]> git.proxmox.com Git - mirror_frr.git/blob - bgpd/bgp_evpn.c
Merge pull request #5473 from yasuhiro-ohara-ntt/ospf6d-self-orig-maxage-fix
[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 #include "bgpd/bgp_mac.h"
52 #include "bgpd/bgp_vty.h"
53
54 /*
55 * Definitions and external declarations.
56 */
57 extern struct zclient *zclient;
58
59 DEFINE_QOBJ_TYPE(bgpevpn)
60 DEFINE_QOBJ_TYPE(evpnes)
61
62
63 /*
64 * Static function declarations
65 */
66 static void delete_evpn_route_entry(struct bgp *bgp, afi_t afi, safi_t safi,
67 struct bgp_node *rn,
68 struct bgp_path_info **pi);
69 static int delete_all_vni_routes(struct bgp *bgp, struct bgpevpn *vpn);
70
71 /*
72 * Private functions.
73 */
74
75 /* compare two IPV4 VTEP IPs */
76 static int evpn_vtep_ip_cmp(void *p1, void *p2)
77 {
78 const struct in_addr *ip1 = p1;
79 const struct in_addr *ip2 = p2;
80
81 return ip1->s_addr - ip2->s_addr;
82 }
83
84 /*
85 * Make hash key for ESI.
86 */
87 static unsigned int esi_hash_keymake(const void *p)
88 {
89 const struct evpnes *pes = p;
90 const void *pnt = (void *)pes->esi.val;
91
92 return jhash(pnt, ESI_BYTES, 0xa5a5a55a);
93 }
94
95 /*
96 * Compare two ESIs.
97 */
98 static bool esi_cmp(const void *p1, const void *p2)
99 {
100 const struct evpnes *pes1 = p1;
101 const struct evpnes *pes2 = p2;
102
103 if (pes1 == NULL && pes2 == NULL)
104 return true;
105
106 if (pes1 == NULL || pes2 == NULL)
107 return false;
108
109 return (memcmp(pes1->esi.val, pes2->esi.val, ESI_BYTES) == 0);
110 }
111
112 /*
113 * Make vni hash key.
114 */
115 static unsigned int vni_hash_key_make(const void *p)
116 {
117 const struct bgpevpn *vpn = p;
118 return (jhash_1word(vpn->vni, 0));
119 }
120
121 /*
122 * Comparison function for vni hash
123 */
124 static bool vni_hash_cmp(const void *p1, const void *p2)
125 {
126 const struct bgpevpn *vpn1 = p1;
127 const struct bgpevpn *vpn2 = p2;
128
129 if (!vpn1 && !vpn2)
130 return true;
131 if (!vpn1 || !vpn2)
132 return false;
133 return (vpn1->vni == vpn2->vni);
134 }
135
136 static int vni_list_cmp(void *p1, void *p2)
137 {
138 const struct bgpevpn *vpn1 = p1;
139 const struct bgpevpn *vpn2 = p2;
140
141 return vpn1->vni - vpn2->vni;
142 }
143
144 /*
145 * Make vrf import route target hash key.
146 */
147 static unsigned int vrf_import_rt_hash_key_make(const void *p)
148 {
149 const struct vrf_irt_node *irt = p;
150 const char *pnt = irt->rt.val;
151
152 return jhash(pnt, 8, 0x5abc1234);
153 }
154
155 /*
156 * Comparison function for vrf import rt hash
157 */
158 static bool vrf_import_rt_hash_cmp(const void *p1, const void *p2)
159 {
160 const struct vrf_irt_node *irt1 = p1;
161 const struct vrf_irt_node *irt2 = p2;
162
163 if (irt1 == NULL && irt2 == NULL)
164 return true;
165
166 if (irt1 == NULL || irt2 == NULL)
167 return false;
168
169 return (memcmp(irt1->rt.val, irt2->rt.val, ECOMMUNITY_SIZE) == 0);
170 }
171
172 /*
173 * Create a new vrf import_rt in evpn instance
174 */
175 static struct vrf_irt_node *vrf_import_rt_new(struct ecommunity_val *rt)
176 {
177 struct bgp *bgp_evpn = NULL;
178 struct vrf_irt_node *irt;
179
180 bgp_evpn = bgp_get_evpn();
181 if (!bgp_evpn) {
182 flog_err(EC_BGP_NO_DFLT,
183 "vrf import rt new - evpn instance not created yet");
184 return NULL;
185 }
186
187 irt = XCALLOC(MTYPE_BGP_EVPN_VRF_IMPORT_RT,
188 sizeof(struct vrf_irt_node));
189
190 irt->rt = *rt;
191 irt->vrfs = list_new();
192
193 /* Add to hash */
194 if (!hash_get(bgp_evpn->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_evpn = NULL;
208
209 bgp_evpn = bgp_get_evpn();
210 if (!bgp_evpn) {
211 flog_err(EC_BGP_NO_DFLT,
212 "vrf import rt free - evpn instance not created yet");
213 return;
214 }
215
216 hash_release(bgp_evpn->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_evpn = NULL;
228 struct vrf_irt_node *irt;
229 struct vrf_irt_node tmp;
230
231 bgp_evpn = bgp_get_evpn();
232 if (!bgp_evpn) {
233 flog_err(
234 EC_BGP_NO_DFLT,
235 "vrf import rt lookup - evpn instance not created yet");
236 return NULL;
237 }
238
239 memset(&tmp, 0, sizeof(struct vrf_irt_node));
240 memcpy(&tmp.rt, rt, ECOMMUNITY_SIZE);
241 irt = hash_lookup(bgp_evpn->vrf_import_rt_hash, &tmp);
242 return irt;
243 }
244
245 /*
246 * Is specified VRF present on the RT's list of "importing" VRFs?
247 */
248 static int is_vrf_present_in_irt_vrfs(struct list *vrfs, struct bgp *bgp_vrf)
249 {
250 struct listnode *node = NULL, *nnode = NULL;
251 struct bgp *tmp_bgp_vrf = NULL;
252
253 for (ALL_LIST_ELEMENTS(vrfs, node, nnode, tmp_bgp_vrf)) {
254 if (tmp_bgp_vrf == bgp_vrf)
255 return 1;
256 }
257 return 0;
258 }
259
260 /*
261 * Make import route target hash key.
262 */
263 static unsigned int import_rt_hash_key_make(const void *p)
264 {
265 const struct irt_node *irt = p;
266 const char *pnt = irt->rt.val;
267
268 return jhash(pnt, 8, 0xdeadbeef);
269 }
270
271 /*
272 * Comparison function for import rt hash
273 */
274 static bool import_rt_hash_cmp(const void *p1, const void *p2)
275 {
276 const struct irt_node *irt1 = p1;
277 const struct irt_node *irt2 = p2;
278
279 if (irt1 == NULL && irt2 == NULL)
280 return true;
281
282 if (irt1 == NULL || irt2 == NULL)
283 return false;
284
285 return (memcmp(irt1->rt.val, irt2->rt.val, ECOMMUNITY_SIZE) == 0);
286 }
287
288 /*
289 * Create a new import_rt
290 */
291 static struct irt_node *import_rt_new(struct bgp *bgp,
292 struct ecommunity_val *rt)
293 {
294 struct irt_node *irt;
295
296 if (!bgp)
297 return NULL;
298
299 irt = XCALLOC(MTYPE_BGP_EVPN_IMPORT_RT, sizeof(struct irt_node));
300
301 irt->rt = *rt;
302 irt->vnis = list_new();
303
304 /* Add to hash */
305 if (!hash_get(bgp->import_rt_hash, irt, hash_alloc_intern)) {
306 XFREE(MTYPE_BGP_EVPN_IMPORT_RT, irt);
307 return NULL;
308 }
309
310 return irt;
311 }
312
313 /*
314 * Free the import rt node
315 */
316 static void import_rt_free(struct bgp *bgp, struct irt_node *irt)
317 {
318 hash_release(bgp->import_rt_hash, irt);
319 list_delete(&irt->vnis);
320 XFREE(MTYPE_BGP_EVPN_IMPORT_RT, irt);
321 }
322
323 /*
324 * Function to lookup Import RT node - used to map a RT to set of
325 * VNIs importing routes with that RT.
326 */
327 static struct irt_node *lookup_import_rt(struct bgp *bgp,
328 struct ecommunity_val *rt)
329 {
330 struct irt_node *irt;
331 struct irt_node tmp;
332
333 memset(&tmp, 0, sizeof(struct irt_node));
334 memcpy(&tmp.rt, rt, ECOMMUNITY_SIZE);
335 irt = hash_lookup(bgp->import_rt_hash, &tmp);
336 return irt;
337 }
338
339 /*
340 * Is specified VNI present on the RT's list of "importing" VNIs?
341 */
342 static int is_vni_present_in_irt_vnis(struct list *vnis, struct bgpevpn *vpn)
343 {
344 struct listnode *node, *nnode;
345 struct bgpevpn *tmp_vpn;
346
347 for (ALL_LIST_ELEMENTS(vnis, node, nnode, tmp_vpn)) {
348 if (tmp_vpn == vpn)
349 return 1;
350 }
351
352 return 0;
353 }
354
355 /*
356 * Compare Route Targets.
357 */
358 static int evpn_route_target_cmp(struct ecommunity *ecom1,
359 struct ecommunity *ecom2)
360 {
361 if (ecom1 && !ecom2)
362 return -1;
363
364 if (!ecom1 && ecom2)
365 return 1;
366
367 if (!ecom1 && !ecom2)
368 return 0;
369
370 if (ecom1->str && !ecom2->str)
371 return -1;
372
373 if (!ecom1->str && ecom2->str)
374 return 1;
375
376 if (!ecom1->str && !ecom2->str)
377 return 0;
378
379 return strcmp(ecom1->str, ecom2->str);
380 }
381
382 static void evpn_xxport_delete_ecomm(void *val)
383 {
384 struct ecommunity *ecomm = val;
385 ecommunity_free(&ecomm);
386 }
387
388 /*
389 * Mask off global-admin field of specified extended community (RT),
390 * just retain the local-admin field.
391 */
392 static inline void mask_ecom_global_admin(struct ecommunity_val *dst,
393 struct ecommunity_val *src)
394 {
395 uint8_t type;
396
397 type = src->val[0];
398 dst->val[0] = 0;
399 if (type == ECOMMUNITY_ENCODE_AS) {
400 dst->val[2] = dst->val[3] = 0;
401 } else if (type == ECOMMUNITY_ENCODE_AS4
402 || type == ECOMMUNITY_ENCODE_IP) {
403 dst->val[2] = dst->val[3] = 0;
404 dst->val[4] = dst->val[5] = 0;
405 }
406 }
407
408 /*
409 * Map one RT to specified VRF.
410 * bgp_vrf = BGP vrf instance
411 */
412 static void map_vrf_to_rt(struct bgp *bgp_vrf, struct ecommunity_val *eval)
413 {
414 struct vrf_irt_node *irt = NULL;
415 struct ecommunity_val eval_tmp;
416
417 /* If using "automatic" RT,
418 * we only care about the local-admin sub-field.
419 * This is to facilitate using L3VNI(VRF-VNI)
420 * as the RT for EBGP peering too.
421 */
422 memcpy(&eval_tmp, eval, ECOMMUNITY_SIZE);
423 if (!CHECK_FLAG(bgp_vrf->vrf_flags, BGP_VRF_IMPORT_RT_CFGD))
424 mask_ecom_global_admin(&eval_tmp, eval);
425
426 irt = lookup_vrf_import_rt(&eval_tmp);
427 if (irt && is_vrf_present_in_irt_vrfs(irt->vrfs, bgp_vrf))
428 /* Already mapped. */
429 return;
430
431 if (!irt)
432 irt = vrf_import_rt_new(&eval_tmp);
433
434 /* Add VRF to the list for this RT. */
435 listnode_add(irt->vrfs, bgp_vrf);
436 }
437
438 /*
439 * Unmap specified VRF from specified RT. If there are no other
440 * VRFs for this RT, then the RT hash is deleted.
441 * bgp_vrf: BGP VRF specific instance
442 */
443 static void unmap_vrf_from_rt(struct bgp *bgp_vrf, struct vrf_irt_node *irt)
444 {
445 /* Delete VRF from list for this RT. */
446 listnode_delete(irt->vrfs, bgp_vrf);
447 if (!listnode_head(irt->vrfs)) {
448 vrf_import_rt_free(irt);
449 }
450 }
451
452 /*
453 * Map one RT to specified VNI.
454 */
455 static void map_vni_to_rt(struct bgp *bgp, struct bgpevpn *vpn,
456 struct ecommunity_val *eval)
457 {
458 struct irt_node *irt;
459 struct ecommunity_val eval_tmp;
460
461 /* If using "automatic" RT, we only care about the local-admin
462 * sub-field.
463 * This is to facilitate using VNI as the RT for EBGP peering too.
464 */
465 memcpy(&eval_tmp, eval, ECOMMUNITY_SIZE);
466 if (!is_import_rt_configured(vpn))
467 mask_ecom_global_admin(&eval_tmp, eval);
468
469 irt = lookup_import_rt(bgp, &eval_tmp);
470 if (irt)
471 if (is_vni_present_in_irt_vnis(irt->vnis, vpn))
472 /* Already mapped. */
473 return;
474
475 if (!irt) {
476 irt = import_rt_new(bgp, &eval_tmp);
477 assert(irt);
478 }
479
480 /* Add VNI to the hash list for this RT. */
481 listnode_add(irt->vnis, vpn);
482 }
483
484 /*
485 * Unmap specified VNI from specified RT. If there are no other
486 * VNIs for this RT, then the RT hash is deleted.
487 */
488 static void unmap_vni_from_rt(struct bgp *bgp, struct bgpevpn *vpn,
489 struct irt_node *irt)
490 {
491 /* Delete VNI from hash list for this RT. */
492 listnode_delete(irt->vnis, vpn);
493 if (!listnode_head(irt->vnis)) {
494 import_rt_free(bgp, irt);
495 }
496 }
497
498 static void bgp_evpn_get_rmac_nexthop(struct bgpevpn *vpn,
499 struct prefix_evpn *p,
500 struct attr *attr, uint8_t flags)
501 {
502 struct bgp *bgp_vrf = vpn->bgp_vrf;
503
504 memset(&attr->rmac, 0, sizeof(struct ethaddr));
505 if (!bgp_vrf)
506 return;
507
508 if (p->prefix.route_type != BGP_EVPN_MAC_IP_ROUTE)
509 return;
510
511 /* Copy sys (pip) RMAC and PIP IP as nexthop
512 * in case of route is self MAC-IP,
513 * advertise-pip and advertise-svi-ip features
514 * are enabled.
515 * Otherwise, for all host MAC-IP route's
516 * copy anycast RMAC.
517 */
518 if (CHECK_FLAG(flags, BGP_EVPN_MACIP_TYPE_SVI_IP)
519 && bgp_vrf->evpn_info->advertise_pip &&
520 bgp_vrf->evpn_info->is_anycast_mac) {
521 /* copy sys rmac */
522 memcpy(&attr->rmac, &bgp_vrf->evpn_info->pip_rmac,
523 ETH_ALEN);
524 attr->nexthop = bgp_vrf->evpn_info->pip_ip;
525 attr->mp_nexthop_global_in =
526 bgp_vrf->evpn_info->pip_ip;
527 } else
528 memcpy(&attr->rmac, &bgp_vrf->rmac, ETH_ALEN);
529 }
530
531 /*
532 * Create RT extended community automatically from passed information:
533 * of the form AS:VNI.
534 * NOTE: We use only the lower 16 bits of the AS. This is sufficient as
535 * the need is to get a RT value that will be unique across different
536 * VNIs but the same across routers (in the same AS) for a particular
537 * VNI.
538 */
539 static void form_auto_rt(struct bgp *bgp, vni_t vni, struct list *rtl)
540 {
541 struct ecommunity_val eval;
542 struct ecommunity *ecomadd, *ecom;
543 bool ecom_found = false;
544 struct listnode *node;
545
546 if (bgp->advertise_autort_rfc8365)
547 vni |= EVPN_AUTORT_VXLAN;
548 encode_route_target_as((bgp->as & 0xFFFF), vni, &eval);
549
550 ecomadd = ecommunity_new();
551 ecommunity_add_val(ecomadd, &eval);
552 for (ALL_LIST_ELEMENTS_RO(rtl, node, ecom))
553 if (ecommunity_cmp(ecomadd, ecom))
554 ecom_found = true;
555
556 if (!ecom_found)
557 listnode_add_sort(rtl, ecomadd);
558 }
559
560 /*
561 * Derive RD and RT for a VNI automatically. Invoked at the time of
562 * creation of a VNI.
563 */
564 static void derive_rd_rt_for_vni(struct bgp *bgp, struct bgpevpn *vpn)
565 {
566 bgp_evpn_derive_auto_rd(bgp, vpn);
567 bgp_evpn_derive_auto_rt_import(bgp, vpn);
568 bgp_evpn_derive_auto_rt_export(bgp, vpn);
569 }
570
571 /*
572 * Convert nexthop (remote VTEP IP) into an IPv6 address.
573 */
574 static void evpn_convert_nexthop_to_ipv6(struct attr *attr)
575 {
576 if (BGP_ATTR_NEXTHOP_AFI_IP6(attr))
577 return;
578 ipv4_to_ipv4_mapped_ipv6(&attr->mp_nexthop_global, attr->nexthop);
579 attr->mp_nexthop_len = IPV6_MAX_BYTELEN;
580 }
581
582 /*
583 * Add (update) or delete MACIP from zebra.
584 */
585 static int bgp_zebra_send_remote_macip(struct bgp *bgp, struct bgpevpn *vpn,
586 struct prefix_evpn *p,
587 struct in_addr remote_vtep_ip, int add,
588 uint8_t flags, uint32_t seq)
589 {
590 struct stream *s;
591 int ipa_len;
592 char buf1[ETHER_ADDR_STRLEN];
593 char buf2[INET6_ADDRSTRLEN];
594 char buf3[INET6_ADDRSTRLEN];
595
596 /* Check socket. */
597 if (!zclient || zclient->sock < 0)
598 return 0;
599
600 /* Don't try to register if Zebra doesn't know of this instance. */
601 if (!IS_BGP_INST_KNOWN_TO_ZEBRA(bgp)) {
602 if (BGP_DEBUG(zebra, ZEBRA))
603 zlog_debug("%s: No zebra instance to talk to, not installing remote macip",
604 __PRETTY_FUNCTION__);
605 return 0;
606 }
607 s = zclient->obuf;
608 stream_reset(s);
609
610 zclient_create_header(
611 s, add ? ZEBRA_REMOTE_MACIP_ADD : ZEBRA_REMOTE_MACIP_DEL,
612 bgp->vrf_id);
613 stream_putl(s, vpn->vni);
614 stream_put(s, &p->prefix.macip_addr.mac.octet, ETH_ALEN); /* Mac Addr */
615 /* IP address length and IP address, if any. */
616 if (is_evpn_prefix_ipaddr_none(p))
617 stream_putl(s, 0);
618 else {
619 ipa_len = is_evpn_prefix_ipaddr_v4(p) ? IPV4_MAX_BYTELEN
620 : IPV6_MAX_BYTELEN;
621 stream_putl(s, ipa_len);
622 stream_put(s, &p->prefix.macip_addr.ip.ip.addr, ipa_len);
623 }
624 stream_put_in_addr(s, &remote_vtep_ip);
625
626 /* TX flags - MAC sticky status and/or gateway mac */
627 /* Also TX the sequence number of the best route. */
628 if (add) {
629 stream_putc(s, flags);
630 stream_putl(s, seq);
631 }
632
633 stream_putw_at(s, 0, stream_get_endp(s));
634
635 if (bgp_debug_zebra(NULL))
636 zlog_debug(
637 "Tx %s MACIP, VNI %u MAC %s IP %s flags 0x%x seq %u remote VTEP %s",
638 add ? "ADD" : "DEL", vpn->vni,
639 prefix_mac2str(&p->prefix.macip_addr.mac,
640 buf1, sizeof(buf1)),
641 ipaddr2str(&p->prefix.macip_addr.ip,
642 buf3, sizeof(buf3)), flags, seq,
643 inet_ntop(AF_INET, &remote_vtep_ip, buf2,
644 sizeof(buf2)));
645
646 return zclient_send_message(zclient);
647 }
648
649 /*
650 * Add (update) or delete remote VTEP from zebra.
651 */
652 static int bgp_zebra_send_remote_vtep(struct bgp *bgp, struct bgpevpn *vpn,
653 struct prefix_evpn *p,
654 int flood_control, int add)
655 {
656 struct stream *s;
657
658 /* Check socket. */
659 if (!zclient || zclient->sock < 0)
660 return 0;
661
662 /* Don't try to register if Zebra doesn't know of this instance. */
663 if (!IS_BGP_INST_KNOWN_TO_ZEBRA(bgp)) {
664 if (BGP_DEBUG(zebra, ZEBRA))
665 zlog_debug("%s: No zebra instance to talk to, not installing remote vtep",
666 __PRETTY_FUNCTION__);
667 return 0;
668 }
669
670 s = zclient->obuf;
671 stream_reset(s);
672
673 zclient_create_header(
674 s, add ? ZEBRA_REMOTE_VTEP_ADD : ZEBRA_REMOTE_VTEP_DEL,
675 bgp->vrf_id);
676 stream_putl(s, vpn->vni);
677 if (is_evpn_prefix_ipaddr_v4(p))
678 stream_put_in_addr(s, &p->prefix.imet_addr.ip.ipaddr_v4);
679 else if (is_evpn_prefix_ipaddr_v6(p)) {
680 flog_err(
681 EC_BGP_VTEP_INVALID,
682 "Bad remote IP when trying to %s remote VTEP for VNI %u",
683 add ? "ADD" : "DEL", vpn->vni);
684 return -1;
685 }
686 stream_putl(s, flood_control);
687
688 stream_putw_at(s, 0, stream_get_endp(s));
689
690 if (bgp_debug_zebra(NULL))
691 zlog_debug("Tx %s Remote VTEP, VNI %u remote VTEP %s",
692 add ? "ADD" : "DEL", vpn->vni,
693 inet_ntoa(p->prefix.imet_addr.ip.ipaddr_v4));
694
695 return zclient_send_message(zclient);
696 }
697
698 /*
699 * Build extended community for EVPN ES (type-4) route
700 */
701 static void build_evpn_type4_route_extcomm(struct evpnes *es,
702 struct attr *attr)
703 {
704 struct ecommunity ecom_encap;
705 struct ecommunity ecom_es_rt;
706 struct ecommunity_val eval;
707 struct ecommunity_val eval_es_rt;
708 bgp_encap_types tnl_type;
709 struct ethaddr mac;
710
711 /* Encap */
712 tnl_type = BGP_ENCAP_TYPE_VXLAN;
713 memset(&ecom_encap, 0, sizeof(ecom_encap));
714 encode_encap_extcomm(tnl_type, &eval);
715 ecom_encap.size = 1;
716 ecom_encap.val = (uint8_t *)eval.val;
717 attr->ecommunity = ecommunity_dup(&ecom_encap);
718
719 /* ES import RT */
720 memset(&mac, 0, sizeof(struct ethaddr));
721 memset(&ecom_es_rt, 0, sizeof(ecom_es_rt));
722 es_get_system_mac(&es->esi, &mac);
723 encode_es_rt_extcomm(&eval_es_rt, &mac);
724 ecom_es_rt.size = 1;
725 ecom_es_rt.val = (uint8_t *)eval_es_rt.val;
726 attr->ecommunity =
727 ecommunity_merge(attr->ecommunity, &ecom_es_rt);
728
729 attr->flag |= ATTR_FLAG_BIT(BGP_ATTR_EXT_COMMUNITIES);
730 }
731
732 /*
733 * Build extended communities for EVPN prefix route.
734 */
735 static void build_evpn_type5_route_extcomm(struct bgp *bgp_vrf,
736 struct attr *attr)
737 {
738 struct ecommunity ecom_encap;
739 struct ecommunity ecom_rmac;
740 struct ecommunity_val eval;
741 struct ecommunity_val eval_rmac;
742 bgp_encap_types tnl_type;
743 struct listnode *node, *nnode;
744 struct ecommunity *ecom;
745 struct list *vrf_export_rtl = NULL;
746
747 /* Encap */
748 tnl_type = BGP_ENCAP_TYPE_VXLAN;
749 memset(&ecom_encap, 0, sizeof(ecom_encap));
750 encode_encap_extcomm(tnl_type, &eval);
751 ecom_encap.size = 1;
752 ecom_encap.val = (uint8_t *)eval.val;
753
754 /* Add Encap */
755 attr->ecommunity = ecommunity_dup(&ecom_encap);
756
757 /* Add the export RTs for L3VNI/VRF */
758 vrf_export_rtl = bgp_vrf->vrf_export_rtl;
759 for (ALL_LIST_ELEMENTS(vrf_export_rtl, node, nnode, ecom))
760 attr->ecommunity =
761 ecommunity_merge(attr->ecommunity, ecom);
762
763 /* add the router mac extended community */
764 if (!is_zero_mac(&attr->rmac)) {
765 memset(&ecom_rmac, 0, sizeof(ecom_rmac));
766 encode_rmac_extcomm(&eval_rmac, &attr->rmac);
767 ecom_rmac.size = 1;
768 ecom_rmac.val = (uint8_t *)eval_rmac.val;
769 attr->ecommunity =
770 ecommunity_merge(attr->ecommunity, &ecom_rmac);
771 }
772
773 attr->flag |= ATTR_FLAG_BIT(BGP_ATTR_EXT_COMMUNITIES);
774 }
775
776 /*
777 * Build extended communities for EVPN route.
778 * This function is applicable for type-2 and type-3 routes. The layer-2 RT
779 * and ENCAP extended communities are applicable for all routes.
780 * The default gateway extended community and MAC mobility (sticky) extended
781 * community are added as needed based on passed settings - only for type-2
782 * routes. Likewise, the layer-3 RT and Router MAC extended communities are
783 * added, if present, based on passed settings - only for non-link-local
784 * type-2 routes.
785 */
786 static void build_evpn_route_extcomm(struct bgpevpn *vpn, struct attr *attr,
787 int add_l3_ecomm)
788 {
789 struct ecommunity ecom_encap;
790 struct ecommunity ecom_sticky;
791 struct ecommunity ecom_default_gw;
792 struct ecommunity ecom_rmac;
793 struct ecommunity ecom_na;
794 struct ecommunity_val eval;
795 struct ecommunity_val eval_sticky;
796 struct ecommunity_val eval_default_gw;
797 struct ecommunity_val eval_rmac;
798 struct ecommunity_val eval_na;
799
800 bgp_encap_types tnl_type;
801 struct listnode *node, *nnode;
802 struct ecommunity *ecom;
803 uint32_t seqnum;
804 struct list *vrf_export_rtl = NULL;
805
806 /* Encap */
807 tnl_type = BGP_ENCAP_TYPE_VXLAN;
808 memset(&ecom_encap, 0, sizeof(ecom_encap));
809 encode_encap_extcomm(tnl_type, &eval);
810 ecom_encap.size = 1;
811 ecom_encap.val = (uint8_t *)eval.val;
812
813 /* Add Encap */
814 attr->ecommunity = ecommunity_dup(&ecom_encap);
815
816 /* Add the export RTs for L2VNI */
817 for (ALL_LIST_ELEMENTS(vpn->export_rtl, node, nnode, ecom))
818 attr->ecommunity = ecommunity_merge(attr->ecommunity, ecom);
819
820 /* Add the export RTs for L3VNI if told to - caller determines
821 * when this should be done.
822 */
823 if (add_l3_ecomm) {
824 vrf_export_rtl = bgpevpn_get_vrf_export_rtl(vpn);
825 if (vrf_export_rtl && !list_isempty(vrf_export_rtl)) {
826 for (ALL_LIST_ELEMENTS(vrf_export_rtl, node, nnode,
827 ecom))
828 attr->ecommunity = ecommunity_merge(
829 attr->ecommunity, ecom);
830 }
831 }
832
833 /* Add MAC mobility (sticky) if needed. */
834 if (attr->sticky) {
835 seqnum = 0;
836 memset(&ecom_sticky, 0, sizeof(ecom_sticky));
837 encode_mac_mobility_extcomm(1, seqnum, &eval_sticky);
838 ecom_sticky.size = 1;
839 ecom_sticky.val = (uint8_t *)eval_sticky.val;
840 attr->ecommunity =
841 ecommunity_merge(attr->ecommunity, &ecom_sticky);
842 }
843
844 /* Add RMAC, if told to. */
845 if (add_l3_ecomm) {
846 memset(&ecom_rmac, 0, sizeof(ecom_rmac));
847 encode_rmac_extcomm(&eval_rmac, &attr->rmac);
848 ecom_rmac.size = 1;
849 ecom_rmac.val = (uint8_t *)eval_rmac.val;
850 attr->ecommunity =
851 ecommunity_merge(attr->ecommunity, &ecom_rmac);
852 }
853
854 /* Add default gateway, if needed. */
855 if (attr->default_gw) {
856 memset(&ecom_default_gw, 0, sizeof(ecom_default_gw));
857 encode_default_gw_extcomm(&eval_default_gw);
858 ecom_default_gw.size = 1;
859 ecom_default_gw.val = (uint8_t *)eval_default_gw.val;
860 attr->ecommunity =
861 ecommunity_merge(attr->ecommunity, &ecom_default_gw);
862 }
863
864 if (attr->router_flag) {
865 memset(&ecom_na, 0, sizeof(ecom_na));
866 encode_na_flag_extcomm(&eval_na, attr->router_flag);
867 ecom_na.size = 1;
868 ecom_na.val = (uint8_t *)eval_na.val;
869 attr->ecommunity = ecommunity_merge(attr->ecommunity,
870 &ecom_na);
871 }
872
873 attr->flag |= ATTR_FLAG_BIT(BGP_ATTR_EXT_COMMUNITIES);
874 }
875
876 /*
877 * Add MAC mobility extended community to attribute.
