On Fri, 2017-04-28 at 08:55 -0600, Jens Axboe wrote:
> We only have an internal helper for checking a tag value. Add
> an exported helper that takes a request and hardware queue,
> and check against the driver tags.
>
> Signed-off-by: Jens Axboe <axboe@xxxxxx>
> ---
> block/blk-mq-tag.c | 6 ++++++
> include/linux/blk-mq.h | 1 +
> 2 files changed, 7 insertions(+)
>
> diff --git a/block/blk-mq-tag.c b/block/blk-mq-tag.c
> index d0be72ccb091..0a04b5c14076 100644
> --- a/block/blk-mq-tag.c
> +++ b/block/blk-mq-tag.c
> @@ -476,3 +476,9 @@ u32 blk_mq_unique_tag(struct request *rq)
> (rq->tag & BLK_MQ_UNIQUE_TAG_MASK);
> }
> EXPORT_SYMBOL(blk_mq_unique_tag);
> +
> +bool blk_mq_rq_is_reserved(struct blk_mq_hw_ctx *hctx, struct request *rq)
> +{
> + return blk_mq_tag_is_reserved(hctx->tags, rq->tag);
> +}
> +EXPORT_SYMBOL_GPL(blk_mq_rq_is_reserved);
> diff --git a/include/linux/blk-mq.h b/include/linux/blk-mq.h
> index 32bd8eb5ba67..13e51520fa27 100644
> --- a/include/linux/blk-mq.h
> +++ b/include/linux/blk-mq.h
> @@ -224,6 +224,7 @@ static inline u16 blk_mq_unique_tag_to_tag(u32 unique_tag)
> return unique_tag & BLK_MQ_UNIQUE_TAG_MASK;
> }
>
> +bool blk_mq_rq_is_reserved(struct blk_mq_hw_ctx *hctx, struct request *rq);
>
> int blk_mq_request_started(struct request *rq);
> void blk_mq_start_request(struct request *rq);
Hello Jens,
Have you considered to derive the hctx information from rq->mq_ctx instead of
passing hctx as an argument (hctx = blk_mq_map_queue(rq->q, rq->mq_ctx->cpu))?
Anyway, since otherwise this patch looks fine to me:
Reviewed-by: Bart Van Assche <Bart.VanAssche@xxxxxxxxxxx>
