gvt-linux VS VFIO-Mdev_Samples

Yeah i read that too, I also found this.. https://github.com/intel/gvt-linux/issues/64

Intel has already confirmed that GVT-g is essentially dead and not supported on their Iris/Xe or anything newer graphics.. We can also confirm this via their own drivers source..

https://github.com/intel/gvt-linux/blob/gvt-staging/drivers/...

On Intel iGPUs, there are two methods: GVT-g and GVT-d. GVT-g is basically creating virtual instances of the iGPU for use in VMs, while GVT-d is passing through an entire iGPU to the guest in the same way you would do with a normal GPU.

From the Intel GVTg_Setup_Guide;

Not really pass through, no. If CONFIG_DRM_I915_GVT is enabled in your kernel, you can use Intel's graphics virtualization system... basically a virtio style virtual device that shares the GPU between VM and host. IMO this is way more convenient than real passthrough, where the device is only available either to the VM or the host. The downside is that you don't get full performance in the VM.

"Intel GVT-g is a full GPU virtualization solution with mediated pass-through (VFIO mediated device framework based), starting from 5th generation Intel Core(TM) processors with Intel Graphics processors. GVT-g supports both Xen and KVM (a.k.a XenGT & a.k.a KVMGT). A virtual GPU instance is maintained for each VM, with part of performance critical resources directly assigned. The capability of running native graphics driver inside a VM, without hypervisor intervention in performance critical paths, achieves a good balance among performance, feature, and sharing capability."

https://github.com/intel/gvt-linux/wiki/GVTg_Setup_Guide

Wait a minute. GVT-g with 11th gen iGPUs upwards does work in linux guest? Are you sure about that? See this github issue for reference.

To be clear, I never said it was dead, only a dead end.

As for GVT-g and Xe, according to a post in this[0] issue by one of the Intel devs, Rocket Lake (Xe) is not getting support and only does GVT-d.

Also in the same issue, someone pointed out that Intel themselves have states as much here[1].

I hope I am proven wrong in the end and GVT-g comes to then entire Xe and ARC lineup. Intel's communication on this matter has been...lacking.

0: https://github.com/intel/gvt-linux/issues/190

1: https://www.intel.com/content/www/us/en/support/articles/000...

After days of trial and error, I could not get it to work, maybe one of you knows it. Currently, I try to setup GVT-d with KVM on my Dell XPS 13 2 in 1 7390, which has a i7-1065G7. AFAIK GVT-g is not supported, so I gave GVT-d a chance. The virtual machine is booting without any errors, but the display stays black. I only found this guide, but couldn't get it to work...

https://github.com/OpenMdev/VFIO-Mdev_Samples

> I'm not confused either about how GVM/mdev-gpu works or about its relation to libvf.io. It's not hard to read between the missing lines of your project roadmap.

You define types that are presented in mdevctl via a YAML config file the user can write themselves or via a JSON file. So LibVF.IO might as well be entirely ignorant to the fact that GVM is on the system. All it sees are different arbitrary mdev types that the user decides to create. Not sure where you're going with this comment.

>I read this before I ever wrote a reply, which you should have guessed because there was no other way to get any information. None of it tells anyone WHY this should use this instead of the bindings which have 100 developers on them, which have been battle tested for years, and for which the original author of VFIO wrote exhaustive, excellent manuals on the blog I linked earlier 7.5 years ago.

>What advantages does your system offer?

Well for starters I've never argued LibVF.IO is better than libvirt. In-fact when folks bring it up I generally say there's a lot we can learn from libvirt.