Passthrough VM not initializing

Looking through your VM’s XML, it looks like you don’t have the QXL attached. Are you removing that when you add the RX 580? If so, you could try leaving it attached until you can install the drivers for the RX 580 while it is attached, and then detach QXL. If you’re using a different virtual video display, I don’t think I’ll be much use here.

May not be the prettiest way to get it to work, but I haven’t had any issues with that process yet. Granted, I’m not on 4.15… Maybe worth a shot, though?

If nothing else, it might give you some wiggle room on viewing the logs in the VM without having to install an ssh server.

Leaving QXL attached was one thing I tried. As soon as you attach the GPU, it fails. The problem is that it’s not even making it to the Tianocore logo.

I’ve got a fairly large list of kernels I’m going to be trying to build today. The goal is that I’ll be able to get one of them to work.

I have 4.13.0-32-generic on Ubuntu 16.04 with an RX 580 passthrough. Let me know if you want to see the XML or anything. I have a bit of an odd setup on that one though, since I was using BIOS rather than UEFI.

1 Like

I appreciate it, I’m going to give 4.13.0-37 a go tomorrow when it finishes building.

In the mean time, it’s Friday. Time for a drink.

1 Like

this will sound weird but mem_encrypt=off
does that make a difference as kernel param?

mem_encrypt=off didn’t solve it.

Turns out, this is a distro-agnostic issue. I’m now running into the same issue on Fedora 27.

I’ve been trying different VBIOS on the GPU and it seems that with this asus bios, the GPU will pass through (it seems that way, disk and CPU activity seem to indicate that the system has started to boot), but will not display any output over HDMI the TV I’m using to test.

This is an indication to me that the bios on the GPU is problematic. I’m going to continue to brute-force the BIOS to see what happens, but I’m starting to question the sanity of this card. I’m also going to spend some time compiling that ACS patch I promised to another thread yesterday. All this debugging (rebooting) has made compilation impossible.