Re: [alsa-devel] Asus w5f. ALC660-VD codec not detected

At Wed, 7 Jan 2009 12:57:57 +0100, Salvatore Filippone wrote:

On Wed, Jan 7, 2009 at 12:45 PM, Takashi Iwai tiwai@suse.de wrote:

At Mon, 29 Dec 2008 19:45:25 +0100,
Nigel Henry wrote:
>
> UPDATE: Salvatore has found a workaround, so that both codecs are
detected
> correctly. Any help from alsa developers for a permanent fix would be
> appreciated. See workaround at bottom of post.
>
(snip)
> <Quoting Salvatore>
> Saturday 16:05:58
>
> Ok,  I finally found a workaround.
> According to the thread starting at
> http://thread.gmane.org/gmane.linux.alsa.devel/58849
> I grabbed the latest source code
alsa-driver-1.0.18a.17.g27a32.139.g6e583/
> compiled with --with-debug=verbose and got the message
> Dec 27 15:13:06 localhost kernel: ALSA
> /mnt/travel/ALSA/alsa-driver-1.0.18a.17.g27a32.139.g6e583/pci/hda/../../
alsa-kernel/pci/hda/hda_intel.c:779:
> codec_mask = 0x2
>
> I then figure out that the sound device was at codec#0 when it worked,
> therefore I changed snd_hda_intel.c by forcing codec_mask
>         if (!chip->codec_mask) {
>                 chip->codec_mask = azx_readw(chip, STATESTS);
>                 chip->codec_mask=0x03;
>                 snd_printdd("codec_mask = 0x%x\n", chip->codec_mask);
>         }
>
> And now the sound works correctly.
> Of course this opens the question of why azx_readw  does not return a
> correct mask;

Well, this implies a BIOS bug.  This value is set by BIOS to indicate
which codec slot is available.

> maybe the developers will be interested and/or will suggest a
> fix that is a little more elegant/widely applicable? And why did the
base
> code work just once, for no clear reason ? (I did not have debug=verbose
at
> that time, so I don't have detailed messages in the system log).

Not sure about this as I didn't track this bug.
When did it work once?  Is the device enabled by BIOS properly?

To summarize:

1. It always did work with FC7.

I guess it worked likely just casually. The old driver didn't clear the register bit before probing, so it got uninitialized bits.

Takashi