From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1752822AbXCERlF (ORCPT ); Mon, 5 Mar 2007 12:41:05 -0500 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S1752817AbXCERlE (ORCPT ); Mon, 5 Mar 2007 12:41:04 -0500 Received: from e36.co.us.ibm.com ([32.97.110.154]:55949 "EHLO e36.co.us.ibm.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1752822AbXCERlD (ORCPT ); Mon, 5 Mar 2007 12:41:03 -0500 Date: Mon, 5 Mar 2007 23:17:50 +0530 From: Srivatsa Vaddagiri To: Paul Jackson Cc: Herbert Poetzl , dev@openvz.org, ebiederm@xmission.com, winget@google.com, containers@lists.osdl.org, menage@google.com, ckrm-tech@lists.sourceforge.net, xemul@sw.ru, akpm@linux-foundation.org, linux-kernel@vger.kernel.org Subject: Re: [PATCH 0/2] resource control file system - aka containers on top of nsproxy! Message-ID: <20070305174750.GA14507@in.ibm.com> Reply-To: vatsa@in.ibm.com References: <20070301133543.GK15509@in.ibm.com> <20070301113900.a7dace47.pj@sgi.com> <45E84682.6040500@openvz.org> <20070303174553.GB16051@MAIL.13thfloor.at> <20070303132244.80b5cb40.pj@sgi.com> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <20070303132244.80b5cb40.pj@sgi.com> User-Agent: Mutt/1.5.11 Sender: linux-kernel-owner@vger.kernel.org X-Mailing-List: linux-kernel@vger.kernel.org On Sat, Mar 03, 2007 at 01:22:44PM -0800, Paul Jackson wrote: > I still can't claim to have my head around this, but what you write > here, Herbert, writes here touches on what I suspect is a key > difference between namespaces and resources that would make it > impractical to accomplish both with a shared mechanism for aggregating > tasks. The way nsproxy is structured, its all pointers to actual namespace (or in case of rcfs patch) resource objects. This lets namespaces objects be in a flat hierarchy while resource objects are in tree-like hierarchy. nsproxy itself doesnt decide any hierarchy. Its those objects pointed to by nsproxy which can form different hierarchies. In fact the rcfs patches allows such a combination afaics. > > on every limit accounting or check? I think that > > is quite a lot of overhead ... > > Do either of these dereferences require locks? A rcu_read_lock() should be required, which is not that expensive. -- Regards, vatsa