Bug 215095
Summary: | USB ehci error -110 and soft lockup on kernel 5.15.4 as kvm guest | ||
---|---|---|---|
Product: | Drivers | Reporter: | wangyugui (wangyugui) |
Component: | USB | Assignee: | Default virtual assignee for Drivers/USB (drivers_usb) |
Status: | RESOLVED CODE_FIX | ||
Severity: | normal | ||
Priority: | P1 | ||
Hardware: | x86-64 | ||
OS: | Linux | ||
Kernel Version: | 5.15.4 | Subsystem: | |
Regression: | No | Bisected commit-id: |
Description
wangyugui@e16-tech.com
2021-11-22 01:28:18 UTC
althought we can work around this issue by revert 90d28fb53d4a(usb: core: reduce power-on-good delay time of root hub), but we still need to fix this soft lockup? more test show that this issue is not fully walked around by revert 90d28fb53d4a(usb: core: reduce power-on-good delay time of root hub). still some chance(>20%) to happen. Since this happens only when the kernel is running as a kvm guest but never happens when running on the bare metal, I would think that this is most likely a bug in kvm, not a bug in the kernel. However, if you want to track it down further, you could try different kernel versions between 5.10 and 5.15 to see at what point the trouble begins. But doing this requires that you get reliable testing results, which might not be easy if the problem occurs only 20% of the time. there are 2 problems. 1) USB ehci error -110(timeout) this is hard to happen on the bare metal. 2) soft lockup after USB ehci error -110 In fact, we do not care about 1) on KVM, because the USB device on KVM is not really used. maybe a problem of KVM, maybe a problem of USB. difficult to troubleshoot. Here's something you can do to help diagnose the soft lockup. When the lockup occurs, go to the /sys/kernel/debug/usb/ehci/* directory (fill in the "*" with the pathname for the EHCI controller that's locking up), make copies of the files you find there, and attach those copies to this bug report. Another thing you can do: Before the problem occurs, start a process doing cat /sys/kernel/debug/usb/usbmon/1u >mon.out and after the lockup happens, kill the "cat" process. Then attach the mon.out file to the bug report. There may be additional debugging options available in kvm, but I don't know what they are. This problem yet not happen on kvm/host 5.15.5. so let's mark it as kvm/host problem. although I checked the patches of 5.15.5, but failed to find out the patch which matched this problem directly. |