]>
Commit | Line | Data |
---|---|---|
1 | /* Ethernet-VPN Packet and vty Processing File | |
2 | * Copyright (C) 2016 6WIND | |
3 | * Copyright (C) 2017 Cumulus Networks, Inc. | |
4 | * | |
5 | * This file is part of FRR. | |
6 | * | |
7 | * FRRouting is free software; you can redistribute it and/or modify it | |
8 | * under the terms of the GNU General Public License as published by the | |
9 | * Free Software Foundation; either version 2, or (at your option) any | |
10 | * later version. | |
11 | * | |
12 | * FRRouting is distributed in the hope that it will be useful, but | |
13 | * WITHOUT ANY WARRANTY; without even the implied warranty of | |
14 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU | |
15 | * General Public License for more details. | |
16 | * | |
17 | * You should have received a copy of the GNU General Public License along | |
18 | * with this program; see the file COPYING; if not, write to the Free Software | |
19 | * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA | |
20 | */ | |
21 | ||
22 | #include <zebra.h> | |
23 | ||
24 | #include "command.h" | |
25 | #include "filter.h" | |
26 | #include "prefix.h" | |
27 | #include "log.h" | |
28 | #include "memory.h" | |
29 | #include "stream.h" | |
30 | #include "hash.h" | |
31 | #include "jhash.h" | |
32 | #include "zclient.h" | |
33 | ||
34 | #include "bgpd/bgp_attr_evpn.h" | |
35 | #include "bgpd/bgpd.h" | |
36 | #include "bgpd/bgp_table.h" | |
37 | #include "bgpd/bgp_route.h" | |
38 | #include "bgpd/bgp_attr.h" | |
39 | #include "bgpd/bgp_mplsvpn.h" | |
40 | #include "bgpd/bgp_label.h" | |
41 | #include "bgpd/bgp_evpn.h" | |
42 | #include "bgpd/bgp_evpn_private.h" | |
43 | #include "bgpd/bgp_ecommunity.h" | |
44 | #include "bgpd/bgp_encap_types.h" | |
45 | #include "bgpd/bgp_debug.h" | |
46 | #include "bgpd/bgp_errors.h" | |
47 | #include "bgpd/bgp_aspath.h" | |
48 | #include "bgpd/bgp_zebra.h" | |
49 | #include "bgpd/bgp_nexthop.h" | |
50 | ||
51 | /* | |
52 | * Definitions and external declarations. | |
53 | */ | |
54 | extern struct zclient *zclient; | |
55 | ||
56 | DEFINE_QOBJ_TYPE(bgpevpn) | |
57 | DEFINE_QOBJ_TYPE(evpnes) | |
58 | ||
59 | ||
60 | /* | |
61 | * Static function declarations | |
62 | */ | |
63 | static void delete_evpn_route_entry(struct bgp *bgp, | |
64 | afi_t afi, safi_t safi, | |
65 | struct bgp_node *rn, | |
66 | struct bgp_info **ri); | |
67 | static int delete_all_vni_routes(struct bgp *bgp, struct bgpevpn *vpn); | |
68 | ||
69 | /* | |
70 | * Private functions. | |
71 | */ | |
72 | ||
73 | /* compare two IPV4 VTEP IPs */ | |
74 | static int evpn_vtep_ip_cmp(const void *p1, const void *p2) | |
75 | { | |
76 | const struct in_addr *ip1 = p1; | |
77 | const struct in_addr *ip2 = p2; | |
78 | ||
79 | if (!ip1 && !ip2) | |
80 | return 1; | |
81 | if (!ip1 || !ip2) | |
82 | return 0; | |
83 | return (ip1->s_addr == ip2->s_addr); | |
84 | } | |
85 | ||
86 | /* | |
87 | * Make hash key for ESI. | |
88 | */ | |
89 | static unsigned int esi_hash_keymake(void *p) | |
90 | { | |
91 | struct evpnes *pes = p; | |
92 | const void *pnt = (void *)pes->esi.val; | |
93 | ||
94 | return jhash(pnt, ESI_BYTES, 0xa5a5a55a); | |
95 | } | |
96 | ||
97 | /* | |
98 | * Compare two ESIs. | |
99 | */ | |
100 | static int esi_cmp(const void *p1, const void *p2) | |
101 | { | |
102 | const struct evpnes *pes1 = p1; | |
103 | const struct evpnes *pes2 = p2; | |
104 | ||
105 | if (pes1 == NULL && pes2 == NULL) | |
106 | return 1; | |
107 | ||
108 | if (pes1 == NULL || pes2 == NULL) | |
109 | return 0; | |
110 | ||
111 | return (memcmp(pes1->esi.val, pes2->esi.val, ESI_BYTES) == 0); | |
112 | } | |
113 | ||
114 | /* | |
115 | * Make vni hash key. | |
116 | */ | |
117 | static unsigned int vni_hash_key_make(void *p) | |
118 | { | |
119 | struct bgpevpn *vpn = p; | |
120 | return (jhash_1word(vpn->vni, 0)); | |
121 | } | |
122 | ||
123 | /* | |
124 | * Comparison function for vni hash | |
125 | */ | |
126 | static int vni_hash_cmp(const void *p1, const void *p2) | |
127 | { | |
128 | const struct bgpevpn *vpn1 = p1; | |
129 | const struct bgpevpn *vpn2 = p2; | |
130 | ||
131 | if (!vpn1 && !vpn2) | |
132 | return 1; | |
133 | if (!vpn1 || !vpn2) | |
134 | return 0; | |
135 | return (vpn1->vni == vpn2->vni); | |
136 | } | |
137 | ||
138 | /* | |
139 | * Make vrf import route target hash key. | |
140 | */ | |
141 | static unsigned int vrf_import_rt_hash_key_make(void *p) | |
142 | { | |
143 | struct vrf_irt_node *irt = p; | |
144 | char *pnt = irt->rt.val; | |
145 | ||
146 | return jhash(pnt, 8, 0x5abc1234); | |
147 | } | |
148 | ||
149 | /* | |
150 | * Comparison function for vrf import rt hash | |
151 | */ | |
152 | static int vrf_import_rt_hash_cmp(const void *p1, const void *p2) | |
153 | { | |
154 | const struct vrf_irt_node *irt1 = p1; | |
155 | const struct vrf_irt_node *irt2 = p2; | |
156 | ||
157 | if (irt1 == NULL && irt2 == NULL) | |
158 | return 1; | |
159 | ||
160 | if (irt1 == NULL || irt2 == NULL) | |
161 | return 0; | |
162 | ||
163 | return (memcmp(irt1->rt.val, irt2->rt.val, ECOMMUNITY_SIZE) == 0); | |
164 | } | |
165 | ||
166 | /* | |
167 | * Create a new vrf import_rt in default instance | |
168 | */ | |
169 | static struct vrf_irt_node *vrf_import_rt_new(struct ecommunity_val *rt) | |
170 | { | |
171 | struct bgp *bgp_def = NULL; | |
172 | struct vrf_irt_node *irt; | |
173 | ||
174 | bgp_def = bgp_get_default(); | |
175 | if (!bgp_def) { | |
176 | flog_err(EC_BGP_NO_DFLT, | |
177 | "vrf import rt new - def instance not created yet"); | |
178 | return NULL; | |
179 | } | |
180 | ||
181 | irt = XCALLOC(MTYPE_BGP_EVPN_VRF_IMPORT_RT, | |
182 | sizeof(struct vrf_irt_node)); | |
183 | if (!irt) | |
184 | return NULL; | |
185 | ||
186 | irt->rt = *rt; | |
187 | irt->vrfs = list_new(); | |
188 | ||
189 | /* Add to hash */ | |
190 | if (!hash_get(bgp_def->vrf_import_rt_hash, irt, hash_alloc_intern)) { | |
191 | XFREE(MTYPE_BGP_EVPN_VRF_IMPORT_RT, irt); | |
192 | return NULL; | |
193 | } | |
194 | ||
195 | return irt; | |
196 | } | |
197 | ||
198 | /* | |
199 | * Free the vrf import rt node | |
200 | */ | |
201 | static void vrf_import_rt_free(struct vrf_irt_node *irt) | |
202 | { | |
203 | struct bgp *bgp_def = NULL; | |
204 | ||
205 | bgp_def = bgp_get_default(); | |
206 | if (!bgp_def) { | |
207 | flog_err(EC_BGP_NO_DFLT, | |
208 | "vrf import rt free - def instance not created yet"); | |
209 | return; | |
210 | } | |
211 | ||
212 | hash_release(bgp_def->vrf_import_rt_hash, irt); | |
213 | list_delete_and_null(&irt->vrfs); | |
214 | XFREE(MTYPE_BGP_EVPN_VRF_IMPORT_RT, irt); | |
215 | } | |
216 | ||
217 | /* | |
218 | * Function to lookup Import RT node - used to map a RT to set of | |
219 | * VNIs importing routes with that RT. | |
220 | */ | |
221 | static struct vrf_irt_node *lookup_vrf_import_rt(struct ecommunity_val *rt) | |
222 | { | |
223 | struct bgp *bgp_def = NULL; | |
224 | struct vrf_irt_node *irt; | |
225 | struct vrf_irt_node tmp; | |
226 | ||
227 | bgp_def = bgp_get_default(); | |
228 | if (!bgp_def) { | |
229 | flog_err(EC_BGP_NO_DFLT, | |
230 | "vrf import rt lookup - def instance not created yet"); | |
231 | return NULL; | |
232 | } | |
233 | ||
234 | memset(&tmp, 0, sizeof(struct vrf_irt_node)); | |
235 | memcpy(&tmp.rt, rt, ECOMMUNITY_SIZE); | |
236 | irt = hash_lookup(bgp_def->vrf_import_rt_hash, &tmp); | |
237 | return irt; | |
238 | } | |
239 | ||
240 | /* | |
241 | * Is specified VRF present on the RT's list of "importing" VRFs? | |
242 | */ | |
243 | static int is_vrf_present_in_irt_vrfs(struct list *vrfs, struct bgp *bgp_vrf) | |
244 | { | |
245 | struct listnode *node = NULL, *nnode = NULL; | |
246 | struct bgp *tmp_bgp_vrf = NULL; | |
247 | ||
248 | for (ALL_LIST_ELEMENTS(vrfs, node, nnode, tmp_bgp_vrf)) { | |
249 | if (tmp_bgp_vrf == bgp_vrf) | |
250 | return 1; | |
251 | } | |
252 | return 0; | |
253 | } | |
254 | ||
255 | /* | |
256 | * Make import route target hash key. | |
257 | */ | |
258 | static unsigned int import_rt_hash_key_make(void *p) | |
259 | { | |
260 | struct irt_node *irt = p; | |
261 | char *pnt = irt->rt.val; | |
262 | ||
263 | return jhash(pnt, 8, 0xdeadbeef); | |
264 | } | |
265 | ||
266 | /* | |
267 | * Comparison function for import rt hash | |
268 | */ | |
269 | static int import_rt_hash_cmp(const void *p1, const void *p2) | |
270 | { | |
271 | const struct irt_node *irt1 = p1; | |
272 | const struct irt_node *irt2 = p2; | |
273 | ||
274 | if (irt1 == NULL && irt2 == NULL) | |
275 | return 1; | |
276 | ||
277 | if (irt1 == NULL || irt2 == NULL) | |
278 | return 0; | |
279 | ||
280 | return (memcmp(irt1->rt.val, irt2->rt.val, ECOMMUNITY_SIZE) == 0); | |
281 | } | |
282 | ||
283 | /* | |
284 | * Create a new import_rt | |
285 | */ | |
286 | static struct irt_node *import_rt_new(struct bgp *bgp, | |
287 | struct ecommunity_val *rt) | |
288 | { | |
289 | struct irt_node *irt; | |
290 | ||
291 | if (!bgp) | |
292 | return NULL; | |
293 | ||
294 | irt = XCALLOC(MTYPE_BGP_EVPN_IMPORT_RT, sizeof(struct irt_node)); | |
295 | if (!irt) | |
296 | return NULL; | |
297 | ||
298 | irt->rt = *rt; | |
299 | irt->vnis = list_new(); | |
300 | ||
301 | /* Add to hash */ | |
302 | if (!hash_get(bgp->import_rt_hash, irt, hash_alloc_intern)) { | |
303 | XFREE(MTYPE_BGP_EVPN_IMPORT_RT, irt); | |
304 | return NULL; | |
305 | } | |
306 | ||
307 | return irt; | |
308 | } | |
309 | ||
310 | /* | |
311 | * Free the import rt node | |
312 | */ | |
313 | static void import_rt_free(struct bgp *bgp, struct irt_node *irt) | |
314 | { | |
315 | hash_release(bgp->import_rt_hash, irt); | |
316 | list_delete_and_null(&irt->vnis); | |
317 | XFREE(MTYPE_BGP_EVPN_IMPORT_RT, irt); | |
318 | } | |
319 | ||
320 | /* | |
321 | * Function to lookup Import RT node - used to map a RT to set of | |
322 | * VNIs importing routes with that RT. | |
323 | */ | |
324 | static struct irt_node *lookup_import_rt(struct bgp *bgp, | |
325 | struct ecommunity_val *rt) | |
326 | { | |
327 | struct irt_node *irt; | |
328 | struct irt_node tmp; | |
329 | ||
330 | memset(&tmp, 0, sizeof(struct irt_node)); | |
331 | memcpy(&tmp.rt, rt, ECOMMUNITY_SIZE); | |
332 | irt = hash_lookup(bgp->import_rt_hash, &tmp); | |
333 | return irt; | |
334 | } | |
335 | ||
336 | /* | |
337 | * Is specified VNI present on the RT's list of "importing" VNIs? | |
338 | */ | |
339 | static int is_vni_present_in_irt_vnis(struct list *vnis, struct bgpevpn *vpn) | |
340 | { | |
341 | struct listnode *node, *nnode; | |
342 | struct bgpevpn *tmp_vpn; | |
343 | ||
344 | for (ALL_LIST_ELEMENTS(vnis, node, nnode, tmp_vpn)) { | |
345 | if (tmp_vpn == vpn) | |
346 | return 1; | |
347 | } | |
348 | ||
349 | return 0; | |
350 | } | |
351 | ||
352 | /* | |
353 | * Compare Route Targets. | |
354 | */ | |
355 | static int evpn_route_target_cmp(struct ecommunity *ecom1, | |
356 | struct ecommunity *ecom2) | |
357 | { | |
358 | if (ecom1 && !ecom2) | |
359 | return -1; | |
360 | ||
361 | if (!ecom1 && ecom2) | |
362 | return 1; | |
363 | ||
364 | if (!ecom1 && !ecom2) | |
365 | return 0; | |
366 | ||
367 | if (ecom1->str && !ecom2->str) | |
368 | return -1; | |
369 | ||
370 | if (!ecom1->str && ecom2->str) | |
371 | return 1; | |
372 | ||
373 | if (!ecom1->str && !ecom2->str) | |
374 | return 0; | |
375 | ||
376 | return strcmp(ecom1->str, ecom2->str); | |
377 | } | |
378 | ||
379 | static void evpn_xxport_delete_ecomm(void *val) | |
380 | { | |
381 | struct ecommunity *ecomm = val; | |
382 | ecommunity_free(&ecomm); | |
383 | } | |
384 | ||
385 | /* | |
386 | * Mask off global-admin field of specified extended community (RT), | |
387 | * just retain the local-admin field. | |
388 | */ | |
389 | static inline void mask_ecom_global_admin(struct ecommunity_val *dst, | |
390 | struct ecommunity_val *src) | |
391 | { | |
392 | uint8_t type; | |
393 | ||
394 | type = src->val[0]; | |
395 | dst->val[0] = 0; | |
396 | if (type == ECOMMUNITY_ENCODE_AS) { | |
397 | dst->val[2] = dst->val[3] = 0; | |
398 | } else if (type == ECOMMUNITY_ENCODE_AS4 | |
399 | || type == ECOMMUNITY_ENCODE_IP) { | |
400 | dst->val[2] = dst->val[3] = 0; | |
401 | dst->val[4] = dst->val[5] = 0; | |
402 | } | |
403 | } | |
404 | ||
405 | /* | |
406 | * Map one RT to specified VRF. | |
407 | * bgp_vrf = BGP vrf instance | |
408 | */ | |
409 | static void map_vrf_to_rt(struct bgp *bgp_vrf, struct ecommunity_val *eval) | |
410 | { | |
411 | struct vrf_irt_node *irt = NULL; | |
412 | struct ecommunity_val eval_tmp; | |
413 | ||
414 | /* If using "automatic" RT, | |
415 | * we only care about the local-admin sub-field. | |
416 | * This is to facilitate using L3VNI(VRF-VNI) | |
417 | * as the RT for EBGP peering too. | |
418 | */ | |
419 | memcpy(&eval_tmp, eval, ECOMMUNITY_SIZE); | |
420 | if (!CHECK_FLAG(bgp_vrf->vrf_flags, BGP_VRF_IMPORT_RT_CFGD)) | |
421 | mask_ecom_global_admin(&eval_tmp, eval); | |
422 | ||
423 | irt = lookup_vrf_import_rt(&eval_tmp); | |
424 | if (irt && is_vrf_present_in_irt_vrfs(irt->vrfs, bgp_vrf)) | |
425 | /* Already mapped. */ | |
426 | return; | |
427 | ||
428 | if (!irt) | |
429 | irt = vrf_import_rt_new(&eval_tmp); | |
430 | ||
431 | /* Add VRF to the list for this RT. */ | |
432 | listnode_add(irt->vrfs, bgp_vrf); | |
433 | } | |
434 | ||
435 | /* | |
436 | * Unmap specified VRF from specified RT. If there are no other | |
437 | * VRFs for this RT, then the RT hash is deleted. | |
438 | * bgp_vrf: BGP VRF specific instance | |
439 | */ | |
440 | static void unmap_vrf_from_rt(struct bgp *bgp_vrf, struct vrf_irt_node *irt) | |
441 | { | |
442 | /* Delete VRF from list for this RT. */ | |
443 | listnode_delete(irt->vrfs, bgp_vrf); | |
444 | if (!listnode_head(irt->vrfs)) { | |
445 | vrf_import_rt_free(irt); | |
446 | } | |
447 | } | |
448 | ||
449 | /* | |
450 | * Map one RT to specified VNI. | |
451 | */ | |
452 | static void map_vni_to_rt(struct bgp *bgp, struct bgpevpn *vpn, | |
453 | struct ecommunity_val *eval) | |
454 | { | |
455 | struct irt_node *irt; | |
456 | struct ecommunity_val eval_tmp; | |
457 | ||
458 | /* If using "automatic" RT, we only care about the local-admin | |
459 | * sub-field. | |
460 | * This is to facilitate using VNI as the RT for EBGP peering too. | |
461 | */ | |
462 | memcpy(&eval_tmp, eval, ECOMMUNITY_SIZE); | |
463 | if (!is_import_rt_configured(vpn)) | |
464 | mask_ecom_global_admin(&eval_tmp, eval); | |
465 | ||
466 | irt = lookup_import_rt(bgp, &eval_tmp); | |
467 | if (irt) | |
468 | if (is_vni_present_in_irt_vnis(irt->vnis, vpn)) | |
469 | /* Already mapped. */ | |
470 | return; | |
471 | ||
472 | if (!irt) { | |
473 | irt = import_rt_new(bgp, &eval_tmp); | |
474 | assert(irt); | |
475 | } | |
476 | ||
477 | /* Add VNI to the hash list for this RT. */ | |
478 | listnode_add(irt->vnis, vpn); | |
479 | } | |
480 | ||
481 | /* | |
482 | * Unmap specified VNI from specified RT. If there are no other | |
483 | * VNIs for this RT, then the RT hash is deleted. | |
484 | */ | |
485 | static void unmap_vni_from_rt(struct bgp *bgp, struct bgpevpn *vpn, | |
486 | struct irt_node *irt) | |
487 | { | |
488 | /* Delete VNI from hash list for this RT. */ | |
489 | listnode_delete(irt->vnis, vpn); | |
490 | if (!listnode_head(irt->vnis)) { | |
491 | import_rt_free(bgp, irt); | |
492 | } | |
493 | } | |
494 | ||
495 | /* | |
496 | * Create RT extended community automatically from passed information: | |
497 | * of the form AS:VNI. | |
498 | * NOTE: We use only the lower 16 bits of the AS. This is sufficient as | |
499 | * the need is to get a RT value that will be unique across different | |
500 | * VNIs but the same across routers (in the same AS) for a particular | |
501 | * VNI. | |
502 | */ | |
503 | static void form_auto_rt(struct bgp *bgp, vni_t vni, struct list *rtl) | |
504 | { | |
505 | struct ecommunity_val eval; | |
506 | struct ecommunity *ecomadd; | |
507 | ||
508 | if (bgp->advertise_autort_rfc8365) | |
509 | vni |= EVPN_AUTORT_VXLAN; | |
510 | encode_route_target_as((bgp->as & 0xFFFF), vni, &eval); | |
511 | ||
512 | ecomadd = ecommunity_new(); | |
513 | ecommunity_add_val(ecomadd, &eval); | |
514 | listnode_add_sort(rtl, ecomadd); | |
515 | } | |
516 | ||
517 | /* | |
518 | * Derive RD and RT for a VNI automatically. Invoked at the time of | |
519 | * creation of a VNI. | |
520 | */ | |
521 | static void derive_rd_rt_for_vni(struct bgp *bgp, struct bgpevpn *vpn) | |
522 | { | |
523 | bgp_evpn_derive_auto_rd(bgp, vpn); | |
524 | bgp_evpn_derive_auto_rt_import(bgp, vpn); | |
525 | bgp_evpn_derive_auto_rt_export(bgp, vpn); | |
526 | } | |
527 | ||
528 | /* | |
529 | * Convert nexthop (remote VTEP IP) into an IPv6 address. | |
530 | */ | |
531 | static void evpn_convert_nexthop_to_ipv6(struct attr *attr) | |
532 | { | |
533 | if (BGP_ATTR_NEXTHOP_AFI_IP6(attr)) | |
534 | return; | |
535 | ipv4_to_ipv4_mapped_ipv6(&attr->mp_nexthop_global, attr->nexthop); | |
536 | attr->mp_nexthop_len = IPV6_MAX_BYTELEN; | |
537 | } | |
538 | ||
539 | /* | |
540 | * Add (update) or delete MACIP from zebra. | |
541 | */ | |
542 | static int bgp_zebra_send_remote_macip(struct bgp *bgp, struct bgpevpn *vpn, | |
543 | struct prefix_evpn *p, | |
544 | struct in_addr remote_vtep_ip, int add, | |
545 | uint8_t flags, uint32_t seq) | |
546 | { | |
547 | struct stream *s; | |
548 | int ipa_len; | |
549 | char buf1[ETHER_ADDR_STRLEN]; | |
550 | char buf2[INET6_ADDRSTRLEN]; | |
551 | char buf3[INET6_ADDRSTRLEN]; | |
552 | ||
553 | /* Check socket. */ | |
554 | if (!zclient || zclient->sock < 0) | |
555 | return 0; | |
556 | ||
557 | /* Don't try to register if Zebra doesn't know of this instance. */ | |
558 | if (!IS_BGP_INST_KNOWN_TO_ZEBRA(bgp)) | |
559 | return 0; | |
560 | ||
561 | s = zclient->obuf; | |
562 | stream_reset(s); | |
563 | ||
564 | zclient_create_header( | |
565 | s, add ? ZEBRA_REMOTE_MACIP_ADD : ZEBRA_REMOTE_MACIP_DEL, | |
566 | bgp->vrf_id); | |
567 | stream_putl(s, vpn->vni); | |
568 | stream_put(s, &p->prefix.macip_addr.mac.octet, ETH_ALEN); /* Mac Addr */ | |
569 | /* IP address length and IP address, if any. */ | |
570 | if (is_evpn_prefix_ipaddr_none(p)) | |
571 | stream_putl(s, 0); | |
572 | else { | |
573 | ipa_len = is_evpn_prefix_ipaddr_v4(p) ? IPV4_MAX_BYTELEN | |
574 | : IPV6_MAX_BYTELEN; | |
575 | stream_putl(s, ipa_len); | |
576 | stream_put(s, &p->prefix.macip_addr.ip.ip.addr, ipa_len); | |
577 | } | |
578 | stream_put_in_addr(s, &remote_vtep_ip); | |
579 | ||
580 | /* TX flags - MAC sticky status and/or gateway mac */ | |
581 | /* Also TX the sequence number of the best route. */ | |
582 | if (add) { | |
583 | stream_putc(s, flags); | |
584 | stream_putl(s, seq); | |
585 | } | |
586 | ||
587 | stream_putw_at(s, 0, stream_get_endp(s)); | |
588 | ||
589 | if (bgp_debug_zebra(NULL)) | |
590 | zlog_debug( | |
591 | "Tx %s MACIP, VNI %u MAC %s IP %s flags 0x%x seq %u remote VTEP %s", | |
592 | add ? "ADD" : "DEL", vpn->vni, | |
593 | prefix_mac2str(&p->prefix.macip_addr.mac, | |
594 | buf1, sizeof(buf1)), | |
595 | ipaddr2str(&p->prefix.macip_addr.ip, | |
596 | buf3, sizeof(buf3)), flags, seq, | |
597 | inet_ntop(AF_INET, &remote_vtep_ip, buf2, | |
598 | sizeof(buf2))); | |
599 | ||
600 | return zclient_send_message(zclient); | |
601 | } | |
602 | ||
603 | /* | |
604 | * Add (update) or delete remote VTEP from zebra. | |
605 | */ | |
606 | static int bgp_zebra_send_remote_vtep(struct bgp *bgp, struct bgpevpn *vpn, | |
607 | struct prefix_evpn *p, int add) | |
608 | { | |
609 | struct stream *s; | |
610 | ||
611 | /* Check socket. */ | |
612 | if (!zclient || zclient->sock < 0) | |
613 | return 0; | |
614 | ||
615 | /* Don't try to register if Zebra doesn't know of this instance. */ | |
616 | if (!IS_BGP_INST_KNOWN_TO_ZEBRA(bgp)) | |
617 | return 0; | |
618 | ||
619 | s = zclient->obuf; | |
620 | stream_reset(s); | |
621 | ||
622 | zclient_create_header( | |
623 | s, add ? ZEBRA_REMOTE_VTEP_ADD : ZEBRA_REMOTE_VTEP_DEL, | |
624 | bgp->vrf_id); | |
625 | stream_putl(s, vpn->vni); | |
626 | if (is_evpn_prefix_ipaddr_v4(p)) | |
627 | stream_put_in_addr(s, &p->prefix.imet_addr.ip.ipaddr_v4); | |
628 | else if (is_evpn_prefix_ipaddr_v6(p)) { | |
629 | flog_err( | |
630 | EC_BGP_VTEP_INVALID, | |
631 | "Bad remote IP when trying to %s remote VTEP for VNI %u", | |
632 | add ? "ADD" : "DEL", vpn->vni); | |
633 | return -1; | |
634 | } | |
635 | ||
636 | stream_putw_at(s, 0, stream_get_endp(s)); | |
637 | ||
638 | if (bgp_debug_zebra(NULL)) | |
639 | zlog_debug("Tx %s Remote VTEP, VNI %u remote VTEP %s", | |
640 | add ? "ADD" : "DEL", vpn->vni, | |
641 | inet_ntoa(p->prefix.imet_addr.ip.ipaddr_v4)); | |
642 | ||
643 | return zclient_send_message(zclient); | |
644 | } | |
645 | ||
646 | /* | |
647 | * Build extended community for EVPN ES (type-4) route | |
648 | */ | |
649 | static void build_evpn_type4_route_extcomm(struct evpnes *es, | |
650 | struct attr *attr) | |
651 | { | |
652 | struct ecommunity ecom_encap; | |
653 | struct ecommunity ecom_es_rt; | |
654 | struct ecommunity_val eval; | |
655 | struct ecommunity_val eval_es_rt; | |
656 | bgp_encap_types tnl_type; | |
657 | struct ethaddr mac; | |
658 | ||
659 | /* Encap */ | |
660 | tnl_type = BGP_ENCAP_TYPE_VXLAN; | |
661 | memset(&ecom_encap, 0, sizeof(ecom_encap)); | |
662 | encode_encap_extcomm(tnl_type, &eval); | |
663 | ecom_encap.size = 1; | |
664 | ecom_encap.val = (uint8_t *)eval.val; | |
665 | attr->ecommunity = ecommunity_dup(&ecom_encap); | |
666 | ||
667 | /* ES import RT */ | |
668 | memset(&mac, 0, sizeof(struct ethaddr)); | |
669 | memset(&ecom_es_rt, 0, sizeof(ecom_es_rt)); | |
670 | es_get_system_mac(&es->esi, &mac); | |
671 | encode_es_rt_extcomm(&eval_es_rt, &mac); | |
672 | ecom_es_rt.size = 1; | |
673 | ecom_es_rt.val = (uint8_t *)eval_es_rt.val; | |
674 | attr->ecommunity = | |
675 | ecommunity_merge(attr->ecommunity, &ecom_es_rt); | |
676 | ||
677 | attr->flag |= ATTR_FLAG_BIT(BGP_ATTR_EXT_COMMUNITIES); | |
678 | } | |
679 | ||
680 | /* | |
681 | * Build extended communities for EVPN prefix route. | |
682 | */ | |
683 | static void build_evpn_type5_route_extcomm(struct bgp *bgp_vrf, | |
684 | struct attr *attr) | |
685 | { | |
686 | struct ecommunity ecom_encap; | |
687 | struct ecommunity ecom_rmac; | |
688 | struct ecommunity_val eval; | |
689 | struct ecommunity_val eval_rmac; | |
690 | bgp_encap_types tnl_type; | |
691 | struct listnode *node, *nnode; | |
692 | struct ecommunity *ecom; | |
693 | struct list *vrf_export_rtl = NULL; | |
694 | ||
695 | /* Encap */ | |
696 | tnl_type = BGP_ENCAP_TYPE_VXLAN; | |
697 | memset(&ecom_encap, 0, sizeof(ecom_encap)); | |
698 | encode_encap_extcomm(tnl_type, &eval); | |
699 | ecom_encap.size = 1; | |
700 | ecom_encap.val = (uint8_t *)eval.val; | |
701 | ||
702 | /* Add Encap */ | |
703 | attr->ecommunity = ecommunity_dup(&ecom_encap); | |
704 | ||
705 | /* Add the export RTs for L3VNI/VRF */ | |
706 | vrf_export_rtl = bgp_vrf->vrf_export_rtl; | |
707 | for (ALL_LIST_ELEMENTS(vrf_export_rtl, node, nnode, ecom)) | |
708 | attr->ecommunity = | |
709 | ecommunity_merge(attr->ecommunity, ecom); | |
710 | ||
711 | /* add the router mac extended community */ | |
712 | if (!is_zero_mac(&attr->rmac)) { | |
713 | memset(&ecom_rmac, 0, sizeof(ecom_rmac)); | |
714 | encode_rmac_extcomm(&eval_rmac, &attr->rmac); | |
715 | ecom_rmac.size = 1; | |
716 | ecom_rmac.val = (uint8_t *)eval_rmac.val; | |
717 | attr->ecommunity = | |
718 | ecommunity_merge(attr->ecommunity, &ecom_rmac); | |
719 | } | |
720 | ||
721 | attr->flag |= ATTR_FLAG_BIT(BGP_ATTR_EXT_COMMUNITIES); | |
722 | } | |
723 | ||
724 | /* | |
725 | * Build extended communities for EVPN route. | |
726 | * This function is applicable for type-2 and type-3 routes. The layer-2 RT | |
727 | * and ENCAP extended communities are applicable for all routes. | |
728 | * The default gateway extended community and MAC mobility (sticky) extended | |
729 | * community are added as needed based on passed settings - only for type-2 | |
730 | * routes. Likewise, the layer-3 RT and Router MAC extended communities are | |
731 | * added, if present, based on passed settings - only for non-link-local | |
732 | * type-2 routes. | |
733 | */ | |
734 | static void build_evpn_route_extcomm(struct bgpevpn *vpn, struct attr *attr, | |
735 | int add_l3_ecomm) | |
736 | { | |
737 | struct ecommunity ecom_encap; | |
738 | struct ecommunity ecom_sticky; | |
739 | struct ecommunity ecom_default_gw; | |
740 | struct ecommunity ecom_rmac; | |
741 | struct ecommunity ecom_na; | |
742 | struct ecommunity_val eval; | |
743 | struct ecommunity_val eval_sticky; | |
744 | struct ecommunity_val eval_default_gw; | |
745 | struct ecommunity_val eval_rmac; | |
746 | struct ecommunity_val eval_na; | |
747 | ||
748 | bgp_encap_types tnl_type; | |
749 | struct listnode *node, *nnode; | |
750 | struct ecommunity *ecom; | |
751 | uint32_t seqnum; | |
752 | struct list *vrf_export_rtl = NULL; | |
753 | ||
754 | /* Encap */ | |
755 | tnl_type = BGP_ENCAP_TYPE_VXLAN; | |
756 | memset(&ecom_encap, 0, sizeof(ecom_encap)); | |
757 | encode_encap_extcomm(tnl_type, &eval); | |
758 | ecom_encap.size = 1; | |
759 | ecom_encap.val = (uint8_t *)eval.val; | |
760 | ||
761 | /* Add Encap */ | |
