1 /* BGP Nexthop tracking
2 * Copyright (C) 2013 Cumulus Networks, Inc.
4 * This file is part of GNU Zebra.
6 * GNU Zebra is free software; you can redistribute it and/or modify it
7 * under the terms of the GNU General Public License as published by the
8 * Free Software Foundation; either version 2, or (at your option) any
11 * GNU Zebra is distributed in the hope that it will be useful, but
12 * WITHOUT ANY WARRANTY; without even the implied warranty of
13 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
14 * General Public License for more details.
16 * You should have received a copy of the GNU General Public License along
17 * with this program; see the file COPYING; if not, write to the Free Software
18 * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
34 #include "nexthop_group.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_nexthop.h"
41 #include "bgpd/bgp_debug.h"
42 #include "bgpd/bgp_errors.h"
43 #include "bgpd/bgp_nht.h"
44 #include "bgpd/bgp_fsm.h"
45 #include "bgpd/bgp_zebra.h"
46 #include "bgpd/bgp_flowspec_util.h"
47 #include "bgpd/bgp_evpn.h"
48 #include "bgpd/bgp_rd.h"
50 extern struct zclient
*zclient
;
52 static void register_zebra_rnh(struct bgp_nexthop_cache
*bnc
);
53 static void unregister_zebra_rnh(struct bgp_nexthop_cache
*bnc
);
54 static int make_prefix(int afi
, struct bgp_path_info
*pi
, struct prefix
*p
);
55 static void bgp_nht_ifp_initial(struct thread
*thread
);
57 static int bgp_isvalid_nexthop(struct bgp_nexthop_cache
*bnc
)
59 return (bgp_zebra_num_connects() == 0
60 || (bnc
&& CHECK_FLAG(bnc
->flags
, BGP_NEXTHOP_VALID
)
61 && bnc
->nexthop_num
> 0));
64 static int bgp_isvalid_nexthop_for_ebgp(struct bgp_nexthop_cache
*bnc
,
65 struct bgp_path_info
*path
)
67 struct interface
*ifp
= NULL
;
68 struct nexthop
*nexthop
;
69 struct bgp_interface
*iifp
;
72 if (!path
->extra
|| !path
->extra
->peer_orig
)
75 peer
= path
->extra
->peer_orig
;
77 /* only connected ebgp peers are valid */
78 if (peer
->sort
!= BGP_PEER_EBGP
|| peer
->ttl
!= BGP_DEFAULT_TTL
||
79 CHECK_FLAG(peer
->flags
, PEER_FLAG_DISABLE_CONNECTED_CHECK
) ||
80 CHECK_FLAG(peer
->bgp
->flags
, BGP_FLAG_DISABLE_NH_CONNECTED_CHK
))
83 for (nexthop
= bnc
->nexthop
; nexthop
; nexthop
= nexthop
->next
) {
84 if (nexthop
->type
== NEXTHOP_TYPE_IFINDEX
||
85 nexthop
->type
== NEXTHOP_TYPE_IPV4_IFINDEX
||
86 nexthop
->type
== NEXTHOP_TYPE_IPV6_IFINDEX
) {
87 ifp
= if_lookup_by_index(
88 bnc
->ifindex
? bnc
->ifindex
: nexthop
->ifindex
,
94 if (CHECK_FLAG(iifp
->flags
, BGP_INTERFACE_MPLS_BGP_FORWARDING
))
100 static int bgp_isvalid_nexthop_for_mplsovergre(struct bgp_nexthop_cache
*bnc
,
101 struct bgp_path_info
*path
)
103 struct interface
*ifp
= NULL
;
104 struct nexthop
*nexthop
;
106 for (nexthop
= bnc
->nexthop
; nexthop
; nexthop
= nexthop
->next
) {
107 if (nexthop
->type
!= NEXTHOP_TYPE_BLACKHOLE
) {
108 ifp
= if_lookup_by_index(
109 bnc
->ifindex
? bnc
->ifindex
: nexthop
->ifindex
,
111 if (ifp
&& (ifp
->ll_type
== ZEBRA_LLT_IPGRE
||
112 ifp
->ll_type
== ZEBRA_LLT_IP6GRE
))
119 if (CHECK_FLAG(path
->attr
->rmap_change_flags
,
120 BATTR_RMAP_L3VPN_ACCEPT_GRE
))
126 static int bgp_isvalid_nexthop_for_mpls(struct bgp_nexthop_cache
*bnc
,
127 struct bgp_path_info
*path
)
130 * - In the case of MPLS-VPN, the label is learned from LDP or other
131 * protocols, and nexthop tracking is enabled for the label.
132 * The value is recorded as BGP_NEXTHOP_LABELED_VALID.
133 * - In the case of SRv6-VPN, we need to track the reachability to the
134 * SID (in other words, IPv6 address). As in MPLS, we need to record
135 * the value as BGP_NEXTHOP_SID_VALID. However, this function is
136 * currently not implemented, and this function assumes that all
137 * Transit routes for SRv6-VPN are valid.
