OSDN Git Service

virtio-net migration fix (Mark McLoughlin)
authoraliguori <aliguori@c046a42c-6fe2-441c-8c8c-71466251a162>
Wed, 7 Jan 2009 17:50:45 +0000 (17:50 +0000)
committeraliguori <aliguori@c046a42c-6fe2-441c-8c8c-71466251a162>
Wed, 7 Jan 2009 17:50:45 +0000 (17:50 +0000)
We are failing to save whether the guest will supply us rx
buffers using the new mergeable format; this can cause a
migrated guest to crash with:

   virtio-net header not in first element

Bump the savevm version number and refuse to load v1 saves
just to be on the safe side.

Signed-off-by: Mark McLoughlin <markmc@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
git-svn-id: svn://svn.savannah.nongnu.org/qemu/trunk@6221 c046a42c-6fe2-441c-8c8c-71466251a162

hw/virtio-net.c

index 4443ff5..de1f7ce 100644 (file)
@@ -275,19 +275,21 @@ static void virtio_net_save(QEMUFile *f, void *opaque)
 
     qemu_put_buffer(f, n->mac, 6);
     qemu_put_be32(f, n->tx_timer_active);
+    qemu_put_be32(f, n->mergeable_rx_bufs);
 }
 
 static int virtio_net_load(QEMUFile *f, void *opaque, int version_id)
 {
     VirtIONet *n = opaque;
 
-    if (version_id != 1)
+    if (version_id != 2)
         return -EINVAL;
 
     virtio_load(&n->vdev, f);
 
     qemu_get_buffer(f, n->mac, 6);
     n->tx_timer_active = qemu_get_be32(f);
+    n->mergeable_rx_bufs = qemu_get_be32(f);
 
     if (n->tx_timer_active) {
         qemu_mod_timer(n->tx_timer,
@@ -324,7 +326,7 @@ PCIDevice *virtio_net_init(PCIBus *bus, NICInfo *nd, int devfn)
     n->tx_timer_active = 0;
     n->mergeable_rx_bufs = 0;
 
-    register_savevm("virtio-net", virtio_net_id++, 1,
+    register_savevm("virtio-net", virtio_net_id++, 2,
                     virtio_net_save, virtio_net_load, n);
 
     return (PCIDevice *)n;