After invoking ->dirty_inode(), __mark_inode_dirty() does smp_mb() and
tests inode->i_state locklessly to see whether it already has all the
necessary I_DIRTY bits set.  The comment above the barrier doesn't
contain any useful information - memory barriers can't ensure "changes
are seen by all cpus" by itself.

And it sure enough was broken.  Please consider the following
scenario.

 CPU 0					CPU 1
 -------------------------------------------------------------------------------

					enters __writeback_single_inode()
					grabs inode->i_lock
					tests PAGECACHE_TAG_DIRTY which is clear
 enters __set_page_dirty()
 grabs mapping->tree_lock
 sets PAGECACHE_TAG_DIRTY
 releases mapping->tree_lock
 leaves __set_page_dirty()

 enters __mark_inode_dirty()
 smp_mb()
 sees I_DIRTY_PAGES set
 leaves __mark_inode_dirty()
					clears I_DIRTY_PAGES
					releases inode->i_lock

Now @inode has dirty pages w/ I_DIRTY_PAGES clear.  This doesn't seem
to lead to an immediately critical problem because requeue_inode()
later checks PAGECACHE_TAG_DIRTY instead of I_DIRTY_PAGES when
deciding whether the inode needs to be requeued for IO and there are
enough unintentional memory barriers inbetween, so while the inode
ends up with inconsistent I_DIRTY_PAGES flag, it doesn't fall off the
IO list.

The lack of explicit barrier may also theoretically affect the other
I_DIRTY bits which deal with metadata dirtiness.  There is no
guarantee that a strong enough barrier exists between
I_DIRTY_[DATA]SYNC clearing and write_inode() writing out the dirtied
inode.  Filesystem inode writeout path likely has enough stuff which
can behave as full barrier but it's theoretically possible that the
writeout may not see all the updates from ->dirty_inode().

Fix it by adding an explicit smp_mb() after I_DIRTY clearing.  Note
that I_DIRTY_PAGES needs a special treatment as it always needs to be
cleared to be interlocked with the lockless test on
__mark_inode_dirty() side.  It's cleared unconditionally and
reinstated after smp_mb() if the mapping still has dirty pages.

Also add comments explaining how and why the barriers are paired.

Lightly tested.
Signed-off-by: NTejun Heo <tj@kernel.org>
Cc: Jan Kara <jack@suse.cz>
Cc: Mikulas Patocka <mpatocka@redhat.com>
Cc: Jens Axboe <axboe@kernel.dk>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Cc: stable@vger.kernel.org
Reviewed-by: NJan Kara <jack@suse.cz>
Signed-off-by: NJens Axboe <axboe@fb.com>

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: NNeilBrown <neilb@suse.de>
Acked-by: David Howells <dhowells@redhat.com> (fscache, keys)
Acked-by: Steven Whitehouse <swhiteho@redhat.com> (gfs2)
Acked-by: NPeter 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.brownSigned-off-by: NIngo Molnar <mingo@kernel.org>

After commit 839a8e86 ("writeback: replace custom worker pool
implementation with unbound workqueue") when device is removed while we
are writing to it we crash in bdi_writeback_workfn() ->
set_worker_desc() because bdi->dev is NULL.

This can happen because even though bdi_unregister() cancels all pending
flushing work, nothing really prevents new ones from being queued from
balance_dirty_pages() or other places.

Fix the problem by clearing BDI_registered bit in bdi_unregister() and
checking it before scheduling of any flushing work.

Fixes: 839a8e86Reviewed-by: NTejun Heo <tj@kernel.org>
Signed-off-by: NJan Kara <jack@suse.cz>
Cc: Derek Basehore <dbasehore@chromium.org>
Cc: Jens Axboe <axboe@kernel.dk>
Cc: <stable@vger.kernel.org>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

bdi_wakeup_thread_delayed() used the mod_delayed_work() function to
schedule work to writeback dirty inodes.  The problem with this is that
it can delay work that is scheduled for immediate execution, such as the
work from sync_inodes_sb().  This can happen since mod_delayed_work()
can now steal work from a work_queue.  This fixes the problem by using
queue_delayed_work() instead.  This is a regression caused by commit
839a8e86 ("writeback: replace custom worker pool implementation with
unbound workqueue").

