]>
Commit | Line | Data |
---|---|---|
1 | /* BGP-4 Finite State Machine | |
2 | * From RFC1771 [A Border Gateway Protocol 4 (BGP-4)] | |
3 | * Copyright (C) 1996, 97, 98 Kunihiro Ishiguro | |
4 | * | |
5 | * This file is part of GNU Zebra. | |
6 | * | |
7 | * GNU Zebra is free software; you can redistribute it and/or modify it | |
8 | * under the terms of the GNU General Public License as published by the | |
9 | * Free Software Foundation; either version 2, or (at your option) any | |
10 | * later version. | |
11 | * | |
12 | * GNU Zebra is distributed in the hope that it will be useful, but | |
13 | * WITHOUT ANY WARRANTY; without even the implied warranty of | |
14 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU | |
15 | * General Public License for more details. | |
16 | * | |
17 | * You should have received a copy of the GNU General Public License along | |
18 | * with this program; see the file COPYING; if not, write to the Free Software | |
19 | * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA | |
20 | */ | |
21 | ||
22 | #include <zebra.h> | |
23 | ||
24 | #include "linklist.h" | |
25 | #include "prefix.h" | |
26 | #include "sockunion.h" | |
27 | #include "thread.h" | |
28 | #include "log.h" | |
29 | #include "stream.h" | |
30 | #include "ringbuf.h" | |
31 | #include "memory.h" | |
32 | #include "plist.h" | |
33 | #include "workqueue.h" | |
34 | #include "queue.h" | |
35 | #include "filter.h" | |
36 | #include "command.h" | |
37 | #include "lib_errors.h" | |
38 | ||
39 | #include "lib/json.h" | |
40 | #include "bgpd/bgpd.h" | |
41 | #include "bgpd/bgp_attr.h" | |
42 | #include "bgpd/bgp_debug.h" | |
43 | #include "bgpd/bgp_errors.h" | |
44 | #include "bgpd/bgp_fsm.h" | |
45 | #include "bgpd/bgp_packet.h" | |
46 | #include "bgpd/bgp_network.h" | |
47 | #include "bgpd/bgp_route.h" | |
48 | #include "bgpd/bgp_dump.h" | |
49 | #include "bgpd/bgp_open.h" | |
50 | #include "bgpd/bgp_advertise.h" | |
51 | #include "bgpd/bgp_updgrp.h" | |
52 | #include "bgpd/bgp_nht.h" | |
53 | #include "bgpd/bgp_bfd.h" | |
54 | #include "bgpd/bgp_memory.h" | |
55 | #include "bgpd/bgp_keepalives.h" | |
56 | #include "bgpd/bgp_io.h" | |
57 | #include "bgpd/bgp_zebra.h" | |
58 | ||
59 | DEFINE_HOOK(peer_backward_transition, (struct peer * peer), (peer)) | |
60 | DEFINE_HOOK(peer_established, (struct peer * peer), (peer)) | |
61 | ||
62 | /* Definition of display strings corresponding to FSM events. This should be | |
63 | * kept consistent with the events defined in bgpd.h | |
64 | */ | |
65 | static const char *bgp_event_str[] = { | |
66 | NULL, | |
67 | "BGP_Start", | |
68 | "BGP_Stop", | |
69 | "TCP_connection_open", | |
70 | "TCP_connection_closed", | |
71 | "TCP_connection_open_failed", | |
72 | "TCP_fatal_error", | |
73 | "ConnectRetry_timer_expired", | |
74 | "Hold_Timer_expired", | |
75 | "KeepAlive_timer_expired", | |
76 | "Receive_OPEN_message", | |
77 | "Receive_KEEPALIVE_message", | |
78 | "Receive_UPDATE_message", | |
79 | "Receive_NOTIFICATION_message", | |
80 | "Clearing_Completed", | |
81 | }; | |
82 | ||
83 | /* BGP FSM (finite state machine) has three types of functions. Type | |
84 | one is thread functions. Type two is event functions. Type three | |
85 | is FSM functions. Timer functions are set by bgp_timer_set | |
86 | function. */ | |
87 | ||
88 | /* BGP event function. */ | |
89 | int bgp_event(struct thread *); | |
90 | ||
91 | /* BGP thread functions. */ | |
92 | static int bgp_start_timer(struct thread *); | |
93 | static int bgp_connect_timer(struct thread *); | |
94 | static int bgp_holdtime_timer(struct thread *); | |
95 | ||
96 | /* BGP FSM functions. */ | |
97 | static int bgp_start(struct peer *); | |
98 | ||
99 | static void peer_xfer_stats(struct peer *peer_dst, struct peer *peer_src) | |
100 | { | |
101 | /* Copy stats over. These are only the pre-established state stats */ | |
102 | peer_dst->open_in += peer_src->open_in; | |
103 | peer_dst->open_out += peer_src->open_out; | |
104 | peer_dst->keepalive_in += peer_src->keepalive_in; | |
105 | peer_dst->keepalive_out += peer_src->keepalive_out; | |
106 | peer_dst->notify_in += peer_src->notify_in; | |
107 | peer_dst->notify_out += peer_src->notify_out; | |
108 | peer_dst->dynamic_cap_in += peer_src->dynamic_cap_in; | |
109 | peer_dst->dynamic_cap_out += peer_src->dynamic_cap_out; | |
110 | } | |
111 | ||
112 | static struct peer *peer_xfer_conn(struct peer *from_peer) | |
113 | { | |
114 | struct peer *peer; | |
115 | afi_t afi; | |
116 | safi_t safi; | |
117 | int fd; | |
118 | int status, pstatus; | |
119 | unsigned char last_evt, last_maj_evt; | |
120 | ||
121 | assert(from_peer != NULL); | |
122 | ||
123 | peer = from_peer->doppelganger; | |
124 | ||
125 | if (!peer || !CHECK_FLAG(peer->flags, PEER_FLAG_CONFIG_NODE)) | |
126 | return from_peer; | |
127 | ||
128 | /* | |
129 | * Let's check that we are not going to loose known configuration | |
130 | * state based upon doppelganger rules. | |
131 | */ | |
132 | FOREACH_AFI_SAFI (afi, safi) { | |
133 | if (from_peer->afc[afi][safi] != peer->afc[afi][safi]) { | |
134 | flog_err( | |
135 | EC_BGP_DOPPELGANGER_CONFIG, | |
136 | "from_peer->afc[%d][%d] is not the same as what we are overwriting", | |
137 | afi, safi); | |
138 | return NULL; | |
139 | } | |
140 | } | |
141 | ||
142 | if (bgp_debug_neighbor_events(peer)) | |
143 | zlog_debug("%s: peer transfer %p fd %d -> %p fd %d)", | |
144 | from_peer->host, from_peer, from_peer->fd, peer, | |
145 | peer->fd); | |
146 | ||
147 | bgp_writes_off(peer); | |
148 | bgp_reads_off(peer); | |
149 | bgp_writes_off(from_peer); | |
150 | bgp_reads_off(from_peer); | |
151 | ||
152 | BGP_TIMER_OFF(peer->t_routeadv); | |
153 | BGP_TIMER_OFF(peer->t_connect); | |
154 | BGP_TIMER_OFF(peer->t_connect_check_r); | |
155 | BGP_TIMER_OFF(peer->t_connect_check_w); | |
156 | BGP_TIMER_OFF(from_peer->t_routeadv); | |
157 | BGP_TIMER_OFF(from_peer->t_connect); | |
158 | BGP_TIMER_OFF(from_peer->t_connect_check_r); | |
159 | BGP_TIMER_OFF(from_peer->t_connect_check_w); | |
160 | BGP_TIMER_OFF(from_peer->t_process_packet); | |
161 | ||
162 | /* | |
163 | * At this point in time, it is possible that there are packets pending | |
164 | * on various buffers. Those need to be transferred or dropped, | |
165 | * otherwise we'll get spurious failures during session establishment. | |
166 | */ | |
167 | pthread_mutex_lock(&peer->io_mtx); | |
168 | pthread_mutex_lock(&from_peer->io_mtx); | |
169 | { | |
170 | fd = peer->fd; | |
171 | peer->fd = from_peer->fd; | |
172 | from_peer->fd = fd; | |
173 | ||
174 | stream_fifo_clean(peer->ibuf); | |
175 | stream_fifo_clean(peer->obuf); | |
176 | ||
177 | /* | |
178 | * this should never happen, since bgp_process_packet() is the | |
179 | * only task that sets and unsets the current packet and it | |
180 | * runs in our pthread. | |
181 | */ | |
182 | if (peer->curr) { | |
183 | flog_err( | |
184 | EC_BGP_PKT_PROCESS, | |
185 | "[%s] Dropping pending packet on connection transfer:", | |
186 | peer->host); | |
187 | uint16_t type = stream_getc_from(peer->curr, | |
188 | BGP_MARKER_SIZE + 2); | |
189 | bgp_dump_packet(peer, type, peer->curr); | |
190 | stream_free(peer->curr); | |
191 | peer->curr = NULL; | |
192 | } | |
193 | ||
194 | // copy each packet from old peer's output queue to new peer | |
195 | while (from_peer->obuf->head) | |
196 | stream_fifo_push(peer->obuf, | |
197 | stream_fifo_pop(from_peer->obuf)); | |
198 | ||
199 | // copy each packet from old peer's input queue to new peer | |
200 | while (from_peer->ibuf->head) | |
201 | stream_fifo_push(peer->ibuf, | |
202 | stream_fifo_pop(from_peer->ibuf)); | |
203 | ||
204 | ringbuf_wipe(peer->ibuf_work); | |
205 | ringbuf_copy(peer->ibuf_work, from_peer->ibuf_work, | |
206 | ringbuf_remain(from_peer->ibuf_work)); | |
207 | } | |
208 | pthread_mutex_unlock(&from_peer->io_mtx); | |
209 | pthread_mutex_unlock(&peer->io_mtx); | |
210 | ||
211 | peer->as = from_peer->as; | |
212 | peer->v_holdtime = from_peer->v_holdtime; | |
213 | peer->v_keepalive = from_peer->v_keepalive; | |
214 | peer->v_routeadv = from_peer->v_routeadv; | |
215 | peer->v_gr_restart = from_peer->v_gr_restart; | |
216 | peer->cap = from_peer->cap; | |
217 | status = peer->status; | |
218 | pstatus = peer->ostatus; | |
219 | last_evt = peer->last_event; | |
220 | last_maj_evt = peer->last_major_event; | |
221 | peer->status = from_peer->status; | |
222 | peer->ostatus = from_peer->ostatus; | |
223 | peer->last_event = from_peer->last_event; | |
224 | peer->last_major_event = from_peer->last_major_event; | |
225 | from_peer->status = status; | |
226 | from_peer->ostatus = pstatus; | |
227 | from_peer->last_event = last_evt; | |
228 | from_peer->last_major_event = last_maj_evt; | |
229 | peer->remote_id = from_peer->remote_id; | |
230 | ||
231 | if (from_peer->hostname != NULL) { | |
232 | if (peer->hostname) { | |
233 | XFREE(MTYPE_BGP_PEER_HOST, peer->hostname); | |
234 | peer->hostname = NULL; | |
235 | } | |
236 | ||
237 | peer->hostname = from_peer->hostname; | |
238 | from_peer->hostname = NULL; | |
239 | } | |
240 | ||
241 | if (from_peer->domainname != NULL) { | |
242 | if (peer->domainname) { | |
243 | XFREE(MTYPE_BGP_PEER_HOST, peer->domainname); | |
244 | peer->domainname = NULL; | |
245 | } | |
