diff --git a/block/blk-cgroup.c b/block/blk-cgroup.c
index d6e4555c982fe5bacb54719f1bf442869e9cac6b..d6d59ad105b4dfcfddbc1ca2c9f9907942d2767e 100644
--- a/block/blk-cgroup.c
+++ b/block/blk-cgroup.c
@@ -54,6 +54,17 @@ struct blkio_cgroup *bio_blkio_cgroup(struct bio *bio)
 }
 EXPORT_SYMBOL_GPL(bio_blkio_cgroup);
 
+static bool blkcg_policy_enabled(struct request_queue *q,
+				 const struct blkio_policy_type *pol)
+{
+	return pol && test_bit(pol->plid, q->blkcg_pols);
+}
+
+static size_t blkg_pd_size(const struct blkio_policy_type *pol)
+{
+	return sizeof(struct blkg_policy_data) + pol->pdata_size;
+}
+
 /**
  * blkg_free - free a blkg
  * @blkg: blkg to free
@@ -111,12 +122,11 @@ static struct blkio_group *blkg_alloc(struct blkio_cgroup *blkcg,
 		struct blkio_policy_type *pol = blkio_policy[i];
 		struct blkg_policy_data *pd;
 
-		if (!pol)
+		if (!blkcg_policy_enabled(q, pol))
 			continue;
 
 		/* alloc per-policy data and attach it to blkg */
-		pd = kzalloc_node(sizeof(*pd) + pol->pdata_size, GFP_ATOMIC,
-				  q->node);
+		pd = kzalloc_node(blkg_pd_size(pol), GFP_ATOMIC, q->node);
 		if (!pd) {
 			blkg_free(blkg);
 			return NULL;
@@ -130,7 +140,7 @@ static struct blkio_group *blkg_alloc(struct blkio_cgroup *blkcg,
 	for (i = 0; i < BLKCG_MAX_POLS; i++) {
 		struct blkio_policy_type *pol = blkio_policy[i];
 
-		if (pol)
+		if (blkcg_policy_enabled(blkg->q, pol))
 			pol->ops.blkio_init_group_fn(blkg);
 	}
 
@@ -236,36 +246,6 @@ static void blkg_destroy(struct blkio_group *blkg)
 	blkg_put(blkg);
 }
 
-/*
- * XXX: This updates blkg policy data in-place for root blkg, which is
- * necessary across elevator switch and policy registration as root blkgs
- * aren't shot down.  This broken and racy implementation is temporary.
- * Eventually, blkg shoot down will be replaced by proper in-place update.
- */
-void update_root_blkg_pd(struct request_queue *q,
-			 const struct blkio_policy_type *pol)
-{
-	struct blkio_group *blkg = blkg_lookup(&blkio_root_cgroup, q);
-	struct blkg_policy_data *pd;
-
-	if (!blkg)
-		return;
-
-	kfree(blkg->pd[pol->plid]);
-	blkg->pd[pol->plid] = NULL;
-
-	if (!pol)
-		return;
-
-	pd = kzalloc(sizeof(*pd) + pol->pdata_size, GFP_KERNEL);
-	WARN_ON_ONCE(!pd);
-
-	blkg->pd[pol->plid] = pd;
-	pd->blkg = blkg;
-	pol->ops.blkio_init_group_fn(blkg);
-}
-EXPORT_SYMBOL_GPL(update_root_blkg_pd);
-
 /**
  * blkg_destroy_all - destroy all blkgs associated with a request_queue
  * @q: request_queue of interest
@@ -339,7 +319,8 @@ blkiocg_reset_stats(struct cgroup *cgroup, struct cftype *cftype, u64 val)
 		for (i = 0; i < BLKCG_MAX_POLS; i++) {
 			struct blkio_policy_type *pol = blkio_policy[i];
 
