[PATCH v2] dhcp6c: DHCPv6 client

Leonid Lisovskiy lly.dev at gmail.com
Mon Nov 7 18:57:57 UTC 2011


On Mon, Nov 7, 2011 at 10:23 PM, Denys Vlasenko
<vda.linux at googlemail.com> wrote:
>>>
>>>> One upstream(listen) interface - yes. But simplification to single
>>>> downstream (IA, PD) interface seems to be bad idea.
>>>
>>> Can you elaborate why is it a bad idea?
>>
>> Of course, yes. For a example, take linux(busybox)-based CPE -
>> router/gateway/access point. It has at least two interfaces - WAN &
>> LAN. In case of DHCPv6 used by ISP to provide IPv6 access, we have to:
>>
>> 1) assign address(IA) on WAN interface
>> 2) assing prefix(PD) on LAN interface
>>
>> Thus, we must support two interfaces, at least.
>
> This does not explain the problem for me.
> You have two interfaces? No problem: run two udchpc6's,
> one per interface.

We must issue request/renew/release _both_ IA/PD queries on WAN
interface(which connects to ISP equipment). LAN connected to customer
equipment only. Any broadcast/multicast queries on LAN interface will
never rich ISP without special ip(6)tables rules.

For IPv4 scheme was simple - udhcpc client assign address on WAN, LAN
has predefined private addresses (192.168.0.0/24). Customers works via
NAT.

For IPv6 similar scheme don't work - we hasn't private addresses at
all. ISP must give our CPE /128 address + /64 prefix, at least. And
this info can be in single DHCPv6 packet.


Leonid


More information about the busybox mailing list