diff --git a/block/blk-core.c b/block/blk-core.c
index e447799..d2f23ec 100644
--- a/block/blk-core.c
+++ b/block/blk-core.c
@@ -198,7 +198,8 @@
 	if (blk_queue_stopped(q))
 		return;
 
-	if (!test_and_set_bit(QUEUE_FLAG_PLUGGED, &q->queue_flags)) {
+	if (!test_bit(QUEUE_FLAG_PLUGGED, &q->queue_flags)) {
+		__set_bit(QUEUE_FLAG_PLUGGED, &q->queue_flags);
 		mod_timer(&q->unplug_timer, jiffies + q->unplug_delay);
 		blk_add_trace_generic(q, NULL, 0, BLK_TA_PLUG);
 	}
@@ -213,9 +214,10 @@
 {
 	WARN_ON(!irqs_disabled());
 
-	if (!test_and_clear_bit(QUEUE_FLAG_PLUGGED, &q->queue_flags))
+	if (!test_bit(QUEUE_FLAG_PLUGGED, &q->queue_flags))
 		return 0;
 
+	queue_flag_clear(QUEUE_FLAG_PLUGGED, q);
 	del_timer(&q->unplug_timer);
 	return 1;
 }
@@ -311,15 +313,16 @@
 {
 	WARN_ON(!irqs_disabled());
 
-	clear_bit(QUEUE_FLAG_STOPPED, &q->queue_flags);
+	queue_flag_clear(QUEUE_FLAG_STOPPED, q);
 
 	/*
 	 * one level of recursion is ok and is much faster than kicking
 	 * the unplug handling
 	 */
-	if (!test_and_set_bit(QUEUE_FLAG_REENTER, &q->queue_flags)) {
+	if (!test_bit(QUEUE_FLAG_REENTER, &q->queue_flags)) {
+		queue_flag_set(QUEUE_FLAG_REENTER, q);
 		q->request_fn(q);
-		clear_bit(QUEUE_FLAG_REENTER, &q->queue_flags);
+		queue_flag_clear(QUEUE_FLAG_REENTER, q);
 	} else {
 		blk_plug_device(q);
 		kblockd_schedule_work(&q->unplug_work);
@@ -344,7 +347,7 @@
 void blk_stop_queue(struct request_queue *q)
 {
 	blk_remove_plug(q);
-	set_bit(QUEUE_FLAG_STOPPED, &q->queue_flags);
+	queue_flag_set(QUEUE_FLAG_STOPPED, q);
 }
 EXPORT_SYMBOL(blk_stop_queue);
 
@@ -373,11 +376,8 @@
  * blk_run_queue - run a single device queue
  * @q:	The queue to run
  */
-void blk_run_queue(struct request_queue *q)
+void __blk_run_queue(struct request_queue *q)
 {
-	unsigned long flags;
-
-	spin_lock_irqsave(q->queue_lock, flags);
 	blk_remove_plug(q);
 
 	/*
@@ -385,15 +385,28 @@
 	 * handling reinvoke the handler shortly if we already got there.
 	 */
 	if (!elv_queue_empty(q)) {
-		if (!test_and_set_bit(QUEUE_FLAG_REENTER, &q->queue_flags)) {
+		if (!test_bit(QUEUE_FLAG_REENTER, &q->queue_flags)) {
+			queue_flag_set(QUEUE_FLAG_REENTER, q);
 			q->request_fn(q);
-			clear_bit(QUEUE_FLAG_REENTER, &q->queue_flags);
+			queue_flag_clear(QUEUE_FLAG_REENTER, q);
 		} else {
 			blk_plug_device(q);
 			kblockd_schedule_work(&q->unplug_work);
 		}
 	}
+}
+EXPORT_SYMBOL(__blk_run_queue);
 
