]> git.proxmox.com Git - mirror_frr.git/blob - bgpd/bgp_evpn.c
Merge pull request #12690 from opensourcerouting/feature/deny_merging_prs_with_freeze...
[mirror_frr.git] / bgpd / bgp_evpn.c
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 "lib/printfrr.h"
35
36 #include "bgpd/bgp_attr_evpn.h"
37 #include "bgpd/bgpd.h"
38 #include "bgpd/bgp_table.h"
39 #include "bgpd/bgp_route.h"
40 #include "bgpd/bgp_attr.h"
41 #include "bgpd/bgp_mplsvpn.h"
42 #include "bgpd/bgp_label.h"
43 #include "bgpd/bgp_evpn.h"
44 #include "bgpd/bgp_evpn_private.h"
45 #include "bgpd/bgp_evpn_mh.h"
46 #include "bgpd/bgp_ecommunity.h"
47 #include "bgpd/bgp_encap_types.h"
48 #include "bgpd/bgp_debug.h"
49 #include "bgpd/bgp_errors.h"
50 #include "bgpd/bgp_aspath.h"
51 #include "bgpd/bgp_zebra.h"
52 #include "bgpd/bgp_nexthop.h"
53 #include "bgpd/bgp_addpath.h"
54 #include "bgpd/bgp_mac.h"
55 #include "bgpd/bgp_vty.h"
56 #include "bgpd/bgp_nht.h"
57 #include "bgpd/bgp_trace.h"
58
59 /*
60 * Definitions and external declarations.
61 */
62 DEFINE_QOBJ_TYPE(bgpevpn);
63 DEFINE_QOBJ_TYPE(bgp_evpn_es);
64
65 DEFINE_MTYPE_STATIC(BGPD, VRF_ROUTE_TARGET, "L3 Route Target");
66
67 /*
68 * Static function declarations
69 */
70 static void bgp_evpn_remote_ip_hash_init(struct bgpevpn *evpn);
71 static void bgp_evpn_remote_ip_hash_destroy(struct bgpevpn *evpn);
72 static void bgp_evpn_remote_ip_hash_add(struct bgpevpn *vpn,
73 struct bgp_path_info *pi);
74 static void bgp_evpn_remote_ip_hash_del(struct bgpevpn *vpn,
75 struct bgp_path_info *pi);
76 static void bgp_evpn_remote_ip_hash_iterate(struct bgpevpn *vpn,
77 void (*func)(struct hash_bucket *,
78 void *),
79 void *arg);
80 static void bgp_evpn_link_to_vni_svi_hash(struct bgp *bgp, struct bgpevpn *vpn);
81 static void bgp_evpn_unlink_from_vni_svi_hash(struct bgp *bgp,
82 struct bgpevpn *vpn);
83 static unsigned int vni_svi_hash_key_make(const void *p);
84 static bool vni_svi_hash_cmp(const void *p1, const void *p2);
85 static void bgp_evpn_remote_ip_process_nexthops(struct bgpevpn *vpn,
86 struct ipaddr *addr,
87 bool resolve);
88 static void bgp_evpn_remote_ip_hash_link_nexthop(struct hash_bucket *bucket,
89 void *args);
90 static void bgp_evpn_remote_ip_hash_unlink_nexthop(struct hash_bucket *bucket,
91 void *args);
92 static struct in_addr zero_vtep_ip;
93
94 /*
95 * Private functions.
96 */
97
98 /*
99 * Make vni hash key.
100 */
101 static unsigned int vni_hash_key_make(const void *p)
102 {
103 const struct bgpevpn *vpn = p;
104 return (jhash_1word(vpn->vni, 0));
105 }
106
107 /*
108 * Comparison function for vni hash
109 */
110 static bool vni_hash_cmp(const void *p1, const void *p2)
111 {
112 const struct bgpevpn *vpn1 = p1;
113 const struct bgpevpn *vpn2 = p2;
114
115 return vpn1->vni == vpn2->vni;
116 }
117
118 int vni_list_cmp(void *p1, void *p2)
119 {
120 const struct bgpevpn *vpn1 = p1;
121 const struct bgpevpn *vpn2 = p2;
122
123 return vpn1->vni - vpn2->vni;
124 }
125
126 /*
127 * Make vrf import route target hash key.
128 */
129 static unsigned int vrf_import_rt_hash_key_make(const void *p)
130 {
131 const struct vrf_irt_node *irt = p;
132 const char *pnt = irt->rt.val;
133
134 return jhash(pnt, 8, 0x5abc1234);
135 }
136
137 /*
138 * Comparison function for vrf import rt hash
139 */
140 static bool vrf_import_rt_hash_cmp(const void *p1, const void *p2)
141 {
142 const struct vrf_irt_node *irt1 = p1;
143 const struct vrf_irt_node *irt2 = p2;
144
145 return (memcmp(irt1->rt.val, irt2->rt.val, ECOMMUNITY_SIZE) == 0);
146 }
147
148 /*
149 * Create a new vrf import_rt in evpn instance
150 */
151 static struct vrf_irt_node *vrf_import_rt_new(struct ecommunity_val *rt)
152 {
153 struct bgp *bgp_evpn = NULL;
154 struct vrf_irt_node *irt;
155
156 bgp_evpn = bgp_get_evpn();
157 if (!bgp_evpn) {
158 flog_err(EC_BGP_NO_DFLT,
159 "vrf import rt new - evpn instance not created yet");
160 return NULL;
161 }
162
163 irt = XCALLOC(MTYPE_BGP_EVPN_VRF_IMPORT_RT,
164 sizeof(struct vrf_irt_node));
165
166 irt->rt = *rt;
167 irt->vrfs = list_new();
168
169 /* Add to hash */
170 (void)hash_get(bgp_evpn->vrf_import_rt_hash, irt, hash_alloc_intern);
171
172 return irt;
173 }
174
175 /*
176 * Free the vrf import rt node
177 */
178 static void vrf_import_rt_free(struct vrf_irt_node *irt)
179 {
180 struct bgp *bgp_evpn = NULL;
181
182 bgp_evpn = bgp_get_evpn();
183 if (!bgp_evpn) {
184 flog_err(EC_BGP_NO_DFLT,
185 "vrf import rt free - evpn instance not created yet");
186 return;
187 }
188
189 hash_release(bgp_evpn->vrf_import_rt_hash, irt);
190 list_delete(&irt->vrfs);
191 XFREE(MTYPE_BGP_EVPN_VRF_IMPORT_RT, irt);
192 }
193
194 static void hash_vrf_import_rt_free(struct vrf_irt_node *irt)
195 {
196 XFREE(MTYPE_BGP_EVPN_VRF_IMPORT_RT, irt);
197 }
198
199 /*
200 * Function to lookup Import RT node - used to map a RT to set of
201 * VNIs importing routes with that RT.
202 */
203 static struct vrf_irt_node *lookup_vrf_import_rt(struct ecommunity_val *rt)
204 {
205 struct bgp *bgp_evpn = NULL;
206 struct vrf_irt_node *irt;
207 struct vrf_irt_node tmp;
208
209 bgp_evpn = bgp_get_evpn();
210 if (!bgp_evpn) {
211 flog_err(
212 EC_BGP_NO_DFLT,
213 "vrf import rt lookup - evpn instance not created yet");
214 return NULL;
215 }
216
217 memset(&tmp, 0, sizeof(tmp));
218 memcpy(&tmp.rt, rt, ECOMMUNITY_SIZE);
219 irt = hash_lookup(bgp_evpn->vrf_import_rt_hash, &tmp);
220 return irt;
221 }
222
223 /*
224 * Is specified VRF present on the RT's list of "importing" VRFs?
225 */
226 static int is_vrf_present_in_irt_vrfs(struct list *vrfs, struct bgp *bgp_vrf)
227 {
228 struct listnode *node = NULL, *nnode = NULL;
229 struct bgp *tmp_bgp_vrf = NULL;
230
231 for (ALL_LIST_ELEMENTS(vrfs, node, nnode, tmp_bgp_vrf)) {
232 if (tmp_bgp_vrf == bgp_vrf)
233 return 1;
234 }
235 return 0;
236 }
237
238 /*
239 * Make import route target hash key.
240 */
241 static unsigned int import_rt_hash_key_make(const void *p)
242 {
243 const struct irt_node *irt = p;
244 const char *pnt = irt->rt.val;
245
246 return jhash(pnt, 8, 0xdeadbeef);
247 }
248
249 /*
250 * Comparison function for import rt hash
251 */
252 static bool import_rt_hash_cmp(const void *p1, const void *p2)
253 {
254 const struct irt_node *irt1 = p1;
255 const struct irt_node *irt2 = p2;
256
257 return (memcmp(irt1->rt.val, irt2->rt.val, ECOMMUNITY_SIZE) == 0);
258 }
259
260 /*
261 * Create a new import_rt
262 */
263 static struct irt_node *import_rt_new(struct bgp *bgp,
264 struct ecommunity_val *rt)
265 {
266 struct irt_node *irt;
267
268 irt = XCALLOC(MTYPE_BGP_EVPN_IMPORT_RT, sizeof(struct irt_node));
269
270 irt->rt = *rt;
271 irt->vnis = list_new();
272
273 /* Add to hash */
274 (void)hash_get(bgp->import_rt_hash, irt, hash_alloc_intern);
275
276 return irt;
277 }
278
279 /*
280 * Free the import rt node
281 */
282 static void import_rt_free(struct bgp *bgp, struct irt_node *irt)
283 {
284 hash_release(bgp->import_rt_hash, irt);
285 list_delete(&irt->vnis);
286 XFREE(MTYPE_BGP_EVPN_IMPORT_RT, irt);
287 }
288
289 static void hash_import_rt_free(struct irt_node *irt)
290 {
291 XFREE(MTYPE_BGP_EVPN_IMPORT_RT, irt);
292 }
293
294 /*
295 * Function to lookup Import RT node - used to map a RT to set of
296 * VNIs importing routes with that RT.
297 */
298 static struct irt_node *lookup_import_rt(struct bgp *bgp,
299 struct ecommunity_val *rt)
300 {
301 struct irt_node *irt;
302 struct irt_node tmp;
303
304 memset(&tmp, 0, sizeof(tmp));
305 memcpy(&tmp.rt, rt, ECOMMUNITY_SIZE);
306 irt = hash_lookup(bgp->import_rt_hash, &tmp);
307 return irt;
308 }
309
310 /*
311 * Is specified VNI present on the RT's list of "importing" VNIs?
312 */
313 static int is_vni_present_in_irt_vnis(struct list *vnis, struct bgpevpn *vpn)
314 {
315 struct listnode *node, *nnode;
316 struct bgpevpn *tmp_vpn;
317
318 for (ALL_LIST_ELEMENTS(vnis, node, nnode, tmp_vpn)) {
319 if (tmp_vpn == vpn)
320 return 1;
321 }
322
323 return 0;
324 }
325
326 /*
327 * Compare Route Targets.
328 */
329 int bgp_evpn_route_target_cmp(struct ecommunity *ecom1,
330 struct ecommunity *ecom2)
331 {
332 if (ecom1 && !ecom2)
333 return -1;
334
335 if (!ecom1 && ecom2)
336 return 1;
337
338 if (!ecom1 && !ecom2)
339 return 0;
340
341 if (ecom1->str && !ecom2->str)
342 return -1;
343
344 if (!ecom1->str && ecom2->str)
345 return 1;
346
347 if (!ecom1->str && !ecom2->str)
348 return 0;
349
350 return strcmp(ecom1->str, ecom2->str);
351 }
352
353 /*
354 * Compare L3 Route Targets.
355 */
356 static int evpn_vrf_route_target_cmp(struct vrf_route_target *rt1,
357 struct vrf_route_target *rt2)
358 {
359 return bgp_evpn_route_target_cmp(rt1->ecom, rt2->ecom);
360 }
361
362 void bgp_evpn_xxport_delete_ecomm(void *val)
363 {
364 struct ecommunity *ecomm = val;
365 ecommunity_free(&ecomm);
366 }
367
368 /*
369 * Delete l3 Route Target.
370 */
371 static void evpn_vrf_rt_del(void *val)
372 {
373 struct vrf_route_target *l3rt = val;
374
375 ecommunity_free(&l3rt->ecom);
376
377 XFREE(MTYPE_VRF_ROUTE_TARGET, l3rt);
378 }
379
380 /*
381 * Allocate a new l3 Route Target.
382 */
383 static struct vrf_route_target *evpn_vrf_rt_new(struct ecommunity *ecom)
384 {
385 struct vrf_route_target *l3rt;
386
387 l3rt = XCALLOC(MTYPE_VRF_ROUTE_TARGET, sizeof(struct vrf_route_target));
388
389 l3rt->ecom = ecom;
390
391 return l3rt;
392 }
393
394 /*
395 * Mask off global-admin field of specified extended community (RT),
396 * just retain the local-admin field.
397 */
398 static inline void mask_ecom_global_admin(struct ecommunity_val *dst,
399 const struct ecommunity_val *src)
400 {
401 uint8_t type;
402
403 type = src->val[0];
404 dst->val[0] = 0;
405 if (type == ECOMMUNITY_ENCODE_AS) {
406 dst->val[2] = dst->val[3] = 0;
407 } else if (type == ECOMMUNITY_ENCODE_AS4
408 || type == ECOMMUNITY_ENCODE_IP) {
409 dst->val[2] = dst->val[3] = 0;
410 dst->val[4] = dst->val[5] = 0;
411 }
412 }
413
414 /*
415 * Converts the RT to Ecommunity Value and adjusts masking based
416 * on flags set for RT.
417 */
418 static void vrf_rt2ecom_val(struct ecommunity_val *to_eval,
419 const struct vrf_route_target *l3rt, int iter)
420 {
421 const struct ecommunity_val *eval;
422
423 eval = (const struct ecommunity_val *)(l3rt->ecom->val +
424 (iter * ECOMMUNITY_SIZE));
425 /* If using "automatic" or "wildcard *" RT,
426 * we only care about the local-admin sub-field.
427 * This is to facilitate using L3VNI(VRF-VNI)
428 * as the RT for EBGP peering too and simplify
429 * configurations by allowing any ASN via '*'.
430 */
431 memcpy(to_eval, eval, ECOMMUNITY_SIZE);
432
433 if (CHECK_FLAG(l3rt->flags, BGP_VRF_RT_AUTO) ||
434 CHECK_FLAG(l3rt->flags, BGP_VRF_RT_WILD))
435 mask_ecom_global_admin(to_eval, eval);
436 }
437
438 /*
439 * Map one RT to specified VRF.
440 * bgp_vrf = BGP vrf instance
441 */
442 static void map_vrf_to_rt(struct bgp *bgp_vrf, struct vrf_route_target *l3rt)
443 {
444 uint32_t i = 0;
445
446 for (i = 0; i < l3rt->ecom->size; i++) {
447 struct vrf_irt_node *irt = NULL;
448 struct ecommunity_val eval_tmp;
449
450 /* Adjust masking for value */
451 vrf_rt2ecom_val(&eval_tmp, l3rt, i);
452
453 irt = lookup_vrf_import_rt(&eval_tmp);
454
455 if (irt && is_vrf_present_in_irt_vrfs(irt->vrfs, bgp_vrf))
456 return; /* Already mapped. */
457
458 if (!irt)
459 irt = vrf_import_rt_new(&eval_tmp);
460
461 /* Add VRF to the list for this RT. */
462 listnode_add(irt->vrfs, bgp_vrf);
463 }
464 }
465
466 /*
467 * Unmap specified VRF from specified RT. If there are no other
468 * VRFs for this RT, then the RT hash is deleted.
469 * bgp_vrf: BGP VRF specific instance
470 */
471 static void unmap_vrf_from_rt(struct bgp *bgp_vrf,
472 struct vrf_route_target *l3rt)
473 {
474 uint32_t i;
475
476 for (i = 0; i < l3rt->ecom->size; i++) {
477 struct vrf_irt_node *irt;
478 struct ecommunity_val eval_tmp;
479
480 /* Adjust masking for value */
481 vrf_rt2ecom_val(&eval_tmp, l3rt, i);
482
483 irt = lookup_vrf_import_rt(&eval_tmp);
484
485 if (!irt)
486 return; /* Not mapped */
487
488 /* Delete VRF from list for this RT. */
489 listnode_delete(irt->vrfs, bgp_vrf);
490
491 if (!listnode_head(irt->vrfs))
492 vrf_import_rt_free(irt);
493 }
494 }
495
496 /*
497 * Map one RT to specified VNI.
498 */
499 static void map_vni_to_rt(struct bgp *bgp, struct bgpevpn *vpn,
500 struct ecommunity_val *eval)
501 {
502 struct irt_node *irt;
503 struct ecommunity_val eval_tmp;
504
505 /* If using "automatic" RT, we only care about the local-admin
506 * sub-field.
507 * This is to facilitate using VNI as the RT for EBGP peering too.
508 */
509 memcpy(&eval_tmp, eval, ECOMMUNITY_SIZE);
510 if (!is_import_rt_configured(vpn))
511 mask_ecom_global_admin(&eval_tmp, eval);
512
513 irt = lookup_import_rt(bgp, &eval_tmp);
514 if (irt)
515 if (is_vni_present_in_irt_vnis(irt->vnis, vpn))
516 /* Already mapped. */
517 return;
518
519 if (!irt)
520 irt = import_rt_new(bgp, &eval_tmp);
521
522 /* Add VNI to the hash list for this RT. */
523 listnode_add(irt->vnis, vpn);
524 }
525
526 /*
527 * Unmap specified VNI from specified RT. If there are no other
528 * VNIs for this RT, then the RT hash is deleted.
529 */
530 static void unmap_vni_from_rt(struct bgp *bgp, struct bgpevpn *vpn,
531 struct irt_node *irt)
532 {
533 /* Delete VNI from hash list for this RT. */
534 listnode_delete(irt->vnis, vpn);
535 if (!listnode_head(irt->vnis)) {
536 import_rt_free(bgp, irt);
537 }
538 }
539
540 static void bgp_evpn_get_rmac_nexthop(struct bgpevpn *vpn,
541 const struct prefix_evpn *p,
542 struct attr *attr, uint8_t flags)
543 {
544 struct bgp *bgp_vrf = vpn->bgp_vrf;
545
546 memset(&attr->rmac, 0, sizeof(struct ethaddr));
547 if (!bgp_vrf)
548 return;
549
550 if (p->prefix.route_type != BGP_EVPN_MAC_IP_ROUTE)
551 return;
552
553 /* Copy sys (pip) RMAC and PIP IP as nexthop
554 * in case of route is self MAC-IP,
555 * advertise-pip and advertise-svi-ip features
556 * are enabled.
557 * Otherwise, for all host MAC-IP route's
558 * copy anycast RMAC.
559 */
560 if (CHECK_FLAG(flags, BGP_EVPN_MACIP_TYPE_SVI_IP)
561 && bgp_vrf->evpn_info->advertise_pip &&
562 bgp_vrf->evpn_info->is_anycast_mac) {
563 /* copy sys rmac */
564 memcpy(&attr->rmac, &bgp_vrf->evpn_info->pip_rmac,
565 ETH_ALEN);
566 attr->nexthop = bgp_vrf->evpn_info->pip_ip;
567 attr->mp_nexthop_global_in =
568 bgp_vrf->evpn_info->pip_ip;
569 } else
570 memcpy(&attr->rmac, &bgp_vrf->rmac, ETH_ALEN);
571 }
572
573 /*
574 * Create RT extended community automatically from passed information:
575 * of the form AS:VNI.
576 * NOTE: We use only the lower 16 bits of the AS. This is sufficient as
577 * the need is to get a RT value that will be unique across different
578 * VNIs but the same across routers (in the same AS) for a particular
579 * VNI.
580 */
581 static void form_auto_rt(struct bgp *bgp, vni_t vni, struct list *rtl,
582 bool is_l3)
583 {
584 struct ecommunity_val eval;
585 struct ecommunity *ecomadd;
586 struct ecommunity *ecom;
587 struct vrf_route_target *l3rt;
588 struct vrf_route_target *newrt;
589 bool ecom_found = false;
590 struct listnode *node;
591
592 if (bgp->advertise_autort_rfc8365)
593 vni |= EVPN_AUTORT_VXLAN;
594 encode_route_target_as((bgp->as & 0xFFFF), vni, &eval);
595
596 ecomadd = ecommunity_new();
597 ecommunity_add_val(ecomadd, &eval, false, false);
598
599 if (is_l3) {
600 for (ALL_LIST_ELEMENTS_RO(rtl, node, l3rt))
601 if (ecommunity_cmp(ecomadd, l3rt->ecom)) {
602 ecom_found = true;
603 break;
604 }
605 } else {
606 for (ALL_LIST_ELEMENTS_RO(rtl, node, ecom))
607 if (ecommunity_cmp(ecomadd, ecom)) {
608 ecom_found = true;
609 break;
610 }
611 }
612
613 if (!ecom_found) {
614 if (is_l3) {
615 newrt = evpn_vrf_rt_new(ecomadd);
616 /* Label it as autoderived */
617 SET_FLAG(newrt->flags, BGP_VRF_RT_AUTO);
618 listnode_add_sort(rtl, newrt);
619 } else
620 listnode_add_sort(rtl, ecomadd);
621 } else
622 ecommunity_free(&ecomadd);
623 }
624
625 /*
626 * Derive RD and RT for a VNI automatically. Invoked at the time of
627 * creation of a VNI.
628 */
629 static void derive_rd_rt_for_vni(struct bgp *bgp, struct bgpevpn *vpn)
630 {
631 bgp_evpn_derive_auto_rd(bgp, vpn);
632 bgp_evpn_derive_auto_rt_import(bgp, vpn);
633 bgp_evpn_derive_auto_rt_export(bgp, vpn);
634 }
635
636 /*
637 * Convert nexthop (remote VTEP IP) into an IPv6 address.
638 */
639 static void evpn_convert_nexthop_to_ipv6(struct attr *attr)
640 {
641 if (BGP_ATTR_NEXTHOP_AFI_IP6(attr))
642 return;
643 ipv4_to_ipv4_mapped_ipv6(&attr->mp_nexthop_global, attr->nexthop);
644 attr->mp_nexthop_len = IPV6_MAX_BYTELEN;
645 }
646
647 /*
648 * Wrapper for node get in global table.
649 */
650 struct bgp_dest *bgp_evpn_global_node_get(struct bgp_table *table, afi_t afi,
651 safi_t safi,
652 const struct prefix_evpn *evp,
653 struct prefix_rd *prd,
654 const struct bgp_path_info *local_pi)
655 {
656 struct prefix_evpn global_p;
657
658 if (evp->prefix.route_type == BGP_EVPN_AD_ROUTE) {
659 /* prefix in the global table doesn't include the VTEP-IP so
660 * we need to create a different copy of the prefix
661 */
662 evpn_type1_prefix_global_copy(&global_p, evp);
663 evp = &global_p;
664 } else if (evp->prefix.route_type == BGP_EVPN_MAC_IP_ROUTE &&
665 local_pi) {
666 /*
667 * prefix in the global table needs MAC/IP, ensure they are
668 * present, using one's from local table's path_info.
669 */
670 if (is_evpn_prefix_ipaddr_none(evp)) {
671 /* VNI MAC -> Global */
672 evpn_type2_prefix_global_copy(
673 &global_p, evp, NULL /* mac */,
674 evpn_type2_path_info_get_ip(local_pi));
675 } else {
676 /* VNI IP -> Global */
677 evpn_type2_prefix_global_copy(
678 &global_p, evp,
679 evpn_type2_path_info_get_mac(local_pi),
680 NULL /* ip */);
681 }
682
683 evp = &global_p;
684 }
685 return bgp_afi_node_get(table, afi, safi, (struct prefix *)evp, prd);
686 }
687
688 /*
689 * Wrapper for node lookup in global table.
690 */
691 struct bgp_dest *
692 bgp_evpn_global_node_lookup(struct bgp_table *table, afi_t afi, safi_t safi,
693 const struct prefix_evpn *evp,
694 struct prefix_rd *prd,
695 const struct bgp_path_info *local_pi)
696 {
697 struct prefix_evpn global_p;
698
699 if (evp->prefix.route_type == BGP_EVPN_AD_ROUTE) {
700 /* prefix in the global table doesn't include the VTEP-IP so
701 * we need to create a different copy of the prefix
702 */
703 evpn_type1_prefix_global_copy(&global_p, evp);
704 evp = &global_p;
705 } else if (evp->prefix.route_type == BGP_EVPN_MAC_IP_ROUTE &&
706 local_pi) {
707 /*
708 * prefix in the global table needs MAC/IP, ensure they are
709 * present, using one's from local table's path_info.
710 */
711 if (is_evpn_prefix_ipaddr_none(evp)) {
712 /* VNI MAC -> Global */
713 evpn_type2_prefix_global_copy(
714 &global_p, evp, NULL /* mac */,
715 evpn_type2_path_info_get_ip(local_pi));
716 } else {
717 /* VNI IP -> Global */
718 evpn_type2_prefix_global_copy(
719 &global_p, evp,
720 evpn_type2_path_info_get_mac(local_pi),
721 NULL /* ip */);
722 }
723
724 evp = &global_p;
725 }
726 return bgp_afi_node_lookup(table, afi, safi, (struct prefix *)evp, prd);
727 }
728
729 /*
730 * Wrapper for node get in VNI IP table.
731 */
732 struct bgp_dest *bgp_evpn_vni_ip_node_get(struct bgp_table *const table,
733 const struct prefix_evpn *evp,
734 const struct bgp_path_info *parent_pi)
735 {
736 struct prefix_evpn vni_p;
737
738 if (evp->prefix.route_type == BGP_EVPN_AD_ROUTE && parent_pi) {
739 /* prefix in the global table doesn't include the VTEP-IP so
740 * we need to create a different copy for the VNI
741 */
742 evpn_type1_prefix_vni_ip_copy(&vni_p, evp,
743 parent_pi->attr->nexthop);
744 evp = &vni_p;
745 } else if (evp->prefix.route_type == BGP_EVPN_MAC_IP_ROUTE) {
746 /* Only MAC-IP should go into this table, not mac-only */
747 assert(is_evpn_prefix_ipaddr_none(evp) == false);
748
749 /*
750 * prefix in the vni IP table doesn't include MAC so
751 * we need to create a different copy of the prefix.
752 */
753 evpn_type2_prefix_vni_ip_copy(&vni_p, evp);
754 evp = &vni_p;
755 }
756 return bgp_node_get(table, (struct prefix *)evp);
757 }
758
759 /*
760 * Wrapper for node lookup in VNI IP table.
761 */
762 struct bgp_dest *
763 bgp_evpn_vni_ip_node_lookup(const struct bgp_table *const table,
764 const struct prefix_evpn *evp,
765 const struct bgp_path_info *parent_pi)
766 {
767 struct prefix_evpn vni_p;
768
769 if (evp->prefix.route_type == BGP_EVPN_AD_ROUTE && parent_pi) {
770 /* prefix in the global table doesn't include the VTEP-IP so
771 * we need to create a different copy for the VNI
772 */
773 evpn_type1_prefix_vni_ip_copy(&vni_p, evp,
774 parent_pi->attr->nexthop);
775 evp = &vni_p;
776 } else if (evp->prefix.route_type == BGP_EVPN_MAC_IP_ROUTE) {
777 /* Only MAC-IP should go into this table, not mac-only */
778 assert(is_evpn_prefix_ipaddr_none(evp) == false);
779
780 /*
781 * prefix in the vni IP table doesn't include MAC so
782 * we need to create a different copy of the prefix.
783 */
784 evpn_type2_prefix_vni_ip_copy(&vni_p, evp);
785 evp = &vni_p;
786 }
787 return bgp_node_lookup(table, (struct prefix *)evp);
788 }
789
790 /*
791 * Wrapper for node get in VNI MAC table.
792 */
793 struct bgp_dest *
794 bgp_evpn_vni_mac_node_get(struct bgp_table *const table,
795 const struct prefix_evpn *evp,
796 const struct bgp_path_info *parent_pi)
797 {
798 struct prefix_evpn vni_p;
799
800 /* Only type-2 should ever go into this table */
801 assert(evp->prefix.route_type == BGP_EVPN_MAC_IP_ROUTE);
802
803 /*
804 * prefix in the vni MAC table doesn't include IP so
805 * we need to create a different copy of the prefix.
806 */
807 evpn_type2_prefix_vni_mac_copy(&vni_p, evp);
808 evp = &vni_p;
809 return bgp_node_get(table, (struct prefix *)evp);
810 }
811
812 /*
813 * Wrapper for node lookup in VNI MAC table.
814 */
815 struct bgp_dest *
816 bgp_evpn_vni_mac_node_lookup(const struct bgp_table *const table,
817 const struct prefix_evpn *evp,
818 const struct bgp_path_info *parent_pi)
819 {
820 struct prefix_evpn vni_p;
821
822 /* Only type-2 should ever go into this table */
823 assert(evp->prefix.route_type == BGP_EVPN_MAC_IP_ROUTE);
824
825 /*
826 * prefix in the vni MAC table doesn't include IP so
827 * we need to create a different copy of the prefix.
828 */
829 evpn_type2_prefix_vni_mac_copy(&vni_p, evp);
830 evp = &vni_p;
831 return bgp_node_lookup(table, (struct prefix *)evp);
832 }
833
834 /*
835 * Wrapper for node get in both VNI tables.
836 */
837 struct bgp_dest *bgp_evpn_vni_node_get(struct bgpevpn *vpn,
838 const struct prefix_evpn *p,
839 const struct bgp_path_info *parent_pi)
840 {
841 if ((p->prefix.route_type == BGP_EVPN_MAC_IP_ROUTE) &&
842 (is_evpn_prefix_ipaddr_none(p) == true))
843 return bgp_evpn_vni_mac_node_get(vpn->mac_table, p, parent_pi);
844
845 return bgp_evpn_vni_ip_node_get(vpn->ip_table, p, parent_pi);
846 }
847
848 /*
849 * Wrapper for node lookup in both VNI tables.
850 */
851 struct bgp_dest *bgp_evpn_vni_node_lookup(const struct bgpevpn *vpn,
852 const struct prefix_evpn *p,
853 const struct bgp_path_info *parent_pi)
854 {
855 if ((p->prefix.route_type == BGP_EVPN_MAC_IP_ROUTE) &&
856 (is_evpn_prefix_ipaddr_none(p) == true))
857 return bgp_evpn_vni_mac_node_lookup(vpn->mac_table, p,
858 parent_pi);
859
860 return bgp_evpn_vni_ip_node_lookup(vpn->ip_table, p, parent_pi);
861 }
862
863 /*
864 * Add (update) or delete MACIP from zebra.
865 */
866 static int bgp_zebra_send_remote_macip(struct bgp *bgp, struct bgpevpn *vpn,
867 const struct prefix_evpn *p,
868 const struct ethaddr *mac,
869 struct in_addr remote_vtep_ip, int add,
870 uint8_t flags, uint32_t seq, esi_t *esi)
871 {
872 struct stream *s;
873 uint16_t ipa_len;
874 static struct in_addr zero_remote_vtep_ip;
875 bool esi_valid;
876
877 /* Check socket. */
878 if (!zclient || zclient->sock < 0)
879 return 0;
880
881 /* Don't try to register if Zebra doesn't know of this instance. */
882 if (!IS_BGP_INST_KNOWN_TO_ZEBRA(bgp)) {
883 if (BGP_DEBUG(zebra, ZEBRA))
884 zlog_debug(
885 "%s: No zebra instance to talk to, not installing remote macip",
886 __func__);
887 return 0;
888 }
889
890 if (!esi)
891 esi = zero_esi;
892 s = zclient->obuf;
893 stream_reset(s);
894
895 zclient_create_header(
896 s, add ? ZEBRA_REMOTE_MACIP_ADD : ZEBRA_REMOTE_MACIP_DEL,
897 bgp->vrf_id);
898 stream_putl(s, vpn->vni);
899
900 if (mac) /* Mac Addr */
901 stream_put(s, &mac->octet, ETH_ALEN);
902 else
903 stream_put(s, &p->prefix.macip_addr.mac.octet, ETH_ALEN);
904
905 /* IP address length and IP address, if any. */
906 if (is_evpn_prefix_ipaddr_none(p))
907 stream_putw(s, 0);
908 else {
909 ipa_len = is_evpn_prefix_ipaddr_v4(p) ? IPV4_MAX_BYTELEN
910 : IPV6_MAX_BYTELEN;
911 stream_putw(s, ipa_len);
912 stream_put(s, &p->prefix.macip_addr.ip.ip.addr, ipa_len);
913 }
914 /* If the ESI is valid that becomes the nexthop; tape out the
915 * VTEP-IP for that case
916 */
917 if (bgp_evpn_is_esi_valid(esi)) {
918 esi_valid = true;
919 stream_put_in_addr(s, &zero_remote_vtep_ip);
920 } else {
921 esi_valid = false;
922 stream_put_in_addr(s, &remote_vtep_ip);
923 }
924
925 /* TX flags - MAC sticky status and/or gateway mac */
926 /* Also TX the sequence number of the best route. */
927 if (add) {
928 stream_putc(s, flags);
929 stream_putl(s, seq);
930 stream_put(s, esi, sizeof(esi_t));
931 }
932
933 stream_putw_at(s, 0, stream_get_endp(s));
934
935 if (bgp_debug_zebra(NULL)) {
936 char esi_buf[ESI_STR_LEN];
937
938 if (esi_valid)
939 esi_to_str(esi, esi_buf, sizeof(esi_buf));
940 else
941 snprintf(esi_buf, sizeof(esi_buf), "-");
942 zlog_debug(
943 "Tx %s MACIP, VNI %u MAC %pEA IP %pIA flags 0x%x seq %u remote VTEP %pI4 esi %s",
944 add ? "ADD" : "DEL", vpn->vni,
945 (mac ? mac : &p->prefix.macip_addr.mac),
946 &p->prefix.macip_addr.ip, flags, seq, &remote_vtep_ip,
947 esi_buf);
948 }
949
950 frrtrace(5, frr_bgp, evpn_mac_ip_zsend, add, vpn, p, remote_vtep_ip,
951 esi);
952
953 return zclient_send_message(zclient);
954 }
955
956 /*
957 * Add (update) or delete remote VTEP from zebra.
958 */
959 static int bgp_zebra_send_remote_vtep(struct bgp *bgp, struct bgpevpn *vpn,
960 const struct prefix_evpn *p,
961 int flood_control, int add)
962 {
963 struct stream *s;
964
965 /* Check socket. */
966 if (!zclient || zclient->sock < 0)
967 return 0;
968
969 /* Don't try to register if Zebra doesn't know of this instance. */
970 if (!IS_BGP_INST_KNOWN_TO_ZEBRA(bgp)) {
971 if (BGP_DEBUG(zebra, ZEBRA))
972 zlog_debug(
973 "%s: No zebra instance to talk to, not installing remote vtep",
974 __func__);
975 return 0;
976 }
977
978 s = zclient->obuf;
979 stream_reset(s);
980
981 zclient_create_header(
982 s, add ? ZEBRA_REMOTE_VTEP_ADD : ZEBRA_REMOTE_VTEP_DEL,
983 bgp->vrf_id);
984 stream_putl(s, vpn->vni);
985 if (is_evpn_prefix_ipaddr_v4(p))
986 stream_put_in_addr(s, &p->prefix.imet_addr.ip.ipaddr_v4);
987 else if (is_evpn_prefix_ipaddr_v6(p)) {
988 flog_err(
989 EC_BGP_VTEP_INVALID,
990 "Bad remote IP when trying to %s remote VTEP for VNI %u",
991 add ? "ADD" : "DEL", vpn->vni);
992 return -1;
993 }
994 stream_putl(s, flood_control);
995
996 stream_putw_at(s, 0, stream_get_endp(s));
997
998 if (bgp_debug_zebra(NULL))
999 zlog_debug("Tx %s Remote VTEP, VNI %u remote VTEP %pI4",
1000 add ? "ADD" : "DEL", vpn->vni,
1001 &p->prefix.imet_addr.ip.ipaddr_v4);
1002
1003 frrtrace(3, frr_bgp, evpn_bum_vtep_zsend, add, vpn, p);
1004
1005 return zclient_send_message(zclient);
1006 }
1007
1008 /*
1009 * Build extended communities for EVPN prefix route.
1010 */
1011 static void build_evpn_type5_route_extcomm(struct bgp *bgp_vrf,
1012 struct attr *attr)
1013 {
1014 struct ecommunity ecom_encap;
1015 struct ecommunity_val eval;
1016 struct ecommunity_val eval_rmac;
1017 bgp_encap_types tnl_type;
1018 struct listnode *node, *nnode;
1019 struct vrf_route_target *l3rt;
1020 struct ecommunity *old_ecom;
1021 struct ecommunity *ecom;
1022 struct list *vrf_export_rtl = NULL;
1023
1024 /* Encap */
1025 tnl_type = BGP_ENCAP_TYPE_VXLAN;
1026 memset(&ecom_encap, 0, sizeof(ecom_encap));
1027 encode_encap_extcomm(tnl_type, &eval);
1028 ecom_encap.size = 1;
1029 ecom_encap.unit_size = ECOMMUNITY_SIZE;
1030 ecom_encap.val = (uint8_t *)eval.val;
1031
1032 /* Add Encap */
1033 if (bgp_attr_get_ecommunity(attr)) {
1034 old_ecom = bgp_attr_get_ecommunity(attr);
1035 ecom = ecommunity_merge(ecommunity_dup(old_ecom), &ecom_encap);
1036 if (!old_ecom->refcnt)
1037 ecommunity_free(&old_ecom);
1038 } else
1039 ecom = ecommunity_dup(&ecom_encap);
1040 bgp_attr_set_ecommunity(attr, ecom);
1041 attr->encap_tunneltype = tnl_type;
1042
1043 /* Add the export RTs for L3VNI/VRF */
1044 vrf_export_rtl = bgp_vrf->vrf_export_rtl;
1045 for (ALL_LIST_ELEMENTS(vrf_export_rtl, node, nnode, l3rt))
1046 bgp_attr_set_ecommunity(
1047 attr, ecommunity_merge(bgp_attr_get_ecommunity(attr),
1048 l3rt->ecom));
1049
1050 /* add the router mac extended community */
1051 if (!is_zero_mac(&attr->rmac)) {
1052 encode_rmac_extcomm(&eval_rmac, &attr->rmac);
1053 ecommunity_add_val(bgp_attr_get_ecommunity(attr), &eval_rmac,
1054 true, true);
1055 }
1056 }
1057
1058 /*
1059 * Build extended communities for EVPN route.
