On Tue, Feb 14, 2023 at 08:02:08AM +0100, Eugenio Perez Martin wrote:
> On Tue, Feb 14, 2023 at 7:31 AM Jason Wang <jasow...@redhat.com> wrote:
> >
> > On Tue, Feb 14, 2023 at 3:19 AM Eugenio Pérez <epere...@redhat.com> wrote:
> > >
> > > VIRTIO_F_ORDER_PLATFORM indicates that memory accesses by the driver and
> > > the device are ordered in a way described by the platform. Since vDPA
> > > devices may be backed by a hardware devices, let's allow
> > > VIRTIO_F_ORDER_PLATFORM.
> > >
> > > Signed-off-by: Eugenio Pérez <epere...@redhat.com>
> > > ---
> > > hw/virtio/vhost-shadow-virtqueue.c | 1 +
> > > 1 file changed, 1 insertion(+)
> > >
> > > diff --git a/hw/virtio/vhost-shadow-virtqueue.c
> > > b/hw/virtio/vhost-shadow-virtqueue.c
> > > index 4307296358..6bb1998f12 100644
> > > --- a/hw/virtio/vhost-shadow-virtqueue.c
> > > +++ b/hw/virtio/vhost-shadow-virtqueue.c
> > > @@ -34,6 +34,7 @@ bool vhost_svq_valid_features(uint64_t features, Error
> > > **errp)
> > > switch (b) {
> > > case VIRTIO_F_ANY_LAYOUT:
> > > case VIRTIO_RING_F_EVENT_IDX:
> > > + case VIRTIO_F_ORDER_PLATFORM:
> >
> > Do we need to add this bit to vdpa_feature_bits[] as well?
> >
>
> If we want to pass it to the guest, yes we should. I'll send another
> patch for it.
>
> But I think that should be done on top / in parallel actually.
>
> Open question: Should all vdpa hardware devices offer it? Or this is
> only needed on specific archs?
>
> Thanks!
I don't get what this is doing at all frankly. vdpa has to
pass VIRTIO_F_ORDER_PLATFORM to guest at all times - unless
- it's a x86 host where it kind of works anyway
- it's vduse which frankly is so slow we can do VIRTIO_F_ORDER_PLATFORM anyway.
In short VIRTIO_F_ORDER_PLATFORM has nothing to do with the device
and everything with qemu itself.
Yea we can allow VIRTIO_F_ORDER_PLATFORM from kernel but given
we never did at this point it will need a protocol feature bit.
I don't think it's worth it ..
--
MST
