Re: [MaraDNS list] How to get MaraDNS and Deadwood to talk to each other?
Joshua Kinard <[email protected]> Sun, 04 Mar 2012 19:16:45 -0500
| Newsgroups | gmane.network.dns.maradns.general |
|---|---|
| Message-ID | <[email protected]> |
On 03/04/2012 00:22, Sam Trenholme wrote: >> This is a SQA regression that I run before making a Deadwood release; >> if this regression fails, I don't make a Deadwood release until the >> regression passes again. This test passes both on "bare metal" and in >> a OpenVZ virtual container. > > Note to self: This regression is actually broken. While, if properly > compiled, both MaraDNS and Deadwood can bind to and use IPv6 addresses > like fd4d:6172:6144:4e53::1 and fd4d:6172:6144:4e53::2, the SQA > regression no longer correctly tests for this. > > MaraDNS needs to be compiled with special flags to use IPv6 and this > SQA test was never correctly updated to use MaraDNS 2's new IPv6 > compile-time flags. > > I should fix the IPv6 regressions one of these years. I'm in no real > hurry to do so. Just so I understand, you're saying that Deadwood should bind to a ULA address, but just the SQA regression is broken? The only compile-time flags that I can observe being passed for IPv6 is -DIPV6. That said, though, I pinned the issue down to a little bit of PEBKAC and a possible bug in Gentoo init scripts, or just the Kernel disabling binds while duplicate address detection (DAD) is on-going. My virtual machine appears to boot up *too* fast for several services trying to bind to IPv6 (sshd and Deadwood), thus the binds are getting denied because the kernel disallows binds to "tentative" addresses. I have a workaround in place that mitigates this. So, that problem is solved. Celebrate! How about the interaction between MaraDNS and Deadwood? I did find that the one example of having Deadwood apply a different "root" for a private network to work okay, but it seems like the two daemons should try to link to each other in some format. I believe PowerDNS allows for one to talk to the other over IPv4, IPv6, or even a UNIX socket. I am not certain of the security implications of such a setup, however, and I know security is one of your primary goals with MaraDNS. But it would eliminate the need for Deadwood selectively setting the AA bit, and instead just forward the response it would get from MaraDNS. Cheers, -- Joshua Kinard Gentoo/MIPS [email protected] 4096R/D25D95E3 2011-03-28 "The past tempts us, the present confuses us, the future frightens us. And our lives slip away, moment by moment, lost in that vast, terrible in-between." --Emperor Turhan, Centauri Republic