]>
Commit | Line | Data |
---|---|---|
718e3744 | 1 | /* BGP packet management routine. |
896014f4 DL |
2 | * Copyright (C) 1999 Kunihiro Ishiguro |
3 | * | |
4 | * This file is part of GNU Zebra. | |
5 | * | |
6 | * GNU Zebra is free software; you can redistribute it and/or modify it | |
7 | * under the terms of the GNU General Public License as published by the | |
8 | * Free Software Foundation; either version 2, or (at your option) any | |
9 | * later version. | |
10 | * | |
11 | * GNU Zebra is distributed in the hope that it will be useful, but | |
12 | * WITHOUT ANY WARRANTY; without even the implied warranty of | |
13 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU | |
14 | * General Public License for more details. | |
15 | * | |
16 | * You should have received a copy of the GNU General Public License along | |
17 | * with this program; see the file COPYING; if not, write to the Free Software | |
18 | * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA | |
19 | */ | |
718e3744 | 20 | |
21 | #include <zebra.h> | |
d3ecc69e | 22 | #include <sys/time.h> |
718e3744 | 23 | |
24 | #include "thread.h" | |
25 | #include "stream.h" | |
26 | #include "network.h" | |
27 | #include "prefix.h" | |
28 | #include "command.h" | |
29 | #include "log.h" | |
30 | #include "memory.h" | |
d62a17ae | 31 | #include "sockunion.h" /* for inet_ntop () */ |
baa376fc | 32 | #include "sockopt.h" |
718e3744 | 33 | #include "linklist.h" |
34 | #include "plist.h" | |
3f9c7369 | 35 | #include "queue.h" |
039f3a34 | 36 | #include "filter.h" |
718e3744 | 37 | |
38 | #include "bgpd/bgpd.h" | |
39 | #include "bgpd/bgp_table.h" | |
40 | #include "bgpd/bgp_dump.h" | |
41 | #include "bgpd/bgp_attr.h" | |
42 | #include "bgpd/bgp_debug.h" | |
43 | #include "bgpd/bgp_fsm.h" | |
44 | #include "bgpd/bgp_route.h" | |
45 | #include "bgpd/bgp_packet.h" | |
46 | #include "bgpd/bgp_open.h" | |
47 | #include "bgpd/bgp_aspath.h" | |
48 | #include "bgpd/bgp_community.h" | |
49 | #include "bgpd/bgp_ecommunity.h" | |
57d187bc | 50 | #include "bgpd/bgp_lcommunity.h" |
718e3744 | 51 | #include "bgpd/bgp_network.h" |
52 | #include "bgpd/bgp_mplsvpn.h" | |
7ef5a232 | 53 | #include "bgpd/bgp_evpn.h" |
718e3744 | 54 | #include "bgpd/bgp_advertise.h" |
93406d87 | 55 | #include "bgpd/bgp_vty.h" |
3f9c7369 | 56 | #include "bgpd/bgp_updgrp.h" |
cd1964ff | 57 | #include "bgpd/bgp_label.h" |
718e3744 | 58 | |
d3ecc69e QY |
59 | /* Linked list of active peers */ |
60 | static pthread_mutex_t plist_mtx = PTHREAD_MUTEX_INITIALIZER; | |
61 | static pthread_cond_t write_cond = PTHREAD_COND_INITIALIZER; | |
62 | static struct list *plist; | |
63 | ||
64 | bool bgp_packet_writes_thread_run; | |
65 | ||
718e3744 | 66 | /* Set up BGP packet marker and packet type. */ |
d62a17ae | 67 | int bgp_packet_set_marker(struct stream *s, u_char type) |
718e3744 | 68 | { |
d62a17ae | 69 | int i; |
718e3744 | 70 | |
d62a17ae | 71 | /* Fill in marker. */ |
72 | for (i = 0; i < BGP_MARKER_SIZE; i++) | |
73 | stream_putc(s, 0xff); | |
718e3744 | 74 | |
d62a17ae | 75 | /* Dummy total length. This field is should be filled in later on. */ |
76 | stream_putw(s, 0); | |
718e3744 | 77 | |
d62a17ae | 78 | /* BGP packet type. */ |
79 | stream_putc(s, type); | |
718e3744 | 80 | |
d62a17ae | 81 | /* Return current stream size. */ |
82 | return stream_get_endp(s); | |
718e3744 | 83 | } |
84 | ||
85 | /* Set BGP packet header size entry. If size is zero then use current | |
86 | stream size. */ | |
d62a17ae | 87 | int bgp_packet_set_size(struct stream *s) |
718e3744 | 88 | { |
d62a17ae | 89 | int cp; |
718e3744 | 90 | |
d62a17ae | 91 | /* Preserve current pointer. */ |
92 | cp = stream_get_endp(s); | |
93 | stream_putw_at(s, BGP_MARKER_SIZE, cp); | |
718e3744 | 94 | |
d62a17ae | 95 | return cp; |
718e3744 | 96 | } |
97 | ||
d3ecc69e QY |
98 | /** |
99 | * Push a packet onto the beginning of the peer's output queue. | |
100 | * Must be externally synchronized around 'peer'. | |
101 | */ | |
102 | static void bgp_packet_add_unsafe(struct peer *peer, struct stream *s) | |
718e3744 | 103 | { |
d62a17ae | 104 | /* Add packet to the end of list. */ |
105 | stream_fifo_push(peer->obuf, s); | |
d3ecc69e QY |
106 | peer_writes_wake(); |
107 | } | |
108 | ||
109 | /* | |
110 | * Push a packet onto the beginning of the peer's output queue. | |
111 | * This function acquires the peer's write mutex before proceeding. | |
112 | */ | |
113 | static void bgp_packet_add(struct peer *peer, struct stream *s) | |
114 | { | |
115 | pthread_mutex_lock(&peer->obuf_mtx); | |
116 | bgp_packet_add_unsafe(peer, s); | |
117 | pthread_mutex_unlock(&peer->obuf_mtx); | |
718e3744 | 118 | } |
119 | ||
d3ecc69e QY |
120 | /** |
121 | * Pop a packet off the end of the peer's output queue. | |
122 | * Must be externally synchronized around 'peer'. | |
123 | */ | |
124 | static void bgp_packet_delete_unsafe(struct peer *peer) | |
718e3744 | 125 | { |
d62a17ae | 126 | stream_free(stream_fifo_pop(peer->obuf)); |
718e3744 | 127 | } |
128 | ||
d3ecc69e | 129 | |
718e3744 | 130 | /* Check file descriptor whether connect is established. */ |
d3ecc69e | 131 | static int bgp_connect_check(struct peer *peer, int change_state) |
718e3744 | 132 | { |
d62a17ae | 133 | int status; |
134 | socklen_t slen; | |
135 | int ret; | |
136 | ||
137 | /* Anyway I have to reset read and write thread. */ | |
138 | BGP_READ_OFF(peer->t_read); | |
d62a17ae | 139 | |
140 | /* Check file descriptor. */ | |
141 | slen = sizeof(status); | |
142 | ret = getsockopt(peer->fd, SOL_SOCKET, SO_ERROR, (void *)&status, | |
143 | &slen); | |
144 | ||
145 | /* If getsockopt is fail, this is fatal error. */ | |
146 | if (ret < 0) { | |
147 | zlog_info("can't get sockopt for nonblocking connect"); | |
148 | BGP_EVENT_ADD(peer, TCP_fatal_error); | |
149 | return -1; | |
150 | } | |
151 | ||
152 | /* When status is 0 then TCP connection is established. */ | |
153 | if (status == 0) { | |
154 | BGP_EVENT_ADD(peer, TCP_connection_open); | |
155 | return 1; | |
156 | } else { | |
157 | if (bgp_debug_neighbor_events(peer)) | |
158 | zlog_debug("%s [Event] Connect failed (%s)", peer->host, | |
159 | safe_strerror(errno)); | |
160 | if (change_state) | |
161 | BGP_EVENT_ADD(peer, TCP_connection_open_failed); | |
162 | return 0; | |
163 | } | |
718e3744 | 164 | } |
165 | ||
d62a17ae | 166 | static struct stream *bgp_update_packet_eor(struct peer *peer, afi_t afi, |
167 | safi_t safi) | |
93406d87 | 168 | { |
d62a17ae | 169 | struct stream *s; |
170 | iana_afi_t pkt_afi; | |
5c525538 | 171 | iana_safi_t pkt_safi; |
d62a17ae | 172 | |
173 | if (DISABLE_BGP_ANNOUNCE) | |
174 | return NULL; | |
175 | ||
176 | if (bgp_debug_neighbor_events(peer)) | |
177 | zlog_debug("send End-of-RIB for %s to %s", | |
178 | afi_safi_print(afi, safi), peer->host); | |
179 | ||
180 | s = stream_new(BGP_MAX_PACKET_SIZE); | |
181 | ||
182 | /* Make BGP update packet. */ | |
183 | bgp_packet_set_marker(s, BGP_MSG_UPDATE); | |
184 | ||
185 | /* Unfeasible Routes Length */ | |
186 | stream_putw(s, 0); | |
187 | ||
188 | if (afi == AFI_IP && safi == SAFI_UNICAST) { | |
189 | /* Total Path Attribute Length */ | |
190 | stream_putw(s, 0); | |
191 | } else { | |
192 | /* Convert AFI, SAFI to values for packet. */ | |
193 | bgp_map_afi_safi_int2iana(afi, safi, &pkt_afi, &pkt_safi); | |
194 | ||
195 | /* Total Path Attribute Length */ | |
196 | stream_putw(s, 6); | |
197 | stream_putc(s, BGP_ATTR_FLAG_OPTIONAL); | |
198 | stream_putc(s, BGP_ATTR_MP_UNREACH_NLRI); | |
199 | stream_putc(s, 3); | |
200 | stream_putw(s, pkt_afi); | |
201 | stream_putc(s, pkt_safi); | |
202 | } | |
203 | ||
204 | bgp_packet_set_size(s); | |
d3ecc69e | 205 | bgp_packet_add_unsafe(peer, s); |
d62a17ae | 206 | return s; |
718e3744 | 207 | } |
208 | ||
718e3744 | 209 | /* Get next packet to be written. */ |
d62a17ae | 210 | static struct stream *bgp_write_packet(struct peer *peer) |
718e3744 | 211 | { |
d62a17ae | 212 | struct stream *s = NULL; |
213 | struct peer_af *paf; | |
214 | struct bpacket *next_pkt; | |
215 | afi_t afi; | |
216 | safi_t safi; | |
217 | ||
218 | s = stream_fifo_head(peer->obuf); | |
219 | if (s) | |
220 | return s; | |
221 | ||
222 | /* | |
223 | * The code beyond this part deals with update packets, proceed only | |
224 | * if peer is Established and updates are not on hold (as part of | |
225 | * update-delay post processing). | |
3f9c7369 | 226 | */ |
d62a17ae | 227 | if (peer->status != Established) |
228 | return NULL; | |
229 | ||
230 | if (peer->bgp && peer->bgp->main_peers_update_hold) | |
231 | return NULL; | |
232 | ||
05c7a1cc QY |
233 | FOREACH_AFI_SAFI (afi, safi) { |
234 | paf = peer_af_find(peer, afi, safi); | |
235 | if (!paf || !PAF_SUBGRP(paf)) | |
236 | continue; | |
237 | next_pkt = paf->next_pkt_to_send; | |
238 | ||
239 | /* Try to generate a packet for the peer if we are at | |
240 | * the end of | |
241 | * the list. Always try to push out WITHDRAWs first. */ | |
242 | if (!next_pkt || !next_pkt->buffer) { | |
243 | next_pkt = subgroup_withdraw_packet(PAF_SUBGRP(paf)); | |
244 | if (!next_pkt || !next_pkt->buffer) | |
245 | subgroup_update_packet(PAF_SUBGRP(paf)); | |
d62a17ae | 246 | next_pkt = paf->next_pkt_to_send; |
05c7a1cc | 247 | } |
d62a17ae | 248 | |
05c7a1cc QY |
249 | /* If we still don't have a packet to send to the peer, |
250 | * then | |
251 | * try to find out out if we have to send eor or if not, | |
252 | * skip to | |
253 | * the next AFI, SAFI. | |
254 | * Don't send the EOR prematurely... if the subgroup's | |
255 | * coalesce | |
256 | * timer is running, the adjacency-out structure is not | |
257 | * created | |
258 | * yet. | |
259 | */ | |
260 | if (!next_pkt || !next_pkt->buffer) { | |
261 | if (CHECK_FLAG(peer->cap, PEER_CAP_RESTART_RCV)) { | |
262 | if (!(PAF_SUBGRP(paf))->t_coalesce | |
263 | && peer->afc_nego[afi][safi] | |
264 | && peer->synctime | |
265 | && !CHECK_FLAG(peer->af_sflags[afi][safi], | |
266 | PEER_STATUS_EOR_SEND)) { | |
267 | SET_FLAG(peer->af_sflags[afi][safi], | |
268 | PEER_STATUS_EOR_SEND); | |
269 | return bgp_update_packet_eor(peer, afi, | |
270 | safi); | |
d62a17ae | 271 | } |
d62a17ae | 272 | } |
05c7a1cc QY |
273 | continue; |
274 | } | |
d62a17ae | 275 | |
276 | ||
d3ecc69e QY |
277 | /* Found a packet template to send, overwrite packet |
278 | * with appropriate | |
279 | * attributes from peer and advance peer */ | |
280 | s = bpacket_reformat_for_peer(next_pkt, paf); | |
281 | bgp_packet_add_unsafe(peer, s); | |
282 | bpacket_queue_advance_peer(paf); | |
283 | return s; | |
d62a17ae | 284 | } |
cb1faec9 | 285 | |
d3ecc69e | 286 | return NULL; |
718e3744 | 287 | } |
288 | ||
d3ecc69e QY |
289 | /* |
290 | * Creates a BGP Keepalive packet and appends it to the peer's output queue. | |
291 | */ | |
d62a17ae | 292 | void bgp_keepalive_send(struct peer *peer) |
718e3744 | 293 | { |
d62a17ae | 294 | struct stream *s; |
295 | ||
