luzeal

The port I had in mind is this one [1], and the use case is to make it possible to run 3D accelerated Linux guests on macOS hosts. I may have wrongly believed that it would be useful for macOS guests.

Besides, it is correct that whatever happens on the driver side of things, Apple will have to approve it first (as the other reply pointed out)

[1] https://mail.gnu.org/archive/html/qemu-devel/2021-02/msg0423...

By complicated, I mean that you have to modify your the default behavior of your host OS (which by default will grab any hardware) and pick a GPU that is supported by the guest OS. People with Nvidia GPUs cannot do that for macOS. Ideally, out-of-the-box, with a few clicks, most people including non-tech savy ones should be able to enjoy 3D acceleration in their VM, regardless of their host configuration.

Besides, in the long term, Apple is likely to drop support for AMD or Intel GPU in their OS (but yes, they might also drop support for virtio devices).

Yes, evdev is currently the best way to share your inputs devices to a VM, thank you for pointing that out (alas, it is only available in distributions with very recent virtualization-related packages).

In my (perhaps narrow) view: virtualization should be about making you less reliant on the underlying hardware. I wish that GPU manufacturers would agree on a common standard to allow users to split their beefy GPU into smaller parts (Intel is apparently dropping support for vfio-mdev (Intel GVT-d), adopting SR/IOV instead for their latest offerings. But SR/IOV is only available on professional models for AMD and Nvidia. In summary, there won't be a standard any time soon, leaving virtio-gpu as the sole hardware-agnostic contender)

I believe the most promising route is to not try to do passthrough but to use a "virtual" gpu, aka a paravirtualized device. Thanks to the contributions of an open-source developer*, virtio-gpu has been ported to macOS (it already works on Linux but not on Windows and soon it will be on macOS). Apple is also reportedly working on bringing better support for macOS virtualized environments [2].

It means that users will be able to create any number of macOS instances with 3D acceleration, without having to care about trying to do complicated things (mediated devices or single-gpu passthrough or pci-passthrough of a second GPU).

As of now, the performance hit is high (more than 50%), but it will certainly improve over time, and many users might already find it acceptable (= people NOT playing recent video games).

Icing on the cake, thanks to the EGL-headless display (which is independent of virtio-gpu) and VNC, it is possible to connect to these instances remotely.

Shameless plug: I am working on Phyllome OS [3], a Fedora Remix with the goal to include the necessary plumbing to ease advanced virtualization techniques for end-users, and to offer curated virtual machine models.

I currently only use virtual machines as desktop machines, including locally, and even my poor x230 laptop can do it. I keep the host operating system intact (Phyllome OS), and use VMs as desktop machines. It has drawbacks (difficult USB access; displays models that were not designed for local use; sound support), but I firmly believe it will get better over time and, perhaps, one day, end-users will be able to easily deploy any modern OS on any modern hardware...

[1] Sorry, I could not find the link to the repository [2] https://passthroughpo.st/mac-os-adds-early-support-for-virti... [3] https://phyllo.me/

[edit] Delete mention of KVM/QEMU

It depends on your internal GPU. If it's an Intel GPU, you could use vfio-mdev, which allows you to split your physical GPU into smaller counterparts. It will only work with the macOS version that are compatible with Intel GPUs (probably most of them).

Have a look here for a guide on Archlinux : https://wiki.archlinux.org/title/Intel_GVT-g. And here for a script to deploy macOS on KVM : https://github.com/kholia/OSX-KVM.

Side-note : vfio-mdev can now be unlocked on consumer-grades Nvidia cards too : https://github.com/DualCoder/vgpu_unlock. Sadly, it is not possible for AMD-cards.

By the way, I am working on Phyllome OS (https://phyllo.me/), which is an attempt to make it easier to do such things. But please don't tell anyone :)