]> git.proxmox.com Git - mirror_frr.git/blob - bgpd/bgp_evpn.c
ldpd: changes for code maintainability
[mirror_frr.git] / bgpd / bgp_evpn.c
1 // SPDX-License-Identifier: GPL-2.0-or-later
2 /* Ethernet-VPN Packet and vty Processing File
3 * Copyright (C) 2016 6WIND
4 * Copyright (C) 2017 Cumulus Networks, Inc.
5 */
6
7 #include <zebra.h>
8
9 #include "command.h"
10 #include "filter.h"
11 #include "prefix.h"
12 #include "log.h"
13 #include "memory.h"
14 #include "stream.h"
15 #include "hash.h"
16 #include "jhash.h"
17 #include "zclient.h"
18
19 #include "lib/printfrr.h"
20
21 #include "bgpd/bgp_attr_evpn.h"
22 #include "bgpd/bgpd.h"
23 #include "bgpd/bgp_table.h"
24 #include "bgpd/bgp_route.h"
25 #include "bgpd/bgp_attr.h"
26 #include "bgpd/bgp_mplsvpn.h"
27 #include "bgpd/bgp_label.h"
28 #include "bgpd/bgp_evpn.h"
29 #include "bgpd/bgp_evpn_private.h"
30 #include "bgpd/bgp_evpn_mh.h"
31 #include "bgpd/bgp_ecommunity.h"
32 #include "bgpd/bgp_encap_types.h"
33 #include "bgpd/bgp_debug.h"
34 #include "bgpd/bgp_errors.h"
35 #include "bgpd/bgp_aspath.h"
36 #include "bgpd/bgp_zebra.h"
37 #include "bgpd/bgp_nexthop.h"
38 #include "bgpd/bgp_addpath.h"
39 #include "bgpd/bgp_mac.h"
40 #include "bgpd/bgp_vty.h"
41 #include "bgpd/bgp_nht.h"
42 #include "bgpd/bgp_trace.h"
43 #include "bgpd/bgp_mpath.h"
44
45 /*
46 * Definitions and external declarations.
47 */
48 DEFINE_QOBJ_TYPE(bgpevpn);
49 DEFINE_QOBJ_TYPE(bgp_evpn_es);
50
51 DEFINE_MTYPE_STATIC(BGPD, VRF_ROUTE_TARGET, "L3 Route Target");
52
53 /*
54 * Static function declarations
55 */
56 static void bgp_evpn_remote_ip_hash_init(struct bgpevpn *evpn);
57 static void bgp_evpn_remote_ip_hash_destroy(struct bgpevpn *evpn);
58 static void bgp_evpn_remote_ip_hash_add(struct bgpevpn *vpn,
59 struct bgp_path_info *pi);
60 static void bgp_evpn_remote_ip_hash_del(struct bgpevpn *vpn,
61 struct bgp_path_info *pi);
62 static void bgp_evpn_remote_ip_hash_iterate(struct bgpevpn *vpn,
63 void (*func)(struct hash_bucket *,
64 void *),
65 void *arg);
66 static void bgp_evpn_link_to_vni_svi_hash(struct bgp *bgp, struct bgpevpn *vpn);
67 static void bgp_evpn_unlink_from_vni_svi_hash(struct bgp *bgp,
68 struct bgpevpn *vpn);
69 static unsigned int vni_svi_hash_key_make(const void *p);
70 static bool vni_svi_hash_cmp(const void *p1, const void *p2);
71 static void bgp_evpn_remote_ip_process_nexthops(struct bgpevpn *vpn,
72 struct ipaddr *addr,
73 bool resolve);
74 static void bgp_evpn_remote_ip_hash_link_nexthop(struct hash_bucket *bucket,
75 void *args);
76 static void bgp_evpn_remote_ip_hash_unlink_nexthop(struct hash_bucket *bucket,
77 void *args);
78 static struct in_addr zero_vtep_ip;
79
80 /*
81 * Private functions.
82 */
83
84 /*
85 * Make vni hash key.
86 */
87 static unsigned int vni_hash_key_make(const void *p)
88 {
89 const struct bgpevpn *vpn = p;
90 return (jhash_1word(vpn->vni, 0));
91 }
92
93 /*
94 * Comparison function for vni hash
95 */
96 static bool vni_hash_cmp(const void *p1, const void *p2)
97 {
98 const struct bgpevpn *vpn1 = p1;
99 const struct bgpevpn *vpn2 = p2;
100
101 return vpn1->vni == vpn2->vni;
102 }
103
104 int vni_list_cmp(void *p1, void *p2)
105 {
106 const struct bgpevpn *vpn1 = p1;
107 const struct bgpevpn *vpn2 = p2;
108
109 return vpn1->vni - vpn2->vni;
110 }
111
112 /*
113 * Make vrf import route target hash key.
114 */
115 static unsigned int vrf_import_rt_hash_key_make(const void *p)
116 {
117 const struct vrf_irt_node *irt = p;
118 const char *pnt = irt->rt.val;
119
120 return jhash(pnt, 8, 0x5abc1234);
121 }
122
123 /*
124 * Comparison function for vrf import rt hash
125 */
126 static bool vrf_import_rt_hash_cmp(const void *p1, const void *p2)
127 {
128 const struct vrf_irt_node *irt1 = p1;
129 const struct vrf_irt_node *irt2 = p2;
130
131 return (memcmp(irt1->rt.val, irt2->rt.val, ECOMMUNITY_SIZE) == 0);
132 }
133
134 /*
135 * Create a new vrf import_rt in evpn instance
136 */
137 static struct vrf_irt_node *vrf_import_rt_new(struct ecommunity_val *rt)
138 {
139 struct bgp *bgp_evpn = NULL;
140 struct vrf_irt_node *irt;
141
142 bgp_evpn = bgp_get_evpn();
143 if (!bgp_evpn) {
144 flog_err(EC_BGP_NO_DFLT,
145 "vrf import rt new - evpn instance not created yet");
146 return NULL;
147 }
148
149 irt = XCALLOC(MTYPE_BGP_EVPN_VRF_IMPORT_RT,
150 sizeof(struct vrf_irt_node));
151
152 irt->rt = *rt;
153 irt->vrfs = list_new();
154
155 /* Add to hash */
156 (void)hash_get(bgp_evpn->vrf_import_rt_hash, irt, hash_alloc_intern);
157
158 return irt;
159 }
160
161 /*
162 * Free the vrf import rt node
163 */
164 static void vrf_import_rt_free(struct vrf_irt_node *irt)
165 {
166 struct bgp *bgp_evpn = NULL;
167
168 bgp_evpn = bgp_get_evpn();
169 if (!bgp_evpn) {
170 flog_err(EC_BGP_NO_DFLT,
171 "vrf import rt free - evpn instance not created yet");
172 return;
173 }
174
175 hash_release(bgp_evpn->vrf_import_rt_hash, irt);
176 list_delete(&irt->vrfs);
177 XFREE(MTYPE_BGP_EVPN_VRF_IMPORT_RT, irt);
178 }
179
180 static void hash_vrf_import_rt_free(struct vrf_irt_node *irt)
181 {
182 XFREE(MTYPE_BGP_EVPN_VRF_IMPORT_RT, irt);
183 }
184
185 /*
186 * Function to lookup Import RT node - used to map a RT to set of
187 * VNIs importing routes with that RT.
188 */
189 static struct vrf_irt_node *lookup_vrf_import_rt(struct ecommunity_val *rt)
190 {
191 struct bgp *bgp_evpn = NULL;
192 struct vrf_irt_node *irt;
193 struct vrf_irt_node tmp;
194
195 bgp_evpn = bgp_get_evpn();
196 if (!bgp_evpn) {
197 flog_err(
198 EC_BGP_NO_DFLT,
199 "vrf import rt lookup - evpn instance not created yet");
200 return NULL;
201 }
202
203 memset(&tmp, 0, sizeof(tmp));
204 memcpy(&tmp.rt, rt, ECOMMUNITY_SIZE);
205 irt = hash_lookup(bgp_evpn->vrf_import_rt_hash, &tmp);
206 return irt;
207 }
208
209 /*
210 * Is specified VRF present on the RT's list of "importing" VRFs?
211 */
212 static int is_vrf_present_in_irt_vrfs(struct list *vrfs, struct bgp *bgp_vrf)
213 {
214 struct listnode *node = NULL, *nnode = NULL;
215 struct bgp *tmp_bgp_vrf = NULL;
216
217 for (ALL_LIST_ELEMENTS(vrfs, node, nnode, tmp_bgp_vrf)) {
218 if (tmp_bgp_vrf == bgp_vrf)
219 return 1;
220 }
221 return 0;
222 }
223
224 /*
225 * Make import route target hash key.
226 */
227 static unsigned int import_rt_hash_key_make(const void *p)
228 {
229 const struct irt_node *irt = p;
230 const char *pnt = irt->rt.val;
231
232 return jhash(pnt, 8, 0xdeadbeef);
233 }
234
235 /*
236 * Comparison function for import rt hash
237 */
238 static bool import_rt_hash_cmp(const void *p1, const void *p2)
239 {
240 const struct irt_node *irt1 = p1;
241 const struct irt_node *irt2 = p2;
242
243 return (memcmp(irt1->rt.val, irt2->rt.val, ECOMMUNITY_SIZE) == 0);
244 }
245
246 /*
247 * Create a new import_rt
248 */
249 static struct irt_node *import_rt_new(struct bgp *bgp,
250 struct ecommunity_val *rt)
251 {
252 struct irt_node *irt;
253
254 irt = XCALLOC(MTYPE_BGP_EVPN_IMPORT_RT, sizeof(struct irt_node));
255
256 irt->rt = *rt;
257 irt->vnis = list_new();
258
259 /* Add to hash */
260 (void)hash_get(bgp->import_rt_hash, irt, hash_alloc_intern);
261
262 return irt;
263 }
264
265 /*
266 * Free the import rt node
267 */
268 static void import_rt_free(struct bgp *bgp, struct irt_node *irt)
269 {
270 hash_release(bgp->import_rt_hash, irt);
271 list_delete(&irt->vnis);
272 XFREE(MTYPE_BGP_EVPN_IMPORT_RT, irt);
273 }
274
275 static void hash_import_rt_free(struct irt_node *irt)
276 {
277 XFREE(MTYPE_BGP_EVPN_IMPORT_RT, irt);
278 }
279
280 /*
281 * Function to lookup Import RT node - used to map a RT to set of
282 * VNIs importing routes with that RT.
283 */
284 static struct irt_node *lookup_import_rt(struct bgp *bgp,
285 struct ecommunity_val *rt)
286 {
287 struct irt_node *irt;
288 struct irt_node tmp;
289
290 memset(&tmp, 0, sizeof(tmp));
291 memcpy(&tmp.rt, rt, ECOMMUNITY_SIZE);
292 irt = hash_lookup(bgp->import_rt_hash, &tmp);
293 return irt;
294 }
295
296 /*
297 * Is specified VNI present on the RT's list of "importing" VNIs?
298 */
299 static int is_vni_present_in_irt_vnis(struct list *vnis, struct bgpevpn *vpn)
300 {
301 struct listnode *node, *nnode;
302 struct bgpevpn *tmp_vpn;
303
304 for (ALL_LIST_ELEMENTS(vnis, node, nnode, tmp_vpn)) {
305 if (tmp_vpn == vpn)
306 return 1;
307 }
308
309 return 0;
310 }
311
312 /*
313 * Compare Route Targets.
314 */
315 int bgp_evpn_route_target_cmp(struct ecommunity *ecom1,
316 struct ecommunity *ecom2)
317 {
318 if (ecom1 && !ecom2)
319 return -1;
320
321 if (!ecom1 && ecom2)
322 return 1;
323
324 if (!ecom1 && !ecom2)
325 return 0;
326
327 if (ecom1->str && !ecom2->str)
328 return -1;
329
330 if (!ecom1->str && ecom2->str)
331 return 1;
332
333 if (!ecom1->str && !ecom2->str)
334 return 0;
335
336 return strcmp(ecom1->str, ecom2->str);
337 }
338
339 /*
340 * Compare L3 Route Targets.
341 */
342 static int evpn_vrf_route_target_cmp(struct vrf_route_target *rt1,
343 struct vrf_route_target *rt2)
344 {
345 return bgp_evpn_route_target_cmp(rt1->ecom, rt2->ecom);
346 }
347
348 void bgp_evpn_xxport_delete_ecomm(void *val)
349 {
350 struct ecommunity *ecomm = val;
351 ecommunity_free(&ecomm);
352 }
353
354 /*
355 * Delete l3 Route Target.
356 */
357 static void evpn_vrf_rt_del(void *val)
358 {
359 struct vrf_route_target *l3rt = val;
360
361 ecommunity_free(&l3rt->ecom);
362
363 XFREE(MTYPE_VRF_ROUTE_TARGET, l3rt);
364 }
365
366 /*
367 * Allocate a new l3 Route Target.
368 */
369 static struct vrf_route_target *evpn_vrf_rt_new(struct ecommunity *ecom)
370 {
371 struct vrf_route_target *l3rt;
372
373 l3rt = XCALLOC(MTYPE_VRF_ROUTE_TARGET, sizeof(struct vrf_route_target));
374
375 l3rt->ecom = ecom;
376
377 return l3rt;
378 }
379
380 /*
381 * Mask off global-admin field of specified extended community (RT),
382 * just retain the local-admin field.
383 */
384 static inline void mask_ecom_global_admin(struct ecommunity_val *dst,
385 const struct ecommunity_val *src)
386 {
387 uint8_t type;
388
389 type = src->val[0];
390 dst->val[0] = 0;
391 if (type == ECOMMUNITY_ENCODE_AS) {
392 dst->val[2] = dst->val[3] = 0;
393 } else if (type == ECOMMUNITY_ENCODE_AS4
394 || type == ECOMMUNITY_ENCODE_IP) {
395 dst->val[2] = dst->val[3] = 0;
396 dst->val[4] = dst->val[5] = 0;
397 }
398 }
399
400 /*
401 * Converts the RT to Ecommunity Value and adjusts masking based
402 * on flags set for RT.
403 */
404 static void vrf_rt2ecom_val(struct ecommunity_val *to_eval,
405 const struct vrf_route_target *l3rt, int iter)
406 {
407 const struct ecommunity_val *eval;
408
409 eval = (const struct ecommunity_val *)(l3rt->ecom->val +
410 (iter * ECOMMUNITY_SIZE));
411 /* If using "automatic" or "wildcard *" RT,
412 * we only care about the local-admin sub-field.
413 * This is to facilitate using L3VNI(VRF-VNI)
414 * as the RT for EBGP peering too and simplify
415 * configurations by allowing any ASN via '*'.
416 */
417 memcpy(to_eval, eval, ECOMMUNITY_SIZE);
418
419 if (CHECK_FLAG(l3rt->flags, BGP_VRF_RT_AUTO) ||
420 CHECK_FLAG(l3rt->flags, BGP_VRF_RT_WILD))
421 mask_ecom_global_admin(to_eval, eval);
422 }
423
424 /*
425 * Map one RT to specified VRF.
426 * bgp_vrf = BGP vrf instance
427 */
428 static void map_vrf_to_rt(struct bgp *bgp_vrf, struct vrf_route_target *l3rt)
429 {
430 uint32_t i = 0;
431
432 for (i = 0; i < l3rt->ecom->size; i++) {
433 struct vrf_irt_node *irt = NULL;
434 struct ecommunity_val eval_tmp;
435
436 /* Adjust masking for value */
437 vrf_rt2ecom_val(&eval_tmp, l3rt, i);
438
439 irt = lookup_vrf_import_rt(&eval_tmp);
440
441 if (irt && is_vrf_present_in_irt_vrfs(irt->vrfs, bgp_vrf))
442 return; /* Already mapped. */
443
444 if (!irt)
445 irt = vrf_import_rt_new(&eval_tmp);
446
447 /* Add VRF to the list for this RT. */
448 listnode_add(irt->vrfs, bgp_vrf);
449 }
450 }
451
452 /*
453 * Unmap specified VRF from specified RT. If there are no other
454 * VRFs for this RT, then the RT hash is deleted.
455 * bgp_vrf: BGP VRF specific instance
456 */
457 static void unmap_vrf_from_rt(struct bgp *bgp_vrf,
458 struct vrf_route_target *l3rt)
459 {
460 uint32_t i;
461
462 for (i = 0; i < l3rt->ecom->size; i++) {
463 struct vrf_irt_node *irt;
464 struct ecommunity_val eval_tmp;
465
466 /* Adjust masking for value */
467 vrf_rt2ecom_val(&eval_tmp, l3rt, i);
468
469 irt = lookup_vrf_import_rt(&eval_tmp);
470
471 if (!irt)
472 return; /* Not mapped */
473
474 /* Delete VRF from list for this RT. */
475 listnode_delete(irt->vrfs, bgp_vrf);
476
477 if (!listnode_head(irt->vrfs))
478 vrf_import_rt_free(irt);
479 }
480 }
481
482 /*
483 * Map one RT to specified VNI.
484 */
485 static void map_vni_to_rt(struct bgp *bgp, struct bgpevpn *vpn,
486 struct ecommunity_val *eval)
487 {
488 struct irt_node *irt;
489 struct ecommunity_val eval_tmp;
490
491 /* If using "automatic" RT, we only care about the local-admin
492 * sub-field.
493 * This is to facilitate using VNI as the RT for EBGP peering too.
494 */
495 memcpy(&eval_tmp, eval, ECOMMUNITY_SIZE);
496 if (!is_import_rt_configured(vpn))
497 mask_ecom_global_admin(&eval_tmp, eval);
498
499 irt = lookup_import_rt(bgp, &eval_tmp);
500 if (irt)
501 if (is_vni_present_in_irt_vnis(irt->vnis, vpn))
502 /* Already mapped. */
503 return;
504
505 if (!irt)
506 irt = import_rt_new(bgp, &eval_tmp);
507
508 /* Add VNI to the hash list for this RT. */
509 listnode_add(irt->vnis, vpn);
510 }
511
512 /*
513 * Unmap specified VNI from specified RT. If there are no other
514 * VNIs for this RT, then the RT hash is deleted.
515 */
516 static void unmap_vni_from_rt(struct bgp *bgp, struct bgpevpn *vpn,
517 struct irt_node *irt)
518 {
519 /* Delete VNI from hash list for this RT. */
520 listnode_delete(irt->vnis, vpn);
521 if (!listnode_head(irt->vnis)) {
522 import_rt_free(bgp, irt);
523 }
524 }
525
526 static void bgp_evpn_get_rmac_nexthop(struct bgpevpn *vpn,
527 const struct prefix_evpn *p,
528 struct attr *attr, uint8_t flags)
529 {
530 struct bgp *bgp_vrf = vpn->bgp_vrf;
531
532 memset(&attr->rmac, 0, sizeof(struct ethaddr));
533 if (!bgp_vrf)
534 return;
535
536 if (p->prefix.route_type != BGP_EVPN_MAC_IP_ROUTE)
537 return;
538
539 /* Copy sys (pip) RMAC and PIP IP as nexthop
540 * in case of route is self MAC-IP,
541 * advertise-pip and advertise-svi-ip features
542 * are enabled.
543 * Otherwise, for all host MAC-IP route's
544 * copy anycast RMAC.
545 */
546 if (CHECK_FLAG(flags, BGP_EVPN_MACIP_TYPE_SVI_IP)
547 && bgp_vrf->evpn_info->advertise_pip &&
548 bgp_vrf->evpn_info->is_anycast_mac) {
549 /* copy sys rmac */
550 memcpy(&attr->rmac, &bgp_vrf->evpn_info->pip_rmac,
551 ETH_ALEN);
552 attr->nexthop = bgp_vrf->evpn_info->pip_ip;
553 attr->mp_nexthop_global_in =
554 bgp_vrf->evpn_info->pip_ip;
555 } else
556 memcpy(&attr->rmac, &bgp_vrf->rmac, ETH_ALEN);
557 }
558
559 /*
560 * Create RT extended community automatically from passed information:
561 * of the form AS:VNI.
562 * NOTE: We use only the lower 16 bits of the AS. This is sufficient as
563 * the need is to get a RT value that will be unique across different
564 * VNIs but the same across routers (in the same AS) for a particular
565 * VNI.
566 */
567 static void form_auto_rt(struct bgp *bgp, vni_t vni, struct list *rtl,
568 bool is_l3)
569 {
570 struct ecommunity_val eval;
571 struct ecommunity *ecomadd;
572 struct ecommunity *ecom;
573 struct vrf_route_target *l3rt;
574 struct vrf_route_target *newrt;
575 bool ecom_found = false;
576 struct listnode *node;
577
578 if (bgp->advertise_autort_rfc8365)
579 vni |= EVPN_AUTORT_VXLAN;
580 encode_route_target_as((bgp->as & 0xFFFF), vni, &eval);
581
582 ecomadd = ecommunity_new();
583 ecommunity_add_val(ecomadd, &eval, false, false);
584
585 if (is_l3) {
586 for (ALL_LIST_ELEMENTS_RO(rtl, node, l3rt))
587 if (ecommunity_cmp(ecomadd, l3rt->ecom)) {
588 ecom_found = true;
589 break;
590 }
591 } else {
592 for (ALL_LIST_ELEMENTS_RO(rtl, node, ecom))
593 if (ecommunity_cmp(ecomadd, ecom)) {
594 ecom_found = true;
595 break;
596 }
597 }
598
599 if (!ecom_found) {
600 if (is_l3) {
601 newrt = evpn_vrf_rt_new(ecomadd);
602 /* Label it as autoderived */
603 SET_FLAG(newrt->flags, BGP_VRF_RT_AUTO);
604 listnode_add_sort(rtl, newrt);
605 } else
606 listnode_add_sort(rtl, ecomadd);
607 } else
608 ecommunity_free(&ecomadd);
609 }
610
611 /*
612 * Derive RD and RT for a VNI automatically. Invoked at the time of
613 * creation of a VNI.
614 */
615 static void derive_rd_rt_for_vni(struct bgp *bgp, struct bgpevpn *vpn)
616 {
617 bgp_evpn_derive_auto_rd(bgp, vpn);
618 bgp_evpn_derive_auto_rt_import(bgp, vpn);
619 bgp_evpn_derive_auto_rt_export(bgp, vpn);
620 }
621
622 /*
623 * Convert nexthop (remote VTEP IP) into an IPv6 address.
624 */
625 static void evpn_convert_nexthop_to_ipv6(struct attr *attr)
626 {
627 if (BGP_ATTR_NEXTHOP_AFI_IP6(attr))
628 return;
629 ipv4_to_ipv4_mapped_ipv6(&attr->mp_nexthop_global, attr->nexthop);
630 attr->mp_nexthop_len = IPV6_MAX_BYTELEN;
631 }
632
633 /*
634 * Wrapper for node get in global table.
635 */
636 struct bgp_dest *bgp_evpn_global_node_get(struct bgp_table *table, afi_t afi,
637 safi_t safi,
638 const struct prefix_evpn *evp,
639 struct prefix_rd *prd,
640 const struct bgp_path_info *local_pi)
641 {
642 struct prefix_evpn global_p;
643
644 if (evp->prefix.route_type == BGP_EVPN_AD_ROUTE) {
645 /* prefix in the global table doesn't include the VTEP-IP so
646 * we need to create a different copy of the prefix
647 */
648 evpn_type1_prefix_global_copy(&global_p, evp);
649 evp = &global_p;
650 } else if (evp->prefix.route_type == BGP_EVPN_MAC_IP_ROUTE &&
651 local_pi) {
652 /*
653 * prefix in the global table needs MAC/IP, ensure they are
654 * present, using one's from local table's path_info.
655 */
656 if (is_evpn_prefix_ipaddr_none(evp)) {
657 /* VNI MAC -> Global */
658 evpn_type2_prefix_global_copy(
659 &global_p, evp, NULL /* mac */,
660 evpn_type2_path_info_get_ip(local_pi));
661 } else {
662 /* VNI IP -> Global */
663 evpn_type2_prefix_global_copy(
664 &global_p, evp,
665 evpn_type2_path_info_get_mac(local_pi),
666 NULL /* ip */);
667 }
668
669 evp = &global_p;
670 }
671 return bgp_afi_node_get(table, afi, safi, (struct prefix *)evp, prd);
672 }
673
674 /*
675 * Wrapper for node lookup in global table.
676 */
677 struct bgp_dest *bgp_evpn_global_node_lookup(
678 struct bgp_table *table, safi_t safi, const struct prefix_evpn *evp,
679 struct prefix_rd *prd, const struct bgp_path_info *local_pi)
680 {
681 struct prefix_evpn global_p;
682
683 if (evp->prefix.route_type == BGP_EVPN_AD_ROUTE) {
684 /* prefix in the global table doesn't include the VTEP-IP so
685 * we need to create a different copy of the prefix
686 */
687 evpn_type1_prefix_global_copy(&global_p, evp);
688 evp = &global_p;
689 } else if (evp->prefix.route_type == BGP_EVPN_MAC_IP_ROUTE &&
690 local_pi) {
691 /*
692 * prefix in the global table needs MAC/IP, ensure they are
693 * present, using one's from local table's path_info.
694 */
695 if (is_evpn_prefix_ipaddr_none(evp)) {
696 /* VNI MAC -> Global */
697 evpn_type2_prefix_global_copy(
698 &global_p, evp, NULL /* mac */,
699 evpn_type2_path_info_get_ip(local_pi));
700 } else {
701 /* VNI IP -> Global */
702 evpn_type2_prefix_global_copy(
703 &global_p, evp,
704 evpn_type2_path_info_get_mac(local_pi),
705 NULL /* ip */);
706 }
707
708 evp = &global_p;
709 }
710 return bgp_safi_node_lookup(table, safi, (struct prefix *)evp, prd);
711 }
712
713 /*
714 * Wrapper for node get in VNI IP table.
715 */
716 struct bgp_dest *bgp_evpn_vni_ip_node_get(struct bgp_table *const table,
717 const struct prefix_evpn *evp,
718 const struct bgp_path_info *parent_pi)
719 {
720 struct prefix_evpn vni_p;
721
722 if (evp->prefix.route_type == BGP_EVPN_AD_ROUTE && parent_pi) {
723 /* prefix in the global table doesn't include the VTEP-IP so
724 * we need to create a different copy for the VNI
725 */
726 evpn_type1_prefix_vni_ip_copy(&vni_p, evp,
727 parent_pi->attr->nexthop);
728 evp = &vni_p;
729 } else if (evp->prefix.route_type == BGP_EVPN_MAC_IP_ROUTE) {
730 /* Only MAC-IP should go into this table, not mac-only */
731 assert(is_evpn_prefix_ipaddr_none(evp) == false);
732
733 /*
734 * prefix in the vni IP table doesn't include MAC so
735 * we need to create a different copy of the prefix.
736 */
737 evpn_type2_prefix_vni_ip_copy(&vni_p, evp);
738 evp = &vni_p;
739 }
740 return bgp_node_get(table, (struct prefix *)evp);
741 }
742
743 /*
744 * Wrapper for node lookup in VNI IP table.
745 */
746 struct bgp_dest *
747 bgp_evpn_vni_ip_node_lookup(const struct bgp_table *const table,
748 const struct prefix_evpn *evp,
749 const struct bgp_path_info *parent_pi)
750 {
751 struct prefix_evpn vni_p;
752
753 if (evp->prefix.route_type == BGP_EVPN_AD_ROUTE && parent_pi) {
754 /* prefix in the global table doesn't include the VTEP-IP so
755 * we need to create a different copy for the VNI
756 */
757 evpn_type1_prefix_vni_ip_copy(&vni_p, evp,
758 parent_pi->attr->nexthop);
759 evp = &vni_p;
760 } else if (evp->prefix.route_type == BGP_EVPN_MAC_IP_ROUTE) {
761 /* Only MAC-IP should go into this table, not mac-only */
762 assert(is_evpn_prefix_ipaddr_none(evp) == false);
763
764 /*
765 * prefix in the vni IP table doesn't include MAC so
766 * we need to create a different copy of the prefix.
767 */
768 evpn_type2_prefix_vni_ip_copy(&vni_p, evp);
769 evp = &vni_p;
770 }
771 return bgp_node_lookup(table, (struct prefix *)evp);
772 }
773
774 /*
775 * Wrapper for node get in VNI MAC table.
