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(path
->flags
, BGP_PATH_ACCEPT_OWN
) ||
143 CHECK_FLAG(bnc
->flags
, BGP_NEXTHOP_LABELED_VALID
) ||
144 bnc
->bgp
->srv6_enabled
||
145 bgp_isvalid_nexthop_for_ebgp(bnc
, path
) ||
146 bgp_isvalid_nexthop_for_mplsovergre(bnc
, path
)))));
149 static void bgp_unlink_nexthop_check(struct bgp_nexthop_cache
*bnc
)
151 if (LIST_EMPTY(&(bnc
->paths
)) && !bnc
->nht_info
) {
152 if (BGP_DEBUG(nht
, NHT
))
153 zlog_debug("%s: freeing bnc %pFX(%d)(%u)(%s)", __func__
,
154 &bnc
->prefix
, bnc
->ifindex
, bnc
->srte_color
,
155 bnc
->bgp
->name_pretty
);
156 /* only unregister if this is the last nh for this prefix*/
157 if (!bnc_existing_for_prefix(bnc
))
158 unregister_zebra_rnh(bnc
);
163 void bgp_unlink_nexthop(struct bgp_path_info
*path
)
165 struct bgp_nexthop_cache
*bnc
= path
->nexthop
;
170 path_nh_map(path
, NULL
, false);
172 bgp_unlink_nexthop_check(bnc
);
175 void bgp_replace_nexthop_by_peer(struct peer
*from
, struct peer
*to
)
179 struct bgp_nexthop_cache
*bncp
, *bnct
;
181 ifindex_t ifindex
= 0;
183 if (!sockunion2hostprefix(&from
->su
, &pp
))
187 * Gather the ifindex for if up/down events to be
188 * tagged into this fun
190 if (from
->conf_if
&& IN6_IS_ADDR_LINKLOCAL(&from
->su
.sin6
.sin6_addr
))
191 ifindex
= from
->su
.sin6
.sin6_scope_id
;
193 afi
= family2afi(pp
.family
);
194 bncp
= bnc_find(&from
->bgp
->nexthop_cache_table
[afi
], &pp
, 0, ifindex
);
196 if (!sockunion2hostprefix(&to
->su
, &pt
))
200 * Gather the ifindex for if up/down events to be
201 * tagged into this fun
204 if (to
->conf_if
&& IN6_IS_ADDR_LINKLOCAL(&to
->su
.sin6
.sin6_addr
))
205 ifindex
= to
->su
.sin6
.sin6_scope_id
;
206 bnct
= bnc_find(&to
->bgp
->nexthop_cache_table
[afi
], &pt
, 0, ifindex
);
215 void bgp_unlink_nexthop_by_peer(struct peer
*peer
)
218 struct bgp_nexthop_cache
*bnc
;
219 afi_t afi
= family2afi(peer
->su
.sa
.sa_family
);
220 ifindex_t ifindex
= 0;
222 if (!sockunion2hostprefix(&peer
->su
, &p
))
225 * Gather the ifindex for if up/down events to be
226 * tagged into this fun
228 if (afi
== AFI_IP6
&& IN6_IS_ADDR_LINKLOCAL(&peer
->su
.sin6
.sin6_addr
))
229 ifindex
= peer
->su
.sin6
.sin6_scope_id
;
230 bnc
= bnc_find(&peer
->bgp
->nexthop_cache_table
[afi
], &p
, 0, ifindex
);
234 /* cleanup the peer reference */
235 bnc
->nht_info
= NULL
;
237 bgp_unlink_nexthop_check(bnc
);
241 * A route and its nexthop might belong to different VRFs. Therefore,
242 * we need both the bgp_route and bgp_nexthop pointers.
