]> git.ipfire.org Git - thirdparty/bird.git/blob - proto/bgp/packets.c
projects / thirdparty / bird.git / blob
? search:


c3bd600a10d9b0183c505ed293aa87885113c958
[thirdparty/bird.git] / proto / bgp / packets.c
1 /*
2 * BIRD -- BGP Packet Processing
3 *
4 * (c) 2000 Martin Mares <mj@ucw.cz>
5 * (c) 2008--2016 Ondrej Zajicek <santiago@crfreenet.org>
6 * (c) 2008--2016 CZ.NIC z.s.p.o.
7 *
8 * Can be freely distributed and used under the terms of the GNU GPL.
9 */
10
11 #undef LOCAL_DEBUG
12
13 #include <stdlib.h>
14
15 #include "nest/bird.h"
16 #include "nest/iface.h"
17 #include "nest/protocol.h"
18 #include "nest/route.h"
19 #include "nest/attrs.h"
20 #include "proto/mrt/mrt.h"
21 #include "conf/conf.h"
22 #include "lib/unaligned.h"
23 #include "lib/flowspec.h"
24 #include "lib/socket.h"
25
26 #include "nest/cli.h"
27
28 #include "bgp.h"
29
30
31 #define BGP_RR_REQUEST 0
32 #define BGP_RR_BEGIN 1
33 #define BGP_RR_END 2
34
35 #define BGP_NLRI_MAX (4 + 1 + 32)
36
37 #define BGP_MPLS_BOS 1 /* Bottom-of-stack bit */
38 #define BGP_MPLS_MAX 10 /* Max number of labels that 24*n <= 255 */
39 #define BGP_MPLS_NULL 3 /* Implicit NULL label */
40 #define BGP_MPLS_MAGIC 0x800000 /* Magic withdraw label value, RFC 3107 3 */
41
42
43 static struct tbf rl_rcv_update = TBF_DEFAULT_LOG_LIMITS;
44 static struct tbf rl_snd_update = TBF_DEFAULT_LOG_LIMITS;
45
46 /* Table for state -> RFC 6608 FSM error subcodes */
47 static byte fsm_err_subcode[BS_MAX] = {
48 [BS_OPENSENT] = 1,
49 [BS_OPENCONFIRM] = 2,
50 [BS_ESTABLISHED] = 3
51 };
52
53
54 static struct bgp_channel *
55 bgp_get_channel(struct bgp_proto *p, u32 afi)
56 {
57 uint i;
58
59 for (i = 0; i < p->channel_count; i++)
60 if (p->afi_map[i] == afi)
61 return p->channel_map[i];
62
63 return NULL;
64 }
65
66 static inline void
67 put_af3(byte *buf, u32 id)
68 {
69 put_u16(buf, id >> 16);
70 buf[2] = id & 0xff;
71 }
72
73 static inline void
74 put_af4(byte *buf, u32 id)
75 {
76 put_u16(buf, id >> 16);
77 buf[2] = 0;
78 buf[3] = id & 0xff;
79 }
80
81 static inline u32
82 get_af3(byte *buf)
83 {
84 return (get_u16(buf) << 16) | buf[2];
85 }
86
87 static inline u32
88 get_af4(byte *buf)
89 {
90 return (get_u16(buf) << 16) | buf[3];
91 }
92
93 static void
94 init_mrt_bgp_data(struct bgp_conn *conn, struct mrt_bgp_data *d)
95 {
96 struct bgp_proto *p = conn->bgp;
97 int p_ok = conn->state >= BS_OPENCONFIRM;
98
99 memset(d, 0, sizeof(struct mrt_bgp_data));
100 d->peer_as = p->remote_as;
101 d->local_as = p->local_as;
102 d->index = (p->neigh && p->neigh->iface) ? p->neigh->iface->index : 0;
103 d->af = ipa_is_ip4(p->remote_ip) ? BGP_AFI_IPV4 : BGP_AFI_IPV6;
104 d->peer_ip = conn->sk ? conn->sk->daddr : IPA_NONE;
105 d->local_ip = conn->sk ? conn->sk->saddr : IPA_NONE;
106 d->as4 = p_ok ? p->as4_session : 0;
107 }
108
109 static uint bgp_find_update_afi(byte *pos, uint len);
110
111 static int
112 bgp_estimate_add_path(struct bgp_proto *p, byte *pkt, uint len)
113 {
114 /* No need to estimate it for other messages than UPDATE */
115 if (pkt[18] != PKT_UPDATE)
116 return 0;
117
118 /* 1 -> no channel, 2 -> all channels, 3 -> some channels */
119 if (p->summary_add_path_rx < 3)
120 return p->summary_add_path_rx == 2;
121
122 uint afi = bgp_find_update_afi(pkt, len);
123 struct bgp_channel *c = bgp_get_channel(p, afi);
124 if (!c)
125 {
126 /* Either frame error (if !afi) or unknown AFI/SAFI,
127 will be reported later in regular parsing */
128 BGP_TRACE(D_PACKETS, "MRT processing noticed invalid packet");
129 return 0;
130 }
131
132 return c->add_path_rx;
133 }
134
135 static void
136 bgp_dump_message(struct bgp_conn *conn, byte *pkt, uint len)
137 {
138 struct mrt_bgp_data d;
139 init_mrt_bgp_data(conn, &d);
140
141 d.message = pkt;
142 d.msg_len = len;
143 d.add_path = bgp_estimate_add_path(conn->bgp, pkt, len);
144
145 mrt_dump_bgp_message(&d);
146 }
147
148 void
149 bgp_dump_state_change(struct bgp_conn *conn, uint old, uint new)
150 {
151 struct mrt_bgp_data d;
152 init_mrt_bgp_data(conn, &d);
153
154 d.old_state = old;
155 d.new_state = new;
156
157 mrt_dump_bgp_state_change(&d);
158 }
159
160 static byte *
161 bgp_create_notification(struct bgp_conn *conn, byte *buf)
162 {
163 struct bgp_proto *p = conn->bgp;
164
165 BGP_TRACE(D_PACKETS, "Sending NOTIFICATION(code=%d.%d)", conn->notify_code, conn->notify_subcode);
166 buf[0] = conn->notify_code;
167 buf[1] = conn->notify_subcode;
168 memcpy(buf+2, conn->notify_data, conn->notify_size);
169 return buf + 2 + conn->notify_size;
170 }
171
172
173 /* Capability negotiation as per RFC 5492 */
174
175 const struct bgp_af_caps *
176 bgp_find_af_caps(struct bgp_caps *caps, u32 afi)
177 {
178 struct bgp_af_caps *ac;
179
180 WALK_AF_CAPS(caps, ac)
181 if (ac->afi == afi)
182 return ac;
183
184 return NULL;
185 }
186
187 static struct bgp_af_caps *
188 bgp_get_af_caps(struct bgp_caps **pcaps, u32 afi)
189 {
190 struct bgp_caps *caps = *pcaps;
191 struct bgp_af_caps *ac;
192
193 WALK_AF_CAPS(caps, ac)
194 if (ac->afi == afi)
195 return ac;
196
197 uint n = caps->af_count;
198 if (uint_is_pow2(n))
199 *pcaps = caps = mb_realloc(caps, sizeof(struct bgp_caps) +
200 (2 * n) * sizeof(struct bgp_af_caps));
201
202 ac = &caps->af_data[caps->af_count++];
203 memset(ac, 0, sizeof(struct bgp_af_caps));
204 ac->afi = afi;
205
206 return ac;
207 }
208
209 static int
210 bgp_af_caps_cmp(const void *X, const void *Y)
211 {
212 const struct bgp_af_caps *x = X, *y = Y;
213 return (x->afi < y->afi) ? -1 : (x->afi > y->afi) ? 1 : 0;
214 }
215
216
217 void
218 bgp_prepare_capabilities(struct bgp_conn *conn)
219 {
220 struct bgp_proto *p = conn->bgp;
221 struct bgp_channel *c;
222 struct bgp_caps *caps;
223 struct bgp_af_caps *ac;
224
225 if (!p->cf->capabilities)
226 {
227 /* Just prepare empty local_caps */
228 conn->local_caps = mb_allocz(p->p.pool, sizeof(struct bgp_caps));
229 return;
230 }
231
232 /* Prepare bgp_caps structure */
233 int n = list_length(&p->p.channels);
234 caps = mb_allocz(p->p.pool, sizeof(struct bgp_caps) + n * sizeof(struct bgp_af_caps));
235 conn->local_caps = caps;
236
237 caps->as4_support = p->cf->enable_as4;
238 caps->ext_messages = p->cf->enable_extended_messages;
239 caps->route_refresh = p->cf->enable_refresh;
240 caps->enhanced_refresh = p->cf->enable_refresh;
241
242 if (caps->as4_support)
243 caps->as4_number = p->public_as;
244
245 if (p->cf->gr_mode)
246 {
247 caps->gr_aware = 1;
248 caps->gr_time = p->cf->gr_time;
249 caps->gr_flags = p->p.gr_recovery ? BGP_GRF_RESTART : 0;
250 }
251
252 if (p->cf->llgr_mode)
253 caps->llgr_aware = 1;
254
255 /* Allocate and fill per-AF fields */
256 WALK_LIST(c, p->p.channels)
257 {
258 ac = &caps->af_data[caps->af_count++];
259 ac->afi = c->afi;
260 ac->ready = 1;
261
262 ac->ext_next_hop = bgp_channel_is_ipv4(c) && c->cf->ext_next_hop;
263 caps->any_ext_next_hop |= ac->ext_next_hop;
264
265 ac->add_path = c->cf->add_path;
266 caps->any_add_path |= ac->add_path;
267
268 if (c->cf->gr_able)
269 {
270 ac->gr_able = 1;
271
272 if (p->p.gr_recovery)
273 ac->gr_af_flags |= BGP_GRF_FORWARDING;
274 }
275
276 if (c->cf->llgr_able)
277 {
278 ac->llgr_able = 1;
279 ac->llgr_time = c->cf->llgr_time;
280
281 if (p->p.gr_recovery)
282 ac->llgr_flags |= BGP_LLGRF_FORWARDING;
283 }
284 }
285
286 /* Sort capability fields by AFI/SAFI */
287 qsort(caps->af_data, caps->af_count, sizeof(struct bgp_af_caps), bgp_af_caps_cmp);
288 }
289
290 static byte *
291 bgp_write_capabilities(struct bgp_conn *conn, byte *buf)
292 {
293 struct bgp_proto *p = conn->bgp;
294 struct bgp_caps *caps = conn->local_caps;
295 struct bgp_af_caps *ac;
296 byte *buf_head = buf;
297 byte *data;
298
299 /* Create capability list in buffer */
300
301 /*
302 * Note that max length is ~ 22+21*af_count. With max 12 channels that is
303 * 274. We are limited just by buffer size (4096, minus header), as we support
304 * extended optional parameres. Therefore, we have enough space for expansion.