776 */
777 struct bgp_dest *
778 bgp_evpn_vni_mac_node_get(struct bgp_table *const table,
779 const struct prefix_evpn *evp,
780 const struct bgp_path_info *parent_pi)
781 {
782 struct prefix_evpn vni_p;
783
784 /* Only type-2 should ever go into this table */
785 assert(evp->prefix.route_type == BGP_EVPN_MAC_IP_ROUTE);
786
787 /*
788 * prefix in the vni MAC table doesn't include IP so
789 * we need to create a different copy of the prefix.
790 */
791 evpn_type2_prefix_vni_mac_copy(&vni_p, evp);
792 evp = &vni_p;
793 return bgp_node_get(table, (struct prefix *)evp);
794 }
795
796 /*
797 * Wrapper for node lookup in VNI MAC table.
798 */
799 struct bgp_dest *
800 bgp_evpn_vni_mac_node_lookup(const struct bgp_table *const table,
801 const struct prefix_evpn *evp,
802 const struct bgp_path_info *parent_pi)
803 {
804 struct prefix_evpn vni_p;
805
806 /* Only type-2 should ever go into this table */
807 assert(evp->prefix.route_type == BGP_EVPN_MAC_IP_ROUTE);
808
809 /*
810 * prefix in the vni MAC table doesn't include IP so
811 * we need to create a different copy of the prefix.
812 */
813 evpn_type2_prefix_vni_mac_copy(&vni_p, evp);
814 evp = &vni_p;
815 return bgp_node_lookup(table, (struct prefix *)evp);
816 }
817
818 /*
819 * Wrapper for node get in both VNI tables.
820 */
821 struct bgp_dest *bgp_evpn_vni_node_get(struct bgpevpn *vpn,
822 const struct prefix_evpn *p,
823 const struct bgp_path_info *parent_pi)
824 {
825 if ((p->prefix.route_type == BGP_EVPN_MAC_IP_ROUTE) &&
826 (is_evpn_prefix_ipaddr_none(p) == true))
827 return bgp_evpn_vni_mac_node_get(vpn->mac_table, p, parent_pi);
828
829 return bgp_evpn_vni_ip_node_get(vpn->ip_table, p, parent_pi);
830 }
831
832 /*
833 * Wrapper for node lookup in both VNI tables.
834 */
835 struct bgp_dest *bgp_evpn_vni_node_lookup(const struct bgpevpn *vpn,
836 const struct prefix_evpn *p,
837 const struct bgp_path_info *parent_pi)
838 {
839 if ((p->prefix.route_type == BGP_EVPN_MAC_IP_ROUTE) &&
840 (is_evpn_prefix_ipaddr_none(p) == true))
841 return bgp_evpn_vni_mac_node_lookup(vpn->mac_table, p,
842 parent_pi);
843
844 return bgp_evpn_vni_ip_node_lookup(vpn->ip_table, p, parent_pi);
845 }
846
847 /*
848 * Add (update) or delete MACIP from zebra.
849 */
850 static int bgp_zebra_send_remote_macip(struct bgp *bgp, struct bgpevpn *vpn,
851 const struct prefix_evpn *p,
852 const struct ethaddr *mac,
853 struct in_addr remote_vtep_ip, int add,
854 uint8_t flags, uint32_t seq, esi_t *esi)
855 {
856 struct stream *s;
857 uint16_t ipa_len;
858 static struct in_addr zero_remote_vtep_ip;
859 bool esi_valid;
860
861 /* Check socket. */
862 if (!zclient || zclient->sock < 0)
863 return 0;
864
865 /* Don't try to register if Zebra doesn't know of this instance. */
866 if (!IS_BGP_INST_KNOWN_TO_ZEBRA(bgp)) {
867 if (BGP_DEBUG(zebra, ZEBRA))
868 zlog_debug(
869 "%s: No zebra instance to talk to, not installing remote macip",
870 __func__);
871 return 0;
872 }
873
874 if (!esi)
875 esi = zero_esi;
876 s = zclient->obuf;
877 stream_reset(s);
878
879 zclient_create_header(
880 s, add ? ZEBRA_REMOTE_MACIP_ADD : ZEBRA_REMOTE_MACIP_DEL,
881 bgp->vrf_id);
882 stream_putl(s, vpn->vni);
883
884 if (mac) /* Mac Addr */
885 stream_put(s, &mac->octet, ETH_ALEN);
886 else
887 stream_put(s, &p->prefix.macip_addr.mac.octet, ETH_ALEN);
888
889 /* IP address length and IP address, if any. */
890 if (is_evpn_prefix_ipaddr_none(p))
891 stream_putw(s, 0);
892 else {
893 ipa_len = is_evpn_prefix_ipaddr_v4(p) ? IPV4_MAX_BYTELEN
894 : IPV6_MAX_BYTELEN;
895 stream_putw(s, ipa_len);
896 stream_put(s, &p->prefix.macip_addr.ip.ip.addr, ipa_len);
897 }
898 /* If the ESI is valid that becomes the nexthop; tape out the
899 * VTEP-IP for that case
900 */
901 if (bgp_evpn_is_esi_valid(esi)) {
902 esi_valid = true;
903 stream_put_in_addr(s, &zero_remote_vtep_ip);
904 } else {
905 esi_valid = false;
906 stream_put_in_addr(s, &remote_vtep_ip);
907 }
908
909 /* TX flags - MAC sticky status and/or gateway mac */
910 /* Also TX the sequence number of the best route. */
911 if (add) {
912 stream_putc(s, flags);
913 stream_putl(s, seq);
914 stream_put(s, esi, sizeof(esi_t));
915 }
916
917 stream_putw_at(s, 0, stream_get_endp(s));
918
919 if (bgp_debug_zebra(NULL)) {
920 char esi_buf[ESI_STR_LEN];
921
922 if (esi_valid)
923 esi_to_str(esi, esi_buf, sizeof(esi_buf));
924 else
925 snprintf(esi_buf, sizeof(esi_buf), "-");
926 zlog_debug(
927 "Tx %s MACIP, VNI %u MAC %pEA IP %pIA flags 0x%x seq %u remote VTEP %pI4 esi %s",
928 add ? "ADD" : "DEL", vpn->vni,
929 (mac ? mac : &p->prefix.macip_addr.mac),
930 &p->prefix.macip_addr.ip, flags, seq, &remote_vtep_ip,
931 esi_buf);
932 }
933
934 frrtrace(5, frr_bgp, evpn_mac_ip_zsend, add, vpn, p, remote_vtep_ip,
935 esi);
936
937 return zclient_send_message(zclient);
938 }
939
940 /*
941 * Add (update) or delete remote VTEP from zebra.
942 */
943 static int bgp_zebra_send_remote_vtep(struct bgp *bgp, struct bgpevpn *vpn,
944 const struct prefix_evpn *p,
945 int flood_control, int add)
946 {
947 struct stream *s;
948
949 /* Check socket. */
950 if (!zclient || zclient->sock < 0)
951 return 0;
952
953 /* Don't try to register if Zebra doesn't know of this instance. */
954 if (!IS_BGP_INST_KNOWN_TO_ZEBRA(bgp)) {
955 if (BGP_DEBUG(zebra, ZEBRA))
956 zlog_debug(
957 "%s: No zebra instance to talk to, not installing remote vtep",
958 __func__);
959 return 0;
960 }
961
962 s = zclient->obuf;
963 stream_reset(s);
964
965 zclient_create_header(
966 s, add ? ZEBRA_REMOTE_VTEP_ADD : ZEBRA_REMOTE_VTEP_DEL,
967 bgp->vrf_id);
968 stream_putl(s, vpn->vni);
969 if (is_evpn_prefix_ipaddr_v4(p))
970 stream_put_in_addr(s, &p->prefix.imet_addr.ip.ipaddr_v4);
971 else if (is_evpn_prefix_ipaddr_v6(p)) {
972 flog_err(
973 EC_BGP_VTEP_INVALID,
974 "Bad remote IP when trying to %s remote VTEP for VNI %u",
975 add ? "ADD" : "DEL", vpn->vni);
976 return -1;
977 }
978 stream_putl(s, flood_control);
979
980 stream_putw_at(s, 0, stream_get_endp(s));
981
982 if (bgp_debug_zebra(NULL))
983 zlog_debug("Tx %s Remote VTEP, VNI %u remote VTEP %pI4",
984 add ? "ADD" : "DEL", vpn->vni,
985 &p->prefix.imet_addr.ip.ipaddr_v4);
986
987 frrtrace(3, frr_bgp, evpn_bum_vtep_zsend, add, vpn, p);
988
989 return zclient_send_message(zclient);
990 }
991
992 /*
993 * Build extended communities for EVPN prefix route.
994 */
995 static void build_evpn_type5_route_extcomm(struct bgp *bgp_vrf,
996 struct attr *attr)
997 {
998 struct ecommunity ecom_encap;
999 struct ecommunity_val eval;
1000 struct ecommunity_val eval_rmac;
1001 bgp_encap_types tnl_type;
1002 struct listnode *node, *nnode;
1003 struct vrf_route_target *l3rt;
1004 struct ecommunity *old_ecom;
1005 struct ecommunity *ecom;
1006 struct list *vrf_export_rtl = NULL;
1007
1008 /* Encap */
1009 tnl_type = BGP_ENCAP_TYPE_VXLAN;
1010 memset(&ecom_encap, 0, sizeof(ecom_encap));
1011 encode_encap_extcomm(tnl_type, &eval);
1012 ecom_encap.size = 1;
1013 ecom_encap.unit_size = ECOMMUNITY_SIZE;
1014 ecom_encap.val = (uint8_t *)eval.val;
1015
1016 /* Add Encap */
1017 if (bgp_attr_get_ecommunity(attr)) {
1018 old_ecom = bgp_attr_get_ecommunity(attr);
1019 ecom = ecommunity_merge(ecommunity_dup(old_ecom), &ecom_encap);
1020 if (!old_ecom->refcnt)
1021 ecommunity_free(&old_ecom);
1022 } else
1023 ecom = ecommunity_dup(&ecom_encap);
1024 bgp_attr_set_ecommunity(attr, ecom);
1025 attr->encap_tunneltype = tnl_type;
1026
1027 /* Add the export RTs for L3VNI/VRF */
1028 vrf_export_rtl = bgp_vrf->vrf_export_rtl;
1029 for (ALL_LIST_ELEMENTS(vrf_export_rtl, node, nnode, l3rt))
1030 bgp_attr_set_ecommunity(
1031 attr, ecommunity_merge(bgp_attr_get_ecommunity(attr),
1032 l3rt->ecom));
1033
1034 /* add the router mac extended community */
1035 if (!is_zero_mac(&attr->rmac)) {
1036 encode_rmac_extcomm(&eval_rmac, &attr->rmac);
1037 ecommunity_add_val(bgp_attr_get_ecommunity(attr), &eval_rmac,
1038 true, true);
1039 }
1040 }
1041
1042 /*
1043 * Build extended communities for EVPN route.
1044 * This function is applicable for type-2 and type-3 routes. The layer-2 RT
1045 * and ENCAP extended communities are applicable for all routes.
1046 * The default gateway extended community and MAC mobility (sticky) extended
1047 * community are added as needed based on passed settings - only for type-2
1048 * routes. Likewise, the layer-3 RT and Router MAC extended communities are
1049 * added, if present, based on passed settings - only for non-link-local
1050 * type-2 routes.
1051 */
1052 static void build_evpn_route_extcomm(struct bgpevpn *vpn, struct attr *attr,
1053 int add_l3_ecomm)
1054 {
1055 struct ecommunity ecom_encap;
1056 struct ecommunity ecom_sticky;
1057 struct ecommunity ecom_default_gw;
1058 struct ecommunity ecom_na;
1059 struct ecommunity_val eval;
1060 struct ecommunity_val eval_sticky;
1061 struct ecommunity_val eval_default_gw;
1062 struct ecommunity_val eval_rmac;
1063 struct ecommunity_val eval_na;
1064 bool proxy;
1065
1066 bgp_encap_types tnl_type;
1067 struct listnode *node, *nnode;
1068 struct ecommunity *ecom;
1069 struct vrf_route_target *l3rt;
1070 uint32_t seqnum;
1071 struct list *vrf_export_rtl = NULL;
1072
1073 /* Encap */
1074 tnl_type = BGP_ENCAP_TYPE_VXLAN;
1075 memset(&ecom_encap, 0, sizeof(ecom_encap));
1076 encode_encap_extcomm(tnl_type, &eval);
1077 ecom_encap.size = 1;
1078 ecom_encap.unit_size = ECOMMUNITY_SIZE;
1079 ecom_encap.val = (uint8_t *)eval.val;
1080
1081 /* Add Encap */
1082 bgp_attr_set_ecommunity(attr, ecommunity_dup(&ecom_encap));
1083 attr->encap_tunneltype = tnl_type;
1084
1085 /* Add the export RTs for L2VNI */
1086 for (ALL_LIST_ELEMENTS(vpn->export_rtl, node, nnode, ecom))
1087 bgp_attr_set_ecommunity(
1088 attr,
1089 ecommunity_merge(bgp_attr_get_ecommunity(attr), ecom));
1090
1091 /* Add the export RTs for L3VNI if told to - caller determines
1092 * when this should be done.
1093 */
1094 if (add_l3_ecomm) {
1095 vrf_export_rtl = bgpevpn_get_vrf_export_rtl(vpn);
1096 if (vrf_export_rtl && !list_isempty(vrf_export_rtl)) {
1097 for (ALL_LIST_ELEMENTS(vrf_export_rtl, node, nnode,
1098 l3rt))
1099 bgp_attr_set_ecommunity(
1100 attr,
1101 ecommunity_merge(
1102 bgp_attr_get_ecommunity(attr),
1103 l3rt->ecom));
1104 }
1105 }
1106
1107 /* Add MAC mobility (sticky) if needed. */
1108 if (attr->sticky) {
1109 seqnum = 0;
1110 memset(&ecom_sticky, 0, sizeof(ecom_sticky));
1111 encode_mac_mobility_extcomm(1, seqnum, &eval_sticky);
1112 ecom_sticky.size = 1;
1113 ecom_sticky.unit_size = ECOMMUNITY_SIZE;
1114 ecom_sticky.val = (uint8_t *)eval_sticky.val;
1115 bgp_attr_set_ecommunity(
1116 attr, ecommunity_merge(bgp_attr_get_ecommunity(attr),
1117 &ecom_sticky));
1118 }
1119
1120 /* Add RMAC, if told to. */
1121 if (add_l3_ecomm) {
1122 encode_rmac_extcomm(&eval_rmac, &attr->rmac);
1123 ecommunity_add_val(bgp_attr_get_ecommunity(attr), &eval_rmac,
1124 true, true);
1125 }
1126
1127 /* Add default gateway, if needed. */
1128 if (attr->default_gw) {
1129 memset(&ecom_default_gw, 0, sizeof(ecom_default_gw));
1130 encode_default_gw_extcomm(&eval_default_gw);
1131 ecom_default_gw.size = 1;
1132 ecom_default_gw.unit_size = ECOMMUNITY_SIZE;
1133 ecom_default_gw.val = (uint8_t *)eval_default_gw.val;
1134 bgp_attr_set_ecommunity(
1135 attr, ecommunity_merge(bgp_attr_get_ecommunity(attr),
1136 &ecom_default_gw));
1137 }
1138
1139 proxy = !!(attr->es_flags & ATTR_ES_PROXY_ADVERT);
1140 if (attr->router_flag || proxy) {
1141 memset(&ecom_na, 0, sizeof(ecom_na));
1142 encode_na_flag_extcomm(&eval_na, attr->router_flag, proxy);
1143 ecom_na.size = 1;
1144 ecom_na.unit_size = ECOMMUNITY_SIZE;
1145 ecom_na.val = (uint8_t *)eval_na.val;
1146 bgp_attr_set_ecommunity(
1147 attr, ecommunity_merge(bgp_attr_get_ecommunity(attr),
1148 &ecom_na));
1149 }
1150 }
1151
1152 /*
1153 * Add MAC mobility extended community to attribute.
1154 */
1155 static void add_mac_mobility_to_attr(uint32_t seq_num, struct attr *attr)
1156 {
1157 struct ecommunity ecom_tmp;
1158 struct ecommunity_val eval;
1159 uint8_t *ecom_val_ptr;
1160 uint32_t i;
1161 uint8_t *pnt;
1162 int type = 0;
1163 int sub_type = 0;
1164 struct ecommunity *ecomm = bgp_attr_get_ecommunity(attr);
1165
1166 /* Build MM */
1167 encode_mac_mobility_extcomm(0, seq_num, &eval);
1168
1169 /* Find current MM ecommunity */
1170 ecom_val_ptr = NULL;
1171
1172 if (ecomm) {
1173 for (i = 0; i < ecomm->size; i++) {
1174 pnt = ecomm->val + (i * ecomm->unit_size);
1175 type = *pnt++;
1176 sub_type = *pnt++;
1177
1178 if (type == ECOMMUNITY_ENCODE_EVPN
1179 && sub_type
1180 == ECOMMUNITY_EVPN_SUBTYPE_MACMOBILITY) {
1181 ecom_val_ptr =
1182 (ecomm->val + (i * ecomm->unit_size));
1183 break;
1184 }
1185 }
1186 }
1187
1188 /* Update the existing MM ecommunity */
1189 if (ecom_val_ptr) {
1190 memcpy(ecom_val_ptr, eval.val, sizeof(char) * ecomm->unit_size);
1191 }
1192 /* Add MM to existing */
1193 else {
1194 memset(&ecom_tmp, 0, sizeof(ecom_tmp));
1195 ecom_tmp.size = 1;
1196 ecom_tmp.unit_size = ECOMMUNITY_SIZE;
1197 ecom_tmp.val = (uint8_t *)eval.val;
1198
1199 if (ecomm)
1200 bgp_attr_set_ecommunity(
1201 attr, ecommunity_merge(ecomm, &ecom_tmp));
1202 else
1203 bgp_attr_set_ecommunity(attr,
1204 ecommunity_dup(&ecom_tmp));
1205 }
1206 }
1207
1208 /* Install EVPN route into zebra. */
1209 static int evpn_zebra_install(struct bgp *bgp, struct bgpevpn *vpn,
1210 const struct prefix_evpn *p,
1211 struct bgp_path_info *pi)
1212 {
1213 int ret;
1214 uint8_t flags;
1215 int flood_control = VXLAN_FLOOD_DISABLED;
1216 uint32_t seq;
1217
1218 if (p->prefix.route_type == BGP_EVPN_MAC_IP_ROUTE) {
1219 flags = 0;
1220
1221 if (pi->sub_type == BGP_ROUTE_IMPORTED) {
1222 if (pi->attr->sticky)
1223 SET_FLAG(flags, ZEBRA_MACIP_TYPE_STICKY);
1224 if (pi->attr->default_gw)
1225 SET_FLAG(flags, ZEBRA_MACIP_TYPE_GW);
1226 if (is_evpn_prefix_ipaddr_v6(p) &&
1227 pi->attr->router_flag)
1228 SET_FLAG(flags, ZEBRA_MACIP_TYPE_ROUTER_FLAG);
1229
1230 seq = mac_mobility_seqnum(pi->attr);
1231 /* if local ES notify zebra that this is a sync path */
1232 if (bgp_evpn_attr_is_local_es(pi->attr)) {
1233 SET_FLAG(flags, ZEBRA_MACIP_TYPE_SYNC_PATH);
1234 if (bgp_evpn_attr_is_proxy(pi->attr))
1235 SET_FLAG(flags,
1236 ZEBRA_MACIP_TYPE_PROXY_ADVERT);
1237 }
1238 } else {
1239 if (!bgp_evpn_attr_is_sync(pi->attr))
1240 return 0;
1241
1242 /* if a local path is being turned around and sent
1243 * to zebra it is because it is a sync path on
1244 * a local ES
1245 */
1246 SET_FLAG(flags, ZEBRA_MACIP_TYPE_SYNC_PATH);
1247 /* supply the highest peer seq number to zebra
1248 * for MM seq syncing
1249 */
1250 seq = bgp_evpn_attr_get_sync_seq(pi->attr);
1251 /* if any of the paths from the peer have the ROUTER
1252 * flag set install the local entry as a router entry
1253 */
1254 if (is_evpn_prefix_ipaddr_v6(p) &&
1255 (pi->attr->es_flags &
1256 ATTR_ES_PEER_ROUTER))
1257 SET_FLAG(flags,
1258 ZEBRA_MACIP_TYPE_ROUTER_FLAG);
1259
1260 if (!(pi->attr->es_flags & ATTR_ES_PEER_ACTIVE))
1261 SET_FLAG(flags,
1262 ZEBRA_MACIP_TYPE_PROXY_ADVERT);
1263 }
1264
1265 ret = bgp_zebra_send_remote_macip(
1266 bgp, vpn, p,
1267 (is_evpn_prefix_ipaddr_none(p)
1268 ? NULL /* MAC update */
1269 : evpn_type2_path_info_get_mac(
1270 pi) /* MAC-IP update */),
1271 pi->attr->nexthop, 1, flags, seq,
1272 bgp_evpn_attr_get_esi(pi->attr));
1273 } else if (p->prefix.route_type == BGP_EVPN_AD_ROUTE) {
1274 ret = bgp_evpn_remote_es_evi_add(bgp, vpn, p);
1275 } else {
1276 switch (bgp_attr_get_pmsi_tnl_type(pi->attr)) {
1277 case PMSI_TNLTYPE_INGR_REPL:
1278 flood_control = VXLAN_FLOOD_HEAD_END_REPL;
1279 break;
1280
1281 case PMSI_TNLTYPE_PIM_SM:
1282 flood_control = VXLAN_FLOOD_PIM_SM;
1283 break;
1284
1285 case PMSI_TNLTYPE_NO_INFO:
1286 case PMSI_TNLTYPE_RSVP_TE_P2MP:
1287 case PMSI_TNLTYPE_MLDP_P2MP:
1288 case PMSI_TNLTYPE_PIM_SSM:
1289 case PMSI_TNLTYPE_PIM_BIDIR:
1290 case PMSI_TNLTYPE_MLDP_MP2MP:
1291 flood_control = VXLAN_FLOOD_DISABLED;
1292 break;
1293 }
1294 ret = bgp_zebra_send_remote_vtep(bgp, vpn, p, flood_control, 1);
1295 }
1296
1297 return ret;
1298 }
1299
1300 /* Uninstall EVPN route from zebra. */
1301 static int evpn_zebra_uninstall(struct bgp *bgp, struct bgpevpn *vpn,
1302 const struct prefix_evpn *p,
1303 struct bgp_path_info *pi, bool is_sync)
1304 {
1305 int ret;
1306
1307 if (p->prefix.route_type == BGP_EVPN_MAC_IP_ROUTE)
1308 ret = bgp_zebra_send_remote_macip(
1309 bgp, vpn, p,
1310 (is_evpn_prefix_ipaddr_none(p)
1311 ? NULL /* MAC update */
1312 : evpn_type2_path_info_get_mac(
1313 pi) /* MAC-IP update */),
1314 (is_sync ? zero_vtep_ip : pi->attr->nexthop), 0, 0, 0,
1315 NULL);
1316 else if (p->prefix.route_type == BGP_EVPN_AD_ROUTE)
1317 ret = bgp_evpn_remote_es_evi_del(bgp, vpn, p);
1318 else
1319 ret = bgp_zebra_send_remote_vtep(bgp, vpn, p,
1320 VXLAN_FLOOD_DISABLED, 0);
1321
1322 return ret;
1323 }
1324
1325 /*
1326 * Due to MAC mobility, the prior "local" best route has been supplanted
1327 * by a "remote" best route. The prior route has to be deleted and withdrawn
1328 * from peers.
1329 */
1330 static void evpn_delete_old_local_route(struct bgp *bgp, struct bgpevpn *vpn,
1331 struct bgp_dest *dest,
1332 struct bgp_path_info *old_local,
1333 struct bgp_path_info *new_select)
1334 {
1335 struct bgp_dest *global_dest;
1336 struct bgp_path_info *pi;
1337 afi_t afi = AFI_L2VPN;
1338 safi_t safi = SAFI_EVPN;
1339
1340 if (BGP_DEBUG(evpn_mh, EVPN_MH_RT)) {
1341 char esi_buf[ESI_STR_LEN];
1342 char esi_buf2[ESI_STR_LEN];
1343 struct prefix_evpn *evp =
1344 (struct prefix_evpn *)bgp_dest_get_prefix(dest);
1345
1346 zlog_debug("local path deleted %pFX es %s; new-path-es %s", evp,
1347 esi_to_str(&old_local->attr->esi, esi_buf,
1348 sizeof(esi_buf)),
1349 new_select ? esi_to_str(&new_select->attr->esi,
1350 esi_buf2, sizeof(esi_buf2))
1351 : "");
1352 }
1353
1354 /* Locate route node in the global EVPN routing table. Note that
1355 * this table is a 2-level tree (RD-level + Prefix-level) similar to
1356 * L3VPN routes.
1357 */
1358 global_dest = bgp_evpn_global_node_lookup(
1359 bgp->rib[afi][safi], safi,
1360 (const struct prefix_evpn *)bgp_dest_get_prefix(dest),
1361 &vpn->prd, old_local);
1362 if (global_dest) {
1363 /* Delete route entry in the global EVPN table. */
1364 delete_evpn_route_entry(bgp, afi, safi, global_dest, &pi);
1365
1366 /* Schedule for processing - withdraws to peers happen from
1367 * this table.
1368 */
1369 if (pi)
1370 bgp_process(bgp, global_dest, afi, safi);
1371 bgp_dest_unlock_node(global_dest);
1372 }
1373
1374 /* Delete route entry in the VNI route table, caller to remove. */
1375 bgp_path_info_delete(dest, old_local);
1376 }
1377
1378 /*
1379 * Calculate the best path for an EVPN route. Install/update best path in zebra,
1380 * if appropriate.
1381 * Note: vpn is NULL for local EAD-ES routes.
1382 */
1383 int evpn_route_select_install(struct bgp *bgp, struct bgpevpn *vpn,
1384 struct bgp_dest *dest)
1385 {
1386 struct bgp_path_info *old_select, *new_select;
1387 struct bgp_path_info_pair old_and_new;
1388 afi_t afi = AFI_L2VPN;
1389 safi_t safi = SAFI_EVPN;
1390 int ret = 0;
1391
1392 /* Compute the best path. */
1393 bgp_best_selection(bgp, dest, &bgp->maxpaths[afi][safi], &old_and_new,
1394 afi, safi);
1395 old_select = old_and_new.old;
1396 new_select = old_and_new.new;
1397
1398 /* If the best path hasn't changed - see if there is still something to
1399 * update to zebra RIB.
1400 * Remote routes and SYNC route (i.e. local routes with
1401 * SYNCED_FROM_PEER flag) need to updated to zebra on any attr
1402 * change.
1403 */
1404 if (old_select && old_select == new_select
1405 && old_select->type == ZEBRA_ROUTE_BGP
1406 && (old_select->sub_type == BGP_ROUTE_IMPORTED ||
1407 bgp_evpn_attr_is_sync(old_select->attr))
1408 && !CHECK_FLAG(dest->flags, BGP_NODE_USER_CLEAR)
1409 && !CHECK_FLAG(old_select->flags, BGP_PATH_ATTR_CHANGED)
1410 && !bgp_addpath_is_addpath_used(&bgp->tx_addpath, afi, safi)) {
1411 if (bgp_zebra_has_route_changed(old_select))
1412 ret = evpn_zebra_install(
1413 bgp, vpn,
1414 (const struct prefix_evpn *)bgp_dest_get_prefix(
1415 dest),
1416 old_select);
1417 UNSET_FLAG(old_select->flags, BGP_PATH_MULTIPATH_CHG);
1418 UNSET_FLAG(old_select->flags, BGP_PATH_LINK_BW_CHG);
1419 bgp_zebra_clear_route_change_flags(dest);
1420 return ret;
1421 }
1422
1423 /* If the user did a "clear" this flag will be set */
1424 UNSET_FLAG(dest->flags, BGP_NODE_USER_CLEAR);
1425
1426 /* bestpath has changed; update relevant fields and install or uninstall
1427 * into the zebra RIB.
