Lina reported that if throttle limits are initially very high and then
dropped, then no new bio might be dispatched for a long time. And the
reason being that after dropping the limits we don't reset the existing
slice and do the rate calculation with new low rate and account the bios
dispatched at high rate. To fix it, reset the slice upon rate change.
https://lkml.org/lkml/2011/3/10/298
Another problem with very high limit is that we never queued the
bio on throtl service tree. That means we kept on extending the
group slice but never trimmed it. Fix that also by regulary
trimming the slice even if bio is not being queued up.
Reported-by: Lina Lu <lulina_nuaa@foxmail.com>
Signed-off-by: Vivek Goyal <vgoyal@redhat.com>
Signed-off-by: Jens Axboe <jaxboe@fusionio.com>
Use plug in throttle dispatch also as we are dispatching a bunch of
bios in throttle context and some of them might merge.
Signed-off-by: Vivek Goyal <vgoyal@redhat.com>
Signed-off-by: Jens Axboe <jaxboe@fusionio.com>
Code has been converted over to the new explicit on-stack plugging,
and delay users have been converted to use the new API for that.
So lets kill off the old plugging along with aops->sync_page().
Signed-off-by: Jens Axboe <jaxboe@fusionio.com>
When throttle group limits are updated through cgroups, a thread is
woken up to process these updates. While reviewing that code, oleg noted
couple of race conditions existed in the code and he also suggested that
code can be simplified.
This patch fixes the races simplifies the code based on Oleg's suggestions:
- Use xchg().
- Introduced a common function throtl_update_blkio_group_common()
which is shared now by all iops/bps update functions.
Reviewed-by: Oleg Nesterov <oleg@redhat.com>
Reviewed-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Signed-off-by: Vivek Goyal <vgoyal@redhat.com>
Fixed a merge issue, throtl_schedule_delayed_work() takes throtl_data
as the argument now, not the queue.
Signed-off-by: Jens Axboe <jaxboe@fusionio.com>
With the help of cgroup interface one can go and upate the bps/iops
limits of existing group. Once the limits are udpated, a thread is
woken up to see if some blocked group needs recalculation based on new
limits and needs to be requeued.
There was also a piece of code where I was checking for group limit
update when a fresh bio comes in. This patch gets rid of that piece of
code and keeps processing the limit change at one place
throtl_process_limit_change(). It just keeps the code simple and easy
to understand.
Signed-off-by: Vivek Goyal <vgoyal@redhat.com>
Signed-off-by: Jens Axboe <jaxboe@fusionio.com>
This merge creates two set of conflicts. One is simple context
conflicts caused by removal of throtl_scheduled_delayed_work() in
for-linus and removal of throtl_shutdown_timer_wq() in
for-2.6.39/core.
The other is caused by commit 255bb490c8 (block: blk-flush shouldn't
call directly into q->request_fn() __blk_run_queue()) in for-linus
crashing with FLUSH reimplementation in for-2.6.39/core. The conflict
isn't trivial but the resolution is straight-forward.
* __blk_run_queue() calls in flush_end_io() and flush_data_end_io()
should be called with @force_kblockd set to %true.
* elv_insert() in blk_kick_flush() should use
%ELEVATOR_INSERT_REQUEUE.
Both changes are to avoid invoking ->request_fn() directly from
request completion path and closely match the changes in the commit
255bb490c8.
Signed-off-by: Tejun Heo <tj@kernel.org>
Move blk_throtl_exit() in blk_cleanup_queue() as blk_throtl_exit() is
written in such a way that it needs queue lock. In blk_release_queue()
there is no gurantee that ->queue_lock is still around.
Initially blk_throtl_exit() was in blk_cleanup_queue() but Ingo reported
one problem.
https://lkml.org/lkml/2010/10/23/86
And a quick fix moved blk_throtl_exit() to blk_release_queue().
commit 7ad58c0286
Author: Jens Axboe <jaxboe@fusionio.com>
Date: Sat Oct 23 20:40:26 2010 +0200
block: fix use-after-free bug in blk throttle code
This patch reverts above change and does not try to shutdown the
throtl work in blk_sync_queue(). By avoiding call to
throtl_shutdown_timer_wq() from blk_sync_queue(), we should also avoid
the problem reported by Ingo.
blk_sync_queue() seems to be used only by md driver and it seems to be
using it to make sure q->unplug_fn is not called as md registers its
own unplug functions and it is about to free up the data structures
used by unplug_fn(). Block throttle does not call back into unplug_fn()
or into md. So there is no need to cancel blk throttle work.
