Merge tag 'v4.12-rc5' into for-4.13/block

We've already got a few conflicts and upcoming work depends on some of the
changes that have gone into mainline as regression fixes for this series.

Pull in 4.12-rc5 to resolve these conflicts and make it easier on down stream
trees to continue working on 4.13 changes.

Signed-off-by: Jens Axboe <axboe@kernel.dk>

11 files changed, 179 insertions, 57 deletions

diff --git a/block/bfq-cgroup.c b/block/bfq-cgroup.c
index c8a32fb345cf..78b2e0db4fb2 100644
--- a/block/bfq-cgroup.c
+++ b/block/bfq-cgroup.c
@@ -52,7 +52,7 @@ BFQG_FLAG_FNS(idling)
 BFQG_FLAG_FNS(empty)
 #undef BFQG_FLAG_FNS
 
-/* This should be called with the queue_lock held. */
+/* This should be called with the scheduler lock held. */
 static void bfqg_stats_update_group_wait_time(struct bfqg_stats *stats)
 {
 	unsigned long long now;
@@ -67,7 +67,7 @@ static void bfqg_stats_update_group_wait_time(struct bfqg_stats *stats)
 	bfqg_stats_clear_waiting(stats);
 }
 
-/* This should be called with the queue_lock held. */
+/* This should be called with the scheduler lock held. */
 static void bfqg_stats_set_start_group_wait_time(struct bfq_group *bfqg,
 						 struct bfq_group *curr_bfqg)
 {
@@ -81,7 +81,7 @@ static void bfqg_stats_set_start_group_wait_time(struct bfq_group *bfqg,
 	bfqg_stats_mark_waiting(stats);
 }
 
-/* This should be called with the queue_lock held. */
+/* This should be called with the scheduler lock held. */
 static void bfqg_stats_end_empty_time(struct bfqg_stats *stats)
 {
 	unsigned long long now;
@@ -203,12 +203,30 @@ struct bfq_group *bfqq_group(struct bfq_queue *bfqq)
 
 static void bfqg_get(struct bfq_group *bfqg)
 {
-	return blkg_get(bfqg_to_blkg(bfqg));
+	bfqg->ref++;
 }
 
 void bfqg_put(struct bfq_group *bfqg)
 {
-	return blkg_put(bfqg_to_blkg(bfqg));
+	bfqg->ref--;
+
+	if (bfqg->ref == 0)
+		kfree(bfqg);
+}
+
+static void bfqg_and_blkg_get(struct bfq_group *bfqg)
+{
+	/* see comments in bfq_bic_update_cgroup for why refcounting bfqg */
+	bfqg_get(bfqg);
+
+	blkg_get(bfqg_to_blkg(bfqg));
+}
+
+void bfqg_and_blkg_put(struct bfq_group *bfqg)
+{
+	bfqg_put(bfqg);
+
+	blkg_put(bfqg_to_blkg(bfqg));
 }
 
 void bfqg_stats_update_io_add(struct bfq_group *bfqg, struct bfq_queue *bfqq,
@@ -312,7 +330,11 @@ void bfq_init_entity(struct bfq_entity *entity, struct bfq_group *bfqg)
 	if (bfqq) {
 		bfqq->ioprio = bfqq->new_ioprio;
 		bfqq->ioprio_class = bfqq->new_ioprio_class;
-		bfqg_get(bfqg);
+		/*
+		 * Make sure that bfqg and its associated blkg do not
+		 * disappear before entity.
+		 */
+		bfqg_and_blkg_get(bfqg);
 	}
 	entity->parent = bfqg->my_entity; /* NULL for root group */
 	entity->sched_data = &bfqg->sched_data;
@@ -399,6 +421,8 @@ struct blkg_policy_data *bfq_pd_alloc(gfp_t gfp, int node)
 		return NULL;
 	}
 
+	/* see comments in bfq_bic_update_cgroup for why refcounting */
+	bfqg_get(bfqg);
 	return &bfqg->pd;
 }
 
