Compensate the task's think time when computing the final pause time,
so that ->dirty_ratelimit can be executed accurately.
think time := time spend outside of balance_dirty_pages()
In the rare case that the task slept longer than the 200ms period time
(result in negative pause time), the sleep time will be compensated in
the following periods, too, if it's less than 1 second.
Accumulated errors are carefully avoided as long as the max pause area
is not hitted.
Pseudo code:
period = pages_dirtied / task_ratelimit;
think = jiffies - dirty_paused_when;
pause = period - think;
1) normal case: period > think
pause = period - think
dirty_paused_when = jiffies + pause
nr_dirtied = 0
period time
|===============================>|
think time pause time
|===============>|==============>|
------|----------------|---------------|------------------------
dirty_paused_when jiffies
2) no pause case: period <= think
don't pause; reduce future pause time by:
dirty_paused_when += period
nr_dirtied = 0
period time
|===============================>|
think time
|===================================================>|
------|--------------------------------+-------------------|----
dirty_paused_when jiffies
Acked-by: Jan Kara <jack@suse.cz>
Acked-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Signed-off-by: Wu Fengguang <fengguang.wu@intel.com>
This makes the binary trace understandable by trace-cmd.
CC: Dave Chinner <david@fromorbit.com>
CC: Curt Wohlgemuth <curtw@google.com>
CC: Steven Rostedt <rostedt@goodmis.org>
Signed-off-by: Wu Fengguang <fengguang.wu@intel.com>
This creates a new 'reason' field in a wb_writeback_work
structure, which unambiguously identifies who initiates
writeback activity. A 'wb_reason' enumeration has been
added to writeback.h, to enumerate the possible reasons.
The 'writeback_work_class' and tracepoint event class and
'writeback_queue_io' tracepoints are updated to include the
symbolic 'reason' in all trace events.
And the 'writeback_inodes_sbXXX' family of routines has had
a wb_stats parameter added to them, so callers can specify
why writeback is being started.
Acked-by: Jan Kara <jack@suse.cz>
Signed-off-by: Curt Wohlgemuth <curtw@google.com>
Signed-off-by: Wu Fengguang <fengguang.wu@intel.com>
Instead of sending ->older_than_this to queue_io() and
move_expired_inodes(), send the entire wb_writeback_work
structure. There are other fields of a work item that are
useful in these routines and in tracepoints.
Acked-by: Jan Kara <jack@suse.cz>
Signed-off-by: Curt Wohlgemuth <curtw@google.com>
Signed-off-by: Wu Fengguang <fengguang.wu@intel.com>
Useful for analyzing the dynamics of the throttling algorithms and
debugging user reported problems.
Signed-off-by: Wu Fengguang <fengguang.wu@intel.com>
As proposed by Chris, Dave and Jan, don't start foreground writeback IO
inside balance_dirty_pages(). Instead, simply let it idle sleep for some
time to throttle the dirtying task. In the mean while, kick off the
per-bdi flusher thread to do background writeback IO.
RATIONALS
=========
- disk seeks on concurrent writeback of multiple inodes (Dave Chinner)
If every thread doing writes and being throttled start foreground
writeback, it leads to N IO submitters from at least N different
inodes at the same time, end up with N different sets of IO being
issued with potentially zero locality to each other, resulting in
much lower elevator sort/merge efficiency and hence we seek the disk
all over the place to service the different sets of IO.
OTOH, if there is only one submission thread, it doesn't jump between
inodes in the same way when congestion clears - it keeps writing to
the same inode, resulting in large related chunks of sequential IOs
being issued to the disk. This is more efficient than the above
foreground writeback because the elevator works better and the disk
seeks less.
- lock contention and cache bouncing on concurrent IO submitters (Dave Chinner)
With this patchset, the fs_mark benchmark on a 12-drive software RAID0 goes
from CPU bound to IO bound, freeing "3-4 CPUs worth of spinlock contention".
* "CPU usage has dropped by ~55%", "it certainly appears that most of
the CPU time saving comes from the removal of contention on the
inode_wb_list_lock" (IMHO at least 10% comes from the reduction of
cacheline bouncing, because the new code is able to call much less
frequently into balance_dirty_pages() and hence access the global
page states)
* the user space "App overhead" is reduced by 20%, by avoiding the
cacheline pollution by the complex writeback code path
* "for a ~5% throughput reduction", "the number of write IOs have
dropped by ~25%", and the elapsed time reduced from 41:42.17 to
40:53.23.
