BGP redesign

Integrated and extensible BGP with generalized AFI handling,
support for IPv4+IPv6 AFI and unicast+multicast SAFI.

1 files changed, 314 insertions, 105 deletions

diff --git a/proto/bgp/bgp.h b/proto/bgp/bgp.h
index b4067f3a..4ecb86a0 100644
--- a/proto/bgp/bgp.h
+++ b/proto/bgp/bgp.h
@@ -2,6 +2,8 @@
  *	BIRD -- The Border Gateway Protocol
  *
  *	(c) 2000 Martin Mares <mj@ucw.cz>
+ *	(c) 2008--2016 Ondrej Zajicek <santiago@crfreenet.org>
+ *	(c) 2008--2016 CZ.NIC z.s.p.o.
  *
  *	Can be freely distributed and used under the terms of the GNU GPL.
  */
@@ -10,26 +12,66 @@
 #define _BIRD_BGP_H_
 
 #include <stdint.h>
+#include <setjmp.h>
+#include "nest/bird.h"
 #include "nest/route.h"
 #include "nest/bfd.h"
+//#include "lib/lists.h"
 #include "lib/hash.h"
+#include "lib/socket.h"
 
 struct linpool;
 struct eattr;
 
+
+/* Address families */
+
+#define BGP_AFI_IPV4		1
+#define BGP_AFI_IPV6		2
+
+#define BGP_SAFI_UNICAST	1
+#define BGP_SAFI_MULTICAST	2
+
+/* Internal AF codes */
+
+#define BGP_AF(A, B)		(((u32)(A) << 16) | (u32)(B))
+#define BGP_AFI(A)		((u32)(A) >> 16)
+#define BGP_SAFI(A)		((u32)(A) & 0xFFFF)
+
+#define BGP_AF_IPV4		BGP_AF( BGP_AFI_IPV4, BGP_SAFI_UNICAST )
+#define BGP_AF_IPV6		BGP_AF( BGP_AFI_IPV6, BGP_SAFI_UNICAST )
+#define BGP_AF_IPV4_MC		BGP_AF( BGP_AFI_IPV4, BGP_SAFI_MULTICAST )
+#define BGP_AF_IPV6_MC		BGP_AF( BGP_AFI_IPV6, BGP_SAFI_MULTICAST )
+
+
+struct bgp_write_state;
+struct bgp_parse_state;
+struct bgp_export_state;
+struct bgp_bucket;
+
+struct bgp_af_desc {
+  u32 afi;
+  u32 net;
+  const char *name;
+  uint (*encode_nlri)(struct bgp_write_state *s, struct bgp_bucket *buck, byte *buf, uint size);
+  void (*decode_nlri)(struct bgp_parse_state *s, byte *pos, uint len, rta *a);
+  void (*update_next_hop)(struct bgp_export_state *s, eattr *nh, ea_list **to);
+  uint (*encode_next_hop)(struct bgp_write_state *s, eattr *nh, byte *buf, uint size);
+  void (*decode_next_hop)(struct bgp_parse_state *s, byte *pos, uint len, rta *a);
+};
+
+
 struct bgp_config {
   struct proto_config c;
   u32 local_as, remote_as;
+  ip_addr local_ip;			/* Source address to use */
   ip_addr remote_ip;
-  ip_addr source_addr;			/* Source address to use */
   struct iface *iface;			/* Interface for link-local addresses */
+  u16 local_port;			/* Local listening port */
   u16 remote_port; 			/* Neighbor destination port */
   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 strict_bind;			/* Bind listening socket to local address XXXX */
+  int ttl_security;			/* Enable TTL security [RFC 5082] */
   int compare_path_lengths;		/* Use path lengths when selecting best route */
   int med_metric;			/* Compare MULTI_EXIT_DISC even between routes from differen ASes */
   int igp_metric;			/* Use IGP metrics when selecting best route */
@@ -37,18 +79,17 @@ struct bgp_config {
   int deterministic_med;		/* Use more complicated algo to have strict RFC 4271 MED comparison */
   u32 default_local_pref;		/* Default value for LOCAL_PREF attribute */
   u32 default_med;			/* Default value for MULTI_EXIT_DISC attribute */
-  int capabilities;			/* Enable capability handshake [RFC3392] */
-  int enable_refresh;			/* Enable local support for route refresh [RFC2918] */
-  int enable_as4;			/* Enable local support for 4B AS numbers [RFC4893] */
+  int capabilities;			/* Enable capability handshake [RFC 3392] */
+  int enable_refresh;			/* Enable local support for route refresh [RFC 2918] */
+  int enable_as4;			/* Enable local support for 4B AS numbers [RFC 4893] */
   int enable_extended_messages;		/* Enable local support for extended messages [draft] */
   u32 rr_cluster_id;			/* Route reflector cluster ID, if different from local ID */
   int rr_client;			/* Whether neighbor is RR client of me */
   int rs_client;			/* Whether neighbor is RS client of me */
-  int advertise_ipv4;			/* Whether we should add IPv4 capability advertisement to OPEN message */
+  u32 confederation;			/* Confederation ID, or zero if confeds not active */
+  int confederation_member;		/* Whether neighbor AS is member of our confederation */
   int passive;				/* Do not initiate outgoing connection */
   int interpret_communities;		/* Hardwired handling of well-known communities */
-  int secondary;			/* Accept also non-best routes (i.e. RA_ACCEPTED) */
-  int add_path;				/* Use ADD-PATH extension [draft] */
   int allow_local_as;			/* Allow that number of local ASNs in incoming AS_PATHs */
   int gr_mode;				/* Graceful restart mode (BGP_GR_*) */
   int setkey;				/* Set MD5 password to system SA/SP database */
@@ -63,11 +104,27 @@ struct bgp_config {
   unsigned disable_after_error;		/* Disable the protocol when error is detected */
 