1428 */
1429 if (old_select || new_select)
1430 bgp_bump_version(dest);
1431
1432 if (old_select)
1433 bgp_path_info_unset_flag(dest, old_select, BGP_PATH_SELECTED);
1434 if (new_select) {
1435 bgp_path_info_set_flag(dest, new_select, BGP_PATH_SELECTED);
1436 bgp_path_info_unset_flag(dest, new_select,
1437 BGP_PATH_ATTR_CHANGED);
1438 UNSET_FLAG(new_select->flags, BGP_PATH_MULTIPATH_CHG);
1439 UNSET_FLAG(new_select->flags, BGP_PATH_LINK_BW_CHG);
1440 }
1441
1442 /* a local entry with the SYNC flag also results in a MAC-IP update
1443 * to zebra
1444 */
1445 if (new_select && new_select->type == ZEBRA_ROUTE_BGP
1446 && (new_select->sub_type == BGP_ROUTE_IMPORTED ||
1447 bgp_evpn_attr_is_sync(new_select->attr))) {
1448 ret = evpn_zebra_install(
1449 bgp, vpn,
1450 (struct prefix_evpn *)bgp_dest_get_prefix(dest),
1451 new_select);
1452
1453 /* If an old best existed and it was a "local" route, the only
1454 * reason
1455 * it would be supplanted is due to MAC mobility procedures. So,
1456 * we
1457 * need to do an implicit delete and withdraw that route from
1458 * peers.
1459 */
1460 if (new_select->sub_type == BGP_ROUTE_IMPORTED &&
1461 old_select && old_select->peer == bgp->peer_self
1462 && old_select->type == ZEBRA_ROUTE_BGP
1463 && old_select->sub_type == BGP_ROUTE_STATIC
1464 && vpn)
1465 evpn_delete_old_local_route(bgp, vpn, dest,
1466 old_select, new_select);
1467 } else {
1468 if (old_select && old_select->type == ZEBRA_ROUTE_BGP
1469 && old_select->sub_type == BGP_ROUTE_IMPORTED)
1470 ret = evpn_zebra_uninstall(
1471 bgp, vpn,
1472 (const struct prefix_evpn *)bgp_dest_get_prefix(
1473 dest),
1474 old_select, false);
1475 }
1476
1477 /* Clear any route change flags. */
1478 bgp_zebra_clear_route_change_flags(dest);
1479
1480 /* Reap old select bgp_path_info, if it has been removed */
1481 if (old_select && CHECK_FLAG(old_select->flags, BGP_PATH_REMOVED))
1482 bgp_path_info_reap(dest, old_select);
1483
1484 return ret;
1485 }
1486
1487 static struct bgp_path_info *bgp_evpn_route_get_local_path(
1488 struct bgp *bgp, struct bgp_dest *dest)
1489 {
1490 struct bgp_path_info *tmp_pi;
1491 struct bgp_path_info *local_pi = NULL;
1492
1493 for (tmp_pi = bgp_dest_get_bgp_path_info(dest); tmp_pi;
1494 tmp_pi = tmp_pi->next) {
1495 if (bgp_evpn_is_path_local(bgp, tmp_pi)) {
1496 local_pi = tmp_pi;
1497 break;
1498 }
1499 }
1500
1501 return local_pi;
1502 }
1503
1504 static int update_evpn_type5_route_entry(struct bgp *bgp_evpn,
1505 struct bgp *bgp_vrf, afi_t afi,
1506 safi_t safi, struct bgp_dest *dest,
1507 struct attr *attr, int *route_changed)
1508 {
1509 struct attr *attr_new = NULL;
1510 struct bgp_path_info *pi = NULL;
1511 mpls_label_t label = MPLS_INVALID_LABEL;
1512 struct bgp_path_info *local_pi = NULL;
1513 struct bgp_path_info *tmp_pi = NULL;
1514
1515 *route_changed = 0;
1516 /* locate the local route entry if any */
1517 for (tmp_pi = bgp_dest_get_bgp_path_info(dest); tmp_pi;
1518 tmp_pi = tmp_pi->next) {
1519 if (tmp_pi->peer == bgp_evpn->peer_self
1520 && tmp_pi->type == ZEBRA_ROUTE_BGP
1521 && tmp_pi->sub_type == BGP_ROUTE_STATIC)
1522 local_pi = tmp_pi;
1523 }
1524
1525 /*
1526 * create a new route entry if one doesn't exist.
1527 * Otherwise see if route attr has changed
1528 */
1529 if (!local_pi) {
1530
1531 /* route has changed as this is the first entry */
1532 *route_changed = 1;
1533
1534 /* Add (or update) attribute to hash. */
1535 attr_new = bgp_attr_intern(attr);
1536
1537 /* create the route info from attribute */
1538 pi = info_make(ZEBRA_ROUTE_BGP, BGP_ROUTE_STATIC, 0,
1539 bgp_evpn->peer_self, attr_new, dest);
1540 SET_FLAG(pi->flags, BGP_PATH_VALID);
1541
1542 /* Type-5 routes advertise the L3-VNI */
1543 bgp_path_info_extra_get(pi);
1544 vni2label(bgp_vrf->l3vni, &label);
1545 memcpy(&pi->extra->label, &label, sizeof(label));
1546 pi->extra->num_labels = 1;
1547
1548 /* add the route entry to route node*/
1549 bgp_path_info_add(dest, pi);
1550 } else {
1551
1552 tmp_pi = local_pi;
1553 if (!attrhash_cmp(tmp_pi->attr, attr)) {
1554
1555 /* attribute changed */
1556 *route_changed = 1;
1557
1558 /* The attribute has changed. */
1559 /* Add (or update) attribute to hash. */
1560 attr_new = bgp_attr_intern(attr);
1561 bgp_path_info_set_flag(dest, tmp_pi,
1562 BGP_PATH_ATTR_CHANGED);
1563
1564 /* Restore route, if needed. */
1565 if (CHECK_FLAG(tmp_pi->flags, BGP_PATH_REMOVED))
1566 bgp_path_info_restore(dest, tmp_pi);
1567
1568 /* Unintern existing, set to new. */
1569 bgp_attr_unintern(&tmp_pi->attr);
1570 tmp_pi->attr = attr_new;
1571 tmp_pi->uptime = monotime(NULL);
1572 }
1573 }
1574 return 0;
1575 }
1576
1577 /* update evpn type-5 route entry */
1578 static int update_evpn_type5_route(struct bgp *bgp_vrf, struct prefix_evpn *evp,
1579 struct attr *src_attr, afi_t src_afi,
1580 safi_t src_safi)
1581 {
1582 afi_t afi = AFI_L2VPN;
1583 safi_t safi = SAFI_EVPN;
1584 struct attr attr;
1585 struct bgp_dest *dest = NULL;
1586 struct bgp *bgp_evpn = NULL;
1587 int route_changed = 0;
1588
1589 bgp_evpn = bgp_get_evpn();
1590 if (!bgp_evpn)
1591 return 0;
1592
1593 /* Build path attribute for this route - use the source attr, if
1594 * present, else treat as locally originated.
1595 */
1596 if (src_attr)
1597 attr = *src_attr;
1598 else {
1599 memset(&attr, 0, sizeof(attr));
1600 bgp_attr_default_set(&attr, bgp_vrf, BGP_ORIGIN_IGP);
1601 }
1602
1603 /* Advertise Primary IP (PIP) is enabled, send individual
1604 * IP (default instance router-id) as nexthop.
1605 * PIP is disabled or vrr interface is not present
1606 * use anycast-IP as nexthop and anycast RMAC.
1607 */
1608 if (!bgp_vrf->evpn_info->advertise_pip ||
1609 (!bgp_vrf->evpn_info->is_anycast_mac)) {
1610 attr.nexthop = bgp_vrf->originator_ip;
1611 attr.mp_nexthop_global_in = bgp_vrf->originator_ip;
1612 memcpy(&attr.rmac, &bgp_vrf->rmac, ETH_ALEN);
1613 } else {
1614 /* copy sys rmac */
1615 memcpy(&attr.rmac, &bgp_vrf->evpn_info->pip_rmac, ETH_ALEN);
1616 if (bgp_vrf->evpn_info->pip_ip.s_addr != INADDR_ANY) {
1617 attr.nexthop = bgp_vrf->evpn_info->pip_ip;
1618 attr.mp_nexthop_global_in = bgp_vrf->evpn_info->pip_ip;
1619 } else if (bgp_vrf->evpn_info->pip_ip.s_addr == INADDR_ANY)
1620 if (bgp_debug_zebra(NULL))
1621 zlog_debug(
1622 "VRF %s evp %pFX advertise-pip primary ip is not configured",
1623 vrf_id_to_name(bgp_vrf->vrf_id), evp);
1624 }
1625
1626 if (bgp_debug_zebra(NULL))
1627 zlog_debug(
1628 "VRF %s type-5 route evp %pFX RMAC %pEA nexthop %pI4",
1629 vrf_id_to_name(bgp_vrf->vrf_id), evp, &attr.rmac,
1630 &attr.nexthop);
1631
1632 attr.mp_nexthop_len = BGP_ATTR_NHLEN_IPV4;
1633
1634 if (src_afi == AFI_IP6 &&
1635 CHECK_FLAG(bgp_vrf->af_flags[AFI_L2VPN][SAFI_EVPN],
1636 BGP_L2VPN_EVPN_ADV_IPV6_UNICAST_GW_IP)) {
1637 if (src_attr &&
1638 !IN6_IS_ADDR_UNSPECIFIED(&src_attr->mp_nexthop_global)) {
1639 attr.evpn_overlay.type = OVERLAY_INDEX_GATEWAY_IP;
1640 SET_IPADDR_V6(&attr.evpn_overlay.gw_ip);
1641 memcpy(&attr.evpn_overlay.gw_ip.ipaddr_v6,
1642 &src_attr->mp_nexthop_global,
1643 sizeof(struct in6_addr));
1644 }
1645 } else if (src_afi == AFI_IP &&
1646 CHECK_FLAG(bgp_vrf->af_flags[AFI_L2VPN][SAFI_EVPN],
1647 BGP_L2VPN_EVPN_ADV_IPV4_UNICAST_GW_IP)) {
1648 if (src_attr && src_attr->nexthop.s_addr != 0) {
1649 attr.evpn_overlay.type = OVERLAY_INDEX_GATEWAY_IP;
1650 SET_IPADDR_V4(&attr.evpn_overlay.gw_ip);
1651 memcpy(&attr.evpn_overlay.gw_ip.ipaddr_v4,
1652 &src_attr->nexthop, sizeof(struct in_addr));
1653 }
1654 }
1655
1656 /* Setup RT and encap extended community */
1657 build_evpn_type5_route_extcomm(bgp_vrf, &attr);
1658
1659 /* get the route node in global table */
1660 dest = bgp_evpn_global_node_get(bgp_evpn->rib[afi][safi], afi, safi,
1661 evp, &bgp_vrf->vrf_prd, NULL);
1662 assert(dest);
1663
1664 /* create or update the route entry within the route node */
1665 update_evpn_type5_route_entry(bgp_evpn, bgp_vrf, afi, safi, dest, &attr,
1666 &route_changed);
1667
1668 /* schedule for processing and unlock node */
1669 if (route_changed) {
1670 bgp_process(bgp_evpn, dest, afi, safi);
1671 bgp_dest_unlock_node(dest);
1672 }
1673
1674 /* uninten temporary */
1675 if (!src_attr)
1676 aspath_unintern(&attr.aspath);
1677 return 0;
1678 }
1679
1680 static void bgp_evpn_get_sync_info(struct bgp *bgp, esi_t *esi,
1681 struct bgp_dest *dest, uint32_t loc_seq,
1682 uint32_t *max_sync_seq, bool *active_on_peer,
1683 bool *peer_router, bool *proxy_from_peer,
1684 const struct ethaddr *mac)
1685 {
1686 struct bgp_path_info *tmp_pi;
1687 struct bgp_path_info *second_best_path = NULL;
1688 uint32_t tmp_mm_seq = 0;
1689 esi_t *tmp_esi;
1690 int paths_eq;
1691 struct ethaddr *tmp_mac;
1692 bool mac_cmp = false;
1693 struct prefix_evpn *evp = (struct prefix_evpn *)&dest->p;
1694
1695
1696 /* mac comparison is not needed for MAC-only routes */
1697 if (mac && !is_evpn_prefix_ipaddr_none(evp))
1698 mac_cmp = true;
1699
1700 /* find the best non-local path. a local path can only be present
1701 * as best path
1702 */
1703 for (tmp_pi = bgp_dest_get_bgp_path_info(dest); tmp_pi;
1704 tmp_pi = tmp_pi->next) {
1705 if (tmp_pi->sub_type != BGP_ROUTE_IMPORTED ||
1706 !CHECK_FLAG(tmp_pi->flags, BGP_PATH_VALID))
1707 continue;
1708
1709 /* ignore paths that have a different mac */
1710 if (mac_cmp) {
1711 tmp_mac = evpn_type2_path_info_get_mac(tmp_pi);
1712 if (memcmp(mac, tmp_mac, sizeof(*mac)))
1713 continue;
1714 }
1715
1716 if (bgp_evpn_path_info_cmp(bgp, tmp_pi,
1717 second_best_path, &paths_eq))
1718 second_best_path = tmp_pi;
1719 }
1720
1721 if (!second_best_path)
1722 return;
1723
1724 tmp_esi = bgp_evpn_attr_get_esi(second_best_path->attr);
1725 /* if this has the same ES desination as the local path
1726 * it is a sync path
1727 */
1728 if (!memcmp(esi, tmp_esi, sizeof(esi_t))) {
1729 tmp_mm_seq = mac_mobility_seqnum(second_best_path->attr);
1730 if (tmp_mm_seq < loc_seq)
1731 return;
1732
1733 /* we have a non-proxy path from the ES peer. */
1734 if (second_best_path->attr->es_flags &
1735 ATTR_ES_PROXY_ADVERT) {
1736 *proxy_from_peer = true;
1737 } else {
1738 *active_on_peer = true;
1739 }
1740
1741 if (second_best_path->attr->router_flag)
1742 *peer_router = true;
1743
1744 /* we use both proxy and non-proxy imports to
1745 * determine the max sync sequence
1746 */
1747 if (tmp_mm_seq > *max_sync_seq)
1748 *max_sync_seq = tmp_mm_seq;
1749 }
1750 }
1751
1752 /* Bubble up sync-info from all paths (non-best) to the local-path.
1753 * This is need for MM sequence number syncing and proxy advertisement.
1754 * Note: The local path can only exist as a best path in the
1755 * VPN route table. It will take precedence over all sync paths.
1756 */
1757 static void update_evpn_route_entry_sync_info(struct bgp *bgp,
1758 struct bgp_dest *dest,
1759 struct attr *attr,
1760 uint32_t loc_seq, bool setup_sync,
1761 const struct ethaddr *mac)
1762 {
1763 esi_t *esi;
1764 struct prefix_evpn *evp =
1765 (struct prefix_evpn *)bgp_dest_get_prefix(dest);
1766
1767 if (evp->prefix.route_type != BGP_EVPN_MAC_IP_ROUTE)
1768 return;
1769
1770 esi = bgp_evpn_attr_get_esi(attr);
1771 if (bgp_evpn_is_esi_valid(esi)) {
1772 if (setup_sync) {
1773 uint32_t max_sync_seq = 0;
1774 bool active_on_peer = false;
1775 bool peer_router = false;
1776 bool proxy_from_peer = false;
1777
1778 bgp_evpn_get_sync_info(bgp, esi, dest, loc_seq,
1779 &max_sync_seq, &active_on_peer,
1780 &peer_router, &proxy_from_peer,
1781 mac);
1782 attr->mm_sync_seqnum = max_sync_seq;
1783 if (active_on_peer)
1784 attr->es_flags |= ATTR_ES_PEER_ACTIVE;
1785 else
1786 attr->es_flags &= ~ATTR_ES_PEER_ACTIVE;
1787 if (proxy_from_peer)
1788 attr->es_flags |= ATTR_ES_PEER_PROXY;
1789 else
1790 attr->es_flags &= ~ATTR_ES_PEER_PROXY;
1791 if (peer_router)
1792 attr->es_flags |= ATTR_ES_PEER_ROUTER;
1793 else
1794 attr->es_flags &= ~ATTR_ES_PEER_ROUTER;
1795
1796 if (BGP_DEBUG(evpn_mh, EVPN_MH_RT)) {
1797 char esi_buf[ESI_STR_LEN];
1798
1799 zlog_debug(
1800 "setup sync info for %pFX es %s max_seq %d %s%s%s",
1801 evp,
1802 esi_to_str(esi, esi_buf,
1803 sizeof(esi_buf)),
1804 max_sync_seq,
1805 (attr->es_flags & ATTR_ES_PEER_ACTIVE)
1806 ? "peer-active "
1807 : "",
1808 (attr->es_flags & ATTR_ES_PEER_PROXY)
1809 ? "peer-proxy "
1810 : "",
1811 (attr->es_flags & ATTR_ES_PEER_ROUTER)
1812 ? "peer-router "
1813 : "");
1814 }
1815 }
1816 } else {
1817 attr->mm_sync_seqnum = 0;
1818 attr->es_flags &= ~ATTR_ES_PEER_ACTIVE;
1819 attr->es_flags &= ~ATTR_ES_PEER_PROXY;
1820 }
1821 }
1822
1823 /*
1824 * Create or update EVPN route entry. This could be in the VNI route tables
1825 * or the global route table.
1826 */
1827 static int update_evpn_route_entry(struct bgp *bgp, struct bgpevpn *vpn,
1828 afi_t afi, safi_t safi,
1829 struct bgp_dest *dest, struct attr *attr,
1830 const struct ethaddr *mac,
1831 const struct ipaddr *ip, int add,
1832 struct bgp_path_info **pi, uint8_t flags,
1833 uint32_t seq, bool vpn_rt, bool *old_is_sync)
1834 {
1835 struct bgp_path_info *tmp_pi;
1836 struct bgp_path_info *local_pi;
1837 struct attr *attr_new;
1838 struct attr local_attr;
1839 mpls_label_t label[BGP_MAX_LABELS];
1840 uint32_t num_labels = 1;
1841 int route_change = 1;
1842 uint8_t sticky = 0;
1843 const struct prefix_evpn *evp;
1844
1845 *pi = NULL;
1846 evp = (const struct prefix_evpn *)bgp_dest_get_prefix(dest);
1847 memset(&label, 0, sizeof(label));
1848
1849 /* See if this is an update of an existing route, or a new add. */
1850 local_pi = bgp_evpn_route_get_local_path(bgp, dest);
1851
1852 /* If route doesn't exist already, create a new one, if told to.
1853 * Otherwise act based on whether the attributes of the route have
1854 * changed or not.
1855 */
1856 if (!local_pi && !add)
1857 return 0;
1858
1859 if (old_is_sync && local_pi)
1860 *old_is_sync = bgp_evpn_attr_is_sync(local_pi->attr);
1861
1862 /* if a local path is being added with a non-zero esi look
1863 * for SYNC paths from ES peers and bubble up the sync-info
1864 */
1865 update_evpn_route_entry_sync_info(bgp, dest, attr, seq, vpn_rt, mac);
1866
1867 /* For non-GW MACs, update MAC mobility seq number, if needed. */
1868 if (seq && !CHECK_FLAG(flags, ZEBRA_MACIP_TYPE_GW))
1869 add_mac_mobility_to_attr(seq, attr);
1870
1871 if (!local_pi) {
1872 local_attr = *attr;
1873
1874 /* Extract MAC mobility sequence number, if any. */
1875 local_attr.mm_seqnum =
1876 bgp_attr_mac_mobility_seqnum(&local_attr, &sticky);
1877 local_attr.sticky = sticky;
1878
1879 /* Add (or update) attribute to hash. */
1880 attr_new = bgp_attr_intern(&local_attr);
1881
1882 /* Create new route with its attribute. */
1883 tmp_pi = info_make(ZEBRA_ROUTE_BGP, BGP_ROUTE_STATIC, 0,
1884 bgp->peer_self, attr_new, dest);
1885 SET_FLAG(tmp_pi->flags, BGP_PATH_VALID);
1886 bgp_path_info_extra_get(tmp_pi);
1887
1888 /* The VNI goes into the 'label' field of the route */
1889 vni2label(vpn->vni, &label[0]);
1890
1891 /* Type-2 routes may carry a second VNI - the L3-VNI.
1892 * Only attach second label if we are advertising two labels for
1893 * type-2 routes.
1894 */
1895 if (evp->prefix.route_type == BGP_EVPN_MAC_IP_ROUTE
1896 && CHECK_FLAG(vpn->flags, VNI_FLAG_USE_TWO_LABELS)) {
1897 vni_t l3vni;
1898
1899 l3vni = bgpevpn_get_l3vni(vpn);
1900 if (l3vni) {
1901 vni2label(l3vni, &label[1]);
1902 num_labels++;
1903 }
1904 }
1905
1906 memcpy(&tmp_pi->extra->label, label, sizeof(label));
1907 tmp_pi->extra->num_labels = num_labels;
1908
1909 if (evp->prefix.route_type == BGP_EVPN_MAC_IP_ROUTE) {
1910 if (mac)
1911 evpn_type2_path_info_set_mac(tmp_pi, *mac);
1912 else if (ip)
1913 evpn_type2_path_info_set_ip(tmp_pi, *ip);
1914 }
1915
1916 /* Mark route as self type-2 route */
1917 if (flags && CHECK_FLAG(flags, ZEBRA_MACIP_TYPE_SVI_IP))
1918 tmp_pi->extra->af_flags = BGP_EVPN_MACIP_TYPE_SVI_IP;
1919 bgp_path_info_add(dest, tmp_pi);
1920 } else {
1921 tmp_pi = local_pi;
1922 if (attrhash_cmp(tmp_pi->attr, attr)
1923 && !CHECK_FLAG(tmp_pi->flags, BGP_PATH_REMOVED))
1924 route_change = 0;
1925 else {
1926 /*
1927 * The attributes have changed, type-2 routes needs to
1928 * be advertised with right labels.
1929 */
1930 vni2label(vpn->vni, &label[0]);
1931 if (evp->prefix.route_type == BGP_EVPN_MAC_IP_ROUTE
1932 && CHECK_FLAG(vpn->flags,
1933 VNI_FLAG_USE_TWO_LABELS)) {
1934 vni_t l3vni;
1935
1936 l3vni = bgpevpn_get_l3vni(vpn);
1937 if (l3vni) {
1938 vni2label(l3vni, &label[1]);
1939 num_labels++;
1940 }
1941 }
1942 memcpy(&tmp_pi->extra->label, label, sizeof(label));
1943 tmp_pi->extra->num_labels = num_labels;
1944
1945 if (evp->prefix.route_type == BGP_EVPN_MAC_IP_ROUTE) {
1946 if (mac)
1947 evpn_type2_path_info_set_mac(tmp_pi,
1948 *mac);
1949 else if (ip)
1950 evpn_type2_path_info_set_ip(tmp_pi,
1951 *ip);
1952 }
1953
1954 /* The attribute has changed. */
1955 /* Add (or update) attribute to hash. */
1956 local_attr = *attr;
1957 bgp_path_info_set_flag(dest, tmp_pi,
1958 BGP_PATH_ATTR_CHANGED);
1959
1960 /* Extract MAC mobility sequence number, if any. */
1961 local_attr.mm_seqnum = bgp_attr_mac_mobility_seqnum(
1962 &local_attr, &sticky);
1963 local_attr.sticky = sticky;
1964
1965 attr_new = bgp_attr_intern(&local_attr);
1966
1967 /* Restore route, if needed. */
1968 if (CHECK_FLAG(tmp_pi->flags, BGP_PATH_REMOVED))
1969 bgp_path_info_restore(dest, tmp_pi);
1970
1971 /* Unintern existing, set to new. */
1972 bgp_attr_unintern(&tmp_pi->attr);
1973 tmp_pi->attr = attr_new;
1974 tmp_pi->uptime = monotime(NULL);
1975 }
1976 }
1977
1978 /* local MAC-IP routes in the VNI table are linked to
1979 * the destination ES
1980 */
1981 if (route_change && vpn_rt
1982 && (evp->prefix.route_type == BGP_EVPN_MAC_IP_ROUTE))
1983 bgp_evpn_path_es_link(tmp_pi, vpn->vni,
1984 bgp_evpn_attr_get_esi(tmp_pi->attr));
1985
1986 /* Return back the route entry. */
1987 *pi = tmp_pi;
1988 return route_change;
1989 }
1990
1991 static void evpn_zebra_reinstall_best_route(struct bgp *bgp,
1992 struct bgpevpn *vpn,
1993 struct bgp_dest *dest)
1994 {
1995 struct bgp_path_info *tmp_ri;
1996 struct bgp_path_info *curr_select = NULL;
1997
1998 for (tmp_ri = bgp_dest_get_bgp_path_info(dest); tmp_ri;
1999 tmp_ri = tmp_ri->next) {
2000 if (CHECK_FLAG(tmp_ri->flags, BGP_PATH_SELECTED)) {
2001 curr_select = tmp_ri;
2002 break;
2003 }
2004 }
2005
2006 if (curr_select && curr_select->type == ZEBRA_ROUTE_BGP
2007 && (curr_select->sub_type == BGP_ROUTE_IMPORTED ||
2008 bgp_evpn_attr_is_sync(curr_select->attr)))
2009 evpn_zebra_install(bgp, vpn,
2010 (const struct prefix_evpn *)bgp_dest_get_prefix(dest),
2011 curr_select);
2012 }
2013
2014 /*
2015 * If the local route was not selected evict it and tell zebra to re-add
2016 * the best remote dest.
2017 *
2018 * Typically a local path added by zebra is expected to be selected as
2019 * best. In which case when a remote path wins as best (later)
2020 * evpn_route_select_install itself evicts the older-local-best path.
2021 *
2022 * However if bgp's add and zebra's add cross paths (race condition) it
2023 * is possible that the local path is no longer the "older" best path.
2024 * It is a path that was never designated as best and hence requires
2025 * additional handling to prevent bgp from injecting and holding on to a
2026 * non-best local path.
2027 */
2028 static void evpn_cleanup_local_non_best_route(struct bgp *bgp,
2029 struct bgpevpn *vpn,
2030 struct bgp_dest *dest,
2031 struct bgp_path_info *local_pi)
2032 {
2033 /* local path was not picked as the winner; kick it out */
2034 if (bgp_debug_zebra(NULL))
2035 zlog_debug("evicting local evpn prefix %pBD as remote won",
2036 dest);
2037
2038 evpn_delete_old_local_route(bgp, vpn, dest, local_pi, NULL);
2039 bgp_path_info_reap(dest, local_pi);
2040
2041 /* tell zebra to re-add the best remote path */
2042 evpn_zebra_reinstall_best_route(bgp, vpn, dest);
2043 }
2044
2045 static inline bool bgp_evpn_route_add_l3_ecomm_ok(struct bgpevpn *vpn,
2046 const struct prefix_evpn *p,
2047 esi_t *esi)
2048 {
2049 return p->prefix.route_type == BGP_EVPN_MAC_IP_ROUTE
2050 && (is_evpn_prefix_ipaddr_v4(p)
2051 || (is_evpn_prefix_ipaddr_v6(p)
2052 && !IN6_IS_ADDR_LINKLOCAL(
2053 &p->prefix.macip_addr.ip.ipaddr_v6)))
2054 && CHECK_FLAG(vpn->flags, VNI_FLAG_USE_TWO_LABELS)
2055 && bgpevpn_get_l3vni(vpn) && bgp_evpn_es_add_l3_ecomm_ok(esi);
2056 }
2057
2058 /*
2059 * Create or update EVPN route (of type based on prefix) for specified VNI
2060 * and schedule for processing.
