[BUG]
There has an issue of io delayed dispatch caused by io splitting. Consider
the following scenario:
1) If we set a BPS limit of 1MB/s and restrict the maximum IO size per
dispatch to 4KB, submitting -two- 1MB IO requests results in completion
times of 1s and 2s, which is expected.
2) However, if we additionally set an IOPS limit of 1,000,000/s with the
same BPS limit of 1MB/s, submitting -two- 1MB IO requests again results in
both completing in 2s, even though the IOPS constraint is being met.
[CAUSE]
This issue arises because BPS and IOPS currently share the same queue in
the blkthrotl mechanism:
1) This issue does not occur when only BPS is limited because the split IOs
return false in blk_should_throtl() and do not go through to throtl again.
2) For split IOs, even if they have been tagged with BIO_BPS_THROTTLED,
they still get queued alternately in the same list due to continuous
splitting and reordering. As a result, the two IO requests are both
completed at the 2-second mark, causing an unintended delay.
3) It is not difficult to imagine that in this scenario, if N 1MB IOs are
issued at once, all IOs will eventually complete together in N seconds.
[FIX]
With the queue separation introduced in the previous patches, we now have
separate BPS and IOPS queues. For IOs that have already passed the BPS
limitation, they do not need to re-enter the BPS queue and can directly
placed to the IOPS queue.
Since we have split the queues, when the IOPS queue is previously empty
and a new bio is added to the first qnode->bios_iops list in the
service_queue, we also need to update the disptime. This patch introduces
"THROTL_TG_IOPS_WAS_EMPTY" flag to mark it.
Signed-off-by: Zizhi Wo <wozizhi@huawei.com>
Reviewed-by: Yu Kuai <yukuai3@huawei.com>
Signed-off-by: Zizhi Wo <wozizhi@huaweicloud.com>
Link: https://lore.kernel.org/r/20250506020935.655574-8-wozizhi@huaweicloud.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
This patch splits throtl_service_queue->nr_queued into "nr_queued_bps" and
"nr_queued_iops", allowing separate accounting of BPS and IOPS queued bios.
This prepares for future changes that need to check whether the BPS or IOPS
queues are empty.
To facilitate updating the number of IOs in the BPS and IOPS queues, the
addition logic will be moved from throtl_add_bio_tg() to
throtl_qnode_add_bio(), and similarly, the removal logic will be moved from
tg_dispatch_one_bio() to throtl_pop_queued().
And introduce sq_queued() to calculate the total sum of sq->nr_queued.
Signed-off-by: Zizhi Wo <wozizhi@huawei.com>
Reviewed-by: Yu Kuai <yukuai3@huawei.com>
Signed-off-by: Zizhi Wo <wozizhi@huaweicloud.com>
Link: https://lore.kernel.org/r/20250506020935.655574-7-wozizhi@huaweicloud.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
This patch splits the single queue into separate bps and iops queues. Now,
an IO request must first pass through the bps queue, then the iops queue,
and finally be dispatched. Due to the queue splitting, we need to modify
the throtl add/peek/pop function.
Additionally, the patch modifies the logic related to tg_dispatch_time().
If bio needs to wait for bps, function directly returns the bps wait time;
otherwise, it charges bps and returns the iops wait time so that bio can be
directly placed into the iops queue afterward. Note that this may lead to
more frequent updates to disptime, but the overhead is negligible for the
slow path.
Signed-off-by: Zizhi Wo <wozizhi@huawei.com>
Reviewed-by: Yu Kuai <yukuai3@huawei.com>
Signed-off-by: Zizhi Wo <wozizhi@huaweicloud.com>
Link: https://lore.kernel.org/r/20250506020935.655574-6-wozizhi@huaweicloud.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
We no longer need carryover_[bytes/ios] in tg, so it is removed. The
related comments about carryover in tg are also merged into
[bytes/io]_disp, and modify other related comments.
Signed-off-by: Zizhi Wo <wozizhi@huawei.com>
Reviewed-by: Ming Lei <ming.lei@redhat.com>
Reviewed-by: Yu Kuai <yukuai3@huawei.com>
Link: https://lore.kernel.org/r/20250417132054.2866409-3-wozizhi@huaweicloud.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Add an SPDX license identifier line to blk-throttle.h
Use 'GPL-2.0' as the identifier, since blk-throttle.c uses
that, and blk.h (from which some material was copied when
blk-throttle.h was created) also uses that identifier.
Signed-off-by: Tim Bird <tim.bird@sony.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Link: https://lore.kernel.org/r/MW5PR13MB5632EE4645BCA24ED111EC0EFDB62@MW5PR13MB5632.namprd13.prod.outlook.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Now ->carryover_bytes[] and ->carryover_ios[] only covers limit/config
update.
