1 /* Ethernet-VPN Packet and vty Processing File
2 * Copyright (C) 2016 6WIND
3 * Copyright (C) 2017 Cumulus Networks, Inc.
5 * This file is part of FRR.
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
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.
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
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"
52 * Definitions and external declarations.
54 extern struct zclient
*zclient
;
56 DEFINE_QOBJ_TYPE(bgpevpn
)
57 DEFINE_QOBJ_TYPE(evpnes
)
61 * Static function declarations
63 static void delete_evpn_route_entry(struct bgp
*bgp
, afi_t afi
, safi_t safi
,
65 struct bgp_path_info
**pi
);
66 static int delete_all_vni_routes(struct bgp
*bgp
, struct bgpevpn
*vpn
);
72 /* compare two IPV4 VTEP IPs */
73 static int evpn_vtep_ip_cmp(const void *p1
, const void *p2
)
75 const struct in_addr
*ip1
= p1
;
76 const struct in_addr
*ip2
= p2
;
82 return (ip1
->s_addr
== ip2
->s_addr
);
86 * Make hash key for ESI.
88 static unsigned int esi_hash_keymake(void *p
)
90 struct evpnes
*pes
= p
;
91 const void *pnt
= (void *)pes
->esi
.val
;
93 return jhash(pnt
, ESI_BYTES
, 0xa5a5a55a);
99 static int esi_cmp(const void *p1
, const void *p2
)
101 const struct evpnes
*pes1
= p1
;
102 const struct evpnes
*pes2
= p2
;
104 if (pes1
== NULL
&& pes2
== NULL
)
107 if (pes1
== NULL
|| pes2
== NULL
)
110 return (memcmp(pes1
->esi
.val
, pes2
->esi
.val
, ESI_BYTES
) == 0);
116 static unsigned int vni_hash_key_make(void *p
)
118 struct bgpevpn
*vpn
= p
;
119 return (jhash_1word(vpn
->vni
, 0));
123 * Comparison function for vni hash
125 static int vni_hash_cmp(const void *p1
, const void *p2
)
127 const struct bgpevpn
*vpn1
= p1
;
128 const struct bgpevpn
*vpn2
= p2
;
134 return (vpn1
->vni
== vpn2
->vni
);
138 * Make vrf import route target hash key.
140 static unsigned int vrf_import_rt_hash_key_make(void *p
)
142 struct vrf_irt_node
*irt
= p
;
143 char *pnt
= irt
->rt
.val
;
145 return jhash(pnt
, 8, 0x5abc1234);
149 * Comparison function for vrf import rt hash
151 static int vrf_import_rt_hash_cmp(const void *p1
, const void *p2
)
153 const struct vrf_irt_node
*irt1
= p1
;
154 const struct vrf_irt_node
*irt2
= p2
;
156 if (irt1
== NULL
&& irt2
== NULL
)
159 if (irt1
== NULL
|| irt2
== NULL
)
162 return (memcmp(irt1
->rt
.val
, irt2
->rt
.val
, ECOMMUNITY_SIZE
) == 0);
166 * Create a new vrf import_rt in default instance
168 static struct vrf_irt_node
*vrf_import_rt_new(struct ecommunity_val
*rt
)
170 struct bgp
*bgp_def
= NULL
;
171 struct vrf_irt_node
*irt
;
173 bgp_def
= bgp_get_default();
175 flog_err(EC_BGP_NO_DFLT
,
176 "vrf import rt new - def instance not created yet");
180 irt
= XCALLOC(MTYPE_BGP_EVPN_VRF_IMPORT_RT
,
181 sizeof(struct vrf_irt_node
));
186 irt
->vrfs
= list_new();
189 if (!hash_get(bgp_def
->vrf_import_rt_hash
, irt
, hash_alloc_intern
)) {
190 XFREE(MTYPE_BGP_EVPN_VRF_IMPORT_RT
, irt
);
198 * Free the vrf import rt node
200 static void vrf_import_rt_free(struct vrf_irt_node
*irt
)
202 struct bgp
*bgp_def
= NULL
;
204 bgp_def
= bgp_get_default();
206 flog_err(EC_BGP_NO_DFLT
,
207 "vrf import rt free - def instance not created yet");
211 hash_release(bgp_def
->vrf_import_rt_hash
, irt
);
212 list_delete(&irt
->vrfs
);
213 XFREE(MTYPE_BGP_EVPN_VRF_IMPORT_RT
, irt
);
217 * Function to lookup Import RT node - used to map a RT to set of
218 * VNIs importing routes with that RT.
220 static struct vrf_irt_node
*lookup_vrf_import_rt(struct ecommunity_val
*rt
)
222 struct bgp
*bgp_def
= NULL
;
223 struct vrf_irt_node
*irt
;
224 struct vrf_irt_node tmp
;
226 bgp_def
= bgp_get_default();
228 flog_err(EC_BGP_NO_DFLT
,
229 "vrf import rt lookup - def instance not created yet");
233 memset(&tmp
, 0, sizeof(struct vrf_irt_node
));
234 memcpy(&tmp
.rt
, rt
, ECOMMUNITY_SIZE
);
235 irt
= hash_lookup(bgp_def
->vrf_import_rt_hash
, &tmp
);
240 * Is specified VRF present on the RT's list of "importing" VRFs?
242 static int is_vrf_present_in_irt_vrfs(struct list
*vrfs
, struct bgp
*bgp_vrf
)
244 struct listnode
*node
= NULL
, *nnode
= NULL
;
245 struct bgp
*tmp_bgp_vrf
= NULL
;
247 for (ALL_LIST_ELEMENTS(vrfs
, node
, nnode
, tmp_bgp_vrf
)) {
248 if (tmp_bgp_vrf
== bgp_vrf
)
255 * Make import route target hash key.
257 static unsigned int import_rt_hash_key_make(void *p
)
259 struct irt_node
*irt
= p
;
260 char *pnt
= irt
->rt
.val
;
262 return jhash(pnt
, 8, 0xdeadbeef);
266 * Comparison function for import rt hash
268 static int import_rt_hash_cmp(const void *p1
, const void *p2
)
270 const struct irt_node
*irt1
= p1
;
271 const struct irt_node
*irt2
= p2
;
273 if (irt1
== NULL
&& irt2
== NULL
)
276 if (irt1
== NULL
|| irt2
== NULL
)
279 return (memcmp(irt1
->rt
.val
, irt2
->rt
.val
, ECOMMUNITY_SIZE
) == 0);
283 * Create a new import_rt
285 static struct irt_node
*import_rt_new(struct bgp
*bgp
,
286 struct ecommunity_val
*rt
)
288 struct irt_node
*irt
;
293 irt
= XCALLOC(MTYPE_BGP_EVPN_IMPORT_RT
, sizeof(struct irt_node
));
298 irt
->vnis
= list_new();
301 if (!hash_get(bgp
->import_rt_hash
, irt
, hash_alloc_intern
)) {
302 XFREE(MTYPE_BGP_EVPN_IMPORT_RT
, irt
);
310 * Free the import rt node
312 static void import_rt_free(struct bgp
*bgp
, struct irt_node
*irt
)
314 hash_release(bgp
->import_rt_hash
, irt
);
315 list_delete(&irt
->vnis
);
316 XFREE(MTYPE_BGP_EVPN_IMPORT_RT
, irt
);
320 * Function to lookup Import RT node - used to map a RT to set of
321 * VNIs importing routes with that RT.
323 static struct irt_node
*lookup_import_rt(struct bgp
*bgp
,
324 struct ecommunity_val
*rt
)
326 struct irt_node
*irt
;
329 memset(&tmp
, 0, sizeof(struct irt_node
));
330 memcpy(&tmp
.rt
, rt
, ECOMMUNITY_SIZE
);
331 irt
= hash_lookup(bgp
->import_rt_hash
, &tmp
);
336 * Is specified VNI present on the RT's list of "importing" VNIs?
338 static int is_vni_present_in_irt_vnis(struct list
*vnis
, struct bgpevpn
*vpn
)
340 struct listnode
*node
, *nnode
;
341 struct bgpevpn
*tmp_vpn
;
343 for (ALL_LIST_ELEMENTS(vnis
, node
, nnode
, tmp_vpn
)) {
352 * Compare Route Targets.
354 static int evpn_route_target_cmp(struct ecommunity
*ecom1
,
355 struct ecommunity
*ecom2
)
363 if (!ecom1
&& !ecom2
)
366 if (ecom1
->str
&& !ecom2
->str
)
369 if (!ecom1
->str
&& ecom2
->str
)
372 if (!ecom1
->str
&& !ecom2
->str
)
375 return strcmp(ecom1
->str
, ecom2
->str
);
378 static void evpn_xxport_delete_ecomm(void *val
)
380 struct ecommunity
*ecomm
= val
;
381 ecommunity_free(&ecomm
);
385 * Mask off global-admin field of specified extended community (RT),
386 * just retain the local-admin field.
388 static inline void mask_ecom_global_admin(struct ecommunity_val
*dst
,
389 struct ecommunity_val
*src
)
395 if (type
== ECOMMUNITY_ENCODE_AS
) {
396 dst
->val
[2] = dst
->val
[3] = 0;
397 } else if (type
== ECOMMUNITY_ENCODE_AS4
398 || type
== ECOMMUNITY_ENCODE_IP
) {
399 dst
->val
[2] = dst
->val
[3] = 0;
400 dst
->val
[4] = dst
->val
[5] = 0;
405 * Map one RT to specified VRF.
406 * bgp_vrf = BGP vrf instance
408 static void map_vrf_to_rt(struct bgp
*bgp_vrf
, struct ecommunity_val
*eval
)
410 struct vrf_irt_node
*irt
= NULL
;
411 struct ecommunity_val eval_tmp
;
413 /* If using "automatic" RT,
414 * we only care about the local-admin sub-field.
415 * This is to facilitate using L3VNI(VRF-VNI)
416 * as the RT for EBGP peering too.
418 memcpy(&eval_tmp
, eval
, ECOMMUNITY_SIZE
);
419 if (!CHECK_FLAG(bgp_vrf
->vrf_flags
, BGP_VRF_IMPORT_RT_CFGD
))
420 mask_ecom_global_admin(&eval_tmp
, eval
);
422 irt
= lookup_vrf_import_rt(&eval_tmp
);
423 if (irt
&& is_vrf_present_in_irt_vrfs(irt
->vrfs
, bgp_vrf
))
424 /* Already mapped. */
428 irt
= vrf_import_rt_new(&eval_tmp
);
430 /* Add VRF to the list for this RT. */
431 listnode_add(irt
->vrfs
, bgp_vrf
);
435 * Unmap specified VRF from specified RT. If there are no other
436 * VRFs for this RT, then the RT hash is deleted.
437 * bgp_vrf: BGP VRF specific instance
439 static void unmap_vrf_from_rt(struct bgp
*bgp_vrf
, struct vrf_irt_node
*irt
)
441 /* Delete VRF from list for this RT. */
442 listnode_delete(irt
->vrfs
, bgp_vrf
);
443 if (!listnode_head(irt
->vrfs
)) {
444 vrf_import_rt_free(irt
);
449 * Map one RT to specified VNI.
451 static void map_vni_to_rt(struct bgp
*bgp
, struct bgpevpn
*vpn
,
452 struct ecommunity_val
*eval
)
454 struct irt_node
*irt
;
455 struct ecommunity_val eval_tmp
;
457 /* If using "automatic" RT, we only care about the local-admin
459 * This is to facilitate using VNI as the RT for EBGP peering too.
461 memcpy(&eval_tmp
, eval
, ECOMMUNITY_SIZE
);
462 if (!is_import_rt_configured(vpn
))
463 mask_ecom_global_admin(&eval_tmp
, eval
);
465 irt
= lookup_import_rt(bgp
, &eval_tmp
);
467 if (is_vni_present_in_irt_vnis(irt
->vnis
, vpn
))
468 /* Already mapped. */
472 irt
= import_rt_new(bgp
, &eval_tmp
);
476 /* Add VNI to the hash list for this RT. */
477 listnode_add(irt
->vnis
, vpn
);
481 * Unmap specified VNI from specified RT. If there are no other
482 * VNIs for this RT, then the RT hash is deleted.
484 static void unmap_vni_from_rt(struct bgp
*bgp
, struct bgpevpn
*vpn
,
485 struct irt_node
*irt
)
487 /* Delete VNI from hash list for this RT. */
488 listnode_delete(irt
->vnis
, vpn
);
489 if (!listnode_head(irt
->vnis
)) {
490 import_rt_free(bgp
, irt
);
495 * Create RT extended community automatically from passed information:
496 * of the form AS:VNI.
497 * NOTE: We use only the lower 16 bits of the AS. This is sufficient as
498 * the need is to get a RT value that will be unique across different
499 * VNIs but the same across routers (in the same AS) for a particular
502 static void form_auto_rt(struct bgp
*bgp
, vni_t vni
, struct list
*rtl
)
504 struct ecommunity_val eval
;
505 struct ecommunity
*ecomadd
;
507 if (bgp
->advertise_autort_rfc8365
)
508 vni
|= EVPN_AUTORT_VXLAN
;
509 encode_route_target_as((bgp
->as
& 0xFFFF), vni
, &eval
);
511 ecomadd
= ecommunity_new();
512 ecommunity_add_val(ecomadd
, &eval
);
513 listnode_add_sort(rtl
, ecomadd
);
517 * Derive RD and RT for a VNI automatically. Invoked at the time of
520 static void derive_rd_rt_for_vni(struct bgp
*bgp
, struct bgpevpn
*vpn
)
522 bgp_evpn_derive_auto_rd(bgp
, vpn
);
523 bgp_evpn_derive_auto_rt_import(bgp
, vpn
);
524 bgp_evpn_derive_auto_rt_export(bgp
, vpn
);
528 * Convert nexthop (remote VTEP IP) into an IPv6 address.
530 static void evpn_convert_nexthop_to_ipv6(struct attr
*attr
)
532 if (BGP_ATTR_NEXTHOP_AFI_IP6(attr
))
534 ipv4_to_ipv4_mapped_ipv6(&attr
->mp_nexthop_global
, attr
->nexthop
);
535 attr
->mp_nexthop_len
= IPV6_MAX_BYTELEN
;
539 * Add (update) or delete MACIP from zebra.
541 static int bgp_zebra_send_remote_macip(struct bgp
*bgp
, struct bgpevpn
*vpn
,
542 struct prefix_evpn
*p
,
543 struct in_addr remote_vtep_ip
, int add
,
544 uint8_t flags
, uint32_t seq
)
548 char buf1
[ETHER_ADDR_STRLEN
];
549 char buf2
[INET6_ADDRSTRLEN
];
550 char buf3
[INET6_ADDRSTRLEN
];
553 if (!zclient
|| zclient
->sock
< 0)
556 /* Don't try to register if Zebra doesn't know of this instance. */
557 if (!IS_BGP_INST_KNOWN_TO_ZEBRA(bgp
))
563 zclient_create_header(
564 s
, add
? ZEBRA_REMOTE_MACIP_ADD
: ZEBRA_REMOTE_MACIP_DEL
,
566 stream_putl(s
, vpn
->vni
);
567 stream_put(s
, &p
->prefix
.macip_addr
.mac
.octet
, ETH_ALEN
); /* Mac Addr */
568 /* IP address length and IP address, if any. */
569 if (is_evpn_prefix_ipaddr_none(p
))
572 ipa_len
= is_evpn_prefix_ipaddr_v4(p
) ? IPV4_MAX_BYTELEN
574 stream_putl(s
, ipa_len
);
575 stream_put(s
, &p
->prefix
.macip_addr
.ip
.ip
.addr
, ipa_len
);
577 stream_put_in_addr(s
, &remote_vtep_ip
);
579 /* TX flags - MAC sticky status and/or gateway mac */
580 /* Also TX the sequence number of the best route. */
582 stream_putc(s
, flags
);
586 stream_putw_at(s
, 0, stream_get_endp(s
));
588 if (bgp_debug_zebra(NULL
))
590 "Tx %s MACIP, VNI %u MAC %s IP %s flags 0x%x seq %u remote VTEP %s",
591 add
? "ADD" : "DEL", vpn
->vni
,
592 prefix_mac2str(&p
->prefix
.macip_addr
.mac
,
594 ipaddr2str(&p
->prefix
.macip_addr
.ip
,
595 buf3
, sizeof(buf3
)), flags
, seq
,
596 inet_ntop(AF_INET
, &remote_vtep_ip
, buf2
,
599 return zclient_send_message(zclient
);
603 * Add (update) or delete remote VTEP from zebra.
605 static int bgp_zebra_send_remote_vtep(struct bgp
*bgp
, struct bgpevpn
*vpn
,
606 struct prefix_evpn
*p
, int add
)
611 if (!zclient
|| zclient
->sock
< 0)
614 /* Don't try to register if Zebra doesn't know of this instance. */
615 if (!IS_BGP_INST_KNOWN_TO_ZEBRA(bgp
))
621 zclient_create_header(
622 s
, add
? ZEBRA_REMOTE_VTEP_ADD
: ZEBRA_REMOTE_VTEP_DEL
,
624 stream_putl(s
, vpn
->vni
);
625 if (is_evpn_prefix_ipaddr_v4(p
))
626 stream_put_in_addr(s
, &p
->prefix
.imet_addr
.ip
.ipaddr_v4
);
627 else if (is_evpn_prefix_ipaddr_v6(p
)) {
630 "Bad remote IP when trying to %s remote VTEP for VNI %u",
631 add
? "ADD" : "DEL", vpn
->vni
);
635 stream_putw_at(s
, 0, stream_get_endp(s
));
637 if (bgp_debug_zebra(NULL
))
638 zlog_debug("Tx %s Remote VTEP, VNI %u remote VTEP %s",
639 add
? "ADD" : "DEL", vpn
->vni
,
640 inet_ntoa(p
->prefix
.imet_addr
.ip
.ipaddr_v4
));
642 return zclient_send_message(zclient
);
646 * Build extended community for EVPN ES (type-4) route
648 static void build_evpn_type4_route_extcomm(struct evpnes
*es
,
651 struct ecommunity ecom_encap
;
652 struct ecommunity ecom_es_rt
;
653 struct ecommunity_val eval
;
654 struct ecommunity_val eval_es_rt
;
655 bgp_encap_types tnl_type
;
659 tnl_type
= BGP_ENCAP_TYPE_VXLAN
;
660 memset(&ecom_encap
, 0, sizeof(ecom_encap
));
661 encode_encap_extcomm(tnl_type
, &eval
);
663 ecom_encap
.val
= (uint8_t *)eval
.val
;
664 attr
->ecommunity
= ecommunity_dup(&ecom_encap
);
667 memset(&mac
, 0, sizeof(struct ethaddr
));
668 memset(&ecom_es_rt
, 0, sizeof(ecom_es_rt
));
669 es_get_system_mac(&es
->esi
, &mac
);
670 encode_es_rt_extcomm(&eval_es_rt
, &mac
);
672 ecom_es_rt
.val
= (uint8_t *)eval_es_rt
.val
;
674 ecommunity_merge(attr
->ecommunity
, &ecom_es_rt
);
676 attr
->flag
|= ATTR_FLAG_BIT(BGP_ATTR_EXT_COMMUNITIES
);
680 * Build extended communities for EVPN prefix route.
682 static void build_evpn_type5_route_extcomm(struct bgp
*bgp_vrf
,
685 struct ecommunity ecom_encap
;
686 struct ecommunity ecom_rmac
;
687 struct ecommunity_val eval
;
688 struct ecommunity_val eval_rmac
;
689 bgp_encap_types tnl_type
;
690 struct listnode
*node
, *nnode
;
691 struct ecommunity
*ecom
;
692 struct list
*vrf_export_rtl
= NULL
;
695 tnl_type
= BGP_ENCAP_TYPE_VXLAN
;
696 memset(&ecom_encap
, 0, sizeof(ecom_encap
));
697 encode_encap_extcomm(tnl_type
, &eval
);
699 ecom_encap
.val
= (uint8_t *)eval
.val
;
702 attr
->ecommunity
= ecommunity_dup(&ecom_encap
);
704 /* Add the export RTs for L3VNI/VRF */
705 vrf_export_rtl
= bgp_vrf
->vrf_export_rtl
;
706 for (ALL_LIST_ELEMENTS(vrf_export_rtl
, node
, nnode
, ecom
))
708 ecommunity_merge(attr
->ecommunity
, ecom
);
710 /* add the router mac extended community */
711 if (!is_zero_mac(&attr
->rmac
)) {
712 memset(&ecom_rmac
, 0, sizeof(ecom_rmac
));
713 encode_rmac_extcomm(&eval_rmac
, &attr
->rmac
);
715 ecom_rmac
.val
= (uint8_t *)eval_rmac
.val
;
717 ecommunity_merge(attr
->ecommunity
, &ecom_rmac
);
720 attr
->flag
|= ATTR_FLAG_BIT(BGP_ATTR_EXT_COMMUNITIES
);
724 * Build extended communities for EVPN route.
725 * This function is applicable for type-2 and type-3 routes. The layer-2 RT
726 * and ENCAP extended communities are applicable for all routes.
727 * The default gateway extended community and MAC mobility (sticky) extended
728 * community are added as needed based on passed settings - only for type-2
729 * routes. Likewise, the layer-3 RT and Router MAC extended communities are
730 * added, if present, based on passed settings - only for non-link-local
733 static void build_evpn_route_extcomm(struct bgpevpn
*vpn
, struct attr
*attr
,
736 struct ecommunity ecom_encap
;
737 struct ecommunity ecom_sticky
;
738 struct ecommunity ecom_default_gw
;
739 struct ecommunity ecom_rmac
;
740 struct ecommunity ecom_na
;
741 struct ecommunity_val eval
;
742 struct ecommunity_val eval_sticky
;
743 struct ecommunity_val eval_default_gw
;
744 struct ecommunity_val eval_rmac
;
745 struct ecommunity_val eval_na
;
747 bgp_encap_types tnl_type
;
748 struct listnode
*node
, *nnode
;
749 struct ecommunity
*ecom
;
751 struct list
*vrf_export_rtl
= NULL
;
754 tnl_type
= BGP_ENCAP_TYPE_VXLAN
;
755 memset(&ecom_encap
, 0, sizeof(ecom_encap
));
756 encode_encap_extcomm(tnl_type
, &eval
);
758 ecom_encap
.val
= (uint8_t *)eval
.val
;
761 attr
->ecommunity
= ecommunity_dup(&ecom_encap
);
763 /* Add the export RTs for L2VNI */
764 for (ALL_LIST_ELEMENTS(vpn
->export_rtl
, node
, nnode
, ecom
))
765 attr
->ecommunity
= ecommunity_merge(attr
->ecommunity
, ecom
);
767 /* Add the export RTs for L3VNI if told to - caller determines
768 * when this should be done.
771 vrf_export_rtl
= bgpevpn_get_vrf_export_rtl(vpn
);
772 if (vrf_export_rtl
&& !list_isempty(vrf_export_rtl
)) {
773 for (ALL_LIST_ELEMENTS(vrf_export_rtl
, node
, nnode
,
775 attr
->ecommunity
= ecommunity_merge(
776 attr
->ecommunity
, ecom
);
780 /* Add MAC mobility (sticky) if needed. */
783 memset(&ecom_sticky
, 0, sizeof(ecom_sticky
));
784 encode_mac_mobility_extcomm(1, seqnum
, &eval_sticky
);
785 ecom_sticky
.size
= 1;
786 ecom_sticky
.val
= (uint8_t *)eval_sticky
.val
;
788 ecommunity_merge(attr
->ecommunity
, &ecom_sticky
);
791 /* Add RMAC, if told to. */
793 memset(&ecom_rmac
, 0, sizeof(ecom_rmac
));
794 encode_rmac_extcomm(&eval_rmac
, &attr
->rmac
);
796 ecom_rmac
.val
= (uint8_t *)eval_rmac
.val
;
798 ecommunity_merge(attr
->ecommunity
, &ecom_rmac
);
801 /* Add default gateway, if needed. */
802 if (attr
->default_gw
) {
803 memset(&ecom_default_gw
, 0, sizeof(ecom_default_gw
));
804 encode_default_gw_extcomm(&eval_default_gw
);
805 ecom_default_gw
.size
= 1;
806 ecom_default_gw
.val
= (uint8_t *)eval_default_gw
.val
;
808 ecommunity_merge(attr
->ecommunity
, &ecom_default_gw
);
811 if (attr
->router_flag
) {
812 memset(&ecom_na
, 0, sizeof(ecom_na
));
813 encode_na_flag_extcomm(&eval_na
, attr
->router_flag
);
815 ecom_na
.val
= (uint8_t *)eval_na
.val
;
816 attr
->ecommunity
= ecommunity_merge(attr
->ecommunity
,
820 attr
->flag
|= ATTR_FLAG_BIT(BGP_ATTR_EXT_COMMUNITIES
);
824 * Add MAC mobility extended community to attribute.