296 | s = stream_new(BGP_MAX_PACKET_SIZE); | |
718e3744 | 297 | |
d62a17ae | 298 | /* Make keepalive packet. */ |
299 | bgp_packet_set_marker(s, BGP_MSG_KEEPALIVE); | |
718e3744 | 300 | |
d62a17ae | 301 | /* Set packet size. */ |
302 | (void)bgp_packet_set_size(s); | |
718e3744 | 303 | |
d62a17ae | 304 | /* Dump packet if debug option is set. */ |
305 | /* bgp_packet_dump (s); */ | |
718e3744 | 306 | |
d62a17ae | 307 | if (bgp_debug_keepalive(peer)) |
308 | zlog_debug("%s sending KEEPALIVE", peer->host); | |
718e3744 | 309 | |
d62a17ae | 310 | /* Add packet to the peer. */ |
311 | bgp_packet_add(peer, s); | |
718e3744 | 312 | } |
313 | ||
d3ecc69e QY |
314 | /* |
315 | * Creates a BGP Open packet and appends it to the peer's output queue. | |
316 | * Sets capabilities as necessary. | |
317 | */ | |
d62a17ae | 318 | void bgp_open_send(struct peer *peer) |
718e3744 | 319 | { |
d62a17ae | 320 | struct stream *s; |
321 | u_int16_t send_holdtime; | |
322 | as_t local_as; | |
718e3744 | 323 | |
d25e4efc | 324 | if (PEER_OR_GROUP_TIMER_SET(peer)) |
d62a17ae | 325 | send_holdtime = peer->holdtime; |
326 | else | |
327 | send_holdtime = peer->bgp->default_holdtime; | |
718e3744 | 328 | |
d62a17ae | 329 | /* local-as Change */ |
330 | if (peer->change_local_as) | |
331 | local_as = peer->change_local_as; | |
332 | else | |
333 | local_as = peer->local_as; | |
718e3744 | 334 | |
d62a17ae | 335 | s = stream_new(BGP_MAX_PACKET_SIZE); |
718e3744 | 336 | |
d62a17ae | 337 | /* Make open packet. */ |
338 | bgp_packet_set_marker(s, BGP_MSG_OPEN); | |
718e3744 | 339 | |
d62a17ae | 340 | /* Set open packet values. */ |
341 | stream_putc(s, BGP_VERSION_4); /* BGP version */ | |
9d303b37 DL |
342 | stream_putw(s, |
343 | (local_as <= BGP_AS_MAX) ? (u_int16_t)local_as | |
344 | : BGP_AS_TRANS); | |
d62a17ae | 345 | stream_putw(s, send_holdtime); /* Hold Time */ |
346 | stream_put_in_addr(s, &peer->local_id); /* BGP Identifier */ | |
718e3744 | 347 | |
d62a17ae | 348 | /* Set capability code. */ |
349 | bgp_open_capability(s, peer); | |
718e3744 | 350 | |
d62a17ae | 351 | /* Set BGP packet length. */ |
352 | (void)bgp_packet_set_size(s); | |
718e3744 | 353 | |
d62a17ae | 354 | if (bgp_debug_neighbor_events(peer)) |
355 | zlog_debug( | |
356 | "%s sending OPEN, version %d, my as %u, holdtime %d, id %s", | |
357 | peer->host, BGP_VERSION_4, local_as, send_holdtime, | |
358 | inet_ntoa(peer->local_id)); | |
718e3744 | 359 | |
d62a17ae | 360 | /* Dump packet if debug option is set. */ |
361 | /* bgp_packet_dump (s); */ | |
718e3744 | 362 | |
d62a17ae | 363 | /* Add packet to the peer. */ |
364 | bgp_packet_add(peer, s); | |
718e3744 | 365 | } |
366 | ||
d3ecc69e QY |
367 | /* |
368 | * Creates a BGP Notify and appends it to the peer's output queue. | |
369 | * | |
370 | * This function awakens the write thread to ensure the packet | |
371 | * gets out ASAP. | |
372 | * | |
373 | * @param peer | |
374 | * @param code BGP error code | |
375 | * @param sub_code BGP error subcode | |
376 | * @param data Data portion | |
377 | * @param datalen length of data portion | |
378 | */ | |
d62a17ae | 379 | void bgp_notify_send_with_data(struct peer *peer, u_char code, u_char sub_code, |
380 | u_char *data, size_t datalen) | |
718e3744 | 381 | { |
d62a17ae | 382 | struct stream *s; |
383 | int length; | |
384 | ||
385 | /* Allocate new stream. */ | |
386 | s = stream_new(BGP_MAX_PACKET_SIZE); | |
387 | ||
d3ecc69e | 388 | /* Make notify packet. */ |
d62a17ae | 389 | bgp_packet_set_marker(s, BGP_MSG_NOTIFY); |
390 | ||
391 | /* Set notify packet values. */ | |
392 | stream_putc(s, code); /* BGP notify code */ | |
393 | stream_putc(s, sub_code); /* BGP notify sub_code */ | |
394 | ||
395 | /* If notify data is present. */ | |
396 | if (data) | |
397 | stream_write(s, data, datalen); | |
398 | ||
399 | /* Set BGP packet length. */ | |
400 | length = bgp_packet_set_size(s); | |
401 | ||
402 | /* Add packet to the peer. */ | |
d3ecc69e | 403 | pthread_mutex_lock(&peer->obuf_mtx); |
d62a17ae | 404 | stream_fifo_clean(peer->obuf); |
d3ecc69e | 405 | pthread_mutex_unlock(&peer->obuf_mtx); |
d62a17ae | 406 | |
407 | /* For debug */ | |
408 | { | |
409 | struct bgp_notify bgp_notify; | |
410 | int first = 0; | |
411 | int i; | |
412 | char c[4]; | |
413 | ||
414 | bgp_notify.code = code; | |
415 | bgp_notify.subcode = sub_code; | |
416 | bgp_notify.data = NULL; | |
417 | bgp_notify.length = length - BGP_MSG_NOTIFY_MIN_SIZE; | |
418 | bgp_notify.raw_data = data; | |
419 | ||
420 | peer->notify.code = bgp_notify.code; | |
421 | peer->notify.subcode = bgp_notify.subcode; | |
422 | ||
423 | if (bgp_notify.length) { | |
424 | bgp_notify.data = | |
425 | XMALLOC(MTYPE_TMP, bgp_notify.length * 3); | |
426 | for (i = 0; i < bgp_notify.length; i++) | |
427 | if (first) { | |
428 | sprintf(c, " %02x", data[i]); | |
429 | strcat(bgp_notify.data, c); | |
430 | } else { | |
431 | first = 1; | |
432 | sprintf(c, "%02x", data[i]); | |
433 | strcpy(bgp_notify.data, c); | |
434 | } | |
435 | } | |
436 | bgp_notify_print(peer, &bgp_notify, "sending"); | |
437 | ||
438 | if (bgp_notify.data) { | |
439 | XFREE(MTYPE_TMP, bgp_notify.data); | |
440 | bgp_notify.data = NULL; | |
441 | bgp_notify.length = 0; | |
442 | } | |
443 | } | |
444 | ||
445 | /* peer reset cause */ | |
446 | if (code == BGP_NOTIFY_CEASE) { | |
447 | if (sub_code == BGP_NOTIFY_CEASE_ADMIN_RESET) | |
448 | peer->last_reset = PEER_DOWN_USER_RESET; | |
449 | else if (sub_code == BGP_NOTIFY_CEASE_ADMIN_SHUTDOWN) | |
450 | peer->last_reset = PEER_DOWN_USER_SHUTDOWN; | |
451 | else | |
452 | peer->last_reset = PEER_DOWN_NOTIFY_SEND; | |
453 | } else | |
454 | peer->last_reset = PEER_DOWN_NOTIFY_SEND; | |
455 | ||
d3ecc69e QY |
456 | /* Add packet to peer's output queue */ |
457 | bgp_packet_add(peer, s); | |
458 | /* Wake up the write thread to get the notify out ASAP */ | |
459 | peer_writes_wake(); | |
718e3744 | 460 | } |
461 | ||
d3ecc69e QY |
462 | /* |
463 | * Creates a BGP Notify and appends it to the peer's output queue. | |
464 | * | |
465 | * This function awakens the write thread to ensure the packet | |
466 | * gets out ASAP. | |
467 | * | |
468 | * @param peer | |
469 | * @param code BGP error code | |
470 | * @param sub_code BGP error subcode | |
471 | */ | |
d62a17ae | 472 | void bgp_notify_send(struct peer *peer, u_char code, u_char sub_code) |
718e3744 | 473 | { |
d62a17ae | 474 | bgp_notify_send_with_data(peer, code, sub_code, NULL, 0); |
718e3744 | 475 | } |
476 | ||
d3ecc69e QY |
477 | /* |
478 | * Creates BGP Route Refresh packet and appends it to the peer's output queue. | |
479 | * | |
480 | * @param peer | |
481 | * @param afi Address Family Identifier | |
482 | * @param safi Subsequent Address Family Identifier | |
483 | * @param orf_type Outbound Route Filtering type | |
484 | * @param when_to_refresh Whether to refresh immediately or defer | |
485 | * @param remove Whether to remove ORF for specified AFI/SAFI | |
486 | */ | |
d62a17ae | 487 | void bgp_route_refresh_send(struct peer *peer, afi_t afi, safi_t safi, |
488 | u_char orf_type, u_char when_to_refresh, int remove) | |
718e3744 | 489 | { |
d62a17ae | 490 | struct stream *s; |
491 | struct bgp_filter *filter; | |
492 | int orf_refresh = 0; | |
493 | iana_afi_t pkt_afi; | |
5c525538 | 494 | iana_safi_t pkt_safi; |
d62a17ae | 495 | |
496 | if (DISABLE_BGP_ANNOUNCE) | |
497 | return; | |
498 | ||
499 | filter = &peer->filter[afi][safi]; | |
500 | ||
501 | /* Convert AFI, SAFI to values for packet. */ | |
502 | bgp_map_afi_safi_int2iana(afi, safi, &pkt_afi, &pkt_safi); | |
503 | ||
504 | s = stream_new(BGP_MAX_PACKET_SIZE); | |
505 | ||
506 | /* Make BGP update packet. */ | |
507 | if (CHECK_FLAG(peer->cap, PEER_CAP_REFRESH_NEW_RCV)) | |
508 | bgp_packet_set_marker(s, BGP_MSG_ROUTE_REFRESH_NEW); | |
718e3744 | 509 | else |
d62a17ae | 510 | bgp_packet_set_marker(s, BGP_MSG_ROUTE_REFRESH_OLD); |
511 | ||
512 | /* Encode Route Refresh message. */ | |
513 | stream_putw(s, pkt_afi); | |
514 | stream_putc(s, 0); | |
515 | stream_putc(s, pkt_safi); | |
516 | ||
517 | if (orf_type == ORF_TYPE_PREFIX || orf_type == ORF_TYPE_PREFIX_OLD) | |
518 | if (remove || filter->plist[FILTER_IN].plist) { | |
519 | u_int16_t orf_len; | |
520 | unsigned long orfp; | |
521 | ||
522 | orf_refresh = 1; | |
523 | stream_putc(s, when_to_refresh); | |
524 | stream_putc(s, orf_type); | |
525 | orfp = stream_get_endp(s); | |
526 | stream_putw(s, 0); | |
527 | ||
528 | if (remove) { | |
529 | UNSET_FLAG(peer->af_sflags[afi][safi], | |
530 | PEER_STATUS_ORF_PREFIX_SEND); | |
531 | stream_putc(s, ORF_COMMON_PART_REMOVE_ALL); | |
532 | if (bgp_debug_neighbor_events(peer)) | |
533 | zlog_debug( | |
534 | "%s sending REFRESH_REQ to remove ORF(%d) (%s) for afi/safi: %d/%d", | |
535 | peer->host, orf_type, | |
536 | (when_to_refresh == REFRESH_DEFER | |
537 | ? "defer" | |
538 | : "immediate"), | |
539 | pkt_afi, pkt_safi); | |
540 | } else { | |
541 | SET_FLAG(peer->af_sflags[afi][safi], | |
542 | PEER_STATUS_ORF_PREFIX_SEND); | |
543 | prefix_bgp_orf_entry( | |
544 | s, filter->plist[FILTER_IN].plist, | |
545 | ORF_COMMON_PART_ADD, | |
546 | ORF_COMMON_PART_PERMIT, | |
547 | ORF_COMMON_PART_DENY); | |
548 | if (bgp_debug_neighbor_events(peer)) | |
549 | zlog_debug( | |
550 | "%s sending REFRESH_REQ with pfxlist ORF(%d) (%s) for afi/safi: %d/%d", | |
551 | peer->host, orf_type, | |
552 | (when_to_refresh == REFRESH_DEFER | |
553 | ? "defer" | |
554 | : "immediate"), | |
555 | pkt_afi, pkt_safi); | |
556 | } | |
557 | ||
558 | /* Total ORF Entry Len. */ | |
559 | orf_len = stream_get_endp(s) - orfp - 2; | |
560 | stream_putw_at(s, orfp, orf_len); | |
561 | } | |
562 | ||
563 | /* Set packet size. */ | |
564 | (void)bgp_packet_set_size(s); | |
565 | ||
566 | if (bgp_debug_neighbor_events(peer)) { | |
567 | if (!orf_refresh) | |
568 | zlog_debug("%s sending REFRESH_REQ for afi/safi: %d/%d", | |
569 | peer->host, pkt_afi, pkt_safi); | |
570 | } | |
571 | ||
572 | /* Add packet to the peer. */ | |
573 | bgp_packet_add(peer, s); | |
718e3744 | 574 | } |
575 | ||
d3ecc69e QY |
576 | /* |
577 | * Create a BGP Capability packet and append it to the peer's output queue. | |
578 | * | |
579 | * @param peer | |
580 | * @param afi Address Family Identifier | |
581 | * @param safi Subsequent Address Family Identifier | |
582 | * @param capability_code BGP Capability Code | |
583 | * @param action Set or Remove capability | |
584 | */ | |
d62a17ae | 585 | void bgp_capability_send(struct peer *peer, afi_t afi, safi_t safi, |
586 | int capability_code, int action) | |
718e3744 | 587 | { |
d62a17ae | 588 | struct stream *s; |
589 | iana_afi_t pkt_afi; | |
5c525538 | 590 | iana_safi_t pkt_safi; |
d62a17ae | 591 | |
592 | /* Convert AFI, SAFI to values for packet. */ | |
593 | bgp_map_afi_safi_int2iana(afi, safi, &pkt_afi, &pkt_safi); | |
594 | ||
595 | s = stream_new(BGP_MAX_PACKET_SIZE); | |
596 | ||