Actually the carryover bytes/ios can be carried to ->bytes_disp[] and
->io_disp[] directly, since the carryover is one-shot thing and only valid
in current slice.
Then we can remove the two fields and simplify code much.
Type of ->bytes_disp[] and ->io_disp[] has to change as signed because the
two fields may become negative when updating limits or config, but both are
big enough for holding bytes/ios dispatched in single slice
Cc: Tejun Heo <tj@kernel.org>
Cc: Josef Bacik <josef@toxicpanda.com>
Cc: Yu Kuai <yukuai3@huawei.com>
Signed-off-by: Ming Lei <ming.lei@redhat.com>
Acked-by: Tejun Heo <tj@kernel.org>
Link: https://lore.kernel.org/r/20250305043123.3938491-4-ming.lei@redhat.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
The two fields are not used any more, so remove them.
Cc: Tejun Heo <tj@kernel.org>
Cc: Josef Bacik <josef@toxicpanda.com>
Cc: Yu Kuai <yukuai3@huawei.com>
Signed-off-by: Ming Lei <ming.lei@redhat.com>
Acked-by: Tejun Heo <tj@kernel.org>
Link: https://lore.kernel.org/r/20250305043123.3938491-2-ming.lei@redhat.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Commit bf20ab538c ("blk-throttle: remove CONFIG_BLK_DEV_THROTTLING_LOW")
attempts to revert the code change introduced by commit cd5ab1b0fc
("blk-throttle: add .low interface"). However, it leaves behind the
bps_conf[] and iops_conf[] fields in the throtl_grp structure which
aren't set anywhere in the new blk-throttle.c code but are still being
used by tg_prfill_limit() to display the limits in io.max. Now io.max
always displays the following values if a block queue is used:
<m>:<n> rbps=0 wbps=0 riops=0 wiops=0
Fix this problem by removing bps_conf[] and iops_conf[] and use bps[]
and iops[] instead to complete the revert.
Fixes: bf20ab538c ("blk-throttle: remove CONFIG_BLK_DEV_THROTTLING_LOW")
Reported-by: Justin Forbes <jforbes@redhat.com>
Closes: https://github.com/containers/podman/issues/22701#issuecomment-2120627789
Signed-off-by: Waiman Long <longman@redhat.com>
Acked-by: Tejun Heo <tj@kernel.org>
Reviewed-by: Yu Kuai <yukuai3@huawei.com>
Link: https://lore.kernel.org/r/20240530134547.970075-1-longman@redhat.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Other cgroup policy like bfq, iocost are lazy-initialized when they are
configured for the first time for the device, but blk-throttle is
initialized unconditionally from blkcg_init_disk().
Delay initialization of blk-throttle as well, to save some cpu and
memory overhead if it's not configured.
Noted that once it's initialized, it can't be destroyed until disk
removal, even if it's disabled.
Signed-off-by: Yu Kuai <yukuai3@huawei.com>
Link: https://lore.kernel.org/r/20240509121107.3195568-3-yukuai1@huaweicloud.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
One the one hand, it's marked EXPERIMENTAL since 2017, and looks like
there are no users since then, and no testers and no developers, it's
just not active at all.
On the other hand, even if the config is disabled, there are still many
fields in throtl_grp and throtl_data and many functions that are only
used for throtl low.
At last, currently blk-throtl is initialized during disk initialization,
and destroyed during disk removal, and it exposes many functions to be
called directly from block layer.
Remove throtl low to make code much more cleaner and follow up work much
easier.
Signed-off-by: Yu Kuai <yukuai3@huawei.com>
Acked-by: Tejun Heo <tj@kernel.org>
Link: https://lore.kernel.org/r/20240509121107.3195568-2-yukuai1@huaweicloud.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
'carryover_bytes/ios' can be negative, indicate that some bio is
dispatched in advance within slice while configuration is updated.
Print a huge value is not user-friendly.
Signed-off-by: Yu Kuai <yukuai3@huawei.com>
Acked-by: Tejun Heo <tj@kernel.org>
Link: https://lore.kernel.org/r/20230816012708.1193747-2-yukuai1@huaweicloud.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
After commit f382fb0bce ("block: remove legacy IO schedulers"),
blkio.throttle.io_serviced and blkio.throttle.io_service_bytes become
the only stable io stats interface of cgroup v1, and these statistics
are done in the blk-throttle code. But the current code only counts the
bios that are actually throttled. When the user does not add the throttle
limit, the io stats for cgroup v1 has nothing. I fix it according to the
statistical method of v2, and made it count all ios accurately.