-			if (pol && pol->ops.blkio_reset_group_stats_fn)
+			if (blkcg_policy_enabled(blkg->q, pol) &&
+			    pol->ops.blkio_reset_group_stats_fn)
 				pol->ops.blkio_reset_group_stats_fn(blkg);
 		}
 	}
@@ -385,7 +366,7 @@ void blkcg_print_blkgs(struct seq_file *sf, struct blkio_cgroup *blkcg,
 
 	spin_lock_irq(&blkcg->lock);
 	hlist_for_each_entry(blkg, n, &blkcg->blkg_list, blkcg_node)
-		if (blkg->pd[pol->plid])
+		if (blkcg_policy_enabled(blkg->q, pol))
 			total += prfill(sf, blkg->pd[pol->plid]->pdata, data);
 	spin_unlock_irq(&blkcg->lock);
 
@@ -510,7 +491,10 @@ int blkg_conf_prep(struct blkio_cgroup *blkcg,
 	rcu_read_lock();
 	spin_lock_irq(disk->queue->queue_lock);
 
-	blkg = blkg_lookup_create(blkcg, disk->queue, false);
+	if (blkcg_policy_enabled(disk->queue, pol))
+		blkg = blkg_lookup_create(blkcg, disk->queue, false);
+	else
+		blkg = ERR_PTR(-EINVAL);
 
 	if (IS_ERR(blkg)) {
 		ret = PTR_ERR(blkg);
@@ -712,30 +696,6 @@ static int blkiocg_can_attach(struct cgroup *cgrp, struct cgroup_taskset *tset)
 	return ret;
 }
 
-static void blkcg_bypass_start(void)
-	__acquires(&all_q_mutex)
-{
-	struct request_queue *q;
-
-	mutex_lock(&all_q_mutex);
-
-	list_for_each_entry(q, &all_q_list, all_q_node) {
-		blk_queue_bypass_start(q);
-		blkg_destroy_all(q, false);
-	}
-}
-
-static void blkcg_bypass_end(void)
-	__releases(&all_q_mutex)
-{
-	struct request_queue *q;
-
-	list_for_each_entry(q, &all_q_list, all_q_node)
-		blk_queue_bypass_end(q);
-
-	mutex_unlock(&all_q_mutex);
-}
-
 struct cgroup_subsys blkio_subsys = {
 	.name = "blkio",
 	.create = blkiocg_create,
@@ -748,6 +708,139 @@ struct cgroup_subsys blkio_subsys = {
 };
 EXPORT_SYMBOL_GPL(blkio_subsys);
 