246 | ||
247 | peer->domainname = from_peer->domainname; | |
248 | from_peer->domainname = NULL; | |
249 | } | |
250 | ||
251 | FOREACH_AFI_SAFI (afi, safi) { | |
252 | peer->af_flags[afi][safi] = from_peer->af_flags[afi][safi]; | |
253 | peer->af_sflags[afi][safi] = from_peer->af_sflags[afi][safi]; | |
254 | peer->af_cap[afi][safi] = from_peer->af_cap[afi][safi]; | |
255 | peer->afc_nego[afi][safi] = from_peer->afc_nego[afi][safi]; | |
256 | peer->afc_adv[afi][safi] = from_peer->afc_adv[afi][safi]; | |
257 | peer->afc_recv[afi][safi] = from_peer->afc_recv[afi][safi]; | |
258 | peer->orf_plist[afi][safi] = from_peer->orf_plist[afi][safi]; | |
259 | } | |
260 | ||
261 | if (bgp_getsockname(peer) < 0) { | |
262 | flog_err( | |
263 | EC_LIB_SOCKET, | |
264 | "%%bgp_getsockname() failed for %s peer %s fd %d (from_peer fd %d)", | |
265 | (CHECK_FLAG(peer->sflags, PEER_STATUS_ACCEPT_PEER) | |
266 | ? "accept" | |
267 | : ""), | |
268 | peer->host, peer->fd, from_peer->fd); | |
269 | bgp_stop(peer); | |
270 | bgp_stop(from_peer); | |
271 | return NULL; | |
272 | } | |
273 | if (from_peer->status > Active) { | |
274 | if (bgp_getsockname(from_peer) < 0) { | |
275 | flog_err( | |
276 | EC_LIB_SOCKET, | |
277 | "%%bgp_getsockname() failed for %s from_peer %s fd %d (peer fd %d)", | |
278 | ||
279 | (CHECK_FLAG(from_peer->sflags, | |
280 | PEER_STATUS_ACCEPT_PEER) | |
281 | ? "accept" | |
282 | : ""), | |
283 | from_peer->host, from_peer->fd, peer->fd); | |
284 | bgp_stop(from_peer); | |
285 | from_peer = NULL; | |
286 | } | |
287 | } | |
288 | ||
289 | ||
290 | // Note: peer_xfer_stats() must be called with I/O turned OFF | |
291 | if (from_peer) | |
292 | peer_xfer_stats(peer, from_peer); | |
293 | ||
294 | bgp_reads_on(peer); | |
295 | bgp_writes_on(peer); | |
296 | thread_add_timer_msec(bm->master, bgp_process_packet, peer, 0, | |
297 | &peer->t_process_packet); | |
298 | ||
299 | return (peer); | |
300 | } | |
301 | ||
302 | /* Hook function called after bgp event is occered. And vty's | |
303 | neighbor command invoke this function after making neighbor | |
304 | structure. */ | |
305 | void bgp_timer_set(struct peer *peer) | |
306 | { | |
307 | switch (peer->status) { | |
308 | case Idle: | |
309 | /* First entry point of peer's finite state machine. In Idle | |
310 | status start timer is on unless peer is shutdown or peer is | |
311 | inactive. All other timer must be turned off */ | |
312 | if (BGP_PEER_START_SUPPRESSED(peer) || !peer_active(peer) | |
313 | || peer->bgp->vrf_id == VRF_UNKNOWN) { | |
314 | BGP_TIMER_OFF(peer->t_start); | |
315 | } else { | |
316 | BGP_TIMER_ON(peer->t_start, bgp_start_timer, | |
317 | peer->v_start); | |
318 | } | |
319 | BGP_TIMER_OFF(peer->t_connect); | |
320 | BGP_TIMER_OFF(peer->t_holdtime); | |
321 | bgp_keepalives_off(peer); | |
322 | BGP_TIMER_OFF(peer->t_routeadv); | |
323 | break; | |
324 | ||
325 | case Connect: | |
326 | /* After start timer is expired, the peer moves to Connect | |
327 | status. Make sure start timer is off and connect timer is | |
328 | on. */ | |
329 | BGP_TIMER_OFF(peer->t_start); | |
330 | BGP_TIMER_ON(peer->t_connect, bgp_connect_timer, | |
331 | peer->v_connect); | |
332 | BGP_TIMER_OFF(peer->t_holdtime); | |
333 | bgp_keepalives_off(peer); | |
334 | BGP_TIMER_OFF(peer->t_routeadv); | |
335 | break; | |
336 | ||
337 | case Active: | |
338 | /* Active is waiting connection from remote peer. And if | |
339 | connect timer is expired, change status to Connect. */ | |
340 | BGP_TIMER_OFF(peer->t_start); | |
341 | /* If peer is passive mode, do not set connect timer. */ | |
342 | if (CHECK_FLAG(peer->flags, PEER_FLAG_PASSIVE) | |
343 | || CHECK_FLAG(peer->sflags, PEER_STATUS_NSF_WAIT)) { | |
344 | BGP_TIMER_OFF(peer->t_connect); | |
345 | } else { | |
346 | BGP_TIMER_ON(peer->t_connect, bgp_connect_timer, | |
347 | peer->v_connect); | |
348 | } | |
349 | BGP_TIMER_OFF(peer->t_holdtime); | |
350 | bgp_keepalives_off(peer); | |
351 | BGP_TIMER_OFF(peer->t_routeadv); | |
352 | break; | |
353 | ||
354 | case OpenSent: | |
355 | /* OpenSent status. */ | |
356 | BGP_TIMER_OFF(peer->t_start); | |
357 | BGP_TIMER_OFF(peer->t_connect); | |
358 | if (peer->v_holdtime != 0) { | |
359 | BGP_TIMER_ON(peer->t_holdtime, bgp_holdtime_timer, | |
360 | peer->v_holdtime); | |
361 | } else { | |
362 | BGP_TIMER_OFF(peer->t_holdtime); | |
363 | } | |
364 | bgp_keepalives_off(peer); | |
365 | BGP_TIMER_OFF(peer->t_routeadv); | |
366 | break; | |
367 | ||
368 | case OpenConfirm: | |
369 | /* OpenConfirm status. */ | |
370 | BGP_TIMER_OFF(peer->t_start); | |
371 | BGP_TIMER_OFF(peer->t_connect); | |
372 | ||
373 | /* If the negotiated Hold Time value is zero, then the Hold Time | |
374 | timer and KeepAlive timers are not started. */ | |
375 | if (peer->v_holdtime == 0) { | |
376 | BGP_TIMER_OFF(peer->t_holdtime); | |
377 | bgp_keepalives_off(peer); | |
378 | } else { | |
379 | BGP_TIMER_ON(peer->t_holdtime, bgp_holdtime_timer, | |
380 | peer->v_holdtime); | |
381 | bgp_keepalives_on(peer); | |
382 | } | |
383 | BGP_TIMER_OFF(peer->t_routeadv); | |
384 | break; | |
385 | ||
386 | case Established: | |
387 | /* In Established status start and connect timer is turned | |
388 | off. */ | |
389 | BGP_TIMER_OFF(peer->t_start); | |
390 | BGP_TIMER_OFF(peer->t_connect); | |
391 | ||
392 | /* Same as OpenConfirm, if holdtime is zero then both holdtime | |
393 | and keepalive must be turned off. */ | |
394 | if (peer->v_holdtime == 0) { | |
395 | BGP_TIMER_OFF(peer->t_holdtime); | |
396 | bgp_keepalives_off(peer); | |
397 | } else { | |
398 | BGP_TIMER_ON(peer->t_holdtime, bgp_holdtime_timer, | |
399 | peer->v_holdtime); | |
400 | bgp_keepalives_on(peer); | |
401 | } | |
402 | break; | |
403 | case Deleted: | |
404 | BGP_TIMER_OFF(peer->t_gr_restart); | |
405 | BGP_TIMER_OFF(peer->t_gr_stale); | |
406 | BGP_TIMER_OFF(peer->t_pmax_restart); | |
407 | /* fallthru */ | |
408 | case Clearing: | |
409 | BGP_TIMER_OFF(peer->t_start); | |
410 | BGP_TIMER_OFF(peer->t_connect); | |
411 | BGP_TIMER_OFF(peer->t_holdtime); | |
412 | bgp_keepalives_off(peer); | |
413 | BGP_TIMER_OFF(peer->t_routeadv); | |
414 | break; | |
415 | } | |
416 | } | |
417 | ||
418 | /* BGP start timer. This function set BGP_Start event to thread value | |
419 | and process event. */ | |
420 | static int bgp_start_timer(struct thread *thread) | |
421 | { | |
422 | struct peer *peer; | |
423 | ||
424 | peer = THREAD_ARG(thread); | |
425 | peer->t_start = NULL; | |
426 | ||
427 | if (bgp_debug_neighbor_events(peer)) | |
428 | zlog_debug("%s [FSM] Timer (start timer expire).", peer->host); | |
429 | ||
430 | THREAD_VAL(thread) = BGP_Start; | |
431 | bgp_event(thread); /* bgp_event unlocks peer */ | |
432 | ||
433 | return 0; | |
434 | } | |
435 | ||
436 | /* BGP connect retry timer. */ | |
437 | static int bgp_connect_timer(struct thread *thread) | |
438 | { | |
439 | struct peer *peer; | |
440 | int ret; | |
441 | ||
442 | peer = THREAD_ARG(thread); | |
443 | ||
444 | assert(!peer->t_write); | |
445 | assert(!peer->t_read); | |
446 | ||
447 | peer->t_connect = NULL; | |
448 | ||
449 | if (bgp_debug_neighbor_events(peer)) | |
450 | zlog_debug("%s [FSM] Timer (connect timer expire)", peer->host); | |
451 | ||
452 | if (CHECK_FLAG(peer->sflags, PEER_STATUS_ACCEPT_PEER)) { | |
453 | bgp_stop(peer); | |
454 | ret = -1; | |
455 | } else { | |
456 | THREAD_VAL(thread) = ConnectRetry_timer_expired; | |
457 | bgp_event(thread); /* bgp_event unlocks peer */ | |
458 | ret = 0; | |
459 | } | |
460 | ||
461 | return ret; | |
462 | } | |
463 | ||
464 | /* BGP holdtime timer. */ | |
465 | static int bgp_holdtime_timer(struct thread *thread) | |
466 | { | |
467 | struct peer *peer; | |
468 | ||
469 | peer = THREAD_ARG(thread); | |
470 | peer->t_holdtime = NULL; | |
471 | ||
472 | if (bgp_debug_neighbor_events(peer)) | |
473 | zlog_debug("%s [FSM] Timer (holdtime timer expire)", | |
474 | peer->host); | |
475 | ||
476 | THREAD_VAL(thread) = Hold_Timer_expired; | |
477 | bgp_event(thread); /* bgp_event unlocks peer */ | |
478 | ||
479 | return 0; | |
480 | } | |
481 | ||
482 | int bgp_routeadv_timer(struct thread *thread) | |
483 | { | |
484 | struct peer *peer; | |
485 | ||
486 | peer = THREAD_ARG(thread); | |
487 | peer->t_routeadv = NULL; | |
488 | ||
489 | if (bgp_debug_neighbor_events(peer)) | |
490 | zlog_debug("%s [FSM] Timer (routeadv timer expire)", | |
491 | peer->host); | |
492 | ||
493 | peer->synctime = bgp_clock(); | |
494 | ||
495 | thread_add_timer_msec(bm->master, bgp_generate_updgrp_packets, peer, 0, | |
496 | &peer->t_generate_updgrp_packets); | |
497 | ||
498 | /* MRAI timer will be started again when FIFO is built, no need to | |
499 | * do it here. | |
500 | */ | |
501 | return 0; | |
502 | } | |
503 | ||
504 | /* BGP Peer Down Cause */ | |
505 | const char *peer_down_str[] = {"", | |
506 | "Router ID changed", | |
507 | "Remote AS changed", | |
508 | "Local AS change", | |
509 | "Cluster ID changed", | |
510 | "Confederation identifier changed", | |
511 | "Confederation peer changed", | |
512 | "RR client config change", | |
513 | "RS client config change", | |
514 | "Update source change", | |
515 | "Address family activated", | |
516 | "Admin. shutdown", | |
517 | "User reset", | |
518 | "BGP Notification received", | |