Fixes: a7b36ee6ba ("block: move blk-throtl fast path inline")
Tested-by: Andrea Righi <andrea.righi@canonical.com>
Signed-off-by: Jinke Han <hanjinke.666@bytedance.com>
Acked-by: Muchun Song <songmuchun@bytedance.com>
Acked-by: Tejun Heo <tj@kernel.org>
Link: https://lore.kernel.org/r/20230507170631.89607-1-hanjinke.666@bytedance.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Pass the gendisk to blk_throtl_cancel_bios as part of moving the
blk-cgroup infrastructure to be gendisk based.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Andreas Herrmann <aherrmann@suse.de>
Acked-by: Tejun Heo <tj@kernel.org>
Link: https://lore.kernel.org/r/20220921180501.1539876-15-hch@lst.de
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Pass the gendisk to blk_throtl_register_queue as part of moving the
blk-cgroup infrastructure to be gendisk based.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Andreas Herrmann <aherrmann@suse.de>
Acked-by: Tejun Heo <tj@kernel.org>
Link: https://lore.kernel.org/r/20220921180501.1539876-14-hch@lst.de
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Pass the gendisk to blk_throtl_init and blk_throtl_exit as part of moving
the blk-cgroup infrastructure to be gendisk based.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Andreas Herrmann <aherrmann@suse.de>
Acked-by: Tejun Heo <tj@kernel.org>
Link: https://lore.kernel.org/r/20220921180501.1539876-13-hch@lst.de
Signed-off-by: Jens Axboe <axboe@kernel.dk>
"tg->has_rules" is extended to "tg->has_rules_iops/bps", thus bios that
don't need to be throttled can be checked accurately.
With this patch, bio will be throttled if:
1) Bio is read/write, and corresponding read/write iops limit exist.
2) If corresponding doesn't exist, corresponding bps limit exist and
bio is not throttled before.
Signed-off-by: Yu Kuai <yukuai3@huawei.com>
Acked-by: Tejun Heo <tj@kernel.org>
Link: https://lore.kernel.org/r/20220921095309.1481289-3-yukuai1@huaweicloud.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Currently, "tg->has_rules" and "tg->flags & THROTL_TG_HAS_IOPS_LIMIT"
both try to bypass bios that don't need to be throttled, however, they are
a little redundant and both not perfect:
1) "tg->has_rules" only distinguish read and write, but not iops and bps
limit.
2) "tg->flags & THROTL_TG_HAS_IOPS_LIMIT" only check if iops limit
exist, read and write is not distinguished, and bps limit is not
checked.
tg->has_rules will extended to distinguish bps and iops in the following
patch. There is no need to keep the flag.
Signed-off-by: Yu Kuai <yukuai3@huawei.com>
Acked-by: Tejun Heo <tj@kernel.org>
Link: https://lore.kernel.org/r/20220921095309.1481289-2-yukuai1@huaweicloud.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
If new configuration is submitted while a bio is throttled, then new
waiting time is recalculated regardless that the bio might already wait
for some time:
tg_conf_updated
throtl_start_new_slice
tg_update_disptime
throtl_schedule_next_dispatch
Then io hung can be triggered by always submmiting new configuration
before the throttled bio is dispatched.
Fix the problem by respecting the time that throttled bio already waited.
In order to do that, add new fields to record how many bytes/io are
waited, and use it to calculate wait time for throttled bio under new
configuration.
Some simple test:
1)
cd /sys/fs/cgroup/blkio/
echo $$ > cgroup.procs
echo "8:0 2048" > blkio.throttle.write_bps_device
{
sleep 2
echo "8:0 1024" > blkio.throttle.write_bps_device
} &
dd if=/dev/zero of=/dev/sda bs=8k count=1 oflag=direct
2)
cd /sys/fs/cgroup/blkio/
echo $$ > cgroup.procs
echo "8:0 1024" > blkio.throttle.write_bps_device
{
sleep 4
echo "8:0 2048" > blkio.throttle.write_bps_device
} &
dd if=/dev/zero of=/dev/sda bs=8k count=1 oflag=direct
test results: io finish time
before this patch with this patch
1) 10s 6s
2) 8s 6s
Signed-off-by: Yu Kuai <yukuai3@huawei.com>
Reviewed-by: Michal Koutný <mkoutny@suse.com>
Acked-by: Tejun Heo <tj@kernel.org>
Link: https://lore.kernel.org/r/20220829022240.3348319-5-yukuai1@huaweicloud.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Test scripts:
cd /sys/fs/cgroup/blkio/
echo "8:0 1024" > blkio.throttle.write_bps_device
echo $$ > cgroup.procs
dd if=/dev/zero of=/dev/sda bs=10k count=1 oflag=direct &
dd if=/dev/zero of=/dev/sda bs=10k count=1 oflag=direct &
Test result:
10240 bytes (10 kB, 10 KiB) copied, 10.0134 s, 1.0 kB/s
10240 bytes (10 kB, 10 KiB) copied, 10.0135 s, 1.0 kB/s
The problem is that the second bio is finished after 10s instead of 20s.
