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
35 #include "bgpd/bgp_attr_evpn.h"
36 #include "bgpd/bgpd.h"
37 #include "bgpd/bgp_table.h"
38 #include "bgpd/bgp_route.h"
39 #include "bgpd/bgp_attr.h"
40 #include "bgpd/bgp_mplsvpn.h"
41 #include "bgpd/bgp_label.h"
42 #include "bgpd/bgp_evpn.h"
43 #include "bgpd/bgp_evpn_private.h"
44 #include "bgpd/bgp_ecommunity.h"
45 #include "bgpd/bgp_encap_types.h"
46 #include "bgpd/bgp_debug.h"
47 #include "bgpd/bgp_aspath.h"
48 #include "bgpd/bgp_zebra.h"
49 #include "bgpd/bgp_nexthop.h"
52 * Definitions and external declarations.
54 extern struct zclient
*zclient
;
56 DEFINE_QOBJ_TYPE(bgpevpn
)
60 * Static function declarations
62 static void delete_evpn_route_entry(struct bgp
*bgp
, struct bgpevpn
*vpn
,
63 afi_t afi
, safi_t safi
, struct bgp_node
*rn
,
64 struct bgp_info
**ri
);
65 static int delete_all_vni_routes(struct bgp
*bgp
, struct bgpevpn
*vpn
);
74 static unsigned int vni_hash_key_make(void *p
)
76 struct bgpevpn
*vpn
= p
;
77 return (jhash_1word(vpn
->vni
, 0));
81 * Comparison function for vni hash
83 static int vni_hash_cmp(const void *p1
, const void *p2
)
85 const struct bgpevpn
*vpn1
= p1
;
86 const struct bgpevpn
*vpn2
= p2
;
92 return (vpn1
->vni
== vpn2
->vni
);
96 * Make vrf import route target hash key.
98 static unsigned int vrf_import_rt_hash_key_make(void *p
)
100 struct vrf_irt_node
*irt
= p
;
101 char *pnt
= irt
->rt
.val
;
103 return jhash(pnt
, 8, 0x5abc1234);
107 * Comparison function for vrf import rt hash
109 static int vrf_import_rt_hash_cmp(const void *p1
, const void *p2
)
111 const struct vrf_irt_node
*irt1
= p1
;
112 const struct vrf_irt_node
*irt2
= p2
;
114 if (irt1
== NULL
&& irt2
== NULL
)
117 if (irt1
== NULL
|| irt2
== NULL
)
120 return (memcmp(irt1
->rt
.val
, irt2
->rt
.val
, ECOMMUNITY_SIZE
) == 0);
124 * Create a new vrf import_rt in default instance
126 static struct vrf_irt_node
*vrf_import_rt_new(struct ecommunity_val
*rt
)
128 struct bgp
*bgp_def
= NULL
;
129 struct vrf_irt_node
*irt
;
131 bgp_def
= bgp_get_default();
133 zlog_err("vrf import rt new - def instance not created yet");
137 irt
= XCALLOC(MTYPE_BGP_EVPN_VRF_IMPORT_RT
,
138 sizeof(struct vrf_irt_node
));
143 irt
->vrfs
= list_new();
146 if (!hash_get(bgp_def
->vrf_import_rt_hash
, irt
, hash_alloc_intern
)) {
147 XFREE(MTYPE_BGP_EVPN_VRF_IMPORT_RT
, irt
);
155 * Free the vrf import rt node
157 static void vrf_import_rt_free(struct vrf_irt_node
*irt
)
159 struct bgp
*bgp_def
= NULL
;
161 bgp_def
= bgp_get_default();
163 zlog_err("vrf import rt free - def instance not created yet");
167 hash_release(bgp_def
->vrf_import_rt_hash
, irt
);
168 XFREE(MTYPE_BGP_EVPN_VRF_IMPORT_RT
, irt
);
172 * Function to lookup Import RT node - used to map a RT to set of
173 * VNIs importing routes with that RT.
175 static struct vrf_irt_node
*lookup_vrf_import_rt(struct ecommunity_val
*rt
)
177 struct bgp
*bgp_def
= NULL
;
178 struct vrf_irt_node
*irt
;
179 struct vrf_irt_node tmp
;
181 bgp_def
= bgp_get_default();
183 zlog_err("vrf import rt lookup - def instance not created yet");
187 memset(&tmp
, 0, sizeof(struct vrf_irt_node
));
188 memcpy(&tmp
.rt
, rt
, ECOMMUNITY_SIZE
);
189 irt
= hash_lookup(bgp_def
->vrf_import_rt_hash
, &tmp
);
194 * Is specified VRF present on the RT's list of "importing" VRFs?
196 static int is_vrf_present_in_irt_vrfs(struct list
*vrfs
,
199 struct listnode
*node
= NULL
, *nnode
= NULL
;
200 struct bgp
*tmp_bgp_vrf
= NULL
;
202 for (ALL_LIST_ELEMENTS(vrfs
, node
, nnode
, tmp_bgp_vrf
)) {
203 if (tmp_bgp_vrf
== bgp_vrf
)
210 * Make import route target hash key.
212 static unsigned int import_rt_hash_key_make(void *p
)
214 struct irt_node
*irt
= p
;
215 char *pnt
= irt
->rt
.val
;
217 return jhash(pnt
, 8, 0xdeadbeef);
221 * Comparison function for import rt hash
223 static int import_rt_hash_cmp(const void *p1
, const void *p2
)
225 const struct irt_node
*irt1
= p1
;
226 const struct irt_node
*irt2
= p2
;
228 if (irt1
== NULL
&& irt2
== NULL
)
231 if (irt1
== NULL
|| irt2
== NULL
)
234 return (memcmp(irt1
->rt
.val
, irt2
->rt
.val
, ECOMMUNITY_SIZE
) == 0);
238 * Create a new import_rt
240 static struct irt_node
*import_rt_new(struct bgp
*bgp
,
241 struct ecommunity_val
*rt
)
243 struct irt_node
*irt
;
248 irt
= XCALLOC(MTYPE_BGP_EVPN_IMPORT_RT
, sizeof(struct irt_node
));
253 irt
->vnis
= list_new();
256 if (!hash_get(bgp
->import_rt_hash
, irt
, hash_alloc_intern
)) {
257 XFREE(MTYPE_BGP_EVPN_IMPORT_RT
, irt
);
265 * Free the import rt node
267 static void import_rt_free(struct bgp
*bgp
, struct irt_node
*irt
)
269 hash_release(bgp
->import_rt_hash
, irt
);
270 XFREE(MTYPE_BGP_EVPN_IMPORT_RT
, irt
);
274 * Function to lookup Import RT node - used to map a RT to set of
275 * VNIs importing routes with that RT.
277 static struct irt_node
*lookup_import_rt(struct bgp
*bgp
,
278 struct ecommunity_val
*rt
)
280 struct irt_node
*irt
;
283 memset(&tmp
, 0, sizeof(struct irt_node
));
284 memcpy(&tmp
.rt
, rt
, ECOMMUNITY_SIZE
);
285 irt
= hash_lookup(bgp
->import_rt_hash
, &tmp
);
290 * Is specified VNI present on the RT's list of "importing" VNIs?
292 static int is_vni_present_in_irt_vnis(struct list
*vnis
, struct bgpevpn
*vpn
)
294 struct listnode
*node
, *nnode
;
295 struct bgpevpn
*tmp_vpn
;
297 for (ALL_LIST_ELEMENTS(vnis
, node
, nnode
, tmp_vpn
)) {
306 * Compare Route Targets.
308 static int evpn_route_target_cmp(struct ecommunity
*ecom1
,
309 struct ecommunity
*ecom2
)
317 if (!ecom1
&& !ecom2
)
320 if (ecom1
->str
&& !ecom2
->str
)
323 if (!ecom1
->str
&& ecom2
->str
)
326 if (!ecom1
->str
&& !ecom2
->str
)
329 return strcmp(ecom1
->str
, ecom2
->str
);
333 * Mask off global-admin field of specified extended community (RT),
334 * just retain the local-admin field.
336 static inline void mask_ecom_global_admin(struct ecommunity_val
*dst
,
337 struct ecommunity_val
*src
)
343 if (type
== ECOMMUNITY_ENCODE_AS
) {
344 dst
->val
[2] = dst
->val
[3] = 0;
345 } else if (type
== ECOMMUNITY_ENCODE_AS4
346 || type
== ECOMMUNITY_ENCODE_IP
) {
347 dst
->val
[2] = dst
->val
[3] = 0;
348 dst
->val
[4] = dst
->val
[5] = 0;
353 * Map one RT to specified VRF.
354 * bgp_vrf = BGP vrf instance
356 static void map_vrf_to_rt(struct bgp
*bgp_vrf
,
357 struct ecommunity_val
*eval
)
359 struct vrf_irt_node
*irt
= NULL
;
360 struct ecommunity_val eval_tmp
;
362 /* If using "automatic" RT,
363 * we only care about the local-admin sub-field.
364 * This is to facilitate using L3VNI(VRF-VNI)
365 * as the RT for EBGP peering too.
367 memcpy(&eval_tmp
, eval
, ECOMMUNITY_SIZE
);
368 if (!CHECK_FLAG(bgp_vrf
->vrf_flags
,
369 BGP_VRF_IMPORT_RT_CFGD
))
370 mask_ecom_global_admin(&eval_tmp
, eval
);
372 irt
= lookup_vrf_import_rt(&eval_tmp
);
373 if (irt
&& irt
->vrfs
)
374 if (is_vrf_present_in_irt_vrfs(irt
->vrfs
, bgp_vrf
))
375 /* Already mapped. */
379 irt
= vrf_import_rt_new(&eval_tmp
);
383 /* Add VRF to the list for this RT. */
384 listnode_add(irt
->vrfs
, bgp_vrf
);
388 * Unmap specified VRF from specified RT. If there are no other
389 * VRFs for this RT, then the RT hash is deleted.
390 * bgp_vrf: BGP VRF specific instance
392 static void unmap_vrf_from_rt(struct bgp
*bgp_vrf
,
393 struct vrf_irt_node
*irt
)
395 /* Delete VRF from list for this RT. */
396 listnode_delete(irt
->vrfs
, bgp_vrf
);
397 if (!listnode_head(irt
->vrfs
)) {
398 list_delete_and_null(&irt
->vrfs
);
399 vrf_import_rt_free(irt
);
404 * Map one RT to specified VNI.
406 static void map_vni_to_rt(struct bgp
*bgp
, struct bgpevpn
*vpn
,
407 struct ecommunity_val
*eval
)
409 struct irt_node
*irt
;
410 struct ecommunity_val eval_tmp
;
412 /* If using "automatic" RT, we only care about the local-admin
414 * This is to facilitate using VNI as the RT for EBGP peering too.
416 memcpy(&eval_tmp
, eval
, ECOMMUNITY_SIZE
);
417 if (!is_import_rt_configured(vpn
))
418 mask_ecom_global_admin(&eval_tmp
, eval
);
420 irt
= lookup_import_rt(bgp
, &eval_tmp
);
421 if (irt
&& irt
->vnis
)
422 if (is_vni_present_in_irt_vnis(irt
->vnis
, vpn
))
423 /* Already mapped. */
427 irt
= import_rt_new(bgp
, &eval_tmp
);
431 /* Add VNI to the hash list for this RT. */
432 listnode_add(irt
->vnis
, vpn
);
436 * Unmap specified VNI from specified RT. If there are no other
437 * VNIs for this RT, then the RT hash is deleted.
439 static void unmap_vni_from_rt(struct bgp
*bgp
, struct bgpevpn
*vpn
,
440 struct irt_node
*irt
)
442 /* Delete VNI from hash list for this RT. */
443 listnode_delete(irt
->vnis
, vpn
);
444 if (!listnode_head(irt
->vnis
)) {
445 list_delete_and_null(&irt
->vnis
);
446 import_rt_free(bgp
, irt
);
451 * Create RT extended community automatically from passed information:
452 * of the form AS:VNI.
453 * NOTE: We use only the lower 16 bits of the AS. This is sufficient as
454 * the need is to get a RT value that will be unique across different
455 * VNIs but the same across routers (in the same AS) for a particular
458 static void form_auto_rt(struct bgp
*bgp
, vni_t vni
, struct list
*rtl
)
460 struct ecommunity_val eval
;
461 struct ecommunity
*ecomadd
;
463 encode_route_target_as((bgp
->as
& 0xFFFF), vni
, &eval
);
465 ecomadd
= ecommunity_new();
466 ecommunity_add_val(ecomadd
, &eval
);
467 listnode_add_sort(rtl
, ecomadd
);
471 * Derive RD and RT for a VNI automatically. Invoked at the time of
474 static void derive_rd_rt_for_vni(struct bgp
*bgp
, struct bgpevpn
*vpn
)
476 bgp_evpn_derive_auto_rd(bgp
, vpn
);
477 bgp_evpn_derive_auto_rt_import(bgp
, vpn
);
478 bgp_evpn_derive_auto_rt_export(bgp
, vpn
);
482 * Add (update) or delete MACIP from zebra.
484 static int bgp_zebra_send_remote_macip(struct bgp
*bgp
, struct bgpevpn
*vpn
,
485 struct prefix_evpn
*p
,
486 struct in_addr remote_vtep_ip
, int add
,
491 char buf1
[ETHER_ADDR_STRLEN
];
492 char buf2
[INET6_ADDRSTRLEN
];
493 char buf3
[INET6_ADDRSTRLEN
];
496 if (!zclient
|| zclient
->sock
< 0)
499 /* Don't try to register if Zebra doesn't know of this instance. */
500 if (!IS_BGP_INST_KNOWN_TO_ZEBRA(bgp
))
506 zclient_create_header(s
, add
? ZEBRA_REMOTE_MACIP_ADD
507 : ZEBRA_REMOTE_MACIP_DEL
,
509 stream_putl(s
, vpn
->vni
);
510 stream_put(s
, &p
->prefix
.mac
.octet
, ETH_ALEN
); /* Mac Addr */
511 /* IP address length and IP address, if any. */
512 if (IS_EVPN_PREFIX_IPADDR_NONE(p
))
515 ipa_len
= IS_EVPN_PREFIX_IPADDR_V4(p
) ? IPV4_MAX_BYTELEN
517 stream_putl(s
, ipa_len
);
518 stream_put(s
, &p
->prefix
.ip
.ip
.addr
, ipa_len
);
520 stream_put_in_addr(s
, &remote_vtep_ip
);
522 /* TX flags - MAC sticky status and/or gateway mac */
524 stream_putc(s
, flags
);
526 stream_putw_at(s
, 0, stream_get_endp(s
));
528 if (bgp_debug_zebra(NULL
))
529 zlog_debug("Tx %s MACIP, VNI %u MAC %s IP %s (flags: 0x%x) remote VTEP %s",
530 add
? "ADD" : "DEL", vpn
->vni
,
531 prefix_mac2str(&p
->prefix
.mac
, buf1
, sizeof(buf1
)),
532 ipaddr2str(&p
->prefix
.ip
, buf3
, sizeof(buf3
)),
534 inet_ntop(AF_INET
, &remote_vtep_ip
, buf2
,
537 return zclient_send_message(zclient
);
541 * Add (update) or delete remote VTEP from zebra.
543 static int bgp_zebra_send_remote_vtep(struct bgp
*bgp
, struct bgpevpn
*vpn
,
544 struct prefix_evpn
*p
, int add
)
549 if (!zclient
|| zclient
->sock
< 0)
552 /* Don't try to register if Zebra doesn't know of this instance. */
553 if (!IS_BGP_INST_KNOWN_TO_ZEBRA(bgp
))
559 zclient_create_header(s
, add
? ZEBRA_REMOTE_VTEP_ADD
560 : ZEBRA_REMOTE_VTEP_DEL
,
562 stream_putl(s
, vpn
->vni
);
563 if (IS_EVPN_PREFIX_IPADDR_V4(p
))
564 stream_put_in_addr(s
, &p
->prefix
.ip
.ipaddr_v4
);
565 else if (IS_EVPN_PREFIX_IPADDR_V6(p
)) {
567 "Bad remote IP when trying to %s remote VTEP for VNI %u",
568 add
? "ADD" : "DEL", vpn
->vni
);
572 stream_putw_at(s
, 0, stream_get_endp(s
));
574 if (bgp_debug_zebra(NULL
))
575 zlog_debug("Tx %s Remote VTEP, VNI %u remote VTEP %s",
576 add
? "ADD" : "DEL", vpn
->vni
,
577 inet_ntoa(p
->prefix
.ip
.ipaddr_v4
));
579 return zclient_send_message(zclient
);
583 * Build extended communities for EVPN prefix route.
585 static void build_evpn_type5_route_extcomm(struct bgp
*bgp_vrf
,
588 struct ecommunity ecom_encap
;
589 struct ecommunity ecom_rmac
;
590 struct ecommunity_val eval
;
591 struct ecommunity_val eval_rmac
;
592 bgp_encap_types tnl_type
;
593 struct listnode
*node
, *nnode
;
594 struct ecommunity
*ecom
;
595 struct list
*vrf_export_rtl
= NULL
;
598 tnl_type
= BGP_ENCAP_TYPE_VXLAN
;
599 memset(&ecom_encap
, 0, sizeof(ecom_encap
));
600 encode_encap_extcomm(tnl_type
, &eval
);
602 ecom_encap
.val
= (u_int8_t
*)eval
.val
;
605 attr
->ecommunity
= ecommunity_dup(&ecom_encap
);
607 /* Add the export RTs for L3VNI/VRF */
608 vrf_export_rtl
= bgp_vrf
->vrf_export_rtl
;
609 if (vrf_export_rtl
&& !list_isempty(vrf_export_rtl
)) {
610 for (ALL_LIST_ELEMENTS(vrf_export_rtl
, node
, nnode
, ecom
))
611 attr
->ecommunity
= ecommunity_merge(attr
->ecommunity
,
615 /* add the router mac extended community */
616 if (!is_zero_mac(&attr
->rmac
)) {
617 memset(&ecom_rmac
, 0, sizeof(ecom_rmac
));
618 encode_rmac_extcomm(&eval_rmac
, &attr
->rmac
);
620 ecom_rmac
.val
= (uint8_t *)eval_rmac
.val
;
621 attr
->ecommunity
= ecommunity_merge(attr
->ecommunity
,
625 attr
->flag
|= ATTR_FLAG_BIT(BGP_ATTR_EXT_COMMUNITIES
);
629 * Build extended communities for EVPN route. RT and ENCAP are
630 * applicable to all routes.
631 * TODO: currently kernel doesnt support ipv6 routes with ipv4 nexthops.
632 * This means that we can't do symmetric routing for ipv6 hosts routes
633 * in the same way as ipv4 host routes.
634 * We wont attach l3-vni related RTs for ipv6 routes.
635 * For now, We will only adevrtise ipv4 host routes
636 * with L3-VNI related ext-comm.
638 static void build_evpn_route_extcomm(struct bgpevpn
*vpn
, struct attr
*attr
,
641 struct ecommunity ecom_encap
;
642 struct ecommunity ecom_sticky
;
643 struct ecommunity ecom_default_gw
;
644 struct ecommunity ecom_rmac
;
645 struct ecommunity_val eval
;
646 struct ecommunity_val eval_sticky
;
647 struct ecommunity_val eval_default_gw
;
648 struct ecommunity_val eval_rmac
;
649 bgp_encap_types tnl_type
;
650 struct listnode
*node
, *nnode
;
651 struct ecommunity
*ecom
;
653 struct list
*vrf_export_rtl
= NULL
;
656 tnl_type
= BGP_ENCAP_TYPE_VXLAN
;
657 memset(&ecom_encap
, 0, sizeof(ecom_encap
));
658 encode_encap_extcomm(tnl_type
, &eval
);
660 ecom_encap
.val
= (u_int8_t
*)eval
.val
;
663 attr
->ecommunity
= ecommunity_dup(&ecom_encap
);
665 /* Add the export RTs for L2VNI */
666 for (ALL_LIST_ELEMENTS(vpn
->export_rtl
, node
, nnode
, ecom
))
667 attr
->ecommunity
= ecommunity_merge(attr
->ecommunity
, ecom
);
670 * only attach l3-vni export rts for ipv4 address family and if we are
671 * advertising both the labels in type-2 routes
673 if (afi
== AFI_IP
&& CHECK_FLAG(vpn
->flags
, VNI_FLAG_USE_TWO_LABELS
)) {
674 vrf_export_rtl
= bgpevpn_get_vrf_export_rtl(vpn
);
675 if (vrf_export_rtl
&& !list_isempty(vrf_export_rtl
)) {
676 for (ALL_LIST_ELEMENTS(vrf_export_rtl
, node
, nnode
,
679 ecommunity_merge(attr
->ecommunity
,
686 memset(&ecom_sticky
, 0, sizeof(ecom_sticky
));
687 encode_mac_mobility_extcomm(1, seqnum
, &eval_sticky
);
688 ecom_sticky
.size
= 1;
689 ecom_sticky
.val
= (u_int8_t
*)eval_sticky
.val
;
691 ecommunity_merge(attr
->ecommunity
, &ecom_sticky
);
695 * only attach l3-vni rmac for ipv4 address family and if we are
696 * advertising both the labels in type-2 routes
698 if (afi
== AFI_IP
&& !is_zero_mac(&attr
->rmac
) &&
699 CHECK_FLAG(vpn
->flags
, VNI_FLAG_USE_TWO_LABELS
)) {
700 memset(&ecom_rmac
, 0, sizeof(ecom_rmac
));
701 encode_rmac_extcomm(&eval_rmac
, &attr
->rmac
);
703 ecom_rmac
.val
= (uint8_t *)eval_rmac
.val
;
704 attr
->ecommunity
= ecommunity_merge(attr
->ecommunity
,
708 if (attr
->default_gw
) {
709 memset(&ecom_default_gw
, 0, sizeof(ecom_default_gw
));
710 encode_default_gw_extcomm(&eval_default_gw
);
711 ecom_default_gw
.size
= 1;
712 ecom_default_gw
.val
= (uint8_t *)eval_default_gw
.val
;
713 attr
->ecommunity
= ecommunity_merge(attr
->ecommunity
,
717 attr
->flag
|= ATTR_FLAG_BIT(BGP_ATTR_EXT_COMMUNITIES
);
721 * Add MAC mobility extended community to attribute.
