On Wed, 6 Apr 2016 14:09:12 +0200
Paolo Bonzini <pbonz...@redhat.com> wrote:
> On 06/04/2016 13:11, Cornelia Huck wrote:
> >> > +static void virtio_queue_notify_aio_vq(VirtQueue *vq)
> >> > +{
> >> > + if (vq->vring.desc && vq->handle_aio_output) {
> >> > + VirtIODevice *vdev = vq->vdev;
> >> > +
> >> > + trace_virtio_queue_notify(vdev, vq - vdev->vq, vq);
> >> > + vq->handle_aio_output(vdev, vq);
> >> > + }
> >> > +}
> >> > +
> > So this avoids reentrancy, but might we miss one notify if
> > ->handle_aio_output has already been unset? What am I missing?
>
> Calling the notifier just before unset is handled by using "false,
> false" when unsetting the notifier, and only setting
> vq->handle_aio_output after the notifier has been unset.
Actually, the code is not quite right until the next two patches have
been applied, but I think we can live with that.
> Patch 7 makes things clearer.
That as well.
