From mboxrd@z Thu Jan  1 00:00:00 1970
From: Glauber Costa <glommer-bzQdu9zFT3WakBO8gow8eQ@public.gmane.org>
Subject: Re: [PATCH v3 19/28] slab: per-memcg accounting of slab caches
Date: Tue, 29 May 2012 20:13:28 +0400
Message-ID: <4FC4F5A8.9060506@parallels.com>
References: <1337951028-3427-1-git-send-email-glommer@parallels.com> <1337951028-3427-20-git-send-email-glommer@parallels.com> <alpine.DEB.2.00.1205290951520.4666@router.home> <4FC4F42D.6060601@parallels.com>
Mime-Version: 1.0
Content-Transfer-Encoding: 7bit
Return-path: <cgroups-owner-u79uwXL29TY76Z2rM5mHXA@public.gmane.org>
In-Reply-To: <4FC4F42D.6060601-bzQdu9zFT3WakBO8gow8eQ@public.gmane.org>
Sender: cgroups-owner-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
List-ID: <cgroups.vger.kernel.org>
Content-Type: text/plain; charset="us-ascii"; format="flowed"
To: Christoph Lameter <cl-vYTEC60ixJUAvxtiuMwx3w@public.gmane.org>
Cc: linux-kernel-u79uwXL29TY76Z2rM5mHXA@public.gmane.org, cgroups-u79uwXL29TY76Z2rM5mHXA@public.gmane.org, linux-mm-Bw31MaZKKs3YtjvyW6yDsg@public.gmane.org, kamezawa.hiroyu-+CUm20s59erQFUHtdCDX3A@public.gmane.org, Tejun Heo <tj-DgEjT+Ai2ygdnm+yROfE0A@public.gmane.org>, Li Zefan <lizefan-hv44wF8Li93QT0dZR+AlfA@public.gmane.org>, Greg Thelen <gthelen-hpIqsD4AKlfQT0dZR+AlfA@public.gmane.org>, Suleiman Souhlal <suleiman-hpIqsD4AKlfQT0dZR+AlfA@public.gmane.org>, Michal Hocko <mhocko-AlSwsSmVLrQ@public.gmane.org>, Johannes Weiner <hannes-druUgvl0LCNAfugRpC6u6w@public.gmane.org>, devel-GEFAQzZX7r8dnm+yROfE0A@public.gmane.org, David Rientjes <rientjes-hpIqsD4AKlfQT0dZR+AlfA@public.gmane.org>, Pekka Enberg <penberg-bbCR+/B0CizivPeTLB3BmA@public.gmane.org>

On 05/29/2012 08:07 PM, Glauber Costa wrote:
> On 05/29/2012 06:52 PM, Christoph Lameter wrote:
>> On Fri, 25 May 2012, Glauber Costa wrote:
>>
>>> > This patch charges allocation of a slab object to a particular
>>> > memcg.
>> Ok so a requirement is to support tracking of individual slab
>> objects to cgroups? That is going to be quite expensive since it will
>> touch the hotpaths.
>>
>
> No, we track pages. But all the objects in the page belong to the same
> cgroup.
>

Also, please note the following:

The code that relays us to the right cache, is wrapped inside a static 
branch. Whoever is not using more than the root cgroup, will not suffer 
a single bit.

If you are, but your process is in the right cgroup, you will 
unfortunately pay function call penalty(*), but the code will make and 
effort to detect that as early as possible and resume.


(*) Not even then if you fall in the following categories, that are 
resolved inline:

+       if (!current->mm)
+               return cachep;
+       if (in_interrupt())
+               return cachep;
+       if (gfp & __GFP_NOFAIL)
+               return cachep;