From 34e0a279a993debaff03158fc2fbf6a00c093643 Mon Sep 17 00:00:00 2001 From: Jan Kara Date: Mon, 13 Mar 2023 10:30:02 +0100 Subject: block: do not reverse request order when flushing plug list Commit 26fed4ac4eab ("block: flush plug based on hardware and software queue order") changed flushing of plug list to submit requests one device at a time. However while doing that it also started using list_add_tail() instead of list_add() used previously thus effectively submitting requests in reverse order. Also when forming a rq_list with remaining requests (in case two or more devices are used), we effectively reverse the ordering of the plug list for each device we process. Submitting requests in reverse order has negative impact on performance for rotational disks (when BFQ is not in use). We observe 10-25% regression in random 4k write throughput, as well as ~20% regression in MariaDB OLTP benchmark on rotational storage on btrfs filesystem. Fix the problem by preserving ordering of the plug list when inserting requests into the queuelist as well as by appending to requeue_list instead of prepending to it. Fixes: 26fed4ac4eab ("block: flush plug based on hardware and software queue order") Signed-off-by: Jan Kara Reviewed-by: Christoph Hellwig Link: https://lore.kernel.org/r/20230313093002.11756-1-jack@suse.cz Signed-off-by: Jens Axboe --- block/blk-mq.c | 5 +++-- 1 file changed, 3 insertions(+), 2 deletions(-) (limited to 'block') diff --git a/block/blk-mq.c b/block/blk-mq.c index d0cb2ef18fe2..cf1a39adf9a5 100644 --- a/block/blk-mq.c +++ b/block/blk-mq.c @@ -2725,6 +2725,7 @@ static void blk_mq_dispatch_plug_list(struct blk_plug *plug, bool from_sched) struct blk_mq_hw_ctx *this_hctx = NULL; struct blk_mq_ctx *this_ctx = NULL; struct request *requeue_list = NULL; + struct request **requeue_lastp = &requeue_list; unsigned int depth = 0; LIST_HEAD(list); @@ -2735,10 +2736,10 @@ static void blk_mq_dispatch_plug_list(struct blk_plug *plug, bool from_sched) this_hctx = rq->mq_hctx; this_ctx = rq->mq_ctx; } else if (this_hctx != rq->mq_hctx || this_ctx != rq->mq_ctx) { - rq_list_add(&requeue_list, rq); + rq_list_add_tail(&requeue_lastp, rq); continue; } - list_add_tail(&rq->queuelist, &list); + list_add(&rq->queuelist, &list); depth++; } while (!rq_list_empty(plug->mq_list)); -- cgit From 00e885efcfbb8712d3e1bfc1ae30639c15ca1d3b Mon Sep 17 00:00:00 2001 From: Chris Leech Date: Fri, 10 Mar 2023 09:09:13 +0800 Subject: blk-mq: fix "bad unlock balance detected" on q->srcu in __blk_mq_run_dispatch_ops The 'q' parameter of the macro __blk_mq_run_dispatch_ops may not be one local variable, such as, it is rq->q, then request queue pointed by this variable could be changed to another queue in case of BLK_MQ_F_TAG_QUEUE_SHARED after 'dispatch_ops' returns, then 'bad unlock balance' is triggered. Fixes the issue by adding one local variable for doing srcu lock/unlock. Fixes: 2a904d00855f ("blk-mq: remove hctx_lock and hctx_unlock") Cc: Marco Patalano Signed-off-by: Chris Leech Signed-off-by: Ming Lei Link: https://lore.kernel.org/r/20230310010913.1014789-1-ming.lei@redhat.com Signed-off-by: Jens Axboe --- block/blk-mq.h | 5 +++-- 1 file changed, 3 insertions(+), 2 deletions(-) (limited to 'block') diff --git a/block/blk-mq.h b/block/blk-mq.h index ef59fee62780..a7482d2cc82e 100644 --- a/block/blk-mq.h +++ b/block/blk-mq.h @@ -378,12 +378,13 @@ static inline bool hctx_may_queue(struct blk_mq_hw_ctx *hctx, #define __blk_mq_run_dispatch_ops(q, check_sleep, dispatch_ops) \ do { \ if ((q)->tag_set->flags & BLK_MQ_F_BLOCKING) { \ + struct blk_mq_tag_set *__tag_set = (q)->tag_set; \ int srcu_idx; \ \ might_sleep_if(check_sleep); \ - srcu_idx = srcu_read_lock((q)->tag_set->srcu); \ + srcu_idx = srcu_read_lock(__tag_set->srcu); \ (dispatch_ops); \ - srcu_read_unlock((q)->tag_set->srcu, srcu_idx); \ + srcu_read_unlock(__tag_set->srcu, srcu_idx); \ } else { \ rcu_read_lock(); \ (dispatch_ops); \ -- cgit From 