You could try, but it might break other tweaks they have made to the qemu settings.

You can try it, but backup your libvirt.img before doing anything.

 

well i tried it and it does seem to resolve the problem. However, I tried this on an osx VM that booted past the SEABIOS before and now it doesnt, grrr.

Hey everyone,

 

I have a Supermicro server with onboard Matrox adapter that unRAID is using as its primary display adapter. I have 3 Geforce GTX 750ti GPU's that I am passing through to WIndows 10 VM's. Each adapter is in it's own IOMMU group. Two out of the three Geforce cards pass through without issue. The third produced the error that this hotfix describes. I have applied the hotfix and now the third VM does start but I get no graphical output and the output of the onboard Matrox adapter corrupts. Everything else still works and unRAID continues to function. Other VM's behave fine as well. My Geforce GPU's are in IOMMU groups 16, 17 and 34. The Matrox adpater is in group 13. Any one have any idea what is happening?

 

I should add, I have updated the BIOS and I have double checked that the Matrox adapter is set as the primary adapter. I have tried using PCI ACS override and I have tried using pci stub to claim the Geforce cards but all to no avail!

 

Any help would be appreciated!

 

John

Hey everyone,

 

I have a Supermicro server with onboard Matrox adapter that unRAID is using as its primary display adapter. I have 3 Geforce GTX 750ti GPU's that I am passing through to WIndows 10 VM's. Each adapter is in it's own IOMMU group. Two out of the three Geforce cards pass through without issue. The third produced the error that this hotfix describes. I have applied the hotfix and now the third VM does start but I get no graphical output and the output of the onboard Matrox adapter corrupts. Everything else still works and unRAID continues to function. Other VM's behave fine as well. My Geforce GPU's are in IOMMU groups 16, 17 and 34. The Matrox adpater is in group 13. Any one have any idea what is happening?

 

I should add, I have updated the BIOS and I have double checked that the Matrox adapter is set as the primary adapter. I have tried using PCI ACS override and I have tried using pci stub to claim the Geforce cards but all to no avail!

 

Any help would be appreciated!

 

John

If by corrupts you mean the screen goes black, you most likely have set the last VM to use seabios. Then you will have no more access to the console. This is normal behaviour caused by the VGA arbitration.

Thank you saarg,

 

Switching from seabios did the trick. Just to clarify in case anyone else reads this, by corrupt I meant that the text on the screen became garbled and unreadable. Works fine now though! 

 

Thank you!

 

John

Thank you saarg,

 

Switching from seabios did the trick. Just to clarify in case anyone else reads this, by corrupt I meant that the text on the screen became garbled and unreadable. Works fine now though! 

 

Thank you!

 

John

Glad it helped 

I thought you ment garbled, but I thought the console just went black. I learned something also 

Does anyone know if the official fix is included in 6.2.0 beta?

Can we get an update on this?

 

I'm on beta 21. My stock qemu.conf does not contain any of the iommu groups listed in the cgroup_device_acl section.

 

Passthrough using "hostdev" works fine. But if I try to use the "qemu-arg" method, I get the error from the first post. Manually adding the gpu iommu groups to the qemu.conf works.

 

So, was this technically fixed in beta 21? or not? If so, perhaps the original post should be updated since it is a sticky and is referred to by many people.

 

Thanks

Was this addressed in 6.1.9?  I am unable to get my GTX570 working in my VM.  I see the same message.

Was this addressed in 6.1.9?  I am unable to get my GTX570 working in my VM.  I see the same message.

 

No, not addressed in 6.1.9

This issue has been resolved in 6.2.  If you're still experiencing a problem after this release, please post a new support thread.