+/**
+ * blkcg_activate_policy - activate a blkcg policy on a request_queue
+ * @q: request_queue of interest
+ * @pol: blkcg policy to activate
+ *
+ * Activate @pol on @q.  Requires %GFP_KERNEL context.  @q goes through
+ * bypass mode to populate its blkgs with policy_data for @pol.
+ *
+ * Activation happens with @q bypassed, so nobody would be accessing blkgs
+ * from IO path.  Update of each blkg is protected by both queue and blkcg
+ * locks so that holding either lock and testing blkcg_policy_enabled() is
+ * always enough for dereferencing policy data.
+ *
+ * The caller is responsible for synchronizing [de]activations and policy
+ * [un]registerations.  Returns 0 on success, -errno on failure.
+ */
+int blkcg_activate_policy(struct request_queue *q,
+			  const struct blkio_policy_type *pol)
+{
+	LIST_HEAD(pds);
+	struct blkio_group *blkg;
+	struct blkg_policy_data *pd, *n;
+	int cnt = 0, ret;
+
+	if (blkcg_policy_enabled(q, pol))
+		return 0;
+
+	blk_queue_bypass_start(q);
+
+	/* make sure the root blkg exists and count the existing blkgs */
+	spin_lock_irq(q->queue_lock);
+
+	rcu_read_lock();
+	blkg = blkg_lookup_create(&blkio_root_cgroup, q, true);
+	rcu_read_unlock();
+
+	if (IS_ERR(blkg)) {
+		ret = PTR_ERR(blkg);
+		goto out_unlock;
+	}
+	q->root_blkg = blkg;
+
+	list_for_each_entry(blkg, &q->blkg_list, q_node)
+		cnt++;
+
+	spin_unlock_irq(q->queue_lock);
+
+	/* allocate policy_data for all existing blkgs */
+	while (cnt--) {
+		pd = kzalloc_node(blkg_pd_size(pol), GFP_KERNEL, q->node);
+		if (!pd) {
+			ret = -ENOMEM;
+			goto out_free;
+		}
+		list_add_tail(&pd->alloc_node, &pds);
+	}
+
+	/*
+	 * Install the allocated pds.  With @q bypassing, no new blkg
+	 * should have been created while the queue lock was dropped.
+	 */
+	spin_lock_irq(q->queue_lock);
+
+	list_for_each_entry(blkg, &q->blkg_list, q_node) {
+		if (WARN_ON(list_empty(&pds))) {
+			/* umm... this shouldn't happen, just abort */
+			ret = -ENOMEM;
+			goto out_unlock;
+		}
+		pd = list_first_entry(&pds, struct blkg_policy_data, alloc_node);
+		list_del_init(&pd->alloc_node);
+
+		/* grab blkcg lock too while installing @pd on @blkg */
+		spin_lock(&blkg->blkcg->lock);
+
+		blkg->pd[pol->plid] = pd;
+		pd->blkg = blkg;
+		pol->ops.blkio_init_group_fn(blkg);
+
+		spin_unlock(&blkg->blkcg->lock);
+	}
+
+	__set_bit(pol->plid, q->blkcg_pols);
+	ret = 0;
+out_unlock:
+	spin_unlock_irq(q->queue_lock);
+out_free:
+	blk_queue_bypass_end(q);
+	list_for_each_entry_safe(pd, n, &pds, alloc_node)
+		kfree(pd);
+	return ret;
+}
+EXPORT_SYMBOL_GPL(blkcg_activate_policy);
+
+/**
+ * blkcg_deactivate_policy - deactivate a blkcg policy on a request_queue
+ * @q: request_queue of interest
+ * @pol: blkcg policy to deactivate
+ *
+ * Deactivate @pol on @q.  Follows the same synchronization rules as
+ * blkcg_activate_policy().
+ */
+void blkcg_deactivate_policy(struct request_queue *q,
+			     const struct blkio_policy_type *pol)
+{
+	struct blkio_group *blkg;
+
+	if (!blkcg_policy_enabled(q, pol))
+		return;
+
+	blk_queue_bypass_start(q);
+	spin_lock_irq(q->queue_lock);
+
+	__clear_bit(pol->plid, q->blkcg_pols);
+
+	list_for_each_entry(blkg, &q->blkg_list, q_node) {
+		/* grab blkcg lock too while removing @pd from @blkg */
+		spin_lock(&blkg->blkcg->lock);
+
+		if (pol->ops.blkio_exit_group_fn)
+			pol->ops.blkio_exit_group_fn(blkg);
+
+		kfree(blkg->pd[pol->plid]);
+		blkg->pd[pol->plid] = NULL;
+
+		spin_unlock(&blkg->blkcg->lock);
+	}
+
+	spin_unlock_irq(q->queue_lock);
+	blk_queue_bypass_end(q);
+}
+EXPORT_SYMBOL_GPL(blkcg_deactivate_policy);
+
 /**
  * blkio_policy_register - register a blkcg policy
  * @blkiop: blkcg policy to register
@@ -758,7 +851,6 @@ EXPORT_SYMBOL_GPL(blkio_subsys);
  */
 int blkio_policy_register(struct blkio_policy_type *blkiop)
 {
-	struct request_queue *q;
 	int i, ret;
 
 	mutex_lock(&blkcg_pol_mutex);
@@ -775,11 +867,6 @@ int blkio_policy_register(struct blkio_policy_type *blkiop)
 	blkiop->plid = i;
 	blkio_policy[i] = blkiop;
 
