If the main loop cancels all block jobs while the block layer is not
drained, this cancelling may not happen instantaneously.  We can start a
drained section before vm_shutdown(), which entails another
bdrv_drain_all(); this nested bdrv_drain_all() will thus be a no-op,
basically.

We do not have to end the drained section, because we actually do not
want any requests to happen from this point on.

Signed-off-by: Max Reitz <mreitz@redhat.com>
---
I don't know whether it actually makes sense to never end this drained
section.  It makes sense to me.  Please correct me if I'm wrong.
---
 vl.c | 11 +++++++++++
 1 file changed, 11 insertions(+)

diff --git a/vl.c b/vl.c
index cd1fbc4cdc..3f8b3f74f5 100644
--- a/vl.c
+++ b/vl.c
@@ -4538,6 +4538,17 @@ int main(int argc, char **argv, char **envp)
      */
     migration_shutdown();
 
+    /*
+     * We must cancel all block jobs while the block layer is drained,
+     * or cancelling will be affected by throttling and thus may block
+     * for an extended period of time.
+     * vm_shutdown() will bdrv_drain_all(), so we may as well include
+     * it in the drained section.
+     * We do not need to end this section, because we do not want any
+     * requests happening from here on anyway.
+     */
+    bdrv_drain_all_begin();
+
     /* No more vcpu or device emulation activity beyond this point */
     vm_shutdown();
 
-- 
2.21.0