Signed-off-by: NGu Zheng <guz.fnst@cn.fujitsu.com>
Signed-off-by: NJaegeuk Kim <jaegeuk.kim@samsung.com>

Integrated a couple of minor changes for better readability suggested by
Chao Yu.
Signed-off-by: NGu Zheng <guz.fnst@cn.fujitsu.com>
Reviewed-by: NChao Yu <chao2.yu@samsung.com>
Signed-off-by: NJaegeuk Kim <jaegeuk.kim@samsung.com>

This patch includes a simple fix to adjust coding style.
Signed-off-by: NJaegeuk Kim <jaegeuk.kim@samsung.com>

This patch fixes performance regression of dbench reported by
Alex <hbx7d@yandex.com>.

This issue was revealed by Phoronix tests results:
http://www.phoronix.com/scan.php?page=article&item=linux_314_ssdfs&num=2

It turns out that we need to assign WRITE_SYNC to the node writes, if
fsync is triggered.

The performance numbers are like below, which is measured by Alex.
1. 355MB/s       ext4
2. 225MB/s       f2fs : WRITE for node writes
3. 525MB/s       f2fs : WRITE_SYNC for node writes

Reported-And-Tested-by: Alex <hbx7d@yandex.com>.
Signed-off-by: NJaegeuk Kim <jaegeuk.kim@samsung.com>

We should de-account dirty counters for page when redirty in ->writepage().

Wu Fengguang described in 'commit 971767ca':
"writeback: fix dirtied pages accounting on redirty
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)."
Signed-off-by: NChao Yu <chao2.yu@samsung.com>
Signed-off-by: NJaegeuk Kim <jaegeuk.kim@samsung.com>

This patch use existing macro F2FS_INODE/NEXT_FREE_BLKADDR to clean up some
codes.
Signed-off-by: NChao Yu <chao2.yu@samsung.com>
Signed-off-by: NJaegeuk Kim <jaegeuk.kim@samsung.com>

If there are multi segments in one section, we will read those SSA blocks which
have contiguous address one by one in f2fs_gc. It may lost performance, let's
read ahead SSA blocks by merge multi read request.
Signed-off-by: NChao Yu <chao2.yu@samsung.com>
Signed-off-by: NJaegeuk Kim <jaegeuk.kim@samsung.com>

This patch adds an sysfs entry to control dir_level used by the large directory.

The description of this entry is:

 dir_level                    This parameter controls the directory level to
			      support large directory. If a directory has a
			      number of files, it can reduce the file lookup
			      latency by increasing this dir_level value.
			      Otherwise, it needs to decrease this value to
			      reduce the space overhead. The default value is 0.
Signed-off-by: NJaegeuk Kim <jaegeuk.kim@samsung.com>

This patch introduces an i_dir_level field to support large directory.

Previously, f2fs maintains multi-level hash tables to find a dentry quickly
from a bunch of chiild dentries in a directory, and the hash tables consist of
the following tree structure as below.

In Documentation/filesystems/f2fs.txt,

----------------------
A : bucket
B : block
N : MAX_DIR_HASH_DEPTH
----------------------

level #0   | A(2B)
           |
level #1   | A(2B) - A(2B)
           |
level #2   | A(2B) - A(2B) - A(2B) - A(2B)
     .     |   .       .       .       .
level #N/2 | A(2B) - A(2B) - A(2B) - A(2B) - A(2B) - ... - A(2B)
     .     |   .       .       .       .
level #N   | A(4B) - A(4B) - A(4B) - A(4B) - A(4B) - ... - A(4B)

But, if we can guess that a directory will handle a number of child files,
we don't need to traverse the tree from level #0 to #N all the time.
Since the lower level tables contain relatively small number of dentries,
the miss ratio of the target dentry is likely to be high.

