The amount of available memory to a memcg wb_domain can change as
memcg configuration changes. A domain's ->dirty_limit exists to
smooth out sudden drops in dirty threshold; however, when a domain's
size actually drops significantly, it hinders the dirty throttling
from adjusting to the new configuration leading to unexpected
behaviors including unnecessary OOM kills.
This patch resolves the issue by adding wb_domain_size_changed() which
resets ->dirty_limit[_tstmp] and making memcg call it on configuration
changes.
Signed-off-by: Tejun Heo <tj@kernel.org>
Cc: Jens Axboe <axboe@kernel.dk>
Cc: Jan Kara <jack@suse.cz>
Cc: Wu Fengguang <fengguang.wu@intel.com>
Cc: Greg Thelen <gthelen@google.com>
Signed-off-by: Jens Axboe <axboe@fb.com>
Dirtyable memory is distributed to a wb (bdi_writeback) according to
the relative bandwidth the wb is writing out in the whole system.
This distribution is global - each wb is measured against all other
wb's and gets the proportinately sized portion of the memory in the
whole system.
For cgroup writeback, the amount of dirtyable memory is scoped by
memcg and thus each wb would need to be measured and controlled in its
memcg. IOW, a wb will belong to two writeback domains - the global
and memcg domains.
The previous patches laid the groundwork to support the two wb_domains
and this patch implements memcg wb_domain. memcg->cgwb_domain is
initialized on css online and destroyed on css release,
wb->memcg_completions is added, and __wb_writeout_inc() is updated to
increment completions against both global and memcg wb_domains.
The following patches will update balance_dirty_pages() and its
subroutines to actually consider memcg wb_domain for throttling.
Signed-off-by: Tejun Heo <tj@kernel.org>
Cc: Jens Axboe <axboe@kernel.dk>
Cc: Jan Kara <jack@suse.cz>
Cc: Wu Fengguang <fengguang.wu@intel.com>
Cc: Greg Thelen <gthelen@google.com>
Signed-off-by: Jens Axboe <axboe@fb.com>
and rename it to wb_over_bg_thresh(). The function is closely tied to
the dirty throttling mechanism implemented in page-writeback.c. This
relocation will allow future updates necessary for cgroup writeback
support.
While at it, add function comment.
This is pure reorganization and doesn't introduce any behavioral
changes.
Signed-off-by: Tejun Heo <tj@kernel.org>
Cc: Jens Axboe <axboe@kernel.dk>
Cc: Jan Kara <jack@suse.cz>
Cc: Wu Fengguang <fengguang.wu@intel.com>
Cc: Greg Thelen <gthelen@google.com>
Signed-off-by: Jens Axboe <axboe@fb.com>
This patch is a part of the series to define wb_domain which
represents a domain that wb's (bdi_writeback's) belong to and are
measured against each other in. This will enable IO backpressure
propagation for cgroup writeback.
global_dirty_limit exists to regulate the global dirty threshold which
is a property of the wb_domain. This patch moves hard_dirty_limit,
dirty_lock, and update_time into wb_domain.
This is pure reorganization and doesn't introduce any behavioral
changes.
Signed-off-by: Tejun Heo <tj@kernel.org>
Cc: Jens Axboe <axboe@kernel.dk>
Cc: Jan Kara <jack@suse.cz>
Cc: Wu Fengguang <fengguang.wu@intel.com>
Cc: Greg Thelen <gthelen@google.com>
Signed-off-by: Jens Axboe <axboe@fb.com>
Dirtyable memory is distributed to a wb (bdi_writeback) according to
the relative bandwidth the wb is writing out in the whole system.
This distribution is global - each wb is measured against all other
wb's and gets the proportinately sized portion of the memory in the
whole system.
For cgroup writeback, the amount of dirtyable memory is scoped by
memcg and thus each wb would need to be measured and controlled in its
memcg. IOW, a wb will belong to two writeback domains - the global
and memcg domains.
Currently, what constitutes the global writeback domain are scattered
across a number of global states. This patch starts collecting them
into struct wb_domain.
* fprop_global which serves as the basis for proportional bandwidth
measurement and its period timer are moved into struct wb_domain.
* global_wb_domain hosts the states for the global domain.
* While at it, flatten wb_writeout_fraction() into its callers. This
thin wrapper doesn't provide any actual benefits while getting in
the way.
This is pure reorganization and doesn't introduce any behavioral
changes.
Signed-off-by: Tejun Heo <tj@kernel.org>
Cc: Jens Axboe <axboe@kernel.dk>
Cc: Jan Kara <jack@suse.cz>
Cc: Wu Fengguang <fengguang.wu@intel.com>
Cc: Greg Thelen <gthelen@google.com>
Signed-off-by: Jens Axboe <axboe@fb.com>
__wb_update_bandwidth() is called from two places -
fs/fs-writeback.c::balance_dirty_pages() and
mm/page-writeback.c::wb_writeback(). The latter updates only the
write bandwidth while the former also deals with the dirty ratelimit.
The two callsites are distinguished by whether @thresh parameter is
zero or not, which is cryptic. In addition, the two files define
their own different versions of wb_update_bandwidth() on top of
__wb_update_bandwidth(), which is confusing to say the least. This
patch cleans up [__]wb_update_bandwidth() in the following ways.
* __wb_update_bandwidth() now takes explicit @update_ratelimit
parameter to gate dirty ratelimit handling.
* mm/page-writeback.c::wb_update_bandwidth() is flattened into its
caller - balance_dirty_pages().
