Better check for IPv6 compatibility with the BGP peer.
authorBenjamin Cama <benoar@dolka.fr>
Sat, 23 Jul 2011 16:53:12 +0000 (18:53 +0200)
committerBenjamin Cama <benoar@dolka.fr>
Thu, 28 Jul 2011 13:01:57 +0000 (15:01 +0200)
We would advertise IPv6 routes to non-multiprotocol aware peers. Fix that.
Also, fix the way we parse options, to handle multiple optional parameters with
one capability in it (or many; it's just the way quagga send them).

Signed-off-by: Benjamin Cama <benoar@dolka.fr>
bgp.c

diff --git a/bgp.c b/bgp.c
index f0ff879..795ffc5 100644 (file)
--- a/bgp.c
+++ b/bgp.c
@@ -556,7 +556,8 @@ int bgp_add_route6(struct in6_addr ip, int prefixlen)
 
     /* flag established peers for update */
     for (i = 0; i < BGP_NUM_PEERS; i++)
-       if (bgp_peers[i].state == Established)
+       if (bgp_peers[i].state == Established
+               && bgp_peers[i].mp_handling == HandleIPv6Routes)
            bgp_peers[i].update_routes6 = 1;
 
     LOG(4, 0, 0, "Registered BGP route %s/%d\n",
@@ -656,7 +657,8 @@ int bgp_del_route6(struct in6_addr ip, int prefixlen)
 
     /* flag established peers for update */
     for (i = 0; i < BGP_NUM_PEERS; i++)
-       if (bgp_peers[i].state == Established)
+       if (bgp_peers[i].state == Established
+               && bgp_peers[i].mp_handling == HandleIPv6Routes)
            bgp_peers[i].update_routes6 = 1;
 
     LOG(4, 0, 0, "Removed BGP route %s/%d\n",
@@ -1185,7 +1187,7 @@ static int bgp_handle_input(struct bgp_peer *peer)
                    param_offset < data.opt_len;
                    param_offset += 2 + param->len)
            {
-               param = (struct bgp_opt_param *)(&data.opt_params + param_offset);
+               param = (struct bgp_opt_param *)((char *)&data.opt_params + param_offset);
 
                /* sensible check */
                if (data.opt_len - param_offset < 2
@@ -1210,11 +1212,8 @@ static int bgp_handle_input(struct bgp_peer *peer)
 
                capabilities_len = param->len;
                capabilities = (char *)&param->value;
-           }
 
-           /* look for BGP multiprotocol capability */
-           if (capabilities)
-           {
+               /* look for BGP multiprotocol capability */
                for (capability_offset = 0;
                        capability_offset < capabilities_len;
                        capability_offset += 2 + capability->len)
@@ -1258,10 +1257,19 @@ static int bgp_handle_input(struct bgp_peer *peer)
                        continue;
                    }
 
+                   /* yes it can! */
                    peer->mp_handling = HandleIPv6Routes;
                }
            }
 
+           if (peer->mp_handling != HandleIPv6Routes)
+           {
+               peer->mp_handling = DoesntHandleIPv6Routes;
+               if (config->ipv6_prefix.s6_addr[0])
+                   LOG(1, 0, 0, "Warning: BGP peer %s doesn't handle IPv6 prefixes updates\n",
+                           peer->name);
+           }
+
            /* next transition requires an exchange of keepalives */
            bgp_send_keepalive(peer);
        }
@@ -1363,8 +1371,9 @@ static int bgp_send_open(struct bgp_peer *peer)
 
     /* if we know peer doesn't support MP (mp_handling == DoesntHandleIPv6Routes)
        then don't add this parameter */
-    if (peer->mp_handling == HandlingUnknown
-           || peer->mp_handling == HandleIPv6Routes)
+    if (config->ipv6_prefix.s6_addr[0]
+           && (peer->mp_handling == HandlingUnknown
+               || peer->mp_handling == HandleIPv6Routes))
     {
        /* construct the param and capability */
        cap_mp_ipv6.code = BGP_CAP_CODE_MP;