In order to avoid that, we can configure the hash tables sparsely from level #0
like this.

level #0   | A(2B) - A(2B) - A(2B) - A(2B)

level #1   | A(2B) - A(2B) - A(2B) - A(2B) - A(2B) - ... - A(2B)
     .     |   .       .       .       .
level #N/2 | A(2B) - A(2B) - A(2B) - A(2B) - A(2B) - ... - A(2B)
     .     |   .       .       .       .
level #N   | A(4B) - A(4B) - A(4B) - A(4B) - A(4B) - ... - A(4B)

With this structure, we can skip the ineffective tree searches in lower level
hash tables.

This patch adds just a facility for this by introducing i_dir_level in
f2fs_inode.
Signed-off-by: NJaegeuk Kim <jaegeuk.kim@samsung.com>

It turns out that a bit operation like find_next_bit is not always fast enough
for f2fs_find_entry.
Instead, it is pretty much simple and fast to traverse each dentries.
Signed-off-by: NJaegeuk Kim <jaegeuk.kim@samsung.com>

The stat_show is just to show the current status of f2fs.
So, we can remove all the there-in locks.
Signed-off-by: NJaegeuk Kim <jaegeuk.kim@samsung.com>

This patch introduces a radix tree for the list of free_nids, which enhances
the performance on free nid management.
Signed-off-by: NJaegeuk Kim <jaegeuk.kim@samsung.com>

Introduce help macro on_build_free_nids() which just uses build_lock
to judge whether the building free nid is going, so that we can remove
the on_build_free_nids field from f2fs_sb_info.
Signed-off-by: NGu Zheng <guz.fnst@cn.fujitsu.com>
[Jaegeuk Kim: remove an unnecessary white line removal]
Signed-off-by: NJaegeuk Kim <jaegeuk.kim@samsung.com>

The nat cache entry maintains a status whether it is checkpointed or not.
So, if a new cache entry is loaded from the last checkpoint,
nat_entry->checkpointed should be true.
If the cache entry is modified as being dirty, nat_entry->checkpoint should
be false.
Signed-off-by: NJaegeuk Kim <jaegeuk.kim@samsung.com>

At the end of the recovery procedure, write_checkpoint is called and updates
the cp count which is managed by f2fs stat.
But, previously build_stat() is called after the recovery procedure, which
results in:

BUG: unable to handle kernel NULL pointer dereference at 000000000000012c
IP: [<ffffffffa03b1030>] write_checkpoint+0x720/0xbc0 [f2fs]
Call Trace:
 [<ffffffff810a6b44>] ? mark_held_locks+0x74/0x140
 [<ffffffff8109a3e0>] ? __init_waitqueue_head+0x60/0x60
 [<ffffffffa03bf036>] recover_fsync_data+0x656/0xf20 [f2fs]
 [<ffffffff812ee3eb>] ? security_d_instantiate+0x1b/0x30
 [<ffffffffa03aeb4d>] f2fs_fill_super+0x94d/0xa00 [f2fs]
 [<ffffffff811a9825>] mount_bdev+0x1a5/0x1f0
 [<ffffffff8114915e>] ? __get_free_pages+0xe/0x40
 [<ffffffffa03ae200>] ? f2fs_remount+0x130/0x130 [f2fs]
 [<ffffffffa03aa575>] f2fs_mount+0x15/0x20 [f2fs]
 [<ffffffff811aa713>] mount_fs+0x43/0x1b0
 [<ffffffff811c7124>] vfs_kern_mount+0x74/0x160
 [<ffffffff811c5cb1>] ? __get_fs_type+0x51/0x60
 [<ffffffff811c9727>] do_mount+0x237/0xb50
 [<ffffffff811c936a>] ? copy_mount_options+0x3a/0x170

So, this patche changes the order of recovery_fsync_data() and
f2fs_build_stats().
Signed-off-by: NJaegeuk Kim <jaegeuk.kim@samsung.com>

Even if f2fs_write_data_page is called by the page reclaiming path, we should
not write the page to provide enough free segments for the worst case scenario.
Otherwise, f2fs can face with no free segment while gc is conducted, resulting
in:

