Live data from Hacker News

Microsoft submits Linux kernel patches to enable complete Hyper-V on Linux

theregister.com

171–180 of 236 posts

Re: Microsoft submits Linux kernel patches to enable complete Hyper-V on Linux

#171

Earlier quoted context omitted.

>Everyone should be afraid Please don't FUD like this. Proprietary systems on top of Linux are nothing new and Microsoft certainly isn't the first company to have done it. If you have some real proof of their business plans then let's hear that rather than encouraging speculation and fearmongering.

It's not FUD. It's the Microsoft's never-changing strategy. I don't know how many times people need to experience it to remember this.

To be clear, the GP wasn’t making a claim about Microsoft’s strategy; what they’re claiming is that Linux is robust against such threats. Linux is constantly “Embraced” and “Extended” by all sorts of people, but it’s far too large a tent at this point for any entity, no matter how large, to gain the control necessary to “Extinguish” it.

(In this specific case, it’s because Linux has many in-kernel hypervisor systems, not just Microsoft’s. And at no point would they get rid of any of them, even if Microsoft’s is “better” in some sense. That’s not how Linux works: those systems are there to allow various groups to scratch their own itches, not as some central ideologically-driven top-down design. Microsoft can certainly add to Linux—anyone can—but nobody can force Linux to take away alternatives until it becomes dependent in some way on their particular code.)

Re: Microsoft submits Linux kernel patches to enable complete Hyper-V on Linux

#172
post #51

If we assume this is technically sound and so on, is there a political dimension that will be considered before merging this? Such as "this will decrease Linux marketplace penetration in this-and-this niche"? I know that Linux is not a company etc so a direct comparison is awkward, but there can still be such considerations weighing in, perhaps...?

How does enabling software to run on Linux decrease Linux marketplace penetration? If anything it would have the opposite effect as anyone running Hyper-V could now choose to host on Linux.

Re: Microsoft submits Linux kernel patches to enable complete Hyper-V on Linux

#173

>Windows 10 is on a path to becoming a hybrid Windows/Linux system This sounds promising, something closer to OSX would give apple competition for devs like myself that want a linux based system and a fluffy "it-just-works" gui

>> This sounds promising, something closer to OSX would give apple competition for devs like myself that want a linux based system and a fluffy "it-just-works" gui

No, don't. That seems to be what they want. It's not about running linux apps on Windows. It's about blending the two so that even Linux apps are dependent on Microsft APIs and such. What is the point of bringing native DX12 to linux if not that? BTW they floated that a while back and the kernel folks said no.

It's the "extend" phase they are in now. Once they extend linux and you jump on with that gui, then the work to extinguish the other options, or just dont care because they get to take in money for every user of MS/linux hybrid.

This is painfully obvious and a bunch of people on here are saying "STFU" unless you have proof of their intentions, which is a stupid argument given the history and current efforts to "extend".

Re: Microsoft submits Linux kernel patches to enable complete Hyper-V on Linux

#174

Sorry for the ignorance, but what does this mean? I understand from the article that the patch allows the full Hyper-V stack to run on a linux-based machine (as opposed to previously needing windows to run the root partition). But what does _that_ mean to have the ability to run Hyper-V entirely on linux? Is it just a "good to have options" thing?

It allows them to run a bare metal (type 1) hypervisor instead of a VM on top of Windows.

Hyper-V already is a bare metal (type 1) hypervisor. Admittedly it does use one (or more) of the guests, called the "parent partition", to handle some of the work for it. This change allows that parent partition to be Linux instead of Windows, but it does not make it any more of a bare metal hypervisor than it was before, unless you know something in addition to what the article says.

Re: Microsoft submits Linux kernel patches to enable complete Hyper-V on Linux

#175
post #51

If we assume this is technically sound and so on, is there a political dimension that will be considered before merging this? Such as "this will decrease Linux marketplace penetration in this-and-this niche"? I know that Linux is not a company etc so a direct comparison is awkward, but there can still be such considerations weighing in, perhaps...?

This only increases Linux's marketplace penetration as this allows Linux to be used in situations where previously on Windows could be used.

But Linux developers do not care about market penetration -- they simply want to create the best OS kernel. If this makes the kernel better, they will merge it.

Re: Microsoft submits Linux kernel patches to enable complete Hyper-V on Linux

#176

>Windows 10 is on a path to becoming a hybrid Windows/Linux system This sounds promising, something closer to OSX would give apple competition for devs like myself that want a linux based system and a fluffy "it-just-works" gui

>> This sounds promising, something closer to OSX would give apple competition for devs like myself that want a linux based system and a fluffy "it-just-works" gui No, don't. That seems to be what they want. It's not about running linux apps on Windows. It's about blending the two so that even Linux apps are dependent on Microsft APIs and such. What is the point of bringing native DX12 to linux if not that? BTW they…

> What is the point of bringing native DX12 to linux if not that?

