* [Qemu-devel] [PATCH] virtio-balloon: stats vq migration spec compliance
@ 2016-08-15 21:32 Michael S. Tsirkin
2016-08-16 7:31 ` Ladi Prosek
2016-08-16 10:30 ` Stefan Hajnoczi
0 siblings, 2 replies; 4+ messages in thread
From: Michael S. Tsirkin @ 2016-08-15 21:32 UTC (permalink / raw)
To: qemu-devel; +Cc: Ladi Prosek
virtio spec says that devices must not touch VQs
before DRIVER_OK is set.
Additionally, balloon must not touch stats VQ
unless the stats feature bit has been negotiated.
Cc: Ladi Prosek <lprosek@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
---
hw/virtio/virtio-balloon.c | 4 +++-
1 file changed, 3 insertions(+), 1 deletion(-)
diff --git a/hw/virtio/virtio-balloon.c b/hw/virtio/virtio-balloon.c
index 65457e9..7189260 100644
--- a/hw/virtio/virtio-balloon.c
+++ b/hw/virtio/virtio-balloon.c
@@ -427,6 +427,7 @@ static void virtio_balloon_vmstate_cb(void *opaque, int running,
RunState state)
{
VirtIOBalloon *s = opaque;
+ VirtIODevice *vdev = VIRTIO_DEVICE(s);
if (!running) {
/* put the stats element back if the VM is not running */
@@ -436,7 +437,8 @@ static void virtio_balloon_vmstate_cb(void *opaque, int running,
s->stats_vq_elem = NULL;
}
- } else {
+ } else if (balloon_stats_supported(s) &&
+ (vdev->status & VIRTIO_CONFIG_S_DRIVER_OK)) {
/* poll stats queue for the element we may have discarded
* when the VM was stopped */
virtio_balloon_receive_stats(VIRTIO_DEVICE(s), s->svq);
--
MST
^ permalink raw reply related [flat|nested] 4+ messages in thread
* Re: [Qemu-devel] [PATCH] virtio-balloon: stats vq migration spec compliance
2016-08-15 21:32 [Qemu-devel] [PATCH] virtio-balloon: stats vq migration spec compliance Michael S. Tsirkin
@ 2016-08-16 7:31 ` Ladi Prosek
2016-08-16 10:30 ` Stefan Hajnoczi
1 sibling, 0 replies; 4+ messages in thread
From: Ladi Prosek @ 2016-08-16 7:31 UTC (permalink / raw)
To: Michael S. Tsirkin; +Cc: qemu-devel
On Mon, Aug 15, 2016 at 11:32 PM, Michael S. Tsirkin <mst@redhat.com> wrote:
> virtio spec says that devices must not touch VQs
> before DRIVER_OK is set.
>
> Additionally, balloon must not touch stats VQ
> unless the stats feature bit has been negotiated.
>
> Cc: Ladi Prosek <lprosek@redhat.com>
> Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Looks good! Do you want me to add this to the patch that introduces
virtio_balloon_vmstate_cb or will you apply it on top of it?
Thanks,
Ladi
^ permalink raw reply [flat|nested] 4+ messages in thread
* Re: [Qemu-devel] [PATCH] virtio-balloon: stats vq migration spec compliance
2016-08-15 21:32 [Qemu-devel] [PATCH] virtio-balloon: stats vq migration spec compliance Michael S. Tsirkin
2016-08-16 7:31 ` Ladi Prosek
@ 2016-08-16 10:30 ` Stefan Hajnoczi
2016-08-23 19:12 ` Michael S. Tsirkin
1 sibling, 1 reply; 4+ messages in thread
From: Stefan Hajnoczi @ 2016-08-16 10:30 UTC (permalink / raw)
To: Michael S. Tsirkin; +Cc: qemu-devel, Ladi Prosek
[-- Attachment #1: Type: text/plain, Size: 1854 bytes --]
On Tue, Aug 16, 2016 at 12:32:55AM +0300, Michael S. Tsirkin wrote:
> virtio spec says that devices must not touch VQs
> before DRIVER_OK is set.
>
> Additionally, balloon must not touch stats VQ
> unless the stats feature bit has been negotiated.
