]>
Commit | Line | Data |
---|---|---|
718e3744 | 1 | /* BGP packet management routine. |
56257a44 QY |
2 | * Contains utility functions for constructing and consuming BGP messages. |
3 | * Copyright (C) 2017 Cumulus Networks | |
896014f4 DL |
4 | * Copyright (C) 1999 Kunihiro Ishiguro |
5 | * | |
6 | * This file is part of GNU Zebra. | |
7 | * | |
8 | * GNU Zebra is free software; you can redistribute it and/or modify it | |
9 | * under the terms of the GNU General Public License as published by the | |
10 | * Free Software Foundation; either version 2, or (at your option) any | |
11 | * later version. | |
12 | * | |
13 | * GNU Zebra is distributed in the hope that it will be useful, but | |
14 | * WITHOUT ANY WARRANTY; without even the implied warranty of | |
15 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU | |
16 | * General Public License for more details. | |
17 | * | |
18 | * You should have received a copy of the GNU General Public License along | |
19 | * with this program; see the file COPYING; if not, write to the Free Software | |
20 | * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA | |
21 | */ | |
718e3744 | 22 | |
23 | #include <zebra.h> | |
d3ecc69e | 24 | #include <sys/time.h> |
718e3744 | 25 | |
26 | #include "thread.h" | |
27 | #include "stream.h" | |
28 | #include "network.h" | |
29 | #include "prefix.h" | |
30 | #include "command.h" | |
31 | #include "log.h" | |
32 | #include "memory.h" | |
d62a17ae | 33 | #include "sockunion.h" /* for inet_ntop () */ |
baa376fc | 34 | #include "sockopt.h" |
718e3744 | 35 | #include "linklist.h" |
36 | #include "plist.h" | |
3f9c7369 | 37 | #include "queue.h" |
039f3a34 | 38 | #include "filter.h" |
718e3744 | 39 | |
40 | #include "bgpd/bgpd.h" | |
41 | #include "bgpd/bgp_table.h" | |
42 | #include "bgpd/bgp_dump.h" | |
43 | #include "bgpd/bgp_attr.h" | |
44 | #include "bgpd/bgp_debug.h" | |
45 | #include "bgpd/bgp_fsm.h" | |
46 | #include "bgpd/bgp_route.h" | |
47 | #include "bgpd/bgp_packet.h" | |
48 | #include "bgpd/bgp_open.h" | |
49 | #include "bgpd/bgp_aspath.h" | |
50 | #include "bgpd/bgp_community.h" | |
51 | #include "bgpd/bgp_ecommunity.h" | |
57d187bc | 52 | #include "bgpd/bgp_lcommunity.h" |
718e3744 | 53 | #include "bgpd/bgp_network.h" |
54 | #include "bgpd/bgp_mplsvpn.h" | |
7ef5a232 | 55 | #include "bgpd/bgp_evpn.h" |
718e3744 | 56 | #include "bgpd/bgp_advertise.h" |
93406d87 | 57 | #include "bgpd/bgp_vty.h" |
3f9c7369 | 58 | #include "bgpd/bgp_updgrp.h" |
cd1964ff | 59 | #include "bgpd/bgp_label.h" |
56257a44 | 60 | #include "bgpd/bgp_io.h" |
934af458 | 61 | #include "bgpd/bgp_keepalives.h" |
7c40bf39 | 62 | #include "bgpd/bgp_flowspec.h" |
718e3744 | 63 | |
d8151687 QY |
64 | /** |
65 | * Sets marker and type fields for a BGP message. | |
66 | * | |
67 | * @param s the stream containing the packet | |
68 | * @param type the packet type | |
69 | * @return the size of the stream | |
70 | */ | |
d7c0a89a | 71 | int bgp_packet_set_marker(struct stream *s, uint8_t type) |
718e3744 | 72 | { |
d62a17ae | 73 | int i; |
718e3744 | 74 | |
d62a17ae | 75 | /* Fill in marker. */ |
76 | for (i = 0; i < BGP_MARKER_SIZE; i++) | |
77 | stream_putc(s, 0xff); | |
718e3744 | 78 | |
d62a17ae | 79 | /* Dummy total length. This field is should be filled in later on. */ |
80 | stream_putw(s, 0); | |
718e3744 | 81 | |
d62a17ae | 82 | /* BGP packet type. */ |
83 | stream_putc(s, type); | |
718e3744 | 84 | |
d62a17ae | 85 | /* Return current stream size. */ |
86 | return stream_get_endp(s); | |
718e3744 | 87 | } |
88 | ||
d8151687 QY |
89 | /** |
90 | * Sets size field for a BGP message. | |
91 | * | |
92 | * Size field is set to the size of the stream passed. | |
93 | * | |
94 | * @param s the stream containing the packet | |
95 | * @return the size of the stream | |
96 | */ | |
d62a17ae | 97 | int bgp_packet_set_size(struct stream *s) |
718e3744 | 98 | { |
d62a17ae | 99 | int cp; |
718e3744 | 100 | |
d62a17ae | 101 | /* Preserve current pointer. */ |
102 | cp = stream_get_endp(s); | |
103 | stream_putw_at(s, BGP_MARKER_SIZE, cp); | |
718e3744 | 104 | |
d62a17ae | 105 | return cp; |
718e3744 | 106 | } |
107 | ||
d3ecc69e QY |
108 | /* |
109 | * Push a packet onto the beginning of the peer's output queue. | |
110 | * This function acquires the peer's write mutex before proceeding. | |
111 | */ | |
112 | static void bgp_packet_add(struct peer *peer, struct stream *s) | |
113 | { | |
424ab01d | 114 | pthread_mutex_lock(&peer->io_mtx); |
56257a44 | 115 | stream_fifo_push(peer->obuf, s); |
424ab01d | 116 | pthread_mutex_unlock(&peer->io_mtx); |
718e3744 | 117 | } |
118 | ||
d62a17ae | 119 | static struct stream *bgp_update_packet_eor(struct peer *peer, afi_t afi, |
120 | safi_t safi) | |
93406d87 | 121 | { |
d62a17ae | 122 | struct stream *s; |
123 | iana_afi_t pkt_afi; | |
5c525538 | 124 | iana_safi_t pkt_safi; |
d62a17ae | 125 | |
126 | if (DISABLE_BGP_ANNOUNCE) | |
127 | return NULL; | |
128 | ||
129 | if (bgp_debug_neighbor_events(peer)) | |
130 | zlog_debug("send End-of-RIB for %s to %s", | |
131 | afi_safi_print(afi, safi), peer->host); | |
132 | ||
133 | s = stream_new(BGP_MAX_PACKET_SIZE); | |
134 | ||
135 | /* Make BGP update packet. */ | |
136 | bgp_packet_set_marker(s, BGP_MSG_UPDATE); | |
137 | ||
138 | /* Unfeasible Routes Length */ | |
139 | stream_putw(s, 0); | |
140 | ||
141 | if (afi == AFI_IP && safi == SAFI_UNICAST) { | |
142 | /* Total Path Attribute Length */ | |
143 | stream_putw(s, 0); | |
144 | } else { | |
145 | /* Convert AFI, SAFI to values for packet. */ | |
146 | bgp_map_afi_safi_int2iana(afi, safi, &pkt_afi, &pkt_safi); | |
147 | ||
148 | /* Total Path Attribute Length */ | |
149 | stream_putw(s, 6); | |
150 | stream_putc(s, BGP_ATTR_FLAG_OPTIONAL); | |
151 | stream_putc(s, BGP_ATTR_MP_UNREACH_NLRI); | |
152 | stream_putc(s, 3); | |
153 | stream_putw(s, pkt_afi); | |
154 | stream_putc(s, pkt_safi); | |
155 | } | |
156 | ||
157 | bgp_packet_set_size(s); | |
d62a17ae | 158 | return s; |
718e3744 | 159 | } |
160 | ||
d8151687 QY |
161 | /* Called when there is a change in the EOR(implicit or explicit) status of a |
162 | * peer. Ends the update-delay if all expected peers are done with EORs. */ | |
163 | void bgp_check_update_delay(struct bgp *bgp) | |
164 | { | |
165 | struct listnode *node, *nnode; | |
166 | struct peer *peer = NULL; | |
167 | ||
168 | if (bgp_debug_neighbor_events(peer)) | |
169 | zlog_debug("Checking update delay, T: %d R: %d I:%d E: %d", | |
170 | bgp->established, bgp->restarted_peers, | |
171 | bgp->implicit_eors, bgp->explicit_eors); | |
172 | ||
173 | if (bgp->established | |
174 | <= bgp->restarted_peers + bgp->implicit_eors + bgp->explicit_eors) { | |
becedef6 QY |
175 | /* |
176 | * This is an extra sanity check to make sure we wait for all | |
177 | * the eligible configured peers. This check is performed if | |
178 | * establish wait timer is on, or establish wait option is not | |
179 | * given with the update-delay command | |
180 | */ | |
d8151687 QY |
181 | if (bgp->t_establish_wait |
182 | || (bgp->v_establish_wait == bgp->v_update_delay)) | |
183 | for (ALL_LIST_ELEMENTS(bgp->peer, node, nnode, peer)) { | |
184 | if (CHECK_FLAG(peer->flags, | |
185 | PEER_FLAG_CONFIG_NODE) | |
186 | && !CHECK_FLAG(peer->flags, | |
187 | PEER_FLAG_SHUTDOWN) | |
188 | && !peer->update_delay_over) { | |
189 | if (bgp_debug_neighbor_events(peer)) | |
190 | zlog_debug( | |
191 | " Peer %s pending, continuing read-only mode", | |
192 | peer->host); | |
193 | return; | |
194 | } | |
195 | } | |
196 | ||
197 | zlog_info( | |
198 | "Update delay ended, restarted: %d, EORs implicit: %d, explicit: %d", | |
199 | bgp->restarted_peers, bgp->implicit_eors, | |
200 | bgp->explicit_eors); | |
201 | bgp_update_delay_end(bgp); | |
202 | } | |
203 | } | |
204 | ||
becedef6 QY |
205 | /* |
206 | * Called if peer is known to have restarted. The restart-state bit in | |
207 | * Graceful-Restart capability is used for that | |
208 | */ | |
d8151687 QY |
209 | void bgp_update_restarted_peers(struct peer *peer) |
210 | { | |
211 | if (!bgp_update_delay_active(peer->bgp)) | |
212 | return; /* BGP update delay has ended */ | |
213 | if (peer->update_delay_over) | |
214 | return; /* This peer has already been considered */ | |
215 | ||
216 | if (bgp_debug_neighbor_events(peer)) | |
217 | zlog_debug("Peer %s: Checking restarted", peer->host); | |
218 | ||
219 | if (peer->status == Established) { | |
220 | peer->update_delay_over = 1; | |
221 | peer->bgp->restarted_peers++; | |
222 | bgp_check_update_delay(peer->bgp); | |
223 | } | |
224 | } | |
225 | ||
becedef6 QY |
226 | /* |
227 | * Called as peer receives a keep-alive. Determines if this occurence can be | |
228 | * taken as an implicit EOR for this peer. | |
229 | * NOTE: The very first keep-alive after the Established state of a peer is | |
230 | * considered implicit EOR for the update-delay purposes | |
231 | */ | |
d8151687 QY |
232 | void bgp_update_implicit_eors(struct peer *peer) |
233 | { | |
234 | if (!bgp_update_delay_active(peer->bgp)) | |
235 | return; /* BGP update delay has ended */ | |
236 | if (peer->update_delay_over) | |
237 | return; /* This peer has already been considered */ | |
238 | ||
239 | if (bgp_debug_neighbor_events(peer)) | |
240 | zlog_debug("Peer %s: Checking implicit EORs", peer->host); | |
241 | ||
242 | if (peer->status == Established) { | |
243 | peer->update_delay_over = 1; | |
244 | peer->bgp->implicit_eors++; | |
245 | bgp_check_update_delay(peer->bgp); | |
246 | } | |
247 | } | |
248 | ||
becedef6 QY |
249 | /* |
250 | * Should be called only when there is a change in the EOR_RECEIVED status | |
251 | * for any afi/safi on a peer. | |
252 | */ | |
d8151687 QY |
253 | static void bgp_update_explicit_eors(struct peer *peer) |
254 | { | |
255 | afi_t afi; | |
256 | safi_t safi; | |
257 | ||
258 | if (!bgp_update_delay_active(peer->bgp)) | |
259 | return; /* BGP update delay has ended */ | |
260 | if (peer->update_delay_over) | |
261 | return; /* This peer has already been considered */ | |
262 | ||
263 | if (bgp_debug_neighbor_events(peer)) | |
264 | zlog_debug("Peer %s: Checking explicit EORs", peer->host); | |
265 | ||
266 | for (afi = AFI_IP; afi < AFI_MAX; afi++) | |
267 | for (safi = SAFI_UNICAST; safi < SAFI_MAX; safi++) { | |
268 | if (peer->afc_nego[afi][safi] | |
269 | && !CHECK_FLAG(peer->af_sflags[afi][safi], | |
270 | PEER_STATUS_EOR_RECEIVED)) { | |
271 | if (bgp_debug_neighbor_events(peer)) | |
272 | zlog_debug( | |
273 | " afi %d safi %d didnt receive EOR", | |
274 | afi, safi); | |
275 | return; | |
276 | } | |
277 | } | |
278 | ||
279 | peer->update_delay_over = 1; | |
280 | peer->bgp->explicit_eors++; | |
281 | bgp_check_update_delay(peer->bgp); | |
282 | } | |
283 | ||
284 | /** | |
285 | * Frontend for NLRI parsing, to fan-out to AFI/SAFI specific parsers. | |
286 | * | |
287 | * mp_withdraw, if set, is used to nullify attr structure on most of the | |
