Re: Regression: webcam freezing since Linux 6.15

[Michal Pecio <michal.pecio@gmail.com>]

On Wed, 5 Aug 2026 14:11:20 -0700, Bart Nagel wrote:
> OK. Attached are the first ~700 lines of the trace file.
> 
> That whole thing is 17M, and is here for 48 hours from now in case it
> is more illuminating: https://tmpfiles.org/wAwAIQxuuahs/trace.txt

It seems that Missed Service Error pointers are about equally likely to
be 0x10 or 0x20 after the previous event.

And the events are usually Success, Short Packet is less common.

Considering that also roughly half of TDs are 0x20 long, we can choose
to be optimists and assume that (when it chooses not to report zero),
your HC actually reports correct addresses, except for the known narrow
problem case of MSE after Short Packet on an early TRB.

Or we can choose to be pessimists and say that 0x20 can be bogus too,
and then we wrongly skip two TDs instead of one, but three are always
missed at a time (for some reason) so no worry, until something will
change and suddenly only one TD is missed.

Not sure if I have any answer which option is true without another test
patch and compilation cycle. Maybe let's be optimistic if it works and
no errors are logged, only dynamic debug noise.

Regards,
Michal