On Tue, Mar 06, 2012 at 03:59:53PM +0100, Paolo Bonzini wrote: > Il 06/03/2012 15:55, Michael S. Tsirkin ha scritto: > > > It is not a problem if the destination does not have > > > VIRTIO_BALLOON_F_MUST_TELL_HOST. In that case, the guest > > > will simply do useless virtqueue traffic, but the destination > > > does not have a problem. > > > (In fact, it _is_ a problem if the destination has > > > VIRTIO_BALLOON_F_MUST_TELL_HOST but the source does not. > > > The feature bit should have been backwards! Luckily, > > > our implementation is free from this problem). > > > > > > Signed-off-by: Paolo Bonzini <pbonz...@redhat.com> > > > > Can't we just add a flag to control this feature? > > I think the sane thing here would be to remove it from the spec. > > Paolo
I guess we could but what does it buy us? -- MST