The reason that this causes a problem is that laptop-mode will change
the delay, dirty_writeback_centisecs, to 60000 (10 minutes) by default.
In the case that bdi_wakeup_thread_delayed() races with
sync_inodes_sb(), sync will be stopped for 10 minutes and trigger a hung
task.  Even if dirty_writeback_centisecs is not long enough to cause a
hung task, we still don't want to delay sync for that long.

We fix the problem by using queue_delayed_work() when we want to
schedule writeback sometime in future.  This function doesn't change the
timer if it is already armed.

For the same reason, we also change bdi_writeback_workfn() to
immediately queue the work again in the case that the work_list is not
empty.  The same problem can happen if the sync work is run on the
rescue worker.

[jack@suse.cz: update changelog, add comment, use bdi_wakeup_thread_delayed()]
Signed-off-by: NDerek Basehore <dbasehore@chromium.org>
Reviewed-by: NJan Kara <jack@suse.cz>
Cc: Alexander Viro <viro@zento.linux.org.uk>
Reviewed-by: NTejun Heo <tj@kernel.org>
Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Cc: "Darrick J. Wong" <darrick.wong@oracle.com>
Cc: Derek Basehore <dbasehore@chromium.org>
Cc: Kees Cook <keescook@chromium.org>
Cc: Benson Leung <bleung@chromium.org>
Cc: Sonny Rao <sonnyrao@chromium.org>
Cc: Luigi Semenzato <semenzato@chromium.org>
Cc: Jens Axboe <axboe@kernel.dk>
Cc: Dave Chinner <david@fromorbit.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

This reverts commit c4a391b5. 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: NJan Kara <jack@suse.cz>

GFS2 has carried what is more or less a copy of the
write_cache_pages() for some time. It seems that this
copy has slipped behind the core code over time. This
patch brings it back uptodate, and in addition adds the
tracepoint which would otherwise be missing.

We could go further, and eliminate some or all of the
code duplication here. The issue is that if we do that,
then the function we need to split out from the existing
write_cache_pages(), which will look a lot like
gfs2_jdata_write_pagevec(), would land up putting quite a
lot of extra variables on the stack. I know that has been
a problem in the past in the writeback code path, which
is why I've hesitated to do it here.
Signed-off-by: NSteven Whitehouse <swhiteho@redhat.com>

Commit 4f8ad655 "writeback: Refactor writeback_single_inode()" added
a condition to skip clean inode. However this is wrong in WB_SYNC_ALL
mode because there we also want to wait for outstanding writeback on
possibly clean inode. This was causing occasional data corruption issues
on NFS because it uses sync_inode() to make sure all outstanding writes
are flushed to the server before truncating the inode and with
sync_inode() returning prematurely file was sometimes extended back
by an outstanding write after it was truncated.

So modify the test to also check for pages under writeback in
WB_SYNC_ALL mode.

CC: stable@vger.kernel.org # >= 3.5
Fixes: 4f8ad655Reported-and-tested-by: NDan Duval <dan.duval@oracle.com>
Signed-off-by: NJan 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: NJan Kara <jack@suse.cz>
Reviewed-by: NFengguang Wu <fengguang.wu@intel.com>
Reviewed-by: NDave Chinner <dchinner@redhat.com>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

aka br_write_{lock,unlock} of vfsmount_lock.  Inlines in fs/mount.h,
vfsmount_lock extern moved over there as well.
Signed-off-by: NAl Viro <viro@zeniv.linux.org.uk>

There is a race between mark inode dirty and writeback thread, see the
following scenario.  In this case, writeback thread will not run though
there is dirty_io.

__mark_inode_dirty()                                          bdi_writeback_workfn()
	...                                                       	...
	spin_lock(&inode->i_lock);
	...
	if (bdi_cap_writeback_dirty(bdi)) {
	    <<< assume wb has dirty_io, so wakeup_bdi is false.
	    <<< the following inode_dirty also have wakeup_bdi false.
	    if (!wb_has_dirty_io(&bdi->wb))
		    wakeup_bdi = true;
	}
	spin_unlock(&inode->i_lock);
	                                                            <<< assume last dirty_io is removed here.
	                                                            pages_written = wb_do_writeback(wb);
	                                                            ...
	                                                            <<< work_list empty and wb has no dirty_io,
	                                                            <<< delayed_work will not be queued.
	                                                            if (!list_empty(&bdi->work_list) ||
	                                                                (wb_has_dirty_io(wb) && dirty_writeback_interval))
	                                                                queue_delayed_work(bdi_wq, &wb->dwork,
	                                                                    msecs_to_jiffies(dirty_writeback_interval * 10));
	spin_lock(&bdi->wb.list_lock);
	inode->dirtied_when = jiffies;
	<<< new dirty_io is added.
	list_move(&inode->i_wb_list, &bdi->wb.b_dirty);
	spin_unlock(&bdi->wb.list_lock);