288 | * calling safi function and for evpn, passed as parameter | |
289 | */ | |
290 | int bgp_nlri_parse(struct peer *peer, struct attr *attr, | |
291 | struct bgp_nlri *packet, int mp_withdraw) | |
292 | { | |
293 | switch (packet->safi) { | |
294 | case SAFI_UNICAST: | |
295 | case SAFI_MULTICAST: | |
296 | return bgp_nlri_parse_ip(peer, mp_withdraw ? NULL : attr, | |
297 | packet); | |
298 | case SAFI_LABELED_UNICAST: | |
299 | return bgp_nlri_parse_label(peer, mp_withdraw ? NULL : attr, | |
300 | packet); | |
301 | case SAFI_MPLS_VPN: | |
302 | return bgp_nlri_parse_vpn(peer, mp_withdraw ? NULL : attr, | |
303 | packet); | |
304 | case SAFI_EVPN: | |
305 | return bgp_nlri_parse_evpn(peer, attr, packet, mp_withdraw); | |
7c40bf39 | 306 | case SAFI_FLOWSPEC: |
307 | return bgp_nlri_parse_flowspec(peer, attr, packet, mp_withdraw); | |
d8151687 QY |
308 | } |
309 | return -1; | |
310 | } | |
311 | ||
6ec98a2f QY |
312 | /* |
313 | * Checks a variety of conditions to determine whether the peer needs to be | |
314 | * rescheduled for packet generation again, and does so if necessary. | |
315 | * | |
316 | * @param peer to check for rescheduling | |
317 | */ | |
af1e1dc6 QY |
318 | static void bgp_write_proceed_actions(struct peer *peer) |
319 | { | |
320 | afi_t afi; | |
321 | safi_t safi; | |
322 | struct peer_af *paf; | |
323 | struct bpacket *next_pkt; | |
324 | struct update_subgroup *subgrp; | |
325 | ||
c58b0f46 QY |
326 | FOREACH_AFI_SAFI (afi, safi) { |
327 | paf = peer_af_find(peer, afi, safi); | |
328 | if (!paf) | |
329 | continue; | |
330 | subgrp = paf->subgroup; | |
331 | if (!subgrp) | |
332 | continue; | |
333 | ||
334 | next_pkt = paf->next_pkt_to_send; | |
335 | if (next_pkt && next_pkt->buffer) { | |
336 | BGP_TIMER_ON(peer->t_generate_updgrp_packets, | |
337 | bgp_generate_updgrp_packets, 0); | |
338 | return; | |
339 | } | |
af1e1dc6 | 340 | |
c58b0f46 QY |
341 | /* No packets readily available for AFI/SAFI, are there |
342 | * subgroup packets | |
343 | * that need to be generated? */ | |
344 | if (bpacket_queue_is_full(SUBGRP_INST(subgrp), | |
345 | SUBGRP_PKTQ(subgrp)) | |
346 | || subgroup_packets_to_build(subgrp)) { | |
347 | BGP_TIMER_ON(peer->t_generate_updgrp_packets, | |
348 | bgp_generate_updgrp_packets, 0); | |
349 | return; | |
350 | } | |
af1e1dc6 | 351 | |
c58b0f46 QY |
352 | /* No packets to send, see if EOR is pending */ |
353 | if (CHECK_FLAG(peer->cap, PEER_CAP_RESTART_RCV)) { | |
354 | if (!subgrp->t_coalesce && peer->afc_nego[afi][safi] | |
355 | && peer->synctime | |
356 | && !CHECK_FLAG(peer->af_sflags[afi][safi], | |
357 | PEER_STATUS_EOR_SEND) | |
358 | && safi != SAFI_MPLS_VPN) { | |
af1e1dc6 QY |
359 | BGP_TIMER_ON(peer->t_generate_updgrp_packets, |
360 | bgp_generate_updgrp_packets, 0); | |
361 | return; | |
362 | } | |
af1e1dc6 | 363 | } |
c58b0f46 | 364 | } |
af1e1dc6 QY |
365 | } |
366 | ||
6ec98a2f QY |
367 | /* |
368 | * Generate advertisement information (withdraws, updates, EOR) from each | |
369 | * update group a peer belongs to, encode this information into packets, and | |
370 | * enqueue the packets onto the peer's output buffer. | |
56257a44 QY |
371 | */ |
372 | int bgp_generate_updgrp_packets(struct thread *thread) | |
718e3744 | 373 | { |
56257a44 | 374 | struct peer *peer = THREAD_ARG(thread); |
56257a44 QY |
375 | |
376 | struct stream *s; | |
d62a17ae | 377 | struct peer_af *paf; |
378 | struct bpacket *next_pkt; | |
d0ad6d8e QY |
379 | uint32_t wpq; |
380 | uint32_t generated = 0; | |
d62a17ae | 381 | afi_t afi; |
382 | safi_t safi; | |
383 | ||
d0ad6d8e QY |
384 | wpq = atomic_load_explicit(&peer->bgp->wpkt_quanta, |
385 | memory_order_relaxed); | |
386 | ||
d62a17ae | 387 | /* |
388 | * The code beyond this part deals with update packets, proceed only | |
389 | * if peer is Established and updates are not on hold (as part of | |
390 | * update-delay post processing). | |
3f9c7369 | 391 | */ |
d62a17ae | 392 | if (peer->status != Established) |
56257a44 | 393 | return 0; |
d62a17ae | 394 | |
5410015a | 395 | if (peer->bgp->main_peers_update_hold) |
56257a44 | 396 | return 0; |
d62a17ae | 397 | |
56257a44 QY |
398 | do { |
399 | s = NULL; | |
c58b0f46 QY |
400 | FOREACH_AFI_SAFI (afi, safi) { |
401 | paf = peer_af_find(peer, afi, safi); | |
402 | if (!paf || !PAF_SUBGRP(paf)) | |
403 | continue; | |
404 | next_pkt = paf->next_pkt_to_send; | |
80bd61c4 | 405 | |
c58b0f46 QY |
406 | /* |
407 | * Try to generate a packet for the peer if we are at | |
408 | * the end of the list. Always try to push out | |
409 | * WITHDRAWs first. | |
410 | */ | |
411 | if (!next_pkt || !next_pkt->buffer) { | |
412 | next_pkt = subgroup_withdraw_packet( | |
413 | PAF_SUBGRP(paf)); | |
414 | if (!next_pkt || !next_pkt->buffer) | |
415 | subgroup_update_packet(PAF_SUBGRP(paf)); | |
416 | next_pkt = paf->next_pkt_to_send; | |
417 | } | |
80bd61c4 | 418 | |
c58b0f46 QY |
419 | /* |
420 | * If we still don't have a packet to send to the peer, | |
421 | * then try to find out out if we have to send eor or | |
422 | * if not, skip to the next AFI, SAFI. Don't send the | |
423 | * EOR prematurely; if the subgroup's coalesce timer is | |
424 | * running, the adjacency-out structure is not created | |
425 | * yet. | |
426 | */ | |
427 | if (!next_pkt || !next_pkt->buffer) { | |
428 | if (CHECK_FLAG(peer->cap, | |
429 | PEER_CAP_RESTART_RCV)) { | |
430 | if (!(PAF_SUBGRP(paf))->t_coalesce | |
431 | && peer->afc_nego[afi][safi] | |
432 | && peer->synctime | |
433 | && !CHECK_FLAG( | |
434 | peer->af_sflags[afi] | |
435 | [safi], | |
436 | PEER_STATUS_EOR_SEND)) { | |
437 | SET_FLAG(peer->af_sflags[afi] | |
56257a44 | 438 | [safi], |
c58b0f46 QY |
439 | PEER_STATUS_EOR_SEND); |
440 | ||
441 | if ((s = bgp_update_packet_eor( | |
442 | peer, afi, | |
443 | safi))) { | |
444 | bgp_packet_add(peer, s); | |
56257a44 | 445 | } |
80bd61c4 | 446 | } |
d62a17ae | 447 | } |
c58b0f46 QY |
448 | continue; |
449 | } | |
d62a17ae | 450 | |
718e3744 | 451 | |
c58b0f46 QY |
452 | /* Found a packet template to send, overwrite |
453 | * packet with appropriate attributes from peer | |
454 | * and advance peer */ | |
455 | s = bpacket_reformat_for_peer(next_pkt, paf); | |
456 | bgp_packet_add(peer, s); | |
c58b0f46 QY |
457 | bpacket_queue_advance_peer(paf); |
458 | } | |
d0ad6d8e | 459 | } while (s && (++generated < wpq)); |
80bd61c4 | 460 | |
6ec98a2f QY |
461 | if (generated) |
462 | bgp_writes_on(peer); | |
463 | ||
af1e1dc6 QY |
464 | bgp_write_proceed_actions(peer); |
465 | ||
80bd61c4 QY |
466 | return 0; |
467 | } | |
468 | ||
d3ecc69e QY |
469 | /* |
470 | * Creates a BGP Keepalive packet and appends it to the peer's output queue. | |
471 | */ | |
d62a17ae | 472 | void bgp_keepalive_send(struct peer *peer) |
718e3744 | 473 | { |
d62a17ae | 474 | struct stream *s; |
475 | ||
476 | s = stream_new(BGP_MAX_PACKET_SIZE); | |
718e3744 | 477 | |
d62a17ae | 478 | /* Make keepalive packet. */ |
479 | bgp_packet_set_marker(s, BGP_MSG_KEEPALIVE); | |
718e3744 | 480 | |
d62a17ae | 481 | /* Set packet size. */ |
482 | (void)bgp_packet_set_size(s); | |
718e3744 | 483 | |
d62a17ae | 484 | /* Dump packet if debug option is set. */ |
485 | /* bgp_packet_dump (s); */ | |
718e3744 | 486 | |
d62a17ae | 487 | if (bgp_debug_keepalive(peer)) |
488 | zlog_debug("%s sending KEEPALIVE", peer->host); | |
718e3744 | 489 | |
d62a17ae | 490 | /* Add packet to the peer. */ |
491 | bgp_packet_add(peer, s); | |
424ab01d QY |
492 | |
493 | bgp_writes_on(peer); | |
718e3744 | 494 | } |
495 | ||
d3ecc69e QY |
496 | /* |
497 | * Creates a BGP Open packet and appends it to the peer's output queue. | |
498 | * Sets capabilities as necessary. | |
499 | */ | |
d62a17ae | 500 | void bgp_open_send(struct peer *peer) |
718e3744 | 501 | { |
d62a17ae | 502 | struct stream *s; |
d7c0a89a | 503 | uint16_t send_holdtime; |
d62a17ae | 504 | as_t local_as; |
718e3744 | 505 | |
d25e4efc | 506 | if (PEER_OR_GROUP_TIMER_SET(peer)) |
d62a17ae | 507 | send_holdtime = peer->holdtime; |
508 | else | |
509 | send_holdtime = peer->bgp->default_holdtime; | |
718e3744 | 510 | |
d62a17ae | 511 | /* local-as Change */ |
512 | if (peer->change_local_as) | |
513 | local_as = peer->change_local_as; | |
514 | else | |
515 | local_as = peer->local_as; | |
718e3744 | 516 | |
d62a17ae | 517 | s = stream_new(BGP_MAX_PACKET_SIZE); |
718e3744 | 518 | |
d62a17ae | 519 | /* Make open packet. */ |
520 | bgp_packet_set_marker(s, BGP_MSG_OPEN); | |
718e3744 | 521 | |
d62a17ae | 522 | /* Set open packet values. */ |
523 | stream_putc(s, BGP_VERSION_4); /* BGP version */ | |
d7c0a89a QY |
524 | stream_putw(s, (local_as <= BGP_AS_MAX) ? (uint16_t)local_as |
525 | : BGP_AS_TRANS); | |
d62a17ae | 526 | stream_putw(s, send_holdtime); /* Hold Time */ |
527 | stream_put_in_addr(s, &peer->local_id); /* BGP Identifier */ | |
718e3744 | 528 | |
d62a17ae | 529 | /* Set capability code. */ |
530 | bgp_open_capability(s, peer); | |
718e3744 | 531 | |
d62a17ae | 532 | /* Set BGP packet length. */ |
533 | (void)bgp_packet_set_size(s); | |
718e3744 | 534 | |
d62a17ae | 535 | if (bgp_debug_neighbor_events(peer)) |
536 | zlog_debug( | |
537 | "%s sending OPEN, version %d, my as %u, holdtime %d, id %s", | |
538 | peer->host, BGP_VERSION_4, local_as, send_holdtime, | |
539 | inet_ntoa(peer->local_id)); | |
718e3744 | 540 | |
d62a17ae | 541 | /* Dump packet if debug option is set. */ |
542 | /* bgp_packet_dump (s); */ | |
718e3744 | 543 | |
d62a17ae | 544 | /* Add packet to the peer. */ |
545 | bgp_packet_add(peer, s); | |
424ab01d QY |
546 | |
547 | bgp_writes_on(peer); | |
548 | } | |
549 | ||
a127f33b QY |
550 | /* |
551 | * Writes NOTIFICATION message directly to a peer socket without waiting for | |
552 | * the I/O thread. | |
553 | * | |
554 | * There must be exactly one stream on the peer->obuf FIFO, and the data within | |
555 | * this stream must match the format of a BGP NOTIFICATION message. | |
556 | * Transmission is best-effort. | |
557 | * | |
558 | * @requires peer->io_mtx | |
559 | * @param peer | |
560 | * @return 0 | |
561 | */ | |
424ab01d QY |
562 | static int bgp_write_notify(struct peer *peer) |
563 | { | |
564 | int ret, val; | |
d7c0a89a | 565 | uint8_t type; |
424ab01d QY |
566 | struct stream *s; |
567 | ||
a127f33b QY |
568 | /* There should be at least one packet. */ |
569 | s = stream_fifo_pop(peer->obuf); | |
424ab01d | 570 | |
8ec586b0 QY |
571 | if (!s) |
572 | return 0; | |
573 | ||
574 | assert(stream_get_endp(s) >= BGP_HEADER_SIZE); | |
575 | ||
424ab01d QY |
576 | /* Stop collecting data within the socket */ |
577 | sockopt_cork(peer->fd, 0); | |
578 | ||
becedef6 QY |
579 | /* |
580 | * socket is in nonblocking mode, if we can't deliver the NOTIFY, well, | |
581 | * we only care about getting a clean shutdown at this point. | |
582 | */ | |
424ab01d QY |
583 | ret = write(peer->fd, STREAM_DATA(s), stream_get_endp(s)); |
584 | ||
becedef6 QY |
585 | /* |
586 | * only connection reset/close gets counted as TCP_fatal_error, failure | |
