From: Jan Blunck Subject: Re: [PATCH 00/32] VFS based Union Mount (V3) Date: Tue, 19 May 2009 14:15:04 +0200 Message-ID: <20090519121504.GO16526@bolzano.suse.de> References: <1242662968-11684-1-git-send-email-jblunck@suse.de> <20090519103924.GL16526@bolzano.suse.de> <200905191354.08701.arnd@arndb.de> Mime-Version: 1.0 Content-Type: text/plain; charset=iso-8859-1 Content-Transfer-Encoding: QUOTED-PRINTABLE Cc: Miklos Szeredi , linux-kernel@vger.kernel.org, linux-fsdevel@vger.kernel.org, viro@zeniv.linux.org.uk, bharata@in.ibm.com, dwmw2@infradead.org, mszeredi@suse.cz, vaurora@redhat.com To: Arnd Bergmann Return-path: Content-Disposition: inline In-Reply-To: <200905191354.08701.arnd@arndb.de> Sender: linux-kernel-owner@vger.kernel.org List-Id: linux-fsdevel.vger.kernel.org On Tue, May 19, Arnd Bergmann wrote: > On Tuesday 19 May 2009, Jan Blunck wrote: > > > So this means that the topmost branch always needs to be writable= , > > > right? =A0It isn't possible to make a union of two iso9660 filesy= stems, > > > for example? > >=20 > > Exactly. Although, you can do that with the help of tmpfs on top of= the two > > iso9660 filesystems. >=20 > But how do you get there? You can mount the tmpfs on top of two iso96= 60 > file systems, but it seems that you wouldn't be able to get the two > stacked on top of each other in the first place. Well, at the moment you can stack them but readdir will fail every time= you call it ... I think this is just a question of policy if we want to all= ow that or not. > Also, by mounting a tmpfs on top, wouldn't you you violate the requir= ement > for persistent inode numbers again? There is no requirement for persistent unique inode numbers except if y= ou want to export the union again. This is something that is out of scope of th= is implementation. If you are going to export a union mounted filesystem, = you only export the topmost filesystem. > > Or by adding fake write support to iso9660 ...=20 >=20 > This would work, but you'd have to do this for each file system if yo= u want > to be able to use it as the top of the union while backed by a read-o= nly > block device or when you don't want it to be written. I know that the requirement for the topmost filesystem to be able to cr= eate directories and fill them with fallthrus is an unattractive one. On the= other hand this is the cost that you have to pay at the moment to get this ki= nd of functionality. This implementation will not help with all use-cases. It= s focus is to get certain use-cases right.