block: serialize all debugfs operations using q->debugfs_mutex

Various places like I/O schedulers or the QOS infrastructure try to
register debugfs files on demans, which can race with creating and
removing the main queue debugfs directory.  Use the existing
debugfs_mutex to serialize all debugfs operations that rely on
q->debugfs_dir or the directories hanging off it.

To make the teardown code a little simpler declare all debugfs dentry
pointers and not just the main one uncoditionally in blkdev.h.

Move debugfs_mutex next to the dentries that it protects and document
what it is used for.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Link: https://lore.kernel.org/r/20220614074827.458955-3-hch@lst.deSigned-off-by: Jens Axboe <axboe@kernel.dk>

block/blk-mq-debugfs.c

@@ -711,11 +711,6 @@ void blk_mq_debugfs_register(struct request_queue *q)
 	}
 }
 
-void blk_mq_debugfs_unregister(struct request_queue *q)
-{
-	q->sched_debugfs_dir = NULL;
-}
-
 static void blk_mq_debugfs_register_ctx(struct blk_mq_hw_ctx *hctx,
 					struct blk_mq_ctx *ctx)
 {
@@ -746,6 +741,8 @@ void blk_mq_debugfs_register_hctx(struct request_queue *q,
 
 void blk_mq_debugfs_unregister_hctx(struct blk_mq_hw_ctx *hctx)
 {
+	if (!hctx->queue->debugfs_dir)
+		return;
 	debugfs_remove_recursive(hctx->debugfs_dir);
 	hctx->sched_debugfs_dir = NULL;
 	hctx->debugfs_dir = NULL;
@@ -773,6 +770,8 @@ void blk_mq_debugfs_register_sched(struct request_queue *q)
 {
 	struct elevator_type *e = q->elevator->type;
 
+	lockdep_assert_held(&q->debugfs_mutex);
+
 	/*
 	 * If the parent directory has not been created yet, return, we will be
 	 * called again later on and the directory/files will be created then.
@@ -790,6 +789,8 @@ void blk_mq_debugfs_register_sched(struct request_queue *q)
 
 void blk_mq_debugfs_unregister_sched(struct request_queue *q)
 {
+	lockdep_assert_held(&q->debugfs_mutex);
+
 	debugfs_remove_recursive(q->sched_debugfs_dir);
 	q->sched_debugfs_dir = NULL;
 }
@@ -811,6 +812,10 @@ static const char *rq_qos_id_to_name(enum rq_qos_id id)
 
 void blk_mq_debugfs_unregister_rqos(struct rq_qos *rqos)
 {
+	lockdep_assert_held(&rqos->q->debugfs_mutex);
+
+	if (!rqos->q->debugfs_dir)
+		return;
 	debugfs_remove_recursive(rqos->debugfs_dir);
 	rqos->debugfs_dir = NULL;
 }
@@ -820,6 +825,8 @@ void blk_mq_debugfs_register_rqos(struct rq_qos *rqos)
 	struct request_queue *q = rqos->q;
 	const char *dir_name = rq_qos_id_to_name(rqos->id);
 
+	lockdep_assert_held(&q->debugfs_mutex);
+
 	if (rqos->debugfs_dir || !rqos->ops->debugfs_attrs)
 		return;
 
@@ -835,6 +842,8 @@ void blk_mq_debugfs_register_rqos(struct rq_qos *rqos)
 
 void blk_mq_debugfs_unregister_queue_rqos(struct request_queue *q)
 {
+	lockdep_assert_held(&q->debugfs_mutex);
+
 	debugfs_remove_recursive(q->rqos_debugfs_dir);
 	q->rqos_debugfs_dir = NULL;
 }
@@ -844,6 +853,8 @@ void blk_mq_debugfs_register_sched_hctx(struct request_queue *q,
 {
 	struct elevator_type *e = q->elevator->type;
 
