Re: /tmp

[Michal Zalewski <lcamtuf () DIONE IDS PL>]

On Thu, 21 Dec 2000, Dan Stromberg wrote:

> 2) Provide a library function, perhaps mkytemp(), which calls a setuid
> root program that creates /tmp/secy/username/templateXXXXXX. /tmp/secy
> is only writable by root, and mkdir'd at system boot time. Pass back
> the filename and a file handle.  The setuid program should eliminate
> the DOS, but is slower (exec's hurt).  You could just use /sec or
> /stmp, but then temp scrubbers have to be modified slightly, slowing
> migration.

Please tell me why are you considering /tmp as the only one solution?
Moving runtime temporary files that do not *have* to be shared from /tmp
to eg. ~/tmp is pretty good solution, as well. Unfortunately, this won't
solve numerous problems of programs that are not following mk*temp()
convention, creating eg. pid-based temporary files ;) On the other hand,
most of context pseudo-filesystem / redirection solutions (like making
real location of /tmp entries for every UID different) might broke eg. X
server / clients functionality etc.

The truth is obvious: /tmp isn't anything good, and there are no easy
workarounds for programs using it. We should modify these programs,
instead, moving ALL unnecessary /tmp files (which are not used for
interprocess communication, like Xserver sockets) to ~/tmp or so. You
cannot feel safe without it, no matter how complicated system will be used
to guard /tmp.

It is not really difficult to grep all system sources against /tmp,
mkstemp, mktemp, tmpnam, tempnam, tmpfile and similar stuff...

--
_______________________________________________________
Michal Zalewski [lcamtuf () tpi pl] [tp.internet/security]
[http://lcamtuf.na.export.pl] <=--=> bash$ :(){ :|:&};:
=--=> Did you know that clones never use mirrors? <=--=