On 6/6/21 9:41 PM, Ricardo Ribalda Delgado wrote:
Hi Hans
Could you try running git bisect to figure out exactly at what commit
the device stops working?
https://git-scm.com/docs/git-bisect
Usually in 7 steps or so you will get an answer and it will really
help the debug process.
Thanks!
Hi Ricardo,
I did a quick bisect back to Linux 4.2 (Torvald's Linux), but was unable
to find the offending commit. All revisions didn't work. This was done
on FreeBSD. I'm pretty sure this has worked before, because I have some
microscope pictures saved on my computer from 7th of February 2020. That
UVC code was on Linux 5.10-rc6 (Torvald's Linux).
Then I didn't use the USB microscope for a year. And when I plugged it
again recently, I only got it working with MacOSX.
Then I made a USB trace, and found that Linux sent a lot of default
value parameters during attach.
When I removed those, the device started working again.
So maybe some timing issue then :-(
Anyways optimizing away those redundant parameters settings should be a
good thing to reduce stress on UVC devices during attach. See separate
patch I posted.
Thank you for your time!
--HPS
