As the title indicates, I’m wondering what happened to the series on Qubes.
Did looking-glass eat it? Did the first video not do well enough? Do @wendell and co. not know that R4.0 came out months ago?
I was really looking forward to it as I have started dabbling with Qubes but my knowledge of systems is nowhere near as deep. Really hoping to benefit from some of their insights.
I would guess that is on hold because of some problems with the current version of qubes. At least I had some stability issues on my laptop.
Do you know if that’s more wide spread than your laptop? I’m embarking on a build that I wanted Qubes to be the software foundation of, it would be terrible news that it’s widely unstable.
No idea. That is just what happened to proxmox, freenas and other systems when they had subpar releases. So I assumed …
We tried 4.0rc, but it had a lot of bugs. 4.1 may be worth a revisit.
Since the video I’ve learned that at least some of the Qubes developers do not believe it is possible to securely do stuff like vfio.
I do want qubes + vfio, but there are some wrinkles.
Another thing that’s come up is some games do not like running in any vm – CS:go for example – cannot be run on software like esxi or unraid because you’re “always” inside a VM. This would be same w/Qubes.
Also, rootless/seamless virtualization is an unmitigated dumpster fire at present.
I sincerely believe that, if we are to witness the “year of the linux desktop” it will happen when a user can run a program to “pack their (windows) bags” and then just boot right up on linux without missing a beat. Yes, for a time, there will still be licensing fees involved to MS. But it buys time. This is, I am fairly sure, the flashpoint of Linux on the desktop.
When it is easy for a user to run their software, with no performance tradeoffs on cpu or gpu computations, simply and easily, inside a container that can be created and destroyed as needed, then we will see the near instant adoption of linux everywhere.
Qubes 4.0.1 is out now and, according to the news post on Qubes-OS.org, is “the first stable point release of Qubes (4.0).”
This isn’t the number you pointed out but it may be worth taking a peek if you are looking for an additional distraction 
Q:
Does Qubes support 3d acceleration via open source AMD drivers?
I guess what i’m trying to determine is whether or not i can replace ubuntu with qubes as a general purpose OS including Linux gaming… (VM assumptions mentioned by @wendell above aside…)
Sure but you’d have to set up passthrough.
Cheers, no so inbuilt support then.
Is it still the case that:
Vfio is not really explicitly supported because dma means no security?
And
3d acceleration isn’t really supported because… Dma means no security?
Thanks for confirming @wendell.
It’s a shame, because 3d acceleration is a requirement for many people including myself. I get the security concerns, but for a lot of end users i suspect that some sort of compromise for 3d acceleration would be a lot better than a general purpose OS.
If i could run applications with comparable driver support to Linux, i’d be all over this, but alas.
Even if the part of Qubes to just “open in disposable VM” and the API to enable file copy between VMs easily (so you could spin say firefox off into a VM sandbox) could be ported to general purpose Linux, that would help.
edit:
just realised (I mis-parsed “is it” as “it is” 
) you actually asked a question. i guess i have exactly the same query…
I did find this thread, including posts from Joanna Rutkowska
https://groups.google.com/forum/#!topic/qubes-devel/MfHy2jmXhXM
She mentions that GPU passthrough should work? (and one user in thread did claim to get it working, some time ago?)
It is, admittedly, quite old though. Maybe something to try out?
I might give it a go myself, but i’m a bit of a GPU-passthrough noob (and only ran Qubes once a couple of years ago to play with on an old inadequate machine to see what it was doing and how it worked) and don’t have it working on anything properly yet
Also, someone asking about it within the last 2 weeks and also getting responses indicating AMD should work (reset bug aside):
https://groups.google.com/forum/#!topic/qubes-devel/wbvr-_YrEFI
Passthrough should work, but it’s not the
kind of feature that you can set up in a GUI at this point Xen supports it so Qubes does as well.
3D acceleration works, but you can only assign a GPU to one VM so have multiple, although xfce shouldn’t require one.
@wendell Any chance that a workstation class card with SR-IOV would help with this situation? I don’t have one myself but I would love to see if that would work.
Pretty sure Xen supports SRIOV so it should work although it may be in a more recent version than Qubes is based off of. Would require more configuration finagleing regardless.
I recently checked up on the Qubes project, as I used to use it. I am led to believe that GPU passthough should work? As Qubes 4.x supports HVM with PCIE passthrough.
I believe that it may be tricky to passthrough an SSD or HDD directly to the VM, as I do not believe Xen Supports SATA passthrough. You could try to get the windows disk in an isolated IOMMU group and pass it through, likely using a PCIE to SATA card with JBOD passthrough.
If I am wrong about any of the above, someone please let me know. I was planning to try this myself soon after I work on some hardware kinks in my current rig, so seriously if its not remotely possible save me some trouble 
VFIO AMD driver bug of note for anyone trying AMD passthrough:
None of the 2019 radeon drivers have worked for me at all in VFIO with a Vega 64 (18.12.2+). That is in typical kvm situations, not qubes. Can confirm it is the AMD driver as I can boot the exact same SSD in vfio if I pass a nvidia card to it instead.
For testing VFIO passthrough with an AMD card, I would personally say 18.3.3 is a good test, as I used that driver in VFIO for a very long time. Reset bug still applies, however.
So don’t rule out AMD passthrough in Qubes if it doesn’t work, and you happened to have been using any driver 18.12.2+.
They claim to have added some feature that lets you do minor timing adjustments that would otherwise be done by reflashing the VGA bios. My guess is this feature is what broke VFIO, but I don’t really know.