blkcg: replace blkcg_policy->pd_size with ->pd_alloc/free_fn() methods

A blkg (blkcg_gq) represents the relationship between a cgroup and
request_queue.  Each active policy has a pd (blkg_policy_data) on each
blkg.  The pd's were allocated by blkcg core and each policy could
request to allocate extra space at the end by setting
blkcg_policy->pd_size larger than the size of pd.

This is a bit unusual but was done this way mostly to simplify error
handling and all the existing use cases could be handled this way;
however, this is becoming too restrictive now that percpu memory can
be allocated without blocking.

This introduces two new mandatory blkcg_policy methods - pd_alloc_fn()
and pd_free_fn() - which are used to allocate and release pd for a
given policy.  As pd allocation is now done from policy side, it can
simply allocate a larger area which embeds pd at the beginning.  This
change makes ->pd_size pointless.  Removed.

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

block/blk-cgroup.c

@@ -68,7 +68,8 @@ static void blkg_free(struct blkcg_gq *blkg)
 		return;
 
 	for (i = 0; i < BLKCG_MAX_POLS; i++)
-		kfree(blkg->pd[i]);
+		if (blkg->pd[i])
+			blkcg_policy[i]->pd_free_fn(blkg->pd[i]);
 
 	if (blkg->blkcg != &blkcg_root)
 		blk_exit_rl(&blkg->rl);
@@ -114,7 +115,7 @@ static struct blkcg_gq *blkg_alloc(struct blkcg *blkcg, struct request_queue *q,
 			continue;
 
 		/* alloc per-policy data and attach it to blkg */
-		pd = kzalloc_node(pol->pd_size, gfp_mask, q->node);
+		pd = pol->pd_alloc_fn(gfp_mask, q->node);
 		if (!pd)
 			goto err_free;
 
@@ -1057,7 +1058,7 @@ int blkcg_activate_policy(struct request_queue *q,
 	blk_queue_bypass_start(q);
 pd_prealloc:
 	if (!pd_prealloc) {
-		pd_prealloc = kzalloc_node(pol->pd_size, GFP_KERNEL, q->node);
+		pd_prealloc = pol->pd_alloc_fn(GFP_KERNEL, q->node);
 		if (!pd_prealloc) {
 			ret = -ENOMEM;
 			goto out_bypass_end;
@@ -1072,7 +1073,7 @@ int blkcg_activate_policy(struct request_queue *q,
 		if (blkg->pd[pol->plid])
 			continue;
 
-		pd = kzalloc_node(pol->pd_size, GFP_NOWAIT, q->node);
+		pd = pol->pd_alloc_fn(GFP_NOWAIT, q->node);
 		if (!pd)
 			swap(pd, pd_prealloc);
 		if (!pd) {
@@ -1093,7 +1094,8 @@ int blkcg_activate_policy(struct request_queue *q,
 	spin_unlock_irq(q->queue_lock);
 out_bypass_end:
 	blk_queue_bypass_end(q);
-	kfree(pd_prealloc);
+	if (pd_prealloc)
+		pol->pd_free_fn(pd_prealloc);
 	return ret;
 }
 EXPORT_SYMBOL_GPL(blkcg_activate_policy);
@@ -1128,8 +1130,10 @@ void blkcg_deactivate_policy(struct request_queue *q,
 		if (pol->pd_exit_fn)
 			pol->pd_exit_fn(blkg);
 
-		kfree(blkg->pd[pol->plid]);
-		blkg->pd[pol->plid] = NULL;
+		if (blkg->pd[pol->plid]) {
+			pol->pd_free_fn(blkg->pd[pol->plid]);
+			blkg->pd[pol->plid] = NULL;
+		}
 
 		spin_unlock(&blkg->blkcg->lock);
 	}
@@ -1151,9 +1155,6 @@ int blkcg_policy_register(struct blkcg_policy *pol)
 	struct blkcg *blkcg;
 	int i, ret;
 
-	if (WARN_ON(pol->pd_size < sizeof(struct blkg_policy_data)))
-		return -EINVAL;
-
 	mutex_lock(&blkcg_pol_register_mutex);
 	mutex_lock(&blkcg_pol_mutex);
 

block/blk-throttle.c

@@ -403,6 +403,11 @@ static void throtl_service_queue_exit(struct throtl_service_queue *sq)
 	del_timer_sync(&sq->pending_timer);
 }
 
+static struct blkg_policy_data *throtl_pd_alloc(gfp_t gfp, int node)
+{
+	return kzalloc_node(sizeof(struct throtl_grp), gfp, node);
+}
+
 static void throtl_pd_init(struct blkcg_gq *blkg)
 {
 	struct throtl_grp *tg = blkg_to_tg(blkg);
@@ -493,6 +498,11 @@ static void throtl_pd_exit(struct blkcg_gq *blkg)
 	throtl_service_queue_exit(&tg->service_queue);
 }
 
+static void throtl_pd_free(struct blkg_policy_data *pd)
+{
+	kfree(pd);
+}
+
 static void throtl_pd_reset_stats(struct blkcg_gq *blkg)
 {
 	struct throtl_grp *tg = blkg_to_tg(blkg);
@@ -1468,12 +1478,13 @@ static void throtl_shutdown_wq(struct request_queue *q)
 }
 
 static struct blkcg_policy blkcg_policy_throtl = {
-	.pd_size		= sizeof(struct throtl_grp),
 	.cftypes		= throtl_files,
 