826 static void add_mac_mobility_to_attr(uint32_t seq_num
, struct attr
*attr
)
828 struct ecommunity ecom_tmp
;
829 struct ecommunity_val eval
;
830 uint8_t *ecom_val_ptr
;
837 encode_mac_mobility_extcomm(0, seq_num
, &eval
);
839 /* Find current MM ecommunity */
842 if (attr
->ecommunity
) {
843 for (i
= 0; i
< attr
->ecommunity
->size
; i
++) {
844 pnt
= attr
->ecommunity
->val
+ (i
* 8);
848 if (type
== ECOMMUNITY_ENCODE_EVPN
850 == ECOMMUNITY_EVPN_SUBTYPE_MACMOBILITY
) {
851 ecom_val_ptr
= (uint8_t *)(attr
->ecommunity
->val
858 /* Update the existing MM ecommunity */
860 memcpy(ecom_val_ptr
, eval
.val
, sizeof(char) * ECOMMUNITY_SIZE
);
862 /* Add MM to existing */
864 memset(&ecom_tmp
, 0, sizeof(ecom_tmp
));
866 ecom_tmp
.val
= (uint8_t *)eval
.val
;
868 if (attr
->ecommunity
)
870 ecommunity_merge(attr
->ecommunity
, &ecom_tmp
);
872 attr
->ecommunity
= ecommunity_dup(&ecom_tmp
);
876 /* Install EVPN route into zebra. */
877 static int evpn_zebra_install(struct bgp
*bgp
, struct bgpevpn
*vpn
,
878 struct prefix_evpn
*p
,
879 struct in_addr remote_vtep_ip
, uint8_t flags
,
884 if (p
->prefix
.route_type
== BGP_EVPN_MAC_IP_ROUTE
)
885 ret
= bgp_zebra_send_remote_macip(bgp
, vpn
, p
, remote_vtep_ip
,
888 ret
= bgp_zebra_send_remote_vtep(bgp
, vpn
, p
, 1);
893 /* Uninstall EVPN route from zebra. */
894 static int evpn_zebra_uninstall(struct bgp
*bgp
, struct bgpevpn
*vpn
,
895 struct prefix_evpn
*p
,
896 struct in_addr remote_vtep_ip
)
900 if (p
->prefix
.route_type
== BGP_EVPN_MAC_IP_ROUTE
)
901 ret
= bgp_zebra_send_remote_macip(bgp
, vpn
, p
, remote_vtep_ip
,
904 ret
= bgp_zebra_send_remote_vtep(bgp
, vpn
, p
, 0);
910 * Due to MAC mobility, the prior "local" best route has been supplanted
911 * by a "remote" best route. The prior route has to be deleted and withdrawn
914 static void evpn_delete_old_local_route(struct bgp
*bgp
, struct bgpevpn
*vpn
,
916 struct bgp_path_info
*old_local
)
918 struct bgp_node
*global_rn
;
919 struct bgp_path_info
*pi
;
920 afi_t afi
= AFI_L2VPN
;
921 safi_t safi
= SAFI_EVPN
;
923 /* Locate route node in the global EVPN routing table. Note that
924 * this table is a 2-level tree (RD-level + Prefix-level) similar to
927 global_rn
= bgp_afi_node_lookup(bgp
->rib
[afi
][safi
], afi
, safi
,
928 (struct prefix
*)&rn
->p
, &vpn
->prd
);
930 /* Delete route entry in the global EVPN table. */
931 delete_evpn_route_entry(bgp
, afi
, safi
, global_rn
, &pi
);
933 /* Schedule for processing - withdraws to peers happen from
937 bgp_process(bgp
, global_rn
, afi
, safi
);
938 bgp_unlock_node(global_rn
);
941 /* Delete route entry in the VNI route table, caller to remove. */
942 bgp_path_info_delete(rn
, old_local
);
945 static struct in_addr
*es_vtep_new(struct in_addr vtep
)
949 ip
= XCALLOC(MTYPE_BGP_EVPN_ES_VTEP
, sizeof(struct in_addr
));
953 ip
->s_addr
= vtep
.s_addr
;
957 static void es_vtep_free(struct in_addr
*ip
)
959 XFREE(MTYPE_BGP_EVPN_ES_VTEP
, ip
);
962 /* check if VTEP is already part of the list */
963 static int is_vtep_present_in_list(struct list
*list
,
966 struct listnode
*node
= NULL
;
969 for (ALL_LIST_ELEMENTS_RO(list
, node
, tmp
)) {
970 if (tmp
->s_addr
== vtep
.s_addr
)
977 * Best path for ES route was changed,
978 * update the list of VTEPs for this ES
980 static int evpn_es_install_vtep(struct bgp
*bgp
,
982 struct prefix_evpn
*p
,
983 struct in_addr rvtep
)
985 struct in_addr
*vtep_ip
;
987 if (is_vtep_present_in_list(es
->vtep_list
, rvtep
))
991 vtep_ip
= es_vtep_new(rvtep
);
993 listnode_add_sort(es
->vtep_list
, vtep_ip
);
998 * Best path for ES route was changed,
999 * update the list of VTEPs for this ES
1001 static int evpn_es_uninstall_vtep(struct bgp
*bgp
,
1003 struct prefix_evpn
*p
,
1004 struct in_addr rvtep
)
1006 struct listnode
*node
, *nnode
, *node_to_del
= NULL
;
1007 struct in_addr
*tmp
;
1009 for (ALL_LIST_ELEMENTS(es
->vtep_list
, node
, nnode
, tmp
)) {
1010 if (tmp
->s_addr
== rvtep
.s_addr
) {
1017 list_delete_node(es
->vtep_list
, node_to_del
);
1023 * Calculate the best path for a ES(type-4) route.
1025 static int evpn_es_route_select_install(struct bgp
*bgp
,
1027 struct bgp_node
*rn
)
1030 afi_t afi
= AFI_L2VPN
;
1031 safi_t safi
= SAFI_EVPN
;
1032 struct bgp_path_info
*old_select
; /* old best */
1033 struct bgp_path_info
*new_select
; /* new best */
1034 struct bgp_path_info_pair old_and_new
;
1036 /* Compute the best path. */
1037 bgp_best_selection(bgp
, rn
, &bgp
->maxpaths
[afi
][safi
],
1038 &old_and_new
, afi
, safi
);
1039 old_select
= old_and_new
.old
;
1040 new_select
= old_and_new
.new;
1043 * If the best path hasn't changed - see if something needs to be
1046 if (old_select
&& old_select
== new_select
1047 && old_select
->type
== ZEBRA_ROUTE_BGP
1048 && old_select
->sub_type
== BGP_ROUTE_IMPORTED
1049 && !CHECK_FLAG(rn
->flags
, BGP_NODE_USER_CLEAR
)
1050 && !CHECK_FLAG(old_select
->flags
, BGP_PATH_ATTR_CHANGED
)
1051 && !bgp
->addpath_tx_used
[afi
][safi
]) {
1052 if (bgp_zebra_has_route_changed(rn
, old_select
)) {
1053 ret
= evpn_es_install_vtep(bgp
, es
,
1054 (struct prefix_evpn
*)&rn
->p
,
1055 old_select
->attr
->nexthop
);
1057 UNSET_FLAG(old_select
->flags
, BGP_PATH_MULTIPATH_CHG
);
1058 bgp_zebra_clear_route_change_flags(rn
);
1062 /* If the user did a "clear" this flag will be set */
1063 UNSET_FLAG(rn
->flags
, BGP_NODE_USER_CLEAR
);
1066 * bestpath has changed; update relevant fields and install or uninstall
1067 * into the zebra RIB.
1069 if (old_select
|| new_select
)
1070 bgp_bump_version(rn
);
1073 bgp_path_info_unset_flag(rn
, old_select
, BGP_PATH_SELECTED
);
1075 bgp_path_info_set_flag(rn
, new_select
, BGP_PATH_SELECTED
);
1076 bgp_path_info_unset_flag(rn
, new_select
, BGP_PATH_ATTR_CHANGED
);
1077 UNSET_FLAG(new_select
->flags
, BGP_PATH_MULTIPATH_CHG
);
1080 if (new_select
&& new_select
->type
== ZEBRA_ROUTE_BGP
1081 && new_select
->sub_type
== BGP_ROUTE_IMPORTED
) {
1082 ret
= evpn_es_install_vtep(bgp
, es
,
1083 (struct prefix_evpn
*)&rn
->p
,
1084 new_select
->attr
->nexthop
);
1086 if (old_select
&& old_select
->type
== ZEBRA_ROUTE_BGP
1087 && old_select
->sub_type
== BGP_ROUTE_IMPORTED
)
1088 ret
= evpn_es_uninstall_vtep(
1089 bgp
, es
, (struct prefix_evpn
*)&rn
->p
,
1090 old_select
->attr
->nexthop
);
1093 /* Clear any route change flags. */
1094 bgp_zebra_clear_route_change_flags(rn
);
1096 /* Reap old select bgp_path_info, if it has been removed */
1097 if (old_select
&& CHECK_FLAG(old_select
->flags
, BGP_PATH_REMOVED
))
1098 bgp_path_info_reap(rn
, old_select
);
1104 * Calculate the best path for an EVPN route. Install/update best path in zebra,
1107 static int evpn_route_select_install(struct bgp
*bgp
, struct bgpevpn
*vpn
,
1108 struct bgp_node
*rn
)
1110 struct bgp_path_info
*old_select
, *new_select
;
1111 struct bgp_path_info_pair old_and_new
;
1112 struct prefix_evpn
*evp
;
1113 afi_t afi
= AFI_L2VPN
;
1114 safi_t safi
= SAFI_EVPN
;
1118 /* Compute the best path. */
1119 bgp_best_selection(bgp
, rn
, &bgp
->maxpaths
[afi
][safi
], &old_and_new
,
1121 old_select
= old_and_new
.old
;
1122 new_select
= old_and_new
.new;
1124 evp
= (struct prefix_evpn
*)&rn
->p
;
1125 /* If the best path hasn't changed - see if there is still something to
1129 if (old_select
&& old_select
== new_select
1130 && old_select
->type
== ZEBRA_ROUTE_BGP
1131 && old_select
->sub_type
== BGP_ROUTE_IMPORTED
1132 && !CHECK_FLAG(rn
->flags
, BGP_NODE_USER_CLEAR
)
1133 && !CHECK_FLAG(old_select
->flags
, BGP_PATH_ATTR_CHANGED
)
1134 && !bgp
->addpath_tx_used
[afi
][safi
]) {
1135 if (bgp_zebra_has_route_changed(rn
, old_select
)) {
1136 if (old_select
->attr
->sticky
)
1137 SET_FLAG(flags
, ZEBRA_MACIP_TYPE_STICKY
);
1138 if (old_select
->attr
->default_gw
)
1139 SET_FLAG(flags
, ZEBRA_MACIP_TYPE_GW
);
1140 if (is_evpn_prefix_ipaddr_v6(evp
) &&
1141 old_select
->attr
->router_flag
)
1142 SET_FLAG(flags
, ZEBRA_MACIP_TYPE_ROUTER_FLAG
);
1144 ret
= evpn_zebra_install(
1145 bgp
, vpn
, (struct prefix_evpn
*)&rn
->p
,
1146 old_select
->attr
->nexthop
, flags
,
1147 mac_mobility_seqnum(old_select
->attr
));
1149 UNSET_FLAG(old_select
->flags
, BGP_PATH_MULTIPATH_CHG
);
1150 bgp_zebra_clear_route_change_flags(rn
);
1154 /* If the user did a "clear" this flag will be set */
1155 UNSET_FLAG(rn
->flags
, BGP_NODE_USER_CLEAR
);
1157 /* bestpath has changed; update relevant fields and install or uninstall
1158 * into the zebra RIB.
1160 if (old_select
|| new_select
)
1161 bgp_bump_version(rn
);
1164 bgp_path_info_unset_flag(rn
, old_select
, BGP_PATH_SELECTED
);
1166 bgp_path_info_set_flag(rn
, new_select
, BGP_PATH_SELECTED
);
1167 bgp_path_info_unset_flag(rn
, new_select
, BGP_PATH_ATTR_CHANGED
);
1168 UNSET_FLAG(new_select
->flags
, BGP_PATH_MULTIPATH_CHG
);
1171 if (new_select
&& new_select
->type
== ZEBRA_ROUTE_BGP
1172 && new_select
->sub_type
== BGP_ROUTE_IMPORTED
) {
1174 if (new_select
->attr
->sticky
)
1175 SET_FLAG(flags
, ZEBRA_MACIP_TYPE_STICKY
);
1176 if (new_select
->attr
->default_gw
)
1177 SET_FLAG(flags
, ZEBRA_MACIP_TYPE_GW
);
1178 if (is_evpn_prefix_ipaddr_v6(evp
) &&
1179 new_select
->attr
->router_flag
)
1180 SET_FLAG(flags
, ZEBRA_MACIP_TYPE_ROUTER_FLAG
);
1182 ret
= evpn_zebra_install(bgp
, vpn
, (struct prefix_evpn
*)&rn
->p
,
1183 new_select
->attr
->nexthop
, flags
,
1184 mac_mobility_seqnum(new_select
->attr
));
1185 /* If an old best existed and it was a "local" route, the only
1187 * it would be supplanted is due to MAC mobility procedures. So,
1189 * need to do an implicit delete and withdraw that route from
1192 if (old_select
&& old_select
->peer
== bgp
->peer_self
1193 && old_select
->type
== ZEBRA_ROUTE_BGP
1194 && old_select
->sub_type
== BGP_ROUTE_STATIC
)
1195 evpn_delete_old_local_route(bgp
, vpn
, rn
, old_select
);
1197 if (old_select
&& old_select
->type
== ZEBRA_ROUTE_BGP
1198 && old_select
->sub_type
== BGP_ROUTE_IMPORTED
)
1199 ret
= evpn_zebra_uninstall(bgp
, vpn
,
1200 (struct prefix_evpn
*)&rn
->p
,
1201 old_select
->attr
->nexthop
);
1204 /* Clear any route change flags. */
1205 bgp_zebra_clear_route_change_flags(rn
);
1207 /* Reap old select bgp_path_info, if it has been removed */
1208 if (old_select
&& CHECK_FLAG(old_select
->flags
, BGP_PATH_REMOVED
))
1209 bgp_path_info_reap(rn
, old_select
);
1215 * Return true if the local ri for this rn is of type gateway mac
1217 static int evpn_route_is_def_gw(struct bgp
*bgp
, struct bgp_node
*rn
)
1219 struct bgp_path_info
*tmp_pi
= NULL
;
1220 struct bgp_path_info
*local_pi
= NULL
;
1223 for (tmp_pi
= rn
->info
; tmp_pi
; tmp_pi
= tmp_pi
->next
) {
1224 if (tmp_pi
->peer
== bgp
->peer_self
1225 && tmp_pi
->type
== ZEBRA_ROUTE_BGP
1226 && tmp_pi
->sub_type
== BGP_ROUTE_STATIC
)
1233 return local_pi
->attr
->default_gw
;
1238 * Return true if the local ri for this rn has sticky set
1240 static int evpn_route_is_sticky(struct bgp
*bgp
, struct bgp_node
*rn
)
1242 struct bgp_path_info
*tmp_pi
;
1243 struct bgp_path_info
*local_pi
;
1246 for (tmp_pi
= rn
->info
; tmp_pi
; tmp_pi
= tmp_pi
->next
) {
1247 if (tmp_pi
->peer
== bgp
->peer_self
1248 && tmp_pi
->type
== ZEBRA_ROUTE_BGP
1249 && tmp_pi
->sub_type
== BGP_ROUTE_STATIC
)
1256 return local_pi
->attr
->sticky
;
1260 * create or update EVPN type4 route entry.
1261 * This could be in the ES table or the global table.
1262 * TODO: handle remote ES (type4) routes as well
1264 static int update_evpn_type4_route_entry(struct bgp
*bgp
, struct evpnes
*es
,
1265 afi_t afi
, safi_t safi
,
1266 struct bgp_node
*rn
, struct attr
*attr
,
1267 int add
, struct bgp_path_info
**ri
,
1270 char buf
[ESI_STR_LEN
];
1271 char buf1
[INET6_ADDRSTRLEN
];
1272 struct bgp_path_info
*tmp_pi
= NULL
;
1273 struct bgp_path_info
*local_pi
= NULL
; /* local route entry if any */
1274 struct bgp_path_info
*remote_pi
= NULL
; /* remote route entry if any */
1275 struct attr
*attr_new
= NULL
;
1276 struct prefix_evpn
*evp
= NULL
;
1280 evp
= (struct prefix_evpn
*)&rn
->p
;
1282 /* locate the local and remote entries if any */
1283 for (tmp_pi
= rn
->info
; tmp_pi
; tmp_pi
= tmp_pi
->next
) {
1284 if (tmp_pi
->peer
== bgp
->peer_self
1285 && tmp_pi
->type
== ZEBRA_ROUTE_BGP
1286 && tmp_pi
->sub_type
== BGP_ROUTE_STATIC
)
1288 if (tmp_pi
->type
== ZEBRA_ROUTE_BGP
1289 && tmp_pi
->sub_type
== BGP_ROUTE_IMPORTED
1290 && CHECK_FLAG(tmp_pi
->flags
, BGP_PATH_VALID
))
1294 /* we don't expect to see a remote_pi at this point.
1295 * An ES route has esi + vtep_ip as the key,
1296 * We shouldn't see the same route from any other vtep.
1301 "%u ERROR: local es route for ESI: %s Vtep %s also learnt from remote",
1303 esi_to_str(&evp
->prefix
.es_addr
.esi
, buf
, sizeof(buf
)),
1304 ipaddr2str(&es
->originator_ip
, buf1
, sizeof(buf1
)));
1308 if (!local_pi
&& !add
)
1311 /* create or update the entry */
1314 /* Add or update attribute to hash */
1315 attr_new
= bgp_attr_intern(attr
);
1317 /* Create new route with its attribute. */
1318 tmp_pi
= info_make(ZEBRA_ROUTE_BGP
, BGP_ROUTE_STATIC
, 0,
1319 bgp
->peer_self
, attr_new
, rn
);
1320 SET_FLAG(tmp_pi
->flags
, BGP_PATH_VALID
);
1322 /* add the newly created path to the route-node */
1323 bgp_path_info_add(rn
, tmp_pi
);
1326 if (attrhash_cmp(tmp_pi
->attr
, attr
)
1327 && !CHECK_FLAG(tmp_pi
->flags
, BGP_PATH_REMOVED
))
1330 /* The attribute has changed.
1331 * Add (or update) attribute to hash. */
1332 attr_new
= bgp_attr_intern(attr
);
1333 bgp_path_info_set_flag(rn
, tmp_pi
,
1334 BGP_PATH_ATTR_CHANGED
);
1336 /* Restore route, if needed. */
1337 if (CHECK_FLAG(tmp_pi
->flags
, BGP_PATH_REMOVED
))
1338 bgp_path_info_restore(rn
, tmp_pi
);
1340 /* Unintern existing, set to new. */
1341 bgp_attr_unintern(&tmp_pi
->attr
);
1342 tmp_pi
->attr
= attr_new
;
1343 tmp_pi
->uptime
= bgp_clock();
1347 /* Return back the route entry. */
1352 /* update evpn es (type-4) route */
1353 static int update_evpn_type4_route(struct bgp
*bgp
,
1355 struct prefix_evpn
*p
)
1358 int route_changed
= 0;
1359 char buf
[ESI_STR_LEN
];
1360 char buf1
[INET6_ADDRSTRLEN
];
1361 afi_t afi
= AFI_L2VPN
;
1362 safi_t safi
= SAFI_EVPN
;
1364 struct attr
*attr_new
= NULL
;
1365 struct bgp_node
*rn
= NULL
;
1366 struct bgp_path_info
*pi
= NULL
;
1368 memset(&attr
, 0, sizeof(struct attr
));
1370 /* Build path-attribute for this route. */
1371 bgp_attr_default_set(&attr
, BGP_ORIGIN_IGP
);
1372 attr
.nexthop
= es
->originator_ip
.ipaddr_v4
;
1373 attr
.mp_nexthop_global_in
= es
->originator_ip
.ipaddr_v4
;
1374 attr
.mp_nexthop_len
= BGP_ATTR_NHLEN_IPV4
;
1376 /* Set up extended community. */
1377 build_evpn_type4_route_extcomm(es
, &attr
);
1379 /* First, create (or fetch) route node within the ESI. */
1380 /* NOTE: There is no RD here. */
1381 rn
= bgp_node_get(es
->route_table
, (struct prefix
*)p
);
1383 /* Create or update route entry. */
1384 ret
= update_evpn_type4_route_entry(bgp
, es
, afi
, safi
, rn
, &attr
, 1,
1385 &pi
, &route_changed
);
1387 flog_err(EC_BGP_ES_INVALID
,
1388 "%u ERROR: Failed to updated ES route ESI: %s VTEP %s",
1390 esi_to_str(&p
->prefix
.es_addr
.esi
, buf
, sizeof(buf
)),
1391 ipaddr2str(&es
->originator_ip
, buf1
, sizeof(buf1
)));
1395 attr_new
= pi
->attr
;
1397 /* Perform route selection;
1398 * this is just to set the flags correctly
1399 * as local route in the ES always wins.
1401 evpn_es_route_select_install(bgp
, es
, rn
);
1402 bgp_unlock_node(rn
);
1404 /* If this is a new route or some attribute has changed, export the
1405 * route to the global table. The route will be advertised to peers
1406 * from there. Note that this table is a 2-level tree (RD-level +
1407 * Prefix-level) similar to L3VPN routes.
1409 if (route_changed
) {
1410 struct bgp_path_info
*global_pi
;
1412 rn
= bgp_afi_node_get(bgp
->rib
[afi
][safi
], afi
, safi
,
1413 (struct prefix
*)p
, &es
->prd
);
1414 update_evpn_type4_route_entry(bgp
, es
, afi
, safi
, rn
, attr_new
,
1415 1, &global_pi
, &route_changed
);
1417 /* Schedule for processing and unlock node. */
1418 bgp_process(bgp
, rn
, afi
, safi
);
1419 bgp_unlock_node(rn
);
1422 /* Unintern temporary. */
1423 aspath_unintern(&attr
.aspath
);
1427 static int update_evpn_type5_route_entry(struct bgp
*bgp_def
,
1428 struct bgp
*bgp_vrf
, afi_t afi
,
1429 safi_t safi
, struct bgp_node
*rn
,
1430 struct attr
*attr
, int *route_changed
)
1432 struct attr
*attr_new
= NULL
;
1433 struct bgp_path_info
*pi
= NULL
;
1434 mpls_label_t label
= MPLS_INVALID_LABEL
;
1435 struct bgp_path_info
*local_pi
= NULL
;
1436 struct bgp_path_info
*tmp_pi
= NULL
;
1439 /* locate the local route entry if any */
1440 for (tmp_pi
= rn
->info
; tmp_pi
; tmp_pi
= tmp_pi
->next
) {
1441 if (tmp_pi
->peer
== bgp_def
->peer_self
1442 && tmp_pi
->type
== ZEBRA_ROUTE_BGP
1443 && tmp_pi
->sub_type
== BGP_ROUTE_STATIC
)
1448 * create a new route entry if one doesnt exist.
1449 * Otherwise see if route attr has changed
1453 /* route has changed as this is the first entry */
1456 /* Add (or update) attribute to hash. */
1457 attr_new
= bgp_attr_intern(attr
);
1459 /* create the route info from attribute */
1460 pi
= info_make(ZEBRA_ROUTE_BGP
, BGP_ROUTE_STATIC
, 0,
1461 bgp_def
->peer_self
, attr_new
, rn
);
1462 SET_FLAG(pi
->flags
, BGP_PATH_VALID
);
1464 /* Type-5 routes advertise the L3-VNI */
1465 bgp_path_info_extra_get(pi
);
1466 vni2label(bgp_vrf
->l3vni
, &label
);
1467 memcpy(&pi
->extra
->label
, &label
, sizeof(label
));
1468 pi
->extra
->num_labels
= 1;
1470 /* add the route entry to route node*/
1471 bgp_path_info_add(rn
, pi
);
1475 if (!attrhash_cmp(tmp_pi
->attr
, attr
)) {
1477 /* attribute changed */
1480 /* The attribute has changed. */
1481 /* Add (or update) attribute to hash. */
1482 attr_new
= bgp_attr_intern(attr
);
1483 bgp_path_info_set_flag(rn
, tmp_pi
,
1484 BGP_PATH_ATTR_CHANGED
);
1486 /* Restore route, if needed. */
1487 if (CHECK_FLAG(tmp_pi
->flags
, BGP_PATH_REMOVED
))
1488 bgp_path_info_restore(rn
, tmp_pi
);
1490 /* Unintern existing, set to new. */
1491 bgp_attr_unintern(&tmp_pi
->attr
);
1492 tmp_pi
->attr
= attr_new
;
1493 tmp_pi
->uptime
= bgp_clock();
1499 /* update evpn type-5 route entry */
1500 static int update_evpn_type5_route(struct bgp
*bgp_vrf
, struct prefix_evpn
*evp
,
1501 struct attr
*src_attr
)
1503 afi_t afi
= AFI_L2VPN
;
1504 safi_t safi
= SAFI_EVPN
;
1506 struct bgp_node
*rn
= NULL
;
1507 struct bgp
*bgp_def
= NULL
;
1508 int route_changed
= 0;
1510 bgp_def
= bgp_get_default();
1514 /* Build path attribute for this route - use the source attr, if
1515 * present, else treat as locally originated.