519 | "BGP Notification send", | |
520 | "Peer closed the session", | |
521 | "Neighbor deleted", | |
522 | "Peer-group add member", | |
523 | "Peer-group delete member", | |
524 | "Capability changed", | |
525 | "Passive config change", | |
526 | "Multihop config change", | |
527 | "NSF peer closed the session", | |
528 | "Intf peering v6only config change", | |
529 | "BFD down received", | |
530 | "Interface down", | |
531 | "Neighbor address lost"}; | |
532 | ||
533 | static int bgp_graceful_restart_timer_expire(struct thread *thread) | |
534 | { | |
535 | struct peer *peer; | |
536 | afi_t afi; | |
537 | safi_t safi; | |
538 | ||
539 | peer = THREAD_ARG(thread); | |
540 | peer->t_gr_restart = NULL; | |
541 | ||
542 | /* NSF delete stale route */ | |
543 | for (afi = AFI_IP; afi < AFI_MAX; afi++) | |
544 | for (safi = SAFI_UNICAST; safi <= SAFI_MPLS_VPN; safi++) | |
545 | if (peer->nsf[afi][safi]) | |
546 | bgp_clear_stale_route(peer, afi, safi); | |
547 | ||
548 | UNSET_FLAG(peer->sflags, PEER_STATUS_NSF_WAIT); | |
549 | BGP_TIMER_OFF(peer->t_gr_stale); | |
550 | ||
551 | if (bgp_debug_neighbor_events(peer)) { | |
552 | zlog_debug("%s graceful restart timer expired", peer->host); | |
553 | zlog_debug("%s graceful restart stalepath timer stopped", | |
554 | peer->host); | |
555 | } | |
556 | ||
557 | bgp_timer_set(peer); | |
558 | ||
559 | return 0; | |
560 | } | |
561 | ||
562 | static int bgp_graceful_stale_timer_expire(struct thread *thread) | |
563 | { | |
564 | struct peer *peer; | |
565 | afi_t afi; | |
566 | safi_t safi; | |
567 | ||
568 | peer = THREAD_ARG(thread); | |
569 | peer->t_gr_stale = NULL; | |
570 | ||
571 | if (bgp_debug_neighbor_events(peer)) | |
572 | zlog_debug("%s graceful restart stalepath timer expired", | |
573 | peer->host); | |
574 | ||
575 | /* NSF delete stale route */ | |
576 | for (afi = AFI_IP; afi < AFI_MAX; afi++) | |
577 | for (safi = SAFI_UNICAST; safi <= SAFI_MPLS_VPN; safi++) | |
578 | if (peer->nsf[afi][safi]) | |
579 | bgp_clear_stale_route(peer, afi, safi); | |
580 | ||
581 | return 0; | |
582 | } | |
583 | ||
584 | static int bgp_update_delay_applicable(struct bgp *bgp) | |
585 | { | |
586 | /* update_delay_over flag should be reset (set to 0) for any new | |
587 | applicability of the update-delay during BGP process lifetime. | |
588 | And it should be set after an occurence of the update-delay is | |
589 | over)*/ | |
590 | if (!bgp->update_delay_over) | |
591 | return 1; | |
592 | ||
593 | return 0; | |
594 | } | |
595 | ||
596 | int bgp_update_delay_active(struct bgp *bgp) | |
597 | { | |
598 | if (bgp->t_update_delay) | |
599 | return 1; | |
600 | ||
601 | return 0; | |
602 | } | |
603 | ||
604 | int bgp_update_delay_configured(struct bgp *bgp) | |
605 | { | |
606 | if (bgp->v_update_delay) | |
607 | return 1; | |
608 | ||
609 | return 0; | |
610 | } | |
611 | ||
612 | /* Do the post-processing needed when bgp comes out of the read-only mode | |
613 | on ending the update delay. */ | |
614 | void bgp_update_delay_end(struct bgp *bgp) | |
615 | { | |
616 | THREAD_TIMER_OFF(bgp->t_update_delay); | |
617 | THREAD_TIMER_OFF(bgp->t_establish_wait); | |
618 | ||
619 | /* Reset update-delay related state */ | |
620 | bgp->update_delay_over = 1; | |
621 | bgp->established = 0; | |
622 | bgp->restarted_peers = 0; | |
623 | bgp->implicit_eors = 0; | |
624 | bgp->explicit_eors = 0; | |
625 | ||
626 | quagga_timestamp(3, bgp->update_delay_end_time, | |
627 | sizeof(bgp->update_delay_end_time)); | |
628 | ||
629 | /* | |
630 | * Add an end-of-initial-update marker to the main process queues so | |
631 | * that | |
632 | * the route advertisement timer for the peers can be started. Also set | |
633 | * the zebra and peer update hold flags. These flags are used to achieve | |
634 | * three stages in the update-delay post processing: | |
635 | * 1. Finish best-path selection for all the prefixes held on the | |
636 | * queues. | |
637 | * (routes in BGP are updated, and peers sync queues are populated | |
638 | * too) | |
639 | * 2. As the eoiu mark is reached in the bgp process routine, ship all | |
640 | * the | |
641 | * routes to zebra. With that zebra should see updates from BGP | |
642 | * close | |
643 | * to each other. | |
644 | * 3. Unblock the peer update writes. With that peer update packing | |
645 | * with | |
646 | * the prefixes should be at its maximum. | |
647 | */ | |
648 | bgp_add_eoiu_mark(bgp); | |
649 | bgp->main_zebra_update_hold = 1; | |
650 | bgp->main_peers_update_hold = 1; | |
651 | ||
652 | /* Resume the queue processing. This should trigger the event that would | |
653 | take | |
654 | care of processing any work that was queued during the read-only | |
655 | mode. */ | |
656 | work_queue_unplug(bm->process_main_queue); | |
657 | } | |
658 | ||
659 | /** | |
660 | * see bgp_fsm.h | |
661 | */ | |
662 | void bgp_start_routeadv(struct bgp *bgp) | |
663 | { | |
664 | struct listnode *node, *nnode; | |
665 | struct peer *peer; | |
666 | ||
667 | zlog_info("bgp_start_routeadv(), update hold status %d", | |
668 | bgp->main_peers_update_hold); | |
669 | ||
670 | if (bgp->main_peers_update_hold) | |
671 | return; | |
672 | ||
673 | quagga_timestamp(3, bgp->update_delay_peers_resume_time, | |
674 | sizeof(bgp->update_delay_peers_resume_time)); | |
675 | ||
676 | for (ALL_LIST_ELEMENTS(bgp->peer, node, nnode, peer)) { | |
677 | if (peer->status != Established) | |
678 | continue; | |
679 | BGP_TIMER_OFF(peer->t_routeadv); | |
680 | BGP_TIMER_ON(peer->t_routeadv, bgp_routeadv_timer, 0); | |
681 | } | |
682 | } | |
683 | ||
684 | /** | |
685 | * see bgp_fsm.h | |
686 | */ | |
687 | void bgp_adjust_routeadv(struct peer *peer) | |
688 | { | |
689 | time_t nowtime = bgp_clock(); | |
690 | double diff; | |
691 | unsigned long remain; | |
692 | ||
693 | /* Bypass checks for special case of MRAI being 0 */ | |
694 | if (peer->v_routeadv == 0) { | |
695 | /* Stop existing timer, just in case it is running for a | |
696 | * different | |
697 | * duration and schedule write thread immediately. | |
698 | */ | |
699 | if (peer->t_routeadv) | |
700 | BGP_TIMER_OFF(peer->t_routeadv); | |
701 | ||
702 | peer->synctime = bgp_clock(); | |
703 | thread_add_timer_msec(bm->master, bgp_generate_updgrp_packets, | |
704 | peer, 0, | |
705 | &peer->t_generate_updgrp_packets); | |
706 | return; | |
707 | } | |
708 | ||
709 | ||
710 | /* | |
711 | * CASE I: | |
712 | * If the last update was written more than MRAI back, expire the timer | |
713 | * instantly so that we can send the update out sooner. | |
714 | * | |
715 | * <------- MRAI ---------> | |
716 | * |-----------------|-----------------------| | |
717 | * <------------- m ------------> | |
718 | * ^ ^ ^ | |
719 | * | | | | |
720 | * | | current time | |
721 | * | timer start | |
722 | * last write | |
723 | * | |
724 | * m > MRAI | |
725 | */ | |
726 | diff = difftime(nowtime, peer->last_update); | |
727 | if (diff > (double)peer->v_routeadv) { | |
728 | BGP_TIMER_OFF(peer->t_routeadv); | |
729 | BGP_TIMER_ON(peer->t_routeadv, bgp_routeadv_timer, 0); | |
730 | return; | |
731 | } | |
732 | ||
733 | /* | |
734 | * CASE II: | |
735 | * - Find when to expire the MRAI timer. | |
736 | * If MRAI timer is not active, assume we can start it now. | |
737 | * | |
738 | * <------- MRAI ---------> | |
739 | * |------------|-----------------------| | |
740 | * <-------- m ----------><----- r -----> | |
741 | * ^ ^ ^ | |
742 | * | | | | |
743 | * | | current time | |
744 | * | timer start | |
745 | * last write | |
746 | * | |
747 | * (MRAI - m) < r | |
748 | */ | |
749 | if (peer->t_routeadv) | |
750 | remain = thread_timer_remain_second(peer->t_routeadv); | |
751 | else | |
752 | remain = peer->v_routeadv; | |
753 | diff = peer->v_routeadv - diff; | |
754 | if (diff <= (double)remain) { | |
755 | BGP_TIMER_OFF(peer->t_routeadv); | |
756 | BGP_TIMER_ON(peer->t_routeadv, bgp_routeadv_timer, diff); | |
757 | } | |
758 | } | |
759 | ||
760 | static int bgp_maxmed_onstartup_applicable(struct bgp *bgp) | |
761 | { | |
762 | if (!bgp->maxmed_onstartup_over) | |
763 | return 1; | |
764 | ||
765 | return 0; | |
766 | } | |
767 | ||
768 | int bgp_maxmed_onstartup_configured(struct bgp *bgp) | |
769 | { | |
770 | if (bgp->v_maxmed_onstartup != BGP_MAXMED_ONSTARTUP_UNCONFIGURED) | |
771 | return 1; | |
772 | ||
773 | return 0; | |
774 | } | |
775 | ||
776 | int bgp_maxmed_onstartup_active(struct bgp *bgp) | |
777 | { | |
778 | if (bgp->t_maxmed_onstartup) | |
779 | return 1; | |
780 | ||
781 | return 0; | |
782 | } | |
783 | ||
784 | void bgp_maxmed_update(struct bgp *bgp) | |
785 | { | |
786 | uint8_t maxmed_active; | |
787 | uint32_t maxmed_value; | |
788 | ||
789 | if (bgp->v_maxmed_admin) { | |
790 | maxmed_active = 1; | |
791 | maxmed_value = bgp->maxmed_admin_value; | |
792 | } else if (bgp->t_maxmed_onstartup) { | |
793 | maxmed_active = 1; | |
794 | maxmed_value = bgp->maxmed_onstartup_value; | |
795 | } else { | |
796 | maxmed_active = 0; | |
797 | maxmed_value = BGP_MAXMED_VALUE_DEFAULT; | |
798 | } | |
799 | ||
800 | if (bgp->maxmed_active != maxmed_active | |
801 | || bgp->maxmed_value != maxmed_value) { | |
802 | bgp->maxmed_active = maxmed_active; | |
803 | bgp->maxmed_value = maxmed_value; | |
804 | ||
805 | update_group_announce(bgp); | |