+/**
+ * blk_run_queue - run a single device queue
+ * @q: The queue to run
+ */
+void blk_run_queue(struct request_queue *q)
+{
+	unsigned long flags;
+
+	spin_lock_irqsave(q->queue_lock, flags);
+	__blk_run_queue(q);
 	spin_unlock_irqrestore(q->queue_lock, flags);
 }
 EXPORT_SYMBOL(blk_run_queue);
@@ -406,7 +419,7 @@
 void blk_cleanup_queue(struct request_queue *q)
 {
 	mutex_lock(&q->sysfs_lock);
-	set_bit(QUEUE_FLAG_DEAD, &q->queue_flags);
+	queue_flag_set_unlocked(QUEUE_FLAG_DEAD, q);
 	mutex_unlock(&q->sysfs_lock);
 
 	if (q->elevator)
diff --git a/block/blk-merge.c b/block/blk-merge.c
index b5c5c4a..73b2356 100644
--- a/block/blk-merge.c
+++ b/block/blk-merge.c
@@ -55,7 +55,7 @@
 	if (!rq->bio)
 		return;
 
-	cluster = q->queue_flags & (1 << QUEUE_FLAG_CLUSTER);
+	cluster = test_bit(QUEUE_FLAG_CLUSTER, &q->queue_flags);
 	hw_seg_size = seg_size = 0;
 	phys_size = hw_size = nr_phys_segs = nr_hw_segs = 0;
 	rq_for_each_segment(bv, rq, iter) {
@@ -128,7 +128,7 @@
 static int blk_phys_contig_segment(struct request_queue *q, struct bio *bio,
 				   struct bio *nxt)
 {
-	if (!(q->queue_flags & (1 << QUEUE_FLAG_CLUSTER)))
+	if (!test_bit(QUEUE_FLAG_CLUSTER, &q->queue_flags))
 		return 0;
 
 	if (!BIOVEC_PHYS_MERGEABLE(__BVEC_END(bio), __BVEC_START(nxt)))
@@ -175,7 +175,7 @@
 	int nsegs, cluster;
 
 	nsegs = 0;
-	cluster = q->queue_flags & (1 << QUEUE_FLAG_CLUSTER);
+	cluster = test_bit(QUEUE_FLAG_CLUSTER, &q->queue_flags);
 
 	/*
 	 * for each bio in rq
diff --git a/block/blk-settings.c b/block/blk-settings.c
index 77b51dc..6089384 100644
--- a/block/blk-settings.c
+++ b/block/blk-settings.c
@@ -287,7 +287,7 @@
 	t->max_segment_size = min(t->max_segment_size, b->max_segment_size);
 	t->hardsect_size = max(t->hardsect_size, b->hardsect_size);
 	if (!test_bit(QUEUE_FLAG_CLUSTER, &b->queue_flags))
-		clear_bit(QUEUE_FLAG_CLUSTER, &t->queue_flags);
+		queue_flag_clear(QUEUE_FLAG_CLUSTER, t);
 }
 EXPORT_SYMBOL(blk_queue_stack_limits);
 
diff --git a/block/blk-tag.c b/block/blk-tag.c
index 4780a46..e176ddb 100644
--- a/block/blk-tag.c
+++ b/block/blk-tag.c
@@ -70,7 +70,7 @@
 	__blk_free_tags(bqt);
 
 	q->queue_tags = NULL;
-	q->queue_flags &= ~(1 << QUEUE_FLAG_QUEUED);
+	queue_flag_clear(QUEUE_FLAG_QUEUED, q);
 }
 
 /**
@@ -98,7 +98,7 @@
  **/
 void blk_queue_free_tags(struct request_queue *q)
 {
-	clear_bit(QUEUE_FLAG_QUEUED, &q->queue_flags);
+	queue_flag_clear(QUEUE_FLAG_QUEUED, q);
 }
 EXPORT_SYMBOL(blk_queue_free_tags);
 
