On Fri, Jul 03, 2015 at 03:50:19PM +0200, Paolo Bonzini wrote: > > > On 03/07/2015 15:08, Don Slutz wrote: > >>> > >> Creating it at the pc level and propagating makes code > >> messy but I'd go along with it if it made sense from > >> user's point of view, but it does not seem to make sense: > >> to me this looks more like a CPU feature than a machine property. > >> Or the property of the vmport rpc device that you created. > > > > Well this is not a clear area. It does look more like a CPU feature. I > > went with a machine feature because that is where "vmport" currently is. > > I think it should be a vmport property. > > It would be cleaner to implement it using the MemTxAttrs mechanism, but > for KVM I'm afraid it would be too slow (and the slowness would affect > all I/O accesses, not just vmport ones, unless really ugly hacks are done). > > Paolo
Can it ever be implemented for kvm? I couldn't think of a clean way to do it. -- MST