1 /*
2 * BIRD -- The Border Gateway Protocol
3 *
4 * (c) 2000 Martin Mares <mj@ucw.cz>
5 *
6 * Can be freely distributed and used under the terms of the GNU GPL.
7 */
8
9 #ifndef _BIRD_BGP_H_
10 #define _BIRD_BGP_H_
11
12 #include <stdint.h>
13 #include "nest/route.h"
14 #include "nest/bfd.h"
15 #include "lib/hash.h"
16
17 struct linpool;
18 struct eattr;
19
20 struct bgp_config {
21 struct proto_config c;
22 u32 local_as, remote_as;
23 ip_addr remote_ip;
24 ip_addr source_addr; /* Source address to use */
25 struct iface *iface; /* Interface for link-local addresses */
26 u16 remote_port; /* Neighbor destination port */
27 int multihop; /* Number of hops if multihop */
28 int ttl_security; /* Enable TTL security [RFC5082] */
29 int next_hop_self; /* Always set next hop to local IP address */
30 int next_hop_keep; /* Do not touch next hop attribute */
31 int missing_lladdr; /* What we will do when we don' know link-local addr, see MLL_* */
32 int gw_mode; /* How we compute route gateway from next_hop attr, see GW_* */
33 int compare_path_lengths; /* Use path lengths when selecting best route */
34 int med_metric; /* Compare MULTI_EXIT_DISC even between routes from differen ASes */
35 int igp_metric; /* Use IGP metrics when selecting best route */
36 int prefer_older; /* Prefer older routes according to RFC 5004 */
37 int deterministic_med; /* Use more complicated algo to have strict RFC 4271 MED comparison */
38 u32 default_local_pref; /* Default value for LOCAL_PREF attribute */
39 u32 default_med; /* Default value for MULTI_EXIT_DISC attribute */
40 int capabilities; /* Enable capability handshake [RFC3392] */
41 int enable_refresh; /* Enable local support for route refresh [RFC2918] */
42 int enable_as4; /* Enable local support for 4B AS numbers [RFC4893] */
43 int enable_extended_messages; /* Enable local support for extended messages [draft] */
44 u32 rr_cluster_id; /* Route reflector cluster ID, if different from local ID */
45 int rr_client; /* Whether neighbor is RR client of me */
46 int rs_client; /* Whether neighbor is RS client of me */
47 int advertise_ipv4; /* Whether we should add IPv4 capability advertisement to OPEN message */
48 int passive; /* Do not initiate outgoing connection */
49 int interpret_communities; /* Hardwired handling of well-known communities */
50 int secondary; /* Accept also non-best routes (i.e. RA_ACCEPTED) */
51 int add_path; /* Use ADD-PATH extension [RFC7911] */
52 int allow_local_as; /* Allow that number of local ASNs in incoming AS_PATHs */
53 int allow_local_pref; /* Allow LOCAL_PREF in EBGP sessions */
54 int gr_mode; /* Graceful restart mode (BGP_GR_*) */
55 int llgr_mode; /* Long-lived graceful restart mode (BGP_LLGR_*) */
56 int setkey; /* Set MD5 password to system SA/SP database */
57 unsigned gr_time; /* Graceful restart timeout */
58 unsigned llgr_time; /* Long-lived graceful restart timeout */
59 unsigned connect_delay_time; /* Minimum delay between connect attempts */
60 unsigned connect_retry_time; /* Timeout for connect attempts */
61 unsigned hold_time, initial_hold_time;
62 unsigned keepalive_time;
63 unsigned error_amnesia_time; /* Errors are forgotten after */
64 unsigned error_delay_time_min; /* Time to wait after an error is detected */