@@ -188,7 +188,7 @@
 		rc = blk_queue_resize_tags(q, depth);
 		if (rc)
 			return rc;
-		set_bit(QUEUE_FLAG_QUEUED, &q->queue_flags);
+		queue_flag_set(QUEUE_FLAG_QUEUED, q);
 		return 0;
 	} else
 		atomic_inc(&tags->refcnt);
@@ -197,7 +197,7 @@
 	 * assign it, all done
 	 */
 	q->queue_tags = tags;
-	q->queue_flags |= (1 << QUEUE_FLAG_QUEUED);
+	queue_flag_set(QUEUE_FLAG_QUEUED, q);
 	INIT_LIST_HEAD(&q->tag_busy_list);
 	return 0;
 fail:
diff --git a/block/elevator.c b/block/elevator.c
index 88318c3..e8a90fe 100644
--- a/block/elevator.c
+++ b/block/elevator.c
@@ -1070,7 +1070,7 @@
 	 */
 	spin_lock_irq(q->queue_lock);
 
-	set_bit(QUEUE_FLAG_ELVSWITCH, &q->queue_flags);
+	queue_flag_set(QUEUE_FLAG_ELVSWITCH, q);
 
 	elv_drain_elevator(q);
 
@@ -1104,7 +1104,10 @@
 	 * finally exit old elevator and turn off BYPASS.
 	 */
 	elevator_exit(old_elevator);
-	clear_bit(QUEUE_FLAG_ELVSWITCH, &q->queue_flags);
+	spin_lock_irq(q->queue_lock);
+	queue_flag_clear(QUEUE_FLAG_ELVSWITCH, q);
+	spin_unlock_irq(q->queue_lock);
+
 	return 1;
 
 fail_register:
@@ -1115,7 +1118,11 @@
 	elevator_exit(e);
 	q->elevator = old_elevator;
 	elv_register_queue(q);
-	clear_bit(QUEUE_FLAG_ELVSWITCH, &q->queue_flags);
+
+	spin_lock_irq(q->queue_lock);
+	queue_flag_clear(QUEUE_FLAG_ELVSWITCH, q);
+	spin_unlock_irq(q->queue_lock);
+
 	return 0;
 }
 
diff --git a/drivers/block/loop.c b/drivers/block/loop.c
index f7f1635..d3a25b0 100644
--- a/drivers/block/loop.c
+++ b/drivers/block/loop.c
@@ -546,7 +546,7 @@
 {
 	struct loop_device *lo = q->queuedata;
 
-	clear_bit(QUEUE_FLAG_PLUGGED, &q->queue_flags);
+	queue_flag_clear_unlocked(QUEUE_FLAG_PLUGGED, q);
 	blk_run_address_space(lo->lo_backing_file->f_mapping);
 }
 
diff --git a/drivers/block/ub.c b/drivers/block/ub.c
index 27bfe72..e322cce 100644
--- a/drivers/block/ub.c
+++ b/drivers/block/ub.c
@@ -2399,7 +2399,7 @@
 		del_gendisk(lun->disk);
 		/*
 		 * I wish I could do:
-		 *    set_bit(QUEUE_FLAG_DEAD, &q->queue_flags);
+		 *    queue_flag_set(QUEUE_FLAG_DEAD, q);
 		 * As it is, we rely on our internal poisoning and let
 		 * the upper levels to spin furiously failing all the I/O.
 		 */
diff --git a/drivers/md/dm-table.c b/drivers/md/dm-table.c
index 51be533..73326e7 100644
--- a/drivers/md/dm-table.c
+++ b/drivers/md/dm-table.c
@@ -873,10 +873,13 @@
 	q->max_hw_sectors = t->limits.max_hw_sectors;
 	q->seg_boundary_mask = t->limits.seg_boundary_mask;
 	q->bounce_pfn = t->limits.bounce_pfn;
+	/* XXX: the below will probably go bug. must ensure there can be no
+	 * concurrency on queue_flags, and use the unlocked versions...
+	 */
 	if (t->limits.no_cluster)
-		q->queue_flags &= ~(1 << QUEUE_FLAG_CLUSTER);
+		queue_flag_clear(QUEUE_FLAG_CLUSTER, q);
 	else
-		q->queue_flags |= (1 << QUEUE_FLAG_CLUSTER);
+		queue_flag_set(QUEUE_FLAG_CLUSTER, q);
 
