Re: draft-wbeebee-ipv6-cpe-router-03

[Brian E Carpenter <brian.e.carpenter@gmail.com>]

Pekka,

On 2008-11-19 19:35, Pekka Savola wrote:
> On Wed, 19 Nov 2008, Brian E Carpenter wrote:
>>> How about auto-enabling 6to4 assuming it can actually reach a
>>> 192.88.99l.1 gateway?
>>
>> That doesn't guarantee connectivity; you actually need to see an
>> IPv6 reply before you know that 6to4 is working. So you really
>> need to enable 6to4, run a reachability test, and disable it if
>> no response.
> 
> Well, while this is one possible approach, and it may be the simplest
> test from a human perspective, an implementation could do a test without
> enabling 6to4.
> 
> You just need to craft a packet using a raw socket that resembles that
> kind of 6to4 packet, and watch for replies.  As an example of how
> Microsoft has done it, the implementation sends a proto-41 packet with
> IPv6 TTL=1 to an IPv6 address corresponding relay's v4 address, so the
> relay will return _something_ if connectivity at least up to the relay
> is working, the relay has v6 enabled, etc.

Sure, but it is not sure that the 6to4 path is symmetric - the
reply packets might come from a completely different relay.

   Brian

> 
> Live example from a relay:
> 
> 08:31:01.202671 IP (tos 0x0, ttl 119, id 174, offset 0, flags [none],
> proto IPv6 (41), length 68) 93.106.184.x > 192.88.99.1: IP6 (hlim 1,
> next-header ICMPv6 (58) payload length: 8) 2002:5d6a:b8xx::5d6a:b8xx >
> 2002:c058:6301::c058:6301: [icmp6 sum ok] ICMP6, echo request, length 8,
> seq 2
> 08:31:01.202752 IP (tos 0x0, ttl 40, id 32297, offset 0, flags [none],
> proto IPv6 (41), length 116) 192.88.99.1 > 93.106.184.x: IP6 (hlim 64,
> next-header ICMPv6 (58) payload length: 56) 2002:c058:6301::1741 >
> 2002:5d6a:b8xx::5d6a:b8xx: [icmp6 sum ok] ICMP6, time exceeded
> in-transit, length 56 for 2002:c058:6301::c058:6301
>