Create a Gaming HVM

Community Guides
, , ,
302

The issue does not seem to be a problem with passthrough, it seems to be an issue with the driver

If sudo lspci | grep -i nvidia in the VM shows the GPU, passthrough is working

You should forget about Qubes - your issue is with Linux and nvidia

What you need to do:

  • Read the Arch docs (even if not using Arch, their docs have excellent details and context l)
  • Create a new VM from a clean template (e.g. debian-12-xfce)
  • Follow the most authoritative documentation for setting up the drivers (you can choose between nvidia docs and debian docs)
303

Thanks, I’ll try that.

I’m currently can’t retrieve the IOMMU groups at all for some reason. I’m booting a live distribution with iommu=1 and iommu_amd=on params, I check afterwards that these params are loaded, but when I check for ls /sys/kernel/iommu_groups it shows nothing. I checked the dmesg and it says operation not permitted. I definitely remember it worked last time when I checked this, not much have changed I think.
Any ideas what I missed or forgot?

304

OK, I managed to retrieve the IOMMU groups. It says that I have my CPUs and dGPU combined in the same IOMMU group:

IOMMU Group 2:
  ... PCI bridge ...: Intel Corporation ... PCIe Controller (x16) ... (rev 05)
  ... 3D controller ...: NVIDIA Corporation ... (rev a1)

I did the ... to remove potential fingerprinting info.

I’m almost certain it shouldn’t be an issue, as the last letters (rev a1) appears only once in the whole list of devices and the (rev 05) appears multiple times.
But please, can someone just clarify this for me, so I won’t spend enormous time just to find out it’s impossible for me to passthrough this GPU?

305

Is this a config error on my part, or does switching to a TTY (inside the HVM) not work?

306

Hi,
So I have been spending this entire time trying and trying to get the GPU to work with support from AMD on the whole driver issue and still just a big mess. Not working.
Ive recently experienced today error when starting a qube with the connected gpu and get unkown PCI header type 127.
Now my device has not shutdown until after this error when attempting multiple reboots(6 to be percise) still persisting with this error.
When I got my issues with the Windows standalone sorted 2 days ago I tried with the GPU connected and didnt detect the device either.

307

Power the system off rather than reboot

For the error 127, that may be a result of it not handling a reset well. When attaching the GPU with qvm-pci, use

-o permissive=True -o no-strict-reset

Have you read the guide on this from start to finish?

308

Don’t pass through the bridge, pass through all of the devices in that group

309

Yes I have done this guide to the full extent and beyond.
This is a recent issue. Have not had this happen prior however then with prior tests for attaching the gpu there were just driver issues that is still not resolved and even with trying official support from AMD for last two months results still to no success.
The extent I get is I able able to get a display via a additional montior per the point of the post but not beyond that, even with VirtualGL it will only render with the virtual cpus.
This error is recent. I am aware if the system shutdowns there can be errors which may require multiple restarts to eventually avoid error based on what I read from others experience hence why I would only suspend. However such issue remains as a new matter.