* fs/fs-writeback.c::wb_update_bandwidth() is moved to
mm/page-writeback.c and __wb_update_bandwidth() is made static.
* While at it, add a lockdep assertion to __wb_update_bandwidth().
Except for the lockdep addition, this is pure reorganization and
doesn't introduce any behavioral changes.
Signed-off-by: Tejun Heo <tj@kernel.org>
Cc: Jens Axboe <axboe@kernel.dk>
Cc: Jan Kara <jack@suse.cz>
Cc: Wu Fengguang <fengguang.wu@intel.com>
Cc: Greg Thelen <gthelen@google.com>
Signed-off-by: Jens Axboe <axboe@fb.com>
The function name wb_dirty_limit(), its argument @dirty and the local
variable @wb_dirty are mortally confusing given that the function
calculates per-wb threshold value not dirty pages, especially given
that @dirty and @wb_dirty are used elsewhere for dirty pages.
Let's rename the function to wb_calc_thresh() and wb_dirty to
wb_thresh.
Signed-off-by: Tejun Heo <tj@kernel.org>
Cc: Jens Axboe <axboe@kernel.dk>
Cc: Jan Kara <jack@suse.cz>
Cc: Wu Fengguang <fengguang.wu@intel.com>
Cc: Greg Thelen <gthelen@google.com>
Signed-off-by: Jens Axboe <axboe@fb.com>
try_writeback_inodes_sb_nr() wraps writeback_inodes_sb_nr() so that it
handles s_umount locking and skips if writeback is already in
progress. The in progress test is performed on the root wb
(bdi_writeback) which isn't sufficient for cgroup writeback support.
The test must be done per-wb.
To prepare for the change, this patch factors out
__writeback_inodes_sb_nr() from writeback_inodes_sb_nr() and adds
@skip_if_busy and moves the in progress test right before queueing the
wb_writeback_work. try_writeback_inodes_sb_nr() now just grabs
s_umount and invokes __writeback_inodes_sb_nr() with asserted
@skip_if_busy. This way, later addition of multiple wb handling can
skip only the wb's which already have writeback in progress.
This swaps the order between in progress test and s_umount test which
can flip the return value when writeback is in progress and s_umount
is being held by someone else but this shouldn't cause any meaningful
difference. It's a fringe condition and the return value is an
unsynchronized hint anyway.
Signed-off-by: Tejun Heo <tj@kernel.org>
Cc: Jens Axboe <axboe@kernel.dk>
Cc: Jan Kara <jack@suse.cz>
Signed-off-by: Jens Axboe <axboe@fb.com>
Currently, a bdi (backing_dev_info) embeds single wb (bdi_writeback)
and the role of the separation is unclear. For cgroup support for
writeback IOs, a bdi will be updated to host multiple wb's where each
wb serves writeback IOs of a different cgroup on the bdi. To achieve
that, a wb should carry all states necessary for servicing writeback
IOs for a cgroup independently.
This patch moves bandwidth related fields from backing_dev_info into
bdi_writeback.
* The moved fields are: bw_time_stamp, dirtied_stamp, written_stamp,
write_bandwidth, avg_write_bandwidth, dirty_ratelimit,
balanced_dirty_ratelimit, completions and dirty_exceeded.
* writeback_chunk_size() and over_bground_thresh() now take @wb
instead of @bdi.
* bdi_writeout_fraction(bdi, ...) -> wb_writeout_fraction(wb, ...)
bdi_dirty_limit(bdi, ...) -> wb_dirty_limit(wb, ...)
bdi_position_ration(bdi, ...) -> wb_position_ratio(wb, ...)
bdi_update_writebandwidth(bdi, ...) -> wb_update_write_bandwidth(wb, ...)
[__]bdi_update_bandwidth(bdi, ...) -> [__]wb_update_bandwidth(wb, ...)
bdi_{max|min}_pause(bdi, ...) -> wb_{max|min}_pause(wb, ...)
bdi_dirty_limits(bdi, ...) -> wb_dirty_limits(wb, ...)
* Init/exits of the relocated fields are moved to bdi_wb_init/exit()
respectively. Note that explicit zeroing is dropped in the process
as wb's are cleared in entirety anyway.
* As there's still only one bdi_writeback per backing_dev_info, all
uses of bdi->stat[] are mechanically replaced with bdi->wb.stat[]
introducing no behavior changes.
v2: Typo in description fixed as suggested by Jan.
Signed-off-by: Tejun Heo <tj@kernel.org>
Reviewed-by: Jan Kara <jack@suse.cz>
Cc: Jens Axboe <axboe@kernel.dk>
Cc: Wu Fengguang <fengguang.wu@intel.com>
Cc: Jaegeuk Kim <jaegeuk@kernel.org>
Cc: Steven Whitehouse <swhiteho@redhat.com>
Signed-off-by: Jens Axboe <axboe@fb.com>
Add a tuning knob so we can adjust the dirtytime expiration timeout,
which is very useful for testing lazytime.
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Reviewed-by: Jan Kara <jack@suse.cz>
Tejun, while reviewing the code, spotted the following race condition
between the dirtying and truncation of a page:
__set_page_dirty_nobuffers() __delete_from_page_cache()
if (TestSetPageDirty(page))
page->mapping = NULL
if (PageDirty())
dec_zone_page_state(page, NR_FILE_DIRTY);
dec_bdi_stat(mapping->backing_dev_info, BDI_RECLAIMABLE);
if (page->mapping)
account_page_dirtied(page)
__inc_zone_page_state(page, NR_FILE_DIRTY);
__inc_bdi_stat(mapping->backing_dev_info, BDI_RECLAIMABLE);
which results in an imbalance of NR_FILE_DIRTY and BDI_RECLAIMABLE.
