3 files changed, 32 insertions, 6 deletions

diff --git a/block/blk-flush.c b/block/blk-flush.c
index 6c9b5e189e62..bb21e4c36f70 100644
--- a/block/blk-flush.c
+++ b/block/blk-flush.c
@@ -212,13 +212,19 @@ static void flush_end_io(struct request *flush_rq, int error)
         }
 
         /*
-         * Moving a request silently to empty queue_head may stall the
-         * queue.  Kick the queue in those cases.  This function is called
-         * from request completion path and calling directly into
-         * request_fn may confuse the driver.  Always use kblockd.
+         * Kick the queue to avoid stall for two cases:
+         * 1. Moving a request silently to empty queue_head may stall the
+         * queue.
+         * 2. When flush request is running in non-queueable queue, the
+         * queue is hold. Restart the queue after flush request is finished
+         * to avoid stall.
+         * This function is called from request completion path and calling
+         * directly into request_fn may confuse the driver.  Always use
+         * kblockd.
          */
-        if (queued)
+        if (queued || q->flush_queue_delayed)
                 blk_run_queue_async(q);
+        q->flush_queue_delayed = 0;
 }
 
 /**
diff --git a/block/blk.h b/block/blk.h
index c9df8fc3c999..83e4bff36201 100644
--- a/block/blk.h
+++ b/block/blk.h
@@ -62,7 +62,26 @@ static inline struct request *__elv_next_request(struct request_queue *q)
                         rq = list_entry_rq(q->queue_head.next);
                         return rq;
                 }
-
+                /*
+                 * Flush request is running and flush request isn't queueable
+                 * in the drive, we can hold the queue till flush request is
+                 * finished. Even we don't do this, driver can't dispatch next
+                 * requests and will requeue them. And this can improve
+                 * throughput too. For example, we have request flush1, write1,
+                 * flush 2. flush1 is dispatched, then queue is hold, write1
+                 * isn't inserted to queue. After flush1 is finished, flush2
+                 * will be dispatched. Since disk cache is already clean,
+                 * flush2 will be finished very soon, so looks like flush2 is
+                 * folded to flush1.
+                 * Since the queue is hold, a flag is set to indicate the queue
+                 * should be restarted later. Please see flush_end_io() for
+                 * details.
+                 */
+                if (q->flush_pending_idx != q->flush_running_idx &&
+                                !queue_flush_queueable(q)) {
+                        q->flush_queue_delayed = 1;
+                        return NULL;
+                }
                 if (!q->elevator->ops->elevator_dispatch_fn(q, 0))
                         return NULL;
         }
diff --git a/include/linux/blkdev.h b/include/linux/blkdev.h
index 8bd2a271b2d8..9f921bf4bf8c 100644
--- a/include/linux/blkdev.h
+++ b/include/linux/blkdev.h
@@ -365,6 +365,7 @@ struct request_queue
          */
         unsigned int            flush_flags;
         unsigned int            flush_not_queueable:1;
+        unsigned int            flush_queue_delayed:1;
         unsigned int            flush_pending_idx:1;
         unsigned int            flush_running_idx:1;
         unsigned long           flush_pending_since;