-	blkcg_bypass_start();
-	list_for_each_entry(q, &all_q_list, all_q_node)
-		update_root_blkg_pd(q, blkiop);
-	blkcg_bypass_end();
-
 	/* everything is in place, add intf files for the new policy */
 	if (blkiop->cftypes)
 		WARN_ON(cgroup_add_cftypes(&blkio_subsys, blkiop->cftypes));
@@ -798,8 +885,6 @@ EXPORT_SYMBOL_GPL(blkio_policy_register);
  */
 void blkio_policy_unregister(struct blkio_policy_type *blkiop)
 {
-	struct request_queue *q;
-
 	mutex_lock(&blkcg_pol_mutex);
 
 	if (WARN_ON(blkio_policy[blkiop->plid] != blkiop))
@@ -811,11 +896,6 @@ void blkio_policy_unregister(struct blkio_policy_type *blkiop)
 
 	/* unregister and update blkgs */
 	blkio_policy[blkiop->plid] = NULL;
-
-	blkcg_bypass_start();
-	list_for_each_entry(q, &all_q_list, all_q_node)
-		update_root_blkg_pd(q, blkiop);
-	blkcg_bypass_end();
 out_unlock:
 	mutex_unlock(&blkcg_pol_mutex);
 }
diff --git a/block/blk-cgroup.h b/block/blk-cgroup.h
index df1c7b290c22b5344fcb47fe334c8cbd0671a623..66253a7c8ff4cb90683b2092edb8239cc7eb5124 100644
--- a/block/blk-cgroup.h
+++ b/block/blk-cgroup.h
@@ -64,6 +64,9 @@ struct blkg_policy_data {
 	/* the blkg this per-policy data belongs to */
 	struct blkio_group *blkg;
 
+	/* used during policy activation */
+	struct list_head alloc_node;
+
 	/* pol->pdata_size bytes of private data used by policy impl */
 	char pdata[] __aligned(__alignof__(unsigned long long));
 };
@@ -108,9 +111,11 @@ extern void blkcg_exit_queue(struct request_queue *q);
 /* Blkio controller policy registration */
 extern int blkio_policy_register(struct blkio_policy_type *);
 extern void blkio_policy_unregister(struct blkio_policy_type *);
+extern int blkcg_activate_policy(struct request_queue *q,
+				 const struct blkio_policy_type *pol);
+extern void blkcg_deactivate_policy(struct request_queue *q,
+				    const struct blkio_policy_type *pol);
 extern void blkg_destroy_all(struct request_queue *q, bool destroy_root);
-extern void update_root_blkg_pd(struct request_queue *q,
-				const struct blkio_policy_type *pol);
 
 void blkcg_print_blkgs(struct seq_file *sf, struct blkio_cgroup *blkcg,
 		       u64 (*prfill)(struct seq_file *, void *, int),
@@ -325,10 +330,12 @@ static inline void blkcg_drain_queue(struct request_queue *q) { }
 static inline void blkcg_exit_queue(struct request_queue *q) { }
 static inline int blkio_policy_register(struct blkio_policy_type *blkiop) { return 0; }
 static inline void blkio_policy_unregister(struct blkio_policy_type *blkiop) { }
+static inline int blkcg_activate_policy(struct request_queue *q,
+					const struct blkio_policy_type *pol) { return 0; }
+static inline void blkcg_deactivate_policy(struct request_queue *q,
+					   const struct blkio_policy_type *pol) { }
 static inline void blkg_destroy_all(struct request_queue *q,
 				    bool destory_root) { }
-static inline void update_root_blkg_pd(struct request_queue *q,
-				       const struct blkio_policy_type *pol) { }
 
 static inline void *blkg_to_pdata(struct blkio_group *blkg,
 				struct blkio_policy_type *pol) { return NULL; }
diff --git a/block/blk-throttle.c b/block/blk-throttle.c
index 8c520fad6885411465ba0ce0e02389fef6ffed42..2fc964e06ea427c760d59bbd6a9aff8738411e25 100644
--- a/block/blk-throttle.c
+++ b/block/blk-throttle.c
@@ -995,35 +995,31 @@ static int tg_set_conf(struct cgroup *cgrp, struct cftype *cft, const char *buf,
 	struct blkio_cgroup *blkcg = cgroup_to_blkio_cgroup(cgrp);
 	struct blkg_conf_ctx ctx;
 	struct throtl_grp *tg;
+	struct throtl_data *td;
 	int ret;
 
 	ret = blkg_conf_prep(blkcg, &blkio_policy_throtl, buf, &ctx);
 	if (ret)
 		return ret;
 