587 | * to write the entire NOTIFY doesn't get different FSM treatment | |
588 | */ | |
424ab01d | 589 | if (ret <= 0) { |
3735936b | 590 | stream_free(s); |
424ab01d QY |
591 | BGP_EVENT_ADD(peer, TCP_fatal_error); |
592 | return 0; | |
593 | } | |
594 | ||
595 | /* Disable Nagle, make NOTIFY packet go out right away */ | |
596 | val = 1; | |
597 | (void)setsockopt(peer->fd, IPPROTO_TCP, TCP_NODELAY, (char *)&val, | |
598 | sizeof(val)); | |
599 | ||
600 | /* Retrieve BGP packet type. */ | |
601 | stream_set_getp(s, BGP_MARKER_SIZE + 2); | |
602 | type = stream_getc(s); | |
603 | ||
604 | assert(type == BGP_MSG_NOTIFY); | |
605 | ||
606 | /* Type should be notify. */ | |
0112e9e0 | 607 | atomic_fetch_add_explicit(&peer->notify_out, 1, memory_order_relaxed); |
424ab01d QY |
608 | peer->notify_out++; |
609 | ||
610 | /* Double start timer. */ | |
611 | peer->v_start *= 2; | |
612 | ||
613 | /* Overflow check. */ | |
614 | if (peer->v_start >= (60 * 2)) | |
615 | peer->v_start = (60 * 2); | |
616 | ||
becedef6 QY |
617 | /* |
618 | * Handle Graceful Restart case where the state changes to | |
619 | * Connect instead of Idle | |
620 | */ | |
424ab01d QY |
621 | BGP_EVENT_ADD(peer, BGP_Stop); |
622 | ||
3735936b QY |
623 | stream_free(s); |
624 | ||
424ab01d | 625 | return 0; |
718e3744 | 626 | } |
627 | ||
d3ecc69e QY |
628 | /* |
629 | * Creates a BGP Notify and appends it to the peer's output queue. | |
630 | * | |
becedef6 QY |
631 | * This function attempts to write the packet from the thread it is called |
632 | * from, to ensure the packet gets out ASAP. | |
d3ecc69e | 633 | * |
a127f33b QY |
634 | * This function may be called from multiple threads. Since the function |
635 | * modifies I/O buffer(s) in the peer, these are locked for the duration of the | |
636 | * call to prevent tampering from other threads. | |
637 | * | |
638 | * Delivery of the NOTIFICATION is attempted once and is best-effort. After | |
639 | * return, the peer structure *must* be reset; no assumptions about session | |
640 | * state are valid. | |
641 | * | |
d3ecc69e QY |
642 | * @param peer |
643 | * @param code BGP error code | |
644 | * @param sub_code BGP error subcode | |
645 | * @param data Data portion | |
646 | * @param datalen length of data portion | |
647 | */ | |
d7c0a89a QY |
648 | void bgp_notify_send_with_data(struct peer *peer, uint8_t code, |
649 | uint8_t sub_code, uint8_t *data, size_t datalen) | |
718e3744 | 650 | { |
d62a17ae | 651 | struct stream *s; |
d62a17ae | 652 | |
a127f33b QY |
653 | /* Lock I/O mutex to prevent other threads from pushing packets */ |
654 | pthread_mutex_lock(&peer->io_mtx); | |
655 | /* ============================================== */ | |
656 | ||
d62a17ae | 657 | /* Allocate new stream. */ |
658 | s = stream_new(BGP_MAX_PACKET_SIZE); | |
659 | ||
d3ecc69e | 660 | /* Make notify packet. */ |
d62a17ae | 661 | bgp_packet_set_marker(s, BGP_MSG_NOTIFY); |
662 | ||
663 | /* Set notify packet values. */ | |
664 | stream_putc(s, code); /* BGP notify code */ | |
665 | stream_putc(s, sub_code); /* BGP notify sub_code */ | |
666 | ||
667 | /* If notify data is present. */ | |
668 | if (data) | |
669 | stream_write(s, data, datalen); | |
670 | ||
671 | /* Set BGP packet length. */ | |
bd6b2706 | 672 | bgp_packet_set_size(s); |
d62a17ae | 673 | |
424ab01d | 674 | /* wipe output buffer */ |
a127f33b | 675 | stream_fifo_clean(peer->obuf); |
d62a17ae | 676 | |
becedef6 QY |
677 | /* |
678 | * If possible, store last packet for debugging purposes. This check is | |
679 | * in place because we are sometimes called with a doppelganger peer, | |
680 | * who tends to have a plethora of fields nulled out. | |
681 | */ | |
d8151687 QY |
682 | if (peer->curr && peer->last_reset_cause_size) { |
683 | size_t packetsize = stream_get_endp(peer->curr); | |
684 | assert(packetsize <= peer->last_reset_cause_size); | |
685 | memcpy(peer->last_reset_cause, peer->curr->data, packetsize); | |
686 | peer->last_reset_cause_size = packetsize; | |
687 | } | |
688 | ||
d62a17ae | 689 | /* For debug */ |
690 | { | |
691 | struct bgp_notify bgp_notify; | |
692 | int first = 0; | |
693 | int i; | |
694 | char c[4]; | |
695 | ||
696 | bgp_notify.code = code; | |
697 | bgp_notify.subcode = sub_code; | |
698 | bgp_notify.data = NULL; | |
e0981960 | 699 | bgp_notify.length = datalen; |
d62a17ae | 700 | bgp_notify.raw_data = data; |
701 | ||
702 | peer->notify.code = bgp_notify.code; | |
703 | peer->notify.subcode = bgp_notify.subcode; | |
704 | ||
e0981960 | 705 | if (bgp_notify.length && data) { |
d62a17ae | 706 | bgp_notify.data = |
707 | XMALLOC(MTYPE_TMP, bgp_notify.length * 3); | |
708 | for (i = 0; i < bgp_notify.length; i++) | |
709 | if (first) { | |
710 | sprintf(c, " %02x", data[i]); | |
711 | strcat(bgp_notify.data, c); | |
712 | } else { | |
713 | first = 1; | |
714 | sprintf(c, "%02x", data[i]); | |
715 | strcpy(bgp_notify.data, c); | |
716 | } | |
717 | } | |
718 | bgp_notify_print(peer, &bgp_notify, "sending"); | |
719 | ||
720 | if (bgp_notify.data) { | |
721 | XFREE(MTYPE_TMP, bgp_notify.data); | |
722 | bgp_notify.data = NULL; | |
723 | bgp_notify.length = 0; | |
724 | } | |
725 | } | |
726 | ||
727 | /* peer reset cause */ | |
728 | if (code == BGP_NOTIFY_CEASE) { | |
729 | if (sub_code == BGP_NOTIFY_CEASE_ADMIN_RESET) | |
730 | peer->last_reset = PEER_DOWN_USER_RESET; | |
731 | else if (sub_code == BGP_NOTIFY_CEASE_ADMIN_SHUTDOWN) | |
732 | peer->last_reset = PEER_DOWN_USER_SHUTDOWN; | |
733 | else | |
734 | peer->last_reset = PEER_DOWN_NOTIFY_SEND; | |
735 | } else | |
736 | peer->last_reset = PEER_DOWN_NOTIFY_SEND; | |
737 | ||
d3ecc69e | 738 | /* Add packet to peer's output queue */ |
a127f33b | 739 | stream_fifo_push(peer->obuf, s); |
424ab01d QY |
740 | |
741 | bgp_write_notify(peer); | |
a127f33b QY |
742 | |
743 | /* ============================================== */ | |
744 | pthread_mutex_unlock(&peer->io_mtx); | |
718e3744 | 745 | } |
746 | ||
d3ecc69e QY |
747 | /* |
748 | * Creates a BGP Notify and appends it to the peer's output queue. | |
749 | * | |
becedef6 QY |
750 | * This function attempts to write the packet from the thread it is called |
751 | * from, to ensure the packet gets out ASAP. | |
d3ecc69e QY |
752 | * |
753 | * @param peer | |
754 | * @param code BGP error code | |
755 | * @param sub_code BGP error subcode | |
756 | */ | |
d7c0a89a | 757 | void bgp_notify_send(struct peer *peer, uint8_t code, uint8_t sub_code) |
718e3744 | 758 | { |
d62a17ae | 759 | bgp_notify_send_with_data(peer, code, sub_code, NULL, 0); |
718e3744 | 760 | } |
761 | ||
d3ecc69e QY |
762 | /* |
763 | * Creates BGP Route Refresh packet and appends it to the peer's output queue. | |
764 | * | |
765 | * @param peer | |
766 | * @param afi Address Family Identifier | |
767 | * @param safi Subsequent Address Family Identifier | |
768 | * @param orf_type Outbound Route Filtering type | |
769 | * @param when_to_refresh Whether to refresh immediately or defer | |
770 | * @param remove Whether to remove ORF for specified AFI/SAFI | |
771 | */ | |
d62a17ae | 772 | void bgp_route_refresh_send(struct peer *peer, afi_t afi, safi_t safi, |
d7c0a89a QY |
773 | uint8_t orf_type, uint8_t when_to_refresh, |
774 | int remove) | |
718e3744 | 775 | { |
d62a17ae | 776 | struct stream *s; |
777 | struct bgp_filter *filter; | |
778 | int orf_refresh = 0; | |
779 | iana_afi_t pkt_afi; | |
5c525538 | 780 | iana_safi_t pkt_safi; |
d62a17ae | 781 | |
782 | if (DISABLE_BGP_ANNOUNCE) | |
783 | return; | |
784 | ||
785 | filter = &peer->filter[afi][safi]; | |
786 | ||
787 | /* Convert AFI, SAFI to values for packet. */ | |
788 | bgp_map_afi_safi_int2iana(afi, safi, &pkt_afi, &pkt_safi); | |
789 | ||
790 | s = stream_new(BGP_MAX_PACKET_SIZE); | |
791 | ||
792 | /* Make BGP update packet. */ | |
793 | if (CHECK_FLAG(peer->cap, PEER_CAP_REFRESH_NEW_RCV)) | |
794 | bgp_packet_set_marker(s, BGP_MSG_ROUTE_REFRESH_NEW); | |
718e3744 | 795 | else |
d62a17ae | 796 | bgp_packet_set_marker(s, BGP_MSG_ROUTE_REFRESH_OLD); |
797 | ||
798 | /* Encode Route Refresh message. */ | |
799 | stream_putw(s, pkt_afi); | |
800 | stream_putc(s, 0); | |
801 | stream_putc(s, pkt_safi); | |
802 | ||
803 | if (orf_type == ORF_TYPE_PREFIX || orf_type == ORF_TYPE_PREFIX_OLD) | |
804 | if (remove || filter->plist[FILTER_IN].plist) { | |
d7c0a89a | 805 | uint16_t orf_len; |
d62a17ae | 806 | unsigned long orfp; |
807 | ||
808 | orf_refresh = 1; | |
809 | stream_putc(s, when_to_refresh); | |
810 | stream_putc(s, orf_type); | |
811 | orfp = stream_get_endp(s); | |
812 | stream_putw(s, 0); | |
813 | ||
814 | if (remove) { | |
815 | UNSET_FLAG(peer->af_sflags[afi][safi], | |
816 | PEER_STATUS_ORF_PREFIX_SEND); | |
817 | stream_putc(s, ORF_COMMON_PART_REMOVE_ALL); | |
818 | if (bgp_debug_neighbor_events(peer)) | |
819 | zlog_debug( | |
820 | "%s sending REFRESH_REQ to remove ORF(%d) (%s) for afi/safi: %d/%d", | |
821 | peer->host, orf_type, | |
822 | (when_to_refresh == REFRESH_DEFER | |
823 | ? "defer" | |
824 | : "immediate"), | |
825 | pkt_afi, pkt_safi); | |
826 | } else { | |
827 | SET_FLAG(peer->af_sflags[afi][safi], | |
828 | PEER_STATUS_ORF_PREFIX_SEND); | |
829 | prefix_bgp_orf_entry( | |
830 | s, filter->plist[FILTER_IN].plist, | |
831 | ORF_COMMON_PART_ADD, | |
832 | ORF_COMMON_PART_PERMIT, | |
833 | ORF_COMMON_PART_DENY); | |
834 | if (bgp_debug_neighbor_events(peer)) | |
835 | zlog_debug( | |
836 | "%s sending REFRESH_REQ with pfxlist ORF(%d) (%s) for afi/safi: %d/%d", | |
837 | peer->host, orf_type, | |
838 | (when_to_refresh == REFRESH_DEFER | |
839 | ? "defer" | |
840 | : "immediate"), | |
841 | pkt_afi, pkt_safi); | |
842 | } | |
843 | ||
844 | /* Total ORF Entry Len. */ | |
845 | orf_len = stream_get_endp(s) - orfp - 2; | |
846 | stream_putw_at(s, orfp, orf_len); | |
847 | } | |
848 | ||
849 | /* Set packet size. */ | |
850 | (void)bgp_packet_set_size(s); | |
851 | ||
852 | if (bgp_debug_neighbor_events(peer)) { | |
853 | if (!orf_refresh) | |
854 | zlog_debug("%s sending REFRESH_REQ for afi/safi: %d/%d", | |
855 | peer->host, pkt_afi, pkt_safi); | |
856 | } | |
857 | ||
858 | /* Add packet to the peer. */ | |
859 | bgp_packet_add(peer, s); | |
424ab01d QY |
860 | |
861 | bgp_writes_on(peer); | |
718e3744 | 862 | } |
863 | ||
d3ecc69e QY |
864 | /* |
865 | * Create a BGP Capability packet and append it to the peer's output queue. | |
866 | * | |
867 | * @param peer | |
868 | * @param afi Address Family Identifier | |
869 | * @param safi Subsequent Address Family Identifier | |
870 | * @param capability_code BGP Capability Code | |
871 | * @param action Set or Remove capability | |
872 | */ | |
d62a17ae | 873 | void bgp_capability_send(struct peer *peer, afi_t afi, safi_t safi, |
874 | int capability_code, int action) | |
718e3744 | 875 | { |
d62a17ae | 876 | struct stream *s; |
877 | iana_afi_t pkt_afi; | |
5c525538 | 878 | iana_safi_t pkt_safi; |
d62a17ae | 879 | |
