Hi Jens, This patchset avoids to allocate driver tag beforehand for flush rq in case of I/O scheduler, then flush rq isn't treated specially wrt. get/put driver tag, code gets cleanup much, such as, reorder_tags_to_front() is removed, and we needn't to worry about request order in dispatch list for avoiding I/O deadlock. 'dbench -t 30 -s 64' has been run on different devices(shared tag, multi-queue, singele queue, ...), and no issues are observed, even very low queue depth test are run, debench still works well. Please consider it for V4.15, thanks! V3: - rebase on the latest for-next of block tree - clean up commit log and comment - include Jianchao's fix on put driver tag V2: - release driver tag before requeue - warning on inserting a req with driver tag Jianchao Wang (1): blk-mq: put the driver tag of nxt rq before first one is requeued Ming Lei (6): blk-flush: don't run queue for requests of bypassing flush block: pass 'run_queue' to blk_mq_request_bypass_insert blk-flush: use blk_mq_request_bypass_insert() blk-mq-sched: decide how to handle flush rq via RQF_FLUSH_SEQ blk-mq: move blk_mq_put_driver_tag*() into blk-mq.h blk-mq: don't allocate driver tag beforehand for flush rq block/blk-core.c | 2 +- block/blk-flush.c | 37 ++++++++++++++------ block/blk-mq-sched.c | 63 +++++++++------------------------- block/blk-mq.c | 97 ++++++++++------------------------------------------ block/blk-mq.h | 35 ++++++++++++++++++- 5 files changed, 97 insertions(+), 137 deletions(-) -- 2.9.5
