Re: Default value of net.ipv6.bindv6only should revert to 0
Hendrik Sattler <firstname.lastname@example.org> writes:
> I guess not all of them. OTOH, not all of them might use AF_INET6
> sockets for IPv4 connections (the two-sockets approach is always
> possible when using IPV6_V6ONLY socket option).
> At least xinetd uses IPV6_ADDRFORM although that socket option is
> already deprecated again but still useful for such cases.
> openbsd-inetd explicitely doesn't support IPv4-mapped addresses,
> configuration needs seperate IPv4 and IPv6 setup.
> I didn't look at the other two.
Thank you very much for the investigation! That's really useful
information (at least to me) and I appreciate you taking the effort.
I took a quick look at the source of rlinetd, and so far as I can tell it
will do whatever the default IPv6 stack behavior is. It iterates through
every return from getaddrinfo for AI_PASSIVE for a particular family, and
doesn't appear to do anything special with the resulting socket before
passing it along to the underlying server. But I'm not sure that I'm
following the code.
Russ Allbery (email@example.com) <http://www.eyrie.org/~eagle/>