Currently we scan the entire plug list, which is potentially very
expensive. In an IOPS bound workload, we can drive about 5.6M IOPS with
merging enabled, and profiling shows that the plug merge check is the
(by far) most expensive thing we're doing:
Overhead Command Shared Object Symbol
+ 20.89% io_uring [kernel.vmlinux] [k] blk_attempt_plug_merge
+ 4.98% io_uring [kernel.vmlinux] [k] io_submit_sqes
+ 4.78% io_uring [kernel.vmlinux] [k] blkdev_direct_IO
+ 4.61% io_uring [kernel.vmlinux] [k] blk_mq_submit_bio
Instead of browsing the whole list, just check the previously inserted
entry. That is enough for a naive merge check and will catch most cases,
and for devices that need full merging, the IO scheduler attached to
such devices will do that anyway. The plug merge is meant to be an
inexpensive check to avoid getting a request, but if we repeatedly
scan the list for every single insert, it is very much not a cheap
check.
With this patch, the workload instead runs at ~7.0M IOPS, providing
a 25% improvement. Disabling merging entirely yields another 5%
improvement.
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Every object under block/ depends on CONFIG_BLOCK.
Move the guard to the top Makefile since there is no point to
descend into block/ if CONFIG_BLOCK=n.
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Link: https://lore.kernel.org/r/20210927140000.866249-5-masahiroy@kernel.org
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Everything under block/ depends on BLOCK. BLOCK_HOLDER_DEPRECATED is
selected from drivers/md/Kconfig, which is entirely dependent on BLOCK.
Extend the 'if BLOCK' ... 'endif' so it covers the whole block/Kconfig.
Also, clean up the definition of BLOCK_COMPAT and BLK_MQ_PCI because
COMPAT and PCI are boolean.
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Link: https://lore.kernel.org/r/20210927140000.866249-3-masahiroy@kernel.org
Signed-off-by: Jens Axboe <axboe@kernel.dk>
CONFIG_BLK_CGROUP is a boolean option, that is, its value is 'y' or 'n'.
The comparison to 'y' is redundant.
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Link: https://lore.kernel.org/r/20210927140000.866249-2-masahiroy@kernel.org
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Add a blk_mq_get_tags() helper, which uses the new sbitmap API for
allocating a batch of tags all at once. This both simplifies the block
code for batched allocation, and it is also more efficient than just
doing repeated calls into __sbitmap_queue_get().
This reduces the sbitmap overhead in peak runs from ~3% to ~1% and
yields a performanc increase from 6.6M IOPS to 6.8M IOPS for a single
CPU core.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
The block layer tag allocation batching still calls into sbitmap to get
each tag, but we can improve on that. Add __sbitmap_queue_get_batch(),
which returns a mask of tags all at once, along with an offset for
those tags.
An example return would be 0xff, where bits 0..7 are set, with
tag_offset == 128. The valid tags in this case would be 128..135.
A batch is specific to an individual sbitmap_map, hence it cannot be
larger than that. The requested number of tags is automatically reduced
to the max that can be satisfied with a single map.
On failure, 0 is returned. Caller should fall back to single tag
allocation at that point/
Signed-off-by: Jens Axboe <axboe@kernel.dk>
We already have a blk_mq_need_time_stamp() check in
__blk_mq_end_request() to get a timestamp, hide all the statistics
accounting under it. It cuts some cycles for requests that don't need
stats, and is free otherwise.
Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Link: https://lore.kernel.org/r/e0f2ea812e93a8adcd07101212e7d7e70ca304e7.1634115360.git.asml.silence@gmail.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
bio_get_first_bvec and bio_get_last_bvec are only used in blk-merge.c,
so move them there.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Link: https://lore.kernel.org/r/20211012161804.991559-8-hch@lst.de
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Mark __bio_try_merge_page static and move it up a bit to avoid the need
for a forward declaration.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Link: https://lore.kernel.org/r/20211012161804.991559-7-hch@lst.de
Signed-off-by: Jens Axboe <axboe@kernel.dk>
BIO_DEBUG is always defined, so just switch the two instances to use
BUG_ON directly.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Link: https://lore.kernel.org/r/20211012161804.991559-2-hch@lst.de
Signed-off-by: Jens Axboe <axboe@kernel.dk>
We could have a race here, where the request gets freed before we call
into blk_mq_run_hw_queue(). If this happens, we cannot rely on the state
of the request.
