Hi Takashi,
Takashi Iwai writes:
This patch itself is safe and good to have, so I'm going to take it as is.
Thanks!
But I'm still curious in which code path the problem happens. That is, we should address such unnecessary repeats if possible. Do you have some more data?
Unfortunately not, I've been running with a kernel testing some other mm changes recently so haven't had a chance to debug what's going on here. But it starts like this:
[Fri Apr 21 13:21:10 2023] usb 3-7: USB disconnect, device number 39 [Fri Apr 21 13:21:10 2023] usb 3-7.1: USB disconnect, device number 40 [Fri Apr 21 13:21:10 2023] usb 3-7.1.1: USB disconnect, device number 42 [Fri Apr 21 13:21:10 2023] usb 3-7.4: cannot submit urb (err = -19) [Fri Apr 21 13:21:10 2023] usb 3-7.4: Unable to submit urb #2: -19 at snd_usb_queue_pending_output_urbs [Fri Apr 21 13:21:10 2023] usb 3-7.4: cannot submit urb 0, error -19: no device [Fri Apr 21 13:21:10 2023] usb 3-7.4: cannot submit urb 0, error -19: no device [Fri Apr 21 13:21:10 2023] usb 3-7.4: cannot submit urb 0, error -19: no device [Fri Apr 21 13:21:10 2023] usb 3-7.4: 1:0: usb_set_interface failed (-19) [Fri Apr 21 13:21:10 2023] usb 3-7.4: 2:0: usb_set_interface failed (-19) [Fri Apr 21 13:21:10 2023] usb 3-7.1.2: USB disconnect, device number 44 [Fri Apr 21 13:21:10 2023] usb 3-7.1.4: 1:1: usb_set_interface failed (-19) [Fri Apr 21 13:21:10 2023] usb 3-7.1.4: 1:1: usb_set_interface failed (-19) [Fri Apr 21 13:21:10 2023] usb 3-7.1.4: 1:1: usb_set_interface failed (-19) [Fri Apr 21 13:21:10 2023] usb 3-7.1.4: 1:1: usb_set_interface failed (-19) [Fri Apr 21 13:21:10 2023] usb 3-7.1.4: 1:1: usb_set_interface failed (-19) [Fri Apr 21 13:21:10 2023] usb 3-7.1.4: 1:1: usb_set_interface failed (-19) [Fri Apr 21 13:21:10 2023] usb 3-7.1.4: 1:1: usb_set_interface failed (-19) [Fri Apr 21 13:21:10 2023] usb 3-7.1.4: 1:1: usb_set_interface failed (-19) [... thousands of messages ensue ...]
So it looks like maybe some ordering is not quite right in destruction, perhaps only on multi-level USB topologies?
Thanks,
Chris