Patchwork [v2] ipv6: fix ecmp lookup when oif is specified

login
register
mail settings
Submitter Nicolas Dichtel
Date June 14, 2013, 8:32 a.m.
Message ID <1371198778-26099-1-git-send-email-nicolas.dichtel@6wind.com>
Download mbox | patch
Permalink /patch/251265/
State Rejected
Delegated to: David Miller
Headers show

Comments

Nicolas Dichtel - June 14, 2013, 8:32 a.m.
There is no reason to skip ECMP lookup when oif is specified, but this implies
to check oif given by user when selecting another route.
When the new route does not match oif requirement, we simply keep the initial
one.

Spotted-by: dingzhi <zhi.ding@6wind.com>
Signed-off-by: Nicolas Dichtel <nicolas.dichtel@6wind.com>
---

v2: check oif requirement in rt6_multipath_select()

 net/ipv6/route.c | 14 +++++++++-----
 1 file changed, 9 insertions(+), 5 deletions(-)
David Miller - June 24, 2013, 1:40 a.m.
From: Nicolas Dichtel <nicolas.dichtel@6wind.com>
Date: Fri, 14 Jun 2013 10:32:58 +0200

> There is no reason to skip ECMP lookup when oif is specified, but this implies
> to check oif given by user when selecting another route.
> When the new route does not match oif requirement, we simply keep the initial
> one.
> 
> Spotted-by: dingzhi <zhi.ding@6wind.com>
> Signed-off-by: Nicolas Dichtel <nicolas.dichtel@6wind.com>

IPV4 routing has elided the multipath lookup when the interface is specified
for nearly two decades.

So two things:

1) A decision that old must have a good reason, and you must do some
   research to figure out exactly what that reason is.

2) If ipv4 is found to be wrong too, we must bring both ipv4 and ipv6
   into the same behavior at the same time.

But I do not want to make this change for another reason, it potentially
breaks things for someone, somewhere.  And I see very miniscule, if any,
value to this new behavior.  The risk far outweighs the gains.

Sorry, I really have to reject this patch for now.
--
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Nicolas Dichtel - June 24, 2013, 4:44 p.m.
Le 24/06/2013 03:40, David Miller a écrit :
> From: Nicolas Dichtel <nicolas.dichtel@6wind.com>
> Date: Fri, 14 Jun 2013 10:32:58 +0200
>
>> There is no reason to skip ECMP lookup when oif is specified, but this implies
>> to check oif given by user when selecting another route.
>> When the new route does not match oif requirement, we simply keep the initial
>> one.
>>
>> Spotted-by: dingzhi <zhi.ding@6wind.com>
>> Signed-off-by: Nicolas Dichtel <nicolas.dichtel@6wind.com>
>
> IPV4 routing has elided the multipath lookup when the interface is specified
> for nearly two decades.
>
> So two things:
>
> 1) A decision that old must have a good reason, and you must do some
>     research to figure out exactly what that reason is.
>
> 2) If ipv4 is found to be wrong too, we must bring both ipv4 and ipv6
>     into the same behavior at the same time.
In fact, routing engines are different in forwarding case:

IPv4 case:
fl4->flowi4_oif is 0 when a packet is forwarded and this value is checked to 
select or not multipath functions.

IPv6 case:
The check is done against the argument oif (not fl6->flowi6_oif) of the function 
ip6_pol_route(). And this argument is set to the input interface by 
ip6_pol_route_input().

Functions call:
   - ip6_rcv_finish() -> ip6_route_input() -> ip6_route_input_lookup() -> 
fib6_rule_lookup() -> ip6_pol_route_input() -> ip6_pol_route()

Will you accept the patch if I remove only the check of oif argument in 
ip6_pol_route() (and leave ip6_pol_route_lookup() untouched).

My goal was only to have ECMP working in the forwarding case.
--
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
David Miller - June 26, 2013, 10:19 p.m.
From: Nicolas Dichtel <nicolas.dichtel@6wind.com>
Date: Mon, 24 Jun 2013 18:44:24 +0200

> Will you accept the patch if I remove only the check of oif argument
> in ip6_pol_route() (and leave ip6_pol_route_lookup() untouched).
> 
> My goal was only to have ECMP working in the forwarding case.

Yes, if you make it such that, like ipv4, oif is respected for output
route lookups.
--
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Patch

diff --git a/net/ipv6/route.c b/net/ipv6/route.c
index ad0aa6b..4a5d259 100644
--- a/net/ipv6/route.c
+++ b/net/ipv6/route.c
@@ -83,6 +83,7 @@  static void		ip6_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
 					   struct sk_buff *skb, u32 mtu);
 static void		rt6_do_redirect(struct dst_entry *dst, struct sock *sk,
 					struct sk_buff *skb);
+static int rt6_score_route(struct rt6_info *rt, int oif, int strict);
 
 #ifdef CONFIG_IPV6_ROUTE_INFO
 static struct rt6_info *rt6_add_route_info(struct net *net,
@@ -394,7 +395,8 @@  static int rt6_info_hash_nhsfn(unsigned int candidate_count,
 }
 
 static struct rt6_info *rt6_multipath_select(struct rt6_info *match,
-					     struct flowi6 *fl6)
+					     struct flowi6 *fl6, int oif,
+					     int strict)
 {
 	struct rt6_info *sibling, *next_sibling;
 	int route_choosen;
@@ -408,6 +410,8 @@  static struct rt6_info *rt6_multipath_select(struct rt6_info *match,
 				&match->rt6i_siblings, rt6i_siblings) {
 			route_choosen--;
 			if (route_choosen == 0) {
+				if (rt6_score_route(sibling, oif, strict) < 0)
+					break;
 				match = sibling;
 				break;
 			}
@@ -742,8 +746,8 @@  static struct rt6_info *ip6_pol_route_lookup(struct net *net,
 restart:
 	rt = fn->leaf;
 	rt = rt6_device_match(net, rt, &fl6->saddr, fl6->flowi6_oif, flags);
-	if (rt->rt6i_nsiblings && fl6->flowi6_oif == 0)
-		rt = rt6_multipath_select(rt, fl6);
+	if (rt->rt6i_nsiblings)
+		rt = rt6_multipath_select(rt, fl6, fl6->flowi6_oif, flags);
 	BACKTRACK(net, &fl6->saddr);
 out:
 	dst_use(&rt->dst, jiffies);
@@ -875,8 +879,8 @@  restart_2:
 
 restart:
 	rt = rt6_select(fn, oif, strict | reachable);
-	if (rt->rt6i_nsiblings && oif == 0)
-		rt = rt6_multipath_select(rt, fl6);
+	if (rt->rt6i_nsiblings)
+		rt = rt6_multipath_select(rt, fl6, oif, strict | reachable);
 	BACKTRACK(net, &fl6->saddr);
 	if (rt == net->ipv6.ip6_null_entry ||
 	    rt->rt6i_flags & RTF_CACHE)