Microsoft Promotes Windows Subsystem For Linux "WSL" To GA Status

[qarium]

https://lawstreetmedia.com/news/tech/github-microsoft-and-openai-sued-over-ai-tools-unprecedented-scale-of-software-piracy/

Microsoft is a criminal organisation.

[mercster]

https://lawstreetmedia.com/news/tech...ftware-piracy/

Microsoft is a criminal organisation.

You're impressing noone by LARP'ing as an angsty 1990's Linux enthusiast, bro. It's (EDIT: nearly) 2023.

[cynyr]

The question is: what's really missing?

SR-IOV support by mainstream GPUs. (or other natively fast way to get full hardware level access to the GPU from the guest)
An easy and fast (like natively fast) way to share files between host and guest such that windows doesn't really care that they aren't native files.
Anti-cheat in games that works even in a VM.

[Artim]

SR-IOV support by mainstream GPUs. (or other natively fast way to get full hardware level access to the GPU from the guest)
An easy and fast (like natively fast) way to share files between host and guest such that windows doesn't really care that they aren't native files.
Anti-cheat in games that works even in a VM.

Well, Anti-Cheat will probably never work in a VM. It has never been more than some whacky tech doing wild guesses that works more through sheer luck than anything else.

With SR-IOV I don't really see how that would benefit you compared to the already existing stuff like PCIe passthrough, libvirt.

File sharing between host and guest shouldn't be any problem. You should just be able to share folders. How much of native speed they have I can't tell, but if Windows cares about that or not it's entirely up to MS.

[ssokolow]

With SR-IOV I don't really see how that would benefit you compared to the already existing stuff like PCIe passthrough, libvirt.

According to Microsoft's introduction to it, "The SR-IOV specification from PCI-SIG defines the extensions to the PCI Express (PCIe) specification suite that enable multiple virtual machines (VMs) to share the same PCIe physical hardware resources."

In other words, sounds to me like it'd save the cost of buying two GPUs and a motherboard suitable to put them both in.

[Quackdoc]

According to Microsoft's introduction to it, "The SR-IOV specification from PCI-SIG defines the extensions to the PCI Express (PCIe) specification suite that enable multiple virtual machines (VMs) to share the same PCIe physical hardware resources."

In other words, sounds to me like it'd save the cost of buying two GPUs and a motherboard suitable to put them both in.

two or more depending on how many VMs you have

Well, Anti-Cheat will probably never work in a VM. It has never been more than some whacky tech doing wild guesses that works more through sheer luck than anything else.

Actually a few anticheats works in windows 11 when using secure memory protection. not entirely sure why, but I guess it probably adds a degree of protection against the hypervisor but its not SEV SNP.

With SR-IOV I don't really see how that would benefit you compared to the already existing stuff like PCIe passthrough, libvirt.

because PCIe Passthrough sucks, much better to have the window act like a fullscreen window on the guest, with only a single gpu needed, with sriov it can be a lot more integreated at the trade off of a bit of performance.

File sharing between host and guest shouldn't be any problem. You should just be able to share folders. How much of native speed they have I can't tell, but if Windows cares about that or not it's entirely up to MS.

the only solutions right now for this are 9pfs which is kinda slow, and virtiofs which works pretty decently but is a tad buggy.

[Artim]

According to Microsoft's introduction to it, "The SR-IOV specification from PCI-SIG defines the extensions to the PCI Express (PCIe) specification suite that enable multiple virtual machines (VMs) to share the same PCIe physical hardware resources."

In other words, sounds to me like it'd save the cost of buying two GPUs and a motherboard suitable to put them both in.

Well it would actually be a major improvement if you could share the GPU between host and guest with both having native access to it, including access to accelerators. But I'm not sure if that's included since articles are only talking about sharing between VMs. But current support is mainly limited to Windows, though Intel seems to be planning to upstream their drivers. In Q4'23 that is. And it seems to be limited to Xe graphics, ARC GPUs seem to be excluded right now. And indeed, both Nvidia and AMD only enable it in their data center GPUs, even though someone has already managed to activate it on an RTX card.

[Artim]

because PCIe Passthrough sucks, much better to have the window act like a fullscreen window on the guest, with only a single gpu needed, with sriov it can be a lot more integreated at the trade off of a bit of performance.

Ever tried the libvirt solution? It sounds like a passthrough in the opposite direction. E.g. they just added hardware acceleration for h.264 and h.265. Though I don't know if it works only with Linux guests or also with Windows guests.

[Quackdoc]

Ever tried the libvirt solution? It sounds like a passthrough in the opposite direction. E.g. they just added hardware acceleration for h.264 and h.265. Though I don't know if it works only with Linux guests or also with Windows guests.

ah you mean virgl? yeah thats only linux guests at the moment, and the preformance is pretty poor, a collabora dev is working on getting virgl venus working (vulkan instead of opengl) on qemu, when that happens their might be interest in getting that working on windows, but for now virtio-gpu is limited to linux guests.

[Artim]

ah you mean virgl? yeah thats only linux guests at the moment, and the preformance is pretty poor, a collabora dev is working on getting virgl venus working (vulkan instead of opengl) on qemu, when that happens their might be interest in getting that working on windows, but for now virtio-gpu is limited to linux guests.

Good to know. I read about it in the Arch wiki, but got pretty confused between it and LIBVF.IO. But good to know there's progress. Depending on when machines with Ryzens with integrated RDNA3 will become a thing, I'll be off of Windows and I hope that by then I'll be able to use it in a VM with decent performance. Sure, there's a hack that allows you to detach the host from an iGPU and attach the guest to it, but that seems much more complicated than necessary. And I don't know yet which apps I'll need to run on Windows because there's no way to stay on Linux.