1518 bgp_attr_dup(&attr
, src_attr
);
1520 memset(&attr
, 0, sizeof(struct attr
));
1521 bgp_attr_default_set(&attr
, BGP_ORIGIN_IGP
);
1523 /* Set nexthop to ourselves and fill in the Router MAC. */
1524 attr
.nexthop
= bgp_vrf
->originator_ip
;
1525 attr
.mp_nexthop_global_in
= bgp_vrf
->originator_ip
;
1526 attr
.mp_nexthop_len
= BGP_ATTR_NHLEN_IPV4
;
1527 memcpy(&attr
.rmac
, &bgp_vrf
->rmac
, sizeof(struct ethaddr
));
1529 /* Setup RT and encap extended community */
1530 build_evpn_type5_route_extcomm(bgp_vrf
, &attr
);
1532 /* get the route node in global table */
1533 rn
= bgp_afi_node_get(bgp_def
->rib
[afi
][safi
], afi
, safi
,
1534 (struct prefix
*)evp
, &bgp_vrf
->vrf_prd
);
1537 /* create or update the route entry within the route node */
1538 update_evpn_type5_route_entry(bgp_def
, bgp_vrf
, afi
, safi
, rn
, &attr
,
1541 /* schedule for processing and unlock node */
1542 if (route_changed
) {
1543 bgp_process(bgp_def
, rn
, afi
, safi
);
1544 bgp_unlock_node(rn
);
1547 /* uninten temporary */
1549 aspath_unintern(&attr
.aspath
);
1554 * Create or update EVPN route entry. This could be in the VNI route table
1555 * or the global route table.
1557 static int update_evpn_route_entry(struct bgp
*bgp
, struct bgpevpn
*vpn
,
1558 afi_t afi
, safi_t safi
, struct bgp_node
*rn
,
1559 struct attr
*attr
, int add
,
1560 struct bgp_path_info
**pi
, uint8_t flags
,
1563 struct bgp_path_info
*tmp_pi
;
1564 struct bgp_path_info
*local_pi
;
1565 struct attr
*attr_new
;
1566 mpls_label_t label
[BGP_MAX_LABELS
];
1567 uint32_t num_labels
= 1;
1568 int route_change
= 1;
1570 struct prefix_evpn
*evp
;
1573 evp
= (struct prefix_evpn
*)&rn
->p
;
1574 memset(&label
, 0, sizeof(label
));
1576 /* See if this is an update of an existing route, or a new add. */
1578 for (tmp_pi
= rn
->info
; tmp_pi
; tmp_pi
= tmp_pi
->next
) {
1579 if (tmp_pi
->peer
== bgp
->peer_self
1580 && tmp_pi
->type
== ZEBRA_ROUTE_BGP
1581 && tmp_pi
->sub_type
== BGP_ROUTE_STATIC
)
1585 /* If route doesn't exist already, create a new one, if told to.
1586 * Otherwise act based on whether the attributes of the route have
1589 if (!local_pi
&& !add
)
1592 /* For non-GW MACs, update MAC mobility seq number, if needed. */
1593 if (seq
&& !CHECK_FLAG(flags
, ZEBRA_MACIP_TYPE_GW
))
1594 add_mac_mobility_to_attr(seq
, attr
);
1597 /* Add (or update) attribute to hash. */
1598 attr_new
= bgp_attr_intern(attr
);
1600 /* Extract MAC mobility sequence number, if any. */
1601 attr_new
->mm_seqnum
=
1602 bgp_attr_mac_mobility_seqnum(attr_new
, &sticky
);
1603 attr_new
->sticky
= sticky
;
1605 /* Create new route with its attribute. */
1606 tmp_pi
= info_make(ZEBRA_ROUTE_BGP
, BGP_ROUTE_STATIC
, 0,
1607 bgp
->peer_self
, attr_new
, rn
);
1608 SET_FLAG(tmp_pi
->flags
, BGP_PATH_VALID
);
1609 bgp_path_info_extra_get(tmp_pi
);
1611 /* The VNI goes into the 'label' field of the route */
1612 vni2label(vpn
->vni
, &label
[0]);
1614 /* Type-2 routes may carry a second VNI - the L3-VNI.
1615 * Only attach second label if we are advertising two labels for
1618 if (evp
->prefix
.route_type
== BGP_EVPN_MAC_IP_ROUTE
1619 && CHECK_FLAG(vpn
->flags
, VNI_FLAG_USE_TWO_LABELS
)) {
1622 l3vni
= bgpevpn_get_l3vni(vpn
);
1624 vni2label(l3vni
, &label
[1]);
1629 memcpy(&tmp_pi
->extra
->label
, label
, sizeof(label
));
1630 tmp_pi
->extra
->num_labels
= num_labels
;
1631 bgp_path_info_add(rn
, tmp_pi
);
1634 if (attrhash_cmp(tmp_pi
->attr
, attr
)
1635 && !CHECK_FLAG(tmp_pi
->flags
, BGP_PATH_REMOVED
))
1639 * The attributes have changed, type-2 routes needs to
1640 * be advertised with right labels.
1642 vni2label(vpn
->vni
, &label
[0]);
1643 if (evp
->prefix
.route_type
== BGP_EVPN_MAC_IP_ROUTE
1644 && CHECK_FLAG(vpn
->flags
,
1645 VNI_FLAG_USE_TWO_LABELS
)) {
1648 l3vni
= bgpevpn_get_l3vni(vpn
);
1650 vni2label(l3vni
, &label
[1]);
1654 memcpy(&tmp_pi
->extra
->label
, label
, sizeof(label
));
1655 tmp_pi
->extra
->num_labels
= num_labels
;
1657 /* The attribute has changed. */
1658 /* Add (or update) attribute to hash. */
1659 attr_new
= bgp_attr_intern(attr
);
1660 bgp_path_info_set_flag(rn
, tmp_pi
,
1661 BGP_PATH_ATTR_CHANGED
);
1663 /* Extract MAC mobility sequence number, if any. */
1664 attr_new
->mm_seqnum
=
1665 bgp_attr_mac_mobility_seqnum(attr_new
, &sticky
);
1666 attr_new
->sticky
= sticky
;
1668 /* Restore route, if needed. */
1669 if (CHECK_FLAG(tmp_pi
->flags
, BGP_PATH_REMOVED
))
1670 bgp_path_info_restore(rn
, tmp_pi
);
1672 /* Unintern existing, set to new. */
1673 bgp_attr_unintern(&tmp_pi
->attr
);
1674 tmp_pi
->attr
= attr_new
;
1675 tmp_pi
->uptime
= bgp_clock();
1679 /* Return back the route entry. */
1681 return route_change
;
1685 * Create or update EVPN route (of type based on prefix) for specified VNI
1686 * and schedule for processing.
1688 static int update_evpn_route(struct bgp
*bgp
, struct bgpevpn
*vpn
,
1689 struct prefix_evpn
*p
, uint8_t flags
,
1692 struct bgp_node
*rn
;
1694 struct attr
*attr_new
;
1695 int add_l3_ecomm
= 0;
1696 struct bgp_path_info
*pi
;
1697 afi_t afi
= AFI_L2VPN
;
1698 safi_t safi
= SAFI_EVPN
;
1701 memset(&attr
, 0, sizeof(struct attr
));
1703 /* Build path-attribute for this route. */
1704 bgp_attr_default_set(&attr
, BGP_ORIGIN_IGP
);
1705 attr
.nexthop
= vpn
->originator_ip
;
1706 attr
.mp_nexthop_global_in
= vpn
->originator_ip
;
1707 attr
.mp_nexthop_len
= BGP_ATTR_NHLEN_IPV4
;
1708 attr
.sticky
= CHECK_FLAG(flags
, ZEBRA_MACIP_TYPE_STICKY
) ? 1 : 0;
1709 attr
.default_gw
= CHECK_FLAG(flags
, ZEBRA_MACIP_TYPE_GW
) ? 1 : 0;
1710 attr
.router_flag
= CHECK_FLAG(flags
,
1711 ZEBRA_MACIP_TYPE_ROUTER_FLAG
) ? 1 : 0;
1713 /* PMSI is only needed for type-3 routes */
1714 if (p
->prefix
.route_type
== BGP_EVPN_IMET_ROUTE
)
1715 attr
.flag
|= ATTR_FLAG_BIT(BGP_ATTR_PMSI_TUNNEL
);
1717 /* router mac is only needed for type-2 routes here. */
1718 if (p
->prefix
.route_type
== BGP_EVPN_MAC_IP_ROUTE
)
1719 bgpevpn_get_rmac(vpn
, &attr
.rmac
);
1720 vni2label(vpn
->vni
, &(attr
.label
));
1722 /* Include L3 VNI related RTs and RMAC for type-2 routes, if they're
1723 * IPv4 or IPv6 global addresses and we're advertising L3VNI with
1726 if (p
->prefix
.route_type
== BGP_EVPN_MAC_IP_ROUTE
&&
1727 (is_evpn_prefix_ipaddr_v4(p
) ||
1728 !IN6_IS_ADDR_LINKLOCAL(&p
->prefix
.macip_addr
.ip
.ipaddr_v6
)) &&
1729 CHECK_FLAG(vpn
->flags
, VNI_FLAG_USE_TWO_LABELS
) &&
1730 bgpevpn_get_l3vni(vpn
))
1733 /* Set up extended community. */
1734 build_evpn_route_extcomm(vpn
, &attr
, add_l3_ecomm
);
1736 /* First, create (or fetch) route node within the VNI. */
1737 /* NOTE: There is no RD here. */
1738 rn
= bgp_node_get(vpn
->route_table
, (struct prefix
*)p
);
1740 /* Create or update route entry. */
1741 route_change
= update_evpn_route_entry(bgp
, vpn
, afi
, safi
, rn
, &attr
,
1742 1, &pi
, flags
, seq
);
1744 attr_new
= pi
->attr
;
1746 /* Perform route selection; this is just to set the flags correctly
1747 * as local route in the VNI always wins.
1749 evpn_route_select_install(bgp
, vpn
, rn
);
1750 bgp_unlock_node(rn
);
1752 /* If this is a new route or some attribute has changed, export the
1753 * route to the global table. The route will be advertised to peers
1754 * from there. Note that this table is a 2-level tree (RD-level +
1755 * Prefix-level) similar to L3VPN routes.
1758 struct bgp_path_info
*global_pi
;
1760 rn
= bgp_afi_node_get(bgp
->rib
[afi
][safi
], afi
, safi
,
1761 (struct prefix
*)p
, &vpn
->prd
);
1762 update_evpn_route_entry(bgp
, vpn
, afi
, safi
, rn
, attr_new
, 1,
1763 &global_pi
, flags
, seq
);
1765 /* Schedule for processing and unlock node. */
1766 bgp_process(bgp
, rn
, afi
, safi
);
1767 bgp_unlock_node(rn
);
1770 /* Unintern temporary. */
1771 aspath_unintern(&attr
.aspath
);
1777 * Delete EVPN route entry.
1778 * The entry can be in ESI/VNI table or the global table.
1780 static void delete_evpn_route_entry(struct bgp
*bgp
, afi_t afi
, safi_t safi
,
1781 struct bgp_node
*rn
,
1782 struct bgp_path_info
**pi
)
1784 struct bgp_path_info
*tmp_pi
;
1788 /* Now, find matching route. */
1789 for (tmp_pi
= rn
->info
; tmp_pi
; tmp_pi
= tmp_pi
->next
)
1790 if (tmp_pi
->peer
== bgp
->peer_self
1791 && tmp_pi
->type
== ZEBRA_ROUTE_BGP
1792 && tmp_pi
->sub_type
== BGP_ROUTE_STATIC
)
1797 /* Mark route for delete. */
1799 bgp_path_info_delete(rn
, tmp_pi
);
1804 /* Delete EVPN ES (type-4) route */
1805 static int delete_evpn_type4_route(struct bgp
*bgp
,
1807 struct prefix_evpn
*p
)
1809 afi_t afi
= AFI_L2VPN
;
1810 safi_t safi
= SAFI_EVPN
;
1811 struct bgp_path_info
*pi
;
1812 struct bgp_node
*rn
= NULL
; /* rn in esi table */
1813 struct bgp_node
*global_rn
= NULL
; /* rn in global table */
1815 /* First, locate the route node within the ESI.
1816 * If it doesn't exist, ther is nothing to do.
1817 * Note: there is no RD here.
1819 rn
= bgp_node_lookup(es
->route_table
, (struct prefix
*)p
);
1823 /* Next, locate route node in the global EVPN routing table.
1824 * Note that this table is a 2-level tree (RD-level + Prefix-level)
1826 global_rn
= bgp_afi_node_lookup(bgp
->rib
[afi
][safi
], afi
, safi
,
1827 (struct prefix
*)p
, &es
->prd
);
1830 /* Delete route entry in the global EVPN table. */
1831 delete_evpn_route_entry(bgp
, afi
, safi
, global_rn
, &pi
);
1833 /* Schedule for processing - withdraws to peers happen from
1837 bgp_process(bgp
, global_rn
, afi
, safi
);
1838 bgp_unlock_node(global_rn
);
1842 * Delete route entry in the ESI route table.
1843 * This can just be removed.
1845 delete_evpn_route_entry(bgp
, afi
, safi
, rn
, &pi
);
1847 bgp_path_info_reap(rn
, pi
);
1848 bgp_unlock_node(rn
);
1852 /* Delete EVPN type5 route */
1853 static int delete_evpn_type5_route(struct bgp
*bgp_vrf
, struct prefix_evpn
*evp
)
1855 afi_t afi
= AFI_L2VPN
;
1856 safi_t safi
= SAFI_EVPN
;
1857 struct bgp_node
*rn
= NULL
;
1858 struct bgp_path_info
*pi
= NULL
;
1859 struct bgp
*bgp_def
= NULL
; /* default bgp instance */
1861 bgp_def
= bgp_get_default();
1865 /* locate the global route entry for this type-5 prefix */
1866 rn
= bgp_afi_node_lookup(bgp_def
->rib
[afi
][safi
], afi
, safi
,
1867 (struct prefix
*)evp
, &bgp_vrf
->vrf_prd
);
1871 delete_evpn_route_entry(bgp_def
, afi
, safi
, rn
, &pi
);
1873 bgp_process(bgp_def
, rn
, afi
, safi
);
1874 bgp_unlock_node(rn
);
1879 * Delete EVPN route (of type based on prefix) for specified VNI and
1880 * schedule for processing.
1882 static int delete_evpn_route(struct bgp
*bgp
, struct bgpevpn
*vpn
,
1883 struct prefix_evpn
*p
)
1885 struct bgp_node
*rn
, *global_rn
;
1886 struct bgp_path_info
*pi
;
1887 afi_t afi
= AFI_L2VPN
;
1888 safi_t safi
= SAFI_EVPN
;
1890 /* First, locate the route node within the VNI. If it doesn't exist,
1892 * is nothing further to do.
1894 /* NOTE: There is no RD here. */
1895 rn
= bgp_node_lookup(vpn
->route_table
, (struct prefix
*)p
);
1899 /* Next, locate route node in the global EVPN routing table. Note that
1900 * this table is a 2-level tree (RD-level + Prefix-level) similar to
1903 global_rn
= bgp_afi_node_lookup(bgp
->rib
[afi
][safi
], afi
, safi
,
1904 (struct prefix
*)p
, &vpn
->prd
);
1906 /* Delete route entry in the global EVPN table. */
1907 delete_evpn_route_entry(bgp
, afi
, safi
, global_rn
, &pi
);
1909 /* Schedule for processing - withdraws to peers happen from
1913 bgp_process(bgp
, global_rn
, afi
, safi
);
1914 bgp_unlock_node(global_rn
);
1917 /* Delete route entry in the VNI route table. This can just be removed.
1919 delete_evpn_route_entry(bgp
, afi
, safi
, rn
, &pi
);
1921 bgp_path_info_reap(rn
, pi
);
1922 bgp_unlock_node(rn
);
1928 * Update all type-2 (MACIP) local routes for this VNI - these should also
1929 * be scheduled for advertise to peers.
1931 static int update_all_type2_routes(struct bgp
*bgp
, struct bgpevpn
*vpn
)
1935 struct bgp_node
*rn
;
1936 struct bgp_path_info
*pi
, *tmp_pi
;
1938 struct attr
*attr_new
;
1940 int add_l3_ecomm
= 0;
1945 /* Walk this VNI's route table and update local type-2 routes. For any
1946 * routes updated, update corresponding entry in the global table too.
1948 for (rn
= bgp_table_top(vpn
->route_table
); rn
;
1949 rn
= bgp_route_next(rn
)) {
1950 struct prefix_evpn
*evp
= (struct prefix_evpn
*)&rn
->p
;
1951 struct bgp_node
*rd_rn
;
1952 struct bgp_path_info
*global_pi
;
1954 if (evp
->prefix
.route_type
!= BGP_EVPN_MAC_IP_ROUTE
)
1957 /* Identify local route. */
1958 for (tmp_pi
= rn
->info
; tmp_pi
; tmp_pi
= tmp_pi
->next
) {
1959 if (tmp_pi
->peer
== bgp
->peer_self
1960 && tmp_pi
->type
== ZEBRA_ROUTE_BGP
1961 && tmp_pi
->sub_type
== BGP_ROUTE_STATIC
)
1969 * Build attribute per local route as the MAC mobility and
1970 * some other values could differ for different routes. The
1971 * attributes will be shared in the hash table.
1973 bgp_attr_default_set(&attr
, BGP_ORIGIN_IGP
);
1974 attr
.nexthop
= vpn
->originator_ip
;
1975 attr
.mp_nexthop_global_in
= vpn
->originator_ip
;
1976 attr
.mp_nexthop_len
= BGP_ATTR_NHLEN_IPV4
;
1977 bgpevpn_get_rmac(vpn
, &attr
.rmac
);
1979 if (evpn_route_is_sticky(bgp
, rn
))
1981 else if (evpn_route_is_def_gw(bgp
, rn
)) {
1982 attr
.default_gw
= 1;
1983 if (is_evpn_prefix_ipaddr_v6(evp
))
1984 attr
.router_flag
= 1;
1987 /* Add L3 VNI RTs and RMAC for non IPv6 link-local if
1988 * using L3 VNI for type-2 routes also.
1990 if ((is_evpn_prefix_ipaddr_v4(evp
) ||
1991 !IN6_IS_ADDR_LINKLOCAL(
1992 &evp
->prefix
.macip_addr
.ip
.ipaddr_v6
)) &&
1993 CHECK_FLAG(vpn
->flags
, VNI_FLAG_USE_TWO_LABELS
) &&
1994 bgpevpn_get_l3vni(vpn
))
1997 /* Set up extended community. */
1998 build_evpn_route_extcomm(vpn
, &attr
, add_l3_ecomm
);
2000 seq
= mac_mobility_seqnum(tmp_pi
->attr
);
2002 /* Update the route entry. */
2003 update_evpn_route_entry(bgp
, vpn
, afi
, safi
, rn
, &attr
, 0, &pi
,
2006 /* Perform route selection; this is just to set the flags
2007 * correctly as local route in the VNI always wins.
2009 evpn_route_select_install(bgp
, vpn
, rn
);
2011 attr_new
= pi
->attr
;
2013 /* Update route in global routing table. */
2014 rd_rn
= bgp_afi_node_get(bgp
->rib
[afi
][safi
], afi
, safi
,
2015 (struct prefix
*)evp
, &vpn
->prd
);
2017 update_evpn_route_entry(bgp
, vpn
, afi
, safi
, rd_rn
, attr_new
, 0,
2019 mac_mobility_seqnum(attr_new
));
2021 /* Schedule for processing and unlock node. */
2022 bgp_process(bgp
, rd_rn
, afi
, safi
);
2023 bgp_unlock_node(rd_rn
);
2025 /* Unintern temporary. */
2026 aspath_unintern(&attr
.aspath
);
2034 * Delete all type-2 (MACIP) local routes for this VNI - only from the
2035 * global routing table. These are also scheduled for withdraw from peers.
2037 static int delete_global_type2_routes(struct bgp
*bgp
, struct bgpevpn
*vpn
)
2041 struct bgp_node
*rdrn
, *rn
;
2042 struct bgp_table
*table
;
2043 struct bgp_path_info
*pi
;
2048 rdrn
= bgp_node_lookup(bgp
->rib
[afi
][safi
], (struct prefix
*)&vpn
->prd
);
2049 if (rdrn
&& rdrn
->info
) {
2050 table
= (struct bgp_table
*)rdrn
->info
;
2051 for (rn
= bgp_table_top(table
); rn
; rn
= bgp_route_next(rn
)) {
2052 struct prefix_evpn
*evp
= (struct prefix_evpn
*)&rn
->p
;
2054 if (evp
->prefix
.route_type
!= BGP_EVPN_MAC_IP_ROUTE
)
2057 delete_evpn_route_entry(bgp
, afi
, safi
, rn
, &pi
);
2059 bgp_process(bgp
, rn
, afi
, safi
);
2063 /* Unlock RD node. */
2065 bgp_unlock_node(rdrn
);
2071 * Delete all type-2 (MACIP) local routes for this VNI - from the global
2072 * table as well as the per-VNI route table.
2074 static int delete_all_type2_routes(struct bgp
*bgp
, struct bgpevpn
*vpn
)
2078 struct bgp_node
*rn
;
2079 struct bgp_path_info
*pi
;
2084 /* First, walk the global route table for this VNI's type-2 local
2086 * EVPN routes are a 2-level table, first get the RD table.
2088 delete_global_type2_routes(bgp
, vpn
);
2090 /* Next, walk this VNI's route table and delete local type-2 routes. */
2091 for (rn
= bgp_table_top(vpn
->route_table
); rn
;
2092 rn
= bgp_route_next(rn
)) {
2093 struct prefix_evpn
*evp
= (struct prefix_evpn
*)&rn
->p
;
2095 if (evp
->prefix
.route_type
!= BGP_EVPN_MAC_IP_ROUTE
)
2098 delete_evpn_route_entry(bgp
, afi
, safi
, rn
, &pi
);
2100 /* Route entry in local table gets deleted immediately. */
2102 bgp_path_info_reap(rn
, pi
);
2109 * Delete all routes in per ES route-table
2111 static int delete_all_es_routes(struct bgp
*bgp
, struct evpnes
*es
)
2113 struct bgp_node
*rn
;
2114 struct bgp_path_info
*pi
, *nextpi
;
2116 /* Walk this ES's route table and delete all routes. */
2117 for (rn
= bgp_table_top(es
->route_table
); rn
;
2118 rn
= bgp_route_next(rn
)) {
2119 for (pi
= rn
->info
; (pi
!= NULL
) && (nextpi
= pi
->next
, 1);
2121 bgp_path_info_delete(rn
, pi
);
2122 bgp_path_info_reap(rn
, pi
);
2130 * Delete all routes in the per-VNI route table.
2132 static int delete_all_vni_routes(struct bgp
*bgp
, struct bgpevpn
*vpn
)
2134 struct bgp_node
*rn
;
2135 struct bgp_path_info
*pi
, *nextpi
;
2137 /* Walk this VNI's route table and delete all routes. */
2138 for (rn
= bgp_table_top(vpn
->route_table
); rn
;
2139 rn
= bgp_route_next(rn
)) {
2140 for (pi
= rn
->info
; (pi
!= NULL
) && (nextpi
= pi
->next
, 1);
2142 bgp_path_info_delete(rn
, pi
);
2143 bgp_path_info_reap(rn
, pi
);
2151 * Update (and advertise) local routes for a VNI. Invoked upon the VNI
2152 * export RT getting modified or change to tunnel IP. Note that these
2153 * situations need the route in the per-VNI table as well as the global
2154 * table to be updated (as attributes change).
2156 static int update_routes_for_vni(struct bgp
*bgp
, struct bgpevpn
*vpn
)
2159 struct prefix_evpn p
;
2161 /* Update and advertise the type-3 route (only one) followed by the
2162 * locally learnt type-2 routes (MACIP) - for this VNI.
2164 * RT-3 only if doing head-end replication
2166 if (bgp
->vxlan_flood_ctrl
== VXLAN_FLOOD_HEAD_END_REPL
) {
2167 build_evpn_type3_prefix(&p
, vpn
->originator_ip
);
2168 ret
= update_evpn_route(bgp
, vpn
, &p
, 0, 0);
2173 return update_all_type2_routes(bgp
, vpn
);
2176 /* Delete (and withdraw) local routes for specified ES from global and ES table.
2177 * Also remove all other routes from the per ES table.
2178 * Invoked when ES is deleted.
2180 static int delete_routes_for_es(struct bgp
*bgp
, struct evpnes
*es
)
2183 char buf
[ESI_STR_LEN
];
2184 struct prefix_evpn p
;
2186 /* Delete and withdraw locally learnt ES route */
2187 build_evpn_type4_prefix(&p
, &es
->esi
, es
->originator_ip
.ipaddr_v4
);
2188 ret
= delete_evpn_type4_route(bgp
, es
, &p
);
2190 flog_err(EC_BGP_EVPN_ROUTE_DELETE
,
2191 "%u failed to delete type-4 route for ESI %s",
2192 bgp
->vrf_id
, esi_to_str(&es
->esi
, buf
, sizeof(buf
)));
2195 /* Delete all routes from per ES table */
2196 return delete_all_es_routes(bgp
, es
);
2200 * Delete (and withdraw) local routes for specified VNI from the global
2201 * table and per-VNI table. After this, remove all other routes from
2202 * the per-VNI table. Invoked upon the VNI being deleted or EVPN
2203 * (advertise-all-vni) being disabled.
2205 static int delete_routes_for_vni(struct bgp
*bgp
, struct bgpevpn
*vpn
)
2208 struct prefix_evpn p
;
2210 /* Delete and withdraw locally learnt type-2 routes (MACIP)
2211 * followed by type-3 routes (only one) - for this VNI.
