IPv6 Routing Problem

I have built my own dualstack network. There are 1 gateway and several
clients.
All IPv4 are worked fine. Gateway can use IPv6 to get on the Internet
including ping, traceroute, browsing.

All clients can ping6 to gateway and gateway also can forward the
ICMPv6 packet to other gateway(Service Provider). But it only send the
echo request out, my gateway didn't receive any echo reply.

Then, I used the tcpdump to capture packets.
Service Provider Gateway = 2001:ce0:3:63::1/64
My Gateway = 2001:ce0:3:63::f000/64 & 2001:ce0:3:63::1:1100/116
Client = 2001:ce0:3:63::1:1200/116


00:05:07.542900 IP6 2001:ce0:3:63::1:1200 > 2001:ce0:3:63::1: ICMP6,
echo request, seq 20, length 16
00:05:07.543300 IP6 2001:ce0:3:63::1 > ff02::1:ff01:1200: ICMP6,
neighbor solicitation, who has 2001:ce0:3:63::1:1200, length 32
00:05:08.543758 IP6 2001:ce0:3:63::1:1200 > 2001:ce0:3:63::1: ICMP6,
echo request, seq 21, length 16
00:05:08.544309 IP6 2001:ce0:3:63::1 > ff02::1:ff01:1200: ICMP6,
neighbor solicitation, who has 2001:ce0:3:63::1:1200, length 32

It always ask the "who has 2001:ce0:3:63::1:1200, length 32".
How to solve the problem?
Thank you.

Hello,

mclainet wrote :
> I have built my own dualstack network. There are 1 gateway and several
> clients.
> All IPv4 are worked fine. Gateway can use IPv6 to get on the Internet
> including ping, traceroute, browsing.
>
> All clients can ping6 to gateway and gateway also can forward the
> ICMPv6 packet to other gateway(Service Provider). But it only send the
> echo request out, my gateway didn't receive any echo reply.
>
> Then, I used the tcpdump to capture packets.
> Service Provider Gateway = 2001:ce0:3:63::1/64
> My Gateway = 2001:ce0:3:63::f000/64 & 2001:ce0:3:63::1:1100/116
> Client = 2001:ce0:3:63::1:1200/116
>
> 00:05:07.542900 IP6 2001:ce0:3:63::1:1200 > 2001:ce0:3:63::1: ICMP6,
> echo request, seq 20, length 16
> 00:05:07.543300 IP6 2001:ce0:3:63::1 > ff02::1:ff01:1200: ICMP6,
> neighbor solicitation, who has 2001:ce0:3:63::1:1200, length 32
> 00:05:08.543758 IP6 2001:ce0:3:63::1:1200 > 2001:ce0:3:63::1: ICMP6,
> echo request, seq 21, length 16
> 00:05:08.544309 IP6 2001:ce0:3:63::1 > ff02::1:ff01:1200: ICMP6,
> neighbor solicitation, who has 2001:ce0:3:63::1:1200, length 32
>
> It always ask the "who has 2001:ce0:3:63::1:1200, length 32".

The ISP's router looks for 2001:ce0:3:63::1:1200 directly on its link
layer, which is logical since 2001:ce0:3:63::1:1200 belongs to
2001:ce0:3:63::/64. Are you sure that 2001:ce0:3:63::1:1000/116 is
allocated to you, and that routing is the proper way to use this block
instead of bridging ?

Thanks for your reply.
They assign 2001:ce0:3:63::1:1/64 ~ 2001:ce0:3:63::1:ffff/64 to me.

1. Is it mean that 2001:ce0:3:63::1:1000/116 is allocated to me?
2. I can't get your idea "routing is the proper way to use this block
instead of bridging"

