getaddrinfo bug in glibc (CVE-2015-7547) questions?
 

Hello

I just about the bug. As far as I understand it's this:

If you do reverse DNS lookups, and the result comes from an evil DNS server, bad things can happen. (if length > 2048) Now people are patching and rebooting.

My question is if it's only the the DNS servers need fixing. If I boot with a old live cd and reverse lookup some nasty IP address, what will happen?

It's not specifically DNS that needs patching (we're talking about an adversary exploiting what is "allowed" reply size-wise within the DNS RFCs) but any machine using an affected Glibc version that uses AF_UNSPEC-type lookups and uses unfiltered replies (doesn't say anything about the type) or receives unfiltered lookup results from other systems.

Ok. We have a DNS server using BIND, and as far as I understand, it's doing the actual resolving of reverse DNS lookups. Or am I incorrect?

The DNS server has been patched, but I was thinking it could maybe do the filtering of replies? A live CD gets the DNS from DHCP, so I was wondering if it was a way to protect them against this.

A Domain Name server can be configured in many way: authoritative name server, slave, caching name server, forwarder. Then there's devices running Linux who have a resolver stub as libraries like Glibc provide them (uClibc, dietlibc and musl seem unaffected AFAIK). These are separate things ..


.. so while you may patch DNS as part of your mitigation strategy this only combats symptoms but does not solve the cause IMHO.