2213 ret
= delete_all_type2_routes(bgp
, vpn
);
2217 build_evpn_type3_prefix(&p
, vpn
->originator_ip
);
2218 ret
= delete_evpn_route(bgp
, vpn
, &p
);
2222 /* Delete all routes from the per-VNI table. */
2223 return delete_all_vni_routes(bgp
, vpn
);
2227 * There is a tunnel endpoint IP address change for this VNI, delete
2228 * prior type-3 route (if needed) and update.
2229 * Note: Route re-advertisement happens elsewhere after other processing
2232 static int handle_tunnel_ip_change(struct bgp
*bgp
, struct bgpevpn
*vpn
,
2233 struct in_addr originator_ip
)
2235 struct prefix_evpn p
;
2237 /* If VNI is not live, we only need to update the originator ip */
2238 if (!is_vni_live(vpn
)) {
2239 vpn
->originator_ip
= originator_ip
;
2243 /* Update the tunnel-ip hash */
2244 bgp_tip_del(bgp
, &vpn
->originator_ip
);
2245 bgp_tip_add(bgp
, &originator_ip
);
2247 /* filter routes as martian nexthop db has changed */
2248 bgp_filter_evpn_routes_upon_martian_nh_change(bgp
);
2250 /* Need to withdraw type-3 route as the originator IP is part
2253 build_evpn_type3_prefix(&p
, vpn
->originator_ip
);
2254 delete_evpn_route(bgp
, vpn
, &p
);
2256 /* Update the tunnel IP and re-advertise all routes for this VNI. */
2257 vpn
->originator_ip
= originator_ip
;
2261 /* Install EVPN route entry in ES */
2262 static int install_evpn_route_entry_in_es(struct bgp
*bgp
, struct evpnes
*es
,
2263 struct prefix_evpn
*p
,
2264 struct bgp_path_info
*parent_pi
)
2267 struct bgp_node
*rn
= NULL
;
2268 struct bgp_path_info
*pi
= NULL
;
2269 struct attr
*attr_new
= NULL
;
2271 /* Create (or fetch) route within the VNI.
2272 * NOTE: There is no RD here.
2274 rn
= bgp_node_get(es
->route_table
, (struct prefix
*)p
);
2276 /* Check if route entry is already present. */
2277 for (pi
= rn
->info
; pi
; pi
= pi
->next
)
2279 && (struct bgp_path_info
*)pi
->extra
->parent
== parent_pi
)
2283 /* Add (or update) attribute to hash. */
2284 attr_new
= bgp_attr_intern(parent_pi
->attr
);
2286 /* Create new route with its attribute. */
2287 pi
= info_make(parent_pi
->type
, BGP_ROUTE_IMPORTED
, 0,
2288 parent_pi
->peer
, attr_new
, rn
);
2289 SET_FLAG(pi
->flags
, BGP_PATH_VALID
);
2290 bgp_path_info_extra_get(pi
);
2291 pi
->extra
->parent
= parent_pi
;
2292 bgp_path_info_add(rn
, pi
);
2294 if (attrhash_cmp(pi
->attr
, parent_pi
->attr
)
2295 && !CHECK_FLAG(pi
->flags
, BGP_PATH_REMOVED
)) {
2296 bgp_unlock_node(rn
);
2299 /* The attribute has changed. */
2300 /* Add (or update) attribute to hash. */
2301 attr_new
= bgp_attr_intern(parent_pi
->attr
);
2303 /* Restore route, if needed. */
2304 if (CHECK_FLAG(pi
->flags
, BGP_PATH_REMOVED
))
2305 bgp_path_info_restore(rn
, pi
);
2307 /* Mark if nexthop has changed. */
2308 if (!IPV4_ADDR_SAME(&pi
->attr
->nexthop
, &attr_new
->nexthop
))
2309 SET_FLAG(pi
->flags
, BGP_PATH_IGP_CHANGED
);
2311 /* Unintern existing, set to new. */
2312 bgp_attr_unintern(&pi
->attr
);
2313 pi
->attr
= attr_new
;
2314 pi
->uptime
= bgp_clock();
2317 /* Perform route selection and update zebra, if required. */
2318 ret
= evpn_es_route_select_install(bgp
, es
, rn
);
2323 * Install route entry into the VRF routing table and invoke route selection.
2325 static int install_evpn_route_entry_in_vrf(struct bgp
*bgp_vrf
,
2326 struct prefix_evpn
*evp
,
2327 struct bgp_path_info
*parent_pi
)
2329 struct bgp_node
*rn
;
2330 struct bgp_path_info
*pi
;
2332 struct attr
*attr_new
;
2335 struct prefix
*pp
= &p
;
2338 char buf
[PREFIX_STRLEN
];
2339 char buf1
[PREFIX_STRLEN
];
2341 memset(pp
, 0, sizeof(struct prefix
));
2342 ip_prefix_from_evpn_prefix(evp
, pp
);
2344 if (bgp_debug_zebra(NULL
)) {
2346 "installing evpn prefix %s as ip prefix %s in vrf %s",
2347 prefix2str(evp
, buf
, sizeof(buf
)),
2348 prefix2str(pp
, buf1
, sizeof(buf
)),
2349 vrf_id_to_name(bgp_vrf
->vrf_id
));
2352 /* Create (or fetch) route within the VRF. */
2353 /* NOTE: There is no RD here. */
2354 if (is_evpn_prefix_ipaddr_v4(evp
)) {
2356 safi
= SAFI_UNICAST
;
2357 rn
= bgp_node_get(bgp_vrf
->rib
[afi
][safi
], pp
);
2358 } else if (is_evpn_prefix_ipaddr_v6(evp
)) {
2360 safi
= SAFI_UNICAST
;
2361 rn
= bgp_node_get(bgp_vrf
->rib
[afi
][safi
], pp
);
2365 /* EVPN routes currently only support a IPv4 next hop which corresponds
2366 * to the remote VTEP. When importing into a VRF, if it is IPv6 host
2367 * or prefix route, we have to convert the next hop to an IPv4-mapped
2368 * address for the rest of the code to flow through. In the case of IPv4,
2369 * make sure to set the flag for next hop attribute.
2371 bgp_attr_dup(&attr
, parent_pi
->attr
);
2373 evpn_convert_nexthop_to_ipv6(&attr
);
2375 attr
.flag
|= ATTR_FLAG_BIT(BGP_ATTR_NEXT_HOP
);
2377 /* Check if route entry is already present. */
2378 for (pi
= rn
->info
; pi
; pi
= pi
->next
)
2380 && (struct bgp_path_info
*)pi
->extra
->parent
== parent_pi
)
2384 /* Add (or update) attribute to hash. */
2385 attr_new
= bgp_attr_intern(&attr
);
2387 /* Create new route with its attribute. */
2388 pi
= info_make(parent_pi
->type
, BGP_ROUTE_IMPORTED
, 0,
2389 parent_pi
->peer
, attr_new
, rn
);
2390 SET_FLAG(pi
->flags
, BGP_PATH_VALID
);
2391 bgp_path_info_extra_get(pi
);
2392 pi
->extra
->parent
= bgp_path_info_lock(parent_pi
);
2393 bgp_lock_node((struct bgp_node
*)parent_pi
->net
);
2394 if (parent_pi
->extra
) {
2395 memcpy(&pi
->extra
->label
, &parent_pi
->extra
->label
,
2396 sizeof(pi
->extra
->label
));
2397 pi
->extra
->num_labels
= parent_pi
->extra
->num_labels
;
2399 bgp_path_info_add(rn
, pi
);
2401 if (attrhash_cmp(pi
->attr
, &attr
)
2402 && !CHECK_FLAG(pi
->flags
, BGP_PATH_REMOVED
)) {
2403 bgp_unlock_node(rn
);
2406 /* The attribute has changed. */
2407 /* Add (or update) attribute to hash. */
2408 attr_new
= bgp_attr_intern(&attr
);
2410 /* Restore route, if needed. */
2411 if (CHECK_FLAG(pi
->flags
, BGP_PATH_REMOVED
))
2412 bgp_path_info_restore(rn
, pi
);
2414 /* Mark if nexthop has changed. */
2416 && !IPV4_ADDR_SAME(&pi
->attr
->nexthop
, &attr_new
->nexthop
))
2418 && !IPV6_ADDR_SAME(&pi
->attr
->mp_nexthop_global
,
2419 &attr_new
->mp_nexthop_global
)))
2420 SET_FLAG(pi
->flags
, BGP_PATH_IGP_CHANGED
);
2422 /* Unintern existing, set to new. */
2423 bgp_attr_unintern(&pi
->attr
);
2424 pi
->attr
= attr_new
;
2425 pi
->uptime
= bgp_clock();
2428 bgp_aggregate_increment(bgp_vrf
, &rn
->p
, pi
, afi
, safi
);
2430 /* Perform route selection and update zebra, if required. */
2431 bgp_process(bgp_vrf
, rn
, afi
, safi
);
2437 * Install route entry into the VNI routing table and invoke route selection.
2439 static int install_evpn_route_entry(struct bgp
*bgp
, struct bgpevpn
*vpn
,
2440 struct prefix_evpn
*p
,
2441 struct bgp_path_info
*parent_pi
)
2443 struct bgp_node
*rn
;
2444 struct bgp_path_info
*pi
;
2445 struct attr
*attr_new
;
2448 /* Create (or fetch) route within the VNI. */
2449 /* NOTE: There is no RD here. */
2450 rn
= bgp_node_get(vpn
->route_table
, (struct prefix
*)p
);
2452 /* Check if route entry is already present. */
2453 for (pi
= rn
->info
; pi
; pi
= pi
->next
)
2455 && (struct bgp_path_info
*)pi
->extra
->parent
== parent_pi
)
2459 /* Add (or update) attribute to hash. */
2460 attr_new
= bgp_attr_intern(parent_pi
->attr
);
2462 /* Create new route with its attribute. */
2463 pi
= info_make(parent_pi
->type
, BGP_ROUTE_IMPORTED
, 0,
2464 parent_pi
->peer
, attr_new
, rn
);
2465 SET_FLAG(pi
->flags
, BGP_PATH_VALID
);
2466 bgp_path_info_extra_get(pi
);
2467 pi
->extra
->parent
= bgp_path_info_lock(parent_pi
);
2468 bgp_lock_node((struct bgp_node
*)parent_pi
->net
);
2469 if (parent_pi
->extra
) {
2470 memcpy(&pi
->extra
->label
, &parent_pi
->extra
->label
,
2471 sizeof(pi
->extra
->label
));
2472 pi
->extra
->num_labels
= parent_pi
->extra
->num_labels
;
2474 bgp_path_info_add(rn
, pi
);
2476 if (attrhash_cmp(pi
->attr
, parent_pi
->attr
)
2477 && !CHECK_FLAG(pi
->flags
, BGP_PATH_REMOVED
)) {
2478 bgp_unlock_node(rn
);
2481 /* The attribute has changed. */
2482 /* Add (or update) attribute to hash. */
2483 attr_new
= bgp_attr_intern(parent_pi
->attr
);
2485 /* Restore route, if needed. */
2486 if (CHECK_FLAG(pi
->flags
, BGP_PATH_REMOVED
))
2487 bgp_path_info_restore(rn
, pi
);
2489 /* Mark if nexthop has changed. */
2490 if (!IPV4_ADDR_SAME(&pi
->attr
->nexthop
, &attr_new
->nexthop
))
2491 SET_FLAG(pi
->flags
, BGP_PATH_IGP_CHANGED
);
2493 /* Unintern existing, set to new. */
2494 bgp_attr_unintern(&pi
->attr
);
2495 pi
->attr
= attr_new
;
2496 pi
->uptime
= bgp_clock();
2499 /* Perform route selection and update zebra, if required. */
2500 ret
= evpn_route_select_install(bgp
, vpn
, rn
);
2505 /* Uninstall EVPN route entry from ES route table */
2506 static int uninstall_evpn_route_entry_in_es(struct bgp
*bgp
, struct evpnes
*es
,
2507 struct prefix_evpn
*p
,
2508 struct bgp_path_info
*parent_pi
)
2511 struct bgp_node
*rn
;
2512 struct bgp_path_info
*pi
;
2514 if (!es
->route_table
)
2517 /* Locate route within the ESI.
2518 * NOTE: There is no RD here.
2520 rn
= bgp_node_lookup(es
->route_table
, (struct prefix
*)p
);
2524 /* Find matching route entry. */
2525 for (pi
= rn
->info
; pi
; pi
= pi
->next
)
2527 && (struct bgp_path_info
*)pi
->extra
->parent
== parent_pi
)
2533 /* Mark entry for deletion */
2534 bgp_path_info_delete(rn
, pi
);
2536 /* Perform route selection and update zebra, if required. */
2537 ret
= evpn_es_route_select_install(bgp
, es
, rn
);
2539 /* Unlock route node. */
2540 bgp_unlock_node(rn
);
2546 * Uninstall route entry from the VRF routing table and send message
2547 * to zebra, if appropriate.
2549 static int uninstall_evpn_route_entry_in_vrf(struct bgp
*bgp_vrf
,
2550 struct prefix_evpn
*evp
,
2551 struct bgp_path_info
*parent_pi
)
2553 struct bgp_node
*rn
;
2554 struct bgp_path_info
*pi
;
2557 struct prefix
*pp
= &p
;
2560 char buf
[PREFIX_STRLEN
];
2561 char buf1
[PREFIX_STRLEN
];
2563 memset(pp
, 0, sizeof(struct prefix
));
2564 ip_prefix_from_evpn_prefix(evp
, pp
);
2566 if (bgp_debug_zebra(NULL
)) {
2568 "uninstalling evpn prefix %s as ip prefix %s in vrf %s",
2569 prefix2str(evp
, buf
, sizeof(buf
)),
2570 prefix2str(pp
, buf1
, sizeof(buf
)),
2571 vrf_id_to_name(bgp_vrf
->vrf_id
));
2574 /* Locate route within the VRF. */
2575 /* NOTE: There is no RD here. */
2576 if (is_evpn_prefix_ipaddr_v4(evp
)) {
2578 safi
= SAFI_UNICAST
;
2579 rn
= bgp_node_lookup(bgp_vrf
->rib
[afi
][safi
], pp
);
2582 safi
= SAFI_UNICAST
;
2583 rn
= bgp_node_lookup(bgp_vrf
->rib
[afi
][safi
], pp
);
2589 /* Find matching route entry. */
2590 for (pi
= rn
->info
; pi
; pi
= pi
->next
)
2592 && (struct bgp_path_info
*)pi
->extra
->parent
== parent_pi
)
2598 bgp_aggregate_decrement(bgp_vrf
, &rn
->p
, pi
, afi
, safi
);
2600 /* Mark entry for deletion */
2601 bgp_path_info_delete(rn
, pi
);
2603 /* Perform route selection and update zebra, if required. */
2604 bgp_process(bgp_vrf
, rn
, afi
, safi
);
2606 /* Unlock route node. */
2607 bgp_unlock_node(rn
);
2613 * Uninstall route entry from the VNI routing table and send message
2614 * to zebra, if appropriate.
2616 static int uninstall_evpn_route_entry(struct bgp
*bgp
, struct bgpevpn
*vpn
,
2617 struct prefix_evpn
*p
,
2618 struct bgp_path_info
*parent_pi
)
2620 struct bgp_node
*rn
;
2621 struct bgp_path_info
*pi
;
2624 /* Locate route within the VNI. */
2625 /* NOTE: There is no RD here. */
2626 rn
= bgp_node_lookup(vpn
->route_table
, (struct prefix
*)p
);
2630 /* Find matching route entry. */
2631 for (pi
= rn
->info
; pi
; pi
= pi
->next
)
2633 && (struct bgp_path_info
*)pi
->extra
->parent
== parent_pi
)
2639 /* Mark entry for deletion */
2640 bgp_path_info_delete(rn
, pi
);
2642 /* Perform route selection and update zebra, if required. */
2643 ret
= evpn_route_select_install(bgp
, vpn
, rn
);
2645 /* Unlock route node. */
2646 bgp_unlock_node(rn
);
2652 * Given a prefix, see if it belongs to ES.
2654 static int is_prefix_matching_for_es(struct prefix_evpn
*p
,
2657 /* if not an ES route return false */
2658 if (p
->prefix
.route_type
!= BGP_EVPN_ES_ROUTE
)
2661 if (memcmp(&p
->prefix
.es_addr
.esi
, &es
->esi
, sizeof(esi_t
)) == 0)
2668 * Given a route entry and a VRF, see if this route entry should be
2669 * imported into the VRF i.e., RTs match.
2671 static int is_route_matching_for_vrf(struct bgp
*bgp_vrf
,
2672 struct bgp_path_info
*pi
)
2674 struct attr
*attr
= pi
->attr
;
2675 struct ecommunity
*ecom
;
2679 /* Route should have valid RT to be even considered. */
2680 if (!(attr
->flag
& ATTR_FLAG_BIT(BGP_ATTR_EXT_COMMUNITIES
)))
2683 ecom
= attr
->ecommunity
;
2684 if (!ecom
|| !ecom
->size
)
2687 /* For each extended community RT, see if it matches this VNI. If any RT
2688 * matches, we're done.
2690 for (i
= 0; i
< ecom
->size
; i
++) {
2692 uint8_t type
, sub_type
;
2693 struct ecommunity_val
*eval
;
2694 struct ecommunity_val eval_tmp
;
2695 struct vrf_irt_node
*irt
;
2697 /* Only deal with RTs */
2698 pnt
= (ecom
->val
+ (i
* ECOMMUNITY_SIZE
));
2699 eval
= (struct ecommunity_val
*)(ecom
->val
2700 + (i
* ECOMMUNITY_SIZE
));
2703 if (sub_type
!= ECOMMUNITY_ROUTE_TARGET
)
2706 /* See if this RT matches specified VNIs import RTs */
2707 irt
= lookup_vrf_import_rt(eval
);
2709 if (is_vrf_present_in_irt_vrfs(irt
->vrfs
, bgp_vrf
))
2712 /* Also check for non-exact match. In this, we mask out the AS
2714 * only check on the local-admin sub-field. This is to
2716 * VNI as the RT for EBGP peering too.
2719 if (type
== ECOMMUNITY_ENCODE_AS
2720 || type
== ECOMMUNITY_ENCODE_AS4
2721 || type
== ECOMMUNITY_ENCODE_IP
) {
2722 memcpy(&eval_tmp
, eval
, ECOMMUNITY_SIZE
);
2723 mask_ecom_global_admin(&eval_tmp
, eval
);
2724 irt
= lookup_vrf_import_rt(&eval_tmp
);
2727 if (is_vrf_present_in_irt_vrfs(irt
->vrfs
, bgp_vrf
))
2735 * Given a route entry and a VNI, see if this route entry should be
2736 * imported into the VNI i.e., RTs match.
2738 static int is_route_matching_for_vni(struct bgp
*bgp
, struct bgpevpn
*vpn
,
2739 struct bgp_path_info
*pi
)
2741 struct attr
*attr
= pi
->attr
;
2742 struct ecommunity
*ecom
;
2746 /* Route should have valid RT to be even considered. */
2747 if (!(attr
->flag
& ATTR_FLAG_BIT(BGP_ATTR_EXT_COMMUNITIES
)))
2750 ecom
= attr
->ecommunity
;
2751 if (!ecom
|| !ecom
->size
)
2754 /* For each extended community RT, see if it matches this VNI. If any RT
2755 * matches, we're done.
2757 for (i
= 0; i
< ecom
->size
; i
++) {
2759 uint8_t type
, sub_type
;
2760 struct ecommunity_val
*eval
;
2761 struct ecommunity_val eval_tmp
;
2762 struct irt_node
*irt
;
2764 /* Only deal with RTs */
2765 pnt
= (ecom
->val
+ (i
* ECOMMUNITY_SIZE
));
2766 eval
= (struct ecommunity_val
*)(ecom
->val
2767 + (i
* ECOMMUNITY_SIZE
));
2770 if (sub_type
!= ECOMMUNITY_ROUTE_TARGET
)
2773 /* See if this RT matches specified VNIs import RTs */
2774 irt
= lookup_import_rt(bgp
, eval
);
2776 if (is_vni_present_in_irt_vnis(irt
->vnis
, vpn
))
2779 /* Also check for non-exact match. In this, we mask out the AS
2781 * only check on the local-admin sub-field. This is to
2783 * VNI as the RT for EBGP peering too.
2786 if (type
== ECOMMUNITY_ENCODE_AS
2787 || type
== ECOMMUNITY_ENCODE_AS4
2788 || type
== ECOMMUNITY_ENCODE_IP
) {
2789 memcpy(&eval_tmp
, eval
, ECOMMUNITY_SIZE
);
2790 mask_ecom_global_admin(&eval_tmp
, eval
);
2791 irt
= lookup_import_rt(bgp
, &eval_tmp
);
2794 if (is_vni_present_in_irt_vnis(irt
->vnis
, vpn
))
2801 static int install_uninstall_routes_for_es(struct bgp
*bgp
,
2808 char buf
[PREFIX_STRLEN
];
2809 char buf1
[ESI_STR_LEN
];
2810 struct bgp_node
*rd_rn
, *rn
;
2811 struct bgp_table
*table
;
2812 struct bgp_path_info
*pi
;
2818 * Walk entire global routing table and evaluate routes which could be
2819 * imported into this VRF. Note that we need to loop through all global
2820 * routes to determine which route matches the import rt on vrf
2822 for (rd_rn
= bgp_table_top(bgp
->rib
[afi
][safi
]); rd_rn
;
2823 rd_rn
= bgp_route_next(rd_rn
)) {
2824 table
= (struct bgp_table
*)(rd_rn
->info
);
2828 for (rn
= bgp_table_top(table
); rn
; rn
= bgp_route_next(rn
)) {
2829 struct prefix_evpn
*evp
= (struct prefix_evpn
*)&rn
->p
;
2831 for (pi
= rn
->info
; pi
; pi
= pi
->next
) {
2833 * Consider "valid" remote routes applicable for
2836 if (!(CHECK_FLAG(pi
->flags
, BGP_PATH_VALID
)
2837 && pi
->type
== ZEBRA_ROUTE_BGP
2838 && pi
->sub_type
== BGP_ROUTE_NORMAL
))
2841 if (!is_prefix_matching_for_es(evp
, es
))
2845 ret
= install_evpn_route_entry_in_es(
2848 ret
= uninstall_evpn_route_entry_in_es(
2854 "Failed to %s EVPN %s route in ESI %s",
2857 prefix2str(evp
, buf
,
2859 esi_to_str(&es
->esi
, buf1
,
2870 * Install or uninstall mac-ip routes are appropriate for this
2873 static int install_uninstall_routes_for_vrf(struct bgp
*bgp_vrf
, int install
)
2877 struct bgp_node
*rd_rn
, *rn
;
2878 struct bgp_table
*table
;
2879 struct bgp_path_info
*pi
;
2881 char buf
[PREFIX_STRLEN
];
2882 struct bgp
*bgp_def
= NULL
;
2886 bgp_def
= bgp_get_default();
2890 /* Walk entire global routing table and evaluate routes which could be
2891 * imported into this VRF. Note that we need to loop through all global
2892 * routes to determine which route matches the import rt on vrf
2894 for (rd_rn
= bgp_table_top(bgp_def
->rib
[afi
][safi
]); rd_rn
;
2895 rd_rn
= bgp_route_next(rd_rn
)) {
2896 table
= (struct bgp_table
*)(rd_rn
->info
);
2900 for (rn
= bgp_table_top(table
); rn
; rn
= bgp_route_next(rn
)) {
2901 struct prefix_evpn
*evp
= (struct prefix_evpn
*)&rn
->p
;
2903 /* if not mac-ip route skip this route */
2904 if (!(evp
->prefix
.route_type
== BGP_EVPN_MAC_IP_ROUTE
2905 || evp
->prefix
.route_type
2906 == BGP_EVPN_IP_PREFIX_ROUTE
))
2909 /* if not a mac+ip route skip this route */
2910 if (!(is_evpn_prefix_ipaddr_v4(evp
)
2911 || is_evpn_prefix_ipaddr_v6(evp
)))
2914 for (pi
= rn
->info
; pi
; pi
= pi
->next
) {
2915 /* Consider "valid" remote routes applicable for
2918 if (!(CHECK_FLAG(pi
->flags
, BGP_PATH_VALID
)
2919 && pi
->type
== ZEBRA_ROUTE_BGP
2920 && pi
->sub_type
== BGP_ROUTE_NORMAL
))
2923 if (is_route_matching_for_vrf(bgp_vrf
, pi
)) {
2925 ret
= install_evpn_route_entry_in_vrf(
2928 ret
= uninstall_evpn_route_entry_in_vrf(
2934 "Failed to %s EVPN %s route in VRF %s",
2937 prefix2str(evp
, buf
,
2952 * Install or uninstall routes of specified type that are appropriate for this
2955 static int install_uninstall_routes_for_vni(struct bgp
*bgp
,
2956 struct bgpevpn
*vpn
,
2957 bgp_evpn_route_type rtype
,
2962 struct bgp_node
*rd_rn
, *rn
;
2963 struct bgp_table
*table
;
2964 struct bgp_path_info
*pi
;
2970 /* Walk entire global routing table and evaluate routes which could be
2971 * imported into this VPN. Note that we cannot just look at the routes
2973 * the VNI's RD - remote routes applicable for this VNI could have any
2976 /* EVPN routes are a 2-level table. */
2977 for (rd_rn
= bgp_table_top(bgp
->rib
[afi
][safi
]); rd_rn
;
2978 rd_rn
= bgp_route_next(rd_rn
)) {
2979 table
= (struct bgp_table
*)(rd_rn
->info
);
2983 for (rn
= bgp_table_top(table
); rn
; rn
= bgp_route_next(rn
)) {
2984 struct prefix_evpn
*evp
= (struct prefix_evpn
*)&rn
->p
;
2986 if (evp
->prefix
.route_type
!= rtype
)
2989 for (pi
= rn
->info
; pi
; pi
= pi
->next
) {
2990 /* Consider "valid" remote routes applicable for
2992 if (!(CHECK_FLAG(pi
->flags
, BGP_PATH_VALID
)
2993 && pi
->type
== ZEBRA_ROUTE_BGP
2994 && pi
->sub_type
== BGP_ROUTE_NORMAL
))
2997 if (is_route_matching_for_vni(bgp
, vpn
, pi
)) {
2999 ret
= install_evpn_route_entry(
3002 ret
= uninstall_evpn_route_entry(
3008 "%u: Failed to %s EVPN %s route in VNI %u",
3012 rtype
== BGP_EVPN_MAC_IP_ROUTE
3026 /* Install any existing remote ES routes applicable for this ES into its routing
3027 * table. This is invoked when ES comes up.