762 | attr->ecommunity = ecommunity_dup(&ecom_encap); | |
763 | ||
764 | /* Add the export RTs for L2VNI */ | |
765 | for (ALL_LIST_ELEMENTS(vpn->export_rtl, node, nnode, ecom)) | |
766 | attr->ecommunity = ecommunity_merge(attr->ecommunity, ecom); | |
767 | ||
768 | /* Add the export RTs for L3VNI if told to - caller determines | |
769 | * when this should be done. | |
770 | */ | |
771 | if (add_l3_ecomm) { | |
772 | vrf_export_rtl = bgpevpn_get_vrf_export_rtl(vpn); | |
773 | if (vrf_export_rtl && !list_isempty(vrf_export_rtl)) { | |
774 | for (ALL_LIST_ELEMENTS(vrf_export_rtl, node, nnode, | |
775 | ecom)) | |
776 | attr->ecommunity = ecommunity_merge( | |
777 | attr->ecommunity, ecom); | |
778 | } | |
779 | } | |
780 | ||
781 | /* Add MAC mobility (sticky) if needed. */ | |
782 | if (attr->sticky) { | |
783 | seqnum = 0; | |
784 | memset(&ecom_sticky, 0, sizeof(ecom_sticky)); | |
785 | encode_mac_mobility_extcomm(1, seqnum, &eval_sticky); | |
786 | ecom_sticky.size = 1; | |
787 | ecom_sticky.val = (uint8_t *)eval_sticky.val; | |
788 | attr->ecommunity = | |
789 | ecommunity_merge(attr->ecommunity, &ecom_sticky); | |
790 | } | |
791 | ||
792 | /* Add RMAC, if told to. */ | |
793 | if (add_l3_ecomm) { | |
794 | memset(&ecom_rmac, 0, sizeof(ecom_rmac)); | |
795 | encode_rmac_extcomm(&eval_rmac, &attr->rmac); | |
796 | ecom_rmac.size = 1; | |
797 | ecom_rmac.val = (uint8_t *)eval_rmac.val; | |
798 | attr->ecommunity = | |
799 | ecommunity_merge(attr->ecommunity, &ecom_rmac); | |
800 | } | |
801 | ||
802 | /* Add default gateway, if needed. */ | |
803 | if (attr->default_gw) { | |
804 | memset(&ecom_default_gw, 0, sizeof(ecom_default_gw)); | |
805 | encode_default_gw_extcomm(&eval_default_gw); | |
806 | ecom_default_gw.size = 1; | |
807 | ecom_default_gw.val = (uint8_t *)eval_default_gw.val; | |
808 | attr->ecommunity = | |
809 | ecommunity_merge(attr->ecommunity, &ecom_default_gw); | |
810 | } | |
811 | ||
812 | if (attr->router_flag) { | |
813 | memset(&ecom_na, 0, sizeof(ecom_na)); | |
814 | encode_na_flag_extcomm(&eval_na, attr->router_flag); | |
815 | ecom_na.size = 1; | |
816 | ecom_na.val = (uint8_t *)eval_na.val; | |
817 | attr->ecommunity = ecommunity_merge(attr->ecommunity, | |
818 | &ecom_na); | |
819 | } | |
820 | ||
821 | attr->flag |= ATTR_FLAG_BIT(BGP_ATTR_EXT_COMMUNITIES); | |
822 | } | |
823 | ||
824 | /* | |
825 | * Add MAC mobility extended community to attribute. | |
826 | */ | |
827 | static void add_mac_mobility_to_attr(uint32_t seq_num, struct attr *attr) | |
828 | { | |
829 | struct ecommunity ecom_tmp; | |
830 | struct ecommunity_val eval; | |
831 | uint8_t *ecom_val_ptr; | |
832 | int i; | |
833 | uint8_t *pnt; | |
834 | int type = 0; | |
835 | int sub_type = 0; | |
836 | ||
837 | /* Build MM */ | |
838 | encode_mac_mobility_extcomm(0, seq_num, &eval); | |
839 | ||
840 | /* Find current MM ecommunity */ | |
841 | ecom_val_ptr = NULL; | |
842 | ||
843 | if (attr->ecommunity) { | |
844 | for (i = 0; i < attr->ecommunity->size; i++) { | |
845 | pnt = attr->ecommunity->val + (i * 8); | |
846 | type = *pnt++; | |
847 | sub_type = *pnt++; | |
848 | ||
849 | if (type == ECOMMUNITY_ENCODE_EVPN | |
850 | && sub_type | |
851 | == ECOMMUNITY_EVPN_SUBTYPE_MACMOBILITY) { | |
852 | ecom_val_ptr = (uint8_t *)(attr->ecommunity->val | |
853 | + (i * 8)); | |
854 | break; | |
855 | } | |
856 | } | |
857 | } | |
858 | ||
859 | /* Update the existing MM ecommunity */ | |
860 | if (ecom_val_ptr) { | |
861 | memcpy(ecom_val_ptr, eval.val, sizeof(char) * ECOMMUNITY_SIZE); | |
862 | } | |
863 | /* Add MM to existing */ | |
864 | else { | |
865 | memset(&ecom_tmp, 0, sizeof(ecom_tmp)); | |
866 | ecom_tmp.size = 1; | |
867 | ecom_tmp.val = (uint8_t *)eval.val; | |
868 | ||
869 | if (attr->ecommunity) | |
870 | attr->ecommunity = | |
871 | ecommunity_merge(attr->ecommunity, &ecom_tmp); | |
872 | else | |
873 | attr->ecommunity = ecommunity_dup(&ecom_tmp); | |
874 | } | |
875 | } | |
876 | ||
877 | /* Install EVPN route into zebra. */ | |
878 | static int evpn_zebra_install(struct bgp *bgp, struct bgpevpn *vpn, | |
879 | struct prefix_evpn *p, | |
880 | struct in_addr remote_vtep_ip, uint8_t flags, | |
881 | uint32_t seq) | |
882 | { | |
883 | int ret; | |
884 | ||
885 | if (p->prefix.route_type == BGP_EVPN_MAC_IP_ROUTE) | |
886 | ret = bgp_zebra_send_remote_macip(bgp, vpn, p, remote_vtep_ip, | |
887 | 1, flags, seq); | |
888 | else | |
889 | ret = bgp_zebra_send_remote_vtep(bgp, vpn, p, 1); | |
890 | ||
891 | return ret; | |
892 | } | |
893 | ||
894 | /* Uninstall EVPN route from zebra. */ | |
895 | static int evpn_zebra_uninstall(struct bgp *bgp, struct bgpevpn *vpn, | |
896 | struct prefix_evpn *p, | |
897 | struct in_addr remote_vtep_ip) | |
898 | { | |
899 | int ret; | |
900 | ||
901 | if (p->prefix.route_type == BGP_EVPN_MAC_IP_ROUTE) | |
902 | ret = bgp_zebra_send_remote_macip(bgp, vpn, p, remote_vtep_ip, | |
903 | 0, 0, 0); | |
904 | else | |
905 | ret = bgp_zebra_send_remote_vtep(bgp, vpn, p, 0); | |
906 | ||
907 | return ret; | |
908 | } | |
909 | ||
910 | /* | |
911 | * Due to MAC mobility, the prior "local" best route has been supplanted | |
912 | * by a "remote" best route. The prior route has to be deleted and withdrawn | |
913 | * from peers. | |
914 | */ | |
915 | static void evpn_delete_old_local_route(struct bgp *bgp, struct bgpevpn *vpn, | |
916 | struct bgp_node *rn, | |
917 | struct bgp_info *old_local) | |
918 | { | |
919 | struct bgp_node *global_rn; | |
920 | struct bgp_info *ri; | |
921 | afi_t afi = AFI_L2VPN; | |
922 | safi_t safi = SAFI_EVPN; | |
923 | ||
924 | /* Locate route node in the global EVPN routing table. Note that | |
925 | * this table is a 2-level tree (RD-level + Prefix-level) similar to | |
926 | * L3VPN routes. | |
927 | */ | |
928 | global_rn = bgp_afi_node_lookup(bgp->rib[afi][safi], afi, safi, | |
929 | (struct prefix *)&rn->p, &vpn->prd); | |
930 | if (global_rn) { | |
931 | /* Delete route entry in the global EVPN table. */ | |
932 | delete_evpn_route_entry(bgp, afi, safi, global_rn, &ri); | |
933 | ||
934 | /* Schedule for processing - withdraws to peers happen from | |
935 | * this table. | |
936 | */ | |
937 | if (ri) | |
938 | bgp_process(bgp, global_rn, afi, safi); | |
939 | bgp_unlock_node(global_rn); | |
940 | } | |
941 | ||
942 | /* Delete route entry in the VNI route table, caller to remove. */ | |
943 | bgp_info_delete(rn, old_local); | |
944 | } | |
945 | ||
946 | static struct in_addr *es_vtep_new(struct in_addr vtep) | |
947 | { | |
948 | struct in_addr *ip; | |
949 | ||
950 | ip = XCALLOC(MTYPE_BGP_EVPN_ES_VTEP, sizeof(struct in_addr)); | |
951 | if (!ip) | |
952 | return NULL; | |
953 | ||
954 | ip->s_addr = vtep.s_addr; | |
955 | return ip; | |
956 | } | |
957 | ||
958 | static void es_vtep_free(struct in_addr *ip) | |
959 | { | |
960 | XFREE(MTYPE_BGP_EVPN_ES_VTEP, ip); | |
961 | } | |
962 | ||
963 | /* check if VTEP is already part of the list */ | |
964 | static int is_vtep_present_in_list(struct list *list, | |
965 | struct in_addr vtep) | |
966 | { | |
967 | struct listnode *node = NULL; | |
968 | struct in_addr *tmp; | |
969 | ||
970 | for (ALL_LIST_ELEMENTS_RO(list, node, tmp)) { | |
971 | if (tmp->s_addr == vtep.s_addr) | |
972 | return 1; | |
973 | } | |
974 | return 0; | |
975 | } | |
976 | ||
977 | /* | |
978 | * Best path for ES route was changed, | |
979 | * update the list of VTEPs for this ES | |
980 | */ | |
981 | static int evpn_es_install_vtep(struct bgp *bgp, | |
982 | struct evpnes *es, | |
983 | struct prefix_evpn *p, | |
984 | struct in_addr rvtep) | |
985 | { | |
986 | struct in_addr *vtep_ip; | |
987 | ||
988 | if (is_vtep_present_in_list(es->vtep_list, rvtep)) | |
989 | return 0; | |
990 | ||
991 | ||
992 | vtep_ip = es_vtep_new(rvtep); | |
993 | if (vtep_ip) | |
994 | listnode_add_sort(es->vtep_list, vtep_ip); | |
995 | return 0; | |
996 | } | |
997 | ||
998 | /* | |
999 | * Best path for ES route was changed, | |
1000 | * update the list of VTEPs for this ES | |
1001 | */ | |
1002 | static int evpn_es_uninstall_vtep(struct bgp *bgp, | |
1003 | struct evpnes *es, | |
1004 | struct prefix_evpn *p, | |
1005 | struct in_addr rvtep) | |
1006 | { | |
1007 | struct listnode *node, *nnode, *node_to_del = NULL; | |
1008 | struct in_addr *tmp; | |
1009 | ||
1010 | for (ALL_LIST_ELEMENTS(es->vtep_list, node, nnode, tmp)) { | |
1011 | if (tmp->s_addr == rvtep.s_addr) { | |
1012 | es_vtep_free(tmp); | |
1013 | node_to_del = node; | |
1014 | } | |
1015 | } | |
1016 | ||
1017 | if (node_to_del) | |
1018 | list_delete_node(es->vtep_list, node_to_del); | |
1019 | ||
1020 | return 0; | |
1021 | } | |
1022 | ||
1023 | /* | |
1024 | * Calculate the best path for a ES(type-4) route. | |
1025 | */ | |
1026 | static int evpn_es_route_select_install(struct bgp *bgp, | |
1027 | struct evpnes *es, | |
1028 | struct bgp_node *rn) | |
1029 | { | |
1030 | int ret = 0; | |
1031 | afi_t afi = AFI_L2VPN; | |
1032 | safi_t safi = SAFI_EVPN; | |
1033 | struct bgp_info *old_select; /* old best */ | |
1034 | struct bgp_info *new_select; /* new best */ | |
1035 | struct bgp_info_pair old_and_new; | |
1036 | ||
1037 | /* Compute the best path. */ | |
1038 | bgp_best_selection(bgp, rn, &bgp->maxpaths[afi][safi], | |
1039 | &old_and_new, afi, safi); | |
1040 | old_select = old_and_new.old; | |
1041 | new_select = old_and_new.new; | |
1042 | ||
1043 | /* | |
1044 | * If the best path hasn't changed - see if something needs to be | |
1045 | * updated | |
1046 | */ | |
1047 | if (old_select && old_select == new_select | |
1048 | && old_select->type == ZEBRA_ROUTE_BGP | |
1049 | && old_select->sub_type == BGP_ROUTE_IMPORTED | |
1050 | && !CHECK_FLAG(rn->flags, BGP_NODE_USER_CLEAR) | |
1051 | && !CHECK_FLAG(old_select->flags, BGP_INFO_ATTR_CHANGED) | |
1052 | && !bgp->addpath_tx_used[afi][safi]) { | |
1053 | if (bgp_zebra_has_route_changed(rn, old_select)) { | |
1054 | ret = evpn_es_install_vtep(bgp, es, | |
1055 | (struct prefix_evpn *)&rn->p, | |
1056 | old_select->attr->nexthop); | |
1057 | } | |
1058 | UNSET_FLAG(old_select->flags, BGP_INFO_MULTIPATH_CHG); | |
1059 | bgp_zebra_clear_route_change_flags(rn); | |
1060 | return ret; | |
1061 | } | |
1062 | ||
1063 | /* If the user did a "clear" this flag will be set */ | |
1064 | UNSET_FLAG(rn->flags, BGP_NODE_USER_CLEAR); | |
1065 | ||
1066 | /* | |
1067 | * bestpath has changed; update relevant fields and install or uninstall | |
1068 | * into the zebra RIB. | |
1069 | */ | |
1070 | if (old_select || new_select) | |
1071 | bgp_bump_version(rn); | |
1072 | ||
1073 | if (old_select) | |
1074 | bgp_info_unset_flag(rn, old_select, BGP_INFO_SELECTED); | |
1075 | if (new_select) { | |
1076 | bgp_info_set_flag(rn, new_select, BGP_INFO_SELECTED); | |
1077 | bgp_info_unset_flag(rn, new_select, BGP_INFO_ATTR_CHANGED); | |
1078 | UNSET_FLAG(new_select->flags, BGP_INFO_MULTIPATH_CHG); | |
1079 | } | |
1080 | ||
1081 | if (new_select && new_select->type == ZEBRA_ROUTE_BGP | |
1082 | && new_select->sub_type == BGP_ROUTE_IMPORTED) { | |
1083 | ret = evpn_es_install_vtep(bgp, es, | |
1084 | (struct prefix_evpn *)&rn->p, | |
1085 | new_select->attr->nexthop); | |
1086 | } else { | |
1087 | if (old_select && old_select->type == ZEBRA_ROUTE_BGP | |
1088 | && old_select->sub_type == BGP_ROUTE_IMPORTED) | |
1089 | ret = evpn_es_uninstall_vtep( | |
1090 | bgp, es, (struct prefix_evpn *)&rn->p, | |
1091 | old_select->attr->nexthop); | |
1092 | } | |
1093 | ||
1094 | /* Clear any route change flags. */ | |
1095 | bgp_zebra_clear_route_change_flags(rn); | |
1096 | ||
1097 | /* Reap old select bgp_info, if it has been removed */ | |
1098 | if (old_select && CHECK_FLAG(old_select->flags, BGP_INFO_REMOVED)) | |
1099 | bgp_info_reap(rn, old_select); | |
1100 | ||
1101 | return ret; | |
1102 | } | |
1103 | ||
1104 | /* | |
1105 | * Calculate the best path for an EVPN route. Install/update best path in zebra, | |
1106 | * if appropriate. | |
1107 | */ | |
1108 | static int evpn_route_select_install(struct bgp *bgp, struct bgpevpn *vpn, | |
1109 | struct bgp_node *rn) | |
1110 | { | |
1111 | struct bgp_info *old_select, *new_select; | |
1112 | struct bgp_info_pair old_and_new; | |
1113 | struct prefix_evpn *evp; | |
1114 | afi_t afi = AFI_L2VPN; | |
1115 | safi_t safi = SAFI_EVPN; | |
1116 | int ret = 0; | |
1117 | uint8_t flags = 0; | |
1118 | ||
1119 | /* Compute the best path. */ | |
1120 | bgp_best_selection(bgp, rn, &bgp->maxpaths[afi][safi], &old_and_new, | |
1121 | afi, safi); | |
1122 | old_select = old_and_new.old; | |
1123 | new_select = old_and_new.new; | |
1124 | ||
1125 | evp = (struct prefix_evpn *)&rn->p; | |
1126 | /* If the best path hasn't changed - see if there is still something to | |
1127 | * update | |
1128 | * to zebra RIB. | |
1129 | */ | |
1130 | if (old_select && old_select == new_select | |
1131 | && old_select->type == ZEBRA_ROUTE_BGP | |
1132 | && old_select->sub_type == BGP_ROUTE_IMPORTED | |
1133 | && !CHECK_FLAG(rn->flags, BGP_NODE_USER_CLEAR) | |
1134 | && !CHECK_FLAG(old_select->flags, BGP_INFO_ATTR_CHANGED) | |
1135 | && !bgp->addpath_tx_used[afi][safi]) { | |
1136 | if (bgp_zebra_has_route_changed(rn, old_select)) { | |
1137 | if (old_select->attr->sticky) | |
1138 | SET_FLAG(flags, ZEBRA_MACIP_TYPE_STICKY); | |
1139 | if (old_select->attr->default_gw) | |
1140 | SET_FLAG(flags, ZEBRA_MACIP_TYPE_GW); | |
1141 | if (is_evpn_prefix_ipaddr_v6(evp) && | |
1142 | old_select->attr->router_flag) | |
1143 | SET_FLAG(flags, ZEBRA_MACIP_TYPE_ROUTER_FLAG); | |
1144 | ||
1145 | ret = evpn_zebra_install( | |
1146 | bgp, vpn, (struct prefix_evpn *)&rn->p, | |
1147 | old_select->attr->nexthop, flags, | |
1148 | mac_mobility_seqnum(old_select->attr)); | |
1149 | } | |
1150 | UNSET_FLAG(old_select->flags, BGP_INFO_MULTIPATH_CHG); | |
1151 | bgp_zebra_clear_route_change_flags(rn); | |
1152 | return ret; | |
1153 | } | |
1154 | ||
1155 | /* If the user did a "clear" this flag will be set */ | |
1156 | UNSET_FLAG(rn->flags, BGP_NODE_USER_CLEAR); | |
1157 | ||
1158 | /* bestpath has changed; update relevant fields and install or uninstall | |
1159 | * into the zebra RIB. | |
1160 | */ | |
1161 | if (old_select || new_select) | |
1162 | bgp_bump_version(rn); | |
1163 | ||
1164 | if (old_select) | |
1165 | bgp_info_unset_flag(rn, old_select, BGP_INFO_SELECTED); | |
1166 | if (new_select) { | |
1167 | bgp_info_set_flag(rn, new_select, BGP_INFO_SELECTED); | |
1168 | bgp_info_unset_flag(rn, new_select, BGP_INFO_ATTR_CHANGED); | |
1169 | UNSET_FLAG(new_select->flags, BGP_INFO_MULTIPATH_CHG); | |
1170 | } | |
1171 | ||
1172 | if (new_select && new_select->type == ZEBRA_ROUTE_BGP | |
1173 | && new_select->sub_type == BGP_ROUTE_IMPORTED) { | |
1174 | flags = 0; | |
1175 | if (new_select->attr->sticky) | |
1176 | SET_FLAG(flags, ZEBRA_MACIP_TYPE_STICKY); | |
1177 | if (new_select->attr->default_gw) | |
1178 | SET_FLAG(flags, ZEBRA_MACIP_TYPE_GW); | |
1179 | if (is_evpn_prefix_ipaddr_v6(evp) && | |
1180 | new_select->attr->router_flag) | |
1181 | SET_FLAG(flags, ZEBRA_MACIP_TYPE_ROUTER_FLAG); | |
1182 | ||
1183 | ret = evpn_zebra_install(bgp, vpn, (struct prefix_evpn *)&rn->p, | |
1184 | new_select->attr->nexthop, flags, | |
1185 | mac_mobility_seqnum(new_select->attr)); | |
1186 | /* If an old best existed and it was a "local" route, the only | |
1187 | * reason | |
1188 | * it would be supplanted is due to MAC mobility procedures. So, | |
1189 | * we | |
1190 | * need to do an implicit delete and withdraw that route from | |
1191 | * peers. | |
1192 | */ | |
1193 | if (old_select && old_select->peer == bgp->peer_self | |
1194 | && old_select->type == ZEBRA_ROUTE_BGP | |
1195 | && old_select->sub_type == BGP_ROUTE_STATIC) | |
1196 | evpn_delete_old_local_route(bgp, vpn, rn, old_select); | |
1197 | } else { | |
1198 | if (old_select && old_select->type == ZEBRA_ROUTE_BGP | |
1199 | && old_select->sub_type == BGP_ROUTE_IMPORTED) | |
1200 | ret = evpn_zebra_uninstall(bgp, vpn, | |
1201 | (struct prefix_evpn *)&rn->p, | |
1202 | old_select->attr->nexthop); | |
1203 | } | |
1204 | ||
1205 | /* Clear any route change flags. */ | |
1206 | bgp_zebra_clear_route_change_flags(rn); | |
1207 | ||
1208 | /* Reap old select bgp_info, if it has been removed */ | |
1209 | if (old_select && CHECK_FLAG(old_select->flags, BGP_INFO_REMOVED)) | |
1210 | bgp_info_reap(rn, old_select); | |
1211 | ||
1212 | return ret; | |
1213 | } | |
1214 | ||
1215 | /* | |
1216 | * Return true if the local ri for this rn is of type gateway mac | |
1217 | */ | |
1218 | static int evpn_route_is_def_gw(struct bgp *bgp, struct bgp_node *rn) | |
1219 | { | |
1220 | struct bgp_info *tmp_ri = NULL; | |
1221 | struct bgp_info *local_ri = NULL; | |
1222 | ||
1223 | local_ri = NULL; | |
1224 | for (tmp_ri = rn->info; tmp_ri; tmp_ri = tmp_ri->next) { | |
1225 | if (tmp_ri->peer == bgp->peer_self | |
1226 | && tmp_ri->type == ZEBRA_ROUTE_BGP | |
1227 | && tmp_ri->sub_type == BGP_ROUTE_STATIC) | |
1228 | local_ri = tmp_ri; | |
1229 | } | |
1230 | ||
1231 | if (!local_ri) | |
1232 | return 0; | |
1233 | ||
1234 | return local_ri->attr->default_gw; | |
1235 | } | |
1236 | ||
1237 | ||
1238 | /* | |
1239 | * Return true if the local ri for this rn has sticky set | |
1240 | */ | |
1241 | static int evpn_route_is_sticky(struct bgp *bgp, struct bgp_node *rn) | |
1242 | { | |
1243 | struct bgp_info *tmp_ri; | |
1244 | struct bgp_info *local_ri; | |
1245 | ||
1246 | local_ri = NULL; | |
1247 | for (tmp_ri = rn->info; tmp_ri; tmp_ri = tmp_ri->next) { | |
1248 | if (tmp_ri->peer == bgp->peer_self | |
1249 | && tmp_ri->type == ZEBRA_ROUTE_BGP | |
1250 | && tmp_ri->sub_type == BGP_ROUTE_STATIC) | |
1251 | local_ri = tmp_ri; | |
1252 | } | |
1253 | ||
1254 | if (!local_ri) | |
1255 | return 0; | |
1256 | ||
1257 | return local_ri->attr->sticky; | |
1258 | } | |
1259 | ||
1260 | /* | |
1261 | * create or update EVPN type4 route entry. | |
1262 | * This could be in the ES table or the global table. | |
1263 | * TODO: handle remote ES (type4) routes as well | |
1264 | */ | |
1265 | static int update_evpn_type4_route_entry(struct bgp *bgp, | |
1266 | struct evpnes *es, | |
1267 | afi_t afi, safi_t safi, | |
1268 | struct bgp_node *rn, | |
1269 | struct attr *attr, | |
1270 | int add, | |
1271 | struct bgp_info **ri, | |
1272 | int *route_changed) | |
1273 | { | |
1274 | char buf[ESI_STR_LEN]; | |
1275 | char buf1[INET6_ADDRSTRLEN]; | |
1276 | struct bgp_info *tmp_ri = NULL; | |
1277 | struct bgp_info *local_ri = NULL; /* local route entry if any */ | |
1278 | struct bgp_info *remote_ri = NULL; /* remote route entry if any */ | |
1279 | struct attr *attr_new = NULL; | |
1280 | struct prefix_evpn *evp = NULL; | |
1281 | ||
1282 | *ri = NULL; | |
1283 | *route_changed = 1; | |
1284 | evp = (struct prefix_evpn *)&rn->p; | |
1285 | ||
1286 | /* locate the local and remote entries if any */ | |
1287 | for (tmp_ri = rn->info; tmp_ri; tmp_ri = tmp_ri->next) { | |
1288 | if (tmp_ri->peer == bgp->peer_self && | |
1289 | tmp_ri->type == ZEBRA_ROUTE_BGP && | |
1290 | tmp_ri->sub_type == BGP_ROUTE_STATIC) | |
1291 | local_ri = tmp_ri; | |
1292 | if (tmp_ri->type == ZEBRA_ROUTE_BGP && | |
1293 | tmp_ri->sub_type == BGP_ROUTE_IMPORTED && | |
1294 | CHECK_FLAG(tmp_ri->flags, BGP_INFO_VALID)) | |
1295 | remote_ri = tmp_ri; | |
1296 | } | |
1297 | ||
1298 | /* we don't expect to see a remote_ri at this point. | |
1299 | * An ES route has esi + vtep_ip as the key, | |
1300 | * We shouldn't see the same route from any other vtep. | |
1301 | */ | |
1302 | if (remote_ri) { | |
1303 | flog_err( | |
1304 | EC_BGP_ES_INVALID, | |
1305 | "%u ERROR: local es route for ESI: %s Vtep %s also learnt from remote", | |
1306 | bgp->vrf_id, | |
1307 | esi_to_str(&evp->prefix.es_addr.esi, buf, sizeof(buf)), | |
1308 | ipaddr2str(&es->originator_ip, buf1, sizeof(buf1))); | |
1309 | return -1; | |
1310 | } | |
1311 | ||
1312 | if (!local_ri && !add) | |
1313 | return 0; | |
1314 | ||
1315 | /* create or update the entry */ | |
1316 | if (!local_ri) { | |
1317 | ||
1318 | /* Add or update attribute to hash */ | |
1319 | attr_new = bgp_attr_intern(attr); | |
1320 | ||
1321 | /* Create new route with its attribute. */ | |
1322 | tmp_ri = info_make(ZEBRA_ROUTE_BGP, BGP_ROUTE_STATIC, | |
1323 | 0, bgp->peer_self, attr_new, rn); | |
1324 | SET_FLAG(tmp_ri->flags, BGP_INFO_VALID); | |
1325 | ||
1326 | /* add the newly created path to the route-node */ | |
1327 | bgp_info_add(rn, tmp_ri); | |
1328 | } else { | |
1329 | tmp_ri = local_ri; | |
1330 | if (attrhash_cmp(tmp_ri->attr, attr) | |
1331 | && !CHECK_FLAG(tmp_ri->flags, BGP_INFO_REMOVED)) | |
1332 | *route_changed = 0; | |
1333 | else { | |
1334 | /* The attribute has changed. | |
1335 | * Add (or update) attribute to hash. */ | |
1336 | attr_new = bgp_attr_intern(attr); | |
1337 | bgp_info_set_flag(rn, tmp_ri, BGP_INFO_ATTR_CHANGED); | |
1338 | ||
1339 | /* Restore route, if needed. */ | |
1340 | if (CHECK_FLAG(tmp_ri->flags, BGP_INFO_REMOVED)) | |
1341 | bgp_info_restore(rn, tmp_ri); | |
1342 | ||
1343 | /* Unintern existing, set to new. */ | |
1344 | bgp_attr_unintern(&tmp_ri->attr); | |
1345 | tmp_ri->attr = attr_new; | |
1346 | tmp_ri->uptime = bgp_clock(); | |
1347 | } | |
1348 | } | |
1349 | ||
1350 | /* Return back the route entry. */ | |
1351 | *ri = tmp_ri; | |
1352 | return 0; | |
1353 | } | |
1354 | ||
1355 | /* update evpn es (type-4) route */ | |
1356 | static int update_evpn_type4_route(struct bgp *bgp, | |
1357 | struct evpnes *es, | |
1358 | struct prefix_evpn *p) | |
1359 | { | |
1360 | int ret = 0; | |
1361 | int route_changed = 0; | |
1362 | char buf[ESI_STR_LEN]; | |
1363 | char buf1[INET6_ADDRSTRLEN]; | |
1364 | afi_t afi = AFI_L2VPN; | |
1365 | safi_t safi = SAFI_EVPN; | |
1366 | struct attr attr; | |
1367 | struct attr *attr_new = NULL; | |
1368 | struct bgp_node *rn = NULL; | |
1369 | struct bgp_info *ri = NULL; | |
1370 | ||
1371 | memset(&attr, 0, sizeof(struct attr)); | |
1372 | ||
1373 | /* Build path-attribute for this route. */ | |
1374 | bgp_attr_default_set(&attr, BGP_ORIGIN_IGP); | |
1375 | attr.nexthop = es->originator_ip.ipaddr_v4; | |
1376 | attr.mp_nexthop_global_in = es->originator_ip.ipaddr_v4; | |
1377 | attr.mp_nexthop_len = BGP_ATTR_NHLEN_IPV4; | |
1378 | ||
1379 | /* Set up extended community. */ | |
1380 | build_evpn_type4_route_extcomm(es, &attr); | |
1381 | ||
1382 | /* First, create (or fetch) route node within the ESI. */ | |
1383 | /* NOTE: There is no RD here. */ | |
1384 | rn = bgp_node_get(es->route_table, (struct prefix *)p); | |
1385 | ||
1386 | /* Create or update route entry. */ | |
1387 | ret = update_evpn_type4_route_entry(bgp, es, afi, safi, rn, | |
1388 | &attr, 1, &ri, | |
1389 | &route_changed); | |
1390 | if (ret != 0) { | |
1391 | flog_err(EC_BGP_ES_INVALID, | |
1392 | "%u ERROR: Failed to updated ES route ESI: %s VTEP %s", | |
1393 | bgp->vrf_id, | |
1394 | esi_to_str(&p->prefix.es_addr.esi, buf, sizeof(buf)), | |
1395 | ipaddr2str(&es->originator_ip, buf1, sizeof(buf1))); | |
1396 | } | |
1397 | ||
1398 | assert(ri); | |
1399 | attr_new = ri->attr; | |
1400 | ||
1401 | /* Perform route selection; | |
1402 | * this is just to set the flags correctly | |
1403 | * as local route in the ES always wins. | |
1404 | */ | |
1405 | evpn_es_route_select_install(bgp, es, rn); | |
1406 | bgp_unlock_node(rn); | |
1407 | ||
1408 | /* If this is a new route or some attribute has changed, export the | |
1409 | * route to the global table. The route will be advertised to peers | |
1410 | * from there. Note that this table is a 2-level tree (RD-level + | |
1411 | * Prefix-level) similar to L3VPN routes. | |
1412 | */ | |
1413 | if (route_changed) { | |
1414 | struct bgp_info *global_ri; | |
1415 | ||
1416 | rn = bgp_afi_node_get(bgp->rib[afi][safi], afi, safi, | |
1417 | (struct prefix *)p, &es->prd); | |
1418 | update_evpn_type4_route_entry(bgp, es, afi, safi, | |
1419 | rn, attr_new, | |
1420 | 1, &global_ri, | |
1421 | &route_changed); | |
1422 | ||
1423 | /* Schedule for processing and unlock node. */ | |
1424 | bgp_process(bgp, rn, afi, safi); | |
1425 | bgp_unlock_node(rn); | |
1426 | } | |
1427 | ||
1428 | /* Unintern temporary. */ | |
1429 | aspath_unintern(&attr.aspath); | |
1430 | return 0; | |
1431 | } | |
1432 | ||
1433 | static int update_evpn_type5_route_entry(struct bgp *bgp_def, | |
1434 | struct bgp *bgp_vrf, afi_t afi, | |
1435 | safi_t safi, struct bgp_node *rn, | |
1436 | struct attr *attr, int *route_changed) | |
1437 | { | |
1438 | struct attr *attr_new = NULL; | |
1439 | struct bgp_info *ri = NULL; | |
1440 | mpls_label_t label = MPLS_INVALID_LABEL; | |
1441 | struct bgp_info *local_ri = NULL; | |
1442 | struct bgp_info *tmp_ri = NULL; | |
1443 | ||
1444 | *route_changed = 0; | |
1445 | /* locate the local route entry if any */ | |
1446 | for (tmp_ri = rn->info; tmp_ri; tmp_ri = tmp_ri->next) { | |
1447 | if (tmp_ri->peer == bgp_def->peer_self | |
1448 | && tmp_ri->type == ZEBRA_ROUTE_BGP | |
1449 | && tmp_ri->sub_type == BGP_ROUTE_STATIC) | |
1450 | local_ri = tmp_ri; | |
1451 | } | |
1452 | ||
1453 | /* | |
1454 | * create a new route entry if one doesnt exist. | |