138 * - Otherwise check for mpls-gre acceptance
140 return (bgp_zebra_num_connects() == 0 ||
141 (bnc
&& (bnc
->nexthop_num
> 0 &&
142 (CHECK_FLAG(bnc
->flags
, BGP_NEXTHOP_LABELED_VALID
) ||
143 bnc
->bgp
->srv6_enabled
||
144 bgp_isvalid_nexthop_for_ebgp(bnc
, path
) ||
145 bgp_isvalid_nexthop_for_mplsovergre(bnc
, path
)))));
148 static void bgp_unlink_nexthop_check(struct bgp_nexthop_cache
*bnc
)
150 if (LIST_EMPTY(&(bnc
->paths
)) && !bnc
->nht_info
) {
151 if (BGP_DEBUG(nht
, NHT
))
152 zlog_debug("%s: freeing bnc %pFX(%d)(%u)(%s)", __func__
,
153 &bnc
->prefix
, bnc
->ifindex
, bnc
->srte_color
,
154 bnc
->bgp
->name_pretty
);
155 /* only unregister if this is the last nh for this prefix*/
156 if (!bnc_existing_for_prefix(bnc
))
157 unregister_zebra_rnh(bnc
);
162 void bgp_unlink_nexthop(struct bgp_path_info
*path
)
164 struct bgp_nexthop_cache
*bnc
= path
->nexthop
;
169 path_nh_map(path
, NULL
, false);
171 bgp_unlink_nexthop_check(bnc
);
174 void bgp_replace_nexthop_by_peer(struct peer
*from
, struct peer
*to
)
178 struct bgp_nexthop_cache
*bncp
, *bnct
;
180 ifindex_t ifindex
= 0;
182 if (!sockunion2hostprefix(&from
->su
, &pp
))
186 * Gather the ifindex for if up/down events to be
187 * tagged into this fun
189 if (from
->conf_if
&& IN6_IS_ADDR_LINKLOCAL(&from
->su
.sin6
.sin6_addr
))
190 ifindex
= from
->su
.sin6
.sin6_scope_id
;
192 afi
= family2afi(pp
.family
);
193 bncp
= bnc_find(&from
->bgp
->nexthop_cache_table
[afi
], &pp
, 0, ifindex
);
195 if (!sockunion2hostprefix(&to
->su
, &pt
))
199 * Gather the ifindex for if up/down events to be
200 * tagged into this fun
203 if (to
->conf_if
&& IN6_IS_ADDR_LINKLOCAL(&to
->su
.sin6
.sin6_addr
))
204 ifindex
= to
->su
.sin6
.sin6_scope_id
;
205 bnct
= bnc_find(&to
->bgp
->nexthop_cache_table
[afi
], &pt
, 0, ifindex
);
214 void bgp_unlink_nexthop_by_peer(struct peer
*peer
)
217 struct bgp_nexthop_cache
*bnc
;
218 afi_t afi
= family2afi(peer
->su
.sa
.sa_family
);
219 ifindex_t ifindex
= 0;
221 if (!sockunion2hostprefix(&peer
->su
, &p
))
224 * Gather the ifindex for if up/down events to be
225 * tagged into this fun
227 if (afi
== AFI_IP6
&& IN6_IS_ADDR_LINKLOCAL(&peer
->su
.sin6
.sin6_addr
))
228 ifindex
= peer
->su
.sin6
.sin6_scope_id
;
229 bnc
= bnc_find(&peer
->bgp
->nexthop_cache_table
[afi
], &p
, 0, ifindex
);
233 /* cleanup the peer reference */
234 bnc
->nht_info
= NULL
;
236 bgp_unlink_nexthop_check(bnc
);
240 * A route and its nexthop might belong to different VRFs. Therefore,
241 * we need both the bgp_route and bgp_nexthop pointers.
243 int bgp_find_or_add_nexthop(struct bgp
*bgp_route
, struct bgp
*bgp_nexthop
,
244 afi_t afi
, safi_t safi
, struct bgp_path_info
*pi
,
245 struct peer
*peer
, int connected
,
246 const struct prefix
*orig_prefix
)
248 struct bgp_nexthop_cache_head
*tree
= NULL
;
249 struct bgp_nexthop_cache
*bnc
;
251 uint32_t srte_color
= 0;
252 int is_bgp_static_route
= 0;
253 ifindex_t ifindex
= 0;
256 is_bgp_static_route
= ((pi
->type
== ZEBRA_ROUTE_BGP
)
257 && (pi
->sub_type
== BGP_ROUTE_STATIC
))
261 /* Since Extended Next-hop Encoding (RFC5549) support, we want
263 address-family from the next-hop. */
264 if (!is_bgp_static_route
)
265 afi
= BGP_ATTR_MP_NEXTHOP_LEN_IP6(pi
->attr
) ? AFI_IP6
268 /* Validation for the ipv4 mapped ipv6 nexthop. */
269 if (IS_MAPPED_IPV6(&pi
->attr
->mp_nexthop_global
)) {
273 /* This will return true if the global IPv6 NH is a link local
275 if (make_prefix(afi
, pi
, &p
) < 0)
278 if (!is_bgp_static_route
&& orig_prefix
279 && prefix_same(&p
, orig_prefix
)) {
280 if (BGP_DEBUG(nht
, NHT
)) {
282 "%s(%pFX): prefix loops through itself",
288 srte_color
= pi
->attr
->srte_color
;
291 * Gather the ifindex for if up/down events to be
292 * tagged into this fun
294 if (afi
== AFI_IP6
&&
295 IN6_IS_ADDR_LINKLOCAL(&peer
->su
.sin6
.sin6_addr
)) {
296 ifindex
= peer
->su
.sin6
.sin6_scope_id
;
298 if (BGP_DEBUG(nht
, NHT
)) {
300 "%s: Unable to locate ifindex, waiting till we have one",
307 if (!sockunion2hostprefix(&peer
->su
, &p
)) {
308 if (BGP_DEBUG(nht
, NHT
)) {
310 "%s: Attempting to register with unknown AFI %d (not %d or %d)",
311 __func__
, afi
, AFI_IP
, AFI_IP6
);
318 if (is_bgp_static_route
)
319 tree
= &bgp_nexthop
->import_check_table
[afi
];
321 tree
= &bgp_nexthop
->nexthop_cache_table
[afi
];
323 bnc
= bnc_find(tree
, &p
, srte_color
, ifindex
);
325 bnc
= bnc_new(tree
, &p
, srte_color
, ifindex
);
326 bnc
->bgp
= bgp_nexthop
;
327 if (BGP_DEBUG(nht
, NHT
))
328 zlog_debug("Allocated bnc %pFX(%d)(%u)(%s) peer %p",
329 &bnc
->prefix
, bnc
->ifindex
, bnc
->srte_color
,
330 bnc
->bgp
->name_pretty
, peer
);
332 if (BGP_DEBUG(nht
, NHT
))
334 "Found existing bnc %pFX(%d)(%s) flags 0x%x ifindex %d #paths %d peer %p",
335 &bnc
->prefix
, bnc
->ifindex
,
336 bnc
->bgp
->name_pretty
, bnc
->flags
, bnc
->ifindex
,
337 bnc
->path_count
, bnc
->nht_info
);
340 if (pi
&& is_route_parent_evpn(pi
))
341 bnc
->is_evpn_gwip_nexthop
= true;
343 if (is_bgp_static_route
) {
344 SET_FLAG(bnc
->flags
, BGP_STATIC_ROUTE
);
346 /* If we're toggling the type, re-register */
347 if ((CHECK_FLAG(bgp_route
->flags
, BGP_FLAG_IMPORT_CHECK
))
348 && !CHECK_FLAG(bnc
->flags
, BGP_STATIC_ROUTE_EXACT_MATCH
)) {
349 SET_FLAG(bnc
->flags
, BGP_STATIC_ROUTE_EXACT_MATCH
);
350 UNSET_FLAG(bnc
->flags
, BGP_NEXTHOP_REGISTERED
);
351 UNSET_FLAG(bnc
->flags
, BGP_NEXTHOP_VALID
);
352 } else if ((!CHECK_FLAG(bgp_route
->flags
,
353 BGP_FLAG_IMPORT_CHECK
))
354 && CHECK_FLAG(bnc
->flags
,
355 BGP_STATIC_ROUTE_EXACT_MATCH
)) {
356 UNSET_FLAG(bnc
->flags
, BGP_STATIC_ROUTE_EXACT_MATCH
);
357 UNSET_FLAG(bnc
->flags
, BGP_NEXTHOP_REGISTERED
);
358 UNSET_FLAG(bnc
->flags
, BGP_NEXTHOP_VALID
);
361 /* When nexthop is already known, but now requires 'connected'
363 * re-register it. The reverse scenario where the nexthop currently
365 * 'connected' resolution does not need a re-register (i.e., we treat
366 * 'connected-required' as an override) except in the scenario where
368 * is actually a case of tracking a peer for connectivity (e.g., after
369 * disable connected-check).
