Linux source address selection vs. EUI-64

[Geert Hendrickx]

On Sun, Nov 14, 2010 at 12:58:04AM +1030, Mark Smith wrote:
> On Sat, 13 Nov 2010 13:51:05 +0100
> Geert Hendrickx <ghen at telenet.be> wrote:
> > my VPS hosting provider has now deployed native IPv6 on their network
> > (hooray!), and assigned each customer a /96 for their VPS, together
> > with stateless autoconfiguration for routing and a EUI-64 address.
> 
> SLAAC requires /64 prefix lengths, as Modified EUI-64 Interface
> Identifiers (IIDs) are 64 bits in size. When you say they've given each
> customer a /96, are you saying a /96 range of addresses that you
> can statically configure within a /64 shared by many customers (e.g.
> you get 0x0 - 0xffffffff, next customer gets 0x100000000 -
> 0x1ffffffff etc.), or have they actually tried to route a /96 towards
> you?

No, the /96 assigned to me is part of the same /64 as the SLAAC subnet, so
it's all a single broadcast domain (I actually configure my static address
with /64 as the prefix, but I can only "claim" a /96 subset of them).

The EUI-64 address and my static address are thus in the same /64, and
without other distiction, Linux picks the most recently configured address,
which will (almost) always be the SLAAC address.


> If it is the latter, they should have enough /64s to easily be able to
> give every customer at least one, if not multiple e.g. /60 or /56. That
> is a fairly fundamental thing to get wrong, I'd be wondering if they
> really know all that much about IPv6 at all, and would be hesitant to
> trust them to operate it reliably.

For individual hosts (esp. in a VPS environment), assigning a /64 or
larger makes little sense to me, a /96 is more than enough.


	Geert
 

-- 
Geert Hendrickx  -=-  ghen at telenet.be  -=-  PGP: 0xC4BB9E9F
This e-mail was composed using 100% recycled spam messages!