806 | } | |
807 | } | |
808 | ||
809 | /* The maxmed onstartup timer expiry callback. */ | |
810 | static int bgp_maxmed_onstartup_timer(struct thread *thread) | |
811 | { | |
812 | struct bgp *bgp; | |
813 | ||
814 | zlog_info("Max med on startup ended - timer expired."); | |
815 | ||
816 | bgp = THREAD_ARG(thread); | |
817 | THREAD_TIMER_OFF(bgp->t_maxmed_onstartup); | |
818 | bgp->maxmed_onstartup_over = 1; | |
819 | ||
820 | bgp_maxmed_update(bgp); | |
821 | ||
822 | return 0; | |
823 | } | |
824 | ||
825 | static void bgp_maxmed_onstartup_begin(struct bgp *bgp) | |
826 | { | |
827 | /* Applicable only once in the process lifetime on the startup */ | |
828 | if (bgp->maxmed_onstartup_over) | |
829 | return; | |
830 | ||
831 | zlog_info("Begin maxmed onstartup mode - timer %d seconds", | |
832 | bgp->v_maxmed_onstartup); | |
833 | ||
834 | thread_add_timer(bm->master, bgp_maxmed_onstartup_timer, bgp, | |
835 | bgp->v_maxmed_onstartup, &bgp->t_maxmed_onstartup); | |
836 | ||
837 | if (!bgp->v_maxmed_admin) { | |
838 | bgp->maxmed_active = 1; | |
839 | bgp->maxmed_value = bgp->maxmed_onstartup_value; | |
840 | } | |
841 | ||
842 | /* Route announce to all peers should happen after this in | |
843 | * bgp_establish() */ | |
844 | } | |
845 | ||
846 | static void bgp_maxmed_onstartup_process_status_change(struct peer *peer) | |
847 | { | |
848 | if (peer->status == Established && !peer->bgp->established) { | |
849 | bgp_maxmed_onstartup_begin(peer->bgp); | |
850 | } | |
851 | } | |
852 | ||
853 | /* The update delay timer expiry callback. */ | |
854 | static int bgp_update_delay_timer(struct thread *thread) | |
855 | { | |
856 | struct bgp *bgp; | |
857 | ||
858 | zlog_info("Update delay ended - timer expired."); | |
859 | ||
860 | bgp = THREAD_ARG(thread); | |
861 | THREAD_TIMER_OFF(bgp->t_update_delay); | |
862 | bgp_update_delay_end(bgp); | |
863 | ||
864 | return 0; | |
865 | } | |
866 | ||
867 | /* The establish wait timer expiry callback. */ | |
868 | static int bgp_establish_wait_timer(struct thread *thread) | |
869 | { | |
870 | struct bgp *bgp; | |
871 | ||
872 | zlog_info("Establish wait - timer expired."); | |
873 | ||
874 | bgp = THREAD_ARG(thread); | |
875 | THREAD_TIMER_OFF(bgp->t_establish_wait); | |
876 | bgp_check_update_delay(bgp); | |
877 | ||
878 | return 0; | |
879 | } | |
880 | ||
881 | /* Steps to begin the update delay: | |
882 | - initialize queues if needed | |
883 | - stop the queue processing | |
884 | - start the timer */ | |
885 | static void bgp_update_delay_begin(struct bgp *bgp) | |
886 | { | |
887 | struct listnode *node, *nnode; | |
888 | struct peer *peer; | |
889 | ||
890 | /* Stop the processing of queued work. Enqueue shall continue */ | |
891 | work_queue_plug(bm->process_main_queue); | |
892 | ||
893 | for (ALL_LIST_ELEMENTS(bgp->peer, node, nnode, peer)) | |
894 | peer->update_delay_over = 0; | |
895 | ||
896 | /* Start the update-delay timer */ | |
897 | thread_add_timer(bm->master, bgp_update_delay_timer, bgp, | |
898 | bgp->v_update_delay, &bgp->t_update_delay); | |
899 | ||
900 | if (bgp->v_establish_wait != bgp->v_update_delay) | |
901 | thread_add_timer(bm->master, bgp_establish_wait_timer, bgp, | |
902 | bgp->v_establish_wait, &bgp->t_establish_wait); | |
903 | ||
904 | quagga_timestamp(3, bgp->update_delay_begin_time, | |
905 | sizeof(bgp->update_delay_begin_time)); | |
906 | } | |
907 | ||
908 | static void bgp_update_delay_process_status_change(struct peer *peer) | |
909 | { | |
910 | if (peer->status == Established) { | |
911 | if (!peer->bgp->established++) { | |
912 | bgp_update_delay_begin(peer->bgp); | |
913 | zlog_info( | |
914 | "Begin read-only mode - update-delay timer %d seconds", | |
915 | peer->bgp->v_update_delay); | |
916 | } | |
917 | if (CHECK_FLAG(peer->cap, PEER_CAP_RESTART_BIT_RCV)) | |
918 | bgp_update_restarted_peers(peer); | |
919 | } | |
920 | if (peer->ostatus == Established | |
921 | && bgp_update_delay_active(peer->bgp)) { | |
922 | /* Adjust the update-delay state to account for this flap. | |
923 | NOTE: Intentionally skipping adjusting implicit_eors or | |
924 | explicit_eors | |
925 | counters. Extra sanity check in bgp_check_update_delay() | |
926 | should | |
927 | be enough to take care of any additive discrepancy in bgp eor | |
928 | counters */ | |
929 | peer->bgp->established--; | |
930 | peer->update_delay_over = 0; | |
931 | } | |
932 | } | |
933 | ||
934 | /* Called after event occurred, this function change status and reset | |
935 | read/write and timer thread. */ | |
936 | void bgp_fsm_change_status(struct peer *peer, int status) | |
937 | { | |
938 | struct bgp *bgp; | |
939 | uint32_t peer_count; | |
940 | ||
941 | bgp_dump_state(peer, peer->status, status); | |
942 | ||
943 | bgp = peer->bgp; | |
944 | peer_count = bgp->established_peers; | |
945 | ||
946 | if (status == Established) | |
947 | bgp->established_peers++; | |
948 | else if ((peer->status == Established) && (status != Established)) | |
949 | bgp->established_peers--; | |
950 | ||
951 | if (BGP_DEBUG(neighbor_events, NEIGHBOR_EVENTS)) | |
952 | zlog_debug("%s : vrf %u, established_peers %u", __func__, | |
953 | bgp->vrf_id, bgp->established_peers); | |
954 | /* Set to router ID to the value provided by RIB if there are no peers | |
955 | * in the established state and peer count did not change | |
956 | */ | |
957 | if ((peer_count != bgp->established_peers) && | |
958 | (bgp->established_peers == 0)) | |
959 | bgp_router_id_zebra_bump(bgp->vrf_id, NULL); | |
960 | ||
961 | /* Transition into Clearing or Deleted must /always/ clear all routes.. | |
962 | * (and must do so before actually changing into Deleted.. | |
963 | */ | |
964 | if (status >= Clearing) { | |
965 | bgp_clear_route_all(peer); | |
966 | ||
967 | /* If no route was queued for the clear-node processing, | |
968 | * generate the | |
969 | * completion event here. This is needed because if there are no | |
970 | * routes | |
971 | * to trigger the background clear-node thread, the event won't | |
972 | * get | |
973 | * generated and the peer would be stuck in Clearing. Note that | |
974 | * this | |
975 | * event is for the peer and helps the peer transition out of | |
976 | * Clearing | |
977 | * state; it should not be generated per (AFI,SAFI). The event | |
978 | * is | |
979 | * directly posted here without calling clear_node_complete() as | |
980 | * we | |
981 | * shouldn't do an extra unlock. This event will get processed | |
982 | * after | |
983 | * the state change that happens below, so peer will be in | |
984 | * Clearing | |
985 | * (or Deleted). | |
986 | */ | |
987 | if (!work_queue_is_scheduled(peer->clear_node_queue)) | |
988 | BGP_EVENT_ADD(peer, Clearing_Completed); | |
989 | } | |
990 | ||
991 | /* Preserve old status and change into new status. */ | |
992 | peer->ostatus = peer->status; | |
993 | peer->status = status; | |
994 | ||
995 | /* Save event that caused status change. */ | |
996 | peer->last_major_event = peer->cur_event; | |
997 | ||
998 | if (status == Established) | |
999 | UNSET_FLAG(peer->sflags, PEER_STATUS_ACCEPT_PEER); | |
1000 | ||
1001 | /* If max-med processing is applicable, do the necessary. */ | |
1002 | if (status == Established) { | |
1003 | if (bgp_maxmed_onstartup_configured(peer->bgp) | |
1004 | && bgp_maxmed_onstartup_applicable(peer->bgp)) | |
1005 | bgp_maxmed_onstartup_process_status_change(peer); | |
1006 | else | |
1007 | peer->bgp->maxmed_onstartup_over = 1; | |
1008 | } | |
1009 | ||
1010 | /* If update-delay processing is applicable, do the necessary. */ | |
1011 | if (bgp_update_delay_configured(peer->bgp) | |
1012 | && bgp_update_delay_applicable(peer->bgp)) | |
1013 | bgp_update_delay_process_status_change(peer); | |
1014 | ||
1015 | if (bgp_debug_neighbor_events(peer)) | |
1016 | zlog_debug("%s went from %s to %s", peer->host, | |
1017 | lookup_msg(bgp_status_msg, peer->ostatus, NULL), | |
1018 | lookup_msg(bgp_status_msg, peer->status, NULL)); | |
1019 | } | |
1020 | ||
1021 | /* Flush the event queue and ensure the peer is shut down */ | |
1022 | static int bgp_clearing_completed(struct peer *peer) | |
1023 | { | |
1024 | int rc = bgp_stop(peer); | |
1025 | ||
1026 | if (rc >= 0) | |
1027 | BGP_EVENT_FLUSH(peer); | |
1028 | ||
1029 | return rc; | |
1030 | } | |
1031 | ||
1032 | /* Administrative BGP peer stop event. */ | |
1033 | /* May be called multiple times for the same peer */ | |
1034 | int bgp_stop(struct peer *peer) | |
1035 | { | |
1036 | afi_t afi; | |
1037 | safi_t safi; | |
1038 | char orf_name[BUFSIZ]; | |
1039 | int ret = 0; | |
1040 | ||
1041 | if (peer_dynamic_neighbor(peer) | |
1042 | && !(CHECK_FLAG(peer->flags, PEER_FLAG_DELETE))) { | |
1043 | if (bgp_debug_neighbor_events(peer)) | |
1044 | zlog_debug("%s (dynamic neighbor) deleted", peer->host); | |
1045 | peer_delete(peer); | |
1046 | return -1; | |
1047 | } | |
1048 | ||
1049 | /* Can't do this in Clearing; events are used for state transitions */ | |
1050 | if (peer->status != Clearing) { | |
1051 | /* Delete all existing events of the peer */ | |
1052 | BGP_EVENT_FLUSH(peer); | |
1053 | } | |
1054 | ||
1055 | /* Increment Dropped count. */ | |
1056 | if (peer->status == Established) { | |
1057 | peer->dropped++; | |
1058 | ||
