Implements proper multihop BGP.

Also does some incompatible changes to config file syntax,
like removing 'via IP' from multihop option.

7 files changed, 80 insertions, 52 deletions

diff --git a/doc/bird.sgml b/doc/bird.sgml
index a6fa4df..82ff630 100644
--- a/doc/bird.sgml
+++ b/doc/bird.sgml
@@ -272,7 +272,7 @@ protocol rip {
 	listen to IPv6 connections only. This is needed if you want to
 	run both bird and bird6 on the same port.
 
-	<tag>timeformat route|protocol|base|log "<m/format1/" [<m/limit> "<m/format2/"]</tag>
+	<tag>timeformat route|protocol|base|log "<m/format1/" [<m/limit/ "<m/format2/"]</tag>
 	This option allows to specify a format of date/time used by
 	BIRD.  The first argument specifies for which purpose such
 	format is used. <cf/route/ is a format used in 'show route'
@@ -280,8 +280,8 @@ protocol rip {
 	command output, <cf/base/ is used for other commands and
 	<cf/log/ is used in a log file.
 
-	"<m/format1/" is a format string using <i/strftime(3)/
-	notation (see <i/man strftime/ for details). <m/limit> and
+	"<m/format1/" is a format string using <it/strftime(3)/
+	notation (see <it/man strftime/ for details). <m/limit> and
 	"<m/format2/" allow to specify the second format string for
 	times in past deeper than <m/limit/ seconds. There are two
 	shorthands: <cf/iso long/ is a ISO 8601 date/time format
@@ -976,10 +976,13 @@ This allows to set routing policy and all the other parameters differently
 for each neighbor using the following configuration parameters:
 
 <descrip>
-	<tag>local as <m/number/</tag> Define which AS we are part of. (Note that
-	contrary to other IP routers, BIRD is able to act as a router located
-	in multiple AS'es simultaneously, but in such cases you need to tweak
-	the BGP paths manually in the filters to get consistent behavior.)
+	<tag>local <m/[ip]/] as <m/number/</tag> Define which AS we
+	are part of. (Note that contrary to other IP routers, BIRD is
+	able to act as a router located in multiple AS'es
+	simultaneously, but in such cases you need to tweak the BGP
+	paths manually in the filters to get consistent behavior.)
+	Optional <cf/ip/ argument specifies a source address,
+	equivalent to the <cf/source address/ option (see below).
 	This parameter is mandatory.
 
 	<tag>neighbor <m/ip/ as <m/number/</tag> Define neighboring router
@@ -988,16 +991,27 @@ for each neighbor using the following configuration parameters:
 	of your router's interfaces. In case the neighbor is in the same AS
 	as we are, we automatically switch to iBGP. This parameter is mandatory.
 
-	<tag>multihop <m/number/ via <m/ip/</tag> Configure multihop BGP to a
-	neighbor which is connected at most <m/number/ hops far and to which
-	we should route via our direct neighbor with address <m/ip/.
+	<tag>multihop <m/[number]/]</tag> Configure multihop BGP
+	session to a neighbor that isn't directly connected.
+	Accurately, this option should be used if the configured
+	neighbor IP address does not match with any local network
+	subnets. Such IP address have to be reachable through system
+	routing table. For multihop BGP it is recommended to
+	explicitly configure <cf/source address/ to have it
+	stable. Optional <cf/number/ argument can be used to limit TTL
+	(the number of hops).
 	Default: switched off.
 
+	<tag>source address <m/ip/</tag> Define local address we
+	should use for next hop calculation and as a source address
+	for the BGP session. Default: the address of the local
+	end of the interface our neighbor is connected to.
+
 	<tag>next hop self</tag> Avoid calculation of the Next Hop
-	attribute and always advertise our own source address (see
-	below) as a next hop.  This needs to be used only occasionally
-	to circumvent misconfigurations of other routers.
-	Default: disabled.
+	attribute and always advertise our own source address as a
+	next hop.  This needs to be used only occasionally to
+	circumvent misconfigurations of other routers.  Default:
+	disabled.
 
 	<tag>missing lladdr self|drop|ignore</tag>Next Hop attribute
 	in BGP-IPv6 sometimes contains just the global IPv6 address,
@@ -1015,10 +1029,6 @@ for each neighbor using the following configuration parameters:
 	that case default is <cf/drop/, because route servers usually
 	does not forward packets ifselves.
 	
