block: Reduce the amount of memory used for tag sets

Instead of allocating an array of size nr_cpu_ids for set->tags, allocate
an array of size set->nr_hw_queues. This patch improves behavior that was
introduced by commit 868f2f0b7206 ("blk-mq: dynamic h/w context count").

Reallocating tag sets from inside __blk_mq_update_nr_hw_queues() is safe
because:
- All request queues that share the tag sets are frozen before the tag sets
  are reallocated.
- blk_mq_queue_tag_busy_iter() holds q->q_usage_counter while active and
  hence is serialized against __blk_mq_update_nr_hw_queues().

Cc: Keith Busch <keith.busch@intel.com>
Cc: Christoph Hellwig <hch@infradead.org>
Cc: Ming Lei <ming.lei@redhat.com>
Cc: Hannes Reinecke <hare@suse.com>
Cc: Johannes Thumshirn <jthumshirn@suse.de>
Signed-off-by: Bart Van Assche <bvanassche@acm.org>
Signed-off-by: Jens Axboe <axboe@kernel.dk>

1 files changed, 30 insertions, 17 deletions

diff --git a/block/blk-mq.c b/block/blk-mq.c
index ba09cda49953..df41b2d16261 100644
--- a/block/blk-mq.c
+++ b/block/blk-mq.c
@@ -2833,19 +2833,6 @@ static void blk_mq_realloc_hw_ctxs(struct blk_mq_tag_set *set,
 	mutex_unlock(&q->sysfs_lock);
 }
 
-/*
- * Maximum number of hardware queues we support. For single sets, we'll never
- * have more than the CPUs (software queues). For multiple sets, the tag_set
- * user may have set ->nr_hw_queues larger.
- */
-static unsigned int nr_hw_queues(struct blk_mq_tag_set *set)
-{
-	if (set->nr_maps == 1)
-		return nr_cpu_ids;
-
-	return max(set->nr_hw_queues, nr_cpu_ids);
-}
-
 struct request_queue *blk_mq_init_allocated_queue(struct blk_mq_tag_set *set,
 						  struct request_queue *q,
 						  bool elevator_init)
@@ -3012,6 +2999,29 @@ static int blk_mq_update_queue_map(struct blk_mq_tag_set *set)
 	}
 }
 
+static int blk_mq_realloc_tag_set_tags(struct blk_mq_tag_set *set,
+				  int cur_nr_hw_queues, int new_nr_hw_queues)
+{
+	struct blk_mq_tags **new_tags;
+
+	if (cur_nr_hw_queues >= new_nr_hw_queues)
+		return 0;
+
+	new_tags = kcalloc_node(new_nr_hw_queues, sizeof(struct blk_mq_tags *),
+				GFP_KERNEL, set->numa_node);
+	if (!new_tags)
+		return -ENOMEM;
+
+	if (set->tags)
+		memcpy(new_tags, set->tags, cur_nr_hw_queues *
+		       sizeof(*set->tags));
+	kfree(set->tags);
+	set->tags = new_tags;
+	set->nr_hw_queues = new_nr_hw_queues;
+
+	return 0;
+}
+
 /*
  * Alloc a tag set to be associated with one or more request queues.
  * May fail with EINVAL for various error conditions. May adjust the
@@ -3065,9 +3075,7 @@ int blk_mq_alloc_tag_set(struct blk_mq_tag_set *set)
 	if (set->nr_maps == 1 && set->nr_hw_queues > nr_cpu_ids)
 		set->nr_hw_queues = nr_cpu_ids;
 
-	set->tags = kcalloc_node(nr_hw_queues(set), sizeof(struct blk_mq_tags *),
-				 GFP_KERNEL, set->numa_node);
-	if (!set->tags)
+	if (blk_mq_realloc_tag_set_tags(set, 0, set->nr_hw_queues) < 0)
 		return -ENOMEM;
 
 	ret = -ENOMEM;
@@ -3108,7 +3116,7 @@ void blk_mq_free_tag_set(struct blk_mq_tag_set *set)
 {
 	int i, j;
 
-	for (i = 0; i < nr_hw_queues(set); i++)
+	for (i = 0; i < set->nr_hw_queues; i++)
 		blk_mq_free_map_and_requests(set, i);
 
 	for (j = 0; j < set->nr_maps; j++) {
@@ -3266,6 +3274,10 @@ static void __blk_mq_update_nr_hw_queues(struct blk_mq_tag_set *set,
 		blk_mq_sysfs_unregister(q);
 	}
 
+	if (blk_mq_realloc_tag_set_tags(set, set->nr_hw_queues, nr_hw_queues) <
+	    0)
+		goto reregister;
+
 	prev_nr_hw_queues = set->nr_hw_queues;
 	set->nr_hw_queues = nr_hw_queues;
 	blk_mq_update_queue_map(set);
@@ -3282,6 +3294,7 @@ fallback:
 		blk_mq_map_swqueue(q);
 	}
 
+reregister:
 	list_for_each_entry(q, &set->tag_list, tag_set_list) {
 		blk_mq_sysfs_register(q);
 		blk_mq_debugfs_register_hctxs(q);