1059 | /* bgp log-neighbor-changes of neighbor Down */ | |
1060 | if (bgp_flag_check(peer->bgp, BGP_FLAG_LOG_NEIGHBOR_CHANGES)) { | |
1061 | struct vrf *vrf = vrf_lookup_by_id(peer->bgp->vrf_id); | |
1062 | zlog_info( | |
1063 | "%%ADJCHANGE: neighbor %s(%s) in vrf %s Down %s", | |
1064 | peer->host, | |
1065 | (peer->hostname) ? peer->hostname : "Unknown", | |
1066 | vrf ? ((vrf->vrf_id != VRF_DEFAULT) | |
1067 | ? vrf->name | |
1068 | : VRF_DEFAULT_NAME) | |
1069 | : "", | |
1070 | peer_down_str[(int)peer->last_reset]); | |
1071 | } | |
1072 | ||
1073 | /* graceful restart */ | |
1074 | if (peer->t_gr_stale) { | |
1075 | BGP_TIMER_OFF(peer->t_gr_stale); | |
1076 | if (bgp_debug_neighbor_events(peer)) | |
1077 | zlog_debug( | |
1078 | "%s graceful restart stalepath timer stopped", | |
1079 | peer->host); | |
1080 | } | |
1081 | if (CHECK_FLAG(peer->sflags, PEER_STATUS_NSF_WAIT)) { | |
1082 | if (bgp_debug_neighbor_events(peer)) { | |
1083 | zlog_debug( | |
1084 | "%s graceful restart timer started for %d sec", | |
1085 | peer->host, peer->v_gr_restart); | |
1086 | zlog_debug( | |
1087 | "%s graceful restart stalepath timer started for %d sec", | |
1088 | peer->host, peer->bgp->stalepath_time); | |
1089 | } | |
1090 | BGP_TIMER_ON(peer->t_gr_restart, | |
1091 | bgp_graceful_restart_timer_expire, | |
1092 | peer->v_gr_restart); | |
1093 | BGP_TIMER_ON(peer->t_gr_stale, | |
1094 | bgp_graceful_stale_timer_expire, | |
1095 | peer->bgp->stalepath_time); | |
1096 | } else { | |
1097 | UNSET_FLAG(peer->sflags, PEER_STATUS_NSF_MODE); | |
1098 | ||
1099 | for (afi = AFI_IP; afi < AFI_MAX; afi++) | |
1100 | for (safi = SAFI_UNICAST; safi <= SAFI_MPLS_VPN; | |
1101 | safi++) | |
1102 | peer->nsf[afi][safi] = 0; | |
1103 | } | |
1104 | ||
1105 | /* set last reset time */ | |
1106 | peer->resettime = peer->uptime = bgp_clock(); | |
1107 | ||
1108 | if (BGP_DEBUG(update_groups, UPDATE_GROUPS)) | |
1109 | zlog_debug("%s remove from all update group", | |
1110 | peer->host); | |
1111 | update_group_remove_peer_afs(peer); | |
1112 | ||
1113 | hook_call(peer_backward_transition, peer); | |
1114 | ||
1115 | /* Reset peer synctime */ | |
1116 | peer->synctime = 0; | |
1117 | ||
1118 | bgp_bfd_deregister_peer(peer); | |
1119 | } | |
1120 | ||
1121 | /* stop keepalives */ | |
1122 | bgp_keepalives_off(peer); | |
1123 | ||
1124 | /* Stop read and write threads. */ | |
1125 | bgp_writes_off(peer); | |
1126 | bgp_reads_off(peer); | |
1127 | ||
1128 | THREAD_OFF(peer->t_connect_check_r); | |
1129 | THREAD_OFF(peer->t_connect_check_w); | |
1130 | ||
1131 | /* Stop all timers. */ | |
1132 | BGP_TIMER_OFF(peer->t_start); | |
1133 | BGP_TIMER_OFF(peer->t_connect); | |
1134 | BGP_TIMER_OFF(peer->t_holdtime); | |
1135 | BGP_TIMER_OFF(peer->t_routeadv); | |
1136 | ||
1137 | /* Clear input and output buffer. */ | |
1138 | pthread_mutex_lock(&peer->io_mtx); | |
1139 | { | |
1140 | if (peer->ibuf) | |
1141 | stream_fifo_clean(peer->ibuf); | |
1142 | if (peer->obuf) | |
1143 | stream_fifo_clean(peer->obuf); | |
1144 | ||
1145 | if (peer->ibuf_work) | |
1146 | ringbuf_wipe(peer->ibuf_work); | |
1147 | if (peer->obuf_work) | |
1148 | stream_reset(peer->obuf_work); | |
1149 | ||
1150 | if (peer->curr) { | |
1151 | stream_free(peer->curr); | |
1152 | peer->curr = NULL; | |
1153 | } | |
1154 | } | |
1155 | pthread_mutex_unlock(&peer->io_mtx); | |
1156 | ||
1157 | /* Close of file descriptor. */ | |
1158 | if (peer->fd >= 0) { | |
1159 | close(peer->fd); | |
1160 | peer->fd = -1; | |
1161 | } | |
1162 | ||
1163 | FOREACH_AFI_SAFI (afi, safi) { | |
1164 | /* Reset all negotiated variables */ | |
1165 | peer->afc_nego[afi][safi] = 0; | |
1166 | peer->afc_adv[afi][safi] = 0; | |
1167 | peer->afc_recv[afi][safi] = 0; | |
1168 | ||
1169 | /* peer address family capability flags*/ | |
1170 | peer->af_cap[afi][safi] = 0; | |
1171 | ||
1172 | /* peer address family status flags*/ | |
1173 | peer->af_sflags[afi][safi] = 0; | |
1174 | ||
1175 | /* Received ORF prefix-filter */ | |
1176 | peer->orf_plist[afi][safi] = NULL; | |
1177 | ||
1178 | if ((peer->status == OpenConfirm) | |
1179 | || (peer->status == Established)) { | |
1180 | /* ORF received prefix-filter pnt */ | |
1181 | sprintf(orf_name, "%s.%d.%d", peer->host, afi, safi); | |
1182 | prefix_bgp_orf_remove_all(afi, orf_name); | |
1183 | } | |
1184 | } | |
1185 | ||
1186 | /* Reset keepalive and holdtime */ | |
1187 | if (CHECK_FLAG(peer->flags, PEER_FLAG_TIMER)) { | |
1188 | peer->v_keepalive = peer->keepalive; | |
1189 | peer->v_holdtime = peer->holdtime; | |
1190 | } else { | |
1191 | peer->v_keepalive = peer->bgp->default_keepalive; | |
1192 | peer->v_holdtime = peer->bgp->default_holdtime; | |
1193 | } | |
1194 | ||
1195 | peer->update_time = 0; | |
1196 | ||
1197 | /* Until we are sure that there is no problem about prefix count | |
1198 | this should be commented out.*/ | |
1199 | #if 0 | |
1200 | /* Reset prefix count */ | |
1201 | peer->pcount[AFI_IP][SAFI_UNICAST] = 0; | |
1202 | peer->pcount[AFI_IP][SAFI_MULTICAST] = 0; | |
1203 | peer->pcount[AFI_IP][SAFI_LABELED_UNICAST] = 0; | |
1204 | peer->pcount[AFI_IP][SAFI_MPLS_VPN] = 0; | |
1205 | peer->pcount[AFI_IP6][SAFI_UNICAST] = 0; | |
1206 | peer->pcount[AFI_IP6][SAFI_MULTICAST] = 0; | |
1207 | peer->pcount[AFI_IP6][SAFI_LABELED_UNICAST] = 0; | |
1208 | #endif /* 0 */ | |
1209 | ||
1210 | if (!CHECK_FLAG(peer->flags, PEER_FLAG_CONFIG_NODE) | |
1211 | && !(CHECK_FLAG(peer->flags, PEER_FLAG_DELETE))) { | |
1212 | peer_delete(peer); | |
1213 | ret = -1; | |
1214 | } else { | |
1215 | bgp_peer_conf_if_to_su_update(peer); | |
1216 | } | |
1217 | ||
1218 | return ret; | |
1219 | } | |
1220 | ||
1221 | /* BGP peer is stoped by the error. */ | |
1222 | static int bgp_stop_with_error(struct peer *peer) | |
1223 | { | |
1224 | /* Double start timer. */ | |
1225 | peer->v_start *= 2; | |
1226 | ||
1227 | /* Overflow check. */ | |
1228 | if (peer->v_start >= (60 * 2)) | |
1229 | peer->v_start = (60 * 2); | |
1230 | ||
1231 | if (peer_dynamic_neighbor(peer)) { | |
1232 | if (bgp_debug_neighbor_events(peer)) | |
1233 | zlog_debug("%s (dynamic neighbor) deleted", peer->host); | |
1234 | peer_delete(peer); | |
1235 | return -1; | |
1236 | } | |
1237 | ||
1238 | return (bgp_stop(peer)); | |
1239 | } | |
1240 | ||
1241 | ||
1242 | /* something went wrong, send notify and tear down */ | |
1243 | static int bgp_stop_with_notify(struct peer *peer, uint8_t code, | |
1244 | uint8_t sub_code) | |
1245 | { | |
1246 | /* Send notify to remote peer */ | |
1247 | bgp_notify_send(peer, code, sub_code); | |
1248 | ||
1249 | if (peer_dynamic_neighbor(peer)) { | |
1250 | if (bgp_debug_neighbor_events(peer)) | |
1251 | zlog_debug("%s (dynamic neighbor) deleted", peer->host); | |
1252 | peer_delete(peer); | |
1253 | return -1; | |
1254 | } | |
1255 | ||
1256 | /* Clear start timer value to default. */ | |
1257 | peer->v_start = BGP_INIT_START_TIMER; | |
1258 | ||
1259 | return (bgp_stop(peer)); | |
1260 | } | |
1261 | ||
1262 | /** | |
1263 | * Determines whether a TCP session has successfully established for a peer and | |
1264 | * events as appropriate. | |
1265 | * | |
1266 | * This function is called when setting up a new session. After connect() is | |
1267 | * called on the peer's socket (in bgp_start()), the fd is passed to poll() | |
1268 | * to wait for connection success or failure. When poll() returns, this | |
1269 | * function is called to evaluate the result. | |
1270 | * | |
1271 | * Due to differences in behavior of poll() on Linux and BSD - specifically, | |
1272 | * the value of .revents in the case of a closed connection - this function is | |
1273 | * scheduled both for a read and a write event. The write event is triggered | |
1274 | * when the connection is established. A read event is triggered when the | |
1275 | * connection is closed. Thus we need to cancel whichever one did not occur. | |
1276 | */ | |
1277 | static int bgp_connect_check(struct thread *thread) | |
1278 | { | |
1279 | int status; | |
1280 | socklen_t slen; | |
1281 | int ret; | |
1282 | struct peer *peer; | |
1283 | ||
1284 | peer = THREAD_ARG(thread); | |
1285 | assert(!CHECK_FLAG(peer->thread_flags, PEER_THREAD_READS_ON)); | |
1286 | assert(!CHECK_FLAG(peer->thread_flags, PEER_THREAD_WRITES_ON)); | |
1287 | assert(!peer->t_read); | |
1288 | assert(!peer->t_write); | |
1289 | ||
1290 | THREAD_OFF(peer->t_connect_check_r); | |
1291 | THREAD_OFF(peer->t_connect_check_w); | |
1292 | ||
1293 | /* Check file descriptor. */ | |
1294 | slen = sizeof(status); | |
1295 | ret = getsockopt(peer->fd, SOL_SOCKET, SO_ERROR, (void *)&status, | |
1296 | &slen); | |
1297 | ||
1298 | /* If getsockopt is fail, this is fatal error. */ | |
1299 | if (ret < 0) { | |
1300 | zlog_info("can't get sockopt for nonblocking connect: %d(%s)", | |
1301 | errno, safe_strerror(errno)); | |
1302 | BGP_EVENT_ADD(peer, TCP_fatal_error); | |
1303 | return -1; | |
1304 | } | |
1305 | ||
1306 | /* When status is 0 then TCP connection is established. */ | |
1307 | if (status == 0) { | |
1308 | BGP_EVENT_ADD(peer, TCP_connection_open); | |
1309 | return 1; | |
1310 | } else { | |
1311 | if (bgp_debug_neighbor_events(peer)) | |
1312 | zlog_debug("%s [Event] Connect failed %d(%s)", | |
1313 | peer->host, status, safe_strerror(status)); | |
