On 1/27/22 14:57, Daniel Henrique Barboza wrote: > > > On 1/26/22 06:29, Michal Privoznik wrote: >> v2 of: >> >> https://listman.redhat.com/archives/libvir-list/2022-January/msg01111.html >> >> >> Patches 1/3 and 2/3 were ACKed but I'm sending them for completeness. > > > Back when I've added this UNASSIGNED address type I was also pursuing a > PCI multifunction > hotplug/unplug feature in Libvirt and, apparently, I didn't mind testing > hotplug/unplug > of non-multifunction unassigned hostdevs like the bug mentioned in patch > 1 did. > > > Thanks for fixing this up. yeah, this is more of a workaround than a real fix, because IIUC unassigned PCI address makes sense for devices within one IOMMU group. I mean, you want to passthrough one specific PCI device but because it's in IOMMU group with another one, both have to be detached from the host. Well, and for that we would need the hotplug/hotunplug API understand that and attach/detach two or more devices at once. With my patches you can still hotplug/hotunplug just a single device, but hey, at least we don't error out! Michal