-	<tag>source address <m/ip/</tag> Define local address we should use
-	for next hop calculation. Default: the address of the local end
-	of the interface our neighbor is connected to.
-
 	<tag>password <m/string/</tag> Use this password for MD5 authentication
 	of BGP sessions. Default: no authentication. Password has to be set by
 	external utility (e.g. setkey(8)) on BSD systems.
diff --git a/proto/bgp/attrs.c b/proto/bgp/attrs.c
index 8743358..039828f 100644
--- a/proto/bgp/attrs.c
+++ b/proto/bgp/attrs.c
@@ -752,7 +752,7 @@ bgp_get_bucket(struct bgp_proto *p, net *n, ea_list *attrs, int originate)
 
   /* Check if next hop is valid */
   a = ea_find(new, EA_CODE(EAP_BGP, BA_NEXT_HOP));
-  if (!a || ipa_equal(p->next_hop, *(ip_addr *)a->u.ptr->data))
+  if (!a || ipa_equal(p->cf->remote_ip, *(ip_addr *)a->u.ptr->data))
     {
       log(L_ERR "%s: Invalid NEXT_HOP attribute in route %I/%d", p->p.name, n->n.prefix, n->n.pxlen);
       return NULL;
@@ -808,7 +808,6 @@ bgp_rt_notify(struct proto *P, rtable *tbl UNUSED, net *n, rte *new, rte *old UN
   bgp_schedule_packet(p->conn, PKT_UPDATE);
 }
 
-
 static int
 bgp_create_attrs(struct bgp_proto *p, rte *e, ea_list **attrs, struct linpool *pool)
 {
@@ -834,12 +833,14 @@ bgp_create_attrs(struct bgp_proto *p, rte *e, ea_list **attrs, struct linpool *p
       put_u32(z+2, p->local_as);
     }
 
+  /* iBGP -> use gw, eBGP multi-hop -> use source_addr,
+     eBGP single-hop -> use gw if on the same iface */
   z = bgp_set_attr_wa(ea->attrs+2, pool, BA_NEXT_HOP, NEXT_HOP_LENGTH);
   if (p->cf->next_hop_self ||
       rta->dest != RTD_ROUTER ||
-      ipa_equal(e->attrs->gw, IPA_NONE) ||
+      ipa_equal(rta->gw, IPA_NONE) ||
       ipa_has_link_scope(rta->gw) ||
-      (!p->is_internal && (rta->iface != p->neigh->iface)))
+      (!p->is_internal && (!p->neigh || (rta->iface != p->neigh->iface))))
     set_next_hop(z, p->source_addr);
   else
     set_next_hop(z, rta->gw);
@@ -904,8 +905,11 @@ bgp_update_attrs(struct bgp_proto *p, rte *e, ea_list **attrs, struct linpool *p
 	bgp_attach_attr(attrs, pool, BA_MULTI_EXIT_DISC, 0);
     }
 
+  /* iBGP -> keep next_hop, eBGP multi-hop -> use source_addr,
+     eBGP single-hop -> keep next_hop if on the same iface */
   a = ea_find(e->attrs->eattrs, EA_CODE(EAP_BGP, BA_NEXT_HOP));
-  if (a && !p->cf->next_hop_self && (p->is_internal || (!p->is_internal && e->attrs->iface == p->neigh->iface)))
+  if (a && !p->cf->next_hop_self && 
+      (p->is_internal || (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.c b/proto/bgp/bgp.c
index b36c4a3..3aa8845 100644
--- a/proto/bgp/bgp.c
+++ b/proto/bgp/bgp.c
@@ -145,6 +145,10 @@ bgp_startup_timeout(timer *t)
 static void
 bgp_initiate(struct bgp_proto *p)
 {
+  int rv = bgp_open(p);
+  if (rv < 0)
+    return;
+
   if (p->startup_delay)
     {
       BGP_TRACE(D_EVENTS, "Startup delayed by %d seconds", p->startup_delay);
@@ -347,6 +351,10 @@ bgp_conn_enter_established_state(struct bgp_conn *conn)
   BGP_TRACE(D_EVENTS, "BGP session established");
   DBG("BGP: UP!!!\n");
 
+  /* For multi-hop BGP sessions */
+  if (ipa_zero(p->source_addr))
+    p->source_addr = conn->sk->saddr; 
+
   p->conn = conn;
   p->last_error_class = 0;
   p->last_error_code = 0;
@@ -669,10 +677,11 @@ bgp_setup_listen_sk(ip_addr addr, unsigned port, u32 flags)
 static void
 bgp_start_neighbor(struct bgp_proto *p)
 {
-  p->local_addr = p->neigh->iface->addr->ip;
-  p->source_addr = ipa_nonzero(p->cf->source_addr) ? p->cf->source_addr : p->local_addr;
+  /* Called only for single-hop BGP sessions */
+
+  if (ipa_zero(p->source_addr))
+    p->source_addr = p->neigh->iface->addr->ip; 
 
-  DBG("BGP: local=%I remote=%I\n", p->source_addr, p->next_hop);
 #ifdef IPV6
   {
     struct ifa *a;
@@ -691,10 +700,6 @@ bgp_start_neighbor(struct bgp_proto *p)
   }
 #endif
 