* On a simple test of 100 dd, it reduces the CPU %system time from 30% to 3%,
and improves IO throughput from 38MB/s to 42MB/s.
- IO size too small for fast arrays and too large for slow USB sticks
The write_chunk used by current balance_dirty_pages() cannot be
directly set to some large value (eg. 128MB) for better IO efficiency.
Because it could lead to more than 1 second user perceivable stalls.
Even the current 4MB write size may be too large for slow USB sticks.
The fact that balance_dirty_pages() starts IO on itself couples the
IO size to wait time, which makes it hard to do suitable IO size while
keeping the wait time under control.
Now it's possible to increase writeback chunk size proportional to the
disk bandwidth. In a simple test of 50 dd's on XFS, 1-HDD, 3GB ram,
the larger writeback size dramatically reduces the seek count to 1/10
(far beyond my expectation) and improves the write throughput by 24%.
- long block time in balance_dirty_pages() hurts desktop responsiveness
Many of us may have the experience: it often takes a couple of seconds
or even long time to stop a heavy writing dd/cp/tar command with
Ctrl-C or "kill -9".
- IO pipeline broken by bumpy write() progress
There are a broad class of "loop {read(buf); write(buf);}" applications
whose read() pipeline will be under-utilized or even come to a stop if
the write()s have long latencies _or_ don't progress in a constant rate.
The current threshold based throttling inherently transfers the large
low level IO completion fluctuations to bumpy application write()s,
and further deteriorates with increasing number of dirtiers and/or bdi's.
For example, when doing 50 dd's + 1 remote rsync to an XFS partition,
the rsync progresses very bumpy in legacy kernel, and throughput is
improved by 67% by this patchset. (plus the larger write chunk size,
it will be 93% speedup).
The new rate based throttling can support 1000+ dd's with excellent
smoothness, low latency and low overheads.
For the above reasons, it's much better to do IO-less and low latency
pauses in balance_dirty_pages().
Jan Kara, Dave Chinner and me explored the scheme to let
balance_dirty_pages() wait for enough writeback IO completions to
safeguard the dirty limit. However it's found to have two problems:
- in large NUMA systems, the per-cpu counters may have big accounting
errors, leading to big throttle wait time and jitters.
- NFS may kill large amount of unstable pages with one single COMMIT.
Because NFS server serves COMMIT with expensive fsync() IOs, it is
desirable to delay and reduce the number of COMMITs. So it's not
likely to optimize away such kind of bursty IO completions, and the
resulted large (and tiny) stall times in IO completion based throttling.
So here is a pause time oriented approach, which tries to control the
pause time in each balance_dirty_pages() invocations, by controlling
the number of pages dirtied before calling balance_dirty_pages(), for
smooth and efficient dirty throttling:
- avoid useless (eg. zero pause time) balance_dirty_pages() calls
- avoid too small pause time (less than 4ms, which burns CPU power)
- avoid too large pause time (more than 200ms, which hurts responsiveness)
- avoid big fluctuations of pause times
It can control pause times at will. The default policy (in a followup
patch) will be to do ~10ms pauses in 1-dd case, and increase to ~100ms
in 1000-dd case.
BEHAVIOR CHANGE
===============
(1) dirty threshold
Users will notice that the applications will get throttled once crossing
the global (background + dirty)/2=15% threshold, and then balanced around
17.5%. Before patch, the behavior is to just throttle it at 20% dirtyable
memory in 1-dd case.
Since the task will be soft throttled earlier than before, it may be
perceived by end users as performance "slow down" if his application
happens to dirty more than 15% dirtyable memory.
(2) smoothness/responsiveness
Users will notice a more responsive system during heavy writeback.
"killall dd" will take effect instantly.
Signed-off-by: Wu Fengguang <fengguang.wu@intel.com>
Save inode->dirtied_when in the raw trace output for reliable scripting,
and to also show in formatted output the relative age in seconds for
easy human reading.
