Messages in this thread | | | Date | Thu, 8 Mar 2012 14:36:19 +0900 | From | KAMEZAWA Hiroyuki <> | Subject | Re: [PATCH 6/7] mm/memcg: rework inactive_ratio calculation |
| |
On Fri, 02 Mar 2012 10:24:12 +0400 Konstantin Khlebnikov <khlebnikov@openvz.org> wrote:
> KAMEZAWA Hiroyuki wrote: > > On Wed, 29 Feb 2012 13:16:00 +0400 > > Konstantin Khlebnikov<khlebnikov@openvz.org> wrote: > > > >> This patch removes precalculated zone->inactive_ratio. > >> Now it always calculated in inactive_anon_is_low() from current lru size. > >> After that we can merge memcg and non-memcg cases and drop duplicated code. > >> > >> We can drop precalculated ratio, because its calculation fast enough to do it > >> each time. Plus precalculation uses zone size as basis, this estimation not > >> always match with page lru size, for example if a significant proportion > >> of memory occupied by kernel objects.om memory cgroup which is triggered this memory reclaim. > This is more reason > >> > >> Signed-off-by: Konstantin Khlebnikov<khlebnikov@openvz.org> > > > > Maybe good....but please don't change the user interface /proc/zoneinfo implicitly. > > How about calculating inactive_ratio at reading /proc/zoneinfo ? > > I don't know... Anybody need this?
I don't like changes in user interface without any caution in feature-removal-schedule.txt I just don't Ack. If others say ok, please go ahead with them.
Thanks, -Kame
| |