 ------------[ cut here ]------------
 kernel BUG at /home/zeus/f2fs_test/src/fs/f2fs/segment.c:565!
 RIP: 0010:[<ffffffffa02c3b11>]  [<ffffffffa02c3b11>] new_curseg+0x331/0x340 [f2fs]
 Call Trace:
  allocate_segment_by_default+0x204/0x280 [f2fs]
  allocate_data_block+0x108/0x210 [f2fs]
  write_data_page+0x8a/0xc0 [f2fs]
  do_write_data_page+0xe1/0x2a0 [f2fs]
  move_data_page+0x8a/0xf0 [f2fs]
  f2fs_gc+0x446/0x970 [f2fs]
  f2fs_balance_fs+0xb6/0xd0 [f2fs]
  f2fs_write_begin+0x50/0x350 [f2fs]
  ? unlock_page+0x27/0x30
  ? unlock_page+0x27/0x30
  generic_file_buffered_write+0x10a/0x280
  ? file_update_time+0xa3/0xf0
  __generic_file_aio_write+0x1c8/0x3d0
  ? generic_file_aio_write+0x52/0xb0
  ? generic_file_aio_write+0x52/0xb0
  generic_file_aio_write+0x65/0xb0
  do_sync_write+0x5a/0x90
  vfs_write+0xc5/0x1f0
  SyS_write+0x55/0xa0
  system_call_fastpath+0x16/0x1b
Signed-off-by: NJaegeuk Kim <jaegeuk.kim@samsung.com>

Total nids that f2fs can use should not include 0, nid for node inode, and nid
for meta inode.
Signed-off-by: NJaegeuk Kim <jaegeuk.kim@samsung.com>

This patch shows the counts of checkpoint in f2fs' status.
Signed-off-by: NChangman Lee <cm224.lee@samsung.com>
Signed-off-by: NJaegeuk Kim <jaegeuk.kim@samsung.com>

This patch help us to cleanup the readahead code by merging ra_{sit,nat}_pages
function into ra_meta_pages.
Additionally the new function is used to readahead cp block in
recover_orphan_inodes.

Change log from v1:
 o fix a deadloop bug pointed by Jaegeuk Kim.
Signed-off-by: NChao Yu <chao2.yu@samsung.com>
Signed-off-by: NJaegeuk Kim <jaegeuk.kim@samsung.com>

Previously without protection of inode mutex, f2fs_falloc and other data
correlated operations will interfere with each other.
So let's use inode mutex to keep atomicity of f2fs_falloc.
Signed-off-by: NChao Yu <chao2.yu@samsung.com>
Signed-off-by: NJaegeuk Kim <jaegeuk.kim@samsung.com>

This patch integrates inode_[inc|dec]_dirty_dents with inc_page_count to remove
redundant calls.
Signed-off-by: NJaegeuk Kim <jaegeuk.kim@samsung.com>

If f2fs entered errorneous checkpoint status, it should skip writing meta
pages instead of redirtying the pages out.
Otherwise, it cannot unmount the partition even though f2fs is under read-only
status.
Signed-off-by: NJaegeuk Kim <jaegeuk.kim@samsung.com>

When a new directory is allocated, if an error is occurred, we should truncate
preallocated dentry pages too.

This bug was reported by Andrey Tsyvarev after a while as follows.

mkdir()->
 f2fs_add_link()->
  init_inode_metadata()->
    f2fs_init_acl()->
      f2fs_get_acl()->
        f2fs_getxattr()->
          read_all_xattrs() fails.

Also there was a BUG_ON triggered after the fault in
mkdir()->
 f2fs_add_link()->
   init_inode_metadata()->
    remove_inode_page() ->
      f2fs_bug_on(inode->i_blocks != 0 && inode->i_blocks != 1);

But, previous patch wasn't perfect to resolve that bug, so the following bug
report was also submitted.

kernel BUG at fs/f2fs/inode.c:274!
Call Trace:
 [<ffffffff811fde03>] evict+0xa3/0x1a0
 [<ffffffff811fe615>] iput+0xf5/0x180
 [<ffffffffa01c7f63>] f2fs_mkdir+0xf3/0x150 [f2fs]
 [<ffffffff811f2a77>] vfs_mkdir+0xb7/0x160
 [<ffffffff811f36bf>] SyS_mkdir+0x5f/0xc0
 [<ffffffff81680769>] system_call_fastpath+0x16/0x1b

Finally, this patch resolves all the issues like below.

