Re: [patch net] ipv6: do not create neighbor entries for local delivery

[Marcelo Ricardo Leitner <mleitner@redhat.com>]

Em 12-08-2013 19:26, Hannes Frederic Sowa escreveu:
> Hi Marcelo!
>
> On Mon, Aug 12, 2013 at 03:09:19PM -0300, Marcelo Ricardo Leitner wrote:
>> Hannes, would something like this be acceptable? I'm hoping it's not too
>> ugly/hacky... as far as I could track back, input and output routines were
>> merged mainly due code similarity.
>
> Your idea seems sound and I don't think it is very ugly or hacky. It's
> as minimal as a stable-only patch should be. But we could simplify the
> logic a bit. ;) See below.
>
>> TPROXY scenario needs to not create this neighbor entries on INPUT path,
>> while Debabrata ping test needs it on OUTPUT path. This patch limits my
>> previous patch to INPUT only then.
>
> Yes, agreed. I don't see anything which could break because of this patch.
> So I would go with it.
>
>> Initial testing here seems good, TPROXY seems to be working as expected and
>> also the ping6 test.
>>
>> What do you think?

Aye Hannes, thanks! I'll rework the patch based on your points, do some more 
testings in here and post it probably only by tomorrow.

Thanks!
Marcelo

>> diff --git a/net/ipv6/route.c b/net/ipv6/route.c
>> index 18ea73c..603f9d9 100644
>> --- a/net/ipv6/route.c
>> +++ b/net/ipv6/route.c
>> @@ -791,7 +791,7 @@ static struct rt6_info *rt6_alloc_clone(struct rt6_info *ort,
>>   }
>>
>>   static struct rt6_info *ip6_pol_route(struct net *net, struct fib6_table *table, int oif,
>> -				      struct flowi6 *fl6, int flags)
>> +				      struct flowi6 *fl6, int flags, int output)
>
> 								     bool input
>
>>   {
>>   	struct fib6_node *fn;
>>   	struct rt6_info *rt, *nrt;
>> @@ -799,8 +799,11 @@ static struct rt6_info *ip6_pol_route(struct net *net, struct fib6_table *table,
>>   	int attempts = 3;
>>   	int err;
>>   	int reachable = net->ipv6.devconf_all->forwarding ? 0 : RT6_LOOKUP_F_REACHABLE;
>> +	int local = RTF_NONEXTHOP;
>>
>>   	strict |= flags & RT6_LOOKUP_F_IFACE;
>> +	if (!output)
>> +			local |= RTF_LOCAL;
>
>
> 	if (input)
> 			local |= RTF_LOCAL;
>
>>
>>   relookup:
>>   	read_lock_bh(&table->tb6_lock);
>> @@ -820,7 +823,7 @@ restart:
>>   	read_unlock_bh(&table->tb6_lock);
>>
>>   	if (!dst_get_neighbour_raw(&rt->dst)
>> -	    && !(rt->rt6i_flags & (RTF_NONEXTHOP | RTF_LOCAL)))
>> +	    && !(rt->rt6i_flags & local))
>>   		nrt = rt6_alloc_cow(rt, &fl6->daddr, &fl6->saddr);
>>   	else if (!(rt->dst.flags & DST_HOST))
>>   		nrt = rt6_alloc_clone(rt, &fl6->daddr);
>> @@ -864,7 +867,7 @@ out2:
>>   static struct rt6_info *ip6_pol_route_input(struct net *net, struct fib6_table *table,
>>   					    struct flowi6 *fl6, int flags)
>>   {
>> -	return ip6_pol_route(net, table, fl6->flowi6_iif, fl6, flags);
>> +	return ip6_pol_route(net, table, fl6->flowi6_iif, fl6, flags, 0);
>
> 								      true);
>
>
>>   }
>>
>>   void ip6_route_input(struct sk_buff *skb)
>> @@ -890,7 +893,7 @@ void ip6_route_input(struct sk_buff *skb)
>>   static struct rt6_info *ip6_pol_route_output(struct net *net, struct fib6_table *table,
>>   					     struct flowi6 *fl6, int flags)
>>   {
>> -	return ip6_pol_route(net, table, fl6->flowi6_oif, fl6, flags);
>> +	return ip6_pol_route(net, table, fl6->flowi6_oif, fl6, flags, 1);
>
> 								      false);
>
>>   }
>>
>>   struct dst_entry * ip6_route_output(struct net *net, const struct sock *sk,
>
> Thanks,
>
>    Hannes
>
>