244 int bgp_find_or_add_nexthop(struct bgp
*bgp_route
, struct bgp
*bgp_nexthop
,
245 afi_t afi
, safi_t safi
, struct bgp_path_info
*pi
,
246 struct peer
*peer
, int connected
,
247 const struct prefix
*orig_prefix
)
249 struct bgp_nexthop_cache_head
*tree
= NULL
;
250 struct bgp_nexthop_cache
*bnc
;
252 uint32_t srte_color
= 0;
253 int is_bgp_static_route
= 0;
254 ifindex_t ifindex
= 0;
257 is_bgp_static_route
= ((pi
->type
== ZEBRA_ROUTE_BGP
)
258 && (pi
->sub_type
== BGP_ROUTE_STATIC
))
262 /* Since Extended Next-hop Encoding (RFC5549) support, we want
264 address-family from the next-hop. */
265 if (!is_bgp_static_route
)
266 afi
= BGP_ATTR_MP_NEXTHOP_LEN_IP6(pi
->attr
) ? AFI_IP6
269 /* Validation for the ipv4 mapped ipv6 nexthop. */
270 if (IS_MAPPED_IPV6(&pi
->attr
->mp_nexthop_global
)) {
274 /* This will return true if the global IPv6 NH is a link local
276 if (make_prefix(afi
, pi
, &p
) < 0)
279 if (!is_bgp_static_route
&& orig_prefix
280 && prefix_same(&p
, orig_prefix
)) {
281 if (BGP_DEBUG(nht
, NHT
)) {
283 "%s(%pFX): prefix loops through itself",
289 srte_color
= pi
->attr
->srte_color
;
292 * Gather the ifindex for if up/down events to be
293 * tagged into this fun
295 if (afi
== AFI_IP6
&&
296 IN6_IS_ADDR_LINKLOCAL(&peer
->su
.sin6
.sin6_addr
)) {
297 ifindex
= peer
->su
.sin6
.sin6_scope_id
;
299 if (BGP_DEBUG(nht
, NHT
)) {
301 "%s: Unable to locate ifindex, waiting till we have one",
308 if (!sockunion2hostprefix(&peer
->su
, &p
)) {
309 if (BGP_DEBUG(nht
, NHT
)) {
311 "%s: Attempting to register with unknown AFI %d (not %d or %d)",
312 __func__
, afi
, AFI_IP
, AFI_IP6
);
319 if (is_bgp_static_route
)
320 tree
= &bgp_nexthop
->import_check_table
[afi
];
322 tree
= &bgp_nexthop
->nexthop_cache_table
[afi
];
324 bnc
= bnc_find(tree
, &p
, srte_color
, ifindex
);
326 bnc
= bnc_new(tree
, &p
, srte_color
, ifindex
);
327 bnc
->bgp
= bgp_nexthop
;
328 if (BGP_DEBUG(nht
, NHT
))
329 zlog_debug("Allocated bnc %pFX(%d)(%u)(%s) peer %p",
330 &bnc
->prefix
, bnc
->ifindex
, bnc
->srte_color
,
331 bnc
->bgp
->name_pretty
, peer
);
333 if (BGP_DEBUG(nht
, NHT
))
335 "Found existing bnc %pFX(%d)(%s) flags 0x%x ifindex %d #paths %d peer %p",
336 &bnc
->prefix
, bnc
->ifindex
,
337 bnc
->bgp
->name_pretty
, bnc
->flags
, bnc
->ifindex
,
338 bnc
->path_count
, bnc
->nht_info
);
341 if (pi
&& is_route_parent_evpn(pi
))
342 bnc
->is_evpn_gwip_nexthop
= true;
344 if (is_bgp_static_route
) {
345 SET_FLAG(bnc
->flags
, BGP_STATIC_ROUTE
);
347 /* If we're toggling the type, re-register */
348 if ((CHECK_FLAG(bgp_route
->flags
, BGP_FLAG_IMPORT_CHECK
))
349 && !CHECK_FLAG(bnc
->flags
, BGP_STATIC_ROUTE_EXACT_MATCH
)) {
350 SET_FLAG(bnc
->flags
, BGP_STATIC_ROUTE_EXACT_MATCH
);
351 UNSET_FLAG(bnc
->flags
, BGP_NEXTHOP_REGISTERED
);
352 UNSET_FLAG(bnc
->flags
, BGP_NEXTHOP_VALID
);
353 } else if ((!CHECK_FLAG(bgp_route
->flags
,
354 BGP_FLAG_IMPORT_CHECK
))
355 && CHECK_FLAG(bnc
->flags
,
356 BGP_STATIC_ROUTE_EXACT_MATCH
)) {
357 UNSET_FLAG(bnc
->flags
, BGP_STATIC_ROUTE_EXACT_MATCH
);
358 UNSET_FLAG(bnc
->flags
, BGP_NEXTHOP_REGISTERED
);
359 UNSET_FLAG(bnc
->flags
, BGP_NEXTHOP_VALID
);
362 /* When nexthop is already known, but now requires 'connected'
364 * re-register it. The reverse scenario where the nexthop currently
366 * 'connected' resolution does not need a re-register (i.e., we treat
367 * 'connected-required' as an override) except in the scenario where
369 * is actually a case of tracking a peer for connectivity (e.g., after
370 * disable connected-check).