65 unsigned error_delay_time_max;
66 unsigned disable_after_error; /* Disable the protocol when error is detected */
67 u32 disable_after_cease; /* Disable it when cease is received, bitfield */
68
69 char *password; /* Password used for MD5 authentication */
70 struct rtable_config *igp_table; /* Table used for recursive next hop lookups */
71 int check_link; /* Use iface link state for liveness detection */
72 int bfd; /* Use BFD for liveness detection */
73 };
74
75 #define MLL_SELF 1
76 #define MLL_DROP 2
77 #define MLL_IGNORE 3
78
79 #define GW_DIRECT 1
80 #define GW_RECURSIVE 2
81
82 #define ADD_PATH_RX 1
83 #define ADD_PATH_TX 2
84 #define ADD_PATH_FULL 3
85
86 #define BGP_GR_ABLE 1
87 #define BGP_GR_AWARE 2
88
89 /* For peer_gr_flags */
90 #define BGP_GRF_RESTART 0x80
91
92 /* For peer_gr_aflags */
93 #define BGP_GRF_FORWARDING 0x80
94
95 #define BGP_LLGR_ABLE 1
96 #define BGP_LLGR_AWARE 2
97
98 #define BGP_LLGRF_FORWARDING 0x80
99
100 #define BGP_GRS_NONE 0 /* No GR */
101 #define BGP_GRS_ACTIVE 1 /* Graceful restart per RFC 4724 */
102 #define BGP_GRS_LLGR_1 2 /* Long-lived GR phase 1 (restart time) */
103 #define BGP_GRS_LLGR_2 3 /* Long-lived GR phase 2 (stale time) */
104
105 #define BGP_BFD_GRACEFUL 2 /* BFD down triggers graceful restart */
106
107
108 struct bgp_conn {
109 struct bgp_proto *bgp;
110 struct birdsock *sk;
111 uint state; /* State of connection state machine */
112 struct timer *connect_retry_timer;
113 struct timer *hold_timer;
114 struct timer *keepalive_timer;
115 struct event *tx_ev;
116 int packets_to_send; /* Bitmap of packet types to be sent */
117 int notify_code, notify_subcode, notify_size;
118 byte *notify_data;
119 u32 advertised_as; /* Temporary value for AS number received */
120 int start_state; /* protocol start_state snapshot when connection established */
121 u8 peer_refresh_support; /* Peer supports route refresh [RFC2918] */
122 u8 peer_as4_support; /* Peer supports 4B AS numbers [RFC4893] */
123 u8 peer_add_path; /* Peer supports ADD-PATH [RFC7911] */
124 u8 peer_enhanced_refresh_support; /* Peer supports enhanced refresh [RFC7313] */
125 u8 peer_gr_aware;
126 u8 peer_gr_able;
127 u16 peer_gr_time;
128 u8 peer_gr_flags;
129 u8 peer_gr_aflags;
130 u8 peer_llgr_aware;
131 u8 peer_llgr_able;
132 uint peer_llgr_time;
133 u8 peer_llgr_aflags;
134 u8 peer_ext_messages_support; /* Peer supports extended message length [draft] */
135 unsigned hold_time, keepalive_time; /* Times calculated from my and neighbor's requirements */
136 };
137
138 struct bgp_proto {
139 struct proto p;
140 struct bgp_config *cf; /* Shortcut to BGP configuration */
141 u32 local_as, remote_as;
142 int start_state; /* Substates that partitions BS_START */
143 u8 is_internal; /* Internal BGP connection (local_as == remote_as) */
144 u8 as4_session; /* Session uses 4B AS numbers in AS_PATH (both sides support it) */
145 u8 add_path_rx; /* Session expects receive of ADD-PATH extended NLRI */
146 u8 add_path_tx; /* Session expects transmit of ADD-PATH extended NLRI */
147 u8 ext_messages; /* Session allows to use extended messages (both sides support it) */
148 u32 local_id; /* BGP identifier of this router */
149 u32 remote_id; /* BGP identifier of the neighbor */
150 u32 rr_cluster_id; /* Route reflector cluster ID */
