]>
Commit | Line | Data |
---|---|---|
cd1964ff DS |
1 | /* BGP carrying label information |
2 | * Copyright (C) 2013 Cumulus Networks, Inc. | |
3 | * | |
4 | * This file is part of GNU Zebra. | |
5 | * | |
6 | * GNU Zebra is free software; you can redistribute it and/or modify it | |
7 | * under the terms of the GNU General Public License as published by the | |
8 | * Free Software Foundation; either version 2, or (at your option) any | |
9 | * later version. | |
10 | * | |
11 | * GNU Zebra is distributed in the hope that it will be useful, but | |
12 | * WITHOUT ANY WARRANTY; without even the implied warranty of | |
13 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU | |
14 | * General Public License for more details. | |
15 | * | |
896014f4 DL |
16 | * You should have received a copy of the GNU General Public License along |
17 | * with this program; see the file COPYING; if not, write to the Free Software | |
18 | * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA | |
cd1964ff DS |
19 | */ |
20 | ||
21 | #include <zebra.h> | |
22 | ||
23 | #include "command.h" | |
24 | #include "thread.h" | |
25 | #include "prefix.h" | |
26 | #include "zclient.h" | |
27 | #include "stream.h" | |
28 | #include "network.h" | |
29 | #include "log.h" | |
30 | #include "memory.h" | |
31 | #include "nexthop.h" | |
32 | #include "mpls.h" | |
33 | ||
34 | #include "bgpd/bgpd.h" | |
35 | #include "bgpd/bgp_table.h" | |
36 | #include "bgpd/bgp_route.h" | |
37 | #include "bgpd/bgp_attr.h" | |
38 | #include "bgpd/bgp_label.h" | |
39 | #include "bgpd/bgp_packet.h" | |
40 | #include "bgpd/bgp_debug.h" | |
14454c9f | 41 | #include "bgpd/bgp_errors.h" |
cd1964ff DS |
42 | |
43 | extern struct zclient *zclient; | |
44 | ||
d62a17ae | 45 | int bgp_parse_fec_update(void) |
cd1964ff | 46 | { |
d62a17ae | 47 | struct stream *s; |
48 | struct bgp_node *rn; | |
49 | struct bgp *bgp; | |
50 | struct bgp_table *table; | |
51 | struct prefix p; | |
d7c0a89a | 52 | uint32_t label; |
d62a17ae | 53 | afi_t afi; |
54 | safi_t safi; | |
55 | ||
56 | s = zclient->ibuf; | |
57 | ||
58 | memset(&p, 0, sizeof(struct prefix)); | |
59 | p.family = stream_getw(s); | |
60 | p.prefixlen = stream_getc(s); | |
a85297a7 | 61 | stream_get(p.u.val, s, PSIZE(p.prefixlen)); |
d62a17ae | 62 | label = stream_getl(s); |
63 | ||
64 | /* hack for the bgp instance & SAFI = have to send/receive it */ | |
65 | afi = family2afi(p.family); | |
66 | safi = SAFI_UNICAST; | |
67 | bgp = bgp_get_default(); | |
68 | if (!bgp) { | |
69 | zlog_debug("no default bgp instance"); | |
70 | return -1; | |
71 | } | |
72 | ||
73 | table = bgp->rib[afi][safi]; | |
74 | if (!table) { | |
75 | zlog_debug("no %u unicast table", p.family); | |
76 | return -1; | |
77 | } | |
78 | rn = bgp_node_lookup(table, &p); | |
79 | if (!rn) { | |
80 | zlog_debug("no node for the prefix"); | |
81 | return -1; | |
82 | } | |
83 | ||
84 | /* treat it as implicit withdraw - the label is invalid */ | |