305 */
306
307 WALK_AF_CAPS(caps, ac)
308 if (ac->ready)
309 {
310 *buf++ = 1; /* Capability 1: Multiprotocol extensions */
311 *buf++ = 4; /* Capability data length */
312 put_af4(buf, ac->afi);
313 buf += 4;
314 }
315
316 if (caps->route_refresh)
317 {
318 *buf++ = 2; /* Capability 2: Support for route refresh */
319 *buf++ = 0; /* Capability data length */
320 }
321
322 if (caps->any_ext_next_hop)
323 {
324 *buf++ = 5; /* Capability 5: Support for extended next hop */
325 *buf++ = 0; /* Capability data length, will be fixed later */
326 data = buf;
327
328 WALK_AF_CAPS(caps, ac)
329 if (ac->ext_next_hop)
330 {
331 put_af4(buf, ac->afi);
332 put_u16(buf+4, BGP_AFI_IPV6);
333 buf += 6;
334 }
335
336 data[-1] = buf - data;
337 }
338
339 if (caps->ext_messages)
340 {
341 *buf++ = 6; /* Capability 6: Support for extended messages */
342 *buf++ = 0; /* Capability data length */
343 }
344
345 if (caps->gr_aware)
346 {
347 *buf++ = 64; /* Capability 64: Support for graceful restart */
348 *buf++ = 0; /* Capability data length, will be fixed later */
349 data = buf;
350
351 put_u16(buf, caps->gr_time);
352 buf[0] |= caps->gr_flags;
353 buf += 2;
354
355 WALK_AF_CAPS(caps, ac)
356 if (ac->gr_able)
357 {
358 put_af3(buf, ac->afi);
359 buf[3] = ac->gr_af_flags;
360 buf += 4;
361 }
362
363 data[-1] = buf - data;
364 }
365
366 if (caps->as4_support)
367 {
368 *buf++ = 65; /* Capability 65: Support for 4-octet AS number */
369 *buf++ = 4; /* Capability data length */
370 put_u32(buf, p->public_as);
371 buf += 4;
372 }
373
374 if (caps->any_add_path)
375 {
376 *buf++ = 69; /* Capability 69: Support for ADD-PATH */
377 *buf++ = 0; /* Capability data length, will be fixed later */
378 data = buf;
379
380 WALK_AF_CAPS(caps, ac)
381 if (ac->add_path)
382 {
383 put_af3(buf, ac->afi);
384 buf[3] = ac->add_path;
385 buf += 4;
386 }
387
388 data[-1] = buf - data;
389 }
390
391 if (caps->enhanced_refresh)
392 {
393 *buf++ = 70; /* Capability 70: Support for enhanced route refresh */
394 *buf++ = 0; /* Capability data length */
395 }
396
397 if (caps->llgr_aware)
398 {
399 *buf++ = 71; /* Capability 71: Support for long-lived graceful restart */
400 *buf++ = 0; /* Capability data length, will be fixed later */
401 data = buf;
402
403 WALK_AF_CAPS(caps, ac)
404 if (ac->llgr_able)
405 {
406 put_af3(buf, ac->afi);
407 buf[3] = ac->llgr_flags;
408 put_u24(buf+4, ac->llgr_time);
409 buf += 7;
410 }
411
412 data[-1] = buf - data;
413 }
414
415 caps->length = buf - buf_head;
416
417 return buf;
418 }
419
420 static int
421 bgp_read_capabilities(struct bgp_conn *conn, byte *pos, int len)
422 {
423 struct bgp_proto *p = conn->bgp;
424 struct bgp_caps *caps;
425 struct bgp_af_caps *ac;
426 int i, cl;
427 u32 af;
428
429 if (!conn->remote_caps)
430 caps = mb_allocz(p->p.pool, sizeof(struct bgp_caps) + sizeof(struct bgp_af_caps));
431 else
432 {
433 caps = conn->remote_caps;
434 conn->remote_caps = NULL;
435 }
436
437 caps->length += len;
438
439 while (len > 0)
440 {
441 if (len < 2 || len < (2 + pos[1]))
442 goto err;
443
444 /* Capability length */
445 cl = pos[1];
446
447 /* Capability type */
448 switch (pos[0])
449 {
450 case 1: /* Multiprotocol capability, RFC 4760 */
451 if (cl != 4)
452 goto err;
453
454 af = get_af4(pos+2);
455 ac = bgp_get_af_caps(&caps, af);
456 ac->ready = 1;
457 break;
458
459 case 2: /* Route refresh capability, RFC 2918 */
460 if (cl != 0)
461 goto err;
462
463 caps->route_refresh = 1;
464 break;
465
466 case 5: /* Extended next hop encoding capability, RFC 5549 */
467 if (cl % 6)
468 goto err;
469
470 for (i = 0; i < cl; i += 6)
471 {
472 /* Specified only for IPv4 prefixes with IPv6 next hops */
473 if ((get_u16(pos+2+i+0) != BGP_AFI_IPV4) ||
474 (get_u16(pos+2+i+4) != BGP_AFI_IPV6))
475 continue;
476
477 af = get_af4(pos+2+i);
478 ac = bgp_get_af_caps(&caps, af);
479 ac->ext_next_hop = 1;
480 }
481 break;
482
483 case 6: /* Extended message length capability, RFC draft */
484 if (cl != 0)
485 goto err;
486
487 caps->ext_messages = 1;
488 break;
489
490 case 64: /* Graceful restart capability, RFC 4724 */
491 if (cl % 4 != 2)
492 goto err;
493
494 /* Only the last instance is valid */
495 WALK_AF_CAPS(caps, ac)
496 {
497 ac->gr_able = 0;
498 ac->gr_af_flags = 0;
499 }
500
501 caps->gr_aware = 1;
502 caps->gr_flags = pos[2] & 0xf0;
503 caps->gr_time = get_u16(pos + 2) & 0x0fff;
504
505 for (i = 2; i < cl; i += 4)
506 {
507 af = get_af3(pos+2+i);
508 ac = bgp_get_af_caps(&caps, af);
509 ac->gr_able = 1;
510 ac->gr_af_flags = pos[2+i+3];
511 }
512 break;
513
514 case 65: /* AS4 capability, RFC 6793 */
515 if (cl != 4)
516 goto err;
517
518 caps->as4_support = 1;
519 caps->as4_number = get_u32(pos + 2);
520 break;
521
522 case 69: /* ADD-PATH capability, RFC 7911 */
523 if (cl % 4)
524 goto err;
525
526 for (i = 0; i < cl; i += 4)
527 {
528 byte val = pos[2+i+3];
529 if (!val || (val > BGP_ADD_PATH_FULL))
530 {
531 log(L_WARN "%s: Got ADD-PATH capability with unknown value %u, ignoring",
532 p->p.name, val);
533 break;
534 }
535 }
536
537 for (i = 0; i < cl; i += 4)
538 {
539 af = get_af3(pos+2+i);
540 ac = bgp_get_af_caps(&caps, af);
541 ac->add_path = pos[2+i+3];
542 }
543 break;
544
545 case 70: /* Enhanced route refresh capability, RFC 7313 */
546 if (cl != 0)
547 goto err;
548
549 caps->enhanced_refresh = 1;
550 break;
551
552 case 71: /* Long lived graceful restart capability, RFC draft */
553 if (cl % 7)
554 goto err;
555
556 /* Presumably, only the last instance is valid */
557 WALK_AF_CAPS(caps, ac)
558 {
559 ac->llgr_able = 0;
560 ac->llgr_flags = 0;
561 ac->llgr_time = 0;
562 }
563
564 caps->llgr_aware = 1;
565
566 for (i = 0; i < cl; i += 7)
567 {
568 af = get_af3(pos+2+i);
569 ac = bgp_get_af_caps(&caps, af);
570 ac->llgr_able = 1;
571 ac->llgr_flags = pos[2+i+3];
572 ac->llgr_time = get_u24(pos + 2+i+4);
573 }
574 break;
575
576 /* We can safely ignore all other capabilities */
577 }
578
579 ADVANCE(pos, len, 2 + cl);
580 }
581
582 /* The LLGR capability must be advertised together with the GR capability,
583 otherwise it must be disregarded */
584 if (!caps->gr_aware && caps->llgr_aware)
585 {
586 caps->llgr_aware = 0;
587 WALK_AF_CAPS(caps, ac)
588 {
589 ac->llgr_able = 0;
590 ac->llgr_flags = 0;
591 ac->llgr_time = 0;
592 }
593 }
594
595 conn->remote_caps = caps;
596 return 0;
597
598 err:
599 mb_free(caps);
600 bgp_error(conn, 2, 0, NULL, 0);
601 return -1;
602 }
603
604 static int
605 bgp_check_capabilities(struct bgp_conn *conn)
606 {
607 struct bgp_proto *p = conn->bgp;
608 struct bgp_caps *local = conn->local_caps;
609 struct bgp_caps *remote = conn->remote_caps;
610 struct bgp_channel *c;
611 int count = 0;
612
613 /* This is partially overlapping with bgp_conn_enter_established_state(),
614 but we need to run this just after we receive OPEN message */
615
616 WALK_LIST(c, p->p.channels)
617 {
618 const struct bgp_af_caps *loc = bgp_find_af_caps(local, c->afi);
619 const struct bgp_af_caps *rem = bgp_find_af_caps(remote, c->afi);
620
621 /* Find out whether this channel will be active */
622 int active = loc && loc->ready &&
623 ((rem && rem->ready) || (!remote->length && (c->afi == BGP_AF_IPV4)));
624
625 /* Mandatory must be active */
626 if (c->cf->mandatory && !active)
627 return 0;
628
629 if (active)
630 count++;
631 }
632
633 /* We need at least one channel active */
634 if (!count)
635 return 0;
636
637 return 1;
638 }
639
640 static int
641 bgp_read_options(struct bgp_conn *conn, byte *pos, uint len, uint rest)
642 {
643 struct bgp_proto *p = conn->bgp;
644 int ext = 0;
645
646 /* Handle extended length (draft-ietf-idr-ext-opt-param-07) */
647 if ((len > 0) && (rest > 0) && (pos[0] == 255))
648 {
649 if (rest < 3)
650 goto err;
651
652 /* Update pos/len to describe optional data */
653 len = get_u16(pos+1);
654 ext = 1;
655 pos += 3;
656 rest -= 3;
657 }
658
659 /* Verify that optional data fits into OPEN packet */
660 if (len > rest)
661 goto err;
662
663 /* Length of option parameter header */
664 uint hlen = ext ? 3 : 2;
665
666 while (len > 0)
667 {
668 if (len < hlen)
669 goto err;
670
671 uint otype = get_u8(pos);
672 uint olen = ext ? get_u16(pos+1) : get_u8(pos+1);
673
674 if (len < (hlen + olen))
675 goto err;
676
677 if (otype == 2)
678 {
679 /* BGP capabilities, RFC 5492 */
680 if (p->cf->capabilities)
681 if (bgp_read_capabilities(conn, pos + hlen, olen) < 0)
682 return -1;
683 }
684 else
685 {
686 /* Unknown option */
687 bgp_error(conn, 2, 4, pos, hlen + olen);
688 return -1;
689 }
690
691 ADVANCE(pos, len, hlen + olen);
692 }
693
694 /* Prepare empty caps if no capability option was announced */
695 if (!conn->remote_caps)
696 conn->remote_caps = mb_allocz(p->p.pool, sizeof(struct bgp_caps));
697
698 return 0;
699
700 err:
701 bgp_error(conn, 2, 0, NULL, 0);
702 return -1;
703 }
704
705 static byte *
706 bgp_create_open(struct bgp_conn *conn, byte *buf)
707 {
708 struct bgp_proto *p = conn->bgp;
709
710 BGP_TRACE(D_PACKETS, "Sending OPEN(ver=%d,as=%d,hold=%d,id=%08x)",
711 BGP_VERSION, p->public_as, p->cf->hold_time, p->local_id);
712
713 buf[0] = BGP_VERSION;
714 put_u16(buf+1, (p->public_as < 0xFFFF) ? p->public_as : AS_TRANS);
715 put_u16(buf+3, p->cf->hold_time);
716 put_u32(buf+5, p->local_id);
717
718 if (p->cf->capabilities)
719 {
720 /* Prepare local_caps and write capabilities to buffer */
721 byte *pos = buf+12;
722 byte *end = bgp_write_capabilities(conn, pos);
723 uint len = end - pos;
724
725 if (len < 254)
726 {
727 buf[9] = len + 2; /* Optional parameters length */
728 buf[10] = 2; /* Option 2: Capability list */
729 buf[11] = len; /* Option data length */
730 }
731 else /* draft-ietf-idr-ext-opt-param-07 */
732 {
733 /* Move capabilities 4 B forward */
734 memmove(buf + 16, pos, len);
735 pos = buf + 16;
736 end = pos + len;
737
738 buf[9] = 255; /* Non-ext OP length, fake */
739 buf[10] = 255; /* Non-ext OP type, signals extended length */
740 put_u16(buf+11, len + 3); /* Extended optional parameters length */
741 buf[13] = 2; /* Option 2: Capability list */
742 put_u16(buf+14, len); /* Option extended data length */
743 }
744
745 return end;
746 }
747 else
748 {
749 buf[9] = 0; /* No optional parameters */
750 return buf + 10;
751 }
752
753 return buf;
754 }
755
756 static void
757 bgp_rx_open(struct bgp_conn *conn, byte *pkt, uint len)
758 {
759 struct bgp_proto *p = conn->bgp;
760 struct bgp_conn *other;
761 u32 asn, hold, id;
762
763 /* Check state */
764 if (conn->state != BS_OPENSENT)
765 { bgp_error(conn, 5, fsm_err_subcode[conn->state], NULL, 0); return; }
766
767 /* Check message length */
768 if (len < 29)
769 { bgp_error(conn, 1, 2, pkt+16, 2); return; }
770
771 if (pkt[19] != BGP_VERSION)
772 { u16 val = BGP_VERSION; bgp_error(conn, 2, 1, (byte *) &val, 2); return; }
773
774 asn = get_u16(pkt+20);
775 hold = get_u16(pkt+22);
776 id = get_u32(pkt+24);
777 BGP_TRACE(D_PACKETS, "Got OPEN(as=%d,hold=%d,id=%R)", asn, hold, id);
778
779 if (bgp_read_options(conn, pkt+29, pkt[28], len-29) < 0)
780 return;
781
782 if (hold > 0 && hold < 3)
783 { bgp_error(conn, 2, 6, pkt+22, 2); return; }
784
785 /* RFC 6286 2.2 - router ID is nonzero and AS-wide unique */
786 if (!id || (p->is_internal && id == p->local_id))
787 { bgp_error(conn, 2, 3, pkt+24, -4); return; }
788
789 /* RFC 5492 4 - check for required capabilities */
790 if (p->cf->capabilities && !bgp_check_capabilities(conn))
791 { bgp_error(conn, 2, 7, NULL, 0); return; }
792
793 struct bgp_caps *caps = conn->remote_caps;
794
795 if (caps->as4_support)
796 {
797 u32 as4 = caps->as4_number;
798
799 if ((as4 != asn) && (asn != AS_TRANS))
800 log(L_WARN "%s: Peer advertised inconsistent AS numbers", p->p.name);
801
802 /* When remote ASN is unspecified, it must be external one */
803 if (p->remote_as ? (as4 != p->remote_as) : (as4 == p->local_as))
804 { as4 = htonl(as4); bgp_error(conn, 2, 2, (byte *) &as4, 4); return; }
805
806 conn->received_as = as4;
807 }
808 else
809 {
810 if (p->remote_as ? (asn != p->remote_as) : (asn == p->local_as))
811 { bgp_error(conn, 2, 2, pkt+20, 2); return; }
812
813 conn->received_as = asn;
814 }
815
816 /* Check the other connection */
817 other = (conn == &p->outgoing_conn) ? &p->incoming_conn : &p->outgoing_conn;
818 switch (other->state)
819 {
820 case BS_CONNECT:
821 case BS_ACTIVE:
822 /* Stop outgoing connection attempts */
823 bgp_conn_enter_idle_state(other);
824 break;
825
826 case BS_IDLE:
827 case BS_OPENSENT:
828 case BS_CLOSE:
829 break;
830
831 case BS_OPENCONFIRM:
832 /*
833 * Description of collision detection rules in RFC 4271 is confusing and
834 * contradictory, but it is essentially:
835 *
836 * 1. Router with higher ID is dominant
837 * 2. If both have the same ID, router with higher ASN is dominant [RFC6286]
838 * 3. When both connections are in OpenConfirm state, one initiated by
839 * the dominant router is kept.
840 *
841 * The first line in the expression below evaluates whether the neighbor
842 * is dominant, the second line whether the new connection was initiated
843 * by the neighbor. If both are true (or both are false), we keep the new
844 * connection, otherwise we keep the old one.