597 | /* Make BGP update packet. */ | |
598 | bgp_packet_set_marker(s, BGP_MSG_CAPABILITY); | |
599 | ||
600 | /* Encode MP_EXT capability. */ | |
601 | if (capability_code == CAPABILITY_CODE_MP) { | |
602 | stream_putc(s, action); | |
603 | stream_putc(s, CAPABILITY_CODE_MP); | |
604 | stream_putc(s, CAPABILITY_CODE_MP_LEN); | |
605 | stream_putw(s, pkt_afi); | |
606 | stream_putc(s, 0); | |
607 | stream_putc(s, pkt_safi); | |
608 | ||
609 | if (bgp_debug_neighbor_events(peer)) | |
610 | zlog_debug( | |
611 | "%s sending CAPABILITY has %s MP_EXT CAP for afi/safi: %d/%d", | |
612 | peer->host, | |
613 | action == CAPABILITY_ACTION_SET ? "Advertising" | |
614 | : "Removing", | |
615 | pkt_afi, pkt_safi); | |
616 | } | |
617 | ||
618 | /* Set packet size. */ | |
619 | (void)bgp_packet_set_size(s); | |
718e3744 | 620 | |
d62a17ae | 621 | /* Add packet to the peer. */ |
622 | bgp_packet_add(peer, s); | |
d62a17ae | 623 | } |
718e3744 | 624 | |
d62a17ae | 625 | /* RFC1771 6.8 Connection collision detection. */ |
626 | static int bgp_collision_detect(struct peer *new, struct in_addr remote_id) | |
627 | { | |
628 | struct peer *peer; | |
629 | ||
630 | /* Upon receipt of an OPEN message, the local system must examine | |
631 | all of its connections that are in the OpenConfirm state. A BGP | |
632 | speaker may also examine connections in an OpenSent state if it | |
633 | knows the BGP Identifier of the peer by means outside of the | |
634 | protocol. If among these connections there is a connection to a | |
635 | remote BGP speaker whose BGP Identifier equals the one in the | |
636 | OPEN message, then the local system performs the following | |
637 | collision resolution procedure: */ | |
638 | ||
639 | if ((peer = new->doppelganger) != NULL) { | |
640 | /* Do not accept the new connection in Established or Clearing | |
641 | * states. | |
642 | * Note that a peer GR is handled by closing the existing | |
643 | * connection | |
644 | * upon receipt of new one. | |
645 | */ | |
646 | if (peer->status == Established || peer->status == Clearing) { | |
647 | bgp_notify_send(new, BGP_NOTIFY_CEASE, | |
648 | BGP_NOTIFY_CEASE_COLLISION_RESOLUTION); | |
649 | return (-1); | |
650 | } else if ((peer->status == OpenConfirm) | |
651 | || (peer->status == OpenSent)) { | |
652 | /* 1. The BGP Identifier of the local system is compared | |
653 | to | |
654 | the BGP Identifier of the remote system (as specified | |
655 | in | |
656 | the OPEN message). */ | |
657 | ||
658 | if (ntohl(peer->local_id.s_addr) | |
659 | < ntohl(remote_id.s_addr)) | |
660 | if (!CHECK_FLAG(peer->sflags, | |
661 | PEER_STATUS_ACCEPT_PEER)) { | |
662 | /* 2. If the value of the local BGP | |
663 | Identifier is less | |
664 | than the remote one, the local system | |
665 | closes BGP | |
666 | connection that already exists (the | |
667 | one that is | |
668 | already in the OpenConfirm state), | |
669 | and accepts BGP | |
670 | connection initiated by the remote | |
671 | system. */ | |
672 | bgp_notify_send( | |
673 | peer, BGP_NOTIFY_CEASE, | |
674 | BGP_NOTIFY_CEASE_COLLISION_RESOLUTION); | |
675 | return 1; | |
676 | } else { | |
677 | bgp_notify_send( | |
678 | new, BGP_NOTIFY_CEASE, | |
679 | BGP_NOTIFY_CEASE_COLLISION_RESOLUTION); | |
680 | return -1; | |
681 | } | |
682 | else { | |
683 | /* 3. Otherwise, the local system closes newly | |
684 | created | |
685 | BGP connection (the one associated with the | |
686 | newly | |
687 | received OPEN message), and continues to use | |
688 | the | |
689 | existing one (the one that is already in the | |
690 | OpenConfirm state). */ | |
691 | if (CHECK_FLAG(peer->sflags, | |
692 | PEER_STATUS_ACCEPT_PEER)) { | |
693 | bgp_notify_send( | |
694 | peer, BGP_NOTIFY_CEASE, | |
695 | BGP_NOTIFY_CEASE_COLLISION_RESOLUTION); | |
696 | return 1; | |
697 | } else { | |
698 | bgp_notify_send( | |
699 | new, BGP_NOTIFY_CEASE, | |
700 | BGP_NOTIFY_CEASE_COLLISION_RESOLUTION); | |
701 | return -1; | |
702 | } | |
703 | } | |
704 | } | |
705 | } | |
706 | return 0; | |
707 | } | |
718e3744 | 708 | |
d62a17ae | 709 | static int bgp_open_receive(struct peer *peer, bgp_size_t size) |
710 | { | |
711 | int ret; | |
712 | u_char version; | |
713 | u_char optlen; | |
714 | u_int16_t holdtime; | |
715 | u_int16_t send_holdtime; | |
716 | as_t remote_as; | |
717 | as_t as4 = 0; | |
718 | struct in_addr remote_id; | |
719 | int mp_capability; | |
720 | u_int8_t notify_data_remote_as[2]; | |
721 | u_int8_t notify_data_remote_as4[4]; | |
722 | u_int8_t notify_data_remote_id[4]; | |
723 | u_int16_t *holdtime_ptr; | |
724 | ||
725 | /* Parse open packet. */ | |
726 | version = stream_getc(peer->ibuf); | |
727 | memcpy(notify_data_remote_as, stream_pnt(peer->ibuf), 2); | |
728 | remote_as = stream_getw(peer->ibuf); | |
729 | holdtime_ptr = (u_int16_t *)stream_pnt(peer->ibuf); | |
730 | holdtime = stream_getw(peer->ibuf); | |
731 | memcpy(notify_data_remote_id, stream_pnt(peer->ibuf), 4); | |
732 | remote_id.s_addr = stream_get_ipv4(peer->ibuf); | |
733 | ||
734 | /* Receive OPEN message log */ | |
735 | if (bgp_debug_neighbor_events(peer)) | |
736 | zlog_debug( | |
737 | "%s rcv OPEN, version %d, remote-as (in open) %u," | |
738 | " holdtime %d, id %s", | |
739 | peer->host, version, remote_as, holdtime, | |
740 | inet_ntoa(remote_id)); | |
741 | ||
742 | /* BEGIN to read the capability here, but dont do it yet */ | |
743 | mp_capability = 0; | |
744 | optlen = stream_getc(peer->ibuf); | |
745 | ||
746 | if (optlen != 0) { | |
747 | /* If not enough bytes, it is an error. */ | |
748 | if (STREAM_READABLE(peer->ibuf) < optlen) { | |
749 | bgp_notify_send(peer, BGP_NOTIFY_OPEN_ERR, | |
750 | BGP_NOTIFY_OPEN_MALFORMED_ATTR); | |
751 | return -1; | |
752 | } | |
718e3744 | 753 | |
d62a17ae | 754 | /* We need the as4 capability value *right now* because |
755 | * if it is there, we have not got the remote_as yet, and | |
756 | * without | |
757 | * that we do not know which peer is connecting to us now. | |
758 | */ | |
759 | as4 = peek_for_as4_capability(peer, optlen); | |
760 | memcpy(notify_data_remote_as4, &as4, 4); | |
761 | } | |
718e3744 | 762 | |
d62a17ae | 763 | /* Just in case we have a silly peer who sends AS4 capability set to 0 |
764 | */ | |
765 | if (CHECK_FLAG(peer->cap, PEER_CAP_AS4_RCV) && !as4) { | |
766 | zlog_err("%s bad OPEN, got AS4 capability, but AS4 set to 0", | |
767 | peer->host); | |
768 | bgp_notify_send_with_data(peer, BGP_NOTIFY_OPEN_ERR, | |
769 | BGP_NOTIFY_OPEN_BAD_PEER_AS, | |
770 | notify_data_remote_as4, 4); | |
771 | return -1; | |
772 | } | |
718e3744 | 773 | |
d62a17ae | 774 | if (remote_as == BGP_AS_TRANS) { |
775 | /* Take the AS4 from the capability. We must have received the | |
776 | * capability now! Otherwise we have a asn16 peer who uses | |
777 | * BGP_AS_TRANS, for some unknown reason. | |
778 | */ | |
779 | if (as4 == BGP_AS_TRANS) { | |
780 | zlog_err( | |
781 | "%s [AS4] NEW speaker using AS_TRANS for AS4, not allowed", | |
782 | peer->host); | |
783 | bgp_notify_send_with_data(peer, BGP_NOTIFY_OPEN_ERR, | |
784 | BGP_NOTIFY_OPEN_BAD_PEER_AS, | |
785 | notify_data_remote_as4, 4); | |
786 | return -1; | |
787 | } | |
718e3744 | 788 | |
d62a17ae | 789 | if (!as4 && BGP_DEBUG(as4, AS4)) |
790 | zlog_debug( | |
791 | "%s [AS4] OPEN remote_as is AS_TRANS, but no AS4." | |
792 | " Odd, but proceeding.", | |
793 | peer->host); | |
794 | else if (as4 < BGP_AS_MAX && BGP_DEBUG(as4, AS4)) | |
795 | zlog_debug( | |
796 | "%s [AS4] OPEN remote_as is AS_TRANS, but AS4 (%u) fits " | |
797 | "in 2-bytes, very odd peer.", | |
798 | peer->host, as4); | |
799 | if (as4) | |
800 | remote_as = as4; | |
801 | } else { | |
802 | /* We may have a partner with AS4 who has an asno < BGP_AS_MAX | |
803 | */ | |
804 | /* If we have got the capability, peer->as4cap must match | |
805 | * remote_as */ | |
806 | if (CHECK_FLAG(peer->cap, PEER_CAP_AS4_RCV) | |
807 | && as4 != remote_as) { | |
808 | /* raise error, log this, close session */ | |
809 | zlog_err( | |
810 | "%s bad OPEN, got AS4 capability, but remote_as %u" | |
811 | " mismatch with 16bit 'myasn' %u in open", | |
812 | peer->host, as4, remote_as); | |
813 | bgp_notify_send_with_data(peer, BGP_NOTIFY_OPEN_ERR, | |
814 | BGP_NOTIFY_OPEN_BAD_PEER_AS, | |
815 | notify_data_remote_as4, 4); | |
816 | return -1; | |
817 | } | |
818 | } | |
6b0655a2 | 819 | |
d62a17ae | 820 | /* remote router-id check. */ |
821 | if (remote_id.s_addr == 0 || IPV4_CLASS_DE(ntohl(remote_id.s_addr)) | |
822 | || ntohl(peer->local_id.s_addr) == ntohl(remote_id.s_addr)) { | |
823 | if (bgp_debug_neighbor_events(peer)) | |
824 | zlog_debug("%s bad OPEN, wrong router identifier %s", | |
825 | peer->host, inet_ntoa(remote_id)); | |
826 | bgp_notify_send_with_data(peer, BGP_NOTIFY_OPEN_ERR, | |
827 | BGP_NOTIFY_OPEN_BAD_BGP_IDENT, | |
828 | notify_data_remote_id, 4); | |
829 | return -1; | |
830 | } | |
831 | ||
832 | /* Set remote router-id */ | |
833 | peer->remote_id = remote_id; | |
834 | ||
835 | /* Peer BGP version check. */ | |
836 | if (version != BGP_VERSION_4) { | |
837 | u_int16_t maxver = htons(BGP_VERSION_4); | |
838 | /* XXX this reply may not be correct if version < 4 XXX */ | |
839 | if (bgp_debug_neighbor_events(peer)) | |
840 | zlog_debug( | |
841 | "%s bad protocol version, remote requested %d, local request %d", | |
842 | peer->host, version, BGP_VERSION_4); | |
843 | /* Data must be in network byte order here */ | |
844 | bgp_notify_send_with_data(peer, BGP_NOTIFY_OPEN_ERR, | |
845 | BGP_NOTIFY_OPEN_UNSUP_VERSION, | |
846 | (u_int8_t *)&maxver, 2); | |
847 | return -1; | |
848 | } | |
849 | ||
850 | /* Check neighbor as number. */ | |
851 | if (peer->as_type == AS_UNSPECIFIED) { | |
852 | if (bgp_debug_neighbor_events(peer)) | |
853 | zlog_debug( | |
854 | "%s bad OPEN, remote AS is unspecified currently", | |
855 | peer->host); | |
856 | bgp_notify_send_with_data(peer, BGP_NOTIFY_OPEN_ERR, | |
857 | BGP_NOTIFY_OPEN_BAD_PEER_AS, | |
858 | notify_data_remote_as, 2); | |
1ff9a340 | 859 | return -1; |
d62a17ae | 860 | } else if (peer->as_type == AS_INTERNAL) { |
861 | if (remote_as != peer->bgp->as) { | |
862 | if (bgp_debug_neighbor_events(peer)) | |
863 | zlog_debug( | |
864 | "%s bad OPEN, remote AS is %u, internal specified", | |
865 | peer->host, remote_as); | |
866 | bgp_notify_send_with_data(peer, BGP_NOTIFY_OPEN_ERR, | |
867 | BGP_NOTIFY_OPEN_BAD_PEER_AS, | |
868 | notify_data_remote_as, 2); | |
869 | return -1; | |
1ff9a340 | 870 | } |
d62a17ae | 871 | peer->as = peer->local_as; |
872 | } else if (peer->as_type == AS_EXTERNAL) { | |
873 | if (remote_as == peer->bgp->as) { | |
874 | if (bgp_debug_neighbor_events(peer)) | |
875 | zlog_debug( | |
876 | "%s bad OPEN, remote AS is %u, external specified", | |
877 | peer->host, remote_as); | |
878 | bgp_notify_send_with_data(peer, BGP_NOTIFY_OPEN_ERR, | |
879 | BGP_NOTIFY_OPEN_BAD_PEER_AS, | |
880 | notify_data_remote_as, 2); | |
881 | return -1; | |
1ff9a340 | 882 | } |
d62a17ae | 883 | peer->as = remote_as; |
884 | } else if ((peer->as_type == AS_SPECIFIED) && (remote_as != peer->as)) { | |
885 | if (bgp_debug_neighbor_events(peer)) | |
886 | zlog_debug("%s bad OPEN, remote AS is %u, expected %u", | |
887 | peer->host, remote_as, peer->as); | |
888 | bgp_notify_send_with_data(peer, BGP_NOTIFY_OPEN_ERR, | |
889 | BGP_NOTIFY_OPEN_BAD_PEER_AS, | |
890 | notify_data_remote_as, 2); | |
891 | return -1; | |
eb821189 | 892 | } |
718e3744 | 893 | |
d62a17ae | 894 | /* From the rfc: Upon receipt of an OPEN message, a BGP speaker MUST |
895 | calculate the value of the Hold Timer by using the smaller of its | |
896 | configured Hold Time and the Hold Time received in the OPEN message. | |
897 | The Hold Time MUST be either zero or at least three seconds. An | |
898 | implementation may reject connections on the basis of the Hold Time. | |
0b2aa3a0 | 899 | */ |
d62a17ae | 900 | |
901 | if (holdtime < 3 && holdtime != 0) { | |
902 | bgp_notify_send_with_data(peer, BGP_NOTIFY_OPEN_ERR, | |
903 | BGP_NOTIFY_OPEN_UNACEP_HOLDTIME, | |
904 | (u_char *)holdtime_ptr, 2); | |
905 | return -1; | |
0b2aa3a0 | 906 | } |
d62a17ae | 907 | |
908 | /* From the rfc: A reasonable maximum time between KEEPALIVE messages | |
909 | would be one third of the Hold Time interval. KEEPALIVE messages | |
910 | MUST NOT be sent more frequently than one per second. An | |
911 | implementation MAY adjust the rate at which it sends KEEPALIVE | |
912 | messages as a function of the Hold Time interval. */ | |
913 | ||
d25e4efc | 914 | if (PEER_OR_GROUP_TIMER_SET(peer)) |
d62a17ae | 915 | send_holdtime = peer->holdtime; |
916 | else | |
917 | send_holdtime = peer->bgp->default_holdtime; | |
918 | ||
919 | if (holdtime < send_holdtime) | |
920 | peer->v_holdtime = holdtime; | |
921 | else | |
922 | peer->v_holdtime = send_holdtime; | |
923 | ||
d25e4efc DS |
924 | if ((PEER_OR_GROUP_TIMER_SET(peer)) |
925 | && (peer->keepalive < peer->v_holdtime / 3)) | |
d62a17ae | 926 | peer->v_keepalive = peer->keepalive; |
927 | else | |
928 | peer->v_keepalive = peer->v_holdtime / 3; | |
929 | ||
930 | /* Open option part parse. */ | |
931 | if (optlen != 0) { | |
932 | if ((ret = bgp_open_option_parse(peer, optlen, &mp_capability)) | |
933 | < 0) | |
934 | return ret; | |
935 | } else { | |
936 | if (bgp_debug_neighbor_events(peer)) | |
937 | zlog_debug("%s rcvd OPEN w/ OPTION parameter len: 0", | |
938 | peer->host); | |
0299c004 | 939 | } |
d62a17ae | 940 | |
941 | /* | |
942 | * Assume that the peer supports the locally configured set of | |
943 | * AFI/SAFIs if the peer did not send us any Mulitiprotocol | |
944 | * capabilities, or if 'override-capability' is configured. | |
945 | */ | |
946 | if (!mp_capability | |
947 | || CHECK_FLAG(peer->flags, PEER_FLAG_OVERRIDE_CAPABILITY)) { | |
948 | peer->afc_nego[AFI_IP][SAFI_UNICAST] = | |
949 | peer->afc[AFI_IP][SAFI_UNICAST]; | |
950 | peer->afc_nego[AFI_IP][SAFI_MULTICAST] = | |
951 | peer->afc[AFI_IP][SAFI_MULTICAST]; | |
952 | peer->afc_nego[AFI_IP][SAFI_LABELED_UNICAST] = | |
953 | peer->afc[AFI_IP][SAFI_LABELED_UNICAST]; | |
954 | peer->afc_nego[AFI_IP6][SAFI_UNICAST] = | |
955 | peer->afc[AFI_IP6][SAFI_UNICAST]; | |
956 | peer->afc_nego[AFI_IP6][SAFI_MULTICAST] = | |
957 | peer->afc[AFI_IP6][SAFI_MULTICAST]; | |
958 | peer->afc_nego[AFI_IP6][SAFI_LABELED_UNICAST] = | |
959 | peer->afc[AFI_IP6][SAFI_LABELED_UNICAST]; | |
960 | peer->afc_nego[AFI_L2VPN][SAFI_EVPN] = | |
961 | peer->afc[AFI_L2VPN][SAFI_EVPN]; | |
0299c004 | 962 | } |
d62a17ae | 963 | |
964 | /* When collision is detected and this peer is closed. Retrun | |
965 | immidiately. */ | |
966 | ret = bgp_collision_detect(peer, remote_id); | |
967 | if (ret < 0) | |
968 | return ret; | |
969 | ||
970 | /* Get sockname. */ | |
971 | if ((ret = bgp_getsockname(peer)) < 0) { | |
972 | zlog_err("%s: bgp_getsockname() failed for peer: %s", | |
973 | __FUNCTION__, peer->host); | |
974 | return (ret); | |
975 | } | |
976 | ||
977 | /* Verify valid local address present based on negotiated | |
978 | * address-families. */ | |
979 | if (peer->afc_nego[AFI_IP][SAFI_UNICAST] | |
e5f22b30 | 980 | || peer->afc_nego[AFI_IP][SAFI_LABELED_UNICAST] |
d62a17ae | 981 | || peer->afc_nego[AFI_IP][SAFI_MULTICAST] |
982 | || peer->afc_nego[AFI_IP][SAFI_MPLS_VPN] | |
983 | || peer->afc_nego[AFI_IP][SAFI_ENCAP]) { | |
984 | if (!peer->nexthop.v4.s_addr) { | |
985 | #if defined(HAVE_CUMULUS) | |
986 | zlog_err( | |
987 | "%s: No local IPv4 addr resetting connection, fd %d", | |
988 | peer->host, peer->fd); | |
989 | bgp_notify_send(peer, BGP_NOTIFY_CEASE, | |
990 | BGP_NOTIFY_SUBCODE_UNSPECIFIC); | |
991 | return -1; | |
1d808091 | 992 | #endif |
d62a17ae | 993 | } |
994 | } | |
995 | if (peer->afc_nego[AFI_IP6][SAFI_UNICAST] | |
e5f22b30 | 996 | || peer->afc_nego[AFI_IP6][SAFI_LABELED_UNICAST] |
d62a17ae | 997 | || peer->afc_nego[AFI_IP6][SAFI_MULTICAST] |
998 | || peer->afc_nego[AFI_IP6][SAFI_MPLS_VPN] | |
999 | || peer->afc_nego[AFI_IP6][SAFI_ENCAP]) { | |
1000 | if (IN6_IS_ADDR_UNSPECIFIED(&peer->nexthop.v6_global)) { | |
1001 | #if defined(HAVE_CUMULUS) | |
1002 | zlog_err( | |
1003 | "%s: No local IPv6 addr resetting connection, fd %d", | |
1004 | peer->host, peer->fd); | |
1005 | bgp_notify_send(peer, BGP_NOTIFY_CEASE, | |
1006 | BGP_NOTIFY_SUBCODE_UNSPECIFIC); | |
1007 | return -1; | |
1d808091 | 1008 | #endif |
d62a17ae | 1009 | } |
1010 | } | |
1011 | peer->rtt = sockopt_tcp_rtt(peer->fd); | |
1012 | ||
1013 | if ((ret = bgp_event_update(peer, Receive_OPEN_message)) < 0) { | |
1014 | zlog_err("%s: BGP event update failed for peer: %s", | |
1015 | __FUNCTION__, peer->host); | |
1016 | /* DD: bgp send notify and reset state */ | |
1017 | return (ret); | |
1018 | } | |
1019 | ||
1020 | peer->packet_size = 0; | |
1021 | if (peer->ibuf) | |
1022 | stream_reset(peer->ibuf); | |
1023 | ||
1024 | return 0; | |
718e3744 | 1025 | } |
1026 | ||
d62a17ae | 1027 | /* Called when there is a change in the EOR(implicit or explicit) status of a |
1028 | peer. | |
f188f2c4 | 1029 | Ends the update-delay if all expected peers are done with EORs. */ |
d62a17ae | 1030 | void bgp_check_update_delay(struct bgp *bgp) |
f188f2c4 | 1031 | { |
d62a17ae | 1032 | struct listnode *node, *nnode; |
1033 | struct peer *peer = NULL; | |
1034 | ||
1035 | if (bgp_debug_neighbor_events(peer)) | |
1036 | zlog_debug("Checking update delay, T: %d R: %d I:%d E: %d", | |
1037 | bgp->established, bgp->restarted_peers, | |
1038 | bgp->implicit_eors, bgp->explicit_eors); | |
1039 | ||
1040 | if (bgp->established | |
1041 | <= bgp->restarted_peers + bgp->implicit_eors + bgp->explicit_eors) { | |
1042 | /* This is an extra sanity check to make sure we wait for all | |
1043 | the | |
1044 | eligible configured peers. This check is performed if | |
1045 | establish wait | |
1046 | timer is on, or establish wait option is not given with the | |
1047 | update-delay command */ | |
1048 | if (bgp->t_establish_wait | |
1049 | || (bgp->v_establish_wait == bgp->v_update_delay)) | |
1050 | for (ALL_LIST_ELEMENTS(bgp->peer, node, nnode, peer)) { | |
1051 | if (CHECK_FLAG(peer->flags, | |
1052 | PEER_FLAG_CONFIG_NODE) | |
1053 | && !CHECK_FLAG(peer->flags, | |
1054 | PEER_FLAG_SHUTDOWN) | |
1055 | && !peer->update_delay_over) { | |
1056 | if (bgp_debug_neighbor_events(peer)) | |
1057 | zlog_debug( | |
1058 | " Peer %s pending, continuing read-only mode", | |
1059 | peer->host); | |
1060 | return; | |
1061 | } | |
1062 | } | |
1063 | ||
1064 | zlog_info( | |
1065 | "Update delay ended, restarted: %d, EORs implicit: %d, explicit: %d", | |
1066 | bgp->restarted_peers, bgp->implicit_eors, | |
1067 | bgp->explicit_eors); | |
1068 | bgp_update_delay_end(bgp); | |
1069 | } | |
f188f2c4 DS |
1070 | } |
1071 | ||
1072 | /* Called if peer is known to have restarted. The restart-state bit in | |
1073 | Graceful-Restart capability is used for that */ | |
d62a17ae | 1074 | void bgp_update_restarted_peers(struct peer *peer) |
f188f2c4 | 1075 | { |
d62a17ae | 1076 | if (!bgp_update_delay_active(peer->bgp)) |
1077 | return; /* BGP update delay has ended */ | |
1078 | if (peer->update_delay_over) | |
1079 | return; /* This peer has already been considered */ | |
1080 | ||
1081 | if (bgp_debug_neighbor_events(peer)) | |
1082 | zlog_debug("Peer %s: Checking restarted", peer->host); | |
1083 | ||
1084 | if (peer->status == Established) { | |
1085 | peer->update_delay_over = 1; | |
1086 | peer->bgp->restarted_peers++; | |
1087 | bgp_check_update_delay(peer->bgp); | |
1088 | } | |
f188f2c4 DS |
1089 | } |
1090 | ||
1091 | /* Called as peer receives a keep-alive. Determines if this occurence can be | |
1092 | taken as an implicit EOR for this peer. | |
1093 | NOTE: The very first keep-alive after the Established state of a peer is | |
d62a17ae | 1094 | considered implicit EOR for the update-delay purposes */ |
1095 | void bgp_update_implicit_eors(struct peer *peer) | |
f188f2c4 | 1096 | { |
d62a17ae | 1097 | if (!bgp_update_delay_active(peer->bgp)) |
1098 | return; /* BGP update delay has ended */ | |
1099 | if (peer->update_delay_over) | |
1100 | return; /* This peer has already been considered */ | |
1101 | ||
1102 | if (bgp_debug_neighbor_events(peer)) | |
1103 | zlog_debug("Peer %s: Checking implicit EORs", peer->host); | |
1104 | ||
1105 | if (peer->status == Established) { | |
1106 | peer->update_delay_over = 1; | |
1107 | peer->bgp->implicit_eors++; | |
1108 | bgp_check_update_delay(peer->bgp); | |
1109 | } | |
f188f2c4 DS |
1110 | } |
1111 | ||
1112 | /* Should be called only when there is a change in the EOR_RECEIVED status | |
1113 | for any afi/safi on a peer */ | |
d62a17ae | 1114 | static void bgp_update_explicit_eors(struct peer *peer) |
f188f2c4 | 1115 | { |
d62a17ae | 1116 | afi_t afi; |
1117 | safi_t safi; | |
1118 | ||
1119 | if (!bgp_update_delay_active(peer->bgp)) | |
1120 | return; /* BGP update delay has ended */ | |
1121 | if (peer->update_delay_over) | |
1122 | return; /* This peer has already been considered */ | |
1123 | ||
1124 | if (bgp_debug_neighbor_events(peer)) | |
1125 | zlog_debug("Peer %s: Checking explicit EORs", peer->host); | |
1126 | ||
05c7a1cc QY |
1127 | FOREACH_AFI_SAFI (afi, safi) { |
1128 | if (peer->afc_nego[afi][safi] | |
1129 | && !CHECK_FLAG(peer->af_sflags[afi][safi], | |
1130 | PEER_STATUS_EOR_RECEIVED)) { | |
1131 | if (bgp_debug_neighbor_events(peer)) | |
1132 | zlog_debug( | |
1133 | " afi %d safi %d didnt receive EOR", | |
1134 | afi, safi); | |
1135 | return; | |
d62a17ae | 1136 | } |
05c7a1cc | 1137 | } |
d62a17ae | 1138 | |
1139 | peer->update_delay_over = 1; | |
1140 | peer->bgp->explicit_eors++; | |
1141 | bgp_check_update_delay(peer->bgp); | |
f188f2c4 DS |
1142 | } |
1143 | ||
d62a17ae | 1144 | /* Frontend for NLRI parsing, to fan-out to AFI/SAFI specific parsers |
1145 | * mp_withdraw, if set, is used to nullify attr structure on most of the calling | |
1146 | * safi function | |
7ef5a232 PG |
1147 | * and for evpn, passed as parameter |
1148 | */ | |
d62a17ae | 1149 | int bgp_nlri_parse(struct peer *peer, struct attr *attr, |
1150 | struct bgp_nlri *packet, int mp_withdraw) | |
96e52474 | 1151 | { |
d62a17ae | 1152 | switch (packet->safi) { |
1153 | case SAFI_UNICAST: | |
1154 | case SAFI_MULTICAST: | |
1155 | return bgp_nlri_parse_ip(peer, mp_withdraw ? NULL : attr, | |
1156 | packet); | |
1157 | case SAFI_LABELED_UNICAST: | |
1158 | return bgp_nlri_parse_label(peer, mp_withdraw ? NULL : attr, | |
1159 | packet); | |
1160 | case SAFI_MPLS_VPN: | |
1161 | return bgp_nlri_parse_vpn(peer, mp_withdraw ? NULL : attr, | |
1162 | packet); | |
1163 | case SAFI_EVPN: | |
1164 | return bgp_nlri_parse_evpn(peer, attr, packet, mp_withdraw); | |
5c525538 RW |
1165 | default: |
1166 | return -1; | |
d62a17ae | 1167 | } |
96e52474 PJ |
1168 | } |
1169 | ||
718e3744 | 1170 | /* Parse BGP Update packet and make attribute object. */ |
d62a17ae | 1171 | static int bgp_update_receive(struct peer *peer, bgp_size_t size) |
718e3744 | 1172 | { |
d62a17ae | 1173 | int ret, nlri_ret; |
1174 | u_char *end; | |
1175 | struct stream *s; | |
1176 | struct attr attr; | |
1177 | bgp_size_t attribute_len; | |
1178 | bgp_size_t update_len; | |
1179 | bgp_size_t withdraw_len; | |
1180 | ||
1181 | enum NLRI_TYPES { | |
1182 | NLRI_UPDATE, | |
1183 | NLRI_WITHDRAW, | |
1184 | NLRI_MP_UPDATE, | |
1185 | NLRI_MP_WITHDRAW, | |
1186 | NLRI_TYPE_MAX | |
1187 | }; | |
1188 | struct bgp_nlri nlris[NLRI_TYPE_MAX]; | |
1189 | ||
1190 | /* Status must be Established. */ | |
1191 | if (peer->status != Established) { | |
1192 | zlog_err("%s [FSM] Update packet received under status %s", | |
1193 | peer->host, | |
1194 | lookup_msg(bgp_status_msg, peer->status, NULL)); | |
1195 | bgp_notify_send(peer, BGP_NOTIFY_FSM_ERR, 0); | |
1196 | return -1; | |
1197 | } | |
1198 | ||
1199 | /* Set initial values. */ | |
1200 | memset(&attr, 0, sizeof(struct attr)); | |
1201 | attr.label_index = BGP_INVALID_LABEL_INDEX; | |
1202 | attr.label = MPLS_INVALID_LABEL; | |
1203 | memset(&nlris, 0, sizeof(nlris)); | |
1204 | memset(peer->rcvd_attr_str, 0, BUFSIZ); | |
1205 | peer->rcvd_attr_printed = 0; | |
1206 | ||
1207 | s = peer->ibuf; | |
1208 | end = stream_pnt(s) + size; | |
1209 | ||
1210 | /* RFC1771 6.3 If the Unfeasible Routes Length or Total Attribute | |
1211 | Length is too large (i.e., if Unfeasible Routes Length + Total | |
1212 | Attribute Length + 23 exceeds the message Length), then the Error | |
1213 | Subcode is set to Malformed Attribute List. */ | |
1214 | if (stream_pnt(s) + 2 > end) { | |
1215 | zlog_err( | |
1216 | "%s [Error] Update packet error" | |
1217 | " (packet length is short for unfeasible length)", | |
1218 | peer->host); | |
1219 | bgp_notify_send(peer, BGP_NOTIFY_UPDATE_ERR, | |
1220 | BGP_NOTIFY_UPDATE_MAL_ATTR); | |
1221 | return -1; | |
1222 | } | |
1223 | ||
1224 | /* Unfeasible Route Length. */ | |
1225 | withdraw_len = stream_getw(s); | |
1226 | ||
1227 | /* Unfeasible Route Length check. */ | |
1228 | if (stream_pnt(s) + withdraw_len > end) { | |
1229 | zlog_err( | |
1230 | "%s [Error] Update packet error" | |
1231 | " (packet unfeasible length overflow %d)", | |
1232 | peer->host, withdraw_len); | |
1233 | bgp_notify_send(peer, BGP_NOTIFY_UPDATE_ERR, | |
1234 | BGP_NOTIFY_UPDATE_MAL_ATTR); | |
1235 | return -1; | |
1236 | } | |
1237 | ||
1238 | /* Unfeasible Route packet format check. */ | |
1239 | if (withdraw_len > 0) { | |
1240 | nlris[NLRI_WITHDRAW].afi = AFI_IP; | |
1241 | nlris[NLRI_WITHDRAW].safi = SAFI_UNICAST; | |
1242 | nlris[NLRI_WITHDRAW].nlri = stream_pnt(s); | |
1243 | nlris[NLRI_WITHDRAW].length = withdraw_len; | |
1244 | stream_forward_getp(s, withdraw_len); | |
1245 | } | |
1246 | ||
1247 | /* Attribute total length check. */ | |
1248 | if (stream_pnt(s) + 2 > end) { | |
1249 | zlog_warn( | |
1250 | "%s [Error] Packet Error" | |
1251 | " (update packet is short for attribute length)", | |
1252 | peer->host); | |
1253 | bgp_notify_send(peer, BGP_NOTIFY_UPDATE_ERR, | |
1254 | BGP_NOTIFY_UPDATE_MAL_ATTR); | |
1255 | return -1; | |
1256 | } | |
1257 | ||
1258 | /* Fetch attribute total length. */ | |
1259 | attribute_len = stream_getw(s); | |
1260 | ||
1261 | /* Attribute length check. */ | |
1262 | if (stream_pnt(s) + attribute_len > end) { | |
1263 | zlog_warn( | |
1264 | "%s [Error] Packet Error" | |
1265 | " (update packet attribute length overflow %d)", | |
1266 | peer->host, attribute_len); | |
1267 | bgp_notify_send(peer, BGP_NOTIFY_UPDATE_ERR, | |
1268 | BGP_NOTIFY_UPDATE_MAL_ATTR); | |
1269 | return -1; | |
1270 | } | |
1271 | ||
1272 | /* Certain attribute parsing errors should not be considered bad enough | |
1273 | * to reset the session for, most particularly any partial/optional | |
1274 | * attributes that have 'tunneled' over speakers that don't understand | |
1275 | * them. Instead we withdraw only the prefix concerned. | |
1276 | * | |
1277 | * Complicates the flow a little though.. | |
1278 | */ | |
1279 | bgp_attr_parse_ret_t attr_parse_ret = BGP_ATTR_PARSE_PROCEED; | |
1280 | /* This define morphs the update case into a withdraw when lower levels | |
1281 | * have signalled an error condition where this is best. | |
1282 | */ | |
b881c707 | 1283 | #define NLRI_ATTR_ARG (attr_parse_ret != BGP_ATTR_PARSE_WITHDRAW ? &attr : NULL) |
718e3744 | 1284 | |
d62a17ae | 1285 | /* Parse attribute when it exists. */ |
1286 | if (attribute_len) { | |
1287 | attr_parse_ret = bgp_attr_parse(peer, &attr, attribute_len, | |
1288 | &nlris[NLRI_MP_UPDATE], | |
1289 | &nlris[NLRI_MP_WITHDRAW]); | |
1290 | if (attr_parse_ret == BGP_ATTR_PARSE_ERROR) { | |
1291 | bgp_attr_unintern_sub(&attr); | |
1292 | return -1; | |
1293 | } | |
1294 | } | |
1295 | ||
1296 | /* Logging the attribute. */ | |
1297 | if (attr_parse_ret == BGP_ATTR_PARSE_WITHDRAW | |
1298 | || BGP_DEBUG(update, UPDATE_IN) | |
1299 | || BGP_DEBUG(update, UPDATE_PREFIX)) { | |
1300 | ret = bgp_dump_attr(&attr, peer->rcvd_attr_str, BUFSIZ); | |
1301 | ||
1302 | if (attr_parse_ret == BGP_ATTR_PARSE_WITHDRAW) | |
1303 | zlog_err( | |
1304 | "%s rcvd UPDATE with errors in attr(s)!! Withdrawing route.", | |
1305 | peer->host); | |
1306 | ||
1307 | if (ret && bgp_debug_update(peer, NULL, NULL, 1)) { | |
1308 | zlog_debug("%s rcvd UPDATE w/ attr: %s", peer->host, | |
1309 | peer->rcvd_attr_str); | |
1310 | peer->rcvd_attr_printed = 1; | |
1311 | } | |
1312 | } | |
1313 | ||
1314 | /* Network Layer Reachability Information. */ | |
1315 | update_len = end - stream_pnt(s); | |
1316 | ||
1317 | if (update_len) { | |
1318 | /* Set NLRI portion to structure. */ | |
1319 | nlris[NLRI_UPDATE].afi = AFI_IP; | |
1320 | nlris[NLRI_UPDATE].safi = SAFI_UNICAST; | |
1321 | nlris[NLRI_UPDATE].nlri = stream_pnt(s); | |
1322 | nlris[NLRI_UPDATE].length = update_len; | |
1323 | stream_forward_getp(s, update_len); | |
1324 | } | |
1325 | ||
1326 | if (BGP_DEBUG(update, UPDATE_IN)) | |
1327 | zlog_debug("%s rcvd UPDATE wlen %d attrlen %d alen %d", | |
1328 | peer->host, withdraw_len, attribute_len, update_len); | |
1329 | ||
1330 | /* Parse any given NLRIs */ | |
1331 | for (int i = NLRI_UPDATE; i < NLRI_TYPE_MAX; i++) { | |
1332 | if (!nlris[i].nlri) | |
1333 | continue; | |
1334 | ||
1335 | /* NLRI is processed iff the peer if configured for the specific | |
1336 | * afi/safi */ | |
1337 | if (!peer->afc[nlris[i].afi][nlris[i].safi]) { | |
1338 | zlog_info( | |
1339 | "%s [Info] UPDATE for non-enabled AFI/SAFI %u/%u", | |
1340 | peer->host, nlris[i].afi, nlris[i].safi); | |
1341 | continue; | |
1342 | } | |
1343 | ||
1344 | /* EoR handled later */ | |
1345 | if (nlris[i].length == 0) | |
1346 | continue; | |
1347 | ||
1348 | switch (i) { | |
1349 | case NLRI_UPDATE: | |
1350 | case NLRI_MP_UPDATE: | |
1351 | nlri_ret = bgp_nlri_parse(peer, NLRI_ATTR_ARG, | |
1352 | &nlris[i], 0); | |
1353 | break; | |
1354 | case NLRI_WITHDRAW: | |
1355 | case NLRI_MP_WITHDRAW: | |
1356 | nlri_ret = bgp_nlri_parse(peer, &attr, &nlris[i], 1); | |
1357 | break; | |
1358 | default: | |
1359 | nlri_ret = -1; | |
1360 | } | |
1361 | ||
1362 | if (nlri_ret < 0) { | |
1363 | zlog_err("%s [Error] Error parsing NLRI", peer->host); | |
1364 | if (peer->status == Established) | |
1365 | bgp_notify_send( | |
1366 | peer, BGP_NOTIFY_UPDATE_ERR, | |
1367 | i <= NLRI_WITHDRAW | |
1368 | ? BGP_NOTIFY_UPDATE_INVAL_NETWORK | |
1369 | : BGP_NOTIFY_UPDATE_OPT_ATTR_ERR); | |
1370 | bgp_attr_unintern_sub(&attr); | |
1371 | return -1; | |
1372 | } | |
1373 | } | |
1374 | ||
1375 | /* EoR checks | |
1376 | * | |
1377 | * Non-MP IPv4/Unicast EoR is a completely empty UPDATE | |
1378 | * and MP EoR should have only an empty MP_UNREACH | |
1379 | */ | |
9b9df989 DS |
1380 | if ((!update_len && !withdraw_len && |
1381 | nlris[NLRI_MP_UPDATE].length == 0) || | |
1382 | (attr_parse_ret == BGP_ATTR_PARSE_EOR)) { | |
d62a17ae | 1383 | afi_t afi = 0; |
1384 | safi_t safi; | |
1385 | ||
1386 | /* Non-MP IPv4/Unicast is a completely emtpy UPDATE - already | |
1387 | * checked | |
1388 | * update and withdraw NLRI lengths are 0. | |
1389 | */ | |
1390 | if (!attribute_len) { | |
1391 | afi = AFI_IP; | |
1392 | safi = SAFI_UNICAST; | |
1393 | } else if (attr.flag & ATTR_FLAG_BIT(BGP_ATTR_MP_UNREACH_NLRI) | |
1394 | && nlris[NLRI_MP_WITHDRAW].length == 0) { | |
1395 | afi = nlris[NLRI_MP_WITHDRAW].afi; | |
1396 | safi = nlris[NLRI_MP_WITHDRAW].safi; | |
9b9df989 DS |
1397 | } else if (attr_parse_ret == BGP_ATTR_PARSE_EOR) { |
1398 | afi = nlris[NLRI_MP_UPDATE].afi; | |
1399 | safi = nlris[NLRI_MP_UPDATE].safi; | |
d62a17ae | 1400 | } |
1401 | ||
1402 | if (afi && peer->afc[afi][safi]) { | |
1403 | /* End-of-RIB received */ | |
1404 | if (!CHECK_FLAG(peer->af_sflags[afi][safi], | |
1405 | PEER_STATUS_EOR_RECEIVED)) { | |
1406 | SET_FLAG(peer->af_sflags[afi][safi], | |
1407 | PEER_STATUS_EOR_RECEIVED); | |
1408 | bgp_update_explicit_eors(peer); | |
1409 | } | |
1410 | ||
1411 | /* NSF delete stale route */ | |
1412 | if (peer->nsf[afi][safi]) | |
1413 | bgp_clear_stale_route(peer, afi, safi); | |
1414 | ||
1415 | if (bgp_debug_neighbor_events(peer)) { | |
1416 | zlog_debug("rcvd End-of-RIB for %s from %s", | |
1417 | afi_safi_print(afi, safi), | |
1418 | peer->host); | |
1419 | } | |
1420 | } | |
f80f838b | 1421 | } |
d62a17ae | 1422 | |
1423 | /* Everything is done. We unintern temporary structures which | |
1424 | interned in bgp_attr_parse(). */ | |
1425 | bgp_attr_unintern_sub(&attr); | |
1426 | ||
1427 | /* If peering is stopped due to some reason, do not generate BGP | |
1428 | event. */ | |
1429 | if (peer->status != Established) | |
1430 | return 0; | |
1431 | ||
1432 | /* Increment packet counter. */ | |
1433 | peer->update_in++; | |
1434 | peer->update_time = bgp_clock(); | |
1435 | ||
1436 | /* Rearm holdtime timer */ | |