   char *password;			/* Password used for MD5 authentication */
-  struct rtable_config *igp_table;	/* Table used for recursive next hop lookups */
   int check_link;			/* Use iface link state for liveness detection */
   int bfd;				/* Use BFD for liveness detection */
 };
 
+struct bgp_channel_config {
+  struct channel_config c;
+
+  u32 afi;
+
+  ip_addr next_hop_addr;		/* Local address for NEXT_HOP attribute */
+  u8 next_hop_self;			/* Always set next hop to local IP address */
+  u8 next_hop_keep;			/* Do not touch next hop attribute */
+  u8 missing_lladdr;			/* What we will do when we don' know link-local addr, see MLL_* */
+  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) */
+  u8 gr_able;				/* Allow full graceful restart for the channel */
+  u8 add_path;				/* Use ADD-PATH extension [RFC 7911] */
+
+  struct rtable_config *igp_table;	/* Table used for recursive next hop lookups */
+};
+
 #define MLL_SELF 1
 #define MLL_DROP 2
 #define MLL_IGNORE 3
@@ -75,112 +132,226 @@ struct bgp_config {
 #define GW_DIRECT 1
 #define GW_RECURSIVE 2
 
-#define ADD_PATH_RX 1
-#define ADD_PATH_TX 2
-#define ADD_PATH_FULL 3
+#define BGP_ADD_PATH_RX		1
+#define BGP_ADD_PATH_TX		2
+#define BGP_ADD_PATH_FULL	3
 
-#define BGP_GR_ABLE 1
-#define BGP_GR_AWARE 2
+#define BGP_GR_ABLE		1
+#define BGP_GR_AWARE		2
 
-/* For peer_gr_flags */
+/* For GR capability common flags */
 #define BGP_GRF_RESTART 0x80
 
-/* For peer_gr_aflags */
+/* For GR capability per-AF flags */
 #define BGP_GRF_FORWARDING 0x80
 