845 */
846 if (((p->local_id < id) || ((p->local_id == id) && (p->public_as < p->remote_as)))
847 == (conn == &p->incoming_conn))
848 {
849 /* Should close the other connection */
850 BGP_TRACE(D_EVENTS, "Connection collision, giving up the other connection");
851 bgp_error(other, 6, 7, NULL, 0);
852 break;
853 }
854 /* Fall thru */
855 case BS_ESTABLISHED:
856 /* Should close this connection */
857 BGP_TRACE(D_EVENTS, "Connection collision, giving up this connection");
858 bgp_error(conn, 6, 7, NULL, 0);
859 return;
860
861 default:
862 bug("bgp_rx_open: Unknown state");
863 }
864
865 /* Update our local variables */
866 conn->hold_time = MIN(hold, p->cf->hold_time);
867 conn->keepalive_time = p->cf->keepalive_time ? : conn->hold_time / 3;
868 conn->as4_session = conn->local_caps->as4_support && caps->as4_support;
869 conn->ext_messages = conn->local_caps->ext_messages && caps->ext_messages;
870 p->remote_id = id;
871
872 DBG("BGP: Hold timer set to %d, keepalive to %d, AS to %d, ID to %x, AS4 session to %d\n",
873 conn->hold_time, conn->keepalive_time, p->remote_as, p->remote_id, conn->as4_session);
874
875 bgp_schedule_packet(conn, NULL, PKT_KEEPALIVE);
876 bgp_start_timer(conn->hold_timer, conn->hold_time);
877 bgp_conn_enter_openconfirm_state(conn);
878 }
879
880
881 /*
882 * Next hop handling
883 */
884
885 #define REPORT(msg, args...) \
886 ({ log(L_REMOTE "%s: " msg, s->proto->p.name, ## args); })
887
888 #define DISCARD(msg, args...) \
889 ({ REPORT(msg, ## args); return; })
890
891 #define WITHDRAW(msg, args...) \
892 ({ REPORT(msg, ## args); s->err_withdraw = 1; return; })
893
894 #define BAD_AFI "Unexpected AF <%u/%u> in UPDATE"
895 #define BAD_NEXT_HOP "Invalid NEXT_HOP attribute"
896 #define NO_NEXT_HOP "Missing NEXT_HOP attribute"
897 #define NO_LABEL_STACK "Missing MPLS stack"
898
899
900 static void
901 bgp_apply_next_hop(struct bgp_parse_state *s, rta *a, ip_addr gw, ip_addr ll)
902 {
903 struct bgp_proto *p = s->proto;
904 struct bgp_channel *c = s->channel;
905
906 if (c->cf->gw_mode == GW_DIRECT)
907 {
908 neighbor *nbr = NULL;
909
910 /* GW_DIRECT -> single_hop -> p->neigh != NULL */
911 if (ipa_nonzero(gw))
912 nbr = neigh_find(&p->p, gw, NULL, 0);
913 else if (ipa_nonzero(ll))
914 nbr = neigh_find(&p->p, ll, p->neigh->iface, 0);
915
916 if (!nbr || (nbr->scope == SCOPE_HOST))
917 WITHDRAW(BAD_NEXT_HOP);
918
919 a->dest = RTD_UNICAST;
920 a->nh.gw = nbr->addr;
921 a->nh.iface = nbr->iface;
922 a->igp_metric = c->cf->cost;
923 }
924 else /* GW_RECURSIVE */
925 {
926 if (ipa_zero(gw))
927 WITHDRAW(BAD_NEXT_HOP);
928
929 rtable *tab = ipa_is_ip4(gw) ? c->igp_table_ip4 : c->igp_table_ip6;
930 s->hostentry = rt_get_hostentry(tab, gw, ll, c->c.table);
931
932 if (!s->mpls)
933 rta_apply_hostentry(a, s->hostentry, NULL);
934
935 /* With MPLS, hostentry is applied later in bgp_apply_mpls_labels() */
936 }
937 }
938
939 static void
940 bgp_apply_mpls_labels(struct bgp_parse_state *s, rta *a, u32 *labels, uint lnum)
941 {
942 if (lnum > MPLS_MAX_LABEL_STACK)
943 {
944 REPORT("Too many MPLS labels ($u)", lnum);
945
946 a->dest = RTD_UNREACHABLE;
947 a->hostentry = NULL;
948 a->nh = (struct nexthop) { };
949 return;
950 }
951
952 /* Handle implicit NULL as empty MPLS stack */
953 if ((lnum == 1) && (labels[0] == BGP_MPLS_NULL))
954 lnum = 0;
955
956 if (s->channel->cf->gw_mode == GW_DIRECT)
957 {
958 a->nh.labels = lnum;
959 memcpy(a->nh.label, labels, 4*lnum);
960 }
961 else /* GW_RECURSIVE */
962 {
963 mpls_label_stack ms;
964
965 ms.len = lnum;
966 memcpy(ms.stack, labels, 4*lnum);
967 rta_apply_hostentry(a, s->hostentry, &ms);
968 }
969 }
970
971
972 static int
973 bgp_match_src(struct bgp_export_state *s, int mode)
974 {
975 switch (mode)
976 {
977 case NH_NO: return 0;
978 case NH_ALL: return 1;
979 case NH_IBGP: return s->src && s->src->is_internal;
980 case NH_EBGP: return s->src && !s->src->is_internal;
981 default: return 0;
982 }
983 }
984
985 static inline int
986 bgp_use_next_hop(struct bgp_export_state *s, eattr *a)
987 {
988 struct bgp_proto *p = s->proto;
989 struct bgp_channel *c = s->channel;
990 ip_addr *nh = (void *) a->u.ptr->data;
991
992 /* Handle next hop self option */
993 if (c->cf->next_hop_self && bgp_match_src(s, c->cf->next_hop_self))
994 return 0;
995
996 /* Handle next hop keep option */
997 if (c->cf->next_hop_keep && bgp_match_src(s, c->cf->next_hop_keep))
998 return 1;
999
1000 /* Keep it when explicitly set in export filter */
1001 if (a->type & EAF_FRESH)
1002 return 1;
1003
1004 /* Check for non-matching AF */
1005 if ((ipa_is_ip4(*nh) != bgp_channel_is_ipv4(c)) && !c->ext_next_hop)
1006 return 0;
1007
1008 /* Keep it when exported to internal peers */
1009 if (p->is_interior && ipa_nonzero(*nh))
1010 return 1;
1011
1012 /* Keep it when forwarded between single-hop BGPs on the same iface */
1013 struct iface *ifa = (s->src && s->src->neigh) ? s->src->neigh->iface : NULL;
1014 return p->neigh && (p->neigh->iface == ifa);
1015 }
1016
1017 static inline int
1018 bgp_use_gateway(struct bgp_export_state *s)
1019 {
1020 struct bgp_proto *p = s->proto;
1021 struct bgp_channel *c = s->channel;
1022 rta *ra = s->route->attrs;
1023
1024 /* Handle next hop self option - also applies to gateway */
1025 if (c->cf->next_hop_self && bgp_match_src(s, c->cf->next_hop_self))
1026 return 0;
1027
1028 /* We need one valid global gateway */
1029 if ((ra->dest != RTD_UNICAST) || ra->nh.next || ipa_zero(ra->nh.gw) || ipa_is_link_local(ra->nh.gw))
1030 return 0;
1031
1032 /* Check for non-matching AF */
1033 if ((ipa_is_ip4(ra->nh.gw) != bgp_channel_is_ipv4(c)) && !c->ext_next_hop)
1034 return 0;
1035
1036 /* Use it when exported to internal peers */
1037 if (p->is_interior)
1038 return 1;
1039
1040 /* Use it when forwarded to single-hop BGP peer on on the same iface */
1041 return p->neigh && (p->neigh->iface == ra->nh.iface);
1042 }
1043
1044 static void
1045 bgp_update_next_hop_ip(struct bgp_export_state *s, eattr *a, ea_list **to)
1046 {
1047 if (!a || !bgp_use_next_hop(s, a))
1048 {
1049 if (bgp_use_gateway(s))
1050 {
1051 rta *ra = s->route->attrs;
1052 ip_addr nh[1] = { ra->nh.gw };
1053 bgp_set_attr_data(to, s->pool, BA_NEXT_HOP, 0, nh, 16);
1054
1055 if (s->mpls)
1056 {
1057 u32 implicit_null = BGP_MPLS_NULL;
1058 u32 *labels = ra->nh.labels ? ra->nh.label : &implicit_null;
1059 uint lnum = ra->nh.labels ? ra->nh.labels : 1;
1060 bgp_set_attr_data(to, s->pool, BA_MPLS_LABEL_STACK, 0, labels, lnum * 4);
1061 }
1062 }
1063 else
1064 {
1065 ip_addr nh[2] = { s->channel->next_hop_addr, s->channel->link_addr };
1066 bgp_set_attr_data(to, s->pool, BA_NEXT_HOP, 0, nh, ipa_nonzero(nh[1]) ? 32 : 16);
1067 s->local_next_hop = 1;
1068
1069 /* TODO: Use local MPLS assigned label */
1070 if (s->mpls)
1071 {
1072 u32 implicit_null = BGP_MPLS_NULL;
1073 bgp_set_attr_data(to, s->pool, BA_MPLS_LABEL_STACK, 0, &implicit_null, 4);
1074 }
1075 }
1076 }
1077
1078 /* Check if next hop is valid */
1079 a = bgp_find_attr(*to, BA_NEXT_HOP);
1080 if (!a)
1081 WITHDRAW(NO_NEXT_HOP);
1082
1083 ip_addr *nh = (void *) a->u.ptr->data;
1084 ip_addr peer = s->proto->remote_ip;
1085 uint len = a->u.ptr->length;
1086
1087 /* Forbid zero next hop */
1088 if (ipa_zero(nh[0]) && ((len != 32) || ipa_zero(nh[1])))
1089 WITHDRAW(BAD_NEXT_HOP);
1090
1091 /* Forbid next hop equal to neighbor IP */
1092 if (ipa_equal(peer, nh[0]) || ((len == 32) && ipa_equal(peer, nh[1])))
1093 WITHDRAW(BAD_NEXT_HOP);
1094
1095 /* Forbid next hop with non-matching AF */
1096 if ((ipa_is_ip4(nh[0]) != bgp_channel_is_ipv4(s->channel)) &&
1097 !s->channel->ext_next_hop)
1098 WITHDRAW(BAD_NEXT_HOP);
1099
1100 /* Just check if MPLS stack */
1101 if (s->mpls && !bgp_find_attr(*to, BA_MPLS_LABEL_STACK))
1102 WITHDRAW(NO_LABEL_STACK);
1103 }
1104
1105 static uint
1106 bgp_encode_next_hop_ip(struct bgp_write_state *s, eattr *a, byte *buf, uint size UNUSED)
1107 {
1108 /* This function is used only for MP-BGP, see bgp_encode_next_hop() for IPv4 BGP */
1109 ip_addr *nh = (void *) a->u.ptr->data;
1110 uint len = a->u.ptr->length;
1111
1112 ASSERT((len == 16) || (len == 32));
1113
1114 /*
1115 * Both IPv4 and IPv6 next hops can be used (with ext_next_hop enabled). This
1116 * is specified in RFC 5549 for IPv4 and in RFC 4798 for IPv6. The difference
1117 * is that IPv4 address is directly encoded with IPv4 NLRI, but as IPv4-mapped
1118 * IPv6 address with IPv6 NLRI.
1119 */
1120
1121 if (bgp_channel_is_ipv4(s->channel) && ipa_is_ip4(nh[0]))
1122 {
1123 put_ip4(buf, ipa_to_ip4(nh[0]));
1124 return 4;
1125 }
1126
1127 put_ip6(buf, ipa_to_ip6(nh[0]));
1128
1129 if (len == 32)
1130 put_ip6(buf+16, ipa_to_ip6(nh[1]));
1131
1132 return len;
1133 }
1134
1135 static void
1136 bgp_decode_next_hop_ip(struct bgp_parse_state *s, byte *data, uint len, rta *a)
1137 {
1138 struct bgp_channel *c = s->channel;
1139 struct adata *ad = lp_alloc_adata(s->pool, 32);
1140 ip_addr *nh = (void *) ad->data;
1141
1142 if (len == 4)
1143 {
1144 nh[0] = ipa_from_ip4(get_ip4(data));
1145 nh[1] = IPA_NONE;
1146 }
1147 else if (len == 16)
1148 {
1149 nh[0] = ipa_from_ip6(get_ip6(data));
1150 nh[1] = IPA_NONE;
1151
1152 if (ipa_is_link_local(nh[0]))
1153 { nh[1] = nh[0]; nh[0] = IPA_NONE; }
1154 }
1155 else if (len == 32)
1156 {
1157 nh[0] = ipa_from_ip6(get_ip6(data));
1158 nh[1] = ipa_from_ip6(get_ip6(data+16));
1159
1160 if (ipa_is_ip4(nh[0]) || !ip6_is_link_local(nh[1]))
1161 nh[1] = IPA_NONE;
1162 }
1163 else
1164 bgp_parse_error(s, 9);
1165
1166 if (ipa_zero(nh[1]))
1167 ad->length = 16;
1168
1169 if ((bgp_channel_is_ipv4(c) != ipa_is_ip4(nh[0])) && !c->ext_next_hop)
1170 WITHDRAW(BAD_NEXT_HOP);
1171
1172 // XXXX validate next hop
1173
1174 bgp_set_attr_ptr(&(a->eattrs), s->pool, BA_NEXT_HOP, 0, ad);
1175 bgp_apply_next_hop(s, a, nh[0], nh[1]);
1176 }
1177
1178 static uint
1179 bgp_encode_next_hop_vpn(struct bgp_write_state *s, eattr *a, byte *buf, uint size UNUSED)
1180 {
1181 ip_addr *nh = (void *) a->u.ptr->data;
1182 uint len = a->u.ptr->length;
1183
1184 ASSERT((len == 16) || (len == 32));
1185
1186 /*
1187 * Both IPv4 and IPv6 next hops can be used (with ext_next_hop enabled). This
1188 * is specified in RFC 5549 for VPNv4 and in RFC 4659 for VPNv6. The difference
1189 * is that IPv4 address is directly encoded with VPNv4 NLRI, but as IPv4-mapped
1190 * IPv6 address with VPNv6 NLRI.
