bgpd: IPv6 routes should not be attached to the interface on which they are received

[jharbott]

When an IPv6 route is received from a neighbor, the interface on which the BGP session is connected is attached to the route. This leads to broken forwarding when the announcement contains a next-hop that is reachable via a different next-hop. A common scenario for this is when the neighbor is a route-server or iBGP route-reflector, as those will mostly annouce routes pointing to other routers than themselves. The output of show ipv6 route in that case looks like:

network-node17# show ipv6 route
Codes: K - kernel route, C - connected, S - static, R - RIPng,
       O - OSPFv3, I - IS-IS, B - BGP, N - NHRP, T - Table,
       v - VNC, V - VNC-Direct, A - Babel, D - SHARP,
       > - selected route, * - FIB route

C>* 2001:db8:3020:4::/64 is directly connected, eth0.2, 00:12:43
B   2001:db8:3023::/64 [20/0] via 2001:db8:3020:4::6, eth0.8, 00:00:17
C>* fd84:e2bb:d6c8:1::/64 is directly connected, eth0.8, 00:12:43

Note how the BGP route contains eth0.8 which is the interface via which the BGP neighbor is connected, while the next-hop would be reachable via eth0.2.

This issue seems to have been inherited from quagga, see https://bugzilla.quagga.net/show_bug.cgi?id=851 but I verified it with frr 4.0. The patch in that bug does not solve the issue for me when tested against quagga 1.2.4.

[donaldsharp]

Give this branch a try: https://github.com/donaldsharp/frr/tree/bgp_v6_ifindex

Please note I have done absolutely positively no testing on this as of yet.

[subsecond]

I just tested the patch received through Cumulus GSS and can confirm that this fixes the issue for us.

Before

cumulus@prod-spine1-zrh1:mgmt-vrf:~$ net show route vrf public 2a06:c00::1ab:3353/128
RIB entry for 2a06:c00::1ab:3353/128 in vrf public
==================================================
Routing entry for 2a06:c00::1ab:3353/128
 Known via "bgp", distance 20, metric 0, vrf public
 Last update 01:59:12 ago
   2a06:c01:1:1902::7a0c:12, via public

FIB entry for 2a06:c00::1ab:3353/128 in vrf public
==================================================

After

cumulus@prod-spine1-zrh1:mgmt-vrf:~$ net show route vrf public 2a06:c00::1ab:3353/128
RIB entry for 2a06:c00::1ab:3353/128 in vrf public
==================================================
Routing entry for 2a06:c00::1ab:3353/128
  Known via "bgp", distance 20, metric 0, vrf public, best
  Last update 00:01:00 ago
  * 2a06:c01:1:1902::7a0c:12, via vlan1902

FIB entry for 2a06:c00::1ab:3353/128 in vrf public
==================================================
2a06:c00::1ab:3353 via 2a06:c01:1:1902::7a0c:12 dev vlan1902  proto bgp  metric 20  pref medium

Thanks a lot for the fix! I hope this will make it into CL 3.6.0.

[donaldsharp]

it will.