]> git.proxmox.com Git - mirror_frr.git/blob - bgpd/bgp_evpn.c
bgpd: enable aggregation in evpn
[mirror_frr.git] / bgpd / bgp_evpn.c
1 /* Ethernet-VPN Packet and vty Processing File
2 * Copyright (C) 2016 6WIND
3 * Copyright (C) 2017 Cumulus Networks, Inc.
4 *
5 * This file is part of FRR.
6 *
7 * FRRouting is free software; you can redistribute it and/or modify it
8 * under the terms of the GNU General Public License as published by the
9 * Free Software Foundation; either version 2, or (at your option) any
10 * later version.
11 *
12 * FRRouting is distributed in the hope that it will be useful, but
13 * WITHOUT ANY WARRANTY; without even the implied warranty of
14 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
15 * General Public License for more details.
16 *
17 * You should have received a copy of the GNU General Public License along
18 * with this program; see the file COPYING; if not, write to the Free Software
19 * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
20 */
21
22 #include <zebra.h>
23
24 #include "command.h"
25 #include "filter.h"
26 #include "prefix.h"
27 #include "log.h"
28 #include "memory.h"
29 #include "stream.h"
30 #include "hash.h"
31 #include "jhash.h"
32 #include "zclient.h"
33
34 #include "bgpd/bgp_attr_evpn.h"
35 #include "bgpd/bgpd.h"
36 #include "bgpd/bgp_table.h"
37 #include "bgpd/bgp_route.h"
38 #include "bgpd/bgp_attr.h"
39 #include "bgpd/bgp_mplsvpn.h"
40 #include "bgpd/bgp_label.h"
41 #include "bgpd/bgp_evpn.h"
42 #include "bgpd/bgp_evpn_private.h"
43 #include "bgpd/bgp_ecommunity.h"
44 #include "bgpd/bgp_encap_types.h"
45 #include "bgpd/bgp_debug.h"
46 #include "bgpd/bgp_errors.h"
47 #include "bgpd/bgp_aspath.h"
48 #include "bgpd/bgp_zebra.h"
49 #include "bgpd/bgp_nexthop.h"
50
51 /*
52 * Definitions and external declarations.
53 */
54 extern struct zclient *zclient;
55
56 DEFINE_QOBJ_TYPE(bgpevpn)
57 DEFINE_QOBJ_TYPE(evpnes)
58
59
60 /*
61 * Static function declarations
62 */
63 static void delete_evpn_route_entry(struct bgp *bgp,
64 afi_t afi, safi_t safi,
65 struct bgp_node *rn,
66 struct bgp_info **ri);
67 static int delete_all_vni_routes(struct bgp *bgp, struct bgpevpn *vpn);
68
69 /*
70 * Private functions.
71 */
72
73 /* compare two IPV4 VTEP IPs */
74 static int evpn_vtep_ip_cmp(const void *p1, const void *p2)
75 {
76 const struct in_addr *ip1 = p1;
77 const struct in_addr *ip2 = p2;
78
79 if (!ip1 && !ip2)
80 return 1;
81 if (!ip1 || !ip2)
82 return 0;
83 return (ip1->s_addr == ip2->s_addr);
84 }
85
86 /*
87 * Make hash key for ESI.
88 */
89 static unsigned int esi_hash_keymake(void *p)
90 {
91 struct evpnes *pes = p;
92 const void *pnt = (void *)pes->esi.val;
93
94 return jhash(pnt, ESI_BYTES, 0xa5a5a55a);
95 }
96
97 /*
98 * Compare two ESIs.
99 */
100 static int esi_cmp(const void *p1, const void *p2)
101 {
102 const struct evpnes *pes1 = p1;
103 const struct evpnes *pes2 = p2;
104
105 if (pes1 == NULL && pes2 == NULL)
106 return 1;
107
108 if (pes1 == NULL || pes2 == NULL)
109 return 0;
110
111 return (memcmp(pes1->esi.val, pes2->esi.val, ESI_BYTES) == 0);
112 }
113
114 /*
115 * Make vni hash key.
116 */
117 static unsigned int vni_hash_key_make(void *p)
118 {
119 struct bgpevpn *vpn = p;
120 return (jhash_1word(vpn->vni, 0));
121 }
122
123 /*
124 * Comparison function for vni hash
125 */
126 static int vni_hash_cmp(const void *p1, const void *p2)
127 {
128 const struct bgpevpn *vpn1 = p1;
129 const struct bgpevpn *vpn2 = p2;
130
131 if (!vpn1 && !vpn2)
132 return 1;
133 if (!vpn1 || !vpn2)
134 return 0;
135 return (vpn1->vni == vpn2->vni);
136 }
137
138 /*
139 * Make vrf import route target hash key.
140 */
141 static unsigned int vrf_import_rt_hash_key_make(void *p)
142 {
143 struct vrf_irt_node *irt = p;
144 char *pnt = irt->rt.val;
145
146 return jhash(pnt, 8, 0x5abc1234);
147 }
148
149 /*
150 * Comparison function for vrf import rt hash
151 */
152 static int vrf_import_rt_hash_cmp(const void *p1, const void *p2)
153 {
154 const struct vrf_irt_node *irt1 = p1;
155 const struct vrf_irt_node *irt2 = p2;
156
157 if (irt1 == NULL && irt2 == NULL)
158 return 1;
159
160 if (irt1 == NULL || irt2 == NULL)
161 return 0;
162
163 return (memcmp(irt1->rt.val, irt2->rt.val, ECOMMUNITY_SIZE) == 0);
164 }
165
166 /*
167 * Create a new vrf import_rt in default instance
168 */
169 static struct vrf_irt_node *vrf_import_rt_new(struct ecommunity_val *rt)
170 {
171 struct bgp *bgp_def = NULL;
172 struct vrf_irt_node *irt;
173
174 bgp_def = bgp_get_default();
175 if (!bgp_def) {
176 flog_err(EC_BGP_NO_DFLT,
177 "vrf import rt new - def instance not created yet");
178 return NULL;
179 }
180
181 irt = XCALLOC(MTYPE_BGP_EVPN_VRF_IMPORT_RT,
182 sizeof(struct vrf_irt_node));
183 if (!irt)
184 return NULL;
185
186 irt->rt = *rt;
187 irt->vrfs = list_new();
188
189 /* Add to hash */
190 if (!hash_get(bgp_def->vrf_import_rt_hash, irt, hash_alloc_intern)) {
191 XFREE(MTYPE_BGP_EVPN_VRF_IMPORT_RT, irt);
192 return NULL;
193 }
194
195 return irt;
196 }
197
198 /*
199 * Free the vrf import rt node
200 */
201 static void vrf_import_rt_free(struct vrf_irt_node *irt)
202 {
203 struct bgp *bgp_def = NULL;
204
205 bgp_def = bgp_get_default();
206 if (!bgp_def) {
207 flog_err(EC_BGP_NO_DFLT,
208 "vrf import rt free - def instance not created yet");
209 return;
210 }
211
212 hash_release(bgp_def->vrf_import_rt_hash, irt);
213 list_delete_and_null(&irt->vrfs);
214 XFREE(MTYPE_BGP_EVPN_VRF_IMPORT_RT, irt);
215 }
216
217 /*
218 * Function to lookup Import RT node - used to map a RT to set of
219 * VNIs importing routes with that RT.
220 */
221 static struct vrf_irt_node *lookup_vrf_import_rt(struct ecommunity_val *rt)
222 {
223 struct bgp *bgp_def = NULL;
224 struct vrf_irt_node *irt;
225 struct vrf_irt_node tmp;
226
227 bgp_def = bgp_get_default();
228 if (!bgp_def) {
229 flog_err(EC_BGP_NO_DFLT,
230 "vrf import rt lookup - def instance not created yet");
231 return NULL;
232 }
233
234 memset(&tmp, 0, sizeof(struct vrf_irt_node));
235 memcpy(&tmp.rt, rt, ECOMMUNITY_SIZE);
236 irt = hash_lookup(bgp_def->vrf_import_rt_hash, &tmp);
237 return irt;
238 }
239
240 /*
241 * Is specified VRF present on the RT's list of "importing" VRFs?
242 */
243 static int is_vrf_present_in_irt_vrfs(struct list *vrfs, struct bgp *bgp_vrf)
244 {
245 struct listnode *node = NULL, *nnode = NULL;
246 struct bgp *tmp_bgp_vrf = NULL;
247
248 for (ALL_LIST_ELEMENTS(vrfs, node, nnode, tmp_bgp_vrf)) {
249 if (tmp_bgp_vrf == bgp_vrf)
250 return 1;
251 }
252 return 0;
253 }
254
255 /*
256 * Make import route target hash key.
257 */
258 static unsigned int import_rt_hash_key_make(void *p)
259 {
260 struct irt_node *irt = p;
261 char *pnt = irt->rt.val;
262
263 return jhash(pnt, 8, 0xdeadbeef);
264 }
265
266 /*
267 * Comparison function for import rt hash
268 */
269 static int import_rt_hash_cmp(const void *p1, const void *p2)
270 {
271 const struct irt_node *irt1 = p1;
272 const struct irt_node *irt2 = p2;
273
274 if (irt1 == NULL && irt2 == NULL)
275 return 1;
276
277 if (irt1 == NULL || irt2 == NULL)
278 return 0;
279
280 return (memcmp(irt1->rt.val, irt2->rt.val, ECOMMUNITY_SIZE) == 0);
281 }
282
283 /*
284 * Create a new import_rt
285 */
286 static struct irt_node *import_rt_new(struct bgp *bgp,
287 struct ecommunity_val *rt)
288 {
289 struct irt_node *irt;
290
291 if (!bgp)
292 return NULL;
293
294 irt = XCALLOC(MTYPE_BGP_EVPN_IMPORT_RT, sizeof(struct irt_node));
295 if (!irt)
296 return NULL;
297
298 irt->rt = *rt;
299 irt->vnis = list_new();
300
301 /* Add to hash */
302 if (!hash_get(bgp->import_rt_hash, irt, hash_alloc_intern)) {
303 XFREE(MTYPE_BGP_EVPN_IMPORT_RT, irt);
304 return NULL;
305 }
306
307 return irt;
308 }
309
310 /*
311 * Free the import rt node
312 */
313 static void import_rt_free(struct bgp *bgp, struct irt_node *irt)
314 {
315 hash_release(bgp->import_rt_hash, irt);
316 list_delete_and_null(&irt->vnis);
317 XFREE(MTYPE_BGP_EVPN_IMPORT_RT, irt);
318 }
319
320 /*
321 * Function to lookup Import RT node - used to map a RT to set of
322 * VNIs importing routes with that RT.
323 */
324 static struct irt_node *lookup_import_rt(struct bgp *bgp,
325 struct ecommunity_val *rt)
326 {
327 struct irt_node *irt;
328 struct irt_node tmp;
329
330 memset(&tmp, 0, sizeof(struct irt_node));
331 memcpy(&tmp.rt, rt, ECOMMUNITY_SIZE);
332 irt = hash_lookup(bgp->import_rt_hash, &tmp);
333 return irt;
334 }
335
336 /*
337 * Is specified VNI present on the RT's list of "importing" VNIs?
338 */
339 static int is_vni_present_in_irt_vnis(struct list *vnis, struct bgpevpn *vpn)
340 {
341 struct listnode *node, *nnode;
342 struct bgpevpn *tmp_vpn;
343
344 for (ALL_LIST_ELEMENTS(vnis, node, nnode, tmp_vpn)) {
345 if (tmp_vpn == vpn)
346 return 1;
347 }
348
349 return 0;
350 }
351
352 /*
353 * Compare Route Targets.
354 */
355 static int evpn_route_target_cmp(struct ecommunity *ecom1,
356 struct ecommunity *ecom2)
357 {
358 if (ecom1 && !ecom2)
359 return -1;
360
361 if (!ecom1 && ecom2)
362 return 1;
363
364 if (!ecom1 && !ecom2)
365 return 0;
366
367 if (ecom1->str && !ecom2->str)
368 return -1;
369
370 if (!ecom1->str && ecom2->str)
371 return 1;
372
373 if (!ecom1->str && !ecom2->str)
374 return 0;
375
376 return strcmp(ecom1->str, ecom2->str);
377 }
378
379 static void evpn_xxport_delete_ecomm(void *val)
380 {
381 struct ecommunity *ecomm = val;
382 ecommunity_free(&ecomm);
383 }
384
385 /*
386 * Mask off global-admin field of specified extended community (RT),
387 * just retain the local-admin field.
388 */
389 static inline void mask_ecom_global_admin(struct ecommunity_val *dst,
390 struct ecommunity_val *src)
391 {
392 uint8_t type;
393
394 type = src->val[0];
395 dst->val[0] = 0;
396 if (type == ECOMMUNITY_ENCODE_AS) {
397 dst->val[2] = dst->val[3] = 0;
398 } else if (type == ECOMMUNITY_ENCODE_AS4
399 || type == ECOMMUNITY_ENCODE_IP) {
400 dst->val[2] = dst->val[3] = 0;
401 dst->val[4] = dst->val[5] = 0;
402 }
403 }
404
405 /*
406 * Map one RT to specified VRF.
407 * bgp_vrf = BGP vrf instance
408 */
409 static void map_vrf_to_rt(struct bgp *bgp_vrf, struct ecommunity_val *eval)
410 {
411 struct vrf_irt_node *irt = NULL;
412 struct ecommunity_val eval_tmp;
413
414 /* If using "automatic" RT,
415 * we only care about the local-admin sub-field.
416 * This is to facilitate using L3VNI(VRF-VNI)
417 * as the RT for EBGP peering too.
418 */
419 memcpy(&eval_tmp, eval, ECOMMUNITY_SIZE);
420 if (!CHECK_FLAG(bgp_vrf->vrf_flags, BGP_VRF_IMPORT_RT_CFGD))
421 mask_ecom_global_admin(&eval_tmp, eval);
422
423 irt = lookup_vrf_import_rt(&eval_tmp);
424 if (irt && is_vrf_present_in_irt_vrfs(irt->vrfs, bgp_vrf))
425 /* Already mapped. */
426 return;
427
428 if (!irt)
429 irt = vrf_import_rt_new(&eval_tmp);
430
431 /* Add VRF to the list for this RT. */
432 listnode_add(irt->vrfs, bgp_vrf);
433 }
434
435 /*
436 * Unmap specified VRF from specified RT. If there are no other
437 * VRFs for this RT, then the RT hash is deleted.
438 * bgp_vrf: BGP VRF specific instance
439 */
440 static void unmap_vrf_from_rt(struct bgp *bgp_vrf, struct vrf_irt_node *irt)
441 {
442 /* Delete VRF from list for this RT. */
443 listnode_delete(irt->vrfs, bgp_vrf);
444 if (!listnode_head(irt->vrfs)) {
445 vrf_import_rt_free(irt);
446 }
447 }
448
449 /*
450 * Map one RT to specified VNI.
451 */
452 static void map_vni_to_rt(struct bgp *bgp, struct bgpevpn *vpn,
453 struct ecommunity_val *eval)
454 {
455 struct irt_node *irt;
456 struct ecommunity_val eval_tmp;
457
458 /* If using "automatic" RT, we only care about the local-admin
459 * sub-field.
460 * This is to facilitate using VNI as the RT for EBGP peering too.
461 */
462 memcpy(&eval_tmp, eval, ECOMMUNITY_SIZE);
463 if (!is_import_rt_configured(vpn))
464 mask_ecom_global_admin(&eval_tmp, eval);
465
466 irt = lookup_import_rt(bgp, &eval_tmp);
467 if (irt)
468 if (is_vni_present_in_irt_vnis(irt->vnis, vpn))
469 /* Already mapped. */
470 return;
471
472 if (!irt) {
473 irt = import_rt_new(bgp, &eval_tmp);
474 assert(irt);
475 }
476
477 /* Add VNI to the hash list for this RT. */
478 listnode_add(irt->vnis, vpn);
479 }
480
481 /*
482 * Unmap specified VNI from specified RT. If there are no other
483 * VNIs for this RT, then the RT hash is deleted.
484 */
485 static void unmap_vni_from_rt(struct bgp *bgp, struct bgpevpn *vpn,
486 struct irt_node *irt)
487 {
488 /* Delete VNI from hash list for this RT. */
489 listnode_delete(irt->vnis, vpn);
490 if (!listnode_head(irt->vnis)) {
491 import_rt_free(bgp, irt);
492 }
493 }
494
495 /*
496 * Create RT extended community automatically from passed information:
497 * of the form AS:VNI.
498 * NOTE: We use only the lower 16 bits of the AS. This is sufficient as
499 * the need is to get a RT value that will be unique across different
500 * VNIs but the same across routers (in the same AS) for a particular
501 * VNI.
502 */
503 static void form_auto_rt(struct bgp *bgp, vni_t vni, struct list *rtl)
504 {
505 struct ecommunity_val eval;
506 struct ecommunity *ecomadd;
507
508 if (bgp->advertise_autort_rfc8365)
509 vni |= EVPN_AUTORT_VXLAN;
510 encode_route_target_as((bgp->as & 0xFFFF), vni, &eval);
511
512 ecomadd = ecommunity_new();
513 ecommunity_add_val(ecomadd, &eval);
514 listnode_add_sort(rtl, ecomadd);
515 }
516
517 /*
518 * Derive RD and RT for a VNI automatically. Invoked at the time of
519 * creation of a VNI.
520 */
521 static void derive_rd_rt_for_vni(struct bgp *bgp, struct bgpevpn *vpn)
522 {
523 bgp_evpn_derive_auto_rd(bgp, vpn);
524 bgp_evpn_derive_auto_rt_import(bgp, vpn);
525 bgp_evpn_derive_auto_rt_export(bgp, vpn);
526 }
527
528 /*
529 * Convert nexthop (remote VTEP IP) into an IPv6 address.
530 */
531 static void evpn_convert_nexthop_to_ipv6(struct attr *attr)
532 {
533 if (BGP_ATTR_NEXTHOP_AFI_IP6(attr))
534 return;
535 ipv4_to_ipv4_mapped_ipv6(&attr->mp_nexthop_global, attr->nexthop);
536 attr->mp_nexthop_len = IPV6_MAX_BYTELEN;
537 }
538
539 /*
540 * Add (update) or delete MACIP from zebra.
541 */
542 static int bgp_zebra_send_remote_macip(struct bgp *bgp, struct bgpevpn *vpn,
543 struct prefix_evpn *p,
544 struct in_addr remote_vtep_ip, int add,
545 uint8_t flags, uint32_t seq)
546 {
547 struct stream *s;
548 int ipa_len;
549 char buf1[ETHER_ADDR_STRLEN];
550 char buf2[INET6_ADDRSTRLEN];
551 char buf3[INET6_ADDRSTRLEN];
552
553 /* Check socket. */
554 if (!zclient || zclient->sock < 0)
555 return 0;
556
557 /* Don't try to register if Zebra doesn't know of this instance. */
558 if (!IS_BGP_INST_KNOWN_TO_ZEBRA(bgp))
559 return 0;
560
561 s = zclient->obuf;
562 stream_reset(s);
563
564 zclient_create_header(
565 s, add ? ZEBRA_REMOTE_MACIP_ADD : ZEBRA_REMOTE_MACIP_DEL,
566 bgp->vrf_id);
567 stream_putl(s, vpn->vni);
568 stream_put(s, &p->prefix.macip_addr.mac.octet, ETH_ALEN); /* Mac Addr */
569 /* IP address length and IP address, if any. */
570 if (is_evpn_prefix_ipaddr_none(p))
571 stream_putl(s, 0);
572 else {
573 ipa_len = is_evpn_prefix_ipaddr_v4(p) ? IPV4_MAX_BYTELEN
574 : IPV6_MAX_BYTELEN;
575 stream_putl(s, ipa_len);
576 stream_put(s, &p->prefix.macip_addr.ip.ip.addr, ipa_len);
577 }
578 stream_put_in_addr(s, &remote_vtep_ip);
579
580 /* TX flags - MAC sticky status and/or gateway mac */
581 /* Also TX the sequence number of the best route. */
582 if (add) {
583 stream_putc(s, flags);
584 stream_putl(s, seq);
585 }
586
587 stream_putw_at(s, 0, stream_get_endp(s));
588
589 if (bgp_debug_zebra(NULL))
590 zlog_debug(
591 "Tx %s MACIP, VNI %u MAC %s IP %s flags 0x%x seq %u remote VTEP %s",
592 add ? "ADD" : "DEL", vpn->vni,
593 prefix_mac2str(&p->prefix.macip_addr.mac,
594 buf1, sizeof(buf1)),
595 ipaddr2str(&p->prefix.macip_addr.ip,
596 buf3, sizeof(buf3)), flags, seq,
597 inet_ntop(AF_INET, &remote_vtep_ip, buf2,
598 sizeof(buf2)));
599
600 return zclient_send_message(zclient);
601 }
602
603 /*
604 * Add (update) or delete remote VTEP from zebra.
605 */
606 static int bgp_zebra_send_remote_vtep(struct bgp *bgp, struct bgpevpn *vpn,
607 struct prefix_evpn *p, int add)
608 {
609 struct stream *s;
610
611 /* Check socket. */
612 if (!zclient || zclient->sock < 0)
613 return 0;
614
615 /* Don't try to register if Zebra doesn't know of this instance. */
616 if (!IS_BGP_INST_KNOWN_TO_ZEBRA(bgp))
617 return 0;
618
619 s = zclient->obuf;
620 stream_reset(s);
621
622 zclient_create_header(
623 s, add ? ZEBRA_REMOTE_VTEP_ADD : ZEBRA_REMOTE_VTEP_DEL,
624 bgp->vrf_id);
625 stream_putl(s, vpn->vni);
626 if (is_evpn_prefix_ipaddr_v4(p))
627 stream_put_in_addr(s, &p->prefix.imet_addr.ip.ipaddr_v4);
628 else if (is_evpn_prefix_ipaddr_v6(p)) {
629 flog_err(
630 EC_BGP_VTEP_INVALID,
631 "Bad remote IP when trying to %s remote VTEP for VNI %u",
632 add ? "ADD" : "DEL", vpn->vni);
633 return -1;
634 }
635
636 stream_putw_at(s, 0, stream_get_endp(s));
637
638 if (bgp_debug_zebra(NULL))
639 zlog_debug("Tx %s Remote VTEP, VNI %u remote VTEP %s",
640 add ? "ADD" : "DEL", vpn->vni,
641 inet_ntoa(p->prefix.imet_addr.ip.ipaddr_v4));
642
643 return zclient_send_message(zclient);
644 }
645
646 /*
647 * Build extended community for EVPN ES (type-4) route
648 */
649 static void build_evpn_type4_route_extcomm(struct evpnes *es,
650 struct attr *attr)
651 {
652 struct ecommunity ecom_encap;
653 struct ecommunity ecom_es_rt;
654 struct ecommunity_val eval;
655 struct ecommunity_val eval_es_rt;
656 bgp_encap_types tnl_type;
657 struct ethaddr mac;
658
659 /* Encap */
660 tnl_type = BGP_ENCAP_TYPE_VXLAN;
661 memset(&ecom_encap, 0, sizeof(ecom_encap));
662 encode_encap_extcomm(tnl_type, &eval);
663 ecom_encap.size = 1;
664 ecom_encap.val = (uint8_t *)eval.val;
665 attr->ecommunity = ecommunity_dup(&ecom_encap);
666
667 /* ES import RT */
668 memset(&mac, 0, sizeof(struct ethaddr));
669 memset(&ecom_es_rt, 0, sizeof(ecom_es_rt));
670 es_get_system_mac(&es->esi, &mac);
671 encode_es_rt_extcomm(&eval_es_rt, &mac);
672 ecom_es_rt.size = 1;
673 ecom_es_rt.val = (uint8_t *)eval_es_rt.val;
674 attr->ecommunity =
675 ecommunity_merge(attr->ecommunity, &ecom_es_rt);
676
677 attr->flag |= ATTR_FLAG_BIT(BGP_ATTR_EXT_COMMUNITIES);
678 }
679
680 /*
681 * Build extended communities for EVPN prefix route.
682 */
683 static void build_evpn_type5_route_extcomm(struct bgp *bgp_vrf,
684 struct attr *attr)
685 {
686 struct ecommunity ecom_encap;
687 struct ecommunity ecom_rmac;
688 struct ecommunity_val eval;
689 struct ecommunity_val eval_rmac;
690 bgp_encap_types tnl_type;
691 struct listnode *node, *nnode;
692 struct ecommunity *ecom;
693 struct list *vrf_export_rtl = NULL;
694
695 /* Encap */
696 tnl_type = BGP_ENCAP_TYPE_VXLAN;
697 memset(&ecom_encap, 0, sizeof(ecom_encap));
698 encode_encap_extcomm(tnl_type, &eval);
699 ecom_encap.size = 1;
700 ecom_encap.val = (uint8_t *)eval.val;
701
702 /* Add Encap */
703 attr->ecommunity = ecommunity_dup(&ecom_encap);
704
705 /* Add the export RTs for L3VNI/VRF */
706 vrf_export_rtl = bgp_vrf->vrf_export_rtl;
707 for (ALL_LIST_ELEMENTS(vrf_export_rtl, node, nnode, ecom))
708 attr->ecommunity =
709 ecommunity_merge(attr->ecommunity, ecom);
710
711 /* add the router mac extended community */
712 if (!is_zero_mac(&attr->rmac)) {
713 memset(&ecom_rmac, 0, sizeof(ecom_rmac));
714 encode_rmac_extcomm(&eval_rmac, &attr->rmac);
715 ecom_rmac.size = 1;
716 ecom_rmac.val = (uint8_t *)eval_rmac.val;
717 attr->ecommunity =
718 ecommunity_merge(attr->ecommunity, &ecom_rmac);
719 }
720
721 attr->flag |= ATTR_FLAG_BIT(BGP_ATTR_EXT_COMMUNITIES);
722 }
723
724 /*
725 * Build extended communities for EVPN route.
726 * This function is applicable for type-2 and type-3 routes. The layer-2 RT
727 * and ENCAP extended communities are applicable for all routes.
728 * The default gateway extended community and MAC mobility (sticky) extended
729 * community are added as needed based on passed settings - only for type-2
730 * routes. Likewise, the layer-3 RT and Router MAC extended communities are
731 * added, if present, based on passed settings - only for non-link-local
732 * type-2 routes.
733 */
734 static void build_evpn_route_extcomm(struct bgpevpn *vpn, struct attr *attr,
735 int add_l3_ecomm)
736 {
737 struct ecommunity ecom_encap;
738 struct ecommunity ecom_sticky;
739 struct ecommunity ecom_default_gw;
740 struct ecommunity ecom_rmac;
741 struct ecommunity ecom_na;
742 struct ecommunity_val eval;
743 struct ecommunity_val eval_sticky;
744 struct ecommunity_val eval_default_gw;
745 struct ecommunity_val eval_rmac;
746 struct ecommunity_val eval_na;
747
748 bgp_encap_types tnl_type;
749 struct listnode *node, *nnode;
750 struct ecommunity *ecom;
751 uint32_t seqnum;
752 struct list *vrf_export_rtl = NULL;
753
754 /* Encap */
755 tnl_type = BGP_ENCAP_TYPE_VXLAN;
756 memset(&ecom_encap, 0, sizeof(ecom_encap));
757 encode_encap_extcomm(tnl_type, &eval);
758 ecom_encap.size = 1;
759 ecom_encap.val = (uint8_t *)eval.val;
760
761 /* Add Encap */
762 attr->ecommunity = ecommunity_dup(&ecom_encap);
763
764 /* Add the export RTs for L2VNI */
765 for (ALL_LIST_ELEMENTS(vpn->export_rtl, node, nnode, ecom))
766 attr->ecommunity = ecommunity_merge(attr->ecommunity, ecom);
767
768 /* Add the export RTs for L3VNI if told to - caller determines
769 * when this should be done.
