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"
50 #include "bgpd/bgp_addpath.h"
51 #include "bgpd/bgp_mac.h"
52 #include "bgpd/bgp_vty.h"
55 * Definitions and external declarations.
57 extern struct zclient
*zclient
;
59 DEFINE_QOBJ_TYPE(bgpevpn
)
60 DEFINE_QOBJ_TYPE(evpnes
)
64 * Static function declarations
66 static void delete_evpn_route_entry(struct bgp
*bgp
, afi_t afi
, safi_t safi
,
68 struct bgp_path_info
**pi
);
69 static int delete_all_vni_routes(struct bgp
*bgp
, struct bgpevpn
*vpn
);
75 /* compare two IPV4 VTEP IPs */
76 static int evpn_vtep_ip_cmp(void *p1
, void *p2
)
78 const struct in_addr
*ip1
= p1
;
79 const struct in_addr
*ip2
= p2
;
81 return ip1
->s_addr
- ip2
->s_addr
;
85 * Make hash key for ESI.
87 static unsigned int esi_hash_keymake(const void *p
)
89 const struct evpnes
*pes
= p
;
90 const void *pnt
= (void *)pes
->esi
.val
;
92 return jhash(pnt
, ESI_BYTES
, 0xa5a5a55a);
98 static bool esi_cmp(const void *p1
, const void *p2
)
100 const struct evpnes
*pes1
= p1
;
101 const struct evpnes
*pes2
= p2
;
103 if (pes1
== NULL
&& pes2
== NULL
)
106 if (pes1
== NULL
|| pes2
== NULL
)
109 return (memcmp(pes1
->esi
.val
, pes2
->esi
.val
, ESI_BYTES
) == 0);
115 static unsigned int vni_hash_key_make(const void *p
)
117 const struct bgpevpn
*vpn
= p
;
118 return (jhash_1word(vpn
->vni
, 0));
122 * Comparison function for vni hash
124 static bool vni_hash_cmp(const void *p1
, const void *p2
)
126 const struct bgpevpn
*vpn1
= p1
;
127 const struct bgpevpn
*vpn2
= p2
;
133 return (vpn1
->vni
== vpn2
->vni
);
136 static int vni_list_cmp(void *p1
, void *p2
)
138 const struct bgpevpn
*vpn1
= p1
;
139 const struct bgpevpn
*vpn2
= p2
;
141 return vpn1
->vni
- vpn2
->vni
;
145 * Make vrf import route target hash key.
147 static unsigned int vrf_import_rt_hash_key_make(const void *p
)
149 const struct vrf_irt_node
*irt
= p
;
150 const char *pnt
= irt
->rt
.val
;
152 return jhash(pnt
, 8, 0x5abc1234);
156 * Comparison function for vrf import rt hash
158 static bool vrf_import_rt_hash_cmp(const void *p1
, const void *p2
)
160 const struct vrf_irt_node
*irt1
= p1
;
161 const struct vrf_irt_node
*irt2
= p2
;
163 if (irt1
== NULL
&& irt2
== NULL
)
166 if (irt1
== NULL
|| irt2
== NULL
)
169 return (memcmp(irt1
->rt
.val
, irt2
->rt
.val
, ECOMMUNITY_SIZE
) == 0);
173 * Create a new vrf import_rt in evpn instance
175 static struct vrf_irt_node
*vrf_import_rt_new(struct ecommunity_val
*rt
)
177 struct bgp
*bgp_evpn
= NULL
;
178 struct vrf_irt_node
*irt
;
180 bgp_evpn
= bgp_get_evpn();
182 flog_err(EC_BGP_NO_DFLT
,
183 "vrf import rt new - evpn instance not created yet");
187 irt
= XCALLOC(MTYPE_BGP_EVPN_VRF_IMPORT_RT
,
188 sizeof(struct vrf_irt_node
));
191 irt
->vrfs
= list_new();
194 if (!hash_get(bgp_evpn
->vrf_import_rt_hash
, irt
, hash_alloc_intern
)) {
195 XFREE(MTYPE_BGP_EVPN_VRF_IMPORT_RT
, irt
);
203 * Free the vrf import rt node
205 static void vrf_import_rt_free(struct vrf_irt_node
*irt
)
207 struct bgp
*bgp_evpn
= NULL
;
209 bgp_evpn
= bgp_get_evpn();
211 flog_err(EC_BGP_NO_DFLT
,
212 "vrf import rt free - evpn instance not created yet");
216 hash_release(bgp_evpn
->vrf_import_rt_hash
, irt
);
217 list_delete(&irt
->vrfs
);
218 XFREE(MTYPE_BGP_EVPN_VRF_IMPORT_RT
, irt
);
222 * Function to lookup Import RT node - used to map a RT to set of
223 * VNIs importing routes with that RT.
225 static struct vrf_irt_node
*lookup_vrf_import_rt(struct ecommunity_val
*rt
)
227 struct bgp
*bgp_evpn
= NULL
;
228 struct vrf_irt_node
*irt
;
229 struct vrf_irt_node tmp
;
231 bgp_evpn
= bgp_get_evpn();
235 "vrf import rt lookup - evpn instance not created yet");
239 memset(&tmp
, 0, sizeof(struct vrf_irt_node
));
240 memcpy(&tmp
.rt
, rt
, ECOMMUNITY_SIZE
);
241 irt
= hash_lookup(bgp_evpn
->vrf_import_rt_hash
, &tmp
);
246 * Is specified VRF present on the RT's list of "importing" VRFs?
248 static int is_vrf_present_in_irt_vrfs(struct list
*vrfs
, struct bgp
*bgp_vrf
)
250 struct listnode
*node
= NULL
, *nnode
= NULL
;
251 struct bgp
*tmp_bgp_vrf
= NULL
;
253 for (ALL_LIST_ELEMENTS(vrfs
, node
, nnode
, tmp_bgp_vrf
)) {
254 if (tmp_bgp_vrf
== bgp_vrf
)
261 * Make import route target hash key.
263 static unsigned int import_rt_hash_key_make(const void *p
)
265 const struct irt_node
*irt
= p
;
266 const char *pnt
= irt
->rt
.val
;
268 return jhash(pnt
, 8, 0xdeadbeef);
272 * Comparison function for import rt hash
274 static bool import_rt_hash_cmp(const void *p1
, const void *p2
)
276 const struct irt_node
*irt1
= p1
;
277 const struct irt_node
*irt2
= p2
;
279 if (irt1
== NULL
&& irt2
== NULL
)
282 if (irt1
== NULL
|| irt2
== NULL
)
285 return (memcmp(irt1
->rt
.val
, irt2
->rt
.val
, ECOMMUNITY_SIZE
) == 0);
289 * Create a new import_rt
291 static struct irt_node
*import_rt_new(struct bgp
*bgp
,
292 struct ecommunity_val
*rt
)
294 struct irt_node
*irt
;
299 irt
= XCALLOC(MTYPE_BGP_EVPN_IMPORT_RT
, sizeof(struct irt_node
));
302 irt
->vnis
= list_new();
305 if (!hash_get(bgp
->import_rt_hash
, irt
, hash_alloc_intern
)) {
306 XFREE(MTYPE_BGP_EVPN_IMPORT_RT
, irt
);
314 * Free the import rt node
316 static void import_rt_free(struct bgp
*bgp
, struct irt_node
*irt
)
318 hash_release(bgp
->import_rt_hash
, irt
);
319 list_delete(&irt
->vnis
);
320 XFREE(MTYPE_BGP_EVPN_IMPORT_RT
, irt
);
324 * Function to lookup Import RT node - used to map a RT to set of
325 * VNIs importing routes with that RT.
327 static struct irt_node
*lookup_import_rt(struct bgp
*bgp
,
328 struct ecommunity_val
*rt
)
330 struct irt_node
*irt
;
333 memset(&tmp
, 0, sizeof(struct irt_node
));
334 memcpy(&tmp
.rt
, rt
, ECOMMUNITY_SIZE
);
335 irt
= hash_lookup(bgp
->import_rt_hash
, &tmp
);
340 * Is specified VNI present on the RT's list of "importing" VNIs?
342 static int is_vni_present_in_irt_vnis(struct list
*vnis
, struct bgpevpn
*vpn
)
344 struct listnode
*node
, *nnode
;
345 struct bgpevpn
*tmp_vpn
;
347 for (ALL_LIST_ELEMENTS(vnis
, node
, nnode
, tmp_vpn
)) {
356 * Compare Route Targets.
358 static int evpn_route_target_cmp(struct ecommunity
*ecom1
,
359 struct ecommunity
*ecom2
)
367 if (!ecom1
&& !ecom2
)
370 if (ecom1
->str
&& !ecom2
->str
)
373 if (!ecom1
->str
&& ecom2
->str
)
376 if (!ecom1
->str
&& !ecom2
->str
)
379 return strcmp(ecom1
->str
, ecom2
->str
);
382 static void evpn_xxport_delete_ecomm(void *val
)
384 struct ecommunity
*ecomm
= val
;
385 ecommunity_free(&ecomm
);
389 * Mask off global-admin field of specified extended community (RT),
390 * just retain the local-admin field.
392 static inline void mask_ecom_global_admin(struct ecommunity_val
*dst
,
393 struct ecommunity_val
*src
)
399 if (type
== ECOMMUNITY_ENCODE_AS
) {
400 dst
->val
[2] = dst
->val
[3] = 0;
401 } else if (type
== ECOMMUNITY_ENCODE_AS4
402 || type
== ECOMMUNITY_ENCODE_IP
) {
403 dst
->val
[2] = dst
->val
[3] = 0;
404 dst
->val
[4] = dst
->val
[5] = 0;
409 * Map one RT to specified VRF.
410 * bgp_vrf = BGP vrf instance
412 static void map_vrf_to_rt(struct bgp
*bgp_vrf
, struct ecommunity_val
*eval
)
414 struct vrf_irt_node
*irt
= NULL
;
415 struct ecommunity_val eval_tmp
;
417 /* If using "automatic" RT,
418 * we only care about the local-admin sub-field.
419 * This is to facilitate using L3VNI(VRF-VNI)
420 * as the RT for EBGP peering too.
422 memcpy(&eval_tmp
, eval
, ECOMMUNITY_SIZE
);
423 if (!CHECK_FLAG(bgp_vrf
->vrf_flags
, BGP_VRF_IMPORT_RT_CFGD
))
424 mask_ecom_global_admin(&eval_tmp
, eval
);
426 irt
= lookup_vrf_import_rt(&eval_tmp
);
427 if (irt
&& is_vrf_present_in_irt_vrfs(irt
->vrfs
, bgp_vrf
))
428 /* Already mapped. */
432 irt
= vrf_import_rt_new(&eval_tmp
);
434 /* Add VRF to the list for this RT. */
435 listnode_add(irt
->vrfs
, bgp_vrf
);
439 * Unmap specified VRF from specified RT. If there are no other
440 * VRFs for this RT, then the RT hash is deleted.
441 * bgp_vrf: BGP VRF specific instance
443 static void unmap_vrf_from_rt(struct bgp
*bgp_vrf
, struct vrf_irt_node
*irt
)
445 /* Delete VRF from list for this RT. */
446 listnode_delete(irt
->vrfs
, bgp_vrf
);
447 if (!listnode_head(irt
->vrfs
)) {
448 vrf_import_rt_free(irt
);
453 * Map one RT to specified VNI.
455 static void map_vni_to_rt(struct bgp
*bgp
, struct bgpevpn
*vpn
,
456 struct ecommunity_val
*eval
)
458 struct irt_node
*irt
;
459 struct ecommunity_val eval_tmp
;
461 /* If using "automatic" RT, we only care about the local-admin
463 * This is to facilitate using VNI as the RT for EBGP peering too.
465 memcpy(&eval_tmp
, eval
, ECOMMUNITY_SIZE
);
466 if (!is_import_rt_configured(vpn
))
467 mask_ecom_global_admin(&eval_tmp
, eval
);
469 irt
= lookup_import_rt(bgp
, &eval_tmp
);
471 if (is_vni_present_in_irt_vnis(irt
->vnis
, vpn
))
472 /* Already mapped. */
476 irt
= import_rt_new(bgp
, &eval_tmp
);
480 /* Add VNI to the hash list for this RT. */
481 listnode_add(irt
->vnis
, vpn
);
485 * Unmap specified VNI from specified RT. If there are no other
486 * VNIs for this RT, then the RT hash is deleted.
488 static void unmap_vni_from_rt(struct bgp
*bgp
, struct bgpevpn
*vpn
,
489 struct irt_node
*irt
)
491 /* Delete VNI from hash list for this RT. */
492 listnode_delete(irt
->vnis
, vpn
);
493 if (!listnode_head(irt
->vnis
)) {
494 import_rt_free(bgp
, irt
);
498 static void bgp_evpn_get_rmac_nexthop(struct bgpevpn
*vpn
,
499 struct prefix_evpn
*p
,
500 struct attr
*attr
, uint8_t flags
)
502 struct bgp
*bgp_vrf
= vpn
->bgp_vrf
;
504 memset(&attr
->rmac
, 0, sizeof(struct ethaddr
));
508 if (p
->prefix
.route_type
!= BGP_EVPN_MAC_IP_ROUTE
)
511 /* Copy sys (pip) RMAC and PIP IP as nexthop
512 * in case of route is self MAC-IP,
513 * advertise-pip and advertise-svi-ip features
515 * Otherwise, for all host MAC-IP route's
518 if (CHECK_FLAG(flags
, BGP_EVPN_MACIP_TYPE_SVI_IP
)
519 && bgp_vrf
->evpn_info
->advertise_pip
&&
520 bgp_vrf
->evpn_info
->is_anycast_mac
) {
522 memcpy(&attr
->rmac
, &bgp_vrf
->evpn_info
->pip_rmac
,
524 attr
->nexthop
= bgp_vrf
->evpn_info
->pip_ip
;
525 attr
->mp_nexthop_global_in
=
526 bgp_vrf
->evpn_info
->pip_ip
;
528 memcpy(&attr
->rmac
, &bgp_vrf
->rmac
, ETH_ALEN
);
532 * Create RT extended community automatically from passed information:
533 * of the form AS:VNI.
534 * NOTE: We use only the lower 16 bits of the AS. This is sufficient as
535 * the need is to get a RT value that will be unique across different
536 * VNIs but the same across routers (in the same AS) for a particular
539 static void form_auto_rt(struct bgp
*bgp
, vni_t vni
, struct list
*rtl
)
541 struct ecommunity_val eval
;
542 struct ecommunity
*ecomadd
, *ecom
;
543 bool ecom_found
= false;
544 struct listnode
*node
;
546 if (bgp
->advertise_autort_rfc8365
)
547 vni
|= EVPN_AUTORT_VXLAN
;
548 encode_route_target_as((bgp
->as
& 0xFFFF), vni
, &eval
);
550 ecomadd
= ecommunity_new();
551 ecommunity_add_val(ecomadd
, &eval
);
552 for (ALL_LIST_ELEMENTS_RO(rtl
, node
, ecom
))
553 if (ecommunity_cmp(ecomadd
, ecom
))
557 listnode_add_sort(rtl
, ecomadd
);
561 * Derive RD and RT for a VNI automatically. Invoked at the time of
564 static void derive_rd_rt_for_vni(struct bgp
*bgp
, struct bgpevpn
*vpn
)
566 bgp_evpn_derive_auto_rd(bgp
, vpn
);
567 bgp_evpn_derive_auto_rt_import(bgp
, vpn
);
568 bgp_evpn_derive_auto_rt_export(bgp
, vpn
);
572 * Convert nexthop (remote VTEP IP) into an IPv6 address.
574 static void evpn_convert_nexthop_to_ipv6(struct attr
*attr
)
576 if (BGP_ATTR_NEXTHOP_AFI_IP6(attr
))
578 ipv4_to_ipv4_mapped_ipv6(&attr
->mp_nexthop_global
, attr
->nexthop
);
579 attr
->mp_nexthop_len
= IPV6_MAX_BYTELEN
;
583 * Add (update) or delete MACIP from zebra.
585 static int bgp_zebra_send_remote_macip(struct bgp
*bgp
, struct bgpevpn
*vpn
,
586 struct prefix_evpn
*p
,
587 struct in_addr remote_vtep_ip
, int add
,
588 uint8_t flags
, uint32_t seq
)
592 char buf1
[ETHER_ADDR_STRLEN
];
593 char buf2
[INET6_ADDRSTRLEN
];
594 char buf3
[INET6_ADDRSTRLEN
];
597 if (!zclient
|| zclient
->sock
< 0)
600 /* Don't try to register if Zebra doesn't know of this instance. */
601 if (!IS_BGP_INST_KNOWN_TO_ZEBRA(bgp
)) {
602 if (BGP_DEBUG(zebra
, ZEBRA
))
603 zlog_debug("%s: No zebra instance to talk to, not installing remote macip",
604 __PRETTY_FUNCTION__
);
610 zclient_create_header(
611 s
, add
? ZEBRA_REMOTE_MACIP_ADD
: ZEBRA_REMOTE_MACIP_DEL
,
613 stream_putl(s
, vpn
->vni
);
614 stream_put(s
, &p
->prefix
.macip_addr
.mac
.octet
, ETH_ALEN
); /* Mac Addr */
615 /* IP address length and IP address, if any. */
616 if (is_evpn_prefix_ipaddr_none(p
))
619 ipa_len
= is_evpn_prefix_ipaddr_v4(p
) ? IPV4_MAX_BYTELEN
621 stream_putl(s
, ipa_len
);
622 stream_put(s
, &p
->prefix
.macip_addr
.ip
.ip
.addr
, ipa_len
);
624 stream_put_in_addr(s
, &remote_vtep_ip
);
626 /* TX flags - MAC sticky status and/or gateway mac */
627 /* Also TX the sequence number of the best route. */
629 stream_putc(s
, flags
);
633 stream_putw_at(s
, 0, stream_get_endp(s
));
635 if (bgp_debug_zebra(NULL
))
637 "Tx %s MACIP, VNI %u MAC %s IP %s flags 0x%x seq %u remote VTEP %s",
638 add
? "ADD" : "DEL", vpn
->vni
,
639 prefix_mac2str(&p
->prefix
.macip_addr
.mac
,
641 ipaddr2str(&p
->prefix
.macip_addr
.ip
,
642 buf3
, sizeof(buf3
)), flags
, seq
,
643 inet_ntop(AF_INET
, &remote_vtep_ip
, buf2
,
646 return zclient_send_message(zclient
);
650 * Add (update) or delete remote VTEP from zebra.
652 static int bgp_zebra_send_remote_vtep(struct bgp
*bgp
, struct bgpevpn
*vpn
,
653 struct prefix_evpn
*p
,
654 int flood_control
, int add
)
659 if (!zclient
|| zclient
->sock
< 0)
662 /* Don't try to register if Zebra doesn't know of this instance. */
663 if (!IS_BGP_INST_KNOWN_TO_ZEBRA(bgp
)) {
664 if (BGP_DEBUG(zebra
, ZEBRA
))
665 zlog_debug("%s: No zebra instance to talk to, not installing remote vtep",
666 __PRETTY_FUNCTION__
);
673 zclient_create_header(
674 s
, add
? ZEBRA_REMOTE_VTEP_ADD
: ZEBRA_REMOTE_VTEP_DEL
,
676 stream_putl(s
, vpn
->vni
);
677 if (is_evpn_prefix_ipaddr_v4(p
))
678 stream_put_in_addr(s
, &p
->prefix
.imet_addr
.ip
.ipaddr_v4
);
679 else if (is_evpn_prefix_ipaddr_v6(p
)) {
682 "Bad remote IP when trying to %s remote VTEP for VNI %u",
683 add
? "ADD" : "DEL", vpn
->vni
);
686 stream_putl(s
, flood_control
);
688 stream_putw_at(s
, 0, stream_get_endp(s
));
690 if (bgp_debug_zebra(NULL
))
691 zlog_debug("Tx %s Remote VTEP, VNI %u remote VTEP %s",
692 add
? "ADD" : "DEL", vpn
->vni
,
693 inet_ntoa(p
->prefix
.imet_addr
.ip
.ipaddr_v4
));
695 return zclient_send_message(zclient
);
699 * Build extended community for EVPN ES (type-4) route
701 static void build_evpn_type4_route_extcomm(struct evpnes
*es
,
704 struct ecommunity ecom_encap
;
705 struct ecommunity ecom_es_rt
;
706 struct ecommunity_val eval
;
707 struct ecommunity_val eval_es_rt
;
708 bgp_encap_types tnl_type
;
712 tnl_type
= BGP_ENCAP_TYPE_VXLAN
;
713 memset(&ecom_encap
, 0, sizeof(ecom_encap
));
714 encode_encap_extcomm(tnl_type
, &eval
);
716 ecom_encap
.val
= (uint8_t *)eval
.val
;
717 attr
->ecommunity
= ecommunity_dup(&ecom_encap
);
720 memset(&mac
, 0, sizeof(struct ethaddr
));
721 memset(&ecom_es_rt
, 0, sizeof(ecom_es_rt
));
722 es_get_system_mac(&es
->esi
, &mac
);
723 encode_es_rt_extcomm(&eval_es_rt
, &mac
);
725 ecom_es_rt
.val
= (uint8_t *)eval_es_rt
.val
;
727 ecommunity_merge(attr
->ecommunity
, &ecom_es_rt
);
729 attr
->flag
|= ATTR_FLAG_BIT(BGP_ATTR_EXT_COMMUNITIES
);
733 * Build extended communities for EVPN prefix route.
735 static void build_evpn_type5_route_extcomm(struct bgp
*bgp_vrf
,
738 struct ecommunity ecom_encap
;
739 struct ecommunity ecom_rmac
;
740 struct ecommunity_val eval
;
741 struct ecommunity_val eval_rmac
;
742 bgp_encap_types tnl_type
;
743 struct listnode
*node
, *nnode
;
744 struct ecommunity
*ecom
;
745 struct list
*vrf_export_rtl
= NULL
;
748 tnl_type
= BGP_ENCAP_TYPE_VXLAN
;
749 memset(&ecom_encap
, 0, sizeof(ecom_encap
));
750 encode_encap_extcomm(tnl_type
, &eval
);
752 ecom_encap
.val
= (uint8_t *)eval
.val
;
755 attr
->ecommunity
= ecommunity_dup(&ecom_encap
);
757 /* Add the export RTs for L3VNI/VRF */
758 vrf_export_rtl
= bgp_vrf
->vrf_export_rtl
;
759 for (ALL_LIST_ELEMENTS(vrf_export_rtl
, node
, nnode
, ecom
))
761 ecommunity_merge(attr
->ecommunity
, ecom
);
763 /* add the router mac extended community */
764 if (!is_zero_mac(&attr
->rmac
)) {
765 memset(&ecom_rmac
, 0, sizeof(ecom_rmac
));
766 encode_rmac_extcomm(&eval_rmac
, &attr
->rmac
);
768 ecom_rmac
.val
= (uint8_t *)eval_rmac
.val
;
770 ecommunity_merge(attr
->ecommunity
, &ecom_rmac
);
773 attr
->flag
|= ATTR_FLAG_BIT(BGP_ATTR_EXT_COMMUNITIES
);
777 * Build extended communities for EVPN route.
778 * This function is applicable for type-2 and type-3 routes. The layer-2 RT
779 * and ENCAP extended communities are applicable for all routes.
780 * The default gateway extended community and MAC mobility (sticky) extended
781 * community are added as needed based on passed settings - only for type-2
782 * routes. Likewise, the layer-3 RT and Router MAC extended communities are
783 * added, if present, based on passed settings - only for non-link-local
786 static void build_evpn_route_extcomm(struct bgpevpn
*vpn
, struct attr
*attr
,
789 struct ecommunity ecom_encap
;
790 struct ecommunity ecom_sticky
;
791 struct ecommunity ecom_default_gw
;
792 struct ecommunity ecom_rmac
;
793 struct ecommunity ecom_na
;
794 struct ecommunity_val eval
;
795 struct ecommunity_val eval_sticky
;
796 struct ecommunity_val eval_default_gw
;
797 struct ecommunity_val eval_rmac
;
798 struct ecommunity_val eval_na
;
800 bgp_encap_types tnl_type
;
801 struct listnode
*node
, *nnode
;
802 struct ecommunity
*ecom
;
804 struct list
*vrf_export_rtl
= NULL
;
807 tnl_type
= BGP_ENCAP_TYPE_VXLAN
;
808 memset(&ecom_encap
, 0, sizeof(ecom_encap
));
809 encode_encap_extcomm(tnl_type
, &eval
);
811 ecom_encap
.val
= (uint8_t *)eval
.val
;
814 attr
->ecommunity
= ecommunity_dup(&ecom_encap
);
816 /* Add the export RTs for L2VNI */
817 for (ALL_LIST_ELEMENTS(vpn
->export_rtl
, node
, nnode
, ecom
))
818 attr
->ecommunity
= ecommunity_merge(attr
->ecommunity
, ecom
);
820 /* Add the export RTs for L3VNI if told to - caller determines
821 * when this should be done.
824 vrf_export_rtl
= bgpevpn_get_vrf_export_rtl(vpn
);
825 if (vrf_export_rtl
&& !list_isempty(vrf_export_rtl
)) {
826 for (ALL_LIST_ELEMENTS(vrf_export_rtl
, node
, nnode
,
828 attr
->ecommunity
= ecommunity_merge(
829 attr
->ecommunity
, ecom
);
833 /* Add MAC mobility (sticky) if needed. */
836 memset(&ecom_sticky
, 0, sizeof(ecom_sticky
));
837 encode_mac_mobility_extcomm(1, seqnum
, &eval_sticky
);
838 ecom_sticky
.size
= 1;
839 ecom_sticky
.val
= (uint8_t *)eval_sticky
.val
;
841 ecommunity_merge(attr
->ecommunity
, &ecom_sticky
);
844 /* Add RMAC, if told to. */
846 memset(&ecom_rmac
, 0, sizeof(ecom_rmac
));
847 encode_rmac_extcomm(&eval_rmac
, &attr
->rmac
);
849 ecom_rmac
.val
= (uint8_t *)eval_rmac
.val
;
851 ecommunity_merge(attr
->ecommunity
, &ecom_rmac
);
854 /* Add default gateway, if needed. */
855 if (attr
->default_gw
) {
856 memset(&ecom_default_gw
, 0, sizeof(ecom_default_gw
));
857 encode_default_gw_extcomm(&eval_default_gw
);
858 ecom_default_gw
.size
= 1;
859 ecom_default_gw
.val
= (uint8_t *)eval_default_gw
.val
;
861 ecommunity_merge(attr
->ecommunity
, &ecom_default_gw
);
864 if (attr
->router_flag
) {
865 memset(&ecom_na
, 0, sizeof(ecom_na
));
866 encode_na_flag_extcomm(&eval_na
, attr
->router_flag
);
868 ecom_na
.val
= (uint8_t *)eval_na
.val
;
869 attr
->ecommunity
= ecommunity_merge(attr
->ecommunity
,
873 attr
->flag
|= ATTR_FLAG_BIT(BGP_ATTR_EXT_COMMUNITIES
);
877 * Add MAC mobility extended community to attribute.
