Jens Axboe 8e8320c931 blk-mq: fix performance regression with shared tags ... If we have shared tags enabled, then every IO completion will trigger a full loop of every queue belonging to a tag set, and every hardware queue for each of those queues, even if nothing needs to be done. This causes a massive performance regression if you have a lot of shared devices. Instead of doing this huge full scan on every IO, add an atomic counter to the main queue that tracks how many hardware queues have been marked as needing a restart. With that, we can avoid looking for restartable queues, if we don't have to. Max reports that this restores performance. Before this patch, 4K IOPS was limited to 22-23K IOPS. With the patch, we are running at 950-970K IOPS. Fixes: 6d8c6c0f97 ("blk-mq: Restart a single queue if tag sets are shared") Reported-by: Max Gurtovoy <maxg@mellanox.com> Tested-by: Max Gurtovoy <maxg@mellanox.com> Reviewed-by: Bart Van Assche <bart.vanassche@sandisk.com> Tested-by: Bart Van Assche <bart.vanassche@wdc.com> Signed-off-by: Jens Axboe <axboe@kernel.dk>		2017-06-21 10:17:49 -06:00
..
partitions	partitions/msdos: FreeBSD UFS2 file systems are not recognized	2017-05-23 09:16:07 -06:00
badblocks.c
bfq-cgroup.c	block, bfq: access and cache blkg data only when safe	2017-06-08 09:51:10 -06:00
bfq-iosched.c	block, bfq: access and cache blkg data only when safe	2017-06-08 09:51:10 -06:00
bfq-iosched.h	block, bfq: access and cache blkg data only when safe	2017-06-08 09:51:10 -06:00
bfq-wf2q.c	block, bfq: use pointer entity->sched_data only if set	2017-05-10 07:39:43 -06:00
bio-integrity.c	bio-integrity: Do not allocate integrity context for bio w/o data	2017-06-03 07:36:27 -06:00
bio.c	Merge branch 'md-next' into md-linus	2017-05-01 14:09:21 -07:00
blk-cgroup.c	block: Avoid that blk_exit_rl() triggers a use-after-free	2017-06-01 13:07:55 -06:00
blk-core.c	block: Avoid that blk_exit_rl() triggers a use-after-free	2017-06-01 13:07:55 -06:00
blk-exec.c	block: remove the errors field from struct request	2017-04-20 12:16:10 -06:00
blk-flush.c	block: make __blk_end_bidi_request private	2017-04-19 10:19:47 -06:00
blk-integrity.c	block: fix blk_integrity_register to use template's interval_exp if not 0	2017-04-23 12:59:56 -06:00
blk-ioc.c
blk-lib.c	block: remove the discard_zeroes_data flag	2017-04-08 11:25:38 -06:00
blk-map.c
blk-merge.c	block: implement splitting of REQ_OP_WRITE_ZEROES bios	2017-04-08 11:25:38 -06:00
blk-mq-cpumap.c
blk-mq-debugfs.c	mq-deadline: add debugfs attributes	2017-05-04 08:25:17 -06:00
blk-mq-debugfs.h	mq-deadline: add debugfs attributes	2017-05-04 08:25:17 -06:00
blk-mq-pci.c
blk-mq-sched.c	blk-mq: fix performance regression with shared tags	2017-06-21 10:17:49 -06:00
blk-mq-sched.h	blk-mq: fix performance regression with shared tags	2017-06-21 10:17:49 -06:00
blk-mq-sysfs.c	blk-mq: untangle debugfs and sysfs	2017-05-04 08:24:13 -06:00
blk-mq-tag.c	blk-mq: add shallow depth option for blk_mq_get_tag()	2017-04-14 14:06:54 -06:00
blk-mq-tag.h
blk-mq-virtio.c
blk-mq.c	blk-mq: fix performance regression with shared tags	2017-06-21 10:17:49 -06:00
blk-mq.h	blk-mq: move debugfs declarations to a separate header file	2017-05-04 08:23:44 -06:00
blk-settings.c	block: remove the discard_zeroes_data flag	2017-04-08 11:25:38 -06:00
blk-softirq.c
blk-stat.c	blk-stat: don't use this_cpu_ptr() in a preemptable section	2017-05-10 07:40:18 -06:00
blk-stat.h	blk-stat: kill blk_stat_rq_ddir()	2017-04-21 07:56:23 -06:00
blk-sysfs.c	block: Fix a blk_exit_rl() regression	2017-06-14 13:27:50 -06:00
blk-tag.c
blk-throttle.c	blk-throttle: set default latency baseline for harddisk	2017-06-07 09:09:32 -06:00
blk-timeout.c	block: remove the errors field from struct request	2017-04-20 12:16:10 -06:00
blk-wbt.c	blk-stat: kill blk_stat_rq_ddir()	2017-04-21 07:56:23 -06:00
blk-wbt.h	block: Make writeback throttling defaults consistent for SQ devices	2017-04-19 08:49:03 -06:00
blk-zoned.c
blk.h	block: Avoid that blk_exit_rl() triggers a use-after-free	2017-06-01 13:07:55 -06:00
bounce.c
bsg-lib.c	scsi: introduce a result field in struct scsi_request	2017-04-20 12:16:10 -06:00
bsg.c	Merge branch 'work.uaccess' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs	2017-05-01 14:41:04 -07:00
cfq-iosched.c	cfq-iosched: fix the delay of cfq_group's vdisktime under iops mode	2017-05-31 09:25:21 -06:00
cmdline-parser.c
compat_ioctl.c	block: remove the discard_zeroes_data flag	2017-04-08 11:25:38 -06:00
deadline-iosched.c
elevator.c	elevator: remove redundant warnings on IO scheduler switch	2017-05-10 07:40:04 -06:00
genhd.c	A reasonably busy cycle for documentation this time around. There is a new	2017-05-02 10:21:17 -07:00
ioctl.c	block: remove the discard_zeroes_data flag	2017-04-08 11:25:38 -06:00
ioprio.c	block: Optimize ioprio_best()	2017-04-19 17:38:36 -06:00
Kconfig	Merge branch 'libnvdimm-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/nvdimm/nvdimm	2017-05-12 15:43:10 -07:00
Kconfig.iosched	block, bfq: add full hierarchical scheduling and cgroups support	2017-04-19 08:30:26 -06:00
kyber-iosched.c	kyber: add debugfs attributes	2017-05-04 08:25:17 -06:00
Makefile	block, bfq: split bfq-iosched.c into multiple source files	2017-04-19 08:48:24 -06:00
mq-deadline.c	mq-deadline: add debugfs attributes	2017-05-04 08:25:17 -06:00
noop-iosched.c
opal_proto.h
partition-generic.c	block: fix an error code in add_partition()	2017-05-23 08:41:59 -06:00
scsi_ioctl.c	scsi: introduce a result field in struct scsi_request	2017-04-20 12:16:10 -06:00
sed-opal.c	block: sed-opal: Tone down all the pr_* to debugs	2017-04-07 14:24:16 -06:00
t10-pi.c