723 static void add_mac_mobility_to_attr(u_int32_t seq_num
, struct attr
*attr
)
725 struct ecommunity ecom_tmp
;
726 struct ecommunity_val eval
;
727 u_int8_t
*ecom_val_ptr
;
734 encode_mac_mobility_extcomm(0, seq_num
, &eval
);
736 /* Find current MM ecommunity */
739 if (attr
->ecommunity
) {
740 for (i
= 0; i
< attr
->ecommunity
->size
; i
++) {
741 pnt
= attr
->ecommunity
->val
+ (i
* 8);
745 if (type
== ECOMMUNITY_ENCODE_EVPN
747 == ECOMMUNITY_EVPN_SUBTYPE_MACMOBILITY
) {
749 (u_int8_t
*)(attr
->ecommunity
->val
756 /* Update the existing MM ecommunity */
758 memcpy(ecom_val_ptr
, eval
.val
, sizeof(char) * ECOMMUNITY_SIZE
);
760 /* Add MM to existing */
762 memset(&ecom_tmp
, 0, sizeof(ecom_tmp
));
764 ecom_tmp
.val
= (u_int8_t
*)eval
.val
;
767 ecommunity_merge(attr
->ecommunity
, &ecom_tmp
);
771 /* Install EVPN route into zebra. */
772 static int evpn_zebra_install(struct bgp
*bgp
, struct bgpevpn
*vpn
,
773 struct prefix_evpn
*p
,
774 struct in_addr remote_vtep_ip
, u_char flags
)
778 if (p
->prefix
.route_type
== BGP_EVPN_MAC_IP_ROUTE
)
779 ret
= bgp_zebra_send_remote_macip(bgp
, vpn
, p
, remote_vtep_ip
,
782 ret
= bgp_zebra_send_remote_vtep(bgp
, vpn
, p
, 1);
787 /* Uninstall EVPN route from zebra. */
788 static int evpn_zebra_uninstall(struct bgp
*bgp
, struct bgpevpn
*vpn
,
789 struct prefix_evpn
*p
,
790 struct in_addr remote_vtep_ip
)
794 if (p
->prefix
.route_type
== BGP_EVPN_MAC_IP_ROUTE
)
795 ret
= bgp_zebra_send_remote_macip(bgp
, vpn
, p
, remote_vtep_ip
,
798 ret
= bgp_zebra_send_remote_vtep(bgp
, vpn
, p
, 0);
804 * Due to MAC mobility, the prior "local" best route has been supplanted
805 * by a "remote" best route. The prior route has to be deleted and withdrawn
808 static void evpn_delete_old_local_route(struct bgp
*bgp
, struct bgpevpn
*vpn
,
810 struct bgp_info
*old_local
)
812 struct bgp_node
*global_rn
;
814 afi_t afi
= AFI_L2VPN
;
815 safi_t safi
= SAFI_EVPN
;
817 /* Locate route node in the global EVPN routing table. Note that
818 * this table is a 2-level tree (RD-level + Prefix-level) similar to
821 global_rn
= bgp_afi_node_lookup(bgp
->rib
[afi
][safi
], afi
, safi
,
822 (struct prefix
*)&rn
->p
, &vpn
->prd
);
824 /* Delete route entry in the global EVPN table. */
825 delete_evpn_route_entry(bgp
, vpn
, afi
, safi
, global_rn
, &ri
);
827 /* Schedule for processing - withdraws to peers happen from
831 bgp_process(bgp
, global_rn
, afi
, safi
);
832 bgp_unlock_node(global_rn
);
835 /* Delete route entry in the VNI route table, caller to remove. */
836 bgp_info_delete(rn
, old_local
);
840 * Calculate the best path for an EVPN route. Install/update best path in zebra,
843 static int evpn_route_select_install(struct bgp
*bgp
, struct bgpevpn
*vpn
,
846 struct bgp_info
*old_select
, *new_select
;
847 struct bgp_info_pair old_and_new
;
848 afi_t afi
= AFI_L2VPN
;
849 safi_t safi
= SAFI_EVPN
;
853 /* Compute the best path. */
854 bgp_best_selection(bgp
, rn
, &bgp
->maxpaths
[afi
][safi
], &old_and_new
,
856 old_select
= old_and_new
.old
;
857 new_select
= old_and_new
.new;
859 /* If the best path hasn't changed - see if there is still something to
863 if (old_select
&& old_select
== new_select
864 && old_select
->type
== ZEBRA_ROUTE_BGP
865 && old_select
->sub_type
== BGP_ROUTE_NORMAL
866 && !CHECK_FLAG(rn
->flags
, BGP_NODE_USER_CLEAR
)
867 && !CHECK_FLAG(old_select
->flags
, BGP_INFO_ATTR_CHANGED
)
868 && !bgp
->addpath_tx_used
[afi
][safi
]) {
869 if (bgp_zebra_has_route_changed(rn
, old_select
)) {
870 if (old_select
->attr
->sticky
)
871 SET_FLAG(flags
, ZEBRA_MACIP_TYPE_STICKY
);
872 if (old_select
->attr
->default_gw
)
873 SET_FLAG(flags
, ZEBRA_MACIP_TYPE_GW
);
874 ret
= evpn_zebra_install(bgp
, vpn
,
875 (struct prefix_evpn
*)&rn
->p
,
876 old_select
->attr
->nexthop
,
879 UNSET_FLAG(old_select
->flags
, BGP_INFO_MULTIPATH_CHG
);
880 bgp_zebra_clear_route_change_flags(rn
);
884 /* If the user did a "clear" this flag will be set */
885 UNSET_FLAG(rn
->flags
, BGP_NODE_USER_CLEAR
);
887 /* bestpath has changed; update relevant fields and install or uninstall
888 * into the zebra RIB.
890 if (old_select
|| new_select
)
891 bgp_bump_version(rn
);
894 bgp_info_unset_flag(rn
, old_select
, BGP_INFO_SELECTED
);
896 bgp_info_set_flag(rn
, new_select
, BGP_INFO_SELECTED
);
897 bgp_info_unset_flag(rn
, new_select
, BGP_INFO_ATTR_CHANGED
);
898 UNSET_FLAG(new_select
->flags
, BGP_INFO_MULTIPATH_CHG
);
901 if (new_select
&& new_select
->type
== ZEBRA_ROUTE_BGP
902 && new_select
->sub_type
== BGP_ROUTE_NORMAL
) {
904 if (new_select
->attr
->sticky
)
905 SET_FLAG(flags
, ZEBRA_MACIP_TYPE_STICKY
);
906 if (new_select
->attr
->default_gw
)
907 SET_FLAG(flags
, ZEBRA_MACIP_TYPE_GW
);
908 ret
= evpn_zebra_install(bgp
, vpn
, (struct prefix_evpn
*)&rn
->p
,
909 new_select
->attr
->nexthop
,
911 /* If an old best existed and it was a "local" route, the only
913 * it would be supplanted is due to MAC mobility procedures. So,
915 * need to do an implicit delete and withdraw that route from
918 if (old_select
&& old_select
->peer
== bgp
->peer_self
919 && old_select
->type
== ZEBRA_ROUTE_BGP
920 && old_select
->sub_type
== BGP_ROUTE_STATIC
)
921 evpn_delete_old_local_route(bgp
, vpn
, rn
, old_select
);
923 if (old_select
&& old_select
->type
== ZEBRA_ROUTE_BGP
924 && old_select
->sub_type
== BGP_ROUTE_NORMAL
)
925 ret
= evpn_zebra_uninstall(bgp
, vpn
,
926 (struct prefix_evpn
*)&rn
->p
,
927 old_select
->attr
->nexthop
);
930 /* Clear any route change flags. */
931 bgp_zebra_clear_route_change_flags(rn
);
933 /* Reap old select bgp_info, if it has been removed */
934 if (old_select
&& CHECK_FLAG(old_select
->flags
, BGP_INFO_REMOVED
))
935 bgp_info_reap(rn
, old_select
);
941 * Return true if the local ri for this rn is of type gateway mac
943 static int evpn_route_is_def_gw(struct bgp
*bgp
, struct bgp_node
*rn
)
945 struct bgp_info
*tmp_ri
= NULL
;
946 struct bgp_info
*local_ri
= NULL
;
949 for (tmp_ri
= rn
->info
; tmp_ri
; tmp_ri
= tmp_ri
->next
) {
950 if (tmp_ri
->peer
== bgp
->peer_self
951 && tmp_ri
->type
== ZEBRA_ROUTE_BGP
952 && tmp_ri
->sub_type
== BGP_ROUTE_STATIC
)
959 return local_ri
->attr
->default_gw
;
964 * Return true if the local ri for this rn has sticky set
966 static int evpn_route_is_sticky(struct bgp
*bgp
, struct bgp_node
*rn
)
968 struct bgp_info
*tmp_ri
;
969 struct bgp_info
*local_ri
;
972 for (tmp_ri
= rn
->info
; tmp_ri
; tmp_ri
= tmp_ri
->next
) {
973 if (tmp_ri
->peer
== bgp
->peer_self
974 && tmp_ri
->type
== ZEBRA_ROUTE_BGP
975 && tmp_ri
->sub_type
== BGP_ROUTE_STATIC
)
982 return local_ri
->attr
->sticky
;
985 static int update_evpn_type5_route_entry(struct bgp
*bgp_def
,
986 struct bgp
*bgp_vrf
, afi_t afi
,
987 safi_t safi
, struct bgp_node
*rn
,
988 struct attr
*attr
, int *route_changed
)
990 struct attr
*attr_new
= NULL
;
991 struct bgp_info
*ri
= NULL
;
992 mpls_label_t label
= MPLS_INVALID_LABEL
;
993 struct bgp_info
*local_ri
= NULL
;
994 struct bgp_info
*tmp_ri
= NULL
;
997 /* locate the local route entry if any */
998 for (tmp_ri
= rn
->info
; tmp_ri
; tmp_ri
= tmp_ri
->next
) {
999 if (tmp_ri
->peer
== bgp_def
->peer_self
1000 && tmp_ri
->type
== ZEBRA_ROUTE_BGP
1001 && tmp_ri
->sub_type
== BGP_ROUTE_STATIC
)
1005 /* create a new route entry if one doesnt exist.
1006 Otherwise see if route attr has changed
1010 /* route has changed as this is the first entry */
1013 /* Add (or update) attribute to hash. */
1014 attr_new
= bgp_attr_intern(attr
);
1016 /* create the route info from attribute */
1017 ri
= info_make(ZEBRA_ROUTE_BGP
, BGP_ROUTE_STATIC
, 0,
1018 bgp_def
->peer_self
, attr_new
, rn
);
1019 SET_FLAG(ri
->flags
, BGP_INFO_VALID
);
1021 /* Type-5 routes advertise the L3-VNI */
1022 bgp_info_extra_get(ri
);
1023 vni2label(bgp_vrf
->l3vni
, &label
);
1024 memcpy(&ri
->extra
->label
, &label
, sizeof(label
));
1025 ri
->extra
->num_labels
= 1;
1027 /* add the route entry to route node*/
1028 bgp_info_add(rn
, ri
);
1032 if (!attrhash_cmp(tmp_ri
->attr
, attr
)) {
1034 /* attribute changed */
1037 /* The attribute has changed. */
1038 /* Add (or update) attribute to hash. */
1039 attr_new
= bgp_attr_intern(attr
);
1040 bgp_info_set_flag(rn
, tmp_ri
, BGP_INFO_ATTR_CHANGED
);
1042 /* Restore route, if needed. */
1043 if (CHECK_FLAG(tmp_ri
->flags
, BGP_INFO_REMOVED
))
1044 bgp_info_restore(rn
, tmp_ri
);
1046 /* Unintern existing, set to new. */
1047 bgp_attr_unintern(&tmp_ri
->attr
);
1048 tmp_ri
->attr
= attr_new
;
1049 tmp_ri
->uptime
= bgp_clock();
1055 /* update evpn type-5 route entry */
1056 static int update_evpn_type5_route(struct bgp
*bgp_vrf
,
1057 struct prefix_evpn
*evp
,
1058 struct attr
* src_attr
)
1060 afi_t afi
= AFI_L2VPN
;
1061 safi_t safi
= SAFI_EVPN
;
1063 struct bgp_node
*rn
= NULL
;
1064 struct bgp
*bgp_def
= NULL
;
1065 int route_changed
= 0;
1067 bgp_def
= bgp_get_default();
1071 /* Build path attribute for this route - use the source attr, if
1072 * present, else treat as locally originated.
1075 bgp_attr_dup(&attr
, src_attr
);
1077 memset(&attr
, 0, sizeof(struct attr
));
1078 bgp_attr_default_set(&attr
, BGP_ORIGIN_IGP
);
1080 /* Set nexthop to ourselves and fill in the Router MAC. */
1081 attr
.nexthop
= bgp_vrf
->originator_ip
;
1082 attr
.mp_nexthop_global_in
= bgp_vrf
->originator_ip
;
1083 attr
.mp_nexthop_len
= BGP_ATTR_NHLEN_IPV4
;
1084 memcpy(&attr
.rmac
, &bgp_vrf
->rmac
, sizeof(struct ethaddr
));
1086 /* Setup RT and encap extended community */
1087 build_evpn_type5_route_extcomm(bgp_vrf
, &attr
);
1089 /* get the route node in global table */
1090 rn
= bgp_afi_node_get(bgp_def
->rib
[afi
][safi
], afi
, safi
,
1091 (struct prefix
*)evp
,
1095 /* create or update the route entry within the route node */
1096 update_evpn_type5_route_entry(bgp_def
, bgp_vrf
,
1098 rn
, &attr
, &route_changed
);
1100 /* schedule for processing and unlock node */
1101 if (route_changed
) {
1102 bgp_process(bgp_def
, rn
, afi
, safi
);
1103 bgp_unlock_node(rn
);
1106 /* uninten temporary */
1108 aspath_unintern(&attr
.aspath
);
1113 * Create or update EVPN route entry. This could be in the VNI route table
1114 * or the global route table.
1116 static int update_evpn_route_entry(struct bgp
*bgp
, struct bgpevpn
*vpn
,
1117 afi_t afi
, safi_t safi
, struct bgp_node
*rn
,
1118 struct attr
*attr
, int add
, int vni_table
,
1119 struct bgp_info
**ri
, u_char flags
)
1121 struct bgp_info
*tmp_ri
;
1122 struct bgp_info
*local_ri
, *remote_ri
;
1123 struct attr
*attr_new
;
1124 mpls_label_t label
[BGP_MAX_LABELS
];
1125 u_int32_t num_labels
= 1;
1126 int route_change
= 1;
1128 struct prefix_evpn
*evp
;
1131 evp
= (struct prefix_evpn
*)&rn
->p
;
1132 memset(&label
, 0, sizeof(label
));
1134 /* See if this is an update of an existing route, or a new add. Also,
1135 * identify if already known from remote, and if so, the one with the
1136 * highest sequence number; this is only when adding to the VNI routing
1139 local_ri
= remote_ri
= NULL
;
1140 for (tmp_ri
= rn
->info
; tmp_ri
; tmp_ri
= tmp_ri
->next
) {
1141 if (tmp_ri
->peer
== bgp
->peer_self
1142 && tmp_ri
->type
== ZEBRA_ROUTE_BGP
1143 && tmp_ri
->sub_type
== BGP_ROUTE_STATIC
)
1146 if (tmp_ri
->type
== ZEBRA_ROUTE_BGP
1147 && tmp_ri
->sub_type
== BGP_ROUTE_NORMAL
1148 && CHECK_FLAG(tmp_ri
->flags
, BGP_INFO_VALID
)) {
1151 else if (mac_mobility_seqnum(tmp_ri
->attr
)
1152 > mac_mobility_seqnum(remote_ri
->attr
))
1158 /* If route doesn't exist already, create a new one, if told to.
1159 * Otherwise act based on whether the attributes of the route have
1162 if (!local_ri
&& !add
)
1166 /* When learnt locally for the first time but already known from
1167 * remote, we have to initiate appropriate MAC mobility steps.
1169 * is applicable when updating the VNI routing table.
1170 * We need to skip mobility steps for g/w macs (local mac on g/w
1171 * SVI) advertised in EVPN.
1172 * This will ensure that local routes are preferred for g/w macs
1174 if (remote_ri
&& !CHECK_FLAG(flags
, ZEBRA_MACIP_TYPE_GW
)) {
1175 u_int32_t cur_seqnum
;
1177 /* Add MM extended community to route. */
1178 cur_seqnum
= mac_mobility_seqnum(remote_ri
->attr
);
1179 add_mac_mobility_to_attr(cur_seqnum
+ 1, attr
);
1182 /* Add (or update) attribute to hash. */
1183 attr_new
= bgp_attr_intern(attr
);
1185 /* Extract MAC mobility sequence number, if any. */
1186 attr_new
->mm_seqnum
=
1187 bgp_attr_mac_mobility_seqnum(attr_new
, &sticky
);
1188 attr_new
->sticky
= sticky
;
1190 /* Create new route with its attribute. */
1191 tmp_ri
= info_make(ZEBRA_ROUTE_BGP
, BGP_ROUTE_STATIC
, 0,
1192 bgp
->peer_self
, attr_new
, rn
);
1193 SET_FLAG(tmp_ri
->flags
, BGP_INFO_VALID
);
1194 bgp_info_extra_get(tmp_ri
);
1196 /* The VNI goes into the 'label' field of the route */
1197 vni2label(vpn
->vni
, &label
[0]);
1199 /* Type-2 routes may carry a second VNI - the L3-VNI.
1200 * Only attach second label if we are advertising two labels for
1203 if (evp
->prefix
.route_type
== BGP_EVPN_MAC_IP_ROUTE
&&
1204 CHECK_FLAG(vpn
->flags
, VNI_FLAG_USE_TWO_LABELS
)) {
1207 l3vni
= bgpevpn_get_l3vni(vpn
);
1209 vni2label(l3vni
, &label
[1]);
1214 memcpy(&tmp_ri
->extra
->label
, label
, sizeof(label
));
1215 tmp_ri
->extra
->num_labels
= num_labels
;
1216 bgp_info_add(rn
, tmp_ri
);
1219 if (attrhash_cmp(tmp_ri
->attr
, attr
)
1220 && !CHECK_FLAG(tmp_ri
->flags
, BGP_INFO_REMOVED
))
1224 * The attributes have changed, type-2 routes needs to
1225 * be advertised with right labels.