879 static void add_mac_mobility_to_attr(uint32_t seq_num
, struct attr
*attr
)
881 struct ecommunity ecom_tmp
;
882 struct ecommunity_val eval
;
883 uint8_t *ecom_val_ptr
;
890 encode_mac_mobility_extcomm(0, seq_num
, &eval
);
892 /* Find current MM ecommunity */
895 if (attr
->ecommunity
) {
896 for (i
= 0; i
< attr
->ecommunity
->size
; i
++) {
897 pnt
= attr
->ecommunity
->val
+ (i
* 8);
901 if (type
== ECOMMUNITY_ENCODE_EVPN
903 == ECOMMUNITY_EVPN_SUBTYPE_MACMOBILITY
) {
904 ecom_val_ptr
= (uint8_t *)(attr
->ecommunity
->val
911 /* Update the existing MM ecommunity */
913 memcpy(ecom_val_ptr
, eval
.val
, sizeof(char) * ECOMMUNITY_SIZE
);
915 /* Add MM to existing */
917 memset(&ecom_tmp
, 0, sizeof(ecom_tmp
));
919 ecom_tmp
.val
= (uint8_t *)eval
.val
;
921 if (attr
->ecommunity
)
923 ecommunity_merge(attr
->ecommunity
, &ecom_tmp
);
925 attr
->ecommunity
= ecommunity_dup(&ecom_tmp
);
929 /* Install EVPN route into zebra. */
930 static int evpn_zebra_install(struct bgp
*bgp
, struct bgpevpn
*vpn
,
931 struct prefix_evpn
*p
, struct bgp_path_info
*pi
)
937 if (p
->prefix
.route_type
== BGP_EVPN_MAC_IP_ROUTE
) {
939 if (pi
->attr
->sticky
)
940 SET_FLAG(flags
, ZEBRA_MACIP_TYPE_STICKY
);
941 if (pi
->attr
->default_gw
)
942 SET_FLAG(flags
, ZEBRA_MACIP_TYPE_GW
);
943 if (is_evpn_prefix_ipaddr_v6(p
) &&
944 pi
->attr
->router_flag
)
945 SET_FLAG(flags
, ZEBRA_MACIP_TYPE_ROUTER_FLAG
);
946 ret
= bgp_zebra_send_remote_macip(
947 bgp
, vpn
, p
, pi
->attr
->nexthop
, 1, flags
,
948 mac_mobility_seqnum(pi
->attr
));
950 switch (pi
->attr
->pmsi_tnl_type
) {
951 case PMSI_TNLTYPE_INGR_REPL
:
952 flood_control
= VXLAN_FLOOD_HEAD_END_REPL
;
955 case PMSI_TNLTYPE_PIM_SM
:
956 flood_control
= VXLAN_FLOOD_PIM_SM
;
960 flood_control
= VXLAN_FLOOD_DISABLED
;
963 ret
= bgp_zebra_send_remote_vtep(bgp
, vpn
, p
, flood_control
, 1);
969 /* Uninstall EVPN route from zebra. */
970 static int evpn_zebra_uninstall(struct bgp
*bgp
, struct bgpevpn
*vpn
,
971 struct prefix_evpn
*p
,
972 struct in_addr remote_vtep_ip
)
976 if (p
->prefix
.route_type
== BGP_EVPN_MAC_IP_ROUTE
)
977 ret
= bgp_zebra_send_remote_macip(bgp
, vpn
, p
, remote_vtep_ip
,
980 ret
= bgp_zebra_send_remote_vtep(bgp
, vpn
, p
,
981 VXLAN_FLOOD_DISABLED
, 0);
987 * Due to MAC mobility, the prior "local" best route has been supplanted
988 * by a "remote" best route. The prior route has to be deleted and withdrawn
991 static void evpn_delete_old_local_route(struct bgp
*bgp
, struct bgpevpn
*vpn
,
993 struct bgp_path_info
*old_local
)
995 struct bgp_node
*global_rn
;
996 struct bgp_path_info
*pi
;
997 afi_t afi
= AFI_L2VPN
;
998 safi_t safi
= SAFI_EVPN
;
1000 /* Locate route node in the global EVPN routing table. Note that
1001 * this table is a 2-level tree (RD-level + Prefix-level) similar to
1004 global_rn
= bgp_afi_node_lookup(bgp
->rib
[afi
][safi
], afi
, safi
,
1005 (struct prefix
*)&rn
->p
, &vpn
->prd
);
1007 /* Delete route entry in the global EVPN table. */
1008 delete_evpn_route_entry(bgp
, afi
, safi
, global_rn
, &pi
);
1010 /* Schedule for processing - withdraws to peers happen from
1014 bgp_process(bgp
, global_rn
, afi
, safi
);
1015 bgp_unlock_node(global_rn
);
1018 /* Delete route entry in the VNI route table, caller to remove. */
1019 bgp_path_info_delete(rn
, old_local
);
1022 static struct in_addr
*es_vtep_new(struct in_addr vtep
)
1026 ip
= XCALLOC(MTYPE_BGP_EVPN_ES_VTEP
, sizeof(struct in_addr
));
1028 ip
->s_addr
= vtep
.s_addr
;
1032 static void es_vtep_free(struct in_addr
*ip
)
1034 XFREE(MTYPE_BGP_EVPN_ES_VTEP
, ip
);
1037 /* check if VTEP is already part of the list */
1038 static int is_vtep_present_in_list(struct list
*list
,
1039 struct in_addr vtep
)
1041 struct listnode
*node
= NULL
;
1042 struct in_addr
*tmp
;
1044 for (ALL_LIST_ELEMENTS_RO(list
, node
, tmp
)) {
1045 if (tmp
->s_addr
== vtep
.s_addr
)
1052 * Best path for ES route was changed,
1053 * update the list of VTEPs for this ES
1055 static int evpn_es_install_vtep(struct bgp
*bgp
,
1057 struct prefix_evpn
*p
,
1058 struct in_addr rvtep
)
1060 struct in_addr
*vtep_ip
;
1062 if (is_vtep_present_in_list(es
->vtep_list
, rvtep
))
1066 vtep_ip
= es_vtep_new(rvtep
);
1068 listnode_add_sort(es
->vtep_list
, vtep_ip
);
1073 * Best path for ES route was changed,
1074 * update the list of VTEPs for this ES
1076 static int evpn_es_uninstall_vtep(struct bgp
*bgp
,
1078 struct prefix_evpn
*p
,
1079 struct in_addr rvtep
)
1081 struct listnode
*node
, *nnode
, *node_to_del
= NULL
;
1082 struct in_addr
*tmp
;
1084 for (ALL_LIST_ELEMENTS(es
->vtep_list
, node
, nnode
, tmp
)) {
1085 if (tmp
->s_addr
== rvtep
.s_addr
) {
1092 list_delete_node(es
->vtep_list
, node_to_del
);
1098 * Calculate the best path for a ES(type-4) route.
1100 static int evpn_es_route_select_install(struct bgp
*bgp
,
1102 struct bgp_node
*rn
)
1105 afi_t afi
= AFI_L2VPN
;
1106 safi_t safi
= SAFI_EVPN
;
1107 struct bgp_path_info
*old_select
; /* old best */
1108 struct bgp_path_info
*new_select
; /* new best */
1109 struct bgp_path_info_pair old_and_new
;
1111 /* Compute the best path. */
1112 bgp_best_selection(bgp
, rn
, &bgp
->maxpaths
[afi
][safi
],
1113 &old_and_new
, afi
, safi
);
1114 old_select
= old_and_new
.old
;
1115 new_select
= old_and_new
.new;
1118 * If the best path hasn't changed - see if something needs to be
1121 if (old_select
&& old_select
== new_select
1122 && old_select
->type
== ZEBRA_ROUTE_BGP
1123 && old_select
->sub_type
== BGP_ROUTE_IMPORTED
1124 && !CHECK_FLAG(rn
->flags
, BGP_NODE_USER_CLEAR
)
1125 && !CHECK_FLAG(old_select
->flags
, BGP_PATH_ATTR_CHANGED
)
1126 && !bgp_addpath_is_addpath_used(&bgp
->tx_addpath
, afi
, safi
)) {
1127 if (bgp_zebra_has_route_changed(rn
, old_select
)) {
1128 ret
= evpn_es_install_vtep(bgp
, es
,
1129 (struct prefix_evpn
*)&rn
->p
,
1130 old_select
->attr
->nexthop
);
1132 UNSET_FLAG(old_select
->flags
, BGP_PATH_MULTIPATH_CHG
);
1133 bgp_zebra_clear_route_change_flags(rn
);
1137 /* If the user did a "clear" this flag will be set */
1138 UNSET_FLAG(rn
->flags
, BGP_NODE_USER_CLEAR
);
1141 * bestpath has changed; update relevant fields and install or uninstall
1142 * into the zebra RIB.
1144 if (old_select
|| new_select
)
1145 bgp_bump_version(rn
);
1148 bgp_path_info_unset_flag(rn
, old_select
, BGP_PATH_SELECTED
);
1150 bgp_path_info_set_flag(rn
, new_select
, BGP_PATH_SELECTED
);
1151 bgp_path_info_unset_flag(rn
, new_select
, BGP_PATH_ATTR_CHANGED
);
1152 UNSET_FLAG(new_select
->flags
, BGP_PATH_MULTIPATH_CHG
);
1155 if (new_select
&& new_select
->type
== ZEBRA_ROUTE_BGP
1156 && new_select
->sub_type
== BGP_ROUTE_IMPORTED
) {
1157 ret
= evpn_es_install_vtep(bgp
, es
,
1158 (struct prefix_evpn
*)&rn
->p
,
1159 new_select
->attr
->nexthop
);
1161 if (old_select
&& old_select
->type
== ZEBRA_ROUTE_BGP
1162 && old_select
->sub_type
== BGP_ROUTE_IMPORTED
)
1163 ret
= evpn_es_uninstall_vtep(
1164 bgp
, es
, (struct prefix_evpn
*)&rn
->p
,
1165 old_select
->attr
->nexthop
);
1168 /* Clear any route change flags. */
1169 bgp_zebra_clear_route_change_flags(rn
);
1171 /* Reap old select bgp_path_info, if it has been removed */
1172 if (old_select
&& CHECK_FLAG(old_select
->flags
, BGP_PATH_REMOVED
))
1173 bgp_path_info_reap(rn
, old_select
);
1179 * Calculate the best path for an EVPN route. Install/update best path in zebra,
1182 static int evpn_route_select_install(struct bgp
*bgp
, struct bgpevpn
*vpn
,
1183 struct bgp_node
*rn
)
1185 struct bgp_path_info
*old_select
, *new_select
;
1186 struct bgp_path_info_pair old_and_new
;
1187 afi_t afi
= AFI_L2VPN
;
1188 safi_t safi
= SAFI_EVPN
;
1191 /* Compute the best path. */
1192 bgp_best_selection(bgp
, rn
, &bgp
->maxpaths
[afi
][safi
], &old_and_new
,
1194 old_select
= old_and_new
.old
;
1195 new_select
= old_and_new
.new;
1197 /* If the best path hasn't changed - see if there is still something to
1201 if (old_select
&& old_select
== new_select
1202 && old_select
->type
== ZEBRA_ROUTE_BGP
1203 && old_select
->sub_type
== BGP_ROUTE_IMPORTED
1204 && !CHECK_FLAG(rn
->flags
, BGP_NODE_USER_CLEAR
)
1205 && !CHECK_FLAG(old_select
->flags
, BGP_PATH_ATTR_CHANGED
)
1206 && !bgp_addpath_is_addpath_used(&bgp
->tx_addpath
, afi
, safi
)) {
1207 if (bgp_zebra_has_route_changed(rn
, old_select
))
1208 ret
= evpn_zebra_install(
1209 bgp
, vpn
, (struct prefix_evpn
*)&rn
->p
,
1211 UNSET_FLAG(old_select
->flags
, BGP_PATH_MULTIPATH_CHG
);
1212 bgp_zebra_clear_route_change_flags(rn
);
1216 /* If the user did a "clear" this flag will be set */
1217 UNSET_FLAG(rn
->flags
, BGP_NODE_USER_CLEAR
);
1219 /* bestpath has changed; update relevant fields and install or uninstall
1220 * into the zebra RIB.
1222 if (old_select
|| new_select
)
1223 bgp_bump_version(rn
);
1226 bgp_path_info_unset_flag(rn
, old_select
, BGP_PATH_SELECTED
);
1228 bgp_path_info_set_flag(rn
, new_select
, BGP_PATH_SELECTED
);
1229 bgp_path_info_unset_flag(rn
, new_select
, BGP_PATH_ATTR_CHANGED
);
1230 UNSET_FLAG(new_select
->flags
, BGP_PATH_MULTIPATH_CHG
);
1233 if (new_select
&& new_select
->type
== ZEBRA_ROUTE_BGP
1234 && new_select
->sub_type
== BGP_ROUTE_IMPORTED
) {
1235 ret
= evpn_zebra_install(bgp
, vpn
, (struct prefix_evpn
*)&rn
->p
,
1238 /* If an old best existed and it was a "local" route, the only
1240 * it would be supplanted is due to MAC mobility procedures. So,
1242 * need to do an implicit delete and withdraw that route from
1245 if (old_select
&& old_select
->peer
== bgp
->peer_self
1246 && old_select
->type
== ZEBRA_ROUTE_BGP
1247 && old_select
->sub_type
== BGP_ROUTE_STATIC
)
1248 evpn_delete_old_local_route(bgp
, vpn
, rn
, old_select
);
1250 if (old_select
&& old_select
->type
== ZEBRA_ROUTE_BGP
1251 && old_select
->sub_type
== BGP_ROUTE_IMPORTED
)
1252 ret
= evpn_zebra_uninstall(bgp
, vpn
,
1253 (struct prefix_evpn
*)&rn
->p
,
1254 old_select
->attr
->nexthop
);
1257 /* Clear any route change flags. */
1258 bgp_zebra_clear_route_change_flags(rn
);
1260 /* Reap old select bgp_path_info, if it has been removed */
1261 if (old_select
&& CHECK_FLAG(old_select
->flags
, BGP_PATH_REMOVED
))
1262 bgp_path_info_reap(rn
, old_select
);
1268 * Return true if the local ri for this rn is of type gateway mac
1270 static int evpn_route_is_def_gw(struct bgp
*bgp
, struct bgp_node
*rn
)
1272 struct bgp_path_info
*tmp_pi
= NULL
;
1273 struct bgp_path_info
*local_pi
= NULL
;
1276 for (tmp_pi
= bgp_node_get_bgp_path_info(rn
); tmp_pi
;
1277 tmp_pi
= tmp_pi
->next
) {
1278 if (tmp_pi
->peer
== bgp
->peer_self
1279 && tmp_pi
->type
== ZEBRA_ROUTE_BGP
1280 && tmp_pi
->sub_type
== BGP_ROUTE_STATIC
)
1287 return local_pi
->attr
->default_gw
;
1292 * Return true if the local ri for this rn has sticky set
1294 static int evpn_route_is_sticky(struct bgp
*bgp
, struct bgp_node
*rn
)
1296 struct bgp_path_info
*tmp_pi
;
1297 struct bgp_path_info
*local_pi
;
1300 for (tmp_pi
= bgp_node_get_bgp_path_info(rn
); tmp_pi
;
1301 tmp_pi
= tmp_pi
->next
) {
1302 if (tmp_pi
->peer
== bgp
->peer_self
1303 && tmp_pi
->type
== ZEBRA_ROUTE_BGP
1304 && tmp_pi
->sub_type
== BGP_ROUTE_STATIC
)
1311 return local_pi
->attr
->sticky
;
1315 * create or update EVPN type4 route entry.
1316 * This could be in the ES table or the global table.
1317 * TODO: handle remote ES (type4) routes as well
1319 static int update_evpn_type4_route_entry(struct bgp
*bgp
, struct evpnes
*es
,
1320 afi_t afi
, safi_t safi
,
1321 struct bgp_node
*rn
, struct attr
*attr
,
1322 int add
, struct bgp_path_info
**ri
,
1325 char buf
[ESI_STR_LEN
];
1326 char buf1
[INET6_ADDRSTRLEN
];
1327 struct bgp_path_info
*tmp_pi
= NULL
;
1328 struct bgp_path_info
*local_pi
= NULL
; /* local route entry if any */
1329 struct bgp_path_info
*remote_pi
= NULL
; /* remote route entry if any */
1330 struct attr
*attr_new
= NULL
;
1331 struct prefix_evpn
*evp
= NULL
;
1335 evp
= (struct prefix_evpn
*)&rn
->p
;
1337 /* locate the local and remote entries if any */
1338 for (tmp_pi
= bgp_node_get_bgp_path_info(rn
); tmp_pi
;
1339 tmp_pi
= tmp_pi
->next
) {
1340 if (tmp_pi
->peer
== bgp
->peer_self
1341 && tmp_pi
->type
== ZEBRA_ROUTE_BGP
1342 && tmp_pi
->sub_type
== BGP_ROUTE_STATIC
)
1344 if (tmp_pi
->type
== ZEBRA_ROUTE_BGP
1345 && tmp_pi
->sub_type
== BGP_ROUTE_IMPORTED
1346 && CHECK_FLAG(tmp_pi
->flags
, BGP_PATH_VALID
))
1350 /* we don't expect to see a remote_ri at this point.
1351 * An ES route has esi + vtep_ip as the key,
1352 * We shouldn't see the same route from any other vtep.
1357 "%u ERROR: local es route for ESI: %s Vtep %s also learnt from remote",
1359 esi_to_str(&evp
->prefix
.es_addr
.esi
, buf
, sizeof(buf
)),
1360 ipaddr2str(&es
->originator_ip
, buf1
, sizeof(buf1
)));
1364 if (!local_pi
&& !add
)
1367 /* create or update the entry */
1370 /* Add or update attribute to hash */
1371 attr_new
= bgp_attr_intern(attr
);
1373 /* Create new route with its attribute. */
1374 tmp_pi
= info_make(ZEBRA_ROUTE_BGP
, BGP_ROUTE_STATIC
, 0,
1375 bgp
->peer_self
, attr_new
, rn
);
1376 SET_FLAG(tmp_pi
->flags
, BGP_PATH_VALID
);
1378 /* add the newly created path to the route-node */
1379 bgp_path_info_add(rn
, tmp_pi
);
1382 if (attrhash_cmp(tmp_pi
->attr
, attr
)
1383 && !CHECK_FLAG(tmp_pi
->flags
, BGP_PATH_REMOVED
))
1386 /* The attribute has changed.
1387 * Add (or update) attribute to hash. */
1388 attr_new
= bgp_attr_intern(attr
);
1389 bgp_path_info_set_flag(rn
, tmp_pi
,
1390 BGP_PATH_ATTR_CHANGED
);
1392 /* Restore route, if needed. */
1393 if (CHECK_FLAG(tmp_pi
->flags
, BGP_PATH_REMOVED
))
1394 bgp_path_info_restore(rn
, tmp_pi
);
1396 /* Unintern existing, set to new. */
1397 bgp_attr_unintern(&tmp_pi
->attr
);
1398 tmp_pi
->attr
= attr_new
;
1399 tmp_pi
->uptime
= bgp_clock();
1403 /* Return back the route entry. */
1408 /* update evpn es (type-4) route */
1409 static int update_evpn_type4_route(struct bgp
*bgp
,
1411 struct prefix_evpn
*p
)
1414 int route_changed
= 0;
1415 char buf
[ESI_STR_LEN
];
1416 char buf1
[INET6_ADDRSTRLEN
];
1417 afi_t afi
= AFI_L2VPN
;
1418 safi_t safi
= SAFI_EVPN
;
1420 struct attr
*attr_new
= NULL
;
1421 struct bgp_node
*rn
= NULL
;
1422 struct bgp_path_info
*pi
= NULL
;
1424 memset(&attr
, 0, sizeof(struct attr
));
1426 /* Build path-attribute for this route. */
1427 bgp_attr_default_set(&attr
, BGP_ORIGIN_IGP
);
1428 attr
.nexthop
= es
->originator_ip
.ipaddr_v4
;
1429 attr
.mp_nexthop_global_in
= es
->originator_ip
.ipaddr_v4
;
1430 attr
.mp_nexthop_len
= BGP_ATTR_NHLEN_IPV4
;
1432 /* Set up extended community. */
1433 build_evpn_type4_route_extcomm(es
, &attr
);
1435 /* First, create (or fetch) route node within the ESI. */
1436 /* NOTE: There is no RD here. */
1437 rn
= bgp_node_get(es
->route_table
, (struct prefix
*)p
);
1439 /* Create or update route entry. */
1440 ret
= update_evpn_type4_route_entry(bgp
, es
, afi
, safi
, rn
, &attr
, 1,
1441 &pi
, &route_changed
);
1443 flog_err(EC_BGP_ES_INVALID
,
1444 "%u ERROR: Failed to updated ES route ESI: %s VTEP %s",
1446 esi_to_str(&p
->prefix
.es_addr
.esi
, buf
, sizeof(buf
)),
1447 ipaddr2str(&es
->originator_ip
, buf1
, sizeof(buf1
)));
1451 attr_new
= pi
->attr
;
1453 /* Perform route selection;
1454 * this is just to set the flags correctly
1455 * as local route in the ES always wins.
1457 evpn_es_route_select_install(bgp
, es
, rn
);
1458 bgp_unlock_node(rn
);
1460 /* If this is a new route or some attribute has changed, export the
1461 * route to the global table. The route will be advertised to peers
1462 * from there. Note that this table is a 2-level tree (RD-level +
1463 * Prefix-level) similar to L3VPN routes.
1465 if (route_changed
) {
1466 struct bgp_path_info
*global_pi
;
1468 rn
= bgp_afi_node_get(bgp
->rib
[afi
][safi
], afi
, safi
,
1469 (struct prefix
*)p
, &es
->prd
);
1470 update_evpn_type4_route_entry(bgp
, es
, afi
, safi
, rn
, attr_new
,
1471 1, &global_pi
, &route_changed
);
1473 /* Schedule for processing and unlock node. */
1474 bgp_process(bgp
, rn
, afi
, safi
);
1475 bgp_unlock_node(rn
);
1478 /* Unintern temporary. */
1479 aspath_unintern(&attr
.aspath
);
1483 static int update_evpn_type5_route_entry(struct bgp
*bgp_evpn
,
1484 struct bgp
*bgp_vrf
, afi_t afi
,
1485 safi_t safi
, struct bgp_node
*rn
,
1486 struct attr
*attr
, int *route_changed
)
1488 struct attr
*attr_new
= NULL
;
1489 struct bgp_path_info
*pi
= NULL
;
1490 mpls_label_t label
= MPLS_INVALID_LABEL
;
1491 struct bgp_path_info
*local_pi
= NULL
;
1492 struct bgp_path_info
*tmp_pi
= NULL
;
1495 /* locate the local route entry if any */
1496 for (tmp_pi
= bgp_node_get_bgp_path_info(rn
); tmp_pi
;
1497 tmp_pi
= tmp_pi
->next
) {
1498 if (tmp_pi
->peer
== bgp_evpn
->peer_self
1499 && tmp_pi
->type
== ZEBRA_ROUTE_BGP
1500 && tmp_pi
->sub_type
== BGP_ROUTE_STATIC
)
1505 * create a new route entry if one doesn't exist.
1506 * Otherwise see if route attr has changed
1510 /* route has changed as this is the first entry */
1513 /* Add (or update) attribute to hash. */
1514 attr_new
= bgp_attr_intern(attr
);
1516 /* create the route info from attribute */
1517 pi
= info_make(ZEBRA_ROUTE_BGP
, BGP_ROUTE_STATIC
, 0,
1518 bgp_evpn
->peer_self
, attr_new
, rn
);
1519 SET_FLAG(pi
->flags
, BGP_PATH_VALID
);
1521 /* Type-5 routes advertise the L3-VNI */
1522 bgp_path_info_extra_get(pi
);
1523 vni2label(bgp_vrf
->l3vni
, &label
);
1524 memcpy(&pi
->extra
->label
, &label
, sizeof(label
));
1525 pi
->extra
->num_labels
= 1;
1527 /* add the route entry to route node*/
1528 bgp_path_info_add(rn
, pi
);
1532 if (!attrhash_cmp(tmp_pi
->attr
, attr
)) {
1534 /* attribute changed */
1537 /* The attribute has changed. */
1538 /* Add (or update) attribute to hash. */
1539 attr_new
= bgp_attr_intern(attr
);
1540 bgp_path_info_set_flag(rn
, tmp_pi
,
1541 BGP_PATH_ATTR_CHANGED
);
1543 /* Restore route, if needed. */
1544 if (CHECK_FLAG(tmp_pi
->flags
, BGP_PATH_REMOVED
))
1545 bgp_path_info_restore(rn
, tmp_pi
);
1547 /* Unintern existing, set to new. */
1548 bgp_attr_unintern(&tmp_pi
->attr
);
1549 tmp_pi
->attr
= attr_new
;
1550 tmp_pi
->uptime
= bgp_clock();
1556 /* update evpn type-5 route entry */
1557 static int update_evpn_type5_route(struct bgp
*bgp_vrf
, struct prefix_evpn
*evp
,
1558 struct attr
*src_attr
)
1560 afi_t afi
= AFI_L2VPN
;
1561 safi_t safi
= SAFI_EVPN
;
1563 struct bgp_node
*rn
= NULL
;
1564 struct bgp
*bgp_evpn
= NULL
;
1565 int route_changed
= 0;
1567 bgp_evpn
= bgp_get_evpn();
1571 /* Build path attribute for this route - use the source attr, if
1572 * present, else treat as locally originated.
1577 memset(&attr
, 0, sizeof(struct attr
));
1578 bgp_attr_default_set(&attr
, BGP_ORIGIN_IGP
);
1581 /* Advertise Primary IP (PIP) is enabled, send individual
1582 * IP (default instance router-id) as nexthop.
1583 * PIP is disabled or vrr interface is not present
1584 * use anycast-IP as nexthop and anycast RMAC.
1586 if (!bgp_vrf
->evpn_info
->advertise_pip
||
1587 (!bgp_vrf
->evpn_info
->is_anycast_mac
)) {
1588 attr
.nexthop
= bgp_vrf
->originator_ip
;
1589 attr
.mp_nexthop_global_in
= bgp_vrf
->originator_ip
;
1590 memcpy(&attr
.rmac
, &bgp_vrf
->rmac
, ETH_ALEN
);
1593 memcpy(&attr
.rmac
, &bgp_vrf
->evpn_info
->pip_rmac
, ETH_ALEN
);
1594 if (bgp_vrf
->evpn_info
->pip_ip
.s_addr
!= INADDR_ANY
) {
1595 attr
.nexthop
= bgp_vrf
->evpn_info
->pip_ip
;
1596 attr
.mp_nexthop_global_in
= bgp_vrf
->evpn_info
->pip_ip
;
1597 } else if (bgp_vrf
->evpn_info
->pip_ip
.s_addr
== INADDR_ANY
)
1598 if (bgp_debug_zebra(NULL
)) {
1599 char buf1
[PREFIX_STRLEN
];
1601 zlog_debug("VRF %s evp %s advertise-pip primary ip is not configured",
1602 vrf_id_to_name(bgp_vrf
->vrf_id
),
1603 prefix2str(evp
, buf1
, sizeof(buf1
)));
1607 if (bgp_debug_zebra(NULL
)) {
1608 char buf
[ETHER_ADDR_STRLEN
];
1609 char buf1
[PREFIX_STRLEN
];
1610 char buf2
[INET6_ADDRSTRLEN
];
1612 zlog_debug("VRF %s type-5 route evp %s RMAC %s nexthop %s",
1613 vrf_id_to_name(bgp_vrf
->vrf_id
),
1614 prefix2str(evp
, buf1
, sizeof(buf1
)),
1615 prefix_mac2str(&attr
.rmac
, buf
, sizeof(buf
)),
1616 inet_ntop(AF_INET
, &attr
.nexthop
, buf2
,
1620 attr
.mp_nexthop_len
= BGP_ATTR_NHLEN_IPV4
;
1622 /* Setup RT and encap extended community */
1623 build_evpn_type5_route_extcomm(bgp_vrf
, &attr
);
1625 /* get the route node in global table */
1626 rn
= bgp_afi_node_get(bgp_evpn
->rib
[afi
][safi
], afi
, safi
,
1627 (struct prefix
*)evp
, &bgp_vrf
->vrf_prd
);
1630 /* create or update the route entry within the route node */
1631 update_evpn_type5_route_entry(bgp_evpn
, bgp_vrf
, afi
, safi
, rn
, &attr
,
1634 /* schedule for processing and unlock node */
1635 if (route_changed
) {
1636 bgp_process(bgp_evpn
, rn
, afi
, safi
);
1637 bgp_unlock_node(rn
);
1640 /* uninten temporary */
1642 aspath_unintern(&attr
.aspath
);
1647 * Create or update EVPN route entry. This could be in the VNI route table
1648 * or the global route table.
1650 static int update_evpn_route_entry(struct bgp
*bgp
, struct bgpevpn
*vpn
,
1651 afi_t afi
, safi_t safi
, struct bgp_node
*rn
,
1652 struct attr
*attr
, int add
,
1653 struct bgp_path_info
**pi
, uint8_t flags
,
1656 struct bgp_path_info
*tmp_pi
;
1657 struct bgp_path_info
*local_pi
;
1658 struct attr
*attr_new
;
1659 mpls_label_t label
[BGP_MAX_LABELS
];
1660 uint32_t num_labels
= 1;
1661 int route_change
= 1;
1663 struct prefix_evpn
*evp
;
1666 evp
= (struct prefix_evpn
*)&rn
->p
;
1667 memset(&label
, 0, sizeof(label
));
1669 /* See if this is an update of an existing route, or a new add. */
1671 for (tmp_pi
= bgp_node_get_bgp_path_info(rn
); tmp_pi
;
1672 tmp_pi
= tmp_pi
->next
) {
1673 if (tmp_pi
->peer
== bgp
->peer_self
1674 && tmp_pi
->type
== ZEBRA_ROUTE_BGP
1675 && tmp_pi
->sub_type
== BGP_ROUTE_STATIC
)
1679 /* If route doesn't exist already, create a new one, if told to.
