]> git.proxmox.com Git - mirror_frr.git/blob - bgpd/bgp_evpn.c
Merge pull request #3310 from adeg/bugfix/bgpd-mplsvpn-route-import-check
[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 #include "bgpd/bgp_addpath.h"
51
52 /*
53 * Definitions and external declarations.
54 */
55 extern struct zclient *zclient;
56
57 DEFINE_QOBJ_TYPE(bgpevpn)
58 DEFINE_QOBJ_TYPE(evpnes)
59
60
61 /*
62 * Static function declarations
63 */
64 static void delete_evpn_route_entry(struct bgp *bgp, afi_t afi, safi_t safi,
65 struct bgp_node *rn,
66 struct bgp_path_info **pi);
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(void *p1, void *p2)
75 {
76 const struct in_addr *ip1 = p1;
77 const struct in_addr *ip2 = p2;
78
79 return ip1->s_addr - ip2->s_addr;
80 }
81
82 /*
83 * Make hash key for ESI.
84 */
85 static unsigned int esi_hash_keymake(void *p)
86 {
87 struct evpnes *pes = p;
88 const void *pnt = (void *)pes->esi.val;
89
90 return jhash(pnt, ESI_BYTES, 0xa5a5a55a);
91 }
92
93 /*
94 * Compare two ESIs.
95 */
96 static bool esi_cmp(const void *p1, const void *p2)
97 {
98 const struct evpnes *pes1 = p1;
99 const struct evpnes *pes2 = p2;
100
101 if (pes1 == NULL && pes2 == NULL)
102 return true;
103
104 if (pes1 == NULL || pes2 == NULL)
105 return false;
106
107 return (memcmp(pes1->esi.val, pes2->esi.val, ESI_BYTES) == 0);
108 }
109
110 /*
111 * Make vni hash key.
112 */
113 static unsigned int vni_hash_key_make(void *p)
114 {
115 struct bgpevpn *vpn = p;
116 return (jhash_1word(vpn->vni, 0));
117 }
118
119 /*
120 * Comparison function for vni hash
121 */
122 static bool vni_hash_cmp(const void *p1, const void *p2)
123 {
124 const struct bgpevpn *vpn1 = p1;
125 const struct bgpevpn *vpn2 = p2;
126
127 if (!vpn1 && !vpn2)
128 return true;
129 if (!vpn1 || !vpn2)
130 return false;
131 return (vpn1->vni == vpn2->vni);
132 }
133
134 static int vni_list_cmp(void *p1, void *p2)
135 {
136 const struct bgpevpn *vpn1 = p1;
137 const struct bgpevpn *vpn2 = p2;
138
139 return vpn1->vni - vpn2->vni;
140 }
141
142 /*
143 * Make vrf import route target hash key.
144 */
145 static unsigned int vrf_import_rt_hash_key_make(void *p)
146 {
147 struct vrf_irt_node *irt = p;
148 char *pnt = irt->rt.val;
149
150 return jhash(pnt, 8, 0x5abc1234);
151 }
152
153 /*
154 * Comparison function for vrf import rt hash
155 */
156 static bool vrf_import_rt_hash_cmp(const void *p1, const void *p2)
157 {
158 const struct vrf_irt_node *irt1 = p1;
159 const struct vrf_irt_node *irt2 = p2;
160
161 if (irt1 == NULL && irt2 == NULL)
162 return true;
163
164 if (irt1 == NULL || irt2 == NULL)
165 return false;
166
167 return (memcmp(irt1->rt.val, irt2->rt.val, ECOMMUNITY_SIZE) == 0);
168 }
169
170 /*
171 * Create a new vrf import_rt in default instance
172 */
173 static struct vrf_irt_node *vrf_import_rt_new(struct ecommunity_val *rt)
174 {
175 struct bgp *bgp_def = NULL;
176 struct vrf_irt_node *irt;
177
178 bgp_def = bgp_get_default();
179 if (!bgp_def) {
180 flog_err(EC_BGP_NO_DFLT,
181 "vrf import rt new - def instance not created yet");
182 return NULL;
183 }
184
185 irt = XCALLOC(MTYPE_BGP_EVPN_VRF_IMPORT_RT,
186 sizeof(struct vrf_irt_node));
187 if (!irt)
188 return NULL;
189
190 irt->rt = *rt;
191 irt->vrfs = list_new();
192
193 /* Add to hash */
194 if (!hash_get(bgp_def->vrf_import_rt_hash, irt, hash_alloc_intern)) {
195 XFREE(MTYPE_BGP_EVPN_VRF_IMPORT_RT, irt);
196 return NULL;
197 }
198
199 return irt;
200 }
201
202 /*
203 * Free the vrf import rt node
204 */
205 static void vrf_import_rt_free(struct vrf_irt_node *irt)
206 {
207 struct bgp *bgp_def = NULL;
208
209 bgp_def = bgp_get_default();
210 if (!bgp_def) {
211 flog_err(EC_BGP_NO_DFLT,
212 "vrf import rt free - def instance not created yet");
213 return;
214 }
215
216 hash_release(bgp_def->vrf_import_rt_hash, irt);
217 list_delete(&irt->vrfs);
218 XFREE(MTYPE_BGP_EVPN_VRF_IMPORT_RT, irt);
219 }
220
221 /*
222 * Function to lookup Import RT node - used to map a RT to set of
223 * VNIs importing routes with that RT.
224 */
225 static struct vrf_irt_node *lookup_vrf_import_rt(struct ecommunity_val *rt)
226 {
227 struct bgp *bgp_def = NULL;
228 struct vrf_irt_node *irt;
229 struct vrf_irt_node tmp;
230
231 bgp_def = bgp_get_default();
232 if (!bgp_def) {
233 flog_err(EC_BGP_NO_DFLT,
234 "vrf import rt lookup - def instance not created yet");
235 return NULL;
236 }
237
238 memset(&tmp, 0, sizeof(struct vrf_irt_node));
239 memcpy(&tmp.rt, rt, ECOMMUNITY_SIZE);
240 irt = hash_lookup(bgp_def->vrf_import_rt_hash, &tmp);
241 return irt;
242 }
243
244 /*
245 * Is specified VRF present on the RT's list of "importing" VRFs?
246 */
247 static int is_vrf_present_in_irt_vrfs(struct list *vrfs, struct bgp *bgp_vrf)
248 {
249 struct listnode *node = NULL, *nnode = NULL;
250 struct bgp *tmp_bgp_vrf = NULL;
251
252 for (ALL_LIST_ELEMENTS(vrfs, node, nnode, tmp_bgp_vrf)) {
253 if (tmp_bgp_vrf == bgp_vrf)
254 return 1;
255 }
256 return 0;
257 }
258
259 /*
260 * Make import route target hash key.
261 */
262 static unsigned int import_rt_hash_key_make(void *p)
263 {
264 struct irt_node *irt = p;
265 char *pnt = irt->rt.val;
266
267 return jhash(pnt, 8, 0xdeadbeef);
268 }
269
270 /*
271 * Comparison function for import rt hash
272 */
273 static bool import_rt_hash_cmp(const void *p1, const void *p2)
274 {
275 const struct irt_node *irt1 = p1;
276 const struct irt_node *irt2 = p2;
277
278 if (irt1 == NULL && irt2 == NULL)
279 return true;
280
281 if (irt1 == NULL || irt2 == NULL)
282 return false;
283
284 return (memcmp(irt1->rt.val, irt2->rt.val, ECOMMUNITY_SIZE) == 0);
285 }
286
287 /*
288 * Create a new import_rt
289 */
290 static struct irt_node *import_rt_new(struct bgp *bgp,
291 struct ecommunity_val *rt)
292 {
293 struct irt_node *irt;
294
295 if (!bgp)
296 return NULL;
297
298 irt = XCALLOC(MTYPE_BGP_EVPN_IMPORT_RT, sizeof(struct irt_node));
299 if (!irt)
300 return NULL;
301
302 irt->rt = *rt;
303 irt->vnis = list_new();
304
305 /* Add to hash */
306 if (!hash_get(bgp->import_rt_hash, irt, hash_alloc_intern)) {
307 XFREE(MTYPE_BGP_EVPN_IMPORT_RT, irt);
308 return NULL;
309 }
310
311 return irt;
312 }
313
314 /*
315 * Free the import rt node
316 */
317 static void import_rt_free(struct bgp *bgp, struct irt_node *irt)
318 {
319 hash_release(bgp->import_rt_hash, irt);
320 list_delete(&irt->vnis);
321 XFREE(MTYPE_BGP_EVPN_IMPORT_RT, irt);
322 }
323
324 /*
325 * Function to lookup Import RT node - used to map a RT to set of
326 * VNIs importing routes with that RT.
327 */
328 static struct irt_node *lookup_import_rt(struct bgp *bgp,
329 struct ecommunity_val *rt)
330 {
331 struct irt_node *irt;
332 struct irt_node tmp;
333
334 memset(&tmp, 0, sizeof(struct irt_node));
335 memcpy(&tmp.rt, rt, ECOMMUNITY_SIZE);
336 irt = hash_lookup(bgp->import_rt_hash, &tmp);
337 return irt;
338 }
339
340 /*
341 * Is specified VNI present on the RT's list of "importing" VNIs?
342 */
343 static int is_vni_present_in_irt_vnis(struct list *vnis, struct bgpevpn *vpn)
344 {
345 struct listnode *node, *nnode;
346 struct bgpevpn *tmp_vpn;
347
348 for (ALL_LIST_ELEMENTS(vnis, node, nnode, tmp_vpn)) {
349 if (tmp_vpn == vpn)
350 return 1;
351 }
352
353 return 0;
354 }
355
356 /*
357 * Compare Route Targets.
358 */
359 static int evpn_route_target_cmp(struct ecommunity *ecom1,
360 struct ecommunity *ecom2)
361 {
362 if (ecom1 && !ecom2)
363 return -1;
364
365 if (!ecom1 && ecom2)
366 return 1;
367
368 if (!ecom1 && !ecom2)
369 return 0;
370
371 if (ecom1->str && !ecom2->str)
372 return -1;
373
374 if (!ecom1->str && ecom2->str)
375 return 1;
376
377 if (!ecom1->str && !ecom2->str)
378 return 0;
379
380 return strcmp(ecom1->str, ecom2->str);
381 }
382
383 static void evpn_xxport_delete_ecomm(void *val)
384 {
385 struct ecommunity *ecomm = val;
386 ecommunity_free(&ecomm);
387 }
388
389 /*
390 * Mask off global-admin field of specified extended community (RT),
391 * just retain the local-admin field.
392 */
393 static inline void mask_ecom_global_admin(struct ecommunity_val *dst,
394 struct ecommunity_val *src)
395 {
396 uint8_t type;
397
398 type = src->val[0];
399 dst->val[0] = 0;
400 if (type == ECOMMUNITY_ENCODE_AS) {
401 dst->val[2] = dst->val[3] = 0;
402 } else if (type == ECOMMUNITY_ENCODE_AS4
403 || type == ECOMMUNITY_ENCODE_IP) {
404 dst->val[2] = dst->val[3] = 0;
405 dst->val[4] = dst->val[5] = 0;
406 }
407 }
408
409 /*
410 * Map one RT to specified VRF.
411 * bgp_vrf = BGP vrf instance
412 */
413 static void map_vrf_to_rt(struct bgp *bgp_vrf, struct ecommunity_val *eval)
414 {
415 struct vrf_irt_node *irt = NULL;
416 struct ecommunity_val eval_tmp;
417
418 /* If using "automatic" RT,
419 * we only care about the local-admin sub-field.
420 * This is to facilitate using L3VNI(VRF-VNI)
421 * as the RT for EBGP peering too.
422 */
423 memcpy(&eval_tmp, eval, ECOMMUNITY_SIZE);
424 if (!CHECK_FLAG(bgp_vrf->vrf_flags, BGP_VRF_IMPORT_RT_CFGD))
425 mask_ecom_global_admin(&eval_tmp, eval);
426
427 irt = lookup_vrf_import_rt(&eval_tmp);
428 if (irt && is_vrf_present_in_irt_vrfs(irt->vrfs, bgp_vrf))
429 /* Already mapped. */
430 return;
431
432 if (!irt)
433 irt = vrf_import_rt_new(&eval_tmp);
434
435 /* Add VRF to the list for this RT. */
436 listnode_add(irt->vrfs, bgp_vrf);
437 }
438
439 /*
440 * Unmap specified VRF from specified RT. If there are no other
441 * VRFs for this RT, then the RT hash is deleted.
442 * bgp_vrf: BGP VRF specific instance
443 */
444 static void unmap_vrf_from_rt(struct bgp *bgp_vrf, struct vrf_irt_node *irt)
445 {
446 /* Delete VRF from list for this RT. */
447 listnode_delete(irt->vrfs, bgp_vrf);
448 if (!listnode_head(irt->vrfs)) {
449 vrf_import_rt_free(irt);
450 }
451 }
452
453 /*
454 * Map one RT to specified VNI.
455 */
456 static void map_vni_to_rt(struct bgp *bgp, struct bgpevpn *vpn,
457 struct ecommunity_val *eval)
458 {
459 struct irt_node *irt;
460 struct ecommunity_val eval_tmp;
461
462 /* If using "automatic" RT, we only care about the local-admin
463 * sub-field.
464 * This is to facilitate using VNI as the RT for EBGP peering too.
465 */
466 memcpy(&eval_tmp, eval, ECOMMUNITY_SIZE);
467 if (!is_import_rt_configured(vpn))
468 mask_ecom_global_admin(&eval_tmp, eval);
469
470 irt = lookup_import_rt(bgp, &eval_tmp);
471 if (irt)
472 if (is_vni_present_in_irt_vnis(irt->vnis, vpn))
473 /* Already mapped. */
474 return;
475
476 if (!irt) {
477 irt = import_rt_new(bgp, &eval_tmp);
478 assert(irt);
479 }
480
481 /* Add VNI to the hash list for this RT. */
482 listnode_add(irt->vnis, vpn);
483 }
484
485 /*
486 * Unmap specified VNI from specified RT. If there are no other
487 * VNIs for this RT, then the RT hash is deleted.
488 */
489 static void unmap_vni_from_rt(struct bgp *bgp, struct bgpevpn *vpn,
490 struct irt_node *irt)
491 {
492 /* Delete VNI from hash list for this RT. */
493 listnode_delete(irt->vnis, vpn);
494 if (!listnode_head(irt->vnis)) {
495 import_rt_free(bgp, irt);
496 }
497 }
498
499 /*
500 * Create RT extended community automatically from passed information:
501 * of the form AS:VNI.
502 * NOTE: We use only the lower 16 bits of the AS. This is sufficient as
503 * the need is to get a RT value that will be unique across different
504 * VNIs but the same across routers (in the same AS) for a particular
505 * VNI.
506 */
507 static void form_auto_rt(struct bgp *bgp, vni_t vni, struct list *rtl)
508 {
509 struct ecommunity_val eval;
510 struct ecommunity *ecomadd;
511
512 if (bgp->advertise_autort_rfc8365)
513 vni |= EVPN_AUTORT_VXLAN;
514 encode_route_target_as((bgp->as & 0xFFFF), vni, &eval);
515
516 ecomadd = ecommunity_new();
517 ecommunity_add_val(ecomadd, &eval);
518 listnode_add_sort(rtl, ecomadd);
519 }
520
521 /*
522 * Derive RD and RT for a VNI automatically. Invoked at the time of
523 * creation of a VNI.
524 */
525 static void derive_rd_rt_for_vni(struct bgp *bgp, struct bgpevpn *vpn)
526 {
527 bgp_evpn_derive_auto_rd(bgp, vpn);
528 bgp_evpn_derive_auto_rt_import(bgp, vpn);
529 bgp_evpn_derive_auto_rt_export(bgp, vpn);
530 }
531
532 /*
533 * Convert nexthop (remote VTEP IP) into an IPv6 address.
534 */
535 static void evpn_convert_nexthop_to_ipv6(struct attr *attr)
536 {
537 if (BGP_ATTR_NEXTHOP_AFI_IP6(attr))
538 return;
539 ipv4_to_ipv4_mapped_ipv6(&attr->mp_nexthop_global, attr->nexthop);
540 attr->mp_nexthop_len = IPV6_MAX_BYTELEN;
541 }
542
543 /*
544 * Add (update) or delete MACIP from zebra.
545 */
546 static int bgp_zebra_send_remote_macip(struct bgp *bgp, struct bgpevpn *vpn,
547 struct prefix_evpn *p,
548 struct in_addr remote_vtep_ip, int add,
549 uint8_t flags, uint32_t seq)
550 {
551 struct stream *s;
552 int ipa_len;
553 char buf1[ETHER_ADDR_STRLEN];
554 char buf2[INET6_ADDRSTRLEN];
555 char buf3[INET6_ADDRSTRLEN];
556
557 /* Check socket. */
558 if (!zclient || zclient->sock < 0)
559 return 0;
560
561 /* Don't try to register if Zebra doesn't know of this instance. */
562 if (!IS_BGP_INST_KNOWN_TO_ZEBRA(bgp)) {
563 if (BGP_DEBUG(zebra, ZEBRA))
564 zlog_debug("%s: No zebra instance to talk to, not installing remote macip",
565 __PRETTY_FUNCTION__);
566 return 0;
567 }
568 s = zclient->obuf;
569 stream_reset(s);
570
571 zclient_create_header(
572 s, add ? ZEBRA_REMOTE_MACIP_ADD : ZEBRA_REMOTE_MACIP_DEL,
573 bgp->vrf_id);
574 stream_putl(s, vpn->vni);
575 stream_put(s, &p->prefix.macip_addr.mac.octet, ETH_ALEN); /* Mac Addr */
576 /* IP address length and IP address, if any. */
577 if (is_evpn_prefix_ipaddr_none(p))
578 stream_putl(s, 0);
579 else {
580 ipa_len = is_evpn_prefix_ipaddr_v4(p) ? IPV4_MAX_BYTELEN
581 : IPV6_MAX_BYTELEN;
582 stream_putl(s, ipa_len);
583 stream_put(s, &p->prefix.macip_addr.ip.ip.addr, ipa_len);
584 }
585 stream_put_in_addr(s, &remote_vtep_ip);
586
587 /* TX flags - MAC sticky status and/or gateway mac */
588 /* Also TX the sequence number of the best route. */
589 if (add) {
590 stream_putc(s, flags);
591 stream_putl(s, seq);
592 }
593
594 stream_putw_at(s, 0, stream_get_endp(s));
595
596 if (bgp_debug_zebra(NULL))
597 zlog_debug(
598 "Tx %s MACIP, VNI %u MAC %s IP %s flags 0x%x seq %u remote VTEP %s",
599 add ? "ADD" : "DEL", vpn->vni,
600 prefix_mac2str(&p->prefix.macip_addr.mac,
601 buf1, sizeof(buf1)),
602 ipaddr2str(&p->prefix.macip_addr.ip,
603 buf3, sizeof(buf3)), flags, seq,
604 inet_ntop(AF_INET, &remote_vtep_ip, buf2,
605 sizeof(buf2)));
606
607 return zclient_send_message(zclient);
608 }
609
610 /*
611 * Add (update) or delete remote VTEP from zebra.
612 */
613 static int bgp_zebra_send_remote_vtep(struct bgp *bgp, struct bgpevpn *vpn,
614 struct prefix_evpn *p, int add)
615 {
616 struct stream *s;
617
618 /* Check socket. */
619 if (!zclient || zclient->sock < 0)
620 return 0;
621
622 /* Don't try to register if Zebra doesn't know of this instance. */
623 if (!IS_BGP_INST_KNOWN_TO_ZEBRA(bgp)) {
624 if (BGP_DEBUG(zebra, ZEBRA))
625 zlog_debug("%s: No zebra instance to talk to, not installing remote vtep",
626 __PRETTY_FUNCTION__);
627 return 0;
628 }
629
630 s = zclient->obuf;
631 stream_reset(s);
632
633 zclient_create_header(
634 s, add ? ZEBRA_REMOTE_VTEP_ADD : ZEBRA_REMOTE_VTEP_DEL,
635 bgp->vrf_id);
636 stream_putl(s, vpn->vni);
637 if (is_evpn_prefix_ipaddr_v4(p))
638 stream_put_in_addr(s, &p->prefix.imet_addr.ip.ipaddr_v4);
639 else if (is_evpn_prefix_ipaddr_v6(p)) {
640 flog_err(
641 EC_BGP_VTEP_INVALID,
642 "Bad remote IP when trying to %s remote VTEP for VNI %u",
643 add ? "ADD" : "DEL", vpn->vni);
644 return -1;
645 }
646
647 stream_putw_at(s, 0, stream_get_endp(s));
648
649 if (bgp_debug_zebra(NULL))
650 zlog_debug("Tx %s Remote VTEP, VNI %u remote VTEP %s",
651 add ? "ADD" : "DEL", vpn->vni,
652 inet_ntoa(p->prefix.imet_addr.ip.ipaddr_v4));
653
654 return zclient_send_message(zclient);
655 }
656
657 /*
658 * Build extended community for EVPN ES (type-4) route
659 */
660 static void build_evpn_type4_route_extcomm(struct evpnes *es,
661 struct attr *attr)
662 {
663 struct ecommunity ecom_encap;
664 struct ecommunity ecom_es_rt;
665 struct ecommunity_val eval;
666 struct ecommunity_val eval_es_rt;
667 bgp_encap_types tnl_type;
668 struct ethaddr mac;
669
670 /* Encap */
671 tnl_type = BGP_ENCAP_TYPE_VXLAN;
672 memset(&ecom_encap, 0, sizeof(ecom_encap));
673 encode_encap_extcomm(tnl_type, &eval);
674 ecom_encap.size = 1;
675 ecom_encap.val = (uint8_t *)eval.val;
676 attr->ecommunity = ecommunity_dup(&ecom_encap);
677
678 /* ES import RT */
679 memset(&mac, 0, sizeof(struct ethaddr));
680 memset(&ecom_es_rt, 0, sizeof(ecom_es_rt));
681 es_get_system_mac(&es->esi, &mac);
682 encode_es_rt_extcomm(&eval_es_rt, &mac);
683 ecom_es_rt.size = 1;
684 ecom_es_rt.val = (uint8_t *)eval_es_rt.val;
685 attr->ecommunity =
686 ecommunity_merge(attr->ecommunity, &ecom_es_rt);
687
688 attr->flag |= ATTR_FLAG_BIT(BGP_ATTR_EXT_COMMUNITIES);
689 }
690
691 /*
692 * Build extended communities for EVPN prefix route.
693 */
694 static void build_evpn_type5_route_extcomm(struct bgp *bgp_vrf,
695 struct attr *attr)
696 {
697 struct ecommunity ecom_encap;
698 struct ecommunity ecom_rmac;
699 struct ecommunity_val eval;
700 struct ecommunity_val eval_rmac;
701 bgp_encap_types tnl_type;
702 struct listnode *node, *nnode;
703 struct ecommunity *ecom;
704 struct list *vrf_export_rtl = NULL;
705
706 /* Encap */
707 tnl_type = BGP_ENCAP_TYPE_VXLAN;
708 memset(&ecom_encap, 0, sizeof(ecom_encap));
709 encode_encap_extcomm(tnl_type, &eval);
710 ecom_encap.size = 1;
711 ecom_encap.val = (uint8_t *)eval.val;
712
713 /* Add Encap */
714 attr->ecommunity = ecommunity_dup(&ecom_encap);
715
716 /* Add the export RTs for L3VNI/VRF */
717 vrf_export_rtl = bgp_vrf->vrf_export_rtl;
718 for (ALL_LIST_ELEMENTS(vrf_export_rtl, node, nnode, ecom))
719 attr->ecommunity =
720 ecommunity_merge(attr->ecommunity, ecom);
721
722 /* add the router mac extended community */
723 if (!is_zero_mac(&attr->rmac)) {
724 memset(&ecom_rmac, 0, sizeof(ecom_rmac));
725 encode_rmac_extcomm(&eval_rmac, &attr->rmac);
726 ecom_rmac.size = 1;
727 ecom_rmac.val = (uint8_t *)eval_rmac.val;
728 attr->ecommunity =
729 ecommunity_merge(attr->ecommunity, &ecom_rmac);
730 }
731
732 attr->flag |= ATTR_FLAG_BIT(BGP_ATTR_EXT_COMMUNITIES);
733 }
734
735 /*
736 * Build extended communities for EVPN route.
737 * This function is applicable for type-2 and type-3 routes. The layer-2 RT
738 * and ENCAP extended communities are applicable for all routes.
739 * The default gateway extended community and MAC mobility (sticky) extended
740 * community are added as needed based on passed settings - only for type-2
741 * routes. Likewise, the layer-3 RT and Router MAC extended communities are
742 * added, if present, based on passed settings - only for non-link-local
743 * type-2 routes.
744 */
745 static void build_evpn_route_extcomm(struct bgpevpn *vpn, struct attr *attr,
746 int add_l3_ecomm)
747 {
748 struct ecommunity ecom_encap;
749 struct ecommunity ecom_sticky;
750 struct ecommunity ecom_default_gw;
751 struct ecommunity ecom_rmac;
752 struct ecommunity ecom_na;
753 struct ecommunity_val eval;
754 struct ecommunity_val eval_sticky;
755 struct ecommunity_val eval_default_gw;
756 struct ecommunity_val eval_rmac;
757 struct ecommunity_val eval_na;
758
759 bgp_encap_types tnl_type;
760 struct listnode *node, *nnode;
761 struct ecommunity *ecom;
762 uint32_t seqnum;
763 struct list *vrf_export_rtl = NULL;
764
765 /* Encap */
766 tnl_type = BGP_ENCAP_TYPE_VXLAN;
767 memset(&ecom_encap, 0, sizeof(ecom_encap));
768 encode_encap_extcomm(tnl_type, &eval);
769 ecom_encap.size = 1;
770 ecom_encap.val = (uint8_t *)eval.val;
771
772 /* Add Encap */
773 attr->ecommunity = ecommunity_dup(&ecom_encap);
774
775 /* Add the export RTs for L2VNI */
776 for (ALL_LIST_ELEMENTS(vpn->export_rtl, node, nnode, ecom))
777 attr->ecommunity = ecommunity_merge(attr->ecommunity, ecom);
778
779 /* Add the export RTs for L3VNI if told to - caller determines
780 * when this should be done.
781 */
782 if (add_l3_ecomm) {
783 vrf_export_rtl = bgpevpn_get_vrf_export_rtl(vpn);
784 if (vrf_export_rtl && !list_isempty(vrf_export_rtl)) {
785 for (ALL_LIST_ELEMENTS(vrf_export_rtl, node, nnode,
786 ecom))
787 attr->ecommunity = ecommunity_merge(
788 attr->ecommunity, ecom);
789 }
790 }
791
792 /* Add MAC mobility (sticky) if needed. */
793 if (attr->sticky) {
794 seqnum = 0;
795 memset(&ecom_sticky, 0, sizeof(ecom_sticky));
796 encode_mac_mobility_extcomm(1, seqnum, &eval_sticky);
797 ecom_sticky.size = 1;
798 ecom_sticky.val = (uint8_t *)eval_sticky.val;
799 attr->ecommunity =
800 ecommunity_merge(attr->ecommunity, &ecom_sticky);
801 }
802
803 /* Add RMAC, if told to. */
804 if (add_l3_ecomm) {
805 memset(&ecom_rmac, 0, sizeof(ecom_rmac));
806 encode_rmac_extcomm(&eval_rmac, &attr->rmac);
807 ecom_rmac.size = 1;
808 ecom_rmac.val = (uint8_t *)eval_rmac.val;
809 attr->ecommunity =
810 ecommunity_merge(attr->ecommunity, &ecom_rmac);
811 }
812
813 /* Add default gateway, if needed. */
814 if (attr->default_gw) {
815 memset(&ecom_default_gw, 0, sizeof(ecom_default_gw));
816 encode_default_gw_extcomm(&eval_default_gw);
817 ecom_default_gw.size = 1;
818 ecom_default_gw.val = (uint8_t *)eval_default_gw.val;
819 attr->ecommunity =
820 ecommunity_merge(attr->ecommunity, &ecom_default_gw);
821 }
822
823 if (attr->router_flag) {
824 memset(&ecom_na, 0, sizeof(ecom_na));
825 encode_na_flag_extcomm(&eval_na, attr->router_flag);
826 ecom_na.size = 1;
827 ecom_na.val = (uint8_t *)eval_na.val;
828 attr->ecommunity = ecommunity_merge(attr->ecommunity,
829 &ecom_na);
830 }
831
832 attr->flag |= ATTR_FLAG_BIT(BGP_ATTR_EXT_COMMUNITIES);
833 }
834
835 /*
836 * Add MAC mobility extended community to attribute.