CC: Jan Kara <jack@suse.cz>
Acked-by: Christoph Hellwig <hch@infradead.org>
Signed-off-by: Wu Fengguang <fengguang.wu@intel.com>
Add trace event balance_dirty_state for showing the global dirty page
counts and thresholds at each global_dirty_limits() invocation. This
will cover the callers throttle_vm_writeout(), over_bground_thresh()
and each balance_dirty_pages() loop.
Signed-off-by: Wu Fengguang <fengguang.wu@intel.com>
Pass struct wb_writeback_work all the way down to writeback_sb_inodes(),
and initialize the struct writeback_control there.
struct writeback_control is basically designed to control writeback of a
single file, but we keep abuse it for writing multiple files in
writeback_sb_inodes() and its callers.
It immediately clean things up, e.g. suddenly wbc.nr_to_write vs
work->nr_pages starts to make sense, and instead of saving and restoring
pages_skipped in writeback_sb_inodes it can always start with a clean
zero value.
It also makes a neat IO pattern change: large dirty files are now
written in the full 4MB writeback chunk size, rather than whatever
remained quota in wbc->nr_to_write.
Acked-by: Jan Kara <jack@suse.cz>
Proposed-by: Christoph Hellwig <hch@infradead.org>
Signed-off-by: Wu Fengguang <fengguang.wu@intel.com>
Note that it adds a little overheads to account the moved/enqueued
inodes from b_dirty to b_io. The "moved" accounting may be later used to
limit the number of inodes that can be moved in one shot, in order to
keep spinlock hold time under control.
Signed-off-by: Wu Fengguang <fengguang.wu@intel.com>
It is valuable to know how the dirty inodes are iterated and their IO size.
"writeback_single_inode: bdi 8:0: ino=134246746 state=I_DIRTY_SYNC|I_SYNC age=414 index=0 to_write=1024 wrote=0"
- "state" reflects inode->i_state at the end of writeback_single_inode()
- "index" reflects mapping->writeback_index after the ->writepages() call
- "to_write" is the wbc->nr_to_write at entrance of writeback_single_inode()
- "wrote" is the number of pages actually written
v2: add trace event writeback_single_inode_requeue as proposed by Dave.
CC: Dave Chinner <david@fromorbit.com>
Signed-off-by: Wu Fengguang <fengguang.wu@intel.com>
When wbc.more_io was first introduced, it indicates whether there are
at least one superblock whose s_more_io contains more IO work. Now with
the per-bdi writeback, it can be replaced with a simple b_more_io test.
Acked-by: Jan Kara <jack@suse.cz>
Acked-by: Mel Gorman <mel@csn.ul.ie>
Reviewed-by: Minchan Kim <minchan.kim@gmail.com>
Signed-off-by: Wu Fengguang <fengguang.wu@intel.com>
This tracks when balance_dirty_pages() tries to wakeup the flusher thread
for background writeback (if it was not started already).
Suggested-by: Christoph Hellwig <hch@infradead.org>
Signed-off-by: Wu Fengguang <fengguang.wu@intel.com>
Cc: Jan Kara <jack@suse.cz>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Dave Chinner <david@fromorbit.com>
Cc: Jan Engelhardt <jengelh@medozas.de>
Cc: Jens Axboe <axboe@kernel.dk>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
If congestion_wait() is called with no BDI congested, the caller will
sleep for the full timeout and this may be an unnecessary sleep. This
patch adds a wait_iff_congested() that checks congestion and only sleeps
if a BDI is congested else, it calls cond_resched() to ensure the caller
is not hogging the CPU longer than its quota but otherwise will not sleep.
This is aimed at reducing some of the major desktop stalls reported during
IO. For example, while kswapd is operating, it calls congestion_wait()
but it could just have been reclaiming clean page cache pages with no
congestion. Without this patch, it would sleep for a full timeout but
after this patch, it'll just call schedule() if it has been on the CPU too
long. Similar logic applies to direct reclaimers that are not making
enough progress.
Signed-off-by: Mel Gorman <mel@csn.ul.ie>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Minchan Kim <minchan.kim@gmail.com>
Cc: Wu Fengguang <fengguang.wu@intel.com>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Cc: Rik van Riel <riel@redhat.com>
Cc: Jens Axboe <axboe@kernel.dk>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
There is strong evidence to indicate a lot of time is being spent in
congestion_wait(), some of it unnecessarily. This patch adds a tracepoint
for congestion_wait to record when congestion_wait() was called, how long
the timeout was for and how long it actually slept.
