Messages in this thread | | | Date | Tue, 15 Mar 2011 11:54:51 +0900 | From | KAMEZAWA Hiroyuki <> | Subject | Re: [PATCH v6 0/9] memcg: per cgroup dirty page accounting |
| |
On Mon, 14 Mar 2011 19:51:22 -0700 Greg Thelen <gthelen@google.com> wrote:
> On Mon, Mar 14, 2011 at 6:56 PM, KAMEZAWA Hiroyuki > <kamezawa.hiroyu@jp.fujitsu.com> wrote: > > On Mon, 14 Mar 2011 11:29:17 -0700 > > Greg Thelen <gthelen@google.com> wrote: > > > >> On Fri, Mar 11, 2011 at 5:10 PM, Andrew Morton > The foreign dirtier issue is all about identifying the memcg (or > possibly multiple bdi) that need balancing. If the foreign dirtier > issue is not important then we can focus on identifying inodes to > writeback that will lower the current's memcg dirty usage. I am fine > ignoring the foreign dirtier issue for now and breaking the problem > into smaller pieces. > ok.
> I think this can be done with out any additional state. Can just scan > the memcg lru to find dirty file pages and thus inodes to pass to > sync_inode(), or some other per-inode writeback routine? >
I think it works, finding inodes to be cleaned by LRU scanning.
Thanks, -Kame
| |