370 * NOTE: We don't track the number of paths separately for 'connected-
371 * required' vs 'connected-not-required' as this change is not a common
374 else if (connected
&& !CHECK_FLAG(bnc
->flags
, BGP_NEXTHOP_CONNECTED
)) {
375 SET_FLAG(bnc
->flags
, BGP_NEXTHOP_CONNECTED
);
376 UNSET_FLAG(bnc
->flags
, BGP_NEXTHOP_REGISTERED
);
377 UNSET_FLAG(bnc
->flags
, BGP_NEXTHOP_VALID
);
378 } else if (peer
&& !connected
379 && CHECK_FLAG(bnc
->flags
, BGP_NEXTHOP_CONNECTED
)) {
380 UNSET_FLAG(bnc
->flags
, BGP_NEXTHOP_CONNECTED
);
381 UNSET_FLAG(bnc
->flags
, BGP_NEXTHOP_REGISTERED
);
382 UNSET_FLAG(bnc
->flags
, BGP_NEXTHOP_VALID
);
384 if (peer
&& (bnc
->ifindex
!= ifindex
)) {
385 UNSET_FLAG(bnc
->flags
, BGP_NEXTHOP_REGISTERED
);
386 UNSET_FLAG(bnc
->flags
, BGP_NEXTHOP_VALID
);
387 bnc
->ifindex
= ifindex
;
389 if (bgp_route
->inst_type
== BGP_INSTANCE_TYPE_VIEW
) {
390 SET_FLAG(bnc
->flags
, BGP_NEXTHOP_REGISTERED
);
391 SET_FLAG(bnc
->flags
, BGP_NEXTHOP_VALID
);
392 } else if (!CHECK_FLAG(bnc
->flags
, BGP_NEXTHOP_REGISTERED
)
393 && !is_default_host_route(&bnc
->prefix
))
394 register_zebra_rnh(bnc
);
396 if (pi
&& pi
->nexthop
!= bnc
) {
397 /* Unlink from existing nexthop cache, if any. This will also
399 * the nexthop cache entry, if appropriate.
401 bgp_unlink_nexthop(pi
);
403 /* updates NHT pi list reference */
404 path_nh_map(pi
, bnc
, true);
406 if (CHECK_FLAG(bnc
->flags
, BGP_NEXTHOP_VALID
) && bnc
->metric
)
407 (bgp_path_info_extra_get(pi
))->igpmetric
= bnc
->metric
;
409 pi
->extra
->igpmetric
= 0;
412 * Let's not accidentally save the peer data for a peer
413 * we are going to throw away in a second or so.
414 * When we come back around we'll fix up this
415 * data properly in replace_nexthop_by_peer
417 if (CHECK_FLAG(peer
->flags
, PEER_FLAG_CONFIG_NODE
))
418 bnc
->nht_info
= (void *)peer
; /* NHT peer reference */
422 * We are cheating here. Views have no associated underlying
423 * ability to detect nexthops. So when we have a view
424 * just tell everyone the nexthop is valid
426 if (bgp_route
->inst_type
== BGP_INSTANCE_TYPE_VIEW
)
428 else if (safi
== SAFI_UNICAST
&& pi
&&
429 pi
->sub_type
== BGP_ROUTE_IMPORTED
&& pi
->extra
&&
430 pi
->extra
->num_labels
&& !bnc
->is_evpn_gwip_nexthop
)
431 return bgp_isvalid_nexthop_for_mpls(bnc
, pi
);
433 return (bgp_isvalid_nexthop(bnc
));
436 void bgp_delete_connected_nexthop(afi_t afi
, struct peer
*peer
)
438 struct bgp_nexthop_cache
*bnc
;
440 ifindex_t ifindex
= 0;
445 if (!sockunion2hostprefix(&peer
->su
, &p
))
448 * Gather the ifindex for if up/down events to be
449 * tagged into this fun
451 if (afi
== AFI_IP6
&& IN6_IS_ADDR_LINKLOCAL(&peer
->su
.sin6
.sin6_addr
))
452 ifindex
= peer
->su
.sin6
.sin6_scope_id
;
453 bnc
= bnc_find(&peer
->bgp
->nexthop_cache_table
[family2afi(p
.family
)],
456 if (BGP_DEBUG(nht
, NHT
))
458 "Cannot find connected NHT node for peer %s(%s)",
459 peer
->host
, peer
->bgp
->name_pretty
);
463 if (bnc
->nht_info
!= peer
) {
464 if (BGP_DEBUG(nht
, NHT
))
466 "Connected NHT %p node for peer %s(%s) points to %p",
467 bnc
, peer
->host
, bnc
->bgp
->name_pretty
,
472 bnc
->nht_info
= NULL
;
474 if (LIST_EMPTY(&(bnc
->paths
))) {
475 if (BGP_DEBUG(nht
, NHT
))
477 "Freeing connected NHT node %p for peer %s(%s)",
478 bnc
, peer
->host
, bnc
->bgp
->name_pretty
);
479 unregister_zebra_rnh(bnc
);
484 static void bgp_process_nexthop_update(struct bgp_nexthop_cache
*bnc
,
485 struct zapi_route
*nhr
,
488 struct nexthop
*nexthop
;
489 struct nexthop
*oldnh
;
490 struct nexthop
*nhlist_head
= NULL
;
491 struct nexthop
*nhlist_tail
= NULL
;
493 bool evpn_resolved
= false;
495 bnc
->last_update
= monotime(NULL
);
496 bnc
->change_flags
= 0;
498 /* debug print the input */
499 if (BGP_DEBUG(nht
, NHT
)) {
500 char bnc_buf
[BNC_FLAG_DUMP_SIZE
];
503 "%s(%u): Rcvd NH update %pFX(%u)%u) - metric %d/%d #nhops %d/%d flags %s",
504 bnc
->bgp
->name_pretty
, bnc
->bgp
->vrf_id
, &nhr
->prefix
,
505 bnc
->ifindex
, bnc
->srte_color
, nhr
->metric
, bnc
->metric
,
506 nhr
->nexthop_num
, bnc
->nexthop_num
,
507 bgp_nexthop_dump_bnc_flags(bnc
, bnc_buf
,
511 if (nhr
->metric
!= bnc
->metric
)
512 bnc
