block: fix module reference leak in mq-deadline I/O scheduler

During probe, when the block layer registers a request queue, it
defaults to the mq-deadline I/O scheduler if the device is single-queue
and the mq-deadline module is available. To determine availability, the
elevator_set_default() invokes elevator_find_get(), which increments the
module's reference count. However, this reference is never released,
resulting in a module reference leak that prevents the mq-deadline module
from being unloaded.

This patch fixes the issue by ensuring the acquired module reference is
properly released.

Fixes: 1e44bed ("block: unifying elevator change")
Signed-off-by: Nilay Shroff <nilay@linux.ibm.com>
Reviewed-by: Ming Lei <ming.lei@redhat.com>
Link: https://lore.kernel.org/r/20250719132722.769536-1-nilay@linux.ibm.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>

Author: shroffni

block/elevator.c

@@ -719,7 +719,8 @@ void elevator_set_default(struct request_queue *q)
 		.name = "mq-deadline",
 		.no_uevent = true,
 	};
-	int err = 0;
+	int err;
+	struct elevator_type *e;
 
 	/* now we allow to switch elevator */
 	blk_queue_flag_clear(QUEUE_FLAG_NO_ELV_SWITCH, q);
@@ -732,12 +733,18 @@ void elevator_set_default(struct request_queue *q)
 	 * have multiple queues or mq-deadline is not available, default
 	 * to "none".
 	 */
-	if (elevator_find_get(ctx.name) && (q->nr_hw_queues == 1 ||
-			 blk_mq_is_shared_tags(q->tag_set->flags)))
+	e = elevator_find_get(ctx.name);
+	if (!e)
+		return;
+
+	if ((q->nr_hw_queues == 1 ||
+			blk_mq_is_shared_tags(q->tag_set->flags))) {
 		err = elevator_change(q, &ctx);
-	if (err < 0)
-		pr_warn("\"%s\" elevator initialization, failed %d, "
-			"falling back to \"none\"\n", ctx.name, err);
+		if (err < 0)
+			pr_warn("\"%s\" elevator initialization, failed %d, falling back to \"none\"\n",
+					ctx.name, err);
+	}
+	elevator_put(e);
 }
 
 void elevator_set_none(struct request_queue *q)