ghost82

Consider, when searching for gpus, the amd reset bug, i.e. older amd gpus (<series 6000) could not reset properly on vm shutdown/restart and this needs the whole server to be restarted.

Not when the vm is running. Moreover, when the vm is shutdown you need to detach the gpu from vfio driver to make it available again in the host.

Can you send me, in pm if you want, the reddit link?Just curious about what it's written. Update: found, but it doesn't seem sponsored in any way. It seems a simple review, not good, not too bad.

Sorry ignore this, the position you wrote is the right one! As far as the other issue, I'm sorry I didn't try but only reported some findings

It's normal that you have some video output when unraid boots, vfio attaches after. Yes, multifunction is applied correctly You didn't Attach diagnostics and the vbios file you are using. Note that if you dump the vbios using gpuz you still needs to remove the header.

Is the issue that you don't have internet inside the vm or the message guest agent not installed? If it's the latter, just install qemu guest agent into your linux vm. Package name could differ for different linux distribution, for example it can be qemu-guest-agent. After installation enable it, for example: systemctl enable qemu-guest-agent systemctl start qemu-guest-agent Then check if it's running correctly, for example: systemctl status qemu-guest-agent If you don't have internet change network type from e1000 to virtio or virtio-net or e1000-82545em

See if this applies: https://forums.unraid.net/topic/125626-fehler-mit-einer-win-10-vm-bei-unraid-seit-ver-6103/?do=findComment&comment=1145374

attach diagnostics

You edit it from the unraid gui, but editing the file is right too..can you attach the file?I think there could be some incorrect format.

As you can see from your command output efifb attaches to your gpu. By looking at your syslog you are not applying the correct kernel arguments you pasted in your post #1

Are you asking if it will work if you hybernate the vm from inside the guest instead of from the host?My reply is...I don't know But several users reported it working with virsh commands; dompmsuspend and dompmwakeup are virsh commands to be given from the host and the guest requires the guest agent installed. here are posts where I get some info: https://www.reddit.com/r/VFIO/comments/568mmt/saving_vm_state_with_gpu_passthrough/

mmm...this shouldn't happen..if hibernation is set to disk, the vm should report as shutdown and the gpu should be free for other uses. Did you enable sustend to disk in the xml? Check this, might help: https://forums.unraid.net/topic/130134-switching-from-gpu-passthrough-local-to-vnc-in-linux-pop/?do=findComment&comment=1184943

Attach a diagnostics and the output of command "cat /proc/iomem"

In addition to hot22shot suggestion, which I think it's necessary, otherwise you could get a code 12 error in windows, pay attention to the layout in the guest os; you can't have the audio of the gpu 2 in the same bus and slot of the video of gpu 1. Moreover addresses and multifunction are in the wrong place So change with this: <hostdev mode='subsystem' type='pci' managed='yes'> <driver name='vfio'/> <source> <address domain='0x0000' bus='0x01' slot='0x00' function='0x0'/> </source> <address type='pci' domain='0x0000' bus='0x00' slot='0x05' function='0x0' multifunction='on'/> </hostdev> <hostdev mode='subsystem' type='pci' managed='yes'> <driver name='vfio'/> <source> <address domain='0x0000' bus='0x02' slot='0x00' function='0x0'/> </source> <address type='pci' domain='0x0000' bus='0x00' slot='0x08' function='0x0' multifunction='on'/> </hostdev> <hostdev mode='subsystem' type='pci' managed='yes'> <driver name='vfio'/> <source> <address domain='0x0000' bus='0x01' slot='0x00' function='0x1'/> </source> <address type='pci' domain='0x0000' bus='0x00' slot='0x05' function='0x1'/> </hostdev> <hostdev mode='subsystem' type='pci' managed='yes'> <driver name='vfio'/> <source> <address domain='0x0000' bus='0x02' slot='0x00' function='0x1'/> </source> <address type='pci' domain='0x0000' bus='0x00' slot='0x08' function='0x1'/> </hostdev>

This could suggest a kernel issue: 6.10.3 --> kernel 5.15.46 6.11.x --> kernel 5.19.x

me too, and it could be a great idea to share results anonimously. For what it worth all I wrote has already been written here in the forum.

Restore the syslinux parameters you need and in particular video=efifb:off Set Pcie acs override to "both", restart the server and see if iommu group 16 is splitted with your video and audio splitted in a group without anything else.

Attach new diagnostics Fix the syslinux line, video=efifb:off is repeated 2 times.

Read above, it seems virtualization has been resetted to off in your bios, or you changed it.

One step back...sorry, looking at your diagnostics it seems iommu is not available (?), make sure virtualization is enabled in bios (vtx+vtd/svm+amd-v)

Tools -> System Devices; put checkmarks and reboot

your gpu is flagged as boot vga and it's not isolated, that's why it probably doesn't show. Attach it to vfio and use vbios file. Adding kernel parameters to syslinux conf may be required too.

Thanks for pointing it: it should be something introduced in newer os, as I always had a popup, but this happened in Catalina if I remember well. Weird that the code doesn't popup with a gui.

Seems related: https://www.reddit.com/r/FlowZ13/comments/vplgfg/tutorial_compeletely_get_rid_of_the_dumb_cpu/ that was for a z13, it could be the same issue.

If you are unsure if the issue comes from vm configuration or from a hardware failure I could suggest to test the card bare metal.