Virglrenderer 0.8 Offers Better Open-Source OpenGL Support To KVM/QEMU Guests

[tildearrow]

They are already working on that

https://gitlab.freedesktop.org/virgl...-/milestones/2

Awesome!

Also, is it called VirGL or Virgil? It seems that only the features page calls it Virgil.

[torbido]

Somebody should start a project to support DirectX 11 on VM via Vulkan based on DXVK.

They are already working on that

https://gitlab.freedesktop.org/virgl/virglrenderer/-/milestones/2

[GreenByte]

I don't agree with your first point.



Vanilla CS:GO doesn't really fit my definition of a game with client-side anti-cheat software. Most others have issue (see below).

Here are my findings: VMware, Virtualbox, Unraid(KVM) and some others are detected by popular Anti-cheat for the past 1-3 years. Stock KVM with VFIO on Ubuntu/Fedora/Arch runs Rust (EAC), CS:GO (ESEA), Arma3/PUBG (BattlEye), and Battlefield 1 (FairFight?) in KVM. It should be quite easy to detect my gaming VM, but it would seem Anti-cheat software is only going after commercial solutions for now.

I have tried to reached out to ESEA. I explained use my system the way that I do and wanted their take on the situation, but have yet to receive feedback from them. That said, vanilla CS:GO is the game that has the most cheaters in my experience and or region. I watch a ton of overwatch and most competitive games that I play with this config: https://steamcommunity.com/sharedfil.../?id=226647583 it really helps to spot the script kiddies.

I don't want to admit this... In theory I agree with Valve about training A.I. to detect cheaters. In current reality though it seems that generally speaking client-side anti-cheat is working better. There are obviously an array of issues that I have with client-side anti-cheat, like backdoor/privacy issues, false positives that have led people to loose their accounts and general bad experiences like game disconnections or out of date EAC Linux clients that prevents you from playing legal games that you have payed for!

That makes total sense, I wasn't aware ACs blocked other VM systems, as I was only using KVM/QEMU.

And I agree that valve's approach doesn't currently work, but it will with time. At the moment they seem to be still playing it safe and putting the player to overwatch queue instead of insta-bannimg them. One possible solution would be to make the detection real-time and have a realtime overwatch. That would allow to ban the player while in game, but we have no idea how much processing time the detection NN needs.

[Jabberwocky]

Anti-cheats usually pretty good in detecting the fact that program is running in virtual machine, so virgl not gonna help here anyway. We need either native releases with native EAC/BattleEye, or some solution from Valve.

I don't agree with your first point.

Protected games work just fine in VMs. It was only VAC that disallowed csgo VM play for a while, but that was only because VMs were abused for account boosting and the restriction was sort of removed once they solved the issue in other ways.

Vanilla CS:GO doesn't really fit my definition of a game with client-side anti-cheat software. Most others have issue (see below).

Here are my findings: VMware, Virtualbox, Unraid(KVM) and some others are detected by popular Anti-cheat for the past 1-3 years. Stock KVM with VFIO on Ubuntu/Fedora/Arch runs Rust (EAC), CS:GO (ESEA), Arma3/PUBG (BattlEye), and Battlefield 1 (FairFight?) in KVM. It should be quite easy to detect my gaming VM, but it would seem Anti-cheat software is only going after commercial solutions for now.

I have tried to reached out to ESEA. I explained use my system the way that I do and wanted their take on the situation, but have yet to receive feedback from them. That said, vanilla CS:GO is the game that has the most cheaters in my experience and or region. I watch a ton of overwatch and most competitive games that I play with this config: https://steamcommunity.com/sharedfil.../?id=226647583 it really helps to spot the script kiddies.

I don't want to admit this... In theory I agree with Valve about training A.I. to detect cheaters. In current reality though it seems that generally speaking client-side anti-cheat is working better. There are obviously an array of issues that I have with client-side anti-cheat, like backdoor/privacy issues, false positives that have led people to loose their accounts and general bad experiences like game disconnections or out of date EAC Linux clients that prevents you from playing legal games that you have payed for!

[pal666]

I don't see anything stating VirGL is doing DX.

it doesn't. it does gallium, nine can convert dx to gallium

it probably will through DX-over-Vulkan layers imho

virgl knows no vulkan and gallium knows no vulkan

[starshipeleven]

i think it's gallium, so dx should be provided by nine

I don't see anything stating VirGL is doing DX. All I see (as is also this test in the article/blog) is for OpenGL.
Again, I'm not saying this can't happen (it probably will through DX-over-Vulkan layers imho), just that right now It does not support DX.

[RussianNeuroMancer]

i think it's gallium, so dx should be provided by nine

It's a possibility, but it's not implemented at this moment.

[pal666]

You aren't missing much, virgl is (at the moment) worthless for Windows as afaik there is no client drivers for Windows, and even if there are it's OpenGL 3.3 only, no DX at all.

i think it's gallium, so dx should be provided by nine

[starshipeleven]

lol I actually forgot about virgl, and just went ahead and used GPU passthrough for my VM. Obviously that takes more work, but, at least I get pretty much bare-metal GPU performance.

You aren't missing much, virgl is (at the moment) worthless for Windows as afaik there is no client drivers for Windows, and even if there are it's OpenGL 3.3 only, no DX at all.

(yes, making the big stab in the dark assumption that the VM is Windows as that's the most likely thing you would need a true GPU for)

[pabloski]

Any update on QEMU Windows Direct3D guest support?

There is zero support.

Parallels recently got Direct3D 11 implemented in Metal for Mac OS w/ Windows 10 guests. Perhaps VMWare Workstation has something planned next for Linux hosts?

?? Vmware already has Direct3D acceleration for guest under Linux hosts. Performances aren't good. I tried to use it to run Unity Editor. Now I run it under Qemu/Kvm with GPU passthrough.