85 | if (label == MPLS_INVALID_LABEL) | |
86 | bgp_unset_valid_label(&rn->local_label); | |
87 | else { | |
88 | label_ntop(label, 1, &rn->local_label); | |
89 | bgp_set_valid_label(&rn->local_label); | |
90 | } | |
91 | SET_FLAG(rn->flags, BGP_NODE_LABEL_CHANGED); | |
92 | bgp_unlock_node(rn); | |
93 | bgp_process(bgp, rn, afi, safi); | |
94 | return 1; | |
cd1964ff DS |
95 | } |
96 | ||
40381db7 | 97 | mpls_label_t bgp_adv_label(struct bgp_node *rn, struct bgp_path_info *pi, |
d62a17ae | 98 | struct peer *to, afi_t afi, safi_t safi) |
cd1964ff | 99 | { |
d62a17ae | 100 | struct peer *from; |
101 | mpls_label_t remote_label; | |
102 | int reflect; | |
cd1964ff | 103 | |
40381db7 | 104 | if (!rn || !pi || !to) |
d62a17ae | 105 | return MPLS_INVALID_LABEL; |
cd1964ff | 106 | |
40381db7 DS |
107 | remote_label = pi->extra ? pi->extra->label[0] : MPLS_INVALID_LABEL; |
108 | from = pi->peer; | |
d62a17ae | 109 | reflect = |
110 | ((from->sort == BGP_PEER_IBGP) && (to->sort == BGP_PEER_IBGP)); | |
cd1964ff | 111 | |
d62a17ae | 112 | if (reflect |
113 | && !CHECK_FLAG(to->af_flags[afi][safi], | |
114 | PEER_FLAG_FORCE_NEXTHOP_SELF)) | |
115 | return remote_label; | |
cd1964ff | 116 | |
d62a17ae | 117 | if (CHECK_FLAG(to->af_flags[afi][safi], PEER_FLAG_NEXTHOP_UNCHANGED)) |
118 | return remote_label; | |
cd1964ff | 119 | |
d62a17ae | 120 | return rn->local_label; |
cd1964ff DS |
121 | } |
122 | ||
57592a53 AD |
123 | /** |
124 | * This is passed as the callback function to bgp_labelpool.c:bgp_lp_get() | |
125 | * by bgp_reg_dereg_for_label() when a label needs to be obtained from | |
126 | * label pool. | |
127 | * Note that it will reject the allocated label if a label index is found, | |
128 | * because the label index supposes predictable labels | |
129 | */ | |
130 | int bgp_reg_for_label_callback(mpls_label_t new_label, void *labelid, | |
131 | bool allocated) | |
132 | { | |
ea63ff6b EDP |
133 | struct bgp_path_info *pi; |
134 | struct bgp_node *rn; | |
57592a53 AD |
135 | char addr[PREFIX_STRLEN]; |
136 | ||
ea63ff6b EDP |
137 | pi = labelid; |
138 | /* Is this path still valid? */ | |
139 | if (!bgp_path_info_unlock(pi)) { | |
140 | if (BGP_DEBUG(labelpool, LABELPOOL)) | |
141 | zlog_debug( | |
142 | "%s: bgp_path_info is no longer valid, ignoring", | |
143 | __func__); | |
144 | return -1; | |
145 | } | |
146 | ||
147 | rn = pi->net; | |
57592a53 AD |
148 | prefix2str(&rn->p, addr, PREFIX_STRLEN); |
149 | ||
150 | if (BGP_DEBUG(labelpool, LABELPOOL)) | |
151 | zlog_debug("%s: FEC %s label=%u, allocated=%d", __func__, addr, | |
152 | new_label, allocated); | |
153 | ||
154 | if (!allocated) { | |
155 | /* | |
156 | * previously-allocated label is now invalid | |
157 | */ | |
158 | if (pi->attr->label_index == MPLS_INVALID_LABEL_INDEX | |
159 | && pi->attr->label != MPLS_LABEL_NONE | |
160 | && CHECK_FLAG(rn->flags, BGP_NODE_REGISTERED_FOR_LABEL)) { | |
