Follow @Openwall on Twitter for new release announcements and other news
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Date: Wed, 8 Oct 2014 14:51:50 +0100
From: Stephane Chazelas <>
Subject: Re: Thoughts on Shellshock and beyond

2014-10-07 21:06:14 -0700, Michal Zalewski:
> 4) Following the original find but before the end of the embargo,
> there was no immediate realization that the underlying parser is
> complex and likely not designed with security in mind, and therefore,
> that it will very likely follow a well-established pattern and come
> apart under closer scrutiny. I'm not sure if this is an argument for
> not having embargoes (perhaps) or for sanctioning more thoughtful
> reviews of the proposed fixes. Or perhaps it's just a fluke.

That's not totally true. Florian's (Debian/Red Hat security)
initial reply the day after the report already brought up
possible hardening solution (delay parsing until the function is
called (like in rc)), prefix, require newline in the value, list
the functions to import in a fixed env var...) to avoid exposing
the parser. The fact that it was wrong not to use the BASH_
prefix/namespace was also brought up.

There were some voices against those, like it was not
necessarily needed, that it could break compatibility,
functionality, that doing it properly was possibly going to take
a long time, that such a radical change was not for a patch
release, some misunderstanding as well, and after a few patch
iterations and some review of the parser code (not by me, not
that I would have been able to assess if it was safe enough or
not), no compelling reason to say that the stopping from
interpreting past the function definition (and before which the
fix addresses) wasn't enough.

Everyone was also in agreement that we shouldn't sit on it too
long for fear of a leak (one of the mistakes on my part was that
I put PoC code in the subject of the email for instance, and it
was sent to quite a large party).

It was quite obvious that the interpreting past the function
definition was a bug and needed fixed asap given what it
allowed and more importantly there was no doubt on how it should
be fixed (though the simple fix did break backward
compatibility) not so much for the exposing of the parser and
how to address it for which it would have taken longer to reach
an agreement.

So more than a misunderstanding, I'd say that was a rushed bad
decision (to address the obvious immediate issue now, and harden
(the "right" way) later, assuming (wrongly) that the parser
would hold).

Note that the hardening fix that was provided in a rush
post-disclosure is not the best one. It addresses the security
concern, but it cripples the feature and clutters the

It breaks backward compatibility because it restricts which
functions can be exported. For instance, you can't export a
/bin/rm function anymore (was useful as a debugging tool).

Some tools (like pdksh/mksh) don't propagate env vars with `()`
or `%%` in their name, so you can no longer export functions
accross them (POSIX also leaves the behaviour there undefined
which means there's a portability dimension, bash is ported to
almost everything out there).

The solution I was in favour of was a:

BASH_FUNCDEFS='f() { f-function; }
g() { g-function; }...'

That also stops exposing the parser, that's *one* env var (no
cluttering, easy to understand the implications or blacklist it
as it's the same as BASH_ENV or BASHOPTS, PS4, PATH... that are
equally dangerous).


BASH_FUNCDEFS='() {(format=4.4)}
f() { f-function; }
g() { g-function; }...'

That would even be one-way compatible with an older bash
(actually exploiting the bug), and updating sudo's blacklist
would not have been necessary (because the "()" prefix is kept)..

Another thing that was mentionned is that bash is not consistent
in what name it allows for function names (and that's also
locale-dependant), so ideally, that should be fixed as well.

Now all of that would have been a bigger change that would have
taken more than the 2 week embargo to write, test and release
and is not the kind of thing you ship with a patch release.


Powered by blists - more mailing lists

Please check out the Open Source Software Security Wiki, which is counterpart to this mailing list.

Confused about mailing lists and their use? Read about mailing lists on Wikipedia and check out these guidelines on proper formatting of your messages.