You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
This repository has been archived by the owner on Oct 3, 2024. It is now read-only.
When a guest VM uses a vGPU created by GVT-g, i915 send multiple "add" udev events for the intel gpu itself on the host.
This isn't only at launch or at the creation of the virtual GPU either, it happens randomly during operation of the VM.
This confuses some Wayland Compositors, because they think that the same GPU got added another time.
Here the output of udevadm monitor -k on the host while a VM is started, vgpu is created, the VM is running for a minute or two, and then being shutdown again and the vpu removed.
Notice the 4x add /devices/pci0000:00/0000:00:02.0/drm/card0 (drm) that happen while the VM is running.
I'm currently running into an issue that I can't easily search about it but maybe related to this issue.
I enabled virtualisation.kvmgt in NixOS 24.05 and use qemu-system-x86_64 to run a Windows VM that installed GVT-g related GPU driver which was created long time ago. Yes, I used to have this setup running.
But now, every time this VM passed the Windows logo, from when it started to use GVT-g GPU.
All application window in my Sway desktop becomes invisible.
I noticed that Sway now recognized two output devices which have exactly same name (output name like eDP-1) and make (vendor model like Sharp Corporation) .
All my current window goes into that invisible output device. I need to drag them back.
After a while, it sometimes continues to create more and more invisible output devices.
Sign up for freeto subscribe to this conversation on GitHub.
Already have an account?
Sign in.
When a guest VM uses a vGPU created by GVT-g, i915 send multiple "add" udev events for the intel gpu itself on the host.
This isn't only at launch or at the creation of the virtual GPU either, it happens randomly during operation of the VM.
This confuses some Wayland Compositors, because they think that the same GPU got added another time.
Here the output of
udevadm monitor -k
on the host while a VM is started, vgpu is created, the VM is running for a minute or two, and then being shutdown again and the vpu removed.Notice the 4x
add /devices/pci0000:00/0000:00:02.0/drm/card0 (drm)
that happen while the VM is running.The text was updated successfully, but these errors were encountered: