]>
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_status_changed, (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 | /* Register peer with NHT */ | |
100 | static int bgp_peer_reg_with_nht(struct peer *peer) | |
101 | { | |
102 | int connected = 0; | |
103 | ||
104 | if (peer->sort == BGP_PEER_EBGP && peer->ttl == BGP_DEFAULT_TTL | |
105 | && !CHECK_FLAG(peer->flags, PEER_FLAG_DISABLE_CONNECTED_CHECK) | |
106 | && !bgp_flag_check(peer->bgp, BGP_FLAG_DISABLE_NH_CONNECTED_CHK)) | |
107 | connected = 1; | |
108 | ||
109 | return bgp_find_or_add_nexthop( | |
110 | peer->bgp, peer->bgp, family2afi(peer->su.sa.sa_family), | |
111 | NULL, peer, connected); | |
112 | } | |
113 | ||
114 | static void peer_xfer_stats(struct peer *peer_dst, struct peer *peer_src) | |
115 | { | |
116 | /* Copy stats over. These are only the pre-established state stats */ | |
117 | peer_dst->open_in += peer_src->open_in; | |
118 | peer_dst->open_out += peer_src->open_out; | |
119 | peer_dst->keepalive_in += peer_src->keepalive_in; | |
120 | peer_dst->keepalive_out += peer_src->keepalive_out; | |
121 | peer_dst->notify_in += peer_src->notify_in; | |
122 | peer_dst->notify_out += peer_src->notify_out; | |
123 | peer_dst->dynamic_cap_in += peer_src->dynamic_cap_in; | |
124 | peer_dst->dynamic_cap_out += peer_src->dynamic_cap_out; | |
125 | } | |
126 | ||
127 | static struct peer *peer_xfer_conn(struct peer *from_peer) | |
128 | { | |
129 | struct peer *peer; | |
130 | afi_t afi; | |
131 | safi_t safi; | |
132 | int fd; | |
133 | int status, pstatus; | |
134 | unsigned char last_evt, last_maj_evt; | |
135 | ||
136 | assert(from_peer != NULL); | |
137 | ||
138 | peer = from_peer->doppelganger; | |
139 | ||
140 | if (!peer || !CHECK_FLAG(peer->flags, PEER_FLAG_CONFIG_NODE)) | |
141 | return from_peer; | |
142 | ||
143 | /* | |
144 | * Let's check that we are not going to loose known configuration | |
145 | * state based upon doppelganger rules. | |
146 | */ | |
147 | FOREACH_AFI_SAFI (afi, safi) { | |
148 | if (from_peer->afc[afi][safi] != peer->afc[afi][safi]) { | |
149 | flog_err( | |
150 | EC_BGP_DOPPELGANGER_CONFIG, | |
151 | "from_peer->afc[%d][%d] is not the same as what we are overwriting", | |
152 | afi, safi); | |
153 | return NULL; | |
154 | } | |
155 | } | |
156 | ||
157 | if (bgp_debug_neighbor_events(peer)) | |
158 | zlog_debug("%s: peer transfer %p fd %d -> %p fd %d)", | |
159 | from_peer->host, from_peer, from_peer->fd, peer, | |
160 | peer->fd); | |
161 | ||
162 | bgp_writes_off(peer); | |
163 | bgp_reads_off(peer); | |
164 | bgp_writes_off(from_peer); | |
165 | bgp_reads_off(from_peer); | |
166 | ||
167 | BGP_TIMER_OFF(peer->t_routeadv); | |
168 | BGP_TIMER_OFF(peer->t_connect); | |
169 | BGP_TIMER_OFF(peer->t_connect_check_r); | |
170 | BGP_TIMER_OFF(peer->t_connect_check_w); | |
171 | BGP_TIMER_OFF(from_peer->t_routeadv); | |
172 | BGP_TIMER_OFF(from_peer->t_connect); | |
173 | BGP_TIMER_OFF(from_peer->t_connect_check_r); | |
174 | BGP_TIMER_OFF(from_peer->t_connect_check_w); | |
175 | BGP_TIMER_OFF(from_peer->t_process_packet); | |
176 | ||
177 | /* | |
178 | * At this point in time, it is possible that there are packets pending | |
179 | * on various buffers. Those need to be transferred or dropped, | |
180 | * otherwise we'll get spurious failures during session establishment. | |
181 | */ | |
182 | frr_with_mutex(&peer->io_mtx, &from_peer->io_mtx) { | |
183 | fd = peer->fd; | |
184 | peer->fd = from_peer->fd; | |
185 | from_peer->fd = fd; | |
186 | ||
187 | stream_fifo_clean(peer->ibuf); | |
188 | stream_fifo_clean(peer->obuf); | |
189 | ||
190 | /* | |
191 | * this should never happen, since bgp_process_packet() is the | |
192 | * only task that sets and unsets the current packet and it | |
193 | * runs in our pthread. | |
194 | */ | |
195 | if (peer->curr) { | |
196 | flog_err( | |
197 | EC_BGP_PKT_PROCESS, | |
198 | "[%s] Dropping pending packet on connection transfer:", | |
199 | peer->host); | |
200 | /* there used to be a bgp_packet_dump call here, but | |
201 | * that's extremely confusing since there's no way to | |
202 | * identify the packet in MRT dumps or BMP as dropped | |
203 | * due to connection transfer. | |
204 | */ | |
205 | stream_free(peer->curr); | |
206 | peer->curr = NULL; | |
207 | } | |
208 | ||
209 | // copy each packet from old peer's output queue to new peer | |
210 | while (from_peer->obuf->head) | |
211 | stream_fifo_push(peer->obuf, | |
212 | stream_fifo_pop(from_peer->obuf)); | |
213 | ||
214 | // copy each packet from old peer's input queue to new peer | |
215 | while (from_peer->ibuf->head) | |
216 | stream_fifo_push(peer->ibuf, | |
217 | stream_fifo_pop(from_peer->ibuf)); | |
218 | ||
219 | ringbuf_wipe(peer->ibuf_work); | |
220 | ringbuf_copy(peer->ibuf_work, from_peer->ibuf_work, | |
221 | ringbuf_remain(from_peer->ibuf_work)); | |
222 | } | |
223 | ||
224 | peer->as = from_peer->as; | |
225 | peer->v_holdtime = from_peer->v_holdtime; | |
226 | peer->v_keepalive = from_peer->v_keepalive; | |
227 | peer->v_routeadv = from_peer->v_routeadv; | |
228 | peer->v_gr_restart = from_peer->v_gr_restart; | |
229 | peer->cap = from_peer->cap; | |
230 | status = peer->status; | |
231 | pstatus = peer->ostatus; | |
232 | last_evt = peer->last_event; | |
233 | last_maj_evt = peer->last_major_event; | |
234 | peer->status = from_peer->status; | |
235 | peer->ostatus = from_peer->ostatus; | |
236 | peer->last_event = from_peer->last_event; | |
237 | peer->last_major_event = from_peer->last_major_event; | |
238 | from_peer->status = status; | |
239 | from_peer->ostatus = pstatus; | |
240 | from_peer->last_event = last_evt; | |
241 | from_peer->last_major_event = last_maj_evt; | |
242 | peer->remote_id = from_peer->remote_id; | |
243 | peer->last_reset = from_peer->last_reset; | |
244 | ||
245 | if (from_peer->hostname != NULL) { | |
246 | if (peer->hostname) { | |
247 | XFREE(MTYPE_BGP_PEER_HOST, peer->hostname); | |
248 | peer->hostname = NULL; | |
249 | } | |
250 | ||
251 | peer->hostname = from_peer->hostname; | |
252 | from_peer->hostname = NULL; | |
253 | } | |
254 | ||
255 | if (from_peer->domainname != NULL) { | |
256 | if (peer->domainname) { | |
257 | XFREE(MTYPE_BGP_PEER_HOST, peer->domainname); | |
258 | peer->domainname = NULL; | |
259 | } | |
260 | ||
261 | peer->domainname = from_peer->domainname; | |
262 | from_peer->domainname = NULL; | |
263 | } | |
264 | ||
265 | FOREACH_AFI_SAFI (afi, safi) { | |
266 | peer->af_flags[afi][safi] = from_peer->af_flags[afi][safi]; | |
267 | peer->af_sflags[afi][safi] = from_peer->af_sflags[afi][safi]; | |
268 | peer->af_cap[afi][safi] = from_peer->af_cap[afi][safi]; | |
269 | peer->afc_nego[afi][safi] = from_peer->afc_nego[afi][safi]; | |
270 | peer->afc_adv[afi][safi] = from_peer->afc_adv[afi][safi]; | |
271 | peer->afc_recv[afi][safi] = from_peer->afc_recv[afi][safi]; | |
272 | peer->orf_plist[afi][safi] = from_peer->orf_plist[afi][safi]; | |
273 | } | |
274 | ||
275 | if (bgp_getsockname(peer) < 0) { | |
276 | flog_err( | |
277 | EC_LIB_SOCKET, | |
278 | "%%bgp_getsockname() failed for %s peer %s fd %d (from_peer fd %d)", | |
279 | (CHECK_FLAG(peer->sflags, PEER_STATUS_ACCEPT_PEER) | |
280 | ? "accept" | |
281 | : ""), | |
282 | peer->host, peer->fd, from_peer->fd); | |
283 | bgp_stop(peer); | |
284 | bgp_stop(from_peer); | |
285 | return NULL; | |
286 | } | |
287 | if (from_peer->status > Active) { | |
288 | if (bgp_getsockname(from_peer) < 0) { | |
289 | flog_err( | |
290 | EC_LIB_SOCKET, | |
291 | "%%bgp_getsockname() failed for %s from_peer %s fd %d (peer fd %d)", | |
292 | ||
293 | (CHECK_FLAG(from_peer->sflags, | |
294 | PEER_STATUS_ACCEPT_PEER) | |
295 | ? "accept" | |
296 | : ""), | |
297 | from_peer->host, from_peer->fd, peer->fd); | |
298 | bgp_stop(from_peer); | |
299 | from_peer = NULL; | |
300 | } | |
301 | } | |
302 | ||
303 | ||
304 | // Note: peer_xfer_stats() must be called with I/O turned OFF | |
305 | if (from_peer) | |
306 | peer_xfer_stats(peer, from_peer); | |
307 | ||
308 | /* Register peer for NHT. This is to allow RAs to be enabled when | |
309 | * needed, even on a passive connection. | |
310 | */ | |
311 | bgp_peer_reg_with_nht(peer); | |
312 | ||
313 | bgp_reads_on(peer); | |
314 | bgp_writes_on(peer); | |
315 | thread_add_timer_msec(bm->master, bgp_process_packet, peer, 0, | |
316 | &peer->t_process_packet); | |
317 | ||
318 | return (peer); | |
319 | } | |
320 | ||
321 | /* Hook function called after bgp event is occered. And vty's | |
322 | neighbor command invoke this function after making neighbor | |
323 | structure. */ | |
324 | void bgp_timer_set(struct peer *peer) | |
325 | { | |
326 | switch (peer->status) { | |
327 | case Idle: | |
328 | /* First entry point of peer's finite state machine. In Idle | |
329 | status start timer is on unless peer is shutdown or peer is | |
330 | inactive. All other timer must be turned off */ | |
331 | if (BGP_PEER_START_SUPPRESSED(peer) || !peer_active(peer) | |
332 | || (peer->bgp->inst_type != BGP_INSTANCE_TYPE_VIEW && | |
333 | peer->bgp->vrf_id == VRF_UNKNOWN)) { | |
334 | BGP_TIMER_OFF(peer->t_start); | |
335 | } else { | |
336 | BGP_TIMER_ON(peer->t_start, bgp_start_timer, | |
337 | peer->v_start); | |
338 | } | |
339 | BGP_TIMER_OFF(peer->t_connect); | |
340 | BGP_TIMER_OFF(peer->t_holdtime); | |
341 | bgp_keepalives_off(peer); | |
342 | BGP_TIMER_OFF(peer->t_routeadv); | |
343 | break; | |
344 | ||
345 | case Connect: | |
346 | /* After start timer is expired, the peer moves to Connect | |
347 | status. Make sure start timer is off and connect timer is | |
348 | on. */ | |
349 | BGP_TIMER_OFF(peer->t_start); | |
350 | BGP_TIMER_ON(peer->t_connect, bgp_connect_timer, | |
351 | peer->v_connect); | |
352 | BGP_TIMER_OFF(peer->t_holdtime); | |
353 | bgp_keepalives_off(peer); | |
354 | BGP_TIMER_OFF(peer->t_routeadv); | |
355 | break; | |
356 | ||
357 | case Active: | |
358 | /* Active is waiting connection from remote peer. And if | |
359 | connect timer is expired, change status to Connect. */ | |
360 | BGP_TIMER_OFF(peer->t_start); | |
361 | /* If peer is passive mode, do not set connect timer. */ | |
362 | if (CHECK_FLAG(peer->flags, PEER_FLAG_PASSIVE) | |
363 | || CHECK_FLAG(peer->sflags, PEER_STATUS_NSF_WAIT)) { | |
364 | BGP_TIMER_OFF(peer->t_connect); | |
365 | } else { | |
366 | BGP_TIMER_ON(peer->t_connect, bgp_connect_timer, | |
367 | peer->v_connect); | |
368 | } | |
369 | BGP_TIMER_OFF(peer->t_holdtime); | |
370 | bgp_keepalives_off(peer); | |
371 | BGP_TIMER_OFF(peer->t_routeadv); | |
372 | break; | |
373 | ||
374 | case OpenSent: | |
375 | /* OpenSent status. */ | |
376 | BGP_TIMER_OFF(peer->t_start); | |
377 | BGP_TIMER_OFF(peer->t_connect); | |
378 | if (peer->v_holdtime != 0) { | |
379 | BGP_TIMER_ON(peer->t_holdtime, bgp_holdtime_timer, | |
380 | peer->v_holdtime); | |
381 | } else { | |
382 | BGP_TIMER_OFF(peer->t_holdtime); | |
383 | } | |
384 | bgp_keepalives_off(peer); | |
385 | BGP_TIMER_OFF(peer->t_routeadv); | |
386 | break; | |
387 | ||
388 | case OpenConfirm: | |
389 | /* OpenConfirm status. */ | |
390 | BGP_TIMER_OFF(peer->t_start); | |
391 | BGP_TIMER_OFF(peer->t_connect); | |
392 | ||
393 | /* If the negotiated Hold Time value is zero, then the Hold Time | |
394 | timer and KeepAlive timers are not started. */ | |
395 | if (peer->v_holdtime == 0) { | |
396 | BGP_TIMER_OFF(peer->t_holdtime); | |
397 | bgp_keepalives_off(peer); | |
398 | } else { | |
399 | BGP_TIMER_ON(peer->t_holdtime, bgp_holdtime_timer, | |
400 | peer->v_holdtime); | |
401 | bgp_keepalives_on(peer); | |
402 | } | |
403 | BGP_TIMER_OFF(peer->t_routeadv); | |
404 | break; | |
405 | ||
406 | case Established: | |
407 | /* In Established status start and connect timer is turned | |
408 | off. */ | |
409 | BGP_TIMER_OFF(peer->t_start); | |
410 | BGP_TIMER_OFF(peer->t_connect); | |
411 | ||
412 | /* Same as OpenConfirm, if holdtime is zero then both holdtime | |
413 | and keepalive must be turned off. */ | |
414 | if (peer->v_holdtime == 0) { | |
415 | BGP_TIMER_OFF(peer->t_holdtime); | |
416 | bgp_keepalives_off(peer); | |
417 | } else { | |
418 | BGP_TIMER_ON(peer->t_holdtime, bgp_holdtime_timer, | |
419 | peer->v_holdtime); | |
420 | bgp_keepalives_on(peer); | |
421 | } | |
422 | break; | |
423 | case Deleted: | |
424 | BGP_TIMER_OFF(peer->t_gr_restart); | |
425 | BGP_TIMER_OFF(peer->t_gr_stale); | |
426 | BGP_TIMER_OFF(peer->t_pmax_restart); | |
427 | /* fallthru */ | |
428 | case Clearing: | |
429 | BGP_TIMER_OFF(peer->t_start); | |
430 | BGP_TIMER_OFF(peer->t_connect); | |
431 | BGP_TIMER_OFF(peer->t_holdtime); | |
432 | bgp_keepalives_off(peer); | |
433 | BGP_TIMER_OFF(peer->t_routeadv); | |
434 | break; | |
435 | } | |
436 | } | |
437 | ||
438 | /* BGP start timer. This function set BGP_Start event to thread value | |
439 | and process event. */ | |
440 | static int bgp_start_timer(struct thread *thread) | |
441 | { | |
442 | struct peer *peer; | |
443 | ||
444 | peer = THREAD_ARG(thread); | |
445 | peer->t_start = NULL; | |
446 | ||
447 | if (bgp_debug_neighbor_events(peer)) | |
448 | zlog_debug("%s [FSM] Timer (start timer expire).", peer->host); | |
449 | ||
450 | THREAD_VAL(thread) = BGP_Start; | |
451 | bgp_event(thread); /* bgp_event unlocks peer */ | |
452 | ||
453 | return 0; | |
454 | } | |
455 | ||
456 | /* BGP connect retry timer. */ | |
457 | static int bgp_connect_timer(struct thread *thread) | |
458 | { | |
459 | struct peer *peer; | |
460 | int ret; | |
461 | ||
462 | peer = THREAD_ARG(thread); | |
463 | ||
464 | assert(!peer->t_write); | |
465 | assert(!peer->t_read); | |
466 | ||
467 | peer->t_connect = NULL; | |
468 | ||
469 | if (bgp_debug_neighbor_events(peer)) | |
470 | zlog_debug("%s [FSM] Timer (connect timer expire)", peer->host); | |
471 | ||
472 | if (CHECK_FLAG(peer->sflags, PEER_STATUS_ACCEPT_PEER)) { | |
473 | bgp_stop(peer); | |
474 | ret = -1; | |
475 | } else { | |
476 | THREAD_VAL(thread) = ConnectRetry_timer_expired; | |
477 | bgp_event(thread); /* bgp_event unlocks peer */ | |
478 | ret = 0; | |
479 | } | |
480 | ||
481 | return ret; | |
482 | } | |
483 | ||
484 | /* BGP holdtime timer. */ | |
485 | static int bgp_holdtime_timer(struct thread *thread) | |
486 | { | |
487 | struct peer *peer; | |
488 | ||
489 | peer = THREAD_ARG(thread); | |
490 | peer->t_holdtime = NULL; | |
491 | ||
492 | if (bgp_debug_neighbor_events(peer)) | |
493 | zlog_debug("%s [FSM] Timer (holdtime timer expire)", | |
494 | peer->host); | |
495 | ||
496 | THREAD_VAL(thread) = Hold_Timer_expired; | |
497 | bgp_event(thread); /* bgp_event unlocks peer */ | |
498 | ||
499 | return 0; | |
500 | } | |
501 | ||
502 | int bgp_routeadv_timer(struct thread *thread) | |
503 | { | |
504 | struct peer *peer; | |
505 | ||
506 | peer = THREAD_ARG(thread); | |
507 | peer->t_routeadv = NULL; | |
508 | ||
509 | if (bgp_debug_neighbor_events(peer)) | |
510 | zlog_debug("%s [FSM] Timer (routeadv timer expire)", | |
511 | peer->host); | |
512 | ||
513 | peer->synctime = bgp_clock(); | |
514 | ||
515 | thread_add_timer_msec(bm->master, bgp_generate_updgrp_packets, peer, 0, | |
516 | &peer->t_generate_updgrp_packets); | |
517 | ||
518 | /* MRAI timer will be started again when FIFO is built, no need to | |
519 | * do it here. | |
520 | */ | |
521 | return 0; | |
522 | } | |
523 | ||
524 | /* BGP Peer Down Cause */ | |
525 | const char *peer_down_str[] = {"", | |
526 | "Router ID changed", | |
527 | "Remote AS changed", | |
528 | "Local AS change", | |
529 | "Cluster ID changed", | |
530 | "Confederation identifier changed", | |
531 | "Confederation peer changed", | |
532 | "RR client config change", | |
533 | "RS client config change", | |
534 | "Update source change", | |
535 | "Address family activated", | |
536 | "Admin. shutdown", | |
537 | "User reset", | |
538 | "BGP Notification received", | |
539 | "BGP Notification send", | |
540 | "Peer closed the session", | |
541 | "Neighbor deleted", | |
542 | "Peer-group add member", | |
543 | "Peer-group delete member", | |
544 | "Capability changed", | |
545 | "Passive config change", | |
546 | "Multihop config change", | |
547 | "NSF peer closed the session", | |
548 | "Intf peering v6only config change", | |
549 | "BFD down received", | |
550 | "Interface down", | |
551 | "Neighbor address lost", | |
552 | "Waiting for NHT", | |
553 | "Waiting for Peer IPv6 LLA", | |
554 | "Waiting for VRF to be initialized", | |
555 | "No AFI/SAFI activated for peer"}; | |
556 | ||
557 | static int bgp_graceful_restart_timer_expire(struct thread *thread) | |
558 | { | |
559 | struct peer *peer; | |
560 | afi_t afi; | |
561 | safi_t safi; | |
562 | ||
563 | peer = THREAD_ARG(thread); | |
564 | peer->t_gr_restart = NULL; | |
565 | ||
566 | /* NSF delete stale route */ | |
567 | for (afi = AFI_IP; afi < AFI_MAX; afi++) | |
568 | for (safi = SAFI_UNICAST; safi <= SAFI_MPLS_VPN; safi++) | |
569 | if (peer->nsf[afi][safi]) | |
570 | bgp_clear_stale_route(peer, afi, safi); | |
571 | ||
572 | UNSET_FLAG(peer->sflags, PEER_STATUS_NSF_WAIT); | |
573 | BGP_TIMER_OFF(peer->t_gr_stale); | |
574 | ||
575 | if (bgp_debug_neighbor_events(peer)) { | |
576 | zlog_debug("%s graceful restart timer expired", peer->host); | |
577 | zlog_debug("%s graceful restart stalepath timer stopped", | |
578 | peer->host); | |
579 | } | |
580 | ||
581 | bgp_timer_set(peer); | |
582 | ||
583 | return 0; | |
584 | } | |
585 | ||
586 | static int bgp_graceful_stale_timer_expire(struct thread *thread) | |
587 | { | |
588 | struct peer *peer; | |
589 | afi_t afi; | |
590 | safi_t safi; | |
591 | ||
592 | peer = THREAD_ARG(thread); | |
593 | peer->t_gr_stale = NULL; | |
594 | ||
595 | if (bgp_debug_neighbor_events(peer)) | |
596 | zlog_debug("%s graceful restart stalepath timer expired", | |
597 | peer->host); | |
598 | ||
599 | /* NSF delete stale route */ | |
600 | for (afi = AFI_IP; afi < AFI_MAX; afi++) | |
601 | for (safi = SAFI_UNICAST; safi <= SAFI_MPLS_VPN; safi++) | |