770 */
771 if (add_l3_ecomm) {
772 vrf_export_rtl = bgpevpn_get_vrf_export_rtl(vpn);
773 if (vrf_export_rtl && !list_isempty(vrf_export_rtl)) {
774 for (ALL_LIST_ELEMENTS(vrf_export_rtl, node, nnode,
775 ecom))
776 attr->ecommunity = ecommunity_merge(
777 attr->ecommunity, ecom);
778 }
779 }
780
781 /* Add MAC mobility (sticky) if needed. */
782 if (attr->sticky) {
783 seqnum = 0;
784 memset(&ecom_sticky, 0, sizeof(ecom_sticky));
785 encode_mac_mobility_extcomm(1, seqnum, &eval_sticky);
786 ecom_sticky.size = 1;
787 ecom_sticky.val = (uint8_t *)eval_sticky.val;
788 attr->ecommunity =
789 ecommunity_merge(attr->ecommunity, &ecom_sticky);
790 }
791
792 /* Add RMAC, if told to. */
793 if (add_l3_ecomm) {
794 memset(&ecom_rmac, 0, sizeof(ecom_rmac));
795 encode_rmac_extcomm(&eval_rmac, &attr->rmac);
796 ecom_rmac.size = 1;
797 ecom_rmac.val = (uint8_t *)eval_rmac.val;
798 attr->ecommunity =
799 ecommunity_merge(attr->ecommunity, &ecom_rmac);
800 }
801
802 /* Add default gateway, if needed. */
803 if (attr->default_gw) {
804 memset(&ecom_default_gw, 0, sizeof(ecom_default_gw));
805 encode_default_gw_extcomm(&eval_default_gw);
806 ecom_default_gw.size = 1;
807 ecom_default_gw.val = (uint8_t *)eval_default_gw.val;
808 attr->ecommunity =
809 ecommunity_merge(attr->ecommunity, &ecom_default_gw);
810 }
811
812 if (attr->router_flag) {
813 memset(&ecom_na, 0, sizeof(ecom_na));
814 encode_na_flag_extcomm(&eval_na, attr->router_flag);
815 ecom_na.size = 1;
816 ecom_na.val = (uint8_t *)eval_na.val;
817 attr->ecommunity = ecommunity_merge(attr->ecommunity,
818 &ecom_na);
819 }
820
821 attr->flag |= ATTR_FLAG_BIT(BGP_ATTR_EXT_COMMUNITIES);
822 }
823
824 /*
825 * Add MAC mobility extended community to attribute.
826 */
827 static void add_mac_mobility_to_attr(uint32_t seq_num, struct attr *attr)
828 {
829 struct ecommunity ecom_tmp;
830 struct ecommunity_val eval;
831 uint8_t *ecom_val_ptr;
832 int i;
833 uint8_t *pnt;
834 int type = 0;
835 int sub_type = 0;
836
837 /* Build MM */
838 encode_mac_mobility_extcomm(0, seq_num, &eval);
839
840 /* Find current MM ecommunity */
841 ecom_val_ptr = NULL;
842
843 if (attr->ecommunity) {
844 for (i = 0; i < attr->ecommunity->size; i++) {
845 pnt = attr->ecommunity->val + (i * 8);
846 type = *pnt++;
847 sub_type = *pnt++;
848
849 if (type == ECOMMUNITY_ENCODE_EVPN
850 && sub_type
851 == ECOMMUNITY_EVPN_SUBTYPE_MACMOBILITY) {
852 ecom_val_ptr = (uint8_t *)(attr->ecommunity->val
853 + (i * 8));
854 break;
855 }
856 }
857 }
858
859 /* Update the existing MM ecommunity */
860 if (ecom_val_ptr) {
861 memcpy(ecom_val_ptr, eval.val, sizeof(char) * ECOMMUNITY_SIZE);
862 }
863 /* Add MM to existing */
864 else {
865 memset(&ecom_tmp, 0, sizeof(ecom_tmp));
866 ecom_tmp.size = 1;
867 ecom_tmp.val = (uint8_t *)eval.val;
868
869 if (attr->ecommunity)
870 attr->ecommunity =
871 ecommunity_merge(attr->ecommunity, &ecom_tmp);
872 else
873 attr->ecommunity = ecommunity_dup(&ecom_tmp);
874 }
875 }
876
877 /* Install EVPN route into zebra. */
878 static int evpn_zebra_install(struct bgp *bgp, struct bgpevpn *vpn,
879 struct prefix_evpn *p,
880 struct in_addr remote_vtep_ip, uint8_t flags,
881 uint32_t seq)
882 {
883 int ret;
884
885 if (p->prefix.route_type == BGP_EVPN_MAC_IP_ROUTE)
886 ret = bgp_zebra_send_remote_macip(bgp, vpn, p, remote_vtep_ip,
887 1, flags, seq);
888 else
889 ret = bgp_zebra_send_remote_vtep(bgp, vpn, p, 1);
890
891 return ret;
892 }
893
894 /* Uninstall EVPN route from zebra. */
895 static int evpn_zebra_uninstall(struct bgp *bgp, struct bgpevpn *vpn,
896 struct prefix_evpn *p,
897 struct in_addr remote_vtep_ip)
898 {
899 int ret;
900
901 if (p->prefix.route_type == BGP_EVPN_MAC_IP_ROUTE)
902 ret = bgp_zebra_send_remote_macip(bgp, vpn, p, remote_vtep_ip,
903 0, 0, 0);
904 else
905 ret = bgp_zebra_send_remote_vtep(bgp, vpn, p, 0);
906
907 return ret;
908 }
909
910 /*
911 * Due to MAC mobility, the prior "local" best route has been supplanted
912 * by a "remote" best route. The prior route has to be deleted and withdrawn
913 * from peers.
914 */
915 static void evpn_delete_old_local_route(struct bgp *bgp, struct bgpevpn *vpn,
916 struct bgp_node *rn,
917 struct bgp_info *old_local)
918 {
919 struct bgp_node *global_rn;
920 struct bgp_info *ri;
921 afi_t afi = AFI_L2VPN;
922 safi_t safi = SAFI_EVPN;
923
924 /* Locate route node in the global EVPN routing table. Note that
925 * this table is a 2-level tree (RD-level + Prefix-level) similar to
926 * L3VPN routes.
927 */
928 global_rn = bgp_afi_node_lookup(bgp->rib[afi][safi], afi, safi,
929 (struct prefix *)&rn->p, &vpn->prd);
930 if (global_rn) {
931 /* Delete route entry in the global EVPN table. */
932 delete_evpn_route_entry(bgp, afi, safi, global_rn, &ri);
933
934 /* Schedule for processing - withdraws to peers happen from
935 * this table.
936 */
937 if (ri)
938 bgp_process(bgp, global_rn, afi, safi);
939 bgp_unlock_node(global_rn);
940 }
941
942 /* Delete route entry in the VNI route table, caller to remove. */
943 bgp_info_delete(rn, old_local);
944 }
945
946 static struct in_addr *es_vtep_new(struct in_addr vtep)
947 {
948 struct in_addr *ip;
949
950 ip = XCALLOC(MTYPE_BGP_EVPN_ES_VTEP, sizeof(struct in_addr));
951 if (!ip)
952 return NULL;
953
954 ip->s_addr = vtep.s_addr;
955 return ip;
956 }
957
958 static void es_vtep_free(struct in_addr *ip)
959 {
960 XFREE(MTYPE_BGP_EVPN_ES_VTEP, ip);
961 }
962
963 /* check if VTEP is already part of the list */
964 static int is_vtep_present_in_list(struct list *list,
965 struct in_addr vtep)
966 {
967 struct listnode *node = NULL;
968 struct in_addr *tmp;
969
970 for (ALL_LIST_ELEMENTS_RO(list, node, tmp)) {
971 if (tmp->s_addr == vtep.s_addr)
972 return 1;
973 }
974 return 0;
975 }
976
977 /*
978 * Best path for ES route was changed,
979 * update the list of VTEPs for this ES
980 */
981 static int evpn_es_install_vtep(struct bgp *bgp,
982 struct evpnes *es,
983 struct prefix_evpn *p,
984 struct in_addr rvtep)
985 {
986 struct in_addr *vtep_ip;
987
988 if (is_vtep_present_in_list(es->vtep_list, rvtep))
989 return 0;
990
991
992 vtep_ip = es_vtep_new(rvtep);
993 if (vtep_ip)
994 listnode_add_sort(es->vtep_list, vtep_ip);
995 return 0;
996 }
997
998 /*
999 * Best path for ES route was changed,
1000 * update the list of VTEPs for this ES
1001 */
1002 static int evpn_es_uninstall_vtep(struct bgp *bgp,
1003 struct evpnes *es,
1004 struct prefix_evpn *p,
1005 struct in_addr rvtep)
1006 {
1007 struct listnode *node, *nnode, *node_to_del = NULL;
1008 struct in_addr *tmp;
1009
1010 for (ALL_LIST_ELEMENTS(es->vtep_list, node, nnode, tmp)) {
1011 if (tmp->s_addr == rvtep.s_addr) {
1012 es_vtep_free(tmp);
1013 node_to_del = node;
1014 }
1015 }
1016
1017 if (node_to_del)
1018 list_delete_node(es->vtep_list, node_to_del);
1019
1020 return 0;
1021 }
1022
1023 /*
1024 * Calculate the best path for a ES(type-4) route.
1025 */
1026 static int evpn_es_route_select_install(struct bgp *bgp,
1027 struct evpnes *es,
1028 struct bgp_node *rn)
1029 {
1030 int ret = 0;
1031 afi_t afi = AFI_L2VPN;
1032 safi_t safi = SAFI_EVPN;
1033 struct bgp_info *old_select; /* old best */
1034 struct bgp_info *new_select; /* new best */
1035 struct bgp_info_pair old_and_new;
1036
1037 /* Compute the best path. */
1038 bgp_best_selection(bgp, rn, &bgp->maxpaths[afi][safi],
1039 &old_and_new, afi, safi);
1040 old_select = old_and_new.old;
1041 new_select = old_and_new.new;
1042
1043 /*
1044 * If the best path hasn't changed - see if something needs to be
1045 * updated
1046 */
1047 if (old_select && old_select == new_select
1048 && old_select->type == ZEBRA_ROUTE_BGP
1049 && old_select->sub_type == BGP_ROUTE_IMPORTED
1050 && !CHECK_FLAG(rn->flags, BGP_NODE_USER_CLEAR)
1051 && !CHECK_FLAG(old_select->flags, BGP_INFO_ATTR_CHANGED)
1052 && !bgp->addpath_tx_used[afi][safi]) {
1053 if (bgp_zebra_has_route_changed(rn, old_select)) {
1054 ret = evpn_es_install_vtep(bgp, es,
1055 (struct prefix_evpn *)&rn->p,
1056 old_select->attr->nexthop);
1057 }
1058 UNSET_FLAG(old_select->flags, BGP_INFO_MULTIPATH_CHG);
1059 bgp_zebra_clear_route_change_flags(rn);
1060 return ret;
1061 }
1062
1063 /* If the user did a "clear" this flag will be set */
1064 UNSET_FLAG(rn->flags, BGP_NODE_USER_CLEAR);
1065
1066 /*
1067 * bestpath has changed; update relevant fields and install or uninstall
1068 * into the zebra RIB.
1069 */
1070 if (old_select || new_select)
1071 bgp_bump_version(rn);
1072
1073 if (old_select)
1074 bgp_info_unset_flag(rn, old_select, BGP_INFO_SELECTED);
1075 if (new_select) {
1076 bgp_info_set_flag(rn, new_select, BGP_INFO_SELECTED);
1077 bgp_info_unset_flag(rn, new_select, BGP_INFO_ATTR_CHANGED);
1078 UNSET_FLAG(new_select->flags, BGP_INFO_MULTIPATH_CHG);
1079 }
1080
1081 if (new_select && new_select->type == ZEBRA_ROUTE_BGP
1082 && new_select->sub_type == BGP_ROUTE_IMPORTED) {
1083 ret = evpn_es_install_vtep(bgp, es,
1084 (struct prefix_evpn *)&rn->p,
1085 new_select->attr->nexthop);
1086 } else {
1087 if (old_select && old_select->type == ZEBRA_ROUTE_BGP
1088 && old_select->sub_type == BGP_ROUTE_IMPORTED)
1089 ret = evpn_es_uninstall_vtep(
1090 bgp, es, (struct prefix_evpn *)&rn->p,
1091 old_select->attr->nexthop);
1092 }
1093
1094 /* Clear any route change flags. */
1095 bgp_zebra_clear_route_change_flags(rn);
1096
1097 /* Reap old select bgp_info, if it has been removed */
1098 if (old_select && CHECK_FLAG(old_select->flags, BGP_INFO_REMOVED))
1099 bgp_info_reap(rn, old_select);
1100
1101 return ret;
1102 }
1103
1104 /*
1105 * Calculate the best path for an EVPN route. Install/update best path in zebra,
1106 * if appropriate.
1107 */
1108 static int evpn_route_select_install(struct bgp *bgp, struct bgpevpn *vpn,
1109 struct bgp_node *rn)
1110 {
1111 struct bgp_info *old_select, *new_select;
1112 struct bgp_info_pair old_and_new;
1113 struct prefix_evpn *evp;
1114 afi_t afi = AFI_L2VPN;
1115 safi_t safi = SAFI_EVPN;
1116 int ret = 0;
1117 uint8_t flags = 0;
1118
1119 /* Compute the best path. */
1120 bgp_best_selection(bgp, rn, &bgp->maxpaths[afi][safi], &old_and_new,
1121 afi, safi);
1122 old_select = old_and_new.old;
1123 new_select = old_and_new.new;
1124
1125 evp = (struct prefix_evpn *)&rn->p;
1126 /* If the best path hasn't changed - see if there is still something to
1127 * update
1128 * to zebra RIB.
1129 */
1130 if (old_select && old_select == new_select
1131 && old_select->type == ZEBRA_ROUTE_BGP
1132 && old_select->sub_type == BGP_ROUTE_IMPORTED
1133 && !CHECK_FLAG(rn->flags, BGP_NODE_USER_CLEAR)
1134 && !CHECK_FLAG(old_select->flags, BGP_INFO_ATTR_CHANGED)
1135 && !bgp->addpath_tx_used[afi][safi]) {
1136 if (bgp_zebra_has_route_changed(rn, old_select)) {
1137 if (old_select->attr->sticky)
1138 SET_FLAG(flags, ZEBRA_MACIP_TYPE_STICKY);
1139 if (old_select->attr->default_gw)
1140 SET_FLAG(flags, ZEBRA_MACIP_TYPE_GW);
1141 if (is_evpn_prefix_ipaddr_v6(evp) &&
1142 old_select->attr->router_flag)
1143 SET_FLAG(flags, ZEBRA_MACIP_TYPE_ROUTER_FLAG);
1144
1145 ret = evpn_zebra_install(
1146 bgp, vpn, (struct prefix_evpn *)&rn->p,
1147 old_select->attr->nexthop, flags,
1148 mac_mobility_seqnum(old_select->attr));
1149 }
1150 UNSET_FLAG(old_select->flags, BGP_INFO_MULTIPATH_CHG);
1151 bgp_zebra_clear_route_change_flags(rn);
1152 return ret;
1153 }
1154
1155 /* If the user did a "clear" this flag will be set */
1156 UNSET_FLAG(rn->flags, BGP_NODE_USER_CLEAR);
1157
1158 /* bestpath has changed; update relevant fields and install or uninstall
1159 * into the zebra RIB.
1160 */
1161 if (old_select || new_select)
1162 bgp_bump_version(rn);
1163
1164 if (old_select)
1165 bgp_info_unset_flag(rn, old_select, BGP_INFO_SELECTED);
1166 if (new_select) {
1167 bgp_info_set_flag(rn, new_select, BGP_INFO_SELECTED);
1168 bgp_info_unset_flag(rn, new_select, BGP_INFO_ATTR_CHANGED);
1169 UNSET_FLAG(new_select->flags, BGP_INFO_MULTIPATH_CHG);
1170 }
1171
1172 if (new_select && new_select->type == ZEBRA_ROUTE_BGP
1173 && new_select->sub_type == BGP_ROUTE_IMPORTED) {
1174 flags = 0;
1175 if (new_select->attr->sticky)
1176 SET_FLAG(flags, ZEBRA_MACIP_TYPE_STICKY);
1177 if (new_select->attr->default_gw)
1178 SET_FLAG(flags, ZEBRA_MACIP_TYPE_GW);
1179 if (is_evpn_prefix_ipaddr_v6(evp) &&
1180 new_select->attr->router_flag)
1181 SET_FLAG(flags, ZEBRA_MACIP_TYPE_ROUTER_FLAG);
1182
1183 ret = evpn_zebra_install(bgp, vpn, (struct prefix_evpn *)&rn->p,
1184 new_select->attr->nexthop, flags,
1185 mac_mobility_seqnum(new_select->attr));
1186 /* If an old best existed and it was a "local" route, the only
1187 * reason
1188 * it would be supplanted is due to MAC mobility procedures. So,
1189 * we
1190 * need to do an implicit delete and withdraw that route from
1191 * peers.
1192 */
1193 if (old_select && old_select->peer == bgp->peer_self
1194 && old_select->type == ZEBRA_ROUTE_BGP
1195 && old_select->sub_type == BGP_ROUTE_STATIC)
1196 evpn_delete_old_local_route(bgp, vpn, rn, old_select);
1197 } else {
1198 if (old_select && old_select->type == ZEBRA_ROUTE_BGP
1199 && old_select->sub_type == BGP_ROUTE_IMPORTED)
1200 ret = evpn_zebra_uninstall(bgp, vpn,
1201 (struct prefix_evpn *)&rn->p,
1202 old_select->attr->nexthop);
1203 }
1204
1205 /* Clear any route change flags. */
1206 bgp_zebra_clear_route_change_flags(rn);
1207
1208 /* Reap old select bgp_info, if it has been removed */
1209 if (old_select && CHECK_FLAG(old_select->flags, BGP_INFO_REMOVED))
1210 bgp_info_reap(rn, old_select);
1211
1212 return ret;
1213 }
1214
1215 /*
1216 * Return true if the local ri for this rn is of type gateway mac
1217 */
1218 static int evpn_route_is_def_gw(struct bgp *bgp, struct bgp_node *rn)
1219 {
1220 struct bgp_info *tmp_ri = NULL;
1221 struct bgp_info *local_ri = NULL;
1222
1223 local_ri = NULL;
1224 for (tmp_ri = rn->info; tmp_ri; tmp_ri = tmp_ri->next) {
1225 if (tmp_ri->peer == bgp->peer_self
1226 && tmp_ri->type == ZEBRA_ROUTE_BGP
1227 && tmp_ri->sub_type == BGP_ROUTE_STATIC)
1228 local_ri = tmp_ri;
1229 }
1230
1231 if (!local_ri)
1232 return 0;
1233
1234 return local_ri->attr->default_gw;
1235 }
1236
1237
1238 /*
1239 * Return true if the local ri for this rn has sticky set
1240 */
1241 static int evpn_route_is_sticky(struct bgp *bgp, struct bgp_node *rn)
1242 {
1243 struct bgp_info *tmp_ri;
1244 struct bgp_info *local_ri;
1245
1246 local_ri = NULL;
1247 for (tmp_ri = rn->info; tmp_ri; tmp_ri = tmp_ri->next) {
1248 if (tmp_ri->peer == bgp->peer_self
1249 && tmp_ri->type == ZEBRA_ROUTE_BGP
1250 && tmp_ri->sub_type == BGP_ROUTE_STATIC)
1251 local_ri = tmp_ri;
1252 }
1253
1254 if (!local_ri)
1255 return 0;
1256
1257 return local_ri->attr->sticky;
1258 }
1259
1260 /*
1261 * create or update EVPN type4 route entry.
1262 * This could be in the ES table or the global table.
1263 * TODO: handle remote ES (type4) routes as well
1264 */
1265 static int update_evpn_type4_route_entry(struct bgp *bgp,
1266 struct evpnes *es,
1267 afi_t afi, safi_t safi,
1268 struct bgp_node *rn,
1269 struct attr *attr,
1270 int add,
1271 struct bgp_info **ri,
1272 int *route_changed)
1273 {
1274 char buf[ESI_STR_LEN];
1275 char buf1[INET6_ADDRSTRLEN];
1276 struct bgp_info *tmp_ri = NULL;
1277 struct bgp_info *local_ri = NULL; /* local route entry if any */
1278 struct bgp_info *remote_ri = NULL; /* remote route entry if any */
1279 struct attr *attr_new = NULL;
1280 struct prefix_evpn *evp = NULL;
1281
1282 *ri = NULL;
1283 *route_changed = 1;
1284 evp = (struct prefix_evpn *)&rn->p;
1285
1286 /* locate the local and remote entries if any */
1287 for (tmp_ri = rn->info; tmp_ri; tmp_ri = tmp_ri->next) {
1288 if (tmp_ri->peer == bgp->peer_self &&
1289 tmp_ri->type == ZEBRA_ROUTE_BGP &&
1290 tmp_ri->sub_type == BGP_ROUTE_STATIC)
1291 local_ri = tmp_ri;
1292 if (tmp_ri->type == ZEBRA_ROUTE_BGP &&
1293 tmp_ri->sub_type == BGP_ROUTE_IMPORTED &&
1294 CHECK_FLAG(tmp_ri->flags, BGP_INFO_VALID))
1295 remote_ri = tmp_ri;
1296 }
1297
1298 /* we don't expect to see a remote_ri at this point.
1299 * An ES route has esi + vtep_ip as the key,
1300 * We shouldn't see the same route from any other vtep.
1301 */
1302 if (remote_ri) {
1303 flog_err(
1304 EC_BGP_ES_INVALID,
1305 "%u ERROR: local es route for ESI: %s Vtep %s also learnt from remote",
1306 bgp->vrf_id,
1307 esi_to_str(&evp->prefix.es_addr.esi, buf, sizeof(buf)),
1308 ipaddr2str(&es->originator_ip, buf1, sizeof(buf1)));
1309 return -1;
1310 }
1311
1312 if (!local_ri && !add)
1313 return 0;
1314
1315 /* create or update the entry */
1316 if (!local_ri) {
1317
1318 /* Add or update attribute to hash */
1319 attr_new = bgp_attr_intern(attr);
1320
1321 /* Create new route with its attribute. */
1322 tmp_ri = info_make(ZEBRA_ROUTE_BGP, BGP_ROUTE_STATIC,
1323 0, bgp->peer_self, attr_new, rn);
1324 SET_FLAG(tmp_ri->flags, BGP_INFO_VALID);
1325
1326 /* add the newly created path to the route-node */
1327 bgp_info_add(rn, tmp_ri);
1328 } else {
1329 tmp_ri = local_ri;
1330 if (attrhash_cmp(tmp_ri->attr, attr)
1331 && !CHECK_FLAG(tmp_ri->flags, BGP_INFO_REMOVED))
1332 *route_changed = 0;
1333 else {
1334 /* The attribute has changed.
1335 * Add (or update) attribute to hash. */
1336 attr_new = bgp_attr_intern(attr);
1337 bgp_info_set_flag(rn, tmp_ri, BGP_INFO_ATTR_CHANGED);
1338
1339 /* Restore route, if needed. */
1340 if (CHECK_FLAG(tmp_ri->flags, BGP_INFO_REMOVED))
1341 bgp_info_restore(rn, tmp_ri);
1342
1343 /* Unintern existing, set to new. */
1344 bgp_attr_unintern(&tmp_ri->attr);
1345 tmp_ri->attr = attr_new;
1346 tmp_ri->uptime = bgp_clock();
1347 }
1348 }
1349
1350 /* Return back the route entry. */
1351 *ri = tmp_ri;
1352 return 0;
1353 }
1354
1355 /* update evpn es (type-4) route */
1356 static int update_evpn_type4_route(struct bgp *bgp,
1357 struct evpnes *es,
1358 struct prefix_evpn *p)
1359 {
1360 int ret = 0;
1361 int route_changed = 0;
1362 char buf[ESI_STR_LEN];
1363 char buf1[INET6_ADDRSTRLEN];
1364 afi_t afi = AFI_L2VPN;
1365 safi_t safi = SAFI_EVPN;
1366 struct attr attr;
1367 struct attr *attr_new = NULL;
1368 struct bgp_node *rn = NULL;
1369 struct bgp_info *ri = NULL;
1370
1371 memset(&attr, 0, sizeof(struct attr));
1372
1373 /* Build path-attribute for this route. */
1374 bgp_attr_default_set(&attr, BGP_ORIGIN_IGP);
1375 attr.nexthop = es->originator_ip.ipaddr_v4;
1376 attr.mp_nexthop_global_in = es->originator_ip.ipaddr_v4;
1377 attr.mp_nexthop_len = BGP_ATTR_NHLEN_IPV4;
1378
1379 /* Set up extended community. */
1380 build_evpn_type4_route_extcomm(es, &attr);
1381
1382 /* First, create (or fetch) route node within the ESI. */
1383 /* NOTE: There is no RD here. */
1384 rn = bgp_node_get(es->route_table, (struct prefix *)p);
1385
1386 /* Create or update route entry. */
1387 ret = update_evpn_type4_route_entry(bgp, es, afi, safi, rn,
1388 &attr, 1, &ri,
1389 &route_changed);
1390 if (ret != 0) {
1391 flog_err(EC_BGP_ES_INVALID,
1392 "%u ERROR: Failed to updated ES route ESI: %s VTEP %s",
1393 bgp->vrf_id,
1394 esi_to_str(&p->prefix.es_addr.esi, buf, sizeof(buf)),
1395 ipaddr2str(&es->originator_ip, buf1, sizeof(buf1)));
1396 }
1397
1398 assert(ri);
1399 attr_new = ri->attr;
1400
1401 /* Perform route selection;
1402 * this is just to set the flags correctly
1403 * as local route in the ES always wins.
1404 */
1405 evpn_es_route_select_install(bgp, es, rn);
1406 bgp_unlock_node(rn);
1407
1408 /* If this is a new route or some attribute has changed, export the
1409 * route to the global table. The route will be advertised to peers
1410 * from there. Note that this table is a 2-level tree (RD-level +
1411 * Prefix-level) similar to L3VPN routes.
1412 */
1413 if (route_changed) {
1414 struct bgp_info *global_ri;
1415
1416 rn = bgp_afi_node_get(bgp->rib[afi][safi], afi, safi,
1417 (struct prefix *)p, &es->prd);
1418 update_evpn_type4_route_entry(bgp, es, afi, safi,
1419 rn, attr_new,
1420 1, &global_ri,
1421 &route_changed);
1422
1423 /* Schedule for processing and unlock node. */
1424 bgp_process(bgp, rn, afi, safi);
1425 bgp_unlock_node(rn);
1426 }
1427
1428 /* Unintern temporary. */
1429 aspath_unintern(&attr.aspath);
1430 return 0;
1431 }
1432
1433 static int update_evpn_type5_route_entry(struct bgp *bgp_def,
1434 struct bgp *bgp_vrf, afi_t afi,
1435 safi_t safi, struct bgp_node *rn,
1436 struct attr *attr, int *route_changed)
1437 {
1438 struct attr *attr_new = NULL;
1439 struct bgp_info *ri = NULL;
1440 mpls_label_t label = MPLS_INVALID_LABEL;
1441 struct bgp_info *local_ri = NULL;
1442 struct bgp_info *tmp_ri = NULL;
1443
1444 *route_changed = 0;
1445 /* locate the local route entry if any */
1446 for (tmp_ri = rn->info; tmp_ri; tmp_ri = tmp_ri->next) {
1447 if (tmp_ri->peer == bgp_def->peer_self
1448 && tmp_ri->type == ZEBRA_ROUTE_BGP
1449 && tmp_ri->sub_type == BGP_ROUTE_STATIC)
1450 local_ri = tmp_ri;
1451 }
1452
1453 /*
1454 * create a new route entry if one doesnt exist.
1455 * Otherwise see if route attr has changed
1456 */
1457 if (!local_ri) {
1458
1459 /* route has changed as this is the first entry */
1460 *route_changed = 1;
1461
1462 /* Add (or update) attribute to hash. */
1463 attr_new = bgp_attr_intern(attr);
1464
1465 /* create the route info from attribute */
1466 ri = info_make(ZEBRA_ROUTE_BGP, BGP_ROUTE_STATIC, 0,
1467 bgp_def->peer_self, attr_new, rn);
1468 SET_FLAG(ri->flags, BGP_INFO_VALID);
1469
1470 /* Type-5 routes advertise the L3-VNI */
1471 bgp_info_extra_get(ri);
1472 vni2label(bgp_vrf->l3vni, &label);
1473 memcpy(&ri->extra->label, &label, sizeof(label));
1474 ri->extra->num_labels = 1;
1475
1476 /* add the route entry to route node*/
1477 bgp_info_add(rn, ri);
1478 } else {
1479
1480 tmp_ri = local_ri;
1481 if (!attrhash_cmp(tmp_ri->attr, attr)) {
1482
1483 /* attribute changed */
1484 *route_changed = 1;
1485
1486 /* The attribute has changed. */
1487 /* Add (or update) attribute to hash. */
1488 attr_new = bgp_attr_intern(attr);
1489 bgp_info_set_flag(rn, tmp_ri, BGP_INFO_ATTR_CHANGED);
1490
1491 /* Restore route, if needed. */
1492 if (CHECK_FLAG(tmp_ri->flags, BGP_INFO_REMOVED))
1493 bgp_info_restore(rn, tmp_ri);
1494
1495 /* Unintern existing, set to new. */
1496 bgp_attr_unintern(&tmp_ri->attr);
1497 tmp_ri->attr = attr_new;
1498 tmp_ri->uptime = bgp_clock();
1499 }
1500 }
1501 return 0;
1502 }
1503
1504 /* update evpn type-5 route entry */
1505 static int update_evpn_type5_route(struct bgp *bgp_vrf, struct prefix_evpn *evp,
1506 struct attr *src_attr)
1507 {
1508 afi_t afi = AFI_L2VPN;
1509 safi_t safi = SAFI_EVPN;
1510 struct attr attr;
1511 struct bgp_node *rn = NULL;
1512 struct bgp *bgp_def = NULL;
1513 int route_changed = 0;
1514
1515 bgp_def = bgp_get_default();
1516 if (!bgp_def)
1517 return 0;
1518
1519 /* Build path attribute for this route - use the source attr, if
1520 * present, else treat as locally originated.