>
> Cc: Ladi Prosek <lprosek@redhat.com>
> Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
> ---
> hw/virtio/virtio-balloon.c | 4 +++-
> 1 file changed, 3 insertions(+), 1 deletion(-)
>
> diff --git a/hw/virtio/virtio-balloon.c b/hw/virtio/virtio-balloon.c
> index 65457e9..7189260 100644
> --- a/hw/virtio/virtio-balloon.c
> +++ b/hw/virtio/virtio-balloon.c
> @@ -427,6 +427,7 @@ static void virtio_balloon_vmstate_cb(void *opaque, int running,
> RunState state)
> {
> VirtIOBalloon *s = opaque;
> + VirtIODevice *vdev = VIRTIO_DEVICE(s);
>
> if (!running) {
> /* put the stats element back if the VM is not running */
> @@ -436,7 +437,8 @@ static void virtio_balloon_vmstate_cb(void *opaque, int running,
> s->stats_vq_elem = NULL;
> }
>
> - } else {
> + } else if (balloon_stats_supported(s) &&
> + (vdev->status & VIRTIO_CONFIG_S_DRIVER_OK)) {
> /* poll stats queue for the element we may have discarded
> * when the VM was stopped */
> virtio_balloon_receive_stats(VIRTIO_DEVICE(s), s->svq);
Another suspect case is when the feature bit was negotiated,
VIRTIO_CONFIG_S_DRIVER_OK is set, but the virtqueue PFN wasn't set up.
In that case virtqueue_pop() will access junk values.
I'm not sure if anything terrible can happen but a check that
vq->vring.desc != 0 would be nice - just like we do in
virtio_queue_notify_vq().
Can you add a virtio_queue_get_addr() call to see if the PFN has been
set?
Stefan
[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 473 bytes --]
^ permalink raw reply [flat|nested] 4+ messages in thread
* Re: [Qemu-devel] [PATCH] virtio-balloon: stats vq migration spec compliance
2016-08-16 10:30 ` Stefan Hajnoczi
@ 2016-08-23 19:12 ` Michael S. Tsirkin
0 siblings, 0 replies; 4+ messages in thread
From: Michael S. Tsirkin @ 2016-08-23 19:12 UTC (permalink / raw)
To: Stefan Hajnoczi; +Cc: qemu-devel, Ladi Prosek
On Tue, Aug 16, 2016 at 11:30:16AM +0100, Stefan Hajnoczi wrote:
> On Tue, Aug 16, 2016 at 12:32:55AM +0300, Michael S. Tsirkin wrote:
> > virtio spec says that devices must not touch VQs
> > before DRIVER_OK is set.
> >
> > Additionally, balloon must not touch stats VQ
> > unless the stats feature bit has been negotiated.
> >
> > Cc: Ladi Prosek <lprosek@redhat.com>
> > Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
> > ---
> > hw/virtio/virtio-balloon.c | 4 +++-
> > 1 file changed, 3 insertions(+), 1 deletion(-)
> >
> > diff --git a/hw/virtio/virtio-balloon.c b/hw/virtio/virtio-balloon.c
> > index 65457e9..7189260 100644
> > --- a/hw/virtio/virtio-balloon.c
> > +++ b/hw/virtio/virtio-balloon.c
> > @@ -427,6 +427,7 @@ static void virtio_balloon_vmstate_cb(void *opaque, int running,
> > RunState state)
> > {
> > VirtIOBalloon *s = opaque;
> > + VirtIODevice *vdev = VIRTIO_DEVICE(s);
> >
> > if (!running) {
> > /* put the stats element back if the VM is not running */
> > @@ -436,7 +437,8 @@ static void virtio_balloon_vmstate_cb(void *opaque, int running,
> > s->stats_vq_elem = NULL;
> > }
> >
> > - } else {
> > + } else if (balloon_stats_supported(s) &&
> > + (vdev->status & VIRTIO_CONFIG_S_DRIVER_OK)) {
> > /* poll stats queue for the element we may have discarded
> > * when the VM was stopped */
> > virtio_balloon_receive_stats(VIRTIO_DEVICE(s), s->svq);
>
> Another suspect case is when the feature bit was negotiated,
> VIRTIO_CONFIG_S_DRIVER_OK is set, but the virtqueue PFN wasn't set up.
>
> In that case virtqueue_pop() will access junk values.
>
> I'm not sure if anything terrible can happen but a check that
> vq->vring.desc != 0 would be nice - just like we do in
> virtio_queue_notify_vq().
>
> Can you add a virtio_queue_get_addr() call to see if the PFN has been
> set?
>
> Stefan
DRIVER_OK is set after VQs are initialized so I think that's enough.
^ permalink raw reply [flat|nested] 4+ messages in thread
end of thread, other threads:[~2016-08-23 19:12 UTC | newest]
Thread overview: 4+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2016-08-15 21:32 [Qemu-devel] [PATCH] virtio-balloon: stats vq migration spec compliance Michael S. Tsirkin
2016-08-16 7:31 ` Ladi Prosek
2016-08-16 10:30 ` Stefan Hajnoczi
2016-08-23 19:12 ` Michael S. Tsirkin
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.