->change_flags
|= BGP_NEXTHOP_METRIC_CHANGED
;
514 if (nhr
->nexthop_num
!= bnc
->nexthop_num
)
515 bnc
->change_flags
|= BGP_NEXTHOP_CHANGED
;
517 if (import_check
&& (nhr
->type
== ZEBRA_ROUTE_BGP
||
518 !prefix_same(&bnc
->prefix
, &nhr
->prefix
))) {
519 SET_FLAG(bnc
->change_flags
, BGP_NEXTHOP_CHANGED
);
520 UNSET_FLAG(bnc
->flags
, BGP_NEXTHOP_VALID
);
521 UNSET_FLAG(bnc
->flags
, BGP_NEXTHOP_LABELED_VALID
);
522 UNSET_FLAG(bnc
->flags
, BGP_NEXTHOP_EVPN_INCOMPLETE
);
524 bnc_nexthop_free(bnc
);
527 if (BGP_DEBUG(nht
, NHT
))
529 "%s: Import Check does not resolve to the same prefix for %pFX received %pFX or matching route is BGP",
530 __func__
, &bnc
->prefix
, &nhr
->prefix
);
531 } else if (nhr
->nexthop_num
) {
532 struct peer
*peer
= bnc
->nht_info
;
534 /* notify bgp fsm if nbr ip goes from invalid->valid */
535 if (!bnc
->nexthop_num
)
536 UNSET_FLAG(bnc
->flags
, BGP_NEXTHOP_PEER_NOTIFIED
);
538 if (!bnc
->is_evpn_gwip_nexthop
)
539 bnc
->flags
|= BGP_NEXTHOP_VALID
;
540 bnc
->metric
= nhr
->metric
;
541 bnc
->nexthop_num
= nhr
->nexthop_num
;
543 bnc
->flags
&= ~BGP_NEXTHOP_LABELED_VALID
; /* check below */
545 for (i
= 0; i
< nhr
->nexthop_num
; i
++) {
548 nexthop
= nexthop_from_zapi_nexthop(&nhr
->nexthops
[i
]);
551 * Turn on RA for the v6 nexthops
552 * we receive from bgp. This is to allow us
553 * to work with v4 routing over v6 nexthops
555 if (peer
&& !peer
->ifp
556 && CHECK_FLAG(peer
->flags
,
557 PEER_FLAG_CAPABILITY_ENHE
)
558 && nhr
->prefix
.family
== AF_INET6
559 && nexthop
->type
!= NEXTHOP_TYPE_BLACKHOLE
) {
560 struct interface
*ifp
;
562 ifp
= if_lookup_by_index(nexthop
->ifindex
,
565 zclient_send_interface_radv_req(
566 zclient
, nexthop
->vrf_id
, ifp
,
568 BGP_UNNUM_DEFAULT_RA_INTERVAL
);
570 /* There is at least one label-switched path */
571 if (nexthop
->nh_label
&&
572 nexthop
->nh_label
->num_labels
) {
574 bnc
->flags
|= BGP_NEXTHOP_LABELED_VALID
;
575 num_labels
= nexthop
->nh_label
->num_labels
;
578 if (BGP_DEBUG(nht
, NHT
)) {
579 char buf
[NEXTHOP_STRLEN
];
581 " nhop via %s (%d labels)",
582 nexthop2str(nexthop
, buf
, sizeof(buf
)),
587 nhlist_tail
->next
= nexthop
;
588 nhlist_tail
= nexthop
;
590 nhlist_tail
= nexthop
;
591 nhlist_head
= nexthop
;
594 /* No need to evaluate the nexthop if we have already
596 * that there has been a change.
598 if (bnc
->change_flags
& BGP_NEXTHOP_CHANGED
)
601 for (oldnh
= bnc
->nexthop
; oldnh
; oldnh
= oldnh
->next
)
602 if (nexthop_same(oldnh
, nexthop
))
606 bnc
->change_flags
|= BGP_NEXTHOP_CHANGED
;
608 bnc_nexthop_free(bnc
);
609 bnc
->nexthop
= nhlist_head
;
612 * Gateway IP nexthop is L3 reachable. Mark it as
613 * BGP_NEXTHOP_VALID only if it is recursively resolved with a
615 * Else, mark it as BGP_NEXTHOP_EVPN_INCOMPLETE.
616 * When its mapping with EVPN RT-2 is established, unset
617 * BGP_NEXTHOP_EVPN_INCOMPLETE and set BGP_NEXTHOP_VALID.
619 if (bnc
->is_evpn_gwip_nexthop
) {
620 evpn_resolved
= bgp_evpn_is_gateway_ip_resolved(bnc
);
622 if (BGP_DEBUG(nht
, NHT
))
624 "EVPN gateway IP %pFX recursive MAC/IP lookup %s",
626 (evpn_resolved
? "successful"
630 bnc
->flags
|= BGP_NEXTHOP_VALID
;
631 bnc
->flags
&= ~BGP_NEXTHOP_EVPN_INCOMPLETE
;
632 bnc
->change_flags
|= BGP_NEXTHOP_MACIP_CHANGED
;
634 bnc
->flags
|= BGP_NEXTHOP_EVPN_INCOMPLETE
;
635 bnc
->flags
&= ~BGP_NEXTHOP_VALID
;
639 bnc
->flags
&= ~BGP_NEXTHOP_EVPN_INCOMPLETE
;
640 bnc
->flags
&= ~BGP_NEXTHOP_VALID
;
641 bnc
->flags
&= ~BGP_NEXTHOP_LABELED_VALID
;
642 bnc
->nexthop_num
= nhr
->nexthop_num
;
644 /* notify bgp fsm if nbr ip goes from valid->invalid */
645 UNSET_FLAG(bnc
->flags
, BGP_NEXTHOP_PEER_NOTIFIED
);
647 bnc_nexthop_free(bnc
);
654 static void bgp_nht_ifp_table_handle(struct bgp
*bgp
,
655 struct bgp_nexthop_cache_head
*table
,
656 struct interface
*ifp
, bool up
)
658 struct bgp_nexthop_cache
*bnc
;
660 frr_each (bgp_nexthop_cache
, table
, bnc
) {
661 if (bnc
->ifindex
!= ifp
->ifindex
)
664 bnc
->last_update
= monotime(NULL
);
665 bnc
->change_flags
= 0;
668 * For interface based routes ( ala the v6 LL routes
669 * that this was written for ) the metric received
670 * for the connected route is 0 not 1.