878 */
879 static void add_mac_mobility_to_attr(uint32_t seq_num, struct attr *attr)
880 {
881 struct ecommunity ecom_tmp;
882 struct ecommunity_val eval;
883 uint8_t *ecom_val_ptr;
884 int i;
885 uint8_t *pnt;
886 int type = 0;
887 int sub_type = 0;
888
889 /* Build MM */
890 encode_mac_mobility_extcomm(0, seq_num, &eval);
891
892 /* Find current MM ecommunity */
893 ecom_val_ptr = NULL;
894
895 if (attr->ecommunity) {
896 for (i = 0; i < attr->ecommunity->size; i++) {
897 pnt = attr->ecommunity->val + (i * 8);
898 type = *pnt++;
899 sub_type = *pnt++;
900
901 if (type == ECOMMUNITY_ENCODE_EVPN
902 && sub_type
903 == ECOMMUNITY_EVPN_SUBTYPE_MACMOBILITY) {
904 ecom_val_ptr = (uint8_t *)(attr->ecommunity->val
905 + (i * 8));
906 break;
907 }
908 }
909 }
910
911 /* Update the existing MM ecommunity */
912 if (ecom_val_ptr) {
913 memcpy(ecom_val_ptr, eval.val, sizeof(char) * ECOMMUNITY_SIZE);
914 }
915 /* Add MM to existing */
916 else {
917 memset(&ecom_tmp, 0, sizeof(ecom_tmp));
918 ecom_tmp.size = 1;
919 ecom_tmp.val = (uint8_t *)eval.val;
920
921 if (attr->ecommunity)
922 attr->ecommunity =
923 ecommunity_merge(attr->ecommunity, &ecom_tmp);
924 else
925 attr->ecommunity = ecommunity_dup(&ecom_tmp);
926 }
927 }
928
929 /* Install EVPN route into zebra. */
930 static int evpn_zebra_install(struct bgp *bgp, struct bgpevpn *vpn,
931 struct prefix_evpn *p, struct bgp_path_info *pi)
932 {
933 int ret;
934 uint8_t flags;
935 int flood_control;
936
937 if (p->prefix.route_type == BGP_EVPN_MAC_IP_ROUTE) {
938 flags = 0;
939 if (pi->attr->sticky)
940 SET_FLAG(flags, ZEBRA_MACIP_TYPE_STICKY);
941 if (pi->attr->default_gw)
942 SET_FLAG(flags, ZEBRA_MACIP_TYPE_GW);
943 if (is_evpn_prefix_ipaddr_v6(p) &&
944 pi->attr->router_flag)
945 SET_FLAG(flags, ZEBRA_MACIP_TYPE_ROUTER_FLAG);
946 ret = bgp_zebra_send_remote_macip(
947 bgp, vpn, p, pi->attr->nexthop, 1, flags,
948 mac_mobility_seqnum(pi->attr));
949 } else {
950 switch (pi->attr->pmsi_tnl_type) {
951 case PMSI_TNLTYPE_INGR_REPL:
952 flood_control = VXLAN_FLOOD_HEAD_END_REPL;
953 break;
954
955 case PMSI_TNLTYPE_PIM_SM:
956 flood_control = VXLAN_FLOOD_PIM_SM;
957 break;
958
959 default:
960 flood_control = VXLAN_FLOOD_DISABLED;
961 break;
962 }
963 ret = bgp_zebra_send_remote_vtep(bgp, vpn, p, flood_control, 1);
964 }
965
966 return ret;
967 }
968
969 /* Uninstall EVPN route from zebra. */
970 static int evpn_zebra_uninstall(struct bgp *bgp, struct bgpevpn *vpn,
971 struct prefix_evpn *p,
972 struct in_addr remote_vtep_ip)
973 {
974 int ret;
975
976 if (p->prefix.route_type == BGP_EVPN_MAC_IP_ROUTE)
977 ret = bgp_zebra_send_remote_macip(bgp, vpn, p, remote_vtep_ip,
978 0, 0, 0);
979 else
980 ret = bgp_zebra_send_remote_vtep(bgp, vpn, p,
981 VXLAN_FLOOD_DISABLED, 0);
982
983 return ret;
984 }
985
986 /*
987 * Due to MAC mobility, the prior "local" best route has been supplanted
988 * by a "remote" best route. The prior route has to be deleted and withdrawn
989 * from peers.
990 */
991 static void evpn_delete_old_local_route(struct bgp *bgp, struct bgpevpn *vpn,
992 struct bgp_node *rn,
993 struct bgp_path_info *old_local)
994 {
995 struct bgp_node *global_rn;
996 struct bgp_path_info *pi;
997 afi_t afi = AFI_L2VPN;
998 safi_t safi = SAFI_EVPN;
999
1000 /* Locate route node in the global EVPN routing table. Note that
1001 * this table is a 2-level tree (RD-level + Prefix-level) similar to
1002 * L3VPN routes.
1003 */
1004 global_rn = bgp_afi_node_lookup(bgp->rib[afi][safi], afi, safi,
1005 (struct prefix *)&rn->p, &vpn->prd);
1006 if (global_rn) {
1007 /* Delete route entry in the global EVPN table. */
1008 delete_evpn_route_entry(bgp, afi, safi, global_rn, &pi);
1009
1010 /* Schedule for processing - withdraws to peers happen from
1011 * this table.
1012 */
1013 if (pi)
1014 bgp_process(bgp, global_rn, afi, safi);
1015 bgp_unlock_node(global_rn);
1016 }
1017
1018 /* Delete route entry in the VNI route table, caller to remove. */
1019 bgp_path_info_delete(rn, old_local);
1020 }
1021
1022 static struct in_addr *es_vtep_new(struct in_addr vtep)
1023 {
1024 struct in_addr *ip;
1025
1026 ip = XCALLOC(MTYPE_BGP_EVPN_ES_VTEP, sizeof(struct in_addr));
1027
1028 ip->s_addr = vtep.s_addr;
1029 return ip;
1030 }
1031
1032 static void es_vtep_free(struct in_addr *ip)
1033 {
1034 XFREE(MTYPE_BGP_EVPN_ES_VTEP, ip);
1035 }
1036
1037 /* check if VTEP is already part of the list */
1038 static int is_vtep_present_in_list(struct list *list,
1039 struct in_addr vtep)
1040 {
1041 struct listnode *node = NULL;
1042 struct in_addr *tmp;
1043
1044 for (ALL_LIST_ELEMENTS_RO(list, node, tmp)) {
1045 if (tmp->s_addr == vtep.s_addr)
1046 return 1;
1047 }
1048 return 0;
1049 }
1050
1051 /*
1052 * Best path for ES route was changed,
1053 * update the list of VTEPs for this ES
1054 */
1055 static int evpn_es_install_vtep(struct bgp *bgp,
1056 struct evpnes *es,
1057 struct prefix_evpn *p,
1058 struct in_addr rvtep)
1059 {
1060 struct in_addr *vtep_ip;
1061
1062 if (is_vtep_present_in_list(es->vtep_list, rvtep))
1063 return 0;
1064
1065
1066 vtep_ip = es_vtep_new(rvtep);
1067 if (vtep_ip)
1068 listnode_add_sort(es->vtep_list, vtep_ip);
1069 return 0;
1070 }
1071
1072 /*
1073 * Best path for ES route was changed,
1074 * update the list of VTEPs for this ES
1075 */
1076 static int evpn_es_uninstall_vtep(struct bgp *bgp,
1077 struct evpnes *es,
1078 struct prefix_evpn *p,
1079 struct in_addr rvtep)
1080 {
1081 struct listnode *node, *nnode, *node_to_del = NULL;
1082 struct in_addr *tmp;
1083
1084 for (ALL_LIST_ELEMENTS(es->vtep_list, node, nnode, tmp)) {
1085 if (tmp->s_addr == rvtep.s_addr) {
1086 es_vtep_free(tmp);
1087 node_to_del = node;
1088 }
1089 }
1090
1091 if (node_to_del)
1092 list_delete_node(es->vtep_list, node_to_del);
1093
1094 return 0;
1095 }
1096
1097 /*
1098 * Calculate the best path for a ES(type-4) route.
1099 */
1100 static int evpn_es_route_select_install(struct bgp *bgp,
1101 struct evpnes *es,
1102 struct bgp_node *rn)
1103 {
1104 int ret = 0;
1105 afi_t afi = AFI_L2VPN;
1106 safi_t safi = SAFI_EVPN;
1107 struct bgp_path_info *old_select; /* old best */
1108 struct bgp_path_info *new_select; /* new best */
1109 struct bgp_path_info_pair old_and_new;
1110
1111 /* Compute the best path. */
1112 bgp_best_selection(bgp, rn, &bgp->maxpaths[afi][safi],
1113 &old_and_new, afi, safi);
1114 old_select = old_and_new.old;
1115 new_select = old_and_new.new;
1116
1117 /*
1118 * If the best path hasn't changed - see if something needs to be
1119 * updated
1120 */
1121 if (old_select && old_select == new_select
1122 && old_select->type == ZEBRA_ROUTE_BGP
1123 && old_select->sub_type == BGP_ROUTE_IMPORTED
1124 && !CHECK_FLAG(rn->flags, BGP_NODE_USER_CLEAR)
1125 && !CHECK_FLAG(old_select->flags, BGP_PATH_ATTR_CHANGED)
1126 && !bgp_addpath_is_addpath_used(&bgp->tx_addpath, afi, safi)) {
1127 if (bgp_zebra_has_route_changed(rn, old_select)) {
1128 ret = evpn_es_install_vtep(bgp, es,
1129 (struct prefix_evpn *)&rn->p,
1130 old_select->attr->nexthop);
1131 }
1132 UNSET_FLAG(old_select->flags, BGP_PATH_MULTIPATH_CHG);
1133 bgp_zebra_clear_route_change_flags(rn);
1134 return ret;
1135 }
1136
1137 /* If the user did a "clear" this flag will be set */
1138 UNSET_FLAG(rn->flags, BGP_NODE_USER_CLEAR);
1139
1140 /*
1141 * bestpath has changed; update relevant fields and install or uninstall
1142 * into the zebra RIB.
1143 */
1144 if (old_select || new_select)
1145 bgp_bump_version(rn);
1146
1147 if (old_select)
1148 bgp_path_info_unset_flag(rn, old_select, BGP_PATH_SELECTED);
1149 if (new_select) {
1150 bgp_path_info_set_flag(rn, new_select, BGP_PATH_SELECTED);
1151 bgp_path_info_unset_flag(rn, new_select, BGP_PATH_ATTR_CHANGED);
1152 UNSET_FLAG(new_select->flags, BGP_PATH_MULTIPATH_CHG);
1153 }
1154
1155 if (new_select && new_select->type == ZEBRA_ROUTE_BGP
1156 && new_select->sub_type == BGP_ROUTE_IMPORTED) {
1157 ret = evpn_es_install_vtep(bgp, es,
1158 (struct prefix_evpn *)&rn->p,
1159 new_select->attr->nexthop);
1160 } else {
1161 if (old_select && old_select->type == ZEBRA_ROUTE_BGP
1162 && old_select->sub_type == BGP_ROUTE_IMPORTED)
1163 ret = evpn_es_uninstall_vtep(
1164 bgp, es, (struct prefix_evpn *)&rn->p,
1165 old_select->attr->nexthop);
1166 }
1167
1168 /* Clear any route change flags. */
1169 bgp_zebra_clear_route_change_flags(rn);
1170
1171 /* Reap old select bgp_path_info, if it has been removed */
1172 if (old_select && CHECK_FLAG(old_select->flags, BGP_PATH_REMOVED))
1173 bgp_path_info_reap(rn, old_select);
1174
1175 return ret;
1176 }
1177
1178 /*
1179 * Calculate the best path for an EVPN route. Install/update best path in zebra,
1180 * if appropriate.
1181 */
1182 static int evpn_route_select_install(struct bgp *bgp, struct bgpevpn *vpn,
1183 struct bgp_node *rn)
1184 {
1185 struct bgp_path_info *old_select, *new_select;
1186 struct bgp_path_info_pair old_and_new;
1187 afi_t afi = AFI_L2VPN;
1188 safi_t safi = SAFI_EVPN;
1189 int ret = 0;
1190
1191 /* Compute the best path. */
1192 bgp_best_selection(bgp, rn, &bgp->maxpaths[afi][safi], &old_and_new,
1193 afi, safi);
1194 old_select = old_and_new.old;
1195 new_select = old_and_new.new;
1196
1197 /* If the best path hasn't changed - see if there is still something to
1198 * update
1199 * to zebra RIB.
1200 */
1201 if (old_select && old_select == new_select
1202 && old_select->type == ZEBRA_ROUTE_BGP
1203 && old_select->sub_type == BGP_ROUTE_IMPORTED
1204 && !CHECK_FLAG(rn->flags, BGP_NODE_USER_CLEAR)
1205 && !CHECK_FLAG(old_select->flags, BGP_PATH_ATTR_CHANGED)
1206 && !bgp_addpath_is_addpath_used(&bgp->tx_addpath, afi, safi)) {
1207 if (bgp_zebra_has_route_changed(rn, old_select))
1208 ret = evpn_zebra_install(
1209 bgp, vpn, (struct prefix_evpn *)&rn->p,
1210 old_select);
1211 UNSET_FLAG(old_select->flags, BGP_PATH_MULTIPATH_CHG);
1212 bgp_zebra_clear_route_change_flags(rn);
1213 return ret;
1214 }
1215
1216 /* If the user did a "clear" this flag will be set */
1217 UNSET_FLAG(rn->flags, BGP_NODE_USER_CLEAR);
1218
1219 /* bestpath has changed; update relevant fields and install or uninstall
1220 * into the zebra RIB.
1221 */
1222 if (old_select || new_select)
1223 bgp_bump_version(rn);
1224
1225 if (old_select)
1226 bgp_path_info_unset_flag(rn, old_select, BGP_PATH_SELECTED);
1227 if (new_select) {
1228 bgp_path_info_set_flag(rn, new_select, BGP_PATH_SELECTED);
1229 bgp_path_info_unset_flag(rn, new_select, BGP_PATH_ATTR_CHANGED);
1230 UNSET_FLAG(new_select->flags, BGP_PATH_MULTIPATH_CHG);
1231 }
1232
1233 if (new_select && new_select->type == ZEBRA_ROUTE_BGP
1234 && new_select->sub_type == BGP_ROUTE_IMPORTED) {
1235 ret = evpn_zebra_install(bgp, vpn, (struct prefix_evpn *)&rn->p,
1236 new_select);
1237
1238 /* If an old best existed and it was a "local" route, the only
1239 * reason
1240 * it would be supplanted is due to MAC mobility procedures. So,
1241 * we
1242 * need to do an implicit delete and withdraw that route from
1243 * peers.
1244 */
1245 if (old_select && old_select->peer == bgp->peer_self
1246 && old_select->type == ZEBRA_ROUTE_BGP
1247 && old_select->sub_type == BGP_ROUTE_STATIC)
1248 evpn_delete_old_local_route(bgp, vpn, rn, old_select);
1249 } else {
1250 if (old_select && old_select->type == ZEBRA_ROUTE_BGP
1251 && old_select->sub_type == BGP_ROUTE_IMPORTED)
1252 ret = evpn_zebra_uninstall(bgp, vpn,
1253 (struct prefix_evpn *)&rn->p,
1254 old_select->attr->nexthop);
1255 }
1256
1257 /* Clear any route change flags. */
1258 bgp_zebra_clear_route_change_flags(rn);
1259
1260 /* Reap old select bgp_path_info, if it has been removed */
1261 if (old_select && CHECK_FLAG(old_select->flags, BGP_PATH_REMOVED))
1262 bgp_path_info_reap(rn, old_select);
1263
1264 return ret;
1265 }
1266
1267 /*
1268 * Return true if the local ri for this rn is of type gateway mac
1269 */
1270 static int evpn_route_is_def_gw(struct bgp *bgp, struct bgp_node *rn)
1271 {
1272 struct bgp_path_info *tmp_pi = NULL;
1273 struct bgp_path_info *local_pi = NULL;
1274
1275 local_pi = NULL;
1276 for (tmp_pi = bgp_node_get_bgp_path_info(rn); tmp_pi;
1277 tmp_pi = tmp_pi->next) {
1278 if (tmp_pi->peer == bgp->peer_self
1279 && tmp_pi->type == ZEBRA_ROUTE_BGP
1280 && tmp_pi->sub_type == BGP_ROUTE_STATIC)
1281 local_pi = tmp_pi;
1282 }
1283
1284 if (!local_pi)
1285 return 0;
1286
1287 return local_pi->attr->default_gw;
1288 }
1289
1290
1291 /*
1292 * Return true if the local ri for this rn has sticky set
1293 */
1294 static int evpn_route_is_sticky(struct bgp *bgp, struct bgp_node *rn)
1295 {
1296 struct bgp_path_info *tmp_pi;
1297 struct bgp_path_info *local_pi;
1298
1299 local_pi = NULL;
1300 for (tmp_pi = bgp_node_get_bgp_path_info(rn); tmp_pi;
1301 tmp_pi = tmp_pi->next) {
1302 if (tmp_pi->peer == bgp->peer_self
1303 && tmp_pi->type == ZEBRA_ROUTE_BGP
1304 && tmp_pi->sub_type == BGP_ROUTE_STATIC)
1305 local_pi = tmp_pi;
1306 }
1307
1308 if (!local_pi)
1309 return 0;
1310
1311 return local_pi->attr->sticky;
1312 }
1313
1314 /*
1315 * create or update EVPN type4 route entry.
1316 * This could be in the ES table or the global table.
1317 * TODO: handle remote ES (type4) routes as well
1318 */
1319 static int update_evpn_type4_route_entry(struct bgp *bgp, struct evpnes *es,
1320 afi_t afi, safi_t safi,
1321 struct bgp_node *rn, struct attr *attr,
1322 int add, struct bgp_path_info **ri,
1323 int *route_changed)
1324 {
1325 char buf[ESI_STR_LEN];
1326 char buf1[INET6_ADDRSTRLEN];
1327 struct bgp_path_info *tmp_pi = NULL;
1328 struct bgp_path_info *local_pi = NULL; /* local route entry if any */
1329 struct bgp_path_info *remote_pi = NULL; /* remote route entry if any */
1330 struct attr *attr_new = NULL;
1331 struct prefix_evpn *evp = NULL;
1332
1333 *ri = NULL;
1334 *route_changed = 1;
1335 evp = (struct prefix_evpn *)&rn->p;
1336
1337 /* locate the local and remote entries if any */
1338 for (tmp_pi = bgp_node_get_bgp_path_info(rn); tmp_pi;
1339 tmp_pi = tmp_pi->next) {
1340 if (tmp_pi->peer == bgp->peer_self
1341 && tmp_pi->type == ZEBRA_ROUTE_BGP
1342 && tmp_pi->sub_type == BGP_ROUTE_STATIC)
1343 local_pi = tmp_pi;
1344 if (tmp_pi->type == ZEBRA_ROUTE_BGP
1345 && tmp_pi->sub_type == BGP_ROUTE_IMPORTED
1346 && CHECK_FLAG(tmp_pi->flags, BGP_PATH_VALID))
1347 remote_pi = tmp_pi;
1348 }
1349
1350 /* we don't expect to see a remote_ri at this point.
1351 * An ES route has esi + vtep_ip as the key,
1352 * We shouldn't see the same route from any other vtep.
1353 */
1354 if (remote_pi) {
1355 flog_err(
1356 EC_BGP_ES_INVALID,
1357 "%u ERROR: local es route for ESI: %s Vtep %s also learnt from remote",
1358 bgp->vrf_id,
1359 esi_to_str(&evp->prefix.es_addr.esi, buf, sizeof(buf)),
1360 ipaddr2str(&es->originator_ip, buf1, sizeof(buf1)));
1361 return -1;
1362 }
1363
1364 if (!local_pi && !add)
1365 return 0;
1366
1367 /* create or update the entry */
1368 if (!local_pi) {
1369
1370 /* Add or update attribute to hash */
1371 attr_new = bgp_attr_intern(attr);
1372
1373 /* Create new route with its attribute. */
1374 tmp_pi = info_make(ZEBRA_ROUTE_BGP, BGP_ROUTE_STATIC, 0,
1375 bgp->peer_self, attr_new, rn);
1376 SET_FLAG(tmp_pi->flags, BGP_PATH_VALID);
1377
1378 /* add the newly created path to the route-node */
1379 bgp_path_info_add(rn, tmp_pi);
1380 } else {
1381 tmp_pi = local_pi;
1382 if (attrhash_cmp(tmp_pi->attr, attr)
1383 && !CHECK_FLAG(tmp_pi->flags, BGP_PATH_REMOVED))
1384 *route_changed = 0;
1385 else {
1386 /* The attribute has changed.
1387 * Add (or update) attribute to hash. */
1388 attr_new = bgp_attr_intern(attr);
1389 bgp_path_info_set_flag(rn, tmp_pi,
1390 BGP_PATH_ATTR_CHANGED);
1391
1392 /* Restore route, if needed. */
1393 if (CHECK_FLAG(tmp_pi->flags, BGP_PATH_REMOVED))
1394 bgp_path_info_restore(rn, tmp_pi);
1395
1396 /* Unintern existing, set to new. */
1397 bgp_attr_unintern(&tmp_pi->attr);
1398 tmp_pi->attr = attr_new;
1399 tmp_pi->uptime = bgp_clock();
1400 }
1401 }
1402
1403 /* Return back the route entry. */
1404 *ri = tmp_pi;
1405 return 0;
1406 }
1407
1408 /* update evpn es (type-4) route */
1409 static int update_evpn_type4_route(struct bgp *bgp,
1410 struct evpnes *es,
1411 struct prefix_evpn *p)
1412 {
1413 int ret = 0;
1414 int route_changed = 0;
1415 char buf[ESI_STR_LEN];
1416 char buf1[INET6_ADDRSTRLEN];
1417 afi_t afi = AFI_L2VPN;
1418 safi_t safi = SAFI_EVPN;
1419 struct attr attr;
1420 struct attr *attr_new = NULL;
1421 struct bgp_node *rn = NULL;
1422 struct bgp_path_info *pi = NULL;
1423
1424 memset(&attr, 0, sizeof(struct attr));
1425
1426 /* Build path-attribute for this route. */
1427 bgp_attr_default_set(&attr, BGP_ORIGIN_IGP);
1428 attr.nexthop = es->originator_ip.ipaddr_v4;
1429 attr.mp_nexthop_global_in = es->originator_ip.ipaddr_v4;
1430 attr.mp_nexthop_len = BGP_ATTR_NHLEN_IPV4;
1431
1432 /* Set up extended community. */
1433 build_evpn_type4_route_extcomm(es, &attr);
1434
1435 /* First, create (or fetch) route node within the ESI. */
1436 /* NOTE: There is no RD here. */
1437 rn = bgp_node_get(es->route_table, (struct prefix *)p);
1438
1439 /* Create or update route entry. */
1440 ret = update_evpn_type4_route_entry(bgp, es, afi, safi, rn, &attr, 1,
1441 &pi, &route_changed);
1442 if (ret != 0) {
1443 flog_err(EC_BGP_ES_INVALID,
1444 "%u ERROR: Failed to updated ES route ESI: %s VTEP %s",
1445 bgp->vrf_id,
1446 esi_to_str(&p->prefix.es_addr.esi, buf, sizeof(buf)),
1447 ipaddr2str(&es->originator_ip, buf1, sizeof(buf1)));
1448 }
1449
1450 assert(pi);
1451 attr_new = pi->attr;
1452
1453 /* Perform route selection;
1454 * this is just to set the flags correctly
1455 * as local route in the ES always wins.
1456 */
1457 evpn_es_route_select_install(bgp, es, rn);
1458 bgp_unlock_node(rn);
1459
1460 /* If this is a new route or some attribute has changed, export the
1461 * route to the global table. The route will be advertised to peers
1462 * from there. Note that this table is a 2-level tree (RD-level +
1463 * Prefix-level) similar to L3VPN routes.
1464 */
1465 if (route_changed) {
1466 struct bgp_path_info *global_pi;
1467
1468 rn = bgp_afi_node_get(bgp->rib[afi][safi], afi, safi,
1469 (struct prefix *)p, &es->prd);
1470 update_evpn_type4_route_entry(bgp, es, afi, safi, rn, attr_new,
1471 1, &global_pi, &route_changed);
1472
1473 /* Schedule for processing and unlock node. */
1474 bgp_process(bgp, rn, afi, safi);
1475 bgp_unlock_node(rn);
1476 }
1477
1478 /* Unintern temporary. */
1479 aspath_unintern(&attr.aspath);
1480 return 0;
1481 }
1482
1483 static int update_evpn_type5_route_entry(struct bgp *bgp_evpn,
1484 struct bgp *bgp_vrf, afi_t afi,
1485 safi_t safi, struct bgp_node *rn,
1486 struct attr *attr, int *route_changed)
1487 {
1488 struct attr *attr_new = NULL;
1489 struct bgp_path_info *pi = NULL;
1490 mpls_label_t label = MPLS_INVALID_LABEL;
1491 struct bgp_path_info *local_pi = NULL;
1492 struct bgp_path_info *tmp_pi = NULL;
1493
1494 *route_changed = 0;
1495 /* locate the local route entry if any */
1496 for (tmp_pi = bgp_node_get_bgp_path_info(rn); tmp_pi;
1497 tmp_pi = tmp_pi->next) {
1498 if (tmp_pi->peer == bgp_evpn->peer_self
1499 && tmp_pi->type == ZEBRA_ROUTE_BGP
1500 && tmp_pi->sub_type == BGP_ROUTE_STATIC)
1501 local_pi = tmp_pi;
1502 }
1503
1504 /*
1505 * create a new route entry if one doesn't exist.
1506 * Otherwise see if route attr has changed
1507 */
1508 if (!local_pi) {
1509
1510 /* route has changed as this is the first entry */
1511 *route_changed = 1;
1512
1513 /* Add (or update) attribute to hash. */
1514 attr_new = bgp_attr_intern(attr);
1515
1516 /* create the route info from attribute */
1517 pi = info_make(ZEBRA_ROUTE_BGP, BGP_ROUTE_STATIC, 0,
1518 bgp_evpn->peer_self, attr_new, rn);
1519 SET_FLAG(pi->flags, BGP_PATH_VALID);
1520
1521 /* Type-5 routes advertise the L3-VNI */
1522 bgp_path_info_extra_get(pi);
1523 vni2label(bgp_vrf->l3vni, &label);
1524 memcpy(&pi->extra->label, &label, sizeof(label));
1525 pi->extra->num_labels = 1;
1526
1527 /* add the route entry to route node*/
1528 bgp_path_info_add(rn, pi);
1529 } else {
1530
1531 tmp_pi = local_pi;
1532 if (!attrhash_cmp(tmp_pi->attr, attr)) {
1533
1534 /* attribute changed */
1535 *route_changed = 1;
1536
1537 /* The attribute has changed. */
1538 /* Add (or update) attribute to hash. */
1539 attr_new = bgp_attr_intern(attr);
1540 bgp_path_info_set_flag(rn, tmp_pi,
1541 BGP_PATH_ATTR_CHANGED);
1542
1543 /* Restore route, if needed. */
1544 if (CHECK_FLAG(tmp_pi->flags, BGP_PATH_REMOVED))
1545 bgp_path_info_restore(rn, tmp_pi);
1546
1547 /* Unintern existing, set to new. */
1548 bgp_attr_unintern(&tmp_pi->attr);
1549 tmp_pi->attr = attr_new;
1550 tmp_pi->uptime = bgp_clock();
1551 }
1552 }
1553 return 0;
1554 }
1555
1556 /* update evpn type-5 route entry */
1557 static int update_evpn_type5_route(struct bgp *bgp_vrf, struct prefix_evpn *evp,
1558 struct attr *src_attr)
1559 {
1560 afi_t afi = AFI_L2VPN;
1561 safi_t safi = SAFI_EVPN;
1562 struct attr attr;
1563 struct bgp_node *rn = NULL;
1564 struct bgp *bgp_evpn = NULL;
1565 int route_changed = 0;
1566
1567 bgp_evpn = bgp_get_evpn();
1568 if (!bgp_evpn)
1569 return 0;
1570
1571 /* Build path attribute for this route - use the source attr, if
1572 * present, else treat as locally originated.
1573 */
1574 if (src_attr)
1575 attr = *src_attr;
1576 else {
1577 memset(&attr, 0, sizeof(struct attr));
1578 bgp_attr_default_set(&attr, BGP_ORIGIN_IGP);
1579 }
1580
1581 /* Advertise Primary IP (PIP) is enabled, send individual
1582 * IP (default instance router-id) as nexthop.
1583 * PIP is disabled or vrr interface is not present
1584 * use anycast-IP as nexthop and anycast RMAC.
1585 */
1586 if (!bgp_vrf->evpn_info->advertise_pip ||
1587 (!bgp_vrf->evpn_info->is_anycast_mac)) {
1588 attr.nexthop = bgp_vrf->originator_ip;
1589 attr.mp_nexthop_global_in = bgp_vrf->originator_ip;
1590 memcpy(&attr.rmac, &bgp_vrf->rmac, ETH_ALEN);
1591 } else {
1592 /* copy sys rmac */
1593 memcpy(&attr.rmac, &bgp_vrf->evpn_info->pip_rmac, ETH_ALEN);
1594 if (bgp_vrf->evpn_info->pip_ip.s_addr != INADDR_ANY) {
1595 attr.nexthop = bgp_vrf->evpn_info->pip_ip;
1596 attr.mp_nexthop_global_in = bgp_vrf->evpn_info->pip_ip;
1597 } else if (bgp_vrf->evpn_info->pip_ip.s_addr == INADDR_ANY)
1598 if (bgp_debug_zebra(NULL)) {
1599 char buf1[PREFIX_STRLEN];
1600
1601 zlog_debug("VRF %s evp %s advertise-pip primary ip is not configured",
1602 vrf_id_to_name(bgp_vrf->vrf_id),
1603 prefix2str(evp, buf1, sizeof(buf1)));
1604 }
1605 }
1606
1607 if (bgp_debug_zebra(NULL)) {
1608 char buf[ETHER_ADDR_STRLEN];
1609 char buf1[PREFIX_STRLEN];
1610 char buf2[INET6_ADDRSTRLEN];
1611
1612 zlog_debug("VRF %s type-5 route evp %s RMAC %s nexthop %s",
1613 vrf_id_to_name(bgp_vrf->vrf_id),
1614 prefix2str(evp, buf1, sizeof(buf1)),
1615 prefix_mac2str(&attr.rmac, buf, sizeof(buf)),
1616 inet_ntop(AF_INET, &attr.nexthop, buf2,
1617 INET_ADDRSTRLEN));
1618 }
1619
1620 attr.mp_nexthop_len = BGP_ATTR_NHLEN_IPV4;
1621
1622 /* Setup RT and encap extended community */
1623 build_evpn_type5_route_extcomm(bgp_vrf, &attr);
1624
1625 /* get the route node in global table */
1626 rn = bgp_afi_node_get(bgp_evpn->rib[afi][safi], afi, safi,
1627 (struct prefix *)evp, &bgp_vrf->vrf_prd);
1628 assert(rn);
1629
1630 /* create or update the route entry within the route node */
1631 update_evpn_type5_route_entry(bgp_evpn, bgp_vrf, afi, safi, rn, &attr,
1632 &route_changed);
1633
1634 /* schedule for processing and unlock node */
1635 if (route_changed) {
1636 bgp_process(bgp_evpn, rn, afi, safi);
1637 bgp_unlock_node(rn);
1638 }
1639
1640 /* uninten temporary */
1641 if (!src_attr)
1642 aspath_unintern(&attr.aspath);
1643 return 0;
1644 }
1645
1646 /*
1647 * Create or update EVPN route entry. This could be in the VNI route table
1648 * or the global route table.
1649 */
1650 static int update_evpn_route_entry(struct bgp *bgp, struct bgpevpn *vpn,
1651 afi_t afi, safi_t safi, struct bgp_node *rn,
1652 struct attr *attr, int add,
1653 struct bgp_path_info **pi, uint8_t flags,
1654 uint32_t seq)
1655 {
1656 struct bgp_path_info *tmp_pi;
1657 struct bgp_path_info *local_pi;
1658 struct attr *attr_new;
1659 mpls_label_t label[BGP_MAX_LABELS];
1660 uint32_t num_labels = 1;
1661 int route_change = 1;
1662 uint8_t sticky = 0;
1663 struct prefix_evpn *evp;
1664
1665 *pi = NULL;
1666 evp = (struct prefix_evpn *)&rn->p;
1667 memset(&label, 0, sizeof(label));
1668
1669 /* See if this is an update of an existing route, or a new add. */
1670 local_pi = NULL;
1671 for (tmp_pi = bgp_node_get_bgp_path_info(rn); tmp_pi;
1672 tmp_pi = tmp_pi->next) {
1673 if (tmp_pi->peer == bgp->peer_self
1674 && tmp_pi->type == ZEBRA_ROUTE_BGP
1675 && tmp_pi->sub_type == BGP_ROUTE_STATIC)
1676 local_pi = tmp_pi;
1677 }
1678
1679 /* If route doesn't exist already, create a new one, if told to.
1680 * Otherwise act based on whether the attributes of the route have
1681 * changed or not.