2061 */
2062 static int update_evpn_route(struct bgp *bgp, struct bgpevpn *vpn,
2063 struct prefix_evpn *p, uint8_t flags,
2064 uint32_t seq, esi_t *esi)
2065 {
2066 struct bgp_dest *dest;
2067 struct attr attr;
2068 struct attr *attr_new;
2069 int add_l3_ecomm = 0;
2070 struct bgp_path_info *pi;
2071 afi_t afi = AFI_L2VPN;
2072 safi_t safi = SAFI_EVPN;
2073 int route_change;
2074 bool old_is_sync = false;
2075 bool mac_only = false;
2076
2077 memset(&attr, 0, sizeof(attr));
2078
2079 /* Build path-attribute for this route. */
2080 bgp_attr_default_set(&attr, bgp, BGP_ORIGIN_IGP);
2081 attr.nexthop = vpn->originator_ip;
2082 attr.mp_nexthop_global_in = vpn->originator_ip;
2083 attr.mp_nexthop_len = BGP_ATTR_NHLEN_IPV4;
2084 attr.sticky = CHECK_FLAG(flags, ZEBRA_MACIP_TYPE_STICKY) ? 1 : 0;
2085 attr.default_gw = CHECK_FLAG(flags, ZEBRA_MACIP_TYPE_GW) ? 1 : 0;
2086 attr.router_flag = CHECK_FLAG(flags,
2087 ZEBRA_MACIP_TYPE_ROUTER_FLAG) ? 1 : 0;
2088 if (CHECK_FLAG(flags, ZEBRA_MACIP_TYPE_PROXY_ADVERT))
2089 attr.es_flags |= ATTR_ES_PROXY_ADVERT;
2090
2091 if (esi && bgp_evpn_is_esi_valid(esi)) {
2092 memcpy(&attr.esi, esi, sizeof(esi_t));
2093 attr.es_flags |= ATTR_ES_IS_LOCAL;
2094 }
2095
2096 /* PMSI is only needed for type-3 routes */
2097 if (p->prefix.route_type == BGP_EVPN_IMET_ROUTE) {
2098 attr.flag |= ATTR_FLAG_BIT(BGP_ATTR_PMSI_TUNNEL);
2099 bgp_attr_set_pmsi_tnl_type(&attr, PMSI_TNLTYPE_INGR_REPL);
2100 }
2101
2102 /* router mac is only needed for type-2 routes here. */
2103 if (p->prefix.route_type == BGP_EVPN_MAC_IP_ROUTE) {
2104 uint8_t af_flags = 0;
2105
2106 if (CHECK_FLAG(flags, ZEBRA_MACIP_TYPE_SVI_IP))
2107 SET_FLAG(af_flags, BGP_EVPN_MACIP_TYPE_SVI_IP);
2108
2109 bgp_evpn_get_rmac_nexthop(vpn, p, &attr, af_flags);
2110 }
2111
2112 if (bgp_debug_zebra(NULL)) {
2113 char buf3[ESI_STR_LEN];
2114
2115 zlog_debug(
2116 "VRF %s vni %u type-%u route evp %pFX RMAC %pEA nexthop %pI4 esi %s",
2117 vpn->bgp_vrf ? vrf_id_to_name(vpn->bgp_vrf->vrf_id)
2118 : "None",
2119 vpn->vni, p->prefix.route_type, p, &attr.rmac,
2120 &attr.mp_nexthop_global_in,
2121 esi_to_str(esi, buf3, sizeof(buf3)));
2122 }
2123
2124 vni2label(vpn->vni, &(attr.label));
2125
2126 /* Include L3 VNI related RTs and RMAC for type-2 routes, if they're
2127 * IPv4 or IPv6 global addresses and we're advertising L3VNI with
2128 * these routes.
2129 */
2130 add_l3_ecomm = bgp_evpn_route_add_l3_ecomm_ok(
2131 vpn, p, (attr.es_flags & ATTR_ES_IS_LOCAL) ? &attr.esi : NULL);
2132
2133 /* Set up extended community. */
2134 build_evpn_route_extcomm(vpn, &attr, add_l3_ecomm);
2135
2136 /* First, create (or fetch) route node within the VNI.
2137 * NOTE: There is no RD here.
2138 */
2139 dest = bgp_evpn_vni_node_get(vpn, p, NULL);
2140
2141 if ((p->prefix.route_type == BGP_EVPN_MAC_IP_ROUTE) &&
2142 (is_evpn_prefix_ipaddr_none(p) == true))
2143 mac_only = true;
2144
2145 /* Create or update route entry. */
2146 route_change = update_evpn_route_entry(
2147 bgp, vpn, afi, safi, dest, &attr,
2148 (mac_only ? NULL : &p->prefix.macip_addr.mac), NULL /* ip */, 1,
2149 &pi, flags, seq, true /* setup_sync */, &old_is_sync);
2150 assert(pi);
2151 attr_new = pi->attr;
2152
2153 /* lock ri to prevent freeing in evpn_route_select_install */
2154 bgp_path_info_lock(pi);
2155
2156 /* Perform route selection. Normally, the local route in the
2157 * VNI is expected to win and be the best route. However, if
2158 * there is a race condition where a host moved from local to
2159 * remote and the remote route was received in BGP just prior
2160 * to the local MACIP notification from zebra, the remote
2161 * route would win, and we should evict the defunct local route
2162 * and (re)install the remote route into zebra.
2163 */
2164 evpn_route_select_install(bgp, vpn, dest);
2165 /*
2166 * If the new local route was not selected evict it and tell zebra
2167 * to re-add the best remote dest. BGP doesn't retain non-best local
2168 * routes.
2169 */
2170 if (CHECK_FLAG(pi->flags, BGP_PATH_REMOVED)) {
2171 route_change = 0;
2172 } else {
2173 if (!CHECK_FLAG(pi->flags, BGP_PATH_SELECTED)) {
2174 route_change = 0;
2175 evpn_cleanup_local_non_best_route(bgp, vpn, dest, pi);
2176 } else {
2177 bool new_is_sync;
2178
2179 /* If the local path already existed and is still the
2180 * best path we need to also check if it transitioned
2181 * from being a sync path to a non-sync path. If it
2182 * it did we need to notify zebra that the sync-path
2183 * has been removed.
2184 */
2185 new_is_sync = bgp_evpn_attr_is_sync(pi->attr);
2186 if (!new_is_sync && old_is_sync)
2187 evpn_zebra_uninstall(bgp, vpn, p, pi, true);
2188 }
2189 }
2190 bgp_path_info_unlock(pi);
2191
2192 bgp_dest_unlock_node(dest);
2193
2194 /* If this is a new route or some attribute has changed, export the
2195 * route to the global table. The route will be advertised to peers
2196 * from there. Note that this table is a 2-level tree (RD-level +
2197 * Prefix-level) similar to L3VPN routes.
2198 */
2199 if (route_change) {
2200 struct bgp_path_info *global_pi;
2201
2202 dest = bgp_evpn_global_node_get(bgp->rib[afi][safi], afi, safi,
2203 p, &vpn->prd, NULL);
2204 update_evpn_route_entry(
2205 bgp, vpn, afi, safi, dest, attr_new, NULL /* mac */,
2206 NULL /* ip */, 1, &global_pi, flags, seq,
2207 false /* setup_sync */, NULL /* old_is_sync */);
2208
2209 /* Schedule for processing and unlock node. */
2210 bgp_process(bgp, dest, afi, safi);
2211 bgp_dest_unlock_node(dest);
2212 }
2213
2214 /* Unintern temporary. */
2215 aspath_unintern(&attr.aspath);
2216
2217 return 0;
2218 }
2219
2220 /*
2221 * Delete EVPN route entry.
2222 * The entry can be in ESI/VNI table or the global table.
2223 */
2224 void delete_evpn_route_entry(struct bgp *bgp, afi_t afi, safi_t safi,
2225 struct bgp_dest *dest,
2226 struct bgp_path_info **pi)
2227 {
2228 struct bgp_path_info *tmp_pi;
2229
2230 *pi = NULL;
2231
2232 /* Now, find matching route. */
2233 for (tmp_pi = bgp_dest_get_bgp_path_info(dest); tmp_pi;
2234 tmp_pi = tmp_pi->next)
2235 if (tmp_pi->peer == bgp->peer_self
2236 && tmp_pi->type == ZEBRA_ROUTE_BGP
2237 && tmp_pi->sub_type == BGP_ROUTE_STATIC)
2238 break;
2239
2240 *pi = tmp_pi;
2241
2242 /* Mark route for delete. */
2243 if (tmp_pi)
2244 bgp_path_info_delete(dest, tmp_pi);
2245 }
2246
2247 /* Delete EVPN type5 route */
2248 static int delete_evpn_type5_route(struct bgp *bgp_vrf, struct prefix_evpn *evp)
2249 {
2250 afi_t afi = AFI_L2VPN;
2251 safi_t safi = SAFI_EVPN;
2252 struct bgp_dest *dest = NULL;
2253 struct bgp_path_info *pi = NULL;
2254 struct bgp *bgp_evpn = NULL; /* evpn bgp instance */
2255
2256 bgp_evpn = bgp_get_evpn();
2257 if (!bgp_evpn)
2258 return 0;
2259
2260 /* locate the global route entry for this type-5 prefix */
2261 dest = bgp_evpn_global_node_lookup(bgp_evpn->rib[afi][safi], safi, evp,
2262 &bgp_vrf->vrf_prd, NULL);
2263 if (!dest)
2264 return 0;
2265
2266 delete_evpn_route_entry(bgp_evpn, afi, safi, dest, &pi);
2267 if (pi)
2268 bgp_process(bgp_evpn, dest, afi, safi);
2269 bgp_dest_unlock_node(dest);
2270 return 0;
2271 }
2272
2273 /*
2274 * Delete EVPN route (of type based on prefix) for specified VNI and
2275 * schedule for processing.
2276 */
2277 static int delete_evpn_route(struct bgp *bgp, struct bgpevpn *vpn,
2278 struct prefix_evpn *p)
2279 {
2280 struct bgp_dest *dest, *global_dest;
2281 struct bgp_path_info *pi;
2282 afi_t afi = AFI_L2VPN;
2283 safi_t safi = SAFI_EVPN;
2284
2285 /* First, locate the route node within the VNI. If it doesn't exist,
2286 * there
2287 * is nothing further to do.
2288 * NOTE: There is no RD here.
2289 */
2290 dest = bgp_evpn_vni_node_lookup(vpn, p, NULL);
2291 if (!dest)
2292 return 0;
2293
2294 /* Next, locate route node in the global EVPN routing table. Note that
2295 * this table is a 2-level tree (RD-level + Prefix-level) similar to
2296 * L3VPN routes.
2297 */
2298 global_dest = bgp_evpn_global_node_lookup(bgp->rib[afi][safi], safi, p,
2299 &vpn->prd, NULL);
2300 if (global_dest) {
2301 /* Delete route entry in the global EVPN table. */
2302 delete_evpn_route_entry(bgp, afi, safi, global_dest, &pi);
2303
2304 /* Schedule for processing - withdraws to peers happen from
2305 * this table.
2306 */
2307 if (pi)
2308 bgp_process(bgp, global_dest, afi, safi);
2309 bgp_dest_unlock_node(global_dest);
2310 }
2311
2312 /* Delete route entry in the VNI route table. This can just be removed.
2313 */
2314 delete_evpn_route_entry(bgp, afi, safi, dest, &pi);
2315 if (pi) {
2316 bgp_path_info_reap(dest, pi);
2317 evpn_route_select_install(bgp, vpn, dest);
2318 }
2319 bgp_dest_unlock_node(dest);
2320
2321 return 0;
2322 }
2323
2324 void bgp_evpn_update_type2_route_entry(struct bgp *bgp, struct bgpevpn *vpn,
2325 struct bgp_dest *dest,
2326 struct bgp_path_info *local_pi,
2327 const char *caller)
2328 {
2329 afi_t afi = AFI_L2VPN;
2330 safi_t safi = SAFI_EVPN;
2331 struct bgp_path_info *pi;
2332 struct attr attr;
2333 struct attr *attr_new;
2334 uint32_t seq;
2335 int add_l3_ecomm = 0;
2336 struct bgp_dest *global_dest;
2337 struct bgp_path_info *global_pi;
2338 struct prefix_evpn evp;
2339 int route_change;
2340 bool old_is_sync = false;
2341
2342 if (CHECK_FLAG(local_pi->flags, BGP_PATH_REMOVED))
2343 return;
2344
2345 /*
2346 * VNI table MAC-IP prefixes don't have MAC so make sure it's set from
2347 * path info here.
2348 */
2349 if (is_evpn_prefix_ipaddr_none((struct prefix_evpn *)&dest->p)) {
2350 /* VNI MAC -> Global */
2351 evpn_type2_prefix_global_copy(
2352 &evp, (struct prefix_evpn *)&dest->p, NULL /* mac */,
2353 evpn_type2_path_info_get_ip(local_pi));
2354 } else {
2355 /* VNI IP -> Global */
2356 evpn_type2_prefix_global_copy(
2357 &evp, (struct prefix_evpn *)&dest->p,
2358 evpn_type2_path_info_get_mac(local_pi), NULL /* ip */);
2359 }
2360
2361 /*
2362 * Build attribute per local route as the MAC mobility and
2363 * some other values could differ for different routes. The
2364 * attributes will be shared in the hash table.
2365 */
2366 bgp_attr_default_set(&attr, bgp, BGP_ORIGIN_IGP);
2367 attr.nexthop = vpn->originator_ip;
2368 attr.mp_nexthop_global_in = vpn->originator_ip;
2369 attr.mp_nexthop_len = BGP_ATTR_NHLEN_IPV4;
2370 attr.sticky = (local_pi->attr->sticky) ? 1 : 0;
2371 attr.router_flag = (local_pi->attr->router_flag) ? 1 : 0;
2372 attr.es_flags = local_pi->attr->es_flags;
2373 if (local_pi->attr->default_gw) {
2374 attr.default_gw = 1;
2375 if (is_evpn_prefix_ipaddr_v6(&evp))
2376 attr.router_flag = 1;
2377 }
2378 memcpy(&attr.esi, &local_pi->attr->esi, sizeof(esi_t));
2379 bgp_evpn_get_rmac_nexthop(vpn, &evp, &attr, local_pi->extra->af_flags);
2380 vni2label(vpn->vni, &(attr.label));
2381 /* Add L3 VNI RTs and RMAC for non IPv6 link-local if
2382 * using L3 VNI for type-2 routes also.
2383 */
2384 add_l3_ecomm = bgp_evpn_route_add_l3_ecomm_ok(
2385 vpn, &evp,
2386 (attr.es_flags & ATTR_ES_IS_LOCAL) ? &attr.esi : NULL);
2387
2388 /* Set up extended community. */
2389 build_evpn_route_extcomm(vpn, &attr, add_l3_ecomm);
2390 seq = mac_mobility_seqnum(local_pi->attr);
2391
2392 if (bgp_debug_zebra(NULL)) {
2393 char buf3[ESI_STR_LEN];
2394
2395 zlog_debug(
2396 "VRF %s vni %u evp %pFX RMAC %pEA nexthop %pI4 esi %s esf 0x%x from %s",
2397 vpn->bgp_vrf ? vrf_id_to_name(vpn->bgp_vrf->vrf_id)
2398 : " ",
2399 vpn->vni, &evp, &attr.rmac, &attr.mp_nexthop_global_in,
2400 esi_to_str(&attr.esi, buf3, sizeof(buf3)),
2401 attr.es_flags, caller);
2402 }
2403
2404 /* Update the route entry. */
2405 route_change = update_evpn_route_entry(
2406 bgp, vpn, afi, safi, dest, &attr, NULL /* mac */, NULL /* ip */,
2407 0, &pi, 0, seq, true /* setup_sync */, &old_is_sync);
2408
2409 assert(pi);
2410 attr_new = pi->attr;
2411 /* lock ri to prevent freeing in evpn_route_select_install */
2412 bgp_path_info_lock(pi);
2413
2414 /* Perform route selection. Normally, the local route in the
2415 * VNI is expected to win and be the best route. However,
2416 * under peculiar situations (e.g., tunnel (next hop) IP change
2417 * that causes best selection to be based on next hop), a
2418 * remote route could win. If the local route is the best,
2419 * ensure it is updated in the global EVPN route table and
2420 * advertised to peers; otherwise, ensure it is evicted and
2421 * (re)install the remote route into zebra.
2422 */
2423 evpn_route_select_install(bgp, vpn, dest);
2424
2425 if (CHECK_FLAG(pi->flags, BGP_PATH_REMOVED)) {
2426 route_change = 0;
2427 } else {
2428 if (!CHECK_FLAG(pi->flags, BGP_PATH_SELECTED)) {
2429 route_change = 0;
2430 evpn_cleanup_local_non_best_route(bgp, vpn, dest, pi);
2431 } else {
2432 bool new_is_sync;
2433
2434 /* If the local path already existed and is still the
2435 * best path we need to also check if it transitioned
2436 * from being a sync path to a non-sync path. If it
2437 * it did we need to notify zebra that the sync-path
2438 * has been removed.
2439 */
2440 new_is_sync = bgp_evpn_attr_is_sync(pi->attr);
2441 if (!new_is_sync && old_is_sync)
2442 evpn_zebra_uninstall(bgp, vpn, &evp, pi, true);
2443 }
2444 }
2445
2446
2447 /* unlock pi */
2448 bgp_path_info_unlock(pi);
2449
2450 if (route_change) {
2451 /* Update route in global routing table. */
2452 global_dest = bgp_evpn_global_node_get(
2453 bgp->rib[afi][safi], afi, safi, &evp, &vpn->prd, NULL);
2454 assert(global_dest);
2455 update_evpn_route_entry(
2456 bgp, vpn, afi, safi, global_dest, attr_new,
2457 NULL /* mac */, NULL /* ip */, 0, &global_pi, 0,
2458 mac_mobility_seqnum(attr_new), false /* setup_sync */,
2459 NULL /* old_is_sync */);
2460
2461 /* Schedule for processing and unlock node. */
2462 bgp_process(bgp, global_dest, afi, safi);
2463 bgp_dest_unlock_node(global_dest);
2464 }
2465
2466 /* Unintern temporary. */
2467 aspath_unintern(&attr.aspath);
2468 }
2469
2470 static void update_type2_route(struct bgp *bgp, struct bgpevpn *vpn,
2471 struct bgp_dest *dest)
2472 {
2473 struct bgp_path_info *tmp_pi;
2474
2475 const struct prefix_evpn *evp =
2476 (const struct prefix_evpn *)bgp_dest_get_prefix(dest);
2477
2478 if (evp->prefix.route_type != BGP_EVPN_MAC_IP_ROUTE)
2479 return;
2480
2481 /* Identify local route. */
2482 for (tmp_pi = bgp_dest_get_bgp_path_info(dest); tmp_pi;
2483 tmp_pi = tmp_pi->next) {
2484 if (tmp_pi->peer == bgp->peer_self &&
2485 tmp_pi->type == ZEBRA_ROUTE_BGP &&
2486 tmp_pi->sub_type == BGP_ROUTE_STATIC)
2487 break;
2488 }
2489
2490 if (!tmp_pi)
2491 return;
2492
2493 bgp_evpn_update_type2_route_entry(bgp, vpn, dest, tmp_pi, __func__);
2494 }
2495
2496 /*
2497 * Update all type-2 (MACIP) local routes for this VNI - these should also
2498 * be scheduled for advertise to peers.
2499 */
2500 static void update_all_type2_routes(struct bgp *bgp, struct bgpevpn *vpn)
2501 {
2502 struct bgp_dest *dest;
2503
2504 /* Walk this VNI's route MAC & IP table and update local type-2
2505 * routes. For any routes updated, update corresponding entry in the
2506 * global table too.
2507 */
2508 for (dest = bgp_table_top(vpn->mac_table); dest;
2509 dest = bgp_route_next(dest))
2510 update_type2_route(bgp, vpn, dest);
2511
2512 for (dest = bgp_table_top(vpn->ip_table); dest;
2513 dest = bgp_route_next(dest))
2514 update_type2_route(bgp, vpn, dest);
2515 }
2516
2517 /*
2518 * Delete all type-2 (MACIP) local routes for this VNI - only from the
2519 * global routing table. These are also scheduled for withdraw from peers.
2520 */
2521 static void delete_global_type2_routes(struct bgp *bgp, struct bgpevpn *vpn)
2522 {
2523 afi_t afi;
2524 safi_t safi;
2525 struct bgp_dest *rddest, *dest;
2526 struct bgp_table *table;
2527 struct bgp_path_info *pi;
2528
2529 afi = AFI_L2VPN;
2530 safi = SAFI_EVPN;
2531
2532 rddest = bgp_node_lookup(bgp->rib[afi][safi],
2533 (struct prefix *)&vpn->prd);
2534 if (rddest) {
2535 table = bgp_dest_get_bgp_table_info(rddest);
2536 for (dest = bgp_table_top(table); dest;
2537 dest = bgp_route_next(dest)) {
2538 const struct prefix_evpn *evp =
2539 (const struct prefix_evpn *)bgp_dest_get_prefix(
2540 dest);
2541
2542 if (evp->prefix.route_type != BGP_EVPN_MAC_IP_ROUTE)
2543 continue;
2544
2545 delete_evpn_route_entry(bgp, afi, safi, dest, &pi);
2546 if (pi)
2547 bgp_process(bgp, dest, afi, safi);
2548 }
2549
2550 /* Unlock RD node. */
2551 bgp_dest_unlock_node(rddest);
2552 }
2553 }
2554
2555 static void delete_vni_type2_route(struct bgp *bgp, struct bgp_dest *dest)
2556 {
2557 struct bgp_path_info *pi;
2558 afi_t afi = AFI_L2VPN;
2559 safi_t safi = SAFI_EVPN;
2560
2561 const struct prefix_evpn *evp =
2562 (const struct prefix_evpn *)bgp_dest_get_prefix(dest);
2563
2564 if (evp->prefix.route_type != BGP_EVPN_MAC_IP_ROUTE)
2565 return;
2566
2567 delete_evpn_route_entry(bgp, afi, safi, dest, &pi);
2568
2569 /* Route entry in local table gets deleted immediately. */
2570 if (pi)
2571 bgp_path_info_reap(dest, pi);
2572 }
2573
2574 static void delete_vni_type2_routes(struct bgp *bgp, struct bgpevpn *vpn)
2575 {
2576 struct bgp_dest *dest;
2577
2578 /* Next, walk this VNI's MAC & IP route table and delete local type-2
2579 * routes.
2580 */
2581 for (dest = bgp_table_top(vpn->mac_table); dest;
2582 dest = bgp_route_next(dest))
2583 delete_vni_type2_route(bgp, dest);
2584
2585 for (dest = bgp_table_top(vpn->ip_table); dest;
2586 dest = bgp_route_next(dest))
2587 delete_vni_type2_route(bgp, dest);
2588 }
2589
2590 /*
2591 * Delete all type-2 (MACIP) local routes for this VNI - from the global
2592 * table as well as the per-VNI route table.
2593 */
2594 static void delete_all_type2_routes(struct bgp *bgp, struct bgpevpn *vpn)
2595 {
2596 /* First, walk the global route table for this VNI's type-2 local
2597 * routes.
2598 * EVPN routes are a 2-level table, first get the RD table.
2599 */
2600 delete_global_type2_routes(bgp, vpn);
2601 delete_vni_type2_routes(bgp, vpn);
2602 }
2603
2604 /*
2605 * Delete all routes in the per-VNI route table.
2606 */
2607 static void delete_all_vni_routes(struct bgp *bgp, struct bgpevpn *vpn)
2608 {
2609 struct bgp_dest *dest;
2610 struct bgp_path_info *pi, *nextpi;
2611
2612 /* Walk this VNI's MAC & IP route table and delete all routes. */
2613 for (dest = bgp_table_top(vpn->mac_table); dest;
2614 dest = bgp_route_next(dest)) {
2615 for (pi = bgp_dest_get_bgp_path_info(dest);
2616 (pi != NULL) && (nextpi = pi->next, 1); pi = nextpi) {
2617 bgp_evpn_remote_ip_hash_del(vpn, pi);
2618 bgp_path_info_delete(dest, pi);
2619 bgp_path_info_reap(dest, pi);
2620 }
2621 }
2622
2623 for (dest = bgp_table_top(vpn->ip_table); dest;
2624 dest = bgp_route_next(dest)) {
2625 for (pi = bgp_dest_get_bgp_path_info(dest);
2626 (pi != NULL) && (nextpi = pi->next, 1); pi = nextpi) {
2627 bgp_path_info_delete(dest, pi);
2628 bgp_path_info_reap(dest, pi);
2629 }
2630 }
2631 }
2632
2633 /* BUM traffic flood mode per-l2-vni */
2634 static int bgp_evpn_vni_flood_mode_get(struct bgp *bgp,
2635 struct bgpevpn *vpn)
2636 {
2637 /* if flooding has been globally disabled per-vni mode is
2638 * not relevant
2639 */
2640 if (bgp->vxlan_flood_ctrl == VXLAN_FLOOD_DISABLED)
2641 return VXLAN_FLOOD_DISABLED;
2642
2643 /* if mcast group ip has been specified we use a PIM-SM MDT */
2644 if (vpn->mcast_grp.s_addr != INADDR_ANY)
2645 return VXLAN_FLOOD_PIM_SM;
2646
2647 /* default is ingress replication */
2648 return VXLAN_FLOOD_HEAD_END_REPL;
2649 }
2650
2651 /*
2652 * Update (and advertise) local routes for a VNI. Invoked upon the VNI
2653 * export RT getting modified or change to tunnel IP. Note that these
2654 * situations need the route in the per-VNI table as well as the global
2655 * table to be updated (as attributes change).
2656 */
2657 int update_routes_for_vni(struct bgp *bgp, struct bgpevpn *vpn)
2658 {
2659 int ret;
2660 struct prefix_evpn p;
2661
2662 update_type1_routes_for_evi(bgp, vpn);
2663
2664 /* Update and advertise the type-3 route (only one) followed by the
2665 * locally learnt type-2 routes (MACIP) - for this VNI.
2666 *
2667 * RT-3 only if doing head-end replication
2668 */
2669 if (bgp_evpn_vni_flood_mode_get(bgp, vpn)
2670 == VXLAN_FLOOD_HEAD_END_REPL) {
2671 build_evpn_type3_prefix(&p, vpn->originator_ip);
2672 ret = update_evpn_route(bgp, vpn, &p, 0, 0, NULL);
2673 if (ret)
2674 return ret;
2675 }
2676
2677 update_all_type2_routes(bgp, vpn);
2678 return 0;
2679 }
2680
2681 /*
2682 * Delete (and withdraw) local routes for specified VNI from the global
2683 * table and per-VNI table. After this, remove all other routes from
2684 * the per-VNI table. Invoked upon the VNI being deleted or EVPN
2685 * (advertise-all-vni) being disabled.
2686 */
2687 static int delete_routes_for_vni(struct bgp *bgp, struct bgpevpn *vpn)
2688 {
2689 int ret;
2690 struct prefix_evpn p;
2691
2692 /* Delete and withdraw locally learnt type-2 routes (MACIP)
2693 * followed by type-3 routes (only one) - for this VNI.
2694 */
2695 delete_all_type2_routes(bgp, vpn);
2696
2697 build_evpn_type3_prefix(&p, vpn->originator_ip);
2698 ret = delete_evpn_route(bgp, vpn, &p);
2699 if (ret)
2700 return ret;
2701
2702 /* Delete all routes from the per-VNI table. */
2703 delete_all_vni_routes(bgp, vpn);
2704 return 0;
2705 }
2706
2707 /*
2708 * There is a flood mcast IP address change. Update the mcast-grp and
2709 * remove the type-3 route if any. A new type-3 route will be generated
2710 * post tunnel_ip update if the new flood mode is head-end-replication.
2711 */
2712 static int bgp_evpn_mcast_grp_change(struct bgp *bgp, struct bgpevpn *vpn,
2713 struct in_addr mcast_grp)
2714 {
2715 struct prefix_evpn p;
2716
2717 vpn->mcast_grp = mcast_grp;
2718
2719 if (is_vni_live(vpn)) {
2720 build_evpn_type3_prefix(&p, vpn->originator_ip);
2721 delete_evpn_route(bgp, vpn, &p);
2722 }
2723
2724 return 0;
2725 }
2726
2727 /*
2728 * There is a tunnel endpoint IP address change for this VNI, delete
2729 * prior type-3 route (if needed) and update.
2730 * Note: Route re-advertisement happens elsewhere after other processing
2731 * other changes.