3029 static int install_routes_for_es(struct bgp
*bgp
, struct evpnes
*es
)
3031 return install_uninstall_routes_for_es(bgp
, es
, 1);
3035 /* Install any existing remote routes applicable for this VRF into VRF RIB. This
3036 * is invoked upon l3vni-add or l3vni import rt change
3038 static int install_routes_for_vrf(struct bgp
*bgp_vrf
)
3040 install_uninstall_routes_for_vrf(bgp_vrf
, 1);
3045 * Install any existing remote routes applicable for this VNI into its
3046 * routing table. This is invoked when a VNI becomes "live" or its Import
3049 static int install_routes_for_vni(struct bgp
*bgp
, struct bgpevpn
*vpn
)
3053 /* Install type-3 routes followed by type-2 routes - the ones applicable
3056 ret
= install_uninstall_routes_for_vni(bgp
, vpn
, BGP_EVPN_IMET_ROUTE
,
3061 return install_uninstall_routes_for_vni(bgp
, vpn
, BGP_EVPN_MAC_IP_ROUTE
,
3065 /* uninstall routes from l3vni vrf. */
3066 static int uninstall_routes_for_vrf(struct bgp
*bgp_vrf
)
3068 install_uninstall_routes_for_vrf(bgp_vrf
, 0);
3073 * Uninstall any existing remote routes for this VNI. One scenario in which
3074 * this is invoked is upon an import RT change.
3076 static int uninstall_routes_for_vni(struct bgp
*bgp
, struct bgpevpn
*vpn
)
3080 /* Uninstall type-2 routes followed by type-3 routes - the ones
3084 ret
= install_uninstall_routes_for_vni(bgp
, vpn
, BGP_EVPN_MAC_IP_ROUTE
,
3089 return install_uninstall_routes_for_vni(bgp
, vpn
, BGP_EVPN_IMET_ROUTE
,
3093 /* Install or unistall route in ES */
3094 static int install_uninstall_route_in_es(struct bgp
*bgp
, struct evpnes
*es
,
3095 afi_t afi
, safi_t safi
,
3096 struct prefix_evpn
*evp
,
3097 struct bgp_path_info
*pi
, int install
)
3100 char buf
[ESI_STR_LEN
];
3103 ret
= install_evpn_route_entry_in_es(bgp
, es
, evp
, pi
);
3105 ret
= uninstall_evpn_route_entry_in_es(bgp
, es
, evp
, pi
);
3110 "%u: Failed to %s EVPN %s route in ESI %s", bgp
->vrf_id
,
3111 install
? "install" : "uninstall", "ES",
3112 esi_to_str(&evp
->prefix
.es_addr
.esi
, buf
, sizeof(buf
)));
3119 * Install or uninstall route in matching VRFs (list).
3121 static int install_uninstall_route_in_vrfs(struct bgp
*bgp_def
, afi_t afi
,
3122 safi_t safi
, struct prefix_evpn
*evp
,
3123 struct bgp_path_info
*pi
,
3124 struct list
*vrfs
, int install
)
3126 char buf
[PREFIX2STR_BUFFER
];
3127 struct bgp
*bgp_vrf
;
3128 struct listnode
*node
, *nnode
;
3130 /* Only type-2/type-5 routes go into a VRF */
3131 if (!(evp
->prefix
.route_type
== BGP_EVPN_MAC_IP_ROUTE
3132 || evp
->prefix
.route_type
== BGP_EVPN_IP_PREFIX_ROUTE
))
3135 /* if it is type-2 route and not a mac+ip route skip this route */
3136 if ((evp
->prefix
.route_type
== BGP_EVPN_MAC_IP_ROUTE
)
3137 && !(is_evpn_prefix_ipaddr_v4(evp
)
3138 || is_evpn_prefix_ipaddr_v6(evp
)))
3141 for (ALL_LIST_ELEMENTS(vrfs
, node
, nnode
, bgp_vrf
)) {
3145 ret
= install_evpn_route_entry_in_vrf(bgp_vrf
, evp
, pi
);
3147 ret
= uninstall_evpn_route_entry_in_vrf(bgp_vrf
, evp
,
3151 flog_err(EC_BGP_EVPN_FAIL
,
3152 "%u: Failed to %s prefix %s in VRF %s",
3154 install
? "install" : "uninstall",
3155 prefix2str(evp
, buf
, sizeof(buf
)),
3156 vrf_id_to_name(bgp_vrf
->vrf_id
));
3165 * Install or uninstall route in matching VNIs (list).
3167 static int install_uninstall_route_in_vnis(struct bgp
*bgp
, afi_t afi
,
3168 safi_t safi
, struct prefix_evpn
*evp
,
3169 struct bgp_path_info
*pi
,
3170 struct list
*vnis
, int install
)
3172 struct bgpevpn
*vpn
;
3173 struct listnode
*node
, *nnode
;
3175 for (ALL_LIST_ELEMENTS(vnis
, node
, nnode
, vpn
)) {
3178 if (!is_vni_live(vpn
))
3182 ret
= install_evpn_route_entry(bgp
, vpn
, evp
, pi
);
3184 ret
= uninstall_evpn_route_entry(bgp
, vpn
, evp
, pi
);
3187 flog_err(EC_BGP_EVPN_FAIL
,
3188 "%u: Failed to %s EVPN %s route in VNI %u",
3189 bgp
->vrf_id
, install
? "install" : "uninstall",
3190 evp
->prefix
.route_type
== BGP_EVPN_MAC_IP_ROUTE
3202 * Install or uninstall route for appropriate VNIs/ESIs.
3204 static int install_uninstall_evpn_route(struct bgp
*bgp
, afi_t afi
, safi_t safi
,
3206 struct bgp_path_info
*pi
, int import
)
3208 struct prefix_evpn
*evp
= (struct prefix_evpn
*)p
;
3209 struct attr
*attr
= pi
->attr
;
3210 struct ecommunity
*ecom
;
3215 /* Only type-2, type-3, type-4 and type-5 are supported currently */
3216 if (!(evp
->prefix
.route_type
== BGP_EVPN_MAC_IP_ROUTE
3217 || evp
->prefix
.route_type
== BGP_EVPN_IMET_ROUTE
3218 || evp
->prefix
.route_type
== BGP_EVPN_ES_ROUTE
3219 || evp
->prefix
.route_type
== BGP_EVPN_IP_PREFIX_ROUTE
))
3222 /* If we don't have Route Target, nothing much to do. */
3223 if (!(attr
->flag
& ATTR_FLAG_BIT(BGP_ATTR_EXT_COMMUNITIES
)))
3226 ecom
= attr
->ecommunity
;
3227 if (!ecom
|| !ecom
->size
)
3230 /* An EVPN route belongs to a VNI or a VRF or an ESI based on the RTs
3231 * attached to the route */
3232 for (i
= 0; i
< ecom
->size
; i
++) {
3234 uint8_t type
, sub_type
;
3235 struct ecommunity_val
*eval
;
3236 struct ecommunity_val eval_tmp
;
3237 struct irt_node
*irt
; /* import rt for l2vni */
3238 struct vrf_irt_node
*vrf_irt
; /* import rt for l3vni */
3241 /* Only deal with RTs */
3242 pnt
= (ecom
->val
+ (i
* ECOMMUNITY_SIZE
));
3243 eval
= (struct ecommunity_val
*)(ecom
->val
3244 + (i
* ECOMMUNITY_SIZE
));
3247 if (sub_type
!= ECOMMUNITY_ROUTE_TARGET
)
3251 * macip routes (type-2) are imported into VNI and VRF tables.
3252 * IMET route is imported into VNI table.
3253 * prefix routes are imported into VRF table.
3255 if (evp
->prefix
.route_type
== BGP_EVPN_MAC_IP_ROUTE
||
3256 evp
->prefix
.route_type
== BGP_EVPN_IMET_ROUTE
||
3257 evp
->prefix
.route_type
== BGP_EVPN_IP_PREFIX_ROUTE
) {
3259 irt
= lookup_import_rt(bgp
, eval
);
3261 install_uninstall_route_in_vnis(
3262 bgp
, afi
, safi
, evp
, pi
, irt
->vnis
,
3265 vrf_irt
= lookup_vrf_import_rt(eval
);
3267 install_uninstall_route_in_vrfs(
3268 bgp
, afi
, safi
, evp
, pi
, vrf_irt
->vrfs
,
3271 /* Also check for non-exact match.
3272 * In this, we mask out the AS and
3273 * only check on the local-admin sub-field.
3274 * This is to facilitate using
3275 * VNI as the RT for EBGP peering too.
3279 if (type
== ECOMMUNITY_ENCODE_AS
3280 || type
== ECOMMUNITY_ENCODE_AS4
3281 || type
== ECOMMUNITY_ENCODE_IP
) {
3282 memcpy(&eval_tmp
, eval
, ECOMMUNITY_SIZE
);
3283 mask_ecom_global_admin(&eval_tmp
, eval
);
3284 irt
= lookup_import_rt(bgp
, &eval_tmp
);
3285 vrf_irt
= lookup_vrf_import_rt(&eval_tmp
);
3289 install_uninstall_route_in_vnis(
3290 bgp
, afi
, safi
, evp
, pi
, irt
->vnis
,
3293 install_uninstall_route_in_vrfs(
3294 bgp
, afi
, safi
, evp
, pi
, vrf_irt
->vrfs
,
3298 /* es route is imported into the es table */
3299 if (evp
->prefix
.route_type
== BGP_EVPN_ES_ROUTE
) {
3301 /* we will match based on the entire esi to avoid
3302 * imoort of an es route for esi2 into esi1
3304 es
= bgp_evpn_lookup_es(bgp
, &evp
->prefix
.es_addr
.esi
);
3305 if (es
&& is_es_local(es
))
3306 install_uninstall_route_in_es(
3307 bgp
, es
, afi
, safi
, evp
, pi
, import
);
3315 * delete and withdraw all ipv4 and ipv6 routes in the vrf table as type-5
3318 static void delete_withdraw_vrf_routes(struct bgp
*bgp_vrf
)
3320 /* delete all ipv4 routes and withdraw from peers */
3321 if (advertise_type5_routes(bgp_vrf
, AFI_IP
))
3322 bgp_evpn_withdraw_type5_routes(bgp_vrf
, AFI_IP
, SAFI_UNICAST
);
3324 /* delete all ipv6 routes and withdraw from peers */
3325 if (advertise_type5_routes(bgp_vrf
, AFI_IP6
))
3326 bgp_evpn_withdraw_type5_routes(bgp_vrf
, AFI_IP6
, SAFI_UNICAST
);
3330 * update and advertise all ipv4 and ipv6 routes in thr vrf table as type-5
3333 static void update_advertise_vrf_routes(struct bgp
*bgp_vrf
)
3335 /* update all ipv4 routes */
3336 if (advertise_type5_routes(bgp_vrf
, AFI_IP
))
3337 bgp_evpn_advertise_type5_routes(bgp_vrf
, AFI_IP
, SAFI_UNICAST
);
3339 /* update all ipv6 routes */
3340 if (advertise_type5_routes(bgp_vrf
, AFI_IP6
))
3341 bgp_evpn_advertise_type5_routes(bgp_vrf
, AFI_IP6
, SAFI_UNICAST
);
3345 * update and advertise local routes for a VRF as type-5 routes.
3346 * This is invoked upon RD change for a VRF. Note taht the processing is only
3347 * done in the global route table using the routes which already exist in the
3350 static void update_router_id_vrf(struct bgp
*bgp_vrf
)
3352 /* skip if the RD is configured */
3353 if (is_vrf_rd_configured(bgp_vrf
))
3356 /* derive the RD for the VRF based on new router-id */
3357 bgp_evpn_derive_auto_rd_for_vrf(bgp_vrf
);
3359 /* update advertise ipv4|ipv6 routes as type-5 routes */
3360 update_advertise_vrf_routes(bgp_vrf
);
3364 * Delete and withdraw all type-5 routes for the RD corresponding to VRF.
3365 * This is invoked upon VRF RD change. The processing is done only from global
3368 static void withdraw_router_id_vrf(struct bgp
*bgp_vrf
)
3370 /* skip if the RD is configured */
3371 if (is_vrf_rd_configured(bgp_vrf
))
3374 /* delete/withdraw ipv4|ipv6 routes as type-5 routes */
3375 delete_withdraw_vrf_routes(bgp_vrf
);
3379 * Update and advertise local routes for a VNI. Invoked upon router-id
3380 * change. Note that the processing is done only on the global route table
3381 * using routes that already exist in the per-VNI table.
3383 static int update_advertise_vni_routes(struct bgp
*bgp
, struct bgpevpn
*vpn
)
3385 struct prefix_evpn p
;
3386 struct bgp_node
*rn
, *global_rn
;
3387 struct bgp_path_info
*pi
, *global_pi
;
3389 afi_t afi
= AFI_L2VPN
;
3390 safi_t safi
= SAFI_EVPN
;
3392 /* Locate type-3 route for VNI in the per-VNI table and use its
3393 * attributes to create and advertise the type-3 route for this VNI
3394 * in the global table.
3396 * RT-3 only if doing head-end replication
3398 if (bgp
->vxlan_flood_ctrl
== VXLAN_FLOOD_HEAD_END_REPL
) {
3399 build_evpn_type3_prefix(&p
, vpn
->originator_ip
);
3400 rn
= bgp_node_lookup(vpn
->route_table
, (struct prefix
*)&p
);
3401 if (!rn
) /* unexpected */
3403 for (pi
= rn
->info
; pi
; pi
= pi
->next
)
3404 if (pi
->peer
== bgp
->peer_self
&&
3405 pi
->type
== ZEBRA_ROUTE_BGP
3406 && pi
->sub_type
== BGP_ROUTE_STATIC
)
3408 if (!pi
) /* unexpected */
3412 global_rn
= bgp_afi_node_get(bgp
->rib
[afi
][safi
], afi
, safi
,
3413 (struct prefix
*)&p
, &vpn
->prd
);
3414 update_evpn_route_entry(bgp
, vpn
, afi
, safi
, global_rn
, attr
,
3415 1, &pi
, 0, mac_mobility_seqnum(attr
));
3417 /* Schedule for processing and unlock node. */
3418 bgp_process(bgp
, global_rn
, afi
, safi
);
3419 bgp_unlock_node(global_rn
);
3422 /* Now, walk this VNI's route table and use the route and its attribute
3423 * to create and schedule route in global table.
3425 for (rn
= bgp_table_top(vpn
->route_table
); rn
;
3426 rn
= bgp_route_next(rn
)) {
3427 struct prefix_evpn
*evp
= (struct prefix_evpn
*)&rn
->p
;
3429 /* Identify MAC-IP local routes. */
3430 if (evp
->prefix
.route_type
!= BGP_EVPN_MAC_IP_ROUTE
)
3433 for (pi
= rn
->info
; pi
; pi
= pi
->next
)
3434 if (pi
->peer
== bgp
->peer_self
3435 && pi
->type
== ZEBRA_ROUTE_BGP
3436 && pi
->sub_type
== BGP_ROUTE_STATIC
)
3441 /* Create route in global routing table using this route entry's
3445 global_rn
= bgp_afi_node_get(bgp
->rib
[afi
][safi
], afi
, safi
,
3446 (struct prefix
*)evp
, &vpn
->prd
);
3448 update_evpn_route_entry(bgp
, vpn
, afi
, safi
, global_rn
, attr
, 1,
3450 mac_mobility_seqnum(attr
));
3452 /* Schedule for processing and unlock node. */
3453 bgp_process(bgp
, global_rn
, afi
, safi
);
3454 bgp_unlock_node(global_rn
);
3461 * Delete (and withdraw) local routes for a VNI - only from the global
3462 * table. Invoked upon router-id change.
3464 static int delete_withdraw_vni_routes(struct bgp
*bgp
, struct bgpevpn
*vpn
)
3467 struct prefix_evpn p
;
3468 struct bgp_node
*global_rn
;
3469 struct bgp_path_info
*pi
;
3470 afi_t afi
= AFI_L2VPN
;
3471 safi_t safi
= SAFI_EVPN
;
3473 /* Delete and withdraw locally learnt type-2 routes (MACIP)
3474 * for this VNI - from the global table.
3476 ret
= delete_global_type2_routes(bgp
, vpn
);
3480 /* Remove type-3 route for this VNI from global table. */
3481 build_evpn_type3_prefix(&p
, vpn
->originator_ip
);
3482 global_rn
= bgp_afi_node_lookup(bgp
->rib
[afi
][safi
], afi
, safi
,
3483 (struct prefix
*)&p
, &vpn
->prd
);
3485 /* Delete route entry in the global EVPN table. */
3486 delete_evpn_route_entry(bgp
, afi
, safi
, global_rn
, &pi
);
3488 /* Schedule for processing - withdraws to peers happen from
3492 bgp_process(bgp
, global_rn
, afi
, safi
);
3493 bgp_unlock_node(global_rn
);
3500 * Handle router-id change. Update and advertise local routes corresponding
3501 * to this VNI from peers. Note that this is invoked after updating the
3502 * router-id. The routes in the per-VNI table are used to create routes in
3503 * the global table and schedule them.
3505 static void update_router_id_vni(struct hash_backet
*backet
, struct bgp
*bgp
)
3507 struct bgpevpn
*vpn
= (struct bgpevpn
*)backet
->data
;
3509 /* Skip VNIs with configured RD. */
3510 if (is_rd_configured(vpn
))
3513 bgp_evpn_derive_auto_rd(bgp
, vpn
);
3514 update_advertise_vni_routes(bgp
, vpn
);
3518 * Handle router-id change. Delete and withdraw local routes corresponding
3519 * to this VNI from peers. Note that this is invoked prior to updating
3520 * the router-id and is done only on the global route table, the routes
3521 * are needed in the per-VNI table to re-advertise with new router id.
3523 static void withdraw_router_id_vni(struct hash_backet
*backet
, struct bgp
*bgp
)
3525 struct bgpevpn
*vpn
= (struct bgpevpn
*)backet
->data
;
3527 /* Skip VNIs with configured RD. */
3528 if (is_rd_configured(vpn
))
3531 delete_withdraw_vni_routes(bgp
, vpn
);
3535 * Create RT-3 for a VNI and schedule for processing and advertisement.
3536 * This is invoked upon flooding mode changing to head-end replication.
3538 static void create_advertise_type3(struct hash_backet
*backet
, void *data
)
3540 struct bgpevpn
*vpn
= backet
->data
;
3541 struct bgp
*bgp
= data
;
3542 struct prefix_evpn p
;
3544 if (!vpn
|| !is_vni_live(vpn
))
3547 build_evpn_type3_prefix(&p
, vpn
->originator_ip
);
3548 if (update_evpn_route(bgp
, vpn
, &p
, 0, 0))
3549 flog_err(EC_BGP_EVPN_ROUTE_CREATE
,
3550 "Type3 route creation failure for VNI %u", vpn
->vni
);
3554 * Delete RT-3 for a VNI and schedule for processing and withdrawal.
3555 * This is invoked upon flooding mode changing to drop BUM packets.
3557 static void delete_withdraw_type3(struct hash_backet
*backet
, void *data
)
3559 struct bgpevpn
*vpn
= backet
->data
;
3560 struct bgp
*bgp
= data
;
3561 struct prefix_evpn p
;
3563 if (!vpn
|| !is_vni_live(vpn
))
3566 build_evpn_type3_prefix(&p
, vpn
->originator_ip
);
3567 delete_evpn_route(bgp
, vpn
, &p
);
3571 * Process received EVPN type-2 route (advertise or withdraw).
3573 static int process_type2_route(struct peer
*peer
, afi_t afi
, safi_t safi
,
3574 struct attr
*attr
, uint8_t *pfx
, int psize
,
3575 uint32_t addpath_id
)
3577 struct prefix_rd prd
;
3578 struct prefix_evpn p
;
3579 struct bgp_route_evpn evpn
;
3581 uint8_t macaddr_len
;
3582 mpls_label_t label
[BGP_MAX_LABELS
]; /* holds the VNI(s) as in packet */
3583 uint32_t num_labels
= 0;
3587 /* Type-2 route should be either 33, 37 or 49 bytes or an
3588 * additional 3 bytes if there is a second label (VNI):
3589 * RD (8), ESI (10), Eth Tag (4), MAC Addr Len (1),
3590 * MAC Addr (6), IP len (1), IP (0, 4 or 16),
3591 * MPLS Lbl1 (3), MPLS Lbl2 (0 or 3)
3593 if (psize
!= 33 && psize
!= 37 && psize
!= 49 && psize
!= 36
3594 && psize
!= 40 && psize
!= 52) {
3595 flog_err(EC_BGP_EVPN_ROUTE_INVALID
,
3596 "%u:%s - Rx EVPN Type-2 NLRI with invalid length %d",
3597 peer
->bgp
->vrf_id
, peer
->host
, psize
);
3601 memset(&evpn
, 0, sizeof(evpn
));
3603 /* Make prefix_rd */
3604 prd
.family
= AF_UNSPEC
;
3606 memcpy(&prd
.val
, pfx
, 8);
3609 /* Make EVPN prefix. */
3610 memset(&p
, 0, sizeof(struct prefix_evpn
));
3612 p
.prefixlen
= EVPN_ROUTE_PREFIXLEN
;
3613 p
.prefix
.route_type
= BGP_EVPN_MAC_IP_ROUTE
;
3615 /* Copy Ethernet Seg Identifier */
3616 memcpy(&evpn
.eth_s_id
.val
, pfx
, ESI_LEN
);
3619 /* Copy Ethernet Tag */
3620 memcpy(ð_tag
, pfx
, 4);
3621 p
.prefix
.macip_addr
.eth_tag
= ntohl(eth_tag
);
3624 /* Get the MAC Addr len */
3625 macaddr_len
= *pfx
++;
3627 /* Get the MAC Addr */
3628 if (macaddr_len
== (ETH_ALEN
* 8)) {
3629 memcpy(&p
.prefix
.macip_addr
.mac
.octet
, pfx
, ETH_ALEN
);
3633 EC_BGP_EVPN_ROUTE_INVALID
,
3634 "%u:%s - Rx EVPN Type-2 NLRI with unsupported MAC address length %d",
3635 peer
->bgp
->vrf_id
, peer
->host
, macaddr_len
);
3641 ipaddr_len
= *pfx
++;
3642 if (ipaddr_len
!= 0 && ipaddr_len
!= IPV4_MAX_BITLEN
3643 && ipaddr_len
!= IPV6_MAX_BITLEN
) {
3645 EC_BGP_EVPN_ROUTE_INVALID
,
3646 "%u:%s - Rx EVPN Type-2 NLRI with unsupported IP address length %d",
3647 peer
->bgp
->vrf_id
, peer
->host
, ipaddr_len
);
3652 ipaddr_len
/= 8; /* Convert to bytes. */
3653 p
.prefix
.macip_addr
.ip
.ipa_type
= (ipaddr_len
== IPV4_MAX_BYTELEN
)
3656 memcpy(&p
.prefix
.macip_addr
.ip
.ip
.addr
, pfx
, ipaddr_len
);
3660 /* Get the VNI(s). Stored as bytes here. */
3662 memset(label
, 0, sizeof(label
));
3663 memcpy(&label
[0], pfx
, BGP_LABEL_BYTES
);
3664 pfx
+= BGP_LABEL_BYTES
;
3665 psize
-= (33 + ipaddr_len
);
3666 /* Do we have a second VNI? */
3669 memcpy(&label
[1], pfx
, BGP_LABEL_BYTES
);
3671 * If in future, we are required to access additional fields,
3672 * we MUST increment pfx by BGP_LABEL_BYTES in before reading
3677 /* Process the route. */
3679 ret
= bgp_update(peer
, (struct prefix
*)&p
, addpath_id
, attr
,
3680 afi
, safi
, ZEBRA_ROUTE_BGP
, BGP_ROUTE_NORMAL
,
3681 &prd
, &label
[0], num_labels
, 0, &evpn
);
3683 ret
= bgp_withdraw(peer
, (struct prefix
*)&p
, addpath_id
, attr
,
3684 afi
, safi
, ZEBRA_ROUTE_BGP
, BGP_ROUTE_NORMAL
,
3685 &prd
, &label
[0], num_labels
, &evpn
);
3690 * Process received EVPN type-3 route (advertise or withdraw).