	<<< though there is dirty_io, but wakeup_bdi is false,
	<<< so writeback thread will not be waked up and
	<<< the new dirty_io will not be flushed.
	if (wakeup_bdi)
	    bdi_wakeup_thread_delayed(bdi);

Writeback will run until there is a new flush work queued.  This may cause
a lot of dirty pages stay in memory for a long time.
Signed-off-by: NJunxiao Bi <junxiao.bi@oracle.com>
Reviewed-by: NJan Kara <jack@suse.cz>
Cc: Fengguang Wu <fengguang.wu@intel.com>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

It's not used globally and could be static.
Signed-off-by: NWanpeng 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: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

In case when system contains no dirty pages, wakeup_flusher_threads() will
submit WB_SYNC_NONE writeback for 0 pages so wb_writeback() exits
immediately without doing anything, even though there are dirty inodes in
the system.  Thus sync(1) will write all the dirty inodes from a
WB_SYNC_ALL writeback pass which is slow.

Fix the problem by using get_nr_dirty_pages() in wakeup_flusher_threads()
instead of calculating number of dirty pages manually.  That function also
takes number of dirty inodes into account.
Signed-off-by: NJan Kara <jack@suse.cz>
Reported-by: NPaul Taysom <taysom@chromium.org>
Cc: Wu Fengguang <fengguang.wu@intel.com>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

After commit 839a8e86 ("writeback: replace custom worker pool
implementation with unbound workqueue"), bdi_writeback_workfn runs off
bdi_writeback->dwork, on each execution, it processes bdi->work_list and
reschedules if there are more things to do instead of flush any work
that race with us existing.  It is unecessary to check force_wait in
wb_do_writeback since it is always 0 after the mentioned commit.  This
patch remove the force_wait in wb_do_writeback.
Signed-off-by: NWanpeng Li <liwanp@linux.vnet.ibm.com>
Reviewed-by: NTejun Heo <tj@kernel.org>
Reviewed-by: NFengguang Wu <fengguang.wu@intel.com>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

It's not used globally and could be static.
Signed-off-by: NHaicheng 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: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

It is very likely that block device inode will be part of BDI dirty list
as well. However it doesn't make sence to sort inodes on the b_io list
just because of this inode (as it contains buffers all over the device
anyway). So save some CPU cycles which is valuable since we hold relatively
contented wb->list_lock.
Signed-off-by: NJan Kara <jack@suse.cz>

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: NDave Chinner <dchinner@redhat.com>
Reviewed-by: NJan Kara <jack@suse.cz>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

Writeback has been recently converted to use workqueue instead of its
private thread pool implementation.  One negative side effect of this
conversion is that there's no easy to tell which backing device a
writeback work item was working on at the time of task dump, be it
sysrq-t, BUG, WARN or whatever, which, according to our writeback
brethren, is important in tracking down issues with a lot of mounted
file systems on a lot of different devices.

This patch restores that information using the new worker description
facility.  bdi_writeback_workfn() calls set_work_desc() to identify
which bdi it's working on.  The description is printed out together with
the worqueue name and worker function as in the following example dump.