Grab the hardware context before inserting the flush.
Fixes: 0f38d76646 ("blk-mq: cleanup blk_mq_submit_bio")
Reviewed-by: Ming Lei <ming.lei@redhat.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Move the blk_mq_alloc_data stack allocation only into the branch
that actually needs it, and use rq->mq_hctx instead of data.hctx
to refer to the hctx.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Link: https://lore.kernel.org/r/20211012104045.658051-3-hch@lst.de
Signed-off-by: Jens Axboe <axboe@kernel.dk>
The newly added loop for the cached requests in __blk_mq_alloc_request
is a little too convoluted for my taste, so unwind it a bit. Also
rename the function to __blk_mq_alloc_requests now that it can allocate
more than a single request.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Link: https://lore.kernel.org/r/20211012104045.658051-2-hch@lst.de
Signed-off-by: Jens Axboe <axboe@kernel.dk>
The caller typically has a good (or even exact) idea of how many requests
it needs to submit. We can make the request/tag allocation a lot more
efficient if we just allocate N requests/tags upfront when we queue the
first bio from the batch.
Provide a new plug start helper that allows the caller to specify how many
IOs are expected. This sets plug->nr_ios, and we can use that for smarter
request allocation. The plug provides a holding spot for requests, and
request allocation will check it before calling into the normal request
allocation path.
The blk_finish_plug() is called, check if there are unused requests and
free them. This should not happen in normal operations. The exception is
if we get merging, then we may be left with requests that need freeing
when done.
This raises the per-core performance on my setup from ~5.8M to ~6.1M
IOPS.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Particularly for NVMe with efficient deferred submission for many
requests, there are nice benefits to be seen by bumping the default max
plug count from 16 to 32. This is especially true for virtualized setups,
where the submit part is more expensive. But can be noticed even on
native hardware.
Reduce the multiple queue factor from 4 to 2, since we're changing the
default size.
While changing it, move the defines into the block layer private header.
These aren't values that anyone outside of the block layer uses, or
should use.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Doing high IOPS testing with blk-cgroups enabled spends ~15-20% of the
time just doing ktime_get_ns() -> readtsc. We essentially read and
set the start time twice, one for the bio and then again when that bio
is mapped to a request.
Given that the time between the two is very short, inherit the bio
start time instead of reading it again. This cuts 1/3rd of the overhead
of the time keeping.
Acked-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Even if no policies are defined, we spend ~2% of the total IO time
checking. Move the fast path inline.
Acked-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Now that we use shared tags for shared sbitmap support, we don't require
the tags sbitmap pointers, so drop them.
This essentially reverts commit 222a5ae03c ("blk-mq: Use pointers for
blk_mq_tags bitmap tags").
Function blk_mq_init_bitmap_tags() is removed also, since it would be only
a wrappper for blk_mq_init_bitmaps().
Reviewed-by: Ming Lei <ming.lei@redhat.com>
Reviewed-by: Hannes Reinecke <hare@suse.de>
Signed-off-by: John Garry <john.garry@huawei.com>
Link: https://lore.kernel.org/r/1633429419-228500-14-git-send-email-john.garry@huawei.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Currently we use separate sbitmap pairs and active_queues atomic_t for
shared sbitmap support.
However a full sets of static requests are used per HW queue, which is
quite wasteful, considering that the total number of requests usable at
any given time across all HW queues is limited by the shared sbitmap depth.
As such, it is considerably more memory efficient in the case of shared
sbitmap to allocate a set of static rqs per tag set or request queue, and
not per HW queue.
So replace the sbitmap pairs and active_queues atomic_t with a shared
tags per tagset and request queue, which will hold a set of shared static
rqs.
Since there is now no valid HW queue index to be passed to the blk_mq_ops
.init and .exit_request callbacks, pass an invalid index token. This
changes the semantics of the APIs, such that the callback would need to
validate the HW queue index before using it. Currently no user of shared
sbitmap actually uses the HW queue index (as would be expected).
Signed-off-by: John Garry <john.garry@huawei.com>
Reviewed-by: Ming Lei <ming.lei@redhat.com>
Link: https://lore.kernel.org/r/1633429419-228500-13-git-send-email-john.garry@huawei.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Refactor blk_mq_free_map_and_requests() such that it can be used at many
sites at which the tag map and rqs are freed.