@@ -426,7 +450,7 @@ void bfq_pd_free(struct blkg_policy_data *pd)
 	struct bfq_group *bfqg = pd_to_bfqg(pd);
 
 	bfqg_stats_exit(&bfqg->stats);
-	return kfree(bfqg);
+	bfqg_put(bfqg);
 }
 
 void bfq_pd_reset_stats(struct blkg_policy_data *pd)
@@ -496,9 +520,10 @@ struct bfq_group *bfq_find_set_group(struct bfq_data *bfqd,
  * Move @bfqq to @bfqg, deactivating it from its old group and reactivating
  * it on the new one.  Avoid putting the entity on the old group idle tree.
  *
- * Must be called under the queue lock; the cgroup owning @bfqg must
- * not disappear (by now this just means that we are called under
- * rcu_read_lock()).
+ * Must be called under the scheduler lock, to make sure that the blkg
+ * owning @bfqg does not disappear (see comments in
+ * bfq_bic_update_cgroup on guaranteeing the consistency of blkg
+ * objects).
  */
 void bfq_bfqq_move(struct bfq_data *bfqd, struct bfq_queue *bfqq,
 		   struct bfq_group *bfqg)
@@ -519,16 +544,12 @@ void bfq_bfqq_move(struct bfq_data *bfqd, struct bfq_queue *bfqq,
 		bfq_deactivate_bfqq(bfqd, bfqq, false, false);
 	else if (entity->on_st)
 		bfq_put_idle_entity(bfq_entity_service_tree(entity), entity);
-	bfqg_put(bfqq_group(bfqq));
+	bfqg_and_blkg_put(bfqq_group(bfqq));
 
-	/*
-	 * Here we use a reference to bfqg.  We don't need a refcounter
-	 * as the cgroup reference will not be dropped, so that its
-	 * destroy() callback will not be invoked.
-	 */
 	entity->parent = bfqg->my_entity;
 	entity->sched_data = &bfqg->sched_data;
-	bfqg_get(bfqg);
+	/* pin down bfqg and its associated blkg  */
+	bfqg_and_blkg_get(bfqg);
 
 	if (bfq_bfqq_busy(bfqq)) {
 		bfq_pos_tree_add_move(bfqd, bfqq);
@@ -545,8 +566,9 @@ void bfq_bfqq_move(struct bfq_data *bfqd, struct bfq_queue *bfqq,
  * @bic: the bic to move.
  * @blkcg: the blk-cgroup to move to.
  *
- * Move bic to blkcg, assuming that bfqd->queue is locked; the caller
- * has to make sure that the reference to cgroup is valid across the call.
+ * Move bic to blkcg, assuming that bfqd->lock is held; which makes
+ * sure that the reference to cgroup is valid across the call (see
+ * comments in bfq_bic_update_cgroup on this issue)
  *
  * NOTE: an alternative approach might have been to store the current
  * cgroup in bfqq and getting a reference to it, reducing the lookup
@@ -604,6 +626,57 @@ void bfq_bic_update_cgroup(struct bfq_io_cq *bic, struct bio *bio)
 		goto out;
 
