block: Reduce the amount of memory required per request queue

Instead of always allocating at least nr_cpu_ids hardware queues per request
queue, reallocate q->queue_hw_ctx if it has to grow. This patch improves
behavior that was introduced by commit 868f2f0b ("blk-mq: dynamic h/w
context count").

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>

block/blk-mq.c

@@ -2761,6 +2761,23 @@ static void blk_mq_realloc_hw_ctxs(struct blk_mq_tag_set *set,
 	int i, j, end;
 	struct blk_mq_hw_ctx **hctxs = q->queue_hw_ctx;
 
+	if (q->nr_hw_queues < set->nr_hw_queues) {
+		struct blk_mq_hw_ctx **new_hctxs;
+
+		new_hctxs = kcalloc_node(set->nr_hw_queues,
+				       sizeof(*new_hctxs), GFP_KERNEL,
+				       set->numa_node);
+		if (!new_hctxs)
+			return;
+		if (hctxs)
+			memcpy(new_hctxs, hctxs, q->nr_hw_queues *
+			       sizeof(*hctxs));
+		q->queue_hw_ctx = new_hctxs;
+		q->nr_hw_queues = set->nr_hw_queues;
+		kfree(hctxs);
+		hctxs = new_hctxs;
+	}
+
 	/* protect against switching io scheduler  */
 	mutex_lock(&q->sysfs_lock);
 	for (i = 0; i < set->nr_hw_queues; i++) {
@@ -2848,12 +2865,6 @@ struct request_queue *blk_mq_init_allocated_queue(struct blk_mq_tag_set *set,
 	/* init q->mq_kobj and sw queues' kobjects */
 	blk_mq_sysfs_init(q);
 
-	q->queue_hw_ctx = kcalloc_node(nr_hw_queues(set),
-				       sizeof(*(q->queue_hw_ctx)), GFP_KERNEL,
-				       set->numa_node);
-	if (!q->queue_hw_ctx)
-		goto err_sys_init;
-
 	INIT_LIST_HEAD(&q->unused_hctx_list);
 	spin_lock_init(&q->unused_hctx_lock);
 
@@ -2901,7 +2912,6 @@ struct request_queue *blk_mq_init_allocated_queue(struct blk_mq_tag_set *set,
 err_hctxs:
 	kfree(q->queue_hw_ctx);
 	q->nr_hw_queues = 0;
-err_sys_init:
 	blk_mq_sysfs_deinit(q);
 err_poll:
 	blk_stat_free_callback(q->poll_cb);