Zeroing freed blocks

[Keld Jørn Simonsen]

On Sun, Apr 02, 2006 at 06:07:21PM +0100, Ron Yorston wrote:
> A couple of years ago there was a discussion on lkml under the thread
> 'PATCH - ext2fs privacy (i.e. secure deletion) patch' about zapping
> deleted data in the filesystem as a security mechanism.  The discussion
> wandered off into how 'chattr +s' could be implemented and whether
> encrypting filesystems wouldn't be a better solution to the problem.
> 
> I've been maintaining a simplified version of the patch for a different
> reason:  to keep filesystems in files sparse.  Filesystem images for use
> by things like user-mode Linux and Xen are often created as sparse files.
> After they've been in use for a while their sparseness is reduced even
> though they may have lots of free space.  Having the guest kernel fill
> deleted blocks with zeros doesn't make the underlying file sparse,
> but it does help.  I've got a page with more details:
> 
>    http://intgat.tigress.co.uk/rmy/uml/sparsify.html
> 
> Anyway, a couple of things:
> 
> 1. The patch (see below) is pretty simple.  I've been using it for some
>    time in UML build systems for old versions of software (rh62, anyone?),
>    and today I even tried it for several seconds in a Xen domU kernel.
>    It seems to do what I want, but is it any good?
> 
> 2. The patch is now for ext2 only, the original ext3 version having 
>    succumbed to bitrot.  What would it take to implement something
>    similar for ext3 these days?

Well, I think this should be optional, if included. It does directly
counteract the patch I recently sent to salvage files from their data
blocks in ext2/ext3. 

Best regards
keld