diff options
| author | Miao Xie <miaox@cn.fujitsu.com> | 2013-01-29 10:09:20 (GMT) |
|---|---|---|
| committer | Josef Bacik <jbacik@fusionio.com> | 2013-02-20 17:59:04 (GMT) |
| commit | e2d845211eda9cf296e8edf6724b3d541f4fbfd5 (patch) | |
| tree | 69acdf4bfe81d173673fee8d8104393e4b73681e /fs/btrfs/extent_io.c | |
| parent | c018daecead7a46a575e2a1397fea850b83396c8 (diff) | |
| download | linux-fsl-qoriq-e2d845211eda9cf296e8edf6724b3d541f4fbfd5.tar.xz | |
Btrfs: use percpu counter for dirty metadata count
->dirty_metadata_bytes is accessed very frequently, so use percpu
counter instead of the u64 variant to reduce the contention of
the lock.
This patch also fixed the problem that we access it without
lock protection in __btrfs_btree_balance_dirty(), which may
cause we skip the dirty pages flush.
Signed-off-by: Miao Xie <miaox@cn.fujitsu.com>
Signed-off-by: Josef Bacik <jbacik@fusionio.com>
Diffstat (limited to 'fs/btrfs/extent_io.c')
| -rw-r--r-- | fs/btrfs/extent_io.c | 9 |
1 files changed, 3 insertions, 6 deletions
diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c index bda36fe..b9019c8 100644 --- a/fs/btrfs/extent_io.c +++ b/fs/btrfs/extent_io.c @@ -3122,12 +3122,9 @@ static int lock_extent_buffer_for_io(struct extent_buffer *eb, set_bit(EXTENT_BUFFER_WRITEBACK, &eb->bflags); spin_unlock(&eb->refs_lock); btrfs_set_header_flag(eb, BTRFS_HEADER_FLAG_WRITTEN); - spin_lock(&fs_info->delalloc_lock); - if (fs_info->dirty_metadata_bytes >= eb->len) - fs_info->dirty_metadata_bytes -= eb->len; - else - WARN_ON(1); - spin_unlock(&fs_info->delalloc_lock); + __percpu_counter_add(&fs_info->dirty_metadata_bytes, + -eb->len, + fs_info->dirty_metadata_batch); ret = 1; } else { spin_unlock(&eb->refs_lock); |