1455 | * Otherwise see if route attr has changed | |
1456 | */ | |
1457 | if (!local_ri) { | |
1458 | ||
1459 | /* route has changed as this is the first entry */ | |
1460 | *route_changed = 1; | |
1461 | ||
1462 | /* Add (or update) attribute to hash. */ | |
1463 | attr_new = bgp_attr_intern(attr); | |
1464 | ||
1465 | /* create the route info from attribute */ | |
1466 | ri = info_make(ZEBRA_ROUTE_BGP, BGP_ROUTE_STATIC, 0, | |
1467 | bgp_def->peer_self, attr_new, rn); | |
1468 | SET_FLAG(ri->flags, BGP_INFO_VALID); | |
1469 | ||
1470 | /* Type-5 routes advertise the L3-VNI */ | |
1471 | bgp_info_extra_get(ri); | |
1472 | vni2label(bgp_vrf->l3vni, &label); | |
1473 | memcpy(&ri->extra->label, &label, sizeof(label)); | |
1474 | ri->extra->num_labels = 1; | |
1475 | ||
1476 | /* add the route entry to route node*/ | |
1477 | bgp_info_add(rn, ri); | |
1478 | } else { | |
1479 | ||
1480 | tmp_ri = local_ri; | |
1481 | if (!attrhash_cmp(tmp_ri->attr, attr)) { | |
1482 | ||
1483 | /* attribute changed */ | |
1484 | *route_changed = 1; | |
1485 | ||
1486 | /* The attribute has changed. */ | |
1487 | /* Add (or update) attribute to hash. */ | |
1488 | attr_new = bgp_attr_intern(attr); | |
1489 | bgp_info_set_flag(rn, tmp_ri, BGP_INFO_ATTR_CHANGED); | |
1490 | ||
1491 | /* Restore route, if needed. */ | |
1492 | if (CHECK_FLAG(tmp_ri->flags, BGP_INFO_REMOVED)) | |
1493 | bgp_info_restore(rn, tmp_ri); | |
1494 | ||
1495 | /* Unintern existing, set to new. */ | |
1496 | bgp_attr_unintern(&tmp_ri->attr); | |
1497 | tmp_ri->attr = attr_new; | |
1498 | tmp_ri->uptime = bgp_clock(); | |
1499 | } | |
1500 | } | |
1501 | return 0; | |
1502 | } | |
1503 | ||
1504 | /* update evpn type-5 route entry */ | |
1505 | static int update_evpn_type5_route(struct bgp *bgp_vrf, struct prefix_evpn *evp, | |
1506 | struct attr *src_attr) | |
1507 | { | |
1508 | afi_t afi = AFI_L2VPN; | |
1509 | safi_t safi = SAFI_EVPN; | |
1510 | struct attr attr; | |
1511 | struct bgp_node *rn = NULL; | |
1512 | struct bgp *bgp_def = NULL; | |
1513 | int route_changed = 0; | |
1514 | ||
1515 | bgp_def = bgp_get_default(); | |
1516 | if (!bgp_def) | |
1517 | return 0; | |
1518 | ||
1519 | /* Build path attribute for this route - use the source attr, if | |
1520 | * present, else treat as locally originated. | |
1521 | */ | |
1522 | if (src_attr) | |
1523 | bgp_attr_dup(&attr, src_attr); | |
1524 | else { | |
1525 | memset(&attr, 0, sizeof(struct attr)); | |
1526 | bgp_attr_default_set(&attr, BGP_ORIGIN_IGP); | |
1527 | } | |
1528 | /* Set nexthop to ourselves and fill in the Router MAC. */ | |
1529 | attr.nexthop = bgp_vrf->originator_ip; | |
1530 | attr.mp_nexthop_global_in = bgp_vrf->originator_ip; | |
1531 | attr.mp_nexthop_len = BGP_ATTR_NHLEN_IPV4; | |
1532 | memcpy(&attr.rmac, &bgp_vrf->rmac, sizeof(struct ethaddr)); | |
1533 | ||
1534 | /* Setup RT and encap extended community */ | |
1535 | build_evpn_type5_route_extcomm(bgp_vrf, &attr); | |
1536 | ||
1537 | /* get the route node in global table */ | |
1538 | rn = bgp_afi_node_get(bgp_def->rib[afi][safi], afi, safi, | |
1539 | (struct prefix *)evp, &bgp_vrf->vrf_prd); | |
1540 | assert(rn); | |
1541 | ||
1542 | /* create or update the route entry within the route node */ | |
1543 | update_evpn_type5_route_entry(bgp_def, bgp_vrf, afi, safi, rn, &attr, | |
1544 | &route_changed); | |
1545 | ||
1546 | /* schedule for processing and unlock node */ | |
1547 | if (route_changed) { | |
1548 | bgp_process(bgp_def, rn, afi, safi); | |
1549 | bgp_unlock_node(rn); | |
1550 | } | |
1551 | ||
1552 | /* uninten temporary */ | |
1553 | if (!src_attr) | |
1554 | aspath_unintern(&attr.aspath); | |
1555 | return 0; | |
1556 | } | |
1557 | ||
1558 | /* | |
1559 | * Create or update EVPN route entry. This could be in the VNI route table | |
1560 | * or the global route table. | |
1561 | */ | |
1562 | static int update_evpn_route_entry(struct bgp *bgp, struct bgpevpn *vpn, | |
1563 | afi_t afi, safi_t safi, struct bgp_node *rn, | |
1564 | struct attr *attr, int add, | |
1565 | struct bgp_info **ri, uint8_t flags, | |
1566 | uint32_t seq) | |
1567 | { | |
1568 | struct bgp_info *tmp_ri; | |
1569 | struct bgp_info *local_ri; | |
1570 | struct attr *attr_new; | |
1571 | mpls_label_t label[BGP_MAX_LABELS]; | |
1572 | uint32_t num_labels = 1; | |
1573 | int route_change = 1; | |
1574 | uint8_t sticky = 0; | |
1575 | struct prefix_evpn *evp; | |
1576 | ||
1577 | *ri = NULL; | |
1578 | evp = (struct prefix_evpn *)&rn->p; | |
1579 | memset(&label, 0, sizeof(label)); | |
1580 | ||
1581 | /* See if this is an update of an existing route, or a new add. */ | |
1582 | local_ri = NULL; | |
1583 | for (tmp_ri = rn->info; tmp_ri; tmp_ri = tmp_ri->next) { | |
1584 | if (tmp_ri->peer == bgp->peer_self | |
1585 | && tmp_ri->type == ZEBRA_ROUTE_BGP | |
1586 | && tmp_ri->sub_type == BGP_ROUTE_STATIC) | |
1587 | local_ri = tmp_ri; | |
1588 | } | |
1589 | ||
1590 | /* If route doesn't exist already, create a new one, if told to. | |
1591 | * Otherwise act based on whether the attributes of the route have | |
1592 | * changed or not. | |
1593 | */ | |
1594 | if (!local_ri && !add) | |
1595 | return 0; | |
1596 | ||
1597 | /* For non-GW MACs, update MAC mobility seq number, if needed. */ | |
1598 | if (seq && !CHECK_FLAG(flags, ZEBRA_MACIP_TYPE_GW)) | |
1599 | add_mac_mobility_to_attr(seq, attr); | |
1600 | ||
1601 | if (!local_ri) { | |
1602 | /* Add (or update) attribute to hash. */ | |
1603 | attr_new = bgp_attr_intern(attr); | |
1604 | ||
1605 | /* Extract MAC mobility sequence number, if any. */ | |
1606 | attr_new->mm_seqnum = | |
1607 | bgp_attr_mac_mobility_seqnum(attr_new, &sticky); | |
1608 | attr_new->sticky = sticky; | |
1609 | ||
1610 | /* Create new route with its attribute. */ | |
1611 | tmp_ri = info_make(ZEBRA_ROUTE_BGP, BGP_ROUTE_STATIC, 0, | |
1612 | bgp->peer_self, attr_new, rn); | |
1613 | SET_FLAG(tmp_ri->flags, BGP_INFO_VALID); | |
1614 | bgp_info_extra_get(tmp_ri); | |
1615 | ||
1616 | /* The VNI goes into the 'label' field of the route */ | |
1617 | vni2label(vpn->vni, &label[0]); | |
1618 | ||
1619 | /* Type-2 routes may carry a second VNI - the L3-VNI. | |
1620 | * Only attach second label if we are advertising two labels for | |
1621 | * type-2 routes. | |
1622 | */ | |
1623 | if (evp->prefix.route_type == BGP_EVPN_MAC_IP_ROUTE | |
1624 | && CHECK_FLAG(vpn->flags, VNI_FLAG_USE_TWO_LABELS)) { | |
1625 | vni_t l3vni; | |
1626 | ||
1627 | l3vni = bgpevpn_get_l3vni(vpn); | |
1628 | if (l3vni) { | |
1629 | vni2label(l3vni, &label[1]); | |
1630 | num_labels++; | |
1631 | } | |
1632 | } | |
1633 | ||
1634 | memcpy(&tmp_ri->extra->label, label, sizeof(label)); | |
1635 | tmp_ri->extra->num_labels = num_labels; | |
1636 | bgp_info_add(rn, tmp_ri); | |
1637 | } else { | |
1638 | tmp_ri = local_ri; | |
1639 | if (attrhash_cmp(tmp_ri->attr, attr) | |
1640 | && !CHECK_FLAG(tmp_ri->flags, BGP_INFO_REMOVED)) | |
1641 | route_change = 0; | |
1642 | else { | |
1643 | /* | |
1644 | * The attributes have changed, type-2 routes needs to | |
1645 | * be advertised with right labels. | |
1646 | */ | |
1647 | vni2label(vpn->vni, &label[0]); | |
1648 | if (evp->prefix.route_type == BGP_EVPN_MAC_IP_ROUTE | |
1649 | && CHECK_FLAG(vpn->flags, | |
1650 | VNI_FLAG_USE_TWO_LABELS)) { | |
1651 | vni_t l3vni; | |
1652 | ||
1653 | l3vni = bgpevpn_get_l3vni(vpn); | |
1654 | if (l3vni) { | |
1655 | vni2label(l3vni, &label[1]); | |
1656 | num_labels++; | |
1657 | } | |
1658 | } | |
1659 | memcpy(&tmp_ri->extra->label, label, sizeof(label)); | |
1660 | tmp_ri->extra->num_labels = num_labels; | |
1661 | ||
1662 | /* The attribute has changed. */ | |
1663 | /* Add (or update) attribute to hash. */ | |
1664 | attr_new = bgp_attr_intern(attr); | |
1665 | bgp_info_set_flag(rn, tmp_ri, BGP_INFO_ATTR_CHANGED); | |
1666 | ||
1667 | /* Extract MAC mobility sequence number, if any. */ | |
1668 | attr_new->mm_seqnum = | |
1669 | bgp_attr_mac_mobility_seqnum(attr_new, &sticky); | |
1670 | attr_new->sticky = sticky; | |
1671 | ||
1672 | /* Restore route, if needed. */ | |
1673 | if (CHECK_FLAG(tmp_ri->flags, BGP_INFO_REMOVED)) | |
1674 | bgp_info_restore(rn, tmp_ri); | |
1675 | ||
1676 | /* Unintern existing, set to new. */ | |
1677 | bgp_attr_unintern(&tmp_ri->attr); | |
1678 | tmp_ri->attr = attr_new; | |
1679 | tmp_ri->uptime = bgp_clock(); | |
1680 | } | |
1681 | } | |
1682 | ||
1683 | /* Return back the route entry. */ | |
1684 | *ri = tmp_ri; | |
1685 | return route_change; | |
1686 | } | |
1687 | ||
1688 | /* | |
1689 | * Create or update EVPN route (of type based on prefix) for specified VNI | |
1690 | * and schedule for processing. | |
1691 | */ | |
1692 | static int update_evpn_route(struct bgp *bgp, struct bgpevpn *vpn, | |
1693 | struct prefix_evpn *p, uint8_t flags, | |
1694 | uint32_t seq) | |
1695 | { | |
1696 | struct bgp_node *rn; | |
1697 | struct attr attr; | |
1698 | struct attr *attr_new; | |
1699 | int add_l3_ecomm = 0; | |
1700 | struct bgp_info *ri; | |
1701 | afi_t afi = AFI_L2VPN; | |
1702 | safi_t safi = SAFI_EVPN; | |
1703 | int route_change; | |
1704 | ||
1705 | memset(&attr, 0, sizeof(struct attr)); | |
1706 | ||
1707 | /* Build path-attribute for this route. */ | |
1708 | bgp_attr_default_set(&attr, BGP_ORIGIN_IGP); | |
1709 | attr.nexthop = vpn->originator_ip; | |
1710 | attr.mp_nexthop_global_in = vpn->originator_ip; | |
1711 | attr.mp_nexthop_len = BGP_ATTR_NHLEN_IPV4; | |
1712 | attr.sticky = CHECK_FLAG(flags, ZEBRA_MACIP_TYPE_STICKY) ? 1 : 0; | |
1713 | attr.default_gw = CHECK_FLAG(flags, ZEBRA_MACIP_TYPE_GW) ? 1 : 0; | |
1714 | attr.router_flag = CHECK_FLAG(flags, | |
1715 | ZEBRA_MACIP_TYPE_ROUTER_FLAG) ? 1 : 0; | |
1716 | ||
1717 | /* PMSI is only needed for type-3 routes */ | |
1718 | if (p->prefix.route_type == BGP_EVPN_IMET_ROUTE) | |
1719 | attr.flag |= ATTR_FLAG_BIT(BGP_ATTR_PMSI_TUNNEL); | |
1720 | ||
1721 | /* router mac is only needed for type-2 routes here. */ | |
1722 | if (p->prefix.route_type == BGP_EVPN_MAC_IP_ROUTE) | |
1723 | bgpevpn_get_rmac(vpn, &attr.rmac); | |
1724 | vni2label(vpn->vni, &(attr.label)); | |
1725 | ||
1726 | /* Include L3 VNI related RTs and RMAC for type-2 routes, if they're | |
1727 | * IPv4 or IPv6 global addresses and we're advertising L3VNI with | |
1728 | * these routes. | |
1729 | */ | |
1730 | if (p->prefix.route_type == BGP_EVPN_MAC_IP_ROUTE && | |
1731 | (is_evpn_prefix_ipaddr_v4(p) || | |
1732 | !IN6_IS_ADDR_LINKLOCAL(&p->prefix.macip_addr.ip.ipaddr_v6)) && | |
1733 | CHECK_FLAG(vpn->flags, VNI_FLAG_USE_TWO_LABELS) && | |
1734 | bgpevpn_get_l3vni(vpn)) | |
1735 | add_l3_ecomm = 1; | |
1736 | ||
1737 | /* Set up extended community. */ | |
1738 | build_evpn_route_extcomm(vpn, &attr, add_l3_ecomm); | |
1739 | ||
1740 | /* First, create (or fetch) route node within the VNI. */ | |
1741 | /* NOTE: There is no RD here. */ | |
1742 | rn = bgp_node_get(vpn->route_table, (struct prefix *)p); | |
1743 | ||
1744 | /* Create or update route entry. */ | |
1745 | route_change = update_evpn_route_entry(bgp, vpn, afi, safi, rn, &attr, | |
1746 | 1, &ri, flags, seq); | |
1747 | assert(ri); | |
1748 | attr_new = ri->attr; | |
1749 | ||
1750 | /* Perform route selection; this is just to set the flags correctly | |
1751 | * as local route in the VNI always wins. | |
1752 | */ | |
1753 | evpn_route_select_install(bgp, vpn, rn); | |
1754 | bgp_unlock_node(rn); | |
1755 | ||
1756 | /* If this is a new route or some attribute has changed, export the | |
1757 | * route to the global table. The route will be advertised to peers | |
1758 | * from there. Note that this table is a 2-level tree (RD-level + | |
1759 | * Prefix-level) similar to L3VPN routes. | |
1760 | */ | |
1761 | if (route_change) { | |
1762 | struct bgp_info *global_ri; | |
1763 | ||
1764 | rn = bgp_afi_node_get(bgp->rib[afi][safi], afi, safi, | |
1765 | (struct prefix *)p, &vpn->prd); | |
1766 | update_evpn_route_entry(bgp, vpn, afi, safi, rn, attr_new, 1, | |
1767 | &global_ri, flags, seq); | |
1768 | ||
1769 | /* Schedule for processing and unlock node. */ | |
1770 | bgp_process(bgp, rn, afi, safi); | |
1771 | bgp_unlock_node(rn); | |
1772 | } | |
1773 | ||
1774 | /* Unintern temporary. */ | |
1775 | aspath_unintern(&attr.aspath); | |
1776 | ||
1777 | return 0; | |
1778 | } | |
1779 | ||
1780 | /* | |
1781 | * Delete EVPN route entry. | |
1782 | * The entry can be in ESI/VNI table or the global table. | |
1783 | */ | |
1784 | static void delete_evpn_route_entry(struct bgp *bgp, | |
1785 | afi_t afi, safi_t safi, | |
1786 | struct bgp_node *rn, | |
1787 | struct bgp_info **ri) | |
1788 | { | |
1789 | struct bgp_info *tmp_ri; | |
1790 | ||
1791 | *ri = NULL; | |
1792 | ||
1793 | /* Now, find matching route. */ | |
1794 | for (tmp_ri = rn->info; tmp_ri; tmp_ri = tmp_ri->next) | |
1795 | if (tmp_ri->peer == bgp->peer_self | |
1796 | && tmp_ri->type == ZEBRA_ROUTE_BGP | |
1797 | && tmp_ri->sub_type == BGP_ROUTE_STATIC) | |
1798 | break; | |
1799 | ||
1800 | *ri = tmp_ri; | |
1801 | ||
1802 | /* Mark route for delete. */ | |
1803 | if (tmp_ri) | |
1804 | bgp_info_delete(rn, tmp_ri); | |
1805 | } | |
1806 | ||
1807 | ||
1808 | ||
1809 | /* Delete EVPN ES (type-4) route */ | |
1810 | static int delete_evpn_type4_route(struct bgp *bgp, | |
1811 | struct evpnes *es, | |
1812 | struct prefix_evpn *p) | |
1813 | { | |
1814 | afi_t afi = AFI_L2VPN; | |
1815 | safi_t safi = SAFI_EVPN; | |
1816 | struct bgp_info *ri; | |
1817 | struct bgp_node *rn = NULL; /* rn in esi table */ | |
1818 | struct bgp_node *global_rn = NULL; /* rn in global table */ | |
1819 | ||
1820 | /* First, locate the route node within the ESI. | |
1821 | * If it doesn't exist, ther is nothing to do. | |
1822 | * Note: there is no RD here. | |
1823 | */ | |
1824 | rn = bgp_node_lookup(es->route_table, (struct prefix *)p); | |
1825 | if (!rn) | |
1826 | return 0; | |
1827 | ||
1828 | /* Next, locate route node in the global EVPN routing table. | |
1829 | * Note that this table is a 2-level tree (RD-level + Prefix-level) | |
1830 | */ | |
1831 | global_rn = bgp_afi_node_lookup(bgp->rib[afi][safi], afi, safi, | |
1832 | (struct prefix *)p, &es->prd); | |
1833 | if (global_rn) { | |
1834 | ||
1835 | /* Delete route entry in the global EVPN table. */ | |
1836 | delete_evpn_route_entry(bgp, afi, safi, | |
1837 | global_rn, &ri); | |
1838 | ||
1839 | /* Schedule for processing - withdraws to peers happen from | |
1840 | * this table. | |
1841 | */ | |
1842 | if (ri) | |
1843 | bgp_process(bgp, global_rn, afi, safi); | |
1844 | bgp_unlock_node(global_rn); | |
1845 | } | |
1846 | ||
1847 | /* | |
1848 | * Delete route entry in the ESI route table. | |
1849 | * This can just be removed. | |
1850 | */ | |
1851 | delete_evpn_route_entry(bgp, afi, safi, rn, &ri); | |
1852 | if (ri) | |
1853 | bgp_info_reap(rn, ri); | |
1854 | bgp_unlock_node(rn); | |
1855 | return 0; | |
1856 | } | |
1857 | ||
1858 | /* Delete EVPN type5 route */ | |
1859 | static int delete_evpn_type5_route(struct bgp *bgp_vrf, struct prefix_evpn *evp) | |
1860 | { | |
1861 | afi_t afi = AFI_L2VPN; | |
1862 | safi_t safi = SAFI_EVPN; | |
1863 | struct bgp_node *rn = NULL; | |
1864 | struct bgp_info *ri = NULL; | |
1865 | struct bgp *bgp_def = NULL; /* default bgp instance */ | |
1866 | ||
1867 | bgp_def = bgp_get_default(); | |
1868 | if (!bgp_def) | |
1869 | return 0; | |
1870 | ||
1871 | /* locate the global route entry for this type-5 prefix */ | |
1872 | rn = bgp_afi_node_lookup(bgp_def->rib[afi][safi], afi, safi, | |
1873 | (struct prefix *)evp, &bgp_vrf->vrf_prd); | |
1874 | if (!rn) | |
1875 | return 0; | |
1876 | ||
1877 | delete_evpn_route_entry(bgp_def, afi, safi, rn, &ri); | |
1878 | if (ri) | |
1879 | bgp_process(bgp_def, rn, afi, safi); | |
1880 | bgp_unlock_node(rn); | |
1881 | return 0; | |
1882 | } | |
1883 | ||
1884 | /* | |
1885 | * Delete EVPN route (of type based on prefix) for specified VNI and | |
1886 | * schedule for processing. | |
1887 | */ | |
1888 | static int delete_evpn_route(struct bgp *bgp, struct bgpevpn *vpn, | |
1889 | struct prefix_evpn *p) | |
1890 | { | |
1891 | struct bgp_node *rn, *global_rn; | |
1892 | struct bgp_info *ri; | |
1893 | afi_t afi = AFI_L2VPN; | |
1894 | safi_t safi = SAFI_EVPN; | |
1895 | ||
1896 | /* First, locate the route node within the VNI. If it doesn't exist, | |
1897 | * there | |
1898 | * is nothing further to do. | |
1899 | */ | |
1900 | /* NOTE: There is no RD here. */ | |
1901 | rn = bgp_node_lookup(vpn->route_table, (struct prefix *)p); | |
1902 | if (!rn) | |
1903 | return 0; | |
1904 | ||
1905 | /* Next, locate route node in the global EVPN routing table. Note that | |
1906 | * this table is a 2-level tree (RD-level + Prefix-level) similar to | |
1907 | * L3VPN routes. | |
1908 | */ | |
1909 | global_rn = bgp_afi_node_lookup(bgp->rib[afi][safi], afi, safi, | |
1910 | (struct prefix *)p, &vpn->prd); | |
1911 | if (global_rn) { | |
1912 | /* Delete route entry in the global EVPN table. */ | |
1913 | delete_evpn_route_entry(bgp, afi, safi, global_rn, &ri); | |
1914 | ||
1915 | /* Schedule for processing - withdraws to peers happen from | |
1916 | * this table. | |
1917 | */ | |
1918 | if (ri) | |
1919 | bgp_process(bgp, global_rn, afi, safi); | |
1920 | bgp_unlock_node(global_rn); | |
1921 | } | |
1922 | ||
1923 | /* Delete route entry in the VNI route table. This can just be removed. | |
1924 | */ | |
1925 | delete_evpn_route_entry(bgp, afi, safi, rn, &ri); | |
1926 | if (ri) | |
1927 | bgp_info_reap(rn, ri); | |
1928 | bgp_unlock_node(rn); | |
1929 | ||
1930 | return 0; | |
1931 | } | |
1932 | ||
1933 | /* | |
1934 | * Update all type-2 (MACIP) local routes for this VNI - these should also | |
1935 | * be scheduled for advertise to peers. | |
1936 | */ | |
1937 | static int update_all_type2_routes(struct bgp *bgp, struct bgpevpn *vpn) | |
1938 | { | |
1939 | afi_t afi; | |
1940 | safi_t safi; | |
1941 | struct bgp_node *rn; | |
1942 | struct bgp_info *ri, *tmp_ri; | |
1943 | struct attr attr; | |
1944 | struct attr *attr_new; | |
1945 | uint32_t seq; | |
1946 | int add_l3_ecomm = 0; | |
1947 | ||
1948 | afi = AFI_L2VPN; | |
1949 | safi = SAFI_EVPN; | |
1950 | ||
1951 | /* Walk this VNI's route table and update local type-2 routes. For any | |
1952 | * routes updated, update corresponding entry in the global table too. | |
1953 | */ | |
1954 | for (rn = bgp_table_top(vpn->route_table); rn; | |
1955 | rn = bgp_route_next(rn)) { | |
1956 | struct prefix_evpn *evp = (struct prefix_evpn *)&rn->p; | |
1957 | struct bgp_node *rd_rn; | |
1958 | struct bgp_info *global_ri; | |
1959 | ||
1960 | if (evp->prefix.route_type != BGP_EVPN_MAC_IP_ROUTE) | |
1961 | continue; | |
1962 | ||
1963 | /* Identify local route. */ | |
1964 | for (tmp_ri = rn->info; tmp_ri; tmp_ri = tmp_ri->next) { | |
1965 | if (tmp_ri->peer == bgp->peer_self | |
1966 | && tmp_ri->type == ZEBRA_ROUTE_BGP | |
1967 | && tmp_ri->sub_type == BGP_ROUTE_STATIC) | |
1968 | break; | |
1969 | } | |
1970 | ||
1971 | if (!tmp_ri) | |
1972 | continue; | |
1973 | ||
1974 | /* | |
1975 | * Build attribute per local route as the MAC mobility and | |
1976 | * some other values could differ for different routes. The | |
1977 | * attributes will be shared in the hash table. | |
1978 | */ | |
1979 | bgp_attr_default_set(&attr, BGP_ORIGIN_IGP); | |
1980 | attr.nexthop = vpn->originator_ip; | |
1981 | attr.mp_nexthop_global_in = vpn->originator_ip; | |
1982 | attr.mp_nexthop_len = BGP_ATTR_NHLEN_IPV4; | |
1983 | bgpevpn_get_rmac(vpn, &attr.rmac); | |
1984 | ||
1985 | if (evpn_route_is_sticky(bgp, rn)) | |
1986 | attr.sticky = 1; | |
1987 | else if (evpn_route_is_def_gw(bgp, rn)) { | |
1988 | attr.default_gw = 1; | |
1989 | if (is_evpn_prefix_ipaddr_v6(evp)) | |
1990 | attr.router_flag = 1; | |
1991 | } | |
1992 | ||
1993 | /* Add L3 VNI RTs and RMAC for non IPv6 link-local if | |
1994 | * using L3 VNI for type-2 routes also. | |
1995 | */ | |
1996 | if ((is_evpn_prefix_ipaddr_v4(evp) || | |
1997 | !IN6_IS_ADDR_LINKLOCAL( | |
1998 | &evp->prefix.macip_addr.ip.ipaddr_v6)) && | |
1999 | CHECK_FLAG(vpn->flags, VNI_FLAG_USE_TWO_LABELS) && | |
2000 | bgpevpn_get_l3vni(vpn)) | |
2001 | add_l3_ecomm = 1; | |
2002 | ||
2003 | /* Set up extended community. */ | |
2004 | build_evpn_route_extcomm(vpn, &attr, add_l3_ecomm); | |
2005 | ||
2006 | seq = mac_mobility_seqnum(tmp_ri->attr); | |
2007 | ||
2008 | /* Update the route entry. */ | |
2009 | update_evpn_route_entry(bgp, vpn, afi, safi, rn, | |
2010 | &attr, 0, &ri, 0, seq); | |
2011 | ||
2012 | /* Perform route selection; this is just to set the flags | |
2013 | * correctly as local route in the VNI always wins. | |
2014 | */ | |
2015 | evpn_route_select_install(bgp, vpn, rn); | |
2016 | ||
2017 | attr_new = ri->attr; | |
2018 | ||
2019 | /* Update route in global routing table. */ | |
2020 | rd_rn = bgp_afi_node_get(bgp->rib[afi][safi], afi, safi, | |
2021 | (struct prefix *)evp, &vpn->prd); | |
2022 | assert(rd_rn); | |
2023 | update_evpn_route_entry(bgp, vpn, afi, safi, rd_rn, attr_new, 0, | |
2024 | &global_ri, 0, | |
2025 | mac_mobility_seqnum(attr_new)); | |
2026 | ||
2027 | /* Schedule for processing and unlock node. */ | |
2028 | bgp_process(bgp, rd_rn, afi, safi); | |
2029 | bgp_unlock_node(rd_rn); | |
2030 | ||
2031 | /* Unintern temporary. */ | |
2032 | aspath_unintern(&attr.aspath); | |
2033 | ||
2034 | } | |
2035 | ||
2036 | return 0; | |
2037 | } | |
2038 | ||
2039 | /* | |
2040 | * Delete all type-2 (MACIP) local routes for this VNI - only from the | |
2041 | * global routing table. These are also scheduled for withdraw from peers. | |
2042 | */ | |
2043 | static int delete_global_type2_routes(struct bgp *bgp, struct bgpevpn *vpn) | |
2044 | { | |
2045 | afi_t afi; | |
2046 | safi_t safi; | |
2047 | struct bgp_node *rdrn, *rn; | |
2048 | struct bgp_table *table; | |
2049 | struct bgp_info *ri; | |
2050 | ||
2051 | afi = AFI_L2VPN; | |
2052 | safi = SAFI_EVPN; | |
2053 | ||
2054 | rdrn = bgp_node_lookup(bgp->rib[afi][safi], (struct prefix *)&vpn->prd); | |
2055 | if (rdrn && rdrn->info) { | |
2056 | table = (struct bgp_table *)rdrn->info; | |
2057 | for (rn = bgp_table_top(table); rn; rn = bgp_route_next(rn)) { | |
2058 | struct prefix_evpn *evp = (struct prefix_evpn *)&rn->p; | |
2059 | ||
2060 | if (evp->prefix.route_type != BGP_EVPN_MAC_IP_ROUTE) | |
2061 | continue; | |
2062 | ||
2063 | delete_evpn_route_entry(bgp, afi, safi, rn, &ri); | |
2064 | if (ri) | |
2065 | bgp_process(bgp, rn, afi, safi); | |
2066 | } | |
2067 | } | |
2068 | ||
2069 | /* Unlock RD node. */ | |
2070 | if (rdrn) | |
2071 | bgp_unlock_node(rdrn); | |
2072 | ||
2073 | return 0; | |
2074 | } | |
2075 | ||
2076 | /* | |
2077 | * Delete all type-2 (MACIP) local routes for this VNI - from the global | |
2078 | * table as well as the per-VNI route table. | |
2079 | */ | |
2080 | static int delete_all_type2_routes(struct bgp *bgp, struct bgpevpn *vpn) | |
2081 | { | |
2082 | afi_t afi; | |
2083 | safi_t safi; | |
2084 | struct bgp_node *rn; | |
2085 | struct bgp_info *ri; | |
2086 | ||
2087 | afi = AFI_L2VPN; | |
2088 | safi = SAFI_EVPN; | |
2089 | ||
2090 | /* First, walk the global route table for this VNI's type-2 local | |
2091 | * routes. | |
2092 | * EVPN routes are a 2-level table, first get the RD table. | |
2093 | */ | |
2094 | delete_global_type2_routes(bgp, vpn); | |
2095 | ||
2096 | /* Next, walk this VNI's route table and delete local type-2 routes. */ | |
2097 | for (rn = bgp_table_top(vpn->route_table); rn; | |
2098 | rn = bgp_route_next(rn)) { | |
2099 | struct prefix_evpn *evp = (struct prefix_evpn *)&rn->p; | |
2100 | ||
2101 | if (evp->prefix.route_type != BGP_EVPN_MAC_IP_ROUTE) | |
2102 | continue; | |
2103 | ||
2104 | delete_evpn_route_entry(bgp, afi, safi, rn, &ri); | |
2105 | ||
2106 | /* Route entry in local table gets deleted immediately. */ | |
2107 | if (ri) | |
2108 | bgp_info_reap(rn, ri); | |
2109 | } | |
2110 | ||
2111 | return 0; | |
2112 | } | |
2113 | ||
2114 | /* | |
2115 | * Delete all routes in per ES route-table | |
2116 | */ | |
2117 | static int delete_all_es_routes(struct bgp *bgp, struct evpnes *es) | |
2118 | { | |
2119 | struct bgp_node *rn; | |
2120 | struct bgp_info *ri, *nextri; | |
2121 | ||
2122 | /* Walk this ES's route table and delete all routes. */ | |
2123 | for (rn = bgp_table_top(es->route_table); rn; | |
2124 | rn = bgp_route_next(rn)) { | |
2125 | for (ri = rn->info; (ri != NULL) && (nextri = ri->next, 1); | |
2126 | ri = nextri) { | |
2127 | bgp_info_delete(rn, ri); | |
2128 | bgp_info_reap(rn, ri); | |
2129 | } | |
2130 | } | |
2131 | ||
2132 | return 0; | |
2133 | } | |
2134 | ||
2135 | /* | |
2136 | * Delete all routes in the per-VNI route table. | |
2137 | */ | |
2138 | static int delete_all_vni_routes(struct bgp *bgp, struct bgpevpn *vpn) | |
2139 | { | |
2140 | struct bgp_node *rn; | |
2141 | struct bgp_info *ri, *nextri; | |
2142 | ||
2143 | /* Walk this VNI's route table and delete all routes. */ | |
2144 | for (rn = bgp_table_top(vpn->route_table); rn; | |
2145 | rn = bgp_route_next(rn)) { | |
2146 | for (ri = rn->info; (ri != NULL) && (nextri = ri->next, 1); | |
2147 | ri = nextri) { | |
