]>
Commit | Line | Data |
---|---|---|
acddc0ed | 1 | // SPDX-License-Identifier: GPL-2.0-or-later |
718e3744 | 2 | /* BGP packet management routine. |
56257a44 QY |
3 | * Contains utility functions for constructing and consuming BGP messages. |
4 | * Copyright (C) 2017 Cumulus Networks | |
896014f4 | 5 | * Copyright (C) 1999 Kunihiro Ishiguro |
896014f4 | 6 | */ |
718e3744 | 7 | |
8 | #include <zebra.h> | |
d3ecc69e | 9 | #include <sys/time.h> |
718e3744 | 10 | |
11 | #include "thread.h" | |
12 | #include "stream.h" | |
13 | #include "network.h" | |
14 | #include "prefix.h" | |
15 | #include "command.h" | |
16 | #include "log.h" | |
17 | #include "memory.h" | |
d62a17ae | 18 | #include "sockunion.h" /* for inet_ntop () */ |
baa376fc | 19 | #include "sockopt.h" |
718e3744 | 20 | #include "linklist.h" |
21 | #include "plist.h" | |
3f9c7369 | 22 | #include "queue.h" |
039f3a34 | 23 | #include "filter.h" |
02705213 | 24 | #include "lib_errors.h" |
718e3744 | 25 | |
26 | #include "bgpd/bgpd.h" | |
27 | #include "bgpd/bgp_table.h" | |
28 | #include "bgpd/bgp_dump.h" | |
6c29258c | 29 | #include "bgpd/bgp_bmp.h" |
718e3744 | 30 | #include "bgpd/bgp_attr.h" |
31 | #include "bgpd/bgp_debug.h" | |
14454c9f | 32 | #include "bgpd/bgp_errors.h" |
718e3744 | 33 | #include "bgpd/bgp_fsm.h" |
34 | #include "bgpd/bgp_route.h" | |
35 | #include "bgpd/bgp_packet.h" | |
36 | #include "bgpd/bgp_open.h" | |
37 | #include "bgpd/bgp_aspath.h" | |
38 | #include "bgpd/bgp_community.h" | |
39 | #include "bgpd/bgp_ecommunity.h" | |
57d187bc | 40 | #include "bgpd/bgp_lcommunity.h" |
718e3744 | 41 | #include "bgpd/bgp_network.h" |
42 | #include "bgpd/bgp_mplsvpn.h" | |
7ef5a232 | 43 | #include "bgpd/bgp_evpn.h" |
718e3744 | 44 | #include "bgpd/bgp_advertise.h" |
93406d87 | 45 | #include "bgpd/bgp_vty.h" |
3f9c7369 | 46 | #include "bgpd/bgp_updgrp.h" |
cd1964ff | 47 | #include "bgpd/bgp_label.h" |
56257a44 | 48 | #include "bgpd/bgp_io.h" |
934af458 | 49 | #include "bgpd/bgp_keepalives.h" |
7c40bf39 | 50 | #include "bgpd/bgp_flowspec.h" |
d9a03c57 | 51 | #include "bgpd/bgp_trace.h" |
718e3744 | 52 | |
584470fb DL |
53 | DEFINE_HOOK(bgp_packet_dump, |
54 | (struct peer *peer, uint8_t type, bgp_size_t size, | |
55 | struct stream *s), | |
8451921b | 56 | (peer, type, size, s)); |
584470fb | 57 | |
6fd04594 DL |
58 | DEFINE_HOOK(bgp_packet_send, |
59 | (struct peer *peer, uint8_t type, bgp_size_t size, | |
60 | struct stream *s), | |
8451921b | 61 | (peer, type, size, s)); |
6fd04594 | 62 | |
d8151687 QY |
63 | /** |
64 | * Sets marker and type fields for a BGP message. | |
65 | * | |
66 | * @param s the stream containing the packet | |
67 | * @param type the packet type | |
68 | * @return the size of the stream | |
69 | */ | |
d7c0a89a | 70 | int bgp_packet_set_marker(struct stream *s, uint8_t type) |
718e3744 | 71 | { |
d62a17ae | 72 | int i; |
718e3744 | 73 | |
d62a17ae | 74 | /* Fill in marker. */ |
75 | for (i = 0; i < BGP_MARKER_SIZE; i++) | |
76 | stream_putc(s, 0xff); | |
718e3744 | 77 | |
d62a17ae | 78 | /* Dummy total length. This field is should be filled in later on. */ |
79 | stream_putw(s, 0); | |
718e3744 | 80 | |
d62a17ae | 81 | /* BGP packet type. */ |
82 | stream_putc(s, type); | |
718e3744 | 83 | |
d62a17ae | 84 | /* Return current stream size. */ |
85 | return stream_get_endp(s); | |
718e3744 | 86 | } |
87 | ||
d8151687 QY |
88 | /** |
89 | * Sets size field for a BGP message. | |
90 | * | |
91 | * Size field is set to the size of the stream passed. | |
92 | * | |
93 | * @param s the stream containing the packet | |
d8151687 | 94 | */ |
65baedca | 95 | void bgp_packet_set_size(struct stream *s) |
718e3744 | 96 | { |
d62a17ae | 97 | int cp; |
718e3744 | 98 | |
d62a17ae | 99 | /* Preserve current pointer. */ |
100 | cp = stream_get_endp(s); | |
101 | stream_putw_at(s, BGP_MARKER_SIZE, cp); | |
718e3744 | 102 | } |
103 | ||
d3ecc69e QY |
104 | /* |
105 | * Push a packet onto the beginning of the peer's output queue. | |
106 | * This function acquires the peer's write mutex before proceeding. | |
107 | */ | |
108 | static void bgp_packet_add(struct peer *peer, struct stream *s) | |
109 | { | |
bd9fb6f3 DL |
110 | intmax_t delta; |
111 | uint32_t holdtime; | |
460ed839 | 112 | intmax_t sendholdtime; |
bd9fb6f3 | 113 | |
cb1991af | 114 | frr_with_mutex (&peer->io_mtx) { |
bd9fb6f3 DL |
115 | /* if the queue is empty, reset the "last OK" timestamp to |
116 | * now, otherwise if we write another packet immediately | |
117 | * after it'll get confused | |
118 | */ | |
119 | if (!stream_fifo_count_safe(peer->obuf)) | |
083ec940 | 120 | peer->last_sendq_ok = monotime(NULL); |
bd9fb6f3 | 121 | |
00dffa8c | 122 | stream_fifo_push(peer->obuf, s); |
bd9fb6f3 | 123 | |
083ec940 | 124 | delta = monotime(NULL) - peer->last_sendq_ok; |
460ed839 DA |
125 | |
126 | if (CHECK_FLAG(peer->flags, PEER_FLAG_TIMER)) | |
127 | holdtime = atomic_load_explicit(&peer->holdtime, | |
128 | memory_order_relaxed); | |
129 | else | |
130 | holdtime = peer->bgp->default_holdtime; | |
131 | ||
132 | sendholdtime = holdtime * 2; | |
bd9fb6f3 DL |
133 | |
134 | /* Note that when we're here, we're adding some packet to the | |
135 | * OutQ. That includes keepalives when there is nothing to | |
136 | * do, so there's a guarantee we pass by here once in a while. | |
137 | * | |
138 | * That implies there is no need to go set up another separate | |
139 | * timer that ticks down SendHoldTime, as we'll be here sooner | |
140 | * or later anyway and will see the checks below failing. | |
141 | */ | |
382268f4 DL |
142 | if (!holdtime) { |
143 | /* no holdtime, do nothing. */ | |
460ed839 | 144 | } else if (delta > sendholdtime) { |
bd9fb6f3 DL |
145 | flog_err( |
146 | EC_BGP_SENDQ_STUCK_PROPER, | |
460ed839 DA |
147 | "%pBP has not made any SendQ progress for 2 holdtimes (%jds), terminating session", |
148 | peer, sendholdtime); | |
bd9fb6f3 DL |
149 | BGP_EVENT_ADD(peer, TCP_fatal_error); |
150 | } else if (delta > (intmax_t)holdtime && | |
083ec940 | 151 | monotime(NULL) - peer->last_sendq_warn > 5) { |
bd9fb6f3 DL |
152 | flog_warn( |
153 | EC_BGP_SENDQ_STUCK_WARN, | |
460ed839 DA |
154 | "%pBP has not made any SendQ progress for 1 holdtime (%us), peer overloaded?", |
155 | peer, holdtime); | |
083ec940 | 156 | peer->last_sendq_warn = monotime(NULL); |
bd9fb6f3 | 157 | } |
00dffa8c | 158 | } |
718e3744 | 159 | } |
160 | ||
d62a17ae | 161 | static struct stream *bgp_update_packet_eor(struct peer *peer, afi_t afi, |
162 | safi_t safi) | |
93406d87 | 163 | { |
d62a17ae | 164 | struct stream *s; |
617975d1 DS |
165 | iana_afi_t pkt_afi = IANA_AFI_IPV4; |
166 | iana_safi_t pkt_safi = IANA_SAFI_UNICAST; | |
d62a17ae | 167 | |
168 | if (DISABLE_BGP_ANNOUNCE) | |
169 | return NULL; | |
170 | ||
171 | if (bgp_debug_neighbor_events(peer)) | |
172 | zlog_debug("send End-of-RIB for %s to %s", | |
5cb5f4d0 | 173 | get_afi_safi_str(afi, safi, false), peer->host); |
d62a17ae | 174 | |
ef56aee4 | 175 | s = stream_new(peer->max_packet_size); |
d62a17ae | 176 | |
177 | /* Make BGP update packet. */ | |
178 | bgp_packet_set_marker(s, BGP_MSG_UPDATE); | |
179 | ||
180 | /* Unfeasible Routes Length */ | |
181 | stream_putw(s, 0); | |
182 | ||
183 | if (afi == AFI_IP && safi == SAFI_UNICAST) { | |
184 | /* Total Path Attribute Length */ | |
185 | stream_putw(s, 0); | |
186 | } else { | |
187 | /* Convert AFI, SAFI to values for packet. */ | |
188 | bgp_map_afi_safi_int2iana(afi, safi, &pkt_afi, &pkt_safi); | |
189 | ||
190 | /* Total Path Attribute Length */ | |
191 | stream_putw(s, 6); | |
192 | stream_putc(s, BGP_ATTR_FLAG_OPTIONAL); | |
193 | stream_putc(s, BGP_ATTR_MP_UNREACH_NLRI); | |
194 | stream_putc(s, 3); | |
195 | stream_putw(s, pkt_afi); | |
196 | stream_putc(s, pkt_safi); | |
197 | } | |
198 | ||
199 | bgp_packet_set_size(s); | |
d62a17ae | 200 | return s; |
718e3744 | 201 | } |
202 | ||
d8151687 QY |
203 | /* Called when there is a change in the EOR(implicit or explicit) status of a |
204 | * peer. Ends the update-delay if all expected peers are done with EORs. */ | |
205 | void bgp_check_update_delay(struct bgp *bgp) | |
206 | { | |
207 | struct listnode *node, *nnode; | |
208 | struct peer *peer = NULL; | |
209 | ||
210 | if (bgp_debug_neighbor_events(peer)) | |
211 | zlog_debug("Checking update delay, T: %d R: %d I:%d E: %d", | |
212 | bgp->established, bgp->restarted_peers, | |
213 | bgp->implicit_eors, bgp->explicit_eors); | |
214 | ||
215 | if (bgp->established | |
216 | <= bgp->restarted_peers + bgp->implicit_eors + bgp->explicit_eors) { | |
becedef6 QY |
217 | /* |
218 | * This is an extra sanity check to make sure we wait for all | |
219 | * the eligible configured peers. This check is performed if | |
220 | * establish wait timer is on, or establish wait option is not | |
221 | * given with the update-delay command | |
222 | */ | |
d8151687 QY |
223 | if (bgp->t_establish_wait |
224 | || (bgp->v_establish_wait == bgp->v_update_delay)) | |
225 | for (ALL_LIST_ELEMENTS(bgp->peer, node, nnode, peer)) { | |
226 | if (CHECK_FLAG(peer->flags, | |
227 | PEER_FLAG_CONFIG_NODE) | |
228 | && !CHECK_FLAG(peer->flags, | |
229 | PEER_FLAG_SHUTDOWN) | |
cb9196e7 DS |
230 | && !CHECK_FLAG(peer->bgp->flags, |
231 | BGP_FLAG_SHUTDOWN) | |
d8151687 QY |
232 | && !peer->update_delay_over) { |
233 | if (bgp_debug_neighbor_events(peer)) | |
234 | zlog_debug( | |
235 | " Peer %s pending, continuing read-only mode", | |
236 | peer->host); | |
237 | return; | |
238 | } | |
239 | } | |
240 | ||
241 | zlog_info( | |
242 | "Update delay ended, restarted: %d, EORs implicit: %d, explicit: %d", | |
243 | bgp->restarted_peers, bgp->implicit_eors, | |
244 | bgp->explicit_eors); | |
245 | bgp_update_delay_end(bgp); | |
246 | } | |
247 | } | |
248 | ||
becedef6 QY |
249 | /* |
250 | * Called if peer is known to have restarted. The restart-state bit in | |
251 | * Graceful-Restart capability is used for that | |
252 | */ | |
d8151687 QY |
253 | void bgp_update_restarted_peers(struct peer *peer) |
254 | { | |
255 | if (!bgp_update_delay_active(peer->bgp)) | |
256 | return; /* BGP update delay has ended */ | |
257 | if (peer->update_delay_over) | |
258 | return; /* This peer has already been considered */ | |
259 | ||
260 | if (bgp_debug_neighbor_events(peer)) | |
261 | zlog_debug("Peer %s: Checking restarted", peer->host); | |
262 | ||
feb17238 | 263 | if (peer_established(peer)) { |
d8151687 QY |
264 | peer->update_delay_over = 1; |
265 | peer->bgp->restarted_peers++; | |
266 | bgp_check_update_delay(peer->bgp); | |
267 | } | |
268 | } | |
269 | ||
becedef6 QY |
270 | /* |
271 | * Called as peer receives a keep-alive. Determines if this occurence can be | |
272 | * taken as an implicit EOR for this peer. | |
273 | * NOTE: The very first keep-alive after the Established state of a peer is | |
274 | * considered implicit EOR for the update-delay purposes | |
275 | */ | |
d8151687 QY |
276 | void bgp_update_implicit_eors(struct peer *peer) |
277 | { | |
278 | if (!bgp_update_delay_active(peer->bgp)) | |
279 | return; /* BGP update delay has ended */ | |
280 | if (peer->update_delay_over) | |
281 | return; /* This peer has already been considered */ | |
282 | ||
283 | if (bgp_debug_neighbor_events(peer)) | |
284 | zlog_debug("Peer %s: Checking implicit EORs", peer->host); | |
285 | ||
feb17238 | 286 | if (peer_established(peer)) { |
d8151687 QY |
287 | peer->update_delay_over = 1; |
288 | peer->bgp->implicit_eors++; | |
289 | bgp_check_update_delay(peer->bgp); | |
290 | } | |
291 | } | |
292 | ||
becedef6 QY |
293 | /* |
294 | * Should be called only when there is a change in the EOR_RECEIVED status | |
295 | * for any afi/safi on a peer. | |
296 | */ | |
d8151687 QY |
297 | static void bgp_update_explicit_eors(struct peer *peer) |
298 | { | |
299 | afi_t afi; | |
300 | safi_t safi; | |
301 | ||
302 | if (!bgp_update_delay_active(peer->bgp)) | |
303 | return; /* BGP update delay has ended */ | |
304 | if (peer->update_delay_over) | |
305 | return; /* This peer has already been considered */ | |
306 | ||
307 | if (bgp_debug_neighbor_events(peer)) | |
308 | zlog_debug("Peer %s: Checking explicit EORs", peer->host); | |
309 | ||
f18ba3cd DS |
310 | FOREACH_AFI_SAFI (afi, safi) { |
311 | if (peer->afc_nego[afi][safi] | |
312 | && !CHECK_FLAG(peer->af_sflags[afi][safi], | |
313 | PEER_STATUS_EOR_RECEIVED)) { | |
314 | if (bgp_debug_neighbor_events(peer)) | |
315 | zlog_debug( | |
316 | " afi %d safi %d didn't receive EOR", | |
317 | afi, safi); | |
318 | return; | |
d8151687 | 319 | } |
f18ba3cd | 320 | } |
d8151687 QY |
321 | |
322 | peer->update_delay_over = 1; | |
323 | peer->bgp->explicit_eors++; | |
324 | bgp_check_update_delay(peer->bgp); | |
325 | } | |
326 | ||
327 | /** | |
328 | * Frontend for NLRI parsing, to fan-out to AFI/SAFI specific parsers. | |
329 | * | |
330 | * mp_withdraw, if set, is used to nullify attr structure on most of the | |
331 | * calling safi function and for evpn, passed as parameter | |
332 | */ | |
333 | int bgp_nlri_parse(struct peer *peer, struct attr *attr, | |
334 | struct bgp_nlri *packet, int mp_withdraw) | |
335 | { | |
336 | switch (packet->safi) { | |
337 | case SAFI_UNICAST: | |
338 | case SAFI_MULTICAST: | |
339 | return bgp_nlri_parse_ip(peer, mp_withdraw ? NULL : attr, | |
340 | packet); | |
341 | case SAFI_LABELED_UNICAST: | |
342 | return bgp_nlri_parse_label(peer, mp_withdraw ? NULL : attr, | |
343 | packet); | |
344 | case SAFI_MPLS_VPN: | |
345 | return bgp_nlri_parse_vpn(peer, mp_withdraw ? NULL : attr, | |
346 | packet); | |
347 | case SAFI_EVPN: | |
348 | return bgp_nlri_parse_evpn(peer, attr, packet, mp_withdraw); | |
7c40bf39 | 349 | case SAFI_FLOWSPEC: |
350 | return bgp_nlri_parse_flowspec(peer, attr, packet, mp_withdraw); | |
d8151687 | 351 | } |
513386b5 | 352 | return BGP_NLRI_PARSE_ERROR; |
d8151687 QY |
353 | } |
354 | ||
a783cc05 XL |
355 | |
356 | /* | |
357 | * Check if route-refresh request from peer is pending (received before EoR), | |
358 | * and process it now. | |
359 | */ | |
360 | static void bgp_process_pending_refresh(struct peer *peer, afi_t afi, | |
361 | safi_t safi) | |
362 | { | |
363 | if (CHECK_FLAG(peer->af_sflags[afi][safi], | |
364 | PEER_STATUS_REFRESH_PENDING)) { | |
365 | UNSET_FLAG(peer->af_sflags[afi][safi], | |
366 | PEER_STATUS_REFRESH_PENDING); | |
367 | bgp_route_refresh_send(peer, afi, safi, 0, 0, 0, | |
368 | BGP_ROUTE_REFRESH_BORR); | |
369 | if (bgp_debug_neighbor_events(peer)) | |
370 | zlog_debug( | |
371 | "%pBP sending route-refresh (BoRR) for %s/%s (for pending REQUEST)", | |
372 | peer, afi2str(afi), safi2str(safi)); | |
373 | ||
374 | SET_FLAG(peer->af_sflags[afi][safi], PEER_STATUS_BORR_SEND); | |
375 | UNSET_FLAG(peer->af_sflags[afi][safi], PEER_STATUS_EORR_SEND); | |
376 | bgp_announce_route(peer, afi, safi, true); | |
377 | } | |
378 | } | |
379 | ||
6ec98a2f QY |
380 | /* |
381 | * Checks a variety of conditions to determine whether the peer needs to be | |
382 | * rescheduled for packet generation again, and does so if necessary. | |
383 | * | |
384 | * @param peer to check for rescheduling | |
385 | */ | |
af1e1dc6 QY |
386 | static void bgp_write_proceed_actions(struct peer *peer) |
387 | { | |
388 | afi_t afi; | |
