Re: USB device enumeration failure after a certain amount of uptime

["Michał Pecio" <michal.pecio@gmail.com>]

> > Is this a complete log? And the device was really working during
> > those six minutes before "setup address" failure? Usually, address
> > setup happens once immediately after device connection, not sure
> > what would lead to it being repeated later.

Nevermind, it looks like address setup is also done on device reset,
so that's what may be happening here.

Although now I'm not sure what would request a reset of an idle device
and why. Apparently not uvcvideo/snd-usb-audio drivers because they
only ever seem to reset a few specific USB IDs. Maybe USB core did it
for some reason.

You could try enabling usbcore dynamic debug:
echo 'module usbcore +p' >/proc/dynamic_debug/control

and see if there is some noise in dmesg when the device is idle or
shortly before it fails. Note that this enables debug logging on all
USB buses in the system. Change +p to -p to turn it off.

> I also discovered that apparently if I stream from the capture card
> to some application it never disconnects, or at least I have never
> encountered it disconnecting.

I guess a simple test is to leave it running overnight with some webcam
viewer software or similar. If it survives, and then fails shortly
after turning off the software, that looks like an idle state problem.

Maybe autosuspend is involved? Totally guessing. It can be disabled:
echo -1 >/sys/bus/usb/devices/4-1/power/autosuspend_delay_ms

This is quite unlikely, I think, but sometimes those USB controllers
can break and stop performing transfers correctly. You could see if
things return to normal after re-initializing the controller:

echo 0000:08:00.0 >/sys/bus/pci/drivers/xhci_hcd/unbind
echo 0000:08:00.0 >/sys/bus/pci/drivers/xhci_hcd/bind

assuming that this card is 08:00.0 and not 09:00.0 (or something else).


Michal