+	lockdep_assert_held(&q->debugfs_mutex);
+
 	/*
 	 * If the parent debugfs directory has not been created yet, return;
 	 * We will be called again later on with appropriate parent debugfs
@@ -863,6 +874,10 @@ void blk_mq_debugfs_register_sched_hctx(struct request_queue *q,
 
 void blk_mq_debugfs_unregister_sched_hctx(struct blk_mq_hw_ctx *hctx)
 {
+	lockdep_assert_held(&hctx->queue->debugfs_mutex);
+
+	if (!hctx->queue->debugfs_dir)
+		return;
 	debugfs_remove_recursive(hctx->sched_debugfs_dir);
 	hctx->sched_debugfs_dir = NULL;
 }

block/blk-mq-debugfs.h

@@ -21,7 +21,6 @@ int __blk_mq_debugfs_rq_show(struct seq_file *m, struct request *rq);
 int blk_mq_debugfs_rq_show(struct seq_file *m, void *v);
 
 void blk_mq_debugfs_register(struct request_queue *q);
-void blk_mq_debugfs_unregister(struct request_queue *q);
 void blk_mq_debugfs_register_hctx(struct request_queue *q,
 				  struct blk_mq_hw_ctx *hctx);
 void blk_mq_debugfs_unregister_hctx(struct blk_mq_hw_ctx *hctx);
@@ -42,10 +41,6 @@ static inline void blk_mq_debugfs_register(struct request_queue *q)
 {
 }
 
-static inline void blk_mq_debugfs_unregister(struct request_queue *q)
-{
-}
-
 static inline void blk_mq_debugfs_register_hctx(struct request_queue *q,
 						struct blk_mq_hw_ctx *hctx)
 {

block/blk-mq-sched.c

@@ -594,7 +594,9 @@ int blk_mq_init_sched(struct request_queue *q, struct elevator_type *e)
 	if (ret)
 		goto err_free_map_and_rqs;
 
+	mutex_lock(&q->debugfs_mutex);
 	blk_mq_debugfs_register_sched(q);
+	mutex_unlock(&q->debugfs_mutex);
 
 	queue_for_each_hw_ctx(q, hctx, i) {
 		if (e->ops.init_hctx) {
@@ -607,7 +609,9 @@ int blk_mq_init_sched(struct request_queue *q, struct elevator_type *e)
 				return ret;
 			}
 		}
+		mutex_lock(&q->debugfs_mutex);
 		blk_mq_debugfs_register_sched_hctx(q, hctx);
+		mutex_unlock(&q->debugfs_mutex);
 	}
 
 	return 0;
@@ -648,14 +652,21 @@ void blk_mq_exit_sched(struct request_queue *q, struct elevator_queue *e)
 	unsigned int flags = 0;
 
 	queue_for_each_hw_ctx(q, hctx, i) {
+		mutex_lock(&q->debugfs_mutex);
 		blk_mq_debugfs_unregister_sched_hctx(hctx);
+		mutex_unlock(&q->debugfs_mutex);
+
 		if (e->type->ops.exit_hctx && hctx->sched_data) {
 			e->type->ops.exit_hctx(hctx, i);
 			hctx->sched_data = NULL;
 		}
 		flags = hctx->flags;
 	}
+
+	mutex_lock(&q->debugfs_mutex);
 	blk_mq_debugfs_unregister_sched(q);
+	mutex_unlock(&q->debugfs_mutex);
+
 	if (e->type->ops.exit_sched)
 		e->type->ops.exit_sched(e);
 	blk_mq_sched_tags_teardown(q, flags);

block/blk-rq-qos.c

@@ -294,7 +294,9 @@ void rq_qos_wait(struct rq_wait *rqw, void *private_data,
 
 void rq_qos_exit(struct request_queue *q)
 {
+	mutex_lock(&q->debugfs_mutex);
 	blk_mq_debugfs_unregister_queue_rqos(q);
+	mutex_unlock(&q->debugfs_mutex);
 
 	while (q->rq_qos) {
 		struct rq_qos *rqos = q->rq_qos;

block/blk-rq-qos.h

@@ -104,8 +104,11 @@ static inline void rq_qos_add(struct request_queue *q, struct rq_qos *rqos)
 
 	blk_mq_unfreeze_queue(q);
 