-  int rv = bgp_open(p);
-  if (rv < 0)
-    return;
-
   bgp_initiate(p);
 }
 
@@ -742,25 +747,22 @@ bgp_start_locked(struct object_lock *lock)
   if (p->p.proto_state != PS_START)
     {
       DBG("BGP: Got lock in different state %d\n", p->p.proto_state);
-    return;
+      return;
     }
 
   DBG("BGP: Got lock\n");
-  p->local_id = proto_get_router_id(&cf->c);
-  p->next_hop = cf->multihop ? cf->multihop_via : cf->remote_ip;
-  p->neigh = neigh_find(&p->p, &p->next_hop, NEF_STICKY);
 
-  if (cf->rr_client)
+  if (cf->multihop)
     {
-      p->rr_cluster_id = cf->rr_cluster_id ? cf->rr_cluster_id : p->local_id;
-      p->rr_client = cf->rr_client;
+      /* Multi-hop sessions do not use neighbor entries */
+      bgp_initiate(p);
+      return;
     }
 
-  p->rs_client = cf->rs_client;
-
-  if (!p->neigh)
+  p->neigh = neigh_find(&p->p, &cf->remote_ip, NEF_STICKY);
+  if (!p->neigh || (p->neigh->scope == SCOPE_HOST))
     {
-      log(L_ERR "%s: Invalid next hop %I", p->p.name, p->next_hop);
+      log(L_ERR "%s: Invalid remote address %I", p->p.name, cf->remote_ip);
       /* As we do not start yet, we can just disable protocol */
       p->p.disabled = 1;
       bgp_store_error(p, NULL, BE_MISC, BEM_INVALID_NEXT_HOP);
@@ -771,7 +773,7 @@ bgp_start_locked(struct object_lock *lock)
   if (p->neigh->iface)
     bgp_start_neighbor(p);
   else
-    BGP_TRACE(D_EVENTS, "Waiting for %I to become my neighbor", p->next_hop);
+    BGP_TRACE(D_EVENTS, "Waiting for %I to become my neighbor", cf->remote_ip);
 }
 
 static int
@@ -796,6 +798,9 @@ bgp_start(struct proto *P)
   p->startup_timer->hook = bgp_startup_timeout;
   p->startup_timer->data = p;
 
+  p->remote_id = 0;
+  p->source_addr = p->cf->source_addr;
+
   /*
    *  Before attempting to create the connection, we need to lock the
    *  port, so that are sure we're the only instance attempting to talk
@@ -869,7 +874,13 @@ bgp_init(struct proto_config *C)
   p->local_as = c->local_as;
   p->remote_as = c->remote_as;
   p->is_internal = (c->local_as == c->remote_as);
+  p->local_id = proto_get_router_id(C);
+  p->rs_client = c->rs_client;
+  p->rr_client = c->rr_client;
+  if (p->rr_client)
+    p->rr_cluster_id = c->rr_cluster_id ? c->rr_cluster_id : p->local_id;
   p->igp_table = get_igp_table(c);
+
   return P;
 }
 
@@ -1046,7 +1057,6 @@ bgp_show_proto_info(struct proto *P)
       cli_msg(-1006, "    Neighbor AS:      %u", p->remote_as);
       cli_msg(-1006, "    Neighbor ID:      %R", p->remote_id);
       cli_msg(-1006, "    Neighbor address: %I", p->cf->remote_ip);
-      cli_msg(-1006, "    Nexthop address:  %I", p->next_hop);
       cli_msg(-1006, "    Source address:   %I", p->source_addr);
       cli_msg(-1006, "    Neighbor caps:   %s%s",
 	      c->peer_refresh_support ? " refresh" : "",
diff --git a/proto/bgp/bgp.h b/proto/bgp/bgp.h
index 76844af..160aa3d 100644
--- a/proto/bgp/bgp.h
+++ b/proto/bgp/bgp.h
@@ -20,7 +20,6 @@ struct bgp_config {
   u32 local_as, remote_as;
   ip_addr remote_ip;
   int multihop;				/* Number of hops if multihop */
-  ip_addr multihop_via;			/* Multihop: address to route to */
   ip_addr source_addr;			/* Source address to use */
   int next_hop_self;			/* Always set next hop to local IP address */
   int missing_lladdr;			/* What we will do when we don' know link-local addr, see MLL_* */
@@ -89,10 +88,8 @@ struct bgp_proto {
   struct bgp_conn outgoing_conn;	/* Outgoing connection we're working with */
   struct bgp_conn incoming_conn;	/* Incoming connection we have neither accepted nor rejected yet */
   struct object_lock *lock;		/* Lock for neighbor connection */
-  ip_addr next_hop;			/* Either the peer or multihop_via */
-  struct neighbor *neigh;		/* Neighbor entry corresponding to next_hop */
-  ip_addr local_addr;			/* Address of the local end of the link to next_hop */
-  ip_addr source_addr;			/* Address used as advertised next hop, usually local_addr */
+  struct neighbor *neigh;		/* Neighbor entry corresponding to remote ip, NULL if multihop */
+  ip_addr source_addr;			/* Local address used as an advertised next hop */
   rtable *igp_table;			/* Table used for recursive next hop lookups */
   struct event *event;			/* Event for respawning and shutting process */
   struct timer *startup_timer;		/* Timer used to delay protocol startup due to previous errors (startup_delay) */
@@ -109,7 +106,7 @@ struct bgp_proto {
 #ifdef IPV6
   byte *mp_reach_start, *mp_unreach_start; /* Multiprotocol BGP attribute notes */
   unsigned mp_reach_len, mp_unreach_len;
-  ip_addr local_link;			/* Link-level version of local_addr */
+  ip_addr local_link;			/* Link-level version of source_addr */
 #endif
 };
 