837 */
838 static void add_mac_mobility_to_attr(uint32_t seq_num, struct attr *attr)
839 {
840 struct ecommunity ecom_tmp;
841 struct ecommunity_val eval;
842 uint8_t *ecom_val_ptr;
843 int i;
844 uint8_t *pnt;
845 int type = 0;
846 int sub_type = 0;
847
848 /* Build MM */
849 encode_mac_mobility_extcomm(0, seq_num, &eval);
850
851 /* Find current MM ecommunity */
852 ecom_val_ptr = NULL;
853
854 if (attr->ecommunity) {
855 for (i = 0; i < attr->ecommunity->size; i++) {
856 pnt = attr->ecommunity->val + (i * 8);
857 type = *pnt++;
858 sub_type = *pnt++;
859
860 if (type == ECOMMUNITY_ENCODE_EVPN
861 && sub_type
862 == ECOMMUNITY_EVPN_SUBTYPE_MACMOBILITY) {
863 ecom_val_ptr = (uint8_t *)(attr->ecommunity->val
864 + (i * 8));
865 break;
866 }
867 }
868 }
869
870 /* Update the existing MM ecommunity */
871 if (ecom_val_ptr) {
872 memcpy(ecom_val_ptr, eval.val, sizeof(char) * ECOMMUNITY_SIZE);
873 }
874 /* Add MM to existing */
875 else {
876 memset(&ecom_tmp, 0, sizeof(ecom_tmp));
877 ecom_tmp.size = 1;
878 ecom_tmp.val = (uint8_t *)eval.val;
879
880 if (attr->ecommunity)
881 attr->ecommunity =
882 ecommunity_merge(attr->ecommunity, &ecom_tmp);
883 else
884 attr->ecommunity = ecommunity_dup(&ecom_tmp);
885 }
886 }
887
888 /* Install EVPN route into zebra. */
889 static int evpn_zebra_install(struct bgp *bgp, struct bgpevpn *vpn,
890 struct prefix_evpn *p,
891 struct in_addr remote_vtep_ip, uint8_t flags,
892 uint32_t seq)
893 {
894 int ret;
895
896 if (p->prefix.route_type == BGP_EVPN_MAC_IP_ROUTE)
897 ret = bgp_zebra_send_remote_macip(bgp, vpn, p, remote_vtep_ip,
898 1, flags, seq);
899 else
900 ret = bgp_zebra_send_remote_vtep(bgp, vpn, p, 1);
901
902 return ret;
903 }
904
905 /* Uninstall EVPN route from zebra. */
906 static int evpn_zebra_uninstall(struct bgp *bgp, struct bgpevpn *vpn,
907 struct prefix_evpn *p,
908 struct in_addr remote_vtep_ip)
909 {
910 int ret;
911
912 if (p->prefix.route_type == BGP_EVPN_MAC_IP_ROUTE)
913 ret = bgp_zebra_send_remote_macip(bgp, vpn, p, remote_vtep_ip,
914 0, 0, 0);
915 else
916 ret = bgp_zebra_send_remote_vtep(bgp, vpn, p, 0);
917
918 return ret;
919 }
920
921 /*
922 * Due to MAC mobility, the prior "local" best route has been supplanted
923 * by a "remote" best route. The prior route has to be deleted and withdrawn
924 * from peers.
925 */
926 static void evpn_delete_old_local_route(struct bgp *bgp, struct bgpevpn *vpn,
927 struct bgp_node *rn,
928 struct bgp_path_info *old_local)
929 {
930 struct bgp_node *global_rn;
931 struct bgp_path_info *pi;
932 afi_t afi = AFI_L2VPN;
933 safi_t safi = SAFI_EVPN;
934
935 /* Locate route node in the global EVPN routing table. Note that
936 * this table is a 2-level tree (RD-level + Prefix-level) similar to
937 * L3VPN routes.
938 */
939 global_rn = bgp_afi_node_lookup(bgp->rib[afi][safi], afi, safi,
940 (struct prefix *)&rn->p, &vpn->prd);
941 if (global_rn) {
942 /* Delete route entry in the global EVPN table. */
943 delete_evpn_route_entry(bgp, afi, safi, global_rn, &pi);
944
945 /* Schedule for processing - withdraws to peers happen from
946 * this table.
947 */
948 if (pi)
949 bgp_process(bgp, global_rn, afi, safi);
950 bgp_unlock_node(global_rn);
951 }
952
953 /* Delete route entry in the VNI route table, caller to remove. */
954 bgp_path_info_delete(rn, old_local);
955 }
956
957 static struct in_addr *es_vtep_new(struct in_addr vtep)
958 {
959 struct in_addr *ip;
960
961 ip = XCALLOC(MTYPE_BGP_EVPN_ES_VTEP, sizeof(struct in_addr));
962 if (!ip)
963 return NULL;
964
965 ip->s_addr = vtep.s_addr;
966 return ip;
967 }
968
969 static void es_vtep_free(struct in_addr *ip)
970 {
971 XFREE(MTYPE_BGP_EVPN_ES_VTEP, ip);
972 }
973
974 /* check if VTEP is already part of the list */
975 static int is_vtep_present_in_list(struct list *list,
976 struct in_addr vtep)
977 {
978 struct listnode *node = NULL;
979 struct in_addr *tmp;
980
981 for (ALL_LIST_ELEMENTS_RO(list, node, tmp)) {
982 if (tmp->s_addr == vtep.s_addr)
983 return 1;
984 }
985 return 0;
986 }
987
988 /*
989 * Best path for ES route was changed,
990 * update the list of VTEPs for this ES
991 */
992 static int evpn_es_install_vtep(struct bgp *bgp,
993 struct evpnes *es,
994 struct prefix_evpn *p,
995 struct in_addr rvtep)
996 {
997 struct in_addr *vtep_ip;
998
999 if (is_vtep_present_in_list(es->vtep_list, rvtep))
1000 return 0;
1001
1002
1003 vtep_ip = es_vtep_new(rvtep);
1004 if (vtep_ip)
1005 listnode_add_sort(es->vtep_list, vtep_ip);
1006 return 0;
1007 }
1008
1009 /*
1010 * Best path for ES route was changed,
1011 * update the list of VTEPs for this ES
1012 */
1013 static int evpn_es_uninstall_vtep(struct bgp *bgp,
1014 struct evpnes *es,
1015 struct prefix_evpn *p,
1016 struct in_addr rvtep)
1017 {
1018 struct listnode *node, *nnode, *node_to_del = NULL;
1019 struct in_addr *tmp;
1020
1021 for (ALL_LIST_ELEMENTS(es->vtep_list, node, nnode, tmp)) {
1022 if (tmp->s_addr == rvtep.s_addr) {
1023 es_vtep_free(tmp);
1024 node_to_del = node;
1025 }
1026 }
1027
1028 if (node_to_del)
1029 list_delete_node(es->vtep_list, node_to_del);
1030
1031 return 0;
1032 }
1033
1034 /*
1035 * Calculate the best path for a ES(type-4) route.
1036 */
1037 static int evpn_es_route_select_install(struct bgp *bgp,
1038 struct evpnes *es,
1039 struct bgp_node *rn)
1040 {
1041 int ret = 0;
1042 afi_t afi = AFI_L2VPN;
1043 safi_t safi = SAFI_EVPN;
1044 struct bgp_path_info *old_select; /* old best */
1045 struct bgp_path_info *new_select; /* new best */
1046 struct bgp_path_info_pair old_and_new;
1047
1048 /* Compute the best path. */
1049 bgp_best_selection(bgp, rn, &bgp->maxpaths[afi][safi],
1050 &old_and_new, afi, safi);
1051 old_select = old_and_new.old;
1052 new_select = old_and_new.new;
1053
1054 /*
1055 * If the best path hasn't changed - see if something needs to be
1056 * updated
1057 */
1058 if (old_select && old_select == new_select
1059 && old_select->type == ZEBRA_ROUTE_BGP
1060 && old_select->sub_type == BGP_ROUTE_IMPORTED
1061 && !CHECK_FLAG(rn->flags, BGP_NODE_USER_CLEAR)
1062 && !CHECK_FLAG(old_select->flags, BGP_PATH_ATTR_CHANGED)
1063 && !bgp_addpath_is_addpath_used(&bgp->tx_addpath, afi, safi)) {
1064 if (bgp_zebra_has_route_changed(rn, old_select)) {
1065 ret = evpn_es_install_vtep(bgp, es,
1066 (struct prefix_evpn *)&rn->p,
1067 old_select->attr->nexthop);
1068 }
1069 UNSET_FLAG(old_select->flags, BGP_PATH_MULTIPATH_CHG);
1070 bgp_zebra_clear_route_change_flags(rn);
1071 return ret;
1072 }
1073
1074 /* If the user did a "clear" this flag will be set */
1075 UNSET_FLAG(rn->flags, BGP_NODE_USER_CLEAR);
1076
1077 /*
1078 * bestpath has changed; update relevant fields and install or uninstall
1079 * into the zebra RIB.
1080 */
1081 if (old_select || new_select)
1082 bgp_bump_version(rn);
1083
1084 if (old_select)
1085 bgp_path_info_unset_flag(rn, old_select, BGP_PATH_SELECTED);
1086 if (new_select) {
1087 bgp_path_info_set_flag(rn, new_select, BGP_PATH_SELECTED);
1088 bgp_path_info_unset_flag(rn, new_select, BGP_PATH_ATTR_CHANGED);
1089 UNSET_FLAG(new_select->flags, BGP_PATH_MULTIPATH_CHG);
1090 }
1091
1092 if (new_select && new_select->type == ZEBRA_ROUTE_BGP
1093 && new_select->sub_type == BGP_ROUTE_IMPORTED) {
1094 ret = evpn_es_install_vtep(bgp, es,
1095 (struct prefix_evpn *)&rn->p,
1096 new_select->attr->nexthop);
1097 } else {
1098 if (old_select && old_select->type == ZEBRA_ROUTE_BGP
1099 && old_select->sub_type == BGP_ROUTE_IMPORTED)
1100 ret = evpn_es_uninstall_vtep(
1101 bgp, es, (struct prefix_evpn *)&rn->p,
1102 old_select->attr->nexthop);
1103 }
1104
1105 /* Clear any route change flags. */
1106 bgp_zebra_clear_route_change_flags(rn);
1107
1108 /* Reap old select bgp_path_info, if it has been removed */
1109 if (old_select && CHECK_FLAG(old_select->flags, BGP_PATH_REMOVED))
1110 bgp_path_info_reap(rn, old_select);
1111
1112 return ret;
1113 }
1114
1115 /*
1116 * Calculate the best path for an EVPN route. Install/update best path in zebra,
1117 * if appropriate.
1118 */
1119 static int evpn_route_select_install(struct bgp *bgp, struct bgpevpn *vpn,
1120 struct bgp_node *rn)
1121 {
1122 struct bgp_path_info *old_select, *new_select;
1123 struct bgp_path_info_pair old_and_new;
1124 struct prefix_evpn *evp;
1125 afi_t afi = AFI_L2VPN;
1126 safi_t safi = SAFI_EVPN;
1127 int ret = 0;
1128 uint8_t flags = 0;
1129
1130 /* Compute the best path. */
1131 bgp_best_selection(bgp, rn, &bgp->maxpaths[afi][safi], &old_and_new,
1132 afi, safi);
1133 old_select = old_and_new.old;
1134 new_select = old_and_new.new;
1135
1136 evp = (struct prefix_evpn *)&rn->p;
1137 /* If the best path hasn't changed - see if there is still something to
1138 * update
1139 * to zebra RIB.
1140 */
1141 if (old_select && old_select == new_select
1142 && old_select->type == ZEBRA_ROUTE_BGP
1143 && old_select->sub_type == BGP_ROUTE_IMPORTED
1144 && !CHECK_FLAG(rn->flags, BGP_NODE_USER_CLEAR)
1145 && !CHECK_FLAG(old_select->flags, BGP_PATH_ATTR_CHANGED)
1146 && !bgp_addpath_is_addpath_used(&bgp->tx_addpath, afi, safi)) {
1147 if (bgp_zebra_has_route_changed(rn, old_select)) {
1148 if (old_select->attr->sticky)
1149 SET_FLAG(flags, ZEBRA_MACIP_TYPE_STICKY);
1150 if (old_select->attr->default_gw)
1151 SET_FLAG(flags, ZEBRA_MACIP_TYPE_GW);
1152 if (is_evpn_prefix_ipaddr_v6(evp) &&
1153 old_select->attr->router_flag)
1154 SET_FLAG(flags, ZEBRA_MACIP_TYPE_ROUTER_FLAG);
1155
1156 ret = evpn_zebra_install(
1157 bgp, vpn, (struct prefix_evpn *)&rn->p,
1158 old_select->attr->nexthop, flags,
1159 mac_mobility_seqnum(old_select->attr));
1160 }
1161 UNSET_FLAG(old_select->flags, BGP_PATH_MULTIPATH_CHG);
1162 bgp_zebra_clear_route_change_flags(rn);
1163 return ret;
1164 }
1165
1166 /* If the user did a "clear" this flag will be set */
1167 UNSET_FLAG(rn->flags, BGP_NODE_USER_CLEAR);
1168
1169 /* bestpath has changed; update relevant fields and install or uninstall
1170 * into the zebra RIB.
1171 */
1172 if (old_select || new_select)
1173 bgp_bump_version(rn);
1174
1175 if (old_select)
1176 bgp_path_info_unset_flag(rn, old_select, BGP_PATH_SELECTED);
1177 if (new_select) {
1178 bgp_path_info_set_flag(rn, new_select, BGP_PATH_SELECTED);
1179 bgp_path_info_unset_flag(rn, new_select, BGP_PATH_ATTR_CHANGED);
1180 UNSET_FLAG(new_select->flags, BGP_PATH_MULTIPATH_CHG);
1181 }
1182
1183 if (new_select && new_select->type == ZEBRA_ROUTE_BGP
1184 && new_select->sub_type == BGP_ROUTE_IMPORTED) {
1185 flags = 0;
1186 if (new_select->attr->sticky)
1187 SET_FLAG(flags, ZEBRA_MACIP_TYPE_STICKY);
1188 if (new_select->attr->default_gw)
1189 SET_FLAG(flags, ZEBRA_MACIP_TYPE_GW);
1190 if (is_evpn_prefix_ipaddr_v6(evp) &&
1191 new_select->attr->router_flag)
1192 SET_FLAG(flags, ZEBRA_MACIP_TYPE_ROUTER_FLAG);
1193
1194 ret = evpn_zebra_install(bgp, vpn, (struct prefix_evpn *)&rn->p,
1195 new_select->attr->nexthop, flags,
1196 mac_mobility_seqnum(new_select->attr));
1197 /* If an old best existed and it was a "local" route, the only
1198 * reason
1199 * it would be supplanted is due to MAC mobility procedures. So,
1200 * we
1201 * need to do an implicit delete and withdraw that route from
1202 * peers.
1203 */
1204 if (old_select && old_select->peer == bgp->peer_self
1205 && old_select->type == ZEBRA_ROUTE_BGP
1206 && old_select->sub_type == BGP_ROUTE_STATIC)
1207 evpn_delete_old_local_route(bgp, vpn, rn, old_select);
1208 } else {
1209 if (old_select && old_select->type == ZEBRA_ROUTE_BGP
1210 && old_select->sub_type == BGP_ROUTE_IMPORTED)
1211 ret = evpn_zebra_uninstall(bgp, vpn,
1212 (struct prefix_evpn *)&rn->p,
1213 old_select->attr->nexthop);
1214 }
1215
1216 /* Clear any route change flags. */
1217 bgp_zebra_clear_route_change_flags(rn);
1218
1219 /* Reap old select bgp_path_info, if it has been removed */
1220 if (old_select && CHECK_FLAG(old_select->flags, BGP_PATH_REMOVED))
1221 bgp_path_info_reap(rn, old_select);
1222
1223 return ret;
1224 }
1225
1226 /*
1227 * Return true if the local ri for this rn is of type gateway mac
1228 */
1229 static int evpn_route_is_def_gw(struct bgp *bgp, struct bgp_node *rn)
1230 {
1231 struct bgp_path_info *tmp_pi = NULL;
1232 struct bgp_path_info *local_pi = NULL;
1233
1234 local_pi = NULL;
1235 for (tmp_pi = rn->info; tmp_pi; tmp_pi = tmp_pi->next) {
1236 if (tmp_pi->peer == bgp->peer_self
1237 && tmp_pi->type == ZEBRA_ROUTE_BGP
1238 && tmp_pi->sub_type == BGP_ROUTE_STATIC)
1239 local_pi = tmp_pi;
1240 }
1241
1242 if (!local_pi)
1243 return 0;
1244
1245 return local_pi->attr->default_gw;
1246 }
1247
1248
1249 /*
1250 * Return true if the local ri for this rn has sticky set
1251 */
1252 static int evpn_route_is_sticky(struct bgp *bgp, struct bgp_node *rn)
1253 {
1254 struct bgp_path_info *tmp_pi;
1255 struct bgp_path_info *local_pi;
1256
1257 local_pi = NULL;
1258 for (tmp_pi = rn->info; tmp_pi; tmp_pi = tmp_pi->next) {
1259 if (tmp_pi->peer == bgp->peer_self
1260 && tmp_pi->type == ZEBRA_ROUTE_BGP
1261 && tmp_pi->sub_type == BGP_ROUTE_STATIC)
1262 local_pi = tmp_pi;
1263 }
1264
1265 if (!local_pi)
1266 return 0;
1267
1268 return local_pi->attr->sticky;
1269 }
1270
1271 /*
1272 * create or update EVPN type4 route entry.
1273 * This could be in the ES table or the global table.
1274 * TODO: handle remote ES (type4) routes as well
1275 */
1276 static int update_evpn_type4_route_entry(struct bgp *bgp, struct evpnes *es,
1277 afi_t afi, safi_t safi,
1278 struct bgp_node *rn, struct attr *attr,
1279 int add, struct bgp_path_info **ri,
1280 int *route_changed)
1281 {
1282 char buf[ESI_STR_LEN];
1283 char buf1[INET6_ADDRSTRLEN];
1284 struct bgp_path_info *tmp_pi = NULL;
1285 struct bgp_path_info *local_pi = NULL; /* local route entry if any */
1286 struct bgp_path_info *remote_pi = NULL; /* remote route entry if any */
1287 struct attr *attr_new = NULL;
1288 struct prefix_evpn *evp = NULL;
1289
1290 *ri = NULL;
1291 *route_changed = 1;
1292 evp = (struct prefix_evpn *)&rn->p;
1293
1294 /* locate the local and remote entries if any */
1295 for (tmp_pi = rn->info; tmp_pi; tmp_pi = tmp_pi->next) {
1296 if (tmp_pi->peer == bgp->peer_self
1297 && tmp_pi->type == ZEBRA_ROUTE_BGP
1298 && tmp_pi->sub_type == BGP_ROUTE_STATIC)
1299 local_pi = tmp_pi;
1300 if (tmp_pi->type == ZEBRA_ROUTE_BGP
1301 && tmp_pi->sub_type == BGP_ROUTE_IMPORTED
1302 && CHECK_FLAG(tmp_pi->flags, BGP_PATH_VALID))
1303 remote_pi = tmp_pi;
1304 }
1305
1306 /* we don't expect to see a remote_pi at this point.
1307 * An ES route has esi + vtep_ip as the key,
1308 * We shouldn't see the same route from any other vtep.
1309 */
1310 if (remote_pi) {
1311 flog_err(
1312 EC_BGP_ES_INVALID,
1313 "%u ERROR: local es route for ESI: %s Vtep %s also learnt from remote",
1314 bgp->vrf_id,
1315 esi_to_str(&evp->prefix.es_addr.esi, buf, sizeof(buf)),
1316 ipaddr2str(&es->originator_ip, buf1, sizeof(buf1)));
1317 return -1;
1318 }
1319
1320 if (!local_pi && !add)
1321 return 0;
1322
1323 /* create or update the entry */
1324 if (!local_pi) {
1325
1326 /* Add or update attribute to hash */
1327 attr_new = bgp_attr_intern(attr);
1328
1329 /* Create new route with its attribute. */
1330 tmp_pi = info_make(ZEBRA_ROUTE_BGP, BGP_ROUTE_STATIC, 0,
1331 bgp->peer_self, attr_new, rn);
1332 SET_FLAG(tmp_pi->flags, BGP_PATH_VALID);
1333
1334 /* add the newly created path to the route-node */
1335 bgp_path_info_add(rn, tmp_pi);
1336 } else {
1337 tmp_pi = local_pi;
1338 if (attrhash_cmp(tmp_pi->attr, attr)
1339 && !CHECK_FLAG(tmp_pi->flags, BGP_PATH_REMOVED))
1340 *route_changed = 0;
1341 else {
1342 /* The attribute has changed.
1343 * Add (or update) attribute to hash. */
1344 attr_new = bgp_attr_intern(attr);
1345 bgp_path_info_set_flag(rn, tmp_pi,
1346 BGP_PATH_ATTR_CHANGED);
1347
1348 /* Restore route, if needed. */
1349 if (CHECK_FLAG(tmp_pi->flags, BGP_PATH_REMOVED))
1350 bgp_path_info_restore(rn, tmp_pi);
1351
1352 /* Unintern existing, set to new. */
1353 bgp_attr_unintern(&tmp_pi->attr);
1354 tmp_pi->attr = attr_new;
1355 tmp_pi->uptime = bgp_clock();
1356 }
1357 }
1358
1359 /* Return back the route entry. */
1360 *ri = tmp_pi;
1361 return 0;
1362 }
1363
1364 /* update evpn es (type-4) route */
1365 static int update_evpn_type4_route(struct bgp *bgp,
1366 struct evpnes *es,
1367 struct prefix_evpn *p)
1368 {
1369 int ret = 0;
1370 int route_changed = 0;
1371 char buf[ESI_STR_LEN];
1372 char buf1[INET6_ADDRSTRLEN];
1373 afi_t afi = AFI_L2VPN;
1374 safi_t safi = SAFI_EVPN;
1375 struct attr attr;
1376 struct attr *attr_new = NULL;
1377 struct bgp_node *rn = NULL;
1378 struct bgp_path_info *pi = NULL;
1379
1380 memset(&attr, 0, sizeof(struct attr));
1381
1382 /* Build path-attribute for this route. */
1383 bgp_attr_default_set(&attr, BGP_ORIGIN_IGP);
1384 attr.nexthop = es->originator_ip.ipaddr_v4;
1385 attr.mp_nexthop_global_in = es->originator_ip.ipaddr_v4;
1386 attr.mp_nexthop_len = BGP_ATTR_NHLEN_IPV4;
1387
1388 /* Set up extended community. */
1389 build_evpn_type4_route_extcomm(es, &attr);
1390
1391 /* First, create (or fetch) route node within the ESI. */
1392 /* NOTE: There is no RD here. */
1393 rn = bgp_node_get(es->route_table, (struct prefix *)p);
1394
1395 /* Create or update route entry. */
1396 ret = update_evpn_type4_route_entry(bgp, es, afi, safi, rn, &attr, 1,
1397 &pi, &route_changed);
1398 if (ret != 0) {
1399 flog_err(EC_BGP_ES_INVALID,
1400 "%u ERROR: Failed to updated ES route ESI: %s VTEP %s",
1401 bgp->vrf_id,
1402 esi_to_str(&p->prefix.es_addr.esi, buf, sizeof(buf)),
1403 ipaddr2str(&es->originator_ip, buf1, sizeof(buf1)));
1404 }
1405
1406 assert(pi);
1407 attr_new = pi->attr;
1408
1409 /* Perform route selection;
1410 * this is just to set the flags correctly
1411 * as local route in the ES always wins.
1412 */
1413 evpn_es_route_select_install(bgp, es, rn);
1414 bgp_unlock_node(rn);
1415
1416 /* If this is a new route or some attribute has changed, export the
1417 * route to the global table. The route will be advertised to peers
1418 * from there. Note that this table is a 2-level tree (RD-level +
1419 * Prefix-level) similar to L3VPN routes.
1420 */
1421 if (route_changed) {
1422 struct bgp_path_info *global_pi;
1423
1424 rn = bgp_afi_node_get(bgp->rib[afi][safi], afi, safi,
1425 (struct prefix *)p, &es->prd);
1426 update_evpn_type4_route_entry(bgp, es, afi, safi, rn, attr_new,
1427 1, &global_pi, &route_changed);
1428
1429 /* Schedule for processing and unlock node. */
1430 bgp_process(bgp, rn, afi, safi);
1431 bgp_unlock_node(rn);
1432 }
1433
1434 /* Unintern temporary. */
1435 aspath_unintern(&attr.aspath);
1436 return 0;
1437 }
1438
1439 static int update_evpn_type5_route_entry(struct bgp *bgp_def,
1440 struct bgp *bgp_vrf, afi_t afi,
1441 safi_t safi, struct bgp_node *rn,
1442 struct attr *attr, int *route_changed)
1443 {
1444 struct attr *attr_new = NULL;
1445 struct bgp_path_info *pi = NULL;
1446 mpls_label_t label = MPLS_INVALID_LABEL;
1447 struct bgp_path_info *local_pi = NULL;
1448 struct bgp_path_info *tmp_pi = NULL;
1449
1450 *route_changed = 0;
1451 /* locate the local route entry if any */
1452 for (tmp_pi = rn->info; tmp_pi; tmp_pi = tmp_pi->next) {
1453 if (tmp_pi->peer == bgp_def->peer_self
1454 && tmp_pi->type == ZEBRA_ROUTE_BGP
1455 && tmp_pi->sub_type == BGP_ROUTE_STATIC)
1456 local_pi = tmp_pi;
1457 }
1458
1459 /*
1460 * create a new route entry if one doesn't exist.
1461 * Otherwise see if route attr has changed
1462 */
1463 if (!local_pi) {
1464
1465 /* route has changed as this is the first entry */
1466 *route_changed = 1;
1467
1468 /* Add (or update) attribute to hash. */
1469 attr_new = bgp_attr_intern(attr);
1470
1471 /* create the route info from attribute */
1472 pi = info_make(ZEBRA_ROUTE_BGP, BGP_ROUTE_STATIC, 0,
1473 bgp_def->peer_self, attr_new, rn);
1474 SET_FLAG(pi->flags, BGP_PATH_VALID);
1475
1476 /* Type-5 routes advertise the L3-VNI */
1477 bgp_path_info_extra_get(pi);
1478 vni2label(bgp_vrf->l3vni, &label);
1479 memcpy(&pi->extra->label, &label, sizeof(label));
1480 pi->extra->num_labels = 1;
1481
1482 /* add the route entry to route node*/
1483 bgp_path_info_add(rn, pi);
1484 } else {
1485
1486 tmp_pi = local_pi;
1487 if (!attrhash_cmp(tmp_pi->attr, attr)) {
1488
1489 /* attribute changed */
1490 *route_changed = 1;
1491
1492 /* The attribute has changed. */
1493 /* Add (or update) attribute to hash. */
1494 attr_new = bgp_attr_intern(attr);
1495 bgp_path_info_set_flag(rn, tmp_pi,
1496 BGP_PATH_ATTR_CHANGED);
1497
1498 /* Restore route, if needed. */
1499 if (CHECK_FLAG(tmp_pi->flags, BGP_PATH_REMOVED))
1500 bgp_path_info_restore(rn, tmp_pi);
1501
1502 /* Unintern existing, set to new. */
1503 bgp_attr_unintern(&tmp_pi->attr);
1504 tmp_pi->attr = attr_new;
1505 tmp_pi->uptime = bgp_clock();
1506 }
1507 }
1508 return 0;
1509 }
1510
1511 /* update evpn type-5 route entry */
1512 static int update_evpn_type5_route(struct bgp *bgp_vrf, struct prefix_evpn *evp,
1513 struct attr *src_attr)
1514 {
1515 afi_t afi = AFI_L2VPN;
1516 safi_t safi = SAFI_EVPN;
1517 struct attr attr;
1518 struct bgp_node *rn = NULL;
1519 struct bgp *bgp_def = NULL;
1520 int route_changed = 0;
1521
1522 bgp_def = bgp_get_default();
1523 if (!bgp_def)
1524 return 0;
1525
1526 /* Build path attribute for this route - use the source attr, if
1527 * present, else treat as locally originated.
1528 */
1529 if (src_attr)
1530 bgp_attr_dup(&attr, src_attr);
1531 else {
1532 memset(&attr, 0, sizeof(struct attr));
1533 bgp_attr_default_set(&attr, BGP_ORIGIN_IGP);
1534 }
1535 /* Set nexthop to ourselves and fill in the Router MAC. */
1536 attr.nexthop = bgp_vrf->originator_ip;
1537 attr.mp_nexthop_global_in = bgp_vrf->originator_ip;
1538 attr.mp_nexthop_len = BGP_ATTR_NHLEN_IPV4;
1539 memcpy(&attr.rmac, &bgp_vrf->rmac, sizeof(struct ethaddr));
1540
1541 /* Setup RT and encap extended community */
1542 build_evpn_type5_route_extcomm(bgp_vrf, &attr);
1543
1544 /* get the route node in global table */
1545 rn = bgp_afi_node_get(bgp_def->rib[afi][safi], afi, safi,
1546 (struct prefix *)evp, &bgp_vrf->vrf_prd);
1547 assert(rn);
1548
1549 /* create or update the route entry within the route node */
1550 update_evpn_type5_route_entry(bgp_def, bgp_vrf, afi, safi, rn, &attr,
1551 &route_changed);
1552
1553 /* schedule for processing and unlock node */
1554 if (route_changed) {
1555 bgp_process(bgp_def, rn, afi, safi);
1556 bgp_unlock_node(rn);
1557 }
1558
1559 /* uninten temporary */
1560 if (!src_attr)
1561 aspath_unintern(&attr.aspath);
1562 return 0;
1563 }
1564
1565 /*
1566 * Create or update EVPN route entry. This could be in the VNI route table
1567 * or the global route table.