 }
 
diff --git a/drivers/md/md.c b/drivers/md/md.c
index 87620b7..acd716b 100644
--- a/drivers/md/md.c
+++ b/drivers/md/md.c
@@ -282,7 +282,8 @@
 		kfree(new);
 		return NULL;
 	}
-	set_bit(QUEUE_FLAG_CLUSTER, &new->queue->queue_flags);
+	/* Can be unlocked because the queue is new: no concurrency */
+	queue_flag_set_unlocked(QUEUE_FLAG_CLUSTER, new->queue);
 
 	blk_queue_make_request(new->queue, md_fail_request);
 
diff --git a/drivers/scsi/scsi_debug.c b/drivers/scsi/scsi_debug.c
index 07103c3..f6600bf 100644
--- a/drivers/scsi/scsi_debug.c
+++ b/drivers/scsi/scsi_debug.c
@@ -1773,7 +1773,7 @@
 	if (SCSI_DEBUG_OPT_NOISE & scsi_debug_opts)
 		printk(KERN_INFO "scsi_debug: slave_alloc <%u %u %u %u>\n",
 		       sdp->host->host_no, sdp->channel, sdp->id, sdp->lun);
-	set_bit(QUEUE_FLAG_BIDI, &sdp->request_queue->queue_flags);
+	queue_flag_set_unlocked(QUEUE_FLAG_BIDI, sdp->request_queue);
 	return 0;
 }
 
diff --git a/drivers/scsi/scsi_lib.c b/drivers/scsi/scsi_lib.c
index 67f412b..d545ad1 100644
--- a/drivers/scsi/scsi_lib.c
+++ b/drivers/scsi/scsi_lib.c
@@ -536,6 +536,9 @@
 	       !shost->host_blocked && !shost->host_self_blocked &&
 		!((shost->can_queue > 0) &&
 		  (shost->host_busy >= shost->can_queue))) {
+
+		int flagset;
+
 		/*
 		 * As long as shost is accepting commands and we have
 		 * starved queues, call blk_run_queue. scsi_request_fn
@@ -549,19 +552,20 @@
 		sdev = list_entry(shost->starved_list.next,
 					  struct scsi_device, starved_entry);
 		list_del_init(&sdev->starved_entry);
-		spin_unlock_irqrestore(shost->host_lock, flags);
+		spin_unlock(shost->host_lock);
 
+		spin_lock(sdev->request_queue->queue_lock);
+		flagset = test_bit(QUEUE_FLAG_REENTER, &q->queue_flags) &&
+				!test_bit(QUEUE_FLAG_REENTER,
+					&sdev->request_queue->queue_flags);
+		if (flagset)
+			queue_flag_set(QUEUE_FLAG_REENTER, sdev->request_queue);
+		__blk_run_queue(sdev->request_queue);
+		if (flagset)
+			queue_flag_clear(QUEUE_FLAG_REENTER, sdev->request_queue);
+		spin_unlock(sdev->request_queue->queue_lock);
 
-		if (test_bit(QUEUE_FLAG_REENTER, &q->queue_flags) &&
-		    !test_and_set_bit(QUEUE_FLAG_REENTER,
-				      &sdev->request_queue->queue_flags)) {
-			blk_run_queue(sdev->request_queue);
-			clear_bit(QUEUE_FLAG_REENTER,
-				  &sdev->request_queue->queue_flags);
-		} else
-			blk_run_queue(sdev->request_queue);
-
-		spin_lock_irqsave(shost->host_lock, flags);
+		spin_lock(shost->host_lock);
 		if (unlikely(!list_empty(&sdev->starved_entry)))
 			/*
 			 * sdev lost a race, and was put back on the
@@ -1585,8 +1589,9 @@
 
 	blk_queue_max_segment_size(q, dma_get_max_seg_size(dev));
 