1521 */
1522 if (src_attr)
1523 bgp_attr_dup(&attr, src_attr);
1524 else {
1525 memset(&attr, 0, sizeof(struct attr));
1526 bgp_attr_default_set(&attr, BGP_ORIGIN_IGP);
1527 }
1528 /* Set nexthop to ourselves and fill in the Router MAC. */
1529 attr.nexthop = bgp_vrf->originator_ip;
1530 attr.mp_nexthop_global_in = bgp_vrf->originator_ip;
1531 attr.mp_nexthop_len = BGP_ATTR_NHLEN_IPV4;
1532 memcpy(&attr.rmac, &bgp_vrf->rmac, sizeof(struct ethaddr));
1533
1534 /* Setup RT and encap extended community */
1535 build_evpn_type5_route_extcomm(bgp_vrf, &attr);
1536
1537 /* get the route node in global table */
1538 rn = bgp_afi_node_get(bgp_def->rib[afi][safi], afi, safi,
1539 (struct prefix *)evp, &bgp_vrf->vrf_prd);
1540 assert(rn);
1541
1542 /* create or update the route entry within the route node */
1543 update_evpn_type5_route_entry(bgp_def, bgp_vrf, afi, safi, rn, &attr,
1544 &route_changed);
1545
1546 /* schedule for processing and unlock node */
1547 if (route_changed) {
1548 bgp_process(bgp_def, rn, afi, safi);
1549 bgp_unlock_node(rn);
1550 }
1551
1552 /* uninten temporary */
1553 if (!src_attr)
1554 aspath_unintern(&attr.aspath);
1555 return 0;
1556 }
1557
1558 /*
1559 * Create or update EVPN route entry. This could be in the VNI route table
1560 * or the global route table.
1561 */
1562 static int update_evpn_route_entry(struct bgp *bgp, struct bgpevpn *vpn,
1563 afi_t afi, safi_t safi, struct bgp_node *rn,
1564 struct attr *attr, int add,
1565 struct bgp_info **ri, uint8_t flags,
1566 uint32_t seq)
1567 {
1568 struct bgp_info *tmp_ri;
1569 struct bgp_info *local_ri;
1570 struct attr *attr_new;
1571 mpls_label_t label[BGP_MAX_LABELS];
1572 uint32_t num_labels = 1;
1573 int route_change = 1;
1574 uint8_t sticky = 0;
1575 struct prefix_evpn *evp;
1576
1577 *ri = NULL;
1578 evp = (struct prefix_evpn *)&rn->p;
1579 memset(&label, 0, sizeof(label));
1580
1581 /* See if this is an update of an existing route, or a new add. */
1582 local_ri = NULL;
1583 for (tmp_ri = rn->info; tmp_ri; tmp_ri = tmp_ri->next) {
1584 if (tmp_ri->peer == bgp->peer_self
1585 && tmp_ri->type == ZEBRA_ROUTE_BGP
1586 && tmp_ri->sub_type == BGP_ROUTE_STATIC)
1587 local_ri = tmp_ri;
1588 }
1589
1590 /* If route doesn't exist already, create a new one, if told to.
1591 * Otherwise act based on whether the attributes of the route have
1592 * changed or not.
1593 */
1594 if (!local_ri && !add)
1595 return 0;
1596
1597 /* For non-GW MACs, update MAC mobility seq number, if needed. */
1598 if (seq && !CHECK_FLAG(flags, ZEBRA_MACIP_TYPE_GW))
1599 add_mac_mobility_to_attr(seq, attr);
1600
1601 if (!local_ri) {
1602 /* Add (or update) attribute to hash. */
1603 attr_new = bgp_attr_intern(attr);
1604
1605 /* Extract MAC mobility sequence number, if any. */
1606 attr_new->mm_seqnum =
1607 bgp_attr_mac_mobility_seqnum(attr_new, &sticky);
1608 attr_new->sticky = sticky;
1609
1610 /* Create new route with its attribute. */
1611 tmp_ri = info_make(ZEBRA_ROUTE_BGP, BGP_ROUTE_STATIC, 0,
1612 bgp->peer_self, attr_new, rn);
1613 SET_FLAG(tmp_ri->flags, BGP_INFO_VALID);
1614 bgp_info_extra_get(tmp_ri);
1615
1616 /* The VNI goes into the 'label' field of the route */
1617 vni2label(vpn->vni, &label[0]);
1618
1619 /* Type-2 routes may carry a second VNI - the L3-VNI.
1620 * Only attach second label if we are advertising two labels for
1621 * type-2 routes.
1622 */
1623 if (evp->prefix.route_type == BGP_EVPN_MAC_IP_ROUTE
1624 && CHECK_FLAG(vpn->flags, VNI_FLAG_USE_TWO_LABELS)) {
1625 vni_t l3vni;
1626
1627 l3vni = bgpevpn_get_l3vni(vpn);
1628 if (l3vni) {
1629 vni2label(l3vni, &label[1]);
1630 num_labels++;
1631 }
1632 }
1633
1634 memcpy(&tmp_ri->extra->label, label, sizeof(label));
1635 tmp_ri->extra->num_labels = num_labels;
1636 bgp_info_add(rn, tmp_ri);
1637 } else {
1638 tmp_ri = local_ri;
1639 if (attrhash_cmp(tmp_ri->attr, attr)
1640 && !CHECK_FLAG(tmp_ri->flags, BGP_INFO_REMOVED))
1641 route_change = 0;
1642 else {
1643 /*
1644 * The attributes have changed, type-2 routes needs to
1645 * be advertised with right labels.
1646 */
1647 vni2label(vpn->vni, &label[0]);
1648 if (evp->prefix.route_type == BGP_EVPN_MAC_IP_ROUTE
1649 && CHECK_FLAG(vpn->flags,
1650 VNI_FLAG_USE_TWO_LABELS)) {
1651 vni_t l3vni;
1652
1653 l3vni = bgpevpn_get_l3vni(vpn);
1654 if (l3vni) {
1655 vni2label(l3vni, &label[1]);
1656 num_labels++;
1657 }
1658 }
1659 memcpy(&tmp_ri->extra->label, label, sizeof(label));
1660 tmp_ri->extra->num_labels = num_labels;
1661
1662 /* The attribute has changed. */
1663 /* Add (or update) attribute to hash. */
1664 attr_new = bgp_attr_intern(attr);
1665 bgp_info_set_flag(rn, tmp_ri, BGP_INFO_ATTR_CHANGED);
1666
1667 /* Extract MAC mobility sequence number, if any. */
1668 attr_new->mm_seqnum =
1669 bgp_attr_mac_mobility_seqnum(attr_new, &sticky);
1670 attr_new->sticky = sticky;
1671
1672 /* Restore route, if needed. */
1673 if (CHECK_FLAG(tmp_ri->flags, BGP_INFO_REMOVED))
1674 bgp_info_restore(rn, tmp_ri);
1675
1676 /* Unintern existing, set to new. */
1677 bgp_attr_unintern(&tmp_ri->attr);
1678 tmp_ri->attr = attr_new;
1679 tmp_ri->uptime = bgp_clock();
1680 }
1681 }
1682
1683 /* Return back the route entry. */
1684 *ri = tmp_ri;
1685 return route_change;
1686 }
1687
1688 /*
1689 * Create or update EVPN route (of type based on prefix) for specified VNI
1690 * and schedule for processing.
1691 */
1692 static int update_evpn_route(struct bgp *bgp, struct bgpevpn *vpn,
1693 struct prefix_evpn *p, uint8_t flags,
1694 uint32_t seq)
1695 {
1696 struct bgp_node *rn;
1697 struct attr attr;
1698 struct attr *attr_new;
1699 int add_l3_ecomm = 0;
1700 struct bgp_info *ri;
1701 afi_t afi = AFI_L2VPN;
1702 safi_t safi = SAFI_EVPN;
1703 int route_change;
1704
1705 memset(&attr, 0, sizeof(struct attr));
1706
1707 /* Build path-attribute for this route. */
1708 bgp_attr_default_set(&attr, BGP_ORIGIN_IGP);
1709 attr.nexthop = vpn->originator_ip;
1710 attr.mp_nexthop_global_in = vpn->originator_ip;
1711 attr.mp_nexthop_len = BGP_ATTR_NHLEN_IPV4;
1712 attr.sticky = CHECK_FLAG(flags, ZEBRA_MACIP_TYPE_STICKY) ? 1 : 0;
1713 attr.default_gw = CHECK_FLAG(flags, ZEBRA_MACIP_TYPE_GW) ? 1 : 0;
1714 attr.router_flag = CHECK_FLAG(flags,
1715 ZEBRA_MACIP_TYPE_ROUTER_FLAG) ? 1 : 0;
1716
1717 /* PMSI is only needed for type-3 routes */
1718 if (p->prefix.route_type == BGP_EVPN_IMET_ROUTE)
1719 attr.flag |= ATTR_FLAG_BIT(BGP_ATTR_PMSI_TUNNEL);
1720
1721 /* router mac is only needed for type-2 routes here. */
1722 if (p->prefix.route_type == BGP_EVPN_MAC_IP_ROUTE)
1723 bgpevpn_get_rmac(vpn, &attr.rmac);
1724 vni2label(vpn->vni, &(attr.label));
1725
1726 /* Include L3 VNI related RTs and RMAC for type-2 routes, if they're
1727 * IPv4 or IPv6 global addresses and we're advertising L3VNI with
1728 * these routes.
1729 */
1730 if (p->prefix.route_type == BGP_EVPN_MAC_IP_ROUTE &&
1731 (is_evpn_prefix_ipaddr_v4(p) ||
1732 !IN6_IS_ADDR_LINKLOCAL(&p->prefix.macip_addr.ip.ipaddr_v6)) &&
1733 CHECK_FLAG(vpn->flags, VNI_FLAG_USE_TWO_LABELS) &&
1734 bgpevpn_get_l3vni(vpn))
1735 add_l3_ecomm = 1;
1736
1737 /* Set up extended community. */
1738 build_evpn_route_extcomm(vpn, &attr, add_l3_ecomm);
1739
1740 /* First, create (or fetch) route node within the VNI. */
1741 /* NOTE: There is no RD here. */
1742 rn = bgp_node_get(vpn->route_table, (struct prefix *)p);
1743
1744 /* Create or update route entry. */
1745 route_change = update_evpn_route_entry(bgp, vpn, afi, safi, rn, &attr,
1746 1, &ri, flags, seq);
1747 assert(ri);
1748 attr_new = ri->attr;
1749
1750 /* Perform route selection; this is just to set the flags correctly
1751 * as local route in the VNI always wins.
1752 */
1753 evpn_route_select_install(bgp, vpn, rn);
1754 bgp_unlock_node(rn);
1755
1756 /* If this is a new route or some attribute has changed, export the
1757 * route to the global table. The route will be advertised to peers
1758 * from there. Note that this table is a 2-level tree (RD-level +
1759 * Prefix-level) similar to L3VPN routes.
1760 */
1761 if (route_change) {
1762 struct bgp_info *global_ri;
1763
1764 rn = bgp_afi_node_get(bgp->rib[afi][safi], afi, safi,
1765 (struct prefix *)p, &vpn->prd);
1766 update_evpn_route_entry(bgp, vpn, afi, safi, rn, attr_new, 1,
1767 &global_ri, flags, seq);
1768
1769 /* Schedule for processing and unlock node. */
1770 bgp_process(bgp, rn, afi, safi);
1771 bgp_unlock_node(rn);
1772 }
1773
1774 /* Unintern temporary. */
1775 aspath_unintern(&attr.aspath);
1776
1777 return 0;
1778 }
1779
1780 /*
1781 * Delete EVPN route entry.
1782 * The entry can be in ESI/VNI table or the global table.
1783 */
1784 static void delete_evpn_route_entry(struct bgp *bgp,
1785 afi_t afi, safi_t safi,
1786 struct bgp_node *rn,
1787 struct bgp_info **ri)
1788 {
1789 struct bgp_info *tmp_ri;
1790
1791 *ri = NULL;
1792
1793 /* Now, find matching route. */
1794 for (tmp_ri = rn->info; tmp_ri; tmp_ri = tmp_ri->next)
1795 if (tmp_ri->peer == bgp->peer_self
1796 && tmp_ri->type == ZEBRA_ROUTE_BGP
1797 && tmp_ri->sub_type == BGP_ROUTE_STATIC)
1798 break;
1799
1800 *ri = tmp_ri;
1801
1802 /* Mark route for delete. */
1803 if (tmp_ri)
1804 bgp_info_delete(rn, tmp_ri);
1805 }
1806
1807
1808
1809 /* Delete EVPN ES (type-4) route */
1810 static int delete_evpn_type4_route(struct bgp *bgp,
1811 struct evpnes *es,
1812 struct prefix_evpn *p)
1813 {
1814 afi_t afi = AFI_L2VPN;
1815 safi_t safi = SAFI_EVPN;
1816 struct bgp_info *ri;
1817 struct bgp_node *rn = NULL; /* rn in esi table */
1818 struct bgp_node *global_rn = NULL; /* rn in global table */
1819
1820 /* First, locate the route node within the ESI.
1821 * If it doesn't exist, ther is nothing to do.
1822 * Note: there is no RD here.
1823 */
1824 rn = bgp_node_lookup(es->route_table, (struct prefix *)p);
1825 if (!rn)
1826 return 0;
1827
1828 /* Next, locate route node in the global EVPN routing table.
1829 * Note that this table is a 2-level tree (RD-level + Prefix-level)
1830 */
1831 global_rn = bgp_afi_node_lookup(bgp->rib[afi][safi], afi, safi,
1832 (struct prefix *)p, &es->prd);
1833 if (global_rn) {
1834
1835 /* Delete route entry in the global EVPN table. */
1836 delete_evpn_route_entry(bgp, afi, safi,
1837 global_rn, &ri);
1838
1839 /* Schedule for processing - withdraws to peers happen from
1840 * this table.
1841 */
1842 if (ri)
1843 bgp_process(bgp, global_rn, afi, safi);
1844 bgp_unlock_node(global_rn);
1845 }
1846
1847 /*
1848 * Delete route entry in the ESI route table.
1849 * This can just be removed.
1850 */
1851 delete_evpn_route_entry(bgp, afi, safi, rn, &ri);
1852 if (ri)
1853 bgp_info_reap(rn, ri);
1854 bgp_unlock_node(rn);
1855 return 0;
1856 }
1857
1858 /* Delete EVPN type5 route */
1859 static int delete_evpn_type5_route(struct bgp *bgp_vrf, struct prefix_evpn *evp)
1860 {
1861 afi_t afi = AFI_L2VPN;
1862 safi_t safi = SAFI_EVPN;
1863 struct bgp_node *rn = NULL;
1864 struct bgp_info *ri = NULL;
1865 struct bgp *bgp_def = NULL; /* default bgp instance */
1866
1867 bgp_def = bgp_get_default();
1868 if (!bgp_def)
1869 return 0;
1870
1871 /* locate the global route entry for this type-5 prefix */
1872 rn = bgp_afi_node_lookup(bgp_def->rib[afi][safi], afi, safi,
1873 (struct prefix *)evp, &bgp_vrf->vrf_prd);
1874 if (!rn)
1875 return 0;
1876
1877 delete_evpn_route_entry(bgp_def, afi, safi, rn, &ri);
1878 if (ri)
1879 bgp_process(bgp_def, rn, afi, safi);
1880 bgp_unlock_node(rn);
1881 return 0;
1882 }
1883
1884 /*
1885 * Delete EVPN route (of type based on prefix) for specified VNI and
1886 * schedule for processing.
1887 */
1888 static int delete_evpn_route(struct bgp *bgp, struct bgpevpn *vpn,
1889 struct prefix_evpn *p)
1890 {
1891 struct bgp_node *rn, *global_rn;
1892 struct bgp_info *ri;
1893 afi_t afi = AFI_L2VPN;
1894 safi_t safi = SAFI_EVPN;
1895
1896 /* First, locate the route node within the VNI. If it doesn't exist,
1897 * there
1898 * is nothing further to do.
1899 */
1900 /* NOTE: There is no RD here. */
1901 rn = bgp_node_lookup(vpn->route_table, (struct prefix *)p);
1902 if (!rn)
1903 return 0;
1904
1905 /* Next, locate route node in the global EVPN routing table. Note that
1906 * this table is a 2-level tree (RD-level + Prefix-level) similar to
1907 * L3VPN routes.
1908 */
1909 global_rn = bgp_afi_node_lookup(bgp->rib[afi][safi], afi, safi,
1910 (struct prefix *)p, &vpn->prd);
1911 if (global_rn) {
1912 /* Delete route entry in the global EVPN table. */
1913 delete_evpn_route_entry(bgp, afi, safi, global_rn, &ri);
1914
1915 /* Schedule for processing - withdraws to peers happen from
1916 * this table.
1917 */
1918 if (ri)
1919 bgp_process(bgp, global_rn, afi, safi);
1920 bgp_unlock_node(global_rn);
1921 }
1922
1923 /* Delete route entry in the VNI route table. This can just be removed.
1924 */
1925 delete_evpn_route_entry(bgp, afi, safi, rn, &ri);
1926 if (ri)
1927 bgp_info_reap(rn, ri);
1928 bgp_unlock_node(rn);
1929
1930 return 0;
1931 }
1932
1933 /*
1934 * Update all type-2 (MACIP) local routes for this VNI - these should also
1935 * be scheduled for advertise to peers.
1936 */
1937 static int update_all_type2_routes(struct bgp *bgp, struct bgpevpn *vpn)
1938 {
1939 afi_t afi;
1940 safi_t safi;
1941 struct bgp_node *rn;
1942 struct bgp_info *ri, *tmp_ri;
1943 struct attr attr;
1944 struct attr *attr_new;
1945 uint32_t seq;
1946 int add_l3_ecomm = 0;
1947
1948 afi = AFI_L2VPN;
1949 safi = SAFI_EVPN;
1950
1951 /* Walk this VNI's route table and update local type-2 routes. For any
1952 * routes updated, update corresponding entry in the global table too.
1953 */
1954 for (rn = bgp_table_top(vpn->route_table); rn;
1955 rn = bgp_route_next(rn)) {
1956 struct prefix_evpn *evp = (struct prefix_evpn *)&rn->p;
1957 struct bgp_node *rd_rn;
1958 struct bgp_info *global_ri;
1959
1960 if (evp->prefix.route_type != BGP_EVPN_MAC_IP_ROUTE)
1961 continue;
1962
1963 /* Identify local route. */
1964 for (tmp_ri = rn->info; tmp_ri; tmp_ri = tmp_ri->next) {
1965 if (tmp_ri->peer == bgp->peer_self
1966 && tmp_ri->type == ZEBRA_ROUTE_BGP
1967 && tmp_ri->sub_type == BGP_ROUTE_STATIC)
1968 break;
1969 }
1970
1971 if (!tmp_ri)
1972 continue;
1973
1974 /*
1975 * Build attribute per local route as the MAC mobility and
1976 * some other values could differ for different routes. The
1977 * attributes will be shared in the hash table.
1978 */
1979 bgp_attr_default_set(&attr, BGP_ORIGIN_IGP);
1980 attr.nexthop = vpn->originator_ip;
1981 attr.mp_nexthop_global_in = vpn->originator_ip;
1982 attr.mp_nexthop_len = BGP_ATTR_NHLEN_IPV4;
1983 bgpevpn_get_rmac(vpn, &attr.rmac);
1984
1985 if (evpn_route_is_sticky(bgp, rn))
1986 attr.sticky = 1;
1987 else if (evpn_route_is_def_gw(bgp, rn)) {
1988 attr.default_gw = 1;
1989 if (is_evpn_prefix_ipaddr_v6(evp))
1990 attr.router_flag = 1;
1991 }
1992
1993 /* Add L3 VNI RTs and RMAC for non IPv6 link-local if
1994 * using L3 VNI for type-2 routes also.
1995 */
1996 if ((is_evpn_prefix_ipaddr_v4(evp) ||
1997 !IN6_IS_ADDR_LINKLOCAL(
1998 &evp->prefix.macip_addr.ip.ipaddr_v6)) &&
1999 CHECK_FLAG(vpn->flags, VNI_FLAG_USE_TWO_LABELS) &&
2000 bgpevpn_get_l3vni(vpn))
2001 add_l3_ecomm = 1;
2002
2003 /* Set up extended community. */
2004 build_evpn_route_extcomm(vpn, &attr, add_l3_ecomm);
2005
2006 seq = mac_mobility_seqnum(tmp_ri->attr);
2007
2008 /* Update the route entry. */
2009 update_evpn_route_entry(bgp, vpn, afi, safi, rn,
2010 &attr, 0, &ri, 0, seq);
2011
2012 /* Perform route selection; this is just to set the flags
2013 * correctly as local route in the VNI always wins.
2014 */
2015 evpn_route_select_install(bgp, vpn, rn);
2016
2017 attr_new = ri->attr;
2018
2019 /* Update route in global routing table. */
2020 rd_rn = bgp_afi_node_get(bgp->rib[afi][safi], afi, safi,
2021 (struct prefix *)evp, &vpn->prd);
2022 assert(rd_rn);
2023 update_evpn_route_entry(bgp, vpn, afi, safi, rd_rn, attr_new, 0,
2024 &global_ri, 0,
2025 mac_mobility_seqnum(attr_new));
2026
2027 /* Schedule for processing and unlock node. */
2028 bgp_process(bgp, rd_rn, afi, safi);
2029 bgp_unlock_node(rd_rn);
2030
2031 /* Unintern temporary. */
2032 aspath_unintern(&attr.aspath);
2033
2034 }
2035
2036 return 0;
2037 }
2038
2039 /*
2040 * Delete all type-2 (MACIP) local routes for this VNI - only from the
2041 * global routing table. These are also scheduled for withdraw from peers.
2042 */
2043 static int delete_global_type2_routes(struct bgp *bgp, struct bgpevpn *vpn)
2044 {
2045 afi_t afi;
2046 safi_t safi;
2047 struct bgp_node *rdrn, *rn;
2048 struct bgp_table *table;
2049 struct bgp_info *ri;
2050
2051 afi = AFI_L2VPN;
2052 safi = SAFI_EVPN;
2053
2054 rdrn = bgp_node_lookup(bgp->rib[afi][safi], (struct prefix *)&vpn->prd);
2055 if (rdrn && rdrn->info) {
2056 table = (struct bgp_table *)rdrn->info;
2057 for (rn = bgp_table_top(table); rn; rn = bgp_route_next(rn)) {
2058 struct prefix_evpn *evp = (struct prefix_evpn *)&rn->p;
2059
2060 if (evp->prefix.route_type != BGP_EVPN_MAC_IP_ROUTE)
2061 continue;
2062
2063 delete_evpn_route_entry(bgp, afi, safi, rn, &ri);
2064 if (ri)
2065 bgp_process(bgp, rn, afi, safi);
2066 }
2067 }
2068
2069 /* Unlock RD node. */
2070 if (rdrn)
2071 bgp_unlock_node(rdrn);
2072
2073 return 0;
2074 }
2075
2076 /*
2077 * Delete all type-2 (MACIP) local routes for this VNI - from the global
2078 * table as well as the per-VNI route table.
2079 */
2080 static int delete_all_type2_routes(struct bgp *bgp, struct bgpevpn *vpn)
2081 {
2082 afi_t afi;
2083 safi_t safi;
2084 struct bgp_node *rn;
2085 struct bgp_info *ri;
2086
2087 afi = AFI_L2VPN;
2088 safi = SAFI_EVPN;
2089
2090 /* First, walk the global route table for this VNI's type-2 local
2091 * routes.
2092 * EVPN routes are a 2-level table, first get the RD table.
2093 */
2094 delete_global_type2_routes(bgp, vpn);
2095
2096 /* Next, walk this VNI's route table and delete local type-2 routes. */
2097 for (rn = bgp_table_top(vpn->route_table); rn;
2098 rn = bgp_route_next(rn)) {
2099 struct prefix_evpn *evp = (struct prefix_evpn *)&rn->p;
2100
2101 if (evp->prefix.route_type != BGP_EVPN_MAC_IP_ROUTE)
2102 continue;
2103
2104 delete_evpn_route_entry(bgp, afi, safi, rn, &ri);
2105
2106 /* Route entry in local table gets deleted immediately. */
2107 if (ri)
2108 bgp_info_reap(rn, ri);
2109 }
2110
2111 return 0;
2112 }
2113
2114 /*
2115 * Delete all routes in per ES route-table
2116 */
2117 static int delete_all_es_routes(struct bgp *bgp, struct evpnes *es)
2118 {
2119 struct bgp_node *rn;
2120 struct bgp_info *ri, *nextri;
2121
2122 /* Walk this ES's route table and delete all routes. */
2123 for (rn = bgp_table_top(es->route_table); rn;
2124 rn = bgp_route_next(rn)) {
2125 for (ri = rn->info; (ri != NULL) && (nextri = ri->next, 1);
2126 ri = nextri) {
2127 bgp_info_delete(rn, ri);
2128 bgp_info_reap(rn, ri);
2129 }
2130 }
2131
2132 return 0;
2133 }
2134
2135 /*
2136 * Delete all routes in the per-VNI route table.
2137 */
2138 static int delete_all_vni_routes(struct bgp *bgp, struct bgpevpn *vpn)
2139 {
2140 struct bgp_node *rn;
2141 struct bgp_info *ri, *nextri;
2142
2143 /* Walk this VNI's route table and delete all routes. */
2144 for (rn = bgp_table_top(vpn->route_table); rn;
2145 rn = bgp_route_next(rn)) {
2146 for (ri = rn->info; (ri != NULL) && (nextri = ri->next, 1);
2147 ri = nextri) {
2148 bgp_info_delete(rn, ri);
2149 bgp_info_reap(rn, ri);
2150 }
2151 }
2152
2153 return 0;
2154 }
2155
2156 /*
2157 * Update (and advertise) local routes for a VNI. Invoked upon the VNI
2158 * export RT getting modified or change to tunnel IP. Note that these
2159 * situations need the route in the per-VNI table as well as the global
2160 * table to be updated (as attributes change).
2161 */
2162 static int update_routes_for_vni(struct bgp *bgp, struct bgpevpn *vpn)
2163 {
2164 int ret;
2165 struct prefix_evpn p;
2166
2167 /* Update and advertise the type-3 route (only one) followed by the
2168 * locally learnt type-2 routes (MACIP) - for this VNI.
2169 */
2170 build_evpn_type3_prefix(&p, vpn->originator_ip);
2171 ret = update_evpn_route(bgp, vpn, &p, 0, 0);
2172 if (ret)
2173 return ret;
2174
2175 return update_all_type2_routes(bgp, vpn);
2176 }
2177
2178 /* Delete (and withdraw) local routes for specified ES from global and ES table.
2179 * Also remove all other routes from the per ES table.
2180 * Invoked when ES is deleted.