1568 */
1569 static int update_evpn_route_entry(struct bgp *bgp, struct bgpevpn *vpn,
1570 afi_t afi, safi_t safi, struct bgp_node *rn,
1571 struct attr *attr, int add,
1572 struct bgp_path_info **pi, uint8_t flags,
1573 uint32_t seq)
1574 {
1575 struct bgp_path_info *tmp_pi;
1576 struct bgp_path_info *local_pi;
1577 struct attr *attr_new;
1578 mpls_label_t label[BGP_MAX_LABELS];
1579 uint32_t num_labels = 1;
1580 int route_change = 1;
1581 uint8_t sticky = 0;
1582 struct prefix_evpn *evp;
1583
1584 *pi = NULL;
1585 evp = (struct prefix_evpn *)&rn->p;
1586 memset(&label, 0, sizeof(label));
1587
1588 /* See if this is an update of an existing route, or a new add. */
1589 local_pi = NULL;
1590 for (tmp_pi = rn->info; tmp_pi; tmp_pi = tmp_pi->next) {
1591 if (tmp_pi->peer == bgp->peer_self
1592 && tmp_pi->type == ZEBRA_ROUTE_BGP
1593 && tmp_pi->sub_type == BGP_ROUTE_STATIC)
1594 local_pi = tmp_pi;
1595 }
1596
1597 /* If route doesn't exist already, create a new one, if told to.
1598 * Otherwise act based on whether the attributes of the route have
1599 * changed or not.
1600 */
1601 if (!local_pi && !add)
1602 return 0;
1603
1604 /* For non-GW MACs, update MAC mobility seq number, if needed. */
1605 if (seq && !CHECK_FLAG(flags, ZEBRA_MACIP_TYPE_GW))
1606 add_mac_mobility_to_attr(seq, attr);
1607
1608 if (!local_pi) {
1609 /* Add (or update) attribute to hash. */
1610 attr_new = bgp_attr_intern(attr);
1611
1612 /* Extract MAC mobility sequence number, if any. */
1613 attr_new->mm_seqnum =
1614 bgp_attr_mac_mobility_seqnum(attr_new, &sticky);
1615 attr_new->sticky = sticky;
1616
1617 /* Create new route with its attribute. */
1618 tmp_pi = info_make(ZEBRA_ROUTE_BGP, BGP_ROUTE_STATIC, 0,
1619 bgp->peer_self, attr_new, rn);
1620 SET_FLAG(tmp_pi->flags, BGP_PATH_VALID);
1621 bgp_path_info_extra_get(tmp_pi);
1622
1623 /* The VNI goes into the 'label' field of the route */
1624 vni2label(vpn->vni, &label[0]);
1625
1626 /* Type-2 routes may carry a second VNI - the L3-VNI.
1627 * Only attach second label if we are advertising two labels for
1628 * type-2 routes.
1629 */
1630 if (evp->prefix.route_type == BGP_EVPN_MAC_IP_ROUTE
1631 && CHECK_FLAG(vpn->flags, VNI_FLAG_USE_TWO_LABELS)) {
1632 vni_t l3vni;
1633
1634 l3vni = bgpevpn_get_l3vni(vpn);
1635 if (l3vni) {
1636 vni2label(l3vni, &label[1]);
1637 num_labels++;
1638 }
1639 }
1640
1641 memcpy(&tmp_pi->extra->label, label, sizeof(label));
1642 tmp_pi->extra->num_labels = num_labels;
1643 bgp_path_info_add(rn, tmp_pi);
1644 } else {
1645 tmp_pi = local_pi;
1646 if (attrhash_cmp(tmp_pi->attr, attr)
1647 && !CHECK_FLAG(tmp_pi->flags, BGP_PATH_REMOVED))
1648 route_change = 0;
1649 else {
1650 /*
1651 * The attributes have changed, type-2 routes needs to
1652 * be advertised with right labels.
1653 */
1654 vni2label(vpn->vni, &label[0]);
1655 if (evp->prefix.route_type == BGP_EVPN_MAC_IP_ROUTE
1656 && CHECK_FLAG(vpn->flags,
1657 VNI_FLAG_USE_TWO_LABELS)) {
1658 vni_t l3vni;
1659
1660 l3vni = bgpevpn_get_l3vni(vpn);
1661 if (l3vni) {
1662 vni2label(l3vni, &label[1]);
1663 num_labels++;
1664 }
1665 }
1666 memcpy(&tmp_pi->extra->label, label, sizeof(label));
1667 tmp_pi->extra->num_labels = num_labels;
1668
1669 /* The attribute has changed. */
1670 /* Add (or update) attribute to hash. */
1671 attr_new = bgp_attr_intern(attr);
1672 bgp_path_info_set_flag(rn, tmp_pi,
1673 BGP_PATH_ATTR_CHANGED);
1674
1675 /* Extract MAC mobility sequence number, if any. */
1676 attr_new->mm_seqnum =
1677 bgp_attr_mac_mobility_seqnum(attr_new, &sticky);
1678 attr_new->sticky = sticky;
1679
1680 /* Restore route, if needed. */
1681 if (CHECK_FLAG(tmp_pi->flags, BGP_PATH_REMOVED))
1682 bgp_path_info_restore(rn, tmp_pi);
1683
1684 /* Unintern existing, set to new. */
1685 bgp_attr_unintern(&tmp_pi->attr);
1686 tmp_pi->attr = attr_new;
1687 tmp_pi->uptime = bgp_clock();
1688 }
1689 }
1690
1691 /* Return back the route entry. */
1692 *pi = tmp_pi;
1693 return route_change;
1694 }
1695
1696 /*
1697 * If the local route was not selected evict it and tell zebra to re-add
1698 * the best remote dest.
1699 *
1700 * Typically a local path added by zebra is expected to be selected as
1701 * best. In which case when a remote path wins as best (later)
1702 * evpn_route_select_install itself evicts the older-local-best path.
1703 *
1704 * However if bgp's add and zebra's add cross paths (race condition) it
1705 * is possible that the local path is no longer the "older" best path.
1706 * It is a path that was never designated as best and hence requires
1707 * additional handling to prevent bgp from injecting and holding on to a
1708 * non-best local path.
1709 */
1710 static void evpn_cleanup_local_non_best_route(struct bgp *bgp,
1711 struct bgpevpn *vpn,
1712 struct bgp_node *rn,
1713 struct bgp_path_info *local_pi)
1714 {
1715 struct bgp_path_info *tmp_pi;
1716 struct bgp_path_info *curr_select = NULL;
1717 uint8_t flags = 0;
1718 char buf[PREFIX_STRLEN];
1719
1720 /* local path was not picked as the winner; kick it out */
1721 if (bgp_debug_zebra(NULL)) {
1722 zlog_debug("evicting local evpn prefix %s as remote won",
1723 prefix2str(&rn->p, buf, sizeof(buf)));
1724 }
1725 evpn_delete_old_local_route(bgp, vpn, rn, local_pi);
1726 bgp_path_info_reap(rn, local_pi);
1727
1728 /* tell zebra to re-add the best remote path */
1729 for (tmp_pi = rn->info; tmp_pi; tmp_pi = tmp_pi->next) {
1730 if (CHECK_FLAG(tmp_pi->flags, BGP_PATH_SELECTED)) {
1731 curr_select = tmp_pi;
1732 break;
1733 }
1734 }
1735 if (curr_select &&
1736 curr_select->type == ZEBRA_ROUTE_BGP
1737 && curr_select->sub_type == BGP_ROUTE_IMPORTED) {
1738 if (curr_select->attr->sticky)
1739 SET_FLAG(flags, ZEBRA_MACIP_TYPE_STICKY);
1740 if (curr_select->attr->default_gw)
1741 SET_FLAG(flags, ZEBRA_MACIP_TYPE_GW);
1742 evpn_zebra_install(bgp, vpn, (struct prefix_evpn *)&rn->p,
1743 curr_select->attr->nexthop, flags,
1744 mac_mobility_seqnum(curr_select->attr));
1745 }
1746 }
1747
1748 /*
1749 * Create or update EVPN route (of type based on prefix) for specified VNI
1750 * and schedule for processing.
1751 */
1752 static int update_evpn_route(struct bgp *bgp, struct bgpevpn *vpn,
1753 struct prefix_evpn *p, uint8_t flags,
1754 uint32_t seq)
1755 {
1756 struct bgp_node *rn;
1757 struct attr attr;
1758 struct attr *attr_new;
1759 int add_l3_ecomm = 0;
1760 struct bgp_path_info *pi;
1761 afi_t afi = AFI_L2VPN;
1762 safi_t safi = SAFI_EVPN;
1763 int route_change;
1764
1765 memset(&attr, 0, sizeof(struct attr));
1766
1767 /* Build path-attribute for this route. */
1768 bgp_attr_default_set(&attr, BGP_ORIGIN_IGP);
1769 attr.nexthop = vpn->originator_ip;
1770 attr.mp_nexthop_global_in = vpn->originator_ip;
1771 attr.mp_nexthop_len = BGP_ATTR_NHLEN_IPV4;
1772 attr.sticky = CHECK_FLAG(flags, ZEBRA_MACIP_TYPE_STICKY) ? 1 : 0;
1773 attr.default_gw = CHECK_FLAG(flags, ZEBRA_MACIP_TYPE_GW) ? 1 : 0;
1774 attr.router_flag = CHECK_FLAG(flags,
1775 ZEBRA_MACIP_TYPE_ROUTER_FLAG) ? 1 : 0;
1776
1777 /* PMSI is only needed for type-3 routes */
1778 if (p->prefix.route_type == BGP_EVPN_IMET_ROUTE)
1779 attr.flag |= ATTR_FLAG_BIT(BGP_ATTR_PMSI_TUNNEL);
1780
1781 /* router mac is only needed for type-2 routes here. */
1782 if (p->prefix.route_type == BGP_EVPN_MAC_IP_ROUTE)
1783 bgpevpn_get_rmac(vpn, &attr.rmac);
1784 vni2label(vpn->vni, &(attr.label));
1785
1786 /* Include L3 VNI related RTs and RMAC for type-2 routes, if they're
1787 * IPv4 or IPv6 global addresses and we're advertising L3VNI with
1788 * these routes.
1789 */
1790 if (p->prefix.route_type == BGP_EVPN_MAC_IP_ROUTE &&
1791 (is_evpn_prefix_ipaddr_v4(p) ||
1792 !IN6_IS_ADDR_LINKLOCAL(&p->prefix.macip_addr.ip.ipaddr_v6)) &&
1793 CHECK_FLAG(vpn->flags, VNI_FLAG_USE_TWO_LABELS) &&
1794 bgpevpn_get_l3vni(vpn))
1795 add_l3_ecomm = 1;
1796
1797 /* Set up extended community. */
1798 build_evpn_route_extcomm(vpn, &attr, add_l3_ecomm);
1799
1800 /* First, create (or fetch) route node within the VNI. */
1801 /* NOTE: There is no RD here. */
1802 rn = bgp_node_get(vpn->route_table, (struct prefix *)p);
1803
1804 /* Create or update route entry. */
1805 route_change = update_evpn_route_entry(bgp, vpn, afi, safi, rn, &attr,
1806 1, &pi, flags, seq);
1807 assert(pi);
1808 attr_new = pi->attr;
1809
1810 /* lock ri to prevent freeing in evpn_route_select_install */
1811 bgp_path_info_lock(pi);
1812 /* Perform route selection; this is just to set the flags correctly
1813 * as local route in the VNI always wins.
1814 */
1815 evpn_route_select_install(bgp, vpn, rn);
1816 /*
1817 * If the new local route was not selected evict it and tell zebra
1818 * to re-add the best remote dest. BGP doesn't retain non-best local
1819 * routes.
1820 */
1821 if (!CHECK_FLAG(pi->flags, BGP_PATH_SELECTED)) {
1822 route_change = 0;
1823 evpn_cleanup_local_non_best_route(bgp, vpn, rn, pi);
1824 }
1825 bgp_path_info_unlock(pi);
1826
1827 bgp_unlock_node(rn);
1828
1829 /* If this is a new route or some attribute has changed, export the
1830 * route to the global table. The route will be advertised to peers
1831 * from there. Note that this table is a 2-level tree (RD-level +
1832 * Prefix-level) similar to L3VPN routes.
1833 */
1834 if (route_change) {
1835 struct bgp_path_info *global_pi;
1836
1837 rn = bgp_afi_node_get(bgp->rib[afi][safi], afi, safi,
1838 (struct prefix *)p, &vpn->prd);
1839 update_evpn_route_entry(bgp, vpn, afi, safi, rn, attr_new, 1,
1840 &global_pi, flags, seq);
1841
1842 /* Schedule for processing and unlock node. */
1843 bgp_process(bgp, rn, afi, safi);
1844 bgp_unlock_node(rn);
1845 }
1846
1847 /* Unintern temporary. */
1848 aspath_unintern(&attr.aspath);
1849
1850 return 0;
1851 }
1852
1853 /*
1854 * Delete EVPN route entry.
1855 * The entry can be in ESI/VNI table or the global table.
1856 */
1857 static void delete_evpn_route_entry(struct bgp *bgp, afi_t afi, safi_t safi,
1858 struct bgp_node *rn,
1859 struct bgp_path_info **pi)
1860 {
1861 struct bgp_path_info *tmp_pi;
1862
1863 *pi = NULL;
1864
1865 /* Now, find matching route. */
1866 for (tmp_pi = rn->info; tmp_pi; tmp_pi = tmp_pi->next)
1867 if (tmp_pi->peer == bgp->peer_self
1868 && tmp_pi->type == ZEBRA_ROUTE_BGP
1869 && tmp_pi->sub_type == BGP_ROUTE_STATIC)
1870 break;
1871
1872 *pi = tmp_pi;
1873
1874 /* Mark route for delete. */
1875 if (tmp_pi)
1876 bgp_path_info_delete(rn, tmp_pi);
1877 }
1878
1879
1880
1881 /* Delete EVPN ES (type-4) route */
1882 static int delete_evpn_type4_route(struct bgp *bgp,
1883 struct evpnes *es,
1884 struct prefix_evpn *p)
1885 {
1886 afi_t afi = AFI_L2VPN;
1887 safi_t safi = SAFI_EVPN;
1888 struct bgp_path_info *pi;
1889 struct bgp_node *rn = NULL; /* rn in esi table */
1890 struct bgp_node *global_rn = NULL; /* rn in global table */
1891
1892 /* First, locate the route node within the ESI.
1893 * If it doesn't exist, ther is nothing to do.
1894 * Note: there is no RD here.
1895 */
1896 rn = bgp_node_lookup(es->route_table, (struct prefix *)p);
1897 if (!rn)
1898 return 0;
1899
1900 /* Next, locate route node in the global EVPN routing table.
1901 * Note that this table is a 2-level tree (RD-level + Prefix-level)
1902 */
1903 global_rn = bgp_afi_node_lookup(bgp->rib[afi][safi], afi, safi,
1904 (struct prefix *)p, &es->prd);
1905 if (global_rn) {
1906
1907 /* Delete route entry in the global EVPN table. */
1908 delete_evpn_route_entry(bgp, afi, safi, global_rn, &pi);
1909
1910 /* Schedule for processing - withdraws to peers happen from
1911 * this table.
1912 */
1913 if (pi)
1914 bgp_process(bgp, global_rn, afi, safi);
1915 bgp_unlock_node(global_rn);
1916 }
1917
1918 /*
1919 * Delete route entry in the ESI route table.
1920 * This can just be removed.
1921 */
1922 delete_evpn_route_entry(bgp, afi, safi, rn, &pi);
1923 if (pi)
1924 bgp_path_info_reap(rn, pi);
1925 bgp_unlock_node(rn);
1926 return 0;
1927 }
1928
1929 /* Delete EVPN type5 route */
1930 static int delete_evpn_type5_route(struct bgp *bgp_vrf, struct prefix_evpn *evp)
1931 {
1932 afi_t afi = AFI_L2VPN;
1933 safi_t safi = SAFI_EVPN;
1934 struct bgp_node *rn = NULL;
1935 struct bgp_path_info *pi = NULL;
1936 struct bgp *bgp_def = NULL; /* default bgp instance */
1937
1938 bgp_def = bgp_get_default();
1939 if (!bgp_def)
1940 return 0;
1941
1942 /* locate the global route entry for this type-5 prefix */
1943 rn = bgp_afi_node_lookup(bgp_def->rib[afi][safi], afi, safi,
1944 (struct prefix *)evp, &bgp_vrf->vrf_prd);
1945 if (!rn)
1946 return 0;
1947
1948 delete_evpn_route_entry(bgp_def, afi, safi, rn, &pi);
1949 if (pi)
1950 bgp_process(bgp_def, rn, afi, safi);
1951 bgp_unlock_node(rn);
1952 return 0;
1953 }
1954
1955 /*
1956 * Delete EVPN route (of type based on prefix) for specified VNI and
1957 * schedule for processing.
1958 */
1959 static int delete_evpn_route(struct bgp *bgp, struct bgpevpn *vpn,
1960 struct prefix_evpn *p)
1961 {
1962 struct bgp_node *rn, *global_rn;
1963 struct bgp_path_info *pi;
1964 afi_t afi = AFI_L2VPN;
1965 safi_t safi = SAFI_EVPN;
1966
1967 /* First, locate the route node within the VNI. If it doesn't exist,
1968 * there
1969 * is nothing further to do.
1970 */
1971 /* NOTE: There is no RD here. */
1972 rn = bgp_node_lookup(vpn->route_table, (struct prefix *)p);
1973 if (!rn)
1974 return 0;
1975
1976 /* Next, locate route node in the global EVPN routing table. Note that
1977 * this table is a 2-level tree (RD-level + Prefix-level) similar to
1978 * L3VPN routes.
1979 */
1980 global_rn = bgp_afi_node_lookup(bgp->rib[afi][safi], afi, safi,
1981 (struct prefix *)p, &vpn->prd);
1982 if (global_rn) {
1983 /* Delete route entry in the global EVPN table. */
1984 delete_evpn_route_entry(bgp, afi, safi, global_rn, &pi);
1985
1986 /* Schedule for processing - withdraws to peers happen from
1987 * this table.
1988 */
1989 if (pi)
1990 bgp_process(bgp, global_rn, afi, safi);
1991 bgp_unlock_node(global_rn);
1992 }
1993
1994 /* Delete route entry in the VNI route table. This can just be removed.
1995 */
1996 delete_evpn_route_entry(bgp, afi, safi, rn, &pi);
1997 if (pi) {
1998 bgp_path_info_reap(rn, pi);
1999 evpn_route_select_install(bgp, vpn, rn);
2000 }
2001 bgp_unlock_node(rn);
2002
2003 return 0;
2004 }
2005
2006 /*
2007 * Update all type-2 (MACIP) local routes for this VNI - these should also
2008 * be scheduled for advertise to peers.
2009 */
2010 static int update_all_type2_routes(struct bgp *bgp, struct bgpevpn *vpn)
2011 {
2012 afi_t afi;
2013 safi_t safi;
2014 struct bgp_node *rn;
2015 struct bgp_path_info *pi, *tmp_pi;
2016 struct attr attr;
2017 struct attr *attr_new;
2018 uint32_t seq;
2019 int add_l3_ecomm = 0;
2020
2021 afi = AFI_L2VPN;
2022 safi = SAFI_EVPN;
2023
2024 /* Walk this VNI's route table and update local type-2 routes. For any
2025 * routes updated, update corresponding entry in the global table too.
2026 */
2027 for (rn = bgp_table_top(vpn->route_table); rn;
2028 rn = bgp_route_next(rn)) {
2029 struct prefix_evpn *evp = (struct prefix_evpn *)&rn->p;
2030 struct bgp_node *rd_rn;
2031 struct bgp_path_info *global_pi;
2032
2033 if (evp->prefix.route_type != BGP_EVPN_MAC_IP_ROUTE)
2034 continue;
2035
2036 /* Identify local route. */
2037 for (tmp_pi = rn->info; tmp_pi; tmp_pi = tmp_pi->next) {
2038 if (tmp_pi->peer == bgp->peer_self
2039 && tmp_pi->type == ZEBRA_ROUTE_BGP
2040 && tmp_pi->sub_type == BGP_ROUTE_STATIC)
2041 break;
2042 }
2043
2044 if (!tmp_pi)
2045 continue;
2046
2047 /*
2048 * Build attribute per local route as the MAC mobility and
2049 * some other values could differ for different routes. The
2050 * attributes will be shared in the hash table.
2051 */
2052 bgp_attr_default_set(&attr, BGP_ORIGIN_IGP);
2053 attr.nexthop = vpn->originator_ip;
2054 attr.mp_nexthop_global_in = vpn->originator_ip;
2055 attr.mp_nexthop_len = BGP_ATTR_NHLEN_IPV4;
2056 bgpevpn_get_rmac(vpn, &attr.rmac);
2057
2058 if (evpn_route_is_sticky(bgp, rn))
2059 attr.sticky = 1;
2060 else if (evpn_route_is_def_gw(bgp, rn)) {
2061 attr.default_gw = 1;
2062 if (is_evpn_prefix_ipaddr_v6(evp))
2063 attr.router_flag = 1;
2064 }
2065
2066 /* Add L3 VNI RTs and RMAC for non IPv6 link-local if
2067 * using L3 VNI for type-2 routes also.
2068 */
2069 if ((is_evpn_prefix_ipaddr_v4(evp) ||
2070 !IN6_IS_ADDR_LINKLOCAL(
2071 &evp->prefix.macip_addr.ip.ipaddr_v6)) &&
2072 CHECK_FLAG(vpn->flags, VNI_FLAG_USE_TWO_LABELS) &&
2073 bgpevpn_get_l3vni(vpn))
2074 add_l3_ecomm = 1;
2075
2076 /* Set up extended community. */
2077 build_evpn_route_extcomm(vpn, &attr, add_l3_ecomm);
2078
2079 seq = mac_mobility_seqnum(tmp_pi->attr);
2080
2081 /* Update the route entry. */
2082 update_evpn_route_entry(bgp, vpn, afi, safi, rn, &attr, 0, &pi,
2083 0, seq);
2084
2085 /* Perform route selection; this is just to set the flags
2086 * correctly as local route in the VNI always wins.
2087 */
2088 evpn_route_select_install(bgp, vpn, rn);
2089
2090 attr_new = pi->attr;
2091
2092 /* Update route in global routing table. */
2093 rd_rn = bgp_afi_node_get(bgp->rib[afi][safi], afi, safi,
2094 (struct prefix *)evp, &vpn->prd);
2095 assert(rd_rn);
2096 update_evpn_route_entry(bgp, vpn, afi, safi, rd_rn, attr_new, 0,
2097 &global_pi, 0,
2098 mac_mobility_seqnum(attr_new));
2099
2100 /* Schedule for processing and unlock node. */
2101 bgp_process(bgp, rd_rn, afi, safi);
2102 bgp_unlock_node(rd_rn);
2103
2104 /* Unintern temporary. */
2105 aspath_unintern(&attr.aspath);
2106
2107 }
2108
2109 return 0;
2110 }
2111
2112 /*
2113 * Delete all type-2 (MACIP) local routes for this VNI - only from the
2114 * global routing table. These are also scheduled for withdraw from peers.
2115 */
2116 static int delete_global_type2_routes(struct bgp *bgp, struct bgpevpn *vpn)
2117 {
2118 afi_t afi;
2119 safi_t safi;
2120 struct bgp_node *rdrn, *rn;
2121 struct bgp_table *table;
2122 struct bgp_path_info *pi;
2123
2124 afi = AFI_L2VPN;
2125 safi = SAFI_EVPN;
2126
2127 rdrn = bgp_node_lookup(bgp->rib[afi][safi], (struct prefix *)&vpn->prd);
2128 if (rdrn && rdrn->info) {
2129 table = (struct bgp_table *)rdrn->info;
2130 for (rn = bgp_table_top(table); rn; rn = bgp_route_next(rn)) {
2131 struct prefix_evpn *evp = (struct prefix_evpn *)&rn->p;
2132
2133 if (evp->prefix.route_type != BGP_EVPN_MAC_IP_ROUTE)
2134 continue;
2135
2136 delete_evpn_route_entry(bgp, afi, safi, rn, &pi);
2137 if (pi)
2138 bgp_process(bgp, rn, afi, safi);
2139 }
2140 }
2141
2142 /* Unlock RD node. */
2143 if (rdrn)
2144 bgp_unlock_node(rdrn);
2145
2146 return 0;
2147 }
2148
2149 /*
2150 * Delete all type-2 (MACIP) local routes for this VNI - from the global
2151 * table as well as the per-VNI route table.
2152 */
2153 static int delete_all_type2_routes(struct bgp *bgp, struct bgpevpn *vpn)
2154 {
2155 afi_t afi;
2156 safi_t safi;
2157 struct bgp_node *rn;
2158 struct bgp_path_info *pi;
2159
2160 afi = AFI_L2VPN;
2161 safi = SAFI_EVPN;
2162
2163 /* First, walk the global route table for this VNI's type-2 local
2164 * routes.
2165 * EVPN routes are a 2-level table, first get the RD table.
2166 */
2167 delete_global_type2_routes(bgp, vpn);
2168
2169 /* Next, walk this VNI's route table and delete local type-2 routes. */
2170 for (rn = bgp_table_top(vpn->route_table); rn;
2171 rn = bgp_route_next(rn)) {
2172 struct prefix_evpn *evp = (struct prefix_evpn *)&rn->p;
2173
2174 if (evp->prefix.route_type != BGP_EVPN_MAC_IP_ROUTE)
2175 continue;
2176
2177 delete_evpn_route_entry(bgp, afi, safi, rn, &pi);
2178
2179 /* Route entry in local table gets deleted immediately. */
2180 if (pi)
2181 bgp_path_info_reap(rn, pi);
2182 }
2183
2184 return 0;
2185 }
2186
2187 /*
2188 * Delete all routes in per ES route-table
2189 */
2190 static int delete_all_es_routes(struct bgp *bgp, struct evpnes *es)
2191 {
2192 struct bgp_node *rn;
2193 struct bgp_path_info *pi, *nextpi;
2194
2195 /* Walk this ES's route table and delete all routes. */
2196 for (rn = bgp_table_top(es->route_table); rn;
2197 rn = bgp_route_next(rn)) {
2198 for (pi = rn->info; (pi != NULL) && (nextpi = pi->next, 1);
2199 pi = nextpi) {
2200 bgp_path_info_delete(rn, pi);
2201 bgp_path_info_reap(rn, pi);
2202 }
2203 }
2204
2205 return 0;
2206 }
2207
2208 /*
2209 * Delete all routes in the per-VNI route table.
2210 */
2211 static int delete_all_vni_routes(struct bgp *bgp, struct bgpevpn *vpn)
2212 {
2213 struct bgp_node *rn;
2214 struct bgp_path_info *pi, *nextpi;
2215
2216 /* Walk this VNI's route table and delete all routes. */
2217 for (rn = bgp_table_top(vpn->route_table); rn;
2218 rn = bgp_route_next(rn)) {
2219 for (pi = rn->info; (pi != NULL) && (nextpi = pi->next, 1);
2220 pi = nextpi) {
2221 bgp_path_info_delete(rn, pi);
2222 bgp_path_info_reap(rn, pi);
2223 }
2224 }
2225
2226 return 0;
2227 }
2228
2229 /*
2230 * Update (and advertise) local routes for a VNI. Invoked upon the VNI
2231 * export RT getting modified or change to tunnel IP. Note that these
2232 * situations need the route in the per-VNI table as well as the global
2233 * table to be updated (as attributes change).
2234 */
2235 static int update_routes_for_vni(struct bgp *bgp, struct bgpevpn *vpn)
2236 {
2237 int ret;
2238 struct prefix_evpn p;
2239
2240 /* Update and advertise the type-3 route (only one) followed by the
2241 * locally learnt type-2 routes (MACIP) - for this VNI.
2242 *
2243 * RT-3 only if doing head-end replication
2244 */
2245 if (bgp->vxlan_flood_ctrl == VXLAN_FLOOD_HEAD_END_REPL) {
2246 build_evpn_type3_prefix(&p, vpn->originator_ip);
2247 ret = update_evpn_route(bgp, vpn, &p, 0, 0);
2248 if (ret)
2249 return ret;
2250 }
2251
2252 return update_all_type2_routes(bgp, vpn);
2253 }
2254
2255 /* Delete (and withdraw) local routes for specified ES from global and ES table.
2256 * Also remove all other routes from the per ES table.
2257 * Invoked when ES is deleted.
