[Dnsmasq-discuss] dhcpv6 @ rc1
simon at thekelleys.org.uk
Wed Feb 29 12:17:19 GMT 2012
On 29/02/12 08:00, Vladislav Grishenko wrote:
> Hi, Simon
> In latest 60rc1, it seems need to use enable-ra along with
> Without it default link local gateway can’t be assigned to clients,
> because it’s always done via RA.
That's correct: I suppose we could make --enable-ra the default, and
have --disable-ra, but that's a bit dangerous for people who are already
running an RA-server.
> About dhcp-range, prefix6::, ra-only syntax,
> According sources, preffered and valid lifetime can be set with ra-only
> token, but not mentioned in sample conf, a bit confusing
> Next thing is dnsv6 option, it works a bit abnormally (comparing with v4)
> Suppose we need to assign several dns servers, where one is dnsmasq itself.
> With dhcpv4 it could be done with
> But following for dhcpv6 passes [::] in reply message, instead of
> interface global/link local address
Fixed, too. I wonder if there are any other bits of functionality in the
DHCP4 code that should have been done in IPv6 as well. I can't see any
others at the moment.
> As for dnsv6 without dhcpv6, list can be passed via RA RDNSS option, see
> RFC 6106.
> Windows clients don’t support it, and, according technet, will never do.
> However, there’s 3^rd party win32 daemons, which can be used, like
The rationale is that the RA functionality is just what's needed to make
DHCPv6 work. (the RDNSS isn't complete either: if your DNS server isn't
the dnsmasq server, there's no way to change the address sent.) It's
expected that you will configure this in DHCPv6, or if you want a
complex RA setup (DNS search, mobile IPv6, different timeouts, etc) use
a fully configurable RA server like radvd.
I'm happy to consider arguments that the subset I've chosen isn't
optimal: it was mainly chosen becasue it's what could be done with no
extra config over DHCPv6, apart from "enable-ra"
Many thanks for you useful suggestions. I'll push rc2 this evening after
I have a chance to check it.
More information about the Dnsmasq-discuss