 	bfqg = __bfq_bic_change_cgroup(bfqd, bic, bio_blkcg(bio));
+	/*
+	 * Update blkg_path for bfq_log_* functions. We cache this
+	 * path, and update it here, for the following
+	 * reasons. Operations on blkg objects in blk-cgroup are
+	 * protected with the request_queue lock, and not with the
+	 * lock that protects the instances of this scheduler
+	 * (bfqd->lock). This exposes BFQ to the following sort of
+	 * race.
+	 *
+	 * The blkg_lookup performed in bfq_get_queue, protected
+	 * through rcu, may happen to return the address of a copy of
+	 * the original blkg. If this is the case, then the
+	 * bfqg_and_blkg_get performed in bfq_get_queue, to pin down
+	 * the blkg, is useless: it does not prevent blk-cgroup code
+	 * from destroying both the original blkg and all objects
+	 * directly or indirectly referred by the copy of the
+	 * blkg.
+	 *
+	 * On the bright side, destroy operations on a blkg invoke, as
+	 * a first step, hooks of the scheduler associated with the
+	 * blkg. And these hooks are executed with bfqd->lock held for
+	 * BFQ. As a consequence, for any blkg associated with the
+	 * request queue this instance of the scheduler is attached
+	 * to, we are guaranteed that such a blkg is not destroyed, and
+	 * that all the pointers it contains are consistent, while we
+	 * are holding bfqd->lock. A blkg_lookup performed with
+	 * bfqd->lock held then returns a fully consistent blkg, which
+	 * remains consistent until this lock is held.
+	 *
+	 * Thanks to the last fact, and to the fact that: (1) bfqg has
+	 * been obtained through a blkg_lookup in the above
+	 * assignment, and (2) bfqd->lock is being held, here we can
+	 * safely use the policy data for the involved blkg (i.e., the
+	 * field bfqg->pd) to get to the blkg associated with bfqg,
+	 * and then we can safely use any field of blkg. After we
+	 * release bfqd->lock, even just getting blkg through this
+	 * bfqg may cause dangling references to be traversed, as
+	 * bfqg->pd may not exist any more.
+	 *
+	 * In view of the above facts, here we cache, in the bfqg, any
+	 * blkg data we may need for this bic, and for its associated
+	 * bfq_queue. As of now, we need to cache only the path of the
+	 * blkg, which is used in the bfq_log_* functions.
+	 *
+	 * Finally, note that bfqg itself needs to be protected from
+	 * destruction on the blkg_free of the original blkg (which
+	 * invokes bfq_pd_free). We use an additional private
+	 * refcounter for bfqg, to let it disappear only after no
+	 * bfq_queue refers to it any longer.
+	 */
+	blkg_path(bfqg_to_blkg(bfqg), bfqg->blkg_path, sizeof(bfqg->blkg_path));
 	bic->blkcg_serial_nr = serial_nr;
 out:
 	rcu_read_unlock();
@@ -640,8 +713,6 @@ static void bfq_reparent_leaf_entity(struct bfq_data *bfqd,
  * @bfqd: the device data structure with the root group.
  * @bfqg: the group to move from.
  * @st: the service tree with the entities.
- *
- * Needs queue_lock to be taken and reference to be valid over the call.
  */
 static void bfq_reparent_active_entities(struct bfq_data *bfqd,
 					 struct bfq_group *bfqg,
@@ -692,8 +763,7 @@ void bfq_pd_offline(struct blkg_policy_data *pd)
 		/*
 		 * The idle tree may still contain bfq_queues belonging
 		 * to exited task because they never migrated to a different
-		 * cgroup from the one being destroyed now.  No one else
-		 * can access them so it's safe to act without any lock.
+		 * cgroup from the one being destroyed now.
 		 */
 		bfq_flush_idle_tree(st);
 
diff --git a/block/bfq-iosched.c b/block/bfq-iosched.c
index 08ce45096350..ed93da2462ab 100644
--- a/block/bfq-iosched.c
+++ b/block/bfq-iosched.c
@@ -3665,7 +3665,7 @@ void bfq_put_queue(struct bfq_queue *bfqq)
 
 	kmem_cache_free(bfq_pool, bfqq);
 #ifdef CONFIG_BFQ_GROUP_IOSCHED
-	bfqg_put(bfqg);
+	bfqg_and_blkg_put(bfqg);
 #endif
 }
 
diff --git a/block/bfq-iosched.h b/block/bfq-iosched.h
index ae783c06dfd9..5c3bf9861492 100644
--- a/block/bfq-iosched.h
+++ b/block/bfq-iosched.h
@@ -759,6 +759,12 @@ struct bfq_group {
 	/* must be the first member */
 	struct blkg_policy_data pd;
 
+	/* cached path for this blkg (see comments in bfq_bic_update_cgroup) */
+	char blkg_path[128];
+
+	/* reference counter (see comments in bfq_bic_update_cgroup) */
+	int ref;
+
 	struct bfq_entity entity;
 	struct bfq_sched_data sched_data;
 