1680 * Otherwise act based on whether the attributes of the route have
1683 if (!local_pi
&& !add
)
1686 /* For non-GW MACs, update MAC mobility seq number, if needed. */
1687 if (seq
&& !CHECK_FLAG(flags
, ZEBRA_MACIP_TYPE_GW
))
1688 add_mac_mobility_to_attr(seq
, attr
);
1691 /* Add (or update) attribute to hash. */
1692 attr_new
= bgp_attr_intern(attr
);
1694 /* Extract MAC mobility sequence number, if any. */
1695 attr_new
->mm_seqnum
=
1696 bgp_attr_mac_mobility_seqnum(attr_new
, &sticky
);
1697 attr_new
->sticky
= sticky
;
1699 /* Create new route with its attribute. */
1700 tmp_pi
= info_make(ZEBRA_ROUTE_BGP
, BGP_ROUTE_STATIC
, 0,
1701 bgp
->peer_self
, attr_new
, rn
);
1702 SET_FLAG(tmp_pi
->flags
, BGP_PATH_VALID
);
1703 bgp_path_info_extra_get(tmp_pi
);
1705 /* The VNI goes into the 'label' field of the route */
1706 vni2label(vpn
->vni
, &label
[0]);
1708 /* Type-2 routes may carry a second VNI - the L3-VNI.
1709 * Only attach second label if we are advertising two labels for
1712 if (evp
->prefix
.route_type
== BGP_EVPN_MAC_IP_ROUTE
1713 && CHECK_FLAG(vpn
->flags
, VNI_FLAG_USE_TWO_LABELS
)) {
1716 l3vni
= bgpevpn_get_l3vni(vpn
);
1718 vni2label(l3vni
, &label
[1]);
1723 memcpy(&tmp_pi
->extra
->label
, label
, sizeof(label
));
1724 tmp_pi
->extra
->num_labels
= num_labels
;
1725 /* Mark route as self type-2 route */
1726 if (flags
&& CHECK_FLAG(flags
, ZEBRA_MACIP_TYPE_SVI_IP
))
1727 tmp_pi
->extra
->af_flags
= BGP_EVPN_MACIP_TYPE_SVI_IP
;
1728 bgp_path_info_add(rn
, tmp_pi
);
1731 if (attrhash_cmp(tmp_pi
->attr
, attr
)
1732 && !CHECK_FLAG(tmp_pi
->flags
, BGP_PATH_REMOVED
))
1736 * The attributes have changed, type-2 routes needs to
1737 * be advertised with right labels.
1739 vni2label(vpn
->vni
, &label
[0]);
1740 if (evp
->prefix
.route_type
== BGP_EVPN_MAC_IP_ROUTE
1741 && CHECK_FLAG(vpn
->flags
,
1742 VNI_FLAG_USE_TWO_LABELS
)) {
1745 l3vni
= bgpevpn_get_l3vni(vpn
);
1747 vni2label(l3vni
, &label
[1]);
1751 memcpy(&tmp_pi
->extra
->label
, label
, sizeof(label
));
1752 tmp_pi
->extra
->num_labels
= num_labels
;
1754 /* The attribute has changed. */
1755 /* Add (or update) attribute to hash. */
1756 attr_new
= bgp_attr_intern(attr
);
1757 bgp_path_info_set_flag(rn
, tmp_pi
,
1758 BGP_PATH_ATTR_CHANGED
);
1760 /* Extract MAC mobility sequence number, if any. */
1761 attr_new
->mm_seqnum
=
1762 bgp_attr_mac_mobility_seqnum(attr_new
, &sticky
);
1763 attr_new
->sticky
= sticky
;
1765 /* Restore route, if needed. */
1766 if (CHECK_FLAG(tmp_pi
->flags
, BGP_PATH_REMOVED
))
1767 bgp_path_info_restore(rn
, tmp_pi
);
1769 /* Unintern existing, set to new. */
1770 bgp_attr_unintern(&tmp_pi
->attr
);
1771 tmp_pi
->attr
= attr_new
;
1772 tmp_pi
->uptime
= bgp_clock();
1776 /* Return back the route entry. */
1778 return route_change
;
1781 static void evpn_zebra_reinstall_best_route(struct bgp
*bgp
,
1782 struct bgpevpn
*vpn
, struct bgp_node
*rn
)
1784 struct bgp_path_info
*tmp_ri
;
1785 struct bgp_path_info
*curr_select
= NULL
;
1787 for (tmp_ri
= bgp_node_get_bgp_path_info(rn
);
1788 tmp_ri
; tmp_ri
= tmp_ri
->next
) {
1789 if (CHECK_FLAG(tmp_ri
->flags
, BGP_PATH_SELECTED
)) {
1790 curr_select
= tmp_ri
;
1795 if (curr_select
&& curr_select
->type
== ZEBRA_ROUTE_BGP
1796 && curr_select
->sub_type
== BGP_ROUTE_IMPORTED
)
1797 evpn_zebra_install(bgp
, vpn
,
1798 (struct prefix_evpn
*)&rn
->p
,
1803 * If the local route was not selected evict it and tell zebra to re-add
1804 * the best remote dest.
1806 * Typically a local path added by zebra is expected to be selected as
1807 * best. In which case when a remote path wins as best (later)
1808 * evpn_route_select_install itself evicts the older-local-best path.
1810 * However if bgp's add and zebra's add cross paths (race condition) it
1811 * is possible that the local path is no longer the "older" best path.
1812 * It is a path that was never designated as best and hence requires
1813 * additional handling to prevent bgp from injecting and holding on to a
1814 * non-best local path.
1816 static void evpn_cleanup_local_non_best_route(struct bgp
*bgp
,
1817 struct bgpevpn
*vpn
,
1818 struct bgp_node
*rn
,
1819 struct bgp_path_info
*local_pi
)
1821 char buf
[PREFIX_STRLEN
];
1823 /* local path was not picked as the winner; kick it out */
1824 if (bgp_debug_zebra(NULL
)) {
1825 zlog_debug("evicting local evpn prefix %s as remote won",
1826 prefix2str(&rn
->p
, buf
, sizeof(buf
)));
1828 evpn_delete_old_local_route(bgp
, vpn
, rn
, local_pi
);
1829 bgp_path_info_reap(rn
, local_pi
);
1831 /* tell zebra to re-add the best remote path */
1832 evpn_zebra_reinstall_best_route(bgp
, vpn
, rn
);
1836 * Create or update EVPN route (of type based on prefix) for specified VNI
1837 * and schedule for processing.
1839 static int update_evpn_route(struct bgp
*bgp
, struct bgpevpn
*vpn
,
1840 struct prefix_evpn
*p
, uint8_t flags
,
1843 struct bgp_node
*rn
;
1845 struct attr
*attr_new
;
1846 int add_l3_ecomm
= 0;
1847 struct bgp_path_info
*pi
;
1848 afi_t afi
= AFI_L2VPN
;
1849 safi_t safi
= SAFI_EVPN
;
1852 memset(&attr
, 0, sizeof(struct attr
));
1854 /* Build path-attribute for this route. */
1855 bgp_attr_default_set(&attr
, BGP_ORIGIN_IGP
);
1856 attr
.nexthop
= vpn
->originator_ip
;
1857 attr
.mp_nexthop_global_in
= vpn
->originator_ip
;
1858 attr
.mp_nexthop_len
= BGP_ATTR_NHLEN_IPV4
;
1859 attr
.sticky
= CHECK_FLAG(flags
, ZEBRA_MACIP_TYPE_STICKY
) ? 1 : 0;
1860 attr
.default_gw
= CHECK_FLAG(flags
, ZEBRA_MACIP_TYPE_GW
) ? 1 : 0;
1861 attr
.router_flag
= CHECK_FLAG(flags
,
1862 ZEBRA_MACIP_TYPE_ROUTER_FLAG
) ? 1 : 0;
1864 /* PMSI is only needed for type-3 routes */
1865 if (p
->prefix
.route_type
== BGP_EVPN_IMET_ROUTE
) {
1866 attr
.flag
|= ATTR_FLAG_BIT(BGP_ATTR_PMSI_TUNNEL
);
1867 attr
.pmsi_tnl_type
= PMSI_TNLTYPE_INGR_REPL
;
1870 /* router mac is only needed for type-2 routes here. */
1871 if (p
->prefix
.route_type
== BGP_EVPN_MAC_IP_ROUTE
) {
1872 uint8_t af_flags
= 0;
1874 if (CHECK_FLAG(flags
, ZEBRA_MACIP_TYPE_SVI_IP
))
1875 SET_FLAG(af_flags
, BGP_EVPN_MACIP_TYPE_SVI_IP
);
1877 bgp_evpn_get_rmac_nexthop(vpn
, p
, &attr
, af_flags
);
1879 if (bgp_debug_zebra(NULL
)) {
1880 char buf
[ETHER_ADDR_STRLEN
];
1881 char buf1
[PREFIX_STRLEN
];
1883 zlog_debug("VRF %s vni %u type-2 route evp %s RMAC %s nexthop %s",
1885 vrf_id_to_name(vpn
->bgp_vrf
->vrf_id
) : " ",
1887 prefix2str(p
, buf1
, sizeof(buf1
)),
1888 prefix_mac2str(&attr
.rmac
, buf
,
1890 inet_ntoa(attr
.mp_nexthop_global_in
));
1894 vni2label(vpn
->vni
, &(attr
.label
));
1896 /* Include L3 VNI related RTs and RMAC for type-2 routes, if they're
1897 * IPv4 or IPv6 global addresses and we're advertising L3VNI with
1900 if (p
->prefix
.route_type
== BGP_EVPN_MAC_IP_ROUTE
&&
1901 (is_evpn_prefix_ipaddr_v4(p
) ||
1902 !IN6_IS_ADDR_LINKLOCAL(&p
->prefix
.macip_addr
.ip
.ipaddr_v6
)) &&
1903 CHECK_FLAG(vpn
->flags
, VNI_FLAG_USE_TWO_LABELS
) &&
1904 bgpevpn_get_l3vni(vpn
))
1907 /* Set up extended community. */
1908 build_evpn_route_extcomm(vpn
, &attr
, add_l3_ecomm
);
1910 /* First, create (or fetch) route node within the VNI. */
1911 /* NOTE: There is no RD here. */
1912 rn
= bgp_node_get(vpn
->route_table
, (struct prefix
*)p
);
1914 /* Create or update route entry. */
1915 route_change
= update_evpn_route_entry(bgp
, vpn
, afi
, safi
, rn
, &attr
,
1916 1, &pi
, flags
, seq
);
1918 attr_new
= pi
->attr
;
1920 /* lock ri to prevent freeing in evpn_route_select_install */
1921 bgp_path_info_lock(pi
);
1923 /* Perform route selection. Normally, the local route in the
1924 * VNI is expected to win and be the best route. However, if
1925 * there is a race condition where a host moved from local to
1926 * remote and the remote route was received in BGP just prior
1927 * to the local MACIP notification from zebra, the remote
1928 * route would win, and we should evict the defunct local route
1929 * and (re)install the remote route into zebra.
1931 evpn_route_select_install(bgp
, vpn
, rn
);
1933 * If the new local route was not selected evict it and tell zebra
1934 * to re-add the best remote dest. BGP doesn't retain non-best local
1937 if (!CHECK_FLAG(pi
->flags
, BGP_PATH_SELECTED
)) {
1939 evpn_cleanup_local_non_best_route(bgp
, vpn
, rn
, pi
);
1941 bgp_path_info_unlock(pi
);
1943 bgp_unlock_node(rn
);
1945 /* If this is a new route or some attribute has changed, export the
1946 * route to the global table. The route will be advertised to peers
1947 * from there. Note that this table is a 2-level tree (RD-level +
1948 * Prefix-level) similar to L3VPN routes.
1951 struct bgp_path_info
*global_pi
;
1953 rn
= bgp_afi_node_get(bgp
->rib
[afi
][safi
], afi
, safi
,
1954 (struct prefix
*)p
, &vpn
->prd
);
1955 update_evpn_route_entry(bgp
, vpn
, afi
, safi
, rn
, attr_new
, 1,
1956 &global_pi
, flags
, seq
);
1958 /* Schedule for processing and unlock node. */
1959 bgp_process(bgp
, rn
, afi
, safi
);
1960 bgp_unlock_node(rn
);
1963 /* Unintern temporary. */
1964 aspath_unintern(&attr
.aspath
);
1970 * Delete EVPN route entry.
1971 * The entry can be in ESI/VNI table or the global table.
1973 static void delete_evpn_route_entry(struct bgp
*bgp
, afi_t afi
, safi_t safi
,
1974 struct bgp_node
*rn
,
1975 struct bgp_path_info
**pi
)
1977 struct bgp_path_info
*tmp_pi
;
1981 /* Now, find matching route. */
1982 for (tmp_pi
= bgp_node_get_bgp_path_info(rn
); tmp_pi
;
1983 tmp_pi
= tmp_pi
->next
)
1984 if (tmp_pi
->peer
== bgp
->peer_self
1985 && tmp_pi
->type
== ZEBRA_ROUTE_BGP
1986 && tmp_pi
->sub_type
== BGP_ROUTE_STATIC
)
1991 /* Mark route for delete. */
1993 bgp_path_info_delete(rn
, tmp_pi
);
1998 /* Delete EVPN ES (type-4) route */
1999 static int delete_evpn_type4_route(struct bgp
*bgp
,
2001 struct prefix_evpn
*p
)
2003 afi_t afi
= AFI_L2VPN
;
2004 safi_t safi
= SAFI_EVPN
;
2005 struct bgp_path_info
*pi
;
2006 struct bgp_node
*rn
= NULL
; /* rn in esi table */
2007 struct bgp_node
*global_rn
= NULL
; /* rn in global table */
2009 /* First, locate the route node within the ESI.
2010 * If it doesn't exist, ther is nothing to do.
2011 * Note: there is no RD here.
2013 rn
= bgp_node_lookup(es
->route_table
, (struct prefix
*)p
);
2017 /* Next, locate route node in the global EVPN routing table.
2018 * Note that this table is a 2-level tree (RD-level + Prefix-level)
2020 global_rn
= bgp_afi_node_lookup(bgp
->rib
[afi
][safi
], afi
, safi
,
2021 (struct prefix
*)p
, &es
->prd
);
2024 /* Delete route entry in the global EVPN table. */
2025 delete_evpn_route_entry(bgp
, afi
, safi
, global_rn
, &pi
);
2027 /* Schedule for processing - withdraws to peers happen from
2031 bgp_process(bgp
, global_rn
, afi
, safi
);
2032 bgp_unlock_node(global_rn
);
2036 * Delete route entry in the ESI route table.
2037 * This can just be removed.
2039 delete_evpn_route_entry(bgp
, afi
, safi
, rn
, &pi
);
2041 bgp_path_info_reap(rn
, pi
);
2042 bgp_unlock_node(rn
);
2046 /* Delete EVPN type5 route */
2047 static int delete_evpn_type5_route(struct bgp
*bgp_vrf
, struct prefix_evpn
*evp
)
2049 afi_t afi
= AFI_L2VPN
;
2050 safi_t safi
= SAFI_EVPN
;
2051 struct bgp_node
*rn
= NULL
;
2052 struct bgp_path_info
*pi
= NULL
;
2053 struct bgp
*bgp_evpn
= NULL
; /* evpn bgp instance */
2055 bgp_evpn
= bgp_get_evpn();
2059 /* locate the global route entry for this type-5 prefix */
2060 rn
= bgp_afi_node_lookup(bgp_evpn
->rib
[afi
][safi
], afi
, safi
,
2061 (struct prefix
*)evp
, &bgp_vrf
->vrf_prd
);
2065 delete_evpn_route_entry(bgp_evpn
, afi
, safi
, rn
, &pi
);
2067 bgp_process(bgp_evpn
, rn
, afi
, safi
);
2068 bgp_unlock_node(rn
);
2073 * Delete EVPN route (of type based on prefix) for specified VNI and
2074 * schedule for processing.
2076 static int delete_evpn_route(struct bgp
*bgp
, struct bgpevpn
*vpn
,
2077 struct prefix_evpn
*p
)
2079 struct bgp_node
*rn
, *global_rn
;
2080 struct bgp_path_info
*pi
;
2081 afi_t afi
= AFI_L2VPN
;
2082 safi_t safi
= SAFI_EVPN
;
2084 /* First, locate the route node within the VNI. If it doesn't exist,
2086 * is nothing further to do.
2088 /* NOTE: There is no RD here. */
2089 rn
= bgp_node_lookup(vpn
->route_table
, (struct prefix
*)p
);
2093 /* Next, locate route node in the global EVPN routing table. Note that
2094 * this table is a 2-level tree (RD-level + Prefix-level) similar to
2097 global_rn
= bgp_afi_node_lookup(bgp
->rib
[afi
][safi
], afi
, safi
,
2098 (struct prefix
*)p
, &vpn
->prd
);
2100 /* Delete route entry in the global EVPN table. */
2101 delete_evpn_route_entry(bgp
, afi
, safi
, global_rn
, &pi
);
2103 /* Schedule for processing - withdraws to peers happen from
2107 bgp_process(bgp
, global_rn
, afi
, safi
);
2108 bgp_unlock_node(global_rn
);
2111 /* Delete route entry in the VNI route table. This can just be removed.
2113 delete_evpn_route_entry(bgp
, afi
, safi
, rn
, &pi
);
2115 bgp_path_info_reap(rn
, pi
);
2116 evpn_route_select_install(bgp
, vpn
, rn
);
2118 bgp_unlock_node(rn
);
2124 * Update all type-2 (MACIP) local routes for this VNI - these should also
2125 * be scheduled for advertise to peers.
2127 static int update_all_type2_routes(struct bgp
*bgp
, struct bgpevpn
*vpn
)
2131 struct bgp_node
*rn
;
2132 struct bgp_path_info
*pi
, *tmp_pi
;
2134 struct attr
*attr_new
;
2136 int add_l3_ecomm
= 0;
2141 /* Walk this VNI's route table and update local type-2 routes. For any
2142 * routes updated, update corresponding entry in the global table too.
2144 for (rn
= bgp_table_top(vpn
->route_table
); rn
;
2145 rn
= bgp_route_next(rn
)) {
2146 struct prefix_evpn
*evp
= (struct prefix_evpn
*)&rn
->p
;
2147 struct bgp_node
*rd_rn
;
2148 struct bgp_path_info
*global_pi
;
2150 if (evp
->prefix
.route_type
!= BGP_EVPN_MAC_IP_ROUTE
)
2153 /* Identify local route. */
2154 for (tmp_pi
= bgp_node_get_bgp_path_info(rn
); tmp_pi
;
2155 tmp_pi
= tmp_pi
->next
) {
2156 if (tmp_pi
->peer
== bgp
->peer_self
2157 && tmp_pi
->type
== ZEBRA_ROUTE_BGP
2158 && tmp_pi
->sub_type
== BGP_ROUTE_STATIC
)
2166 * Build attribute per local route as the MAC mobility and
2167 * some other values could differ for different routes. The
2168 * attributes will be shared in the hash table.
2170 bgp_attr_default_set(&attr
, BGP_ORIGIN_IGP
);
2171 attr
.nexthop
= vpn
->originator_ip
;
2172 attr
.mp_nexthop_global_in
= vpn
->originator_ip
;
2173 attr
.mp_nexthop_len
= BGP_ATTR_NHLEN_IPV4
;
2174 bgp_evpn_get_rmac_nexthop(vpn
, evp
, &attr
,
2175 tmp_pi
->extra
->af_flags
);
2177 if (evpn_route_is_sticky(bgp
, rn
))
2179 else if (evpn_route_is_def_gw(bgp
, rn
)) {
2180 attr
.default_gw
= 1;
2181 if (is_evpn_prefix_ipaddr_v6(evp
))
2182 attr
.router_flag
= 1;
2185 if (bgp_debug_zebra(NULL
)) {
2186 char buf
[ETHER_ADDR_STRLEN
];
2187 char buf1
[PREFIX_STRLEN
];
2189 zlog_debug("VRF %s vni %u evp %s RMAC %s nexthop %s",
2191 vrf_id_to_name(vpn
->bgp_vrf
->vrf_id
) : " ",
2193 prefix2str(evp
, buf1
, sizeof(buf1
)),
2194 prefix_mac2str(&attr
.rmac
, buf
, sizeof(buf
)),
2195 inet_ntoa(attr
.mp_nexthop_global_in
));
2198 /* Add L3 VNI RTs and RMAC for non IPv6 link-local if
2199 * using L3 VNI for type-2 routes also.
2201 if ((is_evpn_prefix_ipaddr_v4(evp
) ||
2202 !IN6_IS_ADDR_LINKLOCAL(
2203 &evp
->prefix
.macip_addr
.ip
.ipaddr_v6
)) &&
2204 CHECK_FLAG(vpn
->flags
, VNI_FLAG_USE_TWO_LABELS
) &&
2205 bgpevpn_get_l3vni(vpn
))
2208 /* Set up extended community. */
2209 build_evpn_route_extcomm(vpn
, &attr
, add_l3_ecomm
);
2211 seq
= mac_mobility_seqnum(tmp_pi
->attr
);
2213 /* Update the route entry. */
2214 update_evpn_route_entry(bgp
, vpn
, afi
, safi
, rn
, &attr
, 0, &pi
,
2217 /* lock ri to prevent freeing in evpn_route_select_install */
2218 bgp_path_info_lock(pi
);
2220 /* Perform route selection. Normally, the local route in the
2221 * VNI is expected to win and be the best route. However,
2222 * under peculiar situations (e.g., tunnel (next hop) IP change
2223 * that causes best selection to be based on next hop), a
2224 * remote route could win. If the local route is the best,
2225 * ensure it is updated in the global EVPN route table and
2226 * advertised to peers; otherwise, ensure it is evicted and
2227 * (re)install the remote route into zebra.
2229 evpn_route_select_install(bgp
, vpn
, rn
);
2230 if (!CHECK_FLAG(pi
->flags
, BGP_PATH_SELECTED
)) {
2231 evpn_cleanup_local_non_best_route(bgp
, vpn
, rn
, pi
);
2233 bgp_path_info_unlock(pi
);
2235 attr_new
= pi
->attr
;
2237 bgp_path_info_unlock(pi
);
2239 /* Update route in global routing table. */
2240 rd_rn
= bgp_afi_node_get(bgp
->rib
[afi
][safi
], afi
, safi
,
2241 (struct prefix
*)evp
, &vpn
->prd
);
2243 update_evpn_route_entry(bgp
, vpn
, afi
, safi
, rd_rn
,
2244 attr_new
, 0, &global_pi
, 0,
2245 mac_mobility_seqnum(attr_new
));
2247 /* Schedule for processing and unlock node. */
2248 bgp_process(bgp
, rd_rn
, afi
, safi
);
2249 bgp_unlock_node(rd_rn
);
2252 /* Unintern temporary. */
2253 aspath_unintern(&attr
.aspath
);
2261 * Delete all type-2 (MACIP) local routes for this VNI - only from the
2262 * global routing table. These are also scheduled for withdraw from peers.
2264 static int delete_global_type2_routes(struct bgp
*bgp
, struct bgpevpn
*vpn
)
2268 struct bgp_node
*rdrn
, *rn
;
2269 struct bgp_table
*table
;
2270 struct bgp_path_info
*pi
;
2275 rdrn
= bgp_node_lookup(bgp
->rib
[afi
][safi
], (struct prefix
*)&vpn
->prd
);
2276 if (rdrn
&& bgp_node_has_bgp_path_info_data(rdrn
)) {
2277 table
= bgp_node_get_bgp_table_info(rdrn
);
2278 for (rn
= bgp_table_top(table
); rn
; rn
= bgp_route_next(rn
)) {
2279 struct prefix_evpn
*evp
= (struct prefix_evpn
*)&rn
->p
;
2281 if (evp
->prefix
.route_type
!= BGP_EVPN_MAC_IP_ROUTE
)
2284 delete_evpn_route_entry(bgp
, afi
, safi
, rn
, &pi
);
2286 bgp_process(bgp
, rn
, afi
, safi
);
2290 /* Unlock RD node. */
2292 bgp_unlock_node(rdrn
);
2298 * Delete all type-2 (MACIP) local routes for this VNI - from the global
2299 * table as well as the per-VNI route table.
2301 static int delete_all_type2_routes(struct bgp
*bgp
, struct bgpevpn
*vpn
)
2305 struct bgp_node
*rn
;
2306 struct bgp_path_info
*pi
;
2311 /* First, walk the global route table for this VNI's type-2 local
2313 * EVPN routes are a 2-level table, first get the RD table.
2315 delete_global_type2_routes(bgp
, vpn
);
2317 /* Next, walk this VNI's route table and delete local type-2 routes. */
2318 for (rn
= bgp_table_top(vpn
->route_table
); rn
;
2319 rn
= bgp_route_next(rn
)) {
2320 struct prefix_evpn
*evp
= (struct prefix_evpn
*)&rn
->p
;
2322 if (evp
->prefix
.route_type
!= BGP_EVPN_MAC_IP_ROUTE
)
2325 delete_evpn_route_entry(bgp
, afi
, safi
, rn
, &pi
);
2327 /* Route entry in local table gets deleted immediately. */
2329 bgp_path_info_reap(rn
, pi
);
2336 * Delete all routes in per ES route-table
2338 static int delete_all_es_routes(struct bgp
*bgp
, struct evpnes
*es
)
2340 struct bgp_node
*rn
;
2341 struct bgp_path_info
*pi
, *nextpi
;
2343 /* Walk this ES's route table and delete all routes. */
2344 for (rn
= bgp_table_top(es
->route_table
); rn
;
2345 rn
= bgp_route_next(rn
)) {
2346 for (pi
= bgp_node_get_bgp_path_info(rn
);
2347 (pi
!= NULL
) && (nextpi
= pi
->next
, 1); pi
= nextpi
) {
2348 bgp_path_info_delete(rn
, pi
);
2349 bgp_path_info_reap(rn
, pi
);
2357 * Delete all routes in the per-VNI route table.
2359 static int delete_all_vni_routes(struct bgp
*bgp
, struct bgpevpn
*vpn
)
2361 struct bgp_node
*rn
;
2362 struct bgp_path_info
*pi
, *nextpi
;
2364 /* Walk this VNI's route table and delete all routes. */
2365 for (rn
= bgp_table_top(vpn
->route_table
); rn
;
2366 rn
= bgp_route_next(rn
)) {
2367 for (pi
= bgp_node_get_bgp_path_info(rn
);
2368 (pi
!= NULL
) && (nextpi
= pi
->next
, 1); pi
= nextpi
) {
2369 bgp_path_info_delete(rn
, pi
);
2370 bgp_path_info_reap(rn
, pi
);
2377 /* BUM traffic flood mode per-l2-vni */
2378 static int bgp_evpn_vni_flood_mode_get(struct bgp
*bgp
,
2379 struct bgpevpn
*vpn
)
2381 /* if flooding has been globally disabled per-vni mode is
2384 if (bgp
->vxlan_flood_ctrl
== VXLAN_FLOOD_DISABLED
)
2385 return VXLAN_FLOOD_DISABLED
;
2387 /* if mcast group ip has been specified we use a PIM-SM MDT */
2388 if (vpn
->mcast_grp
.s_addr
!= INADDR_ANY
)
2389 return VXLAN_FLOOD_PIM_SM
;
2391 /* default is ingress replication */
2392 return VXLAN_FLOOD_HEAD_END_REPL
;
2396 * Update (and advertise) local routes for a VNI. Invoked upon the VNI
2397 * export RT getting modified or change to tunnel IP. Note that these
2398 * situations need the route in the per-VNI table as well as the global
2399 * table to be updated (as attributes change).
2401 int update_routes_for_vni(struct bgp
*bgp
, struct bgpevpn
*vpn
)
2404 struct prefix_evpn p
;
2406 /* Update and advertise the type-3 route (only one) followed by the
2407 * locally learnt type-2 routes (MACIP) - for this VNI.
2409 * RT-3 only if doing head-end replication
2411 if (bgp_evpn_vni_flood_mode_get(bgp
, vpn
)
2412 == VXLAN_FLOOD_HEAD_END_REPL
) {
2413 build_evpn_type3_prefix(&p
, vpn
->originator_ip
);
2414 ret
= update_evpn_route(bgp
, vpn
, &p
, 0, 0);
2419 return update_all_type2_routes(bgp
, vpn
);
2422 /* Delete (and withdraw) local routes for specified ES from global and ES table.