161 | bgp_unregister_for_label(rn); | |
162 | label_ntop(MPLS_LABEL_IMPLICIT_NULL, 1, | |
163 | &rn->local_label); | |
164 | bgp_set_valid_label(&rn->local_label); | |
165 | } | |
166 | return 0; | |
167 | } | |
168 | ||
169 | /* | |
170 | * label index is assigned, this should be handled by SR-related code, | |
171 | * so retry FEC registration and then reject label allocation for | |
172 | * it to be released to label pool | |
173 | */ | |
174 | if (pi->attr->label_index != MPLS_INVALID_LABEL_INDEX) { | |
175 | flog_err( | |
176 | EC_BGP_LABEL, | |
177 | "%s: FEC %s Rejecting allocated label %u as Label Index is %u", | |
178 | __func__, addr, new_label, pi->attr->label_index); | |
179 | ||
180 | bgp_register_for_label(pi->net, pi); | |
181 | ||
182 | return -1; | |
183 | } | |
184 | ||
185 | if (pi->attr->label != MPLS_INVALID_LABEL) { | |
186 | if (new_label == pi->attr->label) { | |
187 | /* already have same label, accept but do nothing */ | |
188 | return 0; | |
189 | } | |
190 | /* Shouldn't happen: different label allocation */ | |
191 | flog_err(EC_BGP_LABEL, | |
192 | "%s: %s had label %u but got new assignment %u", | |
193 | __func__, addr, pi->attr->label, new_label); | |
194 | /* continue means use new one */ | |
195 | } | |
196 | ||
197 | label_ntop(new_label, 1, &rn->local_label); | |
198 | bgp_set_valid_label(&rn->local_label); | |
199 | ||
200 | /* | |
201 | * Get back to registering the FEC | |
202 | */ | |
203 | bgp_register_for_label(pi->net, pi); | |
204 | ||
205 | return 0; | |
206 | } | |
207 | ||
40381db7 | 208 | void bgp_reg_dereg_for_label(struct bgp_node *rn, struct bgp_path_info *pi, |
57592a53 | 209 | bool reg) |
cd1964ff | 210 | { |
57592a53 | 211 | bool with_label_index = false; |
d62a17ae | 212 | struct stream *s; |
213 | struct prefix *p; | |
57592a53 | 214 | mpls_label_t *local_label; |
d62a17ae | 215 | int command; |
d7c0a89a | 216 | uint16_t flags = 0; |
d62a17ae | 217 | size_t flags_pos = 0; |
57592a53 AD |
218 | char addr[PREFIX_STRLEN]; |
219 | ||
220 | p = &(rn->p); | |
221 | local_label = &(rn->local_label); | |
222 | /* this prevents the loop when we're called by | |
223 | * bgp_reg_for_label_callback() | |
224 | */ | |
225 | bool have_label_to_reg = bgp_is_valid_label(local_label) | |
226 | && label_pton(local_label) != MPLS_LABEL_IMPLICIT_NULL; | |
227 | ||
228 | if (reg) { | |
229 | assert(pi); | |
230 | /* | |
231 | * Determine if we will let zebra should derive label from | |
232 | * label index instead of bgpd requesting from label pool | |
233 | */ | |
234 | if (CHECK_FLAG(pi->attr->flag, | |
235 | ATTR_FLAG_BIT(BGP_ATTR_PREFIX_SID)) | |
236 | && pi->attr->label_index != BGP_INVALID_LABEL_INDEX) { | |
237 | with_label_index = true; | |
238 | } else { | |
239 | /* | |
240 | * If no label index was provided -- assume any label | |
241 | * from label pool will do. This means that label index | |
242 | * always takes precedence over auto-assigned labels. | |