389 | safi_t safi; | |
390 | struct peer_af *paf; | |
391 | struct bpacket *next_pkt; | |
392 | struct update_subgroup *subgrp; | |
0e5cdd59 | 393 | enum bgp_af_index index; |
af1e1dc6 | 394 | |
0e5cdd59 DS |
395 | for (index = BGP_AF_START; index < BGP_AF_MAX; index++) { |
396 | paf = peer->peer_af_array[index]; | |
c58b0f46 QY |
397 | if (!paf) |
398 | continue; | |
0e5cdd59 | 399 | |
c58b0f46 QY |
400 | subgrp = paf->subgroup; |
401 | if (!subgrp) | |
402 | continue; | |
403 | ||
404 | next_pkt = paf->next_pkt_to_send; | |
405 | if (next_pkt && next_pkt->buffer) { | |
406 | BGP_TIMER_ON(peer->t_generate_updgrp_packets, | |
407 | bgp_generate_updgrp_packets, 0); | |
408 | return; | |
409 | } | |
af1e1dc6 | 410 | |
c58b0f46 QY |
411 | /* No packets readily available for AFI/SAFI, are there |
412 | * subgroup packets | |
413 | * that need to be generated? */ | |
414 | if (bpacket_queue_is_full(SUBGRP_INST(subgrp), | |
415 | SUBGRP_PKTQ(subgrp)) | |
416 | || subgroup_packets_to_build(subgrp)) { | |
417 | BGP_TIMER_ON(peer->t_generate_updgrp_packets, | |
418 | bgp_generate_updgrp_packets, 0); | |
419 | return; | |
420 | } | |
af1e1dc6 | 421 | |
0e5cdd59 DS |
422 | afi = paf->afi; |
423 | safi = paf->safi; | |
424 | ||
c58b0f46 QY |
425 | /* No packets to send, see if EOR is pending */ |
426 | if (CHECK_FLAG(peer->cap, PEER_CAP_RESTART_RCV)) { | |
427 | if (!subgrp->t_coalesce && peer->afc_nego[afi][safi] | |
428 | && peer->synctime | |
429 | && !CHECK_FLAG(peer->af_sflags[afi][safi], | |
430 | PEER_STATUS_EOR_SEND) | |
431 | && safi != SAFI_MPLS_VPN) { | |
af1e1dc6 QY |
432 | BGP_TIMER_ON(peer->t_generate_updgrp_packets, |
433 | bgp_generate_updgrp_packets, 0); | |
434 | return; | |
435 | } | |
af1e1dc6 | 436 | } |
c58b0f46 | 437 | } |
af1e1dc6 QY |
438 | } |
439 | ||
6ec98a2f QY |
440 | /* |
441 | * Generate advertisement information (withdraws, updates, EOR) from each | |
442 | * update group a peer belongs to, encode this information into packets, and | |
443 | * enqueue the packets onto the peer's output buffer. | |
56257a44 | 444 | */ |
cc9f21da | 445 | void bgp_generate_updgrp_packets(struct thread *thread) |
718e3744 | 446 | { |
56257a44 | 447 | struct peer *peer = THREAD_ARG(thread); |
56257a44 QY |
448 | |
449 | struct stream *s; | |
d62a17ae | 450 | struct peer_af *paf; |
451 | struct bpacket *next_pkt; | |
d0ad6d8e QY |
452 | uint32_t wpq; |
453 | uint32_t generated = 0; | |
d62a17ae | 454 | afi_t afi; |
455 | safi_t safi; | |
456 | ||
d0ad6d8e QY |
457 | wpq = atomic_load_explicit(&peer->bgp->wpkt_quanta, |
458 | memory_order_relaxed); | |
459 | ||
d62a17ae | 460 | /* |
461 | * The code beyond this part deals with update packets, proceed only | |
462 | * if peer is Established and updates are not on hold (as part of | |
f4d2dd84 | 463 | * update-delay processing). |
3f9c7369 | 464 | */ |
feb17238 | 465 | if (!peer_established(peer)) |
cc9f21da | 466 | return; |
d62a17ae | 467 | |
f4d2dd84 DS |
468 | if ((peer->bgp->main_peers_update_hold) |
469 | || bgp_update_delay_active(peer->bgp)) | |
cc9f21da | 470 | return; |
d62a17ae | 471 | |
b10b6d52 | 472 | if (peer->t_routeadv) |
cc9f21da | 473 | return; |
b10b6d52 | 474 | |
963b7ee4 DS |
475 | /* |
476 | * Since the following is a do while loop | |
477 | * let's stop adding to the outq if we are | |
478 | * already at the limit. | |
479 | */ | |
480 | if (peer->obuf->count >= bm->outq_limit) { | |
481 | bgp_write_proceed_actions(peer); | |
482 | return; | |
483 | } | |
484 | ||
56257a44 | 485 | do { |
0e5cdd59 DS |
486 | enum bgp_af_index index; |
487 | ||
56257a44 | 488 | s = NULL; |
0e5cdd59 DS |
489 | for (index = BGP_AF_START; index < BGP_AF_MAX; index++) { |
490 | paf = peer->peer_af_array[index]; | |
c58b0f46 QY |
491 | if (!paf || !PAF_SUBGRP(paf)) |
492 | continue; | |
0e5cdd59 DS |
493 | |
494 | afi = paf->afi; | |
495 | safi = paf->safi; | |
c58b0f46 | 496 | next_pkt = paf->next_pkt_to_send; |
80bd61c4 | 497 | |
c58b0f46 QY |
498 | /* |
499 | * Try to generate a packet for the peer if we are at | |
500 | * the end of the list. Always try to push out | |
501 | * WITHDRAWs first. | |
502 | */ | |
503 | if (!next_pkt || !next_pkt->buffer) { | |
504 | next_pkt = subgroup_withdraw_packet( | |
505 | PAF_SUBGRP(paf)); | |
506 | if (!next_pkt || !next_pkt->buffer) | |
507 | subgroup_update_packet(PAF_SUBGRP(paf)); | |
508 | next_pkt = paf->next_pkt_to_send; | |
509 | } | |
80bd61c4 | 510 | |
c58b0f46 QY |
511 | /* |
512 | * If we still don't have a packet to send to the peer, | |
513 | * then try to find out out if we have to send eor or | |
514 | * if not, skip to the next AFI, SAFI. Don't send the | |
515 | * EOR prematurely; if the subgroup's coalesce timer is | |
516 | * running, the adjacency-out structure is not created | |
517 | * yet. | |
518 | */ | |
519 | if (!next_pkt || !next_pkt->buffer) { | |
9af52ccf DA |
520 | if (!paf->t_announce_route) { |
521 | /* Make sure we supress BGP UPDATES | |
522 | * for normal processing later again. | |
523 | */ | |
2adac256 DA |
524 | UNSET_FLAG(paf->subgroup->sflags, |
525 | SUBGRP_STATUS_FORCE_UPDATES); | |
526 | ||
9af52ccf DA |
527 | /* If route-refresh BoRR message was |
528 | * already sent and we are done with | |
529 | * re-announcing tables for a decent | |
530 | * afi/safi, we ready to send | |
531 | * EoRR request. | |
532 | */ | |
533 | if (CHECK_FLAG( | |
534 | peer->af_sflags[afi][safi], | |
535 | PEER_STATUS_BORR_SEND)) { | |
536 | bgp_route_refresh_send( | |
537 | peer, afi, safi, 0, 0, | |
538 | 0, | |
539 | BGP_ROUTE_REFRESH_EORR); | |
540 | ||
541 | SET_FLAG(peer->af_sflags[afi] | |
542 | [safi], | |
543 | PEER_STATUS_EORR_SEND); | |
544 | UNSET_FLAG( | |
545 | peer->af_sflags[afi] | |
546 | [safi], | |
547 | PEER_STATUS_BORR_SEND); | |
548 | ||
549 | if (bgp_debug_neighbor_events( | |
550 | peer)) | |
551 | zlog_debug( | |
f70c91dc DA |
552 | "%pBP sending route-refresh (EoRR) for %s/%s", |
553 | peer, | |
9af52ccf DA |
554 | afi2str(afi), |
555 | safi2str(safi)); | |
556 | } | |
557 | } | |
558 | ||
c58b0f46 | 559 | if (CHECK_FLAG(peer->cap, |
36235319 | 560 | PEER_CAP_RESTART_RCV)) { |
c58b0f46 | 561 | if (!(PAF_SUBGRP(paf))->t_coalesce |
36235319 QY |
562 | && peer->afc_nego[afi][safi] |
563 | && peer->synctime | |
564 | && !CHECK_FLAG( | |
565 | peer->af_sflags[afi][safi], | |
566 | PEER_STATUS_EOR_SEND)) { | |
d6e3c15b | 567 | /* If EOR is disabled, |
568 | * the message is not sent | |
569 | */ | |
36235319 QY |
570 | if (BGP_SEND_EOR(peer->bgp, afi, |
571 | safi)) { | |
d6e3c15b | 572 | SET_FLAG( |
36235319 QY |
573 | peer->af_sflags |
574 | [afi] | |
575 | [safi], | |
576 | PEER_STATUS_EOR_SEND); | |
d6e3c15b | 577 | |
9e3b51a7 | 578 | /* Update EOR |
579 | * send time | |
580 | */ | |
36235319 QY |
581 | peer->eor_stime[afi] |
582 | [safi] = | |
583 | monotime(NULL); | |
9e3b51a7 | 584 | |
d6e3c15b | 585 | BGP_UPDATE_EOR_PKT( |
36235319 QY |
586 | peer, afi, safi, |
587 | s); | |
a783cc05 XL |
588 | bgp_process_pending_refresh( |
589 | peer, afi, | |
590 | safi); | |
56257a44 | 591 | } |
80bd61c4 | 592 | } |
d62a17ae | 593 | } |
c58b0f46 QY |
594 | continue; |
595 | } | |
9e3b51a7 | 596 | |
597 | /* Update packet send time */ | |
598 | peer->pkt_stime[afi][safi] = monotime(NULL); | |
599 | ||
c58b0f46 QY |
600 | /* Found a packet template to send, overwrite |
601 | * packet with appropriate attributes from peer | |
602 | * and advance peer */ | |
603 | s = bpacket_reformat_for_peer(next_pkt, paf); | |
604 | bgp_packet_add(peer, s); | |
c58b0f46 QY |
605 | bpacket_queue_advance_peer(paf); |
606 | } | |
963b7ee4 DS |
607 | } while (s && (++generated < wpq) && |
608 | (peer->obuf->count <= bm->outq_limit)); | |
80bd61c4 | 609 | |
6ec98a2f QY |
610 | if (generated) |
611 | bgp_writes_on(peer); | |
612 | ||
af1e1dc6 | 613 | bgp_write_proceed_actions(peer); |
80bd61c4 QY |
614 | } |
615 | ||
d3ecc69e QY |
616 | /* |
617 | * Creates a BGP Keepalive packet and appends it to the peer's output queue. | |
618 | */ | |
d62a17ae | 619 | void bgp_keepalive_send(struct peer *peer) |
718e3744 | 620 | { |
d62a17ae | 621 | struct stream *s; |
622 | ||
556beacf | 623 | s = stream_new(BGP_STANDARD_MESSAGE_MAX_PACKET_SIZE); |
718e3744 | 624 | |
d62a17ae | 625 | /* Make keepalive packet. */ |
626 | bgp_packet_set_marker(s, BGP_MSG_KEEPALIVE); | |
718e3744 | 627 | |
d62a17ae | 628 | /* Set packet size. */ |
65baedca | 629 | bgp_packet_set_size(s); |
718e3744 | 630 | |
d62a17ae | 631 | /* Dump packet if debug option is set. */ |
632 | /* bgp_packet_dump (s); */ | |
718e3744 | 633 | |
d62a17ae | 634 | if (bgp_debug_keepalive(peer)) |
635 | zlog_debug("%s sending KEEPALIVE", peer->host); | |
718e3744 | 636 | |
d62a17ae | 637 | /* Add packet to the peer. */ |
638 | bgp_packet_add(peer, s); | |
424ab01d QY |
639 | |
640 | bgp_writes_on(peer); | |
718e3744 | 641 | } |
642 | ||
d3ecc69e QY |
643 | /* |
644 | * Creates a BGP Open packet and appends it to the peer's output queue. | |
645 | * Sets capabilities as necessary. | |
646 | */ | |
d62a17ae | 647 | void bgp_open_send(struct peer *peer) |
718e3744 | 648 | { |
d62a17ae | 649 | struct stream *s; |
d7c0a89a | 650 | uint16_t send_holdtime; |
d62a17ae | 651 | as_t local_as; |
718e3744 | 652 | |
b90a8e13 | 653 | if (CHECK_FLAG(peer->flags, PEER_FLAG_TIMER)) |
d62a17ae | 654 | send_holdtime = peer->holdtime; |
655 | else | |
656 | send_holdtime = peer->bgp->default_holdtime; | |
718e3744 | 657 | |
d62a17ae | 658 | /* local-as Change */ |
659 | if (peer->change_local_as) | |
660 | local_as = peer->change_local_as; | |
661 | else | |
662 | local_as = peer->local_as; | |
718e3744 | 663 | |
556beacf | 664 | s = stream_new(BGP_STANDARD_MESSAGE_MAX_PACKET_SIZE); |
718e3744 | 665 | |
d62a17ae | 666 | /* Make open packet. */ |
667 | bgp_packet_set_marker(s, BGP_MSG_OPEN); | |
718e3744 | 668 | |
d62a17ae | 669 | /* Set open packet values. */ |
670 | stream_putc(s, BGP_VERSION_4); /* BGP version */ | |
d7c0a89a QY |
671 | stream_putw(s, (local_as <= BGP_AS_MAX) ? (uint16_t)local_as |
672 | : BGP_AS_TRANS); | |
d62a17ae | 673 | stream_putw(s, send_holdtime); /* Hold Time */ |
674 | stream_put_in_addr(s, &peer->local_id); /* BGP Identifier */ | |
718e3744 | 675 | |
d08c0c80 DA |
676 | /* Set capabilities */ |
677 | if (CHECK_FLAG(peer->flags, PEER_FLAG_EXTENDED_OPT_PARAMS)) { | |
678 | (void)bgp_open_capability(s, peer, true); | |
679 | } else { | |
680 | struct stream *tmp = stream_new(STREAM_SIZE(s)); | |
681 | ||
682 | stream_copy(tmp, s); | |
683 | if (bgp_open_capability(tmp, peer, false) | |
684 | > BGP_OPEN_NON_EXT_OPT_LEN) { | |
685 | stream_free(tmp); | |
686 | (void)bgp_open_capability(s, peer, true); | |
687 | } else { | |
688 | stream_copy(s, tmp); | |
689 | stream_free(tmp); | |
690 | } | |
691 | } | |
718e3744 | 692 | |
d62a17ae | 693 | /* Set BGP packet length. */ |
65baedca | 694 | bgp_packet_set_size(s); |
718e3744 | 695 | |
d62a17ae | 696 | if (bgp_debug_neighbor_events(peer)) |
697 | zlog_debug( | |
23d0a753 | 698 | "%s sending OPEN, version %d, my as %u, holdtime %d, id %pI4", |
d62a17ae | 699 | peer->host, BGP_VERSION_4, local_as, send_holdtime, |
23d0a753 | 700 | &peer->local_id); |
718e3744 | 701 | |
d62a17ae | 702 | /* Dump packet if debug option is set. */ |
703 | /* bgp_packet_dump (s); */ | |
6fd04594 | 704 | hook_call(bgp_packet_send, peer, BGP_MSG_OPEN, stream_get_endp(s), s); |
718e3744 | 705 | |
d62a17ae | 706 | /* Add packet to the peer. */ |
707 | bgp_packet_add(peer, s); | |
424ab01d QY |
708 | |
709 | bgp_writes_on(peer); | |
710 | } | |
711 | ||
a127f33b QY |
712 | /* |
713 | * Writes NOTIFICATION message directly to a peer socket without waiting for | |
714 | * the I/O thread. | |
715 | * | |
716 | * There must be exactly one stream on the peer->obuf FIFO, and the data within | |
717 | * this stream must match the format of a BGP NOTIFICATION message. | |
718 | * Transmission is best-effort. | |
719 | * | |
720 | * @requires peer->io_mtx | |
721 | * @param peer | |
722 | * @return 0 | |
723 | */ | |
3dc339cd | 724 | static void bgp_write_notify(struct peer *peer) |
424ab01d QY |
725 | { |
726 | int ret, val; | |
d7c0a89a | 727 | uint8_t type; |
424ab01d QY |
728 | struct stream *s; |
729 | ||
a127f33b QY |
730 | /* There should be at least one packet. */ |
731 | s = stream_fifo_pop(peer->obuf); | |
424ab01d | 732 | |
8ec586b0 | 733 | if (!s) |
3dc339cd | 734 | return; |
8ec586b0 QY |
735 | |
736 | assert(stream_get_endp(s) >= BGP_HEADER_SIZE); | |
737 | ||
becedef6 QY |
738 | /* |
739 | * socket is in nonblocking mode, if we can't deliver the NOTIFY, well, | |
740 | * we only care about getting a clean shutdown at this point. | |
741 | */ | |
424ab01d QY |
742 | ret = write(peer->fd, STREAM_DATA(s), stream_get_endp(s)); |
743 | ||
becedef6 QY |
744 | /* |
745 | * only connection reset/close gets counted as TCP_fatal_error, failure | |
746 | * to write the entire NOTIFY doesn't get different FSM treatment | |
747 | */ | |
424ab01d | 748 | if (ret <= 0) { |
3735936b | 749 | stream_free(s); |
424ab01d | 750 | BGP_EVENT_ADD(peer, TCP_fatal_error); |
3dc339cd | 751 | return; |
424ab01d QY |
752 | } |
753 | ||
754 | /* Disable Nagle, make NOTIFY packet go out right away */ | |
755 | val = 1; | |
756 | (void)setsockopt(peer->fd, IPPROTO_TCP, TCP_NODELAY, (char *)&val, | |
757 | sizeof(val)); | |
758 | ||
759 | /* Retrieve BGP packet type. */ | |
760 | stream_set_getp(s, BGP_MARKER_SIZE + 2); | |
761 | type = stream_getc(s); | |
762 | ||
763 | assert(type == BGP_MSG_NOTIFY); | |
764 | ||
765 | /* Type should be notify. */ | |
0112e9e0 | 766 | atomic_fetch_add_explicit(&peer->notify_out, 1, memory_order_relaxed); |
424ab01d QY |
767 | |
768 | /* Double start timer. */ | |
769 | peer->v_start *= 2; | |
770 | ||
771 | /* Overflow check. */ | |
772 | if (peer->v_start >= (60 * 2)) | |
773 | peer->v_start = (60 * 2); | |
774 | ||
becedef6 QY |
775 | /* |
776 | * Handle Graceful Restart case where the state changes to | |
777 | * Connect instead of Idle | |
778 | */ | |
424ab01d QY |
779 | BGP_EVENT_ADD(peer, BGP_Stop); |
780 | ||
3735936b | 781 | stream_free(s); |
718e3744 | 782 | } |
783 | ||
eea685b6 DA |
784 | /* |
785 | * Encapsulate an original BGP CEASE Notification into Hard Reset | |
786 | */ | |
787 | static uint8_t *bgp_notify_encapsulate_hard_reset(uint8_t code, uint8_t subcode, | |
788 | uint8_t *data, size_t datalen) | |
789 | { | |
790 | uint8_t *message = XCALLOC(MTYPE_BGP_NOTIFICATION, datalen + 2); | |
791 | ||
792 | /* ErrCode */ | |
793 | message[0] = code; | |
794 | /* Subcode */ | |
795 | message[1] = subcode; | |
796 | /* Data */ | |
797 | if (datalen) | |
798 | memcpy(message + 2, data, datalen); | |
799 | ||
800 | return message; | |
801 | } | |
802 | ||
803 | /* | |
804 | * Decapsulate an original BGP CEASE Notification from Hard Reset | |
805 | */ | |
806 | struct bgp_notify bgp_notify_decapsulate_hard_reset(struct bgp_notify *notify) | |
807 | { | |
808 | struct bgp_notify bn = {}; | |
809 | ||
810 | bn.code = notify->raw_data[0]; | |
811 | bn.subcode = notify->raw_data[1]; | |
812 | bn.length = notify->length - 2; | |
813 | ||
c73d2363 | 814 | bn.raw_data = XMALLOC(MTYPE_BGP_NOTIFICATION, bn.length); |
eea685b6 DA |