1227 vni2label(vpn
->vni
, &label
[0]);
1228 if (evp
->prefix
.route_type
== BGP_EVPN_MAC_IP_ROUTE
&&
1229 CHECK_FLAG(vpn
->flags
, VNI_FLAG_USE_TWO_LABELS
)) {
1232 l3vni
= bgpevpn_get_l3vni(vpn
);
1234 vni2label(l3vni
, &label
[1]);
1238 memcpy(&tmp_ri
->extra
->label
, label
, sizeof(label
));
1239 tmp_ri
->extra
->num_labels
= num_labels
;
1241 /* The attribute has changed. */
1242 /* Add (or update) attribute to hash. */
1243 attr_new
= bgp_attr_intern(attr
);
1244 bgp_info_set_flag(rn
, tmp_ri
, BGP_INFO_ATTR_CHANGED
);
1246 /* Restore route, if needed. */
1247 if (CHECK_FLAG(tmp_ri
->flags
, BGP_INFO_REMOVED
))
1248 bgp_info_restore(rn
, tmp_ri
);
1250 /* Unintern existing, set to new. */
1251 bgp_attr_unintern(&tmp_ri
->attr
);
1252 tmp_ri
->attr
= attr_new
;
1253 tmp_ri
->uptime
= bgp_clock();
1257 /* Return back the route entry. */
1259 return route_change
;
1263 * Create or update EVPN route (of type based on prefix) for specified VNI
1264 * and schedule for processing.
1266 static int update_evpn_route(struct bgp
*bgp
, struct bgpevpn
*vpn
,
1267 struct prefix_evpn
*p
, u_char flags
)
1269 struct bgp_node
*rn
;
1271 struct attr
*attr_new
;
1272 struct bgp_info
*ri
;
1273 afi_t afi
= AFI_L2VPN
;
1274 safi_t safi
= SAFI_EVPN
;
1277 memset(&attr
, 0, sizeof(struct attr
));
1279 /* Build path-attribute for this route. */
1280 bgp_attr_default_set(&attr
, BGP_ORIGIN_IGP
);
1281 attr
.nexthop
= vpn
->originator_ip
;
1282 attr
.mp_nexthop_global_in
= vpn
->originator_ip
;
1283 attr
.mp_nexthop_len
= BGP_ATTR_NHLEN_IPV4
;
1284 attr
.sticky
= CHECK_FLAG(flags
, ZEBRA_MACIP_TYPE_STICKY
) ? 1 : 0;
1285 attr
.default_gw
= CHECK_FLAG(flags
, ZEBRA_MACIP_TYPE_GW
) ? 1 : 0;
1287 /* PMSI is only needed for type-3 routes */
1288 if (p
->prefix
.route_type
== BGP_EVPN_IMET_ROUTE
)
1289 attr
.flag
|= ATTR_FLAG_BIT(BGP_ATTR_PMSI_TUNNEL
);
1291 /* router mac is only needed for type-2 and type-5 routes */
1292 if (p
->prefix
.route_type
== BGP_EVPN_MAC_IP_ROUTE
)
1293 bgpevpn_get_rmac(vpn
, &attr
.rmac
);
1294 vni2label(vpn
->vni
, &(attr
.label
));
1296 /* Set up RT and ENCAP extended community. */
1297 build_evpn_route_extcomm(vpn
, &attr
,
1298 IS_EVPN_PREFIX_IPADDR_V4(p
) ?
1301 /* First, create (or fetch) route node within the VNI. */
1302 /* NOTE: There is no RD here. */
1303 rn
= bgp_node_get(vpn
->route_table
, (struct prefix
*)p
);
1305 /* Create or update route entry. */
1306 route_change
= update_evpn_route_entry(bgp
, vpn
, afi
, safi
, rn
, &attr
,
1309 attr_new
= ri
->attr
;
1311 /* Perform route selection; this is just to set the flags correctly
1312 * as local route in the VNI always wins.
1314 evpn_route_select_install(bgp
, vpn
, rn
);
1315 bgp_unlock_node(rn
);
1317 /* If this is a new route or some attribute has changed, export the
1318 * route to the global table. The route will be advertised to peers
1319 * from there. Note that this table is a 2-level tree (RD-level +
1320 * Prefix-level) similar to L3VPN routes.
1323 struct bgp_info
*global_ri
;
1325 rn
= bgp_afi_node_get(bgp
->rib
[afi
][safi
], afi
, safi
,
1326 (struct prefix
*)p
, &vpn
->prd
);
1327 update_evpn_route_entry(bgp
, vpn
, afi
, safi
, rn
, attr_new
, 1, 0,
1330 /* Schedule for processing and unlock node. */
1331 bgp_process(bgp
, rn
, afi
, safi
);
1332 bgp_unlock_node(rn
);
1335 /* Unintern temporary. */
1336 aspath_unintern(&attr
.aspath
);
1341 /* Delete EVPN type5 route entry from global table */
1342 static void delete_evpn_type5_route_entry(struct bgp
*bgp_def
,
1343 struct bgp
*bgp_vrf
,
1344 afi_t afi
, safi_t safi
,
1345 struct bgp_node
*rn
,
1346 struct bgp_info
**ri
)
1348 struct bgp_info
*tmp_ri
= NULL
;
1352 /* find the matching route entry */
1353 for (tmp_ri
= rn
->info
; tmp_ri
; tmp_ri
= tmp_ri
->next
)
1354 if (tmp_ri
->peer
== bgp_def
->peer_self
1355 && tmp_ri
->type
== ZEBRA_ROUTE_BGP
1356 && tmp_ri
->sub_type
== BGP_ROUTE_STATIC
)
1361 /* Mark route for delete. */
1363 bgp_info_delete(rn
, tmp_ri
);
1366 /* Delete EVPN type5 route */
1367 static int delete_evpn_type5_route(struct bgp
*bgp_vrf
,
1368 struct prefix_evpn
*evp
)
1370 afi_t afi
= AFI_L2VPN
;
1371 safi_t safi
= SAFI_EVPN
;
1372 struct bgp_node
*rn
= NULL
;
1373 struct bgp_info
*ri
= NULL
;
1374 struct bgp
*bgp_def
= NULL
; /* default bgp instance */
1376 bgp_def
= bgp_get_default();
1380 /* locate the global route entry for this type-5 prefix */
1381 rn
= bgp_afi_node_lookup(bgp_def
->rib
[afi
][safi
], afi
, safi
,
1382 (struct prefix
*)evp
, &bgp_vrf
->vrf_prd
);
1386 delete_evpn_type5_route_entry(bgp_def
, bgp_vrf
, afi
, safi
, rn
, &ri
);
1388 bgp_process(bgp_def
, rn
, afi
, safi
);
1389 bgp_unlock_node(rn
);
1394 * Delete EVPN route entry. This could be in the VNI route table
1395 * or the global route table.
1397 static void delete_evpn_route_entry(struct bgp
*bgp
, struct bgpevpn
*vpn
,
1398 afi_t afi
, safi_t safi
, struct bgp_node
*rn
,
1399 struct bgp_info
**ri
)
1401 struct bgp_info
*tmp_ri
;
1405 /* Now, find matching route. */
1406 for (tmp_ri
= rn
->info
; tmp_ri
; tmp_ri
= tmp_ri
->next
)
1407 if (tmp_ri
->peer
== bgp
->peer_self
1408 && tmp_ri
->type
== ZEBRA_ROUTE_BGP
1409 && tmp_ri
->sub_type
== BGP_ROUTE_STATIC
)
1414 /* Mark route for delete. */
1416 bgp_info_delete(rn
, tmp_ri
);
1420 * Delete EVPN route (of type based on prefix) for specified VNI and
1421 * schedule for processing.
1423 static int delete_evpn_route(struct bgp
*bgp
, struct bgpevpn
*vpn
,
1424 struct prefix_evpn
*p
)
1426 struct bgp_node
*rn
, *global_rn
;
1427 struct bgp_info
*ri
;
1428 afi_t afi
= AFI_L2VPN
;
1429 safi_t safi
= SAFI_EVPN
;
1431 /* First, locate the route node within the VNI. If it doesn't exist,
1433 * is nothing further to do.
1435 /* NOTE: There is no RD here. */
1436 rn
= bgp_node_lookup(vpn
->route_table
, (struct prefix
*)p
);
1440 /* Next, locate route node in the global EVPN routing table. Note that
1441 * this table is a 2-level tree (RD-level + Prefix-level) similar to
1444 global_rn
= bgp_afi_node_lookup(bgp
->rib
[afi
][safi
], afi
, safi
,
1445 (struct prefix
*)p
, &vpn
->prd
);
1447 /* Delete route entry in the global EVPN table. */
1448 delete_evpn_route_entry(bgp
, vpn
, afi
, safi
, global_rn
, &ri
);
1450 /* Schedule for processing - withdraws to peers happen from
1454 bgp_process(bgp
, global_rn
, afi
, safi
);
1455 bgp_unlock_node(global_rn
);
1458 /* Delete route entry in the VNI route table. This can just be removed.
1460 delete_evpn_route_entry(bgp
, vpn
, afi
, safi
, rn
, &ri
);
1462 bgp_info_reap(rn
, ri
);
1463 bgp_unlock_node(rn
);
1469 * Update all type-2 (MACIP) local routes for this VNI - these should also
1470 * be scheduled for advertise to peers.
1472 static int update_all_type2_routes(struct bgp
*bgp
, struct bgpevpn
*vpn
)
1476 struct bgp_node
*rn
;
1477 struct bgp_info
*ri
;
1479 struct attr attr_sticky
;
1480 struct attr attr_def_gw
;
1481 struct attr attr_ip6
;
1482 struct attr attr_sticky_ip6
;
1483 struct attr attr_def_gw_ip6
;
1484 struct attr
*attr_new
;
1488 memset(&attr
, 0, sizeof(struct attr
));
1489 memset(&attr_sticky
, 0, sizeof(struct attr
));
1490 memset(&attr_def_gw
, 0, sizeof(struct attr
));
1491 memset(&attr_ip6
, 0, sizeof(struct attr
));
1492 memset(&attr_sticky_ip6
, 0, sizeof(struct attr
));
1493 memset(&attr_def_gw_ip6
, 0, sizeof(struct attr
));
1495 /* Build path-attribute - all type-2 routes for this VNI will share the
1496 * same path attribute.
1498 bgp_attr_default_set(&attr
, BGP_ORIGIN_IGP
);
1499 bgp_attr_default_set(&attr_sticky
, BGP_ORIGIN_IGP
);
1500 bgp_attr_default_set(&attr_def_gw
, BGP_ORIGIN_IGP
);
1501 attr
.nexthop
= vpn
->originator_ip
;
1502 attr
.mp_nexthop_global_in
= vpn
->originator_ip
;
1503 attr
.mp_nexthop_len
= BGP_ATTR_NHLEN_IPV4
;
1504 bgpevpn_get_rmac(vpn
, &attr
.rmac
);
1505 attr_sticky
.nexthop
= vpn
->originator_ip
;
1506 attr_sticky
.mp_nexthop_global_in
= vpn
->originator_ip
;
1507 attr_sticky
.mp_nexthop_len
= BGP_ATTR_NHLEN_IPV4
;
1508 attr_sticky
.sticky
= 1;
1509 bgpevpn_get_rmac(vpn
, &attr_sticky
.rmac
);
1510 attr_def_gw
.nexthop
= vpn
->originator_ip
;
1511 attr_def_gw
.mp_nexthop_global_in
= vpn
->originator_ip
;
1512 attr_def_gw
.mp_nexthop_len
= BGP_ATTR_NHLEN_IPV4
;
1513 attr_def_gw
.default_gw
= 1;
1514 bgpevpn_get_rmac(vpn
, &attr_def_gw
.rmac
);
1515 bgp_attr_default_set(&attr_ip6
, BGP_ORIGIN_IGP
);
1516 bgp_attr_default_set(&attr_sticky_ip6
, BGP_ORIGIN_IGP
);
1517 bgp_attr_default_set(&attr_def_gw_ip6
, BGP_ORIGIN_IGP
);
1518 attr_ip6
.nexthop
= vpn
->originator_ip
;
1519 attr_ip6
.mp_nexthop_global_in
= vpn
->originator_ip
;
1520 attr_ip6
.mp_nexthop_len
= BGP_ATTR_NHLEN_IPV4
;
1521 bgpevpn_get_rmac(vpn
, &attr_ip6
.rmac
);
1522 attr_sticky_ip6
.nexthop
= vpn
->originator_ip
;
1523 attr_sticky_ip6
.mp_nexthop_global_in
= vpn
->originator_ip
;
1524 attr_sticky_ip6
.mp_nexthop_len
= BGP_ATTR_NHLEN_IPV4
;
1525 attr_sticky_ip6
.sticky
= 1;
1526 bgpevpn_get_rmac(vpn
, &attr_sticky_ip6
.rmac
);
1527 attr_def_gw_ip6
.nexthop
= vpn
->originator_ip
;
1528 attr_def_gw_ip6
.mp_nexthop_global_in
= vpn
->originator_ip
;
1529 attr_def_gw_ip6
.mp_nexthop_len
= BGP_ATTR_NHLEN_IPV4
;
1530 attr_def_gw_ip6
.default_gw
= 1;
1531 bgpevpn_get_rmac(vpn
, &attr_def_gw_ip6
.rmac
);
1533 /* Set up RT, ENCAP and sticky MAC extended community. */
1534 build_evpn_route_extcomm(vpn
, &attr
, AFI_IP
);
1535 build_evpn_route_extcomm(vpn
, &attr_sticky
, AFI_IP
);
1536 build_evpn_route_extcomm(vpn
, &attr_def_gw
, AFI_IP
);
1537 build_evpn_route_extcomm(vpn
, &attr_ip6
, AFI_IP6
);
1538 build_evpn_route_extcomm(vpn
, &attr_sticky_ip6
, AFI_IP6
);
1539 build_evpn_route_extcomm(vpn
, &attr_def_gw_ip6
, AFI_IP
);
1541 /* Walk this VNI's route table and update local type-2 routes. For any
1542 * routes updated, update corresponding entry in the global table too.
1544 for (rn
= bgp_table_top(vpn
->route_table
); rn
;
1545 rn
= bgp_route_next(rn
)) {
1546 struct prefix_evpn
*evp
= (struct prefix_evpn
*)&rn
->p
;
1547 struct bgp_node
*rd_rn
;
1548 struct bgp_info
*global_ri
;
1550 if (evp
->prefix
.route_type
!= BGP_EVPN_MAC_IP_ROUTE
)
1553 if (IS_EVPN_PREFIX_IPADDR_V4(evp
)) {
1554 if (evpn_route_is_sticky(bgp
, rn
))
1555 update_evpn_route_entry(bgp
, vpn
, afi
, safi
, rn
,
1558 else if (evpn_route_is_def_gw(bgp
, rn
))
1559 update_evpn_route_entry(bgp
, vpn
, afi
, safi
, rn
,
1563 update_evpn_route_entry(bgp
, vpn
, afi
, safi
, rn
,
1564 &attr
, 0, 1, &ri
, 0);
1566 if (evpn_route_is_sticky(bgp
, rn
))
1567 update_evpn_route_entry(bgp
, vpn
, afi
, safi
, rn
,
1568 &attr_sticky_ip6
, 0, 1,
1570 else if (evpn_route_is_def_gw(bgp
, rn
))
1571 update_evpn_route_entry(bgp
, vpn
, afi
, safi
, rn
,
1572 &attr_def_gw_ip6
, 0, 1,
1575 update_evpn_route_entry(bgp
, vpn
, afi
, safi
, rn
,
1580 /* If a local route exists for this prefix, we need to update
1581 * the global routing table too.
1586 /* Perform route selection; this is just to set the flags
1588 * as local route in the VNI always wins.
1590 evpn_route_select_install(bgp
, vpn
, rn
);
1592 attr_new
= ri
->attr
;
1594 /* Update route in global routing table. */
1595 rd_rn
= bgp_afi_node_get(bgp
->rib
[afi
][safi
], afi
, safi
,
1596 (struct prefix
*)evp
, &vpn
->prd
);
1598 update_evpn_route_entry(bgp
, vpn
, afi
, safi
, rd_rn
, attr_new
, 0,
1601 /* Schedule for processing and unlock node. */
1602 bgp_process(bgp
, rd_rn
, afi
, safi
);
1603 bgp_unlock_node(rd_rn
);
1606 /* Unintern temporary. */
1607 aspath_unintern(&attr
.aspath
);
1608 aspath_unintern(&attr_ip6
.aspath
);
1609 aspath_unintern(&attr_sticky
.aspath
);
1610 aspath_unintern(&attr_sticky_ip6
.aspath
);
1611 aspath_unintern(&attr_def_gw
.aspath
);
1612 aspath_unintern(&attr_def_gw_ip6
.aspath
);
1618 * Delete all type-2 (MACIP) local routes for this VNI - only from the
1619 * global routing table. These are also scheduled for withdraw from peers.
1621 static int delete_global_type2_routes(struct bgp
*bgp
, struct bgpevpn
*vpn
)
1625 struct bgp_node
*rdrn
, *rn
;
1626 struct bgp_table
*table
;
1627 struct bgp_info
*ri
;
1632 rdrn
= bgp_node_lookup(bgp
->rib
[afi
][safi
], (struct prefix
*)&vpn
->prd
);
1633 if (rdrn
&& rdrn
->info
) {
1634 table
= (struct bgp_table
*)rdrn
->info
;
1635 for (rn
= bgp_table_top(table
); rn
; rn
= bgp_route_next(rn
)) {
1636 struct prefix_evpn
*evp
= (struct prefix_evpn
*)&rn
->p
;
1638 if (evp
->prefix
.route_type
!= BGP_EVPN_MAC_IP_ROUTE
)
1641 delete_evpn_route_entry(bgp
, vpn
, afi
, safi
, rn
, &ri
);
1643 bgp_process(bgp
, rn
, afi
, safi
);
1647 /* Unlock RD node. */
1649 bgp_unlock_node(rdrn
);
1655 * Delete all type-2 (MACIP) local routes for this VNI - from the global
1656 * table as well as the per-VNI route table.
1658 static int delete_all_type2_routes(struct bgp
*bgp
, struct bgpevpn
*vpn
)
1662 struct bgp_node
*rn
;
1663 struct bgp_info
*ri
;
1668 /* First, walk the global route table for this VNI's type-2 local
1670 * EVPN routes are a 2-level table, first get the RD table.
1672 delete_global_type2_routes(bgp
, vpn
);
1674 /* Next, walk this VNI's route table and delete local type-2 routes. */
1675 for (rn
= bgp_table_top(vpn
->route_table
); rn
;
1676 rn
= bgp_route_next(rn
)) {
1677 struct prefix_evpn
*evp
= (struct prefix_evpn
*)&rn
->p
;
1679 if (evp
->prefix
.route_type
!= BGP_EVPN_MAC_IP_ROUTE
)
1682 delete_evpn_route_entry(bgp
, vpn
, afi
, safi
, rn
, &ri
);
1684 /* Route entry in local table gets deleted immediately. */
1686 bgp_info_reap(rn
, ri
);
1693 * Delete all routes in the per-VNI route table.
1695 static int delete_all_vni_routes(struct bgp
*bgp
, struct bgpevpn
*vpn
)
1697 struct bgp_node
*rn
;
1698 struct bgp_info
*ri
, *nextri
;
1700 /* Walk this VNI's route table and delete all routes. */
1701 for (rn
= bgp_table_top(vpn
->route_table
); rn
;
1702 rn
= bgp_route_next(rn
)) {
1703 for (ri
= rn
->info
; (ri
!= NULL
) && (nextri
= ri
->next
, 1);
1705 bgp_info_delete(rn
, ri
);
1706 bgp_info_reap(rn
, ri
);
1714 * Update (and advertise) local routes for a VNI. Invoked upon the VNI
1715 * export RT getting modified or change to tunnel IP. Note that these
1716 * situations need the route in the per-VNI table as well as the global
1717 * table to be updated (as attributes change).
1719 static int update_routes_for_vni(struct bgp
*bgp
, struct bgpevpn
*vpn
)
1722 struct prefix_evpn p
;
1724 /* Update and advertise the type-3 route (only one) followed by the
1725 * locally learnt type-2 routes (MACIP) - for this VNI.
1727 build_evpn_type3_prefix(&p
, vpn
->originator_ip
);
1728 ret
= update_evpn_route(bgp
, vpn
, &p
, 0);
1732 return update_all_type2_routes(bgp
, vpn
);
1736 * Delete (and withdraw) local routes for specified VNI from the global
1737 * table and per-VNI table. After this, remove all other routes from
1738 * the per-VNI table. Invoked upon the VNI being deleted or EVPN
1739 * (advertise-all-vni) being disabled.
1741 static int delete_routes_for_vni(struct bgp
*bgp
, struct bgpevpn
*vpn
)
1744 struct prefix_evpn p
;
1746 /* Delete and withdraw locally learnt type-2 routes (MACIP)
1747 * followed by type-3 routes (only one) - for this VNI.
1749 ret
= delete_all_type2_routes(bgp
, vpn
);
1753 build_evpn_type3_prefix(&p
, vpn
->originator_ip
);
1754 ret
= delete_evpn_route(bgp
, vpn
, &p
);
1758 /* Delete all routes from the per-VNI table. */
1759 return delete_all_vni_routes(bgp
, vpn
);
1763 * There is a tunnel endpoint IP address change for this VNI, delete
1764 * prior type-3 route (if needed) and update.