@@ -838,7 +844,7 @@ struct bfq_group *bfq_find_set_group(struct bfq_data *bfqd,
 struct blkcg_gq *bfqg_to_blkg(struct bfq_group *bfqg);
 struct bfq_group *bfqq_group(struct bfq_queue *bfqq);
 struct bfq_group *bfq_create_group_hierarchy(struct bfq_data *bfqd, int node);
-void bfqg_put(struct bfq_group *bfqg);
+void bfqg_and_blkg_put(struct bfq_group *bfqg);
 
 #ifdef CONFIG_BFQ_GROUP_IOSCHED
 extern struct cftype bfq_blkcg_legacy_files[];
@@ -910,20 +916,13 @@ void bfq_add_bfqq_busy(struct bfq_data *bfqd, struct bfq_queue *bfqq);
 struct bfq_group *bfqq_group(struct bfq_queue *bfqq);
 
 #define bfq_log_bfqq(bfqd, bfqq, fmt, args...)	do {			\
-	char __pbuf[128];						\
-									\
-	blkg_path(bfqg_to_blkg(bfqq_group(bfqq)), __pbuf, sizeof(__pbuf)); \
-	blk_add_trace_msg((bfqd)->queue, "bfq%d%c %s " fmt, (bfqq)->pid, \
+	blk_add_trace_msg((bfqd)->queue, "bfq%d%c %s " fmt, (bfqq)->pid,\
 			bfq_bfqq_sync((bfqq)) ? 'S' : 'A',		\
-			  __pbuf, ##args);				\
+			bfqq_group(bfqq)->blkg_path, ##args);		\
 } while (0)
 
-#define bfq_log_bfqg(bfqd, bfqg, fmt, args...)	do {			\
-	char __pbuf[128];						\
-									\
-	blkg_path(bfqg_to_blkg(bfqg), __pbuf, sizeof(__pbuf));		\
-	blk_add_trace_msg((bfqd)->queue, "%s " fmt, __pbuf, ##args);	\
-} while (0)
+#define bfq_log_bfqg(bfqd, bfqg, fmt, args...)	\
+	blk_add_trace_msg((bfqd)->queue, "%s " fmt, (bfqg)->blkg_path, ##args)
 
 #else /* CONFIG_BFQ_GROUP_IOSCHED */
 
diff --git a/block/bio-integrity.c b/block/bio-integrity.c
index 17b9740e138b..b8a3a65f7364 100644
--- a/block/bio-integrity.c
+++ b/block/bio-integrity.c
@@ -175,6 +175,9 @@ bool bio_integrity_enabled(struct bio *bio)
 	if (bio_op(bio) != REQ_OP_READ && bio_op(bio) != REQ_OP_WRITE)
 		return false;
 
+	if (!bio_sectors(bio))
+		return false;
+
 	/* Already protected? */
 	if (bio_integrity(bio))
 		return false;
diff --git a/block/blk-cgroup.c b/block/blk-cgroup.c
index 7c2947128f58..0480892e97e5 100644
--- a/block/blk-cgroup.c
+++ b/block/blk-cgroup.c
@@ -74,7 +74,7 @@ static void blkg_free(struct blkcg_gq *blkg)
 			blkcg_policy[i]->pd_free_fn(blkg->pd[i]);
 
 	if (blkg->blkcg != &blkcg_root)
-		blk_exit_rl(&blkg->rl);
+		blk_exit_rl(blkg->q, &blkg->rl);
 
 	blkg_rwstat_exit(&blkg->stat_ios);
 	blkg_rwstat_exit(&blkg->stat_bytes);
diff --git a/block/blk-core.c b/block/blk-core.c
index 3d84820ace9e..8592409db272 100644
--- a/block/blk-core.c
+++ b/block/blk-core.c
@@ -706,13 +706,19 @@ int blk_init_rl(struct request_list *rl, struct request_queue *q,
 	if (!rl->rq_pool)
 		return -ENOMEM;
 
