blkcg: unify blkg's for blkcg policies

Currently, blkg is per cgroup-queue-policy combination.  This is
unnatural and leads to various convolutions in partially used
duplicate fields in blkg, config / stat access, and general management
of blkgs.

This patch make blkg's per cgroup-queue and let them serve all
policies.  blkgs are now created and destroyed by blkcg core proper.
This will allow further consolidation of common management logic into
blkcg core and API with better defined semantics and layering.

As a transitional step to untangle blkg management, elvswitch and
policy [de]registration, all blkgs except the root blkg are being shot
down during elvswitch and bypass.  This patch adds blkg_root_update()
to update root blkg in place on policy change.  This is hacky and racy
but should be good enough as interim step until we get locking
simplified and switch over to proper in-place update for all blkgs.

-v2: Root blkgs need to be updated on elvswitch too and blkg_alloc()
     comment wasn't updated according to the function change.  Fixed.
     Both pointed out by Vivek.

-v3: v2 updated blkg_destroy_all() to invoke update_root_blkg_pd() for
     all policies.  This freed root pd during elvswitch before the
     last queue finished exiting and led to oops.  Directly invoke
     update_root_blkg_pd() only on BLKIO_POLICY_PROP from
     cfq_exit_queue().  This also is closer to what will be done with
     proper in-place blkg update.  Reported by Vivek.

Signed-off-by: Tejun Heo <tj@kernel.org>
Cc: Vivek Goyal <vgoyal@redhat.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>

1 files changed, 137 insertions, 87 deletions

diff --git a/block/blk-cgroup.c b/block/blk-cgroup.c
index 2ca9a15db0f..cad5f15cf49 100644
--- a/block/blk-cgroup.c
+++ b/block/blk-cgroup.c
@@ -461,16 +461,20 @@ EXPORT_SYMBOL_GPL(blkiocg_update_io_merged_stats);
  */
 static void blkg_free(struct blkio_group *blkg)
 {
-	struct blkg_policy_data *pd;
+	int i;
 
 	if (!blkg)
 		return;
 
-	pd = blkg->pd[blkg->plid];
-	if (pd) {
-		free_percpu(pd->stats_cpu);
-		kfree(pd);
+	for (i = 0; i < BLKIO_NR_POLICIES; i++) {
+		struct blkg_policy_data *pd = blkg->pd[i];
+
+		if (pd) {
+			free_percpu(pd->stats_cpu);
+			kfree(pd);
+		}
 	}
+
 	kfree(blkg);
 }
 
@@ -478,19 +482,17 @@ static void blkg_free(struct blkio_group *blkg)
  * blkg_alloc - allocate a blkg
  * @blkcg: block cgroup the new blkg is associated with
  * @q: request_queue the new blkg is associated with
- * @pol: policy the new blkg is associated with
  *
- * Allocate a new blkg assocating @blkcg and @q for @pol.
+ * Allocate a new blkg assocating @blkcg and @q.
  *
  * FIXME: Should be called with queue locked but currently isn't due to
  *        percpu stat breakage.
  */
 static struct blkio_group *blkg_alloc(struct blkio_cgroup *blkcg,
-				      struct request_queue *q,
-				      struct blkio_policy_type *pol)
+				      struct request_queue *q)
 {
 	struct blkio_group *blkg;
-	struct blkg_policy_data *pd;
+	int i;
 
 	/* alloc and init base part */
 	blkg = kzalloc_node(sizeof(*blkg), GFP_ATOMIC, q->node);
@@ -499,34 +501,45 @@ static struct blkio_group *blkg_alloc(struct blkio_cgroup *blkcg,
 
 	spin_lock_init(&blkg->stats_lock);
 	rcu_assign_pointer(blkg->q, q);
-	INIT_LIST_HEAD(&blkg->q_node[0]);
-	INIT_LIST_HEAD(&blkg->q_node[1]);
+	INIT_LIST_HEAD(&blkg->q_node);
 	blkg->blkcg = blkcg;
-	blkg->plid = pol->plid;
 	blkg->refcnt = 1;
 	cgroup_path(blkcg->css.cgroup, blkg->path, sizeof(blkg->path));
 