880 | /* Convert AFI, SAFI to values for packet. */ | |
881 | bgp_map_afi_safi_int2iana(afi, safi, &pkt_afi, &pkt_safi); | |
882 | ||
883 | s = stream_new(BGP_MAX_PACKET_SIZE); | |
884 | ||
885 | /* Make BGP update packet. */ | |
886 | bgp_packet_set_marker(s, BGP_MSG_CAPABILITY); | |
887 | ||
888 | /* Encode MP_EXT capability. */ | |
889 | if (capability_code == CAPABILITY_CODE_MP) { | |
890 | stream_putc(s, action); | |
891 | stream_putc(s, CAPABILITY_CODE_MP); | |
892 | stream_putc(s, CAPABILITY_CODE_MP_LEN); | |
893 | stream_putw(s, pkt_afi); | |
894 | stream_putc(s, 0); | |
895 | stream_putc(s, pkt_safi); | |
896 | ||
897 | if (bgp_debug_neighbor_events(peer)) | |
898 | zlog_debug( | |
899 | "%s sending CAPABILITY has %s MP_EXT CAP for afi/safi: %d/%d", | |
900 | peer->host, | |
901 | action == CAPABILITY_ACTION_SET ? "Advertising" | |
902 | : "Removing", | |
903 | pkt_afi, pkt_safi); | |
904 | } | |
905 | ||
906 | /* Set packet size. */ | |
907 | (void)bgp_packet_set_size(s); | |
718e3744 | 908 | |
d62a17ae | 909 | /* Add packet to the peer. */ |
910 | bgp_packet_add(peer, s); | |
424ab01d QY |
911 | |
912 | bgp_writes_on(peer); | |
d62a17ae | 913 | } |
718e3744 | 914 | |
d62a17ae | 915 | /* RFC1771 6.8 Connection collision detection. */ |
916 | static int bgp_collision_detect(struct peer *new, struct in_addr remote_id) | |
917 | { | |
918 | struct peer *peer; | |
919 | ||
920 | /* Upon receipt of an OPEN message, the local system must examine | |
921 | all of its connections that are in the OpenConfirm state. A BGP | |
922 | speaker may also examine connections in an OpenSent state if it | |
923 | knows the BGP Identifier of the peer by means outside of the | |
924 | protocol. If among these connections there is a connection to a | |
925 | remote BGP speaker whose BGP Identifier equals the one in the | |
926 | OPEN message, then the local system performs the following | |
927 | collision resolution procedure: */ | |
928 | ||
929 | if ((peer = new->doppelganger) != NULL) { | |
930 | /* Do not accept the new connection in Established or Clearing | |
931 | * states. | |
932 | * Note that a peer GR is handled by closing the existing | |
933 | * connection | |
934 | * upon receipt of new one. | |
935 | */ | |
936 | if (peer->status == Established || peer->status == Clearing) { | |
937 | bgp_notify_send(new, BGP_NOTIFY_CEASE, | |
938 | BGP_NOTIFY_CEASE_COLLISION_RESOLUTION); | |
939 | return (-1); | |
940 | } else if ((peer->status == OpenConfirm) | |
941 | || (peer->status == OpenSent)) { | |
942 | /* 1. The BGP Identifier of the local system is compared | |
943 | to | |
944 | the BGP Identifier of the remote system (as specified | |
945 | in | |
946 | the OPEN message). */ | |
947 | ||
948 | if (ntohl(peer->local_id.s_addr) | |
949 | < ntohl(remote_id.s_addr)) | |
950 | if (!CHECK_FLAG(peer->sflags, | |
951 | PEER_STATUS_ACCEPT_PEER)) { | |
952 | /* 2. If the value of the local BGP | |
953 | Identifier is less | |
954 | than the remote one, the local system | |
955 | closes BGP | |
956 | connection that already exists (the | |
957 | one that is | |
958 | already in the OpenConfirm state), | |
959 | and accepts BGP | |
960 | connection initiated by the remote | |
961 | system. */ | |
962 | bgp_notify_send( | |
963 | peer, BGP_NOTIFY_CEASE, | |
964 | BGP_NOTIFY_CEASE_COLLISION_RESOLUTION); | |
965 | return 1; | |
966 | } else { | |
967 | bgp_notify_send( | |
968 | new, BGP_NOTIFY_CEASE, | |
969 | BGP_NOTIFY_CEASE_COLLISION_RESOLUTION); | |
970 | return -1; | |
971 | } | |
972 | else { | |
973 | /* 3. Otherwise, the local system closes newly | |
974 | created | |
975 | BGP connection (the one associated with the | |
976 | newly | |
977 | received OPEN message), and continues to use | |
978 | the | |
979 | existing one (the one that is already in the | |
980 | OpenConfirm state). */ | |
981 | if (CHECK_FLAG(peer->sflags, | |
982 | PEER_STATUS_ACCEPT_PEER)) { | |
983 | bgp_notify_send( | |
984 | peer, BGP_NOTIFY_CEASE, | |
985 | BGP_NOTIFY_CEASE_COLLISION_RESOLUTION); | |
986 | return 1; | |
987 | } else { | |
988 | bgp_notify_send( | |
989 | new, BGP_NOTIFY_CEASE, | |
990 | BGP_NOTIFY_CEASE_COLLISION_RESOLUTION); | |
991 | return -1; | |
992 | } | |
993 | } | |
994 | } | |
995 | } | |
996 | return 0; | |
997 | } | |
718e3744 | 998 | |
d8151687 QY |
999 | /* Packet processing routines ---------------------------------------------- */ |
1000 | /* | |
1001 | * This is a family of functions designed to be called from | |
1002 | * bgp_process_packet(). These functions all share similar behavior and should | |
1003 | * adhere to the following invariants and restrictions: | |
1004 | * | |
1005 | * Return codes | |
1006 | * ------------ | |
1007 | * The return code of any one of those functions should be one of the FSM event | |
1008 | * codes specified in bgpd.h. If a NOTIFY was sent, this event code MUST be | |
1009 | * BGP_Stop. Otherwise, the code SHOULD correspond to the function's expected | |
1010 | * packet type. For example, bgp_open_receive() should return BGP_Stop upon | |
1011 | * error and Receive_OPEN_message otherwise. | |
1012 | * | |
1013 | * If no action is necessary, the correct return code is BGP_PACKET_NOOP as | |
1014 | * defined below. | |
1015 | * | |
1016 | * Side effects | |
1017 | * ------------ | |
1018 | * - May send NOTIFY messages | |
1019 | * - May not modify peer->status | |
1020 | * - May not call bgp_event_update() | |
1021 | */ | |
1022 | ||
1023 | #define BGP_PACKET_NOOP 0 | |
1024 | ||
1025 | /** | |
1026 | * Process BGP OPEN message for peer. | |
1027 | * | |
1028 | * If any errors are encountered in the OPEN message, immediately sends NOTIFY | |
1029 | * and returns BGP_Stop. | |
1030 | * | |
1031 | * @param peer | |
1032 | * @param size size of the packet | |
1033 | * @return as in summary | |
1034 | */ | |
d62a17ae | 1035 | static int bgp_open_receive(struct peer *peer, bgp_size_t size) |
1036 | { | |
1037 | int ret; | |
d7c0a89a QY |
1038 | uint8_t version; |
1039 | uint8_t optlen; | |
1040 | uint16_t holdtime; | |
1041 | uint16_t send_holdtime; | |
d62a17ae | 1042 | as_t remote_as; |
1043 | as_t as4 = 0; | |
1044 | struct in_addr remote_id; | |
1045 | int mp_capability; | |
d7c0a89a QY |
1046 | uint8_t notify_data_remote_as[2]; |
1047 | uint8_t notify_data_remote_as4[4]; | |
1048 | uint8_t notify_data_remote_id[4]; | |
1049 | uint16_t *holdtime_ptr; | |
d62a17ae | 1050 | |
1051 | /* Parse open packet. */ | |
424ab01d QY |
1052 | version = stream_getc(peer->curr); |
1053 | memcpy(notify_data_remote_as, stream_pnt(peer->curr), 2); | |
1054 | remote_as = stream_getw(peer->curr); | |
d7c0a89a | 1055 | holdtime_ptr = (uint16_t *)stream_pnt(peer->curr); |
424ab01d QY |
1056 | holdtime = stream_getw(peer->curr); |
1057 | memcpy(notify_data_remote_id, stream_pnt(peer->curr), 4); | |
1058 | remote_id.s_addr = stream_get_ipv4(peer->curr); | |
d62a17ae | 1059 | |
1060 | /* Receive OPEN message log */ | |
1061 | if (bgp_debug_neighbor_events(peer)) | |
1062 | zlog_debug( | |
1063 | "%s rcv OPEN, version %d, remote-as (in open) %u," | |
1064 | " holdtime %d, id %s", | |
1065 | peer->host, version, remote_as, holdtime, | |
1066 | inet_ntoa(remote_id)); | |
1067 | ||
1068 | /* BEGIN to read the capability here, but dont do it yet */ | |
1069 | mp_capability = 0; | |
424ab01d | 1070 | optlen = stream_getc(peer->curr); |
d62a17ae | 1071 | |
1072 | if (optlen != 0) { | |
1073 | /* If not enough bytes, it is an error. */ | |
424ab01d | 1074 | if (STREAM_READABLE(peer->curr) < optlen) { |
d62a17ae | 1075 | bgp_notify_send(peer, BGP_NOTIFY_OPEN_ERR, |
1076 | BGP_NOTIFY_OPEN_MALFORMED_ATTR); | |
d8151687 | 1077 | return BGP_Stop; |
d62a17ae | 1078 | } |
718e3744 | 1079 | |
d62a17ae | 1080 | /* We need the as4 capability value *right now* because |
1081 | * if it is there, we have not got the remote_as yet, and | |
1082 | * without | |
1083 | * that we do not know which peer is connecting to us now. | |
1084 | */ | |
1085 | as4 = peek_for_as4_capability(peer, optlen); | |
1086 | memcpy(notify_data_remote_as4, &as4, 4); | |
1087 | } | |
718e3744 | 1088 | |
d62a17ae | 1089 | /* Just in case we have a silly peer who sends AS4 capability set to 0 |
1090 | */ | |
1091 | if (CHECK_FLAG(peer->cap, PEER_CAP_AS4_RCV) && !as4) { | |
1092 | zlog_err("%s bad OPEN, got AS4 capability, but AS4 set to 0", | |
1093 | peer->host); | |
1094 | bgp_notify_send_with_data(peer, BGP_NOTIFY_OPEN_ERR, | |
1095 | BGP_NOTIFY_OPEN_BAD_PEER_AS, | |
1096 | notify_data_remote_as4, 4); | |
d8151687 | 1097 | return BGP_Stop; |
d62a17ae | 1098 | } |
718e3744 | 1099 | |
d62a17ae | 1100 | if (remote_as == BGP_AS_TRANS) { |
1101 | /* Take the AS4 from the capability. We must have received the | |
1102 | * capability now! Otherwise we have a asn16 peer who uses | |
1103 | * BGP_AS_TRANS, for some unknown reason. | |
1104 | */ | |
1105 | if (as4 == BGP_AS_TRANS) { | |
1106 | zlog_err( | |
1107 | "%s [AS4] NEW speaker using AS_TRANS for AS4, not allowed", | |
1108 | peer->host); | |
1109 | bgp_notify_send_with_data(peer, BGP_NOTIFY_OPEN_ERR, | |
1110 | BGP_NOTIFY_OPEN_BAD_PEER_AS, | |
1111 | notify_data_remote_as4, 4); | |
d8151687 | 1112 | return BGP_Stop; |
d62a17ae | 1113 | } |
718e3744 | 1114 | |
d62a17ae | 1115 | if (!as4 && BGP_DEBUG(as4, AS4)) |
1116 | zlog_debug( | |
1117 | "%s [AS4] OPEN remote_as is AS_TRANS, but no AS4." | |
1118 | " Odd, but proceeding.", | |
1119 | peer->host); | |
1120 | else if (as4 < BGP_AS_MAX && BGP_DEBUG(as4, AS4)) | |
1121 | zlog_debug( | |
1122 | "%s [AS4] OPEN remote_as is AS_TRANS, but AS4 (%u) fits " | |
1123 | "in 2-bytes, very odd peer.", | |
1124 | peer->host, as4); | |
1125 | if (as4) | |
1126 | remote_as = as4; | |
1127 | } else { | |
1128 | /* We may have a partner with AS4 who has an asno < BGP_AS_MAX | |
1129 | */ | |
1130 | /* If we have got the capability, peer->as4cap must match | |
1131 | * remote_as */ | |
1132 | if (CHECK_FLAG(peer->cap, PEER_CAP_AS4_RCV) | |
1133 | && as4 != remote_as) { | |
1134 | /* raise error, log this, close session */ | |
1135 | zlog_err( | |
1136 | "%s bad OPEN, got AS4 capability, but remote_as %u" | |
1137 | " mismatch with 16bit 'myasn' %u in open", | |
1138 | peer->host, as4, remote_as); | |
1139 | bgp_notify_send_with_data(peer, BGP_NOTIFY_OPEN_ERR, | |
1140 | BGP_NOTIFY_OPEN_BAD_PEER_AS, | |
1141 | notify_data_remote_as4, 4); | |
d8151687 | 1142 | return BGP_Stop; |
d62a17ae | 1143 | } |
1144 | } | |
6b0655a2 | 1145 | |
d62a17ae | 1146 | /* remote router-id check. */ |
1147 | if (remote_id.s_addr == 0 || IPV4_CLASS_DE(ntohl(remote_id.s_addr)) | |
1148 | || ntohl(peer->local_id.s_addr) == ntohl(remote_id.s_addr)) { | |
1149 | if (bgp_debug_neighbor_events(peer)) | |
1150 | zlog_debug("%s bad OPEN, wrong router identifier %s", | |
1151 | peer->host, inet_ntoa(remote_id)); | |
1152 | bgp_notify_send_with_data(peer, BGP_NOTIFY_OPEN_ERR, | |
1153 | BGP_NOTIFY_OPEN_BAD_BGP_IDENT, | |
1154 | notify_data_remote_id, 4); | |