2148 | bgp_info_delete(rn, ri); | |
2149 | bgp_info_reap(rn, ri); | |
2150 | } | |
2151 | } | |
2152 | ||
2153 | return 0; | |
2154 | } | |
2155 | ||
2156 | /* | |
2157 | * Update (and advertise) local routes for a VNI. Invoked upon the VNI | |
2158 | * export RT getting modified or change to tunnel IP. Note that these | |
2159 | * situations need the route in the per-VNI table as well as the global | |
2160 | * table to be updated (as attributes change). | |
2161 | */ | |
2162 | static int update_routes_for_vni(struct bgp *bgp, struct bgpevpn *vpn) | |
2163 | { | |
2164 | int ret; | |
2165 | struct prefix_evpn p; | |
2166 | ||
2167 | /* Update and advertise the type-3 route (only one) followed by the | |
2168 | * locally learnt type-2 routes (MACIP) - for this VNI. | |
2169 | */ | |
2170 | build_evpn_type3_prefix(&p, vpn->originator_ip); | |
2171 | ret = update_evpn_route(bgp, vpn, &p, 0, 0); | |
2172 | if (ret) | |
2173 | return ret; | |
2174 | ||
2175 | return update_all_type2_routes(bgp, vpn); | |
2176 | } | |
2177 | ||
2178 | /* Delete (and withdraw) local routes for specified ES from global and ES table. | |
2179 | * Also remove all other routes from the per ES table. | |
2180 | * Invoked when ES is deleted. | |
2181 | */ | |
2182 | static int delete_routes_for_es(struct bgp *bgp, struct evpnes *es) | |
2183 | { | |
2184 | int ret; | |
2185 | char buf[ESI_STR_LEN]; | |
2186 | struct prefix_evpn p; | |
2187 | ||
2188 | /* Delete and withdraw locally learnt ES route */ | |
2189 | build_evpn_type4_prefix(&p, &es->esi, es->originator_ip.ipaddr_v4); | |
2190 | ret = delete_evpn_type4_route(bgp, es, &p); | |
2191 | if (ret) { | |
2192 | flog_err(EC_BGP_EVPN_ROUTE_DELETE, | |
2193 | "%u failed to delete type-4 route for ESI %s", | |
2194 | bgp->vrf_id, esi_to_str(&es->esi, buf, sizeof(buf))); | |
2195 | } | |
2196 | ||
2197 | /* Delete all routes from per ES table */ | |
2198 | return delete_all_es_routes(bgp, es); | |
2199 | } | |
2200 | ||
2201 | /* | |
2202 | * Delete (and withdraw) local routes for specified VNI from the global | |
2203 | * table and per-VNI table. After this, remove all other routes from | |
2204 | * the per-VNI table. Invoked upon the VNI being deleted or EVPN | |
2205 | * (advertise-all-vni) being disabled. | |
2206 | */ | |
2207 | static int delete_routes_for_vni(struct bgp *bgp, struct bgpevpn *vpn) | |
2208 | { | |
2209 | int ret; | |
2210 | struct prefix_evpn p; | |
2211 | ||
2212 | /* Delete and withdraw locally learnt type-2 routes (MACIP) | |
2213 | * followed by type-3 routes (only one) - for this VNI. | |
2214 | */ | |
2215 | ret = delete_all_type2_routes(bgp, vpn); | |
2216 | if (ret) | |
2217 | return ret; | |
2218 | ||
2219 | build_evpn_type3_prefix(&p, vpn->originator_ip); | |
2220 | ret = delete_evpn_route(bgp, vpn, &p); | |
2221 | if (ret) | |
2222 | return ret; | |
2223 | ||
2224 | /* Delete all routes from the per-VNI table. */ | |
2225 | return delete_all_vni_routes(bgp, vpn); | |
2226 | } | |
2227 | ||
2228 | /* | |
2229 | * There is a tunnel endpoint IP address change for this VNI, delete | |
2230 | * prior type-3 route (if needed) and update. | |
2231 | * Note: Route re-advertisement happens elsewhere after other processing | |
2232 | * other changes. | |
2233 | */ | |
2234 | static int handle_tunnel_ip_change(struct bgp *bgp, struct bgpevpn *vpn, | |
2235 | struct in_addr originator_ip) | |
2236 | { | |
2237 | struct prefix_evpn p; | |
2238 | ||
2239 | /* If VNI is not live, we only need to update the originator ip */ | |
2240 | if (!is_vni_live(vpn)) { | |
2241 | vpn->originator_ip = originator_ip; | |
2242 | return 0; | |
2243 | } | |
2244 | ||
2245 | /* Update the tunnel-ip hash */ | |
2246 | bgp_tip_del(bgp, &vpn->originator_ip); | |
2247 | bgp_tip_add(bgp, &originator_ip); | |
2248 | ||
2249 | /* filter routes as martian nexthop db has changed */ | |
2250 | bgp_filter_evpn_routes_upon_martian_nh_change(bgp); | |
2251 | ||
2252 | /* Need to withdraw type-3 route as the originator IP is part | |
2253 | * of the key. | |
2254 | */ | |
2255 | build_evpn_type3_prefix(&p, vpn->originator_ip); | |
2256 | delete_evpn_route(bgp, vpn, &p); | |
2257 | ||
2258 | /* Update the tunnel IP and re-advertise all routes for this VNI. */ | |
2259 | vpn->originator_ip = originator_ip; | |
2260 | return 0; | |
2261 | } | |
2262 | ||
2263 | /* Install EVPN route entry in ES */ | |
2264 | static int install_evpn_route_entry_in_es(struct bgp *bgp, | |
2265 | struct evpnes *es, | |
2266 | struct prefix_evpn *p, | |
2267 | struct bgp_info *parent_ri) | |
2268 | { | |
2269 | int ret = 0; | |
2270 | struct bgp_node *rn = NULL; | |
2271 | struct bgp_info *ri = NULL; | |
2272 | struct attr *attr_new = NULL; | |
2273 | ||
2274 | /* Create (or fetch) route within the VNI. | |
2275 | * NOTE: There is no RD here. | |
2276 | */ | |
2277 | rn = bgp_node_get(es->route_table, (struct prefix *)p); | |
2278 | ||
2279 | /* Check if route entry is already present. */ | |
2280 | for (ri = rn->info; ri; ri = ri->next) | |
2281 | if (ri->extra && | |
2282 | (struct bgp_info *)ri->extra->parent == parent_ri) | |
2283 | break; | |
2284 | ||
2285 | if (!ri) { | |
2286 | /* Add (or update) attribute to hash. */ | |
2287 | attr_new = bgp_attr_intern(parent_ri->attr); | |
2288 | ||
2289 | /* Create new route with its attribute. */ | |
2290 | ri = info_make(parent_ri->type, BGP_ROUTE_IMPORTED, 0, | |
2291 | parent_ri->peer, attr_new, rn); | |
2292 | SET_FLAG(ri->flags, BGP_INFO_VALID); | |
2293 | bgp_info_extra_get(ri); | |
2294 | ri->extra->parent = parent_ri; | |
2295 | bgp_info_add(rn, ri); | |
2296 | } else { | |
2297 | if (attrhash_cmp(ri->attr, parent_ri->attr) | |
2298 | && !CHECK_FLAG(ri->flags, BGP_INFO_REMOVED)) { | |
2299 | bgp_unlock_node(rn); | |
2300 | return 0; | |
2301 | } | |
2302 | /* The attribute has changed. */ | |
2303 | /* Add (or update) attribute to hash. */ | |
2304 | attr_new = bgp_attr_intern(parent_ri->attr); | |
2305 | ||
2306 | /* Restore route, if needed. */ | |
2307 | if (CHECK_FLAG(ri->flags, BGP_INFO_REMOVED)) | |
2308 | bgp_info_restore(rn, ri); | |
2309 | ||
2310 | /* Mark if nexthop has changed. */ | |
2311 | if (!IPV4_ADDR_SAME(&ri->attr->nexthop, &attr_new->nexthop)) | |
2312 | SET_FLAG(ri->flags, BGP_INFO_IGP_CHANGED); | |
2313 | ||
2314 | /* Unintern existing, set to new. */ | |
2315 | bgp_attr_unintern(&ri->attr); | |
2316 | ri->attr = attr_new; | |
2317 | ri->uptime = bgp_clock(); | |
2318 | } | |
2319 | ||
2320 | /* Perform route selection and update zebra, if required. */ | |
2321 | ret = evpn_es_route_select_install(bgp, es, rn); | |
2322 | return ret; | |
2323 | } | |
2324 | ||
2325 | /* | |
2326 | * Install route entry into the VRF routing table and invoke route selection. | |
2327 | */ | |
2328 | static int install_evpn_route_entry_in_vrf(struct bgp *bgp_vrf, | |
2329 | struct prefix_evpn *evp, | |
2330 | struct bgp_info *parent_ri) | |
2331 | { | |
2332 | struct bgp_node *rn; | |
2333 | struct bgp_info *ri; | |
2334 | struct attr attr; | |
2335 | struct attr *attr_new; | |
2336 | int ret = 0; | |
2337 | struct prefix p; | |
2338 | struct prefix *pp = &p; | |
2339 | afi_t afi = 0; | |
2340 | safi_t safi = 0; | |
2341 | char buf[PREFIX_STRLEN]; | |
2342 | char buf1[PREFIX_STRLEN]; | |
2343 | ||
2344 | memset(pp, 0, sizeof(struct prefix)); | |
2345 | ip_prefix_from_evpn_prefix(evp, pp); | |
2346 | ||
2347 | if (bgp_debug_zebra(NULL)) { | |
2348 | zlog_debug( | |
2349 | "installing evpn prefix %s as ip prefix %s in vrf %s", | |
2350 | prefix2str(evp, buf, sizeof(buf)), | |
2351 | prefix2str(pp, buf1, sizeof(buf)), | |
2352 | vrf_id_to_name(bgp_vrf->vrf_id)); | |
2353 | } | |
2354 | ||
2355 | /* Create (or fetch) route within the VRF. */ | |
2356 | /* NOTE: There is no RD here. */ | |
2357 | if (is_evpn_prefix_ipaddr_v4(evp)) { | |
2358 | afi = AFI_IP; | |
2359 | safi = SAFI_UNICAST; | |
2360 | rn = bgp_node_get(bgp_vrf->rib[afi][safi], pp); | |
2361 | } else if (is_evpn_prefix_ipaddr_v6(evp)) { | |
2362 | afi = AFI_IP6; | |
2363 | safi = SAFI_UNICAST; | |
2364 | rn = bgp_node_get(bgp_vrf->rib[afi][safi], pp); | |
2365 | } else | |
2366 | return 0; | |
2367 | ||
2368 | /* EVPN routes currently only support a IPv4 next hop which corresponds | |
2369 | * to the remote VTEP. When importing into a VRF, if it is IPv6 host | |
2370 | * or prefix route, we have to convert the next hop to an IPv4-mapped | |
2371 | * address for the rest of the code to flow through. In the case of IPv4, | |
2372 | * make sure to set the flag for next hop attribute. | |
2373 | */ | |
2374 | bgp_attr_dup(&attr, parent_ri->attr); | |
2375 | if (afi == AFI_IP6) | |
2376 | evpn_convert_nexthop_to_ipv6(&attr); | |
2377 | else | |
2378 | attr.flag |= ATTR_FLAG_BIT(BGP_ATTR_NEXT_HOP); | |
2379 | ||
2380 | /* Check if route entry is already present. */ | |
2381 | for (ri = rn->info; ri; ri = ri->next) | |
2382 | if (ri->extra | |
2383 | && (struct bgp_info *)ri->extra->parent == parent_ri) | |
2384 | break; | |
2385 | ||
2386 | if (!ri) { | |
2387 | /* Add (or update) attribute to hash. */ | |
2388 | attr_new = bgp_attr_intern(&attr); | |
2389 | ||
2390 | /* Create new route with its attribute. */ | |
2391 | ri = info_make(parent_ri->type, BGP_ROUTE_IMPORTED, 0, | |
2392 | parent_ri->peer, attr_new, rn); | |
2393 | SET_FLAG(ri->flags, BGP_INFO_VALID); | |
2394 | bgp_info_extra_get(ri); | |
2395 | ri->extra->parent = bgp_info_lock(parent_ri); | |
2396 | bgp_lock_node((struct bgp_node *)parent_ri->net); | |
2397 | if (parent_ri->extra) { | |
2398 | memcpy(&ri->extra->label, &parent_ri->extra->label, | |
2399 | sizeof(ri->extra->label)); | |
2400 | ri->extra->num_labels = parent_ri->extra->num_labels; | |
2401 | } | |
2402 | bgp_info_add(rn, ri); | |
2403 | } else { | |
2404 | if (attrhash_cmp(ri->attr, &attr) | |
2405 | && !CHECK_FLAG(ri->flags, BGP_INFO_REMOVED)) { | |
2406 | bgp_unlock_node(rn); | |
2407 | return 0; | |
2408 | } | |
2409 | /* The attribute has changed. */ | |
2410 | /* Add (or update) attribute to hash. */ | |
2411 | attr_new = bgp_attr_intern(&attr); | |
2412 | ||
2413 | /* Restore route, if needed. */ | |
2414 | if (CHECK_FLAG(ri->flags, BGP_INFO_REMOVED)) | |
2415 | bgp_info_restore(rn, ri); | |
2416 | ||
2417 | /* Mark if nexthop has changed. */ | |
2418 | if ((afi == AFI_IP && | |
2419 | !IPV4_ADDR_SAME(&ri->attr->nexthop, &attr_new->nexthop)) || | |
2420 | (afi == AFI_IP6 && | |
2421 | !IPV6_ADDR_SAME(&ri->attr->mp_nexthop_global, | |
2422 | &attr_new->mp_nexthop_global))) | |
2423 | SET_FLAG(ri->flags, BGP_INFO_IGP_CHANGED); | |
2424 | ||
2425 | /* Unintern existing, set to new. */ | |
2426 | bgp_attr_unintern(&ri->attr); | |
2427 | ri->attr = attr_new; | |
2428 | ri->uptime = bgp_clock(); | |
2429 | } | |
2430 | ||
2431 | /* Perform route selection and update zebra, if required. */ | |
2432 | bgp_process(bgp_vrf, rn, afi, safi); | |
2433 | ||
2434 | return ret; | |
2435 | } | |
2436 | ||
2437 | /* | |
2438 | * Install route entry into the VNI routing table and invoke route selection. | |
2439 | */ | |
2440 | static int install_evpn_route_entry(struct bgp *bgp, struct bgpevpn *vpn, | |
2441 | struct prefix_evpn *p, | |
2442 | struct bgp_info *parent_ri) | |
2443 | { | |
2444 | struct bgp_node *rn; | |
2445 | struct bgp_info *ri; | |
2446 | struct attr *attr_new; | |
2447 | int ret; | |
2448 | ||
2449 | /* Create (or fetch) route within the VNI. */ | |
2450 | /* NOTE: There is no RD here. */ | |
2451 | rn = bgp_node_get(vpn->route_table, (struct prefix *)p); | |
2452 | ||
2453 | /* Check if route entry is already present. */ | |
2454 | for (ri = rn->info; ri; ri = ri->next) | |
2455 | if (ri->extra | |
2456 | && (struct bgp_info *)ri->extra->parent == parent_ri) | |
2457 | break; | |
2458 | ||
2459 | if (!ri) { | |
2460 | /* Add (or update) attribute to hash. */ | |
2461 | attr_new = bgp_attr_intern(parent_ri->attr); | |
2462 | ||
2463 | /* Create new route with its attribute. */ | |
2464 | ri = info_make(parent_ri->type, BGP_ROUTE_IMPORTED, 0, | |
2465 | parent_ri->peer, attr_new, rn); | |
2466 | SET_FLAG(ri->flags, BGP_INFO_VALID); | |
2467 | bgp_info_extra_get(ri); | |
2468 | ri->extra->parent = bgp_info_lock(parent_ri); | |
2469 | bgp_lock_node((struct bgp_node *)parent_ri->net); | |
2470 | if (parent_ri->extra) { | |
2471 | memcpy(&ri->extra->label, &parent_ri->extra->label, | |
2472 | sizeof(ri->extra->label)); | |
2473 | ri->extra->num_labels = parent_ri->extra->num_labels; | |
2474 | } | |
2475 | bgp_info_add(rn, ri); | |
2476 | } else { | |
2477 | if (attrhash_cmp(ri->attr, parent_ri->attr) | |
2478 | && !CHECK_FLAG(ri->flags, BGP_INFO_REMOVED)) { | |
2479 | bgp_unlock_node(rn); | |
2480 | return 0; | |
2481 | } | |
2482 | /* The attribute has changed. */ | |
2483 | /* Add (or update) attribute to hash. */ | |
2484 | attr_new = bgp_attr_intern(parent_ri->attr); | |
2485 | ||
2486 | /* Restore route, if needed. */ | |
2487 | if (CHECK_FLAG(ri->flags, BGP_INFO_REMOVED)) | |
2488 | bgp_info_restore(rn, ri); | |
2489 | ||
2490 | /* Mark if nexthop has changed. */ | |
2491 | if (!IPV4_ADDR_SAME(&ri->attr->nexthop, &attr_new->nexthop)) | |
2492 | SET_FLAG(ri->flags, BGP_INFO_IGP_CHANGED); | |
2493 | ||
2494 | /* Unintern existing, set to new. */ | |
2495 | bgp_attr_unintern(&ri->attr); | |
2496 | ri->attr = attr_new; | |
2497 | ri->uptime = bgp_clock(); | |
2498 | } | |
2499 | ||
2500 | /* Perform route selection and update zebra, if required. */ | |
2501 | ret = evpn_route_select_install(bgp, vpn, rn); | |
2502 | ||
2503 | return ret; | |
2504 | } | |
2505 | ||
2506 | /* Uninstall EVPN route entry from ES route table */ | |
2507 | static int uninstall_evpn_route_entry_in_es(struct bgp *bgp, | |
2508 | struct evpnes *es, | |
2509 | struct prefix_evpn *p, | |
2510 | struct bgp_info *parent_ri) | |
2511 | { | |
2512 | int ret; | |
2513 | struct bgp_node *rn; | |
2514 | struct bgp_info *ri; | |
2515 | ||
2516 | if (!es->route_table) | |
2517 | return 0; | |
2518 | ||
2519 | /* Locate route within the ESI. | |
2520 | * NOTE: There is no RD here. | |
2521 | */ | |
2522 | rn = bgp_node_lookup(es->route_table, (struct prefix *)p); | |
2523 | if (!rn) | |
2524 | return 0; | |
2525 | ||
2526 | /* Find matching route entry. */ | |
2527 | for (ri = rn->info; ri; ri = ri->next) | |
2528 | if (ri->extra && | |
2529 | (struct bgp_info *)ri->extra->parent == parent_ri) | |
2530 | break; | |
2531 | ||
2532 | if (!ri) | |
2533 | return 0; | |
2534 | ||
2535 | /* Mark entry for deletion */ | |
2536 | bgp_info_delete(rn, ri); | |
2537 | ||
2538 | /* Perform route selection and update zebra, if required. */ | |
2539 | ret = evpn_es_route_select_install(bgp, es, rn); | |
2540 | ||
2541 | /* Unlock route node. */ | |
2542 | bgp_unlock_node(rn); | |
2543 | ||
2544 | return ret; | |
2545 | } | |
2546 | ||
2547 | /* | |
2548 | * Uninstall route entry from the VRF routing table and send message | |
2549 | * to zebra, if appropriate. | |
2550 | */ | |
2551 | static int uninstall_evpn_route_entry_in_vrf(struct bgp *bgp_vrf, | |
2552 | struct prefix_evpn *evp, | |
2553 | struct bgp_info *parent_ri) | |
2554 | { | |
2555 | struct bgp_node *rn; | |
2556 | struct bgp_info *ri; | |
2557 | int ret = 0; | |
2558 | struct prefix p; | |
2559 | struct prefix *pp = &p; | |
2560 | afi_t afi = 0; | |
2561 | safi_t safi = 0; | |
2562 | char buf[PREFIX_STRLEN]; | |
2563 | char buf1[PREFIX_STRLEN]; | |
2564 | ||
2565 | memset(pp, 0, sizeof(struct prefix)); | |
2566 | ip_prefix_from_evpn_prefix(evp, pp); | |
2567 | ||
2568 | if (bgp_debug_zebra(NULL)) { | |
2569 | zlog_debug( | |
2570 | "uninstalling evpn prefix %s as ip prefix %s in vrf %s", | |
2571 | prefix2str(evp, buf, sizeof(buf)), | |
2572 | prefix2str(pp, buf1, sizeof(buf)), | |
2573 | vrf_id_to_name(bgp_vrf->vrf_id)); | |
2574 | } | |
2575 | ||
2576 | /* Locate route within the VRF. */ | |
2577 | /* NOTE: There is no RD here. */ | |
2578 | if (is_evpn_prefix_ipaddr_v4(evp)) { | |
2579 | afi = AFI_IP; | |
2580 | safi = SAFI_UNICAST; | |
2581 | rn = bgp_node_lookup(bgp_vrf->rib[afi][safi], pp); | |
2582 | } else { | |
2583 | afi = AFI_IP6; | |
2584 | safi = SAFI_UNICAST; | |
2585 | rn = bgp_node_lookup(bgp_vrf->rib[afi][safi], pp); | |
2586 | } | |
2587 | ||
2588 | if (!rn) | |
2589 | return 0; | |
2590 | ||
2591 | /* Find matching route entry. */ | |
2592 | for (ri = rn->info; ri; ri = ri->next) | |
2593 | if (ri->extra | |
2594 | && (struct bgp_info *)ri->extra->parent == parent_ri) | |
2595 | break; | |
2596 | ||
2597 | if (!ri) | |
2598 | return 0; | |
2599 | ||
2600 | /* Mark entry for deletion */ | |
2601 | bgp_info_delete(rn, ri); | |
2602 | ||
2603 | /* Perform route selection and update zebra, if required. */ | |
2604 | bgp_process(bgp_vrf, rn, afi, safi); | |
2605 | ||
2606 | /* Unlock route node. */ | |
2607 | bgp_unlock_node(rn); | |
2608 | ||
2609 | return ret; | |
2610 | } | |
2611 | ||
2612 | /* | |
2613 | * Uninstall route entry from the VNI routing table and send message | |
2614 | * to zebra, if appropriate. | |
2615 | */ | |
2616 | static int uninstall_evpn_route_entry(struct bgp *bgp, struct bgpevpn *vpn, | |
2617 | struct prefix_evpn *p, | |
2618 | struct bgp_info *parent_ri) | |
2619 | { | |
2620 | struct bgp_node *rn; | |
2621 | struct bgp_info *ri; | |
2622 | int ret; | |
2623 | ||
2624 | /* Locate route within the VNI. */ | |
2625 | /* NOTE: There is no RD here. */ | |
2626 | rn = bgp_node_lookup(vpn->route_table, (struct prefix *)p); | |
2627 | if (!rn) | |
2628 | return 0; | |
2629 | ||
2630 | /* Find matching route entry. */ | |
2631 | for (ri = rn->info; ri; ri = ri->next) | |
2632 | if (ri->extra | |
2633 | && (struct bgp_info *)ri->extra->parent == parent_ri) | |
2634 | break; | |
2635 | ||
2636 | if (!ri) | |
2637 | return 0; | |
2638 | ||
2639 | /* Mark entry for deletion */ | |
2640 | bgp_info_delete(rn, ri); | |
2641 | ||
2642 | /* Perform route selection and update zebra, if required. */ | |
2643 | ret = evpn_route_select_install(bgp, vpn, rn); | |
2644 | ||
2645 | /* Unlock route node. */ | |
2646 | bgp_unlock_node(rn); | |
2647 | ||
2648 | return ret; | |
2649 | } | |
2650 | ||
2651 | /* | |
2652 | * Given a prefix, see if it belongs to ES. | |
2653 | */ | |
2654 | static int is_prefix_matching_for_es(struct prefix_evpn *p, | |
2655 | struct evpnes *es) | |
2656 | { | |
2657 | /* if not an ES route return false */ | |
2658 | if (p->prefix.route_type != BGP_EVPN_ES_ROUTE) | |
2659 | return 0; | |
2660 | ||
2661 | if (memcmp(&p->prefix.es_addr.esi, &es->esi, sizeof(esi_t)) == 0) | |
2662 | return 1; | |
2663 | ||
2664 | return 0; | |
2665 | } | |
2666 | ||
2667 | /* | |
2668 | * Given a route entry and a VRF, see if this route entry should be | |
2669 | * imported into the VRF i.e., RTs match. | |
2670 | */ | |
2671 | static int is_route_matching_for_vrf(struct bgp *bgp_vrf, struct bgp_info *ri) | |
2672 | { | |
2673 | struct attr *attr = ri->attr; | |
2674 | struct ecommunity *ecom; | |
2675 | int i; | |
2676 | ||
2677 | assert(attr); | |
2678 | /* Route should have valid RT to be even considered. */ | |
2679 | if (!(attr->flag & ATTR_FLAG_BIT(BGP_ATTR_EXT_COMMUNITIES))) | |
2680 | return 0; | |
2681 | ||
2682 | ecom = attr->ecommunity; | |
2683 | if (!ecom || !ecom->size) | |
2684 | return 0; | |
2685 | ||
2686 | /* For each extended community RT, see if it matches this VNI. If any RT | |
2687 | * matches, we're done. | |
2688 | */ | |
2689 | for (i = 0; i < ecom->size; i++) { | |
2690 | uint8_t *pnt; | |
2691 | uint8_t type, sub_type; | |
2692 | struct ecommunity_val *eval; | |
2693 | struct ecommunity_val eval_tmp; | |
2694 | struct vrf_irt_node *irt; | |
2695 | ||
2696 | /* Only deal with RTs */ | |
2697 | pnt = (ecom->val + (i * ECOMMUNITY_SIZE)); | |
2698 | eval = (struct ecommunity_val *)(ecom->val | |
2699 | + (i * ECOMMUNITY_SIZE)); | |
2700 | type = *pnt++; | |
2701 | sub_type = *pnt++; | |
2702 | if (sub_type != ECOMMUNITY_ROUTE_TARGET) | |
2703 | continue; | |
2704 | ||
2705 | /* See if this RT matches specified VNIs import RTs */ | |
2706 | irt = lookup_vrf_import_rt(eval); | |
2707 | if (irt) | |
2708 | if (is_vrf_present_in_irt_vrfs(irt->vrfs, bgp_vrf)) | |
2709 | return 1; | |
2710 | ||
2711 | /* Also check for non-exact match. In this, we mask out the AS | |
2712 | * and | |
2713 | * only check on the local-admin sub-field. This is to | |
2714 | * facilitate using | |
2715 | * VNI as the RT for EBGP peering too. | |
2716 | */ | |
2717 | irt = NULL; | |
2718 | if (type == ECOMMUNITY_ENCODE_AS | |
2719 | || type == ECOMMUNITY_ENCODE_AS4 | |
2720 | || type == ECOMMUNITY_ENCODE_IP) { | |
2721 | memcpy(&eval_tmp, eval, ECOMMUNITY_SIZE); | |
2722 | mask_ecom_global_admin(&eval_tmp, eval); | |
2723 | irt = lookup_vrf_import_rt(&eval_tmp); | |
2724 | } | |
2725 | if (irt) | |
2726 | if (is_vrf_present_in_irt_vrfs(irt->vrfs, bgp_vrf)) | |
2727 | return 1; | |
2728 | } | |
2729 | ||
2730 | return 0; | |
2731 | } | |
2732 | ||
2733 | /* | |
2734 | * Given a route entry and a VNI, see if this route entry should be | |
2735 | * imported into the VNI i.e., RTs match. | |
2736 | */ | |
2737 | static int is_route_matching_for_vni(struct bgp *bgp, struct bgpevpn *vpn, | |
2738 | struct bgp_info *ri) | |
2739 | { | |
2740 | struct attr *attr = ri->attr; | |
2741 | struct ecommunity *ecom; | |
2742 | int i; | |
2743 | ||
2744 | assert(attr); | |
2745 | /* Route should have valid RT to be even considered. */ | |
2746 | if (!(attr->flag & ATTR_FLAG_BIT(BGP_ATTR_EXT_COMMUNITIES))) | |
2747 | return 0; | |
2748 | ||
2749 | ecom = attr->ecommunity; | |
2750 | if (!ecom || !ecom->size) | |
2751 | return 0; | |
2752 | ||
2753 | /* For each extended community RT, see if it matches this VNI. If any RT | |
2754 | * matches, we're done. | |
2755 | */ | |
2756 | for (i = 0; i < ecom->size; i++) { | |
2757 | uint8_t *pnt; | |
2758 | uint8_t type, sub_type; | |
2759 | struct ecommunity_val *eval; | |
2760 | struct ecommunity_val eval_tmp; | |
2761 | struct irt_node *irt; | |
2762 | ||
2763 | /* Only deal with RTs */ | |
2764 | pnt = (ecom->val + (i * ECOMMUNITY_SIZE)); | |
2765 | eval = (struct ecommunity_val *)(ecom->val | |
2766 | + (i * ECOMMUNITY_SIZE)); | |
2767 | type = *pnt++; | |
2768 | sub_type = *pnt++; | |
2769 | if (sub_type != ECOMMUNITY_ROUTE_TARGET) | |
2770 | continue; | |
2771 | ||
2772 | /* See if this RT matches specified VNIs import RTs */ | |
2773 | irt = lookup_import_rt(bgp, eval); | |
2774 | if (irt) | |
2775 | if (is_vni_present_in_irt_vnis(irt->vnis, vpn)) | |
2776 | return 1; | |
2777 | ||
2778 | /* Also check for non-exact match. In this, we mask out the AS | |
2779 | * and | |
2780 | * only check on the local-admin sub-field. This is to | |
2781 | * facilitate using | |
2782 | * VNI as the RT for EBGP peering too. | |
2783 | */ | |
2784 | irt = NULL; | |
2785 | if (type == ECOMMUNITY_ENCODE_AS | |
2786 | || type == ECOMMUNITY_ENCODE_AS4 | |
2787 | || type == ECOMMUNITY_ENCODE_IP) { | |
2788 | memcpy(&eval_tmp, eval, ECOMMUNITY_SIZE); | |
2789 | mask_ecom_global_admin(&eval_tmp, eval); | |
2790 | irt = lookup_import_rt(bgp, &eval_tmp); | |
2791 | } | |
2792 | if (irt) | |
2793 | if (is_vni_present_in_irt_vnis(irt->vnis, vpn)) | |
2794 | return 1; | |
2795 | } | |
2796 | ||
2797 | return 0; | |
2798 | } | |
2799 | ||
2800 | static int install_uninstall_routes_for_es(struct bgp *bgp, | |
2801 | struct evpnes *es, | |
2802 | int install) | |
2803 | { | |
2804 | int ret; | |
2805 | afi_t afi; | |
2806 | safi_t safi; | |
2807 | char buf[PREFIX_STRLEN]; | |
2808 | char buf1[ESI_STR_LEN]; | |
2809 | struct bgp_node *rd_rn, *rn; | |
2810 | struct bgp_table *table; | |
2811 | struct bgp_info *ri; | |
2812 | ||
2813 | afi = AFI_L2VPN; | |
2814 | safi = SAFI_EVPN; | |
2815 | ||
2816 | /* | |
2817 | * Walk entire global routing table and evaluate routes which could be | |
2818 | * imported into this VRF. Note that we need to loop through all global | |
2819 | * routes to determine which route matches the import rt on vrf | |
2820 | */ | |
2821 | for (rd_rn = bgp_table_top(bgp->rib[afi][safi]); rd_rn; | |
2822 | rd_rn = bgp_route_next(rd_rn)) { | |
2823 | table = (struct bgp_table *)(rd_rn->info); | |
2824 | if (!table) | |
2825 | continue; | |
2826 | ||
2827 | for (rn = bgp_table_top(table); rn; rn = bgp_route_next(rn)) { | |
2828 | struct prefix_evpn *evp = (struct prefix_evpn *)&rn->p; | |
2829 | ||
2830 | for (ri = rn->info; ri; ri = ri->next) { | |
2831 | /* | |
2832 | * Consider "valid" remote routes applicable for | |
2833 | * this ES. | |
2834 | */ | |
2835 | if (!(CHECK_FLAG(ri->flags, BGP_INFO_VALID) | |
2836 | && ri->type == ZEBRA_ROUTE_BGP | |
2837 | && ri->sub_type == BGP_ROUTE_NORMAL)) | |
2838 | continue; | |
2839 | ||
2840 | if (!is_prefix_matching_for_es(evp, es)) | |
2841 | continue; | |
2842 | ||
2843 | if (install) | |
2844 | ret = install_evpn_route_entry_in_es( | |
2845 | bgp, es, evp, ri); | |
2846 | else | |
2847 | ret = uninstall_evpn_route_entry_in_es( | |
2848 | bgp, es, evp, ri); | |
2849 | ||
2850 | if (ret) { | |
2851 | flog_err( | |
2852 | EC_BGP_EVPN_FAIL, | |
2853 | "Failed to %s EVPN %s route in ESI %s", | |
2854 | install ? "install" | |