1314 | BGP_EVENT_ADD(peer, TCP_connection_open_failed); | |
1315 | return 0; | |
1316 | } | |
1317 | } | |
1318 | ||
1319 | /* TCP connection open. Next we send open message to remote peer. And | |
1320 | add read thread for reading open message. */ | |
1321 | static int bgp_connect_success(struct peer *peer) | |
1322 | { | |
1323 | if (peer->fd < 0) { | |
1324 | flog_err(EC_BGP_CONNECT, | |
1325 | "bgp_connect_success peer's fd is negative value %d", | |
1326 | peer->fd); | |
1327 | bgp_stop(peer); | |
1328 | return -1; | |
1329 | } | |
1330 | ||
1331 | if (bgp_getsockname(peer) < 0) { | |
1332 | flog_err_sys(EC_LIB_SOCKET, | |
1333 | "%s: bgp_getsockname(): failed for peer %s, fd %d", | |
1334 | __FUNCTION__, peer->host, peer->fd); | |
1335 | bgp_notify_send(peer, BGP_NOTIFY_FSM_ERR, | |
1336 | 0); /* internal error */ | |
1337 | bgp_writes_on(peer); | |
1338 | return -1; | |
1339 | } | |
1340 | ||
1341 | bgp_reads_on(peer); | |
1342 | ||
1343 | if (bgp_debug_neighbor_events(peer)) { | |
1344 | char buf1[SU_ADDRSTRLEN]; | |
1345 | ||
1346 | if (!CHECK_FLAG(peer->sflags, PEER_STATUS_ACCEPT_PEER)) | |
1347 | zlog_debug("%s open active, local address %s", | |
1348 | peer->host, | |
1349 | sockunion2str(peer->su_local, buf1, | |
1350 | SU_ADDRSTRLEN)); | |
1351 | else | |
1352 | zlog_debug("%s passive open", peer->host); | |
1353 | } | |
1354 | ||
1355 | bgp_open_send(peer); | |
1356 | ||
1357 | return 0; | |
1358 | } | |
1359 | ||
1360 | /* TCP connect fail */ | |
1361 | static int bgp_connect_fail(struct peer *peer) | |
1362 | { | |
1363 | if (peer_dynamic_neighbor(peer)) { | |
1364 | if (bgp_debug_neighbor_events(peer)) | |
1365 | zlog_debug("%s (dynamic neighbor) deleted", peer->host); | |
1366 | peer_delete(peer); | |
1367 | return -1; | |
1368 | } | |
1369 | ||
1370 | return (bgp_stop(peer)); | |
1371 | } | |
1372 | ||
1373 | /* This function is the first starting point of all BGP connection. It | |
1374 | try to connect to remote peer with non-blocking IO. */ | |
1375 | int bgp_start(struct peer *peer) | |
1376 | { | |
1377 | int status; | |
1378 | int connected = 0; | |
1379 | ||
1380 | bgp_peer_conf_if_to_su_update(peer); | |
1381 | ||
1382 | if (peer->su.sa.sa_family == AF_UNSPEC) { | |
1383 | if (bgp_debug_neighbor_events(peer)) | |
1384 | zlog_debug( | |
1385 | "%s [FSM] Unable to get neighbor's IP address, waiting...", | |
1386 | peer->host); | |
1387 | return -1; | |
1388 | } | |
1389 | ||
1390 | if (BGP_PEER_START_SUPPRESSED(peer)) { | |
1391 | if (bgp_debug_neighbor_events(peer)) | |
1392 | flog_err(EC_BGP_FSM, | |
1393 | "%s [FSM] Trying to start suppressed peer" | |
1394 | " - this is never supposed to happen!", | |
1395 | peer->host); | |
1396 | return -1; | |
1397 | } | |
1398 | ||
1399 | /* Scrub some information that might be left over from a previous, | |
1400 | * session | |
1401 | */ | |
1402 | /* Connection information. */ | |
1403 | if (peer->su_local) { | |
1404 | sockunion_free(peer->su_local); | |
1405 | peer->su_local = NULL; | |
1406 | } | |
1407 | ||
1408 | if (peer->su_remote) { | |
1409 | sockunion_free(peer->su_remote); | |
1410 | peer->su_remote = NULL; | |
1411 | } | |
1412 | ||
1413 | /* Clear remote router-id. */ | |
1414 | peer->remote_id.s_addr = 0; | |
1415 | ||
1416 | /* Clear peer capability flag. */ | |
1417 | peer->cap = 0; | |
1418 | ||
1419 | /* If the peer is passive mode, force to move to Active mode. */ | |
1420 | if (CHECK_FLAG(peer->flags, PEER_FLAG_PASSIVE)) { | |
1421 | BGP_EVENT_ADD(peer, TCP_connection_open_failed); | |
1422 | return 0; | |
1423 | } | |
1424 | ||
1425 | if (peer->bgp->vrf_id == VRF_UNKNOWN) { | |
1426 | if (bgp_debug_neighbor_events(peer)) | |
1427 | flog_err( | |
1428 | EC_BGP_FSM, | |
1429 | "%s [FSM] In a VRF that is not initialised yet", | |
1430 | peer->host); | |
1431 | return -1; | |
1432 | } | |
1433 | ||
1434 | /* Register to be notified on peer up */ | |
1435 | if (peer->sort == BGP_PEER_EBGP && peer->ttl == 1 | |
1436 | && !CHECK_FLAG(peer->flags, PEER_FLAG_DISABLE_CONNECTED_CHECK) | |
1437 | && !bgp_flag_check(peer->bgp, BGP_FLAG_DISABLE_NH_CONNECTED_CHK)) | |
1438 | connected = 1; | |
1439 | else | |
1440 | connected = 0; | |
1441 | ||
1442 | if (!bgp_find_or_add_nexthop(peer->bgp, peer->bgp, | |
1443 | family2afi(peer->su.sa.sa_family), NULL, | |
1444 | peer, connected)) { | |
1445 | if (bgp_zebra_num_connects()) { | |
1446 | if (bgp_debug_neighbor_events(peer)) | |
1447 | zlog_debug("%s [FSM] Waiting for NHT", | |
1448 | peer->host); | |
1449 | ||
1450 | BGP_EVENT_ADD(peer, TCP_connection_open_failed); | |
1451 | return 0; | |
1452 | } | |
1453 | } | |
1454 | ||
1455 | assert(!peer->t_write); | |
1456 | assert(!peer->t_read); | |
1457 | assert(!CHECK_FLAG(peer->thread_flags, PEER_THREAD_WRITES_ON)); | |
1458 | assert(!CHECK_FLAG(peer->thread_flags, PEER_THREAD_READS_ON)); | |
1459 | status = bgp_connect(peer); | |
1460 | ||
1461 | switch (status) { | |
1462 | case connect_error: | |
1463 | if (bgp_debug_neighbor_events(peer)) | |
1464 | zlog_debug("%s [FSM] Connect error", peer->host); | |
1465 | BGP_EVENT_ADD(peer, TCP_connection_open_failed); | |
1466 | break; | |
1467 | case connect_success: | |
1468 | if (bgp_debug_neighbor_events(peer)) | |
1469 | zlog_debug( | |
1470 | "%s [FSM] Connect immediately success, fd %d", | |
1471 | peer->host, peer->fd); | |
1472 | BGP_EVENT_ADD(peer, TCP_connection_open); | |
1473 | break; | |
1474 | case connect_in_progress: | |
1475 | /* To check nonblocking connect, we wait until socket is | |
1476 | readable or writable. */ | |
1477 | if (bgp_debug_neighbor_events(peer)) | |
1478 | zlog_debug( | |
1479 | "%s [FSM] Non blocking connect waiting result, fd %d", | |
1480 | peer->host, peer->fd); | |
1481 | if (peer->fd < 0) { | |
1482 | flog_err(EC_BGP_FSM, | |
1483 | "bgp_start peer's fd is negative value %d", | |
1484 | peer->fd); | |
1485 | return -1; | |
1486 | } | |
1487 | /* | |
1488 | * - when the socket becomes ready, poll() will signify POLLOUT | |
1489 | * - if it fails to connect, poll() will signify POLLHUP | |
1490 | * - POLLHUP is handled as a 'read' event by thread.c | |
1491 | * | |
1492 | * therefore, we schedule both a read and a write event with | |
1493 | * bgp_connect_check() as the handler for each and cancel the | |
1494 | * unused event in that function. | |
1495 | */ | |
1496 | thread_add_read(bm->master, bgp_connect_check, peer, peer->fd, | |
1497 | &peer->t_connect_check_r); | |
1498 | thread_add_write(bm->master, bgp_connect_check, peer, peer->fd, | |
1499 | &peer->t_connect_check_w); | |
1500 | break; | |
1501 | } | |
1502 | return 0; | |
1503 | } | |
1504 | ||
1505 | /* Connect retry timer is expired when the peer status is Connect. */ | |
1506 | static int bgp_reconnect(struct peer *peer) | |
1507 | { | |
1508 | if (bgp_stop(peer) < 0) | |
1509 | return -1; | |
1510 | ||
1511 | bgp_start(peer); | |
1512 | return 0; | |
1513 | } | |
1514 | ||
1515 | static int bgp_fsm_open(struct peer *peer) | |
1516 | { | |
1517 | /* Send keepalive and make keepalive timer */ | |
1518 | bgp_keepalive_send(peer); | |
1519 | ||
1520 | /* Reset holdtimer value. */ | |
1521 | BGP_TIMER_OFF(peer->t_holdtime); | |
1522 | ||
1523 | return 0; | |
1524 | } | |
1525 | ||
1526 | /* FSM error, unexpected event. This is error of BGP connection. So cut the | |
1527 | peer and change to Idle status. */ | |
1528 | static int bgp_fsm_event_error(struct peer *peer) | |
1529 | { | |
1530 | flog_err(EC_BGP_FSM, "%s [FSM] unexpected packet received in state %s", | |
1531 | peer->host, lookup_msg(bgp_status_msg, peer->status, NULL)); | |
1532 | ||
1533 | return bgp_stop_with_notify(peer, BGP_NOTIFY_FSM_ERR, 0); | |
1534 | } | |
1535 | ||
1536 | /* Hold timer expire. This is error of BGP connection. So cut the | |
1537 | peer and change to Idle status. */ | |
1538 | static int bgp_fsm_holdtime_expire(struct peer *peer) | |
1539 | { | |
1540 | if (bgp_debug_neighbor_events(peer)) | |
1541 | zlog_debug("%s [FSM] Hold timer expire", peer->host); | |
1542 | ||
1543 | return bgp_stop_with_notify(peer, BGP_NOTIFY_HOLD_ERR, 0); | |
1544 | } | |
1545 | ||
1546 | /** | |
1547 | * Transition to Established state. | |
1548 | * | |
1549 | * Convert peer from stub to full fledged peer, set some timers, and generate | |
1550 | * initial updates. | |
1551 | */ | |
1552 | static int bgp_establish(struct peer *peer) | |
1553 | { | |
1554 | afi_t afi; | |
1555 | safi_t safi; | |
1556 | int nsf_af_count = 0; | |
1557 | int ret = 0; | |
1558 | struct peer *other; | |
1559 | ||
1560 | other = peer->doppelganger; | |
1561 | peer = peer_xfer_conn(peer); | |
1562 | if (!peer) { | |
1563 | flog_err(EC_BGP_CONNECT, "%%Neighbor failed in xfer_conn"); | |
1564 | return -1; | |
1565 | } | |
1566 | ||
1567 | if (other == peer) | |
1568 | ret = 1; /* bgp_establish specific code when xfer_conn | |
1569 | happens. */ | |
1570 | ||
1571 | /* Reset capability open status flag. */ | |
1572 | if (!CHECK_FLAG(peer->sflags, PEER_STATUS_CAPABILITY_OPEN)) | |
1573 | SET_FLAG(peer->sflags, PEER_STATUS_CAPABILITY_OPEN); | |
1574 | ||
1575 | /* Clear start timer value to default. */ | |
1576 | peer->v_start = BGP_INIT_START_TIMER; | |
1577 | ||
1578 | /* Increment established count. */ | |
1579 | peer->established++; | |
1580 | bgp_fsm_change_status(peer, Established); | |