1437 | BGP_TIMER_OFF(peer->t_holdtime); | |
1438 | bgp_timer_set(peer); | |
1439 | ||
1440 | return 0; | |
718e3744 | 1441 | } |
1442 | ||
1443 | /* Notify message treatment function. */ | |
d62a17ae | 1444 | static void bgp_notify_receive(struct peer *peer, bgp_size_t size) |
718e3744 | 1445 | { |
d62a17ae | 1446 | struct bgp_notify bgp_notify; |
1447 | ||
1448 | if (peer->notify.data) { | |
1449 | XFREE(MTYPE_TMP, peer->notify.data); | |
1450 | peer->notify.data = NULL; | |
1451 | peer->notify.length = 0; | |
1452 | } | |
1453 | ||
1454 | bgp_notify.code = stream_getc(peer->ibuf); | |
1455 | bgp_notify.subcode = stream_getc(peer->ibuf); | |
1456 | bgp_notify.length = size - 2; | |
1457 | bgp_notify.data = NULL; | |
1458 | ||
1459 | /* Preserv notify code and sub code. */ | |
1460 | peer->notify.code = bgp_notify.code; | |
1461 | peer->notify.subcode = bgp_notify.subcode; | |
1462 | /* For further diagnostic record returned Data. */ | |
1463 | if (bgp_notify.length) { | |
1464 | peer->notify.length = size - 2; | |
1465 | peer->notify.data = XMALLOC(MTYPE_TMP, size - 2); | |
1466 | memcpy(peer->notify.data, stream_pnt(peer->ibuf), size - 2); | |
1467 | } | |
1468 | ||
1469 | /* For debug */ | |
1470 | { | |
1471 | int i; | |
1472 | int first = 0; | |
1473 | char c[4]; | |
1474 | ||
1475 | if (bgp_notify.length) { | |
1476 | bgp_notify.data = | |
1477 | XMALLOC(MTYPE_TMP, bgp_notify.length * 3); | |
1478 | for (i = 0; i < bgp_notify.length; i++) | |
1479 | if (first) { | |
1480 | sprintf(c, " %02x", | |
1481 | stream_getc(peer->ibuf)); | |
1482 | strcat(bgp_notify.data, c); | |
1483 | } else { | |
1484 | first = 1; | |
1485 | sprintf(c, "%02x", | |
1486 | stream_getc(peer->ibuf)); | |
1487 | strcpy(bgp_notify.data, c); | |
1488 | } | |
1489 | bgp_notify.raw_data = (u_char *)peer->notify.data; | |
1490 | } | |
1491 | ||
1492 | bgp_notify_print(peer, &bgp_notify, "received"); | |
1493 | if (bgp_notify.data) { | |
1494 | XFREE(MTYPE_TMP, bgp_notify.data); | |
1495 | bgp_notify.data = NULL; | |
1496 | bgp_notify.length = 0; | |
1497 | } | |
1498 | } | |
1499 | ||
1500 | /* peer count update */ | |
1501 | peer->notify_in++; | |
1502 | ||
1503 | peer->last_reset = PEER_DOWN_NOTIFY_RECEIVED; | |
1504 | ||
1505 | /* We have to check for Notify with Unsupported Optional Parameter. | |
1506 | in that case we fallback to open without the capability option. | |
1507 | But this done in bgp_stop. We just mark it here to avoid changing | |
1508 | the fsm tables. */ | |
1509 | if (bgp_notify.code == BGP_NOTIFY_OPEN_ERR | |
1510 | && bgp_notify.subcode == BGP_NOTIFY_OPEN_UNSUP_PARAM) | |
1511 | UNSET_FLAG(peer->sflags, PEER_STATUS_CAPABILITY_OPEN); | |
1512 | ||
1513 | BGP_EVENT_ADD(peer, Receive_NOTIFICATION_message); | |
718e3744 | 1514 | } |
1515 | ||
1516 | /* Keepalive treatment function -- get keepalive send keepalive */ | |
d62a17ae | 1517 | static void bgp_keepalive_receive(struct peer *peer, bgp_size_t size) |
718e3744 | 1518 | { |
d62a17ae | 1519 | if (bgp_debug_keepalive(peer)) |
1520 | zlog_debug("%s KEEPALIVE rcvd", peer->host); | |
1521 | ||
1522 | BGP_EVENT_ADD(peer, Receive_KEEPALIVE_message); | |
718e3744 | 1523 | } |
1524 | ||
1525 | /* Route refresh message is received. */ | |
d62a17ae | 1526 | static void bgp_route_refresh_receive(struct peer *peer, bgp_size_t size) |
718e3744 | 1527 | { |
d62a17ae | 1528 | iana_afi_t pkt_afi; |
1529 | afi_t afi; | |
5c525538 RW |
1530 | iana_safi_t pkt_safi; |
1531 | safi_t safi; | |
d62a17ae | 1532 | struct stream *s; |
1533 | struct peer_af *paf; | |
1534 | struct update_group *updgrp; | |
1535 | struct peer *updgrp_peer; | |
1536 | ||
1537 | /* If peer does not have the capability, send notification. */ | |
1538 | if (!CHECK_FLAG(peer->cap, PEER_CAP_REFRESH_ADV)) { | |
1539 | zlog_err("%s [Error] BGP route refresh is not enabled", | |
1540 | peer->host); | |
1541 | bgp_notify_send(peer, BGP_NOTIFY_HEADER_ERR, | |
1542 | BGP_NOTIFY_HEADER_BAD_MESTYPE); | |
1543 | return; | |
1544 | } | |
1545 | ||
1546 | /* Status must be Established. */ | |
1547 | if (peer->status != Established) { | |
1548 | zlog_err( | |
1549 | "%s [Error] Route refresh packet received under status %s", | |
1550 | peer->host, | |
1551 | lookup_msg(bgp_status_msg, peer->status, NULL)); | |
1552 | bgp_notify_send(peer, BGP_NOTIFY_FSM_ERR, 0); | |
1553 | return; | |
1554 | } | |
1555 | ||
1556 | s = peer->ibuf; | |
1557 | ||
1558 | /* Parse packet. */ | |
1559 | pkt_afi = stream_getw(s); | |
1560 | (void)stream_getc(s); | |
1561 | pkt_safi = stream_getc(s); | |
1562 | ||
1563 | if (bgp_debug_update(peer, NULL, NULL, 0)) | |
1564 | zlog_debug("%s rcvd REFRESH_REQ for afi/safi: %d/%d", | |
1565 | peer->host, pkt_afi, pkt_safi); | |
1566 | ||
1567 | /* Convert AFI, SAFI to internal values and check. */ | |
1568 | if (bgp_map_afi_safi_iana2int(pkt_afi, pkt_safi, &afi, &safi)) { | |
1569 | zlog_info( | |
1570 | "%s REFRESH_REQ for unrecognized afi/safi: %d/%d - ignored", | |
1571 | peer->host, pkt_afi, pkt_safi); | |
1572 | return; | |
1573 | } | |
1574 | ||
1575 | if (size != BGP_MSG_ROUTE_REFRESH_MIN_SIZE - BGP_HEADER_SIZE) { | |
1576 | u_char *end; | |
1577 | u_char when_to_refresh; | |
1578 | u_char orf_type; | |
1579 | u_int16_t orf_len; | |
1580 | ||
1581 | if (size - (BGP_MSG_ROUTE_REFRESH_MIN_SIZE - BGP_HEADER_SIZE) | |
1582 | < 5) { | |
1583 | zlog_info("%s ORF route refresh length error", | |
1584 | peer->host); | |
1585 | bgp_notify_send(peer, BGP_NOTIFY_CEASE, 0); | |
1586 | return; | |
718e3744 | 1587 | } |
1588 | ||
d62a17ae | 1589 | when_to_refresh = stream_getc(s); |
1590 | end = stream_pnt(s) + (size - 5); | |
1591 | ||
1592 | while ((stream_pnt(s) + 2) < end) { | |
1593 | orf_type = stream_getc(s); | |
1594 | orf_len = stream_getw(s); | |
1595 | ||
1596 | /* orf_len in bounds? */ | |
1597 | if ((stream_pnt(s) + orf_len) > end) | |
1598 | break; /* XXX: Notify instead?? */ | |
1599 | if (orf_type == ORF_TYPE_PREFIX | |
1600 | || orf_type == ORF_TYPE_PREFIX_OLD) { | |
1601 | uint8_t *p_pnt = stream_pnt(s); | |
1602 | uint8_t *p_end = stream_pnt(s) + orf_len; | |
1603 | struct orf_prefix orfp; | |
1604 | u_char common = 0; | |
1605 | u_int32_t seq; | |
1606 | int psize; | |
1607 | char name[BUFSIZ]; | |
1608 | int ret = CMD_SUCCESS; | |
1609 | ||
1610 | if (bgp_debug_neighbor_events(peer)) { | |
1611 | zlog_debug( | |
1612 | "%s rcvd Prefixlist ORF(%d) length %d", | |
1613 | peer->host, orf_type, orf_len); | |
1614 | } | |
1615 | ||
1616 | /* we're going to read at least 1 byte of common | |
1617 | * ORF header, | |
1618 | * and 7 bytes of ORF Address-filter entry from | |
1619 | * the stream | |
1620 | */ | |
1621 | if (orf_len < 7) | |
1622 | break; | |
1623 | ||
1624 | /* ORF prefix-list name */ | |
1625 | sprintf(name, "%s.%d.%d", peer->host, afi, | |
1626 | safi); | |
1627 | ||
1628 | while (p_pnt < p_end) { | |
1629 | /* If the ORF entry is malformed, want | |
1630 | * to read as much of it | |
1631 | * as possible without going beyond the | |
1632 | * bounds of the entry, | |
1633 | * to maximise debug information. | |
1634 | */ | |
1635 | int ok; | |
1636 | memset(&orfp, 0, | |
1637 | sizeof(struct orf_prefix)); | |
1638 | common = *p_pnt++; | |
1639 | /* after ++: p_pnt <= p_end */ | |
1640 | if (common | |
1641 | & ORF_COMMON_PART_REMOVE_ALL) { | |
1642 | if (bgp_debug_neighbor_events( | |
1643 | peer)) | |
1644 | zlog_debug( | |
1645 | "%s rcvd Remove-All pfxlist ORF request", | |
1646 | peer->host); | |
1647 | prefix_bgp_orf_remove_all(afi, | |
1648 | name); | |
1649 | break; | |
1650 | } | |
1651 | ok = ((u_int32_t)(p_end - p_pnt) | |
1652 | >= sizeof(u_int32_t)); | |
1653 | if (ok) { | |
1654 | memcpy(&seq, p_pnt, | |
1655 | sizeof(u_int32_t)); | |
1656 | p_pnt += sizeof(u_int32_t); | |
1657 | orfp.seq = ntohl(seq); | |
1658 | } else | |
1659 | p_pnt = p_end; | |
1660 | ||
1661 | if ((ok = (p_pnt < p_end))) | |
1662 | orfp.ge = | |
1663 | *p_pnt++; /* value | |
1664 | checked in | |
1665 | prefix_bgp_orf_set() | |
1666 | */ | |
1667 | if ((ok = (p_pnt < p_end))) | |
1668 | orfp.le = | |
1669 | *p_pnt++; /* value | |
1670 | checked in | |
1671 | prefix_bgp_orf_set() | |
1672 | */ | |
1673 | if ((ok = (p_pnt < p_end))) | |
1674 | orfp.p.prefixlen = *p_pnt++; | |
1675 | orfp.p.family = afi2family( | |
1676 | afi); /* afi checked already */ | |
1677 | ||
1678 | psize = PSIZE( | |
1679 | orfp.p.prefixlen); /* 0 if not | |
1680 | ok */ | |
1681 | if (psize | |
1682 | > prefix_blen( | |
1683 | &orfp.p)) /* valid for | |
1684 | family ? */ | |
1685 | { | |
1686 | ok = 0; | |
1687 | psize = prefix_blen(&orfp.p); | |
1688 | } | |
1689 | if (psize | |
1690 | > (p_end - p_pnt)) /* valid for | |
1691 | packet ? */ | |
1692 | { | |
1693 | ok = 0; | |
1694 | psize = p_end - p_pnt; | |
1695 | } | |
1696 | ||
1697 | if (psize > 0) | |
1698 | memcpy(&orfp.p.u.prefix, p_pnt, | |
1699 | psize); | |
1700 | p_pnt += psize; | |
1701 | ||
1702 | if (bgp_debug_neighbor_events(peer)) { | |
1703 | char buf[INET6_BUFSIZ]; | |
1704 | ||
1705 | zlog_debug( | |
1706 | "%s rcvd %s %s seq %u %s/%d ge %d le %d%s", | |
1707 | peer->host, | |
1708 | (common & ORF_COMMON_PART_REMOVE | |
1709 | ? "Remove" | |
1710 | : "Add"), | |
1711 | (common & ORF_COMMON_PART_DENY | |
1712 | ? "deny" | |
1713 | : "permit"), | |
1714 | orfp.seq, | |
1715 | inet_ntop( | |
1716 | orfp.p.family, | |
1717 | &orfp.p.u.prefix, | |
1718 | buf, | |
1719 | INET6_BUFSIZ), | |
1720 | orfp.p.prefixlen, | |
1721 | orfp.ge, orfp.le, | |
1722 | ok ? "" : " MALFORMED"); | |
1723 | } | |
1724 | ||
1725 | if (ok) | |
1726 | ret = prefix_bgp_orf_set( | |
1727 | name, afi, &orfp, | |
1728 | (common & ORF_COMMON_PART_DENY | |
1729 | ? 0 | |
1730 | : 1), | |
1731 | (common & ORF_COMMON_PART_REMOVE | |
1732 | ? 0 | |
1733 | : 1)); | |
1734 | ||
1735 | if (!ok || (ok && ret != CMD_SUCCESS)) { | |
1736 | zlog_info( | |
1737 | "%s Received misformatted prefixlist ORF." | |
1738 | " Remove All pfxlist", | |
1739 | peer->host); | |
1740 | prefix_bgp_orf_remove_all(afi, | |
1741 | name); | |
1742 | break; | |
1743 | } | |
1744 | } | |
1745 | ||
1746 | peer->orf_plist[afi][safi] = | |
1747 | prefix_bgp_orf_lookup(afi, name); | |
1748 | } | |
1749 | stream_forward_getp(s, orf_len); | |
718e3744 | 1750 | } |
d62a17ae | 1751 | if (bgp_debug_neighbor_events(peer)) |
1752 | zlog_debug("%s rcvd Refresh %s ORF request", peer->host, | |
1753 | when_to_refresh == REFRESH_DEFER | |
1754 | ? "Defer" | |
1755 | : "Immediate"); | |
1756 | if (when_to_refresh == REFRESH_DEFER) | |
1757 | return; | |
1758 | } | |
40d2700d | 1759 | |
d62a17ae | 1760 | /* First update is deferred until ORF or ROUTE-REFRESH is received */ |
1761 | if (CHECK_FLAG(peer->af_sflags[afi][safi], | |
1762 | PEER_STATUS_ORF_WAIT_REFRESH)) | |
1763 | UNSET_FLAG(peer->af_sflags[afi][safi], | |
1764 | PEER_STATUS_ORF_WAIT_REFRESH); | |
1765 | ||
1766 | paf = peer_af_find(peer, afi, safi); | |