d8151687 | 1155 | return BGP_Stop; |
d62a17ae | 1156 | } |
1157 | ||
1158 | /* Set remote router-id */ | |
1159 | peer->remote_id = remote_id; | |
1160 | ||
1161 | /* Peer BGP version check. */ | |
1162 | if (version != BGP_VERSION_4) { | |
d7c0a89a | 1163 | uint16_t maxver = htons(BGP_VERSION_4); |
d62a17ae | 1164 | /* XXX this reply may not be correct if version < 4 XXX */ |
1165 | if (bgp_debug_neighbor_events(peer)) | |
1166 | zlog_debug( | |
1167 | "%s bad protocol version, remote requested %d, local request %d", | |
1168 | peer->host, version, BGP_VERSION_4); | |
1169 | /* Data must be in network byte order here */ | |
1170 | bgp_notify_send_with_data(peer, BGP_NOTIFY_OPEN_ERR, | |
1171 | BGP_NOTIFY_OPEN_UNSUP_VERSION, | |
d7c0a89a | 1172 | (uint8_t *)&maxver, 2); |
d8151687 | 1173 | return BGP_Stop; |
d62a17ae | 1174 | } |
1175 | ||
1176 | /* Check neighbor as number. */ | |
1177 | if (peer->as_type == AS_UNSPECIFIED) { | |
1178 | if (bgp_debug_neighbor_events(peer)) | |
1179 | zlog_debug( | |
1180 | "%s bad OPEN, remote AS is unspecified currently", | |
1181 | peer->host); | |
1182 | bgp_notify_send_with_data(peer, BGP_NOTIFY_OPEN_ERR, | |
1183 | BGP_NOTIFY_OPEN_BAD_PEER_AS, | |
1184 | notify_data_remote_as, 2); | |
d8151687 | 1185 | return BGP_Stop; |
d62a17ae | 1186 | } else if (peer->as_type == AS_INTERNAL) { |
1187 | if (remote_as != peer->bgp->as) { | |
1188 | if (bgp_debug_neighbor_events(peer)) | |
1189 | zlog_debug( | |
1190 | "%s bad OPEN, remote AS is %u, internal specified", | |
1191 | peer->host, remote_as); | |
1192 | bgp_notify_send_with_data(peer, BGP_NOTIFY_OPEN_ERR, | |
1193 | BGP_NOTIFY_OPEN_BAD_PEER_AS, | |
1194 | notify_data_remote_as, 2); | |
d8151687 | 1195 | return BGP_Stop; |
1ff9a340 | 1196 | } |
d62a17ae | 1197 | peer->as = peer->local_as; |
1198 | } else if (peer->as_type == AS_EXTERNAL) { | |
1199 | if (remote_as == peer->bgp->as) { | |
1200 | if (bgp_debug_neighbor_events(peer)) | |
1201 | zlog_debug( | |
1202 | "%s bad OPEN, remote AS is %u, external specified", | |
1203 | peer->host, remote_as); | |
1204 | bgp_notify_send_with_data(peer, BGP_NOTIFY_OPEN_ERR, | |
1205 | BGP_NOTIFY_OPEN_BAD_PEER_AS, | |
1206 | notify_data_remote_as, 2); | |
d8151687 | 1207 | return BGP_Stop; |
1ff9a340 | 1208 | } |
d62a17ae | 1209 | peer->as = remote_as; |
1210 | } else if ((peer->as_type == AS_SPECIFIED) && (remote_as != peer->as)) { | |
1211 | if (bgp_debug_neighbor_events(peer)) | |
1212 | zlog_debug("%s bad OPEN, remote AS is %u, expected %u", | |
1213 | peer->host, remote_as, peer->as); | |
1214 | bgp_notify_send_with_data(peer, BGP_NOTIFY_OPEN_ERR, | |
1215 | BGP_NOTIFY_OPEN_BAD_PEER_AS, | |
1216 | notify_data_remote_as, 2); | |
d8151687 | 1217 | return BGP_Stop; |
eb821189 | 1218 | } |
718e3744 | 1219 | |
d62a17ae | 1220 | /* From the rfc: Upon receipt of an OPEN message, a BGP speaker MUST |
1221 | calculate the value of the Hold Timer by using the smaller of its | |
1222 | configured Hold Time and the Hold Time received in the OPEN message. | |
1223 | The Hold Time MUST be either zero or at least three seconds. An | |
1224 | implementation may reject connections on the basis of the Hold Time. | |
0b2aa3a0 | 1225 | */ |
d62a17ae | 1226 | |
1227 | if (holdtime < 3 && holdtime != 0) { | |
1228 | bgp_notify_send_with_data(peer, BGP_NOTIFY_OPEN_ERR, | |
1229 | BGP_NOTIFY_OPEN_UNACEP_HOLDTIME, | |
d7c0a89a | 1230 | (uint8_t *)holdtime_ptr, 2); |
d8151687 | 1231 | return BGP_Stop; |
0b2aa3a0 | 1232 | } |
d62a17ae | 1233 | |
1234 | /* From the rfc: A reasonable maximum time between KEEPALIVE messages | |
1235 | would be one third of the Hold Time interval. KEEPALIVE messages | |
1236 | MUST NOT be sent more frequently than one per second. An | |
1237 | implementation MAY adjust the rate at which it sends KEEPALIVE | |
1238 | messages as a function of the Hold Time interval. */ | |
1239 | ||
d25e4efc | 1240 | if (PEER_OR_GROUP_TIMER_SET(peer)) |
d62a17ae | 1241 | send_holdtime = peer->holdtime; |
1242 | else | |
1243 | send_holdtime = peer->bgp->default_holdtime; | |
1244 | ||
1245 | if (holdtime < send_holdtime) | |
1246 | peer->v_holdtime = holdtime; | |
1247 | else | |
1248 | peer->v_holdtime = send_holdtime; | |
1249 | ||
d25e4efc DS |
1250 | if ((PEER_OR_GROUP_TIMER_SET(peer)) |
1251 | && (peer->keepalive < peer->v_holdtime / 3)) | |
d62a17ae | 1252 | peer->v_keepalive = peer->keepalive; |
1253 | else | |
1254 | peer->v_keepalive = peer->v_holdtime / 3; | |
1255 | ||
1256 | /* Open option part parse. */ | |
1257 | if (optlen != 0) { | |
1258 | if ((ret = bgp_open_option_parse(peer, optlen, &mp_capability)) | |
1259 | < 0) | |
d8151687 | 1260 | return BGP_Stop; |
d62a17ae | 1261 | } else { |
1262 | if (bgp_debug_neighbor_events(peer)) | |
1263 | zlog_debug("%s rcvd OPEN w/ OPTION parameter len: 0", | |
1264 | peer->host); | |
0299c004 | 1265 | } |
d62a17ae | 1266 | |
1267 | /* | |
1268 | * Assume that the peer supports the locally configured set of | |
1269 | * AFI/SAFIs if the peer did not send us any Mulitiprotocol | |
1270 | * capabilities, or if 'override-capability' is configured. | |
1271 | */ | |
1272 | if (!mp_capability | |
1273 | || CHECK_FLAG(peer->flags, PEER_FLAG_OVERRIDE_CAPABILITY)) { | |
1274 | peer->afc_nego[AFI_IP][SAFI_UNICAST] = | |
1275 | peer->afc[AFI_IP][SAFI_UNICAST]; | |
1276 | peer->afc_nego[AFI_IP][SAFI_MULTICAST] = | |
1277 | peer->afc[AFI_IP][SAFI_MULTICAST]; | |
1278 | peer->afc_nego[AFI_IP][SAFI_LABELED_UNICAST] = | |
1279 | peer->afc[AFI_IP][SAFI_LABELED_UNICAST]; | |
7c40bf39 | 1280 | peer->afc_nego[AFI_IP][SAFI_FLOWSPEC] = |
1281 | peer->afc[AFI_IP][SAFI_FLOWSPEC]; | |
d62a17ae | 1282 | peer->afc_nego[AFI_IP6][SAFI_UNICAST] = |
1283 | peer->afc[AFI_IP6][SAFI_UNICAST]; | |
1284 | peer->afc_nego[AFI_IP6][SAFI_MULTICAST] = | |
1285 | peer->afc[AFI_IP6][SAFI_MULTICAST]; | |
1286 | peer->afc_nego[AFI_IP6][SAFI_LABELED_UNICAST] = | |
1287 | peer->afc[AFI_IP6][SAFI_LABELED_UNICAST]; | |
1288 | peer->afc_nego[AFI_L2VPN][SAFI_EVPN] = | |
1289 | peer->afc[AFI_L2VPN][SAFI_EVPN]; | |
7c40bf39 | 1290 | peer->afc_nego[AFI_IP6][SAFI_FLOWSPEC] = |
1291 | peer->afc[AFI_IP6][SAFI_FLOWSPEC]; | |
0299c004 | 1292 | } |
d62a17ae | 1293 | |
1294 | /* When collision is detected and this peer is closed. Retrun | |
1295 | immidiately. */ | |
1296 | ret = bgp_collision_detect(peer, remote_id); | |
1297 | if (ret < 0) | |
d8151687 | 1298 | return BGP_Stop; |
d62a17ae | 1299 | |
1300 | /* Get sockname. */ | |
1301 | if ((ret = bgp_getsockname(peer)) < 0) { | |
1302 | zlog_err("%s: bgp_getsockname() failed for peer: %s", | |
1303 | __FUNCTION__, peer->host); | |
d8151687 | 1304 | return BGP_Stop; |
d62a17ae | 1305 | } |
1306 | ||
1307 | /* Verify valid local address present based on negotiated | |
1308 | * address-families. */ | |
1309 | if (peer->afc_nego[AFI_IP][SAFI_UNICAST] | |
e5f22b30 | 1310 | || peer->afc_nego[AFI_IP][SAFI_LABELED_UNICAST] |
d62a17ae | 1311 | || peer->afc_nego[AFI_IP][SAFI_MULTICAST] |
1312 | || peer->afc_nego[AFI_IP][SAFI_MPLS_VPN] | |
1313 | || peer->afc_nego[AFI_IP][SAFI_ENCAP]) { | |
1314 | if (!peer->nexthop.v4.s_addr) { | |
1315 | #if defined(HAVE_CUMULUS) | |
1316 | zlog_err( | |
1317 | "%s: No local IPv4 addr resetting connection, fd %d", | |
1318 | peer->host, peer->fd); | |
1319 | bgp_notify_send(peer, BGP_NOTIFY_CEASE, | |
1320 | BGP_NOTIFY_SUBCODE_UNSPECIFIC); | |
d8151687 | 1321 | return BGP_Stop; |
1d808091 | 1322 | #endif |
d62a17ae | 1323 | } |
1324 | } | |
1325 | if (peer->afc_nego[AFI_IP6][SAFI_UNICAST] | |
e5f22b30 | 1326 | || peer->afc_nego[AFI_IP6][SAFI_LABELED_UNICAST] |
d62a17ae | 1327 | || peer->afc_nego[AFI_IP6][SAFI_MULTICAST] |
1328 | || peer->afc_nego[AFI_IP6][SAFI_MPLS_VPN] | |
1329 | || peer->afc_nego[AFI_IP6][SAFI_ENCAP]) { | |
1330 | if (IN6_IS_ADDR_UNSPECIFIED(&peer->nexthop.v6_global)) { | |
1331 | #if defined(HAVE_CUMULUS) | |
1332 | zlog_err( | |
1333 | "%s: No local IPv6 addr resetting connection, fd %d", | |
1334 | peer->host, peer->fd); | |
1335 | bgp_notify_send(peer, BGP_NOTIFY_CEASE, | |
1336 | BGP_NOTIFY_SUBCODE_UNSPECIFIC); | |
d8151687 | 1337 | return BGP_Stop; |
1d808091 | 1338 | #endif |
d62a17ae | 1339 | } |
1340 | } | |
1341 | peer->rtt = sockopt_tcp_rtt(peer->fd); | |
1342 | ||
d8151687 | 1343 | return Receive_OPEN_message; |
718e3744 | 1344 | } |
1345 | ||
d8151687 QY |
1346 | /** |
1347 | * Process BGP KEEPALIVE message for peer. | |
1348 | * | |
1349 | * @param peer | |
1350 | * @param size size of the packet | |
1351 | * @return as in summary | |
1352 | */ | |
1353 | static int bgp_keepalive_receive(struct peer *peer, bgp_size_t size) | |
f188f2c4 | 1354 | { |
d8151687 QY |
1355 | if (bgp_debug_keepalive(peer)) |
1356 | zlog_debug("%s KEEPALIVE rcvd", peer->host); | |
d62a17ae | 1357 | |
d8151687 | 1358 | bgp_update_implicit_eors(peer); |
d62a17ae | 1359 | |
d8151687 | 1360 | return Receive_KEEPALIVE_message; |
f188f2c4 DS |
1361 | } |
1362 | ||
d62a17ae | 1363 | |
d8151687 QY |
1364 | /** |
1365 | * Process BGP UPDATE message for peer. | |
1366 | * | |
1367 | * Parses UPDATE and creates attribute object. | |
1368 | * | |
1369 | * @param peer | |
1370 | * @param size size of the packet | |
1371 | * @return as in summary | |
7ef5a232 | 1372 | */ |
d62a17ae | 1373 | static int bgp_update_receive(struct peer *peer, bgp_size_t size) |
718e3744 | 1374 | { |
d62a17ae | 1375 | int ret, nlri_ret; |
d7c0a89a | 1376 | uint8_t *end; |
d62a17ae | 1377 | struct stream *s; |
1378 | struct attr attr; | |
1379 | bgp_size_t attribute_len; | |
1380 | bgp_size_t update_len; | |
1381 | bgp_size_t withdraw_len; | |
1382 | ||
1383 | enum NLRI_TYPES { | |
1384 | NLRI_UPDATE, | |
1385 | NLRI_WITHDRAW, | |
1386 | NLRI_MP_UPDATE, | |
1387 | NLRI_MP_WITHDRAW, | |
1388 | NLRI_TYPE_MAX | |
1389 | }; | |
1390 | struct bgp_nlri nlris[NLRI_TYPE_MAX]; | |
1391 | ||
1392 | /* Status must be Established. */ | |
1393 | if (peer->status != Established) { | |
1394 | zlog_err("%s [FSM] Update packet received under status %s", | |
1395 | peer->host, | |
1396 | lookup_msg(bgp_status_msg, peer->status, NULL)); | |
1397 | bgp_notify_send(peer, BGP_NOTIFY_FSM_ERR, 0); | |
d8151687 | 1398 | return BGP_Stop; |
d62a17ae | 1399 | } |
1400 | ||
1401 | /* Set initial values. */ | |
1402 | memset(&attr, 0, sizeof(struct attr)); | |
1403 | attr.label_index = BGP_INVALID_LABEL_INDEX; | |
1404 | attr.label = MPLS_INVALID_LABEL; | |
1405 | memset(&nlris, 0, sizeof(nlris)); | |
1406 | memset(peer->rcvd_attr_str, 0, BUFSIZ); | |
1407 | peer->rcvd_attr_printed = 0; | |
1408 | ||
424ab01d | 1409 | s = peer->curr; |
d62a17ae | 1410 | end = stream_pnt(s) + size; |
1411 | ||
1412 | /* RFC1771 6.3 If the Unfeasible Routes Length or Total Attribute | |
1413 | Length is too large (i.e., if Unfeasible Routes Length + Total | |
1414 | Attribute Length + 23 exceeds the message Length), then the Error | |
1415 | Subcode is set to Malformed Attribute List. */ | |
1416 | if (stream_pnt(s) + 2 > end) { | |