674 SET_FLAG(bnc
->flags
, BGP_NEXTHOP_VALID
);
675 SET_FLAG(bnc
->change_flags
, BGP_NEXTHOP_CHANGED
);
676 bnc
->nexthop_num
= 1;
678 UNSET_FLAG(bnc
->flags
, BGP_NEXTHOP_PEER_NOTIFIED
);
679 UNSET_FLAG(bnc
->flags
, BGP_NEXTHOP_VALID
);
680 SET_FLAG(bnc
->change_flags
, BGP_NEXTHOP_CHANGED
);
681 bnc
->nexthop_num
= 0;
687 static void bgp_nht_ifp_handle(struct interface
*ifp
, bool up
)
691 bgp
= ifp
->vrf
->info
;
695 bgp_nht_ifp_table_handle(bgp
, &bgp
->nexthop_cache_table
[AFI_IP
], ifp
,
697 bgp_nht_ifp_table_handle(bgp
, &bgp
->import_check_table
[AFI_IP
], ifp
,
699 bgp_nht_ifp_table_handle(bgp
, &bgp
->nexthop_cache_table
[AFI_IP6
], ifp
,
701 bgp_nht_ifp_table_handle(bgp
, &bgp
->import_check_table
[AFI_IP6
], ifp
,
705 void bgp_nht_ifp_up(struct interface
*ifp
)
707 bgp_nht_ifp_handle(ifp
, true);
710 void bgp_nht_ifp_down(struct interface
*ifp
)
712 bgp_nht_ifp_handle(ifp
, false);
715 static void bgp_nht_ifp_initial(struct thread
*thread
)
717 ifindex_t ifindex
= THREAD_VAL(thread
);
718 struct bgp
*bgp
= THREAD_ARG(thread
);
719 struct interface
*ifp
= if_lookup_by_index(ifindex
, bgp
->vrf_id
);
724 if (BGP_DEBUG(nht
, NHT
))
726 "Handle NHT initial update for Intf %s(%d) status %s",
727 ifp
->name
, ifp
->ifindex
, if_is_up(ifp
) ? "up" : "down");
732 bgp_nht_ifp_down(ifp
);
736 * So the bnc code has the ability to handle interface up/down
737 * events to properly handle v6 LL peering.
738 * What is happening here:
739 * The event system for peering expects the nht code to
740 * report on the tracking events after we move to active
741 * So let's give the system a chance to report on that event
742 * in a manner that is expected.
744 void bgp_nht_interface_events(struct peer
*peer
)
746 struct bgp
*bgp
= peer
->bgp
;
747 struct bgp_nexthop_cache_head
*table
;
748 struct bgp_nexthop_cache
*bnc
;
750 ifindex_t ifindex
= 0;
752 if (!IN6_IS_ADDR_LINKLOCAL(&peer
->su
.sin6
.sin6_addr
))
755 if (!sockunion2hostprefix(&peer
->su
, &p
))
758 * Gather the ifindex for if up/down events to be
759 * tagged into this fun
761 if (peer
->conf_if
&& IN6_IS_ADDR_LINKLOCAL(&peer
->su
.sin6
.sin6_addr
))
762 ifindex
= peer
->su
.sin6
.sin6_scope_id
;
764 table
= &bgp
->nexthop_cache_table
[AFI_IP6
];
765 bnc
= bnc_find(table
, &p
, 0, ifindex
);
770 thread_add_event(bm
->master
, bgp_nht_ifp_initial
, bnc
->bgp
,
774 void bgp_parse_nexthop_update(int command
, vrf_id_t vrf_id
)
776 struct bgp_nexthop_cache_head
*tree
= NULL
;
777 struct bgp_nexthop_cache
*bnc_nhc
, *bnc_import
;
780 struct zapi_route nhr
;
783 bgp
= bgp_lookup_by_vrf_id(vrf_id
);
787 "parse nexthop update: instance not found for vrf_id %u",
792 if (!zapi_nexthop_update_decode(zclient
->ibuf
, &match
, &nhr
)) {
793 zlog_err("%s[%s]: Failure to decode nexthop update", __func__
,
798 afi
= family2afi(match
.family
);
799 tree
= &bgp
->nexthop_cache_table
[afi
];
801 bnc_nhc
= bnc_find(tree
, &match
, nhr
.srte_color
, 0);
803 if (BGP_DEBUG(nht
, NHT
))
805 "parse nexthop update(%pFX(%u)(%s)): bnc info not found for nexthop cache",
806 &nhr
.prefix
, nhr
.srte_color
, bgp
->name_pretty
);
808 bgp_process_nexthop_update(bnc_nhc
, &nhr
, false);
810 tree
= &bgp
->import_check_table
[afi
];
812 bnc_import
= bnc_find(tree
, &match
, nhr
.srte_color
, 0);
814 if (BGP_DEBUG(nht
, NHT
))
816 "parse nexthop update(%pFX(%u)(%s)): bnc info not found for import check",
817 &nhr
.prefix
, nhr
.srte_color
, bgp
->name_pretty
);
819 bgp_process_nexthop_update(bnc_import
, &nhr
, true);
822 * HACK: if any BGP route is dependant on an SR-policy that doesn't
823 * exist, zebra will never send NH updates relative to that policy. In
824 * that case, whenever we receive an update about a colorless NH, update
825 * the corresponding colorful NHs that share the same endpoint but that
826 * are inactive. This ugly hack should work around the problem at the
827 * cost of a performance pernalty. Long term, what should be done is to
828 * make zebra's RNH subsystem aware of SR-TE colors (like bgpd is),
829 * which should provide a better infrastructure to solve this issue in
830 * a more efficient and elegant way.
832 if (nhr
.srte_color
== 0 && bnc_nhc
) {
833 struct bgp_nexthop_cache
*bnc_iter
;
835 frr_each (bgp_nexthop_cache
, &bgp
->nexthop_cache_table
[afi
],
837 if (!prefix_same(&bnc_nhc
->prefix
, &bnc_iter
->prefix
) ||
838 bnc_iter
->srte_color
== 0 ||
839 CHECK_FLAG(bnc_iter
->flags
, BGP_NEXTHOP_VALID
))
842 bgp_process_nexthop_update(bnc_iter
, &nhr
, false);
848 * Cleanup nexthop registration and status information for BGP nexthops
849 * pertaining to this VRF. This is invoked upon VRF deletion.
