VFIO GPU Passthrough stops working when vGPU host driver is installed

Host OS: RHEL 8.4
vGPU host driver: 460.91.03
libvirt: 6.0.0-35.1
qemu-kvm: 15:4.2.0-48

I am attempting to run multiple VMs on the same host. This host contains two Titan RTX cards and two Tesla V100 cards. Prior to setting up vGPU, I was only running two VMs. Each of those two VMs had a Titan RTX passed through to it. When I install the vGPU host driver, the passthrough of the Titan RTX cards to the existing VMs stops working. The VMs won’t start and the libvirtd process hangs with an input/output error message when I run “systemctl status libvirtd”. It seems like the vGPU host driver is grabbing my Titan RTX cards but I have passed the VFIO PCI dummy driver to them at boot time. See /etc/default/grub file below. Any assistance would be greatly appreciated.

GRUB_DISTRIBUTOR="$(sed ‘s, release .*$,g’ /etc/system-release)"
GRUB_CMDLINE_LINUX="crashkernel=auto resume=/dev/mapper/rhel-swap rd.lvm.lv=rhel/root rd.lvm.lv=rhel/swap rhgb quiet modprobe.blacklist=nouveau intel_iommu=on iommu=pt vfio-pci.ids=10de:1e02,10de:10f7,10de:1ad6

Hi Brad,

Thanks for the question. This is not a configuration we would ever QA. The vGPU host driver is likely detecting that the Titan RTX GPU is not compatible with vGPU and not initializing correctly on them as a result.