| Message ID | 20231018175602.2148415-4-bvanassche@acm.org |
|---|---|
| State | New |
| Headers | show |
| Series | Improve write performance for zoned UFS devices | expand |
On 10/19/23 02:54, Bart Van Assche wrote: > blk_mq_requeue_work() inserts requeued requests in front of other > requests. This is fine for all request types except for sequential zoned > writes. Hence this patch. > > Note: moving this functionality into the mq-deadline I/O scheduler is > not an option because we want to be able to use zoned storage without > I/O scheduler. > > Cc: Christoph Hellwig <hch@lst.de> > Cc: Damien Le Moal <dlemoal@kernel.org> > Cc: Ming Lei <ming.lei@redhat.com> > Cc: Hannes Reinecke <hare@suse.de> > Signed-off-by: Bart Van Assche <bvanassche@acm.org> > --- > block/blk-mq.c | 4 +++- > 1 file changed, 3 insertions(+), 1 deletion(-) > > diff --git a/block/blk-mq.c b/block/blk-mq.c > index 502dafa76716..ce6ddb249959 100644 > --- a/block/blk-mq.c > +++ b/block/blk-mq.c > @@ -1485,7 +1485,9 @@ static void blk_mq_requeue_work(struct work_struct *work) > blk_mq_request_bypass_insert(rq, 0); > } else { > list_del_init(&rq->queuelist); > - blk_mq_insert_request(rq, BLK_MQ_INSERT_AT_HEAD); > + blk_mq_insert_request(rq, > + !blk_rq_is_seq_zoned_write(rq) ? > + BLK_MQ_INSERT_AT_HEAD : 0); Something like: } else { blk_insert_t flags = BLK_MQ_INSERT_AT_HEAD; if (blk_rq_is_seq_zoned_write(rq)) flags = 0; blk_mq_insert_request(rq, flags); } would be a lot easier to read in my opinion.
On 10/18/23 17:15, Damien Le Moal wrote: > On 10/19/23 02:54, Bart Van Assche wrote: >> blk_mq_requeue_work() inserts requeued requests in front of other >> requests. This is fine for all request types except for sequential zoned >> writes. Hence this patch. >> >> Note: moving this functionality into the mq-deadline I/O scheduler is >> not an option because we want to be able to use zoned storage without >> I/O scheduler. >> >> Cc: Christoph Hellwig <hch@lst.de> >> Cc: Damien Le Moal <dlemoal@kernel.org> >> Cc: Ming Lei <ming.lei@redhat.com> >> Cc: Hannes Reinecke <hare@suse.de> >> Signed-off-by: Bart Van Assche <bvanassche@acm.org> >> --- >> block/blk-mq.c | 4 +++- >> 1 file changed, 3 insertions(+), 1 deletion(-) >> >> diff --git a/block/blk-mq.c b/block/blk-mq.c >> index 502dafa76716..ce6ddb249959 100644 >> --- a/block/blk-mq.c >> +++ b/block/blk-mq.c >> @@ -1485,7 +1485,9 @@ static void blk_mq_requeue_work(struct work_struct *work) >> blk_mq_request_bypass_insert(rq, 0); >> } else { >> list_del_init(&rq->queuelist); >> - blk_mq_insert_request(rq, BLK_MQ_INSERT_AT_HEAD); >> + blk_mq_insert_request(rq, >> + !blk_rq_is_seq_zoned_write(rq) ? >> + BLK_MQ_INSERT_AT_HEAD : 0); > > Something like: > > } else { > blk_insert_t flags = BLK_MQ_INSERT_AT_HEAD; > > if (blk_rq_is_seq_zoned_write(rq)) > flags = 0; > blk_mq_insert_request(rq, flags); > } > > would be a lot easier to read in my opinion. Hi Damien, I will make this change. Thanks, Bart.
diff --git a/block/blk-mq.c b/block/blk-mq.c index 502dafa76716..ce6ddb249959 100644 --- a/block/blk-mq.c +++ b/block/blk-mq.c @@ -1485,7 +1485,9 @@ static void blk_mq_requeue_work(struct work_struct *work) blk_mq_request_bypass_insert(rq, 0); } else { list_del_init(&rq->queuelist); - blk_mq_insert_request(rq, BLK_MQ_INSERT_AT_HEAD); + blk_mq_insert_request(rq, + !blk_rq_is_seq_zoned_write(rq) ? + BLK_MQ_INSERT_AT_HEAD : 0); } }
blk_mq_requeue_work() inserts requeued requests in front of other requests. This is fine for all request types except for sequential zoned writes. Hence this patch. Note: moving this functionality into the mq-deadline I/O scheduler is not an option because we want to be able to use zoned storage without I/O scheduler. Cc: Christoph Hellwig <hch@lst.de> Cc: Damien Le Moal <dlemoal@kernel.org> Cc: Ming Lei <ming.lei@redhat.com> Cc: Hannes Reinecke <hare@suse.de> Signed-off-by: Bart Van Assche <bvanassche@acm.org> --- block/blk-mq.c | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-)