602 | if (peer->nsf[afi][safi]) | |
603 | bgp_clear_stale_route(peer, afi, safi); | |
604 | ||
605 | return 0; | |
606 | } | |
607 | ||
608 | static int bgp_update_delay_applicable(struct bgp *bgp) | |
609 | { | |
610 | /* update_delay_over flag should be reset (set to 0) for any new | |
611 | applicability of the update-delay during BGP process lifetime. | |
612 | And it should be set after an occurence of the update-delay is | |
613 | over)*/ | |
614 | if (!bgp->update_delay_over) | |
615 | return 1; | |
616 | ||
617 | return 0; | |
618 | } | |
619 | ||
620 | int bgp_update_delay_active(struct bgp *bgp) | |
621 | { | |
622 | if (bgp->t_update_delay) | |
623 | return 1; | |
624 | ||
625 | return 0; | |
626 | } | |
627 | ||
628 | int bgp_update_delay_configured(struct bgp *bgp) | |
629 | { | |
630 | if (bgp->v_update_delay) | |
631 | return 1; | |
632 | ||
633 | return 0; | |
634 | } | |
635 | ||
636 | /* Do the post-processing needed when bgp comes out of the read-only mode | |
637 | on ending the update delay. */ | |
638 | void bgp_update_delay_end(struct bgp *bgp) | |
639 | { | |
640 | THREAD_TIMER_OFF(bgp->t_update_delay); | |
641 | THREAD_TIMER_OFF(bgp->t_establish_wait); | |
642 | ||
643 | /* Reset update-delay related state */ | |
644 | bgp->update_delay_over = 1; | |
645 | bgp->established = 0; | |
646 | bgp->restarted_peers = 0; | |
647 | bgp->implicit_eors = 0; | |
648 | bgp->explicit_eors = 0; | |
649 | ||
650 | quagga_timestamp(3, bgp->update_delay_end_time, | |
651 | sizeof(bgp->update_delay_end_time)); | |
652 | ||
653 | /* | |
654 | * Add an end-of-initial-update marker to the main process queues so | |
655 | * that | |
656 | * the route advertisement timer for the peers can be started. Also set | |
657 | * the zebra and peer update hold flags. These flags are used to achieve | |
658 | * three stages in the update-delay post processing: | |
659 | * 1. Finish best-path selection for all the prefixes held on the | |
660 | * queues. | |
661 | * (routes in BGP are updated, and peers sync queues are populated | |
662 | * too) | |
663 | * 2. As the eoiu mark is reached in the bgp process routine, ship all | |
664 | * the | |
665 | * routes to zebra. With that zebra should see updates from BGP | |
666 | * close | |
667 | * to each other. | |
668 | * 3. Unblock the peer update writes. With that peer update packing | |
669 | * with | |
670 | * the prefixes should be at its maximum. | |
671 | */ | |
672 | bgp_add_eoiu_mark(bgp); | |
673 | bgp->main_zebra_update_hold = 1; | |
674 | bgp->main_peers_update_hold = 1; | |
675 | ||
676 | /* Resume the queue processing. This should trigger the event that would | |
677 | take | |
678 | care of processing any work that was queued during the read-only | |
679 | mode. */ | |
680 | work_queue_unplug(bm->process_main_queue); | |
681 | } | |
682 | ||
683 | /** | |
684 | * see bgp_fsm.h | |
685 | */ | |
686 | void bgp_start_routeadv(struct bgp *bgp) | |
687 | { | |
688 | struct listnode *node, *nnode; | |
689 | struct peer *peer; | |
690 | ||
691 | zlog_info("bgp_start_routeadv(), update hold status %d", | |
692 | bgp->main_peers_update_hold); | |
693 | ||
694 | if (bgp->main_peers_update_hold) | |
695 | return; | |
696 | ||
697 | quagga_timestamp(3, bgp->update_delay_peers_resume_time, | |
698 | sizeof(bgp->update_delay_peers_resume_time)); | |
699 | ||
700 | for (ALL_LIST_ELEMENTS(bgp->peer, node, nnode, peer)) { | |
701 | if (peer->status != Established) | |
702 | continue; | |
703 | BGP_TIMER_OFF(peer->t_routeadv); | |
704 | BGP_TIMER_ON(peer->t_routeadv, bgp_routeadv_timer, 0); | |
705 | } | |
706 | } | |
707 | ||
708 | /** | |
709 | * see bgp_fsm.h | |
710 | */ | |
711 | void bgp_adjust_routeadv(struct peer *peer) | |
712 | { | |
713 | time_t nowtime = bgp_clock(); | |
714 | double diff; | |
715 | unsigned long remain; | |
716 | ||
717 | /* Bypass checks for special case of MRAI being 0 */ | |
718 | if (peer->v_routeadv == 0) { | |
719 | /* Stop existing timer, just in case it is running for a | |
720 | * different | |
721 | * duration and schedule write thread immediately. | |
722 | */ | |
723 | if (peer->t_routeadv) | |
724 | BGP_TIMER_OFF(peer->t_routeadv); | |
725 | ||
726 | peer->synctime = bgp_clock(); | |
727 | thread_add_timer_msec(bm->master, bgp_generate_updgrp_packets, | |
728 | peer, 0, | |
729 | &peer->t_generate_updgrp_packets); | |
730 | return; | |
731 | } | |
732 | ||
733 | ||
734 | /* | |
735 | * CASE I: | |
736 | * If the last update was written more than MRAI back, expire the timer | |
737 | * instantly so that we can send the update out sooner. | |
738 | * | |
739 | * <------- MRAI ---------> | |
740 | * |-----------------|-----------------------| | |
741 | * <------------- m ------------> | |
742 | * ^ ^ ^ | |
743 | * | | | | |
744 | * | | current time | |
745 | * | timer start | |
746 | * last write | |
747 | * | |
748 | * m > MRAI | |
749 | */ | |
750 | diff = difftime(nowtime, peer->last_update); | |
751 | if (diff > (double)peer->v_routeadv) { | |
752 | BGP_TIMER_OFF(peer->t_routeadv); | |
753 | BGP_TIMER_ON(peer->t_routeadv, bgp_routeadv_timer, 0); | |
754 | return; | |
755 | } | |
756 | ||
757 | /* | |
758 | * CASE II: | |
759 | * - Find when to expire the MRAI timer. | |
760 | * If MRAI timer is not active, assume we can start it now. | |
761 | * | |
762 | * <------- MRAI ---------> | |
763 | * |------------|-----------------------| | |
764 | * <-------- m ----------><----- r -----> | |
765 | * ^ ^ ^ | |
766 | * | | | | |
767 | * | | current time | |
768 | * | timer start | |
769 | * last write | |
770 | * | |
771 | * (MRAI - m) < r | |
772 | */ | |
773 | if (peer->t_routeadv) | |
774 | remain = thread_timer_remain_second(peer->t_routeadv); | |
775 | else | |
776 | remain = peer->v_routeadv; | |
777 | diff = peer->v_routeadv - diff; | |
778 | if (diff <= (double)remain) { | |
779 | BGP_TIMER_OFF(peer->t_routeadv); | |
780 | BGP_TIMER_ON(peer->t_routeadv, bgp_routeadv_timer, diff); | |
781 | } | |
782 | } | |
783 | ||
784 | static int bgp_maxmed_onstartup_applicable(struct bgp *bgp) | |
785 | { | |
786 | if (!bgp->maxmed_onstartup_over) | |
787 | return 1; | |
788 | ||
789 | return 0; | |
790 | } | |
791 | ||
792 | int bgp_maxmed_onstartup_configured(struct bgp *bgp) | |
793 | { | |
794 | if (bgp->v_maxmed_onstartup != BGP_MAXMED_ONSTARTUP_UNCONFIGURED) | |
795 | return 1; | |
796 | ||
797 | return 0; | |
798 | } | |
799 | ||
800 | int bgp_maxmed_onstartup_active(struct bgp *bgp) | |
801 | { | |
802 | if (bgp->t_maxmed_onstartup) | |
803 | return 1; | |
804 | ||
805 | return 0; | |
806 | } | |
807 | ||
808 | void bgp_maxmed_update(struct bgp *bgp) | |
809 | { | |
810 | uint8_t maxmed_active; | |
811 | uint32_t maxmed_value; | |
812 | ||
813 | if (bgp->v_maxmed_admin) { | |
814 | maxmed_active = 1; | |
815 | maxmed_value = bgp->maxmed_admin_value; | |
816 | } else if (bgp->t_maxmed_onstartup) { | |
817 | maxmed_active = 1; | |
818 | maxmed_value = bgp->maxmed_onstartup_value; | |
819 | } else { | |
820 | maxmed_active = 0; | |
821 | maxmed_value = BGP_MAXMED_VALUE_DEFAULT; | |
822 | } | |
823 | ||
824 | if (bgp->maxmed_active != maxmed_active | |
825 | || bgp->maxmed_value != maxmed_value) { | |
826 | bgp->maxmed_active = maxmed_active; | |
827 | bgp->maxmed_value = maxmed_value; | |
828 | ||
829 | update_group_announce(bgp); | |
830 | } | |
831 | } | |
832 | ||
833 | /* The maxmed onstartup timer expiry callback. */ | |
834 | static int bgp_maxmed_onstartup_timer(struct thread *thread) | |
835 | { | |
836 | struct bgp *bgp; | |
837 | ||
838 | zlog_info("Max med on startup ended - timer expired."); | |
839 | ||
840 | bgp = THREAD_ARG(thread); | |
841 | THREAD_TIMER_OFF(bgp->t_maxmed_onstartup); | |
842 | bgp->maxmed_onstartup_over = 1; | |
843 | ||
844 | bgp_maxmed_update(bgp); | |
845 | ||
846 | return 0; | |
847 | } | |
848 | ||
849 | static void bgp_maxmed_onstartup_begin(struct bgp *bgp) | |
850 | { | |
851 | /* Applicable only once in the process lifetime on the startup */ | |
852 | if (bgp->maxmed_onstartup_over) | |
853 | return; | |
854 | ||
855 | zlog_info("Begin maxmed onstartup mode - timer %d seconds", | |
856 | bgp->v_maxmed_onstartup); | |
857 | ||
858 | thread_add_timer(bm->master, bgp_maxmed_onstartup_timer, bgp, | |
859 | bgp->v_maxmed_onstartup, &bgp->t_maxmed_onstartup); | |
860 | ||
861 | if (!bgp->v_maxmed_admin) { | |
862 | bgp->maxmed_active = 1; | |
863 | bgp->maxmed_value = bgp->maxmed_onstartup_value; | |
864 | } | |
865 | ||
866 | /* Route announce to all peers should happen after this in | |
867 | * bgp_establish() */ | |
868 | } | |
869 | ||
870 | static void bgp_maxmed_onstartup_process_status_change(struct peer *peer) | |
871 | { | |
872 | if (peer->status == Established && !peer->bgp->established) { | |
873 | bgp_maxmed_onstartup_begin(peer->bgp); | |
874 | } | |
875 | } | |
876 | ||
877 | /* The update delay timer expiry callback. */ | |