+	if (rl != &q->root_rl)
+		WARN_ON_ONCE(!blk_get_queue(q));
+
 	return 0;
 }
 
-void blk_exit_rl(struct request_list *rl)
+void blk_exit_rl(struct request_queue *q, struct request_list *rl)
 {
-	if (rl->rq_pool)
+	if (rl->rq_pool) {
 		mempool_destroy(rl->rq_pool);
+		if (rl != &q->root_rl)
+			blk_put_queue(q);
+	}
 }
 
 struct request_queue *blk_alloc_queue(gfp_t gfp_mask)
diff --git a/block/blk-mq.c b/block/blk-mq.c
index 7af78b1e9db9..da2f21961525 100644
--- a/block/blk-mq.c
+++ b/block/blk-mq.c
@@ -1396,22 +1396,28 @@ static blk_qc_t request_to_qc_t(struct blk_mq_hw_ctx *hctx, struct request *rq)
 	return blk_tag_to_qc_t(rq->internal_tag, hctx->queue_num, true);
 }
 
-static void __blk_mq_try_issue_directly(struct request *rq, blk_qc_t *cookie,
-				      bool may_sleep)
+static void __blk_mq_try_issue_directly(struct blk_mq_hw_ctx *hctx,
+					struct request *rq,
+					blk_qc_t *cookie, bool may_sleep)
 {
 	struct request_queue *q = rq->q;
 	struct blk_mq_queue_data bd = {
 		.rq = rq,
 		.last = true,
 	};
-	struct blk_mq_hw_ctx *hctx;
 	blk_qc_t new_cookie;
-	blk_status_t ret;
+	int ret;
+	bool run_queue = true;
+
+	if (blk_mq_hctx_stopped(hctx)) {
+		run_queue = false;
+		goto insert;
+	}
 
 	if (q->elevator)
 		goto insert;
 
-	if (!blk_mq_get_driver_tag(rq, &hctx, false))
+	if (!blk_mq_get_driver_tag(rq, NULL, false))
 		goto insert;
 
 	new_cookie = request_to_qc_t(hctx, rq);
@@ -1436,7 +1442,7 @@ static void __blk_mq_try_issue_directly(struct request *rq, blk_qc_t *cookie,
 	}
 
 insert:
-	blk_mq_sched_insert_request(rq, false, true, false, may_sleep);
+	blk_mq_sched_insert_request(rq, false, run_queue, false, may_sleep);
 }
 
 static void blk_mq_try_issue_directly(struct blk_mq_hw_ctx *hctx,
@@ -1444,7 +1450,7 @@ static void blk_mq_try_issue_directly(struct blk_mq_hw_ctx *hctx,
 {
 	if (!(hctx->flags & BLK_MQ_F_BLOCKING)) {
 		rcu_read_lock();
-		__blk_mq_try_issue_directly(rq, cookie, false);
+		__blk_mq_try_issue_directly(hctx, rq, cookie, false);
 		rcu_read_unlock();
 	} else {
 		unsigned int srcu_idx;
@@ -1452,7 +1458,7 @@ static void blk_mq_try_issue_directly(struct blk_mq_hw_ctx *hctx,
 		might_sleep();
 
 		srcu_idx = srcu_read_lock(&hctx->queue_rq_srcu);
-		__blk_mq_try_issue_directly(rq, cookie, true);
+		__blk_mq_try_issue_directly(hctx, rq, cookie, true);
 		srcu_read_unlock(&hctx->queue_rq_srcu, srcu_idx);
 	}
 }
@@ -1555,9 +1561,12 @@ static blk_qc_t blk_mq_make_request(struct request_queue *q, struct bio *bio)
 
 		blk_mq_put_ctx(data.ctx);
 
-		if (same_queue_rq)
+		if (same_queue_rq) {
+			data.hctx = blk_mq_map_queue(q,
+					same_queue_rq->mq_ctx->cpu);
 			blk_mq_try_issue_directly(data.hctx, same_queue_rq,
 					&cookie);
+		}
 	} else if (q->nr_hw_queues > 1 && is_sync) {
 		blk_mq_put_ctx(data.ctx);
 		blk_mq_bio_to_request(rq, bio);
@@ -2578,7 +2587,8 @@ int blk_mq_update_nr_requests(struct request_queue *q, unsigned int nr)
 	return ret;
 }
 
