Re: [Qemu-devel] [RFC 4/6] virtio/migration: Add VMStateDescription to VirtioDeviceClass

[Cornelia Huck]

On Wed, 24 Aug 2016 16:42:05 +0100
"Dr. David Alan Gilbert" <address@hidden> wrote:

> * Cornelia Huck (address@hidden) wrote:
> > On Wed, 24 Aug 2016 14:42:31 +0100
> > "Dr. David Alan Gilbert (git)" <address@hidden> wrote:

> > > diff --git a/include/hw/virtio/virtio.h b/include/hw/virtio/virtio.h
> > > index d2490c1..fd386ac 100644
> > > --- a/include/hw/virtio/virtio.h
> > > +++ b/include/hw/virtio/virtio.h
> > > @@ -124,8 +124,10 @@ typedef struct VirtioDeviceClass {
> > >       * must mask in frontend instead.
> > >       */
> > >      void (*guest_notifier_mask)(VirtIODevice *vdev, int n, bool mask);
> > > +    /* Saving and loading of a device; use *either* save/load OR vmsd */
> > 
> > Should we try to enforce this in some way? Then virtio_{save,load} can
> > call either/or instead of fallthrough which may have unintended
> > consequences...
> 
> I was thinking of doing that; but my intention is to kill off the save/load
> methods ASAP.

Fair enough, but doing a check for "did you by accident supply both?"
might be helpful during development.