Re: dhcpcd and static IPv6 addresses
Roy Marples
Fri Feb 05 21:52:53 2016
On Friday 05 February 2016 15:38:31 Nicholas Williams wrote:
> On Fri, Feb 5, 2016 at 3:06 PM, Roy Marples <roy@xxxxxxxxxxxx> wrote:
> > Hi List!
> >
> > So over the years, several people have asked for dhcpcd to support
> > static IPv6 addresses - I even have an open ticket for it. Now, IPv6 is
> > designed around many addresses, whereas IPv4 is designed around a
> > singular address. So some questions :)
> >
> > * Is there a need to configure >1 static IPv6 address?
>
> I can imagine a whole host of reasons for needing >1 static IPv6 addresses,
> as I can also imagine many reasons for >1 static IPv4s (none of them for
> VyOS, because we use a different mechanism for static addresses). I'm
> guessing you'd also need static v6 routers, like with IPv4?
So far no-one has asked for static routes.
>
> > * Should a static IPv6 address disable IPv6RS and DHCPv6?
>
> It seems invalid for an interface to have both static and dynamic
> addresses, and I can't imagine a reason why one would need it. Disabling
> RS and DHCPv6 would also be consistent with static IPv4 support's disabling
> DHCPv4, wouldn't it?
Yes, it would be consistent, but since all supported kernels now allow >1 IPv4
address per interface it could also be views as a bug.
> I wasn't going to bug the list with this question until you sent an email
> about it, but now that the subject has been breached, I'm quite curious:
> Why does a program designed to be a DHCP client have support for static
> addresses at all? Seems counter-intuitive.
DHCP INFORM requires a static IP address, as does the DHCPv6 equivalent (but a
lladdr will also suffice).
Another good reason could be space - no need for ifconfig/route/ip on the
machine if fully working.
It could though be argued that they are also good debugging tools for when
dhcpcd goes tits up ;)
Roy
Archive administrator: postmaster@marples.name