1060 * This function is applicable for type-2 and type-3 routes. The layer-2 RT
1061 * and ENCAP extended communities are applicable for all routes.
1062 * The default gateway extended community and MAC mobility (sticky) extended
1063 * community are added as needed based on passed settings - only for type-2
1064 * routes. Likewise, the layer-3 RT and Router MAC extended communities are
1065 * added, if present, based on passed settings - only for non-link-local
1066 * type-2 routes.
1067 */
1068 static void build_evpn_route_extcomm(struct bgpevpn *vpn, struct attr *attr,
1069 int add_l3_ecomm)
1070 {
1071 struct ecommunity ecom_encap;
1072 struct ecommunity ecom_sticky;
1073 struct ecommunity ecom_default_gw;
1074 struct ecommunity ecom_na;
1075 struct ecommunity_val eval;
1076 struct ecommunity_val eval_sticky;
1077 struct ecommunity_val eval_default_gw;
1078 struct ecommunity_val eval_rmac;
1079 struct ecommunity_val eval_na;
1080 bool proxy;
1081
1082 bgp_encap_types tnl_type;
1083 struct listnode *node, *nnode;
1084 struct ecommunity *ecom;
1085 struct vrf_route_target *l3rt;
1086 uint32_t seqnum;
1087 struct list *vrf_export_rtl = NULL;
1088
1089 /* Encap */
1090 tnl_type = BGP_ENCAP_TYPE_VXLAN;
1091 memset(&ecom_encap, 0, sizeof(ecom_encap));
1092 encode_encap_extcomm(tnl_type, &eval);
1093 ecom_encap.size = 1;
1094 ecom_encap.unit_size = ECOMMUNITY_SIZE;
1095 ecom_encap.val = (uint8_t *)eval.val;
1096
1097 /* Add Encap */
1098 bgp_attr_set_ecommunity(attr, ecommunity_dup(&ecom_encap));
1099 attr->encap_tunneltype = tnl_type;
1100
1101 /* Add the export RTs for L2VNI */
1102 for (ALL_LIST_ELEMENTS(vpn->export_rtl, node, nnode, ecom))
1103 bgp_attr_set_ecommunity(
1104 attr,
1105 ecommunity_merge(bgp_attr_get_ecommunity(attr), ecom));
1106
1107 /* Add the export RTs for L3VNI if told to - caller determines
1108 * when this should be done.
1109 */
1110 if (add_l3_ecomm) {
1111 vrf_export_rtl = bgpevpn_get_vrf_export_rtl(vpn);
1112 if (vrf_export_rtl && !list_isempty(vrf_export_rtl)) {
1113 for (ALL_LIST_ELEMENTS(vrf_export_rtl, node, nnode,
1114 l3rt))
1115 bgp_attr_set_ecommunity(
1116 attr,
1117 ecommunity_merge(
1118 bgp_attr_get_ecommunity(attr),
1119 l3rt->ecom));
1120 }
1121 }
1122
1123 /* Add MAC mobility (sticky) if needed. */
1124 if (attr->sticky) {
1125 seqnum = 0;
1126 memset(&ecom_sticky, 0, sizeof(ecom_sticky));
1127 encode_mac_mobility_extcomm(1, seqnum, &eval_sticky);
1128 ecom_sticky.size = 1;
1129 ecom_sticky.unit_size = ECOMMUNITY_SIZE;
1130 ecom_sticky.val = (uint8_t *)eval_sticky.val;
1131 bgp_attr_set_ecommunity(
1132 attr, ecommunity_merge(bgp_attr_get_ecommunity(attr),
1133 &ecom_sticky));
1134 }
1135
1136 /* Add RMAC, if told to. */
1137 if (add_l3_ecomm) {
1138 encode_rmac_extcomm(&eval_rmac, &attr->rmac);
1139 ecommunity_add_val(bgp_attr_get_ecommunity(attr), &eval_rmac,
1140 true, true);
1141 }
1142
1143 /* Add default gateway, if needed. */
1144 if (attr->default_gw) {
1145 memset(&ecom_default_gw, 0, sizeof(ecom_default_gw));
1146 encode_default_gw_extcomm(&eval_default_gw);
1147 ecom_default_gw.size = 1;
1148 ecom_default_gw.unit_size = ECOMMUNITY_SIZE;
1149 ecom_default_gw.val = (uint8_t *)eval_default_gw.val;
1150 bgp_attr_set_ecommunity(
1151 attr, ecommunity_merge(bgp_attr_get_ecommunity(attr),
1152 &ecom_default_gw));
1153 }
1154
1155 proxy = !!(attr->es_flags & ATTR_ES_PROXY_ADVERT);
1156 if (attr->router_flag || proxy) {
1157 memset(&ecom_na, 0, sizeof(ecom_na));
1158 encode_na_flag_extcomm(&eval_na, attr->router_flag, proxy);
1159 ecom_na.size = 1;
1160 ecom_na.unit_size = ECOMMUNITY_SIZE;
1161 ecom_na.val = (uint8_t *)eval_na.val;
1162 bgp_attr_set_ecommunity(
1163 attr, ecommunity_merge(bgp_attr_get_ecommunity(attr),
1164 &ecom_na));
1165 }
1166 }
1167
1168 /*
1169 * Add MAC mobility extended community to attribute.
1170 */
1171 static void add_mac_mobility_to_attr(uint32_t seq_num, struct attr *attr)
1172 {
1173 struct ecommunity ecom_tmp;
1174 struct ecommunity_val eval;
1175 uint8_t *ecom_val_ptr;
1176 uint32_t i;
1177 uint8_t *pnt;
1178 int type = 0;
1179 int sub_type = 0;
1180 struct ecommunity *ecomm = bgp_attr_get_ecommunity(attr);
1181
1182 /* Build MM */
1183 encode_mac_mobility_extcomm(0, seq_num, &eval);
1184
1185 /* Find current MM ecommunity */
1186 ecom_val_ptr = NULL;
1187
1188 if (ecomm) {
1189 for (i = 0; i < ecomm->size; i++) {
1190 pnt = ecomm->val + (i * ecomm->unit_size);
1191 type = *pnt++;
1192 sub_type = *pnt++;
1193
1194 if (type == ECOMMUNITY_ENCODE_EVPN
1195 && sub_type
1196 == ECOMMUNITY_EVPN_SUBTYPE_MACMOBILITY) {
1197 ecom_val_ptr =
1198 (ecomm->val + (i * ecomm->unit_size));
1199 break;
1200 }
1201 }
1202 }
1203
1204 /* Update the existing MM ecommunity */
1205 if (ecom_val_ptr) {
1206 memcpy(ecom_val_ptr, eval.val, sizeof(char) * ecomm->unit_size);
1207 }
1208 /* Add MM to existing */
1209 else {
1210 memset(&ecom_tmp, 0, sizeof(ecom_tmp));
1211 ecom_tmp.size = 1;
1212 ecom_tmp.unit_size = ECOMMUNITY_SIZE;
1213 ecom_tmp.val = (uint8_t *)eval.val;
1214
1215 if (ecomm)
1216 bgp_attr_set_ecommunity(
1217 attr, ecommunity_merge(ecomm, &ecom_tmp));
1218 else
1219 bgp_attr_set_ecommunity(attr,
1220 ecommunity_dup(&ecom_tmp));
1221 }
1222 }
1223
1224 /* Install EVPN route into zebra. */
1225 static int evpn_zebra_install(struct bgp *bgp, struct bgpevpn *vpn,
1226 const struct prefix_evpn *p,
1227 struct bgp_path_info *pi)
1228 {
1229 int ret;
1230 uint8_t flags;
1231 int flood_control;
1232 uint32_t seq;
1233
1234 if (p->prefix.route_type == BGP_EVPN_MAC_IP_ROUTE) {
1235 flags = 0;
1236
1237 if (pi->sub_type == BGP_ROUTE_IMPORTED) {
1238 if (pi->attr->sticky)
1239 SET_FLAG(flags, ZEBRA_MACIP_TYPE_STICKY);
1240 if (pi->attr->default_gw)
1241 SET_FLAG(flags, ZEBRA_MACIP_TYPE_GW);
1242 if (is_evpn_prefix_ipaddr_v6(p) &&
1243 pi->attr->router_flag)
1244 SET_FLAG(flags, ZEBRA_MACIP_TYPE_ROUTER_FLAG);
1245
1246 seq = mac_mobility_seqnum(pi->attr);
1247 /* if local ES notify zebra that this is a sync path */
1248 if (bgp_evpn_attr_is_local_es(pi->attr)) {
1249 SET_FLAG(flags, ZEBRA_MACIP_TYPE_SYNC_PATH);
1250 if (bgp_evpn_attr_is_proxy(pi->attr))
1251 SET_FLAG(flags,
1252 ZEBRA_MACIP_TYPE_PROXY_ADVERT);
1253 }
1254 } else {
1255 if (!bgp_evpn_attr_is_sync(pi->attr))
1256 return 0;
1257
1258 /* if a local path is being turned around and sent
1259 * to zebra it is because it is a sync path on
1260 * a local ES
1261 */
1262 SET_FLAG(flags, ZEBRA_MACIP_TYPE_SYNC_PATH);
1263 /* supply the highest peer seq number to zebra
1264 * for MM seq syncing
1265 */
1266 seq = bgp_evpn_attr_get_sync_seq(pi->attr);
1267 /* if any of the paths from the peer have the ROUTER
1268 * flag set install the local entry as a router entry
1269 */
1270 if (is_evpn_prefix_ipaddr_v6(p) &&
1271 (pi->attr->es_flags &
1272 ATTR_ES_PEER_ROUTER))
1273 SET_FLAG(flags,
1274 ZEBRA_MACIP_TYPE_ROUTER_FLAG);
1275
1276 if (!(pi->attr->es_flags & ATTR_ES_PEER_ACTIVE))
1277 SET_FLAG(flags,
1278 ZEBRA_MACIP_TYPE_PROXY_ADVERT);
1279 }
1280
1281 ret = bgp_zebra_send_remote_macip(
1282 bgp, vpn, p,
1283 (is_evpn_prefix_ipaddr_none(p)
1284 ? NULL /* MAC update */
1285 : evpn_type2_path_info_get_mac(
1286 pi) /* MAC-IP update */),
1287 pi->attr->nexthop, 1, flags, seq,
1288 bgp_evpn_attr_get_esi(pi->attr));
1289 } else if (p->prefix.route_type == BGP_EVPN_AD_ROUTE) {
1290 ret = bgp_evpn_remote_es_evi_add(bgp, vpn, p);
1291 } else {
1292 switch (bgp_attr_get_pmsi_tnl_type(pi->attr)) {
1293 case PMSI_TNLTYPE_INGR_REPL:
1294 flood_control = VXLAN_FLOOD_HEAD_END_REPL;
1295 break;
1296
1297 case PMSI_TNLTYPE_PIM_SM:
1298 flood_control = VXLAN_FLOOD_PIM_SM;
1299 break;
1300
1301 default:
1302 flood_control = VXLAN_FLOOD_DISABLED;
1303 break;
1304 }
1305 ret = bgp_zebra_send_remote_vtep(bgp, vpn, p, flood_control, 1);
1306 }
1307
1308 return ret;
1309 }
1310
1311 /* Uninstall EVPN route from zebra. */
1312 static int evpn_zebra_uninstall(struct bgp *bgp, struct bgpevpn *vpn,
1313 const struct prefix_evpn *p,
1314 struct bgp_path_info *pi, bool is_sync)
1315 {
1316 int ret;
1317
1318 if (p->prefix.route_type == BGP_EVPN_MAC_IP_ROUTE)
1319 ret = bgp_zebra_send_remote_macip(
1320 bgp, vpn, p,
1321 (is_evpn_prefix_ipaddr_none(p)
1322 ? NULL /* MAC update */
1323 : evpn_type2_path_info_get_mac(
1324 pi) /* MAC-IP update */),
1325 (is_sync ? zero_vtep_ip : pi->attr->nexthop), 0, 0, 0,
1326 NULL);
1327 else if (p->prefix.route_type == BGP_EVPN_AD_ROUTE)
1328 ret = bgp_evpn_remote_es_evi_del(bgp, vpn, p);
1329 else
1330 ret = bgp_zebra_send_remote_vtep(bgp, vpn, p,
1331 VXLAN_FLOOD_DISABLED, 0);
1332
1333 return ret;
1334 }
1335
1336 /*
1337 * Due to MAC mobility, the prior "local" best route has been supplanted
1338 * by a "remote" best route. The prior route has to be deleted and withdrawn
1339 * from peers.
1340 */
1341 static void evpn_delete_old_local_route(struct bgp *bgp, struct bgpevpn *vpn,
1342 struct bgp_dest *dest,
1343 struct bgp_path_info *old_local,
1344 struct bgp_path_info *new_select)
1345 {
1346 struct bgp_dest *global_dest;
1347 struct bgp_path_info *pi;
1348 afi_t afi = AFI_L2VPN;
1349 safi_t safi = SAFI_EVPN;
1350
1351 if (BGP_DEBUG(evpn_mh, EVPN_MH_RT)) {
1352 char esi_buf[ESI_STR_LEN];
1353 char esi_buf2[ESI_STR_LEN];
1354 struct prefix_evpn *evp =
1355 (struct prefix_evpn *)bgp_dest_get_prefix(dest);
1356
1357 zlog_debug("local path deleted %pFX es %s; new-path-es %s", evp,
1358 esi_to_str(&old_local->attr->esi, esi_buf,
1359 sizeof(esi_buf)),
1360 new_select ? esi_to_str(&new_select->attr->esi,
1361 esi_buf2, sizeof(esi_buf2))
1362 : "");
1363 }
1364
1365 /* Locate route node in the global EVPN routing table. Note that
1366 * this table is a 2-level tree (RD-level + Prefix-level) similar to
1367 * L3VPN routes.
1368 */
1369 global_dest = bgp_evpn_global_node_lookup(
1370 bgp->rib[afi][safi], afi, safi,
1371 (const struct prefix_evpn *)bgp_dest_get_prefix(dest),
1372 &vpn->prd, old_local);
1373 if (global_dest) {
1374 /* Delete route entry in the global EVPN table. */
1375 delete_evpn_route_entry(bgp, afi, safi, global_dest, &pi);
1376
1377 /* Schedule for processing - withdraws to peers happen from
1378 * this table.
1379 */
1380 if (pi)
1381 bgp_process(bgp, global_dest, afi, safi);
1382 bgp_dest_unlock_node(global_dest);
1383 }
1384
1385 /* Delete route entry in the VNI route table, caller to remove. */
1386 bgp_path_info_delete(dest, old_local);
1387 }
1388
1389 /*
1390 * Calculate the best path for an EVPN route. Install/update best path in zebra,
1391 * if appropriate.
1392 * Note: vpn is NULL for local EAD-ES routes.
1393 */
1394 int evpn_route_select_install(struct bgp *bgp, struct bgpevpn *vpn,
1395 struct bgp_dest *dest)
1396 {
1397 struct bgp_path_info *old_select, *new_select;
1398 struct bgp_path_info_pair old_and_new;
1399 afi_t afi = AFI_L2VPN;
1400 safi_t safi = SAFI_EVPN;
1401 int ret = 0;
1402
1403 /* Compute the best path. */
1404 bgp_best_selection(bgp, dest, &bgp->maxpaths[afi][safi], &old_and_new,
1405 afi, safi);
1406 old_select = old_and_new.old;
1407 new_select = old_and_new.new;
1408
1409 /* If the best path hasn't changed - see if there is still something to
1410 * update to zebra RIB.
1411 * Remote routes and SYNC route (i.e. local routes with
1412 * SYNCED_FROM_PEER flag) need to updated to zebra on any attr
1413 * change.
1414 */
1415 if (old_select && old_select == new_select
1416 && old_select->type == ZEBRA_ROUTE_BGP
1417 && (old_select->sub_type == BGP_ROUTE_IMPORTED ||
1418 bgp_evpn_attr_is_sync(old_select->attr))
1419 && !CHECK_FLAG(dest->flags, BGP_NODE_USER_CLEAR)
1420 && !CHECK_FLAG(old_select->flags, BGP_PATH_ATTR_CHANGED)
1421 && !bgp_addpath_is_addpath_used(&bgp->tx_addpath, afi, safi)) {
1422 if (bgp_zebra_has_route_changed(old_select))
1423 ret = evpn_zebra_install(
1424 bgp, vpn,
1425 (const struct prefix_evpn *)bgp_dest_get_prefix(
1426 dest),
1427 old_select);
1428 UNSET_FLAG(old_select->flags, BGP_PATH_MULTIPATH_CHG);
1429 UNSET_FLAG(old_select->flags, BGP_PATH_LINK_BW_CHG);
1430 bgp_zebra_clear_route_change_flags(dest);
1431 return ret;
1432 }
1433
1434 /* If the user did a "clear" this flag will be set */
1435 UNSET_FLAG(dest->flags, BGP_NODE_USER_CLEAR);
1436
1437 /* bestpath has changed; update relevant fields and install or uninstall
1438 * into the zebra RIB.
1439 */
1440 if (old_select || new_select)
1441 bgp_bump_version(dest);
1442
1443 if (old_select)
1444 bgp_path_info_unset_flag(dest, old_select, BGP_PATH_SELECTED);
1445 if (new_select) {
1446 bgp_path_info_set_flag(dest, new_select, BGP_PATH_SELECTED);
1447 bgp_path_info_unset_flag(dest, new_select,
1448 BGP_PATH_ATTR_CHANGED);
1449 UNSET_FLAG(new_select->flags, BGP_PATH_MULTIPATH_CHG);
1450 UNSET_FLAG(new_select->flags, BGP_PATH_LINK_BW_CHG);
1451 }
1452
1453 /* a local entry with the SYNC flag also results in a MAC-IP update
1454 * to zebra
1455 */
1456 if (new_select && new_select->type == ZEBRA_ROUTE_BGP
1457 && (new_select->sub_type == BGP_ROUTE_IMPORTED ||
1458 bgp_evpn_attr_is_sync(new_select->attr))) {
1459 ret = evpn_zebra_install(
1460 bgp, vpn,
1461 (struct prefix_evpn *)bgp_dest_get_prefix(dest),
1462 new_select);
1463
1464 /* If an old best existed and it was a "local" route, the only
1465 * reason
1466 * it would be supplanted is due to MAC mobility procedures. So,
1467 * we
1468 * need to do an implicit delete and withdraw that route from
1469 * peers.
1470 */
1471 if (new_select->sub_type == BGP_ROUTE_IMPORTED &&
1472 old_select && old_select->peer == bgp->peer_self
1473 && old_select->type == ZEBRA_ROUTE_BGP
1474 && old_select->sub_type == BGP_ROUTE_STATIC
1475 && vpn)
1476 evpn_delete_old_local_route(bgp, vpn, dest,
1477 old_select, new_select);
1478 } else {
1479 if (old_select && old_select->type == ZEBRA_ROUTE_BGP
1480 && old_select->sub_type == BGP_ROUTE_IMPORTED)
1481 ret = evpn_zebra_uninstall(
1482 bgp, vpn,
1483 (const struct prefix_evpn *)bgp_dest_get_prefix(
1484 dest),
1485 old_select, false);
1486 }
1487
1488 /* Clear any route change flags. */
1489 bgp_zebra_clear_route_change_flags(dest);
1490
1491 /* Reap old select bgp_path_info, if it has been removed */
1492 if (old_select && CHECK_FLAG(old_select->flags, BGP_PATH_REMOVED))
1493 bgp_path_info_reap(dest, old_select);
1494
1495 return ret;
1496 }
1497
1498 static struct bgp_path_info *bgp_evpn_route_get_local_path(
1499 struct bgp *bgp, struct bgp_dest *dest)
1500 {
1501 struct bgp_path_info *tmp_pi;
1502 struct bgp_path_info *local_pi = NULL;
1503
1504 for (tmp_pi = bgp_dest_get_bgp_path_info(dest); tmp_pi;
1505 tmp_pi = tmp_pi->next) {
1506 if (bgp_evpn_is_path_local(bgp, tmp_pi)) {
1507 local_pi = tmp_pi;
1508 break;
1509 }
1510 }
1511
1512 return local_pi;
1513 }
1514
1515 static int update_evpn_type5_route_entry(struct bgp *bgp_evpn,
1516 struct bgp *bgp_vrf, afi_t afi,
1517 safi_t safi, struct bgp_dest *dest,
1518 struct attr *attr, int *route_changed)
1519 {
1520 struct attr *attr_new = NULL;
1521 struct bgp_path_info *pi = NULL;
1522 mpls_label_t label = MPLS_INVALID_LABEL;
1523 struct bgp_path_info *local_pi = NULL;
1524 struct bgp_path_info *tmp_pi = NULL;
1525
1526 *route_changed = 0;
1527 /* locate the local route entry if any */
1528 for (tmp_pi = bgp_dest_get_bgp_path_info(dest); tmp_pi;
1529 tmp_pi = tmp_pi->next) {
1530 if (tmp_pi->peer == bgp_evpn->peer_self
1531 && tmp_pi->type == ZEBRA_ROUTE_BGP
1532 && tmp_pi->sub_type == BGP_ROUTE_STATIC)
1533 local_pi = tmp_pi;
1534 }
1535
1536 /*
1537 * create a new route entry if one doesn't exist.
1538 * Otherwise see if route attr has changed
1539 */
1540 if (!local_pi) {
1541
1542 /* route has changed as this is the first entry */
1543 *route_changed = 1;
1544
1545 /* Add (or update) attribute to hash. */
1546 attr_new = bgp_attr_intern(attr);
1547
1548 /* create the route info from attribute */
1549 pi = info_make(ZEBRA_ROUTE_BGP, BGP_ROUTE_STATIC, 0,
1550 bgp_evpn->peer_self, attr_new, dest);
1551 SET_FLAG(pi->flags, BGP_PATH_VALID);
1552
1553 /* Type-5 routes advertise the L3-VNI */
1554 bgp_path_info_extra_get(pi);
1555 vni2label(bgp_vrf->l3vni, &label);
1556 memcpy(&pi->extra->label, &label, sizeof(label));
1557 pi->extra->num_labels = 1;
1558
1559 /* add the route entry to route node*/
1560 bgp_path_info_add(dest, pi);
1561 } else {
1562
1563 tmp_pi = local_pi;
1564 if (!attrhash_cmp(tmp_pi->attr, attr)) {
1565
1566 /* attribute changed */
1567 *route_changed = 1;
1568
1569 /* The attribute has changed. */
1570 /* Add (or update) attribute to hash. */
1571 attr_new = bgp_attr_intern(attr);
1572 bgp_path_info_set_flag(dest, tmp_pi,
1573 BGP_PATH_ATTR_CHANGED);
1574
1575 /* Restore route, if needed. */
1576 if (CHECK_FLAG(tmp_pi->flags, BGP_PATH_REMOVED))
1577 bgp_path_info_restore(dest, tmp_pi);
1578
1579 /* Unintern existing, set to new. */
1580 bgp_attr_unintern(&tmp_pi->attr);
1581 tmp_pi->attr = attr_new;
1582 tmp_pi->uptime = monotime(NULL);
1583 }
1584 }
1585 return 0;
1586 }
1587
1588 /* update evpn type-5 route entry */
1589 static int update_evpn_type5_route(struct bgp *bgp_vrf, struct prefix_evpn *evp,
1590 struct attr *src_attr, afi_t src_afi,
1591 safi_t src_safi)
1592 {
1593 afi_t afi = AFI_L2VPN;
1594 safi_t safi = SAFI_EVPN;
1595 struct attr attr;
1596 struct bgp_dest *dest = NULL;
1597 struct bgp *bgp_evpn = NULL;
1598 int route_changed = 0;
1599
1600 bgp_evpn = bgp_get_evpn();
1601 if (!bgp_evpn)
1602 return 0;
1603
1604 /* Build path attribute for this route - use the source attr, if
1605 * present, else treat as locally originated.
1606 */
1607 if (src_attr)
1608 attr = *src_attr;
1609 else {
1610 memset(&attr, 0, sizeof(attr));
1611 bgp_attr_default_set(&attr, bgp_vrf, BGP_ORIGIN_IGP);
1612 }
1613
1614 /* Advertise Primary IP (PIP) is enabled, send individual
1615 * IP (default instance router-id) as nexthop.
1616 * PIP is disabled or vrr interface is not present
1617 * use anycast-IP as nexthop and anycast RMAC.
1618 */
1619 if (!bgp_vrf->evpn_info->advertise_pip ||
1620 (!bgp_vrf->evpn_info->is_anycast_mac)) {
1621 attr.nexthop = bgp_vrf->originator_ip;
1622 attr.mp_nexthop_global_in = bgp_vrf->originator_ip;
1623 memcpy(&attr.rmac, &bgp_vrf->rmac, ETH_ALEN);
1624 } else {
1625 /* copy sys rmac */
1626 memcpy(&attr.rmac, &bgp_vrf->evpn_info->pip_rmac, ETH_ALEN);
1627 if (bgp_vrf->evpn_info->pip_ip.s_addr != INADDR_ANY) {
1628 attr.nexthop = bgp_vrf->evpn_info->pip_ip;
1629 attr.mp_nexthop_global_in = bgp_vrf->evpn_info->pip_ip;
1630 } else if (bgp_vrf->evpn_info->pip_ip.s_addr == INADDR_ANY)
1631 if (bgp_debug_zebra(NULL))
1632 zlog_debug(
1633 "VRF %s evp %pFX advertise-pip primary ip is not configured",
1634 vrf_id_to_name(bgp_vrf->vrf_id), evp);
1635 }
1636
1637 if (bgp_debug_zebra(NULL))
1638 zlog_debug(
1639 "VRF %s type-5 route evp %pFX RMAC %pEA nexthop %pI4",
1640 vrf_id_to_name(bgp_vrf->vrf_id), evp, &attr.rmac,
1641 &attr.nexthop);
1642
1643 attr.mp_nexthop_len = BGP_ATTR_NHLEN_IPV4;
1644
1645 if (src_afi == AFI_IP6 &&
1646 CHECK_FLAG(bgp_vrf->af_flags[AFI_L2VPN][SAFI_EVPN],
1647 BGP_L2VPN_EVPN_ADV_IPV6_UNICAST_GW_IP)) {
1648 if (src_attr &&
1649 !IN6_IS_ADDR_UNSPECIFIED(&src_attr->mp_nexthop_global)) {
1650 attr.evpn_overlay.type = OVERLAY_INDEX_GATEWAY_IP;
1651 SET_IPADDR_V6(&attr.evpn_overlay.gw_ip);
1652 memcpy(&attr.evpn_overlay.gw_ip.ipaddr_v6,
1653 &src_attr->mp_nexthop_global,
1654 sizeof(struct in6_addr));
1655 }
1656 } else if (src_afi == AFI_IP &&
1657 CHECK_FLAG(bgp_vrf->af_flags[AFI_L2VPN][SAFI_EVPN],
1658 BGP_L2VPN_EVPN_ADV_IPV4_UNICAST_GW_IP)) {
1659 if (src_attr && src_attr->nexthop.s_addr != 0) {
1660 attr.evpn_overlay.type = OVERLAY_INDEX_GATEWAY_IP;
1661 SET_IPADDR_V4(&attr.evpn_overlay.gw_ip);
1662 memcpy(&attr.evpn_overlay.gw_ip.ipaddr_v4,
1663 &src_attr->nexthop, sizeof(struct in_addr));
1664 }
1665 }
1666
1667 /* Setup RT and encap extended community */
1668 build_evpn_type5_route_extcomm(bgp_vrf, &attr);
1669
1670 /* get the route node in global table */
1671 dest = bgp_evpn_global_node_get(bgp_evpn->rib[afi][safi], afi, safi,
1672 evp, &bgp_vrf->vrf_prd, NULL);
1673 assert(dest);
1674
1675 /* create or update the route entry within the route node */
1676 update_evpn_type5_route_entry(bgp_evpn, bgp_vrf, afi, safi, dest, &attr,
1677 &route_changed);
1678
1679 /* schedule for processing and unlock node */
1680 if (route_changed) {
1681 bgp_process(bgp_evpn, dest, afi, safi);
1682 bgp_dest_unlock_node(dest);
1683 }
1684
1685 /* uninten temporary */
1686 if (!src_attr)
1687 aspath_unintern(&attr.aspath);
1688 return 0;
1689 }
1690
1691 static void bgp_evpn_get_sync_info(struct bgp *bgp, esi_t *esi,
1692 struct bgp_dest *dest, uint32_t loc_seq,
1693 uint32_t *max_sync_seq, bool *active_on_peer,
1694 bool *peer_router, bool *proxy_from_peer,
1695 const struct ethaddr *mac)
1696 {
1697 struct bgp_path_info *tmp_pi;
1698 struct bgp_path_info *second_best_path = NULL;
1699 uint32_t tmp_mm_seq = 0;
1700 esi_t *tmp_esi;
1701 int paths_eq;
1702 struct ethaddr *tmp_mac;
1703 bool mac_cmp = false;
1704 struct prefix_evpn *evp = (struct prefix_evpn *)&dest->p;
1705
1706
1707 /* mac comparison is not needed for MAC-only routes */
1708 if (mac && !is_evpn_prefix_ipaddr_none(evp))
1709 mac_cmp = true;
1710
1711 /* find the best non-local path. a local path can only be present
1712 * as best path
1713 */
1714 for (tmp_pi = bgp_dest_get_bgp_path_info(dest); tmp_pi;
1715 tmp_pi = tmp_pi->next) {
1716 if (tmp_pi->sub_type != BGP_ROUTE_IMPORTED ||
1717 !CHECK_FLAG(tmp_pi->flags, BGP_PATH_VALID))
1718 continue;
1719
1720 /* ignore paths that have a different mac */
1721 if (mac_cmp) {
1722 tmp_mac = evpn_type2_path_info_get_mac(tmp_pi);
1723 if (memcmp(mac, tmp_mac, sizeof(*mac)))
1724 continue;
1725 }
1726
1727 if (bgp_evpn_path_info_cmp(bgp, tmp_pi,
1728 second_best_path, &paths_eq))
1729 second_best_path = tmp_pi;
1730 }
1731
1732 if (!second_best_path)
1733 return;
1734
1735 tmp_esi = bgp_evpn_attr_get_esi(second_best_path->attr);
1736 /* if this has the same ES desination as the local path
1737 * it is a sync path
1738 */
1739 if (!memcmp(esi, tmp_esi, sizeof(esi_t))) {
1740 tmp_mm_seq = mac_mobility_seqnum(second_best_path->attr);
1741 if (tmp_mm_seq < loc_seq)
1742 return;
1743
1744 /* we have a non-proxy path from the ES peer. */
1745 if (second_best_path->attr->es_flags &
1746 ATTR_ES_PROXY_ADVERT) {
1747 *proxy_from_peer = true;
1748 } else {
1749 *active_on_peer = true;
1750 }
1751
1752 if (second_best_path->attr->router_flag)
1753 *peer_router = true;
1754
1755 /* we use both proxy and non-proxy imports to
1756 * determine the max sync sequence
1757 */
1758 if (tmp_mm_seq > *max_sync_seq)
1759 *max_sync_seq = tmp_mm_seq;
1760 }
1761 }
1762
1763 /* Bubble up sync-info from all paths (non-best) to the local-path.
1764 * This is need for MM sequence number syncing and proxy advertisement.
1765 * Note: The local path can only exist as a best path in the
1766 * VPN route table. It will take precedence over all sync paths.
1767 */
1768 static void update_evpn_route_entry_sync_info(struct bgp *bgp,
1769 struct bgp_dest *dest,
1770 struct attr *attr,
1771 uint32_t loc_seq, bool setup_sync,
1772 const struct ethaddr *mac)
1773 {
1774 esi_t *esi;
1775 struct prefix_evpn *evp =
1776 (struct prefix_evpn *)bgp_dest_get_prefix(dest);
1777
1778 if (evp->prefix.route_type != BGP_EVPN_MAC_IP_ROUTE)
1779 return;
1780
1781 esi = bgp_evpn_attr_get_esi(attr);
1782 if (bgp_evpn_is_esi_valid(esi)) {
1783 if (setup_sync) {
1784 uint32_t max_sync_seq = 0;
1785 bool active_on_peer = false;
1786 bool peer_router = false;
1787 bool proxy_from_peer = false;
1788
1789 bgp_evpn_get_sync_info(bgp, esi, dest, loc_seq,
1790 &max_sync_seq, &active_on_peer,
1791 &peer_router, &proxy_from_peer,
1792 mac);
1793 attr->mm_sync_seqnum = max_sync_seq;
1794 if (active_on_peer)
1795 attr->es_flags |= ATTR_ES_PEER_ACTIVE;
1796 else
1797 attr->es_flags &= ~ATTR_ES_PEER_ACTIVE;
1798 if (proxy_from_peer)
1799 attr->es_flags |= ATTR_ES_PEER_PROXY;
1800 else
1801 attr->es_flags &= ~ATTR_ES_PEER_PROXY;
1802 if (peer_router)
1803 attr->es_flags |= ATTR_ES_PEER_ROUTER;
1804 else
1805 attr->es_flags &= ~ATTR_ES_PEER_ROUTER;
1806
1807 if (BGP_DEBUG(evpn_mh, EVPN_MH_RT)) {
1808 char esi_buf[ESI_STR_LEN];
1809
1810 zlog_debug(
1811 "setup sync info for %pFX es %s max_seq %d %s%s%s",
1812 evp,
1813 esi_to_str(esi, esi_buf,
1814 sizeof(esi_buf)),
1815 max_sync_seq,
1816 (attr->es_flags & ATTR_ES_PEER_ACTIVE)
1817 ? "peer-active "
1818 : "",
1819 (attr->es_flags & ATTR_ES_PEER_PROXY)
1820 ? "peer-proxy "
1821 : "",
1822 (attr->es_flags & ATTR_ES_PEER_ROUTER)
1823 ? "peer-router "
1824 : "");
1825 }
1826 }
1827 } else {
1828 attr->mm_sync_seqnum = 0;
1829 attr->es_flags &= ~ATTR_ES_PEER_ACTIVE;
1830 attr->es_flags &= ~ATTR_ES_PEER_PROXY;
1831 }
1832 }
1833
1834 /*
1835 * Create or update EVPN route entry. This could be in the VNI route tables
1836 * or the global route table.
1837 */
1838 static int update_evpn_route_entry(struct bgp *bgp, struct bgpevpn *vpn,
1839 afi_t afi, safi_t safi,
1840 struct bgp_dest *dest, struct attr *attr,
1841 const struct ethaddr *mac,
1842 const struct ipaddr *ip, int add,
1843 struct bgp_path_info **pi, uint8_t flags,
1844 uint32_t seq, bool vpn_rt, bool *old_is_sync)
1845 {
1846 struct bgp_path_info *tmp_pi;
1847 struct bgp_path_info *local_pi;
1848 struct attr *attr_new;
1849 mpls_label_t label[BGP_MAX_LABELS];
1850 uint32_t num_labels = 1;
1851 int route_change = 1;
1852 uint8_t sticky = 0;
1853 const struct prefix_evpn *evp;
1854
1855 *pi = NULL;
1856 evp = (const struct prefix_evpn *)bgp_dest_get_prefix(dest);
1857 memset(&label, 0, sizeof(label));
1858
1859 /* See if this is an update of an existing route, or a new add. */
1860 local_pi = bgp_evpn_route_get_local_path(bgp, dest);
1861
1862 /* If route doesn't exist already, create a new one, if told to.
1863 * Otherwise act based on whether the attributes of the route have
1864 * changed or not.
1865 */
1866 if (!local_pi && !add)
1867 return 0;
1868
1869 if (old_is_sync && local_pi)
1870 *old_is_sync = bgp_evpn_attr_is_sync(local_pi->attr);
1871
1872 /* if a local path is being added with a non-zero esi look
1873 * for SYNC paths from ES peers and bubble up the sync-info
1874 */
1875 update_evpn_route_entry_sync_info(bgp, dest, attr, seq, vpn_rt, mac);
1876
1877 /* For non-GW MACs, update MAC mobility seq number, if needed. */
1878 if (seq && !CHECK_FLAG(flags, ZEBRA_MACIP_TYPE_GW))
1879 add_mac_mobility_to_attr(seq, attr);
1880
1881 if (!local_pi) {
1882 /* Add (or update) attribute to hash. */
1883 attr_new = bgp_attr_intern(attr);
1884
1885 /* Extract MAC mobility sequence number, if any. */
1886 attr_new->mm_seqnum =
1887 bgp_attr_mac_mobility_seqnum(attr_new, &sticky);
1888 attr_new->sticky = sticky;
1889
1890 /* Create new route with its attribute. */
1891 tmp_pi = info_make(ZEBRA_ROUTE_BGP, BGP_ROUTE_STATIC, 0,
1892 bgp->peer_self, attr_new, dest);
1893 SET_FLAG(tmp_pi->flags, BGP_PATH_VALID);
1894 bgp_path_info_extra_get(tmp_pi);
1895
1896 /* The VNI goes into the 'label' field of the route */
1897 vni2label(vpn->vni, &label[0]);
1898
1899 /* Type-2 routes may carry a second VNI - the L3-VNI.