2258 */
2259 static int delete_routes_for_es(struct bgp *bgp, struct evpnes *es)
2260 {
2261 int ret;
2262 char buf[ESI_STR_LEN];
2263 struct prefix_evpn p;
2264
2265 /* Delete and withdraw locally learnt ES route */
2266 build_evpn_type4_prefix(&p, &es->esi, es->originator_ip.ipaddr_v4);
2267 ret = delete_evpn_type4_route(bgp, es, &p);
2268 if (ret) {
2269 flog_err(EC_BGP_EVPN_ROUTE_DELETE,
2270 "%u failed to delete type-4 route for ESI %s",
2271 bgp->vrf_id, esi_to_str(&es->esi, buf, sizeof(buf)));
2272 }
2273
2274 /* Delete all routes from per ES table */
2275 return delete_all_es_routes(bgp, es);
2276 }
2277
2278 /*
2279 * Delete (and withdraw) local routes for specified VNI from the global
2280 * table and per-VNI table. After this, remove all other routes from
2281 * the per-VNI table. Invoked upon the VNI being deleted or EVPN
2282 * (advertise-all-vni) being disabled.
2283 */
2284 static int delete_routes_for_vni(struct bgp *bgp, struct bgpevpn *vpn)
2285 {
2286 int ret;
2287 struct prefix_evpn p;
2288
2289 /* Delete and withdraw locally learnt type-2 routes (MACIP)
2290 * followed by type-3 routes (only one) - for this VNI.
2291 */
2292 ret = delete_all_type2_routes(bgp, vpn);
2293 if (ret)
2294 return ret;
2295
2296 build_evpn_type3_prefix(&p, vpn->originator_ip);
2297 ret = delete_evpn_route(bgp, vpn, &p);
2298 if (ret)
2299 return ret;
2300
2301 /* Delete all routes from the per-VNI table. */
2302 return delete_all_vni_routes(bgp, vpn);
2303 }
2304
2305 /*
2306 * There is a tunnel endpoint IP address change for this VNI, delete
2307 * prior type-3 route (if needed) and update.
2308 * Note: Route re-advertisement happens elsewhere after other processing
2309 * other changes.
2310 */
2311 static int handle_tunnel_ip_change(struct bgp *bgp, struct bgpevpn *vpn,
2312 struct in_addr originator_ip)
2313 {
2314 struct prefix_evpn p;
2315
2316 /* If VNI is not live, we only need to update the originator ip */
2317 if (!is_vni_live(vpn)) {
2318 vpn->originator_ip = originator_ip;
2319 return 0;
2320 }
2321
2322 /* Update the tunnel-ip hash */
2323 bgp_tip_del(bgp, &vpn->originator_ip);
2324 bgp_tip_add(bgp, &originator_ip);
2325
2326 /* filter routes as martian nexthop db has changed */
2327 bgp_filter_evpn_routes_upon_martian_nh_change(bgp);
2328
2329 /* Need to withdraw type-3 route as the originator IP is part
2330 * of the key.
2331 */
2332 build_evpn_type3_prefix(&p, vpn->originator_ip);
2333 delete_evpn_route(bgp, vpn, &p);
2334
2335 /* Update the tunnel IP and re-advertise all routes for this VNI. */
2336 vpn->originator_ip = originator_ip;
2337 return 0;
2338 }
2339
2340 /* Install EVPN route entry in ES */
2341 static int install_evpn_route_entry_in_es(struct bgp *bgp, struct evpnes *es,
2342 struct prefix_evpn *p,
2343 struct bgp_path_info *parent_pi)
2344 {
2345 int ret = 0;
2346 struct bgp_node *rn = NULL;
2347 struct bgp_path_info *pi = NULL;
2348 struct attr *attr_new = NULL;
2349
2350 /* Create (or fetch) route within the VNI.
2351 * NOTE: There is no RD here.
2352 */
2353 rn = bgp_node_get(es->route_table, (struct prefix *)p);
2354
2355 /* Check if route entry is already present. */
2356 for (pi = rn->info; pi; pi = pi->next)
2357 if (pi->extra
2358 && (struct bgp_path_info *)pi->extra->parent == parent_pi)
2359 break;
2360
2361 if (!pi) {
2362 /* Add (or update) attribute to hash. */
2363 attr_new = bgp_attr_intern(parent_pi->attr);
2364
2365 /* Create new route with its attribute. */
2366 pi = info_make(parent_pi->type, BGP_ROUTE_IMPORTED, 0,
2367 parent_pi->peer, attr_new, rn);
2368 SET_FLAG(pi->flags, BGP_PATH_VALID);
2369 bgp_path_info_extra_get(pi);
2370 pi->extra->parent = parent_pi;
2371 bgp_path_info_add(rn, pi);
2372 } else {
2373 if (attrhash_cmp(pi->attr, parent_pi->attr)
2374 && !CHECK_FLAG(pi->flags, BGP_PATH_REMOVED)) {
2375 bgp_unlock_node(rn);
2376 return 0;
2377 }
2378 /* The attribute has changed. */
2379 /* Add (or update) attribute to hash. */
2380 attr_new = bgp_attr_intern(parent_pi->attr);
2381
2382 /* Restore route, if needed. */
2383 if (CHECK_FLAG(pi->flags, BGP_PATH_REMOVED))
2384 bgp_path_info_restore(rn, pi);
2385
2386 /* Mark if nexthop has changed. */
2387 if (!IPV4_ADDR_SAME(&pi->attr->nexthop, &attr_new->nexthop))
2388 SET_FLAG(pi->flags, BGP_PATH_IGP_CHANGED);
2389
2390 /* Unintern existing, set to new. */
2391 bgp_attr_unintern(&pi->attr);
2392 pi->attr = attr_new;
2393 pi->uptime = bgp_clock();
2394 }
2395
2396 /* Perform route selection and update zebra, if required. */
2397 ret = evpn_es_route_select_install(bgp, es, rn);
2398 return ret;
2399 }
2400
2401 /*
2402 * Install route entry into the VRF routing table and invoke route selection.
2403 */
2404 static int install_evpn_route_entry_in_vrf(struct bgp *bgp_vrf,
2405 struct prefix_evpn *evp,
2406 struct bgp_path_info *parent_pi)
2407 {
2408 struct bgp_node *rn;
2409 struct bgp_path_info *pi;
2410 struct attr attr;
2411 struct attr *attr_new;
2412 int ret = 0;
2413 struct prefix p;
2414 struct prefix *pp = &p;
2415 afi_t afi = 0;
2416 safi_t safi = 0;
2417 char buf[PREFIX_STRLEN];
2418 char buf1[PREFIX_STRLEN];
2419
2420 memset(pp, 0, sizeof(struct prefix));
2421 ip_prefix_from_evpn_prefix(evp, pp);
2422
2423 if (bgp_debug_zebra(NULL)) {
2424 zlog_debug(
2425 "installing evpn prefix %s as ip prefix %s in vrf %s",
2426 prefix2str(evp, buf, sizeof(buf)),
2427 prefix2str(pp, buf1, sizeof(buf)),
2428 vrf_id_to_name(bgp_vrf->vrf_id));
2429 }
2430
2431 /* Create (or fetch) route within the VRF. */
2432 /* NOTE: There is no RD here. */
2433 if (is_evpn_prefix_ipaddr_v4(evp)) {
2434 afi = AFI_IP;
2435 safi = SAFI_UNICAST;
2436 rn = bgp_node_get(bgp_vrf->rib[afi][safi], pp);
2437 } else if (is_evpn_prefix_ipaddr_v6(evp)) {
2438 afi = AFI_IP6;
2439 safi = SAFI_UNICAST;
2440 rn = bgp_node_get(bgp_vrf->rib[afi][safi], pp);
2441 } else
2442 return 0;
2443
2444 /* EVPN routes currently only support a IPv4 next hop which corresponds
2445 * to the remote VTEP. When importing into a VRF, if it is IPv6 host
2446 * or prefix route, we have to convert the next hop to an IPv4-mapped
2447 * address for the rest of the code to flow through. In the case of IPv4,
2448 * make sure to set the flag for next hop attribute.
2449 */
2450 bgp_attr_dup(&attr, parent_pi->attr);
2451 if (afi == AFI_IP6)
2452 evpn_convert_nexthop_to_ipv6(&attr);
2453 else
2454 attr.flag |= ATTR_FLAG_BIT(BGP_ATTR_NEXT_HOP);
2455
2456 /* Check if route entry is already present. */
2457 for (pi = rn->info; pi; pi = pi->next)
2458 if (pi->extra
2459 && (struct bgp_path_info *)pi->extra->parent == parent_pi)
2460 break;
2461
2462 if (!pi) {
2463 /* Add (or update) attribute to hash. */
2464 attr_new = bgp_attr_intern(&attr);
2465
2466 /* Create new route with its attribute. */
2467 pi = info_make(parent_pi->type, BGP_ROUTE_IMPORTED, 0,
2468 parent_pi->peer, attr_new, rn);
2469 SET_FLAG(pi->flags, BGP_PATH_VALID);
2470 bgp_path_info_extra_get(pi);
2471 pi->extra->parent = bgp_path_info_lock(parent_pi);
2472 bgp_lock_node((struct bgp_node *)parent_pi->net);
2473 if (parent_pi->extra) {
2474 memcpy(&pi->extra->label, &parent_pi->extra->label,
2475 sizeof(pi->extra->label));
2476 pi->extra->num_labels = parent_pi->extra->num_labels;
2477 }
2478 bgp_path_info_add(rn, pi);
2479 } else {
2480 if (attrhash_cmp(pi->attr, &attr)
2481 && !CHECK_FLAG(pi->flags, BGP_PATH_REMOVED)) {
2482 bgp_unlock_node(rn);
2483 return 0;
2484 }
2485 /* The attribute has changed. */
2486 /* Add (or update) attribute to hash. */
2487 attr_new = bgp_attr_intern(&attr);
2488
2489 /* Restore route, if needed. */
2490 if (CHECK_FLAG(pi->flags, BGP_PATH_REMOVED))
2491 bgp_path_info_restore(rn, pi);
2492
2493 /* Mark if nexthop has changed. */
2494 if ((afi == AFI_IP
2495 && !IPV4_ADDR_SAME(&pi->attr->nexthop, &attr_new->nexthop))
2496 || (afi == AFI_IP6
2497 && !IPV6_ADDR_SAME(&pi->attr->mp_nexthop_global,
2498 &attr_new->mp_nexthop_global)))
2499 SET_FLAG(pi->flags, BGP_PATH_IGP_CHANGED);
2500
2501 /* Unintern existing, set to new. */
2502 bgp_attr_unintern(&pi->attr);
2503 pi->attr = attr_new;
2504 pi->uptime = bgp_clock();
2505 }
2506
2507 bgp_aggregate_increment(bgp_vrf, &rn->p, pi, afi, safi);
2508
2509 /* Perform route selection and update zebra, if required. */
2510 bgp_process(bgp_vrf, rn, afi, safi);
2511
2512 return ret;
2513 }
2514
2515 /*
2516 * Install route entry into the VNI routing table and invoke route selection.
2517 */
2518 static int install_evpn_route_entry(struct bgp *bgp, struct bgpevpn *vpn,
2519 struct prefix_evpn *p,
2520 struct bgp_path_info *parent_pi)
2521 {
2522 struct bgp_node *rn;
2523 struct bgp_path_info *pi;
2524 struct attr *attr_new;
2525 int ret;
2526
2527 /* Create (or fetch) route within the VNI. */
2528 /* NOTE: There is no RD here. */
2529 rn = bgp_node_get(vpn->route_table, (struct prefix *)p);
2530
2531 /* Check if route entry is already present. */
2532 for (pi = rn->info; pi; pi = pi->next)
2533 if (pi->extra
2534 && (struct bgp_path_info *)pi->extra->parent == parent_pi)
2535 break;
2536
2537 if (!pi) {
2538 /* Add (or update) attribute to hash. */
2539 attr_new = bgp_attr_intern(parent_pi->attr);
2540
2541 /* Create new route with its attribute. */
2542 pi = info_make(parent_pi->type, BGP_ROUTE_IMPORTED, 0,
2543 parent_pi->peer, attr_new, rn);
2544 SET_FLAG(pi->flags, BGP_PATH_VALID);
2545 bgp_path_info_extra_get(pi);
2546 pi->extra->parent = bgp_path_info_lock(parent_pi);
2547 bgp_lock_node((struct bgp_node *)parent_pi->net);
2548 if (parent_pi->extra) {
2549 memcpy(&pi->extra->label, &parent_pi->extra->label,
2550 sizeof(pi->extra->label));
2551 pi->extra->num_labels = parent_pi->extra->num_labels;
2552 }
2553 bgp_path_info_add(rn, pi);
2554 } else {
2555 if (attrhash_cmp(pi->attr, parent_pi->attr)
2556 && !CHECK_FLAG(pi->flags, BGP_PATH_REMOVED)) {
2557 bgp_unlock_node(rn);
2558 return 0;
2559 }
2560 /* The attribute has changed. */
2561 /* Add (or update) attribute to hash. */
2562 attr_new = bgp_attr_intern(parent_pi->attr);
2563
2564 /* Restore route, if needed. */
2565 if (CHECK_FLAG(pi->flags, BGP_PATH_REMOVED))
2566 bgp_path_info_restore(rn, pi);
2567
2568 /* Mark if nexthop has changed. */
2569 if (!IPV4_ADDR_SAME(&pi->attr->nexthop, &attr_new->nexthop))
2570 SET_FLAG(pi->flags, BGP_PATH_IGP_CHANGED);
2571
2572 /* Unintern existing, set to new. */
2573 bgp_attr_unintern(&pi->attr);
2574 pi->attr = attr_new;
2575 pi->uptime = bgp_clock();
2576 }
2577
2578 /* Perform route selection and update zebra, if required. */
2579 ret = evpn_route_select_install(bgp, vpn, rn);
2580
2581 return ret;
2582 }
2583
2584 /* Uninstall EVPN route entry from ES route table */
2585 static int uninstall_evpn_route_entry_in_es(struct bgp *bgp, struct evpnes *es,
2586 struct prefix_evpn *p,
2587 struct bgp_path_info *parent_pi)
2588 {
2589 int ret;
2590 struct bgp_node *rn;
2591 struct bgp_path_info *pi;
2592
2593 if (!es->route_table)
2594 return 0;
2595
2596 /* Locate route within the ESI.
2597 * NOTE: There is no RD here.
2598 */
2599 rn = bgp_node_lookup(es->route_table, (struct prefix *)p);
2600 if (!rn)
2601 return 0;
2602
2603 /* Find matching route entry. */
2604 for (pi = rn->info; pi; pi = pi->next)
2605 if (pi->extra
2606 && (struct bgp_path_info *)pi->extra->parent == parent_pi)
2607 break;
2608
2609 if (!pi)
2610 return 0;
2611
2612 /* Mark entry for deletion */
2613 bgp_path_info_delete(rn, pi);
2614
2615 /* Perform route selection and update zebra, if required. */
2616 ret = evpn_es_route_select_install(bgp, es, rn);
2617
2618 /* Unlock route node. */
2619 bgp_unlock_node(rn);
2620
2621 return ret;
2622 }
2623
2624 /*
2625 * Uninstall route entry from the VRF routing table and send message
2626 * to zebra, if appropriate.
2627 */
2628 static int uninstall_evpn_route_entry_in_vrf(struct bgp *bgp_vrf,
2629 struct prefix_evpn *evp,
2630 struct bgp_path_info *parent_pi)
2631 {
2632 struct bgp_node *rn;
2633 struct bgp_path_info *pi;
2634 int ret = 0;
2635 struct prefix p;
2636 struct prefix *pp = &p;
2637 afi_t afi = 0;
2638 safi_t safi = 0;
2639 char buf[PREFIX_STRLEN];
2640 char buf1[PREFIX_STRLEN];
2641
2642 memset(pp, 0, sizeof(struct prefix));
2643 ip_prefix_from_evpn_prefix(evp, pp);
2644
2645 if (bgp_debug_zebra(NULL)) {
2646 zlog_debug(
2647 "uninstalling evpn prefix %s as ip prefix %s in vrf %s",
2648 prefix2str(evp, buf, sizeof(buf)),
2649 prefix2str(pp, buf1, sizeof(buf)),
2650 vrf_id_to_name(bgp_vrf->vrf_id));
2651 }
2652
2653 /* Locate route within the VRF. */
2654 /* NOTE: There is no RD here. */
2655 if (is_evpn_prefix_ipaddr_v4(evp)) {
2656 afi = AFI_IP;
2657 safi = SAFI_UNICAST;
2658 rn = bgp_node_lookup(bgp_vrf->rib[afi][safi], pp);
2659 } else {
2660 afi = AFI_IP6;
2661 safi = SAFI_UNICAST;
2662 rn = bgp_node_lookup(bgp_vrf->rib[afi][safi], pp);
2663 }
2664
2665 if (!rn)
2666 return 0;
2667
2668 /* Find matching route entry. */
2669 for (pi = rn->info; pi; pi = pi->next)
2670 if (pi->extra
2671 && (struct bgp_path_info *)pi->extra->parent == parent_pi)
2672 break;
2673
2674 if (!pi)
2675 return 0;
2676
2677 bgp_aggregate_decrement(bgp_vrf, &rn->p, pi, afi, safi);
2678
2679 /* Mark entry for deletion */
2680 bgp_path_info_delete(rn, pi);
2681
2682 /* Perform route selection and update zebra, if required. */
2683 bgp_process(bgp_vrf, rn, afi, safi);
2684
2685 /* Unlock route node. */
2686 bgp_unlock_node(rn);
2687
2688 return ret;
2689 }
2690
2691 /*
2692 * Uninstall route entry from the VNI routing table and send message
2693 * to zebra, if appropriate.
2694 */
2695 static int uninstall_evpn_route_entry(struct bgp *bgp, struct bgpevpn *vpn,
2696 struct prefix_evpn *p,
2697 struct bgp_path_info *parent_pi)
2698 {
2699 struct bgp_node *rn;
2700 struct bgp_path_info *pi;
2701 int ret;
2702
2703 /* Locate route within the VNI. */
2704 /* NOTE: There is no RD here. */
2705 rn = bgp_node_lookup(vpn->route_table, (struct prefix *)p);
2706 if (!rn)
2707 return 0;
2708
2709 /* Find matching route entry. */
2710 for (pi = rn->info; pi; pi = pi->next)
2711 if (pi->extra
2712 && (struct bgp_path_info *)pi->extra->parent == parent_pi)
2713 break;
2714
2715 if (!pi)
2716 return 0;
2717
2718 /* Mark entry for deletion */
2719 bgp_path_info_delete(rn, pi);
2720
2721 /* Perform route selection and update zebra, if required. */
2722 ret = evpn_route_select_install(bgp, vpn, rn);
2723
2724 /* Unlock route node. */
2725 bgp_unlock_node(rn);
2726
2727 return ret;
2728 }
2729
2730 /*
2731 * Given a prefix, see if it belongs to ES.
2732 */
2733 static int is_prefix_matching_for_es(struct prefix_evpn *p,
2734 struct evpnes *es)
2735 {
2736 /* if not an ES route return false */
2737 if (p->prefix.route_type != BGP_EVPN_ES_ROUTE)
2738 return 0;
2739
2740 if (memcmp(&p->prefix.es_addr.esi, &es->esi, sizeof(esi_t)) == 0)
2741 return 1;
2742
2743 return 0;
2744 }
2745
2746 /*
2747 * Given a route entry and a VRF, see if this route entry should be
2748 * imported into the VRF i.e., RTs match.
2749 */
2750 static int is_route_matching_for_vrf(struct bgp *bgp_vrf,
2751 struct bgp_path_info *pi)
2752 {
2753 struct attr *attr = pi->attr;
2754 struct ecommunity *ecom;
2755 int i;
2756
2757 assert(attr);
2758 /* Route should have valid RT to be even considered. */
2759 if (!(attr->flag & ATTR_FLAG_BIT(BGP_ATTR_EXT_COMMUNITIES)))
2760 return 0;
2761
2762 ecom = attr->ecommunity;
2763 if (!ecom || !ecom->size)
2764 return 0;
2765
2766 /* For each extended community RT, see if it matches this VNI. If any RT
2767 * matches, we're done.
2768 */
2769 for (i = 0; i < ecom->size; i++) {
2770 uint8_t *pnt;
2771 uint8_t type, sub_type;
2772 struct ecommunity_val *eval;
2773 struct ecommunity_val eval_tmp;
2774 struct vrf_irt_node *irt;
2775
2776 /* Only deal with RTs */
2777 pnt = (ecom->val + (i * ECOMMUNITY_SIZE));
2778 eval = (struct ecommunity_val *)(ecom->val
2779 + (i * ECOMMUNITY_SIZE));
2780 type = *pnt++;
2781 sub_type = *pnt++;
2782 if (sub_type != ECOMMUNITY_ROUTE_TARGET)
2783 continue;
2784
2785 /* See if this RT matches specified VNIs import RTs */
2786 irt = lookup_vrf_import_rt(eval);
2787 if (irt)
2788 if (is_vrf_present_in_irt_vrfs(irt->vrfs, bgp_vrf))
2789 return 1;
2790
2791 /* Also check for non-exact match. In this, we mask out the AS
2792 * and
2793 * only check on the local-admin sub-field. This is to
2794 * facilitate using
2795 * VNI as the RT for EBGP peering too.
2796 */
2797 irt = NULL;
2798 if (type == ECOMMUNITY_ENCODE_AS
2799 || type == ECOMMUNITY_ENCODE_AS4
2800 || type == ECOMMUNITY_ENCODE_IP) {
2801 memcpy(&eval_tmp, eval, ECOMMUNITY_SIZE);
2802 mask_ecom_global_admin(&eval_tmp, eval);
2803 irt = lookup_vrf_import_rt(&eval_tmp);
2804 }
2805 if (irt)
2806 if (is_vrf_present_in_irt_vrfs(irt->vrfs, bgp_vrf))
2807 return 1;
2808 }
2809
2810 return 0;
2811 }
2812
2813 /*
2814 * Given a route entry and a VNI, see if this route entry should be
2815 * imported into the VNI i.e., RTs match.
2816 */
2817 static int is_route_matching_for_vni(struct bgp *bgp, struct bgpevpn *vpn,
2818 struct bgp_path_info *pi)
2819 {
2820 struct attr *attr = pi->attr;
2821 struct ecommunity *ecom;
2822 int i;
2823
2824 assert(attr);
2825 /* Route should have valid RT to be even considered. */
2826 if (!(attr->flag & ATTR_FLAG_BIT(BGP_ATTR_EXT_COMMUNITIES)))
2827 return 0;
2828
2829 ecom = attr->ecommunity;
2830 if (!ecom || !ecom->size)
2831 return 0;
2832
2833 /* For each extended community RT, see if it matches this VNI. If any RT
2834 * matches, we're done.
2835 */
2836 for (i = 0; i < ecom->size; i++) {
2837 uint8_t *pnt;
2838 uint8_t type, sub_type;
2839 struct ecommunity_val *eval;
2840 struct ecommunity_val eval_tmp;
2841 struct irt_node *irt;
2842
2843 /* Only deal with RTs */
2844 pnt = (ecom->val + (i * ECOMMUNITY_SIZE));
2845 eval = (struct ecommunity_val *)(ecom->val
2846 + (i * ECOMMUNITY_SIZE));
2847 type = *pnt++;
2848 sub_type = *pnt++;
2849 if (sub_type != ECOMMUNITY_ROUTE_TARGET)
2850 continue;
2851
2852 /* See if this RT matches specified VNIs import RTs */
2853 irt = lookup_import_rt(bgp, eval);
2854 if (irt)
2855 if (is_vni_present_in_irt_vnis(irt->vnis, vpn))
2856 return 1;
2857
2858 /* Also check for non-exact match. In this, we mask out the AS
2859 * and
2860 * only check on the local-admin sub-field. This is to
2861 * facilitate using
2862 * VNI as the RT for EBGP peering too.
2863 */
2864 irt = NULL;
2865 if (type == ECOMMUNITY_ENCODE_AS
2866 || type == ECOMMUNITY_ENCODE_AS4
2867 || type == ECOMMUNITY_ENCODE_IP) {
2868 memcpy(&eval_tmp, eval, ECOMMUNITY_SIZE);
2869 mask_ecom_global_admin(&eval_tmp, eval);
2870 irt = lookup_import_rt(bgp, &eval_tmp);
2871 }
2872 if (irt)
2873 if (is_vni_present_in_irt_vnis(irt->vnis, vpn))
2874 return 1;
2875 }
2876
2877 return 0;
2878 }
2879
2880 static int install_uninstall_routes_for_es(struct bgp *bgp,
2881 struct evpnes *es,
2882 int install)
2883 {
2884 int ret;
2885 afi_t afi;
2886 safi_t safi;
2887 char buf[PREFIX_STRLEN];
2888 char buf1[ESI_STR_LEN];
2889 struct bgp_node *rd_rn, *rn;
2890 struct bgp_table *table;
2891 struct bgp_path_info *pi;
2892
2893 afi = AFI_L2VPN;
2894 safi = SAFI_EVPN;
2895
2896 /*
2897 * Walk entire global routing table and evaluate routes which could be
2898 * imported into this VRF. Note that we need to loop through all global
2899 * routes to determine which route matches the import rt on vrf
2900 */
2901 for (rd_rn = bgp_table_top(bgp->rib[afi][safi]); rd_rn;
2902 rd_rn = bgp_route_next(rd_rn)) {
2903 table = (struct bgp_table *)(rd_rn->info);
2904 if (!table)
2905 continue;
2906
2907 for (rn = bgp_table_top(table); rn; rn = bgp_route_next(rn)) {
2908 struct prefix_evpn *evp = (struct prefix_evpn *)&rn->p;
2909
2910 for (pi = rn->info; pi; pi = pi->next) {
2911 /*
2912 * Consider "valid" remote routes applicable for
2913 * this ES.
2914 */
2915 if (!(CHECK_FLAG(pi->flags, BGP_PATH_VALID)
2916 && pi->type == ZEBRA_ROUTE_BGP
2917 && pi->sub_type == BGP_ROUTE_NORMAL))
2918 continue;
2919
2920 if (!is_prefix_matching_for_es(evp, es))
2921 continue;
2922
2923 if (install)
2924 ret = install_evpn_route_entry_in_es(
2925 bgp, es, evp, pi);
2926 else
2927 ret = uninstall_evpn_route_entry_in_es(
2928 bgp, es, evp, pi);
2929
2930 if (ret) {
2931 flog_err(
2932 EC_BGP_EVPN_FAIL,
2933 "Failed to %s EVPN %s route in ESI %s",
2934 install ? "install"
2935 : "uninstall",
2936 prefix2str(evp, buf,
2937 sizeof(buf)),
2938 esi_to_str(&es->esi, buf1,
2939 sizeof(buf1)));
2940 return ret;
2941 }
2942 }
2943 }
2944 }
2945 return 0;
2946 }
2947
2948 /*
2949 * Install or uninstall mac-ip routes are appropriate for this
2950 * particular VRF.
2951 */
2952 static int install_uninstall_routes_for_vrf(struct bgp *bgp_vrf, int install)
2953 {
2954 afi_t afi;
2955 safi_t safi;
2956 struct bgp_node *rd_rn, *rn;
2957 struct bgp_table *table;
2958 struct bgp_path_info *pi;
2959 int ret;
2960 char buf[PREFIX_STRLEN];
2961 struct bgp *bgp_def = NULL;
2962
2963 afi = AFI_L2VPN;
2964 safi = SAFI_EVPN;
2965 bgp_def = bgp_get_default();
2966 if (!bgp_def)
2967 return -1;
2968
2969 /* Walk entire global routing table and evaluate routes which could be
2970 * imported into this VRF. Note that we need to loop through all global
2971 * routes to determine which route matches the import rt on vrf
2972 */
2973 for (rd_rn = bgp_table_top(bgp_def->rib[afi][safi]); rd_rn;
2974 rd_rn = bgp_route_next(rd_rn)) {
2975 table = (struct bgp_table *)(rd_rn->info);
2976 if (!table)
2977 continue;
2978
2979 for (rn = bgp_table_top(table); rn; rn = bgp_route_next(rn)) {
2980 struct prefix_evpn *evp = (struct prefix_evpn *)&rn->p;
2981
2982 /* if not mac-ip route skip this route */
2983 if (!(evp->prefix.route_type == BGP_EVPN_MAC_IP_ROUTE
2984 || evp->prefix.route_type
2985 == BGP_EVPN_IP_PREFIX_ROUTE))
2986 continue;
2987
2988 /* if not a mac+ip route skip this route */
2989 if (!(is_evpn_prefix_ipaddr_v4(evp)
2990 || is_evpn_prefix_ipaddr_v6(evp)))
2991 continue;
2992
2993 for (pi = rn->info; pi; pi = pi->next) {
2994 /* Consider "valid" remote routes applicable for
2995 * this VRF.
2996 */
2997 if (!(CHECK_FLAG(pi->flags, BGP_PATH_VALID)
2998 && pi->type == ZEBRA_ROUTE_BGP
2999 && pi->sub_type == BGP_ROUTE_NORMAL))
3000 continue;
3001
3002 if (is_route_matching_for_vrf(bgp_vrf, pi)) {
3003 if (install)
3004 ret = install_evpn_route_entry_in_vrf(
3005 bgp_vrf, evp, pi);
3006 else
3007 ret = uninstall_evpn_route_entry_in_vrf(
3008 bgp_vrf, evp, pi);
3009
3010 if (ret) {
3011 flog_err(
3012 EC_BGP_EVPN_FAIL,
3013 "Failed to %s EVPN %s route in VRF %s",
3014 install ? "install"
3015 : "uninstall",
3016 prefix2str(evp, buf,
3017 sizeof(buf)),
3018 vrf_id_to_name(
3019 bgp_vrf->vrf_id));
3020 return ret;
3021 }
3022 }
3023 }
3024 }
3025 }
3026
3027 return 0;
3028 }
3029
3030 /*
3031 * Install or uninstall routes of specified type that are appropriate for this
3032 * particular VNI.