2423 * Also remove all other routes from the per ES table.
2424 * Invoked when ES is deleted.
2426 static int delete_routes_for_es(struct bgp
*bgp
, struct evpnes
*es
)
2429 char buf
[ESI_STR_LEN
];
2430 struct prefix_evpn p
;
2432 /* Delete and withdraw locally learnt ES route */
2433 build_evpn_type4_prefix(&p
, &es
->esi
, es
->originator_ip
.ipaddr_v4
);
2434 ret
= delete_evpn_type4_route(bgp
, es
, &p
);
2436 flog_err(EC_BGP_EVPN_ROUTE_DELETE
,
2437 "%u failed to delete type-4 route for ESI %s",
2438 bgp
->vrf_id
, esi_to_str(&es
->esi
, buf
, sizeof(buf
)));
2441 /* Delete all routes from per ES table */
2442 return delete_all_es_routes(bgp
, es
);
2446 * Delete (and withdraw) local routes for specified VNI from the global
2447 * table and per-VNI table. After this, remove all other routes from
2448 * the per-VNI table. Invoked upon the VNI being deleted or EVPN
2449 * (advertise-all-vni) being disabled.
2451 static int delete_routes_for_vni(struct bgp
*bgp
, struct bgpevpn
*vpn
)
2454 struct prefix_evpn p
;
2456 /* Delete and withdraw locally learnt type-2 routes (MACIP)
2457 * followed by type-3 routes (only one) - for this VNI.
2459 ret
= delete_all_type2_routes(bgp
, vpn
);
2463 build_evpn_type3_prefix(&p
, vpn
->originator_ip
);
2464 ret
= delete_evpn_route(bgp
, vpn
, &p
);
2468 /* Delete all routes from the per-VNI table. */
2469 return delete_all_vni_routes(bgp
, vpn
);
2473 * There is a flood mcast IP address change. Update the mcast-grp and
2474 * remove the type-3 route if any. A new type-3 route will be generated
2475 * post tunnel_ip update if the new flood mode is head-end-replication.
2477 static int bgp_evpn_mcast_grp_change(struct bgp
*bgp
, struct bgpevpn
*vpn
,
2478 struct in_addr mcast_grp
)
2480 struct prefix_evpn p
;
2482 vpn
->mcast_grp
= mcast_grp
;
2484 if (is_vni_live(vpn
)) {
2485 build_evpn_type3_prefix(&p
, vpn
->originator_ip
);
2486 delete_evpn_route(bgp
, vpn
, &p
);
2493 * There is a tunnel endpoint IP address change for this VNI, delete
2494 * prior type-3 route (if needed) and update.
2495 * Note: Route re-advertisement happens elsewhere after other processing
2498 static int handle_tunnel_ip_change(struct bgp
*bgp
, struct bgpevpn
*vpn
,
2499 struct in_addr originator_ip
)
2501 struct prefix_evpn p
;
2503 /* If VNI is not live, we only need to update the originator ip */
2504 if (!is_vni_live(vpn
)) {
2505 vpn
->originator_ip
= originator_ip
;
2509 /* Update the tunnel-ip hash */
2510 bgp_tip_del(bgp
, &vpn
->originator_ip
);
2511 bgp_tip_add(bgp
, &originator_ip
);
2513 /* filter routes as martian nexthop db has changed */
2514 bgp_filter_evpn_routes_upon_martian_nh_change(bgp
);
2516 /* Need to withdraw type-3 route as the originator IP is part
2519 build_evpn_type3_prefix(&p
, vpn
->originator_ip
);
2520 delete_evpn_route(bgp
, vpn
, &p
);
2522 /* Update the tunnel IP and re-advertise all routes for this VNI. */
2523 vpn
->originator_ip
= originator_ip
;
2527 static struct bgp_path_info
*
2528 bgp_create_evpn_bgp_path_info(struct bgp_path_info
*parent_pi
,
2529 struct bgp_node
*rn
)
2531 struct attr
*attr_new
;
2532 struct bgp_path_info
*pi
;
2534 /* Add (or update) attribute to hash. */
2535 attr_new
= bgp_attr_intern(parent_pi
->attr
);
2537 /* Create new route with its attribute. */
2538 pi
= info_make(parent_pi
->type
, BGP_ROUTE_IMPORTED
, 0, parent_pi
->peer
,
2540 SET_FLAG(pi
->flags
, BGP_PATH_VALID
);
2541 bgp_path_info_extra_get(pi
);
2542 pi
->extra
->parent
= bgp_path_info_lock(parent_pi
);
2543 bgp_lock_node((struct bgp_node
*)parent_pi
->net
);
2544 if (parent_pi
->extra
) {
2545 memcpy(&pi
->extra
->label
, &parent_pi
->extra
->label
,
2546 sizeof(pi
->extra
->label
));
2547 pi
->extra
->num_labels
= parent_pi
->extra
->num_labels
;
2549 bgp_path_info_add(rn
, pi
);
2554 /* Install EVPN route entry in ES */
2555 static int install_evpn_route_entry_in_es(struct bgp
*bgp
, struct evpnes
*es
,
2556 struct prefix_evpn
*p
,
2557 struct bgp_path_info
*parent_pi
)
2560 struct bgp_node
*rn
= NULL
;
2561 struct bgp_path_info
*pi
= NULL
;
2562 struct attr
*attr_new
= NULL
;
2564 /* Create (or fetch) route within the VNI.
2565 * NOTE: There is no RD here.
2567 rn
= bgp_node_get(es
->route_table
, (struct prefix
*)p
);
2569 /* Check if route entry is already present. */
2570 for (pi
= bgp_node_get_bgp_path_info(rn
); pi
; pi
= pi
->next
)
2572 && (struct bgp_path_info
*)pi
->extra
->parent
== parent_pi
)
2576 /* Add (or update) attribute to hash. */
2577 attr_new
= bgp_attr_intern(parent_pi
->attr
);
2579 /* Create new route with its attribute. */
2580 pi
= info_make(parent_pi
->type
, BGP_ROUTE_IMPORTED
, 0,
2581 parent_pi
->peer
, attr_new
, rn
);
2582 SET_FLAG(pi
->flags
, BGP_PATH_VALID
);
2583 bgp_path_info_extra_get(pi
);
2584 pi
->extra
->parent
= bgp_path_info_lock(parent_pi
);
2585 bgp_lock_node((struct bgp_node
*)parent_pi
->net
);
2586 bgp_path_info_add(rn
, pi
);
2588 if (attrhash_cmp(pi
->attr
, parent_pi
->attr
)
2589 && !CHECK_FLAG(pi
->flags
, BGP_PATH_REMOVED
)) {
2590 bgp_unlock_node(rn
);
2593 /* The attribute has changed. */
2594 /* Add (or update) attribute to hash. */
2595 attr_new
= bgp_attr_intern(parent_pi
->attr
);
2597 /* Restore route, if needed. */
2598 if (CHECK_FLAG(pi
->flags
, BGP_PATH_REMOVED
))
2599 bgp_path_info_restore(rn
, pi
);
2601 /* Mark if nexthop has changed. */
2602 if (!IPV4_ADDR_SAME(&pi
->attr
->nexthop
, &attr_new
->nexthop
))
2603 SET_FLAG(pi
->flags
, BGP_PATH_IGP_CHANGED
);
2605 /* Unintern existing, set to new. */
2606 bgp_attr_unintern(&pi
->attr
);
2607 pi
->attr
= attr_new
;
2608 pi
->uptime
= bgp_clock();
2611 /* Perform route selection and update zebra, if required. */
2612 ret
= evpn_es_route_select_install(bgp
, es
, rn
);
2617 * Install route entry into the VRF routing table and invoke route selection.
2619 static int install_evpn_route_entry_in_vrf(struct bgp
*bgp_vrf
,
2620 struct prefix_evpn
*evp
,
2621 struct bgp_path_info
*parent_pi
)
2623 struct bgp_node
*rn
;
2624 struct bgp_path_info
*pi
;
2626 struct attr
*attr_new
;
2629 struct prefix
*pp
= &p
;
2632 char buf
[PREFIX_STRLEN
];
2633 char buf1
[PREFIX_STRLEN
];
2635 memset(pp
, 0, sizeof(struct prefix
));
2636 ip_prefix_from_evpn_prefix(evp
, pp
);
2638 if (bgp_debug_zebra(NULL
)) {
2640 "import evpn prefix %s as ip prefix %s in vrf %s",
2641 prefix2str(evp
, buf
, sizeof(buf
)),
2642 prefix2str(pp
, buf1
, sizeof(buf
)),
2643 vrf_id_to_name(bgp_vrf
->vrf_id
));
2646 /* Create (or fetch) route within the VRF. */
2647 /* NOTE: There is no RD here. */
2648 if (is_evpn_prefix_ipaddr_v4(evp
)) {
2650 safi
= SAFI_UNICAST
;
2651 rn
= bgp_node_get(bgp_vrf
->rib
[afi
][safi
], pp
);
2652 } else if (is_evpn_prefix_ipaddr_v6(evp
)) {
2654 safi
= SAFI_UNICAST
;
2655 rn
= bgp_node_get(bgp_vrf
->rib
[afi
][safi
], pp
);
2659 /* EVPN routes currently only support a IPv4 next hop which corresponds
2660 * to the remote VTEP. When importing into a VRF, if it is IPv6 host
2661 * or prefix route, we have to convert the next hop to an IPv4-mapped
2662 * address for the rest of the code to flow through. In the case of IPv4,
2663 * make sure to set the flag for next hop attribute.
2665 attr
= *parent_pi
->attr
;
2667 evpn_convert_nexthop_to_ipv6(&attr
);
2669 attr
.flag
|= ATTR_FLAG_BIT(BGP_ATTR_NEXT_HOP
);
2671 /* Check if route entry is already present. */
2672 for (pi
= bgp_node_get_bgp_path_info(rn
); pi
; pi
= pi
->next
)
2674 && (struct bgp_path_info
*)pi
->extra
->parent
== parent_pi
)
2678 pi
= bgp_create_evpn_bgp_path_info(parent_pi
, rn
);
2680 if (attrhash_cmp(pi
->attr
, &attr
)
2681 && !CHECK_FLAG(pi
->flags
, BGP_PATH_REMOVED
)) {
2682 bgp_unlock_node(rn
);
2685 /* The attribute has changed. */
2686 /* Add (or update) attribute to hash. */
2687 attr_new
= bgp_attr_intern(&attr
);
2689 /* Restore route, if needed. */
2690 if (CHECK_FLAG(pi
->flags
, BGP_PATH_REMOVED
))
2691 bgp_path_info_restore(rn
, pi
);
2693 /* Mark if nexthop has changed. */
2695 && !IPV4_ADDR_SAME(&pi
->attr
->nexthop
, &attr_new
->nexthop
))
2697 && !IPV6_ADDR_SAME(&pi
->attr
->mp_nexthop_global
,
2698 &attr_new
->mp_nexthop_global
)))
2699 SET_FLAG(pi
->flags
, BGP_PATH_IGP_CHANGED
);
2701 bgp_path_info_set_flag(rn
, pi
, BGP_PATH_ATTR_CHANGED
);
2702 /* Unintern existing, set to new. */
2703 bgp_attr_unintern(&pi
->attr
);
2704 pi
->attr
= attr_new
;
2705 pi
->uptime
= bgp_clock();
2707 /* as it is an importation, change nexthop */
2708 bgp_path_info_set_flag(rn
, pi
, BGP_PATH_ANNC_NH_SELF
);
2710 bgp_aggregate_increment(bgp_vrf
, &rn
->p
, pi
, afi
, safi
);
2712 /* Perform route selection and update zebra, if required. */
2713 bgp_process(bgp_vrf
, rn
, afi
, safi
);
2715 /* Process for route leaking. */
2716 vpn_leak_from_vrf_update(bgp_get_default(), bgp_vrf
, pi
);
2718 bgp_unlock_node(rn
);
2724 * Install route entry into the VNI routing table and invoke route selection.
2726 static int install_evpn_route_entry(struct bgp
*bgp
, struct bgpevpn
*vpn
,
2727 struct prefix_evpn
*p
,
2728 struct bgp_path_info
*parent_pi
)
2730 struct bgp_node
*rn
;
2731 struct bgp_path_info
*pi
;
2732 struct attr
*attr_new
;
2735 /* Create (or fetch) route within the VNI. */
2736 /* NOTE: There is no RD here. */
2737 rn
= bgp_node_get(vpn
->route_table
, (struct prefix
*)p
);
2739 /* Check if route entry is already present. */
2740 for (pi
= bgp_node_get_bgp_path_info(rn
); pi
; pi
= pi
->next
)
2742 && (struct bgp_path_info
*)pi
->extra
->parent
== parent_pi
)
2746 /* Create an info */
2747 (void)bgp_create_evpn_bgp_path_info(parent_pi
, rn
);
2749 if (attrhash_cmp(pi
->attr
, parent_pi
->attr
)
2750 && !CHECK_FLAG(pi
->flags
, BGP_PATH_REMOVED
)) {
2751 bgp_unlock_node(rn
);
2754 /* The attribute has changed. */
2755 /* Add (or update) attribute to hash. */
2756 attr_new
= bgp_attr_intern(parent_pi
->attr
);
2758 /* Restore route, if needed. */
2759 if (CHECK_FLAG(pi
->flags
, BGP_PATH_REMOVED
))
2760 bgp_path_info_restore(rn
, pi
);
2762 /* Mark if nexthop has changed. */
2763 if (!IPV4_ADDR_SAME(&pi
->attr
->nexthop
, &attr_new
->nexthop
))
2764 SET_FLAG(pi
->flags
, BGP_PATH_IGP_CHANGED
);
2766 /* Unintern existing, set to new. */
2767 bgp_attr_unintern(&pi
->attr
);
2768 pi
->attr
= attr_new
;
2769 pi
->uptime
= bgp_clock();
2772 /* Perform route selection and update zebra, if required. */
2773 ret
= evpn_route_select_install(bgp
, vpn
, rn
);
2775 bgp_unlock_node(rn
);
2780 /* Uninstall EVPN route entry from ES route table */
2781 static int uninstall_evpn_route_entry_in_es(struct bgp
*bgp
, struct evpnes
*es
,
2782 struct prefix_evpn
*p
,
2783 struct bgp_path_info
*parent_pi
)
2786 struct bgp_node
*rn
;
2787 struct bgp_path_info
*pi
;
2789 if (!es
->route_table
)
2792 /* Locate route within the ESI.
2793 * NOTE: There is no RD here.
2795 rn
= bgp_node_lookup(es
->route_table
, (struct prefix
*)p
);
2799 /* Find matching route entry. */
2800 for (pi
= bgp_node_get_bgp_path_info(rn
); pi
; pi
= pi
->next
)
2802 && (struct bgp_path_info
*)pi
->extra
->parent
== parent_pi
)
2808 /* Mark entry for deletion */
2809 bgp_path_info_delete(rn
, pi
);
2811 /* Perform route selection and update zebra, if required. */
2812 ret
= evpn_es_route_select_install(bgp
, es
, rn
);
2814 /* Unlock route node. */
2815 bgp_unlock_node(rn
);
2821 * Uninstall route entry from the VRF routing table and send message
2822 * to zebra, if appropriate.
2824 static int uninstall_evpn_route_entry_in_vrf(struct bgp
*bgp_vrf
,
2825 struct prefix_evpn
*evp
,
2826 struct bgp_path_info
*parent_pi
)
2828 struct bgp_node
*rn
;
2829 struct bgp_path_info
*pi
;
2832 struct prefix
*pp
= &p
;
2835 char buf
[PREFIX_STRLEN
];
2836 char buf1
[PREFIX_STRLEN
];
2838 memset(pp
, 0, sizeof(struct prefix
));
2839 ip_prefix_from_evpn_prefix(evp
, pp
);
2841 if (bgp_debug_zebra(NULL
)) {
2843 "uninstalling evpn prefix %s as ip prefix %s in vrf %s",
2844 prefix2str(evp
, buf
, sizeof(buf
)),
2845 prefix2str(pp
, buf1
, sizeof(buf
)),
2846 vrf_id_to_name(bgp_vrf
->vrf_id
));
2849 /* Locate route within the VRF. */
2850 /* NOTE: There is no RD here. */
2851 if (is_evpn_prefix_ipaddr_v4(evp
)) {
2853 safi
= SAFI_UNICAST
;
2854 rn
= bgp_node_lookup(bgp_vrf
->rib
[afi
][safi
], pp
);
2857 safi
= SAFI_UNICAST
;
2858 rn
= bgp_node_lookup(bgp_vrf
->rib
[afi
][safi
], pp
);
2864 /* Find matching route entry. */
2865 for (pi
= bgp_node_get_bgp_path_info(rn
); pi
; pi
= pi
->next
)
2867 && (struct bgp_path_info
*)pi
->extra
->parent
== parent_pi
)
2873 /* Process for route leaking. */
2874 vpn_leak_from_vrf_withdraw(bgp_get_default(), bgp_vrf
, pi
);
2876 bgp_aggregate_decrement(bgp_vrf
, &rn
->p
, pi
, afi
, safi
);
2878 /* Mark entry for deletion */
2879 bgp_path_info_delete(rn
, pi
);
2881 /* Perform route selection and update zebra, if required. */
2882 bgp_process(bgp_vrf
, rn
, afi
, safi
);
2884 /* Unlock route node. */
2885 bgp_unlock_node(rn
);
2891 * Uninstall route entry from the VNI routing table and send message
2892 * to zebra, if appropriate.
2894 static int uninstall_evpn_route_entry(struct bgp
*bgp
, struct bgpevpn
*vpn
,
2895 struct prefix_evpn
*p
,
2896 struct bgp_path_info
*parent_pi
)
2898 struct bgp_node
*rn
;
2899 struct bgp_path_info
*pi
;
2902 /* Locate route within the VNI. */
2903 /* NOTE: There is no RD here. */
2904 rn
= bgp_node_lookup(vpn
->route_table
, (struct prefix
*)p
);
2908 /* Find matching route entry. */
2909 for (pi
= bgp_node_get_bgp_path_info(rn
); pi
; pi
= pi
->next
)
2911 && (struct bgp_path_info
*)pi
->extra
->parent
== parent_pi
)
2917 /* Mark entry for deletion */
2918 bgp_path_info_delete(rn
, pi
);
2920 /* Perform route selection and update zebra, if required. */
2921 ret
= evpn_route_select_install(bgp
, vpn
, rn
);
2923 /* Unlock route node. */
2924 bgp_unlock_node(rn
);
2930 * Given a prefix, see if it belongs to ES.
2932 static int is_prefix_matching_for_es(struct prefix_evpn
*p
,
2935 /* if not an ES route return false */
2936 if (p
->prefix
.route_type
!= BGP_EVPN_ES_ROUTE
)
2939 if (memcmp(&p
->prefix
.es_addr
.esi
, &es
->esi
, sizeof(esi_t
)) == 0)
2946 * Given a route entry and a VRF, see if this route entry should be
2947 * imported into the VRF i.e., RTs match.
2949 static int is_route_matching_for_vrf(struct bgp
*bgp_vrf
,
2950 struct bgp_path_info
*pi
)
2952 struct attr
*attr
= pi
->attr
;
2953 struct ecommunity
*ecom
;
2957 /* Route should have valid RT to be even considered. */
2958 if (!(attr
->flag
& ATTR_FLAG_BIT(BGP_ATTR_EXT_COMMUNITIES
)))
2961 ecom
= attr
->ecommunity
;
2962 if (!ecom
|| !ecom
->size
)
2965 /* For each extended community RT, see if it matches this VNI. If any RT
2966 * matches, we're done.
2968 for (i
= 0; i
< ecom
->size
; i
++) {
2970 uint8_t type
, sub_type
;
2971 struct ecommunity_val
*eval
;
2972 struct ecommunity_val eval_tmp
;
2973 struct vrf_irt_node
*irt
;
2975 /* Only deal with RTs */
2976 pnt
= (ecom
->val
+ (i
* ECOMMUNITY_SIZE
));
2977 eval
= (struct ecommunity_val
*)(ecom
->val
2978 + (i
* ECOMMUNITY_SIZE
));
2981 if (sub_type
!= ECOMMUNITY_ROUTE_TARGET
)
2984 /* See if this RT matches specified VNIs import RTs */
2985 irt
= lookup_vrf_import_rt(eval
);
2987 if (is_vrf_present_in_irt_vrfs(irt
->vrfs
, bgp_vrf
))
2990 /* Also check for non-exact match. In this, we mask out the AS
2992 * only check on the local-admin sub-field. This is to
2994 * VNI as the RT for EBGP peering too.
2997 if (type
== ECOMMUNITY_ENCODE_AS
2998 || type
== ECOMMUNITY_ENCODE_AS4
2999 || type
== ECOMMUNITY_ENCODE_IP
) {
3000 memcpy(&eval_tmp
, eval
, ECOMMUNITY_SIZE
);
3001 mask_ecom_global_admin(&eval_tmp
, eval
);
3002 irt
= lookup_vrf_import_rt(&eval_tmp
);
3005 if (is_vrf_present_in_irt_vrfs(irt
->vrfs
, bgp_vrf
))
3013 * Given a route entry and a VNI, see if this route entry should be
3014 * imported into the VNI i.e., RTs match.
3016 static int is_route_matching_for_vni(struct bgp
*bgp
, struct bgpevpn
*vpn
,
3017 struct bgp_path_info
*pi
)
3019 struct attr
*attr
= pi
->attr
;
3020 struct ecommunity
*ecom
;
3024 /* Route should have valid RT to be even considered. */
3025 if (!(attr
->flag
& ATTR_FLAG_BIT(BGP_ATTR_EXT_COMMUNITIES
)))
3028 ecom
= attr
->ecommunity
;
3029 if (!ecom
|| !ecom
->size
)
3032 /* For each extended community RT, see if it matches this VNI. If any RT
3033 * matches, we're done.
3035 for (i
= 0; i
< ecom
->size
; i
++) {
3037 uint8_t type
, sub_type
;
3038 struct ecommunity_val
*eval
;
3039 struct ecommunity_val eval_tmp
;
3040 struct irt_node
*irt
;
3042 /* Only deal with RTs */
3043 pnt
= (ecom
->val
+ (i
* ECOMMUNITY_SIZE
));
3044 eval
= (struct ecommunity_val
*)(ecom
->val
3045 + (i
* ECOMMUNITY_SIZE
));
3048 if (sub_type
!= ECOMMUNITY_ROUTE_TARGET
)
3051 /* See if this RT matches specified VNIs import RTs */
3052 irt
= lookup_import_rt(bgp
, eval
);
3054 if (is_vni_present_in_irt_vnis(irt
->vnis
, vpn
))
3057 /* Also check for non-exact match. In this, we mask out the AS
3059 * only check on the local-admin sub-field. This is to
3061 * VNI as the RT for EBGP peering too.
3064 if (type
== ECOMMUNITY_ENCODE_AS
3065 || type
== ECOMMUNITY_ENCODE_AS4
3066 || type
== ECOMMUNITY_ENCODE_IP
) {
3067 memcpy(&eval_tmp
, eval
, ECOMMUNITY_SIZE
);
3068 mask_ecom_global_admin(&eval_tmp
, eval
);
3069 irt
= lookup_import_rt(bgp
, &eval_tmp
);
3072 if (is_vni_present_in_irt_vnis(irt
->vnis
, vpn
))
3079 static int install_uninstall_routes_for_es(struct bgp
*bgp
,
3086 char buf
[PREFIX_STRLEN
];
3087 char buf1
[ESI_STR_LEN
];
3088 struct bgp_node
*rd_rn
, *rn
;
3089 struct bgp_table
*table
;
3090 struct bgp_path_info
*pi
;
3096 * Walk entire global routing table and evaluate routes which could be
3097 * imported into this VRF. Note that we need to loop through all global
3098 * routes to determine which route matches the import rt on vrf
3100 for (rd_rn
= bgp_table_top(bgp
->rib
[afi
][safi
]); rd_rn
;
3101 rd_rn
= bgp_route_next(rd_rn
)) {
3102 table
= bgp_node_get_bgp_table_info(rd_rn
);
3106 for (rn
= bgp_table_top(table
); rn
; rn
= bgp_route_next(rn
)) {
3107 struct prefix_evpn
*evp
= (struct prefix_evpn
*)&rn
->p
;
3109 for (pi
= bgp_node_get_bgp_path_info(rn
); pi
;
3112 * Consider "valid" remote routes applicable for
3115 if (!(CHECK_FLAG(pi
->flags
, BGP_PATH_VALID
)
3116 && pi
->type
== ZEBRA_ROUTE_BGP
3117 && pi
->sub_type
== BGP_ROUTE_NORMAL
))
3120 if (!is_prefix_matching_for_es(evp
, es
))
3124 ret
= install_evpn_route_entry_in_es(
3127 ret
= uninstall_evpn_route_entry_in_es(
3133 "Failed to %s EVPN %s route in ESI %s",
3136 prefix2str(evp
, buf
,
3138 esi_to_str(&es
->esi
, buf1
,
3148 /* This API will scan evpn routes for checking attribute's rmac
3149 * macthes with bgp instance router mac. It avoid installing
3150 * route into bgp vrf table and remote rmac in bridge table.
3152 static int bgp_evpn_route_rmac_self_check(struct bgp
*bgp_vrf
,
3153 struct prefix_evpn
*evp
,
3154 struct bgp_path_info
*pi
)
3156 /* evpn route could have learnt prior to L3vni has come up,
3157 * perform rmac check before installing route and
3158 * remote router mac.
3159 * The route will be removed from global bgp table once
3160 * SVI comes up with MAC and stored in hash, triggers
3161 * bgp_mac_rescan_all_evpn_tables.
3163 if (memcmp(&bgp_vrf
->rmac
, &pi
->attr
->rmac
, ETH_ALEN
) == 0) {
3164 if (bgp_debug_update(pi
->peer
, NULL
, NULL
, 1)) {
3165 char buf1
[PREFIX_STRLEN
];
3166 char attr_str
[BUFSIZ
] = {0};
3168 bgp_dump_attr(pi
->attr
, attr_str
, BUFSIZ
);
3170 zlog_debug("%s: bgp %u prefix %s with attr %s - DENIED due to self mac",
3171 __func__
, bgp_vrf
->vrf_id
,
3172 prefix2str(evp
, buf1
, sizeof(buf1
)),
3183 * Install or uninstall mac-ip routes are appropriate for this
3186 static int install_uninstall_routes_for_vrf(struct bgp
*bgp_vrf
, int install
)
3190 struct bgp_node
*rd_rn
, *rn
;
3191 struct bgp_table
*table
;
3192 struct bgp_path_info
*pi
;
3194 char buf
[PREFIX_STRLEN
];
3195 struct bgp
*bgp_evpn
= NULL
;
3199 bgp_evpn
= bgp_get_evpn();
3203 /* Walk entire global routing table and evaluate routes which could be
3204 * imported into this VRF. Note that we need to loop through all global
3205 * routes to determine which route matches the import rt on vrf
3207 for (rd_rn
= bgp_table_top(bgp_evpn
->rib
[afi
][safi
]); rd_rn
;
3208 rd_rn
= bgp_route_next(rd_rn
)) {
3209 table
= bgp_node_get_bgp_table_info(rd_rn
);
3213 for (rn
= bgp_table_top(table
); rn
; rn
= bgp_route_next(rn
)) {
3214 struct prefix_evpn
*evp
= (struct prefix_evpn
*)&rn
->p
;
3216 /* if not mac-ip route skip this route */
3217 if (!(evp
->prefix
.route_type
== BGP_EVPN_MAC_IP_ROUTE
3218 || evp
->prefix
.route_type
3219 == BGP_EVPN_IP_PREFIX_ROUTE
))
3222 /* if not a mac+ip route skip this route */
3223 if (!(is_evpn_prefix_ipaddr_v4(evp
)
3224 || is_evpn_prefix_ipaddr_v6(evp
)))
3227 for (pi
= bgp_node_get_bgp_path_info(rn
); pi
;
3229 /* Consider "valid" remote routes applicable for
3232 if (!(CHECK_FLAG(pi
->flags
, BGP_PATH_VALID
)
3233 && pi
->type
== ZEBRA_ROUTE_BGP
3234 && pi
->sub_type
== BGP_ROUTE_NORMAL
))
3237 if (is_route_matching_for_vrf(bgp_vrf
, pi
)) {
3238 if (bgp_evpn_route_rmac_self_check(
3243 ret
= install_evpn_route_entry_in_vrf(
3246 ret
= uninstall_evpn_route_entry_in_vrf(
3252 "Failed to %s EVPN %s route in VRF %s",
3255 prefix2str(evp
, buf
,
3270 * Install or uninstall routes of specified type that are appropriate for this
3273 static int install_uninstall_routes_for_vni(struct bgp
*bgp
,
3274 struct bgpevpn
*vpn
,
3275 bgp_evpn_route_type rtype
,
3280 struct bgp_node
*rd_rn
, *rn
;
3281 struct bgp_table
*table
;
3282 struct bgp_path_info
*pi
;
3288 /* Walk entire global routing table and evaluate routes which could be
3289 * imported into this VPN. Note that we cannot just look at the routes
3291 * the VNI's RD - remote routes applicable for this VNI could have any
3294 /* EVPN routes are a 2-level table. */
3295 for (rd_rn
= bgp_table_top(bgp
->rib
[afi
][safi
]); rd_rn
;
3296 rd_rn
= bgp_route_next(rd_rn
)) {
3297 table
= bgp_node_get_bgp_table_info(rd_rn
);
3301 for (rn
= bgp_table_top(table
); rn
; rn
= bgp_route_next(rn
)) {
3302 struct prefix_evpn
*evp
= (struct prefix_evpn
*)&rn
->p
;
3304 if (evp
->prefix
.route_type
!= rtype
)
3307 for (pi
= bgp_node_get_bgp_path_info(rn
); pi
;
3309 /* Consider "valid" remote routes applicable for
3311 if (!(CHECK_FLAG(pi
->flags
, BGP_PATH_VALID
)
3312 && pi
->type
== ZEBRA_ROUTE_BGP
3313 && pi
->sub_type
== BGP_ROUTE_NORMAL
))
3316 if (is_route_matching_for_vni(bgp
, vpn
, pi
)) {
3318 ret
= install_evpn_route_entry(
3321 ret
= uninstall_evpn_route_entry(
3327 "%u: Failed to %s EVPN %s route in VNI %u",
3331 rtype
== BGP_EVPN_MAC_IP_ROUTE
3345 /* Install any existing remote ES routes applicable for this ES into its routing
3346 * table. This is invoked when ES comes up.