815 | memcpy(bn.raw_data, notify->raw_data + 2, bn.length); |
816 | ||
817 | return bn; | |
818 | } | |
819 | ||
20170775 DA |
820 | /* Check if Graceful-Restart N-bit is exchanged */ |
821 | bool bgp_has_graceful_restart_notification(struct peer *peer) | |
822 | { | |
823 | return CHECK_FLAG(peer->cap, PEER_CAP_GRACEFUL_RESTART_N_BIT_RCV) && | |
824 | CHECK_FLAG(peer->cap, PEER_CAP_GRACEFUL_RESTART_N_BIT_ADV); | |
825 | } | |
826 | ||
eea685b6 DA |
827 | /* |
828 | * Check if to send BGP CEASE Notification/Hard Reset? | |
829 | */ | |
1ae314be DA |
830 | bool bgp_notify_send_hard_reset(struct peer *peer, uint8_t code, |
831 | uint8_t subcode) | |
eea685b6 DA |
832 | { |
833 | /* When the "N" bit has been exchanged, a Hard Reset message is used to | |
834 | * indicate to the peer that the session is to be fully terminated. | |
835 | */ | |
20170775 | 836 | if (!bgp_has_graceful_restart_notification(peer)) |
eea685b6 DA |
837 | return false; |
838 | ||
839 | /* | |
840 | * https://datatracker.ietf.org/doc/html/rfc8538#section-5.1 | |
841 | */ | |
1ae314be | 842 | if (code == BGP_NOTIFY_CEASE) { |
eea685b6 DA |
843 | switch (subcode) { |
844 | case BGP_NOTIFY_CEASE_MAX_PREFIX: | |
845 | case BGP_NOTIFY_CEASE_ADMIN_SHUTDOWN: | |
846 | case BGP_NOTIFY_CEASE_PEER_UNCONFIG: | |
847 | case BGP_NOTIFY_CEASE_HARD_RESET: | |
aebe2e37 | 848 | case BGP_NOTIFY_CEASE_BFD_DOWN: |
eea685b6 | 849 | return true; |
1ae314be DA |
850 | case BGP_NOTIFY_CEASE_ADMIN_RESET: |
851 | /* Provide user control: | |
852 | * `bgp hard-adminstrative-reset` | |
853 | */ | |
854 | if (CHECK_FLAG(peer->bgp->flags, | |
855 | BGP_FLAG_HARD_ADMIN_RESET)) | |
856 | return true; | |
857 | else | |
858 | return false; | |
eea685b6 DA |
859 | default: |
860 | break; | |
861 | } | |
862 | } | |
863 | ||
864 | return false; | |
865 | } | |
866 | ||
1ae314be DA |
867 | /* |
868 | * Check if received BGP CEASE Notification/Hard Reset? | |
869 | */ | |
870 | bool bgp_notify_received_hard_reset(struct peer *peer, uint8_t code, | |
871 | uint8_t subcode) | |
872 | { | |
873 | /* When the "N" bit has been exchanged, a Hard Reset message is used to | |
874 | * indicate to the peer that the session is to be fully terminated. | |
875 | */ | |
20170775 | 876 | if (!bgp_has_graceful_restart_notification(peer)) |
1ae314be DA |
877 | return false; |
878 | ||
879 | if (code == BGP_NOTIFY_CEASE && subcode == BGP_NOTIFY_CEASE_HARD_RESET) | |
880 | return true; | |
881 | ||
882 | return false; | |
883 | } | |
884 | ||
d3ecc69e QY |
885 | /* |
886 | * Creates a BGP Notify and appends it to the peer's output queue. | |
887 | * | |
becedef6 QY |
888 | * This function attempts to write the packet from the thread it is called |
889 | * from, to ensure the packet gets out ASAP. | |
d3ecc69e | 890 | * |
a127f33b QY |
891 | * This function may be called from multiple threads. Since the function |
892 | * modifies I/O buffer(s) in the peer, these are locked for the duration of the | |
893 | * call to prevent tampering from other threads. | |
894 | * | |
895 | * Delivery of the NOTIFICATION is attempted once and is best-effort. After | |
896 | * return, the peer structure *must* be reset; no assumptions about session | |
897 | * state are valid. | |
898 | * | |
d3ecc69e QY |
899 | * @param peer |
900 | * @param code BGP error code | |
901 | * @param sub_code BGP error subcode | |
902 | * @param data Data portion | |
903 | * @param datalen length of data portion | |
904 | */ | |
71ca5b09 MS |
905 | static void bgp_notify_send_internal(struct peer *peer, uint8_t code, |
906 | uint8_t sub_code, uint8_t *data, | |
907 | size_t datalen, bool use_curr) | |
718e3744 | 908 | { |
d62a17ae | 909 | struct stream *s; |
1ae314be | 910 | bool hard_reset = bgp_notify_send_hard_reset(peer, code, sub_code); |
d62a17ae | 911 | |
a127f33b | 912 | /* Lock I/O mutex to prevent other threads from pushing packets */ |
00dffa8c | 913 | frr_mutex_lock_autounlock(&peer->io_mtx); |
a127f33b QY |
914 | /* ============================================== */ |
915 | ||
d62a17ae | 916 | /* Allocate new stream. */ |
ef56aee4 | 917 | s = stream_new(peer->max_packet_size); |
d62a17ae | 918 | |
d3ecc69e | 919 | /* Make notify packet. */ |
d62a17ae | 920 | bgp_packet_set_marker(s, BGP_MSG_NOTIFY); |
921 | ||
eea685b6 DA |
922 | /* Check if we should send Hard Reset Notification or not */ |
923 | if (hard_reset) { | |
924 | uint8_t *hard_reset_message = bgp_notify_encapsulate_hard_reset( | |
925 | code, sub_code, data, datalen); | |
d62a17ae | 926 | |
eea685b6 DA |
927 | /* Hard Reset encapsulates another NOTIFICATION message |
928 | * in its data portion. | |
929 | */ | |
930 | stream_putc(s, BGP_NOTIFY_CEASE); | |
931 | stream_putc(s, BGP_NOTIFY_CEASE_HARD_RESET); | |
932 | stream_write(s, hard_reset_message, datalen + 2); | |
933 | ||
934 | XFREE(MTYPE_BGP_NOTIFICATION, hard_reset_message); | |
935 | } else { | |
936 | stream_putc(s, code); | |
937 | stream_putc(s, sub_code); | |
938 | if (data) | |
939 | stream_write(s, data, datalen); | |
940 | } | |
d62a17ae | 941 | |
942 | /* Set BGP packet length. */ | |
bd6b2706 | 943 | bgp_packet_set_size(s); |
d62a17ae | 944 | |
424ab01d | 945 | /* wipe output buffer */ |
a127f33b | 946 | stream_fifo_clean(peer->obuf); |
d62a17ae | 947 | |
becedef6 QY |
948 | /* |
949 | * If possible, store last packet for debugging purposes. This check is | |
950 | * in place because we are sometimes called with a doppelganger peer, | |
951 | * who tends to have a plethora of fields nulled out. | |
71ca5b09 MS |
952 | * |
953 | * Some callers should not attempt this - the io pthread for example | |
954 | * should not touch internals of the peer struct. | |
becedef6 | 955 | */ |
71ca5b09 | 956 | if (use_curr && peer->curr) { |
d8151687 | 957 | size_t packetsize = stream_get_endp(peer->curr); |
556beacf | 958 | assert(packetsize <= peer->max_packet_size); |
d8151687 QY |
959 | memcpy(peer->last_reset_cause, peer->curr->data, packetsize); |
960 | peer->last_reset_cause_size = packetsize; | |
961 | } | |
962 | ||
d62a17ae | 963 | /* For debug */ |
964 | { | |
965 | struct bgp_notify bgp_notify; | |
966 | int first = 0; | |
967 | int i; | |
968 | char c[4]; | |
969 | ||
970 | bgp_notify.code = code; | |
971 | bgp_notify.subcode = sub_code; | |
972 | bgp_notify.data = NULL; | |
e0981960 | 973 | bgp_notify.length = datalen; |
d62a17ae | 974 | bgp_notify.raw_data = data; |
975 | ||
976 | peer->notify.code = bgp_notify.code; | |
977 | peer->notify.subcode = bgp_notify.subcode; | |
3573b171 | 978 | peer->notify.length = bgp_notify.length; |
d62a17ae | 979 | |
e0981960 | 980 | if (bgp_notify.length && data) { |
3573b171 DA |
981 | bgp_notify.data = XMALLOC(MTYPE_BGP_NOTIFICATION, |
982 | bgp_notify.length * 3); | |
d62a17ae | 983 | for (i = 0; i < bgp_notify.length; i++) |
984 | if (first) { | |
552d6491 QY |
985 | snprintf(c, sizeof(c), " %02x", |
986 | data[i]); | |
f009ff26 | 987 | |
552d6491 | 988 | strlcat(bgp_notify.data, c, |
2ba1fe69 | 989 | bgp_notify.length); |
f009ff26 | 990 | |
d62a17ae | 991 | } else { |
992 | first = 1; | |
552d6491 | 993 | snprintf(c, sizeof(c), "%02x", data[i]); |
f009ff26 | 994 | |
552d6491 | 995 | strlcpy(bgp_notify.data, c, |
2ba1fe69 | 996 | bgp_notify.length); |
d62a17ae | 997 | } |
998 | } | |
eea685b6 | 999 | bgp_notify_print(peer, &bgp_notify, "sending", hard_reset); |
d62a17ae | 1000 | |
1001 | if (bgp_notify.data) { | |
3573b171 DA |
1002 | if (data) { |
1003 | XFREE(MTYPE_BGP_NOTIFICATION, | |
1004 | peer->notify.data); | |
1005 | peer->notify.data = XCALLOC( | |
1006 | MTYPE_BGP_NOTIFICATION, datalen); | |
1007 | memcpy(peer->notify.data, data, datalen); | |
1008 | } | |
1009 | ||
1010 | XFREE(MTYPE_BGP_NOTIFICATION, bgp_notify.data); | |
d62a17ae | 1011 | bgp_notify.length = 0; |
1012 | } | |
1013 | } | |
1014 | ||
1015 | /* peer reset cause */ | |
1016 | if (code == BGP_NOTIFY_CEASE) { | |
1017 | if (sub_code == BGP_NOTIFY_CEASE_ADMIN_RESET) | |
1018 | peer->last_reset = PEER_DOWN_USER_RESET; | |
5597214c DA |
1019 | else if (sub_code == BGP_NOTIFY_CEASE_ADMIN_SHUTDOWN) { |
1020 | if (CHECK_FLAG(peer->sflags, PEER_STATUS_RTT_SHUTDOWN)) | |
1021 | peer->last_reset = PEER_DOWN_RTT_SHUTDOWN; | |
1022 | else | |
1023 | peer->last_reset = PEER_DOWN_USER_SHUTDOWN; | |
1024 | } else | |
d62a17ae | 1025 | peer->last_reset = PEER_DOWN_NOTIFY_SEND; |
1026 | } else | |
1027 | peer->last_reset = PEER_DOWN_NOTIFY_SEND; | |
1028 | ||
d3ecc69e | 1029 | /* Add packet to peer's output queue */ |
a127f33b | 1030 | stream_fifo_push(peer->obuf, s); |
424ab01d | 1031 | |
5cce3f05 | 1032 | bgp_peer_gr_flags_update(peer); |
36235319 QY |
1033 | BGP_GR_ROUTER_DETECT_AND_SEND_CAPABILITY_TO_ZEBRA(peer->bgp, |
1034 | peer->bgp->peer); | |
5cce3f05 | 1035 | |
424ab01d | 1036 | bgp_write_notify(peer); |
718e3744 | 1037 | } |
1038 | ||
d3ecc69e QY |
1039 | /* |
1040 | * Creates a BGP Notify and appends it to the peer's output queue. | |
1041 | * | |
becedef6 QY |
1042 | * This function attempts to write the packet from the thread it is called |
1043 | * from, to ensure the packet gets out ASAP. | |
d3ecc69e QY |
1044 | * |
1045 | * @param peer | |
1046 | * @param code BGP error code | |
1047 | * @param sub_code BGP error subcode | |
1048 | */ | |
d7c0a89a | 1049 | void bgp_notify_send(struct peer *peer, uint8_t code, uint8_t sub_code) |
718e3744 | 1050 | { |
71ca5b09 MS |
1051 | bgp_notify_send_internal(peer, code, sub_code, NULL, 0, true); |
1052 | } | |
1053 | ||
1054 | /* | |
1055 | * Enqueue notification; called from the main pthread, peer object access is ok. | |
1056 | */ | |
1057 | void bgp_notify_send_with_data(struct peer *peer, uint8_t code, | |
1058 | uint8_t sub_code, uint8_t *data, size_t datalen) | |
1059 | { | |
1060 | bgp_notify_send_internal(peer, code, sub_code, data, datalen, true); | |
1061 | } | |
1062 | ||
1063 | /* | |
1064 | * For use by the io pthread, queueing a notification but avoiding access to | |
1065 | * the peer object. | |
1066 | */ | |
1067 | void bgp_notify_io_invalid(struct peer *peer, uint8_t code, uint8_t sub_code, | |
1068 | uint8_t *data, size_t datalen) | |
1069 | { | |
1070 | /* Avoid touching the peer object */ | |
1071 | bgp_notify_send_internal(peer, code, sub_code, data, datalen, false); | |
718e3744 | 1072 | } |
1073 | ||
d3ecc69e QY |
1074 | /* |
1075 | * Creates BGP Route Refresh packet and appends it to the peer's output queue. | |
1076 | * | |
1077 | * @param peer | |
1078 | * @param afi Address Family Identifier | |
1079 | * @param safi Subsequent Address Family Identifier | |
1080 | * @param orf_type Outbound Route Filtering type | |
1081 | * @param when_to_refresh Whether to refresh immediately or defer | |
1082 | * @param remove Whether to remove ORF for specified AFI/SAFI | |
1083 | */ | |
d62a17ae | 1084 | void bgp_route_refresh_send(struct peer *peer, afi_t afi, safi_t safi, |
d7c0a89a | 1085 | uint8_t orf_type, uint8_t when_to_refresh, |
9af52ccf | 1086 | int remove, uint8_t subtype) |
718e3744 | 1087 | { |
d62a17ae | 1088 | struct stream *s; |
1089 | struct bgp_filter *filter; | |
1090 | int orf_refresh = 0; | |
617975d1 DS |
1091 | iana_afi_t pkt_afi = IANA_AFI_IPV4; |
1092 | iana_safi_t pkt_safi = IANA_SAFI_UNICAST; | |
d62a17ae | 1093 | |
1094 | if (DISABLE_BGP_ANNOUNCE) | |
1095 | return; | |
1096 | ||
1097 | filter = &peer->filter[afi][safi]; | |
1098 | ||
1099 | /* Convert AFI, SAFI to values for packet. */ | |
1100 | bgp_map_afi_safi_int2iana(afi, safi, &pkt_afi, &pkt_safi); | |
1101 | ||
ef56aee4 | 1102 | s = stream_new(peer->max_packet_size); |
d62a17ae | 1103 | |
1104 | /* Make BGP update packet. */ | |
1105 | if (CHECK_FLAG(peer->cap, PEER_CAP_REFRESH_NEW_RCV)) | |
1106 | bgp_packet_set_marker(s, BGP_MSG_ROUTE_REFRESH_NEW); | |
718e3744 | 1107 | else |
d62a17ae | 1108 | bgp_packet_set_marker(s, BGP_MSG_ROUTE_REFRESH_OLD); |
1109 | ||
1110 | /* Encode Route Refresh message. */ | |
1111 | stream_putw(s, pkt_afi); | |
9af52ccf DA |
1112 | if (subtype) |
1113 | stream_putc(s, subtype); | |
1114 | else | |
1115 | stream_putc(s, 0); | |
d62a17ae | 1116 | stream_putc(s, pkt_safi); |
1117 | ||
1118 | if (orf_type == ORF_TYPE_PREFIX || orf_type == ORF_TYPE_PREFIX_OLD) | |
1119 | if (remove || filter->plist[FILTER_IN].plist) { | |
d7c0a89a | 1120 | uint16_t orf_len; |
d62a17ae | 1121 | unsigned long orfp; |
1122 | ||
1123 | orf_refresh = 1; | |
1124 | stream_putc(s, when_to_refresh); | |
1125 | stream_putc(s, orf_type); | |
1126 | orfp = stream_get_endp(s); | |
1127 | stream_putw(s, 0); | |
1128 | ||
1129 | if (remove) { | |
1130 | UNSET_FLAG(peer->af_sflags[afi][safi], | |
1131 | PEER_STATUS_ORF_PREFIX_SEND); | |
1132 | stream_putc(s, ORF_COMMON_PART_REMOVE_ALL); | |
1133 | if (bgp_debug_neighbor_events(peer)) | |
1134 | zlog_debug( | |
f70c91dc DA |
1135 | "%pBP sending REFRESH_REQ to remove ORF(%d) (%s) for afi/safi: %s/%s", |
1136 | peer, orf_type, | |
a7d91a8c DA |
1137 | (when_to_refresh == |
1138 | REFRESH_DEFER | |
d62a17ae | 1139 | ? "defer" |
1140 | : "immediate"), | |
748a041f DS |
1141 | iana_afi2str(pkt_afi), |
1142 | iana_safi2str(pkt_safi)); | |
d62a17ae | 1143 | } else { |
1144 | SET_FLAG(peer->af_sflags[afi][safi], | |
1145 | PEER_STATUS_ORF_PREFIX_SEND); | |
1146 | prefix_bgp_orf_entry( | |
1147 | s, filter->plist[FILTER_IN].plist, | |
1148 | ORF_COMMON_PART_ADD, | |
1149 | ORF_COMMON_PART_PERMIT, | |
1150 | ORF_COMMON_PART_DENY); | |
1151 | if (bgp_debug_neighbor_events(peer)) | |
1152 | zlog_debug( | |
f70c91dc DA |
1153 | "%pBP sending REFRESH_REQ with pfxlist ORF(%d) (%s) for afi/safi: %s/%s", |
1154 | peer, orf_type, | |
a7d91a8c DA |
1155 | (when_to_refresh == |
1156 | REFRESH_DEFER | |
d62a17ae | 1157 | ? "defer" |
1158 | : "immediate"), | |
748a041f DS |
1159 | iana_afi2str(pkt_afi), |
1160 | iana_safi2str(pkt_safi)); | |
d62a17ae | 1161 | } |
1162 | ||
1163 | /* Total ORF Entry Len. */ | |
1164 | orf_len = stream_get_endp(s) - orfp - 2; | |
1165 | stream_putw_at(s, orfp, orf_len); | |
1166 | } | |
1167 | ||
1168 | /* Set packet size. */ | |
65baedca | 1169 | bgp_packet_set_size(s); |
d62a17ae | 1170 | |
1171 | if (bgp_debug_neighbor_events(peer)) { | |
1172 | if (!orf_refresh) | |
a7d91a8c | 1173 | zlog_debug( |
f70c91dc DA |
1174 | "%pBP sending REFRESH_REQ for afi/safi: %s/%s", |
1175 | peer, iana_afi2str(pkt_afi), | |
1176 | iana_safi2str(pkt_safi)); | |
d62a17ae | 1177 | } |
1178 | ||
1179 | /* Add packet to the peer. */ | |
1180 | bgp_packet_add(peer, s); | |
424ab01d QY |
1181 | |
1182 | bgp_writes_on(peer); | |
718e3744 | 1183 | } |
1184 | ||
d3ecc69e QY |
1185 | /* |
1186 | * Create a BGP Capability packet and append it to the peer's output queue. | |
1187 | * | |
1188 | * @param peer | |
1189 | * @param afi Address Family Identifier | |
1190 | * @param safi Subsequent Address Family Identifier | |
1191 | * @param capability_code BGP Capability Code | |
1192 | * @param action Set or Remove capability | |
1193 | */ | |
d62a17ae | 1194 | void bgp_capability_send(struct peer *peer, afi_t afi, safi_t safi, |
1195 | int capability_code, int action) | |
718e3744 | 1196 | { |
d62a17ae | 1197 | struct stream *s; |
617975d1 DS |
1198 | iana_afi_t pkt_afi = IANA_AFI_IPV4; |
1199 | iana_safi_t pkt_safi = IANA_SAFI_UNICAST; | |
d62a17ae | 1200 | |