 
+struct bgp_af_caps {
+  u32 afi;
+  u8 ready;				/* Multiprotocol capability, RFC 4760 */
+  u8 gr_able;				/* Graceful restart support, RFC 4724 */
+  u8 gr_af_flags;			/* Graceful restart per-AF flags */
+  u8 add_path;				/* Multiple paths support,   RFC 7911 */
+};
+
+struct bgp_caps {
+  u32 as4_number;			/* Announced ASN */
+
+  u8 as4_support;			/* Four-octet AS capability, RFC 6793 */
+  u8 ext_messages;			/* Extended message length,  RFC draft */
+  u8 route_refresh;			/* Route refresh capability, RFC 2918 */
+  u8 enhanced_refresh;			/* Enhanced route refresh,   RFC 7313 */
+
+  u8 gr_aware;				/* Graceful restart capability, RFC 4724 */
+  u8 gr_flags;				/* Graceful restart flags */
+  u16 gr_time;				/* Graceful restart time in seconds */
+
+  u16 af_count;				/* Number of af_data items */
+
+  struct bgp_af_caps af_data[0];	/* Per-AF capability data */
+};
+
+struct bgp_socket {
+  node n;				/* Node in global bgp_sockets */
+  sock *sk;				/* Real listening socket */
+  u32 uc;				/* Use count */
+};
+
 struct bgp_conn {
   struct bgp_proto *bgp;
   struct birdsock *sk;
-  uint state;				/* State of connection state machine */
-  struct timer *connect_retry_timer;
+  u8 state;				/* State of connection state machine */
+  u8 as4_session;			/* Session uses 4B AS numbers in AS_PATH (both sides support it) */
+  u8 ext_messages;			/* Session uses extended message length */
+
+  struct bgp_caps *local_caps;
+  struct bgp_caps *remote_caps;
+  struct timer *connect_timer;
   struct timer *hold_timer;
   struct timer *keepalive_timer;
   struct event *tx_ev;
-  int packets_to_send;			/* Bitmap of packet types to be sent */
+  u32 packets_to_send;			/* Bitmap of packet types to be sent */
+  u32 channels_to_send;			/* Bitmap of channels with packets to be sent */
+  u8 last_channel;			/* Channel used last time for TX */
+  u8 last_channel_count;		/* Number of times the last channel was used in succession */
   int notify_code, notify_subcode, notify_size;
   byte *notify_data;
-  u32 advertised_as;			/* Temporary value for AS number received */
-  int start_state;			/* protocol start_state snapshot when connection established */
-  u8 peer_refresh_support;		/* Peer supports route refresh [RFC2918] */
-  u8 peer_as4_support;			/* Peer supports 4B AS numbers [RFC4893] */
-  u8 peer_add_path;			/* Peer supports ADD-PATH [draft] */
-  u8 peer_enhanced_refresh_support;	/* Peer supports enhanced refresh [RFC7313] */
-  u8 peer_gr_aware;
-  u8 peer_gr_able;
-  u16 peer_gr_time;
-  u8 peer_gr_flags;
-  u8 peer_gr_aflags;
-  u8 peer_ext_messages_support;		/* Peer supports extended message length [draft] */
-  unsigned hold_time, keepalive_time;	/* Times calculated from my and neighbor's requirements */
+
+  uint hold_time, keepalive_time;	/* Times calculated from my and neighbor's requirements */
 };
 