151 int rr_client; /* Whether neighbor is RR client of me */
152 int rs_client; /* Whether neighbor is RS client of me */
153 u8 gr_ready; /* Neighbor could do graceful restart */
154 u8 gr_active; /* Neighbor is doing graceful restart (BGP_GRS_*) */
155 u8 feed_state; /* Feed state (TX) for EoR, RR packets, see BFS_* */
156 u8 load_state; /* Load state (RX) for EoR, RR packets, see BFS_* */
157 uint stale_time; /* Long-lived stale time for LLGR */
158 struct bgp_conn *conn; /* Connection we have established */
159 struct bgp_conn outgoing_conn; /* Outgoing connection we're working with */
160 struct bgp_conn incoming_conn; /* Incoming connection we have neither accepted nor rejected yet */
161 struct object_lock *lock; /* Lock for neighbor connection */
162 struct neighbor *neigh; /* Neighbor entry corresponding to remote ip, NULL if multihop */
163 struct bfd_request *bfd_req; /* BFD request, if BFD is used */
164 ip_addr source_addr; /* Local address used as an advertised next hop */
165 rtable *igp_table; /* Table used for recursive next hop lookups */
166 struct event *event; /* Event for respawning and shutting process */
167 struct timer *startup_timer; /* Timer used to delay protocol startup due to previous errors (startup_delay) */
168 struct timer *gr_timer; /* Timer waiting for reestablishment after graceful restart */
169 struct bgp_bucket **bucket_hash; /* Hash table of attribute buckets */
170 uint hash_size, hash_count, hash_limit;
171 HASH(struct bgp_prefix) prefix_hash; /* Prefixes to be sent */
172 slab *prefix_slab; /* Slab holding prefix nodes */
173 list bucket_queue; /* Queue of buckets to send */
174 struct bgp_bucket *withdraw_bucket; /* Withdrawn routes */
175 unsigned startup_delay; /* Time to delay protocol startup by due to errors */
176 bird_clock_t last_proto_error; /* Time of last error that leads to protocol stop */
177 u8 last_error_class; /* Error class of last error */
178 u32 last_error_code; /* Error code of last error. BGP protocol errors
179 are encoded as (bgp_err_code << 16 | bgp_err_subcode) */
180 #ifdef IPV6
181 byte *mp_reach_start, *mp_unreach_start; /* Multiprotocol BGP attribute notes */
182 unsigned mp_reach_len, mp_unreach_len;
183 ip_addr local_link; /* Link-level version of source_addr */
184 #endif
185 };
186
187 struct bgp_prefix {
188 struct {
189 ip_addr prefix;
190 int pxlen;
191 } n;
192 u32 path_id;
193 struct bgp_prefix *next;
194 node bucket_node; /* Node in per-bucket list */
195 };
196
197 struct bgp_bucket {
198 node send_node; /* Node in send queue */
199 struct bgp_bucket *hash_next, *hash_prev; /* Node in bucket hash table */
200 unsigned hash; /* Hash over extended attributes */
201 list prefixes; /* Prefixes in this buckets */
202 ea_list eattrs[0]; /* Per-bucket extended attributes */
203 };
204
205 #define BGP_PORT 179
206 #define BGP_VERSION 4
207 #define BGP_HEADER_LENGTH 19
208 #define BGP_MAX_MESSAGE_LENGTH 4096
209 #define BGP_MAX_EXT_MSG_LENGTH 65535
210 #define BGP_RX_BUFFER_SIZE 4096
211 #define BGP_TX_BUFFER_SIZE 4096
212 #define BGP_RX_BUFFER_EXT_SIZE 65535
213 #define BGP_TX_BUFFER_EXT_SIZE 65535
214
bgp_max_packet_length(struct bgp_proto * p)215 static inline uint bgp_max_packet_length(struct bgp_proto *p)
