Re: ALSA: hda/realtek: quirk for D945GCLF2 mainboard

[Takashi Iwai <tiwai@suse.de>]

At Fri, 7 May 2010 02:59:01 +0200,
Stefan Lippers-Hollmann wrote:
> 
> [1  <text/plain; iso-8859-1 (7bit)>]
> Hi
> 
> On Friday 07 May 2010, Takashi Iwai wrote:
> > At Thu, 29 Apr 2010 23:00:42 +0200,
> > Stefan Lippers-Hollmann wrote:
> > > 
> > > Hi
> > > 
> > > Since kernel 2.6.33 (and including 2.6.34-rc5-git10), sound output is 
> > > broken on my Intel D945GCLF2 mainboard, while it did work well on earlier
> > > kernels up to and including 2.6.32.12.
> > 
> > Could you give alsa-info.sh output (run with --no-upload option)?
> > I would revert in the worst case, but before that, I'd like to analyze
> > the real issue.
> 
> Attached are alsa-info.sh and full, as far as the buffer allows, dmesg 
> outputs from:
> - plain 2.6.34-rc6-git5, no sound:
>   + alsa-info_plain-2.6.34-rc6-git5.log.gz
>   + dmesg_plain-2.6.34-rc6-git5.log.gz
> - 2.6.34-rc6-git5 with the proposed 
>   "revert-ALSA_hda-realtek_quirk-for-D945GCLF2-mainboard" applied:
>   + alsa-info_revert-ALSA_hda-realtek_quirk-for-D945GCLF2-mainboard.log.gz
>   + dmesg_revert-ALSA_hda-realtek_quirk-for-D945GCLF2-mainboard.log.gz
> the kernel config is config-2.6.34-rc6-sidux-amd64.gz in both cases.
> 
> Additionally I've tested a current OpenSuSE factory live CD 
> (openSUSE-KDE-LiveCD-x86_64-Build0589-Media.iso) daily with kernel 
> 2.6.34-rc - sound not working:
>   + alsa-info_openSUSE-KDE-LiveCD-x86_64-Build0589-Media.log.gz
>   + dmesg_openSUSE-KDE-LiveCD-x86_64-Build0589-Media.log.gz
> and a few days old Fedora rawhide nightly with exactly the same symptoms,
> no logs attached.
> 
> Up to and including kernel 2.6.32 (tested since ~2.6.26) ALSA always worked
> out-of-the-box, once I revert 7aee67466536bbf8bb44a95712c848a61c5a0acd
> (ALSA: hda/realtek: quirk for D945GCLF2 mainboard) from 2.6.33(.3) or 
> 2.6.34-rc6-git5, snd-hda-intel/ snd_hda_codec_realtek have managed to 
> initialize again in all my my tests so far. I'd be happy to test any 
> potential alternatives to that patch. Unfortunately I usually don't notice 
> graphics or audio regressions quickly, as that system is running quasi 
> headless/ speakerless most of the time, but I can test it any time of 
> course.

Could you try probe_mask=4 option without reverting?
I don't find any wrong verb with the quirk, so this looks more like a
problem in the slot probing.


thanks,

Takashi