Re: [PATCH 3/4] blk-mq: check bio_mergeable() early before merging

[Jeff Moyer]

Ming Lei <ming.lei@xxxxxxxxxxxxx> writes:

> It isn't necessary to try to merge the bio which is marked
> as NOMERGE.
>
> Signed-off-by: Ming Lei <ming.lei@xxxxxxxxxxxxx>
> ---
>  block/blk-mq.c | 5 ++++-
>  1 file changed, 4 insertions(+), 1 deletion(-)
>
> diff --git a/block/blk-mq.c b/block/blk-mq.c
> index 546b3b8..deb5f4c 100644
> --- a/block/blk-mq.c
> +++ b/block/blk-mq.c
> @@ -671,6 +671,9 @@ static bool blk_mq_attempt_merge(struct request_queue *q,
>  	struct request *rq;
>  	int checked = 8;
>  
> +	if (!bio_mergeable(bio))
> +		return false;
> +
>  	list_for_each_entry_reverse(rq, &ctx->rq_list, queuelist) {
>  		int el_ret;
>  
> @@ -1140,7 +1143,7 @@ static inline bool blk_mq_merge_queue_io(struct blk_mq_hw_ctx *hctx,
>  					 struct blk_mq_ctx *ctx,
>  					 struct request *rq, struct bio *bio)
>  {
> -	if (!hctx_allow_merges(hctx)) {
> +	if (!hctx_allow_merges(hctx) || !bio_mergeable(bio)) {
>  		blk_mq_bio_to_request(rq, bio);
>  		spin_lock(&ctx->lock);
>  insert_rq:

blk_mq_attempt_merge is only called from blk_mq_merge_queue_io.  So, by
adding the conditional in blk_mq_merge_queue_io, you don't need any
change in blk_mq_attempt_merge.

Also, why haven't you updated the non-multiqueue code paths similarly?

Cheers,
Jeff
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/