Date: Fri, 01 May 2009 06:48:43 -0400
From: Jon Oberheide <jon@...rheide.org>
To: oss-security@...ts.openwall.com
Subject: Re: CVE request (sort of): Quagga BGP crasher

Florian,

On Fri, 2009-05-01 at 11:02 +0200, Florian Weimer wrote:
> There's a crasher bug in Quagga's bgpd which can allegedly be
> triggered by routes present in the global table.  See:
> 
>   <http://thread.gmane.org/gmane.network.quagga.devel/6513>
> 
> I think we need a CVE for that, but I don't understand the problem yet
> (and I can't reproduce it), so I can't come up with a concise
> vulnerability description.

Looks like the Quagga code in bgp_aspath.c is assuming that converting
each ASN of the AS path to a string will be 5 bytes plus a space
(#define ASN_STR_LEN (5 + 1)).  Therefore, it allocates (ASN_STR_LEN *
the number of ASNs in the path segment) bytes to snprintf into when
creating the pretty-print version of the AS path.

This is all fine and dandy until we hit an AS path with 32-bit ASNs
whose string representation can of course be longer than 5 bytes.  The
len += snprintf()'s will return more bytes written than expected and
then we'll hit our assert since we've written more bytes than originally
allocated (str_size).

Have you tried reproducing it with an AS path containing 32-bit ASNs (of
value at least 100000)?

Regards,
Jon Oberheide

-- 
Jon Oberheide <jon@...rheide.org>
GnuPG Key: 1024D/F47C17FE
Fingerprint: B716 DA66 8173 6EDD 28F6  F184 5842 1C89 F47C 17FE

[ CONTENT OF TYPE application/pgp-signature SKIPPED ]

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

Powered by Openwall GNU/*/Linux - Powered by OpenVZ