[alsa-devel] snd_had_intel turning off internal speakers by going to power state D3

Takashi Iwai tiwai at suse.de
Mon Sep 24 10:06:14 CEST 2012


At Sat, 22 Sep 2012 13:03:37 -0700,
Jeff Robins wrote:
> 
> Hello,
> 
> I have a Toshiba Satellite C655D with Mageia 2 and I have a problem
> where the internal speakers eventually stop producing sound, but the
> headphones still work.  After doing a bit of research and running
> alsa-info (attached) I figured out that one of the parts of the sound
> card
> had gone into a D3 power state.
> 
> Removing and then reloading the snd_hda_intel module fixed the
> problem, but then the sound would turn off again.  I was able to find
> a post
> in alsa-users where someone else came up with a better solution using
> hda-verb to reset the power state:
> http://www.mail-archive.com/alsa-user@lists.sourceforge.net/msg28769.html
> 
> I also found a thread in the gentoo forums where other people are
> having the same problem.  One of them has kernel 3.5.0:
> http://forums.gentoo.org/viewtopic-t-932038.html?sid=2bee6543c63f2342057c71bc876e7273
> 
> I'm open to compiling my own alsa and helping out in any way I can.
> Please let me know what I can do.
> 
> Mageia bug report for completeness:
> https://bugs.mageia.org/show_bug.cgi?id=7292

OK, it's getting interesting now.
We've got similar reports but so far only on Lenovo laptops with the
Conexant codecs.  (I'm not quite sure whether all are with CX20585.
Might be with other Conexant variants?)

I asked Conexant guys to check this issue, and they informed that
there is no automatic-go-to-D3 feature in the codec hardware itself.
And judging from the hardware state, it looks like the class-D speaker
goes down due to the over temperature.  So, we concluded that this
should be very likely Lenovo-specific things.

But now you reported about Toshiba laptop.  It brings up again the
possibility of the sound driver side...  Weird.

Can it be a regression since some kernel version?  The kernel
regression doesn't mean necessarily a sound driver regression but
might be a problem in another part like ACPI, too.


thanks,

Takashi


More information about the Alsa-devel mailing list