1682 */
1683 if (!local_pi && !add)
1684 return 0;
1685
1686 /* For non-GW MACs, update MAC mobility seq number, if needed. */
1687 if (seq && !CHECK_FLAG(flags, ZEBRA_MACIP_TYPE_GW))
1688 add_mac_mobility_to_attr(seq, attr);
1689
1690 if (!local_pi) {
1691 /* Add (or update) attribute to hash. */
1692 attr_new = bgp_attr_intern(attr);
1693
1694 /* Extract MAC mobility sequence number, if any. */
1695 attr_new->mm_seqnum =
1696 bgp_attr_mac_mobility_seqnum(attr_new, &sticky);
1697 attr_new->sticky = sticky;
1698
1699 /* Create new route with its attribute. */
1700 tmp_pi = info_make(ZEBRA_ROUTE_BGP, BGP_ROUTE_STATIC, 0,
1701 bgp->peer_self, attr_new, rn);
1702 SET_FLAG(tmp_pi->flags, BGP_PATH_VALID);
1703 bgp_path_info_extra_get(tmp_pi);
1704
1705 /* The VNI goes into the 'label' field of the route */
1706 vni2label(vpn->vni, &label[0]);
1707
1708 /* Type-2 routes may carry a second VNI - the L3-VNI.
1709 * Only attach second label if we are advertising two labels for
1710 * type-2 routes.
1711 */
1712 if (evp->prefix.route_type == BGP_EVPN_MAC_IP_ROUTE
1713 && CHECK_FLAG(vpn->flags, VNI_FLAG_USE_TWO_LABELS)) {
1714 vni_t l3vni;
1715
1716 l3vni = bgpevpn_get_l3vni(vpn);
1717 if (l3vni) {
1718 vni2label(l3vni, &label[1]);
1719 num_labels++;
1720 }
1721 }
1722
1723 memcpy(&tmp_pi->extra->label, label, sizeof(label));
1724 tmp_pi->extra->num_labels = num_labels;
1725 /* Mark route as self type-2 route */
1726 if (flags && CHECK_FLAG(flags, ZEBRA_MACIP_TYPE_SVI_IP))
1727 tmp_pi->extra->af_flags = BGP_EVPN_MACIP_TYPE_SVI_IP;
1728 bgp_path_info_add(rn, tmp_pi);
1729 } else {
1730 tmp_pi = local_pi;
1731 if (attrhash_cmp(tmp_pi->attr, attr)
1732 && !CHECK_FLAG(tmp_pi->flags, BGP_PATH_REMOVED))
1733 route_change = 0;
1734 else {
1735 /*
1736 * The attributes have changed, type-2 routes needs to
1737 * be advertised with right labels.
1738 */
1739 vni2label(vpn->vni, &label[0]);
1740 if (evp->prefix.route_type == BGP_EVPN_MAC_IP_ROUTE
1741 && CHECK_FLAG(vpn->flags,
1742 VNI_FLAG_USE_TWO_LABELS)) {
1743 vni_t l3vni;
1744
1745 l3vni = bgpevpn_get_l3vni(vpn);
1746 if (l3vni) {
1747 vni2label(l3vni, &label[1]);
1748 num_labels++;
1749 }
1750 }
1751 memcpy(&tmp_pi->extra->label, label, sizeof(label));
1752 tmp_pi->extra->num_labels = num_labels;
1753
1754 /* The attribute has changed. */
1755 /* Add (or update) attribute to hash. */
1756 attr_new = bgp_attr_intern(attr);
1757 bgp_path_info_set_flag(rn, tmp_pi,
1758 BGP_PATH_ATTR_CHANGED);
1759
1760 /* Extract MAC mobility sequence number, if any. */
1761 attr_new->mm_seqnum =
1762 bgp_attr_mac_mobility_seqnum(attr_new, &sticky);
1763 attr_new->sticky = sticky;
1764
1765 /* Restore route, if needed. */
1766 if (CHECK_FLAG(tmp_pi->flags, BGP_PATH_REMOVED))
1767 bgp_path_info_restore(rn, tmp_pi);
1768
1769 /* Unintern existing, set to new. */
1770 bgp_attr_unintern(&tmp_pi->attr);
1771 tmp_pi->attr = attr_new;
1772 tmp_pi->uptime = bgp_clock();
1773 }
1774 }
1775
1776 /* Return back the route entry. */
1777 *pi = tmp_pi;
1778 return route_change;
1779 }
1780
1781 static void evpn_zebra_reinstall_best_route(struct bgp *bgp,
1782 struct bgpevpn *vpn, struct bgp_node *rn)
1783 {
1784 struct bgp_path_info *tmp_ri;
1785 struct bgp_path_info *curr_select = NULL;
1786
1787 for (tmp_ri = bgp_node_get_bgp_path_info(rn);
1788 tmp_ri; tmp_ri = tmp_ri->next) {
1789 if (CHECK_FLAG(tmp_ri->flags, BGP_PATH_SELECTED)) {
1790 curr_select = tmp_ri;
1791 break;
1792 }
1793 }
1794
1795 if (curr_select && curr_select->type == ZEBRA_ROUTE_BGP
1796 && curr_select->sub_type == BGP_ROUTE_IMPORTED)
1797 evpn_zebra_install(bgp, vpn,
1798 (struct prefix_evpn *)&rn->p,
1799 curr_select);
1800 }
1801
1802 /*
1803 * If the local route was not selected evict it and tell zebra to re-add
1804 * the best remote dest.
1805 *
1806 * Typically a local path added by zebra is expected to be selected as
1807 * best. In which case when a remote path wins as best (later)
1808 * evpn_route_select_install itself evicts the older-local-best path.
1809 *
1810 * However if bgp's add and zebra's add cross paths (race condition) it
1811 * is possible that the local path is no longer the "older" best path.
1812 * It is a path that was never designated as best and hence requires
1813 * additional handling to prevent bgp from injecting and holding on to a
1814 * non-best local path.
1815 */
1816 static void evpn_cleanup_local_non_best_route(struct bgp *bgp,
1817 struct bgpevpn *vpn,
1818 struct bgp_node *rn,
1819 struct bgp_path_info *local_pi)
1820 {
1821 char buf[PREFIX_STRLEN];
1822
1823 /* local path was not picked as the winner; kick it out */
1824 if (bgp_debug_zebra(NULL)) {
1825 zlog_debug("evicting local evpn prefix %s as remote won",
1826 prefix2str(&rn->p, buf, sizeof(buf)));
1827 }
1828 evpn_delete_old_local_route(bgp, vpn, rn, local_pi);
1829 bgp_path_info_reap(rn, local_pi);
1830
1831 /* tell zebra to re-add the best remote path */
1832 evpn_zebra_reinstall_best_route(bgp, vpn, rn);
1833 }
1834
1835 /*
1836 * Create or update EVPN route (of type based on prefix) for specified VNI
1837 * and schedule for processing.
1838 */
1839 static int update_evpn_route(struct bgp *bgp, struct bgpevpn *vpn,
1840 struct prefix_evpn *p, uint8_t flags,
1841 uint32_t seq)
1842 {
1843 struct bgp_node *rn;
1844 struct attr attr;
1845 struct attr *attr_new;
1846 int add_l3_ecomm = 0;
1847 struct bgp_path_info *pi;
1848 afi_t afi = AFI_L2VPN;
1849 safi_t safi = SAFI_EVPN;
1850 int route_change;
1851
1852 memset(&attr, 0, sizeof(struct attr));
1853
1854 /* Build path-attribute for this route. */
1855 bgp_attr_default_set(&attr, BGP_ORIGIN_IGP);
1856 attr.nexthop = vpn->originator_ip;
1857 attr.mp_nexthop_global_in = vpn->originator_ip;
1858 attr.mp_nexthop_len = BGP_ATTR_NHLEN_IPV4;
1859 attr.sticky = CHECK_FLAG(flags, ZEBRA_MACIP_TYPE_STICKY) ? 1 : 0;
1860 attr.default_gw = CHECK_FLAG(flags, ZEBRA_MACIP_TYPE_GW) ? 1 : 0;
1861 attr.router_flag = CHECK_FLAG(flags,
1862 ZEBRA_MACIP_TYPE_ROUTER_FLAG) ? 1 : 0;
1863
1864 /* PMSI is only needed for type-3 routes */
1865 if (p->prefix.route_type == BGP_EVPN_IMET_ROUTE) {
1866 attr.flag |= ATTR_FLAG_BIT(BGP_ATTR_PMSI_TUNNEL);
1867 attr.pmsi_tnl_type = PMSI_TNLTYPE_INGR_REPL;
1868 }
1869
1870 /* router mac is only needed for type-2 routes here. */
1871 if (p->prefix.route_type == BGP_EVPN_MAC_IP_ROUTE) {
1872 uint8_t af_flags = 0;
1873
1874 if (CHECK_FLAG(flags, ZEBRA_MACIP_TYPE_SVI_IP))
1875 SET_FLAG(af_flags, BGP_EVPN_MACIP_TYPE_SVI_IP);
1876
1877 bgp_evpn_get_rmac_nexthop(vpn, p, &attr, af_flags);
1878
1879 if (bgp_debug_zebra(NULL)) {
1880 char buf[ETHER_ADDR_STRLEN];
1881 char buf1[PREFIX_STRLEN];
1882
1883 zlog_debug("VRF %s vni %u type-2 route evp %s RMAC %s nexthop %s",
1884 vpn->bgp_vrf ?
1885 vrf_id_to_name(vpn->bgp_vrf->vrf_id) : " ",
1886 vpn->vni,
1887 prefix2str(p, buf1, sizeof(buf1)),
1888 prefix_mac2str(&attr.rmac, buf,
1889 sizeof(buf)),
1890 inet_ntoa(attr.mp_nexthop_global_in));
1891 }
1892 }
1893
1894 vni2label(vpn->vni, &(attr.label));
1895
1896 /* Include L3 VNI related RTs and RMAC for type-2 routes, if they're
1897 * IPv4 or IPv6 global addresses and we're advertising L3VNI with
1898 * these routes.
1899 */
1900 if (p->prefix.route_type == BGP_EVPN_MAC_IP_ROUTE &&
1901 (is_evpn_prefix_ipaddr_v4(p) ||
1902 !IN6_IS_ADDR_LINKLOCAL(&p->prefix.macip_addr.ip.ipaddr_v6)) &&
1903 CHECK_FLAG(vpn->flags, VNI_FLAG_USE_TWO_LABELS) &&
1904 bgpevpn_get_l3vni(vpn))
1905 add_l3_ecomm = 1;
1906
1907 /* Set up extended community. */
1908 build_evpn_route_extcomm(vpn, &attr, add_l3_ecomm);
1909
1910 /* First, create (or fetch) route node within the VNI. */
1911 /* NOTE: There is no RD here. */
1912 rn = bgp_node_get(vpn->route_table, (struct prefix *)p);
1913
1914 /* Create or update route entry. */
1915 route_change = update_evpn_route_entry(bgp, vpn, afi, safi, rn, &attr,
1916 1, &pi, flags, seq);
1917 assert(pi);
1918 attr_new = pi->attr;
1919
1920 /* lock ri to prevent freeing in evpn_route_select_install */
1921 bgp_path_info_lock(pi);
1922
1923 /* Perform route selection. Normally, the local route in the
1924 * VNI is expected to win and be the best route. However, if
1925 * there is a race condition where a host moved from local to
1926 * remote and the remote route was received in BGP just prior
1927 * to the local MACIP notification from zebra, the remote
1928 * route would win, and we should evict the defunct local route
1929 * and (re)install the remote route into zebra.
1930 */
1931 evpn_route_select_install(bgp, vpn, rn);
1932 /*
1933 * If the new local route was not selected evict it and tell zebra
1934 * to re-add the best remote dest. BGP doesn't retain non-best local
1935 * routes.
1936 */
1937 if (!CHECK_FLAG(pi->flags, BGP_PATH_SELECTED)) {
1938 route_change = 0;
1939 evpn_cleanup_local_non_best_route(bgp, vpn, rn, pi);
1940 }
1941 bgp_path_info_unlock(pi);
1942
1943 bgp_unlock_node(rn);
1944
1945 /* If this is a new route or some attribute has changed, export the
1946 * route to the global table. The route will be advertised to peers
1947 * from there. Note that this table is a 2-level tree (RD-level +
1948 * Prefix-level) similar to L3VPN routes.
1949 */
1950 if (route_change) {
1951 struct bgp_path_info *global_pi;
1952
1953 rn = bgp_afi_node_get(bgp->rib[afi][safi], afi, safi,
1954 (struct prefix *)p, &vpn->prd);
1955 update_evpn_route_entry(bgp, vpn, afi, safi, rn, attr_new, 1,
1956 &global_pi, flags, seq);
1957
1958 /* Schedule for processing and unlock node. */
1959 bgp_process(bgp, rn, afi, safi);
1960 bgp_unlock_node(rn);
1961 }
1962
1963 /* Unintern temporary. */
1964 aspath_unintern(&attr.aspath);
1965
1966 return 0;
1967 }
1968
1969 /*
1970 * Delete EVPN route entry.
1971 * The entry can be in ESI/VNI table or the global table.
1972 */
1973 static void delete_evpn_route_entry(struct bgp *bgp, afi_t afi, safi_t safi,
1974 struct bgp_node *rn,
1975 struct bgp_path_info **pi)
1976 {
1977 struct bgp_path_info *tmp_pi;
1978
1979 *pi = NULL;
1980
1981 /* Now, find matching route. */
1982 for (tmp_pi = bgp_node_get_bgp_path_info(rn); tmp_pi;
1983 tmp_pi = tmp_pi->next)
1984 if (tmp_pi->peer == bgp->peer_self
1985 && tmp_pi->type == ZEBRA_ROUTE_BGP
1986 && tmp_pi->sub_type == BGP_ROUTE_STATIC)
1987 break;
1988
1989 *pi = tmp_pi;
1990
1991 /* Mark route for delete. */
1992 if (tmp_pi)
1993 bgp_path_info_delete(rn, tmp_pi);
1994 }
1995
1996
1997
1998 /* Delete EVPN ES (type-4) route */
1999 static int delete_evpn_type4_route(struct bgp *bgp,
2000 struct evpnes *es,
2001 struct prefix_evpn *p)
2002 {
2003 afi_t afi = AFI_L2VPN;
2004 safi_t safi = SAFI_EVPN;
2005 struct bgp_path_info *pi;
2006 struct bgp_node *rn = NULL; /* rn in esi table */
2007 struct bgp_node *global_rn = NULL; /* rn in global table */
2008
2009 /* First, locate the route node within the ESI.
2010 * If it doesn't exist, ther is nothing to do.
2011 * Note: there is no RD here.
2012 */
2013 rn = bgp_node_lookup(es->route_table, (struct prefix *)p);
2014 if (!rn)
2015 return 0;
2016
2017 /* Next, locate route node in the global EVPN routing table.
2018 * Note that this table is a 2-level tree (RD-level + Prefix-level)
2019 */
2020 global_rn = bgp_afi_node_lookup(bgp->rib[afi][safi], afi, safi,
2021 (struct prefix *)p, &es->prd);
2022 if (global_rn) {
2023
2024 /* Delete route entry in the global EVPN table. */
2025 delete_evpn_route_entry(bgp, afi, safi, global_rn, &pi);
2026
2027 /* Schedule for processing - withdraws to peers happen from
2028 * this table.
2029 */
2030 if (pi)
2031 bgp_process(bgp, global_rn, afi, safi);
2032 bgp_unlock_node(global_rn);
2033 }
2034
2035 /*
2036 * Delete route entry in the ESI route table.
2037 * This can just be removed.
2038 */
2039 delete_evpn_route_entry(bgp, afi, safi, rn, &pi);
2040 if (pi)
2041 bgp_path_info_reap(rn, pi);
2042 bgp_unlock_node(rn);
2043 return 0;
2044 }
2045
2046 /* Delete EVPN type5 route */
2047 static int delete_evpn_type5_route(struct bgp *bgp_vrf, struct prefix_evpn *evp)
2048 {
2049 afi_t afi = AFI_L2VPN;
2050 safi_t safi = SAFI_EVPN;
2051 struct bgp_node *rn = NULL;
2052 struct bgp_path_info *pi = NULL;
2053 struct bgp *bgp_evpn = NULL; /* evpn bgp instance */
2054
2055 bgp_evpn = bgp_get_evpn();
2056 if (!bgp_evpn)
2057 return 0;
2058
2059 /* locate the global route entry for this type-5 prefix */
2060 rn = bgp_afi_node_lookup(bgp_evpn->rib[afi][safi], afi, safi,
2061 (struct prefix *)evp, &bgp_vrf->vrf_prd);
2062 if (!rn)
2063 return 0;
2064
2065 delete_evpn_route_entry(bgp_evpn, afi, safi, rn, &pi);
2066 if (pi)
2067 bgp_process(bgp_evpn, rn, afi, safi);
2068 bgp_unlock_node(rn);
2069 return 0;
2070 }
2071
2072 /*
2073 * Delete EVPN route (of type based on prefix) for specified VNI and
2074 * schedule for processing.
2075 */
2076 static int delete_evpn_route(struct bgp *bgp, struct bgpevpn *vpn,
2077 struct prefix_evpn *p)
2078 {
2079 struct bgp_node *rn, *global_rn;
2080 struct bgp_path_info *pi;
2081 afi_t afi = AFI_L2VPN;
2082 safi_t safi = SAFI_EVPN;
2083
2084 /* First, locate the route node within the VNI. If it doesn't exist,
2085 * there
2086 * is nothing further to do.
2087 */
2088 /* NOTE: There is no RD here. */
2089 rn = bgp_node_lookup(vpn->route_table, (struct prefix *)p);
2090 if (!rn)
2091 return 0;
2092
2093 /* Next, locate route node in the global EVPN routing table. Note that
2094 * this table is a 2-level tree (RD-level + Prefix-level) similar to
2095 * L3VPN routes.
2096 */
2097 global_rn = bgp_afi_node_lookup(bgp->rib[afi][safi], afi, safi,
2098 (struct prefix *)p, &vpn->prd);
2099 if (global_rn) {
2100 /* Delete route entry in the global EVPN table. */
2101 delete_evpn_route_entry(bgp, afi, safi, global_rn, &pi);
2102
2103 /* Schedule for processing - withdraws to peers happen from
2104 * this table.
2105 */
2106 if (pi)
2107 bgp_process(bgp, global_rn, afi, safi);
2108 bgp_unlock_node(global_rn);
2109 }
2110
2111 /* Delete route entry in the VNI route table. This can just be removed.
2112 */
2113 delete_evpn_route_entry(bgp, afi, safi, rn, &pi);
2114 if (pi) {
2115 bgp_path_info_reap(rn, pi);
2116 evpn_route_select_install(bgp, vpn, rn);
2117 }
2118 bgp_unlock_node(rn);
2119
2120 return 0;
2121 }
2122
2123 /*
2124 * Update all type-2 (MACIP) local routes for this VNI - these should also
2125 * be scheduled for advertise to peers.
2126 */
2127 static int update_all_type2_routes(struct bgp *bgp, struct bgpevpn *vpn)
2128 {
2129 afi_t afi;
2130 safi_t safi;
2131 struct bgp_node *rn;
2132 struct bgp_path_info *pi, *tmp_pi;
2133 struct attr attr;
2134 struct attr *attr_new;
2135 uint32_t seq;
2136 int add_l3_ecomm = 0;
2137
2138 afi = AFI_L2VPN;
2139 safi = SAFI_EVPN;
2140
2141 /* Walk this VNI's route table and update local type-2 routes. For any
2142 * routes updated, update corresponding entry in the global table too.
2143 */
2144 for (rn = bgp_table_top(vpn->route_table); rn;
2145 rn = bgp_route_next(rn)) {
2146 struct prefix_evpn *evp = (struct prefix_evpn *)&rn->p;
2147 struct bgp_node *rd_rn;
2148 struct bgp_path_info *global_pi;
2149
2150 if (evp->prefix.route_type != BGP_EVPN_MAC_IP_ROUTE)
2151 continue;
2152
2153 /* Identify local route. */
2154 for (tmp_pi = bgp_node_get_bgp_path_info(rn); tmp_pi;
2155 tmp_pi = tmp_pi->next) {
2156 if (tmp_pi->peer == bgp->peer_self
2157 && tmp_pi->type == ZEBRA_ROUTE_BGP
2158 && tmp_pi->sub_type == BGP_ROUTE_STATIC)
2159 break;
2160 }
2161
2162 if (!tmp_pi)
2163 continue;
2164
2165 /*
2166 * Build attribute per local route as the MAC mobility and
2167 * some other values could differ for different routes. The
2168 * attributes will be shared in the hash table.
2169 */
2170 bgp_attr_default_set(&attr, BGP_ORIGIN_IGP);
2171 attr.nexthop = vpn->originator_ip;
2172 attr.mp_nexthop_global_in = vpn->originator_ip;
2173 attr.mp_nexthop_len = BGP_ATTR_NHLEN_IPV4;
2174 bgp_evpn_get_rmac_nexthop(vpn, evp, &attr,
2175 tmp_pi->extra->af_flags);
2176
2177 if (evpn_route_is_sticky(bgp, rn))
2178 attr.sticky = 1;
2179 else if (evpn_route_is_def_gw(bgp, rn)) {
2180 attr.default_gw = 1;
2181 if (is_evpn_prefix_ipaddr_v6(evp))
2182 attr.router_flag = 1;
2183 }
2184
2185 if (bgp_debug_zebra(NULL)) {
2186 char buf[ETHER_ADDR_STRLEN];
2187 char buf1[PREFIX_STRLEN];
2188
2189 zlog_debug("VRF %s vni %u evp %s RMAC %s nexthop %s",
2190 vpn->bgp_vrf ?
2191 vrf_id_to_name(vpn->bgp_vrf->vrf_id) : " ",
2192 vpn->vni,
2193 prefix2str(evp, buf1, sizeof(buf1)),
2194 prefix_mac2str(&attr.rmac, buf, sizeof(buf)),
2195 inet_ntoa(attr.mp_nexthop_global_in));
2196 }
2197
2198 /* Add L3 VNI RTs and RMAC for non IPv6 link-local if
2199 * using L3 VNI for type-2 routes also.
2200 */
2201 if ((is_evpn_prefix_ipaddr_v4(evp) ||
2202 !IN6_IS_ADDR_LINKLOCAL(
2203 &evp->prefix.macip_addr.ip.ipaddr_v6)) &&
2204 CHECK_FLAG(vpn->flags, VNI_FLAG_USE_TWO_LABELS) &&
2205 bgpevpn_get_l3vni(vpn))
2206 add_l3_ecomm = 1;
2207
2208 /* Set up extended community. */
2209 build_evpn_route_extcomm(vpn, &attr, add_l3_ecomm);
2210
2211 seq = mac_mobility_seqnum(tmp_pi->attr);
2212
2213 /* Update the route entry. */
2214 update_evpn_route_entry(bgp, vpn, afi, safi, rn, &attr, 0, &pi,
2215 0, seq);
2216
2217 /* lock ri to prevent freeing in evpn_route_select_install */
2218 bgp_path_info_lock(pi);
2219
2220 /* Perform route selection. Normally, the local route in the
2221 * VNI is expected to win and be the best route. However,
2222 * under peculiar situations (e.g., tunnel (next hop) IP change
2223 * that causes best selection to be based on next hop), a
2224 * remote route could win. If the local route is the best,
2225 * ensure it is updated in the global EVPN route table and
2226 * advertised to peers; otherwise, ensure it is evicted and
2227 * (re)install the remote route into zebra.
2228 */
2229 evpn_route_select_install(bgp, vpn, rn);
2230 if (!CHECK_FLAG(pi->flags, BGP_PATH_SELECTED)) {
2231 evpn_cleanup_local_non_best_route(bgp, vpn, rn, pi);
2232 /* unlock pi */
2233 bgp_path_info_unlock(pi);
2234 } else {
2235 attr_new = pi->attr;
2236 /* unlock pi */
2237 bgp_path_info_unlock(pi);
2238
2239 /* Update route in global routing table. */
2240 rd_rn = bgp_afi_node_get(bgp->rib[afi][safi], afi, safi,
2241 (struct prefix *)evp, &vpn->prd);
2242 assert(rd_rn);
2243 update_evpn_route_entry(bgp, vpn, afi, safi, rd_rn,
2244 attr_new, 0, &global_pi, 0,
2245 mac_mobility_seqnum(attr_new));
2246
2247 /* Schedule for processing and unlock node. */
2248 bgp_process(bgp, rd_rn, afi, safi);
2249 bgp_unlock_node(rd_rn);
2250 }
2251
2252 /* Unintern temporary. */
2253 aspath_unintern(&attr.aspath);
2254
2255 }
2256
2257 return 0;
2258 }
2259
2260 /*
2261 * Delete all type-2 (MACIP) local routes for this VNI - only from the
2262 * global routing table. These are also scheduled for withdraw from peers.
2263 */
2264 static int delete_global_type2_routes(struct bgp *bgp, struct bgpevpn *vpn)
2265 {
2266 afi_t afi;
2267 safi_t safi;
2268 struct bgp_node *rdrn, *rn;
2269 struct bgp_table *table;
2270 struct bgp_path_info *pi;
2271
2272 afi = AFI_L2VPN;
2273 safi = SAFI_EVPN;
2274
2275 rdrn = bgp_node_lookup(bgp->rib[afi][safi], (struct prefix *)&vpn->prd);
2276 if (rdrn && bgp_node_has_bgp_path_info_data(rdrn)) {
2277 table = bgp_node_get_bgp_table_info(rdrn);
2278 for (rn = bgp_table_top(table); rn; rn = bgp_route_next(rn)) {
2279 struct prefix_evpn *evp = (struct prefix_evpn *)&rn->p;
2280
2281 if (evp->prefix.route_type != BGP_EVPN_MAC_IP_ROUTE)
2282 continue;
2283
2284 delete_evpn_route_entry(bgp, afi, safi, rn, &pi);
2285 if (pi)
2286 bgp_process(bgp, rn, afi, safi);
2287 }
2288 }
2289
2290 /* Unlock RD node. */
2291 if (rdrn)
2292 bgp_unlock_node(rdrn);
2293
2294 return 0;
2295 }
2296
2297 /*
2298 * Delete all type-2 (MACIP) local routes for this VNI - from the global
2299 * table as well as the per-VNI route table.
2300 */
2301 static int delete_all_type2_routes(struct bgp *bgp, struct bgpevpn *vpn)
2302 {
2303 afi_t afi;
2304 safi_t safi;
2305 struct bgp_node *rn;
2306 struct bgp_path_info *pi;
2307
2308 afi = AFI_L2VPN;
2309 safi = SAFI_EVPN;
2310
2311 /* First, walk the global route table for this VNI's type-2 local
2312 * routes.
2313 * EVPN routes are a 2-level table, first get the RD table.
2314 */
2315 delete_global_type2_routes(bgp, vpn);
2316
2317 /* Next, walk this VNI's route table and delete local type-2 routes. */
2318 for (rn = bgp_table_top(vpn->route_table); rn;
2319 rn = bgp_route_next(rn)) {
2320 struct prefix_evpn *evp = (struct prefix_evpn *)&rn->p;
2321
2322 if (evp->prefix.route_type != BGP_EVPN_MAC_IP_ROUTE)
2323 continue;
2324
2325 delete_evpn_route_entry(bgp, afi, safi, rn, &pi);
2326
2327 /* Route entry in local table gets deleted immediately. */
2328 if (pi)
2329 bgp_path_info_reap(rn, pi);
2330 }
2331
2332 return 0;
2333 }
2334
2335 /*
2336 * Delete all routes in per ES route-table
2337 */
2338 static int delete_all_es_routes(struct bgp *bgp, struct evpnes *es)
2339 {
2340 struct bgp_node *rn;
2341 struct bgp_path_info *pi, *nextpi;
2342
2343 /* Walk this ES's route table and delete all routes. */
2344 for (rn = bgp_table_top(es->route_table); rn;
2345 rn = bgp_route_next(rn)) {
2346 for (pi = bgp_node_get_bgp_path_info(rn);
2347 (pi != NULL) && (nextpi = pi->next, 1); pi = nextpi) {
2348 bgp_path_info_delete(rn, pi);
2349 bgp_path_info_reap(rn, pi);
2350 }
2351 }
2352
2353 return 0;
2354 }
2355
2356 /*
2357 * Delete all routes in the per-VNI route table.
2358 */
2359 static int delete_all_vni_routes(struct bgp *bgp, struct bgpevpn *vpn)
2360 {
2361 struct bgp_node *rn;
2362 struct bgp_path_info *pi, *nextpi;
2363
2364 /* Walk this VNI's route table and delete all routes. */
2365 for (rn = bgp_table_top(vpn->route_table); rn;
2366 rn = bgp_route_next(rn)) {
2367 for (pi = bgp_node_get_bgp_path_info(rn);
2368 (pi != NULL) && (nextpi = pi->next, 1); pi = nextpi) {
2369 bgp_path_info_delete(rn, pi);
2370 bgp_path_info_reap(rn, pi);
2371 }
2372 }
2373
2374 return 0;
2375 }
2376
2377 /* BUM traffic flood mode per-l2-vni */
2378 static int bgp_evpn_vni_flood_mode_get(struct bgp *bgp,
2379 struct bgpevpn *vpn)
2380 {
2381 /* if flooding has been globally disabled per-vni mode is
2382 * not relevant
2383 */
2384 if (bgp->vxlan_flood_ctrl == VXLAN_FLOOD_DISABLED)
2385 return VXLAN_FLOOD_DISABLED;
2386
2387 /* if mcast group ip has been specified we use a PIM-SM MDT */
2388 if (vpn->mcast_grp.s_addr != INADDR_ANY)
2389 return VXLAN_FLOOD_PIM_SM;
2390
2391 /* default is ingress replication */
2392 return VXLAN_FLOOD_HEAD_END_REPL;
2393 }
2394
2395 /*
2396 * Update (and advertise) local routes for a VNI. Invoked upon the VNI
2397 * export RT getting modified or change to tunnel IP. Note that these
2398 * situations need the route in the per-VNI table as well as the global
2399 * table to be updated (as attributes change).
2400 */
2401 int update_routes_for_vni(struct bgp *bgp, struct bgpevpn *vpn)
2402 {
2403 int ret;
2404 struct prefix_evpn p;
2405
2406 /* Update and advertise the type-3 route (only one) followed by the
2407 * locally learnt type-2 routes (MACIP) - for this VNI.
2408 *
2409 * RT-3 only if doing head-end replication
2410 */
2411 if (bgp_evpn_vni_flood_mode_get(bgp, vpn)
2412 == VXLAN_FLOOD_HEAD_END_REPL) {
2413 build_evpn_type3_prefix(&p, vpn->originator_ip);
2414 ret = update_evpn_route(bgp, vpn, &p, 0, 0);
2415 if (ret)
2416 return ret;
2417 }
2418
2419 return update_all_type2_routes(bgp, vpn);
2420 }
2421
2422 /* Delete (and withdraw) local routes for specified ES from global and ES table.
2423 * Also remove all other routes from the per ES table.
2424 * Invoked when ES is deleted.
2425 */
2426 static int delete_routes_for_es(struct bgp *bgp, struct evpnes *es)
2427 {
2428 int ret;
2429 char buf[ESI_STR_LEN];
2430 struct prefix_evpn p;
2431
2432 /* Delete and withdraw locally learnt ES route */
2433 build_evpn_type4_prefix(&p, &es->esi, es->originator_ip.ipaddr_v4);
2434 ret = delete_evpn_type4_route(bgp, es, &p);
2435 if (ret) {
2436 flog_err(EC_BGP_EVPN_ROUTE_DELETE,
2437 "%u failed to delete type-4 route for ESI %s",
2438 bgp->vrf_id, esi_to_str(&es->esi, buf, sizeof(buf)));
2439 }
2440
2441 /* Delete all routes from per ES table */
2442 return delete_all_es_routes(bgp, es);
2443 }
2444
2445 /*
2446 * Delete (and withdraw) local routes for specified VNI from the global
2447 * table and per-VNI table. After this, remove all other routes from
2448 * the per-VNI table. Invoked upon the VNI being deleted or EVPN
2449 * (advertise-all-vni) being disabled.
2450 */
2451 static int delete_routes_for_vni(struct bgp *bgp, struct bgpevpn *vpn)
2452 {
2453 int ret;
2454 struct prefix_evpn p;
2455
2456 /* Delete and withdraw locally learnt type-2 routes (MACIP)
2457 * followed by type-3 routes (only one) - for this VNI.