1765 * Note: Route re-advertisement happens elsewhere after other processing
1768 static int handle_tunnel_ip_change(struct bgp
*bgp
, struct bgpevpn
*vpn
,
1769 struct in_addr originator_ip
)
1771 struct prefix_evpn p
;
1773 /* If VNI is not live, we only need to update the originator ip */
1774 if (!is_vni_live(vpn
)) {
1775 vpn
->originator_ip
= originator_ip
;
1779 /* Update the tunnel-ip hash */
1780 bgp_tip_del(bgp
, &vpn
->originator_ip
);
1781 bgp_tip_add(bgp
, &originator_ip
);
1783 /* filter routes as martian nexthop db has changed */
1784 bgp_filter_evpn_routes_upon_martian_nh_change(bgp
);
1786 /* Need to withdraw type-3 route as the originator IP is part
1789 build_evpn_type3_prefix(&p
, vpn
->originator_ip
);
1790 delete_evpn_route(bgp
, vpn
, &p
);
1792 /* Update the tunnel IP and re-advertise all routes for this VNI. */
1793 vpn
->originator_ip
= originator_ip
;
1798 * Install route entry into the VRF routing table and invoke route selection.
1800 static int install_evpn_route_entry_in_vrf(struct bgp
*bgp_vrf
,
1801 struct prefix_evpn
*evp
,
1802 struct bgp_info
*parent_ri
)
1804 struct bgp_node
*rn
;
1805 struct bgp_info
*ri
;
1806 struct attr
*attr_new
;
1809 struct prefix
*pp
= &p
;
1812 char buf
[PREFIX_STRLEN
];
1813 char buf1
[PREFIX_STRLEN
];
1815 memset(pp
, 0, sizeof(struct prefix
));
1816 if (evp
->prefix
.route_type
== BGP_EVPN_MAC_IP_ROUTE
)
1817 ip_prefix_from_type2_prefix(evp
, pp
);
1818 else if (evp
->prefix
.route_type
== BGP_EVPN_IP_PREFIX_ROUTE
)
1819 ip_prefix_from_type5_prefix(evp
, pp
);
1821 if (bgp_debug_zebra(NULL
)) {
1822 zlog_debug("installing evpn prefix %s as ip prefix %s in vrf %s",
1823 prefix2str(evp
, buf
, sizeof(buf
)),
1824 prefix2str(pp
, buf1
, sizeof(buf
)),
1825 vrf_id_to_name(bgp_vrf
->vrf_id
));
1828 /* Create (or fetch) route within the VRF. */
1829 /* NOTE: There is no RD here. */
1830 if (IS_EVPN_PREFIX_IPADDR_V4(evp
)) {
1832 safi
= SAFI_UNICAST
;
1833 rn
= bgp_node_get(bgp_vrf
->rib
[afi
][safi
], pp
);
1834 } else if (IS_EVPN_PREFIX_IPADDR_V6(evp
)) {
1836 safi
= SAFI_UNICAST
;
1837 rn
= bgp_node_get(bgp_vrf
->rib
[afi
][safi
], pp
);
1841 /* Check if route entry is already present. */
1842 for (ri
= rn
->info
; ri
; ri
= ri
->next
)
1844 && (struct bgp_info
*)ri
->extra
->parent
== parent_ri
)
1848 /* Add (or update) attribute to hash. */
1849 attr_new
= bgp_attr_intern(parent_ri
->attr
);
1851 /* Create new route with its attribute. */
1852 ri
= info_make(parent_ri
->type
, parent_ri
->sub_type
, 0,
1853 parent_ri
->peer
, attr_new
, rn
);
1854 SET_FLAG(ri
->flags
, BGP_INFO_VALID
);
1855 bgp_info_extra_get(ri
);
1856 ri
->extra
->parent
= parent_ri
;
1857 if (parent_ri
->extra
) {
1858 memcpy(&ri
->extra
->label
, &parent_ri
->extra
->label
,
1859 sizeof(ri
->extra
->label
));
1860 ri
->extra
->num_labels
= parent_ri
->extra
->num_labels
;
1862 bgp_info_add(rn
, ri
);
1864 if (attrhash_cmp(ri
->attr
, parent_ri
->attr
)
1865 && !CHECK_FLAG(ri
->flags
, BGP_INFO_REMOVED
)) {
1866 bgp_unlock_node(rn
);
1869 /* The attribute has changed. */
1870 /* Add (or update) attribute to hash. */
1871 attr_new
= bgp_attr_intern(parent_ri
->attr
);
1873 /* Restore route, if needed. */
1874 if (CHECK_FLAG(ri
->flags
, BGP_INFO_REMOVED
))
1875 bgp_info_restore(rn
, ri
);
1877 /* Mark if nexthop has changed. */
1878 if (!IPV4_ADDR_SAME(&ri
->attr
->nexthop
, &attr_new
->nexthop
))
1879 SET_FLAG(ri
->flags
, BGP_INFO_IGP_CHANGED
);
1881 /* Unintern existing, set to new. */
1882 bgp_attr_unintern(&ri
->attr
);
1883 ri
->attr
= attr_new
;
1884 ri
->uptime
= bgp_clock();
1887 /* Perform route selection and update zebra, if required. */
1888 bgp_process(bgp_vrf
, rn
, afi
, safi
);
1894 * Install route entry into the VNI routing table and invoke route selection.
1896 static int install_evpn_route_entry(struct bgp
*bgp
, struct bgpevpn
*vpn
,
1897 struct prefix_evpn
*p
,
1898 struct bgp_info
*parent_ri
)
1900 struct bgp_node
*rn
;
1901 struct bgp_info
*ri
;
1902 struct attr
*attr_new
;
1905 /* Create (or fetch) route within the VNI. */
1906 /* NOTE: There is no RD here. */
1907 rn
= bgp_node_get(vpn
->route_table
, (struct prefix
*)p
);
1909 /* Check if route entry is already present. */
1910 for (ri
= rn
->info
; ri
; ri
= ri
->next
)
1912 && (struct bgp_info
*)ri
->extra
->parent
== parent_ri
)
1916 /* Add (or update) attribute to hash. */
1917 attr_new
= bgp_attr_intern(parent_ri
->attr
);
1919 /* Create new route with its attribute. */
1920 ri
= info_make(parent_ri
->type
, parent_ri
->sub_type
, 0,
1921 parent_ri
->peer
, attr_new
, rn
);
1922 SET_FLAG(ri
->flags
, BGP_INFO_VALID
);
1923 bgp_info_extra_get(ri
);
1924 ri
->extra
->parent
= parent_ri
;
1925 if (parent_ri
->extra
) {
1926 memcpy(&ri
->extra
->label
, &parent_ri
->extra
->label
,
1927 sizeof(ri
->extra
->label
));
1928 ri
->extra
->num_labels
= parent_ri
->extra
->num_labels
;
1930 bgp_info_add(rn
, ri
);
1932 if (attrhash_cmp(ri
->attr
, parent_ri
->attr
)
1933 && !CHECK_FLAG(ri
->flags
, BGP_INFO_REMOVED
)) {
1934 bgp_unlock_node(rn
);
1937 /* The attribute has changed. */
1938 /* Add (or update) attribute to hash. */
1939 attr_new
= bgp_attr_intern(parent_ri
->attr
);
1941 /* Restore route, if needed. */
1942 if (CHECK_FLAG(ri
->flags
, BGP_INFO_REMOVED
))
1943 bgp_info_restore(rn
, ri
);
1945 /* Mark if nexthop has changed. */
1946 if (!IPV4_ADDR_SAME(&ri
->attr
->nexthop
, &attr_new
->nexthop
))
1947 SET_FLAG(ri
->flags
, BGP_INFO_IGP_CHANGED
);
1949 /* Unintern existing, set to new. */
1950 bgp_attr_unintern(&ri
->attr
);
1951 ri
->attr
= attr_new
;
1952 ri
->uptime
= bgp_clock();
1955 /* Perform route selection and update zebra, if required. */
1956 ret
= evpn_route_select_install(bgp
, vpn
, rn
);
1962 * Uninstall route entry from the VRF routing table and send message
1963 * to zebra, if appropriate.
1965 static int uninstall_evpn_route_entry_in_vrf(struct bgp
*bgp_vrf
,
1966 struct prefix_evpn
*evp
,
1967 struct bgp_info
*parent_ri
)
1969 struct bgp_node
*rn
;
1970 struct bgp_info
*ri
;
1973 struct prefix
*pp
= &p
;
1976 char buf
[PREFIX_STRLEN
];
1977 char buf1
[PREFIX_STRLEN
];
1979 memset(pp
, 0, sizeof(struct prefix
));
1980 if (evp
->prefix
.route_type
== BGP_EVPN_MAC_IP_ROUTE
)
1981 ip_prefix_from_type2_prefix(evp
, pp
);
1982 else if (evp
->prefix
.route_type
== BGP_EVPN_IP_PREFIX_ROUTE
)
1983 ip_prefix_from_type5_prefix(evp
, pp
);
1985 if (bgp_debug_zebra(NULL
)) {
1986 zlog_debug("uninstalling evpn prefix %s as ip prefix %s in vrf %s",
1987 prefix2str(evp
, buf
, sizeof(buf
)),
1988 prefix2str(pp
, buf1
, sizeof(buf
)),
1989 vrf_id_to_name(bgp_vrf
->vrf_id
));
1992 /* Locate route within the VRF. */
1993 /* NOTE: There is no RD here. */
1994 if (IS_EVPN_PREFIX_IPADDR_V4(evp
)) {
1996 safi
= SAFI_UNICAST
;
1997 rn
= bgp_node_lookup(bgp_vrf
->rib
[afi
][safi
], pp
);
2000 safi
= SAFI_UNICAST
;
2001 rn
= bgp_node_lookup(bgp_vrf
->rib
[afi
][safi
], pp
);
2007 /* Find matching route entry. */
2008 for (ri
= rn
->info
; ri
; ri
= ri
->next
)
2010 && (struct bgp_info
*)ri
->extra
->parent
== parent_ri
)
2016 /* Mark entry for deletion */
2017 bgp_info_delete(rn
, ri
);
2019 /* Perform route selection and update zebra, if required. */
2020 bgp_process(bgp_vrf
, rn
, afi
, safi
);
2022 /* Unlock route node. */
2023 bgp_unlock_node(rn
);
2029 * Uninstall route entry from the VNI routing table and send message
2030 * to zebra, if appropriate.
2032 static int uninstall_evpn_route_entry(struct bgp
*bgp
, struct bgpevpn
*vpn
,
2033 struct prefix_evpn
*p
,
2034 struct bgp_info
*parent_ri
)
2036 struct bgp_node
*rn
;
2037 struct bgp_info
*ri
;
2040 /* Locate route within the VNI. */
2041 /* NOTE: There is no RD here. */
2042 rn
= bgp_node_lookup(vpn
->route_table
, (struct prefix
*)p
);
2046 /* Find matching route entry. */
2047 for (ri
= rn
->info
; ri
; ri
= ri
->next
)
2049 && (struct bgp_info
*)ri
->extra
->parent
== parent_ri
)
2055 /* Mark entry for deletion */
2056 bgp_info_delete(rn
, ri
);
2058 /* Perform route selection and update zebra, if required. */
2059 ret
= evpn_route_select_install(bgp
, vpn
, rn
);
2061 /* Unlock route node. */
2062 bgp_unlock_node(rn
);
2068 * Given a route entry and a VRF, see if this route entry should be
2069 * imported into the VRF i.e., RTs match.
2071 static int is_route_matching_for_vrf(struct bgp
*bgp_vrf
,
2072 struct bgp_info
*ri
)
2074 struct attr
*attr
= ri
->attr
;
2075 struct ecommunity
*ecom
;
2079 /* Route should have valid RT to be even considered. */
2080 if (!(attr
->flag
& ATTR_FLAG_BIT(BGP_ATTR_EXT_COMMUNITIES
)))
2083 ecom
= attr
->ecommunity
;
2084 if (!ecom
|| !ecom
->size
)
2087 /* For each extended community RT, see if it matches this VNI. If any RT
2088 * matches, we're done.
2090 for (i
= 0; i
< ecom
->size
; i
++) {
2092 u_char type
, sub_type
;
2093 struct ecommunity_val
*eval
;
2094 struct ecommunity_val eval_tmp
;
2095 struct vrf_irt_node
*irt
;
2097 /* Only deal with RTs */
2098 pnt
= (ecom
->val
+ (i
* ECOMMUNITY_SIZE
));
2099 eval
= (struct ecommunity_val
*)(ecom
->val
2100 + (i
* ECOMMUNITY_SIZE
));
2103 if (sub_type
!= ECOMMUNITY_ROUTE_TARGET
)
2106 /* See if this RT matches specified VNIs import RTs */
2107 irt
= lookup_vrf_import_rt(eval
);
2108 if (irt
&& irt
->vrfs
)
2109 if (is_vrf_present_in_irt_vrfs(irt
->vrfs
, bgp_vrf
))
2112 /* Also check for non-exact match. In this, we mask out the AS
2114 * only check on the local-admin sub-field. This is to
2116 * VNI as the RT for EBGP peering too.
2119 if (type
== ECOMMUNITY_ENCODE_AS
2120 || type
== ECOMMUNITY_ENCODE_AS4
2121 || type
== ECOMMUNITY_ENCODE_IP
) {
2122 memcpy(&eval_tmp
, eval
, ECOMMUNITY_SIZE
);
2123 mask_ecom_global_admin(&eval_tmp
, eval
);
2124 irt
= lookup_vrf_import_rt(&eval_tmp
);
2126 if (irt
&& irt
->vrfs
)
2127 if (is_vrf_present_in_irt_vrfs(irt
->vrfs
, bgp_vrf
))
2135 * Given a route entry and a VNI, see if this route entry should be
2136 * imported into the VNI i.e., RTs match.
2138 static int is_route_matching_for_vni(struct bgp
*bgp
, struct bgpevpn
*vpn
,
2139 struct bgp_info
*ri
)
2141 struct attr
*attr
= ri
->attr
;
2142 struct ecommunity
*ecom
;
2146 /* Route should have valid RT to be even considered. */
2147 if (!(attr
->flag
& ATTR_FLAG_BIT(BGP_ATTR_EXT_COMMUNITIES
)))
2150 ecom
= attr
->ecommunity
;
2151 if (!ecom
|| !ecom
->size
)
2154 /* For each extended community RT, see if it matches this VNI. If any RT
2155 * matches, we're done.
2157 for (i
= 0; i
< ecom
->size
; i
++) {
2159 u_char type
, sub_type
;
2160 struct ecommunity_val
*eval
;
2161 struct ecommunity_val eval_tmp
;
2162 struct irt_node
*irt
;
2164 /* Only deal with RTs */
2165 pnt
= (ecom
->val
+ (i
* ECOMMUNITY_SIZE
));
2166 eval
= (struct ecommunity_val
*)(ecom
->val
2167 + (i
* ECOMMUNITY_SIZE
));
2170 if (sub_type
!= ECOMMUNITY_ROUTE_TARGET
)
2173 /* See if this RT matches specified VNIs import RTs */
2174 irt
= lookup_import_rt(bgp
, eval
);
2175 if (irt
&& irt
->vnis
)
2176 if (is_vni_present_in_irt_vnis(irt
->vnis
, vpn
))
2179 /* Also check for non-exact match. In this, we mask out the AS
2181 * only check on the local-admin sub-field. This is to
2183 * VNI as the RT for EBGP peering too.
2186 if (type
== ECOMMUNITY_ENCODE_AS
2187 || type
== ECOMMUNITY_ENCODE_AS4
2188 || type
== ECOMMUNITY_ENCODE_IP
) {
2189 memcpy(&eval_tmp
, eval
, ECOMMUNITY_SIZE
);
2190 mask_ecom_global_admin(&eval_tmp
, eval
);
2191 irt
= lookup_import_rt(bgp
, &eval_tmp
);
2193 if (irt
&& irt
->vnis
)
2194 if (is_vni_present_in_irt_vnis(irt
->vnis
, vpn
))
2202 * Install or uninstall mac-ip routes are appropriate for this
2205 static int install_uninstall_routes_for_vrf(struct bgp
*bgp_vrf
,
2210 struct bgp_node
*rd_rn
, *rn
;
2211 struct bgp_table
*table
;
2212 struct bgp_info
*ri
;
2214 char buf
[PREFIX_STRLEN
];
2215 struct bgp
*bgp_def
= NULL
;
2219 bgp_def
= bgp_get_default();
2223 /* Walk entire global routing table and evaluate routes which could be
2224 * imported into this VRF. Note that we need to loop through all global
2225 * routes to determine which route matches the import rt on vrf
2227 for (rd_rn
= bgp_table_top(bgp_def
->rib
[afi
][safi
]); rd_rn
;
2228 rd_rn
= bgp_route_next(rd_rn
)) {
2229 table
= (struct bgp_table
*)(rd_rn
->info
);
2233 for (rn
= bgp_table_top(table
); rn
; rn
= bgp_route_next(rn
)) {
2234 struct prefix_evpn
*evp
= (struct prefix_evpn
*)&rn
->p
;
2236 /* if not mac-ip route skip this route */
2237 if (!(evp
->prefix
.route_type
== BGP_EVPN_MAC_IP_ROUTE
||
2238 evp
->prefix
.route_type
== BGP_EVPN_IP_PREFIX_ROUTE
))
2241 /* if not a mac+ip route skip this route */
2242 if (!(IS_EVPN_PREFIX_IPADDR_V4(evp
) ||
2243 IS_EVPN_PREFIX_IPADDR_V6(evp
)))
2246 for (ri
= rn
->info
; ri
; ri
= ri
->next
) {
2247 /* Consider "valid" remote routes applicable for
2250 if (!(CHECK_FLAG(ri
->flags
, BGP_INFO_VALID
)
2251 && ri
->type
== ZEBRA_ROUTE_BGP
2252 && ri
->sub_type
== BGP_ROUTE_NORMAL
))
2255 if (is_route_matching_for_vrf(bgp_vrf
, ri
)) {
2258 install_evpn_route_entry_in_vrf(
2262 uninstall_evpn_route_entry_in_vrf(
2267 "Failed to %s EVPN %s route in VRF %s",
2270 prefix2str(evp
, buf
,
2272 vrf_id_to_name(bgp_vrf
->vrf_id
));
2284 * Install or uninstall routes of specified type that are appropriate for this
2287 static int install_uninstall_routes_for_vni(struct bgp
*bgp
,
2288 struct bgpevpn
*vpn
,
2289 bgp_evpn_route_type rtype
,
2294 struct bgp_node
*rd_rn
, *rn
;
2295 struct bgp_table
*table
;
2296 struct bgp_info
*ri
;
2302 /* Walk entire global routing table and evaluate routes which could be
2303 * imported into this VPN. Note that we cannot just look at the routes
2305 * the VNI's RD - remote routes applicable for this VNI could have any
2308 /* EVPN routes are a 2-level table. */
2309 for (rd_rn
= bgp_table_top(bgp
->rib
[afi
][safi
]); rd_rn
;
2310 rd_rn
= bgp_route_next(rd_rn
)) {
2311 table
= (struct bgp_table
*)(rd_rn
->info
);
2315 for (rn
= bgp_table_top(table
); rn
; rn
= bgp_route_next(rn
)) {
2316 struct prefix_evpn
*evp
= (struct prefix_evpn
*)&rn
->p
;
2318 if (evp
->prefix
.route_type
!= rtype
)
2321 for (ri
= rn
->info
; ri
; ri
= ri
->next
) {
2322 /* Consider "valid" remote routes applicable for
2324 if (!(CHECK_FLAG(ri
->flags
, BGP_INFO_VALID
)
2325 && ri
->type
== ZEBRA_ROUTE_BGP
2326 && ri
->sub_type
== BGP_ROUTE_NORMAL
))
2329 if (is_route_matching_for_vni(bgp
, vpn
, ri
)) {
2331 ret
= install_evpn_route_entry(
2334 ret
= uninstall_evpn_route_entry(
2339 "%u: Failed to %s EVPN %s route in VNI %u",
2343 rtype
== BGP_EVPN_MAC_IP_ROUTE
2357 /* Install any existing remote routes applicable for this VRF into VRF RIB. This
2358 * is invoked upon l3vni-add or l3vni import rt change
2360 static int install_routes_for_vrf(struct bgp
*bgp_vrf
)
2362 install_uninstall_routes_for_vrf(bgp_vrf
, 1);
2367 * Install any existing remote routes applicable for this VNI into its
2368 * routing table. This is invoked when a VNI becomes "live" or its Import
2371 static int install_routes_for_vni(struct bgp
*bgp
, struct bgpevpn
*vpn
)
2375 /* Install type-3 routes followed by type-2 routes - the ones applicable
2378 ret
= install_uninstall_routes_for_vni(bgp
, vpn
, BGP_EVPN_IMET_ROUTE
,
2383 return install_uninstall_routes_for_vni(bgp
, vpn
, BGP_EVPN_MAC_IP_ROUTE
,
2387 /* uninstall routes from l3vni vrf. */
2388 static int uninstall_routes_for_vrf(struct bgp
*bgp_vrf
)
2390 install_uninstall_routes_for_vrf(bgp_vrf
, 0);
2395 * Uninstall any existing remote routes for this VNI. One scenario in which
2396 * this is invoked is upon an import RT change.
