Re: Regression: webcam freezing since Linux 6.15

[Bart Nagel <bart@tremby.net>]

At 2026-07-23 08:14:40 +0200, Michal Pecio wrote:
> On Wed, 22 Jul 2026 16:23:28 -0700, Bart Nagel wrote:
> > > You will probably see Missed Service Errors being logged
> > > occasionally, likely both on "good" and "bad" kernels. And it seems
> > > that good ones will recover every time, but not sure about bad
> > > ones. Does the first MSE break things forever? Or does it take a
> > > few? Could you post a log when it breaks? Caution: you may see
> > > thousands of lines per second.  
> > 
> > I'm still on the asterisked kernel from the list above at the moment
> > (appears to be "good"). I tried enabling this and I do indeed get a
> > few of those errors and they seem to recover. This is what I have so
> > far after having set it about 5 minutes ago:
> > 
> >     Jul 22 16:14:07 spiral kernel: xhci_hcd 0000:00:14.0: Miss service interval error for slot 18 ep 2, set skip flag
> >     Jul 22 16:14:07 spiral kernel: xhci_hcd 0000:00:14.0: Miss service interval error for slot 18 ep 2, set skip flag
> >     Jul 22 16:14:07 spiral kernel: xhci_hcd 0000:00:14.0: Miss service interval error for slot 18 ep 2, set skip flag
> >     Jul 22 16:14:07 spiral kernel: xhci_hcd 0000:00:14.0: Found td. Clear skip flag for slot 18 ep 2.
> >     Jul 22 16:14:52 spiral kernel: xhci_hcd 0000:00:14.0: Miss service interval error for slot 18 ep 2, set skip flag
> >     Jul 22 16:14:53 spiral kernel: xhci_hcd 0000:00:14.0: Miss service interval error for slot 18 ep 2, set skip flag
> >     Jul 22 16:14:53 spiral kernel: xhci_hcd 0000:00:14.0: Miss service interval error for slot 18 ep 2, set skip flag
> >     Jul 22 16:14:53 spiral kernel: xhci_hcd 0000:00:14.0: Found td. Clear skip flag for slot 18 ep 2.
> >     Jul 22 16:16:06 spiral kernel: xhci_hcd 0000:00:14.0: Miss service interval error for slot 18 ep 2, set skip flag
> >     Jul 22 16:16:06 spiral kernel: xhci_hcd 0000:00:14.0: Miss service interval error for slot 18 ep 2, set skip flag
> >     Jul 22 16:16:06 spiral kernel: xhci_hcd 0000:00:14.0: Found td. Clear skip flag for slot 18 ep 2.
> >     Jul 22 16:17:33 spiral kernel: xhci_hcd 0000:00:14.0: Miss service interval error for slot 18 ep 2, set skip flag
> >     Jul 22 16:17:33 spiral kernel: xhci_hcd 0000:00:14.0: Miss service interval error for slot 18 ep 2, set skip flag
> >     Jul 22 16:17:33 spiral kernel: xhci_hcd 0000:00:14.0: Miss service interval error for slot 18 ep 2, set skip flag
> >     Jul 22 16:17:33 spiral kernel: xhci_hcd 0000:00:14.0: Found td. Clear skip flag for slot 18 ep 2.
> 
> Please use 'journalctl -o short-precise' or simply 'dmesg' next time.

Will do.

> It seems these errors are happening in bursts, sporadically, but much
> more often than your failure rate, so 'bad' kernels handle most of them
> fine and only occasionally something goes wrong.
> 
> Per your other email, affected hardware is Intel 100 series chipset
> (8086:a12f ?) USB 3.0 controller. If you run 'dmesg | grep hci.version'
> I suppose you will see version 0x100, so it's somewhat old HW.

It is indeed old hardware, but still capable enough for my needs.
I get this:

    [    0.391721] xhci_hcd 0000:00:14.0: hcc params 0x200077c1 hci version 0x100 quirks 0x0000000001109810
    [    0.455073] xhci_hcd 0000:05:00.0: hcc params 0x0200eec0 hci version 0x110 quirks 0x0000000000800010

I suppose the other is another controller.

> Maybe don't bother with newer kernels for now, but simply checkout
> d0b619599e52 and also add this patch. Recompilation should be fast.
> Then run it to failure and post the log. You may want to run
> 'dmesg -W | tee logfile.txt' just in case, because the resulting flood
> may overwrite initial messages in the kernel log buffer and I'm not
> sure how much extra buffering journald maintains.

OK. Checked out that commit, applied the patch, and it's compiling.

(I haven't been able to figure out how to compile in-tree and then
have Nixos accept and install its outputs, so I've been compiling via
Nix's toolchain, since it just works and I can point it to a commit
ref on my clone of the repo, but it builds from scratch each time. I'm
sure it's possible to get it faster, just haven't figured it out. Last
time I tried, I spent longer tinkering than a full compilation cycle
would take before giving up.)

I'll report back with logs when the time comes.

At 2026-07-23 08:19:42 +0200, Michal Pecio wrote:
> Please use "Reply to all" and be careful to preserve Cc's, particularly
> maintainers of the affected driver. Re-adding Mathias Nyman.

Oh, sorry! I actually did that intentionally, since it was you
replying and not Mathias, in an attempt to not bother an extra person.
I see now that was a misstep. I'm new to the kernel mailing list ways.

Thanks again.