-	ret = -EINVAL;
 	tg = blkg_to_tg(ctx.blkg);
-	if (tg) {
-		struct throtl_data *td = ctx.blkg->q->td;
-
-		if (!ctx.v)
-			ctx.v = -1;
+	td = ctx.blkg->q->td;
 
-		if (is_u64)
-			*(u64 *)((void *)tg + cft->private) = ctx.v;
-		else
-			*(unsigned int *)((void *)tg + cft->private) = ctx.v;
+	if (!ctx.v)
+		ctx.v = -1;
 
-		/* XXX: we don't need the following deferred processing */
-		xchg(&tg->limits_changed, true);
-		xchg(&td->limits_changed, true);
-		throtl_schedule_delayed_work(td, 0);
+	if (is_u64)
+		*(u64 *)((void *)tg + cft->private) = ctx.v;
+	else
+		*(unsigned int *)((void *)tg + cft->private) = ctx.v;
 
-		ret = 0;
-	}
+	/* XXX: we don't need the following deferred processing */
+	xchg(&tg->limits_changed, true);
+	xchg(&td->limits_changed, true);
+	throtl_schedule_delayed_work(td, 0);
 
 	blkg_conf_finish(&ctx);
-	return ret;
+	return 0;
 }
 
 static int tg_set_conf_u64(struct cgroup *cgrp, struct cftype *cft,
@@ -1230,7 +1226,7 @@ void blk_throtl_drain(struct request_queue *q)
 int blk_throtl_init(struct request_queue *q)
 {
 	struct throtl_data *td;
-	struct blkio_group *blkg;
+	int ret;
 
 	td = kzalloc_node(sizeof(*td), GFP_KERNEL, q->node);
 	if (!td)
@@ -1243,28 +1239,18 @@ int blk_throtl_init(struct request_queue *q)
 	q->td = td;
 	td->queue = q;
 
-	/* alloc and init root group. */
-	rcu_read_lock();
-	spin_lock_irq(q->queue_lock);
-
-	blkg = blkg_lookup_create(&blkio_root_cgroup, q, true);
-	if (!IS_ERR(blkg))
-		q->root_blkg = blkg;
-
-	spin_unlock_irq(q->queue_lock);
-	rcu_read_unlock();
-
-	if (!q->root_blkg) {
+	/* activate policy */
+	ret = blkcg_activate_policy(q, &blkio_policy_throtl);
+	if (ret)
 		kfree(td);
-		return -ENOMEM;
-	}
-	return 0;
+	return ret;
 }
 
 void blk_throtl_exit(struct request_queue *q)
 {
 	BUG_ON(!q->td);
 	throtl_shutdown_wq(q);
+	blkcg_deactivate_policy(q, &blkio_policy_throtl);
 	kfree(q->td);
 }
 
diff --git a/block/cfq-iosched.c b/block/cfq-iosched.c
index 86440e04f3ee72e8e4ce854411492c359f1ae847..0203652e1f3488bd1a50e58e2ddc3fc928b2a2d0 100644
--- a/block/cfq-iosched.c
+++ b/block/cfq-iosched.c
@@ -1406,8 +1406,7 @@ static int cfqg_set_weight_device(struct cgroup *cgrp, struct cftype *cft,
 
 	ret = -EINVAL;
 	cfqg = blkg_to_cfqg(ctx.blkg);
-	if (cfqg && (!ctx.v || (ctx.v >= CFQ_WEIGHT_MIN &&
-				ctx.v <= CFQ_WEIGHT_MAX))) {
+	if (!ctx.v || (ctx.v >= CFQ_WEIGHT_MIN && ctx.v <= CFQ_WEIGHT_MAX)) {
 		cfqg->dev_weight = ctx.v;
 		cfqg->new_weight = cfqg->dev_weight ?: blkcg->cfq_weight;
 		ret = 0;
@@ -3938,7 +3937,7 @@ static void cfq_exit_queue(struct elevator_queue *e)
 #ifndef CONFIG_CFQ_GROUP_IOSCHED
 	kfree(cfqd->root_group);
 #endif
-	update_root_blkg_pd(q, &blkio_policy_cfq);
+	blkcg_deactivate_policy(q, &blkio_policy_cfq);
 	kfree(cfqd);
 }
 