1201 | /* Convert AFI, SAFI to values for packet. */ | |
1202 | bgp_map_afi_safi_int2iana(afi, safi, &pkt_afi, &pkt_safi); | |
1203 | ||
ef56aee4 | 1204 | s = stream_new(peer->max_packet_size); |
d62a17ae | 1205 | |
1206 | /* Make BGP update packet. */ | |
1207 | bgp_packet_set_marker(s, BGP_MSG_CAPABILITY); | |
1208 | ||
1209 | /* Encode MP_EXT capability. */ | |
1210 | if (capability_code == CAPABILITY_CODE_MP) { | |
1211 | stream_putc(s, action); | |
1212 | stream_putc(s, CAPABILITY_CODE_MP); | |
1213 | stream_putc(s, CAPABILITY_CODE_MP_LEN); | |
1214 | stream_putw(s, pkt_afi); | |
1215 | stream_putc(s, 0); | |
1216 | stream_putc(s, pkt_safi); | |
1217 | ||
1218 | if (bgp_debug_neighbor_events(peer)) | |
1219 | zlog_debug( | |
f70c91dc DA |
1220 | "%pBP sending CAPABILITY has %s MP_EXT CAP for afi/safi: %s/%s", |
1221 | peer, | |
d62a17ae | 1222 | action == CAPABILITY_ACTION_SET ? "Advertising" |
1223 | : "Removing", | |
748a041f | 1224 | iana_afi2str(pkt_afi), iana_safi2str(pkt_safi)); |
d62a17ae | 1225 | } |
1226 | ||
1227 | /* Set packet size. */ | |
65baedca | 1228 | bgp_packet_set_size(s); |
718e3744 | 1229 | |
d62a17ae | 1230 | /* Add packet to the peer. */ |
1231 | bgp_packet_add(peer, s); | |
424ab01d QY |
1232 | |
1233 | bgp_writes_on(peer); | |
d62a17ae | 1234 | } |
718e3744 | 1235 | |
d62a17ae | 1236 | /* RFC1771 6.8 Connection collision detection. */ |
1237 | static int bgp_collision_detect(struct peer *new, struct in_addr remote_id) | |
1238 | { | |
1239 | struct peer *peer; | |
1240 | ||
f88221f3 DS |
1241 | /* |
1242 | * Upon receipt of an OPEN message, the local system must examine | |
1243 | * all of its connections that are in the OpenConfirm state. A BGP | |
1244 | * speaker may also examine connections in an OpenSent state if it | |
1245 | * knows the BGP Identifier of the peer by means outside of the | |
1246 | * protocol. If among these connections there is a connection to a | |
1247 | * remote BGP speaker whose BGP Identifier equals the one in the | |
1248 | * OPEN message, then the local system performs the following | |
1249 | * collision resolution procedure: | |
1250 | */ | |
1251 | peer = new->doppelganger; | |
1252 | if (peer == NULL) | |
1253 | return 0; | |
1254 | ||
1255 | /* | |
1256 | * Do not accept the new connection in Established or Clearing | |
1257 | * states. Note that a peer GR is handled by closing the existing | |
1258 | * connection upon receipt of new one. | |
1259 | */ | |
feb17238 | 1260 | if (peer_established(peer) || peer->status == Clearing) { |
f88221f3 DS |
1261 | bgp_notify_send(new, BGP_NOTIFY_CEASE, |
1262 | BGP_NOTIFY_CEASE_COLLISION_RESOLUTION); | |
1263 | return -1; | |
1264 | } | |
1265 | ||
1266 | if ((peer->status != OpenConfirm) && (peer->status != OpenSent)) | |
1267 | return 0; | |
1268 | ||
1269 | /* | |
1270 | * 1. The BGP Identifier of the local system is | |
1271 | * compared to the BGP Identifier of the remote | |
1272 | * system (as specified in the OPEN message). | |
1273 | * | |
1274 | * If the BGP Identifiers of the peers | |
1275 | * involved in the connection collision | |
1276 | * are identical, then the connection | |
1277 | * initiated by the BGP speaker with the | |
1278 | * larger AS number is preserved. | |
1279 | */ | |
1280 | if (ntohl(peer->local_id.s_addr) < ntohl(remote_id.s_addr) | |
1281 | || (ntohl(peer->local_id.s_addr) == ntohl(remote_id.s_addr) | |
1282 | && peer->local_as < peer->as)) | |
1283 | if (!CHECK_FLAG(peer->sflags, PEER_STATUS_ACCEPT_PEER)) { | |
1284 | /* | |
1285 | * 2. If the value of the local BGP | |
1286 | * Identifier is less than the remote one, | |
1287 | * the local system closes BGP connection | |
1288 | * that already exists (the one that is | |
1289 | * already in the OpenConfirm state), | |
1290 | * and accepts BGP connection initiated by | |
1291 | * the remote system. | |
1292 | */ | |
1293 | bgp_notify_send(peer, BGP_NOTIFY_CEASE, | |
1294 | BGP_NOTIFY_CEASE_COLLISION_RESOLUTION); | |
1295 | return 1; | |
1296 | } else { | |
1297 | bgp_notify_send(new, BGP_NOTIFY_CEASE, | |
1298 | BGP_NOTIFY_CEASE_COLLISION_RESOLUTION); | |
1299 | return -1; | |
1300 | } | |
1301 | else { | |
1302 | if (ntohl(peer->local_id.s_addr) == ntohl(remote_id.s_addr) | |
1303 | && peer->local_as == peer->as) | |
1304 | flog_err(EC_BGP_ROUTER_ID_SAME, | |
1305 | "Peer's router-id %pI4 is the same as ours", | |
1306 | &remote_id); | |
1307 | ||
1308 | /* | |
1309 | * 3. Otherwise, the local system closes newly | |
1310 | * created BGP connection (the one associated with the | |
1311 | * newly received OPEN message), and continues to use | |
1312 | * the existing one (the one that is already in the | |
1313 | * OpenConfirm state). | |
d62a17ae | 1314 | */ |
f88221f3 DS |
1315 | if (CHECK_FLAG(peer->sflags, PEER_STATUS_ACCEPT_PEER)) { |
1316 | bgp_notify_send(peer, BGP_NOTIFY_CEASE, | |
1317 | BGP_NOTIFY_CEASE_COLLISION_RESOLUTION); | |
1318 | return 1; | |
1319 | } else { | |
d62a17ae | 1320 | bgp_notify_send(new, BGP_NOTIFY_CEASE, |
1321 | BGP_NOTIFY_CEASE_COLLISION_RESOLUTION); | |
95f7965d | 1322 | return -1; |
d62a17ae | 1323 | } |
1324 | } | |
d62a17ae | 1325 | } |
718e3744 | 1326 | |
d8151687 QY |
1327 | /* Packet processing routines ---------------------------------------------- */ |
1328 | /* | |
1329 | * This is a family of functions designed to be called from | |
1330 | * bgp_process_packet(). These functions all share similar behavior and should | |
1331 | * adhere to the following invariants and restrictions: | |
1332 | * | |
1333 | * Return codes | |
1334 | * ------------ | |
1335 | * The return code of any one of those functions should be one of the FSM event | |
1336 | * codes specified in bgpd.h. If a NOTIFY was sent, this event code MUST be | |
1337 | * BGP_Stop. Otherwise, the code SHOULD correspond to the function's expected | |
1338 | * packet type. For example, bgp_open_receive() should return BGP_Stop upon | |
1339 | * error and Receive_OPEN_message otherwise. | |
1340 | * | |
1341 | * If no action is necessary, the correct return code is BGP_PACKET_NOOP as | |
1342 | * defined below. | |
1343 | * | |
1344 | * Side effects | |
1345 | * ------------ | |
1346 | * - May send NOTIFY messages | |
1347 | * - May not modify peer->status | |
1348 | * - May not call bgp_event_update() | |
1349 | */ | |
1350 | ||
1351 | #define BGP_PACKET_NOOP 0 | |
1352 | ||
1353 | /** | |
1354 | * Process BGP OPEN message for peer. | |
1355 | * | |
1356 | * If any errors are encountered in the OPEN message, immediately sends NOTIFY | |
1357 | * and returns BGP_Stop. | |
1358 | * | |
1359 | * @param peer | |
1360 | * @param size size of the packet | |
1361 | * @return as in summary | |
1362 | */ | |
d62a17ae | 1363 | static int bgp_open_receive(struct peer *peer, bgp_size_t size) |
1364 | { | |
1365 | int ret; | |
d7c0a89a | 1366 | uint8_t version; |
d08c0c80 | 1367 | uint16_t optlen; |
d7c0a89a QY |
1368 | uint16_t holdtime; |
1369 | uint16_t send_holdtime; | |
d62a17ae | 1370 | as_t remote_as; |
6dcef54c | 1371 | as_t as4 = 0, as4_be; |
d62a17ae | 1372 | struct in_addr remote_id; |
1373 | int mp_capability; | |
d7c0a89a QY |
1374 | uint8_t notify_data_remote_as[2]; |
1375 | uint8_t notify_data_remote_as4[4]; | |
1376 | uint8_t notify_data_remote_id[4]; | |
1377 | uint16_t *holdtime_ptr; | |
d62a17ae | 1378 | |
1379 | /* Parse open packet. */ | |
424ab01d QY |
1380 | version = stream_getc(peer->curr); |
1381 | memcpy(notify_data_remote_as, stream_pnt(peer->curr), 2); | |
1382 | remote_as = stream_getw(peer->curr); | |
d7c0a89a | 1383 | holdtime_ptr = (uint16_t *)stream_pnt(peer->curr); |
424ab01d QY |
1384 | holdtime = stream_getw(peer->curr); |
1385 | memcpy(notify_data_remote_id, stream_pnt(peer->curr), 4); | |
1386 | remote_id.s_addr = stream_get_ipv4(peer->curr); | |
d62a17ae | 1387 | |
d62a17ae | 1388 | /* BEGIN to read the capability here, but dont do it yet */ |
1389 | mp_capability = 0; | |
424ab01d | 1390 | optlen = stream_getc(peer->curr); |
d62a17ae | 1391 | |
d08c0c80 DA |
1392 | /* Extended Optional Parameters Length for BGP OPEN Message */ |
1393 | if (optlen == BGP_OPEN_NON_EXT_OPT_LEN | |
1394 | || CHECK_FLAG(peer->flags, PEER_FLAG_EXTENDED_OPT_PARAMS)) { | |
1395 | uint8_t opttype; | |
1396 | ||
766eec1b DS |
1397 | if (STREAM_READABLE(peer->curr) < 1) { |
1398 | flog_err( | |
1399 | EC_BGP_PKT_OPEN, | |
1400 | "%s: stream does not have enough bytes for extended optional parameters", | |
1401 | peer->host); | |
1402 | bgp_notify_send(peer, BGP_NOTIFY_OPEN_ERR, | |
1403 | BGP_NOTIFY_OPEN_MALFORMED_ATTR); | |
1404 | return BGP_Stop; | |
1405 | } | |
1406 | ||
d08c0c80 DA |
1407 | opttype = stream_getc(peer->curr); |
1408 | if (opttype == BGP_OPEN_NON_EXT_OPT_TYPE_EXTENDED_LENGTH) { | |
766eec1b DS |
1409 | if (STREAM_READABLE(peer->curr) < 2) { |
1410 | flog_err( | |
1411 | EC_BGP_PKT_OPEN, | |
1412 | "%s: stream does not have enough bytes to read the extended optional parameters optlen", | |
1413 | peer->host); | |
1414 | bgp_notify_send(peer, BGP_NOTIFY_OPEN_ERR, | |
1415 | BGP_NOTIFY_OPEN_MALFORMED_ATTR); | |
1416 | return BGP_Stop; | |
1417 | } | |
d08c0c80 DA |
1418 | optlen = stream_getw(peer->curr); |
1419 | SET_FLAG(peer->sflags, | |
1420 | PEER_STATUS_EXT_OPT_PARAMS_LENGTH); | |
1421 | } | |
1422 | } | |
1423 | ||
1424 | /* Receive OPEN message log */ | |
1425 | if (bgp_debug_neighbor_events(peer)) | |
1426 | zlog_debug( | |
1427 | "%s rcv OPEN%s, version %d, remote-as (in open) %u, holdtime %d, id %pI4", | |
1428 | peer->host, | |
1429 | CHECK_FLAG(peer->sflags, | |
1430 | PEER_STATUS_EXT_OPT_PARAMS_LENGTH) | |
1431 | ? " (Extended)" | |
1432 | : "", | |
1433 | version, remote_as, holdtime, &remote_id); | |
1434 | ||
d62a17ae | 1435 | if (optlen != 0) { |
1436 | /* If not enough bytes, it is an error. */ | |
424ab01d | 1437 | if (STREAM_READABLE(peer->curr) < optlen) { |
d08c0c80 DA |
1438 | flog_err(EC_BGP_PKT_OPEN, |
1439 | "%s: stream has not enough bytes (%u)", | |
1440 | peer->host, optlen); | |
d62a17ae | 1441 | bgp_notify_send(peer, BGP_NOTIFY_OPEN_ERR, |
1442 | BGP_NOTIFY_OPEN_MALFORMED_ATTR); | |
d8151687 | 1443 | return BGP_Stop; |
d62a17ae | 1444 | } |
718e3744 | 1445 | |
d62a17ae | 1446 | /* We need the as4 capability value *right now* because |
1447 | * if it is there, we have not got the remote_as yet, and | |
1448 | * without | |
1449 | * that we do not know which peer is connecting to us now. | |
1450 | */ | |
1451 | as4 = peek_for_as4_capability(peer, optlen); | |
d62a17ae | 1452 | } |
718e3744 | 1453 | |
6dcef54c DL |
1454 | as4_be = htonl(as4); |
1455 | memcpy(notify_data_remote_as4, &as4_be, 4); | |
1456 | ||
d62a17ae | 1457 | /* Just in case we have a silly peer who sends AS4 capability set to 0 |
1458 | */ | |
1459 | if (CHECK_FLAG(peer->cap, PEER_CAP_AS4_RCV) && !as4) { | |
e50f7cfd | 1460 | flog_err(EC_BGP_PKT_OPEN, |
1c50c1c0 QY |
1461 | "%s bad OPEN, got AS4 capability, but AS4 set to 0", |
1462 | peer->host); | |
d62a17ae | 1463 | bgp_notify_send_with_data(peer, BGP_NOTIFY_OPEN_ERR, |
1464 | BGP_NOTIFY_OPEN_BAD_PEER_AS, | |
1465 | notify_data_remote_as4, 4); | |
d8151687 | 1466 | return BGP_Stop; |
d62a17ae | 1467 | } |
718e3744 | 1468 | |
33d022bc DA |
1469 | /* Codification of AS 0 Processing */ |
1470 | if (remote_as == BGP_AS_ZERO) { | |
1471 | flog_err(EC_BGP_PKT_OPEN, "%s bad OPEN, got AS set to 0", | |
1472 | peer->host); | |
1473 | bgp_notify_send(peer, BGP_NOTIFY_OPEN_ERR, | |
1474 | BGP_NOTIFY_OPEN_BAD_PEER_AS); | |
1475 | return BGP_Stop; | |
1476 | } | |
1477 | ||
d62a17ae | 1478 | if (remote_as == BGP_AS_TRANS) { |
1479 | /* Take the AS4 from the capability. We must have received the | |
1480 | * capability now! Otherwise we have a asn16 peer who uses | |
1481 | * BGP_AS_TRANS, for some unknown reason. | |
1482 | */ | |
1483 | if (as4 == BGP_AS_TRANS) { | |
af4c2728 | 1484 | flog_err( |
e50f7cfd | 1485 | EC_BGP_PKT_OPEN, |
d62a17ae | 1486 | "%s [AS4] NEW speaker using AS_TRANS for AS4, not allowed", |
1487 | peer->host); | |
1488 | bgp_notify_send_with_data(peer, BGP_NOTIFY_OPEN_ERR, | |
1489 | BGP_NOTIFY_OPEN_BAD_PEER_AS, | |
1490 | notify_data_remote_as4, 4); | |
d8151687 | 1491 | return BGP_Stop; |
d62a17ae | 1492 | } |
718e3744 | 1493 | |
d62a17ae | 1494 | if (!as4 && BGP_DEBUG(as4, AS4)) |
1495 | zlog_debug( | |
3efd0893 | 1496 | "%s [AS4] OPEN remote_as is AS_TRANS, but no AS4. Odd, but proceeding.", |
d62a17ae | 1497 | peer->host); |
1498 | else if (as4 < BGP_AS_MAX && BGP_DEBUG(as4, AS4)) | |
1499 | zlog_debug( | |
3efd0893 | 1500 | "%s [AS4] OPEN remote_as is AS_TRANS, but AS4 (%u) fits in 2-bytes, very odd peer.", |
d62a17ae | 1501 | peer->host, as4); |
1502 | if (as4) | |
1503 | remote_as = as4; | |
1504 | } else { | |
1505 | /* We may have a partner with AS4 who has an asno < BGP_AS_MAX | |
1506 | */ | |
1507 | /* If we have got the capability, peer->as4cap must match | |
1508 | * remote_as */ | |
1509 | if (CHECK_FLAG(peer->cap, PEER_CAP_AS4_RCV) | |
1510 | && as4 != remote_as) { | |
1511 | /* raise error, log this, close session */ | |
af4c2728 | 1512 | flog_err( |
e50f7cfd | 1513 | EC_BGP_PKT_OPEN, |
3efd0893 | 1514 | "%s bad OPEN, got AS4 capability, but remote_as %u mismatch with 16bit 'myasn' %u in open", |
d62a17ae | 1515 | peer->host, as4, remote_as); |
1516 | bgp_notify_send_with_data(peer, BGP_NOTIFY_OPEN_ERR, | |
1517 | BGP_NOTIFY_OPEN_BAD_PEER_AS, | |
1518 | notify_data_remote_as4, 4); | |
d8151687 | 1519 | return BGP_Stop; |
d62a17ae | 1520 | } |
1521 | } | |
6b0655a2 | 1522 | |
787c3020 DA |
1523 | /* rfc6286: |
1524 | * If the BGP Identifier field of the OPEN message | |
1525 | * is zero, or if it is the same as the BGP Identifier | |
1526 | * of the local BGP speaker and the message is from an | |
1527 | * internal peer, then the Error Subcode is set to | |
1528 | * "Bad BGP Identifier". | |
1529 | */ | |
975a328e | 1530 | if (remote_id.s_addr == INADDR_ANY |
787c3020 DA |
1531 | || (peer->sort == BGP_PEER_IBGP |
1532 | && ntohl(peer->local_id.s_addr) == ntohl(remote_id.s_addr))) { | |
d62a17ae | 1533 | if (bgp_debug_neighbor_events(peer)) |
23d0a753 DA |
1534 | zlog_debug("%s bad OPEN, wrong router identifier %pI4", |
1535 | peer->host, &remote_id); | |
d62a17ae | 1536 | bgp_notify_send_with_data(peer, BGP_NOTIFY_OPEN_ERR, |
1537 | BGP_NOTIFY_OPEN_BAD_BGP_IDENT, | |
1538 | notify_data_remote_id, 4); | |
d8151687 | 1539 | return BGP_Stop; |
d62a17ae | 1540 | } |
1541 | ||
d62a17ae | 1542 | /* Peer BGP version check. */ |
1543 | if (version != BGP_VERSION_4) { | |
d7c0a89a | 1544 | uint16_t maxver = htons(BGP_VERSION_4); |
d62a17ae | 1545 | /* XXX this reply may not be correct if version < 4 XXX */ |
1546 | if (bgp_debug_neighbor_events(peer)) | |
1547 | zlog_debug( | |
1548 | "%s bad protocol version, remote requested %d, local request %d", | |
1549 | peer->host, version, BGP_VERSION_4); | |
1550 | /* Data must be in network byte order here */ | |
1551 | bgp_notify_send_with_data(peer, BGP_NOTIFY_OPEN_ERR, | |
1552 | BGP_NOTIFY_OPEN_UNSUP_VERSION, | |
d7c0a89a | 1553 | (uint8_t *)&maxver, 2); |
d8151687 | 1554 | return BGP_Stop; |
d62a17ae | 1555 | } |
1556 | ||
1557 | /* Check neighbor as number. */ | |
1558 | if (peer->as_type == AS_UNSPECIFIED) { | |