1191 */
1192
1193 if (bgp_channel_is_ipv4(s->channel) && ipa_is_ip4(nh[0]))
1194 {
1195 put_u64(buf, 0); /* VPN RD is 0 */
1196 put_ip4(buf+8, ipa_to_ip4(nh[0]));
1197 return 12;
1198 }
1199
1200 put_u64(buf, 0); /* VPN RD is 0 */
1201 put_ip6(buf+8, ipa_to_ip6(nh[0]));
1202
1203 if (len == 16)
1204 return 24;
1205
1206 put_u64(buf+24, 0); /* VPN RD is 0 */
1207 put_ip6(buf+32, ipa_to_ip6(nh[1]));
1208
1209 return 48;
1210 }
1211
1212 static void
1213 bgp_decode_next_hop_vpn(struct bgp_parse_state *s, byte *data, uint len, rta *a)
1214 {
1215 struct bgp_channel *c = s->channel;
1216 struct adata *ad = lp_alloc_adata(s->pool, 32);
1217 ip_addr *nh = (void *) ad->data;
1218
1219 if (len == 12)
1220 {
1221 nh[0] = ipa_from_ip4(get_ip4(data+8));
1222 nh[1] = IPA_NONE;
1223 }
1224 else if (len == 24)
1225 {
1226 nh[0] = ipa_from_ip6(get_ip6(data+8));
1227 nh[1] = IPA_NONE;
1228
1229 if (ipa_is_link_local(nh[0]))
1230 { nh[1] = nh[0]; nh[0] = IPA_NONE; }
1231 }
1232 else if (len == 48)
1233 {
1234 nh[0] = ipa_from_ip6(get_ip6(data+8));
1235 nh[1] = ipa_from_ip6(get_ip6(data+32));
1236
1237 if (ipa_is_ip4(nh[0]) || !ip6_is_link_local(nh[1]))
1238 nh[1] = IPA_NONE;
1239 }
1240 else
1241 bgp_parse_error(s, 9);
1242
1243 if (ipa_zero(nh[1]))
1244 ad->length = 16;
1245
1246 /* XXXX which error */
1247 if ((get_u64(data) != 0) || ((len == 48) && (get_u64(data+24) != 0)))
1248 bgp_parse_error(s, 9);
1249
1250 if ((bgp_channel_is_ipv4(c) != ipa_is_ip4(nh[0])) && !c->ext_next_hop)
1251 WITHDRAW(BAD_NEXT_HOP);
1252
1253 // XXXX validate next hop
1254
1255 bgp_set_attr_ptr(&(a->eattrs), s->pool, BA_NEXT_HOP, 0, ad);
1256 bgp_apply_next_hop(s, a, nh[0], nh[1]);
1257 }
1258
1259
1260
1261 static uint
1262 bgp_encode_next_hop_none(struct bgp_write_state *s UNUSED, eattr *a UNUSED, byte *buf UNUSED, uint size UNUSED)
1263 {
1264 return 0;
1265 }
1266
1267 static void
1268 bgp_decode_next_hop_none(struct bgp_parse_state *s UNUSED, byte *data UNUSED, uint len UNUSED, rta *a UNUSED)
1269 {
1270 /*
1271 * Although we expect no next hop and RFC 7606 7.11 states that attribute
1272 * MP_REACH_NLRI with unexpected next hop length is considered malformed,
1273 * FlowSpec RFC 5575 4 states that next hop shall be ignored on receipt.
1274 */
1275
1276 return;
1277 }
1278
1279 static void
1280 bgp_update_next_hop_none(struct bgp_export_state *s, eattr *a, ea_list **to)
1281 {
1282 /* NEXT_HOP shall not pass */
1283 if (a)
1284 bgp_unset_attr(to, s->pool, BA_NEXT_HOP);
1285 }
1286
1287
1288 /*
1289 * UPDATE
1290 */
1291
1292 static void
1293 bgp_rte_update(struct bgp_parse_state *s, net_addr *n, u32 path_id, rta *a0)
1294 {
1295 if (path_id != s->last_id)
1296 {
1297 s->last_src = rt_get_source(&s->proto->p, path_id);
1298 s->last_id = path_id;
1299
1300 rta_free(s->cached_rta);
1301 s->cached_rta = NULL;
1302 }
1303
1304 if (!a0)
1305 {
1306 /* Route withdraw */
1307 rte_update3(&s->channel->c, n, NULL, s->last_src);
1308 return;
1309 }
1310
1311 /* Prepare cached route attributes */
1312 if (s->cached_rta == NULL)
1313 {
1314 a0->src = s->last_src;
1315
1316 /* Workaround for rta_lookup() breaking eattrs */
1317 ea_list *ea = a0->eattrs;
1318 s->cached_rta = rta_lookup(a0);
1319 a0->eattrs = ea;
1320 }
1321
1322 rta *a = rta_clone(s->cached_rta);
1323 rte *e = rte_get_temp(a);
1324
1325 e->pflags = 0;
1326 e->u.bgp.suppressed = 0;
1327 e->u.bgp.stale = -1;
1328 rte_update3(&s->channel->c, n, e, s->last_src);
1329 }
1330
1331 static void
1332 bgp_encode_mpls_labels(struct bgp_write_state *s UNUSED, const adata *mpls, byte **pos, uint *size, byte *pxlen)
1333 {
1334 const u32 dummy = 0;
1335 const u32 *labels = mpls ? (const u32 *) mpls->data : &dummy;
1336 uint lnum = mpls ? (mpls->length / 4) : 1;
1337
1338 for (uint i = 0; i < lnum; i++)
1339 {
1340 put_u24(*pos, labels[i] << 4);
1341 ADVANCE(*pos, *size, 3);
1342 }
1343
1344 /* Add bottom-of-stack flag */
1345 (*pos)[-1] |= BGP_MPLS_BOS;
1346
1347 *pxlen += 24 * lnum;
1348 }
1349
1350 static void
1351 bgp_decode_mpls_labels(struct bgp_parse_state *s, byte **pos, uint *len, uint *pxlen, rta *a)
1352 {
1353 u32 labels[BGP_MPLS_MAX], label;
1354 uint lnum = 0;
1355
1356 do {
1357 if (*pxlen < 24)
1358 bgp_parse_error(s, 1);
1359
1360 label = get_u24(*pos);
1361 labels[lnum++] = label >> 4;
1362 ADVANCE(*pos, *len, 3);
1363 *pxlen -= 24;
1364
1365 /* RFC 8277 2.4 - withdraw does not have variable-size MPLS stack but
1366 fixed-size 24-bit Compatibility field, which MUST be ignored */
1367 if (!a && !s->err_withdraw)
1368 return;
1369 }
1370 while (!(label & BGP_MPLS_BOS));
1371
1372 if (!a)
1373 return;
1374
1375 /* Attach MPLS attribute unless we already have one */
1376 if (!s->mpls_labels)
1377 {
1378 s->mpls_labels = lp_alloc_adata(s->pool, 4*BGP_MPLS_MAX);
1379 bgp_set_attr_ptr(&(a->eattrs), s->pool, BA_MPLS_LABEL_STACK, 0, s->mpls_labels);
1380 }
1381
1382 /* Overwrite data in the attribute */
1383 s->mpls_labels->length = 4*lnum;
1384 memcpy(s->mpls_labels->data, labels, 4*lnum);
1385
1386 /* Update next hop entry in rta */
1387 bgp_apply_mpls_labels(s, a, labels, lnum);
1388
1389 /* Attributes were changed, invalidate cached entry */
1390 rta_free(s->cached_rta);
1391 s->cached_rta = NULL;
1392
1393 return;
1394 }
1395
1396 static uint
1397 bgp_encode_nlri_ip4(struct bgp_write_state *s, struct bgp_bucket *buck, byte *buf, uint size)
1398 {
1399 byte *pos = buf;
1400
1401 while (!EMPTY_LIST(buck->prefixes) && (size >= BGP_NLRI_MAX))
1402 {
1403 struct bgp_prefix *px = HEAD(buck->prefixes);
1404 struct net_addr_ip4 *net = (void *) px->net;
1405
1406 /* Encode path ID */
1407 if (s->add_path)
1408 {
1409 put_u32(pos, px->path_id);
1410 ADVANCE(pos, size, 4);
1411 }
1412
1413 /* Encode prefix length */
1414 *pos = net->pxlen;
1415 ADVANCE(pos, size, 1);
1416
1417 /* Encode MPLS labels */
1418 if (s->mpls)
1419 bgp_encode_mpls_labels(s, s->mpls_labels, &pos, &size, pos - 1);
1420
1421 /* Encode prefix body */
1422 ip4_addr a = ip4_hton(net->prefix);
1423 uint b = (net->pxlen + 7) / 8;
1424 memcpy(pos, &a, b);
1425 ADVANCE(pos, size, b);
1426
1427 bgp_free_prefix(s->channel, px);
1428 }
1429
1430 return pos - buf;
1431 }
1432
1433 static void
1434 bgp_decode_nlri_ip4(struct bgp_parse_state *s, byte *pos, uint len, rta *a)
1435 {
1436 while (len)
1437 {
1438 net_addr_ip4 net;
1439 u32 path_id = 0;
1440
1441 /* Decode path ID */
1442 if (s->add_path)
1443 {
1444 if (len < 5)
1445 bgp_parse_error(s, 1);
1446
1447 path_id = get_u32(pos);
1448 ADVANCE(pos, len, 4);
1449 }
1450
1451 /* Decode prefix length */
1452 uint l = *pos;
1453 ADVANCE(pos, len, 1);
1454
1455 if (len < ((l + 7) / 8))
1456 bgp_parse_error(s, 1);
1457
1458 /* Decode MPLS labels */
1459 if (s->mpls)
1460 bgp_decode_mpls_labels(s, &pos, &len, &l, a);
1461
1462 if (l > IP4_MAX_PREFIX_LENGTH)
1463 bgp_parse_error(s, 10);
1464
1465 /* Decode prefix body */
1466 ip4_addr addr = IP4_NONE;
1467 uint b = (l + 7) / 8;
1468 memcpy(&addr, pos, b);
1469 ADVANCE(pos, len, b);
1470
1471 net = NET_ADDR_IP4(ip4_ntoh(addr), l);
1472 net_normalize_ip4(&net);
1473
1474 // XXXX validate prefix
1475
1476 bgp_rte_update(s, (net_addr *) &net, path_id, a);
1477 }
1478 }
1479
1480
1481 static uint
1482 bgp_encode_nlri_ip6(struct bgp_write_state *s, struct bgp_bucket *buck, byte *buf, uint size)
1483 {
1484 byte *pos = buf;
1485
1486 while (!EMPTY_LIST(buck->prefixes) && (size >= BGP_NLRI_MAX))
1487 {
1488 struct bgp_prefix *px = HEAD(buck->prefixes);
1489 struct net_addr_ip6 *net = (void *) px->net;
1490
1491 /* Encode path ID */
1492 if (s->add_path)
1493 {
1494 put_u32(pos, px->path_id);
1495 ADVANCE(pos, size, 4);
1496 }
1497
1498 /* Encode prefix length */
1499 *pos = net->pxlen;
1500 ADVANCE(pos, size, 1);
1501
1502 /* Encode MPLS labels */
1503 if (s->mpls)
1504 bgp_encode_mpls_labels(s, s->mpls_labels, &pos, &size, pos - 1);
1505
1506 /* Encode prefix body */
1507 ip6_addr a = ip6_hton(net->prefix);
1508 uint b = (net->pxlen + 7) / 8;
1509 memcpy(pos, &a, b);
1510 ADVANCE(pos, size, b);
1511
1512 bgp_free_prefix(s->channel, px);
1513 }
1514
1515 return pos - buf;
1516 }
1517
1518 static void
1519 bgp_decode_nlri_ip6(struct bgp_parse_state *s, byte *pos, uint len, rta *a)
1520 {
1521 while (len)
1522 {
1523 net_addr_ip6 net;
1524 u32 path_id = 0;
1525
1526 /* Decode path ID */
1527 if (s->add_path)
1528 {
1529 if (len < 5)
1530 bgp_parse_error(s, 1);
1531
1532 path_id = get_u32(pos);
1533 ADVANCE(pos, len, 4);
1534 }
1535
1536 /* Decode prefix length */
1537 uint l = *pos;
1538 ADVANCE(pos, len, 1);
1539
1540 if (len < ((l + 7) / 8))
1541 bgp_parse_error(s, 1);
1542
1543 /* Decode MPLS labels */
1544 if (s->mpls)
1545 bgp_decode_mpls_labels(s, &pos, &len, &l, a);
1546
1547 if (l > IP6_MAX_PREFIX_LENGTH)
1548 bgp_parse_error(s, 10);
1549
1550 /* Decode prefix body */
1551 ip6_addr addr = IP6_NONE;
1552 uint b = (l + 7) / 8;
1553 memcpy(&addr, pos, b);
1554 ADVANCE(pos, len, b);
1555
1556 net = NET_ADDR_IP6(ip6_ntoh(addr), l);
1557 net_normalize_ip6(&net);
1558
1559 // XXXX validate prefix
1560
1561 bgp_rte_update(s, (net_addr *) &net, path_id, a);
1562 }
1563 }
1564
1565 static uint
1566 bgp_encode_nlri_vpn4(struct bgp_write_state *s, struct bgp_bucket *buck, byte *buf, uint size)
1567 {
1568 byte *pos = buf;
1569
1570 while (!EMPTY_LIST(buck->prefixes) && (size >= BGP_NLRI_MAX))
1571 {
1572 struct bgp_prefix *px = HEAD(buck->prefixes);
1573 struct net_addr_vpn4 *net = (void *) px->net;
1574
1575 /* Encode path ID */
1576 if (s->add_path)
1577 {
1578 put_u32(pos, px->path_id);
1579 ADVANCE(pos, size, 4);
1580 }
1581
1582 /* Encode prefix length */
1583 *pos = 64 + net->pxlen;
1584 ADVANCE(pos, size, 1);
1585
1586 /* Encode MPLS labels */
1587 if (s->mpls)
1588 bgp_encode_mpls_labels(s, s->mpls_labels, &pos, &size, pos - 1);
1589
1590 /* Encode route distinguisher */
1591 put_u64(pos, net->rd);
1592 ADVANCE(pos, size, 8);
1593
1594 /* Encode prefix body */
1595 ip4_addr a = ip4_hton(net->prefix);
1596 uint b = (net->pxlen + 7) / 8;
1597 memcpy(pos, &a, b);
1598 ADVANCE(pos, size, b);
1599
1600 bgp_free_prefix(s->channel, px);
1601 }
1602
1603 return pos - buf;
1604 }
1605
1606 static void
1607 bgp_decode_nlri_vpn4(struct bgp_parse_state *s, byte *pos, uint len, rta *a)
1608 {
1609 while (len)
1610 {
1611 net_addr_vpn4 net;
1612 u32 path_id = 0;
1613
1614 /* Decode path ID */
1615 if (s->add_path)
1616 {