1767 | if (paf && paf->subgroup) { | |
1768 | if (peer->orf_plist[afi][safi]) { | |
1769 | updgrp = PAF_UPDGRP(paf); | |
1770 | updgrp_peer = UPDGRP_PEER(updgrp); | |
1771 | updgrp_peer->orf_plist[afi][safi] = | |
1772 | peer->orf_plist[afi][safi]; | |
1773 | } | |
1774 | ||
1775 | /* If the peer is configured for default-originate clear the | |
1776 | * SUBGRP_STATUS_DEFAULT_ORIGINATE flag so that we will | |
1777 | * re-advertise the | |
1778 | * default | |
1779 | */ | |
1780 | if (CHECK_FLAG(paf->subgroup->sflags, | |
1781 | SUBGRP_STATUS_DEFAULT_ORIGINATE)) | |
1782 | UNSET_FLAG(paf->subgroup->sflags, | |
1783 | SUBGRP_STATUS_DEFAULT_ORIGINATE); | |
718e3744 | 1784 | } |
d62a17ae | 1785 | |
1786 | /* Perform route refreshment to the peer */ | |
1787 | bgp_announce_route(peer, afi, safi); | |
718e3744 | 1788 | } |
1789 | ||
d62a17ae | 1790 | static int bgp_capability_msg_parse(struct peer *peer, u_char *pnt, |
1791 | bgp_size_t length) | |
718e3744 | 1792 | { |
d62a17ae | 1793 | u_char *end; |
1794 | struct capability_mp_data mpc; | |
1795 | struct capability_header *hdr; | |
1796 | u_char action; | |
1797 | iana_afi_t pkt_afi; | |
1798 | afi_t afi; | |
5c525538 RW |
1799 | iana_safi_t pkt_safi; |
1800 | safi_t safi; | |
d62a17ae | 1801 | |
1802 | end = pnt + length; | |
1803 | ||
1804 | while (pnt < end) { | |
1805 | /* We need at least action, capability code and capability | |
1806 | * length. */ | |
1807 | if (pnt + 3 > end) { | |
1808 | zlog_info("%s Capability length error", peer->host); | |
1809 | bgp_notify_send(peer, BGP_NOTIFY_CEASE, 0); | |
1810 | return -1; | |
1811 | } | |
1812 | action = *pnt; | |
1813 | hdr = (struct capability_header *)(pnt + 1); | |
1814 | ||
1815 | /* Action value check. */ | |
1816 | if (action != CAPABILITY_ACTION_SET | |
1817 | && action != CAPABILITY_ACTION_UNSET) { | |
1818 | zlog_info("%s Capability Action Value error %d", | |
1819 | peer->host, action); | |
1820 | bgp_notify_send(peer, BGP_NOTIFY_CEASE, 0); | |
1821 | return -1; | |
1822 | } | |
1823 | ||
1824 | if (bgp_debug_neighbor_events(peer)) | |
1825 | zlog_debug( | |
1826 | "%s CAPABILITY has action: %d, code: %u, length %u", | |
1827 | peer->host, action, hdr->code, hdr->length); | |
1828 | ||
1829 | /* Capability length check. */ | |
1830 | if ((pnt + hdr->length + 3) > end) { | |
1831 | zlog_info("%s Capability length error", peer->host); | |
1832 | bgp_notify_send(peer, BGP_NOTIFY_CEASE, 0); | |
1833 | return -1; | |
1834 | } | |
1835 | ||
1836 | /* Fetch structure to the byte stream. */ | |
1837 | memcpy(&mpc, pnt + 3, sizeof(struct capability_mp_data)); | |
d2b6417b | 1838 | pnt += hdr->length + 3; |
d62a17ae | 1839 | |
1840 | /* We know MP Capability Code. */ | |
1841 | if (hdr->code == CAPABILITY_CODE_MP) { | |
1842 | pkt_afi = ntohs(mpc.afi); | |
1843 | pkt_safi = mpc.safi; | |
1844 | ||
1845 | /* Ignore capability when override-capability is set. */ | |
1846 | if (CHECK_FLAG(peer->flags, | |
1847 | PEER_FLAG_OVERRIDE_CAPABILITY)) | |
1848 | continue; | |
1849 | ||
1850 | /* Convert AFI, SAFI to internal values. */ | |
1851 | if (bgp_map_afi_safi_iana2int(pkt_afi, pkt_safi, &afi, | |
1852 | &safi)) { | |
1853 | if (bgp_debug_neighbor_events(peer)) | |
1854 | zlog_debug( | |
1855 | "%s Dynamic Capability MP_EXT afi/safi invalid " | |
1856 | "(%u/%u)", | |
1857 | peer->host, pkt_afi, pkt_safi); | |
1858 | continue; | |
1859 | } | |
1860 | ||
1861 | /* Address family check. */ | |
1862 | if (bgp_debug_neighbor_events(peer)) | |
1863 | zlog_debug( | |
1864 | "%s CAPABILITY has %s MP_EXT CAP for afi/safi: %u/%u", | |
1865 | peer->host, | |
1866 | action == CAPABILITY_ACTION_SET | |
1867 | ? "Advertising" | |
1868 | : "Removing", | |
1869 | pkt_afi, pkt_safi); | |
1870 | ||
1871 | if (action == CAPABILITY_ACTION_SET) { | |
1872 | peer->afc_recv[afi][safi] = 1; | |
1873 | if (peer->afc[afi][safi]) { | |
1874 | peer->afc_nego[afi][safi] = 1; | |
1875 | bgp_announce_route(peer, afi, safi); | |
1876 | } | |
1877 | } else { | |
1878 | peer->afc_recv[afi][safi] = 0; | |
1879 | peer->afc_nego[afi][safi] = 0; | |
1880 | ||
1881 | if (peer_active_nego(peer)) | |
1882 | bgp_clear_route(peer, afi, safi); | |
1883 | else | |
1884 | BGP_EVENT_ADD(peer, BGP_Stop); | |
1885 | } | |
1886 | } else { | |
1887 | zlog_warn( | |
1888 | "%s unrecognized capability code: %d - ignored", | |
1889 | peer->host, hdr->code); | |
1890 | } | |
d62a17ae | 1891 | } |
1892 | return 0; | |
718e3744 | 1893 | } |
1894 | ||
d62a17ae | 1895 | /* Dynamic Capability is received. |
01b7ce2d PJ |
1896 | * |
1897 | * This is exported for unit-test purposes | |
1898 | */ | |
d62a17ae | 1899 | int bgp_capability_receive(struct peer *peer, bgp_size_t size) |
718e3744 | 1900 | { |
d62a17ae | 1901 | u_char *pnt; |
1902 | ||
1903 | /* Fetch pointer. */ | |
1904 | pnt = stream_pnt(peer->ibuf); | |
1905 | ||
1906 | if (bgp_debug_neighbor_events(peer)) | |
1907 | zlog_debug("%s rcv CAPABILITY", peer->host); | |
1908 | ||
1909 | /* If peer does not have the capability, send notification. */ | |
1910 | if (!CHECK_FLAG(peer->cap, PEER_CAP_DYNAMIC_ADV)) { | |
1911 | zlog_err("%s [Error] BGP dynamic capability is not enabled", | |
1912 | peer->host); | |
1913 | bgp_notify_send(peer, BGP_NOTIFY_HEADER_ERR, | |
1914 | BGP_NOTIFY_HEADER_BAD_MESTYPE); | |
1915 | return -1; | |
1916 | } | |
1917 | ||
1918 | /* Status must be Established. */ | |
1919 | if (peer->status != Established) { | |
1920 | zlog_err( | |
1921 | "%s [Error] Dynamic capability packet received under status %s", | |
1922 | peer->host, | |
1923 | lookup_msg(bgp_status_msg, peer->status, NULL)); | |
1924 | bgp_notify_send(peer, BGP_NOTIFY_FSM_ERR, 0); | |
1925 | return -1; | |
1926 | } | |
1927 | ||
1928 | /* Parse packet. */ | |
1929 | return bgp_capability_msg_parse(peer, pnt, size); | |
718e3744 | 1930 | } |
6b0655a2 | 1931 | |
718e3744 | 1932 | /* BGP read utility function. */ |
d62a17ae | 1933 | static int bgp_read_packet(struct peer *peer) |
718e3744 | 1934 | { |
d62a17ae | 1935 | int nbytes; |
1936 | int readsize; | |
718e3744 | 1937 | |
d62a17ae | 1938 | readsize = peer->packet_size - stream_get_endp(peer->ibuf); |
718e3744 | 1939 | |
d62a17ae | 1940 | /* If size is zero then return. */ |
1941 | if (!readsize) | |
1942 | return 0; | |
718e3744 | 1943 | |
d62a17ae | 1944 | /* Read packet from fd. */ |
1945 | nbytes = stream_read_try(peer->ibuf, peer->fd, readsize); | |
718e3744 | 1946 | |
d62a17ae | 1947 | /* If read byte is smaller than zero then error occured. */ |
1948 | if (nbytes < 0) { | |
1949 | /* Transient error should retry */ | |
1950 | if (nbytes == -2) | |
1951 | return -1; | |
718e3744 | 1952 | |
d62a17ae | 1953 | zlog_err("%s [Error] bgp_read_packet error: %s", peer->host, |
1954 | safe_strerror(errno)); | |
93406d87 | 1955 | |
d62a17ae | 1956 | if (peer->status == Established) { |
1957 | if (CHECK_FLAG(peer->sflags, PEER_STATUS_NSF_MODE)) { | |
1958 | peer->last_reset = PEER_DOWN_NSF_CLOSE_SESSION; | |
1959 | SET_FLAG(peer->sflags, PEER_STATUS_NSF_WAIT); | |
1960 | } else | |
1961 | peer->last_reset = PEER_DOWN_CLOSE_SESSION; | |
1962 | } | |
93406d87 | 1963 | |
d62a17ae | 1964 | BGP_EVENT_ADD(peer, TCP_fatal_error); |
1965 | return -1; | |
1966 | } | |
718e3744 | 1967 | |
d62a17ae | 1968 | /* When read byte is zero : clear bgp peer and return */ |
1969 | if (nbytes == 0) { | |
1970 | if (bgp_debug_neighbor_events(peer)) | |
1971 | zlog_debug("%s [Event] BGP connection closed fd %d", | |
1972 | peer->host, peer->fd); | |
1973 | ||
1974 | if (peer->status == Established) { | |
1975 | if (CHECK_FLAG(peer->sflags, PEER_STATUS_NSF_MODE)) { | |
1976 | peer->last_reset = PEER_DOWN_NSF_CLOSE_SESSION; | |
1977 | SET_FLAG(peer->sflags, PEER_STATUS_NSF_WAIT); | |
1978 | } else | |
1979 | peer->last_reset = PEER_DOWN_CLOSE_SESSION; | |
1980 | } | |
e0701b79 | 1981 | |
d62a17ae | 1982 | BGP_EVENT_ADD(peer, TCP_connection_closed); |
1983 | return -1; | |
93406d87 | 1984 | } |
e0701b79 | 1985 | |
d62a17ae | 1986 | /* We read partial packet. */ |
1987 | if (stream_get_endp(peer->ibuf) != peer->packet_size) | |
1988 | return -1; | |
718e3744 | 1989 | |
d62a17ae | 1990 | return 0; |
718e3744 | 1991 | } |
1992 | ||
1993 | /* Marker check. */ | |
d62a17ae | 1994 | static int bgp_marker_all_one(struct stream *s, int length) |
718e3744 | 1995 | { |
d62a17ae | 1996 | int i; |
718e3744 | 1997 | |
d62a17ae | 1998 | for (i = 0; i < length; i++) |
1999 | if (s->data[i] != 0xff) | |
2000 | return 0; | |
718e3744 | 2001 | |
d62a17ae | 2002 | return 1; |
718e3744 | 2003 | } |
2004 | ||
2005 | /* Starting point of packet process function. */ | |
d62a17ae | 2006 | int bgp_read(struct thread *thread) |
718e3744 | 2007 | { |
d62a17ae | 2008 | int ret; |
2009 | u_char type = 0; | |
2010 | struct peer *peer; | |
2011 | bgp_size_t size; | |
2012 | char notify_data_length[2]; | |
2013 | u_int32_t notify_out; | |
2014 | ||
2015 | /* Yes first of all get peer pointer. */ | |
2016 | peer = THREAD_ARG(thread); | |
2017 | peer->t_read = NULL; | |
2018 | ||
2019 | /* Note notify_out so we can check later to see if we sent another one | |
2020 | */ | |
2021 | notify_out = peer->notify_out; | |
2022 | ||
2023 | /* For non-blocking IO check. */ | |
2024 | if (peer->status == Connect) { | |
2025 | bgp_connect_check(peer, 1); | |
2026 | goto done; | |
2027 | } else { | |
2028 | if (peer->fd < 0) { | |
2029 | zlog_err("bgp_read peer's fd is negative value %d", | |
2030 | peer->fd); | |
2031 | return -1; | |
2032 | } | |
2033 | BGP_READ_ON(peer->t_read, bgp_read, peer->fd); | |
718e3744 | 2034 | } |
d62a17ae | 2035 | |
2036 | /* Read packet header to determine type of the packet */ | |
2037 | if (peer->packet_size == 0) | |
2038 | peer->packet_size = BGP_HEADER_SIZE; | |
2039 | ||
2040 | if (stream_get_endp(peer->ibuf) < BGP_HEADER_SIZE) { | |
2041 | ret = bgp_read_packet(peer); | |
2042 | ||
2043 | /* Header read error or partial read packet. */ | |
2044 | if (ret < 0) | |
2045 | goto done; | |
2046 | ||
2047 | /* Get size and type. */ | |
2048 | stream_forward_getp(peer->ibuf, BGP_MARKER_SIZE); | |
2049 | memcpy(notify_data_length, stream_pnt(peer->ibuf), 2); | |
2050 | size = stream_getw(peer->ibuf); | |
2051 | type = stream_getc(peer->ibuf); | |
2052 | ||
2053 | /* Marker check */ | |
2054 | if (((type == BGP_MSG_OPEN) || (type == BGP_MSG_KEEPALIVE)) | |
2055 | && !bgp_marker_all_one(peer->ibuf, BGP_MARKER_SIZE)) { | |
2056 | bgp_notify_send(peer, BGP_NOTIFY_HEADER_ERR, | |
2057 | BGP_NOTIFY_HEADER_NOT_SYNC); | |
2058 | goto done; | |
2059 | } | |
2060 | ||
2061 | /* BGP type check. */ | |
2062 | if (type != BGP_MSG_OPEN && type != BGP_MSG_UPDATE | |
2063 | && type != BGP_MSG_NOTIFY && type != BGP_MSG_KEEPALIVE | |
2064 | && type != BGP_MSG_ROUTE_REFRESH_NEW | |
2065 | && type != BGP_MSG_ROUTE_REFRESH_OLD | |
2066 | && type != BGP_MSG_CAPABILITY) { | |
2067 | if (bgp_debug_neighbor_events(peer)) | |
