[quagga-users 14807] Re: IPv6 BGP routes from a directly connected peer are not accepted in the routing table

> Hi!
>
> > ???I've got exactly the same problem (Quagga 1.2.1 on FreeBSD
> > 11.1-BETA2):
> >
> > R2# sh bgp neighbors 2001:db8:23::3 received-routes
> > BGP table version is 0, local router ID is 192.168.56.12
> > Status codes: s suppressed, d damped, h history, * valid, > best, =
> > multipath,
> >               i internal, r RIB-failure, S Stale, R Removed
> > Origin codes: i - IGP, e - EGP, ? - incomplete
> >
> >    Network          Next Hop            Metric LocPrf Weight Path
> > *> 2001:db8:3::/64  2001:db8:23::3           0     100       0 i
> > *> 2001:db8:34::/64 2001:db8:23::3           0     100       0 i
> >
> > Total number of prefixes 2
> > R2# sh ipv6 bgp 2001:db8:34::/64
> > BGP routing table entry for 2001:db8:34::/64
> > Paths: (1 available, no best path)
> >   Not advertised to any peer
> >   Local
> >     2001:db8:23::3 (inaccessible) from 2001:db8:23::3 (10.0.3.3)
> >     (fe80::5a9c:fcff:fe02:303)
> >       Origin IGP, metric 0, localpref 100, invalid, internal
> >       Last update: Fri Jun 23 17:55:43 2017
>
> Please check, there's this well-known bug:
>
> https://bugzilla.quagga.net/show_bug.cgi?id=870
>
> https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=214481


Hi,

not sure if this is the same as quagga bz#870 - I haven't (yet) seen
OutQ stuck at >0.

But I have encountered the same "inaccessible" ipv6 nexthops problem as
above, but on Linux.

Suspect log entries:

2017/09/14 21:06:28 [10070] BGP: bgp_ensure_nexthop: NHT could not ensure, failed to get rn!
2017/09/14 21:06:28 [10070] BGP: bgp_update_main(0.0.0.0): NH unresolved

I could resolve that by comparing the bgp_ensure_nexthop() function in
bgp/bgp_nht.c to the counterpart bgp_find_or_add_nexthop() function from
FRR, and then cherrypicking something that seemed related to ipv6 link
local nexthops.

I don't claim to fully understand the code, and I hope I am not
infringing on the copyright of the author of that code. Patch is
attached.


Regards
Matthias

_______________________________________________
Quagga-users mailing list
Quagga-users-UOy77sIEA+cAd7ICUelF/[email protected]
https://lists.quagga.net/mailman/listinfo/quagga-users

--- a/bgpd/bgp_nht.c	2017-03-10 13:55:06.000000000 +0100
+++ b/bgpd/bgp_nht.c	2017-09-14 22:23:27.310570973 +0200
@@ -192,12 +192,44 @@ bgp_unlink_nexthop_by_peer (struct peer
   bgp_unlink_nexthop_check (bnc);
 }
 
+
+#define BGP_ATTR_NEXTHOP_AFI_IP6(attr) \
+  (! CHECK_FLAG (attr->flag, ATTR_FLAG_BIT (BGP_ATTR_NEXT_HOP)) && \
+   (attr)->extra && ((attr)->extra->mp_nexthop_len == 16 || \
+    (attr)->extra->mp_nexthop_len == 32))
+
 int
 bgp_ensure_nexthop (struct bgp_info *ri, struct peer *peer,
                     int connected)
 {
   struct bgp_node *rn;
   struct bgp_nexthop_cache *bnc;
+  struct prefix p;
+  int is_bgp_static_route = 0;
+
+  afi_t afi = 0;
+
+  if (ri)
+    {
+      is_bgp_static_route = ((ri->type == ZEBRA_ROUTE_BGP) &&
+                            (ri->sub_type == BGP_ROUTE_STATIC)) ? 1 : 0;
+
+      /* Since Extended Next-hop Encoding (RFC5549) support, we want to derive
+         address-family from the next-hop. */
+      if (!is_bgp_static_route)
+        afi = BGP_ATTR_NEXTHOP_AFI_IP6(ri->attr) ? AFI_IP6 : AFI_IP;
+
+      /* This will return TRUE if the global IPv6 NH is a link local addr */
+      if (make_prefix(afi, ri, &p) < 0)
+       return 1;
+    }
+  else if (peer)
+    {
+      /* Don't register link local NH */
+      if (afi == AFI_IP6 && IN6_IS_ADDR_LINKLOCAL (&peer->su.sin6.sin6_addr))
+       return 1;
+    }
+
   
   rn = bgp_get_nexthop_rn (ri, peer);