Hi Jens, In Red Hat internal storage test wrt. blk-mq scheduler, we found that I/O performance is much bad with mq-deadline, especially about sequential I/O on some multi-queue SCSI devcies(lpfc, qla2xxx, SRP...) Turns out one big issue causes the performance regression: requests are still dequeued from sw queue/scheduler queue even when ldd's queue is busy, so I/O merge becomes quite difficult to make, then sequential IO degrades a lot. This issue becomes one of mains reasons for reverting default SCSI_MQ in V4.13. This six patches improve this situation, and brings back performance loss. With this change, SCSI-MQ sequential I/O performance is improved much, Paolo reported that mq-deadline performance improved much[2] in his dbench test wrt V2. Also performance improvement on lpfc/qla2xx was observed with V1.[1] [1] http://marc.info/?l=linux-block&m=150151989915776&w=2 [2] https://marc.info/?l=linux-block&m=150217980602843&w=2 gitweb: https://github.com/ming1/linux/commits/blk_mq_improve_scsi_mpath_perf_V7_part1 V7: - introduce .get_budget/.put_budget, and IO merge gets improved compared with V6, and in theory this approach is better than the way taken in block legacy - drop patch of 'blk-mq-sched: don't dequeue request until all in ->dispatch are flushed' V6: - address comments from Christoph - drop the 1st patch which changes blk_mq_request_bypass_insert(), which should belong to dm-mpath's improvement - move ' blk-mq-sched: move actual dispatching into one helper' as 2nd patch, and use the introduced helper to simplify dispatch logic - merge two previous patches into one for improving dispatch from sw queue - make comment/commit log line width as ~70, as suggested by Christoph V5: - address some comments from Omar - add Tested-by & Reveiewed-by tag - use direct issue for blk_mq_request_bypass_insert(), and start to consider to improve sequential I/O for dm-mpath - only include part 1(the original patch 1 ~ 6), as suggested by Omar V4: - add Reviewed-by tag - some trival change: typo fix in commit log or comment, variable name, no actual functional change V3: - totally round robin for picking req from ctx, as suggested by Bart - remove one local variable in __sbitmap_for_each_set() - drop patches of single dispatch list, which can improve performance on mq-deadline, but cause a bit degrade on none because all hctxs need to be checked after ->dispatch is flushed. Will post it again once it is mature. - rebase on v4.13-rc6 with block for-next V2: - dequeue request from sw queues in round roubin's style as suggested by Bart, and introduces one helper in sbitmap for this purpose - improve bio merge via hash table from sw queue - add comments about using DISPATCH_BUSY state in lockless way, simplifying handling on busy state, - hold ctx->lock when clearing ctx busy bit as suggested by Bart Ming Lei (6): blk-mq-sched: fix scheduler bad performance blk-mq-sched: move actual dispatching into one helper sbitmap: introduce __sbitmap_for_each_set() blk-mq: introduce .get_budget and .put_budget in blk_mq_ops blk-mq-sched: improve dispatching from sw queue SCSI: implement .get_budget and .put_budget for blk-mq block/blk-mq-sched.c | 121 +++++++++++++++++++++++++++++++++++++++++------- block/blk-mq.c | 74 +++++++++++++++++++++++++++-- block/blk-mq.h | 4 +- drivers/scsi/scsi_lib.c | 42 +++++++++++++---- include/linux/blk-mq.h | 12 +++++ include/linux/sbitmap.h | 64 ++++++++++++++++++------- 6 files changed, 268 insertions(+), 49 deletions(-) -- 2.9.5
