binding tcp socket error with listen-on-v6

[Mark_Andrews at isc.org]

> When enabling listen-on-v6 {any; }; in named.conf, I get the following 
> errors in syslog:
> 
> Jan 26 18:52:58 rdb named[12395]: starting BIND 9.2.0 -u daemon -t 
> /var/named
> Jan 26 18:52:58 rdb named[12395]: using 1 CPU
> Jan 26 18:52:58 rdb named[12400]: loading configuration from 
> '/etc/named.conf'
> Jan 26 18:52:58 rdb named[12400]: listening on IPv6 interfaces, port 53
> Jan 26 18:52:58 rdb named[12400]: listening on IPv4 interface eth0, 
> xxx.xxx.xxx.xxx#53
> Jan 26 18:52:58 rdb named[12400]: binding TCP socket: address in use
> Jan 26 18:52:58 rdb named[12400]: listening on IPv4 interface eth1, 
> 192.168.1.1#53
> Jan 26 18:52:58 rdb named[12400]: binding TCP socket: address in use
> 
> Named responds on all ipv6 interfaces, so why the error messages?
> If I comment out the v6 listen line, I don't get the errors.

	You have a broken IPv6 stack.  You should be able to bind to
	::#53 and then bind to specific addresses and port 53 especially
	as you can bind to 0.0.0.0#53 and then to specific addresses and
	port 53.  IPv6 should be no different to IPv4 in this respect.
 
> And why can't specific ipv6 interfaces be listed as opposed to none or any?

	Because IPv6 interface scanning is not (well) documented.
	Just about every vendor decided to do things slightly
	differently.  Linux is worst of all in that it didn't extend
	the existing ioctl() / sysctl() interfaces but decided to
	invent yet another method.  Talk about "not invented here"
	syndrome.

	BIND 9.3 will attempt to support IPv6 interface scanning.

	Mark
--
Mark Andrews, Internet Software Consortium
1 Seymour St., Dundas Valley, NSW 2117, Australia
PHONE: +61 2 9871 4742                 INTERNET: Mark.Andrews at isc.org