878 | static int bgp_update_delay_timer(struct thread *thread) | |
879 | { | |
880 | struct bgp *bgp; | |
881 | ||
882 | zlog_info("Update delay ended - timer expired."); | |
883 | ||
884 | bgp = THREAD_ARG(thread); | |
885 | THREAD_TIMER_OFF(bgp->t_update_delay); | |
886 | bgp_update_delay_end(bgp); | |
887 | ||
888 | return 0; | |
889 | } | |
890 | ||
891 | /* The establish wait timer expiry callback. */ | |
892 | static int bgp_establish_wait_timer(struct thread *thread) | |
893 | { | |
894 | struct bgp *bgp; | |
895 | ||
896 | zlog_info("Establish wait - timer expired."); | |
897 | ||
898 | bgp = THREAD_ARG(thread); | |
899 | THREAD_TIMER_OFF(bgp->t_establish_wait); | |
900 | bgp_check_update_delay(bgp); | |
901 | ||
902 | return 0; | |
903 | } | |
904 | ||
905 | /* Steps to begin the update delay: | |
906 | - initialize queues if needed | |
907 | - stop the queue processing | |
908 | - start the timer */ | |
909 | static void bgp_update_delay_begin(struct bgp *bgp) | |
910 | { | |
911 | struct listnode *node, *nnode; | |
912 | struct peer *peer; | |
913 | ||
914 | /* Stop the processing of queued work. Enqueue shall continue */ | |
915 | work_queue_plug(bm->process_main_queue); | |
916 | ||
917 | for (ALL_LIST_ELEMENTS(bgp->peer, node, nnode, peer)) | |
918 | peer->update_delay_over = 0; | |
919 | ||
920 | /* Start the update-delay timer */ | |
921 | thread_add_timer(bm->master, bgp_update_delay_timer, bgp, | |
922 | bgp->v_update_delay, &bgp->t_update_delay); | |
923 | ||
924 | if (bgp->v_establish_wait != bgp->v_update_delay) | |
925 | thread_add_timer(bm->master, bgp_establish_wait_timer, bgp, | |
926 | bgp->v_establish_wait, &bgp->t_establish_wait); | |
927 | ||
928 | quagga_timestamp(3, bgp->update_delay_begin_time, | |
929 | sizeof(bgp->update_delay_begin_time)); | |
930 | } | |
931 | ||
932 | static void bgp_update_delay_process_status_change(struct peer *peer) | |
933 | { | |
934 | if (peer->status == Established) { | |
935 | if (!peer->bgp->established++) { | |
936 | bgp_update_delay_begin(peer->bgp); | |
937 | zlog_info( | |
938 | "Begin read-only mode - update-delay timer %d seconds", | |
939 | peer->bgp->v_update_delay); | |
940 | } | |
941 | if (CHECK_FLAG(peer->cap, PEER_CAP_RESTART_BIT_RCV)) | |
942 | bgp_update_restarted_peers(peer); | |
943 | } | |
944 | if (peer->ostatus == Established | |
945 | && bgp_update_delay_active(peer->bgp)) { | |
946 | /* Adjust the update-delay state to account for this flap. | |
947 | NOTE: Intentionally skipping adjusting implicit_eors or | |
948 | explicit_eors | |
949 | counters. Extra sanity check in bgp_check_update_delay() | |
950 | should | |
951 | be enough to take care of any additive discrepancy in bgp eor | |
952 | counters */ | |
953 | peer->bgp->established--; | |
954 | peer->update_delay_over = 0; | |
955 | } | |
956 | } | |
957 | ||
958 | /* Called after event occurred, this function change status and reset | |
959 | read/write and timer thread. */ | |
960 | void bgp_fsm_change_status(struct peer *peer, int status) | |
961 | { | |
962 | struct bgp *bgp; | |
963 | uint32_t peer_count; | |
964 | ||
965 | bgp = peer->bgp; | |
966 | peer_count = bgp->established_peers; | |
967 | ||
968 | if (status == Established) | |
969 | bgp->established_peers++; | |
970 | else if ((peer->status == Established) && (status != Established)) | |
971 | bgp->established_peers--; | |
972 | ||
973 | if (bgp_debug_neighbor_events(peer)) { | |
974 | struct vrf *vrf = vrf_lookup_by_id(bgp->vrf_id); | |
975 | ||
976 | zlog_debug("%s : vrf %s(%u), Status: %s established_peers %u", __func__, | |
977 | vrf ? vrf->name : "Unknown", bgp->vrf_id, | |
978 | lookup_msg(bgp_status_msg, status, NULL), | |
979 | bgp->established_peers); | |
980 | } | |
981 | ||
982 | /* Set to router ID to the value provided by RIB if there are no peers | |
983 | * in the established state and peer count did not change | |
984 | */ | |
985 | if ((peer_count != bgp->established_peers) && | |
986 | (bgp->established_peers == 0)) | |
987 | bgp_router_id_zebra_bump(bgp->vrf_id, NULL); | |
988 | ||
989 | /* Transition into Clearing or Deleted must /always/ clear all routes.. | |
990 | * (and must do so before actually changing into Deleted.. | |
991 | */ | |
992 | if (status >= Clearing) { | |
993 | bgp_clear_route_all(peer); | |
994 | ||
995 | /* If no route was queued for the clear-node processing, | |
996 | * generate the | |
997 | * completion event here. This is needed because if there are no | |
998 | * routes | |
999 | * to trigger the background clear-node thread, the event won't | |
1000 | * get | |
1001 | * generated and the peer would be stuck in Clearing. Note that | |
1002 | * this | |
1003 | * event is for the peer and helps the peer transition out of | |
1004 | * Clearing | |
1005 | * state; it should not be generated per (AFI,SAFI). The event | |
1006 | * is | |
1007 | * directly posted here without calling clear_node_complete() as | |
1008 | * we | |
1009 | * shouldn't do an extra unlock. This event will get processed | |
1010 | * after | |
1011 | * the state change that happens below, so peer will be in | |
1012 | * Clearing | |
1013 | * (or Deleted). | |
1014 | */ | |
1015 | if (!work_queue_is_scheduled(peer->clear_node_queue)) | |
1016 | BGP_EVENT_ADD(peer, Clearing_Completed); | |
1017 | } | |
1018 | ||
1019 | /* Preserve old status and change into new status. */ | |
1020 | peer->ostatus = peer->status; | |
1021 | peer->status = status; | |
1022 | ||
1023 | /* Save event that caused status change. */ | |
1024 | peer->last_major_event = peer->cur_event; | |
1025 | ||
1026 | /* Operations after status change */ | |
1027 | hook_call(peer_status_changed, peer); | |
1028 | ||
1029 | if (status == Established) | |
1030 | UNSET_FLAG(peer->sflags, PEER_STATUS_ACCEPT_PEER); | |
1031 | ||
1032 | /* If max-med processing is applicable, do the necessary. */ | |
1033 | if (status == Established) { | |
1034 | if (bgp_maxmed_onstartup_configured(peer->bgp) | |
1035 | && bgp_maxmed_onstartup_applicable(peer->bgp)) | |
1036 | bgp_maxmed_onstartup_process_status_change(peer); | |
1037 | else | |
1038 | peer->bgp->maxmed_onstartup_over = 1; | |
1039 | } | |
1040 | ||
1041 | /* If update-delay processing is applicable, do the necessary. */ | |
1042 | if (bgp_update_delay_configured(peer->bgp) | |
1043 | && bgp_update_delay_applicable(peer->bgp)) | |
1044 | bgp_update_delay_process_status_change(peer); | |
1045 | ||
1046 | if (bgp_debug_neighbor_events(peer)) | |
1047 | zlog_debug("%s went from %s to %s", peer->host, | |
1048 | lookup_msg(bgp_status_msg, peer->ostatus, NULL), | |
1049 | lookup_msg(bgp_status_msg, peer->status, NULL)); | |
1050 | } | |
1051 | ||
1052 | /* Flush the event queue and ensure the peer is shut down */ | |
1053 | static int bgp_clearing_completed(struct peer *peer) | |
1054 | { | |
1055 | int rc = bgp_stop(peer); | |
1056 | ||
1057 | if (rc >= 0) | |
1058 | BGP_EVENT_FLUSH(peer); | |
1059 | ||
1060 | return rc; | |
1061 | } | |
1062 | ||
1063 | /* Administrative BGP peer stop event. */ | |
1064 | /* May be called multiple times for the same peer */ | |
1065 | int bgp_stop(struct peer *peer) | |
1066 | { | |
1067 | afi_t afi; | |
1068 | safi_t safi; | |
1069 | char orf_name[BUFSIZ]; | |
1070 | int ret = 0; | |
1071 | ||
1072 | if (peer_dynamic_neighbor(peer) | |
1073 | && !(CHECK_FLAG(peer->flags, PEER_FLAG_DELETE))) { | |
1074 | if (bgp_debug_neighbor_events(peer)) | |
1075 | zlog_debug("%s (dynamic neighbor) deleted", peer->host); | |
1076 | peer_delete(peer); | |
1077 | return -1; | |
1078 | } | |
1079 | ||
1080 | /* Can't do this in Clearing; events are used for state transitions */ | |
1081 | if (peer->status != Clearing) { | |
1082 | /* Delete all existing events of the peer */ | |
1083 | BGP_EVENT_FLUSH(peer); | |
1084 | } | |
1085 | ||
1086 | /* Increment Dropped count. */ | |
1087 | if (peer->status == Established) { | |
1088 | peer->dropped++; | |
1089 | ||
1090 | /* bgp log-neighbor-changes of neighbor Down */ | |
1091 | if (bgp_flag_check(peer->bgp, BGP_FLAG_LOG_NEIGHBOR_CHANGES)) { | |
1092 | struct vrf *vrf = vrf_lookup_by_id(peer->bgp->vrf_id); | |
1093 | zlog_info( | |
1094 | "%%ADJCHANGE: neighbor %s(%s) in vrf %s Down %s", | |
1095 | peer->host, | |
1096 | (peer->hostname) ? peer->hostname : "Unknown", | |
1097 | vrf ? ((vrf->vrf_id != VRF_DEFAULT) | |
1098 | ? vrf->name | |
1099 | : VRF_DEFAULT_NAME) | |
1100 | : "", | |
1101 | peer_down_str[(int)peer->last_reset]); | |
1102 | } | |
1103 | ||
1104 | /* graceful restart */ | |
1105 | if (peer->t_gr_stale) { | |
1106 | BGP_TIMER_OFF(peer->t_gr_stale); | |
1107 | if (bgp_debug_neighbor_events(peer)) | |
1108 | zlog_debug( | |
1109 | "%s graceful restart stalepath timer stopped", | |
1110 | peer->host); | |
1111 | } | |
1112 | if (CHECK_FLAG(peer->sflags, PEER_STATUS_NSF_WAIT)) { | |
1113 | if (bgp_debug_neighbor_events(peer)) { | |
1114 | zlog_debug( | |
1115 | "%s graceful restart timer started for %d sec", | |