+	.pd_alloc_fn		= throtl_pd_alloc,
 	.pd_init_fn		= throtl_pd_init,
 	.pd_online_fn		= throtl_pd_online,
 	.pd_exit_fn		= throtl_pd_exit,
+	.pd_free_fn		= throtl_pd_free,
 	.pd_reset_stats_fn	= throtl_pd_reset_stats,
 };
 

block/cfq-iosched.c

@@ -1582,6 +1582,11 @@ static void cfq_cpd_init(const struct blkcg *blkcg)
 	}
 }
 
+static struct blkg_policy_data *cfq_pd_alloc(gfp_t gfp, int node)
+{
+	return kzalloc_node(sizeof(struct cfq_group), gfp, node);
+}
+
 static void cfq_pd_init(struct blkcg_gq *blkg)
 {
 	struct cfq_group *cfqg = blkg_to_cfqg(blkg);
@@ -1618,6 +1623,11 @@ static void cfq_pd_offline(struct blkcg_gq *blkg)
 	cfqg_stats_xfer_dead(cfqg);
 }
 
+static void cfq_pd_free(struct blkg_policy_data *pd)
+{
+	return kfree(pd);
+}
+
 /* offset delta from cfqg->stats to cfqg->dead_stats */
 static const int dead_stats_off_delta = offsetof(struct cfq_group, dead_stats) -
 					offsetof(struct cfq_group, stats);
@@ -4633,13 +4643,14 @@ static struct elevator_type iosched_cfq = {
 
 #ifdef CONFIG_CFQ_GROUP_IOSCHED
 static struct blkcg_policy blkcg_policy_cfq = {
-	.pd_size		= sizeof(struct cfq_group),
 	.cpd_size		= sizeof(struct cfq_group_data),
 	.cftypes		= cfq_blkcg_files,
 
 	.cpd_init_fn		= cfq_cpd_init,
+	.pd_alloc_fn		= cfq_pd_alloc,
 	.pd_init_fn		= cfq_pd_init,
 	.pd_offline_fn		= cfq_pd_offline,
+	.pd_free_fn		= cfq_pd_free,
 	.pd_reset_stats_fn	= cfq_pd_reset_stats,
 };
 #endif

include/linux/blk-cgroup.h

@@ -68,13 +68,11 @@ struct blkg_rwstat {
  * request_queue (q).  This is used by blkcg policies which need to track
  * information per blkcg - q pair.
  *
- * There can be multiple active blkcg policies and each has its private
- * data on each blkg, the size of which is determined by
- * blkcg_policy->pd_size.  blkcg core allocates and frees such areas
- * together with blkg and invokes pd_init/exit_fn() methods.
- *
- * Such private data must embed struct blkg_policy_data (pd) at the
- * beginning and pd_size can't be smaller than pd.
+ * There can be multiple active blkcg policies and each blkg:policy pair is
+ * represented by a blkg_policy_data which is allocated and freed by each
+ * policy's pd_alloc/free_fn() methods.  A policy can allocate private data
+ * area by allocating larger data structure which embeds blkg_policy_data
+ * at the beginning.
  */
 struct blkg_policy_data {
 	/* the blkg and policy id this per-policy data belongs to */
@@ -126,16 +124,16 @@ struct blkcg_gq {
 };
 
 typedef void (blkcg_pol_init_cpd_fn)(const struct blkcg *blkcg);
+typedef struct blkg_policy_data *(blkcg_pol_alloc_pd_fn)(gfp_t gfp, int node);
 typedef void (blkcg_pol_init_pd_fn)(struct blkcg_gq *blkg);
 typedef void (blkcg_pol_online_pd_fn)(struct blkcg_gq *blkg);
 typedef void (blkcg_pol_offline_pd_fn)(struct blkcg_gq *blkg);
 typedef void (blkcg_pol_exit_pd_fn)(struct blkcg_gq *blkg);
+typedef void (blkcg_pol_free_pd_fn)(struct blkg_policy_data *pd);
 typedef void (blkcg_pol_reset_pd_stats_fn)(struct blkcg_gq *blkg);
 
 struct blkcg_policy {
 	int				plid;
-	/* policy specific private data size */
-	size_t				pd_size;
 	/* policy specific per-blkcg data size */
 	size_t				cpd_size;
 	/* cgroup files for the policy */
@@ -143,10 +141,12 @@ struct blkcg_policy {
 
 	/* operations */
 	blkcg_pol_init_cpd_fn		*cpd_init_fn;
+	blkcg_pol_alloc_pd_fn		*pd_alloc_fn;
 	blkcg_pol_init_pd_fn		*pd_init_fn;
 	blkcg_pol_online_pd_fn		*pd_online_fn;
 	blkcg_pol_offline_pd_fn		*pd_offline_fn;
 	blkcg_pol_exit_pd_fn		*pd_exit_fn;
+	blkcg_pol_free_pd_fn		*pd_free_fn;
 	blkcg_pol_reset_pd_stats_fn	*pd_reset_stats_fn;
 };