If an error is occurred after make_empty_dir(),
 1. truncate_inode_pages()
   The make_bad_inode() prior to iput() will change i_mode to S_IFREG, which
   means that f2fs will not decrement fi->dirty_dents during f2fs_evict_inode.
   But, by calling it here, we can do that.

 2. truncate_blocks()
   Preallocated dentry pages are trucated here to sync i_blocks.

 3. remove_dirty_dir_inode()
   Remove this directory inode from the list.
Reported-and-Tested-by: NAndrey Tsyvarev <tsyvarev@ispras.ru>
Signed-off-by: NJaegeuk Kim <jaegeuk.kim@samsung.com>

This patch modifies flow a little bit to avoid the following build warnings.

src/fs/f2fs/recovery.c: In function ‘check_index_in_prev_nodes’:
src/fs/f2fs/recovery.c:288:51: warning: ‘sum.<U5390>.<U52f8>.ofs_in_node’ may
	be used uninitialized in this function [-Wmaybe-uninitialized]
src/fs/f2fs/recovery.c:260:23: warning: ‘sum.nid’ may be used uninitialized
	in this function [-Wmaybe-uninitialized]
Signed-off-by: NJaegeuk Kim <jaegeuk.kim@samsung.com>

This patch adds GET_BLKOFF_FROM_SEG0 to clean up some codes.
Signed-off-by: NJaegeuk Kim <jaegeuk.kim@samsung.com>

This is the erroneous scenario.

                             i_size    on-disk i_size    i_blocks
__f2fs_add_link()             4096           4096           2
 get_new_data_page            8192           4096           3
 -ENOSPC = init_inode_metadata
 checkpoint                     -            4096           3
 POR and reboot

__f2fs_add_link()             4096           4096           3
 page = get_new_data_page (page->index = 1 by NEW_ADDR)
 add a dentry to the page successfully

f2fs_rmdir()
 f2fs_empty_dir()             4096           4096           3
 f2fs_unlink() goes, since there is no valid dentry due to i_size = 4096.
 But, still there is one dentry in page->index = 1.

So this patch moves the code to write dir->i_size into on-disk i_size in order
to sync dir's i_size, on-disk i_size, and its i_blocks.
Signed-off-by: NJaegeuk Kim <jaegeuk.kim@samsung.com>

This patch modifies the use of bi_private to remove pointer chasing for sbi.
Previously, we had a bi_private structure, but it needs memory allocation.
So this patch uses bi_private by the sbi pointer and adds a completion pointer
into the sbi.
This can achieve no memory allocation and nice use of the bi_private.
Signed-off-by: NJaegeuk Kim <jaegeuk.kim@samsung.com>

If a new xattr node page was allocated and its inode is fsynced, we should
recover the xattr node page during the roll-forward process after power-cut.
But, previously, f2fs didn't handle that case, resulting in kernel panic as
follows reported by Tom Li.

BUG: unable to handle kernel paging request at ffffc9001c861a98
IP: [<ffffffffa0295236>] check_index_in_prev_nodes+0x86/0x2d0 [f2fs]
Call Trace:
 [<ffffffff815ece9b>] ? printk+0x48/0x4a
 [<ffffffffa029626a>] recover_fsync_data+0xdca/0xf50 [f2fs]
 [<ffffffffa02873ae>] f2fs_fill_super+0x92e/0x970 [f2fs]
 [<ffffffff8112c9f8>] mount_bdev+0x1b8/0x200
 [<ffffffffa0286a80>] ? f2fs_remount+0x130/0x130 [f2fs]
 [<ffffffffa0285e40>] f2fs_mount+0x10/0x20 [f2fs]
 [<ffffffff8112d4de>] mount_fs+0x3e/0x1b0
 [<ffffffff810ef4eb>] ? __alloc_percpu+0xb/0x10
 [<ffffffff8114761f>] vfs_kern_mount+0x6f/0x120
 [<ffffffff811497b9>] do_mount+0x259/0xa90
 [<ffffffff810ead1d>] ? memdup_user+0x3d/0x80
 [<ffffffff810eadb3>] ? strndup_user+0x53/0x70
 [<ffffffff8114a2c9>] SyS_mount+0x89/0xd0
 [<ffffffff815feae2>] system_call_fastpath+0x16/0x1b