3033 */
3034 static int install_uninstall_routes_for_vni(struct bgp *bgp,
3035 struct bgpevpn *vpn,
3036 bgp_evpn_route_type rtype,
3037 int install)
3038 {
3039 afi_t afi;
3040 safi_t safi;
3041 struct bgp_node *rd_rn, *rn;
3042 struct bgp_table *table;
3043 struct bgp_path_info *pi;
3044 int ret;
3045
3046 afi = AFI_L2VPN;
3047 safi = SAFI_EVPN;
3048
3049 /* Walk entire global routing table and evaluate routes which could be
3050 * imported into this VPN. Note that we cannot just look at the routes
3051 * for
3052 * the VNI's RD - remote routes applicable for this VNI could have any
3053 * RD.
3054 */
3055 /* EVPN routes are a 2-level table. */
3056 for (rd_rn = bgp_table_top(bgp->rib[afi][safi]); rd_rn;
3057 rd_rn = bgp_route_next(rd_rn)) {
3058 table = (struct bgp_table *)(rd_rn->info);
3059 if (!table)
3060 continue;
3061
3062 for (rn = bgp_table_top(table); rn; rn = bgp_route_next(rn)) {
3063 struct prefix_evpn *evp = (struct prefix_evpn *)&rn->p;
3064
3065 if (evp->prefix.route_type != rtype)
3066 continue;
3067
3068 for (pi = rn->info; pi; pi = pi->next) {
3069 /* Consider "valid" remote routes applicable for
3070 * this VNI. */
3071 if (!(CHECK_FLAG(pi->flags, BGP_PATH_VALID)
3072 && pi->type == ZEBRA_ROUTE_BGP
3073 && pi->sub_type == BGP_ROUTE_NORMAL))
3074 continue;
3075
3076 if (is_route_matching_for_vni(bgp, vpn, pi)) {
3077 if (install)
3078 ret = install_evpn_route_entry(
3079 bgp, vpn, evp, pi);
3080 else
3081 ret = uninstall_evpn_route_entry(
3082 bgp, vpn, evp, pi);
3083
3084 if (ret) {
3085 flog_err(
3086 EC_BGP_EVPN_FAIL,
3087 "%u: Failed to %s EVPN %s route in VNI %u",
3088 bgp->vrf_id,
3089 install ? "install"
3090 : "uninstall",
3091 rtype == BGP_EVPN_MAC_IP_ROUTE
3092 ? "MACIP"
3093 : "IMET",
3094 vpn->vni);
3095 return ret;
3096 }
3097 }
3098 }
3099 }
3100 }
3101
3102 return 0;
3103 }
3104
3105 /* Install any existing remote ES routes applicable for this ES into its routing
3106 * table. This is invoked when ES comes up.
3107 */
3108 static int install_routes_for_es(struct bgp *bgp, struct evpnes *es)
3109 {
3110 return install_uninstall_routes_for_es(bgp, es, 1);
3111 }
3112
3113
3114 /* Install any existing remote routes applicable for this VRF into VRF RIB. This
3115 * is invoked upon l3vni-add or l3vni import rt change
3116 */
3117 static int install_routes_for_vrf(struct bgp *bgp_vrf)
3118 {
3119 install_uninstall_routes_for_vrf(bgp_vrf, 1);
3120 return 0;
3121 }
3122
3123 /*
3124 * Install any existing remote routes applicable for this VNI into its
3125 * routing table. This is invoked when a VNI becomes "live" or its Import
3126 * RT is changed.
3127 */
3128 static int install_routes_for_vni(struct bgp *bgp, struct bgpevpn *vpn)
3129 {
3130 int ret;
3131
3132 /* Install type-3 routes followed by type-2 routes - the ones applicable
3133 * for this VNI.
3134 */
3135 ret = install_uninstall_routes_for_vni(bgp, vpn, BGP_EVPN_IMET_ROUTE,
3136 1);
3137 if (ret)
3138 return ret;
3139
3140 return install_uninstall_routes_for_vni(bgp, vpn, BGP_EVPN_MAC_IP_ROUTE,
3141 1);
3142 }
3143
3144 /* uninstall routes from l3vni vrf. */
3145 static int uninstall_routes_for_vrf(struct bgp *bgp_vrf)
3146 {
3147 install_uninstall_routes_for_vrf(bgp_vrf, 0);
3148 return 0;
3149 }
3150
3151 /*
3152 * Uninstall any existing remote routes for this VNI. One scenario in which
3153 * this is invoked is upon an import RT change.
3154 */
3155 static int uninstall_routes_for_vni(struct bgp *bgp, struct bgpevpn *vpn)
3156 {
3157 int ret;
3158
3159 /* Uninstall type-2 routes followed by type-3 routes - the ones
3160 * applicable
3161 * for this VNI.
3162 */
3163 ret = install_uninstall_routes_for_vni(bgp, vpn, BGP_EVPN_MAC_IP_ROUTE,
3164 0);
3165 if (ret)
3166 return ret;
3167
3168 return install_uninstall_routes_for_vni(bgp, vpn, BGP_EVPN_IMET_ROUTE,
3169 0);
3170 }
3171
3172 /* Install or unistall route in ES */
3173 static int install_uninstall_route_in_es(struct bgp *bgp, struct evpnes *es,
3174 afi_t afi, safi_t safi,
3175 struct prefix_evpn *evp,
3176 struct bgp_path_info *pi, int install)
3177 {
3178 int ret = 0;
3179 char buf[ESI_STR_LEN];
3180
3181 if (install)
3182 ret = install_evpn_route_entry_in_es(bgp, es, evp, pi);
3183 else
3184 ret = uninstall_evpn_route_entry_in_es(bgp, es, evp, pi);
3185
3186 if (ret) {
3187 flog_err(
3188 EC_BGP_EVPN_FAIL,
3189 "%u: Failed to %s EVPN %s route in ESI %s", bgp->vrf_id,
3190 install ? "install" : "uninstall", "ES",
3191 esi_to_str(&evp->prefix.es_addr.esi, buf, sizeof(buf)));
3192 return ret;
3193 }
3194 return 0;
3195 }
3196
3197 /*
3198 * Install or uninstall route in matching VRFs (list).
3199 */
3200 static int install_uninstall_route_in_vrfs(struct bgp *bgp_def, afi_t afi,
3201 safi_t safi, struct prefix_evpn *evp,
3202 struct bgp_path_info *pi,
3203 struct list *vrfs, int install)
3204 {
3205 char buf[PREFIX2STR_BUFFER];
3206 struct bgp *bgp_vrf;
3207 struct listnode *node, *nnode;
3208
3209 /* Only type-2/type-5 routes go into a VRF */
3210 if (!(evp->prefix.route_type == BGP_EVPN_MAC_IP_ROUTE
3211 || evp->prefix.route_type == BGP_EVPN_IP_PREFIX_ROUTE))
3212 return 0;
3213
3214 /* if it is type-2 route and not a mac+ip route skip this route */
3215 if ((evp->prefix.route_type == BGP_EVPN_MAC_IP_ROUTE)
3216 && !(is_evpn_prefix_ipaddr_v4(evp)
3217 || is_evpn_prefix_ipaddr_v6(evp)))
3218 return 0;
3219
3220 for (ALL_LIST_ELEMENTS(vrfs, node, nnode, bgp_vrf)) {
3221 int ret;
3222
3223 if (install)
3224 ret = install_evpn_route_entry_in_vrf(bgp_vrf, evp, pi);
3225 else
3226 ret = uninstall_evpn_route_entry_in_vrf(bgp_vrf, evp,
3227 pi);
3228
3229 if (ret) {
3230 flog_err(EC_BGP_EVPN_FAIL,
3231 "%u: Failed to %s prefix %s in VRF %s",
3232 bgp_def->vrf_id,
3233 install ? "install" : "uninstall",
3234 prefix2str(evp, buf, sizeof(buf)),
3235 vrf_id_to_name(bgp_vrf->vrf_id));
3236 return ret;
3237 }
3238 }
3239
3240 return 0;
3241 }
3242
3243 /*
3244 * Install or uninstall route in matching VNIs (list).
3245 */
3246 static int install_uninstall_route_in_vnis(struct bgp *bgp, afi_t afi,
3247 safi_t safi, struct prefix_evpn *evp,
3248 struct bgp_path_info *pi,
3249 struct list *vnis, int install)
3250 {
3251 struct bgpevpn *vpn;
3252 struct listnode *node, *nnode;
3253
3254 for (ALL_LIST_ELEMENTS(vnis, node, nnode, vpn)) {
3255 int ret;
3256
3257 if (!is_vni_live(vpn))
3258 continue;
3259
3260 if (install)
3261 ret = install_evpn_route_entry(bgp, vpn, evp, pi);
3262 else
3263 ret = uninstall_evpn_route_entry(bgp, vpn, evp, pi);
3264
3265 if (ret) {
3266 flog_err(EC_BGP_EVPN_FAIL,
3267 "%u: Failed to %s EVPN %s route in VNI %u",
3268 bgp->vrf_id, install ? "install" : "uninstall",
3269 evp->prefix.route_type == BGP_EVPN_MAC_IP_ROUTE
3270 ? "MACIP"
3271 : "IMET",
3272 vpn->vni);
3273 return ret;
3274 }
3275 }
3276
3277 return 0;
3278 }
3279
3280 /*
3281 * Install or uninstall route for appropriate VNIs/ESIs.
3282 */
3283 static int install_uninstall_evpn_route(struct bgp *bgp, afi_t afi, safi_t safi,
3284 struct prefix *p,
3285 struct bgp_path_info *pi, int import)
3286 {
3287 struct prefix_evpn *evp = (struct prefix_evpn *)p;
3288 struct attr *attr = pi->attr;
3289 struct ecommunity *ecom;
3290 int i;
3291
3292 assert(attr);
3293
3294 /* Only type-2, type-3, type-4 and type-5 are supported currently */
3295 if (!(evp->prefix.route_type == BGP_EVPN_MAC_IP_ROUTE
3296 || evp->prefix.route_type == BGP_EVPN_IMET_ROUTE
3297 || evp->prefix.route_type == BGP_EVPN_ES_ROUTE
3298 || evp->prefix.route_type == BGP_EVPN_IP_PREFIX_ROUTE))
3299 return 0;
3300
3301 /* If we don't have Route Target, nothing much to do. */
3302 if (!(attr->flag & ATTR_FLAG_BIT(BGP_ATTR_EXT_COMMUNITIES)))
3303 return 0;
3304
3305 ecom = attr->ecommunity;
3306 if (!ecom || !ecom->size)
3307 return -1;
3308
3309 /* An EVPN route belongs to a VNI or a VRF or an ESI based on the RTs
3310 * attached to the route */
3311 for (i = 0; i < ecom->size; i++) {
3312 uint8_t *pnt;
3313 uint8_t type, sub_type;
3314 struct ecommunity_val *eval;
3315 struct ecommunity_val eval_tmp;
3316 struct irt_node *irt; /* import rt for l2vni */
3317 struct vrf_irt_node *vrf_irt; /* import rt for l3vni */
3318 struct evpnes *es;
3319
3320 /* Only deal with RTs */
3321 pnt = (ecom->val + (i * ECOMMUNITY_SIZE));
3322 eval = (struct ecommunity_val *)(ecom->val
3323 + (i * ECOMMUNITY_SIZE));
3324 type = *pnt++;
3325 sub_type = *pnt++;
3326 if (sub_type != ECOMMUNITY_ROUTE_TARGET)
3327 continue;
3328
3329 /*
3330 * macip routes (type-2) are imported into VNI and VRF tables.
3331 * IMET route is imported into VNI table.
3332 * prefix routes are imported into VRF table.
3333 */
3334 if (evp->prefix.route_type == BGP_EVPN_MAC_IP_ROUTE ||
3335 evp->prefix.route_type == BGP_EVPN_IMET_ROUTE ||
3336 evp->prefix.route_type == BGP_EVPN_IP_PREFIX_ROUTE) {
3337
3338 irt = lookup_import_rt(bgp, eval);
3339 if (irt)
3340 install_uninstall_route_in_vnis(
3341 bgp, afi, safi, evp, pi, irt->vnis,
3342 import);
3343
3344 vrf_irt = lookup_vrf_import_rt(eval);
3345 if (vrf_irt)
3346 install_uninstall_route_in_vrfs(
3347 bgp, afi, safi, evp, pi, vrf_irt->vrfs,
3348 import);
3349
3350 /* Also check for non-exact match.
3351 * In this, we mask out the AS and
3352 * only check on the local-admin sub-field.
3353 * This is to facilitate using
3354 * VNI as the RT for EBGP peering too.
3355 */
3356 irt = NULL;
3357 vrf_irt = NULL;
3358 if (type == ECOMMUNITY_ENCODE_AS
3359 || type == ECOMMUNITY_ENCODE_AS4
3360 || type == ECOMMUNITY_ENCODE_IP) {
3361 memcpy(&eval_tmp, eval, ECOMMUNITY_SIZE);
3362 mask_ecom_global_admin(&eval_tmp, eval);
3363 irt = lookup_import_rt(bgp, &eval_tmp);
3364 vrf_irt = lookup_vrf_import_rt(&eval_tmp);
3365 }
3366
3367 if (irt)
3368 install_uninstall_route_in_vnis(
3369 bgp, afi, safi, evp, pi, irt->vnis,
3370 import);
3371 if (vrf_irt)
3372 install_uninstall_route_in_vrfs(
3373 bgp, afi, safi, evp, pi, vrf_irt->vrfs,
3374 import);
3375 }
3376
3377 /* es route is imported into the es table */
3378 if (evp->prefix.route_type == BGP_EVPN_ES_ROUTE) {
3379
3380 /* we will match based on the entire esi to avoid
3381 * imoort of an es route for esi2 into esi1
3382 */
3383 es = bgp_evpn_lookup_es(bgp, &evp->prefix.es_addr.esi);
3384 if (es && is_es_local(es))
3385 install_uninstall_route_in_es(
3386 bgp, es, afi, safi, evp, pi, import);
3387 }
3388 }
3389
3390 return 0;
3391 }
3392
3393 /*
3394 * delete and withdraw all ipv4 and ipv6 routes in the vrf table as type-5
3395 * routes
3396 */
3397 static void delete_withdraw_vrf_routes(struct bgp *bgp_vrf)
3398 {
3399 /* delete all ipv4 routes and withdraw from peers */
3400 if (advertise_type5_routes(bgp_vrf, AFI_IP))
3401 bgp_evpn_withdraw_type5_routes(bgp_vrf, AFI_IP, SAFI_UNICAST);
3402
3403 /* delete all ipv6 routes and withdraw from peers */
3404 if (advertise_type5_routes(bgp_vrf, AFI_IP6))
3405 bgp_evpn_withdraw_type5_routes(bgp_vrf, AFI_IP6, SAFI_UNICAST);
3406 }
3407
3408 /*
3409 * update and advertise all ipv4 and ipv6 routes in thr vrf table as type-5
3410 * routes
3411 */
3412 static void update_advertise_vrf_routes(struct bgp *bgp_vrf)
3413 {
3414 /* update all ipv4 routes */
3415 if (advertise_type5_routes(bgp_vrf, AFI_IP))
3416 bgp_evpn_advertise_type5_routes(bgp_vrf, AFI_IP, SAFI_UNICAST);
3417
3418 /* update all ipv6 routes */
3419 if (advertise_type5_routes(bgp_vrf, AFI_IP6))
3420 bgp_evpn_advertise_type5_routes(bgp_vrf, AFI_IP6, SAFI_UNICAST);
3421 }
3422
3423 /*
3424 * update and advertise local routes for a VRF as type-5 routes.
3425 * This is invoked upon RD change for a VRF. Note taht the processing is only
3426 * done in the global route table using the routes which already exist in the
3427 * VRF routing table
3428 */
3429 static void update_router_id_vrf(struct bgp *bgp_vrf)
3430 {
3431 /* skip if the RD is configured */
3432 if (is_vrf_rd_configured(bgp_vrf))
3433 return;
3434
3435 /* derive the RD for the VRF based on new router-id */
3436 bgp_evpn_derive_auto_rd_for_vrf(bgp_vrf);
3437
3438 /* update advertise ipv4|ipv6 routes as type-5 routes */
3439 update_advertise_vrf_routes(bgp_vrf);
3440 }
3441
3442 /*
3443 * Delete and withdraw all type-5 routes for the RD corresponding to VRF.
3444 * This is invoked upon VRF RD change. The processing is done only from global
3445 * table.
3446 */
3447 static void withdraw_router_id_vrf(struct bgp *bgp_vrf)
3448 {
3449 /* skip if the RD is configured */
3450 if (is_vrf_rd_configured(bgp_vrf))
3451 return;
3452
3453 /* delete/withdraw ipv4|ipv6 routes as type-5 routes */
3454 delete_withdraw_vrf_routes(bgp_vrf);
3455 }
3456
3457 /*
3458 * Update and advertise local routes for a VNI. Invoked upon router-id
3459 * change. Note that the processing is done only on the global route table
3460 * using routes that already exist in the per-VNI table.
3461 */
3462 static int update_advertise_vni_routes(struct bgp *bgp, struct bgpevpn *vpn)
3463 {
3464 struct prefix_evpn p;
3465 struct bgp_node *rn, *global_rn;
3466 struct bgp_path_info *pi, *global_pi;
3467 struct attr *attr;
3468 afi_t afi = AFI_L2VPN;
3469 safi_t safi = SAFI_EVPN;
3470
3471 /* Locate type-3 route for VNI in the per-VNI table and use its
3472 * attributes to create and advertise the type-3 route for this VNI
3473 * in the global table.
3474 *
3475 * RT-3 only if doing head-end replication
3476 */
3477 if (bgp->vxlan_flood_ctrl == VXLAN_FLOOD_HEAD_END_REPL) {
3478 build_evpn_type3_prefix(&p, vpn->originator_ip);
3479 rn = bgp_node_lookup(vpn->route_table, (struct prefix *)&p);
3480 if (!rn) /* unexpected */
3481 return 0;
3482 for (pi = rn->info; pi; pi = pi->next)
3483 if (pi->peer == bgp->peer_self &&
3484 pi->type == ZEBRA_ROUTE_BGP
3485 && pi->sub_type == BGP_ROUTE_STATIC)
3486 break;
3487 if (!pi) /* unexpected */
3488 return 0;
3489 attr = pi->attr;
3490
3491 global_rn = bgp_afi_node_get(bgp->rib[afi][safi], afi, safi,
3492 (struct prefix *)&p, &vpn->prd);
3493 update_evpn_route_entry(bgp, vpn, afi, safi, global_rn, attr,
3494 1, &pi, 0, mac_mobility_seqnum(attr));
3495
3496 /* Schedule for processing and unlock node. */
3497 bgp_process(bgp, global_rn, afi, safi);
3498 bgp_unlock_node(global_rn);
3499 }
3500
3501 /* Now, walk this VNI's route table and use the route and its attribute
3502 * to create and schedule route in global table.
3503 */
3504 for (rn = bgp_table_top(vpn->route_table); rn;
3505 rn = bgp_route_next(rn)) {
3506 struct prefix_evpn *evp = (struct prefix_evpn *)&rn->p;
3507
3508 /* Identify MAC-IP local routes. */
3509 if (evp->prefix.route_type != BGP_EVPN_MAC_IP_ROUTE)
3510 continue;
3511
3512 for (pi = rn->info; pi; pi = pi->next)
3513 if (pi->peer == bgp->peer_self
3514 && pi->type == ZEBRA_ROUTE_BGP
3515 && pi->sub_type == BGP_ROUTE_STATIC)
3516 break;
3517 if (!pi)
3518 continue;
3519
3520 /* Create route in global routing table using this route entry's
3521 * attribute.
3522 */
3523 attr = pi->attr;
3524 global_rn = bgp_afi_node_get(bgp->rib[afi][safi], afi, safi,
3525 (struct prefix *)evp, &vpn->prd);
3526 assert(global_rn);
3527 update_evpn_route_entry(bgp, vpn, afi, safi, global_rn, attr, 1,
3528 &global_pi, 0,
3529 mac_mobility_seqnum(attr));
3530
3531 /* Schedule for processing and unlock node. */
3532 bgp_process(bgp, global_rn, afi, safi);
3533 bgp_unlock_node(global_rn);
3534 }
3535
3536 return 0;
3537 }
3538
3539 /*
3540 * Delete (and withdraw) local routes for a VNI - only from the global
3541 * table. Invoked upon router-id change.
3542 */
3543 static int delete_withdraw_vni_routes(struct bgp *bgp, struct bgpevpn *vpn)
3544 {
3545 int ret;
3546 struct prefix_evpn p;
3547 struct bgp_node *global_rn;
3548 struct bgp_path_info *pi;
3549 afi_t afi = AFI_L2VPN;
3550 safi_t safi = SAFI_EVPN;
3551
3552 /* Delete and withdraw locally learnt type-2 routes (MACIP)
3553 * for this VNI - from the global table.
3554 */
3555 ret = delete_global_type2_routes(bgp, vpn);
3556 if (ret)
3557 return ret;
3558
3559 /* Remove type-3 route for this VNI from global table. */
3560 build_evpn_type3_prefix(&p, vpn->originator_ip);
3561 global_rn = bgp_afi_node_lookup(bgp->rib[afi][safi], afi, safi,
3562 (struct prefix *)&p, &vpn->prd);
3563 if (global_rn) {
3564 /* Delete route entry in the global EVPN table. */
3565 delete_evpn_route_entry(bgp, afi, safi, global_rn, &pi);
3566
3567 /* Schedule for processing - withdraws to peers happen from
3568 * this table.
3569 */
3570 if (pi)
3571 bgp_process(bgp, global_rn, afi, safi);
3572 bgp_unlock_node(global_rn);
3573 }
3574
3575 return 0;
3576 }
3577
3578 /*
3579 * Handle router-id change. Update and advertise local routes corresponding
3580 * to this VNI from peers. Note that this is invoked after updating the
3581 * router-id. The routes in the per-VNI table are used to create routes in
3582 * the global table and schedule them.
3583 */
3584 static void update_router_id_vni(struct hash_backet *backet, struct bgp *bgp)
3585 {
3586 struct bgpevpn *vpn = (struct bgpevpn *)backet->data;
3587
3588 /* Skip VNIs with configured RD. */
3589 if (is_rd_configured(vpn))
3590 return;
3591
3592 bgp_evpn_derive_auto_rd(bgp, vpn);
3593 update_advertise_vni_routes(bgp, vpn);
3594 }
3595
3596 /*
3597 * Handle router-id change. Delete and withdraw local routes corresponding
3598 * to this VNI from peers. Note that this is invoked prior to updating
3599 * the router-id and is done only on the global route table, the routes
3600 * are needed in the per-VNI table to re-advertise with new router id.
3601 */
3602 static void withdraw_router_id_vni(struct hash_backet *backet, struct bgp *bgp)
3603 {
3604 struct bgpevpn *vpn = (struct bgpevpn *)backet->data;
3605
3606 /* Skip VNIs with configured RD. */
3607 if (is_rd_configured(vpn))
3608 return;
3609
3610 delete_withdraw_vni_routes(bgp, vpn);
3611 }
3612
3613 /*
3614 * Create RT-3 for a VNI and schedule for processing and advertisement.
3615 * This is invoked upon flooding mode changing to head-end replication.
3616 */
3617 static void create_advertise_type3(struct hash_backet *backet, void *data)
3618 {
3619 struct bgpevpn *vpn = backet->data;
3620 struct bgp *bgp = data;
3621 struct prefix_evpn p;
3622
3623 if (!vpn || !is_vni_live(vpn))
3624 return;
3625
3626 build_evpn_type3_prefix(&p, vpn->originator_ip);
3627 if (update_evpn_route(bgp, vpn, &p, 0, 0))
3628 flog_err(EC_BGP_EVPN_ROUTE_CREATE,
3629 "Type3 route creation failure for VNI %u", vpn->vni);
3630 }
3631
3632 /*
3633 * Delete RT-3 for a VNI and schedule for processing and withdrawal.
3634 * This is invoked upon flooding mode changing to drop BUM packets.
3635 */
3636 static void delete_withdraw_type3(struct hash_backet *backet, void *data)
3637 {
3638 struct bgpevpn *vpn = backet->data;
3639 struct bgp *bgp = data;
3640 struct prefix_evpn p;
3641
3642 if (!vpn || !is_vni_live(vpn))
3643 return;
3644
3645 build_evpn_type3_prefix(&p, vpn->originator_ip);
3646 delete_evpn_route(bgp, vpn, &p);
3647 }
3648
3649 /*
3650 * Process received EVPN type-2 route (advertise or withdraw).
3651 */
3652 static int process_type2_route(struct peer *peer, afi_t afi, safi_t safi,
3653 struct attr *attr, uint8_t *pfx, int psize,
3654 uint32_t addpath_id)
3655 {
3656 struct prefix_rd prd;
3657 struct prefix_evpn p;
3658 struct bgp_route_evpn evpn;
3659 uint8_t ipaddr_len;
3660 uint8_t macaddr_len;
3661 mpls_label_t label[BGP_MAX_LABELS]; /* holds the VNI(s) as in packet */
3662 uint32_t num_labels = 0;
3663 uint32_t eth_tag;
3664 int ret;
3665
3666 /* Type-2 route should be either 33, 37 or 49 bytes or an
3667 * additional 3 bytes if there is a second label (VNI):
3668 * RD (8), ESI (10), Eth Tag (4), MAC Addr Len (1),
3669 * MAC Addr (6), IP len (1), IP (0, 4 or 16),
3670 * MPLS Lbl1 (3), MPLS Lbl2 (0 or 3)
3671 */
3672 if (psize != 33 && psize != 37 && psize != 49 && psize != 36
3673 && psize != 40 && psize != 52) {
3674 flog_err(EC_BGP_EVPN_ROUTE_INVALID,
3675 "%u:%s - Rx EVPN Type-2 NLRI with invalid length %d",
3676 peer->bgp->vrf_id, peer->host, psize);
3677 return -1;
3678 }
3679
3680 memset(&evpn, 0, sizeof(evpn));
3681
3682 /* Make prefix_rd */
3683 prd.family = AF_UNSPEC;
3684 prd.prefixlen = 64;
3685 memcpy(&prd.val, pfx, 8);
3686 pfx += 8;
3687
3688 /* Make EVPN prefix. */
3689 memset(&p, 0, sizeof(struct prefix_evpn));
3690 p.family = AF_EVPN;
3691 p.prefixlen = EVPN_ROUTE_PREFIXLEN;
3692 p.prefix.route_type = BGP_EVPN_MAC_IP_ROUTE;
3693
3694 /* Copy Ethernet Seg Identifier */
3695 memcpy(&evpn.eth_s_id.val, pfx, ESI_LEN);
3696 pfx += ESI_LEN;
3697
3698 /* Copy Ethernet Tag */
3699 memcpy(&eth_tag, pfx, 4);
3700 p.prefix.macip_addr.eth_tag = ntohl(eth_tag);
3701 pfx += 4;
3702
3703 /* Get the MAC Addr len */
3704 macaddr_len = *pfx++;
3705
3706 /* Get the MAC Addr */
3707 if (macaddr_len == (ETH_ALEN * 8)) {
3708 memcpy(&p.prefix.macip_addr.mac.octet, pfx, ETH_ALEN);
3709 pfx += ETH_ALEN;
3710 } else {
3711 flog_err(
3712 EC_BGP_EVPN_ROUTE_INVALID,
3713 "%u:%s - Rx EVPN Type-2 NLRI with unsupported MAC address length %d",
3714 peer->bgp->vrf_id, peer->host, macaddr_len);
3715 return -1;
3716 }
3717
3718
3719 /* Get the IP. */
3720 ipaddr_len = *pfx++;
3721 if (ipaddr_len != 0 && ipaddr_len != IPV4_MAX_BITLEN
3722 && ipaddr_len != IPV6_MAX_BITLEN) {
3723 flog_err(
3724 EC_BGP_EVPN_ROUTE_INVALID,
3725 "%u:%s - Rx EVPN Type-2 NLRI with unsupported IP address length %d",
3726 peer->bgp->vrf_id, peer->host, ipaddr_len);
3727 return -1;
3728 }
3729
3730 if (ipaddr_len) {
3731 ipaddr_len /= 8; /* Convert to bytes. */
3732 p.prefix.macip_addr.ip.ipa_type = (ipaddr_len == IPV4_MAX_BYTELEN)
3733 ? IPADDR_V4
3734 : IPADDR_V6;
3735 memcpy(&p.prefix.macip_addr.ip.ip.addr, pfx, ipaddr_len);
3736 }
3737 pfx += ipaddr_len;
3738
3739 /* Get the VNI(s). Stored as bytes here. */
3740 num_labels++;
3741 memset(label, 0, sizeof(label));
3742 memcpy(&label[0], pfx, BGP_LABEL_BYTES);
3743 pfx += BGP_LABEL_BYTES;
3744 psize -= (33 + ipaddr_len);
3745 /* Do we have a second VNI? */
3746 if (psize) {
3747 num_labels++;
3748 memcpy(&label[1], pfx, BGP_LABEL_BYTES);
3749 /*
3750 * If in future, we are required to access additional fields,
3751 * we MUST increment pfx by BGP_LABEL_BYTES in before reading
3752 * the next field
3753 */
3754 }
3755
3756 /* Process the route. */
3757 if (attr)
3758 ret = bgp_update(peer, (struct prefix *)&p, addpath_id, attr,
3759 afi, safi, ZEBRA_ROUTE_BGP, BGP_ROUTE_NORMAL,
3760 &prd, &label[0], num_labels, 0, &evpn);
3761 else
3762 ret = bgp_withdraw(peer, (struct prefix *)&p, addpath_id, attr,
3763 afi, safi, ZEBRA_ROUTE_BGP, BGP_ROUTE_NORMAL,
3764 &prd, &label[0], num_labels, &evpn);
3765 return ret;
3766 }
3767
3768 /*
3769 * Process received EVPN type-3 route (advertise or withdraw).