In fact I think cancelling block throttle work is bad because it might
happen that some bios are throttled and scheduled to be dispatched later
with the help of pending work and if work is cancelled, these bios might
never be dispatched.
Block layer also uses blk_sync_queue() during blk_cleanup_queue() and
blk_release_queue() time. That should be safe as we are also calling
blk_throtl_exit() which should make sure all the throttling related
data structures are cleaned up.
Signed-off-by: Vivek Goyal <vgoyal@redhat.com>
Signed-off-by: Jens Axboe <jaxboe@fusionio.com>
o Dominik Klein reported a system hang issue while doing some blkio
throttling testing.
https://lkml.org/lkml/2011/2/24/173
o Some tracing revealed that CFQ was not dispatching any more jobs as
queue unplug was not happening. And queue unplug was not happening
because unplug work was not being called as there was one throttling
work on same cpu which as not finished yet. And throttling work had not
finished as it was tyring to dispatch a bio to CFQ but all the request
descriptors were consume to it was put to sleep.
o So basically it is a cyclic dependecny between CFQ unplug work and
throtl dispatch work. Tejun suggested that use separate workqueue for
such cases.
o This patch uses a separate workqueue for throttle related work and
does not rely on kblockd workqueue anymore.
Cc: stable@kernel.org
Reported-by: Dominik Klein <dk@in-telegence.net>
Signed-off-by: Vivek Goyal <vgoyal@redhat.com>
Acked-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Jens Axboe <jaxboe@fusionio.com>
o Jeff Moyer was doing some testing on a RAM backed disk and
blkiocg_lookup_group() showed up high overhead after memcpy(). Similarly
somebody else reported that blkiocg_lookup_group() is eating 6% extra
cpu. Though looking at the code I can't think why the overhead of
this function is so high. One thing is that it is called with very high
frequency (once for every IO).
o For lot of folks blkio controller will be compiled in but they might
not have actually created cgroups. Hence optimize the case of root
cgroup where we can avoid calling blkiocg_lookup_group() if IO is happening
in root group (common case).
Reported-by: Jeff Moyer <jmoyer@redhat.com>
Signed-off-by: Vivek Goyal <vgoyal@redhat.com>
Acked-by: Jeff Moyer <jmoyer@redhat.com>
Signed-off-by: Jens Axboe <jaxboe@fusionio.com>
o I was discussing what are the variable being updated without spin lock and
why do we need barriers and Oleg pointed out that location of smp_rmb()
should be between read of td->limits_changed and tg->limits_changed. This
patch fixes it.
o Following is one possible sequence of events. Say cpu0 is executing
throtl_update_blkio_group_read_bps() and cpu1 is executing
throtl_process_limit_change().
cpu0 cpu1
tg->limits_changed = true;
smp_mb__before_atomic_inc();
atomic_inc(&td->limits_changed);
if (!atomic_read(&td->limits_changed))
return;
if (tg->limits_changed)
do_something;
If cpu0 has updated tg->limits_changed and td->limits_changed, we want to
make sure that if update to td->limits_changed is visible on cpu1, then
update to tg->limits_changed should also be visible.
Oleg pointed out to ensure that we need to insert an smp_rmb() between
td->limits_changed read and tg->limits_changed read.
o I had erroneously put smp_rmb() before atomic_read(&td->limits_changed).
This patch fixes it.
Reported-by: Oleg Nesterov <oleg@redhat.com>
Signed-off-by: Vivek Goyal <vgoyal@redhat.com>
Signed-off-by: Jens Axboe <jaxboe@fusionio.com>
o During some testing I did following and noticed throttling stops working.
- Put a very low limit on a cgroup, say 1 byte per second.
- Start some reads, this will set slice_end to a very high value.
- Change the limit to higher value say 1MB/s
- Now IO unthrottles and finishes as expected.
- Try to do the read again but IO is not limited to 1MB/s as expected.
o What is happening.
- Initially low value of limit sets slice_end to a very high value.
