On 12/01/18 10:21, Peter Maydell wrote: > On 10 January 2018 at 11:25, Jean-Philippe Brucker > <jean-philippe.brucker@xxxxxxx> wrote: >> Hi Peter, >> >> On 10/01/18 11:19, Peter Maydell wrote: >>> Are there uses that make it worthwhile to get virtio-1 >>> support added to virtio-mmio, rather than just getting >>> people to move over to virtio-pci instead ? >> >> virtio-iommu uses virtio-mmio transport. It makes little sense to have an >> IOMMU presented as a PCI endpoint. > > Having an entire transport just for the IOMMU doesn't make > a great deal of sense either though :-) If we didn't already > have virtio-mmio kicking around would we really have designed > it that way? Possibly. It certainly was on the table during early investigations. It does beat the alternative, having to redesign firmware interfaces and rewrite core driver code to cater for unrealistic device topologies. Thanks, Jean _______________________________________________ kvmarm mailing list kvmarm@xxxxxxxxxxxxxxxxxxxxx https://lists.cs.columbia.edu/mailman/listinfo/kvmarm
