mirror: Go through ready -> complete process for 0 len image

When mirroring or active committing a zero length image, BLOCK_JOB_READY
is not reported now, instead the job completes because we short circuit
the mirror job loop.

This is inconsistent with non-zero length images, and only confuses
management software.

Let's do the same thing when seeing a 0-length image: report ready
immediately; wait for block-job-cancel or block-job-complete; clear the
cancel flag as existing non-zero image synced case (cancelled after
ready); then jump to the exit.

Reported-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Fam Zheng <famz@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
(cherry picked from commit 9e48b025400b2d284e17860862b0a4aa02c6032d)

*removed dependency on bcada37b

Signed-off-by: Michael Roth <mdroth@linux.vnet.ibm.com>

diff --git a/block/mirror.c b/block/mirror.c
index 9a0a55940c741fb487894fdb39df0f8c49088452..21f2d8dcfdbbb93b877970933801c75e79155ac0 100644 (file)
--- a/block/mirror.c
+++ b/block/mirror.c
@@ -324,9 +324,18 @@ static void coroutine_fn mirror_run(void *opaque)
     }
 
     s->common.len = bdrv_getlength(bs);
-    if (s->common.len <= 0) {
+    if (s->common.len < 0) {
         ret = s->common.len;
         goto immediate_exit;
+    } else if (s->common.len == 0) {
+        /* Report BLOCK_JOB_READY and wait for complete. */
+        block_job_ready(&s->common);
+        s->synced = true;
+        while (!block_job_is_cancelled(&s->common) && !s->should_complete) {
+            block_job_yield(&s->common);
+        }
+        s->common.cancelled = false;
+        goto immediate_exit;
     }
 
     length = (bdrv_getlength(bs) + s->granularity - 1) / s->granularity;