]> bbs.cooldavid.org Git - net-next-2.6.git/commit
mm: compaction: handle active and inactive fairly in too_many_isolated
authorMinchan Kim <minchan.kim@gmail.com>
Thu, 9 Sep 2010 23:38:00 +0000 (16:38 -0700)
committerLinus Torvalds <torvalds@linux-foundation.org>
Fri, 10 Sep 2010 01:57:24 +0000 (18:57 -0700)
commitbc6930457460788e14b2c0808ed4632a1592bd61
tree78e562bc90b3b99f1ba1dbc32c846127f10a67e3
parent1c24de60e50fb19b94d94225458da17c720f0729
mm: compaction: handle active and inactive fairly in too_many_isolated

Iram reported that compaction's too_many_isolated() loops forever.
(http://www.spinics.net/lists/linux-mm/msg08123.html)

The meminfo when the situation happened was inactive anon is zero.  That's
because the system has no memory pressure until then.  While all anon
pages were in the active lru, compaction could select active lru as well
as inactive lru.  That's a different thing from vmscan's isolated.  So we
has been two too_many_isolated.

While compaction can isolate pages in both active and inactive, current
implementation of too_many_isolated only considers inactive.  It made
Iram's problem.

This patch handles active and inactive fairly.  That's because we can't
expect where from and how many compaction would isolated pages.

This patch changes (nr_isolated > nr_inactive) with
nr_isolated > (nr_active + nr_inactive) / 2.

Signed-off-by: Minchan Kim <minchan.kim@gmail.com>
Reported-by: Iram Shahzad <iram.shahzad@jp.fujitsu.com>
Acked-by: Mel Gorman <mel@csn.ul.ie>
Acked-by: Wu Fengguang <fengguang.wu@intel.com>
Cc: <stable@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
mm/compaction.c