Also rename to blk_mq_free_map_and_rqs(), which is shorter and matches the
alloc equivalent.
Suggested-by: Ming Lei <ming.lei@redhat.com>
Signed-off-by: John Garry <john.garry@huawei.com>
Reviewed-by: Hannes Reinecke <hare@suse.de>
Link: https://lore.kernel.org/r/1633429419-228500-12-git-send-email-john.garry@huawei.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Add a function to combine allocating tags and the associated requests,
and factor out common patterns to use this new function.
Some function only call blk_mq_alloc_map_and_rqs() now, but more
functionality will be added later.
Also make blk_mq_alloc_rq_map() and blk_mq_alloc_rqs() static since they
are only used in blk-mq.c, and finally rename some functions for
conciseness and consistency with other function names:
- __blk_mq_alloc_map_and_{request -> rqs}()
- blk_mq_alloc_{map_and_requests -> set_map_and_rqs}()
Suggested-by: Ming Lei <ming.lei@redhat.com>
Signed-off-by: John Garry <john.garry@huawei.com>
Reviewed-by: Hannes Reinecke <hare@suse.de>
Reviewed-by: Ming Lei <ming.lei@redhat.com>
Link: https://lore.kernel.org/r/1633429419-228500-11-git-send-email-john.garry@huawei.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Put the functionality to update the sched shared sbitmap size in a common
function.
Since the same formula is always used to resize, and it can be got from
the request queue argument, so just pass the request queue pointer.
Signed-off-by: John Garry <john.garry@huawei.com>
Reviewed-by: Ming Lei <ming.lei@redhat.com>
Reviewed-by: Hannes Reinecke <hare@suse.de>
Link: https://lore.kernel.org/r/1633429419-228500-10-git-send-email-john.garry@huawei.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Function blk_mq_clear_rq_mapping() is required to clear the sched tags
mappings in driver tags rqs[].
But there is no need for a driver tags to clear its own mapping, so skip
clearing the mapping in this scenario.
Signed-off-by: John Garry <john.garry@huawei.com>
Reviewed-by: Hannes Reinecke <hare@suse.de>
Reviewed-by: Ming Lei <ming.lei@redhat.com>
Link: https://lore.kernel.org/r/1633429419-228500-9-git-send-email-john.garry@huawei.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Function blk_mq_clear_rq_mapping() will be used for shared sbitmap tags
in future, so pass a driver tags pointer instead of the tagset container
and HW queue index.
Signed-off-by: John Garry <john.garry@huawei.com>
Reviewed-by: Hannes Reinecke <hare@suse.de>
Reviewed-by: Ming Lei <ming.lei@redhat.com>
Link: https://lore.kernel.org/r/1633429419-228500-8-git-send-email-john.garry@huawei.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
To be more concise and consistent in naming, rename
blk_mq_sched_free_requests() -> blk_mq_sched_free_rqs().
Signed-off-by: John Garry <john.garry@huawei.com>
Reviewed-by: Hannes Reinecke <hare@suse.de>
Reviewed-by: Ming Lei <ming.lei@redhat.com>
Link: https://lore.kernel.org/r/1633429419-228500-7-git-send-email-john.garry@huawei.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Function blk_mq_sched_alloc_tags() does same as
__blk_mq_alloc_map_and_request(), so give a similar name to be consistent.
Similarly rename label err_free_tags -> err_free_map_and_rqs.
Signed-off-by: John Garry <john.garry@huawei.com>
Reviewed-by: Ming Lei <ming.lei@redhat.com>
Reviewed-by: Hannes Reinecke <hare@suse.de>
Link: https://lore.kernel.org/r/1633429419-228500-6-git-send-email-john.garry@huawei.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
It's easier to read:
if (x)
X;
else
Y;
over:
if (!x)
Y;
else
X;
No functional change intended.
Signed-off-by: John Garry <john.garry@huawei.com>
Reviewed-by: Ming Lei <ming.lei@redhat.com>
Reviewed-by: Hannes Reinecke <hare@suse.de>
Link: https://lore.kernel.org/r/1633429419-228500-5-git-send-email-john.garry@huawei.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
For shared sbitmap, if the call to blk_mq_tag_update_depth() was
successful for any hctx when hctx->sched_tags is not set, then it would be
successful for all (due to nature in which blk_mq_tag_update_depth()
fails).