 struct bgp_proto {
   struct proto p;
   struct bgp_config *cf;		/* Shortcut to BGP configuration */
   u32 local_as, remote_as;
-  int start_state;			/* Substates that partitions BS_START */
-  u8 is_internal;			/* Internal BGP connection (local_as == remote_as) */
-  u8 as4_session;			/* Session uses 4B AS numbers in AS_PATH (both sides support it) */
-  u8 add_path_rx;			/* Session expects receive of ADD-PATH extended NLRI */
-  u8 add_path_tx;			/* Session expects transmit of ADD-PATH extended NLRI */
-  u8 ext_messages;			/* Session allows to use extended messages (both sides support it) */
+  u32 public_as;			/* Externally visible ASN (local_as or confederation id) */
   u32 local_id;				/* BGP identifier of this router */
   u32 remote_id;			/* BGP identifier of the neighbor */
   u32 rr_cluster_id;			/* Route reflector cluster ID */
-  int rr_client;			/* Whether neighbor is RR client of me */
-  int rs_client;			/* Whether neighbor is RS client of me */
+  int start_state;			/* Substates that partitions BS_START */
+  u8 is_internal;			/* Internal BGP session (local_as == remote_as) */
+  u8 is_interior;			/* Internal or intra-confederation BGP session */
+  u8 as4_session;			/* Session uses 4B AS numbers in AS_PATH (both sides support it) */
+  u8 rr_client;				/* Whether neighbor is RR client of me */
+  u8 rs_client;				/* Whether neighbor is RS client of me */
+  u8 route_refresh;			/* Route refresh allowed to send [RFC 2918] */
+  u8 enhanced_refresh;			/* Enhanced refresh is negotiated [RFC 7313] */
   u8 gr_ready;				/* Neighbor could do graceful restart */
-  u8 gr_active;				/* Neighbor is doing graceful restart */
-  u8 feed_state;			/* Feed state (TX) for EoR, RR packets, see BFS_* */
-  u8 load_state;			/* Load state (RX) for EoR, RR packets, see BFS_* */
+  u8 gr_active_num;			/* Neighbor is doing GR, number of active channels */
+  u8 channel_count;			/* Number of active channels */
+  u32 *afi_map;				/* Map channel index -> AFI */
+  struct bgp_channel **channel_map;	/* Map channel index -> channel */
   struct bgp_conn *conn;		/* Connection we have established */
   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 */
   struct neighbor *neigh;		/* Neighbor entry corresponding to remote ip, NULL if multihop */
+  struct bgp_socket *sock;		/* Shared listening socket */
   struct bfd_request *bfd_req;		/* BFD request, if BFD is used */
   ip_addr source_addr;			/* Local address used as an advertised next hop */
-  rtable *igp_table;			/* Table used for recursive next hop lookups */
+  ip_addr link_addr;			/* Link-local version of source_addr */
   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) */
   struct timer *gr_timer;		/* Timer waiting for reestablishment after graceful restart */
-  struct bgp_bucket **bucket_hash;	/* Hash table of attribute buckets */
-  uint hash_size, hash_count, hash_limit;
-  HASH(struct bgp_prefix) prefix_hash;	/* Prefixes to be sent */
-  slab *prefix_slab;			/* Slab holding prefix nodes */
-  list bucket_queue;			/* Queue of buckets to send */
-  struct bgp_bucket *withdraw_bucket;	/* Withdrawn routes */
   unsigned startup_delay;		/* Time to delay protocol startup by due to errors */
   bird_clock_t last_proto_error;	/* Time of last error that leads to protocol stop */
   u8 last_error_class; 			/* Error class of last error */
   u32 last_error_code;			/* Error code of last error. BGP protocol errors
 					   are encoded as (bgp_err_code << 16 | bgp_err_subcode) */
-#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 source_addr */
-#endif
+};
+
+struct bgp_channel {
+  struct channel c;
+
+  /* Rest are BGP specific data */
+  struct bgp_channel_config *cf;
+  pool *pool; /* XXXX */
+
+  u32 afi;
+  u32 index;
+  const struct bgp_af_desc *desc;
+
+  HASH(struct bgp_bucket) bucket_hash;	/* Hash table of route buckets */
+  struct bgp_bucket *withdraw_bucket;	/* Withdrawn routes */
+  list bucket_queue;			/* Queue of buckets to send (struct bgp_bucket) */
+
+  HASH(struct bgp_prefix) prefix_hash;	/* Prefixes to be sent */
+  slab *prefix_slab;			/* Slab holding prefix nodes */
+
+  rtable *igp_table;			/* Table used for recursive next hop lookups */
+  ip_addr next_hop_addr;		/* Local address for NEXT_HOP attribute */
+  ip_addr link_addr;			/* Link-local version of next_hop_addr */
+
+  u32 packets_to_send;			/* Bitmap of packet types to be sent */
+
+  u8 gr_ready;				/* Neighbor could do GR on this AF */
+  u8 gr_active;				/* Neighbor is doing GR and keeping fwd state */
+
+  u8 add_path_rx;			/* Session expects receive of ADD-PATH extended NLRI */
+  u8 add_path_tx;			/* Session expects transmit of ADD-PATH extended NLRI */
+
+  u8 feed_state;			/* Feed state (TX) for EoR, RR packets, see BFS_* */
+  u8 load_state;			/* Load state (RX) for EoR, RR packets, see BFS_* */
 };
 
 struct bgp_prefix {
-  struct {
-    ip_addr prefix;
-    int pxlen;
-  } n;
+  node buck_node;			/* Node in per-bucket list */
+  struct bgp_prefix *next;		/* Node in prefix hash table */
+  u32 hash;
   u32 path_id;
-  struct bgp_prefix *next;
-  node bucket_node;			/* Node in per-bucket list */
+  net_addr net[0];
 };
 
 struct bgp_bucket {
   node send_node;			/* Node in send queue */
-  struct bgp_bucket *hash_next, *hash_prev;	/* Node in bucket hash table */
-  unsigned hash;			/* Hash over extended attributes */
-  list prefixes;			/* Prefixes in this buckets */
+  struct bgp_bucket *next;		/* Node in bucket hash table */
+  list prefixes;			/* Prefixes in this bucket (struct bgp_prefix) */
+  u32 hash;				/* Hash over extended attributes */
   ea_list eattrs[0];			/* Per-bucket extended attributes */
 };
 