1116 | peer->host, peer->v_gr_restart); | |
1117 | zlog_debug( | |
1118 | "%s graceful restart stalepath timer started for %d sec", | |
1119 | peer->host, peer->bgp->stalepath_time); | |
1120 | } | |
1121 | BGP_TIMER_ON(peer->t_gr_restart, | |
1122 | bgp_graceful_restart_timer_expire, | |
1123 | peer->v_gr_restart); | |
1124 | BGP_TIMER_ON(peer->t_gr_stale, | |
1125 | bgp_graceful_stale_timer_expire, | |
1126 | peer->bgp->stalepath_time); | |
1127 | } else { | |
1128 | UNSET_FLAG(peer->sflags, PEER_STATUS_NSF_MODE); | |
1129 | ||
1130 | for (afi = AFI_IP; afi < AFI_MAX; afi++) | |
1131 | for (safi = SAFI_UNICAST; safi <= SAFI_MPLS_VPN; | |
1132 | safi++) | |
1133 | peer->nsf[afi][safi] = 0; | |
1134 | } | |
1135 | ||
1136 | /* set last reset time */ | |
1137 | peer->resettime = peer->uptime = bgp_clock(); | |
1138 | ||
1139 | if (BGP_DEBUG(update_groups, UPDATE_GROUPS)) | |
1140 | zlog_debug("%s remove from all update group", | |
1141 | peer->host); | |
1142 | update_group_remove_peer_afs(peer); | |
1143 | ||
1144 | hook_call(peer_backward_transition, peer); | |
1145 | ||
1146 | /* Reset peer synctime */ | |
1147 | peer->synctime = 0; | |
1148 | } | |
1149 | ||
1150 | /* stop keepalives */ | |
1151 | bgp_keepalives_off(peer); | |
1152 | ||
1153 | /* Stop read and write threads. */ | |
1154 | bgp_writes_off(peer); | |
1155 | bgp_reads_off(peer); | |
1156 | ||
1157 | THREAD_OFF(peer->t_connect_check_r); | |
1158 | THREAD_OFF(peer->t_connect_check_w); | |
1159 | ||
1160 | /* Stop all timers. */ | |
1161 | BGP_TIMER_OFF(peer->t_start); | |
1162 | BGP_TIMER_OFF(peer->t_connect); | |
1163 | BGP_TIMER_OFF(peer->t_holdtime); | |
1164 | BGP_TIMER_OFF(peer->t_routeadv); | |
1165 | ||
1166 | /* Clear input and output buffer. */ | |
1167 | frr_with_mutex(&peer->io_mtx) { | |
1168 | if (peer->ibuf) | |
1169 | stream_fifo_clean(peer->ibuf); | |
1170 | if (peer->obuf) | |
1171 | stream_fifo_clean(peer->obuf); | |
1172 | ||
1173 | if (peer->ibuf_work) | |
1174 | ringbuf_wipe(peer->ibuf_work); | |
1175 | if (peer->obuf_work) | |
1176 | stream_reset(peer->obuf_work); | |
1177 | ||
1178 | if (peer->curr) { | |
1179 | stream_free(peer->curr); | |
1180 | peer->curr = NULL; | |
1181 | } | |
1182 | } | |
1183 | ||
1184 | /* Close of file descriptor. */ | |
1185 | if (peer->fd >= 0) { | |
1186 | close(peer->fd); | |
1187 | peer->fd = -1; | |
1188 | } | |
1189 | ||
1190 | FOREACH_AFI_SAFI (afi, safi) { | |
1191 | /* Reset all negotiated variables */ | |
1192 | peer->afc_nego[afi][safi] = 0; | |
1193 | peer->afc_adv[afi][safi] = 0; | |
1194 | peer->afc_recv[afi][safi] = 0; | |
1195 | ||
1196 | /* peer address family capability flags*/ | |
1197 | peer->af_cap[afi][safi] = 0; | |
1198 | ||
1199 | /* peer address family status flags*/ | |
1200 | peer->af_sflags[afi][safi] = 0; | |
1201 | ||
1202 | /* Received ORF prefix-filter */ | |
1203 | peer->orf_plist[afi][safi] = NULL; | |
1204 | ||
1205 | if ((peer->status == OpenConfirm) | |
1206 | || (peer->status == Established)) { | |
1207 | /* ORF received prefix-filter pnt */ | |
1208 | sprintf(orf_name, "%s.%d.%d", peer->host, afi, safi); | |
1209 | prefix_bgp_orf_remove_all(afi, orf_name); | |
1210 | } | |
1211 | } | |
1212 | ||
1213 | /* Reset keepalive and holdtime */ | |
1214 | if (CHECK_FLAG(peer->flags, PEER_FLAG_TIMER)) { | |
1215 | peer->v_keepalive = peer->keepalive; | |
1216 | peer->v_holdtime = peer->holdtime; | |
1217 | } else { | |
1218 | peer->v_keepalive = peer->bgp->default_keepalive; | |
1219 | peer->v_holdtime = peer->bgp->default_holdtime; | |
1220 | } | |
1221 | ||
1222 | peer->update_time = 0; | |
1223 | ||
1224 | /* Until we are sure that there is no problem about prefix count | |
1225 | this should be commented out.*/ | |
1226 | #if 0 | |
1227 | /* Reset prefix count */ | |
1228 | peer->pcount[AFI_IP][SAFI_UNICAST] = 0; | |
1229 | peer->pcount[AFI_IP][SAFI_MULTICAST] = 0; | |
1230 | peer->pcount[AFI_IP][SAFI_LABELED_UNICAST] = 0; | |
1231 | peer->pcount[AFI_IP][SAFI_MPLS_VPN] = 0; | |
1232 | peer->pcount[AFI_IP6][SAFI_UNICAST] = 0; | |
1233 | peer->pcount[AFI_IP6][SAFI_MULTICAST] = 0; | |
1234 | peer->pcount[AFI_IP6][SAFI_LABELED_UNICAST] = 0; | |
1235 | #endif /* 0 */ | |
1236 | ||
1237 | if (!CHECK_FLAG(peer->flags, PEER_FLAG_CONFIG_NODE) | |
1238 | && !(CHECK_FLAG(peer->flags, PEER_FLAG_DELETE))) { | |
1239 | peer_delete(peer); | |
1240 | ret = -1; | |
1241 | } else { | |
1242 | bgp_peer_conf_if_to_su_update(peer); | |
1243 | } | |
1244 | ||
1245 | return ret; | |
1246 | } | |
1247 | ||
1248 | /* BGP peer is stoped by the error. */ | |
1249 | static int bgp_stop_with_error(struct peer *peer) | |
1250 | { | |
1251 | /* Double start timer. */ | |
1252 | peer->v_start *= 2; | |
1253 | ||
1254 | /* Overflow check. */ | |
1255 | if (peer->v_start >= (60 * 2)) | |
1256 | peer->v_start = (60 * 2); | |
1257 | ||
1258 | if (peer_dynamic_neighbor(peer)) { | |
1259 | if (bgp_debug_neighbor_events(peer)) | |
1260 | zlog_debug("%s (dynamic neighbor) deleted", peer->host); | |
1261 | peer_delete(peer); | |
1262 | return -1; | |
1263 | } | |
1264 | ||
1265 | return (bgp_stop(peer)); | |
1266 | } | |
1267 | ||
1268 | ||
1269 | /* something went wrong, send notify and tear down */ | |
1270 | static int bgp_stop_with_notify(struct peer *peer, uint8_t code, | |
1271 | uint8_t sub_code) | |
1272 | { | |
1273 | /* Send notify to remote peer */ | |
1274 | bgp_notify_send(peer, code, sub_code); | |
1275 | ||
1276 | if (peer_dynamic_neighbor(peer)) { | |
1277 | if (bgp_debug_neighbor_events(peer)) | |
1278 | zlog_debug("%s (dynamic neighbor) deleted", peer->host); | |
1279 | peer_delete(peer); | |
1280 | return -1; | |
1281 | } | |
1282 | ||
1283 | /* Clear start timer value to default. */ | |
1284 | peer->v_start = BGP_INIT_START_TIMER; | |
1285 | ||
1286 | return (bgp_stop(peer)); | |
1287 | } | |
1288 | ||
1289 | /** | |
1290 | * Determines whether a TCP session has successfully established for a peer and | |
1291 | * events as appropriate. | |
1292 | * | |
1293 | * This function is called when setting up a new session. After connect() is | |
1294 | * called on the peer's socket (in bgp_start()), the fd is passed to poll() | |
1295 | * to wait for connection success or failure. When poll() returns, this | |
1296 | * function is called to evaluate the result. | |
1297 | * | |
1298 | * Due to differences in behavior of poll() on Linux and BSD - specifically, | |
1299 | * the value of .revents in the case of a closed connection - this function is | |
1300 | * scheduled both for a read and a write event. The write event is triggered | |
1301 | * when the connection is established. A read event is triggered when the | |
1302 | * connection is closed. Thus we need to cancel whichever one did not occur. | |
1303 | */ | |
1304 | static int bgp_connect_check(struct thread *thread) | |
1305 | { | |
1306 | int status; | |
1307 | socklen_t slen; | |
1308 | int ret; | |
1309 | struct peer *peer; | |
1310 | ||
1311 | peer = THREAD_ARG(thread); | |
1312 | assert(!CHECK_FLAG(peer->thread_flags, PEER_THREAD_READS_ON)); | |
1313 | assert(!CHECK_FLAG(peer->thread_flags, PEER_THREAD_WRITES_ON)); | |
1314 | assert(!peer->t_read); | |
1315 | assert(!peer->t_write); | |
1316 | ||
1317 | THREAD_OFF(peer->t_connect_check_r); | |
1318 | THREAD_OFF(peer->t_connect_check_w); | |
1319 | ||
1320 | /* Check file descriptor. */ | |
1321 | slen = sizeof(status); | |
1322 | ret = getsockopt(peer->fd, SOL_SOCKET, SO_ERROR, (void *)&status, | |
1323 | &slen); | |
1324 | ||
1325 | /* If getsockopt is fail, this is fatal error. */ | |
1326 | if (ret < 0) { | |
1327 | zlog_err("can't get sockopt for nonblocking connect: %d(%s)", | |
1328 | errno, safe_strerror(errno)); | |
1329 | BGP_EVENT_ADD(peer, TCP_fatal_error); | |
1330 | return -1; | |
1331 | } | |
1332 | ||
1333 | /* When status is 0 then TCP connection is established. */ | |
1334 | if (status == 0) { | |
1335 | BGP_EVENT_ADD(peer, TCP_connection_open); | |
1336 | return 1; | |
1337 | } else { | |
1338 | if (bgp_debug_neighbor_events(peer)) | |
1339 | zlog_debug("%s [Event] Connect failed %d(%s)", | |
1340 | peer->host, status, safe_strerror(status)); | |
1341 | BGP_EVENT_ADD(peer, TCP_connection_open_failed); | |
1342 | return 0; | |
1343 | } | |
1344 | } | |
1345 | ||
1346 | /* TCP connection open. Next we send open message to remote peer. And | |
1347 | add read thread for reading open message. */ | |
1348 | static int bgp_connect_success(struct peer *peer) | |
1349 | { | |
1350 | if (peer->fd < 0) { | |
1351 | flog_err(EC_BGP_CONNECT, | |
1352 | "bgp_connect_success peer's fd is negative value %d", | |
1353 | peer->fd); | |
1354 | bgp_stop(peer); | |
1355 | return -1; | |
1356 | } | |
1357 | ||
1358 | if (bgp_getsockname(peer) < 0) { | |
1359 | flog_err_sys(EC_LIB_SOCKET, | |
1360 | "%s: bgp_getsockname(): failed for peer %s, fd %d", | |
1361 | __FUNCTION__, peer->host, peer->fd); | |
1362 | bgp_notify_send( | |
1363 | peer, BGP_NOTIFY_FSM_ERR, | |