Dirtiers usually lock out truncation, either by holding the page lock
directly, or in case of zap_pte_range(), by pinning the mapcount with
the page table lock held. The notable exception to this rule, though,
is do_wp_page(), for which this race exists. However, do_wp_page()
already waits for a locked page to unlock before setting the dirty bit,
in order to prevent a race where clear_page_dirty() misses the page bit
in the presence of dirty ptes. Upgrade that wait to a fully locked
set_page_dirty() to also cover the situation explained above.
Afterwards, the code in set_page_dirty() dealing with a truncation race
is no longer needed. Remove it.
Reported-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Johannes Weiner <hannes@cmpxchg.org>
Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Reviewed-by: Jan Kara <jack@suse.cz>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
The current "wait_on_bit" interface requires an 'action'
function to be provided which does the actual waiting.
There are over 20 such functions, many of them identical.
Most cases can be satisfied by one of just two functions, one
which uses io_schedule() and one which just uses schedule().
So:
Rename wait_on_bit and wait_on_bit_lock to
wait_on_bit_action and wait_on_bit_lock_action
to make it explicit that they need an action function.
Introduce new wait_on_bit{,_lock} and wait_on_bit{,_lock}_io
which are *not* given an action function but implicitly use
a standard one.
The decision to error-out if a signal is pending is now made
based on the 'mode' argument rather than being encoded in the action
function.
All instances of the old wait_on_bit and wait_on_bit_lock which
can use the new version have been changed accordingly and their
action functions have been discarded.
wait_on_bit{_lock} does not return any specific error code in the
event of a signal so the caller must check for non-zero and
interpolate their own error code as appropriate.
The wait_on_bit() call in __fscache_wait_on_invalidate() was
ambiguous as it specified TASK_UNINTERRUPTIBLE but used
fscache_wait_bit_interruptible as an action function.
David Howells confirms this should be uniformly
"uninterruptible"
The main remaining user of wait_on_bit{,_lock}_action is NFS
which needs to use a freezer-aware schedule() call.
A comment in fs/gfs2/glock.c notes that having multiple 'action'
functions is useful as they display differently in the 'wchan'
field of 'ps'. (and /proc/$PID/wchan).
As the new bit_wait{,_io} functions are tagged "__sched", they
will not show up at all, but something higher in the stack. So
the distinction will still be visible, only with different
function names (gds2_glock_wait versus gfs2_glock_dq_wait in the
gfs2/glock.c case).
Since first version of this patch (against 3.15) two new action
functions appeared, on in NFS and one in CIFS. CIFS also now
uses an action function that makes the same freezer aware
schedule call as NFS.
Signed-off-by: NeilBrown <neilb@suse.de>
Acked-by: David Howells <dhowells@redhat.com> (fscache, keys)
Acked-by: Steven Whitehouse <swhiteho@redhat.com> (gfs2)
Acked-by: Peter Zijlstra <peterz@infradead.org>
Cc: Oleg Nesterov <oleg@redhat.com>
Cc: Steve French <sfrench@samba.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Link: http://lkml.kernel.org/r/20140707051603.28027.72349.stgit@notabene.brown
Signed-off-by: Ingo Molnar <mingo@kernel.org>
There's only one caller of set_page_dirty_balance() and that will call it
with page_mkwrite == 0.
The page_mkwrite argument was unused since commit b827e496c8 "mm: close
page_mkwrite races".
Signed-off-by: Miklos Szeredi <mszeredi@suse.cz>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
This reverts commit c4a391b53a. Dave
Chinner <david@fromorbit.com> has reported the commit may cause some
inodes to be left out from sync(2). This is because we can call
redirty_tail() for some inode (which sets i_dirtied_when to current time)
after sync(2) has started or similarly requeue_inode() can set
i_dirtied_when to current time if writeback had to skip some pages. The
real problem is in the functions clobbering i_dirtied_when but fixing
that isn't trivial so revert is a safer choice for now.
CC: stable@vger.kernel.org # >= 3.13
Signed-off-by: Jan Kara <jack@suse.cz>
When there are processes heavily creating small files while sync(2) is
running, it can easily happen that quite some new files are created
between WB_SYNC_NONE and WB_SYNC_ALL pass of sync(2). That can happen
especially if there are several busy filesystems (remember that sync
traverses filesystems sequentially and waits in WB_SYNC_ALL phase on one
fs before starting it on another fs). Because WB_SYNC_ALL pass is slow
(e.g. causes a transaction commit and cache flush for each inode in
ext3), resulting sync(2) times are rather large.
The following script reproduces the problem:
function run_writers
{
for (( i = 0; i < 10; i++ )); do
mkdir $1/dir$i
for (( j = 0; j < 40000; j++ )); do
dd if=/dev/zero of=$1/dir$i/$j bs=4k count=4 &>/dev/null
done &
done
}
for dir in "$@"; do
run_writers $dir
done
sleep 40
time sync
Fix the problem by disregarding inodes dirtied after sync(2) was called
in the WB_SYNC_ALL pass. To allow for this, sync_inodes_sb() now takes
a time stamp when sync has started which is used for setting up work for
flusher threads.
To give some numbers, when above script is run on two ext4 filesystems
on simple SATA drive, the average sync time from 10 runs is 267.549
seconds with standard deviation 104.799426. With the patched kernel,
the average sync time from 10 runs is 2.995 seconds with standard
deviation 0.096.
