Received: from mx.ams1.isc.org (mx.ams1.isc.org [199.6.1.65]) by bugs.isc.org (Postfix) with ESMTP id 5EE0220EE26D for ; Tue, 15 Feb 2011 23:43:09 +0000 (UTC) Received: from order.stressinduktion.org (order.stressinduktion.org [87.106.68.36]) (using TLSv1 with cipher DHE-RSA-AES256-SHA (256/256 bits)) (Client CN "order.stressinduktion.org", Issuer "order.stressinduktion.org" (not verified)) by mx.ams1.isc.org (Postfix) with ESMTPS id 71A545F984C for ; Tue, 15 Feb 2011 23:42:54 +0000 (UTC) (envelope-from hannes@stressinduktion.org) Received: by order.stressinduktion.org (Postfix, from userid 500) id 9D65F1A0C3AB; Wed, 16 Feb 2011 00:42:53 +0100 (CET) Delivered-To: dhcp-suggest@bugs.isc.org MIME-Version: 1.0 Subject: rfc compliant dhcpv6 prefix length X-Spam-Status: No, score=-1.9 required=5.0 tests=BAYES_00,T_RP_MATCHES_RCVD autolearn=ham version=3.3.1 Return-Path: X-Original-To: dhcp-suggest@bugs.isc.org Content-Disposition: inline Date: Wed, 16 Feb 2011 00:42:53 +0100 X-Spam-Checker-Version: SpamAssassin 3.3.1 (2010-03-16) on mx.ams1.isc.org Content-Type: text/plain; charset=utf-8 Message-ID: <20110215234253.GV14589@order.stressinduktion.org> To: dhcp-suggest@isc.org From: Hannes Frederic Sowa X-RT-Original-Encoding: utf-8 Content-Length: 1343 Hello! Some time ago (last year) we had a thread on linux-net about handling prefix lengths in dhcpv6. David Hankins idea [1] was to use a custom prefix-length option in the VSIO space. I like this idea a lot, especially because you can deploy networks without being dependent on router advertisments. In the case where such an option is missing from a dhcp-offer one could fall back to the following solution (I also posted it as a reply [2]): Scan the routingtable via libnl or rtnetlink and apply the prefix-length of the route with the longest-matching prefix marked with RTF_ADDRCONF | RTF_PREFIX_RT (this is of couse a linux-only solution, solaris and freebsd should have proper apis to query prefix-length information from router advertisments). I haven't tried it if the flags are correclty exported to userspace. But if you are interested in such a fall-back solution I would be happy to test this case at the weekend in my testbed. Btw: I have seen setups where they currently use dibbler-dhcp to advertise /96-prefixes, as the "upstream" has only received a /64 itself. So non-/64 length prefixes will be an issue in near future, I think (I know it is so wrong). Greetings from Germany, Hannes [1] http://article.gmane.org/gmane.linux.network.general/14099 [2] http://article.gmane.org/gmane.linux.network.general/14572