+struct bgp_export_state {
+  struct bgp_proto *proto;
+  struct bgp_channel *channel;
+  struct linpool *pool;
+
+  struct bgp_proto *src;
+  rte *route;
+
+  u32 attrs_seen[1];
+  uint err_withdraw;
+};
+
+struct bgp_write_state {
+  struct bgp_proto *proto;
+  struct bgp_channel *channel;
+  struct linpool *pool;
+
+  int as4_session;
+  int add_path;
+
+  eattr *mp_next_hop;
+};
+
+struct bgp_parse_state {
+  struct bgp_proto *proto;
+  struct bgp_channel *channel;
+  struct linpool *pool;
+
+  int as4_session;
+  int add_path;
+
+  u32 attrs_seen[256/32];
+
+  u32 mp_reach_af;
+  u32 mp_unreach_af;
+
+  uint attr_len;
+  uint ip_reach_len;
+  uint ip_unreach_len;
+  uint ip_next_hop_len;
+  uint mp_reach_len;
+  uint mp_unreach_len;
+  uint mp_next_hop_len;
+
+  byte *attrs;
+  byte *ip_reach_nlri;
+  byte *ip_unreach_nlri;
+  byte *ip_next_hop_data;
+  byte *mp_reach_nlri;
+  byte *mp_unreach_nlri;
+  byte *mp_next_hop_data;
+
+  uint err_withdraw;
+  uint err_subcode;
+  jmp_buf err_jmpbuf;
+
+  /* Cached state for bgp_rte_update() */
+  u32 last_id;
+  struct rte_src *last_src;
+  rta *cached_rta;
+};
+
 #define BGP_PORT		179
 #define BGP_VERSION		4
 #define BGP_HEADER_LENGTH	19
@@ -191,8 +362,21 @@ struct bgp_bucket {
 #define BGP_RX_BUFFER_EXT_SIZE	65535
 #define BGP_TX_BUFFER_EXT_SIZE	65535
 
-static inline uint bgp_max_packet_length(struct bgp_proto *p)
-{ return p->ext_messages ? BGP_MAX_EXT_MSG_LENGTH : BGP_MAX_MESSAGE_LENGTH; }
+static inline int bgp_channel_is_ipv4(struct bgp_channel *c)
+{ return BGP_AFI(c->afi) == BGP_AFI_IPV4; }
+
+static inline int bgp_channel_is_ipv6(struct bgp_channel *c)
+{ return BGP_AFI(c->afi) == BGP_AFI_IPV6; }
+
+static inline uint bgp_max_packet_length(struct bgp_conn *conn)
+{ return conn->ext_messages ? BGP_MAX_EXT_MSG_LENGTH : BGP_MAX_MESSAGE_LENGTH; }
+
+static inline void
+bgp_parse_error(struct bgp_parse_state *s, uint subcode)
+{
+  s->err_subcode = subcode;
+  longjmp(s->err_jmpbuf, 1);
+}
 
 extern struct linpool *bgp_linpool;
 
@@ -207,9 +391,9 @@ void bgp_conn_enter_established_state(struct bgp_conn *conn);
 void bgp_conn_enter_close_state(struct bgp_conn *conn);
 void bgp_conn_enter_idle_state(struct bgp_conn *conn);
 void bgp_handle_graceful_restart(struct bgp_proto *p);
-void bgp_graceful_restart_done(struct bgp_proto *p);
-void bgp_refresh_begin(struct bgp_proto *p);
-void bgp_refresh_end(struct bgp_proto *p);
+void bgp_graceful_restart_done(struct bgp_channel *c);
+void bgp_refresh_begin(struct bgp_channel *c);
+void bgp_refresh_end(struct bgp_channel *c);
 void bgp_store_error(struct bgp_proto *p, struct bgp_conn *c, u8 class, u32 code);
 void bgp_stop(struct bgp_proto *p, unsigned subcode);
 
@@ -232,46 +416,83 @@ struct rte_source *bgp_get_source(struct bgp_proto *p, u32 path_id);
 