851 void bgp_cleanup_nexthops(struct bgp
*bgp
)
853 for (afi_t afi
= AFI_IP
; afi
< AFI_MAX
; afi
++) {
854 struct bgp_nexthop_cache
*bnc
;
856 frr_each (bgp_nexthop_cache
, &bgp
->nexthop_cache_table
[afi
],
858 /* Clear relevant flags. */
859 UNSET_FLAG(bnc
->flags
, BGP_NEXTHOP_VALID
);
860 UNSET_FLAG(bnc
->flags
, BGP_NEXTHOP_REGISTERED
);
861 UNSET_FLAG(bnc
->flags
, BGP_NEXTHOP_PEER_NOTIFIED
);
862 UNSET_FLAG(bnc
->flags
, BGP_NEXTHOP_EVPN_INCOMPLETE
);
868 * make_prefix - make a prefix structure from the path (essentially
871 static int make_prefix(int afi
, struct bgp_path_info
*pi
, struct prefix
*p
)
874 int is_bgp_static
= ((pi
->type
== ZEBRA_ROUTE_BGP
)
875 && (pi
->sub_type
== BGP_ROUTE_STATIC
))
878 struct bgp_dest
*net
= pi
->net
;
879 const struct prefix
*p_orig
= bgp_dest_get_prefix(net
);
882 if (p_orig
->family
== AF_FLOWSPEC
) {
885 return bgp_flowspec_get_first_nh(pi
->peer
->bgp
,
888 memset(p
, 0, sizeof(struct prefix
));
893 p
->u
.prefix4
= p_orig
->u
.prefix4
;
894 p
->prefixlen
= p_orig
->prefixlen
;
896 if (IS_MAPPED_IPV6(&pi
->attr
->mp_nexthop_global
)) {
897 ipv4_mapped_ipv6_to_ipv4(
898 &pi
->attr
->mp_nexthop_global
, &ipv4
);
900 p
->prefixlen
= IPV4_MAX_BITLEN
;
902 if (p_orig
->family
== AF_EVPN
)
904 pi
->attr
->mp_nexthop_global_in
;
906 p
->u
.prefix4
= pi
->attr
->nexthop
;
907 p
->prefixlen
= IPV4_MAX_BITLEN
;
912 p
->family
= AF_INET6
;
915 p
->u
.prefix6
= p_orig
->u
.prefix6
;
916 p
->prefixlen
= p_orig
->prefixlen
;
918 /* If we receive MP_REACH nexthop with ::(LL)
919 * or LL(LL), use LL address as nexthop cache.
921 if (pi
->attr
->mp_nexthop_len
922 == BGP_ATTR_NHLEN_IPV6_GLOBAL_AND_LL
923 && (IN6_IS_ADDR_UNSPECIFIED(
924 &pi
->attr
->mp_nexthop_global
)
925 || IN6_IS_ADDR_LINKLOCAL(
926 &pi
->attr
->mp_nexthop_global
)))
927 p
->u
.prefix6
= pi
->attr
->mp_nexthop_local
;
928 /* If we receive MR_REACH with (GA)::(LL)
929 * then check for route-map to choose GA or LL
931 else if (pi
->attr
->mp_nexthop_len
932 == BGP_ATTR_NHLEN_IPV6_GLOBAL_AND_LL
) {
933 if (pi
->attr
->mp_nexthop_prefer_global
)
935 pi
->attr
->mp_nexthop_global
;
938 pi
->attr
->mp_nexthop_local
;
940 p
->u
.prefix6
= pi
->attr
->mp_nexthop_global
;
941 p
->prefixlen
= IPV6_MAX_BITLEN
;
945 if (BGP_DEBUG(nht
, NHT
)) {
947 "%s: Attempting to make prefix with unknown AFI %d (not %d or %d)",
948 __func__
, afi
, AFI_IP
, AFI_IP6
);
956 * sendmsg_zebra_rnh -- Format and send a nexthop register/Unregister
959 * struct bgp_nexthop_cache *bnc -- the nexthop structure.
960 * int command -- command to send to zebra
964 static void sendmsg_zebra_rnh(struct bgp_nexthop_cache
*bnc
, int command
)
966 bool exact_match
= false;
967 bool resolve_via_default
= false;
973 /* Don't try to register if Zebra doesn't know of this instance. */
974 if (!IS_BGP_INST_KNOWN_TO_ZEBRA(bnc
->bgp
)) {
975 if (BGP_DEBUG(zebra
, ZEBRA
))
977 "%s: No zebra instance to talk to, not installing NHT entry",
982 if (!bgp_zebra_num_connects()) {
983 if (BGP_DEBUG(zebra
, ZEBRA
))
985 "%s: We have not connected yet, cannot send nexthops",
988 if (command
== ZEBRA_NEXTHOP_REGISTER
) {
989 if (CHECK_FLAG(bnc
->flags
, BGP_NEXTHOP_CONNECTED
))
991 if (CHECK_FLAG(bnc
->flags
, BGP_STATIC_ROUTE_EXACT_MATCH
))
992 resolve_via_default
= true;
995 if (BGP_DEBUG(zebra
, ZEBRA
))
996 zlog_debug("%s: sending cmd %s for %pFX (vrf %s)", __func__
,
997 zserv_command_string(command
), &bnc
->prefix
,
998 bnc
->bgp
->name_pretty
);
1000 ret
= zclient_send_rnh(zclient
, command
, &bnc
->prefix
, SAFI_UNICAST
,
1001 exact_match
, resolve_via_default
,
1003 if (ret
== ZCLIENT_SEND_FAILURE
) {
1004 flog_warn(EC_BGP_ZEBRA_SEND
,
1005 "sendmsg_nexthop: zclient_send_message() failed");
1009 if (command
== ZEBRA_NEXTHOP_REGISTER
)
1010 SET_FLAG(bnc
->flags
, BGP_NEXTHOP_REGISTERED
);
1011 else if (command
== ZEBRA_NEXTHOP_UNREGISTER
)
1012 UNSET_FLAG(bnc
->flags
, BGP_NEXTHOP_REGISTERED
);
1017 * register_zebra_rnh - register a NH/route with Zebra for notification
1018 * when the route or the route to the nexthop changes.
1020 * struct bgp_nexthop_cache *bnc
1024 static void register_zebra_rnh(struct bgp_nexthop_cache
*bnc
)
1026 /* Check if we have already registered */
1027 if (bnc
->flags
& BGP_NEXTHOP_REGISTERED
)
1031 SET_FLAG(bnc
->flags
, BGP_NEXTHOP_REGISTERED
);
1035 sendmsg_zebra_rnh(bnc
, ZEBRA_NEXTHOP_REGISTER
);
1039 * unregister_zebra_rnh -- Unregister the route/nexthop from Zebra.
1041 * struct bgp_nexthop_cache *bnc
1045 static void unregister_zebra_rnh(struct bgp_nexthop_cache
*bnc
)
1047 /* Check if we have already registered */
1048 if (!CHECK_FLAG(bnc
->flags
, BGP_NEXTHOP_REGISTERED
))
1052 UNSET_FLAG(bnc
->flags
, BGP_NEXTHOP_REGISTERED
);
1056 sendmsg_zebra_rnh(bnc
, ZEBRA_NEXTHOP_UNREGISTER
);
1060 * evaluate_paths - Evaluate the paths/nets associated with a nexthop.
1062 * struct bgp_nexthop_cache *bnc -- the nexthop structure.