5f27571382ca42daa3e3d40d1b252bf18c2b61d2 Mon Sep 17 00:00:00 2001 From: Yu Kuai Date: Thu, 23 Feb 2023 17:12:26 +0800 Subject: block: count 'ios' and 'sectors' when io is done for bio-based device While using iostat for raid, I observed very strange 'await' occasionally, and turns out it's due to that 'ios' and 'sectors' is counted in bdev_start_io_acct(), while 'nsecs' is counted in bdev_end_io_acct(). I'm not sure why they are ccounted like that but I think this behaviour is obviously wrong because user will get wrong disk stats. Fix the problem by counting 'ios' and 'sectors' when io is done, like what rq-based device does. Fixes: 394ffa503bc4 ("blk: introduce generic io stat accounting help function") Signed-off-by: Yu Kuai Reviewed-by: Christoph Hellwig Link: https://lore.kernel.org/r/20230223091226.1135678-1-yukuai1@huaweicloud.com Signed-off-by: Jens Axboe --- block/blk-core.c | 16 ++++++---------- 1 file changed, 6 insertions(+), 10 deletions(-) (limited to 'block') diff --git a/block/blk-core.c b/block/blk-core.c index 9e5e0277a4d9..42926e6cb83c 100644 --- a/block/blk-core.c +++ b/block/blk-core.c @@ -959,16 +959,11 @@ again: } } -unsigned long bdev_start_io_acct(struct block_device *bdev, - unsigned int sectors, enum req_op op, +unsigned long bdev_start_io_acct(struct block_device *bdev, enum req_op op, unsigned long start_time) { - const int sgrp = op_stat_group(op); - part_stat_lock(); update_io_ticks(bdev, start_time, false); - part_stat_inc(bdev, ios[sgrp]); - part_stat_add(bdev, sectors[sgrp], sectors); part_stat_local_inc(bdev, in_flight[op_is_write(op)]); part_stat_unlock(); @@ -984,13 +979,12 @@ EXPORT_SYMBOL(bdev_start_io_acct); */ unsigned long bio_start_io_acct(struct bio *bio) { - return bdev_start_io_acct(bio->bi_bdev, bio_sectors(bio), - bio_op(bio), jiffies); + return bdev_start_io_acct(bio->bi_bdev, bio_op(bio), jiffies); } EXPORT_SYMBOL_GPL(bio_start_io_acct); void bdev_end_io_acct(struct block_device *bdev, enum req_op op, - unsigned long start_time) + unsigned int sectors, unsigned long start_time) { const int sgrp = op_stat_group(op); unsigned long now = READ_ONCE(jiffies); @@ -998,6 +992,8 @@ void bdev_end_io_acct(struct block_device *bdev, enum req_op op, part_stat_lock(); update_io_ticks(bdev, now, true); + part_stat_inc(bdev, ios[sgrp]); + part_stat_add(bdev, sectors[sgrp], sectors); part_stat_add(bdev, nsecs[sgrp], jiffies_to_nsecs(duration)); part_stat_local_dec(bdev, in_flight[op_is_write(op)]); part_stat_unlock(); @@ -1007,7 +1003,7 @@ EXPORT_SYMBOL(bdev_end_io_acct); void bio_end_io_acct_remapped(struct bio *bio, unsigned long start_time, struct block_device *orig_bdev) { - bdev_end_io_acct(orig_bdev, bio_op(bio), start_time); + bdev_end_io_acct(orig_bdev, bio_op(bio), bio_sectors(bio), start_time); } EXPORT_SYMBOL_GPL(bio_end_io_acct_remapped); -- cgit From 8f0d196e4dc137470bbd5de98278d941c8002fcb Mon Sep 17 00:00:00 2001 From: Lukas Bulwahn Date: Thu, 16 Mar 2023 12:16:30 +0100 Subject: block: remove obsolete config BLOCK_COMPAT Before commit bdc1ddad3e5f ("compat_ioctl: block: move blkdev_compat_ioctl() into ioctl.c"), the config BLOCK_COMPAT was used to include compat_ioctl.c into the kernel build. With this commit, the code is moved into ioctl.c and included with the config COMPAT. So, since then, the config BLOCK_COMPAT has no effect and any further purpose. Remove this obsolete config BLOCK_COMPAT. Signed-off-by: Lukas Bulwahn Reviewed-by: Christoph Hellwig Acked-by: Arnd Bergmann Link: https://lore.kernel.org/r/20230316111630.4897-1-lukas.bulwahn@gmail.com Signed-off-by: Jens Axboe --- block/Kconfig | 3 --- 1 file changed, 3 deletions(-) (limited to 'block') diff --git a/block/Kconfig b/block/Kconfig index 5d9d9c84d516..941b2dca70db 100644 --- a/block/Kconfig +++ b/block/Kconfig @@ -204,9 +204,6 @@ config BLK_INLINE_ENCRYPTION_FALLBACK source "block/partitions/Kconfig" -config BLOCK_COMPAT - def_bool COMPAT - config BLK_MQ_PCI def_bool PCI -- cgit