-void blk_mq_update_nr_hw_queues(struct blk_mq_tag_set *set, int nr_hw_queues)
+static void __blk_mq_update_nr_hw_queues(struct blk_mq_tag_set *set,
+							int nr_hw_queues)
 {
 	struct request_queue *q;
 
@@ -2602,6 +2612,13 @@ void blk_mq_update_nr_hw_queues(struct blk_mq_tag_set *set, int nr_hw_queues)
 	list_for_each_entry(q, &set->tag_list, tag_set_list)
 		blk_mq_unfreeze_queue(q);
 }
+
+void blk_mq_update_nr_hw_queues(struct blk_mq_tag_set *set, int nr_hw_queues)
+{
+	mutex_lock(&set->tag_list_lock);
+	__blk_mq_update_nr_hw_queues(set, nr_hw_queues);
+	mutex_unlock(&set->tag_list_lock);
+}
 EXPORT_SYMBOL_GPL(blk_mq_update_nr_hw_queues);
 
 /* Enable polling stats and return whether they were already enabled. */
diff --git a/block/blk-sysfs.c b/block/blk-sysfs.c
index 712b018e9f54..283da7fbe034 100644
--- a/block/blk-sysfs.c
+++ b/block/blk-sysfs.c
@@ -809,7 +809,7 @@ static void blk_release_queue(struct kobject *kobj)
 
 	blk_free_queue_stats(q->stats);
 
-	blk_exit_rl(&q->root_rl);
+	blk_exit_rl(q, &q->root_rl);
 
 	if (q->queue_tags)
 		__blk_queue_free_tags(q);
diff --git a/block/blk-throttle.c b/block/blk-throttle.c
index fc13dd0c6e39..a7285bf2831c 100644
--- a/block/blk-throttle.c
+++ b/block/blk-throttle.c
@@ -27,6 +27,13 @@ static int throtl_quantum = 32;
 #define MIN_THROTL_IOPS (10)
 #define DFL_LATENCY_TARGET (-1L)
 #define DFL_IDLE_THRESHOLD (0)
+#define DFL_HD_BASELINE_LATENCY (4000L) /* 4ms */
+#define LATENCY_FILTERED_SSD (0)
+/*
+ * For HD, very small latency comes from sequential IO. Such IO is helpless to
+ * help determine if its IO is impacted by others, hence we ignore the IO
+ */
+#define LATENCY_FILTERED_HD (1000L) /* 1ms */
 
 #define SKIP_LATENCY (((u64)1) << BLK_STAT_RES_SHIFT)
 
@@ -212,6 +219,7 @@ struct throtl_data
 	struct avg_latency_bucket avg_buckets[LATENCY_BUCKET_SIZE];
 	struct latency_bucket __percpu *latency_buckets;
 	unsigned long last_calculate_time;
+	unsigned long filtered_latency;
 
 	bool track_bio_latency;
 };
@@ -698,7 +706,7 @@ static void throtl_dequeue_tg(struct throtl_grp *tg)
 static void throtl_schedule_pending_timer(struct throtl_service_queue *sq,
 					  unsigned long expires)
 {
-	unsigned long max_expire = jiffies + 8 * sq_to_tg(sq)->td->throtl_slice;
+	unsigned long max_expire = jiffies + 8 * sq_to_td(sq)->throtl_slice;
 
 	/*
 	 * Since we are adjusting the throttle limit dynamically, the sleep
@@ -2281,7 +2289,7 @@ void blk_throtl_bio_endio(struct bio *bio)
 		throtl_track_latency(tg->td, blk_stat_size(&bio->bi_issue_stat),
 			bio_op(bio), lat);
 
-	if (tg->latency_target) {
+	if (tg->latency_target && lat >= tg->td->filtered_latency) {
 		int bucket;
 		unsigned int threshold;
 