Oh, easy: it’s so games written for DX12 on Windows—that are currently run on Linux using Wine—can have a direct, low-overhead GPU driver path, rather than Wine needing to translate DX12 calls into OpenGL/Vulkan calls first.

Of course, if you’re a game-engine developer, you know you can achieve the same by just writing your engine to target Vulkan instead of DX12.

But if you’re e.g. Steam, you’d love to see more games “automatically made” multi-platform (with comparable performance), so that you can sell them to your Linux userbase.

Microsoft themselves are—maybe surprisingly—in the same boat as Steam here. The various game studios they own publish multi-platform games. Those platforms don’t necessarily include Linux, but they do often include Android. And merging code into the upstream Linux kernel is one way to get it to appear in Android.

—————

This change would also allow the graphics of Linux programs running in a VM on Windows to be accelerated through virtualization (i.e. having the X/Wayland server on Linux act as a Windows DX12 client.) It’s analogous to the reason that some VM software offers special-purpose “host-guest filesystem” drivers to the guest, that allows the guest to just pass through filesystem requests at a high level, rather than needing a virtual block device or a networked-filesystem protocol.

This is the reason Microsoft themselves offered. I feel like there’s a lot more money on being able to easily port Windows games to Android, though.

Re: Microsoft submits Linux kernel patches to enable complete Hyper-V on Linux

#177
post #170
post #6

Earlier quoted context omitted.

Well, according to TFA, the patches were submitted as an RFC. Short version: it's not in Linux, this is basically the very first step in getting it into Linux, it's still gonna be probably several months at least before it makes it into the mainline kernel, and then however long after that for it to make it into your distro's kernel. FWIW, I think it was a little over 3 years from the first "RFC" for Wireguard until…

unrelated: what does TFA mean? someone else used it in another comment

The Freaking Article, probably.

Re: Microsoft submits Linux kernel patches to enable complete Hyper-V on Linux

#178

Earlier quoted context omitted.

You got it completely wrong. WSL1 was a system call translation layer, and everything worked directly on the Windows kernel, like Windows programs do under WINE. WSL2 is an standard Hyper-V virtual machine, no more no less, with a custom Linux kernel build and a ext4 root filesystem stored in a .vhdx image. Every process runs as a native Linux process under the Microsoft's Linux kernel, _inside_ the VM. What makes it…

And you got it all wrong :p. It’s not a “standard Hyper-V virtual machine” you can’t make one from scratch.

With "standard Hyper-V virtual machine" I mean that's like every other VM, because it runs under its own instance of vmwp.exe, spawned by vmms.exe, like every other one you may have in the system.

Re: Microsoft submits Linux kernel patches to enable complete Hyper-V on Linux

#179
post #176

Earlier quoted context omitted.

>> This sounds promising, something closer to OSX would give apple competition for devs like myself that want a linux based system and a fluffy "it-just-works" gui No, don't. That seems to be what they want. It's not about running linux apps on Windows. It's about blending the two so that even Linux apps are dependent on Microsft APIs and such. What is the point of bringing native DX12 to linux if not that? BTW they…

> What is the point of bringing native DX12 to linux if not that? Oh, easy: it’s so games written for DX12 on Windows—that are currently run on Linux using Wine—can have a direct, low-overhead GPU driver path, rather than Wine needing to translate DX12 calls into OpenGL/Vulkan calls first. Of course, if you’re a game-engine developer, you know you can achieve the same by just writing your engine to target Vulkan inst…

Right but the larger point is that it's less healthy long term since Microsoft controls DX, full stop. So it makes direct Vulcan support less appealing.

Re: Microsoft submits Linux kernel patches to enable complete Hyper-V on Linux

#180
If this reaches production status (e.g. with a new Hyper-V Hardware Qualification List and HHQL logo program like WHQL), it could be the biggest news for Linux since Dell started shipping a laptop with pre-installed Linux.

Such a logo program could provide OEMs with a test suite for Hyper-V with Linux hardware drivers, which means Microsoft could start contributing test cases to upstream Linux projects like CKI.

Most importantly, a Hyper-V for Linux logo qualification program could require that OEM/ODMs pass HHQL before shipping supported hardware, when they still have engineering resources allocated for system and device firmware fixes.

In short, Hyper-V for Linux has the potential for positive ripple effects throughout the supply chain for Linux "secured core" hardware. When combined with WSL2, Direct X paravirtualized graphics for Linux, and Azure Sphere (based on OpenEmbedded/Yocto), it's a major endorsement for Linux and the flexibility of Type-1, CPU-assisted virtualization pioneered by open-source Xen.

The Microsoft Linux license for IoT/embedded devices includes support for 10+ years of security updates, which will hopefully extend OEM firmware support timelines for edge hardware, https://www.platformsecuritysummit.com/2019/speaker/seay/

> This talk will cover ... device security from the chip to the Linux kernel, user application isolation, network communication, cloud interaction, and what it takes to keep a system secure for 13 years.

Post reply on HN