Thank you very much.
Regards,

Pascal Hambourg ¼g¹D¡G

> Hello,
>
> mclainet wrote :
> > I have built my own dualstack network. There are 1 gateway and several
> > clients.
> > All IPv4 are worked fine. Gateway can use IPv6 to get on the Internet
> > including ping, traceroute, browsing.
> >
> > All clients can ping6 to gateway and gateway also can forward the
> > ICMPv6 packet to other gateway(Service Provider). But it only send the
> > echo request out, my gateway didn't receive any echo reply.
> >
> > Then, I used the tcpdump to capture packets.
> > Service Provider Gateway = 2001:ce0:3:63::1/64
> > My Gateway = 2001:ce0:3:63::f000/64 & 2001:ce0:3:63::1:1100/116
> > Client = 2001:ce0:3:63::1:1200/116
> >
> > 00:05:07.542900 IP6 2001:ce0:3:63::1:1200 > 2001:ce0:3:63::1: ICMP6,
> > echo request, seq 20, length 16
> > 00:05:07.543300 IP6 2001:ce0:3:63::1 > ff02::1:ff01:1200: ICMP6,
> > neighbor solicitation, who has 2001:ce0:3:63::1:1200, length 32
> > 00:05:08.543758 IP6 2001:ce0:3:63::1:1200 > 2001:ce0:3:63::1: ICMP6,
> > echo request, seq 21, length 16
> > 00:05:08.544309 IP6 2001:ce0:3:63::1 > ff02::1:ff01:1200: ICMP6,
> > neighbor solicitation, who has 2001:ce0:3:63::1:1200, length 32
> >
> > It always ask the "who has 2001:ce0:3:63::1:1200, length 32".
>
> The ISP's router looks for 2001:ce0:3:63::1:1200 directly on its link
> layer, which is logical since 2001:ce0:3:63::1:1200 belongs to
> 2001:ce0:3:63::/64. Are you sure that 2001:ce0:3:63::1:1000/116 is
> allocated to you, and that routing is the proper way to use this block
> instead of bridging ?

mclainet wrote:

> They assign 2001:ce0:3:63::1:1/64 ~ 2001:ce0:3:63::1:ffff/64 to me.

This is nonsense. 2001:ce0:3:63::1:1/64 runs from:
2001:0ce0:0003:0063:0000:0000:0000:0000 -
2001:0ce0:0003:0063:ffff:ffff:ffff:ffff

> 1. Is it mean that 2001:ce0:3:63::1:1000/116 is allocated to me?

Yes.

Regards
F.

Thank you for your reply.
2001:ce0:3:63::1:1/64 ~ 2001:ce0:3:63::1:ffff/64
I understand what you are talking, but they restricted this range to
me(Actually, they are my department in my U. Not ISP.). So I am
confusing and don't know why all my client can't go outside world
through my internal gateway.

Does any ports IPv6 needed for routing??
thanks

Regards,

Friedemann Stoyan ¼g¹D¡G

> mclainet wrote:
>
> > They assign 2001:ce0:3:63::1:1/64 ~ 2001:ce0:3:63::1:ffff/64 to me.
>
> This is nonsense. 2001:ce0:3:63::1:1/64 runs from:
> 2001:0ce0:0003:0063:0000:0000:0000:0000 -
> 2001:0ce0:0003:0063:ffff:ffff:ffff:ffff
>
> > 1. Is it mean that 2001:ce0:3:63::1:1000/116 is allocated to me?
>
> Yes.
>
> Regards
> F.

[Please don't top post.]

mclainet wrote :
>>>
>>>The ISP's router looks for 2001:ce0:3:63::1:1200 directly on its link
>>>layer, which is logical since 2001:ce0:3:63::1:1200 belongs to
>>>2001:ce0:3:63::/64. Are you sure that 2001:ce0:3:63::1:1000/116 is
>>>allocated to you, and that routing is the proper way to use this block
>>>instead of bridging ?
>>
> They assign 2001:ce0:3:63::1:1/64 ~ 2001:ce0:3:63::1:ffff/64 to me.
>
> 1. Is it mean that 2001:ce0:3:63::1:1000/116 is allocated to me?

I think so, but it does not mean that this address range is routed to
your router.

> 2. I can't get your idea "routing is the proper way to use this block
> instead of bridging"

I mean that /maybe/ the /112 address range which is allocated to you is
not a subnet but only a part of a larger /64 network on the same link
layer (ethernet network). It is not a separate routed subnet, so other
hosts in the /64 network expect to find your hosts in the same network
and link layer.

So in this situation your gateway should act :
- as a layer 3 router with IPv4 traffic ;
- as a layer 2 transparent bridge with IPv6 traffic.