2458 */
2459 ret = delete_all_type2_routes(bgp, vpn);
2460 if (ret)
2461 return ret;
2462
2463 build_evpn_type3_prefix(&p, vpn->originator_ip);
2464 ret = delete_evpn_route(bgp, vpn, &p);
2465 if (ret)
2466 return ret;
2467
2468 /* Delete all routes from the per-VNI table. */
2469 return delete_all_vni_routes(bgp, vpn);
2470 }
2471
2472 /*
2473 * There is a flood mcast IP address change. Update the mcast-grp and
2474 * remove the type-3 route if any. A new type-3 route will be generated
2475 * post tunnel_ip update if the new flood mode is head-end-replication.
2476 */
2477 static int bgp_evpn_mcast_grp_change(struct bgp *bgp, struct bgpevpn *vpn,
2478 struct in_addr mcast_grp)
2479 {
2480 struct prefix_evpn p;
2481
2482 vpn->mcast_grp = mcast_grp;
2483
2484 if (is_vni_live(vpn)) {
2485 build_evpn_type3_prefix(&p, vpn->originator_ip);
2486 delete_evpn_route(bgp, vpn, &p);
2487 }
2488
2489 return 0;
2490 }
2491
2492 /*
2493 * There is a tunnel endpoint IP address change for this VNI, delete
2494 * prior type-3 route (if needed) and update.
2495 * Note: Route re-advertisement happens elsewhere after other processing
2496 * other changes.
2497 */
2498 static int handle_tunnel_ip_change(struct bgp *bgp, struct bgpevpn *vpn,
2499 struct in_addr originator_ip)
2500 {
2501 struct prefix_evpn p;
2502
2503 /* If VNI is not live, we only need to update the originator ip */
2504 if (!is_vni_live(vpn)) {
2505 vpn->originator_ip = originator_ip;
2506 return 0;
2507 }
2508
2509 /* Update the tunnel-ip hash */
2510 bgp_tip_del(bgp, &vpn->originator_ip);
2511 bgp_tip_add(bgp, &originator_ip);
2512
2513 /* filter routes as martian nexthop db has changed */
2514 bgp_filter_evpn_routes_upon_martian_nh_change(bgp);
2515
2516 /* Need to withdraw type-3 route as the originator IP is part
2517 * of the key.
2518 */
2519 build_evpn_type3_prefix(&p, vpn->originator_ip);
2520 delete_evpn_route(bgp, vpn, &p);
2521
2522 /* Update the tunnel IP and re-advertise all routes for this VNI. */
2523 vpn->originator_ip = originator_ip;
2524 return 0;
2525 }
2526
2527 static struct bgp_path_info *
2528 bgp_create_evpn_bgp_path_info(struct bgp_path_info *parent_pi,
2529 struct bgp_node *rn)
2530 {
2531 struct attr *attr_new;
2532 struct bgp_path_info *pi;
2533
2534 /* Add (or update) attribute to hash. */
2535 attr_new = bgp_attr_intern(parent_pi->attr);
2536
2537 /* Create new route with its attribute. */
2538 pi = info_make(parent_pi->type, BGP_ROUTE_IMPORTED, 0, parent_pi->peer,
2539 attr_new, rn);
2540 SET_FLAG(pi->flags, BGP_PATH_VALID);
2541 bgp_path_info_extra_get(pi);
2542 pi->extra->parent = bgp_path_info_lock(parent_pi);
2543 bgp_lock_node((struct bgp_node *)parent_pi->net);
2544 if (parent_pi->extra) {
2545 memcpy(&pi->extra->label, &parent_pi->extra->label,
2546 sizeof(pi->extra->label));
2547 pi->extra->num_labels = parent_pi->extra->num_labels;
2548 }
2549 bgp_path_info_add(rn, pi);
2550
2551 return pi;
2552 }
2553
2554 /* Install EVPN route entry in ES */
2555 static int install_evpn_route_entry_in_es(struct bgp *bgp, struct evpnes *es,
2556 struct prefix_evpn *p,
2557 struct bgp_path_info *parent_pi)
2558 {
2559 int ret = 0;
2560 struct bgp_node *rn = NULL;
2561 struct bgp_path_info *pi = NULL;
2562 struct attr *attr_new = NULL;
2563
2564 /* Create (or fetch) route within the VNI.
2565 * NOTE: There is no RD here.
2566 */
2567 rn = bgp_node_get(es->route_table, (struct prefix *)p);
2568
2569 /* Check if route entry is already present. */
2570 for (pi = bgp_node_get_bgp_path_info(rn); pi; pi = pi->next)
2571 if (pi->extra
2572 && (struct bgp_path_info *)pi->extra->parent == parent_pi)
2573 break;
2574
2575 if (!pi) {
2576 /* Add (or update) attribute to hash. */
2577 attr_new = bgp_attr_intern(parent_pi->attr);
2578
2579 /* Create new route with its attribute. */
2580 pi = info_make(parent_pi->type, BGP_ROUTE_IMPORTED, 0,
2581 parent_pi->peer, attr_new, rn);
2582 SET_FLAG(pi->flags, BGP_PATH_VALID);
2583 bgp_path_info_extra_get(pi);
2584 pi->extra->parent = bgp_path_info_lock(parent_pi);
2585 bgp_lock_node((struct bgp_node *)parent_pi->net);
2586 bgp_path_info_add(rn, pi);
2587 } else {
2588 if (attrhash_cmp(pi->attr, parent_pi->attr)
2589 && !CHECK_FLAG(pi->flags, BGP_PATH_REMOVED)) {
2590 bgp_unlock_node(rn);
2591 return 0;
2592 }
2593 /* The attribute has changed. */
2594 /* Add (or update) attribute to hash. */
2595 attr_new = bgp_attr_intern(parent_pi->attr);
2596
2597 /* Restore route, if needed. */
2598 if (CHECK_FLAG(pi->flags, BGP_PATH_REMOVED))
2599 bgp_path_info_restore(rn, pi);
2600
2601 /* Mark if nexthop has changed. */
2602 if (!IPV4_ADDR_SAME(&pi->attr->nexthop, &attr_new->nexthop))
2603 SET_FLAG(pi->flags, BGP_PATH_IGP_CHANGED);
2604
2605 /* Unintern existing, set to new. */
2606 bgp_attr_unintern(&pi->attr);
2607 pi->attr = attr_new;
2608 pi->uptime = bgp_clock();
2609 }
2610
2611 /* Perform route selection and update zebra, if required. */
2612 ret = evpn_es_route_select_install(bgp, es, rn);
2613 return ret;
2614 }
2615
2616 /*
2617 * Install route entry into the VRF routing table and invoke route selection.
2618 */
2619 static int install_evpn_route_entry_in_vrf(struct bgp *bgp_vrf,
2620 struct prefix_evpn *evp,
2621 struct bgp_path_info *parent_pi)
2622 {
2623 struct bgp_node *rn;
2624 struct bgp_path_info *pi;
2625 struct attr attr;
2626 struct attr *attr_new;
2627 int ret = 0;
2628 struct prefix p;
2629 struct prefix *pp = &p;
2630 afi_t afi = 0;
2631 safi_t safi = 0;
2632 char buf[PREFIX_STRLEN];
2633 char buf1[PREFIX_STRLEN];
2634
2635 memset(pp, 0, sizeof(struct prefix));
2636 ip_prefix_from_evpn_prefix(evp, pp);
2637
2638 if (bgp_debug_zebra(NULL)) {
2639 zlog_debug(
2640 "import evpn prefix %s as ip prefix %s in vrf %s",
2641 prefix2str(evp, buf, sizeof(buf)),
2642 prefix2str(pp, buf1, sizeof(buf)),
2643 vrf_id_to_name(bgp_vrf->vrf_id));
2644 }
2645
2646 /* Create (or fetch) route within the VRF. */
2647 /* NOTE: There is no RD here. */
2648 if (is_evpn_prefix_ipaddr_v4(evp)) {
2649 afi = AFI_IP;
2650 safi = SAFI_UNICAST;
2651 rn = bgp_node_get(bgp_vrf->rib[afi][safi], pp);
2652 } else if (is_evpn_prefix_ipaddr_v6(evp)) {
2653 afi = AFI_IP6;
2654 safi = SAFI_UNICAST;
2655 rn = bgp_node_get(bgp_vrf->rib[afi][safi], pp);
2656 } else
2657 return 0;
2658
2659 /* EVPN routes currently only support a IPv4 next hop which corresponds
2660 * to the remote VTEP. When importing into a VRF, if it is IPv6 host
2661 * or prefix route, we have to convert the next hop to an IPv4-mapped
2662 * address for the rest of the code to flow through. In the case of IPv4,
2663 * make sure to set the flag for next hop attribute.
2664 */
2665 attr = *parent_pi->attr;
2666 if (afi == AFI_IP6)
2667 evpn_convert_nexthop_to_ipv6(&attr);
2668 else
2669 attr.flag |= ATTR_FLAG_BIT(BGP_ATTR_NEXT_HOP);
2670
2671 /* Check if route entry is already present. */
2672 for (pi = bgp_node_get_bgp_path_info(rn); pi; pi = pi->next)
2673 if (pi->extra
2674 && (struct bgp_path_info *)pi->extra->parent == parent_pi)
2675 break;
2676
2677 if (!pi)
2678 pi = bgp_create_evpn_bgp_path_info(parent_pi, rn);
2679 else {
2680 if (attrhash_cmp(pi->attr, &attr)
2681 && !CHECK_FLAG(pi->flags, BGP_PATH_REMOVED)) {
2682 bgp_unlock_node(rn);
2683 return 0;
2684 }
2685 /* The attribute has changed. */
2686 /* Add (or update) attribute to hash. */
2687 attr_new = bgp_attr_intern(&attr);
2688
2689 /* Restore route, if needed. */
2690 if (CHECK_FLAG(pi->flags, BGP_PATH_REMOVED))
2691 bgp_path_info_restore(rn, pi);
2692
2693 /* Mark if nexthop has changed. */
2694 if ((afi == AFI_IP
2695 && !IPV4_ADDR_SAME(&pi->attr->nexthop, &attr_new->nexthop))
2696 || (afi == AFI_IP6
2697 && !IPV6_ADDR_SAME(&pi->attr->mp_nexthop_global,
2698 &attr_new->mp_nexthop_global)))
2699 SET_FLAG(pi->flags, BGP_PATH_IGP_CHANGED);
2700
2701 bgp_path_info_set_flag(rn, pi, BGP_PATH_ATTR_CHANGED);
2702 /* Unintern existing, set to new. */
2703 bgp_attr_unintern(&pi->attr);
2704 pi->attr = attr_new;
2705 pi->uptime = bgp_clock();
2706 }
2707
2708 bgp_aggregate_increment(bgp_vrf, &rn->p, pi, afi, safi);
2709
2710 /* Perform route selection and update zebra, if required. */
2711 bgp_process(bgp_vrf, rn, afi, safi);
2712
2713 /* Process for route leaking. */
2714 vpn_leak_from_vrf_update(bgp_get_default(), bgp_vrf, pi);
2715
2716 bgp_unlock_node(rn);
2717
2718 return ret;
2719 }
2720
2721 /*
2722 * Install route entry into the VNI routing table and invoke route selection.
2723 */
2724 static int install_evpn_route_entry(struct bgp *bgp, struct bgpevpn *vpn,
2725 struct prefix_evpn *p,
2726 struct bgp_path_info *parent_pi)
2727 {
2728 struct bgp_node *rn;
2729 struct bgp_path_info *pi;
2730 struct attr *attr_new;
2731 int ret;
2732
2733 /* Create (or fetch) route within the VNI. */
2734 /* NOTE: There is no RD here. */
2735 rn = bgp_node_get(vpn->route_table, (struct prefix *)p);
2736
2737 /* Check if route entry is already present. */
2738 for (pi = bgp_node_get_bgp_path_info(rn); pi; pi = pi->next)
2739 if (pi->extra
2740 && (struct bgp_path_info *)pi->extra->parent == parent_pi)
2741 break;
2742
2743 if (!pi) {
2744 /* Create an info */
2745 (void)bgp_create_evpn_bgp_path_info(parent_pi, rn);
2746 } else {
2747 if (attrhash_cmp(pi->attr, parent_pi->attr)
2748 && !CHECK_FLAG(pi->flags, BGP_PATH_REMOVED)) {
2749 bgp_unlock_node(rn);
2750 return 0;
2751 }
2752 /* The attribute has changed. */
2753 /* Add (or update) attribute to hash. */
2754 attr_new = bgp_attr_intern(parent_pi->attr);
2755
2756 /* Restore route, if needed. */
2757 if (CHECK_FLAG(pi->flags, BGP_PATH_REMOVED))
2758 bgp_path_info_restore(rn, pi);
2759
2760 /* Mark if nexthop has changed. */
2761 if (!IPV4_ADDR_SAME(&pi->attr->nexthop, &attr_new->nexthop))
2762 SET_FLAG(pi->flags, BGP_PATH_IGP_CHANGED);
2763
2764 /* Unintern existing, set to new. */
2765 bgp_attr_unintern(&pi->attr);
2766 pi->attr = attr_new;
2767 pi->uptime = bgp_clock();
2768 }
2769
2770 /* Perform route selection and update zebra, if required. */
2771 ret = evpn_route_select_install(bgp, vpn, rn);
2772
2773 bgp_unlock_node(rn);
2774
2775 return ret;
2776 }
2777
2778 /* Uninstall EVPN route entry from ES route table */
2779 static int uninstall_evpn_route_entry_in_es(struct bgp *bgp, struct evpnes *es,
2780 struct prefix_evpn *p,
2781 struct bgp_path_info *parent_pi)
2782 {
2783 int ret;
2784 struct bgp_node *rn;
2785 struct bgp_path_info *pi;
2786
2787 if (!es->route_table)
2788 return 0;
2789
2790 /* Locate route within the ESI.
2791 * NOTE: There is no RD here.
2792 */
2793 rn = bgp_node_lookup(es->route_table, (struct prefix *)p);
2794 if (!rn)
2795 return 0;
2796
2797 /* Find matching route entry. */
2798 for (pi = bgp_node_get_bgp_path_info(rn); pi; pi = pi->next)
2799 if (pi->extra
2800 && (struct bgp_path_info *)pi->extra->parent == parent_pi)
2801 break;
2802
2803 if (!pi)
2804 return 0;
2805
2806 /* Mark entry for deletion */
2807 bgp_path_info_delete(rn, pi);
2808
2809 /* Perform route selection and update zebra, if required. */
2810 ret = evpn_es_route_select_install(bgp, es, rn);
2811
2812 /* Unlock route node. */
2813 bgp_unlock_node(rn);
2814
2815 return ret;
2816 }
2817
2818 /*
2819 * Uninstall route entry from the VRF routing table and send message
2820 * to zebra, if appropriate.
2821 */
2822 static int uninstall_evpn_route_entry_in_vrf(struct bgp *bgp_vrf,
2823 struct prefix_evpn *evp,
2824 struct bgp_path_info *parent_pi)
2825 {
2826 struct bgp_node *rn;
2827 struct bgp_path_info *pi;
2828 int ret = 0;
2829 struct prefix p;
2830 struct prefix *pp = &p;
2831 afi_t afi = 0;
2832 safi_t safi = 0;
2833 char buf[PREFIX_STRLEN];
2834 char buf1[PREFIX_STRLEN];
2835
2836 memset(pp, 0, sizeof(struct prefix));
2837 ip_prefix_from_evpn_prefix(evp, pp);
2838
2839 if (bgp_debug_zebra(NULL)) {
2840 zlog_debug(
2841 "uninstalling evpn prefix %s as ip prefix %s in vrf %s",
2842 prefix2str(evp, buf, sizeof(buf)),
2843 prefix2str(pp, buf1, sizeof(buf)),
2844 vrf_id_to_name(bgp_vrf->vrf_id));
2845 }
2846
2847 /* Locate route within the VRF. */
2848 /* NOTE: There is no RD here. */
2849 if (is_evpn_prefix_ipaddr_v4(evp)) {
2850 afi = AFI_IP;
2851 safi = SAFI_UNICAST;
2852 rn = bgp_node_lookup(bgp_vrf->rib[afi][safi], pp);
2853 } else {
2854 afi = AFI_IP6;
2855 safi = SAFI_UNICAST;
2856 rn = bgp_node_lookup(bgp_vrf->rib[afi][safi], pp);
2857 }
2858
2859 if (!rn)
2860 return 0;
2861
2862 /* Find matching route entry. */
2863 for (pi = bgp_node_get_bgp_path_info(rn); pi; pi = pi->next)
2864 if (pi->extra
2865 && (struct bgp_path_info *)pi->extra->parent == parent_pi)
2866 break;
2867
2868 if (!pi)
2869 return 0;
2870
2871 /* Process for route leaking. */
2872 vpn_leak_from_vrf_withdraw(bgp_get_default(), bgp_vrf, pi);
2873
2874 bgp_aggregate_decrement(bgp_vrf, &rn->p, pi, afi, safi);
2875
2876 /* Mark entry for deletion */
2877 bgp_path_info_delete(rn, pi);
2878
2879 /* Perform route selection and update zebra, if required. */
2880 bgp_process(bgp_vrf, rn, afi, safi);
2881
2882 /* Unlock route node. */
2883 bgp_unlock_node(rn);
2884
2885 return ret;
2886 }
2887
2888 /*
2889 * Uninstall route entry from the VNI routing table and send message
2890 * to zebra, if appropriate.
2891 */
2892 static int uninstall_evpn_route_entry(struct bgp *bgp, struct bgpevpn *vpn,
2893 struct prefix_evpn *p,
2894 struct bgp_path_info *parent_pi)
2895 {
2896 struct bgp_node *rn;
2897 struct bgp_path_info *pi;
2898 int ret;
2899
2900 /* Locate route within the VNI. */
2901 /* NOTE: There is no RD here. */
2902 rn = bgp_node_lookup(vpn->route_table, (struct prefix *)p);
2903 if (!rn)
2904 return 0;
2905
2906 /* Find matching route entry. */
2907 for (pi = bgp_node_get_bgp_path_info(rn); pi; pi = pi->next)
2908 if (pi->extra
2909 && (struct bgp_path_info *)pi->extra->parent == parent_pi)
2910 break;
2911
2912 if (!pi)
2913 return 0;
2914
2915 /* Mark entry for deletion */
2916 bgp_path_info_delete(rn, pi);
2917
2918 /* Perform route selection and update zebra, if required. */
2919 ret = evpn_route_select_install(bgp, vpn, rn);
2920
2921 /* Unlock route node. */
2922 bgp_unlock_node(rn);
2923
2924 return ret;
2925 }
2926
2927 /*
2928 * Given a prefix, see if it belongs to ES.
2929 */
2930 static int is_prefix_matching_for_es(struct prefix_evpn *p,
2931 struct evpnes *es)
2932 {
2933 /* if not an ES route return false */
2934 if (p->prefix.route_type != BGP_EVPN_ES_ROUTE)
2935 return 0;
2936
2937 if (memcmp(&p->prefix.es_addr.esi, &es->esi, sizeof(esi_t)) == 0)
2938 return 1;
2939
2940 return 0;
2941 }
2942
2943 /*
2944 * Given a route entry and a VRF, see if this route entry should be
2945 * imported into the VRF i.e., RTs match.
2946 */
2947 static int is_route_matching_for_vrf(struct bgp *bgp_vrf,
2948 struct bgp_path_info *pi)
2949 {
2950 struct attr *attr = pi->attr;
2951 struct ecommunity *ecom;
2952 int i;
2953
2954 assert(attr);
2955 /* Route should have valid RT to be even considered. */
2956 if (!(attr->flag & ATTR_FLAG_BIT(BGP_ATTR_EXT_COMMUNITIES)))
2957 return 0;
2958
2959 ecom = attr->ecommunity;
2960 if (!ecom || !ecom->size)
2961 return 0;
2962
2963 /* For each extended community RT, see if it matches this VNI. If any RT
2964 * matches, we're done.
2965 */
2966 for (i = 0; i < ecom->size; i++) {
2967 uint8_t *pnt;
2968 uint8_t type, sub_type;
2969 struct ecommunity_val *eval;
2970 struct ecommunity_val eval_tmp;
2971 struct vrf_irt_node *irt;
2972
2973 /* Only deal with RTs */
2974 pnt = (ecom->val + (i * ECOMMUNITY_SIZE));
2975 eval = (struct ecommunity_val *)(ecom->val
2976 + (i * ECOMMUNITY_SIZE));
2977 type = *pnt++;
2978 sub_type = *pnt++;
2979 if (sub_type != ECOMMUNITY_ROUTE_TARGET)
2980 continue;
2981
2982 /* See if this RT matches specified VNIs import RTs */
2983 irt = lookup_vrf_import_rt(eval);
2984 if (irt)
2985 if (is_vrf_present_in_irt_vrfs(irt->vrfs, bgp_vrf))
2986 return 1;
2987
2988 /* Also check for non-exact match. In this, we mask out the AS
2989 * and
2990 * only check on the local-admin sub-field. This is to
2991 * facilitate using
2992 * VNI as the RT for EBGP peering too.
2993 */
2994 irt = NULL;
2995 if (type == ECOMMUNITY_ENCODE_AS
2996 || type == ECOMMUNITY_ENCODE_AS4
2997 || type == ECOMMUNITY_ENCODE_IP) {
2998 memcpy(&eval_tmp, eval, ECOMMUNITY_SIZE);
2999 mask_ecom_global_admin(&eval_tmp, eval);
3000 irt = lookup_vrf_import_rt(&eval_tmp);
3001 }
3002 if (irt)
3003 if (is_vrf_present_in_irt_vrfs(irt->vrfs, bgp_vrf))
3004 return 1;
3005 }
3006
3007 return 0;
3008 }
3009
3010 /*
3011 * Given a route entry and a VNI, see if this route entry should be
3012 * imported into the VNI i.e., RTs match.
3013 */
3014 static int is_route_matching_for_vni(struct bgp *bgp, struct bgpevpn *vpn,
3015 struct bgp_path_info *pi)
3016 {
3017 struct attr *attr = pi->attr;
3018 struct ecommunity *ecom;
3019 int i;
3020
3021 assert(attr);
3022 /* Route should have valid RT to be even considered. */
3023 if (!(attr->flag & ATTR_FLAG_BIT(BGP_ATTR_EXT_COMMUNITIES)))
3024 return 0;
3025
3026 ecom = attr->ecommunity;
3027 if (!ecom || !ecom->size)
3028 return 0;
3029
3030 /* For each extended community RT, see if it matches this VNI. If any RT
3031 * matches, we're done.
3032 */
3033 for (i = 0; i < ecom->size; i++) {
3034 uint8_t *pnt;
3035 uint8_t type, sub_type;
3036 struct ecommunity_val *eval;
3037 struct ecommunity_val eval_tmp;
3038 struct irt_node *irt;
3039
3040 /* Only deal with RTs */
3041 pnt = (ecom->val + (i * ECOMMUNITY_SIZE));
3042 eval = (struct ecommunity_val *)(ecom->val
3043 + (i * ECOMMUNITY_SIZE));
3044 type = *pnt++;
3045 sub_type = *pnt++;
3046 if (sub_type != ECOMMUNITY_ROUTE_TARGET)
3047 continue;
3048
3049 /* See if this RT matches specified VNIs import RTs */
3050 irt = lookup_import_rt(bgp, eval);
3051 if (irt)
3052 if (is_vni_present_in_irt_vnis(irt->vnis, vpn))
3053 return 1;
3054
3055 /* Also check for non-exact match. In this, we mask out the AS
3056 * and
3057 * only check on the local-admin sub-field. This is to
3058 * facilitate using
3059 * VNI as the RT for EBGP peering too.
3060 */
3061 irt = NULL;
3062 if (type == ECOMMUNITY_ENCODE_AS
3063 || type == ECOMMUNITY_ENCODE_AS4
3064 || type == ECOMMUNITY_ENCODE_IP) {
3065 memcpy(&eval_tmp, eval, ECOMMUNITY_SIZE);
3066 mask_ecom_global_admin(&eval_tmp, eval);
3067 irt = lookup_import_rt(bgp, &eval_tmp);
3068 }
3069 if (irt)
3070 if (is_vni_present_in_irt_vnis(irt->vnis, vpn))
3071 return 1;
3072 }
3073
3074 return 0;
3075 }
3076
3077 static int install_uninstall_routes_for_es(struct bgp *bgp,
3078 struct evpnes *es,
3079 int install)
3080 {
3081 int ret;
3082 afi_t afi;
3083 safi_t safi;
3084 char buf[PREFIX_STRLEN];
3085 char buf1[ESI_STR_LEN];
3086 struct bgp_node *rd_rn, *rn;
3087 struct bgp_table *table;
3088 struct bgp_path_info *pi;
3089
3090 afi = AFI_L2VPN;
3091 safi = SAFI_EVPN;
3092
3093 /*
3094 * Walk entire global routing table and evaluate routes which could be
3095 * imported into this VRF. Note that we need to loop through all global
3096 * routes to determine which route matches the import rt on vrf
3097 */
3098 for (rd_rn = bgp_table_top(bgp->rib[afi][safi]); rd_rn;
3099 rd_rn = bgp_route_next(rd_rn)) {
3100 table = bgp_node_get_bgp_table_info(rd_rn);
3101 if (!table)
3102 continue;
3103
3104 for (rn = bgp_table_top(table); rn; rn = bgp_route_next(rn)) {
3105 struct prefix_evpn *evp = (struct prefix_evpn *)&rn->p;
3106
3107 for (pi = bgp_node_get_bgp_path_info(rn); pi;
3108 pi = pi->next) {
3109 /*
3110 * Consider "valid" remote routes applicable for
3111 * this ES.
3112 */
3113 if (!(CHECK_FLAG(pi->flags, BGP_PATH_VALID)
3114 && pi->type == ZEBRA_ROUTE_BGP
3115 && pi->sub_type == BGP_ROUTE_NORMAL))
3116 continue;
3117
3118 if (!is_prefix_matching_for_es(evp, es))
3119 continue;
3120
3121 if (install)
3122 ret = install_evpn_route_entry_in_es(
3123 bgp, es, evp, pi);
3124 else
3125 ret = uninstall_evpn_route_entry_in_es(
3126 bgp, es, evp, pi);
3127
3128 if (ret) {
3129 flog_err(
3130 EC_BGP_EVPN_FAIL,
3131 "Failed to %s EVPN %s route in ESI %s",
3132 install ? "install"
3133 : "uninstall",
3134 prefix2str(evp, buf,
3135 sizeof(buf)),
3136 esi_to_str(&es->esi, buf1,
3137 sizeof(buf1)));
3138 return ret;
3139 }
3140 }
3141 }
3142 }
3143 return 0;
3144 }
3145
3146 /* This API will scan evpn routes for checking attribute's rmac
3147 * macthes with bgp instance router mac. It avoid installing
3148 * route into bgp vrf table and remote rmac in bridge table.
3149 */
3150 static int bgp_evpn_route_rmac_self_check(struct bgp *bgp_vrf,
3151 struct prefix_evpn *evp,
3152 struct bgp_path_info *pi)
3153 {
3154 /* evpn route could have learnt prior to L3vni has come up,
3155 * perform rmac check before installing route and
3156 * remote router mac.
3157 * The route will be removed from global bgp table once
3158 * SVI comes up with MAC and stored in hash, triggers
3159 * bgp_mac_rescan_all_evpn_tables.
3160 */
3161 if (memcmp(&bgp_vrf->rmac, &pi->attr->rmac, ETH_ALEN) == 0) {
3162 if (bgp_debug_update(pi->peer, NULL, NULL, 1)) {
3163 char buf1[PREFIX_STRLEN];
3164 char attr_str[BUFSIZ] = {0};
3165
3166 bgp_dump_attr(pi->attr, attr_str, BUFSIZ);
3167
3168 zlog_debug("%s: bgp %u prefix %s with attr %s - DENIED due to self mac",
3169 __func__, bgp_vrf->vrf_id,
3170 prefix2str(evp, buf1, sizeof(buf1)),
3171 attr_str);
3172 }
3173
3174 return 1;
3175 }
3176
3177 return 0;
3178 }
3179
3180 /*
3181 * Install or uninstall mac-ip routes are appropriate for this
3182 * particular VRF.
3183 */
3184 static int install_uninstall_routes_for_vrf(struct bgp *bgp_vrf, int install)
3185 {
3186 afi_t afi;
3187 safi_t safi;
3188 struct bgp_node *rd_rn, *rn;
3189 struct bgp_table *table;
3190 struct bgp_path_info *pi;
3191 int ret;
3192 char buf[PREFIX_STRLEN];
3193 struct bgp *bgp_evpn = NULL;
3194
3195 afi = AFI_L2VPN;
3196 safi = SAFI_EVPN;
3197 bgp_evpn = bgp_get_evpn();
3198 if (!bgp_evpn)
3199 return -1;
3200
3201 /* Walk entire global routing table and evaluate routes which could be
3202 * imported into this VRF. Note that we need to loop through all global
3203 * routes to determine which route matches the import rt on vrf
3204 */
3205 for (rd_rn = bgp_table_top(bgp_evpn->rib[afi][safi]); rd_rn;
3206 rd_rn = bgp_route_next(rd_rn)) {
3207 table = bgp_node_get_bgp_table_info(rd_rn);
3208 if (!table)
3209 continue;
3210
3211 for (rn = bgp_table_top(table); rn; rn = bgp_route_next(rn)) {
3212 struct prefix_evpn *evp = (struct prefix_evpn *)&rn->p;
3213
3214 /* if not mac-ip route skip this route */
3215 if (!(evp->prefix.route_type == BGP_EVPN_MAC_IP_ROUTE
3216 || evp->prefix.route_type
3217 == BGP_EVPN_IP_PREFIX_ROUTE))
3218 continue;
3219
3220 /* if not a mac+ip route skip this route */
3221 if (!(is_evpn_prefix_ipaddr_v4(evp)
3222 || is_evpn_prefix_ipaddr_v6(evp)))
3223 continue;
3224
3225 for (pi = bgp_node_get_bgp_path_info(rn); pi;
3226 pi = pi->next) {
3227 /* Consider "valid" remote routes applicable for
3228 * this VRF.
3229 */
3230 if (!(CHECK_FLAG(pi->flags, BGP_PATH_VALID)
3231 && pi->type == ZEBRA_ROUTE_BGP
3232 && pi->sub_type == BGP_ROUTE_NORMAL))
3233 continue;
3234
3235 if (is_route_matching_for_vrf(bgp_vrf, pi)) {
3236 if (bgp_evpn_route_rmac_self_check(
3237 bgp_vrf, evp, pi))
3238 continue;
3239
3240 if (install)
3241 ret = install_evpn_route_entry_in_vrf(
3242 bgp_vrf, evp, pi);
3243 else
3244 ret = uninstall_evpn_route_entry_in_vrf(
3245 bgp_vrf, evp, pi);
3246
3247 if (ret) {
3248 flog_err(
3249 EC_BGP_EVPN_FAIL,
3250 "Failed to %s EVPN %s route in VRF %s",
3251 install ? "install"
3252 : "uninstall",
3253 prefix2str(evp, buf,
3254 sizeof(buf)),
3255 vrf_id_to_name(
3256 bgp_vrf->vrf_id));
3257 return ret;
3258 }
3259 }
3260 }
3261 }
3262 }
3263
3264 return 0;
3265 }
3266
3267 /*
3268 * Install or uninstall routes of specified type that are appropriate for this
3269 * particular VNI.