 WARNING: at fs/fs-writeback.c:1015 bdi_writeback_workfn+0x2b4/0x3c0()
 Modules linked in:
 Pid: 28, comm: kworker/u18:0 Not tainted 3.9.0-rc1-work+ #24 empty empty/S3992
 Workqueue: writeback bdi_writeback_workfn (flush-8:16)
  ffffffff820a3a98 ffff88015b927cb8 ffffffff81c61855 ffff88015b927cf8
  ffffffff8108f500 0000000000000000 ffff88007a171948 ffff88007a1716b0
  ffff88015b49df00 ffff88015b8d3940 0000000000000000 ffff88015b927d08
 Call Trace:
  [<ffffffff81c61855>] dump_stack+0x19/0x1b
  [<ffffffff8108f500>] warn_slowpath_common+0x70/0xa0
  [<ffffffff8108f54a>] warn_slowpath_null+0x1a/0x20
  [<ffffffff81200144>] bdi_writeback_workfn+0x2b4/0x3c0
  [<ffffffff810b4c87>] process_one_work+0x1d7/0x660
  [<ffffffff810b5c72>] worker_thread+0x122/0x380
  [<ffffffff810bdfea>] kthread+0xea/0xf0
  [<ffffffff81c6cedc>] ret_from_fork+0x7c/0xb0
Signed-off-by: NTejun Heo <tj@kernel.org>
Cc: Dave Chinner <david@fromorbit.com>
Cc: Jan Kara <jack@suse.cz>
Cc: Jens Axboe <axboe@kernel.dk>
Cc: Oleg Nesterov <oleg@redhat.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

Writeback implements its own worker pool - each bdi can be associated
with a worker thread which is created and destroyed dynamically.  The
worker thread for the default bdi is always present and serves as the
"forker" thread which forks off worker threads for other bdis.

there's no reason for writeback to implement its own worker pool when
using unbound workqueue instead is much simpler and more efficient.
This patch replaces custom worker pool implementation in writeback
with an unbound workqueue.

The conversion isn't too complicated but the followings are worth
mentioning.

* bdi_writeback->last_active, task and wakeup_timer are removed.
  delayed_work ->dwork is added instead.  Explicit timer handling is
  no longer necessary.  Everything works by either queueing / modding
  / flushing / canceling the delayed_work item.

* bdi_writeback_thread() becomes bdi_writeback_workfn() which runs off
  bdi_writeback->dwork.  On each execution, it processes
  bdi->work_list and reschedules itself if there are more things to
  do.

  The function also handles low-mem condition, which used to be
  handled by the forker thread.  If the function is running off a
  rescuer thread, it only writes out limited number of pages so that
  the rescuer can serve other bdis too.  This preserves the flusher
  creation failure behavior of the forker thread.

* INIT_LIST_HEAD(&bdi->bdi_list) is used to tell
  bdi_writeback_workfn() about on-going bdi unregistration so that it
  always drains work_list even if it's running off the rescuer.  Note
  that the original code was broken in this regard.  Under memory
  pressure, a bdi could finish unregistration with non-empty
  work_list.

* The default bdi is no longer special.  It now is treated the same as
  any other bdi and bdi_cap_flush_forker() is removed.

* BDI_pending is no longer used.  Removed.

* Some tracepoints become non-applicable.  The following TPs are
  removed - writeback_nothread, writeback_wake_thread,
  writeback_wake_forker_thread, writeback_thread_start,
  writeback_thread_stop.

Everything, including devices coming and going away and rescuer
operation under simulated memory pressure, seems to work fine in my
test setup.
Signed-off-by: NTejun Heo <tj@kernel.org>
Reviewed-by: NJan Kara <jack@suse.cz>
Cc: Jens Axboe <axboe@kernel.dk>
Cc: Fengguang Wu <fengguang.wu@intel.com>
Cc: Jeff Moyer <jmoyer@redhat.com>

Add tracepoints for page dirtying, writeback_single_inode start, inode
dirtying and writeback.  For the latter two inode events, a pair of
events are defined to denote start and end of the operations (the
starting one has _start suffix and the one w/o suffix happens after
the operation is complete).  These inode ops are FS specific and can
be non-trivial and having enclosing tracepoints is useful for external
tracers.

This is part of tracepoint additions to improve visiblity into
dirtying / writeback operations for io tracer and userland.

v2: writeback_dirty_inode[_start] TPs may be called for files on
    pseudo FSes w/ unregistered bdi.  Check whether bdi->dev is %NULL
    before dereferencing.

v3: buffer dirtying moved to a block TP.
Signed-off-by: NTejun Heo <tj@kernel.org>
Reviewed-by: NJan Kara <jack@suse.cz>
Signed-off-by: NJens Axboe <axboe@kernel.dk>

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: NJan Kara <jack@suse.cz>
Signed-off-by: NMiao Xie <miaox@cn.fujitsu.com>
Signed-off-by: NFengguang Wu <fengguang.wu@intel.com>

Signed-off-by: NYan Hong <clouds.yan@gmail.com>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

Commit 169ebd90 ("writeback: Avoid iput() from flusher thread")
removed iget-iput pair from inode writeback.  As a side effect, inodes
that are dirty during iput_final() call won't be ever added to inode LRU
(iput_final() doesn't add dirty inodes to LRU and later when the inode
is cleaned there's noone to add the inode there).  Thus inodes are
effectively unreclaimable until someone looks them up again.