3348 static int install_routes_for_es(struct bgp
*bgp
, struct evpnes
*es
)
3350 return install_uninstall_routes_for_es(bgp
, es
, 1);
3354 /* Install any existing remote routes applicable for this VRF into VRF RIB. This
3355 * is invoked upon l3vni-add or l3vni import rt change
3357 static int install_routes_for_vrf(struct bgp
*bgp_vrf
)
3359 install_uninstall_routes_for_vrf(bgp_vrf
, 1);
3364 * Install any existing remote routes applicable for this VNI into its
3365 * routing table. This is invoked when a VNI becomes "live" or its Import
3368 static int install_routes_for_vni(struct bgp
*bgp
, struct bgpevpn
*vpn
)
3372 /* Install type-3 routes followed by type-2 routes - the ones applicable
3375 ret
= install_uninstall_routes_for_vni(bgp
, vpn
, BGP_EVPN_IMET_ROUTE
,
3380 return install_uninstall_routes_for_vni(bgp
, vpn
, BGP_EVPN_MAC_IP_ROUTE
,
3384 /* uninstall routes from l3vni vrf. */
3385 static int uninstall_routes_for_vrf(struct bgp
*bgp_vrf
)
3387 install_uninstall_routes_for_vrf(bgp_vrf
, 0);
3392 * Uninstall any existing remote routes for this VNI. One scenario in which
3393 * this is invoked is upon an import RT change.
3395 static int uninstall_routes_for_vni(struct bgp
*bgp
, struct bgpevpn
*vpn
)
3399 /* Uninstall type-2 routes followed by type-3 routes - the ones
3403 ret
= install_uninstall_routes_for_vni(bgp
, vpn
, BGP_EVPN_MAC_IP_ROUTE
,
3408 return install_uninstall_routes_for_vni(bgp
, vpn
, BGP_EVPN_IMET_ROUTE
,
3412 /* Install or unistall route in ES */
3413 static int install_uninstall_route_in_es(struct bgp
*bgp
, struct evpnes
*es
,
3414 afi_t afi
, safi_t safi
,
3415 struct prefix_evpn
*evp
,
3416 struct bgp_path_info
*pi
, int install
)
3419 char buf
[ESI_STR_LEN
];
3422 ret
= install_evpn_route_entry_in_es(bgp
, es
, evp
, pi
);
3424 ret
= uninstall_evpn_route_entry_in_es(bgp
, es
, evp
, pi
);
3429 "%u: Failed to %s EVPN %s route in ESI %s", bgp
->vrf_id
,
3430 install
? "install" : "uninstall", "ES",
3431 esi_to_str(&evp
->prefix
.es_addr
.esi
, buf
, sizeof(buf
)));
3438 * Install or uninstall route in matching VRFs (list).
3440 static int install_uninstall_route_in_vrfs(struct bgp
*bgp_def
, afi_t afi
,
3441 safi_t safi
, struct prefix_evpn
*evp
,
3442 struct bgp_path_info
*pi
,
3443 struct list
*vrfs
, int install
)
3445 char buf
[PREFIX2STR_BUFFER
];
3446 struct bgp
*bgp_vrf
;
3447 struct listnode
*node
, *nnode
;
3449 /* Only type-2/type-5 routes go into a VRF */
3450 if (!(evp
->prefix
.route_type
== BGP_EVPN_MAC_IP_ROUTE
3451 || evp
->prefix
.route_type
== BGP_EVPN_IP_PREFIX_ROUTE
))
3454 /* if it is type-2 route and not a mac+ip route skip this route */
3455 if ((evp
->prefix
.route_type
== BGP_EVPN_MAC_IP_ROUTE
)
3456 && !(is_evpn_prefix_ipaddr_v4(evp
)
3457 || is_evpn_prefix_ipaddr_v6(evp
)))
3460 for (ALL_LIST_ELEMENTS(vrfs
, node
, nnode
, bgp_vrf
)) {
3464 ret
= install_evpn_route_entry_in_vrf(bgp_vrf
, evp
, pi
);
3466 ret
= uninstall_evpn_route_entry_in_vrf(bgp_vrf
, evp
,
3470 flog_err(EC_BGP_EVPN_FAIL
,
3471 "%u: Failed to %s prefix %s in VRF %s",
3473 install
? "install" : "uninstall",
3474 prefix2str(evp
, buf
, sizeof(buf
)),
3475 vrf_id_to_name(bgp_vrf
->vrf_id
));
3484 * Install or uninstall route in matching VNIs (list).
3486 static int install_uninstall_route_in_vnis(struct bgp
*bgp
, afi_t afi
,
3487 safi_t safi
, struct prefix_evpn
*evp
,
3488 struct bgp_path_info
*pi
,
3489 struct list
*vnis
, int install
)
3491 struct bgpevpn
*vpn
;
3492 struct listnode
*node
, *nnode
;
3494 for (ALL_LIST_ELEMENTS(vnis
, node
, nnode
, vpn
)) {
3497 if (!is_vni_live(vpn
))
3501 ret
= install_evpn_route_entry(bgp
, vpn
, evp
, pi
);
3503 ret
= uninstall_evpn_route_entry(bgp
, vpn
, evp
, pi
);
3506 flog_err(EC_BGP_EVPN_FAIL
,
3507 "%u: Failed to %s EVPN %s route in VNI %u",
3508 bgp
->vrf_id
, install
? "install" : "uninstall",
3509 evp
->prefix
.route_type
== BGP_EVPN_MAC_IP_ROUTE
3521 * Install or uninstall route for appropriate VNIs/ESIs.
3523 static int install_uninstall_evpn_route(struct bgp
*bgp
, afi_t afi
, safi_t safi
,
3525 struct bgp_path_info
*pi
, int import
)
3527 struct prefix_evpn
*evp
= (struct prefix_evpn
*)p
;
3528 struct attr
*attr
= pi
->attr
;
3529 struct ecommunity
*ecom
;
3534 /* Only type-2, type-3, type-4 and type-5 are supported currently */
3535 if (!(evp
->prefix
.route_type
== BGP_EVPN_MAC_IP_ROUTE
3536 || evp
->prefix
.route_type
== BGP_EVPN_IMET_ROUTE
3537 || evp
->prefix
.route_type
== BGP_EVPN_ES_ROUTE
3538 || evp
->prefix
.route_type
== BGP_EVPN_IP_PREFIX_ROUTE
))
3541 /* If we don't have Route Target, nothing much to do. */
3542 if (!(attr
->flag
& ATTR_FLAG_BIT(BGP_ATTR_EXT_COMMUNITIES
)))
3545 ecom
= attr
->ecommunity
;
3546 if (!ecom
|| !ecom
->size
)
3549 /* An EVPN route belongs to a VNI or a VRF or an ESI based on the RTs
3550 * attached to the route */
3551 for (i
= 0; i
< ecom
->size
; i
++) {
3553 uint8_t type
, sub_type
;
3554 struct ecommunity_val
*eval
;
3555 struct ecommunity_val eval_tmp
;
3556 struct irt_node
*irt
; /* import rt for l2vni */
3557 struct vrf_irt_node
*vrf_irt
; /* import rt for l3vni */
3560 /* Only deal with RTs */
3561 pnt
= (ecom
->val
+ (i
* ECOMMUNITY_SIZE
));
3562 eval
= (struct ecommunity_val
*)(ecom
->val
3563 + (i
* ECOMMUNITY_SIZE
));
3566 if (sub_type
!= ECOMMUNITY_ROUTE_TARGET
)
3570 * macip routes (type-2) are imported into VNI and VRF tables.
3571 * IMET route is imported into VNI table.
3572 * prefix routes are imported into VRF table.
3574 if (evp
->prefix
.route_type
== BGP_EVPN_MAC_IP_ROUTE
||
3575 evp
->prefix
.route_type
== BGP_EVPN_IMET_ROUTE
||
3576 evp
->prefix
.route_type
== BGP_EVPN_IP_PREFIX_ROUTE
) {
3578 irt
= lookup_import_rt(bgp
, eval
);
3580 install_uninstall_route_in_vnis(
3581 bgp
, afi
, safi
, evp
, pi
, irt
->vnis
,
3584 vrf_irt
= lookup_vrf_import_rt(eval
);
3586 install_uninstall_route_in_vrfs(
3587 bgp
, afi
, safi
, evp
, pi
, vrf_irt
->vrfs
,
3590 /* Also check for non-exact match.
3591 * In this, we mask out the AS and
3592 * only check on the local-admin sub-field.
3593 * This is to facilitate using
3594 * VNI as the RT for EBGP peering too.
3598 if (type
== ECOMMUNITY_ENCODE_AS
3599 || type
== ECOMMUNITY_ENCODE_AS4
3600 || type
== ECOMMUNITY_ENCODE_IP
) {
3601 memcpy(&eval_tmp
, eval
, ECOMMUNITY_SIZE
);
3602 mask_ecom_global_admin(&eval_tmp
, eval
);
3603 irt
= lookup_import_rt(bgp
, &eval_tmp
);
3604 vrf_irt
= lookup_vrf_import_rt(&eval_tmp
);
3608 install_uninstall_route_in_vnis(
3609 bgp
, afi
, safi
, evp
, pi
, irt
->vnis
,
3612 install_uninstall_route_in_vrfs(
3613 bgp
, afi
, safi
, evp
, pi
, vrf_irt
->vrfs
,
3617 /* es route is imported into the es table */
3618 if (evp
->prefix
.route_type
== BGP_EVPN_ES_ROUTE
) {
3620 /* we will match based on the entire esi to avoid
3621 * imoort of an es route for esi2 into esi1
3623 es
= bgp_evpn_lookup_es(bgp
, &evp
->prefix
.es_addr
.esi
);
3624 if (es
&& is_es_local(es
))
3625 install_uninstall_route_in_es(
3626 bgp
, es
, afi
, safi
, evp
, pi
, import
);
3634 * delete and withdraw all ipv4 and ipv6 routes in the vrf table as type-5
3637 static void delete_withdraw_vrf_routes(struct bgp
*bgp_vrf
)
3639 /* Delete ipv4 default route and withdraw from peers */
3640 if (evpn_default_originate_set(bgp_vrf
, AFI_IP
, SAFI_UNICAST
))
3641 bgp_evpn_install_uninstall_default_route(bgp_vrf
, AFI_IP
,
3642 SAFI_UNICAST
, false);
3644 /* delete all ipv4 routes and withdraw from peers */
3645 if (advertise_type5_routes(bgp_vrf
, AFI_IP
))
3646 bgp_evpn_withdraw_type5_routes(bgp_vrf
, AFI_IP
, SAFI_UNICAST
);
3648 /* Delete ipv6 default route and withdraw from peers */
3649 if (evpn_default_originate_set(bgp_vrf
, AFI_IP6
, SAFI_UNICAST
))
3650 bgp_evpn_install_uninstall_default_route(bgp_vrf
, AFI_IP6
,
3651 SAFI_UNICAST
, false);
3653 /* delete all ipv6 routes and withdraw from peers */
3654 if (advertise_type5_routes(bgp_vrf
, AFI_IP6
))
3655 bgp_evpn_withdraw_type5_routes(bgp_vrf
, AFI_IP6
, SAFI_UNICAST
);
3659 * update and advertise all ipv4 and ipv6 routes in thr vrf table as type-5
3662 void update_advertise_vrf_routes(struct bgp
*bgp_vrf
)
3664 struct bgp
*bgp_evpn
= NULL
; /* EVPN bgp instance */
3666 bgp_evpn
= bgp_get_evpn();
3670 /* update all ipv4 routes */
3671 if (advertise_type5_routes(bgp_vrf
, AFI_IP
))
3672 bgp_evpn_advertise_type5_routes(bgp_vrf
, AFI_IP
, SAFI_UNICAST
);
3674 /* update ipv4 default route and withdraw from peers */
3675 if (evpn_default_originate_set(bgp_vrf
, AFI_IP
, SAFI_UNICAST
))
3676 bgp_evpn_install_uninstall_default_route(bgp_vrf
, AFI_IP
,
3677 SAFI_UNICAST
, true);
3679 /* update all ipv6 routes */
3680 if (advertise_type5_routes(bgp_vrf
, AFI_IP6
))
3681 bgp_evpn_advertise_type5_routes(bgp_vrf
, AFI_IP6
, SAFI_UNICAST
);
3683 /* update ipv6 default route and withdraw from peers */
3684 if (evpn_default_originate_set(bgp_vrf
, AFI_IP6
, SAFI_UNICAST
))
3685 bgp_evpn_install_uninstall_default_route(bgp_vrf
, AFI_IP6
,
3686 SAFI_UNICAST
, true);
3691 * update and advertise local routes for a VRF as type-5 routes.
3692 * This is invoked upon RD change for a VRF. Note taht the processing is only
3693 * done in the global route table using the routes which already exist in the
3696 static void update_router_id_vrf(struct bgp
*bgp_vrf
)
3698 /* skip if the RD is configured */
3699 if (is_vrf_rd_configured(bgp_vrf
))
3702 /* derive the RD for the VRF based on new router-id */
3703 bgp_evpn_derive_auto_rd_for_vrf(bgp_vrf
);
3705 /* update advertise ipv4|ipv6 routes as type-5 routes */
3706 update_advertise_vrf_routes(bgp_vrf
);
3710 * Delete and withdraw all type-5 routes for the RD corresponding to VRF.
3711 * This is invoked upon VRF RD change. The processing is done only from global
3714 static void withdraw_router_id_vrf(struct bgp
*bgp_vrf
)
3716 /* skip if the RD is configured */
3717 if (is_vrf_rd_configured(bgp_vrf
))
3720 /* delete/withdraw ipv4|ipv6 routes as type-5 routes */
3721 delete_withdraw_vrf_routes(bgp_vrf
);
3725 * Update and advertise local routes for a VNI. Invoked upon router-id
3726 * change. Note that the processing is done only on the global route table
3727 * using routes that already exist in the per-VNI table.
3729 static int update_advertise_vni_routes(struct bgp
*bgp
, struct bgpevpn
*vpn
)
3731 struct prefix_evpn p
;
3732 struct bgp_node
*rn
, *global_rn
;
3733 struct bgp_path_info
*pi
, *global_pi
;
3735 afi_t afi
= AFI_L2VPN
;
3736 safi_t safi
= SAFI_EVPN
;
3738 /* Locate type-3 route for VNI in the per-VNI table and use its
3739 * attributes to create and advertise the type-3 route for this VNI
3740 * in the global table.
3742 * RT-3 only if doing head-end replication
3744 if (bgp_evpn_vni_flood_mode_get(bgp
, vpn
)
3745 == VXLAN_FLOOD_HEAD_END_REPL
) {
3746 build_evpn_type3_prefix(&p
, vpn
->originator_ip
);
3747 rn
= bgp_node_lookup(vpn
->route_table
, (struct prefix
*)&p
);
3748 if (!rn
) /* unexpected */
3750 for (pi
= bgp_node_get_bgp_path_info(rn
); pi
; pi
= pi
->next
)
3751 if (pi
->peer
== bgp
->peer_self
&&
3752 pi
->type
== ZEBRA_ROUTE_BGP
3753 && pi
->sub_type
== BGP_ROUTE_STATIC
)
3755 if (!pi
) /* unexpected */
3759 global_rn
= bgp_afi_node_get(bgp
->rib
[afi
][safi
], afi
, safi
,
3760 (struct prefix
*)&p
, &vpn
->prd
);
3761 update_evpn_route_entry(bgp
, vpn
, afi
, safi
, global_rn
, attr
,
3762 1, &pi
, 0, mac_mobility_seqnum(attr
));
3764 /* Schedule for processing and unlock node. */
3765 bgp_process(bgp
, global_rn
, afi
, safi
);
3766 bgp_unlock_node(global_rn
);
3769 /* Now, walk this VNI's route table and use the route and its attribute
3770 * to create and schedule route in global table.
3772 for (rn
= bgp_table_top(vpn
->route_table
); rn
;
3773 rn
= bgp_route_next(rn
)) {
3774 struct prefix_evpn
*evp
= (struct prefix_evpn
*)&rn
->p
;
3776 /* Identify MAC-IP local routes. */
3777 if (evp
->prefix
.route_type
!= BGP_EVPN_MAC_IP_ROUTE
)
3780 for (pi
= bgp_node_get_bgp_path_info(rn
); pi
; pi
= pi
->next
)
3781 if (pi
->peer
== bgp
->peer_self
3782 && pi
->type
== ZEBRA_ROUTE_BGP
3783 && pi
->sub_type
== BGP_ROUTE_STATIC
)
3788 /* Create route in global routing table using this route entry's
3792 global_rn
= bgp_afi_node_get(bgp
->rib
[afi
][safi
], afi
, safi
,
3793 (struct prefix
*)evp
, &vpn
->prd
);
3795 update_evpn_route_entry(bgp
, vpn
, afi
, safi
, global_rn
, attr
, 1,
3797 mac_mobility_seqnum(attr
));
3799 /* Schedule for processing and unlock node. */
3800 bgp_process(bgp
, global_rn
, afi
, safi
);
3801 bgp_unlock_node(global_rn
);
3808 * Delete (and withdraw) local routes for a VNI - only from the global
3809 * table. Invoked upon router-id change.
3811 static int delete_withdraw_vni_routes(struct bgp
*bgp
, struct bgpevpn
*vpn
)
3814 struct prefix_evpn p
;
3815 struct bgp_node
*global_rn
;
3816 struct bgp_path_info
*pi
;
3817 afi_t afi
= AFI_L2VPN
;
3818 safi_t safi
= SAFI_EVPN
;
3820 /* Delete and withdraw locally learnt type-2 routes (MACIP)
3821 * for this VNI - from the global table.
3823 ret
= delete_global_type2_routes(bgp
, vpn
);
3827 /* Remove type-3 route for this VNI from global table. */
3828 build_evpn_type3_prefix(&p
, vpn
->originator_ip
);
3829 global_rn
= bgp_afi_node_lookup(bgp
->rib
[afi
][safi
], afi
, safi
,
3830 (struct prefix
*)&p
, &vpn
->prd
);
3832 /* Delete route entry in the global EVPN table. */
3833 delete_evpn_route_entry(bgp
, afi
, safi
, global_rn
, &pi
);
3835 /* Schedule for processing - withdraws to peers happen from
3839 bgp_process(bgp
, global_rn
, afi
, safi
);
3840 bgp_unlock_node(global_rn
);
3847 * Handle router-id change. Update and advertise local routes corresponding
3848 * to this VNI from peers. Note that this is invoked after updating the
3849 * router-id. The routes in the per-VNI table are used to create routes in
3850 * the global table and schedule them.
3852 static void update_router_id_vni(struct hash_bucket
*bucket
, struct bgp
*bgp
)
3854 struct bgpevpn
*vpn
= (struct bgpevpn
*)bucket
->data
;
3856 /* Skip VNIs with configured RD. */
3857 if (is_rd_configured(vpn
))
3860 bgp_evpn_derive_auto_rd(bgp
, vpn
);
3861 update_advertise_vni_routes(bgp
, vpn
);
3865 * Handle router-id change. Delete and withdraw local routes corresponding
3866 * to this VNI from peers. Note that this is invoked prior to updating
3867 * the router-id and is done only on the global route table, the routes
3868 * are needed in the per-VNI table to re-advertise with new router id.
3870 static void withdraw_router_id_vni(struct hash_bucket
*bucket
, struct bgp
*bgp
)
3872 struct bgpevpn
*vpn
= (struct bgpevpn
*)bucket
->data
;
3874 /* Skip VNIs with configured RD. */
3875 if (is_rd_configured(vpn
))
3878 delete_withdraw_vni_routes(bgp
, vpn
);
3882 * Create RT-3 for a VNI and schedule for processing and advertisement.
3883 * This is invoked upon flooding mode changing to head-end replication.
3885 static void create_advertise_type3(struct hash_bucket
*bucket
, void *data
)
3887 struct bgpevpn
*vpn
= bucket
->data
;
3888 struct bgp
*bgp
= data
;
3889 struct prefix_evpn p
;
3891 if (!vpn
|| !is_vni_live(vpn
) ||
3892 bgp_evpn_vni_flood_mode_get(bgp
, vpn
)
3893 != VXLAN_FLOOD_HEAD_END_REPL
)
3896 build_evpn_type3_prefix(&p
, vpn
->originator_ip
);
3897 if (update_evpn_route(bgp
, vpn
, &p
, 0, 0))
3898 flog_err(EC_BGP_EVPN_ROUTE_CREATE
,
3899 "Type3 route creation failure for VNI %u", vpn
->vni
);
3903 * Delete RT-3 for a VNI and schedule for processing and withdrawal.
3904 * This is invoked upon flooding mode changing to drop BUM packets.
3906 static void delete_withdraw_type3(struct hash_bucket
*bucket
, void *data
)
3908 struct bgpevpn
*vpn
= bucket
->data
;
3909 struct bgp
*bgp
= data
;
3910 struct prefix_evpn p
;
3912 if (!vpn
|| !is_vni_live(vpn
))
3915 build_evpn_type3_prefix(&p
, vpn
->originator_ip
);
3916 delete_evpn_route(bgp
, vpn
, &p
);
3920 * Process received EVPN type-2 route (advertise or withdraw).
3922 static int process_type2_route(struct peer
*peer
, afi_t afi
, safi_t safi
,
3923 struct attr
*attr
, uint8_t *pfx
, int psize
,
3924 uint32_t addpath_id
)
3926 struct prefix_rd prd
;
3927 struct prefix_evpn p
;
3928 struct bgp_route_evpn evpn
;
3930 uint8_t macaddr_len
;
3931 mpls_label_t label
[BGP_MAX_LABELS
]; /* holds the VNI(s) as in packet */
3932 uint32_t num_labels
= 0;
3936 /* Type-2 route should be either 33, 37 or 49 bytes or an
3937 * additional 3 bytes if there is a second label (VNI):
3938 * RD (8), ESI (10), Eth Tag (4), MAC Addr Len (1),
3939 * MAC Addr (6), IP len (1), IP (0, 4 or 16),
3940 * MPLS Lbl1 (3), MPLS Lbl2 (0 or 3)
3942 if (psize
!= 33 && psize
!= 37 && psize
!= 49 && psize
!= 36
3943 && psize
!= 40 && psize
!= 52) {
3944 flog_err(EC_BGP_EVPN_ROUTE_INVALID
,
3945 "%u:%s - Rx EVPN Type-2 NLRI with invalid length %d",
3946 peer
->bgp
->vrf_id
, peer
->host
, psize
);
3950 memset(&evpn
, 0, sizeof(evpn
));
3952 /* Make prefix_rd */
3953 prd
.family
= AF_UNSPEC
;
3955 memcpy(&prd
.val
, pfx
, 8);
3958 /* Make EVPN prefix. */
3959 memset(&p
, 0, sizeof(struct prefix_evpn
));
3961 p
.prefixlen
= EVPN_ROUTE_PREFIXLEN
;
3962 p
.prefix
.route_type
= BGP_EVPN_MAC_IP_ROUTE
;
3964 /* Copy Ethernet Seg Identifier */
3965 memcpy(&evpn
.eth_s_id
.val
, pfx
, ESI_LEN
);
3968 /* Copy Ethernet Tag */
3969 memcpy(ð_tag
, pfx
, 4);
3970 p
.prefix
.macip_addr
.eth_tag
= ntohl(eth_tag
);
3973 /* Get the MAC Addr len */
3974 macaddr_len
= *pfx
++;
3976 /* Get the MAC Addr */
3977 if (macaddr_len
== (ETH_ALEN
* 8)) {
3978 memcpy(&p
.prefix
.macip_addr
.mac
.octet
, pfx
, ETH_ALEN
);
3982 EC_BGP_EVPN_ROUTE_INVALID
,
3983 "%u:%s - Rx EVPN Type-2 NLRI with unsupported MAC address length %d",
3984 peer
->bgp
->vrf_id
, peer
->host
, macaddr_len
);
3990 ipaddr_len
= *pfx
++;
3991 if (ipaddr_len
!= 0 && ipaddr_len
!= IPV4_MAX_BITLEN
3992 && ipaddr_len
!= IPV6_MAX_BITLEN
) {
3994 EC_BGP_EVPN_ROUTE_INVALID
,
3995 "%u:%s - Rx EVPN Type-2 NLRI with unsupported IP address length %d",
3996 peer
->bgp
->vrf_id
, peer
->host
, ipaddr_len
);
4001 ipaddr_len
/= 8; /* Convert to bytes. */
4002 p
.prefix
.macip_addr
.ip
.ipa_type
= (ipaddr_len
== IPV4_MAX_BYTELEN
)
4005 memcpy(&p
.prefix
.macip_addr
.ip
.ip
.addr
, pfx
, ipaddr_len
);
4009 /* Get the VNI(s). Stored as bytes here. */
4011 memset(label
, 0, sizeof(label
));
4012 memcpy(&label
[0], pfx
, BGP_LABEL_BYTES
);
4013 pfx
+= BGP_LABEL_BYTES
;
4014 psize
-= (33 + ipaddr_len
);
4015 /* Do we have a second VNI? */
4018 memcpy(&label
[1], pfx
, BGP_LABEL_BYTES
);
4020 * If in future, we are required to access additional fields,
4021 * we MUST increment pfx by BGP_LABEL_BYTES in before reading
4026 /* Process the route. */
4028 ret
= bgp_update(peer
, (struct prefix
*)&p
, addpath_id
, attr
,
4029 afi
, safi
, ZEBRA_ROUTE_BGP
, BGP_ROUTE_NORMAL
,
4030 &prd
, &label
[0], num_labels
, 0, &evpn
);
4032 ret
= bgp_withdraw(peer
, (struct prefix
*)&p
, addpath_id
, attr
,
4033 afi
, safi
, ZEBRA_ROUTE_BGP
, BGP_ROUTE_NORMAL
,
4034 &prd
, &label
[0], num_labels
, &evpn
);
4039 * Process received EVPN type-3 route (advertise or withdraw).
4041 static int process_type3_route(struct peer
*peer
, afi_t afi
, safi_t safi
,
4042 struct attr
*attr
, uint8_t *pfx
, int psize
,
4043 uint32_t addpath_id
)
4045 struct prefix_rd prd
;
4046 struct prefix_evpn p
;
4051 /* Type-3 route should be either 17 or 29 bytes: RD (8), Eth Tag (4),
4052 * IP len (1) and IP (4 or 16).
4054 if (psize
!= 17 && psize
!= 29) {
4055 flog_err(EC_BGP_EVPN_ROUTE_INVALID
,
4056 "%u:%s - Rx EVPN Type-3 NLRI with invalid length %d",
4057 peer
->bgp
->vrf_id
, peer
->host
, psize
);
4061 /* If PMSI is present, log if it is anything other than IR.