+	/* New queue, no concurrency on queue_flags */
 	if (!shost->use_clustering)
-		clear_bit(QUEUE_FLAG_CLUSTER, &q->queue_flags);
+		queue_flag_clear_unlocked(QUEUE_FLAG_CLUSTER, q);
 
 	/*
 	 * set a reasonable default alignment on word boundaries: the
diff --git a/drivers/scsi/scsi_transport_sas.c b/drivers/scsi/scsi_transport_sas.c
index 7899e3d..f4461d3 100644
--- a/drivers/scsi/scsi_transport_sas.c
+++ b/drivers/scsi/scsi_transport_sas.c
@@ -248,8 +248,7 @@
 	else
 		q->queuedata = shost;
 
-	set_bit(QUEUE_FLAG_BIDI, &q->queue_flags);
-
+	queue_flag_set_unlocked(QUEUE_FLAG_BIDI, q);
 	return 0;
 }
 
diff --git a/include/linux/blkdev.h b/include/linux/blkdev.h
index c5065e3..8ca481c 100644
--- a/include/linux/blkdev.h
+++ b/include/linux/blkdev.h
@@ -408,6 +408,30 @@
 #define QUEUE_FLAG_ELVSWITCH	8	/* don't use elevator, just do FIFO */
 #define QUEUE_FLAG_BIDI		9	/* queue supports bidi requests */
 
+static inline void queue_flag_set_unlocked(unsigned int flag,
+					   struct request_queue *q)
+{
+	__set_bit(flag, &q->queue_flags);
+}
+
+static inline void queue_flag_set(unsigned int flag, struct request_queue *q)
+{
+	WARN_ON_ONCE(!spin_is_locked(q->queue_lock));
+	__set_bit(flag, &q->queue_flags);
+}
+
+static inline void queue_flag_clear_unlocked(unsigned int flag,
+					     struct request_queue *q)
+{
+	__clear_bit(flag, &q->queue_flags);
+}
+
+static inline void queue_flag_clear(unsigned int flag, struct request_queue *q)
+{
+	WARN_ON_ONCE(!spin_is_locked(q->queue_lock));
+	__clear_bit(flag, &q->queue_flags);
+}
+
 enum {
 	/*
 	 * Hardbarrier is supported with one of the following methods.
@@ -496,17 +520,17 @@
 static inline void blk_set_queue_full(struct request_queue *q, int rw)
 {
 	if (rw == READ)
-		set_bit(QUEUE_FLAG_READFULL, &q->queue_flags);
+		queue_flag_set(QUEUE_FLAG_READFULL, q);
 	else
-		set_bit(QUEUE_FLAG_WRITEFULL, &q->queue_flags);
+		queue_flag_set(QUEUE_FLAG_WRITEFULL, q);
 }
 
 static inline void blk_clear_queue_full(struct request_queue *q, int rw)
 {
 	if (rw == READ)
-		clear_bit(QUEUE_FLAG_READFULL, &q->queue_flags);
+		queue_flag_clear(QUEUE_FLAG_READFULL, q);
 	else
-		clear_bit(QUEUE_FLAG_WRITEFULL, &q->queue_flags);
+		queue_flag_clear(QUEUE_FLAG_WRITEFULL, q);
 }
 
 
@@ -626,6 +650,7 @@
 extern void blk_stop_queue(struct request_queue *q);
 extern void blk_sync_queue(struct request_queue *q);
 extern void __blk_stop_queue(struct request_queue *q);
+extern void __blk_run_queue(struct request_queue *);
 extern void blk_run_queue(struct request_queue *);
 extern void blk_start_queueing(struct request_queue *);
 extern int blk_rq_map_user(struct request_queue *, struct request *, void __user *, unsigned long);