3770 */
3771 static int process_type3_route(struct peer *peer, afi_t afi, safi_t safi,
3772 struct attr *attr, uint8_t *pfx, int psize,
3773 uint32_t addpath_id)
3774 {
3775 struct prefix_rd prd;
3776 struct prefix_evpn p;
3777 uint8_t ipaddr_len;
3778 uint32_t eth_tag;
3779 int ret;
3780
3781 /* Type-3 route should be either 17 or 29 bytes: RD (8), Eth Tag (4),
3782 * IP len (1) and IP (4 or 16).
3783 */
3784 if (psize != 17 && psize != 29) {
3785 flog_err(EC_BGP_EVPN_ROUTE_INVALID,
3786 "%u:%s - Rx EVPN Type-3 NLRI with invalid length %d",
3787 peer->bgp->vrf_id, peer->host, psize);
3788 return -1;
3789 }
3790
3791 /* If PMSI is present, log if it is anything other than IR.
3792 * Note: We just simply ignore the values as it is not clear if
3793 * doing anything else is better.
3794 */
3795 if (attr &&
3796 (attr->flag & ATTR_FLAG_BIT(BGP_ATTR_PMSI_TUNNEL))) {
3797 if (attr->pmsi_tnl_type != PMSI_TNLTYPE_INGR_REPL) {
3798 flog_warn(
3799 EC_BGP_EVPN_PMSI_PRESENT,
3800 "%u:%s - Rx EVPN Type-3 NLRI with unsupported PTA %d",
3801 peer->bgp->vrf_id, peer->host,
3802 attr->pmsi_tnl_type);
3803 }
3804 }
3805
3806 /* Make prefix_rd */
3807 prd.family = AF_UNSPEC;
3808 prd.prefixlen = 64;
3809 memcpy(&prd.val, pfx, 8);
3810 pfx += 8;
3811
3812 /* Make EVPN prefix. */
3813 memset(&p, 0, sizeof(struct prefix_evpn));
3814 p.family = AF_EVPN;
3815 p.prefixlen = EVPN_ROUTE_PREFIXLEN;
3816 p.prefix.route_type = BGP_EVPN_IMET_ROUTE;
3817
3818 /* Copy Ethernet Tag */
3819 memcpy(&eth_tag, pfx, 4);
3820 p.prefix.imet_addr.eth_tag = ntohl(eth_tag);
3821 pfx += 4;
3822
3823 /* Get the IP. */
3824 ipaddr_len = *pfx++;
3825 if (ipaddr_len == IPV4_MAX_BITLEN) {
3826 p.prefix.imet_addr.ip.ipa_type = IPADDR_V4;
3827 memcpy(&p.prefix.imet_addr.ip.ip.addr, pfx, IPV4_MAX_BYTELEN);
3828 } else {
3829 flog_err(
3830 EC_BGP_EVPN_ROUTE_INVALID,
3831 "%u:%s - Rx EVPN Type-3 NLRI with unsupported IP address length %d",
3832 peer->bgp->vrf_id, peer->host, ipaddr_len);
3833 return -1;
3834 }
3835
3836 /* Process the route. */
3837 if (attr)
3838 ret = bgp_update(peer, (struct prefix *)&p, addpath_id, attr,
3839 afi, safi, ZEBRA_ROUTE_BGP, BGP_ROUTE_NORMAL,
3840 &prd, NULL, 0, 0, NULL);
3841 else
3842 ret = bgp_withdraw(peer, (struct prefix *)&p, addpath_id, attr,
3843 afi, safi, ZEBRA_ROUTE_BGP, BGP_ROUTE_NORMAL,
3844 &prd, NULL, 0, NULL);
3845 return ret;
3846 }
3847
3848 /*
3849 * Process received EVPN type-4 route (advertise or withdraw).
3850 */
3851 static int process_type4_route(struct peer *peer, afi_t afi, safi_t safi,
3852 struct attr *attr, uint8_t *pfx, int psize,
3853 uint32_t addpath_id)
3854 {
3855 int ret;
3856 esi_t esi;
3857 uint8_t ipaddr_len;
3858 struct in_addr vtep_ip;
3859 struct prefix_rd prd;
3860 struct prefix_evpn p;
3861
3862 /* Type-4 route should be either 23 or 35 bytes
3863 * RD (8), ESI (10), ip-len (1), ip (4 or 16)
3864 */
3865 if (psize != 23 && psize != 35) {
3866 flog_err(EC_BGP_EVPN_ROUTE_INVALID,
3867 "%u:%s - Rx EVPN Type-4 NLRI with invalid length %d",
3868 peer->bgp->vrf_id, peer->host, psize);
3869 return -1;
3870 }
3871
3872 /* Make prefix_rd */
3873 prd.family = AF_UNSPEC;
3874 prd.prefixlen = 64;
3875 memcpy(&prd.val, pfx, 8);
3876 pfx += 8;
3877
3878 /* get the ESI */
3879 memcpy(&esi, pfx, ESI_BYTES);
3880 pfx += ESI_BYTES;
3881
3882
3883 /* Get the IP. */
3884 ipaddr_len = *pfx++;
3885 if (ipaddr_len == IPV4_MAX_BITLEN) {
3886 memcpy(&vtep_ip, pfx, IPV4_MAX_BYTELEN);
3887 } else {
3888 flog_err(
3889 EC_BGP_EVPN_ROUTE_INVALID,
3890 "%u:%s - Rx EVPN Type-4 NLRI with unsupported IP address length %d",
3891 peer->bgp->vrf_id, peer->host, ipaddr_len);
3892 return -1;
3893 }
3894
3895 build_evpn_type4_prefix(&p, &esi, vtep_ip);
3896 /* Process the route. */
3897 if (attr) {
3898 ret = bgp_update(peer, (struct prefix *)&p, addpath_id, attr,
3899 afi, safi, ZEBRA_ROUTE_BGP, BGP_ROUTE_NORMAL,
3900 &prd, NULL, 0, 0, NULL);
3901 } else {
3902 ret = bgp_withdraw(peer, (struct prefix *)&p, addpath_id, attr,
3903 afi, safi, ZEBRA_ROUTE_BGP, BGP_ROUTE_NORMAL,
3904 &prd, NULL, 0, NULL);
3905 }
3906 return ret;
3907 }
3908
3909
3910 /*
3911 * Process received EVPN type-5 route (advertise or withdraw).
3912 */
3913 static int process_type5_route(struct peer *peer, afi_t afi, safi_t safi,
3914 struct attr *attr, uint8_t *pfx, int psize,
3915 uint32_t addpath_id, int withdraw)
3916 {
3917 struct prefix_rd prd;
3918 struct prefix_evpn p;
3919 struct bgp_route_evpn evpn;
3920 uint8_t ippfx_len;
3921 uint32_t eth_tag;
3922 mpls_label_t label; /* holds the VNI as in the packet */
3923 int ret;
3924
3925 /* Type-5 route should be 34 or 58 bytes:
3926 * RD (8), ESI (10), Eth Tag (4), IP len (1), IP (4 or 16),
3927 * GW (4 or 16) and VNI (3).
3928 * Note that the IP and GW should both be IPv4 or both IPv6.
3929 */
3930 if (psize != 34 && psize != 58) {
3931 flog_err(EC_BGP_EVPN_ROUTE_INVALID,
3932 "%u:%s - Rx EVPN Type-5 NLRI with invalid length %d",
3933 peer->bgp->vrf_id, peer->host, psize);
3934 return -1;
3935 }
3936
3937 /* Make prefix_rd */
3938 prd.family = AF_UNSPEC;
3939 prd.prefixlen = 64;
3940 memcpy(&prd.val, pfx, 8);
3941 pfx += 8;
3942
3943 /* Make EVPN prefix. */
3944 memset(&p, 0, sizeof(struct prefix_evpn));
3945 p.family = AF_EVPN;
3946 p.prefixlen = EVPN_ROUTE_PREFIXLEN;
3947 p.prefix.route_type = BGP_EVPN_IP_PREFIX_ROUTE;
3948
3949 /* Additional information outside of prefix - ESI and GW IP */
3950 memset(&evpn, 0, sizeof(evpn));
3951
3952 /* Fetch ESI */
3953 memcpy(&evpn.eth_s_id.val, pfx, 10);
3954 pfx += 10;
3955
3956 /* Fetch Ethernet Tag. */
3957 memcpy(&eth_tag, pfx, 4);
3958 p.prefix.prefix_addr.eth_tag = ntohl(eth_tag);
3959 pfx += 4;
3960
3961 /* Fetch IP prefix length. */
3962 ippfx_len = *pfx++;
3963 if (ippfx_len > IPV6_MAX_BITLEN) {
3964 flog_err(
3965 EC_BGP_EVPN_ROUTE_INVALID,
3966 "%u:%s - Rx EVPN Type-5 NLRI with invalid IP Prefix length %d",
3967 peer->bgp->vrf_id, peer->host, ippfx_len);
3968 return -1;
3969 }
3970 p.prefix.prefix_addr.ip_prefix_length = ippfx_len;
3971
3972 /* Determine IPv4 or IPv6 prefix */
3973 /* Since the address and GW are from the same family, this just becomes
3974 * a simple check on the total size.
3975 */
3976 if (psize == 34) {
3977 SET_IPADDR_V4(&p.prefix.prefix_addr.ip);
3978 memcpy(&p.prefix.prefix_addr.ip.ipaddr_v4, pfx, 4);
3979 pfx += 4;
3980 memcpy(&evpn.gw_ip.ipv4, pfx, 4);
3981 pfx += 4;
3982 } else {
3983 SET_IPADDR_V6(&p.prefix.prefix_addr.ip);
3984 memcpy(&p.prefix.prefix_addr.ip.ipaddr_v6, pfx, 16);
3985 pfx += 16;
3986 memcpy(&evpn.gw_ip.ipv6, pfx, 16);
3987 pfx += 16;
3988 }
3989
3990 /* Get the VNI (in MPLS label field). Stored as bytes here. */
3991 memset(&label, 0, sizeof(label));
3992 memcpy(&label, pfx, BGP_LABEL_BYTES);
3993
3994 /*
3995 * If in future, we are required to access additional fields,
3996 * we MUST increment pfx by BGP_LABEL_BYTES in before reading the next
3997 * field
3998 */
3999
4000 /* Process the route. */
4001 if (!withdraw)
4002 ret = bgp_update(peer, (struct prefix *)&p, addpath_id, attr,
4003 afi, safi, ZEBRA_ROUTE_BGP, BGP_ROUTE_NORMAL,
4004 &prd, &label, 1, 0, &evpn);
4005 else
4006 ret = bgp_withdraw(peer, (struct prefix *)&p, addpath_id, attr,
4007 afi, safi, ZEBRA_ROUTE_BGP, BGP_ROUTE_NORMAL,
4008 &prd, &label, 1, &evpn);
4009
4010 return ret;
4011 }
4012
4013 static void evpn_mpattr_encode_type5(struct stream *s, struct prefix *p,
4014 struct prefix_rd *prd, mpls_label_t *label,
4015 uint32_t num_labels, struct attr *attr)
4016 {
4017 int len;
4018 char temp[16];
4019 struct evpn_addr *p_evpn_p;
4020
4021 memset(&temp, 0, 16);
4022 if (p->family != AF_EVPN)
4023 return;
4024 p_evpn_p = &(p->u.prefix_evpn);
4025
4026 /* len denites the total len of IP and GW-IP in the route
4027 IP and GW-IP have to be both ipv4 or ipv6
4028 */
4029 if (IS_IPADDR_V4(&p_evpn_p->prefix_addr.ip))
4030 len = 8; /* IP and GWIP are both ipv4 */
4031 else
4032 len = 32; /* IP and GWIP are both ipv6 */
4033 /* Prefix contains RD, ESI, EthTag, IP length, IP, GWIP and VNI */
4034 stream_putc(s, 8 + 10 + 4 + 1 + len + 3);
4035 stream_put(s, prd->val, 8);
4036 if (attr)
4037 stream_put(s, &(attr->evpn_overlay.eth_s_id), 10);
4038 else
4039 stream_put(s, &temp, 10);
4040 stream_putl(s, p_evpn_p->prefix_addr.eth_tag);
4041 stream_putc(s, p_evpn_p->prefix_addr.ip_prefix_length);
4042 if (IS_IPADDR_V4(&p_evpn_p->prefix_addr.ip))
4043 stream_put_ipv4(s, p_evpn_p->prefix_addr.ip.ipaddr_v4.s_addr);
4044 else
4045 stream_put(s, &p_evpn_p->prefix_addr.ip.ipaddr_v6, 16);
4046 if (attr) {
4047 if (IS_IPADDR_V4(&p_evpn_p->prefix_addr.ip))
4048 stream_put_ipv4(s,
4049 attr->evpn_overlay.gw_ip.ipv4.s_addr);
4050 else
4051 stream_put(s, &(attr->evpn_overlay.gw_ip.ipv6), 16);
4052 } else {
4053 if (IS_IPADDR_V4(&p_evpn_p->prefix_addr.ip))
4054 stream_put_ipv4(s, 0);
4055 else
4056 stream_put(s, &temp, 16);
4057 }
4058
4059 if (num_labels)
4060 stream_put(s, label, 3);
4061 else
4062 stream_put3(s, 0);
4063 }
4064
4065 /*
4066 * Cleanup specific VNI upon EVPN (advertise-all-vni) being disabled.
4067 */
4068 static void cleanup_vni_on_disable(struct hash_backet *backet, struct bgp *bgp)
4069 {
4070 struct bgpevpn *vpn = (struct bgpevpn *)backet->data;
4071
4072 /* Remove EVPN routes and schedule for processing. */
4073 delete_routes_for_vni(bgp, vpn);
4074
4075 /* Clear "live" flag and see if hash needs to be freed. */
4076 UNSET_FLAG(vpn->flags, VNI_FLAG_LIVE);
4077 if (!is_vni_configured(vpn))
4078 bgp_evpn_free(bgp, vpn);
4079 }
4080
4081 /*
4082 * Free a VNI entry; iterator function called during cleanup.
4083 */
4084 static void free_vni_entry(struct hash_backet *backet, struct bgp *bgp)
4085 {
4086 struct bgpevpn *vpn = (struct bgpevpn *)backet->data;
4087
4088 delete_all_vni_routes(bgp, vpn);
4089 bgp_evpn_free(bgp, vpn);
4090 }
4091
4092 /*
4093 * Derive AUTO import RT for BGP VRF - L3VNI
4094 */
4095 static void evpn_auto_rt_import_add_for_vrf(struct bgp *bgp_vrf)
4096 {
4097 struct bgp *bgp_def = NULL;
4098
4099 form_auto_rt(bgp_vrf, bgp_vrf->l3vni, bgp_vrf->vrf_import_rtl);
4100 UNSET_FLAG(bgp_vrf->vrf_flags, BGP_VRF_IMPORT_RT_CFGD);
4101
4102 /* Map RT to VRF */
4103 bgp_def = bgp_get_default();
4104 if (!bgp_def)
4105 return;
4106 bgp_evpn_map_vrf_to_its_rts(bgp_vrf);
4107 }
4108
4109 /*
4110 * Delete AUTO import RT from BGP VRF - L3VNI
4111 */
4112 static void evpn_auto_rt_import_delete_for_vrf(struct bgp *bgp_vrf)
4113 {
4114 evpn_rt_delete_auto(bgp_vrf, bgp_vrf->l3vni, bgp_vrf->vrf_import_rtl);
4115 }
4116
4117 /*
4118 * Derive AUTO export RT for BGP VRF - L3VNI
4119 */
4120 static void evpn_auto_rt_export_add_for_vrf(struct bgp *bgp_vrf)
4121 {
4122 UNSET_FLAG(bgp_vrf->vrf_flags, BGP_VRF_EXPORT_RT_CFGD);
4123 form_auto_rt(bgp_vrf, bgp_vrf->l3vni, bgp_vrf->vrf_export_rtl);
4124 }
4125
4126 /*
4127 * Delete AUTO export RT from BGP VRF - L3VNI
4128 */
4129 static void evpn_auto_rt_export_delete_for_vrf(struct bgp *bgp_vrf)
4130 {
4131 evpn_rt_delete_auto(bgp_vrf, bgp_vrf->l3vni, bgp_vrf->vrf_export_rtl);
4132 }
4133
4134 static void bgp_evpn_handle_export_rt_change_for_vrf(struct bgp *bgp_vrf)
4135 {
4136 struct bgp *bgp_def = NULL;
4137 struct listnode *node = NULL;
4138 struct bgpevpn *vpn = NULL;
4139
4140 bgp_def = bgp_get_default();
4141 if (!bgp_def)
4142 return;
4143
4144 /* update all type-5 routes */
4145 update_advertise_vrf_routes(bgp_vrf);
4146
4147 /* update all type-2 routes */
4148 for (ALL_LIST_ELEMENTS_RO(bgp_vrf->l2vnis, node, vpn))
4149 update_routes_for_vni(bgp_def, vpn);
4150 }
4151
4152 /*
4153 * Handle autort change for a given VNI.
4154 */
4155 static void update_autort_vni(struct hash_backet *backet, struct bgp *bgp)
4156 {
4157 struct bgpevpn *vpn = backet->data;
4158
4159 if (!is_import_rt_configured(vpn)) {
4160 if (is_vni_live(vpn))
4161 bgp_evpn_uninstall_routes(bgp, vpn);
4162 bgp_evpn_unmap_vni_from_its_rts(bgp, vpn);
4163 list_delete_all_node(vpn->import_rtl);
4164 bgp_evpn_derive_auto_rt_import(bgp, vpn);
4165 if (is_vni_live(vpn))
4166 bgp_evpn_install_routes(bgp, vpn);
4167 }
4168 if (!is_export_rt_configured(vpn)) {
4169 list_delete_all_node(vpn->export_rtl);
4170 bgp_evpn_derive_auto_rt_export(bgp, vpn);
4171 if (is_vni_live(vpn))
4172 bgp_evpn_handle_export_rt_change(bgp, vpn);
4173 }
4174 }
4175
4176 /*
4177 * Public functions.
4178 */
4179
4180 /* withdraw type-5 route corresponding to ip prefix */
4181 void bgp_evpn_withdraw_type5_route(struct bgp *bgp_vrf, struct prefix *p,
4182 afi_t afi, safi_t safi)
4183 {
4184 int ret = 0;
4185 struct prefix_evpn evp;
4186 char buf[PREFIX_STRLEN];
4187
4188 build_type5_prefix_from_ip_prefix(&evp, p);
4189 ret = delete_evpn_type5_route(bgp_vrf, &evp);
4190 if (ret) {
4191 flog_err(
4192 EC_BGP_EVPN_ROUTE_DELETE,
4193 "%u failed to delete type-5 route for prefix %s in vrf %s",
4194 bgp_vrf->vrf_id, prefix2str(p, buf, sizeof(buf)),
4195 vrf_id_to_name(bgp_vrf->vrf_id));
4196 }
4197 }
4198
4199 /* withdraw all type-5 routes for an address family */
4200 void bgp_evpn_withdraw_type5_routes(struct bgp *bgp_vrf, afi_t afi, safi_t safi)
4201 {
4202 struct bgp_table *table = NULL;
4203 struct bgp_node *rn = NULL;
4204 struct bgp_path_info *pi;
4205
4206 table = bgp_vrf->rib[afi][safi];
4207 for (rn = bgp_table_top(table); rn; rn = bgp_route_next(rn)) {
4208 /* Only care about "selected" routes - non-imported. */
4209 /* TODO: Support for AddPath for EVPN. */
4210 for (pi = rn->info; pi; pi = pi->next) {
4211 if (CHECK_FLAG(pi->flags, BGP_PATH_SELECTED)
4212 && (!pi->extra || !pi->extra->parent)) {
4213 bgp_evpn_withdraw_type5_route(bgp_vrf, &rn->p,
4214 afi, safi);
4215 break;
4216 }
4217 }
4218 }
4219 }
4220
4221 /*
4222 * Advertise IP prefix as type-5 route. The afi/safi and src_attr passed
4223 * to this function correspond to those of the source IP prefix (best
4224 * path in the case of the attr. In the case of a local prefix (when we
4225 * are advertising local subnets), the src_attr will be NULL.
4226 */
4227 void bgp_evpn_advertise_type5_route(struct bgp *bgp_vrf, struct prefix *p,
4228 struct attr *src_attr, afi_t afi,
4229 safi_t safi)
4230 {
4231 int ret = 0;
4232 struct prefix_evpn evp;
4233 char buf[PREFIX_STRLEN];
4234
4235 build_type5_prefix_from_ip_prefix(&evp, p);
4236 ret = update_evpn_type5_route(bgp_vrf, &evp, src_attr);
4237 if (ret)
4238 flog_err(EC_BGP_EVPN_ROUTE_CREATE,
4239 "%u: Failed to create type-5 route for prefix %s",
4240 bgp_vrf->vrf_id, prefix2str(p, buf, sizeof(buf)));
4241 }
4242
4243 /* Inject all prefixes of a particular address-family (currently, IPv4 or
4244 * IPv6 unicast) into EVPN as type-5 routes. This is invoked when the
4245 * advertisement is enabled.
4246 */
4247 void bgp_evpn_advertise_type5_routes(struct bgp *bgp_vrf, afi_t afi,
4248 safi_t safi)
4249 {
4250 struct bgp_table *table = NULL;
4251 struct bgp_node *rn = NULL;
4252 struct bgp_path_info *pi;
4253
4254 table = bgp_vrf->rib[afi][safi];
4255 for (rn = bgp_table_top(table); rn; rn = bgp_route_next(rn)) {
4256 /* Need to identify the "selected" route entry to use its
4257 * attribute. Also, we only consider "non-imported" routes.
4258 * TODO: Support for AddPath for EVPN.
4259 */
4260 for (pi = rn->info; pi; pi = pi->next) {
4261 if (CHECK_FLAG(pi->flags, BGP_PATH_SELECTED)
4262 && (!pi->extra || !pi->extra->parent)) {
4263
4264 /* apply the route-map */
4265 if (bgp_vrf->adv_cmd_rmap[afi][safi].map) {
4266 int ret = 0;
4267
4268 ret = route_map_apply(
4269 bgp_vrf->adv_cmd_rmap[afi][safi]
4270 .map,
4271 &rn->p, RMAP_BGP, pi);
4272 if (ret == RMAP_DENYMATCH)
4273 continue;
4274 }
4275 bgp_evpn_advertise_type5_route(
4276 bgp_vrf, &rn->p, pi->attr, afi, safi);
4277 break;
4278 }
4279 }
4280 }
4281 }
4282
4283 void evpn_rt_delete_auto(struct bgp *bgp, vni_t vni, struct list *rtl)
4284 {
4285 struct listnode *node, *nnode, *node_to_del;
4286 struct ecommunity *ecom, *ecom_auto;
4287 struct ecommunity_val eval;
4288
4289 if (bgp->advertise_autort_rfc8365)
4290 vni |= EVPN_AUTORT_VXLAN;
4291 encode_route_target_as((bgp->as & 0xFFFF), vni, &eval);
4292
4293 ecom_auto = ecommunity_new();
4294 ecommunity_add_val(ecom_auto, &eval);
4295 node_to_del = NULL;
4296
4297 for (ALL_LIST_ELEMENTS(rtl, node, nnode, ecom)) {
4298 if (ecommunity_match(ecom, ecom_auto)) {
4299 ecommunity_free(&ecom);
4300 node_to_del = node;
4301 }
4302 }
4303
4304 if (node_to_del)
4305 list_delete_node(rtl, node_to_del);
4306
4307 ecommunity_free(&ecom_auto);
4308 }
4309
4310 void bgp_evpn_configure_import_rt_for_vrf(struct bgp *bgp_vrf,
4311 struct ecommunity *ecomadd)
4312 {
4313 /* uninstall routes from vrf */
4314 uninstall_routes_for_vrf(bgp_vrf);
4315
4316 /* Cleanup the RT to VRF mapping */
4317 bgp_evpn_unmap_vrf_from_its_rts(bgp_vrf);
4318
4319 /* Remove auto generated RT */
4320 evpn_auto_rt_import_delete_for_vrf(bgp_vrf);
4321
4322 /* Add the newly configured RT to RT list */
4323 listnode_add_sort(bgp_vrf->vrf_import_rtl, ecomadd);
4324 SET_FLAG(bgp_vrf->vrf_flags, BGP_VRF_IMPORT_RT_CFGD);
4325
4326 /* map VRF to its RTs */
4327 bgp_evpn_map_vrf_to_its_rts(bgp_vrf);
4328
4329 /* install routes matching the new VRF */
4330 install_routes_for_vrf(bgp_vrf);
4331 }
4332
4333 void bgp_evpn_unconfigure_import_rt_for_vrf(struct bgp *bgp_vrf,
4334 struct ecommunity *ecomdel)
4335 {
4336 struct listnode *node = NULL, *nnode = NULL, *node_to_del = NULL;
4337 struct ecommunity *ecom = NULL;
4338
4339 /* uninstall routes from vrf */
4340 uninstall_routes_for_vrf(bgp_vrf);
4341
4342 /* Cleanup the RT to VRF mapping */
4343 bgp_evpn_unmap_vrf_from_its_rts(bgp_vrf);
4344
4345 /* remove the RT from the RT list */
4346 for (ALL_LIST_ELEMENTS(bgp_vrf->vrf_import_rtl, node, nnode, ecom)) {
4347 if (ecommunity_match(ecom, ecomdel)) {
4348 ecommunity_free(&ecom);
4349 node_to_del = node;
4350 break;
4351 }
4352 }
4353
4354 if (node_to_del)
4355 list_delete_node(bgp_vrf->vrf_import_rtl, node_to_del);
4356
4357 assert(bgp_vrf->vrf_import_rtl);
4358 /* fallback to auto import rt, if this was the last RT */
4359 if (bgp_vrf->vrf_import_rtl && list_isempty(bgp_vrf->vrf_import_rtl)) {
4360 UNSET_FLAG(bgp_vrf->vrf_flags, BGP_VRF_IMPORT_RT_CFGD);
4361 evpn_auto_rt_import_add_for_vrf(bgp_vrf);
4362 }
4363
4364 /* map VRFs to its RTs */
4365 bgp_evpn_map_vrf_to_its_rts(bgp_vrf);
4366
4367 /* install routes matching this new RT */
4368 install_routes_for_vrf(bgp_vrf);
4369 }
4370
4371 void bgp_evpn_configure_export_rt_for_vrf(struct bgp *bgp_vrf,
4372 struct ecommunity *ecomadd)
4373 {
4374 /* remove auto-generated RT */
4375 evpn_auto_rt_export_delete_for_vrf(bgp_vrf);
4376
4377 /* Add the new RT to the RT list */
4378 listnode_add_sort(bgp_vrf->vrf_export_rtl, ecomadd);
4379 SET_FLAG(bgp_vrf->vrf_flags, BGP_VRF_EXPORT_RT_CFGD);
4380
4381 bgp_evpn_handle_export_rt_change_for_vrf(bgp_vrf);
4382 }
4383
4384 void bgp_evpn_unconfigure_export_rt_for_vrf(struct bgp *bgp_vrf,
4385 struct ecommunity *ecomdel)
4386 {
4387 struct listnode *node = NULL, *nnode = NULL, *node_to_del = NULL;
4388 struct ecommunity *ecom = NULL;
4389
4390 /* Remove the RT from the RT list */
4391 for (ALL_LIST_ELEMENTS(bgp_vrf->vrf_export_rtl, node, nnode, ecom)) {
4392 if (ecommunity_match(ecom, ecomdel)) {
4393 ecommunity_free(&ecom);
4394 node_to_del = node;
4395 break;
4396 }
4397 }
4398
4399 if (node_to_del)
4400 list_delete_node(bgp_vrf->vrf_export_rtl, node_to_del);
4401
4402 /*
4403 * Temporary assert to make SA happy.