2398 static int uninstall_routes_for_vni(struct bgp
*bgp
, struct bgpevpn
*vpn
)
2402 /* Uninstall type-2 routes followed by type-3 routes - the ones
2406 ret
= install_uninstall_routes_for_vni(bgp
, vpn
, BGP_EVPN_MAC_IP_ROUTE
,
2411 return install_uninstall_routes_for_vni(bgp
, vpn
, BGP_EVPN_IMET_ROUTE
,
2416 * Install or uninstall route in matching VRFs (list).
2418 static int install_uninstall_route_in_vrfs(struct bgp
*bgp_def
, afi_t afi
,
2419 safi_t safi
, struct prefix_evpn
*evp
,
2420 struct bgp_info
*ri
,
2421 struct list
*vrfs
, int install
)
2423 char buf
[PREFIX2STR_BUFFER
];
2424 struct bgp
*bgp_vrf
;
2425 struct listnode
*node
, *nnode
;
2427 /* Only type-2/type-5 routes go into a VRF */
2428 if (!(evp
->prefix
.route_type
== BGP_EVPN_MAC_IP_ROUTE
||
2429 evp
->prefix
.route_type
== BGP_EVPN_IP_PREFIX_ROUTE
))
2432 /* if it is type-2 route and not a mac+ip route skip this route */
2433 if ((evp
->prefix
.route_type
== BGP_EVPN_MAC_IP_ROUTE
) &&
2434 !(IS_EVPN_PREFIX_IPADDR_V4(evp
) || IS_EVPN_PREFIX_IPADDR_V6(evp
)))
2437 for (ALL_LIST_ELEMENTS(vrfs
, node
, nnode
, bgp_vrf
)) {
2441 ret
= install_evpn_route_entry_in_vrf(bgp_vrf
,
2444 ret
= uninstall_evpn_route_entry_in_vrf(bgp_vrf
,
2448 zlog_err("%u: Failed to %s prefix %s in VRF %s",
2450 install
? "install" : "uninstall",
2451 prefix2str(evp
, buf
, sizeof(buf
)),
2452 vrf_id_to_name(bgp_vrf
->vrf_id
));
2461 * Install or uninstall route in matching VNIs (list).
2463 static int install_uninstall_route_in_vnis(struct bgp
*bgp
, afi_t afi
,
2464 safi_t safi
, struct prefix_evpn
*evp
,
2465 struct bgp_info
*ri
,
2466 struct list
*vnis
, int install
)
2468 struct bgpevpn
*vpn
;
2469 struct listnode
*node
, *nnode
;
2471 for (ALL_LIST_ELEMENTS(vnis
, node
, nnode
, vpn
)) {
2474 if (!is_vni_live(vpn
))
2478 ret
= install_evpn_route_entry(bgp
, vpn
, evp
, ri
);
2480 ret
= uninstall_evpn_route_entry(bgp
, vpn
, evp
, ri
);
2483 zlog_err("%u: Failed to %s EVPN %s route in VNI %u",
2484 bgp
->vrf_id
, install
? "install" : "uninstall",
2485 evp
->prefix
.route_type
== BGP_EVPN_MAC_IP_ROUTE
2497 * Install or uninstall route for appropriate VNIs.
2499 static int install_uninstall_evpn_route(struct bgp
*bgp
, afi_t afi
, safi_t safi
,
2500 struct prefix
*p
, struct bgp_info
*ri
,
2503 struct prefix_evpn
*evp
= (struct prefix_evpn
*)p
;
2504 struct attr
*attr
= ri
->attr
;
2505 struct ecommunity
*ecom
;
2510 /* Only type-2 and type-3 and type-5 are supported currently */
2511 if (!(evp
->prefix
.route_type
== BGP_EVPN_MAC_IP_ROUTE
2512 || evp
->prefix
.route_type
== BGP_EVPN_IMET_ROUTE
2513 || evp
->prefix
.route_type
== BGP_EVPN_IP_PREFIX_ROUTE
))
2516 /* If we don't have Route Target, nothing much to do. */
2517 if (!(attr
->flag
& ATTR_FLAG_BIT(BGP_ATTR_EXT_COMMUNITIES
)))
2520 ecom
= attr
->ecommunity
;
2521 if (!ecom
|| !ecom
->size
)
2524 /* For each extended community RT, see which VNIs/VRFs match and import
2525 * the route into matching VNIs/VRFs.
2527 for (i
= 0; i
< ecom
->size
; i
++) {
2529 u_char type
, sub_type
;
2530 struct ecommunity_val
*eval
;
2531 struct ecommunity_val eval_tmp
;
2532 struct irt_node
*irt
; /* import rt for l2vni */
2533 struct vrf_irt_node
*vrf_irt
; /* import rt for l3vni */
2535 /* Only deal with RTs */
2536 pnt
= (ecom
->val
+ (i
* ECOMMUNITY_SIZE
));
2537 eval
= (struct ecommunity_val
*)(ecom
->val
2538 + (i
* ECOMMUNITY_SIZE
));
2541 if (sub_type
!= ECOMMUNITY_ROUTE_TARGET
)
2544 /* Import route into matching l2-vnis (type-2/type-3 routes go
2547 irt
= lookup_import_rt(bgp
, eval
);
2548 if (irt
&& irt
->vnis
)
2549 install_uninstall_route_in_vnis(bgp
, afi
, safi
, evp
, ri
,
2552 /* Import route into matching l3-vnis (type-2/type-5 routes go
2553 * into l3vni/vrf table)
2555 vrf_irt
= lookup_vrf_import_rt(eval
);
2556 if (vrf_irt
&& vrf_irt
->vrfs
)
2557 install_uninstall_route_in_vrfs(bgp
, afi
, safi
, evp
, ri
,
2558 vrf_irt
->vrfs
, import
);
2560 /* Also check for non-exact match. In this,
2561 * we mask out the AS and
2562 * only check on the local-admin sub-field.
2563 * This is to facilitate using
2564 * VNI as the RT for EBGP peering too.
2568 if (type
== ECOMMUNITY_ENCODE_AS
2569 || type
== ECOMMUNITY_ENCODE_AS4
2570 || type
== ECOMMUNITY_ENCODE_IP
) {
2571 memcpy(&eval_tmp
, eval
, ECOMMUNITY_SIZE
);
2572 mask_ecom_global_admin(&eval_tmp
, eval
);
2573 irt
= lookup_import_rt(bgp
, &eval_tmp
);
2574 vrf_irt
= lookup_vrf_import_rt(&eval_tmp
);
2576 if (irt
&& irt
->vnis
)
2577 install_uninstall_route_in_vnis(bgp
, afi
, safi
, evp
, ri
,
2579 if (vrf_irt
&& vrf_irt
->vrfs
)
2580 install_uninstall_route_in_vrfs(bgp
, afi
, safi
, evp
,
2588 /* delete and withdraw all ipv4 and ipv6 routes in the vrf table as type-5
2590 static void delete_withdraw_vrf_routes(struct bgp
*bgp_vrf
)
2592 /* delete all ipv4 routes and withdraw from peers */
2593 bgp_evpn_withdraw_type5_routes(bgp_vrf
, AFI_IP
, SAFI_UNICAST
);
2595 /* delete all ipv6 routes and withdraw from peers */
2596 bgp_evpn_withdraw_type5_routes(bgp_vrf
, AFI_IP6
, SAFI_UNICAST
);
2599 /* update and advertise all ipv4 and ipv6 routes in thr vrf table as type-5
2601 static void update_advertise_vrf_routes(struct bgp
*bgp_vrf
)
2603 /* update all ipv4 routes */
2604 bgp_evpn_advertise_type5_routes(bgp_vrf
, AFI_IP
, SAFI_UNICAST
);
2606 /* update all ipv6 routes */
2607 bgp_evpn_advertise_type5_routes(bgp_vrf
, AFI_IP6
, SAFI_UNICAST
);
2611 * update and advertise local routes for a VRF as type-5 routes.
2612 * This is invoked upon RD change for a VRF. Note taht the processing is only
2613 * done in the global route table using the routes which already exist in the
2616 static void update_router_id_vrf(struct bgp
*bgp_vrf
)
2618 /* skip if the RD is configured */
2619 if (is_vrf_rd_configured(bgp_vrf
))
2622 /* derive the RD for the VRF based on new router-id */
2623 bgp_evpn_derive_auto_rd_for_vrf(bgp_vrf
);
2625 /* update advertise ipv4|ipv6 routes as type-5 routes */
2626 update_advertise_vrf_routes(bgp_vrf
);
2630 * Delete and withdraw all type-5 routes for the RD corresponding to VRF.
2631 * This is invoked upon VRF RD change. The processing is done only from global
2634 static void withdraw_router_id_vrf(struct bgp
*bgp_vrf
)
2636 /* skip if the RD is configured */
2637 if (is_vrf_rd_configured(bgp_vrf
))
2640 /* delete/withdraw ipv4|ipv6 routes as type-5 routes */
2641 delete_withdraw_vrf_routes(bgp_vrf
);
2645 * Update and advertise local routes for a VNI. Invoked upon router-id
2646 * change. Note that the processing is done only on the global route table
2647 * using routes that already exist in the per-VNI table.
2649 static int update_advertise_vni_routes(struct bgp
*bgp
, struct bgpevpn
*vpn
)
2651 struct prefix_evpn p
;
2652 struct bgp_node
*rn
, *global_rn
;
2653 struct bgp_info
*ri
, *global_ri
;
2655 afi_t afi
= AFI_L2VPN
;
2656 safi_t safi
= SAFI_EVPN
;
2658 /* Locate type-3 route for VNI in the per-VNI table and use its
2659 * attributes to create and advertise the type-3 route for this VNI
2660 * in the global table.
2662 build_evpn_type3_prefix(&p
, vpn
->originator_ip
);
2663 rn
= bgp_node_lookup(vpn
->route_table
, (struct prefix
*)&p
);
2664 if (!rn
) /* unexpected */
2666 for (ri
= rn
->info
; ri
; ri
= ri
->next
)
2667 if (ri
->peer
== bgp
->peer_self
&& ri
->type
== ZEBRA_ROUTE_BGP
2668 && ri
->sub_type
== BGP_ROUTE_STATIC
)
2670 if (!ri
) /* unexpected */
2674 global_rn
= bgp_afi_node_get(bgp
->rib
[afi
][safi
], afi
, safi
,
2675 (struct prefix
*)&p
, &vpn
->prd
);
2676 update_evpn_route_entry(bgp
, vpn
, afi
, safi
, global_rn
, attr
, 1, 0, &ri
,
2679 /* Schedule for processing and unlock node. */
2680 bgp_process(bgp
, global_rn
, afi
, safi
);
2681 bgp_unlock_node(global_rn
);
2683 /* Now, walk this VNI's route table and use the route and its attribute
2684 * to create and schedule route in global table.
2686 for (rn
= bgp_table_top(vpn
->route_table
); rn
;
2687 rn
= bgp_route_next(rn
)) {
2688 struct prefix_evpn
*evp
= (struct prefix_evpn
*)&rn
->p
;
2690 /* Identify MAC-IP local routes. */
2691 if (evp
->prefix
.route_type
!= BGP_EVPN_MAC_IP_ROUTE
)
2694 for (ri
= rn
->info
; ri
; ri
= ri
->next
)
2695 if (ri
->peer
== bgp
->peer_self
2696 && ri
->type
== ZEBRA_ROUTE_BGP
2697 && ri
->sub_type
== BGP_ROUTE_STATIC
)
2702 /* Create route in global routing table using this route entry's
2706 global_rn
= bgp_afi_node_get(bgp
->rib
[afi
][safi
], afi
, safi
,
2707 (struct prefix
*)evp
, &vpn
->prd
);
2709 update_evpn_route_entry(bgp
, vpn
, afi
, safi
, global_rn
, attr
, 1,
2712 /* Schedule for processing and unlock node. */
2713 bgp_process(bgp
, global_rn
, afi
, safi
);
2714 bgp_unlock_node(global_rn
);
2721 * Delete (and withdraw) local routes for a VNI - only from the global
2722 * table. Invoked upon router-id change.
2724 static int delete_withdraw_vni_routes(struct bgp
*bgp
, struct bgpevpn
*vpn
)
2727 struct prefix_evpn p
;
2728 struct bgp_node
*global_rn
;
2729 struct bgp_info
*ri
;
2730 afi_t afi
= AFI_L2VPN
;
2731 safi_t safi
= SAFI_EVPN
;
2733 /* Delete and withdraw locally learnt type-2 routes (MACIP)
2734 * for this VNI - from the global table.
2736 ret
= delete_global_type2_routes(bgp
, vpn
);
2740 /* Remove type-3 route for this VNI from global table. */
2741 build_evpn_type3_prefix(&p
, vpn
->originator_ip
);
2742 global_rn
= bgp_afi_node_lookup(bgp
->rib
[afi
][safi
], afi
, safi
,
2743 (struct prefix
*)&p
, &vpn
->prd
);
2745 /* Delete route entry in the global EVPN table. */
2746 delete_evpn_route_entry(bgp
, vpn
, afi
, safi
, global_rn
, &ri
);
2748 /* Schedule for processing - withdraws to peers happen from
2752 bgp_process(bgp
, global_rn
, afi
, safi
);
2753 bgp_unlock_node(global_rn
);
2760 * Handle router-id change. Update and advertise local routes corresponding
2761 * to this VNI from peers. Note that this is invoked after updating the
2762 * router-id. The routes in the per-VNI table are used to create routes in
2763 * the global table and schedule them.
2765 static void update_router_id_vni(struct hash_backet
*backet
, struct bgp
*bgp
)
2767 struct bgpevpn
*vpn
;
2769 vpn
= (struct bgpevpn
*)backet
->data
;
2772 zlog_warn("%s: VNI hash entry for VNI not found", __FUNCTION__
);
2776 /* Skip VNIs with configured RD. */
2777 if (is_rd_configured(vpn
))
2780 bgp_evpn_derive_auto_rd(bgp
, vpn
);
2781 update_advertise_vni_routes(bgp
, vpn
);
2785 * Handle router-id change. Delete and withdraw local routes corresponding
2786 * to this VNI from peers. Note that this is invoked prior to updating
2787 * the router-id and is done only on the global route table, the routes
2788 * are needed in the per-VNI table to re-advertise with new router id.
2790 static void withdraw_router_id_vni(struct hash_backet
*backet
, struct bgp
*bgp
)
2792 struct bgpevpn
*vpn
;
2794 vpn
= (struct bgpevpn
*)backet
->data
;
2797 zlog_warn("%s: VNI hash entry for VNI not found", __FUNCTION__
);
2801 /* Skip VNIs with configured RD. */
2802 if (is_rd_configured(vpn
))
2805 delete_withdraw_vni_routes(bgp
, vpn
);
2809 * Process received EVPN type-2 route (advertise or withdraw).
2811 static int process_type2_route(struct peer
*peer
, afi_t afi
, safi_t safi
,
2812 struct attr
*attr
, u_char
*pfx
, int psize
,
2813 u_int32_t addpath_id
)
2815 struct prefix_rd prd
;
2816 struct prefix_evpn p
;
2819 mpls_label_t label
[BGP_MAX_LABELS
]; /* holds the VNI(s) as in packet */
2820 u_int32_t num_labels
= 0;
2823 /* Type-2 route should be either 33, 37 or 49 bytes or an
2824 * additional 3 bytes if there is a second label (VNI):
2825 * RD (8), ESI (10), Eth Tag (4), MAC Addr Len (1),
2826 * MAC Addr (6), IP len (1), IP (0, 4 or 16),
2827 * MPLS Lbl1 (3), MPLS Lbl2 (0 or 3)
2829 if (psize
!= 33 && psize
!= 37 && psize
!= 49 && psize
!= 36
2830 && psize
!= 40 && psize
!= 52) {
2831 zlog_err("%u:%s - Rx EVPN Type-2 NLRI with invalid length %d",
2832 peer
->bgp
->vrf_id
, peer
->host
, psize
);
2836 /* Make prefix_rd */
2837 prd
.family
= AF_UNSPEC
;
2839 memcpy(&prd
.val
, pfx
, 8);
2842 /* Make EVPN prefix. */
2843 memset(&p
, 0, sizeof(struct prefix_evpn
));
2845 p
.prefixlen
= EVPN_TYPE_2_ROUTE_PREFIXLEN
;
2846 p
.prefix
.route_type
= BGP_EVPN_MAC_IP_ROUTE
;
2848 /* Skip over Ethernet Seg Identifier for now. */
2851 /* Skip over Ethernet Tag for now. */
2854 /* Get the MAC Addr len */
2855 macaddr_len
= *pfx
++;
2857 /* Get the MAC Addr */
2858 if (macaddr_len
== (ETH_ALEN
* 8)) {
2859 memcpy(&p
.prefix
.mac
.octet
, pfx
, ETH_ALEN
);
2863 "%u:%s - Rx EVPN Type-2 NLRI with unsupported MAC address length %d",
2864 peer
->bgp
->vrf_id
, peer
->host
, macaddr_len
);
2870 ipaddr_len
= *pfx
++;
2871 if (ipaddr_len
!= 0 && ipaddr_len
!= IPV4_MAX_BITLEN
2872 && ipaddr_len
!= IPV6_MAX_BITLEN
) {
2874 "%u:%s - Rx EVPN Type-2 NLRI with unsupported IP address length %d",
2875 peer
->bgp
->vrf_id
, peer
->host
, ipaddr_len
);
2880 ipaddr_len
/= 8; /* Convert to bytes. */
2881 p
.prefix
.ip
.ipa_type
= (ipaddr_len
== IPV4_MAX_BYTELEN
)
2884 memcpy(&p
.prefix
.ip
.ip
.addr
, pfx
, ipaddr_len
);
2888 /* Get the VNI(s). Stored as bytes here. */
2890 memset(label
, 0, sizeof(label
));
2891 memcpy(&label
[0], pfx
, BGP_LABEL_BYTES
);
2892 pfx
+= BGP_LABEL_BYTES
;
2893 psize
-= (33 + ipaddr_len
);
2894 /* Do we have a second VNI? */
2897 memcpy(&label
[1], pfx
, BGP_LABEL_BYTES
);
2899 * If in future, we are required to access additional fields,
2900 * we MUST increment pfx by BGP_LABEL_BYTES in before reading the next field
2904 /* Process the route. */
2906 ret
= bgp_update(peer
, (struct prefix
*)&p
, addpath_id
, attr
,
2907 afi
, safi
, ZEBRA_ROUTE_BGP
, BGP_ROUTE_NORMAL
,
2908 &prd
, &label
[0], num_labels
, 0, NULL
);
2910 ret
= bgp_withdraw(peer
, (struct prefix
*)&p
, addpath_id
, attr
,
2911 afi
, safi
, ZEBRA_ROUTE_BGP
, BGP_ROUTE_NORMAL
,
2912 &prd
, &label
[0], num_labels
, NULL
);
2917 * Process received EVPN type-3 route (advertise or withdraw).
2919 static int process_type3_route(struct peer
*peer
, afi_t afi
, safi_t safi
,
2920 struct attr
*attr
, u_char
*pfx
, int psize
,
2921 u_int32_t addpath_id
)
2923 struct prefix_rd prd
;
2924 struct prefix_evpn p
;
2928 /* Type-3 route should be either 17 or 29 bytes: RD (8), Eth Tag (4),
2929 * IP len (1) and IP (4 or 16).