diff --git a/proto/bgp/config.Y b/proto/bgp/config.Y
index a46431c..e591f89 100644
--- a/proto/bgp/config.Y
+++ b/proto/bgp/config.Y
@@ -54,6 +54,7 @@ bgp_proto:
    bgp_proto_start proto_name '{'
  | bgp_proto proto_item ';'
  | bgp_proto LOCAL AS expr ';' { BGP_CFG->local_as = $4; }
+ | bgp_proto LOCAL ipa AS expr ';' { BGP_CFG->source_addr = $3; BGP_CFG->local_as = $5; }
  | bgp_proto NEIGHBOR ipa AS expr ';' {
      if (ipa_nonzero(BGP_CFG->remote_ip)) cf_error("Only one neighbor per BGP instance is allowed");
 
@@ -67,7 +68,8 @@ bgp_proto:
  | bgp_proto STARTUP HOLD TIME expr ';' { BGP_CFG->initial_hold_time = $5; }
  | bgp_proto CONNECT RETRY TIME expr ';' { BGP_CFG->connect_retry_time = $5; }
  | bgp_proto KEEPALIVE TIME expr ';' { BGP_CFG->keepalive_time = $4; }
- | bgp_proto MULTIHOP expr VIA ipa ';' { BGP_CFG->multihop = $3; BGP_CFG->multihop_via = $5; }
+ | bgp_proto MULTIHOP ';' { BGP_CFG->multihop = 64; }
+ | bgp_proto MULTIHOP expr ';' { BGP_CFG->multihop = $3; }
  | bgp_proto NEXT HOP SELF ';' { BGP_CFG->next_hop_self = 1; }
  | bgp_proto MISSING LLADDR SELF ';' { BGP_CFG->missing_lladdr = MLL_SELF; }
  | bgp_proto MISSING LLADDR DROP ';' { BGP_CFG->missing_lladdr = MLL_DROP; }
diff --git a/proto/bgp/packets.c b/proto/bgp/packets.c
index 514e878..ee2c1b0 100644
--- a/proto/bgp/packets.c
+++ b/proto/bgp/packets.c
@@ -58,7 +58,7 @@ mrt_put_bgp4_hdr(byte *buf, struct bgp_conn *conn, int as4)
       buf+=4;
     }
 
-  put_u16(buf+0, p->neigh->iface->index);
+  put_u16(buf+0, p->neigh ? p->neigh->iface->index : 0);
   put_u16(buf+2, BGP_AF);
   buf+=4;
   buf = ipa_put_addr(buf, conn->sk ? conn->sk->daddr : IPA_NONE);
@@ -402,7 +402,7 @@ bgp_create_update(struct bgp_conn *conn, byte *buf)
 	       */
 
 	      n = neigh_find(&p->p, &ip, 0);
-	      if (n && n->iface == p->neigh->iface)
+	      if (n && p->neigh && n->iface == p->neigh->iface)
 		{
 		  if (second && ipa_nonzero(ipp[1]))
 		    ip_ll = ipp[1];
diff --git a/sysdep/unix/io.c b/sysdep/unix/io.c
index 690c8fc..c1837f0 100644
--- a/sysdep/unix/io.c
+++ b/sysdep/unix/io.c
@@ -1004,6 +1004,11 @@ sk_leave_group(sock *s, ip_addr maddr)
 static void
 sk_tcp_connected(sock *s)
 {
+  sockaddr lsa;
+  int lsa_len = sizeof(lsa);
+  if (getsockname(s->fd, (struct sockaddr *) &lsa, &lsa_len) == 0)
+    get_sockaddr(&lsa, &s->saddr, &s->sport, 1);
+
   s->type = SK_TCP;
   sk_alloc_bufs(s);
   s->tx_hook(s);