1900 * Only attach second label if we are advertising two labels for
1901 * type-2 routes.
1902 */
1903 if (evp->prefix.route_type == BGP_EVPN_MAC_IP_ROUTE
1904 && CHECK_FLAG(vpn->flags, VNI_FLAG_USE_TWO_LABELS)) {
1905 vni_t l3vni;
1906
1907 l3vni = bgpevpn_get_l3vni(vpn);
1908 if (l3vni) {
1909 vni2label(l3vni, &label[1]);
1910 num_labels++;
1911 }
1912 }
1913
1914 memcpy(&tmp_pi->extra->label, label, sizeof(label));
1915 tmp_pi->extra->num_labels = num_labels;
1916
1917 if (evp->prefix.route_type == BGP_EVPN_MAC_IP_ROUTE) {
1918 if (mac)
1919 evpn_type2_path_info_set_mac(tmp_pi, *mac);
1920 else if (ip)
1921 evpn_type2_path_info_set_ip(tmp_pi, *ip);
1922 }
1923
1924 /* Mark route as self type-2 route */
1925 if (flags && CHECK_FLAG(flags, ZEBRA_MACIP_TYPE_SVI_IP))
1926 tmp_pi->extra->af_flags = BGP_EVPN_MACIP_TYPE_SVI_IP;
1927 bgp_path_info_add(dest, tmp_pi);
1928 } else {
1929 tmp_pi = local_pi;
1930 if (attrhash_cmp(tmp_pi->attr, attr)
1931 && !CHECK_FLAG(tmp_pi->flags, BGP_PATH_REMOVED))
1932 route_change = 0;
1933 else {
1934 /*
1935 * The attributes have changed, type-2 routes needs to
1936 * be advertised with right labels.
1937 */
1938 vni2label(vpn->vni, &label[0]);
1939 if (evp->prefix.route_type == BGP_EVPN_MAC_IP_ROUTE
1940 && CHECK_FLAG(vpn->flags,
1941 VNI_FLAG_USE_TWO_LABELS)) {
1942 vni_t l3vni;
1943
1944 l3vni = bgpevpn_get_l3vni(vpn);
1945 if (l3vni) {
1946 vni2label(l3vni, &label[1]);
1947 num_labels++;
1948 }
1949 }
1950 memcpy(&tmp_pi->extra->label, label, sizeof(label));
1951 tmp_pi->extra->num_labels = num_labels;
1952
1953 if (evp->prefix.route_type == BGP_EVPN_MAC_IP_ROUTE) {
1954 if (mac)
1955 evpn_type2_path_info_set_mac(tmp_pi,
1956 *mac);
1957 else if (ip)
1958 evpn_type2_path_info_set_ip(tmp_pi,
1959 *ip);
1960 }
1961
1962 /* The attribute has changed. */
1963 /* Add (or update) attribute to hash. */
1964 attr_new = bgp_attr_intern(attr);
1965 bgp_path_info_set_flag(dest, tmp_pi,
1966 BGP_PATH_ATTR_CHANGED);
1967
1968 /* Extract MAC mobility sequence number, if any. */
1969 attr_new->mm_seqnum =
1970 bgp_attr_mac_mobility_seqnum(attr_new, &sticky);
1971 attr_new->sticky = sticky;
1972
1973 /* Restore route, if needed. */
1974 if (CHECK_FLAG(tmp_pi->flags, BGP_PATH_REMOVED))
1975 bgp_path_info_restore(dest, tmp_pi);
1976
1977 /* Unintern existing, set to new. */
1978 bgp_attr_unintern(&tmp_pi->attr);
1979 tmp_pi->attr = attr_new;
1980 tmp_pi->uptime = monotime(NULL);
1981 }
1982 }
1983
1984 /* local MAC-IP routes in the VNI table are linked to
1985 * the destination ES
1986 */
1987 if (route_change && vpn_rt
1988 && (evp->prefix.route_type == BGP_EVPN_MAC_IP_ROUTE))
1989 bgp_evpn_path_es_link(tmp_pi, vpn->vni,
1990 bgp_evpn_attr_get_esi(tmp_pi->attr));
1991
1992 /* Return back the route entry. */
1993 *pi = tmp_pi;
1994 return route_change;
1995 }
1996
1997 static void evpn_zebra_reinstall_best_route(struct bgp *bgp,
1998 struct bgpevpn *vpn,
1999 struct bgp_dest *dest)
2000 {
2001 struct bgp_path_info *tmp_ri;
2002 struct bgp_path_info *curr_select = NULL;
2003
2004 for (tmp_ri = bgp_dest_get_bgp_path_info(dest); tmp_ri;
2005 tmp_ri = tmp_ri->next) {
2006 if (CHECK_FLAG(tmp_ri->flags, BGP_PATH_SELECTED)) {
2007 curr_select = tmp_ri;
2008 break;
2009 }
2010 }
2011
2012 if (curr_select && curr_select->type == ZEBRA_ROUTE_BGP
2013 && (curr_select->sub_type == BGP_ROUTE_IMPORTED ||
2014 bgp_evpn_attr_is_sync(curr_select->attr)))
2015 evpn_zebra_install(bgp, vpn,
2016 (const struct prefix_evpn *)bgp_dest_get_prefix(dest),
2017 curr_select);
2018 }
2019
2020 /*
2021 * If the local route was not selected evict it and tell zebra to re-add
2022 * the best remote dest.
2023 *
2024 * Typically a local path added by zebra is expected to be selected as
2025 * best. In which case when a remote path wins as best (later)
2026 * evpn_route_select_install itself evicts the older-local-best path.
2027 *
2028 * However if bgp's add and zebra's add cross paths (race condition) it
2029 * is possible that the local path is no longer the "older" best path.
2030 * It is a path that was never designated as best and hence requires
2031 * additional handling to prevent bgp from injecting and holding on to a
2032 * non-best local path.
2033 */
2034 static void evpn_cleanup_local_non_best_route(struct bgp *bgp,
2035 struct bgpevpn *vpn,
2036 struct bgp_dest *dest,
2037 struct bgp_path_info *local_pi)
2038 {
2039 /* local path was not picked as the winner; kick it out */
2040 if (bgp_debug_zebra(NULL))
2041 zlog_debug("evicting local evpn prefix %pBD as remote won",
2042 dest);
2043
2044 evpn_delete_old_local_route(bgp, vpn, dest, local_pi, NULL);
2045 bgp_path_info_reap(dest, local_pi);
2046
2047 /* tell zebra to re-add the best remote path */
2048 evpn_zebra_reinstall_best_route(bgp, vpn, dest);
2049 }
2050
2051 static inline bool bgp_evpn_route_add_l3_ecomm_ok(struct bgpevpn *vpn,
2052 const struct prefix_evpn *p,
2053 esi_t *esi)
2054 {
2055 return p->prefix.route_type == BGP_EVPN_MAC_IP_ROUTE
2056 && (is_evpn_prefix_ipaddr_v4(p)
2057 || (is_evpn_prefix_ipaddr_v6(p)
2058 && !IN6_IS_ADDR_LINKLOCAL(
2059 &p->prefix.macip_addr.ip.ipaddr_v6)))
2060 && CHECK_FLAG(vpn->flags, VNI_FLAG_USE_TWO_LABELS)
2061 && bgpevpn_get_l3vni(vpn) && bgp_evpn_es_add_l3_ecomm_ok(esi);
2062 }
2063
2064 /*
2065 * Create or update EVPN route (of type based on prefix) for specified VNI
2066 * and schedule for processing.
2067 */
2068 static int update_evpn_route(struct bgp *bgp, struct bgpevpn *vpn,
2069 struct prefix_evpn *p, uint8_t flags,
2070 uint32_t seq, esi_t *esi)
2071 {
2072 struct bgp_dest *dest;
2073 struct attr attr;
2074 struct attr *attr_new;
2075 int add_l3_ecomm = 0;
2076 struct bgp_path_info *pi;
2077 afi_t afi = AFI_L2VPN;
2078 safi_t safi = SAFI_EVPN;
2079 int route_change;
2080 bool old_is_sync = false;
2081 bool mac_only = false;
2082
2083 memset(&attr, 0, sizeof(attr));
2084
2085 /* Build path-attribute for this route. */
2086 bgp_attr_default_set(&attr, bgp, BGP_ORIGIN_IGP);
2087 attr.nexthop = vpn->originator_ip;
2088 attr.mp_nexthop_global_in = vpn->originator_ip;
2089 attr.mp_nexthop_len = BGP_ATTR_NHLEN_IPV4;
2090 attr.sticky = CHECK_FLAG(flags, ZEBRA_MACIP_TYPE_STICKY) ? 1 : 0;
2091 attr.default_gw = CHECK_FLAG(flags, ZEBRA_MACIP_TYPE_GW) ? 1 : 0;
2092 attr.router_flag = CHECK_FLAG(flags,
2093 ZEBRA_MACIP_TYPE_ROUTER_FLAG) ? 1 : 0;
2094 if (CHECK_FLAG(flags, ZEBRA_MACIP_TYPE_PROXY_ADVERT))
2095 attr.es_flags |= ATTR_ES_PROXY_ADVERT;
2096
2097 if (esi && bgp_evpn_is_esi_valid(esi)) {
2098 memcpy(&attr.esi, esi, sizeof(esi_t));
2099 attr.es_flags |= ATTR_ES_IS_LOCAL;
2100 }
2101
2102 /* PMSI is only needed for type-3 routes */
2103 if (p->prefix.route_type == BGP_EVPN_IMET_ROUTE) {
2104 attr.flag |= ATTR_FLAG_BIT(BGP_ATTR_PMSI_TUNNEL);
2105 bgp_attr_set_pmsi_tnl_type(&attr, PMSI_TNLTYPE_INGR_REPL);
2106 }
2107
2108 /* router mac is only needed for type-2 routes here. */
2109 if (p->prefix.route_type == BGP_EVPN_MAC_IP_ROUTE) {
2110 uint8_t af_flags = 0;
2111
2112 if (CHECK_FLAG(flags, ZEBRA_MACIP_TYPE_SVI_IP))
2113 SET_FLAG(af_flags, BGP_EVPN_MACIP_TYPE_SVI_IP);
2114
2115 bgp_evpn_get_rmac_nexthop(vpn, p, &attr, af_flags);
2116 }
2117
2118 if (bgp_debug_zebra(NULL)) {
2119 char buf3[ESI_STR_LEN];
2120
2121 zlog_debug(
2122 "VRF %s vni %u type-%u route evp %pFX RMAC %pEA nexthop %pI4 esi %s",
2123 vpn->bgp_vrf ? vrf_id_to_name(vpn->bgp_vrf->vrf_id)
2124 : "None",
2125 vpn->vni, p->prefix.route_type, p, &attr.rmac,
2126 &attr.mp_nexthop_global_in,
2127 esi_to_str(esi, buf3, sizeof(buf3)));
2128 }
2129
2130 vni2label(vpn->vni, &(attr.label));
2131
2132 /* Include L3 VNI related RTs and RMAC for type-2 routes, if they're
2133 * IPv4 or IPv6 global addresses and we're advertising L3VNI with
2134 * these routes.
2135 */
2136 add_l3_ecomm = bgp_evpn_route_add_l3_ecomm_ok(
2137 vpn, p, (attr.es_flags & ATTR_ES_IS_LOCAL) ? &attr.esi : NULL);
2138
2139 /* Set up extended community. */
2140 build_evpn_route_extcomm(vpn, &attr, add_l3_ecomm);
2141
2142 /* First, create (or fetch) route node within the VNI.
2143 * NOTE: There is no RD here.
2144 */
2145 dest = bgp_evpn_vni_node_get(vpn, p, NULL);
2146
2147 if ((p->prefix.route_type == BGP_EVPN_MAC_IP_ROUTE) &&
2148 (is_evpn_prefix_ipaddr_none(p) == true))
2149 mac_only = true;
2150
2151 /* Create or update route entry. */
2152 route_change = update_evpn_route_entry(
2153 bgp, vpn, afi, safi, dest, &attr,
2154 (mac_only ? NULL : &p->prefix.macip_addr.mac), NULL /* ip */, 1,
2155 &pi, flags, seq, true /* setup_sync */, &old_is_sync);
2156 assert(pi);
2157 attr_new = pi->attr;
2158
2159 /* lock ri to prevent freeing in evpn_route_select_install */
2160 bgp_path_info_lock(pi);
2161
2162 /* Perform route selection. Normally, the local route in the
2163 * VNI is expected to win and be the best route. However, if
2164 * there is a race condition where a host moved from local to
2165 * remote and the remote route was received in BGP just prior
2166 * to the local MACIP notification from zebra, the remote
2167 * route would win, and we should evict the defunct local route
2168 * and (re)install the remote route into zebra.
2169 */
2170 evpn_route_select_install(bgp, vpn, dest);
2171 /*
2172 * If the new local route was not selected evict it and tell zebra
2173 * to re-add the best remote dest. BGP doesn't retain non-best local
2174 * routes.
2175 */
2176 if (CHECK_FLAG(pi->flags, BGP_PATH_REMOVED)) {
2177 route_change = 0;
2178 } else {
2179 if (!CHECK_FLAG(pi->flags, BGP_PATH_SELECTED)) {
2180 route_change = 0;
2181 evpn_cleanup_local_non_best_route(bgp, vpn, dest, pi);
2182 } else {
2183 bool new_is_sync;
2184
2185 /* If the local path already existed and is still the
2186 * best path we need to also check if it transitioned
2187 * from being a sync path to a non-sync path. If it
2188 * it did we need to notify zebra that the sync-path
2189 * has been removed.
2190 */
2191 new_is_sync = bgp_evpn_attr_is_sync(pi->attr);
2192 if (!new_is_sync && old_is_sync)
2193 evpn_zebra_uninstall(bgp, vpn, p, pi, true);
2194 }
2195 }
2196 bgp_path_info_unlock(pi);
2197
2198 bgp_dest_unlock_node(dest);
2199
2200 /* If this is a new route or some attribute has changed, export the
2201 * route to the global table. The route will be advertised to peers
2202 * from there. Note that this table is a 2-level tree (RD-level +
2203 * Prefix-level) similar to L3VPN routes.
2204 */
2205 if (route_change) {
2206 struct bgp_path_info *global_pi;
2207
2208 dest = bgp_evpn_global_node_get(bgp->rib[afi][safi], afi, safi,
2209 p, &vpn->prd, NULL);
2210 update_evpn_route_entry(
2211 bgp, vpn, afi, safi, dest, attr_new, NULL /* mac */,
2212 NULL /* ip */, 1, &global_pi, flags, seq,
2213 false /* setup_sync */, NULL /* old_is_sync */);
2214
2215 /* Schedule for processing and unlock node. */
2216 bgp_process(bgp, dest, afi, safi);
2217 bgp_dest_unlock_node(dest);
2218 }
2219
2220 /* Unintern temporary. */
2221 aspath_unintern(&attr.aspath);
2222
2223 return 0;
2224 }
2225
2226 /*
2227 * Delete EVPN route entry.
2228 * The entry can be in ESI/VNI table or the global table.
2229 */
2230 void delete_evpn_route_entry(struct bgp *bgp, afi_t afi, safi_t safi,
2231 struct bgp_dest *dest,
2232 struct bgp_path_info **pi)
2233 {
2234 struct bgp_path_info *tmp_pi;
2235
2236 *pi = NULL;
2237
2238 /* Now, find matching route. */
2239 for (tmp_pi = bgp_dest_get_bgp_path_info(dest); tmp_pi;
2240 tmp_pi = tmp_pi->next)
2241 if (tmp_pi->peer == bgp->peer_self
2242 && tmp_pi->type == ZEBRA_ROUTE_BGP
2243 && tmp_pi->sub_type == BGP_ROUTE_STATIC)
2244 break;
2245
2246 *pi = tmp_pi;
2247
2248 /* Mark route for delete. */
2249 if (tmp_pi)
2250 bgp_path_info_delete(dest, tmp_pi);
2251 }
2252
2253 /* Delete EVPN type5 route */
2254 static int delete_evpn_type5_route(struct bgp *bgp_vrf, struct prefix_evpn *evp)
2255 {
2256 afi_t afi = AFI_L2VPN;
2257 safi_t safi = SAFI_EVPN;
2258 struct bgp_dest *dest = NULL;
2259 struct bgp_path_info *pi = NULL;
2260 struct bgp *bgp_evpn = NULL; /* evpn bgp instance */
2261
2262 bgp_evpn = bgp_get_evpn();
2263 if (!bgp_evpn)
2264 return 0;
2265
2266 /* locate the global route entry for this type-5 prefix */
2267 dest = bgp_evpn_global_node_lookup(bgp_evpn->rib[afi][safi], afi, safi,
2268 evp, &bgp_vrf->vrf_prd, NULL);
2269 if (!dest)
2270 return 0;
2271
2272 delete_evpn_route_entry(bgp_evpn, afi, safi, dest, &pi);
2273 if (pi)
2274 bgp_process(bgp_evpn, dest, afi, safi);
2275 bgp_dest_unlock_node(dest);
2276 return 0;
2277 }
2278
2279 /*
2280 * Delete EVPN route (of type based on prefix) for specified VNI and
2281 * schedule for processing.
2282 */
2283 static int delete_evpn_route(struct bgp *bgp, struct bgpevpn *vpn,
2284 struct prefix_evpn *p)
2285 {
2286 struct bgp_dest *dest, *global_dest;
2287 struct bgp_path_info *pi;
2288 afi_t afi = AFI_L2VPN;
2289 safi_t safi = SAFI_EVPN;
2290
2291 /* First, locate the route node within the VNI. If it doesn't exist,
2292 * there
2293 * is nothing further to do.
2294 * NOTE: There is no RD here.
2295 */
2296 dest = bgp_evpn_vni_node_lookup(vpn, p, NULL);
2297 if (!dest)
2298 return 0;
2299
2300 /* Next, locate route node in the global EVPN routing table. Note that
2301 * this table is a 2-level tree (RD-level + Prefix-level) similar to
2302 * L3VPN routes.
2303 */
2304 global_dest = bgp_evpn_global_node_lookup(bgp->rib[afi][safi], afi,
2305 safi, p, &vpn->prd, NULL);
2306 if (global_dest) {
2307 /* Delete route entry in the global EVPN table. */
2308 delete_evpn_route_entry(bgp, afi, safi, global_dest, &pi);
2309
2310 /* Schedule for processing - withdraws to peers happen from
2311 * this table.
2312 */
2313 if (pi)
2314 bgp_process(bgp, global_dest, afi, safi);
2315 bgp_dest_unlock_node(global_dest);
2316 }
2317
2318 /* Delete route entry in the VNI route table. This can just be removed.
2319 */
2320 delete_evpn_route_entry(bgp, afi, safi, dest, &pi);
2321 if (pi) {
2322 bgp_path_info_reap(dest, pi);
2323 evpn_route_select_install(bgp, vpn, dest);
2324 }
2325 bgp_dest_unlock_node(dest);
2326
2327 return 0;
2328 }
2329
2330 void bgp_evpn_update_type2_route_entry(struct bgp *bgp, struct bgpevpn *vpn,
2331 struct bgp_dest *dest,
2332 struct bgp_path_info *local_pi,
2333 const char *caller)
2334 {
2335 afi_t afi = AFI_L2VPN;
2336 safi_t safi = SAFI_EVPN;
2337 struct bgp_path_info *pi;
2338 struct attr attr;
2339 struct attr *attr_new;
2340 uint32_t seq;
2341 int add_l3_ecomm = 0;
2342 struct bgp_dest *global_dest;
2343 struct bgp_path_info *global_pi;
2344 struct prefix_evpn evp;
2345 int route_change;
2346 bool old_is_sync = false;
2347
2348 if (CHECK_FLAG(local_pi->flags, BGP_PATH_REMOVED))
2349 return;
2350
2351 /*
2352 * VNI table MAC-IP prefixes don't have MAC so make sure it's set from
2353 * path info here.
2354 */
2355 if (is_evpn_prefix_ipaddr_none((struct prefix_evpn *)&dest->p)) {
2356 /* VNI MAC -> Global */
2357 evpn_type2_prefix_global_copy(
2358 &evp, (struct prefix_evpn *)&dest->p, NULL /* mac */,
2359 evpn_type2_path_info_get_ip(local_pi));
2360 } else {
2361 /* VNI IP -> Global */
2362 evpn_type2_prefix_global_copy(
2363 &evp, (struct prefix_evpn *)&dest->p,
2364 evpn_type2_path_info_get_mac(local_pi), NULL /* ip */);
2365 }
2366
2367 /*
2368 * Build attribute per local route as the MAC mobility and
2369 * some other values could differ for different routes. The
2370 * attributes will be shared in the hash table.
2371 */
2372 bgp_attr_default_set(&attr, bgp, BGP_ORIGIN_IGP);
2373 attr.nexthop = vpn->originator_ip;
2374 attr.mp_nexthop_global_in = vpn->originator_ip;
2375 attr.mp_nexthop_len = BGP_ATTR_NHLEN_IPV4;
2376 attr.sticky = (local_pi->attr->sticky) ? 1 : 0;
2377 attr.router_flag = (local_pi->attr->router_flag) ? 1 : 0;
2378 attr.es_flags = local_pi->attr->es_flags;
2379 if (local_pi->attr->default_gw) {
2380 attr.default_gw = 1;
2381 if (is_evpn_prefix_ipaddr_v6(&evp))
2382 attr.router_flag = 1;
2383 }
2384 memcpy(&attr.esi, &local_pi->attr->esi, sizeof(esi_t));
2385 bgp_evpn_get_rmac_nexthop(vpn, &evp, &attr, local_pi->extra->af_flags);
2386 vni2label(vpn->vni, &(attr.label));
2387 /* Add L3 VNI RTs and RMAC for non IPv6 link-local if
2388 * using L3 VNI for type-2 routes also.
2389 */
2390 add_l3_ecomm = bgp_evpn_route_add_l3_ecomm_ok(
2391 vpn, &evp,
2392 (attr.es_flags & ATTR_ES_IS_LOCAL) ? &attr.esi : NULL);
2393
2394 /* Set up extended community. */
2395 build_evpn_route_extcomm(vpn, &attr, add_l3_ecomm);
2396 seq = mac_mobility_seqnum(local_pi->attr);
2397
2398 if (bgp_debug_zebra(NULL)) {
2399 char buf3[ESI_STR_LEN];
2400
2401 zlog_debug(
2402 "VRF %s vni %u evp %pFX RMAC %pEA nexthop %pI4 esi %s esf 0x%x from %s",
2403 vpn->bgp_vrf ? vrf_id_to_name(vpn->bgp_vrf->vrf_id)
2404 : " ",
2405 vpn->vni, &evp, &attr.rmac, &attr.mp_nexthop_global_in,
2406 esi_to_str(&attr.esi, buf3, sizeof(buf3)),
2407 attr.es_flags, caller);
2408 }
2409
2410 /* Update the route entry. */
2411 route_change = update_evpn_route_entry(
2412 bgp, vpn, afi, safi, dest, &attr, NULL /* mac */, NULL /* ip */,
2413 0, &pi, 0, seq, true /* setup_sync */, &old_is_sync);
2414
2415 assert(pi);
2416 attr_new = pi->attr;
2417 /* lock ri to prevent freeing in evpn_route_select_install */
2418 bgp_path_info_lock(pi);
2419
2420 /* Perform route selection. Normally, the local route in the
2421 * VNI is expected to win and be the best route. However,
2422 * under peculiar situations (e.g., tunnel (next hop) IP change
2423 * that causes best selection to be based on next hop), a
2424 * remote route could win. If the local route is the best,
2425 * ensure it is updated in the global EVPN route table and
2426 * advertised to peers; otherwise, ensure it is evicted and
2427 * (re)install the remote route into zebra.
2428 */
2429 evpn_route_select_install(bgp, vpn, dest);
2430
2431 if (CHECK_FLAG(pi->flags, BGP_PATH_REMOVED)) {
2432 route_change = 0;
2433 } else {
2434 if (!CHECK_FLAG(pi->flags, BGP_PATH_SELECTED)) {
2435 route_change = 0;
2436 evpn_cleanup_local_non_best_route(bgp, vpn, dest, pi);
2437 } else {
2438 bool new_is_sync;
2439
2440 /* If the local path already existed and is still the
2441 * best path we need to also check if it transitioned
2442 * from being a sync path to a non-sync path. If it
2443 * it did we need to notify zebra that the sync-path
2444 * has been removed.
2445 */
2446 new_is_sync = bgp_evpn_attr_is_sync(pi->attr);
2447 if (!new_is_sync && old_is_sync)
2448 evpn_zebra_uninstall(bgp, vpn, &evp, pi, true);
2449 }
2450 }
2451
2452
2453 /* unlock pi */
2454 bgp_path_info_unlock(pi);
2455
2456 if (route_change) {
2457 /* Update route in global routing table. */
2458 global_dest = bgp_evpn_global_node_get(
2459 bgp->rib[afi][safi], afi, safi, &evp, &vpn->prd, NULL);
2460 assert(global_dest);
2461 update_evpn_route_entry(
2462 bgp, vpn, afi, safi, global_dest, attr_new,
2463 NULL /* mac */, NULL /* ip */, 0, &global_pi, 0,
2464 mac_mobility_seqnum(attr_new), false /* setup_sync */,
2465 NULL /* old_is_sync */);
2466
2467 /* Schedule for processing and unlock node. */
2468 bgp_process(bgp, global_dest, afi, safi);
2469 bgp_dest_unlock_node(global_dest);
2470 }
2471
2472 /* Unintern temporary. */
2473 aspath_unintern(&attr.aspath);
2474 }
2475
2476 static void update_type2_route(struct bgp *bgp, struct bgpevpn *vpn,
2477 struct bgp_dest *dest)
2478 {
2479 struct bgp_path_info *tmp_pi;
2480
2481 const struct prefix_evpn *evp =
2482 (const struct prefix_evpn *)bgp_dest_get_prefix(dest);
2483
2484 if (evp->prefix.route_type != BGP_EVPN_MAC_IP_ROUTE)
2485 return;
2486
2487 /* Identify local route. */
2488 for (tmp_pi = bgp_dest_get_bgp_path_info(dest); tmp_pi;
2489 tmp_pi = tmp_pi->next) {
2490 if (tmp_pi->peer == bgp->peer_self &&
2491 tmp_pi->type == ZEBRA_ROUTE_BGP &&
2492 tmp_pi->sub_type == BGP_ROUTE_STATIC)
2493 break;
2494 }
2495
2496 if (!tmp_pi)
2497 return;
2498
2499 bgp_evpn_update_type2_route_entry(bgp, vpn, dest, tmp_pi, __func__);
2500 }
2501
2502 /*
2503 * Update all type-2 (MACIP) local routes for this VNI - these should also
2504 * be scheduled for advertise to peers.
2505 */
2506 static void update_all_type2_routes(struct bgp *bgp, struct bgpevpn *vpn)
2507 {
2508 struct bgp_dest *dest;
2509
2510 /* Walk this VNI's route MAC & IP table and update local type-2
2511 * routes. For any routes updated, update corresponding entry in the
2512 * global table too.
2513 */
2514 for (dest = bgp_table_top(vpn->mac_table); dest;
2515 dest = bgp_route_next(dest))
2516 update_type2_route(bgp, vpn, dest);
2517
2518 for (dest = bgp_table_top(vpn->ip_table); dest;
2519 dest = bgp_route_next(dest))
2520 update_type2_route(bgp, vpn, dest);
2521 }
2522
2523 /*
2524 * Delete all type-2 (MACIP) local routes for this VNI - only from the
2525 * global routing table. These are also scheduled for withdraw from peers.
2526 */
2527 static void delete_global_type2_routes(struct bgp *bgp, struct bgpevpn *vpn)
2528 {
2529 afi_t afi;
2530 safi_t safi;
2531 struct bgp_dest *rddest, *dest;
2532 struct bgp_table *table;
2533 struct bgp_path_info *pi;
2534
2535 afi = AFI_L2VPN;
2536 safi = SAFI_EVPN;
2537
2538 rddest = bgp_node_lookup(bgp->rib[afi][safi],
2539 (struct prefix *)&vpn->prd);
2540 if (rddest) {
2541 table = bgp_dest_get_bgp_table_info(rddest);
2542 for (dest = bgp_table_top(table); dest;
2543 dest = bgp_route_next(dest)) {
2544 const struct prefix_evpn *evp =
2545 (const struct prefix_evpn *)bgp_dest_get_prefix(
2546 dest);
2547
2548 if (evp->prefix.route_type != BGP_EVPN_MAC_IP_ROUTE)
2549 continue;
2550
2551 delete_evpn_route_entry(bgp, afi, safi, dest, &pi);
2552 if (pi)
2553 bgp_process(bgp, dest, afi, safi);
2554 }
2555
2556 /* Unlock RD node. */
2557 bgp_dest_unlock_node(rddest);
2558 }
2559 }
2560
2561 static void delete_vni_type2_route(struct bgp *bgp, struct bgp_dest *dest)
2562 {
2563 struct bgp_path_info *pi;
2564 afi_t afi = AFI_L2VPN;
2565 safi_t safi = SAFI_EVPN;
2566
2567 const struct prefix_evpn *evp =
2568 (const struct prefix_evpn *)bgp_dest_get_prefix(dest);
2569
2570 if (evp->prefix.route_type != BGP_EVPN_MAC_IP_ROUTE)
2571 return;
2572
2573 delete_evpn_route_entry(bgp, afi, safi, dest, &pi);
2574
2575 /* Route entry in local table gets deleted immediately. */
2576 if (pi)
2577 bgp_path_info_reap(dest, pi);
2578 }
2579
2580 static void delete_vni_type2_routes(struct bgp *bgp, struct bgpevpn *vpn)
2581 {
2582 struct bgp_dest *dest;
2583
2584 /* Next, walk this VNI's MAC & IP route table and delete local type-2
2585 * routes.
2586 */
2587 for (dest = bgp_table_top(vpn->mac_table); dest;
2588 dest = bgp_route_next(dest))
2589 delete_vni_type2_route(bgp, dest);
2590
2591 for (dest = bgp_table_top(vpn->ip_table); dest;
2592 dest = bgp_route_next(dest))
2593 delete_vni_type2_route(bgp, dest);
2594 }
2595
2596 /*
2597 * Delete all type-2 (MACIP) local routes for this VNI - from the global
2598 * table as well as the per-VNI route table.
2599 */
2600 static void delete_all_type2_routes(struct bgp *bgp, struct bgpevpn *vpn)
2601 {
2602 /* First, walk the global route table for this VNI's type-2 local
2603 * routes.
2604 * EVPN routes are a 2-level table, first get the RD table.
2605 */
2606 delete_global_type2_routes(bgp, vpn);
2607 delete_vni_type2_routes(bgp, vpn);
2608 }
2609
2610 /*
2611 * Delete all routes in the per-VNI route table.
2612 */
2613 static void delete_all_vni_routes(struct bgp *bgp, struct bgpevpn *vpn)
2614 {
2615 struct bgp_dest *dest;
2616 struct bgp_path_info *pi, *nextpi;
2617
2618 /* Walk this VNI's MAC & IP route table and delete all routes. */
2619 for (dest = bgp_table_top(vpn->mac_table); dest;
2620 dest = bgp_route_next(dest)) {
2621 for (pi = bgp_dest_get_bgp_path_info(dest);
2622 (pi != NULL) && (nextpi = pi->next, 1); pi = nextpi) {
2623 bgp_evpn_remote_ip_hash_del(vpn, pi);
2624 bgp_path_info_delete(dest, pi);
2625 bgp_path_info_reap(dest, pi);
2626 }
2627 }
2628
2629 for (dest = bgp_table_top(vpn->ip_table); dest;
2630 dest = bgp_route_next(dest)) {
2631 for (pi = bgp_dest_get_bgp_path_info(dest);
2632 (pi != NULL) && (nextpi = pi->next, 1); pi = nextpi) {
2633 bgp_path_info_delete(dest, pi);
2634 bgp_path_info_reap(dest, pi);
2635 }
2636 }
2637 }
2638
2639 /* BUM traffic flood mode per-l2-vni */
2640 static int bgp_evpn_vni_flood_mode_get(struct bgp *bgp,
2641 struct bgpevpn *vpn)
2642 {
2643 /* if flooding has been globally disabled per-vni mode is
2644 * not relevant
2645 */
2646 if (bgp->vxlan_flood_ctrl == VXLAN_FLOOD_DISABLED)
2647 return VXLAN_FLOOD_DISABLED;
2648
2649 /* if mcast group ip has been specified we use a PIM-SM MDT */
2650 if (vpn->mcast_grp.s_addr != INADDR_ANY)
2651 return VXLAN_FLOOD_PIM_SM;
2652
2653 /* default is ingress replication */
2654 return VXLAN_FLOOD_HEAD_END_REPL;
2655 }
2656
2657 /*
2658 * Update (and advertise) local routes for a VNI. Invoked upon the VNI
2659 * export RT getting modified or change to tunnel IP. Note that these
2660 * situations need the route in the per-VNI table as well as the global
2661 * table to be updated (as attributes change).
2662 */
2663 int update_routes_for_vni(struct bgp *bgp, struct bgpevpn *vpn)
2664 {
2665 int ret;
2666 struct prefix_evpn p;
2667
2668 update_type1_routes_for_evi(bgp, vpn);
2669
2670 /* Update and advertise the type-3 route (only one) followed by the
2671 * locally learnt type-2 routes (MACIP) - for this VNI.
2672 *
2673 * RT-3 only if doing head-end replication
2674 */
2675 if (bgp_evpn_vni_flood_mode_get(bgp, vpn)
2676 == VXLAN_FLOOD_HEAD_END_REPL) {
2677 build_evpn_type3_prefix(&p, vpn->originator_ip);
2678 ret = update_evpn_route(bgp, vpn, &p, 0, 0, NULL);
2679 if (ret)
2680 return ret;
2681 }
2682
2683 update_all_type2_routes(bgp, vpn);
2684 return 0;
2685 }
2686
2687 /*
2688 * Delete (and withdraw) local routes for specified VNI from the global
2689 * table and per-VNI table. After this, remove all other routes from
2690 * the per-VNI table. Invoked upon the VNI being deleted or EVPN
2691 * (advertise-all-vni) being disabled.
2692 */
2693 static int delete_routes_for_vni(struct bgp *bgp, struct bgpevpn *vpn)
2694 {
2695 int ret;
2696 struct prefix_evpn p;
2697
2698 /* Delete and withdraw locally learnt type-2 routes (MACIP)
2699 * followed by type-3 routes (only one) - for this VNI.
2700 */
2701 delete_all_type2_routes(bgp, vpn);
2702
2703 build_evpn_type3_prefix(&p, vpn->originator_ip);
2704 ret = delete_evpn_route(bgp, vpn, &p);
2705 if (ret)
2706 return ret;
2707
2708 /* Delete all routes from the per-VNI table. */
2709 delete_all_vni_routes(bgp, vpn);
2710 return 0;
2711 }
2712
2713 /*
2714 * There is a flood mcast IP address change. Update the mcast-grp and
2715 * remove the type-3 route if any. A new type-3 route will be generated
2716 * post tunnel_ip update if the new flood mode is head-end-replication.
2717 */
2718 static int bgp_evpn_mcast_grp_change(struct bgp *bgp, struct bgpevpn *vpn,
2719 struct in_addr mcast_grp)
2720 {
2721 struct prefix_evpn p;
2722
2723 vpn->mcast_grp = mcast_grp;
2724
2725 if (is_vni_live(vpn)) {
2726 build_evpn_type3_prefix(&p, vpn->originator_ip);
2727 delete_evpn_route(bgp, vpn, &p);
2728 }
2729
2730 return 0;
2731 }
2732
2733 /*
2734 * There is a tunnel endpoint IP address change for this VNI, delete
2735 * prior type-3 route (if needed) and update.
2736 * Note: Route re-advertisement happens elsewhere after other processing
2737 * other changes.
2738 */
2739 static int handle_tunnel_ip_change(struct bgp *bgp, struct bgpevpn *vpn,
2740 struct in_addr originator_ip)
2741 {
2742 struct prefix_evpn p;
2743
2744 /* If VNI is not live, we only need to update the originator ip */
2745 if (!is_vni_live(vpn)) {
2746 vpn->originator_ip = originator_ip;
2747 return 0;
2748 }
2749
2750 /* Update the tunnel-ip hash */
2751 bgp_tip_del(bgp, &vpn->originator_ip);
2752 bgp_tip_add(bgp, &originator_ip);
2753
2754 /* filter routes as martian nexthop db has changed */
2755 bgp_filter_evpn_routes_upon_martian_nh_change(bgp);
2756
2757 /* Need to withdraw type-3 route as the originator IP is part
2758 * of the key.