216 { return p->ext_messages ? BGP_MAX_EXT_MSG_LENGTH : BGP_MAX_MESSAGE_LENGTH; }
217
218 extern struct linpool *bgp_linpool;
219
220
221 void bgp_start_timer(struct timer *t, int value);
222 void bgp_check_config(struct bgp_config *c);
223 void bgp_error(struct bgp_conn *c, unsigned code, unsigned subcode, byte *data, int len);
224 void bgp_close_conn(struct bgp_conn *c);
225 void bgp_update_startup_delay(struct bgp_proto *p);
226 void bgp_conn_enter_openconfirm_state(struct bgp_conn *conn);
227 void bgp_conn_enter_established_state(struct bgp_conn *conn);
228 void bgp_conn_enter_close_state(struct bgp_conn *conn);
229 void bgp_conn_enter_idle_state(struct bgp_conn *conn);
230 void bgp_handle_graceful_restart(struct bgp_proto *p);
231 void bgp_graceful_restart_done(struct bgp_proto *p);
232 void bgp_refresh_begin(struct bgp_proto *p);
233 void bgp_refresh_end(struct bgp_proto *p);
234 void bgp_store_error(struct bgp_proto *p, struct bgp_conn *c, u8 class, u32 code);
235 void bgp_stop(struct bgp_proto *p, uint subcode, byte *data, uint len);
236
237 struct rte_source *bgp_find_source(struct bgp_proto *p, u32 path_id);
238 struct rte_source *bgp_get_source(struct bgp_proto *p, u32 path_id);
239
240
241
242 #ifdef LOCAL_DEBUG
243 #define BGP_FORCE_DEBUG 1
244 #else
245 #define BGP_FORCE_DEBUG 0
246 #endif
247 #define BGP_TRACE(flags, msg, args...) do { if ((p->p.debug & flags) || BGP_FORCE_DEBUG) \
248 log(L_TRACE "%s: " msg, p->p.name , ## args ); } while(0)
249
250 #define BGP_TRACE_RL(rl, flags, msg, args...) do { if ((p->p.debug & flags) || BGP_FORCE_DEBUG) \
251 log_rl(rl, L_TRACE "%s: " msg, p->p.name , ## args ); } while(0)
252
253
254 /* attrs.c */
255
256 /* Hack: although BA_NEXT_HOP attribute has type EAF_TYPE_IP_ADDRESS, in IPv6
257 * we store two addesses in it - a global address and a link local address.
258 */
259 #ifdef IPV6
260 #define NEXT_HOP_LENGTH (2*sizeof(ip_addr))
set_next_hop(byte * b,ip_addr addr)261 static inline void set_next_hop(byte *b, ip_addr addr) { ((ip_addr *) b)[0] = addr; ((ip_addr *) b)[1] = IPA_NONE; }
262 #else
263 #define NEXT_HOP_LENGTH sizeof(ip_addr)
set_next_hop(byte * b,ip_addr addr)264 static inline void set_next_hop(byte *b, ip_addr addr) { ((ip_addr *) b)[0] = addr; }
265 #endif
266
267 void bgp_attach_attr(struct ea_list **to, struct linpool *pool, unsigned attr, uintptr_t val);
268 byte *bgp_attach_attr_wa(struct ea_list **to, struct linpool *pool, unsigned attr, unsigned len);
269 struct rta *bgp_decode_attrs(struct bgp_conn *conn, byte *a, uint len, struct linpool *pool, int mandatory);
270 int bgp_get_attr(struct eattr *e, byte *buf, int buflen);
271 int bgp_rte_better(struct rte *, struct rte *);
272 int bgp_rte_mergable(rte *pri, rte *sec);
273 int bgp_rte_recalculate(rtable *table, net *net, rte *new, rte *old, rte *old_best);
274 struct rte *bgp_rte_modify_stale(struct rte *r, struct linpool *pool);
275 void bgp_rt_notify(struct proto *P, rtable *tbl UNUSED, net *n, rte *new, rte *old UNUSED, ea_list *attrs);
276 int bgp_import_control(struct proto *, struct rte **, struct ea_list **, struct linpool *);
277 void bgp_init_bucket_table(struct bgp_proto *);
278 void bgp_free_bucket_table(struct bgp_proto *p);
279 void bgp_free_bucket(struct bgp_proto *p, struct bgp_bucket *buck);
