[CentOS] kernel-2.6.18-128.7.1.el5.i686 update breaks AMD onboard sound

Sun Aug 30 00:01:27 UTC 2009
Mike A. Harris <mharris at mharris.ca>

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

William L. Maltby wrote:

>> The new kernel-2.6.18-128.7.1.el5.i686 kernel breaks audio support on my
>> onboard audio on an AMD Solo motherboard.
>>
>> 00:07.5 Multimedia audio controller: Advanced Micro Devices [AMD]
>> AMD-8111 AC97 Audio (rev 03)
> 
> I know this isn't much help, but the starting point is to narrow
> possible causes. I'm guessing it's a mobo/driver specifc issue. This is
> because I just did a test on my setup and all was good.

This is my assumption as well.  It's looking like it is probably limited
either to the driver(s) for my audio hardware, or perhaps some common
component.  Hard to say yet.  I'm guessing that the security patches
changed some assumptions in a driver and the driver is now
malfunctioning.  Just to try and rule out a compilation issue though,
I'm going to test with the official RHEL kernel of the same version first.

> $ uname -a
> Linux centos501.homegroannetwork 2.6.18-128.7.1.el5 #1 SMP Mon Aug 24
> 08:20:55 EDT 2009 i686 athlon i386 GNU/Linux
> 
> 00:11.5 Multimedia audio controller: VIA Technologies, Inc.
> VT8233/A/8235/8237 AC97 Audio Controller (rev 60)
> 
> This is on an EPOX 8KRAI PRO desktop mobo.

This is a good datapoint to know it isn't across the board failures.

> Since it's the weekend, I don't know how many of the folks that could
> help are around. You might try looking at the kernel menu config file
> and see if that driver is disabled.

Keep in mind, this audio hardware has worked fine with all prior EL5
kernels, and just stopped working on the 127.7 security update.  Backing
it off back to the 128.4 build got audio functioning again.  It's
theoretically possible that an audio driver got disabled, but I'd be
surprised for Red Hat to have intentionally or accidentally disabled a
perfectly functional audio driver in a security update.  My assumption
is that due to the priority of the security update, some parts of the
kernel may have broken due to the changes with no easy way to
methodically test it all within Red Hat's QA framework, and a perhaps
buggy driver that got lucky before, is now broken and needs some love.  ;o)

After I confirm the problem exists in the official RHEL 128.7 kernel as
well, I'll check the kernel configs just to be certain, and try to
binary search for what patch or change introduced the problem.

> Maybe the plus kernel has it enabled, since they turn on lots of
> extra things.

Perhaps, but I've been using this hardware for 8 months without having
to use a 3rd party/unofficial kernel, so I doubt the problem is intentional.

The drivers for my hardware are there, they just aren't functioning like
they did previously.  Several device nodes are now absent from /dev/snd,
so something is definitely unintentionally broken as far as I can tell.
 I'll know more from some testing by the end of the day though.

If anyone else is having any similar broken driver type problems (with
any hardware) in the new kernel, please let me know too, as there might
be a common failure point perhaps.

Thanks for the feedback.





- --
Mike A. Harris
http://mharris.ca  |  https://twitter.com/mikeaharris

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.5 (GNU/Linux)

iD8DBQFKmcFV4RNf2rTIeUARAisgAJ9k30kn8gkpx5rab42q+SEfgySznwCfUByY
yQ2xutEckJTc2IlkDVA5P5k=
=zPke
-----END PGP SIGNATURE-----