-	if (rqos->ops->debugfs_attrs)
+	if (rqos->ops->debugfs_attrs) {
+		mutex_lock(&q->debugfs_mutex);
 		blk_mq_debugfs_register_rqos(rqos);
+		mutex_unlock(&q->debugfs_mutex);
+	}
 }
 
 static inline void rq_qos_del(struct request_queue *q, struct rq_qos *rqos)
@@ -129,7 +132,9 @@ static inline void rq_qos_del(struct request_queue *q, struct rq_qos *rqos)
 
 	blk_mq_unfreeze_queue(q);
 
+	mutex_lock(&q->debugfs_mutex);
 	blk_mq_debugfs_unregister_rqos(rqos);
+	mutex_unlock(&q->debugfs_mutex);
 }
 
 typedef bool (acquire_inflight_cb_t)(struct rq_wait *rqw, void *private_data);

block/blk-sysfs.c

@@ -779,14 +779,13 @@ static void blk_release_queue(struct kobject *kobj)
 	if (queue_is_mq(q))
 		blk_mq_release(q);
 
-	blk_trace_shutdown(q);
 	mutex_lock(&q->debugfs_mutex);
+	blk_trace_shutdown(q);
 	debugfs_remove_recursive(q->debugfs_dir);
+	q->debugfs_dir = NULL;
+	q->sched_debugfs_dir = NULL;
 	mutex_unlock(&q->debugfs_mutex);
 
-	if (queue_is_mq(q))
-		blk_mq_debugfs_unregister(q);
-
 	bioset_exit(&q->bio_split);
 
 	if (blk_queue_has_srcu(q))
@@ -836,17 +835,16 @@ int blk_register_queue(struct gendisk *disk)
 		goto unlock;
 	}
 
+	if (queue_is_mq(q))
+		__blk_mq_register_dev(dev, q);
+	mutex_lock(&q->sysfs_lock);
+
 	mutex_lock(&q->debugfs_mutex);
 	q->debugfs_dir = debugfs_create_dir(kobject_name(q->kobj.parent),
 					    blk_debugfs_root);
-	mutex_unlock(&q->debugfs_mutex);
-
-	if (queue_is_mq(q)) {
-		__blk_mq_register_dev(dev, q);
+	if (queue_is_mq(q))
 		blk_mq_debugfs_register(q);
-	}
-
-	mutex_lock(&q->sysfs_lock);
+	mutex_unlock(&q->debugfs_mutex);
 
 	ret = disk_register_independent_access_ranges(disk, NULL);
 	if (ret)

include/linux/blkdev.h

@@ -482,7 +482,6 @@ struct request_queue {
 #endif /* CONFIG_BLK_DEV_ZONED */
 
 	int			node;
-	struct mutex		debugfs_mutex;
 #ifdef CONFIG_BLK_DEV_IO_TRACE
 	struct blk_trace __rcu	*blk_trace;
 #endif
@@ -526,11 +525,12 @@ struct request_queue {
 	struct bio_set		bio_split;
 
 	struct dentry		*debugfs_dir;
-
-#ifdef CONFIG_BLK_DEBUG_FS
 	struct dentry		*sched_debugfs_dir;
 	struct dentry		*rqos_debugfs_dir;
-#endif
+	/*
+	 * Serializes all debugfs metadata operations using the above dentries.
+	 */
+	struct mutex		debugfs_mutex;
 
 	bool			mq_sysfs_init_done;
 

kernel/trace/blktrace.c

@@ -770,14 +770,11 @@ int blk_trace_ioctl(struct block_device *bdev, unsigned cmd, char __user *arg)
  **/
 void blk_trace_shutdown(struct request_queue *q)
 {
-	mutex_lock(&q->debugfs_mutex);
 	if (rcu_dereference_protected(q->blk_trace,
 				      lockdep_is_held(&q->debugfs_mutex))) {
 		__blk_trace_startstop(q, 0);
 		__blk_trace_remove(q);
 	}
-
-	mutex_unlock(&q->debugfs_mutex);
 }
 
 #ifdef CONFIG_BLK_CGROUP