280 void bgp_init_prefix_table(struct bgp_proto *p, u32 order);
281 void bgp_free_prefix_table(struct bgp_proto *p);
282 void bgp_free_prefix(struct bgp_proto *p, struct bgp_prefix *bp);
283 uint bgp_encode_attrs(struct bgp_proto *p, byte *w, ea_list *attrs, int remains);
284 void bgp_get_route_info(struct rte *, byte *buf, struct ea_list *attrs);
285
bgp_attach_attr_ip(struct ea_list ** to,struct linpool * pool,unsigned attr,ip_addr a)286 inline static void bgp_attach_attr_ip(struct ea_list **to, struct linpool *pool, unsigned attr, ip_addr a)
287 { *(ip_addr *) bgp_attach_attr_wa(to, pool, attr, sizeof(ip_addr)) = a; }
288
289 /* packets.c */
290
291 void bgp_dump_state_change(struct bgp_conn *conn, uint old, uint new);
292 void bgp_schedule_packet(struct bgp_conn *conn, int type);
293 void bgp_kick_tx(void *vconn);
294 void bgp_tx(struct birdsock *sk);
295 int bgp_rx(struct birdsock *sk, uint size);
296 const char * bgp_error_dsc(unsigned code, unsigned subcode);
297 void bgp_log_error(struct bgp_proto *p, u8 class, char *msg, unsigned code, unsigned subcode, byte *data, unsigned len);
298
299 /* Packet types */
300
301 #define PKT_OPEN 0x01
302 #define PKT_UPDATE 0x02
303 #define PKT_NOTIFICATION 0x03
304 #define PKT_KEEPALIVE 0x04
305 #define PKT_ROUTE_REFRESH 0x05 /* [RFC2918] */
306 #define PKT_BEGIN_REFRESH 0x1e /* Dummy type for BoRR packet [RFC7313] */
307 #define PKT_SCHEDULE_CLOSE 0x1f /* Used internally to schedule socket close */
308
309 /* Attributes */
310
311 #define BAF_OPTIONAL 0x80
312 #define BAF_TRANSITIVE 0x40
313 #define BAF_PARTIAL 0x20
314 #define BAF_EXT_LEN 0x10
315
316 #define BA_ORIGIN 0x01 /* [RFC1771] */ /* WM */
317 #define BA_AS_PATH 0x02 /* WM */
318 #define BA_NEXT_HOP 0x03 /* WM */
319 #define BA_MULTI_EXIT_DISC 0x04 /* ON */
320 #define BA_LOCAL_PREF 0x05 /* WD */
321 #define BA_ATOMIC_AGGR 0x06 /* WD */
322 #define BA_AGGREGATOR 0x07 /* OT */
323 #define BA_COMMUNITY 0x08 /* [RFC1997] */ /* OT */
324 #define BA_ORIGINATOR_ID 0x09 /* [RFC1966] */ /* ON */
325 #define BA_CLUSTER_LIST 0x0a /* ON */
326 /* We don't support these: */
327 #define BA_DPA 0x0b /* ??? */
328 #define BA_ADVERTISER 0x0c /* [RFC1863] */
329 #define BA_RCID_PATH 0x0d
330 #define BA_MP_REACH_NLRI 0x0e /* [RFC2283] */
331 #define BA_MP_UNREACH_NLRI 0x0f
332 #define BA_EXT_COMMUNITY 0x10 /* [RFC4360] */
333 #define BA_AS4_PATH 0x11 /* [RFC4893] */
334 #define BA_AS4_AGGREGATOR 0x12
335 #define BA_LARGE_COMMUNITY 0x20 /* [RFC8092] */
336
337 /* BGP connection states */
338
339 #define BS_IDLE 0
340 #define BS_CONNECT 1 /* Attempting to connect */
341 #define BS_ACTIVE 2 /* Waiting for connection retry & listening */
342 #define BS_OPENSENT 3
343 #define BS_OPENCONFIRM 4
344 #define BS_ESTABLISHED 5
345 #define BS_CLOSE 6 /* Used during transition to BS_IDLE */
346
347 #define BS_MAX 7
348
349 /* BGP start states
350 *
351 * Used in PS_START for fine-grained specification of starting state.
352 *
353 * When BGP protocol is started by core, it goes to BSS_PREPARE. When BGP
354 * protocol done what is neccessary to start itself (like acquiring the lock),
355 * it goes to BSS_CONNECT. When some connection attempt failed because of
356 * option or capability error, it goes to BSS_CONNECT_NOCAP.