4404 * The ALL_LIST_ELEMENTS macro above has a NULL check
4405 * which means that SA is going to complain about
4406 * the list_isempty call, which doesn't NULL check.
4407 * So until we get this situation cleaned up, here
4408 * we are.
4409 */
4410 assert(bgp_vrf->vrf_export_rtl);
4411
4412 /* fall back to auto-generated RT if this was the last RT */
4413 if (list_isempty(bgp_vrf->vrf_export_rtl)) {
4414 UNSET_FLAG(bgp_vrf->vrf_flags, BGP_VRF_EXPORT_RT_CFGD);
4415 evpn_auto_rt_export_add_for_vrf(bgp_vrf);
4416 }
4417
4418 bgp_evpn_handle_export_rt_change_for_vrf(bgp_vrf);
4419 }
4420
4421 /*
4422 * Handle change to BGP router id. This is invoked twice by the change
4423 * handler, first before the router id has been changed and then after
4424 * the router id has been changed. The first invocation will result in
4425 * local routes for all VNIs/VRF being deleted and withdrawn and the next
4426 * will result in the routes being re-advertised.
4427 */
4428 void bgp_evpn_handle_router_id_update(struct bgp *bgp, int withdraw)
4429 {
4430 if (withdraw) {
4431
4432 /* delete and withdraw all the type-5 routes
4433 stored in the global table for this vrf
4434 */
4435 withdraw_router_id_vrf(bgp);
4436
4437 /* delete all the VNI routes (type-2/type-3) routes for all the
4438 * L2-VNIs
4439 */
4440 hash_iterate(bgp->vnihash,
4441 (void (*)(struct hash_backet *,
4442 void *))withdraw_router_id_vni,
4443 bgp);
4444 } else {
4445
4446 /* advertise all routes in the vrf as type-5 routes with the new
4447 * RD
4448 */
4449 update_router_id_vrf(bgp);
4450
4451 /* advertise all the VNI routes (type-2/type-3) routes with the
4452 * new RD
4453 */
4454 hash_iterate(bgp->vnihash,
4455 (void (*)(struct hash_backet *,
4456 void *))update_router_id_vni,
4457 bgp);
4458 }
4459 }
4460
4461 /*
4462 * Handle change to auto-RT algorithm - update and advertise local routes.
4463 */
4464 void bgp_evpn_handle_autort_change(struct bgp *bgp)
4465 {
4466 hash_iterate(bgp->vnihash,
4467 (void (*)(struct hash_backet *,
4468 void*))update_autort_vni,
4469 bgp);
4470 }
4471
4472 /*
4473 * Handle change to export RT - update and advertise local routes.
4474 */
4475 int bgp_evpn_handle_export_rt_change(struct bgp *bgp, struct bgpevpn *vpn)
4476 {
4477 return update_routes_for_vni(bgp, vpn);
4478 }
4479
4480 void bgp_evpn_handle_vrf_rd_change(struct bgp *bgp_vrf, int withdraw)
4481 {
4482 if (withdraw)
4483 delete_withdraw_vrf_routes(bgp_vrf);
4484 else
4485 update_advertise_vrf_routes(bgp_vrf);
4486 }
4487
4488 /*
4489 * Handle change to RD. This is invoked twice by the change handler,
4490 * first before the RD has been changed and then after the RD has
4491 * been changed. The first invocation will result in local routes
4492 * of this VNI being deleted and withdrawn and the next will result
4493 * in the routes being re-advertised.
4494 */
4495 void bgp_evpn_handle_rd_change(struct bgp *bgp, struct bgpevpn *vpn,
4496 int withdraw)
4497 {
4498 if (withdraw)
4499 delete_withdraw_vni_routes(bgp, vpn);
4500 else
4501 update_advertise_vni_routes(bgp, vpn);
4502 }
4503
4504 /*
4505 * Install routes for this VNI. Invoked upon change to Import RT.
4506 */
4507 int bgp_evpn_install_routes(struct bgp *bgp, struct bgpevpn *vpn)
4508 {
4509 return install_routes_for_vni(bgp, vpn);
4510 }
4511
4512 /*
4513 * Uninstall all routes installed for this VNI. Invoked upon change
4514 * to Import RT.
4515 */
4516 int bgp_evpn_uninstall_routes(struct bgp *bgp, struct bgpevpn *vpn)
4517 {
4518 return uninstall_routes_for_vni(bgp, vpn);
4519 }
4520
4521 /*
4522 * TODO: Hardcoded for a maximum of 2 VNIs right now
4523 */
4524 char *bgp_evpn_label2str(mpls_label_t *label, uint32_t num_labels, char *buf,
4525 int len)
4526 {
4527 vni_t vni1, vni2;
4528
4529 vni1 = label2vni(label);
4530 if (num_labels == 2) {
4531 vni2 = label2vni(label + 1);
4532 snprintf(buf, len, "%u/%u", vni1, vni2);
4533 } else
4534 snprintf(buf, len, "%u", vni1);
4535 return buf;
4536 }
4537
4538 /*
4539 * Function to convert evpn route to json format.
4540 * NOTE: We don't use prefix2str as the output here is a bit different.
4541 */
4542 void bgp_evpn_route2json(struct prefix_evpn *p, json_object *json)
4543 {
4544 char buf1[ETHER_ADDR_STRLEN];
4545 char buf2[PREFIX2STR_BUFFER];
4546
4547 if (!json)
4548 return;
4549
4550 if (p->prefix.route_type == BGP_EVPN_IMET_ROUTE) {
4551 json_object_int_add(json, "routeType", p->prefix.route_type);
4552 json_object_int_add(json, "ethTag",
4553 p->prefix.imet_addr.eth_tag);
4554 json_object_int_add(json, "ipLen",
4555 is_evpn_prefix_ipaddr_v4(p)
4556 ? IPV4_MAX_BITLEN
4557 : IPV6_MAX_BITLEN);
4558 json_object_string_add(json, "ip",
4559 inet_ntoa(p->prefix.imet_addr.ip.ipaddr_v4));
4560 } else if (p->prefix.route_type == BGP_EVPN_MAC_IP_ROUTE) {
4561 if (is_evpn_prefix_ipaddr_none(p)) {
4562 json_object_int_add(json, "routeType",
4563 p->prefix.route_type);
4564 json_object_int_add(json, "ethTag",
4565 p->prefix.macip_addr.eth_tag);
4566 json_object_int_add(json, "macLen", 8 * ETH_ALEN);
4567 json_object_string_add(json, "mac",
4568 prefix_mac2str(&p->prefix.macip_addr.mac,
4569 buf1,
4570 sizeof(buf1)));
4571 } else {
4572 uint8_t family;
4573
4574 family = is_evpn_prefix_ipaddr_v4(p) ? AF_INET
4575 : AF_INET6;
4576
4577 json_object_int_add(json, "routeType",
4578 p->prefix.route_type);
4579 json_object_int_add(json, "ethTag",
4580 p->prefix.macip_addr.eth_tag);
4581 json_object_int_add(json, "macLen", 8 * ETH_ALEN);
4582 json_object_string_add(json, "mac",
4583 prefix_mac2str(&p->prefix.macip_addr.mac,
4584 buf1,
4585 sizeof(buf1)));
4586 json_object_int_add(json, "ipLen",
4587 is_evpn_prefix_ipaddr_v4(p)
4588 ? IPV4_MAX_BITLEN
4589 : IPV6_MAX_BITLEN);
4590 json_object_string_add(
4591 json, "ip",
4592 inet_ntop(family,
4593 &p->prefix.macip_addr.ip.ip.addr,
4594 buf2,
4595 PREFIX2STR_BUFFER));
4596 }
4597 } else {
4598 /* Currently, this is to cater to other AF_ETHERNET code. */
4599 }
4600 }
4601
4602 /*
4603 * Function to convert evpn route to string.
4604 * NOTE: We don't use prefix2str as the output here is a bit different.
4605 */
4606 char *bgp_evpn_route2str(struct prefix_evpn *p, char *buf, int len)
4607 {
4608 char buf1[ETHER_ADDR_STRLEN];
4609 char buf2[PREFIX2STR_BUFFER];
4610 char buf3[ESI_STR_LEN];
4611
4612 if (p->prefix.route_type == BGP_EVPN_IMET_ROUTE) {
4613 snprintf(buf, len, "[%d]:[%d]:[%d]:[%s]", p->prefix.route_type,
4614 p->prefix.imet_addr.eth_tag,
4615 is_evpn_prefix_ipaddr_v4(p) ? IPV4_MAX_BITLEN
4616 : IPV6_MAX_BITLEN,
4617 inet_ntoa(p->prefix.imet_addr.ip.ipaddr_v4));
4618 } else if (p->prefix.route_type == BGP_EVPN_MAC_IP_ROUTE) {
4619 if (is_evpn_prefix_ipaddr_none(p))
4620 snprintf(buf, len, "[%d]:[%d]:[%d]:[%s]",
4621 p->prefix.route_type,
4622 p->prefix.macip_addr.eth_tag,
4623 8 * ETH_ALEN,
4624 prefix_mac2str(&p->prefix.macip_addr.mac, buf1,
4625 sizeof(buf1)));
4626 else {
4627 uint8_t family;
4628
4629 family = is_evpn_prefix_ipaddr_v4(p) ? AF_INET
4630 : AF_INET6;
4631 snprintf(buf, len, "[%d]:[%d]:[%d]:[%s]:[%d]:[%s]",
4632 p->prefix.route_type,
4633 p->prefix.macip_addr.eth_tag,
4634 8 * ETH_ALEN,
4635 prefix_mac2str(&p->prefix.macip_addr.mac, buf1,
4636 sizeof(buf1)),
4637 family == AF_INET ? IPV4_MAX_BITLEN
4638 : IPV6_MAX_BITLEN,
4639 inet_ntop(family,
4640 &p->prefix.macip_addr.ip.ip.addr,
4641 buf2,
4642 PREFIX2STR_BUFFER));
4643 }
4644 } else if (p->prefix.route_type == BGP_EVPN_IP_PREFIX_ROUTE) {
4645 snprintf(buf, len, "[%d]:[%d]:[%d]:[%s]",
4646 p->prefix.route_type,
4647 p->prefix.prefix_addr.eth_tag,
4648 p->prefix.prefix_addr.ip_prefix_length,
4649 is_evpn_prefix_ipaddr_v4(p)
4650 ? inet_ntoa(p->prefix.prefix_addr.ip.ipaddr_v4)
4651 : inet6_ntoa(p->prefix.prefix_addr.ip.ipaddr_v6));
4652 } else if (p->prefix.route_type == BGP_EVPN_ES_ROUTE) {
4653 snprintf(buf, len, "[%d]:[%s]:[%d]:[%s]",
4654 p->prefix.route_type,
4655 esi_to_str(&p->prefix.es_addr.esi, buf3, sizeof(buf3)),
4656 is_evpn_prefix_ipaddr_v4(p) ? IPV4_MAX_BITLEN
4657 : IPV6_MAX_BITLEN,
4658 inet_ntoa(p->prefix.es_addr.ip.ipaddr_v4));
4659 } else {
4660 /* For EVPN route types not supported yet. */
4661 snprintf(buf, len, "(unsupported route type %d)",
4662 p->prefix.route_type);
4663 }
4664
4665 return (buf);
4666 }
4667
4668 /*
4669 * Encode EVPN prefix in Update (MP_REACH)
4670 */
4671 void bgp_evpn_encode_prefix(struct stream *s, struct prefix *p,
4672 struct prefix_rd *prd, mpls_label_t *label,
4673 uint32_t num_labels, struct attr *attr,
4674 int addpath_encode, uint32_t addpath_tx_id)
4675 {
4676 struct prefix_evpn *evp = (struct prefix_evpn *)p;
4677 int len, ipa_len = 0;
4678
4679 if (addpath_encode)
4680 stream_putl(s, addpath_tx_id);
4681
4682 /* Route type */
4683 stream_putc(s, evp->prefix.route_type);
4684
4685 switch (evp->prefix.route_type) {
4686 case BGP_EVPN_MAC_IP_ROUTE:
4687 if (is_evpn_prefix_ipaddr_v4(evp))
4688 ipa_len = IPV4_MAX_BYTELEN;
4689 else if (is_evpn_prefix_ipaddr_v6(evp))
4690 ipa_len = IPV6_MAX_BYTELEN;
4691 /* RD, ESI, EthTag, MAC+len, IP len, [IP], 1 VNI */
4692 len = 8 + 10 + 4 + 1 + 6 + 1 + ipa_len + 3;
4693 if (ipa_len && num_labels > 1) /* There are 2 VNIs */
4694 len += 3;
4695 stream_putc(s, len);
4696 stream_put(s, prd->val, 8); /* RD */
4697 if (attr)
4698 stream_put(s, &attr->evpn_overlay.eth_s_id, ESI_LEN);
4699 else
4700 stream_put(s, 0, 10);
4701 stream_putl(s, evp->prefix.macip_addr.eth_tag); /* Ethernet Tag ID */
4702 stream_putc(s, 8 * ETH_ALEN); /* Mac Addr Len - bits */
4703 stream_put(s, evp->prefix.macip_addr.mac.octet, 6); /* Mac Addr */
4704 stream_putc(s, 8 * ipa_len); /* IP address Length */
4705 if (ipa_len) /* IP */
4706 stream_put(s, &evp->prefix.macip_addr.ip.ip.addr,
4707 ipa_len);
4708 /* 1st label is the L2 VNI */
4709 stream_put(s, label, BGP_LABEL_BYTES);
4710 /* Include 2nd label (L3 VNI) if advertising MAC+IP */
4711 if (ipa_len && num_labels > 1)
4712 stream_put(s, label + 1, BGP_LABEL_BYTES);
4713 break;
4714
4715 case BGP_EVPN_IMET_ROUTE:
4716 stream_putc(s, 17); // TODO: length - assumes IPv4 address
4717 stream_put(s, prd->val, 8); /* RD */
4718 stream_putl(s, evp->prefix.imet_addr.eth_tag); /* Ethernet Tag ID */
4719 stream_putc(s, IPV4_MAX_BITLEN); /* IP address Length - bits */
4720 /* Originating Router's IP Addr */
4721 stream_put_in_addr(s, &evp->prefix.imet_addr.ip.ipaddr_v4);
4722 break;
4723
4724 case BGP_EVPN_ES_ROUTE:
4725 stream_putc(s, 23); /* TODO: length: assumes ipv4 VTEP */
4726 stream_put(s, prd->val, 8); /* RD */
4727 stream_put(s, evp->prefix.es_addr.esi.val, 10); /* ESI */
4728 stream_putc(s, IPV4_MAX_BITLEN); /* IP address Length - bits */
4729 /* VTEP IP */
4730 stream_put_in_addr(s, &evp->prefix.es_addr.ip.ipaddr_v4);
4731 break;
4732
4733 case BGP_EVPN_IP_PREFIX_ROUTE:
4734 /* TODO: AddPath support. */
4735 evpn_mpattr_encode_type5(s, p, prd, label, num_labels, attr);
4736 break;
4737
4738 default:
4739 break;
4740 }
4741 }
4742
4743 int bgp_nlri_parse_evpn(struct peer *peer, struct attr *attr,
4744 struct bgp_nlri *packet, int withdraw)
4745 {
4746 uint8_t *pnt;
4747 uint8_t *lim;
4748 afi_t afi;
4749 safi_t safi;
4750 uint32_t addpath_id;
4751 int addpath_encoded;
4752 int psize = 0;
4753 uint8_t rtype;
4754 struct prefix p;
4755
4756 /* Start processing the NLRI - there may be multiple in the MP_REACH */
4757 pnt = packet->nlri;
4758 lim = pnt + packet->length;
4759 afi = packet->afi;
4760 safi = packet->safi;
4761 addpath_id = 0;
4762
4763 addpath_encoded =
4764 (CHECK_FLAG(peer->af_cap[afi][safi], PEER_CAP_ADDPATH_AF_RX_ADV)
4765 && CHECK_FLAG(peer->af_cap[afi][safi],
4766 PEER_CAP_ADDPATH_AF_TX_RCV));
4767
4768 for (; pnt < lim; pnt += psize) {
4769 /* Clear prefix structure. */
4770 memset(&p, 0, sizeof(struct prefix));
4771
4772 /* Deal with path-id if AddPath is supported. */
4773 if (addpath_encoded) {
4774 /* When packet overflow occurs return immediately. */
4775 if (pnt + BGP_ADDPATH_ID_LEN > lim)
4776 return -1;
4777
4778 addpath_id = ntohl(*((uint32_t *)pnt));
4779 pnt += BGP_ADDPATH_ID_LEN;
4780 }
4781
4782 /* All EVPN NLRI types start with type and length. */
4783 if (pnt + 2 > lim)
4784 return -1;
4785
4786 rtype = *pnt++;
4787 psize = *pnt++;
4788
4789 /* When packet overflow occur return immediately. */
4790 if (pnt + psize > lim)
4791 return -1;
4792
4793 switch (rtype) {
4794 case BGP_EVPN_MAC_IP_ROUTE:
4795 if (process_type2_route(peer, afi, safi,
4796 withdraw ? NULL : attr, pnt,
4797 psize, addpath_id)) {
4798 flog_err(
4799 EC_BGP_EVPN_FAIL,
4800 "%u:%s - Error in processing EVPN type-2 NLRI size %d",
4801 peer->bgp->vrf_id, peer->host, psize);
4802 return -1;
4803 }
4804 break;
4805
4806 case BGP_EVPN_IMET_ROUTE:
4807 if (process_type3_route(peer, afi, safi,
4808 withdraw ? NULL : attr, pnt,
4809 psize, addpath_id)) {
4810 flog_err(
4811 EC_BGP_PKT_PROCESS,
4812 "%u:%s - Error in processing EVPN type-3 NLRI size %d",
4813 peer->bgp->vrf_id, peer->host, psize);
4814 return -1;
4815 }
4816 break;
4817
4818 case BGP_EVPN_ES_ROUTE:
4819 if (process_type4_route(peer, afi, safi,
4820 withdraw ? NULL : attr, pnt,
4821 psize, addpath_id)) {
4822 flog_err(
4823 EC_BGP_PKT_PROCESS,
4824 "%u:%s - Error in processing EVPN type-4 NLRI size %d",
4825 peer->bgp->vrf_id, peer->host, psize);
4826 return -1;
4827 }
4828 break;
4829
4830 case BGP_EVPN_IP_PREFIX_ROUTE:
4831 if (process_type5_route(peer, afi, safi, attr, pnt,
4832 psize, addpath_id, withdraw)) {
4833 flog_err(
4834 EC_BGP_PKT_PROCESS,
4835 "%u:%s - Error in processing EVPN type-5 NLRI size %d",
4836 peer->bgp->vrf_id, peer->host, psize);
4837 return -1;
4838 }
4839 break;
4840
4841 default:
4842 break;
4843 }
4844 }
4845
4846 /* Packet length consistency check. */
4847 if (pnt != lim)
4848 return -1;
4849
4850 return 0;
4851 }
4852
4853 /*
4854 * Map the RTs (configured or automatically derived) of a VRF to the VRF.
4855 * The mapping will be used during route processing.
4856 * bgp_def: default bgp instance
4857 * bgp_vrf: specific bgp vrf instance on which RT is configured
4858 */
4859 void bgp_evpn_map_vrf_to_its_rts(struct bgp *bgp_vrf)
4860 {
4861 int i = 0;
4862 struct ecommunity_val *eval = NULL;
4863 struct listnode *node = NULL, *nnode = NULL;
4864 struct ecommunity *ecom = NULL;
4865
4866 for (ALL_LIST_ELEMENTS(bgp_vrf->vrf_import_rtl, node, nnode, ecom)) {
4867 for (i = 0; i < ecom->size; i++) {
4868 eval = (struct ecommunity_val *)(ecom->val
4869 + (i
4870 * ECOMMUNITY_SIZE));
4871 map_vrf_to_rt(bgp_vrf, eval);
4872 }
4873 }
4874 }
4875
4876 /*
4877 * Unmap the RTs (configured or automatically derived) of a VRF from the VRF.
4878 */
4879 void bgp_evpn_unmap_vrf_from_its_rts(struct bgp *bgp_vrf)
4880 {
4881 int i;
4882 struct ecommunity_val *eval;
4883 struct listnode *node, *nnode;
4884 struct ecommunity *ecom;
4885
4886 for (ALL_LIST_ELEMENTS(bgp_vrf->vrf_import_rtl, node, nnode, ecom)) {
4887 for (i = 0; i < ecom->size; i++) {
4888 struct vrf_irt_node *irt;
4889 struct ecommunity_val eval_tmp;
4890
4891 eval = (struct ecommunity_val *)(ecom->val
4892 + (i
4893 * ECOMMUNITY_SIZE));
4894 /* If using "automatic" RT, we only care about the
4895 * local-admin sub-field.
4896 * This is to facilitate using VNI as the RT for EBGP
4897 * peering too.
4898 */
4899 memcpy(&eval_tmp, eval, ECOMMUNITY_SIZE);
4900 if (!CHECK_FLAG(bgp_vrf->vrf_flags,
4901 BGP_VRF_IMPORT_RT_CFGD))
4902 mask_ecom_global_admin(&eval_tmp, eval);
4903
4904 irt = lookup_vrf_import_rt(&eval_tmp);
4905 if (irt)
4906 unmap_vrf_from_rt(bgp_vrf, irt);
4907 }
4908 }
4909 }
4910
4911
4912 /*
4913 * Map the RTs (configured or automatically derived) of a VNI to the VNI.
4914 * The mapping will be used during route processing.
4915 */
4916 void bgp_evpn_map_vni_to_its_rts(struct bgp *bgp, struct bgpevpn *vpn)
4917 {
4918 int i;
4919 struct ecommunity_val *eval;
4920 struct listnode *node, *nnode;
4921 struct ecommunity *ecom;
4922
4923 for (ALL_LIST_ELEMENTS(vpn->import_rtl, node, nnode, ecom)) {
4924 for (i = 0; i < ecom->size; i++) {
4925 eval = (struct ecommunity_val *)(ecom->val
4926 + (i
4927 * ECOMMUNITY_SIZE));
4928 map_vni_to_rt(bgp, vpn, eval);
4929 }
4930 }
4931 }
4932
4933 /*
4934 * Unmap the RTs (configured or automatically derived) of a VNI from the VNI.
4935 */
4936 void bgp_evpn_unmap_vni_from_its_rts(struct bgp *bgp, struct bgpevpn *vpn)
4937 {
4938 int i;
4939 struct ecommunity_val *eval;
4940 struct listnode *node, *nnode;
4941 struct ecommunity *ecom;
4942
4943 for (ALL_LIST_ELEMENTS(vpn->import_rtl, node, nnode, ecom)) {
4944 for (i = 0; i < ecom->size; i++) {
4945 struct irt_node *irt;
4946 struct ecommunity_val eval_tmp;
4947
4948 eval = (struct ecommunity_val *)(ecom->val
4949 + (i
4950 * ECOMMUNITY_SIZE));
4951 /* If using "automatic" RT, we only care about the
4952 * local-admin sub-field.
4953 * This is to facilitate using VNI as the RT for EBGP
4954 * peering too.
4955 */
4956 memcpy(&eval_tmp, eval, ECOMMUNITY_SIZE);
4957 if (!is_import_rt_configured(vpn))
4958 mask_ecom_global_admin(&eval_tmp, eval);
4959
4960 irt = lookup_import_rt(bgp, &eval_tmp);
4961 if (irt)
4962 unmap_vni_from_rt(bgp, vpn, irt);
4963 }
4964 }
4965 }
4966
4967 /*
4968 * Derive Import RT automatically for VNI and map VNI to RT.
4969 * The mapping will be used during route processing.
4970 */
4971 void bgp_evpn_derive_auto_rt_import(struct bgp *bgp, struct bgpevpn *vpn)
4972 {
4973 form_auto_rt(bgp, vpn->vni, vpn->import_rtl);
4974 UNSET_FLAG(vpn->flags, VNI_FLAG_IMPRT_CFGD);
4975
4976 /* Map RT to VNI */
4977 bgp_evpn_map_vni_to_its_rts(bgp, vpn);
4978 }
4979
4980 /*
4981 * Derive Export RT automatically for VNI.
4982 */
4983 void bgp_evpn_derive_auto_rt_export(struct bgp *bgp, struct bgpevpn *vpn)
4984 {
4985 form_auto_rt(bgp, vpn->vni, vpn->export_rtl);
4986 UNSET_FLAG(vpn->flags, VNI_FLAG_EXPRT_CFGD);
4987 }
4988
4989 /*
4990 * Derive RD automatically for VNI using passed information - it
4991 * is of the form RouterId:unique-id-for-vni.
4992 */
4993 void bgp_evpn_derive_auto_rd_for_vrf(struct bgp *bgp)
4994 {
4995 form_auto_rd(bgp->router_id, bgp->vrf_rd_id, &bgp->vrf_prd);
4996 }
4997
4998 /*
4999 * Derive RD automatically for VNI using passed information - it
5000 * is of the form RouterId:unique-id-for-vni.
5001 */
5002 void bgp_evpn_derive_auto_rd(struct bgp *bgp, struct bgpevpn *vpn)
5003 {
5004 char buf[100];
5005
5006 vpn->prd.family = AF_UNSPEC;
5007 vpn->prd.prefixlen = 64;
5008 sprintf(buf, "%s:%hu", inet_ntoa(bgp->router_id), vpn->rd_id);
5009 (void)str2prefix_rd(buf, &vpn->prd);
5010 UNSET_FLAG(vpn->flags, VNI_FLAG_RD_CFGD);
5011 }
5012
5013 /*
5014 * Lookup L3-VNI
5015 */
5016 bool bgp_evpn_lookup_l3vni_l2vni_table(vni_t vni)
5017 {
5018 struct list *inst = bm->bgp;
5019 struct listnode *node;
5020 struct bgp *bgp_vrf;
5021
5022 for (ALL_LIST_ELEMENTS_RO(inst, node, bgp_vrf)) {
5023 if (bgp_vrf->l3vni == vni)
5024 return true;
5025 }
5026
5027 return false;
5028 }
5029
5030 /*
5031 * Lookup VNI.
5032 */
5033 struct bgpevpn *bgp_evpn_lookup_vni(struct bgp *bgp, vni_t vni)
5034 {
5035 struct bgpevpn *vpn;
5036 struct bgpevpn tmp;
5037
5038 memset(&tmp, 0, sizeof(struct bgpevpn));
5039 tmp.vni = vni;
5040 vpn = hash_lookup(bgp->vnihash, &tmp);
5041 return vpn;
5042 }
5043
5044 /*
5045 * Create a new vpn - invoked upon configuration or zebra notification.
5046 */
5047 struct bgpevpn *bgp_evpn_new(struct bgp *bgp, vni_t vni,
5048 struct in_addr originator_ip,
5049 vrf_id_t tenant_vrf_id)
5050 {
5051 struct bgpevpn *vpn;
5052
5053 if (!bgp)
5054 return NULL;
5055
5056 vpn = XCALLOC(MTYPE_BGP_EVPN, sizeof(struct bgpevpn));
5057 if (!vpn)
5058 return NULL;
5059
5060 /* Set values - RD and RT set to defaults. */
5061 vpn->vni = vni;
5062 vpn->originator_ip = originator_ip;
5063 vpn->tenant_vrf_id = tenant_vrf_id;
5064
5065 /* Initialize route-target import and export lists */
5066 vpn->import_rtl = list_new();
5067 vpn->import_rtl->cmp = (int (*)(void *, void *))evpn_route_target_cmp;
5068 vpn->import_rtl->del = evpn_xxport_delete_ecomm;
5069 vpn->export_rtl = list_new();
5070 vpn->export_rtl->cmp = (int (*)(void *, void *))evpn_route_target_cmp;
5071 vpn->export_rtl->del = evpn_xxport_delete_ecomm;
5072 bf_assign_index(bm->rd_idspace, vpn->rd_id);
5073 derive_rd_rt_for_vni(bgp, vpn);
5074
5075 /* Initialize EVPN route table. */
5076 vpn->route_table = bgp_table_init(bgp, AFI_L2VPN, SAFI_EVPN);
5077
5078 /* Add to hash */
5079 if (!hash_get(bgp->vnihash, vpn, hash_alloc_intern)) {
5080 XFREE(MTYPE_BGP_EVPN, vpn);
5081 return NULL;
5082 }
5083
5084 /* add to l2vni list on corresponding vrf */
5085 bgpevpn_link_to_l3vni(vpn);
5086
5087 QOBJ_REG(vpn, bgpevpn);
5088 return vpn;
5089 }
5090
5091 /*
5092 * Free a given VPN - called in multiple scenarios such as zebra
5093 * notification, configuration being deleted, advertise-all-vni disabled etc.
5094 * This just frees appropriate memory, caller should have taken other
5095 * needed actions.
5096 */
5097 void bgp_evpn_free(struct bgp *bgp, struct bgpevpn *vpn)
5098 {
5099 bgpevpn_unlink_from_l3vni(vpn);
5100 bgp_table_unlock(vpn->route_table);
5101 bgp_evpn_unmap_vni_from_its_rts(bgp, vpn);
5102 list_delete(&vpn->import_rtl);
5103 list_delete(&vpn->export_rtl);
5104 bf_release_index(bm->rd_idspace, vpn->rd_id);
5105 hash_release(bgp->vnihash, vpn);
5106 QOBJ_UNREG(vpn);
5107 XFREE(MTYPE_BGP_EVPN, vpn);
5108 }
5109
5110 /*
5111 * Lookup local ES.
