Re: vfio usage in a vm

Sent: Friday, June 14, 2024 at 9:33 AM From: "Michal Prívozník" <mprivozn(a)redhat.com&gt; To: "daggs" <daggs(a)gmx.com&gt; Cc: users(a)lists.libvirt.org Subject: Re: vfio usage in a vm On 6/13/24 18:18, daggs wrote: > Greetings Michal, > > I think I might be doing something wrong, all the devices are defined a such: > <interface type='network'> > <mac address='52:54:xx:xx:xx:xx'/> > <source network='mynet'/> > <model type='virtio'/> > <address type='pci' domain='0x0000' bus='0x10' slot='0x0x' function='0x0'/> > </interface> > > and I still cannot get vfio to attached to them. > they are autobinded to virtio_pci so I had to unbind them first > when I reload the vfio module, I get the same errors below > Ah, so I've misunderstood what you meant. Virtio and VFIO are two different approaches. While the former is for emulated devices, the latter is for PCI assignment (aka "PCI passthrough"). Your L1 guest is correctly binding the vNIC to virtio. > I'm trying to attach it inside a running guest, in short, I want to be able to pass a vNIC to a nested guest for development reasons This could be possible, but you'll need to give your L1 guest vIOMMU: <iommu model='intel'> <driver intremap='on'/> </iommu> https://libvirt.org/formatdomain.html#iommu-devices Then, to assign this L1 virtio vNIC to the L2 guest you can either do <hostdev/> or <interface type='hostdev'/>: https://libvirt.org/formatdomain.html#pci-passthrough