From: Andreas Gruenbacher Subject: Re: [PATCH 1/5] fsnotify/vfsmount: add fsnotify fields to struct vfsmount Date: Thu, 10 Dec 2009 17:31:42 +0100 Message-ID: <200912101731.43050.agruen@suse.de> References: <20091203055315.21918.83562.stgit@paris.rdu.redhat.com> <20091204153925.GA17062@infradead.org> Mime-Version: 1.0 Content-Type: Text/Plain; charset="iso-8859-1" Content-Transfer-Encoding: 7bit Cc: Eric Paris , linux-kernel@vger.kernel.org, linux-fsdevel@vger.kernel.org, viro@zeniv.linux.org.uk To: Christoph Hellwig Return-path: In-Reply-To: <20091204153925.GA17062@infradead.org> Sender: linux-kernel-owner@vger.kernel.org List-Id: linux-fsdevel.vger.kernel.org On Friday 04 December 2009 16:39:25 Christoph Hellwig wrote: > What's the rationale for adding them? The idea is to provide a mechanism to watch for "all events", but in a namespace aware way: for that, a process registers interest in all mount points it can reach. With the previous hack, a global "I want it all" flag, per-namespace listeners were not possible. Root can use bind mounts onto themselves (e.g., mount --bind /foo/bar /foo/bar) to watch arbitrary directories with the per-mount-point mechanism. Mount as user will open up additional use cases. It would be nice to be able to register interest in all events below a directory which is not a mount point, too. The problem with that is that the number of cached inodes below a given directory could be huge -- and we would have to visit them all atomically in order to "mark" them, which is not feasible. The per-mount-point mechanism will eventually give us *almost* the same, and so I believe it's good enough. Thanks, Andreas