3270 */
3271 static int install_uninstall_routes_for_vni(struct bgp *bgp,
3272 struct bgpevpn *vpn,
3273 bgp_evpn_route_type rtype,
3274 int install)
3275 {
3276 afi_t afi;
3277 safi_t safi;
3278 struct bgp_node *rd_rn, *rn;
3279 struct bgp_table *table;
3280 struct bgp_path_info *pi;
3281 int ret;
3282
3283 afi = AFI_L2VPN;
3284 safi = SAFI_EVPN;
3285
3286 /* Walk entire global routing table and evaluate routes which could be
3287 * imported into this VPN. Note that we cannot just look at the routes
3288 * for
3289 * the VNI's RD - remote routes applicable for this VNI could have any
3290 * RD.
3291 */
3292 /* EVPN routes are a 2-level table. */
3293 for (rd_rn = bgp_table_top(bgp->rib[afi][safi]); rd_rn;
3294 rd_rn = bgp_route_next(rd_rn)) {
3295 table = bgp_node_get_bgp_table_info(rd_rn);
3296 if (!table)
3297 continue;
3298
3299 for (rn = bgp_table_top(table); rn; rn = bgp_route_next(rn)) {
3300 struct prefix_evpn *evp = (struct prefix_evpn *)&rn->p;
3301
3302 if (evp->prefix.route_type != rtype)
3303 continue;
3304
3305 for (pi = bgp_node_get_bgp_path_info(rn); pi;
3306 pi = pi->next) {
3307 /* Consider "valid" remote routes applicable for
3308 * this VNI. */
3309 if (!(CHECK_FLAG(pi->flags, BGP_PATH_VALID)
3310 && pi->type == ZEBRA_ROUTE_BGP
3311 && pi->sub_type == BGP_ROUTE_NORMAL))
3312 continue;
3313
3314 if (is_route_matching_for_vni(bgp, vpn, pi)) {
3315 if (install)
3316 ret = install_evpn_route_entry(
3317 bgp, vpn, evp, pi);
3318 else
3319 ret = uninstall_evpn_route_entry(
3320 bgp, vpn, evp, pi);
3321
3322 if (ret) {
3323 flog_err(
3324 EC_BGP_EVPN_FAIL,
3325 "%u: Failed to %s EVPN %s route in VNI %u",
3326 bgp->vrf_id,
3327 install ? "install"
3328 : "uninstall",
3329 rtype == BGP_EVPN_MAC_IP_ROUTE
3330 ? "MACIP"
3331 : "IMET",
3332 vpn->vni);
3333 return ret;
3334 }
3335 }
3336 }
3337 }
3338 }
3339
3340 return 0;
3341 }
3342
3343 /* Install any existing remote ES routes applicable for this ES into its routing
3344 * table. This is invoked when ES comes up.
3345 */
3346 static int install_routes_for_es(struct bgp *bgp, struct evpnes *es)
3347 {
3348 return install_uninstall_routes_for_es(bgp, es, 1);
3349 }
3350
3351
3352 /* Install any existing remote routes applicable for this VRF into VRF RIB. This
3353 * is invoked upon l3vni-add or l3vni import rt change
3354 */
3355 static int install_routes_for_vrf(struct bgp *bgp_vrf)
3356 {
3357 install_uninstall_routes_for_vrf(bgp_vrf, 1);
3358 return 0;
3359 }
3360
3361 /*
3362 * Install any existing remote routes applicable for this VNI into its
3363 * routing table. This is invoked when a VNI becomes "live" or its Import
3364 * RT is changed.
3365 */
3366 static int install_routes_for_vni(struct bgp *bgp, struct bgpevpn *vpn)
3367 {
3368 int ret;
3369
3370 /* Install type-3 routes followed by type-2 routes - the ones applicable
3371 * for this VNI.
3372 */
3373 ret = install_uninstall_routes_for_vni(bgp, vpn, BGP_EVPN_IMET_ROUTE,
3374 1);
3375 if (ret)
3376 return ret;
3377
3378 return install_uninstall_routes_for_vni(bgp, vpn, BGP_EVPN_MAC_IP_ROUTE,
3379 1);
3380 }
3381
3382 /* uninstall routes from l3vni vrf. */
3383 static int uninstall_routes_for_vrf(struct bgp *bgp_vrf)
3384 {
3385 install_uninstall_routes_for_vrf(bgp_vrf, 0);
3386 return 0;
3387 }
3388
3389 /*
3390 * Uninstall any existing remote routes for this VNI. One scenario in which
3391 * this is invoked is upon an import RT change.
3392 */
3393 static int uninstall_routes_for_vni(struct bgp *bgp, struct bgpevpn *vpn)
3394 {
3395 int ret;
3396
3397 /* Uninstall type-2 routes followed by type-3 routes - the ones
3398 * applicable
3399 * for this VNI.
3400 */
3401 ret = install_uninstall_routes_for_vni(bgp, vpn, BGP_EVPN_MAC_IP_ROUTE,
3402 0);
3403 if (ret)
3404 return ret;
3405
3406 return install_uninstall_routes_for_vni(bgp, vpn, BGP_EVPN_IMET_ROUTE,
3407 0);
3408 }
3409
3410 /* Install or unistall route in ES */
3411 static int install_uninstall_route_in_es(struct bgp *bgp, struct evpnes *es,
3412 afi_t afi, safi_t safi,
3413 struct prefix_evpn *evp,
3414 struct bgp_path_info *pi, int install)
3415 {
3416 int ret = 0;
3417 char buf[ESI_STR_LEN];
3418
3419 if (install)
3420 ret = install_evpn_route_entry_in_es(bgp, es, evp, pi);
3421 else
3422 ret = uninstall_evpn_route_entry_in_es(bgp, es, evp, pi);
3423
3424 if (ret) {
3425 flog_err(
3426 EC_BGP_EVPN_FAIL,
3427 "%u: Failed to %s EVPN %s route in ESI %s", bgp->vrf_id,
3428 install ? "install" : "uninstall", "ES",
3429 esi_to_str(&evp->prefix.es_addr.esi, buf, sizeof(buf)));
3430 return ret;
3431 }
3432 return 0;
3433 }
3434
3435 /*
3436 * Install or uninstall route in matching VRFs (list).
3437 */
3438 static int install_uninstall_route_in_vrfs(struct bgp *bgp_def, afi_t afi,
3439 safi_t safi, struct prefix_evpn *evp,
3440 struct bgp_path_info *pi,
3441 struct list *vrfs, int install)
3442 {
3443 char buf[PREFIX2STR_BUFFER];
3444 struct bgp *bgp_vrf;
3445 struct listnode *node, *nnode;
3446
3447 /* Only type-2/type-5 routes go into a VRF */
3448 if (!(evp->prefix.route_type == BGP_EVPN_MAC_IP_ROUTE
3449 || evp->prefix.route_type == BGP_EVPN_IP_PREFIX_ROUTE))
3450 return 0;
3451
3452 /* if it is type-2 route and not a mac+ip route skip this route */
3453 if ((evp->prefix.route_type == BGP_EVPN_MAC_IP_ROUTE)
3454 && !(is_evpn_prefix_ipaddr_v4(evp)
3455 || is_evpn_prefix_ipaddr_v6(evp)))
3456 return 0;
3457
3458 for (ALL_LIST_ELEMENTS(vrfs, node, nnode, bgp_vrf)) {
3459 int ret;
3460
3461 if (install)
3462 ret = install_evpn_route_entry_in_vrf(bgp_vrf, evp, pi);
3463 else
3464 ret = uninstall_evpn_route_entry_in_vrf(bgp_vrf, evp,
3465 pi);
3466
3467 if (ret) {
3468 flog_err(EC_BGP_EVPN_FAIL,
3469 "%u: Failed to %s prefix %s in VRF %s",
3470 bgp_def->vrf_id,
3471 install ? "install" : "uninstall",
3472 prefix2str(evp, buf, sizeof(buf)),
3473 vrf_id_to_name(bgp_vrf->vrf_id));
3474 return ret;
3475 }
3476 }
3477
3478 return 0;
3479 }
3480
3481 /*
3482 * Install or uninstall route in matching VNIs (list).
3483 */
3484 static int install_uninstall_route_in_vnis(struct bgp *bgp, afi_t afi,
3485 safi_t safi, struct prefix_evpn *evp,
3486 struct bgp_path_info *pi,
3487 struct list *vnis, int install)
3488 {
3489 struct bgpevpn *vpn;
3490 struct listnode *node, *nnode;
3491
3492 for (ALL_LIST_ELEMENTS(vnis, node, nnode, vpn)) {
3493 int ret;
3494
3495 if (!is_vni_live(vpn))
3496 continue;
3497
3498 if (install)
3499 ret = install_evpn_route_entry(bgp, vpn, evp, pi);
3500 else
3501 ret = uninstall_evpn_route_entry(bgp, vpn, evp, pi);
3502
3503 if (ret) {
3504 flog_err(EC_BGP_EVPN_FAIL,
3505 "%u: Failed to %s EVPN %s route in VNI %u",
3506 bgp->vrf_id, install ? "install" : "uninstall",
3507 evp->prefix.route_type == BGP_EVPN_MAC_IP_ROUTE
3508 ? "MACIP"
3509 : "IMET",
3510 vpn->vni);
3511 return ret;
3512 }
3513 }
3514
3515 return 0;
3516 }
3517
3518 /*
3519 * Install or uninstall route for appropriate VNIs/ESIs.
3520 */
3521 static int install_uninstall_evpn_route(struct bgp *bgp, afi_t afi, safi_t safi,
3522 struct prefix *p,
3523 struct bgp_path_info *pi, int import)
3524 {
3525 struct prefix_evpn *evp = (struct prefix_evpn *)p;
3526 struct attr *attr = pi->attr;
3527 struct ecommunity *ecom;
3528 int i;
3529
3530 assert(attr);
3531
3532 /* Only type-2, type-3, type-4 and type-5 are supported currently */
3533 if (!(evp->prefix.route_type == BGP_EVPN_MAC_IP_ROUTE
3534 || evp->prefix.route_type == BGP_EVPN_IMET_ROUTE
3535 || evp->prefix.route_type == BGP_EVPN_ES_ROUTE
3536 || evp->prefix.route_type == BGP_EVPN_IP_PREFIX_ROUTE))
3537 return 0;
3538
3539 /* If we don't have Route Target, nothing much to do. */
3540 if (!(attr->flag & ATTR_FLAG_BIT(BGP_ATTR_EXT_COMMUNITIES)))
3541 return 0;
3542
3543 ecom = attr->ecommunity;
3544 if (!ecom || !ecom->size)
3545 return -1;
3546
3547 /* An EVPN route belongs to a VNI or a VRF or an ESI based on the RTs
3548 * attached to the route */
3549 for (i = 0; i < ecom->size; i++) {
3550 uint8_t *pnt;
3551 uint8_t type, sub_type;
3552 struct ecommunity_val *eval;
3553 struct ecommunity_val eval_tmp;
3554 struct irt_node *irt; /* import rt for l2vni */
3555 struct vrf_irt_node *vrf_irt; /* import rt for l3vni */
3556 struct evpnes *es;
3557
3558 /* Only deal with RTs */
3559 pnt = (ecom->val + (i * ECOMMUNITY_SIZE));
3560 eval = (struct ecommunity_val *)(ecom->val
3561 + (i * ECOMMUNITY_SIZE));
3562 type = *pnt++;
3563 sub_type = *pnt++;
3564 if (sub_type != ECOMMUNITY_ROUTE_TARGET)
3565 continue;
3566
3567 /*
3568 * macip routes (type-2) are imported into VNI and VRF tables.
3569 * IMET route is imported into VNI table.
3570 * prefix routes are imported into VRF table.
3571 */
3572 if (evp->prefix.route_type == BGP_EVPN_MAC_IP_ROUTE ||
3573 evp->prefix.route_type == BGP_EVPN_IMET_ROUTE ||
3574 evp->prefix.route_type == BGP_EVPN_IP_PREFIX_ROUTE) {
3575
3576 irt = lookup_import_rt(bgp, eval);
3577 if (irt)
3578 install_uninstall_route_in_vnis(
3579 bgp, afi, safi, evp, pi, irt->vnis,
3580 import);
3581
3582 vrf_irt = lookup_vrf_import_rt(eval);
3583 if (vrf_irt)
3584 install_uninstall_route_in_vrfs(
3585 bgp, afi, safi, evp, pi, vrf_irt->vrfs,
3586 import);
3587
3588 /* Also check for non-exact match.
3589 * In this, we mask out the AS and
3590 * only check on the local-admin sub-field.
3591 * This is to facilitate using
3592 * VNI as the RT for EBGP peering too.
3593 */
3594 irt = NULL;
3595 vrf_irt = NULL;
3596 if (type == ECOMMUNITY_ENCODE_AS
3597 || type == ECOMMUNITY_ENCODE_AS4
3598 || type == ECOMMUNITY_ENCODE_IP) {
3599 memcpy(&eval_tmp, eval, ECOMMUNITY_SIZE);
3600 mask_ecom_global_admin(&eval_tmp, eval);
3601 irt = lookup_import_rt(bgp, &eval_tmp);
3602 vrf_irt = lookup_vrf_import_rt(&eval_tmp);
3603 }
3604
3605 if (irt)
3606 install_uninstall_route_in_vnis(
3607 bgp, afi, safi, evp, pi, irt->vnis,
3608 import);
3609 if (vrf_irt)
3610 install_uninstall_route_in_vrfs(
3611 bgp, afi, safi, evp, pi, vrf_irt->vrfs,
3612 import);
3613 }
3614
3615 /* es route is imported into the es table */
3616 if (evp->prefix.route_type == BGP_EVPN_ES_ROUTE) {
3617
3618 /* we will match based on the entire esi to avoid
3619 * imoort of an es route for esi2 into esi1
3620 */
3621 es = bgp_evpn_lookup_es(bgp, &evp->prefix.es_addr.esi);
3622 if (es && is_es_local(es))
3623 install_uninstall_route_in_es(
3624 bgp, es, afi, safi, evp, pi, import);
3625 }
3626 }
3627
3628 return 0;
3629 }
3630
3631 /*
3632 * delete and withdraw all ipv4 and ipv6 routes in the vrf table as type-5
3633 * routes
3634 */
3635 static void delete_withdraw_vrf_routes(struct bgp *bgp_vrf)
3636 {
3637 /* Delete ipv4 default route and withdraw from peers */
3638 if (evpn_default_originate_set(bgp_vrf, AFI_IP, SAFI_UNICAST))
3639 bgp_evpn_install_uninstall_default_route(bgp_vrf, AFI_IP,
3640 SAFI_UNICAST, false);
3641
3642 /* delete all ipv4 routes and withdraw from peers */
3643 if (advertise_type5_routes(bgp_vrf, AFI_IP))
3644 bgp_evpn_withdraw_type5_routes(bgp_vrf, AFI_IP, SAFI_UNICAST);
3645
3646 /* Delete ipv6 default route and withdraw from peers */
3647 if (evpn_default_originate_set(bgp_vrf, AFI_IP6, SAFI_UNICAST))
3648 bgp_evpn_install_uninstall_default_route(bgp_vrf, AFI_IP6,
3649 SAFI_UNICAST, false);
3650
3651 /* delete all ipv6 routes and withdraw from peers */
3652 if (advertise_type5_routes(bgp_vrf, AFI_IP6))
3653 bgp_evpn_withdraw_type5_routes(bgp_vrf, AFI_IP6, SAFI_UNICAST);
3654 }
3655
3656 /*
3657 * update and advertise all ipv4 and ipv6 routes in thr vrf table as type-5
3658 * routes
3659 */
3660 void update_advertise_vrf_routes(struct bgp *bgp_vrf)
3661 {
3662 struct bgp *bgp_evpn = NULL; /* EVPN bgp instance */
3663
3664 bgp_evpn = bgp_get_evpn();
3665 if (!bgp_evpn)
3666 return;
3667
3668 /* update all ipv4 routes */
3669 if (advertise_type5_routes(bgp_vrf, AFI_IP))
3670 bgp_evpn_advertise_type5_routes(bgp_vrf, AFI_IP, SAFI_UNICAST);
3671
3672 /* update ipv4 default route and withdraw from peers */
3673 if (evpn_default_originate_set(bgp_vrf, AFI_IP, SAFI_UNICAST))
3674 bgp_evpn_install_uninstall_default_route(bgp_vrf, AFI_IP,
3675 SAFI_UNICAST, true);
3676
3677 /* update all ipv6 routes */
3678 if (advertise_type5_routes(bgp_vrf, AFI_IP6))
3679 bgp_evpn_advertise_type5_routes(bgp_vrf, AFI_IP6, SAFI_UNICAST);
3680
3681 /* update ipv6 default route and withdraw from peers */
3682 if (evpn_default_originate_set(bgp_vrf, AFI_IP6, SAFI_UNICAST))
3683 bgp_evpn_install_uninstall_default_route(bgp_vrf, AFI_IP6,
3684 SAFI_UNICAST, true);
3685
3686 }
3687
3688 /*
3689 * update and advertise local routes for a VRF as type-5 routes.
3690 * This is invoked upon RD change for a VRF. Note taht the processing is only
3691 * done in the global route table using the routes which already exist in the
3692 * VRF routing table
3693 */
3694 static void update_router_id_vrf(struct bgp *bgp_vrf)
3695 {
3696 /* skip if the RD is configured */
3697 if (is_vrf_rd_configured(bgp_vrf))
3698 return;
3699
3700 /* derive the RD for the VRF based on new router-id */
3701 bgp_evpn_derive_auto_rd_for_vrf(bgp_vrf);
3702
3703 /* update advertise ipv4|ipv6 routes as type-5 routes */
3704 update_advertise_vrf_routes(bgp_vrf);
3705 }
3706
3707 /*
3708 * Delete and withdraw all type-5 routes for the RD corresponding to VRF.
3709 * This is invoked upon VRF RD change. The processing is done only from global
3710 * table.
3711 */
3712 static void withdraw_router_id_vrf(struct bgp *bgp_vrf)
3713 {
3714 /* skip if the RD is configured */
3715 if (is_vrf_rd_configured(bgp_vrf))
3716 return;
3717
3718 /* delete/withdraw ipv4|ipv6 routes as type-5 routes */
3719 delete_withdraw_vrf_routes(bgp_vrf);
3720 }
3721
3722 /*
3723 * Update and advertise local routes for a VNI. Invoked upon router-id
3724 * change. Note that the processing is done only on the global route table
3725 * using routes that already exist in the per-VNI table.
3726 */
3727 static int update_advertise_vni_routes(struct bgp *bgp, struct bgpevpn *vpn)
3728 {
3729 struct prefix_evpn p;
3730 struct bgp_node *rn, *global_rn;
3731 struct bgp_path_info *pi, *global_pi;
3732 struct attr *attr;
3733 afi_t afi = AFI_L2VPN;
3734 safi_t safi = SAFI_EVPN;
3735
3736 /* Locate type-3 route for VNI in the per-VNI table and use its
3737 * attributes to create and advertise the type-3 route for this VNI
3738 * in the global table.
3739 *
3740 * RT-3 only if doing head-end replication
3741 */
3742 if (bgp_evpn_vni_flood_mode_get(bgp, vpn)
3743 == VXLAN_FLOOD_HEAD_END_REPL) {
3744 build_evpn_type3_prefix(&p, vpn->originator_ip);
3745 rn = bgp_node_lookup(vpn->route_table, (struct prefix *)&p);
3746 if (!rn) /* unexpected */
3747 return 0;
3748 for (pi = bgp_node_get_bgp_path_info(rn); pi; pi = pi->next)
3749 if (pi->peer == bgp->peer_self &&
3750 pi->type == ZEBRA_ROUTE_BGP
3751 && pi->sub_type == BGP_ROUTE_STATIC)
3752 break;
3753 if (!pi) /* unexpected */
3754 return 0;
3755 attr = pi->attr;
3756
3757 global_rn = bgp_afi_node_get(bgp->rib[afi][safi], afi, safi,
3758 (struct prefix *)&p, &vpn->prd);
3759 update_evpn_route_entry(bgp, vpn, afi, safi, global_rn, attr,
3760 1, &pi, 0, mac_mobility_seqnum(attr));
3761
3762 /* Schedule for processing and unlock node. */
3763 bgp_process(bgp, global_rn, afi, safi);
3764 bgp_unlock_node(global_rn);
3765 }
3766
3767 /* Now, walk this VNI's route table and use the route and its attribute
3768 * to create and schedule route in global table.
3769 */
3770 for (rn = bgp_table_top(vpn->route_table); rn;
3771 rn = bgp_route_next(rn)) {
3772 struct prefix_evpn *evp = (struct prefix_evpn *)&rn->p;
3773
3774 /* Identify MAC-IP local routes. */
3775 if (evp->prefix.route_type != BGP_EVPN_MAC_IP_ROUTE)
3776 continue;
3777
3778 for (pi = bgp_node_get_bgp_path_info(rn); pi; pi = pi->next)
3779 if (pi->peer == bgp->peer_self
3780 && pi->type == ZEBRA_ROUTE_BGP
3781 && pi->sub_type == BGP_ROUTE_STATIC)
3782 break;
3783 if (!pi)
3784 continue;
3785
3786 /* Create route in global routing table using this route entry's
3787 * attribute.
3788 */
3789 attr = pi->attr;
3790 global_rn = bgp_afi_node_get(bgp->rib[afi][safi], afi, safi,
3791 (struct prefix *)evp, &vpn->prd);
3792 assert(global_rn);
3793 update_evpn_route_entry(bgp, vpn, afi, safi, global_rn, attr, 1,
3794 &global_pi, 0,
3795 mac_mobility_seqnum(attr));
3796
3797 /* Schedule for processing and unlock node. */
3798 bgp_process(bgp, global_rn, afi, safi);
3799 bgp_unlock_node(global_rn);
3800 }
3801
3802 return 0;
3803 }
3804
3805 /*
3806 * Delete (and withdraw) local routes for a VNI - only from the global
3807 * table. Invoked upon router-id change.
3808 */
3809 static int delete_withdraw_vni_routes(struct bgp *bgp, struct bgpevpn *vpn)
3810 {
3811 int ret;
3812 struct prefix_evpn p;
3813 struct bgp_node *global_rn;
3814 struct bgp_path_info *pi;
3815 afi_t afi = AFI_L2VPN;
3816 safi_t safi = SAFI_EVPN;
3817
3818 /* Delete and withdraw locally learnt type-2 routes (MACIP)
3819 * for this VNI - from the global table.
3820 */
3821 ret = delete_global_type2_routes(bgp, vpn);
3822 if (ret)
3823 return ret;
3824
3825 /* Remove type-3 route for this VNI from global table. */
3826 build_evpn_type3_prefix(&p, vpn->originator_ip);
3827 global_rn = bgp_afi_node_lookup(bgp->rib[afi][safi], afi, safi,
3828 (struct prefix *)&p, &vpn->prd);
3829 if (global_rn) {
3830 /* Delete route entry in the global EVPN table. */
3831 delete_evpn_route_entry(bgp, afi, safi, global_rn, &pi);
3832
3833 /* Schedule for processing - withdraws to peers happen from
3834 * this table.
3835 */
3836 if (pi)
3837 bgp_process(bgp, global_rn, afi, safi);
3838 bgp_unlock_node(global_rn);
3839 }
3840
3841 return 0;
3842 }
3843
3844 /*
3845 * Handle router-id change. Update and advertise local routes corresponding
3846 * to this VNI from peers. Note that this is invoked after updating the
3847 * router-id. The routes in the per-VNI table are used to create routes in
3848 * the global table and schedule them.
3849 */
3850 static void update_router_id_vni(struct hash_bucket *bucket, struct bgp *bgp)
3851 {
3852 struct bgpevpn *vpn = (struct bgpevpn *)bucket->data;
3853
3854 /* Skip VNIs with configured RD. */
3855 if (is_rd_configured(vpn))
3856 return;
3857
3858 bgp_evpn_derive_auto_rd(bgp, vpn);
3859 update_advertise_vni_routes(bgp, vpn);
3860 }
3861
3862 /*
3863 * Handle router-id change. Delete and withdraw local routes corresponding
3864 * to this VNI from peers. Note that this is invoked prior to updating
3865 * the router-id and is done only on the global route table, the routes
3866 * are needed in the per-VNI table to re-advertise with new router id.
3867 */
3868 static void withdraw_router_id_vni(struct hash_bucket *bucket, struct bgp *bgp)
3869 {
3870 struct bgpevpn *vpn = (struct bgpevpn *)bucket->data;
3871
3872 /* Skip VNIs with configured RD. */
3873 if (is_rd_configured(vpn))
3874 return;
3875
3876 delete_withdraw_vni_routes(bgp, vpn);
3877 }
3878
3879 /*
3880 * Create RT-3 for a VNI and schedule for processing and advertisement.
3881 * This is invoked upon flooding mode changing to head-end replication.
3882 */
3883 static void create_advertise_type3(struct hash_bucket *bucket, void *data)
3884 {
3885 struct bgpevpn *vpn = bucket->data;
3886 struct bgp *bgp = data;
3887 struct prefix_evpn p;
3888
3889 if (!vpn || !is_vni_live(vpn) ||
3890 bgp_evpn_vni_flood_mode_get(bgp, vpn)
3891 != VXLAN_FLOOD_HEAD_END_REPL)
3892 return;
3893
3894 build_evpn_type3_prefix(&p, vpn->originator_ip);
3895 if (update_evpn_route(bgp, vpn, &p, 0, 0))
3896 flog_err(EC_BGP_EVPN_ROUTE_CREATE,
3897 "Type3 route creation failure for VNI %u", vpn->vni);
3898 }
3899
3900 /*
3901 * Delete RT-3 for a VNI and schedule for processing and withdrawal.
3902 * This is invoked upon flooding mode changing to drop BUM packets.
3903 */
3904 static void delete_withdraw_type3(struct hash_bucket *bucket, void *data)
3905 {
3906 struct bgpevpn *vpn = bucket->data;
3907 struct bgp *bgp = data;
3908 struct prefix_evpn p;
3909
3910 if (!vpn || !is_vni_live(vpn))
3911 return;
3912
3913 build_evpn_type3_prefix(&p, vpn->originator_ip);
3914 delete_evpn_route(bgp, vpn, &p);
3915 }
3916
3917 /*
3918 * Process received EVPN type-2 route (advertise or withdraw).
3919 */
3920 static int process_type2_route(struct peer *peer, afi_t afi, safi_t safi,
3921 struct attr *attr, uint8_t *pfx, int psize,
3922 uint32_t addpath_id)
3923 {
3924 struct prefix_rd prd;
3925 struct prefix_evpn p;
3926 struct bgp_route_evpn evpn;
3927 uint8_t ipaddr_len;
3928 uint8_t macaddr_len;
3929 mpls_label_t label[BGP_MAX_LABELS]; /* holds the VNI(s) as in packet */
3930 uint32_t num_labels = 0;
3931 uint32_t eth_tag;
3932 int ret;
3933
3934 /* Type-2 route should be either 33, 37 or 49 bytes or an
3935 * additional 3 bytes if there is a second label (VNI):
3936 * RD (8), ESI (10), Eth Tag (4), MAC Addr Len (1),
3937 * MAC Addr (6), IP len (1), IP (0, 4 or 16),
3938 * MPLS Lbl1 (3), MPLS Lbl2 (0 or 3)
3939 */
3940 if (psize != 33 && psize != 37 && psize != 49 && psize != 36
3941 && psize != 40 && psize != 52) {
3942 flog_err(EC_BGP_EVPN_ROUTE_INVALID,
3943 "%u:%s - Rx EVPN Type-2 NLRI with invalid length %d",
3944 peer->bgp->vrf_id, peer->host, psize);
3945 return -1;
3946 }
3947
3948 memset(&evpn, 0, sizeof(evpn));
3949
3950 /* Make prefix_rd */
3951 prd.family = AF_UNSPEC;
3952 prd.prefixlen = 64;
3953 memcpy(&prd.val, pfx, 8);
3954 pfx += 8;
3955
3956 /* Make EVPN prefix. */
3957 memset(&p, 0, sizeof(struct prefix_evpn));
3958 p.family = AF_EVPN;
3959 p.prefixlen = EVPN_ROUTE_PREFIXLEN;
3960 p.prefix.route_type = BGP_EVPN_MAC_IP_ROUTE;
3961
3962 /* Copy Ethernet Seg Identifier */
3963 memcpy(&evpn.eth_s_id.val, pfx, ESI_LEN);
3964 pfx += ESI_LEN;
3965
3966 /* Copy Ethernet Tag */
3967 memcpy(&eth_tag, pfx, 4);
3968 p.prefix.macip_addr.eth_tag = ntohl(eth_tag);
3969 pfx += 4;
3970
3971 /* Get the MAC Addr len */
3972 macaddr_len = *pfx++;
3973
3974 /* Get the MAC Addr */
3975 if (macaddr_len == (ETH_ALEN * 8)) {
3976 memcpy(&p.prefix.macip_addr.mac.octet, pfx, ETH_ALEN);
3977 pfx += ETH_ALEN;
3978 } else {
3979 flog_err(
3980 EC_BGP_EVPN_ROUTE_INVALID,
3981 "%u:%s - Rx EVPN Type-2 NLRI with unsupported MAC address length %d",
3982 peer->bgp->vrf_id, peer->host, macaddr_len);
3983 return -1;
3984 }
3985
3986
3987 /* Get the IP. */
3988 ipaddr_len = *pfx++;
3989 if (ipaddr_len != 0 && ipaddr_len != IPV4_MAX_BITLEN
3990 && ipaddr_len != IPV6_MAX_BITLEN) {
3991 flog_err(
3992 EC_BGP_EVPN_ROUTE_INVALID,
3993 "%u:%s - Rx EVPN Type-2 NLRI with unsupported IP address length %d",
3994 peer->bgp->vrf_id, peer->host, ipaddr_len);
3995 return -1;
3996 }
3997
3998 if (ipaddr_len) {
3999 ipaddr_len /= 8; /* Convert to bytes. */
4000 p.prefix.macip_addr.ip.ipa_type = (ipaddr_len == IPV4_MAX_BYTELEN)
4001 ? IPADDR_V4
4002 : IPADDR_V6;
4003 memcpy(&p.prefix.macip_addr.ip.ip.addr, pfx, ipaddr_len);
4004 }
4005 pfx += ipaddr_len;
4006
4007 /* Get the VNI(s). Stored as bytes here. */
4008 num_labels++;
4009 memset(label, 0, sizeof(label));
4010 memcpy(&label[0], pfx, BGP_LABEL_BYTES);
4011 pfx += BGP_LABEL_BYTES;
4012 psize -= (33 + ipaddr_len);
4013 /* Do we have a second VNI? */
4014 if (psize) {
4015 num_labels++;
4016 memcpy(&label[1], pfx, BGP_LABEL_BYTES);
4017 /*
4018 * If in future, we are required to access additional fields,
4019 * we MUST increment pfx by BGP_LABEL_BYTES in before reading
4020 * the next field
4021 */
4022 }
4023
4024 /* Process the route. */
4025 if (attr)
4026 ret = bgp_update(peer, (struct prefix *)&p, addpath_id, attr,
4027 afi, safi, ZEBRA_ROUTE_BGP, BGP_ROUTE_NORMAL,
4028 &prd, &label[0], num_labels, 0, &evpn);
4029 else
4030 ret = bgp_withdraw(peer, (struct prefix *)&p, addpath_id, attr,
4031 afi, safi, ZEBRA_ROUTE_BGP, BGP_ROUTE_NORMAL,
4032 &prd, &label[0], num_labels, &evpn);
4033 return ret;
4034 }
4035
4036 /*
4037 * Process received EVPN type-3 route (advertise or withdraw).
4038 */
4039 static int process_type3_route(struct peer *peer, afi_t afi, safi_t safi,
4040 struct attr *attr, uint8_t *pfx, int psize,
4041 uint32_t addpath_id)
4042 {
4043 struct prefix_rd prd;
4044 struct prefix_evpn p;
4045 uint8_t ipaddr_len;
4046 uint32_t eth_tag;
4047 int ret;
4048
4049 /* Type-3 route should be either 17 or 29 bytes: RD (8), Eth Tag (4),
4050 * IP len (1) and IP (4 or 16).