2931 if (psize
!= 17 && psize
!= 29) {
2932 zlog_err("%u:%s - Rx EVPN Type-3 NLRI with invalid length %d",
2933 peer
->bgp
->vrf_id
, peer
->host
, psize
);
2937 /* Make prefix_rd */
2938 prd
.family
= AF_UNSPEC
;
2940 memcpy(&prd
.val
, pfx
, 8);
2943 /* Make EVPN prefix. */
2944 memset(&p
, 0, sizeof(struct prefix_evpn
));
2946 p
.prefixlen
= EVPN_TYPE_3_ROUTE_PREFIXLEN
;
2947 p
.prefix
.route_type
= BGP_EVPN_IMET_ROUTE
;
2949 /* Skip over Ethernet Tag for now. */
2953 ipaddr_len
= *pfx
++;
2954 if (ipaddr_len
== IPV4_MAX_BITLEN
) {
2955 p
.prefix
.ip
.ipa_type
= IPADDR_V4
;
2956 memcpy(&p
.prefix
.ip
.ip
.addr
, pfx
, IPV4_MAX_BYTELEN
);
2959 "%u:%s - Rx EVPN Type-3 NLRI with unsupported IP address length %d",
2960 peer
->bgp
->vrf_id
, peer
->host
, ipaddr_len
);
2964 /* Process the route. */
2966 ret
= bgp_update(peer
, (struct prefix
*)&p
, addpath_id
, attr
,
2967 afi
, safi
, ZEBRA_ROUTE_BGP
, BGP_ROUTE_NORMAL
,
2968 &prd
, NULL
, 0, 0, NULL
);
2970 ret
= bgp_withdraw(peer
, (struct prefix
*)&p
, addpath_id
, attr
,
2971 afi
, safi
, ZEBRA_ROUTE_BGP
, BGP_ROUTE_NORMAL
,
2972 &prd
, NULL
, 0, NULL
);
2977 * Process received EVPN type-5 route (advertise or withdraw).
2979 static int process_type5_route(struct peer
*peer
, afi_t afi
, safi_t safi
,
2980 struct attr
*attr
, u_char
*pfx
, int psize
,
2981 u_int32_t addpath_id
, int withdraw
)
2983 struct prefix_rd prd
;
2984 struct prefix_evpn p
;
2985 struct bgp_route_evpn evpn
;
2988 mpls_label_t label
; /* holds the VNI as in the packet */
2991 /* Type-5 route should be 34 or 58 bytes:
2992 * RD (8), ESI (10), Eth Tag (4), IP len (1), IP (4 or 16),
2993 * GW (4 or 16) and VNI (3).
2994 * Note that the IP and GW should both be IPv4 or both IPv6.
2996 if (psize
!= 34 && psize
!= 58) {
2997 zlog_err("%u:%s - Rx EVPN Type-5 NLRI with invalid length %d",
2998 peer
->bgp
->vrf_id
, peer
->host
, psize
);
3002 /* Make prefix_rd */
3003 prd
.family
= AF_UNSPEC
;
3005 memcpy(&prd
.val
, pfx
, 8);
3008 /* Make EVPN prefix. */
3009 memset(&p
, 0, sizeof(struct prefix_evpn
));
3011 p
.prefixlen
= EVPN_TYPE_5_ROUTE_PREFIXLEN
;
3012 p
.prefix
.route_type
= BGP_EVPN_IP_PREFIX_ROUTE
;
3014 /* Additional information outside of prefix - ESI and GW IP */
3015 memset(&evpn
, 0, sizeof(evpn
));
3018 memcpy(&evpn
.eth_s_id
.val
, pfx
, 10);
3021 /* Fetch Ethernet Tag. */
3022 memcpy(ð_tag
, pfx
, 4);
3023 p
.prefix
.eth_tag
= ntohl(eth_tag
);
3026 /* Fetch IP prefix length. */
3028 if (ippfx_len
> IPV6_MAX_BITLEN
) {
3030 "%u:%s - Rx EVPN Type-5 NLRI with invalid IP Prefix length %d",
3031 peer
->bgp
->vrf_id
, peer
->host
, ippfx_len
);
3034 p
.prefix
.ip_prefix_length
= ippfx_len
;
3036 /* Determine IPv4 or IPv6 prefix */
3037 /* Since the address and GW are from the same family, this just becomes
3038 * a simple check on the total size.
3041 SET_IPADDR_V4(&p
.prefix
.ip
);
3042 memcpy(&p
.prefix
.ip
.ipaddr_v4
, pfx
, 4);
3044 memcpy(&evpn
.gw_ip
.ipv4
, pfx
, 4);
3047 SET_IPADDR_V6(&p
.prefix
.ip
);
3048 memcpy(&p
.prefix
.ip
.ipaddr_v6
, pfx
, 16);
3050 memcpy(&evpn
.gw_ip
.ipv6
, pfx
, 16);
3054 /* Get the VNI (in MPLS label field). Stored as bytes here. */
3055 memset(&label
, 0, sizeof(label
));
3056 memcpy(&label
, pfx
, BGP_LABEL_BYTES
);
3059 * If in future, we are required to access additional fields,
3060 * we MUST increment pfx by BGP_LABEL_BYTES in before reading the next field
3063 /* Process the route. */
3065 ret
= bgp_update(peer
, (struct prefix
*)&p
, addpath_id
, attr
,
3066 afi
, safi
, ZEBRA_ROUTE_BGP
, BGP_ROUTE_NORMAL
,
3067 &prd
, &label
, 1, 0, &evpn
);
3069 ret
= bgp_withdraw(peer
, (struct prefix
*)&p
, addpath_id
, attr
,
3070 afi
, safi
, ZEBRA_ROUTE_BGP
, BGP_ROUTE_NORMAL
,
3071 &prd
, &label
, 1, &evpn
);
3076 static void evpn_mpattr_encode_type5(struct stream
*s
, struct prefix
*p
,
3077 struct prefix_rd
*prd
,
3078 mpls_label_t
*label
, u_int32_t num_labels
,
3083 struct evpn_addr
*p_evpn_p
;
3085 memset(&temp
, 0, 16);
3086 if (p
->family
!= AF_EVPN
)
3088 p_evpn_p
= &(p
->u
.prefix_evpn
);
3090 /* len denites the total len of IP and GW-IP in the route
3091 IP and GW-IP have to be both ipv4 or ipv6
3093 if (IS_IPADDR_V4(&p_evpn_p
->ip
))
3094 len
= 8; /* IP and GWIP are both ipv4 */
3096 len
= 32; /* IP and GWIP are both ipv6 */
3097 /* Prefix contains RD, ESI, EthTag, IP length, IP, GWIP and VNI */
3098 stream_putc(s
, 8 + 10 + 4 + 1 + len
+ 3);
3099 stream_put(s
, prd
->val
, 8);
3101 stream_put(s
, &(attr
->evpn_overlay
.eth_s_id
), 10);
3103 stream_put(s
, &temp
, 10);
3104 stream_putl(s
, p_evpn_p
->eth_tag
);
3105 stream_putc(s
, p_evpn_p
->ip_prefix_length
);
3106 if (IS_IPADDR_V4(&p_evpn_p
->ip
))
3107 stream_put_ipv4(s
, p_evpn_p
->ip
.ipaddr_v4
.s_addr
);
3109 stream_put(s
, &p_evpn_p
->ip
.ipaddr_v6
, 16);
3111 if (IS_IPADDR_V4(&p_evpn_p
->ip
))
3113 attr
->evpn_overlay
.gw_ip
.ipv4
.s_addr
);
3115 stream_put(s
, &(attr
->evpn_overlay
.gw_ip
.ipv6
), 16);
3117 if (IS_IPADDR_V4(&p_evpn_p
->ip
))
3118 stream_put_ipv4(s
, 0);
3120 stream_put(s
, &temp
, 16);
3124 stream_put(s
, label
, 3);
3130 * Cleanup specific VNI upon EVPN (advertise-all-vni) being disabled.
3132 static void cleanup_vni_on_disable(struct hash_backet
*backet
, struct bgp
*bgp
)
3134 struct bgpevpn
*vpn
= (struct bgpevpn
*)backet
->data
;
3136 /* Remove EVPN routes and schedule for processing. */
3137 delete_routes_for_vni(bgp
, vpn
);
3139 /* Clear "live" flag and see if hash needs to be freed. */
3140 UNSET_FLAG(vpn
->flags
, VNI_FLAG_LIVE
);
3141 if (!is_vni_configured(vpn
))
3142 bgp_evpn_free(bgp
, vpn
);
3146 * Free a VNI entry; iterator function called during cleanup.
3148 static void free_vni_entry(struct hash_backet
*backet
, struct bgp
*bgp
)
3150 struct bgpevpn
*vpn
;
3152 vpn
= (struct bgpevpn
*)backet
->data
;
3153 delete_all_vni_routes(bgp
, vpn
);
3154 bgp_evpn_free(bgp
, vpn
);
3158 * Derive AUTO import RT for BGP VRF - L3VNI
3160 static void evpn_auto_rt_import_add_for_vrf(struct bgp
*bgp_vrf
)
3162 struct bgp
*bgp_def
= NULL
;
3164 form_auto_rt(bgp_vrf
, bgp_vrf
->l3vni
, bgp_vrf
->vrf_import_rtl
);
3165 UNSET_FLAG(bgp_vrf
->vrf_flags
, BGP_VRF_IMPORT_RT_CFGD
);
3168 bgp_def
= bgp_get_default();
3171 bgp_evpn_map_vrf_to_its_rts(bgp_vrf
);
3175 * Delete AUTO import RT from BGP VRF - L3VNI
3177 static void evpn_auto_rt_import_delete_for_vrf(struct bgp
*bgp_vrf
)
3179 evpn_rt_delete_auto(bgp_vrf
, bgp_vrf
->l3vni
, bgp_vrf
->vrf_import_rtl
);
3183 * Derive AUTO export RT for BGP VRF - L3VNI
3185 static void evpn_auto_rt_export_add_for_vrf(struct bgp
*bgp_vrf
)
3187 UNSET_FLAG(bgp_vrf
->vrf_flags
, BGP_VRF_EXPORT_RT_CFGD
);
3188 form_auto_rt(bgp_vrf
, bgp_vrf
->l3vni
, bgp_vrf
->vrf_export_rtl
);
3192 * Delete AUTO export RT from BGP VRF - L3VNI
3194 static void evpn_auto_rt_export_delete_for_vrf(struct bgp
*bgp_vrf
)
3196 evpn_rt_delete_auto(bgp_vrf
, bgp_vrf
->l3vni
, bgp_vrf
->vrf_export_rtl
);
3199 static void bgp_evpn_handle_export_rt_change_for_vrf(struct bgp
*bgp_vrf
)
3201 struct bgp
*bgp_def
= NULL
;
3202 struct listnode
*node
= NULL
;
3203 struct bgpevpn
*vpn
= NULL
;
3205 bgp_def
= bgp_get_default();
3209 /* update all type-5 routes */
3210 update_advertise_vrf_routes(bgp_vrf
);
3212 /* update all type-2 routes */
3213 for (ALL_LIST_ELEMENTS_RO(bgp_vrf
->l2vnis
, node
, vpn
))
3214 update_routes_for_vni(bgp_def
, vpn
);
3221 /* withdraw type-5 route corresponding to ip prefix */
3222 void bgp_evpn_withdraw_type5_route(struct bgp
*bgp_vrf
, struct prefix
*p
,
3223 afi_t afi
, safi_t safi
)
3226 struct prefix_evpn evp
;
3227 char buf
[PREFIX_STRLEN
];
3229 /* NOTE: Check needed as this is called per-route also. */
3230 if (!advertise_type5_routes(bgp_vrf
, afi
))
3233 build_type5_prefix_from_ip_prefix(&evp
, p
);
3234 ret
= delete_evpn_type5_route(bgp_vrf
, &evp
);
3237 "%u failed to delete type-5 route for prefix %s in vrf %s",
3239 prefix2str(p
, buf
, sizeof(buf
)),
3240 vrf_id_to_name(bgp_vrf
->vrf_id
));
3244 /* withdraw all type-5 routes for an address family */
3245 void bgp_evpn_withdraw_type5_routes(struct bgp
*bgp_vrf
,
3246 afi_t afi
, safi_t safi
)
3248 struct bgp_table
*table
= NULL
;
3249 struct bgp_node
*rn
= NULL
;
3250 struct bgp_info
*ri
;
3252 /* Bail out early if we don't have to advertise type-5 routes. */
3253 if (!advertise_type5_routes(bgp_vrf
, afi
))
3256 table
= bgp_vrf
->rib
[afi
][safi
];
3257 for (rn
= bgp_table_top(table
); rn
; rn
= bgp_route_next(rn
)) {
3258 /* Only care about "selected" routes - non-imported. */
3259 /* TODO: Support for AddPath for EVPN. */
3260 for (ri
= rn
->info
; ri
; ri
= ri
->next
) {
3261 if (CHECK_FLAG(ri
->flags
, BGP_INFO_SELECTED
) &&
3262 (!ri
->extra
|| !ri
->extra
->parent
)) {
3263 bgp_evpn_withdraw_type5_route(bgp_vrf
, &rn
->p
,
3272 * Advertise IP prefix as type-5 route. The afi/safi and src_attr passed
3273 * to this function correspond to those of the source IP prefix (best
3274 * path in the case of the attr. In the case of a local prefix (when we
3275 * are advertising local subnets), the src_attr will be NULL.
3277 void bgp_evpn_advertise_type5_route(struct bgp
*bgp_vrf
, struct prefix
*p
,
3278 struct attr
*src_attr
,
3279 afi_t afi
, safi_t safi
)
3282 struct prefix_evpn evp
;
3283 char buf
[PREFIX_STRLEN
];
3285 /* NOTE: Check needed as this is called per-route also. */
3286 if (!advertise_type5_routes(bgp_vrf
, afi
))
3289 build_type5_prefix_from_ip_prefix(&evp
, p
);
3290 ret
= update_evpn_type5_route(bgp_vrf
, &evp
, src_attr
);
3293 "%u: Failed to create type-5 route for prefix %s",
3295 prefix2str(p
, buf
, sizeof(buf
)));
3298 /* Inject all prefixes of a particular address-family (currently, IPv4 or
3299 * IPv6 unicast) into EVPN as type-5 routes. This is invoked when the
3300 * advertisement is enabled.
3302 void bgp_evpn_advertise_type5_routes(struct bgp
*bgp_vrf
,
3303 afi_t afi
, safi_t safi
)
3305 struct bgp_table
*table
= NULL
;
3306 struct bgp_node
*rn
= NULL
;
3307 struct bgp_info
*ri
;
3309 /* Bail out early if we don't have to advertise type-5 routes. */
3310 if (!advertise_type5_routes(bgp_vrf
, afi
))
3313 table
= bgp_vrf
->rib
[afi
][safi
];
3314 for (rn
= bgp_table_top(table
); rn
; rn
= bgp_route_next(rn
)) {
3315 /* Need to identify the "selected" route entry to use its
3316 * attribute. Also, we only consider "non-imported" routes.
3317 * TODO: Support for AddPath for EVPN.
3319 for (ri
= rn
->info
; ri
; ri
= ri
->next
) {
3320 if (CHECK_FLAG(ri
->flags
, BGP_INFO_SELECTED
) &&
3321 (!ri
->extra
|| !ri
->extra
->parent
)) {
3323 /* apply the route-map */
3324 if (bgp_vrf
->adv_cmd_rmap
[afi
][safi
].map
) {
3329 bgp_vrf
->adv_cmd_rmap
[afi
][safi
].map
,
3330 &rn
->p
, RMAP_BGP
, ri
);
3331 if (ret
== RMAP_DENYMATCH
)
3334 bgp_evpn_advertise_type5_route(bgp_vrf
, &rn
->p
,
3343 void evpn_rt_delete_auto(struct bgp
*bgp
, vni_t vni
,
3346 struct listnode
*node
, *nnode
, *node_to_del
;
3347 struct ecommunity
*ecom
, *ecom_auto
;
3348 struct ecommunity_val eval
;
3350 encode_route_target_as((bgp
->as
& 0xFFFF), vni
, &eval
);
3352 ecom_auto
= ecommunity_new();
3353 ecommunity_add_val(ecom_auto
, &eval
);
3356 for (ALL_LIST_ELEMENTS(rtl
, node
, nnode
, ecom
)) {
3357 if (ecommunity_match(ecom
, ecom_auto
)) {
3358 ecommunity_free(&ecom
);
3364 list_delete_node(rtl
, node_to_del
);
3366 ecommunity_free(&ecom_auto
);
3369 void bgp_evpn_configure_import_rt_for_vrf(struct bgp
*bgp_vrf
,
3370 struct ecommunity
*ecomadd
)
3372 /* uninstall routes from vrf */
3373 uninstall_routes_for_vrf(bgp_vrf
);
3375 /* Cleanup the RT to VRF mapping */
3376 bgp_evpn_unmap_vrf_from_its_rts(bgp_vrf
);
3378 /* Remove auto generated RT */
3379 evpn_auto_rt_import_delete_for_vrf(bgp_vrf
);
3381 /* Add the newly configured RT to RT list */
3382 listnode_add_sort(bgp_vrf
->vrf_import_rtl
, ecomadd
);
3383 SET_FLAG(bgp_vrf
->vrf_flags
, BGP_VRF_IMPORT_RT_CFGD
);
3385 /* map VRF to its RTs */
3386 bgp_evpn_map_vrf_to_its_rts(bgp_vrf
);
3388 /* install routes matching the new VRF */
3389 install_routes_for_vrf(bgp_vrf
);
3392 void bgp_evpn_unconfigure_import_rt_for_vrf(struct bgp
*bgp_vrf
,
3393 struct ecommunity
*ecomdel
)
3395 struct listnode
*node
= NULL
, *nnode
= NULL
, *node_to_del
= NULL
;
3396 struct ecommunity
*ecom
= NULL
;
3398 /* uninstall routes from vrf */
3399 uninstall_routes_for_vrf(bgp_vrf
);
3401 /* Cleanup the RT to VRF mapping */
3402 bgp_evpn_unmap_vrf_from_its_rts(bgp_vrf
);
3404 /* remove the RT from the RT list */
3405 for (ALL_LIST_ELEMENTS(bgp_vrf
->vrf_import_rtl
, node
, nnode
, ecom
)) {
3406 if (ecommunity_match(ecom
, ecomdel
)) {
3407 ecommunity_free(&ecom
);
3414 list_delete_node(bgp_vrf
->vrf_import_rtl
, node_to_del
);
3416 /* fallback to auto import rt, if this was the last RT */
3417 if (list_isempty(bgp_vrf
->vrf_import_rtl
)) {
3418 UNSET_FLAG(bgp_vrf
->vrf_flags
, BGP_VRF_IMPORT_RT_CFGD
);
3419 evpn_auto_rt_import_add_for_vrf(bgp_vrf
);
3422 /* map VRFs to its RTs */
3423 bgp_evpn_map_vrf_to_its_rts(bgp_vrf
);
3425 /* install routes matching this new RT */
3426 install_routes_for_vrf(bgp_vrf
);
3429 void bgp_evpn_configure_export_rt_for_vrf(struct bgp
*bgp_vrf
,
3430 struct ecommunity
*ecomadd
)
3432 /* remove auto-generated RT */
3433 evpn_auto_rt_export_delete_for_vrf(bgp_vrf
);
3435 /* Add the new RT to the RT list */
3436 listnode_add_sort(bgp_vrf
->vrf_export_rtl
, ecomadd
);
3437 SET_FLAG(bgp_vrf
->vrf_flags
, BGP_VRF_EXPORT_RT_CFGD
);
3439 bgp_evpn_handle_export_rt_change_for_vrf(bgp_vrf
);
3443 void bgp_evpn_unconfigure_export_rt_for_vrf(struct bgp
*bgp_vrf
,
3444 struct ecommunity
*ecomdel
)
3446 struct listnode
*node
= NULL
, *nnode
= NULL
, *node_to_del
= NULL
;
3447 struct ecommunity
*ecom
= NULL
;
3449 /* Remove the RT from the RT list */
3450 for (ALL_LIST_ELEMENTS(bgp_vrf
->vrf_export_rtl
, node
, nnode
, ecom
)) {
3451 if (ecommunity_match(ecom
, ecomdel
)) {
3452 ecommunity_free(&ecom
);
3459 list_delete_node(bgp_vrf
->vrf_export_rtl
, node_to_del
);
3461 /* fall back to auto-generated RT if this was the last RT */
3462 if (bgp_vrf
->vrf_export_rtl
&& list_isempty(bgp_vrf
->vrf_export_rtl
)) {
3463 UNSET_FLAG(bgp_vrf
->vrf_flags
, BGP_VRF_EXPORT_RT_CFGD
);
3464 evpn_auto_rt_export_add_for_vrf(bgp_vrf
);
3467 bgp_evpn_handle_export_rt_change_for_vrf(bgp_vrf
);
3471 * Handle change to BGP router id. This is invoked twice by the change
3472 * handler, first before the router id has been changed and then after
3473 * the router id has been changed. The first invocation will result in
3474 * local routes for all VNIs/VRF being deleted and withdrawn and the next
3475 * will result in the routes being re-advertised.
