Solutions
-
ghost82's post in [SOLVED] Can't passthrough 980 TI to Win 10 VM was marked as the answerI assume the vm is 'Streaming VM'?
Set it multifunction:
<hostdev mode='subsystem' type='pci' managed='yes'> <driver name='vfio'/> <source> <address domain='0x0000' bus='0x03' slot='0x00' function='0x0'/> </source> <rom file='/mnt/user/isos/NVIDA 980TI dump.rom'/> <address type='pci' domain='0x0000' bus='0x00' slot='0x06' function='0x0' multifunction='on'/> </hostdev> <hostdev mode='subsystem' type='pci' managed='yes'> <driver name='vfio'/> <source> <address domain='0x0000' bus='0x03' slot='0x00' function='0x1'/> </source> <address type='pci' domain='0x0000' bus='0x00' slot='0x06' function='0x1'/> </hostdev>
Dump your own vbios.
Jan 16 21:15:11 ETHAN-UNRAID kernel: pci 0000:03:00.0: can't claim BAR 6 [mem 0xfff80000-0xffffffff pref]: no compatible bridge window Probably in use by efifb
--> modify your syslinux config and add in the append line:
video=efifb:off
-
ghost82's post in I need help getting nVidia passthru for windows VM was marked as the answerCorrect, just add that bit to existing one and edit the correct "block", that with the green label (unraid with gui in this case); note that unless you have another primary gpu for your unraid, if you use attach vfio at boot, you will have no video output, if this is the case you may consider to boot unraid without gui (?)
You need to add the multifunction to the video portion, that with the source function at 0x00; then you need to change the bus of the audio portion, to be the same as that of the video, and you need to change the function of audio too, so that you have both video and audio on the same bus and slot, but different function.
Correct, no gui for unraid if you attach to vfio at boot, you can try without attaching to vfio at boot, it may work!
However I'm not sure what happens if you disable efifb, be prepared to have no video output for the unraid gui.
-
ghost82's post in Display resolution issues (800x600) with Nvidia 2070s GPU Passthrough was marked as the answerDid you installed the nvidia proprietary drivers?
Newest drivers allow consumer cards to be used in a vm.
-
ghost82's post in [Solved]VM stuck in D3 after a while was marked as the answerI think this was the most important thing you did, since you have that errors not on boot, but after a xx amount of time the vm was running, apparently random.
Anyway, glad that you (hopefully) solved your issue.
-
ghost82's post in Crashes after attempting GPU pass through. was marked as the answerIt seems hardware related, I would investigate ram as a first step.
-
ghost82's post in Booting into shell after moving HAOS image on Unraid VM was marked as the answerNo no no...sorry, just saw the other thread and you must change on vm xml from type raw to type qcow2.
Everithing will be fixed.
-
ghost82's post in libvirt crashing was marked as the answerHi, there's no xml in your diagnostics file, maybe you deleted the vm, saved diagnostics and created it again?
However, looking at the qemu log I can give advices about your latest booted vm "Windows 11"
2021-12-22T00:55:07.829477Z qemu-system-x86_64: vfio: Cannot reset device 0000:01:00.0, depends on group 14 which is not owned.
1. You were passing through the asmedia usb controller, but it cannot be resetted, even if it seems it's in a different iommu group (13) as your asmedia sata controller (iommu 14) at 01:00.1
The bridge at 01:00.2 should give no issue if not passed through.
I see you were already using acs override patch, you need to passthrough also the sata controller at 01:00.1 or if you can't, don't passthrough the usb controller.
2. You were not attaching to vfio the 6900 xt, I suggest to attach it, both video and audio at 0a:00.0 and 0a:00.1 (iommu groups 25 and 26).
You were not assigning audio and video correctly, they were currently assigned to 04:00.0 (video) and 05:00.0 (audio), you need to put them in a multifunction device, same bus, same slot, different function, like this (assigned both to bus 0x04):
<hostdev mode='subsystem' type='pci' managed='yes'> <driver name='vfio'/> <source> <address domain='0x0000' bus='0x0a' slot='0x00' function='0x0'/> </source> <address type='pci' domain='0x0000' bus='0x04' slot='0x00' function='0x0' multifunction='on'/> </hostdev> <hostdev mode='subsystem' type='pci' managed='yes'> <driver name='vfio'/> <source> <address domain='0x0000' bus='0x0a' slot='0x00' function='0x1'/> </source> <address type='pci' domain='0x0000' bus='0x04' slot='0x00' function='0x1'/> </hostdev>
-
ghost82's post in Windows 10 VM - GTX 2070 Super passthrough only PCIE 1.1 was marked as the answerI had a similar issue with windows 10 bare metal installation and a nvidia gtx titan black, but in my case gpu-z was reporting it running at gen2.
1. check the bios settings and make sure the pcie slot runs at gen3, mine was already set at gen3
2. tweak the windows registry: I remember I followed the description of the second message here:
https://forums.evga.com/How-to-force-PCIE-Gen-3-in-Windows-10-m2372915.aspx
And after rebooting gpu-z was reporting my gpu running at gen3.