2759 */
2760 build_evpn_type3_prefix(&p, vpn->originator_ip);
2761 delete_evpn_route(bgp, vpn, &p);
2762
2763 /* Update the tunnel IP and re-advertise all routes for this VNI. */
2764 vpn->originator_ip = originator_ip;
2765 return 0;
2766 }
2767
2768 static struct bgp_path_info *
2769 bgp_create_evpn_bgp_path_info(struct bgp_path_info *parent_pi,
2770 struct bgp_dest *dest, struct attr *attr)
2771 {
2772 struct attr *attr_new;
2773 struct bgp_path_info *pi;
2774
2775 /* Add (or update) attribute to hash. */
2776 attr_new = bgp_attr_intern(attr);
2777
2778 /* Create new route with its attribute. */
2779 pi = info_make(parent_pi->type, BGP_ROUTE_IMPORTED, 0, parent_pi->peer,
2780 attr_new, dest);
2781 SET_FLAG(pi->flags, BGP_PATH_VALID);
2782 bgp_path_info_extra_get(pi);
2783 pi->extra->parent = bgp_path_info_lock(parent_pi);
2784 bgp_dest_lock_node((struct bgp_dest *)parent_pi->net);
2785 if (parent_pi->extra) {
2786 memcpy(&pi->extra->label, &parent_pi->extra->label,
2787 sizeof(pi->extra->label));
2788 pi->extra->num_labels = parent_pi->extra->num_labels;
2789 pi->extra->igpmetric = parent_pi->extra->igpmetric;
2790 }
2791
2792 bgp_path_info_add(dest, pi);
2793
2794 return pi;
2795 }
2796
2797 /*
2798 * Install route entry into the VRF routing table and invoke route selection.
2799 */
2800 static int install_evpn_route_entry_in_vrf(struct bgp *bgp_vrf,
2801 const struct prefix_evpn *evp,
2802 struct bgp_path_info *parent_pi)
2803 {
2804 struct bgp_dest *dest;
2805 struct bgp_path_info *pi;
2806 struct attr attr;
2807 struct attr *attr_new;
2808 int ret = 0;
2809 struct prefix p;
2810 struct prefix *pp = &p;
2811 afi_t afi = 0;
2812 safi_t safi = 0;
2813 bool new_pi = false;
2814 bool use_l3nhg = false;
2815 bool is_l3nhg_active = false;
2816 char buf1[INET6_ADDRSTRLEN];
2817
2818 memset(pp, 0, sizeof(struct prefix));
2819 ip_prefix_from_evpn_prefix(evp, pp);
2820
2821 if (bgp_debug_zebra(NULL))
2822 zlog_debug(
2823 "vrf %s: import evpn prefix %pFX parent %p flags 0x%x",
2824 vrf_id_to_name(bgp_vrf->vrf_id), evp, parent_pi,
2825 parent_pi->flags);
2826
2827 /* Create (or fetch) route within the VRF. */
2828 /* NOTE: There is no RD here. */
2829 if (is_evpn_prefix_ipaddr_v4(evp)) {
2830 afi = AFI_IP;
2831 safi = SAFI_UNICAST;
2832 dest = bgp_node_get(bgp_vrf->rib[afi][safi], pp);
2833 } else if (is_evpn_prefix_ipaddr_v6(evp)) {
2834 afi = AFI_IP6;
2835 safi = SAFI_UNICAST;
2836 dest = bgp_node_get(bgp_vrf->rib[afi][safi], pp);
2837 } else
2838 return 0;
2839
2840 /* EVPN routes currently only support a IPv4 next hop which corresponds
2841 * to the remote VTEP. When importing into a VRF, if it is IPv6 host
2842 * or prefix route, we have to convert the next hop to an IPv4-mapped
2843 * address for the rest of the code to flow through. In the case of IPv4,
2844 * make sure to set the flag for next hop attribute.
2845 */
2846 attr = *parent_pi->attr;
2847 if (attr.evpn_overlay.type != OVERLAY_INDEX_GATEWAY_IP) {
2848 if (afi == AFI_IP6)
2849 evpn_convert_nexthop_to_ipv6(&attr);
2850 else {
2851 attr.nexthop = attr.mp_nexthop_global_in;
2852 attr.flag |= ATTR_FLAG_BIT(BGP_ATTR_NEXT_HOP);
2853 }
2854 } else {
2855
2856 /*
2857 * If gateway IP overlay index is specified in the NLRI of
2858 * EVPN RT-5, this gateway IP should be used as the nexthop
2859 * for the prefix in the VRF
2860 */
2861 if (bgp_debug_zebra(NULL)) {
2862 zlog_debug(
2863 "Install gateway IP %s as nexthop for prefix %pFX in vrf %s",
2864 inet_ntop(pp->family, &attr.evpn_overlay.gw_ip,
2865 buf1, sizeof(buf1)), pp,
2866 vrf_id_to_name(bgp_vrf->vrf_id));
2867 }
2868
2869 if (afi == AFI_IP6) {
2870 memcpy(&attr.mp_nexthop_global,
2871 &attr.evpn_overlay.gw_ip.ipaddr_v6,
2872 sizeof(struct in6_addr));
2873 attr.mp_nexthop_len = IPV6_MAX_BYTELEN;
2874 } else {
2875 attr.nexthop = attr.evpn_overlay.gw_ip.ipaddr_v4;
2876 attr.flag |= ATTR_FLAG_BIT(BGP_ATTR_NEXT_HOP);
2877 }
2878 }
2879
2880 bgp_evpn_es_vrf_use_nhg(bgp_vrf, &parent_pi->attr->esi, &use_l3nhg,
2881 &is_l3nhg_active, NULL);
2882 if (use_l3nhg)
2883 attr.es_flags |= ATTR_ES_L3_NHG_USE;
2884 if (is_l3nhg_active)
2885 attr.es_flags |= ATTR_ES_L3_NHG_ACTIVE;
2886
2887 /* Check if route entry is already present. */
2888 for (pi = bgp_dest_get_bgp_path_info(dest); pi; pi = pi->next)
2889 if (pi->extra
2890 && (struct bgp_path_info *)pi->extra->parent == parent_pi)
2891 break;
2892
2893 if (!pi) {
2894 pi = bgp_create_evpn_bgp_path_info(parent_pi, dest, &attr);
2895 new_pi = true;
2896 } else {
2897 if (attrhash_cmp(pi->attr, &attr)
2898 && !CHECK_FLAG(pi->flags, BGP_PATH_REMOVED)) {
2899 bgp_dest_unlock_node(dest);
2900 return 0;
2901 }
2902 /* The attribute has changed. */
2903 /* Add (or update) attribute to hash. */
2904 attr_new = bgp_attr_intern(&attr);
2905
2906 /* Restore route, if needed. */
2907 if (CHECK_FLAG(pi->flags, BGP_PATH_REMOVED))
2908 bgp_path_info_restore(dest, pi);
2909
2910 /* Mark if nexthop has changed. */
2911 if ((afi == AFI_IP
2912 && !IPV4_ADDR_SAME(&pi->attr->nexthop, &attr_new->nexthop))
2913 || (afi == AFI_IP6
2914 && !IPV6_ADDR_SAME(&pi->attr->mp_nexthop_global,
2915 &attr_new->mp_nexthop_global)))
2916 SET_FLAG(pi->flags, BGP_PATH_IGP_CHANGED);
2917
2918 bgp_path_info_set_flag(dest, pi, BGP_PATH_ATTR_CHANGED);
2919 /* Unintern existing, set to new. */
2920 bgp_attr_unintern(&pi->attr);
2921 pi->attr = attr_new;
2922 pi->uptime = monotime(NULL);
2923 }
2924
2925 /* Gateway IP nexthop should be resolved */
2926 if (attr.evpn_overlay.type == OVERLAY_INDEX_GATEWAY_IP) {
2927 if (bgp_find_or_add_nexthop(bgp_vrf, bgp_vrf, afi, safi, pi,
2928 NULL, 0, NULL))
2929 bgp_path_info_set_flag(dest, pi, BGP_PATH_VALID);
2930 else {
2931 if (BGP_DEBUG(nht, NHT)) {
2932 inet_ntop(pp->family,
2933 &attr.evpn_overlay.gw_ip,
2934 buf1, sizeof(buf1));
2935 zlog_debug("%s: gateway IP NH unresolved",
2936 buf1);
2937 }
2938 bgp_path_info_unset_flag(dest, pi, BGP_PATH_VALID);
2939 }
2940 } else {
2941
2942 /* as it is an importation, change nexthop */
2943 bgp_path_info_set_flag(dest, pi, BGP_PATH_ANNC_NH_SELF);
2944 }
2945
2946 /* Link path to evpn nexthop */
2947 bgp_evpn_path_nh_add(bgp_vrf, pi);
2948
2949 bgp_aggregate_increment(bgp_vrf, bgp_dest_get_prefix(dest), pi, afi,
2950 safi);
2951
2952 /* Perform route selection and update zebra, if required. */
2953 bgp_process(bgp_vrf, dest, afi, safi);
2954
2955 /* Process for route leaking. */
2956 vpn_leak_from_vrf_update(bgp_get_default(), bgp_vrf, pi);
2957
2958 bgp_dest_unlock_node(dest);
2959
2960 if (bgp_debug_zebra(NULL))
2961 zlog_debug("... %s pi dest %p (l %d) pi %p (l %d, f 0x%x)",
2962 new_pi ? "new" : "update", dest,
2963 bgp_dest_get_lock_count(dest), pi, pi->lock,
2964 pi->flags);
2965
2966 return ret;
2967 }
2968
2969 /*
2970 * Common handling for vni route tables install/selection.
2971 */
2972 static int install_evpn_route_entry_in_vni_common(
2973 struct bgp *bgp, struct bgpevpn *vpn, const struct prefix_evpn *p,
2974 struct bgp_dest *dest, struct bgp_path_info *parent_pi)
2975 {
2976 struct bgp_path_info *pi;
2977 struct bgp_path_info *local_pi;
2978 struct attr *attr_new;
2979 int ret;
2980 bool old_local_es = false;
2981 bool new_local_es;
2982
2983 /* Check if route entry is already present. */
2984 for (pi = bgp_dest_get_bgp_path_info(dest); pi; pi = pi->next)
2985 if (pi->extra
2986 && (struct bgp_path_info *)pi->extra->parent == parent_pi)
2987 break;
2988
2989 if (!pi) {
2990 /* Create an info */
2991 pi = bgp_create_evpn_bgp_path_info(parent_pi, dest,
2992 parent_pi->attr);
2993
2994 if (p->prefix.route_type == BGP_EVPN_MAC_IP_ROUTE) {
2995 if (is_evpn_type2_dest_ipaddr_none(dest))
2996 evpn_type2_path_info_set_ip(
2997 pi, p->prefix.macip_addr.ip);
2998 else
2999 evpn_type2_path_info_set_mac(
3000 pi, p->prefix.macip_addr.mac);
3001 }
3002
3003 new_local_es = bgp_evpn_attr_is_local_es(pi->attr);
3004 } else {
3005 /* Return early if attributes haven't changed
3006 * and dest isn't flagged for removal.
3007 * dest will be unlocked by either
3008 * install_evpn_route_entry_in_vni_mac() or
3009 * install_evpn_route_entry_in_vni_ip()
3010 */
3011 if (attrhash_cmp(pi->attr, parent_pi->attr) &&
3012 !CHECK_FLAG(pi->flags, BGP_PATH_REMOVED))
3013 return 0;
3014 /* The attribute has changed. */
3015 /* Add (or update) attribute to hash. */
3016 attr_new = bgp_attr_intern(parent_pi->attr);
3017
3018 /* Restore route, if needed. */
3019 if (CHECK_FLAG(pi->flags, BGP_PATH_REMOVED))
3020 bgp_path_info_restore(dest, pi);
3021
3022 /* Mark if nexthop has changed. */
3023 if (!IPV4_ADDR_SAME(&pi->attr->nexthop, &attr_new->nexthop))
3024 SET_FLAG(pi->flags, BGP_PATH_IGP_CHANGED);
3025
3026 old_local_es = bgp_evpn_attr_is_local_es(pi->attr);
3027 new_local_es = bgp_evpn_attr_is_local_es(attr_new);
3028 /* If ESI is different or if its type has changed we
3029 * need to reinstall the path in zebra
3030 */
3031 if ((old_local_es != new_local_es)
3032 || memcmp(&pi->attr->esi, &attr_new->esi,
3033 sizeof(attr_new->esi))) {
3034
3035 if (BGP_DEBUG(evpn_mh, EVPN_MH_RT))
3036 zlog_debug("VNI %d path %pFX chg to %s es",
3037 vpn->vni, &pi->net->p,
3038 new_local_es ? "local"
3039 : "non-local");
3040 bgp_path_info_set_flag(dest, pi, BGP_PATH_ATTR_CHANGED);
3041 }
3042
3043 /* Unintern existing, set to new. */
3044 bgp_attr_unintern(&pi->attr);
3045 pi->attr = attr_new;
3046 pi->uptime = monotime(NULL);
3047 }
3048
3049 /* Add this route to remote IP hashtable */
3050 bgp_evpn_remote_ip_hash_add(vpn, pi);
3051
3052 /* Perform route selection and update zebra, if required. */
3053 ret = evpn_route_select_install(bgp, vpn, dest);
3054
3055 /* if the best path is a local path with a non-zero ES
3056 * sync info against the local path may need to be updated
3057 * when a remote path is added/updated (including changes
3058 * from sync-path to remote-path)
3059 */
3060 local_pi = bgp_evpn_route_get_local_path(bgp, dest);
3061 if (local_pi && (old_local_es || new_local_es))
3062 bgp_evpn_update_type2_route_entry(bgp, vpn, dest, local_pi,
3063 __func__);
3064
3065 return ret;
3066 }
3067
3068 /*
3069 * Common handling for vni route tables uninstall/selection.
3070 */
3071 static int uninstall_evpn_route_entry_in_vni_common(
3072 struct bgp *bgp, struct bgpevpn *vpn, const struct prefix_evpn *p,
3073 struct bgp_dest *dest, struct bgp_path_info *parent_pi)
3074 {
3075 struct bgp_path_info *pi;
3076 struct bgp_path_info *local_pi;
3077 int ret;
3078
3079 /* Find matching route entry. */
3080 for (pi = bgp_dest_get_bgp_path_info(dest); pi; pi = pi->next)
3081 if (pi->extra &&
3082 (struct bgp_path_info *)pi->extra->parent == parent_pi)
3083 break;
3084
3085 if (!pi)
3086 return 0;
3087
3088 bgp_evpn_remote_ip_hash_del(vpn, pi);
3089
3090 /* Mark entry for deletion */
3091 bgp_path_info_delete(dest, pi);
3092
3093 /* Perform route selection and update zebra, if required. */
3094 ret = evpn_route_select_install(bgp, vpn, dest);
3095
3096 /* if the best path is a local path with a non-zero ES
3097 * sync info against the local path may need to be updated
3098 * when a remote path is deleted
3099 */
3100 local_pi = bgp_evpn_route_get_local_path(bgp, dest);
3101 if (local_pi && bgp_evpn_attr_is_local_es(local_pi->attr))
3102 bgp_evpn_update_type2_route_entry(bgp, vpn, dest, local_pi,
3103 __func__);
3104
3105 return ret;
3106 }
3107
3108 /*
3109 * Install route entry into VNI IP table and invoke route selection.
3110 */
3111 static int install_evpn_route_entry_in_vni_ip(struct bgp *bgp,
3112 struct bgpevpn *vpn,
3113 const struct prefix_evpn *p,
3114 struct bgp_path_info *parent_pi)
3115 {
3116 int ret;
3117 struct bgp_dest *dest;
3118
3119 /* Ignore MAC Only Type-2 */
3120 if ((p->prefix.route_type == BGP_EVPN_MAC_IP_ROUTE) &&
3121 (is_evpn_prefix_ipaddr_none(p) == true))
3122 return 0;
3123
3124 /* Create (or fetch) route within the VNI IP table. */
3125 dest = bgp_evpn_vni_ip_node_get(vpn->ip_table, p, parent_pi);
3126
3127 ret = install_evpn_route_entry_in_vni_common(bgp, vpn, p, dest,
3128 parent_pi);
3129
3130 bgp_dest_unlock_node(dest);
3131
3132 return ret;
3133 }
3134
3135 /*
3136 * Install route entry into VNI MAC table and invoke route selection.
3137 */
3138 static int install_evpn_route_entry_in_vni_mac(struct bgp *bgp,
3139 struct bgpevpn *vpn,
3140 const struct prefix_evpn *p,
3141 struct bgp_path_info *parent_pi)
3142 {
3143 int ret;
3144 struct bgp_dest *dest;
3145
3146 /* Only type-2 routes go into this table */
3147 if (p->prefix.route_type != BGP_EVPN_MAC_IP_ROUTE)
3148 return 0;
3149
3150 /* Create (or fetch) route within the VNI MAC table. */
3151 dest = bgp_evpn_vni_mac_node_get(vpn->mac_table, p, parent_pi);
3152
3153 ret = install_evpn_route_entry_in_vni_common(bgp, vpn, p, dest,
3154 parent_pi);
3155
3156 bgp_dest_unlock_node(dest);
3157
3158 return ret;
3159 }
3160
3161 /*
3162 * Uninstall route entry from VNI IP table and invoke route selection.
3163 */
3164 static int uninstall_evpn_route_entry_in_vni_ip(struct bgp *bgp,
3165 struct bgpevpn *vpn,
3166 const struct prefix_evpn *p,
3167 struct bgp_path_info *parent_pi)
3168 {
3169 int ret;
3170 struct bgp_dest *dest;
3171
3172 /* Ignore MAC Only Type-2 */
3173 if ((p->prefix.route_type == BGP_EVPN_MAC_IP_ROUTE) &&
3174 (is_evpn_prefix_ipaddr_none(p) == true))
3175 return 0;
3176
3177 /* Locate route within the VNI IP table. */
3178 dest = bgp_evpn_vni_ip_node_lookup(vpn->ip_table, p, parent_pi);
3179 if (!dest)
3180 return 0;
3181
3182 ret = uninstall_evpn_route_entry_in_vni_common(bgp, vpn, p, dest,
3183 parent_pi);
3184
3185 bgp_dest_unlock_node(dest);
3186
3187 return ret;
3188 }
3189
3190 /*
3191 * Uninstall route entry from VNI IP table and invoke route selection.
3192 */
3193 static int
3194 uninstall_evpn_route_entry_in_vni_mac(struct bgp *bgp, struct bgpevpn *vpn,
3195 const struct prefix_evpn *p,
3196 struct bgp_path_info *parent_pi)
3197 {
3198 int ret;
3199 struct bgp_dest *dest;
3200
3201 /* Only type-2 routes go into this table */
3202 if (p->prefix.route_type != BGP_EVPN_MAC_IP_ROUTE)
3203 return 0;
3204
3205 /* Locate route within the VNI MAC table. */
3206 dest = bgp_evpn_vni_mac_node_lookup(vpn->mac_table, p, parent_pi);
3207 if (!dest)
3208 return 0;
3209
3210 ret = uninstall_evpn_route_entry_in_vni_common(bgp, vpn, p, dest,
3211 parent_pi);
3212
3213 bgp_dest_unlock_node(dest);
3214
3215 return ret;
3216 }
3217 /*
3218 * Uninstall route entry from the VRF routing table and send message
3219 * to zebra, if appropriate.
3220 */
3221 static int uninstall_evpn_route_entry_in_vrf(struct bgp *bgp_vrf,
3222 const struct prefix_evpn *evp,
3223 struct bgp_path_info *parent_pi)
3224 {
3225 struct bgp_dest *dest;
3226 struct bgp_path_info *pi;
3227 int ret = 0;
3228 struct prefix p;
3229 struct prefix *pp = &p;
3230 afi_t afi = 0;
3231 safi_t safi = 0;
3232
3233 memset(pp, 0, sizeof(struct prefix));
3234 ip_prefix_from_evpn_prefix(evp, pp);
3235
3236 if (bgp_debug_zebra(NULL))
3237 zlog_debug(
3238 "vrf %s: unimport evpn prefix %pFX parent %p flags 0x%x",
3239 vrf_id_to_name(bgp_vrf->vrf_id), evp, parent_pi,
3240 parent_pi->flags);
3241
3242 /* Locate route within the VRF. */
3243 /* NOTE: There is no RD here. */
3244 if (is_evpn_prefix_ipaddr_v4(evp)) {
3245 afi = AFI_IP;
3246 safi = SAFI_UNICAST;
3247 dest = bgp_node_lookup(bgp_vrf->rib[afi][safi], pp);
3248 } else {
3249 afi = AFI_IP6;
3250 safi = SAFI_UNICAST;
3251 dest = bgp_node_lookup(bgp_vrf->rib[afi][safi], pp);
3252 }
3253
3254 if (!dest)
3255 return 0;
3256
3257 /* Find matching route entry. */
3258 for (pi = bgp_dest_get_bgp_path_info(dest); pi; pi = pi->next)
3259 if (pi->extra
3260 && (struct bgp_path_info *)pi->extra->parent == parent_pi)
3261 break;
3262
3263 if (!pi) {
3264 bgp_dest_unlock_node(dest);
3265 return 0;
3266 }
3267
3268 if (bgp_debug_zebra(NULL))
3269 zlog_debug("... delete dest %p (l %d) pi %p (l %d, f 0x%x)",
3270 dest, bgp_dest_get_lock_count(dest), pi, pi->lock,
3271 pi->flags);
3272
3273 /* Process for route leaking. */
3274 vpn_leak_from_vrf_withdraw(bgp_get_default(), bgp_vrf, pi);
3275
3276 bgp_aggregate_decrement(bgp_vrf, bgp_dest_get_prefix(dest), pi, afi,
3277 safi);
3278
3279 /* Mark entry for deletion */
3280 bgp_path_info_delete(dest, pi);
3281
3282 /* Unlink path to evpn nexthop */
3283 bgp_evpn_path_nh_del(bgp_vrf, pi);
3284
3285 /* Perform route selection and update zebra, if required. */
3286 bgp_process(bgp_vrf, dest, afi, safi);
3287
3288 /* Unlock route node. */
3289 bgp_dest_unlock_node(dest);
3290
3291 return ret;
3292 }
3293
3294 /*
3295 * Install route entry into the VNI routing tables.
3296 */
3297 static int install_evpn_route_entry(struct bgp *bgp, struct bgpevpn *vpn,
3298 const struct prefix_evpn *p,
3299 struct bgp_path_info *parent_pi)
3300 {
3301 int ret = 0;
3302
3303 if (bgp_debug_update(parent_pi->peer, NULL, NULL, 1))
3304 zlog_debug(
3305 "%s (%u): Installing EVPN %pFX route in VNI %u IP/MAC table",
3306 vrf_id_to_name(bgp->vrf_id), bgp->vrf_id, p, vpn->vni);
3307
3308 ret = install_evpn_route_entry_in_vni_mac(bgp, vpn, p, parent_pi);
3309
3310 if (ret) {
3311 flog_err(
3312 EC_BGP_EVPN_FAIL,
3313 "%s (%u): Failed to install EVPN %pFX route in VNI %u MAC table",
3314 vrf_id_to_name(bgp->vrf_id), bgp->vrf_id, p, vpn->vni);
3315
3316 return ret;
3317 }
3318
3319 ret = install_evpn_route_entry_in_vni_ip(bgp, vpn, p, parent_pi);
3320
3321 if (ret) {
3322 flog_err(
3323 EC_BGP_EVPN_FAIL,
3324 "%s (%u): Failed to install EVPN %pFX route in VNI %u IP table",
3325 vrf_id_to_name(bgp->vrf_id), bgp->vrf_id, p, vpn->vni);
3326
3327 return ret;
3328 }
3329
3330 return ret;
3331 }
3332
3333 /*
3334 * Uninstall route entry from the VNI routing tables.
3335 */
3336 static int uninstall_evpn_route_entry(struct bgp *bgp, struct bgpevpn *vpn,
3337 const struct prefix_evpn *p,
3338 struct bgp_path_info *parent_pi)
3339 {
3340 int ret = 0;
3341
3342 if (bgp_debug_update(parent_pi->peer, NULL, NULL, 1))
3343 zlog_debug(
3344 "%s (%u): Uninstalling EVPN %pFX route from VNI %u IP/MAC table",
3345 vrf_id_to_name(bgp->vrf_id), bgp->vrf_id, p, vpn->vni);
3346
3347 ret = uninstall_evpn_route_entry_in_vni_ip(bgp, vpn, p, parent_pi);
3348
3349 if (ret) {
3350 flog_err(
3351 EC_BGP_EVPN_FAIL,
3352 "%s (%u): Failed to uninstall EVPN %pFX route from VNI %u IP table",
3353 vrf_id_to_name(bgp->vrf_id), bgp->vrf_id, p, vpn->vni);
3354
3355 return ret;
3356 }
3357
3358 ret = uninstall_evpn_route_entry_in_vni_mac(bgp, vpn, p, parent_pi);
3359
3360 if (ret) {
3361 flog_err(
3362 EC_BGP_EVPN_FAIL,
3363 "%s (%u): Failed to uninstall EVPN %pFX route from VNI %u MAC table",
3364 vrf_id_to_name(bgp->vrf_id), bgp->vrf_id, p, vpn->vni);
3365
3366 return ret;
3367 }
3368
3369 return ret;
3370 }
3371
3372 /*
3373 * Given a route entry and a VRF, see if this route entry should be
3374 * imported into the VRF i.e., RTs match.
3375 */
3376 static int is_route_matching_for_vrf(struct bgp *bgp_vrf,
3377 struct bgp_path_info *pi)
3378 {
3379 struct attr *attr = pi->attr;
3380 struct ecommunity *ecom;
3381 uint32_t i;
3382
3383 assert(attr);
3384 /* Route should have valid RT to be even considered. */
3385 if (!(attr->flag & ATTR_FLAG_BIT(BGP_ATTR_EXT_COMMUNITIES)))
3386 return 0;
3387
3388 ecom = bgp_attr_get_ecommunity(attr);
3389 if (!ecom || !ecom->size)
3390 return 0;
3391
3392 /* For each extended community RT, see if it matches this VNI. If any RT
3393 * matches, we're done.
3394 */
3395 for (i = 0; i < ecom->size; i++) {
3396 uint8_t *pnt;
3397 uint8_t type, sub_type;
3398 struct ecommunity_val *eval;
3399 struct ecommunity_val eval_tmp;
3400 struct vrf_irt_node *irt;
3401
3402 /* Only deal with RTs */
3403 pnt = (ecom->val + (i * ecom->unit_size));
3404 eval = (struct ecommunity_val *)(ecom->val
3405 + (i * ecom->unit_size));
3406 type = *pnt++;
3407 sub_type = *pnt++;
3408 if (sub_type != ECOMMUNITY_ROUTE_TARGET)
3409 continue;
3410
3411 /* See if this RT matches specified VNIs import RTs */
3412 irt = lookup_vrf_import_rt(eval);
3413 if (irt)
3414 if (is_vrf_present_in_irt_vrfs(irt->vrfs, bgp_vrf))
3415 return 1;
3416
3417 /* Also check for non-exact match. In this, we mask out the AS
3418 * and
3419 * only check on the local-admin sub-field. This is to
3420 * facilitate using
3421 * VNI as the RT for EBGP peering too.
3422 */
3423 irt = NULL;
3424 if (type == ECOMMUNITY_ENCODE_AS
3425 || type == ECOMMUNITY_ENCODE_AS4
3426 || type == ECOMMUNITY_ENCODE_IP) {
3427 memcpy(&eval_tmp, eval, ecom->unit_size);
3428 mask_ecom_global_admin(&eval_tmp, eval);
3429 irt = lookup_vrf_import_rt(&eval_tmp);
3430 }
3431 if (irt)
3432 if (is_vrf_present_in_irt_vrfs(irt->vrfs, bgp_vrf))
3433 return 1;
3434 }
3435
3436 return 0;
3437 }
3438
3439 /*
3440 * Given a route entry and a VNI, see if this route entry should be
3441 * imported into the VNI i.e., RTs match.
3442 */
3443 static int is_route_matching_for_vni(struct bgp *bgp, struct bgpevpn *vpn,
3444 struct bgp_path_info *pi)
3445 {
3446 struct attr *attr = pi->attr;
3447 struct ecommunity *ecom;
3448 uint32_t i;
3449
3450 assert(attr);
3451 /* Route should have valid RT to be even considered. */
3452 if (!(attr->flag & ATTR_FLAG_BIT(BGP_ATTR_EXT_COMMUNITIES)))
3453 return 0;
3454
3455 ecom = bgp_attr_get_ecommunity(attr);
3456 if (!ecom || !ecom->size)
3457 return 0;
3458
3459 /* For each extended community RT, see if it matches this VNI. If any RT
3460 * matches, we're done.
3461 */
3462 for (i = 0; i < ecom->size; i++) {
3463 uint8_t *pnt;
3464 uint8_t type, sub_type;
3465 struct ecommunity_val *eval;
3466 struct ecommunity_val eval_tmp;
3467 struct irt_node *irt;
3468
3469 /* Only deal with RTs */
3470 pnt = (ecom->val + (i * ecom->unit_size));
3471 eval = (struct ecommunity_val *)(ecom->val
3472 + (i * ecom->unit_size));
3473 type = *pnt++;
3474 sub_type = *pnt++;
3475 if (sub_type != ECOMMUNITY_ROUTE_TARGET)
3476 continue;
3477
3478 /* See if this RT matches specified VNIs import RTs */
3479 irt = lookup_import_rt(bgp, eval);
3480 if (irt)
3481 if (is_vni_present_in_irt_vnis(irt->vnis, vpn))
3482 return 1;
3483
3484 /* Also check for non-exact match. In this, we mask out the AS
3485 * and
3486 * only check on the local-admin sub-field. This is to
3487 * facilitate using
3488 * VNI as the RT for EBGP peering too.
3489 */
3490 irt = NULL;
3491 if (type == ECOMMUNITY_ENCODE_AS
3492 || type == ECOMMUNITY_ENCODE_AS4
3493 || type == ECOMMUNITY_ENCODE_IP) {
3494 memcpy(&eval_tmp, eval, ecom->unit_size);
3495 mask_ecom_global_admin(&eval_tmp, eval);
3496 irt = lookup_import_rt(bgp, &eval_tmp);
3497 }
3498 if (irt)
3499 if (is_vni_present_in_irt_vnis(irt->vnis, vpn))
3500 return 1;
3501 }
3502
3503 return 0;
3504 }
3505
3506 /* This API will scan evpn routes for checking attribute's rmac
3507 * macthes with bgp instance router mac. It avoid installing
3508 * route into bgp vrf table and remote rmac in bridge table.
3509 */
3510 static int bgp_evpn_route_rmac_self_check(struct bgp *bgp_vrf,
3511 const struct prefix_evpn *evp,
3512 struct bgp_path_info *pi)
3513 {
3514 /* evpn route could have learnt prior to L3vni has come up,
3515 * perform rmac check before installing route and
3516 * remote router mac.
3517 * The route will be removed from global bgp table once
3518 * SVI comes up with MAC and stored in hash, triggers
3519 * bgp_mac_rescan_all_evpn_tables.
3520 */
3521 if (memcmp(&bgp_vrf->rmac, &pi->attr->rmac, ETH_ALEN) == 0) {
3522 if (bgp_debug_update(pi->peer, NULL, NULL, 1)) {
3523 char attr_str[BUFSIZ] = {0};
3524
3525 bgp_dump_attr(pi->attr, attr_str, sizeof(attr_str));
3526
3527 zlog_debug(
3528 "%s: bgp %u prefix %pFX with attr %s - DENIED due to self mac",
3529 __func__, bgp_vrf->vrf_id, evp, attr_str);
3530 }
3531
3532 return 1;
3533 }
3534
3535 return 0;
3536 }
3537
3538 /* don't import hosts that are locally attached */
3539 static inline bool
3540 bgp_evpn_skip_vrf_import_of_local_es(struct bgp *bgp_vrf,
3541 const struct prefix_evpn *evp,
3542 struct bgp_path_info *pi, int install)
3543 {
3544 esi_t *esi;
3545
3546 if (evp->prefix.route_type == BGP_EVPN_MAC_IP_ROUTE) {
3547 esi = bgp_evpn_attr_get_esi(pi->attr);
3548
3549 /* Don't import routes that point to a local destination */
3550 if (bgp_evpn_attr_is_local_es(pi->attr)) {
3551 if (BGP_DEBUG(evpn_mh, EVPN_MH_RT)) {
3552 char esi_buf[ESI_STR_LEN];
3553
3554 zlog_debug(
3555 "vrf %s of evpn prefix %pFX skipped, local es %s",
3556 install ? "import" : "unimport", evp,
3557 esi_to_str(esi, esi_buf,
3558 sizeof(esi_buf)));
3559 }
3560 return true;
3561 }
3562 }
3563 return false;
3564 }
3565
3566 /*
3567 * Install or uninstall a mac-ip route in the provided vrf if
3568 * there is a rt match
3569 */
3570 int bgp_evpn_route_entry_install_if_vrf_match(struct bgp *bgp_vrf,
3571 struct bgp_path_info *pi,
3572 int install)
3573 {
3574 int ret = 0;
3575 const struct prefix_evpn *evp =
3576 (const struct prefix_evpn *)bgp_dest_get_prefix(pi->net);
3577
3578 /* Consider "valid" remote routes applicable for
3579 * this VRF.
3580 */
3581 if (!(CHECK_FLAG(pi->flags, BGP_PATH_VALID)
3582 && pi->type == ZEBRA_ROUTE_BGP
3583 && pi->sub_type == BGP_ROUTE_NORMAL))
3584 return 0;
3585
3586 if (is_route_matching_for_vrf(bgp_vrf, pi)) {
3587 if (bgp_evpn_route_rmac_self_check(bgp_vrf, evp, pi))
3588 return 0;
3589
3590 /* don't import hosts that are locally attached */
3591 if (install && bgp_evpn_skip_vrf_import_of_local_es(
3592 bgp_vrf, evp, pi, install))
3593 return 0;
3594
3595 if (install)
3596 ret = install_evpn_route_entry_in_vrf(bgp_vrf, evp, pi);
3597 else
3598 ret = uninstall_evpn_route_entry_in_vrf(bgp_vrf, evp,
3599 pi);
3600
3601 if (ret)
3602 flog_err(EC_BGP_EVPN_FAIL,
3603 "Failed to %s EVPN %pFX route in VRF %s",
3604 install ? "install" : "uninstall", evp,
3605 vrf_id_to_name(bgp_vrf->vrf_id));
3606 }
3607
3608 return ret;
3609 }
3610
3611 /*
3612 * Install or uninstall mac-ip routes are appropriate for this
3613 * particular VRF.
3614 */
3615 static int install_uninstall_routes_for_vrf(struct bgp *bgp_vrf, int install)
3616 {
3617 afi_t afi;
3618 safi_t safi;
3619 struct bgp_dest *rd_dest, *dest;
3620 struct bgp_table *table;
3621 struct bgp_path_info *pi;
3622 int ret;
3623 struct bgp *bgp_evpn = NULL;
3624
3625 afi = AFI_L2VPN;
3626 safi = SAFI_EVPN;
3627 bgp_evpn = bgp_get_evpn();
3628 if (!bgp_evpn)
3629 return -1;
3630
3631 /* Walk entire global routing table and evaluate routes which could be
3632 * imported into this VRF. Note that we need to loop through all global
3633 * routes to determine which route matches the import rt on vrf
3634 */
3635 for (rd_dest = bgp_table_top(bgp_evpn->rib[afi][safi]); rd_dest;
3636 rd_dest = bgp_route_next(rd_dest)) {
3637 table = bgp_dest_get_bgp_table_info(rd_dest);
3638 if (!table)
3639 continue;
3640
3641 for (dest = bgp_table_top(table); dest;
3642 dest = bgp_route_next(dest)) {
3643 const struct prefix_evpn *evp =
3644 (const struct prefix_evpn *)bgp_dest_get_prefix(
3645 dest);
3646
3647 /* if not mac-ip route skip this route */
3648 if (!(evp->prefix.route_type == BGP_EVPN_MAC_IP_ROUTE
3649 || evp->prefix.route_type
3650 == BGP_EVPN_IP_PREFIX_ROUTE))
3651 continue;
3652
3653 /* if not a mac+ip route skip this route */
3654 if (!(is_evpn_prefix_ipaddr_v4(evp)
3655 || is_evpn_prefix_ipaddr_v6(evp)))
3656 continue;
3657
3658 for (pi = bgp_dest_get_bgp_path_info(dest); pi;
3659 pi = pi->next) {
3660 ret = bgp_evpn_route_entry_install_if_vrf_match(
3661 bgp_vrf, pi, install);
3662 if (ret)
3663 return ret;
3664 }
3665 }
3666 }
3667
3668 return 0;
3669 }
3670
3671 /*
3672 * Install or uninstall routes of specified type that are appropriate for this
3673 * particular VNI.
3674 */
3675 static int install_uninstall_routes_for_vni(struct bgp *bgp,
3676 struct bgpevpn *vpn,
3677 bgp_evpn_route_type rtype,
3678 int install)
3679 {
3680 afi_t afi;
3681 safi_t safi;
3682 struct bgp_dest *rd_dest, *dest;
3683 struct bgp_table *table;
3684 struct bgp_path_info *pi;
3685 int ret;
3686
3687 afi = AFI_L2VPN;
3688 safi = SAFI_EVPN;
3689
3690 /* Walk entire global routing table and evaluate routes which could be
3691 * imported into this VPN. Note that we cannot just look at the routes
3692 * for
3693 * the VNI's RD - remote routes applicable for this VNI could have any
3694 * RD.
