writeback: make bdi_has_dirty_io() take multiple bdi_writeback's into account
authorTejun Heo <tj@kernel.org>
Fri, 22 May 2015 21:13:47 +0000 (17:13 -0400)
committerJens Axboe <axboe@fb.com>
Tue, 2 Jun 2015 14:33:36 +0000 (08:33 -0600)
commit95a46c65e3c09edb9f17dabf2dc16670cd328739
tree2ff8187223f5f5671394dd9ee84719f3ff934a14
parent766a9d6e60578f1ef6de71f89f022084f8bffc82
writeback: make bdi_has_dirty_io() take multiple bdi_writeback's into account

bdi_has_dirty_io() used to only reflect whether the root wb
(bdi_writeback) has dirty inodes.  For cgroup writeback support, it
needs to take all active wb's into account.  If any wb on the bdi has
dirty inodes, bdi_has_dirty_io() should return true.

To achieve that, as inode_wb_list_{move|del}_locked() now keep track
of the dirty state transition of each wb, the number of dirty wbs can
be counted in the bdi; however, bdi is already aggregating
wb->avg_write_bandwidth which can easily be guaranteed to be > 0 when
there are any dirty inodes by ensuring wb->avg_write_bandwidth can't
dip below 1.  bdi_has_dirty_io() can simply test whether
bdi->tot_write_bandwidth is zero or not.

While this bumps the value of wb->avg_write_bandwidth to one when it
used to be zero, this shouldn't cause any meaningful behavior
difference.

bdi_has_dirty_io() is made an inline function which tests whether
->tot_write_bandwidth is non-zero.  Also, WARN_ON_ONCE()'s on its
value are added to inode_wb_list_{move|del}_locked().

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>
fs/fs-writeback.c
include/linux/backing-dev-defs.h
include/linux/backing-dev.h
mm/backing-dev.c
mm/page-writeback.c