Re: [regression] ctxfi: no speaker output when headphones connected on >=7.0.x

[Takashi Iwai]

On Thu, 07 May 2026 14:47:03 +0200,
Dutch Ingraham wrote:
> 
> > Sent: Thursday, May 07, 2026 at 1:30 AM
> > From: "Takashi Iwai" <tiwai@xxxxxxx>
> > To: "Harin Lee" <me@xxxxxxxxx>
> > Cc: "Takashi Iwai" <tiwai@xxxxxxx>, "Thorsten Leemhuis" <regressions@xxxxxxxxxxxxx>, "Jaroslav Kysela" <perex@xxxxxxxx>, "Takashi Iwai" <tiwai@xxxxxxxx>, "Linux kernel regressions list" <regressions@xxxxxxxxxxxxxxx>, "Linux Sound System" <linux-sound@xxxxxxxxxxxxxxx>, "LKML" <linux-kernel@xxxxxxxxxxxxxxx>, stoa@xxxxxx
> > Subject: Re: [regression] ctxfi: no speaker output when headphones connected on >=7.0.x
> >
> > On Thu, 07 May 2026 00:06:11 +0200,
> > Harin Lee wrote:
> > > 
> > > On 5/6/26 4:05 PM, Takashi Iwai wrote:
> > > > On Wed, 06 May 2026 08:24:36 +0200,
> > > > Thorsten Leemhuis wrote:
> > > >> 
> > > >> Hi! Just FYI, I noticed a regression report in bugzilla:
> > > >> 
> > > >> https://bugzilla.kernel.org/show_bug.cgi?id=221461
> > > >> 
> > > >> To quote:
> > > >> 
> > > >>> Dutch Ingraham 2026-05-03 18:12:49 UTC
> > > >>> 
> > > >>> Kernels >=7.0.1 contain a regression as to Creative Sound Blaster Z SE
> > > >>> discrete pci sound card, but not integrated. Kernels 7.0.1, 7.0.2, and
> > > >>> 7.0.3 tested; 7.0.0 not tested. Reverting to any 6.18.x or 6.12.x does
> > > >>> not present the issue.
> > > >>> 
> > > >>> Specifically, using pipewire managed by pavucontrol, on both Gentoo and
> > > >>> Arch Linux, when both external speakers and headphones are connected,
> > > >>> all sound is routed through the headphones regardless of whether settings
> > > >>> are set to Line Out/Speaker or Line Out/Headphones; there is no speaker
> > > >>> output. If headphones are disconnected, without reboot, speakers work;
> > > >>> reconnect headphones, without reboot, sound reverts to headphones.
> > > >>> 
> > > >>> Kernel: 7.0.3, Gentoo compiled from source using olddefconfig on 6.18.24
> > > >>> 		.config; Arch standard release
> > > >>> 'lspci -nnk | grep -i audio' attached
> > > >>> 'dmesg | grep -i audio' for 7.0.3 and 6.18.24 attached
> > > >>> pipewire-1.6.4 on both Gentoo and Arch Linux
> > > >>> alsa-firmware-1.2.4 on both Gentoo and Arch Linux
> > > >>> 'cat /proc/sys/kernel/tainted' returns 0 for both Gentoo and Arch Linux
> > > >>> 
> > > >>> Processor		:	AMD Ryzen7 5700G
> > > >>> Motherboard		:	Asus Prime X570-Pro
> > > >>> RAM				:	Patriot Viper Steel DDR-4 @3200mHz
> > > >>> Graphics        :   ASRock AMD Radeon RX 7600 Challenger 8GB OC
> > > >>> Sound Card      :   Creative Sound Blaster Z SE
> > > >>> 
> > > >>> 
> > > >>> Expected outcome: ability to choose output to either speakers or head-
> > > >>> phones while both are plugged-in to sound card.
> > > >>> 
> > > >>> Workaround: none known.
> > > >> 
> > > >> See the ticket for more details. One of them: 7.1-rc2 is affected as
> > > >> well. Reporter is CCed. Forwarded this to ensure this report makes it to
> > > >> the interested parties, as bugzilla might not have forwarded this.
> > > >> 
> > > >> Dutch Ingraham: if nobody within a few days has an idea why this
> > > >> happens, you will likely have to use a git bisection to get this resolved.
> > > > 
> > > > As there have been many significant changes in ctxfi since 6.18, it's
> > > > hard to tell what broke.  Please go git bisect.
> > > > 
> > > > 
> > > > thanks,
> > > > 
> > > > Takashi
> > > > 
> > > 
> > > Creative Sound Blaster Z-series uses the snd_hda_codec_ca0132 driver,
> > > not snd_ctxfi. The dmesg logs in the bugzilla shows only ca0132
> > > entries.
> > 
> > Ah, sorry, then it's easier.  The only major change is the
> > commit 778031e1658d206a52bf9491c91ae5d4f4a2509d
> >     ALSA: hda/ca0132: Set HP/Speaker auto-detect default from headphone pin verb
> > 
> > Try to revert this one.
> > 
> > 
> > thanks,
> > 
> > Takashi
> > 
> 
> I can confirm this seems like a good candidate. I built and tested 6.19.14
> and 7.0. On 6.19.14, speakers work as expected; on 7.0, speakers are not
> working as expected and the offending patch is included in the 7.0 patchset.
> 
> Thank you all for your work on this issue.

OK, then let's put Matt to the party.
Matt, could you help for debugging the regression?


Takashi