From mboxrd@z Thu Jan 1 00:00:00 1970 From: "Serge E. Hallyn" Subject: Re: design of user namespaces Date: Mon, 7 Jul 2008 10:24:06 -0500 Message-ID: <20080707152405.GA11250@us.ibm.com> References: <20080620140510.GA21720@us.ibm.com> <20080620205508.GA8429@us.ibm.com> <20080630211325.GA27738@us.ibm.com> Mime-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit Return-path: Content-Disposition: inline In-Reply-To: List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Sender: containers-bounces-cunTk1MwBs9QetFLy7KEm3xJsTq8ys+cHZ5vskTnxNA@public.gmane.org Errors-To: containers-bounces-cunTk1MwBs9QetFLy7KEm3xJsTq8ys+cHZ5vskTnxNA@public.gmane.org To: "Eric W. Biederman" Cc: Linux Containers List-Id: containers.vger.kernel.org Quoting Eric W. Biederman (ebiederm-aS9lmoZGLiVWk0Htik3J/w@public.gmane.org): > "Serge E. Hallyn" writes: > > > Quoting Eric W. Biederman (ebiederm-aS9lmoZGLiVWk0Htik3J/w@public.gmane.org): > >> > >> The very important points are that it is a remount of an existing mount > >> so that we don't have to worry about corrupted filesystem attacks, and > >> that authentication is performed at mount time. > > > > Conceptually that (making corrupted fs attacks a non-issue) is > > wonderful. Practically, I may be missing something: When you say > > remount, it seems you must either mean a bind mount or a remount. If > > remount, then that will want to change superblock flags. If the > > child userns(+child mntns) does a real remount, then that will change > > the flags for the parent ns as well, right? > > > > If instead we do a bind mount we don't have that problem, but then the > > fs can't be the one doing the user namespace work. > > > > I'm probably missing something. > > Essentially I am creating a new mount operation that is a > cousin of a remount. > > Unlike a real remount you can't change the super flags. > Unlike a bind mount you get the fs involved, and you pass in a string of flags > that the fs can interpret in a standard way. > > I expect the flags you pass in would be a subset of what is allowed > in a normal remount. > > Which is why I was calling it nativemount. Although usernsmount > may be better. > > Eric Ah, ok. Now you haven't started any sort of coding for this yet, right? I'm hoping to get some time later this week to think about/play with this. -serge