3692 static int process_type3_route(struct peer
*peer
, afi_t afi
, safi_t safi
,
3693 struct attr
*attr
, uint8_t *pfx
, int psize
,
3694 uint32_t addpath_id
)
3696 struct prefix_rd prd
;
3697 struct prefix_evpn p
;
3702 /* Type-3 route should be either 17 or 29 bytes: RD (8), Eth Tag (4),
3703 * IP len (1) and IP (4 or 16).
3705 if (psize
!= 17 && psize
!= 29) {
3706 flog_err(EC_BGP_EVPN_ROUTE_INVALID
,
3707 "%u:%s - Rx EVPN Type-3 NLRI with invalid length %d",
3708 peer
->bgp
->vrf_id
, peer
->host
, psize
);
3712 /* If PMSI is present, log if it is anything other than IR.
3713 * Note: We just simply ignore the values as it is not clear if
3714 * doing anything else is better.
3717 (attr
->flag
& ATTR_FLAG_BIT(BGP_ATTR_PMSI_TUNNEL
))) {
3718 if (attr
->pmsi_tnl_type
!= PMSI_TNLTYPE_INGR_REPL
) {
3720 EC_BGP_EVPN_PMSI_PRESENT
,
3721 "%u:%s - Rx EVPN Type-3 NLRI with unsupported PTA %d",
3722 peer
->bgp
->vrf_id
, peer
->host
,
3723 attr
->pmsi_tnl_type
);
3727 /* Make prefix_rd */
3728 prd
.family
= AF_UNSPEC
;
3730 memcpy(&prd
.val
, pfx
, 8);
3733 /* Make EVPN prefix. */
3734 memset(&p
, 0, sizeof(struct prefix_evpn
));
3736 p
.prefixlen
= EVPN_ROUTE_PREFIXLEN
;
3737 p
.prefix
.route_type
= BGP_EVPN_IMET_ROUTE
;
3739 /* Copy Ethernet Tag */
3740 memcpy(ð_tag
, pfx
, 4);
3741 p
.prefix
.imet_addr
.eth_tag
= ntohl(eth_tag
);
3745 ipaddr_len
= *pfx
++;
3746 if (ipaddr_len
== IPV4_MAX_BITLEN
) {
3747 p
.prefix
.imet_addr
.ip
.ipa_type
= IPADDR_V4
;
3748 memcpy(&p
.prefix
.imet_addr
.ip
.ip
.addr
, pfx
, IPV4_MAX_BYTELEN
);
3751 EC_BGP_EVPN_ROUTE_INVALID
,
3752 "%u:%s - Rx EVPN Type-3 NLRI with unsupported IP address length %d",
3753 peer
->bgp
->vrf_id
, peer
->host
, ipaddr_len
);
3757 /* Process the route. */
3759 ret
= bgp_update(peer
, (struct prefix
*)&p
, addpath_id
, attr
,
3760 afi
, safi
, ZEBRA_ROUTE_BGP
, BGP_ROUTE_NORMAL
,
3761 &prd
, NULL
, 0, 0, NULL
);
3763 ret
= bgp_withdraw(peer
, (struct prefix
*)&p
, addpath_id
, attr
,
3764 afi
, safi
, ZEBRA_ROUTE_BGP
, BGP_ROUTE_NORMAL
,
3765 &prd
, NULL
, 0, NULL
);
3770 * Process received EVPN type-4 route (advertise or withdraw).
3772 static int process_type4_route(struct peer
*peer
, afi_t afi
, safi_t safi
,
3773 struct attr
*attr
, uint8_t *pfx
, int psize
,
3774 uint32_t addpath_id
)
3779 struct in_addr vtep_ip
;
3780 struct prefix_rd prd
;
3781 struct prefix_evpn p
;
3783 /* Type-4 route should be either 23 or 35 bytes
3784 * RD (8), ESI (10), ip-len (1), ip (4 or 16)
3786 if (psize
!= 23 && psize
!= 35) {
3787 flog_err(EC_BGP_EVPN_ROUTE_INVALID
,
3788 "%u:%s - Rx EVPN Type-4 NLRI with invalid length %d",
3789 peer
->bgp
->vrf_id
, peer
->host
, psize
);
3793 /* Make prefix_rd */
3794 prd
.family
= AF_UNSPEC
;
3796 memcpy(&prd
.val
, pfx
, 8);
3800 memcpy(&esi
, pfx
, ESI_BYTES
);
3805 ipaddr_len
= *pfx
++;
3806 if (ipaddr_len
== IPV4_MAX_BITLEN
) {
3807 memcpy(&vtep_ip
, pfx
, IPV4_MAX_BYTELEN
);
3810 EC_BGP_EVPN_ROUTE_INVALID
,
3811 "%u:%s - Rx EVPN Type-4 NLRI with unsupported IP address length %d",
3812 peer
->bgp
->vrf_id
, peer
->host
, ipaddr_len
);
3816 build_evpn_type4_prefix(&p
, &esi
, vtep_ip
);
3817 /* Process the route. */
3819 ret
= bgp_update(peer
, (struct prefix
*)&p
, addpath_id
, attr
,
3820 afi
, safi
, ZEBRA_ROUTE_BGP
, BGP_ROUTE_NORMAL
,
3821 &prd
, NULL
, 0, 0, NULL
);
3823 ret
= bgp_withdraw(peer
, (struct prefix
*)&p
, addpath_id
, attr
,
3824 afi
, safi
, ZEBRA_ROUTE_BGP
, BGP_ROUTE_NORMAL
,
3825 &prd
, NULL
, 0, NULL
);
3832 * Process received EVPN type-5 route (advertise or withdraw).
3834 static int process_type5_route(struct peer
*peer
, afi_t afi
, safi_t safi
,
3835 struct attr
*attr
, uint8_t *pfx
, int psize
,
3836 uint32_t addpath_id
, int withdraw
)
3838 struct prefix_rd prd
;
3839 struct prefix_evpn p
;
3840 struct bgp_route_evpn evpn
;
3843 mpls_label_t label
; /* holds the VNI as in the packet */
3846 /* Type-5 route should be 34 or 58 bytes:
3847 * RD (8), ESI (10), Eth Tag (4), IP len (1), IP (4 or 16),
3848 * GW (4 or 16) and VNI (3).
3849 * Note that the IP and GW should both be IPv4 or both IPv6.
3851 if (psize
!= 34 && psize
!= 58) {
3852 flog_err(EC_BGP_EVPN_ROUTE_INVALID
,
3853 "%u:%s - Rx EVPN Type-5 NLRI with invalid length %d",
3854 peer
->bgp
->vrf_id
, peer
->host
, psize
);
3858 /* Make prefix_rd */
3859 prd
.family
= AF_UNSPEC
;
3861 memcpy(&prd
.val
, pfx
, 8);
3864 /* Make EVPN prefix. */
3865 memset(&p
, 0, sizeof(struct prefix_evpn
));
3867 p
.prefixlen
= EVPN_ROUTE_PREFIXLEN
;
3868 p
.prefix
.route_type
= BGP_EVPN_IP_PREFIX_ROUTE
;
3870 /* Additional information outside of prefix - ESI and GW IP */
3871 memset(&evpn
, 0, sizeof(evpn
));
3874 memcpy(&evpn
.eth_s_id
.val
, pfx
, 10);
3877 /* Fetch Ethernet Tag. */
3878 memcpy(ð_tag
, pfx
, 4);
3879 p
.prefix
.prefix_addr
.eth_tag
= ntohl(eth_tag
);
3882 /* Fetch IP prefix length. */
3884 if (ippfx_len
> IPV6_MAX_BITLEN
) {
3886 EC_BGP_EVPN_ROUTE_INVALID
,
3887 "%u:%s - Rx EVPN Type-5 NLRI with invalid IP Prefix length %d",
3888 peer
->bgp
->vrf_id
, peer
->host
, ippfx_len
);
3891 p
.prefix
.prefix_addr
.ip_prefix_length
= ippfx_len
;
3893 /* Determine IPv4 or IPv6 prefix */
3894 /* Since the address and GW are from the same family, this just becomes
3895 * a simple check on the total size.
3898 SET_IPADDR_V4(&p
.prefix
.prefix_addr
.ip
);
3899 memcpy(&p
.prefix
.prefix_addr
.ip
.ipaddr_v4
, pfx
, 4);
3901 memcpy(&evpn
.gw_ip
.ipv4
, pfx
, 4);
3904 SET_IPADDR_V6(&p
.prefix
.prefix_addr
.ip
);
3905 memcpy(&p
.prefix
.prefix_addr
.ip
.ipaddr_v6
, pfx
, 16);
3907 memcpy(&evpn
.gw_ip
.ipv6
, pfx
, 16);
3911 /* Get the VNI (in MPLS label field). Stored as bytes here. */
3912 memset(&label
, 0, sizeof(label
));
3913 memcpy(&label
, pfx
, BGP_LABEL_BYTES
);
3916 * If in future, we are required to access additional fields,
3917 * we MUST increment pfx by BGP_LABEL_BYTES in before reading the next
3921 /* Process the route. */
3923 ret
= bgp_update(peer
, (struct prefix
*)&p
, addpath_id
, attr
,
3924 afi
, safi
, ZEBRA_ROUTE_BGP
, BGP_ROUTE_NORMAL
,
3925 &prd
, &label
, 1, 0, &evpn
);
3927 ret
= bgp_withdraw(peer
, (struct prefix
*)&p
, addpath_id
, attr
,
3928 afi
, safi
, ZEBRA_ROUTE_BGP
, BGP_ROUTE_NORMAL
,
3929 &prd
, &label
, 1, &evpn
);
3934 static void evpn_mpattr_encode_type5(struct stream
*s
, struct prefix
*p
,
3935 struct prefix_rd
*prd
, mpls_label_t
*label
,
3936 uint32_t num_labels
, struct attr
*attr
)
3940 struct evpn_addr
*p_evpn_p
;
3942 memset(&temp
, 0, 16);
3943 if (p
->family
!= AF_EVPN
)
3945 p_evpn_p
= &(p
->u
.prefix_evpn
);
3947 /* len denites the total len of IP and GW-IP in the route
3948 IP and GW-IP have to be both ipv4 or ipv6
3950 if (IS_IPADDR_V4(&p_evpn_p
->prefix_addr
.ip
))
3951 len
= 8; /* IP and GWIP are both ipv4 */
3953 len
= 32; /* IP and GWIP are both ipv6 */
3954 /* Prefix contains RD, ESI, EthTag, IP length, IP, GWIP and VNI */
3955 stream_putc(s
, 8 + 10 + 4 + 1 + len
+ 3);
3956 stream_put(s
, prd
->val
, 8);
3958 stream_put(s
, &(attr
->evpn_overlay
.eth_s_id
), 10);
3960 stream_put(s
, &temp
, 10);
3961 stream_putl(s
, p_evpn_p
->prefix_addr
.eth_tag
);
3962 stream_putc(s
, p_evpn_p
->prefix_addr
.ip_prefix_length
);
3963 if (IS_IPADDR_V4(&p_evpn_p
->prefix_addr
.ip
))
3964 stream_put_ipv4(s
, p_evpn_p
->prefix_addr
.ip
.ipaddr_v4
.s_addr
);
3966 stream_put(s
, &p_evpn_p
->prefix_addr
.ip
.ipaddr_v6
, 16);
3968 if (IS_IPADDR_V4(&p_evpn_p
->prefix_addr
.ip
))
3970 attr
->evpn_overlay
.gw_ip
.ipv4
.s_addr
);
3972 stream_put(s
, &(attr
->evpn_overlay
.gw_ip
.ipv6
), 16);
3974 if (IS_IPADDR_V4(&p_evpn_p
->prefix_addr
.ip
))
3975 stream_put_ipv4(s
, 0);
3977 stream_put(s
, &temp
, 16);
3981 stream_put(s
, label
, 3);
3987 * Cleanup specific VNI upon EVPN (advertise-all-vni) being disabled.
3989 static void cleanup_vni_on_disable(struct hash_backet
*backet
, struct bgp
*bgp
)
3991 struct bgpevpn
*vpn
= (struct bgpevpn
*)backet
->data
;
3993 /* Remove EVPN routes and schedule for processing. */
3994 delete_routes_for_vni(bgp
, vpn
);
3996 /* Clear "live" flag and see if hash needs to be freed. */
3997 UNSET_FLAG(vpn
->flags
, VNI_FLAG_LIVE
);
3998 if (!is_vni_configured(vpn
))
3999 bgp_evpn_free(bgp
, vpn
);
4003 * Free a VNI entry; iterator function called during cleanup.
4005 static void free_vni_entry(struct hash_backet
*backet
, struct bgp
*bgp
)
4007 struct bgpevpn
*vpn
= (struct bgpevpn
*)backet
->data
;
4009 delete_all_vni_routes(bgp
, vpn
);
4010 bgp_evpn_free(bgp
, vpn
);
4014 * Derive AUTO import RT for BGP VRF - L3VNI
4016 static void evpn_auto_rt_import_add_for_vrf(struct bgp
*bgp_vrf
)
4018 struct bgp
*bgp_def
= NULL
;
4020 form_auto_rt(bgp_vrf
, bgp_vrf
->l3vni
, bgp_vrf
->vrf_import_rtl
);
4021 UNSET_FLAG(bgp_vrf
->vrf_flags
, BGP_VRF_IMPORT_RT_CFGD
);
4024 bgp_def
= bgp_get_default();
4027 bgp_evpn_map_vrf_to_its_rts(bgp_vrf
);
4031 * Delete AUTO import RT from BGP VRF - L3VNI
4033 static void evpn_auto_rt_import_delete_for_vrf(struct bgp
*bgp_vrf
)
4035 evpn_rt_delete_auto(bgp_vrf
, bgp_vrf
->l3vni
, bgp_vrf
->vrf_import_rtl
);
4039 * Derive AUTO export RT for BGP VRF - L3VNI
4041 static void evpn_auto_rt_export_add_for_vrf(struct bgp
*bgp_vrf
)
4043 UNSET_FLAG(bgp_vrf
->vrf_flags
, BGP_VRF_EXPORT_RT_CFGD
);
4044 form_auto_rt(bgp_vrf
, bgp_vrf
->l3vni
, bgp_vrf
->vrf_export_rtl
);
4048 * Delete AUTO export RT from BGP VRF - L3VNI
4050 static void evpn_auto_rt_export_delete_for_vrf(struct bgp
*bgp_vrf
)
4052 evpn_rt_delete_auto(bgp_vrf
, bgp_vrf
->l3vni
, bgp_vrf
->vrf_export_rtl
);
4055 static void bgp_evpn_handle_export_rt_change_for_vrf(struct bgp
*bgp_vrf
)
4057 struct bgp
*bgp_def
= NULL
;
4058 struct listnode
*node
= NULL
;
4059 struct bgpevpn
*vpn
= NULL
;
4061 bgp_def
= bgp_get_default();
4065 /* update all type-5 routes */
4066 update_advertise_vrf_routes(bgp_vrf
);
4068 /* update all type-2 routes */
4069 for (ALL_LIST_ELEMENTS_RO(bgp_vrf
->l2vnis
, node
, vpn
))
4070 update_routes_for_vni(bgp_def
, vpn
);
4074 * Handle autort change for a given VNI.
4076 static void update_autort_vni(struct hash_backet
*backet
, struct bgp
*bgp
)
4078 struct bgpevpn
*vpn
= backet
->data
;
4080 if (!is_import_rt_configured(vpn
)) {
4081 if (is_vni_live(vpn
))
4082 bgp_evpn_uninstall_routes(bgp
, vpn
);
4083 bgp_evpn_unmap_vni_from_its_rts(bgp
, vpn
);
4084 list_delete_all_node(vpn
->import_rtl
);
4085 bgp_evpn_derive_auto_rt_import(bgp
, vpn
);
4086 if (is_vni_live(vpn
))
4087 bgp_evpn_install_routes(bgp
, vpn
);
4089 if (!is_export_rt_configured(vpn
)) {
4090 list_delete_all_node(vpn
->export_rtl
);
4091 bgp_evpn_derive_auto_rt_export(bgp
, vpn
);
4092 if (is_vni_live(vpn
))
4093 bgp_evpn_handle_export_rt_change(bgp
, vpn
);
4101 /* withdraw type-5 route corresponding to ip prefix */
4102 void bgp_evpn_withdraw_type5_route(struct bgp
*bgp_vrf
, struct prefix
*p
,
4103 afi_t afi
, safi_t safi
)
4106 struct prefix_evpn evp
;
4107 char buf
[PREFIX_STRLEN
];
4109 build_type5_prefix_from_ip_prefix(&evp
, p
);
4110 ret
= delete_evpn_type5_route(bgp_vrf
, &evp
);
4113 EC_BGP_EVPN_ROUTE_DELETE
,
4114 "%u failed to delete type-5 route for prefix %s in vrf %s",
4115 bgp_vrf
->vrf_id
, prefix2str(p
, buf
, sizeof(buf
)),
4116 vrf_id_to_name(bgp_vrf
->vrf_id
));
4120 /* withdraw all type-5 routes for an address family */
4121 void bgp_evpn_withdraw_type5_routes(struct bgp
*bgp_vrf
, afi_t afi
, safi_t safi
)
4123 struct bgp_table
*table
= NULL
;
4124 struct bgp_node
*rn
= NULL
;
4125 struct bgp_path_info
*pi
;
4127 table
= bgp_vrf
->rib
[afi
][safi
];
4128 for (rn
= bgp_table_top(table
); rn
; rn
= bgp_route_next(rn
)) {
4129 /* Only care about "selected" routes - non-imported. */
4130 /* TODO: Support for AddPath for EVPN. */
4131 for (pi
= rn
->info
; pi
; pi
= pi
->next
) {
4132 if (CHECK_FLAG(pi
->flags
, BGP_PATH_SELECTED
)
4133 && (!pi
->extra
|| !pi
->extra
->parent
)) {
4134 bgp_evpn_withdraw_type5_route(bgp_vrf
, &rn
->p
,
4143 * Advertise IP prefix as type-5 route. The afi/safi and src_attr passed
4144 * to this function correspond to those of the source IP prefix (best
4145 * path in the case of the attr. In the case of a local prefix (when we
4146 * are advertising local subnets), the src_attr will be NULL.
4148 void bgp_evpn_advertise_type5_route(struct bgp
*bgp_vrf
, struct prefix
*p
,
4149 struct attr
*src_attr
, afi_t afi
,
4153 struct prefix_evpn evp
;
4154 char buf
[PREFIX_STRLEN
];
4156 build_type5_prefix_from_ip_prefix(&evp
, p
);
4157 ret
= update_evpn_type5_route(bgp_vrf
, &evp
, src_attr
);
4159 flog_err(EC_BGP_EVPN_ROUTE_CREATE
,
4160 "%u: Failed to create type-5 route for prefix %s",
4161 bgp_vrf
->vrf_id
, prefix2str(p
, buf
, sizeof(buf
)));
4164 /* Inject all prefixes of a particular address-family (currently, IPv4 or
4165 * IPv6 unicast) into EVPN as type-5 routes. This is invoked when the
4166 * advertisement is enabled.
4168 void bgp_evpn_advertise_type5_routes(struct bgp
*bgp_vrf
, afi_t afi
,
4171 struct bgp_table
*table
= NULL
;
4172 struct bgp_node
*rn
= NULL
;
4173 struct bgp_path_info
*pi
;
4175 table
= bgp_vrf
->rib
[afi
][safi
];
4176 for (rn
= bgp_table_top(table
); rn
; rn
= bgp_route_next(rn
)) {
4177 /* Need to identify the "selected" route entry to use its
4178 * attribute. Also, we only consider "non-imported" routes.
4179 * TODO: Support for AddPath for EVPN.
4181 for (pi
= rn
->info
; pi
; pi
= pi
->next
) {
4182 if (CHECK_FLAG(pi
->flags
, BGP_PATH_SELECTED
)
4183 && (!pi
->extra
|| !pi
->extra
->parent
)) {
4185 /* apply the route-map */
4186 if (bgp_vrf
->adv_cmd_rmap
[afi
][safi
].map
) {
4189 ret
= route_map_apply(
4190 bgp_vrf
->adv_cmd_rmap
[afi
][safi
]
4192 &rn
->p
, RMAP_BGP
, pi
);
4193 if (ret
== RMAP_DENYMATCH
)
4196 bgp_evpn_advertise_type5_route(
4197 bgp_vrf
, &rn
->p
, pi
->attr
, afi
, safi
);
4204 void evpn_rt_delete_auto(struct bgp
*bgp
, vni_t vni
, struct list
*rtl
)
4206 struct listnode
*node
, *nnode
, *node_to_del
;
4207 struct ecommunity
*ecom
, *ecom_auto
;
4208 struct ecommunity_val eval
;
4210 if (bgp
->advertise_autort_rfc8365
)
4211 vni
|= EVPN_AUTORT_VXLAN
;
4212 encode_route_target_as((bgp
->as
& 0xFFFF), vni
, &eval
);
4214 ecom_auto
= ecommunity_new();
4215 ecommunity_add_val(ecom_auto
, &eval
);
4218 for (ALL_LIST_ELEMENTS(rtl
, node
, nnode
, ecom
)) {
4219 if (ecommunity_match(ecom
, ecom_auto
)) {
4220 ecommunity_free(&ecom
);
4226 list_delete_node(rtl
, node_to_del
);
4228 ecommunity_free(&ecom_auto
);
4231 void bgp_evpn_configure_import_rt_for_vrf(struct bgp
*bgp_vrf
,
4232 struct ecommunity
*ecomadd
)
4234 /* uninstall routes from vrf */
4235 uninstall_routes_for_vrf(bgp_vrf
);
4237 /* Cleanup the RT to VRF mapping */
4238 bgp_evpn_unmap_vrf_from_its_rts(bgp_vrf
);
4240 /* Remove auto generated RT */
4241 evpn_auto_rt_import_delete_for_vrf(bgp_vrf
);
4243 /* Add the newly configured RT to RT list */
4244 listnode_add_sort(bgp_vrf
->vrf_import_rtl
, ecomadd
);
4245 SET_FLAG(bgp_vrf
->vrf_flags
, BGP_VRF_IMPORT_RT_CFGD
);
4247 /* map VRF to its RTs */
4248 bgp_evpn_map_vrf_to_its_rts(bgp_vrf
);
4250 /* install routes matching the new VRF */
4251 install_routes_for_vrf(bgp_vrf
);
4254 void bgp_evpn_unconfigure_import_rt_for_vrf(struct bgp
*bgp_vrf
,
4255 struct ecommunity
*ecomdel
)
4257 struct listnode
*node
= NULL
, *nnode
= NULL
, *node_to_del
= NULL
;
4258 struct ecommunity
*ecom
= NULL
;
4260 /* uninstall routes from vrf */
4261 uninstall_routes_for_vrf(bgp_vrf
);
4263 /* Cleanup the RT to VRF mapping */
4264 bgp_evpn_unmap_vrf_from_its_rts(bgp_vrf
);
4266 /* remove the RT from the RT list */
4267 for (ALL_LIST_ELEMENTS(bgp_vrf
->vrf_import_rtl
, node
, nnode
, ecom
)) {
4268 if (ecommunity_match(ecom
, ecomdel
)) {
4269 ecommunity_free(&ecom
);
4276 list_delete_node(bgp_vrf
->vrf_import_rtl
, node_to_del
);
4278 assert(bgp_vrf
->vrf_import_rtl
);
4279 /* fallback to auto import rt, if this was the last RT */
4280 if (bgp_vrf
->vrf_import_rtl
&& list_isempty(bgp_vrf
->vrf_import_rtl
)) {
4281 UNSET_FLAG(bgp_vrf
->vrf_flags
, BGP_VRF_IMPORT_RT_CFGD
);
4282 evpn_auto_rt_import_add_for_vrf(bgp_vrf
);
4285 /* map VRFs to its RTs */
4286 bgp_evpn_map_vrf_to_its_rts(bgp_vrf
);
4288 /* install routes matching this new RT */
4289 install_routes_for_vrf(bgp_vrf
);
4292 void bgp_evpn_configure_export_rt_for_vrf(struct bgp
*bgp_vrf
,
4293 struct ecommunity
*ecomadd
)
4295 /* remove auto-generated RT */
4296 evpn_auto_rt_export_delete_for_vrf(bgp_vrf
);
4298 /* Add the new RT to the RT list */
4299 listnode_add_sort(bgp_vrf
->vrf_export_rtl
, ecomadd
);
4300 SET_FLAG(bgp_vrf
->vrf_flags
, BGP_VRF_EXPORT_RT_CFGD
);
4302 bgp_evpn_handle_export_rt_change_for_vrf(bgp_vrf
);
4305 void bgp_evpn_unconfigure_export_rt_for_vrf(struct bgp
*bgp_vrf
,
4306 struct ecommunity
*ecomdel
)
4308 struct listnode
*node
= NULL
, *nnode
= NULL
, *node_to_del
= NULL
;
4309 struct ecommunity
*ecom
= NULL
;
4311 /* Remove the RT from the RT list */
4312 for (ALL_LIST_ELEMENTS(bgp_vrf
->vrf_export_rtl
, node
, nnode
, ecom
)) {
4313 if (ecommunity_match(ecom
, ecomdel
)) {
4314 ecommunity_free(&ecom
);
4321 list_delete_node(bgp_vrf
->vrf_export_rtl
, node_to_del
);
4324 * Temporary assert to make SA happy.
4325 * The ALL_LIST_ELEMENTS macro above has a NULL check
4326 * which means that SA is going to complain about
4327 * the list_isempty call, which doesn't NULL check.
