hey everyone
I decided to go with unraid for my home server and so far so good. I wanted to have an Ubuntu VM that I could use for development that had a GPU passed through for some extra speed.
I got a VM set up with an SSD passed through and my RX 560 passed through. I didn’t realize that AMD GPUs had a re-initialization problem. Is there any way to fix it?
I found this guide for Windows, but I was wondering if anyone knew how to do it in a Linux Guest.
Some AMD GPUs have the reset bug. My MSI RX580 doesn’t, but my R9 Fury does.
I don’t think there’s a Linux Equivalent.
@tkoham might be able to comment better on this, since he’s more of a passthrough guru than I.
I’m running an Asus dual fan card. Probably an OEM PCB.
there’s no documented way of doing this on a linux guest as of yet, but that doesn’t mean it isn’t possible. If the card is new I’d suggest you RMA and look around for one that isn’t affected by the bug.
Good news, I think I got it working.
I switched the Machine field in unraid to i440fx-2.11 from Q35-2.11 and it seemed to let me shut down and restart the system without having to power cycle the host.
Interesting.
@tkoham I haven’t seen this variant of the bug, have you?
@Avendor7 Can you post your output of lspci -knn and your VM’s xml? I’m curious about this one.
Yeah, switching to q35 can fix the problem intermittently, but usually leaves it with lock-on-crash (blackscreen, 100% fans, etc)
Switching to or from?
Looks like OP moved q35 -> i440fx.
either or, it doesn’t consistently fix things enough either way, and it tends to cause other issues with the gpu.
if the issue resurfaces (as is likely) you can also manually remove the card from the virtman gui before shutting down to avoid the reset bug.
ok, good to know. Thanks!