Root cause:
1) second bio will be flagged:
__blk_throtl_bio
while (true) {
...
if (sq->nr_queued[rw]) -> some bio is throttled already
break
};
bio_set_flag(bio, BIO_THROTTLED); -> flag the bio
2) flagged bio will be dispatched without waiting:
throtl_dispatch_tg
tg_may_dispatch
tg_with_in_bps_limit
if (bps_limit == U64_MAX || bio_flagged(bio, BIO_THROTTLED))
*wait = 0; -> wait time is zero
return true;
commit 9f5ede3c01 ("block: throttle split bio in case of iops limit")
support to count split bios for iops limit, thus it adds flagged bio
checking in tg_with_in_bps_limit() so that split bios will only count
once for bps limit, however, it introduce a new problem that io throttle
won't work if multiple bios are throttled.
In order to fix the problem, handle iops/bps limit in different ways:
1) for iops limit, there is no flag to record if the bio is throttled,
and iops is always applied.
2) for bps limit, original bio will be flagged with BIO_BPS_THROTTLED,
and io throttle will ignore bio with the flag.
Noted this patch also remove the code to set flag in __bio_clone(), it's
introduced in commit 111be88398 ("block-throttle: avoid double
charge"), and author thinks split bio can be resubmited and throttled
again, which is wrong because split bio will continue to dispatch from
caller.
Fixes: 9f5ede3c01 ("block: throttle split bio in case of iops limit")
Cc: <stable@vger.kernel.org>
Signed-off-by: Yu Kuai <yukuai3@huawei.com>
Acked-by: Tejun Heo <tj@kernel.org>
Link: https://lore.kernel.org/r/20220829022240.3348319-2-yukuai1@huaweicloud.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Throttled bios can't be issued after del_gendisk() is done, thus
it's better to cancel them immediately rather than waiting for
throttle is done.
For example, if user thread is throttled with low bps while it's
issuing large io, and the device is deleted. The user thread will
wait for a long time for io to return.
Signed-off-by: Yu Kuai <yukuai3@huawei.com>
Signed-off-by: Ming Lei <ming.lei@redhat.com>
Link: https://lore.kernel.org/r/20220318130144.1066064-4-ming.lei@redhat.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Revert commit 4f1e9630af ("blk-throtl: optimize IOPS throttle for large
IO scenarios") since we have another easier way to address this issue and
get better iops throttling result.
Acked-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Ming Lei <ming.lei@redhat.com>
Link: https://lore.kernel.org/r/20220216044514.2903784-9-ming.lei@redhat.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
We need to throttle split bio in case of IOPS limit even though the
split bio has been marked as BIO_THROTTLED since block layer
accounts split bio actually.
If only throughput throttle is setup, no need to throttle any more
if BIO_THROTTLED is set since we have accounted & considered the
whole bio bytes already.
Add one flag of THROTL_TG_HAS_IOPS_LIMIT for serving this purpose.
Acked-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Ming Lei <ming.lei@redhat.com>
Link: https://lore.kernel.org/r/20220216044514.2903784-8-ming.lei@redhat.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Commit 111be88398 ("block-throttle: avoid double charge") marks bio as
BIO_THROTTLED unconditionally if __blk_throtl_bio() is called on this bio,
then this bio won't be called into __blk_throtl_bio() any more. This way
is to avoid double charge in case of bio splitting. It is reasonable for
read/write throughput limit, but not reasonable for IOPS limit because
block layer provides io accounting against split bio.
Chunguang Xu has already observed this issue and fixed it in commit
4f1e9630af ("blk-throtl: optimize IOPS throttle for large IO scenarios").
However, that patch only covers bio splitting in __blk_queue_split(), and
we have other kind of bio splitting, such as bio_split() &
submit_bio_noacct() and other ways.
This patch tries to fix the issue in one generic way by always charging
the bio for iops limit in blk_throtl_bio(). This way is reasonable:
re-submission & fast-cloned bio is charged if it is submitted to same
disk/queue, and BIO_THROTTLED will be cleared if bio->bi_bdev is changed.
This new approach can get much more smooth/stable iops limit compared with
commit 4f1e9630af ("blk-throtl: optimize IOPS throttle for large IO
scenarios") since that commit can't throttle current split bios actually.
Also this way won't cause new double bio iops charge in
blk_throtl_dispatch_work_fn() in which blk_throtl_bio() won't be called
any more.
Reported-by: Ning Li <lining2020x@163.com>
Acked-by: Tejun Heo <tj@kernel.org>
Cc: Chunguang Xu <brookxu@tencent.com>
Signed-off-by: Ming Lei <ming.lei@redhat.com>
Link: https://lore.kernel.org/r/20220216044514.2903784-7-ming.lei@redhat.com
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>