2732 */
2733 static void handle_tunnel_ip_change(struct bgp *bgp, struct bgpevpn *vpn,
2734 struct in_addr originator_ip)
2735 {
2736 struct prefix_evpn p;
2737
2738 if (IPV4_ADDR_SAME(&vpn->originator_ip, &originator_ip))
2739 return;
2740
2741 /* If VNI is not live, we only need to update the originator ip */
2742 if (!is_vni_live(vpn)) {
2743 vpn->originator_ip = originator_ip;
2744 return;
2745 }
2746
2747 /* Update the tunnel-ip hash */
2748 bgp_tip_del(bgp, &vpn->originator_ip);
2749 if (bgp_tip_add(bgp, &originator_ip))
2750 /* The originator_ip was not already present in the
2751 * bgp martian next-hop table as a tunnel-ip, so we
2752 * need to go back and filter routes matching the new
2753 * martian next-hop.
2754 */
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;
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], safi, &p,
4319 &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 = 0;
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 bgp_update(peer, (struct prefix *)&p, addpath_id, attr, afi,
4520 safi, ZEBRA_ROUTE_BGP, BGP_ROUTE_NORMAL, &prd,
4521 &label[0], num_labels, 0, &evpn);
4522 else
4523 bgp_withdraw(peer, (struct prefix *)&p, addpath_id, afi, safi,
4524 ZEBRA_ROUTE_BGP, BGP_ROUTE_NORMAL, &prd, &label[0],
4525 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
4551 /* Type-3 route should be either 17 or 29 bytes: RD (8), Eth Tag (4),
4552 * IP len (1) and IP (4 or 16).
4553 */
4554 if (psize != 17 && psize != 29) {
4555 flog_err(EC_BGP_EVPN_ROUTE_INVALID,
4556 "%u:%s - Rx EVPN Type-3 NLRI with invalid length %d",
4557 peer->bgp->vrf_id, peer->host, psize);
4558 return -1;
4559 }
4560
4561 /* If PMSI is present, log if it is anything other than IR.
4562 * Note: We just simply ignore the values as it is not clear if
4563 * doing anything else is better.
4564 */
4565 if (attr &&
4566 (attr->flag & ATTR_FLAG_BIT(BGP_ATTR_PMSI_TUNNEL))) {
4567 enum pta_type pmsi_tnl_type = bgp_attr_get_pmsi_tnl_type(attr);
4568
4569 if (pmsi_tnl_type != PMSI_TNLTYPE_INGR_REPL
4570 && pmsi_tnl_type != PMSI_TNLTYPE_PIM_SM) {
4571 flog_warn(
4572 EC_BGP_EVPN_PMSI_PRESENT,
4573 "%u:%s - Rx EVPN Type-3 NLRI with unsupported PTA %d",
4574 peer->bgp->vrf_id, peer->host, pmsi_tnl_type);
4575 }
4576 }
4577
4578 /* Make prefix_rd */
4579 prd.family = AF_UNSPEC;
4580 prd.prefixlen = 64;
4581 memcpy(&prd.val, pfx, 8);
4582 pfx += 8;
4583
4584 /* Make EVPN prefix. */
4585 memset(&p, 0, sizeof(p));
4586 p.family = AF_EVPN;
4587 p.prefixlen = EVPN_ROUTE_PREFIXLEN;
4588 p.prefix.route_type = BGP_EVPN_IMET_ROUTE;
4589
4590 /* Copy Ethernet Tag */
4591 memcpy(&eth_tag, pfx, 4);
4592 p.prefix.imet_addr.eth_tag = ntohl(eth_tag);
4593 pfx += 4;
4594
4595 /* Get the IP. */
4596 ipaddr_len = *pfx++;
4597 if (ipaddr_len == IPV4_MAX_BITLEN) {
4598 p.prefix.imet_addr.ip.ipa_type = IPADDR_V4;
4599 memcpy(&p.prefix.imet_addr.ip.ip.addr, pfx, IPV4_MAX_BYTELEN);
4600 } else {
4601 flog_err(
4602 EC_BGP_EVPN_ROUTE_INVALID,
4603 "%u:%s - Rx EVPN Type-3 NLRI with unsupported IP address length %d",
4604 peer->bgp->vrf_id, peer->host, ipaddr_len);
4605 return -1;
4606 }
4607
4608 /* Process the route. */
4609 if (attr)
4610 bgp_update(peer, (struct prefix *)&p, addpath_id, attr, afi,
4611 safi, ZEBRA_ROUTE_BGP, BGP_ROUTE_NORMAL, &prd, NULL,
4612 0, 0, NULL);
4613 else
4614 bgp_withdraw(peer, (struct prefix *)&p, addpath_id, afi, safi,
4615 ZEBRA_ROUTE_BGP, BGP_ROUTE_NORMAL, &prd, NULL, 0,
4616 NULL);
4617 return 0;
4618 }
4619
4620 /*
4621 * Process received EVPN type-5 route (advertise or withdraw).
4622 */
4623 static int process_type5_route(struct peer *peer, afi_t afi, safi_t safi,
4624 struct attr *attr, uint8_t *pfx, int psize,
4625 uint32_t addpath_id)
4626 {
4627 struct prefix_rd prd;
4628 struct prefix_evpn p;
4629 struct bgp_route_evpn evpn;
4630 uint8_t ippfx_len;
4631 uint32_t eth_tag;
4632 mpls_label_t label; /* holds the VNI as in the packet */
4633 bool is_valid_update = true;
4634
4635 /* Type-5 route should be 34 or 58 bytes:
4636 * RD (8), ESI (10), Eth Tag (4), IP len (1), IP (4 or 16),
4637 * GW (4 or 16) and VNI (3).
4638 * Note that the IP and GW should both be IPv4 or both IPv6.
4639 */
4640 if (psize != 34 && psize != 58) {
4641 flog_err(EC_BGP_EVPN_ROUTE_INVALID,
4642 "%u:%s - Rx EVPN Type-5 NLRI with invalid length %d",
4643 peer->bgp->vrf_id, peer->host, psize);
4644 return -1;
4645 }
4646
4647 /* Make prefix_rd */
4648 prd.family = AF_UNSPEC;
4649 prd.prefixlen = 64;
4650 memcpy(&prd.val, pfx, 8);
4651 pfx += 8;
4652
4653 /* Make EVPN prefix. */
4654 memset(&p, 0, sizeof(p));
4655 p.family = AF_EVPN;
4656 p.prefixlen = EVPN_ROUTE_PREFIXLEN;
4657 p.prefix.route_type = BGP_EVPN_IP_PREFIX_ROUTE;
4658
4659 /* Additional information outside of prefix - ESI and GW IP */
4660 memset(&evpn, 0, sizeof(evpn));
4661
4662 /* Fetch ESI overlay index */
4663 if (attr)
4664 memcpy(&evpn.eth_s_id, pfx, sizeof(esi_t));
4665 pfx += ESI_BYTES;
4666
4667 /* Fetch Ethernet Tag. */
4668 memcpy(&eth_tag, pfx, 4);
4669 p.prefix.prefix_addr.eth_tag = ntohl(eth_tag);
4670 pfx += 4;
4671
4672 /* Fetch IP prefix length. */
4673 ippfx_len = *pfx++;
4674 if (ippfx_len > IPV6_MAX_BITLEN) {
4675 flog_err(
4676 EC_BGP_EVPN_ROUTE_INVALID,
4677 "%u:%s - Rx EVPN Type-5 NLRI with invalid IP Prefix length %d",
4678 peer->bgp->vrf_id, peer->host, ippfx_len);
4679 return -1;
4680 }
4681 p.prefix.prefix_addr.ip_prefix_length = ippfx_len;
4682
4683 /* Determine IPv4 or IPv6 prefix */
4684 /* Since the address and GW are from the same family, this just becomes
4685 * a simple check on the total size.
4686 */
4687 if (psize == 34) {
4688 SET_IPADDR_V4(&p.prefix.prefix_addr.ip);
4689 memcpy(&p.prefix.prefix_addr.ip.ipaddr_v4, pfx, 4);
4690 pfx += 4;
4691 SET_IPADDR_V4(&evpn.gw_ip);
4692 memcpy(&evpn.gw_ip.ipaddr_v4, pfx, 4);
4693 pfx += 4;
4694 } else {
4695 SET_IPADDR_V6(&p.prefix.prefix_addr.ip);
4696 memcpy(&p.prefix.prefix_addr.ip.ipaddr_v6, pfx,
4697 IPV6_MAX_BYTELEN);
4698 pfx += IPV6_MAX_BYTELEN;
4699 SET_IPADDR_V6(&evpn.gw_ip);
4700 memcpy(&evpn.gw_ip.ipaddr_v6, pfx, IPV6_MAX_BYTELEN);
4701 pfx += IPV6_MAX_BYTELEN;
4702 }
4703
4704 /* Get the VNI (in MPLS label field). Stored as bytes here. */
4705 memset(&label, 0, sizeof(label));
4706 memcpy(&label, pfx, BGP_LABEL_BYTES);
4707
4708 /*
4709 * If in future, we are required to access additional fields,
4710 * we MUST increment pfx by BGP_LABEL_BYTES in before reading the next
4711 * field
4712 */
4713
4714 /*
4715 * An update containing a non-zero gateway IP and a non-zero ESI
4716 * at the same time is should be treated as withdraw
4717 */
4718 if (bgp_evpn_is_esi_valid(&evpn.eth_s_id) &&
4719 !ipaddr_is_zero(&evpn.gw_ip)) {
4720 flog_err(EC_BGP_EVPN_ROUTE_INVALID,
4721 "%s - Rx EVPN Type-5 ESI and gateway-IP both non-zero.",
4722 peer->host);
4723 is_valid_update = false;
4724 } else if (bgp_evpn_is_esi_valid(&evpn.eth_s_id))
4725 evpn.type = OVERLAY_INDEX_ESI;
4726 else if (!ipaddr_is_zero(&evpn.gw_ip))
4727 evpn.type = OVERLAY_INDEX_GATEWAY_IP;
4728 if (attr) {
4729 if (is_zero_mac(&attr->rmac) &&
4730 !bgp_evpn_is_esi_valid(&evpn.eth_s_id) &&
4731 ipaddr_is_zero(&evpn.gw_ip) && label == 0) {
4732 flog_err(EC_BGP_EVPN_ROUTE_INVALID,
4733 "%s - Rx EVPN Type-5 ESI, gateway-IP, RMAC and label all zero",
4734 peer->host);
4735 is_valid_update = false;
4736 }
4737
4738 if (is_mcast_mac(&attr->rmac) || is_bcast_mac(&attr->rmac))
4739 is_valid_update = false;
4740 }
4741
4742 /* Process the route. */
4743 if (attr && is_valid_update)
4744 bgp_update(peer, (struct prefix *)&p, addpath_id, attr, afi,
4745 safi, ZEBRA_ROUTE_BGP, BGP_ROUTE_NORMAL, &prd,
4746 &label, 1, 0, &evpn);
4747 else {
4748 if (!is_valid_update) {
4749 char attr_str[BUFSIZ] = {0};
4750
4751 bgp_dump_attr(attr, attr_str, BUFSIZ);
4752 zlog_warn(
4753 "Invalid update from peer %s vrf %u prefix %pFX attr %s - treat as withdraw",
4754 peer->hostname, peer->bgp->vrf_id, &p,
4755 attr_str);
4756 }
4757 bgp_withdraw(peer, (struct prefix *)&p, addpath_id, afi, safi,
4758 ZEBRA_ROUTE_BGP, BGP_ROUTE_NORMAL, &prd, &label, 1,
4759 &evpn);
4760 }
4761
4762 return 0;
4763 }
4764
4765 static void evpn_mpattr_encode_type5(struct stream *s, const struct prefix *p,
4766 const struct prefix_rd *prd,
4767 mpls_label_t *label, uint32_t num_labels,
4768 struct attr *attr)
4769 {
4770 int len;
4771 char temp[16];
4772 const struct evpn_addr *p_evpn_p;
4773
4774 memset(&temp, 0, sizeof(temp));
4775 if (p->family != AF_EVPN)
4776 return;
4777 p_evpn_p = &(p->u.prefix_evpn);
4778
4779 /* len denites the total len of IP and GW-IP in the route
4780 IP and GW-IP have to be both ipv4 or ipv6
4781 */
4782 if (IS_IPADDR_V4(&p_evpn_p->prefix_addr.ip))
4783 len = 8; /* IP and GWIP are both ipv4 */
4784 else
4785 len = 32; /* IP and GWIP are both ipv6 */
4786 /* Prefix contains RD, ESI, EthTag, IP length, IP, GWIP and VNI */
4787 stream_putc(s, 8 + 10 + 4 + 1 + len + 3);
4788 stream_put(s, prd->val, 8);
4789 if (attr && attr->evpn_overlay.type == OVERLAY_INDEX_ESI)
4790 stream_put(s, &attr->esi, sizeof(esi_t));
4791 else
4792 stream_put(s, 0, sizeof(esi_t));
4793 stream_putl(s, p_evpn_p->prefix_addr.eth_tag);
4794 stream_putc(s, p_evpn_p->prefix_addr.ip_prefix_length);
4795 if (IS_IPADDR_V4(&p_evpn_p->prefix_addr.ip))
4796 stream_put_ipv4(s, p_evpn_p->prefix_addr.ip.ipaddr_v4.s_addr);
4797 else
4798 stream_put(s, &p_evpn_p->prefix_addr.ip.ipaddr_v6, 16);
4799 if (attr && attr->evpn_overlay.type == OVERLAY_INDEX_GATEWAY_IP) {
4800 const struct bgp_route_evpn *evpn_overlay =
4801 bgp_attr_get_evpn_overlay(attr);
4802
4803 if (IS_IPADDR_V4(&p_evpn_p->prefix_addr.ip))
4804 stream_put_ipv4(s,
4805 evpn_overlay->gw_ip.ipaddr_v4.s_addr);
4806 else
4807 stream_put(s, &(evpn_overlay->gw_ip.ipaddr_v6), 16);
4808 } else {
4809 if (IS_IPADDR_V4(&p_evpn_p->prefix_addr.ip))
4810 stream_put_ipv4(s, 0);
4811 else
4812 stream_put(s, &temp, 16);
4813 }
4814
4815 if (num_labels)
4816 stream_put(s, label, 3);
4817 else
4818 stream_put3(s, 0);
4819 }
4820
4821 /*
4822 * Cleanup specific VNI upon EVPN (advertise-all-vni) being disabled.
4823 */
4824 static void cleanup_vni_on_disable(struct hash_bucket *bucket, struct bgp *bgp)
4825 {
4826 struct bgpevpn *vpn = (struct bgpevpn *)bucket->data;
4827
4828 /* Remove EVPN routes and schedule for processing. */
4829 delete_routes_for_vni(bgp, vpn);
4830
4831 /* Clear "live" flag and see if hash needs to be freed. */
4832 UNSET_FLAG(vpn->flags, VNI_FLAG_LIVE);
4833 if (!is_vni_configured(vpn))
4834 bgp_evpn_free(bgp, vpn);
4835 }
4836
4837 /*
4838 * Free a VNI entry; iterator function called during cleanup.
4839 */
4840 static void free_vni_entry(struct hash_bucket *bucket, struct bgp *bgp)
4841 {
4842 struct bgpevpn *vpn = (struct bgpevpn *)bucket->data;
4843
4844 delete_all_vni_routes(bgp, vpn);
4845 bgp_evpn_free(bgp, vpn);
4846 }
4847
4848 /*
4849 * Derive AUTO import RT for BGP VRF - L3VNI
4850 */
4851 static void evpn_auto_rt_import_add_for_vrf(struct bgp *bgp_vrf)
4852 {
4853 struct bgp *bgp_evpn = NULL;
4854
4855 form_auto_rt(bgp_vrf, bgp_vrf->l3vni, bgp_vrf->vrf_import_rtl, true);
4856
4857 /* Map RT to VRF */
4858 bgp_evpn = bgp_get_evpn();
4859
4860 if (!bgp_evpn)
4861 return;
4862
4863 bgp_evpn_map_vrf_to_its_rts(bgp_vrf);
4864 }
4865
4866 /*
4867 * Delete AUTO import RT from BGP VRF - L3VNI
4868 */
4869 static void evpn_auto_rt_import_delete_for_vrf(struct bgp *bgp_vrf)
4870 {
4871 evpn_rt_delete_auto(bgp_vrf, bgp_vrf->l3vni, bgp_vrf->vrf_import_rtl,
4872 true);
4873 }
4874
4875 /*
4876 * Derive AUTO export RT for BGP VRF - L3VNI
4877 */
4878 static void evpn_auto_rt_export_add_for_vrf(struct bgp *bgp_vrf)
4879 {
4880 form_auto_rt(bgp_vrf, bgp_vrf->l3vni, bgp_vrf->vrf_export_rtl, true);
4881 }
4882
4883 /*
4884 * Delete AUTO export RT from BGP VRF - L3VNI
4885 */
4886 static void evpn_auto_rt_export_delete_for_vrf(struct bgp *bgp_vrf)
4887 {
4888 evpn_rt_delete_auto(bgp_vrf, bgp_vrf->l3vni, bgp_vrf->vrf_export_rtl,
4889 true);
4890 }
4891
4892 static void bgp_evpn_handle_export_rt_change_for_vrf(struct bgp *bgp_vrf)
4893 {
4894 struct bgp *bgp_evpn = NULL;
4895 struct listnode *node = NULL;
4896 struct bgpevpn *vpn = NULL;
4897
4898 bgp_evpn = bgp_get_evpn();
4899 if (!bgp_evpn)
4900 return;
4901
4902 /* update all type-5 routes */
4903 update_advertise_vrf_routes(bgp_vrf);
4904
4905 /* update all type-2 routes */
4906 for (ALL_LIST_ELEMENTS_RO(bgp_vrf->l2vnis, node, vpn))
4907 update_routes_for_vni(bgp_evpn, vpn);
4908 }
4909
4910 /*
4911 * Handle autort change for a given VNI.
4912 */
4913 static void update_autort_vni(struct hash_bucket *bucket, struct bgp *bgp)
4914 {
4915 struct bgpevpn *vpn = bucket->data;
4916
4917 if (!is_import_rt_configured(vpn)) {
4918 if (is_vni_live(vpn))
4919 bgp_evpn_uninstall_routes(bgp, vpn);
4920 bgp_evpn_unmap_vni_from_its_rts(bgp, vpn);
4921 list_delete_all_node(vpn->import_rtl);
4922 bgp_evpn_derive_auto_rt_import(bgp, vpn);
4923 if (is_vni_live(vpn))
4924 bgp_evpn_install_routes(bgp, vpn);
4925 }
4926 if (!is_export_rt_configured(vpn)) {
4927 list_delete_all_node(vpn->export_rtl);
4928 bgp_evpn_derive_auto_rt_export(bgp, vpn);
4929 if (is_vni_live(vpn))
4930 bgp_evpn_handle_export_rt_change(bgp, vpn);
4931 }
4932 }
4933
4934 /*
4935 * Handle autort change for L3VNI.
4936 */
4937 static void update_autort_l3vni(struct bgp *bgp)
4938 {
4939 if ((CHECK_FLAG(bgp->vrf_flags, BGP_VRF_IMPORT_RT_CFGD))
4940 && (CHECK_FLAG(bgp->vrf_flags, BGP_VRF_EXPORT_RT_CFGD)))
4941 return;
4942
4943 if (!CHECK_FLAG(bgp->vrf_flags, BGP_VRF_IMPORT_RT_CFGD)) {
4944 if (is_l3vni_live(bgp))
4945 uninstall_routes_for_vrf(bgp);
4946
4947 /* Cleanup the RT to VRF mapping */
4948 bgp_evpn_unmap_vrf_from_its_rts(bgp);
4949
4950 /* Remove auto generated RT */
4951 evpn_auto_rt_import_delete_for_vrf(bgp);
4952
4953 list_delete_all_node(bgp->vrf_import_rtl);
4954
4955 /* Map auto derive or configured RTs */
4956 evpn_auto_rt_import_add_for_vrf(bgp);
4957 }
4958
4959 if (!CHECK_FLAG(bgp->vrf_flags, BGP_VRF_EXPORT_RT_CFGD)) {
4960 list_delete_all_node(bgp->vrf_export_rtl);
4961
4962 evpn_auto_rt_export_delete_for_vrf(bgp);
4963
4964 evpn_auto_rt_export_add_for_vrf(bgp);
4965
4966 if (is_l3vni_live(bgp))
4967 bgp_evpn_map_vrf_to_its_rts(bgp);
4968 }
4969
4970 if (!is_l3vni_live(bgp))
4971 return;
4972
4973 /* advertise type-5 routes if needed */
4974 update_advertise_vrf_routes(bgp);
4975
4976 /* install all remote routes belonging to this l3vni
4977 * into corresponding vrf
4978 */
4979 install_routes_for_vrf(bgp);
4980 }
4981
4982 /*
4983 * Public functions.
4984 */
4985
4986 /* withdraw type-5 route corresponding to ip prefix */
4987 void bgp_evpn_withdraw_type5_route(struct bgp *bgp_vrf, const struct prefix *p,
4988 afi_t afi, safi_t safi)
4989 {
4990 int ret = 0;
4991 struct prefix_evpn evp;
4992
4993 build_type5_prefix_from_ip_prefix(&evp, p);
4994 ret = delete_evpn_type5_route(bgp_vrf, &evp);
4995 if (ret)
4996 flog_err(
4997 EC_BGP_EVPN_ROUTE_DELETE,
4998 "%u failed to delete type-5 route for prefix %pFX in vrf %s",
4999 bgp_vrf->vrf_id, p, vrf_id_to_name(bgp_vrf->vrf_id));
5000 }
5001
5002 /* withdraw all type-5 routes for an address family */
5003 void bgp_evpn_withdraw_type5_routes(struct bgp *bgp_vrf, afi_t afi, safi_t safi)
5004 {
5005 struct bgp_table *table = NULL;
5006 struct bgp_dest *dest = NULL;
5007 struct bgp_path_info *pi;
5008
5009 table = bgp_vrf->rib[afi][safi];
5010 for (dest = bgp_table_top(table); dest; dest = bgp_route_next(dest)) {
5011 /* Only care about "selected" routes. Also ensure that
5012 * these are routes that are injectable into EVPN.
5013 */
5014 /* TODO: Support for AddPath for EVPN. */
5015 for (pi = bgp_dest_get_bgp_path_info(dest); pi; pi = pi->next) {
5016 if (CHECK_FLAG(pi->flags, BGP_PATH_SELECTED)
5017 && is_route_injectable_into_evpn(pi)) {
5018 bgp_evpn_withdraw_type5_route(
5019 bgp_vrf, bgp_dest_get_prefix(dest), afi,
5020 safi);
5021 break;
5022 }
5023 }
5024 }
5025 }
5026
5027 /*
5028 * evpn - enable advertisement of default g/w
5029 */
5030 void bgp_evpn_install_uninstall_default_route(struct bgp *bgp_vrf, afi_t afi,
5031 safi_t safi, bool add)
5032 {
5033 struct prefix ip_prefix;
5034
5035 /* form the default prefix 0.0.0.0/0 */
5036 memset(&ip_prefix, 0, sizeof(ip_prefix));
5037 ip_prefix.family = afi2family(afi);
5038
5039 if (add) {
5040 bgp_evpn_advertise_type5_route(bgp_vrf, &ip_prefix,
5041 NULL, afi, safi);
5042 } else {
5043 bgp_evpn_withdraw_type5_route(bgp_vrf, &ip_prefix,
5044 afi, safi);
5045 }
5046 }
5047
5048
5049 /*
5050 * Advertise IP prefix as type-5 route. The afi/safi and src_attr passed
5051 * to this function correspond to those of the source IP prefix (best
5052 * path in the case of the attr. In the case of a local prefix (when we
5053 * are advertising local subnets), the src_attr will be NULL.
5054 */
5055 void bgp_evpn_advertise_type5_route(struct bgp *bgp_vrf, const struct prefix *p,
5056 struct attr *src_attr, afi_t afi,
5057 safi_t safi)
5058 {
5059 int ret = 0;
5060 struct prefix_evpn evp;
5061
5062 build_type5_prefix_from_ip_prefix(&evp, p);
5063 ret = update_evpn_type5_route(bgp_vrf, &evp, src_attr, afi, safi);
5064 if (ret)
5065 flog_err(EC_BGP_EVPN_ROUTE_CREATE,
5066 "%u: Failed to create type-5 route for prefix %pFX",
5067 bgp_vrf->vrf_id, p);
5068 }
5069
5070 /* Inject all prefixes of a particular address-family (currently, IPv4 or
5071 * IPv6 unicast) into EVPN as type-5 routes. This is invoked when the
5072 * advertisement is enabled.
5073 */
5074 void bgp_evpn_advertise_type5_routes(struct bgp *bgp_vrf, afi_t afi,
5075 safi_t safi)
5076 {
5077 struct bgp_table *table = NULL;
5078 struct bgp_dest *dest = NULL;
5079 struct bgp_path_info *pi;
5080
5081 table = bgp_vrf->rib[afi][safi];
5082 for (dest = bgp_table_top(table); dest; dest = bgp_route_next(dest)) {
5083 /* Need to identify the "selected" route entry to use its
5084 * attribute. Also, ensure that the route is injectable
5085 * into EVPN.
5086 * TODO: Support for AddPath for EVPN.