-	/* alloc per-policy data and attach it to blkg */
-	pd = kzalloc_node(sizeof(*pd) + pol->pdata_size, GFP_ATOMIC,
-			  q->node);
-	if (!pd) {
-		blkg_free(blkg);
-		return NULL;
-	}
+	for (i = 0; i < BLKIO_NR_POLICIES; i++) {
+		struct blkio_policy_type *pol = blkio_policy[i];
+		struct blkg_policy_data *pd;
 
-	blkg->pd[pol->plid] = pd;
-	pd->blkg = blkg;
+		if (!pol)
+			continue;
+
+		/* alloc per-policy data and attach it to blkg */
+		pd = kzalloc_node(sizeof(*pd) + pol->pdata_size, GFP_ATOMIC,
+				  q->node);
+		if (!pd) {
+			blkg_free(blkg);
+			return NULL;
+		}
 
-	/* broken, read comment in the callsite */
+		blkg->pd[i] = pd;
+		pd->blkg = blkg;
 
-	pd->stats_cpu = alloc_percpu(struct blkio_group_stats_cpu);
-	if (!pd->stats_cpu) {
-		blkg_free(blkg);
-		return NULL;
+		/* broken, read comment in the callsite */
+		pd->stats_cpu = alloc_percpu(struct blkio_group_stats_cpu);
+		if (!pd->stats_cpu) {
+			blkg_free(blkg);
+			return NULL;
+		}
 	}
 
 	/* invoke per-policy init */
-	pol->ops.blkio_init_group_fn(blkg);
+	for (i = 0; i < BLKIO_NR_POLICIES; i++) {
+		struct blkio_policy_type *pol = blkio_policy[i];
+
+		if (pol)
+			pol->ops.blkio_init_group_fn(blkg);
+	}
+
 	return blkg;
 }
 
@@ -536,7 +549,6 @@ struct blkio_group *blkg_lookup_create(struct blkio_cgroup *blkcg,
 				       bool for_root)
 	__releases(q->queue_lock) __acquires(q->queue_lock)
 {
-	struct blkio_policy_type *pol = blkio_policy[plid];
 	struct blkio_group *blkg, *new_blkg;
 
 	WARN_ON_ONCE(!rcu_read_lock_held());
@@ -551,7 +563,7 @@ struct blkio_group *blkg_lookup_create(struct blkio_cgroup *blkcg,
 	if (unlikely(blk_queue_bypass(q)) && !for_root)
 		return ERR_PTR(blk_queue_dead(q) ? -EINVAL : -EBUSY);
 
-	blkg = blkg_lookup(blkcg, q, plid);
+	blkg = blkg_lookup(blkcg, q);
 	if (blkg)
 		return blkg;
 
@@ -571,7 +583,7 @@ struct blkio_group *blkg_lookup_create(struct blkio_cgroup *blkcg,
 	spin_unlock_irq(q->queue_lock);
 	rcu_read_unlock();
 
-	new_blkg = blkg_alloc(blkcg, q, pol);
+	new_blkg = blkg_alloc(blkcg, q);
 
 	rcu_read_lock();
 	spin_lock_irq(q->queue_lock);
@@ -583,7 +595,7 @@ struct blkio_group *blkg_lookup_create(struct blkio_cgroup *blkcg,
 	}
 
 	/* did someone beat us to it? */
-	blkg = blkg_lookup(blkcg, q, plid);
+	blkg = blkg_lookup(blkcg, q);
 	if (unlikely(blkg))
 		goto out;
 
@@ -598,8 +610,8 @@ struct blkio_group *blkg_lookup_create(struct blkio_cgroup *blkcg,
 	swap(blkg, new_blkg);
 
 	hlist_add_head_rcu(&blkg->blkcg_node, &blkcg->blkg_list);
-	list_add(&blkg->q_node[plid], &q->blkg_list[plid]);
-	q->nr_blkgs[plid]++;
+	list_add(&blkg->q_node, &q->blkg_list);
+	q->nr_blkgs++;
 
 	spin_unlock(&blkcg->lock);
 out:
@@ -636,31 +648,30 @@ EXPORT_SYMBOL_GPL(blkiocg_del_blkio_group);
 