2181 */
2182 static int delete_routes_for_es(struct bgp *bgp, struct evpnes *es)
2183 {
2184 int ret;
2185 char buf[ESI_STR_LEN];
2186 struct prefix_evpn p;
2187
2188 /* Delete and withdraw locally learnt ES route */
2189 build_evpn_type4_prefix(&p, &es->esi, es->originator_ip.ipaddr_v4);
2190 ret = delete_evpn_type4_route(bgp, es, &p);
2191 if (ret) {
2192 flog_err(EC_BGP_EVPN_ROUTE_DELETE,
2193 "%u failed to delete type-4 route for ESI %s",
2194 bgp->vrf_id, esi_to_str(&es->esi, buf, sizeof(buf)));
2195 }
2196
2197 /* Delete all routes from per ES table */
2198 return delete_all_es_routes(bgp, es);
2199 }
2200
2201 /*
2202 * Delete (and withdraw) local routes for specified VNI from the global
2203 * table and per-VNI table. After this, remove all other routes from
2204 * the per-VNI table. Invoked upon the VNI being deleted or EVPN
2205 * (advertise-all-vni) being disabled.
2206 */
2207 static int delete_routes_for_vni(struct bgp *bgp, struct bgpevpn *vpn)
2208 {
2209 int ret;
2210 struct prefix_evpn p;
2211
2212 /* Delete and withdraw locally learnt type-2 routes (MACIP)
2213 * followed by type-3 routes (only one) - for this VNI.
2214 */
2215 ret = delete_all_type2_routes(bgp, vpn);
2216 if (ret)
2217 return ret;
2218
2219 build_evpn_type3_prefix(&p, vpn->originator_ip);
2220 ret = delete_evpn_route(bgp, vpn, &p);
2221 if (ret)
2222 return ret;
2223
2224 /* Delete all routes from the per-VNI table. */
2225 return delete_all_vni_routes(bgp, vpn);
2226 }
2227
2228 /*
2229 * There is a tunnel endpoint IP address change for this VNI, delete
2230 * prior type-3 route (if needed) and update.
2231 * Note: Route re-advertisement happens elsewhere after other processing
2232 * other changes.
2233 */
2234 static int handle_tunnel_ip_change(struct bgp *bgp, struct bgpevpn *vpn,
2235 struct in_addr originator_ip)
2236 {
2237 struct prefix_evpn p;
2238
2239 /* If VNI is not live, we only need to update the originator ip */
2240 if (!is_vni_live(vpn)) {
2241 vpn->originator_ip = originator_ip;
2242 return 0;
2243 }
2244
2245 /* Update the tunnel-ip hash */
2246 bgp_tip_del(bgp, &vpn->originator_ip);
2247 bgp_tip_add(bgp, &originator_ip);
2248
2249 /* filter routes as martian nexthop db has changed */
2250 bgp_filter_evpn_routes_upon_martian_nh_change(bgp);
2251
2252 /* Need to withdraw type-3 route as the originator IP is part
2253 * of the key.
2254 */
2255 build_evpn_type3_prefix(&p, vpn->originator_ip);
2256 delete_evpn_route(bgp, vpn, &p);
2257
2258 /* Update the tunnel IP and re-advertise all routes for this VNI. */
2259 vpn->originator_ip = originator_ip;
2260 return 0;
2261 }
2262
2263 /* Install EVPN route entry in ES */
2264 static int install_evpn_route_entry_in_es(struct bgp *bgp,
2265 struct evpnes *es,
2266 struct prefix_evpn *p,
2267 struct bgp_info *parent_ri)
2268 {
2269 int ret = 0;
2270 struct bgp_node *rn = NULL;
2271 struct bgp_info *ri = NULL;
2272 struct attr *attr_new = NULL;
2273
2274 /* Create (or fetch) route within the VNI.
2275 * NOTE: There is no RD here.
2276 */
2277 rn = bgp_node_get(es->route_table, (struct prefix *)p);
2278
2279 /* Check if route entry is already present. */
2280 for (ri = rn->info; ri; ri = ri->next)
2281 if (ri->extra &&
2282 (struct bgp_info *)ri->extra->parent == parent_ri)
2283 break;
2284
2285 if (!ri) {
2286 /* Add (or update) attribute to hash. */
2287 attr_new = bgp_attr_intern(parent_ri->attr);
2288
2289 /* Create new route with its attribute. */
2290 ri = info_make(parent_ri->type, BGP_ROUTE_IMPORTED, 0,
2291 parent_ri->peer, attr_new, rn);
2292 SET_FLAG(ri->flags, BGP_INFO_VALID);
2293 bgp_info_extra_get(ri);
2294 ri->extra->parent = parent_ri;
2295 bgp_info_add(rn, ri);
2296 } else {
2297 if (attrhash_cmp(ri->attr, parent_ri->attr)
2298 && !CHECK_FLAG(ri->flags, BGP_INFO_REMOVED)) {
2299 bgp_unlock_node(rn);
2300 return 0;
2301 }
2302 /* The attribute has changed. */
2303 /* Add (or update) attribute to hash. */
2304 attr_new = bgp_attr_intern(parent_ri->attr);
2305
2306 /* Restore route, if needed. */
2307 if (CHECK_FLAG(ri->flags, BGP_INFO_REMOVED))
2308 bgp_info_restore(rn, ri);
2309
2310 /* Mark if nexthop has changed. */
2311 if (!IPV4_ADDR_SAME(&ri->attr->nexthop, &attr_new->nexthop))
2312 SET_FLAG(ri->flags, BGP_INFO_IGP_CHANGED);
2313
2314 /* Unintern existing, set to new. */
2315 bgp_attr_unintern(&ri->attr);
2316 ri->attr = attr_new;
2317 ri->uptime = bgp_clock();
2318 }
2319
2320 /* Perform route selection and update zebra, if required. */
2321 ret = evpn_es_route_select_install(bgp, es, rn);
2322 return ret;
2323 }
2324
2325 /*
2326 * Install route entry into the VRF routing table and invoke route selection.
2327 */
2328 static int install_evpn_route_entry_in_vrf(struct bgp *bgp_vrf,
2329 struct prefix_evpn *evp,
2330 struct bgp_info *parent_ri)
2331 {
2332 struct bgp_node *rn;
2333 struct bgp_info *ri;
2334 struct attr attr;
2335 struct attr *attr_new;
2336 int ret = 0;
2337 struct prefix p;
2338 struct prefix *pp = &p;
2339 afi_t afi = 0;
2340 safi_t safi = 0;
2341 char buf[PREFIX_STRLEN];
2342 char buf1[PREFIX_STRLEN];
2343
2344 memset(pp, 0, sizeof(struct prefix));
2345 ip_prefix_from_evpn_prefix(evp, pp);
2346
2347 if (bgp_debug_zebra(NULL)) {
2348 zlog_debug(
2349 "installing evpn prefix %s as ip prefix %s in vrf %s",
2350 prefix2str(evp, buf, sizeof(buf)),
2351 prefix2str(pp, buf1, sizeof(buf)),
2352 vrf_id_to_name(bgp_vrf->vrf_id));
2353 }
2354
2355 /* Create (or fetch) route within the VRF. */
2356 /* NOTE: There is no RD here. */
2357 if (is_evpn_prefix_ipaddr_v4(evp)) {
2358 afi = AFI_IP;
2359 safi = SAFI_UNICAST;
2360 rn = bgp_node_get(bgp_vrf->rib[afi][safi], pp);
2361 } else if (is_evpn_prefix_ipaddr_v6(evp)) {
2362 afi = AFI_IP6;
2363 safi = SAFI_UNICAST;
2364 rn = bgp_node_get(bgp_vrf->rib[afi][safi], pp);
2365 } else
2366 return 0;
2367
2368 /* EVPN routes currently only support a IPv4 next hop which corresponds
2369 * to the remote VTEP. When importing into a VRF, if it is IPv6 host
2370 * or prefix route, we have to convert the next hop to an IPv4-mapped
2371 * address for the rest of the code to flow through. In the case of IPv4,
2372 * make sure to set the flag for next hop attribute.
2373 */
2374 bgp_attr_dup(&attr, parent_ri->attr);
2375 if (afi == AFI_IP6)
2376 evpn_convert_nexthop_to_ipv6(&attr);
2377 else
2378 attr.flag |= ATTR_FLAG_BIT(BGP_ATTR_NEXT_HOP);
2379
2380 /* Check if route entry is already present. */
2381 for (ri = rn->info; ri; ri = ri->next)
2382 if (ri->extra
2383 && (struct bgp_info *)ri->extra->parent == parent_ri)
2384 break;
2385
2386 if (!ri) {
2387 /* Add (or update) attribute to hash. */
2388 attr_new = bgp_attr_intern(&attr);
2389
2390 /* Create new route with its attribute. */
2391 ri = info_make(parent_ri->type, BGP_ROUTE_IMPORTED, 0,
2392 parent_ri->peer, attr_new, rn);
2393 SET_FLAG(ri->flags, BGP_INFO_VALID);
2394 bgp_info_extra_get(ri);
2395 ri->extra->parent = bgp_info_lock(parent_ri);
2396 bgp_lock_node((struct bgp_node *)parent_ri->net);
2397 if (parent_ri->extra) {
2398 memcpy(&ri->extra->label, &parent_ri->extra->label,
2399 sizeof(ri->extra->label));
2400 ri->extra->num_labels = parent_ri->extra->num_labels;
2401 }
2402 bgp_info_add(rn, ri);
2403 } else {
2404 if (attrhash_cmp(ri->attr, &attr)
2405 && !CHECK_FLAG(ri->flags, BGP_INFO_REMOVED)) {
2406 bgp_unlock_node(rn);
2407 return 0;
2408 }
2409 /* The attribute has changed. */
2410 /* Add (or update) attribute to hash. */
2411 attr_new = bgp_attr_intern(&attr);
2412
2413 /* Restore route, if needed. */
2414 if (CHECK_FLAG(ri->flags, BGP_INFO_REMOVED))
2415 bgp_info_restore(rn, ri);
2416
2417 /* Mark if nexthop has changed. */
2418 if ((afi == AFI_IP &&
2419 !IPV4_ADDR_SAME(&ri->attr->nexthop, &attr_new->nexthop)) ||
2420 (afi == AFI_IP6 &&
2421 !IPV6_ADDR_SAME(&ri->attr->mp_nexthop_global,
2422 &attr_new->mp_nexthop_global)))
2423 SET_FLAG(ri->flags, BGP_INFO_IGP_CHANGED);
2424
2425 /* Unintern existing, set to new. */
2426 bgp_attr_unintern(&ri->attr);
2427 ri->attr = attr_new;
2428 ri->uptime = bgp_clock();
2429 }
2430
2431 bgp_aggregate_increment(bgp_vrf, &rn->p, ri, afi, safi);
2432
2433 /* Perform route selection and update zebra, if required. */
2434 bgp_process(bgp_vrf, rn, afi, safi);
2435
2436 return ret;
2437 }
2438
2439 /*
2440 * Install route entry into the VNI routing table and invoke route selection.
2441 */
2442 static int install_evpn_route_entry(struct bgp *bgp, struct bgpevpn *vpn,
2443 struct prefix_evpn *p,
2444 struct bgp_info *parent_ri)
2445 {
2446 struct bgp_node *rn;
2447 struct bgp_info *ri;
2448 struct attr *attr_new;
2449 int ret;
2450
2451 /* Create (or fetch) route within the VNI. */
2452 /* NOTE: There is no RD here. */
2453 rn = bgp_node_get(vpn->route_table, (struct prefix *)p);
2454
2455 /* Check if route entry is already present. */
2456 for (ri = rn->info; ri; ri = ri->next)
2457 if (ri->extra
2458 && (struct bgp_info *)ri->extra->parent == parent_ri)
2459 break;
2460
2461 if (!ri) {
2462 /* Add (or update) attribute to hash. */
2463 attr_new = bgp_attr_intern(parent_ri->attr);
2464
2465 /* Create new route with its attribute. */
2466 ri = info_make(parent_ri->type, BGP_ROUTE_IMPORTED, 0,
2467 parent_ri->peer, attr_new, rn);
2468 SET_FLAG(ri->flags, BGP_INFO_VALID);
2469 bgp_info_extra_get(ri);
2470 ri->extra->parent = bgp_info_lock(parent_ri);
2471 bgp_lock_node((struct bgp_node *)parent_ri->net);
2472 if (parent_ri->extra) {
2473 memcpy(&ri->extra->label, &parent_ri->extra->label,
2474 sizeof(ri->extra->label));
2475 ri->extra->num_labels = parent_ri->extra->num_labels;
2476 }
2477 bgp_info_add(rn, ri);
2478 } else {
2479 if (attrhash_cmp(ri->attr, parent_ri->attr)
2480 && !CHECK_FLAG(ri->flags, BGP_INFO_REMOVED)) {
2481 bgp_unlock_node(rn);
2482 return 0;
2483 }
2484 /* The attribute has changed. */
2485 /* Add (or update) attribute to hash. */
2486 attr_new = bgp_attr_intern(parent_ri->attr);
2487
2488 /* Restore route, if needed. */
2489 if (CHECK_FLAG(ri->flags, BGP_INFO_REMOVED))
2490 bgp_info_restore(rn, ri);
2491
2492 /* Mark if nexthop has changed. */
2493 if (!IPV4_ADDR_SAME(&ri->attr->nexthop, &attr_new->nexthop))
2494 SET_FLAG(ri->flags, BGP_INFO_IGP_CHANGED);
2495
2496 /* Unintern existing, set to new. */
2497 bgp_attr_unintern(&ri->attr);
2498 ri->attr = attr_new;
2499 ri->uptime = bgp_clock();
2500 }
2501
2502 /* Perform route selection and update zebra, if required. */
2503 ret = evpn_route_select_install(bgp, vpn, rn);
2504
2505 return ret;
2506 }
2507
2508 /* Uninstall EVPN route entry from ES route table */
2509 static int uninstall_evpn_route_entry_in_es(struct bgp *bgp,
2510 struct evpnes *es,
2511 struct prefix_evpn *p,
2512 struct bgp_info *parent_ri)
2513 {
2514 int ret;
2515 struct bgp_node *rn;
2516 struct bgp_info *ri;
2517
2518 if (!es->route_table)
2519 return 0;
2520
2521 /* Locate route within the ESI.
2522 * NOTE: There is no RD here.
2523 */
2524 rn = bgp_node_lookup(es->route_table, (struct prefix *)p);
2525 if (!rn)
2526 return 0;
2527
2528 /* Find matching route entry. */
2529 for (ri = rn->info; ri; ri = ri->next)
2530 if (ri->extra &&
2531 (struct bgp_info *)ri->extra->parent == parent_ri)
2532 break;
2533
2534 if (!ri)
2535 return 0;
2536
2537 /* Mark entry for deletion */
2538 bgp_info_delete(rn, ri);
2539
2540 /* Perform route selection and update zebra, if required. */
2541 ret = evpn_es_route_select_install(bgp, es, rn);
2542
2543 /* Unlock route node. */
2544 bgp_unlock_node(rn);
2545
2546 return ret;
2547 }
2548
2549 /*
2550 * Uninstall route entry from the VRF routing table and send message
2551 * to zebra, if appropriate.
2552 */
2553 static int uninstall_evpn_route_entry_in_vrf(struct bgp *bgp_vrf,
2554 struct prefix_evpn *evp,
2555 struct bgp_info *parent_ri)
2556 {
2557 struct bgp_node *rn;
2558 struct bgp_info *ri;
2559 int ret = 0;
2560 struct prefix p;
2561 struct prefix *pp = &p;
2562 afi_t afi = 0;
2563 safi_t safi = 0;
2564 char buf[PREFIX_STRLEN];
2565 char buf1[PREFIX_STRLEN];
2566
2567 memset(pp, 0, sizeof(struct prefix));
2568 ip_prefix_from_evpn_prefix(evp, pp);
2569
2570 if (bgp_debug_zebra(NULL)) {
2571 zlog_debug(
2572 "uninstalling evpn prefix %s as ip prefix %s in vrf %s",
2573 prefix2str(evp, buf, sizeof(buf)),
2574 prefix2str(pp, buf1, sizeof(buf)),
2575 vrf_id_to_name(bgp_vrf->vrf_id));
2576 }
2577
2578 /* Locate route within the VRF. */
2579 /* NOTE: There is no RD here. */
2580 if (is_evpn_prefix_ipaddr_v4(evp)) {
2581 afi = AFI_IP;
2582 safi = SAFI_UNICAST;
2583 rn = bgp_node_lookup(bgp_vrf->rib[afi][safi], pp);
2584 } else {
2585 afi = AFI_IP6;
2586 safi = SAFI_UNICAST;
2587 rn = bgp_node_lookup(bgp_vrf->rib[afi][safi], pp);
2588 }
2589
2590 if (!rn)
2591 return 0;
2592
2593 /* Find matching route entry. */
2594 for (ri = rn->info; ri; ri = ri->next)
2595 if (ri->extra
2596 && (struct bgp_info *)ri->extra->parent == parent_ri)
2597 break;
2598
2599 if (!ri)
2600 return 0;
2601
2602 /* Mark entry for deletion */
2603 bgp_info_delete(rn, ri);
2604
2605 bgp_aggregate_decrement(bgp_vrf, &rn->p, ri, afi, safi);
2606
2607 /* Perform route selection and update zebra, if required. */
2608 bgp_process(bgp_vrf, rn, afi, safi);
2609
2610 /* Unlock route node. */
2611 bgp_unlock_node(rn);
2612
2613 return ret;
2614 }
2615
2616 /*
2617 * Uninstall route entry from the VNI routing table and send message
2618 * to zebra, if appropriate.
2619 */
2620 static int uninstall_evpn_route_entry(struct bgp *bgp, struct bgpevpn *vpn,
2621 struct prefix_evpn *p,
2622 struct bgp_info *parent_ri)
2623 {
2624 struct bgp_node *rn;
2625 struct bgp_info *ri;
2626 int ret;
2627
2628 /* Locate route within the VNI. */
2629 /* NOTE: There is no RD here. */
2630 rn = bgp_node_lookup(vpn->route_table, (struct prefix *)p);
2631 if (!rn)
2632 return 0;
2633
2634 /* Find matching route entry. */
2635 for (ri = rn->info; ri; ri = ri->next)
2636 if (ri->extra
2637 && (struct bgp_info *)ri->extra->parent == parent_ri)
2638 break;
2639
2640 if (!ri)
2641 return 0;
2642
2643 /* Mark entry for deletion */
2644 bgp_info_delete(rn, ri);
2645
2646 /* Perform route selection and update zebra, if required. */
2647 ret = evpn_route_select_install(bgp, vpn, rn);
2648
2649 /* Unlock route node. */
2650 bgp_unlock_node(rn);
2651
2652 return ret;
2653 }
2654
2655 /*
2656 * Given a prefix, see if it belongs to ES.
2657 */
2658 static int is_prefix_matching_for_es(struct prefix_evpn *p,
2659 struct evpnes *es)
2660 {
2661 /* if not an ES route return false */
2662 if (p->prefix.route_type != BGP_EVPN_ES_ROUTE)
2663 return 0;
2664
2665 if (memcmp(&p->prefix.es_addr.esi, &es->esi, sizeof(esi_t)) == 0)
2666 return 1;
2667
2668 return 0;
2669 }
2670
2671 /*
2672 * Given a route entry and a VRF, see if this route entry should be
2673 * imported into the VRF i.e., RTs match.
2674 */
2675 static int is_route_matching_for_vrf(struct bgp *bgp_vrf, struct bgp_info *ri)
2676 {
2677 struct attr *attr = ri->attr;
2678 struct ecommunity *ecom;
2679 int i;
2680
2681 assert(attr);
2682 /* Route should have valid RT to be even considered. */
2683 if (!(attr->flag & ATTR_FLAG_BIT(BGP_ATTR_EXT_COMMUNITIES)))
2684 return 0;
2685
2686 ecom = attr->ecommunity;
2687 if (!ecom || !ecom->size)
2688 return 0;
2689
2690 /* For each extended community RT, see if it matches this VNI. If any RT
2691 * matches, we're done.
2692 */
2693 for (i = 0; i < ecom->size; i++) {
2694 uint8_t *pnt;
2695 uint8_t type, sub_type;
2696 struct ecommunity_val *eval;
2697 struct ecommunity_val eval_tmp;
2698 struct vrf_irt_node *irt;
2699
2700 /* Only deal with RTs */
2701 pnt = (ecom->val + (i * ECOMMUNITY_SIZE));
2702 eval = (struct ecommunity_val *)(ecom->val
2703 + (i * ECOMMUNITY_SIZE));
2704 type = *pnt++;
2705 sub_type = *pnt++;
2706 if (sub_type != ECOMMUNITY_ROUTE_TARGET)
2707 continue;
2708
2709 /* See if this RT matches specified VNIs import RTs */
2710 irt = lookup_vrf_import_rt(eval);
2711 if (irt)
2712 if (is_vrf_present_in_irt_vrfs(irt->vrfs, bgp_vrf))
2713 return 1;
2714
2715 /* Also check for non-exact match. In this, we mask out the AS
2716 * and
2717 * only check on the local-admin sub-field. This is to
2718 * facilitate using
2719 * VNI as the RT for EBGP peering too.
2720 */
2721 irt = NULL;
2722 if (type == ECOMMUNITY_ENCODE_AS
2723 || type == ECOMMUNITY_ENCODE_AS4
2724 || type == ECOMMUNITY_ENCODE_IP) {
2725 memcpy(&eval_tmp, eval, ECOMMUNITY_SIZE);
2726 mask_ecom_global_admin(&eval_tmp, eval);
2727 irt = lookup_vrf_import_rt(&eval_tmp);
2728 }
2729 if (irt)
2730 if (is_vrf_present_in_irt_vrfs(irt->vrfs, bgp_vrf))
2731 return 1;
2732 }
2733
2734 return 0;
2735 }
2736
2737 /*
2738 * Given a route entry and a VNI, see if this route entry should be
2739 * imported into the VNI i.e., RTs match.
2740 */
2741 static int is_route_matching_for_vni(struct bgp *bgp, struct bgpevpn *vpn,
2742 struct bgp_info *ri)
2743 {
2744 struct attr *attr = ri->attr;
2745 struct ecommunity *ecom;
2746 int i;
2747
2748 assert(attr);
2749 /* Route should have valid RT to be even considered. */
2750 if (!(attr->flag & ATTR_FLAG_BIT(BGP_ATTR_EXT_COMMUNITIES)))
2751 return 0;
2752
2753 ecom = attr->ecommunity;
2754 if (!ecom || !ecom->size)
2755 return 0;
2756
2757 /* For each extended community RT, see if it matches this VNI. If any RT
2758 * matches, we're done.
2759 */
2760 for (i = 0; i < ecom->size; i++) {
2761 uint8_t *pnt;
2762 uint8_t type, sub_type;
2763 struct ecommunity_val *eval;
2764 struct ecommunity_val eval_tmp;
2765 struct irt_node *irt;
2766
2767 /* Only deal with RTs */
2768 pnt = (ecom->val + (i * ECOMMUNITY_SIZE));
2769 eval = (struct ecommunity_val *)(ecom->val
2770 + (i * ECOMMUNITY_SIZE));
2771 type = *pnt++;
2772 sub_type = *pnt++;
2773 if (sub_type != ECOMMUNITY_ROUTE_TARGET)
2774 continue;
2775
2776 /* See if this RT matches specified VNIs import RTs */
2777 irt = lookup_import_rt(bgp, eval);
2778 if (irt)
2779 if (is_vni_present_in_irt_vnis(irt->vnis, vpn))
2780 return 1;
2781
2782 /* Also check for non-exact match. In this, we mask out the AS
2783 * and
2784 * only check on the local-admin sub-field. This is to
2785 * facilitate using
2786 * VNI as the RT for EBGP peering too.
2787 */
2788 irt = NULL;
2789 if (type == ECOMMUNITY_ENCODE_AS
2790 || type == ECOMMUNITY_ENCODE_AS4
2791 || type == ECOMMUNITY_ENCODE_IP) {
2792 memcpy(&eval_tmp, eval, ECOMMUNITY_SIZE);
2793 mask_ecom_global_admin(&eval_tmp, eval);
2794 irt = lookup_import_rt(bgp, &eval_tmp);
2795 }
2796 if (irt)
2797 if (is_vni_present_in_irt_vnis(irt->vnis, vpn))
2798 return 1;
2799 }
2800
2801 return 0;
2802 }
2803
2804 static int install_uninstall_routes_for_es(struct bgp *bgp,
2805 struct evpnes *es,
2806 int install)
2807 {
2808 int ret;
2809 afi_t afi;
2810 safi_t safi;
2811 char buf[PREFIX_STRLEN];
2812 char buf1[ESI_STR_LEN];
2813 struct bgp_node *rd_rn, *rn;
2814 struct bgp_table *table;
2815 struct bgp_info *ri;
2816
2817 afi = AFI_L2VPN;
2818 safi = SAFI_EVPN;
2819
2820 /*
2821 * Walk entire global routing table and evaluate routes which could be
2822 * imported into this VRF. Note that we need to loop through all global
2823 * routes to determine which route matches the import rt on vrf
2824 */
2825 for (rd_rn = bgp_table_top(bgp->rib[afi][safi]); rd_rn;
2826 rd_rn = bgp_route_next(rd_rn)) {
2827 table = (struct bgp_table *)(rd_rn->info);
2828 if (!table)
2829 continue;
2830
2831 for (rn = bgp_table_top(table); rn; rn = bgp_route_next(rn)) {
2832 struct prefix_evpn *evp = (struct prefix_evpn *)&rn->p;
2833
2834 for (ri = rn->info; ri; ri = ri->next) {
2835 /*
2836 * Consider "valid" remote routes applicable for
2837 * this ES.
2838 */
2839 if (!(CHECK_FLAG(ri->flags, BGP_INFO_VALID)
2840 && ri->type == ZEBRA_ROUTE_BGP
2841 && ri->sub_type == BGP_ROUTE_NORMAL))
2842 continue;
2843
2844 if (!is_prefix_matching_for_es(evp, es))
2845 continue;
2846
2847 if (install)
2848 ret = install_evpn_route_entry_in_es(
2849 bgp, es, evp, ri);
2850 else
2851 ret = uninstall_evpn_route_entry_in_es(
2852 bgp, es, evp, ri);
2853
2854 if (ret) {
2855 flog_err(
2856 EC_BGP_EVPN_FAIL,
2857 "Failed to %s EVPN %s route in ESI %s",
2858 install ? "install"
2859 : "uninstall",
2860 prefix2str(evp, buf,
2861 sizeof(buf)),
2862 esi_to_str(&es->esi, buf1,
2863 sizeof(buf1)));
2864 return ret;
2865 }
2866 }
2867 }
2868 }
2869 return 0;
2870 }
2871
2872 /*
2873 * Install or uninstall mac-ip routes are appropriate for this
2874 * particular VRF.
2875 */
2876 static int install_uninstall_routes_for_vrf(struct bgp *bgp_vrf, int install)
2877 {
2878 afi_t afi;
2879 safi_t safi;
2880 struct bgp_node *rd_rn, *rn;
2881 struct bgp_table *table;
2882 struct bgp_info *ri;
2883 int ret;
2884 char buf[PREFIX_STRLEN];
2885 struct bgp *bgp_def = NULL;
2886
2887 afi = AFI_L2VPN;
2888 safi = SAFI_EVPN;
2889 bgp_def = bgp_get_default();
2890 if (!bgp_def)
2891 return -1;
2892
2893 /* Walk entire global routing table and evaluate routes which could be
2894 * imported into this VRF. Note that we need to loop through all global
2895 * routes to determine which route matches the import rt on vrf
2896 */
2897 for (rd_rn = bgp_table_top(bgp_def->rib[afi][safi]); rd_rn;
2898 rd_rn = bgp_route_next(rd_rn)) {
2899 table = (struct bgp_table *)(rd_rn->info);
2900 if (!table)
2901 continue;
2902
2903 for (rn = bgp_table_top(table); rn; rn = bgp_route_next(rn)) {
2904 struct prefix_evpn *evp = (struct prefix_evpn *)&rn->p;
2905
2906 /* if not mac-ip route skip this route */
2907 if (!(evp->prefix.route_type == BGP_EVPN_MAC_IP_ROUTE
2908 || evp->prefix.route_type
2909 == BGP_EVPN_IP_PREFIX_ROUTE))
2910 continue;
2911
2912 /* if not a mac+ip route skip this route */
2913 if (!(is_evpn_prefix_ipaddr_v4(evp)
2914 || is_evpn_prefix_ipaddr_v6(evp)))
2915 continue;
2916
2917 for (ri = rn->info; ri; ri = ri->next) {
2918 /* Consider "valid" remote routes applicable for
2919 * this VRF.