357 */
358
359 #define BSS_PREPARE 0 /* Used before ordinary BGP started, i. e. waiting for lock */
360 #define BSS_DELAY 1 /* Startup delay due to previous errors */
361 #define BSS_CONNECT 2 /* Ordinary BGP connecting */
362 #define BSS_CONNECT_NOCAP 3 /* Legacy BGP connecting (without capabilities) */
363
364
365 /* BGP feed states (TX)
366 *
367 * RFC 4724 specifies that an initial feed should end with End-of-RIB mark.
368 *
369 * RFC 7313 specifies that a route refresh should be demarcated by BoRR and EoRR packets.
370 *
371 * These states (stored in p->feed_state) are used to keep track of these
372 * requirements. When such feed is started, BFS_LOADING / BFS_REFRESHING is
373 * set. When it ended, BFS_LOADED / BFS_REFRESHED is set to schedule End-of-RIB
374 * or EoRR packet. When the packet is sent, the state returned to BFS_NONE.
375 *
376 * Note that when a non-demarcated feed (e.g. plain RFC 4271 initial load
377 * without End-of-RIB or plain RFC 2918 route refresh without BoRR/EoRR
378 * demarcation) is active, BFS_NONE is set.
379 *
380 * BFS_NONE, BFS_LOADING and BFS_REFRESHING are also used as load states (RX)
381 * with correspondent semantics (-, expecting End-of-RIB, expecting EoRR).
382 */
383
384 #define BFS_NONE 0 /* No feed or original non-demarcated feed */
385 #define BFS_LOADING 1 /* Initial feed active, End-of-RIB planned */
386 #define BFS_LOADED 2 /* Loading done, End-of-RIB marker scheduled */
387 #define BFS_REFRESHING 3 /* Route refresh (introduced by BoRR) active */
388 #define BFS_REFRESHED 4 /* Refresh done, EoRR packet scheduled */
389
390
391 /* Error classes */
392
393 #define BE_NONE 0
394 #define BE_MISC 1 /* Miscellaneous error */
395 #define BE_SOCKET 2 /* Socket error */
396 #define BE_BGP_RX 3 /* BGP protocol error notification received */
397 #define BE_BGP_TX 4 /* BGP protocol error notification sent */
398 #define BE_AUTO_DOWN 5 /* Automatic shutdown */
399 #define BE_MAN_DOWN 6 /* Manual shutdown */
400
401 /* Misc error codes */
402
403 #define BEM_NEIGHBOR_LOST 1
404 #define BEM_INVALID_NEXT_HOP 2
405 #define BEM_INVALID_MD5 3 /* MD5 authentication kernel request failed (possibly not supported) */
406 #define BEM_NO_SOCKET 4
407 #define BEM_LINK_DOWN 5
408 #define BEM_BFD_DOWN 6
409 #define BEM_GRACEFUL_RESTART 7
410
411 /* Automatic shutdown error codes */
412
413 #define BEA_ROUTE_LIMIT_EXCEEDED 1
414
415 /* Well-known communities */
416
417 #define BGP_COMM_NO_EXPORT 0xffffff01 /* Don't export outside local AS / confed. */
418 #define BGP_COMM_NO_ADVERTISE 0xffffff02 /* Don't export at all */
419 #define BGP_COMM_NO_EXPORT_SUBCONFED 0xffffff03 /* NO_EXPORT even in local confederation */
420
421 #define BGP_COMM_LLGR_STALE 0xffff0006 /* Route is stale according to LLGR */
422 #define BGP_COMM_NO_LLGR 0xffff0007 /* Do not treat the route according to LLGR */
423
424 /* Origins */
425
426 #define ORIGIN_IGP 0
427 #define ORIGIN_EGP 1
428 #define ORIGIN_INCOMPLETE 2
429
430 /* Address families */
431
432 #define BGP_AF_IPV4 1
433 #define BGP_AF_IPV6 2
434
435 #ifdef IPV6
436 #define BGP_AF BGP_AF_IPV6
437 #else
438 #define BGP_AF BGP_AF_IPV4
439 #endif
440
441 #endif
442