4062 * Note: We just simply ignore the values as it is not clear if
4063 * doing anything else is better.
4066 (attr
->flag
& ATTR_FLAG_BIT(BGP_ATTR_PMSI_TUNNEL
))) {
4067 if (attr
->pmsi_tnl_type
!= PMSI_TNLTYPE_INGR_REPL
&&
4068 attr
->pmsi_tnl_type
!= PMSI_TNLTYPE_PIM_SM
) {
4069 flog_warn(EC_BGP_EVPN_PMSI_PRESENT
,
4070 "%u:%s - Rx EVPN Type-3 NLRI with unsupported PTA %d",
4071 peer
->bgp
->vrf_id
, peer
->host
,
4072 attr
->pmsi_tnl_type
);
4076 /* Make prefix_rd */
4077 prd
.family
= AF_UNSPEC
;
4079 memcpy(&prd
.val
, pfx
, 8);
4082 /* Make EVPN prefix. */
4083 memset(&p
, 0, sizeof(struct prefix_evpn
));
4085 p
.prefixlen
= EVPN_ROUTE_PREFIXLEN
;
4086 p
.prefix
.route_type
= BGP_EVPN_IMET_ROUTE
;
4088 /* Copy Ethernet Tag */
4089 memcpy(ð_tag
, pfx
, 4);
4090 p
.prefix
.imet_addr
.eth_tag
= ntohl(eth_tag
);
4094 ipaddr_len
= *pfx
++;
4095 if (ipaddr_len
== IPV4_MAX_BITLEN
) {
4096 p
.prefix
.imet_addr
.ip
.ipa_type
= IPADDR_V4
;
4097 memcpy(&p
.prefix
.imet_addr
.ip
.ip
.addr
, pfx
, IPV4_MAX_BYTELEN
);
4100 EC_BGP_EVPN_ROUTE_INVALID
,
4101 "%u:%s - Rx EVPN Type-3 NLRI with unsupported IP address length %d",
4102 peer
->bgp
->vrf_id
, peer
->host
, ipaddr_len
);
4106 /* Process the route. */
4108 ret
= bgp_update(peer
, (struct prefix
*)&p
, addpath_id
, attr
,
4109 afi
, safi
, ZEBRA_ROUTE_BGP
, BGP_ROUTE_NORMAL
,
4110 &prd
, NULL
, 0, 0, NULL
);
4112 ret
= bgp_withdraw(peer
, (struct prefix
*)&p
, addpath_id
, attr
,
4113 afi
, safi
, ZEBRA_ROUTE_BGP
, BGP_ROUTE_NORMAL
,
4114 &prd
, NULL
, 0, NULL
);
4119 * Process received EVPN type-4 route (advertise or withdraw).
4121 static int process_type4_route(struct peer
*peer
, afi_t afi
, safi_t safi
,
4122 struct attr
*attr
, uint8_t *pfx
, int psize
,
4123 uint32_t addpath_id
)
4128 struct in_addr vtep_ip
;
4129 struct prefix_rd prd
;
4130 struct prefix_evpn p
;
4132 /* Type-4 route should be either 23 or 35 bytes
4133 * RD (8), ESI (10), ip-len (1), ip (4 or 16)
4135 if (psize
!= 23 && psize
!= 35) {
4136 flog_err(EC_BGP_EVPN_ROUTE_INVALID
,
4137 "%u:%s - Rx EVPN Type-4 NLRI with invalid length %d",
4138 peer
->bgp
->vrf_id
, peer
->host
, psize
);
4142 /* Make prefix_rd */
4143 prd
.family
= AF_UNSPEC
;
4145 memcpy(&prd
.val
, pfx
, 8);
4149 memcpy(&esi
, pfx
, ESI_BYTES
);
4154 ipaddr_len
= *pfx
++;
4155 if (ipaddr_len
== IPV4_MAX_BITLEN
) {
4156 memcpy(&vtep_ip
, pfx
, IPV4_MAX_BYTELEN
);
4159 EC_BGP_EVPN_ROUTE_INVALID
,
4160 "%u:%s - Rx EVPN Type-4 NLRI with unsupported IP address length %d",
4161 peer
->bgp
->vrf_id
, peer
->host
, ipaddr_len
);
4165 build_evpn_type4_prefix(&p
, &esi
, vtep_ip
);
4166 /* Process the route. */
4168 ret
= bgp_update(peer
, (struct prefix
*)&p
, addpath_id
, attr
,
4169 afi
, safi
, ZEBRA_ROUTE_BGP
, BGP_ROUTE_NORMAL
,
4170 &prd
, NULL
, 0, 0, NULL
);
4172 ret
= bgp_withdraw(peer
, (struct prefix
*)&p
, addpath_id
, attr
,
4173 afi
, safi
, ZEBRA_ROUTE_BGP
, BGP_ROUTE_NORMAL
,
4174 &prd
, NULL
, 0, NULL
);
4181 * Process received EVPN type-5 route (advertise or withdraw).
4183 static int process_type5_route(struct peer
*peer
, afi_t afi
, safi_t safi
,
4184 struct attr
*attr
, uint8_t *pfx
, int psize
,
4185 uint32_t addpath_id
)
4187 struct prefix_rd prd
;
4188 struct prefix_evpn p
;
4189 struct bgp_route_evpn evpn
;
4192 mpls_label_t label
; /* holds the VNI as in the packet */
4195 /* Type-5 route should be 34 or 58 bytes:
4196 * RD (8), ESI (10), Eth Tag (4), IP len (1), IP (4 or 16),
4197 * GW (4 or 16) and VNI (3).
4198 * Note that the IP and GW should both be IPv4 or both IPv6.
4200 if (psize
!= 34 && psize
!= 58) {
4201 flog_err(EC_BGP_EVPN_ROUTE_INVALID
,
4202 "%u:%s - Rx EVPN Type-5 NLRI with invalid length %d",
4203 peer
->bgp
->vrf_id
, peer
->host
, psize
);
4207 /* Make prefix_rd */
4208 prd
.family
= AF_UNSPEC
;
4210 memcpy(&prd
.val
, pfx
, 8);
4213 /* Make EVPN prefix. */
4214 memset(&p
, 0, sizeof(struct prefix_evpn
));
4216 p
.prefixlen
= EVPN_ROUTE_PREFIXLEN
;
4217 p
.prefix
.route_type
= BGP_EVPN_IP_PREFIX_ROUTE
;
4219 /* Additional information outside of prefix - ESI and GW IP */
4220 memset(&evpn
, 0, sizeof(evpn
));
4223 memcpy(&evpn
.eth_s_id
.val
, pfx
, 10);
4226 /* Fetch Ethernet Tag. */
4227 memcpy(ð_tag
, pfx
, 4);
4228 p
.prefix
.prefix_addr
.eth_tag
= ntohl(eth_tag
);
4231 /* Fetch IP prefix length. */
4233 if (ippfx_len
> IPV6_MAX_BITLEN
) {
4235 EC_BGP_EVPN_ROUTE_INVALID
,
4236 "%u:%s - Rx EVPN Type-5 NLRI with invalid IP Prefix length %d",
4237 peer
->bgp
->vrf_id
, peer
->host
, ippfx_len
);
4240 p
.prefix
.prefix_addr
.ip_prefix_length
= ippfx_len
;
4242 /* Determine IPv4 or IPv6 prefix */
4243 /* Since the address and GW are from the same family, this just becomes
4244 * a simple check on the total size.
4247 SET_IPADDR_V4(&p
.prefix
.prefix_addr
.ip
);
4248 memcpy(&p
.prefix
.prefix_addr
.ip
.ipaddr_v4
, pfx
, 4);
4250 memcpy(&evpn
.gw_ip
.ipv4
, pfx
, 4);
4253 SET_IPADDR_V6(&p
.prefix
.prefix_addr
.ip
);
4254 memcpy(&p
.prefix
.prefix_addr
.ip
.ipaddr_v6
, pfx
, 16);
4256 memcpy(&evpn
.gw_ip
.ipv6
, pfx
, 16);
4260 /* Get the VNI (in MPLS label field). Stored as bytes here. */
4261 memset(&label
, 0, sizeof(label
));
4262 memcpy(&label
, pfx
, BGP_LABEL_BYTES
);
4265 * If in future, we are required to access additional fields,
4266 * we MUST increment pfx by BGP_LABEL_BYTES in before reading the next
4270 /* Process the route. */
4272 ret
= bgp_update(peer
, (struct prefix
*)&p
, addpath_id
, attr
,
4273 afi
, safi
, ZEBRA_ROUTE_BGP
, BGP_ROUTE_NORMAL
,
4274 &prd
, &label
, 1, 0, &evpn
);
4276 ret
= bgp_withdraw(peer
, (struct prefix
*)&p
, addpath_id
, attr
,
4277 afi
, safi
, ZEBRA_ROUTE_BGP
, BGP_ROUTE_NORMAL
,
4278 &prd
, &label
, 1, &evpn
);
4283 static void evpn_mpattr_encode_type5(struct stream
*s
, struct prefix
*p
,
4284 struct prefix_rd
*prd
, mpls_label_t
*label
,
4285 uint32_t num_labels
, struct attr
*attr
)
4289 struct evpn_addr
*p_evpn_p
;
4291 memset(&temp
, 0, 16);
4292 if (p
->family
!= AF_EVPN
)
4294 p_evpn_p
= &(p
->u
.prefix_evpn
);
4296 /* len denites the total len of IP and GW-IP in the route
4297 IP and GW-IP have to be both ipv4 or ipv6
4299 if (IS_IPADDR_V4(&p_evpn_p
->prefix_addr
.ip
))
4300 len
= 8; /* IP and GWIP are both ipv4 */
4302 len
= 32; /* IP and GWIP are both ipv6 */
4303 /* Prefix contains RD, ESI, EthTag, IP length, IP, GWIP and VNI */
4304 stream_putc(s
, 8 + 10 + 4 + 1 + len
+ 3);
4305 stream_put(s
, prd
->val
, 8);
4307 stream_put(s
, &(attr
->evpn_overlay
.eth_s_id
), 10);
4309 stream_put(s
, &temp
, 10);
4310 stream_putl(s
, p_evpn_p
->prefix_addr
.eth_tag
);
4311 stream_putc(s
, p_evpn_p
->prefix_addr
.ip_prefix_length
);
4312 if (IS_IPADDR_V4(&p_evpn_p
->prefix_addr
.ip
))
4313 stream_put_ipv4(s
, p_evpn_p
->prefix_addr
.ip
.ipaddr_v4
.s_addr
);
4315 stream_put(s
, &p_evpn_p
->prefix_addr
.ip
.ipaddr_v6
, 16);
4317 if (IS_IPADDR_V4(&p_evpn_p
->prefix_addr
.ip
))
4319 attr
->evpn_overlay
.gw_ip
.ipv4
.s_addr
);
4321 stream_put(s
, &(attr
->evpn_overlay
.gw_ip
.ipv6
), 16);
4323 if (IS_IPADDR_V4(&p_evpn_p
->prefix_addr
.ip
))
4324 stream_put_ipv4(s
, 0);
4326 stream_put(s
, &temp
, 16);
4330 stream_put(s
, label
, 3);
4336 * Cleanup specific VNI upon EVPN (advertise-all-vni) being disabled.
4338 static void cleanup_vni_on_disable(struct hash_bucket
*bucket
, struct bgp
*bgp
)
4340 struct bgpevpn
*vpn
= (struct bgpevpn
*)bucket
->data
;
4342 /* Remove EVPN routes and schedule for processing. */
4343 delete_routes_for_vni(bgp
, vpn
);
4345 /* Clear "live" flag and see if hash needs to be freed. */
4346 UNSET_FLAG(vpn
->flags
, VNI_FLAG_LIVE
);
4347 if (!is_vni_configured(vpn
))
4348 bgp_evpn_free(bgp
, vpn
);
4352 * Free a VNI entry; iterator function called during cleanup.
4354 static void free_vni_entry(struct hash_bucket
*bucket
, struct bgp
*bgp
)
4356 struct bgpevpn
*vpn
= (struct bgpevpn
*)bucket
->data
;
4358 delete_all_vni_routes(bgp
, vpn
);
4359 bgp_evpn_free(bgp
, vpn
);
4363 * Derive AUTO import RT for BGP VRF - L3VNI
4365 static void evpn_auto_rt_import_add_for_vrf(struct bgp
*bgp_vrf
)
4367 struct bgp
*bgp_evpn
= NULL
;
4369 form_auto_rt(bgp_vrf
, bgp_vrf
->l3vni
, bgp_vrf
->vrf_import_rtl
);
4370 UNSET_FLAG(bgp_vrf
->vrf_flags
, BGP_VRF_IMPORT_RT_CFGD
);
4373 bgp_evpn
= bgp_get_evpn();
4376 bgp_evpn_map_vrf_to_its_rts(bgp_vrf
);
4380 * Delete AUTO import RT from BGP VRF - L3VNI
4382 static void evpn_auto_rt_import_delete_for_vrf(struct bgp
*bgp_vrf
)
4384 evpn_rt_delete_auto(bgp_vrf
, bgp_vrf
->l3vni
, bgp_vrf
->vrf_import_rtl
);
4388 * Derive AUTO export RT for BGP VRF - L3VNI
4390 static void evpn_auto_rt_export_add_for_vrf(struct bgp
*bgp_vrf
)
4392 UNSET_FLAG(bgp_vrf
->vrf_flags
, BGP_VRF_EXPORT_RT_CFGD
);
4393 form_auto_rt(bgp_vrf
, bgp_vrf
->l3vni
, bgp_vrf
->vrf_export_rtl
);
4397 * Delete AUTO export RT from BGP VRF - L3VNI
4399 static void evpn_auto_rt_export_delete_for_vrf(struct bgp
*bgp_vrf
)
4401 evpn_rt_delete_auto(bgp_vrf
, bgp_vrf
->l3vni
, bgp_vrf
->vrf_export_rtl
);
4404 static void bgp_evpn_handle_export_rt_change_for_vrf(struct bgp
*bgp_vrf
)
4406 struct bgp
*bgp_evpn
= NULL
;
4407 struct listnode
*node
= NULL
;
4408 struct bgpevpn
*vpn
= NULL
;
4410 bgp_evpn
= bgp_get_evpn();
4414 /* update all type-5 routes */
4415 update_advertise_vrf_routes(bgp_vrf
);
4417 /* update all type-2 routes */
4418 for (ALL_LIST_ELEMENTS_RO(bgp_vrf
->l2vnis
, node
, vpn
))
4419 update_routes_for_vni(bgp_evpn
, vpn
);
4423 * Handle autort change for a given VNI.
4425 static void update_autort_vni(struct hash_bucket
*bucket
, struct bgp
*bgp
)
4427 struct bgpevpn
*vpn
= bucket
->data
;
4429 if (!is_import_rt_configured(vpn
)) {
4430 if (is_vni_live(vpn
))
4431 bgp_evpn_uninstall_routes(bgp
, vpn
);
4432 bgp_evpn_unmap_vni_from_its_rts(bgp
, vpn
);
4433 list_delete_all_node(vpn
->import_rtl
);
4434 bgp_evpn_derive_auto_rt_import(bgp
, vpn
);
4435 if (is_vni_live(vpn
))
4436 bgp_evpn_install_routes(bgp
, vpn
);
4438 if (!is_export_rt_configured(vpn
)) {
4439 list_delete_all_node(vpn
->export_rtl
);
4440 bgp_evpn_derive_auto_rt_export(bgp
, vpn
);
4441 if (is_vni_live(vpn
))
4442 bgp_evpn_handle_export_rt_change(bgp
, vpn
);
4450 /* withdraw type-5 route corresponding to ip prefix */
4451 void bgp_evpn_withdraw_type5_route(struct bgp
*bgp_vrf
, struct prefix
*p
,
4452 afi_t afi
, safi_t safi
)
4455 struct prefix_evpn evp
;
4456 char buf
[PREFIX_STRLEN
];
4458 build_type5_prefix_from_ip_prefix(&evp
, p
);
4459 ret
= delete_evpn_type5_route(bgp_vrf
, &evp
);
4462 EC_BGP_EVPN_ROUTE_DELETE
,
4463 "%u failed to delete type-5 route for prefix %s in vrf %s",
4464 bgp_vrf
->vrf_id
, prefix2str(p
, buf
, sizeof(buf
)),
4465 vrf_id_to_name(bgp_vrf
->vrf_id
));
4469 /* withdraw all type-5 routes for an address family */
4470 void bgp_evpn_withdraw_type5_routes(struct bgp
*bgp_vrf
, afi_t afi
, safi_t safi
)
4472 struct bgp_table
*table
= NULL
;
4473 struct bgp_node
*rn
= NULL
;
4474 struct bgp_path_info
*pi
;
4476 table
= bgp_vrf
->rib
[afi
][safi
];
4477 for (rn
= bgp_table_top(table
); rn
; rn
= bgp_route_next(rn
)) {
4478 /* Only care about "selected" routes. Also ensure that
4479 * these are routes that are injectable into EVPN.
4481 /* TODO: Support for AddPath for EVPN. */
4482 for (pi
= bgp_node_get_bgp_path_info(rn
); pi
; pi
= pi
->next
) {
4483 if (CHECK_FLAG(pi
->flags
, BGP_PATH_SELECTED
)
4484 && is_route_injectable_into_evpn(pi
)) {
4485 bgp_evpn_withdraw_type5_route(bgp_vrf
, &rn
->p
,
4494 * evpn - enable advertisement of default g/w
4496 void bgp_evpn_install_uninstall_default_route(struct bgp
*bgp_vrf
, afi_t afi
,
4497 safi_t safi
, bool add
)
4499 struct prefix ip_prefix
;
4501 /* form the default prefix 0.0.0.0/0 */
4502 memset(&ip_prefix
, 0, sizeof(struct prefix
));
4503 ip_prefix
.family
= afi2family(afi
);
4506 bgp_evpn_advertise_type5_route(bgp_vrf
, &ip_prefix
,
4509 bgp_evpn_withdraw_type5_route(bgp_vrf
, &ip_prefix
,
4516 * Advertise IP prefix as type-5 route. The afi/safi and src_attr passed
4517 * to this function correspond to those of the source IP prefix (best
4518 * path in the case of the attr. In the case of a local prefix (when we
4519 * are advertising local subnets), the src_attr will be NULL.
4521 void bgp_evpn_advertise_type5_route(struct bgp
*bgp_vrf
, struct prefix
*p
,
4522 struct attr
*src_attr
, afi_t afi
,
4526 struct prefix_evpn evp
;
4527 char buf
[PREFIX_STRLEN
];
4529 build_type5_prefix_from_ip_prefix(&evp
, p
);
4530 ret
= update_evpn_type5_route(bgp_vrf
, &evp
, src_attr
);
4532 flog_err(EC_BGP_EVPN_ROUTE_CREATE
,
4533 "%u: Failed to create type-5 route for prefix %s",
4534 bgp_vrf
->vrf_id
, prefix2str(p
, buf
, sizeof(buf
)));
4537 /* Inject all prefixes of a particular address-family (currently, IPv4 or
4538 * IPv6 unicast) into EVPN as type-5 routes. This is invoked when the
4539 * advertisement is enabled.
4541 void bgp_evpn_advertise_type5_routes(struct bgp
*bgp_vrf
, afi_t afi
,
4544 struct bgp_table
*table
= NULL
;
4545 struct bgp_node
*rn
= NULL
;
4546 struct bgp_path_info
*pi
;
4548 table
= bgp_vrf
->rib
[afi
][safi
];
4549 for (rn
= bgp_table_top(table
); rn
; rn
= bgp_route_next(rn
)) {
4550 /* Need to identify the "selected" route entry to use its
4551 * attribute. Also, ensure that the route is injectable
4553 * TODO: Support for AddPath for EVPN.
4555 for (pi
= bgp_node_get_bgp_path_info(rn
); pi
; pi
= pi
->next
) {
4556 if (CHECK_FLAG(pi
->flags
, BGP_PATH_SELECTED
)
4557 && is_route_injectable_into_evpn(pi
)) {
4559 /* apply the route-map */
4560 if (bgp_vrf
->adv_cmd_rmap
[afi
][safi
].map
) {
4561 route_map_result_t ret
;
4563 ret
= route_map_apply(
4564 bgp_vrf
->adv_cmd_rmap
[afi
][safi
]
4566 &rn
->p
, RMAP_BGP
, pi
);
4567 if (ret
== RMAP_DENYMATCH
)
4570 bgp_evpn_advertise_type5_route(
4571 bgp_vrf
, &rn
->p
, pi
->attr
, afi
, safi
);
4578 void evpn_rt_delete_auto(struct bgp
*bgp
, vni_t vni
, struct list
*rtl
)
4580 struct listnode
*node
, *nnode
, *node_to_del
;
4581 struct ecommunity
*ecom
, *ecom_auto
;
4582 struct ecommunity_val eval
;
4584 if (bgp
->advertise_autort_rfc8365
)
4585 vni
|= EVPN_AUTORT_VXLAN
;
4586 encode_route_target_as((bgp
->as
& 0xFFFF), vni
, &eval
);
4588 ecom_auto
= ecommunity_new();
4589 ecommunity_add_val(ecom_auto
, &eval
);
4592 for (ALL_LIST_ELEMENTS(rtl
, node
, nnode
, ecom
)) {
4593 if (ecommunity_match(ecom
, ecom_auto
)) {
4594 ecommunity_free(&ecom
);
4600 list_delete_node(rtl
, node_to_del
);
4602 ecommunity_free(&ecom_auto
);
4605 void bgp_evpn_configure_import_rt_for_vrf(struct bgp
*bgp_vrf
,
4606 struct ecommunity
*ecomadd
)
4608 /* uninstall routes from vrf */
4609 if (is_l3vni_live(bgp_vrf
))
4610 uninstall_routes_for_vrf(bgp_vrf
);
4612 /* Cleanup the RT to VRF mapping */
4613 bgp_evpn_unmap_vrf_from_its_rts(bgp_vrf
);
4615 /* Remove auto generated RT */
4616 evpn_auto_rt_import_delete_for_vrf(bgp_vrf
);
4618 /* Add the newly configured RT to RT list */
4619 listnode_add_sort(bgp_vrf
->vrf_import_rtl
, ecomadd
);
4620 SET_FLAG(bgp_vrf
->vrf_flags
, BGP_VRF_IMPORT_RT_CFGD
);
4622 /* map VRF to its RTs and install routes matching the new RTs */
4623 if (is_l3vni_live(bgp_vrf
)) {
4624 bgp_evpn_map_vrf_to_its_rts(bgp_vrf
);
4625 install_routes_for_vrf(bgp_vrf
);
4629 void bgp_evpn_unconfigure_import_rt_for_vrf(struct bgp
*bgp_vrf
,
4630 struct ecommunity
*ecomdel
)
4632 struct listnode
*node
= NULL
, *nnode
= NULL
, *node_to_del
= NULL
;
4633 struct ecommunity
*ecom
= NULL
;
4635 /* uninstall routes from vrf */
4636 if (is_l3vni_live(bgp_vrf
))
4637 uninstall_routes_for_vrf(bgp_vrf
);
4639 /* Cleanup the RT to VRF mapping */
4640 bgp_evpn_unmap_vrf_from_its_rts(bgp_vrf
);
4642 /* remove the RT from the RT list */
4643 for (ALL_LIST_ELEMENTS(bgp_vrf
->vrf_import_rtl
, node
, nnode
, ecom
)) {
4644 if (ecommunity_match(ecom
, ecomdel
)) {
4645 ecommunity_free(&ecom
);
4652 list_delete_node(bgp_vrf
->vrf_import_rtl
, node_to_del
);
4654 assert(bgp_vrf
->vrf_import_rtl
);
4655 /* fallback to auto import rt, if this was the last RT */
4656 if (bgp_vrf
->vrf_import_rtl
&& list_isempty(bgp_vrf
->vrf_import_rtl
)) {
4657 UNSET_FLAG(bgp_vrf
->vrf_flags
, BGP_VRF_IMPORT_RT_CFGD
);
4658 evpn_auto_rt_import_add_for_vrf(bgp_vrf
);
4661 /* map VRFs to its RTs and install routes matching this new RT */
4662 if (is_l3vni_live(bgp_vrf
)) {
4663 bgp_evpn_map_vrf_to_its_rts(bgp_vrf
);
4664 install_routes_for_vrf(bgp_vrf
);
4668 void bgp_evpn_configure_export_rt_for_vrf(struct bgp
*bgp_vrf
,
4669 struct ecommunity
*ecomadd
)
4671 /* remove auto-generated RT */
4672 evpn_auto_rt_export_delete_for_vrf(bgp_vrf
);
4674 /* Add the new RT to the RT list */
4675 listnode_add_sort(bgp_vrf
->vrf_export_rtl
, ecomadd
);
4676 SET_FLAG(bgp_vrf
->vrf_flags
, BGP_VRF_EXPORT_RT_CFGD
);
4678 bgp_evpn_handle_export_rt_change_for_vrf(bgp_vrf
);
4681 void bgp_evpn_unconfigure_export_rt_for_vrf(struct bgp
*bgp_vrf
,
4682 struct ecommunity
*ecomdel
)
4684 struct listnode
*node
= NULL
, *nnode
= NULL
, *node_to_del
= NULL
;
4685 struct ecommunity
*ecom
= NULL
;
4687 /* Remove the RT from the RT list */
4688 for (ALL_LIST_ELEMENTS(bgp_vrf
->vrf_export_rtl
, node
, nnode
, ecom
)) {
4689 if (ecommunity_match(ecom
, ecomdel
)) {
4690 ecommunity_free(&ecom
);
4697 list_delete_node(bgp_vrf
->vrf_export_rtl
, node_to_del
);
4700 * Temporary assert to make SA happy.
4701 * The ALL_LIST_ELEMENTS macro above has a NULL check
4702 * which means that SA is going to complain about
4703 * the list_isempty call, which doesn't NULL check.
4704 * So until we get this situation cleaned up, here
4707 assert(bgp_vrf
->vrf_export_rtl
);
4709 /* fall back to auto-generated RT if this was the last RT */
4710 if (list_isempty(bgp_vrf
->vrf_export_rtl
)) {
4711 UNSET_FLAG(bgp_vrf
->vrf_flags
, BGP_VRF_EXPORT_RT_CFGD
);
4712 evpn_auto_rt_export_add_for_vrf(bgp_vrf
);
4715 bgp_evpn_handle_export_rt_change_for_vrf(bgp_vrf
);
4719 * Handle change to BGP router id. This is invoked twice by the change
4720 * handler, first before the router id has been changed and then after
4721 * the router id has been changed. The first invocation will result in
4722 * local routes for all VNIs/VRF being deleted and withdrawn and the next
4723 * will result in the routes being re-advertised.
4725 void bgp_evpn_handle_router_id_update(struct bgp
*bgp
, int withdraw
)
4727 struct listnode
*node
;
4728 struct bgp
*bgp_vrf
;
4732 /* delete and withdraw all the type-5 routes
4733 stored in the global table for this vrf
4735 withdraw_router_id_vrf(bgp
);
4737 /* delete all the VNI routes (type-2/type-3) routes for all the
4740 hash_iterate(bgp
->vnihash
,
4741 (void (*)(struct hash_bucket
*,
4742 void *))withdraw_router_id_vni
,
4745 if (bgp
->inst_type
== BGP_INSTANCE_TYPE_DEFAULT
) {
4746 for (ALL_LIST_ELEMENTS_RO(bm
->bgp
, node
, bgp_vrf
)) {
4747 if (bgp_vrf
->evpn_info
->advertise_pip
&&
4748 (bgp_vrf
->evpn_info
->pip_ip_static
.s_addr
4750 bgp_vrf
->evpn_info
->pip_ip
.s_addr
4756 /* Assign new default instance router-id */
4757 if (bgp
->inst_type
== BGP_INSTANCE_TYPE_DEFAULT
) {
4758 for (ALL_LIST_ELEMENTS_RO(bm
->bgp
, node
, bgp_vrf
)) {
4759 if (bgp_vrf
->evpn_info
->advertise_pip
&&
4760 (bgp_vrf
->evpn_info
->pip_ip_static
.s_addr
4762 bgp_vrf
->evpn_info
->pip_ip
=
4764 /* advertise type-5 routes with
4767 update_advertise_vrf_routes(bgp_vrf
);
4772 /* advertise all routes in the vrf as type-5 routes with the new
4775 update_router_id_vrf(bgp
);
4777 /* advertise all the VNI routes (type-2/type-3) routes with the
4780 hash_iterate(bgp
->vnihash
,
4781 (void (*)(struct hash_bucket
*,
4782 void *))update_router_id_vni
,
4788 * Handle change to auto-RT algorithm - update and advertise local routes.