2855 | : "uninstall", | |
2856 | prefix2str(evp, buf, | |
2857 | sizeof(buf)), | |
2858 | esi_to_str(&es->esi, buf1, | |
2859 | sizeof(buf1))); | |
2860 | return ret; | |
2861 | } | |
2862 | } | |
2863 | } | |
2864 | } | |
2865 | return 0; | |
2866 | } | |
2867 | ||
2868 | /* | |
2869 | * Install or uninstall mac-ip routes are appropriate for this | |
2870 | * particular VRF. | |
2871 | */ | |
2872 | static int install_uninstall_routes_for_vrf(struct bgp *bgp_vrf, int install) | |
2873 | { | |
2874 | afi_t afi; | |
2875 | safi_t safi; | |
2876 | struct bgp_node *rd_rn, *rn; | |
2877 | struct bgp_table *table; | |
2878 | struct bgp_info *ri; | |
2879 | int ret; | |
2880 | char buf[PREFIX_STRLEN]; | |
2881 | struct bgp *bgp_def = NULL; | |
2882 | ||
2883 | afi = AFI_L2VPN; | |
2884 | safi = SAFI_EVPN; | |
2885 | bgp_def = bgp_get_default(); | |
2886 | if (!bgp_def) | |
2887 | return -1; | |
2888 | ||
2889 | /* Walk entire global routing table and evaluate routes which could be | |
2890 | * imported into this VRF. Note that we need to loop through all global | |
2891 | * routes to determine which route matches the import rt on vrf | |
2892 | */ | |
2893 | for (rd_rn = bgp_table_top(bgp_def->rib[afi][safi]); rd_rn; | |
2894 | rd_rn = bgp_route_next(rd_rn)) { | |
2895 | table = (struct bgp_table *)(rd_rn->info); | |
2896 | if (!table) | |
2897 | continue; | |
2898 | ||
2899 | for (rn = bgp_table_top(table); rn; rn = bgp_route_next(rn)) { | |
2900 | struct prefix_evpn *evp = (struct prefix_evpn *)&rn->p; | |
2901 | ||
2902 | /* if not mac-ip route skip this route */ | |
2903 | if (!(evp->prefix.route_type == BGP_EVPN_MAC_IP_ROUTE | |
2904 | || evp->prefix.route_type | |
2905 | == BGP_EVPN_IP_PREFIX_ROUTE)) | |
2906 | continue; | |
2907 | ||
2908 | /* if not a mac+ip route skip this route */ | |
2909 | if (!(is_evpn_prefix_ipaddr_v4(evp) | |
2910 | || is_evpn_prefix_ipaddr_v6(evp))) | |
2911 | continue; | |
2912 | ||
2913 | for (ri = rn->info; ri; ri = ri->next) { | |
2914 | /* Consider "valid" remote routes applicable for | |
2915 | * this VRF. | |
2916 | */ | |
2917 | if (!(CHECK_FLAG(ri->flags, BGP_INFO_VALID) | |
2918 | && ri->type == ZEBRA_ROUTE_BGP | |
2919 | && ri->sub_type == BGP_ROUTE_NORMAL)) | |
2920 | continue; | |
2921 | ||
2922 | if (is_route_matching_for_vrf(bgp_vrf, ri)) { | |
2923 | if (install) | |
2924 | ret = install_evpn_route_entry_in_vrf( | |
2925 | bgp_vrf, evp, ri); | |
2926 | else | |
2927 | ret = uninstall_evpn_route_entry_in_vrf( | |
2928 | bgp_vrf, evp, ri); | |
2929 | ||
2930 | if (ret) { | |
2931 | flog_err( | |
2932 | EC_BGP_EVPN_FAIL, | |
2933 | "Failed to %s EVPN %s route in VRF %s", | |
2934 | install ? "install" | |
2935 | : "uninstall", | |
2936 | prefix2str(evp, buf, | |
2937 | sizeof(buf)), | |
2938 | vrf_id_to_name( | |
2939 | bgp_vrf->vrf_id)); | |
2940 | return ret; | |
2941 | } | |
2942 | } | |
2943 | } | |
2944 | } | |
2945 | } | |
2946 | ||
2947 | return 0; | |
2948 | } | |
2949 | ||
2950 | /* | |
2951 | * Install or uninstall routes of specified type that are appropriate for this | |
2952 | * particular VNI. | |
2953 | */ | |
2954 | static int install_uninstall_routes_for_vni(struct bgp *bgp, | |
2955 | struct bgpevpn *vpn, | |
2956 | bgp_evpn_route_type rtype, | |
2957 | int install) | |
2958 | { | |
2959 | afi_t afi; | |
2960 | safi_t safi; | |
2961 | struct bgp_node *rd_rn, *rn; | |
2962 | struct bgp_table *table; | |
2963 | struct bgp_info *ri; | |
2964 | int ret; | |
2965 | ||
2966 | afi = AFI_L2VPN; | |
2967 | safi = SAFI_EVPN; | |
2968 | ||
2969 | /* Walk entire global routing table and evaluate routes which could be | |
2970 | * imported into this VPN. Note that we cannot just look at the routes | |
2971 | * for | |
2972 | * the VNI's RD - remote routes applicable for this VNI could have any | |
2973 | * RD. | |
2974 | */ | |
2975 | /* EVPN routes are a 2-level table. */ | |
2976 | for (rd_rn = bgp_table_top(bgp->rib[afi][safi]); rd_rn; | |
2977 | rd_rn = bgp_route_next(rd_rn)) { | |
2978 | table = (struct bgp_table *)(rd_rn->info); | |
2979 | if (!table) | |
2980 | continue; | |
2981 | ||
2982 | for (rn = bgp_table_top(table); rn; rn = bgp_route_next(rn)) { | |
2983 | struct prefix_evpn *evp = (struct prefix_evpn *)&rn->p; | |
2984 | ||
2985 | if (evp->prefix.route_type != rtype) | |
2986 | continue; | |
2987 | ||
2988 | for (ri = rn->info; ri; ri = ri->next) { | |
2989 | /* Consider "valid" remote routes applicable for | |
2990 | * this VNI. */ | |
2991 | if (!(CHECK_FLAG(ri->flags, BGP_INFO_VALID) | |
2992 | && ri->type == ZEBRA_ROUTE_BGP | |
2993 | && ri->sub_type == BGP_ROUTE_NORMAL)) | |
2994 | continue; | |
2995 | ||
2996 | if (is_route_matching_for_vni(bgp, vpn, ri)) { | |
2997 | if (install) | |
2998 | ret = install_evpn_route_entry( | |
2999 | bgp, vpn, evp, ri); | |
3000 | else | |
3001 | ret = uninstall_evpn_route_entry( | |
3002 | bgp, vpn, evp, ri); | |
3003 | ||
3004 | if (ret) { | |
3005 | flog_err( | |
3006 | EC_BGP_EVPN_FAIL, | |
3007 | "%u: Failed to %s EVPN %s route in VNI %u", | |
3008 | bgp->vrf_id, | |
3009 | install ? "install" | |
3010 | : "uninstall", | |
3011 | rtype == BGP_EVPN_MAC_IP_ROUTE | |
3012 | ? "MACIP" | |
3013 | : "IMET", | |
3014 | vpn->vni); | |
3015 | return ret; | |
3016 | } | |
3017 | } | |
3018 | } | |
3019 | } | |
3020 | } | |
3021 | ||
3022 | return 0; | |
3023 | } | |
3024 | ||
3025 | /* Install any existing remote ES routes applicable for this ES into its routing | |
3026 | * table. This is invoked when ES comes up. | |
3027 | */ | |
3028 | static int install_routes_for_es(struct bgp *bgp, struct evpnes *es) | |
3029 | { | |
3030 | return install_uninstall_routes_for_es(bgp, es, 1); | |
3031 | } | |
3032 | ||
3033 | ||
3034 | /* Install any existing remote routes applicable for this VRF into VRF RIB. This | |
3035 | * is invoked upon l3vni-add or l3vni import rt change | |
3036 | */ | |
3037 | static int install_routes_for_vrf(struct bgp *bgp_vrf) | |
3038 | { | |
3039 | install_uninstall_routes_for_vrf(bgp_vrf, 1); | |
3040 | return 0; | |
3041 | } | |
3042 | ||
3043 | /* | |
3044 | * Install any existing remote routes applicable for this VNI into its | |
3045 | * routing table. This is invoked when a VNI becomes "live" or its Import | |
3046 | * RT is changed. | |
3047 | */ | |
3048 | static int install_routes_for_vni(struct bgp *bgp, struct bgpevpn *vpn) | |
3049 | { | |
3050 | int ret; | |
3051 | ||
3052 | /* Install type-3 routes followed by type-2 routes - the ones applicable | |
3053 | * for this VNI. | |
3054 | */ | |
3055 | ret = install_uninstall_routes_for_vni(bgp, vpn, BGP_EVPN_IMET_ROUTE, | |
3056 | 1); | |
3057 | if (ret) | |
3058 | return ret; | |
3059 | ||
3060 | return install_uninstall_routes_for_vni(bgp, vpn, BGP_EVPN_MAC_IP_ROUTE, | |
3061 | 1); | |
3062 | } | |
3063 | ||
3064 | /* uninstall routes from l3vni vrf. */ | |
3065 | static int uninstall_routes_for_vrf(struct bgp *bgp_vrf) | |
3066 | { | |
3067 | install_uninstall_routes_for_vrf(bgp_vrf, 0); | |
3068 | return 0; | |
3069 | } | |
3070 | ||
3071 | /* | |
3072 | * Uninstall any existing remote routes for this VNI. One scenario in which | |
3073 | * this is invoked is upon an import RT change. | |
3074 | */ | |
3075 | static int uninstall_routes_for_vni(struct bgp *bgp, struct bgpevpn *vpn) | |
3076 | { | |
3077 | int ret; | |
3078 | ||
3079 | /* Uninstall type-2 routes followed by type-3 routes - the ones | |
3080 | * applicable | |
3081 | * for this VNI. | |
3082 | */ | |
3083 | ret = install_uninstall_routes_for_vni(bgp, vpn, BGP_EVPN_MAC_IP_ROUTE, | |
3084 | 0); | |
3085 | if (ret) | |
3086 | return ret; | |
3087 | ||
3088 | return install_uninstall_routes_for_vni(bgp, vpn, BGP_EVPN_IMET_ROUTE, | |
3089 | 0); | |
3090 | } | |
3091 | ||
3092 | /* Install or unistall route in ES */ | |
3093 | static int install_uninstall_route_in_es(struct bgp *bgp, struct evpnes *es, | |
3094 | afi_t afi, safi_t safi, | |
3095 | struct prefix_evpn *evp, | |
3096 | struct bgp_info *ri, | |
3097 | int install) | |
3098 | { | |
3099 | int ret = 0; | |
3100 | char buf[ESI_STR_LEN]; | |
3101 | ||
3102 | if (install) | |
3103 | ret = install_evpn_route_entry_in_es(bgp, es, evp, ri); | |
3104 | else | |
3105 | ret = uninstall_evpn_route_entry_in_es(bgp, es, evp, ri); | |
3106 | ||
3107 | if (ret) { | |
3108 | flog_err( | |
3109 | EC_BGP_EVPN_FAIL, | |
3110 | "%u: Failed to %s EVPN %s route in ESI %s", bgp->vrf_id, | |
3111 | install ? "install" : "uninstall", "ES", | |
3112 | esi_to_str(&evp->prefix.es_addr.esi, buf, sizeof(buf))); | |
3113 | return ret; | |
3114 | } | |
3115 | return 0; | |
3116 | } | |
3117 | ||
3118 | /* | |
3119 | * Install or uninstall route in matching VRFs (list). | |
3120 | */ | |
3121 | static int install_uninstall_route_in_vrfs(struct bgp *bgp_def, afi_t afi, | |
3122 | safi_t safi, struct prefix_evpn *evp, | |
3123 | struct bgp_info *ri, | |
3124 | struct list *vrfs, int install) | |
3125 | { | |
3126 | char buf[PREFIX2STR_BUFFER]; | |
3127 | struct bgp *bgp_vrf; | |
3128 | struct listnode *node, *nnode; | |
3129 | ||
3130 | /* Only type-2/type-5 routes go into a VRF */ | |
3131 | if (!(evp->prefix.route_type == BGP_EVPN_MAC_IP_ROUTE | |
3132 | || evp->prefix.route_type == BGP_EVPN_IP_PREFIX_ROUTE)) | |
3133 | return 0; | |
3134 | ||
3135 | /* if it is type-2 route and not a mac+ip route skip this route */ | |
3136 | if ((evp->prefix.route_type == BGP_EVPN_MAC_IP_ROUTE) | |
3137 | && !(is_evpn_prefix_ipaddr_v4(evp) | |
3138 | || is_evpn_prefix_ipaddr_v6(evp))) | |
3139 | return 0; | |
3140 | ||
3141 | for (ALL_LIST_ELEMENTS(vrfs, node, nnode, bgp_vrf)) { | |
3142 | int ret; | |
3143 | ||
3144 | if (install) | |
3145 | ret = install_evpn_route_entry_in_vrf(bgp_vrf, evp, ri); | |
3146 | else | |
3147 | ret = uninstall_evpn_route_entry_in_vrf(bgp_vrf, evp, | |
3148 | ri); | |
3149 | ||
3150 | if (ret) { | |
3151 | flog_err(EC_BGP_EVPN_FAIL, | |
3152 | "%u: Failed to %s prefix %s in VRF %s", | |
3153 | bgp_def->vrf_id, | |
3154 | install ? "install" : "uninstall", | |
3155 | prefix2str(evp, buf, sizeof(buf)), | |
3156 | vrf_id_to_name(bgp_vrf->vrf_id)); | |
3157 | return ret; | |
3158 | } | |
3159 | } | |
3160 | ||
3161 | return 0; | |
3162 | } | |
3163 | ||
3164 | /* | |
3165 | * Install or uninstall route in matching VNIs (list). | |
3166 | */ | |
3167 | static int install_uninstall_route_in_vnis(struct bgp *bgp, afi_t afi, | |
3168 | safi_t safi, struct prefix_evpn *evp, | |
3169 | struct bgp_info *ri, | |
3170 | struct list *vnis, int install) | |
3171 | { | |
3172 | struct bgpevpn *vpn; | |
3173 | struct listnode *node, *nnode; | |
3174 | ||
3175 | for (ALL_LIST_ELEMENTS(vnis, node, nnode, vpn)) { | |
3176 | int ret; | |
3177 | ||
3178 | if (!is_vni_live(vpn)) | |
3179 | continue; | |
3180 | ||
3181 | if (install) | |
3182 | ret = install_evpn_route_entry(bgp, vpn, evp, ri); | |
3183 | else | |
3184 | ret = uninstall_evpn_route_entry(bgp, vpn, evp, ri); | |
3185 | ||
3186 | if (ret) { | |
3187 | flog_err(EC_BGP_EVPN_FAIL, | |
3188 | "%u: Failed to %s EVPN %s route in VNI %u", | |
3189 | bgp->vrf_id, install ? "install" : "uninstall", | |
3190 | evp->prefix.route_type == BGP_EVPN_MAC_IP_ROUTE | |
3191 | ? "MACIP" | |
3192 | : "IMET", | |
3193 | vpn->vni); | |
3194 | return ret; | |
3195 | } | |
3196 | } | |
3197 | ||
3198 | return 0; | |
3199 | } | |
3200 | ||
3201 | /* | |
3202 | * Install or uninstall route for appropriate VNIs/ESIs. | |
3203 | */ | |
3204 | static int install_uninstall_evpn_route(struct bgp *bgp, afi_t afi, safi_t safi, | |
3205 | struct prefix *p, struct bgp_info *ri, | |
3206 | int import) | |
3207 | { | |
3208 | struct prefix_evpn *evp = (struct prefix_evpn *)p; | |
3209 | struct attr *attr = ri->attr; | |
3210 | struct ecommunity *ecom; | |
3211 | int i; | |
3212 | ||
3213 | assert(attr); | |
3214 | ||
3215 | /* Only type-2, type-3, type-4 and type-5 are supported currently */ | |
3216 | if (!(evp->prefix.route_type == BGP_EVPN_MAC_IP_ROUTE | |
3217 | || evp->prefix.route_type == BGP_EVPN_IMET_ROUTE | |
3218 | || evp->prefix.route_type == BGP_EVPN_ES_ROUTE | |
3219 | || evp->prefix.route_type == BGP_EVPN_IP_PREFIX_ROUTE)) | |
3220 | return 0; | |
3221 | ||
3222 | /* If we don't have Route Target, nothing much to do. */ | |
3223 | if (!(attr->flag & ATTR_FLAG_BIT(BGP_ATTR_EXT_COMMUNITIES))) | |
3224 | return 0; | |
3225 | ||
3226 | ecom = attr->ecommunity; | |
3227 | if (!ecom || !ecom->size) | |
3228 | return -1; | |
3229 | ||
3230 | /* An EVPN route belongs to a VNI or a VRF or an ESI based on the RTs | |
3231 | * attached to the route */ | |
3232 | for (i = 0; i < ecom->size; i++) { | |
3233 | uint8_t *pnt; | |
3234 | uint8_t type, sub_type; | |
3235 | struct ecommunity_val *eval; | |
3236 | struct ecommunity_val eval_tmp; | |
3237 | struct irt_node *irt; /* import rt for l2vni */ | |
3238 | struct vrf_irt_node *vrf_irt; /* import rt for l3vni */ | |
3239 | struct evpnes *es; | |
3240 | ||
3241 | /* Only deal with RTs */ | |
3242 | pnt = (ecom->val + (i * ECOMMUNITY_SIZE)); | |
3243 | eval = (struct ecommunity_val *)(ecom->val | |
3244 | + (i * ECOMMUNITY_SIZE)); | |
3245 | type = *pnt++; | |
3246 | sub_type = *pnt++; | |
3247 | if (sub_type != ECOMMUNITY_ROUTE_TARGET) | |
3248 | continue; | |
3249 | ||
3250 | /* | |
3251 | * macip routes (type-2) are imported into VNI and VRF tables. | |
3252 | * IMET route is imported into VNI table. | |
3253 | * prefix routes are imported into VRF table. | |
3254 | */ | |
3255 | if (evp->prefix.route_type == BGP_EVPN_MAC_IP_ROUTE || | |
3256 | evp->prefix.route_type == BGP_EVPN_IMET_ROUTE || | |
3257 | evp->prefix.route_type == BGP_EVPN_IP_PREFIX_ROUTE) { | |
3258 | ||
3259 | irt = lookup_import_rt(bgp, eval); | |
3260 | if (irt) | |
3261 | install_uninstall_route_in_vnis(bgp, afi, safi, | |
3262 | evp, ri, | |
3263 | irt->vnis, | |
3264 | import); | |
3265 | ||
3266 | vrf_irt = lookup_vrf_import_rt(eval); | |
3267 | if (vrf_irt) | |
3268 | install_uninstall_route_in_vrfs(bgp, afi, safi, | |
3269 | evp, ri, | |
3270 | vrf_irt->vrfs, | |
3271 | import); | |
3272 | ||
3273 | /* Also check for non-exact match. | |
3274 | * In this, we mask out the AS and | |
3275 | * only check on the local-admin sub-field. | |
3276 | * This is to facilitate using | |
3277 | * VNI as the RT for EBGP peering too. | |
3278 | */ | |
3279 | irt = NULL; | |
3280 | vrf_irt = NULL; | |
3281 | if (type == ECOMMUNITY_ENCODE_AS | |
3282 | || type == ECOMMUNITY_ENCODE_AS4 | |
3283 | || type == ECOMMUNITY_ENCODE_IP) { | |
3284 | memcpy(&eval_tmp, eval, ECOMMUNITY_SIZE); | |
3285 | mask_ecom_global_admin(&eval_tmp, eval); | |
3286 | irt = lookup_import_rt(bgp, &eval_tmp); | |
3287 | vrf_irt = lookup_vrf_import_rt(&eval_tmp); | |
3288 | } | |
3289 | ||
3290 | if (irt) | |
3291 | install_uninstall_route_in_vnis(bgp, afi, safi, | |
3292 | evp, ri, | |
3293 | irt->vnis, | |
3294 | import); | |
3295 | if (vrf_irt) | |
3296 | install_uninstall_route_in_vrfs(bgp, afi, safi, | |
3297 | evp, ri, | |
3298 | vrf_irt->vrfs, | |
3299 | import); | |
3300 | } | |
3301 | ||
3302 | /* es route is imported into the es table */ | |
3303 | if (evp->prefix.route_type == BGP_EVPN_ES_ROUTE) { | |
3304 | ||
3305 | /* we will match based on the entire esi to avoid | |
3306 | * imoort of an es route for esi2 into esi1 | |
3307 | */ | |
3308 | es = bgp_evpn_lookup_es(bgp, &evp->prefix.es_addr.esi); | |
3309 | if (es && is_es_local(es)) | |
3310 | install_uninstall_route_in_es(bgp, es, | |
3311 | afi, safi, | |
3312 | evp, ri, import); | |
3313 | } | |
3314 | } | |
3315 | ||
3316 | return 0; | |
3317 | } | |
3318 | ||
3319 | /* | |
3320 | * delete and withdraw all ipv4 and ipv6 routes in the vrf table as type-5 | |
3321 | * routes | |
3322 | */ | |
3323 | static void delete_withdraw_vrf_routes(struct bgp *bgp_vrf) | |
3324 | { | |
3325 | /* delete all ipv4 routes and withdraw from peers */ | |
3326 | if (advertise_type5_routes(bgp_vrf, AFI_IP)) | |
3327 | bgp_evpn_withdraw_type5_routes(bgp_vrf, AFI_IP, SAFI_UNICAST); | |
3328 | ||
3329 | /* delete all ipv6 routes and withdraw from peers */ | |
3330 | if (advertise_type5_routes(bgp_vrf, AFI_IP6)) | |
3331 | bgp_evpn_withdraw_type5_routes(bgp_vrf, AFI_IP6, SAFI_UNICAST); | |
3332 | } | |
3333 | ||
3334 | /* | |
3335 | * update and advertise all ipv4 and ipv6 routes in thr vrf table as type-5 | |
3336 | * routes | |
3337 | */ | |
3338 | static void update_advertise_vrf_routes(struct bgp *bgp_vrf) | |
3339 | { | |
3340 | /* update all ipv4 routes */ | |
3341 | if (advertise_type5_routes(bgp_vrf, AFI_IP)) | |
3342 | bgp_evpn_advertise_type5_routes(bgp_vrf, AFI_IP, SAFI_UNICAST); | |
3343 | ||
3344 | /* update all ipv6 routes */ | |
3345 | if (advertise_type5_routes(bgp_vrf, AFI_IP6)) | |
3346 | bgp_evpn_advertise_type5_routes(bgp_vrf, AFI_IP6, SAFI_UNICAST); | |
3347 | } | |
3348 | ||
3349 | /* | |
3350 | * update and advertise local routes for a VRF as type-5 routes. | |
3351 | * This is invoked upon RD change for a VRF. Note taht the processing is only | |
3352 | * done in the global route table using the routes which already exist in the | |
3353 | * VRF routing table | |
3354 | */ | |
3355 | static void update_router_id_vrf(struct bgp *bgp_vrf) | |
3356 | { | |
3357 | /* skip if the RD is configured */ | |
3358 | if (is_vrf_rd_configured(bgp_vrf)) | |
3359 | return; | |
3360 | ||
3361 | /* derive the RD for the VRF based on new router-id */ | |
3362 | bgp_evpn_derive_auto_rd_for_vrf(bgp_vrf); | |
3363 | ||
3364 | /* update advertise ipv4|ipv6 routes as type-5 routes */ | |
3365 | update_advertise_vrf_routes(bgp_vrf); | |
3366 | } | |
3367 | ||
3368 | /* | |
3369 | * Delete and withdraw all type-5 routes for the RD corresponding to VRF. | |
3370 | * This is invoked upon VRF RD change. The processing is done only from global | |
3371 | * table. | |
3372 | */ | |
3373 | static void withdraw_router_id_vrf(struct bgp *bgp_vrf) | |
3374 | { | |
3375 | /* skip if the RD is configured */ | |
3376 | if (is_vrf_rd_configured(bgp_vrf)) | |
3377 | return; | |
3378 | ||
3379 | /* delete/withdraw ipv4|ipv6 routes as type-5 routes */ | |
3380 | delete_withdraw_vrf_routes(bgp_vrf); | |
3381 | } | |
3382 | ||
3383 | /* | |
3384 | * Update and advertise local routes for a VNI. Invoked upon router-id | |
3385 | * change. Note that the processing is done only on the global route table | |
3386 | * using routes that already exist in the per-VNI table. | |
3387 | */ | |
3388 | static int update_advertise_vni_routes(struct bgp *bgp, struct bgpevpn *vpn) | |
3389 | { | |
3390 | struct prefix_evpn p; | |
3391 | struct bgp_node *rn, *global_rn; | |
3392 | struct bgp_info *ri, *global_ri; | |
3393 | struct attr *attr; | |
3394 | afi_t afi = AFI_L2VPN; | |
3395 | safi_t safi = SAFI_EVPN; | |
3396 | ||
3397 | /* Locate type-3 route for VNI in the per-VNI table and use its | |
3398 | * attributes to create and advertise the type-3 route for this VNI | |
3399 | * in the global table. | |
3400 | */ | |
3401 | build_evpn_type3_prefix(&p, vpn->originator_ip); | |
3402 | rn = bgp_node_lookup(vpn->route_table, (struct prefix *)&p); | |
3403 | if (!rn) /* unexpected */ | |
3404 | return 0; | |
3405 | for (ri = rn->info; ri; ri = ri->next) | |
3406 | if (ri->peer == bgp->peer_self && ri->type == ZEBRA_ROUTE_BGP | |
3407 | && ri->sub_type == BGP_ROUTE_STATIC) | |
3408 | break; | |
3409 | if (!ri) /* unexpected */ | |
3410 | return 0; | |
3411 | attr = ri->attr; | |
3412 | ||
3413 | global_rn = bgp_afi_node_get(bgp->rib[afi][safi], afi, safi, | |
3414 | (struct prefix *)&p, &vpn->prd); | |
3415 | update_evpn_route_entry(bgp, vpn, afi, safi, global_rn, attr, 1, &ri, | |
3416 | 0, mac_mobility_seqnum(attr)); | |
3417 | ||
3418 | /* Schedule for processing and unlock node. */ | |
3419 | bgp_process(bgp, global_rn, afi, safi); | |
3420 | bgp_unlock_node(global_rn); | |
3421 | ||
3422 | /* Now, walk this VNI's route table and use the route and its attribute | |
3423 | * to create and schedule route in global table. | |
3424 | */ | |
3425 | for (rn = bgp_table_top(vpn->route_table); rn; | |
3426 | rn = bgp_route_next(rn)) { | |
3427 | struct prefix_evpn *evp = (struct prefix_evpn *)&rn->p; | |
3428 | ||
3429 | /* Identify MAC-IP local routes. */ | |
3430 | if (evp->prefix.route_type != BGP_EVPN_MAC_IP_ROUTE) | |
3431 | continue; | |
3432 | ||
3433 | for (ri = rn->info; ri; ri = ri->next) | |
3434 | if (ri->peer == bgp->peer_self | |
3435 | && ri->type == ZEBRA_ROUTE_BGP | |
3436 | && ri->sub_type == BGP_ROUTE_STATIC) | |
3437 | break; | |
3438 | if (!ri) | |
3439 | continue; | |
3440 | ||
3441 | /* Create route in global routing table using this route entry's | |
3442 | * attribute. | |
3443 | */ | |
3444 | attr = ri->attr; | |
3445 | global_rn = bgp_afi_node_get(bgp->rib[afi][safi], afi, safi, | |
3446 | (struct prefix *)evp, &vpn->prd); | |
3447 | assert(global_rn); | |
3448 | update_evpn_route_entry(bgp, vpn, afi, safi, global_rn, attr, 1, | |
3449 | &global_ri, 0, | |
3450 | mac_mobility_seqnum(attr)); | |
3451 | ||
3452 | /* Schedule for processing and unlock node. */ | |
3453 | bgp_process(bgp, global_rn, afi, safi); | |
3454 | bgp_unlock_node(global_rn); | |
3455 | } | |
3456 | ||
3457 | return 0; | |
3458 | } | |
3459 | ||
3460 | /* | |
3461 | * Delete (and withdraw) local routes for a VNI - only from the global | |
3462 | * table. Invoked upon router-id change. | |
3463 | */ | |
3464 | static int delete_withdraw_vni_routes(struct bgp *bgp, struct bgpevpn *vpn) | |
3465 | { | |
3466 | int ret; | |
3467 | struct prefix_evpn p; | |
3468 | struct bgp_node *global_rn; | |
3469 | struct bgp_info *ri; | |
3470 | afi_t afi = AFI_L2VPN; | |
3471 | safi_t safi = SAFI_EVPN; | |
3472 | ||
3473 | /* Delete and withdraw locally learnt type-2 routes (MACIP) | |
3474 | * for this VNI - from the global table. | |
3475 | */ | |
3476 | ret = delete_global_type2_routes(bgp, vpn); | |
3477 | if (ret) | |
3478 | return ret; | |
3479 | ||
3480 | /* Remove type-3 route for this VNI from global table. */ | |
3481 | build_evpn_type3_prefix(&p, vpn->originator_ip); | |
3482 | global_rn = bgp_afi_node_lookup(bgp->rib[afi][safi], afi, safi, | |
3483 | (struct prefix *)&p, &vpn->prd); | |
3484 | if (global_rn) { | |
3485 | /* Delete route entry in the global EVPN table. */ | |
3486 | delete_evpn_route_entry(bgp, afi, safi, global_rn, &ri); | |
3487 | ||
3488 | /* Schedule for processing - withdraws to peers happen from | |
3489 | * this table. | |
3490 | */ | |
3491 | if (ri) | |
3492 | bgp_process(bgp, global_rn, afi, safi); | |
3493 | bgp_unlock_node(global_rn); | |
3494 | } | |
3495 | ||
3496 | return 0; | |
3497 | } | |
3498 | ||
3499 | /* | |
3500 | * Handle router-id change. Update and advertise local routes corresponding | |
3501 | * to this VNI from peers. Note that this is invoked after updating the | |
3502 | * router-id. The routes in the per-VNI table are used to create routes in | |
3503 | * the global table and schedule them. | |
3504 | */ | |
3505 | static void update_router_id_vni(struct hash_backet *backet, struct bgp *bgp) | |
3506 | { | |
3507 | struct bgpevpn *vpn = (struct bgpevpn *)backet->data; | |
3508 | ||
3509 | /* Skip VNIs with configured RD. */ | |
3510 | if (is_rd_configured(vpn)) | |
3511 | return; | |
3512 | ||
3513 | bgp_evpn_derive_auto_rd(bgp, vpn); | |
3514 | update_advertise_vni_routes(bgp, vpn); | |
3515 | } | |
3516 | ||
3517 | /* | |
3518 | * Handle router-id change. Delete and withdraw local routes corresponding | |
3519 | * to this VNI from peers. Note that this is invoked prior to updating | |
3520 | * the router-id and is done only on the global route table, the routes | |
3521 | * are needed in the per-VNI table to re-advertise with new router id. | |
3522 | */ | |
3523 | static void withdraw_router_id_vni(struct hash_backet *backet, struct bgp *bgp) | |
3524 | { | |
3525 | struct bgpevpn *vpn = (struct bgpevpn *)backet->data; | |
3526 | ||
3527 | /* Skip VNIs with configured RD. */ | |
3528 | if (is_rd_configured(vpn)) | |
3529 | return; | |
3530 | ||
3531 | delete_withdraw_vni_routes(bgp, vpn); | |
3532 | } | |
3533 | ||
3534 | /* | |
3535 | * Process received EVPN type-2 route (advertise or withdraw). | |
3536 | */ | |
3537 | static int process_type2_route(struct peer *peer, afi_t afi, safi_t safi, | |
3538 | struct attr *attr, uint8_t *pfx, int psize, | |
3539 | uint32_t addpath_id) | |
3540 | { | |
3541 | struct prefix_rd prd; | |
3542 | struct prefix_evpn p; | |
3543 | struct bgp_route_evpn evpn; | |
3544 | uint8_t ipaddr_len; | |
3545 | uint8_t macaddr_len; | |
3546 | mpls_label_t label[BGP_MAX_LABELS]; /* holds the VNI(s) as in packet */ | |
3547 | uint32_t num_labels = 0; | |
3548 | uint32_t eth_tag; | |
3549 | int ret; | |
3550 | ||
3551 | /* Type-2 route should be either 33, 37 or 49 bytes or an | |
3552 | * additional 3 bytes if there is a second label (VNI): | |
3553 | * RD (8), ESI (10), Eth Tag (4), MAC Addr Len (1), | |
3554 | * MAC Addr (6), IP len (1), IP (0, 4 or 16), | |
3555 | * MPLS Lbl1 (3), MPLS Lbl2 (0 or 3) | |
3556 | */ | |
3557 | if (psize != 33 && psize != 37 && psize != 49 && psize != 36 | |
3558 | && psize != 40 && psize != 52) { | |
