2021-04-15 03:57:56

by Lin Feng

Subject: [RFC PATCH 2/2] bfq/mq-deadline: remove redundant check for passthrough request

Since commit 01e99aeca39796003 'blk-mq: insert passthrough request into
hctx->dispatch directly', passthrough request should not appear in
IO-scheduler any more, so blk_rq_is_passthrough checking in addon IO
schedulers is redundant.

(Notes: this patch passes generic IO load test with hdds under SAS
controller and hdds under AHCI controller but obviously not covers all.
Not sure if passthrough request can still escape into IO scheduler from
blk_mq_sched_insert_requests, which is used by blk_mq_flush_plug_list and
has lots of indirect callers.)

Signed-off-by: Lin Feng <[email protected]>
---
block/bfq-iosched.c | 2 +-
block/mq-deadline.c | 7 ++-----
2 files changed, 3 insertions(+), 6 deletions(-)

diff --git a/block/bfq-iosched.c b/block/bfq-iosched.c
index 95586137194e..b827c9212b02 100644
--- a/block/bfq-iosched.c
+++ b/block/bfq-iosched.c
@@ -5627,7 +5627,7 @@ static void bfq_insert_request(struct blk_mq_hw_ctx *hctx, struct request *rq,

spin_lock_irq(&bfqd->lock);
bfqq = bfq_init_rq(rq);
- if (!bfqq || at_head || blk_rq_is_passthrough(rq)) {
+ if (!bfqq || at_head) {
if (at_head)
list_add(&rq->queuelist, &bfqd->dispatch);
else
diff --git a/block/mq-deadline.c b/block/mq-deadline.c
index f3631a287466..04aded71ead2 100644
--- a/block/mq-deadline.c
+++ b/block/mq-deadline.c
@@ -500,11 +500,8 @@ static void dd_insert_request(struct blk_mq_hw_ctx *hctx, struct request *rq,

trace_block_rq_insert(rq);

- if (at_head || blk_rq_is_passthrough(rq)) {
- if (at_head)
- list_add(&rq->queuelist, &dd->dispatch);
- else
- list_add_tail(&rq->queuelist, &dd->dispatch);
+ if (at_head) {
+ list_add(&rq->queuelist, &dd->dispatch);
} else {
deadline_add_rq_rb(dd, rq);

--
2.30.2

2021-04-16 02:52:37

by Ming Lei

Subject: Re: [RFC PATCH 2/2] bfq/mq-deadline: remove redundant check for passthrough request

On Thu, Apr 15, 2021 at 11:43:26AM +0800, Lin Feng wrote:
> Since commit 01e99aeca39796003 'blk-mq: insert passthrough request into
> hctx->dispatch directly', passthrough request should not appear in
> IO-scheduler any more, so blk_rq_is_passthrough checking in addon IO
> schedulers is redundant.
>
> (Notes: this patch passes generic IO load test with hdds under SAS
> controller and hdds under AHCI controller but obviously not covers all.
> Not sure if passthrough request can still escape into IO scheduler from
> blk_mq_sched_insert_requests, which is used by blk_mq_flush_plug_list and
> has lots of indirect callers.)
>
> Signed-off-by: Lin Feng <[email protected]>
> ---
> block/bfq-iosched.c | 2 +-
> block/mq-deadline.c | 7 ++-----
> 2 files changed, 3 insertions(+), 6 deletions(-)
>
> diff --git a/block/bfq-iosched.c b/block/bfq-iosched.c
> index 95586137194e..b827c9212b02 100644
> --- a/block/bfq-iosched.c
> +++ b/block/bfq-iosched.c
> @@ -5627,7 +5627,7 @@ static void bfq_insert_request(struct blk_mq_hw_ctx *hctx, struct request *rq,
>
> spin_lock_irq(&bfqd->lock);
> bfqq = bfq_init_rq(rq);
> - if (!bfqq || at_head || blk_rq_is_passthrough(rq)) {
> + if (!bfqq || at_head) {
> if (at_head)
> list_add(&rq->queuelist, &bfqd->dispatch);
> else
> diff --git a/block/mq-deadline.c b/block/mq-deadline.c
> index f3631a287466..04aded71ead2 100644
> --- a/block/mq-deadline.c
> +++ b/block/mq-deadline.c
> @@ -500,11 +500,8 @@ static void dd_insert_request(struct blk_mq_hw_ctx *hctx, struct request *rq,
>
> trace_block_rq_insert(rq);
>
> - if (at_head || blk_rq_is_passthrough(rq)) {
> - if (at_head)
> - list_add(&rq->queuelist, &dd->dispatch);
> - else
> - list_add_tail(&rq->queuelist, &dd->dispatch);
> + if (at_head) {
> + list_add(&rq->queuelist, &dd->dispatch);
> } else {
> deadline_add_rq_rb(dd, rq);
>
> --
> 2.30.2
>

Looks fine:

Reviewed-by: Ming Lei <[email protected]>

Thanks,
Ming

2021-04-16 12:48:26

by Jens Axboe

Subject: Re: [RFC PATCH 2/2] bfq/mq-deadline: remove redundant check for passthrough request

On 4/14/21 9:43 PM, Lin Feng wrote:
> Since commit 01e99aeca39796003 'blk-mq: insert passthrough request into
> hctx->dispatch directly', passthrough request should not appear in
> IO-scheduler any more, so blk_rq_is_passthrough checking in addon IO
> schedulers is redundant.
>
> (Notes: this patch passes generic IO load test with hdds under SAS
> controller and hdds under AHCI controller but obviously not covers all.
> Not sure if passthrough request can still escape into IO scheduler from
> blk_mq_sched_insert_requests, which is used by blk_mq_flush_plug_list and
> has lots of indirect callers.)

Applied, with the bfq bits hand edited to apply for 5.13.

--
Jens Axboe