<<< Date Index >>>     <<< Thread Index >>>

Re: debian dhcpd, old format string bug



infamous41md@xxxxxxxxxx wrote:

On Thu, 28 Oct 2004 10:31:38 +1000
Tarragon Allen <tarragon@xxxxxxxxxxxx> wrote:
On Tuesday 26 October 2004 10:37, infamous41md@xxxxxxxxxx wrote:
Firstly, good etiquette would have been for you to actually report the bug with Debian. I don't see any bugs raised against any of the appropriate packages regarding this.



I've tried contacting the person in charge of the debian security audit project
numerous times to try and co-ordinate audits, and he doesn't respond.  I have
better things to do with my time.  I don't provide notice when people disregard
my emails.  If you don't like, I don't care.  My mother already taught me all
the etiquette I need, but thanks for the moral support.  Btw, is it salad fork
left, or dinner fork left?

The Debian audit people are not the same as the official Debian security team, which is the one in charge of preparing security advisories and fixing security bugs in the stable release.
Please read: http://www.debian.org/security/faq

The first group is an internal project that is reviewing parts of Debian and submitting bugs to the stable and unstable release, the second group fixes only the former, while the later is fixed by the package maintainers themselves.
http://www.debian.org/security/audit/

As to this vulnerability, CA-2002-12 is referenced as CAN-2002-0702 [1], that, based on the page that lists lists vulnerabilities that do not affect the current Debian stable release [2] does not apply to the dhcp3-server packages. And, indeed, reviewing the comon/print.c file in dhcp3-server's source code you can see:

        if (errorp)
                log_error ("%s", obuf);
        else
                log_info ("%s", obuf);

instead of the (vulnerable):

        if (errorp)
                log_error (obuf);
        else
                log_info (obuf);

Which fixes the issue (see [3]). The code is _not_ present in the dhcp packages (version 2.0pl5-11), so they aren't vulnerable to _this_ issue either.

I'm saying, grep -rn syslog * | grep -v \". Soon after I found that, I googled
and found the CERT detailing a format string in logging code.  I assumed it was
the exact same thing I just found.  I spoke with some debian person about this
yesterday, or day before, and they can release an advisory to clear it up.

That grep line brings a lot of code, some of it might be vulnerable to format string attacks, but it's not related to the CERT advisory at all. For those not having the code at hand:

$ grep -B 2 -A 2  -rn syslog * | grep -v \"
(...)
common/errwarn.c-73-#ifndef DEBUG
common/errwarn.c:74:  syslog (log_priority | LOG_ERR, mbuf);
common/errwarn.c-75-#endif

Which could be easily fixed to prevent a format string attack (but is not and is indeed vulnerable). Maybe this bug is related to CAN-2001-0181 (BID-2215). I don't have access to Caldera's code so I can review that...

Regards

Javier



[1] http://cve.mitre.org/cgi-bin/cvename.cgi?name=CAN-2002-0702
[2] http://www.debian.org/security/nonvulns-wood
[3] http://archives.neohapsis.com/archives/vulnwatch/2002-q2/0063.html