Re: [RFC PATCH v2 4/4] virtio-net: pass Device-TLB enable/disable events to vhost

[Viktor Prutyanov <viktor@daynix.com>]

On Mon, May 8, 2023 at 8:26 AM Jason Wang <jasowang@redhat.com> wrote:
>
> On Mon, May 1, 2023 at 10:02 AM Viktor Prutyanov <viktor@daynix.com> wrote:
> >
> > If vhost is enabled for virtio-net, Device-TLB enable/disable events
> > must be passed to vhost for proper IOMMU unmap flag selection.
>
> The patch looks good, just wonder if you have tested it with vhost-user?

Not yet, because I don't know which use case to test. Actually, the
device_iotlb_enabled flag logic doesn't rely on a backend. And that's
really enough to fix the problem for all the guests I know at the
moment. The trigger logic which turned out to be backend-aware is
needed only if ATS is enabled/disabled in runtime and I create such a
situation manually and test that IOMMU flag re-registration is done.

Thanks,
Viktor Prutyanov

>
> (It looks to me like it should work there).
>
> Thanks
>
> >
> > Signed-off-by: Viktor Prutyanov <viktor@daynix.com>
> > ---
> >  hw/net/vhost_net.c      | 11 +++++++++++
> >  hw/net/virtio-net.c     |  8 ++++++++
> >  include/net/vhost_net.h |  2 ++
> >  3 files changed, 21 insertions(+)
> >
> > diff --git a/hw/net/vhost_net.c b/hw/net/vhost_net.c
> > index c4eecc6f36..2364c8de99 100644
> > --- a/hw/net/vhost_net.c
> > +++ b/hw/net/vhost_net.c
> > @@ -552,6 +552,17 @@ int vhost_net_set_mtu(struct vhost_net *net, uint16_t mtu)
> >      return vhost_ops->vhost_net_set_mtu(&net->dev, mtu);
> >  }
> >
> > +void vhost_net_toggle_device_iotlb(struct vhost_dev *dev, bool enable)
> > +{
> > +    const VhostOps *vhost_ops = dev->vhost_ops;
> > +
> > +    if (!vhost_ops->vhost_toggle_device_iotlb) {
> > +        return;
> > +    }
> > +
> > +    vhost_ops->vhost_toggle_device_iotlb(dev, enable);
> > +}
> > +
> >  void vhost_net_virtqueue_reset(VirtIODevice *vdev, NetClientState *nc,
> >                                 int vq_index)
> >  {
> > diff --git a/hw/net/virtio-net.c b/hw/net/virtio-net.c
> > index 53e1c32643..e6851b885c 100644
> > --- a/hw/net/virtio-net.c
> > +++ b/hw/net/virtio-net.c
> > @@ -3843,6 +3843,13 @@ static struct vhost_dev *virtio_net_get_vhost(VirtIODevice *vdev)
> >      return &net->dev;
> >  }
> >
> > +static void virtio_net_toggle_device_iotlb(VirtIODevice *vdev,
> > +                                           bool enable)
> > +{
> > +    if (vdev->vhost_started)
> > +        vhost_net_toggle_device_iotlb(virtio_net_get_vhost(vdev), enable);
> > +}
> > +
> >  static const VMStateDescription vmstate_virtio_net = {
> >      .name = "virtio-net",
> >      .minimum_version_id = VIRTIO_NET_VM_VERSION,
> > @@ -3948,6 +3955,7 @@ static void virtio_net_class_init(ObjectClass *klass, void *data)
> >      vdc->vmsd = &vmstate_virtio_net_device;
> >      vdc->primary_unplug_pending = primary_unplug_pending;
> >      vdc->get_vhost = virtio_net_get_vhost;
> > +    vdc->toggle_device_iotlb = virtio_net_toggle_device_iotlb;
> >  }
> >
> >  static const TypeInfo virtio_net_info = {
> > diff --git a/include/net/vhost_net.h b/include/net/vhost_net.h
> > index c37aba35e6..36d527f321 100644
> > --- a/include/net/vhost_net.h
> > +++ b/include/net/vhost_net.h
> > @@ -56,4 +56,6 @@ int vhost_net_virtqueue_restart(VirtIODevice *vdev, NetClientState *nc,
> >                                  int vq_index);
> >
> >  void vhost_net_save_acked_features(NetClientState *nc);
> > +
> > +void vhost_net_toggle_device_iotlb(struct vhost_dev *dev, bool enable);
> >  #endif
> > --
> > 2.35.1
> >
>