Secondary gpu passthrough of amd rx580 to Linux vm problem

Software & Operating Systems Linux
1

Hi there,
I have a problem with vga passthrough of my amd rx580 gpu. It is about getting Error starting domain: Internal error: Unknown PCI header type ‘127’ when starting linux virtual machine in a virt-manager.
I believe it is a software problem of incorrect deinitialization in linux vm, because windows vm makes no problem for secondary passthrough.
Here is a video demonstration: https://youtu.be/Zg3cJHJtNzg

Any suggestions?
I do not know where exactly to report this issue, so created this post here hoping to get help.
Also I have created a post in reddit, no solution found there though.

2

Does it work the first time you start that VM after you reboot the host?

This error is improper deinitialization 99% of the time.

3

Yes, the first time it works correctly even in linux vm.
I also noticed, that this error disappears if suspend host to ram, but there is no output on passed through gpu.
Yes, seems like deinitialization issue. Is there a way how to defeat it?
Also checked with manjaro - the same result (unable to do secondary passthrough)
Also checked with nvidia gpu - no such problem there (secondary passthrough works normal)

4

Can anybody confirm my issue?
Maybe there is a proper place for reporting this (something like a vfio-pci driver bugtracking system or else component that is failing)?

5

The problem still exists. I have found identical thread here: https://forums.unraid.net/topic/76510-execution-error-internal-error-unknown-pci-header-type-127/
but without a solution.

6

Hey there Ashark I have added the “helpdesk” tag to your title. This helps with discoverability of your thread.

I can’t help you with your problem I don’t have experience with passthrough but I can help with that -

https://www.qemu.org/contribute/report-a-bug/
https://www.linux-kvm.org/page/Bugs

Depending on your bug you should file a bug report in one of those places. You can check if it was reported by other people first.

7

There is no solution. It’s a hardware problem.

Take it up with AMD, since they failed to properly implement bus resets.