3477 void bgp_evpn_handle_router_id_update(struct bgp
*bgp
, int withdraw
)
3481 /* delete and withdraw all the type-5 routes
3482 stored in the global table for this vrf
3484 withdraw_router_id_vrf(bgp
);
3486 /* delete all the VNI routes (type-2/type-3) routes for all the
3489 hash_iterate(bgp
->vnihash
,
3490 (void (*)(struct hash_backet
*,
3491 void *))withdraw_router_id_vni
,
3495 /* advertise all routes in the vrf as type-5 routes with the new
3498 update_router_id_vrf(bgp
);
3500 /* advertise all the VNI routes (type-2/type-3) routes with the
3503 hash_iterate(bgp
->vnihash
,
3504 (void (*)(struct hash_backet
*,
3505 void *))update_router_id_vni
,
3511 * Handle change to export RT - update and advertise local routes.
3513 int bgp_evpn_handle_export_rt_change(struct bgp
*bgp
, struct bgpevpn
*vpn
)
3515 return update_routes_for_vni(bgp
, vpn
);
3518 void bgp_evpn_handle_vrf_rd_change(struct bgp
*bgp_vrf
,
3522 delete_withdraw_vrf_routes(bgp_vrf
);
3524 update_advertise_vrf_routes(bgp_vrf
);
3528 * Handle change to RD. This is invoked twice by the change handler,
3529 * first before the RD has been changed and then after the RD has
3530 * been changed. The first invocation will result in local routes
3531 * of this VNI being deleted and withdrawn and the next will result
3532 * in the routes being re-advertised.
3534 void bgp_evpn_handle_rd_change(struct bgp
*bgp
, struct bgpevpn
*vpn
,
3538 delete_withdraw_vni_routes(bgp
, vpn
);
3540 update_advertise_vni_routes(bgp
, vpn
);
3544 * Install routes for this VNI. Invoked upon change to Import RT.
3546 int bgp_evpn_install_routes(struct bgp
*bgp
, struct bgpevpn
*vpn
)
3548 return install_routes_for_vni(bgp
, vpn
);
3552 * Uninstall all routes installed for this VNI. Invoked upon change
3555 int bgp_evpn_uninstall_routes(struct bgp
*bgp
, struct bgpevpn
*vpn
)
3557 return uninstall_routes_for_vni(bgp
, vpn
);
3561 * TODO: Hardcoded for a maximum of 2 VNIs right now
3563 char *bgp_evpn_label2str(mpls_label_t
*label
, u_int32_t num_labels
,
3568 vni1
= label2vni(label
);
3569 if (num_labels
== 2) {
3570 vni2
= label2vni(label
+1);
3571 snprintf(buf
, len
, "%u/%u", vni1
, vni2
);
3573 snprintf(buf
, len
, "%u", vni1
);
3578 * Function to convert evpn route to json format.
3579 * NOTE: We don't use prefix2str as the output here is a bit different.
3581 void bgp_evpn_route2json(struct prefix_evpn
*p
, json_object
*json
)
3583 char buf1
[ETHER_ADDR_STRLEN
];
3584 char buf2
[PREFIX2STR_BUFFER
];
3589 if (p
->prefix
.route_type
== BGP_EVPN_IMET_ROUTE
) {
3590 json_object_int_add(json
, "routeType", p
->prefix
.route_type
);
3591 json_object_int_add(json
, "ethTag", 0);
3592 json_object_int_add(json
, "ipLen",
3593 IS_EVPN_PREFIX_IPADDR_V4(p
)
3596 json_object_string_add(json
, "ip",
3597 inet_ntoa(p
->prefix
.ip
.ipaddr_v4
));
3598 } else if (p
->prefix
.route_type
== BGP_EVPN_MAC_IP_ROUTE
) {
3599 if (IS_EVPN_PREFIX_IPADDR_NONE(p
)) {
3600 json_object_int_add(json
, "routeType",
3601 p
->prefix
.route_type
);
3602 json_object_int_add(
3604 0); /* TODO: we don't support esi yet */
3605 json_object_int_add(json
, "ethTag", 0);
3606 json_object_int_add(json
, "macLen", 8 * ETH_ALEN
);
3607 json_object_string_add(json
, "mac",
3608 prefix_mac2str(&p
->prefix
.mac
,
3614 family
= IS_EVPN_PREFIX_IPADDR_V4(p
) ? AF_INET
3617 json_object_int_add(json
, "routeType",
3618 p
->prefix
.route_type
);
3619 json_object_int_add(
3621 0); /* TODO: we don't support esi yet */
3622 json_object_int_add(json
, "ethTag", 0);
3623 json_object_int_add(json
, "macLen", 8 * ETH_ALEN
);
3624 json_object_string_add(json
, "mac",
3625 prefix_mac2str(&p
->prefix
.mac
,
3628 json_object_int_add(json
, "ipLen",
3629 IS_EVPN_PREFIX_IPADDR_V4(p
)
3632 json_object_string_add(
3634 inet_ntop(family
, &p
->prefix
.ip
.ip
.addr
, buf2
,
3635 PREFIX2STR_BUFFER
));
3638 /* Currently, this is to cater to other AF_ETHERNET code. */
3643 * Function to convert evpn route to string.
3644 * NOTE: We don't use prefix2str as the output here is a bit different.
3646 char *bgp_evpn_route2str(struct prefix_evpn
*p
, char *buf
, int len
)
3648 char buf1
[ETHER_ADDR_STRLEN
];
3649 char buf2
[PREFIX2STR_BUFFER
];
3651 if (p
->prefix
.route_type
== BGP_EVPN_IMET_ROUTE
) {
3652 snprintf(buf
, len
, "[%d]:[0]:[%d]:[%s]", p
->prefix
.route_type
,
3653 IS_EVPN_PREFIX_IPADDR_V4(p
) ? IPV4_MAX_BITLEN
3655 inet_ntoa(p
->prefix
.ip
.ipaddr_v4
));
3656 } else if (p
->prefix
.route_type
== BGP_EVPN_MAC_IP_ROUTE
) {
3657 if (IS_EVPN_PREFIX_IPADDR_NONE(p
))
3658 snprintf(buf
, len
, "[%d]:[0]:[0]:[%d]:[%s]",
3659 p
->prefix
.route_type
, 8 * ETH_ALEN
,
3660 prefix_mac2str(&p
->prefix
.mac
, buf1
,
3665 family
= IS_EVPN_PREFIX_IPADDR_V4(p
) ? AF_INET
3667 snprintf(buf
, len
, "[%d]:[0]:[0]:[%d]:[%s]:[%d]:[%s]",
3668 p
->prefix
.route_type
, 8 * ETH_ALEN
,
3669 prefix_mac2str(&p
->prefix
.mac
, buf1
,
3671 family
== AF_INET
? IPV4_MAX_BITLEN
3673 inet_ntop(family
, &p
->prefix
.ip
.ip
.addr
, buf2
,
3674 PREFIX2STR_BUFFER
));
3676 } else if (p
->prefix
.route_type
== BGP_EVPN_IP_PREFIX_ROUTE
) {
3677 snprintf(buf
, len
, "[%d]:[0]:[0]:[%d]:[%s]",
3678 p
->prefix
.route_type
,
3679 p
->prefix
.ip_prefix_length
,
3680 IS_EVPN_PREFIX_IPADDR_V4(p
) ?
3681 inet_ntoa(p
->prefix
.ip
.ipaddr_v4
) :
3682 inet6_ntoa(p
->prefix
.ip
.ipaddr_v6
));
3684 /* For EVPN route types not supported yet. */
3685 snprintf(buf
, len
, "(unsupported route type %d)",
3686 p
->prefix
.route_type
);
3693 * Encode EVPN prefix in Update (MP_REACH)
3695 void bgp_evpn_encode_prefix(struct stream
*s
, struct prefix
*p
,
3696 struct prefix_rd
*prd
,
3697 mpls_label_t
*label
, u_int32_t num_labels
,
3698 struct attr
*attr
, int addpath_encode
,
3699 u_int32_t addpath_tx_id
)
3701 struct prefix_evpn
*evp
= (struct prefix_evpn
*)p
;
3702 int len
, ipa_len
= 0;
3705 stream_putl(s
, addpath_tx_id
);
3708 stream_putc(s
, evp
->prefix
.route_type
);
3710 switch (evp
->prefix
.route_type
) {
3711 case BGP_EVPN_MAC_IP_ROUTE
:
3712 if (IS_EVPN_PREFIX_IPADDR_V4(evp
))
3713 ipa_len
= IPV4_MAX_BYTELEN
;
3714 else if (IS_EVPN_PREFIX_IPADDR_V6(evp
))
3715 ipa_len
= IPV6_MAX_BYTELEN
;
3716 /* RD, ESI, EthTag, MAC+len, IP len, [IP], 1 VNI */
3717 len
= 8 + 10 + 4 + 1 + 6 + 1 + ipa_len
+ 3;
3718 if (ipa_len
&& num_labels
> 1) /* There are 2 VNIs */
3720 stream_putc(s
, len
);
3721 stream_put(s
, prd
->val
, 8); /* RD */
3722 stream_put(s
, 0, 10); /* ESI */
3723 stream_putl(s
, 0); /* Ethernet Tag ID */
3724 stream_putc(s
, 8 * ETH_ALEN
); /* Mac Addr Len - bits */
3725 stream_put(s
, evp
->prefix
.mac
.octet
, 6); /* Mac Addr */
3726 stream_putc(s
, 8 * ipa_len
); /* IP address Length */
3727 if (ipa_len
) /* IP */
3728 stream_put(s
, &evp
->prefix
.ip
.ip
.addr
, ipa_len
);
3729 /* 1st label is the L2 VNI */
3730 stream_put(s
, label
, BGP_LABEL_BYTES
);
3731 /* Include 2nd label (L3 VNI) if advertising MAC+IP */
3732 if (ipa_len
&& num_labels
> 1)
3733 stream_put(s
, label
+1, BGP_LABEL_BYTES
);
3736 case BGP_EVPN_IMET_ROUTE
:
3737 stream_putc(s
, 17); // TODO: length - assumes IPv4 address
3738 stream_put(s
, prd
->val
, 8); /* RD */
3739 stream_putl(s
, 0); /* Ethernet Tag ID */
3740 stream_putc(s
, IPV4_MAX_BITLEN
); /* IP address Length - bits */
3741 /* Originating Router's IP Addr */
3742 stream_put_in_addr(s
, &evp
->prefix
.ip
.ipaddr_v4
);
3745 case BGP_EVPN_IP_PREFIX_ROUTE
:
3746 /* TODO: AddPath support. */
3747 evpn_mpattr_encode_type5(s
, p
, prd
, label
, num_labels
, attr
);
3755 int bgp_nlri_parse_evpn(struct peer
*peer
, struct attr
*attr
,
3756 struct bgp_nlri
*packet
, int withdraw
)
3762 u_int32_t addpath_id
;
3763 int addpath_encoded
;
3769 /* Check peer status. */
3770 if (peer
->status
!= Established
) {
3771 zlog_err("%u:%s - EVPN update received in state %d",
3772 peer
->bgp
->vrf_id
, peer
->host
, peer
->status
);
3776 /* Start processing the NLRI - there may be multiple in the MP_REACH */
3778 lim
= pnt
+ packet
->length
;
3780 safi
= packet
->safi
;
3784 (CHECK_FLAG(peer
->af_cap
[afi
][safi
], PEER_CAP_ADDPATH_AF_RX_ADV
)
3785 && CHECK_FLAG(peer
->af_cap
[afi
][safi
],
3786 PEER_CAP_ADDPATH_AF_TX_RCV
));
3788 for (; pnt
< lim
; pnt
+= psize
) {
3789 /* Clear prefix structure. */
3790 memset(&p
, 0, sizeof(struct prefix
));
3792 /* Deal with path-id if AddPath is supported. */
3793 if (addpath_encoded
) {
3794 /* When packet overflow occurs return immediately. */
3795 if (pnt
+ BGP_ADDPATH_ID_LEN
> lim
)
3798 addpath_id
= ntohl(*((uint32_t *)pnt
));
3799 pnt
+= BGP_ADDPATH_ID_LEN
;
3802 /* All EVPN NLRI types start with type and length. */
3807 psize
= rlen
= *pnt
++;
3809 /* When packet overflow occur return immediately. */
3810 if (pnt
+ psize
> lim
)
3814 case BGP_EVPN_MAC_IP_ROUTE
:
3815 if (process_type2_route(peer
, afi
, safi
,
3816 withdraw
? NULL
: attr
, pnt
,
3817 psize
, addpath_id
)) {
3819 "%u:%s - Error in processing EVPN type-2 NLRI size %d",
3820 peer
->bgp
->vrf_id
, peer
->host
, psize
);
3825 case BGP_EVPN_IMET_ROUTE
:
3826 if (process_type3_route(peer
, afi
, safi
,
3827 withdraw
? NULL
: attr
, pnt
,
3828 psize
, addpath_id
)) {
3830 "%u:%s - Error in processing EVPN type-3 NLRI size %d",
3831 peer
->bgp
->vrf_id
, peer
->host
, psize
);
3836 case BGP_EVPN_IP_PREFIX_ROUTE
:
3837 if (process_type5_route(peer
, afi
, safi
, attr
, pnt
,
3838 psize
, addpath_id
, withdraw
)) {
3840 "%u:%s - Error in processing EVPN type-5 NLRI size %d",
3841 peer
->bgp
->vrf_id
, peer
->host
, psize
);
3851 /* Packet length consistency check. */
3859 * Map the RTs (configured or automatically derived) of a VRF to the VRF.
3860 * The mapping will be used during route processing.
3861 * bgp_def: default bgp instance
3862 * bgp_vrf: specific bgp vrf instance on which RT is configured
3864 void bgp_evpn_map_vrf_to_its_rts(struct bgp
*bgp_vrf
)
3867 struct ecommunity_val
*eval
= NULL
;
3868 struct listnode
*node
= NULL
, *nnode
= NULL
;
3869 struct ecommunity
*ecom
= NULL
;
3871 for (ALL_LIST_ELEMENTS(bgp_vrf
->vrf_import_rtl
, node
, nnode
, ecom
)) {
3872 for (i
= 0; i
< ecom
->size
; i
++) {
3873 eval
= (struct ecommunity_val
*)(ecom
->val
3875 * ECOMMUNITY_SIZE
));
3876 map_vrf_to_rt(bgp_vrf
, eval
);
3882 * Unmap the RTs (configured or automatically derived) of a VRF from the VRF.
3884 void bgp_evpn_unmap_vrf_from_its_rts(struct bgp
*bgp_vrf
)
3887 struct ecommunity_val
*eval
;
3888 struct listnode
*node
, *nnode
;
3889 struct ecommunity
*ecom
;
3891 for (ALL_LIST_ELEMENTS(bgp_vrf
->vrf_import_rtl
, node
, nnode
, ecom
)) {
3892 for (i
= 0; i
< ecom
->size
; i
++) {
3893 struct vrf_irt_node
*irt
;
3894 struct ecommunity_val eval_tmp
;
3896 eval
= (struct ecommunity_val
*)(ecom
->val
3898 * ECOMMUNITY_SIZE
));
3899 /* If using "automatic" RT, we only care about the
3900 * local-admin sub-field.
3901 * This is to facilitate using VNI as the RT for EBGP
3904 memcpy(&eval_tmp
, eval
, ECOMMUNITY_SIZE
);
3905 if (!CHECK_FLAG(bgp_vrf
->vrf_flags
,
3906 BGP_VRF_IMPORT_RT_CFGD
))
3907 mask_ecom_global_admin(&eval_tmp
, eval
);
3909 irt
= lookup_vrf_import_rt(&eval_tmp
);
3911 unmap_vrf_from_rt(bgp_vrf
, irt
);
3919 * Map the RTs (configured or automatically derived) of a VNI to the VNI.
3920 * The mapping will be used during route processing.
3922 void bgp_evpn_map_vni_to_its_rts(struct bgp
*bgp
, struct bgpevpn
*vpn
)
3925 struct ecommunity_val
*eval
;
3926 struct listnode
*node
, *nnode
;
3927 struct ecommunity
*ecom
;
3929 for (ALL_LIST_ELEMENTS(vpn
->import_rtl
, node
, nnode
, ecom
)) {
3930 for (i
= 0; i
< ecom
->size
; i
++) {
3931 eval
= (struct ecommunity_val
*)(ecom
->val
3933 * ECOMMUNITY_SIZE
));
3934 map_vni_to_rt(bgp
, vpn
, eval
);
3940 * Unmap the RTs (configured or automatically derived) of a VNI from the VNI.
3942 void bgp_evpn_unmap_vni_from_its_rts(struct bgp
*bgp
, struct bgpevpn
*vpn
)
3945 struct ecommunity_val
*eval
;
3946 struct listnode
*node
, *nnode
;
3947 struct ecommunity
*ecom
;
3949 for (ALL_LIST_ELEMENTS(vpn
->import_rtl
, node
, nnode
, ecom
)) {
3950 for (i
= 0; i
< ecom
->size
; i
++) {
3951 struct irt_node
*irt
;
3952 struct ecommunity_val eval_tmp
;
3954 eval
= (struct ecommunity_val
*)(ecom
->val
3956 * ECOMMUNITY_SIZE
));
3957 /* If using "automatic" RT, we only care about the
3958 * local-admin sub-field.
3959 * This is to facilitate using VNI as the RT for EBGP
3962 memcpy(&eval_tmp
, eval
, ECOMMUNITY_SIZE
);
3963 if (!is_import_rt_configured(vpn
))
3964 mask_ecom_global_admin(&eval_tmp
, eval
);
3966 irt
= lookup_import_rt(bgp
, &eval_tmp
);
3968 unmap_vni_from_rt(bgp
, vpn
, irt
);
3974 * Derive Import RT automatically for VNI and map VNI to RT.
3975 * The mapping will be used during route processing.
3977 void bgp_evpn_derive_auto_rt_import(struct bgp
*bgp
, struct bgpevpn
*vpn
)
3979 form_auto_rt(bgp
, vpn
->vni
, vpn
->import_rtl
);
3980 UNSET_FLAG(vpn
->flags
, VNI_FLAG_IMPRT_CFGD
);
3983 bgp_evpn_map_vni_to_its_rts(bgp
, vpn
);
3987 * Derive Export RT automatically for VNI.
3989 void bgp_evpn_derive_auto_rt_export(struct bgp
*bgp
, struct bgpevpn
*vpn
)
3991 form_auto_rt(bgp
, vpn
->vni
, vpn
->export_rtl
);
3992 UNSET_FLAG(vpn
->flags
, VNI_FLAG_EXPRT_CFGD
);
3996 * Derive RD automatically for VNI using passed information - it
3997 * is of the form RouterId:unique-id-for-vni.
3999 void bgp_evpn_derive_auto_rd_for_vrf(struct bgp
*bgp
)
4003 bgp
->vrf_prd
.family
= AF_UNSPEC
;
4004 bgp
->vrf_prd
.prefixlen
= 64;
4005 sprintf(buf
, "%s:%hu", inet_ntoa(bgp
->router_id
), bgp
->vrf_rd_id
);
4006 str2prefix_rd(buf
, &bgp
->vrf_prd
);
4010 * Derive RD automatically for VNI using passed information - it
4011 * is of the form RouterId:unique-id-for-vni.
4013 void bgp_evpn_derive_auto_rd(struct bgp
*bgp
, struct bgpevpn
*vpn
)
4017 vpn
->prd
.family
= AF_UNSPEC
;
4018 vpn
->prd
.prefixlen
= 64;
4019 sprintf(buf
, "%s:%hu", inet_ntoa(bgp
->router_id
), vpn
->rd_id
);
4020 (void)str2prefix_rd(buf
, &vpn
->prd
);
4021 UNSET_FLAG(vpn
->flags
, VNI_FLAG_RD_CFGD
);
4027 struct bgpevpn
*bgp_evpn_lookup_vni(struct bgp
*bgp
, vni_t vni
)
4029 struct bgpevpn
*vpn
;
4032 memset(&tmp
, 0, sizeof(struct bgpevpn
));
4034 vpn
= hash_lookup(bgp
->vnihash
, &tmp
);
4039 * Create a new vpn - invoked upon configuration or zebra notification.
