[lvs-users] Problem with LVS and IPV6 DR

Simon Horman horms at verge.net.au
Tue Oct 15 10:42:21 BST 2013


On Mon, Oct 14, 2013 at 02:32:21PM +0100, Mark Brooks wrote:
> Hi All - We are having a problem with IPV6 DR, I seem unable to get it to work
> using the latest kernels. So far I have tried 3.9.7,3.10.14 and 3.11.4
> and all appear to produce the same results I see SYN_RECV in the
> connection table for all connections from my test client. Looking at
> packet captures from our appliance it looks like the packet is not
> getting forwarded. You can see SYN packet come in from the client
> making the request but no data is forwarded to the real server. Using
> the same server and the same configuration all I do is reboot to
> switch between the kernels It works straight away on 2.6.35. My
> configuration is as follows -
> 
> APPLIANCE CONFIG ---------------------------------
> 
> loadbalancer eth0 - fde6:d14c:3089::6061/123 scope global
> loadbalancer eth0 vip - fde6:d14c:3089::6063/123 scope global
> 
> ipvsadm config - IP Virtual Server version 1.2.1 (size=32768)
> Prot LocalAddress:Port Scheduler Flags
>   -> RemoteAddress:Port           Forward Weight ActiveConn InActConn
> TCP  [fde6:d14c:3089::6063]:80 wlc
>   -> [fde6:d14c:3089::6070]:80    Route   100    0          0
>   -> [fde6:d14c:3089::6071]:80    Route   100    0          0
> 
> --------------------------------------------------
> 
> Real server 1 -
> eth1 - inet6 fde6:d14c:3089::6070/123 scope global
> lo - inet6 fde6:d14c:3089::6063/128 scope global
> 
> Real server 2 -
> eth1 - inet6 fde6:d14c:3089::6071/123 scope global
> lo - inet6 fde6:d14c:3089::6063/128 scope global

It seems that this problem was introduced between v3.8 and v3.9
by the following commit.

For 3.9 at least a possible work around is to reverse the commit.

But before I make any further progress I must go to dinner.


>From 6fd6ce2056de27090f4723138e0ce0a816f4ba0d Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?YOSHIFUJI=20Hideaki=20/=20=E5=90=89=E8=97=A4=E8=8B=B1?=
 =?UTF-8?q?=E6=98=8E?= <yoshfuji at linux-ipv6.org>
Date: Thu, 17 Jan 2013 12:54:00 +0000
Subject: [PATCH] ipv6: Do not depend on rt->n in ip6_finish_output2().

If neigh is not found, create new one.

Signed-off-by: YOSHIFUJI Hideaki <yoshfuji at linux-ipv6.org>
Signed-off-by: David S. Miller <davem at davemloft.net>
---
 net/ipv6/ip6_output.c | 18 +++++++++++++-----
 1 file changed, 13 insertions(+), 5 deletions(-)

diff --git a/net/ipv6/ip6_output.c b/net/ipv6/ip6_output.c
index 7c00edb..b0895f5 100644
--- a/net/ipv6/ip6_output.c
+++ b/net/ipv6/ip6_output.c
@@ -88,7 +88,8 @@ static int ip6_finish_output2(struct sk_buff *skb)
 	struct dst_entry *dst = skb_dst(skb);
 	struct net_device *dev = dst->dev;
 	struct neighbour *neigh;
-	struct rt6_info *rt;
+	struct in6_addr *nexthop;
+	int ret;
 
 	skb->protocol = htons(ETH_P_IPV6);
 	skb->dev = dev;
@@ -123,10 +124,17 @@ static int ip6_finish_output2(struct sk_buff *skb)
 				skb->len);
 	}
 
-	rt = (struct rt6_info *) dst;
-	neigh = rt->n;
-	if (neigh)
-		return dst_neigh_output(dst, neigh, skb);
+	rcu_read_lock_bh();
+	nexthop = rt6_nexthop((struct rt6_info *)dst, &ipv6_hdr(skb)->daddr);
+	neigh = __ipv6_neigh_lookup_noref(dst->dev, nexthop);
+	if (unlikely(!neigh))
+		neigh = __neigh_create(&nd_tbl, nexthop, dst->dev, false);
+	if (!IS_ERR(neigh)) {
+		ret = dst_neigh_output(dst, neigh, skb);
+		rcu_read_unlock_bh();
+		return ret;
+	}
+	rcu_read_unlock_bh();
 
 	IP6_INC_STATS_BH(dev_net(dst->dev),
 			 ip6_dst_idev(dst), IPSTATS_MIB_OUTNOROUTES);
-- 
1.8.4




More information about the lvs-users mailing list