Signed-off-by: Jan Kara <jack@suse.cz>
Reviewed-by: Fengguang Wu <fengguang.wu@intel.com>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
It's not used globally and could be static.
Signed-off-by: Wanpeng Li <liwanp@linux.vnet.ibm.com>
Cc: Dave Hansen <dave.hansen@linux.intel.com>
Cc: Rik van Riel <riel@redhat.com>
Cc: Fengguang Wu <fengguang.wu@intel.com>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Tejun Heo <tj@kernel.org>
Cc: Yasuaki Ishimatsu <isimatu.yasuaki@jp.fujitsu.com>
Cc: David Rientjes <rientjes@google.com>
Cc: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Cc: Jiri Kosina <jkosina@suse.cz>
Cc: Wanpeng Li <liwanp@linux.vnet.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
After commit 839a8e8660 ("writeback: replace custom worker pool
implementation with unbound workqueue"), there is no bdi forker thread
any more. However, WB_REASON_FORKER_THREAD is still used due to it is
TPs userland visible and we won't be exposing exactly the same
information with just a different name.
Signed-off-by: Wanpeng Li <liwanp@linux.vnet.ibm.com>
Reviewed-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
wb_reason_name is not used any more - remove it.
Signed-off-by: Wanpeng Li <liwanp@linux.vnet.ibm.com>
Reviewed-by: Tejun Heo <tj@kernel.org>
Reviewed-by: Fengguang Wu <fengguang.wu@intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
It's not used globally and could be static.
Signed-off-by: Haicheng Li <haicheng.li@linux.intel.com>
Cc: Jan Kara <jack@suse.cz>
Cc: Wu Fengguang <fengguang.wu@intel.com>
Cc: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
When sync does it's WB_SYNC_ALL writeback, it issues data Io and
then immediately waits for IO completion. This is done in the
context of the flusher thread, and hence completely ties up the
flusher thread for the backing device until all the dirty inodes
have been synced. On filesystems that are dirtying inodes constantly
and quickly, this means the flusher thread can be tied up for
minutes per sync call and hence badly affect system level write IO
performance as the page cache cannot be cleaned quickly.
We already have a wait loop for IO completion for sync(2), so cut
this out of the flusher thread and delegate it to wait_sb_inodes().
Hence we can do rapid IO submission, and then wait for it all to
complete.
Effect of sync on fsmark before the patch:
FSUse% Count Size Files/sec App Overhead
.....
0 640000 4096 35154.6 1026984
0 720000 4096 36740.3 1023844
0 800000 4096 36184.6 916599
0 880000 4096 1282.7 1054367
0 960000 4096 3951.3 918773
0 1040000 4096 40646.2 996448
0 1120000 4096 43610.1 895647
0 1200000 4096 40333.1 921048
And a single sync pass took:
real 0m52.407s
user 0m0.000s
sys 0m0.090s
After the patch, there is no impact on fsmark results, and each
individual sync(2) operation run concurrently with the same fsmark
workload takes roughly 7s:
real 0m6.930s
user 0m0.000s
sys 0m0.039s
IOWs, sync is 7-8x faster on a busy filesystem and does not have an
adverse impact on ongoing async data write operations.
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Jan Kara <jack@suse.cz>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
writeback_inodes_sb(_nr)_if_idle() is re-implemented by replacing down_read()
with down_read_trylock() because
- If ->s_umount is write locked, then the sb is not idle. That is
writeback_inodes_sb(_nr)_if_idle() needn't wait for the lock.
- writeback_inodes_sb(_nr)_if_idle() grabs s_umount lock when it want to start
writeback, it may bring us deadlock problem when doing umount. In order to
fix the problem, ext4 and btrfs implemented their own writeback functions
instead of writeback_inodes_sb(_nr)_if_idle(), but it introduced the redundant
code, it is better to implement a new writeback_inodes_sb(_nr)_if_idle().
The name of these two functions is cumbersome, so rename them to
try_to_writeback_inodes_sb(_nr).
This idea came from Christoph Hellwig.
Some code is from the patch of Kamal Mostafa.
Reviewed-by: Jan Kara <jack@suse.cz>
Signed-off-by: Miao Xie <miaox@cn.fujitsu.com>
Signed-off-by: Fengguang Wu <fengguang.wu@intel.com>
There is no reason to pass the nr_pages_dirtied argument, because
nr_pages_dirtied value from the caller is unused in
balance_dirty_pages_ratelimited_nr().
Signed-off-by: Namjae Jeon <linkinjeon@gmail.com>
Signed-off-by: Vivek Trivedi <vtrivedi018@gmail.com>
Cc: Wu Fengguang <fengguang.wu@intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
The pdflush thread is long gone, so this patch removes references to pdflush
from vfs comments.
Signed-off-by: Artem Bityutskiy <artem.bityutskiy@linux.intel.com>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Since per-BDI flusher threads were introduced in 2.6, the pdflush
mechanism is not used any more. But the old interface exported through
/proc/sys/vm/nr_pdflush_threads still exists and is obviously useless.
For back-compatibility, printk warning information and return 2 to notify
the users that the interface is removed.
Signed-off-by: Wanpeng Li <liwp@linux.vnet.ibm.com>
Cc: Wu Fengguang <fengguang.wu@intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Doing iput() from flusher thread (writeback_sb_inodes()) can create problems
because iput() can do a lot of work - for example truncate the inode if it's
the last iput on unlinked file. Some filesystems depend on flusher thread
progressing (e.g. because they need to flush delay allocated blocks to reduce
allocation uncertainty) and so flusher thread doing truncate creates
interesting dependencies and possibilities for deadlocks.