371 * NOTE: We don't track the number of paths separately for 'connected-
372 * required' vs 'connected-not-required' as this change is not a common
375 else if (connected
&& !CHECK_FLAG(bnc
->flags
, BGP_NEXTHOP_CONNECTED
)) {
376 SET_FLAG(bnc
->flags
, BGP_NEXTHOP_CONNECTED
);
377 UNSET_FLAG(bnc
->flags
, BGP_NEXTHOP_REGISTERED
);
378 UNSET_FLAG(bnc
->flags
, BGP_NEXTHOP_VALID
);
379 } else if (peer
&& !connected
380 && CHECK_FLAG(bnc
->flags
, BGP_NEXTHOP_CONNECTED
)) {
381 UNSET_FLAG(bnc
->flags
, BGP_NEXTHOP_CONNECTED
);
382 UNSET_FLAG(bnc
->flags
, BGP_NEXTHOP_REGISTERED
);
383 UNSET_FLAG(bnc
->flags
, BGP_NEXTHOP_VALID
);
385 if (peer
&& (bnc
->ifindex
!= ifindex
)) {
386 UNSET_FLAG(bnc
->flags
, BGP_NEXTHOP_REGISTERED
);
387 UNSET_FLAG(bnc
->flags
, BGP_NEXTHOP_VALID
);
388 bnc
->ifindex
= ifindex
;
390 if (bgp_route
->inst_type
== BGP_INSTANCE_TYPE_VIEW
) {
391 SET_FLAG(bnc
->flags
, BGP_NEXTHOP_REGISTERED
);
392 SET_FLAG(bnc
->flags
, BGP_NEXTHOP_VALID
);
393 } else if (!CHECK_FLAG(bnc
->flags
, BGP_NEXTHOP_REGISTERED
)
394 && !is_default_host_route(&bnc
->prefix
))
395 register_zebra_rnh(bnc
);
397 if (pi
&& pi
->nexthop
!= bnc
) {
398 /* Unlink from existing nexthop cache, if any. This will also
400 * the nexthop cache entry, if appropriate.
402 bgp_unlink_nexthop(pi
);
404 /* updates NHT pi list reference */
405 path_nh_map(pi
, bnc
, true);
407 if (CHECK_FLAG(bnc
->flags
, BGP_NEXTHOP_VALID
) && bnc
->metric
)
408 (bgp_path_info_extra_get(pi
))->igpmetric
= bnc
->metric
;
410 pi
->extra
->igpmetric
= 0;
413 * Let's not accidentally save the peer data for a peer
414 * we are going to throw away in a second or so.