3559 | flog_err(EC_BGP_EVPN_ROUTE_INVALID, | |
3560 | "%u:%s - Rx EVPN Type-2 NLRI with invalid length %d", | |
3561 | peer->bgp->vrf_id, peer->host, psize); | |
3562 | return -1; | |
3563 | } | |
3564 | ||
3565 | memset(&evpn, 0, sizeof(evpn)); | |
3566 | ||
3567 | /* Make prefix_rd */ | |
3568 | prd.family = AF_UNSPEC; | |
3569 | prd.prefixlen = 64; | |
3570 | memcpy(&prd.val, pfx, 8); | |
3571 | pfx += 8; | |
3572 | ||
3573 | /* Make EVPN prefix. */ | |
3574 | memset(&p, 0, sizeof(struct prefix_evpn)); | |
3575 | p.family = AF_EVPN; | |
3576 | p.prefixlen = EVPN_ROUTE_PREFIXLEN; | |
3577 | p.prefix.route_type = BGP_EVPN_MAC_IP_ROUTE; | |
3578 | ||
3579 | /* Copy Ethernet Seg Identifier */ | |
3580 | memcpy(&evpn.eth_s_id.val, pfx, ESI_LEN); | |
3581 | pfx += ESI_LEN; | |
3582 | ||
3583 | /* Copy Ethernet Tag */ | |
3584 | memcpy(ð_tag, pfx, 4); | |
3585 | p.prefix.macip_addr.eth_tag = ntohl(eth_tag); | |
3586 | pfx += 4; | |
3587 | ||
3588 | /* Get the MAC Addr len */ | |
3589 | macaddr_len = *pfx++; | |
3590 | ||
3591 | /* Get the MAC Addr */ | |
3592 | if (macaddr_len == (ETH_ALEN * 8)) { | |
3593 | memcpy(&p.prefix.macip_addr.mac.octet, pfx, ETH_ALEN); | |
3594 | pfx += ETH_ALEN; | |
3595 | } else { | |
3596 | flog_err( | |
3597 | EC_BGP_EVPN_ROUTE_INVALID, | |
3598 | "%u:%s - Rx EVPN Type-2 NLRI with unsupported MAC address length %d", | |
3599 | peer->bgp->vrf_id, peer->host, macaddr_len); | |
3600 | return -1; | |
3601 | } | |
3602 | ||
3603 | ||
3604 | /* Get the IP. */ | |
3605 | ipaddr_len = *pfx++; | |
3606 | if (ipaddr_len != 0 && ipaddr_len != IPV4_MAX_BITLEN | |
3607 | && ipaddr_len != IPV6_MAX_BITLEN) { | |
3608 | flog_err( | |
3609 | EC_BGP_EVPN_ROUTE_INVALID, | |
3610 | "%u:%s - Rx EVPN Type-2 NLRI with unsupported IP address length %d", | |
3611 | peer->bgp->vrf_id, peer->host, ipaddr_len); | |
3612 | return -1; | |
3613 | } | |
3614 | ||
3615 | if (ipaddr_len) { | |
3616 | ipaddr_len /= 8; /* Convert to bytes. */ | |
3617 | p.prefix.macip_addr.ip.ipa_type = (ipaddr_len == IPV4_MAX_BYTELEN) | |
3618 | ? IPADDR_V4 | |
3619 | : IPADDR_V6; | |
3620 | memcpy(&p.prefix.macip_addr.ip.ip.addr, pfx, ipaddr_len); | |
3621 | } | |
3622 | pfx += ipaddr_len; | |
3623 | ||
3624 | /* Get the VNI(s). Stored as bytes here. */ | |
3625 | num_labels++; | |
3626 | memset(label, 0, sizeof(label)); | |
3627 | memcpy(&label[0], pfx, BGP_LABEL_BYTES); | |
3628 | pfx += BGP_LABEL_BYTES; | |
3629 | psize -= (33 + ipaddr_len); | |
3630 | /* Do we have a second VNI? */ | |
3631 | if (psize) { | |
3632 | num_labels++; | |
3633 | memcpy(&label[1], pfx, BGP_LABEL_BYTES); | |
3634 | /* | |
3635 | * If in future, we are required to access additional fields, | |
3636 | * we MUST increment pfx by BGP_LABEL_BYTES in before reading | |
3637 | * the next field | |
3638 | */ | |
3639 | } | |
3640 | ||
3641 | /* Process the route. */ | |
3642 | if (attr) | |
3643 | ret = bgp_update(peer, (struct prefix *)&p, addpath_id, attr, | |
3644 | afi, safi, ZEBRA_ROUTE_BGP, BGP_ROUTE_NORMAL, | |
3645 | &prd, &label[0], num_labels, 0, &evpn); | |
3646 | else | |
3647 | ret = bgp_withdraw(peer, (struct prefix *)&p, addpath_id, attr, | |
3648 | afi, safi, ZEBRA_ROUTE_BGP, BGP_ROUTE_NORMAL, | |
3649 | &prd, &label[0], num_labels, &evpn); | |
3650 | return ret; | |
3651 | } | |
3652 | ||
3653 | /* | |
3654 | * Process received EVPN type-3 route (advertise or withdraw). | |
3655 | */ | |
3656 | static int process_type3_route(struct peer *peer, afi_t afi, safi_t safi, | |
3657 | struct attr *attr, uint8_t *pfx, int psize, | |
3658 | uint32_t addpath_id) | |
3659 | { | |
3660 | struct prefix_rd prd; | |
3661 | struct prefix_evpn p; | |
3662 | uint8_t ipaddr_len; | |
3663 | uint32_t eth_tag; | |
3664 | int ret; | |
3665 | ||
3666 | /* Type-3 route should be either 17 or 29 bytes: RD (8), Eth Tag (4), | |
3667 | * IP len (1) and IP (4 or 16). | |
3668 | */ | |
3669 | if (psize != 17 && psize != 29) { | |
3670 | flog_err(EC_BGP_EVPN_ROUTE_INVALID, | |
3671 | "%u:%s - Rx EVPN Type-3 NLRI with invalid length %d", | |
3672 | peer->bgp->vrf_id, peer->host, psize); | |
3673 | return -1; | |
3674 | } | |
3675 | ||
3676 | /* If PMSI is present, log if it is anything other than IR. | |
3677 | * Note: We just simply ignore the values as it is not clear if | |
3678 | * doing anything else is better. | |
3679 | */ | |
3680 | if (attr && | |
3681 | (attr->flag & ATTR_FLAG_BIT(BGP_ATTR_PMSI_TUNNEL))) { | |
3682 | if (attr->pmsi_tnl_type != PMSI_TNLTYPE_INGR_REPL) { | |
3683 | flog_warn( | |
3684 | EC_BGP_EVPN_PMSI_PRESENT, | |
3685 | "%u:%s - Rx EVPN Type-3 NLRI with unsupported PTA %d", | |
3686 | peer->bgp->vrf_id, peer->host, | |
3687 | attr->pmsi_tnl_type); | |
3688 | } | |
3689 | } | |
3690 | ||
3691 | /* Make prefix_rd */ | |
3692 | prd.family = AF_UNSPEC; | |
3693 | prd.prefixlen = 64; | |
3694 | memcpy(&prd.val, pfx, 8); | |
3695 | pfx += 8; | |
3696 | ||
3697 | /* Make EVPN prefix. */ | |
3698 | memset(&p, 0, sizeof(struct prefix_evpn)); | |
3699 | p.family = AF_EVPN; | |
3700 | p.prefixlen = EVPN_ROUTE_PREFIXLEN; | |
3701 | p.prefix.route_type = BGP_EVPN_IMET_ROUTE; | |
3702 | ||
3703 | /* Copy Ethernet Tag */ | |
3704 | memcpy(ð_tag, pfx, 4); | |
3705 | p.prefix.imet_addr.eth_tag = ntohl(eth_tag); | |
3706 | pfx += 4; | |
3707 | ||
3708 | /* Get the IP. */ | |
3709 | ipaddr_len = *pfx++; | |
3710 | if (ipaddr_len == IPV4_MAX_BITLEN) { | |
3711 | p.prefix.imet_addr.ip.ipa_type = IPADDR_V4; | |
3712 | memcpy(&p.prefix.imet_addr.ip.ip.addr, pfx, IPV4_MAX_BYTELEN); | |
3713 | } else { | |
3714 | flog_err( | |
3715 | EC_BGP_EVPN_ROUTE_INVALID, | |
3716 | "%u:%s - Rx EVPN Type-3 NLRI with unsupported IP address length %d", | |
3717 | peer->bgp->vrf_id, peer->host, ipaddr_len); | |
3718 | return -1; | |
3719 | } | |
3720 | ||
3721 | /* Process the route. */ | |
3722 | if (attr) | |
3723 | ret = bgp_update(peer, (struct prefix *)&p, addpath_id, attr, | |
3724 | afi, safi, ZEBRA_ROUTE_BGP, BGP_ROUTE_NORMAL, | |
3725 | &prd, NULL, 0, 0, NULL); | |
3726 | else | |
3727 | ret = bgp_withdraw(peer, (struct prefix *)&p, addpath_id, attr, | |
3728 | afi, safi, ZEBRA_ROUTE_BGP, BGP_ROUTE_NORMAL, | |
3729 | &prd, NULL, 0, NULL); | |
3730 | return ret; | |
3731 | } | |
3732 | ||
3733 | /* | |
3734 | * Process received EVPN type-4 route (advertise or withdraw). | |
3735 | */ | |
3736 | static int process_type4_route(struct peer *peer, afi_t afi, safi_t safi, | |
3737 | struct attr *attr, uint8_t *pfx, int psize, | |
3738 | uint32_t addpath_id) | |
3739 | { | |
3740 | int ret; | |
3741 | esi_t esi; | |
3742 | uint8_t ipaddr_len; | |
3743 | struct in_addr vtep_ip; | |
3744 | struct prefix_rd prd; | |
3745 | struct prefix_evpn p; | |
3746 | ||
3747 | /* Type-4 route should be either 23 or 35 bytes | |
3748 | * RD (8), ESI (10), ip-len (1), ip (4 or 16) | |
3749 | */ | |
3750 | if (psize != 23 && psize != 35) { | |
3751 | flog_err(EC_BGP_EVPN_ROUTE_INVALID, | |
3752 | "%u:%s - Rx EVPN Type-4 NLRI with invalid length %d", | |
3753 | peer->bgp->vrf_id, peer->host, psize); | |
3754 | return -1; | |
3755 | } | |
3756 | ||
3757 | /* Make prefix_rd */ | |
3758 | prd.family = AF_UNSPEC; | |
3759 | prd.prefixlen = 64; | |
3760 | memcpy(&prd.val, pfx, 8); | |
3761 | pfx += 8; | |
3762 | ||
3763 | /* get the ESI */ | |
3764 | memcpy(&esi, pfx, ESI_BYTES); | |
3765 | pfx += ESI_BYTES; | |
3766 | ||
3767 | ||
3768 | /* Get the IP. */ | |
3769 | ipaddr_len = *pfx++; | |
3770 | if (ipaddr_len == IPV4_MAX_BITLEN) { | |
3771 | memcpy(&vtep_ip, pfx, IPV4_MAX_BYTELEN); | |
3772 | } else { | |
3773 | flog_err( | |
3774 | EC_BGP_EVPN_ROUTE_INVALID, | |
3775 | "%u:%s - Rx EVPN Type-4 NLRI with unsupported IP address length %d", | |
3776 | peer->bgp->vrf_id, peer->host, ipaddr_len); | |
3777 | return -1; | |
3778 | } | |
3779 | ||
3780 | build_evpn_type4_prefix(&p, &esi, vtep_ip); | |
3781 | /* Process the route. */ | |
3782 | if (attr) { | |
3783 | ret = bgp_update(peer, (struct prefix *)&p, addpath_id, attr, | |
3784 | afi, safi, ZEBRA_ROUTE_BGP, BGP_ROUTE_NORMAL, | |
3785 | &prd, NULL, 0, 0, NULL); | |
3786 | } else { | |
3787 | ret = bgp_withdraw(peer, (struct prefix *)&p, addpath_id, attr, | |
3788 | afi, safi, ZEBRA_ROUTE_BGP, BGP_ROUTE_NORMAL, | |
3789 | &prd, NULL, 0, NULL); | |
3790 | } | |
3791 | return ret; | |
3792 | } | |
3793 | ||
3794 | ||
3795 | /* | |
3796 | * Process received EVPN type-5 route (advertise or withdraw). | |
3797 | */ | |
3798 | static int process_type5_route(struct peer *peer, afi_t afi, safi_t safi, | |
3799 | struct attr *attr, uint8_t *pfx, int psize, | |
3800 | uint32_t addpath_id, int withdraw) | |
3801 | { | |
3802 | struct prefix_rd prd; | |
3803 | struct prefix_evpn p; | |
3804 | struct bgp_route_evpn evpn; | |
3805 | uint8_t ippfx_len; | |
3806 | uint32_t eth_tag; | |
3807 | mpls_label_t label; /* holds the VNI as in the packet */ | |
3808 | int ret; | |
3809 | ||
3810 | /* Type-5 route should be 34 or 58 bytes: | |
3811 | * RD (8), ESI (10), Eth Tag (4), IP len (1), IP (4 or 16), | |
3812 | * GW (4 or 16) and VNI (3). | |
3813 | * Note that the IP and GW should both be IPv4 or both IPv6. | |
3814 | */ | |
3815 | if (psize != 34 && psize != 58) { | |
3816 | flog_err(EC_BGP_EVPN_ROUTE_INVALID, | |
3817 | "%u:%s - Rx EVPN Type-5 NLRI with invalid length %d", | |
3818 | peer->bgp->vrf_id, peer->host, psize); | |
3819 | return -1; | |
3820 | } | |
3821 | ||
3822 | /* Make prefix_rd */ | |
3823 | prd.family = AF_UNSPEC; | |
3824 | prd.prefixlen = 64; | |
3825 | memcpy(&prd.val, pfx, 8); | |
3826 | pfx += 8; | |
3827 | ||
3828 | /* Make EVPN prefix. */ | |
3829 | memset(&p, 0, sizeof(struct prefix_evpn)); | |
3830 | p.family = AF_EVPN; | |
3831 | p.prefixlen = EVPN_ROUTE_PREFIXLEN; | |
3832 | p.prefix.route_type = BGP_EVPN_IP_PREFIX_ROUTE; | |
3833 | ||
3834 | /* Additional information outside of prefix - ESI and GW IP */ | |
3835 | memset(&evpn, 0, sizeof(evpn)); | |
3836 | ||
3837 | /* Fetch ESI */ | |
3838 | memcpy(&evpn.eth_s_id.val, pfx, 10); | |
3839 | pfx += 10; | |
3840 | ||
3841 | /* Fetch Ethernet Tag. */ | |
3842 | memcpy(ð_tag, pfx, 4); | |
3843 | p.prefix.prefix_addr.eth_tag = ntohl(eth_tag); | |
3844 | pfx += 4; | |
3845 | ||
3846 | /* Fetch IP prefix length. */ | |
3847 | ippfx_len = *pfx++; | |
3848 | if (ippfx_len > IPV6_MAX_BITLEN) { | |
3849 | flog_err( | |
3850 | EC_BGP_EVPN_ROUTE_INVALID, | |
3851 | "%u:%s - Rx EVPN Type-5 NLRI with invalid IP Prefix length %d", | |
3852 | peer->bgp->vrf_id, peer->host, ippfx_len); | |
3853 | return -1; | |
3854 | } | |
3855 | p.prefix.prefix_addr.ip_prefix_length = ippfx_len; | |
3856 | ||
3857 | /* Determine IPv4 or IPv6 prefix */ | |
3858 | /* Since the address and GW are from the same family, this just becomes | |
3859 | * a simple check on the total size. | |
3860 | */ | |
3861 | if (psize == 34) { | |
3862 | SET_IPADDR_V4(&p.prefix.prefix_addr.ip); | |
3863 | memcpy(&p.prefix.prefix_addr.ip.ipaddr_v4, pfx, 4); | |
3864 | pfx += 4; | |
3865 | memcpy(&evpn.gw_ip.ipv4, pfx, 4); | |
3866 | pfx += 4; | |
3867 | } else { | |
3868 | SET_IPADDR_V6(&p.prefix.prefix_addr.ip); | |
3869 | memcpy(&p.prefix.prefix_addr.ip.ipaddr_v6, pfx, 16); | |
3870 | pfx += 16; | |
3871 | memcpy(&evpn.gw_ip.ipv6, pfx, 16); | |
3872 | pfx += 16; | |
3873 | } | |
3874 | ||
3875 | /* Get the VNI (in MPLS label field). Stored as bytes here. */ | |
3876 | memset(&label, 0, sizeof(label)); | |
3877 | memcpy(&label, pfx, BGP_LABEL_BYTES); | |
3878 | ||
3879 | /* | |
3880 | * If in future, we are required to access additional fields, | |
3881 | * we MUST increment pfx by BGP_LABEL_BYTES in before reading the next | |
3882 | * field | |
3883 | */ | |
3884 | ||
3885 | /* Process the route. */ | |
3886 | if (!withdraw) | |
3887 | ret = bgp_update(peer, (struct prefix *)&p, addpath_id, attr, | |
3888 | afi, safi, ZEBRA_ROUTE_BGP, BGP_ROUTE_NORMAL, | |
3889 | &prd, &label, 1, 0, &evpn); | |
3890 | else | |
3891 | ret = bgp_withdraw(peer, (struct prefix *)&p, addpath_id, attr, | |
3892 | afi, safi, ZEBRA_ROUTE_BGP, BGP_ROUTE_NORMAL, | |
3893 | &prd, &label, 1, &evpn); | |
3894 | ||
3895 | return ret; | |
3896 | } | |
3897 | ||
3898 | static void evpn_mpattr_encode_type5(struct stream *s, struct prefix *p, | |
3899 | struct prefix_rd *prd, mpls_label_t *label, | |
3900 | uint32_t num_labels, struct attr *attr) | |
3901 | { | |
3902 | int len; | |
3903 | char temp[16]; | |
3904 | struct evpn_addr *p_evpn_p; | |
3905 | ||
3906 | memset(&temp, 0, 16); | |
3907 | if (p->family != AF_EVPN) | |
3908 | return; | |
3909 | p_evpn_p = &(p->u.prefix_evpn); | |
3910 | ||
3911 | /* len denites the total len of IP and GW-IP in the route | |
3912 | IP and GW-IP have to be both ipv4 or ipv6 | |
3913 | */ | |
3914 | if (IS_IPADDR_V4(&p_evpn_p->prefix_addr.ip)) | |
3915 | len = 8; /* IP and GWIP are both ipv4 */ | |
3916 | else | |
3917 | len = 32; /* IP and GWIP are both ipv6 */ | |
3918 | /* Prefix contains RD, ESI, EthTag, IP length, IP, GWIP and VNI */ | |
3919 | stream_putc(s, 8 + 10 + 4 + 1 + len + 3); | |
3920 | stream_put(s, prd->val, 8); | |
3921 | if (attr) | |
3922 | stream_put(s, &(attr->evpn_overlay.eth_s_id), 10); | |
3923 | else | |
3924 | stream_put(s, &temp, 10); | |
3925 | stream_putl(s, p_evpn_p->prefix_addr.eth_tag); | |
3926 | stream_putc(s, p_evpn_p->prefix_addr.ip_prefix_length); | |
3927 | if (IS_IPADDR_V4(&p_evpn_p->prefix_addr.ip)) | |
3928 | stream_put_ipv4(s, p_evpn_p->prefix_addr.ip.ipaddr_v4.s_addr); | |
3929 | else | |
3930 | stream_put(s, &p_evpn_p->prefix_addr.ip.ipaddr_v6, 16); | |
3931 | if (attr) { | |
3932 | if (IS_IPADDR_V4(&p_evpn_p->prefix_addr.ip)) | |
3933 | stream_put_ipv4(s, | |
3934 | attr->evpn_overlay.gw_ip.ipv4.s_addr); | |
3935 | else | |
3936 | stream_put(s, &(attr->evpn_overlay.gw_ip.ipv6), 16); | |
3937 | } else { | |
3938 | if (IS_IPADDR_V4(&p_evpn_p->prefix_addr.ip)) | |
3939 | stream_put_ipv4(s, 0); | |
3940 | else | |
3941 | stream_put(s, &temp, 16); | |
3942 | } | |
3943 | ||
3944 | if (num_labels) | |
3945 | stream_put(s, label, 3); | |
3946 | else | |
3947 | stream_put3(s, 0); | |
3948 | } | |
3949 | ||
3950 | /* | |
3951 | * Cleanup specific VNI upon EVPN (advertise-all-vni) being disabled. | |
3952 | */ | |
3953 | static void cleanup_vni_on_disable(struct hash_backet *backet, struct bgp *bgp) | |
3954 | { | |
3955 | struct bgpevpn *vpn = (struct bgpevpn *)backet->data; | |
3956 | ||
3957 | /* Remove EVPN routes and schedule for processing. */ | |
3958 | delete_routes_for_vni(bgp, vpn); | |
3959 | ||
3960 | /* Clear "live" flag and see if hash needs to be freed. */ | |
3961 | UNSET_FLAG(vpn->flags, VNI_FLAG_LIVE); | |
3962 | if (!is_vni_configured(vpn)) | |
3963 | bgp_evpn_free(bgp, vpn); | |
3964 | } | |
3965 | ||
3966 | /* | |
3967 | * Free a VNI entry; iterator function called during cleanup. | |
3968 | */ | |
3969 | static void free_vni_entry(struct hash_backet *backet, struct bgp *bgp) | |
3970 | { | |
3971 | struct bgpevpn *vpn = (struct bgpevpn *)backet->data; | |
3972 | ||
3973 | delete_all_vni_routes(bgp, vpn); | |
3974 | bgp_evpn_free(bgp, vpn); | |
3975 | } | |
3976 | ||
3977 | /* | |
3978 | * Derive AUTO import RT for BGP VRF - L3VNI | |
3979 | */ | |
3980 | static void evpn_auto_rt_import_add_for_vrf(struct bgp *bgp_vrf) | |
3981 | { | |
3982 | struct bgp *bgp_def = NULL; | |
3983 | ||
3984 | form_auto_rt(bgp_vrf, bgp_vrf->l3vni, bgp_vrf->vrf_import_rtl); | |
3985 | UNSET_FLAG(bgp_vrf->vrf_flags, BGP_VRF_IMPORT_RT_CFGD); | |
3986 | ||
3987 | /* Map RT to VRF */ | |
3988 | bgp_def = bgp_get_default(); | |
3989 | if (!bgp_def) | |
3990 | return; | |
3991 | bgp_evpn_map_vrf_to_its_rts(bgp_vrf); | |
3992 | } | |
3993 | ||
3994 | /* | |
3995 | * Delete AUTO import RT from BGP VRF - L3VNI | |
3996 | */ | |
3997 | static void evpn_auto_rt_import_delete_for_vrf(struct bgp *bgp_vrf) | |
3998 | { | |
3999 | evpn_rt_delete_auto(bgp_vrf, bgp_vrf->l3vni, bgp_vrf->vrf_import_rtl); | |
4000 | } | |
4001 | ||
4002 | /* | |
4003 | * Derive AUTO export RT for BGP VRF - L3VNI | |
4004 | */ | |
4005 | static void evpn_auto_rt_export_add_for_vrf(struct bgp *bgp_vrf) | |
4006 | { | |
4007 | UNSET_FLAG(bgp_vrf->vrf_flags, BGP_VRF_EXPORT_RT_CFGD); | |
4008 | form_auto_rt(bgp_vrf, bgp_vrf->l3vni, bgp_vrf->vrf_export_rtl); | |
4009 | } | |
4010 | ||
4011 | /* | |
4012 | * Delete AUTO export RT from BGP VRF - L3VNI | |
4013 | */ | |
4014 | static void evpn_auto_rt_export_delete_for_vrf(struct bgp *bgp_vrf) | |
4015 | { | |
4016 | evpn_rt_delete_auto(bgp_vrf, bgp_vrf->l3vni, bgp_vrf->vrf_export_rtl); | |
4017 | } | |
4018 | ||
4019 | static void bgp_evpn_handle_export_rt_change_for_vrf(struct bgp *bgp_vrf) | |
4020 | { | |
4021 | struct bgp *bgp_def = NULL; | |
4022 | struct listnode *node = NULL; | |
4023 | struct bgpevpn *vpn = NULL; | |
4024 | ||
4025 | bgp_def = bgp_get_default(); | |
4026 | if (!bgp_def) | |
4027 | return; | |
4028 | ||
4029 | /* update all type-5 routes */ | |
4030 | update_advertise_vrf_routes(bgp_vrf); | |
4031 | ||
4032 | /* update all type-2 routes */ | |
4033 | for (ALL_LIST_ELEMENTS_RO(bgp_vrf->l2vnis, node, vpn)) | |
4034 | update_routes_for_vni(bgp_def, vpn); | |
4035 | } | |
4036 | ||
4037 | /* | |
4038 | * Handle autort change for a given VNI. | |
4039 | */ | |
4040 | static void update_autort_vni(struct hash_backet *backet, struct bgp *bgp) | |
4041 | { | |
4042 | struct bgpevpn *vpn = backet->data; | |
4043 | ||
4044 | if (!is_import_rt_configured(vpn)) { | |
4045 | if (is_vni_live(vpn)) | |
4046 | bgp_evpn_uninstall_routes(bgp, vpn); | |
4047 | bgp_evpn_unmap_vni_from_its_rts(bgp, vpn); | |
4048 | list_delete_all_node(vpn->import_rtl); | |
4049 | bgp_evpn_derive_auto_rt_import(bgp, vpn); | |
4050 | if (is_vni_live(vpn)) | |
4051 | bgp_evpn_install_routes(bgp, vpn); | |
4052 | } | |
4053 | if (!is_export_rt_configured(vpn)) { | |
4054 | list_delete_all_node(vpn->export_rtl); | |
4055 | bgp_evpn_derive_auto_rt_export(bgp, vpn); | |
4056 | if (is_vni_live(vpn)) | |
4057 | bgp_evpn_handle_export_rt_change(bgp, vpn); | |
4058 | } | |
4059 | } | |
4060 | ||
4061 | /* | |
4062 | * Public functions. | |
4063 | */ | |
4064 | ||
4065 | /* withdraw type-5 route corresponding to ip prefix */ | |
4066 | void bgp_evpn_withdraw_type5_route(struct bgp *bgp_vrf, struct prefix *p, | |
4067 | afi_t afi, safi_t safi) | |
4068 | { | |
4069 | int ret = 0; | |
4070 | struct prefix_evpn evp; | |
4071 | char buf[PREFIX_STRLEN]; | |
4072 | ||
4073 | build_type5_prefix_from_ip_prefix(&evp, p); | |
4074 | ret = delete_evpn_type5_route(bgp_vrf, &evp); | |
4075 | if (ret) { | |
4076 | flog_err( | |
4077 | EC_BGP_EVPN_ROUTE_DELETE, | |
4078 | "%u failed to delete type-5 route for prefix %s in vrf %s", | |
4079 | bgp_vrf->vrf_id, prefix2str(p, buf, sizeof(buf)), | |
4080 | vrf_id_to_name(bgp_vrf->vrf_id)); | |
4081 | } | |
4082 | } | |
4083 | ||
4084 | /* withdraw all type-5 routes for an address family */ | |
4085 | void bgp_evpn_withdraw_type5_routes(struct bgp *bgp_vrf, afi_t afi, safi_t safi) | |
4086 | { | |
4087 | struct bgp_table *table = NULL; | |
4088 | struct bgp_node *rn = NULL; | |
4089 | struct bgp_info *ri; | |
4090 | ||
4091 | table = bgp_vrf->rib[afi][safi]; | |
4092 | for (rn = bgp_table_top(table); rn; rn = bgp_route_next(rn)) { | |
4093 | /* Only care about "selected" routes - non-imported. */ | |
4094 | /* TODO: Support for AddPath for EVPN. */ | |
4095 | for (ri = rn->info; ri; ri = ri->next) { | |
4096 | if (CHECK_FLAG(ri->flags, BGP_INFO_SELECTED) | |
4097 | && (!ri->extra || !ri->extra->parent)) { | |
4098 | bgp_evpn_withdraw_type5_route(bgp_vrf, &rn->p, | |
4099 | afi, safi); | |
4100 | break; | |
4101 | } | |
4102 | } | |
4103 | } | |
4104 | } | |
4105 | ||
4106 | /* | |
4107 | * Advertise IP prefix as type-5 route. The afi/safi and src_attr passed | |
4108 | * to this function correspond to those of the source IP prefix (best | |
4109 | * path in the case of the attr. In the case of a local prefix (when we | |
4110 | * are advertising local subnets), the src_attr will be NULL. | |
4111 | */ | |
4112 | void bgp_evpn_advertise_type5_route(struct bgp *bgp_vrf, struct prefix *p, | |
4113 | struct attr *src_attr, afi_t afi, | |
4114 | safi_t safi) | |
4115 | { | |
4116 | int ret = 0; | |
4117 | struct prefix_evpn evp; | |
4118 | char buf[PREFIX_STRLEN]; | |
4119 | ||
4120 | build_type5_prefix_from_ip_prefix(&evp, p); | |
4121 | ret = update_evpn_type5_route(bgp_vrf, &evp, src_attr); | |
4122 | if (ret) | |
4123 | flog_err(EC_BGP_EVPN_ROUTE_CREATE, | |
4124 | "%u: Failed to create type-5 route for prefix %s", | |
4125 | bgp_vrf->vrf_id, prefix2str(p, buf, sizeof(buf))); | |
4126 | } | |
4127 | ||
4128 | /* Inject all prefixes of a particular address-family (currently, IPv4 or | |
4129 | * IPv6 unicast) into EVPN as type-5 routes. This is invoked when the | |
4130 | * advertisement is enabled. | |
4131 | */ | |
4132 | void bgp_evpn_advertise_type5_routes(struct bgp *bgp_vrf, afi_t afi, | |
4133 | safi_t safi) | |
4134 | { | |
4135 | struct bgp_table *table = NULL; | |
4136 | struct bgp_node *rn = NULL; | |
4137 | struct bgp_info *ri; | |
4138 | ||
4139 | table = bgp_vrf->rib[afi][safi]; | |
4140 | for (rn = bgp_table_top(table); rn; rn = bgp_route_next(rn)) { | |
4141 | /* Need to identify the "selected" route entry to use its | |
4142 | * attribute. Also, we only consider "non-imported" routes. | |
4143 | * TODO: Support for AddPath for EVPN. | |
4144 | */ | |
4145 | for (ri = rn->info; ri; ri = ri->next) { | |
4146 | if (CHECK_FLAG(ri->flags, BGP_INFO_SELECTED) | |
4147 | && (!ri->extra || !ri->extra->parent)) { | |
4148 | ||
4149 | /* apply the route-map */ | |
4150 | if (bgp_vrf->adv_cmd_rmap[afi][safi].map) { | |
4151 | int ret = 0; | |
4152 | ||
4153 | ret = route_map_apply( | |
4154 | bgp_vrf->adv_cmd_rmap[afi][safi] | |
4155 | .map, | |
4156 | &rn->p, RMAP_BGP, ri); | |
4157 | if (ret == RMAP_DENYMATCH) | |
4158 | continue; | |
4159 | } | |
4160 | bgp_evpn_advertise_type5_route( | |
4161 | bgp_vrf, &rn->p, ri->attr, afi, safi); | |
4162 | break; | |
4163 | } | |
4164 | } | |
4165 | } | |
4166 | } | |
4167 | ||
4168 | void evpn_rt_delete_auto(struct bgp *bgp, vni_t vni, struct list *rtl) | |
4169 | { | |
4170 | struct listnode *node, *nnode, *node_to_del; | |
4171 | struct ecommunity *ecom, *ecom_auto; | |
4172 | struct ecommunity_val eval; | |
4173 | ||
4174 | if (bgp->advertise_autort_rfc8365) | |
4175 | vni |= EVPN_AUTORT_VXLAN; | |
4176 | encode_route_target_as((bgp->as & 0xFFFF), vni, &eval); | |
4177 | ||
4178 | ecom_auto = ecommunity_new(); | |
4179 | ecommunity_add_val(ecom_auto, &eval); | |
4180 | node_to_del = NULL; | |
4181 | ||
4182 | for (ALL_LIST_ELEMENTS(rtl, node, nnode, ecom)) { | |
4183 | if (ecommunity_match(ecom, ecom_auto)) { | |
4184 | ecommunity_free(&ecom); | |
4185 | node_to_del = node; | |
4186 | } | |
4187 | } | |
4188 | ||
4189 | if (node_to_del) | |
4190 | list_delete_node(rtl, node_to_del); | |
4191 | ||
4192 | ecommunity_free(&ecom_auto); | |
4193 | } | |
4194 | ||
4195 | void bgp_evpn_configure_import_rt_for_vrf(struct bgp *bgp_vrf, | |
4196 | struct ecommunity *ecomadd) | |
4197 | { | |
4198 | /* uninstall routes from vrf */ | |
4199 | uninstall_routes_for_vrf(bgp_vrf); | |
4200 | ||
4201 | /* Cleanup the RT to VRF mapping */ | |
4202 | bgp_evpn_unmap_vrf_from_its_rts(bgp_vrf); | |
4203 | ||
4204 | /* Remove auto generated RT */ | |
4205 | evpn_auto_rt_import_delete_for_vrf(bgp_vrf); | |
4206 | ||
4207 | /* Add the newly configured RT to RT list */ | |
4208 | listnode_add_sort(bgp_vrf->vrf_import_rtl, ecomadd); | |
4209 | SET_FLAG(bgp_vrf->vrf_flags, BGP_VRF_IMPORT_RT_CFGD); | |
4210 | ||
4211 | /* map VRF to its RTs */ | |
4212 | bgp_evpn_map_vrf_to_its_rts(bgp_vrf); | |
4213 | ||
4214 | /* install routes matching the new VRF */ | |
4215 | install_routes_for_vrf(bgp_vrf); | |
4216 | } | |
4217 | ||
4218 | void bgp_evpn_unconfigure_import_rt_for_vrf(struct bgp *bgp_vrf, | |
4219 | struct ecommunity *ecomdel) | |
4220 | { | |
4221 | struct listnode *node = NULL, *nnode = NULL, *node_to_del = NULL; | |
4222 | struct ecommunity *ecom = NULL; | |
4223 | ||
4224 | /* uninstall routes from vrf */ | |
4225 | uninstall_routes_for_vrf(bgp_vrf); | |
4226 | ||
4227 | /* Cleanup the RT to VRF mapping */ | |
4228 | bgp_evpn_unmap_vrf_from_its_rts(bgp_vrf); | |
4229 | ||
4230 | /* remove the RT from the RT list */ | |
4231 | for (ALL_LIST_ELEMENTS(bgp_vrf->vrf_import_rtl, node, nnode, ecom)) { | |
4232 | if (ecommunity_match(ecom, ecomdel)) { | |
4233 | ecommunity_free(&ecom); | |
4234 | node_to_del = node; | |
4235 | break; | |
4236 | } | |
4237 | } | |
4238 | ||
4239 | if (node_to_del) | |
4240 | list_delete_node(bgp_vrf->vrf_import_rtl, node_to_del); | |
4241 | ||
4242 | assert(bgp_vrf->vrf_import_rtl); | |
4243 | /* fallback to auto import rt, if this was the last RT */ | |
4244 | if (bgp_vrf->vrf_import_rtl && list_isempty(bgp_vrf->vrf_import_rtl)) { | |
4245 | UNSET_FLAG(bgp_vrf->vrf_flags, BGP_VRF_IMPORT_RT_CFGD); | |
4246 | evpn_auto_rt_import_add_for_vrf(bgp_vrf); | |
4247 | } | |
4248 | ||
4249 | /* map VRFs to its RTs */ | |
4250 | bgp_evpn_map_vrf_to_its_rts(bgp_vrf); | |
4251 | ||