1617 if (len < 5)
1618 bgp_parse_error(s, 1);
1619
1620 path_id = get_u32(pos);
1621 ADVANCE(pos, len, 4);
1622 }
1623
1624 /* Decode prefix length */
1625 uint l = *pos;
1626 ADVANCE(pos, len, 1);
1627
1628 if (len < ((l + 7) / 8))
1629 bgp_parse_error(s, 1);
1630
1631 /* Decode MPLS labels */
1632 if (s->mpls)
1633 bgp_decode_mpls_labels(s, &pos, &len, &l, a);
1634
1635 /* Decode route distinguisher */
1636 if (l < 64)
1637 bgp_parse_error(s, 1);
1638
1639 u64 rd = get_u64(pos);
1640 ADVANCE(pos, len, 8);
1641 l -= 64;
1642
1643 if (l > IP4_MAX_PREFIX_LENGTH)
1644 bgp_parse_error(s, 10);
1645
1646 /* Decode prefix body */
1647 ip4_addr addr = IP4_NONE;
1648 uint b = (l + 7) / 8;
1649 memcpy(&addr, pos, b);
1650 ADVANCE(pos, len, b);
1651
1652 net = NET_ADDR_VPN4(ip4_ntoh(addr), l, rd);
1653 net_normalize_vpn4(&net);
1654
1655 // XXXX validate prefix
1656
1657 bgp_rte_update(s, (net_addr *) &net, path_id, a);
1658 }
1659 }
1660
1661
1662 static uint
1663 bgp_encode_nlri_vpn6(struct bgp_write_state *s, struct bgp_bucket *buck, byte *buf, uint size)
1664 {
1665 byte *pos = buf;
1666
1667 while (!EMPTY_LIST(buck->prefixes) && (size >= BGP_NLRI_MAX))
1668 {
1669 struct bgp_prefix *px = HEAD(buck->prefixes);
1670 struct net_addr_vpn6 *net = (void *) px->net;
1671
1672 /* Encode path ID */
1673 if (s->add_path)
1674 {
1675 put_u32(pos, px->path_id);
1676 ADVANCE(pos, size, 4);
1677 }
1678
1679 /* Encode prefix length */
1680 *pos = 64 + net->pxlen;
1681 ADVANCE(pos, size, 1);
1682
1683 /* Encode MPLS labels */
1684 if (s->mpls)
1685 bgp_encode_mpls_labels(s, s->mpls_labels, &pos, &size, pos - 1);
1686
1687 /* Encode route distinguisher */
1688 put_u64(pos, net->rd);
1689 ADVANCE(pos, size, 8);
1690
1691 /* Encode prefix body */
1692 ip6_addr a = ip6_hton(net->prefix);
1693 uint b = (net->pxlen + 7) / 8;
1694 memcpy(pos, &a, b);
1695 ADVANCE(pos, size, b);
1696
1697 bgp_free_prefix(s->channel, px);
1698 }
1699
1700 return pos - buf;
1701 }
1702
1703 static void
1704 bgp_decode_nlri_vpn6(struct bgp_parse_state *s, byte *pos, uint len, rta *a)
1705 {
1706 while (len)
1707 {
1708 net_addr_vpn6 net;
1709 u32 path_id = 0;
1710
1711 /* Decode path ID */
1712 if (s->add_path)
1713 {
1714 if (len < 5)
1715 bgp_parse_error(s, 1);
1716
1717 path_id = get_u32(pos);
1718 ADVANCE(pos, len, 4);
1719 }
1720
1721 /* Decode prefix length */
1722 uint l = *pos;
1723 ADVANCE(pos, len, 1);
1724
1725 if (len < ((l + 7) / 8))
1726 bgp_parse_error(s, 1);
1727
1728 /* Decode MPLS labels */
1729 if (s->mpls)
1730 bgp_decode_mpls_labels(s, &pos, &len, &l, a);
1731
1732 /* Decode route distinguisher */
1733 if (l < 64)
1734 bgp_parse_error(s, 1);
1735
1736 u64 rd = get_u64(pos);
1737 ADVANCE(pos, len, 8);
1738 l -= 64;
1739
1740 if (l > IP6_MAX_PREFIX_LENGTH)
1741 bgp_parse_error(s, 10);
1742
1743 /* Decode prefix body */
1744 ip6_addr addr = IP6_NONE;
1745 uint b = (l + 7) / 8;
1746 memcpy(&addr, pos, b);
1747 ADVANCE(pos, len, b);
1748
1749 net = NET_ADDR_VPN6(ip6_ntoh(addr), l, rd);
1750 net_normalize_vpn6(&net);
1751
1752 // XXXX validate prefix
1753
1754 bgp_rte_update(s, (net_addr *) &net, path_id, a);
1755 }
1756 }
1757
1758
1759 static uint
1760 bgp_encode_nlri_flow4(struct bgp_write_state *s, struct bgp_bucket *buck, byte *buf, uint size)
1761 {
1762 byte *pos = buf;
1763
1764 while (!EMPTY_LIST(buck->prefixes) && (size >= 4))
1765 {
1766 struct bgp_prefix *px = HEAD(buck->prefixes);
1767 struct net_addr_flow4 *net = (void *) px->net;
1768 uint flen = net->length - sizeof(net_addr_flow4);
1769
1770 /* Encode path ID */
1771 if (s->add_path)
1772 {
1773 put_u32(pos, px->path_id);
1774 ADVANCE(pos, size, 4);
1775 }
1776
1777 if (flen > size)
1778 break;
1779
1780 /* Copy whole flow data including length */
1781 memcpy(pos, net->data, flen);
1782 ADVANCE(pos, size, flen);
1783
1784 bgp_free_prefix(s->channel, px);
1785 }
1786
1787 return pos - buf;
1788 }
1789
1790 static void
1791 bgp_decode_nlri_flow4(struct bgp_parse_state *s, byte *pos, uint len, rta *a)
1792 {
1793 while (len)
1794 {
1795 u32 path_id = 0;
1796
1797 /* Decode path ID */
1798 if (s->add_path)
1799 {
1800 if (len < 4)
1801 bgp_parse_error(s, 1);
1802
1803 path_id = get_u32(pos);
1804 ADVANCE(pos, len, 4);
1805 }
1806
1807 if (len < 2)
1808 bgp_parse_error(s, 1);
1809
1810 /* Decode flow length */
1811 uint hlen = flow_hdr_length(pos);
1812 uint dlen = flow_read_length(pos);
1813 uint flen = hlen + dlen;
1814 byte *data = pos + hlen;
1815
1816 if (len < flen)
1817 bgp_parse_error(s, 1);
1818
1819 /* Validate flow data */
1820 enum flow_validated_state r = flow4_validate(data, dlen);
1821 if (r != FLOW_ST_VALID)
1822 {
1823 log(L_REMOTE "%s: Invalid flow route: %s", s->proto->p.name, flow_validated_state_str(r));
1824 bgp_parse_error(s, 1);
1825 }
1826
1827 if (data[0] != FLOW_TYPE_DST_PREFIX)
1828 {
1829 log(L_REMOTE "%s: No dst prefix at first pos", s->proto->p.name);
1830 bgp_parse_error(s, 1);
1831 }
1832
1833 /* Decode dst prefix */
1834 ip4_addr px = IP4_NONE;
1835 uint pxlen = data[1];
1836
1837 // FIXME: Use some generic function
1838 memcpy(&px, data+2, BYTES(pxlen));
1839 px = ip4_and(ip4_ntoh(px), ip4_mkmask(pxlen));
1840
1841 /* Prepare the flow */
1842 net_addr *n = alloca(sizeof(struct net_addr_flow4) + flen);
1843 net_fill_flow4(n, px, pxlen, pos, flen);
1844 ADVANCE(pos, len, flen);
1845
1846 bgp_rte_update(s, n, path_id, a);
1847 }
1848 }
1849
1850
1851 static uint
1852 bgp_encode_nlri_flow6(struct bgp_write_state *s, struct bgp_bucket *buck, byte *buf, uint size)
1853 {
1854 byte *pos = buf;
1855
1856 while (!EMPTY_LIST(buck->prefixes) && (size >= 4))
1857 {
1858 struct bgp_prefix *px = HEAD(buck->prefixes);
1859 struct net_addr_flow6 *net = (void *) px->net;
1860 uint flen = net->length - sizeof(net_addr_flow6);
1861
1862 /* Encode path ID */
1863 if (s->add_path)
1864 {
1865 put_u32(pos, px->path_id);
1866 ADVANCE(pos, size, 4);
1867 }
1868
1869 if (flen > size)
1870 break;
1871
1872 /* Copy whole flow data including length */
1873 memcpy(pos, net->data, flen);
1874 ADVANCE(pos, size, flen);
1875
1876 bgp_free_prefix(s->channel, px);
1877 }
1878
1879 return pos - buf;
1880 }
1881
1882 static void
1883 bgp_decode_nlri_flow6(struct bgp_parse_state *s, byte *pos, uint len, rta *a)
1884 {
1885 while (len)
1886 {
1887 u32 path_id = 0;
1888
1889 /* Decode path ID */
1890 if (s->add_path)
1891 {
1892 if (len < 4)
1893 bgp_parse_error(s, 1);
1894
1895 path_id = get_u32(pos);
1896 ADVANCE(pos, len, 4);
1897 }
1898
1899 if (len < 2)
1900 bgp_parse_error(s, 1);
1901
1902 /* Decode flow length */
1903 uint hlen = flow_hdr_length(pos);
1904 uint dlen = flow_read_length(pos);
1905 uint flen = hlen + dlen;
1906 byte *data = pos + hlen;
1907
1908 if (len < flen)
1909 bgp_parse_error(s, 1);
1910
1911 /* Validate flow data */
1912 enum flow_validated_state r = flow6_validate(data, dlen);
1913 if (r != FLOW_ST_VALID)
1914 {
1915 log(L_REMOTE "%s: Invalid flow route: %s", s->proto->p.name, flow_validated_state_str(r));
1916 bgp_parse_error(s, 1);
1917 }
1918
1919 if (data[0] != FLOW_TYPE_DST_PREFIX)
1920 {
1921 log(L_REMOTE "%s: No dst prefix at first pos", s->proto->p.name);
1922 bgp_parse_error(s, 1);
1923 }
1924
1925 /* Decode dst prefix */
1926 ip6_addr px = IP6_NONE;
1927 uint pxlen = data[1];
1928
1929 // FIXME: Use some generic function
1930 memcpy(&px, data+2, BYTES(pxlen));
1931 px = ip6_and(ip6_ntoh(px), ip6_mkmask(pxlen));
1932
1933 /* Prepare the flow */
1934 net_addr *n = alloca(sizeof(struct net_addr_flow6) + flen);
1935 net_fill_flow6(n, px, pxlen, pos, flen);
1936 ADVANCE(pos, len, flen);
1937
1938 bgp_rte_update(s, n, path_id, a);
1939 }
1940 }
1941
1942
1943 static const struct bgp_af_desc bgp_af_table[] = {
1944 {
1945 .afi = BGP_AF_IPV4,
1946 .net = NET_IP4,
1947 .name = "ipv4",
1948 .encode_nlri = bgp_encode_nlri_ip4,
1949 .decode_nlri = bgp_decode_nlri_ip4,
1950 .encode_next_hop = bgp_encode_next_hop_ip,
1951 .decode_next_hop = bgp_decode_next_hop_ip,
1952 .update_next_hop = bgp_update_next_hop_ip,
1953 },
1954 {
1955 .afi = BGP_AF_IPV4_MC,
1956 .net = NET_IP4,
1957 .name = "ipv4-mc",
1958 .encode_nlri = bgp_encode_nlri_ip4,
1959 .decode_nlri = bgp_decode_nlri_ip4,
1960 .encode_next_hop = bgp_encode_next_hop_ip,
1961 .decode_next_hop = bgp_decode_next_hop_ip,
1962 .update_next_hop = bgp_update_next_hop_ip,
1963 },
1964 {
1965 .afi = BGP_AF_IPV4_MPLS,
1966 .net = NET_IP4,
1967 .mpls = 1,
1968 .name = "ipv4-mpls",
1969 .encode_nlri = bgp_encode_nlri_ip4,
1970 .decode_nlri = bgp_decode_nlri_ip4,
1971 .encode_next_hop = bgp_encode_next_hop_ip,
1972 .decode_next_hop = bgp_decode_next_hop_ip,
1973 .update_next_hop = bgp_update_next_hop_ip,
1974 },
1975 {
1976 .afi = BGP_AF_IPV6,
1977 .net = NET_IP6,
1978 .name = "ipv6",
1979 .encode_nlri = bgp_encode_nlri_ip6,
1980 .decode_nlri = bgp_decode_nlri_ip6,
1981 .encode_next_hop = bgp_encode_next_hop_ip,
1982 .decode_next_hop = bgp_decode_next_hop_ip,
1983 .update_next_hop = bgp_update_next_hop_ip,
1984 },
1985 {
1986 .afi = BGP_AF_IPV6_MC,
1987 .net = NET_IP6,
1988 .name = "ipv6-mc",
1989 .encode_nlri = bgp_encode_nlri_ip6,
1990 .decode_nlri = bgp_decode_nlri_ip6,
1991 .encode_next_hop = bgp_encode_next_hop_ip,
1992 .decode_next_hop = bgp_decode_next_hop_ip,
1993 .update_next_hop = bgp_update_next_hop_ip,
1994 },
1995 {
1996 .afi = BGP_AF_IPV6_MPLS,
1997 .net = NET_IP6,
1998 .mpls = 1,
1999 .name = "ipv6-mpls",
2000 .encode_nlri = bgp_encode_nlri_ip6,
2001 .decode_nlri = bgp_decode_nlri_ip6,
2002 .encode_next_hop = bgp_encode_next_hop_ip,
2003 .decode_next_hop = bgp_decode_next_hop_ip,
2004 .update_next_hop = bgp_update_next_hop_ip,
2005 },
2006 {
2007 .afi = BGP_AF_VPN4_MPLS,
2008 .net = NET_VPN4,
2009 .mpls = 1,
2010 .name = "vpn4-mpls",
2011 .encode_nlri = bgp_encode_nlri_vpn4,
2012 .decode_nlri = bgp_decode_nlri_vpn4,
2013 .encode_next_hop = bgp_encode_next_hop_vpn,
2014 .decode_next_hop = bgp_decode_next_hop_vpn,
2015 .update_next_hop = bgp_update_next_hop_ip,
2016 },
2017 {
2018 .afi = BGP_AF_VPN6_MPLS,
2019 .net = NET_VPN6,
2020 .mpls = 1,
2021 .name = "vpn6-mpls",
2022 .encode_nlri = bgp_encode_nlri_vpn6,
2023 .decode_nlri = bgp_decode_nlri_vpn6,
2024 .encode_next_hop = bgp_encode_next_hop_vpn,
2025 .decode_next_hop = bgp_decode_next_hop_vpn,
2026 .update_next_hop = bgp_update_next_hop_ip,
2027 },
2028 {
2029 .afi = BGP_AF_VPN4_MC,
2030 .net = NET_VPN4,
