On Sun, Apr 29, 2018 at 07:17:08AM -0400, Joseph Mayer wrote: > Hi Mike, > > First, thanks for vmm! > > One practical question: > > Did you think about PCI passthrough? If so, what's your perspective on > it? > > There were no previous mentions of it on misc@ so here it is. Also I > have no idea what degree of complexity this is. > > E.g., a VM guest can get exclusive control of a PCI device to lower PCI > device access latency (by cutting out the hypervisor/host OS as active > participant in the VM-PCI device interaction) or to enable the VM to > use a PCI device at all for the case that the host OS lacks support for > the device and hence can't virtualize it. > > I guess PCI passthrough also could work as a robustness or security > measure where you have a VM drive a device so the host OS would be less > likely to crash and burn for any reason that pertains to touching > that device. > > Thanks, > Joseph >
This requires IOMMU support first, which we don't have. There have been diffs floating around for that, but they are not committed yet. In other words, it's not something coming RSN. -ml