The practical effect of this bug is limited by the fact that inodes are
pinned by a dentry for long enough that the inode gets cleaned.  But
still the bug can have nasty consequences leading up to OOM conditions
under certain circumstances.  Following can easily reproduce the
problem:

  for (( i = 0; i < 1000; i++ )); do
    mkdir $i
    for (( j = 0; j < 1000; j++ )); do
      touch $i/$j
      echo 2 > /proc/sys/vm/drop_caches
    done
  done

then one needs to run 'sync; ls -lR' to make inodes reclaimable again.

We fix the issue by inserting unused clean inodes into the LRU after
writeback finishes in inode_sync_complete().
Signed-off-by: NJan Kara <jack@suse.cz>
Reported-by: NOGAWA Hirofumi <hirofumi@mail.parknet.co.jp>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Cc: OGAWA Hirofumi <hirofumi@mail.parknet.co.jp>
Cc: Wu Fengguang <fengguang.wu@intel.com>
Cc: Dave Chinner <david@fromorbit.com>
Cc: <stable@vger.kernel.org>		[3.5+]
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

The parameter 'wb' is never used in this function.
Signed-off-by: NYan Hong <clouds.yan@gmail.com>
Acked-by: NWu Fengguang <fengguang.wu@intel.com>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

Argument @only_this_sb has been removed.
Signed-off-by: NLiu Bo <liub.liubo@gmail.com>
Signed-off-by: NJiri Kosina <jkosina@suse.cz>

In ext4_nonda_switch(), if the file system is getting full we used to
call writeback_inodes_sb_if_idle().  The problem is that we can be
holding i_mutex already, and this causes a potential deadlock when
writeback_inodes_sb_if_idle() when it tries to take s_umount.  (See
lockdep output below).

As it turns out we don't need need to hold s_umount; the fact that we
are in the middle of the write(2) system call will keep the superblock
pinned.  Unfortunately writeback_inodes_sb() checks to make sure
s_umount is taken, and the VFS uses a different mechanism for making
sure the file system doesn't get unmounted out from under us.  The
simplest way of dealing with this is to just simply grab s_umount
using a trylock, and skip kicking the writeback flusher thread in the
very unlikely case that we can't take a read lock on s_umount without
blocking.

Also, we now check the cirteria for kicking the writeback thread
before we decide to whether to fall back to non-delayed writeback, so
if there are any outstanding delayed allocation writes, we try to get
them resolved as soon as possible.

   [ INFO: possible circular locking dependency detected ]
   3.6.0-rc1-00042-gce894ca #367 Not tainted
   -------------------------------------------------------
   dd/8298 is trying to acquire lock:
    (&type->s_umount_key#18){++++..}, at: [<c02277d4>] writeback_inodes_sb_if_idle+0x28/0x46

   but task is already holding lock:
    (&sb->s_type->i_mutex_key#8){+.+...}, at: [<c01ddcce>] generic_file_aio_write+0x5f/0xd3

   which lock already depends on the new lock.

   2 locks held by dd/8298:
    #0:  (sb_writers#2){.+.+.+}, at: [<c01ddcc5>] generic_file_aio_write+0x56/0xd3
    #1:  (&sb->s_type->i_mutex_key#8){+.+...}, at: [<c01ddcce>] generic_file_aio_write+0x5f/0xd3