415 * When we come back around we'll fix up this
416 * data properly in replace_nexthop_by_peer
418 if (CHECK_FLAG(peer
->flags
, PEER_FLAG_CONFIG_NODE
))
419 bnc
->nht_info
= (void *)peer
; /* NHT peer reference */
423 * We are cheating here. Views have no associated underlying
424 * ability to detect nexthops. So when we have a view
425 * just tell everyone the nexthop is valid
427 if (bgp_route
->inst_type
== BGP_INSTANCE_TYPE_VIEW
)
429 else if (safi
== SAFI_UNICAST
&& pi
&&
430 pi
->sub_type
== BGP_ROUTE_IMPORTED
&& pi
->extra
&&
431 pi
->extra
->num_labels
&& !bnc
->is_evpn_gwip_nexthop
)
432 return bgp_isvalid_nexthop_for_mpls(bnc
, pi
);
434 return (bgp_isvalid_nexthop(bnc
));
437 void bgp_delete_connected_nexthop(afi_t afi
, struct peer
*peer
)
439 struct bgp_nexthop_cache
*bnc
;
441 ifindex_t ifindex
= 0;
446 if (!sockunion2hostprefix(&peer
->su
, &p
))
449 * Gather the ifindex for if up/down events to be
450 * tagged into this fun
452 if (afi
== AFI_IP6
&& IN6_IS_ADDR_LINKLOCAL(&peer
->su
.sin6
.sin6_addr
))
453 ifindex
= peer
->su
.sin6
.sin6_scope_id
;
454 bnc
= bnc_find(&peer
->bgp
->nexthop_cache_table
[family2afi(p
.family
)],
457 if (BGP_DEBUG(nht
, NHT
))
459 "Cannot find connected NHT node for peer %s(%s)",
460 peer
->host
, peer
->bgp
->name_pretty
);
464 if (bnc
->nht_info
!= peer
) {
465 if (BGP_DEBUG(nht
, NHT
))
467 "Connected NHT %p node for peer %s(%s) points to %p",
468 bnc
, peer
->host
, bnc
->bgp
->name_pretty
,
473 bnc
->nht_info
= NULL
;
475 if (LIST_EMPTY(&(bnc
->paths
))) {
476 if (BGP_DEBUG(nht
, NHT
))
478 "Freeing connected NHT node %p for peer %s(%s)",
479 bnc
, peer
->host
, bnc
->bgp
->name_pretty
);
480 unregister_zebra_rnh(bnc
);
485 static void bgp_process_nexthop_update(struct bgp_nexthop_cache
*bnc
,
486 struct zapi_route
*nhr
,
489 struct nexthop
*nexthop
;
490 struct nexthop
*oldnh
;
491 struct nexthop
*nhlist_head
= NULL
;
492 struct nexthop
*nhlist_tail
= NULL
;
494 bool evpn_resolved
= false;
496 bnc
->last_update
= monotime(NULL
);
497 bnc
->change_flags
= 0;
499 /* debug print the input */
500 if (BGP_DEBUG(nht
, NHT
)) {
501 char bnc_buf
[BNC_FLAG_DUMP_SIZE
];
504 "%s(%u): Rcvd NH update %pFX(%u)%u) - metric %d/%d #nhops %d/%d flags %s",
505 bnc
->bgp
->name_pretty
, bnc
->bgp
->vrf_id
, &nhr
->prefix
,
506 bnc
->ifindex
, bnc
->srte_color
, nhr
->metric
, bnc
->metric
,
507 nhr
->nexthop_num
, bnc
->nexthop_num
,
508 bgp_nexthop_dump_bnc_flags(bnc
, bnc_buf
,
512 if (nhr
->metric
!= bnc
->metric
)
513 bnc
->change_flags
|= BGP_NEXTHOP_METRIC_CHANGED
;
515 if (nhr
->nexthop_num
!= bnc
->nexthop_num
)
516 bnc
->change_flags
|= BGP_NEXTHOP_CHANGED
;
518 if (import_check
&& (nhr
->type
== ZEBRA_ROUTE_BGP