1559 | if (bgp_debug_neighbor_events(peer)) | |
1560 | zlog_debug( | |
1561 | "%s bad OPEN, remote AS is unspecified currently", | |
1562 | peer->host); | |
1563 | bgp_notify_send_with_data(peer, BGP_NOTIFY_OPEN_ERR, | |
1564 | BGP_NOTIFY_OPEN_BAD_PEER_AS, | |
1565 | notify_data_remote_as, 2); | |
d8151687 | 1566 | return BGP_Stop; |
d62a17ae | 1567 | } else if (peer->as_type == AS_INTERNAL) { |
1568 | if (remote_as != peer->bgp->as) { | |
1569 | if (bgp_debug_neighbor_events(peer)) | |
1570 | zlog_debug( | |
1571 | "%s bad OPEN, remote AS is %u, internal specified", | |
1572 | peer->host, remote_as); | |
1573 | bgp_notify_send_with_data(peer, BGP_NOTIFY_OPEN_ERR, | |
1574 | BGP_NOTIFY_OPEN_BAD_PEER_AS, | |
1575 | notify_data_remote_as, 2); | |
d8151687 | 1576 | return BGP_Stop; |
1ff9a340 | 1577 | } |
d62a17ae | 1578 | peer->as = peer->local_as; |
1579 | } else if (peer->as_type == AS_EXTERNAL) { | |
1580 | if (remote_as == peer->bgp->as) { | |
1581 | if (bgp_debug_neighbor_events(peer)) | |
1582 | zlog_debug( | |
1583 | "%s bad OPEN, remote AS is %u, external specified", | |
1584 | peer->host, remote_as); | |
1585 | bgp_notify_send_with_data(peer, BGP_NOTIFY_OPEN_ERR, | |
1586 | BGP_NOTIFY_OPEN_BAD_PEER_AS, | |
1587 | notify_data_remote_as, 2); | |
d8151687 | 1588 | return BGP_Stop; |
1ff9a340 | 1589 | } |
d62a17ae | 1590 | peer->as = remote_as; |
1591 | } else if ((peer->as_type == AS_SPECIFIED) && (remote_as != peer->as)) { | |
1592 | if (bgp_debug_neighbor_events(peer)) | |
1593 | zlog_debug("%s bad OPEN, remote AS is %u, expected %u", | |
1594 | peer->host, remote_as, peer->as); | |
1595 | bgp_notify_send_with_data(peer, BGP_NOTIFY_OPEN_ERR, | |
1596 | BGP_NOTIFY_OPEN_BAD_PEER_AS, | |
1597 | notify_data_remote_as, 2); | |
d8151687 | 1598 | return BGP_Stop; |
eb821189 | 1599 | } |
718e3744 | 1600 | |
7a75470f DS |
1601 | /* |
1602 | * When collision is detected and this peer is closed. | |
1603 | * Return immediately. | |
1604 | */ | |
1605 | ret = bgp_collision_detect(peer, remote_id); | |
1606 | if (ret < 0) | |
1607 | return BGP_Stop; | |
1608 | ||
1609 | /* Get sockname. */ | |
1610 | if (bgp_getsockname(peer) < 0) { | |
1611 | flog_err_sys(EC_LIB_SOCKET, | |
1612 | "%s: bgp_getsockname() failed for peer: %s", | |
1613 | __func__, peer->host); | |
1614 | return BGP_Stop; | |
1615 | } | |
1616 | ||
1617 | /* Set remote router-id */ | |
1618 | peer->remote_id = remote_id; | |
1619 | ||
d62a17ae | 1620 | /* From the rfc: Upon receipt of an OPEN message, a BGP speaker MUST |
1621 | calculate the value of the Hold Timer by using the smaller of its | |
1622 | configured Hold Time and the Hold Time received in the OPEN message. | |
1623 | The Hold Time MUST be either zero or at least three seconds. An | |
1624 | implementation may reject connections on the basis of the Hold Time. | |
0b2aa3a0 | 1625 | */ |
d62a17ae | 1626 | |
1627 | if (holdtime < 3 && holdtime != 0) { | |
b042667a TI |
1628 | bgp_notify_send_with_data(peer, BGP_NOTIFY_OPEN_ERR, |
1629 | BGP_NOTIFY_OPEN_UNACEP_HOLDTIME, | |
1630 | (uint8_t *)holdtime_ptr, 2); | |
1631 | return BGP_Stop; | |
1632 | } | |
1633 | ||
1634 | /* Send notification message when Hold Time received in the OPEN message | |
1635 | * is smaller than configured minimum Hold Time. */ | |
1636 | if (holdtime < peer->bgp->default_min_holdtime | |
1637 | && peer->bgp->default_min_holdtime != 0) { | |
d62a17ae | 1638 | bgp_notify_send_with_data(peer, BGP_NOTIFY_OPEN_ERR, |
1639 | BGP_NOTIFY_OPEN_UNACEP_HOLDTIME, | |
d7c0a89a | 1640 | (uint8_t *)holdtime_ptr, 2); |
d8151687 | 1641 | return BGP_Stop; |
0b2aa3a0 | 1642 | } |
d62a17ae | 1643 | |
1644 | /* From the rfc: A reasonable maximum time between KEEPALIVE messages | |
1645 | would be one third of the Hold Time interval. KEEPALIVE messages | |
1646 | MUST NOT be sent more frequently than one per second. An | |
1647 | implementation MAY adjust the rate at which it sends KEEPALIVE | |
1648 | messages as a function of the Hold Time interval. */ | |
1649 | ||
b90a8e13 | 1650 | if (CHECK_FLAG(peer->flags, PEER_FLAG_TIMER)) |
d62a17ae | 1651 | send_holdtime = peer->holdtime; |
1652 | else | |
1653 | send_holdtime = peer->bgp->default_holdtime; | |
1654 | ||
1655 | if (holdtime < send_holdtime) | |
1656 | peer->v_holdtime = holdtime; | |
1657 | else | |
1658 | peer->v_holdtime = send_holdtime; | |
1659 | ||
7aa4fd5b TA |
1660 | /* Set effective keepalive to 1/3 the effective holdtime. |
1661 | * Use configured keeplive when < effective keepalive. | |
1662 | */ | |
1663 | peer->v_keepalive = peer->v_holdtime / 3; | |
1664 | if (CHECK_FLAG(peer->flags, PEER_FLAG_TIMER)) { | |
1665 | if (peer->keepalive && peer->keepalive < peer->v_keepalive) | |
1666 | peer->v_keepalive = peer->keepalive; | |
1667 | } else { | |
1668 | if (peer->bgp->default_keepalive | |
1669 | && peer->bgp->default_keepalive < peer->v_keepalive) | |
1670 | peer->v_keepalive = peer->bgp->default_keepalive; | |
1671 | } | |
d62a17ae | 1672 | |
234f6fd4 DA |
1673 | /* If another side disabled sending Software Version capability, |
1674 | * we MUST drop the previous from showing in the outputs to avoid | |
1675 | * stale information and due to security reasons. | |
1676 | */ | |
1677 | if (peer->soft_version) | |
1678 | XFREE(MTYPE_BGP_SOFT_VERSION, peer->soft_version); | |
1679 | ||
d62a17ae | 1680 | /* Open option part parse. */ |
1681 | if (optlen != 0) { | |
1bb379bf | 1682 | if (bgp_open_option_parse(peer, optlen, &mp_capability) < 0) |
d8151687 | 1683 | return BGP_Stop; |
d62a17ae | 1684 | } else { |
1685 | if (bgp_debug_neighbor_events(peer)) | |
1686 | zlog_debug("%s rcvd OPEN w/ OPTION parameter len: 0", | |
1687 | peer->host); | |
0299c004 | 1688 | } |
d62a17ae | 1689 | |
1690 | /* | |
1691 | * Assume that the peer supports the locally configured set of | |
1692 | * AFI/SAFIs if the peer did not send us any Mulitiprotocol | |
1693 | * capabilities, or if 'override-capability' is configured. | |
1694 | */ | |
1695 | if (!mp_capability | |
1696 | || CHECK_FLAG(peer->flags, PEER_FLAG_OVERRIDE_CAPABILITY)) { | |
1697 | peer->afc_nego[AFI_IP][SAFI_UNICAST] = | |
1698 | peer->afc[AFI_IP][SAFI_UNICAST]; | |
1699 | peer->afc_nego[AFI_IP][SAFI_MULTICAST] = | |
1700 | peer->afc[AFI_IP][SAFI_MULTICAST]; | |
1701 | peer->afc_nego[AFI_IP][SAFI_LABELED_UNICAST] = | |
1702 | peer->afc[AFI_IP][SAFI_LABELED_UNICAST]; | |
7c40bf39 | 1703 | peer->afc_nego[AFI_IP][SAFI_FLOWSPEC] = |
1704 | peer->afc[AFI_IP][SAFI_FLOWSPEC]; | |
d62a17ae | 1705 | peer->afc_nego[AFI_IP6][SAFI_UNICAST] = |
1706 | peer->afc[AFI_IP6][SAFI_UNICAST]; | |
1707 | peer->afc_nego[AFI_IP6][SAFI_MULTICAST] = | |
1708 | peer->afc[AFI_IP6][SAFI_MULTICAST]; | |
1709 | peer->afc_nego[AFI_IP6][SAFI_LABELED_UNICAST] = | |
1710 | peer->afc[AFI_IP6][SAFI_LABELED_UNICAST]; | |
1711 | peer->afc_nego[AFI_L2VPN][SAFI_EVPN] = | |
1712 | peer->afc[AFI_L2VPN][SAFI_EVPN]; | |
7c40bf39 | 1713 | peer->afc_nego[AFI_IP6][SAFI_FLOWSPEC] = |
1714 | peer->afc[AFI_IP6][SAFI_FLOWSPEC]; | |
0299c004 | 1715 | } |
d62a17ae | 1716 | |
d62a17ae | 1717 | /* Verify valid local address present based on negotiated |
1718 | * address-families. */ | |
1719 | if (peer->afc_nego[AFI_IP][SAFI_UNICAST] | |
e5f22b30 | 1720 | || peer->afc_nego[AFI_IP][SAFI_LABELED_UNICAST] |
d62a17ae | 1721 | || peer->afc_nego[AFI_IP][SAFI_MULTICAST] |
1722 | || peer->afc_nego[AFI_IP][SAFI_MPLS_VPN] | |
1723 | || peer->afc_nego[AFI_IP][SAFI_ENCAP]) { | |
975a328e | 1724 | if (peer->nexthop.v4.s_addr == INADDR_ANY) { |
d62a17ae | 1725 | #if defined(HAVE_CUMULUS) |
50121ac0 DS |
1726 | zlog_warn("%s: No local IPv4 addr, BGP routing may not work", |
1727 | peer->host); | |
1d808091 | 1728 | #endif |
d62a17ae | 1729 | } |
1730 | } | |
1731 | if (peer->afc_nego[AFI_IP6][SAFI_UNICAST] | |
e5f22b30 | 1732 | || peer->afc_nego[AFI_IP6][SAFI_LABELED_UNICAST] |
d62a17ae | 1733 | || peer->afc_nego[AFI_IP6][SAFI_MULTICAST] |
1734 | || peer->afc_nego[AFI_IP6][SAFI_MPLS_VPN] | |
1735 | || peer->afc_nego[AFI_IP6][SAFI_ENCAP]) { | |
1736 | if (IN6_IS_ADDR_UNSPECIFIED(&peer->nexthop.v6_global)) { | |
1737 | #if defined(HAVE_CUMULUS) | |
50121ac0 DS |
1738 | zlog_warn("%s: No local IPv6 address, BGP routing may not work", |
1739 | peer->host); | |
1d808091 | 1740 | #endif |
d62a17ae | 1741 | } |
1742 | } | |
1743 | peer->rtt = sockopt_tcp_rtt(peer->fd); | |
1744 | ||
d8151687 | 1745 | return Receive_OPEN_message; |
718e3744 | 1746 | } |
1747 | ||
d8151687 QY |
1748 | /** |
1749 | * Process BGP KEEPALIVE message for peer. | |
1750 | * | |
1751 | * @param peer | |
1752 | * @param size size of the packet | |
1753 | * @return as in summary | |
1754 | */ | |
1755 | static int bgp_keepalive_receive(struct peer *peer, bgp_size_t size) | |
f188f2c4 | 1756 | { |
d8151687 QY |
1757 | if (bgp_debug_keepalive(peer)) |
1758 | zlog_debug("%s KEEPALIVE rcvd", peer->host); | |
d62a17ae | 1759 | |
d8151687 | 1760 | bgp_update_implicit_eors(peer); |
d62a17ae | 1761 | |
e410d563 DA |
1762 | peer->rtt = sockopt_tcp_rtt(peer->fd); |
1763 | ||
8336c896 DA |
1764 | /* If the peer's RTT is higher than expected, shutdown |
1765 | * the peer automatically. | |
1766 | */ | |
5597214c DA |
1767 | if (!CHECK_FLAG(peer->flags, PEER_FLAG_RTT_SHUTDOWN)) |
1768 | return Receive_KEEPALIVE_message; | |
8336c896 | 1769 | |
5597214c | 1770 | if (peer->rtt > peer->rtt_expected) { |
8336c896 DA |
1771 | peer->rtt_keepalive_rcv++; |
1772 | ||
1773 | if (peer->rtt_keepalive_rcv > peer->rtt_keepalive_conf) { | |
5597214c DA |
1774 | char rtt_shutdown_reason[BUFSIZ] = {}; |
1775 | ||
1776 | snprintfrr( | |
1777 | rtt_shutdown_reason, | |
1778 | sizeof(rtt_shutdown_reason), | |
1779 | "shutdown due to high round-trip-time (%dms > %dms, hit %u times)", | |
1780 | peer->rtt, peer->rtt_expected, | |
1781 | peer->rtt_keepalive_rcv); | |
1782 | zlog_warn("%s %s", peer->host, rtt_shutdown_reason); | |
1783 | SET_FLAG(peer->sflags, PEER_STATUS_RTT_SHUTDOWN); | |
1784 | peer_tx_shutdown_message_set(peer, rtt_shutdown_reason); | |
8336c896 DA |
1785 | peer_flag_set(peer, PEER_FLAG_SHUTDOWN); |
1786 | } | |
1787 | } else { | |
1788 | if (peer->rtt_keepalive_rcv) | |
1789 | peer->rtt_keepalive_rcv--; | |
1790 | } | |
1791 | ||
d8151687 | 1792 | return Receive_KEEPALIVE_message; |
f188f2c4 DS |
1793 | } |
1794 | ||
cc9f21da | 1795 | static void bgp_refresh_stalepath_timer_expire(struct thread *thread) |
9af52ccf DA |
1796 | { |
1797 | struct peer_af *paf; | |
1798 | ||
1799 | paf = THREAD_ARG(thread); | |
1800 | ||
1801 | afi_t afi = paf->afi; | |
1802 | safi_t safi = paf->safi; | |
1803 | struct peer *peer = paf->peer; | |
1804 | ||
1805 | peer->t_refresh_stalepath = NULL; | |
1806 | ||
1807 | if (peer->nsf[afi][safi]) | |
1808 | bgp_clear_stale_route(peer, afi, safi); | |
1809 | ||
1810 | if (bgp_debug_neighbor_events(peer)) | |
a7d91a8c | 1811 | zlog_debug( |
f70c91dc DA |
1812 | "%pBP route-refresh (BoRR) timer expired for afi/safi: %d/%d", |
1813 | peer, afi, safi); | |
9af52ccf DA |
1814 | |
1815 | bgp_timer_set(peer); | |
9af52ccf | 1816 | } |
d62a17ae | 1817 | |
d8151687 QY |
1818 | /** |
1819 | * Process BGP UPDATE message for peer. | |
1820 | * | |
1821 | * Parses UPDATE and creates attribute object. | |
1822 | * | |
1823 | * @param peer | |
1824 | * @param size size of the packet | |
1825 | * @return as in summary | |
7ef5a232 | 1826 | */ |
d62a17ae | 1827 | static int bgp_update_receive(struct peer *peer, bgp_size_t size) |
718e3744 | 1828 | { |
d62a17ae | 1829 | int ret, nlri_ret; |
d7c0a89a | 1830 | uint8_t *end; |
d62a17ae | 1831 | struct stream *s; |
1832 | struct attr attr; | |
1833 | bgp_size_t attribute_len; | |
1834 | bgp_size_t update_len; | |
1835 | bgp_size_t withdraw_len; | |
f009ff26 | 1836 | bool restart = false; |
d62a17ae | 1837 | |
1838 | enum NLRI_TYPES { | |
1839 | NLRI_UPDATE, | |
1840 | NLRI_WITHDRAW, | |
1841 | NLRI_MP_UPDATE, | |
1842 | NLRI_MP_WITHDRAW, | |
1843 | NLRI_TYPE_MAX | |
1844 | }; | |
1845 | struct bgp_nlri nlris[NLRI_TYPE_MAX]; | |
1846 | ||
1847 | /* Status must be Established. */ | |
feb17238 | 1848 | if (!peer_established(peer)) { |
e50f7cfd | 1849 | flog_err(EC_BGP_INVALID_STATUS, |
1c50c1c0 QY |
1850 | "%s [FSM] Update packet received under status %s", |
1851 | peer->host, | |
1852 | lookup_msg(bgp_status_msg, peer->status, NULL)); | |
0e35025e | 1853 | bgp_notify_send(peer, BGP_NOTIFY_FSM_ERR, |
3893aeee | 1854 | bgp_fsm_error_subcode(peer->status)); |
d8151687 | 1855 | return BGP_Stop; |
d62a17ae | 1856 | } |
1857 | ||
1858 | /* Set initial values. */ | |
6006b807 | 1859 | memset(&attr, 0, sizeof(attr)); |
d62a17ae | 1860 | attr.label_index = BGP_INVALID_LABEL_INDEX; |
1861 | attr.label = MPLS_INVALID_LABEL; | |
1862 | memset(&nlris, 0, sizeof(nlris)); | |
1863 | memset(peer->rcvd_attr_str, 0, BUFSIZ); | |
1864 | peer->rcvd_attr_printed = 0; | |
1865 | ||
424ab01d | 1866 | s = peer->curr; |
d62a17ae | 1867 | end = stream_pnt(s) + size; |
1868 | ||
1869 | /* RFC1771 6.3 If the Unfeasible Routes Length or Total Attribute | |
1870 | Length is too large (i.e., if Unfeasible Routes Length + Total | |
1871 | Attribute Length + 23 exceeds the message Length), then the Error | |
1872 | Subcode is set to Malformed Attribute List. */ | |
1873 | if (stream_pnt(s) + 2 > end) { | |
e50f7cfd | 1874 | flog_err(EC_BGP_UPDATE_RCV, |
3efd0893 | 1875 | "%s [Error] Update packet error (packet length is short for unfeasible length)", |
1c50c1c0 | 1876 | peer->host); |
d62a17ae | 1877 | bgp_notify_send(peer, BGP_NOTIFY_UPDATE_ERR, |
1878 | BGP_NOTIFY_UPDATE_MAL_ATTR); | |
d8151687 | 1879 | return BGP_Stop; |
d62a17ae | 1880 | } |
1881 | ||
1882 | /* Unfeasible Route Length. */ | |
1883 | withdraw_len = stream_getw(s); | |
1884 | ||
1885 | /* Unfeasible Route Length check. */ | |
1886 | if (stream_pnt(s) + withdraw_len > end) { | |
e50f7cfd | 1887 | flog_err(EC_BGP_UPDATE_RCV, |
3efd0893 | 1888 | "%s [Error] Update packet error (packet unfeasible length overflow %d)", |
1c50c1c0 | 1889 | peer->host, withdraw_len); |
d62a17ae | 1890 | bgp_notify_send(peer, BGP_NOTIFY_UPDATE_ERR, |
1891 | BGP_NOTIFY_UPDATE_MAL_ATTR); | |
d8151687 | 1892 | return BGP_Stop; |
d62a17ae | 1893 | } |
1894 | ||
1895 | /* Unfeasible Route packet format check. */ | |
1896 | if (withdraw_len > 0) { | |
1897 | nlris[NLRI_WITHDRAW].afi = AFI_IP; | |
1898 | nlris[NLRI_WITHDRAW].safi = SAFI_UNICAST; | |
1899 | nlris[NLRI_WITHDRAW].nlri = stream_pnt(s); | |
1900 | nlris[NLRI_WITHDRAW].length = withdraw_len; | |
1901 | stream_forward_getp(s, withdraw_len); | |
1902 | } | |
1903 | ||
1904 | /* Attribute total length check. */ | |
1905 | if (stream_pnt(s) + 2 > end) { | |
ade6974d | 1906 | flog_warn( |
e50f7cfd | 1907 | EC_BGP_UPDATE_PACKET_SHORT, |
ade6974d QY |
1908 | "%s [Error] Packet Error (update packet is short for attribute length)", |
1909 | peer->host); | |
d62a17ae | 1910 | bgp_notify_send(peer, BGP_NOTIFY_UPDATE_ERR, |
1911 | BGP_NOTIFY_UPDATE_MAL_ATTR); | |
d8151687 | 1912 | return BGP_Stop; |
d62a17ae | 1913 | } |
1914 | ||
1915 | /* Fetch attribute total length. */ | |
1916 | attribute_len = stream_getw(s); | |
1917 | ||
1918 | /* Attribute length check. */ | |
1919 | if (stream_pnt(s) + attribute_len > end) { | |
ade6974d | 1920 | flog_warn( |