We get rid of iput() in flusher thread by using the fact that I_SYNC inode
flag effectively pins the inode in memory. So if we take care to either hold
i_lock or have I_SYNC set, we can get away without taking inode reference
in writeback_sb_inodes().
As a side effect of these changes, we also fix possible use-after-free in
wb_writeback() because inode_wait_for_writeback() call could try to reacquire
i_lock on the inode that was already free.
Signed-off-by: Jan Kara <jack@suse.cz>
Signed-off-by: Fengguang Wu <fengguang.wu@intel.com>
Reorder structure writeback_control to remove 8 bytes of padding on 64
bit builds, this shrinks its size from 48 to 40 bytes.
This structure is always on the stack and uses C99 named initialisation,
so should be safe and have a small impact on stack usage.
Signed-off-by: Richard Kennedy <richard@rsk.demon.co.uk>
Signed-off-by: Fengguang Wu <fengguang.wu@intel.com>
The maximum number of dirty pages that exist in the system at any time is
determined by a number of pages considered dirtyable and a user-configured
percentage of those, or an absolute number in bytes.
This number of dirtyable pages is the sum of memory provided by all the
zones in the system minus their lowmem reserves and high watermarks, so
that the system can retain a healthy number of free pages without having
to reclaim dirty pages.
But there is a flaw in that we have a zoned page allocator which does not
care about the global state but rather the state of individual memory
zones. And right now there is nothing that prevents one zone from filling
up with dirty pages while other zones are spared, which frequently leads
to situations where kswapd, in order to restore the watermark of free
pages, does indeed have to write pages from that zone's LRU list. This
can interfere so badly with IO from the flusher threads that major
filesystems (btrfs, xfs, ext4) mostly ignore write requests from reclaim
already, taking away the VM's only possibility to keep such a zone
balanced, aside from hoping the flushers will soon clean pages from that
zone.
Enter per-zone dirty limits. They are to a zone's dirtyable memory what
the global limit is to the global amount of dirtyable memory, and try to
make sure that no single zone receives more than its fair share of the
globally allowed dirty pages in the first place. As the number of pages
considered dirtyable excludes the zones' lowmem reserves and high
watermarks, the maximum number of dirty pages in a zone is such that the
zone can always be balanced without requiring page cleaning.
As this is a placement decision in the page allocator and pages are
dirtied only after the allocation, this patch allows allocators to pass
__GFP_WRITE when they know in advance that the page will be written to and
become dirty soon. The page allocator will then attempt to allocate from
the first zone of the zonelist - which on NUMA is determined by the task's
NUMA memory policy - that has not exceeded its dirty limit.
At first glance, it would appear that the diversion to lower zones can
increase pressure on them, but this is not the case. With a full high
zone, allocations will be diverted to lower zones eventually, so it is
more of a shift in timing of the lower zone allocations. Workloads that
previously could fit their dirty pages completely in the higher zone may
be forced to allocate from lower zones, but the amount of pages that
"spill over" are limited themselves by the lower zones' dirty constraints,
and thus unlikely to become a problem.
For now, the problem of unfair dirty page distribution remains for NUMA
configurations where the zones allowed for allocation are in sum not big
enough to trigger the global dirty limits, wake up the flusher threads and
remedy the situation. Because of this, an allocation that could not
succeed on any of the considered zones is allowed to ignore the dirty
limits before going into direct reclaim or even failing the allocation,
until a future patch changes the global dirty throttling and flusher
thread activation so that they take individual zone states into account.
Test results
15M DMA + 3246M DMA32 + 504 Normal = 3765M memory
40% dirty ratio
16G USB thumb drive
10 runs of dd if=/dev/zero of=disk/zeroes bs=32k count=$((10 << 15))
seconds nr_vmscan_write
(stddev) min| median| max
xfs
vanilla: 549.747( 3.492) 0.000| 0.000| 0.000
patched: 550.996( 3.802) 0.000| 0.000| 0.000
fuse-ntfs
vanilla: 1183.094(53.178) 54349.000| 59341.000| 65163.000
patched: 558.049(17.914) 0.000| 0.000| 43.000
btrfs
vanilla: 573.679(14.015) 156657.000| 460178.000| 606926.000
patched: 563.365(11.368) 0.000| 0.000| 1362.000
ext4
vanilla: 561.197(15.782) 0.000|2725438.000|4143837.000
patched: 568.806(17.496) 0.000| 0.000| 0.000
Signed-off-by: Johannes Weiner <jweiner@redhat.com>
Reviewed-by: Minchan Kim <minchan.kim@gmail.com>
Acked-by: Mel Gorman <mgorman@suse.de>
Reviewed-by: Michal Hocko <mhocko@suse.cz>
Tested-by: Wu Fengguang <fengguang.wu@intel.com>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Christoph Hellwig <hch@infradead.org>
Cc: Dave Chinner <david@fromorbit.com>
Cc: Jan Kara <jack@suse.cz>
Cc: Shaohua Li <shaohua.li@intel.com>
Cc: Rik van Riel <riel@redhat.com>
Cc: Chris Mason <chris.mason@oracle.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
The tracing ring-buffer used this function briefly, but not anymore.
Make it local to the writeback code again.
Also, move the function so that no forward declaration needs to be
reintroduced.