2920 */
2921 if (!(CHECK_FLAG(ri->flags, BGP_INFO_VALID)
2922 && ri->type == ZEBRA_ROUTE_BGP
2923 && ri->sub_type == BGP_ROUTE_NORMAL))
2924 continue;
2925
2926 if (is_route_matching_for_vrf(bgp_vrf, ri)) {
2927 if (install)
2928 ret = install_evpn_route_entry_in_vrf(
2929 bgp_vrf, evp, ri);
2930 else
2931 ret = uninstall_evpn_route_entry_in_vrf(
2932 bgp_vrf, evp, ri);
2933
2934 if (ret) {
2935 flog_err(
2936 EC_BGP_EVPN_FAIL,
2937 "Failed to %s EVPN %s route in VRF %s",
2938 install ? "install"
2939 : "uninstall",
2940 prefix2str(evp, buf,
2941 sizeof(buf)),
2942 vrf_id_to_name(
2943 bgp_vrf->vrf_id));
2944 return ret;
2945 }
2946 }
2947 }
2948 }
2949 }
2950
2951 return 0;
2952 }
2953
2954 /*
2955 * Install or uninstall routes of specified type that are appropriate for this
2956 * particular VNI.
2957 */
2958 static int install_uninstall_routes_for_vni(struct bgp *bgp,
2959 struct bgpevpn *vpn,
2960 bgp_evpn_route_type rtype,
2961 int install)
2962 {
2963 afi_t afi;
2964 safi_t safi;
2965 struct bgp_node *rd_rn, *rn;
2966 struct bgp_table *table;
2967 struct bgp_info *ri;
2968 int ret;
2969
2970 afi = AFI_L2VPN;
2971 safi = SAFI_EVPN;
2972
2973 /* Walk entire global routing table and evaluate routes which could be
2974 * imported into this VPN. Note that we cannot just look at the routes
2975 * for
2976 * the VNI's RD - remote routes applicable for this VNI could have any
2977 * RD.
2978 */
2979 /* EVPN routes are a 2-level table. */
2980 for (rd_rn = bgp_table_top(bgp->rib[afi][safi]); rd_rn;
2981 rd_rn = bgp_route_next(rd_rn)) {
2982 table = (struct bgp_table *)(rd_rn->info);
2983 if (!table)
2984 continue;
2985
2986 for (rn = bgp_table_top(table); rn; rn = bgp_route_next(rn)) {
2987 struct prefix_evpn *evp = (struct prefix_evpn *)&rn->p;
2988
2989 if (evp->prefix.route_type != rtype)
2990 continue;
2991
2992 for (ri = rn->info; ri; ri = ri->next) {
2993 /* Consider "valid" remote routes applicable for
2994 * this VNI. */
2995 if (!(CHECK_FLAG(ri->flags, BGP_INFO_VALID)
2996 && ri->type == ZEBRA_ROUTE_BGP
2997 && ri->sub_type == BGP_ROUTE_NORMAL))
2998 continue;
2999
3000 if (is_route_matching_for_vni(bgp, vpn, ri)) {
3001 if (install)
3002 ret = install_evpn_route_entry(
3003 bgp, vpn, evp, ri);
3004 else
3005 ret = uninstall_evpn_route_entry(
3006 bgp, vpn, evp, ri);
3007
3008 if (ret) {
3009 flog_err(
3010 EC_BGP_EVPN_FAIL,
3011 "%u: Failed to %s EVPN %s route in VNI %u",
3012 bgp->vrf_id,
3013 install ? "install"
3014 : "uninstall",
3015 rtype == BGP_EVPN_MAC_IP_ROUTE
3016 ? "MACIP"
3017 : "IMET",
3018 vpn->vni);
3019 return ret;
3020 }
3021 }
3022 }
3023 }
3024 }
3025
3026 return 0;
3027 }
3028
3029 /* Install any existing remote ES routes applicable for this ES into its routing
3030 * table. This is invoked when ES comes up.
3031 */
3032 static int install_routes_for_es(struct bgp *bgp, struct evpnes *es)
3033 {
3034 return install_uninstall_routes_for_es(bgp, es, 1);
3035 }
3036
3037
3038 /* Install any existing remote routes applicable for this VRF into VRF RIB. This
3039 * is invoked upon l3vni-add or l3vni import rt change
3040 */
3041 static int install_routes_for_vrf(struct bgp *bgp_vrf)
3042 {
3043 install_uninstall_routes_for_vrf(bgp_vrf, 1);
3044 return 0;
3045 }
3046
3047 /*
3048 * Install any existing remote routes applicable for this VNI into its
3049 * routing table. This is invoked when a VNI becomes "live" or its Import
3050 * RT is changed.
3051 */
3052 static int install_routes_for_vni(struct bgp *bgp, struct bgpevpn *vpn)
3053 {
3054 int ret;
3055
3056 /* Install type-3 routes followed by type-2 routes - the ones applicable
3057 * for this VNI.
3058 */
3059 ret = install_uninstall_routes_for_vni(bgp, vpn, BGP_EVPN_IMET_ROUTE,
3060 1);
3061 if (ret)
3062 return ret;
3063
3064 return install_uninstall_routes_for_vni(bgp, vpn, BGP_EVPN_MAC_IP_ROUTE,
3065 1);
3066 }
3067
3068 /* uninstall routes from l3vni vrf. */
3069 static int uninstall_routes_for_vrf(struct bgp *bgp_vrf)
3070 {
3071 install_uninstall_routes_for_vrf(bgp_vrf, 0);
3072 return 0;
3073 }
3074
3075 /*
3076 * Uninstall any existing remote routes for this VNI. One scenario in which
3077 * this is invoked is upon an import RT change.
3078 */
3079 static int uninstall_routes_for_vni(struct bgp *bgp, struct bgpevpn *vpn)
3080 {
3081 int ret;
3082
3083 /* Uninstall type-2 routes followed by type-3 routes - the ones
3084 * applicable
3085 * for this VNI.
3086 */
3087 ret = install_uninstall_routes_for_vni(bgp, vpn, BGP_EVPN_MAC_IP_ROUTE,
3088 0);
3089 if (ret)
3090 return ret;
3091
3092 return install_uninstall_routes_for_vni(bgp, vpn, BGP_EVPN_IMET_ROUTE,
3093 0);
3094 }
3095
3096 /* Install or unistall route in ES */
3097 static int install_uninstall_route_in_es(struct bgp *bgp, struct evpnes *es,
3098 afi_t afi, safi_t safi,
3099 struct prefix_evpn *evp,
3100 struct bgp_info *ri,
3101 int install)
3102 {
3103 int ret = 0;
3104 char buf[ESI_STR_LEN];
3105
3106 if (install)
3107 ret = install_evpn_route_entry_in_es(bgp, es, evp, ri);
3108 else
3109 ret = uninstall_evpn_route_entry_in_es(bgp, es, evp, ri);
3110
3111 if (ret) {
3112 flog_err(
3113 EC_BGP_EVPN_FAIL,
3114 "%u: Failed to %s EVPN %s route in ESI %s", bgp->vrf_id,
3115 install ? "install" : "uninstall", "ES",
3116 esi_to_str(&evp->prefix.es_addr.esi, buf, sizeof(buf)));
3117 return ret;
3118 }
3119 return 0;
3120 }
3121
3122 /*
3123 * Install or uninstall route in matching VRFs (list).
3124 */
3125 static int install_uninstall_route_in_vrfs(struct bgp *bgp_def, afi_t afi,
3126 safi_t safi, struct prefix_evpn *evp,
3127 struct bgp_info *ri,
3128 struct list *vrfs, int install)
3129 {
3130 char buf[PREFIX2STR_BUFFER];
3131 struct bgp *bgp_vrf;
3132 struct listnode *node, *nnode;
3133
3134 /* Only type-2/type-5 routes go into a VRF */
3135 if (!(evp->prefix.route_type == BGP_EVPN_MAC_IP_ROUTE
3136 || evp->prefix.route_type == BGP_EVPN_IP_PREFIX_ROUTE))
3137 return 0;
3138
3139 /* if it is type-2 route and not a mac+ip route skip this route */
3140 if ((evp->prefix.route_type == BGP_EVPN_MAC_IP_ROUTE)
3141 && !(is_evpn_prefix_ipaddr_v4(evp)
3142 || is_evpn_prefix_ipaddr_v6(evp)))
3143 return 0;
3144
3145 for (ALL_LIST_ELEMENTS(vrfs, node, nnode, bgp_vrf)) {
3146 int ret;
3147
3148 if (install)
3149 ret = install_evpn_route_entry_in_vrf(bgp_vrf, evp, ri);
3150 else
3151 ret = uninstall_evpn_route_entry_in_vrf(bgp_vrf, evp,
3152 ri);
3153
3154 if (ret) {
3155 flog_err(EC_BGP_EVPN_FAIL,
3156 "%u: Failed to %s prefix %s in VRF %s",
3157 bgp_def->vrf_id,
3158 install ? "install" : "uninstall",
3159 prefix2str(evp, buf, sizeof(buf)),
3160 vrf_id_to_name(bgp_vrf->vrf_id));
3161 return ret;
3162 }
3163 }
3164
3165 return 0;
3166 }
3167
3168 /*
3169 * Install or uninstall route in matching VNIs (list).
3170 */
3171 static int install_uninstall_route_in_vnis(struct bgp *bgp, afi_t afi,
3172 safi_t safi, struct prefix_evpn *evp,
3173 struct bgp_info *ri,
3174 struct list *vnis, int install)
3175 {
3176 struct bgpevpn *vpn;
3177 struct listnode *node, *nnode;
3178
3179 for (ALL_LIST_ELEMENTS(vnis, node, nnode, vpn)) {
3180 int ret;
3181
3182 if (!is_vni_live(vpn))
3183 continue;
3184
3185 if (install)
3186 ret = install_evpn_route_entry(bgp, vpn, evp, ri);
3187 else
3188 ret = uninstall_evpn_route_entry(bgp, vpn, evp, ri);
3189
3190 if (ret) {
3191 flog_err(EC_BGP_EVPN_FAIL,
3192 "%u: Failed to %s EVPN %s route in VNI %u",
3193 bgp->vrf_id, install ? "install" : "uninstall",
3194 evp->prefix.route_type == BGP_EVPN_MAC_IP_ROUTE
3195 ? "MACIP"
3196 : "IMET",
3197 vpn->vni);
3198 return ret;
3199 }
3200 }
3201
3202 return 0;
3203 }
3204
3205 /*
3206 * Install or uninstall route for appropriate VNIs/ESIs.
3207 */
3208 static int install_uninstall_evpn_route(struct bgp *bgp, afi_t afi, safi_t safi,
3209 struct prefix *p, struct bgp_info *ri,
3210 int import)
3211 {
3212 struct prefix_evpn *evp = (struct prefix_evpn *)p;
3213 struct attr *attr = ri->attr;
3214 struct ecommunity *ecom;
3215 int i;
3216
3217 assert(attr);
3218
3219 /* Only type-2, type-3, type-4 and type-5 are supported currently */
3220 if (!(evp->prefix.route_type == BGP_EVPN_MAC_IP_ROUTE
3221 || evp->prefix.route_type == BGP_EVPN_IMET_ROUTE
3222 || evp->prefix.route_type == BGP_EVPN_ES_ROUTE
3223 || evp->prefix.route_type == BGP_EVPN_IP_PREFIX_ROUTE))
3224 return 0;
3225
3226 /* If we don't have Route Target, nothing much to do. */
3227 if (!(attr->flag & ATTR_FLAG_BIT(BGP_ATTR_EXT_COMMUNITIES)))
3228 return 0;
3229
3230 ecom = attr->ecommunity;
3231 if (!ecom || !ecom->size)
3232 return -1;
3233
3234 /* An EVPN route belongs to a VNI or a VRF or an ESI based on the RTs
3235 * attached to the route */
3236 for (i = 0; i < ecom->size; i++) {
3237 uint8_t *pnt;
3238 uint8_t type, sub_type;
3239 struct ecommunity_val *eval;
3240 struct ecommunity_val eval_tmp;
3241 struct irt_node *irt; /* import rt for l2vni */
3242 struct vrf_irt_node *vrf_irt; /* import rt for l3vni */
3243 struct evpnes *es;
3244
3245 /* Only deal with RTs */
3246 pnt = (ecom->val + (i * ECOMMUNITY_SIZE));
3247 eval = (struct ecommunity_val *)(ecom->val
3248 + (i * ECOMMUNITY_SIZE));
3249 type = *pnt++;
3250 sub_type = *pnt++;
3251 if (sub_type != ECOMMUNITY_ROUTE_TARGET)
3252 continue;
3253
3254 /*
3255 * macip routes (type-2) are imported into VNI and VRF tables.
3256 * IMET route is imported into VNI table.
3257 * prefix routes are imported into VRF table.
3258 */
3259 if (evp->prefix.route_type == BGP_EVPN_MAC_IP_ROUTE ||
3260 evp->prefix.route_type == BGP_EVPN_IMET_ROUTE ||
3261 evp->prefix.route_type == BGP_EVPN_IP_PREFIX_ROUTE) {
3262
3263 irt = lookup_import_rt(bgp, eval);
3264 if (irt)
3265 install_uninstall_route_in_vnis(bgp, afi, safi,
3266 evp, ri,
3267 irt->vnis,
3268 import);
3269
3270 vrf_irt = lookup_vrf_import_rt(eval);
3271 if (vrf_irt)
3272 install_uninstall_route_in_vrfs(bgp, afi, safi,
3273 evp, ri,
3274 vrf_irt->vrfs,
3275 import);
3276
3277 /* Also check for non-exact match.
3278 * In this, we mask out the AS and
3279 * only check on the local-admin sub-field.
3280 * This is to facilitate using
3281 * VNI as the RT for EBGP peering too.
3282 */
3283 irt = NULL;
3284 vrf_irt = NULL;
3285 if (type == ECOMMUNITY_ENCODE_AS
3286 || type == ECOMMUNITY_ENCODE_AS4
3287 || type == ECOMMUNITY_ENCODE_IP) {
3288 memcpy(&eval_tmp, eval, ECOMMUNITY_SIZE);
3289 mask_ecom_global_admin(&eval_tmp, eval);
3290 irt = lookup_import_rt(bgp, &eval_tmp);
3291 vrf_irt = lookup_vrf_import_rt(&eval_tmp);
3292 }
3293
3294 if (irt)
3295 install_uninstall_route_in_vnis(bgp, afi, safi,
3296 evp, ri,
3297 irt->vnis,
3298 import);
3299 if (vrf_irt)
3300 install_uninstall_route_in_vrfs(bgp, afi, safi,
3301 evp, ri,
3302 vrf_irt->vrfs,
3303 import);
3304 }
3305
3306 /* es route is imported into the es table */
3307 if (evp->prefix.route_type == BGP_EVPN_ES_ROUTE) {
3308
3309 /* we will match based on the entire esi to avoid
3310 * imoort of an es route for esi2 into esi1
3311 */
3312 es = bgp_evpn_lookup_es(bgp, &evp->prefix.es_addr.esi);
3313 if (es && is_es_local(es))
3314 install_uninstall_route_in_es(bgp, es,
3315 afi, safi,
3316 evp, ri, import);
3317 }
3318 }
3319
3320 return 0;
3321 }
3322
3323 /*
3324 * delete and withdraw all ipv4 and ipv6 routes in the vrf table as type-5
3325 * routes
3326 */
3327 static void delete_withdraw_vrf_routes(struct bgp *bgp_vrf)
3328 {
3329 /* delete all ipv4 routes and withdraw from peers */
3330 if (advertise_type5_routes(bgp_vrf, AFI_IP))
3331 bgp_evpn_withdraw_type5_routes(bgp_vrf, AFI_IP, SAFI_UNICAST);
3332
3333 /* delete all ipv6 routes and withdraw from peers */
3334 if (advertise_type5_routes(bgp_vrf, AFI_IP6))
3335 bgp_evpn_withdraw_type5_routes(bgp_vrf, AFI_IP6, SAFI_UNICAST);
3336 }
3337
3338 /*
3339 * update and advertise all ipv4 and ipv6 routes in thr vrf table as type-5
3340 * routes
3341 */
3342 static void update_advertise_vrf_routes(struct bgp *bgp_vrf)
3343 {
3344 /* update all ipv4 routes */
3345 if (advertise_type5_routes(bgp_vrf, AFI_IP))
3346 bgp_evpn_advertise_type5_routes(bgp_vrf, AFI_IP, SAFI_UNICAST);
3347
3348 /* update all ipv6 routes */
3349 if (advertise_type5_routes(bgp_vrf, AFI_IP6))
3350 bgp_evpn_advertise_type5_routes(bgp_vrf, AFI_IP6, SAFI_UNICAST);
3351 }
3352
3353 /*
3354 * update and advertise local routes for a VRF as type-5 routes.
3355 * This is invoked upon RD change for a VRF. Note taht the processing is only
3356 * done in the global route table using the routes which already exist in the
3357 * VRF routing table
3358 */
3359 static void update_router_id_vrf(struct bgp *bgp_vrf)
3360 {
3361 /* skip if the RD is configured */
3362 if (is_vrf_rd_configured(bgp_vrf))
3363 return;
3364
3365 /* derive the RD for the VRF based on new router-id */
3366 bgp_evpn_derive_auto_rd_for_vrf(bgp_vrf);
3367
3368 /* update advertise ipv4|ipv6 routes as type-5 routes */
3369 update_advertise_vrf_routes(bgp_vrf);
3370 }
3371
3372 /*
3373 * Delete and withdraw all type-5 routes for the RD corresponding to VRF.
3374 * This is invoked upon VRF RD change. The processing is done only from global
3375 * table.
3376 */
3377 static void withdraw_router_id_vrf(struct bgp *bgp_vrf)
3378 {
3379 /* skip if the RD is configured */
3380 if (is_vrf_rd_configured(bgp_vrf))
3381 return;
3382
3383 /* delete/withdraw ipv4|ipv6 routes as type-5 routes */
3384 delete_withdraw_vrf_routes(bgp_vrf);
3385 }
3386
3387 /*
3388 * Update and advertise local routes for a VNI. Invoked upon router-id
3389 * change. Note that the processing is done only on the global route table
3390 * using routes that already exist in the per-VNI table.
3391 */
3392 static int update_advertise_vni_routes(struct bgp *bgp, struct bgpevpn *vpn)
3393 {
3394 struct prefix_evpn p;
3395 struct bgp_node *rn, *global_rn;
3396 struct bgp_info *ri, *global_ri;
3397 struct attr *attr;
3398 afi_t afi = AFI_L2VPN;
3399 safi_t safi = SAFI_EVPN;
3400
3401 /* Locate type-3 route for VNI in the per-VNI table and use its
3402 * attributes to create and advertise the type-3 route for this VNI
3403 * in the global table.
3404 */
3405 build_evpn_type3_prefix(&p, vpn->originator_ip);
3406 rn = bgp_node_lookup(vpn->route_table, (struct prefix *)&p);
3407 if (!rn) /* unexpected */
3408 return 0;
3409 for (ri = rn->info; ri; ri = ri->next)
3410 if (ri->peer == bgp->peer_self && ri->type == ZEBRA_ROUTE_BGP
3411 && ri->sub_type == BGP_ROUTE_STATIC)
3412 break;
3413 if (!ri) /* unexpected */
3414 return 0;
3415 attr = ri->attr;
3416
3417 global_rn = bgp_afi_node_get(bgp->rib[afi][safi], afi, safi,
3418 (struct prefix *)&p, &vpn->prd);
3419 update_evpn_route_entry(bgp, vpn, afi, safi, global_rn, attr, 1, &ri,
3420 0, mac_mobility_seqnum(attr));
3421
3422 /* Schedule for processing and unlock node. */
3423 bgp_process(bgp, global_rn, afi, safi);
3424 bgp_unlock_node(global_rn);
3425
3426 /* Now, walk this VNI's route table and use the route and its attribute
3427 * to create and schedule route in global table.
3428 */
3429 for (rn = bgp_table_top(vpn->route_table); rn;
3430 rn = bgp_route_next(rn)) {
3431 struct prefix_evpn *evp = (struct prefix_evpn *)&rn->p;
3432
3433 /* Identify MAC-IP local routes. */
3434 if (evp->prefix.route_type != BGP_EVPN_MAC_IP_ROUTE)
3435 continue;
3436
3437 for (ri = rn->info; ri; ri = ri->next)
3438 if (ri->peer == bgp->peer_self
3439 && ri->type == ZEBRA_ROUTE_BGP
3440 && ri->sub_type == BGP_ROUTE_STATIC)
3441 break;
3442 if (!ri)
3443 continue;
3444
3445 /* Create route in global routing table using this route entry's
3446 * attribute.
3447 */
3448 attr = ri->attr;
3449 global_rn = bgp_afi_node_get(bgp->rib[afi][safi], afi, safi,
3450 (struct prefix *)evp, &vpn->prd);
3451 assert(global_rn);
3452 update_evpn_route_entry(bgp, vpn, afi, safi, global_rn, attr, 1,
3453 &global_ri, 0,
3454 mac_mobility_seqnum(attr));
3455
3456 /* Schedule for processing and unlock node. */
3457 bgp_process(bgp, global_rn, afi, safi);
3458 bgp_unlock_node(global_rn);
3459 }
3460
3461 return 0;
3462 }
3463
3464 /*
3465 * Delete (and withdraw) local routes for a VNI - only from the global
3466 * table. Invoked upon router-id change.
3467 */
3468 static int delete_withdraw_vni_routes(struct bgp *bgp, struct bgpevpn *vpn)
3469 {
3470 int ret;
3471 struct prefix_evpn p;
3472 struct bgp_node *global_rn;
3473 struct bgp_info *ri;
3474 afi_t afi = AFI_L2VPN;
3475 safi_t safi = SAFI_EVPN;
3476
3477 /* Delete and withdraw locally learnt type-2 routes (MACIP)
3478 * for this VNI - from the global table.
3479 */
3480 ret = delete_global_type2_routes(bgp, vpn);
3481 if (ret)
3482 return ret;
3483
3484 /* Remove type-3 route for this VNI from global table. */
3485 build_evpn_type3_prefix(&p, vpn->originator_ip);
3486 global_rn = bgp_afi_node_lookup(bgp->rib[afi][safi], afi, safi,
3487 (struct prefix *)&p, &vpn->prd);
3488 if (global_rn) {
3489 /* Delete route entry in the global EVPN table. */
3490 delete_evpn_route_entry(bgp, afi, safi, global_rn, &ri);
3491
3492 /* Schedule for processing - withdraws to peers happen from
3493 * this table.
3494 */
3495 if (ri)
3496 bgp_process(bgp, global_rn, afi, safi);
3497 bgp_unlock_node(global_rn);
3498 }
3499
3500 return 0;
3501 }
3502
3503 /*
3504 * Handle router-id change. Update and advertise local routes corresponding
3505 * to this VNI from peers. Note that this is invoked after updating the
3506 * router-id. The routes in the per-VNI table are used to create routes in
3507 * the global table and schedule them.
3508 */
3509 static void update_router_id_vni(struct hash_backet *backet, struct bgp *bgp)
3510 {
3511 struct bgpevpn *vpn = (struct bgpevpn *)backet->data;
3512
3513 /* Skip VNIs with configured RD. */
3514 if (is_rd_configured(vpn))
3515 return;
3516
3517 bgp_evpn_derive_auto_rd(bgp, vpn);
3518 update_advertise_vni_routes(bgp, vpn);
3519 }
3520
3521 /*
3522 * Handle router-id change. Delete and withdraw local routes corresponding
3523 * to this VNI from peers. Note that this is invoked prior to updating
3524 * the router-id and is done only on the global route table, the routes
3525 * are needed in the per-VNI table to re-advertise with new router id.
3526 */
3527 static void withdraw_router_id_vni(struct hash_backet *backet, struct bgp *bgp)
3528 {
3529 struct bgpevpn *vpn = (struct bgpevpn *)backet->data;
3530
3531 /* Skip VNIs with configured RD. */
3532 if (is_rd_configured(vpn))
3533 return;
3534
3535 delete_withdraw_vni_routes(bgp, vpn);
3536 }
3537
3538 /*
3539 * Process received EVPN type-2 route (advertise or withdraw).
3540 */
3541 static int process_type2_route(struct peer *peer, afi_t afi, safi_t safi,
3542 struct attr *attr, uint8_t *pfx, int psize,
3543 uint32_t addpath_id)
3544 {
3545 struct prefix_rd prd;
3546 struct prefix_evpn p;
3547 struct bgp_route_evpn evpn;
3548 uint8_t ipaddr_len;
3549 uint8_t macaddr_len;
3550 mpls_label_t label[BGP_MAX_LABELS]; /* holds the VNI(s) as in packet */
3551 uint32_t num_labels = 0;
3552 uint32_t eth_tag;
3553 int ret;
3554
3555 /* Type-2 route should be either 33, 37 or 49 bytes or an
3556 * additional 3 bytes if there is a second label (VNI):
3557 * RD (8), ESI (10), Eth Tag (4), MAC Addr Len (1),
3558 * MAC Addr (6), IP len (1), IP (0, 4 or 16),
3559 * MPLS Lbl1 (3), MPLS Lbl2 (0 or 3)
3560 */
3561 if (psize != 33 && psize != 37 && psize != 49 && psize != 36
3562 && psize != 40 && psize != 52) {
3563 flog_err(EC_BGP_EVPN_ROUTE_INVALID,
3564 "%u:%s - Rx EVPN Type-2 NLRI with invalid length %d",
3565 peer->bgp->vrf_id, peer->host, psize);
3566 return -1;
3567 }
3568
3569 memset(&evpn, 0, sizeof(evpn));
3570
3571 /* Make prefix_rd */
3572 prd.family = AF_UNSPEC;
3573 prd.prefixlen = 64;
3574 memcpy(&prd.val, pfx, 8);
3575 pfx += 8;
3576
3577 /* Make EVPN prefix. */
3578 memset(&p, 0, sizeof(struct prefix_evpn));
3579 p.family = AF_EVPN;
3580 p.prefixlen = EVPN_ROUTE_PREFIXLEN;
3581 p.prefix.route_type = BGP_EVPN_MAC_IP_ROUTE;
3582
3583 /* Copy Ethernet Seg Identifier */
3584 memcpy(&evpn.eth_s_id.val, pfx, ESI_LEN);
3585 pfx += ESI_LEN;
3586
3587 /* Copy Ethernet Tag */
3588 memcpy(&eth_tag, pfx, 4);
3589 p.prefix.macip_addr.eth_tag = ntohl(eth_tag);
3590 pfx += 4;
3591
3592 /* Get the MAC Addr len */
3593 macaddr_len = *pfx++;
3594
3595 /* Get the MAC Addr */
3596 if (macaddr_len == (ETH_ALEN * 8)) {
3597 memcpy(&p.prefix.macip_addr.mac.octet, pfx, ETH_ALEN);
3598 pfx += ETH_ALEN;
3599 } else {
3600 flog_err(
3601 EC_BGP_EVPN_ROUTE_INVALID,
3602 "%u:%s - Rx EVPN Type-2 NLRI with unsupported MAC address length %d",
3603 peer->bgp->vrf_id, peer->host, macaddr_len);
3604 return -1;
3605 }
3606
3607
3608 /* Get the IP. */
3609 ipaddr_len = *pfx++;
3610 if (ipaddr_len != 0 && ipaddr_len != IPV4_MAX_BITLEN
3611 && ipaddr_len != IPV6_MAX_BITLEN) {
3612 flog_err(
3613 EC_BGP_EVPN_ROUTE_INVALID,
3614 "%u:%s - Rx EVPN Type-2 NLRI with unsupported IP address length %d",
3615 peer->bgp->vrf_id, peer->host, ipaddr_len);
3616 return -1;
3617 }
3618
3619 if (ipaddr_len) {
3620 ipaddr_len /= 8; /* Convert to bytes. */
3621 p.prefix.macip_addr.ip.ipa_type = (ipaddr_len == IPV4_MAX_BYTELEN)
3622 ? IPADDR_V4
3623 : IPADDR_V6;
3624 memcpy(&p.prefix.macip_addr.ip.ip.addr, pfx, ipaddr_len);
3625 }
3626 pfx += ipaddr_len;
3627
3628 /* Get the VNI(s). Stored as bytes here. */
3629 num_labels++;
3630 memset(label, 0, sizeof(label));
3631 memcpy(&label[0], pfx, BGP_LABEL_BYTES);
3632 pfx += BGP_LABEL_BYTES;
3633 psize -= (33 + ipaddr_len);
3634 /* Do we have a second VNI? */
3635 if (psize) {
3636 num_labels++;
3637 memcpy(&label[1], pfx, BGP_LABEL_BYTES);
3638 /*
3639 * If in future, we are required to access additional fields,
3640 * we MUST increment pfx by BGP_LABEL_BYTES in before reading
3641 * the next field
3642 */
3643 }
3644
3645 /* Process the route. */
3646 if (attr)
3647 ret = bgp_update(peer, (struct prefix *)&p, addpath_id, attr,
3648 afi, safi, ZEBRA_ROUTE_BGP, BGP_ROUTE_NORMAL,
3649 &prd, &label[0], num_labels, 0, &evpn);
3650 else
3651 ret = bgp_withdraw(peer, (struct prefix *)&p, addpath_id, attr,
3652 afi, safi, ZEBRA_ROUTE_BGP, BGP_ROUTE_NORMAL,
3653 &prd, &label[0], num_labels, &evpn);
3654 return ret;
3655 }
3656
3657 /*
3658 * Process received EVPN type-3 route (advertise or withdraw).
