memcg: oom notifier at el. (v3)

Updated against mmotm-Mar9.

This patch set's feature is
 - add filter to memcg's oom waitq.
 - oom kill notifier for memcg.
 - oom kill disable for memcg.

Major changes since previous one are
 - add filter to wakeup queue.
 - use its own function and logic rather than reusing thresholds.
 - some minor fixes.

If oom-killer disabled, all tasks under memcg will sleep in memcg_oom_waitq.
What users can do when memcg-oom-killer is disabled is:
 - enlarge limit.
 - kill some task. ---(*)
 - move some task to other cgroup. (with account migration)
   (This patchset doesn't handle a case when account migraion isn't set.)

The benefit of (*) is that the user can save information of all tasks before
killing and he can take coredump (by gcore.) of troublesome process.

I'm now wondering when I remove RFC...but I think this will not have
much HUNK with dirty_ratio sets.

If some codes are unclear, feel free to request me.

Thanks,
-Kame

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/