As such, there is no need to call blk_mq_tag_resize_shared_sbitmap() for
each hctx. So relocate the call until after the hctx iteration under the
!q->elevator check, which is equivalent (to !hctx->sched_tags).
Signed-off-by: John Garry <john.garry@huawei.com>
Reviewed-by: Ming Lei <ming.lei@redhat.com>
Reviewed-by: Hannes Reinecke <hare@suse.de>
Link: https://lore.kernel.org/r/1633429419-228500-4-git-send-email-john.garry@huawei.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
It is a bit confusing that there is BLKDEV_MAX_RQ and MAX_SCHED_RQ, as
the name BLKDEV_MAX_RQ would imply the max requests always, which it is
not.
Rename to BLKDEV_MAX_RQ to BLKDEV_DEFAULT_RQ, matching its usage - that being
the default number of requests assigned when allocating a request queue.
Signed-off-by: John Garry <john.garry@huawei.com>
Reviewed-by: Ming Lei <ming.lei@redhat.com>
Reviewed-by: Hannes Reinecke <hare@suse.de>
Link: https://lore.kernel.org/r/1633429419-228500-3-git-send-email-john.garry@huawei.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
The original code in commit 24d2f90309 ("blk-mq: split out tag
initialization, support shared tags") would check tags->rqs is non-NULL and
then dereference tags->rqs[].
Then in commit 2af8cbe305 ("blk-mq: split tag ->rqs[] into two"), we
started to dereference tags->static_rqs[], but continued to check non-NULL
tags->rqs.
Check tags->static_rqs as non-NULL instead, which is more logical.
Signed-off-by: John Garry <john.garry@huawei.com>
Reviewed-by: Ming Lei <ming.lei@redhat.com>
Reviewed-by: Hannes Reinecke <hare@suse.de>
Link: https://lore.kernel.org/r/1633429419-228500-2-git-send-email-john.garry@huawei.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Make the bad sector information a little more useful by printing
current->comm to identify the caller.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Damien Le Moal <damien.lemoal@wdc.com>
Reviewed-by: Chaitanya Kulkarni <kch@nvidia.com>
Link: https://lore.kernel.org/r/20210928052755.113016-1-hch@lst.de
Signed-off-by: Jens Axboe <axboe@kernel.dk>
In addition to reverting commit 7b05bf7710 ("Revert "block/mq-deadline:
Prioritize high-priority requests""), this patch uses 'jiffies' instead
of ktime_get() in the code for aging lower priority requests.
This patch has been tested as follows:
Measured QD=1/jobs=1 IOPS for nullb with the mq-deadline scheduler.
Result without and with this patch: 555 K IOPS.
Measured QD=1/jobs=8 IOPS for nullb with the mq-deadline scheduler.
Result without and with this patch: about 380 K IOPS.
Ran the following script:
set -e
scriptdir=$(dirname "$0")
if [ -e /sys/module/scsi_debug ]; then modprobe -r scsi_debug; fi
modprobe scsi_debug ndelay=1000000 max_queue=16
sd=''
while [ -z "$sd" ]; do
sd=$(basename /sys/bus/pseudo/drivers/scsi_debug/adapter*/host*/target*/*/block/*)
done
echo $((100*1000)) > "/sys/block/$sd/queue/iosched/prio_aging_expire"
if [ -e /sys/fs/cgroup/io.prio.class ]; then
cd /sys/fs/cgroup
echo restrict-to-be >io.prio.class
echo +io > cgroup.subtree_control
else
cd /sys/fs/cgroup/blkio/
echo restrict-to-be >blkio.prio.class
fi
echo $$ >cgroup.procs
mkdir -p hipri
cd hipri
if [ -e io.prio.class ]; then
echo none-to-rt >io.prio.class
else
echo none-to-rt >blkio.prio.class
fi
{ "${scriptdir}/max-iops" -a1 -d32 -j1 -e mq-deadline "/dev/$sd" >& ~/low-pri.txt & }
echo $$ >cgroup.procs
"${scriptdir}/max-iops" -a1 -d32 -j1 -e mq-deadline "/dev/$sd" >& ~/hi-pri.txt
Result:
* 11000 IOPS for the high-priority job
* 40 IOPS for the low-priority job
If the prio aging expiry time is changed from 100s into 0, the IOPS results
change into 6712 and 6796 IOPS.
