Re: [PATCH v3] block: assign caller-specific lockdep class to disk->open_mutex

[Tetsuo Handa]

On 2026/06/05 16:54, Andreas Hindborg wrote:
> Tetsuo Handa <penguin-kernel@xxxxxxxxxxxxxxxxxxx> writes:
> 
>> The block core currently allocates a single monolithic lockdep key for
>> disk->open_mutex across all callers. This single key conflates locking
>> hierarchies between independent block streams. For example, if a stacked
>> driver like loop flushes its internal workqueues inside lo_release() while
>> holding its own open_mutex, lockdep views this as a potential ABBA deadlock
>> against the underlying storage stack, leading to numerous circular
>> dependency splats.
>>
>> To structurally reduce false positives, this patch splits the global
>> monolithic lock class into distinct, per-caller instances during disk
>> allocation. This is done by replacing "struct lock_class_key" with
>> "struct gendisk_lkclass", which contains two instances of
>> "struct lock_class_key" for the legacy "(bio completion)" map and
>> disk->open_mutex respectively.
>>
>> Signed-off-by: Tetsuo Handa <penguin-kernel@xxxxxxxxxxxxxxxxxxx>
> 
> For the Rust part, we have existing infrastructure for lock class keys
> [1]. Please take a look how we generate lock class keys elsewhere [2].

My understanding is that we don't have infrastructure for lock class keys
that can be applied to

  +struct gendisk_lkclass {
  +	struct lock_class_key bio_lkclass;
  +	struct lock_class_key open_mutex_lkclass;
  +};

  -	static struct lock_class_key __key;
  +	static struct gendisk_lkclass __key;

change. Alternative approach is welcomed if you have one.

> 
> Best regards,
> Andreas Hindborg
> 
> [1] https://rust.docs.kernel.org/kernel/sync/struct.LockClassKey.html
> [2] https://github.com/torvalds/linux/blob/ddd664bbff63e09e7a7f9acae9c43605d4cf185f/rust/kernel/sync/lock/mutex.rs#L12
>