- During updation of limit, slice_end is not being truncated.
- Very high value of slice_end leads to keeping the existing slice
valid for a very long time and new slice does not start.
- tg_may_dispatch() is called in blk_throtle_bio(), and trim_slice()
is not called in this path. So slice_start is some old value and
practically we are able to do huge amount of IO.
o There are many ways it can be fixed. I have fixed it by trying to
adjust/cleanup slice_end in trim_slice(). Generally we extend slices if bio
is big and can't be dispatched in one slice. After dispatch of bio, readjust
the slice_end to make sure we don't end up with huge values.
Signed-off-by: Vivek Goyal <vgoyal@redhat.com>
Signed-off-by: Jens Axboe <jaxboe@fusionio.com>
o Currently we try to dispatch more READS and less WRITES (75%, 25%) in one
dispatch round. ummy pointed out that there is a bug in max_nr_writes
calculation. This patch fixes it.
Reported-by: ummy y <yummylln@yahoo.com.cn>
Signed-off-by: Vivek Goyal <vgoyal@redhat.com>
Signed-off-by: Jens Axboe <jaxboe@fusionio.com>
o User can specify max iops value of 32bit (UINT_MAX), through cgroup
interface. If a user has specified say 4294967294 (UNIT_MAX - 2), then
on 32bit platform, following multiplication can overflow.
io_allowed = (tg->iops[rw] * jiffy_elapsed_rnd)
o Explicitly cast the multiplication to 64bit and then perform division and
then check whether result is still great then UNINT_MAX.
Signed-off-by: Vivek Goyal <vgoyal@redhat.com>
Signed-off-by: Jens Axboe <jaxboe@fusionio.com>
o Do not convert jiffies to mili seconds as it is not required. Just work
with jiffies and HZ.
Signed-off-by: Vivek Goyal <vgoyal@redhat.com>
Signed-off-by: Jens Axboe <jaxboe@fusionio.com>
o Randy Dunlap reported following linux-next failure. This patch fixes it.
on i386:
blk-throttle.c:(.text+0x1abb8): undefined reference to `__udivdi3'
blk-throttle.c:(.text+0x1b1dc): undefined reference to `__udivdi3'
o bytes_per_second interface is 64bit and I was continuing to do 64 bit
division even on 32bit platform without help of special macros/functions
hence the failure.
Signed-off-by: Vivek Goyal <vgoyal@redhat.com>
Reported-by: Randy Dunlap <randy.dunlap@oracle.com>
Signed-off-by: Jens Axboe <jaxboe@fusionio.com>
o Currently any cgroup throttle limit changes are processed asynchronousy and
the change does not take affect till a new bio is dispatched from same group.
o It might happen that a user sets a redicuously low limit on throttling.
Say 1 bytes per second on reads. In such cases simple operations like mount
a disk can wait for a very long time.
o Once bio is throttled, there is no easy way to come out of that wait even if
user increases the read limit later.
o This patch fixes it. Now if a user changes the cgroup limits, we recalculate
the bio dispatch time according to new limits.
o Can't take queueu lock under blkcg_lock, hence after the change I wake
up the dispatch thread again which recalculates the time. So there are some
variables being synchronized across two threads without lock and I had to
make use of barriers. Hoping I have used barriers correctly. Any review of
memory barrier code especially will help.
Signed-off-by: Vivek Goyal <vgoyal@redhat.com>
Signed-off-by: Jens Axboe <jaxboe@fusionio.com>
o Currently all the dynamically allocated groups, except root grp is added
to td->tg_list. This was not a problem so far but in next patch I will
travel through td->tg_list to process any updates of limits on the group.
If root group is not in tg_list, then root group's updates are not
processed.
o It is better to root group also to tg_list instead of doing special
processing for it during limit updates.
Signed-off-by: Vivek Goyal <vgoyal@redhat.com>
Signed-off-by: Jens Axboe <jaxboe@fusionio.com>
o Actual implementation of throttling policy in block layer. Currently it
implements READ and WRITE bytes per second throttling logic. IOPS throttling
comes in later patches.
Signed-off-by: Vivek Goyal <vgoyal@redhat.com>
Signed-off-by: Jens Axboe <jaxboe@fusionio.com>