||
519 !prefix_same(&bnc
->prefix
, &nhr
->prefix
))) {
520 SET_FLAG(bnc
->change_flags
, BGP_NEXTHOP_CHANGED
);
521 UNSET_FLAG(bnc
->flags
, BGP_NEXTHOP_VALID
);
522 UNSET_FLAG(bnc
->flags
, BGP_NEXTHOP_LABELED_VALID
);
523 UNSET_FLAG(bnc
->flags
, BGP_NEXTHOP_EVPN_INCOMPLETE
);
525 bnc_nexthop_free(bnc
);
528 if (BGP_DEBUG(nht
, NHT
))
530 "%s: Import Check does not resolve to the same prefix for %pFX received %pFX or matching route is BGP",
531 __func__
, &bnc
->prefix
, &nhr
->prefix
);
532 } else if (nhr
->nexthop_num
) {
533 struct peer
*peer
= bnc
->nht_info
;
535 /* notify bgp fsm if nbr ip goes from invalid->valid */
536 if (!bnc
->nexthop_num
)
537 UNSET_FLAG(bnc
->flags
, BGP_NEXTHOP_PEER_NOTIFIED
);
539 if (!bnc
->is_evpn_gwip_nexthop
)
540 bnc
->flags
|= BGP_NEXTHOP_VALID
;
541 bnc
->metric
= nhr
->metric
;
542 bnc
->nexthop_num
= nhr
->nexthop_num
;
544 bnc
->flags
&= ~BGP_NEXTHOP_LABELED_VALID
; /* check below */
546 for (i
= 0; i
< nhr
->nexthop_num
; i
++) {
549 nexthop
= nexthop_from_zapi_nexthop(&nhr
->nexthops
[i
]);
552 * Turn on RA for the v6 nexthops
553 * we receive from bgp. This is to allow us
554 * to work with v4 routing over v6 nexthops
556 if (peer
&& !peer
->ifp
557 && CHECK_FLAG(peer
->flags
,
558 PEER_FLAG_CAPABILITY_ENHE
)
559 && nhr
->prefix
.family
== AF_INET6
560 && nexthop
->type
!= NEXTHOP_TYPE_BLACKHOLE
) {
561 struct interface
*ifp
;
563 ifp
= if_lookup_by_index(nexthop
->ifindex
,
566 zclient_send_interface_radv_req(
567 zclient
, nexthop
->vrf_id
, ifp
,
569 BGP_UNNUM_DEFAULT_RA_INTERVAL
);
571 /* There is at least one label-switched path */
572 if (nexthop
->nh_label
&&
573 nexthop
->nh_label
->num_labels
) {
575 bnc
->flags
|= BGP_NEXTHOP_LABELED_VALID
;
576 num_labels
= nexthop
->nh_label
->num_labels
;
579 if (BGP_DEBUG(nht
, NHT
)) {
580 char buf
[NEXTHOP_STRLEN
];
582 " nhop via %s (%d labels)",
583 nexthop2str(nexthop
, buf
, sizeof(buf
)),
588 nhlist_tail
->next
= nexthop
;
589 nhlist_tail
= nexthop
;
591 nhlist_tail
= nexthop
;
592 nhlist_head
= nexthop
;
595 /* No need to evaluate the nexthop if we have already
597 * that there has been a change.
599 if (bnc
->change_flags
& BGP_NEXTHOP_CHANGED
)
602 for (oldnh
= bnc
->nexthop
; oldnh
; oldnh
= oldnh
->next
)
603 if (nexthop_same(oldnh
, nexthop
))
607 bnc
->change_flags
|= BGP_NEXTHOP_CHANGED
;
609 bnc_nexthop_free(bnc
);
610 bnc
->nexthop
= nhlist_head
;
613 * Gateway IP nexthop is L3 reachable. Mark it as
614 * BGP_NEXTHOP_VALID only if it is recursively resolved with a
616 * Else, mark it as BGP_NEXTHOP_EVPN_INCOMPLETE.
617 * When its mapping with EVPN RT-2 is established, unset
618 * BGP_NEXTHOP_EVPN_INCOMPLETE and set BGP_NEXTHOP_VALID.
