From f4a94644d0d6a4ecdf5fcebb0062bc463fad2a28 Mon Sep 17 00:00:00 2001
From: Ondrej Zajicek <santiago@crfreenet.org>
Date: Mon, 24 Mar 2025 17:09:25 +0100
Subject: [PATCH] BGP: Add option to specify format of link-local next hop

When a BGP session is established using link-local next hop addresses,
there is no global IPv6 address for next hop. Implementations differ on
how to encode such next hop. This leads to interoperability problems.

Add the option 'link local next hop format' to specify which format to
use when encoding such next hops.

Based on a patch from Andrey V. Elsukov and Georgy Kirichenko, submitted
by Aleksandr Stepanov, thanks!
---
 doc/bird.sgml       | 13 +++++++++++++
 lib/alloca.h        |  2 ++
 proto/bgp/bgp.c     |  1 +
 proto/bgp/bgp.h     |  6 ++++++
 proto/bgp/config.Y  | 13 +++++++++++--
 proto/bgp/packets.c | 46 +++++++++++++++++++++++++++++++++++++++++++++
 6 files changed, 79 insertions(+), 2 deletions(-)

diff --git a/doc/bird.sgml b/doc/bird.sgml
index 4efca00dd..87317f266 100644
--- a/doc/bird.sgml
+++ b/doc/bird.sgml
@@ -3604,6 +3604,18 @@ be used in explicit configuration.
 	enabled a route may end with a link-local immediate next hop when the
 	IGP route has one. Default: disabled.
 
+	<tag><label id="bgp-link-local-next-hop-format">link local next hop format native|single|double</tag>
+	For IPv6 routes, BGP assumes that the Next Hop attribute contains a
+	global IPv6 address (in the first position) and an optional link-local
+	IPv6 address (in the second position): [<m/global/, <m/link-local/].
+	When a BGP session is established using just link-local addresses, there
+	may be no global IPv6 address for the next hop. BGP implementations
+	differ on how to encode such next hops. BIRD <cf/native/ format is to
+	send [zero, <m/link-local/], <cf/single/ format is [<m/link-local/],
+	<cf/double/ format is [<m/link-local/, <m/link-local/]. BIRD accepts all
+	these variants when decoding received routes, but this option controls
+	which one it uses to encode such next hops. Default: native.
+
 	<tag><label id="bgp-gateway">gateway direct|recursive</tag>
 	For received routes, their <cf/gw/ (immediate next hop) attribute is
 	computed from received <cf/bgp_next_hop/ attribute. This option
@@ -3827,6 +3839,7 @@ direction (re-export of routes to the BGP neighbor):
 	<item><cf/next hop address/
 	<item><cf/next hop self/
 	<item><cf/next hop keep/
+	<item><cf/link local next hop format/
 	<item><cf/aigp/
 	<item><cf/aigp originate/
 </itemize>
diff --git a/lib/alloca.h b/lib/alloca.h
index e5557cdbc..3491d5a6a 100644
--- a/lib/alloca.h
+++ b/lib/alloca.h
@@ -17,4 +17,6 @@
 
 #define allocz(len) ({ void *_x = alloca(len); memset(_x, 0, len); _x; })
 
+#define alloca_copy(src,len) ({ void *_x = alloca(len); memcpy(_x, src, len); })
+
 #endif
diff --git a/proto/bgp/bgp.c b/proto/bgp/bgp.c
index 7ae03a06a..b5327a693 100644
--- a/proto/bgp/bgp.c
+++ b/proto/bgp/bgp.c
@@ -2324,6 +2324,7 @@ bgp_channel_reconfigure(struct channel *C, struct channel_config *CC, int *impor
   if (!ipa_equal(new->next_hop_addr, old->next_hop_addr) ||
       (new->next_hop_self != old->next_hop_self) ||
       (new->next_hop_keep != old->next_hop_keep) ||
+      (new->llnh_format != old->llnh_format) ||
       (new->aigp != old->aigp) ||
       (new->aigp_originate != old->aigp_originate))
     *export_changed = 1;
diff --git a/proto/bgp/bgp.h b/proto/bgp/bgp.h
index 1585f6800..5ec7a9f94 100644
--- a/proto/bgp/bgp.h
+++ b/proto/bgp/bgp.h
@@ -164,6 +164,7 @@ struct bgp_channel_config {
   u8 next_hop_self;			/* Always set next hop to local IP address (NH_*) */
   u8 next_hop_keep;			/* Do not modify next hop attribute (NH_*) */
   u8 next_hop_prefer;			/* Prefer global or link-local next hop (NHP_*) */
+  u8 llnh_format;			/* Requested link-local next hop format (LLNH_*) */
   u8 mandatory;				/* Channel is mandatory in capability negotiation */
   u8 gw_mode;				/* How we compute route gateway from next_hop attr, see GW_* */
   u8 secondary;				/* Accept also non-best routes (i.e. RA_ACCEPTED) */
@@ -207,6 +208,10 @@ struct bgp_channel_config {
 #define MLL_DROP		2
 #define MLL_IGNORE		3
 
+#define LLNH_NATIVE		0
+#define LLNH_SINGLE		1
+#define LLNH_DOUBLE		2
+
 #define GW_DIRECT		1
 #define GW_RECURSIVE		2
 