Signed-off-by: Johannes Weiner <hannes@cmpxchg.org>
Acked-by: Mel Gorman <mgorman@suse.de>
Reviewed-by: Michal Hocko <mhocko@suse.cz>
Cc: Wu Fengguang <fengguang.wu@intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Fix compile error
fs/fs-writeback.c:515:33: error: ‘PAGE_CACHE_SHIFT’ undeclared (first use in this function)
Reported-by: Randy Dunlap <rdunlap@xenotime.net>
Acked-by: Randy Dunlap <rdunlap@xenotime.net>
Signed-off-by: Wu Fengguang <fengguang.wu@intel.com>
De-account the accumulative dirty counters on page redirty.
Page redirties (very common in ext4) will introduce mismatch between
counters (a) and (b)
a) NR_DIRTIED, BDI_DIRTIED, tsk->nr_dirtied
b) NR_WRITTEN, BDI_WRITTEN
This will introduce systematic errors in balanced_rate and result in
dirty page position errors (ie. the dirty pages are no longer balanced
around the global/bdi setpoints).
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>
It's a years long problem that a large number of short-lived dirtiers
(eg. gcc instances in a fast kernel build) may starve long-run dirtiers
(eg. dd) as well as pushing the dirty pages to the global hard limit.
The solution is to charge the pages dirtied by the exited gcc to the
other random dirtying tasks. It sounds not perfect, however should
behave good enough in practice, seeing as that throttled tasks aren't
actually running so those that are running are more likely to pick it up
and get throttled, therefore promoting an equal spread.
Randy: fix compile error: 'dirty_throttle_leaks' undeclared in exit.c
Acked-by: Jan Kara <jack@suse.cz>
Acked-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Signed-off-by: Randy Dunlap <rdunlap@xenotime.net>
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>
Revert the pass-good area introduced in ffd1f609ab ("writeback:
introduce max-pause and pass-good dirty limits") and make the max-pause
area smaller and safe.
This fixes ~30% performance regression in the ext3 data=writeback
fio_mmap_randwrite_64k/fio_mmap_randrw_64k test cases, where there are
12 JBOD disks, on each disk runs 8 concurrent tasks doing reads+writes.
Using deadline scheduler also has a regression, but not that big as CFQ,
so this suggests we have some write starvation.
The test logs show that
- the disks are sometimes under utilized
- global dirty pages sometimes rush high to the pass-good area for
several hundred seconds, while in the mean time some bdi dirty pages
drop to very low value (bdi_dirty << bdi_thresh). Then suddenly the
global dirty pages dropped under global dirty threshold and bdi_dirty
rush very high (for example, 2 times higher than bdi_thresh). During
which time balance_dirty_pages() is not called at all.
So the problems are
1) The random writes progress so slow that they break the assumption of
the max-pause logic that "8 pages per 200ms is typically more than
enough to curb heavy dirtiers".
2) The max-pause logic ignored task_bdi_thresh and thus opens the possibility
for some bdi's to over dirty pages, leading to (bdi_dirty >> bdi_thresh)
and then (bdi_thresh >> bdi_dirty) for others.
3) The higher max-pause/pass-good thresholds somehow leads to the bad
swing of dirty pages.
The fix is to allow the task to slightly dirty over task_bdi_thresh, but
no way to exceed bdi_dirty and/or global dirty_thresh.
Tests show that it fixed the JBOD regression completely (both behavior
and performance), while still being able to cut down large pause times
in balance_dirty_pages() for single-disk cases.
Reported-by: Li Shaohua <shaohua.li@intel.com>
Tested-by: Li Shaohua <shaohua.li@intel.com>
Acked-by: Jan Kara <jack@suse.cz>
Signed-off-by: Wu Fengguang <fengguang.wu@intel.com>
Originally, MAX_WRITEBACK_PAGES was hard-coded to 1024 because of a
concern of not holding I_SYNC for too long. (At least, that was the
comment previously.) This doesn't make sense now because the only
time we wait for I_SYNC is if we are calling sync or fsync, and in
that case we need to write out all of the data anyway. Previously
there may have been other code paths that waited on I_SYNC, but not
any more. -- Theodore Ts'o
So remove the MAX_WRITEBACK_PAGES constraint. The writeback pages
will adapt to as large as the storage device can write within 500ms.
XFS is observed to do IO completions in a batch, and the batch size is
equal to the write chunk size. To avoid dirty pages to suddenly drop
out of balance_dirty_pages()'s dirty control scope and create large
fluctuations, the chunk size is also limited to half the control scope.
The balance_dirty_pages() control scrope is
[(background_thresh + dirty_thresh) / 2, dirty_thresh]
which is by default [15%, 20%] of global dirty pages, whose range size
is dirty_thresh / DIRTY_FULL_SCOPE.
The adpative write chunk size will be rounded to the nearest 4MB
boundary.
http://bugzilla.kernel.org/show_bug.cgi?id=13930
CC: Theodore Ts'o <tytso@mit.edu>
CC: Dave Chinner <david@fromorbit.com>
CC: Chris Mason <chris.mason@oracle.com>
CC: Peter Zijlstra <a.p.zijlstra@chello.nl>
Signed-off-by: Wu Fengguang <fengguang.wu@intel.com>
The max-pause limit helps to keep the sleep time inside
balance_dirty_pages() within MAX_PAUSE=200ms. The 200ms max sleep means
per task rate limit of 8pages/200ms=160KB/s when dirty exceeded, which
normally is enough to stop dirtiers from continue pushing the dirty
pages high, unless there are a sufficient large number of slow dirtiers
(eg. 500 tasks doing 160KB/s will still sum up to 80MB/s, exceeding the
write bandwidth of a slow disk and hence accumulating more and more dirty
pages).
