| Message ID | 1540546116-1620-1-git-send-email-jianchao.w.wang@oracle.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | [V4] block: fix the DISCARD request merge | expand |
On 10/26/18 3:28 AM, Jianchao Wang wrote: > There are two cases when handle DISCARD merge. > If max_discard_segments == 1, the bios/requests need to be contiguous > to merge. If max_discard_segments > 1, it takes every bio as a range > and different range needn't to be contiguous. > > But now, attempt_merge screws this up. It always consider contiguity > for DISCARD for the case max_discard_segments > 1 and cannot merge > contiguous DISCARD for the case max_discard_segments == 1, because > rq_attempt_discard_merge always returns false in this case. > This patch fixes both of the two cases above. Looks good to me, but: > diff --git a/block/blk-merge.c b/block/blk-merge.c > index 42a4674..cf817c7a 100644 > --- a/block/blk-merge.c > +++ b/block/blk-merge.c > @@ -714,6 +714,32 @@ static void blk_account_io_merge(struct request *req) > part_stat_unlock(); > } > } > +/* > + * Two cases of handling DISCARD merge: > + * If max_discard_segments > 1, the driver takes every bio > + * as a range and send them to controller together. The ranges > + * needn't to be contiguous. > + * Otherwise, the bios/requests will be handled as same as > + * others which should be contiguous. > + */ > +static inline bool blk_discard_mergable(struct request *req) > +{ > + if (req_op(req) == REQ_OP_DISCARD && > + queue_max_discard_segments(req->q) > 1) > + return true; > + else > + return false; > +} Please get rid of the redundant else.
Hi Jens On 10/26/18 10:25 PM, Jens Axboe wrote: > On 10/26/18 3:28 AM, Jianchao Wang wrote: >> There are two cases when handle DISCARD merge. >> If max_discard_segments == 1, the bios/requests need to be contiguous >> to merge. If max_discard_segments > 1, it takes every bio as a range >> and different range needn't to be contiguous. >> >> But now, attempt_merge screws this up. It always consider contiguity >> for DISCARD for the case max_discard_segments > 1 and cannot merge >> contiguous DISCARD for the case max_discard_segments == 1, because >> rq_attempt_discard_merge always returns false in this case. >> This patch fixes both of the two cases above. > > Looks good to me, but: > >> diff --git a/block/blk-merge.c b/block/blk-merge.c >> index 42a4674..cf817c7a 100644 >> --- a/block/blk-merge.c >> +++ b/block/blk-merge.c >> @@ -714,6 +714,32 @@ static void blk_account_io_merge(struct request *req) >> part_stat_unlock(); >> } >> } >> +/* >> + * Two cases of handling DISCARD merge: >> + * If max_discard_segments > 1, the driver takes every bio >> + * as a range and send them to controller together. The ranges >> + * needn't to be contiguous. >> + * Otherwise, the bios/requests will be handled as same as >> + * others which should be contiguous. >> + */ >> +static inline bool blk_discard_mergable(struct request *req) >> +{ >> + if (req_op(req) == REQ_OP_DISCARD && >> + queue_max_discard_segments(req->q) > 1) >> + return true; >> + else >> + return false; >> +} > > Please get rid of the redundant else. > OK, I will change it and post next version. Thanks Jianchao
diff --git a/block/blk-merge.c b/block/blk-merge.c index 42a4674..cf817c7a 100644 --- a/block/blk-merge.c +++ b/block/blk-merge.c @@ -714,6 +714,32 @@ static void blk_account_io_merge(struct request *req) part_stat_unlock(); } } +/* + * Two cases of handling DISCARD merge: + * If max_discard_segments > 1, the driver takes every bio + * as a range and send them to controller together. The ranges + * needn't to be contiguous. + * Otherwise, the bios/requests will be handled as same as + * others which should be contiguous. + */ +static inline bool blk_discard_mergable(struct request *req) +{ + if (req_op(req) == REQ_OP_DISCARD && + queue_max_discard_segments(req->q) > 1) + return true; + else + return false; +} + +enum elv_merge blk_try_req_merge(struct request *req, struct request *next) +{ + if (blk_discard_mergable(req)) + return ELEVATOR_DISCARD_MERGE; + else if (blk_rq_pos(req) + blk_rq_sectors(req) == blk_rq_pos(next)) + return ELEVATOR_BACK_MERGE; + + return ELEVATOR_NO_MERGE; +} /* * For non-mq, this has to be called with the request spinlock acquired. @@ -731,12 +757,6 @@ static struct request *attempt_merge(struct request_queue *q, if (req_op(req) != req_op(next)) return NULL; - /* - * not contiguous - */ - if (blk_rq_pos(req) + blk_rq_sectors(req) != blk_rq_pos(next)) - return NULL; - if (rq_data_dir(req) != rq_data_dir(next) || req->rq_disk != next->rq_disk || req_no_special_merge(next)) @@ -760,11 +780,19 @@ static struct request *attempt_merge(struct request_queue *q, * counts here. Handle DISCARDs separately, as they * have separate settings. */ - if (req_op(req) == REQ_OP_DISCARD) { + + switch (blk_try_req_merge(req, next)) { + case ELEVATOR_DISCARD_MERGE: if (!req_attempt_discard_merge(q, req, next)) return NULL; - } else if (!ll_merge_requests_fn(q, req, next)) + break; + case ELEVATOR_BACK_MERGE: + if (!ll_merge_requests_fn(q, req, next)) + return NULL; + break; + default: return NULL; + } /* * If failfast settings disagree or any of the two is already @@ -888,8 +916,7 @@ bool blk_rq_merge_ok(struct request *rq, struct bio *bio) enum elv_merge blk_try_merge(struct request *rq, struct bio *bio) { - if (req_op(rq) == REQ_OP_DISCARD && - queue_max_discard_segments(rq->q) > 1) + if (blk_discard_mergable(rq)) return ELEVATOR_DISCARD_MERGE; else if (blk_rq_pos(rq) + blk_rq_sectors(rq) == bio->bi_iter.bi_sector) return ELEVATOR_BACK_MERGE;
There are two cases when handle DISCARD merge. If max_discard_segments == 1, the bios/requests need to be contiguous to merge. If max_discard_segments > 1, it takes every bio as a range and different range needn't to be contiguous. But now, attempt_merge screws this up. It always consider contiguity for DISCARD for the case max_discard_segments > 1 and cannot merge contiguous DISCARD for the case max_discard_segments == 1, because rq_attempt_discard_merge always returns false in this case. This patch fixes both of the two cases above. Signed-off-by: Jianchao Wang <jianchao.w.wang@oracle.com> --- V4: - introduce blk_try_req_merge as suggestion of Christoph. V3: - Introduce blk_discard_mergable into attempt_merge and blk_try_merge. - Some comment changes. V2: - Add max_discard_segments > 1 checking in attempt_merge. - Change patch title and comment. - Add more comment in attempt_merge block/blk-merge.c | 47 +++++++++++++++++++++++++++++++++++++---------- 1 file changed, 37 insertions(+), 10 deletions(-)