   stack backtrace:
   Pid: 8298, comm: dd Not tainted 3.6.0-rc1-00042-gce894ca #367
   Call Trace:
    [<c015b79c>] ? console_unlock+0x345/0x372
    [<c06d62a1>] print_circular_bug+0x190/0x19d
    [<c019906c>] __lock_acquire+0x86d/0xb6c
    [<c01999db>] ? mark_held_locks+0x5c/0x7b
    [<c0199724>] lock_acquire+0x66/0xb9
    [<c02277d4>] ? writeback_inodes_sb_if_idle+0x28/0x46
    [<c06db935>] down_read+0x28/0x58
    [<c02277d4>] ? writeback_inodes_sb_if_idle+0x28/0x46
    [<c02277d4>] writeback_inodes_sb_if_idle+0x28/0x46
    [<c026f3b2>] ext4_nonda_switch+0xe1/0xf4
    [<c0271ece>] ext4_da_write_begin+0x27/0x193
    [<c01dcdb0>] generic_file_buffered_write+0xc8/0x1bb
    [<c01ddc47>] __generic_file_aio_write+0x1dd/0x205
    [<c01ddce7>] generic_file_aio_write+0x78/0xd3
    [<c026d336>] ext4_file_write+0x480/0x4a6
    [<c0198c1d>] ? __lock_acquire+0x41e/0xb6c
    [<c0180944>] ? sched_clock_cpu+0x11a/0x13e
    [<c01967e9>] ? trace_hardirqs_off+0xb/0xd
    [<c018099f>] ? local_clock+0x37/0x4e
    [<c0209f2c>] do_sync_write+0x67/0x9d
    [<c0209ec5>] ? wait_on_retry_sync_kiocb+0x44/0x44
    [<c020a7b9>] vfs_write+0x7b/0xe6
    [<c020a9a6>] sys_write+0x3b/0x64
    [<c06dd4bd>] syscall_call+0x7/0xb
Signed-off-by: N"Theodore Ts'o" <tytso@mit.edu>
Cc: stable@vger.kernel.org

The function scans @delaying_queue and stops at the first inode
whose dirtied_when is after *work->older_than_this. So the expired
ones being moved are those before *work->older_than_this. Correct
the comment here.
Reviewed-by: NJan Kara <jack@suse.cz>
Signed-off-by: NWang Sheng-Hui <shhuiw@gmail.com>

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: NWanpeng Li <liwp@linux.vnet.ibm.com>
Cc: Wu Fengguang <fengguang.wu@intel.com>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

In principle, a filesystem may want to have ->sync_fs() called during sync(1)
although it does not have a bdi (i.e. s_bdi is set to noop_backing_dev_info).
Only writeback code really needs bdi set to something reasonable. So move the
checks where they are more logical.
Reviewed-by: NChristoph Hellwig <hch@lst.de>
Signed-off-by: NJan Kara <jack@suse.cz>
Signed-off-by: NAl Viro <viro@zeniv.linux.org.uk>

Signed-off-by: NWanpeng Li <liwp@linux.vnet.ibm.com>
Signed-off-by: NFengguang Wu <fengguang.wu@intel.com>

Fix bug introduced by 169ebd90.  We have to have wb_list_lock locked when
restarting writeback loop after having waited for inode writeback.

Bug description by Ted Tso:

  I can reproduce this fairly easily by using ext4 w/o a journal, running
  under KVM with 1024megs memory, with fsstress (xfstests #13):

  [   45.153294] =====================================
  [   45.154784] [ BUG: bad unlock balance detected! ]
  [   45.155591] 3.5.0-rc1-00002-gb22b1f17 #124 Not tainted
  [   45.155591] -------------------------------------
  [   45.155591] flush-254:16/2499 is trying to release lock (&(&wb->list_lock)->rlock) at:
  [   45.155591] [<c022c3da>] writeback_sb_inodes+0x160/0x327
  [   45.155591] but there are no more locks to release!
Reported-by: NTheodore Ts'o <tytso@mit.edu>
Tested-by: NTheodore Ts'o <tytso@mit.edu>
Signed-off-by: NJan Kara <jack@suse.cz>
Signed-off-by: NFengguang Wu <fengguang.wu@intel.com>

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: NJan Kara <jack@suse.cz>
Signed-off-by: NFengguang Wu <fengguang.wu@intel.com>

The code in writeback_single_inode() is relatively complex. The list requeing
logic makes sense only for flusher thread but not really for sync_inode() or
write_inode_now() callers. Also when we want to get rid of inode references
held by flusher thread, we will need a special I_SYNC handling there.