1417 | zlog_err( | |
1418 | "%s [Error] Update packet error" | |
1419 | " (packet length is short for unfeasible length)", | |
1420 | peer->host); | |
1421 | bgp_notify_send(peer, BGP_NOTIFY_UPDATE_ERR, | |
1422 | BGP_NOTIFY_UPDATE_MAL_ATTR); | |
d8151687 | 1423 | return BGP_Stop; |
d62a17ae | 1424 | } |
1425 | ||
1426 | /* Unfeasible Route Length. */ | |
1427 | withdraw_len = stream_getw(s); | |
1428 | ||
1429 | /* Unfeasible Route Length check. */ | |
1430 | if (stream_pnt(s) + withdraw_len > end) { | |
1431 | zlog_err( | |
1432 | "%s [Error] Update packet error" | |
1433 | " (packet unfeasible length overflow %d)", | |
1434 | peer->host, withdraw_len); | |
1435 | bgp_notify_send(peer, BGP_NOTIFY_UPDATE_ERR, | |
1436 | BGP_NOTIFY_UPDATE_MAL_ATTR); | |
d8151687 | 1437 | return BGP_Stop; |
d62a17ae | 1438 | } |
1439 | ||
1440 | /* Unfeasible Route packet format check. */ | |
1441 | if (withdraw_len > 0) { | |
1442 | nlris[NLRI_WITHDRAW].afi = AFI_IP; | |
1443 | nlris[NLRI_WITHDRAW].safi = SAFI_UNICAST; | |
1444 | nlris[NLRI_WITHDRAW].nlri = stream_pnt(s); | |
1445 | nlris[NLRI_WITHDRAW].length = withdraw_len; | |
1446 | stream_forward_getp(s, withdraw_len); | |
1447 | } | |
1448 | ||
1449 | /* Attribute total length check. */ | |
1450 | if (stream_pnt(s) + 2 > end) { | |
1451 | zlog_warn( | |
1452 | "%s [Error] Packet Error" | |
1453 | " (update packet is short for attribute length)", | |
1454 | peer->host); | |
1455 | bgp_notify_send(peer, BGP_NOTIFY_UPDATE_ERR, | |
1456 | BGP_NOTIFY_UPDATE_MAL_ATTR); | |
d8151687 | 1457 | return BGP_Stop; |
d62a17ae | 1458 | } |
1459 | ||
1460 | /* Fetch attribute total length. */ | |
1461 | attribute_len = stream_getw(s); | |
1462 | ||
1463 | /* Attribute length check. */ | |
1464 | if (stream_pnt(s) + attribute_len > end) { | |
1465 | zlog_warn( | |
1466 | "%s [Error] Packet Error" | |
1467 | " (update packet attribute length overflow %d)", | |
1468 | peer->host, attribute_len); | |
1469 | bgp_notify_send(peer, BGP_NOTIFY_UPDATE_ERR, | |
1470 | BGP_NOTIFY_UPDATE_MAL_ATTR); | |
d8151687 | 1471 | return BGP_Stop; |
d62a17ae | 1472 | } |
1473 | ||
1474 | /* Certain attribute parsing errors should not be considered bad enough | |
1475 | * to reset the session for, most particularly any partial/optional | |
1476 | * attributes that have 'tunneled' over speakers that don't understand | |
1477 | * them. Instead we withdraw only the prefix concerned. | |
1478 | * | |
1479 | * Complicates the flow a little though.. | |
1480 | */ | |
1481 | bgp_attr_parse_ret_t attr_parse_ret = BGP_ATTR_PARSE_PROCEED; | |
1482 | /* This define morphs the update case into a withdraw when lower levels | |
1483 | * have signalled an error condition where this is best. | |
1484 | */ | |
b881c707 | 1485 | #define NLRI_ATTR_ARG (attr_parse_ret != BGP_ATTR_PARSE_WITHDRAW ? &attr : NULL) |
718e3744 | 1486 | |
d62a17ae | 1487 | /* Parse attribute when it exists. */ |
1488 | if (attribute_len) { | |
1489 | attr_parse_ret = bgp_attr_parse(peer, &attr, attribute_len, | |
1490 | &nlris[NLRI_MP_UPDATE], | |
1491 | &nlris[NLRI_MP_WITHDRAW]); | |
1492 | if (attr_parse_ret == BGP_ATTR_PARSE_ERROR) { | |
1493 | bgp_attr_unintern_sub(&attr); | |
d8151687 | 1494 | return BGP_Stop; |
d62a17ae | 1495 | } |
1496 | } | |
1497 | ||
1498 | /* Logging the attribute. */ | |
1499 | if (attr_parse_ret == BGP_ATTR_PARSE_WITHDRAW | |
1500 | || BGP_DEBUG(update, UPDATE_IN) | |
1501 | || BGP_DEBUG(update, UPDATE_PREFIX)) { | |
1502 | ret = bgp_dump_attr(&attr, peer->rcvd_attr_str, BUFSIZ); | |
1503 | ||
1504 | if (attr_parse_ret == BGP_ATTR_PARSE_WITHDRAW) | |
1505 | zlog_err( | |
1506 | "%s rcvd UPDATE with errors in attr(s)!! Withdrawing route.", | |
1507 | peer->host); | |
1508 | ||
1509 | if (ret && bgp_debug_update(peer, NULL, NULL, 1)) { | |
1510 | zlog_debug("%s rcvd UPDATE w/ attr: %s", peer->host, | |
1511 | peer->rcvd_attr_str); | |
1512 | peer->rcvd_attr_printed = 1; | |
1513 | } | |
1514 | } | |
1515 | ||
1516 | /* Network Layer Reachability Information. */ | |
1517 | update_len = end - stream_pnt(s); | |
1518 | ||
1519 | if (update_len) { | |
1520 | /* Set NLRI portion to structure. */ | |
1521 | nlris[NLRI_UPDATE].afi = AFI_IP; | |
1522 | nlris[NLRI_UPDATE].safi = SAFI_UNICAST; | |
1523 | nlris[NLRI_UPDATE].nlri = stream_pnt(s); | |
1524 | nlris[NLRI_UPDATE].length = update_len; | |
1525 | stream_forward_getp(s, update_len); | |
1526 | } | |
1527 | ||
1528 | if (BGP_DEBUG(update, UPDATE_IN)) | |
1529 | zlog_debug("%s rcvd UPDATE wlen %d attrlen %d alen %d", | |
1530 | peer->host, withdraw_len, attribute_len, update_len); | |
1531 | ||
1532 | /* Parse any given NLRIs */ | |
1533 | for (int i = NLRI_UPDATE; i < NLRI_TYPE_MAX; i++) { | |
1534 | if (!nlris[i].nlri) | |
1535 | continue; | |
1536 | ||
1537 | /* NLRI is processed iff the peer if configured for the specific | |
1538 | * afi/safi */ | |
1539 | if (!peer->afc[nlris[i].afi][nlris[i].safi]) { | |
1540 | zlog_info( | |
1541 | "%s [Info] UPDATE for non-enabled AFI/SAFI %u/%u", | |
1542 | peer->host, nlris[i].afi, nlris[i].safi); | |
1543 | continue; | |
1544 | } | |
1545 | ||
1546 | /* EoR handled later */ | |
1547 | if (nlris[i].length == 0) | |
1548 | continue; | |
1549 | ||
1550 | switch (i) { | |
1551 | case NLRI_UPDATE: | |
1552 | case NLRI_MP_UPDATE: | |
1553 | nlri_ret = bgp_nlri_parse(peer, NLRI_ATTR_ARG, | |
1554 | &nlris[i], 0); | |
1555 | break; | |
1556 | case NLRI_WITHDRAW: | |
1557 | case NLRI_MP_WITHDRAW: | |
1558 | nlri_ret = bgp_nlri_parse(peer, &attr, &nlris[i], 1); | |
1559 | break; | |
1560 | default: | |
1561 | nlri_ret = -1; | |
1562 | } | |
1563 | ||
1564 | if (nlri_ret < 0) { | |
1565 | zlog_err("%s [Error] Error parsing NLRI", peer->host); | |
1566 | if (peer->status == Established) | |
1567 | bgp_notify_send( | |
1568 | peer, BGP_NOTIFY_UPDATE_ERR, | |
1569 | i <= NLRI_WITHDRAW | |
1570 | ? BGP_NOTIFY_UPDATE_INVAL_NETWORK | |
1571 | : BGP_NOTIFY_UPDATE_OPT_ATTR_ERR); | |
1572 | bgp_attr_unintern_sub(&attr); | |
d8151687 | 1573 | return BGP_Stop; |
d62a17ae | 1574 | } |
1575 | } | |
1576 | ||
1577 | /* EoR checks | |
1578 | * | |
1579 | * Non-MP IPv4/Unicast EoR is a completely empty UPDATE | |
1580 | * and MP EoR should have only an empty MP_UNREACH | |
1581 | */ | |
996c9314 LB |
1582 | if ((!update_len && !withdraw_len && nlris[NLRI_MP_UPDATE].length == 0) |
1583 | || (attr_parse_ret == BGP_ATTR_PARSE_EOR)) { | |
d62a17ae | 1584 | afi_t afi = 0; |
1585 | safi_t safi; | |
1586 | ||
1587 | /* Non-MP IPv4/Unicast is a completely emtpy UPDATE - already | |
1588 | * checked | |
1589 | * update and withdraw NLRI lengths are 0. | |
1590 | */ | |
1591 | if (!attribute_len) { | |
1592 | afi = AFI_IP; | |
1593 | safi = SAFI_UNICAST; | |
1594 | } else if (attr.flag & ATTR_FLAG_BIT(BGP_ATTR_MP_UNREACH_NLRI) | |
1595 | && nlris[NLRI_MP_WITHDRAW].length == 0) { | |
1596 | afi = nlris[NLRI_MP_WITHDRAW].afi; | |
1597 | safi = nlris[NLRI_MP_WITHDRAW].safi; | |
9b9df989 DS |
1598 | } else if (attr_parse_ret == BGP_ATTR_PARSE_EOR) { |
1599 | afi = nlris[NLRI_MP_UPDATE].afi; | |
1600 | safi = nlris[NLRI_MP_UPDATE].safi; | |
d62a17ae | 1601 | } |
1602 | ||
1603 | if (afi && peer->afc[afi][safi]) { | |
1604 | /* End-of-RIB received */ | |
1605 | if (!CHECK_FLAG(peer->af_sflags[afi][safi], | |
1606 | PEER_STATUS_EOR_RECEIVED)) { | |
1607 | SET_FLAG(peer->af_sflags[afi][safi], | |
1608 | PEER_STATUS_EOR_RECEIVED); | |
1609 | bgp_update_explicit_eors(peer); | |
1610 | } | |
1611 | ||
1612 | /* NSF delete stale route */ | |
1613 | if (peer->nsf[afi][safi]) | |
1614 | bgp_clear_stale_route(peer, afi, safi); | |
1615 | ||
1616 | if (bgp_debug_neighbor_events(peer)) { | |
1617 | zlog_debug("rcvd End-of-RIB for %s from %s", | |
1618 | afi_safi_print(afi, safi), | |
1619 | peer->host); | |
1620 | } | |
1621 | } | |
f80f838b | 1622 | } |
d62a17ae | 1623 | |
1624 | /* Everything is done. We unintern temporary structures which | |
1625 | interned in bgp_attr_parse(). */ | |
1626 | bgp_attr_unintern_sub(&attr); | |
1627 | ||
d62a17ae | 1628 | peer->update_time = bgp_clock(); |
1629 | ||
1630 | /* Rearm holdtime timer */ | |
1631 | BGP_TIMER_OFF(peer->t_holdtime); | |
1632 | bgp_timer_set(peer); | |
1633 | ||
d8151687 | 1634 | return Receive_UPDATE_message; |
718e3744 | 1635 | } |
1636 | ||
d8151687 QY |
1637 | /** |
1638 | * Process BGP NOTIFY message for peer. | |
1639 | * | |
1640 | * @param peer | |
1641 | * @param size size of the packet | |
1642 | * @return as in summary | |
1643 | */ | |
1644 | static int bgp_notify_receive(struct peer *peer, bgp_size_t size) | |
718e3744 | 1645 | { |
d62a17ae | 1646 | struct bgp_notify bgp_notify; |
1647 | ||
1648 | if (peer->notify.data) { | |
1649 | XFREE(MTYPE_TMP, peer->notify.data); | |
1650 | peer->notify.data = NULL; | |
1651 | peer->notify.length = 0; | |
1652 | } | |
1653 | ||
424ab01d QY |
1654 | bgp_notify.code = stream_getc(peer->curr); |
1655 | bgp_notify.subcode = stream_getc(peer->curr); | |
d62a17ae | 1656 | bgp_notify.length = size - 2; |
1657 | bgp_notify.data = NULL; | |
1658 | ||
1659 | /* Preserv notify code and sub code. */ | |
1660 | peer->notify.code = bgp_notify.code; | |
1661 | peer->notify.subcode = bgp_notify.subcode; | |
1662 | /* For further diagnostic record returned Data. */ | |
1663 | if (bgp_notify.length) { | |
1664 | peer->notify.length = size - 2; | |
1665 | peer->notify.data = XMALLOC(MTYPE_TMP, size - 2); | |
424ab01d | 1666 | memcpy(peer->notify.data, stream_pnt(peer->curr), size - 2); |
d62a17ae | 1667 | } |
1668 | ||
1669 | /* For debug */ | |
1670 | { | |
1671 | int i; | |
1672 | int first = 0; | |
1673 | char c[4]; | |
1674 | ||
1675 | if (bgp_notify.length) { | |
1676 | bgp_notify.data = | |
1677 | XMALLOC(MTYPE_TMP, bgp_notify.length * 3); | |
1678 | for (i = 0; i < bgp_notify.length; i++) | |
1679 | if (first) { | |
1680 | sprintf(c, " %02x", | |
424ab01d | 1681 | stream_getc(peer->curr)); |
d62a17ae | 1682 | strcat(bgp_notify.data, c); |
1683 | } else { | |
1684 | first = 1; | |
1685 | sprintf(c, "%02x", | |
424ab01d | 1686 | stream_getc(peer->curr)); |
d62a17ae | 1687 | strcpy(bgp_notify.data, c); |
1688 | } | |
d7c0a89a | 1689 | bgp_notify.raw_data = (uint8_t *)peer->notify.data; |
d62a17ae | 1690 | } |
1691 | ||
1692 | bgp_notify_print(peer, &bgp_notify, "received"); | |
1693 | if (bgp_notify.data) { | |
1694 | XFREE(MTYPE_TMP, bgp_notify.data); | |
1695 | bgp_notify.data = NULL; | |
1696 | bgp_notify.length = 0; | |
1697 | } | |
1698 | } | |
1699 | ||
1700 | /* peer count update */ | |
0112e9e0 | 1701 | atomic_fetch_add_explicit(&peer->notify_in, 1, memory_order_relaxed); |
d62a17ae | 1702 | |
1703 | peer->last_reset = PEER_DOWN_NOTIFY_RECEIVED; | |
1704 | ||
1705 | /* We have to check for Notify with Unsupported Optional Parameter. | |
1706 | in that case we fallback to open without the capability option. | |
1707 | But this done in bgp_stop. We just mark it here to avoid changing | |
1708 | the fsm tables. */ | |