3695 */
3696 /* EVPN routes are a 2-level table. */
3697 for (rd_dest = bgp_table_top(bgp->rib[afi][safi]); rd_dest;
3698 rd_dest = bgp_route_next(rd_dest)) {
3699 table = bgp_dest_get_bgp_table_info(rd_dest);
3700 if (!table)
3701 continue;
3702
3703 for (dest = bgp_table_top(table); dest;
3704 dest = bgp_route_next(dest)) {
3705 const struct prefix_evpn *evp =
3706 (const struct prefix_evpn *)bgp_dest_get_prefix(
3707 dest);
3708
3709 if (evp->prefix.route_type != rtype)
3710 continue;
3711
3712 for (pi = bgp_dest_get_bgp_path_info(dest); pi;
3713 pi = pi->next) {
3714 /* Consider "valid" remote routes applicable for
3715 * this VNI. */
3716 if (!(CHECK_FLAG(pi->flags, BGP_PATH_VALID)
3717 && pi->type == ZEBRA_ROUTE_BGP
3718 && pi->sub_type == BGP_ROUTE_NORMAL))
3719 continue;
3720
3721 if (is_route_matching_for_vni(bgp, vpn, pi)) {
3722 if (install)
3723 ret = install_evpn_route_entry(
3724 bgp, vpn, evp, pi);
3725 else
3726 ret = uninstall_evpn_route_entry(
3727 bgp, vpn, evp, pi);
3728
3729 if (ret) {
3730 flog_err(
3731 EC_BGP_EVPN_FAIL,
3732 "%u: Failed to %s EVPN %s route in VNI %u",
3733 bgp->vrf_id,
3734 install ? "install"
3735 : "uninstall",
3736 rtype == BGP_EVPN_MAC_IP_ROUTE
3737 ? "MACIP"
3738 : "IMET",
3739 vpn->vni);
3740
3741 bgp_dest_unlock_node(rd_dest);
3742 bgp_dest_unlock_node(dest);
3743 return ret;
3744 }
3745 }
3746 }
3747 }
3748 }
3749
3750 return 0;
3751 }
3752
3753 /* Install any existing remote routes applicable for this VRF into VRF RIB. This
3754 * is invoked upon l3vni-add or l3vni import rt change
3755 */
3756 static int install_routes_for_vrf(struct bgp *bgp_vrf)
3757 {
3758 install_uninstall_routes_for_vrf(bgp_vrf, 1);
3759 return 0;
3760 }
3761
3762 /*
3763 * Install any existing remote routes applicable for this VNI into its
3764 * routing table. This is invoked when a VNI becomes "live" or its Import
3765 * RT is changed.
3766 */
3767 static int install_routes_for_vni(struct bgp *bgp, struct bgpevpn *vpn)
3768 {
3769 int ret;
3770
3771 /* Install type-3 routes followed by type-2 routes - the ones applicable
3772 * for this VNI.
3773 */
3774 ret = install_uninstall_routes_for_vni(bgp, vpn, BGP_EVPN_IMET_ROUTE,
3775 1);
3776 if (ret)
3777 return ret;
3778
3779 ret = install_uninstall_routes_for_vni(bgp, vpn, BGP_EVPN_AD_ROUTE,
3780 1);
3781 if (ret)
3782 return ret;
3783
3784 return install_uninstall_routes_for_vni(bgp, vpn, BGP_EVPN_MAC_IP_ROUTE,
3785 1);
3786 }
3787
3788 /* uninstall routes from l3vni vrf. */
3789 static int uninstall_routes_for_vrf(struct bgp *bgp_vrf)
3790 {
3791 install_uninstall_routes_for_vrf(bgp_vrf, 0);
3792 return 0;
3793 }
3794
3795 /*
3796 * Uninstall any existing remote routes for this VNI. One scenario in which
3797 * this is invoked is upon an import RT change.
3798 */
3799 static int uninstall_routes_for_vni(struct bgp *bgp, struct bgpevpn *vpn)
3800 {
3801 int ret;
3802
3803 /* Uninstall type-2 routes followed by type-3 routes - the ones
3804 * applicable
3805 * for this VNI.
3806 */
3807 ret = install_uninstall_routes_for_vni(bgp, vpn, BGP_EVPN_MAC_IP_ROUTE,
3808 0);
3809 if (ret)
3810 return ret;
3811
3812 ret = install_uninstall_routes_for_vni(bgp, vpn, BGP_EVPN_AD_ROUTE,
3813 0);
3814 if (ret)
3815 return ret;
3816
3817
3818 return install_uninstall_routes_for_vni(bgp, vpn, BGP_EVPN_IMET_ROUTE,
3819 0);
3820 }
3821
3822 /*
3823 * Install or uninstall route in matching VRFs (list).
3824 */
3825 static int install_uninstall_route_in_vrfs(struct bgp *bgp_def, afi_t afi,
3826 safi_t safi, struct prefix_evpn *evp,
3827 struct bgp_path_info *pi,
3828 struct list *vrfs, int install)
3829 {
3830 struct bgp *bgp_vrf;
3831 struct listnode *node, *nnode;
3832
3833 /* Only type-2/type-5 routes go into a VRF */
3834 if (!(evp->prefix.route_type == BGP_EVPN_MAC_IP_ROUTE
3835 || evp->prefix.route_type == BGP_EVPN_IP_PREFIX_ROUTE))
3836 return 0;
3837
3838 /* if it is type-2 route and not a mac+ip route skip this route */
3839 if ((evp->prefix.route_type == BGP_EVPN_MAC_IP_ROUTE)
3840 && !(is_evpn_prefix_ipaddr_v4(evp)
3841 || is_evpn_prefix_ipaddr_v6(evp)))
3842 return 0;
3843
3844 for (ALL_LIST_ELEMENTS(vrfs, node, nnode, bgp_vrf)) {
3845 int ret;
3846
3847 /* don't import hosts that are locally attached */
3848 if (install && bgp_evpn_skip_vrf_import_of_local_es(
3849 bgp_vrf, evp, pi, install))
3850 return 0;
3851
3852 if (install)
3853 ret = install_evpn_route_entry_in_vrf(bgp_vrf, evp, pi);
3854 else
3855 ret = uninstall_evpn_route_entry_in_vrf(bgp_vrf, evp,
3856 pi);
3857
3858 if (ret) {
3859 flog_err(EC_BGP_EVPN_FAIL,
3860 "%u: Failed to %s prefix %pFX in VRF %s",
3861 bgp_def->vrf_id,
3862 install ? "install" : "uninstall", evp,
3863 vrf_id_to_name(bgp_vrf->vrf_id));
3864 return ret;
3865 }
3866 }
3867
3868 return 0;
3869 }
3870
3871 /*
3872 * Install or uninstall route in matching VNIs (list).
3873 */
3874 static int install_uninstall_route_in_vnis(struct bgp *bgp, afi_t afi,
3875 safi_t safi, struct prefix_evpn *evp,
3876 struct bgp_path_info *pi,
3877 struct list *vnis, int install)
3878 {
3879 struct bgpevpn *vpn;
3880 struct listnode *node, *nnode;
3881
3882 for (ALL_LIST_ELEMENTS(vnis, node, nnode, vpn)) {
3883 int ret;
3884
3885 if (!is_vni_live(vpn))
3886 continue;
3887
3888 if (install)
3889 ret = install_evpn_route_entry(bgp, vpn, evp, pi);
3890 else
3891 ret = uninstall_evpn_route_entry(bgp, vpn, evp, pi);
3892
3893 if (ret) {
3894 flog_err(EC_BGP_EVPN_FAIL,
3895 "%u: Failed to %s EVPN %s route in VNI %u",
3896 bgp->vrf_id, install ? "install" : "uninstall",
3897 evp->prefix.route_type == BGP_EVPN_MAC_IP_ROUTE
3898 ? "MACIP"
3899 : "IMET",
3900 vpn->vni);
3901 return ret;
3902 }
3903 }
3904
3905 return 0;
3906 }
3907
3908 /*
3909 * Install or uninstall route for appropriate VNIs/ESIs.
3910 */
3911 static int bgp_evpn_install_uninstall_table(struct bgp *bgp, afi_t afi,
3912 safi_t safi, const struct prefix *p,
3913 struct bgp_path_info *pi,
3914 int import, bool in_vni_rt,
3915 bool in_vrf_rt)
3916 {
3917 struct prefix_evpn *evp = (struct prefix_evpn *)p;
3918 struct attr *attr = pi->attr;
3919 struct ecommunity *ecom;
3920 uint32_t i;
3921 struct prefix_evpn ad_evp;
3922
3923 assert(attr);
3924
3925 /* Only type-1, type-2, type-3, type-4 and type-5
3926 * are supported currently
3927 */
3928 if (!(evp->prefix.route_type == BGP_EVPN_MAC_IP_ROUTE
3929 || evp->prefix.route_type == BGP_EVPN_IMET_ROUTE
3930 || evp->prefix.route_type == BGP_EVPN_ES_ROUTE
3931 || evp->prefix.route_type == BGP_EVPN_AD_ROUTE
3932 || evp->prefix.route_type == BGP_EVPN_IP_PREFIX_ROUTE))
3933 return 0;
3934
3935 /* If we don't have Route Target, nothing much to do. */
3936 if (!(attr->flag & ATTR_FLAG_BIT(BGP_ATTR_EXT_COMMUNITIES)))
3937 return 0;
3938
3939 /* EAD prefix in the global table doesn't include the VTEP-IP so
3940 * we need to create a different copy for the VNI
3941 */
3942 if (evp->prefix.route_type == BGP_EVPN_AD_ROUTE)
3943 evp = evpn_type1_prefix_vni_ip_copy(&ad_evp, evp,
3944 attr->nexthop);
3945
3946 ecom = bgp_attr_get_ecommunity(attr);
3947 if (!ecom || !ecom->size)
3948 return -1;
3949
3950 /* An EVPN route belongs to a VNI or a VRF or an ESI based on the RTs
3951 * attached to the route */
3952 for (i = 0; i < ecom->size; i++) {
3953 uint8_t *pnt;
3954 uint8_t type, sub_type;
3955 struct ecommunity_val *eval;
3956 struct ecommunity_val eval_tmp;
3957 struct irt_node *irt; /* import rt for l2vni */
3958 struct vrf_irt_node *vrf_irt; /* import rt for l3vni */
3959 struct bgp_evpn_es *es;
3960
3961 /* Only deal with RTs */
3962 pnt = (ecom->val + (i * ecom->unit_size));
3963 eval = (struct ecommunity_val *)(ecom->val
3964 + (i * ecom->unit_size));
3965 type = *pnt++;
3966 sub_type = *pnt++;
3967 if (sub_type != ECOMMUNITY_ROUTE_TARGET)
3968 continue;
3969
3970 /* non-local MAC-IP routes in the global route table are linked
3971 * to the destination ES
3972 */
3973 if (evp->prefix.route_type == BGP_EVPN_MAC_IP_ROUTE)
3974 bgp_evpn_path_es_link(pi, 0,
3975 bgp_evpn_attr_get_esi(pi->attr));
3976
3977 /*
3978 * macip routes (type-2) are imported into VNI and VRF tables.
3979 * IMET route is imported into VNI table.
3980 * prefix routes are imported into VRF table.
3981 */
3982 if (evp->prefix.route_type == BGP_EVPN_MAC_IP_ROUTE ||
3983 evp->prefix.route_type == BGP_EVPN_IMET_ROUTE ||
3984 evp->prefix.route_type == BGP_EVPN_AD_ROUTE ||
3985 evp->prefix.route_type == BGP_EVPN_IP_PREFIX_ROUTE) {
3986
3987 irt = in_vni_rt ? lookup_import_rt(bgp, eval) : NULL;
3988 if (irt)
3989 install_uninstall_route_in_vnis(
3990 bgp, afi, safi, evp, pi, irt->vnis,
3991 import);
3992
3993 vrf_irt = in_vrf_rt ? lookup_vrf_import_rt(eval) : NULL;
3994 if (vrf_irt)
3995 install_uninstall_route_in_vrfs(
3996 bgp, afi, safi, evp, pi, vrf_irt->vrfs,
3997 import);
3998
3999 /* Also check for non-exact match.
4000 * In this, we mask out the AS and
4001 * only check on the local-admin sub-field.
4002 * This is to facilitate using
4003 * VNI as the RT for EBGP peering too.
4004 */
4005 irt = NULL;
4006 vrf_irt = NULL;
4007 if (type == ECOMMUNITY_ENCODE_AS
4008 || type == ECOMMUNITY_ENCODE_AS4
4009 || type == ECOMMUNITY_ENCODE_IP) {
4010 memcpy(&eval_tmp, eval, ecom->unit_size);
4011 mask_ecom_global_admin(&eval_tmp, eval);
4012 if (in_vni_rt)
4013 irt = lookup_import_rt(bgp, &eval_tmp);
4014 if (in_vrf_rt)
4015 vrf_irt =
4016 lookup_vrf_import_rt(&eval_tmp);
4017 }
4018
4019 if (irt)
4020 install_uninstall_route_in_vnis(
4021 bgp, afi, safi, evp, pi, irt->vnis,
4022 import);
4023 if (vrf_irt)
4024 install_uninstall_route_in_vrfs(
4025 bgp, afi, safi, evp, pi, vrf_irt->vrfs,
4026 import);
4027 }
4028
4029 /* es route is imported into the es table */
4030 if (evp->prefix.route_type == BGP_EVPN_ES_ROUTE) {
4031
4032 /* we will match based on the entire esi to avoid
4033 * import of an es route for esi2 into esi1
4034 */
4035 es = bgp_evpn_es_find(&evp->prefix.es_addr.esi);
4036 if (es && bgp_evpn_is_es_local(es))
4037 bgp_evpn_es_route_install_uninstall(
4038 bgp, es, afi, safi, evp, pi, import);
4039 }
4040 }
4041
4042 return 0;
4043 }
4044
4045 /*
4046 * Install or uninstall route for appropriate VNIs/ESIs.
4047 */
4048 static int install_uninstall_evpn_route(struct bgp *bgp, afi_t afi, safi_t safi,
4049 const struct prefix *p,
4050 struct bgp_path_info *pi, int import)
4051 {
4052 return bgp_evpn_install_uninstall_table(bgp, afi, safi, p, pi, import,
4053 true, true);
4054 }
4055
4056 void bgp_evpn_import_type2_route(struct bgp_path_info *pi, int import)
4057 {
4058 struct bgp *bgp_evpn;
4059
4060 bgp_evpn = bgp_get_evpn();
4061 if (!bgp_evpn)
4062 return;
4063
4064 install_uninstall_evpn_route(bgp_evpn, AFI_L2VPN, SAFI_EVPN,
4065 &pi->net->p, pi, import);
4066 }
4067
4068 /*
4069 * delete and withdraw all ipv4 and ipv6 routes in the vrf table as type-5
4070 * routes
4071 */
4072 static void delete_withdraw_vrf_routes(struct bgp *bgp_vrf)
4073 {
4074 /* Delete ipv4 default route and withdraw from peers */
4075 if (evpn_default_originate_set(bgp_vrf, AFI_IP, SAFI_UNICAST))
4076 bgp_evpn_install_uninstall_default_route(bgp_vrf, AFI_IP,
4077 SAFI_UNICAST, false);
4078
4079 /* delete all ipv4 routes and withdraw from peers */
4080 if (advertise_type5_routes(bgp_vrf, AFI_IP))
4081 bgp_evpn_withdraw_type5_routes(bgp_vrf, AFI_IP, SAFI_UNICAST);
4082
4083 /* Delete ipv6 default route and withdraw from peers */
4084 if (evpn_default_originate_set(bgp_vrf, AFI_IP6, SAFI_UNICAST))
4085 bgp_evpn_install_uninstall_default_route(bgp_vrf, AFI_IP6,
4086 SAFI_UNICAST, false);
4087
4088 /* delete all ipv6 routes and withdraw from peers */
4089 if (advertise_type5_routes(bgp_vrf, AFI_IP6))
4090 bgp_evpn_withdraw_type5_routes(bgp_vrf, AFI_IP6, SAFI_UNICAST);
4091 }
4092
4093 /*
4094 * update and advertise all ipv4 and ipv6 routes in thr vrf table as type-5
4095 * routes
4096 */
4097 void update_advertise_vrf_routes(struct bgp *bgp_vrf)
4098 {
4099 struct bgp *bgp_evpn = NULL; /* EVPN bgp instance */
4100
4101 bgp_evpn = bgp_get_evpn();
4102 if (!bgp_evpn)
4103 return;
4104
4105 /* update all ipv4 routes */
4106 if (advertise_type5_routes(bgp_vrf, AFI_IP))
4107 bgp_evpn_advertise_type5_routes(bgp_vrf, AFI_IP, SAFI_UNICAST);
4108
4109 /* update ipv4 default route and withdraw from peers */
4110 if (evpn_default_originate_set(bgp_vrf, AFI_IP, SAFI_UNICAST))
4111 bgp_evpn_install_uninstall_default_route(bgp_vrf, AFI_IP,
4112 SAFI_UNICAST, true);
4113
4114 /* update all ipv6 routes */
4115 if (advertise_type5_routes(bgp_vrf, AFI_IP6))
4116 bgp_evpn_advertise_type5_routes(bgp_vrf, AFI_IP6, SAFI_UNICAST);
4117
4118 /* update ipv6 default route and withdraw from peers */
4119 if (evpn_default_originate_set(bgp_vrf, AFI_IP6, SAFI_UNICAST))
4120 bgp_evpn_install_uninstall_default_route(bgp_vrf, AFI_IP6,
4121 SAFI_UNICAST, true);
4122
4123 }
4124
4125 /*
4126 * update and advertise local routes for a VRF as type-5 routes.
4127 * This is invoked upon RD change for a VRF. Note taht the processing is only
4128 * done in the global route table using the routes which already exist in the
4129 * VRF routing table
4130 */
4131 static void update_router_id_vrf(struct bgp *bgp_vrf)
4132 {
4133 /* skip if the RD is configured */
4134 if (is_vrf_rd_configured(bgp_vrf))
4135 return;
4136
4137 /* derive the RD for the VRF based on new router-id */
4138 bgp_evpn_derive_auto_rd_for_vrf(bgp_vrf);
4139
4140 /* update advertise ipv4|ipv6 routes as type-5 routes */
4141 update_advertise_vrf_routes(bgp_vrf);
4142 }
4143
4144 /*
4145 * Delete and withdraw all type-5 routes for the RD corresponding to VRF.
4146 * This is invoked upon VRF RD change. The processing is done only from global
4147 * table.
4148 */
4149 static void withdraw_router_id_vrf(struct bgp *bgp_vrf)
4150 {
4151 /* skip if the RD is configured */
4152 if (is_vrf_rd_configured(bgp_vrf))
4153 return;
4154
4155 /* delete/withdraw ipv4|ipv6 routes as type-5 routes */
4156 delete_withdraw_vrf_routes(bgp_vrf);
4157 }
4158
4159 static void update_advertise_vni_route(struct bgp *bgp, struct bgpevpn *vpn,
4160 struct bgp_dest *dest)
4161 {
4162 struct bgp_dest *global_dest;
4163 struct bgp_path_info *pi, *global_pi;
4164 struct attr *attr;
4165 afi_t afi = AFI_L2VPN;
4166 safi_t safi = SAFI_EVPN;
4167
4168 struct prefix_evpn tmp_evp;
4169 const struct prefix_evpn *evp =
4170 (const struct prefix_evpn *)bgp_dest_get_prefix(dest);
4171
4172 /*
4173 * We have already processed type-3 routes.
4174 * Process only type-1 and type-2 routes here.
4175 */
4176 if (evp->prefix.route_type != BGP_EVPN_MAC_IP_ROUTE &&
4177 evp->prefix.route_type != BGP_EVPN_AD_ROUTE)
4178 return;
4179
4180 for (pi = bgp_dest_get_bgp_path_info(dest); pi; pi = pi->next)
4181 if (pi->peer == bgp->peer_self && pi->type == ZEBRA_ROUTE_BGP &&
4182 pi->sub_type == BGP_ROUTE_STATIC)
4183 break;
4184 if (!pi)
4185 return;
4186
4187 /*
4188 * VNI table MAC-IP prefixes don't have MAC so make sure it's
4189 * set from path info here.
4190 */
4191 if (evp->prefix.route_type == BGP_EVPN_MAC_IP_ROUTE) {
4192 if (is_evpn_prefix_ipaddr_none(evp)) {
4193 /* VNI MAC -> Global */
4194 evpn_type2_prefix_global_copy(
4195 &tmp_evp, evp, NULL /* mac */,
4196 evpn_type2_path_info_get_ip(pi));
4197 } else {
4198 /* VNI IP -> Global */
4199 evpn_type2_prefix_global_copy(
4200 &tmp_evp, evp, evpn_type2_path_info_get_mac(pi),
4201 NULL /* ip */);
4202 }
4203 } else {
4204 memcpy(&tmp_evp, evp, sizeof(tmp_evp));
4205 }
4206
4207 /* Create route in global routing table using this route entry's
4208 * attribute.
4209 */
4210 attr = pi->attr;
4211 global_dest = bgp_evpn_global_node_get(bgp->rib[afi][safi], afi, safi,
4212 &tmp_evp, &vpn->prd, NULL);
4213 assert(global_dest);
4214
4215 if (evp->prefix.route_type == BGP_EVPN_MAC_IP_ROUTE) {
4216 /* Type-2 route */
4217 update_evpn_route_entry(
4218 bgp, vpn, afi, safi, global_dest, attr, NULL /* mac */,
4219 NULL /* ip */, 1, &global_pi, 0,
4220 mac_mobility_seqnum(attr), false /* setup_sync */,
4221 NULL /* old_is_sync */);
4222 } else {
4223 /* Type-1 route */
4224 struct bgp_evpn_es *es;
4225 int route_changed = 0;
4226
4227 es = bgp_evpn_es_find(&evp->prefix.ead_addr.esi);
4228 bgp_evpn_mh_route_update(bgp, es, vpn, afi, safi, global_dest,
4229 attr, &global_pi, &route_changed);
4230 }
4231
4232 /* Schedule for processing and unlock node. */
4233 bgp_process(bgp, global_dest, afi, safi);
4234 bgp_dest_unlock_node(global_dest);
4235 }
4236
4237 /*
4238 * Update and advertise local routes for a VNI. Invoked upon router-id
4239 * change. Note that the processing is done only on the global route table
4240 * using routes that already exist in the per-VNI table.
4241 */
4242 static void update_advertise_vni_routes(struct bgp *bgp, struct bgpevpn *vpn)
4243 {
4244 struct prefix_evpn p;
4245 struct bgp_dest *dest, *global_dest;
4246 struct bgp_path_info *pi;
4247 struct attr *attr;
4248 afi_t afi = AFI_L2VPN;
4249 safi_t safi = SAFI_EVPN;
4250
4251 /* Locate type-3 route for VNI in the per-VNI table and use its
4252 * attributes to create and advertise the type-3 route for this VNI
4253 * in the global table.
4254 *
4255 * RT-3 only if doing head-end replication
4256 */
4257 if (bgp_evpn_vni_flood_mode_get(bgp, vpn)
4258 == VXLAN_FLOOD_HEAD_END_REPL) {
4259 build_evpn_type3_prefix(&p, vpn->originator_ip);
4260 dest = bgp_evpn_vni_node_lookup(vpn, &p, NULL);
4261 if (!dest) /* unexpected */
4262 return;
4263 for (pi = bgp_dest_get_bgp_path_info(dest); pi; pi = pi->next)
4264 if (pi->peer == bgp->peer_self &&
4265 pi->type == ZEBRA_ROUTE_BGP
4266 && pi->sub_type == BGP_ROUTE_STATIC)
4267 break;
4268 if (!pi) {
4269 bgp_dest_unlock_node(dest);
4270 return;
4271 }
4272
4273 attr = pi->attr;
4274
4275 global_dest = bgp_evpn_global_node_get(
4276 bgp->rib[afi][safi], afi, safi, &p, &vpn->prd, NULL);
4277 update_evpn_route_entry(
4278 bgp, vpn, afi, safi, global_dest, attr, NULL /* mac */,
4279 NULL /* ip */, 1, &pi, 0, mac_mobility_seqnum(attr),
4280 false /* setup_sync */, NULL /* old_is_sync */);
4281
4282 /* Schedule for processing and unlock node. */
4283 bgp_process(bgp, global_dest, afi, safi);
4284 bgp_dest_unlock_node(global_dest);
4285 }
4286
4287 /* Now, walk this VNI's MAC & IP route table and use the route and its
4288 * attribute to create and schedule route in global table.
4289 */
4290 for (dest = bgp_table_top(vpn->mac_table); dest;
4291 dest = bgp_route_next(dest))
4292 update_advertise_vni_route(bgp, vpn, dest);
4293
4294 for (dest = bgp_table_top(vpn->ip_table); dest;
4295 dest = bgp_route_next(dest))
4296 update_advertise_vni_route(bgp, vpn, dest);
4297 }
4298
4299 /*
4300 * Delete (and withdraw) local routes for a VNI - only from the global
4301 * table. Invoked upon router-id change.
4302 */
4303 static int delete_withdraw_vni_routes(struct bgp *bgp, struct bgpevpn *vpn)
4304 {
4305 struct prefix_evpn p;
4306 struct bgp_dest *global_dest;
4307 struct bgp_path_info *pi;
4308 afi_t afi = AFI_L2VPN;
4309 safi_t safi = SAFI_EVPN;
4310
4311 /* Delete and withdraw locally learnt type-2 routes (MACIP)
4312 * for this VNI - from the global table.
4313 */
4314 delete_global_type2_routes(bgp, vpn);
4315
4316 /* Remove type-3 route for this VNI from global table. */
4317 build_evpn_type3_prefix(&p, vpn->originator_ip);
4318 global_dest = bgp_evpn_global_node_lookup(bgp->rib[afi][safi], afi,
4319 safi, &p, &vpn->prd, NULL);
4320 if (global_dest) {
4321 /* Delete route entry in the global EVPN table. */
4322 delete_evpn_route_entry(bgp, afi, safi, global_dest, &pi);
4323
4324 /* Schedule for processing - withdraws to peers happen from
4325 * this table.
4326 */
4327 if (pi)
4328 bgp_process(bgp, global_dest, afi, safi);
4329 bgp_dest_unlock_node(global_dest);
4330 }
4331
4332
4333 delete_global_ead_evi_routes(bgp, vpn);
4334 return 0;
4335 }
4336
4337 /*
4338 * Handle router-id change. Update and advertise local routes corresponding
4339 * to this VNI from peers. Note that this is invoked after updating the
4340 * router-id. The routes in the per-VNI table are used to create routes in
4341 * the global table and schedule them.
4342 */
4343 static void update_router_id_vni(struct hash_bucket *bucket, struct bgp *bgp)
4344 {
4345 struct bgpevpn *vpn = (struct bgpevpn *)bucket->data;
4346
4347 /* Skip VNIs with configured RD. */
4348 if (is_rd_configured(vpn))
4349 return;
4350
4351 bgp_evpn_derive_auto_rd(bgp, vpn);
4352 update_advertise_vni_routes(bgp, vpn);
4353 }
4354
4355 /*
4356 * Handle router-id change. Delete and withdraw local routes corresponding
4357 * to this VNI from peers. Note that this is invoked prior to updating
4358 * the router-id and is done only on the global route table, the routes
4359 * are needed in the per-VNI table to re-advertise with new router id.
4360 */
4361 static void withdraw_router_id_vni(struct hash_bucket *bucket, struct bgp *bgp)
4362 {
4363 struct bgpevpn *vpn = (struct bgpevpn *)bucket->data;
4364
4365 /* Skip VNIs with configured RD. */
4366 if (is_rd_configured(vpn))
4367 return;
4368
4369 delete_withdraw_vni_routes(bgp, vpn);
4370 }
4371
4372 /*
4373 * Create RT-3 for a VNI and schedule for processing and advertisement.
4374 * This is invoked upon flooding mode changing to head-end replication.
4375 */
4376 static void create_advertise_type3(struct hash_bucket *bucket, void *data)
4377 {
4378 struct bgpevpn *vpn = bucket->data;
4379 struct bgp *bgp = data;
4380 struct prefix_evpn p;
4381
4382 if (!vpn || !is_vni_live(vpn) ||
4383 bgp_evpn_vni_flood_mode_get(bgp, vpn)
4384 != VXLAN_FLOOD_HEAD_END_REPL)
4385 return;
4386
4387 build_evpn_type3_prefix(&p, vpn->originator_ip);
4388 if (update_evpn_route(bgp, vpn, &p, 0, 0, NULL))
4389 flog_err(EC_BGP_EVPN_ROUTE_CREATE,
4390 "Type3 route creation failure for VNI %u", vpn->vni);
4391 }
4392
4393 /*
4394 * Delete RT-3 for a VNI and schedule for processing and withdrawal.
4395 * This is invoked upon flooding mode changing to drop BUM packets.
4396 */
4397 static void delete_withdraw_type3(struct hash_bucket *bucket, void *data)
4398 {
4399 struct bgpevpn *vpn = bucket->data;
4400 struct bgp *bgp = data;
4401 struct prefix_evpn p;
4402
4403 if (!vpn || !is_vni_live(vpn))
4404 return;
4405
4406 build_evpn_type3_prefix(&p, vpn->originator_ip);
4407 delete_evpn_route(bgp, vpn, &p);
4408 }
4409
4410 /*
4411 * Process received EVPN type-2 route (advertise or withdraw).
4412 */
4413 static int process_type2_route(struct peer *peer, afi_t afi, safi_t safi,
4414 struct attr *attr, uint8_t *pfx, int psize,
4415 uint32_t addpath_id)
4416 {
4417 struct prefix_rd prd;
4418 struct prefix_evpn p = {};
4419 struct bgp_route_evpn evpn = {};
4420 uint8_t ipaddr_len;
4421 uint8_t macaddr_len;
4422 /* holds the VNI(s) as in packet */
4423 mpls_label_t label[BGP_MAX_LABELS] = {};
4424 uint32_t num_labels = 0;
4425 uint32_t eth_tag;
4426 int ret;
4427
4428 /* Type-2 route should be either 33, 37 or 49 bytes or an
4429 * additional 3 bytes if there is a second label (VNI):
4430 * RD (8), ESI (10), Eth Tag (4), MAC Addr Len (1),
4431 * MAC Addr (6), IP len (1), IP (0, 4 or 16),
4432 * MPLS Lbl1 (3), MPLS Lbl2 (0 or 3)
4433 */
4434 if (psize != 33 && psize != 37 && psize != 49 && psize != 36
4435 && psize != 40 && psize != 52) {
4436 flog_err(EC_BGP_EVPN_ROUTE_INVALID,
4437 "%u:%s - Rx EVPN Type-2 NLRI with invalid length %d",
4438 peer->bgp->vrf_id, peer->host, psize);
4439 return -1;
4440 }
4441
4442 struct stream *pkt = stream_new(psize);
4443 stream_put(pkt, pfx, psize);
4444
4445 /* Make prefix_rd */
4446 prd.family = AF_UNSPEC;
4447 prd.prefixlen = 64;
4448
4449 STREAM_GET(&prd.val, pkt, 8);
4450
4451 /* Make EVPN prefix. */
4452 p.family = AF_EVPN;
4453 p.prefixlen = EVPN_ROUTE_PREFIXLEN;
4454 p.prefix.route_type = BGP_EVPN_MAC_IP_ROUTE;
4455
4456 /* Copy Ethernet Seg Identifier */
4457 if (attr) {
4458 STREAM_GET(&attr->esi, pkt, sizeof(esi_t));
4459
4460 if (bgp_evpn_is_esi_local_and_non_bypass(&attr->esi))
4461 attr->es_flags |= ATTR_ES_IS_LOCAL;
4462 else
4463 attr->es_flags &= ~ATTR_ES_IS_LOCAL;
4464 } else {
4465 STREAM_FORWARD_GETP(pkt, sizeof(esi_t));
4466 }
4467
4468 /* Copy Ethernet Tag */
4469 STREAM_GET(&eth_tag, pkt, 4);
4470 p.prefix.macip_addr.eth_tag = ntohl(eth_tag);
4471
4472 /* Get the MAC Addr len */
4473 STREAM_GETC(pkt, macaddr_len);
4474
4475 /* Get the MAC Addr */
4476 if (macaddr_len == (ETH_ALEN * 8)) {
4477 STREAM_GET(&p.prefix.macip_addr.mac.octet, pkt, ETH_ALEN);
4478 } else {
4479 flog_err(
4480 EC_BGP_EVPN_ROUTE_INVALID,
4481 "%u:%s - Rx EVPN Type-2 NLRI with unsupported MAC address length %d",
4482 peer->bgp->vrf_id, peer->host, macaddr_len);
4483 goto fail;
4484 }
4485
4486
4487 /* Get the IP. */
4488 STREAM_GETC(pkt, ipaddr_len);
4489
4490 if (ipaddr_len != 0 && ipaddr_len != IPV4_MAX_BITLEN
4491 && ipaddr_len != IPV6_MAX_BITLEN) {
4492 flog_err(
4493 EC_BGP_EVPN_ROUTE_INVALID,
4494 "%u:%s - Rx EVPN Type-2 NLRI with unsupported IP address length %d",
4495 peer->bgp->vrf_id, peer->host, ipaddr_len);
4496 goto fail;
4497 }
4498
4499 if (ipaddr_len) {
4500 ipaddr_len /= 8; /* Convert to bytes. */
4501 p.prefix.macip_addr.ip.ipa_type = (ipaddr_len == IPV4_MAX_BYTELEN)
4502 ? IPADDR_V4
4503 : IPADDR_V6;
4504 STREAM_GET(&p.prefix.macip_addr.ip.ip.addr, pkt, ipaddr_len);
4505 }
4506
4507 /* Get the VNI(s). Stored as bytes here. */
4508 STREAM_GET(&label[0], pkt, BGP_LABEL_BYTES);
4509 num_labels++;
4510
4511 /* Do we have a second VNI? */
4512 if (STREAM_READABLE(pkt)) {
4513 num_labels++;
4514 STREAM_GET(&label[1], pkt, BGP_LABEL_BYTES);
4515 }
4516
4517 /* Process the route. */
4518 if (attr)
4519 ret = bgp_update(peer, (struct prefix *)&p, addpath_id, attr,
4520 afi, safi, ZEBRA_ROUTE_BGP, BGP_ROUTE_NORMAL,
4521 &prd, &label[0], num_labels, 0, &evpn);
4522 else
4523 ret = bgp_withdraw(peer, (struct prefix *)&p, addpath_id, attr,
4524 afi, safi, ZEBRA_ROUTE_BGP, BGP_ROUTE_NORMAL,
4525 &prd, &label[0], num_labels, &evpn);
4526 goto done;
4527
4528 fail:
4529 stream_failure:
4530 flog_err(EC_BGP_EVPN_ROUTE_INVALID,
4531 "%u:%s - Rx EVPN Type-2 NLRI - corrupt, discarding",
4532 peer->bgp->vrf_id, peer->host);
4533 ret = -1;
4534 done:
4535 stream_free(pkt);
4536 return ret;
4537 }
4538
4539 /*
4540 * Process received EVPN type-3 route (advertise or withdraw).
4541 */
4542 static int process_type3_route(struct peer *peer, afi_t afi, safi_t safi,
4543 struct attr *attr, uint8_t *pfx, int psize,
4544 uint32_t addpath_id)
4545 {
4546 struct prefix_rd prd;
4547 struct prefix_evpn p;
4548 uint8_t ipaddr_len;
4549 uint32_t eth_tag;
4550 int ret;
4551
4552 /* Type-3 route should be either 17 or 29 bytes: RD (8), Eth Tag (4),
4553 * IP len (1) and IP (4 or 16).
4554 */
4555 if (psize != 17 && psize != 29) {
4556 flog_err(EC_BGP_EVPN_ROUTE_INVALID,
4557 "%u:%s - Rx EVPN Type-3 NLRI with invalid length %d",
4558 peer->bgp->vrf_id, peer->host, psize);
4559 return -1;
4560 }
4561
4562 /* If PMSI is present, log if it is anything other than IR.
4563 * Note: We just simply ignore the values as it is not clear if
4564 * doing anything else is better.
