Re: [RFC] [PATCH] Cgroup based OOM killer controller

[KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>]

> On Tue, 27 Jan 2009, KOSAKI Motohiro wrote:
> 
> > Confused.
> > 
> > As far as I know, people want the method of flexible cache treating.
> > but oom seems less flexible than userland notification.
> > 
> > Why do you think notification is bad?
> > 
> 
> There're a couple of proposals that have been discussed recently that 
> share some functional behavior.
> 
> One is the cgroup oom notifier that allows you to attach a task to wait on 
> an oom condition for a collection of tasks.  That allows userspace to 
> respond to the condition by droping caches, adding nodes to a cpuset, 
> elevating memory controller limits, sending a signal, etc.  It can also 
> defer to the kernel oom killer as a last resort.
> 
> The other is /dev/mem_notify that allows you to poll() on a device file 
> and be informed of low memory events.  This can include the cgroup oom 
> notifier behavior when a collection of tasks is completely out of memory, 
> but can also warn when such a condition may be imminent.  I suggested that 
> this be implemented as a client of cgroups so that different handlers can 
> be responsible for different aggregates of tasks.
>
> I think the latter is a much more powerful tool and includes all the 
> behavior of the former.  It preserves the oom killer as a last resort for 
> the kernel and defers all preference killing or lowmem responses to 
> userspace.

Yup, indeed. :)
honestly, I talked about the same thingk recently "lowmemory android driver not needed?" thread.