DHCID calculation bug

René GARCIA rene.garcia at sogeti.com
Fri Jul 17 07:42:11 UTC 2009


Le jeudi 16 juillet 2009 19:17 de la part de David W. Hankins
> On Thu, Jul 16, 2009 at 03:32:10PM +0200, René GARCIA wrote:
> > I'm working on IPv4 to IPv6 transition. As part of my job I've been 
> > testing ISC DHCP 4.1.1b (and cvs repository) and I'm unhappy with 
> > DHCID handling when sharing IPv4 DHCP+DDNS and IPv6 DHCP+DDNS in the 
> > same domainname. So looking to the source code I've found two bugs. 
> > This post is about the fist bug, I'll describe the second one in another
post.
> 
> I also noticed the bugs in the dhcid generator function, I'm not really
worried 
> about the type range check because it is an internal consistency check,
and the
> error in calculating the first bytes of the text string means that current

> DHCID/TXT RR's in zones actively using this feature would suddenly become 
> incompatible on a maintenance upgrade.
>
> So resolving these DHCID issues has been elided until we migrate from our
TXT
> record to RFC standard DHCID RR's (so we have one migration to manage).

Yes I agree, changing the way DHCID is calculated would make all existing
TXT records to be obsolete. This may generate a lot of migration work on
huge ddns systems.

Is there any reflexion to make DHCID values match for the same client using
DHCPv6 for IPv6 and DHCP for IPv4 ? I know that this has been discussed many
years ago (on 2003 I think) but I haven't seen any conclusion. Another idea
was to create a DHCIDv4 and a DHCIDv6 to make both protocols independant but
I can't find any RFC fixing this.

Regards,
René GARCIA




More information about the dhcp-workers mailing list