f2fs: let checkpoint flush dnode page of regular
authorChao Yu <yuchao0@huawei.com>
Wed, 25 Jul 2018 11:16:21 +0000 (19:16 +0800)
committerJaegeuk Kim <jaegeuk@kernel.org>
Wed, 1 Aug 2018 18:52:36 +0000 (11:52 -0700)
commitfd8c8caf7e7c8261a92ce0f7f2cd0adb8afd9e0d
treef538e2fbc1c2f2ff34e239f0b9b46d8f7103dbcd
parentad6672bbc527727dc8968e8d92687f55ae928ce5
f2fs: let checkpoint flush dnode page of regular

Fsyncer will wait on all dnode pages of regular writeback before flushing,
if there are async dnode pages blocked by IO scheduler, it may decrease
fsync's performance.

In this patch, we choose to let f2fs_balance_fs_bg() to trigger checkpoint
to flush these dnode pages of regular, so async IO of dnode page can be
elimitnated, making fsyncer only need to wait for sync IO.

Signed-off-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
fs/f2fs/node.c
fs/f2fs/node.h
fs/f2fs/segment.c