Sys-usb 53% CPU usage if left open for more than 7 days

Testing Team 4.2 Testing
1

Basically out of nowhere (it would seem), sys-usb and sys-net as a USB Qube seems to suddenly spike once it’s been left open after several days (7 days, I think….)

This is where it gets weird:

  • Fans spin up all of a sudden and stay constantly on.
  • Any USB HID devices “stutter” (the cursor moves around the screen, similar to watching a movie with dropped frames, that’s the best way I can describe it…)
  • htop in sys-usb says that everything is normal, and CPU usage is at 5%
  • The process with the highest usage seems to be qrexec-client
  • It should be noted that systemd-journald is the next-highest process
  • xl top in dom0 says that CPU usage is ~50%
  • After this happening a few times over the past couple of weeks, I got curious and left the machine on overnight this time, to see whether it would sort itself out. The fans were still continuously running at the same speed in the morning, with the same reported CPU usage.
  • qvm-kill sys-usb; sleep 10; qvm-start sys-usb seems to fix it (disposable VM)

Whilst I have already discovered a “fix” (a very hacky fix…), my intention is to assist in finding the part of the codebase that is causing this, to assist the devs.

I’m happy to provide any machine info and/or logs (it is a testing machine, after all), but I honestly have no idea where to start….

4 Likes
2

Is it sys-net that is using CPU or is it the stub domain (sys-net-dm) that went to high usage?

B

3

I have a feeling that it’s the stub domain.

top inside sys-usb shows that everything is normal, but xl top in dom0 shows ~50% CPU usage.

4

It happened again, and I have discovered that it is NOT the stubdom.

It appears to be kworker inside sys-net.

I will post outputs when they happen again.

5

It may be worth looking at dmesg in sys-net when that happens. Maybe some device is malfunctioning?

6

I have dmesg dumping to a text file. I’ll upload it the next time it happens.

I suspect you’re right about it being a faulty piece of hardware, but I’ll post my logs just in case.