@@ -2417,14 +2425,20 @@ void blk_throtl_exit(struct request_queue *q)
 void blk_throtl_register_queue(struct request_queue *q)
 {
 	struct throtl_data *td;
+	int i;
 
 	td = q->td;
 	BUG_ON(!td);
 
-	if (blk_queue_nonrot(q))
+	if (blk_queue_nonrot(q)) {
 		td->throtl_slice = DFL_THROTL_SLICE_SSD;
-	else
+		td->filtered_latency = LATENCY_FILTERED_SSD;
+	} else {
 		td->throtl_slice = DFL_THROTL_SLICE_HD;
+		td->filtered_latency = LATENCY_FILTERED_HD;
+		for (i = 0; i < LATENCY_BUCKET_SIZE; i++)
+			td->avg_buckets[i].latency = DFL_HD_BASELINE_LATENCY;
+	}
 #ifndef CONFIG_BLK_DEV_THROTTLING_LOW
 	/* if no low limit, use previous default */
 	td->throtl_slice = DFL_THROTL_SLICE_HD;
diff --git a/block/blk.h b/block/blk.h
index 2ed70228e44f..83c8e1100525 100644
--- a/block/blk.h
+++ b/block/blk.h
@@ -59,7 +59,7 @@ void blk_free_flush_queue(struct blk_flush_queue *q);
 
 int blk_init_rl(struct request_list *rl, struct request_queue *q,
 		gfp_t gfp_mask);
-void blk_exit_rl(struct request_list *rl);
+void blk_exit_rl(struct request_queue *q, struct request_list *rl);
 void blk_rq_bio_prep(struct request_queue *q, struct request *rq,
 			struct bio *bio);
 void blk_queue_bypass_start(struct request_queue *q);
diff --git a/block/cfq-iosched.c b/block/cfq-iosched.c
index f57bc7d5c483..3d5c28945719 100644
--- a/block/cfq-iosched.c
+++ b/block/cfq-iosched.c
@@ -38,9 +38,13 @@ static const u64 cfq_target_latency = (u64)NSEC_PER_SEC * 3/10; /* 300 ms */
 static const int cfq_hist_divisor = 4;
 
 /*
- * offset from end of service tree
+ * offset from end of queue service tree for idle class
  */
 #define CFQ_IDLE_DELAY		(NSEC_PER_SEC / 5)
+/* offset from end of group service tree under time slice mode */
+#define CFQ_SLICE_MODE_GROUP_DELAY (NSEC_PER_SEC / 5)
+/* offset from end of group service under IOPS mode */
+#define CFQ_IOPS_MODE_GROUP_DELAY (HZ / 5)
 
 /*
  * below this threshold, we consider thinktime immediate
@@ -1353,6 +1357,14 @@ cfq_group_service_tree_add(struct cfq_rb_root *st, struct cfq_group *cfqg)
 	cfqg->vfraction = max_t(unsigned, vfr, 1);
 }
 
+static inline u64 cfq_get_cfqg_vdisktime_delay(struct cfq_data *cfqd)
+{
+	if (!iops_mode(cfqd))
+		return CFQ_SLICE_MODE_GROUP_DELAY;
+	else
+		return CFQ_IOPS_MODE_GROUP_DELAY;
+}
+
 static void
 cfq_group_notify_queue_add(struct cfq_data *cfqd, struct cfq_group *cfqg)
 {
@@ -1372,7 +1384,8 @@ cfq_group_notify_queue_add(struct cfq_data *cfqd, struct cfq_group *cfqg)
 	n = rb_last(&st->rb);
 	if (n) {
 		__cfqg = rb_entry_cfqg(n);
-		cfqg->vdisktime = __cfqg->vdisktime + CFQ_IDLE_DELAY;
+		cfqg->vdisktime = __cfqg->vdisktime +
+			cfq_get_cfqg_vdisktime_delay(cfqd);
 	} else
 		cfqg->vdisktime = st->min_vdisktime;
 	cfq_group_service_tree_add(st, cfqg);