4051 */
4052 if (psize != 17 && psize != 29) {
4053 flog_err(EC_BGP_EVPN_ROUTE_INVALID,
4054 "%u:%s - Rx EVPN Type-3 NLRI with invalid length %d",
4055 peer->bgp->vrf_id, peer->host, psize);
4056 return -1;
4057 }
4058
4059 /* If PMSI is present, log if it is anything other than IR.
4060 * Note: We just simply ignore the values as it is not clear if
4061 * doing anything else is better.
4062 */
4063 if (attr &&
4064 (attr->flag & ATTR_FLAG_BIT(BGP_ATTR_PMSI_TUNNEL))) {
4065 if (attr->pmsi_tnl_type != PMSI_TNLTYPE_INGR_REPL &&
4066 attr->pmsi_tnl_type != PMSI_TNLTYPE_PIM_SM) {
4067 flog_warn(EC_BGP_EVPN_PMSI_PRESENT,
4068 "%u:%s - Rx EVPN Type-3 NLRI with unsupported PTA %d",
4069 peer->bgp->vrf_id, peer->host,
4070 attr->pmsi_tnl_type);
4071 }
4072 }
4073
4074 /* Make prefix_rd */
4075 prd.family = AF_UNSPEC;
4076 prd.prefixlen = 64;
4077 memcpy(&prd.val, pfx, 8);
4078 pfx += 8;
4079
4080 /* Make EVPN prefix. */
4081 memset(&p, 0, sizeof(struct prefix_evpn));
4082 p.family = AF_EVPN;
4083 p.prefixlen = EVPN_ROUTE_PREFIXLEN;
4084 p.prefix.route_type = BGP_EVPN_IMET_ROUTE;
4085
4086 /* Copy Ethernet Tag */
4087 memcpy(&eth_tag, pfx, 4);
4088 p.prefix.imet_addr.eth_tag = ntohl(eth_tag);
4089 pfx += 4;
4090
4091 /* Get the IP. */
4092 ipaddr_len = *pfx++;
4093 if (ipaddr_len == IPV4_MAX_BITLEN) {
4094 p.prefix.imet_addr.ip.ipa_type = IPADDR_V4;
4095 memcpy(&p.prefix.imet_addr.ip.ip.addr, pfx, IPV4_MAX_BYTELEN);
4096 } else {
4097 flog_err(
4098 EC_BGP_EVPN_ROUTE_INVALID,
4099 "%u:%s - Rx EVPN Type-3 NLRI with unsupported IP address length %d",
4100 peer->bgp->vrf_id, peer->host, ipaddr_len);
4101 return -1;
4102 }
4103
4104 /* Process the route. */
4105 if (attr)
4106 ret = bgp_update(peer, (struct prefix *)&p, addpath_id, attr,
4107 afi, safi, ZEBRA_ROUTE_BGP, BGP_ROUTE_NORMAL,
4108 &prd, NULL, 0, 0, NULL);
4109 else
4110 ret = bgp_withdraw(peer, (struct prefix *)&p, addpath_id, attr,
4111 afi, safi, ZEBRA_ROUTE_BGP, BGP_ROUTE_NORMAL,
4112 &prd, NULL, 0, NULL);
4113 return ret;
4114 }
4115
4116 /*
4117 * Process received EVPN type-4 route (advertise or withdraw).
4118 */
4119 static int process_type4_route(struct peer *peer, afi_t afi, safi_t safi,
4120 struct attr *attr, uint8_t *pfx, int psize,
4121 uint32_t addpath_id)
4122 {
4123 int ret;
4124 esi_t esi;
4125 uint8_t ipaddr_len;
4126 struct in_addr vtep_ip;
4127 struct prefix_rd prd;
4128 struct prefix_evpn p;
4129
4130 /* Type-4 route should be either 23 or 35 bytes
4131 * RD (8), ESI (10), ip-len (1), ip (4 or 16)
4132 */
4133 if (psize != 23 && psize != 35) {
4134 flog_err(EC_BGP_EVPN_ROUTE_INVALID,
4135 "%u:%s - Rx EVPN Type-4 NLRI with invalid length %d",
4136 peer->bgp->vrf_id, peer->host, psize);
4137 return -1;
4138 }
4139
4140 /* Make prefix_rd */
4141 prd.family = AF_UNSPEC;
4142 prd.prefixlen = 64;
4143 memcpy(&prd.val, pfx, 8);
4144 pfx += 8;
4145
4146 /* get the ESI */
4147 memcpy(&esi, pfx, ESI_BYTES);
4148 pfx += ESI_BYTES;
4149
4150
4151 /* Get the IP. */
4152 ipaddr_len = *pfx++;
4153 if (ipaddr_len == IPV4_MAX_BITLEN) {
4154 memcpy(&vtep_ip, pfx, IPV4_MAX_BYTELEN);
4155 } else {
4156 flog_err(
4157 EC_BGP_EVPN_ROUTE_INVALID,
4158 "%u:%s - Rx EVPN Type-4 NLRI with unsupported IP address length %d",
4159 peer->bgp->vrf_id, peer->host, ipaddr_len);
4160 return -1;
4161 }
4162
4163 build_evpn_type4_prefix(&p, &esi, vtep_ip);
4164 /* Process the route. */
4165 if (attr) {
4166 ret = bgp_update(peer, (struct prefix *)&p, addpath_id, attr,
4167 afi, safi, ZEBRA_ROUTE_BGP, BGP_ROUTE_NORMAL,
4168 &prd, NULL, 0, 0, NULL);
4169 } else {
4170 ret = bgp_withdraw(peer, (struct prefix *)&p, addpath_id, attr,
4171 afi, safi, ZEBRA_ROUTE_BGP, BGP_ROUTE_NORMAL,
4172 &prd, NULL, 0, NULL);
4173 }
4174 return ret;
4175 }
4176
4177
4178 /*
4179 * Process received EVPN type-5 route (advertise or withdraw).
4180 */
4181 static int process_type5_route(struct peer *peer, afi_t afi, safi_t safi,
4182 struct attr *attr, uint8_t *pfx, int psize,
4183 uint32_t addpath_id)
4184 {
4185 struct prefix_rd prd;
4186 struct prefix_evpn p;
4187 struct bgp_route_evpn evpn;
4188 uint8_t ippfx_len;
4189 uint32_t eth_tag;
4190 mpls_label_t label; /* holds the VNI as in the packet */
4191 int ret;
4192
4193 /* Type-5 route should be 34 or 58 bytes:
4194 * RD (8), ESI (10), Eth Tag (4), IP len (1), IP (4 or 16),
4195 * GW (4 or 16) and VNI (3).
4196 * Note that the IP and GW should both be IPv4 or both IPv6.
4197 */
4198 if (psize != 34 && psize != 58) {
4199 flog_err(EC_BGP_EVPN_ROUTE_INVALID,
4200 "%u:%s - Rx EVPN Type-5 NLRI with invalid length %d",
4201 peer->bgp->vrf_id, peer->host, psize);
4202 return -1;
4203 }
4204
4205 /* Make prefix_rd */
4206 prd.family = AF_UNSPEC;
4207 prd.prefixlen = 64;
4208 memcpy(&prd.val, pfx, 8);
4209 pfx += 8;
4210
4211 /* Make EVPN prefix. */
4212 memset(&p, 0, sizeof(struct prefix_evpn));
4213 p.family = AF_EVPN;
4214 p.prefixlen = EVPN_ROUTE_PREFIXLEN;
4215 p.prefix.route_type = BGP_EVPN_IP_PREFIX_ROUTE;
4216
4217 /* Additional information outside of prefix - ESI and GW IP */
4218 memset(&evpn, 0, sizeof(evpn));
4219
4220 /* Fetch ESI */
4221 memcpy(&evpn.eth_s_id.val, pfx, 10);
4222 pfx += 10;
4223
4224 /* Fetch Ethernet Tag. */
4225 memcpy(&eth_tag, pfx, 4);
4226 p.prefix.prefix_addr.eth_tag = ntohl(eth_tag);
4227 pfx += 4;
4228
4229 /* Fetch IP prefix length. */
4230 ippfx_len = *pfx++;
4231 if (ippfx_len > IPV6_MAX_BITLEN) {
4232 flog_err(
4233 EC_BGP_EVPN_ROUTE_INVALID,
4234 "%u:%s - Rx EVPN Type-5 NLRI with invalid IP Prefix length %d",
4235 peer->bgp->vrf_id, peer->host, ippfx_len);
4236 return -1;
4237 }
4238 p.prefix.prefix_addr.ip_prefix_length = ippfx_len;
4239
4240 /* Determine IPv4 or IPv6 prefix */
4241 /* Since the address and GW are from the same family, this just becomes
4242 * a simple check on the total size.
4243 */
4244 if (psize == 34) {
4245 SET_IPADDR_V4(&p.prefix.prefix_addr.ip);
4246 memcpy(&p.prefix.prefix_addr.ip.ipaddr_v4, pfx, 4);
4247 pfx += 4;
4248 memcpy(&evpn.gw_ip.ipv4, pfx, 4);
4249 pfx += 4;
4250 } else {
4251 SET_IPADDR_V6(&p.prefix.prefix_addr.ip);
4252 memcpy(&p.prefix.prefix_addr.ip.ipaddr_v6, pfx, 16);
4253 pfx += 16;
4254 memcpy(&evpn.gw_ip.ipv6, pfx, 16);
4255 pfx += 16;
4256 }
4257
4258 /* Get the VNI (in MPLS label field). Stored as bytes here. */
4259 memset(&label, 0, sizeof(label));
4260 memcpy(&label, pfx, BGP_LABEL_BYTES);
4261
4262 /*
4263 * If in future, we are required to access additional fields,
4264 * we MUST increment pfx by BGP_LABEL_BYTES in before reading the next
4265 * field
4266 */
4267
4268 /* Process the route. */
4269 if (attr)
4270 ret = bgp_update(peer, (struct prefix *)&p, addpath_id, attr,
4271 afi, safi, ZEBRA_ROUTE_BGP, BGP_ROUTE_NORMAL,
4272 &prd, &label, 1, 0, &evpn);
4273 else
4274 ret = bgp_withdraw(peer, (struct prefix *)&p, addpath_id, attr,
4275 afi, safi, ZEBRA_ROUTE_BGP, BGP_ROUTE_NORMAL,
4276 &prd, &label, 1, &evpn);
4277
4278 return ret;
4279 }
4280
4281 static void evpn_mpattr_encode_type5(struct stream *s, struct prefix *p,
4282 struct prefix_rd *prd, mpls_label_t *label,
4283 uint32_t num_labels, struct attr *attr)
4284 {
4285 int len;
4286 char temp[16];
4287 struct evpn_addr *p_evpn_p;
4288
4289 memset(&temp, 0, 16);
4290 if (p->family != AF_EVPN)
4291 return;
4292 p_evpn_p = &(p->u.prefix_evpn);
4293
4294 /* len denites the total len of IP and GW-IP in the route
4295 IP and GW-IP have to be both ipv4 or ipv6
4296 */
4297 if (IS_IPADDR_V4(&p_evpn_p->prefix_addr.ip))
4298 len = 8; /* IP and GWIP are both ipv4 */
4299 else
4300 len = 32; /* IP and GWIP are both ipv6 */
4301 /* Prefix contains RD, ESI, EthTag, IP length, IP, GWIP and VNI */
4302 stream_putc(s, 8 + 10 + 4 + 1 + len + 3);
4303 stream_put(s, prd->val, 8);
4304 if (attr)
4305 stream_put(s, &(attr->evpn_overlay.eth_s_id), 10);
4306 else
4307 stream_put(s, &temp, 10);
4308 stream_putl(s, p_evpn_p->prefix_addr.eth_tag);
4309 stream_putc(s, p_evpn_p->prefix_addr.ip_prefix_length);
4310 if (IS_IPADDR_V4(&p_evpn_p->prefix_addr.ip))
4311 stream_put_ipv4(s, p_evpn_p->prefix_addr.ip.ipaddr_v4.s_addr);
4312 else
4313 stream_put(s, &p_evpn_p->prefix_addr.ip.ipaddr_v6, 16);
4314 if (attr) {
4315 if (IS_IPADDR_V4(&p_evpn_p->prefix_addr.ip))
4316 stream_put_ipv4(s,
4317 attr->evpn_overlay.gw_ip.ipv4.s_addr);
4318 else
4319 stream_put(s, &(attr->evpn_overlay.gw_ip.ipv6), 16);
4320 } else {
4321 if (IS_IPADDR_V4(&p_evpn_p->prefix_addr.ip))
4322 stream_put_ipv4(s, 0);
4323 else
4324 stream_put(s, &temp, 16);
4325 }
4326
4327 if (num_labels)
4328 stream_put(s, label, 3);
4329 else
4330 stream_put3(s, 0);
4331 }
4332
4333 /*
4334 * Cleanup specific VNI upon EVPN (advertise-all-vni) being disabled.
4335 */
4336 static void cleanup_vni_on_disable(struct hash_bucket *bucket, struct bgp *bgp)
4337 {
4338 struct bgpevpn *vpn = (struct bgpevpn *)bucket->data;
4339
4340 /* Remove EVPN routes and schedule for processing. */
4341 delete_routes_for_vni(bgp, vpn);
4342
4343 /* Clear "live" flag and see if hash needs to be freed. */
4344 UNSET_FLAG(vpn->flags, VNI_FLAG_LIVE);
4345 if (!is_vni_configured(vpn))
4346 bgp_evpn_free(bgp, vpn);
4347 }
4348
4349 /*
4350 * Free a VNI entry; iterator function called during cleanup.
4351 */
4352 static void free_vni_entry(struct hash_bucket *bucket, struct bgp *bgp)
4353 {
4354 struct bgpevpn *vpn = (struct bgpevpn *)bucket->data;
4355
4356 delete_all_vni_routes(bgp, vpn);
4357 bgp_evpn_free(bgp, vpn);
4358 }
4359
4360 /*
4361 * Derive AUTO import RT for BGP VRF - L3VNI
4362 */
4363 static void evpn_auto_rt_import_add_for_vrf(struct bgp *bgp_vrf)
4364 {
4365 struct bgp *bgp_evpn = NULL;
4366
4367 form_auto_rt(bgp_vrf, bgp_vrf->l3vni, bgp_vrf->vrf_import_rtl);
4368 UNSET_FLAG(bgp_vrf->vrf_flags, BGP_VRF_IMPORT_RT_CFGD);
4369
4370 /* Map RT to VRF */
4371 bgp_evpn = bgp_get_evpn();
4372 if (!bgp_evpn)
4373 return;
4374 bgp_evpn_map_vrf_to_its_rts(bgp_vrf);
4375 }
4376
4377 /*
4378 * Delete AUTO import RT from BGP VRF - L3VNI
4379 */
4380 static void evpn_auto_rt_import_delete_for_vrf(struct bgp *bgp_vrf)
4381 {
4382 evpn_rt_delete_auto(bgp_vrf, bgp_vrf->l3vni, bgp_vrf->vrf_import_rtl);
4383 }
4384
4385 /*
4386 * Derive AUTO export RT for BGP VRF - L3VNI
4387 */
4388 static void evpn_auto_rt_export_add_for_vrf(struct bgp *bgp_vrf)
4389 {
4390 UNSET_FLAG(bgp_vrf->vrf_flags, BGP_VRF_EXPORT_RT_CFGD);
4391 form_auto_rt(bgp_vrf, bgp_vrf->l3vni, bgp_vrf->vrf_export_rtl);
4392 }
4393
4394 /*
4395 * Delete AUTO export RT from BGP VRF - L3VNI
4396 */
4397 static void evpn_auto_rt_export_delete_for_vrf(struct bgp *bgp_vrf)
4398 {
4399 evpn_rt_delete_auto(bgp_vrf, bgp_vrf->l3vni, bgp_vrf->vrf_export_rtl);
4400 }
4401
4402 static void bgp_evpn_handle_export_rt_change_for_vrf(struct bgp *bgp_vrf)
4403 {
4404 struct bgp *bgp_evpn = NULL;
4405 struct listnode *node = NULL;
4406 struct bgpevpn *vpn = NULL;
4407
4408 bgp_evpn = bgp_get_evpn();
4409 if (!bgp_evpn)
4410 return;
4411
4412 /* update all type-5 routes */
4413 update_advertise_vrf_routes(bgp_vrf);
4414
4415 /* update all type-2 routes */
4416 for (ALL_LIST_ELEMENTS_RO(bgp_vrf->l2vnis, node, vpn))
4417 update_routes_for_vni(bgp_evpn, vpn);
4418 }
4419
4420 /*
4421 * Handle autort change for a given VNI.
4422 */
4423 static void update_autort_vni(struct hash_bucket *bucket, struct bgp *bgp)
4424 {
4425 struct bgpevpn *vpn = bucket->data;
4426
4427 if (!is_import_rt_configured(vpn)) {
4428 if (is_vni_live(vpn))
4429 bgp_evpn_uninstall_routes(bgp, vpn);
4430 bgp_evpn_unmap_vni_from_its_rts(bgp, vpn);
4431 list_delete_all_node(vpn->import_rtl);
4432 bgp_evpn_derive_auto_rt_import(bgp, vpn);
4433 if (is_vni_live(vpn))
4434 bgp_evpn_install_routes(bgp, vpn);
4435 }
4436 if (!is_export_rt_configured(vpn)) {
4437 list_delete_all_node(vpn->export_rtl);
4438 bgp_evpn_derive_auto_rt_export(bgp, vpn);
4439 if (is_vni_live(vpn))
4440 bgp_evpn_handle_export_rt_change(bgp, vpn);
4441 }
4442 }
4443
4444 /*
4445 * Public functions.
4446 */
4447
4448 /* withdraw type-5 route corresponding to ip prefix */
4449 void bgp_evpn_withdraw_type5_route(struct bgp *bgp_vrf, struct prefix *p,
4450 afi_t afi, safi_t safi)
4451 {
4452 int ret = 0;
4453 struct prefix_evpn evp;
4454 char buf[PREFIX_STRLEN];
4455
4456 build_type5_prefix_from_ip_prefix(&evp, p);
4457 ret = delete_evpn_type5_route(bgp_vrf, &evp);
4458 if (ret) {
4459 flog_err(
4460 EC_BGP_EVPN_ROUTE_DELETE,
4461 "%u failed to delete type-5 route for prefix %s in vrf %s",
4462 bgp_vrf->vrf_id, prefix2str(p, buf, sizeof(buf)),
4463 vrf_id_to_name(bgp_vrf->vrf_id));
4464 }
4465 }
4466
4467 /* withdraw all type-5 routes for an address family */
4468 void bgp_evpn_withdraw_type5_routes(struct bgp *bgp_vrf, afi_t afi, safi_t safi)
4469 {
4470 struct bgp_table *table = NULL;
4471 struct bgp_node *rn = NULL;
4472 struct bgp_path_info *pi;
4473
4474 table = bgp_vrf->rib[afi][safi];
4475 for (rn = bgp_table_top(table); rn; rn = bgp_route_next(rn)) {
4476 /* Only care about "selected" routes. Also ensure that
4477 * these are routes that are injectable into EVPN.
4478 */
4479 /* TODO: Support for AddPath for EVPN. */
4480 for (pi = bgp_node_get_bgp_path_info(rn); pi; pi = pi->next) {
4481 if (CHECK_FLAG(pi->flags, BGP_PATH_SELECTED)
4482 && is_route_injectable_into_evpn(pi)) {
4483 bgp_evpn_withdraw_type5_route(bgp_vrf, &rn->p,
4484 afi, safi);
4485 break;
4486 }
4487 }
4488 }
4489 }
4490
4491 /*
4492 * evpn - enable advertisement of default g/w
4493 */
4494 void bgp_evpn_install_uninstall_default_route(struct bgp *bgp_vrf, afi_t afi,
4495 safi_t safi, bool add)
4496 {
4497 struct prefix ip_prefix;
4498
4499 /* form the default prefix 0.0.0.0/0 */
4500 memset(&ip_prefix, 0, sizeof(struct prefix));
4501 ip_prefix.family = afi2family(afi);
4502
4503 if (add) {
4504 bgp_evpn_advertise_type5_route(bgp_vrf, &ip_prefix,
4505 NULL, afi, safi);
4506 } else {
4507 bgp_evpn_withdraw_type5_route(bgp_vrf, &ip_prefix,
4508 afi, safi);
4509 }
4510 }
4511
4512
4513 /*
4514 * Advertise IP prefix as type-5 route. The afi/safi and src_attr passed
4515 * to this function correspond to those of the source IP prefix (best
4516 * path in the case of the attr. In the case of a local prefix (when we
4517 * are advertising local subnets), the src_attr will be NULL.
4518 */
4519 void bgp_evpn_advertise_type5_route(struct bgp *bgp_vrf, struct prefix *p,
4520 struct attr *src_attr, afi_t afi,
4521 safi_t safi)
4522 {
4523 int ret = 0;
4524 struct prefix_evpn evp;
4525 char buf[PREFIX_STRLEN];
4526
4527 build_type5_prefix_from_ip_prefix(&evp, p);
4528 ret = update_evpn_type5_route(bgp_vrf, &evp, src_attr);
4529 if (ret)
4530 flog_err(EC_BGP_EVPN_ROUTE_CREATE,
4531 "%u: Failed to create type-5 route for prefix %s",
4532 bgp_vrf->vrf_id, prefix2str(p, buf, sizeof(buf)));
4533 }
4534
4535 /* Inject all prefixes of a particular address-family (currently, IPv4 or
4536 * IPv6 unicast) into EVPN as type-5 routes. This is invoked when the
4537 * advertisement is enabled.
4538 */
4539 void bgp_evpn_advertise_type5_routes(struct bgp *bgp_vrf, afi_t afi,
4540 safi_t safi)
4541 {
4542 struct bgp_table *table = NULL;
4543 struct bgp_node *rn = NULL;
4544 struct bgp_path_info *pi;
4545
4546 table = bgp_vrf->rib[afi][safi];
4547 for (rn = bgp_table_top(table); rn; rn = bgp_route_next(rn)) {
4548 /* Need to identify the "selected" route entry to use its
4549 * attribute. Also, ensure that the route is injectable
4550 * into EVPN.
4551 * TODO: Support for AddPath for EVPN.
4552 */
4553 for (pi = bgp_node_get_bgp_path_info(rn); pi; pi = pi->next) {
4554 if (CHECK_FLAG(pi->flags, BGP_PATH_SELECTED)
4555 && is_route_injectable_into_evpn(pi)) {
4556
4557 /* apply the route-map */
4558 if (bgp_vrf->adv_cmd_rmap[afi][safi].map) {
4559 route_map_result_t ret;
4560
4561 ret = route_map_apply(
4562 bgp_vrf->adv_cmd_rmap[afi][safi]
4563 .map,
4564 &rn->p, RMAP_BGP, pi);
4565 if (ret == RMAP_DENYMATCH)
4566 continue;
4567 }
4568 bgp_evpn_advertise_type5_route(
4569 bgp_vrf, &rn->p, pi->attr, afi, safi);
4570 break;
4571 }
4572 }
4573 }
4574 }
4575
4576 void evpn_rt_delete_auto(struct bgp *bgp, vni_t vni, struct list *rtl)
4577 {
4578 struct listnode *node, *nnode, *node_to_del;
4579 struct ecommunity *ecom, *ecom_auto;
4580 struct ecommunity_val eval;
4581
4582 if (bgp->advertise_autort_rfc8365)
4583 vni |= EVPN_AUTORT_VXLAN;
4584 encode_route_target_as((bgp->as & 0xFFFF), vni, &eval);
4585
4586 ecom_auto = ecommunity_new();
4587 ecommunity_add_val(ecom_auto, &eval);
4588 node_to_del = NULL;
4589
4590 for (ALL_LIST_ELEMENTS(rtl, node, nnode, ecom)) {
4591 if (ecommunity_match(ecom, ecom_auto)) {
4592 ecommunity_free(&ecom);
4593 node_to_del = node;
4594 }
4595 }
4596
4597 if (node_to_del)
4598 list_delete_node(rtl, node_to_del);
4599
4600 ecommunity_free(&ecom_auto);
4601 }
4602
4603 void bgp_evpn_configure_import_rt_for_vrf(struct bgp *bgp_vrf,
4604 struct ecommunity *ecomadd)
4605 {
4606 /* uninstall routes from vrf */
4607 if (is_l3vni_live(bgp_vrf))
4608 uninstall_routes_for_vrf(bgp_vrf);
4609
4610 /* Cleanup the RT to VRF mapping */
4611 bgp_evpn_unmap_vrf_from_its_rts(bgp_vrf);
4612
4613 /* Remove auto generated RT */
4614 evpn_auto_rt_import_delete_for_vrf(bgp_vrf);
4615
4616 /* Add the newly configured RT to RT list */
4617 listnode_add_sort(bgp_vrf->vrf_import_rtl, ecomadd);
4618 SET_FLAG(bgp_vrf->vrf_flags, BGP_VRF_IMPORT_RT_CFGD);
4619
4620 /* map VRF to its RTs and install routes matching the new RTs */
4621 if (is_l3vni_live(bgp_vrf)) {
4622 bgp_evpn_map_vrf_to_its_rts(bgp_vrf);
4623 install_routes_for_vrf(bgp_vrf);
4624 }
4625 }
4626
4627 void bgp_evpn_unconfigure_import_rt_for_vrf(struct bgp *bgp_vrf,
4628 struct ecommunity *ecomdel)
4629 {
4630 struct listnode *node = NULL, *nnode = NULL, *node_to_del = NULL;
4631 struct ecommunity *ecom = NULL;
4632
4633 /* uninstall routes from vrf */
4634 if (is_l3vni_live(bgp_vrf))
4635 uninstall_routes_for_vrf(bgp_vrf);
4636
4637 /* Cleanup the RT to VRF mapping */
4638 bgp_evpn_unmap_vrf_from_its_rts(bgp_vrf);
4639
4640 /* remove the RT from the RT list */
4641 for (ALL_LIST_ELEMENTS(bgp_vrf->vrf_import_rtl, node, nnode, ecom)) {
4642 if (ecommunity_match(ecom, ecomdel)) {
4643 ecommunity_free(&ecom);
4644 node_to_del = node;
4645 break;
4646 }
4647 }
4648
4649 if (node_to_del)
4650 list_delete_node(bgp_vrf->vrf_import_rtl, node_to_del);
4651
4652 assert(bgp_vrf->vrf_import_rtl);
4653 /* fallback to auto import rt, if this was the last RT */
4654 if (bgp_vrf->vrf_import_rtl && list_isempty(bgp_vrf->vrf_import_rtl)) {
4655 UNSET_FLAG(bgp_vrf->vrf_flags, BGP_VRF_IMPORT_RT_CFGD);
4656 evpn_auto_rt_import_add_for_vrf(bgp_vrf);
4657 }
4658
4659 /* map VRFs to its RTs and install routes matching this new RT */
4660 if (is_l3vni_live(bgp_vrf)) {
4661 bgp_evpn_map_vrf_to_its_rts(bgp_vrf);
4662 install_routes_for_vrf(bgp_vrf);
4663 }
4664 }
4665
4666 void bgp_evpn_configure_export_rt_for_vrf(struct bgp *bgp_vrf,
4667 struct ecommunity *ecomadd)
4668 {
4669 /* remove auto-generated RT */
4670 evpn_auto_rt_export_delete_for_vrf(bgp_vrf);
4671
4672 /* Add the new RT to the RT list */
4673 listnode_add_sort(bgp_vrf->vrf_export_rtl, ecomadd);
4674 SET_FLAG(bgp_vrf->vrf_flags, BGP_VRF_EXPORT_RT_CFGD);
4675
4676 bgp_evpn_handle_export_rt_change_for_vrf(bgp_vrf);
4677 }
4678
4679 void bgp_evpn_unconfigure_export_rt_for_vrf(struct bgp *bgp_vrf,
4680 struct ecommunity *ecomdel)
4681 {
4682 struct listnode *node = NULL, *nnode = NULL, *node_to_del = NULL;
4683 struct ecommunity *ecom = NULL;
4684
4685 /* Remove the RT from the RT list */
4686 for (ALL_LIST_ELEMENTS(bgp_vrf->vrf_export_rtl, node, nnode, ecom)) {
4687 if (ecommunity_match(ecom, ecomdel)) {
4688 ecommunity_free(&ecom);
4689 node_to_del = node;
4690 break;
4691 }
4692 }
4693
4694 if (node_to_del)
4695 list_delete_node(bgp_vrf->vrf_export_rtl, node_to_del);
4696
4697 /*
4698 * Temporary assert to make SA happy.
4699 * The ALL_LIST_ELEMENTS macro above has a NULL check
4700 * which means that SA is going to complain about
4701 * the list_isempty call, which doesn't NULL check.
4702 * So until we get this situation cleaned up, here
4703 * we are.
4704 */
4705 assert(bgp_vrf->vrf_export_rtl);
4706
4707 /* fall back to auto-generated RT if this was the last RT */
4708 if (list_isempty(bgp_vrf->vrf_export_rtl)) {
4709 UNSET_FLAG(bgp_vrf->vrf_flags, BGP_VRF_EXPORT_RT_CFGD);
4710 evpn_auto_rt_export_add_for_vrf(bgp_vrf);
4711 }
4712
4713 bgp_evpn_handle_export_rt_change_for_vrf(bgp_vrf);
4714 }
4715
4716 /*
4717 * Handle change to BGP router id. This is invoked twice by the change
4718 * handler, first before the router id has been changed and then after
4719 * the router id has been changed. The first invocation will result in
4720 * local routes for all VNIs/VRF being deleted and withdrawn and the next
4721 * will result in the routes being re-advertised.
4722 */
4723 void bgp_evpn_handle_router_id_update(struct bgp *bgp, int withdraw)
4724 {
4725 struct listnode *node;
4726 struct bgp *bgp_vrf;
4727
4728 if (withdraw) {
4729
4730 /* delete and withdraw all the type-5 routes
4731 stored in the global table for this vrf
4732 */
4733 withdraw_router_id_vrf(bgp);
4734
4735 /* delete all the VNI routes (type-2/type-3) routes for all the
4736 * L2-VNIs
4737 */
4738 hash_iterate(bgp->vnihash,
4739 (void (*)(struct hash_bucket *,
4740 void *))withdraw_router_id_vni,
4741 bgp);
4742
4743 if (bgp->inst_type == BGP_INSTANCE_TYPE_DEFAULT) {
4744 for (ALL_LIST_ELEMENTS_RO(bm->bgp, node, bgp_vrf)) {
4745 if (bgp_vrf->evpn_info->advertise_pip &&
4746 (bgp_vrf->evpn_info->pip_ip_static.s_addr
4747 == INADDR_ANY))
4748 bgp_vrf->evpn_info->pip_ip.s_addr
4749 = INADDR_ANY;
4750 }
4751 }
4752 } else {
4753
4754 /* Assign new default instance router-id */
4755 if (bgp->inst_type == BGP_INSTANCE_TYPE_DEFAULT) {
4756 for (ALL_LIST_ELEMENTS_RO(bm->bgp, node, bgp_vrf)) {
4757 if (bgp_vrf->evpn_info->advertise_pip &&
4758 (bgp_vrf->evpn_info->pip_ip_static.s_addr
4759 == INADDR_ANY)) {
4760 bgp_vrf->evpn_info->pip_ip =
4761 bgp->router_id;
4762 /* advertise type-5 routes with
4763 * new nexthop
4764 */
4765 update_advertise_vrf_routes(bgp_vrf);
4766 }
4767 }
4768 }
4769
4770 /* advertise all routes in the vrf as type-5 routes with the new
4771 * RD
4772 */
4773 update_router_id_vrf(bgp);
4774
4775 /* advertise all the VNI routes (type-2/type-3) routes with the
4776 * new RD
4777 */
4778 hash_iterate(bgp->vnihash,
4779 (void (*)(struct hash_bucket *,
4780 void *))update_router_id_vni,
4781 bgp);
4782 }
4783 }
4784
4785 /*
4786 * Handle change to auto-RT algorithm - update and advertise local routes.