This patch adds a recovery function of xattr node pages.
Reported-by: NTom Li <biergaizi@members.fsf.org>
Signed-off-by: NJaegeuk Kim <jaegeuk.kim@samsung.com>

This patch cleans up the refresh_sit_entry to handle locate_dirty_segments.
Signed-off-by: NJaegeuk Kim <jaegeuk.kim@samsung.com>

In order to make fs consistency, update_inode_page should not be failed all
the time. Otherwise, it is possible to lose some metadata in the inode like
a link count.
Signed-off-by: NJaegeuk Kim <jaegeuk.kim@samsung.com>

Pull btrfs fixes from Chris Mason:
 "We have a small collection of fixes in my for-linus branch.

  The big thing that stands out is a revert of a new ioctl.  Users
  haven't shipped yet in btrfs-progs, and Dave Sterba found a better way
  to export the information"

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mason/linux-btrfs:
  Btrfs: use right clone root offset for compressed extents
  btrfs: fix null pointer deference at btrfs_sysfs_add_one+0x105
  Btrfs: unset DCACHE_DISCONNECTED when mounting default subvol
  Btrfs: fix max_inline mount option
  Btrfs: fix a lockdep warning when cleaning up aborted transaction
  Revert "btrfs: add ioctl to export size of global metadata reservation"

Pull devicetree fixes from Rob Herring:
 "Fix booting on PPC boards.  Changes to of_match_node matching caused
  the serial port on some PPC boards to stop working.  Reverted the
  change and reimplement to split matching between new style compatible
  only matching and fallback to old matching algorithm"

* tag 'dt-fixes-for-3.14' of git://git.kernel.org/pub/scm/linux/kernel/git/robh/linux:
  of: search the best compatible match first in __of_match_node()
  Revert "OF: base: match each node compatible against all given matches first"

Currently, of_match_node compares each given match against all node's
compatible strings with of_device_is_compatible.

To achieve multiple compatible strings per node with ordering from
specific to generic, this requires given matches to be ordered from
specific to generic. For most of the drivers this is not true and also
an alphabetical ordering is more sane there.

Therefore, this patch introduces a function to match each of the node's
compatible strings against all given compatible matches without type and
name first, before checking the next compatible string. This implies
that node's compatibles are ordered from specific to generic while
given matches can be in any order. If we fail to find such a match
entry, then fall-back to the old method in order to keep compatibility.

Cc: Sebastian Hesselbarth <sebastian.hesselbarth@gmail.com>
Signed-off-by: NKevin Hao <haokexin@gmail.com>
Tested-by: NStephen Chivers <schivers@csc.com>
Signed-off-by: NRob Herring <robh@kernel.org>

Pull SCSI target fixes from Nicholas Bellinger:
 "Mostly minor fixes this time to v3.14-rc1 related changes.  Also
  included is one fix for a free after use regression in persistent
  reservations UNREGISTER logic that is CC'ed to >= v3.11.y stable"

