Re: [RFC 5/8] vdpa: Add vdpa memory listener

[Eugenio Perez Martin]

On Tue, Aug 23, 2022 at 5:58 AM Jason Wang <jasowang@redhat.com> wrote:
>
> On Fri, Aug 19, 2022 at 6:35 PM Eugenio Perez Martin
> <eperezma@redhat.com> wrote:
> >
> > On Fri, Aug 19, 2022 at 11:01 AM Jason Wang <jasowang@redhat.com> wrote:
> > >
> > > On Fri, Aug 19, 2022 at 4:30 PM Eugenio Perez Martin
> > > <eperezma@redhat.com> wrote:
> > > >
> > > > On Fri, Aug 19, 2022 at 8:29 AM Jason Wang <jasowang@redhat.com> wrote:
> > > > >
> > > > > On Thu, Aug 11, 2022 at 2:42 AM Eugenio Pérez <eperezma@redhat.com> 
> > > > > wrote:
> > > > > >
> > > > > > This enable net/vdpa to restart the full device when a migration is
> > > > > > started or stopped.
> > > > > >
> > > > > > Signed-off-by: Eugenio Pérez <eperezma@redhat.com>
> > > > >
> > > > > I have the following questions
> > > > >
> > > > > 1) any reason that we need to make this net specific? The dirty page
> > > > > tracking via shadow virtqueue is pretty general. And the net specific
> > > > > part was done via NetClientInfo anyhow.
> > > >
> > > > The listener is only used to know when migration is started / stopped,
> > > > no need for actual memory tracking. Maybe there is a better way to do
> > > > so?
> > >
> > > Not sure, SaveVMHandlers?
> > >
> >
> > I'm fine with investigating this, but the only entry in the doc says
> > it's the "legacy way". I assume the "modern way" is through
> > VMStateDescription, which is in virtio-net.
>
> Right.
>
> >
> > The "pre_save" member already assumes the vhost backend is stopped, so
> > I'm not sure if this way is valid.
>
> I wonder if we can
>
> 1) new VhostOps
> 2) call that ops in vhost_log_gloabal_start/stop?
>

Bringing this thread up as I'm rebasing on top of the latest asid version.

We can detect that moment when set_features is called to set _F_LOG.
The problem is that it makes us duplicate the startup and teardown
code of the backend.

SVQ was proposed originally that way [1].

[1] https://lists.gnu.org/archive/html/qemu-devel/2021-05/msg06049.html

Thanks!