scsi-disk: release AioContext in unaligned WRITE SAME case

scsi_write_same_complete() can retry the write if the request was
unaligned.  Make sure to release the AioContext when that code path is
taken!

This patch fixes a hang when QEMU terminates after an unaligned WRITE
SAME request has been processed with dataplane.  The hang occurs because
iothread_stop_all() cannot acquire the AioContext lock that was leaked
by the IOThread in scsi_write_same_complete().

Fixes: b9e413dd37 ("block: explicitly acquire aiocontext in aio callbacks that need it").
Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: qemu-stable@nongnu.org
Reported-by: Cong Li <coli@redhat.com>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Message-Id: <20180104142502.15175-1-stefanha@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
(cherry picked from commit 24355b79bdaf6ab12f7c610b032fc35ec045cd55)
Signed-off-by: Michael Roth <mdroth@linux.vnet.ibm.com>

diff --git a/hw/scsi/scsi-disk.c b/hw/scsi/scsi-disk.c
index 12431177a722445818c0ad7c160b48eb262a665f..5e4d2c38f0975671be71dad018772e212fe2fdbb 100644 (file)
--- a/hw/scsi/scsi-disk.c
+++ b/hw/scsi/scsi-disk.c
@@ -1755,6 +1755,7 @@ static void scsi_write_same_complete(void *opaque, int ret)
                                        data->sector << BDRV_SECTOR_BITS,
                                        &data->qiov, 0,
                                        scsi_write_same_complete, data);
+        aio_context_release(blk_get_aio_context(s->qdev.conf.blk));
         return;
     }