1066 void evaluate_paths(struct bgp_nexthop_cache
*bnc
)
1068 struct bgp_dest
*dest
;
1069 struct bgp_path_info
*path
;
1071 struct peer
*peer
= (struct peer
*)bnc
->nht_info
;
1072 struct bgp_table
*table
;
1074 struct bgp
*bgp_path
;
1075 const struct prefix
*p
;
1077 if (BGP_DEBUG(nht
, NHT
)) {
1078 char bnc_buf
[BNC_FLAG_DUMP_SIZE
];
1079 char chg_buf
[BNC_FLAG_DUMP_SIZE
];
1082 "NH update for %pFX(%d)(%u)(%s) - flags %s chgflags %s- evaluate paths",
1083 &bnc
->prefix
, bnc
->ifindex
, bnc
->srte_color
,
1084 bnc
->bgp
->name_pretty
,
1085 bgp_nexthop_dump_bnc_flags(bnc
, bnc_buf
,
1087 bgp_nexthop_dump_bnc_change_flags(bnc
, chg_buf
,
1091 LIST_FOREACH (path
, &(bnc
->paths
), nh_thread
) {
1092 if (!(path
->type
== ZEBRA_ROUTE_BGP
1093 && ((path
->sub_type
== BGP_ROUTE_NORMAL
)
1094 || (path
->sub_type
== BGP_ROUTE_STATIC
)
1095 || (path
->sub_type
== BGP_ROUTE_IMPORTED
))))
1099 assert(dest
&& bgp_dest_table(dest
));
1100 p
= bgp_dest_get_prefix(dest
);
1101 afi
= family2afi(p
->family
);
1102 table
= bgp_dest_table(dest
);
1106 * handle routes from other VRFs (they can have a
1107 * nexthop in THIS VRF). bgp_path is the bgp instance
1108 * that owns the route referencing this nexthop.
1110 bgp_path
= table
->bgp
;
1113 * Path becomes valid/invalid depending on whether the nexthop
1114 * reachable/unreachable.
1116 * In case of unicast routes that were imported from vpn
1117 * and that have labels, they are valid only if there are
1118 * nexthops with labels
1120 * If the nexthop is EVPN gateway-IP,
1121 * do not check for a valid label.
1124 bool bnc_is_valid_nexthop
= false;
1125 bool path_valid
= false;
1127 if (safi
== SAFI_UNICAST
&& path
->sub_type
== BGP_ROUTE_IMPORTED
1128 && path
->extra
&& path
->extra
->num_labels
1129 && (path
->attr
->evpn_overlay
.type
1130 != OVERLAY_INDEX_GATEWAY_IP
)) {
1131 bnc_is_valid_nexthop
=
1132 bgp_isvalid_nexthop_for_mpls(bnc
, path
) ? true
1135 if (bgp_update_martian_nexthop(
1136 bnc
->bgp
, afi
, safi
, path
->type
,
1137 path
->sub_type
, path
->attr
, dest
)) {
1138 if (BGP_DEBUG(nht
, NHT
))
1140 "%s: prefix %pBD (vrf %s), ignoring path due to martian or self-next-hop",
1141 __func__
, dest
, bgp_path
->name
);
1143 bnc_is_valid_nexthop
=
1144 bgp_isvalid_nexthop(bnc
) ? true : false;
1147 if (BGP_DEBUG(nht
, NHT
)) {
1150 "... eval path %d/%d %pBD RD %pRD %s flags 0x%x",
1152 (struct prefix_rd
*)bgp_dest_get_prefix(
1154 bgp_path
->name_pretty
, path
->flags
);
1157 "... eval path %d/%d %pBD %s flags 0x%x",
1158 afi
, safi
, dest
, bgp_path
->name_pretty
,
1162 /* Skip paths marked for removal or as history. */
1163 if (CHECK_FLAG(path
->flags
, BGP_PATH_REMOVED
)
1164 || CHECK_FLAG(path
->flags
, BGP_PATH_HISTORY
))
1167 /* Copy the metric to the path. Will be used for bestpath
1169 if (bgp_isvalid_nexthop(bnc
) && bnc
->metric
)
1170 (bgp_path_info_extra_get(path
))->igpmetric
=
1172 else if (path
->extra
)
1173 path
->extra
->igpmetric
= 0;
1175 if (CHECK_FLAG(bnc
->change_flags
, BGP_NEXTHOP_METRIC_CHANGED
)
1176 || CHECK_FLAG(bnc
->change_flags
, BGP_NEXTHOP_CHANGED
)
1177 || path
->attr
->srte_color
!= 0)
1178 SET_FLAG(path
->flags
, BGP_PATH_IGP_CHANGED
);
1180 path_valid
= CHECK_FLAG(path
->flags
, BGP_PATH_VALID
);
1181 if (path_valid
!= bnc_is_valid_nexthop
) {
1183 /* No longer valid, clear flag; also for EVPN
1184 * routes, unimport from VRFs if needed.
1186 bgp_aggregate_decrement(bgp_path
, p
, path
, afi
,
1188 bgp_path_info_unset_flag(dest
, path
,
1190 if (safi
== SAFI_EVPN
&&
1191 bgp_evpn_is_prefix_nht_supported(bgp_dest_get_prefix(dest
)))
1192 bgp_evpn_unimport_route(bgp_path
,
1193 afi
, safi
, bgp_dest_get_prefix(dest
), path
);
1195 /* Path becomes valid, set flag; also for EVPN
1196 * routes, import from VRFs if needed.
1198 bgp_path_info_set_flag(dest
, path
,
1200 bgp_aggregate_increment(bgp_path
, p
, path
, afi
,
1202 if (safi
== SAFI_EVPN
&&
1203 bgp_evpn_is_prefix_nht_supported(bgp_dest_get_prefix(dest
)))
1204 bgp_evpn_import_route(bgp_path
,
1205 afi
, safi
, bgp_dest_get_prefix(dest
), path
);
1209 bgp_process(bgp_path
, dest
, afi
, safi
);
1213 int valid_nexthops
= bgp_isvalid_nexthop(bnc
);
1215 if (valid_nexthops
) {
1217 * Peering cannot occur across a blackhole nexthop
1219 if (bnc
->nexthop_num
== 1 && bnc
->nexthop
1220 && bnc
->nexthop
->type
== NEXTHOP_TYPE_BLACKHOLE
) {
1221 peer
->last_reset
= PEER_DOWN_WAITING_NHT
;
1224 peer
->last_reset
= PEER_DOWN_WAITING_OPEN
;
1226 peer
->last_reset
= PEER_DOWN_WAITING_NHT
;
1228 if (!CHECK_FLAG(bnc
->flags
, BGP_NEXTHOP_PEER_NOTIFIED
)) {
1229 if (BGP_DEBUG(nht
, NHT
))
1231 "%s: Updating peer (%s(%s)) status with NHT nexthops %d",
1232 __func__
, peer
->host
,
1233 peer
->bgp
->name_pretty
,
1235 bgp_fsm_nht_update(peer
, !!valid_nexthops
);
1236 SET_FLAG(bnc
->flags
, BGP_NEXTHOP_PEER_NOTIFIED
);
1240 RESET_FLAG(bnc
->change_flags
);
1244 * path_nh_map - make or break path-to-nexthop association.
