diff --git a/block/blk-core.c b/block/blk-core.c
index b850bed..7c59b0f 100644
--- a/block/blk-core.c
+++ b/block/blk-core.c
@@ -1700,6 +1700,7 @@
 int blk_insert_cloned_request(struct request_queue *q, struct request *rq)
 {
 	unsigned long flags;
+	int where = ELEVATOR_INSERT_BACK;
 
 	if (blk_rq_check_limits(q, rq))
 		return -EIO;
@@ -1716,7 +1717,10 @@
 	 */
 	BUG_ON(blk_queued_rq(rq));
 
-	add_acct_request(q, rq, ELEVATOR_INSERT_BACK);
+	if (rq->cmd_flags & (REQ_FLUSH|REQ_FUA))
+		where = ELEVATOR_INSERT_FLUSH;
+
+	add_acct_request(q, rq, where);
 	spin_unlock_irqrestore(q->queue_lock, flags);
 
 	return 0;
@@ -2273,7 +2277,7 @@
  *     %false - we are done with this request
  *     %true  - still buffers pending for this request
  **/
-static bool __blk_end_bidi_request(struct request *rq, int error,
+bool __blk_end_bidi_request(struct request *rq, int error,
 				   unsigned int nr_bytes, unsigned int bidi_bytes)
 {
 	if (blk_update_bidi_request(rq, error, nr_bytes, bidi_bytes))
diff --git a/block/blk-flush.c b/block/blk-flush.c
index 2d162bd..491eb30 100644
--- a/block/blk-flush.c
+++ b/block/blk-flush.c
@@ -123,7 +123,7 @@
 
 	/* make @rq a normal request */
 	rq->cmd_flags &= ~REQ_FLUSH_SEQ;
-	rq->end_io = NULL;
+	rq->end_io = rq->flush.saved_end_io;
 }
 
 /**
@@ -301,9 +301,6 @@
 	unsigned int fflags = q->flush_flags;	/* may change, cache */
 	unsigned int policy = blk_flush_policy(fflags, rq);
 
-	BUG_ON(rq->end_io);
-	BUG_ON(!rq->bio || rq->bio != rq->biotail);
-
 	/*
 	 * @policy now records what operations need to be done.  Adjust
 	 * REQ_FLUSH and FUA for the driver.
@@ -313,6 +310,19 @@
 		rq->cmd_flags &= ~REQ_FUA;
 
 	/*
+	 * An empty flush handed down from a stacking driver may
+	 * translate into nothing if the underlying device does not
+	 * advertise a write-back cache.  In this case, simply
+	 * complete the request.
+	 */
+	if (!policy) {
+		__blk_end_bidi_request(rq, 0, 0, 0);
+		return;
+	}
+
+	BUG_ON(!rq->bio || rq->bio != rq->biotail);
+
+	/*
 	 * If there's data but flush is not necessary, the request can be
 	 * processed directly without going through flush machinery.  Queue
 	 * for normal execution.
@@ -320,6 +330,7 @@
 	if ((policy & REQ_FSEQ_DATA) &&
 	    !(policy & (REQ_FSEQ_PREFLUSH | REQ_FSEQ_POSTFLUSH))) {
 		list_add_tail(&rq->queuelist, &q->queue_head);
+		blk_run_queue_async(q);
 		return;
 	}
 
@@ -330,6 +341,7 @@
 	memset(&rq->flush, 0, sizeof(rq->flush));
 	INIT_LIST_HEAD(&rq->flush.list);
 	rq->cmd_flags |= REQ_FLUSH_SEQ;
+	rq->flush.saved_end_io = rq->end_io; /* Usually NULL */
 	rq->end_io = flush_data_end_io;
 
 	blk_flush_complete_seq(rq, REQ_FSEQ_ACTIONS & ~policy, 0);
diff --git a/block/blk.h b/block/blk.h
index d658628..20b900a 100644
--- a/block/blk.h
+++ b/block/blk.h
@@ -17,6 +17,8 @@
 		      struct bio *bio);
 void blk_dequeue_request(struct request *rq);
 void __blk_queue_free_tags(struct request_queue *q);
+bool __blk_end_bidi_request(struct request *rq, int error,
+			    unsigned int nr_bytes, unsigned int bidi_bytes);
 
 void blk_rq_timed_out_timer(unsigned long data);
 void blk_delete_timer(struct request *);
diff --git a/include/linux/blkdev.h b/include/linux/blkdev.h
index 8479285..84b15d5 100644
--- a/include/linux/blkdev.h
+++ b/include/linux/blkdev.h
@@ -118,6 +118,7 @@
 		struct {
 			unsigned int		seq;
 			struct list_head	list;
+			rq_end_io_fn		*saved_end_io;
 		} flush;
 	};