From: "Dr. David Alan Gilbert" <dgilbert@redhat.com>

  This is a bunch of RDMA fixes, the first is a race
I spotted a while ago that you don't hit during normal
operation; the rest are to do with migration failure and
cancellation that I started looking at because of lp1545052 which
is a failure to recover on the source if the destination
fails.
 
  I'm pretty sure there are other cases where the source
might hang waiting for a failed destination; particularly
if the destination hangs rather than fails completely;
one I know of is waiting for the event after the rdma_disconnect
but I don't have a good fix for it.  Suggestions welcome.
 
v4
  Fix more error paths in qemu_rdma_block_for_wrid - often
  the 'ret' wasn't set. Fix that and also fix the case we
  add to also set the ret. Spotted by Peter.

Dr. David Alan Gilbert (6):
  migration/rdma: Fix race on source
  migration: Close file on failed migration load
  migration/rdma: fix qemu_rdma_block_for_wrid error paths
  migration/rdma: Allow cancelling while waiting for wrid
  migration/rdma: Safely convert control types
  migration/rdma: Send error during cancelling

 migration/migration.c |   1 +
 migration/rdma.c      | 137 ++++++++++++++++++++++++++++++++++++--------------
 2 files changed, 101 insertions(+), 37 deletions(-)

-- 
2.13.0