4565 */
4566 if (attr &&
4567 (attr->flag & ATTR_FLAG_BIT(BGP_ATTR_PMSI_TUNNEL))) {
4568 enum pta_type pmsi_tnl_type = bgp_attr_get_pmsi_tnl_type(attr);
4569
4570 if (pmsi_tnl_type != PMSI_TNLTYPE_INGR_REPL
4571 && pmsi_tnl_type != PMSI_TNLTYPE_PIM_SM) {
4572 flog_warn(
4573 EC_BGP_EVPN_PMSI_PRESENT,
4574 "%u:%s - Rx EVPN Type-3 NLRI with unsupported PTA %d",
4575 peer->bgp->vrf_id, peer->host, pmsi_tnl_type);
4576 }
4577 }
4578
4579 /* Make prefix_rd */
4580 prd.family = AF_UNSPEC;
4581 prd.prefixlen = 64;
4582 memcpy(&prd.val, pfx, 8);
4583 pfx += 8;
4584
4585 /* Make EVPN prefix. */
4586 memset(&p, 0, sizeof(p));
4587 p.family = AF_EVPN;
4588 p.prefixlen = EVPN_ROUTE_PREFIXLEN;
4589 p.prefix.route_type = BGP_EVPN_IMET_ROUTE;
4590
4591 /* Copy Ethernet Tag */
4592 memcpy(&eth_tag, pfx, 4);
4593 p.prefix.imet_addr.eth_tag = ntohl(eth_tag);
4594 pfx += 4;
4595
4596 /* Get the IP. */
4597 ipaddr_len = *pfx++;
4598 if (ipaddr_len == IPV4_MAX_BITLEN) {
4599 p.prefix.imet_addr.ip.ipa_type = IPADDR_V4;
4600 memcpy(&p.prefix.imet_addr.ip.ip.addr, pfx, IPV4_MAX_BYTELEN);
4601 } else {
4602 flog_err(
4603 EC_BGP_EVPN_ROUTE_INVALID,
4604 "%u:%s - Rx EVPN Type-3 NLRI with unsupported IP address length %d",
4605 peer->bgp->vrf_id, peer->host, ipaddr_len);
4606 return -1;
4607 }
4608
4609 /* Process the route. */
4610 if (attr)
4611 ret = bgp_update(peer, (struct prefix *)&p, addpath_id, attr,
4612 afi, safi, ZEBRA_ROUTE_BGP, BGP_ROUTE_NORMAL,
4613 &prd, NULL, 0, 0, NULL);
4614 else
4615 ret = bgp_withdraw(peer, (struct prefix *)&p, addpath_id, attr,
4616 afi, safi, ZEBRA_ROUTE_BGP, BGP_ROUTE_NORMAL,
4617 &prd, NULL, 0, NULL);
4618 return ret;
4619 }
4620
4621 /*
4622 * Process received EVPN type-5 route (advertise or withdraw).
4623 */
4624 static int process_type5_route(struct peer *peer, afi_t afi, safi_t safi,
4625 struct attr *attr, uint8_t *pfx, int psize,
4626 uint32_t addpath_id)
4627 {
4628 struct prefix_rd prd;
4629 struct prefix_evpn p;
4630 struct bgp_route_evpn evpn;
4631 uint8_t ippfx_len;
4632 uint32_t eth_tag;
4633 mpls_label_t label; /* holds the VNI as in the packet */
4634 int ret;
4635 bool is_valid_update = true;
4636
4637 /* Type-5 route should be 34 or 58 bytes:
4638 * RD (8), ESI (10), Eth Tag (4), IP len (1), IP (4 or 16),
4639 * GW (4 or 16) and VNI (3).
4640 * Note that the IP and GW should both be IPv4 or both IPv6.
4641 */
4642 if (psize != 34 && psize != 58) {
4643 flog_err(EC_BGP_EVPN_ROUTE_INVALID,
4644 "%u:%s - Rx EVPN Type-5 NLRI with invalid length %d",
4645 peer->bgp->vrf_id, peer->host, psize);
4646 return -1;
4647 }
4648
4649 /* Make prefix_rd */
4650 prd.family = AF_UNSPEC;
4651 prd.prefixlen = 64;
4652 memcpy(&prd.val, pfx, 8);
4653 pfx += 8;
4654
4655 /* Make EVPN prefix. */
4656 memset(&p, 0, sizeof(p));
4657 p.family = AF_EVPN;
4658 p.prefixlen = EVPN_ROUTE_PREFIXLEN;
4659 p.prefix.route_type = BGP_EVPN_IP_PREFIX_ROUTE;
4660
4661 /* Additional information outside of prefix - ESI and GW IP */
4662 memset(&evpn, 0, sizeof(evpn));
4663
4664 /* Fetch ESI overlay index */
4665 if (attr)
4666 memcpy(&evpn.eth_s_id, pfx, sizeof(esi_t));
4667 pfx += ESI_BYTES;
4668
4669 /* Fetch Ethernet Tag. */
4670 memcpy(&eth_tag, pfx, 4);
4671 p.prefix.prefix_addr.eth_tag = ntohl(eth_tag);
4672 pfx += 4;
4673
4674 /* Fetch IP prefix length. */
4675 ippfx_len = *pfx++;
4676 if (ippfx_len > IPV6_MAX_BITLEN) {
4677 flog_err(
4678 EC_BGP_EVPN_ROUTE_INVALID,
4679 "%u:%s - Rx EVPN Type-5 NLRI with invalid IP Prefix length %d",
4680 peer->bgp->vrf_id, peer->host, ippfx_len);
4681 return -1;
4682 }
4683 p.prefix.prefix_addr.ip_prefix_length = ippfx_len;
4684
4685 /* Determine IPv4 or IPv6 prefix */
4686 /* Since the address and GW are from the same family, this just becomes
4687 * a simple check on the total size.
4688 */
4689 if (psize == 34) {
4690 SET_IPADDR_V4(&p.prefix.prefix_addr.ip);
4691 memcpy(&p.prefix.prefix_addr.ip.ipaddr_v4, pfx, 4);
4692 pfx += 4;
4693 SET_IPADDR_V4(&evpn.gw_ip);
4694 memcpy(&evpn.gw_ip.ipaddr_v4, pfx, 4);
4695 pfx += 4;
4696 } else {
4697 SET_IPADDR_V6(&p.prefix.prefix_addr.ip);
4698 memcpy(&p.prefix.prefix_addr.ip.ipaddr_v6, pfx,
4699 IPV6_MAX_BYTELEN);
4700 pfx += IPV6_MAX_BYTELEN;
4701 SET_IPADDR_V6(&evpn.gw_ip);
4702 memcpy(&evpn.gw_ip.ipaddr_v6, pfx, IPV6_MAX_BYTELEN);
4703 pfx += IPV6_MAX_BYTELEN;
4704 }
4705
4706 /* Get the VNI (in MPLS label field). Stored as bytes here. */
4707 memset(&label, 0, sizeof(label));
4708 memcpy(&label, pfx, BGP_LABEL_BYTES);
4709
4710 /*
4711 * If in future, we are required to access additional fields,
4712 * we MUST increment pfx by BGP_LABEL_BYTES in before reading the next
4713 * field
4714 */
4715
4716 /*
4717 * An update containing a non-zero gateway IP and a non-zero ESI
4718 * at the same time is should be treated as withdraw
4719 */
4720 if (bgp_evpn_is_esi_valid(&evpn.eth_s_id) &&
4721 !ipaddr_is_zero(&evpn.gw_ip)) {
4722 flog_err(EC_BGP_EVPN_ROUTE_INVALID,
4723 "%s - Rx EVPN Type-5 ESI and gateway-IP both non-zero.",
4724 peer->host);
4725 is_valid_update = false;
4726 } else if (bgp_evpn_is_esi_valid(&evpn.eth_s_id))
4727 evpn.type = OVERLAY_INDEX_ESI;
4728 else if (!ipaddr_is_zero(&evpn.gw_ip))
4729 evpn.type = OVERLAY_INDEX_GATEWAY_IP;
4730 if (attr) {
4731 if (is_zero_mac(&attr->rmac) &&
4732 !bgp_evpn_is_esi_valid(&evpn.eth_s_id) &&
4733 ipaddr_is_zero(&evpn.gw_ip) && label == 0) {
4734 flog_err(EC_BGP_EVPN_ROUTE_INVALID,
4735 "%s - Rx EVPN Type-5 ESI, gateway-IP, RMAC and label all zero",
4736 peer->host);
4737 is_valid_update = false;
4738 }
4739
4740 if (is_mcast_mac(&attr->rmac) || is_bcast_mac(&attr->rmac))
4741 is_valid_update = false;
4742 }
4743
4744 /* Process the route. */
4745 if (attr && is_valid_update)
4746 ret = bgp_update(peer, (struct prefix *)&p, addpath_id, attr,
4747 afi, safi, ZEBRA_ROUTE_BGP, BGP_ROUTE_NORMAL,
4748 &prd, &label, 1, 0, &evpn);
4749 else {
4750 if (!is_valid_update) {
4751 char attr_str[BUFSIZ] = {0};
4752
4753 bgp_dump_attr(attr, attr_str, BUFSIZ);
4754 zlog_warn(
4755 "Invalid update from peer %s vrf %u prefix %pFX attr %s - treat as withdraw",
4756 peer->hostname, peer->bgp->vrf_id, &p,
4757 attr_str);
4758 }
4759 ret = bgp_withdraw(peer, (struct prefix *)&p, addpath_id, attr,
4760 afi, safi, ZEBRA_ROUTE_BGP, BGP_ROUTE_NORMAL,
4761 &prd, &label, 1, &evpn);
4762 }
4763
4764 return ret;
4765 }
4766
4767 static void evpn_mpattr_encode_type5(struct stream *s, const struct prefix *p,
4768 const struct prefix_rd *prd,
4769 mpls_label_t *label, uint32_t num_labels,
4770 struct attr *attr)
4771 {
4772 int len;
4773 char temp[16];
4774 const struct evpn_addr *p_evpn_p;
4775
4776 memset(&temp, 0, sizeof(temp));
4777 if (p->family != AF_EVPN)
4778 return;
4779 p_evpn_p = &(p->u.prefix_evpn);
4780
4781 /* len denites the total len of IP and GW-IP in the route
4782 IP and GW-IP have to be both ipv4 or ipv6
4783 */
4784 if (IS_IPADDR_V4(&p_evpn_p->prefix_addr.ip))
4785 len = 8; /* IP and GWIP are both ipv4 */
4786 else
4787 len = 32; /* IP and GWIP are both ipv6 */
4788 /* Prefix contains RD, ESI, EthTag, IP length, IP, GWIP and VNI */
4789 stream_putc(s, 8 + 10 + 4 + 1 + len + 3);
4790 stream_put(s, prd->val, 8);
4791 if (attr && attr->evpn_overlay.type == OVERLAY_INDEX_ESI)
4792 stream_put(s, &attr->esi, sizeof(esi_t));
4793 else
4794 stream_put(s, 0, sizeof(esi_t));
4795 stream_putl(s, p_evpn_p->prefix_addr.eth_tag);
4796 stream_putc(s, p_evpn_p->prefix_addr.ip_prefix_length);
4797 if (IS_IPADDR_V4(&p_evpn_p->prefix_addr.ip))
4798 stream_put_ipv4(s, p_evpn_p->prefix_addr.ip.ipaddr_v4.s_addr);
4799 else
4800 stream_put(s, &p_evpn_p->prefix_addr.ip.ipaddr_v6, 16);
4801 if (attr && attr->evpn_overlay.type == OVERLAY_INDEX_GATEWAY_IP) {
4802 const struct bgp_route_evpn *evpn_overlay =
4803 bgp_attr_get_evpn_overlay(attr);
4804
4805 if (IS_IPADDR_V4(&p_evpn_p->prefix_addr.ip))
4806 stream_put_ipv4(s,
4807 evpn_overlay->gw_ip.ipaddr_v4.s_addr);
4808 else
4809 stream_put(s, &(evpn_overlay->gw_ip.ipaddr_v6), 16);
4810 } else {
4811 if (IS_IPADDR_V4(&p_evpn_p->prefix_addr.ip))
4812 stream_put_ipv4(s, 0);
4813 else
4814 stream_put(s, &temp, 16);
4815 }
4816
4817 if (num_labels)
4818 stream_put(s, label, 3);
4819 else
4820 stream_put3(s, 0);
4821 }
4822
4823 /*
4824 * Cleanup specific VNI upon EVPN (advertise-all-vni) being disabled.
4825 */
4826 static void cleanup_vni_on_disable(struct hash_bucket *bucket, struct bgp *bgp)
4827 {
4828 struct bgpevpn *vpn = (struct bgpevpn *)bucket->data;
4829
4830 /* Remove EVPN routes and schedule for processing. */
4831 delete_routes_for_vni(bgp, vpn);
4832
4833 /* Clear "live" flag and see if hash needs to be freed. */
4834 UNSET_FLAG(vpn->flags, VNI_FLAG_LIVE);
4835 if (!is_vni_configured(vpn))
4836 bgp_evpn_free(bgp, vpn);
4837 }
4838
4839 /*
4840 * Free a VNI entry; iterator function called during cleanup.
4841 */
4842 static void free_vni_entry(struct hash_bucket *bucket, struct bgp *bgp)
4843 {
4844 struct bgpevpn *vpn = (struct bgpevpn *)bucket->data;
4845
4846 delete_all_vni_routes(bgp, vpn);
4847 bgp_evpn_free(bgp, vpn);
4848 }
4849
4850 /*
4851 * Derive AUTO import RT for BGP VRF - L3VNI
4852 */
4853 static void evpn_auto_rt_import_add_for_vrf(struct bgp *bgp_vrf)
4854 {
4855 struct bgp *bgp_evpn = NULL;
4856
4857 form_auto_rt(bgp_vrf, bgp_vrf->l3vni, bgp_vrf->vrf_import_rtl, true);
4858
4859 /* Map RT to VRF */
4860 bgp_evpn = bgp_get_evpn();
4861
4862 if (!bgp_evpn)
4863 return;
4864
4865 bgp_evpn_map_vrf_to_its_rts(bgp_vrf);
4866 }
4867
4868 /*
4869 * Delete AUTO import RT from BGP VRF - L3VNI
4870 */
4871 static void evpn_auto_rt_import_delete_for_vrf(struct bgp *bgp_vrf)
4872 {
4873 evpn_rt_delete_auto(bgp_vrf, bgp_vrf->l3vni, bgp_vrf->vrf_import_rtl,
4874 true);
4875 }
4876
4877 /*
4878 * Derive AUTO export RT for BGP VRF - L3VNI
4879 */
4880 static void evpn_auto_rt_export_add_for_vrf(struct bgp *bgp_vrf)
4881 {
4882 form_auto_rt(bgp_vrf, bgp_vrf->l3vni, bgp_vrf->vrf_export_rtl, true);
4883 }
4884
4885 /*
4886 * Delete AUTO export RT from BGP VRF - L3VNI
4887 */
4888 static void evpn_auto_rt_export_delete_for_vrf(struct bgp *bgp_vrf)
4889 {
4890 evpn_rt_delete_auto(bgp_vrf, bgp_vrf->l3vni, bgp_vrf->vrf_export_rtl,
4891 true);
4892 }
4893
4894 static void bgp_evpn_handle_export_rt_change_for_vrf(struct bgp *bgp_vrf)
4895 {
4896 struct bgp *bgp_evpn = NULL;
4897 struct listnode *node = NULL;
4898 struct bgpevpn *vpn = NULL;
4899
4900 bgp_evpn = bgp_get_evpn();
4901 if (!bgp_evpn)
4902 return;
4903
4904 /* update all type-5 routes */
4905 update_advertise_vrf_routes(bgp_vrf);
4906
4907 /* update all type-2 routes */
4908 for (ALL_LIST_ELEMENTS_RO(bgp_vrf->l2vnis, node, vpn))
4909 update_routes_for_vni(bgp_evpn, vpn);
4910 }
4911
4912 /*
4913 * Handle autort change for a given VNI.
4914 */
4915 static void update_autort_vni(struct hash_bucket *bucket, struct bgp *bgp)
4916 {
4917 struct bgpevpn *vpn = bucket->data;
4918
4919 if (!is_import_rt_configured(vpn)) {
4920 if (is_vni_live(vpn))
4921 bgp_evpn_uninstall_routes(bgp, vpn);
4922 bgp_evpn_unmap_vni_from_its_rts(bgp, vpn);
4923 list_delete_all_node(vpn->import_rtl);
4924 bgp_evpn_derive_auto_rt_import(bgp, vpn);
4925 if (is_vni_live(vpn))
4926 bgp_evpn_install_routes(bgp, vpn);
4927 }
4928 if (!is_export_rt_configured(vpn)) {
4929 list_delete_all_node(vpn->export_rtl);
4930 bgp_evpn_derive_auto_rt_export(bgp, vpn);
4931 if (is_vni_live(vpn))
4932 bgp_evpn_handle_export_rt_change(bgp, vpn);
4933 }
4934 }
4935
4936 /*
4937 * Handle autort change for L3VNI.
4938 */
4939 static void update_autort_l3vni(struct bgp *bgp)
4940 {
4941 if ((CHECK_FLAG(bgp->vrf_flags, BGP_VRF_IMPORT_RT_CFGD))
4942 && (CHECK_FLAG(bgp->vrf_flags, BGP_VRF_EXPORT_RT_CFGD)))
4943 return;
4944
4945 if (!CHECK_FLAG(bgp->vrf_flags, BGP_VRF_IMPORT_RT_CFGD)) {
4946 if (is_l3vni_live(bgp))
4947 uninstall_routes_for_vrf(bgp);
4948
4949 /* Cleanup the RT to VRF mapping */
4950 bgp_evpn_unmap_vrf_from_its_rts(bgp);
4951
4952 /* Remove auto generated RT */
4953 evpn_auto_rt_import_delete_for_vrf(bgp);
4954
4955 list_delete_all_node(bgp->vrf_import_rtl);
4956
4957 /* Map auto derive or configured RTs */
4958 evpn_auto_rt_import_add_for_vrf(bgp);
4959 }
4960
4961 if (!CHECK_FLAG(bgp->vrf_flags, BGP_VRF_EXPORT_RT_CFGD)) {
4962 list_delete_all_node(bgp->vrf_export_rtl);
4963
4964 evpn_auto_rt_export_delete_for_vrf(bgp);
4965
4966 evpn_auto_rt_export_add_for_vrf(bgp);
4967
4968 if (is_l3vni_live(bgp))
4969 bgp_evpn_map_vrf_to_its_rts(bgp);
4970 }
4971
4972 if (!is_l3vni_live(bgp))
4973 return;
4974
4975 /* advertise type-5 routes if needed */
4976 update_advertise_vrf_routes(bgp);
4977
4978 /* install all remote routes belonging to this l3vni
4979 * into corresponding vrf
4980 */
4981 install_routes_for_vrf(bgp);
4982 }
4983
4984 /*
4985 * Public functions.
4986 */
4987
4988 /* withdraw type-5 route corresponding to ip prefix */
4989 void bgp_evpn_withdraw_type5_route(struct bgp *bgp_vrf, const struct prefix *p,
4990 afi_t afi, safi_t safi)
4991 {
4992 int ret = 0;
4993 struct prefix_evpn evp;
4994
4995 build_type5_prefix_from_ip_prefix(&evp, p);
4996 ret = delete_evpn_type5_route(bgp_vrf, &evp);
4997 if (ret)
4998 flog_err(
4999 EC_BGP_EVPN_ROUTE_DELETE,
5000 "%u failed to delete type-5 route for prefix %pFX in vrf %s",
5001 bgp_vrf->vrf_id, p, vrf_id_to_name(bgp_vrf->vrf_id));
5002 }
5003
5004 /* withdraw all type-5 routes for an address family */
5005 void bgp_evpn_withdraw_type5_routes(struct bgp *bgp_vrf, afi_t afi, safi_t safi)
5006 {
5007 struct bgp_table *table = NULL;
5008 struct bgp_dest *dest = NULL;
5009 struct bgp_path_info *pi;
5010
5011 table = bgp_vrf->rib[afi][safi];
5012 for (dest = bgp_table_top(table); dest; dest = bgp_route_next(dest)) {
5013 /* Only care about "selected" routes. Also ensure that
5014 * these are routes that are injectable into EVPN.
5015 */
5016 /* TODO: Support for AddPath for EVPN. */
5017 for (pi = bgp_dest_get_bgp_path_info(dest); pi; pi = pi->next) {
5018 if (CHECK_FLAG(pi->flags, BGP_PATH_SELECTED)
5019 && is_route_injectable_into_evpn(pi)) {
5020 bgp_evpn_withdraw_type5_route(
5021 bgp_vrf, bgp_dest_get_prefix(dest), afi,
5022 safi);
5023 break;
5024 }
5025 }
5026 }
5027 }
5028
5029 /*
5030 * evpn - enable advertisement of default g/w
5031 */
5032 void bgp_evpn_install_uninstall_default_route(struct bgp *bgp_vrf, afi_t afi,
5033 safi_t safi, bool add)
5034 {
5035 struct prefix ip_prefix;
5036
5037 /* form the default prefix 0.0.0.0/0 */
5038 memset(&ip_prefix, 0, sizeof(ip_prefix));
5039 ip_prefix.family = afi2family(afi);
5040
5041 if (add) {
5042 bgp_evpn_advertise_type5_route(bgp_vrf, &ip_prefix,
5043 NULL, afi, safi);
5044 } else {
5045 bgp_evpn_withdraw_type5_route(bgp_vrf, &ip_prefix,
5046 afi, safi);
5047 }
5048 }
5049
5050
5051 /*
5052 * Advertise IP prefix as type-5 route. The afi/safi and src_attr passed
5053 * to this function correspond to those of the source IP prefix (best
5054 * path in the case of the attr. In the case of a local prefix (when we
5055 * are advertising local subnets), the src_attr will be NULL.
5056 */
5057 void bgp_evpn_advertise_type5_route(struct bgp *bgp_vrf, const struct prefix *p,
5058 struct attr *src_attr, afi_t afi,
5059 safi_t safi)
5060 {
5061 int ret = 0;
5062 struct prefix_evpn evp;
5063
5064 build_type5_prefix_from_ip_prefix(&evp, p);
5065 ret = update_evpn_type5_route(bgp_vrf, &evp, src_attr, afi, safi);
5066 if (ret)
5067 flog_err(EC_BGP_EVPN_ROUTE_CREATE,
5068 "%u: Failed to create type-5 route for prefix %pFX",
5069 bgp_vrf->vrf_id, p);
5070 }
5071
5072 /* Inject all prefixes of a particular address-family (currently, IPv4 or
5073 * IPv6 unicast) into EVPN as type-5 routes. This is invoked when the
5074 * advertisement is enabled.
5075 */
5076 void bgp_evpn_advertise_type5_routes(struct bgp *bgp_vrf, afi_t afi,
5077 safi_t safi)
5078 {
5079 struct bgp_table *table = NULL;
5080 struct bgp_dest *dest = NULL;
5081 struct bgp_path_info *pi;
5082
5083 table = bgp_vrf->rib[afi][safi];
5084 for (dest = bgp_table_top(table); dest; dest = bgp_route_next(dest)) {
5085 /* Need to identify the "selected" route entry to use its
5086 * attribute. Also, ensure that the route is injectable
5087 * into EVPN.
5088 * TODO: Support for AddPath for EVPN.
5089 */
5090 for (pi = bgp_dest_get_bgp_path_info(dest); pi; pi = pi->next) {
5091 if (CHECK_FLAG(pi->flags, BGP_PATH_SELECTED)
5092 && is_route_injectable_into_evpn(pi)) {
5093
5094 /* apply the route-map */
5095 if (bgp_vrf->adv_cmd_rmap[afi][safi].map) {
5096 route_map_result_t ret;
5097 struct bgp_path_info tmp_pi;
5098 struct bgp_path_info_extra tmp_pie;
5099 struct attr tmp_attr;
5100
5101 tmp_attr = *pi->attr;
5102
5103 /* Fill temp path_info */
5104 prep_for_rmap_apply(&tmp_pi, &tmp_pie,
5105 dest, pi, pi->peer,
5106 &tmp_attr);
5107
5108 RESET_FLAG(tmp_attr.rmap_change_flags);
5109
5110 ret = route_map_apply(
5111 bgp_vrf->adv_cmd_rmap[afi][safi]
5112 .map,
5113 bgp_dest_get_prefix(dest),
5114 &tmp_pi);
5115 if (ret == RMAP_DENYMATCH) {
5116 bgp_attr_flush(&tmp_attr);
5117 continue;
5118 }
5119 bgp_evpn_advertise_type5_route(
5120 bgp_vrf,
5121 bgp_dest_get_prefix(dest),
5122 &tmp_attr, afi, safi);
5123 } else
5124 bgp_evpn_advertise_type5_route(
5125 bgp_vrf,
5126 bgp_dest_get_prefix(dest),
5127 pi->attr, afi, safi);
5128 break;
5129 }
5130 }
5131 }
5132 }
5133
5134 static void rt_list_remove_node(struct list *rt_list,
5135 struct ecommunity *ecomdel, bool is_l3)
5136 {
5137 struct listnode *node = NULL, *nnode = NULL, *node_to_del = NULL;
5138 struct vrf_route_target *l3rt = NULL;
5139 struct ecommunity *ecom = NULL;
5140
5141 if (is_l3) {
5142 for (ALL_LIST_ELEMENTS(rt_list, node, nnode, l3rt)) {
5143 if (ecommunity_match(l3rt->ecom, ecomdel)) {
5144 evpn_vrf_rt_del(l3rt);
5145 node_to_del = node;
5146 break;
5147 }
5148 }
5149 } else {
5150 for (ALL_LIST_ELEMENTS(rt_list, node, nnode, ecom)) {
5151 if (ecommunity_match(ecom, ecomdel)) {
5152 ecommunity_free(&ecom);
5153 node_to_del = node;
5154 break;
5155 }
5156 }
5157 }
5158
5159
5160 if (node_to_del)
5161 list_delete_node(rt_list, node_to_del);
5162 }
5163
5164 void evpn_rt_delete_auto(struct bgp *bgp, vni_t vni, struct list *rtl,
5165 bool is_l3)
5166 {
5167 struct ecommunity *ecom_auto;
5168 struct ecommunity_val eval;
5169
5170 if (bgp->advertise_autort_rfc8365)
5171 vni |= EVPN_AUTORT_VXLAN;
5172
5173 encode_route_target_as((bgp->as & 0xFFFF), vni, &eval);
5174
5175 ecom_auto = ecommunity_new();
5176 ecommunity_add_val(ecom_auto, &eval, false, false);
5177
5178 rt_list_remove_node(rtl, ecom_auto, is_l3);
5179
5180 ecommunity_free(&ecom_auto);
5181 }
5182
5183 static void evpn_vrf_rt_routes_map(struct bgp *bgp_vrf)
5184 {
5185 /* map VRFs to its RTs and install routes matching this new RT */
5186 if (is_l3vni_live(bgp_vrf)) {
5187 bgp_evpn_map_vrf_to_its_rts(bgp_vrf);
5188 install_routes_for_vrf(bgp_vrf);
5189 }
5190 }
5191
5192 static void evpn_vrf_rt_routes_unmap(struct bgp *bgp_vrf)
5193 {
5194 /* uninstall routes from vrf */
5195 if (is_l3vni_live(bgp_vrf))
5196 uninstall_routes_for_vrf(bgp_vrf);
5197
5198 /* Cleanup the RT to VRF mapping */
5199 bgp_evpn_unmap_vrf_from_its_rts(bgp_vrf);
5200 }
5201
5202 static bool rt_list_has_cfgd_rt(struct list *rt_list)
5203 {
5204 struct listnode *node = NULL, *nnode = NULL;
5205 struct vrf_route_target *l3rt = NULL;
5206
5207 for (ALL_LIST_ELEMENTS(rt_list, node, nnode, l3rt)) {
5208 if (!CHECK_FLAG(l3rt->flags, BGP_VRF_RT_AUTO))
5209 return true;
5210 }
5211
5212 return false;
5213 }
5214
5215 static void unconfigure_import_rt_for_vrf_fini(struct bgp *bgp_vrf)
5216 {
5217 if (!bgp_vrf->vrf_import_rtl)
5218 return; /* this should never fail */
5219
5220 if (!is_l3vni_live(bgp_vrf))
5221 return; /* Nothing to do if no vni */
5222
5223 /* fall back to auto-generated RT if this was the last RT */
5224 if (list_isempty(bgp_vrf->vrf_import_rtl))
5225 evpn_auto_rt_import_add_for_vrf(bgp_vrf);
5226 }
5227
5228 static void unconfigure_export_rt_for_vrf_fini(struct bgp *bgp_vrf)
5229 {
5230
5231 if (!bgp_vrf->vrf_export_rtl)
5232 return; /* this should never fail */
5233
5234 if (!is_l3vni_live(bgp_vrf))
5235 return; /* Nothing to do if no vni */
5236
5237 /* fall back to auto-generated RT if this was the last RT */
5238 if (list_isempty(bgp_vrf->vrf_export_rtl))
5239 evpn_auto_rt_export_add_for_vrf(bgp_vrf);
5240
5241 bgp_evpn_handle_export_rt_change_for_vrf(bgp_vrf);
5242 }
5243
5244 void bgp_evpn_configure_import_rt_for_vrf(struct bgp *bgp_vrf,
5245 struct ecommunity *ecomadd,
5246 bool is_wildcard)
5247 {
5248 struct vrf_route_target *newrt;
5249
5250 newrt = evpn_vrf_rt_new(ecomadd);
5251
5252 if (is_wildcard)
5253 SET_FLAG(newrt->flags, BGP_VRF_RT_WILD);
5254
5255 evpn_vrf_rt_routes_unmap(bgp_vrf);
5256
5257 /* Remove auto generated RT if not configured */
5258 if (!CHECK_FLAG(bgp_vrf->vrf_flags, BGP_VRF_IMPORT_AUTO_RT_CFGD))
5259 evpn_auto_rt_import_delete_for_vrf(bgp_vrf);
5260
5261 /* Add the newly configured RT to RT list */
5262 listnode_add_sort(bgp_vrf->vrf_import_rtl, newrt);
5263
5264 SET_FLAG(bgp_vrf->vrf_flags, BGP_VRF_IMPORT_RT_CFGD);
5265
5266 evpn_vrf_rt_routes_map(bgp_vrf);
5267 }
5268
5269 void bgp_evpn_configure_import_auto_rt_for_vrf(struct bgp *bgp_vrf)
5270 {
5271 if (CHECK_FLAG(bgp_vrf->vrf_flags, BGP_VRF_IMPORT_AUTO_RT_CFGD))
5272 return; /* Already configured */
5273
5274 SET_FLAG(bgp_vrf->vrf_flags, BGP_VRF_IMPORT_AUTO_RT_CFGD);
5275
5276 if (!is_l3vni_live(bgp_vrf))
5277 return; /* Wait for VNI before adding rts */
5278
5279 evpn_vrf_rt_routes_unmap(bgp_vrf);
5280
5281 evpn_auto_rt_import_add_for_vrf(bgp_vrf);
5282
5283 evpn_vrf_rt_routes_map(bgp_vrf);
5284 }
5285
5286 void bgp_evpn_unconfigure_import_rt_for_vrf(struct bgp *bgp_vrf,
5287 struct ecommunity *ecomdel)
5288 {
5289 if (!CHECK_FLAG(bgp_vrf->vrf_flags, BGP_VRF_IMPORT_RT_CFGD))
5290 return; /* Already un-configured */
5291
5292 evpn_vrf_rt_routes_unmap(bgp_vrf);
5293
5294 /* Remove rt */
5295 rt_list_remove_node(bgp_vrf->vrf_import_rtl, ecomdel, true);
5296
5297 if (!rt_list_has_cfgd_rt(bgp_vrf->vrf_import_rtl))
5298 UNSET_FLAG(bgp_vrf->vrf_flags, BGP_VRF_IMPORT_RT_CFGD);
5299
5300 unconfigure_import_rt_for_vrf_fini(bgp_vrf);
5301
5302 evpn_vrf_rt_routes_map(bgp_vrf);
5303 }
5304
5305 void bgp_evpn_unconfigure_import_auto_rt_for_vrf(struct bgp *bgp_vrf)
5306 {
5307 if (!CHECK_FLAG(bgp_vrf->vrf_flags, BGP_VRF_IMPORT_AUTO_RT_CFGD))
5308 return; /* Already un-configured */
5309
5310 evpn_vrf_rt_routes_unmap(bgp_vrf);
5311
5312 /* remove auto-generated RT */
5313 evpn_auto_rt_import_delete_for_vrf(bgp_vrf);
5314
5315 UNSET_FLAG(bgp_vrf->vrf_flags, BGP_VRF_IMPORT_AUTO_RT_CFGD);
5316
5317 unconfigure_import_rt_for_vrf_fini(bgp_vrf);
5318
5319 evpn_vrf_rt_routes_map(bgp_vrf);
5320 }
5321
5322 void bgp_evpn_configure_export_rt_for_vrf(struct bgp *bgp_vrf,
5323 struct ecommunity *ecomadd)
5324 {
5325 struct vrf_route_target *newrt;
5326
5327 newrt = evpn_vrf_rt_new(ecomadd);
5328
5329 /* Remove auto generated RT if not configured */
5330 if (!CHECK_FLAG(bgp_vrf->vrf_flags, BGP_VRF_EXPORT_AUTO_RT_CFGD))
5331 evpn_auto_rt_export_delete_for_vrf(bgp_vrf);
5332
5333 /* Add the new RT to the RT list */
5334 listnode_add_sort(bgp_vrf->vrf_export_rtl, newrt);
5335
5336 SET_FLAG(bgp_vrf->vrf_flags, BGP_VRF_EXPORT_RT_CFGD);
5337
5338 if (is_l3vni_live(bgp_vrf))
5339 bgp_evpn_handle_export_rt_change_for_vrf(bgp_vrf);
5340 }
5341
5342 void bgp_evpn_configure_export_auto_rt_for_vrf(struct bgp *bgp_vrf)
5343 {
5344 if (CHECK_FLAG(bgp_vrf->vrf_flags, BGP_VRF_EXPORT_AUTO_RT_CFGD))
5345 return; /* Already configured */
5346
5347 SET_FLAG(bgp_vrf->vrf_flags, BGP_VRF_EXPORT_AUTO_RT_CFGD);
5348
5349 if (!is_l3vni_live(bgp_vrf))
5350 return; /* Wait for VNI before adding rts */
5351
5352 evpn_auto_rt_export_add_for_vrf(bgp_vrf);
5353
5354 bgp_evpn_handle_export_rt_change_for_vrf(bgp_vrf);
5355 }
5356
5357 void bgp_evpn_unconfigure_export_rt_for_vrf(struct bgp *bgp_vrf,
5358 struct ecommunity *ecomdel)
5359 {
5360 if (!CHECK_FLAG(bgp_vrf->vrf_flags, BGP_VRF_EXPORT_RT_CFGD))
5361 return; /* Already un-configured */
5362
5363 /* Remove rt */
5364 rt_list_remove_node(bgp_vrf->vrf_export_rtl, ecomdel, true);
5365
5366 if (!rt_list_has_cfgd_rt(bgp_vrf->vrf_export_rtl))
5367 UNSET_FLAG(bgp_vrf->vrf_flags, BGP_VRF_EXPORT_RT_CFGD);
5368
5369 unconfigure_export_rt_for_vrf_fini(bgp_vrf);
5370 }
5371
5372 void bgp_evpn_unconfigure_export_auto_rt_for_vrf(struct bgp *bgp_vrf)
5373 {
5374 if (!CHECK_FLAG(bgp_vrf->vrf_flags, BGP_VRF_EXPORT_AUTO_RT_CFGD))
5375 return; /* Already un-configured */
5376
5377 /* remove auto-generated RT */
5378 evpn_auto_rt_export_delete_for_vrf(bgp_vrf);
5379
5380 UNSET_FLAG(bgp_vrf->vrf_flags, BGP_VRF_EXPORT_AUTO_RT_CFGD);
5381
5382 unconfigure_export_rt_for_vrf_fini(bgp_vrf);
5383 }
5384
5385 /*
5386 * Handle change to BGP router id. This is invoked twice by the change
5387 * handler, first before the router id has been changed and then after
5388 * the router id has been changed. The first invocation will result in
5389 * local routes for all VNIs/VRF being deleted and withdrawn and the next
5390 * will result in the routes being re-advertised.
5391 */
5392 void bgp_evpn_handle_router_id_update(struct bgp *bgp, int withdraw)
5393 {
5394 struct listnode *node;
5395 struct bgp *bgp_vrf;
5396
5397 if (withdraw) {
5398
5399 /* delete and withdraw all the type-5 routes
5400 stored in the global table for this vrf
5401 */
5402 withdraw_router_id_vrf(bgp);
5403
5404 /* delete all the VNI routes (type-2/type-3) routes for all the
5405 * L2-VNIs
5406 */
5407 hash_iterate(bgp->vnihash,
5408 (void (*)(struct hash_bucket *,
5409 void *))withdraw_router_id_vni,
5410 bgp);
5411
5412 if (bgp->inst_type == BGP_INSTANCE_TYPE_DEFAULT) {
5413 for (ALL_LIST_ELEMENTS_RO(bm->bgp, node, bgp_vrf)) {
5414 if (bgp_vrf->evpn_info->advertise_pip &&
5415 (bgp_vrf->evpn_info->pip_ip_static.s_addr
5416 == INADDR_ANY))
5417 bgp_vrf->evpn_info->pip_ip.s_addr
5418 = INADDR_ANY;
5419 }
5420 }
5421 } else {
5422
5423 /* Assign new default instance router-id */
5424 if (bgp->inst_type == BGP_INSTANCE_TYPE_DEFAULT) {
5425 for (ALL_LIST_ELEMENTS_RO(bm->bgp, node, bgp_vrf)) {
5426 if (bgp_vrf->evpn_info->advertise_pip &&
5427 (bgp_vrf->evpn_info->pip_ip_static.s_addr
5428 == INADDR_ANY)) {
5429 bgp_vrf->evpn_info->pip_ip =
5430 bgp->router_id;
5431 /* advertise type-5 routes with
5432 * new nexthop
5433 */
5434 update_advertise_vrf_routes(bgp_vrf);
5435 }
5436 }
5437 }
5438
5439 /* advertise all routes in the vrf as type-5 routes with the new
5440 * RD
5441 */
5442 update_router_id_vrf(bgp);
5443
5444 /* advertise all the VNI routes (type-2/type-3) routes with the
5445 * new RD
5446 */
5447 hash_iterate(bgp->vnihash,
5448 (void (*)(struct hash_bucket *,
5449 void *))update_router_id_vni,
5450 bgp);
5451 }
5452 }
5453
5454 /*
5455 * Handle change to auto-RT algorithm - update and advertise local routes.