1709 | if (bgp_notify.code == BGP_NOTIFY_OPEN_ERR | |
1710 | && bgp_notify.subcode == BGP_NOTIFY_OPEN_UNSUP_PARAM) | |
1711 | UNSET_FLAG(peer->sflags, PEER_STATUS_CAPABILITY_OPEN); | |
1712 | ||
d8151687 | 1713 | return Receive_NOTIFICATION_message; |
718e3744 | 1714 | } |
1715 | ||
d8151687 QY |
1716 | /** |
1717 | * Process BGP ROUTEREFRESH message for peer. | |
1718 | * | |
1719 | * @param peer | |
1720 | * @param size size of the packet | |
1721 | * @return as in summary | |
1722 | */ | |
1723 | static int bgp_route_refresh_receive(struct peer *peer, bgp_size_t size) | |
718e3744 | 1724 | { |
d62a17ae | 1725 | iana_afi_t pkt_afi; |
1726 | afi_t afi; | |
5c525538 RW |
1727 | iana_safi_t pkt_safi; |
1728 | safi_t safi; | |
d62a17ae | 1729 | struct stream *s; |
1730 | struct peer_af *paf; | |
1731 | struct update_group *updgrp; | |
1732 | struct peer *updgrp_peer; | |
1733 | ||
1734 | /* If peer does not have the capability, send notification. */ | |
1735 | if (!CHECK_FLAG(peer->cap, PEER_CAP_REFRESH_ADV)) { | |
1736 | zlog_err("%s [Error] BGP route refresh is not enabled", | |
1737 | peer->host); | |
1738 | bgp_notify_send(peer, BGP_NOTIFY_HEADER_ERR, | |
1739 | BGP_NOTIFY_HEADER_BAD_MESTYPE); | |
d8151687 | 1740 | return BGP_Stop; |
d62a17ae | 1741 | } |
1742 | ||
1743 | /* Status must be Established. */ | |
1744 | if (peer->status != Established) { | |
1745 | zlog_err( | |
1746 | "%s [Error] Route refresh packet received under status %s", | |
1747 | peer->host, | |
1748 | lookup_msg(bgp_status_msg, peer->status, NULL)); | |
1749 | bgp_notify_send(peer, BGP_NOTIFY_FSM_ERR, 0); | |
d8151687 | 1750 | return BGP_Stop; |
d62a17ae | 1751 | } |
1752 | ||
424ab01d | 1753 | s = peer->curr; |
d62a17ae | 1754 | |
1755 | /* Parse packet. */ | |
1756 | pkt_afi = stream_getw(s); | |
1757 | (void)stream_getc(s); | |
1758 | pkt_safi = stream_getc(s); | |
1759 | ||
1760 | if (bgp_debug_update(peer, NULL, NULL, 0)) | |
1761 | zlog_debug("%s rcvd REFRESH_REQ for afi/safi: %d/%d", | |
1762 | peer->host, pkt_afi, pkt_safi); | |
1763 | ||
1764 | /* Convert AFI, SAFI to internal values and check. */ | |
1765 | if (bgp_map_afi_safi_iana2int(pkt_afi, pkt_safi, &afi, &safi)) { | |
1766 | zlog_info( | |
1767 | "%s REFRESH_REQ for unrecognized afi/safi: %d/%d - ignored", | |
1768 | peer->host, pkt_afi, pkt_safi); | |
d8151687 | 1769 | return BGP_PACKET_NOOP; |
d62a17ae | 1770 | } |
1771 | ||
1772 | if (size != BGP_MSG_ROUTE_REFRESH_MIN_SIZE - BGP_HEADER_SIZE) { | |
d7c0a89a QY |
1773 | uint8_t *end; |
1774 | uint8_t when_to_refresh; | |
1775 | uint8_t orf_type; | |
1776 | uint16_t orf_len; | |
d62a17ae | 1777 | |
1778 | if (size - (BGP_MSG_ROUTE_REFRESH_MIN_SIZE - BGP_HEADER_SIZE) | |
1779 | < 5) { | |
1780 | zlog_info("%s ORF route refresh length error", | |
1781 | peer->host); | |
1782 | bgp_notify_send(peer, BGP_NOTIFY_CEASE, 0); | |
d8151687 | 1783 | return BGP_Stop; |
718e3744 | 1784 | } |
1785 | ||
d62a17ae | 1786 | when_to_refresh = stream_getc(s); |
1787 | end = stream_pnt(s) + (size - 5); | |
1788 | ||
1789 | while ((stream_pnt(s) + 2) < end) { | |
1790 | orf_type = stream_getc(s); | |
1791 | orf_len = stream_getw(s); | |
1792 | ||
1793 | /* orf_len in bounds? */ | |
1794 | if ((stream_pnt(s) + orf_len) > end) | |
1795 | break; /* XXX: Notify instead?? */ | |
1796 | if (orf_type == ORF_TYPE_PREFIX | |
1797 | || orf_type == ORF_TYPE_PREFIX_OLD) { | |
1798 | uint8_t *p_pnt = stream_pnt(s); | |
1799 | uint8_t *p_end = stream_pnt(s) + orf_len; | |
1800 | struct orf_prefix orfp; | |
d7c0a89a QY |
1801 | uint8_t common = 0; |
1802 | uint32_t seq; | |
d62a17ae | 1803 | int psize; |
1804 | char name[BUFSIZ]; | |
1805 | int ret = CMD_SUCCESS; | |
1806 | ||
1807 | if (bgp_debug_neighbor_events(peer)) { | |
1808 | zlog_debug( | |
1809 | "%s rcvd Prefixlist ORF(%d) length %d", | |
1810 | peer->host, orf_type, orf_len); | |
1811 | } | |
1812 | ||
1813 | /* we're going to read at least 1 byte of common | |
1814 | * ORF header, | |
1815 | * and 7 bytes of ORF Address-filter entry from | |
1816 | * the stream | |
1817 | */ | |
1818 | if (orf_len < 7) | |
1819 | break; | |
1820 | ||
1821 | /* ORF prefix-list name */ | |
1822 | sprintf(name, "%s.%d.%d", peer->host, afi, | |
1823 | safi); | |
1824 | ||
1825 | while (p_pnt < p_end) { | |
1826 | /* If the ORF entry is malformed, want | |
1827 | * to read as much of it | |
1828 | * as possible without going beyond the | |
1829 | * bounds of the entry, | |
1830 | * to maximise debug information. | |
1831 | */ | |
1832 | int ok; | |
1833 | memset(&orfp, 0, | |
1834 | sizeof(struct orf_prefix)); | |
1835 | common = *p_pnt++; | |
1836 | /* after ++: p_pnt <= p_end */ | |
1837 | if (common | |
1838 | & ORF_COMMON_PART_REMOVE_ALL) { | |
1839 | if (bgp_debug_neighbor_events( | |
1840 | peer)) | |
1841 | zlog_debug( | |
1842 | "%s rcvd Remove-All pfxlist ORF request", | |
1843 | peer->host); | |
1844 | prefix_bgp_orf_remove_all(afi, | |
1845 | name); | |
1846 | break; | |
1847 | } | |
d7c0a89a QY |
1848 | ok = ((uint32_t)(p_end - p_pnt) |
1849 | >= sizeof(uint32_t)); | |
d62a17ae | 1850 | if (ok) { |
1851 | memcpy(&seq, p_pnt, | |
d7c0a89a QY |
1852 | sizeof(uint32_t)); |
1853 | p_pnt += sizeof(uint32_t); | |
d62a17ae | 1854 | orfp.seq = ntohl(seq); |
1855 | } else | |
1856 | p_pnt = p_end; | |
1857 | ||
1858 | if ((ok = (p_pnt < p_end))) | |
1859 | orfp.ge = | |
1860 | *p_pnt++; /* value | |
1861 | checked in | |
1862 | prefix_bgp_orf_set() | |
1863 | */ | |
1864 | if ((ok = (p_pnt < p_end))) | |
1865 | orfp.le = | |
1866 | *p_pnt++; /* value | |
1867 | checked in | |
1868 | prefix_bgp_orf_set() | |
1869 | */ | |
1870 | if ((ok = (p_pnt < p_end))) | |
1871 | orfp.p.prefixlen = *p_pnt++; | |
1872 | orfp.p.family = afi2family( | |
1873 | afi); /* afi checked already */ | |
1874 | ||
1875 | psize = PSIZE( | |
1876 | orfp.p.prefixlen); /* 0 if not | |
1877 | ok */ | |
1878 | if (psize | |
1879 | > prefix_blen( | |
1880 | &orfp.p)) /* valid for | |
1881 | family ? */ | |
1882 | { | |
1883 | ok = 0; | |
1884 | psize = prefix_blen(&orfp.p); | |
1885 | } | |
1886 | if (psize | |
1887 | > (p_end - p_pnt)) /* valid for | |
1888 | packet ? */ | |
1889 | { | |
1890 | ok = 0; | |
1891 | psize = p_end - p_pnt; | |
1892 | } | |
1893 | ||
1894 | if (psize > 0) | |
1895 | memcpy(&orfp.p.u.prefix, p_pnt, | |
1896 | psize); | |
1897 | p_pnt += psize; | |
1898 | ||
1899 | if (bgp_debug_neighbor_events(peer)) { | |
1900 | char buf[INET6_BUFSIZ]; | |
1901 | ||
1902 | zlog_debug( | |
1903 | "%s rcvd %s %s seq %u %s/%d ge %d le %d%s", | |
1904 | peer->host, | |
1905 | (common & ORF_COMMON_PART_REMOVE | |
1906 | ? "Remove" | |
1907 | : "Add"), | |
1908 | (common & ORF_COMMON_PART_DENY | |
1909 | ? "deny" | |
1910 | : "permit"), | |
1911 | orfp.seq, | |
1912 | inet_ntop( | |
1913 | orfp.p.family, | |
1914 | &orfp.p.u.prefix, | |
1915 | buf, | |
1916 | INET6_BUFSIZ), | |
1917 | orfp.p.prefixlen, | |
1918 | orfp.ge, orfp.le, | |
1919 | ok ? "" : " MALFORMED"); | |
1920 | } | |
1921 | ||
1922 | if (ok) | |
1923 | ret = prefix_bgp_orf_set( | |
1924 | name, afi, &orfp, | |
1925 | (common & ORF_COMMON_PART_DENY | |
1926 | ? 0 | |
1927 | : 1), | |
1928 | (common & ORF_COMMON_PART_REMOVE | |
1929 | ? 0 | |
1930 | : 1)); | |
1931 | ||
1932 | if (!ok || (ok && ret != CMD_SUCCESS)) { | |
1933 | zlog_info( | |
1934 | "%s Received misformatted prefixlist ORF." | |
1935 | " Remove All pfxlist", | |
1936 | peer->host); | |
1937 | prefix_bgp_orf_remove_all(afi, | |
1938 | name); | |
1939 | break; | |
1940 | } | |
1941 | } | |
1942 | ||
1943 | peer->orf_plist[afi][safi] = | |
1944 | prefix_bgp_orf_lookup(afi, name); | |
1945 | } | |
1946 | stream_forward_getp(s, orf_len); | |
718e3744 | 1947 | } |
d62a17ae | 1948 | if (bgp_debug_neighbor_events(peer)) |
1949 | zlog_debug("%s rcvd Refresh %s ORF request", peer->host, | |
1950 | when_to_refresh == REFRESH_DEFER | |
1951 | ? "Defer" | |
1952 | : "Immediate"); | |
1953 | if (when_to_refresh == REFRESH_DEFER) | |
d8151687 | 1954 | return BGP_PACKET_NOOP; |
d62a17ae | 1955 | } |
40d2700d | 1956 | |
d62a17ae | 1957 | /* First update is deferred until ORF or ROUTE-REFRESH is received */ |
1958 | if (CHECK_FLAG(peer->af_sflags[afi][safi], | |
1959 | PEER_STATUS_ORF_WAIT_REFRESH)) | |
1960 | UNSET_FLAG(peer->af_sflags[afi][safi], | |
1961 | PEER_STATUS_ORF_WAIT_REFRESH); | |
1962 | ||
1963 | paf = peer_af_find(peer, afi, safi); | |
1964 | if (paf && paf->subgroup) { | |
1965 | if (peer->orf_plist[afi][safi]) { | |
1966 | updgrp = PAF_UPDGRP(paf); | |
1967 | updgrp_peer = UPDGRP_PEER(updgrp); | |
1968 | updgrp_peer->orf_plist[afi][safi] = | |
1969 | peer->orf_plist[afi][safi]; | |
1970 | } | |
1971 | ||
1972 | /* If the peer is configured for default-originate clear the | |
1973 | * SUBGRP_STATUS_DEFAULT_ORIGINATE flag so that we will | |
1974 | * re-advertise the | |
1975 | * default | |
1976 | */ | |
1977 | if (CHECK_FLAG(paf->subgroup->sflags, | |
1978 | SUBGRP_STATUS_DEFAULT_ORIGINATE)) | |
1979 | UNSET_FLAG(paf->subgroup->sflags, | |
1980 | SUBGRP_STATUS_DEFAULT_ORIGINATE); | |
718e3744 | 1981 | } |
d62a17ae | 1982 | |
1983 | /* Perform route refreshment to the peer */ | |
1984 | bgp_announce_route(peer, afi, safi); | |
d8151687 QY |
1985 | |
1986 | /* No FSM action necessary */ | |
1987 | return BGP_PACKET_NOOP; | |
718e3744 | 1988 | } |
1989 | ||
d8151687 QY |
1990 | /** |
1991 | * Parse BGP CAPABILITY message for peer. | |
1992 | * | |
1993 | * @param peer | |
1994 | * @param size size of the packet | |
1995 | * @return as in summary | |
1996 | */ | |
d7c0a89a | 1997 | static int bgp_capability_msg_parse(struct peer *peer, uint8_t *pnt, |
d62a17ae | 1998 | bgp_size_t length) |
718e3744 | 1999 | { |
d7c0a89a | 2000 | uint8_t *end; |
d62a17ae | 2001 | struct capability_mp_data mpc; |
2002 | struct capability_header *hdr; | |
d7c0a89a | 2003 | uint8_t action; |
d62a17ae | 2004 | iana_afi_t pkt_afi; |
2005 | afi_t afi; | |
5c525538 RW |
2006 | iana_safi_t pkt_safi; |
2007 | safi_t safi; | |
d62a17ae | 2008 | |
2009 | end = pnt + length; | |
2010 | ||
2011 | while (pnt < end) { | |
2012 | /* We need at least action, capability code and capability | |
2013 | * length. */ | |
2014 | if (pnt + 3 > end) { | |
2015 | zlog_info("%s Capability length error", peer->host); | |
2016 | bgp_notify_send(peer, BGP_NOTIFY_CEASE, 0); | |
d8151687 | 2017 | return BGP_Stop; |
d62a17ae | 2018 | } |
2019 | action = *pnt; | |
2020 | hdr = (struct capability_header *)(pnt + 1); | |
2021 | ||
2022 | /* Action value check. */ | |
2023 | if (action != CAPABILITY_ACTION_SET | |
2024 | && action != CAPABILITY_ACTION_UNSET) { | |
2025 | zlog_info("%s Capability Action Value error %d", | |
2026 | peer->host, action); | |
2027 | bgp_notify_send(peer, BGP_NOTIFY_CEASE, 0); | |
d8151687 | 2028 | return BGP_Stop; |
d62a17ae | 2029 | } |
2030 | ||
2031 | if (bgp_debug_neighbor_events(peer)) | |
2032 | zlog_debug( | |
2033 | "%s CAPABILITY has action: %d, code: %u, length %u", | |