 /* called under rcu_read_lock(). */
 struct blkio_group *blkg_lookup(struct blkio_cgroup *blkcg,
-				struct request_queue *q,
-				enum blkio_policy_id plid)
+				struct request_queue *q)
 {
 	struct blkio_group *blkg;
 	struct hlist_node *n;
 
 	hlist_for_each_entry_rcu(blkg, n, &blkcg->blkg_list, blkcg_node)
-		if (blkg->q == q && blkg->plid == plid)
+		if (blkg->q == q)
 			return blkg;
 	return NULL;
 }
 EXPORT_SYMBOL_GPL(blkg_lookup);
 
-static void blkg_destroy(struct blkio_group *blkg, enum blkio_policy_id plid)
+static void blkg_destroy(struct blkio_group *blkg)
 {
 	struct request_queue *q = blkg->q;
 
 	lockdep_assert_held(q->queue_lock);
 
 	/* Something wrong if we are trying to remove same group twice */
-	WARN_ON_ONCE(list_empty(&blkg->q_node[plid]));
-	list_del_init(&blkg->q_node[plid]);
+	WARN_ON_ONCE(list_empty(&blkg->q_node));
+	list_del_init(&blkg->q_node);
 
-	WARN_ON_ONCE(q->nr_blkgs[plid] <= 0);
-	q->nr_blkgs[plid]--;
+	WARN_ON_ONCE(q->nr_blkgs <= 0);
+	q->nr_blkgs--;
 
 	/*
 	 * Put the reference taken at the time of creation so that when all
@@ -669,8 +680,40 @@ static void blkg_destroy(struct blkio_group *blkg, enum blkio_policy_id plid)
 	blkg_put(blkg);
 }
 
-void blkg_destroy_all(struct request_queue *q, enum blkio_policy_id plid,
-		      bool destroy_root)
+/*
+ * 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, enum blkio_policy_id plid)
+{
+	struct blkio_policy_type *pol = blkio_policy[plid];
+	struct blkio_group *blkg = blkg_lookup(&blkio_root_cgroup, q);
+	struct blkg_policy_data *pd;
+
+	if (!blkg)
+		return;
+
+	kfree(blkg->pd[plid]);
+	blkg->pd[plid] = NULL;
+
+	if (!pol)
+		return;
+
+	pd = kzalloc(sizeof(*pd) + pol->pdata_size, GFP_KERNEL);
+	WARN_ON_ONCE(!pd);
+
+	pd->stats_cpu = alloc_percpu(struct blkio_group_stats_cpu);
+	WARN_ON_ONCE(!pd->stats_cpu);
+
+	blkg->pd[plid] = pd;
+	pd->blkg = blkg;
+	pol->ops.blkio_init_group_fn(blkg);
+}
+EXPORT_SYMBOL_GPL(update_root_blkg_pd);
+
+void blkg_destroy_all(struct request_queue *q, bool destroy_root)
 {
 	struct blkio_group *blkg, *n;
 
@@ -679,8 +722,7 @@ void blkg_destroy_all(struct request_queue *q, enum blkio_policy_id plid,
 
 		spin_lock_irq(q->queue_lock);
 
-		list_for_each_entry_safe(blkg, n, &q->blkg_list[plid],
-					 q_node[plid]) {
+		list_for_each_entry_safe(blkg, n, &q->blkg_list, q_node) {
 			/* skip root? */
 			if (!destroy_root && blkg->blkcg == &blkio_root_cgroup)
 				continue;
@@ -691,7 +733,7 @@ void blkg_destroy_all(struct request_queue *q, enum blkio_policy_id plid,
 			 * take care of destroying cfqg also.
 			 */
 			if (!blkiocg_del_blkio_group(blkg))
-				blkg_destroy(blkg, plid);
+				blkg_destroy(blkg);
 			else
 				done = false;
 		}
@@ -776,43 +818,49 @@ blkiocg_reset_stats(struct cgroup *cgroup, struct cftype *cftype, u64 val)
 #endif
 
 	blkcg = cgroup_to_blkio_cgroup(cgroup);
+	spin_lock(&blkio_list_lock);
 	spin_lock_irq(&blkcg->lock);
 	hlist_for_each_entry(blkg, n, &blkcg->blkg_list, blkcg_node) {
-		struct blkg_policy_data *pd = blkg->pd[blkg->plid];
+		struct blkio_policy_type *pol;
 
