You are not logged in.
So I've been having an issue where USB hubs on my KVMs die occasionally, and only resume working after a reboot. Reconnecting them doesn't help.
This appears in the dmesg log when it happens:
kernel: xhci_hcd 0000:0f:00.3: xHCI host not responding to stop endpoint command
xhci_hcd 0000:0f:00.3: xHCI host controller not responding, assume dead
xhci_hcd 0000:0f:00.3: HC died; cleaning upFrom what I've found it may caused by the hubs misbehaving; but I don't consider buying new hardware a solution to a problem that can be solved in software.
A workaround I've found sufficient is that instead of rebooting the system, I can just reload the xhci_hcd module like so:
rmmod xhci_hcd
modprobe xhci_hcdSo I have a script running that greps dmesg for "xhci_hcd 0000:0f:00.3: HC died; cleaning up" and runs those commands when it founds that line.
Unfortunately it doesn't work anymore, because xhci_hcd is a built in module now so I can't rmmod it. I've postponed thinking about it until that change hits the LTS kernel and, well, it did.
So is there something else I can do to achieve the same effect?
Last edited by GOKOP (2025-02-12 10:41:32)
Offline
Tomorrow is today huh?
I don't know of any way to 'unload' that portion of the kernel aside from removing the module. As you know, its the point of a module. So your two options are a new kernel or a new device.
--- But this is undesirable behavior and I'm curious if its a hardware problem and if software could learn from it.
Are you willing to share some details about the device and its logs? I suspect a gremlin is burrowed between the KVM and handling of the USB host so some details of the virtual machine and the settings it runs at would be needed too.
- my nerd tuple is arch/systemd/ext4/nano/go/lurk
Offline
Post output of 'lsmod'.
Offline
You can unbind/rebind the corresponding PCI device, i.e. googling the thread title brings up: https://stackoverflow.com/questions/667 … nto-kernel does that not work?
@xerxes_ built-in modules won't show up there
Offline
A DDG search for 'xHCI host controller not responding, assume dead' lead to a thread on our forum where bind/unbind method was used to solve the issue.
Disliking systemd intensely, but not satisfied with alternatives so focusing on taming systemd.
clean chroot building not flexible enough ?
Try clean chroot manager by graysky
Offline
@V1del @Lone_Wolf
Thanks, I didn't see this. It's exactly what I wanted
@a7 Unfortunately I'm not sure what device logs are you referring to.
I can add though that when researching the issue originally I've found a suggested solution to disable USB autosuspend, which I'm sure I've checked that it was disabled. But now I've checked again out of curiosity and I see that it isn't - I'll try properly disabling it and see if the problem reoccurs.
Offline
@V1del @Lone_Wolf
Thanks, I didn't see this. It's exactly what I wanted@a7 Unfortunately I'm not sure what device logs are you referring to.
I can add though that when researching the issue originally I've found a suggested solution to disable USB autosuspend, which I'm sure I've checked that it was disabled. But now I've checked again out of curiosity and I see that it isn't - I'll try properly disabling it and see if the problem reoccurs.
I was just looking for hints about how the USB devices went away but suspending is suspect for sure. I wasn't aware of unbinding and rebinding in the kernel like that, so today I learned. Neat!
- my nerd tuple is arch/systemd/ext4/nano/go/lurk
Offline
Please always remember to mark resolved threads by editing your initial posts subject - so others will know that there's no task left, but maybe a solution to find.
Thanks.
Online
Done
Offline