The max-iops script is a script that runs fio with the following arguments:
--bs=4K --gtod_reduce=1 --ioengine=libaio --ioscheduler=${arg_e} --runtime=60
--norandommap --rw=read --thread --buffered=0 --numjobs=${arg_j}
--iodepth=${arg_d} --iodepth_batch_submit=${arg_a}
--iodepth_batch_complete=$((arg_d / 2)) --name=${positional_argument_1}
--filename=${positional_argument_1}
Cc: Damien Le Moal <damien.lemoal@wdc.com>
Cc: Niklas Cassel <Niklas.Cassel@wdc.com>
Cc: Hannes Reinecke <hare@suse.de>
Signed-off-by: Bart Van Assche <bvanassche@acm.org>
Reviewed-by: Damien Le Moal <damien.lemoal@opensource.wdc.com>
Link: https://lore.kernel.org/r/20210927220328.1410161-5-bvanassche@acm.org
[axboe: @latest -> @latest_start]
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Calculating the sum over all CPUs of per-CPU counters frequently is
inefficient. Hence switch from per-CPU to individual counters. Three
counters are protected by the mq-deadline spinlock since these are
only accessed from contexts that already hold that spinlock. The fourth
counter is atomic because protecting it with the mq-deadline spinlock
would trigger lock contention.
Reviewed-by: Damien Le Moal <damien.lemoal@wdc.com>
Reviewed-by: Niklas Cassel <Niklas.Cassel@wdc.com>
Cc: Hannes Reinecke <hare@suse.de>
Signed-off-by: Bart Van Assche <bvanassche@acm.org>
Link: https://lore.kernel.org/r/20210927220328.1410161-4-bvanassche@acm.org
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Check a statistics invariant at module unload time. When running
blktests, the invariant is verified every time a request queue is
removed and hence is verified at least once per test.
Reviewed-by: Damien Le Moal <damien.lemoal@wdc.com>
Reviewed-by: Niklas Cassel <Niklas.Cassel@wdc.com>
Cc: Hannes Reinecke <hare@suse.de>
Signed-off-by: Bart Van Assche <bvanassche@acm.org>
Link: https://lore.kernel.org/r/20210927220328.1410161-3-bvanassche@acm.org
Signed-off-by: Jens Axboe <axboe@kernel.dk>
The scheduler .insert_requests() callback is called when a request is
queued for the first time and also when it is requeued. Only count a
request the first time it is queued. Additionally, since the mq-deadline
scheduler only performs zone locking for requests that have been
inserted, skip the zone unlock code for requests that have not been
inserted into the mq-deadline scheduler.
Fixes: 38ba64d12d ("block/mq-deadline: Track I/O statistics")
Reviewed-by: Damien Le Moal <damien.lemoal@wdc.com>
Reviewed-by: Niklas Cassel <Niklas.Cassel@wdc.com>
Cc: Hannes Reinecke <hare@suse.de>
Signed-off-by: Bart Van Assche <bvanassche@acm.org>
Link: https://lore.kernel.org/r/20210927220328.1410161-2-bvanassche@acm.org
Signed-off-by: Jens Axboe <axboe@kernel.dk>
struct request is only used by blk-mq drivers, so move it and all
related declarations to blk-mq.h.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Johannes Thumshirn <johannes.thumshirn@wdc.com>
Link: https://lore.kernel.org/r/20210920123328.1399408-18-hch@lst.de
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Split the integrity/metadata handling definitions out into a new header.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Johannes Thumshirn <johannes.thumshirn@wdc.com>
Link: https://lore.kernel.org/r/20210920123328.1399408-17-hch@lst.de
Signed-off-by: Jens Axboe <axboe@kernel.dk>
These are block-layer internal helpers, so move them to block/blk.h and
block/blk-merge.c. Also update a comment a bit to use better grammar.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Johannes Thumshirn <johannes.thumshirn@wdc.com>
Link: https://lore.kernel.org/r/20210920123328.1399408-16-hch@lst.de
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Jens Axboe