4252 | /* install routes matching this new RT */ | |
4253 | install_routes_for_vrf(bgp_vrf); | |
4254 | } | |
4255 | ||
4256 | void bgp_evpn_configure_export_rt_for_vrf(struct bgp *bgp_vrf, | |
4257 | struct ecommunity *ecomadd) | |
4258 | { | |
4259 | /* remove auto-generated RT */ | |
4260 | evpn_auto_rt_export_delete_for_vrf(bgp_vrf); | |
4261 | ||
4262 | /* Add the new RT to the RT list */ | |
4263 | listnode_add_sort(bgp_vrf->vrf_export_rtl, ecomadd); | |
4264 | SET_FLAG(bgp_vrf->vrf_flags, BGP_VRF_EXPORT_RT_CFGD); | |
4265 | ||
4266 | bgp_evpn_handle_export_rt_change_for_vrf(bgp_vrf); | |
4267 | } | |
4268 | ||
4269 | void bgp_evpn_unconfigure_export_rt_for_vrf(struct bgp *bgp_vrf, | |
4270 | struct ecommunity *ecomdel) | |
4271 | { | |
4272 | struct listnode *node = NULL, *nnode = NULL, *node_to_del = NULL; | |
4273 | struct ecommunity *ecom = NULL; | |
4274 | ||
4275 | /* Remove the RT from the RT list */ | |
4276 | for (ALL_LIST_ELEMENTS(bgp_vrf->vrf_export_rtl, node, nnode, ecom)) { | |
4277 | if (ecommunity_match(ecom, ecomdel)) { | |
4278 | ecommunity_free(&ecom); | |
4279 | node_to_del = node; | |
4280 | break; | |
4281 | } | |
4282 | } | |
4283 | ||
4284 | if (node_to_del) | |
4285 | list_delete_node(bgp_vrf->vrf_export_rtl, node_to_del); | |
4286 | ||
4287 | /* | |
4288 | * Temporary assert to make SA happy. | |
4289 | * The ALL_LIST_ELEMENTS macro above has a NULL check | |
4290 | * which means that SA is going to complain about | |
4291 | * the list_isempty call, which doesn't NULL check. | |
4292 | * So until we get this situation cleaned up, here | |
4293 | * we are. | |
4294 | */ | |
4295 | assert(bgp_vrf->vrf_export_rtl); | |
4296 | ||
4297 | /* fall back to auto-generated RT if this was the last RT */ | |
4298 | if (list_isempty(bgp_vrf->vrf_export_rtl)) { | |
4299 | UNSET_FLAG(bgp_vrf->vrf_flags, BGP_VRF_EXPORT_RT_CFGD); | |
4300 | evpn_auto_rt_export_add_for_vrf(bgp_vrf); | |
4301 | } | |
4302 | ||
4303 | bgp_evpn_handle_export_rt_change_for_vrf(bgp_vrf); | |
4304 | } | |
4305 | ||
4306 | /* | |
4307 | * Handle change to BGP router id. This is invoked twice by the change | |
4308 | * handler, first before the router id has been changed and then after | |
4309 | * the router id has been changed. The first invocation will result in | |
4310 | * local routes for all VNIs/VRF being deleted and withdrawn and the next | |
4311 | * will result in the routes being re-advertised. | |
4312 | */ | |
4313 | void bgp_evpn_handle_router_id_update(struct bgp *bgp, int withdraw) | |
4314 | { | |
4315 | if (withdraw) { | |
4316 | ||
4317 | /* delete and withdraw all the type-5 routes | |
4318 | stored in the global table for this vrf | |
4319 | */ | |
4320 | withdraw_router_id_vrf(bgp); | |
4321 | ||
4322 | /* delete all the VNI routes (type-2/type-3) routes for all the | |
4323 | * L2-VNIs | |
4324 | */ | |
4325 | hash_iterate(bgp->vnihash, | |
4326 | (void (*)(struct hash_backet *, | |
4327 | void *))withdraw_router_id_vni, | |
4328 | bgp); | |
4329 | } else { | |
4330 | ||
4331 | /* advertise all routes in the vrf as type-5 routes with the new | |
4332 | * RD | |
4333 | */ | |
4334 | update_router_id_vrf(bgp); | |
4335 | ||
4336 | /* advertise all the VNI routes (type-2/type-3) routes with the | |
4337 | * new RD | |
4338 | */ | |
4339 | hash_iterate(bgp->vnihash, | |
4340 | (void (*)(struct hash_backet *, | |
4341 | void *))update_router_id_vni, | |
4342 | bgp); | |
4343 | } | |
4344 | } | |
4345 | ||
4346 | /* | |
4347 | * Handle change to auto-RT algorithm - update and advertise local routes. | |
4348 | */ | |
4349 | void bgp_evpn_handle_autort_change(struct bgp *bgp) | |
4350 | { | |
4351 | hash_iterate(bgp->vnihash, | |
4352 | (void (*)(struct hash_backet *, | |
4353 | void*))update_autort_vni, | |
4354 | bgp); | |
4355 | } | |
4356 | ||
4357 | /* | |
4358 | * Handle change to export RT - update and advertise local routes. | |
4359 | */ | |
4360 | int bgp_evpn_handle_export_rt_change(struct bgp *bgp, struct bgpevpn *vpn) | |
4361 | { | |
4362 | return update_routes_for_vni(bgp, vpn); | |
4363 | } | |
4364 | ||
4365 | void bgp_evpn_handle_vrf_rd_change(struct bgp *bgp_vrf, int withdraw) | |
4366 | { | |
4367 | if (withdraw) | |
4368 | delete_withdraw_vrf_routes(bgp_vrf); | |
4369 | else | |
4370 | update_advertise_vrf_routes(bgp_vrf); | |
4371 | } | |
4372 | ||
4373 | /* | |
4374 | * Handle change to RD. This is invoked twice by the change handler, | |
4375 | * first before the RD has been changed and then after the RD has | |
4376 | * been changed. The first invocation will result in local routes | |
4377 | * of this VNI being deleted and withdrawn and the next will result | |
4378 | * in the routes being re-advertised. | |
4379 | */ | |
4380 | void bgp_evpn_handle_rd_change(struct bgp *bgp, struct bgpevpn *vpn, | |
4381 | int withdraw) | |
4382 | { | |
4383 | if (withdraw) | |
4384 | delete_withdraw_vni_routes(bgp, vpn); | |
4385 | else | |
4386 | update_advertise_vni_routes(bgp, vpn); | |
4387 | } | |
4388 | ||
4389 | /* | |
4390 | * Install routes for this VNI. Invoked upon change to Import RT. | |
4391 | */ | |
4392 | int bgp_evpn_install_routes(struct bgp *bgp, struct bgpevpn *vpn) | |
4393 | { | |
4394 | return install_routes_for_vni(bgp, vpn); | |
4395 | } | |
4396 | ||
4397 | /* | |
4398 | * Uninstall all routes installed for this VNI. Invoked upon change | |
4399 | * to Import RT. | |
4400 | */ | |
4401 | int bgp_evpn_uninstall_routes(struct bgp *bgp, struct bgpevpn *vpn) | |
4402 | { | |
4403 | return uninstall_routes_for_vni(bgp, vpn); | |
4404 | } | |
4405 | ||
4406 | /* | |
4407 | * TODO: Hardcoded for a maximum of 2 VNIs right now | |
4408 | */ | |
4409 | char *bgp_evpn_label2str(mpls_label_t *label, uint32_t num_labels, char *buf, | |
4410 | int len) | |
4411 | { | |
4412 | vni_t vni1, vni2; | |
4413 | ||
4414 | vni1 = label2vni(label); | |
4415 | if (num_labels == 2) { | |
4416 | vni2 = label2vni(label + 1); | |
4417 | snprintf(buf, len, "%u/%u", vni1, vni2); | |
4418 | } else | |
4419 | snprintf(buf, len, "%u", vni1); | |
4420 | return buf; | |
4421 | } | |
4422 | ||
4423 | /* | |
4424 | * Function to convert evpn route to json format. | |
4425 | * NOTE: We don't use prefix2str as the output here is a bit different. | |
4426 | */ | |
4427 | void bgp_evpn_route2json(struct prefix_evpn *p, json_object *json) | |
4428 | { | |
4429 | char buf1[ETHER_ADDR_STRLEN]; | |
4430 | char buf2[PREFIX2STR_BUFFER]; | |
4431 | ||
4432 | if (!json) | |
4433 | return; | |
4434 | ||
4435 | if (p->prefix.route_type == BGP_EVPN_IMET_ROUTE) { | |
4436 | json_object_int_add(json, "routeType", p->prefix.route_type); | |
4437 | json_object_int_add(json, "ethTag", | |
4438 | p->prefix.imet_addr.eth_tag); | |
4439 | json_object_int_add(json, "ipLen", | |
4440 | is_evpn_prefix_ipaddr_v4(p) | |
4441 | ? IPV4_MAX_BITLEN | |
4442 | : IPV6_MAX_BITLEN); | |
4443 | json_object_string_add(json, "ip", | |
4444 | inet_ntoa(p->prefix.imet_addr.ip.ipaddr_v4)); | |
4445 | } else if (p->prefix.route_type == BGP_EVPN_MAC_IP_ROUTE) { | |
4446 | if (is_evpn_prefix_ipaddr_none(p)) { | |
4447 | json_object_int_add(json, "routeType", | |
4448 | p->prefix.route_type); | |
4449 | json_object_int_add(json, "ethTag", | |
4450 | p->prefix.macip_addr.eth_tag); | |
4451 | json_object_int_add(json, "macLen", 8 * ETH_ALEN); | |
4452 | json_object_string_add(json, "mac", | |
4453 | prefix_mac2str(&p->prefix.macip_addr.mac, | |
4454 | buf1, | |
4455 | sizeof(buf1))); | |
4456 | } else { | |
4457 | uint8_t family; | |
4458 | ||
4459 | family = is_evpn_prefix_ipaddr_v4(p) ? AF_INET | |
4460 | : AF_INET6; | |
4461 | ||
4462 | json_object_int_add(json, "routeType", | |
4463 | p->prefix.route_type); | |
4464 | json_object_int_add(json, "ethTag", | |
4465 | p->prefix.macip_addr.eth_tag); | |
4466 | json_object_int_add(json, "macLen", 8 * ETH_ALEN); | |
4467 | json_object_string_add(json, "mac", | |
4468 | prefix_mac2str(&p->prefix.macip_addr.mac, | |
4469 | buf1, | |
4470 | sizeof(buf1))); | |
4471 | json_object_int_add(json, "ipLen", | |
4472 | is_evpn_prefix_ipaddr_v4(p) | |
4473 | ? IPV4_MAX_BITLEN | |
4474 | : IPV6_MAX_BITLEN); | |
4475 | json_object_string_add( | |
4476 | json, "ip", | |
4477 | inet_ntop(family, | |
4478 | &p->prefix.macip_addr.ip.ip.addr, | |
4479 | buf2, | |
4480 | PREFIX2STR_BUFFER)); | |
4481 | } | |
4482 | } else { | |
4483 | /* Currently, this is to cater to other AF_ETHERNET code. */ | |
4484 | } | |
4485 | } | |
4486 | ||
4487 | /* | |
4488 | * Function to convert evpn route to string. | |
4489 | * NOTE: We don't use prefix2str as the output here is a bit different. | |
4490 | */ | |
4491 | char *bgp_evpn_route2str(struct prefix_evpn *p, char *buf, int len) | |
4492 | { | |
4493 | char buf1[ETHER_ADDR_STRLEN]; | |
4494 | char buf2[PREFIX2STR_BUFFER]; | |
4495 | char buf3[ESI_STR_LEN]; | |
4496 | ||
4497 | if (p->prefix.route_type == BGP_EVPN_IMET_ROUTE) { | |
4498 | snprintf(buf, len, "[%d]:[%d]:[%d]:[%s]", p->prefix.route_type, | |
4499 | p->prefix.imet_addr.eth_tag, | |
4500 | is_evpn_prefix_ipaddr_v4(p) ? IPV4_MAX_BITLEN | |
4501 | : IPV6_MAX_BITLEN, | |
4502 | inet_ntoa(p->prefix.imet_addr.ip.ipaddr_v4)); | |
4503 | } else if (p->prefix.route_type == BGP_EVPN_MAC_IP_ROUTE) { | |
4504 | if (is_evpn_prefix_ipaddr_none(p)) | |
4505 | snprintf(buf, len, "[%d]:[%d]:[%d]:[%s]", | |
4506 | p->prefix.route_type, | |
4507 | p->prefix.macip_addr.eth_tag, | |
4508 | 8 * ETH_ALEN, | |
4509 | prefix_mac2str(&p->prefix.macip_addr.mac, buf1, | |
4510 | sizeof(buf1))); | |
4511 | else { | |
4512 | uint8_t family; | |
4513 | ||
4514 | family = is_evpn_prefix_ipaddr_v4(p) ? AF_INET | |
4515 | : AF_INET6; | |
4516 | snprintf(buf, len, "[%d]:[%d]:[%d]:[%s]:[%d]:[%s]", | |
4517 | p->prefix.route_type, | |
4518 | p->prefix.macip_addr.eth_tag, | |
4519 | 8 * ETH_ALEN, | |
4520 | prefix_mac2str(&p->prefix.macip_addr.mac, buf1, | |
4521 | sizeof(buf1)), | |
4522 | family == AF_INET ? IPV4_MAX_BITLEN | |
4523 | : IPV6_MAX_BITLEN, | |
4524 | inet_ntop(family, | |
4525 | &p->prefix.macip_addr.ip.ip.addr, | |
4526 | buf2, | |
4527 | PREFIX2STR_BUFFER)); | |
4528 | } | |
4529 | } else if (p->prefix.route_type == BGP_EVPN_IP_PREFIX_ROUTE) { | |
4530 | snprintf(buf, len, "[%d]:[%d]:[%d]:[%s]", | |
4531 | p->prefix.route_type, | |
4532 | p->prefix.prefix_addr.eth_tag, | |
4533 | p->prefix.prefix_addr.ip_prefix_length, | |
4534 | is_evpn_prefix_ipaddr_v4(p) | |
4535 | ? inet_ntoa(p->prefix.prefix_addr.ip.ipaddr_v4) | |
4536 | : inet6_ntoa(p->prefix.prefix_addr.ip.ipaddr_v6)); | |
4537 | } else if (p->prefix.route_type == BGP_EVPN_ES_ROUTE) { | |
4538 | snprintf(buf, len, "[%d]:[%s]:[%d]:[%s]", | |
4539 | p->prefix.route_type, | |
4540 | esi_to_str(&p->prefix.es_addr.esi, buf3, sizeof(buf3)), | |
4541 | is_evpn_prefix_ipaddr_v4(p) ? IPV4_MAX_BITLEN | |
4542 | : IPV6_MAX_BITLEN, | |
4543 | inet_ntoa(p->prefix.es_addr.ip.ipaddr_v4)); | |
4544 | } else { | |
4545 | /* For EVPN route types not supported yet. */ | |
4546 | snprintf(buf, len, "(unsupported route type %d)", | |
4547 | p->prefix.route_type); | |
4548 | } | |
4549 | ||
4550 | return (buf); | |
4551 | } | |
4552 | ||
4553 | /* | |
4554 | * Encode EVPN prefix in Update (MP_REACH) | |
4555 | */ | |
4556 | void bgp_evpn_encode_prefix(struct stream *s, struct prefix *p, | |
4557 | struct prefix_rd *prd, mpls_label_t *label, | |
4558 | uint32_t num_labels, struct attr *attr, | |
4559 | int addpath_encode, uint32_t addpath_tx_id) | |
4560 | { | |
4561 | struct prefix_evpn *evp = (struct prefix_evpn *)p; | |
4562 | int len, ipa_len = 0; | |
4563 | ||
4564 | if (addpath_encode) | |
4565 | stream_putl(s, addpath_tx_id); | |
4566 | ||
4567 | /* Route type */ | |
4568 | stream_putc(s, evp->prefix.route_type); | |
4569 | ||
4570 | switch (evp->prefix.route_type) { | |
4571 | case BGP_EVPN_MAC_IP_ROUTE: | |
4572 | if (is_evpn_prefix_ipaddr_v4(evp)) | |
4573 | ipa_len = IPV4_MAX_BYTELEN; | |
4574 | else if (is_evpn_prefix_ipaddr_v6(evp)) | |
4575 | ipa_len = IPV6_MAX_BYTELEN; | |
4576 | /* RD, ESI, EthTag, MAC+len, IP len, [IP], 1 VNI */ | |
4577 | len = 8 + 10 + 4 + 1 + 6 + 1 + ipa_len + 3; | |
4578 | if (ipa_len && num_labels > 1) /* There are 2 VNIs */ | |
4579 | len += 3; | |
4580 | stream_putc(s, len); | |
4581 | stream_put(s, prd->val, 8); /* RD */ | |
4582 | if (attr) | |
4583 | stream_put(s, &attr->evpn_overlay.eth_s_id, ESI_LEN); | |
4584 | else | |
4585 | stream_put(s, 0, 10); | |
4586 | stream_putl(s, evp->prefix.macip_addr.eth_tag); /* Ethernet Tag ID */ | |
4587 | stream_putc(s, 8 * ETH_ALEN); /* Mac Addr Len - bits */ | |
4588 | stream_put(s, evp->prefix.macip_addr.mac.octet, 6); /* Mac Addr */ | |
4589 | stream_putc(s, 8 * ipa_len); /* IP address Length */ | |
4590 | if (ipa_len) /* IP */ | |
4591 | stream_put(s, &evp->prefix.macip_addr.ip.ip.addr, | |
4592 | ipa_len); | |
4593 | /* 1st label is the L2 VNI */ | |
4594 | stream_put(s, label, BGP_LABEL_BYTES); | |
4595 | /* Include 2nd label (L3 VNI) if advertising MAC+IP */ | |
4596 | if (ipa_len && num_labels > 1) | |
4597 | stream_put(s, label + 1, BGP_LABEL_BYTES); | |
4598 | break; | |
4599 | ||
4600 | case BGP_EVPN_IMET_ROUTE: | |
4601 | stream_putc(s, 17); // TODO: length - assumes IPv4 address | |
4602 | stream_put(s, prd->val, 8); /* RD */ | |
4603 | stream_putl(s, evp->prefix.imet_addr.eth_tag); /* Ethernet Tag ID */ | |
4604 | stream_putc(s, IPV4_MAX_BITLEN); /* IP address Length - bits */ | |
4605 | /* Originating Router's IP Addr */ | |
4606 | stream_put_in_addr(s, &evp->prefix.imet_addr.ip.ipaddr_v4); | |
4607 | break; | |
4608 | ||
4609 | case BGP_EVPN_ES_ROUTE: | |
4610 | stream_putc(s, 23); /* TODO: length: assumes ipv4 VTEP */ | |
4611 | stream_put(s, prd->val, 8); /* RD */ | |
4612 | stream_put(s, evp->prefix.es_addr.esi.val, 10); /* ESI */ | |
4613 | stream_putc(s, IPV4_MAX_BITLEN); /* IP address Length - bits */ | |
4614 | /* VTEP IP */ | |
4615 | stream_put_in_addr(s, &evp->prefix.es_addr.ip.ipaddr_v4); | |
4616 | break; | |
4617 | ||
4618 | case BGP_EVPN_IP_PREFIX_ROUTE: | |
4619 | /* TODO: AddPath support. */ | |
4620 | evpn_mpattr_encode_type5(s, p, prd, label, num_labels, attr); | |
4621 | break; | |
4622 | ||
4623 | default: | |
4624 | break; | |
4625 | } | |
4626 | } | |
4627 | ||
4628 | int bgp_nlri_parse_evpn(struct peer *peer, struct attr *attr, | |
4629 | struct bgp_nlri *packet, int withdraw) | |
4630 | { | |
4631 | uint8_t *pnt; | |
4632 | uint8_t *lim; | |
4633 | afi_t afi; | |
4634 | safi_t safi; | |
4635 | uint32_t addpath_id; | |
4636 | int addpath_encoded; | |
4637 | int psize = 0; | |
4638 | uint8_t rtype; | |
4639 | struct prefix p; | |
4640 | ||
4641 | /* Start processing the NLRI - there may be multiple in the MP_REACH */ | |
4642 | pnt = packet->nlri; | |
4643 | lim = pnt + packet->length; | |
4644 | afi = packet->afi; | |
4645 | safi = packet->safi; | |
4646 | addpath_id = 0; | |
4647 | ||
4648 | addpath_encoded = | |
4649 | (CHECK_FLAG(peer->af_cap[afi][safi], PEER_CAP_ADDPATH_AF_RX_ADV) | |
4650 | && CHECK_FLAG(peer->af_cap[afi][safi], | |
4651 | PEER_CAP_ADDPATH_AF_TX_RCV)); | |
4652 | ||
4653 | for (; pnt < lim; pnt += psize) { | |
4654 | /* Clear prefix structure. */ | |
4655 | memset(&p, 0, sizeof(struct prefix)); | |
4656 | ||
4657 | /* Deal with path-id if AddPath is supported. */ | |
4658 | if (addpath_encoded) { | |
4659 | /* When packet overflow occurs return immediately. */ | |
4660 | if (pnt + BGP_ADDPATH_ID_LEN > lim) | |
4661 | return -1; | |
4662 | ||
4663 | addpath_id = ntohl(*((uint32_t *)pnt)); | |
4664 | pnt += BGP_ADDPATH_ID_LEN; | |
4665 | } | |
4666 | ||
4667 | /* All EVPN NLRI types start with type and length. */ | |
4668 | if (pnt + 2 > lim) | |
4669 | return -1; | |
4670 | ||
4671 | rtype = *pnt++; | |
4672 | psize = *pnt++; | |
4673 | ||
4674 | /* When packet overflow occur return immediately. */ | |
4675 | if (pnt + psize > lim) | |
4676 | return -1; | |
4677 | ||
4678 | switch (rtype) { | |
4679 | case BGP_EVPN_MAC_IP_ROUTE: | |
4680 | if (process_type2_route(peer, afi, safi, | |
4681 | withdraw ? NULL : attr, pnt, | |
4682 | psize, addpath_id)) { | |
4683 | flog_err( | |
4684 | EC_BGP_EVPN_FAIL, | |
4685 | "%u:%s - Error in processing EVPN type-2 NLRI size %d", | |
4686 | peer->bgp->vrf_id, peer->host, psize); | |
4687 | return -1; | |
4688 | } | |
4689 | break; | |
4690 | ||
4691 | case BGP_EVPN_IMET_ROUTE: | |
4692 | if (process_type3_route(peer, afi, safi, | |
4693 | withdraw ? NULL : attr, pnt, | |
4694 | psize, addpath_id)) { | |
4695 | flog_err( | |
4696 | EC_BGP_PKT_PROCESS, | |
4697 | "%u:%s - Error in processing EVPN type-3 NLRI size %d", | |
4698 | peer->bgp->vrf_id, peer->host, psize); | |
4699 | return -1; | |
4700 | } | |
4701 | break; | |
4702 | ||
4703 | case BGP_EVPN_ES_ROUTE: | |
4704 | if (process_type4_route(peer, afi, safi, | |
4705 | withdraw ? NULL : attr, pnt, | |
4706 | psize, addpath_id)) { | |
4707 | flog_err( | |
4708 | EC_BGP_PKT_PROCESS, | |
4709 | "%u:%s - Error in processing EVPN type-4 NLRI size %d", | |
4710 | peer->bgp->vrf_id, peer->host, psize); | |
4711 | return -1; | |
4712 | } | |
4713 | break; | |
4714 | ||
4715 | case BGP_EVPN_IP_PREFIX_ROUTE: | |
4716 | if (process_type5_route(peer, afi, safi, attr, pnt, | |
4717 | psize, addpath_id, withdraw)) { | |
4718 | flog_err( | |
4719 | EC_BGP_PKT_PROCESS, | |
4720 | "%u:%s - Error in processing EVPN type-5 NLRI size %d", | |
4721 | peer->bgp->vrf_id, peer->host, psize); | |
4722 | return -1; | |
4723 | } | |
4724 | break; | |
4725 | ||
4726 | default: | |
4727 | break; | |
4728 | } | |
4729 | } | |
4730 | ||
4731 | /* Packet length consistency check. */ | |
4732 | if (pnt != lim) | |
4733 | return -1; | |
4734 | ||
4735 | return 0; | |
4736 | } | |
4737 | ||
4738 | /* | |
4739 | * Map the RTs (configured or automatically derived) of a VRF to the VRF. | |
4740 | * The mapping will be used during route processing. | |
4741 | * bgp_def: default bgp instance | |
4742 | * bgp_vrf: specific bgp vrf instance on which RT is configured | |
4743 | */ | |
4744 | void bgp_evpn_map_vrf_to_its_rts(struct bgp *bgp_vrf) | |
4745 | { | |
4746 | int i = 0; | |
4747 | struct ecommunity_val *eval = NULL; | |
4748 | struct listnode *node = NULL, *nnode = NULL; | |
4749 | struct ecommunity *ecom = NULL; | |
4750 | ||
4751 | for (ALL_LIST_ELEMENTS(bgp_vrf->vrf_import_rtl, node, nnode, ecom)) { | |
4752 | for (i = 0; i < ecom->size; i++) { | |
4753 | eval = (struct ecommunity_val *)(ecom->val | |
4754 | + (i | |
4755 | * ECOMMUNITY_SIZE)); | |
4756 | map_vrf_to_rt(bgp_vrf, eval); | |
4757 | } | |
4758 | } | |
4759 | } | |
4760 | ||
4761 | /* | |
4762 | * Unmap the RTs (configured or automatically derived) of a VRF from the VRF. | |
4763 | */ | |
4764 | void bgp_evpn_unmap_vrf_from_its_rts(struct bgp *bgp_vrf) | |
4765 | { | |
4766 | int i; | |
4767 | struct ecommunity_val *eval; | |
4768 | struct listnode *node, *nnode; | |
4769 | struct ecommunity *ecom; | |
4770 | ||
4771 | for (ALL_LIST_ELEMENTS(bgp_vrf->vrf_import_rtl, node, nnode, ecom)) { | |
4772 | for (i = 0; i < ecom->size; i++) { | |
4773 | struct vrf_irt_node *irt; | |
4774 | struct ecommunity_val eval_tmp; | |
4775 | ||
4776 | eval = (struct ecommunity_val *)(ecom->val | |
4777 | + (i | |
4778 | * ECOMMUNITY_SIZE)); | |
4779 | /* If using "automatic" RT, we only care about the | |
4780 | * local-admin sub-field. | |
4781 | * This is to facilitate using VNI as the RT for EBGP | |
4782 | * peering too. | |
4783 | */ | |
4784 | memcpy(&eval_tmp, eval, ECOMMUNITY_SIZE); | |
4785 | if (!CHECK_FLAG(bgp_vrf->vrf_flags, | |
4786 | BGP_VRF_IMPORT_RT_CFGD)) | |
4787 | mask_ecom_global_admin(&eval_tmp, eval); | |
4788 | ||
4789 | irt = lookup_vrf_import_rt(&eval_tmp); | |
4790 | if (irt) | |
4791 | unmap_vrf_from_rt(bgp_vrf, irt); | |
4792 | } | |
4793 | } | |
4794 | } | |
4795 | ||
4796 | ||
4797 | /* | |
4798 | * Map the RTs (configured or automatically derived) of a VNI to the VNI. | |
4799 | * The mapping will be used during route processing. | |
4800 | */ | |
4801 | void bgp_evpn_map_vni_to_its_rts(struct bgp *bgp, struct bgpevpn *vpn) | |
4802 | { | |
4803 | int i; | |
4804 | struct ecommunity_val *eval; | |
4805 | struct listnode *node, *nnode; | |
4806 | struct ecommunity *ecom; | |
4807 | ||
4808 | for (ALL_LIST_ELEMENTS(vpn->import_rtl, node, nnode, ecom)) { | |
4809 | for (i = 0; i < ecom->size; i++) { | |
4810 | eval = (struct ecommunity_val *)(ecom->val | |
4811 | + (i | |
4812 | * ECOMMUNITY_SIZE)); | |
4813 | map_vni_to_rt(bgp, vpn, eval); | |
4814 | } | |
4815 | } | |
4816 | } | |
4817 | ||
4818 | /* | |
4819 | * Unmap the RTs (configured or automatically derived) of a VNI from the VNI. | |
4820 | */ | |
4821 | void bgp_evpn_unmap_vni_from_its_rts(struct bgp *bgp, struct bgpevpn *vpn) | |
4822 | { | |
4823 | int i; | |
4824 | struct ecommunity_val *eval; | |
4825 | struct listnode *node, *nnode; | |
4826 | struct ecommunity *ecom; | |
4827 | ||
4828 | for (ALL_LIST_ELEMENTS(vpn->import_rtl, node, nnode, ecom)) { | |
4829 | for (i = 0; i < ecom->size; i++) { | |
4830 | struct irt_node *irt; | |
4831 | struct ecommunity_val eval_tmp; | |
4832 | ||
4833 | eval = (struct ecommunity_val *)(ecom->val | |
4834 | + (i | |
4835 | * ECOMMUNITY_SIZE)); | |
4836 | /* If using "automatic" RT, we only care about the | |
4837 | * local-admin sub-field. | |
4838 | * This is to facilitate using VNI as the RT for EBGP | |
4839 | * peering too. | |
4840 | */ | |
4841 | memcpy(&eval_tmp, eval, ECOMMUNITY_SIZE); | |
4842 | if (!is_import_rt_configured(vpn)) | |
4843 | mask_ecom_global_admin(&eval_tmp, eval); | |
4844 | ||
4845 | irt = lookup_import_rt(bgp, &eval_tmp); | |
4846 | if (irt) | |
4847 | unmap_vni_from_rt(bgp, vpn, irt); | |
4848 | } | |
4849 | } | |
4850 | } | |
4851 | ||
4852 | /* | |
4853 | * Derive Import RT automatically for VNI and map VNI to RT. | |
4854 | * The mapping will be used during route processing. | |
4855 | */ | |
4856 | void bgp_evpn_derive_auto_rt_import(struct bgp *bgp, struct bgpevpn *vpn) | |
4857 | { | |
4858 | form_auto_rt(bgp, vpn->vni, vpn->import_rtl); | |
4859 | UNSET_FLAG(vpn->flags, VNI_FLAG_IMPRT_CFGD); | |
4860 | ||
4861 | /* Map RT to VNI */ | |
4862 | bgp_evpn_map_vni_to_its_rts(bgp, vpn); | |
4863 | } | |
4864 | ||
4865 | /* | |
4866 | * Derive Export RT automatically for VNI. | |
4867 | */ | |
4868 | void bgp_evpn_derive_auto_rt_export(struct bgp *bgp, struct bgpevpn *vpn) | |
4869 | { | |
4870 | form_auto_rt(bgp, vpn->vni, vpn->export_rtl); | |
4871 | UNSET_FLAG(vpn->flags, VNI_FLAG_EXPRT_CFGD); | |
4872 | } | |
4873 | ||
4874 | /* | |
4875 | * Derive RD automatically for VNI using passed information - it | |
4876 | * is of the form RouterId:unique-id-for-vni. | |
4877 | */ | |
4878 | void bgp_evpn_derive_auto_rd_for_vrf(struct bgp *bgp) | |
4879 | { | |
4880 | form_auto_rd(bgp->router_id, bgp->vrf_rd_id, &bgp->vrf_prd); | |
4881 | } | |
4882 | ||
4883 | /* | |
4884 | * Derive RD automatically for VNI using passed information - it | |
4885 | * is of the form RouterId:unique-id-for-vni. | |
4886 | */ | |
4887 | void bgp_evpn_derive_auto_rd(struct bgp *bgp, struct bgpevpn *vpn) | |
4888 | { | |
4889 | char buf[100]; | |
4890 | ||
4891 | vpn->prd.family = AF_UNSPEC; | |
4892 | vpn->prd.prefixlen = 64; | |
4893 | sprintf(buf, "%s:%hu", inet_ntoa(bgp->router_id), vpn->rd_id); | |
4894 | (void)str2prefix_rd(buf, &vpn->prd); | |
4895 | UNSET_FLAG(vpn->flags, VNI_FLAG_RD_CFGD); | |
4896 | } | |
4897 | ||
4898 | /* | |
4899 | * Lookup L3-VNI | |
4900 | */ | |
4901 | bool bgp_evpn_lookup_l3vni_l2vni_table(vni_t vni) | |
4902 | { | |
4903 | struct list *inst = bm->bgp; | |
4904 | struct listnode *node; | |
4905 | struct bgp *bgp_vrf; | |
4906 | ||
4907 | for (ALL_LIST_ELEMENTS_RO(inst, node, bgp_vrf)) { | |
4908 | if (bgp_vrf->l3vni == vni) | |
4909 | return true; | |
4910 | } | |
4911 | ||
4912 | return false; | |
4913 | } | |
4914 | ||
4915 | /* | |
4916 | * Lookup VNI. | |
4917 | */ | |
4918 | struct bgpevpn *bgp_evpn_lookup_vni(struct bgp *bgp, vni_t vni) | |
4919 | { | |
4920 | struct bgpevpn *vpn; | |
4921 | struct bgpevpn tmp; | |
4922 | ||
4923 | memset(&tmp, 0, sizeof(struct bgpevpn)); | |
4924 | tmp.vni = vni; | |
4925 | vpn = hash_lookup(bgp->vnihash, &tmp); | |
4926 | return vpn; | |
4927 | } | |
4928 | ||
4929 | /* | |
4930 | * Create a new vpn - invoked upon configuration or zebra notification. | |
4931 | */ | |
4932 | struct bgpevpn *bgp_evpn_new(struct bgp *bgp, vni_t vni, | |
4933 | struct in_addr originator_ip, | |
4934 | vrf_id_t tenant_vrf_id) | |
4935 | { | |
4936 | struct bgpevpn *vpn; | |
4937 | ||
4938 | if (!bgp) | |
4939 | return NULL; | |
4940 | ||
4941 | vpn = XCALLOC(MTYPE_BGP_EVPN, sizeof(struct bgpevpn)); | |
4942 | if (!vpn) | |
4943 | return NULL; | |
4944 | ||
4945 | /* Set values - RD and RT set to defaults. */ | |
4946 | vpn->vni = vni; | |
4947 | vpn->originator_ip = originator_ip; | |
4948 | vpn->tenant_vrf_id = tenant_vrf_id; | |
4949 | ||
4950 | /* Initialize route-target import and export lists */ | |
4951 | vpn->import_rtl = list_new(); | |
4952 | vpn->import_rtl->cmp = (int (*)(void *, void *))evpn_route_target_cmp; | |
