fix: unrealize virtio device if we fail to hotplug it
commit7abea552aba6e85b338015726648974d6d6f19c8
authorlinzhecheng <linzhecheng@huawei.com>
Tue, 31 Oct 2017 08:03:03 +0000 (31 16:03 +0800)
committerMichael S. Tsirkin <mst@redhat.com>
Thu, 16 Nov 2017 15:46:53 +0000 (16 17:46 +0200)
tree6a3c7a933426e2a77d5af5a317f1ca34b6dd272d
parent6a7cb8c3d674815cab08d884740d203fded12249
fix: unrealize virtio device if we fail to hotplug it

If we fail to hotplug virtio-blk device and then suspend
or shutdown VM, qemu is likely to crash.

Re-production steps:
1. Run VM named vm001
2. Create a virtio-blk.xml which contains wrong configurations:
<disk device="lun" rawio="yes" type="block">
  <driver cache="none" io="native" name="qemu" type="raw" />
  <source dev="/dev/mapper/11-dm" />
  <target bus="virtio" dev="vdx" />
</disk>
3. Run command : virsh attach-device vm001 virtio-blk.xml
error: Failed to attach device from blk-scsi.xml
error: internal error: unable to execute QEMU command 'device_add': Please set scsi=off for virtio-blk devices in order to use virtio 1.0
it means hotplug virtio-blk device failed.
4. Suspend or shutdown VM will leads to qemu crash

Problem happens in virtio_vmstate_change which is called by
vm_state_notify:
vdev’s parent_bus is NULL, so qdev_get_parent_bus(DEVICE(vdev)) will crash.
virtio_vmstate_change is added to the list vm_change_state_head at virtio_blk_device_realize(virtio_init),
but after hotplug virtio-blk failed, virtio_vmstate_change will not be removed from vm_change_state_head.
Adding unrealize function of virtio-blk device can solve this problem.

Signed-off-by: linzhecheng <linzhecheng@huawei.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
hw/virtio/virtio.c