* git://git.kernel.org/pub/scm/linux/kernel/git/nab/target-pending:
  Target/sbc: Fix protection copy routine
  IB/srpt: replace strict_strtoul() with kstrtoul()
  target: Simplify command completion by removing CMD_T_FAILED flag
  iser-target: Fix leak on failure in isert_conn_create_fastreg_pool
  iscsi-target: Fix SNACK Type 1 + BegRun=0 handling
  target: Fix missing length check in spc_emulate_evpd_83()
  qla2xxx: Remove last vestiges of qla_tgt_cmd.cmd_list
  target: Fix 32-bit + CONFIG_LBDAF=n link error w/ sector_div
  target: Fix free-after-use regression in PR unregister

Pull i2c fixes from Wolfram Sang:
 "i2c has a bugfix and documentation improvements for you"

* 'i2c/for-current' of git://git.kernel.org/pub/scm/linux/kernel/git/wsa/linux:
  Documentation: i2c: mention ACPI method for instantiating devices
  Documentation: i2c: describe devicetree method for instantiating devices
  i2c: mv64xxx: refactor message start to ensure proper initialization

Merge branches 'irq-urgent-for-linus' and 'irq-core-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip

Pull irq update from Thomas Gleixner:
 "Fix from the urgent branch: a trivial oneliner adding the missing
  Kconfig dependency curing build failures which have been discovered by
  several build robots.

  The update in the irq-core branch provides a new function in the
  irq/devres code, which is a prerequisite for driver developers to get
  rid of boilerplate code all over the place.

  Not a bugfix, but it has zero impact on the current kernel due to the
  lack of users.  It's simpler to provide the infrastructure to
  interested parties via your tree than fulfilling the wishlist of
  driver maintainers on which particular commit or tag this should be
  based on"

* 'irq-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  genirq: Add missing irq_to_desc export for CONFIG_SPARSE_IRQ=n

* 'irq-core-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  genirq: Add devm_request_any_context_irq()

Pull timer fixes from Thomas Gleixner:
 "The following trilogy of patches brings you:

   - fix for a long standing math overflow issue with HZ < 60

   - an onliner fix for a corner case in the dreaded tick broadcast
     mechanism affecting a certain range of AMD machines which are
     infested with the infamous automagic C1E power control misfeature

   - a fix for one of the ARM platforms which allows the kernel to
     proceed and boot instead of stupidly panicing for no good reason.
     The patch is slightly larger than necessary, but it's less ugly
     than the alternative 5 liner"

* 'timers-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  tick: Clear broadcast pending bit when switching to oneshot
  clocksource: Kona: Print warning rather than panic
  time: Fix overflow when HZ is smaller than 60

Merge tag 'trace-fixes-v3.14-rc2' of git://git.kernel.org/pub/scm/linux/kernel/git/rostedt/linux-trace

Pull twi tracing fixes from Steven Rostedt:
 "Two urgent fixes in the tracing utility.

  The first is a fix for the way the ring buffer stores timestamps.
  After a restructure of the code was done, the ring buffer timestamp
  logic missed the fact that the first event on a sub buffer is to have
  a zero delta, as the full timestamp is stored on the sub buffer
  itself.  But because the delta was not cleared to zero, the timestamp
  for that event will be calculated as the real timestamp + the delta
  from the last timestamp.  This can skew the timestamps of the events
  and have them say they happened when they didn't really happen.
  That's bad.

  The second fix is for modifying the function graph caller site.  When
  the stop machine was removed from updating the function tracing code,
  it missed updating the function graph call site location.  It is still
  modified as if it is being done via stop machine.  But it's not.  This
  can lead to a GPF and kernel crash if the function graph call site
  happens to lie between cache lines and one CPU is executing it while
  another CPU is doing the update.  It would be a very hard condition to
  hit, but the result is severe enough to have it fixed ASAP"

* tag 'trace-fixes-v3.14-rc2' of git://git.kernel.org/pub/scm/linux/kernel/git/rostedt/linux-trace:
  ftrace/x86: Use breakpoints for converting function graph caller
  ring-buffer: Fix first commit on sub-buffer having non-zero delta

Pull x86 EFI fixes from Peter Anvin:
 "A few more EFI-related fixes"

* 'x86-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  x86/efi: Check status field to validate BGRT header
  x86/efi: Fix 32-bit fallout