620 if (bnc
->is_evpn_gwip_nexthop
) {
621 evpn_resolved
= bgp_evpn_is_gateway_ip_resolved(bnc
);
623 if (BGP_DEBUG(nht
, NHT
))
625 "EVPN gateway IP %pFX recursive MAC/IP lookup %s",
627 (evpn_resolved
? "successful"
631 bnc
->flags
|= BGP_NEXTHOP_VALID
;
632 bnc
->flags
&= ~BGP_NEXTHOP_EVPN_INCOMPLETE
;
633 bnc
->change_flags
|= BGP_NEXTHOP_MACIP_CHANGED
;
635 bnc
->flags
|= BGP_NEXTHOP_EVPN_INCOMPLETE
;
636 bnc
->flags
&= ~BGP_NEXTHOP_VALID
;
640 bnc
->flags
&= ~BGP_NEXTHOP_EVPN_INCOMPLETE
;
641 bnc
->flags
&= ~BGP_NEXTHOP_VALID
;
642 bnc
->flags
&= ~BGP_NEXTHOP_LABELED_VALID
;
643 bnc
->nexthop_num
= nhr
->nexthop_num
;
645 /* notify bgp fsm if nbr ip goes from valid->invalid */
646 UNSET_FLAG(bnc
->flags
, BGP_NEXTHOP_PEER_NOTIFIED
);
648 bnc_nexthop_free(bnc
);
655 static void bgp_nht_ifp_table_handle(struct bgp
*bgp
,
656 struct bgp_nexthop_cache_head
*table
,
657 struct interface
*ifp
, bool up
)
659 struct bgp_nexthop_cache
*bnc
;
661 frr_each (bgp_nexthop_cache
, table
, bnc
) {
662 if (bnc
->ifindex
!= ifp
->ifindex
)
665 bnc
->last_update
= monotime(NULL
);
666 bnc
->change_flags
= 0;
669 * For interface based routes ( ala the v6 LL routes
670 * that this was written for ) the metric received
671 * for the connected route is 0 not 1.
675 SET_FLAG(bnc
->flags
, BGP_NEXTHOP_VALID
);
676 SET_FLAG(bnc
->change_flags
, BGP_NEXTHOP_CHANGED
);
677 bnc
->nexthop_num
= 1;
679 UNSET_FLAG(bnc
->flags
, BGP_NEXTHOP_PEER_NOTIFIED
);
680 UNSET_FLAG(bnc
->flags
, BGP_NEXTHOP_VALID
);
681 SET_FLAG(bnc
->change_flags
, BGP_NEXTHOP_CHANGED
);
682 bnc
->nexthop_num
= 0;
688 static void bgp_nht_ifp_handle(struct interface
*ifp
, bool up
)
692 bgp
= ifp
->vrf
->info
;
696 bgp_nht_ifp_table_handle(bgp
, &bgp
->nexthop_cache_table
[AFI_IP
], ifp
,
698 bgp_nht_ifp_table_handle(bgp
, &bgp
->import_check_table
[AFI_IP
], ifp
,
700 bgp_nht_ifp_table_handle(bgp
, &bgp
->nexthop_cache_table
[AFI_IP6
], ifp
,
702 bgp_nht_ifp_table_handle(bgp
, &bgp
->import_check_table
[AFI_IP6
], ifp
,
706 void bgp_nht_ifp_up(struct interface
*ifp
)
708 bgp_nht_ifp_handle(ifp
, true);
711 void bgp_nht_ifp_down(struct interface
*ifp
)
713 bgp_nht_ifp_handle(ifp
, false);
716 static void bgp_nht_ifp_initial(struct thread
*thread
)
718 ifindex_t ifindex
= THREAD_VAL(thread
);
719 struct bgp
*bgp
= THREAD_ARG(thread
);
720 struct interface
*ifp
= if_lookup_by_index(ifindex
, bgp
->vrf_id
);
725 if (BGP_DEBUG(nht
, NHT
))
727 "Handle NHT initial update for Intf %s(%d) status %s",
728 ifp
->name
, ifp
->ifindex
, if_is_up(ifp
) ? "up" : "down");
733 bgp_nht_ifp_down(ifp
);
737 * So the bnc code has the ability to handle interface up/down
738 * events to properly handle v6 LL peering.