-		spin_lock(&blkg->stats_lock);
-		stats = &pd->stats;
+		list_for_each_entry(pol, &blkio_list, list) {
+			struct blkg_policy_data *pd = blkg->pd[pol->plid];
+
+			spin_lock(&blkg->stats_lock);
+			stats = &pd->stats;
 #ifdef CONFIG_DEBUG_BLK_CGROUP
-		idling = blkio_blkg_idling(stats);
-		waiting = blkio_blkg_waiting(stats);
-		empty = blkio_blkg_empty(stats);
+			idling = blkio_blkg_idling(stats);
+			waiting = blkio_blkg_waiting(stats);
+			empty = blkio_blkg_empty(stats);
 #endif
-		for (i = 0; i < BLKIO_STAT_TOTAL; i++)
-			queued[i] = stats->stat_arr[BLKIO_STAT_QUEUED][i];
-		memset(stats, 0, sizeof(struct blkio_group_stats));
-		for (i = 0; i < BLKIO_STAT_TOTAL; i++)
-			stats->stat_arr[BLKIO_STAT_QUEUED][i] = queued[i];
+			for (i = 0; i < BLKIO_STAT_TOTAL; i++)
+				queued[i] = stats->stat_arr[BLKIO_STAT_QUEUED][i];
+			memset(stats, 0, sizeof(struct blkio_group_stats));
+			for (i = 0; i < BLKIO_STAT_TOTAL; i++)
+				stats->stat_arr[BLKIO_STAT_QUEUED][i] = queued[i];
 #ifdef CONFIG_DEBUG_BLK_CGROUP
-		if (idling) {
-			blkio_mark_blkg_idling(stats);
-			stats->start_idle_time = now;
-		}
-		if (waiting) {
-			blkio_mark_blkg_waiting(stats);
-			stats->start_group_wait_time = now;
-		}
-		if (empty) {
-			blkio_mark_blkg_empty(stats);
-			stats->start_empty_time = now;
-		}
+			if (idling) {
+				blkio_mark_blkg_idling(stats);
+				stats->start_idle_time = now;
+			}
+			if (waiting) {
+				blkio_mark_blkg_waiting(stats);
+				stats->start_group_wait_time = now;
+			}
+			if (empty) {
+				blkio_mark_blkg_empty(stats);
+				stats->start_empty_time = now;
+			}
 #endif
-		spin_unlock(&blkg->stats_lock);
+			spin_unlock(&blkg->stats_lock);
 
-		/* Reset Per cpu stats which don't take blkg->stats_lock */
-		blkio_reset_stats_cpu(blkg, blkg->plid);
+			/* Reset Per cpu stats which don't take blkg->stats_lock */
+			blkio_reset_stats_cpu(blkg, pol->plid);
+		}
 	}
 
 	spin_unlock_irq(&blkcg->lock);
+	spin_unlock(&blkio_list_lock);
 	return 0;
 }
 
@@ -1168,8 +1216,7 @@ static void blkio_read_conf(struct cftype *cft, struct blkio_cgroup *blkcg,
 
 	spin_lock_irq(&blkcg->lock);
 	hlist_for_each_entry(blkg, n, &blkcg->blkg_list, blkcg_node)
-		if (BLKIOFILE_POLICY(cft->private) == blkg->plid)
-			blkio_print_group_conf(cft, blkg, m);
+		blkio_print_group_conf(cft, blkg, m);
 	spin_unlock_irq(&blkcg->lock);
 }
 
@@ -1224,7 +1271,7 @@ static int blkio_read_blkg_stats(struct blkio_cgroup *blkcg,
 		const char *dname = blkg_dev_name(blkg);
 		int plid = BLKIOFILE_POLICY(cft->private);
 