e50f7cfd | 1921 | EC_BGP_UPDATE_PACKET_LONG, |
ade6974d QY |
1922 | "%s [Error] Packet Error (update packet attribute length overflow %d)", |
1923 | peer->host, attribute_len); | |
d62a17ae | 1924 | bgp_notify_send(peer, BGP_NOTIFY_UPDATE_ERR, |
1925 | BGP_NOTIFY_UPDATE_MAL_ATTR); | |
d8151687 | 1926 | return BGP_Stop; |
d62a17ae | 1927 | } |
1928 | ||
1929 | /* Certain attribute parsing errors should not be considered bad enough | |
1930 | * to reset the session for, most particularly any partial/optional | |
1931 | * attributes that have 'tunneled' over speakers that don't understand | |
1932 | * them. Instead we withdraw only the prefix concerned. | |
1933 | * | |
1934 | * Complicates the flow a little though.. | |
1935 | */ | |
79288e4c | 1936 | enum bgp_attr_parse_ret attr_parse_ret = BGP_ATTR_PARSE_PROCEED; |
d62a17ae | 1937 | /* This define morphs the update case into a withdraw when lower levels |
1938 | * have signalled an error condition where this is best. | |
1939 | */ | |
b881c707 | 1940 | #define NLRI_ATTR_ARG (attr_parse_ret != BGP_ATTR_PARSE_WITHDRAW ? &attr : NULL) |
718e3744 | 1941 | |
d62a17ae | 1942 | /* Parse attribute when it exists. */ |
1943 | if (attribute_len) { | |
1944 | attr_parse_ret = bgp_attr_parse(peer, &attr, attribute_len, | |
1945 | &nlris[NLRI_MP_UPDATE], | |
1946 | &nlris[NLRI_MP_WITHDRAW]); | |
1947 | if (attr_parse_ret == BGP_ATTR_PARSE_ERROR) { | |
1948 | bgp_attr_unintern_sub(&attr); | |
d8151687 | 1949 | return BGP_Stop; |
d62a17ae | 1950 | } |
1951 | } | |
1952 | ||
1953 | /* Logging the attribute. */ | |
1954 | if (attr_parse_ret == BGP_ATTR_PARSE_WITHDRAW | |
1955 | || BGP_DEBUG(update, UPDATE_IN) | |
1956 | || BGP_DEBUG(update, UPDATE_PREFIX)) { | |
5022c833 DA |
1957 | ret = bgp_dump_attr(&attr, peer->rcvd_attr_str, |
1958 | sizeof(peer->rcvd_attr_str)); | |
d62a17ae | 1959 | |
b4d46cc9 DL |
1960 | peer->stat_upd_7606++; |
1961 | ||
d62a17ae | 1962 | if (attr_parse_ret == BGP_ATTR_PARSE_WITHDRAW) |
af4c2728 | 1963 | flog_err( |
e50f7cfd | 1964 | EC_BGP_UPDATE_RCV, |
f70c91dc DA |
1965 | "%pBP rcvd UPDATE with errors in attr(s)!! Withdrawing route.", |
1966 | peer); | |
d62a17ae | 1967 | |
1968 | if (ret && bgp_debug_update(peer, NULL, NULL, 1)) { | |
f70c91dc | 1969 | zlog_debug("%pBP rcvd UPDATE w/ attr: %s", peer, |
d62a17ae | 1970 | peer->rcvd_attr_str); |
1971 | peer->rcvd_attr_printed = 1; | |
1972 | } | |
1973 | } | |
1974 | ||
1975 | /* Network Layer Reachability Information. */ | |
1976 | update_len = end - stream_pnt(s); | |
1977 | ||
1978 | if (update_len) { | |
1979 | /* Set NLRI portion to structure. */ | |
1980 | nlris[NLRI_UPDATE].afi = AFI_IP; | |
1981 | nlris[NLRI_UPDATE].safi = SAFI_UNICAST; | |
1982 | nlris[NLRI_UPDATE].nlri = stream_pnt(s); | |
1983 | nlris[NLRI_UPDATE].length = update_len; | |
1984 | stream_forward_getp(s, update_len); | |
9738e9aa | 1985 | |
1986 | if (CHECK_FLAG(attr.flag, ATTR_FLAG_BIT(BGP_ATTR_MP_REACH_NLRI))) { | |
1987 | /* | |
1988 | * We skipped nexthop attribute validation earlier so | |
1989 | * validate the nexthop now. | |
1990 | */ | |
1991 | if (bgp_attr_nexthop_valid(peer, &attr) < 0) { | |
1992 | bgp_attr_unintern_sub(&attr); | |
1993 | return BGP_Stop; | |
1994 | } | |
1995 | } | |
d62a17ae | 1996 | } |
1997 | ||
1998 | if (BGP_DEBUG(update, UPDATE_IN)) | |
f70c91dc DA |
1999 | zlog_debug("%pBP rcvd UPDATE wlen %d attrlen %d alen %d", peer, |
2000 | withdraw_len, attribute_len, update_len); | |
d62a17ae | 2001 | |
2002 | /* Parse any given NLRIs */ | |
2003 | for (int i = NLRI_UPDATE; i < NLRI_TYPE_MAX; i++) { | |
2004 | if (!nlris[i].nlri) | |
2005 | continue; | |
2006 | ||
2007 | /* NLRI is processed iff the peer if configured for the specific | |
2008 | * afi/safi */ | |
2009 | if (!peer->afc[nlris[i].afi][nlris[i].safi]) { | |
2010 | zlog_info( | |
2011 | "%s [Info] UPDATE for non-enabled AFI/SAFI %u/%u", | |
2012 | peer->host, nlris[i].afi, nlris[i].safi); | |
2013 | continue; | |
2014 | } | |
2015 | ||
2016 | /* EoR handled later */ | |
2017 | if (nlris[i].length == 0) | |
2018 | continue; | |
2019 | ||
2020 | switch (i) { | |
2021 | case NLRI_UPDATE: | |
2022 | case NLRI_MP_UPDATE: | |
2023 | nlri_ret = bgp_nlri_parse(peer, NLRI_ATTR_ARG, | |
2024 | &nlris[i], 0); | |
2025 | break; | |
2026 | case NLRI_WITHDRAW: | |
2027 | case NLRI_MP_WITHDRAW: | |
b1d7888f DA |
2028 | nlri_ret = bgp_nlri_parse(peer, NLRI_ATTR_ARG, |
2029 | &nlris[i], 1); | |
d62a17ae | 2030 | break; |
2031 | default: | |
513386b5 | 2032 | nlri_ret = BGP_NLRI_PARSE_ERROR; |
d62a17ae | 2033 | } |
2034 | ||
513386b5 DA |
2035 | if (nlri_ret < BGP_NLRI_PARSE_OK |
2036 | && nlri_ret != BGP_NLRI_PARSE_ERROR_PREFIX_OVERFLOW) { | |
e50f7cfd | 2037 | flog_err(EC_BGP_UPDATE_RCV, |
1c50c1c0 | 2038 | "%s [Error] Error parsing NLRI", peer->host); |
feb17238 | 2039 | if (peer_established(peer)) |
d62a17ae | 2040 | bgp_notify_send( |
2041 | peer, BGP_NOTIFY_UPDATE_ERR, | |
2042 | i <= NLRI_WITHDRAW | |
2043 | ? BGP_NOTIFY_UPDATE_INVAL_NETWORK | |
2044 | : BGP_NOTIFY_UPDATE_OPT_ATTR_ERR); | |
2045 | bgp_attr_unintern_sub(&attr); | |
d8151687 | 2046 | return BGP_Stop; |
d62a17ae | 2047 | } |
2048 | } | |
2049 | ||
2050 | /* EoR checks | |
2051 | * | |
2052 | * Non-MP IPv4/Unicast EoR is a completely empty UPDATE | |
2053 | * and MP EoR should have only an empty MP_UNREACH | |
2054 | */ | |
996c9314 LB |
2055 | if ((!update_len && !withdraw_len && nlris[NLRI_MP_UPDATE].length == 0) |
2056 | || (attr_parse_ret == BGP_ATTR_PARSE_EOR)) { | |
d62a17ae | 2057 | afi_t afi = 0; |
2058 | safi_t safi; | |
f009ff26 | 2059 | struct graceful_restart_info *gr_info; |
2060 | ||
2061 | /* Restarting router */ | |
36235319 QY |
2062 | if (BGP_PEER_GRACEFUL_RESTART_CAPABLE(peer) |
2063 | && BGP_PEER_RESTARTING_MODE(peer)) | |
f009ff26 | 2064 | restart = true; |
d62a17ae | 2065 | |
2066 | /* Non-MP IPv4/Unicast is a completely emtpy UPDATE - already | |
2067 | * checked | |
2068 | * update and withdraw NLRI lengths are 0. | |
2069 | */ | |
2070 | if (!attribute_len) { | |
2071 | afi = AFI_IP; | |
2072 | safi = SAFI_UNICAST; | |
2073 | } else if (attr.flag & ATTR_FLAG_BIT(BGP_ATTR_MP_UNREACH_NLRI) | |
2074 | && nlris[NLRI_MP_WITHDRAW].length == 0) { | |
2075 | afi = nlris[NLRI_MP_WITHDRAW].afi; | |
2076 | safi = nlris[NLRI_MP_WITHDRAW].safi; | |
9b9df989 DS |
2077 | } else if (attr_parse_ret == BGP_ATTR_PARSE_EOR) { |
2078 | afi = nlris[NLRI_MP_UPDATE].afi; | |
2079 | safi = nlris[NLRI_MP_UPDATE].safi; | |
d62a17ae | 2080 | } |
2081 | ||
2082 | if (afi && peer->afc[afi][safi]) { | |
e82d19a3 DS |
2083 | struct vrf *vrf = vrf_lookup_by_id(peer->bgp->vrf_id); |
2084 | ||
d62a17ae | 2085 | /* End-of-RIB received */ |
2086 | if (!CHECK_FLAG(peer->af_sflags[afi][safi], | |
2087 | PEER_STATUS_EOR_RECEIVED)) { | |
2088 | SET_FLAG(peer->af_sflags[afi][safi], | |
2089 | PEER_STATUS_EOR_RECEIVED); | |
2090 | bgp_update_explicit_eors(peer); | |
f009ff26 | 2091 | /* Update graceful restart information */ |
2092 | gr_info = &(peer->bgp->gr_info[afi][safi]); | |
2093 | if (restart) | |
2094 | gr_info->eor_received++; | |
2095 | /* If EOR received from all peers and selection | |
2096 | * deferral timer is running, cancel the timer | |
2097 | * and invoke the best path calculation | |
2098 | */ | |
36235319 QY |
2099 | if (gr_info->eor_required |
2100 | == gr_info->eor_received) { | |
2101 | if (bgp_debug_neighbor_events(peer)) | |
2102 | zlog_debug( | |
2103 | "%s %d, %s %d", | |
f009ff26 | 2104 | "EOR REQ", |
2105 | gr_info->eor_required, | |
2106 | "EOR RCV", | |
2107 | gr_info->eor_received); | |
b96b4f1c SB |
2108 | if (gr_info->t_select_deferral) { |
2109 | void *info = THREAD_ARG( | |
2110 | gr_info->t_select_deferral); | |
2111 | XFREE(MTYPE_TMP, info); | |
2112 | } | |
fa5806c3 | 2113 | THREAD_OFF(gr_info->t_select_deferral); |
f009ff26 | 2114 | gr_info->eor_required = 0; |
2115 | gr_info->eor_received = 0; | |
2116 | /* Best path selection */ | |
42c93837 DA |
2117 | bgp_best_path_select_defer(peer->bgp, |
2118 | afi, safi); | |
f009ff26 | 2119 | } |
d62a17ae | 2120 | } |
2121 | ||
2122 | /* NSF delete stale route */ | |
2123 | if (peer->nsf[afi][safi]) | |
2124 | bgp_clear_stale_route(peer, afi, safi); | |
2125 | ||
1479ed2f DA |
2126 | zlog_info( |
2127 | "%s: rcvd End-of-RIB for %s from %s in vrf %s", | |
2128 | __func__, get_afi_safi_str(afi, safi, false), | |
2129 | peer->host, vrf ? vrf->name : VRF_DEFAULT_NAME); | |
2130 | } | |
f80f838b | 2131 | } |
d62a17ae | 2132 | |
2133 | /* Everything is done. We unintern temporary structures which | |
2134 | interned in bgp_attr_parse(). */ | |
2135 | bgp_attr_unintern_sub(&attr); | |
2136 | ||
083ec940 | 2137 | peer->update_time = monotime(NULL); |
d62a17ae | 2138 | |
c385f82a MK |
2139 | /* Notify BGP Conditional advertisement scanner process */ |
2140 | peer->advmap_table_change = true; | |
2141 | ||
d8151687 | 2142 | return Receive_UPDATE_message; |
718e3744 | 2143 | } |
2144 | ||
d8151687 QY |
2145 | /** |
2146 | * Process BGP NOTIFY message for peer. | |
2147 | * | |
2148 | * @param peer | |
2149 | * @param size size of the packet | |
2150 | * @return as in summary | |
2151 | */ | |
2152 | static int bgp_notify_receive(struct peer *peer, bgp_size_t size) | |
718e3744 | 2153 | { |
10d476d4 DA |
2154 | struct bgp_notify outer = {}; |
2155 | struct bgp_notify inner = {}; | |
eea685b6 | 2156 | bool hard_reset = false; |
d62a17ae | 2157 | |
2158 | if (peer->notify.data) { | |
eea685b6 | 2159 | XFREE(MTYPE_BGP_NOTIFICATION, peer->notify.data); |
d62a17ae | 2160 | peer->notify.length = 0; |
e7ce634f | 2161 | peer->notify.hard_reset = false; |
d62a17ae | 2162 | } |
2163 | ||
eea685b6 DA |
2164 | outer.code = stream_getc(peer->curr); |
2165 | outer.subcode = stream_getc(peer->curr); | |
2166 | outer.length = size - 2; | |
2167 | outer.data = NULL; | |
2168 | outer.raw_data = NULL; | |
2169 | if (outer.length) { | |
2170 | outer.raw_data = XMALLOC(MTYPE_BGP_NOTIFICATION, outer.length); | |
2171 | memcpy(outer.raw_data, stream_pnt(peer->curr), outer.length); | |
2172 | } | |
2173 | ||
1ae314be DA |
2174 | hard_reset = |
2175 | bgp_notify_received_hard_reset(peer, outer.code, outer.subcode); | |
eea685b6 DA |
2176 | if (hard_reset && outer.length) { |
2177 | inner = bgp_notify_decapsulate_hard_reset(&outer); | |
2178 | peer->notify.hard_reset = true; | |
2179 | } else { | |
2180 | inner = outer; | |
2181 | } | |
d62a17ae | 2182 | |
2183 | /* Preserv notify code and sub code. */ | |
eea685b6 DA |
2184 | peer->notify.code = inner.code; |
2185 | peer->notify.subcode = inner.subcode; | |
d62a17ae | 2186 | /* For further diagnostic record returned Data. */ |
eea685b6 DA |
2187 | if (inner.length) { |
2188 | peer->notify.length = inner.length; | |
2189 | peer->notify.data = | |
2190 | XMALLOC(MTYPE_BGP_NOTIFICATION, inner.length); | |
2191 | memcpy(peer->notify.data, inner.raw_data, inner.length); | |
d62a17ae | 2192 | } |
2193 | ||
2194 | /* For debug */ | |
2195 | { | |
2196 | int i; | |
2197 | int first = 0; | |
2198 | char c[4]; | |
2199 | ||
eea685b6 DA |
2200 | if (inner.length) { |
2201 | inner.data = XMALLOC(MTYPE_BGP_NOTIFICATION, | |
2202 | inner.length * 3); | |
2203 | for (i = 0; i < inner.length; i++) | |
d62a17ae | 2204 | if (first) { |
552d6491 | 2205 | snprintf(c, sizeof(c), " %02x", |
424ab01d | 2206 | stream_getc(peer->curr)); |
f009ff26 | 2207 | |
eea685b6 DA |
2208 | strlcat(inner.data, c, |
2209 | inner.length * 3); | |
f009ff26 | 2210 | |
d62a17ae | 2211 | } else { |
2212 | first = 1; | |
552d6491 QY |
2213 | snprintf(c, sizeof(c), "%02x", |
2214 | stream_getc(peer->curr)); | |
f009ff26 | 2215 | |
eea685b6 DA |
2216 | strlcpy(inner.data, c, |
2217 | inner.length * 3); | |
d62a17ae | 2218 | } |
d62a17ae | 2219 | } |
2220 | ||
eea685b6 | 2221 | bgp_notify_print(peer, &inner, "received", hard_reset); |
10d476d4 | 2222 | if (inner.length) { |
eea685b6 DA |
2223 | XFREE(MTYPE_BGP_NOTIFICATION, inner.data); |
2224 | inner.length = 0; | |
2225 | } | |
2226 | if (outer.length) { | |
2227 | XFREE(MTYPE_BGP_NOTIFICATION, outer.data); | |
10d476d4 | 2228 | XFREE(MTYPE_BGP_NOTIFICATION, outer.raw_data); |
c73d2363 DA |
2229 | |
2230 | /* If this is a Hard Reset notification, we MUST free | |
2231 | * the inner (encapsulated) notification too. | |
2232 | */ | |
2233 | if (hard_reset) | |
2234 | XFREE(MTYPE_BGP_NOTIFICATION, inner.raw_data); | |
eea685b6 | 2235 | outer.length = 0; |
d62a17ae | 2236 | } |
2237 | } | |
2238 | ||
2239 | /* peer count update */ | |
0112e9e0 | 2240 | atomic_fetch_add_explicit(&peer->notify_in, 1, memory_order_relaxed); |
d62a17ae | 2241 | |
2242 | peer->last_reset = PEER_DOWN_NOTIFY_RECEIVED; | |
2243 | ||
2244 | /* We have to check for Notify with Unsupported Optional Parameter. | |
2245 | in that case we fallback to open without the capability option. | |
2246 | But this done in bgp_stop. We just mark it here to avoid changing | |
2247 | the fsm tables. */ | |
eea685b6 DA |
2248 | if (inner.code == BGP_NOTIFY_OPEN_ERR && |
2249 | inner.subcode == BGP_NOTIFY_OPEN_UNSUP_PARAM) | |
d62a17ae | 2250 | UNSET_FLAG(peer->sflags, PEER_STATUS_CAPABILITY_OPEN); |
2251 | ||
20170775 DA |
2252 | /* If Graceful-Restart N-bit (Notification) is exchanged, |
2253 | * and it's not a Hard Reset, let's retain the routes. | |
2254 | */ | |
2255 | if (bgp_has_graceful_restart_notification(peer) && !hard_reset && | |
2256 | CHECK_FLAG(peer->sflags, PEER_STATUS_NSF_MODE)) | |
2257 | SET_FLAG(peer->sflags, PEER_STATUS_NSF_WAIT); | |
2258 | ||
5cce3f05 | 2259 | bgp_peer_gr_flags_update(peer); |
36235319 QY |
2260 | BGP_GR_ROUTER_DETECT_AND_SEND_CAPABILITY_TO_ZEBRA(peer->bgp, |
2261 | peer->bgp->peer); | |
5cce3f05 | 2262 | |
d8151687 | 2263 | return Receive_NOTIFICATION_message; |
718e3744 | 2264 | } |
2265 | ||
d8151687 QY |
2266 | /** |
2267 | * Process BGP ROUTEREFRESH message for peer. | |
2268 | * | |
2269 | * @param peer | |
2270 | * @param size size of the packet | |
2271 | * @return as in summary | |
2272 | */ | |
2273 | static int bgp_route_refresh_receive(struct peer *peer, bgp_size_t size) | |
718e3744 | 2274 | { |
d62a17ae | 2275 | iana_afi_t pkt_afi; |
2276 | afi_t afi; | |
5c525538 RW |
2277 | iana_safi_t pkt_safi; |
2278 | safi_t safi; | |
d62a17ae | 2279 | struct stream *s; |
2280 | struct peer_af *paf; | |
2281 | struct update_group *updgrp; | |
2282 | struct peer *updgrp_peer; | |
9af52ccf | 2283 | uint8_t subtype; |
e1a32ec1 | 2284 | bool force_update = false; |
9af52ccf DA |
2285 | bgp_size_t msg_length = |
2286 | size - (BGP_MSG_ROUTE_REFRESH_MIN_SIZE - BGP_HEADER_SIZE); | |
d62a17ae | 2287 | |
2288 | /* If peer does not have the capability, send notification. */ | |
2289 | if (!CHECK_FLAG(peer->cap, PEER_CAP_REFRESH_ADV)) { | |
e50f7cfd | 2290 | flog_err(EC_BGP_NO_CAP, |
1c50c1c0 QY |
2291 | "%s [Error] BGP route refresh is not enabled", |
2292 | peer->host); | |
d62a17ae | 2293 | bgp_notify_send(peer, BGP_NOTIFY_HEADER_ERR, |
2294 | BGP_NOTIFY_HEADER_BAD_MESTYPE); | |
d8151687 | 2295 | return BGP_Stop; |
d62a17ae | 2296 | } |
2297 | ||
2298 | /* Status must be Established. */ | |
feb17238 | 2299 | if (!peer_established(peer)) { |
af4c2728 | 2300 | flog_err( |
e50f7cfd | 2301 | EC_BGP_INVALID_STATUS, |