1364 | BGP_NOTIFY_SUBCODE_UNSPECIFIC); /* internal error */ | |
1365 | bgp_writes_on(peer); | |
1366 | return -1; | |
1367 | } | |
1368 | ||
1369 | bgp_reads_on(peer); | |
1370 | ||
1371 | if (bgp_debug_neighbor_events(peer)) { | |
1372 | char buf1[SU_ADDRSTRLEN]; | |
1373 | ||
1374 | if (!CHECK_FLAG(peer->sflags, PEER_STATUS_ACCEPT_PEER)) | |
1375 | zlog_debug("%s open active, local address %s", | |
1376 | peer->host, | |
1377 | sockunion2str(peer->su_local, buf1, | |
1378 | SU_ADDRSTRLEN)); | |
1379 | else | |
1380 | zlog_debug("%s passive open", peer->host); | |
1381 | } | |
1382 | ||
1383 | bgp_open_send(peer); | |
1384 | ||
1385 | return 0; | |
1386 | } | |
1387 | ||
1388 | /* TCP connect fail */ | |
1389 | static int bgp_connect_fail(struct peer *peer) | |
1390 | { | |
1391 | if (peer_dynamic_neighbor(peer)) { | |
1392 | if (bgp_debug_neighbor_events(peer)) | |
1393 | zlog_debug("%s (dynamic neighbor) deleted", peer->host); | |
1394 | peer_delete(peer); | |
1395 | return -1; | |
1396 | } | |
1397 | ||
1398 | return (bgp_stop(peer)); | |
1399 | } | |
1400 | ||
1401 | /* This function is the first starting point of all BGP connection. It | |
1402 | try to connect to remote peer with non-blocking IO. */ | |
1403 | int bgp_start(struct peer *peer) | |
1404 | { | |
1405 | int status; | |
1406 | ||
1407 | bgp_peer_conf_if_to_su_update(peer); | |
1408 | ||
1409 | if (peer->su.sa.sa_family == AF_UNSPEC) { | |
1410 | if (bgp_debug_neighbor_events(peer)) | |
1411 | zlog_debug( | |
1412 | "%s [FSM] Unable to get neighbor's IP address, waiting...", | |
1413 | peer->host); | |
1414 | peer->last_reset = PEER_DOWN_NBR_ADDR; | |
1415 | return -1; | |
1416 | } | |
1417 | ||
1418 | if (BGP_PEER_START_SUPPRESSED(peer)) { | |
1419 | if (bgp_debug_neighbor_events(peer)) | |
1420 | flog_err(EC_BGP_FSM, | |
1421 | "%s [FSM] Trying to start suppressed peer" | |
1422 | " - this is never supposed to happen!", | |
1423 | peer->host); | |
1424 | return -1; | |
1425 | } | |
1426 | ||
1427 | /* Scrub some information that might be left over from a previous, | |
1428 | * session | |
1429 | */ | |
1430 | /* Connection information. */ | |
1431 | if (peer->su_local) { | |
1432 | sockunion_free(peer->su_local); | |
1433 | peer->su_local = NULL; | |
1434 | } | |
1435 | ||
1436 | if (peer->su_remote) { | |
1437 | sockunion_free(peer->su_remote); | |
1438 | peer->su_remote = NULL; | |
1439 | } | |
1440 | ||
1441 | /* Clear remote router-id. */ | |
1442 | peer->remote_id.s_addr = 0; | |
1443 | ||
1444 | /* Clear peer capability flag. */ | |
1445 | peer->cap = 0; | |
1446 | ||
1447 | /* If the peer is passive mode, force to move to Active mode. */ | |
1448 | if (CHECK_FLAG(peer->flags, PEER_FLAG_PASSIVE)) { | |
1449 | BGP_EVENT_ADD(peer, TCP_connection_open_failed); | |
1450 | return 0; | |
1451 | } | |
1452 | ||
1453 | if (peer->bgp->inst_type != BGP_INSTANCE_TYPE_VIEW && | |
1454 | peer->bgp->vrf_id == VRF_UNKNOWN) { | |
1455 | if (bgp_debug_neighbor_events(peer)) | |
1456 | flog_err( | |
1457 | EC_BGP_FSM, | |
1458 | "%s [FSM] In a VRF that is not initialised yet", | |
1459 | peer->host); | |
1460 | peer->last_reset = PEER_DOWN_VRF_UNINIT; | |
1461 | return -1; | |
1462 | } | |
1463 | ||
1464 | /* Register peer for NHT. If next hop is already resolved, proceed | |
1465 | * with connection setup, else wait. | |
1466 | */ | |
1467 | if (!bgp_peer_reg_with_nht(peer)) { | |
1468 | if (bgp_zebra_num_connects()) { | |
1469 | if (bgp_debug_neighbor_events(peer)) | |
1470 | zlog_debug("%s [FSM] Waiting for NHT", | |
1471 | peer->host); | |
1472 | peer->last_reset = PEER_DOWN_WAITING_NHT; | |
1473 | BGP_EVENT_ADD(peer, TCP_connection_open_failed); | |
1474 | return 0; | |
1475 | } | |
1476 | } | |
1477 | ||
1478 | assert(!peer->t_write); | |
1479 | assert(!peer->t_read); | |
1480 | assert(!CHECK_FLAG(peer->thread_flags, PEER_THREAD_WRITES_ON)); | |
1481 | assert(!CHECK_FLAG(peer->thread_flags, PEER_THREAD_READS_ON)); | |
1482 | status = bgp_connect(peer); | |
1483 | ||
1484 | switch (status) { | |
1485 | case connect_error: | |
1486 | if (bgp_debug_neighbor_events(peer)) | |
1487 | zlog_debug("%s [FSM] Connect error", peer->host); | |
1488 | BGP_EVENT_ADD(peer, TCP_connection_open_failed); | |
1489 | break; | |
1490 | case connect_success: | |
1491 | if (bgp_debug_neighbor_events(peer)) | |
1492 | zlog_debug( | |
1493 | "%s [FSM] Connect immediately success, fd %d", | |
1494 | peer->host, peer->fd); | |
1495 | BGP_EVENT_ADD(peer, TCP_connection_open); | |
1496 | break; | |
1497 | case connect_in_progress: | |
1498 | /* To check nonblocking connect, we wait until socket is | |
1499 | readable or writable. */ | |
1500 | if (bgp_debug_neighbor_events(peer)) | |
1501 | zlog_debug( | |
1502 | "%s [FSM] Non blocking connect waiting result, fd %d", | |
1503 | peer->host, peer->fd); | |
1504 | if (peer->fd < 0) { | |
1505 | flog_err(EC_BGP_FSM, | |
1506 | "bgp_start peer's fd is negative value %d", | |
1507 | peer->fd); | |
1508 | return -1; | |
1509 | } | |
1510 | /* | |
1511 | * - when the socket becomes ready, poll() will signify POLLOUT | |
1512 | * - if it fails to connect, poll() will signify POLLHUP | |
1513 | * - POLLHUP is handled as a 'read' event by thread.c | |
1514 | * | |
1515 | * therefore, we schedule both a read and a write event with | |
1516 | * bgp_connect_check() as the handler for each and cancel the | |
1517 | * unused event in that function. | |
1518 | */ | |
1519 | thread_add_read(bm->master, bgp_connect_check, peer, peer->fd, | |
1520 | &peer->t_connect_check_r); | |
1521 | thread_add_write(bm->master, bgp_connect_check, peer, peer->fd, | |
1522 | &peer->t_connect_check_w); | |
1523 | break; | |
1524 | } | |
1525 | return 0; | |
1526 | } | |
1527 | ||
1528 | /* Connect retry timer is expired when the peer status is Connect. */ | |
1529 | static int bgp_reconnect(struct peer *peer) | |
1530 | { | |
1531 | if (bgp_stop(peer) < 0) | |
1532 | return -1; | |
1533 | ||
1534 | bgp_start(peer); | |
1535 | return 0; | |
1536 | } | |
1537 | ||
1538 | static int bgp_fsm_open(struct peer *peer) | |
1539 | { | |
1540 | /* Send keepalive and make keepalive timer */ | |
1541 | bgp_keepalive_send(peer); | |
1542 | ||
1543 | /* Reset holdtimer value. */ | |
1544 | BGP_TIMER_OFF(peer->t_holdtime); | |
1545 | ||
1546 | return 0; | |
1547 | } | |
1548 | ||
1549 | /* FSM error, unexpected event. This is error of BGP connection. So cut the | |
1550 | peer and change to Idle status. */ | |
1551 | static int bgp_fsm_event_error(struct peer *peer) | |
1552 | { | |
1553 | flog_err(EC_BGP_FSM, "%s [FSM] unexpected packet received in state %s", | |
1554 | peer->host, lookup_msg(bgp_status_msg, peer->status, NULL)); | |
1555 | ||
1556 | return bgp_stop_with_notify(peer, BGP_NOTIFY_FSM_ERR, 0); | |
1557 | } | |
1558 | ||
1559 | /* Hold timer expire. This is error of BGP connection. So cut the | |
1560 | peer and change to Idle status. */ | |
1561 | static int bgp_fsm_holdtime_expire(struct peer *peer) | |
1562 | { | |
1563 | if (bgp_debug_neighbor_events(peer)) | |
1564 | zlog_debug("%s [FSM] Hold timer expire", peer->host); | |
1565 | ||
1566 | return bgp_stop_with_notify(peer, BGP_NOTIFY_HOLD_ERR, 0); | |
1567 | } | |
1568 | ||
1569 | /** | |
1570 | * Transition to Established state. | |
1571 | * | |
1572 | * Convert peer from stub to full fledged peer, set some timers, and generate | |
1573 | * initial updates. | |
1574 | */ | |
1575 | static int bgp_establish(struct peer *peer) | |
1576 | { | |
1577 | afi_t afi; | |
1578 | safi_t safi; | |
1579 | int nsf_af_count = 0; | |
1580 | int ret = 0; | |
1581 | struct peer *other; | |
1582 | ||
1583 | other = peer->doppelganger; | |
1584 | peer = peer_xfer_conn(peer); | |
1585 | if (!peer) { | |
1586 | flog_err(EC_BGP_CONNECT, "%%Neighbor failed in xfer_conn"); | |
1587 | return -1; | |
1588 | } | |
1589 | ||
1590 | if (other == peer) | |
1591 | ret = 1; /* bgp_establish specific code when xfer_conn | |
1592 | happens. */ | |
1593 | ||
1594 | /* Reset capability open status flag. */ | |
1595 | if (!CHECK_FLAG(peer->sflags, PEER_STATUS_CAPABILITY_OPEN)) | |
1596 | SET_FLAG(peer->sflags, PEER_STATUS_CAPABILITY_OPEN); | |
1597 | ||
1598 | /* Clear start timer value to default. */ | |
1599 | peer->v_start = BGP_INIT_START_TIMER; | |
1600 | ||
1601 | /* Increment established count. */ | |
1602 | peer->established++; | |
1603 | bgp_fsm_change_status(peer, Established); | |
1604 | ||
1605 | /* bgp log-neighbor-changes of neighbor Up */ | |
1606 | if (bgp_flag_check(peer->bgp, BGP_FLAG_LOG_NEIGHBOR_CHANGES)) { | |
1607 | struct vrf *vrf = vrf_lookup_by_id(peer->bgp->vrf_id); | |
1608 | zlog_info("%%ADJCHANGE: neighbor %s(%s) in vrf %s Up", | |
1609 | peer->host, | |
1610 | (peer->hostname) ? peer->hostname : "Unknown", | |
1611 | vrf ? ((vrf->vrf_id != VRF_DEFAULT) | |
1612 | ? vrf->name | |
1613 | : VRF_DEFAULT_NAME) | |
1614 | : ""); | |
1615 | } | |
1616 | /* assign update-group/subgroup */ | |
1617 | update_group_adjust_peer_afs(peer); | |
