4 ms·

If you're privacy conscious and weren't aware, you can virtualize your gaming OS of choice with near native performance using VFIO. With this, you can contain W

If you're privacy conscious and weren't aware, you can virtualize your gaming OS of choice with near native performance using VFIO. With this, you can contain Windows telemetry and unwanted spyware into a disposable VM (if it weren't infringing, I'd vote a community created Windows gaming container be called "gladware"), freeing your host from unwanted reporting, data collection, and general "Microsoft has root" fuckery. All you need is a Linux host, some capable entry-level hardware, and two hours in the morning before the house is awake for the initial setup.

bmn__ 8y ago
> VFIO Stupid question (since there is no Wikipedia lemma for that term): is that the same as IOMMU? http://enwp.org/IOMMU_hardware_list http://enwp.org/IOMMU_hardware_list
- washadjeffmad 8y ago
  VFIO relies on IOMMU capability to allow devices to be passed through to guests, but it isn't just IOMMU. The common modern arrangement for VFIO is to use KVM+qemu with libvirt. Of practical concern is finding out which devices are in which IOMMU groups, because devices within the same group can't be split. For instance, if you had an ethernet controller with two ports, they'd both likely be in the same group, so you couldn't give one to the host and one to a VM. IOMMU info: https://heiko-sieger.info/iommu-groups-what-you-need-to-consider/ https://heiko-sieger.info/iommu-groups-what-you-need-to-cons... VFIO explanation: https://www.kernel.org/doc/Documentation/vfio.txt https://www.kernel.org/doc/Documentation/vfio.txt VFIO community: https://www.reddit.com/r/VFIO/ https://www.reddit.com/r/VFIO/ Archwiki: https://wiki.archlinux.org/index.php/PCI_passthrough_via_OVMF https://wiki.archlinux.org/index.php/PCI_passthrough_via_OVM...
deckard1 8y ago
This only works if you have two graphics cards, no? Either onboard graphics + dedicated GPU, or dual GPU. And some CPUs, such as first gen Ryzen (e.g. 1700), do not have built in graphics. And that assumes you always want your better GPU to run Windows rather than Linux.
- washadjeffmad 8y ago
  It can be run with a single graphics device. Just SSH in, then after you've done the other configurations and added the GPU's PCI ids to vfio-pci, create the VM with virsh, attaching the GPU. Run "virsh autostart [vmname]", and it'll spawn automatically at each boot. If you need a host with graphics, you can always unbind the GPU from the host and allow it to be re-attached to VMs if you don't need a headless hypervisor. I know I've seen a few small bash scripts on github for reference.

bmn__ 8y ago

> VFIO Stupid question (since there is no Wikipedia lemma for that term): is that the same as IOMMU? http://enwp.org/IOMMU_hardware_list http://enwp.org/IOMMU_hardware_list

washadjeffmad 8y ago
VFIO relies on IOMMU capability to allow devices to be passed through to guests, but it isn't just IOMMU. The common modern arrangement for VFIO is to use KVM+qemu with libvirt. Of practical concern is finding out which devices are in which IOMMU groups, because devices within the same group can't be split. For instance, if you had an ethernet controller with two ports, they'd both likely be in the same group, so you couldn't give one to the host and one to a VM. IOMMU info: https://heiko-sieger.info/iommu-groups-what-you-need-to-consider/ https://heiko-sieger.info/iommu-groups-what-you-need-to-cons... VFIO explanation: https://www.kernel.org/doc/Documentation/vfio.txt https://www.kernel.org/doc/Documentation/vfio.txt VFIO community: https://www.reddit.com/r/VFIO/ https://www.reddit.com/r/VFIO/ Archwiki: https://wiki.archlinux.org/index.php/PCI_passthrough_via_OVMF https://wiki.archlinux.org/index.php/PCI_passthrough_via_OVM...

deckard1 8y ago

This only works if you have two graphics cards, no? Either onboard graphics + dedicated GPU, or dual GPU. And some CPUs, such as first gen Ryzen (e.g. 1700), do not have built in graphics. And that assumes you always want your better GPU to run Windows rather than Linux.

washadjeffmad 8y ago
It can be run with a single graphics device. Just SSH in, then after you've done the other configurations and added the GPU's PCI ids to vfio-pci, create the VM with virsh, attaching the GPU. Run "virsh autostart [vmname]", and it'll spawn automatically at each boot. If you need a host with graphics, you can always unbind the GPU from the host and allow it to be re-attached to VMs if you don't need a headless hypervisor. I know I've seen a few small bash scripts on github for reference.