3659 */
3660 static int process_type3_route(struct peer *peer, afi_t afi, safi_t safi,
3661 struct attr *attr, uint8_t *pfx, int psize,
3662 uint32_t addpath_id)
3663 {
3664 struct prefix_rd prd;
3665 struct prefix_evpn p;
3666 uint8_t ipaddr_len;
3667 uint32_t eth_tag;
3668 int ret;
3669
3670 /* Type-3 route should be either 17 or 29 bytes: RD (8), Eth Tag (4),
3671 * IP len (1) and IP (4 or 16).
3672 */
3673 if (psize != 17 && psize != 29) {
3674 flog_err(EC_BGP_EVPN_ROUTE_INVALID,
3675 "%u:%s - Rx EVPN Type-3 NLRI with invalid length %d",
3676 peer->bgp->vrf_id, peer->host, psize);
3677 return -1;
3678 }
3679
3680 /* If PMSI is present, log if it is anything other than IR.
3681 * Note: We just simply ignore the values as it is not clear if
3682 * doing anything else is better.
3683 */
3684 if (attr &&
3685 (attr->flag & ATTR_FLAG_BIT(BGP_ATTR_PMSI_TUNNEL))) {
3686 if (attr->pmsi_tnl_type != PMSI_TNLTYPE_INGR_REPL) {
3687 flog_warn(
3688 EC_BGP_EVPN_PMSI_PRESENT,
3689 "%u:%s - Rx EVPN Type-3 NLRI with unsupported PTA %d",
3690 peer->bgp->vrf_id, peer->host,
3691 attr->pmsi_tnl_type);
3692 }
3693 }
3694
3695 /* Make prefix_rd */
3696 prd.family = AF_UNSPEC;
3697 prd.prefixlen = 64;
3698 memcpy(&prd.val, pfx, 8);
3699 pfx += 8;
3700
3701 /* Make EVPN prefix. */
3702 memset(&p, 0, sizeof(struct prefix_evpn));
3703 p.family = AF_EVPN;
3704 p.prefixlen = EVPN_ROUTE_PREFIXLEN;
3705 p.prefix.route_type = BGP_EVPN_IMET_ROUTE;
3706
3707 /* Copy Ethernet Tag */
3708 memcpy(&eth_tag, pfx, 4);
3709 p.prefix.imet_addr.eth_tag = ntohl(eth_tag);
3710 pfx += 4;
3711
3712 /* Get the IP. */
3713 ipaddr_len = *pfx++;
3714 if (ipaddr_len == IPV4_MAX_BITLEN) {
3715 p.prefix.imet_addr.ip.ipa_type = IPADDR_V4;
3716 memcpy(&p.prefix.imet_addr.ip.ip.addr, pfx, IPV4_MAX_BYTELEN);
3717 } else {
3718 flog_err(
3719 EC_BGP_EVPN_ROUTE_INVALID,
3720 "%u:%s - Rx EVPN Type-3 NLRI with unsupported IP address length %d",
3721 peer->bgp->vrf_id, peer->host, ipaddr_len);
3722 return -1;
3723 }
3724
3725 /* Process the route. */
3726 if (attr)
3727 ret = bgp_update(peer, (struct prefix *)&p, addpath_id, attr,
3728 afi, safi, ZEBRA_ROUTE_BGP, BGP_ROUTE_NORMAL,
3729 &prd, NULL, 0, 0, NULL);
3730 else
3731 ret = bgp_withdraw(peer, (struct prefix *)&p, addpath_id, attr,
3732 afi, safi, ZEBRA_ROUTE_BGP, BGP_ROUTE_NORMAL,
3733 &prd, NULL, 0, NULL);
3734 return ret;
3735 }
3736
3737 /*
3738 * Process received EVPN type-4 route (advertise or withdraw).
3739 */
3740 static int process_type4_route(struct peer *peer, afi_t afi, safi_t safi,
3741 struct attr *attr, uint8_t *pfx, int psize,
3742 uint32_t addpath_id)
3743 {
3744 int ret;
3745 esi_t esi;
3746 uint8_t ipaddr_len;
3747 struct in_addr vtep_ip;
3748 struct prefix_rd prd;
3749 struct prefix_evpn p;
3750
3751 /* Type-4 route should be either 23 or 35 bytes
3752 * RD (8), ESI (10), ip-len (1), ip (4 or 16)
3753 */
3754 if (psize != 23 && psize != 35) {
3755 flog_err(EC_BGP_EVPN_ROUTE_INVALID,
3756 "%u:%s - Rx EVPN Type-4 NLRI with invalid length %d",
3757 peer->bgp->vrf_id, peer->host, psize);
3758 return -1;
3759 }
3760
3761 /* Make prefix_rd */
3762 prd.family = AF_UNSPEC;
3763 prd.prefixlen = 64;
3764 memcpy(&prd.val, pfx, 8);
3765 pfx += 8;
3766
3767 /* get the ESI */
3768 memcpy(&esi, pfx, ESI_BYTES);
3769 pfx += ESI_BYTES;
3770
3771
3772 /* Get the IP. */
3773 ipaddr_len = *pfx++;
3774 if (ipaddr_len == IPV4_MAX_BITLEN) {
3775 memcpy(&vtep_ip, pfx, IPV4_MAX_BYTELEN);
3776 } else {
3777 flog_err(
3778 EC_BGP_EVPN_ROUTE_INVALID,
3779 "%u:%s - Rx EVPN Type-4 NLRI with unsupported IP address length %d",
3780 peer->bgp->vrf_id, peer->host, ipaddr_len);
3781 return -1;
3782 }
3783
3784 build_evpn_type4_prefix(&p, &esi, vtep_ip);
3785 /* Process the route. */
3786 if (attr) {
3787 ret = bgp_update(peer, (struct prefix *)&p, addpath_id, attr,
3788 afi, safi, ZEBRA_ROUTE_BGP, BGP_ROUTE_NORMAL,
3789 &prd, NULL, 0, 0, NULL);
3790 } else {
3791 ret = bgp_withdraw(peer, (struct prefix *)&p, addpath_id, attr,
3792 afi, safi, ZEBRA_ROUTE_BGP, BGP_ROUTE_NORMAL,
3793 &prd, NULL, 0, NULL);
3794 }
3795 return ret;
3796 }
3797
3798
3799 /*
3800 * Process received EVPN type-5 route (advertise or withdraw).
3801 */
3802 static int process_type5_route(struct peer *peer, afi_t afi, safi_t safi,
3803 struct attr *attr, uint8_t *pfx, int psize,
3804 uint32_t addpath_id, int withdraw)
3805 {
3806 struct prefix_rd prd;
3807 struct prefix_evpn p;
3808 struct bgp_route_evpn evpn;
3809 uint8_t ippfx_len;
3810 uint32_t eth_tag;
3811 mpls_label_t label; /* holds the VNI as in the packet */
3812 int ret;
3813
3814 /* Type-5 route should be 34 or 58 bytes:
3815 * RD (8), ESI (10), Eth Tag (4), IP len (1), IP (4 or 16),
3816 * GW (4 or 16) and VNI (3).
3817 * Note that the IP and GW should both be IPv4 or both IPv6.
3818 */
3819 if (psize != 34 && psize != 58) {
3820 flog_err(EC_BGP_EVPN_ROUTE_INVALID,
3821 "%u:%s - Rx EVPN Type-5 NLRI with invalid length %d",
3822 peer->bgp->vrf_id, peer->host, psize);
3823 return -1;
3824 }
3825
3826 /* Make prefix_rd */
3827 prd.family = AF_UNSPEC;
3828 prd.prefixlen = 64;
3829 memcpy(&prd.val, pfx, 8);
3830 pfx += 8;
3831
3832 /* Make EVPN prefix. */
3833 memset(&p, 0, sizeof(struct prefix_evpn));
3834 p.family = AF_EVPN;
3835 p.prefixlen = EVPN_ROUTE_PREFIXLEN;
3836 p.prefix.route_type = BGP_EVPN_IP_PREFIX_ROUTE;
3837
3838 /* Additional information outside of prefix - ESI and GW IP */
3839 memset(&evpn, 0, sizeof(evpn));
3840
3841 /* Fetch ESI */
3842 memcpy(&evpn.eth_s_id.val, pfx, 10);
3843 pfx += 10;
3844
3845 /* Fetch Ethernet Tag. */
3846 memcpy(&eth_tag, pfx, 4);
3847 p.prefix.prefix_addr.eth_tag = ntohl(eth_tag);
3848 pfx += 4;
3849
3850 /* Fetch IP prefix length. */
3851 ippfx_len = *pfx++;
3852 if (ippfx_len > IPV6_MAX_BITLEN) {
3853 flog_err(
3854 EC_BGP_EVPN_ROUTE_INVALID,
3855 "%u:%s - Rx EVPN Type-5 NLRI with invalid IP Prefix length %d",
3856 peer->bgp->vrf_id, peer->host, ippfx_len);
3857 return -1;
3858 }
3859 p.prefix.prefix_addr.ip_prefix_length = ippfx_len;
3860
3861 /* Determine IPv4 or IPv6 prefix */
3862 /* Since the address and GW are from the same family, this just becomes
3863 * a simple check on the total size.
3864 */
3865 if (psize == 34) {
3866 SET_IPADDR_V4(&p.prefix.prefix_addr.ip);
3867 memcpy(&p.prefix.prefix_addr.ip.ipaddr_v4, pfx, 4);
3868 pfx += 4;
3869 memcpy(&evpn.gw_ip.ipv4, pfx, 4);
3870 pfx += 4;
3871 } else {
3872 SET_IPADDR_V6(&p.prefix.prefix_addr.ip);
3873 memcpy(&p.prefix.prefix_addr.ip.ipaddr_v6, pfx, 16);
3874 pfx += 16;
3875 memcpy(&evpn.gw_ip.ipv6, pfx, 16);
3876 pfx += 16;
3877 }
3878
3879 /* Get the VNI (in MPLS label field). Stored as bytes here. */
3880 memset(&label, 0, sizeof(label));
3881 memcpy(&label, pfx, BGP_LABEL_BYTES);
3882
3883 /*
3884 * If in future, we are required to access additional fields,
3885 * we MUST increment pfx by BGP_LABEL_BYTES in before reading the next
3886 * field
3887 */
3888
3889 /* Process the route. */
3890 if (!withdraw)
3891 ret = bgp_update(peer, (struct prefix *)&p, addpath_id, attr,
3892 afi, safi, ZEBRA_ROUTE_BGP, BGP_ROUTE_NORMAL,
3893 &prd, &label, 1, 0, &evpn);
3894 else
3895 ret = bgp_withdraw(peer, (struct prefix *)&p, addpath_id, attr,
3896 afi, safi, ZEBRA_ROUTE_BGP, BGP_ROUTE_NORMAL,
3897 &prd, &label, 1, &evpn);
3898
3899 return ret;
3900 }
3901
3902 static void evpn_mpattr_encode_type5(struct stream *s, struct prefix *p,
3903 struct prefix_rd *prd, mpls_label_t *label,
3904 uint32_t num_labels, struct attr *attr)
3905 {
3906 int len;
3907 char temp[16];
3908 struct evpn_addr *p_evpn_p;
3909
3910 memset(&temp, 0, 16);
3911 if (p->family != AF_EVPN)
3912 return;
3913 p_evpn_p = &(p->u.prefix_evpn);
3914
3915 /* len denites the total len of IP and GW-IP in the route
3916 IP and GW-IP have to be both ipv4 or ipv6
3917 */
3918 if (IS_IPADDR_V4(&p_evpn_p->prefix_addr.ip))
3919 len = 8; /* IP and GWIP are both ipv4 */
3920 else
3921 len = 32; /* IP and GWIP are both ipv6 */
3922 /* Prefix contains RD, ESI, EthTag, IP length, IP, GWIP and VNI */
3923 stream_putc(s, 8 + 10 + 4 + 1 + len + 3);
3924 stream_put(s, prd->val, 8);
3925 if (attr)
3926 stream_put(s, &(attr->evpn_overlay.eth_s_id), 10);
3927 else
3928 stream_put(s, &temp, 10);
3929 stream_putl(s, p_evpn_p->prefix_addr.eth_tag);
3930 stream_putc(s, p_evpn_p->prefix_addr.ip_prefix_length);
3931 if (IS_IPADDR_V4(&p_evpn_p->prefix_addr.ip))
3932 stream_put_ipv4(s, p_evpn_p->prefix_addr.ip.ipaddr_v4.s_addr);
3933 else
3934 stream_put(s, &p_evpn_p->prefix_addr.ip.ipaddr_v6, 16);
3935 if (attr) {
3936 if (IS_IPADDR_V4(&p_evpn_p->prefix_addr.ip))
3937 stream_put_ipv4(s,
3938 attr->evpn_overlay.gw_ip.ipv4.s_addr);
3939 else
3940 stream_put(s, &(attr->evpn_overlay.gw_ip.ipv6), 16);
3941 } else {
3942 if (IS_IPADDR_V4(&p_evpn_p->prefix_addr.ip))
3943 stream_put_ipv4(s, 0);
3944 else
3945 stream_put(s, &temp, 16);
3946 }
3947
3948 if (num_labels)
3949 stream_put(s, label, 3);
3950 else
3951 stream_put3(s, 0);
3952 }
3953
3954 /*
3955 * Cleanup specific VNI upon EVPN (advertise-all-vni) being disabled.
3956 */
3957 static void cleanup_vni_on_disable(struct hash_backet *backet, struct bgp *bgp)
3958 {
3959 struct bgpevpn *vpn = (struct bgpevpn *)backet->data;
3960
3961 /* Remove EVPN routes and schedule for processing. */
3962 delete_routes_for_vni(bgp, vpn);
3963
3964 /* Clear "live" flag and see if hash needs to be freed. */
3965 UNSET_FLAG(vpn->flags, VNI_FLAG_LIVE);
3966 if (!is_vni_configured(vpn))
3967 bgp_evpn_free(bgp, vpn);
3968 }
3969
3970 /*
3971 * Free a VNI entry; iterator function called during cleanup.
3972 */
3973 static void free_vni_entry(struct hash_backet *backet, struct bgp *bgp)
3974 {
3975 struct bgpevpn *vpn = (struct bgpevpn *)backet->data;
3976
3977 delete_all_vni_routes(bgp, vpn);
3978 bgp_evpn_free(bgp, vpn);
3979 }
3980
3981 /*
3982 * Derive AUTO import RT for BGP VRF - L3VNI
3983 */
3984 static void evpn_auto_rt_import_add_for_vrf(struct bgp *bgp_vrf)
3985 {
3986 struct bgp *bgp_def = NULL;
3987
3988 form_auto_rt(bgp_vrf, bgp_vrf->l3vni, bgp_vrf->vrf_import_rtl);
3989 UNSET_FLAG(bgp_vrf->vrf_flags, BGP_VRF_IMPORT_RT_CFGD);
3990
3991 /* Map RT to VRF */
3992 bgp_def = bgp_get_default();
3993 if (!bgp_def)
3994 return;
3995 bgp_evpn_map_vrf_to_its_rts(bgp_vrf);
3996 }
3997
3998 /*
3999 * Delete AUTO import RT from BGP VRF - L3VNI
4000 */
4001 static void evpn_auto_rt_import_delete_for_vrf(struct bgp *bgp_vrf)
4002 {
4003 evpn_rt_delete_auto(bgp_vrf, bgp_vrf->l3vni, bgp_vrf->vrf_import_rtl);
4004 }
4005
4006 /*
4007 * Derive AUTO export RT for BGP VRF - L3VNI
4008 */
4009 static void evpn_auto_rt_export_add_for_vrf(struct bgp *bgp_vrf)
4010 {
4011 UNSET_FLAG(bgp_vrf->vrf_flags, BGP_VRF_EXPORT_RT_CFGD);
4012 form_auto_rt(bgp_vrf, bgp_vrf->l3vni, bgp_vrf->vrf_export_rtl);
4013 }
4014
4015 /*
4016 * Delete AUTO export RT from BGP VRF - L3VNI
4017 */
4018 static void evpn_auto_rt_export_delete_for_vrf(struct bgp *bgp_vrf)
4019 {
4020 evpn_rt_delete_auto(bgp_vrf, bgp_vrf->l3vni, bgp_vrf->vrf_export_rtl);
4021 }
4022
4023 static void bgp_evpn_handle_export_rt_change_for_vrf(struct bgp *bgp_vrf)
4024 {
4025 struct bgp *bgp_def = NULL;
4026 struct listnode *node = NULL;
4027 struct bgpevpn *vpn = NULL;
4028
4029 bgp_def = bgp_get_default();
4030 if (!bgp_def)
4031 return;
4032
4033 /* update all type-5 routes */
4034 update_advertise_vrf_routes(bgp_vrf);
4035
4036 /* update all type-2 routes */
4037 for (ALL_LIST_ELEMENTS_RO(bgp_vrf->l2vnis, node, vpn))
4038 update_routes_for_vni(bgp_def, vpn);
4039 }
4040
4041 /*
4042 * Handle autort change for a given VNI.
4043 */
4044 static void update_autort_vni(struct hash_backet *backet, struct bgp *bgp)
4045 {
4046 struct bgpevpn *vpn = backet->data;
4047
4048 if (!is_import_rt_configured(vpn)) {
4049 if (is_vni_live(vpn))
4050 bgp_evpn_uninstall_routes(bgp, vpn);
4051 bgp_evpn_unmap_vni_from_its_rts(bgp, vpn);
4052 list_delete_all_node(vpn->import_rtl);
4053 bgp_evpn_derive_auto_rt_import(bgp, vpn);
4054 if (is_vni_live(vpn))
4055 bgp_evpn_install_routes(bgp, vpn);
4056 }
4057 if (!is_export_rt_configured(vpn)) {
4058 list_delete_all_node(vpn->export_rtl);
4059 bgp_evpn_derive_auto_rt_export(bgp, vpn);
4060 if (is_vni_live(vpn))
4061 bgp_evpn_handle_export_rt_change(bgp, vpn);
4062 }
4063 }
4064
4065 /*
4066 * Public functions.
4067 */
4068
4069 /* withdraw type-5 route corresponding to ip prefix */
4070 void bgp_evpn_withdraw_type5_route(struct bgp *bgp_vrf, struct prefix *p,
4071 afi_t afi, safi_t safi)
4072 {
4073 int ret = 0;
4074 struct prefix_evpn evp;
4075 char buf[PREFIX_STRLEN];
4076
4077 build_type5_prefix_from_ip_prefix(&evp, p);
4078 ret = delete_evpn_type5_route(bgp_vrf, &evp);
4079 if (ret) {
4080 flog_err(
4081 EC_BGP_EVPN_ROUTE_DELETE,
4082 "%u failed to delete type-5 route for prefix %s in vrf %s",
4083 bgp_vrf->vrf_id, prefix2str(p, buf, sizeof(buf)),
4084 vrf_id_to_name(bgp_vrf->vrf_id));
4085 }
4086 }
4087
4088 /* withdraw all type-5 routes for an address family */
4089 void bgp_evpn_withdraw_type5_routes(struct bgp *bgp_vrf, afi_t afi, safi_t safi)
4090 {
4091 struct bgp_table *table = NULL;
4092 struct bgp_node *rn = NULL;
4093 struct bgp_info *ri;
4094
4095 table = bgp_vrf->rib[afi][safi];
4096 for (rn = bgp_table_top(table); rn; rn = bgp_route_next(rn)) {
4097 /* Only care about "selected" routes - non-imported. */
4098 /* TODO: Support for AddPath for EVPN. */
4099 for (ri = rn->info; ri; ri = ri->next) {
4100 if (CHECK_FLAG(ri->flags, BGP_INFO_SELECTED)
4101 && (!ri->extra || !ri->extra->parent)) {
4102 bgp_evpn_withdraw_type5_route(bgp_vrf, &rn->p,
4103 afi, safi);
4104 break;
4105 }
4106 }
4107 }
4108 }
4109
4110 /*
4111 * Advertise IP prefix as type-5 route. The afi/safi and src_attr passed
4112 * to this function correspond to those of the source IP prefix (best
4113 * path in the case of the attr. In the case of a local prefix (when we
4114 * are advertising local subnets), the src_attr will be NULL.
4115 */
4116 void bgp_evpn_advertise_type5_route(struct bgp *bgp_vrf, struct prefix *p,
4117 struct attr *src_attr, afi_t afi,
4118 safi_t safi)
4119 {
4120 int ret = 0;
4121 struct prefix_evpn evp;
4122 char buf[PREFIX_STRLEN];
4123
4124 build_type5_prefix_from_ip_prefix(&evp, p);
4125 ret = update_evpn_type5_route(bgp_vrf, &evp, src_attr);
4126 if (ret)
4127 flog_err(EC_BGP_EVPN_ROUTE_CREATE,
4128 "%u: Failed to create type-5 route for prefix %s",
4129 bgp_vrf->vrf_id, prefix2str(p, buf, sizeof(buf)));
4130 }
4131
4132 /* Inject all prefixes of a particular address-family (currently, IPv4 or
4133 * IPv6 unicast) into EVPN as type-5 routes. This is invoked when the
4134 * advertisement is enabled.
4135 */
4136 void bgp_evpn_advertise_type5_routes(struct bgp *bgp_vrf, afi_t afi,
4137 safi_t safi)
4138 {
4139 struct bgp_table *table = NULL;
4140 struct bgp_node *rn = NULL;
4141 struct bgp_info *ri;
4142
4143 table = bgp_vrf->rib[afi][safi];
4144 for (rn = bgp_table_top(table); rn; rn = bgp_route_next(rn)) {
4145 /* Need to identify the "selected" route entry to use its
4146 * attribute. Also, we only consider "non-imported" routes.
4147 * TODO: Support for AddPath for EVPN.
4148 */
4149 for (ri = rn->info; ri; ri = ri->next) {
4150 if (CHECK_FLAG(ri->flags, BGP_INFO_SELECTED)
4151 && (!ri->extra || !ri->extra->parent)) {
4152
4153 /* apply the route-map */
4154 if (bgp_vrf->adv_cmd_rmap[afi][safi].map) {
4155 int ret = 0;
4156
4157 ret = route_map_apply(
4158 bgp_vrf->adv_cmd_rmap[afi][safi]
4159 .map,
4160 &rn->p, RMAP_BGP, ri);
4161 if (ret == RMAP_DENYMATCH)
4162 continue;
4163 }
4164 bgp_evpn_advertise_type5_route(
4165 bgp_vrf, &rn->p, ri->attr, afi, safi);
4166 break;
4167 }
4168 }
4169 }
4170 }
4171
4172 void evpn_rt_delete_auto(struct bgp *bgp, vni_t vni, struct list *rtl)
4173 {
4174 struct listnode *node, *nnode, *node_to_del;
4175 struct ecommunity *ecom, *ecom_auto;
4176 struct ecommunity_val eval;
4177
4178 if (bgp->advertise_autort_rfc8365)
4179 vni |= EVPN_AUTORT_VXLAN;
4180 encode_route_target_as((bgp->as & 0xFFFF), vni, &eval);
4181
4182 ecom_auto = ecommunity_new();
4183 ecommunity_add_val(ecom_auto, &eval);
4184 node_to_del = NULL;
4185
4186 for (ALL_LIST_ELEMENTS(rtl, node, nnode, ecom)) {
4187 if (ecommunity_match(ecom, ecom_auto)) {
4188 ecommunity_free(&ecom);
4189 node_to_del = node;
4190 }
4191 }
4192
4193 if (node_to_del)
4194 list_delete_node(rtl, node_to_del);
4195
4196 ecommunity_free(&ecom_auto);
4197 }
4198
4199 void bgp_evpn_configure_import_rt_for_vrf(struct bgp *bgp_vrf,
4200 struct ecommunity *ecomadd)
4201 {
4202 /* uninstall routes from vrf */
4203 uninstall_routes_for_vrf(bgp_vrf);
4204
4205 /* Cleanup the RT to VRF mapping */
4206 bgp_evpn_unmap_vrf_from_its_rts(bgp_vrf);
4207
4208 /* Remove auto generated RT */
4209 evpn_auto_rt_import_delete_for_vrf(bgp_vrf);
4210
4211 /* Add the newly configured RT to RT list */
4212 listnode_add_sort(bgp_vrf->vrf_import_rtl, ecomadd);
4213 SET_FLAG(bgp_vrf->vrf_flags, BGP_VRF_IMPORT_RT_CFGD);
4214
4215 /* map VRF to its RTs */
4216 bgp_evpn_map_vrf_to_its_rts(bgp_vrf);
4217
4218 /* install routes matching the new VRF */
4219 install_routes_for_vrf(bgp_vrf);
4220 }
4221
4222 void bgp_evpn_unconfigure_import_rt_for_vrf(struct bgp *bgp_vrf,
4223 struct ecommunity *ecomdel)
4224 {
4225 struct listnode *node = NULL, *nnode = NULL, *node_to_del = NULL;
4226 struct ecommunity *ecom = NULL;
4227
4228 /* uninstall routes from vrf */
4229 uninstall_routes_for_vrf(bgp_vrf);
4230
4231 /* Cleanup the RT to VRF mapping */
4232 bgp_evpn_unmap_vrf_from_its_rts(bgp_vrf);
4233
4234 /* remove the RT from the RT list */
4235 for (ALL_LIST_ELEMENTS(bgp_vrf->vrf_import_rtl, node, nnode, ecom)) {
4236 if (ecommunity_match(ecom, ecomdel)) {
4237 ecommunity_free(&ecom);
4238 node_to_del = node;
4239 break;
4240 }
4241 }
4242
4243 if (node_to_del)
4244 list_delete_node(bgp_vrf->vrf_import_rtl, node_to_del);
4245
4246 assert(bgp_vrf->vrf_import_rtl);
4247 /* fallback to auto import rt, if this was the last RT */
4248 if (bgp_vrf->vrf_import_rtl && list_isempty(bgp_vrf->vrf_import_rtl)) {
4249 UNSET_FLAG(bgp_vrf->vrf_flags, BGP_VRF_IMPORT_RT_CFGD);
4250 evpn_auto_rt_import_add_for_vrf(bgp_vrf);
4251 }
4252
4253 /* map VRFs to its RTs */
4254 bgp_evpn_map_vrf_to_its_rts(bgp_vrf);
4255
4256 /* install routes matching this new RT */
4257 install_routes_for_vrf(bgp_vrf);
4258 }
4259
4260 void bgp_evpn_configure_export_rt_for_vrf(struct bgp *bgp_vrf,
4261 struct ecommunity *ecomadd)
4262 {
4263 /* remove auto-generated RT */
4264 evpn_auto_rt_export_delete_for_vrf(bgp_vrf);
4265
4266 /* Add the new RT to the RT list */
4267 listnode_add_sort(bgp_vrf->vrf_export_rtl, ecomadd);
4268 SET_FLAG(bgp_vrf->vrf_flags, BGP_VRF_EXPORT_RT_CFGD);
4269
4270 bgp_evpn_handle_export_rt_change_for_vrf(bgp_vrf);
4271 }
4272
4273 void bgp_evpn_unconfigure_export_rt_for_vrf(struct bgp *bgp_vrf,
4274 struct ecommunity *ecomdel)
4275 {
4276 struct listnode *node = NULL, *nnode = NULL, *node_to_del = NULL;
4277 struct ecommunity *ecom = NULL;
4278
4279 /* Remove the RT from the RT list */
4280 for (ALL_LIST_ELEMENTS(bgp_vrf->vrf_export_rtl, node, nnode, ecom)) {
4281 if (ecommunity_match(ecom, ecomdel)) {
4282 ecommunity_free(&ecom);
4283 node_to_del = node;
4284 break;
4285 }
4286 }
4287
4288 if (node_to_del)
4289 list_delete_node(bgp_vrf->vrf_export_rtl, node_to_del);
4290
4291 /*
4292 * Temporary assert to make SA happy.