4790 void bgp_evpn_handle_autort_change(struct bgp
*bgp
)
4792 hash_iterate(bgp
->vnihash
,
4793 (void (*)(struct hash_bucket
*,
4794 void*))update_autort_vni
,
4799 * Handle change to export RT - update and advertise local routes.
4801 int bgp_evpn_handle_export_rt_change(struct bgp
*bgp
, struct bgpevpn
*vpn
)
4803 return update_routes_for_vni(bgp
, vpn
);
4806 void bgp_evpn_handle_vrf_rd_change(struct bgp
*bgp_vrf
, int withdraw
)
4809 delete_withdraw_vrf_routes(bgp_vrf
);
4811 update_advertise_vrf_routes(bgp_vrf
);
4815 * Handle change to RD. This is invoked twice by the change handler,
4816 * first before the RD has been changed and then after the RD has
4817 * been changed. The first invocation will result in local routes
4818 * of this VNI being deleted and withdrawn and the next will result
4819 * in the routes being re-advertised.
4821 void bgp_evpn_handle_rd_change(struct bgp
*bgp
, struct bgpevpn
*vpn
,
4825 delete_withdraw_vni_routes(bgp
, vpn
);
4827 update_advertise_vni_routes(bgp
, vpn
);
4831 * Install routes for this VNI. Invoked upon change to Import RT.
4833 int bgp_evpn_install_routes(struct bgp
*bgp
, struct bgpevpn
*vpn
)
4835 return install_routes_for_vni(bgp
, vpn
);
4839 * Uninstall all routes installed for this VNI. Invoked upon change
4842 int bgp_evpn_uninstall_routes(struct bgp
*bgp
, struct bgpevpn
*vpn
)
4844 return uninstall_routes_for_vni(bgp
, vpn
);
4848 * TODO: Hardcoded for a maximum of 2 VNIs right now
4850 char *bgp_evpn_label2str(mpls_label_t
*label
, uint32_t num_labels
, char *buf
,
4855 vni1
= label2vni(label
);
4856 if (num_labels
== 2) {
4857 vni2
= label2vni(label
+ 1);
4858 snprintf(buf
, len
, "%u/%u", vni1
, vni2
);
4860 snprintf(buf
, len
, "%u", vni1
);
4865 * Function to convert evpn route to json format.
4866 * NOTE: We don't use prefix2str as the output here is a bit different.
4868 void bgp_evpn_route2json(struct prefix_evpn
*p
, json_object
*json
)
4870 char buf1
[ETHER_ADDR_STRLEN
];
4871 char buf2
[PREFIX2STR_BUFFER
];
4878 json_object_int_add(json
, "routeType", p
->prefix
.route_type
);
4880 switch (p
->prefix
.route_type
) {
4881 case BGP_EVPN_MAC_IP_ROUTE
:
4882 json_object_int_add(json
, "ethTag",
4883 p
->prefix
.macip_addr
.eth_tag
);
4884 json_object_int_add(json
, "macLen", 8 * ETH_ALEN
);
4885 json_object_string_add(json
, "mac",
4886 prefix_mac2str(&p
->prefix
.macip_addr
.mac
, buf1
,
4889 if (!is_evpn_prefix_ipaddr_none(p
)) {
4890 family
= is_evpn_prefix_ipaddr_v4(p
) ? AF_INET
:
4892 prefixlen
= (family
== AF_INET
) ?
4893 IPV4_MAX_BITLEN
: IPV6_MAX_BITLEN
;
4894 inet_ntop(family
, &p
->prefix
.macip_addr
.ip
.ip
.addr
,
4895 buf2
, PREFIX2STR_BUFFER
);
4896 json_object_int_add(json
, "ipLen", prefixlen
);
4897 json_object_string_add(json
, "ip", buf2
);
4901 case BGP_EVPN_IMET_ROUTE
:
4902 json_object_int_add(json
, "ethTag",
4903 p
->prefix
.imet_addr
.eth_tag
);
4904 family
= is_evpn_prefix_ipaddr_v4(p
) ? AF_INET
: AF_INET6
;
4905 prefixlen
= (family
== AF_INET
) ? IPV4_MAX_BITLEN
:
4907 inet_ntop(family
, &p
->prefix
.imet_addr
.ip
.ip
.addr
, buf2
,
4909 json_object_int_add(json
, "ipLen", prefixlen
);
4910 json_object_string_add(json
, "ip", buf2
);
4913 case BGP_EVPN_IP_PREFIX_ROUTE
:
4914 json_object_int_add(json
, "ethTag",
4915 p
->prefix
.prefix_addr
.eth_tag
);
4916 family
= is_evpn_prefix_ipaddr_v4(p
) ? AF_INET
: AF_INET6
;
4917 inet_ntop(family
, &p
->prefix
.prefix_addr
.ip
.ip
.addr
,
4918 buf2
, sizeof(buf2
));
4919 json_object_int_add(json
, "ipLen",
4920 p
->prefix
.prefix_addr
.ip_prefix_length
);
4921 json_object_string_add(json
, "ip", buf2
);
4930 * Function to convert evpn route to string.
4931 * NOTE: We don't use prefix2str as the output here is a bit different.
4933 char *bgp_evpn_route2str(struct prefix_evpn
*p
, char *buf
, int len
)
4935 char buf1
[ETHER_ADDR_STRLEN
];
4936 char buf2
[PREFIX2STR_BUFFER
];
4937 char buf3
[ESI_STR_LEN
];
4939 if (p
->prefix
.route_type
== BGP_EVPN_IMET_ROUTE
) {
4940 snprintf(buf
, len
, "[%d]:[%d]:[%d]:[%s]", p
->prefix
.route_type
,
4941 p
->prefix
.imet_addr
.eth_tag
,
4942 is_evpn_prefix_ipaddr_v4(p
) ? IPV4_MAX_BITLEN
4944 inet_ntoa(p
->prefix
.imet_addr
.ip
.ipaddr_v4
));
4945 } else if (p
->prefix
.route_type
== BGP_EVPN_MAC_IP_ROUTE
) {
4946 if (is_evpn_prefix_ipaddr_none(p
))
4947 snprintf(buf
, len
, "[%d]:[%d]:[%d]:[%s]",
4948 p
->prefix
.route_type
,
4949 p
->prefix
.macip_addr
.eth_tag
,
4951 prefix_mac2str(&p
->prefix
.macip_addr
.mac
, buf1
,
4956 family
= is_evpn_prefix_ipaddr_v4(p
) ? AF_INET
4958 snprintf(buf
, len
, "[%d]:[%d]:[%d]:[%s]:[%d]:[%s]",
4959 p
->prefix
.route_type
,
4960 p
->prefix
.macip_addr
.eth_tag
,
4962 prefix_mac2str(&p
->prefix
.macip_addr
.mac
, buf1
,
4964 family
== AF_INET
? IPV4_MAX_BITLEN
4967 &p
->prefix
.macip_addr
.ip
.ip
.addr
,
4969 PREFIX2STR_BUFFER
));
4971 } else if (p
->prefix
.route_type
== BGP_EVPN_IP_PREFIX_ROUTE
) {
4972 snprintf(buf
, len
, "[%d]:[%d]:[%d]:[%s]",
4973 p
->prefix
.route_type
,
4974 p
->prefix
.prefix_addr
.eth_tag
,
4975 p
->prefix
.prefix_addr
.ip_prefix_length
,
4976 is_evpn_prefix_ipaddr_v4(p
)
4977 ? inet_ntoa(p
->prefix
.prefix_addr
.ip
.ipaddr_v4
)
4978 : inet6_ntoa(p
->prefix
.prefix_addr
.ip
.ipaddr_v6
));
4979 } else if (p
->prefix
.route_type
== BGP_EVPN_ES_ROUTE
) {
4980 snprintf(buf
, len
, "[%d]:[%s]:[%d]:[%s]",
4981 p
->prefix
.route_type
,
4982 esi_to_str(&p
->prefix
.es_addr
.esi
, buf3
, sizeof(buf3
)),
4983 is_evpn_prefix_ipaddr_v4(p
) ? IPV4_MAX_BITLEN
4985 inet_ntoa(p
->prefix
.es_addr
.ip
.ipaddr_v4
));
4987 /* For EVPN route types not supported yet. */
4988 snprintf(buf
, len
, "(unsupported route type %d)",
4989 p
->prefix
.route_type
);
4996 * Encode EVPN prefix in Update (MP_REACH)
4998 void bgp_evpn_encode_prefix(struct stream
*s
, struct prefix
*p
,
4999 struct prefix_rd
*prd
, mpls_label_t
*label
,
5000 uint32_t num_labels
, struct attr
*attr
,
5001 int addpath_encode
, uint32_t addpath_tx_id
)
5003 struct prefix_evpn
*evp
= (struct prefix_evpn
*)p
;
5004 int len
, ipa_len
= 0;
5007 stream_putl(s
, addpath_tx_id
);
5010 stream_putc(s
, evp
->prefix
.route_type
);
5012 switch (evp
->prefix
.route_type
) {
5013 case BGP_EVPN_MAC_IP_ROUTE
:
5014 if (is_evpn_prefix_ipaddr_v4(evp
))
5015 ipa_len
= IPV4_MAX_BYTELEN
;
5016 else if (is_evpn_prefix_ipaddr_v6(evp
))
5017 ipa_len
= IPV6_MAX_BYTELEN
;
5018 /* RD, ESI, EthTag, MAC+len, IP len, [IP], 1 VNI */
5019 len
= 8 + 10 + 4 + 1 + 6 + 1 + ipa_len
+ 3;
5020 if (ipa_len
&& num_labels
> 1) /* There are 2 VNIs */
5022 stream_putc(s
, len
);
5023 stream_put(s
, prd
->val
, 8); /* RD */
5025 stream_put(s
, &attr
->evpn_overlay
.eth_s_id
, ESI_LEN
);
5027 stream_put(s
, 0, 10);
5028 stream_putl(s
, evp
->prefix
.macip_addr
.eth_tag
); /* Ethernet Tag ID */
5029 stream_putc(s
, 8 * ETH_ALEN
); /* Mac Addr Len - bits */
5030 stream_put(s
, evp
->prefix
.macip_addr
.mac
.octet
, 6); /* Mac Addr */
5031 stream_putc(s
, 8 * ipa_len
); /* IP address Length */
5032 if (ipa_len
) /* IP */
5033 stream_put(s
, &evp
->prefix
.macip_addr
.ip
.ip
.addr
,
5035 /* 1st label is the L2 VNI */
5036 stream_put(s
, label
, BGP_LABEL_BYTES
);
5037 /* Include 2nd label (L3 VNI) if advertising MAC+IP */
5038 if (ipa_len
&& num_labels
> 1)
5039 stream_put(s
, label
+ 1, BGP_LABEL_BYTES
);
5042 case BGP_EVPN_IMET_ROUTE
:
5043 stream_putc(s
, 17); // TODO: length - assumes IPv4 address
5044 stream_put(s
, prd
->val
, 8); /* RD */
5045 stream_putl(s
, evp
->prefix
.imet_addr
.eth_tag
); /* Ethernet Tag ID */
5046 stream_putc(s
, IPV4_MAX_BITLEN
); /* IP address Length - bits */
5047 /* Originating Router's IP Addr */
5048 stream_put_in_addr(s
, &evp
->prefix
.imet_addr
.ip
.ipaddr_v4
);
5051 case BGP_EVPN_ES_ROUTE
:
5052 stream_putc(s
, 23); /* TODO: length: assumes ipv4 VTEP */
5053 stream_put(s
, prd
->val
, 8); /* RD */
5054 stream_put(s
, evp
->prefix
.es_addr
.esi
.val
, 10); /* ESI */
5055 stream_putc(s
, IPV4_MAX_BITLEN
); /* IP address Length - bits */
5057 stream_put_in_addr(s
, &evp
->prefix
.es_addr
.ip
.ipaddr_v4
);
5060 case BGP_EVPN_IP_PREFIX_ROUTE
:
5061 /* TODO: AddPath support. */
5062 evpn_mpattr_encode_type5(s
, p
, prd
, label
, num_labels
, attr
);
5070 int bgp_nlri_parse_evpn(struct peer
*peer
, struct attr
*attr
,
5071 struct bgp_nlri
*packet
, int withdraw
)
5077 uint32_t addpath_id
;
5078 int addpath_encoded
;
5083 /* Start processing the NLRI - there may be multiple in the MP_REACH */
5085 lim
= pnt
+ packet
->length
;
5087 safi
= packet
->safi
;
5091 (CHECK_FLAG(peer
->af_cap
[afi
][safi
], PEER_CAP_ADDPATH_AF_RX_ADV
)
5092 && CHECK_FLAG(peer
->af_cap
[afi
][safi
],
5093 PEER_CAP_ADDPATH_AF_TX_RCV
));
5095 for (; pnt
< lim
; pnt
+= psize
) {
5096 /* Clear prefix structure. */
5097 memset(&p
, 0, sizeof(struct prefix
));
5099 /* Deal with path-id if AddPath is supported. */
5100 if (addpath_encoded
) {
5101 /* When packet overflow occurs return immediately. */
5102 if (pnt
+ BGP_ADDPATH_ID_LEN
> lim
)
5103 return BGP_NLRI_PARSE_ERROR_PACKET_OVERFLOW
;
5105 memcpy(&addpath_id
, pnt
, BGP_ADDPATH_ID_LEN
);
5106 addpath_id
= ntohl(addpath_id
);
5107 pnt
+= BGP_ADDPATH_ID_LEN
;
5110 /* All EVPN NLRI types start with type and length. */
5112 return BGP_NLRI_PARSE_ERROR_EVPN_MISSING_TYPE
;
5117 /* When packet overflow occur return immediately. */
5118 if (pnt
+ psize
> lim
)
5119 return BGP_NLRI_PARSE_ERROR_PACKET_OVERFLOW
;
5122 case BGP_EVPN_MAC_IP_ROUTE
:
5123 if (process_type2_route(peer
, afi
, safi
,
5124 withdraw
? NULL
: attr
, pnt
,
5125 psize
, addpath_id
)) {
5128 "%u:%s - Error in processing EVPN type-2 NLRI size %d",
5129 peer
->bgp
->vrf_id
, peer
->host
, psize
);
5130 return BGP_NLRI_PARSE_ERROR_EVPN_TYPE2_SIZE
;
5134 case BGP_EVPN_IMET_ROUTE
:
5135 if (process_type3_route(peer
, afi
, safi
,
5136 withdraw
? NULL
: attr
, pnt
,
5137 psize
, addpath_id
)) {
5140 "%u:%s - Error in processing EVPN type-3 NLRI size %d",
5141 peer
->bgp
->vrf_id
, peer
->host
, psize
);
5142 return BGP_NLRI_PARSE_ERROR_EVPN_TYPE3_SIZE
;
5146 case BGP_EVPN_ES_ROUTE
:
5147 if (process_type4_route(peer
, afi
, safi
,
5148 withdraw
? NULL
: attr
, pnt
,
5149 psize
, addpath_id
)) {
5152 "%u:%s - Error in processing EVPN type-4 NLRI size %d",
5153 peer
->bgp
->vrf_id
, peer
->host
, psize
);
5154 return BGP_NLRI_PARSE_ERROR_EVPN_TYPE4_SIZE
;
5158 case BGP_EVPN_IP_PREFIX_ROUTE
:
5159 if (process_type5_route(peer
, afi
, safi
,
5160 withdraw
? NULL
: attr
, pnt
,
5161 psize
, addpath_id
)) {
5164 "%u:%s - Error in processing EVPN type-5 NLRI size %d",
5165 peer
->bgp
->vrf_id
, peer
->host
, psize
);
5166 return BGP_NLRI_PARSE_ERROR_EVPN_TYPE5_SIZE
;
5175 /* Packet length consistency check. */
5177 return BGP_NLRI_PARSE_ERROR_PACKET_LENGTH
;
5179 return BGP_NLRI_PARSE_OK
;
5183 * Map the RTs (configured or automatically derived) of a VRF to the VRF.
5184 * The mapping will be used during route processing.
5185 * bgp_def: default bgp instance
5186 * bgp_vrf: specific bgp vrf instance on which RT is configured
5188 void bgp_evpn_map_vrf_to_its_rts(struct bgp
*bgp_vrf
)
5191 struct ecommunity_val
*eval
= NULL
;
5192 struct listnode
*node
= NULL
, *nnode
= NULL
;
5193 struct ecommunity
*ecom
= NULL
;
5195 for (ALL_LIST_ELEMENTS(bgp_vrf
->vrf_import_rtl
, node
, nnode
, ecom
)) {
5196 for (i
= 0; i
< ecom
->size
; i
++) {
5197 eval
= (struct ecommunity_val
*)(ecom
->val
5199 * ECOMMUNITY_SIZE
));
5200 map_vrf_to_rt(bgp_vrf
, eval
);
5206 * Unmap the RTs (configured or automatically derived) of a VRF from the VRF.
5208 void bgp_evpn_unmap_vrf_from_its_rts(struct bgp
*bgp_vrf
)
5211 struct ecommunity_val
*eval
;
5212 struct listnode
*node
, *nnode
;
5213 struct ecommunity
*ecom
;
5215 for (ALL_LIST_ELEMENTS(bgp_vrf
->vrf_import_rtl
, node
, nnode
, ecom
)) {
5216 for (i
= 0; i
< ecom
->size
; i
++) {
5217 struct vrf_irt_node
*irt
;
5218 struct ecommunity_val eval_tmp
;
5220 eval
= (struct ecommunity_val
*)(ecom
->val
5222 * ECOMMUNITY_SIZE
));
5223 /* If using "automatic" RT, we only care about the
5224 * local-admin sub-field.
5225 * This is to facilitate using VNI as the RT for EBGP
5228 memcpy(&eval_tmp
, eval
, ECOMMUNITY_SIZE
);
5229 if (!CHECK_FLAG(bgp_vrf
->vrf_flags
,
5230 BGP_VRF_IMPORT_RT_CFGD
))
5231 mask_ecom_global_admin(&eval_tmp
, eval
);
5233 irt
= lookup_vrf_import_rt(&eval_tmp
);
5235 unmap_vrf_from_rt(bgp_vrf
, irt
);
5242 * Map the RTs (configured or automatically derived) of a VNI to the VNI.
5243 * The mapping will be used during route processing.
5245 void bgp_evpn_map_vni_to_its_rts(struct bgp
*bgp
, struct bgpevpn
*vpn
)
5248 struct ecommunity_val
*eval
;
5249 struct listnode
*node
, *nnode
;
5250 struct ecommunity
*ecom
;
5252 for (ALL_LIST_ELEMENTS(vpn
->import_rtl
, node
, nnode
, ecom
)) {
5253 for (i
= 0; i
< ecom
->size
; i
++) {
5254 eval
= (struct ecommunity_val
*)(ecom
->val
5256 * ECOMMUNITY_SIZE
));
5257 map_vni_to_rt(bgp
, vpn
, eval
);
5263 * Unmap the RTs (configured or automatically derived) of a VNI from the VNI.
5265 void bgp_evpn_unmap_vni_from_its_rts(struct bgp
*bgp
, struct bgpevpn
*vpn
)
5268 struct ecommunity_val
*eval
;
5269 struct listnode
*node
, *nnode
;
5270 struct ecommunity
*ecom
;
5272 for (ALL_LIST_ELEMENTS(vpn
->import_rtl
, node
, nnode
, ecom
)) {
5273 for (i
= 0; i
< ecom
->size
; i
++) {
5274 struct irt_node
*irt
;
5275 struct ecommunity_val eval_tmp
;
5277 eval
= (struct ecommunity_val
*)(ecom
->val
5279 * ECOMMUNITY_SIZE
));
5280 /* If using "automatic" RT, we only care about the
5281 * local-admin sub-field.
5282 * This is to facilitate using VNI as the RT for EBGP
5285 memcpy(&eval_tmp
, eval
, ECOMMUNITY_SIZE
);
5286 if (!is_import_rt_configured(vpn
))
5287 mask_ecom_global_admin(&eval_tmp
, eval
);
5289 irt
= lookup_import_rt(bgp
, &eval_tmp
);
5291 unmap_vni_from_rt(bgp
, vpn
, irt
);
5297 * Derive Import RT automatically for VNI and map VNI to RT.
5298 * The mapping will be used during route processing.
5300 void bgp_evpn_derive_auto_rt_import(struct bgp
*bgp
, struct bgpevpn
*vpn
)
5302 form_auto_rt(bgp
, vpn
->vni
, vpn
->import_rtl
);
5303 UNSET_FLAG(vpn
->flags
, VNI_FLAG_IMPRT_CFGD
);
5306 bgp_evpn_map_vni_to_its_rts(bgp
, vpn
);
5310 * Derive Export RT automatically for VNI.
5312 void bgp_evpn_derive_auto_rt_export(struct bgp
*bgp
, struct bgpevpn
*vpn
)
5314 form_auto_rt(bgp
, vpn
->vni
, vpn
->export_rtl
);
5315 UNSET_FLAG(vpn
->flags
, VNI_FLAG_EXPRT_CFGD
);
5319 * Derive RD automatically for VNI using passed information - it
5320 * is of the form RouterId:unique-id-for-vni.
5322 void bgp_evpn_derive_auto_rd_for_vrf(struct bgp
*bgp
)
5324 if (is_vrf_rd_configured(bgp
))
5327 form_auto_rd(bgp
->router_id
, bgp
->vrf_rd_id
, &bgp
->vrf_prd
);
5331 * Derive RD automatically for VNI using passed information - it
5332 * is of the form RouterId:unique-id-for-vni.
5334 void bgp_evpn_derive_auto_rd(struct bgp
*bgp
, struct bgpevpn
*vpn
)
5338 vpn
->prd
.family
= AF_UNSPEC
;
5339 vpn
->prd
.prefixlen
= 64;
5340 sprintf(buf
, "%s:%hu", inet_ntoa(bgp
->router_id
), vpn
->rd_id
);
5341 (void)str2prefix_rd(buf
, &vpn
->prd
);
5342 UNSET_FLAG(vpn
->flags
, VNI_FLAG_RD_CFGD
);
5348 bool bgp_evpn_lookup_l3vni_l2vni_table(vni_t vni
)
5350 struct list
*inst
= bm
->bgp
;
5351 struct listnode
*node
;
5352 struct bgp
*bgp_vrf
;
5354 for (ALL_LIST_ELEMENTS_RO(inst
, node
, bgp_vrf
)) {
5355 if (bgp_vrf
->l3vni
== vni
)
5365 struct bgpevpn
*bgp_evpn_lookup_vni(struct bgp
*bgp
, vni_t vni
)
5367 struct bgpevpn
*vpn
;
5370 memset(&tmp
, 0, sizeof(struct bgpevpn
));
5372 vpn
= hash_lookup(bgp
->vnihash
, &tmp
);
5377 * Create a new vpn - invoked upon configuration or zebra notification.
5379 struct bgpevpn
*bgp_evpn_new(struct bgp
*bgp
, vni_t vni
,
5380 struct in_addr originator_ip
,
5381 vrf_id_t tenant_vrf_id
,
5382 struct in_addr mcast_grp
)
5384 struct bgpevpn
*vpn
;
5389 vpn
= XCALLOC(MTYPE_BGP_EVPN
, sizeof(struct bgpevpn
));
5391 /* Set values - RD and RT set to defaults. */
5393 vpn
->originator_ip
= originator_ip
;
5394 vpn
->tenant_vrf_id
= tenant_vrf_id
;
5395 vpn
->mcast_grp
= mcast_grp
;
5397 /* Initialize route-target import and export lists */
5398 vpn
->import_rtl
= list_new();
5399 vpn
->import_rtl
->cmp
= (int (*)(void *, void *))evpn_route_target_cmp
;
5400 vpn
->import_rtl
->del
= evpn_xxport_delete_ecomm
;
5401 vpn
->export_rtl
= list_new();
5402 vpn
->export_rtl
->cmp
= (int (*)(void *, void *))evpn_route_target_cmp
;
5403 vpn
->export_rtl
->del
= evpn_xxport_delete_ecomm
;
5404 bf_assign_index(bm
->rd_idspace
, vpn
->rd_id
);
5405 derive_rd_rt_for_vni(bgp
, vpn
);
5407 /* Initialize EVPN route table. */
5408 vpn
->route_table
= bgp_table_init(bgp
, AFI_L2VPN
, SAFI_EVPN
);
5411 if (!hash_get(bgp
->vnihash
, vpn
, hash_alloc_intern
)) {
5412 XFREE(MTYPE_BGP_EVPN
, vpn
);
5416 /* add to l2vni list on corresponding vrf */
5417 bgpevpn_link_to_l3vni(vpn
);
5419 QOBJ_REG(vpn
, bgpevpn
);
5424 * Free a given VPN - called in multiple scenarios such as zebra
5425 * notification, configuration being deleted, advertise-all-vni disabled etc.
5426 * This just frees appropriate memory, caller should have taken other
5429 void bgp_evpn_free(struct bgp
*bgp
, struct bgpevpn
*vpn
)
5431 bgpevpn_unlink_from_l3vni(vpn
);
5432 bgp_table_unlock(vpn
->route_table
);
5433 bgp_evpn_unmap_vni_from_its_rts(bgp
, vpn
);
5434 list_delete(&vpn
->import_rtl
);
5435 list_delete(&vpn
->export_rtl
);
5436 bf_release_index(bm
->rd_idspace
, vpn
->rd_id
);
5437 hash_release(bgp
->vnihash
, vpn
);
5439 XFREE(MTYPE_BGP_EVPN
, vpn
);
5445 struct evpnes
*bgp_evpn_lookup_es(struct bgp
*bgp
, esi_t
*esi
)
5450 memset(&tmp
, 0, sizeof(struct evpnes
));
5451 memcpy(&tmp
.esi
, esi
, sizeof(esi_t
));
5452 es
= hash_lookup(bgp
->esihash
, &tmp
);
5457 * Create a new local es - invoked upon zebra notification.
5459 struct evpnes
*bgp_evpn_es_new(struct bgp
*bgp
,
5461 struct ipaddr
*originator_ip
)
5469 es
= XCALLOC(MTYPE_BGP_EVPN_ES
, sizeof(struct evpnes
));
5471 /* set the ESI and originator_ip */
5472 memcpy(&es
->esi
, esi
, sizeof(esi_t
));
5473 memcpy(&es
->originator_ip
, originator_ip
, sizeof(struct ipaddr
));
5475 /* Initialise the VTEP list */
5476 es
->vtep_list
= list_new();
5477 es
->vtep_list
->cmp
= evpn_vtep_ip_cmp
;
5479 /* auto derive RD for this es */
5480 bf_assign_index(bm
->rd_idspace
, es
->rd_id
);
5481 es
->prd
.family
= AF_UNSPEC
;
5482 es
->prd
.prefixlen
= 64;
5483 sprintf(buf
, "%s:%hu", inet_ntoa(bgp
->router_id
), es
->rd_id
);
5484 (void)str2prefix_rd(buf
, &es
->prd
);
5486 /* Initialize the ES route table */
5487 es
->route_table
= bgp_table_init(bgp
, AFI_L2VPN
, SAFI_EVPN
);
5490 if (!hash_get(bgp
->esihash
, es
, hash_alloc_intern
)) {
5491 XFREE(MTYPE_BGP_EVPN_ES
, es
);
5495 QOBJ_REG(es
, evpnes
);
5501 * This just frees appropriate memory, caller should have taken other
5504 void bgp_evpn_es_free(struct bgp
*bgp
, struct evpnes
*es
)
5506 list_delete(&es
->vtep_list
);
5507 bgp_table_unlock(es
->route_table
);
5508 bf_release_index(bm
->rd_idspace
, es
->rd_id
);
5509 hash_release(bgp
->esihash
, es
);
5511 XFREE(MTYPE_BGP_EVPN_ES
, es
);
5515 * Import evpn route from global table to VNI/VRF/ESI.
5517 int bgp_evpn_import_route(struct bgp
*bgp
, afi_t afi
, safi_t safi
,
5518 struct prefix
*p
, struct bgp_path_info
*pi
)
5520 return install_uninstall_evpn_route(bgp
, afi
, safi
, p
, pi
, 1);
5524 * Unimport evpn route from VNI/VRF/ESI.