 /* attrs.c */
 
+static inline eattr *
+bgp_find_attr(ea_list *attrs, uint code)
+{
+  return ea_find(attrs, EA_CODE(EAP_BGP, code));
+}
+
+eattr *
+bgp_set_attr(ea_list **attrs, struct linpool *pool, uint code, uint flags, uintptr_t val);
+
+static inline void
+bgp_set_attr_u32(ea_list **to, struct linpool *pool, uint code, uint flags, u32 val)
+{ bgp_set_attr(to, pool, code, flags, (uintptr_t) val); }
+
+static inline void
+bgp_set_attr_ptr(ea_list **to, struct linpool *pool, uint code, uint flags, struct adata *val)
+{ bgp_set_attr(to, pool, code, flags, (uintptr_t) val); }
+
+static inline void
+bgp_set_attr_data(ea_list **to, struct linpool *pool, uint code, uint flags, void *data, uint len)
+{
+  struct adata *a = lp_alloc_adata(pool, len);
+  memcpy(a->data, data, len);
+  bgp_set_attr(to, pool, code, flags, (uintptr_t) a);
+}
+
+static inline void
+bgp_unset_attr(ea_list **to, struct linpool *pool, uint code)
+{ eattr *e = bgp_set_attr(to, pool, code, 0, 0); e->type = EAF_TYPE_UNDEF; }
+
+
+
+
 /* Hack: although BA_NEXT_HOP attribute has type EAF_TYPE_IP_ADDRESS, in IPv6
  * we store two addesses in it - a global address and a link local address.
  */
-#ifdef IPV6
+#ifdef XXX
 #define NEXT_HOP_LENGTH (2*sizeof(ip_addr))
 static inline void set_next_hop(byte *b, ip_addr addr) { ((ip_addr *) b)[0] = addr; ((ip_addr *) b)[1] = IPA_NONE; }
-#else
 #define NEXT_HOP_LENGTH sizeof(ip_addr)
 static inline void set_next_hop(byte *b, ip_addr addr) { ((ip_addr *) b)[0] = addr; }
 #endif
 
-void bgp_attach_attr(struct ea_list **to, struct linpool *pool, unsigned attr, uintptr_t val);
-byte *bgp_attach_attr_wa(struct ea_list **to, struct linpool *pool, unsigned attr, unsigned len);
-struct rta *bgp_decode_attrs(struct bgp_conn *conn, byte *a, uint len, struct linpool *pool, int mandatory);
-int bgp_get_attr(struct eattr *e, byte *buf, int buflen);
+int bgp_encode_attrs(struct bgp_write_state *s, ea_list *attrs, byte *buf, byte *end);
+ea_list * bgp_decode_attrs(struct bgp_parse_state *s, byte *data, uint len);
+
+void bgp_init_bucket_table(struct bgp_channel *c);
+void bgp_free_bucket(struct bgp_channel *c, struct bgp_bucket *b);
+void bgp_defer_bucket(struct bgp_channel *c, struct bgp_bucket *b);
+void bgp_withdraw_bucket(struct bgp_channel *c, struct bgp_bucket *b);
+
+void bgp_init_prefix_table(struct bgp_channel *c);
+void bgp_free_prefix(struct bgp_channel *c, struct bgp_prefix *bp);
+
 int bgp_rte_better(struct rte *, struct rte *);
 int bgp_rte_mergable(rte *pri, rte *sec);
 int bgp_rte_recalculate(rtable *table, net *net, rte *new, rte *old, rte *old_best);
-void bgp_rt_notify(struct proto *P, rtable *tbl UNUSED, net *n, rte *new, rte *old UNUSED, ea_list *attrs);
+void bgp_rt_notify(struct proto *P, struct channel *C, net *n, rte *new, rte *old, ea_list *attrs);
 int bgp_import_control(struct proto *, struct rte **, struct ea_list **, struct linpool *);
-void bgp_init_bucket_table(struct bgp_proto *);
-void bgp_free_bucket(struct bgp_proto *p, struct bgp_bucket *buck);
-void bgp_init_prefix_table(struct bgp_proto *p, u32 order);
-void bgp_free_prefix(struct bgp_proto *p, struct bgp_prefix *bp);
-uint bgp_encode_attrs(struct bgp_proto *p, byte *w, ea_list *attrs, int remains);
+int bgp_get_attr(struct eattr *e, byte *buf, int buflen);
 void bgp_get_route_info(struct rte *, byte *buf, struct ea_list *attrs);
 
