Am Mittwoch, den 06.01.2021, 17:13 +0000 schrieb David Mytton: > I'd like to report a bug in the Realtek driver for a RTL8153 10/100/1000 > Ethernet adapter built-in to a USB-C connected Philips 272B7QUPBEB 27" QHD LCD > Monitor. It is not limited to the monitor as I have seen other similar reports of > USB Ethernet adapters with the same issue using the same driver. > > I have reproduced this on Kernel 5.10.2 and 5.9.11, both Manjaro, but have seen > reports across multiple distros. > > Symptoms: > > - After a period of time following a reboot, ethernet connectivity suddenly > drops. Wi-Fi might take over, but not reliably. > - There is a kernel trace in the journalctl output (see below). The trace indicates that runtime power management is waiting for something, but not for what. > - Attempting to do anything related to the network causes a hang e.g. running > inxi hangs after CPU output. > - The only way to recover is to hard reset. reboot command also hangs. The processes are in state D (uninterruptible wait) > - Set usbcore.autosuspend=-1 pci_aspm=off kernel params. This has solved the > issue for me; the system has been stable with several days of uptime. That switches off runtime PM. The unsolved question is what the kernel is waiting for. Or whether we have a deadlock. Have you tried a kernel with lockdep enabled? Could you try to get a lock with dynamic debugging enabled for the module usbcore? In this case I don't think a usbmon trace is going to help much, as the kernel is stuck. Regards Oliver