1618 | ||
1619 | /* graceful restart */ | |
1620 | UNSET_FLAG(peer->sflags, PEER_STATUS_NSF_WAIT); | |
1621 | for (afi = AFI_IP; afi < AFI_MAX; afi++) | |
1622 | for (safi = SAFI_UNICAST; safi <= SAFI_MPLS_VPN; safi++) { | |
1623 | if (peer->afc_nego[afi][safi] | |
1624 | && CHECK_FLAG(peer->cap, PEER_CAP_RESTART_ADV) | |
1625 | && CHECK_FLAG(peer->af_cap[afi][safi], | |
1626 | PEER_CAP_RESTART_AF_RCV)) { | |
1627 | if (peer->nsf[afi][safi] | |
1628 | && !CHECK_FLAG( | |
1629 | peer->af_cap[afi][safi], | |
1630 | PEER_CAP_RESTART_AF_PRESERVE_RCV)) | |
1631 | bgp_clear_stale_route(peer, afi, safi); | |
1632 | ||
1633 | peer->nsf[afi][safi] = 1; | |
1634 | nsf_af_count++; | |
1635 | } else { | |
1636 | if (peer->nsf[afi][safi]) | |
1637 | bgp_clear_stale_route(peer, afi, safi); | |
1638 | peer->nsf[afi][safi] = 0; | |
1639 | } | |
1640 | } | |
1641 | ||
1642 | if (nsf_af_count) | |
1643 | SET_FLAG(peer->sflags, PEER_STATUS_NSF_MODE); | |
1644 | else { | |
1645 | UNSET_FLAG(peer->sflags, PEER_STATUS_NSF_MODE); | |
1646 | if (peer->t_gr_stale) { | |
1647 | BGP_TIMER_OFF(peer->t_gr_stale); | |
1648 | if (bgp_debug_neighbor_events(peer)) | |
1649 | zlog_debug( | |
1650 | "%s graceful restart stalepath timer stopped", | |
1651 | peer->host); | |
1652 | } | |
1653 | } | |
1654 | ||
1655 | if (peer->t_gr_restart) { | |
1656 | BGP_TIMER_OFF(peer->t_gr_restart); | |
1657 | if (bgp_debug_neighbor_events(peer)) | |
1658 | zlog_debug("%s graceful restart timer stopped", | |
1659 | peer->host); | |
1660 | } | |
1661 | ||
1662 | /* Reset uptime, turn on keepalives, send current table. */ | |
1663 | if (!peer->v_holdtime) | |
1664 | bgp_keepalives_on(peer); | |
1665 | ||
1666 | peer->uptime = bgp_clock(); | |
1667 | ||
1668 | /* Send route-refresh when ORF is enabled */ | |
1669 | FOREACH_AFI_SAFI (afi, safi) { | |
1670 | if (CHECK_FLAG(peer->af_cap[afi][safi], | |
1671 | PEER_CAP_ORF_PREFIX_SM_ADV)) { | |
1672 | if (CHECK_FLAG(peer->af_cap[afi][safi], | |
1673 | PEER_CAP_ORF_PREFIX_RM_RCV)) | |
1674 | bgp_route_refresh_send(peer, afi, safi, | |
1675 | ORF_TYPE_PREFIX, | |
1676 | REFRESH_IMMEDIATE, 0); | |
1677 | else if (CHECK_FLAG(peer->af_cap[afi][safi], | |
1678 | PEER_CAP_ORF_PREFIX_RM_OLD_RCV)) | |
1679 | bgp_route_refresh_send(peer, afi, safi, | |
1680 | ORF_TYPE_PREFIX_OLD, | |
1681 | REFRESH_IMMEDIATE, 0); | |
1682 | } | |
1683 | } | |
1684 | ||
1685 | /* First update is deferred until ORF or ROUTE-REFRESH is received */ | |
1686 | FOREACH_AFI_SAFI (afi, safi) { | |
1687 | if (CHECK_FLAG(peer->af_cap[afi][safi], | |
1688 | PEER_CAP_ORF_PREFIX_RM_ADV)) | |
1689 | if (CHECK_FLAG(peer->af_cap[afi][safi], | |
1690 | PEER_CAP_ORF_PREFIX_SM_RCV) | |
1691 | || CHECK_FLAG(peer->af_cap[afi][safi], | |
1692 | PEER_CAP_ORF_PREFIX_SM_OLD_RCV)) | |
1693 | SET_FLAG(peer->af_sflags[afi][safi], | |
1694 | PEER_STATUS_ORF_WAIT_REFRESH); | |
1695 | } | |
1696 | ||
1697 | bgp_announce_peer(peer); | |
1698 | ||
1699 | /* Start the route advertisement timer to send updates to the peer - if | |
1700 | * BGP | |
1701 | * is not in read-only mode. If it is, the timer will be started at the | |
1702 | * end | |
1703 | * of read-only mode. | |
1704 | */ | |
1705 | if (!bgp_update_delay_active(peer->bgp)) { | |
1706 | BGP_TIMER_OFF(peer->t_routeadv); | |
1707 | BGP_TIMER_ON(peer->t_routeadv, bgp_routeadv_timer, 0); | |
1708 | } | |
1709 | ||
1710 | if (peer->doppelganger && (peer->doppelganger->status != Deleted)) { | |
1711 | if (bgp_debug_neighbor_events(peer)) | |
1712 | zlog_debug( | |
1713 | "[Event] Deleting stub connection for peer %s", | |
1714 | peer->host); | |
1715 | ||
1716 | if (peer->doppelganger->status > Active) | |
1717 | bgp_notify_send(peer->doppelganger, BGP_NOTIFY_CEASE, | |
1718 | BGP_NOTIFY_CEASE_COLLISION_RESOLUTION); | |
1719 | else | |
1720 | peer_delete(peer->doppelganger); | |
1721 | } | |
1722 | ||
1723 | /* | |
1724 | * If we are replacing the old peer for a doppelganger | |
1725 | * then switch it around in the bgp->peerhash | |
1726 | * the doppelgangers su and this peer's su are the same | |
1727 | * so the hash_release is the same for either. | |
1728 | */ | |
1729 | hash_release(peer->bgp->peerhash, peer); | |
1730 | hash_get(peer->bgp->peerhash, peer, hash_alloc_intern); | |
1731 | ||
1732 | bgp_bfd_register_peer(peer); | |
1733 | return ret; | |
1734 | } | |
1735 | ||
1736 | /* Keepalive packet is received. */ | |
1737 | static int bgp_fsm_keepalive(struct peer *peer) | |
1738 | { | |
1739 | BGP_TIMER_OFF(peer->t_holdtime); | |
1740 | return 0; | |
1741 | } | |
1742 | ||
1743 | /* Update packet is received. */ | |
1744 | static int bgp_fsm_update(struct peer *peer) | |
1745 | { | |
1746 | BGP_TIMER_OFF(peer->t_holdtime); | |
1747 | return 0; | |
1748 | } | |
1749 | ||
1750 | /* This is empty event. */ | |
1751 | static int bgp_ignore(struct peer *peer) | |
1752 | { | |
1753 | flog_err( | |
1754 | EC_BGP_FSM, | |
1755 | "%s [FSM] Ignoring event %s in state %s, prior events %s, %s, fd %d", | |
1756 | peer->host, bgp_event_str[peer->cur_event], | |
1757 | lookup_msg(bgp_status_msg, peer->status, NULL), | |
1758 | bgp_event_str[peer->last_event], | |
1759 | bgp_event_str[peer->last_major_event], peer->fd); | |
1760 | return 0; | |
1761 | } | |
1762 | ||
1763 | /* This is to handle unexpected events.. */ | |
1764 | static int bgp_fsm_exeption(struct peer *peer) | |
1765 | { | |
1766 | flog_err( | |
1767 | EC_BGP_FSM, | |
1768 | "%s [FSM] Unexpected event %s in state %s, prior events %s, %s, fd %d", | |
1769 | peer->host, bgp_event_str[peer->cur_event], | |
1770 | lookup_msg(bgp_status_msg, peer->status, NULL), | |
1771 | bgp_event_str[peer->last_event], | |
1772 | bgp_event_str[peer->last_major_event], peer->fd); | |
1773 | return (bgp_stop(peer)); | |
1774 | } | |
1775 | ||
1776 | void bgp_fsm_event_update(struct peer *peer, int valid) | |
1777 | { | |
1778 | if (!peer) | |
1779 | return; | |
1780 | ||
1781 | switch (peer->status) { | |
1782 | case Idle: | |
1783 | if (valid) | |
1784 | BGP_EVENT_ADD(peer, BGP_Start); | |
1785 | break; | |
1786 | case Connect: | |
1787 | if (!valid) { | |
1788 | BGP_TIMER_OFF(peer->t_connect); | |
1789 | BGP_EVENT_ADD(peer, TCP_fatal_error); | |
1790 | } | |
1791 | break; | |
1792 | case Active: | |
1793 | if (valid) { | |
1794 | BGP_TIMER_OFF(peer->t_connect); | |
1795 | BGP_EVENT_ADD(peer, ConnectRetry_timer_expired); | |
1796 | } | |
1797 | break; | |
1798 | case OpenSent: | |
1799 | case OpenConfirm: | |
1800 | case Established: | |
1801 | if (!valid && (peer->gtsm_hops == 1)) | |
1802 | BGP_EVENT_ADD(peer, TCP_fatal_error); | |
1803 | case Clearing: | |
1804 | case Deleted: | |
1805 | default: | |
1806 | break; | |
1807 | } | |
1808 | } | |
1809 | ||
1810 | /* Finite State Machine structure */ | |
1811 | static const struct { | |
1812 | int (*func)(struct peer *); | |
1813 | int next_state; | |
1814 | } FSM[BGP_STATUS_MAX - 1][BGP_EVENTS_MAX - 1] = { | |
1815 | { | |
1816 | /* Idle state: In Idle state, all events other than BGP_Start is | |
1817 | ignored. With BGP_Start event, finite state machine calls | |
1818 | bgp_start(). */ | |
1819 | {bgp_start, Connect}, /* BGP_Start */ | |
1820 | {bgp_stop, Idle}, /* BGP_Stop */ | |
1821 | {bgp_stop, Idle}, /* TCP_connection_open */ | |
1822 | {bgp_stop, Idle}, /* TCP_connection_closed */ | |
1823 | {bgp_ignore, Idle}, /* TCP_connection_open_failed */ | |
1824 | {bgp_stop, Idle}, /* TCP_fatal_error */ | |
1825 | {bgp_ignore, Idle}, /* ConnectRetry_timer_expired */ | |
1826 | {bgp_ignore, Idle}, /* Hold_Timer_expired */ | |
1827 | {bgp_ignore, Idle}, /* KeepAlive_timer_expired */ | |
1828 | {bgp_ignore, Idle}, /* Receive_OPEN_message */ | |
1829 | {bgp_ignore, Idle}, /* Receive_KEEPALIVE_message */ | |
1830 | {bgp_ignore, Idle}, /* Receive_UPDATE_message */ | |
1831 | {bgp_ignore, Idle}, /* Receive_NOTIFICATION_message */ | |
1832 | {bgp_ignore, Idle}, /* Clearing_Completed */ | |
1833 | }, | |
1834 | { | |
1835 | /* Connect */ | |
1836 | {bgp_ignore, Connect}, /* BGP_Start */ | |
1837 | {bgp_stop, Idle}, /* BGP_Stop */ | |
1838 | {bgp_connect_success, OpenSent}, /* TCP_connection_open */ | |
1839 | {bgp_stop, Idle}, /* TCP_connection_closed */ | |
1840 | {bgp_connect_fail, Active}, /* TCP_connection_open_failed */ | |
1841 | {bgp_connect_fail, Idle}, /* TCP_fatal_error */ | |
1842 | {bgp_reconnect, Connect}, /* ConnectRetry_timer_expired */ | |
1843 | {bgp_fsm_exeption, Idle}, /* Hold_Timer_expired */ | |
1844 | {bgp_fsm_exeption, Idle}, /* KeepAlive_timer_expired */ | |
1845 | {bgp_fsm_exeption, Idle}, /* Receive_OPEN_message */ | |
1846 | {bgp_fsm_exeption, Idle}, /* Receive_KEEPALIVE_message */ | |
1847 | {bgp_fsm_exeption, Idle}, /* Receive_UPDATE_message */ | |
1848 | {bgp_stop, Idle}, /* Receive_NOTIFICATION_message */ | |
1849 | {bgp_fsm_exeption, Idle}, /* Clearing_Completed */ | |
1850 | }, | |
1851 | { | |
1852 | /* Active, */ | |
1853 | {bgp_ignore, Active}, /* BGP_Start */ | |