5087 */
5088 for (pi = bgp_dest_get_bgp_path_info(dest); pi; pi = pi->next) {
5089 if (CHECK_FLAG(pi->flags, BGP_PATH_SELECTED)
5090 && is_route_injectable_into_evpn(pi)) {
5091
5092 /* apply the route-map */
5093 if (bgp_vrf->adv_cmd_rmap[afi][safi].map) {
5094 route_map_result_t ret;
5095 struct bgp_path_info tmp_pi;
5096 struct bgp_path_info_extra tmp_pie;
5097 struct attr tmp_attr;
5098
5099 tmp_attr = *pi->attr;
5100
5101 /* Fill temp path_info */
5102 prep_for_rmap_apply(&tmp_pi, &tmp_pie,
5103 dest, pi, pi->peer,
5104 &tmp_attr);
5105
5106 RESET_FLAG(tmp_attr.rmap_change_flags);
5107
5108 ret = route_map_apply(
5109 bgp_vrf->adv_cmd_rmap[afi][safi]
5110 .map,
5111 bgp_dest_get_prefix(dest),
5112 &tmp_pi);
5113 if (ret == RMAP_DENYMATCH) {
5114 bgp_attr_flush(&tmp_attr);
5115 continue;
5116 }
5117 bgp_evpn_advertise_type5_route(
5118 bgp_vrf,
5119 bgp_dest_get_prefix(dest),
5120 &tmp_attr, afi, safi);
5121 } else
5122 bgp_evpn_advertise_type5_route(
5123 bgp_vrf,
5124 bgp_dest_get_prefix(dest),
5125 pi->attr, afi, safi);
5126 break;
5127 }
5128 }
5129 }
5130 }
5131
5132 static void rt_list_remove_node(struct list *rt_list,
5133 struct ecommunity *ecomdel, bool is_l3)
5134 {
5135 struct listnode *node = NULL, *nnode = NULL, *node_to_del = NULL;
5136 struct vrf_route_target *l3rt = NULL;
5137 struct ecommunity *ecom = NULL;
5138
5139 if (is_l3) {
5140 for (ALL_LIST_ELEMENTS(rt_list, node, nnode, l3rt)) {
5141 if (ecommunity_match(l3rt->ecom, ecomdel)) {
5142 evpn_vrf_rt_del(l3rt);
5143 node_to_del = node;
5144 break;
5145 }
5146 }
5147 } else {
5148 for (ALL_LIST_ELEMENTS(rt_list, node, nnode, ecom)) {
5149 if (ecommunity_match(ecom, ecomdel)) {
5150 ecommunity_free(&ecom);
5151 node_to_del = node;
5152 break;
5153 }
5154 }
5155 }
5156
5157
5158 if (node_to_del)
5159 list_delete_node(rt_list, node_to_del);
5160 }
5161
5162 void evpn_rt_delete_auto(struct bgp *bgp, vni_t vni, struct list *rtl,
5163 bool is_l3)
5164 {
5165 struct ecommunity *ecom_auto;
5166 struct ecommunity_val eval;
5167
5168 if (bgp->advertise_autort_rfc8365)
5169 vni |= EVPN_AUTORT_VXLAN;
5170
5171 encode_route_target_as((bgp->as & 0xFFFF), vni, &eval);
5172
5173 ecom_auto = ecommunity_new();
5174 ecommunity_add_val(ecom_auto, &eval, false, false);
5175
5176 rt_list_remove_node(rtl, ecom_auto, is_l3);
5177
5178 ecommunity_free(&ecom_auto);
5179 }
5180
5181 static void evpn_vrf_rt_routes_map(struct bgp *bgp_vrf)
5182 {
5183 /* map VRFs to its RTs and install routes matching this new RT */
5184 if (is_l3vni_live(bgp_vrf)) {
5185 bgp_evpn_map_vrf_to_its_rts(bgp_vrf);
5186 install_routes_for_vrf(bgp_vrf);
5187 }
5188 }
5189
5190 static void evpn_vrf_rt_routes_unmap(struct bgp *bgp_vrf)
5191 {
5192 /* uninstall routes from vrf */
5193 if (is_l3vni_live(bgp_vrf))
5194 uninstall_routes_for_vrf(bgp_vrf);
5195
5196 /* Cleanup the RT to VRF mapping */
5197 bgp_evpn_unmap_vrf_from_its_rts(bgp_vrf);
5198 }
5199
5200 static bool rt_list_has_cfgd_rt(struct list *rt_list)
5201 {
5202 struct listnode *node = NULL, *nnode = NULL;
5203 struct vrf_route_target *l3rt = NULL;
5204
5205 for (ALL_LIST_ELEMENTS(rt_list, node, nnode, l3rt)) {
5206 if (!CHECK_FLAG(l3rt->flags, BGP_VRF_RT_AUTO))
5207 return true;
5208 }
5209
5210 return false;
5211 }
5212
5213 static void unconfigure_import_rt_for_vrf_fini(struct bgp *bgp_vrf)
5214 {
5215 if (!bgp_vrf->vrf_import_rtl)
5216 return; /* this should never fail */
5217
5218 if (!is_l3vni_live(bgp_vrf))
5219 return; /* Nothing to do if no vni */
5220
5221 /* fall back to auto-generated RT if this was the last RT */
5222 if (list_isempty(bgp_vrf->vrf_import_rtl))
5223 evpn_auto_rt_import_add_for_vrf(bgp_vrf);
5224 }
5225
5226 static void unconfigure_export_rt_for_vrf_fini(struct bgp *bgp_vrf)
5227 {
5228
5229 if (!bgp_vrf->vrf_export_rtl)
5230 return; /* this should never fail */
5231
5232 if (!is_l3vni_live(bgp_vrf))
5233 return; /* Nothing to do if no vni */
5234
5235 /* fall back to auto-generated RT if this was the last RT */
5236 if (list_isempty(bgp_vrf->vrf_export_rtl))
5237 evpn_auto_rt_export_add_for_vrf(bgp_vrf);
5238
5239 bgp_evpn_handle_export_rt_change_for_vrf(bgp_vrf);
5240 }
5241
5242 void bgp_evpn_configure_import_rt_for_vrf(struct bgp *bgp_vrf,
5243 struct ecommunity *ecomadd,
5244 bool is_wildcard)
5245 {
5246 struct vrf_route_target *newrt;
5247
5248 newrt = evpn_vrf_rt_new(ecomadd);
5249
5250 if (is_wildcard)
5251 SET_FLAG(newrt->flags, BGP_VRF_RT_WILD);
5252
5253 evpn_vrf_rt_routes_unmap(bgp_vrf);
5254
5255 /* Remove auto generated RT if not configured */
5256 if (!CHECK_FLAG(bgp_vrf->vrf_flags, BGP_VRF_IMPORT_AUTO_RT_CFGD))
5257 evpn_auto_rt_import_delete_for_vrf(bgp_vrf);
5258
5259 /* Add the newly configured RT to RT list */
5260 listnode_add_sort(bgp_vrf->vrf_import_rtl, newrt);
5261
5262 SET_FLAG(bgp_vrf->vrf_flags, BGP_VRF_IMPORT_RT_CFGD);
5263
5264 evpn_vrf_rt_routes_map(bgp_vrf);
5265 }
5266
5267 void bgp_evpn_configure_import_auto_rt_for_vrf(struct bgp *bgp_vrf)
5268 {
5269 if (CHECK_FLAG(bgp_vrf->vrf_flags, BGP_VRF_IMPORT_AUTO_RT_CFGD))
5270 return; /* Already configured */
5271
5272 SET_FLAG(bgp_vrf->vrf_flags, BGP_VRF_IMPORT_AUTO_RT_CFGD);
5273
5274 if (!is_l3vni_live(bgp_vrf))
5275 return; /* Wait for VNI before adding rts */
5276
5277 evpn_vrf_rt_routes_unmap(bgp_vrf);
5278
5279 evpn_auto_rt_import_add_for_vrf(bgp_vrf);
5280
5281 evpn_vrf_rt_routes_map(bgp_vrf);
5282 }
5283
5284 void bgp_evpn_unconfigure_import_rt_for_vrf(struct bgp *bgp_vrf,
5285 struct ecommunity *ecomdel)
5286 {
5287 if (!CHECK_FLAG(bgp_vrf->vrf_flags, BGP_VRF_IMPORT_RT_CFGD))
5288 return; /* Already un-configured */
5289
5290 evpn_vrf_rt_routes_unmap(bgp_vrf);
5291
5292 /* Remove rt */
5293 rt_list_remove_node(bgp_vrf->vrf_import_rtl, ecomdel, true);
5294
5295 if (!rt_list_has_cfgd_rt(bgp_vrf->vrf_import_rtl))
5296 UNSET_FLAG(bgp_vrf->vrf_flags, BGP_VRF_IMPORT_RT_CFGD);
5297
5298 unconfigure_import_rt_for_vrf_fini(bgp_vrf);
5299
5300 evpn_vrf_rt_routes_map(bgp_vrf);
5301 }
5302
5303 void bgp_evpn_unconfigure_import_auto_rt_for_vrf(struct bgp *bgp_vrf)
5304 {
5305 if (!CHECK_FLAG(bgp_vrf->vrf_flags, BGP_VRF_IMPORT_AUTO_RT_CFGD))
5306 return; /* Already un-configured */
5307
5308 evpn_vrf_rt_routes_unmap(bgp_vrf);
5309
5310 /* remove auto-generated RT */
5311 evpn_auto_rt_import_delete_for_vrf(bgp_vrf);
5312
5313 UNSET_FLAG(bgp_vrf->vrf_flags, BGP_VRF_IMPORT_AUTO_RT_CFGD);
5314
5315 unconfigure_import_rt_for_vrf_fini(bgp_vrf);
5316
5317 evpn_vrf_rt_routes_map(bgp_vrf);
5318 }
5319
5320 void bgp_evpn_configure_export_rt_for_vrf(struct bgp *bgp_vrf,
5321 struct ecommunity *ecomadd)
5322 {
5323 struct vrf_route_target *newrt;
5324
5325 newrt = evpn_vrf_rt_new(ecomadd);
5326
5327 /* Remove auto generated RT if not configured */
5328 if (!CHECK_FLAG(bgp_vrf->vrf_flags, BGP_VRF_EXPORT_AUTO_RT_CFGD))
5329 evpn_auto_rt_export_delete_for_vrf(bgp_vrf);
5330
5331 /* Add the new RT to the RT list */
5332 listnode_add_sort(bgp_vrf->vrf_export_rtl, newrt);
5333
5334 SET_FLAG(bgp_vrf->vrf_flags, BGP_VRF_EXPORT_RT_CFGD);
5335
5336 if (is_l3vni_live(bgp_vrf))
5337 bgp_evpn_handle_export_rt_change_for_vrf(bgp_vrf);
5338 }
5339
5340 void bgp_evpn_configure_export_auto_rt_for_vrf(struct bgp *bgp_vrf)
5341 {
5342 if (CHECK_FLAG(bgp_vrf->vrf_flags, BGP_VRF_EXPORT_AUTO_RT_CFGD))
5343 return; /* Already configured */
5344
5345 SET_FLAG(bgp_vrf->vrf_flags, BGP_VRF_EXPORT_AUTO_RT_CFGD);
5346
5347 if (!is_l3vni_live(bgp_vrf))
5348 return; /* Wait for VNI before adding rts */
5349
5350 evpn_auto_rt_export_add_for_vrf(bgp_vrf);
5351
5352 bgp_evpn_handle_export_rt_change_for_vrf(bgp_vrf);
5353 }
5354
5355 void bgp_evpn_unconfigure_export_rt_for_vrf(struct bgp *bgp_vrf,
5356 struct ecommunity *ecomdel)
5357 {
5358 if (!CHECK_FLAG(bgp_vrf->vrf_flags, BGP_VRF_EXPORT_RT_CFGD))
5359 return; /* Already un-configured */
5360
5361 /* Remove rt */
5362 rt_list_remove_node(bgp_vrf->vrf_export_rtl, ecomdel, true);
5363
5364 if (!rt_list_has_cfgd_rt(bgp_vrf->vrf_export_rtl))
5365 UNSET_FLAG(bgp_vrf->vrf_flags, BGP_VRF_EXPORT_RT_CFGD);
5366
5367 unconfigure_export_rt_for_vrf_fini(bgp_vrf);
5368 }
5369
5370 void bgp_evpn_unconfigure_export_auto_rt_for_vrf(struct bgp *bgp_vrf)
5371 {
5372 if (!CHECK_FLAG(bgp_vrf->vrf_flags, BGP_VRF_EXPORT_AUTO_RT_CFGD))
5373 return; /* Already un-configured */
5374
5375 /* remove auto-generated RT */
5376 evpn_auto_rt_export_delete_for_vrf(bgp_vrf);
5377
5378 UNSET_FLAG(bgp_vrf->vrf_flags, BGP_VRF_EXPORT_AUTO_RT_CFGD);
5379
5380 unconfigure_export_rt_for_vrf_fini(bgp_vrf);
5381 }
5382
5383 /*
5384 * Handle change to BGP router id. This is invoked twice by the change
5385 * handler, first before the router id has been changed and then after
5386 * the router id has been changed. The first invocation will result in
5387 * local routes for all VNIs/VRF being deleted and withdrawn and the next
5388 * will result in the routes being re-advertised.
5389 */
5390 void bgp_evpn_handle_router_id_update(struct bgp *bgp, int withdraw)
5391 {
5392 struct listnode *node;
5393 struct bgp *bgp_vrf;
5394
5395 if (withdraw) {
5396
5397 /* delete and withdraw all the type-5 routes
5398 stored in the global table for this vrf
5399 */
5400 withdraw_router_id_vrf(bgp);
5401
5402 /* delete all the VNI routes (type-2/type-3) routes for all the
5403 * L2-VNIs
5404 */
5405 hash_iterate(bgp->vnihash,
5406 (void (*)(struct hash_bucket *,
5407 void *))withdraw_router_id_vni,
5408 bgp);
5409
5410 if (bgp->inst_type == BGP_INSTANCE_TYPE_DEFAULT) {
5411 for (ALL_LIST_ELEMENTS_RO(bm->bgp, node, bgp_vrf)) {
5412 if (bgp_vrf->evpn_info->advertise_pip &&
5413 (bgp_vrf->evpn_info->pip_ip_static.s_addr
5414 == INADDR_ANY))
5415 bgp_vrf->evpn_info->pip_ip.s_addr
5416 = INADDR_ANY;
5417 }
5418 }
5419 } else {
5420
5421 /* Assign new default instance router-id */
5422 if (bgp->inst_type == BGP_INSTANCE_TYPE_DEFAULT) {
5423 for (ALL_LIST_ELEMENTS_RO(bm->bgp, node, bgp_vrf)) {
5424 if (bgp_vrf->evpn_info->advertise_pip &&
5425 (bgp_vrf->evpn_info->pip_ip_static.s_addr
5426 == INADDR_ANY)) {
5427 bgp_vrf->evpn_info->pip_ip =
5428 bgp->router_id;
5429 /* advertise type-5 routes with
5430 * new nexthop
5431 */
5432 update_advertise_vrf_routes(bgp_vrf);
5433 }
5434 }
5435 }
5436
5437 /* advertise all routes in the vrf as type-5 routes with the new
5438 * RD
5439 */
5440 update_router_id_vrf(bgp);
5441
5442 /* advertise all the VNI routes (type-2/type-3) routes with the
5443 * new RD
5444 */
5445 hash_iterate(bgp->vnihash,
5446 (void (*)(struct hash_bucket *,
5447 void *))update_router_id_vni,
5448 bgp);
5449 }
5450 }
5451
5452 /*
5453 * Handle change to auto-RT algorithm - update and advertise local routes.
5454 */
5455 void bgp_evpn_handle_autort_change(struct bgp *bgp)
5456 {
5457 hash_iterate(bgp->vnihash,
5458 (void (*)(struct hash_bucket *,
5459 void*))update_autort_vni,
5460 bgp);
5461 if (bgp->l3vni)
5462 update_autort_l3vni(bgp);
5463 }
5464
5465 /*
5466 * Handle change to export RT - update and advertise local routes.
5467 */
5468 int bgp_evpn_handle_export_rt_change(struct bgp *bgp, struct bgpevpn *vpn)
5469 {
5470 return update_routes_for_vni(bgp, vpn);
5471 }
5472
5473 void bgp_evpn_handle_vrf_rd_change(struct bgp *bgp_vrf, int withdraw)
5474 {
5475 if (withdraw)
5476 delete_withdraw_vrf_routes(bgp_vrf);
5477 else
5478 update_advertise_vrf_routes(bgp_vrf);
5479 }
5480
5481 /*
5482 * Handle change to RD. This is invoked twice by the change handler,
5483 * first before the RD has been changed and then after the RD has
5484 * been changed. The first invocation will result in local routes
5485 * of this VNI being deleted and withdrawn and the next will result
5486 * in the routes being re-advertised.
5487 */
5488 void bgp_evpn_handle_rd_change(struct bgp *bgp, struct bgpevpn *vpn,
5489 int withdraw)
5490 {
5491 if (withdraw)
5492 delete_withdraw_vni_routes(bgp, vpn);
5493 else
5494 update_advertise_vni_routes(bgp, vpn);
5495 }
5496
5497 /*
5498 * Install routes for this VNI. Invoked upon change to Import RT.
5499 */
5500 int bgp_evpn_install_routes(struct bgp *bgp, struct bgpevpn *vpn)
5501 {
5502 return install_routes_for_vni(bgp, vpn);
5503 }
5504
5505 /*
5506 * Uninstall all routes installed for this VNI. Invoked upon change
5507 * to Import RT.
5508 */
5509 int bgp_evpn_uninstall_routes(struct bgp *bgp, struct bgpevpn *vpn)
5510 {
5511 return uninstall_routes_for_vni(bgp, vpn);
5512 }
5513
5514 /*
5515 * TODO: Hardcoded for a maximum of 2 VNIs right now
5516 */
5517 char *bgp_evpn_label2str(mpls_label_t *label, uint32_t num_labels, char *buf,
5518 int len)
5519 {
5520 vni_t vni1, vni2;
5521
5522 vni1 = label2vni(label);
5523 if (num_labels == 2) {
5524 vni2 = label2vni(label + 1);
5525 snprintf(buf, len, "%u/%u", vni1, vni2);
5526 } else
5527 snprintf(buf, len, "%u", vni1);
5528 return buf;
5529 }
5530
5531 /*
5532 * Function to convert evpn route to json format.
5533 * NOTE: We don't use prefix2str as the output here is a bit different.
5534 */
5535 void bgp_evpn_route2json(const struct prefix_evpn *p, json_object *json)
5536 {
5537 char buf1[ETHER_ADDR_STRLEN];
5538 char buf2[PREFIX2STR_BUFFER];
5539 uint8_t family;
5540 uint8_t prefixlen;
5541
5542 if (!json)
5543 return;
5544
5545 json_object_int_add(json, "routeType", p->prefix.route_type);
5546
5547 switch (p->prefix.route_type) {
5548 case BGP_EVPN_MAC_IP_ROUTE:
5549 json_object_int_add(json, "ethTag",
5550 p->prefix.macip_addr.eth_tag);
5551 json_object_int_add(json, "macLen", 8 * ETH_ALEN);
5552 json_object_string_add(json, "mac",
5553 prefix_mac2str(&p->prefix.macip_addr.mac, buf1,
5554 sizeof(buf1)));
5555
5556 if (!is_evpn_prefix_ipaddr_none(p)) {
5557 family = is_evpn_prefix_ipaddr_v4(p) ? AF_INET :
5558 AF_INET6;
5559 prefixlen = (family == AF_INET) ?
5560 IPV4_MAX_BITLEN : IPV6_MAX_BITLEN;
5561 inet_ntop(family, &p->prefix.macip_addr.ip.ip.addr,
5562 buf2, PREFIX2STR_BUFFER);
5563 json_object_int_add(json, "ipLen", prefixlen);
5564 json_object_string_add(json, "ip", buf2);
5565 }
5566 break;
5567
5568 case BGP_EVPN_IMET_ROUTE:
5569 json_object_int_add(json, "ethTag",
5570 p->prefix.imet_addr.eth_tag);
5571 family = is_evpn_prefix_ipaddr_v4(p) ? AF_INET : AF_INET6;
5572 prefixlen = (family == AF_INET) ? IPV4_MAX_BITLEN :
5573 IPV6_MAX_BITLEN;
5574 inet_ntop(family, &p->prefix.imet_addr.ip.ip.addr, buf2,
5575 PREFIX2STR_BUFFER);
5576 json_object_int_add(json, "ipLen", prefixlen);
5577 json_object_string_add(json, "ip", buf2);
5578 break;
5579
5580 case BGP_EVPN_IP_PREFIX_ROUTE:
5581 json_object_int_add(json, "ethTag",
5582 p->prefix.prefix_addr.eth_tag);
5583 family = is_evpn_prefix_ipaddr_v4(p) ? AF_INET : AF_INET6;
5584 inet_ntop(family, &p->prefix.prefix_addr.ip.ip.addr,
5585 buf2, sizeof(buf2));
5586 json_object_int_add(json, "ipLen",
5587 p->prefix.prefix_addr.ip_prefix_length);
5588 json_object_string_add(json, "ip", buf2);
5589 break;
5590
5591 default:
5592 break;
5593 }
5594 }
5595
5596 /*
5597 * Encode EVPN prefix in Update (MP_REACH)
5598 */
5599 void bgp_evpn_encode_prefix(struct stream *s, const struct prefix *p,
5600 const struct prefix_rd *prd, mpls_label_t *label,
5601 uint32_t num_labels, struct attr *attr,
5602 bool addpath_capable, uint32_t addpath_tx_id)
5603 {
5604 struct prefix_evpn *evp = (struct prefix_evpn *)p;
5605 int len, ipa_len = 0;
5606
5607 if (addpath_capable)
5608 stream_putl(s, addpath_tx_id);
5609
5610 /* Route type */
5611 stream_putc(s, evp->prefix.route_type);
5612
5613 switch (evp->prefix.route_type) {
5614 case BGP_EVPN_MAC_IP_ROUTE:
5615 if (is_evpn_prefix_ipaddr_v4(evp))
5616 ipa_len = IPV4_MAX_BYTELEN;
5617 else if (is_evpn_prefix_ipaddr_v6(evp))
5618 ipa_len = IPV6_MAX_BYTELEN;
5619 /* RD, ESI, EthTag, MAC+len, IP len, [IP], 1 VNI */
5620 len = 8 + 10 + 4 + 1 + 6 + 1 + ipa_len + 3;
5621 if (ipa_len && num_labels > 1) /* There are 2 VNIs */
5622 len += 3;
5623 stream_putc(s, len);
5624 stream_put(s, prd->val, 8); /* RD */
5625 if (attr)
5626 stream_put(s, &attr->esi, ESI_BYTES);
5627 else
5628 stream_put(s, 0, 10);
5629 stream_putl(s, evp->prefix.macip_addr.eth_tag); /* Ethernet Tag ID */
5630 stream_putc(s, 8 * ETH_ALEN); /* Mac Addr Len - bits */
5631 stream_put(s, evp->prefix.macip_addr.mac.octet, 6); /* Mac Addr */
5632 stream_putc(s, 8 * ipa_len); /* IP address Length */
5633 if (ipa_len) /* IP */
5634 stream_put(s, &evp->prefix.macip_addr.ip.ip.addr,
5635 ipa_len);
5636 /* 1st label is the L2 VNI */
5637 stream_put(s, label, BGP_LABEL_BYTES);
5638 /* Include 2nd label (L3 VNI) if advertising MAC+IP */
5639 if (ipa_len && num_labels > 1)
5640 stream_put(s, label + 1, BGP_LABEL_BYTES);
5641 break;
5642
5643 case BGP_EVPN_IMET_ROUTE:
5644 stream_putc(s, 17); // TODO: length - assumes IPv4 address
5645 stream_put(s, prd->val, 8); /* RD */
5646 stream_putl(s, evp->prefix.imet_addr.eth_tag); /* Ethernet Tag ID */
5647 stream_putc(s, IPV4_MAX_BITLEN); /* IP address Length - bits */
5648 /* Originating Router's IP Addr */
5649 stream_put_in_addr(s, &evp->prefix.imet_addr.ip.ipaddr_v4);
5650 break;
5651
5652 case BGP_EVPN_ES_ROUTE:
5653 stream_putc(s, 23); /* TODO: length: assumes ipv4 VTEP */
5654 stream_put(s, prd->val, 8); /* RD */
5655 stream_put(s, evp->prefix.es_addr.esi.val, 10); /* ESI */
5656 stream_putc(s, IPV4_MAX_BITLEN); /* IP address Length - bits */
5657 /* VTEP IP */
5658 stream_put_in_addr(s, &evp->prefix.es_addr.ip.ipaddr_v4);
5659 break;
5660
5661 case BGP_EVPN_AD_ROUTE:
5662 /* RD, ESI, EthTag, 1 VNI */
5663 len = RD_BYTES + ESI_BYTES + EVPN_ETH_TAG_BYTES + BGP_LABEL_BYTES;
5664 stream_putc(s, len);
5665 stream_put(s, prd->val, RD_BYTES); /* RD */
5666 stream_put(s, evp->prefix.ead_addr.esi.val, ESI_BYTES); /* ESI */
5667 stream_putl(s, evp->prefix.ead_addr.eth_tag); /* Ethernet Tag */
5668 stream_put(s, label, BGP_LABEL_BYTES);
5669 break;
5670
5671 case BGP_EVPN_IP_PREFIX_ROUTE:
5672 /* TODO: AddPath support. */
5673 evpn_mpattr_encode_type5(s, p, prd, label, num_labels, attr);
5674 break;
5675
5676 default:
5677 break;
5678 }
5679 }
5680
5681 int bgp_nlri_parse_evpn(struct peer *peer, struct attr *attr,
5682 struct bgp_nlri *packet, int withdraw)
5683 {
5684 uint8_t *pnt;
5685 uint8_t *lim;
5686 afi_t afi;
5687 safi_t safi;
5688 uint32_t addpath_id;
5689 bool addpath_capable;
5690 int psize = 0;
5691 uint8_t rtype;
5692 struct prefix p;
5693
5694 /* Start processing the NLRI - there may be multiple in the MP_REACH */
5695 pnt = packet->nlri;
5696 lim = pnt + packet->length;
5697 afi = packet->afi;
5698 safi = packet->safi;
5699 addpath_id = 0;
5700
5701 addpath_capable = bgp_addpath_encode_rx(peer, afi, safi);
5702
5703 for (; pnt < lim; pnt += psize) {
5704 /* Clear prefix structure. */
5705 memset(&p, 0, sizeof(p));
5706
5707 /* Deal with path-id if AddPath is supported. */
5708 if (addpath_capable) {
5709 /* When packet overflow occurs return immediately. */
5710 if (pnt + BGP_ADDPATH_ID_LEN > lim)
5711 return BGP_NLRI_PARSE_ERROR_PACKET_OVERFLOW;
5712
5713 memcpy(&addpath_id, pnt, BGP_ADDPATH_ID_LEN);
5714 addpath_id = ntohl(addpath_id);
5715 pnt += BGP_ADDPATH_ID_LEN;
5716 }
5717
5718 /* All EVPN NLRI types start with type and length. */
5719 if (pnt + 2 > lim)
5720 return BGP_NLRI_PARSE_ERROR_EVPN_MISSING_TYPE;
5721
5722 rtype = *pnt++;
5723 psize = *pnt++;
5724
5725 /* When packet overflow occur return immediately. */
5726 if (pnt + psize > lim)
5727 return BGP_NLRI_PARSE_ERROR_PACKET_OVERFLOW;
5728
5729 switch (rtype) {
5730 case BGP_EVPN_MAC_IP_ROUTE:
5731 if (process_type2_route(peer, afi, safi,
5732 withdraw ? NULL : attr, pnt,
5733 psize, addpath_id)) {
5734 flog_err(
5735 EC_BGP_EVPN_FAIL,
5736 "%u:%s - Error in processing EVPN type-2 NLRI size %d",
5737 peer->bgp->vrf_id, peer->host, psize);
5738 return BGP_NLRI_PARSE_ERROR_EVPN_TYPE2_SIZE;
5739 }
5740 break;
5741
5742 case BGP_EVPN_IMET_ROUTE:
5743 if (process_type3_route(peer, afi, safi,
5744 withdraw ? NULL : attr, pnt,
5745 psize, addpath_id)) {
5746 flog_err(
5747 EC_BGP_PKT_PROCESS,
5748 "%u:%s - Error in processing EVPN type-3 NLRI size %d",
5749 peer->bgp->vrf_id, peer->host, psize);
5750 return BGP_NLRI_PARSE_ERROR_EVPN_TYPE3_SIZE;
5751 }
5752 break;
5753
5754 case BGP_EVPN_ES_ROUTE:
5755 if (bgp_evpn_type4_route_process(peer, afi, safi,
5756 withdraw ? NULL : attr, pnt,
5757 psize, addpath_id)) {
5758 flog_err(
5759 EC_BGP_PKT_PROCESS,
5760 "%u:%s - Error in processing EVPN type-4 NLRI size %d",
5761 peer->bgp->vrf_id, peer->host, psize);
5762 return BGP_NLRI_PARSE_ERROR_EVPN_TYPE4_SIZE;
5763 }
5764 break;
5765
5766 case BGP_EVPN_AD_ROUTE:
5767 if (bgp_evpn_type1_route_process(peer, afi, safi,
5768 withdraw ? NULL : attr, pnt,
5769 psize, addpath_id)) {
5770 flog_err(
5771 EC_BGP_PKT_PROCESS,
5772 "%u:%s - Error in processing EVPN type-1 NLRI size %d",
5773 peer->bgp->vrf_id, peer->host, psize);
5774 return BGP_NLRI_PARSE_ERROR_EVPN_TYPE1_SIZE;
5775 }
5776 break;
5777
5778 case BGP_EVPN_IP_PREFIX_ROUTE:
5779 if (process_type5_route(peer, afi, safi,
5780 withdraw ? NULL : attr, pnt,
5781 psize, addpath_id)) {
5782 flog_err(
5783 EC_BGP_PKT_PROCESS,
5784 "%u:%s - Error in processing EVPN type-5 NLRI size %d",
5785 peer->bgp->vrf_id, peer->host, psize);
5786 return BGP_NLRI_PARSE_ERROR_EVPN_TYPE5_SIZE;
5787 }
5788 break;
5789
5790 default:
5791 break;
5792 }
5793 }
5794
5795 /* Packet length consistency check. */
5796 if (pnt != lim)
5797 return BGP_NLRI_PARSE_ERROR_PACKET_LENGTH;
5798
5799 return BGP_NLRI_PARSE_OK;
5800 }
5801
5802 /*
5803 * Map the RTs (configured or automatically derived) of a VRF to the VRF.
5804 * The mapping will be used during route processing.
5805 * bgp_vrf: specific bgp vrf instance on which RT is configured
5806 */
5807 void bgp_evpn_map_vrf_to_its_rts(struct bgp *bgp_vrf)
5808 {
5809 struct listnode *node, *nnode;
5810 struct vrf_route_target *l3rt;
5811
5812 for (ALL_LIST_ELEMENTS(bgp_vrf->vrf_import_rtl, node, nnode, l3rt))
5813 map_vrf_to_rt(bgp_vrf, l3rt);
5814 }
5815
5816 /*
5817 * Unmap the RTs (configured or automatically derived) of a VRF from the VRF.
