Re: Quiet crackling noise at boot and after stopping music

[Takashi Iwai <tiwai@suse.de>]

On Wed, 06 Aug 2025 18:12:20 +0200,
Paul Menzel wrote:
> 
> Dear Takashi,
> 
> 
> Thank you for your reply.
> 
> Am 06.08.25 um 16:42 schrieb Takashi Iwai:
> > On Wed, 06 Aug 2025 15:57:51 +0200, Paul Menzel wrote:
> 
> >> Am 06.08.25 um 10:46 schrieb Takashi Iwai:
> >>> On Wed, 06 Aug 2025 10:34:57 +0200, Paul Menzel wrote:
> >> 
> >>>> Am 06.08.25 um 10:20 schrieb Takashi Iwai:
> >>>>> On Wed, 06 Aug 2025 10:15:08 +0200, Paul Menzel wrote:
> >>>> 
> >>>>>> I believe the Realtek refactoring caused a regression on my Intel Kaby
> >>>>>> Lake laptop Dell XPS 13 9360. There is quiet crackling noise during
> >>>>>> boot, when GDM is started, and also after playing music is
> >>>>>> stopped. It’s reproducible with 6.16.0-04055-g14bed9bc81ba, that
> >>>>>> contains the sound subsystem pull request.
> >>>>>> 
> >>>>>> 177bf8620cf4 Merge tag 'sound-6.17-rc1' of
> >>>>>> git://git.kernel.org/pub/scm/linux/kernel/git/tiwai/sound
> >>>>>> 
> >>>>>> I uploaded the run of alsa-info [1]. I try to verify with plain 6.16 later.
> >>>>> 
> >>>>> It's likely because of the missing CONFIG_SND_HDA_CODEC_* selections.
> >>>>> Please set CONFIG_SND_HDA_CODEC_ALC269=m and
> >>>> 
> >>>> Indeed, I didn’t select this.
> >>>> 
> >>>>> CONFIG_SND_HDA_CODEC_HDMI_INTEL=m if not done yet.
> >>>> 
> >>>> I had selected this.
> >>>> 
> >>>>> Those were already set to default=y in the later PR for avoiding such
> >>>>> a pitfall.
> >>>> 
> >>>> I remember I tried to check which to select, and I ran `alsa-info` to
> >>>> find the codecs, but only found:
> >>>> 
> >>>>       Codec: Realtek ALC3246
> >>>> 
> >>>> As this didn’t match to any of the offerings, I didn’t select
> >>>> it. May I suggest to extend the Kconfig help texts with notes how to
> >>>> determine the correct module?
> >>> 
> >>> There are quite lots of different codec names, so we can't put all
> >>> them, unfortunately...
> >> 
> >> Understood. Then maybe just the command how to find the id, that
> >> people can use to grep the correct module for? From the script
> >> `alsa-info.sh`, maybe, I found:
> >> 
> >>      $ grep -e Codec -e "Vendor Id" /proc/asound/card*/codec#*
> >>      /proc/asound/card0/codec#0:Codec: Realtek ALC3246
> >>      /proc/asound/card0/codec#0:Vendor Id: 0x10ec0256
> >>      /proc/asound/card0/codec#2:Codec: Intel Kabylake HDMI
> >>      /proc/asound/card0/codec#2:Vendor Id: 0x8086280b
> >> 
> >> So mention `/proc/asound/card*/codec#*`?
> > 
> > Honestly speaking, I don't want that; it won't scale, and new codecs
> > may appear at any moment.  Then we'd have to manage multiple places
> > for adding or removing ids.
> 
> Sorry, I didn’t mean to add the id’s to the description, but something like:
> 
> You can check the vendor id in `/proc/asound/card*/codec#*`

Well, I'm afraid that it's still too cryptic to judge which driver to
be enabled.

> > Maybe an alternative option would be to mark prompt of each Realtek
> > codec driver with EXPERT, so that all Realtek drivers will be enabled
> > (when CONFIG_SND_HDA_CODEC_REALTEK is set) unless you say you're an
> > expert :)
> 
> At least I wouldn’t have bothered you. ;-) My gut prefers, improving
> the documentation instead of hiding it, but you are the maintainer.
> 
>     $ ls -lh
> /lib/modules/6.16.0-11852-g479058002c32/kernel/sound/hda/codecs/realtek/
> insgesamt 464K
>     -rw-r--r-- 1 root root 300K  6. Aug 10:22 snd-hda-codec-alc269.ko
>     -rw-r--r-- 1 root root 164K  6. Aug 10:22 snd-hda-codec-realtek-lib.ko
> 
> The size is small compared to other drivers, but not nothing.

> 
> 
> Kind regards,
> 
> Paul
> 
> 
> PS: Despite just building in `CONFIG_SND_HDA_CODEC_ALC269=m`,
> 6.16.0-11852-g479058002c32 is 172 kB bigger:
> 
>     $ du -s /lib/modules/6.16.0/kernel/sound/
>     10992	/lib/modules/6.16.0/kernel/sound/
>     $ du -s /lib/modules/6.16.0-11852-g479058002c32/kernel/sound/
>     11164	/lib/modules/6.16.0-11852-g479058002c32/kernel/sound/

The previous kernels had all Realtek codecs enabled in a single
driver, so I believe it's safer to enable all again.
If you really care about the size, you have still choices (with
CONFIG_EXPERT=y).

Hans also agreed with this approach in this thread, so this looks like
the way to go, for avoiding further stray sheep after 6.17-rc1.

I'm going to submit a patch set.


thanks,

Takashi