d62a17ae | 2302 | "%s [Error] Route refresh packet received under status %s", |
2303 | peer->host, | |
2304 | lookup_msg(bgp_status_msg, peer->status, NULL)); | |
0e35025e | 2305 | bgp_notify_send(peer, BGP_NOTIFY_FSM_ERR, |
3893aeee | 2306 | bgp_fsm_error_subcode(peer->status)); |
d8151687 | 2307 | return BGP_Stop; |
d62a17ae | 2308 | } |
2309 | ||
424ab01d | 2310 | s = peer->curr; |
d62a17ae | 2311 | |
2312 | /* Parse packet. */ | |
2313 | pkt_afi = stream_getw(s); | |
9af52ccf | 2314 | subtype = stream_getc(s); |
d62a17ae | 2315 | pkt_safi = stream_getc(s); |
2316 | ||
d62a17ae | 2317 | /* Convert AFI, SAFI to internal values and check. */ |
2318 | if (bgp_map_afi_safi_iana2int(pkt_afi, pkt_safi, &afi, &safi)) { | |
2319 | zlog_info( | |
748a041f DS |
2320 | "%s REFRESH_REQ for unrecognized afi/safi: %s/%s - ignored", |
2321 | peer->host, iana_afi2str(pkt_afi), | |
2322 | iana_safi2str(pkt_safi)); | |
d8151687 | 2323 | return BGP_PACKET_NOOP; |
d62a17ae | 2324 | } |
2325 | ||
2326 | if (size != BGP_MSG_ROUTE_REFRESH_MIN_SIZE - BGP_HEADER_SIZE) { | |
d7c0a89a QY |
2327 | uint8_t *end; |
2328 | uint8_t when_to_refresh; | |
2329 | uint8_t orf_type; | |
2330 | uint16_t orf_len; | |
d62a17ae | 2331 | |
9af52ccf DA |
2332 | if (subtype) { |
2333 | /* If the length, excluding the fixed-size message | |
2334 | * header, of the received ROUTE-REFRESH message with | |
2335 | * Message Subtype 1 and 2 is not 4, then the BGP | |
2336 | * speaker MUST send a NOTIFICATION message with the | |
2337 | * Error Code of "ROUTE-REFRESH Message Error" and the | |
2338 | * subcode of "Invalid Message Length". | |
2339 | */ | |
2340 | if (msg_length != 4) { | |
2341 | zlog_err( | |
2342 | "%s Enhanced Route Refresh message length error", | |
2343 | peer->host); | |
2344 | bgp_notify_send( | |
2345 | peer, BGP_NOTIFY_ROUTE_REFRESH_ERR, | |
2346 | BGP_NOTIFY_ROUTE_REFRESH_INVALID_MSG_LEN); | |
2347 | } | |
2348 | ||
2349 | /* When the BGP speaker receives a ROUTE-REFRESH message | |
2350 | * with a "Message Subtype" field other than 0, 1, or 2, | |
2351 | * it MUST ignore the received ROUTE-REFRESH message. | |
2352 | */ | |
2353 | if (subtype > 2) | |
2354 | zlog_err( | |
2355 | "%s Enhanced Route Refresh invalid subtype", | |
2356 | peer->host); | |
2357 | } | |
2358 | ||
2359 | if (msg_length < 5) { | |
d62a17ae | 2360 | zlog_info("%s ORF route refresh length error", |
2361 | peer->host); | |
0e35025e DA |
2362 | bgp_notify_send(peer, BGP_NOTIFY_CEASE, |
2363 | BGP_NOTIFY_SUBCODE_UNSPECIFIC); | |
d8151687 | 2364 | return BGP_Stop; |
718e3744 | 2365 | } |
2366 | ||
d62a17ae | 2367 | when_to_refresh = stream_getc(s); |
2368 | end = stream_pnt(s) + (size - 5); | |
2369 | ||
2370 | while ((stream_pnt(s) + 2) < end) { | |
2371 | orf_type = stream_getc(s); | |
2372 | orf_len = stream_getw(s); | |
2373 | ||
2374 | /* orf_len in bounds? */ | |
2375 | if ((stream_pnt(s) + orf_len) > end) | |
2376 | break; /* XXX: Notify instead?? */ | |
2377 | if (orf_type == ORF_TYPE_PREFIX | |
2378 | || orf_type == ORF_TYPE_PREFIX_OLD) { | |
2379 | uint8_t *p_pnt = stream_pnt(s); | |
2380 | uint8_t *p_end = stream_pnt(s) + orf_len; | |
2381 | struct orf_prefix orfp; | |
d7c0a89a QY |
2382 | uint8_t common = 0; |
2383 | uint32_t seq; | |
d62a17ae | 2384 | int psize; |
2385 | char name[BUFSIZ]; | |
2386 | int ret = CMD_SUCCESS; | |
2387 | ||
2388 | if (bgp_debug_neighbor_events(peer)) { | |
2389 | zlog_debug( | |
f70c91dc DA |
2390 | "%pBP rcvd Prefixlist ORF(%d) length %d", |
2391 | peer, orf_type, orf_len); | |
d62a17ae | 2392 | } |
2393 | ||
f1aa4929 DA |
2394 | /* ORF prefix-list name */ |
2395 | snprintf(name, sizeof(name), "%s.%d.%d", | |
2396 | peer->host, afi, safi); | |
2397 | ||
d62a17ae | 2398 | /* we're going to read at least 1 byte of common |
2399 | * ORF header, | |
2400 | * and 7 bytes of ORF Address-filter entry from | |
2401 | * the stream | |
2402 | */ | |
f1aa4929 DA |
2403 | if (*p_pnt & ORF_COMMON_PART_REMOVE_ALL) { |
2404 | if (bgp_debug_neighbor_events(peer)) | |
2405 | zlog_debug( | |
2406 | "%pBP rcvd Remove-All pfxlist ORF request", | |
2407 | peer); | |
2408 | prefix_bgp_orf_remove_all(afi, name); | |
d62a17ae | 2409 | break; |
f1aa4929 | 2410 | } |
d62a17ae | 2411 | |
f1aa4929 DA |
2412 | if (orf_len < 7) |
2413 | break; | |
d62a17ae | 2414 | |
2415 | while (p_pnt < p_end) { | |
2416 | /* If the ORF entry is malformed, want | |
2417 | * to read as much of it | |
2418 | * as possible without going beyond the | |
2419 | * bounds of the entry, | |
2420 | * to maximise debug information. | |
2421 | */ | |
2422 | int ok; | |
6006b807 | 2423 | memset(&orfp, 0, sizeof(orfp)); |
d62a17ae | 2424 | common = *p_pnt++; |
2425 | /* after ++: p_pnt <= p_end */ | |
d7c0a89a QY |
2426 | ok = ((uint32_t)(p_end - p_pnt) |
2427 | >= sizeof(uint32_t)); | |
d62a17ae | 2428 | if (ok) { |
2429 | memcpy(&seq, p_pnt, | |
d7c0a89a QY |
2430 | sizeof(uint32_t)); |
2431 | p_pnt += sizeof(uint32_t); | |
d62a17ae | 2432 | orfp.seq = ntohl(seq); |
2433 | } else | |
2434 | p_pnt = p_end; | |
2435 | ||
5ca840a3 | 2436 | /* val checked in prefix_bgp_orf_set */ |
1bb379bf | 2437 | if (p_pnt < p_end) |
5ca840a3 DS |
2438 | orfp.ge = *p_pnt++; |
2439 | ||
2440 | /* val checked in prefix_bgp_orf_set */ | |
1bb379bf | 2441 | if (p_pnt < p_end) |
5ca840a3 DS |
2442 | orfp.le = *p_pnt++; |
2443 | ||
d62a17ae | 2444 | if ((ok = (p_pnt < p_end))) |
2445 | orfp.p.prefixlen = *p_pnt++; | |
5ca840a3 DS |
2446 | |
2447 | /* afi checked already */ | |
2448 | orfp.p.family = afi2family(afi); | |
2449 | ||
2450 | /* 0 if not ok */ | |
2451 | psize = PSIZE(orfp.p.prefixlen); | |
2452 | /* valid for family ? */ | |
2453 | if (psize > prefix_blen(&orfp.p)) { | |
d62a17ae | 2454 | ok = 0; |
2455 | psize = prefix_blen(&orfp.p); | |
2456 | } | |
5ca840a3 DS |
2457 | /* valid for packet ? */ |
2458 | if (psize > (p_end - p_pnt)) { | |
d62a17ae | 2459 | ok = 0; |
2460 | psize = p_end - p_pnt; | |
2461 | } | |
2462 | ||
2463 | if (psize > 0) | |
2464 | memcpy(&orfp.p.u.prefix, p_pnt, | |
2465 | psize); | |
2466 | p_pnt += psize; | |
2467 | ||
2468 | if (bgp_debug_neighbor_events(peer)) { | |
2469 | char buf[INET6_BUFSIZ]; | |
2470 | ||
2471 | zlog_debug( | |
f70c91dc DA |
2472 | "%pBP rcvd %s %s seq %u %s/%d ge %d le %d%s", |
2473 | peer, | |
d62a17ae | 2474 | (common & ORF_COMMON_PART_REMOVE |
2475 | ? "Remove" | |
2476 | : "Add"), | |
2477 | (common & ORF_COMMON_PART_DENY | |
2478 | ? "deny" | |
2479 | : "permit"), | |
2480 | orfp.seq, | |
2481 | inet_ntop( | |
2482 | orfp.p.family, | |
2483 | &orfp.p.u.prefix, | |
2484 | buf, | |
2485 | INET6_BUFSIZ), | |
2486 | orfp.p.prefixlen, | |
2487 | orfp.ge, orfp.le, | |
2488 | ok ? "" : " MALFORMED"); | |
2489 | } | |
2490 | ||
2491 | if (ok) | |
2492 | ret = prefix_bgp_orf_set( | |
2493 | name, afi, &orfp, | |
2494 | (common & ORF_COMMON_PART_DENY | |
2495 | ? 0 | |
2496 | : 1), | |
2497 | (common & ORF_COMMON_PART_REMOVE | |
2498 | ? 0 | |
2499 | : 1)); | |
2500 | ||
2501 | if (!ok || (ok && ret != CMD_SUCCESS)) { | |
2502 | zlog_info( | |
f70c91dc DA |
2503 | "%pBP Received misformatted prefixlist ORF. Remove All pfxlist", |
2504 | peer); | |
d62a17ae | 2505 | prefix_bgp_orf_remove_all(afi, |
2506 | name); | |
2507 | break; | |
2508 | } | |
2509 | } | |
2510 | ||
2511 | peer->orf_plist[afi][safi] = | |
2512 | prefix_bgp_orf_lookup(afi, name); | |
2513 | } | |
2514 | stream_forward_getp(s, orf_len); | |
718e3744 | 2515 | } |
d62a17ae | 2516 | if (bgp_debug_neighbor_events(peer)) |
f70c91dc | 2517 | zlog_debug("%pBP rcvd Refresh %s ORF request", peer, |
d62a17ae | 2518 | when_to_refresh == REFRESH_DEFER |
2519 | ? "Defer" | |
2520 | : "Immediate"); | |
2521 | if (when_to_refresh == REFRESH_DEFER) | |
d8151687 | 2522 | return BGP_PACKET_NOOP; |
d62a17ae | 2523 | } |
40d2700d | 2524 | |
d62a17ae | 2525 | /* First update is deferred until ORF or ROUTE-REFRESH is received */ |
2526 | if (CHECK_FLAG(peer->af_sflags[afi][safi], | |
2527 | PEER_STATUS_ORF_WAIT_REFRESH)) | |
2528 | UNSET_FLAG(peer->af_sflags[afi][safi], | |
2529 | PEER_STATUS_ORF_WAIT_REFRESH); | |
2530 | ||
2531 | paf = peer_af_find(peer, afi, safi); | |
2532 | if (paf && paf->subgroup) { | |
2533 | if (peer->orf_plist[afi][safi]) { | |
2534 | updgrp = PAF_UPDGRP(paf); | |
2535 | updgrp_peer = UPDGRP_PEER(updgrp); | |
2536 | updgrp_peer->orf_plist[afi][safi] = | |
2537 | peer->orf_plist[afi][safi]; | |
2538 | } | |
2539 | ||
2adac256 DA |
2540 | /* Avoid supressing duplicate routes later |
2541 | * when processing in subgroup_announce_table(). | |
2542 | */ | |
e1a32ec1 | 2543 | force_update = true; |
2adac256 | 2544 | |
d62a17ae | 2545 | /* If the peer is configured for default-originate clear the |
2546 | * SUBGRP_STATUS_DEFAULT_ORIGINATE flag so that we will | |
2547 | * re-advertise the | |
2548 | * default | |
2549 | */ | |
2550 | if (CHECK_FLAG(paf->subgroup->sflags, | |
2551 | SUBGRP_STATUS_DEFAULT_ORIGINATE)) | |
2552 | UNSET_FLAG(paf->subgroup->sflags, | |
2553 | SUBGRP_STATUS_DEFAULT_ORIGINATE); | |
718e3744 | 2554 | } |
d62a17ae | 2555 | |
9af52ccf DA |
2556 | if (subtype == BGP_ROUTE_REFRESH_BORR) { |
2557 | /* A BGP speaker that has received the Graceful Restart | |
2558 | * Capability from its neighbor MUST ignore any BoRRs for | |
2559 | * an <AFI, SAFI> from the neighbor before the speaker | |
2560 | * receives the EoR for the given <AFI, SAFI> from the | |
2561 | * neighbor. | |
2562 | */ | |
2563 | if (CHECK_FLAG(peer->cap, PEER_CAP_RESTART_RCV) | |
2564 | && !CHECK_FLAG(peer->af_sflags[afi][safi], | |
2565 | PEER_STATUS_EOR_RECEIVED)) { | |
2566 | if (bgp_debug_neighbor_events(peer)) | |
2567 | zlog_debug( | |
f70c91dc DA |
2568 | "%pBP rcvd route-refresh (BoRR) for %s/%s before EoR", |
2569 | peer, afi2str(afi), safi2str(safi)); | |
9af52ccf DA |
2570 | return BGP_PACKET_NOOP; |
2571 | } | |
2572 | ||
2573 | if (peer->t_refresh_stalepath) { | |
2574 | if (bgp_debug_neighbor_events(peer)) | |
2575 | zlog_debug( | |
f70c91dc DA |
2576 | "%pBP rcvd route-refresh (BoRR) for %s/%s, whereas BoRR already received", |
2577 | peer, afi2str(afi), safi2str(safi)); | |
9af52ccf DA |
2578 | return BGP_PACKET_NOOP; |
2579 | } | |
2580 | ||
2581 | SET_FLAG(peer->af_sflags[afi][safi], PEER_STATUS_BORR_RECEIVED); | |
2582 | UNSET_FLAG(peer->af_sflags[afi][safi], | |
2583 | PEER_STATUS_EORR_RECEIVED); | |
2584 | ||
2585 | /* When a BGP speaker receives a BoRR message from | |
2586 | * a peer, it MUST mark all the routes with the given | |
2587 | * Address Family Identifier and Subsequent Address | |
2588 | * Family Identifier, <AFI, SAFI> [RFC2918], from | |
2589 | * that peer as stale. | |
2590 | */ | |
2591 | if (peer_active_nego(peer)) { | |
2592 | SET_FLAG(peer->af_sflags[afi][safi], | |
2593 | PEER_STATUS_ENHANCED_REFRESH); | |
2594 | bgp_set_stale_route(peer, afi, safi); | |
2595 | } | |
2596 | ||
feb17238 | 2597 | if (peer_established(peer)) |
9af52ccf DA |
2598 | thread_add_timer(bm->master, |
2599 | bgp_refresh_stalepath_timer_expire, | |
2600 | paf, peer->bgp->stalepath_time, | |
2601 | &peer->t_refresh_stalepath); | |
2602 | ||
2603 | if (bgp_debug_neighbor_events(peer)) | |
2604 | zlog_debug( | |
f70c91dc DA |
2605 | "%pBP rcvd route-refresh (BoRR) for %s/%s, triggering timer for %u seconds", |
2606 | peer, afi2str(afi), safi2str(safi), | |
9af52ccf DA |
2607 | peer->bgp->stalepath_time); |
2608 | } else if (subtype == BGP_ROUTE_REFRESH_EORR) { | |
2609 | if (!peer->t_refresh_stalepath) { | |
2610 | zlog_err( | |
f70c91dc DA |
2611 | "%pBP rcvd route-refresh (EoRR) for %s/%s, whereas no BoRR received", |
2612 | peer, afi2str(afi), safi2str(safi)); | |
9af52ccf DA |
2613 | return BGP_PACKET_NOOP; |
2614 | } | |
2615 | ||
fa5806c3 | 2616 | THREAD_OFF(peer->t_refresh_stalepath); |
9af52ccf DA |
2617 | |
2618 | SET_FLAG(peer->af_sflags[afi][safi], PEER_STATUS_EORR_RECEIVED); | |
2619 | UNSET_FLAG(peer->af_sflags[afi][safi], | |
2620 | PEER_STATUS_BORR_RECEIVED); | |
2621 | ||
2622 | if (bgp_debug_neighbor_events(peer)) | |
2623 | zlog_debug( | |
f70c91dc DA |
2624 | "%pBP rcvd route-refresh (EoRR) for %s/%s, stopping BoRR timer", |
2625 | peer, afi2str(afi), safi2str(safi)); | |
9af52ccf DA |
2626 | |
2627 | if (peer->nsf[afi][safi]) | |
2628 | bgp_clear_stale_route(peer, afi, safi); | |
2629 | } else { | |
bcbeb3f9 | 2630 | if (bgp_debug_neighbor_events(peer)) |
a7d91a8c | 2631 | zlog_debug( |
f70c91dc DA |
2632 | "%pBP rcvd route-refresh (REQUEST) for %s/%s", |
2633 | peer, afi2str(afi), safi2str(safi)); | |
bcbeb3f9 | 2634 | |
9af52ccf DA |
2635 | /* In response to a "normal route refresh request" from the |
2636 | * peer, the speaker MUST send a BoRR message. | |
2637 | */ | |
2638 | if (CHECK_FLAG(peer->cap, PEER_CAP_ENHANCED_RR_RCV)) { | |
2639 | /* For a BGP speaker that supports the BGP Graceful | |
2640 | * Restart, it MUST NOT send a BoRR for an <AFI, SAFI> | |
2641 | * to a neighbor before it sends the EoR for the | |
2642 | * <AFI, SAFI> to the neighbor. | |
2643 | */ | |
2644 | if (!CHECK_FLAG(peer->af_sflags[afi][safi], | |
2645 | PEER_STATUS_EOR_SEND)) { | |
2646 | if (bgp_debug_neighbor_events(peer)) | |
2647 | zlog_debug( | |
f70c91dc DA |
2648 | "%pBP rcvd route-refresh (REQUEST) for %s/%s before EoR", |
2649 | peer, afi2str(afi), | |
2650 | safi2str(safi)); | |
a783cc05 XL |
2651 | /* Can't send BoRR now, postpone after EoR */ |
2652 | SET_FLAG(peer->af_sflags[afi][safi], | |
2653 | PEER_STATUS_REFRESH_PENDING); | |
9af52ccf DA |
2654 | return BGP_PACKET_NOOP; |
2655 | } | |
2656 | ||
2657 | bgp_route_refresh_send(peer, afi, safi, 0, 0, 0, | |
2658 | BGP_ROUTE_REFRESH_BORR); | |
2659 | ||
2660 | if (bgp_debug_neighbor_events(peer)) | |
2661 | zlog_debug( | |
f70c91dc DA |
2662 | "%pBP sending route-refresh (BoRR) for %s/%s", |
2663 | peer, afi2str(afi), safi2str(safi)); | |
9af52ccf DA |
2664 | |
2665 | /* Set flag Ready-To-Send to know when we can send EoRR | |
2666 | * message. | |
2667 | */ | |
2668 | SET_FLAG(peer->af_sflags[afi][safi], | |
2669 | PEER_STATUS_BORR_SEND); | |
2670 | UNSET_FLAG(peer->af_sflags[afi][safi], | |
2671 | PEER_STATUS_EORR_SEND); | |
2672 | } | |
2673 | } | |
2674 | ||
d62a17ae | 2675 | /* Perform route refreshment to the peer */ |
e1a32ec1 | 2676 | bgp_announce_route(peer, afi, safi, force_update); |
d8151687 QY |
2677 | |
2678 | /* No FSM action necessary */ | |
2679 | return BGP_PACKET_NOOP; | |
718e3744 | 2680 | } |
2681 | ||
d8151687 QY |
2682 | /** |
2683 | * Parse BGP CAPABILITY message for peer. | |
2684 | * | |
2685 | * @param peer | |
2686 | * @param size size of the packet | |
2687 | * @return as in summary | |
2688 | */ | |
d7c0a89a | 2689 | static int bgp_capability_msg_parse(struct peer *peer, uint8_t *pnt, |
d62a17ae | 2690 | bgp_size_t length) |
718e3744 | 2691 | { |
d7c0a89a | 2692 | uint8_t *end; |
d62a17ae | 2693 | struct capability_mp_data mpc; |
2694 | struct capability_header *hdr; | |
d7c0a89a | 2695 | uint8_t action; |
d62a17ae | 2696 | iana_afi_t pkt_afi; |
2697 | afi_t afi; | |
5c525538 RW |
2698 | iana_safi_t pkt_safi; |
2699 | safi_t safi; | |
d62a17ae | 2700 | |
2701 | end = pnt + length; | |
2702 | ||
2703 | while (pnt < end) { | |
2704 | /* We need at least action, capability code and capability | |
2705 | * length. */ | |
