From mboxrd@z Thu Jan 1 00:00:00 1970 From: ebiederm@xmission.com (Eric W. Biederman) Subject: [PATCH 0/4] Sysctl namespace support Date: Thu, 29 Nov 2007 10:40:24 -0700 Message-ID: References: <4742C73C.3010904@openvz.org> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Cc: Serge Hallyn , Daniel Lezcano , Cedric Le Goater , Linux Containers , Pavel Emelyanov , , , David Miller To: Herbert Xu , Andrew Morton Return-path: Received: from ebiederm.dsl.xmission.com ([166.70.28.69]:56780 "EHLO ebiederm.dsl.xmission.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1756542AbXK2RpX (ORCPT ); Thu, 29 Nov 2007 12:45:23 -0500 In-Reply-To: <4742C73C.3010904@openvz.org> (Pavel Emelyanov's message of "Tue, 20 Nov 2007 14:38:36 +0300") Sender: netdev-owner@vger.kernel.org List-Id: netdev.vger.kernel.org Currently the network namespace work has gotten about as far as we can without the ability to make sysctls that are per network namespace. The techniques we have been using for other namespace of examining current and replacing the ctl_table.data field depending on the namespace instance that current->nsproxy refers to are both ugly and do not work for the network sysctls. The case in handling the networking sysctls that does not work with the existing ugly pointer munging techniques are directories like /proc/sys/net/ipv4/conf/ and /proc/sys/net/ipv4/neigh/ whose contents vary depending on the networking devices present in the network namespace. Adding support to the sysctl infrastructure to allow to register a sysctl table for a particular instance of a particular namespace removes the need for magic sysctl methods, and allows the use of the techniques for managing dynamic sysctl tables used for years in the network stack. Herbert we need this infrastructure most in net-2.6.25 (as not having it is a current bottleneck to further development of the network namespace) so these patches are against net-2.6.25. Andrew also need this infrastructure in -mm so that we can take advantage of this new infrastructure when implementing other namespaces. So I expect the sane way to deal with this patchset is to merge into both net-2.6.25 and -mm and then Andrew can drop or disable the patches once he pulls bases -mm on a version of net-2.6.25 with the changes. Eric