On 4/30/25 10:05 AM, Ming Lei wrote:
> Move blk_queue_registered() check into elv_iosched_store() and prepare
> for using elevator_change() for covering any kind of elevator change in
> adding/deleting disk and updating nr_hw_queue.
>
> Signed-off-by: Ming Lei <ming.lei@xxxxxxxxxx>
> ---
> block/elevator.c | 8 ++++----
> 1 file changed, 4 insertions(+), 4 deletions(-)
>
> diff --git a/block/elevator.c b/block/elevator.c
> index df3e59107a2a..ac72c4f5a542 100644
> --- a/block/elevator.c
> +++ b/block/elevator.c
> @@ -676,10 +676,6 @@ int elevator_switch(struct request_queue *q, const char *name)
> */
> static int elevator_change(struct request_queue *q, const char *elevator_name)
> {
> - /* Make sure queue is not in the middle of being removed */
> - if (!blk_queue_registered(q))
> - return -ENOENT;
> -
> if (q->elevator && elevator_match(q->elevator->type, elevator_name))
> return 0;
>
> @@ -708,6 +704,10 @@ ssize_t elv_iosched_store(struct gendisk *disk, const char *buf,
> struct request_queue *q = disk->queue;
> struct blk_mq_tag_set *set = q->tag_set;
>
> + /* Make sure queue is not in the middle of being removed */
> + if (!blk_queue_registered(q))
> + return -ENOENT;
> +
> /*
> * If the attribute needs to load a module, do it before freezing the
> * queue to ensure that the module file can be read when the request
Shouldn't blk_queue_registered needs protection? For instance, I see here a race
while queue is being removed from del_gendisk and at the same time elevator is
being updated from elv_iosched_store. BTW, I saw that your patch 19/24 adds that
protection using disable_elv_switch before deleting disk (and so you may want to
fold this patch 11/24 into patch 19/24 ?) but with that also I foresee a thin race
window. I will comment on it while reviewing the patch 19/24.
Thanks,
--Nilay
