice of Paul Menzel I did a "diff" of the sources of the
>>>>> 2.6.32.9 kernel and of the 2.6.33. In patch_realtek.c, we see that there
>>>>> is an new snd_hda_jack_detect funtion with 2 arguments. I have noticed
>>>>> that the second argument is "normaly" the second argument of
>>>>> snd_hda_codec_read. In the alc269_speaker_automute function there is
>>>>> this new snd_hda_jack_detect function, but the second argument is "nid".
>>>>> But in the old alc269_speaker_automute of the 2.6.32.9 the second
>>>>> argument of snd_hda_codec_read is 0x15. So I think there is perhaps a
>>>>> bug here...But I hesitate to modify the source...so I'm waiting the
>>>>> anwser of the dev.
>>>>>
>>>>> --- patch_realtek.c	2010-02-27 14:58:06.000000000 +0100
>>>>> +++ patch_realtek_modif.c	2010-02-27 14:58:54.000000000 +0100
>>>>> @@ -13381,7 +13381,7 @@
>>>>>  	unsigned int present;
>>>>>  	unsigned char bits;
>>>>>
>>>>> -	present = snd_hda_jack_detect(codec, nid);
>>>>> +	present = snd_hda_jack_detect(codec, 0x15);
>>>>>  	bits = present ? AMP_IN_MUTE(0) : 0;
>>>>>  	snd_hda_codec_amp_stereo(codec, 0x0c, HDA_INPUT, 0,
>>>>>  				AMP_IN_MUTE(0), bits);
>>>>>
>>>> I tested it. And with this modif in the kernel, after recompilation and
>>>> reboot, my problem is headphones and speaker is gone.
>>> It implies that the setup of the headphone pin is wrong.  This doesn't
>>> mean a bug of the driver, though.  It might have uncovered the BIOS
>>> bug.
>>>
>> ok. I don't have any knowledge in C or in bios...so you're problably
>> right :)
>>
>> but it seems strange that for all the new line with snd_hda_jack_detect
>> the second argument is always the second argument of snd_hda_codec_read
>> except for the alc269.
>>
>>> Could you provide alsa-info.sh without your patch to check the setup?
>>>
>>>
>> In attachment of the first post on alsa-devel or on the kernel bugzilla
>> http://bugzilla.kernel.org/show_bug.cgi?id=15399, there are the output
>> of alsa-info.sh with the 2.6.32.9 (no problem) and the output of
>> alsa-info.sh with the 2.6.33 without any patch (problem occurs).
> 
> OK, thanks.  This looks actually like a bug :)
> 

a bug in the driver or a bug in the bios ?

I have taken a look under the gitweb :
in the commit of the introduction of snd_hda_jack_detect()
http://git.kernel.org/?p=linux/kernel/git/next/linux-next.git;a=commit;h=864f92be7e8d4a0ba11d912e3f03d1a92a031dee
I see :
@@ -13183,8  +13098,7  @@ static void alc269_speaker_automute(struct
hda_codec *codec)
        unsigned int present;
        unsigned char bits;

-       present = snd_hda_codec_read(codec, 0x15, 0,
-                               AC_VERB_GET_PIN_SENSE, 0) & 0x80000000;
+       present = snd_hda_jack_detect(codec, 0x15);
        bits = present ? AMP_IN_MUTE(0) : 0;
        snd_hda_codec_amp_stereo(codec, 0x0c, HDA_INPUT, 0,
                                AMP_IN_MUTE(0), bits);

But in the commit :
http://git.kernel.org/?p=linux/kernel/git/next/linux-next.git;a=commitdiff;h=2a321e1defa266ad1b4ded29cb6cfb0ed5fce39d
the 0x15 is replaced with nid.

Cheers,
GiGGz