4787 */
4788 void bgp_evpn_handle_autort_change(struct bgp *bgp)
4789 {
4790 hash_iterate(bgp->vnihash,
4791 (void (*)(struct hash_bucket *,
4792 void*))update_autort_vni,
4793 bgp);
4794 }
4795
4796 /*
4797 * Handle change to export RT - update and advertise local routes.
4798 */
4799 int bgp_evpn_handle_export_rt_change(struct bgp *bgp, struct bgpevpn *vpn)
4800 {
4801 return update_routes_for_vni(bgp, vpn);
4802 }
4803
4804 void bgp_evpn_handle_vrf_rd_change(struct bgp *bgp_vrf, int withdraw)
4805 {
4806 if (withdraw)
4807 delete_withdraw_vrf_routes(bgp_vrf);
4808 else
4809 update_advertise_vrf_routes(bgp_vrf);
4810 }
4811
4812 /*
4813 * Handle change to RD. This is invoked twice by the change handler,
4814 * first before the RD has been changed and then after the RD has
4815 * been changed. The first invocation will result in local routes
4816 * of this VNI being deleted and withdrawn and the next will result
4817 * in the routes being re-advertised.
4818 */
4819 void bgp_evpn_handle_rd_change(struct bgp *bgp, struct bgpevpn *vpn,
4820 int withdraw)
4821 {
4822 if (withdraw)
4823 delete_withdraw_vni_routes(bgp, vpn);
4824 else
4825 update_advertise_vni_routes(bgp, vpn);
4826 }
4827
4828 /*
4829 * Install routes for this VNI. Invoked upon change to Import RT.
4830 */
4831 int bgp_evpn_install_routes(struct bgp *bgp, struct bgpevpn *vpn)
4832 {
4833 return install_routes_for_vni(bgp, vpn);
4834 }
4835
4836 /*
4837 * Uninstall all routes installed for this VNI. Invoked upon change
4838 * to Import RT.
4839 */
4840 int bgp_evpn_uninstall_routes(struct bgp *bgp, struct bgpevpn *vpn)
4841 {
4842 return uninstall_routes_for_vni(bgp, vpn);
4843 }
4844
4845 /*
4846 * TODO: Hardcoded for a maximum of 2 VNIs right now
4847 */
4848 char *bgp_evpn_label2str(mpls_label_t *label, uint32_t num_labels, char *buf,
4849 int len)
4850 {
4851 vni_t vni1, vni2;
4852
4853 vni1 = label2vni(label);
4854 if (num_labels == 2) {
4855 vni2 = label2vni(label + 1);
4856 snprintf(buf, len, "%u/%u", vni1, vni2);
4857 } else
4858 snprintf(buf, len, "%u", vni1);
4859 return buf;
4860 }
4861
4862 /*
4863 * Function to convert evpn route to json format.
4864 * NOTE: We don't use prefix2str as the output here is a bit different.
4865 */
4866 void bgp_evpn_route2json(struct prefix_evpn *p, json_object *json)
4867 {
4868 char buf1[ETHER_ADDR_STRLEN];
4869 char buf2[PREFIX2STR_BUFFER];
4870 uint8_t family;
4871 uint8_t prefixlen;
4872
4873 if (!json)
4874 return;
4875
4876 json_object_int_add(json, "routeType", p->prefix.route_type);
4877
4878 switch (p->prefix.route_type) {
4879 case BGP_EVPN_MAC_IP_ROUTE:
4880 json_object_int_add(json, "ethTag",
4881 p->prefix.macip_addr.eth_tag);
4882 json_object_int_add(json, "macLen", 8 * ETH_ALEN);
4883 json_object_string_add(json, "mac",
4884 prefix_mac2str(&p->prefix.macip_addr.mac, buf1,
4885 sizeof(buf1)));
4886
4887 if (!is_evpn_prefix_ipaddr_none(p)) {
4888 family = is_evpn_prefix_ipaddr_v4(p) ? AF_INET :
4889 AF_INET6;
4890 prefixlen = (family == AF_INET) ?
4891 IPV4_MAX_BITLEN : IPV6_MAX_BITLEN;
4892 inet_ntop(family, &p->prefix.macip_addr.ip.ip.addr,
4893 buf2, PREFIX2STR_BUFFER);
4894 json_object_int_add(json, "ipLen", prefixlen);
4895 json_object_string_add(json, "ip", buf2);
4896 }
4897 break;
4898
4899 case BGP_EVPN_IMET_ROUTE:
4900 json_object_int_add(json, "ethTag",
4901 p->prefix.imet_addr.eth_tag);
4902 family = is_evpn_prefix_ipaddr_v4(p) ? AF_INET : AF_INET6;
4903 prefixlen = (family == AF_INET) ? IPV4_MAX_BITLEN :
4904 IPV6_MAX_BITLEN;
4905 inet_ntop(family, &p->prefix.imet_addr.ip.ip.addr, buf2,
4906 PREFIX2STR_BUFFER);
4907 json_object_int_add(json, "ipLen", prefixlen);
4908 json_object_string_add(json, "ip", buf2);
4909 break;
4910
4911 case BGP_EVPN_IP_PREFIX_ROUTE:
4912 json_object_int_add(json, "ethTag",
4913 p->prefix.prefix_addr.eth_tag);
4914 family = is_evpn_prefix_ipaddr_v4(p) ? AF_INET : AF_INET6;
4915 inet_ntop(family, &p->prefix.prefix_addr.ip.ip.addr,
4916 buf2, sizeof(buf2));
4917 json_object_int_add(json, "ipLen",
4918 p->prefix.prefix_addr.ip_prefix_length);
4919 json_object_string_add(json, "ip", buf2);
4920 break;
4921
4922 default:
4923 break;
4924 }
4925 }
4926
4927 /*
4928 * Function to convert evpn route to string.
4929 * NOTE: We don't use prefix2str as the output here is a bit different.
4930 */
4931 char *bgp_evpn_route2str(struct prefix_evpn *p, char *buf, int len)
4932 {
4933 char buf1[ETHER_ADDR_STRLEN];
4934 char buf2[PREFIX2STR_BUFFER];
4935 char buf3[ESI_STR_LEN];
4936
4937 if (p->prefix.route_type == BGP_EVPN_IMET_ROUTE) {
4938 snprintf(buf, len, "[%d]:[%d]:[%d]:[%s]", p->prefix.route_type,
4939 p->prefix.imet_addr.eth_tag,
4940 is_evpn_prefix_ipaddr_v4(p) ? IPV4_MAX_BITLEN
4941 : IPV6_MAX_BITLEN,
4942 inet_ntoa(p->prefix.imet_addr.ip.ipaddr_v4));
4943 } else if (p->prefix.route_type == BGP_EVPN_MAC_IP_ROUTE) {
4944 if (is_evpn_prefix_ipaddr_none(p))
4945 snprintf(buf, len, "[%d]:[%d]:[%d]:[%s]",
4946 p->prefix.route_type,
4947 p->prefix.macip_addr.eth_tag,
4948 8 * ETH_ALEN,
4949 prefix_mac2str(&p->prefix.macip_addr.mac, buf1,
4950 sizeof(buf1)));
4951 else {
4952 uint8_t family;
4953
4954 family = is_evpn_prefix_ipaddr_v4(p) ? AF_INET
4955 : AF_INET6;
4956 snprintf(buf, len, "[%d]:[%d]:[%d]:[%s]:[%d]:[%s]",
4957 p->prefix.route_type,
4958 p->prefix.macip_addr.eth_tag,
4959 8 * ETH_ALEN,
4960 prefix_mac2str(&p->prefix.macip_addr.mac, buf1,
4961 sizeof(buf1)),
4962 family == AF_INET ? IPV4_MAX_BITLEN
4963 : IPV6_MAX_BITLEN,
4964 inet_ntop(family,
4965 &p->prefix.macip_addr.ip.ip.addr,
4966 buf2,
4967 PREFIX2STR_BUFFER));
4968 }
4969 } else if (p->prefix.route_type == BGP_EVPN_IP_PREFIX_ROUTE) {
4970 snprintf(buf, len, "[%d]:[%d]:[%d]:[%s]",
4971 p->prefix.route_type,
4972 p->prefix.prefix_addr.eth_tag,
4973 p->prefix.prefix_addr.ip_prefix_length,
4974 is_evpn_prefix_ipaddr_v4(p)
4975 ? inet_ntoa(p->prefix.prefix_addr.ip.ipaddr_v4)
4976 : inet6_ntoa(p->prefix.prefix_addr.ip.ipaddr_v6));
4977 } else if (p->prefix.route_type == BGP_EVPN_ES_ROUTE) {
4978 snprintf(buf, len, "[%d]:[%s]:[%d]:[%s]",
4979 p->prefix.route_type,
4980 esi_to_str(&p->prefix.es_addr.esi, buf3, sizeof(buf3)),
4981 is_evpn_prefix_ipaddr_v4(p) ? IPV4_MAX_BITLEN
4982 : IPV6_MAX_BITLEN,
4983 inet_ntoa(p->prefix.es_addr.ip.ipaddr_v4));
4984 } else {
4985 /* For EVPN route types not supported yet. */
4986 snprintf(buf, len, "(unsupported route type %d)",
4987 p->prefix.route_type);
4988 }
4989
4990 return (buf);
4991 }
4992
4993 /*
4994 * Encode EVPN prefix in Update (MP_REACH)
4995 */
4996 void bgp_evpn_encode_prefix(struct stream *s, struct prefix *p,
4997 struct prefix_rd *prd, mpls_label_t *label,
4998 uint32_t num_labels, struct attr *attr,
4999 int addpath_encode, uint32_t addpath_tx_id)
5000 {
5001 struct prefix_evpn *evp = (struct prefix_evpn *)p;
5002 int len, ipa_len = 0;
5003
5004 if (addpath_encode)
5005 stream_putl(s, addpath_tx_id);
5006
5007 /* Route type */
5008 stream_putc(s, evp->prefix.route_type);
5009
5010 switch (evp->prefix.route_type) {
5011 case BGP_EVPN_MAC_IP_ROUTE:
5012 if (is_evpn_prefix_ipaddr_v4(evp))
5013 ipa_len = IPV4_MAX_BYTELEN;
5014 else if (is_evpn_prefix_ipaddr_v6(evp))
5015 ipa_len = IPV6_MAX_BYTELEN;
5016 /* RD, ESI, EthTag, MAC+len, IP len, [IP], 1 VNI */
5017 len = 8 + 10 + 4 + 1 + 6 + 1 + ipa_len + 3;
5018 if (ipa_len && num_labels > 1) /* There are 2 VNIs */
5019 len += 3;
5020 stream_putc(s, len);
5021 stream_put(s, prd->val, 8); /* RD */
5022 if (attr)
5023 stream_put(s, &attr->evpn_overlay.eth_s_id, ESI_LEN);
5024 else
5025 stream_put(s, 0, 10);
5026 stream_putl(s, evp->prefix.macip_addr.eth_tag); /* Ethernet Tag ID */
5027 stream_putc(s, 8 * ETH_ALEN); /* Mac Addr Len - bits */
5028 stream_put(s, evp->prefix.macip_addr.mac.octet, 6); /* Mac Addr */
5029 stream_putc(s, 8 * ipa_len); /* IP address Length */
5030 if (ipa_len) /* IP */
5031 stream_put(s, &evp->prefix.macip_addr.ip.ip.addr,
5032 ipa_len);
5033 /* 1st label is the L2 VNI */
5034 stream_put(s, label, BGP_LABEL_BYTES);
5035 /* Include 2nd label (L3 VNI) if advertising MAC+IP */
5036 if (ipa_len && num_labels > 1)
5037 stream_put(s, label + 1, BGP_LABEL_BYTES);
5038 break;
5039
5040 case BGP_EVPN_IMET_ROUTE:
5041 stream_putc(s, 17); // TODO: length - assumes IPv4 address
5042 stream_put(s, prd->val, 8); /* RD */
5043 stream_putl(s, evp->prefix.imet_addr.eth_tag); /* Ethernet Tag ID */
5044 stream_putc(s, IPV4_MAX_BITLEN); /* IP address Length - bits */
5045 /* Originating Router's IP Addr */
5046 stream_put_in_addr(s, &evp->prefix.imet_addr.ip.ipaddr_v4);
5047 break;
5048
5049 case BGP_EVPN_ES_ROUTE:
5050 stream_putc(s, 23); /* TODO: length: assumes ipv4 VTEP */
5051 stream_put(s, prd->val, 8); /* RD */
5052 stream_put(s, evp->prefix.es_addr.esi.val, 10); /* ESI */
5053 stream_putc(s, IPV4_MAX_BITLEN); /* IP address Length - bits */
5054 /* VTEP IP */
5055 stream_put_in_addr(s, &evp->prefix.es_addr.ip.ipaddr_v4);
5056 break;
5057
5058 case BGP_EVPN_IP_PREFIX_ROUTE:
5059 /* TODO: AddPath support. */
5060 evpn_mpattr_encode_type5(s, p, prd, label, num_labels, attr);
5061 break;
5062
5063 default:
5064 break;
5065 }
5066 }
5067
5068 int bgp_nlri_parse_evpn(struct peer *peer, struct attr *attr,
5069 struct bgp_nlri *packet, int withdraw)
5070 {
5071 uint8_t *pnt;
5072 uint8_t *lim;
5073 afi_t afi;
5074 safi_t safi;
5075 uint32_t addpath_id;
5076 int addpath_encoded;
5077 int psize = 0;
5078 uint8_t rtype;
5079 struct prefix p;
5080
5081 /* Start processing the NLRI - there may be multiple in the MP_REACH */
5082 pnt = packet->nlri;
5083 lim = pnt + packet->length;
5084 afi = packet->afi;
5085 safi = packet->safi;
5086 addpath_id = 0;
5087
5088 addpath_encoded =
5089 (CHECK_FLAG(peer->af_cap[afi][safi], PEER_CAP_ADDPATH_AF_RX_ADV)
5090 && CHECK_FLAG(peer->af_cap[afi][safi],
5091 PEER_CAP_ADDPATH_AF_TX_RCV));
5092
5093 for (; pnt < lim; pnt += psize) {
5094 /* Clear prefix structure. */
5095 memset(&p, 0, sizeof(struct prefix));
5096
5097 /* Deal with path-id if AddPath is supported. */
5098 if (addpath_encoded) {
5099 /* When packet overflow occurs return immediately. */
5100 if (pnt + BGP_ADDPATH_ID_LEN > lim)
5101 return BGP_NLRI_PARSE_ERROR_PACKET_OVERFLOW;
5102
5103 memcpy(&addpath_id, pnt, BGP_ADDPATH_ID_LEN);
5104 addpath_id = ntohl(addpath_id);
5105 pnt += BGP_ADDPATH_ID_LEN;
5106 }
5107
5108 /* All EVPN NLRI types start with type and length. */
5109 if (pnt + 2 > lim)
5110 return BGP_NLRI_PARSE_ERROR_EVPN_MISSING_TYPE;
5111
5112 rtype = *pnt++;
5113 psize = *pnt++;
5114
5115 /* When packet overflow occur return immediately. */
5116 if (pnt + psize > lim)
5117 return BGP_NLRI_PARSE_ERROR_PACKET_OVERFLOW;
5118
5119 switch (rtype) {
5120 case BGP_EVPN_MAC_IP_ROUTE:
5121 if (process_type2_route(peer, afi, safi,
5122 withdraw ? NULL : attr, pnt,
5123 psize, addpath_id)) {
5124 flog_err(
5125 EC_BGP_EVPN_FAIL,
5126 "%u:%s - Error in processing EVPN type-2 NLRI size %d",
5127 peer->bgp->vrf_id, peer->host, psize);
5128 return BGP_NLRI_PARSE_ERROR_EVPN_TYPE2_SIZE;
5129 }
5130 break;
5131
5132 case BGP_EVPN_IMET_ROUTE:
5133 if (process_type3_route(peer, afi, safi,
5134 withdraw ? NULL : attr, pnt,
5135 psize, addpath_id)) {
5136 flog_err(
5137 EC_BGP_PKT_PROCESS,
5138 "%u:%s - Error in processing EVPN type-3 NLRI size %d",
5139 peer->bgp->vrf_id, peer->host, psize);
5140 return BGP_NLRI_PARSE_ERROR_EVPN_TYPE3_SIZE;
5141 }
5142 break;
5143
5144 case BGP_EVPN_ES_ROUTE:
5145 if (process_type4_route(peer, afi, safi,
5146 withdraw ? NULL : attr, pnt,
5147 psize, addpath_id)) {
5148 flog_err(
5149 EC_BGP_PKT_PROCESS,
5150 "%u:%s - Error in processing EVPN type-4 NLRI size %d",
5151 peer->bgp->vrf_id, peer->host, psize);
5152 return BGP_NLRI_PARSE_ERROR_EVPN_TYPE4_SIZE;
5153 }
5154 break;
5155
5156 case BGP_EVPN_IP_PREFIX_ROUTE:
5157 if (process_type5_route(peer, afi, safi,
5158 withdraw ? NULL : attr, pnt,
5159 psize, addpath_id)) {
5160 flog_err(
5161 EC_BGP_PKT_PROCESS,
5162 "%u:%s - Error in processing EVPN type-5 NLRI size %d",
5163 peer->bgp->vrf_id, peer->host, psize);
5164 return BGP_NLRI_PARSE_ERROR_EVPN_TYPE5_SIZE;
5165 }
5166 break;
5167
5168 default:
5169 break;
5170 }
5171 }
5172
5173 /* Packet length consistency check. */
5174 if (pnt != lim)
5175 return BGP_NLRI_PARSE_ERROR_PACKET_LENGTH;
5176
5177 return BGP_NLRI_PARSE_OK;
5178 }
5179
5180 /*
5181 * Map the RTs (configured or automatically derived) of a VRF to the VRF.
5182 * The mapping will be used during route processing.
5183 * bgp_def: default bgp instance
5184 * bgp_vrf: specific bgp vrf instance on which RT is configured
5185 */
5186 void bgp_evpn_map_vrf_to_its_rts(struct bgp *bgp_vrf)
5187 {
5188 int i = 0;
5189 struct ecommunity_val *eval = NULL;
5190 struct listnode *node = NULL, *nnode = NULL;
5191 struct ecommunity *ecom = NULL;
5192
5193 for (ALL_LIST_ELEMENTS(bgp_vrf->vrf_import_rtl, node, nnode, ecom)) {
5194 for (i = 0; i < ecom->size; i++) {
5195 eval = (struct ecommunity_val *)(ecom->val
5196 + (i
5197 * ECOMMUNITY_SIZE));
5198 map_vrf_to_rt(bgp_vrf, eval);
5199 }
5200 }
5201 }
5202
5203 /*
5204 * Unmap the RTs (configured or automatically derived) of a VRF from the VRF.
5205 */
5206 void bgp_evpn_unmap_vrf_from_its_rts(struct bgp *bgp_vrf)
5207 {
5208 int i;
5209 struct ecommunity_val *eval;
5210 struct listnode *node, *nnode;
5211 struct ecommunity *ecom;
5212
5213 for (ALL_LIST_ELEMENTS(bgp_vrf->vrf_import_rtl, node, nnode, ecom)) {
5214 for (i = 0; i < ecom->size; i++) {
5215 struct vrf_irt_node *irt;
5216 struct ecommunity_val eval_tmp;
5217
5218 eval = (struct ecommunity_val *)(ecom->val
5219 + (i
5220 * ECOMMUNITY_SIZE));
5221 /* If using "automatic" RT, we only care about the
5222 * local-admin sub-field.
5223 * This is to facilitate using VNI as the RT for EBGP
5224 * peering too.
5225 */
5226 memcpy(&eval_tmp, eval, ECOMMUNITY_SIZE);
5227 if (!CHECK_FLAG(bgp_vrf->vrf_flags,
5228 BGP_VRF_IMPORT_RT_CFGD))
5229 mask_ecom_global_admin(&eval_tmp, eval);
5230
5231 irt = lookup_vrf_import_rt(&eval_tmp);
5232 if (irt)
5233 unmap_vrf_from_rt(bgp_vrf, irt);
5234 }
5235 }
5236 }
5237
5238
5239 /*
5240 * Map the RTs (configured or automatically derived) of a VNI to the VNI.
5241 * The mapping will be used during route processing.
5242 */
5243 void bgp_evpn_map_vni_to_its_rts(struct bgp *bgp, struct bgpevpn *vpn)
5244 {
5245 int i;
5246 struct ecommunity_val *eval;
5247 struct listnode *node, *nnode;
5248 struct ecommunity *ecom;
5249
5250 for (ALL_LIST_ELEMENTS(vpn->import_rtl, node, nnode, ecom)) {
5251 for (i = 0; i < ecom->size; i++) {
5252 eval = (struct ecommunity_val *)(ecom->val
5253 + (i
5254 * ECOMMUNITY_SIZE));
5255 map_vni_to_rt(bgp, vpn, eval);
5256 }
5257 }
5258 }
5259
5260 /*
5261 * Unmap the RTs (configured or automatically derived) of a VNI from the VNI.
5262 */
5263 void bgp_evpn_unmap_vni_from_its_rts(struct bgp *bgp, struct bgpevpn *vpn)
5264 {
5265 int i;
5266 struct ecommunity_val *eval;
5267 struct listnode *node, *nnode;
5268 struct ecommunity *ecom;
5269
5270 for (ALL_LIST_ELEMENTS(vpn->import_rtl, node, nnode, ecom)) {
5271 for (i = 0; i < ecom->size; i++) {
5272 struct irt_node *irt;
5273 struct ecommunity_val eval_tmp;
5274
5275 eval = (struct ecommunity_val *)(ecom->val
5276 + (i
5277 * ECOMMUNITY_SIZE));
5278 /* If using "automatic" RT, we only care about the
5279 * local-admin sub-field.
5280 * This is to facilitate using VNI as the RT for EBGP
5281 * peering too.
5282 */
5283 memcpy(&eval_tmp, eval, ECOMMUNITY_SIZE);
5284 if (!is_import_rt_configured(vpn))
5285 mask_ecom_global_admin(&eval_tmp, eval);
5286
5287 irt = lookup_import_rt(bgp, &eval_tmp);
5288 if (irt)
5289 unmap_vni_from_rt(bgp, vpn, irt);
5290 }
5291 }
5292 }
5293
5294 /*
5295 * Derive Import RT automatically for VNI and map VNI to RT.
5296 * The mapping will be used during route processing.
5297 */
5298 void bgp_evpn_derive_auto_rt_import(struct bgp *bgp, struct bgpevpn *vpn)
5299 {
5300 form_auto_rt(bgp, vpn->vni, vpn->import_rtl);
5301 UNSET_FLAG(vpn->flags, VNI_FLAG_IMPRT_CFGD);
5302
5303 /* Map RT to VNI */
5304 bgp_evpn_map_vni_to_its_rts(bgp, vpn);
5305 }
5306
5307 /*
5308 * Derive Export RT automatically for VNI.
5309 */
5310 void bgp_evpn_derive_auto_rt_export(struct bgp *bgp, struct bgpevpn *vpn)
5311 {
5312 form_auto_rt(bgp, vpn->vni, vpn->export_rtl);
5313 UNSET_FLAG(vpn->flags, VNI_FLAG_EXPRT_CFGD);
5314 }
5315
5316 /*
5317 * Derive RD automatically for VNI using passed information - it
5318 * is of the form RouterId:unique-id-for-vni.
5319 */
5320 void bgp_evpn_derive_auto_rd_for_vrf(struct bgp *bgp)
5321 {
5322 if (is_vrf_rd_configured(bgp))
5323 return;
5324
5325 form_auto_rd(bgp->router_id, bgp->vrf_rd_id, &bgp->vrf_prd);
5326 }
5327
5328 /*
5329 * Derive RD automatically for VNI using passed information - it
5330 * is of the form RouterId:unique-id-for-vni.
5331 */
5332 void bgp_evpn_derive_auto_rd(struct bgp *bgp, struct bgpevpn *vpn)
5333 {
5334 char buf[100];
5335
5336 vpn->prd.family = AF_UNSPEC;
5337 vpn->prd.prefixlen = 64;
5338 sprintf(buf, "%s:%hu", inet_ntoa(bgp->router_id), vpn->rd_id);
5339 (void)str2prefix_rd(buf, &vpn->prd);
5340 UNSET_FLAG(vpn->flags, VNI_FLAG_RD_CFGD);
5341 }
5342
5343 /*
5344 * Lookup L3-VNI
5345 */
5346 bool bgp_evpn_lookup_l3vni_l2vni_table(vni_t vni)
5347 {
5348 struct list *inst = bm->bgp;
5349 struct listnode *node;
5350 struct bgp *bgp_vrf;
5351
5352 for (ALL_LIST_ELEMENTS_RO(inst, node, bgp_vrf)) {
5353 if (bgp_vrf->l3vni == vni)
5354 return true;
5355 }
5356
5357 return false;
5358 }
5359
5360 /*
5361 * Lookup VNI.
5362 */
5363 struct bgpevpn *bgp_evpn_lookup_vni(struct bgp *bgp, vni_t vni)
5364 {
5365 struct bgpevpn *vpn;
5366 struct bgpevpn tmp;
5367
5368 memset(&tmp, 0, sizeof(struct bgpevpn));
5369 tmp.vni = vni;
5370 vpn = hash_lookup(bgp->vnihash, &tmp);
5371 return vpn;
5372 }
5373
5374 /*
5375 * Create a new vpn - invoked upon configuration or zebra notification.
5376 */
5377 struct bgpevpn *bgp_evpn_new(struct bgp *bgp, vni_t vni,
5378 struct in_addr originator_ip,
5379 vrf_id_t tenant_vrf_id,
5380 struct in_addr mcast_grp)
5381 {
5382 struct bgpevpn *vpn;
5383
5384 if (!bgp)
5385 return NULL;
5386
5387 vpn = XCALLOC(MTYPE_BGP_EVPN, sizeof(struct bgpevpn));
5388
5389 /* Set values - RD and RT set to defaults. */
5390 vpn->vni = vni;
5391 vpn->originator_ip = originator_ip;
5392 vpn->tenant_vrf_id = tenant_vrf_id;
5393 vpn->mcast_grp = mcast_grp;
5394
5395 /* Initialize route-target import and export lists */
5396 vpn->import_rtl = list_new();
5397 vpn->import_rtl->cmp = (int (*)(void *, void *))evpn_route_target_cmp;
5398 vpn->import_rtl->del = evpn_xxport_delete_ecomm;
5399 vpn->export_rtl = list_new();
5400 vpn->export_rtl->cmp = (int (*)(void *, void *))evpn_route_target_cmp;
5401 vpn->export_rtl->del = evpn_xxport_delete_ecomm;
5402 bf_assign_index(bm->rd_idspace, vpn->rd_id);
5403 derive_rd_rt_for_vni(bgp, vpn);
5404
5405 /* Initialize EVPN route table. */
5406 vpn->route_table = bgp_table_init(bgp, AFI_L2VPN, SAFI_EVPN);
5407
5408 /* Add to hash */
5409 if (!hash_get(bgp->vnihash, vpn, hash_alloc_intern)) {
5410 XFREE(MTYPE_BGP_EVPN, vpn);
5411 return NULL;
5412 }
5413
5414 /* add to l2vni list on corresponding vrf */
5415 bgpevpn_link_to_l3vni(vpn);
5416
5417 QOBJ_REG(vpn, bgpevpn);
5418 return vpn;
5419 }
5420
5421 /*
5422 * Free a given VPN - called in multiple scenarios such as zebra
5423 * notification, configuration being deleted, advertise-all-vni disabled etc.
5424 * This just frees appropriate memory, caller should have taken other
5425 * needed actions.
5426 */
5427 void bgp_evpn_free(struct bgp *bgp, struct bgpevpn *vpn)
5428 {
5429 bgpevpn_unlink_from_l3vni(vpn);
5430 bgp_table_unlock(vpn->route_table);
5431 bgp_evpn_unmap_vni_from_its_rts(bgp, vpn);
5432 list_delete(&vpn->import_rtl);
5433 list_delete(&vpn->export_rtl);
5434 bf_release_index(bm->rd_idspace, vpn->rd_id);
5435 hash_release(bgp->vnihash, vpn);
5436 QOBJ_UNREG(vpn);
5437 XFREE(MTYPE_BGP_EVPN, vpn);
5438 }
5439
5440 /*
5441 * Lookup local ES.
5442 */
5443 struct evpnes *bgp_evpn_lookup_es(struct bgp *bgp, esi_t *esi)
5444 {
5445 struct evpnes *es;
5446 struct evpnes tmp;
5447
5448 memset(&tmp, 0, sizeof(struct evpnes));
5449 memcpy(&tmp.esi, esi, sizeof(esi_t));
5450 es = hash_lookup(bgp->esihash, &tmp);
5451 return es;
5452 }
5453
5454 /*
5455 * Create a new local es - invoked upon zebra notification.
5456 */
5457 struct evpnes *bgp_evpn_es_new(struct bgp *bgp,
5458 esi_t *esi,
5459 struct ipaddr *originator_ip)
5460 {
5461 char buf[100];
5462 struct evpnes *es;
5463
5464 if (!bgp)
5465 return NULL;
5466
5467 es = XCALLOC(MTYPE_BGP_EVPN_ES, sizeof(struct evpnes));
5468
5469 /* set the ESI and originator_ip */
5470 memcpy(&es->esi, esi, sizeof(esi_t));
5471 memcpy(&es->originator_ip, originator_ip, sizeof(struct ipaddr));
5472
5473 /* Initialise the VTEP list */
5474 es->vtep_list = list_new();
5475 es->vtep_list->cmp = evpn_vtep_ip_cmp;
5476
5477 /* auto derive RD for this es */
5478 bf_assign_index(bm->rd_idspace, es->rd_id);
5479 es->prd.family = AF_UNSPEC;
5480 es->prd.prefixlen = 64;
5481 sprintf(buf, "%s:%hu", inet_ntoa(bgp->router_id), es->rd_id);
5482 (void)str2prefix_rd(buf, &es->prd);
5483
5484 /* Initialize the ES route table */
5485 es->route_table = bgp_table_init(bgp, AFI_L2VPN, SAFI_EVPN);
5486
5487 /* Add to hash */
5488 if (!hash_get(bgp->esihash, es, hash_alloc_intern)) {
5489 XFREE(MTYPE_BGP_EVPN_ES, es);
5490 return NULL;
5491 }
5492
5493 QOBJ_REG(es, evpnes);
5494 return es;
5495 }
5496
5497 /*
5498 * Free a given ES -
5499 * This just frees appropriate memory, caller should have taken other
5500 * needed actions.
5501 */
5502 void bgp_evpn_es_free(struct bgp *bgp, struct evpnes *es)
5503 {
5504 list_delete(&es->vtep_list);
5505 bgp_table_unlock(es->route_table);
5506 bf_release_index(bm->rd_idspace, es->rd_id);
5507 hash_release(bgp->esihash, es);
5508 QOBJ_UNREG(es);
5509 XFREE(MTYPE_BGP_EVPN_ES, es);
5510 }
5511
5512 /*
5513 * Import evpn route from global table to VNI/VRF/ESI.
5514 */
5515 int bgp_evpn_import_route(struct bgp *bgp, afi_t afi, safi_t safi,
5516 struct prefix *p, struct bgp_path_info *pi)
5517 {
5518 return install_uninstall_evpn_route(bgp, afi, safi, p, pi, 1);
5519 }
5520
5521 /*
5522 * Unimport evpn route from VNI/VRF/ESI.