5818 */
5819 void bgp_evpn_unmap_vrf_from_its_rts(struct bgp *bgp_vrf)
5820 {
5821 struct listnode *node, *nnode;
5822 struct vrf_route_target *l3rt;
5823
5824 for (ALL_LIST_ELEMENTS(bgp_vrf->vrf_import_rtl, node, nnode, l3rt))
5825 unmap_vrf_from_rt(bgp_vrf, l3rt);
5826 }
5827
5828 /*
5829 * Map the RTs (configured or automatically derived) of a VNI to the VNI.
5830 * The mapping will be used during route processing.
5831 */
5832 void bgp_evpn_map_vni_to_its_rts(struct bgp *bgp, struct bgpevpn *vpn)
5833 {
5834 uint32_t i;
5835 struct ecommunity_val *eval;
5836 struct listnode *node, *nnode;
5837 struct ecommunity *ecom;
5838
5839 for (ALL_LIST_ELEMENTS(vpn->import_rtl, node, nnode, ecom)) {
5840 for (i = 0; i < ecom->size; i++) {
5841 eval = (struct ecommunity_val *)(ecom->val
5842 + (i
5843 * ECOMMUNITY_SIZE));
5844 map_vni_to_rt(bgp, vpn, eval);
5845 }
5846 }
5847 }
5848
5849 /*
5850 * Unmap the RTs (configured or automatically derived) of a VNI from the VNI.
5851 */
5852 void bgp_evpn_unmap_vni_from_its_rts(struct bgp *bgp, struct bgpevpn *vpn)
5853 {
5854 uint32_t i;
5855 struct ecommunity_val *eval;
5856 struct listnode *node, *nnode;
5857 struct ecommunity *ecom;
5858
5859 for (ALL_LIST_ELEMENTS(vpn->import_rtl, node, nnode, ecom)) {
5860 for (i = 0; i < ecom->size; i++) {
5861 struct irt_node *irt;
5862 struct ecommunity_val eval_tmp;
5863
5864 eval = (struct ecommunity_val *)(ecom->val
5865 + (i
5866 * ECOMMUNITY_SIZE));
5867 /* If using "automatic" RT, we only care about the
5868 * local-admin sub-field.
5869 * This is to facilitate using VNI as the RT for EBGP
5870 * peering too.
5871 */
5872 memcpy(&eval_tmp, eval, ECOMMUNITY_SIZE);
5873 if (!is_import_rt_configured(vpn))
5874 mask_ecom_global_admin(&eval_tmp, eval);
5875
5876 irt = lookup_import_rt(bgp, &eval_tmp);
5877 if (irt)
5878 unmap_vni_from_rt(bgp, vpn, irt);
5879 }
5880 }
5881 }
5882
5883 /*
5884 * Derive Import RT automatically for VNI and map VNI to RT.
5885 * The mapping will be used during route processing.
5886 */
5887 void bgp_evpn_derive_auto_rt_import(struct bgp *bgp, struct bgpevpn *vpn)
5888 {
5889 form_auto_rt(bgp, vpn->vni, vpn->import_rtl, false);
5890 UNSET_FLAG(vpn->flags, VNI_FLAG_IMPRT_CFGD);
5891
5892 /* Map RT to VNI */
5893 bgp_evpn_map_vni_to_its_rts(bgp, vpn);
5894 }
5895
5896 /*
5897 * Derive Export RT automatically for VNI.
5898 */
5899 void bgp_evpn_derive_auto_rt_export(struct bgp *bgp, struct bgpevpn *vpn)
5900 {
5901 form_auto_rt(bgp, vpn->vni, vpn->export_rtl, false);
5902 UNSET_FLAG(vpn->flags, VNI_FLAG_EXPRT_CFGD);
5903 }
5904
5905 /*
5906 * Derive RD automatically for VNI using passed information - it
5907 * is of the form RouterId:unique-id-for-vni.
5908 */
5909 void bgp_evpn_derive_auto_rd_for_vrf(struct bgp *bgp)
5910 {
5911 if (is_vrf_rd_configured(bgp))
5912 return;
5913
5914 form_auto_rd(bgp->router_id, bgp->vrf_rd_id, &bgp->vrf_prd);
5915 }
5916
5917 /*
5918 * Derive RD automatically for VNI using passed information - it
5919 * is of the form RouterId:unique-id-for-vni.
5920 */
5921 void bgp_evpn_derive_auto_rd(struct bgp *bgp, struct bgpevpn *vpn)
5922 {
5923 char buf[BGP_EVPN_PREFIX_RD_LEN];
5924
5925 vpn->prd.family = AF_UNSPEC;
5926 vpn->prd.prefixlen = 64;
5927 snprintfrr(buf, sizeof(buf), "%pI4:%hu", &bgp->router_id, vpn->rd_id);
5928 (void)str2prefix_rd(buf, &vpn->prd);
5929 if (vpn->prd_pretty)
5930 XFREE(MTYPE_BGP, vpn->prd_pretty);
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 if (vpn->prd_pretty)
6036 XFREE(MTYPE_BGP, vpn->prd_pretty);
6037 QOBJ_UNREG(vpn);
6038 XFREE(MTYPE_BGP_EVPN, vpn);
6039 }
6040
6041 static void hash_evpn_free(struct bgpevpn *vpn)
6042 {
6043 XFREE(MTYPE_BGP_EVPN, vpn);
6044 }
6045
6046 /*
6047 * Import evpn route from global table to VNI/VRF/ESI.
6048 */
6049 int bgp_evpn_import_route(struct bgp *bgp, afi_t afi, safi_t safi,
6050 const struct prefix *p, struct bgp_path_info *pi)
6051 {
6052 return install_uninstall_evpn_route(bgp, afi, safi, p, pi, 1);
6053 }
6054
6055 /*
6056 * Unimport evpn route from VNI/VRF/ESI.
6057 */
6058 int bgp_evpn_unimport_route(struct bgp *bgp, afi_t afi, safi_t safi,
6059 const struct prefix *p, struct bgp_path_info *pi)
6060 {
6061 return install_uninstall_evpn_route(bgp, afi, safi, p, pi, 0);
6062 }
6063
6064 /* filter routes which have martian next hops */
6065 int bgp_filter_evpn_routes_upon_martian_nh_change(struct bgp *bgp)
6066 {
6067 afi_t afi;
6068 safi_t safi;
6069 struct bgp_dest *rd_dest, *dest;
6070 struct bgp_table *table;
6071 struct bgp_path_info *pi;
6072
6073 afi = AFI_L2VPN;
6074 safi = SAFI_EVPN;
6075
6076 /* Walk entire global routing table and evaluate routes which could be
6077 * imported into this VPN. Note that we cannot just look at the routes
6078 * for the VNI's RD -
6079 * remote routes applicable for this VNI could have any RD.
6080 */
6081 /* EVPN routes are a 2-level table. */
6082 for (rd_dest = bgp_table_top(bgp->rib[afi][safi]); rd_dest;
6083 rd_dest = bgp_route_next(rd_dest)) {
6084 table = bgp_dest_get_bgp_table_info(rd_dest);
6085 if (!table)
6086 continue;
6087
6088 for (dest = bgp_table_top(table); dest;
6089 dest = bgp_route_next(dest)) {
6090
6091 for (pi = bgp_dest_get_bgp_path_info(dest); pi;
6092 pi = pi->next) {
6093
6094 /* Consider "valid" remote routes applicable for
6095 * this VNI. */
6096 if (!(pi->type == ZEBRA_ROUTE_BGP
6097 && pi->sub_type == BGP_ROUTE_NORMAL))
6098 continue;
6099 if (bgp_nexthop_self(bgp, afi, pi->type,
6100 pi->sub_type, pi->attr,
6101 dest)) {
6102 const struct prefix *p =
6103 bgp_dest_get_prefix(dest);
6104
6105 if (bgp_debug_update(pi->peer, p, NULL,
6106 1)) {
6107 char attr_str[BUFSIZ] = {0};
6108
6109 bgp_dump_attr(pi->attr,
6110 attr_str,
6111 sizeof(attr_str));
6112
6113 zlog_debug(
6114 "%u: prefix %pBD with attr %s - DENIED due to martian or self nexthop",
6115 bgp->vrf_id, dest,
6116 attr_str);
6117 }
6118 bgp_evpn_unimport_route(bgp, afi, safi,
6119 p, pi);
6120
6121 bgp_rib_remove(dest, pi, pi->peer, afi,
6122 safi);
6123 }
6124 }
6125 }
6126 }
6127
6128 return 0;
6129 }
6130
6131 /*
6132 * Handle del of a local MACIP.
6133 */
6134 int bgp_evpn_local_macip_del(struct bgp *bgp, vni_t vni, struct ethaddr *mac,
6135 struct ipaddr *ip, int state)
6136 {
6137 struct bgpevpn *vpn;
6138 struct prefix_evpn p;
6139 struct bgp_dest *dest;
6140
6141 /* Lookup VNI hash - should exist. */
6142 vpn = bgp_evpn_lookup_vni(bgp, vni);
6143 if (!vpn || !is_vni_live(vpn)) {
6144 flog_warn(EC_BGP_EVPN_VPN_VNI,
6145 "%u: VNI hash entry for VNI %u %s at MACIP DEL",
6146 bgp->vrf_id, vni, vpn ? "not live" : "not found");
6147 return -1;
6148 }
6149
6150 build_evpn_type2_prefix(&p, mac, ip);
6151 if (state == ZEBRA_NEIGH_ACTIVE) {
6152 /* Remove EVPN type-2 route and schedule for processing. */
6153 delete_evpn_route(bgp, vpn, &p);
6154 } else {
6155 /* Re-instate the current remote best path if any */
6156 dest = bgp_evpn_vni_node_lookup(vpn, &p, NULL);
6157 if (dest) {
6158 evpn_zebra_reinstall_best_route(bgp, vpn, dest);
6159 bgp_dest_unlock_node(dest);
6160 }
6161 }
6162
6163 return 0;
6164 }
6165
6166 /*
6167 * Handle add of a local MACIP.
6168 */
6169 int bgp_evpn_local_macip_add(struct bgp *bgp, vni_t vni, struct ethaddr *mac,
6170 struct ipaddr *ip, uint8_t flags, uint32_t seq, esi_t *esi)
6171 {
6172 struct bgpevpn *vpn;
6173 struct prefix_evpn p;
6174
6175 /* Lookup VNI hash - should exist. */
6176 vpn = bgp_evpn_lookup_vni(bgp, vni);
6177 if (!vpn || !is_vni_live(vpn)) {
6178 flog_warn(EC_BGP_EVPN_VPN_VNI,
6179 "%u: VNI hash entry for VNI %u %s at MACIP ADD",
6180 bgp->vrf_id, vni, vpn ? "not live" : "not found");
6181 return -1;
6182 }
6183
6184 /* Create EVPN type-2 route and schedule for processing. */
6185 build_evpn_type2_prefix(&p, mac, ip);
6186 if (update_evpn_route(bgp, vpn, &p, flags, seq, esi)) {
6187 flog_err(
6188 EC_BGP_EVPN_ROUTE_CREATE,
6189 "%u:Failed to create Type-2 route, VNI %u %s MAC %pEA IP %pIA (flags: 0x%x)",
6190 bgp->vrf_id, vpn->vni,
6191 CHECK_FLAG(flags, ZEBRA_MACIP_TYPE_STICKY)
6192 ? "sticky gateway"
6193 : "",
6194 mac, ip, flags);
6195 return -1;
6196 }
6197
6198 return 0;
6199 }
6200
6201 static void link_l2vni_hash_to_l3vni(struct hash_bucket *bucket,
6202 struct bgp *bgp_vrf)
6203 {
6204 struct bgpevpn *vpn = (struct bgpevpn *)bucket->data;
6205 struct bgp *bgp_evpn = NULL;
6206
6207 bgp_evpn = bgp_get_evpn();
6208 assert(bgp_evpn);
6209
6210 if (vpn->tenant_vrf_id == bgp_vrf->vrf_id)
6211 bgpevpn_link_to_l3vni(vpn);
6212 }
6213
6214 int bgp_evpn_local_l3vni_add(vni_t l3vni, vrf_id_t vrf_id,
6215 struct ethaddr *svi_rmac,
6216 struct ethaddr *vrr_rmac,
6217 struct in_addr originator_ip, int filter,
6218 ifindex_t svi_ifindex,
6219 bool is_anycast_mac)
6220 {
6221 struct bgp *bgp_vrf = NULL; /* bgp VRF instance */
6222 struct bgp *bgp_evpn = NULL; /* EVPN bgp instance */
6223 struct listnode *node = NULL;
6224 struct bgpevpn *vpn = NULL;
6225 as_t as = 0;
6226
6227 /* get the EVPN instance - required to get the AS number for VRF
6228 * auto-creatio
6229 */
6230 bgp_evpn = bgp_get_evpn();
6231 if (!bgp_evpn) {
6232 flog_err(
6233 EC_BGP_NO_DFLT,
6234 "Cannot process L3VNI %u ADD - EVPN BGP instance not yet created",
6235 l3vni);
6236 return -1;
6237 }
6238 as = bgp_evpn->as;
6239
6240 /* if the BGP vrf instance doesn't exist - create one */
6241 bgp_vrf = bgp_lookup_by_vrf_id(vrf_id);
6242 if (!bgp_vrf) {
6243
6244 int ret = 0;
6245
6246 ret = bgp_get_vty(&bgp_vrf, &as, vrf_id_to_name(vrf_id),
6247 vrf_id == VRF_DEFAULT
6248 ? BGP_INSTANCE_TYPE_DEFAULT
6249 : BGP_INSTANCE_TYPE_VRF,
6250 NULL, ASNOTATION_UNDEFINED);
6251 switch (ret) {
6252 case BGP_ERR_AS_MISMATCH:
6253 flog_err(EC_BGP_EVPN_AS_MISMATCH,
6254 "BGP instance is already running; AS is %s",
6255 bgp_vrf->as_pretty);
6256 return -1;
6257 case BGP_ERR_INSTANCE_MISMATCH:
6258 flog_err(EC_BGP_EVPN_INSTANCE_MISMATCH,
6259 "BGP instance type mismatch");
6260 return -1;
6261 }
6262
6263 /* mark as auto created */
6264 SET_FLAG(bgp_vrf->vrf_flags, BGP_VRF_AUTO);
6265 }
6266
6267 /* associate the vrf with l3vni and related parameters */
6268 bgp_vrf->l3vni = l3vni;
6269 bgp_vrf->originator_ip = originator_ip;
6270 bgp_vrf->l3vni_svi_ifindex = svi_ifindex;
6271 bgp_vrf->evpn_info->is_anycast_mac = is_anycast_mac;
6272
6273 /* copy anycast MAC from VRR MAC */
6274 memcpy(&bgp_vrf->rmac, vrr_rmac, ETH_ALEN);
6275 /* copy sys RMAC from SVI MAC */
6276 memcpy(&bgp_vrf->evpn_info->pip_rmac_zebra, svi_rmac, ETH_ALEN);
6277 /* PIP user configured mac is not present use svi mac as sys mac */
6278 if (is_zero_mac(&bgp_vrf->evpn_info->pip_rmac_static))
6279 memcpy(&bgp_vrf->evpn_info->pip_rmac, svi_rmac, ETH_ALEN);
6280
6281 if (bgp_debug_zebra(NULL))
6282 zlog_debug(
6283 "VRF %s vni %u pip %s RMAC %pEA sys RMAC %pEA static RMAC %pEA is_anycast_mac %s",
6284 vrf_id_to_name(bgp_vrf->vrf_id), bgp_vrf->l3vni,
6285 bgp_vrf->evpn_info->advertise_pip ? "enable"
6286 : "disable",
6287 &bgp_vrf->rmac, &bgp_vrf->evpn_info->pip_rmac,
6288 &bgp_vrf->evpn_info->pip_rmac_static,
6289 is_anycast_mac ? "Enable" : "Disable");
6290
6291 /* set the right filter - are we using l3vni only for prefix routes? */
6292 if (filter) {
6293 SET_FLAG(bgp_vrf->vrf_flags, BGP_VRF_L3VNI_PREFIX_ROUTES_ONLY);
6294
6295 /*
6296 * VNI_FLAG_USE_TWO_LABELS flag for linked L2VNIs should not be
6297 * set before linking vrf to L3VNI. Thus, no need to clear
6298 * that explicitly.
6299 */
6300 } else {
6301 UNSET_FLAG(bgp_vrf->vrf_flags,
6302 BGP_VRF_L3VNI_PREFIX_ROUTES_ONLY);
6303
6304 for (ALL_LIST_ELEMENTS_RO(bgp_vrf->l2vnis, node, vpn)) {
6305 if (!CHECK_FLAG(vpn->flags, VNI_FLAG_USE_TWO_LABELS)) {
6306
6307 /*
6308 * If we are flapping VNI_FLAG_USE_TWO_LABELS
6309 * flag, update all MACIP routes in this VNI
6310 */
6311 SET_FLAG(vpn->flags, VNI_FLAG_USE_TWO_LABELS);
6312 update_all_type2_routes(bgp_evpn, vpn);
6313 }
6314 }
6315 }
6316
6317 /* Map auto derive or configured RTs */
6318 if (!CHECK_FLAG(bgp_vrf->vrf_flags, BGP_VRF_IMPORT_RT_CFGD) ||
6319 CHECK_FLAG(bgp_vrf->vrf_flags, BGP_VRF_IMPORT_AUTO_RT_CFGD))
6320 evpn_auto_rt_import_add_for_vrf(bgp_vrf);
6321 else
6322 bgp_evpn_map_vrf_to_its_rts(bgp_vrf);
6323
6324 if (!CHECK_FLAG(bgp_vrf->vrf_flags, BGP_VRF_EXPORT_RT_CFGD) ||
6325 CHECK_FLAG(bgp_vrf->vrf_flags, BGP_VRF_EXPORT_AUTO_RT_CFGD))
6326 evpn_auto_rt_export_add_for_vrf(bgp_vrf);
6327
6328 /* auto derive RD */
6329 bgp_evpn_derive_auto_rd_for_vrf(bgp_vrf);
6330
6331 /* link all corresponding l2vnis */
6332 hash_iterate(bgp_evpn->vnihash,
6333 (void (*)(struct hash_bucket *,
6334 void *))link_l2vni_hash_to_l3vni,
6335 bgp_vrf);
6336
6337 /* Only update all corresponding type-2 routes if we are advertising two
6338 * labels along with type-2 routes
6339 */
6340 if (!filter)
6341 for (ALL_LIST_ELEMENTS_RO(bgp_vrf->l2vnis, node, vpn))
6342 update_routes_for_vni(bgp_evpn, vpn);
6343
6344 /* advertise type-5 routes if needed */
6345 update_advertise_vrf_routes(bgp_vrf);
6346
6347 /* install all remote routes belonging to this l3vni into correspondng
6348 * vrf */
6349 install_routes_for_vrf(bgp_vrf);
6350
6351 return 0;
6352 }
6353
6354 int bgp_evpn_local_l3vni_del(vni_t l3vni, vrf_id_t vrf_id)
6355 {
6356 struct bgp *bgp_vrf = NULL; /* bgp vrf instance */
6357 struct bgp *bgp_evpn = NULL; /* EVPN bgp instance */
6358 struct listnode *node = NULL;
6359 struct listnode *next = NULL;
6360 struct bgpevpn *vpn = NULL;
6361
6362 bgp_vrf = bgp_lookup_by_vrf_id(vrf_id);
6363 if (!bgp_vrf) {
6364 flog_err(
6365 EC_BGP_NO_DFLT,
6366 "Cannot process L3VNI %u Del - Could not find BGP instance",
6367 l3vni);
6368 return -1;
6369 }
6370
6371 bgp_evpn = bgp_get_evpn();
6372 if (!bgp_evpn) {
6373 flog_err(
6374 EC_BGP_NO_DFLT,
6375 "Cannot process L3VNI %u Del - Could not find EVPN BGP instance",
6376 l3vni);
6377 return -1;
6378 }
6379
6380 /* Remove remote routes from BGT VRF even if BGP_VRF_AUTO is configured,
6381 * bgp_delete would not remove/decrement bgp_path_info of the ip_prefix
6382 * routes. This will uninstalling the routes from zebra and decremnt the
6383 * bgp info count.
6384 */
6385 uninstall_routes_for_vrf(bgp_vrf);
6386
6387 /* delete/withdraw all type-5 routes */
6388 delete_withdraw_vrf_routes(bgp_vrf);
6389
6390 /* remove the l3vni from vrf instance */
6391 bgp_vrf->l3vni = 0;
6392
6393 /* remove the Rmac from the BGP vrf */
6394 memset(&bgp_vrf->rmac, 0, sizeof(struct ethaddr));
6395 memset(&bgp_vrf->evpn_info->pip_rmac_zebra, 0, ETH_ALEN);
6396 if (is_zero_mac(&bgp_vrf->evpn_info->pip_rmac_static) &&
6397 !is_zero_mac(&bgp_vrf->evpn_info->pip_rmac))
6398 memset(&bgp_vrf->evpn_info->pip_rmac, 0, ETH_ALEN);
6399
6400 /* remove default import RT or Unmap non-default import RT */
6401 if (!list_isempty(bgp_vrf->vrf_import_rtl)) {
6402 bgp_evpn_unmap_vrf_from_its_rts(bgp_vrf);
6403 if (!CHECK_FLAG(bgp_vrf->vrf_flags, BGP_VRF_IMPORT_RT_CFGD))
6404 list_delete_all_node(bgp_vrf->vrf_import_rtl);
6405 }
6406
6407 /* remove default export RT */
6408 if (!list_isempty(bgp_vrf->vrf_export_rtl) &&
6409 !CHECK_FLAG(bgp_vrf->vrf_flags, BGP_VRF_EXPORT_RT_CFGD)) {
6410 list_delete_all_node(bgp_vrf->vrf_export_rtl);
6411 }
6412
6413 /* update all corresponding local mac-ip routes */
6414 if (!CHECK_FLAG(bgp_vrf->vrf_flags, BGP_VRF_L3VNI_PREFIX_ROUTES_ONLY)) {
6415 for (ALL_LIST_ELEMENTS_RO(bgp_vrf->l2vnis, node, vpn)) {
6416 UNSET_FLAG(vpn->flags, VNI_FLAG_USE_TWO_LABELS);
6417 update_routes_for_vni(bgp_evpn, vpn);
6418 }
6419 }
6420
6421 /* If any L2VNIs point to this instance, unlink them. */
6422 for (ALL_LIST_ELEMENTS(bgp_vrf->l2vnis, node, next, vpn))
6423 bgpevpn_unlink_from_l3vni(vpn);
6424
6425 UNSET_FLAG(bgp_vrf->vrf_flags, BGP_VRF_L3VNI_PREFIX_ROUTES_ONLY);
6426
6427 /* Delete the instance if it was autocreated */
6428 if (CHECK_FLAG(bgp_vrf->vrf_flags, BGP_VRF_AUTO))
6429 bgp_delete(bgp_vrf);
6430
6431 return 0;
6432 }
6433
6434 /*
6435 * Handle del of a local VNI.
6436 */
6437 int bgp_evpn_local_vni_del(struct bgp *bgp, vni_t vni)
6438 {
6439 struct bgpevpn *vpn;
6440
6441 /* Locate VNI hash */
6442 vpn = bgp_evpn_lookup_vni(bgp, vni);
6443 if (!vpn)
6444 return 0;
6445
6446 /* Remove all local EVPN routes and schedule for processing (to
6447 * withdraw from peers).
6448 */
6449 delete_routes_for_vni(bgp, vpn);
6450
6451 bgp_evpn_unlink_from_vni_svi_hash(bgp, vpn);
6452
6453 vpn->svi_ifindex = 0;
6454 /*
6455 * tunnel is no longer active, del tunnel ip address from tip_hash
6456 */
6457 bgp_tip_del(bgp, &vpn->originator_ip);
6458
6459 /* Clear "live" flag and see if hash needs to be freed. */
6460 UNSET_FLAG(vpn->flags, VNI_FLAG_LIVE);
6461 if (!is_vni_configured(vpn))
6462 bgp_evpn_free(bgp, vpn);
6463
6464 return 0;
6465 }
6466
6467 /*
6468 * Handle add (or update) of a local VNI. The VNI changes we care
6469 * about are for the local-tunnel-ip and the (tenant) VRF.
6470 */
6471 int bgp_evpn_local_vni_add(struct bgp *bgp, vni_t vni,
6472 struct in_addr originator_ip,
6473 vrf_id_t tenant_vrf_id,
6474 struct in_addr mcast_grp,
6475 ifindex_t svi_ifindex)
6476 {
6477 struct bgpevpn *vpn;
6478 struct prefix_evpn p;
6479
6480 /* Lookup VNI. If present and no change, exit. */
6481 vpn = bgp_evpn_lookup_vni(bgp, vni);
6482 if (vpn) {
6483
6484 if (is_vni_live(vpn)
6485 && IPV4_ADDR_SAME(&vpn->originator_ip, &originator_ip)
6486 && IPV4_ADDR_SAME(&vpn->mcast_grp, &mcast_grp)
6487 && vpn->tenant_vrf_id == tenant_vrf_id
6488 && vpn->svi_ifindex == svi_ifindex)
6489 /* Probably some other param has changed that we don't
6490 * care about. */
6491 return 0;
6492
6493 bgp_evpn_mcast_grp_change(bgp, vpn, mcast_grp);
6494
6495 if (vpn->svi_ifindex != svi_ifindex) {
6496
6497 /*
6498 * Unresolve all the gateway IP nexthops for this VNI
6499 * for old SVI
6500 */
6501 bgp_evpn_remote_ip_hash_iterate(
6502 vpn,
6503 (void (*)(struct hash_bucket *, void *))
6504 bgp_evpn_remote_ip_hash_unlink_nexthop,
6505 vpn);
6506 bgp_evpn_unlink_from_vni_svi_hash(bgp, vpn);
6507 vpn->svi_ifindex = svi_ifindex;
6508 bgp_evpn_link_to_vni_svi_hash(bgp, vpn);
6509
6510 /*
6511 * Resolve all the gateway IP nexthops for this VNI
6512 * for new SVI
6513 */
6514 bgp_evpn_remote_ip_hash_iterate(
6515 vpn,
6516 (void (*)(struct hash_bucket *, void *))
6517 bgp_evpn_remote_ip_hash_link_nexthop,
6518 vpn);
6519 }
6520
6521 /* Update tenant_vrf_id if it has changed. */
6522 if (vpn->tenant_vrf_id != tenant_vrf_id) {
6523
6524 /*
6525 * Unresolve all the gateway IP nexthops for this VNI
6526 * in old tenant vrf
6527 */
6528 bgp_evpn_remote_ip_hash_iterate(
6529 vpn,
6530 (void (*)(struct hash_bucket *, void *))
6531 bgp_evpn_remote_ip_hash_unlink_nexthop,
6532 vpn);
6533 bgpevpn_unlink_from_l3vni(vpn);
6534 vpn->tenant_vrf_id = tenant_vrf_id;
6535 bgpevpn_link_to_l3vni(vpn);
6536
6537 /*
6538 * Resolve all the gateway IP nexthops for this VNI
6539 * in new tenant vrf
6540 */
6541 bgp_evpn_remote_ip_hash_iterate(
6542 vpn,
6543 (void (*)(struct hash_bucket *, void *))
6544 bgp_evpn_remote_ip_hash_link_nexthop,
6545 vpn);
6546 }
6547
6548 /* If tunnel endpoint IP has changed, update (and delete prior
6549 * type-3 route, if needed.)
6550 */
6551 handle_tunnel_ip_change(bgp, vpn, originator_ip);
6552
6553 /* Update all routes with new endpoint IP and/or export RT
6554 * for VRFs
6555 */
6556 if (is_vni_live(vpn))
6557 update_routes_for_vni(bgp, vpn);
6558 } else {
6559 /* Create or update as appropriate. */
6560 vpn = bgp_evpn_new(bgp, vni, originator_ip, tenant_vrf_id,
6561 mcast_grp, svi_ifindex);
6562 }
6563
6564 /* if the VNI is live already, there is nothing more to do */
6565 if (is_vni_live(vpn))
6566 return 0;
6567
6568 /* Mark as "live" */
6569 SET_FLAG(vpn->flags, VNI_FLAG_LIVE);
6570
6571 /* tunnel is now active, add tunnel-ip to db */
6572 if (bgp_tip_add(bgp, &originator_ip))
6573 /* The originator_ip was not already present in the
6574 * bgp martian next-hop table as a tunnel-ip, so we
6575 * need to go back and filter routes matching the new
6576 * martian next-hop.