5456 */
5457 void bgp_evpn_handle_autort_change(struct bgp *bgp)
5458 {
5459 hash_iterate(bgp->vnihash,
5460 (void (*)(struct hash_bucket *,
5461 void*))update_autort_vni,
5462 bgp);
5463 if (bgp->l3vni)
5464 update_autort_l3vni(bgp);
5465 }
5466
5467 /*
5468 * Handle change to export RT - update and advertise local routes.
5469 */
5470 int bgp_evpn_handle_export_rt_change(struct bgp *bgp, struct bgpevpn *vpn)
5471 {
5472 return update_routes_for_vni(bgp, vpn);
5473 }
5474
5475 void bgp_evpn_handle_vrf_rd_change(struct bgp *bgp_vrf, int withdraw)
5476 {
5477 if (withdraw)
5478 delete_withdraw_vrf_routes(bgp_vrf);
5479 else
5480 update_advertise_vrf_routes(bgp_vrf);
5481 }
5482
5483 /*
5484 * Handle change to RD. This is invoked twice by the change handler,
5485 * first before the RD has been changed and then after the RD has
5486 * been changed. The first invocation will result in local routes
5487 * of this VNI being deleted and withdrawn and the next will result
5488 * in the routes being re-advertised.
5489 */
5490 void bgp_evpn_handle_rd_change(struct bgp *bgp, struct bgpevpn *vpn,
5491 int withdraw)
5492 {
5493 if (withdraw)
5494 delete_withdraw_vni_routes(bgp, vpn);
5495 else
5496 update_advertise_vni_routes(bgp, vpn);
5497 }
5498
5499 /*
5500 * Install routes for this VNI. Invoked upon change to Import RT.
5501 */
5502 int bgp_evpn_install_routes(struct bgp *bgp, struct bgpevpn *vpn)
5503 {
5504 return install_routes_for_vni(bgp, vpn);
5505 }
5506
5507 /*
5508 * Uninstall all routes installed for this VNI. Invoked upon change
5509 * to Import RT.
5510 */
5511 int bgp_evpn_uninstall_routes(struct bgp *bgp, struct bgpevpn *vpn)
5512 {
5513 return uninstall_routes_for_vni(bgp, vpn);
5514 }
5515
5516 /*
5517 * TODO: Hardcoded for a maximum of 2 VNIs right now
5518 */
5519 char *bgp_evpn_label2str(mpls_label_t *label, uint32_t num_labels, char *buf,
5520 int len)
5521 {
5522 vni_t vni1, vni2;
5523
5524 vni1 = label2vni(label);
5525 if (num_labels == 2) {
5526 vni2 = label2vni(label + 1);
5527 snprintf(buf, len, "%u/%u", vni1, vni2);
5528 } else
5529 snprintf(buf, len, "%u", vni1);
5530 return buf;
5531 }
5532
5533 /*
5534 * Function to convert evpn route to json format.
5535 * NOTE: We don't use prefix2str as the output here is a bit different.
5536 */
5537 void bgp_evpn_route2json(const struct prefix_evpn *p, json_object *json)
5538 {
5539 char buf1[ETHER_ADDR_STRLEN];
5540 char buf2[PREFIX2STR_BUFFER];
5541 uint8_t family;
5542 uint8_t prefixlen;
5543
5544 if (!json)
5545 return;
5546
5547 json_object_int_add(json, "routeType", p->prefix.route_type);
5548
5549 switch (p->prefix.route_type) {
5550 case BGP_EVPN_MAC_IP_ROUTE:
5551 json_object_int_add(json, "ethTag",
5552 p->prefix.macip_addr.eth_tag);
5553 json_object_int_add(json, "macLen", 8 * ETH_ALEN);
5554 json_object_string_add(json, "mac",
5555 prefix_mac2str(&p->prefix.macip_addr.mac, buf1,
5556 sizeof(buf1)));
5557
5558 if (!is_evpn_prefix_ipaddr_none(p)) {
5559 family = is_evpn_prefix_ipaddr_v4(p) ? AF_INET :
5560 AF_INET6;
5561 prefixlen = (family == AF_INET) ?
5562 IPV4_MAX_BITLEN : IPV6_MAX_BITLEN;
5563 inet_ntop(family, &p->prefix.macip_addr.ip.ip.addr,
5564 buf2, PREFIX2STR_BUFFER);
5565 json_object_int_add(json, "ipLen", prefixlen);
5566 json_object_string_add(json, "ip", buf2);
5567 }
5568 break;
5569
5570 case BGP_EVPN_IMET_ROUTE:
5571 json_object_int_add(json, "ethTag",
5572 p->prefix.imet_addr.eth_tag);
5573 family = is_evpn_prefix_ipaddr_v4(p) ? AF_INET : AF_INET6;
5574 prefixlen = (family == AF_INET) ? IPV4_MAX_BITLEN :
5575 IPV6_MAX_BITLEN;
5576 inet_ntop(family, &p->prefix.imet_addr.ip.ip.addr, buf2,
5577 PREFIX2STR_BUFFER);
5578 json_object_int_add(json, "ipLen", prefixlen);
5579 json_object_string_add(json, "ip", buf2);
5580 break;
5581
5582 case BGP_EVPN_IP_PREFIX_ROUTE:
5583 json_object_int_add(json, "ethTag",
5584 p->prefix.prefix_addr.eth_tag);
5585 family = is_evpn_prefix_ipaddr_v4(p) ? AF_INET : AF_INET6;
5586 inet_ntop(family, &p->prefix.prefix_addr.ip.ip.addr,
5587 buf2, sizeof(buf2));
5588 json_object_int_add(json, "ipLen",
5589 p->prefix.prefix_addr.ip_prefix_length);
5590 json_object_string_add(json, "ip", buf2);
5591 break;
5592
5593 default:
5594 break;
5595 }
5596 }
5597
5598 /*
5599 * Encode EVPN prefix in Update (MP_REACH)
5600 */
5601 void bgp_evpn_encode_prefix(struct stream *s, const struct prefix *p,
5602 const struct prefix_rd *prd, mpls_label_t *label,
5603 uint32_t num_labels, struct attr *attr,
5604 bool addpath_capable, uint32_t addpath_tx_id)
5605 {
5606 struct prefix_evpn *evp = (struct prefix_evpn *)p;
5607 int len, ipa_len = 0;
5608
5609 if (addpath_capable)
5610 stream_putl(s, addpath_tx_id);
5611
5612 /* Route type */
5613 stream_putc(s, evp->prefix.route_type);
5614
5615 switch (evp->prefix.route_type) {
5616 case BGP_EVPN_MAC_IP_ROUTE:
5617 if (is_evpn_prefix_ipaddr_v4(evp))
5618 ipa_len = IPV4_MAX_BYTELEN;
5619 else if (is_evpn_prefix_ipaddr_v6(evp))
5620 ipa_len = IPV6_MAX_BYTELEN;
5621 /* RD, ESI, EthTag, MAC+len, IP len, [IP], 1 VNI */
5622 len = 8 + 10 + 4 + 1 + 6 + 1 + ipa_len + 3;
5623 if (ipa_len && num_labels > 1) /* There are 2 VNIs */
5624 len += 3;
5625 stream_putc(s, len);
5626 stream_put(s, prd->val, 8); /* RD */
5627 if (attr)
5628 stream_put(s, &attr->esi, ESI_BYTES);
5629 else
5630 stream_put(s, 0, 10);
5631 stream_putl(s, evp->prefix.macip_addr.eth_tag); /* Ethernet Tag ID */
5632 stream_putc(s, 8 * ETH_ALEN); /* Mac Addr Len - bits */
5633 stream_put(s, evp->prefix.macip_addr.mac.octet, 6); /* Mac Addr */
5634 stream_putc(s, 8 * ipa_len); /* IP address Length */
5635 if (ipa_len) /* IP */
5636 stream_put(s, &evp->prefix.macip_addr.ip.ip.addr,
5637 ipa_len);
5638 /* 1st label is the L2 VNI */
5639 stream_put(s, label, BGP_LABEL_BYTES);
5640 /* Include 2nd label (L3 VNI) if advertising MAC+IP */
5641 if (ipa_len && num_labels > 1)
5642 stream_put(s, label + 1, BGP_LABEL_BYTES);
5643 break;
5644
5645 case BGP_EVPN_IMET_ROUTE:
5646 stream_putc(s, 17); // TODO: length - assumes IPv4 address
5647 stream_put(s, prd->val, 8); /* RD */
5648 stream_putl(s, evp->prefix.imet_addr.eth_tag); /* Ethernet Tag ID */
5649 stream_putc(s, IPV4_MAX_BITLEN); /* IP address Length - bits */
5650 /* Originating Router's IP Addr */
5651 stream_put_in_addr(s, &evp->prefix.imet_addr.ip.ipaddr_v4);
5652 break;
5653
5654 case BGP_EVPN_ES_ROUTE:
5655 stream_putc(s, 23); /* TODO: length: assumes ipv4 VTEP */
5656 stream_put(s, prd->val, 8); /* RD */
5657 stream_put(s, evp->prefix.es_addr.esi.val, 10); /* ESI */
5658 stream_putc(s, IPV4_MAX_BITLEN); /* IP address Length - bits */
5659 /* VTEP IP */
5660 stream_put_in_addr(s, &evp->prefix.es_addr.ip.ipaddr_v4);
5661 break;
5662
5663 case BGP_EVPN_AD_ROUTE:
5664 /* RD, ESI, EthTag, 1 VNI */
5665 len = RD_BYTES + ESI_BYTES + EVPN_ETH_TAG_BYTES + BGP_LABEL_BYTES;
5666 stream_putc(s, len);
5667 stream_put(s, prd->val, RD_BYTES); /* RD */
5668 stream_put(s, evp->prefix.ead_addr.esi.val, ESI_BYTES); /* ESI */
5669 stream_putl(s, evp->prefix.ead_addr.eth_tag); /* Ethernet Tag */
5670 stream_put(s, label, BGP_LABEL_BYTES);
5671 break;
5672
5673 case BGP_EVPN_IP_PREFIX_ROUTE:
5674 /* TODO: AddPath support. */
5675 evpn_mpattr_encode_type5(s, p, prd, label, num_labels, attr);
5676 break;
5677
5678 default:
5679 break;
5680 }
5681 }
5682
5683 int bgp_nlri_parse_evpn(struct peer *peer, struct attr *attr,
5684 struct bgp_nlri *packet, int withdraw)
5685 {
5686 uint8_t *pnt;
5687 uint8_t *lim;
5688 afi_t afi;
5689 safi_t safi;
5690 uint32_t addpath_id;
5691 bool addpath_capable;
5692 int psize = 0;
5693 uint8_t rtype;
5694 struct prefix p;
5695
5696 /* Start processing the NLRI - there may be multiple in the MP_REACH */
5697 pnt = packet->nlri;
5698 lim = pnt + packet->length;
5699 afi = packet->afi;
5700 safi = packet->safi;
5701 addpath_id = 0;
5702
5703 addpath_capable = bgp_addpath_encode_rx(peer, afi, safi);
5704
5705 for (; pnt < lim; pnt += psize) {
5706 /* Clear prefix structure. */
5707 memset(&p, 0, sizeof(p));
5708
5709 /* Deal with path-id if AddPath is supported. */
5710 if (addpath_capable) {
5711 /* When packet overflow occurs return immediately. */
5712 if (pnt + BGP_ADDPATH_ID_LEN > lim)
5713 return BGP_NLRI_PARSE_ERROR_PACKET_OVERFLOW;
5714
5715 memcpy(&addpath_id, pnt, BGP_ADDPATH_ID_LEN);
5716 addpath_id = ntohl(addpath_id);
5717 pnt += BGP_ADDPATH_ID_LEN;
5718 }
5719
5720 /* All EVPN NLRI types start with type and length. */
5721 if (pnt + 2 > lim)
5722 return BGP_NLRI_PARSE_ERROR_EVPN_MISSING_TYPE;
5723
5724 rtype = *pnt++;
5725 psize = *pnt++;
5726
5727 /* When packet overflow occur return immediately. */
5728 if (pnt + psize > lim)
5729 return BGP_NLRI_PARSE_ERROR_PACKET_OVERFLOW;
5730
5731 switch (rtype) {
5732 case BGP_EVPN_MAC_IP_ROUTE:
5733 if (process_type2_route(peer, afi, safi,
5734 withdraw ? NULL : attr, pnt,
5735 psize, addpath_id)) {
5736 flog_err(
5737 EC_BGP_EVPN_FAIL,
5738 "%u:%s - Error in processing EVPN type-2 NLRI size %d",
5739 peer->bgp->vrf_id, peer->host, psize);
5740 return BGP_NLRI_PARSE_ERROR_EVPN_TYPE2_SIZE;
5741 }
5742 break;
5743
5744 case BGP_EVPN_IMET_ROUTE:
5745 if (process_type3_route(peer, afi, safi,
5746 withdraw ? NULL : attr, pnt,
5747 psize, addpath_id)) {
5748 flog_err(
5749 EC_BGP_PKT_PROCESS,
5750 "%u:%s - Error in processing EVPN type-3 NLRI size %d",
5751 peer->bgp->vrf_id, peer->host, psize);
5752 return BGP_NLRI_PARSE_ERROR_EVPN_TYPE3_SIZE;
5753 }
5754 break;
5755
5756 case BGP_EVPN_ES_ROUTE:
5757 if (bgp_evpn_type4_route_process(peer, afi, safi,
5758 withdraw ? NULL : attr, pnt,
5759 psize, addpath_id)) {
5760 flog_err(
5761 EC_BGP_PKT_PROCESS,
5762 "%u:%s - Error in processing EVPN type-4 NLRI size %d",
5763 peer->bgp->vrf_id, peer->host, psize);
5764 return BGP_NLRI_PARSE_ERROR_EVPN_TYPE4_SIZE;
5765 }
5766 break;
5767
5768 case BGP_EVPN_AD_ROUTE:
5769 if (bgp_evpn_type1_route_process(peer, afi, safi,
5770 withdraw ? NULL : attr, pnt,
5771 psize, addpath_id)) {
5772 flog_err(
5773 EC_BGP_PKT_PROCESS,
5774 "%u:%s - Error in processing EVPN type-1 NLRI size %d",
5775 peer->bgp->vrf_id, peer->host, psize);
5776 return BGP_NLRI_PARSE_ERROR_EVPN_TYPE1_SIZE;
5777 }
5778 break;
5779
5780 case BGP_EVPN_IP_PREFIX_ROUTE:
5781 if (process_type5_route(peer, afi, safi,
5782 withdraw ? NULL : attr, pnt,
5783 psize, addpath_id)) {
5784 flog_err(
5785 EC_BGP_PKT_PROCESS,
5786 "%u:%s - Error in processing EVPN type-5 NLRI size %d",
5787 peer->bgp->vrf_id, peer->host, psize);
5788 return BGP_NLRI_PARSE_ERROR_EVPN_TYPE5_SIZE;
5789 }
5790 break;
5791
5792 default:
5793 break;
5794 }
5795 }
5796
5797 /* Packet length consistency check. */
5798 if (pnt != lim)
5799 return BGP_NLRI_PARSE_ERROR_PACKET_LENGTH;
5800
5801 return BGP_NLRI_PARSE_OK;
5802 }
5803
5804 /*
5805 * Map the RTs (configured or automatically derived) of a VRF to the VRF.
5806 * The mapping will be used during route processing.
5807 * bgp_vrf: specific bgp vrf instance on which RT is configured
5808 */
5809 void bgp_evpn_map_vrf_to_its_rts(struct bgp *bgp_vrf)
5810 {
5811 struct listnode *node, *nnode;
5812 struct vrf_route_target *l3rt;
5813
5814 for (ALL_LIST_ELEMENTS(bgp_vrf->vrf_import_rtl, node, nnode, l3rt))
5815 map_vrf_to_rt(bgp_vrf, l3rt);
5816 }
5817
5818 /*
5819 * Unmap the RTs (configured or automatically derived) of a VRF from the VRF.
5820 */
5821 void bgp_evpn_unmap_vrf_from_its_rts(struct bgp *bgp_vrf)
5822 {
5823 struct listnode *node, *nnode;
5824 struct vrf_route_target *l3rt;
5825
5826 for (ALL_LIST_ELEMENTS(bgp_vrf->vrf_import_rtl, node, nnode, l3rt))
5827 unmap_vrf_from_rt(bgp_vrf, l3rt);
5828 }
5829
5830 /*
5831 * Map the RTs (configured or automatically derived) of a VNI to the VNI.
5832 * The mapping will be used during route processing.
5833 */
5834 void bgp_evpn_map_vni_to_its_rts(struct bgp *bgp, struct bgpevpn *vpn)
5835 {
5836 uint32_t i;
5837 struct ecommunity_val *eval;
5838 struct listnode *node, *nnode;
5839 struct ecommunity *ecom;
5840
5841 for (ALL_LIST_ELEMENTS(vpn->import_rtl, node, nnode, ecom)) {
5842 for (i = 0; i < ecom->size; i++) {
5843 eval = (struct ecommunity_val *)(ecom->val
5844 + (i
5845 * ECOMMUNITY_SIZE));
5846 map_vni_to_rt(bgp, vpn, eval);
5847 }
5848 }
5849 }
5850
5851 /*
5852 * Unmap the RTs (configured or automatically derived) of a VNI from the VNI.
5853 */
5854 void bgp_evpn_unmap_vni_from_its_rts(struct bgp *bgp, struct bgpevpn *vpn)
5855 {
5856 uint32_t i;
5857 struct ecommunity_val *eval;
5858 struct listnode *node, *nnode;
5859 struct ecommunity *ecom;
5860
5861 for (ALL_LIST_ELEMENTS(vpn->import_rtl, node, nnode, ecom)) {
5862 for (i = 0; i < ecom->size; i++) {
5863 struct irt_node *irt;
5864 struct ecommunity_val eval_tmp;
5865
5866 eval = (struct ecommunity_val *)(ecom->val
5867 + (i
5868 * ECOMMUNITY_SIZE));
5869 /* If using "automatic" RT, we only care about the
5870 * local-admin sub-field.
5871 * This is to facilitate using VNI as the RT for EBGP
5872 * peering too.
5873 */
5874 memcpy(&eval_tmp, eval, ECOMMUNITY_SIZE);
5875 if (!is_import_rt_configured(vpn))
5876 mask_ecom_global_admin(&eval_tmp, eval);
5877
5878 irt = lookup_import_rt(bgp, &eval_tmp);
5879 if (irt)
5880 unmap_vni_from_rt(bgp, vpn, irt);
5881 }
5882 }
5883 }
5884
5885 /*
5886 * Derive Import RT automatically for VNI and map VNI to RT.
5887 * The mapping will be used during route processing.
5888 */
5889 void bgp_evpn_derive_auto_rt_import(struct bgp *bgp, struct bgpevpn *vpn)
5890 {
5891 form_auto_rt(bgp, vpn->vni, vpn->import_rtl, false);
5892 UNSET_FLAG(vpn->flags, VNI_FLAG_IMPRT_CFGD);
5893
5894 /* Map RT to VNI */
5895 bgp_evpn_map_vni_to_its_rts(bgp, vpn);
5896 }
5897
5898 /*
5899 * Derive Export RT automatically for VNI.
5900 */
5901 void bgp_evpn_derive_auto_rt_export(struct bgp *bgp, struct bgpevpn *vpn)
5902 {
5903 form_auto_rt(bgp, vpn->vni, vpn->export_rtl, false);
5904 UNSET_FLAG(vpn->flags, VNI_FLAG_EXPRT_CFGD);
5905 }
5906
5907 /*
5908 * Derive RD automatically for VNI using passed information - it
5909 * is of the form RouterId:unique-id-for-vni.
5910 */
5911 void bgp_evpn_derive_auto_rd_for_vrf(struct bgp *bgp)
5912 {
5913 if (is_vrf_rd_configured(bgp))
5914 return;
5915
5916 form_auto_rd(bgp->router_id, bgp->vrf_rd_id, &bgp->vrf_prd);
5917 }
5918
5919 /*
5920 * Derive RD automatically for VNI using passed information - it
5921 * is of the form RouterId:unique-id-for-vni.
5922 */
5923 void bgp_evpn_derive_auto_rd(struct bgp *bgp, struct bgpevpn *vpn)
5924 {
5925 char buf[BGP_EVPN_PREFIX_RD_LEN];
5926
5927 vpn->prd.family = AF_UNSPEC;
5928 vpn->prd.prefixlen = 64;
5929 snprintfrr(buf, sizeof(buf), "%pI4:%hu", &bgp->router_id, vpn->rd_id);
5930 (void)str2prefix_rd(buf, &vpn->prd);
5931 UNSET_FLAG(vpn->flags, VNI_FLAG_RD_CFGD);
5932 }
5933
5934 /*
5935 * Lookup L3-VNI
5936 */
5937 bool bgp_evpn_lookup_l3vni_l2vni_table(vni_t vni)
5938 {
5939 struct list *inst = bm->bgp;
5940 struct listnode *node;
5941 struct bgp *bgp_vrf;
5942
5943 for (ALL_LIST_ELEMENTS_RO(inst, node, bgp_vrf)) {
5944 if (bgp_vrf->l3vni == vni)
5945 return true;
5946 }
5947
5948 return false;
5949 }
5950
5951 /*
5952 * Lookup VNI.
5953 */
5954 struct bgpevpn *bgp_evpn_lookup_vni(struct bgp *bgp, vni_t vni)
5955 {
5956 struct bgpevpn *vpn;
5957 struct bgpevpn tmp;
5958
5959 memset(&tmp, 0, sizeof(tmp));
5960 tmp.vni = vni;
5961 vpn = hash_lookup(bgp->vnihash, &tmp);
5962 return vpn;
5963 }
5964
5965 /*
5966 * Create a new vpn - invoked upon configuration or zebra notification.
5967 */
5968 struct bgpevpn *bgp_evpn_new(struct bgp *bgp, vni_t vni,
5969 struct in_addr originator_ip,
5970 vrf_id_t tenant_vrf_id,
5971 struct in_addr mcast_grp,
5972 ifindex_t svi_ifindex)
5973 {
5974 struct bgpevpn *vpn;
5975
5976 vpn = XCALLOC(MTYPE_BGP_EVPN, sizeof(struct bgpevpn));
5977
5978 /* Set values - RD and RT set to defaults. */
5979 vpn->vni = vni;
5980 vpn->originator_ip = originator_ip;
5981 vpn->tenant_vrf_id = tenant_vrf_id;
5982 vpn->mcast_grp = mcast_grp;
5983 vpn->svi_ifindex = svi_ifindex;
5984
5985 /* Initialize route-target import and export lists */
5986 vpn->import_rtl = list_new();
5987 vpn->import_rtl->cmp =
5988 (int (*)(void *, void *))bgp_evpn_route_target_cmp;
5989 vpn->import_rtl->del = bgp_evpn_xxport_delete_ecomm;
5990 vpn->export_rtl = list_new();
5991 vpn->export_rtl->cmp =
5992 (int (*)(void *, void *))bgp_evpn_route_target_cmp;
5993 vpn->export_rtl->del = bgp_evpn_xxport_delete_ecomm;
5994 bf_assign_index(bm->rd_idspace, vpn->rd_id);
5995 derive_rd_rt_for_vni(bgp, vpn);
5996
5997 /* Initialize EVPN route tables. */
5998 vpn->ip_table = bgp_table_init(bgp, AFI_L2VPN, SAFI_EVPN);
5999 vpn->mac_table = bgp_table_init(bgp, AFI_L2VPN, SAFI_EVPN);
6000
6001 /* Add to hash */
6002 (void)hash_get(bgp->vnihash, vpn, hash_alloc_intern);
6003
6004 bgp_evpn_remote_ip_hash_init(vpn);
6005 bgp_evpn_link_to_vni_svi_hash(bgp, vpn);
6006
6007 /* add to l2vni list on corresponding vrf */
6008 bgpevpn_link_to_l3vni(vpn);
6009
6010 bgp_evpn_vni_es_init(vpn);
6011
6012 QOBJ_REG(vpn, bgpevpn);
6013 return vpn;
6014 }
6015
6016 /*
6017 * Free a given VPN - called in multiple scenarios such as zebra
6018 * notification, configuration being deleted, advertise-all-vni disabled etc.
6019 * This just frees appropriate memory, caller should have taken other
6020 * needed actions.
6021 */
6022 void bgp_evpn_free(struct bgp *bgp, struct bgpevpn *vpn)
6023 {
6024 bgp_evpn_remote_ip_hash_destroy(vpn);
6025 bgp_evpn_vni_es_cleanup(vpn);
6026 bgpevpn_unlink_from_l3vni(vpn);
6027 bgp_table_unlock(vpn->ip_table);
6028 bgp_table_unlock(vpn->mac_table);
6029 bgp_evpn_unmap_vni_from_its_rts(bgp, vpn);
6030 list_delete(&vpn->import_rtl);
6031 list_delete(&vpn->export_rtl);
6032 bf_release_index(bm->rd_idspace, vpn->rd_id);
6033 hash_release(bgp->vni_svi_hash, vpn);
6034 hash_release(bgp->vnihash, vpn);
6035 QOBJ_UNREG(vpn);
6036 XFREE(MTYPE_BGP_EVPN, vpn);
6037 }
6038
6039 static void hash_evpn_free(struct bgpevpn *vpn)
6040 {
6041 XFREE(MTYPE_BGP_EVPN, vpn);
6042 }
6043
6044 /*
6045 * Import evpn route from global table to VNI/VRF/ESI.
6046 */
6047 int bgp_evpn_import_route(struct bgp *bgp, afi_t afi, safi_t safi,
6048 const struct prefix *p, struct bgp_path_info *pi)
6049 {
6050 return install_uninstall_evpn_route(bgp, afi, safi, p, pi, 1);
6051 }
6052
6053 /*
6054 * Unimport evpn route from VNI/VRF/ESI.
6055 */
6056 int bgp_evpn_unimport_route(struct bgp *bgp, afi_t afi, safi_t safi,
6057 const struct prefix *p, struct bgp_path_info *pi)
6058 {
6059 return install_uninstall_evpn_route(bgp, afi, safi, p, pi, 0);
6060 }
6061
6062 /* filter routes which have martian next hops */
6063 int bgp_filter_evpn_routes_upon_martian_nh_change(struct bgp *bgp)
6064 {
6065 afi_t afi;
6066 safi_t safi;
6067 struct bgp_dest *rd_dest, *dest;
6068 struct bgp_table *table;
6069 struct bgp_path_info *pi;
6070
6071 afi = AFI_L2VPN;
6072 safi = SAFI_EVPN;
6073
6074 /* Walk entire global routing table and evaluate routes which could be
6075 * imported into this VPN. Note that we cannot just look at the routes
6076 * for the VNI's RD -
6077 * remote routes applicable for this VNI could have any RD.
6078 */
6079 /* EVPN routes are a 2-level table. */
6080 for (rd_dest = bgp_table_top(bgp->rib[afi][safi]); rd_dest;
6081 rd_dest = bgp_route_next(rd_dest)) {
6082 table = bgp_dest_get_bgp_table_info(rd_dest);
6083 if (!table)
6084 continue;
6085
6086 for (dest = bgp_table_top(table); dest;
6087 dest = bgp_route_next(dest)) {
6088
6089 for (pi = bgp_dest_get_bgp_path_info(dest); pi;
6090 pi = pi->next) {
6091
6092 /* Consider "valid" remote routes applicable for
6093 * this VNI. */
6094 if (!(pi->type == ZEBRA_ROUTE_BGP
6095 && pi->sub_type == BGP_ROUTE_NORMAL))
6096 continue;
6097 if (bgp_nexthop_self(bgp, afi, pi->type,
6098 pi->sub_type, pi->attr,
6099 dest)) {
6100 const struct prefix *p =
6101 bgp_dest_get_prefix(dest);
6102
6103 if (bgp_debug_update(pi->peer, p, NULL,
6104 1)) {
6105 char attr_str[BUFSIZ] = {0};
6106
6107 bgp_dump_attr(pi->attr,
6108 attr_str,
6109 sizeof(attr_str));
6110
6111 zlog_debug(
6112 "%u: prefix %pBD with attr %s - DENIED due to martian or self nexthop",
6113 bgp->vrf_id, dest,
6114 attr_str);
6115 }
6116 bgp_evpn_unimport_route(bgp, afi, safi,
6117 p, pi);
6118
6119 bgp_rib_remove(dest, pi, pi->peer, afi,
6120 safi);
6121 }
6122 }
6123 }
6124 }
6125
6126 return 0;
6127 }
6128
6129 /*
6130 * Handle del of a local MACIP.
6131 */
6132 int bgp_evpn_local_macip_del(struct bgp *bgp, vni_t vni, struct ethaddr *mac,
6133 struct ipaddr *ip, int state)
6134 {
6135 struct bgpevpn *vpn;
6136 struct prefix_evpn p;
6137 struct bgp_dest *dest;
6138
6139 /* Lookup VNI hash - should exist. */
6140 vpn = bgp_evpn_lookup_vni(bgp, vni);
6141 if (!vpn || !is_vni_live(vpn)) {
6142 flog_warn(EC_BGP_EVPN_VPN_VNI,
6143 "%u: VNI hash entry for VNI %u %s at MACIP DEL",
6144 bgp->vrf_id, vni, vpn ? "not live" : "not found");
6145 return -1;
6146 }
6147
6148 build_evpn_type2_prefix(&p, mac, ip);
6149 if (state == ZEBRA_NEIGH_ACTIVE) {
6150 /* Remove EVPN type-2 route and schedule for processing. */
6151 delete_evpn_route(bgp, vpn, &p);
6152 } else {
6153 /* Re-instate the current remote best path if any */
6154 dest = bgp_evpn_vni_node_lookup(vpn, &p, NULL);
6155 if (dest) {
6156 evpn_zebra_reinstall_best_route(bgp, vpn, dest);
6157 bgp_dest_unlock_node(dest);
6158 }
6159 }
6160
6161 return 0;
6162 }
6163
6164 /*
6165 * Handle add of a local MACIP.
6166 */
6167 int bgp_evpn_local_macip_add(struct bgp *bgp, vni_t vni, struct ethaddr *mac,
6168 struct ipaddr *ip, uint8_t flags, uint32_t seq, esi_t *esi)
6169 {
6170 struct bgpevpn *vpn;
6171 struct prefix_evpn p;
6172
6173 /* Lookup VNI hash - should exist. */
6174 vpn = bgp_evpn_lookup_vni(bgp, vni);
6175 if (!vpn || !is_vni_live(vpn)) {
6176 flog_warn(EC_BGP_EVPN_VPN_VNI,
6177 "%u: VNI hash entry for VNI %u %s at MACIP ADD",
6178 bgp->vrf_id, vni, vpn ? "not live" : "not found");
6179 return -1;
6180 }
6181
6182 /* Create EVPN type-2 route and schedule for processing. */
6183 build_evpn_type2_prefix(&p, mac, ip);
6184 if (update_evpn_route(bgp, vpn, &p, flags, seq, esi)) {
6185 flog_err(
6186 EC_BGP_EVPN_ROUTE_CREATE,
6187 "%u:Failed to create Type-2 route, VNI %u %s MAC %pEA IP %pIA (flags: 0x%x)",
6188 bgp->vrf_id, vpn->vni,
6189 CHECK_FLAG(flags, ZEBRA_MACIP_TYPE_STICKY)
6190 ? "sticky gateway"
6191 : "",
6192 mac, ip, flags);
6193 return -1;
6194 }
6195
6196 return 0;
6197 }
6198
6199 static void link_l2vni_hash_to_l3vni(struct hash_bucket *bucket,
6200 struct bgp *bgp_vrf)
6201 {
6202 struct bgpevpn *vpn = (struct bgpevpn *)bucket->data;
6203 struct bgp *bgp_evpn = NULL;
6204
6205 bgp_evpn = bgp_get_evpn();
6206 assert(bgp_evpn);
6207
6208 if (vpn->tenant_vrf_id == bgp_vrf->vrf_id)
6209 bgpevpn_link_to_l3vni(vpn);
6210 }
6211
6212 int bgp_evpn_local_l3vni_add(vni_t l3vni, vrf_id_t vrf_id,
6213 struct ethaddr *svi_rmac,
6214 struct ethaddr *vrr_rmac,
6215 struct in_addr originator_ip, int filter,
6216 ifindex_t svi_ifindex,
6217 bool is_anycast_mac)
6218 {
6219 struct bgp *bgp_vrf = NULL; /* bgp VRF instance */
6220 struct bgp *bgp_evpn = NULL; /* EVPN bgp instance */
6221 struct listnode *node = NULL;
6222 struct bgpevpn *vpn = NULL;
6223 as_t as = 0;
6224
6225 /* get the EVPN instance - required to get the AS number for VRF
6226 * auto-creatio
6227 */
6228 bgp_evpn = bgp_get_evpn();
6229 if (!bgp_evpn) {
6230 flog_err(
6231 EC_BGP_NO_DFLT,
6232 "Cannot process L3VNI %u ADD - EVPN BGP instance not yet created",
6233 l3vni);
6234 return -1;
6235 }
6236 as = bgp_evpn->as;
6237
6238 /* if the BGP vrf instance doesn't exist - create one */
6239 bgp_vrf = bgp_lookup_by_vrf_id(vrf_id);
6240 if (!bgp_vrf) {
6241
6242 int ret = 0;
6243
6244 ret = bgp_get_vty(&bgp_vrf, &as, vrf_id_to_name(vrf_id),
6245 vrf_id == VRF_DEFAULT
6246 ? BGP_INSTANCE_TYPE_DEFAULT
6247 : BGP_INSTANCE_TYPE_VRF);
6248 switch (ret) {
6249 case BGP_ERR_AS_MISMATCH:
6250 flog_err(EC_BGP_EVPN_AS_MISMATCH,
6251 "BGP instance is already running; AS is %u",
6252 as);
6253 return -1;
6254 case BGP_ERR_INSTANCE_MISMATCH:
6255 flog_err(EC_BGP_EVPN_INSTANCE_MISMATCH,
6256 "BGP instance type mismatch");
6257 return -1;
6258 }
6259
6260 /* mark as auto created */
6261 SET_FLAG(bgp_vrf->vrf_flags, BGP_VRF_AUTO);
6262 }
6263
6264 /* associate the vrf with l3vni and related parameters */
6265 bgp_vrf->l3vni = l3vni;
6266 bgp_vrf->originator_ip = originator_ip;
6267 bgp_vrf->l3vni_svi_ifindex = svi_ifindex;
6268 bgp_vrf->evpn_info->is_anycast_mac = is_anycast_mac;
6269
6270 /* copy anycast MAC from VRR MAC */
6271 memcpy(&bgp_vrf->rmac, vrr_rmac, ETH_ALEN);
6272 /* copy sys RMAC from SVI MAC */
6273 memcpy(&bgp_vrf->evpn_info->pip_rmac_zebra, svi_rmac, ETH_ALEN);
6274 /* PIP user configured mac is not present use svi mac as sys mac */
6275 if (is_zero_mac(&bgp_vrf->evpn_info->pip_rmac_static))
6276 memcpy(&bgp_vrf->evpn_info->pip_rmac, svi_rmac, ETH_ALEN);
6277
6278 if (bgp_debug_zebra(NULL))
6279 zlog_debug(
6280 "VRF %s vni %u pip %s RMAC %pEA sys RMAC %pEA static RMAC %pEA is_anycast_mac %s",
6281 vrf_id_to_name(bgp_vrf->vrf_id), bgp_vrf->l3vni,
6282 bgp_vrf->evpn_info->advertise_pip ? "enable"
6283 : "disable",
6284 &bgp_vrf->rmac, &bgp_vrf->evpn_info->pip_rmac,
6285 &bgp_vrf->evpn_info->pip_rmac_static,
6286 is_anycast_mac ? "Enable" : "Disable");
6287
6288 /* set the right filter - are we using l3vni only for prefix routes? */
6289 if (filter) {
6290 SET_FLAG(bgp_vrf->vrf_flags, BGP_VRF_L3VNI_PREFIX_ROUTES_ONLY);
6291
6292 /*
6293 * VNI_FLAG_USE_TWO_LABELS flag for linked L2VNIs should not be
6294 * set before linking vrf to L3VNI. Thus, no need to clear
6295 * that explicitly.