5112 */
5113 struct evpnes *bgp_evpn_lookup_es(struct bgp *bgp, esi_t *esi)
5114 {
5115 struct evpnes *es;
5116 struct evpnes tmp;
5117
5118 memset(&tmp, 0, sizeof(struct evpnes));
5119 memcpy(&tmp.esi, esi, sizeof(esi_t));
5120 es = hash_lookup(bgp->esihash, &tmp);
5121 return es;
5122 }
5123
5124 /*
5125 * Create a new local es - invoked upon zebra notification.
5126 */
5127 struct evpnes *bgp_evpn_es_new(struct bgp *bgp,
5128 esi_t *esi,
5129 struct ipaddr *originator_ip)
5130 {
5131 char buf[100];
5132 struct evpnes *es;
5133
5134 if (!bgp)
5135 return NULL;
5136
5137 es = XCALLOC(MTYPE_BGP_EVPN_ES, sizeof(struct evpnes));
5138 if (!es)
5139 return NULL;
5140
5141 /* set the ESI and originator_ip */
5142 memcpy(&es->esi, esi, sizeof(esi_t));
5143 memcpy(&es->originator_ip, originator_ip, sizeof(struct ipaddr));
5144
5145 /* Initialise the VTEP list */
5146 es->vtep_list = list_new();
5147 es->vtep_list->cmp = evpn_vtep_ip_cmp;
5148
5149 /* auto derive RD for this es */
5150 bf_assign_index(bm->rd_idspace, es->rd_id);
5151 es->prd.family = AF_UNSPEC;
5152 es->prd.prefixlen = 64;
5153 sprintf(buf, "%s:%hu", inet_ntoa(bgp->router_id), es->rd_id);
5154 (void)str2prefix_rd(buf, &es->prd);
5155
5156 /* Initialize the ES route table */
5157 es->route_table = bgp_table_init(bgp, AFI_L2VPN, SAFI_EVPN);
5158
5159 /* Add to hash */
5160 if (!hash_get(bgp->esihash, es, hash_alloc_intern)) {
5161 XFREE(MTYPE_BGP_EVPN_ES, es);
5162 return NULL;
5163 }
5164
5165 QOBJ_REG(es, evpnes);
5166 return es;
5167 }
5168
5169 /*
5170 * Free a given ES -
5171 * This just frees appropriate memory, caller should have taken other
5172 * needed actions.
5173 */
5174 void bgp_evpn_es_free(struct bgp *bgp, struct evpnes *es)
5175 {
5176 list_delete(&es->vtep_list);
5177 bgp_table_unlock(es->route_table);
5178 bf_release_index(bm->rd_idspace, es->rd_id);
5179 hash_release(bgp->esihash, es);
5180 QOBJ_UNREG(es);
5181 XFREE(MTYPE_BGP_EVPN_ES, es);
5182 }
5183
5184 /*
5185 * Import evpn route from global table to VNI/VRF/ESI.
5186 */
5187 int bgp_evpn_import_route(struct bgp *bgp, afi_t afi, safi_t safi,
5188 struct prefix *p, struct bgp_path_info *pi)
5189 {
5190 return install_uninstall_evpn_route(bgp, afi, safi, p, pi, 1);
5191 }
5192
5193 /*
5194 * Unimport evpn route from VNI/VRF/ESI.
5195 */
5196 int bgp_evpn_unimport_route(struct bgp *bgp, afi_t afi, safi_t safi,
5197 struct prefix *p, struct bgp_path_info *pi)
5198 {
5199 return install_uninstall_evpn_route(bgp, afi, safi, p, pi, 0);
5200 }
5201
5202 /* filter routes which have martian next hops */
5203 int bgp_filter_evpn_routes_upon_martian_nh_change(struct bgp *bgp)
5204 {
5205 afi_t afi;
5206 safi_t safi;
5207 struct bgp_node *rd_rn, *rn;
5208 struct bgp_table *table;
5209 struct bgp_path_info *pi;
5210
5211 afi = AFI_L2VPN;
5212 safi = SAFI_EVPN;
5213
5214 /* Walk entire global routing table and evaluate routes which could be
5215 * imported into this VPN. Note that we cannot just look at the routes
5216 * for the VNI's RD -
5217 * remote routes applicable for this VNI could have any RD.
5218 */
5219 /* EVPN routes are a 2-level table. */
5220 for (rd_rn = bgp_table_top(bgp->rib[afi][safi]); rd_rn;
5221 rd_rn = bgp_route_next(rd_rn)) {
5222 table = (struct bgp_table *)(rd_rn->info);
5223 if (!table)
5224 continue;
5225
5226 for (rn = bgp_table_top(table); rn; rn = bgp_route_next(rn)) {
5227
5228 for (pi = rn->info; pi; pi = pi->next) {
5229
5230 /* Consider "valid" remote routes applicable for
5231 * this VNI. */
5232 if (!(pi->type == ZEBRA_ROUTE_BGP
5233 && pi->sub_type == BGP_ROUTE_NORMAL))
5234 continue;
5235
5236 if (bgp_nexthop_self(bgp, pi->attr->nexthop)) {
5237
5238 char attr_str[BUFSIZ];
5239 char pbuf[PREFIX_STRLEN];
5240
5241 bgp_dump_attr(pi->attr, attr_str,
5242 BUFSIZ);
5243
5244 if (bgp_debug_update(pi->peer, &rn->p,
5245 NULL, 1))
5246 zlog_debug(
5247 "%u: prefix %s with attr %s - DENIED due to martian or self nexthop",
5248 bgp->vrf_id,
5249 prefix2str(
5250 &rn->p, pbuf,
5251 sizeof(pbuf)),
5252 attr_str);
5253
5254 bgp_evpn_unimport_route(bgp, afi, safi,
5255 &rn->p, pi);
5256
5257 bgp_rib_remove(rn, pi, pi->peer, afi,
5258 safi);
5259 }
5260 }
5261 }
5262 }
5263
5264 return 0;
5265 }
5266
5267 /*
5268 * Handle del of a local MACIP.
5269 */
5270 int bgp_evpn_local_macip_del(struct bgp *bgp, vni_t vni, struct ethaddr *mac,
5271 struct ipaddr *ip)
5272 {
5273 struct bgpevpn *vpn;
5274 struct prefix_evpn p;
5275
5276 /* Lookup VNI hash - should exist. */
5277 vpn = bgp_evpn_lookup_vni(bgp, vni);
5278 if (!vpn || !is_vni_live(vpn)) {
5279 flog_warn(EC_BGP_EVPN_VPN_VNI,
5280 "%u: VNI hash entry for VNI %u %s at MACIP DEL",
5281 bgp->vrf_id, vni, vpn ? "not live" : "not found");
5282 return -1;
5283 }
5284
5285 /* Remove EVPN type-2 route and schedule for processing. */
5286 build_evpn_type2_prefix(&p, mac, ip);
5287 delete_evpn_route(bgp, vpn, &p);
5288
5289 return 0;
5290 }
5291
5292 /*
5293 * Handle add of a local MACIP.
5294 */
5295 int bgp_evpn_local_macip_add(struct bgp *bgp, vni_t vni, struct ethaddr *mac,
5296 struct ipaddr *ip, uint8_t flags, uint32_t seq)
5297 {
5298 struct bgpevpn *vpn;
5299 struct prefix_evpn p;
5300
5301 /* Lookup VNI hash - should exist. */
5302 vpn = bgp_evpn_lookup_vni(bgp, vni);
5303 if (!vpn || !is_vni_live(vpn)) {
5304 flog_warn(EC_BGP_EVPN_VPN_VNI,
5305 "%u: VNI hash entry for VNI %u %s at MACIP ADD",
5306 bgp->vrf_id, vni, vpn ? "not live" : "not found");
5307 return -1;
5308 }
5309
5310 /* Create EVPN type-2 route and schedule for processing. */
5311 build_evpn_type2_prefix(&p, mac, ip);
5312 if (update_evpn_route(bgp, vpn, &p, flags, seq)) {
5313 char buf[ETHER_ADDR_STRLEN];
5314 char buf2[INET6_ADDRSTRLEN];
5315
5316 flog_err(
5317 EC_BGP_EVPN_ROUTE_CREATE,
5318 "%u:Failed to create Type-2 route, VNI %u %s MAC %s IP %s (flags: 0x%x)",
5319 bgp->vrf_id, vpn->vni,
5320 CHECK_FLAG(flags, ZEBRA_MACIP_TYPE_STICKY)
5321 ? "sticky gateway"
5322 : "",
5323 prefix_mac2str(mac, buf, sizeof(buf)),
5324 ipaddr2str(ip, buf2, sizeof(buf2)), flags);
5325 return -1;
5326 }
5327
5328 return 0;
5329 }
5330
5331 static void link_l2vni_hash_to_l3vni(struct hash_backet *backet,
5332 struct bgp *bgp_vrf)
5333 {
5334 struct bgpevpn *vpn = (struct bgpevpn *)backet->data;
5335 struct bgp *bgp_def = NULL;
5336
5337 bgp_def = bgp_get_default();
5338 assert(bgp_def);
5339
5340 if (vpn->tenant_vrf_id == bgp_vrf->vrf_id)
5341 bgpevpn_link_to_l3vni(vpn);
5342 }
5343
5344 int bgp_evpn_local_l3vni_add(vni_t l3vni, vrf_id_t vrf_id, struct ethaddr *rmac,
5345 struct in_addr originator_ip, int filter)
5346 {
5347 struct bgp *bgp_vrf = NULL; /* bgp VRF instance */
5348 struct bgp *bgp_def = NULL; /* default bgp instance */
5349 struct listnode *node = NULL;
5350 struct bgpevpn *vpn = NULL;
5351 as_t as = 0;
5352
5353 /* get the default instance - required to get the AS number for VRF
5354 * auto-creatio
5355 */
5356 bgp_def = bgp_get_default();
5357 if (!bgp_def) {
5358 flog_err(
5359 EC_BGP_NO_DFLT,
5360 "Cannot process L3VNI %u ADD - default BGP instance not yet created",
5361 l3vni);
5362 return -1;
5363 }
5364 as = bgp_def->as;
5365
5366 /* if the BGP vrf instance doesn't exist - create one */
5367 bgp_vrf = bgp_lookup_by_name(vrf_id_to_name(vrf_id));
5368 if (!bgp_vrf) {
5369
5370 int ret = 0;
5371
5372 ret = bgp_get(&bgp_vrf, &as, vrf_id_to_name(vrf_id),
5373 BGP_INSTANCE_TYPE_VRF);
5374 switch (ret) {
5375 case BGP_ERR_MULTIPLE_INSTANCE_NOT_SET:
5376 flog_err(EC_BGP_MULTI_INSTANCE,
5377 "'bgp multiple-instance' not present\n");
5378 return -1;
5379 case BGP_ERR_AS_MISMATCH:
5380 flog_err(EC_BGP_EVPN_AS_MISMATCH,
5381 "BGP is already running; AS is %u\n", as);
5382 return -1;
5383 case BGP_ERR_INSTANCE_MISMATCH:
5384 flog_err(EC_BGP_EVPN_INSTANCE_MISMATCH,
5385 "BGP instance name and AS number mismatch\n");
5386 return -1;
5387 }
5388
5389 /* mark as auto created */
5390 SET_FLAG(bgp_vrf->vrf_flags, BGP_VRF_AUTO);
5391 }
5392
5393 /* associate with l3vni */
5394 bgp_vrf->l3vni = l3vni;
5395
5396 /* set the router mac - to be used in mac-ip routes for this vrf */
5397 memcpy(&bgp_vrf->rmac, rmac, sizeof(struct ethaddr));
5398
5399 /* set the originator ip */
5400 bgp_vrf->originator_ip = originator_ip;
5401
5402 /* set the right filter - are we using l3vni only for prefix routes? */
5403 if (filter)
5404 SET_FLAG(bgp_vrf->vrf_flags, BGP_VRF_L3VNI_PREFIX_ROUTES_ONLY);
5405
5406 /* auto derive RD/RT */
5407 if (!CHECK_FLAG(bgp_vrf->vrf_flags, BGP_VRF_IMPORT_RT_CFGD))
5408 evpn_auto_rt_import_add_for_vrf(bgp_vrf);
5409 if (!CHECK_FLAG(bgp_vrf->vrf_flags, BGP_VRF_EXPORT_RT_CFGD))
5410 evpn_auto_rt_export_add_for_vrf(bgp_vrf);
5411 bgp_evpn_derive_auto_rd_for_vrf(bgp_vrf);
5412
5413 /* link all corresponding l2vnis */
5414 hash_iterate(bgp_def->vnihash,
5415 (void (*)(struct hash_backet *,
5416 void *))link_l2vni_hash_to_l3vni,
5417 bgp_vrf);
5418
5419 /* Only update all corresponding type-2 routes if we are advertising two
5420 * labels along with type-2 routes
5421 */
5422 if (!filter)
5423 for (ALL_LIST_ELEMENTS_RO(bgp_vrf->l2vnis, node, vpn))
5424 update_routes_for_vni(bgp_def, vpn);
5425
5426 /* advertise type-5 routes if needed */
5427 update_advertise_vrf_routes(bgp_vrf);
5428
5429 /* install all remote routes belonging to this l3vni into correspondng
5430 * vrf */
5431 install_routes_for_vrf(bgp_vrf);
5432
5433 return 0;
5434 }
5435
5436 int bgp_evpn_local_l3vni_del(vni_t l3vni, vrf_id_t vrf_id)
5437 {
5438 struct bgp *bgp_vrf = NULL; /* bgp vrf instance */
5439 struct bgp *bgp_def = NULL; /* default bgp instance */
5440 struct listnode *node = NULL;
5441 struct listnode *next = NULL;
5442 struct bgpevpn *vpn = NULL;
5443
5444 bgp_vrf = bgp_lookup_by_vrf_id(vrf_id);
5445 if (!bgp_vrf) {
5446 flog_err(
5447 EC_BGP_NO_DFLT,
5448 "Cannot process L3VNI %u Del - Could not find BGP instance",
5449 l3vni);
5450 return -1;
5451 }
5452
5453 bgp_def = bgp_get_default();
5454 if (!bgp_def) {
5455 flog_err(
5456 EC_BGP_NO_DFLT,
5457 "Cannot process L3VNI %u Del - Could not find default BGP instance",
5458 l3vni);
5459 return -1;
5460 }
5461
5462 /* Remove remote routes from BGT VRF even if BGP_VRF_AUTO is configured,
5463 * bgp_delete would not remove/decrement bgp_path_info of the ip_prefix
5464 * routes. This will uninstalling the routes from zebra and decremnt the
5465 * bgp info count.
5466 */
5467 uninstall_routes_for_vrf(bgp_vrf);
5468
5469 /* delete/withdraw all type-5 routes */
5470 delete_withdraw_vrf_routes(bgp_vrf);
5471
5472 /* remove the l3vni from vrf instance */
5473 bgp_vrf->l3vni = 0;
5474
5475 /* remove the Rmac from the BGP vrf */
5476 memset(&bgp_vrf->rmac, 0, sizeof(struct ethaddr));
5477
5478 /* delete RD/RT */
5479 if (!list_isempty(bgp_vrf->vrf_import_rtl)) {
5480 bgp_evpn_unmap_vrf_from_its_rts(bgp_vrf);
5481 list_delete_all_node(bgp_vrf->vrf_import_rtl);
5482 }
5483 if (!list_isempty(bgp_vrf->vrf_export_rtl)) {
5484 list_delete_all_node(bgp_vrf->vrf_export_rtl);
5485 }
5486
5487 /* update all corresponding local mac-ip routes */
5488 if (!CHECK_FLAG(bgp_vrf->vrf_flags, BGP_VRF_L3VNI_PREFIX_ROUTES_ONLY)) {
5489 for (ALL_LIST_ELEMENTS_RO(bgp_vrf->l2vnis, node, vpn)) {
5490 UNSET_FLAG(vpn->flags, VNI_FLAG_USE_TWO_LABELS);
5491 update_routes_for_vni(bgp_def, vpn);
5492 }
5493 }
5494
5495 /* If any L2VNIs point to this instance, unlink them. */
5496 for (ALL_LIST_ELEMENTS(bgp_vrf->l2vnis, node, next, vpn))
5497 bgpevpn_unlink_from_l3vni(vpn);
5498
5499 /* Delete the instance if it was autocreated */
5500 if (CHECK_FLAG(bgp_vrf->vrf_flags, BGP_VRF_AUTO))
5501 bgp_delete(bgp_vrf);
5502
5503 return 0;
5504 }
5505
5506 /*
5507 * Handle del of a local VNI.
5508 */
5509 int bgp_evpn_local_vni_del(struct bgp *bgp, vni_t vni)
5510 {
5511 struct bgpevpn *vpn;
5512
5513 /* Locate VNI hash */
5514 vpn = bgp_evpn_lookup_vni(bgp, vni);
5515 if (!vpn) {
5516 if (bgp_debug_zebra(NULL))
5517 flog_warn(
5518 EC_BGP_EVPN_VPN_VNI,
5519 "%u: VNI hash entry for VNI %u not found at DEL",
5520 bgp->vrf_id, vni);
5521 return 0;
5522 }
5523
5524 /* Remove all local EVPN routes and schedule for processing (to
5525 * withdraw from peers).
5526 */
5527 delete_routes_for_vni(bgp, vpn);
5528
5529 /*
5530 * tunnel is no longer active, del tunnel ip address from tip_hash
5531 */
5532 bgp_tip_del(bgp, &vpn->originator_ip);
5533
5534 /* Clear "live" flag and see if hash needs to be freed. */
5535 UNSET_FLAG(vpn->flags, VNI_FLAG_LIVE);
5536 if (!is_vni_configured(vpn))
5537 bgp_evpn_free(bgp, vpn);
5538
5539 return 0;
5540 }
5541
5542 /*
5543 * Handle add (or update) of a local VNI. The VNI changes we care
5544 * about are for the local-tunnel-ip and the (tenant) VRF.
5545 */
5546 int bgp_evpn_local_vni_add(struct bgp *bgp, vni_t vni,
5547 struct in_addr originator_ip, vrf_id_t tenant_vrf_id)
5548 {
5549 struct bgpevpn *vpn;
5550 struct prefix_evpn p;
5551
5552 /* Lookup VNI. If present and no change, exit. */
5553 vpn = bgp_evpn_lookup_vni(bgp, vni);
5554 if (vpn) {
5555
5556 if (is_vni_live(vpn)
5557 && IPV4_ADDR_SAME(&vpn->originator_ip, &originator_ip)
5558 && vpn->tenant_vrf_id == tenant_vrf_id)
5559 /* Probably some other param has changed that we don't
5560 * care about. */
5561 return 0;
5562
5563 /* Update tenant_vrf_id if it has changed. */
5564 if (vpn->tenant_vrf_id != tenant_vrf_id) {
5565 bgpevpn_unlink_from_l3vni(vpn);
5566 vpn->tenant_vrf_id = tenant_vrf_id;
5567 bgpevpn_link_to_l3vni(vpn);
5568 }
5569
5570 /* If tunnel endpoint IP has changed, update (and delete prior
5571 * type-3 route, if needed.)
5572 */
5573 if (!IPV4_ADDR_SAME(&vpn->originator_ip, &originator_ip))
5574 handle_tunnel_ip_change(bgp, vpn, originator_ip);
5575
5576 /* Update all routes with new endpoint IP and/or export RT
5577 * for VRFs
5578 */
5579 if (is_vni_live(vpn))
5580 update_routes_for_vni(bgp, vpn);
5581 }
5582
5583 /* Create or update as appropriate. */
5584 if (!vpn) {
5585 vpn = bgp_evpn_new(bgp, vni, originator_ip, tenant_vrf_id);
5586 if (!vpn) {
5587 flog_err(
5588 EC_BGP_VNI,
5589 "%u: Failed to allocate VNI entry for VNI %u - at Add",
5590 bgp->vrf_id, vni);
5591 return -1;
5592 }
5593 }
5594
5595 /* if the VNI is live already, there is nothing more to do */
5596 if (is_vni_live(vpn))
5597 return 0;
5598
5599 /* Mark as "live" */
5600 SET_FLAG(vpn->flags, VNI_FLAG_LIVE);
5601
5602 /* tunnel is now active, add tunnel-ip to db */
5603 bgp_tip_add(bgp, &originator_ip);
5604
5605 /* filter routes as nexthop database has changed */
5606 bgp_filter_evpn_routes_upon_martian_nh_change(bgp);
5607
5608 /*
5609 * Create EVPN type-3 route and schedule for processing.
5610 *
5611 * RT-3 only if doing head-end replication
5612 */
5613 if (bgp->vxlan_flood_ctrl == VXLAN_FLOOD_HEAD_END_REPL) {
5614 build_evpn_type3_prefix(&p, vpn->originator_ip);
5615 if (update_evpn_route(bgp, vpn, &p, 0, 0)) {
5616 flog_err(EC_BGP_EVPN_ROUTE_CREATE,
5617 "%u: Type3 route creation failure for VNI %u",
5618 bgp->vrf_id, vni);
5619 return -1;
5620 }
5621 }
5622
5623 /* If we have learnt and retained remote routes (VTEPs, MACs) for this
5624 * VNI,
5625 * install them.
5626 */
5627 install_routes_for_vni(bgp, vpn);
5628
5629 /* If we are advertising gateway mac-ip
5630 It needs to be conveyed again to zebra */
5631 bgp_zebra_advertise_gw_macip(bgp, vpn->advertise_gw_macip, vpn->vni);
5632
5633 return 0;
5634 }
5635
5636 /*
5637 * bgp_evpn_local_es_del
5638 */
5639 int bgp_evpn_local_es_del(struct bgp *bgp,
5640 esi_t *esi,
5641 struct ipaddr *originator_ip)
5642 {
5643 char buf[ESI_STR_LEN];
5644 struct evpnes *es = NULL;
5645
5646 if (!bgp->esihash) {
5647 flog_err(EC_BGP_ES_CREATE, "%u: ESI hash not yet created",
5648 bgp->vrf_id);
5649 return -1;
5650 }
5651
5652 /* Lookup ESI hash - should exist. */
5653 es = bgp_evpn_lookup_es(bgp, esi);
5654 if (!es) {
5655 flog_warn(EC_BGP_EVPN_ESI,
5656 "%u: ESI hash entry for ESI %s at Local ES DEL",
5657 bgp->vrf_id, esi_to_str(esi, buf, sizeof(buf)));
5658 return -1;
5659 }
5660
5661 /* Delete all local EVPN ES routes from ESI table
5662 * and schedule for processing (to withdraw from peers))
5663 */
5664 delete_routes_for_es(bgp, es);
5665
5666 /* free the hash entry */
5667 bgp_evpn_es_free(bgp, es);
5668
5669 return 0;
5670 }
5671
5672 /*
5673 * bgp_evpn_local_es_add
5674 */
5675 int bgp_evpn_local_es_add(struct bgp *bgp,
5676 esi_t *esi,
5677 struct ipaddr *originator_ip)
5678 {
5679 char buf[ESI_STR_LEN];
5680 struct evpnes *es = NULL;
5681 struct prefix_evpn p;
5682
5683 if (!bgp->esihash) {
5684 flog_err(EC_BGP_ES_CREATE, "%u: ESI hash not yet created",
5685 bgp->vrf_id);
5686 return -1;
5687 }
5688
5689 /* create the new es */
5690 es = bgp_evpn_lookup_es(bgp, esi);
5691 if (!es) {
5692 es = bgp_evpn_es_new(bgp, esi, originator_ip);
5693 if (!es) {
5694 flog_err(
5695 EC_BGP_ES_CREATE,
5696 "%u: Failed to allocate ES entry for ESI %s - at Local ES Add",
5697 bgp->vrf_id, esi_to_str(esi, buf, sizeof(buf)));
5698 return -1;
5699 }
5700 }
5701 UNSET_FLAG(es->flags, EVPNES_REMOTE);
5702 SET_FLAG(es->flags, EVPNES_LOCAL);
5703
5704 build_evpn_type4_prefix(&p, esi, originator_ip->ipaddr_v4);
5705 if (update_evpn_type4_route(bgp, es, &p)) {
5706 flog_err(EC_BGP_EVPN_ROUTE_CREATE,
5707 "%u: Type4 route creation failure for ESI %s",
5708 bgp->vrf_id, esi_to_str(esi, buf, sizeof(buf)));
5709 return -1;
5710 }
5711
5712 /* import all remote ES routes in th ES table */
5713 install_routes_for_es(bgp, es);
5714
5715 return 0;
5716 }
5717
5718 /*
5719 * Handle change in setting for BUM handling. The supported values
5720 * are head-end replication and dropping all BUM packets. Any change
5721 * should be registered with zebra. Also, if doing head-end replication,
5722 * need to advertise local VNIs as EVPN RT-3 wheras, if BUM packets are
5723 * to be dropped, the RT-3s must be withdrawn.
5724 */
5725 void bgp_evpn_flood_control_change(struct bgp *bgp)
5726 {
5727 zlog_info("L2VPN EVPN BUM handling is %s",
5728 bgp->vxlan_flood_ctrl == VXLAN_FLOOD_HEAD_END_REPL ?
5729 "Flooding" : "Flooding Disabled");
5730
5731 bgp_zebra_vxlan_flood_control(bgp, bgp->vxlan_flood_ctrl);
5732 if (bgp->vxlan_flood_ctrl == VXLAN_FLOOD_HEAD_END_REPL)
5733 hash_iterate(bgp->vnihash, create_advertise_type3, bgp);
5734 else if (bgp->vxlan_flood_ctrl == VXLAN_FLOOD_DISABLED)
5735 hash_iterate(bgp->vnihash, delete_withdraw_type3, bgp);
5736 }
5737
5738 /*
5739 * Cleanup EVPN information on disable - Need to delete and withdraw
5740 * EVPN routes from peers.
5741 */
5742 void bgp_evpn_cleanup_on_disable(struct bgp *bgp)
5743 {
5744 hash_iterate(bgp->vnihash, (void (*)(struct hash_backet *,
5745 void *))cleanup_vni_on_disable,
5746 bgp);
5747 }
5748
5749 /*
5750 * Cleanup EVPN information - invoked at the time of bgpd exit or when the
5751 * BGP instance (default) is being freed.
5752 */
5753 void bgp_evpn_cleanup(struct bgp *bgp)
5754 {
5755 hash_iterate(bgp->vnihash,
5756 (void (*)(struct hash_backet *, void *))free_vni_entry,
5757 bgp);
5758
5759 hash_free(bgp->import_rt_hash);
5760 bgp->import_rt_hash = NULL;
5761
5762 hash_free(bgp->vrf_import_rt_hash);
5763 bgp->vrf_import_rt_hash = NULL;
5764
5765 hash_free(bgp->vnihash);
5766 bgp->vnihash = NULL;
5767 if (bgp->esihash)
5768 hash_free(bgp->esihash);
5769 bgp->esihash = NULL;
5770
5771 list_delete(&bgp->vrf_import_rtl);
5772 list_delete(&bgp->vrf_export_rtl);
5773 list_delete(&bgp->l2vnis);
5774 }
5775
5776 /*
5777 * Initialization for EVPN
5778 * Create
5779 * VNI hash table
5780 * hash for RT to VNI
5781 */
5782 void bgp_evpn_init(struct bgp *bgp)
5783 {
5784 bgp->vnihash =
5785 hash_create(vni_hash_key_make, vni_hash_cmp, "BGP VNI Hash");
5786 bgp->esihash =
5787 hash_create(esi_hash_keymake, esi_cmp,
5788 "BGP EVPN Local ESI Hash");
5789 bgp->import_rt_hash =
5790 hash_create(import_rt_hash_key_make, import_rt_hash_cmp,
5791 "BGP Import RT Hash");
5792 bgp->vrf_import_rt_hash =
5793 hash_create(vrf_import_rt_hash_key_make, vrf_import_rt_hash_cmp,
5794 "BGP VRF Import RT Hash");
5795 bgp->vrf_import_rtl = list_new();
5796 bgp->vrf_import_rtl->cmp =
5797 (int (*)(void *, void *))evpn_route_target_cmp;
5798 bgp->vrf_import_rtl->del = evpn_xxport_delete_ecomm;
5799 bgp->vrf_export_rtl = list_new();
5800 bgp->vrf_export_rtl->cmp =
5801 (int (*)(void *, void *))evpn_route_target_cmp;
5802 bgp->vrf_export_rtl->del = evpn_xxport_delete_ecomm;
5803 bgp->l2vnis = list_new();
5804 bgp->l2vnis->cmp = vni_list_cmp;
5805
5806 /* Default BUM handling is to do head-end replication. */
5807 bgp->vxlan_flood_ctrl = VXLAN_FLOOD_HEAD_END_REPL;
5808 }
5809
5810 void bgp_evpn_vrf_delete(struct bgp *bgp_vrf)
5811 {
5812 bgp_evpn_unmap_vrf_from_its_rts(bgp_vrf);
5813 }