On Tue, Aug 03, 2021 at 10:13:09AM +0300, Alexander Shikov wrote:
Hello!
Yesterday we noticed a strange behavior of Bird 1.6.8 with handling one IPv6 BGP route:
birdc6 show route for 2a09:8387:ffff:: all BIRD 1.6.8 ready. 2a09:8387:ffff::/48 via 2001:7f8:63::f2 on bge0 [SIMTELECOM 2021-08-03 00:07:16] * (100) [AS51972i] Type: BGP unicast univ BGP.origin: IGP BGP.as_path: 51972 BGP.next_hop: :: BGP.med: 0 BGP.local_pref: 100 BGP.community: (31210,31210) BGP.ext_community: (rt, 31210, 31210)
The first (best) route has next-hop '::' and routers of some of our customers started dropping BGP sessions with 'Invalid next-hop' cause.
I dumped packets coming from customer 2001:7f8:63::f2 and it is really sending next-hop as '::'.
I enabled logging of 'gw' and 'from' route attributes for net 2a09:8387:ffff::/48, and I got the following:
Aug 3 09:57:10 <daemon.info> rs1 bird6: DEBUG: gw:2001:7f8:63::f4 ---- from:2001:7f8:63::f4 Aug 3 09:57:10 <daemon.info> rs1 bird6: DEBUG: gw:2001:7f8:63::69 ---- from:2001:7f8:63::69
Thus 'gw' and 'from' variables both contain IPv6 address of the peer. But shouldn't 'gw' contain '::'?
Hello The bgp_next_hop is ::, 'gw' a is different attribute - immediate next hop computed from bgp_next_hop, local routing table and other factors. That is shown after 'via' in 'show route' output. When a route is propagated on a route server or a route reflector, original bgp_next_hop is kept. BTW, bgp_next_hop :: is clearly invalid (without attached link-local bgp_next_hop). -- Elen sila lumenn' omentielvo Ondrej 'Santiago' Zajicek (email: santiago@crfreenet.org) OpenPGP encrypted e-mails preferred (KeyID 0x11DEADC3, wwwkeys.pgp.net) "To err is human -- to blame it on a computer is even more so."