2068 | zlog_debug("%s unknown message type 0x%02x", | |
2069 | peer->host, type); | |
2070 | bgp_notify_send_with_data(peer, BGP_NOTIFY_HEADER_ERR, | |
2071 | BGP_NOTIFY_HEADER_BAD_MESTYPE, | |
2072 | &type, 1); | |
2073 | goto done; | |
2074 | } | |
2075 | /* Mimimum packet length check. */ | |
2076 | if ((size < BGP_HEADER_SIZE) || (size > BGP_MAX_PACKET_SIZE) | |
2077 | || (type == BGP_MSG_OPEN && size < BGP_MSG_OPEN_MIN_SIZE) | |
2078 | || (type == BGP_MSG_UPDATE | |
2079 | && size < BGP_MSG_UPDATE_MIN_SIZE) | |
2080 | || (type == BGP_MSG_NOTIFY | |
2081 | && size < BGP_MSG_NOTIFY_MIN_SIZE) | |
2082 | || (type == BGP_MSG_KEEPALIVE | |
2083 | && size != BGP_MSG_KEEPALIVE_MIN_SIZE) | |
2084 | || (type == BGP_MSG_ROUTE_REFRESH_NEW | |
2085 | && size < BGP_MSG_ROUTE_REFRESH_MIN_SIZE) | |
2086 | || (type == BGP_MSG_ROUTE_REFRESH_OLD | |
2087 | && size < BGP_MSG_ROUTE_REFRESH_MIN_SIZE) | |
2088 | || (type == BGP_MSG_CAPABILITY | |
2089 | && size < BGP_MSG_CAPABILITY_MIN_SIZE)) { | |
2090 | if (bgp_debug_neighbor_events(peer)) | |
2091 | zlog_debug("%s bad message length - %d for %s", | |
2092 | peer->host, size, | |
2093 | type == 128 | |
2094 | ? "ROUTE-REFRESH" | |
2095 | : bgp_type_str[(int)type]); | |
2096 | bgp_notify_send_with_data(peer, BGP_NOTIFY_HEADER_ERR, | |
2097 | BGP_NOTIFY_HEADER_BAD_MESLEN, | |
2098 | (u_char *)notify_data_length, | |
2099 | 2); | |
2100 | goto done; | |
2101 | } | |
2102 | ||
2103 | /* Adjust size to message length. */ | |
2104 | peer->packet_size = size; | |
718e3744 | 2105 | } |
2106 | ||
d62a17ae | 2107 | ret = bgp_read_packet(peer); |
2108 | if (ret < 0) | |
2109 | goto done; | |
2110 | ||
2111 | /* Get size and type again. */ | |
2112 | (void)stream_getw_from(peer->ibuf, BGP_MARKER_SIZE); | |
2113 | type = stream_getc_from(peer->ibuf, BGP_MARKER_SIZE + 2); | |
2114 | ||
2115 | /* BGP packet dump function. */ | |
2116 | bgp_dump_packet(peer, type, peer->ibuf); | |
2117 | ||
2118 | size = (peer->packet_size - BGP_HEADER_SIZE); | |
2119 | ||
2120 | /* Read rest of the packet and call each sort of packet routine */ | |
2121 | switch (type) { | |
2122 | case BGP_MSG_OPEN: | |
2123 | peer->open_in++; | |
2124 | bgp_open_receive(peer, size); /* XXX return value ignored! */ | |
2125 | break; | |
2126 | case BGP_MSG_UPDATE: | |
2127 | peer->readtime = monotime(NULL); | |
2128 | bgp_update_receive(peer, size); | |
2129 | break; | |
2130 | case BGP_MSG_NOTIFY: | |
2131 | bgp_notify_receive(peer, size); | |
2132 | break; | |
2133 | case BGP_MSG_KEEPALIVE: | |
2134 | peer->readtime = monotime(NULL); | |
2135 | bgp_keepalive_receive(peer, size); | |
2136 | break; | |
2137 | case BGP_MSG_ROUTE_REFRESH_NEW: | |
2138 | case BGP_MSG_ROUTE_REFRESH_OLD: | |
2139 | peer->refresh_in++; | |
2140 | bgp_route_refresh_receive(peer, size); | |
2141 | break; | |
2142 | case BGP_MSG_CAPABILITY: | |
2143 | peer->dynamic_cap_in++; | |
2144 | bgp_capability_receive(peer, size); | |
2145 | break; | |
718e3744 | 2146 | } |
d62a17ae | 2147 | |
2148 | /* If reading this packet caused us to send a NOTIFICATION then store a | |
2149 | * copy | |
2150 | * of the packet for troubleshooting purposes | |
2151 | */ | |
2152 | if (notify_out < peer->notify_out) { | |
2153 | memcpy(peer->last_reset_cause, peer->ibuf->data, | |
2154 | peer->packet_size); | |
2155 | peer->last_reset_cause_size = peer->packet_size; | |
2156 | notify_out = peer->notify_out; | |
2157 | } | |
2158 | ||
2159 | /* Clear input buffer. */ | |
2160 | peer->packet_size = 0; | |
2161 | if (peer->ibuf) | |
2162 | stream_reset(peer->ibuf); | |
2163 | ||
2164 | done: | |
2165 | /* If reading this packet caused us to send a NOTIFICATION then store a | |
2166 | * copy | |
2167 | * of the packet for troubleshooting purposes | |
2168 | */ | |
2169 | if (notify_out < peer->notify_out) { | |
2170 | memcpy(peer->last_reset_cause, peer->ibuf->data, | |
2171 | peer->packet_size); | |
2172 | peer->last_reset_cause_size = peer->packet_size; | |
718e3744 | 2173 | } |
2174 | ||
d62a17ae | 2175 | return 0; |
718e3744 | 2176 | } |
d3ecc69e QY |
2177 | |
2178 | /* ------------- write thread ------------------ */ | |
2179 | ||
2180 | /** | |
2181 | * Flush peer output buffer. | |
2182 | * | |
2183 | * This function pops packets off of peer->obuf and writes them to peer->fd. | |
2184 | * The amount of packets written is equal to the minimum of peer->wpkt_quanta | |
2185 | * and the number of packets on the output buffer. | |
2186 | * | |
2187 | * If write() returns an error, the appropriate FSM event is generated. | |
2188 | * | |
2189 | * The return value is equal to the number of packets written | |
2190 | * (which may be zero). | |
2191 | */ | |
2192 | static int bgp_write(struct peer *peer) | |
2193 | { | |
2194 | u_char type; | |
2195 | struct stream *s; | |
2196 | int num; | |
2197 | int update_last_write = 0; | |
2198 | unsigned int count = 0; | |
2199 | unsigned int oc = 0; | |
2200 | ||
2201 | /* For non-blocking IO check. */ | |
2202 | if (peer->status == Connect) { | |
2203 | bgp_connect_check(peer, 1); | |
2204 | return 0; | |
2205 | } | |
2206 | ||
2207 | /* Write packets. The number of packets written is the value of | |
2208 | * bgp->wpkt_quanta or the size of the output buffer, whichever is | |
2209 | * smaller.*/ | |
2210 | while (count < peer->bgp->wpkt_quanta | |
2211 | && (s = bgp_write_packet(peer)) != NULL) { | |
2212 | int writenum; | |
2213 | do { // write a full packet, or return on error | |
2214 | writenum = stream_get_endp(s) - stream_get_getp(s); | |
2215 | num = write(peer->fd, STREAM_PNT(s), writenum); | |
2216 | ||
2217 | if (num < 0) { | |
2218 | if (ERRNO_IO_RETRY(errno)) | |
2219 | continue; | |
2220 | ||
2221 | BGP_EVENT_ADD(peer, TCP_fatal_error); | |
2222 | goto done; | |
2223 | } else if (num != writenum) // incomplete write | |
2224 | stream_forward_getp(s, num); | |
2225 | ||
2226 | } while (num != writenum); | |
2227 | ||
2228 | /* Retrieve BGP packet type. */ | |
2229 | stream_set_getp(s, BGP_MARKER_SIZE + 2); | |
2230 | type = stream_getc(s); | |
2231 | ||
2232 | switch (type) { | |
2233 | case BGP_MSG_OPEN: | |
2234 | peer->open_out++; | |
2235 | break; | |
2236 | case BGP_MSG_UPDATE: | |
2237 | peer->update_out++; | |
2238 | break; | |
2239 | case BGP_MSG_NOTIFY: | |
2240 | peer->notify_out++; | |
2241 | /* Double start timer. */ | |
2242 | peer->v_start *= 2; | |
2243 | ||
2244 | /* Overflow check. */ | |
2245 | if (peer->v_start >= (60 * 2)) | |
2246 | peer->v_start = (60 * 2); | |
2247 | ||
2248 | /* Handle Graceful Restart case where the state changes | |
2249 | to | |
2250 | Connect instead of Idle */ | |
2251 | /* Flush any existing events */ | |
2252 | BGP_EVENT_ADD(peer, BGP_Stop); | |
2253 | goto done; | |
2254 | ||
2255 | case BGP_MSG_KEEPALIVE: | |
2256 | peer->keepalive_out++; | |
2257 | break; | |
2258 | case BGP_MSG_ROUTE_REFRESH_NEW: | |
2259 | case BGP_MSG_ROUTE_REFRESH_OLD: | |
2260 | peer->refresh_out++; | |
2261 | break; | |
2262 | case BGP_MSG_CAPABILITY: | |
2263 | peer->dynamic_cap_out++; | |
2264 | break; | |
2265 | } | |
2266 | ||
2267 | count++; | |
2268 | /* OK we send packet so delete it. */ | |
2269 | bgp_packet_delete_unsafe(peer); | |
2270 | update_last_write = 1; | |
2271 | } | |
2272 | ||
2273 | done : { | |
2274 | /* Update last_update if UPDATEs were written. */ | |
2275 | if (peer->update_out > oc) | |
2276 | peer->last_update = bgp_clock(); | |
2277 | ||
2278 | /* If we TXed any flavor of packet update last_write */ | |
2279 | if (update_last_write) | |
2280 | peer->last_write = bgp_clock(); | |
2281 | } | |
2282 | ||
2283 | return count; | |
2284 | } | |
2285 | ||
2286 | static void cleanup_handler(void *arg) | |
2287 | { | |
2288 | if (plist) | |
2289 | list_delete(plist); | |
2290 | ||
2291 | plist = NULL; | |
2292 | ||
2293 | pthread_mutex_unlock(&plist_mtx); | |
2294 | } | |
2295 | ||
2296 | /** | |
2297 | * Entry function for peer packet flushing pthread. | |
2298 | * | |
2299 | * The plist must be initialized before calling this. | |
2300 | */ | |
2301 | void *peer_writes_start(void *arg) | |
2302 | { | |
2303 | struct timeval currtime = {0, 0}; | |
2304 | struct timeval sleeptime = {0, 500}; | |
2305 | struct timespec next_update = {0, 0}; | |
2306 | ||
2307 | // initialize | |
2308 | pthread_mutex_lock(&plist_mtx); | |
2309 | plist = list_new(); | |
2310 | ||
2311 | struct listnode *ln; | |
2312 | struct peer *peer; | |
2313 | ||
2314 | pthread_cleanup_push(&cleanup_handler, NULL); | |
2315 | ||
2316 | bgp_packet_writes_thread_run = true; | |
2317 | ||
2318 | while (bgp_packet_writes_thread_run) { // wait around until next update | |
2319 | // time | |
2320 | if (plist->count > 0) | |
2321 | pthread_cond_timedwait(&write_cond, &plist_mtx, | |
2322 | &next_update); | |
2323 | else // wait around until we have some peers | |
2324 | while (plist->count == 0 | |
2325 | && bgp_packet_writes_thread_run) | |
2326 | pthread_cond_wait(&write_cond, &plist_mtx); | |
2327 | ||
2328 | for (ALL_LIST_ELEMENTS_RO(plist, ln, peer)) { | |
2329 | pthread_mutex_lock(&peer->obuf_mtx); | |
2330 | { | |
2331 | bgp_write(peer); | |
2332 | } | |
2333 | pthread_mutex_unlock(&peer->obuf_mtx); | |
2334 | } | |
2335 | ||
2336 | gettimeofday(&currtime, NULL); | |
2337 | timeradd(&currtime, &sleeptime, &currtime); | |
2338 | TIMEVAL_TO_TIMESPEC(&currtime, &next_update); | |
2339 | } | |
2340 | ||
2341 | // clean up | |
2342 | pthread_cleanup_pop(1); | |
2343 | ||
2344 | return NULL; | |
2345 | } | |
2346 | ||
2347 | /** | |
2348 | * Turns on packet writing for a peer. | |
2349 | */ | |
2350 | void peer_writes_on(struct peer *peer) | |
2351 | { | |
2352 | if (peer->status == Deleted) | |
2353 | return; | |
2354 | ||
2355 | pthread_mutex_lock(&plist_mtx); | |
2356 | { | |
2357 | struct listnode *ln, *nn; | |
2358 | struct peer *p; | |
2359 | ||
2360 | // make sure this peer isn't already in the list | |
2361 | for (ALL_LIST_ELEMENTS(plist, ln, nn, p)) | |
2362 | if (p == peer) { | |
2363 | pthread_mutex_unlock(&plist_mtx); | |
2364 | return; | |
2365 | } | |
2366 | ||
2367 | peer_lock(peer); | |
2368 | listnode_add(plist, peer); | |
2369 | } | |
2370 | pthread_mutex_unlock(&plist_mtx); | |
2371 | peer_writes_wake(); | |
2372 | } | |
2373 | ||
2374 | /** | |
2375 | * Turns off packet writing for a peer. | |
2376 | */ | |
2377 | void peer_writes_off(struct peer *peer) | |
2378 | { | |
2379 | struct listnode *ln, *nn; | |
2380 | struct peer *p; | |
2381 | pthread_mutex_lock(&plist_mtx); | |
2382 | { | |
2383 | for (ALL_LIST_ELEMENTS(plist, ln, nn, p)) | |
2384 | if (p == peer) { | |
2385 | list_delete_node(plist, ln); | |
2386 | peer_unlock(peer); | |
2387 | break; | |
2388 | } | |
2389 | } | |
2390 | pthread_mutex_unlock(&plist_mtx); | |
2391 | } | |
2392 | ||
2393 | /** | |
2394 | * Wakes up the write thread to do work. | |
2395 | */ | |
2396 | void peer_writes_wake() | |
2397 | { | |
2398 | pthread_cond_signal(&write_cond); | |
2399 | } |