The pass-good limit helps to let go of the good bdi's in the presence of
a blocked bdi (ie. NFS server not responding) or slow USB disk which for
some reason build up a large number of initial dirty pages that refuse
to go away anytime soon.
For example, given two bdi's A and B and the initial state
bdi_thresh_A = dirty_thresh / 2
bdi_thresh_B = dirty_thresh / 2
bdi_dirty_A = dirty_thresh / 2
bdi_dirty_B = dirty_thresh / 2
Then A get blocked, after a dozen seconds
bdi_thresh_A = 0
bdi_thresh_B = dirty_thresh
bdi_dirty_A = dirty_thresh / 2
bdi_dirty_B = dirty_thresh / 2
The (bdi_dirty_B < bdi_thresh_B) test is now useless and the dirty pages
will be effectively throttled by condition (nr_dirty < dirty_thresh).
This has two problems:
(1) we lose the protections for light dirtiers
(2) balance_dirty_pages() effectively becomes IO-less because the
(bdi_nr_reclaimable > bdi_thresh) test won't be true. This is good
for IO, but balance_dirty_pages() loses an important way to break
out of the loop which leads to more spread out throttle delays.
DIRTY_PASSGOOD_AREA can eliminate the above issues. The only problem is,
DIRTY_PASSGOOD_AREA needs to be defined as 2 to fully cover the above
example while this patch uses the more conservative value 8 so as not to
surprise people with too many dirty pages than expected.
The max-pause limit won't noticeably impact the speed dirty pages are
knocked down when there is a sudden drop of global/bdi dirty thresholds.
Because the heavy dirties will be throttled below 160KB/s which is slow
enough. It does help to avoid long dirty throttle delays and especially
will make light dirtiers more responsive.
Signed-off-by: Wu Fengguang <fengguang.wu@intel.com>
The start of a heavy weight application (ie. KVM) may instantly knock
down determine_dirtyable_memory() if the swap is not enabled or full.
global_dirty_limits() and bdi_dirty_limit() will in turn get global/bdi
dirty thresholds that are _much_ lower than the global/bdi dirty pages.
balance_dirty_pages() will then heavily throttle all dirtiers including
the light ones, until the dirty pages drop below the new dirty thresholds.
During this _deep_ dirty-exceeded state, the system may appear rather
unresponsive to the users.
About "deep" dirty-exceeded: task_dirty_limit() assigns 1/8 lower dirty
threshold to heavy dirtiers than light ones, and the dirty pages will
be throttled around the heavy dirtiers' dirty threshold and reasonably
below the light dirtiers' dirty threshold. In this state, only the heavy
dirtiers will be throttled and the dirty pages are carefully controlled
to not exceed the light dirtiers' dirty threshold. However if the
threshold itself suddenly drops below the number of dirty pages, the
light dirtiers will get heavily throttled.
So introduce global_dirty_limit for tracking the global dirty threshold
with policies
- follow downwards slowly
- follow up in one shot
global_dirty_limit can effectively mask out the impact of sudden drop of
dirtyable memory. It will be used in the next patch for two new type of
dirty limits. Note that the new dirty limits are not going to avoid
throttling the light dirtiers, but could limit their sleep time to 200ms.
Signed-off-by: Wu Fengguang <fengguang.wu@intel.com>
The estimation value will start from 100MB/s and adapt to the real
bandwidth in seconds.
It tries to update the bandwidth only when disk is fully utilized.
Any inactive period of more than one second will be skipped.
The estimated bandwidth will be reflecting how fast the device can
writeout when _fully utilized_, and won't drop to 0 when it goes idle.
The value will remain constant at disk idle time. At busy write time, if
not considering fluctuations, it will also remain high unless be knocked
down by possible concurrent reads that compete for the disk time and
bandwidth with async writes.
The estimation is not done purely in the flusher because there is no
guarantee for write_cache_pages() to return timely to update bandwidth.
The bdi->avg_write_bandwidth smoothing is very effective for filtering
out sudden spikes, however may be a little biased in long term.
The overheads are low because the bdi bandwidth update only occurs at
200ms intervals.
The 200ms update interval is suitable, because it's not possible to get
the real bandwidth for the instance at all, due to large fluctuations.
The NFS commits can be as large as seconds worth of data. One XFS
completion may be as large as half second worth of data if we are going
to increase the write chunk to half second worth of data. In ext4,
fluctuations with time period of around 5 seconds is observed. And there
is another pattern of irregular periods of up to 20 seconds on SSD tests.
That's why we are not only doing the estimation at 200ms intervals, but
also averaging them over a period of 3 seconds and then go further to do
another level of smoothing in avg_write_bandwidth.
CC: Li Shaohua <shaohua.li@intel.com>
CC: Peter Zijlstra <a.p.zijlstra@chello.nl>
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>
Remove two unused struct writeback_control fields:
.encountered_congestion (completely unused)
.nonblocking (never set, checked/showed in XFS,NFS/btrfs)
The .for_background check in nfs_write_inode() is also removed btw,
as .for_background implies WB_SYNC_NONE.
Reviewed-by: Jan Kara <jack@suse.cz>
Proposed-by: Christoph Hellwig <hch@infradead.org>
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 removes writeback_control.wb_start and does more straightforward
sync livelock prevention by setting .older_than_this to prevent extra
inodes from being enqueued in the first place.
