dataplane: fix virtio_blk_data_plane_create() op blocker error path
commit745a9bb9cdcdcc50219d14704477c5cd678c4a7e
authorStefan Hajnoczi <stefanha@redhat.com>
Thu, 11 Sep 2014 12:49:39 +0000 (11 13:49 +0100)
committerKevin Wolf <kwolf@redhat.com>
Thu, 11 Sep 2014 14:21:46 +0000 (11 16:21 +0200)
tree583f94f6daa7e7ecc107f5b4b63b4fd14702e4b3
parent466c80f21fa2a69eb5cff8048b824c4812179b8c
dataplane: fix virtio_blk_data_plane_create() op blocker error path

Commit 3718d8ab65f68de2acccbe6a315907805f54e3cc ("block: Replace in_use
with operation blocker") broke the error path because it consumed
local_err instead of propagating it.

The caller has no way to know that the function failed.  This caused
virtio-blk to start "successfully" even though there was a fatal
dataplane error.

Steps to reproduce:

  $ qemu-system-x86_64 -enable-kvm -object iothread,id=iothread0 \
                       -drive if=none,id=drive0,file=a.img \
  (qemu) drive_mirror drive0 /tmp/foo.img
  (qemu) device_add virtio-blk-pci,iothread=iothread0,drive=drive0

Expected result:

  Since the mirror block job is using drive0 it is not possible to start
  virtio-blk data-plane.

  device_add fails and the PCI adapter is not added.

Actual result:

  device_add completes and the PCI adapter is added.

Cc: Fam Zheng <famz@redhat.com>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
hw/block/dataplane/virtio-blk.c