1581 | ||
1582 | /* bgp log-neighbor-changes of neighbor Up */ | |
1583 | if (bgp_flag_check(peer->bgp, BGP_FLAG_LOG_NEIGHBOR_CHANGES)) { | |
1584 | struct vrf *vrf = vrf_lookup_by_id(peer->bgp->vrf_id); | |
1585 | zlog_info("%%ADJCHANGE: neighbor %s(%s) in vrf %s Up", | |
1586 | peer->host, | |
1587 | (peer->hostname) ? peer->hostname : "Unknown", | |
1588 | vrf ? ((vrf->vrf_id != VRF_DEFAULT) | |
1589 | ? vrf->name | |
1590 | : VRF_DEFAULT_NAME) | |
1591 | : ""); | |
1592 | } | |
1593 | /* assign update-group/subgroup */ | |
1594 | update_group_adjust_peer_afs(peer); | |
1595 | ||
1596 | /* graceful restart */ | |
1597 | UNSET_FLAG(peer->sflags, PEER_STATUS_NSF_WAIT); | |
1598 | for (afi = AFI_IP; afi < AFI_MAX; afi++) | |
1599 | for (safi = SAFI_UNICAST; safi <= SAFI_MPLS_VPN; safi++) { | |
1600 | if (peer->afc_nego[afi][safi] | |
1601 | && CHECK_FLAG(peer->cap, PEER_CAP_RESTART_ADV) | |
1602 | && CHECK_FLAG(peer->af_cap[afi][safi], | |
1603 | PEER_CAP_RESTART_AF_RCV)) { | |
1604 | if (peer->nsf[afi][safi] | |
1605 | && !CHECK_FLAG( | |
1606 | peer->af_cap[afi][safi], | |
1607 | PEER_CAP_RESTART_AF_PRESERVE_RCV)) | |
1608 | bgp_clear_stale_route(peer, afi, safi); | |
1609 | ||
1610 | peer->nsf[afi][safi] = 1; | |
1611 | nsf_af_count++; | |
1612 | } else { | |
1613 | if (peer->nsf[afi][safi]) | |
1614 | bgp_clear_stale_route(peer, afi, safi); | |
1615 | peer->nsf[afi][safi] = 0; | |
1616 | } | |
1617 | } | |
1618 | ||
1619 | if (nsf_af_count) | |
1620 | SET_FLAG(peer->sflags, PEER_STATUS_NSF_MODE); | |
1621 | else { | |
1622 | UNSET_FLAG(peer->sflags, PEER_STATUS_NSF_MODE); | |
1623 | if (peer->t_gr_stale) { | |
1624 | BGP_TIMER_OFF(peer->t_gr_stale); | |
1625 | if (bgp_debug_neighbor_events(peer)) | |
1626 | zlog_debug( | |
1627 | "%s graceful restart stalepath timer stopped", | |
1628 | peer->host); | |
1629 | } | |
1630 | } | |
1631 | ||
1632 | if (peer->t_gr_restart) { | |
1633 | BGP_TIMER_OFF(peer->t_gr_restart); | |
1634 | if (bgp_debug_neighbor_events(peer)) | |
1635 | zlog_debug("%s graceful restart timer stopped", | |
1636 | peer->host); | |
1637 | } | |
1638 | ||
1639 | hook_call(peer_established, peer); | |
1640 | ||
1641 | /* Reset uptime, turn on keepalives, send current table. */ | |
1642 | if (!peer->v_holdtime) | |
1643 | bgp_keepalives_on(peer); | |
1644 | ||
1645 | peer->uptime = bgp_clock(); | |
1646 | ||
1647 | /* Send route-refresh when ORF is enabled */ | |
1648 | FOREACH_AFI_SAFI (afi, safi) { | |
1649 | if (CHECK_FLAG(peer->af_cap[afi][safi], | |
1650 | PEER_CAP_ORF_PREFIX_SM_ADV)) { | |
1651 | if (CHECK_FLAG(peer->af_cap[afi][safi], | |
1652 | PEER_CAP_ORF_PREFIX_RM_RCV)) | |
1653 | bgp_route_refresh_send(peer, afi, safi, | |
1654 | ORF_TYPE_PREFIX, | |
1655 | REFRESH_IMMEDIATE, 0); | |
1656 | else if (CHECK_FLAG(peer->af_cap[afi][safi], | |
1657 | PEER_CAP_ORF_PREFIX_RM_OLD_RCV)) | |
1658 | bgp_route_refresh_send(peer, afi, safi, | |
1659 | ORF_TYPE_PREFIX_OLD, | |
1660 | REFRESH_IMMEDIATE, 0); | |
1661 | } | |
1662 | } | |
1663 | ||
1664 | /* First update is deferred until ORF or ROUTE-REFRESH is received */ | |
1665 | FOREACH_AFI_SAFI (afi, safi) { | |
1666 | if (CHECK_FLAG(peer->af_cap[afi][safi], | |
1667 | PEER_CAP_ORF_PREFIX_RM_ADV)) | |
1668 | if (CHECK_FLAG(peer->af_cap[afi][safi], | |
1669 | PEER_CAP_ORF_PREFIX_SM_RCV) | |
1670 | || CHECK_FLAG(peer->af_cap[afi][safi], | |
1671 | PEER_CAP_ORF_PREFIX_SM_OLD_RCV)) | |
1672 | SET_FLAG(peer->af_sflags[afi][safi], | |
1673 | PEER_STATUS_ORF_WAIT_REFRESH); | |
1674 | } | |
1675 | ||
1676 | bgp_announce_peer(peer); | |
1677 | ||
1678 | /* Start the route advertisement timer to send updates to the peer - if | |
1679 | * BGP | |
1680 | * is not in read-only mode. If it is, the timer will be started at the | |
1681 | * end | |
1682 | * of read-only mode. | |
1683 | */ | |
1684 | if (!bgp_update_delay_active(peer->bgp)) { | |
1685 | BGP_TIMER_OFF(peer->t_routeadv); | |
1686 | BGP_TIMER_ON(peer->t_routeadv, bgp_routeadv_timer, 0); | |
1687 | } | |
1688 | ||
1689 | if (peer->doppelganger && (peer->doppelganger->status != Deleted)) { | |
1690 | if (bgp_debug_neighbor_events(peer)) | |
1691 | zlog_debug( | |
1692 | "[Event] Deleting stub connection for peer %s", | |
1693 | peer->host); | |
1694 | ||
1695 | if (peer->doppelganger->status > Active) | |
1696 | bgp_notify_send(peer->doppelganger, BGP_NOTIFY_CEASE, | |
1697 | BGP_NOTIFY_CEASE_COLLISION_RESOLUTION); | |
1698 | else | |
1699 | peer_delete(peer->doppelganger); | |
1700 | } | |
1701 | ||
1702 | /* | |
1703 | * If we are replacing the old peer for a doppelganger | |
1704 | * then switch it around in the bgp->peerhash | |
1705 | * the doppelgangers su and this peer's su are the same | |
1706 | * so the hash_release is the same for either. | |
1707 | */ | |
1708 | hash_release(peer->bgp->peerhash, peer); | |
1709 | hash_get(peer->bgp->peerhash, peer, hash_alloc_intern); | |
1710 | ||
1711 | bgp_bfd_register_peer(peer); | |
1712 | return ret; | |
1713 | } | |
1714 | ||
1715 | /* Keepalive packet is received. */ | |
1716 | static int bgp_fsm_keepalive(struct peer *peer) | |
1717 | { | |
1718 | BGP_TIMER_OFF(peer->t_holdtime); | |
1719 | return 0; | |
1720 | } | |
1721 | ||
1722 | /* Update packet is received. */ | |
1723 | static int bgp_fsm_update(struct peer *peer) | |
1724 | { | |
1725 | BGP_TIMER_OFF(peer->t_holdtime); | |
1726 | return 0; | |
1727 | } | |
1728 | ||
1729 | /* This is empty event. */ | |
1730 | static int bgp_ignore(struct peer *peer) | |
1731 | { | |
1732 | flog_err( | |
1733 | EC_BGP_FSM, | |
1734 | "%s [FSM] Ignoring event %s in state %s, prior events %s, %s, fd %d", | |
1735 | peer->host, bgp_event_str[peer->cur_event], | |
1736 | lookup_msg(bgp_status_msg, peer->status, NULL), | |
1737 | bgp_event_str[peer->last_event], | |
1738 | bgp_event_str[peer->last_major_event], peer->fd); | |
1739 | return 0; | |
1740 | } | |
1741 | ||
1742 | /* This is to handle unexpected events.. */ | |
1743 | static int bgp_fsm_exeption(struct peer *peer) | |
1744 | { | |
1745 | flog_err( | |
1746 | EC_BGP_FSM, | |
1747 | "%s [FSM] Unexpected event %s in state %s, prior events %s, %s, fd %d", | |
1748 | peer->host, bgp_event_str[peer->cur_event], | |
1749 | lookup_msg(bgp_status_msg, peer->status, NULL), | |
1750 | bgp_event_str[peer->last_event], | |
1751 | bgp_event_str[peer->last_major_event], peer->fd); | |
1752 | return (bgp_stop(peer)); | |
1753 | } | |
1754 | ||
1755 | void bgp_fsm_nht_update(struct peer *peer, int valid) | |
1756 | { | |
1757 | if (!peer) | |
1758 | return; | |
1759 | ||
1760 | switch (peer->status) { | |
1761 | case Idle: | |
1762 | if (valid) | |
1763 | BGP_EVENT_ADD(peer, BGP_Start); | |
1764 | break; | |
1765 | case Connect: | |
1766 | if (!valid) { | |
1767 | BGP_TIMER_OFF(peer->t_connect); | |
1768 | BGP_EVENT_ADD(peer, TCP_fatal_error); | |
1769 | } | |
1770 | break; | |
1771 | case Active: | |
1772 | if (valid) { | |
1773 | BGP_TIMER_OFF(peer->t_connect); | |
1774 | BGP_EVENT_ADD(peer, ConnectRetry_timer_expired); | |
1775 | } | |
1776 | break; | |
1777 | case OpenSent: | |
1778 | case OpenConfirm: | |
1779 | case Established: | |
1780 | if (!valid && (peer->gtsm_hops == 1)) | |
1781 | BGP_EVENT_ADD(peer, TCP_fatal_error); | |
1782 | case Clearing: | |
1783 | case Deleted: | |
1784 | default: | |
1785 | break; | |
1786 | } | |
1787 | } | |
1788 | ||
1789 | ||
1790 | /* Finite State Machine structure */ | |
1791 | static const struct { | |
1792 | int (*func)(struct peer *); | |
1793 | int next_state; | |
1794 | } FSM[BGP_STATUS_MAX - 1][BGP_EVENTS_MAX - 1] = { | |
1795 | { | |
1796 | /* Idle state: In Idle state, all events other than BGP_Start is | |
1797 | ignored. With BGP_Start event, finite state machine calls | |
1798 | bgp_start(). */ | |
1799 | {bgp_start, Connect}, /* BGP_Start */ | |
1800 | {bgp_stop, Idle}, /* BGP_Stop */ | |
1801 | {bgp_stop, Idle}, /* TCP_connection_open */ | |
1802 | {bgp_stop, Idle}, /* TCP_connection_closed */ | |
1803 | {bgp_ignore, Idle}, /* TCP_connection_open_failed */ | |
1804 | {bgp_stop, Idle}, /* TCP_fatal_error */ | |
1805 | {bgp_ignore, Idle}, /* ConnectRetry_timer_expired */ | |
1806 | {bgp_ignore, Idle}, /* Hold_Timer_expired */ | |
1807 | {bgp_ignore, Idle}, /* KeepAlive_timer_expired */ | |
1808 | {bgp_ignore, Idle}, /* Receive_OPEN_message */ | |
1809 | {bgp_ignore, Idle}, /* Receive_KEEPALIVE_message */ | |
1810 | {bgp_ignore, Idle}, /* Receive_UPDATE_message */ | |
1811 | {bgp_ignore, Idle}, /* Receive_NOTIFICATION_message */ | |
1812 | {bgp_ignore, Idle}, /* Clearing_Completed */ | |
1813 | }, | |
1814 | { | |
1815 | /* Connect */ | |
1816 | {bgp_ignore, Connect}, /* BGP_Start */ | |
1817 | {bgp_stop, Idle}, /* BGP_Stop */ | |
1818 | {bgp_connect_success, OpenSent}, /* TCP_connection_open */ | |
1819 | {bgp_stop, Idle}, /* TCP_connection_closed */ | |
1820 | {bgp_connect_fail, Active}, /* TCP_connection_open_failed */ | |
1821 | {bgp_connect_fail, Idle}, /* TCP_fatal_error */ | |
1822 | {bgp_reconnect, Connect}, /* ConnectRetry_timer_expired */ | |
1823 | {bgp_fsm_exeption, Idle}, /* Hold_Timer_expired */ | |
1824 | {bgp_fsm_exeption, Idle}, /* KeepAlive_timer_expired */ | |
1825 | {bgp_fsm_exeption, Idle}, /* Receive_OPEN_message */ | |
1826 | {bgp_fsm_exeption, Idle}, /* Receive_KEEPALIVE_message */ | |