@@ -458,6 +463,7 @@ struct bgp_write_state {
   int mp_reach;
   int as4_session;
   int add_path;
+  int llnh_format;
   int mpls;
   int sham;
   int ignore_non_bgp_attrs;
diff --git a/proto/bgp/config.Y b/proto/bgp/config.Y
index 222879e94..11003d4e9 100644
--- a/proto/bgp/config.Y
+++ b/proto/bgp/config.Y
@@ -32,9 +32,10 @@ CF_KEYWORDS(BGP, LOCAL, NEIGHBOR, AS, HOLD, TIME, CONNECT, RETRY, KEEPALIVE,
 	LIVED, STALE, IMPORT, IBGP, EBGP, MANDATORY, INTERNAL, EXTERNAL, SETS,
 	DYNAMIC, RANGE, NAME, DIGITS, BGP_AIGP, AIGP, ORIGINATE, COST, ENFORCE,
 	FIRST, FREE, VALIDATE, BASE, ROLE, ROLES, PEER, PROVIDER, CUSTOMER,
-	RS_SERVER, RS_CLIENT, REQUIRE, BGP_OTC, GLOBAL, SEND, MIN, MAX)
+	RS_SERVER, RS_CLIENT, REQUIRE, BGP_OTC, GLOBAL, SEND, MIN, MAX,
+	FORMAT, NATIVE, SINGLE, DOUBLE)
 
-%type <i> bgp_nh
+%type <i> bgp_nh bgp_llnh
 %type <i32> bgp_afi
 
 CF_KEYWORDS(CEASE, PREFIX, LIMIT, HIT, ADMINISTRATIVE, SHUTDOWN, RESET, PEER,
@@ -281,6 +282,13 @@ bgp_nh:
    bool { $$ = $1; }
  | IBGP { $$ = NH_IBGP; }
  | EBGP { $$ = NH_EBGP; }
+ ;
+
+bgp_llnh:
+   NATIVE { $$ = LLNH_NATIVE; }
+ | SINGLE { $$ = LLNH_SINGLE; }
+ | DOUBLE { $$ = LLNH_DOUBLE; }
+ ;
 
 bgp_lladdr: SELF | DROP | IGNORE;
 
@@ -290,6 +298,7 @@ bgp_channel_item:
  | NEXT HOP SELF bgp_nh { BGP_CC->next_hop_self = $4; }
  | NEXT HOP KEEP bgp_nh { BGP_CC->next_hop_keep = $4; }
  | NEXT HOP PREFER GLOBAL { BGP_CC->next_hop_prefer = NHP_GLOBAL; }
+ | LINK LOCAL NEXT HOP FORMAT bgp_llnh { BGP_CC->llnh_format = $6; }
  | MANDATORY bool { BGP_CC->mandatory = $2; }
  | MISSING LLADDR bgp_lladdr { cf_warn("%s.%s: Missing lladdr option is deprecated and ignored, remove it", this_proto->name, this_channel->name); }
  | GATEWAY DIRECT { BGP_CC->gw_mode = GW_DIRECT; }
diff --git a/proto/bgp/packets.c b/proto/bgp/packets.c
index 4ee7c5a5f..1c846cd50 100644
--- a/proto/bgp/packets.c
+++ b/proto/bgp/packets.c
@@ -1313,6 +1313,37 @@ bgp_update_next_hop_ip(struct bgp_export_state *s, eattr *a, ea_list **to)
     REJECT(NO_LABEL_STACK);
 }
 
+static uint
+bgp_prepare_link_local_next_hop(struct bgp_write_state *s, ip_addr *nh)
+{
+  /*
+   * We have link-local next-hop in nh[1]
+   *
+   * Possible variants:
+   * [ ::, fe80::XX ] - BIRD's default/internal (LLNH_NATIVE)
+   * [ fe80::XX ] - draft-white-linklocal-capability (LLNH_SINGLE)
+   * [ fe80::XX, fe80::XX ] - Used in JunoOS (LLNH_DOUBLE)
+   */
+
+  switch (s->llnh_format)
+  {
+  case LLNH_NATIVE:
+    return 32;
+
+  case LLNH_SINGLE:
+    nh[0] = nh[1];
+    return 16;
+
+  case LLNH_DOUBLE:
+    nh[0] = nh[1];
+    return 32;
+
+  default:
+    ASSERT(0);
+    return 32;
+  }
+}
+
 static uint
 bgp_encode_next_hop_ip(struct bgp_write_state *s, eattr *a, byte *buf, uint size UNUSED)
 {
@@ -1335,6 +1366,13 @@ bgp_encode_next_hop_ip(struct bgp_write_state *s, eattr *a, byte *buf, uint size
     return 4;
   }
 
+  /* Handle variants of link-local-only next hop */
+  if (ipa_zero(nh[0]) && (len == 32))
+  {
+    nh = alloca_copy(nh, 32);
+    len = bgp_prepare_link_local_next_hop(s, nh);
+  }
+
   put_ip6(buf, ipa_to_ip6(nh[0]));
 
   if (len == 32)
@@ -1411,6 +1449,13 @@ bgp_encode_next_hop_vpn(struct bgp_write_state *s, eattr *a, byte *buf, uint siz
     return 12;
   }
 
+  /* Handle variants of link-local-only next hop */
+  if (ipa_zero(nh[0]) && (len == 32))
+  {
+    nh = alloca_copy(nh, 32);
+    len = bgp_prepare_link_local_next_hop(s, nh);
+  }
+
   put_u64(buf, 0); /* VPN RD is 0 */
   put_ip6(buf+8, ipa_to_ip6(nh[0]));
 
@@ -2577,6 +2622,7 @@ again: ;
     .mp_reach = (c->afi != BGP_AF_IPV4) || c->ext_next_hop,
     .as4_session = p->as4_session,
     .add_path = c->add_path_tx,
+    .llnh_format = c->cf->llnh_format,
     .mpls = c->desc->mpls,
   };
 
-- 
2.47.2