-inline static void bgp_attach_attr_ip(struct ea_list **to, struct linpool *pool, unsigned attr, ip_addr a)
-{ *(ip_addr *) bgp_attach_attr_wa(to, pool, attr, sizeof(ip_addr)) = a; }
 
 /* packets.c */
 
 void mrt_dump_bgp_state_change(struct bgp_conn *conn, unsigned old, unsigned new);
-void bgp_schedule_packet(struct bgp_conn *conn, int type);
+const struct bgp_af_desc *bgp_get_af_desc(u32 afi);
+const struct bgp_af_caps *bgp_find_af_caps(struct bgp_caps *caps, u32 afi);
+void bgp_schedule_packet(struct bgp_conn *conn, struct bgp_channel *c, int type);
 void bgp_kick_tx(void *vconn);
 void bgp_tx(struct birdsock *sk);
 int bgp_rx(struct birdsock *sk, uint size);
 const char * bgp_error_dsc(unsigned code, unsigned subcode);
 void bgp_log_error(struct bgp_proto *p, u8 class, char *msg, unsigned code, unsigned subcode, byte *data, unsigned len);
 
+void bgp_update_next_hop(struct bgp_export_state *s, eattr *a, ea_list **to);
+
+
 /* Packet types */
 
 #define PKT_OPEN		0x01
@@ -305,9 +526,9 @@ void bgp_log_error(struct bgp_proto *p, u8 class, char *msg, unsigned code, unsi
 #define BA_RCID_PATH		0x0d
 #define BA_MP_REACH_NLRI	0x0e	/* [RFC2283] */
 #define BA_MP_UNREACH_NLRI	0x0f
-#define BA_EXT_COMMUNITY	0x10	/* [RFC4360] */
-#define BA_AS4_PATH             0x11    /* [RFC4893] */
-#define BA_AS4_AGGREGATOR       0x12
+#define BA_EXT_COMMUNITY	0x10	/* RFC 4360 */
+#define BA_AS4_PATH             0x11	/* RFC 6793 */
+#define BA_AS4_AGGREGATOR       0x12	/* RFC 6793 */
 #define BA_LARGE_COMMUNITY	0x20	/* [draft-ietf-idr-large-community] */
 
 /* BGP connection states */
@@ -328,14 +549,12 @@ void bgp_log_error(struct bgp_proto *p, u8 class, char *msg, unsigned code, unsi
  *
  * When BGP protocol is started by core, it goes to BSS_PREPARE. When BGP
  * protocol done what is neccessary to start itself (like acquiring the lock),
- * it goes to BSS_CONNECT.  When some connection attempt failed because of
- * option or capability error, it goes to BSS_CONNECT_NOCAP.
+ * it goes to BSS_CONNECT.
  */
 
 #define BSS_PREPARE		0	/* Used before ordinary BGP started, i. e. waiting for lock */
 #define BSS_DELAY		1	/* Startup delay due to previous errors */
 #define BSS_CONNECT		2	/* Ordinary BGP connecting */
-#define BSS_CONNECT_NOCAP	3	/* Legacy BGP connecting (without capabilities) */
 
 
 /* BGP feed states (TX)
@@ -344,7 +563,7 @@ void bgp_log_error(struct bgp_proto *p, u8 class, char *msg, unsigned code, unsi
  *
  * RFC 7313 specifies that a route refresh should be demarcated by BoRR and EoRR packets.
  *
- * These states (stored in p->feed_state) are used to keep track of these
+ * These states (stored in c->feed_state) are used to keep track of these
  * requirements. When such feed is started, BFS_LOADING / BFS_REFRESHING is
  * set. When it ended, BFS_LOADED / BFS_REFRESHED is set to schedule End-of-RIB
  * or EoRR packet. When the packet is sent, the state returned to BFS_NONE.
@@ -400,15 +619,5 @@ void bgp_log_error(struct bgp_proto *p, u8 class, char *msg, unsigned code, unsi
 #define ORIGIN_EGP		1
 #define ORIGIN_INCOMPLETE	2
 
-/* Address families */
-
-#define BGP_AF_IPV4		1
-#define BGP_AF_IPV6		2
-
-#ifdef IPV6
-#define BGP_AF BGP_AF_IPV6
-#else
-#define BGP_AF BGP_AF_IPV4
-#endif
 
 #endif