243 | */ | |
244 | if (!have_label_to_reg) { | |
245 | if (BGP_DEBUG(labelpool, LABELPOOL)) { | |
246 | prefix2str(p, addr, PREFIX_STRLEN); | |
247 | zlog_debug("%s: Requesting label from LP for %s", | |
248 | __func__, addr); | |
249 | } | |
250 | /* bgp_reg_for_label_callback() will call back | |
251 | * __func__ when it gets a label from the pool. | |
252 | * This means we'll never register FECs without | |
253 | * valid labels. | |
254 | */ | |
255 | bgp_lp_get(LP_TYPE_BGP_LU, pi, | |
256 | bgp_reg_for_label_callback); | |
257 | return; | |
258 | } | |
259 | } | |
260 | } | |
d62a17ae | 261 | |
262 | /* Check socket. */ | |
263 | if (!zclient || zclient->sock < 0) | |
264 | return; | |
265 | ||
57592a53 AD |
266 | /* If the route node has a local_label assigned or the |
267 | * path node has an MPLS SR label index allowing zebra to | |
268 | * derive the label, proceed with registration. */ | |
d62a17ae | 269 | s = zclient->obuf; |
270 | stream_reset(s); | |
271 | command = (reg) ? ZEBRA_FEC_REGISTER : ZEBRA_FEC_UNREGISTER; | |
272 | zclient_create_header(s, command, VRF_DEFAULT); | |
273 | flags_pos = stream_get_endp(s); /* save position of 'flags' */ | |
274 | stream_putw(s, flags); /* initial flags */ | |
275 | stream_putw(s, PREFIX_FAMILY(p)); | |
276 | stream_put_prefix(s, p); | |
277 | if (reg) { | |
57592a53 AD |
278 | if (have_label_to_reg) { |
279 | flags |= ZEBRA_FEC_REGISTER_LABEL; | |
280 | stream_putl(s, label_pton(local_label)); | |
281 | } else if (with_label_index) { | |
282 | flags |= ZEBRA_FEC_REGISTER_LABEL_INDEX; | |
283 | stream_putl(s, pi->attr->label_index); | |
d62a17ae | 284 | } |
285 | SET_FLAG(rn->flags, BGP_NODE_REGISTERED_FOR_LABEL); | |
286 | } else | |
287 | UNSET_FLAG(rn->flags, BGP_NODE_REGISTERED_FOR_LABEL); | |
288 | ||
289 | /* Set length and flags */ | |
290 | stream_putw_at(s, 0, stream_get_endp(s)); | |
e0b84ba1 DS |
291 | |
292 | /* | |
293 | * We only need to write new flags if this is a register | |
294 | */ | |
295 | if (reg) | |
296 | stream_putw_at(s, flags_pos, flags); | |
d62a17ae | 297 | |
298 | zclient_send_message(zclient); | |
cd1964ff DS |
299 | } |
300 | ||
d7c0a89a | 301 | static int bgp_nlri_get_labels(struct peer *peer, uint8_t *pnt, uint8_t plen, |
d62a17ae | 302 | mpls_label_t *label) |
cd1964ff | 303 | { |
d7c0a89a QY |
304 | uint8_t *data = pnt; |
305 | uint8_t *lim = pnt + plen; | |
306 | uint8_t llen = 0; | |
307 | uint8_t label_depth = 0; | |
d62a17ae | 308 | |
309 | for (; data < lim; data += BGP_LABEL_BYTES) { | |
310 | memcpy(label, data, BGP_LABEL_BYTES); | |
311 | llen += BGP_LABEL_BYTES; | |
312 | ||
313 | bgp_set_valid_label(label); | |
314 | label_depth += 1; | |
315 | ||
316 | if (bgp_is_withdraw_label(label) || label_bos(label)) | |
317 | break; | |
318 | } | |
319 | ||
320 | /* If we RX multiple labels we will end up keeping only the last | |