6577 */
6578 bgp_filter_evpn_routes_upon_martian_nh_change(bgp);
6579
6580 /*
6581 * Create EVPN type-3 route and schedule for processing.
6582 *
6583 * RT-3 only if doing head-end replication
6584 */
6585 if (bgp_evpn_vni_flood_mode_get(bgp, vpn)
6586 == VXLAN_FLOOD_HEAD_END_REPL) {
6587 build_evpn_type3_prefix(&p, vpn->originator_ip);
6588 if (update_evpn_route(bgp, vpn, &p, 0, 0, NULL)) {
6589 flog_err(EC_BGP_EVPN_ROUTE_CREATE,
6590 "%u: Type3 route creation failure for VNI %u",
6591 bgp->vrf_id, vni);
6592 return -1;
6593 }
6594 }
6595
6596 /* If we have learnt and retained remote routes (VTEPs, MACs) for this
6597 * VNI,
6598 * install them.
6599 */
6600 install_routes_for_vni(bgp, vpn);
6601
6602 /* If we are advertising gateway mac-ip
6603 It needs to be conveyed again to zebra */
6604 bgp_zebra_advertise_gw_macip(bgp, vpn->advertise_gw_macip, vpn->vni);
6605
6606 /* advertise svi mac-ip knob to zebra */
6607 bgp_zebra_advertise_svi_macip(bgp, vpn->advertise_svi_macip, vpn->vni);
6608
6609 return 0;
6610 }
6611
6612 /*
6613 * Handle change in setting for BUM handling. The supported values
6614 * are head-end replication and dropping all BUM packets. Any change
6615 * should be registered with zebra. Also, if doing head-end replication,
6616 * need to advertise local VNIs as EVPN RT-3 wheras, if BUM packets are
6617 * to be dropped, the RT-3s must be withdrawn.
6618 */
6619 void bgp_evpn_flood_control_change(struct bgp *bgp)
6620 {
6621 zlog_info("L2VPN EVPN BUM handling is %s",
6622 bgp->vxlan_flood_ctrl == VXLAN_FLOOD_HEAD_END_REPL ?
6623 "Flooding" : "Flooding Disabled");
6624
6625 bgp_zebra_vxlan_flood_control(bgp, bgp->vxlan_flood_ctrl);
6626 if (bgp->vxlan_flood_ctrl == VXLAN_FLOOD_HEAD_END_REPL)
6627 hash_iterate(bgp->vnihash, create_advertise_type3, bgp);
6628 else if (bgp->vxlan_flood_ctrl == VXLAN_FLOOD_DISABLED)
6629 hash_iterate(bgp->vnihash, delete_withdraw_type3, bgp);
6630 }
6631
6632 /*
6633 * Cleanup EVPN information on disable - Need to delete and withdraw
6634 * EVPN routes from peers.
6635 */
6636 void bgp_evpn_cleanup_on_disable(struct bgp *bgp)
6637 {
6638 hash_iterate(bgp->vnihash, (void (*)(struct hash_bucket *,
6639 void *))cleanup_vni_on_disable,
6640 bgp);
6641 }
6642
6643 /*
6644 * Cleanup EVPN information - invoked at the time of bgpd exit or when the
6645 * BGP instance (default) is being freed.
6646 */
6647 void bgp_evpn_cleanup(struct bgp *bgp)
6648 {
6649 hash_iterate(bgp->vnihash,
6650 (void (*)(struct hash_bucket *, void *))free_vni_entry,
6651 bgp);
6652
6653 hash_clean_and_free(&bgp->import_rt_hash,
6654 (void (*)(void *))hash_import_rt_free);
6655
6656 hash_clean_and_free(&bgp->vrf_import_rt_hash,
6657 (void (*)(void *))hash_vrf_import_rt_free);
6658
6659 hash_clean_and_free(&bgp->vni_svi_hash,
6660 (void (*)(void *))hash_evpn_free);
6661
6662 /*
6663 * Why is the vnihash freed at the top of this function and
6664 * then deleted here?
6665 */
6666 hash_clean_and_free(&bgp->vnihash, NULL);
6667
6668 list_delete(&bgp->vrf_import_rtl);
6669 list_delete(&bgp->vrf_export_rtl);
6670 list_delete(&bgp->l2vnis);
6671
6672 if (bgp->vrf_prd_pretty)
6673 XFREE(MTYPE_BGP, bgp->vrf_prd_pretty);
6674 }
6675
6676 /*
6677 * Initialization for EVPN
6678 * Create
6679 * VNI hash table
6680 * hash for RT to VNI
6681 */
6682 void bgp_evpn_init(struct bgp *bgp)
6683 {
6684 bgp->vnihash =
6685 hash_create(vni_hash_key_make, vni_hash_cmp, "BGP VNI Hash");
6686 bgp->vni_svi_hash =
6687 hash_create(vni_svi_hash_key_make, vni_svi_hash_cmp,
6688 "BGP VNI hash based on SVI ifindex");
6689 bgp->import_rt_hash =
6690 hash_create(import_rt_hash_key_make, import_rt_hash_cmp,
6691 "BGP Import RT Hash");
6692 bgp->vrf_import_rt_hash =
6693 hash_create(vrf_import_rt_hash_key_make, vrf_import_rt_hash_cmp,
6694 "BGP VRF Import RT Hash");
6695 bgp->vrf_import_rtl = list_new();
6696 bgp->vrf_import_rtl->cmp =
6697 (int (*)(void *, void *))evpn_vrf_route_target_cmp;
6698 bgp->vrf_import_rtl->del = evpn_vrf_rt_del;
6699 bgp->vrf_export_rtl = list_new();
6700 bgp->vrf_export_rtl->cmp =
6701 (int (*)(void *, void *))evpn_vrf_route_target_cmp;
6702 bgp->vrf_export_rtl->del = evpn_vrf_rt_del;
6703 bgp->l2vnis = list_new();
6704 bgp->l2vnis->cmp = vni_list_cmp;
6705 /* By default Duplicate Address Dection is enabled.
6706 * Max-moves (N) 5, detection time (M) 180
6707 * default action is warning-only
6708 * freeze action permanently freezes address,
6709 * and freeze time (auto-recovery) is disabled.
6710 */
6711 if (bgp->evpn_info) {
6712 bgp->evpn_info->dup_addr_detect = true;
6713 bgp->evpn_info->dad_time = EVPN_DAD_DEFAULT_TIME;
6714 bgp->evpn_info->dad_max_moves = EVPN_DAD_DEFAULT_MAX_MOVES;
6715 bgp->evpn_info->dad_freeze = false;
6716 bgp->evpn_info->dad_freeze_time = 0;
6717 /* Initialize zebra vxlan */
6718 bgp_zebra_dup_addr_detection(bgp);
6719 /* Enable PIP feature by default for bgp vrf instance */
6720 if (bgp->inst_type == BGP_INSTANCE_TYPE_VRF) {
6721 struct bgp *bgp_default;
6722
6723 bgp->evpn_info->advertise_pip = true;
6724 bgp_default = bgp_get_default();
6725 if (bgp_default)
6726 bgp->evpn_info->pip_ip = bgp_default->router_id;
6727 }
6728 }
6729
6730 /* Default BUM handling is to do head-end replication. */
6731 bgp->vxlan_flood_ctrl = VXLAN_FLOOD_HEAD_END_REPL;
6732
6733 bgp_evpn_nh_init(bgp);
6734 }
6735
6736 void bgp_evpn_vrf_delete(struct bgp *bgp_vrf)
6737 {
6738 bgp_evpn_unmap_vrf_from_its_rts(bgp_vrf);
6739 bgp_evpn_nh_finish(bgp_vrf);
6740 }
6741
6742 /*
6743 * Get the prefixlen of the ip prefix carried within the type5 evpn route.
6744 */
6745 int bgp_evpn_get_type5_prefixlen(const struct prefix *pfx)
6746 {
6747 struct prefix_evpn *evp = (struct prefix_evpn *)pfx;
6748
6749 if (!pfx || pfx->family != AF_EVPN)
6750 return 0;
6751
6752 if (evp->prefix.route_type != BGP_EVPN_IP_PREFIX_ROUTE)
6753 return 0;
6754
6755 return evp->prefix.prefix_addr.ip_prefix_length;
6756 }
6757
6758 /*
6759 * Should we register nexthop for this EVPN prefix for nexthop tracking?
6760 */
6761 bool bgp_evpn_is_prefix_nht_supported(const struct prefix *pfx)
6762 {
6763 struct prefix_evpn *evp = (struct prefix_evpn *)pfx;
6764
6765 /*
6766 * EVPN routes should be marked as valid only if the nexthop is
6767 * reachable. Only if this happens, the route should be imported
6768 * (into VNI or VRF routing tables) and/or advertised.
6769 * Note: This is currently applied for EVPN type-1, type-2,
6770 * type-3, type-4 and type-5 routes.
6771 * It may be tweaked later on for other routes, or
6772 * even removed completely when all routes are handled.
6773 */
6774 if (pfx && pfx->family == AF_EVPN
6775 && (evp->prefix.route_type == BGP_EVPN_MAC_IP_ROUTE
6776 || evp->prefix.route_type == BGP_EVPN_AD_ROUTE
6777 || evp->prefix.route_type == BGP_EVPN_ES_ROUTE
6778 || evp->prefix.route_type == BGP_EVPN_IMET_ROUTE
6779 || evp->prefix.route_type == BGP_EVPN_IP_PREFIX_ROUTE))
6780 return true;
6781
6782 return false;
6783 }
6784
6785 static void *bgp_evpn_remote_ip_hash_alloc(void *p)
6786 {
6787 const struct evpn_remote_ip *key = (const struct evpn_remote_ip *)p;
6788 struct evpn_remote_ip *ip;
6789
6790 ip = XMALLOC(MTYPE_EVPN_REMOTE_IP, sizeof(struct evpn_remote_ip));
6791 *ip = *key;
6792 ip->macip_path_list = list_new();
6793
6794 return ip;
6795 }
6796
6797 static unsigned int bgp_evpn_remote_ip_hash_key_make(const void *p)
6798 {
6799 const struct evpn_remote_ip *ip = p;
6800 const struct ipaddr *addr = &ip->addr;
6801
6802 if (IS_IPADDR_V4(addr))
6803 return jhash_1word(addr->ipaddr_v4.s_addr, 0);
6804
6805 return jhash2(addr->ipaddr_v6.s6_addr32,
6806 array_size(addr->ipaddr_v6.s6_addr32), 0);
6807 }
6808
6809 static bool bgp_evpn_remote_ip_hash_cmp(const void *p1, const void *p2)
6810 {
6811 const struct evpn_remote_ip *ip1 = p1;
6812 const struct evpn_remote_ip *ip2 = p2;
6813
6814 return !ipaddr_cmp(&ip1->addr, &ip2->addr);
6815 }
6816
6817 static void bgp_evpn_remote_ip_hash_init(struct bgpevpn *vpn)
6818 {
6819 if (!evpn_resolve_overlay_index())
6820 return;
6821
6822 vpn->remote_ip_hash = hash_create(bgp_evpn_remote_ip_hash_key_make,
6823 bgp_evpn_remote_ip_hash_cmp,
6824 "BGP EVPN remote IP hash");
6825 }
6826
6827 static void bgp_evpn_remote_ip_hash_free(struct hash_bucket *bucket, void *args)
6828 {
6829 struct evpn_remote_ip *ip = (struct evpn_remote_ip *)bucket->data;
6830 struct bgpevpn *vpn = (struct bgpevpn *)args;
6831
6832 bgp_evpn_remote_ip_process_nexthops(vpn, &ip->addr, false);
6833
6834 list_delete(&ip->macip_path_list);
6835
6836 hash_release(vpn->remote_ip_hash, ip);
6837 XFREE(MTYPE_EVPN_REMOTE_IP, ip);
6838 }
6839
6840 static void bgp_evpn_remote_ip_hash_destroy(struct bgpevpn *vpn)
6841 {
6842 if (!evpn_resolve_overlay_index() || vpn->remote_ip_hash == NULL)
6843 return;
6844
6845 hash_iterate(vpn->remote_ip_hash,
6846 (void (*)(struct hash_bucket *, void *))bgp_evpn_remote_ip_hash_free,
6847 vpn);
6848
6849 hash_free(vpn->remote_ip_hash);
6850 vpn->remote_ip_hash = NULL;
6851 }
6852
6853 /* Add a remote MAC/IP route to hash table */
6854 static void bgp_evpn_remote_ip_hash_add(struct bgpevpn *vpn,
6855 struct bgp_path_info *pi)
6856 {
6857 struct evpn_remote_ip tmp;
6858 struct evpn_remote_ip *ip;
6859 struct prefix_evpn *evp;
6860
6861 if (!evpn_resolve_overlay_index())
6862 return;
6863
6864 if (pi->type != ZEBRA_ROUTE_BGP || pi->sub_type != BGP_ROUTE_IMPORTED
6865 || !CHECK_FLAG(pi->flags, BGP_PATH_VALID))
6866 return;
6867
6868 evp = (struct prefix_evpn *)&pi->net->p;
6869
6870 if (evp->family != AF_EVPN
6871 || evp->prefix.route_type != BGP_EVPN_MAC_IP_ROUTE
6872 || is_evpn_prefix_ipaddr_none(evp))
6873 return;
6874
6875 tmp.addr = evp->prefix.macip_addr.ip;
6876 ip = hash_lookup(vpn->remote_ip_hash, &tmp);
6877 if (ip) {
6878 if (listnode_lookup(ip->macip_path_list, pi) != NULL)
6879 return;
6880 (void)listnode_add(ip->macip_path_list, pi);
6881 return;
6882 }
6883
6884 ip = hash_get(vpn->remote_ip_hash, &tmp, bgp_evpn_remote_ip_hash_alloc);
6885 (void)listnode_add(ip->macip_path_list, pi);
6886
6887 bgp_evpn_remote_ip_process_nexthops(vpn, &ip->addr, true);
6888 }
6889
6890 /* Delete a remote MAC/IP route from hash table */
6891 static void bgp_evpn_remote_ip_hash_del(struct bgpevpn *vpn,
6892 struct bgp_path_info *pi)
6893 {
6894 struct evpn_remote_ip tmp;
6895 struct evpn_remote_ip *ip;
6896 struct prefix_evpn *evp;
6897
6898 if (!evpn_resolve_overlay_index())
6899 return;
6900
6901 evp = (struct prefix_evpn *)&pi->net->p;
6902
6903 if (evp->family != AF_EVPN
6904 || evp->prefix.route_type != BGP_EVPN_MAC_IP_ROUTE
6905 || is_evpn_prefix_ipaddr_none(evp))
6906 return;
6907
6908 tmp.addr = evp->prefix.macip_addr.ip;
6909 ip = hash_lookup(vpn->remote_ip_hash, &tmp);
6910 if (ip == NULL)
6911 return;
6912
6913 listnode_delete(ip->macip_path_list, pi);
6914
6915 if (ip->macip_path_list->count == 0) {
6916 bgp_evpn_remote_ip_process_nexthops(vpn, &ip->addr, false);
6917 hash_release(vpn->remote_ip_hash, ip);
6918 list_delete(&ip->macip_path_list);
6919 XFREE(MTYPE_EVPN_REMOTE_IP, ip);
6920 }
6921 }
6922
6923 static void bgp_evpn_remote_ip_hash_iterate(struct bgpevpn *vpn,
6924 void (*func)(struct hash_bucket *,
6925 void *),
6926 void *arg)
6927 {
6928 if (!evpn_resolve_overlay_index())
6929 return;
6930
6931 hash_iterate(vpn->remote_ip_hash, func, arg);
6932 }
6933
6934 static void show_remote_ip_entry(struct hash_bucket *bucket, void *args)
6935 {
6936 char buf[INET6_ADDRSTRLEN];
6937 struct listnode *node = NULL;
6938 struct bgp_path_info *pi = NULL;
6939 struct vty *vty = (struct vty *)args;
6940 struct evpn_remote_ip *ip = (struct evpn_remote_ip *)bucket->data;
6941
6942 vty_out(vty, " Remote IP: %s\n",
6943 ipaddr2str(&ip->addr, buf, sizeof(buf)));
6944 vty_out(vty, " Linked MAC/IP routes:\n");
6945 for (ALL_LIST_ELEMENTS_RO(ip->macip_path_list, node, pi))
6946 vty_out(vty, " %pFX\n", &pi->net->p);
6947 }
6948
6949 void bgp_evpn_show_remote_ip_hash(struct hash_bucket *bucket, void *args)
6950 {
6951 struct bgpevpn *vpn = (struct bgpevpn *)bucket->data;
6952 struct vty *vty = (struct vty *)args;
6953
6954 vty_out(vty, "VNI: %u\n", vpn->vni);
6955 bgp_evpn_remote_ip_hash_iterate(
6956 vpn,
6957 (void (*)(struct hash_bucket *, void *))show_remote_ip_entry,
6958 vty);
6959 vty_out(vty, "\n");
6960 }
6961
6962 static void bgp_evpn_remote_ip_hash_link_nexthop(struct hash_bucket *bucket,
6963 void *args)
6964 {
6965 struct evpn_remote_ip *ip = (struct evpn_remote_ip *)bucket->data;
6966 struct bgpevpn *vpn = (struct bgpevpn *)args;
6967
6968 bgp_evpn_remote_ip_process_nexthops(vpn, &ip->addr, true);
6969 }
6970
6971 static void bgp_evpn_remote_ip_hash_unlink_nexthop(struct hash_bucket *bucket,
6972 void *args)
6973 {
6974 struct evpn_remote_ip *ip = (struct evpn_remote_ip *)bucket->data;
6975 struct bgpevpn *vpn = (struct bgpevpn *)args;
6976
6977 bgp_evpn_remote_ip_process_nexthops(vpn, &ip->addr, false);
6978 }
6979
6980 static unsigned int vni_svi_hash_key_make(const void *p)
6981 {
6982 const struct bgpevpn *vpn = p;
6983
6984 return jhash_1word(vpn->svi_ifindex, 0);
6985 }
6986
6987 static bool vni_svi_hash_cmp(const void *p1, const void *p2)
6988 {
6989 const struct bgpevpn *vpn1 = p1;
6990 const struct bgpevpn *vpn2 = p2;
6991
6992 return (vpn1->svi_ifindex == vpn2->svi_ifindex);
6993 }
6994
6995 static struct bgpevpn *bgp_evpn_vni_svi_hash_lookup(struct bgp *bgp,
6996 ifindex_t svi)
6997 {
6998 struct bgpevpn *vpn;
6999 struct bgpevpn tmp;
7000
7001 memset(&tmp, 0, sizeof(tmp));
7002 tmp.svi_ifindex = svi;
7003 vpn = hash_lookup(bgp->vni_svi_hash, &tmp);
7004 return vpn;
7005 }
7006
7007 static void bgp_evpn_link_to_vni_svi_hash(struct bgp *bgp, struct bgpevpn *vpn)
7008 {
7009 if (vpn->svi_ifindex == 0)
7010 return;
7011
7012 (void)hash_get(bgp->vni_svi_hash, vpn, hash_alloc_intern);
7013 }
7014
7015 static void bgp_evpn_unlink_from_vni_svi_hash(struct bgp *bgp,
7016 struct bgpevpn *vpn)
7017 {
7018 if (vpn->svi_ifindex == 0)
7019 return;
7020
7021 hash_release(bgp->vni_svi_hash, vpn);
7022 }
7023
7024 void bgp_evpn_show_vni_svi_hash(struct hash_bucket *bucket, void *args)
7025 {
7026 struct bgpevpn *evpn = (struct bgpevpn *)bucket->data;
7027 struct vty *vty = (struct vty *)args;
7028
7029 vty_out(vty, "SVI: %u VNI: %u\n", evpn->svi_ifindex, evpn->vni);
7030 }
7031
7032 /*
7033 * This function is called for a bgp_nexthop_cache entry when the nexthop is
7034 * gateway IP overlay index.
7035 * This function returns true if there is a remote MAC/IP route for the gateway
7036 * IP in the EVI of the nexthop SVI.
7037 */
7038 bool bgp_evpn_is_gateway_ip_resolved(struct bgp_nexthop_cache *bnc)
7039 {
7040 struct bgp *bgp_evpn = NULL;
7041 struct bgpevpn *vpn = NULL;
7042 struct evpn_remote_ip tmp;
7043 struct prefix *p;
7044
7045 if (!evpn_resolve_overlay_index())
7046 return false;
7047
7048 if (!bnc->nexthop || bnc->nexthop->ifindex == 0)
7049 return false;
7050
7051 bgp_evpn = bgp_get_evpn();
7052 if (!bgp_evpn)
7053 return false;
7054
7055 /*
7056 * Gateway IP is resolved by nht over SVI interface.
7057 * Use this SVI to find corresponding EVI(L2 context)
7058 */
7059 vpn = bgp_evpn_vni_svi_hash_lookup(bgp_evpn, bnc->nexthop->ifindex);
7060 if (!vpn)
7061 return false;
7062
7063 if (vpn->bgp_vrf != bnc->bgp)
7064 return false;
7065
7066 /*
7067 * Check if the gateway IP is present in the EVI remote_ip_hash table
7068 * which stores all the remote IP addresses received via MAC/IP routes
7069 * in this EVI
7070 */
7071 memset(&tmp, 0, sizeof(tmp));
7072
7073 p = &bnc->prefix;
7074 if (p->family == AF_INET) {
7075 tmp.addr.ipa_type = IPADDR_V4;
7076 memcpy(&(tmp.addr.ipaddr_v4), &(p->u.prefix4),
7077 sizeof(struct in_addr));
7078 } else if (p->family == AF_INET6) {
7079 tmp.addr.ipa_type = IPADDR_V6;
7080 memcpy(&(tmp.addr.ipaddr_v6), &(p->u.prefix6),
7081 sizeof(struct in6_addr));
7082 } else
7083 return false;
7084
7085 if (hash_lookup(vpn->remote_ip_hash, &tmp) == NULL)
7086 return false;
7087
7088 return true;
7089 }
7090
7091 /* Resolve/Unresolve nexthops when a MAC/IP route is added/deleted */
7092 static void bgp_evpn_remote_ip_process_nexthops(struct bgpevpn *vpn,
7093 struct ipaddr *addr,
7094 bool resolve)
7095 {
7096 afi_t afi;
7097 struct prefix p;
7098 struct bgp_nexthop_cache *bnc;
7099 struct bgp_nexthop_cache_head *tree = NULL;
7100
7101 if (!vpn->bgp_vrf || vpn->svi_ifindex == 0)
7102 return;
7103
7104 memset(&p, 0, sizeof(p));
7105
7106 if (addr->ipa_type == IPADDR_V4) {
7107 afi = AFI_IP;
7108 p.family = AF_INET;
7109 memcpy(&(p.u.prefix4), &(addr->ipaddr_v4),
7110 sizeof(struct in_addr));
7111 p.prefixlen = IPV4_MAX_BITLEN;
7112 } else if (addr->ipa_type == IPADDR_V6) {
7113 afi = AFI_IP6;
7114 p.family = AF_INET6;
7115 memcpy(&(p.u.prefix6), &(addr->ipaddr_v6),
7116 sizeof(struct in6_addr));
7117 p.prefixlen = IPV6_MAX_BITLEN;
7118 } else
7119 return;
7120
7121 tree = &vpn->bgp_vrf->nexthop_cache_table[afi];
7122 bnc = bnc_find(tree, &p, 0, 0);
7123
7124 if (!bnc || !bnc->is_evpn_gwip_nexthop)
7125 return;
7126
7127 if (!bnc->nexthop || bnc->nexthop->ifindex != vpn->svi_ifindex)
7128 return;
7129
7130 if (BGP_DEBUG(nht, NHT))
7131 zlog_debug("%s(%u): vni %u mac/ip %s for NH %pFX",
7132 vpn->bgp_vrf->name_pretty, vpn->tenant_vrf_id,
7133 vpn->vni, (resolve ? "add" : "delete"),
7134 &bnc->prefix);
7135
7136 /*
7137 * MAC/IP route or SVI or tenant vrf being added to EVI.
7138 * Set nexthop as valid only if it is already L3 reachable
7139 */
7140 if (resolve && bnc->flags & BGP_NEXTHOP_EVPN_INCOMPLETE) {
7141 bnc->flags &= ~BGP_NEXTHOP_EVPN_INCOMPLETE;
7142 bnc->flags |= BGP_NEXTHOP_VALID;
7143 bnc->change_flags |= BGP_NEXTHOP_MACIP_CHANGED;
7144 evaluate_paths(bnc);
7145 }
7146
7147 /* MAC/IP route or SVI or tenant vrf being deleted from EVI */
7148 if (!resolve && bnc->flags & BGP_NEXTHOP_VALID) {
7149 bnc->flags &= ~BGP_NEXTHOP_VALID;
7150 bnc->flags |= BGP_NEXTHOP_EVPN_INCOMPLETE;
7151 bnc->change_flags |= BGP_NEXTHOP_MACIP_CHANGED;
7152 evaluate_paths(bnc);
7153 }
7154 }
7155
7156 void bgp_evpn_handle_resolve_overlay_index_set(struct hash_bucket *bucket,
7157 void *arg)
7158 {
7159 struct bgpevpn *vpn = (struct bgpevpn *)bucket->data;
7160 struct bgp_dest *dest;
7161 struct bgp_path_info *pi;
7162
7163 bgp_evpn_remote_ip_hash_init(vpn);
7164
7165 for (dest = bgp_table_top(vpn->ip_table); dest;
7166 dest = bgp_route_next(dest))
7167 for (pi = bgp_dest_get_bgp_path_info(dest); pi; pi = pi->next)
7168 bgp_evpn_remote_ip_hash_add(vpn, pi);
7169 }
7170
7171 void bgp_evpn_handle_resolve_overlay_index_unset(struct hash_bucket *bucket,
7172 void *arg)
7173 {
7174 struct bgpevpn *vpn = (struct bgpevpn *)bucket->data;
7175
7176 bgp_evpn_remote_ip_hash_destroy(vpn);
7177 }
7178
7179 /*
7180 * Helper function for getting the correct label index for l3vni.
7181 *
7182 * Returns the label with the l3vni of the path's label stack.
7183 *
7184 * L3vni is always last label. Type5 will only
7185 * have one label, Type2 will have two.
7186 *
7187 */
7188 mpls_label_t *bgp_evpn_path_info_labels_get_l3vni(mpls_label_t *labels,
7189 uint32_t num_labels)
7190 {
7191 if (!labels)
7192 return NULL;
7193
7194 if (!num_labels)
7195 return NULL;
7196
7197 return &labels[num_labels - 1];
7198 }
7199
7200 /*
7201 * Returns the l3vni of the path converted from the label stack.
7202 */
7203 vni_t bgp_evpn_path_info_get_l3vni(const struct bgp_path_info *pi)
7204 {
7205 if (!pi->extra)
7206 return 0;
7207
7208 return label2vni(bgp_evpn_path_info_labels_get_l3vni(
7209 pi->extra->label, pi->extra->num_labels));
7210 }
7211
7212 /*
7213 * Returns true if the l3vni of any of this path doesn't match vrf's l3vni.
7214 */
7215 static bool bgp_evpn_path_is_dvni(const struct bgp *bgp_vrf,
7216 const struct bgp_path_info *pi)
7217 {
7218 vni_t vni = 0;
7219
7220 vni = bgp_evpn_path_info_get_l3vni(pi);
7221
7222 if ((vni > 0) && (vni != bgp_vrf->l3vni))
7223 return true;
7224
7225 return false;
7226 }
7227
7228 /*
7229 * Returns true if the l3vni of any of the mpath's doesn't match vrf's l3vni.
7230 */
7231 bool bgp_evpn_mpath_has_dvni(const struct bgp *bgp_vrf,
7232 struct bgp_path_info *mpinfo)
7233 {
7234 for (; mpinfo; mpinfo = bgp_path_info_mpath_next(mpinfo)) {
7235 if (bgp_evpn_path_is_dvni(bgp_vrf, mpinfo))
7236 return true;
7237 }
7238
7239 return false;
7240 }