[now that the holiday season is over I'd like to get this rolling again] On 19.12.22 18:01, Jaroslav Kysela wrote: > On 19. 12. 22 11:27, Thorsten Leemhuis wrote: >> On 19.12.22 11:00, Hans de Goede wrote: >>> On 12/19/22 10:17, Thorsten Leemhuis wrote: >>>> I noticed a regression report in bugzilla.kernel.org. As many (most?) >>>> kernel developer don't keep an eye on it, I decided to forward it by >>>> mail. Quoting from https://bugzilla.kernel.org/show_bug.cgi?id=216818 : >>>> >>>>> sonic82003@xxxxxxxxx 2022-12-18 08:52:32 UTC >>>>> >>>>> The mic mute led of my ThinkPad X1 Carbon Gen 9 doesn't work >>>>> anymore after updating linux to version 6. >>>>> I can still turn it on by running >>>>> >>>>> echo 1 > /sys/class/leds/platform::micmute/brightness >>>>> >>>>> With linux-lts it still works fine. >>>> >>>> See the ticket for more details. >>>> >>>> Note, I found a similar report that (despite my attempts to prevent >>>> things like this from happening) fell through the cracks here: >>>> https://bugzilla.kernel.org/show_bug.cgi?id=216355 >>>> >>>>> plum 2022-08-13 02:11:01 UTC >>>>> >>>>> I upgrade to kernel 5.19.1 but found my thinkpad x1 carbon 2021's >>>>> mute led stop working. >>>>> >>>>> Function is okay but LED won't light up. >>>>> >>>>> Back to kernel 5.18 and it's normal and working again. >>>>> >>>>> Fedora 36 64 bit >>>>> Gnome-shell 42 >>>> >>>> From a quick research it looks to me like this is an issue for the >>>> sounds maintainers, as the LED itself apparently works. If that is >>>> something for the platform people instead please speak up. >>> >>> Thanks for bringing this up, we recently hit this in Fedora too >>> and we have a fix/workaround there. Let me copy and paste what >>> I just added to bko216355 : >> >> Many thx for sharing these details, really helpful. >> >>> This is caused by a behavior change of the kernel code controlling >>> the LED to only turn on the LED when all inputs, including e.g. the >>> jack mic input are turned off in the alsa-mixer settings. >>> >>> But most userspace code only turns the mic which it is actually using >>> on/off when you hit the mic-mute hotkey. >>> >>> Also see: https://bugzilla.redhat.com/show_bug.cgi?id=2134824 >> >> Ahh, lot's of helpful information and even a bisect there. :-D >> >> #regzbot introduced: 9b014266ef8ad0159 > > It's not a regression from my view. Please elaborate. Afaics it is one, as something that used to work stopped working with a newer kernel version; it doesn't matter it worked by accident beforehand or can be fixed by updating userland, as Linus explained multiple times in the past: https://docs.kernel.org/process/handling-regressions.html#quotes-from-linus-about-regression But maybe I'm missing something. >>> Which is the same bug. >>> >>> There is a set of fixes available in the form of an alsa-ucm update >>> which tells the kernel to ignore the state of the jack mic input >>> restoring the old behavior: >>> >>> https://git.alsa-project.org/?p=alsa-ucm-conf.git;a=commitdiff;h=79a8ec44d3dcf097f4a4492c506cbcf338324175 >>> https://git.alsa-project.org/?p=alsa-ucm-conf.git;a=commitdiff;h=9ce9ddb4a84fb467602b716575ea1d8f2bab0c39 >> >> Hmmm, that's nice, but well, by Linux' "no regressions rule" the issue >> is caused by kernel change and thus must be fixed in the kernel, e.g. >> without forcing users to update anything in userspace. >> >> Jaroslav, are there any plans to do that? > > I wrote all relevant information to > https://bugzilla.redhat.com/show_bug.cgi?id=2134824 . The problem exists > from the initial microphone LED support in the SOF HDA driver, because > two drivers control the microphone LED simultaneously (sof-hda-dsp + > hda-intel). My recent update just made this thing more visible - the LED > state may be updated wrongly in all previous kernels. Original behavior: > last write wins. New behavior: all off = LED ON. The UCM fix (update the > default kernel runtime configuration from the user space) is sufficient > in my eyes for now because even the use case when the microphone LED > follows the state when all internal inputs are turned off makes sense. > > I think that the sof-hda-dsp driver maintainer may decide to change the > default settings in the HDA driver when the digital microphone is > detected. Adding Pierre-Louis to the chain. Pierre-Louis? Ciao, Thorsten (wearing his 'the Linux kernel's regression tracker' hat) -- Everything you wanna know about Linux kernel regression tracking: https://linux-regtracking.leemhuis.info/about/#tldr If I did something stupid, please tell me, as explained on that page. #regzbot poke