321 | * one. We do not yet support a label stack greater than 1. */ | |
322 | if (label_depth > 1) | |
01c4f23c | 323 | zlog_info("%s rcvd UPDATE with label stack %d deep", peer->host, |
d62a17ae | 324 | label_depth); |
325 | ||
326 | if (!(bgp_is_withdraw_label(label) || label_bos(label))) | |
01c4f23c | 327 | flog_warn( |
e50f7cfd | 328 | EC_BGP_INVALID_LABEL_STACK, |
d62a17ae | 329 | "%s rcvd UPDATE with invalid label stack - no bottom of stack", |
330 | peer->host); | |
331 | ||
332 | return llen; | |
cd1964ff DS |
333 | } |
334 | ||
d62a17ae | 335 | int bgp_nlri_parse_label(struct peer *peer, struct attr *attr, |
336 | struct bgp_nlri *packet) | |
cd1964ff | 337 | { |
d7c0a89a QY |
338 | uint8_t *pnt; |
339 | uint8_t *lim; | |
d62a17ae | 340 | struct prefix p; |
341 | int psize = 0; | |
342 | int prefixlen; | |
343 | afi_t afi; | |
344 | safi_t safi; | |
345 | int addpath_encoded; | |
d7c0a89a | 346 | uint32_t addpath_id; |
d62a17ae | 347 | mpls_label_t label = MPLS_INVALID_LABEL; |
d7c0a89a | 348 | uint8_t llen; |
d62a17ae | 349 | |
d62a17ae | 350 | pnt = packet->nlri; |
351 | lim = pnt + packet->length; | |
352 | afi = packet->afi; | |
353 | safi = packet->safi; | |
354 | addpath_id = 0; | |
355 | ||
356 | addpath_encoded = | |
357 | (CHECK_FLAG(peer->af_cap[afi][safi], PEER_CAP_ADDPATH_AF_RX_ADV) | |
358 | && CHECK_FLAG(peer->af_cap[afi][safi], | |
359 | PEER_CAP_ADDPATH_AF_TX_RCV)); | |
360 | ||
361 | for (; pnt < lim; pnt += psize) { | |
362 | /* Clear prefix structure. */ | |
363 | memset(&p, 0, sizeof(struct prefix)); | |
d62a17ae | 364 | |
365 | if (addpath_encoded) { | |
366 | ||
367 | /* When packet overflow occurs return immediately. */ | |
368 | if (pnt + BGP_ADDPATH_ID_LEN > lim) | |
513386b5 | 369 | return BGP_NLRI_PARSE_ERROR_PACKET_OVERFLOW; |
d62a17ae | 370 | |
a3a850a1 S |
371 | memcpy(&addpath_id, pnt, BGP_ADDPATH_ID_LEN); |
372 | addpath_id = ntohl(addpath_id); | |
d62a17ae | 373 | pnt += BGP_ADDPATH_ID_LEN; |
374 | } | |
375 | ||
376 | /* Fetch prefix length. */ | |
377 | prefixlen = *pnt++; | |
378 | p.family = afi2family(packet->afi); | |
379 | psize = PSIZE(prefixlen); | |
380 | ||
381 | /* sanity check against packet data */ | |
382 | if ((pnt + psize) > lim) { | |
af4c2728 | 383 | flog_err( |
e50f7cfd | 384 | EC_BGP_UPDATE_RCV, |
d62a17ae | 385 | "%s [Error] Update packet error / L-U (prefix length %d exceeds packet size %u)", |
386 | peer->host, prefixlen, (uint)(lim - pnt)); | |
513386b5 | 387 | return BGP_NLRI_PARSE_ERROR_PACKET_OVERFLOW; |
d62a17ae | 388 | } |
389 | ||
390 | /* Fill in the labels */ | |
391 | llen = bgp_nlri_get_labels(peer, pnt, psize, &label); | |
392 | p.prefixlen = prefixlen - BSIZE(llen); | |
393 | ||
394 | /* There needs to be at least one label */ | |
395 | if (prefixlen < 24) { | |
e50f7cfd | 396 | flog_err(EC_BGP_UPDATE_RCV, |
1c50c1c0 QY |
397 | "%s [Error] Update packet error" |
398 | " (wrong label length %d)", | |
399 | peer->host, prefixlen); | |