5526 int bgp_evpn_unimport_route(struct bgp
*bgp
, afi_t afi
, safi_t safi
,
5527 struct prefix
*p
, struct bgp_path_info
*pi
)
5529 return install_uninstall_evpn_route(bgp
, afi
, safi
, p
, pi
, 0);
5532 /* filter routes which have martian next hops */
5533 int bgp_filter_evpn_routes_upon_martian_nh_change(struct bgp
*bgp
)
5537 struct bgp_node
*rd_rn
, *rn
;
5538 struct bgp_table
*table
;
5539 struct bgp_path_info
*pi
;
5544 /* Walk entire global routing table and evaluate routes which could be
5545 * imported into this VPN. Note that we cannot just look at the routes
5546 * for the VNI's RD -
5547 * remote routes applicable for this VNI could have any RD.
5549 /* EVPN routes are a 2-level table. */
5550 for (rd_rn
= bgp_table_top(bgp
->rib
[afi
][safi
]); rd_rn
;
5551 rd_rn
= bgp_route_next(rd_rn
)) {
5552 table
= bgp_node_get_bgp_table_info(rd_rn
);
5556 for (rn
= bgp_table_top(table
); rn
; rn
= bgp_route_next(rn
)) {
5558 for (pi
= bgp_node_get_bgp_path_info(rn
); pi
;
5561 /* Consider "valid" remote routes applicable for
5563 if (!(pi
->type
== ZEBRA_ROUTE_BGP
5564 && pi
->sub_type
== BGP_ROUTE_NORMAL
))
5566 if (bgp_nexthop_self(bgp
, afi
,
5567 pi
->type
, pi
->sub_type
,
5570 char attr_str
[BUFSIZ
] = {0};
5571 char pbuf
[PREFIX_STRLEN
];
5573 bgp_dump_attr(pi
->attr
, attr_str
,
5576 if (bgp_debug_update(pi
->peer
, &rn
->p
,
5579 "%u: prefix %s with attr %s - DENIED due to martian or self nexthop",
5586 bgp_evpn_unimport_route(bgp
, afi
, safi
,
5589 bgp_rib_remove(rn
, pi
, pi
->peer
, afi
,
5600 * Handle del of a local MACIP.
5602 int bgp_evpn_local_macip_del(struct bgp
*bgp
, vni_t vni
, struct ethaddr
*mac
,
5603 struct ipaddr
*ip
, int state
)
5605 struct bgpevpn
*vpn
;
5606 struct prefix_evpn p
;
5607 struct bgp_node
*rn
;
5609 /* Lookup VNI hash - should exist. */
5610 vpn
= bgp_evpn_lookup_vni(bgp
, vni
);
5611 if (!vpn
|| !is_vni_live(vpn
)) {
5612 flog_warn(EC_BGP_EVPN_VPN_VNI
,
5613 "%u: VNI hash entry for VNI %u %s at MACIP DEL",
5614 bgp
->vrf_id
, vni
, vpn
? "not live" : "not found");
5618 build_evpn_type2_prefix(&p
, mac
, ip
);
5619 if (state
== ZEBRA_NEIGH_ACTIVE
) {
5620 /* Remove EVPN type-2 route and schedule for processing. */
5621 delete_evpn_route(bgp
, vpn
, &p
);
5623 /* Re-instate the current remote best path if any */
5624 rn
= bgp_node_lookup(vpn
->route_table
, (struct prefix
*)&p
);
5626 evpn_zebra_reinstall_best_route(bgp
, vpn
, rn
);
5633 * Handle add of a local MACIP.
5635 int bgp_evpn_local_macip_add(struct bgp
*bgp
, vni_t vni
, struct ethaddr
*mac
,
5636 struct ipaddr
*ip
, uint8_t flags
, uint32_t seq
)
5638 struct bgpevpn
*vpn
;
5639 struct prefix_evpn p
;
5641 /* Lookup VNI hash - should exist. */
5642 vpn
= bgp_evpn_lookup_vni(bgp
, vni
);
5643 if (!vpn
|| !is_vni_live(vpn
)) {
5644 flog_warn(EC_BGP_EVPN_VPN_VNI
,
5645 "%u: VNI hash entry for VNI %u %s at MACIP ADD",
5646 bgp
->vrf_id
, vni
, vpn
? "not live" : "not found");
5650 /* Create EVPN type-2 route and schedule for processing. */
5651 build_evpn_type2_prefix(&p
, mac
, ip
);
5652 if (update_evpn_route(bgp
, vpn
, &p
, flags
, seq
)) {
5653 char buf
[ETHER_ADDR_STRLEN
];
5654 char buf2
[INET6_ADDRSTRLEN
];
5657 EC_BGP_EVPN_ROUTE_CREATE
,
5658 "%u:Failed to create Type-2 route, VNI %u %s MAC %s IP %s (flags: 0x%x)",
5659 bgp
->vrf_id
, vpn
->vni
,
5660 CHECK_FLAG(flags
, ZEBRA_MACIP_TYPE_STICKY
)
5663 prefix_mac2str(mac
, buf
, sizeof(buf
)),
5664 ipaddr2str(ip
, buf2
, sizeof(buf2
)), flags
);
5671 static void link_l2vni_hash_to_l3vni(struct hash_bucket
*bucket
,
5672 struct bgp
*bgp_vrf
)
5674 struct bgpevpn
*vpn
= (struct bgpevpn
*)bucket
->data
;
5675 struct bgp
*bgp_evpn
= NULL
;
5677 bgp_evpn
= bgp_get_evpn();
5680 if (vpn
->tenant_vrf_id
== bgp_vrf
->vrf_id
)
5681 bgpevpn_link_to_l3vni(vpn
);
5684 int bgp_evpn_local_l3vni_add(vni_t l3vni
, vrf_id_t vrf_id
,
5685 struct ethaddr
*svi_rmac
,
5686 struct ethaddr
*vrr_rmac
,
5687 struct in_addr originator_ip
, int filter
,
5688 ifindex_t svi_ifindex
,
5689 bool is_anycast_mac
)
5691 struct bgp
*bgp_vrf
= NULL
; /* bgp VRF instance */
5692 struct bgp
*bgp_evpn
= NULL
; /* EVPN bgp instance */
5693 struct listnode
*node
= NULL
;
5694 struct bgpevpn
*vpn
= NULL
;
5697 /* get the EVPN instance - required to get the AS number for VRF
5700 bgp_evpn
= bgp_get_evpn();
5704 "Cannot process L3VNI %u ADD - EVPN BGP instance not yet created",
5710 /* if the BGP vrf instance doesn't exist - create one */
5711 bgp_vrf
= bgp_lookup_by_vrf_id(vrf_id
);
5716 ret
= bgp_get_vty(&bgp_vrf
, &as
, vrf_id_to_name(vrf_id
),
5717 vrf_id
== VRF_DEFAULT
5718 ? BGP_INSTANCE_TYPE_DEFAULT
5719 : BGP_INSTANCE_TYPE_VRF
);
5721 case BGP_ERR_AS_MISMATCH
:
5722 flog_err(EC_BGP_EVPN_AS_MISMATCH
,
5723 "BGP is already running; AS is %u\n", as
);
5725 case BGP_ERR_INSTANCE_MISMATCH
:
5726 flog_err(EC_BGP_EVPN_INSTANCE_MISMATCH
,
5727 "BGP instance name and AS number mismatch\n");
5731 /* mark as auto created */
5732 SET_FLAG(bgp_vrf
->vrf_flags
, BGP_VRF_AUTO
);
5735 /* associate the vrf with l3vni and related parameters */
5736 bgp_vrf
->l3vni
= l3vni
;
5737 bgp_vrf
->originator_ip
= originator_ip
;
5738 bgp_vrf
->l3vni_svi_ifindex
= svi_ifindex
;
5739 bgp_vrf
->evpn_info
->is_anycast_mac
= is_anycast_mac
;
5741 /* copy anycast MAC from VRR MAC */
5742 memcpy(&bgp_vrf
->rmac
, vrr_rmac
, ETH_ALEN
);
5743 /* copy sys RMAC from SVI MAC */
5744 memcpy(&bgp_vrf
->evpn_info
->pip_rmac_zebra
, svi_rmac
, ETH_ALEN
);
5745 /* PIP user configured mac is not present use svi mac as sys mac */
5746 if (is_zero_mac(&bgp_vrf
->evpn_info
->pip_rmac_static
))
5747 memcpy(&bgp_vrf
->evpn_info
->pip_rmac
, svi_rmac
, ETH_ALEN
);
5749 if (bgp_debug_zebra(NULL
)) {
5750 char buf
[ETHER_ADDR_STRLEN
];
5751 char buf1
[ETHER_ADDR_STRLEN
];
5752 char buf2
[ETHER_ADDR_STRLEN
];
5754 zlog_debug("VRF %s vni %u pip %s RMAC %s sys RMAC %s static RMAC %s is_anycast_mac %s",
5755 vrf_id_to_name(bgp_vrf
->vrf_id
),
5757 bgp_vrf
->evpn_info
->advertise_pip
? "enable"
5759 prefix_mac2str(&bgp_vrf
->rmac
, buf
, sizeof(buf
)),
5760 prefix_mac2str(&bgp_vrf
->evpn_info
->pip_rmac
,
5761 buf1
, sizeof(buf1
)),
5762 prefix_mac2str(&bgp_vrf
->evpn_info
->pip_rmac_static
,
5763 buf2
, sizeof(buf2
)),
5764 is_anycast_mac
? "Enable" : "Disable");
5766 /* set the right filter - are we using l3vni only for prefix routes? */
5768 SET_FLAG(bgp_vrf
->vrf_flags
, BGP_VRF_L3VNI_PREFIX_ROUTES_ONLY
);
5770 /* Map auto derive or configured RTs */
5771 if (!CHECK_FLAG(bgp_vrf
->vrf_flags
, BGP_VRF_IMPORT_RT_CFGD
))
5772 evpn_auto_rt_import_add_for_vrf(bgp_vrf
);
5774 bgp_evpn_map_vrf_to_its_rts(bgp_vrf
);
5776 if (!CHECK_FLAG(bgp_vrf
->vrf_flags
, BGP_VRF_EXPORT_RT_CFGD
))
5777 evpn_auto_rt_export_add_for_vrf(bgp_vrf
);
5779 /* auto derive RD */
5780 bgp_evpn_derive_auto_rd_for_vrf(bgp_vrf
);
5782 /* link all corresponding l2vnis */
5783 hash_iterate(bgp_evpn
->vnihash
,
5784 (void (*)(struct hash_bucket
*,
5785 void *))link_l2vni_hash_to_l3vni
,
5788 /* Only update all corresponding type-2 routes if we are advertising two
5789 * labels along with type-2 routes
5792 for (ALL_LIST_ELEMENTS_RO(bgp_vrf
->l2vnis
, node
, vpn
))
5793 update_routes_for_vni(bgp_evpn
, vpn
);
5795 /* advertise type-5 routes if needed */
5796 update_advertise_vrf_routes(bgp_vrf
);
5798 /* install all remote routes belonging to this l3vni into correspondng
5800 install_routes_for_vrf(bgp_vrf
);
5805 int bgp_evpn_local_l3vni_del(vni_t l3vni
, vrf_id_t vrf_id
)
5807 struct bgp
*bgp_vrf
= NULL
; /* bgp vrf instance */
5808 struct bgp
*bgp_evpn
= NULL
; /* EVPN bgp instance */
5809 struct listnode
*node
= NULL
;
5810 struct listnode
*next
= NULL
;
5811 struct bgpevpn
*vpn
= NULL
;
5813 bgp_vrf
= bgp_lookup_by_vrf_id(vrf_id
);
5817 "Cannot process L3VNI %u Del - Could not find BGP instance",
5822 bgp_evpn
= bgp_get_evpn();
5826 "Cannot process L3VNI %u Del - Could not find EVPN BGP instance",
5831 /* Remove remote routes from BGT VRF even if BGP_VRF_AUTO is configured,
5832 * bgp_delete would not remove/decrement bgp_path_info of the ip_prefix
5833 * routes. This will uninstalling the routes from zebra and decremnt the
5836 uninstall_routes_for_vrf(bgp_vrf
);
5838 /* delete/withdraw all type-5 routes */
5839 delete_withdraw_vrf_routes(bgp_vrf
);
5841 /* remove the l3vni from vrf instance */
5844 /* remove the Rmac from the BGP vrf */
5845 memset(&bgp_vrf
->rmac
, 0, sizeof(struct ethaddr
));
5846 memset(&bgp_vrf
->evpn_info
->pip_rmac_zebra
, 0, ETH_ALEN
);
5847 if (is_zero_mac(&bgp_vrf
->evpn_info
->pip_rmac_static
) &&
5848 !is_zero_mac(&bgp_vrf
->evpn_info
->pip_rmac
))
5849 memset(&bgp_vrf
->evpn_info
->pip_rmac
, 0, ETH_ALEN
);
5851 /* remove default import RT or Unmap non-default import RT */
5852 if (!list_isempty(bgp_vrf
->vrf_import_rtl
)) {
5853 bgp_evpn_unmap_vrf_from_its_rts(bgp_vrf
);
5854 if (!CHECK_FLAG(bgp_vrf
->vrf_flags
, BGP_VRF_IMPORT_RT_CFGD
))
5855 list_delete_all_node(bgp_vrf
->vrf_import_rtl
);
5858 /* remove default export RT */
5859 if (!list_isempty(bgp_vrf
->vrf_export_rtl
) &&
5860 !CHECK_FLAG(bgp_vrf
->vrf_flags
, BGP_VRF_EXPORT_RT_CFGD
)) {
5861 list_delete_all_node(bgp_vrf
->vrf_export_rtl
);
5864 /* update all corresponding local mac-ip routes */
5865 if (!CHECK_FLAG(bgp_vrf
->vrf_flags
, BGP_VRF_L3VNI_PREFIX_ROUTES_ONLY
)) {
5866 for (ALL_LIST_ELEMENTS_RO(bgp_vrf
->l2vnis
, node
, vpn
)) {
5867 UNSET_FLAG(vpn
->flags
, VNI_FLAG_USE_TWO_LABELS
);
5868 update_routes_for_vni(bgp_evpn
, vpn
);
5872 /* If any L2VNIs point to this instance, unlink them. */
5873 for (ALL_LIST_ELEMENTS(bgp_vrf
->l2vnis
, node
, next
, vpn
))
5874 bgpevpn_unlink_from_l3vni(vpn
);
5876 UNSET_FLAG(bgp_vrf
->vrf_flags
, BGP_VRF_L3VNI_PREFIX_ROUTES_ONLY
);
5878 /* Delete the instance if it was autocreated */
5879 if (CHECK_FLAG(bgp_vrf
->vrf_flags
, BGP_VRF_AUTO
))
5880 bgp_delete(bgp_vrf
);
5886 * Handle del of a local VNI.
5888 int bgp_evpn_local_vni_del(struct bgp
*bgp
, vni_t vni
)
5890 struct bgpevpn
*vpn
;
5892 /* Locate VNI hash */
5893 vpn
= bgp_evpn_lookup_vni(bgp
, vni
);
5895 if (bgp_debug_zebra(NULL
))
5897 EC_BGP_EVPN_VPN_VNI
,
5898 "%u: VNI hash entry for VNI %u not found at DEL",
5903 /* Remove all local EVPN routes and schedule for processing (to
5904 * withdraw from peers).
5906 delete_routes_for_vni(bgp
, vpn
);
5909 * tunnel is no longer active, del tunnel ip address from tip_hash
5911 bgp_tip_del(bgp
, &vpn
->originator_ip
);
5913 /* Clear "live" flag and see if hash needs to be freed. */
5914 UNSET_FLAG(vpn
->flags
, VNI_FLAG_LIVE
);
5915 if (!is_vni_configured(vpn
))
5916 bgp_evpn_free(bgp
, vpn
);
5922 * Handle add (or update) of a local VNI. The VNI changes we care
5923 * about are for the local-tunnel-ip and the (tenant) VRF.
5925 int bgp_evpn_local_vni_add(struct bgp
*bgp
, vni_t vni
,
5926 struct in_addr originator_ip
,
5927 vrf_id_t tenant_vrf_id
,
5928 struct in_addr mcast_grp
)
5931 struct bgpevpn
*vpn
;
5932 struct prefix_evpn p
;
5934 /* Lookup VNI. If present and no change, exit. */
5935 vpn
= bgp_evpn_lookup_vni(bgp
, vni
);
5938 if (is_vni_live(vpn
)
5939 && IPV4_ADDR_SAME(&vpn
->originator_ip
, &originator_ip
)
5940 && IPV4_ADDR_SAME(&vpn
->mcast_grp
, &mcast_grp
)
5941 && vpn
->tenant_vrf_id
== tenant_vrf_id
)
5942 /* Probably some other param has changed that we don't
5946 bgp_evpn_mcast_grp_change(bgp
, vpn
, mcast_grp
);
5948 /* Update tenant_vrf_id if it has changed. */
5949 if (vpn
->tenant_vrf_id
!= tenant_vrf_id
) {
5950 bgpevpn_unlink_from_l3vni(vpn
);
5951 vpn
->tenant_vrf_id
= tenant_vrf_id
;
5952 bgpevpn_link_to_l3vni(vpn
);
5955 /* If tunnel endpoint IP has changed, update (and delete prior
5956 * type-3 route, if needed.)
5958 if (!IPV4_ADDR_SAME(&vpn
->originator_ip
, &originator_ip
))
5959 handle_tunnel_ip_change(bgp
, vpn
, originator_ip
);
5961 /* Update all routes with new endpoint IP and/or export RT
5964 if (is_vni_live(vpn
))
5965 update_routes_for_vni(bgp
, vpn
);
5968 /* Create or update as appropriate. */
5970 vpn
= bgp_evpn_new(bgp
, vni
, originator_ip
, tenant_vrf_id
,
5975 "%u: Failed to allocate VNI entry for VNI %u - at Add",
5981 /* if the VNI is live already, there is nothing more to do */
5982 if (is_vni_live(vpn
))
5985 /* Mark as "live" */
5986 SET_FLAG(vpn
->flags
, VNI_FLAG_LIVE
);
5988 /* tunnel is now active, add tunnel-ip to db */
5989 bgp_tip_add(bgp
, &originator_ip
);
5991 /* filter routes as nexthop database has changed */
5992 bgp_filter_evpn_routes_upon_martian_nh_change(bgp
);
5995 * Create EVPN type-3 route and schedule for processing.
5997 * RT-3 only if doing head-end replication
5999 if (bgp_evpn_vni_flood_mode_get(bgp
, vpn
)
6000 == VXLAN_FLOOD_HEAD_END_REPL
) {
6001 build_evpn_type3_prefix(&p
, vpn
->originator_ip
);
6002 if (update_evpn_route(bgp
, vpn
, &p
, 0, 0)) {
6003 flog_err(EC_BGP_EVPN_ROUTE_CREATE
,
6004 "%u: Type3 route creation failure for VNI %u",
6010 /* If we have learnt and retained remote routes (VTEPs, MACs) for this
6014 install_routes_for_vni(bgp
, vpn
);
6016 /* If we are advertising gateway mac-ip
6017 It needs to be conveyed again to zebra */
6018 bgp_zebra_advertise_gw_macip(bgp
, vpn
->advertise_gw_macip
, vpn
->vni
);
6020 /* advertise svi mac-ip knob to zebra */
6021 bgp_zebra_advertise_svi_macip(bgp
, vpn
->advertise_svi_macip
, vpn
->vni
);
6027 * bgp_evpn_local_es_del
6029 int bgp_evpn_local_es_del(struct bgp
*bgp
,
6031 struct ipaddr
*originator_ip
)
6033 char buf
[ESI_STR_LEN
];
6034 struct evpnes
*es
= NULL
;
6036 if (!bgp
->esihash
) {
6037 flog_err(EC_BGP_ES_CREATE
, "%u: ESI hash not yet created",
6042 /* Lookup ESI hash - should exist. */
6043 es
= bgp_evpn_lookup_es(bgp
, esi
);
6045 flog_warn(EC_BGP_EVPN_ESI
,
6046 "%u: ESI hash entry for ESI %s at Local ES DEL",
6047 bgp
->vrf_id
, esi_to_str(esi
, buf
, sizeof(buf
)));
6051 /* Delete all local EVPN ES routes from ESI table
6052 * and schedule for processing (to withdraw from peers))
6054 delete_routes_for_es(bgp
, es
);
6056 /* free the hash entry */
6057 bgp_evpn_es_free(bgp
, es
);
6063 * bgp_evpn_local_es_add
6065 int bgp_evpn_local_es_add(struct bgp
*bgp
,
6067 struct ipaddr
*originator_ip
)
6069 char buf
[ESI_STR_LEN
];
6070 struct evpnes
*es
= NULL
;
6071 struct prefix_evpn p
;
6073 if (!bgp
->esihash
) {
6074 flog_err(EC_BGP_ES_CREATE
, "%u: ESI hash not yet created",
6079 /* create the new es */
6080 es
= bgp_evpn_lookup_es(bgp
, esi
);
6082 es
= bgp_evpn_es_new(bgp
, esi
, originator_ip
);
6086 "%u: Failed to allocate ES entry for ESI %s - at Local ES Add",
6087 bgp
->vrf_id
, esi_to_str(esi
, buf
, sizeof(buf
)));
6091 UNSET_FLAG(es
->flags
, EVPNES_REMOTE
);
6092 SET_FLAG(es
->flags
, EVPNES_LOCAL
);
6094 build_evpn_type4_prefix(&p
, esi
, originator_ip
->ipaddr_v4
);
6095 if (update_evpn_type4_route(bgp
, es
, &p
)) {
6096 flog_err(EC_BGP_EVPN_ROUTE_CREATE
,
6097 "%u: Type4 route creation failure for ESI %s",
6098 bgp
->vrf_id
, esi_to_str(esi
, buf
, sizeof(buf
)));
6102 /* import all remote ES routes in th ES table */
6103 install_routes_for_es(bgp
, es
);
6109 * Handle change in setting for BUM handling. The supported values
6110 * are head-end replication and dropping all BUM packets. Any change
6111 * should be registered with zebra. Also, if doing head-end replication,
6112 * need to advertise local VNIs as EVPN RT-3 wheras, if BUM packets are
6113 * to be dropped, the RT-3s must be withdrawn.
6115 void bgp_evpn_flood_control_change(struct bgp
*bgp
)
6117 zlog_info("L2VPN EVPN BUM handling is %s",
6118 bgp
->vxlan_flood_ctrl
== VXLAN_FLOOD_HEAD_END_REPL
?
6119 "Flooding" : "Flooding Disabled");
6121 bgp_zebra_vxlan_flood_control(bgp
, bgp
->vxlan_flood_ctrl
);
6122 if (bgp
->vxlan_flood_ctrl
== VXLAN_FLOOD_HEAD_END_REPL
)
6123 hash_iterate(bgp
->vnihash
, create_advertise_type3
, bgp
);
6124 else if (bgp
->vxlan_flood_ctrl
== VXLAN_FLOOD_DISABLED
)
6125 hash_iterate(bgp
->vnihash
, delete_withdraw_type3
, bgp
);
6129 * Cleanup EVPN information on disable - Need to delete and withdraw
6130 * EVPN routes from peers.
6132 void bgp_evpn_cleanup_on_disable(struct bgp
*bgp
)
6134 hash_iterate(bgp
->vnihash
, (void (*)(struct hash_bucket
*,
6135 void *))cleanup_vni_on_disable
,
6140 * Cleanup EVPN information - invoked at the time of bgpd exit or when the
6141 * BGP instance (default) is being freed.
6143 void bgp_evpn_cleanup(struct bgp
*bgp
)
6145 hash_iterate(bgp
->vnihash
,
6146 (void (*)(struct hash_bucket
*, void *))free_vni_entry
,
6149 hash_free(bgp
->import_rt_hash
);
6150 bgp
->import_rt_hash
= NULL
;
6152 hash_free(bgp
->vrf_import_rt_hash
);
6153 bgp
->vrf_import_rt_hash
= NULL
;
6155 hash_free(bgp
->vnihash
);
6156 bgp
->vnihash
= NULL
;
6158 hash_free(bgp
->esihash
);
6159 bgp
->esihash
= NULL
;
6161 list_delete(&bgp
->vrf_import_rtl
);
6162 list_delete(&bgp
->vrf_export_rtl
);
6163 list_delete(&bgp
->l2vnis
);
6167 * Initialization for EVPN
6170 * hash for RT to VNI
6172 void bgp_evpn_init(struct bgp
*bgp
)
6175 hash_create(vni_hash_key_make
, vni_hash_cmp
, "BGP VNI Hash");
6177 hash_create(esi_hash_keymake
, esi_cmp
,
6178 "BGP EVPN Local ESI Hash");
6179 bgp
->import_rt_hash
=
6180 hash_create(import_rt_hash_key_make
, import_rt_hash_cmp
,
6181 "BGP Import RT Hash");
6182 bgp
->vrf_import_rt_hash
=
6183 hash_create(vrf_import_rt_hash_key_make
, vrf_import_rt_hash_cmp
,
6184 "BGP VRF Import RT Hash");
6185 bgp
->vrf_import_rtl
= list_new();
6186 bgp
->vrf_import_rtl
->cmp
=
6187 (int (*)(void *, void *))evpn_route_target_cmp
;
6188 bgp
->vrf_import_rtl
->del
= evpn_xxport_delete_ecomm
;
6189 bgp
->vrf_export_rtl
= list_new();
6190 bgp
->vrf_export_rtl
->cmp
=
6191 (int (*)(void *, void *))evpn_route_target_cmp
;
6192 bgp
->vrf_export_rtl
->del
= evpn_xxport_delete_ecomm
;
6193 bgp
->l2vnis
= list_new();
6194 bgp
->l2vnis
->cmp
= vni_list_cmp
;
6195 /* By default Duplicate Address Dection is enabled.
6196 * Max-moves (N) 5, detection time (M) 180
6197 * default action is warning-only
6198 * freeze action permanently freezes address,
6199 * and freeze time (auto-recovery) is disabled.
6201 if (bgp
->evpn_info
) {
6202 bgp
->evpn_info
->dup_addr_detect
= true;
6203 bgp
->evpn_info
->dad_time
= EVPN_DAD_DEFAULT_TIME
;
6204 bgp
->evpn_info
->dad_max_moves
= EVPN_DAD_DEFAULT_MAX_MOVES
;
6205 bgp
->evpn_info
->dad_freeze
= false;
6206 bgp
->evpn_info
->dad_freeze_time
= 0;
6207 /* Initialize zebra vxlan */
6208 bgp_zebra_dup_addr_detection(bgp
);
6209 /* Enable PIP feature by default for bgp vrf instance */
6210 if (bgp
->inst_type
== BGP_INSTANCE_TYPE_VRF
) {
6211 struct bgp
*bgp_default
;
6213 bgp
->evpn_info
->advertise_pip
= true;
6214 bgp_default
= bgp_get_default();
6216 bgp
->evpn_info
->pip_ip
= bgp_default
->router_id
;
6220 /* Default BUM handling is to do head-end replication. */
6221 bgp
->vxlan_flood_ctrl
= VXLAN_FLOOD_HEAD_END_REPL
;
6224 void bgp_evpn_vrf_delete(struct bgp
*bgp_vrf
)
6226 bgp_evpn_unmap_vrf_from_its_rts(bgp_vrf
);
6230 * Get the prefixlen of the ip prefix carried within the type5 evpn route.
6232 int bgp_evpn_get_type5_prefixlen(struct prefix
*pfx
)
6234 struct prefix_evpn
*evp
= (struct prefix_evpn
*)pfx
;
6236 if (!pfx
|| pfx
->family
!= AF_EVPN
)
6239 if (evp
->prefix
.route_type
!= BGP_EVPN_IP_PREFIX_ROUTE
)
6242 return evp
->prefix
.prefix_addr
.ip_prefix_length
;
6246 * Should we register nexthop for this EVPN prefix for nexthop tracking?
6248 bool bgp_evpn_is_prefix_nht_supported(struct prefix
*pfx
)
6250 struct prefix_evpn
*evp
= (struct prefix_evpn
*)pfx
;
6253 * EVPN RT-5 should not be marked as valid and imported to vrfs if the
6254 * BGP nexthop is not reachable. To check for the nexthop reachability,
6255 * Add nexthop for EVPN RT-5 for nexthop tracking.
6257 * Ideally, a BGP route should be marked as valid only if the
6258 * nexthop is reachable. Thus, other EVPN route types also should be
6259 * added here after testing is performed for them.
6261 if (pfx
&& pfx
->family
== AF_EVPN
&&
6262 evp
->prefix
.route_type
== BGP_EVPN_IP_PREFIX_ROUTE
)