[PATCH] block: use trylock to avoid lockdep circular dependency in sysfs

[PATCH] block: use trylock to avoid lockdep circular dependency in sysfs

[Ming Lei]

Use trylock instead of blocking lock acquisition for update_nr_hwq_lock
in queue_requests_store() and elv_iosched_store() to avoid circular lock
dependency with kernfs active reference during concurrent disk deletion:

  update_nr_hwq_lock -> kn->active (via del_gendisk -> kobject_del)
  kn->active -> update_nr_hwq_lock (via sysfs write path)

Return -EBUSY when the lock is not immediately available.

Reported-and-tested-by: Yi Zhang <yi.zhang@redhat.com>
Closes: https://lore.kernel.org/linux-block/CAHj4cs-em-4acsHabMdT=jJhXkCzjnprD-aQH1OgrZo4nTnmMw@mail.gmail.com/
Fixes: 626ff4f8ebcb ("blk-mq: convert to serialize updating nr_requests with update_nr_hwq_lock")
Signed-off-by: Ming Lei <ming.lei@redhat.com>
---
 block/blk-sysfs.c |  8 +++++++-
 block/elevator.c  | 12 +++++++++++-
 2 files changed, 18 insertions(+), 2 deletions(-)

diff --git a/block/blk-sysfs.c b/block/blk-sysfs.c
index f3b1968c80ce..55a1bbfef7d4 100644
--- a/block/blk-sysfs.c
+++ b/block/blk-sysfs.c
@@ -78,8 +78,14 @@ queue_requests_store(struct gendisk *disk, const char *page, size_t count)
 	/*
 	 * Serialize updating nr_requests with concurrent queue_requests_store()
 	 * and switching elevator.
+	 *
+	 * Use trylock to avoid circular lock dependency with kernfs active
+	 * reference during concurrent disk deletion:
+	 *   update_nr_hwq_lock -> kn->active (via del_gendisk -> kobject_del)
+	 *   kn->active -> update_nr_hwq_lock (via this sysfs write path)
 	 */
-	down_write(&set->update_nr_hwq_lock);
+	if (!down_write_trylock(&set->update_nr_hwq_lock))
+		return -EBUSY;
 
 	if (nr == q->nr_requests)
 		goto unlock;
diff --git a/block/elevator.c b/block/elevator.c
index ebe2a1fcf011..3bcd37c2aa34 100644
--- a/block/elevator.c
+++ b/block/elevator.c
@@ -807,7 +807,16 @@ ssize_t elv_iosched_store(struct gendisk *disk, const char *buf,
 	elv_iosched_load_module(ctx.name);
 	ctx.type = elevator_find_get(ctx.name);
 
-	down_read(&set->update_nr_hwq_lock);
+	/*
+	 * Use trylock to avoid circular lock dependency with kernfs active
+	 * reference during concurrent disk deletion:
+	 *   update_nr_hwq_lock -> kn->active (via del_gendisk -> kobject_del)
+	 *   kn->active -> update_nr_hwq_lock (via this sysfs write path)
+	 */
+	if (!down_read_trylock(&set->update_nr_hwq_lock)) {
+		ret = -EBUSY;
+		goto out;
+	}
 	if (!blk_queue_no_elv_switch(q)) {
 		ret = elevator_change(q, &ctx);
 		if (!ret)
@@ -817,6 +826,7 @@ ssize_t elv_iosched_store(struct gendisk *disk, const char *buf,
 	}
 	up_read(&set->update_nr_hwq_lock);
 
+out:
 	if (ctx.type)
 		elevator_put(ctx.type);
 	return ret;
-- 
2.47.0

Re: [PATCH] block: use trylock to avoid lockdep circular dependency in sysfs

[Jens Axboe]

On Thu, 05 Mar 2026 11:15:50 +0800, Ming Lei wrote:
> Use trylock instead of blocking lock acquisition for update_nr_hwq_lock
> in queue_requests_store() and elv_iosched_store() to avoid circular lock
> dependency with kernfs active reference during concurrent disk deletion:
> 
>   update_nr_hwq_lock -> kn->active (via del_gendisk -> kobject_del)
>   kn->active -> update_nr_hwq_lock (via sysfs write path)
> 
> [...]

Applied, thanks!

[1/1] block: use trylock to avoid lockdep circular dependency in sysfs
      commit: ce8ee8583ed83122405eabaa8fb351be4d9dc65c

Best regards,
-- 
Jens Axboe

Re: [PATCH] block: use trylock to avoid lockdep circular dependency in sysfs

[Jens Axboe]

On 3/4/26 8:15 PM, Ming Lei wrote:
> Use trylock instead of blocking lock acquisition for update_nr_hwq_lock
> in queue_requests_store() and elv_iosched_store() to avoid circular lock
> dependency with kernfs active reference during concurrent disk deletion:
> 
>   update_nr_hwq_lock -> kn->active (via del_gendisk -> kobject_del)
>   kn->active -> update_nr_hwq_lock (via sysfs write path)
> 
> Return -EBUSY when the lock is not immediately available.

This does mean that scripts can get spurious -EBUSY they now need to
handle. Which probably isn't a big real world problem. While it isn't a
hard thing to fix up in a script, it can potentially break existing
setups.

Not sure I see a good way around that however, without queueing these
to a clean context and having the store helpers wait on eg work item
being processed and completed.

Hopefully this will never be a thing and we'll never need to discuss
this issue, I'll get this one applied for now.

-- 
Jens Axboe