Acked-by: Jan Kara <jack@suse.cz>
Signed-off-by: Wu Fengguang <fengguang.wu@intel.com>
Split the global inode_wb_list_lock into a per-bdi_writeback list_lock,
as it's currently the most contended lock in the system for metadata
heavy workloads. It won't help for single-filesystem workloads for
which we'll need the I/O-less balance_dirty_pages, but at least we
can dedicate a cpu to spinning on each bdi now for larger systems.
Based on earlier patches from Nick Piggin and Dave Chinner.
It reduces lock contentions to 1/4 in this test case:
10 HDD JBOD, 100 dd on each disk, XFS, 6GB ram
lock_stat version 0.3
-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
class name con-bounces contentions waittime-min waittime-max waittime-total acq-bounces acquisitions holdtime-min holdtime-max holdtime-total
-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
vanilla 2.6.39-rc3:
inode_wb_list_lock: 42590 44433 0.12 147.74 144127.35 252274 886792 0.08 121.34 917211.23
------------------
inode_wb_list_lock 2 [<ffffffff81165da5>] bdev_inode_switch_bdi+0x29/0x85
inode_wb_list_lock 34 [<ffffffff8115bd0b>] inode_wb_list_del+0x22/0x49
inode_wb_list_lock 12893 [<ffffffff8115bb53>] __mark_inode_dirty+0x170/0x1d0
inode_wb_list_lock 10702 [<ffffffff8115afef>] writeback_single_inode+0x16d/0x20a
------------------
inode_wb_list_lock 2 [<ffffffff81165da5>] bdev_inode_switch_bdi+0x29/0x85
inode_wb_list_lock 19 [<ffffffff8115bd0b>] inode_wb_list_del+0x22/0x49
inode_wb_list_lock 5550 [<ffffffff8115bb53>] __mark_inode_dirty+0x170/0x1d0
inode_wb_list_lock 8511 [<ffffffff8115b4ad>] writeback_sb_inodes+0x10f/0x157
2.6.39-rc3 + patch:
&(&wb->list_lock)->rlock: 11383 11657 0.14 151.69 40429.51 90825 527918 0.11 145.90 556843.37
------------------------
&(&wb->list_lock)->rlock 10 [<ffffffff8115b189>] inode_wb_list_del+0x5f/0x86
&(&wb->list_lock)->rlock 1493 [<ffffffff8115b1ed>] writeback_inodes_wb+0x3d/0x150
&(&wb->list_lock)->rlock 3652 [<ffffffff8115a8e9>] writeback_sb_inodes+0x123/0x16f
&(&wb->list_lock)->rlock 1412 [<ffffffff8115a38e>] writeback_single_inode+0x17f/0x223
------------------------
&(&wb->list_lock)->rlock 3 [<ffffffff8110b5af>] bdi_lock_two+0x46/0x4b
&(&wb->list_lock)->rlock 6 [<ffffffff8115b189>] inode_wb_list_del+0x5f/0x86
&(&wb->list_lock)->rlock 2061 [<ffffffff8115af97>] __mark_inode_dirty+0x173/0x1cf
&(&wb->list_lock)->rlock 2629 [<ffffffff8115a8e9>] writeback_sb_inodes+0x123/0x16f
hughd@google.com: fix recursive lock when bdi_lock_two() is called with new the same as old
akpm@linux-foundation.org: cleanup bdev_inode_switch_bdi() comment
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Hugh Dickins <hughd@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Wu Fengguang <fengguang.wu@intel.com>
The flusher works on dirty inodes in batches, and may quit prematurely
if the batch of inodes happen to be metadata-only dirtied: in this case
wbc->nr_to_write won't be decreased at all, which stands for "no pages
written" but also mis-interpreted as "no progress".
So introduce writeback_control.inodes_written to count the inodes get
cleaned from VFS POV. A non-zero value means there are some progress on
writeback, in which case more writeback can be tried.
Acked-by: Jan Kara <jack@suse.cz>
Acked-by: Mel Gorman <mel@csn.ul.ie>
Signed-off-by: Wu Fengguang <fengguang.wu@intel.com>
sync(2) is performed in two stages: the WB_SYNC_NONE sync and the
WB_SYNC_ALL sync. Identify the first stage with .tagged_writepages and
do livelock prevention for it, too.
Jan's commit f446daaea9 ("mm: implement writeback livelock avoidance
using page tagging") is a partial fix in that it only fixed the
WB_SYNC_ALL phase livelock.
Although ext4 is tested to no longer livelock with commit f446daaea9,
it may due to some "redirty_tail() after pages_skipped" effect which
is by no means a guarantee for _all_ the file systems.
Note that writeback_inodes_sb() is called by not only sync(), they are
treated the same because the other callers also need livelock prevention.
Impact: It changes the order in which pages/inodes are synced to disk.
Now in the WB_SYNC_NONE stage, it won't proceed to write the next inode
until finished with the current inode.
Acked-by: Jan Kara <jack@suse.cz>
CC: Dave Chinner <david@fromorbit.com>
Signed-off-by: Wu Fengguang <fengguang.wu@intel.com>
All that remains of the inode_lock is protecting the inode hash list
manipulation and traversals. Rename the inode_lock to
inode_hash_lock to reflect it's actual function.
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Protect the inode writeback list with a new global lock
inode_wb_list_lock and use it to protect the list manipulations and
traversals. This lock replaces the inode_lock as the inodes on the
list can be validity checked while holding the inode->i_lock and
hence the inode_lock is no longer needed to protect the list.
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Tejun Heo