4328 * So until we get this situation cleaned up, here
4331 assert(bgp_vrf
->vrf_export_rtl
);
4333 /* fall back to auto-generated RT if this was the last RT */
4334 if (list_isempty(bgp_vrf
->vrf_export_rtl
)) {
4335 UNSET_FLAG(bgp_vrf
->vrf_flags
, BGP_VRF_EXPORT_RT_CFGD
);
4336 evpn_auto_rt_export_add_for_vrf(bgp_vrf
);
4339 bgp_evpn_handle_export_rt_change_for_vrf(bgp_vrf
);
4343 * Handle change to BGP router id. This is invoked twice by the change
4344 * handler, first before the router id has been changed and then after
4345 * the router id has been changed. The first invocation will result in
4346 * local routes for all VNIs/VRF being deleted and withdrawn and the next
4347 * will result in the routes being re-advertised.
4349 void bgp_evpn_handle_router_id_update(struct bgp
*bgp
, int withdraw
)
4353 /* delete and withdraw all the type-5 routes
4354 stored in the global table for this vrf
4356 withdraw_router_id_vrf(bgp
);
4358 /* delete all the VNI routes (type-2/type-3) routes for all the
4361 hash_iterate(bgp
->vnihash
,
4362 (void (*)(struct hash_backet
*,
4363 void *))withdraw_router_id_vni
,
4367 /* advertise all routes in the vrf as type-5 routes with the new
4370 update_router_id_vrf(bgp
);
4372 /* advertise all the VNI routes (type-2/type-3) routes with the
4375 hash_iterate(bgp
->vnihash
,
4376 (void (*)(struct hash_backet
*,
4377 void *))update_router_id_vni
,
4383 * Handle change to auto-RT algorithm - update and advertise local routes.
4385 void bgp_evpn_handle_autort_change(struct bgp
*bgp
)
4387 hash_iterate(bgp
->vnihash
,
4388 (void (*)(struct hash_backet
*,
4389 void*))update_autort_vni
,
4394 * Handle change to export RT - update and advertise local routes.
4396 int bgp_evpn_handle_export_rt_change(struct bgp
*bgp
, struct bgpevpn
*vpn
)
4398 return update_routes_for_vni(bgp
, vpn
);
4401 void bgp_evpn_handle_vrf_rd_change(struct bgp
*bgp_vrf
, int withdraw
)
4404 delete_withdraw_vrf_routes(bgp_vrf
);
4406 update_advertise_vrf_routes(bgp_vrf
);
4410 * Handle change to RD. This is invoked twice by the change handler,
4411 * first before the RD has been changed and then after the RD has
4412 * been changed. The first invocation will result in local routes
4413 * of this VNI being deleted and withdrawn and the next will result
4414 * in the routes being re-advertised.
4416 void bgp_evpn_handle_rd_change(struct bgp
*bgp
, struct bgpevpn
*vpn
,
4420 delete_withdraw_vni_routes(bgp
, vpn
);
4422 update_advertise_vni_routes(bgp
, vpn
);
4426 * Install routes for this VNI. Invoked upon change to Import RT.
4428 int bgp_evpn_install_routes(struct bgp
*bgp
, struct bgpevpn
*vpn
)
4430 return install_routes_for_vni(bgp
, vpn
);
4434 * Uninstall all routes installed for this VNI. Invoked upon change
4437 int bgp_evpn_uninstall_routes(struct bgp
*bgp
, struct bgpevpn
*vpn
)
4439 return uninstall_routes_for_vni(bgp
, vpn
);
4443 * TODO: Hardcoded for a maximum of 2 VNIs right now
4445 char *bgp_evpn_label2str(mpls_label_t
*label
, uint32_t num_labels
, char *buf
,
4450 vni1
= label2vni(label
);
4451 if (num_labels
== 2) {
4452 vni2
= label2vni(label
+ 1);
4453 snprintf(buf
, len
, "%u/%u", vni1
, vni2
);
4455 snprintf(buf
, len
, "%u", vni1
);
4460 * Function to convert evpn route to json format.
4461 * NOTE: We don't use prefix2str as the output here is a bit different.
4463 void bgp_evpn_route2json(struct prefix_evpn
*p
, json_object
*json
)
4465 char buf1
[ETHER_ADDR_STRLEN
];
4466 char buf2
[PREFIX2STR_BUFFER
];
4471 if (p
->prefix
.route_type
== BGP_EVPN_IMET_ROUTE
) {
4472 json_object_int_add(json
, "routeType", p
->prefix
.route_type
);
4473 json_object_int_add(json
, "ethTag",
4474 p
->prefix
.imet_addr
.eth_tag
);
4475 json_object_int_add(json
, "ipLen",
4476 is_evpn_prefix_ipaddr_v4(p
)
4479 json_object_string_add(json
, "ip",
4480 inet_ntoa(p
->prefix
.imet_addr
.ip
.ipaddr_v4
));
4481 } else if (p
->prefix
.route_type
== BGP_EVPN_MAC_IP_ROUTE
) {
4482 if (is_evpn_prefix_ipaddr_none(p
)) {
4483 json_object_int_add(json
, "routeType",
4484 p
->prefix
.route_type
);
4485 json_object_int_add(json
, "ethTag",
4486 p
->prefix
.macip_addr
.eth_tag
);
4487 json_object_int_add(json
, "macLen", 8 * ETH_ALEN
);
4488 json_object_string_add(json
, "mac",
4489 prefix_mac2str(&p
->prefix
.macip_addr
.mac
,
4495 family
= is_evpn_prefix_ipaddr_v4(p
) ? AF_INET
4498 json_object_int_add(json
, "routeType",
4499 p
->prefix
.route_type
);
4500 json_object_int_add(json
, "ethTag",
4501 p
->prefix
.macip_addr
.eth_tag
);
4502 json_object_int_add(json
, "macLen", 8 * ETH_ALEN
);
4503 json_object_string_add(json
, "mac",
4504 prefix_mac2str(&p
->prefix
.macip_addr
.mac
,
4507 json_object_int_add(json
, "ipLen",
4508 is_evpn_prefix_ipaddr_v4(p
)
4511 json_object_string_add(
4514 &p
->prefix
.macip_addr
.ip
.ip
.addr
,
4516 PREFIX2STR_BUFFER
));
4519 /* Currently, this is to cater to other AF_ETHERNET code. */
4524 * Function to convert evpn route to string.
4525 * NOTE: We don't use prefix2str as the output here is a bit different.
4527 char *bgp_evpn_route2str(struct prefix_evpn
*p
, char *buf
, int len
)
4529 char buf1
[ETHER_ADDR_STRLEN
];
4530 char buf2
[PREFIX2STR_BUFFER
];
4531 char buf3
[ESI_STR_LEN
];
4533 if (p
->prefix
.route_type
== BGP_EVPN_IMET_ROUTE
) {
4534 snprintf(buf
, len
, "[%d]:[%d]:[%d]:[%s]", p
->prefix
.route_type
,
4535 p
->prefix
.imet_addr
.eth_tag
,
4536 is_evpn_prefix_ipaddr_v4(p
) ? IPV4_MAX_BITLEN
4538 inet_ntoa(p
->prefix
.imet_addr
.ip
.ipaddr_v4
));
4539 } else if (p
->prefix
.route_type
== BGP_EVPN_MAC_IP_ROUTE
) {
4540 if (is_evpn_prefix_ipaddr_none(p
))
4541 snprintf(buf
, len
, "[%d]:[%d]:[%d]:[%s]",
4542 p
->prefix
.route_type
,
4543 p
->prefix
.macip_addr
.eth_tag
,
4545 prefix_mac2str(&p
->prefix
.macip_addr
.mac
, buf1
,
4550 family
= is_evpn_prefix_ipaddr_v4(p
) ? AF_INET
4552 snprintf(buf
, len
, "[%d]:[%d]:[%d]:[%s]:[%d]:[%s]",
4553 p
->prefix
.route_type
,
4554 p
->prefix
.macip_addr
.eth_tag
,
4556 prefix_mac2str(&p
->prefix
.macip_addr
.mac
, buf1
,
4558 family
== AF_INET
? IPV4_MAX_BITLEN
4561 &p
->prefix
.macip_addr
.ip
.ip
.addr
,
4563 PREFIX2STR_BUFFER
));
4565 } else if (p
->prefix
.route_type
== BGP_EVPN_IP_PREFIX_ROUTE
) {
4566 snprintf(buf
, len
, "[%d]:[%d]:[%d]:[%s]",
4567 p
->prefix
.route_type
,
4568 p
->prefix
.prefix_addr
.eth_tag
,
4569 p
->prefix
.prefix_addr
.ip_prefix_length
,
4570 is_evpn_prefix_ipaddr_v4(p
)
4571 ? inet_ntoa(p
->prefix
.prefix_addr
.ip
.ipaddr_v4
)
4572 : inet6_ntoa(p
->prefix
.prefix_addr
.ip
.ipaddr_v6
));
4573 } else if (p
->prefix
.route_type
== BGP_EVPN_ES_ROUTE
) {
4574 snprintf(buf
, len
, "[%d]:[%s]:[%d]:[%s]",
4575 p
->prefix
.route_type
,
4576 esi_to_str(&p
->prefix
.es_addr
.esi
, buf3
, sizeof(buf3
)),
4577 is_evpn_prefix_ipaddr_v4(p
) ? IPV4_MAX_BITLEN
4579 inet_ntoa(p
->prefix
.es_addr
.ip
.ipaddr_v4
));
4581 /* For EVPN route types not supported yet. */
4582 snprintf(buf
, len
, "(unsupported route type %d)",
4583 p
->prefix
.route_type
);
4590 * Encode EVPN prefix in Update (MP_REACH)
4592 void bgp_evpn_encode_prefix(struct stream
*s
, struct prefix
*p
,
4593 struct prefix_rd
*prd
, mpls_label_t
*label
,
4594 uint32_t num_labels
, struct attr
*attr
,
4595 int addpath_encode
, uint32_t addpath_tx_id
)
4597 struct prefix_evpn
*evp
= (struct prefix_evpn
*)p
;
4598 int len
, ipa_len
= 0;
4601 stream_putl(s
, addpath_tx_id
);
4604 stream_putc(s
, evp
->prefix
.route_type
);
4606 switch (evp
->prefix
.route_type
) {
4607 case BGP_EVPN_MAC_IP_ROUTE
:
4608 if (is_evpn_prefix_ipaddr_v4(evp
))
4609 ipa_len
= IPV4_MAX_BYTELEN
;
4610 else if (is_evpn_prefix_ipaddr_v6(evp
))
4611 ipa_len
= IPV6_MAX_BYTELEN
;
4612 /* RD, ESI, EthTag, MAC+len, IP len, [IP], 1 VNI */
4613 len
= 8 + 10 + 4 + 1 + 6 + 1 + ipa_len
+ 3;
4614 if (ipa_len
&& num_labels
> 1) /* There are 2 VNIs */
4616 stream_putc(s
, len
);
4617 stream_put(s
, prd
->val
, 8); /* RD */
4619 stream_put(s
, &attr
->evpn_overlay
.eth_s_id
, ESI_LEN
);
4621 stream_put(s
, 0, 10);
4622 stream_putl(s
, evp
->prefix
.macip_addr
.eth_tag
); /* Ethernet Tag ID */
4623 stream_putc(s
, 8 * ETH_ALEN
); /* Mac Addr Len - bits */
4624 stream_put(s
, evp
->prefix
.macip_addr
.mac
.octet
, 6); /* Mac Addr */
4625 stream_putc(s
, 8 * ipa_len
); /* IP address Length */
4626 if (ipa_len
) /* IP */
4627 stream_put(s
, &evp
->prefix
.macip_addr
.ip
.ip
.addr
,
4629 /* 1st label is the L2 VNI */
4630 stream_put(s
, label
, BGP_LABEL_BYTES
);
4631 /* Include 2nd label (L3 VNI) if advertising MAC+IP */
4632 if (ipa_len
&& num_labels
> 1)
4633 stream_put(s
, label
+ 1, BGP_LABEL_BYTES
);
4636 case BGP_EVPN_IMET_ROUTE
:
4637 stream_putc(s
, 17); // TODO: length - assumes IPv4 address
4638 stream_put(s
, prd
->val
, 8); /* RD */
4639 stream_putl(s
, evp
->prefix
.imet_addr
.eth_tag
); /* Ethernet Tag ID */
4640 stream_putc(s
, IPV4_MAX_BITLEN
); /* IP address Length - bits */
4641 /* Originating Router's IP Addr */
4642 stream_put_in_addr(s
, &evp
->prefix
.imet_addr
.ip
.ipaddr_v4
);
4645 case BGP_EVPN_ES_ROUTE
:
4646 stream_putc(s
, 23); /* TODO: length: assumes ipv4 VTEP */
4647 stream_put(s
, prd
->val
, 8); /* RD */
4648 stream_put(s
, evp
->prefix
.es_addr
.esi
.val
, 10); /* ESI */
4649 stream_putc(s
, IPV4_MAX_BITLEN
); /* IP address Length - bits */
4651 stream_put_in_addr(s
, &evp
->prefix
.es_addr
.ip
.ipaddr_v4
);
4654 case BGP_EVPN_IP_PREFIX_ROUTE
:
4655 /* TODO: AddPath support. */
4656 evpn_mpattr_encode_type5(s
, p
, prd
, label
, num_labels
, attr
);
4664 int bgp_nlri_parse_evpn(struct peer
*peer
, struct attr
*attr
,
4665 struct bgp_nlri
*packet
, int withdraw
)
4671 uint32_t addpath_id
;
4672 int addpath_encoded
;
4677 /* Start processing the NLRI - there may be multiple in the MP_REACH */
4679 lim
= pnt
+ packet
->length
;
4681 safi
= packet
->safi
;
4685 (CHECK_FLAG(peer
->af_cap
[afi
][safi
], PEER_CAP_ADDPATH_AF_RX_ADV
)
4686 && CHECK_FLAG(peer
->af_cap
[afi
][safi
],
4687 PEER_CAP_ADDPATH_AF_TX_RCV
));
4689 for (; pnt
< lim
; pnt
+= psize
) {
4690 /* Clear prefix structure. */
4691 memset(&p
, 0, sizeof(struct prefix
));
4693 /* Deal with path-id if AddPath is supported. */
4694 if (addpath_encoded
) {
4695 /* When packet overflow occurs return immediately. */
4696 if (pnt
+ BGP_ADDPATH_ID_LEN
> lim
)
4699 addpath_id
= ntohl(*((uint32_t *)pnt
));
4700 pnt
+= BGP_ADDPATH_ID_LEN
;
4703 /* All EVPN NLRI types start with type and length. */
4710 /* When packet overflow occur return immediately. */
4711 if (pnt
+ psize
> lim
)
4715 case BGP_EVPN_MAC_IP_ROUTE
:
4716 if (process_type2_route(peer
, afi
, safi
,
4717 withdraw
? NULL
: attr
, pnt
,
4718 psize
, addpath_id
)) {
4721 "%u:%s - Error in processing EVPN type-2 NLRI size %d",
4722 peer
->bgp
->vrf_id
, peer
->host
, psize
);
4727 case BGP_EVPN_IMET_ROUTE
:
4728 if (process_type3_route(peer
, afi
, safi
,
4729 withdraw
? NULL
: attr
, pnt
,
4730 psize
, addpath_id
)) {
4733 "%u:%s - Error in processing EVPN type-3 NLRI size %d",
4734 peer
->bgp
->vrf_id
, peer
->host
, psize
);
4739 case BGP_EVPN_ES_ROUTE
:
4740 if (process_type4_route(peer
, afi
, safi
,
4741 withdraw
? NULL
: attr
, pnt
,
4742 psize
, addpath_id
)) {
4745 "%u:%s - Error in processing EVPN type-4 NLRI size %d",
4746 peer
->bgp
->vrf_id
, peer
->host
, psize
);
4751 case BGP_EVPN_IP_PREFIX_ROUTE
:
4752 if (process_type5_route(peer
, afi
, safi
, attr
, pnt
,
4753 psize
, addpath_id
, withdraw
)) {
4756 "%u:%s - Error in processing EVPN type-5 NLRI size %d",
4757 peer
->bgp
->vrf_id
, peer
->host
, psize
);
4767 /* Packet length consistency check. */
4775 * Map the RTs (configured or automatically derived) of a VRF to the VRF.
4776 * The mapping will be used during route processing.
4777 * bgp_def: default bgp instance
4778 * bgp_vrf: specific bgp vrf instance on which RT is configured
4780 void bgp_evpn_map_vrf_to_its_rts(struct bgp
*bgp_vrf
)
4783 struct ecommunity_val
*eval
= NULL
;
4784 struct listnode
*node
= NULL
, *nnode
= NULL
;
4785 struct ecommunity
*ecom
= NULL
;
4787 for (ALL_LIST_ELEMENTS(bgp_vrf
->vrf_import_rtl
, node
, nnode
, ecom
)) {
4788 for (i
= 0; i
< ecom
->size
; i
++) {
4789 eval
= (struct ecommunity_val
*)(ecom
->val
4791 * ECOMMUNITY_SIZE
));
4792 map_vrf_to_rt(bgp_vrf
, eval
);
4798 * Unmap the RTs (configured or automatically derived) of a VRF from the VRF.
4800 void bgp_evpn_unmap_vrf_from_its_rts(struct bgp
*bgp_vrf
)
4803 struct ecommunity_val
*eval
;
4804 struct listnode
*node
, *nnode
;
4805 struct ecommunity
*ecom
;
4807 for (ALL_LIST_ELEMENTS(bgp_vrf
->vrf_import_rtl
, node
, nnode
, ecom
)) {
4808 for (i
= 0; i
< ecom
->size
; i
++) {
4809 struct vrf_irt_node
*irt
;
4810 struct ecommunity_val eval_tmp
;
4812 eval
= (struct ecommunity_val
*)(ecom
->val
4814 * ECOMMUNITY_SIZE
));
4815 /* If using "automatic" RT, we only care about the
4816 * local-admin sub-field.
4817 * This is to facilitate using VNI as the RT for EBGP
4820 memcpy(&eval_tmp
, eval
, ECOMMUNITY_SIZE
);
4821 if (!CHECK_FLAG(bgp_vrf
->vrf_flags
,
4822 BGP_VRF_IMPORT_RT_CFGD
))
4823 mask_ecom_global_admin(&eval_tmp
, eval
);
4825 irt
= lookup_vrf_import_rt(&eval_tmp
);
4827 unmap_vrf_from_rt(bgp_vrf
, irt
);
4834 * Map the RTs (configured or automatically derived) of a VNI to the VNI.
4835 * The mapping will be used during route processing.
4837 void bgp_evpn_map_vni_to_its_rts(struct bgp
*bgp
, struct bgpevpn
*vpn
)
4840 struct ecommunity_val
*eval
;
4841 struct listnode
*node
, *nnode
;
4842 struct ecommunity
*ecom
;
4844 for (ALL_LIST_ELEMENTS(vpn
->import_rtl
, node
, nnode
, ecom
)) {
4845 for (i
= 0; i
< ecom
->size
; i
++) {
4846 eval
= (struct ecommunity_val
*)(ecom
->val
4848 * ECOMMUNITY_SIZE
));
4849 map_vni_to_rt(bgp
, vpn
, eval
);
4855 * Unmap the RTs (configured or automatically derived) of a VNI from the VNI.
4857 void bgp_evpn_unmap_vni_from_its_rts(struct bgp
*bgp
, struct bgpevpn
*vpn
)
4860 struct ecommunity_val
*eval
;
4861 struct listnode
*node
, *nnode
;
4862 struct ecommunity
*ecom
;
4864 for (ALL_LIST_ELEMENTS(vpn
->import_rtl
, node
, nnode
, ecom
)) {
4865 for (i
= 0; i
< ecom
->size
; i
++) {
4866 struct irt_node
*irt
;
4867 struct ecommunity_val eval_tmp
;
4869 eval
= (struct ecommunity_val
*)(ecom
->val
4871 * ECOMMUNITY_SIZE
));
4872 /* If using "automatic" RT, we only care about the
4873 * local-admin sub-field.
4874 * This is to facilitate using VNI as the RT for EBGP
4877 memcpy(&eval_tmp
, eval
, ECOMMUNITY_SIZE
);
4878 if (!is_import_rt_configured(vpn
))
4879 mask_ecom_global_admin(&eval_tmp
, eval
);
4881 irt
= lookup_import_rt(bgp
, &eval_tmp
);
4883 unmap_vni_from_rt(bgp
, vpn
, irt
);
4889 * Derive Import RT automatically for VNI and map VNI to RT.
4890 * The mapping will be used during route processing.
4892 void bgp_evpn_derive_auto_rt_import(struct bgp
*bgp
, struct bgpevpn
*vpn
)
4894 form_auto_rt(bgp
, vpn
->vni
, vpn
->import_rtl
);
4895 UNSET_FLAG(vpn
->flags
, VNI_FLAG_IMPRT_CFGD
);
4898 bgp_evpn_map_vni_to_its_rts(bgp
, vpn
);
4902 * Derive Export RT automatically for VNI.
4904 void bgp_evpn_derive_auto_rt_export(struct bgp
*bgp
, struct bgpevpn
*vpn
)
4906 form_auto_rt(bgp
, vpn
->vni
, vpn
->export_rtl
);
4907 UNSET_FLAG(vpn
->flags
, VNI_FLAG_EXPRT_CFGD
);
4911 * Derive RD automatically for VNI using passed information - it
4912 * is of the form RouterId:unique-id-for-vni.
4914 void bgp_evpn_derive_auto_rd_for_vrf(struct bgp
*bgp
)
4916 form_auto_rd(bgp
->router_id
, bgp
->vrf_rd_id
, &bgp
->vrf_prd
);
4920 * Derive RD automatically for VNI using passed information - it
4921 * is of the form RouterId:unique-id-for-vni.
4923 void bgp_evpn_derive_auto_rd(struct bgp
*bgp
, struct bgpevpn
*vpn
)
4927 vpn
->prd
.family
= AF_UNSPEC
;
4928 vpn
->prd
.prefixlen
= 64;
4929 sprintf(buf
, "%s:%hu", inet_ntoa(bgp
->router_id
), vpn
->rd_id
);
4930 (void)str2prefix_rd(buf
, &vpn
->prd
);
4931 UNSET_FLAG(vpn
->flags
, VNI_FLAG_RD_CFGD
);
4937 bool bgp_evpn_lookup_l3vni_l2vni_table(vni_t vni
)
4939 struct list
*inst
= bm
->bgp
;
4940 struct listnode
*node
;
4941 struct bgp
*bgp_vrf
;
4943 for (ALL_LIST_ELEMENTS_RO(inst
, node
, bgp_vrf
)) {
4944 if (bgp_vrf
->l3vni
== vni
)
4954 struct bgpevpn
*bgp_evpn_lookup_vni(struct bgp
*bgp
, vni_t vni
)
4956 struct bgpevpn
*vpn
;
4959 memset(&tmp
, 0, sizeof(struct bgpevpn
));
4961 vpn
= hash_lookup(bgp
->vnihash
, &tmp
);
4966 * Create a new vpn - invoked upon configuration or zebra notification.
4968 struct bgpevpn
*bgp_evpn_new(struct bgp
*bgp
, vni_t vni
,
4969 struct in_addr originator_ip
,
4970 vrf_id_t tenant_vrf_id
)
4972 struct bgpevpn
*vpn
;
4977 vpn
= XCALLOC(MTYPE_BGP_EVPN
, sizeof(struct bgpevpn
));
4981 /* Set values - RD and RT set to defaults. */
4983 vpn
->originator_ip
= originator_ip
;
4984 vpn
->tenant_vrf_id
= tenant_vrf_id
;
4986 /* Initialize route-target import and export lists */
4987 vpn
->import_rtl
= list_new();
4988 vpn
->import_rtl
->cmp
= (int (*)(void *, void *))evpn_route_target_cmp
;
4989 vpn
->import_rtl
->del
= evpn_xxport_delete_ecomm
;
4990 vpn
->export_rtl
= list_new();
4991 vpn
->export_rtl
->cmp
= (int (*)(void *, void *))evpn_route_target_cmp
;
4992 vpn
->export_rtl
->del
= evpn_xxport_delete_ecomm
;
4993 bf_assign_index(bm
->rd_idspace
, vpn
->rd_id
);
4994 derive_rd_rt_for_vni(bgp
, vpn
);
4996 /* Initialize EVPN route table. */
4997 vpn
->route_table
= bgp_table_init(bgp
, AFI_L2VPN
, SAFI_EVPN
);
5000 if (!hash_get(bgp
->vnihash
, vpn
, hash_alloc_intern
)) {
5001 XFREE(MTYPE_BGP_EVPN
, vpn
);
5005 /* add to l2vni list on corresponding vrf */
5006 bgpevpn_link_to_l3vni(vpn
);
5008 QOBJ_REG(vpn
, bgpevpn
);
5013 * Free a given VPN - called in multiple scenarios such as zebra
5014 * notification, configuration being deleted, advertise-all-vni disabled etc.