1246 * path - pointer to the path structure
1247 * bnc - pointer to the nexthop structure
1248 * make - if set, make the association. if unset, just break the existing
1251 void path_nh_map(struct bgp_path_info
*path
, struct bgp_nexthop_cache
*bnc
,
1254 if (path
->nexthop
) {
1255 LIST_REMOVE(path
, nh_thread
);
1256 path
->nexthop
->path_count
--;
1257 path
->nexthop
= NULL
;
1260 LIST_INSERT_HEAD(&(bnc
->paths
), path
, nh_thread
);
1261 path
->nexthop
= bnc
;
1262 path
->nexthop
->path_count
++;
1267 * This function is called to register nexthops to zebra
1268 * as that we may have tried to install the nexthops
1269 * before we actually have a zebra connection
1271 void bgp_nht_register_nexthops(struct bgp
*bgp
)
1273 for (afi_t afi
= AFI_IP
; afi
< AFI_MAX
; afi
++) {
1274 struct bgp_nexthop_cache
*bnc
;
1276 frr_each (bgp_nexthop_cache
, &bgp
->nexthop_cache_table
[afi
],
1278 register_zebra_rnh(bnc
);
1283 void bgp_nht_reg_enhe_cap_intfs(struct peer
*peer
)
1286 struct bgp_nexthop_cache
*bnc
;
1287 struct nexthop
*nhop
;
1288 struct interface
*ifp
;
1290 ifindex_t ifindex
= 0;
1296 if (!sockunion2hostprefix(&peer
->su
, &p
)) {
1297 zlog_warn("%s: Unable to convert sockunion to prefix for %s",
1298 __func__
, peer
->host
);
1302 if (p
.family
!= AF_INET6
)
1305 * Gather the ifindex for if up/down events to be
1306 * tagged into this fun
1308 if (peer
->conf_if
&& IN6_IS_ADDR_LINKLOCAL(&peer
->su
.sin6
.sin6_addr
))
1309 ifindex
= peer
->su
.sin6
.sin6_scope_id
;
1311 bnc
= bnc_find(&bgp
->nexthop_cache_table
[AFI_IP6
], &p
, 0, ifindex
);
1315 if (peer
!= bnc
->nht_info
)
1318 for (nhop
= bnc
->nexthop
; nhop
; nhop
= nhop
->next
) {
1319 ifp
= if_lookup_by_index(nhop
->ifindex
, nhop
->vrf_id
);
1324 zclient_send_interface_radv_req(zclient
,
1327 BGP_UNNUM_DEFAULT_RA_INTERVAL
);
1331 void bgp_nht_dereg_enhe_cap_intfs(struct peer
*peer
)
1334 struct bgp_nexthop_cache
*bnc
;
1335 struct nexthop
*nhop
;
1336 struct interface
*ifp
;
1338 ifindex_t ifindex
= 0;
1345 if (!sockunion2hostprefix(&peer
->su
, &p
)) {
1346 zlog_warn("%s: Unable to convert sockunion to prefix for %s",
1347 __func__
, peer
->host
);
1351 if (p
.family
!= AF_INET6
)
1354 * Gather the ifindex for if up/down events to be
1355 * tagged into this fun
1357 if (peer
->conf_if
&& IN6_IS_ADDR_LINKLOCAL(&peer
->su
.sin6
.sin6_addr
))
1358 ifindex
= peer
->su
.sin6
.sin6_scope_id
;
1360 bnc
= bnc_find(&bgp
->nexthop_cache_table
[AFI_IP6
], &p
, 0, ifindex
);
1364 if (peer
!= bnc
->nht_info
)
1367 for (nhop
= bnc
->nexthop
; nhop
; nhop
= nhop
->next
) {
1368 ifp
= if_lookup_by_index(nhop
->ifindex
, nhop
->vrf_id
);
1373 zclient_send_interface_radv_req(zclient
, nhop
->vrf_id
, ifp
, 0,
1378 /****************************************************************************
1379 * L3 NHGs are used for fast failover of nexthops in the dplane. These are
1380 * the APIs for allocating L3 NHG ids. Management of the L3 NHG itself is
1381 * left to the application using it.
1382 * PS: Currently EVPN host routes is the only app using L3 NHG for fast
1383 * failover of remote ES links.
1384 ***************************************************************************/
1385 static bitfield_t bgp_nh_id_bitmap
;
1386 static uint32_t bgp_l3nhg_start
;
1388 /* XXX - currently we do nothing on the callbacks */
1389 static void bgp_l3nhg_add_cb(const char *name
)
1392 static void bgp_l3nhg_add_nexthop_cb(const struct nexthop_group_cmd
*nhgc
,
1393 const struct nexthop
*nhop
)
1396 static void bgp_l3nhg_del_nexthop_cb(const struct nexthop_group_cmd
*nhgc
,
1397 const struct nexthop
*nhop
)
1400 static void bgp_l3nhg_del_cb(const char *name
)
1404 static void bgp_l3nhg_zebra_init(void)
1406 static bool bgp_l3nhg_zebra_inited
;
1407 if (bgp_l3nhg_zebra_inited
)
1410 bgp_l3nhg_zebra_inited
= true;
1411 bgp_l3nhg_start
= zclient_get_nhg_start(ZEBRA_ROUTE_BGP
);
1412 nexthop_group_init(bgp_l3nhg_add_cb
, bgp_l3nhg_add_nexthop_cb
,
1413 bgp_l3nhg_del_nexthop_cb
, bgp_l3nhg_del_cb
);
1417 void bgp_l3nhg_init(void)
1421 id_max
= MIN(ZEBRA_NHG_PROTO_SPACING
- 1, 16 * 1024);
1422 bf_init(bgp_nh_id_bitmap
, id_max
);
1423 bf_assign_zero_index(bgp_nh_id_bitmap
);
1425 if (BGP_DEBUG(nht
, NHT
) || BGP_DEBUG(evpn_mh
, EVPN_MH_ES
))
1426 zlog_debug("bgp l3_nhg range %u - %u", bgp_l3nhg_start
+ 1,
1427 bgp_l3nhg_start
+ id_max
);
1430 void bgp_l3nhg_finish(void)
1432 bf_free(bgp_nh_id_bitmap
);
1435 uint32_t bgp_l3nhg_id_alloc(void)
1437 uint32_t nhg_id
= 0;
1439 bgp_l3nhg_zebra_init();
1440 bf_assign_index(bgp_nh_id_bitmap
, nhg_id
);
1442 nhg_id
+= bgp_l3nhg_start
;
1447 void bgp_l3nhg_id_free(uint32_t nhg_id
)
1449 if (!nhg_id
|| (nhg_id
<= bgp_l3nhg_start
))
1452 nhg_id
-= bgp_l3nhg_start
;
1454 bf_release_index(bgp_nh_id_bitmap
, nhg_id
);