2031 .name = "vpn4-mc",
2032 .encode_nlri = bgp_encode_nlri_vpn4,
2033 .decode_nlri = bgp_decode_nlri_vpn4,
2034 .encode_next_hop = bgp_encode_next_hop_vpn,
2035 .decode_next_hop = bgp_decode_next_hop_vpn,
2036 .update_next_hop = bgp_update_next_hop_ip,
2037 },
2038 {
2039 .afi = BGP_AF_VPN6_MC,
2040 .net = NET_VPN6,
2041 .name = "vpn6-mc",
2042 .encode_nlri = bgp_encode_nlri_vpn6,
2043 .decode_nlri = bgp_decode_nlri_vpn6,
2044 .encode_next_hop = bgp_encode_next_hop_vpn,
2045 .decode_next_hop = bgp_decode_next_hop_vpn,
2046 .update_next_hop = bgp_update_next_hop_ip,
2047 },
2048 {
2049 .afi = BGP_AF_FLOW4,
2050 .net = NET_FLOW4,
2051 .no_igp = 1,
2052 .name = "flow4",
2053 .encode_nlri = bgp_encode_nlri_flow4,
2054 .decode_nlri = bgp_decode_nlri_flow4,
2055 .encode_next_hop = bgp_encode_next_hop_none,
2056 .decode_next_hop = bgp_decode_next_hop_none,
2057 .update_next_hop = bgp_update_next_hop_none,
2058 },
2059 {
2060 .afi = BGP_AF_FLOW6,
2061 .net = NET_FLOW6,
2062 .no_igp = 1,
2063 .name = "flow6",
2064 .encode_nlri = bgp_encode_nlri_flow6,
2065 .decode_nlri = bgp_decode_nlri_flow6,
2066 .encode_next_hop = bgp_encode_next_hop_none,
2067 .decode_next_hop = bgp_decode_next_hop_none,
2068 .update_next_hop = bgp_update_next_hop_none,
2069 },
2070 };
2071
2072 const struct bgp_af_desc *
2073 bgp_get_af_desc(u32 afi)
2074 {
2075 uint i;
2076 for (i = 0; i < ARRAY_SIZE(bgp_af_table); i++)
2077 if (bgp_af_table[i].afi == afi)
2078 return &bgp_af_table[i];
2079
2080 return NULL;
2081 }
2082
2083 static inline uint
2084 bgp_encode_nlri(struct bgp_write_state *s, struct bgp_bucket *buck, byte *buf, byte *end)
2085 {
2086 return s->channel->desc->encode_nlri(s, buck, buf, end - buf);
2087 }
2088
2089 static inline uint
2090 bgp_encode_next_hop(struct bgp_write_state *s, eattr *nh, byte *buf)
2091 {
2092 return s->channel->desc->encode_next_hop(s, nh, buf, 255);
2093 }
2094
2095 void
2096 bgp_update_next_hop(struct bgp_export_state *s, eattr *a, ea_list **to)
2097 {
2098 s->channel->desc->update_next_hop(s, a, to);
2099 }
2100
2101 #define MAX_ATTRS_LENGTH (end-buf+BGP_HEADER_LENGTH - 1024)
2102
2103 static byte *
2104 bgp_create_ip_reach(struct bgp_write_state *s, struct bgp_bucket *buck, byte *buf, byte *end)
2105 {
2106 /*
2107 * 2 B Withdrawn Routes Length (zero)
2108 * --- IPv4 Withdrawn Routes NLRI (unused)
2109 * 2 B Total Path Attribute Length
2110 * var Path Attributes
2111 * var IPv4 Network Layer Reachability Information
2112 */
2113
2114 int lr, la;
2115
2116 la = bgp_encode_attrs(s, buck->eattrs, buf+4, buf + MAX_ATTRS_LENGTH);
2117 if (la < 0)
2118 {
2119 /* Attribute list too long */
2120 bgp_withdraw_bucket(s->channel, buck);
2121 return NULL;
2122 }
2123
2124 put_u16(buf+0, 0);
2125 put_u16(buf+2, la);
2126
2127 lr = bgp_encode_nlri(s, buck, buf+4+la, end);
2128
2129 return buf+4+la+lr;
2130 }
2131
2132 static byte *
2133 bgp_create_mp_reach(struct bgp_write_state *s, struct bgp_bucket *buck, byte *buf, byte *end)
2134 {
2135 /*
2136 * 2 B IPv4 Withdrawn Routes Length (zero)
2137 * --- IPv4 Withdrawn Routes NLRI (unused)
2138 * 2 B Total Path Attribute Length
2139 * 1 B MP_REACH_NLRI hdr - Attribute Flags
2140 * 1 B MP_REACH_NLRI hdr - Attribute Type Code
2141 * 2 B MP_REACH_NLRI hdr - Length of Attribute Data
2142 * 2 B MP_REACH_NLRI data - Address Family Identifier
2143 * 1 B MP_REACH_NLRI data - Subsequent Address Family Identifier
2144 * 1 B MP_REACH_NLRI data - Length of Next Hop Network Address
2145 * var MP_REACH_NLRI data - Network Address of Next Hop
2146 * 1 B MP_REACH_NLRI data - Reserved (zero)
2147 * var MP_REACH_NLRI data - Network Layer Reachability Information
2148 * var Rest of Path Attributes
2149 * --- IPv4 Network Layer Reachability Information (unused)
2150 */
2151
2152 int lh, lr, la; /* Lengths of next hop, NLRI and attributes */
2153
2154 /* Begin of MP_REACH_NLRI atribute */
2155 buf[4] = BAF_OPTIONAL | BAF_EXT_LEN;
2156 buf[5] = BA_MP_REACH_NLRI;
2157 put_u16(buf+6, 0); /* Will be fixed later */
2158 put_af3(buf+8, s->channel->afi);
2159 byte *pos = buf+11;
2160
2161 /* Encode attributes to temporary buffer */
2162 byte *abuf = alloca(MAX_ATTRS_LENGTH);
2163 la = bgp_encode_attrs(s, buck->eattrs, abuf, abuf + MAX_ATTRS_LENGTH);
2164 if (la < 0)
2165 {
2166 /* Attribute list too long */
2167 bgp_withdraw_bucket(s->channel, buck);
2168 return NULL;
2169 }
2170
2171 /* Encode the next hop */
2172 lh = bgp_encode_next_hop(s, s->mp_next_hop, pos+1);
2173 *pos = lh;
2174 pos += 1+lh;
2175
2176 /* Reserved field */
2177 *pos++ = 0;
2178
2179 /* Encode the NLRI */
2180 lr = bgp_encode_nlri(s, buck, pos, end - la);
2181 pos += lr;
2182
2183 /* End of MP_REACH_NLRI atribute, update data length */
2184 put_u16(buf+6, pos-buf-8);
2185
2186 /* Copy remaining attributes */
2187 memcpy(pos, abuf, la);
2188 pos += la;
2189
2190 /* Initial UPDATE fields */
2191 put_u16(buf+0, 0);
2192 put_u16(buf+2, pos-buf-4);
2193
2194 return pos;
2195 }
2196
2197 #undef MAX_ATTRS_LENGTH
2198
2199 static byte *
2200 bgp_create_ip_unreach(struct bgp_write_state *s, struct bgp_bucket *buck, byte *buf, byte *end)
2201 {
2202 /*
2203 * 2 B Withdrawn Routes Length
2204 * var IPv4 Withdrawn Routes NLRI
2205 * 2 B Total Path Attribute Length (zero)
2206 * --- Path Attributes (unused)
2207 * --- IPv4 Network Layer Reachability Information (unused)
2208 */
2209
2210 uint len = bgp_encode_nlri(s, buck, buf+2, end);
2211
2212 put_u16(buf+0, len);
2213 put_u16(buf+2+len, 0);
2214
2215 return buf+4+len;
2216 }
2217
2218 static byte *
2219 bgp_create_mp_unreach(struct bgp_write_state *s, struct bgp_bucket *buck, byte *buf, byte *end)
2220 {
2221 /*
2222 * 2 B Withdrawn Routes Length (zero)
2223 * --- IPv4 Withdrawn Routes NLRI (unused)
2224 * 2 B Total Path Attribute Length
2225 * 1 B MP_UNREACH_NLRI hdr - Attribute Flags
2226 * 1 B MP_UNREACH_NLRI hdr - Attribute Type Code
2227 * 2 B MP_UNREACH_NLRI hdr - Length of Attribute Data
2228 * 2 B MP_UNREACH_NLRI data - Address Family Identifier
2229 * 1 B MP_UNREACH_NLRI data - Subsequent Address Family Identifier
2230 * var MP_UNREACH_NLRI data - Network Layer Reachability Information
2231 * --- IPv4 Network Layer Reachability Information (unused)
2232 */
2233
2234 uint len = bgp_encode_nlri(s, buck, buf+11, end);
2235
2236 put_u16(buf+0, 0);
2237 put_u16(buf+2, 7+len);
2238
2239 /* Begin of MP_UNREACH_NLRI atribute */
2240 buf[4] = BAF_OPTIONAL | BAF_EXT_LEN;
2241 buf[5] = BA_MP_UNREACH_NLRI;
2242 put_u16(buf+6, 3+len);
2243 put_af3(buf+8, s->channel->afi);
2244
2245 return buf+11+len;
2246 }
2247
2248 static byte *
2249 bgp_create_update(struct bgp_channel *c, byte *buf)
2250 {
2251 struct bgp_proto *p = (void *) c->c.proto;
2252 struct bgp_bucket *buck;
2253 byte *end = buf + (bgp_max_packet_length(p->conn) - BGP_HEADER_LENGTH);
2254 byte *res = NULL;
2255
2256 again: ;
2257
2258 /* Initialize write state */
2259 struct bgp_write_state s = {
2260 .proto = p,
2261 .channel = c,
2262 .pool = bgp_linpool,
2263 .mp_reach = (c->afi != BGP_AF_IPV4) || c->ext_next_hop,
2264 .as4_session = p->as4_session,
2265 .add_path = c->add_path_tx,
2266 .mpls = c->desc->mpls,
2267 };
2268
2269 /* Try unreachable bucket */
2270 if ((buck = c->withdraw_bucket) && !EMPTY_LIST(buck->prefixes))
2271 {
2272 res = (c->afi == BGP_AF_IPV4) && !c->ext_next_hop ?
2273 bgp_create_ip_unreach(&s, buck, buf, end):
2274 bgp_create_mp_unreach(&s, buck, buf, end);
2275
2276 goto done;
2277 }
2278
2279 /* Try reachable buckets */
2280 if (!EMPTY_LIST(c->bucket_queue))
2281 {
2282 buck = HEAD(c->bucket_queue);
2283
2284 /* Cleanup empty buckets */
2285 if (EMPTY_LIST(buck->prefixes))
2286 {
2287 bgp_free_bucket(c, buck);
2288 goto again;
2289 }
2290
2291 res = !s.mp_reach ?
2292 bgp_create_ip_reach(&s, buck, buf, end):
2293 bgp_create_mp_reach(&s, buck, buf, end);
2294
2295 if (EMPTY_LIST(buck->prefixes))
2296 bgp_free_bucket(c, buck);
2297 else
2298 bgp_defer_bucket(c, buck);
2299
2300 if (!res)
2301 goto again;
2302
2303 goto done;
2304 }
2305
2306 /* No more prefixes to send */
2307 return NULL;
2308
2309 done:
2310 BGP_TRACE_RL(&rl_snd_update, D_PACKETS, "Sending UPDATE");
2311 lp_flush(s.pool);
2312
2313 return res;
2314 }
2315
2316 static byte *
2317 bgp_create_ip_end_mark(struct bgp_channel *c UNUSED, byte *buf)
2318 {
2319 /* Empty update packet */
2320 put_u32(buf, 0);
2321
2322 return buf+4;
2323 }
2324
2325 static byte *
2326 bgp_create_mp_end_mark(struct bgp_channel *c, byte *buf)
2327 {
2328 put_u16(buf+0, 0);
2329 put_u16(buf+2, 6); /* length 4--9 */
2330
2331 /* Empty MP_UNREACH_NLRI atribute */
2332 buf[4] = BAF_OPTIONAL;
2333 buf[5] = BA_MP_UNREACH_NLRI;
2334 buf[6] = 3; /* Length 7--9 */
2335 put_af3(buf+7, c->afi);
2336
2337 return buf+10;
2338 }
2339
2340 static byte *
2341 bgp_create_end_mark(struct bgp_channel *c, byte *buf)
2342 {
2343 struct bgp_proto *p = (void *) c->c.proto;
2344
2345 BGP_TRACE(D_PACKETS, "Sending END-OF-RIB");
2346
2347 return (c->afi == BGP_AF_IPV4) ?
2348 bgp_create_ip_end_mark(c, buf):
2349 bgp_create_mp_end_mark(c, buf);
2350 }
2351
2352 static inline void
2353 bgp_rx_end_mark(struct bgp_parse_state *s, u32 afi)
2354 {
2355 struct bgp_proto *p = s->proto;
2356 struct bgp_channel *c = bgp_get_channel(p, afi);
2357
2358 BGP_TRACE(D_PACKETS, "Got END-OF-RIB");
2359
2360 if (!c)
2361 DISCARD(BAD_AFI, BGP_AFI(afi), BGP_SAFI(afi));
2362
2363 if (c->load_state == BFS_LOADING)
2364 c->load_state = BFS_NONE;
2365
2366 if (p->p.gr_recovery)
2367 channel_graceful_restart_unlock(&c->c);
2368
2369 if (c->gr_active)
2370 bgp_graceful_restart_done(c);
2371 }
2372
2373 static inline void
2374 bgp_decode_nlri(struct bgp_parse_state *s, u32 afi, byte *nlri, uint len, ea_list *ea, byte *nh, uint nh_len)
2375 {
2376 struct bgp_channel *c = bgp_get_channel(s->proto, afi);
2377 rta *a = NULL;
2378
2379 if (!c)
2380 DISCARD(BAD_AFI, BGP_AFI(afi), BGP_SAFI(afi));
2381
2382 s->channel = c;
2383 s->add_path = c->add_path_rx;
2384 s->mpls = c->desc->mpls;
2385
2386 s->last_id = 0;
2387 s->last_src = s->proto->p.main_source;
2388
2389 /*
2390 * IPv4 BGP and MP-BGP may be used together in one update, therefore we do not
2391 * add BA_NEXT_HOP in bgp_decode_attrs(), but we add it here independently for
2392 * IPv4 BGP and MP-BGP. We undo the attribute (and possibly others attached by
2393 * decode_next_hop hooks) by restoring a->eattrs afterwards.