5015 * This just frees appropriate memory, caller should have taken other
5018 void bgp_evpn_free(struct bgp
*bgp
, struct bgpevpn
*vpn
)
5020 bgpevpn_unlink_from_l3vni(vpn
);
5021 bgp_table_unlock(vpn
->route_table
);
5022 bgp_evpn_unmap_vni_from_its_rts(bgp
, vpn
);
5023 list_delete(&vpn
->import_rtl
);
5024 list_delete(&vpn
->export_rtl
);
5025 bf_release_index(bm
->rd_idspace
, vpn
->rd_id
);
5026 hash_release(bgp
->vnihash
, vpn
);
5028 XFREE(MTYPE_BGP_EVPN
, vpn
);
5034 struct evpnes
*bgp_evpn_lookup_es(struct bgp
*bgp
, esi_t
*esi
)
5039 memset(&tmp
, 0, sizeof(struct evpnes
));
5040 memcpy(&tmp
.esi
, esi
, sizeof(esi_t
));
5041 es
= hash_lookup(bgp
->esihash
, &tmp
);
5046 * Create a new local es - invoked upon zebra notification.
5048 struct evpnes
*bgp_evpn_es_new(struct bgp
*bgp
,
5050 struct ipaddr
*originator_ip
)
5058 es
= XCALLOC(MTYPE_BGP_EVPN_ES
, sizeof(struct evpnes
));
5062 /* set the ESI and originator_ip */
5063 memcpy(&es
->esi
, esi
, sizeof(esi_t
));
5064 memcpy(&es
->originator_ip
, originator_ip
, sizeof(struct ipaddr
));
5066 /* Initialise the VTEP list */
5067 es
->vtep_list
= list_new();
5068 es
->vtep_list
->cmp
= (int (*)(void *, void *))evpn_vtep_ip_cmp
;
5070 /* auto derive RD for this es */
5071 bf_assign_index(bm
->rd_idspace
, es
->rd_id
);
5072 es
->prd
.family
= AF_UNSPEC
;
5073 es
->prd
.prefixlen
= 64;
5074 sprintf(buf
, "%s:%hu", inet_ntoa(bgp
->router_id
), es
->rd_id
);
5075 (void)str2prefix_rd(buf
, &es
->prd
);
5077 /* Initialize the ES route table */
5078 es
->route_table
= bgp_table_init(bgp
, AFI_L2VPN
, SAFI_EVPN
);
5081 if (!hash_get(bgp
->esihash
, es
, hash_alloc_intern
)) {
5082 XFREE(MTYPE_BGP_EVPN_ES
, es
);
5086 QOBJ_REG(es
, evpnes
);
5092 * This just frees appropriate memory, caller should have taken other
5095 void bgp_evpn_es_free(struct bgp
*bgp
, struct evpnes
*es
)
5097 list_delete(&es
->vtep_list
);
5098 bgp_table_unlock(es
->route_table
);
5099 bf_release_index(bm
->rd_idspace
, es
->rd_id
);
5100 hash_release(bgp
->esihash
, es
);
5102 XFREE(MTYPE_BGP_EVPN_ES
, es
);
5106 * Import evpn route from global table to VNI/VRF/ESI.
5108 int bgp_evpn_import_route(struct bgp
*bgp
, afi_t afi
, safi_t safi
,
5109 struct prefix
*p
, struct bgp_path_info
*pi
)
5111 return install_uninstall_evpn_route(bgp
, afi
, safi
, p
, pi
, 1);
5115 * Unimport evpn route from VNI/VRF/ESI.
5117 int bgp_evpn_unimport_route(struct bgp
*bgp
, afi_t afi
, safi_t safi
,
5118 struct prefix
*p
, struct bgp_path_info
*pi
)
5120 return install_uninstall_evpn_route(bgp
, afi
, safi
, p
, pi
, 0);
5123 /* filter routes which have martian next hops */
5124 int bgp_filter_evpn_routes_upon_martian_nh_change(struct bgp
*bgp
)
5128 struct bgp_node
*rd_rn
, *rn
;
5129 struct bgp_table
*table
;
5130 struct bgp_path_info
*pi
;
5135 /* Walk entire global routing table and evaluate routes which could be
5136 * imported into this VPN. Note that we cannot just look at the routes
5137 * for the VNI's RD -
5138 * remote routes applicable for this VNI could have any RD.
5140 /* EVPN routes are a 2-level table. */
5141 for (rd_rn
= bgp_table_top(bgp
->rib
[afi
][safi
]); rd_rn
;
5142 rd_rn
= bgp_route_next(rd_rn
)) {
5143 table
= (struct bgp_table
*)(rd_rn
->info
);
5147 for (rn
= bgp_table_top(table
); rn
; rn
= bgp_route_next(rn
)) {
5149 for (pi
= rn
->info
; pi
; pi
= pi
->next
) {
5151 /* Consider "valid" remote routes applicable for
5153 if (!(pi
->type
== ZEBRA_ROUTE_BGP
5154 && pi
->sub_type
== BGP_ROUTE_NORMAL
))
5157 if (bgp_nexthop_self(bgp
, pi
->attr
->nexthop
)) {
5159 char attr_str
[BUFSIZ
];
5160 char pbuf
[PREFIX_STRLEN
];
5162 bgp_dump_attr(pi
->attr
, attr_str
,
5165 if (bgp_debug_update(pi
->peer
, &rn
->p
,
5168 "%u: prefix %s with attr %s - DENIED due to martian or self nexthop",
5175 bgp_evpn_unimport_route(bgp
, afi
, safi
,
5178 bgp_rib_remove(rn
, pi
, pi
->peer
, afi
,
5189 * Handle del of a local MACIP.
5191 int bgp_evpn_local_macip_del(struct bgp
*bgp
, vni_t vni
, struct ethaddr
*mac
,
5194 struct bgpevpn
*vpn
;
5195 struct prefix_evpn p
;
5197 /* Lookup VNI hash - should exist. */
5198 vpn
= bgp_evpn_lookup_vni(bgp
, vni
);
5199 if (!vpn
|| !is_vni_live(vpn
)) {
5200 flog_warn(EC_BGP_EVPN_VPN_VNI
,
5201 "%u: VNI hash entry for VNI %u %s at MACIP DEL",
5202 bgp
->vrf_id
, vni
, vpn
? "not live" : "not found");
5206 /* Remove EVPN type-2 route and schedule for processing. */
5207 build_evpn_type2_prefix(&p
, mac
, ip
);
5208 delete_evpn_route(bgp
, vpn
, &p
);
5214 * Handle add of a local MACIP.
5216 int bgp_evpn_local_macip_add(struct bgp
*bgp
, vni_t vni
, struct ethaddr
*mac
,
5217 struct ipaddr
*ip
, uint8_t flags
, uint32_t seq
)
5219 struct bgpevpn
*vpn
;
5220 struct prefix_evpn p
;
5222 /* Lookup VNI hash - should exist. */
5223 vpn
= bgp_evpn_lookup_vni(bgp
, vni
);
5224 if (!vpn
|| !is_vni_live(vpn
)) {
5225 flog_warn(EC_BGP_EVPN_VPN_VNI
,
5226 "%u: VNI hash entry for VNI %u %s at MACIP ADD",
5227 bgp
->vrf_id
, vni
, vpn
? "not live" : "not found");
5231 /* Create EVPN type-2 route and schedule for processing. */
5232 build_evpn_type2_prefix(&p
, mac
, ip
);
5233 if (update_evpn_route(bgp
, vpn
, &p
, flags
, seq
)) {
5234 char buf
[ETHER_ADDR_STRLEN
];
5235 char buf2
[INET6_ADDRSTRLEN
];
5238 EC_BGP_EVPN_ROUTE_CREATE
,
5239 "%u:Failed to create Type-2 route, VNI %u %s MAC %s IP %s (flags: 0x%x)",
5240 bgp
->vrf_id
, vpn
->vni
,
5241 CHECK_FLAG(flags
, ZEBRA_MACIP_TYPE_STICKY
)
5244 prefix_mac2str(mac
, buf
, sizeof(buf
)),
5245 ipaddr2str(ip
, buf2
, sizeof(buf2
)), flags
);
5252 static void link_l2vni_hash_to_l3vni(struct hash_backet
*backet
,
5253 struct bgp
*bgp_vrf
)
5255 struct bgpevpn
*vpn
= (struct bgpevpn
*)backet
->data
;
5256 struct bgp
*bgp_def
= NULL
;
5258 bgp_def
= bgp_get_default();
5261 if (vpn
->tenant_vrf_id
== bgp_vrf
->vrf_id
)
5262 bgpevpn_link_to_l3vni(vpn
);
5265 int bgp_evpn_local_l3vni_add(vni_t l3vni
, vrf_id_t vrf_id
, struct ethaddr
*rmac
,
5266 struct in_addr originator_ip
, int filter
)
5268 struct bgp
*bgp_vrf
= NULL
; /* bgp VRF instance */
5269 struct bgp
*bgp_def
= NULL
; /* default bgp instance */
5270 struct listnode
*node
= NULL
;
5271 struct bgpevpn
*vpn
= NULL
;
5274 /* get the default instance - required to get the AS number for VRF
5277 bgp_def
= bgp_get_default();
5281 "Cannot process L3VNI %u ADD - default BGP instance not yet created",
5287 /* if the BGP vrf instance doesnt exist - create one */
5288 bgp_vrf
= bgp_lookup_by_name(vrf_id_to_name(vrf_id
));
5293 ret
= bgp_get(&bgp_vrf
, &as
, vrf_id_to_name(vrf_id
),
5294 BGP_INSTANCE_TYPE_VRF
);
5296 case BGP_ERR_MULTIPLE_INSTANCE_NOT_SET
:
5297 flog_err(EC_BGP_MULTI_INSTANCE
,
5298 "'bgp multiple-instance' not present\n");
5300 case BGP_ERR_AS_MISMATCH
:
5301 flog_err(EC_BGP_EVPN_AS_MISMATCH
,
5302 "BGP is already running; AS is %u\n", as
);
5304 case BGP_ERR_INSTANCE_MISMATCH
:
5305 flog_err(EC_BGP_EVPN_INSTANCE_MISMATCH
,
5306 "BGP instance name and AS number mismatch\n");
5310 /* mark as auto created */
5311 SET_FLAG(bgp_vrf
->vrf_flags
, BGP_VRF_AUTO
);
5314 /* associate with l3vni */
5315 bgp_vrf
->l3vni
= l3vni
;
5317 /* set the router mac - to be used in mac-ip routes for this vrf */
5318 memcpy(&bgp_vrf
->rmac
, rmac
, sizeof(struct ethaddr
));
5320 /* set the originator ip */
5321 bgp_vrf
->originator_ip
= originator_ip
;
5323 /* set the right filter - are we using l3vni only for prefix routes? */
5325 SET_FLAG(bgp_vrf
->vrf_flags
, BGP_VRF_L3VNI_PREFIX_ROUTES_ONLY
);
5327 /* auto derive RD/RT */
5328 if (!CHECK_FLAG(bgp_vrf
->vrf_flags
, BGP_VRF_IMPORT_RT_CFGD
))
5329 evpn_auto_rt_import_add_for_vrf(bgp_vrf
);
5330 if (!CHECK_FLAG(bgp_vrf
->vrf_flags
, BGP_VRF_EXPORT_RT_CFGD
))
5331 evpn_auto_rt_export_add_for_vrf(bgp_vrf
);
5332 bgp_evpn_derive_auto_rd_for_vrf(bgp_vrf
);
5334 /* link all corresponding l2vnis */
5335 hash_iterate(bgp_def
->vnihash
,
5336 (void (*)(struct hash_backet
*,
5337 void *))link_l2vni_hash_to_l3vni
,
5340 /* Only update all corresponding type-2 routes if we are advertising two
5341 * labels along with type-2 routes
5344 for (ALL_LIST_ELEMENTS_RO(bgp_vrf
->l2vnis
, node
, vpn
))
5345 update_routes_for_vni(bgp_def
, vpn
);
5347 /* advertise type-5 routes if needed */
5348 update_advertise_vrf_routes(bgp_vrf
);
5350 /* install all remote routes belonging to this l3vni into correspondng
5352 install_routes_for_vrf(bgp_vrf
);
5357 int bgp_evpn_local_l3vni_del(vni_t l3vni
, vrf_id_t vrf_id
)
5359 struct bgp
*bgp_vrf
= NULL
; /* bgp vrf instance */
5360 struct bgp
*bgp_def
= NULL
; /* default bgp instance */
5361 struct listnode
*node
= NULL
;
5362 struct listnode
*next
= NULL
;
5363 struct bgpevpn
*vpn
= NULL
;
5365 bgp_vrf
= bgp_lookup_by_vrf_id(vrf_id
);
5369 "Cannot process L3VNI %u Del - Could not find BGP instance",
5374 bgp_def
= bgp_get_default();
5378 "Cannot process L3VNI %u Del - Could not find default BGP instance",
5383 /* Remove remote routes from BGT VRF even if BGP_VRF_AUTO is configured,
5384 * bgp_delete would not remove/decrement bgp_path_info of the ip_prefix
5385 * routes. This will uninstalling the routes from zebra and decremnt the
5388 uninstall_routes_for_vrf(bgp_vrf
);
5390 /* delete/withdraw all type-5 routes */
5391 delete_withdraw_vrf_routes(bgp_vrf
);
5393 /* remove the l3vni from vrf instance */
5396 /* remove the Rmac from the BGP vrf */
5397 memset(&bgp_vrf
->rmac
, 0, sizeof(struct ethaddr
));
5400 if (!list_isempty(bgp_vrf
->vrf_import_rtl
)) {
5401 bgp_evpn_unmap_vrf_from_its_rts(bgp_vrf
);
5402 list_delete_all_node(bgp_vrf
->vrf_import_rtl
);
5404 if (!list_isempty(bgp_vrf
->vrf_export_rtl
)) {
5405 list_delete_all_node(bgp_vrf
->vrf_export_rtl
);
5408 /* update all corresponding local mac-ip routes */
5409 if (!CHECK_FLAG(bgp_vrf
->vrf_flags
, BGP_VRF_L3VNI_PREFIX_ROUTES_ONLY
)) {
5410 for (ALL_LIST_ELEMENTS_RO(bgp_vrf
->l2vnis
, node
, vpn
)) {
5411 UNSET_FLAG(vpn
->flags
, VNI_FLAG_USE_TWO_LABELS
);
5412 update_routes_for_vni(bgp_def
, vpn
);
5416 /* If any L2VNIs point to this instance, unlink them. */
5417 for (ALL_LIST_ELEMENTS(bgp_vrf
->l2vnis
, node
, next
, vpn
))
5418 bgpevpn_unlink_from_l3vni(vpn
);
5420 /* Delete the instance if it was autocreated */
5421 if (CHECK_FLAG(bgp_vrf
->vrf_flags
, BGP_VRF_AUTO
))
5422 bgp_delete(bgp_vrf
);
5428 * Handle del of a local VNI.
5430 int bgp_evpn_local_vni_del(struct bgp
*bgp
, vni_t vni
)
5432 struct bgpevpn
*vpn
;
5434 /* Locate VNI hash */
5435 vpn
= bgp_evpn_lookup_vni(bgp
, vni
);
5437 if (bgp_debug_zebra(NULL
))
5439 EC_BGP_EVPN_VPN_VNI
,
5440 "%u: VNI hash entry for VNI %u not found at DEL",
5445 /* Remove all local EVPN routes and schedule for processing (to
5446 * withdraw from peers).
5448 delete_routes_for_vni(bgp
, vpn
);
5451 * tunnel is no longer active, del tunnel ip address from tip_hash
5453 bgp_tip_del(bgp
, &vpn
->originator_ip
);
5455 /* Clear "live" flag and see if hash needs to be freed. */
5456 UNSET_FLAG(vpn
->flags
, VNI_FLAG_LIVE
);
5457 if (!is_vni_configured(vpn
))
5458 bgp_evpn_free(bgp
, vpn
);
5464 * Handle add (or update) of a local VNI. The VNI changes we care
5465 * about are for the local-tunnel-ip and the (tenant) VRF.
5467 int bgp_evpn_local_vni_add(struct bgp
*bgp
, vni_t vni
,
5468 struct in_addr originator_ip
, vrf_id_t tenant_vrf_id
)
5470 struct bgpevpn
*vpn
;
5471 struct prefix_evpn p
;
5473 /* Lookup VNI. If present and no change, exit. */
5474 vpn
= bgp_evpn_lookup_vni(bgp
, vni
);
5477 if (is_vni_live(vpn
)
5478 && IPV4_ADDR_SAME(&vpn
->originator_ip
, &originator_ip
)
5479 && vpn
->tenant_vrf_id
== tenant_vrf_id
)
5480 /* Probably some other param has changed that we don't
5484 /* Update tenant_vrf_id if it has changed. */
5485 if (vpn
->tenant_vrf_id
!= tenant_vrf_id
) {
5486 bgpevpn_unlink_from_l3vni(vpn
);
5487 vpn
->tenant_vrf_id
= tenant_vrf_id
;
5488 bgpevpn_link_to_l3vni(vpn
);
5491 /* If tunnel endpoint IP has changed, update (and delete prior
5492 * type-3 route, if needed.)
5494 if (!IPV4_ADDR_SAME(&vpn
->originator_ip
, &originator_ip
))
5495 handle_tunnel_ip_change(bgp
, vpn
, originator_ip
);
5497 /* Update all routes with new endpoint IP and/or export RT
5500 if (is_vni_live(vpn
))
5501 update_routes_for_vni(bgp
, vpn
);
5504 /* Create or update as appropriate. */
5506 vpn
= bgp_evpn_new(bgp
, vni
, originator_ip
, tenant_vrf_id
);
5510 "%u: Failed to allocate VNI entry for VNI %u - at Add",
5516 /* if the VNI is live already, there is nothing more to do */
5517 if (is_vni_live(vpn
))
5520 /* Mark as "live" */
5521 SET_FLAG(vpn
->flags
, VNI_FLAG_LIVE
);
5523 /* tunnel is now active, add tunnel-ip to db */
5524 bgp_tip_add(bgp
, &originator_ip
);
5526 /* filter routes as nexthop database has changed */
5527 bgp_filter_evpn_routes_upon_martian_nh_change(bgp
);
5530 * Create EVPN type-3 route and schedule for processing.
5532 * RT-3 only if doing head-end replication
5534 if (bgp
->vxlan_flood_ctrl
== VXLAN_FLOOD_HEAD_END_REPL
) {
5535 build_evpn_type3_prefix(&p
, vpn
->originator_ip
);
5536 if (update_evpn_route(bgp
, vpn
, &p
, 0, 0)) {
5537 flog_err(EC_BGP_EVPN_ROUTE_CREATE
,
5538 "%u: Type3 route creation failure for VNI %u",
5544 /* If we have learnt and retained remote routes (VTEPs, MACs) for this
5548 install_routes_for_vni(bgp
, vpn
);
5550 /* If we are advertising gateway mac-ip
5551 It needs to be conveyed again to zebra */
5552 bgp_zebra_advertise_gw_macip(bgp
, vpn
->advertise_gw_macip
, vpn
->vni
);
5558 * bgp_evpn_local_es_del
5560 int bgp_evpn_local_es_del(struct bgp
*bgp
,
5562 struct ipaddr
*originator_ip
)
5564 char buf
[ESI_STR_LEN
];
5565 struct evpnes
*es
= NULL
;
5567 if (!bgp
->esihash
) {
5568 flog_err(EC_BGP_ES_CREATE
, "%u: ESI hash not yet created",
5573 /* Lookup ESI hash - should exist. */
5574 es
= bgp_evpn_lookup_es(bgp
, esi
);
5576 flog_warn(EC_BGP_EVPN_ESI
,
5577 "%u: ESI hash entry for ESI %s at Local ES DEL",
5578 bgp
->vrf_id
, esi_to_str(esi
, buf
, sizeof(buf
)));
5582 /* Delete all local EVPN ES routes from ESI table
5583 * and schedule for processing (to withdraw from peers))
5585 delete_routes_for_es(bgp
, es
);
5587 /* free the hash entry */
5588 bgp_evpn_es_free(bgp
, es
);
5594 * bgp_evpn_local_es_add
5596 int bgp_evpn_local_es_add(struct bgp
*bgp
,
5598 struct ipaddr
*originator_ip
)
5600 char buf
[ESI_STR_LEN
];
5601 struct evpnes
*es
= NULL
;
5602 struct prefix_evpn p
;
5604 if (!bgp
->esihash
) {
5605 flog_err(EC_BGP_ES_CREATE
, "%u: ESI hash not yet created",
5610 /* create the new es */
5611 es
= bgp_evpn_lookup_es(bgp
, esi
);
5613 es
= bgp_evpn_es_new(bgp
, esi
, originator_ip
);
5617 "%u: Failed to allocate ES entry for ESI %s - at Local ES Add",
5618 bgp
->vrf_id
, esi_to_str(esi
, buf
, sizeof(buf
)));
5622 UNSET_FLAG(es
->flags
, EVPNES_REMOTE
);
5623 SET_FLAG(es
->flags
, EVPNES_LOCAL
);
5625 build_evpn_type4_prefix(&p
, esi
, originator_ip
->ipaddr_v4
);
5626 if (update_evpn_type4_route(bgp
, es
, &p
)) {
5627 flog_err(EC_BGP_EVPN_ROUTE_CREATE
,
5628 "%u: Type4 route creation failure for ESI %s",
5629 bgp
->vrf_id
, esi_to_str(esi
, buf
, sizeof(buf
)));
5633 /* import all remote ES routes in th ES table */
5634 install_routes_for_es(bgp
, es
);
5640 * Handle change in setting for BUM handling. The supported values
5641 * are head-end replication and dropping all BUM packets. Any change
5642 * should be registered with zebra. Also, if doing head-end replication,
5643 * need to advertise local VNIs as EVPN RT-3 wheras, if BUM packets are
5644 * to be dropped, the RT-3s must be withdrawn.
5646 void bgp_evpn_flood_control_change(struct bgp
*bgp
)
5648 zlog_info("L2VPN EVPN BUM handling is %s",
5649 bgp
->vxlan_flood_ctrl
== VXLAN_FLOOD_HEAD_END_REPL
?
5650 "Flooding" : "Flooding Disabled");
5652 bgp_zebra_vxlan_flood_control(bgp
, bgp
->vxlan_flood_ctrl
);
5653 if (bgp
->vxlan_flood_ctrl
== VXLAN_FLOOD_HEAD_END_REPL
)
5654 hash_iterate(bgp
->vnihash
, create_advertise_type3
, bgp
);
5655 else if (bgp
->vxlan_flood_ctrl
== VXLAN_FLOOD_DISABLED
)
5656 hash_iterate(bgp
->vnihash
, delete_withdraw_type3
, bgp
);
5660 * Cleanup EVPN information on disable - Need to delete and withdraw
5661 * EVPN routes from peers.
5663 void bgp_evpn_cleanup_on_disable(struct bgp
*bgp
)
5665 hash_iterate(bgp
->vnihash
, (void (*)(struct hash_backet
*,
5666 void *))cleanup_vni_on_disable
,
5671 * Cleanup EVPN information - invoked at the time of bgpd exit or when the
5672 * BGP instance (default) is being freed.
5674 void bgp_evpn_cleanup(struct bgp
*bgp
)
5676 hash_iterate(bgp
->vnihash
,
5677 (void (*)(struct hash_backet
*, void *))free_vni_entry
,
5680 hash_free(bgp
->import_rt_hash
);
5681 bgp
->import_rt_hash
= NULL
;
5683 hash_free(bgp
->vrf_import_rt_hash
);
5684 bgp
->vrf_import_rt_hash
= NULL
;
5686 hash_free(bgp
->vnihash
);
5687 bgp
->vnihash
= NULL
;
5689 hash_free(bgp
->esihash
);
5690 bgp
->esihash
= NULL
;
5692 list_delete(&bgp
->vrf_import_rtl
);
5693 list_delete(&bgp
->vrf_export_rtl
);
5694 list_delete(&bgp
->l2vnis
);
5698 * Initialization for EVPN
5701 * hash for RT to VNI
5703 void bgp_evpn_init(struct bgp
*bgp
)
5706 hash_create(vni_hash_key_make
, vni_hash_cmp
, "BGP VNI Hash");
5708 hash_create(esi_hash_keymake
, esi_cmp
,
5709 "BGP EVPN Local ESI Hash");
5710 bgp
->import_rt_hash
=
5711 hash_create(import_rt_hash_key_make
, import_rt_hash_cmp
,
5712 "BGP Import RT Hash");
5713 bgp
->vrf_import_rt_hash
=
5714 hash_create(vrf_import_rt_hash_key_make
, vrf_import_rt_hash_cmp
,
5715 "BGP VRF Import RT Hash");
5716 bgp
->vrf_import_rtl
= list_new();
5717 bgp
->vrf_import_rtl
->cmp
=
5718 (int (*)(void *, void *))evpn_route_target_cmp
;
5719 bgp
->vrf_import_rtl
->del
= evpn_xxport_delete_ecomm
;
5720 bgp
->vrf_export_rtl
= list_new();
5721 bgp
->vrf_export_rtl
->cmp
=
5722 (int (*)(void *, void *))evpn_route_target_cmp
;
5723 bgp
->vrf_export_rtl
->del
= evpn_xxport_delete_ecomm
;
5724 bgp
->l2vnis
= list_new();
5725 bgp
->l2vnis
->cmp
= (int (*)(void *, void *))vni_hash_cmp
;
5727 /* Default BUM handling is to do head-end replication. */
5728 bgp
->vxlan_flood_ctrl
= VXLAN_FLOOD_HEAD_END_REPL
;
5731 void bgp_evpn_vrf_delete(struct bgp
*bgp_vrf
)
5733 bgp_evpn_unmap_vrf_from_its_rts(bgp_vrf
);