2706 | if (pnt + 3 > end) { | |
2707 | zlog_info("%s Capability length error", peer->host); | |
0e35025e DA |
2708 | bgp_notify_send(peer, BGP_NOTIFY_CEASE, |
2709 | BGP_NOTIFY_SUBCODE_UNSPECIFIC); | |
d8151687 | 2710 | return BGP_Stop; |
d62a17ae | 2711 | } |
2712 | action = *pnt; | |
2713 | hdr = (struct capability_header *)(pnt + 1); | |
2714 | ||
2715 | /* Action value check. */ | |
2716 | if (action != CAPABILITY_ACTION_SET | |
2717 | && action != CAPABILITY_ACTION_UNSET) { | |
2718 | zlog_info("%s Capability Action Value error %d", | |
2719 | peer->host, action); | |
0e35025e DA |
2720 | bgp_notify_send(peer, BGP_NOTIFY_CEASE, |
2721 | BGP_NOTIFY_SUBCODE_UNSPECIFIC); | |
d8151687 | 2722 | return BGP_Stop; |
d62a17ae | 2723 | } |
2724 | ||
2725 | if (bgp_debug_neighbor_events(peer)) | |
2726 | zlog_debug( | |
2727 | "%s CAPABILITY has action: %d, code: %u, length %u", | |
2728 | peer->host, action, hdr->code, hdr->length); | |
2729 | ||
ff6db102 DS |
2730 | if (hdr->length < sizeof(struct capability_mp_data)) { |
2731 | zlog_info( | |
2732 | "%pBP Capability structure is not properly filled out, expected at least %zu bytes but header length specified is %d", | |
2733 | peer, sizeof(struct capability_mp_data), | |
2734 | hdr->length); | |
2735 | return BGP_Stop; | |
2736 | } | |
2737 | ||
d62a17ae | 2738 | /* Capability length check. */ |
2739 | if ((pnt + hdr->length + 3) > end) { | |
2740 | zlog_info("%s Capability length error", peer->host); | |
0e35025e DA |
2741 | bgp_notify_send(peer, BGP_NOTIFY_CEASE, |
2742 | BGP_NOTIFY_SUBCODE_UNSPECIFIC); | |
d8151687 | 2743 | return BGP_Stop; |
d62a17ae | 2744 | } |
2745 | ||
2746 | /* Fetch structure to the byte stream. */ | |
2747 | memcpy(&mpc, pnt + 3, sizeof(struct capability_mp_data)); | |
d2b6417b | 2748 | pnt += hdr->length + 3; |
d62a17ae | 2749 | |
2750 | /* We know MP Capability Code. */ | |
2751 | if (hdr->code == CAPABILITY_CODE_MP) { | |
2752 | pkt_afi = ntohs(mpc.afi); | |
2753 | pkt_safi = mpc.safi; | |
2754 | ||
2755 | /* Ignore capability when override-capability is set. */ | |
2756 | if (CHECK_FLAG(peer->flags, | |
2757 | PEER_FLAG_OVERRIDE_CAPABILITY)) | |
2758 | continue; | |
2759 | ||
2760 | /* Convert AFI, SAFI to internal values. */ | |
2761 | if (bgp_map_afi_safi_iana2int(pkt_afi, pkt_safi, &afi, | |
2762 | &safi)) { | |
2763 | if (bgp_debug_neighbor_events(peer)) | |
2764 | zlog_debug( | |
3efd0893 | 2765 | "%s Dynamic Capability MP_EXT afi/safi invalid (%s/%s)", |
748a041f DS |
2766 | peer->host, |
2767 | iana_afi2str(pkt_afi), | |
2768 | iana_safi2str(pkt_safi)); | |
d62a17ae | 2769 | continue; |
2770 | } | |
2771 | ||
2772 | /* Address family check. */ | |
2773 | if (bgp_debug_neighbor_events(peer)) | |
2774 | zlog_debug( | |
c386cdd8 | 2775 | "%s CAPABILITY has %s MP_EXT CAP for afi/safi: %s/%s", |
d62a17ae | 2776 | peer->host, |
2777 | action == CAPABILITY_ACTION_SET | |
2778 | ? "Advertising" | |
2779 | : "Removing", | |
c386cdd8 DA |
2780 | iana_afi2str(pkt_afi), |
2781 | iana_safi2str(pkt_safi)); | |
d62a17ae | 2782 | |
2783 | if (action == CAPABILITY_ACTION_SET) { | |
2784 | peer->afc_recv[afi][safi] = 1; | |
2785 | if (peer->afc[afi][safi]) { | |
2786 | peer->afc_nego[afi][safi] = 1; | |
e1a32ec1 DS |
2787 | bgp_announce_route(peer, afi, safi, |
2788 | false); | |
d62a17ae | 2789 | } |
2790 | } else { | |
2791 | peer->afc_recv[afi][safi] = 0; | |
2792 | peer->afc_nego[afi][safi] = 0; | |
2793 | ||
2794 | if (peer_active_nego(peer)) | |
2795 | bgp_clear_route(peer, afi, safi); | |
2796 | else | |
d8151687 | 2797 | return BGP_Stop; |
d62a17ae | 2798 | } |
2799 | } else { | |
ade6974d | 2800 | flog_warn( |
e50f7cfd | 2801 | EC_BGP_UNRECOGNIZED_CAPABILITY, |
ade6974d QY |
2802 | "%s unrecognized capability code: %d - ignored", |
2803 | peer->host, hdr->code); | |
d62a17ae | 2804 | } |
d62a17ae | 2805 | } |
d8151687 QY |
2806 | |
2807 | /* No FSM action necessary */ | |
2808 | return BGP_PACKET_NOOP; | |
718e3744 | 2809 | } |
2810 | ||
d8151687 QY |
2811 | /** |
2812 | * Parse BGP CAPABILITY message for peer. | |
01b7ce2d | 2813 | * |
d8151687 QY |
2814 | * Exported for unit testing. |
2815 | * | |
2816 | * @param peer | |
2817 | * @param size size of the packet | |
2818 | * @return as in summary | |
01b7ce2d | 2819 | */ |
d62a17ae | 2820 | int bgp_capability_receive(struct peer *peer, bgp_size_t size) |
718e3744 | 2821 | { |
d7c0a89a | 2822 | uint8_t *pnt; |
d62a17ae | 2823 | |
2824 | /* Fetch pointer. */ | |
424ab01d | 2825 | pnt = stream_pnt(peer->curr); |
d62a17ae | 2826 | |
2827 | if (bgp_debug_neighbor_events(peer)) | |
2828 | zlog_debug("%s rcv CAPABILITY", peer->host); | |
2829 | ||
2830 | /* If peer does not have the capability, send notification. */ | |
2831 | if (!CHECK_FLAG(peer->cap, PEER_CAP_DYNAMIC_ADV)) { | |
e50f7cfd | 2832 | flog_err(EC_BGP_NO_CAP, |
1c50c1c0 QY |
2833 | "%s [Error] BGP dynamic capability is not enabled", |
2834 | peer->host); | |
d62a17ae | 2835 | bgp_notify_send(peer, BGP_NOTIFY_HEADER_ERR, |
2836 | BGP_NOTIFY_HEADER_BAD_MESTYPE); | |
d8151687 | 2837 | return BGP_Stop; |
d62a17ae | 2838 | } |
2839 | ||
2840 | /* Status must be Established. */ | |
feb17238 | 2841 | if (!peer_established(peer)) { |
af4c2728 | 2842 | flog_err( |
e50f7cfd | 2843 | EC_BGP_NO_CAP, |
d62a17ae | 2844 | "%s [Error] Dynamic capability packet received under status %s", |
2845 | peer->host, | |
2846 | lookup_msg(bgp_status_msg, peer->status, NULL)); | |
0e35025e | 2847 | bgp_notify_send(peer, BGP_NOTIFY_FSM_ERR, |
3893aeee | 2848 | bgp_fsm_error_subcode(peer->status)); |
d8151687 | 2849 | return BGP_Stop; |
d62a17ae | 2850 | } |
2851 | ||
2852 | /* Parse packet. */ | |
2853 | return bgp_capability_msg_parse(peer, pnt, size); | |
718e3744 | 2854 | } |
6b0655a2 | 2855 | |
d8151687 QY |
2856 | /** |
2857 | * Processes a peer's input buffer. | |
2858 | * | |
2859 | * This function sidesteps the event loop and directly calls bgp_event_update() | |
2860 | * after processing each BGP message. This is necessary to ensure proper | |
2861 | * ordering of FSM events and unifies the behavior that was present previously, | |
2862 | * whereby some of the packet handling functions would update the FSM and some | |
2863 | * would not, making event flow difficult to understand. Please think twice | |
2864 | * before hacking this. | |
2865 | * | |
2866 | * Thread type: THREAD_EVENT | |
2867 | * @param thread | |
2868 | * @return 0 | |
2869 | */ | |
cc9f21da | 2870 | void bgp_process_packet(struct thread *thread) |
718e3744 | 2871 | { |
424ab01d | 2872 | /* Yes first of all get peer pointer. */ |
d8151687 QY |
2873 | struct peer *peer; // peer |
2874 | uint32_t rpkt_quanta_old; // how many packets to read | |
2875 | int fsm_update_result; // return code of bgp_event_update() | |
2876 | int mprc; // message processing return code | |
555e09d4 | 2877 | |
424ab01d | 2878 | peer = THREAD_ARG(thread); |
555e09d4 QY |
2879 | rpkt_quanta_old = atomic_load_explicit(&peer->bgp->rpkt_quanta, |
2880 | memory_order_relaxed); | |
e3c7270d | 2881 | fsm_update_result = 0; |
555e09d4 | 2882 | |
424ab01d | 2883 | /* Guard against scheduled events that occur after peer deletion. */ |
9eb217ff | 2884 | if (peer->status == Deleted || peer->status == Clearing) |
cc9f21da | 2885 | return; |
718e3744 | 2886 | |
555e09d4 | 2887 | unsigned int processed = 0; |
d62a17ae | 2888 | |
555e09d4 | 2889 | while (processed < rpkt_quanta_old) { |
d7c0a89a | 2890 | uint8_t type = 0; |
9eb217ff QY |
2891 | bgp_size_t size; |
2892 | char notify_data_length[2]; | |
d62a17ae | 2893 | |
cb1991af | 2894 | frr_with_mutex (&peer->io_mtx) { |
9eb217ff QY |
2895 | peer->curr = stream_fifo_pop(peer->ibuf); |
2896 | } | |
d62a17ae | 2897 | |
9eb217ff | 2898 | if (peer->curr == NULL) // no packets to process, hmm... |
cc9f21da | 2899 | return; |
d62a17ae | 2900 | |
9eb217ff QY |
2901 | /* skip the marker and copy the packet length */ |
2902 | stream_forward_getp(peer->curr, BGP_MARKER_SIZE); | |
2903 | memcpy(notify_data_length, stream_pnt(peer->curr), 2); | |
2904 | ||
2905 | /* read in the packet length and type */ | |
2906 | size = stream_getw(peer->curr); | |
2907 | type = stream_getc(peer->curr); | |
2908 | ||
584470fb | 2909 | hook_call(bgp_packet_dump, peer, type, size, peer->curr); |
9eb217ff QY |
2910 | |
2911 | /* adjust size to exclude the marker + length + type */ | |
2912 | size -= BGP_HEADER_SIZE; | |
2913 | ||
2914 | /* Read rest of the packet and call each sort of packet routine | |
2915 | */ | |
2916 | switch (type) { | |
2917 | case BGP_MSG_OPEN: | |
c7bb4f00 | 2918 | frrtrace(2, frr_bgp, open_process, peer, size); |
0112e9e0 QY |
2919 | atomic_fetch_add_explicit(&peer->open_in, 1, |
2920 | memory_order_relaxed); | |
d8151687 QY |
2921 | mprc = bgp_open_receive(peer, size); |
2922 | if (mprc == BGP_Stop) | |
af4c2728 | 2923 | flog_err( |
e50f7cfd | 2924 | EC_BGP_PKT_OPEN, |
d8151687 | 2925 | "%s: BGP OPEN receipt failed for peer: %s", |
0767b4f3 | 2926 | __func__, peer->host); |
9eb217ff QY |
2927 | break; |
2928 | case BGP_MSG_UPDATE: | |
c7bb4f00 | 2929 | frrtrace(2, frr_bgp, update_process, peer, size); |
0112e9e0 QY |
2930 | atomic_fetch_add_explicit(&peer->update_in, 1, |
2931 | memory_order_relaxed); | |
9eb217ff | 2932 | peer->readtime = monotime(NULL); |
d8151687 QY |
2933 | mprc = bgp_update_receive(peer, size); |
2934 | if (mprc == BGP_Stop) | |
af4c2728 | 2935 | flog_err( |
e50f7cfd | 2936 | EC_BGP_UPDATE_RCV, |
d8151687 | 2937 | "%s: BGP UPDATE receipt failed for peer: %s", |
0767b4f3 | 2938 | __func__, peer->host); |
9eb217ff QY |
2939 | break; |
2940 | case BGP_MSG_NOTIFY: | |
c7bb4f00 | 2941 | frrtrace(2, frr_bgp, notification_process, peer, size); |
0112e9e0 QY |
2942 | atomic_fetch_add_explicit(&peer->notify_in, 1, |
2943 | memory_order_relaxed); | |
d8151687 QY |
2944 | mprc = bgp_notify_receive(peer, size); |
2945 | if (mprc == BGP_Stop) | |
af4c2728 | 2946 | flog_err( |
e50f7cfd | 2947 | EC_BGP_NOTIFY_RCV, |
d8151687 | 2948 | "%s: BGP NOTIFY receipt failed for peer: %s", |
0767b4f3 | 2949 | __func__, peer->host); |
9eb217ff QY |
2950 | break; |
2951 | case BGP_MSG_KEEPALIVE: | |
c7bb4f00 | 2952 | frrtrace(2, frr_bgp, keepalive_process, peer, size); |
9eb217ff | 2953 | peer->readtime = monotime(NULL); |
0112e9e0 QY |
2954 | atomic_fetch_add_explicit(&peer->keepalive_in, 1, |
2955 | memory_order_relaxed); | |
d8151687 QY |
2956 | mprc = bgp_keepalive_receive(peer, size); |
2957 | if (mprc == BGP_Stop) | |
af4c2728 | 2958 | flog_err( |
e50f7cfd | 2959 | EC_BGP_KEEP_RCV, |
d8151687 | 2960 | "%s: BGP KEEPALIVE receipt failed for peer: %s", |
0767b4f3 | 2961 | __func__, peer->host); |
9eb217ff QY |
2962 | break; |
2963 | case BGP_MSG_ROUTE_REFRESH_NEW: | |
2964 | case BGP_MSG_ROUTE_REFRESH_OLD: | |
c7bb4f00 | 2965 | frrtrace(2, frr_bgp, refresh_process, peer, size); |
0112e9e0 QY |
2966 | atomic_fetch_add_explicit(&peer->refresh_in, 1, |
2967 | memory_order_relaxed); | |
d8151687 QY |
2968 | mprc = bgp_route_refresh_receive(peer, size); |
2969 | if (mprc == BGP_Stop) | |
af4c2728 | 2970 | flog_err( |
e50f7cfd | 2971 | EC_BGP_RFSH_RCV, |
d8151687 | 2972 | "%s: BGP ROUTEREFRESH receipt failed for peer: %s", |
0767b4f3 | 2973 | __func__, peer->host); |
9eb217ff QY |
2974 | break; |
2975 | case BGP_MSG_CAPABILITY: | |
c7bb4f00 | 2976 | frrtrace(2, frr_bgp, capability_process, peer, size); |
0112e9e0 QY |
2977 | atomic_fetch_add_explicit(&peer->dynamic_cap_in, 1, |
2978 | memory_order_relaxed); | |
d8151687 QY |
2979 | mprc = bgp_capability_receive(peer, size); |
2980 | if (mprc == BGP_Stop) | |
af4c2728 | 2981 | flog_err( |
e50f7cfd | 2982 | EC_BGP_CAP_RCV, |
d8151687 | 2983 | "%s: BGP CAPABILITY receipt failed for peer: %s", |
0767b4f3 | 2984 | __func__, peer->host); |
9eb217ff | 2985 | break; |
e3c7270d | 2986 | default: |
db878db0 QY |
2987 | /* Suppress uninitialized variable warning */ |
2988 | mprc = 0; | |
5041dc4f | 2989 | (void)mprc; |
becedef6 QY |
2990 | /* |
2991 | * The message type should have been sanitized before | |
2992 | * we ever got here. Receipt of a message with an | |
2993 | * invalid header at this point is indicative of a | |
2994 | * security issue. | |
2995 | */ | |
e3c7270d | 2996 | assert (!"Message of invalid type received during input processing"); |
9eb217ff QY |
2997 | } |
2998 | ||
d8151687 QY |
2999 | /* delete processed packet */ |
3000 | stream_free(peer->curr); | |
3001 | peer->curr = NULL; | |
3002 | processed++; | |
9eb217ff | 3003 | |
d8151687 QY |
3004 | /* Update FSM */ |
3005 | if (mprc != BGP_PACKET_NOOP) | |
3006 | fsm_update_result = bgp_event_update(peer, mprc); | |
e3c7270d QY |
3007 | else |
3008 | continue; | |
d8151687 | 3009 | |
becedef6 QY |
3010 | /* |
3011 | * If peer was deleted, do not process any more packets. This | |
3012 | * is usually due to executing BGP_Stop or a stub deletion. | |
3013 | */ | |
d8151687 QY |
3014 | if (fsm_update_result == FSM_PEER_TRANSFERRED |
3015 | || fsm_update_result == FSM_PEER_STOPPED) | |
3016 | break; | |
d62a17ae | 3017 | } |
3018 | ||
d8151687 QY |
3019 | if (fsm_update_result != FSM_PEER_TRANSFERRED |
3020 | && fsm_update_result != FSM_PEER_STOPPED) { | |
cb1991af | 3021 | frr_with_mutex (&peer->io_mtx) { |
becedef6 QY |
3022 | // more work to do, come back later |
3023 | if (peer->ibuf->count > 0) | |
e0d550df | 3024 | thread_add_event( |
4af76660 QY |
3025 | bm->master, bgp_process_packet, peer, 0, |
3026 | &peer->t_process_packet); | |
d8151687 | 3027 | } |
718e3744 | 3028 | } |
718e3744 | 3029 | } |
9e3b51a7 | 3030 | |
3031 | /* Send EOR when routes are processed by selection deferral timer */ | |
3032 | void bgp_send_delayed_eor(struct bgp *bgp) | |
3033 | { | |
3034 | struct peer *peer; | |
3035 | struct listnode *node, *nnode; | |
3036 | ||
3037 | /* EOR message sent in bgp_write_proceed_actions */ | |
3038 | for (ALL_LIST_ELEMENTS(bgp->peer, node, nnode, peer)) | |
3039 | bgp_write_proceed_actions(peer); | |
3040 | } | |
6af96fa3 MS |
3041 | |
3042 | /* | |
3043 | * Task callback to handle socket error encountered in the io pthread. We avoid | |
3044 | * having the io pthread try to enqueue fsm events or mess with the peer | |
3045 | * struct. | |
3046 | */ | |
cc9f21da | 3047 | void bgp_packet_process_error(struct thread *thread) |
6af96fa3 MS |
3048 | { |
3049 | struct peer *peer; | |
3050 | int code; | |
3051 | ||
3052 | peer = THREAD_ARG(thread); | |
3053 | code = THREAD_VAL(thread); | |
3054 | ||
3055 | if (bgp_debug_neighbor_events(peer)) | |
3056 | zlog_debug("%s [Event] BGP error %d on fd %d", | |
046bb347 | 3057 | peer->host, code, peer->fd); |
6af96fa3 MS |
3058 | |
3059 | /* Closed connection or error on the socket */ | |
feb17238 | 3060 | if (peer_established(peer)) { |
6af96fa3 MS |
3061 | if ((CHECK_FLAG(peer->flags, PEER_FLAG_GRACEFUL_RESTART) |
3062 | || CHECK_FLAG(peer->flags, | |
3063 | PEER_FLAG_GRACEFUL_RESTART_HELPER)) | |
3064 | && CHECK_FLAG(peer->sflags, PEER_STATUS_NSF_MODE)) { | |
3065 | peer->last_reset = PEER_DOWN_NSF_CLOSE_SESSION; | |
3066 | SET_FLAG(peer->sflags, PEER_STATUS_NSF_WAIT); | |
3067 | } else | |
3068 | peer->last_reset = PEER_DOWN_CLOSE_SESSION; | |
3069 | } | |
3070 | ||
3071 | bgp_event_update(peer, code); | |
6af96fa3 | 3072 | } |