4953 | vpn->import_rtl->del = evpn_xxport_delete_ecomm; | |
4954 | vpn->export_rtl = list_new(); | |
4955 | vpn->export_rtl->cmp = (int (*)(void *, void *))evpn_route_target_cmp; | |
4956 | vpn->export_rtl->del = evpn_xxport_delete_ecomm; | |
4957 | bf_assign_index(bm->rd_idspace, vpn->rd_id); | |
4958 | derive_rd_rt_for_vni(bgp, vpn); | |
4959 | ||
4960 | /* Initialize EVPN route table. */ | |
4961 | vpn->route_table = bgp_table_init(bgp, AFI_L2VPN, SAFI_EVPN); | |
4962 | ||
4963 | /* Add to hash */ | |
4964 | if (!hash_get(bgp->vnihash, vpn, hash_alloc_intern)) { | |
4965 | XFREE(MTYPE_BGP_EVPN, vpn); | |
4966 | return NULL; | |
4967 | } | |
4968 | ||
4969 | /* add to l2vni list on corresponding vrf */ | |
4970 | bgpevpn_link_to_l3vni(vpn); | |
4971 | ||
4972 | QOBJ_REG(vpn, bgpevpn); | |
4973 | return vpn; | |
4974 | } | |
4975 | ||
4976 | /* | |
4977 | * Free a given VPN - called in multiple scenarios such as zebra | |
4978 | * notification, configuration being deleted, advertise-all-vni disabled etc. | |
4979 | * This just frees appropriate memory, caller should have taken other | |
4980 | * needed actions. | |
4981 | */ | |
4982 | void bgp_evpn_free(struct bgp *bgp, struct bgpevpn *vpn) | |
4983 | { | |
4984 | bgpevpn_unlink_from_l3vni(vpn); | |
4985 | bgp_table_unlock(vpn->route_table); | |
4986 | bgp_evpn_unmap_vni_from_its_rts(bgp, vpn); | |
4987 | list_delete_and_null(&vpn->import_rtl); | |
4988 | list_delete_and_null(&vpn->export_rtl); | |
4989 | bf_release_index(bm->rd_idspace, vpn->rd_id); | |
4990 | hash_release(bgp->vnihash, vpn); | |
4991 | QOBJ_UNREG(vpn); | |
4992 | XFREE(MTYPE_BGP_EVPN, vpn); | |
4993 | } | |
4994 | ||
4995 | /* | |
4996 | * Lookup local ES. | |
4997 | */ | |
4998 | struct evpnes *bgp_evpn_lookup_es(struct bgp *bgp, esi_t *esi) | |
4999 | { | |
5000 | struct evpnes *es; | |
5001 | struct evpnes tmp; | |
5002 | ||
5003 | memset(&tmp, 0, sizeof(struct evpnes)); | |
5004 | memcpy(&tmp.esi, esi, sizeof(esi_t)); | |
5005 | es = hash_lookup(bgp->esihash, &tmp); | |
5006 | return es; | |
5007 | } | |
5008 | ||
5009 | /* | |
5010 | * Create a new local es - invoked upon zebra notification. | |
5011 | */ | |
5012 | struct evpnes *bgp_evpn_es_new(struct bgp *bgp, | |
5013 | esi_t *esi, | |
5014 | struct ipaddr *originator_ip) | |
5015 | { | |
5016 | char buf[100]; | |
5017 | struct evpnes *es; | |
5018 | ||
5019 | if (!bgp) | |
5020 | return NULL; | |
5021 | ||
5022 | es = XCALLOC(MTYPE_BGP_EVPN_ES, sizeof(struct evpnes)); | |
5023 | if (!es) | |
5024 | return NULL; | |
5025 | ||
5026 | /* set the ESI and originator_ip */ | |
5027 | memcpy(&es->esi, esi, sizeof(esi_t)); | |
5028 | memcpy(&es->originator_ip, originator_ip, sizeof(struct ipaddr)); | |
5029 | ||
5030 | /* Initialise the VTEP list */ | |
5031 | es->vtep_list = list_new(); | |
5032 | es->vtep_list->cmp = (int (*)(void *, void *))evpn_vtep_ip_cmp; | |
5033 | ||
5034 | /* auto derive RD for this es */ | |
5035 | bf_assign_index(bm->rd_idspace, es->rd_id); | |
5036 | es->prd.family = AF_UNSPEC; | |
5037 | es->prd.prefixlen = 64; | |
5038 | sprintf(buf, "%s:%hu", inet_ntoa(bgp->router_id), es->rd_id); | |
5039 | (void)str2prefix_rd(buf, &es->prd); | |
5040 | ||
5041 | /* Initialize the ES route table */ | |
5042 | es->route_table = bgp_table_init(bgp, AFI_L2VPN, SAFI_EVPN); | |
5043 | ||
5044 | /* Add to hash */ | |
5045 | if (!hash_get(bgp->esihash, es, hash_alloc_intern)) { | |
5046 | XFREE(MTYPE_BGP_EVPN_ES, es); | |
5047 | return NULL; | |
5048 | } | |
5049 | ||
5050 | QOBJ_REG(es, evpnes); | |
5051 | return es; | |
5052 | } | |
5053 | ||
5054 | /* | |
5055 | * Free a given ES - | |
5056 | * This just frees appropriate memory, caller should have taken other | |
5057 | * needed actions. | |
5058 | */ | |
5059 | void bgp_evpn_es_free(struct bgp *bgp, struct evpnes *es) | |
5060 | { | |
5061 | list_delete_and_null(&es->vtep_list); | |
5062 | bgp_table_unlock(es->route_table); | |
5063 | bf_release_index(bm->rd_idspace, es->rd_id); | |
5064 | hash_release(bgp->esihash, es); | |
5065 | QOBJ_UNREG(es); | |
5066 | XFREE(MTYPE_BGP_EVPN_ES, es); | |
5067 | } | |
5068 | ||
5069 | /* | |
5070 | * Import evpn route from global table to VNI/VRF/ESI. | |
5071 | */ | |
5072 | int bgp_evpn_import_route(struct bgp *bgp, afi_t afi, safi_t safi, | |
5073 | struct prefix *p, struct bgp_info *ri) | |
5074 | { | |
5075 | return install_uninstall_evpn_route(bgp, afi, safi, p, ri, 1); | |
5076 | } | |
5077 | ||
5078 | /* | |
5079 | * Unimport evpn route from VNI/VRF/ESI. | |
5080 | */ | |
5081 | int bgp_evpn_unimport_route(struct bgp *bgp, afi_t afi, safi_t safi, | |
5082 | struct prefix *p, struct bgp_info *ri) | |
5083 | { | |
5084 | return install_uninstall_evpn_route(bgp, afi, safi, p, ri, 0); | |
5085 | } | |
5086 | ||
5087 | /* filter routes which have martian next hops */ | |
5088 | int bgp_filter_evpn_routes_upon_martian_nh_change(struct bgp *bgp) | |
5089 | { | |
5090 | afi_t afi; | |
5091 | safi_t safi; | |
5092 | struct bgp_node *rd_rn, *rn; | |
5093 | struct bgp_table *table; | |
5094 | struct bgp_info *ri; | |
5095 | ||
5096 | afi = AFI_L2VPN; | |
5097 | safi = SAFI_EVPN; | |
5098 | ||
5099 | /* Walk entire global routing table and evaluate routes which could be | |
5100 | * imported into this VPN. Note that we cannot just look at the routes | |
5101 | * for the VNI's RD - | |
5102 | * remote routes applicable for this VNI could have any RD. | |
5103 | */ | |
5104 | /* EVPN routes are a 2-level table. */ | |
5105 | for (rd_rn = bgp_table_top(bgp->rib[afi][safi]); rd_rn; | |
5106 | rd_rn = bgp_route_next(rd_rn)) { | |
5107 | table = (struct bgp_table *)(rd_rn->info); | |
5108 | if (!table) | |
5109 | continue; | |
5110 | ||
5111 | for (rn = bgp_table_top(table); rn; rn = bgp_route_next(rn)) { | |
5112 | ||
5113 | for (ri = rn->info; ri; ri = ri->next) { | |
5114 | ||
5115 | /* Consider "valid" remote routes applicable for | |
5116 | * this VNI. */ | |
5117 | if (!(ri->type == ZEBRA_ROUTE_BGP | |
5118 | && ri->sub_type == BGP_ROUTE_NORMAL)) | |
5119 | continue; | |
5120 | ||
5121 | if (bgp_nexthop_self(bgp, ri->attr->nexthop)) { | |
5122 | ||
5123 | char attr_str[BUFSIZ]; | |
5124 | char pbuf[PREFIX_STRLEN]; | |
5125 | ||
5126 | bgp_dump_attr(ri->attr, attr_str, | |
5127 | BUFSIZ); | |
5128 | ||
5129 | if (bgp_debug_update(ri->peer, &rn->p, | |
5130 | NULL, 1)) | |
5131 | zlog_debug( | |
5132 | "%u: prefix %s with attr %s - DENIED due to martian or self nexthop", | |
5133 | bgp->vrf_id, | |
5134 | prefix2str( | |
5135 | &rn->p, pbuf, | |
5136 | sizeof(pbuf)), | |
5137 | attr_str); | |
5138 | ||
5139 | bgp_evpn_unimport_route(bgp, afi, safi, | |
5140 | &rn->p, ri); | |
5141 | ||
5142 | bgp_rib_remove(rn, ri, ri->peer, afi, | |
5143 | safi); | |
5144 | } | |
5145 | } | |
5146 | } | |
5147 | } | |
5148 | ||
5149 | return 0; | |
5150 | } | |
5151 | ||
5152 | /* | |
5153 | * Handle del of a local MACIP. | |
5154 | */ | |
5155 | int bgp_evpn_local_macip_del(struct bgp *bgp, vni_t vni, struct ethaddr *mac, | |
5156 | struct ipaddr *ip) | |
5157 | { | |
5158 | struct bgpevpn *vpn; | |
5159 | struct prefix_evpn p; | |
5160 | ||
5161 | /* Lookup VNI hash - should exist. */ | |
5162 | vpn = bgp_evpn_lookup_vni(bgp, vni); | |
5163 | if (!vpn || !is_vni_live(vpn)) { | |
5164 | flog_warn(EC_BGP_EVPN_VPN_VNI, | |
5165 | "%u: VNI hash entry for VNI %u %s at MACIP DEL", | |
5166 | bgp->vrf_id, vni, vpn ? "not live" : "not found"); | |
5167 | return -1; | |
5168 | } | |
5169 | ||
5170 | /* Remove EVPN type-2 route and schedule for processing. */ | |
5171 | build_evpn_type2_prefix(&p, mac, ip); | |
5172 | delete_evpn_route(bgp, vpn, &p); | |
5173 | ||
5174 | return 0; | |
5175 | } | |
5176 | ||
5177 | /* | |
5178 | * Handle add of a local MACIP. | |
5179 | */ | |
5180 | int bgp_evpn_local_macip_add(struct bgp *bgp, vni_t vni, struct ethaddr *mac, | |
5181 | struct ipaddr *ip, uint8_t flags, uint32_t seq) | |
5182 | { | |
5183 | struct bgpevpn *vpn; | |
5184 | struct prefix_evpn p; | |
5185 | ||
5186 | /* Lookup VNI hash - should exist. */ | |
5187 | vpn = bgp_evpn_lookup_vni(bgp, vni); | |
5188 | if (!vpn || !is_vni_live(vpn)) { | |
5189 | flog_warn(EC_BGP_EVPN_VPN_VNI, | |
5190 | "%u: VNI hash entry for VNI %u %s at MACIP ADD", | |
5191 | bgp->vrf_id, vni, vpn ? "not live" : "not found"); | |
5192 | return -1; | |
5193 | } | |
5194 | ||
5195 | /* Create EVPN type-2 route and schedule for processing. */ | |
5196 | build_evpn_type2_prefix(&p, mac, ip); | |
5197 | if (update_evpn_route(bgp, vpn, &p, flags, seq)) { | |
5198 | char buf[ETHER_ADDR_STRLEN]; | |
5199 | char buf2[INET6_ADDRSTRLEN]; | |
5200 | ||
5201 | flog_err( | |
5202 | EC_BGP_EVPN_ROUTE_CREATE, | |
5203 | "%u:Failed to create Type-2 route, VNI %u %s MAC %s IP %s (flags: 0x%x)", | |
5204 | bgp->vrf_id, vpn->vni, | |
5205 | CHECK_FLAG(flags, ZEBRA_MACIP_TYPE_STICKY) | |
5206 | ? "sticky gateway" | |
5207 | : "", | |
5208 | prefix_mac2str(mac, buf, sizeof(buf)), | |
5209 | ipaddr2str(ip, buf2, sizeof(buf2)), flags); | |
5210 | return -1; | |
5211 | } | |
5212 | ||
5213 | return 0; | |
5214 | } | |
5215 | ||
5216 | static void link_l2vni_hash_to_l3vni(struct hash_backet *backet, | |
5217 | struct bgp *bgp_vrf) | |
5218 | { | |
5219 | struct bgpevpn *vpn = (struct bgpevpn *)backet->data; | |
5220 | struct bgp *bgp_def = NULL; | |
5221 | ||
5222 | bgp_def = bgp_get_default(); | |
5223 | assert(bgp_def); | |
5224 | ||
5225 | if (vpn->tenant_vrf_id == bgp_vrf->vrf_id) | |
5226 | bgpevpn_link_to_l3vni(vpn); | |
5227 | } | |
5228 | ||
5229 | int bgp_evpn_local_l3vni_add(vni_t l3vni, vrf_id_t vrf_id, struct ethaddr *rmac, | |
5230 | struct in_addr originator_ip, int filter) | |
5231 | { | |
5232 | struct bgp *bgp_vrf = NULL; /* bgp VRF instance */ | |
5233 | struct bgp *bgp_def = NULL; /* default bgp instance */ | |
5234 | struct listnode *node = NULL; | |
5235 | struct bgpevpn *vpn = NULL; | |
5236 | as_t as = 0; | |
5237 | ||
5238 | /* get the default instance - required to get the AS number for VRF | |
5239 | * auto-creatio | |
5240 | */ | |
5241 | bgp_def = bgp_get_default(); | |
5242 | if (!bgp_def) { | |
5243 | flog_err( | |
5244 | EC_BGP_NO_DFLT, | |
5245 | "Cannot process L3VNI %u ADD - default BGP instance not yet created", | |
5246 | l3vni); | |
5247 | return -1; | |
5248 | } | |
5249 | as = bgp_def->as; | |
5250 | ||
5251 | /* if the BGP vrf instance doesnt exist - create one */ | |
5252 | bgp_vrf = bgp_lookup_by_name(vrf_id_to_name(vrf_id)); | |
5253 | if (!bgp_vrf) { | |
5254 | ||
5255 | int ret = 0; | |
5256 | ||
5257 | ret = bgp_get(&bgp_vrf, &as, vrf_id_to_name(vrf_id), | |
5258 | BGP_INSTANCE_TYPE_VRF); | |
5259 | switch (ret) { | |
5260 | case BGP_ERR_MULTIPLE_INSTANCE_NOT_SET: | |
5261 | flog_err(EC_BGP_MULTI_INSTANCE, | |
5262 | "'bgp multiple-instance' not present\n"); | |
5263 | return -1; | |
5264 | case BGP_ERR_AS_MISMATCH: | |
5265 | flog_err(EC_BGP_EVPN_AS_MISMATCH, | |
5266 | "BGP is already running; AS is %u\n", as); | |
5267 | return -1; | |
5268 | case BGP_ERR_INSTANCE_MISMATCH: | |
5269 | flog_err(EC_BGP_EVPN_INSTANCE_MISMATCH, | |
5270 | "BGP instance name and AS number mismatch\n"); | |
5271 | return -1; | |
5272 | } | |
5273 | ||
5274 | /* mark as auto created */ | |
5275 | SET_FLAG(bgp_vrf->vrf_flags, BGP_VRF_AUTO); | |
5276 | } | |
5277 | ||
5278 | /* associate with l3vni */ | |
5279 | bgp_vrf->l3vni = l3vni; | |
5280 | ||
5281 | /* set the router mac - to be used in mac-ip routes for this vrf */ | |
5282 | memcpy(&bgp_vrf->rmac, rmac, sizeof(struct ethaddr)); | |
5283 | ||
5284 | /* set the originator ip */ | |
5285 | bgp_vrf->originator_ip = originator_ip; | |
5286 | ||
5287 | /* set the right filter - are we using l3vni only for prefix routes? */ | |
5288 | if (filter) | |
5289 | SET_FLAG(bgp_vrf->vrf_flags, BGP_VRF_L3VNI_PREFIX_ROUTES_ONLY); | |
5290 | ||
5291 | /* auto derive RD/RT */ | |
5292 | if (!CHECK_FLAG(bgp_vrf->vrf_flags, BGP_VRF_IMPORT_RT_CFGD)) | |
5293 | evpn_auto_rt_import_add_for_vrf(bgp_vrf); | |
5294 | if (!CHECK_FLAG(bgp_vrf->vrf_flags, BGP_VRF_EXPORT_RT_CFGD)) | |
5295 | evpn_auto_rt_export_add_for_vrf(bgp_vrf); | |
5296 | bgp_evpn_derive_auto_rd_for_vrf(bgp_vrf); | |
5297 | ||
5298 | /* link all corresponding l2vnis */ | |
5299 | hash_iterate(bgp_def->vnihash, | |
5300 | (void (*)(struct hash_backet *, | |
5301 | void *))link_l2vni_hash_to_l3vni, | |
5302 | bgp_vrf); | |
5303 | ||
5304 | /* Only update all corresponding type-2 routes if we are advertising two | |
5305 | * labels along with type-2 routes | |
5306 | */ | |
5307 | if (!filter) | |
5308 | for (ALL_LIST_ELEMENTS_RO(bgp_vrf->l2vnis, node, vpn)) | |
5309 | update_routes_for_vni(bgp_def, vpn); | |
5310 | ||
5311 | /* advertise type-5 routes if needed */ | |
5312 | update_advertise_vrf_routes(bgp_vrf); | |
5313 | ||
5314 | /* install all remote routes belonging to this l3vni into correspondng | |
5315 | * vrf */ | |
5316 | install_routes_for_vrf(bgp_vrf); | |
5317 | ||
5318 | return 0; | |
5319 | } | |
5320 | ||
5321 | int bgp_evpn_local_l3vni_del(vni_t l3vni, vrf_id_t vrf_id) | |
5322 | { | |
5323 | struct bgp *bgp_vrf = NULL; /* bgp vrf instance */ | |
5324 | struct bgp *bgp_def = NULL; /* default bgp instance */ | |
5325 | struct listnode *node = NULL; | |
5326 | struct listnode *next = NULL; | |
5327 | struct bgpevpn *vpn = NULL; | |
5328 | ||
5329 | bgp_vrf = bgp_lookup_by_vrf_id(vrf_id); | |
5330 | if (!bgp_vrf) { | |
5331 | flog_err( | |
5332 | EC_BGP_NO_DFLT, | |
5333 | "Cannot process L3VNI %u Del - Could not find BGP instance", | |
5334 | l3vni); | |
5335 | return -1; | |
5336 | } | |
5337 | ||
5338 | bgp_def = bgp_get_default(); | |
5339 | if (!bgp_def) { | |
5340 | flog_err( | |
5341 | EC_BGP_NO_DFLT, | |
5342 | "Cannot process L3VNI %u Del - Could not find default BGP instance", | |
5343 | l3vni); | |
5344 | return -1; | |
5345 | } | |
5346 | ||
5347 | /* Remove remote routes from BGT VRF even if BGP_VRF_AUTO is configured, | |
5348 | * bgp_delete would not remove/decrement bgp_info of the ip_prefix | |
5349 | * routes. This will uninstalling the routes from zebra and decremnt the | |
5350 | * bgp info count. | |
5351 | */ | |
5352 | uninstall_routes_for_vrf(bgp_vrf); | |
5353 | ||
5354 | /* delete/withdraw all type-5 routes */ | |
5355 | delete_withdraw_vrf_routes(bgp_vrf); | |
5356 | ||
5357 | /* remove the l3vni from vrf instance */ | |
5358 | bgp_vrf->l3vni = 0; | |
5359 | ||
5360 | /* remove the Rmac from the BGP vrf */ | |
5361 | memset(&bgp_vrf->rmac, 0, sizeof(struct ethaddr)); | |
5362 | ||
5363 | /* delete RD/RT */ | |
5364 | if (!list_isempty(bgp_vrf->vrf_import_rtl)) { | |
5365 | bgp_evpn_unmap_vrf_from_its_rts(bgp_vrf); | |
5366 | list_delete_all_node(bgp_vrf->vrf_import_rtl); | |
5367 | } | |
5368 | if (!list_isempty(bgp_vrf->vrf_export_rtl)) { | |
5369 | list_delete_all_node(bgp_vrf->vrf_export_rtl); | |
5370 | } | |
5371 | ||
5372 | /* update all corresponding local mac-ip routes */ | |
5373 | if (!CHECK_FLAG(bgp_vrf->vrf_flags, BGP_VRF_L3VNI_PREFIX_ROUTES_ONLY)) { | |
5374 | for (ALL_LIST_ELEMENTS_RO(bgp_vrf->l2vnis, node, vpn)) { | |
5375 | UNSET_FLAG(vpn->flags, VNI_FLAG_USE_TWO_LABELS); | |
5376 | update_routes_for_vni(bgp_def, vpn); | |
5377 | } | |
5378 | } | |
5379 | ||
5380 | /* If any L2VNIs point to this instance, unlink them. */ | |
5381 | for (ALL_LIST_ELEMENTS(bgp_vrf->l2vnis, node, next, vpn)) | |
5382 | bgpevpn_unlink_from_l3vni(vpn); | |
5383 | ||
5384 | /* Delete the instance if it was autocreated */ | |
5385 | if (CHECK_FLAG(bgp_vrf->vrf_flags, BGP_VRF_AUTO)) | |
5386 | bgp_delete(bgp_vrf); | |
5387 | ||
5388 | return 0; | |
5389 | } | |
5390 | ||
5391 | /* | |
5392 | * Handle del of a local VNI. | |
5393 | */ | |
5394 | int bgp_evpn_local_vni_del(struct bgp *bgp, vni_t vni) | |
5395 | { | |
5396 | struct bgpevpn *vpn; | |
5397 | ||
5398 | /* Locate VNI hash */ | |
5399 | vpn = bgp_evpn_lookup_vni(bgp, vni); | |
5400 | if (!vpn) { | |
5401 | if (bgp_debug_zebra(NULL)) | |
5402 | flog_warn( | |
5403 | EC_BGP_EVPN_VPN_VNI, | |
5404 | "%u: VNI hash entry for VNI %u not found at DEL", | |
5405 | bgp->vrf_id, vni); | |
5406 | return 0; | |
5407 | } | |
5408 | ||
5409 | /* Remove all local EVPN routes and schedule for processing (to | |
5410 | * withdraw from peers). | |
5411 | */ | |
5412 | delete_routes_for_vni(bgp, vpn); | |
5413 | ||
5414 | /* | |
5415 | * tunnel is no longer active, del tunnel ip address from tip_hash | |
5416 | */ | |
5417 | bgp_tip_del(bgp, &vpn->originator_ip); | |
5418 | ||
5419 | /* Clear "live" flag and see if hash needs to be freed. */ | |
5420 | UNSET_FLAG(vpn->flags, VNI_FLAG_LIVE); | |
5421 | if (!is_vni_configured(vpn)) | |
5422 | bgp_evpn_free(bgp, vpn); | |
5423 | ||
5424 | return 0; | |
5425 | } | |
5426 | ||
5427 | /* | |
5428 | * Handle add (or update) of a local VNI. The VNI changes we care | |
5429 | * about are for the local-tunnel-ip and the (tenant) VRF. | |
5430 | */ | |
5431 | int bgp_evpn_local_vni_add(struct bgp *bgp, vni_t vni, | |
5432 | struct in_addr originator_ip, vrf_id_t tenant_vrf_id) | |
5433 | { | |
5434 | struct bgpevpn *vpn; | |
5435 | struct prefix_evpn p; | |
5436 | ||
5437 | /* Lookup VNI. If present and no change, exit. */ | |
5438 | vpn = bgp_evpn_lookup_vni(bgp, vni); | |
5439 | if (vpn) { | |
5440 | ||
5441 | if (is_vni_live(vpn) | |
5442 | && IPV4_ADDR_SAME(&vpn->originator_ip, &originator_ip) | |
5443 | && vpn->tenant_vrf_id == tenant_vrf_id) | |
5444 | /* Probably some other param has changed that we don't | |
5445 | * care about. */ | |
5446 | return 0; | |
5447 | ||
5448 | /* Update tenant_vrf_id if it has changed. */ | |
5449 | if (vpn->tenant_vrf_id != tenant_vrf_id) { | |
5450 | bgpevpn_unlink_from_l3vni(vpn); | |
5451 | vpn->tenant_vrf_id = tenant_vrf_id; | |
5452 | bgpevpn_link_to_l3vni(vpn); | |
5453 | } | |
5454 | ||
5455 | /* If tunnel endpoint IP has changed, update (and delete prior | |
5456 | * type-3 route, if needed.) | |
5457 | */ | |
5458 | if (!IPV4_ADDR_SAME(&vpn->originator_ip, &originator_ip)) | |
5459 | handle_tunnel_ip_change(bgp, vpn, originator_ip); | |
5460 | ||
5461 | /* Update all routes with new endpoint IP and/or export RT | |
5462 | * for VRFs | |
5463 | */ | |
5464 | if (is_vni_live(vpn)) | |
5465 | update_routes_for_vni(bgp, vpn); | |
5466 | } | |
5467 | ||
5468 | /* Create or update as appropriate. */ | |
5469 | if (!vpn) { | |
5470 | vpn = bgp_evpn_new(bgp, vni, originator_ip, tenant_vrf_id); | |
5471 | if (!vpn) { | |
5472 | flog_err( | |
5473 | EC_BGP_VNI, | |
5474 | "%u: Failed to allocate VNI entry for VNI %u - at Add", | |
5475 | bgp->vrf_id, vni); | |
5476 | return -1; | |
5477 | } | |
5478 | } | |
5479 | ||
5480 | /* if the VNI is live already, there is nothing more to do */ | |
5481 | if (is_vni_live(vpn)) | |
5482 | return 0; | |
5483 | ||
5484 | /* Mark as "live" */ | |
5485 | SET_FLAG(vpn->flags, VNI_FLAG_LIVE); | |
5486 | ||
5487 | /* tunnel is now active, add tunnel-ip to db */ | |
5488 | bgp_tip_add(bgp, &originator_ip); | |
5489 | ||
5490 | /* filter routes as nexthop database has changed */ | |
5491 | bgp_filter_evpn_routes_upon_martian_nh_change(bgp); | |
5492 | ||
5493 | /* Create EVPN type-3 route and schedule for processing. */ | |
5494 | build_evpn_type3_prefix(&p, vpn->originator_ip); | |
5495 | if (update_evpn_route(bgp, vpn, &p, 0, 0)) { | |
5496 | flog_err(EC_BGP_EVPN_ROUTE_CREATE, | |
5497 | "%u: Type3 route creation failure for VNI %u", | |
5498 | bgp->vrf_id, vni); | |
5499 | return -1; | |
5500 | } | |
5501 | ||
5502 | /* If we have learnt and retained remote routes (VTEPs, MACs) for this | |
5503 | * VNI, | |
5504 | * install them. | |
5505 | */ | |
5506 | install_routes_for_vni(bgp, vpn); | |
5507 | ||
5508 | /* If we are advertising gateway mac-ip | |
5509 | It needs to be conveyed again to zebra */ | |
5510 | bgp_zebra_advertise_gw_macip(bgp, vpn->advertise_gw_macip, vpn->vni); | |
5511 | ||
5512 | return 0; | |
5513 | } | |
5514 | ||
5515 | /* | |
5516 | * bgp_evpn_local_es_del | |
5517 | */ | |
5518 | int bgp_evpn_local_es_del(struct bgp *bgp, | |
5519 | esi_t *esi, | |
5520 | struct ipaddr *originator_ip) | |
5521 | { | |
5522 | char buf[ESI_STR_LEN]; | |
5523 | struct evpnes *es = NULL; | |
5524 | ||
5525 | if (!bgp->esihash) { | |
5526 | flog_err(EC_BGP_ES_CREATE, "%u: ESI hash not yet created", | |
5527 | bgp->vrf_id); | |
5528 | return -1; | |
5529 | } | |
5530 | ||
5531 | /* Lookup ESI hash - should exist. */ | |
5532 | es = bgp_evpn_lookup_es(bgp, esi); | |
5533 | if (!es) { | |
5534 | flog_warn(EC_BGP_EVPN_ESI, | |
5535 | "%u: ESI hash entry for ESI %s at Local ES DEL", | |
5536 | bgp->vrf_id, esi_to_str(esi, buf, sizeof(buf))); | |
5537 | return -1; | |
5538 | } | |
5539 | ||
5540 | /* Delete all local EVPN ES routes from ESI table | |
5541 | * and schedule for processing (to withdraw from peers)) | |
5542 | */ | |
5543 | delete_routes_for_es(bgp, es); | |
5544 | ||
5545 | /* free the hash entry */ | |
5546 | bgp_evpn_es_free(bgp, es); | |
5547 | ||
5548 | return 0; | |
5549 | } | |
5550 | ||
5551 | /* | |
5552 | * bgp_evpn_local_es_add | |
5553 | */ | |
5554 | int bgp_evpn_local_es_add(struct bgp *bgp, | |
5555 | esi_t *esi, | |
5556 | struct ipaddr *originator_ip) | |
5557 | { | |
5558 | char buf[ESI_STR_LEN]; | |
5559 | struct evpnes *es = NULL; | |
5560 | struct prefix_evpn p; | |
5561 | ||
5562 | if (!bgp->esihash) { | |
5563 | flog_err(EC_BGP_ES_CREATE, "%u: ESI hash not yet created", | |
5564 | bgp->vrf_id); | |
5565 | return -1; | |
5566 | } | |
5567 | ||
5568 | /* create the new es */ | |
5569 | es = bgp_evpn_lookup_es(bgp, esi); | |
5570 | if (!es) { | |
5571 | es = bgp_evpn_es_new(bgp, esi, originator_ip); | |
5572 | if (!es) { | |
5573 | flog_err( | |
5574 | EC_BGP_ES_CREATE, | |
5575 | "%u: Failed to allocate ES entry for ESI %s - at Local ES Add", | |
5576 | bgp->vrf_id, esi_to_str(esi, buf, sizeof(buf))); | |
5577 | return -1; | |
5578 | } | |
5579 | } | |
5580 | UNSET_FLAG(es->flags, EVPNES_REMOTE); | |
5581 | SET_FLAG(es->flags, EVPNES_LOCAL); | |
5582 | ||
5583 | build_evpn_type4_prefix(&p, esi, originator_ip->ipaddr_v4); | |
5584 | if (update_evpn_type4_route(bgp, es, &p)) { | |
5585 | flog_err(EC_BGP_EVPN_ROUTE_CREATE, | |
5586 | "%u: Type4 route creation failure for ESI %s", | |
5587 | bgp->vrf_id, esi_to_str(esi, buf, sizeof(buf))); | |
5588 | return -1; | |
5589 | } | |
5590 | ||
5591 | /* import all remote ES routes in th ES table */ | |
5592 | install_routes_for_es(bgp, es); | |
5593 | ||
5594 | return 0; | |
5595 | } | |
5596 | ||
5597 | /* | |
5598 | * Cleanup EVPN information on disable - Need to delete and withdraw | |
5599 | * EVPN routes from peers. | |
5600 | */ | |
5601 | void bgp_evpn_cleanup_on_disable(struct bgp *bgp) | |
5602 | { | |
5603 | hash_iterate(bgp->vnihash, (void (*)(struct hash_backet *, | |
5604 | void *))cleanup_vni_on_disable, | |
5605 | bgp); | |
5606 | } | |
5607 | ||
5608 | /* | |
5609 | * Cleanup EVPN information - invoked at the time of bgpd exit or when the | |
5610 | * BGP instance (default) is being freed. | |
5611 | */ | |
5612 | void bgp_evpn_cleanup(struct bgp *bgp) | |
5613 | { | |
5614 | hash_iterate(bgp->vnihash, | |
5615 | (void (*)(struct hash_backet *, void *))free_vni_entry, | |
5616 | bgp); | |
5617 | ||
5618 | hash_free(bgp->import_rt_hash); | |
5619 | bgp->import_rt_hash = NULL; | |
5620 | ||
5621 | hash_free(bgp->vrf_import_rt_hash); | |
5622 | bgp->vrf_import_rt_hash = NULL; | |
5623 | ||
5624 | hash_free(bgp->vnihash); | |
5625 | bgp->vnihash = NULL; | |
5626 | if (bgp->esihash) | |
5627 | hash_free(bgp->esihash); | |
5628 | bgp->esihash = NULL; | |
5629 | ||
5630 | list_delete_and_null(&bgp->vrf_import_rtl); | |
5631 | list_delete_and_null(&bgp->vrf_export_rtl); | |
5632 | list_delete_and_null(&bgp->l2vnis); | |
5633 | } | |
5634 | ||
5635 | /* | |
5636 | * Initialization for EVPN | |
5637 | * Create | |
5638 | * VNI hash table | |
5639 | * hash for RT to VNI | |
5640 | */ | |
5641 | void bgp_evpn_init(struct bgp *bgp) | |
5642 | { | |
5643 | bgp->vnihash = | |
5644 | hash_create(vni_hash_key_make, vni_hash_cmp, "BGP VNI Hash"); | |
5645 | bgp->esihash = | |
5646 | hash_create(esi_hash_keymake, esi_cmp, | |
5647 | "BGP EVPN Local ESI Hash"); | |
5648 | bgp->import_rt_hash = | |
5649 | hash_create(import_rt_hash_key_make, import_rt_hash_cmp, | |
5650 | "BGP Import RT Hash"); | |
5651 | bgp->vrf_import_rt_hash = | |
5652 | hash_create(vrf_import_rt_hash_key_make, vrf_import_rt_hash_cmp, | |
5653 | "BGP VRF Import RT Hash"); | |
5654 | bgp->vrf_import_rtl = list_new(); | |
5655 | bgp->vrf_import_rtl->cmp = | |
5656 | (int (*)(void *, void *))evpn_route_target_cmp; | |
5657 | bgp->vrf_import_rtl->del = evpn_xxport_delete_ecomm; | |
5658 | bgp->vrf_export_rtl = list_new(); | |
5659 | bgp->vrf_export_rtl->cmp = | |
5660 | (int (*)(void *, void *))evpn_route_target_cmp; | |
5661 | bgp->vrf_export_rtl->del = evpn_xxport_delete_ecomm; | |
5662 | bgp->l2vnis = list_new(); | |
5663 | bgp->l2vnis->cmp = (int (*)(void *, void *))vni_hash_cmp; | |
5664 | } | |
5665 | ||
5666 | void bgp_evpn_vrf_delete(struct bgp *bgp_vrf) | |
5667 | { | |
5668 | bgp_evpn_unmap_vrf_from_its_rts(bgp_vrf); | |
5669 | } |