-		if (!dname || plid != blkg->plid)
+		if (!dname)
 			continue;
 		if (pcpu) {
 			cgroup_total += blkio_get_stat_cpu(blkg, plid,
@@ -1335,9 +1382,9 @@ static int blkio_weight_write(struct blkio_cgroup *blkcg, int plid, u64 val)
 	blkcg->weight = (unsigned int)val;
 
 	hlist_for_each_entry(blkg, n, &blkcg->blkg_list, blkcg_node) {
-		struct blkg_policy_data *pd = blkg->pd[blkg->plid];
+		struct blkg_policy_data *pd = blkg->pd[plid];
 
-		if (blkg->plid == plid && !pd->conf.weight)
+		if (!pd->conf.weight)
 			blkio_update_group_weight(blkg, plid, blkcg->weight);
 	}
 
@@ -1560,7 +1607,6 @@ static int blkiocg_pre_destroy(struct cgroup_subsys *subsys,
 	unsigned long flags;
 	struct blkio_group *blkg;
 	struct request_queue *q;
-	struct blkio_policy_type *blkiop;
 
 	rcu_read_lock();
 
@@ -1586,11 +1632,7 @@ static int blkiocg_pre_destroy(struct cgroup_subsys *subsys,
 		 */
 		spin_lock(&blkio_list_lock);
 		spin_lock_irqsave(q->queue_lock, flags);
-		list_for_each_entry(blkiop, &blkio_list, list) {
-			if (blkiop->plid != blkg->plid)
-				continue;
-			blkg_destroy(blkg, blkiop->plid);
-		}
+		blkg_destroy(blkg);
 		spin_unlock_irqrestore(q->queue_lock, flags);
 		spin_unlock(&blkio_list_lock);
 	} while (1);
@@ -1684,6 +1726,8 @@ void blkcg_exit_queue(struct request_queue *q)
 	list_del_init(&q->all_q_node);
 	mutex_unlock(&all_q_mutex);
 
+	blkg_destroy_all(q, true);
+
 	blk_throtl_exit(q);
 }
 
@@ -1733,14 +1777,12 @@ static void blkcg_bypass_start(void)
 	__acquires(&all_q_mutex)
 {
 	struct request_queue *q;
-	int i;
 
 	mutex_lock(&all_q_mutex);
 
 	list_for_each_entry(q, &all_q_list, all_q_node) {
 		blk_queue_bypass_start(q);
-		for (i = 0; i < BLKIO_NR_POLICIES; i++)
-			blkg_destroy_all(q, i, false);
+		blkg_destroy_all(q, false);
 	}
 }
 
@@ -1757,6 +1799,8 @@ static void blkcg_bypass_end(void)
 
 void blkio_policy_register(struct blkio_policy_type *blkiop)
 {
+	struct request_queue *q;
+
 	blkcg_bypass_start();
 	spin_lock(&blkio_list_lock);
 
@@ -1765,12 +1809,16 @@ void blkio_policy_register(struct blkio_policy_type *blkiop)
 	list_add_tail(&blkiop->list, &blkio_list);
 
 	spin_unlock(&blkio_list_lock);
+	list_for_each_entry(q, &all_q_list, all_q_node)
+		update_root_blkg_pd(q, blkiop->plid);
 	blkcg_bypass_end();
 }
 EXPORT_SYMBOL_GPL(blkio_policy_register);
 
 void blkio_policy_unregister(struct blkio_policy_type *blkiop)
 {
+	struct request_queue *q;
+
 	blkcg_bypass_start();
 	spin_lock(&blkio_list_lock);
 
@@ -1779,6 +1827,8 @@ void blkio_policy_unregister(struct blkio_policy_type *blkiop)
 	list_del_init(&blkiop->list);
 
 	spin_unlock(&blkio_list_lock);
+	list_for_each_entry(q, &all_q_list, all_q_node)
+		update_root_blkg_pd(q, blkiop->plid);
 	blkcg_bypass_end();
 }
 EXPORT_SYMBOL_GPL(blkio_policy_unregister);