6296 */
6297 } else {
6298 UNSET_FLAG(bgp_vrf->vrf_flags,
6299 BGP_VRF_L3VNI_PREFIX_ROUTES_ONLY);
6300
6301 for (ALL_LIST_ELEMENTS_RO(bgp_vrf->l2vnis, node, vpn)) {
6302 if (!CHECK_FLAG(vpn->flags, VNI_FLAG_USE_TWO_LABELS)) {
6303
6304 /*
6305 * If we are flapping VNI_FLAG_USE_TWO_LABELS
6306 * flag, update all MACIP routes in this VNI
6307 */
6308 SET_FLAG(vpn->flags, VNI_FLAG_USE_TWO_LABELS);
6309 update_all_type2_routes(bgp_evpn, vpn);
6310 }
6311 }
6312 }
6313
6314 /* Map auto derive or configured RTs */
6315 if (!CHECK_FLAG(bgp_vrf->vrf_flags, BGP_VRF_IMPORT_RT_CFGD) ||
6316 CHECK_FLAG(bgp_vrf->vrf_flags, BGP_VRF_IMPORT_AUTO_RT_CFGD))
6317 evpn_auto_rt_import_add_for_vrf(bgp_vrf);
6318 else
6319 bgp_evpn_map_vrf_to_its_rts(bgp_vrf);
6320
6321 if (!CHECK_FLAG(bgp_vrf->vrf_flags, BGP_VRF_EXPORT_RT_CFGD) ||
6322 CHECK_FLAG(bgp_vrf->vrf_flags, BGP_VRF_EXPORT_AUTO_RT_CFGD))
6323 evpn_auto_rt_export_add_for_vrf(bgp_vrf);
6324
6325 /* auto derive RD */
6326 bgp_evpn_derive_auto_rd_for_vrf(bgp_vrf);
6327
6328 /* link all corresponding l2vnis */
6329 hash_iterate(bgp_evpn->vnihash,
6330 (void (*)(struct hash_bucket *,
6331 void *))link_l2vni_hash_to_l3vni,
6332 bgp_vrf);
6333
6334 /* Only update all corresponding type-2 routes if we are advertising two
6335 * labels along with type-2 routes
6336 */
6337 if (!filter)
6338 for (ALL_LIST_ELEMENTS_RO(bgp_vrf->l2vnis, node, vpn))
6339 update_routes_for_vni(bgp_evpn, vpn);
6340
6341 /* advertise type-5 routes if needed */
6342 update_advertise_vrf_routes(bgp_vrf);
6343
6344 /* install all remote routes belonging to this l3vni into correspondng
6345 * vrf */
6346 install_routes_for_vrf(bgp_vrf);
6347
6348 return 0;
6349 }
6350
6351 int bgp_evpn_local_l3vni_del(vni_t l3vni, vrf_id_t vrf_id)
6352 {
6353 struct bgp *bgp_vrf = NULL; /* bgp vrf instance */
6354 struct bgp *bgp_evpn = NULL; /* EVPN bgp instance */
6355 struct listnode *node = NULL;
6356 struct listnode *next = NULL;
6357 struct bgpevpn *vpn = NULL;
6358
6359 bgp_vrf = bgp_lookup_by_vrf_id(vrf_id);
6360 if (!bgp_vrf) {
6361 flog_err(
6362 EC_BGP_NO_DFLT,
6363 "Cannot process L3VNI %u Del - Could not find BGP instance",
6364 l3vni);
6365 return -1;
6366 }
6367
6368 bgp_evpn = bgp_get_evpn();
6369 if (!bgp_evpn) {
6370 flog_err(
6371 EC_BGP_NO_DFLT,
6372 "Cannot process L3VNI %u Del - Could not find EVPN BGP instance",
6373 l3vni);
6374 return -1;
6375 }
6376
6377 /* Remove remote routes from BGT VRF even if BGP_VRF_AUTO is configured,
6378 * bgp_delete would not remove/decrement bgp_path_info of the ip_prefix
6379 * routes. This will uninstalling the routes from zebra and decremnt the
6380 * bgp info count.
6381 */
6382 uninstall_routes_for_vrf(bgp_vrf);
6383
6384 /* delete/withdraw all type-5 routes */
6385 delete_withdraw_vrf_routes(bgp_vrf);
6386
6387 /* remove the l3vni from vrf instance */
6388 bgp_vrf->l3vni = 0;
6389
6390 /* remove the Rmac from the BGP vrf */
6391 memset(&bgp_vrf->rmac, 0, sizeof(struct ethaddr));
6392 memset(&bgp_vrf->evpn_info->pip_rmac_zebra, 0, ETH_ALEN);
6393 if (is_zero_mac(&bgp_vrf->evpn_info->pip_rmac_static) &&
6394 !is_zero_mac(&bgp_vrf->evpn_info->pip_rmac))
6395 memset(&bgp_vrf->evpn_info->pip_rmac, 0, ETH_ALEN);
6396
6397 /* remove default import RT or Unmap non-default import RT */
6398 if (!list_isempty(bgp_vrf->vrf_import_rtl)) {
6399 bgp_evpn_unmap_vrf_from_its_rts(bgp_vrf);
6400 if (!CHECK_FLAG(bgp_vrf->vrf_flags, BGP_VRF_IMPORT_RT_CFGD))
6401 list_delete_all_node(bgp_vrf->vrf_import_rtl);
6402 }
6403
6404 /* remove default export RT */
6405 if (!list_isempty(bgp_vrf->vrf_export_rtl) &&
6406 !CHECK_FLAG(bgp_vrf->vrf_flags, BGP_VRF_EXPORT_RT_CFGD)) {
6407 list_delete_all_node(bgp_vrf->vrf_export_rtl);
6408 }
6409
6410 /* update all corresponding local mac-ip routes */
6411 if (!CHECK_FLAG(bgp_vrf->vrf_flags, BGP_VRF_L3VNI_PREFIX_ROUTES_ONLY)) {
6412 for (ALL_LIST_ELEMENTS_RO(bgp_vrf->l2vnis, node, vpn)) {
6413 UNSET_FLAG(vpn->flags, VNI_FLAG_USE_TWO_LABELS);
6414 update_routes_for_vni(bgp_evpn, vpn);
6415 }
6416 }
6417
6418 /* If any L2VNIs point to this instance, unlink them. */
6419 for (ALL_LIST_ELEMENTS(bgp_vrf->l2vnis, node, next, vpn))
6420 bgpevpn_unlink_from_l3vni(vpn);
6421
6422 UNSET_FLAG(bgp_vrf->vrf_flags, BGP_VRF_L3VNI_PREFIX_ROUTES_ONLY);
6423
6424 /* Delete the instance if it was autocreated */
6425 if (CHECK_FLAG(bgp_vrf->vrf_flags, BGP_VRF_AUTO))
6426 bgp_delete(bgp_vrf);
6427
6428 return 0;
6429 }
6430
6431 /*
6432 * Handle del of a local VNI.
6433 */
6434 int bgp_evpn_local_vni_del(struct bgp *bgp, vni_t vni)
6435 {
6436 struct bgpevpn *vpn;
6437
6438 /* Locate VNI hash */
6439 vpn = bgp_evpn_lookup_vni(bgp, vni);
6440 if (!vpn)
6441 return 0;
6442
6443 /* Remove all local EVPN routes and schedule for processing (to
6444 * withdraw from peers).
6445 */
6446 delete_routes_for_vni(bgp, vpn);
6447
6448 bgp_evpn_unlink_from_vni_svi_hash(bgp, vpn);
6449
6450 vpn->svi_ifindex = 0;
6451 /*
6452 * tunnel is no longer active, del tunnel ip address from tip_hash
6453 */
6454 bgp_tip_del(bgp, &vpn->originator_ip);
6455
6456 /* Clear "live" flag and see if hash needs to be freed. */
6457 UNSET_FLAG(vpn->flags, VNI_FLAG_LIVE);
6458 if (!is_vni_configured(vpn))
6459 bgp_evpn_free(bgp, vpn);
6460
6461 return 0;
6462 }
6463
6464 /*
6465 * Handle add (or update) of a local VNI. The VNI changes we care
6466 * about are for the local-tunnel-ip and the (tenant) VRF.
6467 */
6468 int bgp_evpn_local_vni_add(struct bgp *bgp, vni_t vni,
6469 struct in_addr originator_ip,
6470 vrf_id_t tenant_vrf_id,
6471 struct in_addr mcast_grp,
6472 ifindex_t svi_ifindex)
6473 {
6474 struct bgpevpn *vpn;
6475 struct prefix_evpn p;
6476
6477 /* Lookup VNI. If present and no change, exit. */
6478 vpn = bgp_evpn_lookup_vni(bgp, vni);
6479 if (vpn) {
6480
6481 if (is_vni_live(vpn)
6482 && IPV4_ADDR_SAME(&vpn->originator_ip, &originator_ip)
6483 && IPV4_ADDR_SAME(&vpn->mcast_grp, &mcast_grp)
6484 && vpn->tenant_vrf_id == tenant_vrf_id
6485 && vpn->svi_ifindex == svi_ifindex)
6486 /* Probably some other param has changed that we don't
6487 * care about. */
6488 return 0;
6489
6490 bgp_evpn_mcast_grp_change(bgp, vpn, mcast_grp);
6491
6492 if (vpn->svi_ifindex != svi_ifindex) {
6493
6494 /*
6495 * Unresolve all the gateway IP nexthops for this VNI
6496 * for old SVI
6497 */
6498 bgp_evpn_remote_ip_hash_iterate(
6499 vpn,
6500 (void (*)(struct hash_bucket *, void *))
6501 bgp_evpn_remote_ip_hash_unlink_nexthop,
6502 vpn);
6503 bgp_evpn_unlink_from_vni_svi_hash(bgp, vpn);
6504 vpn->svi_ifindex = svi_ifindex;
6505 bgp_evpn_link_to_vni_svi_hash(bgp, vpn);
6506
6507 /*
6508 * Resolve all the gateway IP nexthops for this VNI
6509 * for new SVI
6510 */
6511 bgp_evpn_remote_ip_hash_iterate(
6512 vpn,
6513 (void (*)(struct hash_bucket *, void *))
6514 bgp_evpn_remote_ip_hash_link_nexthop,
6515 vpn);
6516 }
6517
6518 /* Update tenant_vrf_id if it has changed. */
6519 if (vpn->tenant_vrf_id != tenant_vrf_id) {
6520
6521 /*
6522 * Unresolve all the gateway IP nexthops for this VNI
6523 * in old tenant vrf
6524 */
6525 bgp_evpn_remote_ip_hash_iterate(
6526 vpn,
6527 (void (*)(struct hash_bucket *, void *))
6528 bgp_evpn_remote_ip_hash_unlink_nexthop,
6529 vpn);
6530 bgpevpn_unlink_from_l3vni(vpn);
6531 vpn->tenant_vrf_id = tenant_vrf_id;
6532 bgpevpn_link_to_l3vni(vpn);
6533
6534 /*
6535 * Resolve all the gateway IP nexthops for this VNI
6536 * in new tenant vrf
6537 */
6538 bgp_evpn_remote_ip_hash_iterate(
6539 vpn,
6540 (void (*)(struct hash_bucket *, void *))
6541 bgp_evpn_remote_ip_hash_link_nexthop,
6542 vpn);
6543 }
6544
6545 /* If tunnel endpoint IP has changed, update (and delete prior
6546 * type-3 route, if needed.)
6547 */
6548 if (!IPV4_ADDR_SAME(&vpn->originator_ip, &originator_ip))
6549 handle_tunnel_ip_change(bgp, vpn, originator_ip);
6550
6551 /* Update all routes with new endpoint IP and/or export RT
6552 * for VRFs
6553 */
6554 if (is_vni_live(vpn))
6555 update_routes_for_vni(bgp, vpn);
6556 } else {
6557 /* Create or update as appropriate. */
6558 vpn = bgp_evpn_new(bgp, vni, originator_ip, tenant_vrf_id,
6559 mcast_grp, svi_ifindex);
6560 }
6561
6562 /* if the VNI is live already, there is nothing more to do */
6563 if (is_vni_live(vpn))
6564 return 0;
6565
6566 /* Mark as "live" */
6567 SET_FLAG(vpn->flags, VNI_FLAG_LIVE);
6568
6569 /* tunnel is now active, add tunnel-ip to db */
6570 bgp_tip_add(bgp, &originator_ip);
6571
6572 /* filter routes as nexthop database has changed */
6573 bgp_filter_evpn_routes_upon_martian_nh_change(bgp);
6574
6575 /*
6576 * Create EVPN type-3 route and schedule for processing.
6577 *
6578 * RT-3 only if doing head-end replication
6579 */
6580 if (bgp_evpn_vni_flood_mode_get(bgp, vpn)
6581 == VXLAN_FLOOD_HEAD_END_REPL) {
6582 build_evpn_type3_prefix(&p, vpn->originator_ip);
6583 if (update_evpn_route(bgp, vpn, &p, 0, 0, NULL)) {
6584 flog_err(EC_BGP_EVPN_ROUTE_CREATE,
6585 "%u: Type3 route creation failure for VNI %u",
6586 bgp->vrf_id, vni);
6587 return -1;
6588 }
6589 }
6590
6591 /* If we have learnt and retained remote routes (VTEPs, MACs) for this
6592 * VNI,
6593 * install them.
6594 */
6595 install_routes_for_vni(bgp, vpn);
6596
6597 /* If we are advertising gateway mac-ip
6598 It needs to be conveyed again to zebra */
6599 bgp_zebra_advertise_gw_macip(bgp, vpn->advertise_gw_macip, vpn->vni);
6600
6601 /* advertise svi mac-ip knob to zebra */
6602 bgp_zebra_advertise_svi_macip(bgp, vpn->advertise_svi_macip, vpn->vni);
6603
6604 return 0;
6605 }
6606
6607 /*
6608 * Handle change in setting for BUM handling. The supported values
6609 * are head-end replication and dropping all BUM packets. Any change
6610 * should be registered with zebra. Also, if doing head-end replication,
6611 * need to advertise local VNIs as EVPN RT-3 wheras, if BUM packets are
6612 * to be dropped, the RT-3s must be withdrawn.
6613 */
6614 void bgp_evpn_flood_control_change(struct bgp *bgp)
6615 {
6616 zlog_info("L2VPN EVPN BUM handling is %s",
6617 bgp->vxlan_flood_ctrl == VXLAN_FLOOD_HEAD_END_REPL ?
6618 "Flooding" : "Flooding Disabled");
6619
6620 bgp_zebra_vxlan_flood_control(bgp, bgp->vxlan_flood_ctrl);
6621 if (bgp->vxlan_flood_ctrl == VXLAN_FLOOD_HEAD_END_REPL)
6622 hash_iterate(bgp->vnihash, create_advertise_type3, bgp);
6623 else if (bgp->vxlan_flood_ctrl == VXLAN_FLOOD_DISABLED)
6624 hash_iterate(bgp->vnihash, delete_withdraw_type3, bgp);
6625 }
6626
6627 /*
6628 * Cleanup EVPN information on disable - Need to delete and withdraw
6629 * EVPN routes from peers.
6630 */
6631 void bgp_evpn_cleanup_on_disable(struct bgp *bgp)
6632 {
6633 hash_iterate(bgp->vnihash, (void (*)(struct hash_bucket *,
6634 void *))cleanup_vni_on_disable,
6635 bgp);
6636 }
6637
6638 /*
6639 * Cleanup EVPN information - invoked at the time of bgpd exit or when the
6640 * BGP instance (default) is being freed.
6641 */
6642 void bgp_evpn_cleanup(struct bgp *bgp)
6643 {
6644 hash_iterate(bgp->vnihash,
6645 (void (*)(struct hash_bucket *, void *))free_vni_entry,
6646 bgp);
6647
6648 hash_clean(bgp->import_rt_hash, (void (*)(void *))hash_import_rt_free);
6649 hash_free(bgp->import_rt_hash);
6650 bgp->import_rt_hash = NULL;
6651
6652 hash_clean(bgp->vrf_import_rt_hash,
6653 (void (*)(void *))hash_vrf_import_rt_free);
6654 hash_free(bgp->vrf_import_rt_hash);
6655 bgp->vrf_import_rt_hash = NULL;
6656
6657 hash_clean(bgp->vni_svi_hash, (void (*)(void *))hash_evpn_free);
6658 hash_free(bgp->vni_svi_hash);
6659 bgp->vni_svi_hash = NULL;
6660 hash_free(bgp->vnihash);
6661 bgp->vnihash = NULL;
6662
6663 list_delete(&bgp->vrf_import_rtl);
6664 list_delete(&bgp->vrf_export_rtl);
6665 list_delete(&bgp->l2vnis);
6666 }
6667
6668 /*
6669 * Initialization for EVPN
6670 * Create
6671 * VNI hash table
6672 * hash for RT to VNI
6673 */
6674 void bgp_evpn_init(struct bgp *bgp)
6675 {
6676 bgp->vnihash =
6677 hash_create(vni_hash_key_make, vni_hash_cmp, "BGP VNI Hash");
6678 bgp->vni_svi_hash =
6679 hash_create(vni_svi_hash_key_make, vni_svi_hash_cmp,
6680 "BGP VNI hash based on SVI ifindex");
6681 bgp->import_rt_hash =
6682 hash_create(import_rt_hash_key_make, import_rt_hash_cmp,
6683 "BGP Import RT Hash");
6684 bgp->vrf_import_rt_hash =
6685 hash_create(vrf_import_rt_hash_key_make, vrf_import_rt_hash_cmp,
6686 "BGP VRF Import RT Hash");
6687 bgp->vrf_import_rtl = list_new();
6688 bgp->vrf_import_rtl->cmp =
6689 (int (*)(void *, void *))evpn_vrf_route_target_cmp;
6690 bgp->vrf_import_rtl->del = evpn_vrf_rt_del;
6691 bgp->vrf_export_rtl = list_new();
6692 bgp->vrf_export_rtl->cmp =
6693 (int (*)(void *, void *))evpn_vrf_route_target_cmp;
6694 bgp->vrf_export_rtl->del = evpn_vrf_rt_del;
6695 bgp->l2vnis = list_new();
6696 bgp->l2vnis->cmp = vni_list_cmp;
6697 /* By default Duplicate Address Dection is enabled.
6698 * Max-moves (N) 5, detection time (M) 180
6699 * default action is warning-only
6700 * freeze action permanently freezes address,
6701 * and freeze time (auto-recovery) is disabled.
6702 */
6703 if (bgp->evpn_info) {
6704 bgp->evpn_info->dup_addr_detect = true;
6705 bgp->evpn_info->dad_time = EVPN_DAD_DEFAULT_TIME;
6706 bgp->evpn_info->dad_max_moves = EVPN_DAD_DEFAULT_MAX_MOVES;
6707 bgp->evpn_info->dad_freeze = false;
6708 bgp->evpn_info->dad_freeze_time = 0;
6709 /* Initialize zebra vxlan */
6710 bgp_zebra_dup_addr_detection(bgp);
6711 /* Enable PIP feature by default for bgp vrf instance */
6712 if (bgp->inst_type == BGP_INSTANCE_TYPE_VRF) {
6713 struct bgp *bgp_default;
6714
6715 bgp->evpn_info->advertise_pip = true;
6716 bgp_default = bgp_get_default();
6717 if (bgp_default)
6718 bgp->evpn_info->pip_ip = bgp_default->router_id;
6719 }
6720 }
6721
6722 /* Default BUM handling is to do head-end replication. */
6723 bgp->vxlan_flood_ctrl = VXLAN_FLOOD_HEAD_END_REPL;
6724
6725 bgp_evpn_nh_init(bgp);
6726 }
6727
6728 void bgp_evpn_vrf_delete(struct bgp *bgp_vrf)
6729 {
6730 bgp_evpn_unmap_vrf_from_its_rts(bgp_vrf);
6731 bgp_evpn_nh_finish(bgp_vrf);
6732 }
6733
6734 /*
6735 * Get the prefixlen of the ip prefix carried within the type5 evpn route.
6736 */
6737 int bgp_evpn_get_type5_prefixlen(const struct prefix *pfx)
6738 {
6739 struct prefix_evpn *evp = (struct prefix_evpn *)pfx;
6740
6741 if (!pfx || pfx->family != AF_EVPN)
6742 return 0;
6743
6744 if (evp->prefix.route_type != BGP_EVPN_IP_PREFIX_ROUTE)
6745 return 0;
6746
6747 return evp->prefix.prefix_addr.ip_prefix_length;
6748 }
6749
6750 /*
6751 * Should we register nexthop for this EVPN prefix for nexthop tracking?
6752 */
6753 bool bgp_evpn_is_prefix_nht_supported(const struct prefix *pfx)
6754 {
6755 struct prefix_evpn *evp = (struct prefix_evpn *)pfx;
6756
6757 /*
6758 * EVPN routes should be marked as valid only if the nexthop is
6759 * reachable. Only if this happens, the route should be imported
6760 * (into VNI or VRF routing tables) and/or advertised.
6761 * Note: This is currently applied for EVPN type-1, type-2,
6762 * type-3, type-4 and type-5 routes.
6763 * It may be tweaked later on for other routes, or
6764 * even removed completely when all routes are handled.
6765 */
6766 if (pfx && pfx->family == AF_EVPN
6767 && (evp->prefix.route_type == BGP_EVPN_MAC_IP_ROUTE
6768 || evp->prefix.route_type == BGP_EVPN_AD_ROUTE
6769 || evp->prefix.route_type == BGP_EVPN_ES_ROUTE
6770 || evp->prefix.route_type == BGP_EVPN_IMET_ROUTE
6771 || evp->prefix.route_type == BGP_EVPN_IP_PREFIX_ROUTE))
6772 return true;
6773
6774 return false;
6775 }
6776
6777 static void *bgp_evpn_remote_ip_hash_alloc(void *p)
6778 {
6779 const struct evpn_remote_ip *key = (const struct evpn_remote_ip *)p;
6780 struct evpn_remote_ip *ip;
6781
6782 ip = XMALLOC(MTYPE_EVPN_REMOTE_IP, sizeof(struct evpn_remote_ip));
6783 *ip = *key;
6784 ip->macip_path_list = list_new();
6785
6786 return ip;
6787 }
6788
6789 static unsigned int bgp_evpn_remote_ip_hash_key_make(const void *p)
6790 {
6791 const struct evpn_remote_ip *ip = p;
6792 const struct ipaddr *addr = &ip->addr;
6793
6794 if (IS_IPADDR_V4(addr))
6795 return jhash_1word(addr->ipaddr_v4.s_addr, 0);
6796
6797 return jhash2(addr->ipaddr_v6.s6_addr32,
6798 array_size(addr->ipaddr_v6.s6_addr32), 0);
6799 }
6800
6801 static bool bgp_evpn_remote_ip_hash_cmp(const void *p1, const void *p2)
6802 {
6803 const struct evpn_remote_ip *ip1 = p1;
6804 const struct evpn_remote_ip *ip2 = p2;
6805
6806 return !ipaddr_cmp(&ip1->addr, &ip2->addr);
6807 }
6808
6809 static void bgp_evpn_remote_ip_hash_init(struct bgpevpn *vpn)
6810 {
6811 if (!evpn_resolve_overlay_index())
6812 return;
6813
6814 vpn->remote_ip_hash = hash_create(bgp_evpn_remote_ip_hash_key_make,
6815 bgp_evpn_remote_ip_hash_cmp,
6816 "BGP EVPN remote IP hash");
6817 }
6818
6819 static void bgp_evpn_remote_ip_hash_free(struct hash_bucket *bucket, void *args)
6820 {
6821 struct evpn_remote_ip *ip = (struct evpn_remote_ip *)bucket->data;
6822 struct bgpevpn *vpn = (struct bgpevpn *)args;
6823
6824 bgp_evpn_remote_ip_process_nexthops(vpn, &ip->addr, false);
6825
6826 list_delete(&ip->macip_path_list);
6827
6828 hash_release(vpn->remote_ip_hash, ip);
6829 XFREE(MTYPE_EVPN_REMOTE_IP, ip);
6830 }
6831
6832 static void bgp_evpn_remote_ip_hash_destroy(struct bgpevpn *vpn)
6833 {
6834 if (!evpn_resolve_overlay_index() || vpn->remote_ip_hash == NULL)
6835 return;
6836
6837 hash_iterate(vpn->remote_ip_hash,
6838 (void (*)(struct hash_bucket *, void *))bgp_evpn_remote_ip_hash_free,
6839 vpn);
6840
6841 hash_free(vpn->remote_ip_hash);
6842 vpn->remote_ip_hash = NULL;
6843 }
6844
6845 /* Add a remote MAC/IP route to hash table */
6846 static void bgp_evpn_remote_ip_hash_add(struct bgpevpn *vpn,
6847 struct bgp_path_info *pi)
6848 {
6849 struct evpn_remote_ip tmp;
6850 struct evpn_remote_ip *ip;
6851 struct prefix_evpn *evp;
6852
6853 if (!evpn_resolve_overlay_index())
6854 return;
6855
6856 if (pi->type != ZEBRA_ROUTE_BGP || pi->sub_type != BGP_ROUTE_IMPORTED
6857 || !CHECK_FLAG(pi->flags, BGP_PATH_VALID))
6858 return;
6859
6860 evp = (struct prefix_evpn *)&pi->net->p;
6861
6862 if (evp->family != AF_EVPN
6863 || evp->prefix.route_type != BGP_EVPN_MAC_IP_ROUTE
6864 || is_evpn_prefix_ipaddr_none(evp))
6865 return;
6866
6867 tmp.addr = evp->prefix.macip_addr.ip;
6868 ip = hash_lookup(vpn->remote_ip_hash, &tmp);
6869 if (ip) {
6870 if (listnode_lookup(ip->macip_path_list, pi) != NULL)
6871 return;
6872 (void)listnode_add(ip->macip_path_list, pi);
6873 return;
6874 }
6875
6876 ip = hash_get(vpn->remote_ip_hash, &tmp, bgp_evpn_remote_ip_hash_alloc);
6877 (void)listnode_add(ip->macip_path_list, pi);
6878
6879 bgp_evpn_remote_ip_process_nexthops(vpn, &ip->addr, true);
6880 }
6881
6882 /* Delete a remote MAC/IP route from hash table */
6883 static void bgp_evpn_remote_ip_hash_del(struct bgpevpn *vpn,
6884 struct bgp_path_info *pi)
6885 {
6886 struct evpn_remote_ip tmp;
6887 struct evpn_remote_ip *ip;
6888 struct prefix_evpn *evp;
6889
6890 if (!evpn_resolve_overlay_index())
6891 return;
6892
6893 evp = (struct prefix_evpn *)&pi->net->p;
6894
6895 if (evp->family != AF_EVPN
6896 || evp->prefix.route_type != BGP_EVPN_MAC_IP_ROUTE
6897 || is_evpn_prefix_ipaddr_none(evp))
6898 return;
6899
6900 tmp.addr = evp->prefix.macip_addr.ip;
6901 ip = hash_lookup(vpn->remote_ip_hash, &tmp);
6902 if (ip == NULL)
6903 return;
6904
6905 listnode_delete(ip->macip_path_list, pi);
6906
6907 if (ip->macip_path_list->count == 0) {
6908 bgp_evpn_remote_ip_process_nexthops(vpn, &ip->addr, false);
6909 hash_release(vpn->remote_ip_hash, ip);
6910 list_delete(&ip->macip_path_list);
6911 XFREE(MTYPE_EVPN_REMOTE_IP, ip);
6912 }
6913 }
6914
6915 static void bgp_evpn_remote_ip_hash_iterate(struct bgpevpn *vpn,
6916 void (*func)(struct hash_bucket *,
6917 void *),
6918 void *arg)
6919 {
6920 if (!evpn_resolve_overlay_index())
6921 return;
6922
6923 hash_iterate(vpn->remote_ip_hash, func, arg);
6924 }
6925
6926 static void show_remote_ip_entry(struct hash_bucket *bucket, void *args)
6927 {
6928 char buf[INET6_ADDRSTRLEN];
6929 struct listnode *node = NULL;
6930 struct bgp_path_info *pi = NULL;
6931 struct vty *vty = (struct vty *)args;
6932 struct evpn_remote_ip *ip = (struct evpn_remote_ip *)bucket->data;
6933
6934 vty_out(vty, " Remote IP: %s\n",
6935 ipaddr2str(&ip->addr, buf, sizeof(buf)));
6936 vty_out(vty, " Linked MAC/IP routes:\n");
6937 for (ALL_LIST_ELEMENTS_RO(ip->macip_path_list, node, pi))
6938 vty_out(vty, " %pFX\n", &pi->net->p);
6939 }
6940
6941 void bgp_evpn_show_remote_ip_hash(struct hash_bucket *bucket, void *args)
6942 {
6943 struct bgpevpn *vpn = (struct bgpevpn *)bucket->data;
6944 struct vty *vty = (struct vty *)args;
6945
6946 vty_out(vty, "VNI: %u\n", vpn->vni);
6947 bgp_evpn_remote_ip_hash_iterate(
6948 vpn,
6949 (void (*)(struct hash_bucket *, void *))show_remote_ip_entry,
6950 vty);
6951 vty_out(vty, "\n");
6952 }
6953
6954 static void bgp_evpn_remote_ip_hash_link_nexthop(struct hash_bucket *bucket,
6955 void *args)
6956 {
6957 struct evpn_remote_ip *ip = (struct evpn_remote_ip *)bucket->data;
6958 struct bgpevpn *vpn = (struct bgpevpn *)args;
6959
6960 bgp_evpn_remote_ip_process_nexthops(vpn, &ip->addr, true);
6961 }
6962
6963 static void bgp_evpn_remote_ip_hash_unlink_nexthop(struct hash_bucket *bucket,
6964 void *args)
6965 {
6966 struct evpn_remote_ip *ip = (struct evpn_remote_ip *)bucket->data;
6967 struct bgpevpn *vpn = (struct bgpevpn *)args;
6968
6969 bgp_evpn_remote_ip_process_nexthops(vpn, &ip->addr, false);
6970 }
6971
6972 static unsigned int vni_svi_hash_key_make(const void *p)
6973 {
6974 const struct bgpevpn *vpn = p;
6975
6976 return jhash_1word(vpn->svi_ifindex, 0);
6977 }
6978
6979 static bool vni_svi_hash_cmp(const void *p1, const void *p2)
6980 {
6981 const struct bgpevpn *vpn1 = p1;
6982 const struct bgpevpn *vpn2 = p2;
6983
6984 return (vpn1->svi_ifindex == vpn2->svi_ifindex);
6985 }
6986
6987 static struct bgpevpn *bgp_evpn_vni_svi_hash_lookup(struct bgp *bgp,
6988 ifindex_t svi)
6989 {
6990 struct bgpevpn *vpn;
6991 struct bgpevpn tmp;
6992
6993 memset(&tmp, 0, sizeof(tmp));
6994 tmp.svi_ifindex = svi;
6995 vpn = hash_lookup(bgp->vni_svi_hash, &tmp);
6996 return vpn;
6997 }
6998
6999 static void bgp_evpn_link_to_vni_svi_hash(struct bgp *bgp, struct bgpevpn *vpn)
7000 {
7001 if (vpn->svi_ifindex == 0)
7002 return;
7003
7004 (void)hash_get(bgp->vni_svi_hash, vpn, hash_alloc_intern);
7005 }
7006
7007 static void bgp_evpn_unlink_from_vni_svi_hash(struct bgp *bgp,
7008 struct bgpevpn *vpn)
7009 {
7010 if (vpn->svi_ifindex == 0)
7011 return;
7012
7013 hash_release(bgp->vni_svi_hash, vpn);
7014 }
7015
7016 void bgp_evpn_show_vni_svi_hash(struct hash_bucket *bucket, void *args)
7017 {
7018 struct bgpevpn *evpn = (struct bgpevpn *)bucket->data;
7019 struct vty *vty = (struct vty *)args;
7020
7021 vty_out(vty, "SVI: %u VNI: %u\n", evpn->svi_ifindex, evpn->vni);
7022 }
7023
7024 /*
7025 * This function is called for a bgp_nexthop_cache entry when the nexthop is
7026 * gateway IP overlay index.
7027 * This function returns true if there is a remote MAC/IP route for the gateway
7028 * IP in the EVI of the nexthop SVI.
7029 */
7030 bool bgp_evpn_is_gateway_ip_resolved(struct bgp_nexthop_cache *bnc)
7031 {
7032 struct bgp *bgp_evpn = NULL;
7033 struct bgpevpn *vpn = NULL;
7034 struct evpn_remote_ip tmp;
7035 struct prefix *p;
7036
7037 if (!evpn_resolve_overlay_index())
7038 return false;
7039
7040 if (!bnc->nexthop || bnc->nexthop->ifindex == 0)
7041 return false;
7042
7043 bgp_evpn = bgp_get_evpn();
7044 if (!bgp_evpn)
7045 return false;
7046
7047 /*
7048 * Gateway IP is resolved by nht over SVI interface.
7049 * Use this SVI to find corresponding EVI(L2 context)
7050 */
7051 vpn = bgp_evpn_vni_svi_hash_lookup(bgp_evpn, bnc->nexthop->ifindex);
7052 if (!vpn)
7053 return false;
7054
7055 if (vpn->bgp_vrf != bnc->bgp)
7056 return false;
7057
7058 /*
7059 * Check if the gateway IP is present in the EVI remote_ip_hash table
7060 * which stores all the remote IP addresses received via MAC/IP routes
7061 * in this EVI
7062 */
7063 memset(&tmp, 0, sizeof(tmp));
7064
7065 p = &bnc->prefix;
7066 if (p->family == AF_INET) {
7067 tmp.addr.ipa_type = IPADDR_V4;
7068 memcpy(&(tmp.addr.ipaddr_v4), &(p->u.prefix4),
7069 sizeof(struct in_addr));
7070 } else if (p->family == AF_INET6) {
7071 tmp.addr.ipa_type = IPADDR_V6;
7072 memcpy(&(tmp.addr.ipaddr_v6), &(p->u.prefix6),
7073 sizeof(struct in6_addr));
7074 } else
7075 return false;
7076
7077 if (hash_lookup(vpn->remote_ip_hash, &tmp) == NULL)
7078 return false;
7079
7080 return true;
7081 }
7082
7083 /* Resolve/Unresolve nexthops when a MAC/IP route is added/deleted */
7084 static void bgp_evpn_remote_ip_process_nexthops(struct bgpevpn *vpn,
7085 struct ipaddr *addr,
7086 bool resolve)
7087 {
7088 afi_t afi;
7089 struct prefix p;
7090 struct bgp_nexthop_cache *bnc;
7091 struct bgp_nexthop_cache_head *tree = NULL;
7092
7093 if (!vpn->bgp_vrf || vpn->svi_ifindex == 0)
7094 return;
7095
7096 memset(&p, 0, sizeof(p));
7097
7098 if (addr->ipa_type == IPADDR_V4) {
7099 afi = AFI_IP;
7100 p.family = AF_INET;
7101 memcpy(&(p.u.prefix4), &(addr->ipaddr_v4),
7102 sizeof(struct in_addr));
7103 p.prefixlen = IPV4_MAX_BITLEN;
7104 } else if (addr->ipa_type == IPADDR_V6) {
7105 afi = AFI_IP6;
7106 p.family = AF_INET6;
7107 memcpy(&(p.u.prefix6), &(addr->ipaddr_v6),
7108 sizeof(struct in6_addr));
7109 p.prefixlen = IPV6_MAX_BITLEN;
7110 } else
7111 return;
7112
7113 tree = &vpn->bgp_vrf->nexthop_cache_table[afi];
7114 bnc = bnc_find(tree, &p, 0, 0);
7115
7116 if (!bnc || !bnc->is_evpn_gwip_nexthop)
7117 return;
7118
7119 if (!bnc->nexthop || bnc->nexthop->ifindex != vpn->svi_ifindex)
7120 return;
7121
7122 if (BGP_DEBUG(nht, NHT))
7123 zlog_debug("%s(%u): vni %u mac/ip %s for NH %pFX",
7124 vpn->bgp_vrf->name_pretty, vpn->tenant_vrf_id,
7125 vpn->vni, (resolve ? "add" : "delete"),
7126 &bnc->prefix);
7127
7128 /*
7129 * MAC/IP route or SVI or tenant vrf being added to EVI.
7130 * Set nexthop as valid only if it is already L3 reachable
7131 */
7132 if (resolve && bnc->flags & BGP_NEXTHOP_EVPN_INCOMPLETE) {
7133 bnc->flags &= ~BGP_NEXTHOP_EVPN_INCOMPLETE;
7134 bnc->flags |= BGP_NEXTHOP_VALID;
7135 bnc->change_flags |= BGP_NEXTHOP_MACIP_CHANGED;
7136 evaluate_paths(bnc);
7137 }
7138
7139 /* MAC/IP route or SVI or tenant vrf being deleted from EVI */
7140 if (!resolve && bnc->flags & BGP_NEXTHOP_VALID) {
7141 bnc->flags &= ~BGP_NEXTHOP_VALID;
7142 bnc->flags |= BGP_NEXTHOP_EVPN_INCOMPLETE;
7143 bnc->change_flags |= BGP_NEXTHOP_MACIP_CHANGED;
7144 evaluate_paths(bnc);
7145 }
7146 }
7147
7148 void bgp_evpn_handle_resolve_overlay_index_set(struct hash_bucket *bucket,
7149 void *arg)
7150 {
7151 struct bgpevpn *vpn = (struct bgpevpn *)bucket->data;
7152 struct bgp_dest *dest;
7153 struct bgp_path_info *pi;
7154
7155 bgp_evpn_remote_ip_hash_init(vpn);
7156
7157 for (dest = bgp_table_top(vpn->ip_table); dest;
7158 dest = bgp_route_next(dest))
7159 for (pi = bgp_dest_get_bgp_path_info(dest); pi; pi = pi->next)
7160 bgp_evpn_remote_ip_hash_add(vpn, pi);
7161 }
7162
7163 void bgp_evpn_handle_resolve_overlay_index_unset(struct hash_bucket *bucket,
7164 void *arg)
7165 {
7166 struct bgpevpn *vpn = (struct bgpevpn *)bucket->data;
7167
7168 bgp_evpn_remote_ip_hash_destroy(vpn);
7169 }