1827 | {bgp_fsm_exeption, Idle}, /* Receive_UPDATE_message */ | |
1828 | {bgp_stop, Idle}, /* Receive_NOTIFICATION_message */ | |
1829 | {bgp_fsm_exeption, Idle}, /* Clearing_Completed */ | |
1830 | }, | |
1831 | { | |
1832 | /* Active, */ | |
1833 | {bgp_ignore, Active}, /* BGP_Start */ | |
1834 | {bgp_stop, Idle}, /* BGP_Stop */ | |
1835 | {bgp_connect_success, OpenSent}, /* TCP_connection_open */ | |
1836 | {bgp_stop, Idle}, /* TCP_connection_closed */ | |
1837 | {bgp_ignore, Active}, /* TCP_connection_open_failed */ | |
1838 | {bgp_fsm_exeption, Idle}, /* TCP_fatal_error */ | |
1839 | {bgp_start, Connect}, /* ConnectRetry_timer_expired */ | |
1840 | {bgp_fsm_exeption, Idle}, /* Hold_Timer_expired */ | |
1841 | {bgp_fsm_exeption, Idle}, /* KeepAlive_timer_expired */ | |
1842 | {bgp_fsm_exeption, Idle}, /* Receive_OPEN_message */ | |
1843 | {bgp_fsm_exeption, Idle}, /* Receive_KEEPALIVE_message */ | |
1844 | {bgp_fsm_exeption, Idle}, /* Receive_UPDATE_message */ | |
1845 | {bgp_fsm_exeption, Idle}, /* Receive_NOTIFICATION_message */ | |
1846 | {bgp_fsm_exeption, Idle}, /* Clearing_Completed */ | |
1847 | }, | |
1848 | { | |
1849 | /* OpenSent, */ | |
1850 | {bgp_ignore, OpenSent}, /* BGP_Start */ | |
1851 | {bgp_stop, Idle}, /* BGP_Stop */ | |
1852 | {bgp_stop, Active}, /* TCP_connection_open */ | |
1853 | {bgp_stop, Active}, /* TCP_connection_closed */ | |
1854 | {bgp_stop, Active}, /* TCP_connection_open_failed */ | |
1855 | {bgp_stop, Active}, /* TCP_fatal_error */ | |
1856 | {bgp_fsm_exeption, Idle}, /* ConnectRetry_timer_expired */ | |
1857 | {bgp_fsm_holdtime_expire, Idle}, /* Hold_Timer_expired */ | |
1858 | {bgp_fsm_exeption, Idle}, /* KeepAlive_timer_expired */ | |
1859 | {bgp_fsm_open, OpenConfirm}, /* Receive_OPEN_message */ | |
1860 | {bgp_fsm_event_error, Idle}, /* Receive_KEEPALIVE_message */ | |
1861 | {bgp_fsm_event_error, Idle}, /* Receive_UPDATE_message */ | |
1862 | {bgp_stop_with_error, Idle}, /* Receive_NOTIFICATION_message */ | |
1863 | {bgp_fsm_exeption, Idle}, /* Clearing_Completed */ | |
1864 | }, | |
1865 | { | |
1866 | /* OpenConfirm, */ | |
1867 | {bgp_ignore, OpenConfirm}, /* BGP_Start */ | |
1868 | {bgp_stop, Idle}, /* BGP_Stop */ | |
1869 | {bgp_stop, Idle}, /* TCP_connection_open */ | |
1870 | {bgp_stop, Idle}, /* TCP_connection_closed */ | |
1871 | {bgp_stop, Idle}, /* TCP_connection_open_failed */ | |
1872 | {bgp_stop, Idle}, /* TCP_fatal_error */ | |
1873 | {bgp_fsm_exeption, Idle}, /* ConnectRetry_timer_expired */ | |
1874 | {bgp_fsm_holdtime_expire, Idle}, /* Hold_Timer_expired */ | |
1875 | {bgp_ignore, OpenConfirm}, /* KeepAlive_timer_expired */ | |
1876 | {bgp_fsm_exeption, Idle}, /* Receive_OPEN_message */ | |
1877 | {bgp_establish, Established}, /* Receive_KEEPALIVE_message */ | |
1878 | {bgp_fsm_exeption, Idle}, /* Receive_UPDATE_message */ | |
1879 | {bgp_stop_with_error, Idle}, /* Receive_NOTIFICATION_message */ | |
1880 | {bgp_fsm_exeption, Idle}, /* Clearing_Completed */ | |
1881 | }, | |
1882 | { | |
1883 | /* Established, */ | |
1884 | {bgp_ignore, Established}, /* BGP_Start */ | |
1885 | {bgp_stop, Clearing}, /* BGP_Stop */ | |
1886 | {bgp_stop, Clearing}, /* TCP_connection_open */ | |
1887 | {bgp_stop, Clearing}, /* TCP_connection_closed */ | |
1888 | {bgp_stop, Clearing}, /* TCP_connection_open_failed */ | |
1889 | {bgp_stop, Clearing}, /* TCP_fatal_error */ | |
1890 | {bgp_stop, Clearing}, /* ConnectRetry_timer_expired */ | |
1891 | {bgp_fsm_holdtime_expire, Clearing}, /* Hold_Timer_expired */ | |
1892 | {bgp_ignore, Established}, /* KeepAlive_timer_expired */ | |
1893 | {bgp_stop, Clearing}, /* Receive_OPEN_message */ | |
1894 | {bgp_fsm_keepalive, | |
1895 | Established}, /* Receive_KEEPALIVE_message */ | |
1896 | {bgp_fsm_update, Established}, /* Receive_UPDATE_message */ | |
1897 | {bgp_stop_with_error, | |
1898 | Clearing}, /* Receive_NOTIFICATION_message */ | |
1899 | {bgp_fsm_exeption, Idle}, /* Clearing_Completed */ | |
1900 | }, | |
1901 | { | |
1902 | /* Clearing, */ | |
1903 | {bgp_ignore, Clearing}, /* BGP_Start */ | |
1904 | {bgp_stop, Clearing}, /* BGP_Stop */ | |
1905 | {bgp_stop, Clearing}, /* TCP_connection_open */ | |
1906 | {bgp_stop, Clearing}, /* TCP_connection_closed */ | |
1907 | {bgp_stop, Clearing}, /* TCP_connection_open_failed */ | |
1908 | {bgp_stop, Clearing}, /* TCP_fatal_error */ | |
1909 | {bgp_stop, Clearing}, /* ConnectRetry_timer_expired */ | |
1910 | {bgp_stop, Clearing}, /* Hold_Timer_expired */ | |
1911 | {bgp_stop, Clearing}, /* KeepAlive_timer_expired */ | |
1912 | {bgp_stop, Clearing}, /* Receive_OPEN_message */ | |
1913 | {bgp_stop, Clearing}, /* Receive_KEEPALIVE_message */ | |
1914 | {bgp_stop, Clearing}, /* Receive_UPDATE_message */ | |
1915 | {bgp_stop, Clearing}, /* Receive_NOTIFICATION_message */ | |
1916 | {bgp_clearing_completed, Idle}, /* Clearing_Completed */ | |
1917 | }, | |
1918 | { | |
1919 | /* Deleted, */ | |
1920 | {bgp_ignore, Deleted}, /* BGP_Start */ | |
1921 | {bgp_ignore, Deleted}, /* BGP_Stop */ | |
1922 | {bgp_ignore, Deleted}, /* TCP_connection_open */ | |
1923 | {bgp_ignore, Deleted}, /* TCP_connection_closed */ | |
1924 | {bgp_ignore, Deleted}, /* TCP_connection_open_failed */ | |
1925 | {bgp_ignore, Deleted}, /* TCP_fatal_error */ | |
1926 | {bgp_ignore, Deleted}, /* ConnectRetry_timer_expired */ | |
1927 | {bgp_ignore, Deleted}, /* Hold_Timer_expired */ | |
1928 | {bgp_ignore, Deleted}, /* KeepAlive_timer_expired */ | |
1929 | {bgp_ignore, Deleted}, /* Receive_OPEN_message */ | |
1930 | {bgp_ignore, Deleted}, /* Receive_KEEPALIVE_message */ | |
1931 | {bgp_ignore, Deleted}, /* Receive_UPDATE_message */ | |
1932 | {bgp_ignore, Deleted}, /* Receive_NOTIFICATION_message */ | |
1933 | {bgp_ignore, Deleted}, /* Clearing_Completed */ | |
1934 | }, | |
1935 | }; | |
1936 | ||
1937 | /* Execute event process. */ | |
1938 | int bgp_event(struct thread *thread) | |
1939 | { | |
1940 | int event; | |
1941 | struct peer *peer; | |
1942 | int ret; | |
1943 | ||
1944 | peer = THREAD_ARG(thread); | |
1945 | event = THREAD_VAL(thread); | |
1946 | ||
1947 | ret = bgp_event_update(peer, event); | |
1948 | ||
1949 | return (ret); | |
1950 | } | |
1951 | ||
1952 | int bgp_event_update(struct peer *peer, int event) | |
1953 | { | |
1954 | int next; | |
1955 | int ret = 0; | |
1956 | struct peer *other; | |
1957 | int passive_conn = 0; | |
1958 | int dyn_nbr; | |
1959 | ||
1960 | /* default return code */ | |
1961 | ret = FSM_PEER_NOOP; | |
1962 | ||
1963 | other = peer->doppelganger; | |
1964 | passive_conn = | |
1965 | (CHECK_FLAG(peer->sflags, PEER_STATUS_ACCEPT_PEER)) ? 1 : 0; | |
1966 | dyn_nbr = peer_dynamic_neighbor(peer); | |
1967 | ||
1968 | /* Logging this event. */ | |
1969 | next = FSM[peer->status - 1][event - 1].next_state; | |
1970 | ||
1971 | if (bgp_debug_neighbor_events(peer) && peer->status != next) | |
1972 | zlog_debug("%s [FSM] %s (%s->%s), fd %d", peer->host, | |
1973 | bgp_event_str[event], | |
1974 | lookup_msg(bgp_status_msg, peer->status, NULL), | |
1975 | lookup_msg(bgp_status_msg, next, NULL), peer->fd); | |
1976 | ||
1977 | peer->last_event = peer->cur_event; | |
1978 | peer->cur_event = event; | |
1979 | ||
1980 | /* Call function. */ | |
1981 | if (FSM[peer->status - 1][event - 1].func) | |
1982 | ret = (*(FSM[peer->status - 1][event - 1].func))(peer); | |
1983 | ||
1984 | if (ret >= 0) { | |
1985 | if (ret == 1 && next == Established) { | |
1986 | /* The case when doppelganger swap accurred in | |
1987 | bgp_establish. | |
1988 | Update the peer pointer accordingly */ | |
1989 | ret = FSM_PEER_TRANSFERRED; | |
1990 | peer = other; | |
1991 | } | |
1992 | ||
1993 | /* If status is changed. */ | |
1994 | if (next != peer->status) { | |
1995 | bgp_fsm_change_status(peer, next); | |
1996 | ||
1997 | /* | |
1998 | * If we're going to ESTABLISHED then we executed a | |
1999 | * peer transfer. In this case we can either return | |
2000 | * FSM_PEER_TRANSITIONED or FSM_PEER_TRANSFERRED. | |
2001 | * Opting for TRANSFERRED since transfer implies | |
2002 | * session establishment. | |
2003 | */ | |
2004 | if (ret != FSM_PEER_TRANSFERRED) | |
2005 | ret = FSM_PEER_TRANSITIONED; | |
2006 | } | |
2007 | ||
2008 | /* Make sure timer is set. */ | |
2009 | bgp_timer_set(peer); | |
2010 | ||
2011 | } else { | |
2012 | /* | |
2013 | * If we got a return value of -1, that means there was an | |
2014 | * error, restart the FSM. Since bgp_stop() was called on the | |
2015 | * peer. only a few fields are safe to access here. In any case | |
2016 | * we need to indicate that the peer was stopped in the return | |
2017 | * code. | |
2018 | */ | |
2019 | if (!dyn_nbr && !passive_conn && peer->bgp) { | |
2020 | flog_err( | |
2021 | EC_BGP_FSM, | |
2022 | "%s [FSM] Failure handling event %s in state %s, " | |
2023 | "prior events %s, %s, fd %d", | |
2024 | peer->host, bgp_event_str[peer->cur_event], | |
2025 | lookup_msg(bgp_status_msg, peer->status, NULL), | |
2026 | bgp_event_str[peer->last_event], | |
2027 | bgp_event_str[peer->last_major_event], | |
2028 | peer->fd); | |
2029 | bgp_stop(peer); | |
2030 | bgp_fsm_change_status(peer, Idle); | |
2031 | bgp_timer_set(peer); | |
2032 | } | |
2033 | ret = FSM_PEER_STOPPED; | |
2034 | } | |
2035 | ||
2036 | return ret; | |
2037 | } |