2394 */
2395
2396 if (ea)
2397 {
2398 a = allocz(RTA_MAX_SIZE);
2399
2400 a->source = RTS_BGP;
2401 a->scope = SCOPE_UNIVERSE;
2402 a->from = s->proto->remote_ip;
2403 a->eattrs = ea;
2404
2405 c->desc->decode_next_hop(s, nh, nh_len, a);
2406 bgp_finish_attrs(s, a);
2407
2408 /* Handle withdraw during next hop decoding */
2409 if (s->err_withdraw)
2410 a = NULL;
2411 }
2412
2413 c->desc->decode_nlri(s, nlri, len, a);
2414
2415 rta_free(s->cached_rta);
2416 s->cached_rta = NULL;
2417 }
2418
2419 static void
2420 bgp_rx_update(struct bgp_conn *conn, byte *pkt, uint len)
2421 {
2422 struct bgp_proto *p = conn->bgp;
2423 ea_list *ea = NULL;
2424
2425 BGP_TRACE_RL(&rl_rcv_update, D_PACKETS, "Got UPDATE");
2426
2427 /* Workaround for some BGP implementations that skip initial KEEPALIVE */
2428 if (conn->state == BS_OPENCONFIRM)
2429 bgp_conn_enter_established_state(conn);
2430
2431 if (conn->state != BS_ESTABLISHED)
2432 { bgp_error(conn, 5, fsm_err_subcode[conn->state], NULL, 0); return; }
2433
2434 bgp_start_timer(conn->hold_timer, conn->hold_time);
2435
2436 /* Initialize parse state */
2437 struct bgp_parse_state s = {
2438 .proto = p,
2439 .pool = bgp_linpool,
2440 .as4_session = p->as4_session,
2441 };
2442
2443 /* Parse error handler */
2444 if (setjmp(s.err_jmpbuf))
2445 {
2446 bgp_error(conn, 3, s.err_subcode, NULL, 0);
2447 goto done;
2448 }
2449
2450 /* Check minimal length */
2451 if (len < 23)
2452 { bgp_error(conn, 1, 2, pkt+16, 2); return; }
2453
2454 /* Skip fixed header */
2455 uint pos = 19;
2456
2457 /*
2458 * UPDATE message format
2459 *
2460 * 2 B IPv4 Withdrawn Routes Length
2461 * var IPv4 Withdrawn Routes NLRI
2462 * 2 B Total Path Attribute Length
2463 * var Path Attributes
2464 * var IPv4 Reachable Routes NLRI
2465 */
2466
2467 s.ip_unreach_len = get_u16(pkt + pos);
2468 s.ip_unreach_nlri = pkt + pos + 2;
2469 pos += 2 + s.ip_unreach_len;
2470
2471 if (pos + 2 > len)
2472 bgp_parse_error(&s, 1);
2473
2474 s.attr_len = get_u16(pkt + pos);
2475 s.attrs = pkt + pos + 2;
2476 pos += 2 + s.attr_len;
2477
2478 if (pos > len)
2479 bgp_parse_error(&s, 1);
2480
2481 s.ip_reach_len = len - pos;
2482 s.ip_reach_nlri = pkt + pos;
2483
2484
2485 if (s.attr_len)
2486 ea = bgp_decode_attrs(&s, s.attrs, s.attr_len);
2487 else
2488 ea = NULL;
2489
2490 /* Check for End-of-RIB marker */
2491 if (!s.attr_len && !s.ip_unreach_len && !s.ip_reach_len)
2492 { bgp_rx_end_mark(&s, BGP_AF_IPV4); goto done; }
2493
2494 /* Check for MP End-of-RIB marker */
2495 if ((s.attr_len < 8) && !s.ip_unreach_len && !s.ip_reach_len &&
2496 !s.mp_reach_len && !s.mp_unreach_len && s.mp_unreach_af)
2497 { bgp_rx_end_mark(&s, s.mp_unreach_af); goto done; }
2498
2499 if (s.ip_unreach_len)
2500 bgp_decode_nlri(&s, BGP_AF_IPV4, s.ip_unreach_nlri, s.ip_unreach_len, NULL, NULL, 0);
2501
2502 if (s.mp_unreach_len)
2503 bgp_decode_nlri(&s, s.mp_unreach_af, s.mp_unreach_nlri, s.mp_unreach_len, NULL, NULL, 0);
2504
2505 if (s.ip_reach_len)
2506 bgp_decode_nlri(&s, BGP_AF_IPV4, s.ip_reach_nlri, s.ip_reach_len,
2507 ea, s.ip_next_hop_data, s.ip_next_hop_len);
2508
2509 if (s.mp_reach_len)
2510 bgp_decode_nlri(&s, s.mp_reach_af, s.mp_reach_nlri, s.mp_reach_len,
2511 ea, s.mp_next_hop_data, s.mp_next_hop_len);
2512
2513 done:
2514 rta_free(s.cached_rta);
2515 lp_flush(s.pool);
2516 return;
2517 }
2518
2519 static uint
2520 bgp_find_update_afi(byte *pos, uint len)
2521 {
2522 /*
2523 * This is stripped-down version of bgp_rx_update(), bgp_decode_attrs() and
2524 * bgp_decode_mp_[un]reach_nlri() used by MRT code in order to find out which
2525 * AFI/SAFI is associated with incoming UPDATE. Returns 0 for framing errors.
2526 */
2527 if (len < 23)
2528 return 0;
2529
2530 /* Assume there is no withrawn NLRI, read lengths and move to attribute list */
2531 uint wlen = get_u16(pos + 19);
2532 uint alen = get_u16(pos + 21);
2533 ADVANCE(pos, len, 23);
2534
2535 /* Either non-zero withdrawn NLRI, non-zero reachable NLRI, or IPv4 End-of-RIB */
2536 if ((wlen != 0) || (alen < len) || !alen)
2537 return BGP_AF_IPV4;
2538
2539 if (alen > len)
2540 return 0;
2541
2542 /* Process attribute list (alen == len) */
2543 while (len)
2544 {
2545 if (len < 2)
2546 return 0;
2547
2548 uint flags = pos[0];
2549 uint code = pos[1];
2550 ADVANCE(pos, len, 2);
2551
2552 uint ll = !(flags & BAF_EXT_LEN) ? 1 : 2;
2553 if (len < ll)
2554 return 0;
2555
2556 /* Read attribute length and move to attribute body */
2557 alen = (ll == 1) ? get_u8(pos) : get_u16(pos);
2558 ADVANCE(pos, len, ll);
2559
2560 if (len < alen)
2561 return 0;
2562
2563 /* Found MP NLRI */
2564 if ((code == BA_MP_REACH_NLRI) || (code == BA_MP_UNREACH_NLRI))
2565 {
2566 if (alen < 3)
2567 return 0;
2568
2569 return BGP_AF(get_u16(pos), pos[2]);
2570 }
2571
2572 /* Move to the next attribute */
2573 ADVANCE(pos, len, alen);
2574 }
2575
2576 /* No basic or MP NLRI, but there are some attributes -> error */
2577 return 0;
2578 }
2579
2580
2581 /*
2582 * ROUTE-REFRESH
2583 */
2584
2585 static inline byte *
2586 bgp_create_route_refresh(struct bgp_channel *c, byte *buf)
2587 {
2588 struct bgp_proto *p = (void *) c->c.proto;
2589
2590 BGP_TRACE(D_PACKETS, "Sending ROUTE-REFRESH");
2591
2592 /* Original route refresh request, RFC 2918 */
2593 put_af4(buf, c->afi);
2594 buf[2] = BGP_RR_REQUEST;
2595
2596 return buf+4;
2597 }
2598
2599 static inline byte *
2600 bgp_create_begin_refresh(struct bgp_channel *c, byte *buf)
2601 {
2602 struct bgp_proto *p = (void *) c->c.proto;
2603
2604 BGP_TRACE(D_PACKETS, "Sending BEGIN-OF-RR");
2605
2606 /* Demarcation of beginning of route refresh (BoRR), RFC 7313 */
2607 put_af4(buf, c->afi);
2608 buf[2] = BGP_RR_BEGIN;
2609
2610 return buf+4;
2611 }
2612
2613 static inline byte *
2614 bgp_create_end_refresh(struct bgp_channel *c, byte *buf)
2615 {
2616 struct bgp_proto *p = (void *) c->c.proto;
2617
2618 BGP_TRACE(D_PACKETS, "Sending END-OF-RR");
2619
2620 /* Demarcation of ending of route refresh (EoRR), RFC 7313 */
2621 put_af4(buf, c->afi);
2622 buf[2] = BGP_RR_END;
2623
2624 return buf+4;
2625 }
2626
2627 static void
2628 bgp_rx_route_refresh(struct bgp_conn *conn, byte *pkt, uint len)
2629 {
2630 struct bgp_proto *p = conn->bgp;
2631
2632 if (conn->state != BS_ESTABLISHED)
2633 { bgp_error(conn, 5, fsm_err_subcode[conn->state], NULL, 0); return; }
2634
2635 if (!conn->local_caps->route_refresh)
2636 { bgp_error(conn, 1, 3, pkt+18, 1); return; }
2637
2638 if (len < (BGP_HEADER_LENGTH + 4))
2639 { bgp_error(conn, 1, 2, pkt+16, 2); return; }
2640
2641 if (len > (BGP_HEADER_LENGTH + 4))
2642 { bgp_error(conn, 7, 1, pkt, MIN(len, 2048)); return; }
2643
2644 struct bgp_channel *c = bgp_get_channel(p, get_af4(pkt+19));
2645 if (!c)
2646 {
2647 log(L_WARN "%s: Got ROUTE-REFRESH subtype %u for AF %u.%u, ignoring",
2648 p->p.name, pkt[21], get_u16(pkt+19), pkt[22]);
2649 return;
2650 }
2651
2652 /* RFC 7313 redefined reserved field as RR message subtype */
2653 uint subtype = p->enhanced_refresh ? pkt[21] : BGP_RR_REQUEST;
2654
2655 switch (subtype)
2656 {
2657 case BGP_RR_REQUEST:
2658 BGP_TRACE(D_PACKETS, "Got ROUTE-REFRESH");
2659 channel_request_feeding(&c->c);
2660 break;
2661
2662 case BGP_RR_BEGIN:
2663 BGP_TRACE(D_PACKETS, "Got BEGIN-OF-RR");
2664 bgp_refresh_begin(c);
2665 break;
2666
2667 case BGP_RR_END:
2668 BGP_TRACE(D_PACKETS, "Got END-OF-RR");
2669 bgp_refresh_end(c);
2670 break;
2671
2672 default:
2673 log(L_WARN "%s: Got ROUTE-REFRESH message with unknown subtype %u, ignoring",
2674 p->p.name, subtype);
2675 break;
2676 }
2677 }
2678
2679 static inline struct bgp_channel *
2680 bgp_get_channel_to_send(struct bgp_proto *p, struct bgp_conn *conn)
2681 {
2682 uint i = conn->last_channel;
2683
2684 /* Try the last channel, but at most several times */
2685 if ((conn->channels_to_send & (1 << i)) &&
2686 (conn->last_channel_count < 16))
2687 goto found;
2688
2689 /* Find channel with non-zero channels_to_send */
2690 do
2691 {
2692 i++;
2693 if (i >= p->channel_count)
2694 i = 0;
2695 }
2696 while (! (conn->channels_to_send & (1 << i)));
2697
2698 /* Use that channel */
2699 conn->last_channel = i;
2700 conn->last_channel_count = 0;
2701
2702 found:
2703 conn->last_channel_count++;
2704 return p->channel_map[i];
2705 }
2706
2707 static inline int
2708 bgp_send(struct bgp_conn *conn, uint type, uint len)
2709 {
2710 sock *sk = conn->sk;
2711 byte *buf = sk->tbuf;
2712
2713 memset(buf, 0xff, 16); /* Marker */
2714 put_u16(buf+16, len);
2715 buf[18] = type;
2716
2717 return sk_send(sk, len);
2718 }
2719
2720 /**
2721 * bgp_fire_tx - transmit packets
2722 * @conn: connection
2723 *
2724 * Whenever the transmit buffers of the underlying TCP connection
2725 * are free and we have any packets queued for sending, the socket functions
2726 * call bgp_fire_tx() which takes care of selecting the highest priority packet
2727 * queued (Notification > Keepalive > Open > Update), assembling its header
2728 * and body and sending it to the connection.
2729 */
2730 static int
2731 bgp_fire_tx(struct bgp_conn *conn)
2732 {
2733 struct bgp_proto *p = conn->bgp;
2734 struct bgp_channel *c;
2735 byte *buf, *pkt, *end;
2736 uint s;
2737
2738 if (!conn->sk)
2739 return 0;
2740
2741 buf = conn->sk->tbuf;
2742 pkt = buf + BGP_HEADER_LENGTH;
2743 s = conn->packets_to_send;
2744
2745 if (s & (1 << PKT_SCHEDULE_CLOSE))
2746 {
2747 /* We can finally close connection and enter idle state */
2748 bgp_conn_enter_idle_state(conn);
2749 return 0;
2750 }
2751 if (s & (1 << PKT_NOTIFICATION))
2752 {
2753 conn->packets_to_send = 1 << PKT_SCHEDULE_CLOSE;
2754 end = bgp_create_notification(conn, pkt);
2755 return bgp_send(conn, PKT_NOTIFICATION, end - buf);
2756 }
2757 else if (s & (1 << PKT_OPEN))
2758 {
2759 conn->packets_to_send &= ~(1 << PKT_OPEN);
2760 end = bgp_create_open(conn, pkt);
2761 return bgp_send(conn, PKT_OPEN, end - buf);
2762 }
2763 else if (s & (1 << PKT_KEEPALIVE))
2764 {
2765 conn->packets_to_send &= ~(1 << PKT_KEEPALIVE);
2766 BGP_TRACE(D_PACKETS, "Sending KEEPALIVE");
2767 bgp_start_timer(conn->keepalive_timer, conn->keepalive_time);
2768 return bgp_send(conn, PKT_KEEPALIVE, BGP_HEADER_LENGTH);
2769 }
2770 else while (conn->channels_to_send)
2771 {
2772 c = bgp_get_channel_to_send(p, conn);
2773 s = c->packets_to_send;
2774
2775 if (s & (1 << PKT_ROUTE_REFRESH))
2776 {
2777 c->packets_to_send &= ~(1 << PKT_ROUTE_REFRESH);
2778 end = bgp_create_route_refresh(c, pkt);
2779 return bgp_send(conn, PKT_ROUTE_REFRESH, end - buf);
2780 }
2781 else if (s & (1 << PKT_BEGIN_REFRESH))
2782 {
2783 /* BoRR is a subtype of RR, but uses separate bit in packets_to_send */
2784 c->packets_to_send &= ~(1 << PKT_BEGIN_REFRESH);
2785 end = bgp_create_begin_refresh(c, pkt);
2786 return bgp_send(conn, PKT_ROUTE_REFRESH, end - buf);
2787 }
2788 else if (s & (1 << PKT_UPDATE))
2789 {
2790 end = bgp_create_update(c, pkt);
2791 if (end)
2792 return bgp_send(conn, PKT_UPDATE, end - buf);
2793
2794 /* No update to send, perhaps we need to send End-of-RIB or EoRR */
2795 c->packets_to_send = 0;
2796 conn->channels_to_send &= ~(1 << c->index);
2797
2798 if (c->feed_state == BFS_LOADED)
2799 {
2800 c->feed_state = BFS_NONE;
2801 end = bgp_create_end_mark(c, pkt);
2802 return bgp_send(conn, PKT_UPDATE, end - buf);
2803 }
2804
2805 else if (c->feed_state == BFS_REFRESHED)
2806 {
2807 c->feed_state = BFS_NONE;
2808 end = bgp_create_end_refresh(c, pkt);
2809 return bgp_send(conn, PKT_ROUTE_REFRESH, end - buf);
2810 }
2811 }
2812 else if (s)
2813 bug("Channel packets_to_send: %x", s);
2814
2815 c->packets_to_send = 0;
2816 conn->channels_to_send &= ~(1 << c->index);
2817 }
2818
2819 return 0;
2820 }
2821
2822 /**
2823 * bgp_schedule_packet - schedule a packet for transmission
2824 * @conn: connection
2825 * @c: channel
2826 * @type: packet type
2827 *
2828 * Schedule a packet of type @type to be sent as soon as possible.
