Re: [PATCH v3 2/2] netfilter: xt_addrtype: ipv6 support

[Patrick McHardy <kaber@trash.net>]

Am 19.03.2011 00:26, schrieb Florian Westphal:
> Patrick McHardy <kaber@trash.net> wrote:
>> On 15.03.2011 19:49, Florian Westphal wrote:
> [..]
>>> +		rt = rt6_lookup(net, addr, NULL, ifindex, !!dev);
>>
>> Florian, I just noticed that this will pull in the IPv6 module just
>> by loading the xt_addrtype module. Can we convert this to use
>> nf_ip6_afinfo->route() instead?
> 
> I tried this, but i found two issues:
> - no netns support (nf_ip6_route passed init_net)
> - its not possible to ask for RT6_LOOKUP_F_IFACE flag in the
>   underlying fib6_rule_lookup() call.
> 
> But AFAICT the latter is needed to support the '--limit-iface-in/out'
> option.
> 
> Any idea?
> 
> Otherwise I think I'll have a go at extending afinfo->route() to pass
> in struct net* and a 'strict' argument (i.e. what rt6_lookup() has).
> 
> Unfortunately that would have to wait for 2.6.40...

Actually I'd consider that (especially the struct net *) a bugfix
since we shouldn't be pulling in the IPv6 module.