From mboxrd@z Thu Jan 1 00:00:00 1970 From: Dan Smith Subject: Re: [PATCH 2/4] [RFC] Add sock_create_kern_net() Date: Wed, 28 Apr 2010 06:38:33 -0700 Message-ID: <87fx2f1yue.fsf@caffeine.danplanet.com> References: <1272034539-19899-1-git-send-email-danms@us.ibm.com> <1272034539-19899-3-git-send-email-danms@us.ibm.com> <1272455094.14068.15.camel@bigi> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Return-path: In-Reply-To: <1272455094.14068.15.camel@bigi> (jamal's message of "Wed\, 28 Apr 2010 07\:44\:54 -0400") Sender: netdev-owner@vger.kernel.org To: hadi@cyberus.ca Cc: containers@lists.osdl.org, netdev@vger.kernel.org, Daniel Lezcano , "Eric W. Biederman" List-Id: containers.vger.kernel.org j> So ... how does user space know what "other_netns" is? That's the point, userspace doesn't know about and can't use this interface. This is a way for the kernel to open a socket in another netns to talk to that netns' RTNETLINK. I realize in its current form it could be used for something more nefarious, but it would be kernel code doing it. j> Also note Eric's recent patches introduced another way of opening a j> socket in a different namespace - are you using those in the j> abstraction to find what netns is? No. The process doing the checkpoint already has pointers to the tasks and thus their netns pointers. Eric's interface is an abstraction to allow userspace to do something similar, I think that using it from the kernel would be rather messy. -- Dan Smith IBM Linux Technology Center email: danms@us.ibm.com