Re: [Qemu-devel] [PATCH v2 4/6] virtio-balloon: keep collecting stats on save/restore

[Roman Kagan]

On Thu, Sep 01, 2016 at 10:35:49AM +0200, Ladi Prosek wrote:
> On Fri, Aug 19, 2016 at 3:39 PM, Roman Kagan <address@hidden> wrote:
> > Upon save/restore virtio-balloon stats acquisition stops.  The reason is
> > that the fact that the (only) virtqueue element is being used by QEMU is
> > not recorded anywhere on save, so upon restore it's not released to the
> > guest, making further progress impossible.
> >
> > Saving the information about the used element would introduce unjustified
> > vmstate incompatibility.
> >
> > So instead just make sure the element is pushed before save, leaving the
> > ball on the guest side.  For that, add vm state change handler to
> > virtio-ballon which would take care of pushing the element if there is
> > one.
> >
> > Signed-off-by: Roman Kagan <address@hidden>
> > Cc: "Michael S. Tsirkin" <address@hidden>
> > Cc: Ladi Prosek <address@hidden>
> > Cc: Stefan Hajnoczi <address@hidden>
> > ---
> >  hw/virtio/virtio-balloon.c         | 27 ++++++++++++++++++++++-----
> >  include/hw/virtio/virtio-balloon.h |  1 +
> >  2 files changed, 23 insertions(+), 5 deletions(-)
> >
> > diff --git a/hw/virtio/virtio-balloon.c b/hw/virtio/virtio-balloon.c
> > index 6d4c57c..f00ad8e 100644
> > --- a/hw/virtio/virtio-balloon.c
> > +++ b/hw/virtio/virtio-balloon.c
> > @@ -88,10 +88,19 @@ static void balloon_stats_change_timer(VirtIOBalloon 
> > *s, int64_t secs)
> >      timer_mod(s->stats_timer, qemu_clock_get_ms(QEMU_CLOCK_VIRTUAL) + secs 
> > * 1000);
> >  }
> >
> > +static void balloon_stats_push_elem(VirtIOBalloon *s)
> > +{
> > +    VirtIODevice *vdev = VIRTIO_DEVICE(s);
> > +
> > +    virtqueue_push(s->svq, s->stats_vq_elem, s->stats_vq_offset);
> > +    virtio_notify(vdev, s->svq);
> > +    g_free(s->stats_vq_elem);
> > +    s->stats_vq_elem = NULL;
> > +}
> > +
> >  static void balloon_stats_poll_cb(void *opaque)
> >  {
> >      VirtIOBalloon *s = opaque;
> > -    VirtIODevice *vdev = VIRTIO_DEVICE(s);
> >
> >      if (!s->stats_vq_elem) {
> >          /* The guest hasn't sent the stats yet (either not enabled or we 
> > came
> > @@ -100,10 +109,7 @@ static void balloon_stats_poll_cb(void *opaque)
> >          return;
> >      }
> >
> > -    virtqueue_push(s->svq, s->stats_vq_elem, s->stats_vq_offset);
> > -    virtio_notify(vdev, s->svq);
> > -    g_free(s->stats_vq_elem);
> > -    s->stats_vq_elem = NULL;
> > +    balloon_stats_push_elem(s);
> >  }
> >
> >  static void balloon_stats_get_all(Object *obj, Visitor *v, const char 
> > *name,
> > @@ -414,6 +420,15 @@ static int virtio_balloon_load_device(VirtIODevice 
> > *vdev, QEMUFile *f,
> >      return 0;
> >  }
> >
> > +static void balloon_vm_state_change(void *opaque, int running, RunState 
> > state)
> > +{
> > +    VirtIOBalloon *s = opaque;
> > +
> > +    if (!running && s->stats_vq_elem) {
> > +        balloon_stats_push_elem(s);
> > +    }
> > +}
> > +
> >  static void virtio_balloon_device_realize(DeviceState *dev, Error **errp)
> >  {
> >      VirtIODevice *vdev = VIRTIO_DEVICE(dev);
> > @@ -436,6 +451,7 @@ static void virtio_balloon_device_realize(DeviceState 
> > *dev, Error **errp)
> >      s->dvq = virtio_add_queue(vdev, 128, virtio_balloon_handle_output);
> >      s->svq = virtio_add_queue(vdev, 1, virtio_balloon_receive_stats);
> >
> > +    s->change = qemu_add_vm_change_state_handler(balloon_vm_state_change, 
> > s);
> >      reset_stats(s);
> >  }
> >
> > @@ -444,6 +460,7 @@ static void virtio_balloon_device_unrealize(DeviceState 
> > *dev, Error **errp)
> >      VirtIODevice *vdev = VIRTIO_DEVICE(dev);
> >      VirtIOBalloon *s = VIRTIO_BALLOON(dev);
> >
> > +    qemu_del_vm_change_state_handler(s->change);
> >      balloon_stats_destroy_timer(s);
> >      qemu_remove_balloon_handler(s);
> >      virtio_cleanup(vdev);
> > diff --git a/include/hw/virtio/virtio-balloon.h 
> > b/include/hw/virtio/virtio-balloon.h
> > index 1ea13bd..d72ff7f 100644
> > --- a/include/hw/virtio/virtio-balloon.h
> > +++ b/include/hw/virtio/virtio-balloon.h
> > @@ -43,6 +43,7 @@ typedef struct VirtIOBalloon {
> >      int64_t stats_last_update;
> >      int64_t stats_poll_interval;
> >      uint32_t host_features;
> > +    VMChangeStateEntry *change;
> >  } VirtIOBalloon;
> >
> >  #endif
> > --
> > 2.7.4
> >
> 
> Hi Roman,
> 
> I talked to Michael Tsirkin and he agrees with merging this patch for
> 2.7.

I'm not happy with this patch: it tries to solve the problem on the
"save" side and therefore doesn't fix the bug when migrating from an
earlier QEMU version.

I wonder if we can do better and solve it on the "load" side.  (At first
I thought that your patch did that but on a closer look it turned out
not the case).

In particular, with Stefan's patch to restore VirtQueue->inuse, we
should be able to just rewind ->last_avail_idx by ->inuse during "load",
which AFAICS would also fix the bug.  What do you think?

> Could you please resubmit and use the set_status callback instead
> of adding another VM state change handler?
> 
> static void virtio_balloon_set_status(VirtIODevice *vdev, uint8_t status)
> {
>     VirtIOBalloon *s = VIRTIO_BALLOON(vdev);
> 
>     if (!vdev->vm_running && s->stats_vq_elem) {
>         balloon_stats_push_elem(s);
>     }
> }
> 
> and
> 
>     vdc->set_status = virtio_balloon_set_status;
> in virtio_balloon_class_init.

If the scheme I described above works out this won't be needed at all.

> This is somewhat urgent because 2.7 will be out soon. If you're busy
> or if I don't hear from you I'll post it on your behalf.

OK will focus on it now.

Thanks,
Roman.