@@ -3946,7 +3945,7 @@ static int cfq_init_queue(struct request_queue *q)
 {
 	struct cfq_data *cfqd;
 	struct blkio_group *blkg __maybe_unused;
-	int i;
+	int i, ret;
 
 	cfqd = kmalloc_node(sizeof(*cfqd), GFP_KERNEL | __GFP_ZERO, q->node);
 	if (!cfqd)
@@ -3960,28 +3959,20 @@ static int cfq_init_queue(struct request_queue *q)
 
 	/* Init root group and prefer root group over other groups by default */
 #ifdef CONFIG_CFQ_GROUP_IOSCHED
-	rcu_read_lock();
-	spin_lock_irq(q->queue_lock);
-
-	blkg = blkg_lookup_create(&blkio_root_cgroup, q, true);
-	if (!IS_ERR(blkg)) {
-		q->root_blkg = blkg;
-		cfqd->root_group = blkg_to_cfqg(blkg);
-	}
+	ret = blkcg_activate_policy(q, &blkio_policy_cfq);
+	if (ret)
+		goto out_free;
 
-	spin_unlock_irq(q->queue_lock);
-	rcu_read_unlock();
+	cfqd->root_group = blkg_to_cfqg(q->root_blkg);
 #else
+	ret = -ENOMEM;
 	cfqd->root_group = kzalloc_node(sizeof(*cfqd->root_group),
 					GFP_KERNEL, cfqd->queue->node);
-	if (cfqd->root_group)
-		cfq_init_cfqg_base(cfqd->root_group);
-#endif
-	if (!cfqd->root_group) {
-		kfree(cfqd);
-		return -ENOMEM;
-	}
+	if (!cfqd->root_group)
+		goto out_free;
 
+	cfq_init_cfqg_base(cfqd->root_group);
+#endif
 	cfqd->root_group->weight = 2 * CFQ_WEIGHT_DEFAULT;
 
 	/*
@@ -4031,6 +4022,10 @@ static int cfq_init_queue(struct request_queue *q)
 	 */
 	cfqd->last_delayed_sync = jiffies - HZ;
 	return 0;
+
+out_free:
+	kfree(cfqd);
+	return ret;
 }
 
 /*
diff --git a/block/elevator.c b/block/elevator.c
index be3ab6df0feaf4751af2c68781b8a2a538ed0c9a..6a55d418896f5ceee0042da69c0177c495219cbc 100644
--- a/block/elevator.c
+++ b/block/elevator.c
@@ -896,8 +896,6 @@ static int elevator_switch(struct request_queue *q, struct elevator_type *new_e)
 	ioc_clear_queue(q);
 	spin_unlock_irq(q->queue_lock);
 
-	blkg_destroy_all(q, false);
-
 	/* allocate, init and register new elevator */
 	err = -ENOMEM;
 	q->elevator = elevator_alloc(q, new_e);
diff --git a/include/linux/blkdev.h b/include/linux/blkdev.h
index b01c377fd73980e2b38c5b1a37c1a7c17536ff33..68720ab275d4dd13ef79ebb56977a3fdee303b9d 100644
--- a/include/linux/blkdev.h
+++ b/include/linux/blkdev.h
@@ -370,6 +370,7 @@ struct request_queue {
 
 	struct list_head	icq_list;
 #ifdef CONFIG_BLK_CGROUP
+	DECLARE_BITMAP		(blkcg_pols, BLKCG_MAX_POLS);
 	struct blkio_group	*root_blkg;
 	struct list_head	blkg_list;
 #endif