5523 */
5524 int bgp_evpn_unimport_route(struct bgp *bgp, afi_t afi, safi_t safi,
5525 struct prefix *p, struct bgp_path_info *pi)
5526 {
5527 return install_uninstall_evpn_route(bgp, afi, safi, p, pi, 0);
5528 }
5529
5530 /* filter routes which have martian next hops */
5531 int bgp_filter_evpn_routes_upon_martian_nh_change(struct bgp *bgp)
5532 {
5533 afi_t afi;
5534 safi_t safi;
5535 struct bgp_node *rd_rn, *rn;
5536 struct bgp_table *table;
5537 struct bgp_path_info *pi;
5538
5539 afi = AFI_L2VPN;
5540 safi = SAFI_EVPN;
5541
5542 /* Walk entire global routing table and evaluate routes which could be
5543 * imported into this VPN. Note that we cannot just look at the routes
5544 * for the VNI's RD -
5545 * remote routes applicable for this VNI could have any RD.
5546 */
5547 /* EVPN routes are a 2-level table. */
5548 for (rd_rn = bgp_table_top(bgp->rib[afi][safi]); rd_rn;
5549 rd_rn = bgp_route_next(rd_rn)) {
5550 table = bgp_node_get_bgp_table_info(rd_rn);
5551 if (!table)
5552 continue;
5553
5554 for (rn = bgp_table_top(table); rn; rn = bgp_route_next(rn)) {
5555
5556 for (pi = bgp_node_get_bgp_path_info(rn); pi;
5557 pi = pi->next) {
5558
5559 /* Consider "valid" remote routes applicable for
5560 * this VNI. */
5561 if (!(pi->type == ZEBRA_ROUTE_BGP
5562 && pi->sub_type == BGP_ROUTE_NORMAL))
5563 continue;
5564 if (bgp_nexthop_self(bgp, afi,
5565 pi->type, pi->sub_type,
5566 pi->attr, rn)) {
5567
5568 char attr_str[BUFSIZ] = {0};
5569 char pbuf[PREFIX_STRLEN];
5570
5571 bgp_dump_attr(pi->attr, attr_str,
5572 BUFSIZ);
5573
5574 if (bgp_debug_update(pi->peer, &rn->p,
5575 NULL, 1))
5576 zlog_debug(
5577 "%u: prefix %s with attr %s - DENIED due to martian or self nexthop",
5578 bgp->vrf_id,
5579 prefix2str(
5580 &rn->p, pbuf,
5581 sizeof(pbuf)),
5582 attr_str);
5583
5584 bgp_evpn_unimport_route(bgp, afi, safi,
5585 &rn->p, pi);
5586
5587 bgp_rib_remove(rn, pi, pi->peer, afi,
5588 safi);
5589 }
5590 }
5591 }
5592 }
5593
5594 return 0;
5595 }
5596
5597 /*
5598 * Handle del of a local MACIP.
5599 */
5600 int bgp_evpn_local_macip_del(struct bgp *bgp, vni_t vni, struct ethaddr *mac,
5601 struct ipaddr *ip, int state)
5602 {
5603 struct bgpevpn *vpn;
5604 struct prefix_evpn p;
5605 struct bgp_node *rn;
5606
5607 /* Lookup VNI hash - should exist. */
5608 vpn = bgp_evpn_lookup_vni(bgp, vni);
5609 if (!vpn || !is_vni_live(vpn)) {
5610 flog_warn(EC_BGP_EVPN_VPN_VNI,
5611 "%u: VNI hash entry for VNI %u %s at MACIP DEL",
5612 bgp->vrf_id, vni, vpn ? "not live" : "not found");
5613 return -1;
5614 }
5615
5616 build_evpn_type2_prefix(&p, mac, ip);
5617 if (state == ZEBRA_NEIGH_ACTIVE) {
5618 /* Remove EVPN type-2 route and schedule for processing. */
5619 delete_evpn_route(bgp, vpn, &p);
5620 } else {
5621 /* Re-instate the current remote best path if any */
5622 rn = bgp_node_lookup(vpn->route_table, (struct prefix *)&p);
5623 if (rn)
5624 evpn_zebra_reinstall_best_route(bgp, vpn, rn);
5625 }
5626
5627 return 0;
5628 }
5629
5630 /*
5631 * Handle add of a local MACIP.
5632 */
5633 int bgp_evpn_local_macip_add(struct bgp *bgp, vni_t vni, struct ethaddr *mac,
5634 struct ipaddr *ip, uint8_t flags, uint32_t seq)
5635 {
5636 struct bgpevpn *vpn;
5637 struct prefix_evpn p;
5638
5639 /* Lookup VNI hash - should exist. */
5640 vpn = bgp_evpn_lookup_vni(bgp, vni);
5641 if (!vpn || !is_vni_live(vpn)) {
5642 flog_warn(EC_BGP_EVPN_VPN_VNI,
5643 "%u: VNI hash entry for VNI %u %s at MACIP ADD",
5644 bgp->vrf_id, vni, vpn ? "not live" : "not found");
5645 return -1;
5646 }
5647
5648 /* Create EVPN type-2 route and schedule for processing. */
5649 build_evpn_type2_prefix(&p, mac, ip);
5650 if (update_evpn_route(bgp, vpn, &p, flags, seq)) {
5651 char buf[ETHER_ADDR_STRLEN];
5652 char buf2[INET6_ADDRSTRLEN];
5653
5654 flog_err(
5655 EC_BGP_EVPN_ROUTE_CREATE,
5656 "%u:Failed to create Type-2 route, VNI %u %s MAC %s IP %s (flags: 0x%x)",
5657 bgp->vrf_id, vpn->vni,
5658 CHECK_FLAG(flags, ZEBRA_MACIP_TYPE_STICKY)
5659 ? "sticky gateway"
5660 : "",
5661 prefix_mac2str(mac, buf, sizeof(buf)),
5662 ipaddr2str(ip, buf2, sizeof(buf2)), flags);
5663 return -1;
5664 }
5665
5666 return 0;
5667 }
5668
5669 static void link_l2vni_hash_to_l3vni(struct hash_bucket *bucket,
5670 struct bgp *bgp_vrf)
5671 {
5672 struct bgpevpn *vpn = (struct bgpevpn *)bucket->data;
5673 struct bgp *bgp_evpn = NULL;
5674
5675 bgp_evpn = bgp_get_evpn();
5676 assert(bgp_evpn);
5677
5678 if (vpn->tenant_vrf_id == bgp_vrf->vrf_id)
5679 bgpevpn_link_to_l3vni(vpn);
5680 }
5681
5682 int bgp_evpn_local_l3vni_add(vni_t l3vni, vrf_id_t vrf_id,
5683 struct ethaddr *svi_rmac,
5684 struct ethaddr *vrr_rmac,
5685 struct in_addr originator_ip, int filter,
5686 ifindex_t svi_ifindex,
5687 bool is_anycast_mac)
5688 {
5689 struct bgp *bgp_vrf = NULL; /* bgp VRF instance */
5690 struct bgp *bgp_evpn = NULL; /* EVPN bgp instance */
5691 struct listnode *node = NULL;
5692 struct bgpevpn *vpn = NULL;
5693 as_t as = 0;
5694
5695 /* get the EVPN instance - required to get the AS number for VRF
5696 * auto-creatio
5697 */
5698 bgp_evpn = bgp_get_evpn();
5699 if (!bgp_evpn) {
5700 flog_err(
5701 EC_BGP_NO_DFLT,
5702 "Cannot process L3VNI %u ADD - EVPN BGP instance not yet created",
5703 l3vni);
5704 return -1;
5705 }
5706 as = bgp_evpn->as;
5707
5708 /* if the BGP vrf instance doesn't exist - create one */
5709 bgp_vrf = bgp_lookup_by_vrf_id(vrf_id);
5710 if (!bgp_vrf) {
5711
5712 int ret = 0;
5713
5714 ret = bgp_get_vty(&bgp_vrf, &as, vrf_id_to_name(vrf_id),
5715 vrf_id == VRF_DEFAULT
5716 ? BGP_INSTANCE_TYPE_DEFAULT
5717 : BGP_INSTANCE_TYPE_VRF);
5718 switch (ret) {
5719 case BGP_ERR_AS_MISMATCH:
5720 flog_err(EC_BGP_EVPN_AS_MISMATCH,
5721 "BGP is already running; AS is %u\n", as);
5722 return -1;
5723 case BGP_ERR_INSTANCE_MISMATCH:
5724 flog_err(EC_BGP_EVPN_INSTANCE_MISMATCH,
5725 "BGP instance name and AS number mismatch\n");
5726 return -1;
5727 }
5728
5729 /* mark as auto created */
5730 SET_FLAG(bgp_vrf->vrf_flags, BGP_VRF_AUTO);
5731 }
5732
5733 /* associate the vrf with l3vni and related parameters */
5734 bgp_vrf->l3vni = l3vni;
5735 bgp_vrf->originator_ip = originator_ip;
5736 bgp_vrf->l3vni_svi_ifindex = svi_ifindex;
5737 bgp_vrf->evpn_info->is_anycast_mac = is_anycast_mac;
5738
5739 /* copy anycast MAC from VRR MAC */
5740 memcpy(&bgp_vrf->rmac, vrr_rmac, ETH_ALEN);
5741 /* copy sys RMAC from SVI MAC */
5742 memcpy(&bgp_vrf->evpn_info->pip_rmac_zebra, svi_rmac, ETH_ALEN);
5743 /* PIP user configured mac is not present use svi mac as sys mac */
5744 if (is_zero_mac(&bgp_vrf->evpn_info->pip_rmac_static))
5745 memcpy(&bgp_vrf->evpn_info->pip_rmac, svi_rmac, ETH_ALEN);
5746
5747 if (bgp_debug_zebra(NULL)) {
5748 char buf[ETHER_ADDR_STRLEN];
5749 char buf1[ETHER_ADDR_STRLEN];
5750 char buf2[ETHER_ADDR_STRLEN];
5751
5752 zlog_debug("VRF %s vni %u pip %s RMAC %s sys RMAC %s static RMAC %s is_anycast_mac %s",
5753 vrf_id_to_name(bgp_vrf->vrf_id),
5754 bgp_vrf->l3vni,
5755 bgp_vrf->evpn_info->advertise_pip ? "enable"
5756 : "disable",
5757 prefix_mac2str(&bgp_vrf->rmac, buf, sizeof(buf)),
5758 prefix_mac2str(&bgp_vrf->evpn_info->pip_rmac,
5759 buf1, sizeof(buf1)),
5760 prefix_mac2str(&bgp_vrf->evpn_info->pip_rmac_static,
5761 buf2, sizeof(buf2)),
5762 is_anycast_mac ? "Enable" : "Disable");
5763 }
5764 /* set the right filter - are we using l3vni only for prefix routes? */
5765 if (filter)
5766 SET_FLAG(bgp_vrf->vrf_flags, BGP_VRF_L3VNI_PREFIX_ROUTES_ONLY);
5767
5768 /* Map auto derive or configured RTs */
5769 if (!CHECK_FLAG(bgp_vrf->vrf_flags, BGP_VRF_IMPORT_RT_CFGD))
5770 evpn_auto_rt_import_add_for_vrf(bgp_vrf);
5771 else
5772 bgp_evpn_map_vrf_to_its_rts(bgp_vrf);
5773
5774 if (!CHECK_FLAG(bgp_vrf->vrf_flags, BGP_VRF_EXPORT_RT_CFGD))
5775 evpn_auto_rt_export_add_for_vrf(bgp_vrf);
5776
5777 /* auto derive RD */
5778 bgp_evpn_derive_auto_rd_for_vrf(bgp_vrf);
5779
5780 /* link all corresponding l2vnis */
5781 hash_iterate(bgp_evpn->vnihash,
5782 (void (*)(struct hash_bucket *,
5783 void *))link_l2vni_hash_to_l3vni,
5784 bgp_vrf);
5785
5786 /* Only update all corresponding type-2 routes if we are advertising two
5787 * labels along with type-2 routes
5788 */
5789 if (!filter)
5790 for (ALL_LIST_ELEMENTS_RO(bgp_vrf->l2vnis, node, vpn))
5791 update_routes_for_vni(bgp_evpn, vpn);
5792
5793 /* advertise type-5 routes if needed */
5794 update_advertise_vrf_routes(bgp_vrf);
5795
5796 /* install all remote routes belonging to this l3vni into correspondng
5797 * vrf */
5798 install_routes_for_vrf(bgp_vrf);
5799
5800 return 0;
5801 }
5802
5803 int bgp_evpn_local_l3vni_del(vni_t l3vni, vrf_id_t vrf_id)
5804 {
5805 struct bgp *bgp_vrf = NULL; /* bgp vrf instance */
5806 struct bgp *bgp_evpn = NULL; /* EVPN bgp instance */
5807 struct listnode *node = NULL;
5808 struct listnode *next = NULL;
5809 struct bgpevpn *vpn = NULL;
5810
5811 bgp_vrf = bgp_lookup_by_vrf_id(vrf_id);
5812 if (!bgp_vrf) {
5813 flog_err(
5814 EC_BGP_NO_DFLT,
5815 "Cannot process L3VNI %u Del - Could not find BGP instance",
5816 l3vni);
5817 return -1;
5818 }
5819
5820 bgp_evpn = bgp_get_evpn();
5821 if (!bgp_evpn) {
5822 flog_err(
5823 EC_BGP_NO_DFLT,
5824 "Cannot process L3VNI %u Del - Could not find EVPN BGP instance",
5825 l3vni);
5826 return -1;
5827 }
5828
5829 /* Remove remote routes from BGT VRF even if BGP_VRF_AUTO is configured,
5830 * bgp_delete would not remove/decrement bgp_path_info of the ip_prefix
5831 * routes. This will uninstalling the routes from zebra and decremnt the
5832 * bgp info count.
5833 */
5834 uninstall_routes_for_vrf(bgp_vrf);
5835
5836 /* delete/withdraw all type-5 routes */
5837 delete_withdraw_vrf_routes(bgp_vrf);
5838
5839 /* remove the l3vni from vrf instance */
5840 bgp_vrf->l3vni = 0;
5841
5842 /* remove the Rmac from the BGP vrf */
5843 memset(&bgp_vrf->rmac, 0, sizeof(struct ethaddr));
5844 memset(&bgp_vrf->evpn_info->pip_rmac_zebra, 0, ETH_ALEN);
5845 if (is_zero_mac(&bgp_vrf->evpn_info->pip_rmac_static) &&
5846 !is_zero_mac(&bgp_vrf->evpn_info->pip_rmac))
5847 memset(&bgp_vrf->evpn_info->pip_rmac, 0, ETH_ALEN);
5848
5849 /* remove default import RT or Unmap non-default import RT */
5850 if (!list_isempty(bgp_vrf->vrf_import_rtl)) {
5851 bgp_evpn_unmap_vrf_from_its_rts(bgp_vrf);
5852 if (!CHECK_FLAG(bgp_vrf->vrf_flags, BGP_VRF_IMPORT_RT_CFGD))
5853 list_delete_all_node(bgp_vrf->vrf_import_rtl);
5854 }
5855
5856 /* remove default export RT */
5857 if (!list_isempty(bgp_vrf->vrf_export_rtl) &&
5858 !CHECK_FLAG(bgp_vrf->vrf_flags, BGP_VRF_EXPORT_RT_CFGD)) {
5859 list_delete_all_node(bgp_vrf->vrf_export_rtl);
5860 }
5861
5862 /* update all corresponding local mac-ip routes */
5863 if (!CHECK_FLAG(bgp_vrf->vrf_flags, BGP_VRF_L3VNI_PREFIX_ROUTES_ONLY)) {
5864 for (ALL_LIST_ELEMENTS_RO(bgp_vrf->l2vnis, node, vpn)) {
5865 UNSET_FLAG(vpn->flags, VNI_FLAG_USE_TWO_LABELS);
5866 update_routes_for_vni(bgp_evpn, vpn);
5867 }
5868 }
5869
5870 /* If any L2VNIs point to this instance, unlink them. */
5871 for (ALL_LIST_ELEMENTS(bgp_vrf->l2vnis, node, next, vpn))
5872 bgpevpn_unlink_from_l3vni(vpn);
5873
5874 UNSET_FLAG(bgp_vrf->vrf_flags, BGP_VRF_L3VNI_PREFIX_ROUTES_ONLY);
5875
5876 /* Delete the instance if it was autocreated */
5877 if (CHECK_FLAG(bgp_vrf->vrf_flags, BGP_VRF_AUTO))
5878 bgp_delete(bgp_vrf);
5879
5880 return 0;
5881 }
5882
5883 /*
5884 * Handle del of a local VNI.
5885 */
5886 int bgp_evpn_local_vni_del(struct bgp *bgp, vni_t vni)
5887 {
5888 struct bgpevpn *vpn;
5889
5890 /* Locate VNI hash */
5891 vpn = bgp_evpn_lookup_vni(bgp, vni);
5892 if (!vpn) {
5893 if (bgp_debug_zebra(NULL))
5894 flog_warn(
5895 EC_BGP_EVPN_VPN_VNI,
5896 "%u: VNI hash entry for VNI %u not found at DEL",
5897 bgp->vrf_id, vni);
5898 return 0;
5899 }
5900
5901 /* Remove all local EVPN routes and schedule for processing (to
5902 * withdraw from peers).
5903 */
5904 delete_routes_for_vni(bgp, vpn);
5905
5906 /*
5907 * tunnel is no longer active, del tunnel ip address from tip_hash
5908 */
5909 bgp_tip_del(bgp, &vpn->originator_ip);
5910
5911 /* Clear "live" flag and see if hash needs to be freed. */
5912 UNSET_FLAG(vpn->flags, VNI_FLAG_LIVE);
5913 if (!is_vni_configured(vpn))
5914 bgp_evpn_free(bgp, vpn);
5915
5916 return 0;
5917 }
5918
5919 /*
5920 * Handle add (or update) of a local VNI. The VNI changes we care
5921 * about are for the local-tunnel-ip and the (tenant) VRF.
5922 */
5923 int bgp_evpn_local_vni_add(struct bgp *bgp, vni_t vni,
5924 struct in_addr originator_ip,
5925 vrf_id_t tenant_vrf_id,
5926 struct in_addr mcast_grp)
5927
5928 {
5929 struct bgpevpn *vpn;
5930 struct prefix_evpn p;
5931
5932 /* Lookup VNI. If present and no change, exit. */
5933 vpn = bgp_evpn_lookup_vni(bgp, vni);
5934 if (vpn) {
5935
5936 if (is_vni_live(vpn)
5937 && IPV4_ADDR_SAME(&vpn->originator_ip, &originator_ip)
5938 && IPV4_ADDR_SAME(&vpn->mcast_grp, &mcast_grp)
5939 && vpn->tenant_vrf_id == tenant_vrf_id)
5940 /* Probably some other param has changed that we don't
5941 * care about. */
5942 return 0;
5943
5944 bgp_evpn_mcast_grp_change(bgp, vpn, mcast_grp);
5945
5946 /* Update tenant_vrf_id if it has changed. */
5947 if (vpn->tenant_vrf_id != tenant_vrf_id) {
5948 bgpevpn_unlink_from_l3vni(vpn);
5949 vpn->tenant_vrf_id = tenant_vrf_id;
5950 bgpevpn_link_to_l3vni(vpn);
5951 }
5952
5953 /* If tunnel endpoint IP has changed, update (and delete prior
5954 * type-3 route, if needed.)
5955 */
5956 if (!IPV4_ADDR_SAME(&vpn->originator_ip, &originator_ip))
5957 handle_tunnel_ip_change(bgp, vpn, originator_ip);
5958
5959 /* Update all routes with new endpoint IP and/or export RT
5960 * for VRFs
5961 */
5962 if (is_vni_live(vpn))
5963 update_routes_for_vni(bgp, vpn);
5964 }
5965
5966 /* Create or update as appropriate. */
5967 if (!vpn) {
5968 vpn = bgp_evpn_new(bgp, vni, originator_ip, tenant_vrf_id,
5969 mcast_grp);
5970 if (!vpn) {
5971 flog_err(
5972 EC_BGP_VNI,
5973 "%u: Failed to allocate VNI entry for VNI %u - at Add",
5974 bgp->vrf_id, vni);
5975 return -1;
5976 }
5977 }
5978
5979 /* if the VNI is live already, there is nothing more to do */
5980 if (is_vni_live(vpn))
5981 return 0;
5982
5983 /* Mark as "live" */
5984 SET_FLAG(vpn->flags, VNI_FLAG_LIVE);
5985
5986 /* tunnel is now active, add tunnel-ip to db */
5987 bgp_tip_add(bgp, &originator_ip);
5988
5989 /* filter routes as nexthop database has changed */
5990 bgp_filter_evpn_routes_upon_martian_nh_change(bgp);
5991
5992 /*
5993 * Create EVPN type-3 route and schedule for processing.
5994 *
5995 * RT-3 only if doing head-end replication
5996 */
5997 if (bgp_evpn_vni_flood_mode_get(bgp, vpn)
5998 == VXLAN_FLOOD_HEAD_END_REPL) {
5999 build_evpn_type3_prefix(&p, vpn->originator_ip);
6000 if (update_evpn_route(bgp, vpn, &p, 0, 0)) {
6001 flog_err(EC_BGP_EVPN_ROUTE_CREATE,
6002 "%u: Type3 route creation failure for VNI %u",
6003 bgp->vrf_id, vni);
6004 return -1;
6005 }
6006 }
6007
6008 /* If we have learnt and retained remote routes (VTEPs, MACs) for this
6009 * VNI,
6010 * install them.
6011 */
6012 install_routes_for_vni(bgp, vpn);
6013
6014 /* If we are advertising gateway mac-ip
6015 It needs to be conveyed again to zebra */
6016 bgp_zebra_advertise_gw_macip(bgp, vpn->advertise_gw_macip, vpn->vni);
6017
6018 /* advertise svi mac-ip knob to zebra */
6019 bgp_zebra_advertise_svi_macip(bgp, vpn->advertise_svi_macip, vpn->vni);
6020
6021 return 0;
6022 }
6023
6024 /*
6025 * bgp_evpn_local_es_del
6026 */
6027 int bgp_evpn_local_es_del(struct bgp *bgp,
6028 esi_t *esi,
6029 struct ipaddr *originator_ip)
6030 {
6031 char buf[ESI_STR_LEN];
6032 struct evpnes *es = NULL;
6033
6034 if (!bgp->esihash) {
6035 flog_err(EC_BGP_ES_CREATE, "%u: ESI hash not yet created",
6036 bgp->vrf_id);
6037 return -1;
6038 }
6039
6040 /* Lookup ESI hash - should exist. */
6041 es = bgp_evpn_lookup_es(bgp, esi);
6042 if (!es) {
6043 flog_warn(EC_BGP_EVPN_ESI,
6044 "%u: ESI hash entry for ESI %s at Local ES DEL",
6045 bgp->vrf_id, esi_to_str(esi, buf, sizeof(buf)));
6046 return -1;
6047 }
6048
6049 /* Delete all local EVPN ES routes from ESI table
6050 * and schedule for processing (to withdraw from peers))
6051 */
6052 delete_routes_for_es(bgp, es);
6053
6054 /* free the hash entry */
6055 bgp_evpn_es_free(bgp, es);
6056
6057 return 0;
6058 }
6059
6060 /*
6061 * bgp_evpn_local_es_add
6062 */
6063 int bgp_evpn_local_es_add(struct bgp *bgp,
6064 esi_t *esi,
6065 struct ipaddr *originator_ip)
6066 {
6067 char buf[ESI_STR_LEN];
6068 struct evpnes *es = NULL;
6069 struct prefix_evpn p;
6070
6071 if (!bgp->esihash) {
6072 flog_err(EC_BGP_ES_CREATE, "%u: ESI hash not yet created",
6073 bgp->vrf_id);
6074 return -1;
6075 }
6076
6077 /* create the new es */
6078 es = bgp_evpn_lookup_es(bgp, esi);
6079 if (!es) {
6080 es = bgp_evpn_es_new(bgp, esi, originator_ip);
6081 if (!es) {
6082 flog_err(
6083 EC_BGP_ES_CREATE,
6084 "%u: Failed to allocate ES entry for ESI %s - at Local ES Add",
6085 bgp->vrf_id, esi_to_str(esi, buf, sizeof(buf)));
6086 return -1;
6087 }
6088 }
6089 UNSET_FLAG(es->flags, EVPNES_REMOTE);
6090 SET_FLAG(es->flags, EVPNES_LOCAL);
6091
6092 build_evpn_type4_prefix(&p, esi, originator_ip->ipaddr_v4);
6093 if (update_evpn_type4_route(bgp, es, &p)) {
6094 flog_err(EC_BGP_EVPN_ROUTE_CREATE,
6095 "%u: Type4 route creation failure for ESI %s",
6096 bgp->vrf_id, esi_to_str(esi, buf, sizeof(buf)));
6097 return -1;
6098 }
6099
6100 /* import all remote ES routes in th ES table */
6101 install_routes_for_es(bgp, es);
6102
6103 return 0;
6104 }
6105
6106 /*
6107 * Handle change in setting for BUM handling. The supported values
6108 * are head-end replication and dropping all BUM packets. Any change
6109 * should be registered with zebra. Also, if doing head-end replication,
6110 * need to advertise local VNIs as EVPN RT-3 wheras, if BUM packets are
6111 * to be dropped, the RT-3s must be withdrawn.
6112 */
6113 void bgp_evpn_flood_control_change(struct bgp *bgp)
6114 {
6115 zlog_info("L2VPN EVPN BUM handling is %s",
6116 bgp->vxlan_flood_ctrl == VXLAN_FLOOD_HEAD_END_REPL ?
6117 "Flooding" : "Flooding Disabled");
6118
6119 bgp_zebra_vxlan_flood_control(bgp, bgp->vxlan_flood_ctrl);
6120 if (bgp->vxlan_flood_ctrl == VXLAN_FLOOD_HEAD_END_REPL)
6121 hash_iterate(bgp->vnihash, create_advertise_type3, bgp);
6122 else if (bgp->vxlan_flood_ctrl == VXLAN_FLOOD_DISABLED)
6123 hash_iterate(bgp->vnihash, delete_withdraw_type3, bgp);
6124 }
6125
6126 /*
6127 * Cleanup EVPN information on disable - Need to delete and withdraw
6128 * EVPN routes from peers.
6129 */
6130 void bgp_evpn_cleanup_on_disable(struct bgp *bgp)
6131 {
6132 hash_iterate(bgp->vnihash, (void (*)(struct hash_bucket *,
6133 void *))cleanup_vni_on_disable,
6134 bgp);
6135 }
6136
6137 /*
6138 * Cleanup EVPN information - invoked at the time of bgpd exit or when the
6139 * BGP instance (default) is being freed.
6140 */
6141 void bgp_evpn_cleanup(struct bgp *bgp)
6142 {
6143 hash_iterate(bgp->vnihash,
6144 (void (*)(struct hash_bucket *, void *))free_vni_entry,
6145 bgp);
6146
6147 hash_free(bgp->import_rt_hash);
6148 bgp->import_rt_hash = NULL;
6149
6150 hash_free(bgp->vrf_import_rt_hash);
6151 bgp->vrf_import_rt_hash = NULL;
6152
6153 hash_free(bgp->vnihash);
6154 bgp->vnihash = NULL;
6155 if (bgp->esihash)
6156 hash_free(bgp->esihash);
6157 bgp->esihash = NULL;
6158
6159 list_delete(&bgp->vrf_import_rtl);
6160 list_delete(&bgp->vrf_export_rtl);
6161 list_delete(&bgp->l2vnis);
6162 }
6163
6164 /*
6165 * Initialization for EVPN
6166 * Create
6167 * VNI hash table
6168 * hash for RT to VNI
6169 */
6170 void bgp_evpn_init(struct bgp *bgp)
6171 {
6172 bgp->vnihash =
6173 hash_create(vni_hash_key_make, vni_hash_cmp, "BGP VNI Hash");
6174 bgp->esihash =
6175 hash_create(esi_hash_keymake, esi_cmp,
6176 "BGP EVPN Local ESI Hash");
6177 bgp->import_rt_hash =
6178 hash_create(import_rt_hash_key_make, import_rt_hash_cmp,
6179 "BGP Import RT Hash");
6180 bgp->vrf_import_rt_hash =
6181 hash_create(vrf_import_rt_hash_key_make, vrf_import_rt_hash_cmp,
6182 "BGP VRF Import RT Hash");
6183 bgp->vrf_import_rtl = list_new();
6184 bgp->vrf_import_rtl->cmp =
6185 (int (*)(void *, void *))evpn_route_target_cmp;
6186 bgp->vrf_import_rtl->del = evpn_xxport_delete_ecomm;
6187 bgp->vrf_export_rtl = list_new();
6188 bgp->vrf_export_rtl->cmp =
6189 (int (*)(void *, void *))evpn_route_target_cmp;
6190 bgp->vrf_export_rtl->del = evpn_xxport_delete_ecomm;
6191 bgp->l2vnis = list_new();
6192 bgp->l2vnis->cmp = vni_list_cmp;
6193 /* By default Duplicate Address Dection is enabled.
6194 * Max-moves (N) 5, detection time (M) 180
6195 * default action is warning-only
6196 * freeze action permanently freezes address,
6197 * and freeze time (auto-recovery) is disabled.
6198 */
6199 if (bgp->evpn_info) {
6200 bgp->evpn_info->dup_addr_detect = true;
6201 bgp->evpn_info->dad_time = EVPN_DAD_DEFAULT_TIME;
6202 bgp->evpn_info->dad_max_moves = EVPN_DAD_DEFAULT_MAX_MOVES;
6203 bgp->evpn_info->dad_freeze = false;
6204 bgp->evpn_info->dad_freeze_time = 0;
6205 /* Initialize zebra vxlan */
6206 bgp_zebra_dup_addr_detection(bgp);
6207 /* Enable PIP feature by default for bgp vrf instance */
6208 if (bgp->inst_type == BGP_INSTANCE_TYPE_VRF) {
6209 struct bgp *bgp_default;
6210
6211 bgp->evpn_info->advertise_pip = true;
6212 bgp_default = bgp_get_default();
6213 if (bgp_default)
6214 bgp->evpn_info->pip_ip = bgp_default->router_id;
6215 }
6216 }
6217
6218 /* Default BUM handling is to do head-end replication. */
6219 bgp->vxlan_flood_ctrl = VXLAN_FLOOD_HEAD_END_REPL;
6220 }
6221
6222 void bgp_evpn_vrf_delete(struct bgp *bgp_vrf)
6223 {
6224 bgp_evpn_unmap_vrf_from_its_rts(bgp_vrf);
6225 }
6226
6227 /*
6228 * Get the prefixlen of the ip prefix carried within the type5 evpn route.
6229 */
6230 int bgp_evpn_get_type5_prefixlen(struct prefix *pfx)
6231 {
6232 struct prefix_evpn *evp = (struct prefix_evpn *)pfx;
6233
6234 if (!pfx || pfx->family != AF_EVPN)
6235 return 0;
6236
6237 if (evp->prefix.route_type != BGP_EVPN_IP_PREFIX_ROUTE)
6238 return 0;
6239
6240 return evp->prefix.prefix_addr.ip_prefix_length;
6241 }
6242
6243 /*
6244 * Should we register nexthop for this EVPN prefix for nexthop tracking?
6245 */
6246 bool bgp_evpn_is_prefix_nht_supported(struct prefix *pfx)
6247 {
6248 struct prefix_evpn *evp = (struct prefix_evpn *)pfx;
6249
6250 /*
6251 * EVPN RT-5 should not be marked as valid and imported to vrfs if the
6252 * BGP nexthop is not reachable. To check for the nexthop reachability,
6253 * Add nexthop for EVPN RT-5 for nexthop tracking.
6254 *
6255 * Ideally, a BGP route should be marked as valid only if the
6256 * nexthop is reachable. Thus, other EVPN route types also should be
6257 * added here after testing is performed for them.
6258 */
6259 if (pfx && pfx->family == AF_EVPN &&
6260 evp->prefix.route_type == BGP_EVPN_IP_PREFIX_ROUTE)
6261 return true;
6262
6263 return false;
6264 }