4293 * The ALL_LIST_ELEMENTS macro above has a NULL check
4294 * which means that SA is going to complain about
4295 * the list_isempty call, which doesn't NULL check.
4296 * So until we get this situation cleaned up, here
4297 * we are.
4298 */
4299 assert(bgp_vrf->vrf_export_rtl);
4300
4301 /* fall back to auto-generated RT if this was the last RT */
4302 if (list_isempty(bgp_vrf->vrf_export_rtl)) {
4303 UNSET_FLAG(bgp_vrf->vrf_flags, BGP_VRF_EXPORT_RT_CFGD);
4304 evpn_auto_rt_export_add_for_vrf(bgp_vrf);
4305 }
4306
4307 bgp_evpn_handle_export_rt_change_for_vrf(bgp_vrf);
4308 }
4309
4310 /*
4311 * Handle change to BGP router id. This is invoked twice by the change
4312 * handler, first before the router id has been changed and then after
4313 * the router id has been changed. The first invocation will result in
4314 * local routes for all VNIs/VRF being deleted and withdrawn and the next
4315 * will result in the routes being re-advertised.
4316 */
4317 void bgp_evpn_handle_router_id_update(struct bgp *bgp, int withdraw)
4318 {
4319 if (withdraw) {
4320
4321 /* delete and withdraw all the type-5 routes
4322 stored in the global table for this vrf
4323 */
4324 withdraw_router_id_vrf(bgp);
4325
4326 /* delete all the VNI routes (type-2/type-3) routes for all the
4327 * L2-VNIs
4328 */
4329 hash_iterate(bgp->vnihash,
4330 (void (*)(struct hash_backet *,
4331 void *))withdraw_router_id_vni,
4332 bgp);
4333 } else {
4334
4335 /* advertise all routes in the vrf as type-5 routes with the new
4336 * RD
4337 */
4338 update_router_id_vrf(bgp);
4339
4340 /* advertise all the VNI routes (type-2/type-3) routes with the
4341 * new RD
4342 */
4343 hash_iterate(bgp->vnihash,
4344 (void (*)(struct hash_backet *,
4345 void *))update_router_id_vni,
4346 bgp);
4347 }
4348 }
4349
4350 /*
4351 * Handle change to auto-RT algorithm - update and advertise local routes.
4352 */
4353 void bgp_evpn_handle_autort_change(struct bgp *bgp)
4354 {
4355 hash_iterate(bgp->vnihash,
4356 (void (*)(struct hash_backet *,
4357 void*))update_autort_vni,
4358 bgp);
4359 }
4360
4361 /*
4362 * Handle change to export RT - update and advertise local routes.
4363 */
4364 int bgp_evpn_handle_export_rt_change(struct bgp *bgp, struct bgpevpn *vpn)
4365 {
4366 return update_routes_for_vni(bgp, vpn);
4367 }
4368
4369 void bgp_evpn_handle_vrf_rd_change(struct bgp *bgp_vrf, int withdraw)
4370 {
4371 if (withdraw)
4372 delete_withdraw_vrf_routes(bgp_vrf);
4373 else
4374 update_advertise_vrf_routes(bgp_vrf);
4375 }
4376
4377 /*
4378 * Handle change to RD. This is invoked twice by the change handler,
4379 * first before the RD has been changed and then after the RD has
4380 * been changed. The first invocation will result in local routes
4381 * of this VNI being deleted and withdrawn and the next will result
4382 * in the routes being re-advertised.
4383 */
4384 void bgp_evpn_handle_rd_change(struct bgp *bgp, struct bgpevpn *vpn,
4385 int withdraw)
4386 {
4387 if (withdraw)
4388 delete_withdraw_vni_routes(bgp, vpn);
4389 else
4390 update_advertise_vni_routes(bgp, vpn);
4391 }
4392
4393 /*
4394 * Install routes for this VNI. Invoked upon change to Import RT.
4395 */
4396 int bgp_evpn_install_routes(struct bgp *bgp, struct bgpevpn *vpn)
4397 {
4398 return install_routes_for_vni(bgp, vpn);
4399 }
4400
4401 /*
4402 * Uninstall all routes installed for this VNI. Invoked upon change
4403 * to Import RT.
4404 */
4405 int bgp_evpn_uninstall_routes(struct bgp *bgp, struct bgpevpn *vpn)
4406 {
4407 return uninstall_routes_for_vni(bgp, vpn);
4408 }
4409
4410 /*
4411 * TODO: Hardcoded for a maximum of 2 VNIs right now
4412 */
4413 char *bgp_evpn_label2str(mpls_label_t *label, uint32_t num_labels, char *buf,
4414 int len)
4415 {
4416 vni_t vni1, vni2;
4417
4418 vni1 = label2vni(label);
4419 if (num_labels == 2) {
4420 vni2 = label2vni(label + 1);
4421 snprintf(buf, len, "%u/%u", vni1, vni2);
4422 } else
4423 snprintf(buf, len, "%u", vni1);
4424 return buf;
4425 }
4426
4427 /*
4428 * Function to convert evpn route to json format.
4429 * NOTE: We don't use prefix2str as the output here is a bit different.
4430 */
4431 void bgp_evpn_route2json(struct prefix_evpn *p, json_object *json)
4432 {
4433 char buf1[ETHER_ADDR_STRLEN];
4434 char buf2[PREFIX2STR_BUFFER];
4435
4436 if (!json)
4437 return;
4438
4439 if (p->prefix.route_type == BGP_EVPN_IMET_ROUTE) {
4440 json_object_int_add(json, "routeType", p->prefix.route_type);
4441 json_object_int_add(json, "ethTag",
4442 p->prefix.imet_addr.eth_tag);
4443 json_object_int_add(json, "ipLen",
4444 is_evpn_prefix_ipaddr_v4(p)
4445 ? IPV4_MAX_BITLEN
4446 : IPV6_MAX_BITLEN);
4447 json_object_string_add(json, "ip",
4448 inet_ntoa(p->prefix.imet_addr.ip.ipaddr_v4));
4449 } else if (p->prefix.route_type == BGP_EVPN_MAC_IP_ROUTE) {
4450 if (is_evpn_prefix_ipaddr_none(p)) {
4451 json_object_int_add(json, "routeType",
4452 p->prefix.route_type);
4453 json_object_int_add(json, "ethTag",
4454 p->prefix.macip_addr.eth_tag);
4455 json_object_int_add(json, "macLen", 8 * ETH_ALEN);
4456 json_object_string_add(json, "mac",
4457 prefix_mac2str(&p->prefix.macip_addr.mac,
4458 buf1,
4459 sizeof(buf1)));
4460 } else {
4461 uint8_t family;
4462
4463 family = is_evpn_prefix_ipaddr_v4(p) ? AF_INET
4464 : AF_INET6;
4465
4466 json_object_int_add(json, "routeType",
4467 p->prefix.route_type);
4468 json_object_int_add(json, "ethTag",
4469 p->prefix.macip_addr.eth_tag);
4470 json_object_int_add(json, "macLen", 8 * ETH_ALEN);
4471 json_object_string_add(json, "mac",
4472 prefix_mac2str(&p->prefix.macip_addr.mac,
4473 buf1,
4474 sizeof(buf1)));
4475 json_object_int_add(json, "ipLen",
4476 is_evpn_prefix_ipaddr_v4(p)
4477 ? IPV4_MAX_BITLEN
4478 : IPV6_MAX_BITLEN);
4479 json_object_string_add(
4480 json, "ip",
4481 inet_ntop(family,
4482 &p->prefix.macip_addr.ip.ip.addr,
4483 buf2,
4484 PREFIX2STR_BUFFER));
4485 }
4486 } else {
4487 /* Currently, this is to cater to other AF_ETHERNET code. */
4488 }
4489 }
4490
4491 /*
4492 * Function to convert evpn route to string.
4493 * NOTE: We don't use prefix2str as the output here is a bit different.
4494 */
4495 char *bgp_evpn_route2str(struct prefix_evpn *p, char *buf, int len)
4496 {
4497 char buf1[ETHER_ADDR_STRLEN];
4498 char buf2[PREFIX2STR_BUFFER];
4499 char buf3[ESI_STR_LEN];
4500
4501 if (p->prefix.route_type == BGP_EVPN_IMET_ROUTE) {
4502 snprintf(buf, len, "[%d]:[%d]:[%d]:[%s]", p->prefix.route_type,
4503 p->prefix.imet_addr.eth_tag,
4504 is_evpn_prefix_ipaddr_v4(p) ? IPV4_MAX_BITLEN
4505 : IPV6_MAX_BITLEN,
4506 inet_ntoa(p->prefix.imet_addr.ip.ipaddr_v4));
4507 } else if (p->prefix.route_type == BGP_EVPN_MAC_IP_ROUTE) {
4508 if (is_evpn_prefix_ipaddr_none(p))
4509 snprintf(buf, len, "[%d]:[%d]:[%d]:[%s]",
4510 p->prefix.route_type,
4511 p->prefix.macip_addr.eth_tag,
4512 8 * ETH_ALEN,
4513 prefix_mac2str(&p->prefix.macip_addr.mac, buf1,
4514 sizeof(buf1)));
4515 else {
4516 uint8_t family;
4517
4518 family = is_evpn_prefix_ipaddr_v4(p) ? AF_INET
4519 : AF_INET6;
4520 snprintf(buf, len, "[%d]:[%d]:[%d]:[%s]:[%d]:[%s]",
4521 p->prefix.route_type,
4522 p->prefix.macip_addr.eth_tag,
4523 8 * ETH_ALEN,
4524 prefix_mac2str(&p->prefix.macip_addr.mac, buf1,
4525 sizeof(buf1)),
4526 family == AF_INET ? IPV4_MAX_BITLEN
4527 : IPV6_MAX_BITLEN,
4528 inet_ntop(family,
4529 &p->prefix.macip_addr.ip.ip.addr,
4530 buf2,
4531 PREFIX2STR_BUFFER));
4532 }
4533 } else if (p->prefix.route_type == BGP_EVPN_IP_PREFIX_ROUTE) {
4534 snprintf(buf, len, "[%d]:[%d]:[%d]:[%s]",
4535 p->prefix.route_type,
4536 p->prefix.prefix_addr.eth_tag,
4537 p->prefix.prefix_addr.ip_prefix_length,
4538 is_evpn_prefix_ipaddr_v4(p)
4539 ? inet_ntoa(p->prefix.prefix_addr.ip.ipaddr_v4)
4540 : inet6_ntoa(p->prefix.prefix_addr.ip.ipaddr_v6));
4541 } else if (p->prefix.route_type == BGP_EVPN_ES_ROUTE) {
4542 snprintf(buf, len, "[%d]:[%s]:[%d]:[%s]",
4543 p->prefix.route_type,
4544 esi_to_str(&p->prefix.es_addr.esi, buf3, sizeof(buf3)),
4545 is_evpn_prefix_ipaddr_v4(p) ? IPV4_MAX_BITLEN
4546 : IPV6_MAX_BITLEN,
4547 inet_ntoa(p->prefix.es_addr.ip.ipaddr_v4));
4548 } else {
4549 /* For EVPN route types not supported yet. */
4550 snprintf(buf, len, "(unsupported route type %d)",
4551 p->prefix.route_type);
4552 }
4553
4554 return (buf);
4555 }
4556
4557 /*
4558 * Encode EVPN prefix in Update (MP_REACH)
4559 */
4560 void bgp_evpn_encode_prefix(struct stream *s, struct prefix *p,
4561 struct prefix_rd *prd, mpls_label_t *label,
4562 uint32_t num_labels, struct attr *attr,
4563 int addpath_encode, uint32_t addpath_tx_id)
4564 {
4565 struct prefix_evpn *evp = (struct prefix_evpn *)p;
4566 int len, ipa_len = 0;
4567
4568 if (addpath_encode)
4569 stream_putl(s, addpath_tx_id);
4570
4571 /* Route type */
4572 stream_putc(s, evp->prefix.route_type);
4573
4574 switch (evp->prefix.route_type) {
4575 case BGP_EVPN_MAC_IP_ROUTE:
4576 if (is_evpn_prefix_ipaddr_v4(evp))
4577 ipa_len = IPV4_MAX_BYTELEN;
4578 else if (is_evpn_prefix_ipaddr_v6(evp))
4579 ipa_len = IPV6_MAX_BYTELEN;
4580 /* RD, ESI, EthTag, MAC+len, IP len, [IP], 1 VNI */
4581 len = 8 + 10 + 4 + 1 + 6 + 1 + ipa_len + 3;
4582 if (ipa_len && num_labels > 1) /* There are 2 VNIs */
4583 len += 3;
4584 stream_putc(s, len);
4585 stream_put(s, prd->val, 8); /* RD */
4586 if (attr)
4587 stream_put(s, &attr->evpn_overlay.eth_s_id, ESI_LEN);
4588 else
4589 stream_put(s, 0, 10);
4590 stream_putl(s, evp->prefix.macip_addr.eth_tag); /* Ethernet Tag ID */
4591 stream_putc(s, 8 * ETH_ALEN); /* Mac Addr Len - bits */
4592 stream_put(s, evp->prefix.macip_addr.mac.octet, 6); /* Mac Addr */
4593 stream_putc(s, 8 * ipa_len); /* IP address Length */
4594 if (ipa_len) /* IP */
4595 stream_put(s, &evp->prefix.macip_addr.ip.ip.addr,
4596 ipa_len);
4597 /* 1st label is the L2 VNI */
4598 stream_put(s, label, BGP_LABEL_BYTES);
4599 /* Include 2nd label (L3 VNI) if advertising MAC+IP */
4600 if (ipa_len && num_labels > 1)
4601 stream_put(s, label + 1, BGP_LABEL_BYTES);
4602 break;
4603
4604 case BGP_EVPN_IMET_ROUTE:
4605 stream_putc(s, 17); // TODO: length - assumes IPv4 address
4606 stream_put(s, prd->val, 8); /* RD */
4607 stream_putl(s, evp->prefix.imet_addr.eth_tag); /* Ethernet Tag ID */
4608 stream_putc(s, IPV4_MAX_BITLEN); /* IP address Length - bits */
4609 /* Originating Router's IP Addr */
4610 stream_put_in_addr(s, &evp->prefix.imet_addr.ip.ipaddr_v4);
4611 break;
4612
4613 case BGP_EVPN_ES_ROUTE:
4614 stream_putc(s, 23); /* TODO: length: assumes ipv4 VTEP */
4615 stream_put(s, prd->val, 8); /* RD */
4616 stream_put(s, evp->prefix.es_addr.esi.val, 10); /* ESI */
4617 stream_putc(s, IPV4_MAX_BITLEN); /* IP address Length - bits */
4618 /* VTEP IP */
4619 stream_put_in_addr(s, &evp->prefix.es_addr.ip.ipaddr_v4);
4620 break;
4621
4622 case BGP_EVPN_IP_PREFIX_ROUTE:
4623 /* TODO: AddPath support. */
4624 evpn_mpattr_encode_type5(s, p, prd, label, num_labels, attr);
4625 break;
4626
4627 default:
4628 break;
4629 }
4630 }
4631
4632 int bgp_nlri_parse_evpn(struct peer *peer, struct attr *attr,
4633 struct bgp_nlri *packet, int withdraw)
4634 {
4635 uint8_t *pnt;
4636 uint8_t *lim;
4637 afi_t afi;
4638 safi_t safi;
4639 uint32_t addpath_id;
4640 int addpath_encoded;
4641 int psize = 0;
4642 uint8_t rtype;
4643 struct prefix p;
4644
4645 /* Start processing the NLRI - there may be multiple in the MP_REACH */
4646 pnt = packet->nlri;
4647 lim = pnt + packet->length;
4648 afi = packet->afi;
4649 safi = packet->safi;
4650 addpath_id = 0;
4651
4652 addpath_encoded =
4653 (CHECK_FLAG(peer->af_cap[afi][safi], PEER_CAP_ADDPATH_AF_RX_ADV)
4654 && CHECK_FLAG(peer->af_cap[afi][safi],
4655 PEER_CAP_ADDPATH_AF_TX_RCV));
4656
4657 for (; pnt < lim; pnt += psize) {
4658 /* Clear prefix structure. */
4659 memset(&p, 0, sizeof(struct prefix));
4660
4661 /* Deal with path-id if AddPath is supported. */
4662 if (addpath_encoded) {
4663 /* When packet overflow occurs return immediately. */
4664 if (pnt + BGP_ADDPATH_ID_LEN > lim)
4665 return -1;
4666
4667 addpath_id = ntohl(*((uint32_t *)pnt));
4668 pnt += BGP_ADDPATH_ID_LEN;
4669 }
4670
4671 /* All EVPN NLRI types start with type and length. */
4672 if (pnt + 2 > lim)
4673 return -1;
4674
4675 rtype = *pnt++;
4676 psize = *pnt++;
4677
4678 /* When packet overflow occur return immediately. */
4679 if (pnt + psize > lim)
4680 return -1;
4681
4682 switch (rtype) {
4683 case BGP_EVPN_MAC_IP_ROUTE:
4684 if (process_type2_route(peer, afi, safi,
4685 withdraw ? NULL : attr, pnt,
4686 psize, addpath_id)) {
4687 flog_err(
4688 EC_BGP_EVPN_FAIL,
4689 "%u:%s - Error in processing EVPN type-2 NLRI size %d",
4690 peer->bgp->vrf_id, peer->host, psize);
4691 return -1;
4692 }
4693 break;
4694
4695 case BGP_EVPN_IMET_ROUTE:
4696 if (process_type3_route(peer, afi, safi,
4697 withdraw ? NULL : attr, pnt,
4698 psize, addpath_id)) {
4699 flog_err(
4700 EC_BGP_PKT_PROCESS,
4701 "%u:%s - Error in processing EVPN type-3 NLRI size %d",
4702 peer->bgp->vrf_id, peer->host, psize);
4703 return -1;
4704 }
4705 break;
4706
4707 case BGP_EVPN_ES_ROUTE:
4708 if (process_type4_route(peer, afi, safi,
4709 withdraw ? NULL : attr, pnt,
4710 psize, addpath_id)) {
4711 flog_err(
4712 EC_BGP_PKT_PROCESS,
4713 "%u:%s - Error in processing EVPN type-4 NLRI size %d",
4714 peer->bgp->vrf_id, peer->host, psize);
4715 return -1;
4716 }
4717 break;
4718
4719 case BGP_EVPN_IP_PREFIX_ROUTE:
4720 if (process_type5_route(peer, afi, safi, attr, pnt,
4721 psize, addpath_id, withdraw)) {
4722 flog_err(
4723 EC_BGP_PKT_PROCESS,
4724 "%u:%s - Error in processing EVPN type-5 NLRI size %d",
4725 peer->bgp->vrf_id, peer->host, psize);
4726 return -1;
4727 }
4728 break;
4729
4730 default:
4731 break;
4732 }
4733 }
4734
4735 /* Packet length consistency check. */
4736 if (pnt != lim)
4737 return -1;
4738
4739 return 0;
4740 }
4741
4742 /*
4743 * Map the RTs (configured or automatically derived) of a VRF to the VRF.
4744 * The mapping will be used during route processing.
4745 * bgp_def: default bgp instance
4746 * bgp_vrf: specific bgp vrf instance on which RT is configured
4747 */
4748 void bgp_evpn_map_vrf_to_its_rts(struct bgp *bgp_vrf)
4749 {
4750 int i = 0;
4751 struct ecommunity_val *eval = NULL;
4752 struct listnode *node = NULL, *nnode = NULL;
4753 struct ecommunity *ecom = NULL;
4754
4755 for (ALL_LIST_ELEMENTS(bgp_vrf->vrf_import_rtl, node, nnode, ecom)) {
4756 for (i = 0; i < ecom->size; i++) {
4757 eval = (struct ecommunity_val *)(ecom->val
4758 + (i
4759 * ECOMMUNITY_SIZE));
4760 map_vrf_to_rt(bgp_vrf, eval);
4761 }
4762 }
4763 }
4764
4765 /*
4766 * Unmap the RTs (configured or automatically derived) of a VRF from the VRF.
4767 */
4768 void bgp_evpn_unmap_vrf_from_its_rts(struct bgp *bgp_vrf)
4769 {
4770 int i;
4771 struct ecommunity_val *eval;
4772 struct listnode *node, *nnode;
4773 struct ecommunity *ecom;
4774
4775 for (ALL_LIST_ELEMENTS(bgp_vrf->vrf_import_rtl, node, nnode, ecom)) {
4776 for (i = 0; i < ecom->size; i++) {
4777 struct vrf_irt_node *irt;
4778 struct ecommunity_val eval_tmp;
4779
4780 eval = (struct ecommunity_val *)(ecom->val
4781 + (i
4782 * ECOMMUNITY_SIZE));
4783 /* If using "automatic" RT, we only care about the
4784 * local-admin sub-field.
4785 * This is to facilitate using VNI as the RT for EBGP
4786 * peering too.
4787 */
4788 memcpy(&eval_tmp, eval, ECOMMUNITY_SIZE);
4789 if (!CHECK_FLAG(bgp_vrf->vrf_flags,
4790 BGP_VRF_IMPORT_RT_CFGD))
4791 mask_ecom_global_admin(&eval_tmp, eval);
4792
4793 irt = lookup_vrf_import_rt(&eval_tmp);
4794 if (irt)
4795 unmap_vrf_from_rt(bgp_vrf, irt);
4796 }
4797 }
4798 }
4799
4800
4801 /*
4802 * Map the RTs (configured or automatically derived) of a VNI to the VNI.
4803 * The mapping will be used during route processing.
4804 */
4805 void bgp_evpn_map_vni_to_its_rts(struct bgp *bgp, struct bgpevpn *vpn)
4806 {
4807 int i;
4808 struct ecommunity_val *eval;
4809 struct listnode *node, *nnode;
4810 struct ecommunity *ecom;
4811
4812 for (ALL_LIST_ELEMENTS(vpn->import_rtl, node, nnode, ecom)) {
4813 for (i = 0; i < ecom->size; i++) {
4814 eval = (struct ecommunity_val *)(ecom->val
4815 + (i
4816 * ECOMMUNITY_SIZE));
4817 map_vni_to_rt(bgp, vpn, eval);
4818 }
4819 }
4820 }
4821
4822 /*
4823 * Unmap the RTs (configured or automatically derived) of a VNI from the VNI.
4824 */
4825 void bgp_evpn_unmap_vni_from_its_rts(struct bgp *bgp, struct bgpevpn *vpn)
4826 {
4827 int i;
4828 struct ecommunity_val *eval;
4829 struct listnode *node, *nnode;
4830 struct ecommunity *ecom;
4831
4832 for (ALL_LIST_ELEMENTS(vpn->import_rtl, node, nnode, ecom)) {
4833 for (i = 0; i < ecom->size; i++) {
4834 struct irt_node *irt;
4835 struct ecommunity_val eval_tmp;
4836
4837 eval = (struct ecommunity_val *)(ecom->val
4838 + (i
4839 * ECOMMUNITY_SIZE));
4840 /* If using "automatic" RT, we only care about the
4841 * local-admin sub-field.
4842 * This is to facilitate using VNI as the RT for EBGP
4843 * peering too.
4844 */
4845 memcpy(&eval_tmp, eval, ECOMMUNITY_SIZE);
4846 if (!is_import_rt_configured(vpn))
4847 mask_ecom_global_admin(&eval_tmp, eval);
4848
4849 irt = lookup_import_rt(bgp, &eval_tmp);
4850 if (irt)
4851 unmap_vni_from_rt(bgp, vpn, irt);
4852 }
4853 }
4854 }
4855
4856 /*
4857 * Derive Import RT automatically for VNI and map VNI to RT.
4858 * The mapping will be used during route processing.
4859 */
4860 void bgp_evpn_derive_auto_rt_import(struct bgp *bgp, struct bgpevpn *vpn)
4861 {
4862 form_auto_rt(bgp, vpn->vni, vpn->import_rtl);
4863 UNSET_FLAG(vpn->flags, VNI_FLAG_IMPRT_CFGD);
4864
4865 /* Map RT to VNI */
4866 bgp_evpn_map_vni_to_its_rts(bgp, vpn);
4867 }
4868
4869 /*
4870 * Derive Export RT automatically for VNI.
4871 */
4872 void bgp_evpn_derive_auto_rt_export(struct bgp *bgp, struct bgpevpn *vpn)
4873 {
4874 form_auto_rt(bgp, vpn->vni, vpn->export_rtl);
4875 UNSET_FLAG(vpn->flags, VNI_FLAG_EXPRT_CFGD);
4876 }
4877
4878 /*
4879 * Derive RD automatically for VNI using passed information - it
4880 * is of the form RouterId:unique-id-for-vni.
4881 */
4882 void bgp_evpn_derive_auto_rd_for_vrf(struct bgp *bgp)
4883 {
4884 form_auto_rd(bgp->router_id, bgp->vrf_rd_id, &bgp->vrf_prd);
4885 }
4886
4887 /*
4888 * Derive RD automatically for VNI using passed information - it
4889 * is of the form RouterId:unique-id-for-vni.
4890 */
4891 void bgp_evpn_derive_auto_rd(struct bgp *bgp, struct bgpevpn *vpn)
4892 {
4893 char buf[100];
4894
4895 vpn->prd.family = AF_UNSPEC;
4896 vpn->prd.prefixlen = 64;
4897 sprintf(buf, "%s:%hu", inet_ntoa(bgp->router_id), vpn->rd_id);
4898 (void)str2prefix_rd(buf, &vpn->prd);
4899 UNSET_FLAG(vpn->flags, VNI_FLAG_RD_CFGD);
4900 }
4901
4902 /*
4903 * Lookup L3-VNI
4904 */
4905 bool bgp_evpn_lookup_l3vni_l2vni_table(vni_t vni)
4906 {
4907 struct list *inst = bm->bgp;
4908 struct listnode *node;
4909 struct bgp *bgp_vrf;
4910
4911 for (ALL_LIST_ELEMENTS_RO(inst, node, bgp_vrf)) {
4912 if (bgp_vrf->l3vni == vni)
4913 return true;
4914 }
4915
4916 return false;
4917 }
4918
4919 /*
4920 * Lookup VNI.
4921 */
4922 struct bgpevpn *bgp_evpn_lookup_vni(struct bgp *bgp, vni_t vni)
4923 {
4924 struct bgpevpn *vpn;
4925 struct bgpevpn tmp;
4926
4927 memset(&tmp, 0, sizeof(struct bgpevpn));
4928 tmp.vni = vni;
4929 vpn = hash_lookup(bgp->vnihash, &tmp);
4930 return vpn;
4931 }
4932
4933 /*
4934 * Create a new vpn - invoked upon configuration or zebra notification.
4935 */
4936 struct bgpevpn *bgp_evpn_new(struct bgp *bgp, vni_t vni,
4937 struct in_addr originator_ip,
4938 vrf_id_t tenant_vrf_id)
4939 {
4940 struct bgpevpn *vpn;
4941
4942 if (!bgp)
4943 return NULL;
4944
4945 vpn = XCALLOC(MTYPE_BGP_EVPN, sizeof(struct bgpevpn));
4946 if (!vpn)
4947 return NULL;
4948
4949 /* Set values - RD and RT set to defaults. */
4950 vpn->vni = vni;
4951 vpn->originator_ip = originator_ip;
4952 vpn->tenant_vrf_id = tenant_vrf_id;
4953
4954 /* Initialize route-target import and export lists */
4955 vpn->import_rtl = list_new();
4956 vpn->import_rtl->cmp = (int (*)(void *, void *))evpn_route_target_cmp;
4957 vpn->import_rtl->del = evpn_xxport_delete_ecomm;
4958 vpn->export_rtl = list_new();
4959 vpn->export_rtl->cmp = (int (*)(void *, void *))evpn_route_target_cmp;
4960 vpn->export_rtl->del = evpn_xxport_delete_ecomm;
4961 bf_assign_index(bm->rd_idspace, vpn->rd_id);
4962 derive_rd_rt_for_vni(bgp, vpn);
4963
4964 /* Initialize EVPN route table. */
4965 vpn->route_table = bgp_table_init(bgp, AFI_L2VPN, SAFI_EVPN);
4966
4967 /* Add to hash */
4968 if (!hash_get(bgp->vnihash, vpn, hash_alloc_intern)) {
4969 XFREE(MTYPE_BGP_EVPN, vpn);
4970 return NULL;
4971 }
4972
4973 /* add to l2vni list on corresponding vrf */
4974 bgpevpn_link_to_l3vni(vpn);
4975
4976 QOBJ_REG(vpn, bgpevpn);
4977 return vpn;
4978 }
4979
4980 /*
4981 * Free a given VPN - called in multiple scenarios such as zebra
4982 * notification, configuration being deleted, advertise-all-vni disabled etc.