d62a17ae | 400 | bgp_notify_send(peer, BGP_NOTIFY_UPDATE_ERR, |
401 | BGP_NOTIFY_UPDATE_INVAL_NETWORK); | |
513386b5 | 402 | return BGP_NLRI_PARSE_ERROR_LABEL_LENGTH; |
d62a17ae | 403 | } |
404 | ||
405 | if ((afi == AFI_IP && p.prefixlen > 32) | |
406 | || (afi == AFI_IP6 && p.prefixlen > 128)) | |
513386b5 | 407 | return BGP_NLRI_PARSE_ERROR_PREFIX_LENGTH; |
d62a17ae | 408 | |
409 | /* Fetch prefix from NLRI packet */ | |
410 | memcpy(&p.u.prefix, pnt + llen, psize - llen); | |
411 | ||
412 | /* Check address. */ | |
413 | if (afi == AFI_IP && safi == SAFI_LABELED_UNICAST) { | |
414 | if (IN_CLASSD(ntohl(p.u.prefix4.s_addr))) { | |
415 | /* From RFC4271 Section 6.3: | |
416 | * | |
417 | * If a prefix in the NLRI field is semantically | |
418 | * incorrect | |
419 | * (e.g., an unexpected multicast IP address), | |
420 | * an error SHOULD | |
421 | * be logged locally, and the prefix SHOULD be | |
422 | * ignored. | |
9d303b37 | 423 | */ |
af4c2728 | 424 | flog_err( |
e50f7cfd | 425 | EC_BGP_UPDATE_RCV, |
d62a17ae | 426 | "%s: IPv4 labeled-unicast NLRI is multicast address %s, ignoring", |
427 | peer->host, inet_ntoa(p.u.prefix4)); | |
428 | continue; | |
429 | } | |
430 | } | |
431 | ||
432 | /* Check address. */ | |
433 | if (afi == AFI_IP6 && safi == SAFI_LABELED_UNICAST) { | |
434 | if (IN6_IS_ADDR_LINKLOCAL(&p.u.prefix6)) { | |
435 | char buf[BUFSIZ]; | |
436 | ||
af4c2728 | 437 | flog_err( |
e50f7cfd | 438 | EC_BGP_UPDATE_RCV, |
d62a17ae | 439 | "%s: IPv6 labeled-unicast NLRI is link-local address %s, ignoring", |
440 | peer->host, | |
441 | inet_ntop(AF_INET6, &p.u.prefix6, buf, | |
442 | BUFSIZ)); | |
443 | ||
444 | continue; | |
445 | } | |
446 | ||
447 | if (IN6_IS_ADDR_MULTICAST(&p.u.prefix6)) { | |
448 | char buf[BUFSIZ]; | |
449 | ||
af4c2728 | 450 | flog_err( |
e50f7cfd | 451 | EC_BGP_UPDATE_RCV, |
d62a17ae | 452 | "%s: IPv6 unicast NLRI is multicast address %s, ignoring", |
453 | peer->host, | |
454 | inet_ntop(AF_INET6, &p.u.prefix6, buf, | |
455 | BUFSIZ)); | |
456 | ||
457 | continue; | |
458 | } | |
459 | } | |
460 | ||
461 | if (attr) { | |
462 | bgp_update(peer, &p, addpath_id, attr, packet->afi, | |
463 | SAFI_UNICAST, ZEBRA_ROUTE_BGP, | |
b57ba6d2 | 464 | BGP_ROUTE_NORMAL, NULL, &label, 1, 0, NULL); |
d62a17ae | 465 | } else { |
466 | bgp_withdraw(peer, &p, addpath_id, attr, packet->afi, | |
467 | SAFI_UNICAST, ZEBRA_ROUTE_BGP, | |
b57ba6d2 | 468 | BGP_ROUTE_NORMAL, NULL, &label, 1, NULL); |
d62a17ae | 469 | } |
470 | } | |
471 | ||
472 | /* Packet length consistency check. */ | |
473 | if (pnt != lim) { | |
af4c2728 | 474 | flog_err( |
e50f7cfd | 475 | EC_BGP_UPDATE_RCV, |
d62a17ae | 476 | "%s [Error] Update packet error / L-U (%zu data remaining after parsing)", |
477 | peer->host, lim - pnt); | |
513386b5 | 478 | return BGP_NLRI_PARSE_ERROR_PACKET_LENGTH; |
d62a17ae | 479 | } |
480 | ||
513386b5 | 481 | return BGP_NLRI_PARSE_OK; |
cd1964ff | 482 | } |