QubesOS (host) → Some Qube running Debian or Fedora (hypervised by Xen/QubeOS) → Windows VM (hypervised by KVM running within Some Qube w/ a GPU passed to it)
Let me explain, I’m trying to use a program known as “looking glass” to interface with a VM. Here’s how it would work on a regular Linux desktop like Debian or Arch:
-Install the looking glass client
-Create a IVSHMEM file (just a systemd temp file)
-Create a VM in libvirt w/ that passes through a GPU via IOMMU and passes the IVSHMEM file
-Launch the VM and looking glass client, the looking glass will replace the VM’s console and allow you to interface with the VM
This is a lot harder to do with Qubes is seems has you would either need to:
-Create a Qube, pass your GPU to it, then create a KVM/Qemu VM within the Qube that uses looking glass
or
-Somehow create a Qube/Xen VM that runs Windows 10/11 and uses a spice and then create another Qube running a Linux distro and somehow get a spice connection and IVSHMEM shared to that Qube (very unsupported by looking glass).
Which one should I try first?
I should note that I’m in a position where my threat model would allow for violations of what are generally considered good practices for enhanced security of Qubes (i.e. I’m less concerned of vulnerabilities being exploited to break into my system but rather taking advantage of the ability to isolate your online identities with different Qubes for different activities)