Random Freeze when gpu is attached.
noideaman opened this issue · 4 comments
Running Rocky linux 8.4 with a GTX1070
I have vgpu host drivers 11.4 (nvidia 450.124)
Guest windows 10 21h1 Qudaro drivers 452.96
At first i thought this was a macvtap issue as i have had this similar issue on my old environment with out any gpu pass-through of any kind. Already made the network migration to bridged.
When running heaven benchmark the vm will lockup. running a suspend and un-suspend will unlock the vm faster sometimes.
Other times it will unlock it's self.
I have tested with disk read/writing with crystal disk mark. i have tested network performance with iperf3.
only when the gpu is attached (just had it freeze on me with nothing but NVIDIA control panel and parsec running) does the vm lock up.
No outputs from journalctl on any nvidia services
No outputs on dmesg
No outputs on libvirtd service
No change of output on nvidia-smi on host.
See attached recording for the lockups.
https://user-images.githubusercontent.com/39884745/140425062-594d5fe9-dd93-4e99-b2f7-fadf227fd578.mp4
got some output when i manually pause and unpause the vm after a freeze i get hit with a long list of
nvidia-vgpu-mgr[160993]: op_type: 0x90960101 failed
in the nvidia-gpu-mgr service and nothing else
still get random freezes using rocky, updated to latest kernel on 8.5 now.
Any ideas on where to look. only affects vm's with the gpu passed through with vgpu unlock and the nvidia driver running
Hey, I'm having the same issue. I'am using RTX 2080 super with unlock script. When gpu initialized vm freezes. Did you find a way to fix this ? @noideaman
I have not