[Supersedes due to mistakes]

mclainet wrote :
>>>
>>>The ISP's router looks for 2001:ce0:3:63::1:1200 directly on its link
>>>layer, which is logical since 2001:ce0:3:63::1:1200 belongs to
>>>2001:ce0:3:63::/64. Are you sure that 2001:ce0:3:63::1:1000/116 is
>>>allocated to you, and that routing is the proper way to use this block
>>>instead of bridging ?
>>
> They assign 2001:ce0:3:63::1:1/64 ~ 2001:ce0:3:63::1:ffff/64 to me.
>
> 1. Is it mean that 2001:ce0:3:63::1:1000/116 is allocated to me?

Not exactly.
2001:ce0:3:63::1:1000/116 = 2001:ce0:3:63::1:1000-2001:ce0:3:63::1:1fff

2001:ce0:3:63::1:1 - 2001:ce0:3:63::1:ffff = 2001:ce0:3:63::1::/112
minus 2001:ce0:3:63::1::

However, in any case, it does not mean that the address range is routed
via your gateway.

> 2. I can't get your idea "routing is the proper way to use this block
> instead of bridging"

I mean that /maybe/ the /112 or /116 block which is allocated to you is
not a subnet but only a part of a larger /64 network on the same link
layer (ethernet network). It is not a separate routed subnet, so other
hosts in the /64 network expect to find your hosts in the same network
and link layer.

So in this situation your gateway should act :
- as a layer 3 router with IPv4 traffic ;
- as a layer 2 transparent bridge with IPv6 traffic.

P.S.: Please don't top post.

Thanks for your help.
I think it is only a part of a larger /64 network on the same link
layer (ethernet network).
How to build the transparent bridge for IPv6 network? Is it same with
IPv4 that compile with bridge module?

Regards,

Pascal Hambourg ¼g¹D¡G

> [Supersedes due to mistakes]
>
> mclainet wrote :
> >>>
> >>>The ISP's router looks for 2001:ce0:3:63::1:1200 directly on its link
> >>>layer, which is logical since 2001:ce0:3:63::1:1200 belongs to
> >>>2001:ce0:3:63::/64. Are you sure that 2001:ce0:3:63::1:1000/116 is
> >>>allocated to you, and that routing is the proper way to use this block
> >>>instead of bridging ?
> >>
> > They assign 2001:ce0:3:63::1:1/64 ~ 2001:ce0:3:63::1:ffff/64 to me.
> >
> > 1. Is it mean that 2001:ce0:3:63::1:1000/116 is allocated to me?
>
> Not exactly.
> 2001:ce0:3:63::1:1000/116 = 2001:ce0:3:63::1:1000-2001:ce0:3:63::1:1fff
>
> 2001:ce0:3:63::1:1 - 2001:ce0:3:63::1:ffff = 2001:ce0:3:63::1::/112
> minus 2001:ce0:3:63::1::
>
> However, in any case, it does not mean that the address range is routed
> via your gateway.
>
> > 2. I can't get your idea "routing is the proper way to use this block
> > instead of bridging"
>
> I mean that /maybe/ the /112 or /116 block which is allocated to you is
> not a subnet but only a part of a larger /64 network on the same link
> layer (ethernet network). It is not a separate routed subnet, so other
> hosts in the /64 network expect to find your hosts in the same network
> and link layer.
>
> So in this situation your gateway should act :
> - as a layer 3 router with IPv4 traffic ;
> - as a layer 2 transparent bridge with IPv6 traffic.
>
> P.S.: Please don't top post.

mclainet wrote :
>>So in this situation your gateway should act :
>>- as a layer 3 router with IPv4 traffic ;
>>- as a layer 2 transparent bridge with IPv6 traffic.
>
> How to build the transparent bridge for IPv6 network? Is it same with
> IPv4 that compile with bridge module?

Yes. True layer 2 bridging is independent of layer 3 protocols such as
IPv4 or IPv6. You'll also need a kernel which supports ebtables
(ethernet bridge filtering) in order to allow only IPv6 packets through
the bridge. Ebtables support is included in 2.6 kernels, there is a
patch for 2.4 kernels.

However, to be honest, I don't know if and how it is possible to do at
the same time routing and bridging between the same pair of interfaces.
You may need two pairs of interfaces or two boxes in parallel, one for
IPv4 routing and one for IPv6 bridging.