2829 */
2830 void
2831 bgp_schedule_packet(struct bgp_conn *conn, struct bgp_channel *c, int type)
2832 {
2833 ASSERT(conn->sk);
2834
2835 DBG("BGP: Scheduling packet type %d\n", type);
2836
2837 if (c)
2838 {
2839 if (! conn->channels_to_send)
2840 {
2841 conn->last_channel = c->index;
2842 conn->last_channel_count = 0;
2843 }
2844
2845 c->packets_to_send |= 1 << type;
2846 conn->channels_to_send |= 1 << c->index;
2847 }
2848 else
2849 conn->packets_to_send |= 1 << type;
2850
2851 if ((conn->sk->tpos == conn->sk->tbuf) && !ev_active(conn->tx_ev))
2852 ev_schedule(conn->tx_ev);
2853 }
2854 void
2855 bgp_kick_tx(void *vconn)
2856 {
2857 struct bgp_conn *conn = vconn;
2858
2859 DBG("BGP: kicking TX\n");
2860 uint max = 1024;
2861 while (--max && (bgp_fire_tx(conn) > 0))
2862 ;
2863
2864 if (!max && !ev_active(conn->tx_ev))
2865 ev_schedule(conn->tx_ev);
2866 }
2867
2868 void
2869 bgp_tx(sock *sk)
2870 {
2871 struct bgp_conn *conn = sk->data;
2872
2873 DBG("BGP: TX hook\n");
2874 uint max = 1024;
2875 while (--max && (bgp_fire_tx(conn) > 0))
2876 ;
2877
2878 if (!max && !ev_active(conn->tx_ev))
2879 ev_schedule(conn->tx_ev);
2880 }
2881
2882
2883 static struct {
2884 byte major, minor;
2885 byte *msg;
2886 } bgp_msg_table[] = {
2887 { 1, 0, "Invalid message header" },
2888 { 1, 1, "Connection not synchronized" },
2889 { 1, 2, "Bad message length" },
2890 { 1, 3, "Bad message type" },
2891 { 2, 0, "Invalid OPEN message" },
2892 { 2, 1, "Unsupported version number" },
2893 { 2, 2, "Bad peer AS" },
2894 { 2, 3, "Bad BGP identifier" },
2895 { 2, 4, "Unsupported optional parameter" },
2896 { 2, 5, "Authentication failure" },
2897 { 2, 6, "Unacceptable hold time" },
2898 { 2, 7, "Required capability missing" }, /* [RFC5492] */
2899 { 2, 8, "No supported AFI/SAFI" }, /* This error msg is nonstandard */
2900 { 3, 0, "Invalid UPDATE message" },
2901 { 3, 1, "Malformed attribute list" },
2902 { 3, 2, "Unrecognized well-known attribute" },
2903 { 3, 3, "Missing mandatory attribute" },
2904 { 3, 4, "Invalid attribute flags" },
2905 { 3, 5, "Invalid attribute length" },
2906 { 3, 6, "Invalid ORIGIN attribute" },
2907 { 3, 7, "AS routing loop" }, /* Deprecated */
2908 { 3, 8, "Invalid NEXT_HOP attribute" },
2909 { 3, 9, "Optional attribute error" },
2910 { 3, 10, "Invalid network field" },
2911 { 3, 11, "Malformed AS_PATH" },
2912 { 4, 0, "Hold timer expired" },
2913 { 5, 0, "Finite state machine error" }, /* Subcodes are according to [RFC6608] */
2914 { 5, 1, "Unexpected message in OpenSent state" },
2915 { 5, 2, "Unexpected message in OpenConfirm state" },
2916 { 5, 3, "Unexpected message in Established state" },
2917 { 6, 0, "Cease" }, /* Subcodes are according to [RFC4486] */
2918 { 6, 1, "Maximum number of prefixes reached" },
2919 { 6, 2, "Administrative shutdown" },
2920 { 6, 3, "Peer de-configured" },
2921 { 6, 4, "Administrative reset" },
2922 { 6, 5, "Connection rejected" },
2923 { 6, 6, "Other configuration change" },
2924 { 6, 7, "Connection collision resolution" },
2925 { 6, 8, "Out of Resources" },
2926 { 7, 0, "Invalid ROUTE-REFRESH message" }, /* [RFC7313] */
2927 { 7, 1, "Invalid ROUTE-REFRESH message length" } /* [RFC7313] */
2928 };
2929
2930 /**
2931 * bgp_error_dsc - return BGP error description
2932 * @code: BGP error code
2933 * @subcode: BGP error subcode
2934 *
2935 * bgp_error_dsc() returns error description for BGP errors
2936 * which might be static string or given temporary buffer.
2937 */
2938 const char *
2939 bgp_error_dsc(uint code, uint subcode)
2940 {
2941 static char buff[32];
2942 uint i;
2943
2944 for (i=0; i < ARRAY_SIZE(bgp_msg_table); i++)
2945 if (bgp_msg_table[i].major == code && bgp_msg_table[i].minor == subcode)
2946 return bgp_msg_table[i].msg;
2947
2948 bsprintf(buff, "Unknown error %u.%u", code, subcode);
2949 return buff;
2950 }
2951
2952 /* RFC 8203 - shutdown communication message */
2953 static int
2954 bgp_handle_message(struct bgp_proto *p, byte *data, uint len, byte **bp)
2955 {
2956 byte *msg = data + 1;
2957 uint msg_len = data[0];
2958 uint i;
2959
2960 /* Handle zero length message */
2961 if (msg_len == 0)
2962 return 1;
2963
2964 /* Handle proper message */
2965 if (msg_len + 1 > len)
2966 return 0;
2967
2968 /* Some elementary cleanup */
2969 for (i = 0; i < msg_len; i++)
2970 if (msg[i] < ' ')
2971 msg[i] = ' ';
2972
2973 proto_set_message(&p->p, msg, msg_len);
2974 *bp += bsprintf(*bp, ": \"%s\"", p->p.message);
2975 return 1;
2976 }
2977
2978 void
2979 bgp_log_error(struct bgp_proto *p, u8 class, char *msg, uint code, uint subcode, byte *data, uint len)
2980 {
2981 byte argbuf[256+16], *t = argbuf;
2982 uint i;
2983
2984 /* Don't report Cease messages generated by myself */
2985 if (code == 6 && class == BE_BGP_TX)
2986 return;
2987
2988 /* Reset shutdown message */
2989 if ((code == 6) && ((subcode == 2) || (subcode == 4)))
2990 proto_set_message(&p->p, NULL, 0);
2991
2992 if (len)
2993 {
2994 /* Bad peer AS - we would like to print the AS */
2995 if ((code == 2) && (subcode == 2) && ((len == 2) || (len == 4)))
2996 {
2997 t += bsprintf(t, ": %u", (len == 2) ? get_u16(data) : get_u32(data));
2998 goto done;
2999 }
3000
3001 /* RFC 8203 - shutdown communication */
3002 if (((code == 6) && ((subcode == 2) || (subcode == 4))))
3003 if (bgp_handle_message(p, data, len, &t))
3004 goto done;
3005
3006 *t++ = ':';
3007 *t++ = ' ';
3008 if (len > 16)
3009 len = 16;
3010 for (i=0; i<len; i++)
3011 t += bsprintf(t, "%02x", data[i]);
3012 }
3013
3014 done:
3015 *t = 0;
3016 const byte *dsc = bgp_error_dsc(code, subcode);
3017 log(L_REMOTE "%s: %s: %s%s", p->p.name, msg, dsc, argbuf);
3018 }
3019
3020 static void
3021 bgp_rx_notification(struct bgp_conn *conn, byte *pkt, uint len)
3022 {
3023 struct bgp_proto *p = conn->bgp;
3024
3025 if (len < 21)
3026 { bgp_error(conn, 1, 2, pkt+16, 2); return; }
3027
3028 uint code = pkt[19];
3029 uint subcode = pkt[20];
3030 int err = (code != 6);
3031
3032 bgp_log_error(p, BE_BGP_RX, "Received", code, subcode, pkt+21, len-21);
3033 bgp_store_error(p, conn, BE_BGP_RX, (code << 16) | subcode);
3034
3035 bgp_conn_enter_close_state(conn);
3036 bgp_schedule_packet(conn, NULL, PKT_SCHEDULE_CLOSE);
3037
3038 if (err)
3039 {
3040 bgp_update_startup_delay(p);
3041 bgp_stop(p, 0, NULL, 0);
3042 }
3043 else
3044 {
3045 uint subcode_bit = 1 << ((subcode <= 8) ? subcode : 0);
3046 if (p->cf->disable_after_cease & subcode_bit)
3047 {
3048 log(L_INFO "%s: Disabled after Cease notification", p->p.name);
3049 p->startup_delay = 0;
3050 p->p.disabled = 1;
3051 }
3052 }
3053 }
3054
3055 static void
3056 bgp_rx_keepalive(struct bgp_conn *conn)
3057 {
3058 struct bgp_proto *p = conn->bgp;
3059
3060 BGP_TRACE(D_PACKETS, "Got KEEPALIVE");
3061 bgp_start_timer(conn->hold_timer, conn->hold_time);
3062
3063 if (conn->state == BS_OPENCONFIRM)
3064 { bgp_conn_enter_established_state(conn); return; }
3065
3066 if (conn->state != BS_ESTABLISHED)
3067 bgp_error(conn, 5, fsm_err_subcode[conn->state], NULL, 0);
3068 }
3069
3070
3071 /**
3072 * bgp_rx_packet - handle a received packet
3073 * @conn: BGP connection
3074 * @pkt: start of the packet
3075 * @len: packet size
3076 *
3077 * bgp_rx_packet() takes a newly received packet and calls the corresponding
3078 * packet handler according to the packet type.
3079 */
3080 static void
3081 bgp_rx_packet(struct bgp_conn *conn, byte *pkt, uint len)
3082 {
3083 byte type = pkt[18];
3084
3085 DBG("BGP: Got packet %02x (%d bytes)\n", type, len);
3086
3087 if (conn->bgp->p.mrtdump & MD_MESSAGES)
3088 bgp_dump_message(conn, pkt, len);
3089
3090 switch (type)
3091 {
3092 case PKT_OPEN: return bgp_rx_open(conn, pkt, len);
3093 case PKT_UPDATE: return bgp_rx_update(conn, pkt, len);
3094 case PKT_NOTIFICATION: return bgp_rx_notification(conn, pkt, len);
3095 case PKT_KEEPALIVE: return bgp_rx_keepalive(conn);
3096 case PKT_ROUTE_REFRESH: return bgp_rx_route_refresh(conn, pkt, len);
3097 default: bgp_error(conn, 1, 3, pkt+18, 1);
3098 }
3099 }
3100
3101 /**
3102 * bgp_rx - handle received data
3103 * @sk: socket
3104 * @size: amount of data received
3105 *
3106 * bgp_rx() is called by the socket layer whenever new data arrive from
3107 * the underlying TCP connection. It assembles the data fragments to packets,
3108 * checks their headers and framing and passes complete packets to
3109 * bgp_rx_packet().
3110 */
3111 int
3112 bgp_rx(sock *sk, uint size)
3113 {
3114 struct bgp_conn *conn = sk->data;
3115 byte *pkt_start = sk->rbuf;
3116 byte *end = pkt_start + size;
3117 uint i, len;
3118
3119 DBG("BGP: RX hook: Got %d bytes\n", size);
3120 while (end >= pkt_start + BGP_HEADER_LENGTH)
3121 {
3122 if ((conn->state == BS_CLOSE) || (conn->sk != sk))
3123 return 0;
3124 for(i=0; i<16; i++)
3125 if (pkt_start[i] != 0xff)
3126 {
3127 bgp_error(conn, 1, 1, NULL, 0);
3128 break;
3129 }
3130 len = get_u16(pkt_start+16);
3131 if ((len < BGP_HEADER_LENGTH) || (len > bgp_max_packet_length(conn)))
3132 {
3133 bgp_error(conn, 1, 2, pkt_start+16, 2);
3134 break;
3135 }
3136 if (end < pkt_start + len)
3137 break;
3138 bgp_rx_packet(conn, pkt_start, len);
3139 pkt_start += len;
3140 }
3141 if (pkt_start != sk->rbuf)
3142 {
3143 memmove(sk->rbuf, pkt_start, end - pkt_start);
3144 sk->rpos = sk->rbuf + (end - pkt_start);
3145 }
3146 return 0;
3147 }
Mirror of https://gitlab.labs.nic.cz/labs/bird.git