writeback: balance_dirty_pages() shall write more than dirtied pages
authorWu Fengguang <fengguang.wu@intel.com>
Wed, 23 Sep 2009 13:56:00 +0000 (21:56 +0800)
committerJens Axboe <jens.axboe@oracle.com>
Fri, 25 Sep 2009 16:08:24 +0000 (18:08 +0200)
commit3a2e9a5a2afc1a2d2c548b8987f133235cebe933
treecb05d2873b2701ded758a7087de5af2932a97736
parenta5989bdc981ec85e0734ac22519cc0b780813d7b
writeback: balance_dirty_pages() shall write more than dirtied pages

Some filesystem may choose to write much more than ratelimit_pages
before calling balance_dirty_pages_ratelimited_nr(). So it is safer to
determine number to write based on real number of dirtied pages.

Otherwise it is possible that
  loop {
    btrfs_file_write():     dirty 1024 pages
    balance_dirty_pages():  write up to 48 pages (= ratelimit_pages * 1.5)
  }
in which the writeback rate cannot keep up with dirty rate, and the
dirty pages go all the way beyond dirty_thresh.

The increased write_chunk may make the dirtier more bumpy.
So filesystems shall be take care not to dirty too much at
a time (eg. > 4MB) without checking the ratelimit.

Signed-off-by: Wu Fengguang <fengguang.wu@intel.com>
Acked-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
mm/page-writeback.c