4983 * This just frees appropriate memory, caller should have taken other
4984 * needed actions.
4985 */
4986 void bgp_evpn_free(struct bgp *bgp, struct bgpevpn *vpn)
4987 {
4988 bgpevpn_unlink_from_l3vni(vpn);
4989 bgp_table_unlock(vpn->route_table);
4990 bgp_evpn_unmap_vni_from_its_rts(bgp, vpn);
4991 list_delete_and_null(&vpn->import_rtl);
4992 list_delete_and_null(&vpn->export_rtl);
4993 bf_release_index(bm->rd_idspace, vpn->rd_id);
4994 hash_release(bgp->vnihash, vpn);
4995 QOBJ_UNREG(vpn);
4996 XFREE(MTYPE_BGP_EVPN, vpn);
4997 }
4998
4999 /*
5000 * Lookup local ES.
5001 */
5002 struct evpnes *bgp_evpn_lookup_es(struct bgp *bgp, esi_t *esi)
5003 {
5004 struct evpnes *es;
5005 struct evpnes tmp;
5006
5007 memset(&tmp, 0, sizeof(struct evpnes));
5008 memcpy(&tmp.esi, esi, sizeof(esi_t));
5009 es = hash_lookup(bgp->esihash, &tmp);
5010 return es;
5011 }
5012
5013 /*
5014 * Create a new local es - invoked upon zebra notification.
5015 */
5016 struct evpnes *bgp_evpn_es_new(struct bgp *bgp,
5017 esi_t *esi,
5018 struct ipaddr *originator_ip)
5019 {
5020 char buf[100];
5021 struct evpnes *es;
5022
5023 if (!bgp)
5024 return NULL;
5025
5026 es = XCALLOC(MTYPE_BGP_EVPN_ES, sizeof(struct evpnes));
5027 if (!es)
5028 return NULL;
5029
5030 /* set the ESI and originator_ip */
5031 memcpy(&es->esi, esi, sizeof(esi_t));
5032 memcpy(&es->originator_ip, originator_ip, sizeof(struct ipaddr));
5033
5034 /* Initialise the VTEP list */
5035 es->vtep_list = list_new();
5036 es->vtep_list->cmp = (int (*)(void *, void *))evpn_vtep_ip_cmp;
5037
5038 /* auto derive RD for this es */
5039 bf_assign_index(bm->rd_idspace, es->rd_id);
5040 es->prd.family = AF_UNSPEC;
5041 es->prd.prefixlen = 64;
5042 sprintf(buf, "%s:%hu", inet_ntoa(bgp->router_id), es->rd_id);
5043 (void)str2prefix_rd(buf, &es->prd);
5044
5045 /* Initialize the ES route table */
5046 es->route_table = bgp_table_init(bgp, AFI_L2VPN, SAFI_EVPN);
5047
5048 /* Add to hash */
5049 if (!hash_get(bgp->esihash, es, hash_alloc_intern)) {
5050 XFREE(MTYPE_BGP_EVPN_ES, es);
5051 return NULL;
5052 }
5053
5054 QOBJ_REG(es, evpnes);
5055 return es;
5056 }
5057
5058 /*
5059 * Free a given ES -
5060 * This just frees appropriate memory, caller should have taken other
5061 * needed actions.
5062 */
5063 void bgp_evpn_es_free(struct bgp *bgp, struct evpnes *es)
5064 {
5065 list_delete_and_null(&es->vtep_list);
5066 bgp_table_unlock(es->route_table);
5067 bf_release_index(bm->rd_idspace, es->rd_id);
5068 hash_release(bgp->esihash, es);
5069 QOBJ_UNREG(es);
5070 XFREE(MTYPE_BGP_EVPN_ES, es);
5071 }
5072
5073 /*
5074 * Import evpn route from global table to VNI/VRF/ESI.
5075 */
5076 int bgp_evpn_import_route(struct bgp *bgp, afi_t afi, safi_t safi,
5077 struct prefix *p, struct bgp_info *ri)
5078 {
5079 return install_uninstall_evpn_route(bgp, afi, safi, p, ri, 1);
5080 }
5081
5082 /*
5083 * Unimport evpn route from VNI/VRF/ESI.
5084 */
5085 int bgp_evpn_unimport_route(struct bgp *bgp, afi_t afi, safi_t safi,
5086 struct prefix *p, struct bgp_info *ri)
5087 {
5088 return install_uninstall_evpn_route(bgp, afi, safi, p, ri, 0);
5089 }
5090
5091 /* filter routes which have martian next hops */
5092 int bgp_filter_evpn_routes_upon_martian_nh_change(struct bgp *bgp)
5093 {
5094 afi_t afi;
5095 safi_t safi;
5096 struct bgp_node *rd_rn, *rn;
5097 struct bgp_table *table;
5098 struct bgp_info *ri;
5099
5100 afi = AFI_L2VPN;
5101 safi = SAFI_EVPN;
5102
5103 /* Walk entire global routing table and evaluate routes which could be
5104 * imported into this VPN. Note that we cannot just look at the routes
5105 * for the VNI's RD -
5106 * remote routes applicable for this VNI could have any RD.
5107 */
5108 /* EVPN routes are a 2-level table. */
5109 for (rd_rn = bgp_table_top(bgp->rib[afi][safi]); rd_rn;
5110 rd_rn = bgp_route_next(rd_rn)) {
5111 table = (struct bgp_table *)(rd_rn->info);
5112 if (!table)
5113 continue;
5114
5115 for (rn = bgp_table_top(table); rn; rn = bgp_route_next(rn)) {
5116
5117 for (ri = rn->info; ri; ri = ri->next) {
5118
5119 /* Consider "valid" remote routes applicable for
5120 * this VNI. */
5121 if (!(ri->type == ZEBRA_ROUTE_BGP
5122 && ri->sub_type == BGP_ROUTE_NORMAL))
5123 continue;
5124
5125 if (bgp_nexthop_self(bgp, ri->attr->nexthop)) {
5126
5127 char attr_str[BUFSIZ];
5128 char pbuf[PREFIX_STRLEN];
5129
5130 bgp_dump_attr(ri->attr, attr_str,
5131 BUFSIZ);
5132
5133 if (bgp_debug_update(ri->peer, &rn->p,
5134 NULL, 1))
5135 zlog_debug(
5136 "%u: prefix %s with attr %s - DENIED due to martian or self nexthop",
5137 bgp->vrf_id,
5138 prefix2str(
5139 &rn->p, pbuf,
5140 sizeof(pbuf)),
5141 attr_str);
5142
5143 bgp_evpn_unimport_route(bgp, afi, safi,
5144 &rn->p, ri);
5145
5146 bgp_rib_remove(rn, ri, ri->peer, afi,
5147 safi);
5148 }
5149 }
5150 }
5151 }
5152
5153 return 0;
5154 }
5155
5156 /*
5157 * Handle del of a local MACIP.
5158 */
5159 int bgp_evpn_local_macip_del(struct bgp *bgp, vni_t vni, struct ethaddr *mac,
5160 struct ipaddr *ip)
5161 {
5162 struct bgpevpn *vpn;
5163 struct prefix_evpn p;
5164
5165 /* Lookup VNI hash - should exist. */
5166 vpn = bgp_evpn_lookup_vni(bgp, vni);
5167 if (!vpn || !is_vni_live(vpn)) {
5168 flog_warn(EC_BGP_EVPN_VPN_VNI,
5169 "%u: VNI hash entry for VNI %u %s at MACIP DEL",
5170 bgp->vrf_id, vni, vpn ? "not live" : "not found");
5171 return -1;
5172 }
5173
5174 /* Remove EVPN type-2 route and schedule for processing. */
5175 build_evpn_type2_prefix(&p, mac, ip);
5176 delete_evpn_route(bgp, vpn, &p);
5177
5178 return 0;
5179 }
5180
5181 /*
5182 * Handle add of a local MACIP.
5183 */
5184 int bgp_evpn_local_macip_add(struct bgp *bgp, vni_t vni, struct ethaddr *mac,
5185 struct ipaddr *ip, uint8_t flags, uint32_t seq)
5186 {
5187 struct bgpevpn *vpn;
5188 struct prefix_evpn p;
5189
5190 /* Lookup VNI hash - should exist. */
5191 vpn = bgp_evpn_lookup_vni(bgp, vni);
5192 if (!vpn || !is_vni_live(vpn)) {
5193 flog_warn(EC_BGP_EVPN_VPN_VNI,
5194 "%u: VNI hash entry for VNI %u %s at MACIP ADD",
5195 bgp->vrf_id, vni, vpn ? "not live" : "not found");
5196 return -1;
5197 }
5198
5199 /* Create EVPN type-2 route and schedule for processing. */
5200 build_evpn_type2_prefix(&p, mac, ip);
5201 if (update_evpn_route(bgp, vpn, &p, flags, seq)) {
5202 char buf[ETHER_ADDR_STRLEN];
5203 char buf2[INET6_ADDRSTRLEN];
5204
5205 flog_err(
5206 EC_BGP_EVPN_ROUTE_CREATE,
5207 "%u:Failed to create Type-2 route, VNI %u %s MAC %s IP %s (flags: 0x%x)",
5208 bgp->vrf_id, vpn->vni,
5209 CHECK_FLAG(flags, ZEBRA_MACIP_TYPE_STICKY)
5210 ? "sticky gateway"
5211 : "",
5212 prefix_mac2str(mac, buf, sizeof(buf)),
5213 ipaddr2str(ip, buf2, sizeof(buf2)), flags);
5214 return -1;
5215 }
5216
5217 return 0;
5218 }
5219
5220 static void link_l2vni_hash_to_l3vni(struct hash_backet *backet,
5221 struct bgp *bgp_vrf)
5222 {
5223 struct bgpevpn *vpn = (struct bgpevpn *)backet->data;
5224 struct bgp *bgp_def = NULL;
5225
5226 bgp_def = bgp_get_default();
5227 assert(bgp_def);
5228
5229 if (vpn->tenant_vrf_id == bgp_vrf->vrf_id)
5230 bgpevpn_link_to_l3vni(vpn);
5231 }
5232
5233 int bgp_evpn_local_l3vni_add(vni_t l3vni, vrf_id_t vrf_id, struct ethaddr *rmac,
5234 struct in_addr originator_ip, int filter)
5235 {
5236 struct bgp *bgp_vrf = NULL; /* bgp VRF instance */
5237 struct bgp *bgp_def = NULL; /* default bgp instance */
5238 struct listnode *node = NULL;
5239 struct bgpevpn *vpn = NULL;
5240 as_t as = 0;
5241
5242 /* get the default instance - required to get the AS number for VRF
5243 * auto-creatio
5244 */
5245 bgp_def = bgp_get_default();
5246 if (!bgp_def) {
5247 flog_err(
5248 EC_BGP_NO_DFLT,
5249 "Cannot process L3VNI %u ADD - default BGP instance not yet created",
5250 l3vni);
5251 return -1;
5252 }
5253 as = bgp_def->as;
5254
5255 /* if the BGP vrf instance doesnt exist - create one */
5256 bgp_vrf = bgp_lookup_by_name(vrf_id_to_name(vrf_id));
5257 if (!bgp_vrf) {
5258
5259 int ret = 0;
5260
5261 ret = bgp_get(&bgp_vrf, &as, vrf_id_to_name(vrf_id),
5262 BGP_INSTANCE_TYPE_VRF);
5263 switch (ret) {
5264 case BGP_ERR_MULTIPLE_INSTANCE_NOT_SET:
5265 flog_err(EC_BGP_MULTI_INSTANCE,
5266 "'bgp multiple-instance' not present\n");
5267 return -1;
5268 case BGP_ERR_AS_MISMATCH:
5269 flog_err(EC_BGP_EVPN_AS_MISMATCH,
5270 "BGP is already running; AS is %u\n", as);
5271 return -1;
5272 case BGP_ERR_INSTANCE_MISMATCH:
5273 flog_err(EC_BGP_EVPN_INSTANCE_MISMATCH,
5274 "BGP instance name and AS number mismatch\n");
5275 return -1;
5276 }
5277
5278 /* mark as auto created */
5279 SET_FLAG(bgp_vrf->vrf_flags, BGP_VRF_AUTO);
5280 }
5281
5282 /* associate with l3vni */
5283 bgp_vrf->l3vni = l3vni;
5284
5285 /* set the router mac - to be used in mac-ip routes for this vrf */
5286 memcpy(&bgp_vrf->rmac, rmac, sizeof(struct ethaddr));
5287
5288 /* set the originator ip */
5289 bgp_vrf->originator_ip = originator_ip;
5290
5291 /* set the right filter - are we using l3vni only for prefix routes? */
5292 if (filter)
5293 SET_FLAG(bgp_vrf->vrf_flags, BGP_VRF_L3VNI_PREFIX_ROUTES_ONLY);
5294
5295 /* auto derive RD/RT */
5296 if (!CHECK_FLAG(bgp_vrf->vrf_flags, BGP_VRF_IMPORT_RT_CFGD))
5297 evpn_auto_rt_import_add_for_vrf(bgp_vrf);
5298 if (!CHECK_FLAG(bgp_vrf->vrf_flags, BGP_VRF_EXPORT_RT_CFGD))
5299 evpn_auto_rt_export_add_for_vrf(bgp_vrf);
5300 bgp_evpn_derive_auto_rd_for_vrf(bgp_vrf);
5301
5302 /* link all corresponding l2vnis */
5303 hash_iterate(bgp_def->vnihash,
5304 (void (*)(struct hash_backet *,
5305 void *))link_l2vni_hash_to_l3vni,
5306 bgp_vrf);
5307
5308 /* Only update all corresponding type-2 routes if we are advertising two
5309 * labels along with type-2 routes
5310 */
5311 if (!filter)
5312 for (ALL_LIST_ELEMENTS_RO(bgp_vrf->l2vnis, node, vpn))
5313 update_routes_for_vni(bgp_def, vpn);
5314
5315 /* advertise type-5 routes if needed */
5316 update_advertise_vrf_routes(bgp_vrf);
5317
5318 /* install all remote routes belonging to this l3vni into correspondng
5319 * vrf */
5320 install_routes_for_vrf(bgp_vrf);
5321
5322 return 0;
5323 }
5324
5325 int bgp_evpn_local_l3vni_del(vni_t l3vni, vrf_id_t vrf_id)
5326 {
5327 struct bgp *bgp_vrf = NULL; /* bgp vrf instance */
5328 struct bgp *bgp_def = NULL; /* default bgp instance */
5329 struct listnode *node = NULL;
5330 struct listnode *next = NULL;
5331 struct bgpevpn *vpn = NULL;
5332
5333 bgp_vrf = bgp_lookup_by_vrf_id(vrf_id);
5334 if (!bgp_vrf) {
5335 flog_err(
5336 EC_BGP_NO_DFLT,
5337 "Cannot process L3VNI %u Del - Could not find BGP instance",
5338 l3vni);
5339 return -1;
5340 }
5341
5342 bgp_def = bgp_get_default();
5343 if (!bgp_def) {
5344 flog_err(
5345 EC_BGP_NO_DFLT,
5346 "Cannot process L3VNI %u Del - Could not find default BGP instance",
5347 l3vni);
5348 return -1;
5349 }
5350
5351 /* Remove remote routes from BGT VRF even if BGP_VRF_AUTO is configured,
5352 * bgp_delete would not remove/decrement bgp_info of the ip_prefix
5353 * routes. This will uninstalling the routes from zebra and decremnt the
5354 * bgp info count.
5355 */
5356 uninstall_routes_for_vrf(bgp_vrf);
5357
5358 /* delete/withdraw all type-5 routes */
5359 delete_withdraw_vrf_routes(bgp_vrf);
5360
5361 /* remove the l3vni from vrf instance */
5362 bgp_vrf->l3vni = 0;
5363
5364 /* remove the Rmac from the BGP vrf */
5365 memset(&bgp_vrf->rmac, 0, sizeof(struct ethaddr));
5366
5367 /* delete RD/RT */
5368 if (!list_isempty(bgp_vrf->vrf_import_rtl)) {
5369 bgp_evpn_unmap_vrf_from_its_rts(bgp_vrf);
5370 list_delete_all_node(bgp_vrf->vrf_import_rtl);
5371 }
5372 if (!list_isempty(bgp_vrf->vrf_export_rtl)) {
5373 list_delete_all_node(bgp_vrf->vrf_export_rtl);
5374 }
5375
5376 /* update all corresponding local mac-ip routes */
5377 if (!CHECK_FLAG(bgp_vrf->vrf_flags, BGP_VRF_L3VNI_PREFIX_ROUTES_ONLY)) {
5378 for (ALL_LIST_ELEMENTS_RO(bgp_vrf->l2vnis, node, vpn)) {
5379 UNSET_FLAG(vpn->flags, VNI_FLAG_USE_TWO_LABELS);
5380 update_routes_for_vni(bgp_def, vpn);
5381 }
5382 }
5383
5384 /* If any L2VNIs point to this instance, unlink them. */
5385 for (ALL_LIST_ELEMENTS(bgp_vrf->l2vnis, node, next, vpn))
5386 bgpevpn_unlink_from_l3vni(vpn);
5387
5388 /* Delete the instance if it was autocreated */
5389 if (CHECK_FLAG(bgp_vrf->vrf_flags, BGP_VRF_AUTO))
5390 bgp_delete(bgp_vrf);
5391
5392 return 0;
5393 }
5394
5395 /*
5396 * Handle del of a local VNI.
5397 */
5398 int bgp_evpn_local_vni_del(struct bgp *bgp, vni_t vni)
5399 {
5400 struct bgpevpn *vpn;
5401
5402 /* Locate VNI hash */
5403 vpn = bgp_evpn_lookup_vni(bgp, vni);
5404 if (!vpn) {
5405 if (bgp_debug_zebra(NULL))
5406 flog_warn(
5407 EC_BGP_EVPN_VPN_VNI,
5408 "%u: VNI hash entry for VNI %u not found at DEL",
5409 bgp->vrf_id, vni);
5410 return 0;
5411 }
5412
5413 /* Remove all local EVPN routes and schedule for processing (to
5414 * withdraw from peers).
5415 */
5416 delete_routes_for_vni(bgp, vpn);
5417
5418 /*
5419 * tunnel is no longer active, del tunnel ip address from tip_hash
5420 */
5421 bgp_tip_del(bgp, &vpn->originator_ip);
5422
5423 /* Clear "live" flag and see if hash needs to be freed. */
5424 UNSET_FLAG(vpn->flags, VNI_FLAG_LIVE);
5425 if (!is_vni_configured(vpn))
5426 bgp_evpn_free(bgp, vpn);
5427
5428 return 0;
5429 }
5430
5431 /*
5432 * Handle add (or update) of a local VNI. The VNI changes we care
5433 * about are for the local-tunnel-ip and the (tenant) VRF.
5434 */
5435 int bgp_evpn_local_vni_add(struct bgp *bgp, vni_t vni,
5436 struct in_addr originator_ip, vrf_id_t tenant_vrf_id)
5437 {
5438 struct bgpevpn *vpn;
5439 struct prefix_evpn p;
5440
5441 /* Lookup VNI. If present and no change, exit. */
5442 vpn = bgp_evpn_lookup_vni(bgp, vni);
5443 if (vpn) {
5444
5445 if (is_vni_live(vpn)
5446 && IPV4_ADDR_SAME(&vpn->originator_ip, &originator_ip)
5447 && vpn->tenant_vrf_id == tenant_vrf_id)
5448 /* Probably some other param has changed that we don't
5449 * care about. */
5450 return 0;
5451
5452 /* Update tenant_vrf_id if it has changed. */
5453 if (vpn->tenant_vrf_id != tenant_vrf_id) {
5454 bgpevpn_unlink_from_l3vni(vpn);
5455 vpn->tenant_vrf_id = tenant_vrf_id;
5456 bgpevpn_link_to_l3vni(vpn);
5457 }
5458
5459 /* If tunnel endpoint IP has changed, update (and delete prior
5460 * type-3 route, if needed.)
5461 */
5462 if (!IPV4_ADDR_SAME(&vpn->originator_ip, &originator_ip))
5463 handle_tunnel_ip_change(bgp, vpn, originator_ip);
5464
5465 /* Update all routes with new endpoint IP and/or export RT
5466 * for VRFs
5467 */
5468 if (is_vni_live(vpn))
5469 update_routes_for_vni(bgp, vpn);
5470 }
5471
5472 /* Create or update as appropriate. */
5473 if (!vpn) {
5474 vpn = bgp_evpn_new(bgp, vni, originator_ip, tenant_vrf_id);
5475 if (!vpn) {
5476 flog_err(
5477 EC_BGP_VNI,
5478 "%u: Failed to allocate VNI entry for VNI %u - at Add",
5479 bgp->vrf_id, vni);
5480 return -1;
5481 }
5482 }
5483
5484 /* if the VNI is live already, there is nothing more to do */
5485 if (is_vni_live(vpn))
5486 return 0;
5487
5488 /* Mark as "live" */
5489 SET_FLAG(vpn->flags, VNI_FLAG_LIVE);
5490
5491 /* tunnel is now active, add tunnel-ip to db */
5492 bgp_tip_add(bgp, &originator_ip);
5493
5494 /* filter routes as nexthop database has changed */
5495 bgp_filter_evpn_routes_upon_martian_nh_change(bgp);
5496
5497 /* Create EVPN type-3 route and schedule for processing. */
5498 build_evpn_type3_prefix(&p, vpn->originator_ip);
5499 if (update_evpn_route(bgp, vpn, &p, 0, 0)) {
5500 flog_err(EC_BGP_EVPN_ROUTE_CREATE,
5501 "%u: Type3 route creation failure for VNI %u",
5502 bgp->vrf_id, vni);
5503 return -1;
5504 }
5505
5506 /* If we have learnt and retained remote routes (VTEPs, MACs) for this
5507 * VNI,
5508 * install them.
5509 */
5510 install_routes_for_vni(bgp, vpn);
5511
5512 /* If we are advertising gateway mac-ip
5513 It needs to be conveyed again to zebra */
5514 bgp_zebra_advertise_gw_macip(bgp, vpn->advertise_gw_macip, vpn->vni);
5515
5516 return 0;
5517 }
5518
5519 /*
5520 * bgp_evpn_local_es_del
5521 */
5522 int bgp_evpn_local_es_del(struct bgp *bgp,
5523 esi_t *esi,
5524 struct ipaddr *originator_ip)
5525 {
5526 char buf[ESI_STR_LEN];
5527 struct evpnes *es = NULL;
5528
5529 if (!bgp->esihash) {
5530 flog_err(EC_BGP_ES_CREATE, "%u: ESI hash not yet created",
5531 bgp->vrf_id);
5532 return -1;
5533 }
5534
5535 /* Lookup ESI hash - should exist. */
5536 es = bgp_evpn_lookup_es(bgp, esi);
5537 if (!es) {
5538 flog_warn(EC_BGP_EVPN_ESI,
5539 "%u: ESI hash entry for ESI %s at Local ES DEL",
5540 bgp->vrf_id, esi_to_str(esi, buf, sizeof(buf)));
5541 return -1;
5542 }
5543
5544 /* Delete all local EVPN ES routes from ESI table
5545 * and schedule for processing (to withdraw from peers))
5546 */
5547 delete_routes_for_es(bgp, es);
5548
5549 /* free the hash entry */
5550 bgp_evpn_es_free(bgp, es);
5551
5552 return 0;
5553 }
5554
5555 /*
5556 * bgp_evpn_local_es_add
5557 */
5558 int bgp_evpn_local_es_add(struct bgp *bgp,
5559 esi_t *esi,
5560 struct ipaddr *originator_ip)
5561 {
5562 char buf[ESI_STR_LEN];
5563 struct evpnes *es = NULL;
5564 struct prefix_evpn p;
5565
5566 if (!bgp->esihash) {
5567 flog_err(EC_BGP_ES_CREATE, "%u: ESI hash not yet created",
5568 bgp->vrf_id);
5569 return -1;
5570 }
5571
5572 /* create the new es */
5573 es = bgp_evpn_lookup_es(bgp, esi);
5574 if (!es) {
5575 es = bgp_evpn_es_new(bgp, esi, originator_ip);
5576 if (!es) {
5577 flog_err(
5578 EC_BGP_ES_CREATE,
5579 "%u: Failed to allocate ES entry for ESI %s - at Local ES Add",
5580 bgp->vrf_id, esi_to_str(esi, buf, sizeof(buf)));
5581 return -1;
5582 }
5583 }
5584 UNSET_FLAG(es->flags, EVPNES_REMOTE);
5585 SET_FLAG(es->flags, EVPNES_LOCAL);
5586
5587 build_evpn_type4_prefix(&p, esi, originator_ip->ipaddr_v4);
5588 if (update_evpn_type4_route(bgp, es, &p)) {
5589 flog_err(EC_BGP_EVPN_ROUTE_CREATE,
5590 "%u: Type4 route creation failure for ESI %s",
5591 bgp->vrf_id, esi_to_str(esi, buf, sizeof(buf)));
5592 return -1;
5593 }
5594
5595 /* import all remote ES routes in th ES table */
5596 install_routes_for_es(bgp, es);
5597
5598 return 0;
5599 }
5600
5601 /*
5602 * Cleanup EVPN information on disable - Need to delete and withdraw
5603 * EVPN routes from peers.
5604 */
5605 void bgp_evpn_cleanup_on_disable(struct bgp *bgp)
5606 {
5607 hash_iterate(bgp->vnihash, (void (*)(struct hash_backet *,
5608 void *))cleanup_vni_on_disable,
5609 bgp);
5610 }
5611
5612 /*
5613 * Cleanup EVPN information - invoked at the time of bgpd exit or when the
5614 * BGP instance (default) is being freed.
5615 */
5616 void bgp_evpn_cleanup(struct bgp *bgp)
5617 {
5618 hash_iterate(bgp->vnihash,
5619 (void (*)(struct hash_backet *, void *))free_vni_entry,
5620 bgp);
5621
5622 hash_free(bgp->import_rt_hash);
5623 bgp->import_rt_hash = NULL;
5624
5625 hash_free(bgp->vrf_import_rt_hash);
5626 bgp->vrf_import_rt_hash = NULL;
5627
5628 hash_free(bgp->vnihash);
5629 bgp->vnihash = NULL;
5630 if (bgp->esihash)
5631 hash_free(bgp->esihash);
5632 bgp->esihash = NULL;
5633
5634 list_delete_and_null(&bgp->vrf_import_rtl);
5635 list_delete_and_null(&bgp->vrf_export_rtl);
5636 list_delete_and_null(&bgp->l2vnis);
5637 }
5638
5639 /*
5640 * Initialization for EVPN
5641 * Create
5642 * VNI hash table
5643 * hash for RT to VNI
5644 */
5645 void bgp_evpn_init(struct bgp *bgp)
5646 {
5647 bgp->vnihash =
5648 hash_create(vni_hash_key_make, vni_hash_cmp, "BGP VNI Hash");
5649 bgp->esihash =
5650 hash_create(esi_hash_keymake, esi_cmp,
5651 "BGP EVPN Local ESI Hash");
5652 bgp->import_rt_hash =
5653 hash_create(import_rt_hash_key_make, import_rt_hash_cmp,
5654 "BGP Import RT Hash");
5655 bgp->vrf_import_rt_hash =
5656 hash_create(vrf_import_rt_hash_key_make, vrf_import_rt_hash_cmp,
5657 "BGP VRF Import RT Hash");
5658 bgp->vrf_import_rtl = list_new();
5659 bgp->vrf_import_rtl->cmp =
5660 (int (*)(void *, void *))evpn_route_target_cmp;
5661 bgp->vrf_import_rtl->del = evpn_xxport_delete_ecomm;
5662 bgp->vrf_export_rtl = list_new();
5663 bgp->vrf_export_rtl->cmp =
5664 (int (*)(void *, void *))evpn_route_target_cmp;
5665 bgp->vrf_export_rtl->del = evpn_xxport_delete_ecomm;
5666 bgp->l2vnis = list_new();
5667 bgp->l2vnis->cmp = (int (*)(void *, void *))vni_hash_cmp;
5668 }
5669
5670 void bgp_evpn_vrf_delete(struct bgp *bgp_vrf)
5671 {
5672 bgp_evpn_unmap_vrf_from_its_rts(bgp_vrf);
5673 }