Signed-off-by: Mel Gorman <mel@csn.ul.ie>
Reviewed-by: Minchan Kim <minchan.kim@gmail.com>
Reviewed-by: Johannes Weiner <hannes@cmpxchg.org>
Cc: Wu Fengguang <fengguang.wu@intel.com>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Cc: Rik van Riel <riel@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
This removes more dead code that was somehow missed by commit 0d99519efe
(writeback: remove unused nonblocking and congestion checks). There are
no behavior change except for the removal of two entries from one of the
ext4 tracing interface.
The nonblocking checks in ->writepages are no longer used because the
flusher now prefer to block on get_request_wait() than to skip inodes on
IO congestion. The latter will lead to more seeky IO.
The nonblocking checks in ->writepage are no longer used because it's
redundant with the WB_SYNC_NONE check.
We no long set ->nonblocking in VM page out and page migration, because
a) it's effectively redundant with WB_SYNC_NONE in current code
b) it's old semantic of "Don't get stuck on request queues" is mis-behavior:
that would skip some dirty inodes on congestion and page out others, which
is unfair in terms of LRU age.
Inspired by Christoph Hellwig. Thanks!
Signed-off-by: Wu Fengguang <fengguang.wu@intel.com>
Cc: Theodore Ts'o <tytso@mit.edu>
Cc: David Howells <dhowells@redhat.com>
Cc: Sage Weil <sage@newdream.net>
Cc: Steve French <sfrench@samba.org>
Cc: Chris Mason <chris.mason@oracle.com>
Cc: Jens Axboe <axboe@kernel.dk>
Cc: Christoph Hellwig <hch@infradead.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Add 2 new trace points to the periodic write-back wake up case, just like we do
in the 'bdi_queue_work()' function. Namely, introduce:
1. trace_writeback_wake_thread(bdi)
2. trace_writeback_wake_forker_thread(bdi)
The first event is triggered every time we wake up a bdi thread to start
periodic background write-out. The second event is triggered only when the bdi
thread does not exist and should be created by the forker thread.
This patch was suggested by Dave Chinner and Christoph Hellwig.
Signed-off-by: Artem Bityutskiy <Artem.Bityutskiy@nokia.com>
Signed-off-by: Jens Axboe <jaxboe@fusionio.com>
include/trace/events/writeback.h uses dev_name(), so it needs to
include linux/device.h.
include/trace/events/writeback.h:12: error: implicit declaration of function 'dev_name'
Signed-off-by: Randy Dunlap <randy.dunlap@oracle.com>
Signed-off-by: Jens Axboe <jaxboe@fusionio.com>
Add a trace event to the ->writepage loop in write_cache_pages to give
visibility into how the ->writepage call is changing variables within the
writeback control structure. Of most interest is how wbc->nr_to_write changes
from call to call, especially with filesystems that write multiple pages
in ->writepage.
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Jens Axboe <jaxboe@fusionio.com>
Tracing high level background writeback events is good, but it doesn't
give the entire picture. Add visibility into write throttling to catch IO
dispatched by foreground throttling of processing dirtying lots of pages.
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Signed-off-by: Jens Axboe <jaxboe@fusionio.com>
Trace queue/sched/exec parts of the writeback loop. This provides
insight into when and why flusher threads are scheduled to run. e.g
a sync invocation leaves traces like:
sync-[...]: writeback_queue: bdi 8:0: sb_dev 8:1 nr_pages=7712 sync_mode=0 kupdate=0 range_cyclic=0 background=0
flush-8:0-[...]: writeback_exec: bdi 8:0: sb_dev 8:1 nr_pages=7712 sync_mode=0 kupdate=0 range_cyclic=0 background=0
This also lays the foundation for adding more writeback tracing to
provide deeper insight into the whole writeback path.
The original tracing code is from Jens Axboe, though this version is
a rewrite as a result of the code being traced changing
significantly.
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Signed-off-by: Jens Axboe <jaxboe@fusionio.com>
