Implements 'next hop keep' option for BGP.

This option allows to keep the received next hop even in cases when
the route is sent to an interface with a different subnet.

4 files changed, 14 insertions, 3 deletions

diff --git a/doc/bird.sgml b/doc/bird.sgml
index 8e5641e0..300a71f3 100644
--- a/doc/bird.sgml
+++ b/doc/bird.sgml
@@ -1329,6 +1329,11 @@ for each neighbor using the following configuration parameters:
 	circumvent misconfigurations of other routers.  Default:
 	disabled.
 
+	<tag>next hop keep</tag> Forward the received Next Hop
+	attribute even in situations where the local address should be
+	used instead, like when the route is sent to an interface with
+	a different subnet. Default: disabled.
+
 	<tag>missing lladdr self|drop|ignore</tag>Next Hop attribute
 	in BGP-IPv6 sometimes contains just the global IPv6 address,
 	but sometimes it has to contain both global and link-local
diff --git a/proto/bgp/attrs.c b/proto/bgp/attrs.c
index 98b2f2c2..c27a4988 100644
--- a/proto/bgp/attrs.c
+++ b/proto/bgp/attrs.c
@@ -935,7 +935,8 @@ bgp_create_attrs(struct bgp_proto *p, rte *e, ea_list **attrs, struct linpool *p
       rta->dest != RTD_ROUTER ||
       ipa_equal(rta->gw, IPA_NONE) ||
       ipa_has_link_scope(rta->gw) ||
-      (!p->is_internal && (!p->neigh || (rta->iface != p->neigh->iface))))
+      (!p->is_internal && !p->cf->next_hop_keep &&
+       (!p->neigh || (rta->iface != p->neigh->iface))))
     set_next_hop(z, p->source_addr);
   else
     set_next_hop(z, rta->gw);
@@ -1003,10 +1004,13 @@ bgp_update_attrs(struct bgp_proto *p, rte *e, ea_list **attrs, struct linpool *p
   /* iBGP -> keep next_hop, eBGP multi-hop -> use source_addr,
    * eBGP single-hop -> keep next_hop if on the same iface.
    * If the next_hop is zero (i.e. link-local), keep only if on the same iface.
+   *
+   * Note that same-iface-check uses iface from route, which is based on gw.
    */
   a = ea_find(e->attrs->eattrs, EA_CODE(EAP_BGP, BA_NEXT_HOP));
   if (a && !p->cf->next_hop_self && 
-      ((p->is_internal && ipa_nonzero(*((ip_addr *) a->u.ptr->data))) ||
+      (p->cf->next_hop_keep ||
+       (p->is_internal && ipa_nonzero(*((ip_addr *) a->u.ptr->data))) ||
        (p->neigh && (e->attrs->iface == p->neigh->iface))))
     {
       /* Leave the original next hop attribute, will check later where does it point */
diff --git a/proto/bgp/bgp.h b/proto/bgp/bgp.h
index c3adf254..13c7fd80 100644
--- a/proto/bgp/bgp.h
+++ b/proto/bgp/bgp.h
@@ -24,6 +24,7 @@ struct bgp_config {
   int multihop;				/* Number of hops if multihop */
   int ttl_security;			/* Enable TTL security [RFC5082] */
   int next_hop_self;			/* Always set next hop to local IP address */
+  int next_hop_keep;			/* Do not touch next hop attribute */
   int missing_lladdr;			/* What we will do when we don' know link-local addr, see MLL_* */
   int gw_mode;				/* How we compute route gateway from next_hop attr, see GW_* */
   int compare_path_lengths;		/* Use path lengths when selecting best route */
diff --git a/proto/bgp/config.Y b/proto/bgp/config.Y
index 8b80d7fd..d5e5aaca 100644
--- a/proto/bgp/config.Y
+++ b/proto/bgp/config.Y
@@ -76,7 +76,8 @@ bgp_proto:
  | bgp_proto KEEPALIVE TIME expr ';' { BGP_CFG->keepalive_time = $4; }
  | bgp_proto MULTIHOP ';' { BGP_CFG->multihop = 64; }
  | bgp_proto MULTIHOP expr ';' { BGP_CFG->multihop = $3; if (($3<1) || ($3>255)) cf_error("Multihop must be in range 1-255"); }
- | bgp_proto NEXT HOP SELF ';' { BGP_CFG->next_hop_self = 1; }
+ | bgp_proto NEXT HOP SELF ';' { BGP_CFG->next_hop_self = 1; BGP_CFG->next_hop_keep = 0; }
+ | bgp_proto NEXT HOP KEEP ';' { BGP_CFG->next_hop_keep = 1; BGP_CFG->next_hop_self = 0; }
  | bgp_proto MISSING LLADDR SELF ';' { BGP_CFG->missing_lladdr = MLL_SELF; }
  | bgp_proto MISSING LLADDR DROP ';' { BGP_CFG->missing_lladdr = MLL_DROP; }
  | bgp_proto MISSING LLADDR IGNORE ';' { BGP_CFG->missing_lladdr = MLL_IGNORE; }