Re: [PATCH] User chroot

[H. Peter Anvin]

Followup to:  <[e-mail]>
By author:    Paul Menage <[e-mail]>
In newsgroup: linux.dev.kernel
> 
> It could potentially be useful for a network daemon (e.g. a simplified
> anonymous FTP server) that wanted to be absolutely sure that neither it
> nor any of its libraries were being tricked into following a bogus
> symlink, or a "/../" in a passed filename. After initialisation, the
> daemon could chroot() into its data directory, and safely only serve
> the set of files within that directory hierarchy.
> 
> This could be regarded as the wrong way to solve such a problem, but
> this kind of bug seems to be occurring often enough on BugTraq that it
> might be useful if you don't have the resources to do a full security
> audit on your program (or if the source to some of your libraries
> isn't available).
> 

If the source to some of your libraries aren't available, you have no
clue when/why/if they will try to access other files in the
filesystem.  Since libc WILL do this, a random chroot() breaks libc
(unless you can set up a proper root environment), and therefore
pretty much anything else is pointless.

	-hpa

-- 
<[e-mail]> at work, <[e-mail]> in private!
"Unix gives you enough rope to shoot yourself in the foot."
http://www.zytor.com/~hpa/puzzle.txt