2034 | peer->host, action, hdr->code, hdr->length); | |
2035 | ||
2036 | /* Capability length check. */ | |
2037 | if ((pnt + hdr->length + 3) > end) { | |
2038 | zlog_info("%s Capability length error", peer->host); | |
2039 | bgp_notify_send(peer, BGP_NOTIFY_CEASE, 0); | |
d8151687 | 2040 | return BGP_Stop; |
d62a17ae | 2041 | } |
2042 | ||
2043 | /* Fetch structure to the byte stream. */ | |
2044 | memcpy(&mpc, pnt + 3, sizeof(struct capability_mp_data)); | |
d2b6417b | 2045 | pnt += hdr->length + 3; |
d62a17ae | 2046 | |
2047 | /* We know MP Capability Code. */ | |
2048 | if (hdr->code == CAPABILITY_CODE_MP) { | |
2049 | pkt_afi = ntohs(mpc.afi); | |
2050 | pkt_safi = mpc.safi; | |
2051 | ||
2052 | /* Ignore capability when override-capability is set. */ | |
2053 | if (CHECK_FLAG(peer->flags, | |
2054 | PEER_FLAG_OVERRIDE_CAPABILITY)) | |
2055 | continue; | |
2056 | ||
2057 | /* Convert AFI, SAFI to internal values. */ | |
2058 | if (bgp_map_afi_safi_iana2int(pkt_afi, pkt_safi, &afi, | |
2059 | &safi)) { | |
2060 | if (bgp_debug_neighbor_events(peer)) | |
2061 | zlog_debug( | |
2062 | "%s Dynamic Capability MP_EXT afi/safi invalid " | |
2063 | "(%u/%u)", | |
2064 | peer->host, pkt_afi, pkt_safi); | |
2065 | continue; | |
2066 | } | |
2067 | ||
2068 | /* Address family check. */ | |
2069 | if (bgp_debug_neighbor_events(peer)) | |
2070 | zlog_debug( | |
2071 | "%s CAPABILITY has %s MP_EXT CAP for afi/safi: %u/%u", | |
2072 | peer->host, | |
2073 | action == CAPABILITY_ACTION_SET | |
2074 | ? "Advertising" | |
2075 | : "Removing", | |
2076 | pkt_afi, pkt_safi); | |
2077 | ||
2078 | if (action == CAPABILITY_ACTION_SET) { | |
2079 | peer->afc_recv[afi][safi] = 1; | |
2080 | if (peer->afc[afi][safi]) { | |
2081 | peer->afc_nego[afi][safi] = 1; | |
2082 | bgp_announce_route(peer, afi, safi); | |
2083 | } | |
2084 | } else { | |
2085 | peer->afc_recv[afi][safi] = 0; | |
2086 | peer->afc_nego[afi][safi] = 0; | |
2087 | ||
2088 | if (peer_active_nego(peer)) | |
2089 | bgp_clear_route(peer, afi, safi); | |
2090 | else | |
d8151687 | 2091 | return BGP_Stop; |
d62a17ae | 2092 | } |
2093 | } else { | |
2094 | zlog_warn( | |
2095 | "%s unrecognized capability code: %d - ignored", | |
2096 | peer->host, hdr->code); | |
2097 | } | |
d62a17ae | 2098 | } |
d8151687 QY |
2099 | |
2100 | /* No FSM action necessary */ | |
2101 | return BGP_PACKET_NOOP; | |
718e3744 | 2102 | } |
2103 | ||
d8151687 QY |
2104 | /** |
2105 | * Parse BGP CAPABILITY message for peer. | |
01b7ce2d | 2106 | * |
d8151687 QY |
2107 | * Exported for unit testing. |
2108 | * | |
2109 | * @param peer | |
2110 | * @param size size of the packet | |
2111 | * @return as in summary | |
01b7ce2d | 2112 | */ |
d62a17ae | 2113 | int bgp_capability_receive(struct peer *peer, bgp_size_t size) |
718e3744 | 2114 | { |
d7c0a89a | 2115 | uint8_t *pnt; |
d62a17ae | 2116 | |
2117 | /* Fetch pointer. */ | |
424ab01d | 2118 | pnt = stream_pnt(peer->curr); |
d62a17ae | 2119 | |
2120 | if (bgp_debug_neighbor_events(peer)) | |
2121 | zlog_debug("%s rcv CAPABILITY", peer->host); | |
2122 | ||
2123 | /* If peer does not have the capability, send notification. */ | |
2124 | if (!CHECK_FLAG(peer->cap, PEER_CAP_DYNAMIC_ADV)) { | |
2125 | zlog_err("%s [Error] BGP dynamic capability is not enabled", | |
2126 | peer->host); | |
2127 | bgp_notify_send(peer, BGP_NOTIFY_HEADER_ERR, | |
2128 | BGP_NOTIFY_HEADER_BAD_MESTYPE); | |
d8151687 | 2129 | return BGP_Stop; |
d62a17ae | 2130 | } |
2131 | ||
2132 | /* Status must be Established. */ | |
2133 | if (peer->status != Established) { | |
2134 | zlog_err( | |
2135 | "%s [Error] Dynamic capability packet received under status %s", | |
2136 | peer->host, | |
2137 | lookup_msg(bgp_status_msg, peer->status, NULL)); | |
2138 | bgp_notify_send(peer, BGP_NOTIFY_FSM_ERR, 0); | |
d8151687 | 2139 | return BGP_Stop; |
d62a17ae | 2140 | } |
2141 | ||
2142 | /* Parse packet. */ | |
2143 | return bgp_capability_msg_parse(peer, pnt, size); | |
718e3744 | 2144 | } |
6b0655a2 | 2145 | |
d8151687 QY |
2146 | /** |
2147 | * Processes a peer's input buffer. | |
2148 | * | |
2149 | * This function sidesteps the event loop and directly calls bgp_event_update() | |
2150 | * after processing each BGP message. This is necessary to ensure proper | |
2151 | * ordering of FSM events and unifies the behavior that was present previously, | |
2152 | * whereby some of the packet handling functions would update the FSM and some | |
2153 | * would not, making event flow difficult to understand. Please think twice | |
2154 | * before hacking this. | |
2155 | * | |
2156 | * Thread type: THREAD_EVENT | |
2157 | * @param thread | |
2158 | * @return 0 | |
2159 | */ | |
424ab01d | 2160 | int bgp_process_packet(struct thread *thread) |
718e3744 | 2161 | { |
424ab01d | 2162 | /* Yes first of all get peer pointer. */ |
d8151687 QY |
2163 | struct peer *peer; // peer |
2164 | uint32_t rpkt_quanta_old; // how many packets to read | |
2165 | int fsm_update_result; // return code of bgp_event_update() | |
2166 | int mprc; // message processing return code | |
555e09d4 | 2167 | |
424ab01d | 2168 | peer = THREAD_ARG(thread); |
555e09d4 QY |
2169 | rpkt_quanta_old = atomic_load_explicit(&peer->bgp->rpkt_quanta, |
2170 | memory_order_relaxed); | |
e3c7270d | 2171 | fsm_update_result = 0; |
555e09d4 | 2172 | |
424ab01d | 2173 | /* Guard against scheduled events that occur after peer deletion. */ |
9eb217ff | 2174 | if (peer->status == Deleted || peer->status == Clearing) |
d62a17ae | 2175 | return 0; |
718e3744 | 2176 | |
555e09d4 | 2177 | unsigned int processed = 0; |
d62a17ae | 2178 | |
555e09d4 | 2179 | while (processed < rpkt_quanta_old) { |
d7c0a89a | 2180 | uint8_t type = 0; |
9eb217ff QY |
2181 | bgp_size_t size; |
2182 | char notify_data_length[2]; | |
d62a17ae | 2183 | |
9eb217ff QY |
2184 | pthread_mutex_lock(&peer->io_mtx); |
2185 | { | |
2186 | peer->curr = stream_fifo_pop(peer->ibuf); | |
2187 | } | |
2188 | pthread_mutex_unlock(&peer->io_mtx); | |
d62a17ae | 2189 | |
9eb217ff QY |
2190 | if (peer->curr == NULL) // no packets to process, hmm... |
2191 | return 0; | |
d62a17ae | 2192 | |
9eb217ff QY |
2193 | /* skip the marker and copy the packet length */ |
2194 | stream_forward_getp(peer->curr, BGP_MARKER_SIZE); | |
2195 | memcpy(notify_data_length, stream_pnt(peer->curr), 2); | |
2196 | ||
2197 | /* read in the packet length and type */ | |
2198 | size = stream_getw(peer->curr); | |
2199 | type = stream_getc(peer->curr); | |
2200 | ||
2201 | /* BGP packet dump function. */ | |
2202 | bgp_dump_packet(peer, type, peer->curr); | |
2203 | ||
2204 | /* adjust size to exclude the marker + length + type */ | |
2205 | size -= BGP_HEADER_SIZE; | |
2206 | ||
2207 | /* Read rest of the packet and call each sort of packet routine | |
2208 | */ | |
2209 | switch (type) { | |
2210 | case BGP_MSG_OPEN: | |
0112e9e0 QY |
2211 | atomic_fetch_add_explicit(&peer->open_in, 1, |
2212 | memory_order_relaxed); | |
d8151687 QY |
2213 | mprc = bgp_open_receive(peer, size); |
2214 | if (mprc == BGP_Stop) | |
2215 | zlog_err( | |
2216 | "%s: BGP OPEN receipt failed for peer: %s", | |
2217 | __FUNCTION__, peer->host); | |
9eb217ff QY |
2218 | break; |
2219 | case BGP_MSG_UPDATE: | |
0112e9e0 QY |
2220 | atomic_fetch_add_explicit(&peer->update_in, 1, |
2221 | memory_order_relaxed); | |
9eb217ff | 2222 | peer->readtime = monotime(NULL); |
d8151687 QY |
2223 | mprc = bgp_update_receive(peer, size); |
2224 | if (mprc == BGP_Stop) | |
2225 | zlog_err( | |
2226 | "%s: BGP UPDATE receipt failed for peer: %s", | |
2227 | __FUNCTION__, peer->host); | |
9eb217ff QY |
2228 | break; |
2229 | case BGP_MSG_NOTIFY: | |
0112e9e0 QY |
2230 | atomic_fetch_add_explicit(&peer->notify_in, 1, |
2231 | memory_order_relaxed); | |
d8151687 QY |
2232 | mprc = bgp_notify_receive(peer, size); |
2233 | if (mprc == BGP_Stop) | |
2234 | zlog_err( | |
2235 | "%s: BGP NOTIFY receipt failed for peer: %s", | |
2236 | __FUNCTION__, peer->host); | |
9eb217ff QY |
2237 | break; |
2238 | case BGP_MSG_KEEPALIVE: | |
2239 | peer->readtime = monotime(NULL); | |
0112e9e0 QY |
2240 | atomic_fetch_add_explicit(&peer->keepalive_in, 1, |
2241 | memory_order_relaxed); | |
d8151687 QY |
2242 | mprc = bgp_keepalive_receive(peer, size); |
2243 | if (mprc == BGP_Stop) | |
2244 | zlog_err( | |
2245 | "%s: BGP KEEPALIVE receipt failed for peer: %s", | |
2246 | __FUNCTION__, peer->host); | |
9eb217ff QY |
2247 | break; |
2248 | case BGP_MSG_ROUTE_REFRESH_NEW: | |
2249 | case BGP_MSG_ROUTE_REFRESH_OLD: | |
0112e9e0 QY |
2250 | atomic_fetch_add_explicit(&peer->refresh_in, 1, |
2251 | memory_order_relaxed); | |
d8151687 QY |
2252 | mprc = bgp_route_refresh_receive(peer, size); |
2253 | if (mprc == BGP_Stop) | |
2254 | zlog_err( | |
2255 | "%s: BGP ROUTEREFRESH receipt failed for peer: %s", | |
2256 | __FUNCTION__, peer->host); | |
9eb217ff QY |
2257 | break; |
2258 | case BGP_MSG_CAPABILITY: | |
0112e9e0 QY |
2259 | atomic_fetch_add_explicit(&peer->dynamic_cap_in, 1, |
2260 | memory_order_relaxed); | |
d8151687 QY |
2261 | mprc = bgp_capability_receive(peer, size); |
2262 | if (mprc == BGP_Stop) | |
2263 | zlog_err( | |
2264 | "%s: BGP CAPABILITY receipt failed for peer: %s", | |
2265 | __FUNCTION__, peer->host); | |
9eb217ff | 2266 | break; |
e3c7270d | 2267 | default: |
becedef6 QY |
2268 | /* |
2269 | * The message type should have been sanitized before | |
2270 | * we ever got here. Receipt of a message with an | |
2271 | * invalid header at this point is indicative of a | |
2272 | * security issue. | |
2273 | */ | |
e3c7270d | 2274 | assert (!"Message of invalid type received during input processing"); |
9eb217ff QY |
2275 | } |
2276 | ||
d8151687 QY |
2277 | /* delete processed packet */ |
2278 | stream_free(peer->curr); | |
2279 | peer->curr = NULL; | |
2280 | processed++; | |
9eb217ff | 2281 | |
d8151687 QY |
2282 | /* Update FSM */ |
2283 | if (mprc != BGP_PACKET_NOOP) | |
2284 | fsm_update_result = bgp_event_update(peer, mprc); | |
e3c7270d QY |
2285 | else |
2286 | continue; | |
d8151687 | 2287 | |
becedef6 QY |
2288 | /* |
2289 | * If peer was deleted, do not process any more packets. This | |
2290 | * is usually due to executing BGP_Stop or a stub deletion. | |
2291 | */ | |
d8151687 QY |
2292 | if (fsm_update_result == FSM_PEER_TRANSFERRED |
2293 | || fsm_update_result == FSM_PEER_STOPPED) | |
2294 | break; | |
d62a17ae | 2295 | } |
2296 | ||
d8151687 QY |
2297 | if (fsm_update_result != FSM_PEER_TRANSFERRED |
2298 | && fsm_update_result != FSM_PEER_STOPPED) { | |
2299 | pthread_mutex_lock(&peer->io_mtx); | |
2300 | { | |
becedef6 QY |
2301 | // more work to do, come back later |
2302 | if (peer->ibuf->count > 0) | |
4af76660 QY |
2303 | thread_add_timer_msec( |
2304 | bm->master, bgp_process_packet, peer, 0, | |
2305 | &peer->t_process_packet); | |
d8151687 QY |
2306 | } |
2307 | pthread_mutex_unlock(&peer->io_mtx); | |
718e3744 | 2308 | } |
2309 | ||
d62a17ae | 2310 | return 0; |
718e3744 | 2311 | } |