From mboxrd@z Thu Jan  1 00:00:00 1970
From: Pavel Emelyanov <xemul@parallels.com>
Subject: Re: [PATCH v3 3/4] limit nr_dentries per superblock
Date: Mon, 15 Aug 2011 15:32:17 +0400
Message-ID: <4E4903C1.9050207@parallels.com>
References: <1313334832-1150-1-git-send-email-glommer@parallels.com>	<1313334832-1150-4-git-send-email-glommer@parallels.com>	<CAOJsxLFYD9DkEU5R+9fZr-uaznteP8-BUC_Ti+FnNFqtbOCHSw@mail.gmail.com>	<20110815104656.GG26978@dastard>	<CAOJsxLGFJmqO-W=itQbO4Mh4DxSD4wrHOC8gQ5bWL5aE1YYeQw@mail.gmail.com>	<4E48FD8A.90401@parallels.com> <CAOJsxLFjZffC9i5kQsBWjGy3eWdj3VMB5awz=yPcwgeKb5BG0Q@mail.gmail.com>
Mime-Version: 1.0
Content-Type: text/plain; charset=ISO-8859-1
Content-Transfer-Encoding: 7bit
Cc: Dave Chinner <david@fromorbit.com>,
	Glauber Costa <glommer@parallels.com>,
	"linux-kernel@vger.kernel.org" <linux-kernel@vger.kernel.org>,
	"linux-fsdevel@vger.kernel.org" <linux-fsdevel@vger.kernel.org>,
	"containers@lists.linux-foundation.org"
	<containers@lists.linux-foundation.org>,
	Al Viro <viro@zeniv.linux.org.uk>,
	Hugh Dickins <hughd@google.com>,
	Nick Piggin <npiggin@kernel.dk>,
	Andrea Arcangeli <aarcange@redhat.com>,
	Rik van Riel <riel@redhat.com>,
	Dave Hansen <dave@linux.vnet.ibm.com>,
	James Bottomley <jbottomley@parallels.com>,
	Eric Dumazet <eric.dumazet@gmail.com>,
	Christoph Lameter <cl@linux.com>,
	David Rientjes <rientjes@google.com>
To: Pekka Enberg <penberg@kernel.org>
Return-path: <linux-kernel-owner@vger.kernel.org>
In-Reply-To: <CAOJsxLFjZffC9i5kQsBWjGy3eWdj3VMB5awz=yPcwgeKb5BG0Q@mail.gmail.com>
Sender: linux-kernel-owner@vger.kernel.org
List-Id: linux-fsdevel.vger.kernel.org

On 08/15/2011 03:14 PM, Pekka Enberg wrote:
> Hi Pavel,
> 
> On Mon, Aug 15, 2011 at 2:05 PM, Pavel Emelyanov <xemul@parallels.com> wrote:
>> This will make sense, since the kernel memory management per-cgroup is one of the
>> things we'd live to have, but this particular idea will definitely not work in case
>> we keep the containers' files on one partition keeping each container in its own
>> chroot environment.
> 
> And you want a per-container dcache limit? 

To be more specific - we want to protect the node with >1 containers from one of
them growing the dcache infinitely. One of the solutions to this - per container
dcache limit.

> Will the containers share the same superblock? 

Yes, this is typical scenario for both OpenVZ and LXC now.

> Couldn't you simply do per-container "struct kmem_accounted_cache" in struct superblock?

If by this you mean "account for all the kmem associated with particular superblock" then
this is OK for us, but this can't be done in a simple

	if (used + size > limit)
		return -ENOMEM
	else {
		used += size;
		return 0;
	}

manner, since once we hit the limit we should shrink the unused dentries. And most of the
patches are about this.

>                         Pekka
> .
>