So separate part of writeback_single_inode() which does the real writeback work
into __writeback_single_inode() and make writeback_single_inode() do only stuff
necessary for callers writing only one inode, moving the special list handling
into writeback_sb_inodes(). As a sideeffect this fixes a possible race where we
could skip some inode during sync(2) because other writer refiled it from b_io
to b_dirty list. Also I_SYNC handling is moved into the callers of
__writeback_single_inode() to make locking easier.
Reviewed-by: NChristoph Hellwig <hch@lst.de>
Signed-off-by: NJan Kara <jack@suse.cz>
Signed-off-by: NFengguang Wu <fengguang.wu@intel.com>

writeback_single_inode() doesn't need wb->list_lock for anything on entry now.
So remove the requirement. This makes locking of writeback_single_inode()
temporarily awkward (entering with i_lock, returning with i_lock and
wb->list_lock) but it will be sanitized in the next patch.

Also inode_wait_for_writeback() doesn't need wb->list_lock for anything. It was
just taking it to make usage convenient for callers but with
writeback_single_inode() changing it's not very convenient anymore. So remove
the lock from that function.
Reviewed-by: NChristoph Hellwig <hch@lst.de>
Signed-off-by: NJan Kara <jack@suse.cz>
Signed-off-by: NFengguang Wu <fengguang.wu@intel.com>

Move inode requeueing after inode has been written out into a separate
function.
Reviewed-by: NChristoph Hellwig <hch@lst.de>
Signed-off-by: NJan Kara <jack@suse.cz>
Signed-off-by: NFengguang Wu <fengguang.wu@intel.com>

Instead of clearing I_DIRTY_PAGES and resetting it when we didn't succeed in
writing them all, just clear the bit only when we succeeded writing all the
pages. We also move the clearing of the bit close to other i_state handling to
separate it from writeback list handling. This is desirable because list
handling will differ for flusher thread and other writeback_single_inode()
callers in future. No filesystem plays any tricks with I_DIRTY_PAGES (like
checking it in ->writepages or ->write_inode implementation) so this movement
is safe.
Signed-off-by: NJan Kara <jack@suse.cz>
Signed-off-by: NFengguang Wu <fengguang.wu@intel.com>

When writeback_single_inode() is called on inode which has I_SYNC already
set while doing WB_SYNC_NONE, inode is moved to b_more_io list. However
this makes sense only if the caller is flusher thread. For other callers of
writeback_single_inode() it doesn't really make sense and may be even wrong
- flusher thread may be doing WB_SYNC_ALL writeback in parallel.

So we move requeueing from writeback_single_inode() to writeback_sb_inodes().
Reviewed-by: NChristoph Hellwig <hch@lst.de>
Signed-off-by: NJan Kara <jack@suse.cz>
Signed-off-by: NFengguang Wu <fengguang.wu@intel.com>

Move clearing of I_SYNC into inode_sync_complete().  It is more logical to have
clearing of I_SYNC bit and waking of waiters in one place. Also later we will
have two places needing to clear I_SYNC and wake up waiters so this allows them
to use the common helper. Moving of I_SYNC clearing to a later stage of
writeback_single_inode() is safe since we hold i_lock all the time.
Reviewed-by: NChristoph Hellwig <hch@lst.de>
Signed-off-by: NJan Kara <jack@suse.cz>
Signed-off-by: NFengguang Wu <fengguang.wu@intel.com>

The comment is hopelessly outdated and misplaced. We no longer have 'bdi'
part of writeback work, the comment about blockdev super is outdated,
comment about throttling as well. Information about list handling is in
more detail at queue_io(). So just move the bit about older_than_this to
close to move_expired_inodes() and remove the rest.
Reviewed-by: NChristoph Hellwig <hch@lst.de>
Signed-off-by: NJan Kara <jack@suse.cz>
Signed-off-by: NFengguang Wu <fengguang.wu@intel.com>

inode_sync_wait() in write_inode_now() is just bogus. That function waits for
I_SYNC bit to be cleared but writeback_single_inode() clears the bit on return
so the wait is effectivelly a nop unless someone else submits the inode for
writeback again. All the waiting write_inode_now() needs is achieved by using
WB_SYNC_ALL writeback mode.
Signed-off-by: NJan Kara <jack@suse.cz>
Signed-off-by: NChristoph Hellwig <hch@lst.de>
Signed-off-by: NFengguang Wu <fengguang.wu@intel.com>

Signed-off-by: NFengguang Wu <fengguang.wu@intel.com>
Signed-off-by: NJiri Kosina <jkosina@suse.cz>