1854 | {bgp_stop, Idle}, /* BGP_Stop */ | |
1855 | {bgp_connect_success, OpenSent}, /* TCP_connection_open */ | |
1856 | {bgp_stop, Idle}, /* TCP_connection_closed */ | |
1857 | {bgp_ignore, Active}, /* TCP_connection_open_failed */ | |
1858 | {bgp_fsm_exeption, Idle}, /* TCP_fatal_error */ | |
1859 | {bgp_start, Connect}, /* ConnectRetry_timer_expired */ | |
1860 | {bgp_fsm_exeption, Idle}, /* Hold_Timer_expired */ | |
1861 | {bgp_fsm_exeption, Idle}, /* KeepAlive_timer_expired */ | |
1862 | {bgp_fsm_exeption, Idle}, /* Receive_OPEN_message */ | |
1863 | {bgp_fsm_exeption, Idle}, /* Receive_KEEPALIVE_message */ | |
1864 | {bgp_fsm_exeption, Idle}, /* Receive_UPDATE_message */ | |
1865 | {bgp_fsm_exeption, Idle}, /* Receive_NOTIFICATION_message */ | |
1866 | {bgp_fsm_exeption, Idle}, /* Clearing_Completed */ | |
1867 | }, | |
1868 | { | |
1869 | /* OpenSent, */ | |
1870 | {bgp_ignore, OpenSent}, /* BGP_Start */ | |
1871 | {bgp_stop, Idle}, /* BGP_Stop */ | |
1872 | {bgp_stop, Active}, /* TCP_connection_open */ | |
1873 | {bgp_stop, Active}, /* TCP_connection_closed */ | |
1874 | {bgp_stop, Active}, /* TCP_connection_open_failed */ | |
1875 | {bgp_stop, Active}, /* TCP_fatal_error */ | |
1876 | {bgp_fsm_exeption, Idle}, /* ConnectRetry_timer_expired */ | |
1877 | {bgp_fsm_holdtime_expire, Idle}, /* Hold_Timer_expired */ | |
1878 | {bgp_fsm_exeption, Idle}, /* KeepAlive_timer_expired */ | |
1879 | {bgp_fsm_open, OpenConfirm}, /* Receive_OPEN_message */ | |
1880 | {bgp_fsm_event_error, Idle}, /* Receive_KEEPALIVE_message */ | |
1881 | {bgp_fsm_event_error, Idle}, /* Receive_UPDATE_message */ | |
1882 | {bgp_stop_with_error, Idle}, /* Receive_NOTIFICATION_message */ | |
1883 | {bgp_fsm_exeption, Idle}, /* Clearing_Completed */ | |
1884 | }, | |
1885 | { | |
1886 | /* OpenConfirm, */ | |
1887 | {bgp_ignore, OpenConfirm}, /* BGP_Start */ | |
1888 | {bgp_stop, Idle}, /* BGP_Stop */ | |
1889 | {bgp_stop, Idle}, /* TCP_connection_open */ | |
1890 | {bgp_stop, Idle}, /* TCP_connection_closed */ | |
1891 | {bgp_stop, Idle}, /* TCP_connection_open_failed */ | |
1892 | {bgp_stop, Idle}, /* TCP_fatal_error */ | |
1893 | {bgp_fsm_exeption, Idle}, /* ConnectRetry_timer_expired */ | |
1894 | {bgp_fsm_holdtime_expire, Idle}, /* Hold_Timer_expired */ | |
1895 | {bgp_ignore, OpenConfirm}, /* KeepAlive_timer_expired */ | |
1896 | {bgp_fsm_exeption, Idle}, /* Receive_OPEN_message */ | |
1897 | {bgp_establish, Established}, /* Receive_KEEPALIVE_message */ | |
1898 | {bgp_fsm_exeption, Idle}, /* Receive_UPDATE_message */ | |
1899 | {bgp_stop_with_error, Idle}, /* Receive_NOTIFICATION_message */ | |
1900 | {bgp_fsm_exeption, Idle}, /* Clearing_Completed */ | |
1901 | }, | |
1902 | { | |
1903 | /* Established, */ | |
1904 | {bgp_ignore, Established}, /* BGP_Start */ | |
1905 | {bgp_stop, Clearing}, /* BGP_Stop */ | |
1906 | {bgp_stop, Clearing}, /* TCP_connection_open */ | |
1907 | {bgp_stop, Clearing}, /* TCP_connection_closed */ | |
1908 | {bgp_stop, Clearing}, /* TCP_connection_open_failed */ | |
1909 | {bgp_stop, Clearing}, /* TCP_fatal_error */ | |
1910 | {bgp_stop, Clearing}, /* ConnectRetry_timer_expired */ | |
1911 | {bgp_fsm_holdtime_expire, Clearing}, /* Hold_Timer_expired */ | |
1912 | {bgp_ignore, Established}, /* KeepAlive_timer_expired */ | |
1913 | {bgp_stop, Clearing}, /* Receive_OPEN_message */ | |
1914 | {bgp_fsm_keepalive, | |
1915 | Established}, /* Receive_KEEPALIVE_message */ | |
1916 | {bgp_fsm_update, Established}, /* Receive_UPDATE_message */ | |
1917 | {bgp_stop_with_error, | |
1918 | Clearing}, /* Receive_NOTIFICATION_message */ | |
1919 | {bgp_fsm_exeption, Idle}, /* Clearing_Completed */ | |
1920 | }, | |
1921 | { | |
1922 | /* Clearing, */ | |
1923 | {bgp_ignore, Clearing}, /* BGP_Start */ | |
1924 | {bgp_stop, Clearing}, /* BGP_Stop */ | |
1925 | {bgp_stop, Clearing}, /* TCP_connection_open */ | |
1926 | {bgp_stop, Clearing}, /* TCP_connection_closed */ | |
1927 | {bgp_stop, Clearing}, /* TCP_connection_open_failed */ | |
1928 | {bgp_stop, Clearing}, /* TCP_fatal_error */ | |
1929 | {bgp_stop, Clearing}, /* ConnectRetry_timer_expired */ | |
1930 | {bgp_stop, Clearing}, /* Hold_Timer_expired */ | |
1931 | {bgp_stop, Clearing}, /* KeepAlive_timer_expired */ | |
1932 | {bgp_stop, Clearing}, /* Receive_OPEN_message */ | |
1933 | {bgp_stop, Clearing}, /* Receive_KEEPALIVE_message */ | |
1934 | {bgp_stop, Clearing}, /* Receive_UPDATE_message */ | |
1935 | {bgp_stop, Clearing}, /* Receive_NOTIFICATION_message */ | |
1936 | {bgp_clearing_completed, Idle}, /* Clearing_Completed */ | |
1937 | }, | |
1938 | { | |
1939 | /* Deleted, */ | |
1940 | {bgp_ignore, Deleted}, /* BGP_Start */ | |
1941 | {bgp_ignore, Deleted}, /* BGP_Stop */ | |
1942 | {bgp_ignore, Deleted}, /* TCP_connection_open */ | |
1943 | {bgp_ignore, Deleted}, /* TCP_connection_closed */ | |
1944 | {bgp_ignore, Deleted}, /* TCP_connection_open_failed */ | |
1945 | {bgp_ignore, Deleted}, /* TCP_fatal_error */ | |
1946 | {bgp_ignore, Deleted}, /* ConnectRetry_timer_expired */ | |
1947 | {bgp_ignore, Deleted}, /* Hold_Timer_expired */ | |
1948 | {bgp_ignore, Deleted}, /* KeepAlive_timer_expired */ | |
1949 | {bgp_ignore, Deleted}, /* Receive_OPEN_message */ | |
1950 | {bgp_ignore, Deleted}, /* Receive_KEEPALIVE_message */ | |
1951 | {bgp_ignore, Deleted}, /* Receive_UPDATE_message */ | |
1952 | {bgp_ignore, Deleted}, /* Receive_NOTIFICATION_message */ | |
1953 | {bgp_ignore, Deleted}, /* Clearing_Completed */ | |
1954 | }, | |
1955 | }; | |
1956 | ||
1957 | /* Execute event process. */ | |
1958 | int bgp_event(struct thread *thread) | |
1959 | { | |
1960 | int event; | |
1961 | struct peer *peer; | |
1962 | int ret; | |
1963 | ||
1964 | peer = THREAD_ARG(thread); | |
1965 | event = THREAD_VAL(thread); | |
1966 | ||
1967 | ret = bgp_event_update(peer, event); | |
1968 | ||
1969 | return (ret); | |
1970 | } | |
1971 | ||
1972 | int bgp_event_update(struct peer *peer, int event) | |
1973 | { | |
1974 | int next; | |
1975 | int ret = 0; | |
1976 | struct peer *other; | |
1977 | int passive_conn = 0; | |
1978 | int dyn_nbr; | |
1979 | ||
1980 | /* default return code */ | |
1981 | ret = FSM_PEER_NOOP; | |
1982 | ||
1983 | other = peer->doppelganger; | |
1984 | passive_conn = | |
1985 | (CHECK_FLAG(peer->sflags, PEER_STATUS_ACCEPT_PEER)) ? 1 : 0; | |
1986 | dyn_nbr = peer_dynamic_neighbor(peer); | |
1987 | ||
1988 | /* Logging this event. */ | |
1989 | next = FSM[peer->status - 1][event - 1].next_state; | |
1990 | ||
1991 | if (bgp_debug_neighbor_events(peer) && peer->status != next) | |
1992 | zlog_debug("%s [FSM] %s (%s->%s), fd %d", peer->host, | |
1993 | bgp_event_str[event], | |
1994 | lookup_msg(bgp_status_msg, peer->status, NULL), | |
1995 | lookup_msg(bgp_status_msg, next, NULL), peer->fd); | |
1996 | ||
1997 | peer->last_event = peer->cur_event; | |
1998 | peer->cur_event = event; | |
1999 | ||
2000 | /* Call function. */ | |
2001 | if (FSM[peer->status - 1][event - 1].func) | |
2002 | ret = (*(FSM[peer->status - 1][event - 1].func))(peer); | |
2003 | ||
2004 | if (ret >= 0) { | |
2005 | if (ret == 1 && next == Established) { | |
2006 | /* The case when doppelganger swap accurred in | |
2007 | bgp_establish. | |
2008 | Update the peer pointer accordingly */ | |
2009 | ret = FSM_PEER_TRANSFERRED; | |
2010 | peer = other; | |
2011 | } | |
2012 | ||
2013 | /* If status is changed. */ | |
2014 | if (next != peer->status) { | |
2015 | bgp_fsm_change_status(peer, next); | |
2016 | ||
2017 | /* | |
2018 | * If we're going to ESTABLISHED then we executed a | |
2019 | * peer transfer. In this case we can either return | |
2020 | * FSM_PEER_TRANSITIONED or FSM_PEER_TRANSFERRED. | |
2021 | * Opting for TRANSFERRED since transfer implies | |
2022 | * session establishment. | |
2023 | */ | |
2024 | if (ret != FSM_PEER_TRANSFERRED) | |
2025 | ret = FSM_PEER_TRANSITIONED; | |
2026 | } | |
2027 | ||
2028 | /* Make sure timer is set. */ | |
2029 | bgp_timer_set(peer); | |
2030 | ||
2031 | } else { | |
2032 | /* | |
2033 | * If we got a return value of -1, that means there was an | |
2034 | * error, restart the FSM. Since bgp_stop() was called on the | |
2035 | * peer. only a few fields are safe to access here. In any case | |
2036 | * we need to indicate that the peer was stopped in the return | |
2037 | * code. | |
2038 | */ | |
2039 | if (!dyn_nbr && !passive_conn && peer->bgp) { | |
2040 | flog_err( | |
2041 | EC_BGP_FSM, | |
2042 | "%s [FSM] Failure handling event %s in state %s, " | |
2043 | "prior events %s, %s, fd %d", | |
2044 | peer->host, bgp_event_str[peer->cur_event], | |
2045 | lookup_msg(bgp_status_msg, peer->status, NULL), | |
2046 | bgp_event_str[peer->last_event], | |
2047 | bgp_event_str[peer->last_major_event], | |
2048 | peer->fd); | |
2049 | bgp_stop(peer); | |
2050 | bgp_fsm_change_status(peer, Idle); | |
2051 | bgp_timer_set(peer); | |
2052 | } | |
2053 | ret = FSM_PEER_STOPPED; | |
2054 | } | |
2055 | ||
2056 | return ret; | |
2057 | } |