4041 struct bgpevpn
*bgp_evpn_new(struct bgp
*bgp
, vni_t vni
,
4042 struct in_addr originator_ip
,
4043 vrf_id_t tenant_vrf_id
)
4045 struct bgpevpn
*vpn
;
4050 vpn
= XCALLOC(MTYPE_BGP_EVPN
, sizeof(struct bgpevpn
));
4054 /* Set values - RD and RT set to defaults. */
4056 vpn
->originator_ip
= originator_ip
;
4057 vpn
->tenant_vrf_id
= tenant_vrf_id
;
4059 /* Initialize route-target import and export lists */
4060 vpn
->import_rtl
= list_new();
4061 vpn
->import_rtl
->cmp
= (int (*)(void *, void *))evpn_route_target_cmp
;
4062 vpn
->export_rtl
= list_new();
4063 vpn
->export_rtl
->cmp
= (int (*)(void *, void *))evpn_route_target_cmp
;
4064 bf_assign_index(bm
->rd_idspace
, vpn
->rd_id
);
4065 derive_rd_rt_for_vni(bgp
, vpn
);
4067 /* Initialize EVPN route table. */
4068 vpn
->route_table
= bgp_table_init(AFI_L2VPN
, SAFI_EVPN
);
4071 if (!hash_get(bgp
->vnihash
, vpn
, hash_alloc_intern
)) {
4072 XFREE(MTYPE_BGP_EVPN
, vpn
);
4076 /* add to l2vni list on corresponding vrf */
4077 bgpevpn_link_to_l3vni(vpn
);
4079 QOBJ_REG(vpn
, bgpevpn
);
4084 * Free a given VPN - called in multiple scenarios such as zebra
4085 * notification, configuration being deleted, advertise-all-vni disabled etc.
4086 * This just frees appropriate memory, caller should have taken other
4089 void bgp_evpn_free(struct bgp
*bgp
, struct bgpevpn
*vpn
)
4091 bgpevpn_unlink_from_l3vni(vpn
);
4092 bgp_table_unlock(vpn
->route_table
);
4093 bgp_evpn_unmap_vni_from_its_rts(bgp
, vpn
);
4094 list_delete_and_null(&vpn
->import_rtl
);
4095 list_delete_and_null(&vpn
->export_rtl
);
4096 bf_release_index(bm
->rd_idspace
, vpn
->rd_id
);
4097 hash_release(bgp
->vnihash
, vpn
);
4099 XFREE(MTYPE_BGP_EVPN
, vpn
);
4103 * Import route into matching VNI(s).
4105 int bgp_evpn_import_route(struct bgp
*bgp
, afi_t afi
, safi_t safi
,
4106 struct prefix
*p
, struct bgp_info
*ri
)
4108 return install_uninstall_evpn_route(bgp
, afi
, safi
, p
, ri
, 1);
4112 * Unimport route from matching VNI(s).
4114 int bgp_evpn_unimport_route(struct bgp
*bgp
, afi_t afi
, safi_t safi
,
4115 struct prefix
*p
, struct bgp_info
*ri
)
4117 return install_uninstall_evpn_route(bgp
, afi
, safi
, p
, ri
, 0);
4120 /* filter routes which have martian next hops */
4121 int bgp_filter_evpn_routes_upon_martian_nh_change(struct bgp
*bgp
)
4125 struct bgp_node
*rd_rn
, *rn
;
4126 struct bgp_table
*table
;
4127 struct bgp_info
*ri
;
4132 /* Walk entire global routing table and evaluate routes which could be
4133 * imported into this VPN. Note that we cannot just look at the routes
4134 * for the VNI's RD -
4135 * remote routes applicable for this VNI could have any RD.
4137 /* EVPN routes are a 2-level table. */
4138 for (rd_rn
= bgp_table_top(bgp
->rib
[afi
][safi
]); rd_rn
;
4139 rd_rn
= bgp_route_next(rd_rn
)) {
4140 table
= (struct bgp_table
*)(rd_rn
->info
);
4144 for (rn
= bgp_table_top(table
); rn
; rn
= bgp_route_next(rn
)) {
4146 for (ri
= rn
->info
; ri
; ri
= ri
->next
) {
4148 /* Consider "valid" remote routes applicable for
4150 if (!(ri
->type
== ZEBRA_ROUTE_BGP
4151 && ri
->sub_type
== BGP_ROUTE_NORMAL
))
4154 if (bgp_nexthop_self(bgp
, ri
->attr
->nexthop
)) {
4156 char attr_str
[BUFSIZ
];
4157 char pbuf
[PREFIX_STRLEN
];
4159 bgp_dump_attr(ri
->attr
, attr_str
,
4162 if (bgp_debug_update(ri
->peer
, &rn
->p
,
4165 "%u: prefix %s with attr %s - DENIED due to martian or self nexthop",
4172 bgp_evpn_unimport_route(bgp
, afi
, safi
,
4175 bgp_rib_remove(rn
, ri
, ri
->peer
, afi
,
4186 * Handle del of a local MACIP.
4188 int bgp_evpn_local_macip_del(struct bgp
*bgp
, vni_t vni
, struct ethaddr
*mac
,
4191 struct bgpevpn
*vpn
;
4192 struct prefix_evpn p
;
4194 if (!bgp
->vnihash
) {
4195 zlog_err("%u: VNI hash not created", bgp
->vrf_id
);
4199 /* Lookup VNI hash - should exist. */
4200 vpn
= bgp_evpn_lookup_vni(bgp
, vni
);
4201 if (!vpn
|| !is_vni_live(vpn
)) {
4202 zlog_warn("%u: VNI hash entry for VNI %u %s at MACIP DEL",
4203 bgp
->vrf_id
, vni
, vpn
? "not live" : "not found");
4207 /* Remove EVPN type-2 route and schedule for processing. */
4208 build_evpn_type2_prefix(&p
, mac
, ip
);
4209 delete_evpn_route(bgp
, vpn
, &p
);
4215 * Handle add of a local MACIP.
4217 int bgp_evpn_local_macip_add(struct bgp
*bgp
, vni_t vni
, struct ethaddr
*mac
,
4218 struct ipaddr
*ip
, u_char flags
)
4220 struct bgpevpn
*vpn
;
4221 struct prefix_evpn p
;
4223 if (!bgp
->vnihash
) {
4224 zlog_err("%u: VNI hash not created", bgp
->vrf_id
);
4228 /* Lookup VNI hash - should exist. */
4229 vpn
= bgp_evpn_lookup_vni(bgp
, vni
);
4230 if (!vpn
|| !is_vni_live(vpn
)) {
4231 zlog_warn("%u: VNI hash entry for VNI %u %s at MACIP ADD",
4232 bgp
->vrf_id
, vni
, vpn
? "not live" : "not found");
4236 /* Create EVPN type-2 route and schedule for processing. */
4237 build_evpn_type2_prefix(&p
, mac
, ip
);
4238 if (update_evpn_route(bgp
, vpn
, &p
, flags
)) {
4239 char buf
[ETHER_ADDR_STRLEN
];
4240 char buf2
[INET6_ADDRSTRLEN
];
4243 "%u:Failed to create Type-2 route, VNI %u %s MAC %s IP %s (flags: 0x%x)",
4244 bgp
->vrf_id
, vpn
->vni
,
4245 CHECK_FLAG(flags
, ZEBRA_MACIP_TYPE_STICKY
) ? "sticky gateway"
4247 prefix_mac2str(mac
, buf
, sizeof(buf
)),
4248 ipaddr2str(ip
, buf2
, sizeof(buf2
)),
4256 static void link_l2vni_hash_to_l3vni(struct hash_backet
*backet
,
4257 struct bgp
*bgp_vrf
)
4259 struct bgpevpn
*vpn
= NULL
;
4260 struct bgp
*bgp_def
= NULL
;
4262 bgp_def
= bgp_get_default();
4265 vpn
= (struct bgpevpn
*)backet
->data
;
4266 if (vpn
->tenant_vrf_id
== bgp_vrf
->vrf_id
)
4267 bgpevpn_link_to_l3vni(vpn
);
4270 int bgp_evpn_local_l3vni_add(vni_t l3vni
,
4272 struct ethaddr
*rmac
,
4273 struct in_addr originator_ip
,
4276 struct bgp
*bgp_vrf
= NULL
; /* bgp VRF instance */
4277 struct bgp
*bgp_def
= NULL
; /* default bgp instance */
4278 struct listnode
*node
= NULL
;
4279 struct bgpevpn
*vpn
= NULL
;
4282 /* get the default instamce - required to get the AS number for VRF
4285 bgp_def
= bgp_get_default();
4287 zlog_err("Cannot process L3VNI %u ADD - default BGP instance not yet created",
4293 /* if the BGP vrf instance doesnt exist - create one */
4294 bgp_vrf
= bgp_lookup_by_name(vrf_id_to_name(vrf_id
));
4299 ret
= bgp_get(&bgp_vrf
, &as
, vrf_id_to_name(vrf_id
),
4300 BGP_INSTANCE_TYPE_VRF
);
4302 case BGP_ERR_MULTIPLE_INSTANCE_NOT_SET
:
4303 zlog_err("'bgp multiple-instance' not present\n");
4305 case BGP_ERR_AS_MISMATCH
:
4306 zlog_err("BGP is already running; AS is %u\n", as
);
4308 case BGP_ERR_INSTANCE_MISMATCH
:
4309 zlog_err("BGP instance name and AS number mismatch\n");
4313 /* mark as auto created */
4314 SET_FLAG(bgp_vrf
->vrf_flags
, BGP_VRF_AUTO
);
4317 /* associate with l3vni */
4318 bgp_vrf
->l3vni
= l3vni
;
4320 /* set the router mac - to be used in mac-ip routes for this vrf */
4321 memcpy(&bgp_vrf
->rmac
, rmac
, sizeof(struct ethaddr
));
4323 /* set the originator ip */
4324 bgp_vrf
->originator_ip
= originator_ip
;
4326 /* set the right filter - are we using l3vni only for prefix routes? */
4328 SET_FLAG(bgp_vrf
->vrf_flags
, BGP_VRF_L3VNI_PREFIX_ROUTES_ONLY
);
4330 /* auto derive RD/RT */
4331 if (!CHECK_FLAG(bgp_vrf
->vrf_flags
, BGP_VRF_IMPORT_RT_CFGD
))
4332 evpn_auto_rt_import_add_for_vrf(bgp_vrf
);
4333 if (!CHECK_FLAG(bgp_vrf
->vrf_flags
, BGP_VRF_EXPORT_RT_CFGD
))
4334 evpn_auto_rt_export_add_for_vrf(bgp_vrf
);
4335 bgp_evpn_derive_auto_rd_for_vrf(bgp_vrf
);
4337 /* link all corresponding l2vnis */
4338 hash_iterate(bgp_def
->vnihash
,
4339 (void (*)(struct hash_backet
*, void *))
4340 link_l2vni_hash_to_l3vni
,
4343 /* Only update all corresponding type-2 routes if we are advertising two
4344 * labels along with type-2 routes
4347 for (ALL_LIST_ELEMENTS_RO(bgp_vrf
->l2vnis
, node
, vpn
))
4348 update_routes_for_vni(bgp_def
, vpn
);
4350 /* advertise type-5 routes if needed */
4351 update_advertise_vrf_routes(bgp_vrf
);
4353 /* install all remote routes belonging to this l3vni into correspondng
4355 install_routes_for_vrf(bgp_vrf
);
4360 int bgp_evpn_local_l3vni_del(vni_t l3vni
,
4363 struct bgp
*bgp_vrf
= NULL
; /* bgp vrf instance */
4364 struct bgp
*bgp_def
= NULL
; /* default bgp instance */
4365 struct listnode
*node
= NULL
;
4366 struct bgpevpn
*vpn
= NULL
;
4368 bgp_vrf
= bgp_lookup_by_vrf_id(vrf_id
);
4370 zlog_err("Cannot process L3VNI %u Del - Could not find BGP instance",
4375 bgp_def
= bgp_get_default();
4377 zlog_err("Cannot process L3VNI %u Del - Could not find default BGP instance",
4382 /* unimport remote routes from VRF, if it is AUTO vrf bgp_delete will
4383 * take care of uninstalling the routes from zebra
4385 if (!CHECK_FLAG(bgp_vrf
->vrf_flags
, BGP_VRF_AUTO
))
4386 uninstall_routes_for_vrf(bgp_vrf
);
4388 /* delete/withdraw all type-5 routes */
4389 delete_withdraw_vrf_routes(bgp_vrf
);
4391 /* remove the l3vni from vrf instance */
4394 /* remove the Rmac from the BGP vrf */
4395 memset(&bgp_vrf
->rmac
, 0, sizeof(struct ethaddr
));
4398 if (bgp_vrf
->vrf_import_rtl
&& !list_isempty(bgp_vrf
->vrf_import_rtl
)) {
4399 bgp_evpn_unmap_vrf_from_its_rts(bgp_vrf
);
4400 list_delete_all_node(bgp_vrf
->vrf_import_rtl
);
4402 if (bgp_vrf
->vrf_export_rtl
&& !list_isempty(bgp_vrf
->vrf_export_rtl
)) {
4403 list_delete_all_node(bgp_vrf
->vrf_export_rtl
);
4406 /* update all corresponding local mac-ip routes */
4407 if (!CHECK_FLAG(bgp_vrf
->vrf_flags
, BGP_VRF_L3VNI_PREFIX_ROUTES_ONLY
)) {
4408 for (ALL_LIST_ELEMENTS_RO(bgp_vrf
->l2vnis
, node
, vpn
)) {
4409 UNSET_FLAG(vpn
->flags
, VNI_FLAG_USE_TWO_LABELS
);
4410 update_routes_for_vni(bgp_def
, vpn
);
4414 /* Delete the instance if it was autocreated */
4415 if (CHECK_FLAG(bgp_vrf
->vrf_flags
, BGP_VRF_AUTO
))
4416 bgp_delete(bgp_vrf
);
4422 * Handle del of a local VNI.
4424 int bgp_evpn_local_vni_del(struct bgp
*bgp
, vni_t vni
)
4426 struct bgpevpn
*vpn
;
4428 if (!bgp
->vnihash
) {
4429 zlog_err("%u: VNI hash not created", bgp
->vrf_id
);
4433 /* Locate VNI hash */
4434 vpn
= bgp_evpn_lookup_vni(bgp
, vni
);
4436 zlog_warn("%u: VNI hash entry for VNI %u not found at DEL",
4441 /* Remove all local EVPN routes and schedule for processing (to
4442 * withdraw from peers).
4444 delete_routes_for_vni(bgp
, vpn
);
4447 * tunnel is no longer active, del tunnel ip address from tip_hash
4449 bgp_tip_del(bgp
, &vpn
->originator_ip
);
4451 /* Clear "live" flag and see if hash needs to be freed. */
4452 UNSET_FLAG(vpn
->flags
, VNI_FLAG_LIVE
);
4453 if (!is_vni_configured(vpn
))
4454 bgp_evpn_free(bgp
, vpn
);
4460 * Handle add (or update) of a local VNI. The VNI changes we care
4461 * about are for the local-tunnel-ip and the (tenant) VRF.
4463 int bgp_evpn_local_vni_add(struct bgp
*bgp
, vni_t vni
,
4464 struct in_addr originator_ip
,
4465 vrf_id_t tenant_vrf_id
)
4467 struct bgpevpn
*vpn
;
4468 struct prefix_evpn p
;
4470 if (!bgp
->vnihash
) {
4471 zlog_err("%u: VNI hash not created", bgp
->vrf_id
);
4475 /* Lookup VNI. If present and no change, exit. */
4476 vpn
= bgp_evpn_lookup_vni(bgp
, vni
);
4479 if (is_vni_live(vpn
)
4480 && IPV4_ADDR_SAME(&vpn
->originator_ip
, &originator_ip
)
4481 && vpn
->tenant_vrf_id
== tenant_vrf_id
)
4482 /* Probably some other param has changed that we don't
4486 /* Update tenant_vrf_id if it has changed. */
4487 if (vpn
->tenant_vrf_id
!= tenant_vrf_id
) {
4488 bgpevpn_unlink_from_l3vni(vpn
);
4489 vpn
->tenant_vrf_id
= tenant_vrf_id
;
4490 bgpevpn_link_to_l3vni(vpn
);
4493 /* If tunnel endpoint IP has changed, update (and delete prior
4494 * type-3 route, if needed.)
4496 if (!IPV4_ADDR_SAME(&vpn
->originator_ip
, &originator_ip
))
4497 handle_tunnel_ip_change(bgp
, vpn
, originator_ip
);
4499 /* Update all routes with new endpoint IP and/or export RT
4502 if (is_vni_live(vpn
))
4503 update_routes_for_vni(bgp
, vpn
);
4506 /* Create or update as appropriate. */
4508 vpn
= bgp_evpn_new(bgp
, vni
, originator_ip
, tenant_vrf_id
);
4511 "%u: Failed to allocate VNI entry for VNI %u - at Add",
4517 /* if the VNI is live already, there is nothing more to do */
4518 if (is_vni_live(vpn
))
4521 /* Mark as "live" */
4522 SET_FLAG(vpn
->flags
, VNI_FLAG_LIVE
);
4524 /* tunnel is now active, add tunnel-ip to db */
4525 bgp_tip_add(bgp
, &originator_ip
);
4527 /* filter routes as nexthop database has changed */
4528 bgp_filter_evpn_routes_upon_martian_nh_change(bgp
);
4530 /* Create EVPN type-3 route and schedule for processing. */
4531 build_evpn_type3_prefix(&p
, vpn
->originator_ip
);
4532 if (update_evpn_route(bgp
, vpn
, &p
, 0)) {
4533 zlog_err("%u: Type3 route creation failure for VNI %u",
4538 /* If we have learnt and retained remote routes (VTEPs, MACs) for this
4542 install_routes_for_vni(bgp
, vpn
);
4544 /* If we are advertising gateway mac-ip
4545 It needs to be conveyed again to zebra */
4546 bgp_zebra_advertise_gw_macip(bgp
, vpn
->advertise_gw_macip
, vpn
->vni
);
4552 * Cleanup EVPN information on disable - Need to delete and withdraw
4553 * EVPN routes from peers.
4555 void bgp_evpn_cleanup_on_disable(struct bgp
*bgp
)
4557 hash_iterate(bgp
->vnihash
, (void (*)(struct hash_backet
*,
4558 void *))cleanup_vni_on_disable
,
4563 * Cleanup EVPN information - invoked at the time of bgpd exit or when the
4564 * BGP instance (default) is being freed.
4566 void bgp_evpn_cleanup(struct bgp
*bgp
)
4569 hash_iterate(bgp
->vnihash
, (void (*)(struct hash_backet
*,
4570 void *))free_vni_entry
,
4572 if (bgp
->import_rt_hash
)
4573 hash_free(bgp
->import_rt_hash
);
4574 bgp
->import_rt_hash
= NULL
;
4575 if (bgp
->vrf_import_rt_hash
)
4576 hash_free(bgp
->vrf_import_rt_hash
);
4577 bgp
->vrf_import_rt_hash
= NULL
;
4579 hash_free(bgp
->vnihash
);
4580 bgp
->vnihash
= NULL
;
4581 if (bgp
->vrf_import_rtl
)
4582 list_delete_and_null(&bgp
->vrf_import_rtl
);
4583 if (bgp
->vrf_export_rtl
)
4584 list_delete_and_null(&bgp
->vrf_export_rtl
);
4586 list_delete_and_null(&bgp
->l2vnis
);
4587 bf_release_index(bm
->rd_idspace
, bgp
->vrf_rd_id
);
4591 * Initialization for EVPN
4594 * hash for RT to VNI
4595 * assign a unique rd id for auto derivation of vrf_prd
4597 void bgp_evpn_init(struct bgp
*bgp
)
4600 hash_create(vni_hash_key_make
, vni_hash_cmp
, "BGP VNI Hash");
4601 bgp
->import_rt_hash
=
4602 hash_create(import_rt_hash_key_make
, import_rt_hash_cmp
,
4603 "BGP Import RT Hash");
4604 bgp
->vrf_import_rt_hash
=
4605 hash_create(vrf_import_rt_hash_key_make
, vrf_import_rt_hash_cmp
,
4606 "BGP VRF Import RT Hash");
4607 bgp
->vrf_import_rtl
= list_new();
4608 bgp
->vrf_import_rtl
->cmp
=
4609 (int (*)(void *, void *))evpn_route_target_cmp
;
4611 bgp
->vrf_export_rtl
= list_new();
4612 bgp
->vrf_export_rtl
->cmp
=
4613 (int (*)(void *, void *))evpn_route_target_cmp
;
4614 bgp
->l2vnis
= list_new();
4616 (int (*)(void *, void *))vni_hash_cmp
;
4617 bf_assign_index(bm
->rd_idspace
, bgp
->vrf_rd_id
);
4621 void bgp_evpn_vrf_delete(struct bgp
*bgp_vrf
)
4623 bgp_evpn_unmap_vrf_from_its_rts(bgp_vrf
);