739 * What is happening here:
740 * The event system for peering expects the nht code to
741 * report on the tracking events after we move to active
742 * So let's give the system a chance to report on that event
743 * in a manner that is expected.
745 void bgp_nht_interface_events(struct peer
*peer
)
747 struct bgp
*bgp
= peer
->bgp
;
748 struct bgp_nexthop_cache_head
*table
;
749 struct bgp_nexthop_cache
*bnc
;
751 ifindex_t ifindex
= 0;
753 if (!IN6_IS_ADDR_LINKLOCAL(&peer
->su
.sin6
.sin6_addr
))
756 if (!sockunion2hostprefix(&peer
->su
, &p
))
759 * Gather the ifindex for if up/down events to be
760 * tagged into this fun
762 if (peer
->conf_if
&& IN6_IS_ADDR_LINKLOCAL(&peer
->su
.sin6
.sin6_addr
))
763 ifindex
= peer
->su
.sin6
.sin6_scope_id
;
765 table
= &bgp
->nexthop_cache_table
[AFI_IP6
];
766 bnc
= bnc_find(table
, &p
, 0, ifindex
);
771 thread_add_event(bm
->master
, bgp_nht_ifp_initial
, bnc
->bgp
,
775 void bgp_parse_nexthop_update(int command
, vrf_id_t vrf_id
)
777 struct bgp_nexthop_cache_head
*tree
= NULL
;
778 struct bgp_nexthop_cache
*bnc_nhc
, *bnc_import
;
781 struct zapi_route nhr
;
784 bgp
= bgp_lookup_by_vrf_id(vrf_id
);
788 "parse nexthop update: instance not found for vrf_id %u",
793 if (!zapi_nexthop_update_decode(zclient
->ibuf
, &match
, &nhr
)) {
794 zlog_err("%s[%s]: Failure to decode nexthop update", __func__
,
799 afi
= family2afi(match
.family
);
800 tree
= &bgp
->nexthop_cache_table
[afi
];
802 bnc_nhc
= bnc_find(tree
, &match
, nhr
.srte_color
, 0);
804 if (BGP_DEBUG(nht
, NHT
))
806 "parse nexthop update(%pFX(%u)(%s)): bnc info not found for nexthop cache",
807 &nhr
.prefix
, nhr
.srte_color
, bgp
->name_pretty
);
809 bgp_process_nexthop_update(bnc_nhc
, &nhr
, false);
811 tree
= &bgp
->import_check_table
[afi
];
813 bnc_import
= bnc_find(tree
, &match
, nhr
.srte_color
, 0);
815 if (BGP_DEBUG(nht
, NHT
))
817 "parse nexthop update(%pFX(%u)(%s)): bnc info not found for import check",
818 &nhr
.prefix
, nhr
.srte_color
, bgp
->name_pretty
);
820 bgp_process_nexthop_update(bnc_import
, &nhr
, true);
823 * HACK: if any BGP route is dependant on an SR-policy that doesn't
824 * exist, zebra will never send NH updates relative to that policy. In
825 * that case, whenever we receive an update about a colorless NH, update
826 * the corresponding colorful NHs that share the same endpoint but that
827 * are inactive. This ugly hack should work around the problem at the
828 * cost of a performance pernalty. Long term, what should be done is to
829 * make zebra's RNH subsystem aware of SR-TE colors (like bgpd is),
830 * which should provide a better infrastructure to solve this issue in
831 * a more efficient and elegant way.
833 if (nhr
.srte_color
== 0 && bnc_nhc
) {
834 struct bgp_nexthop_cache
*bnc_iter
;
836 frr_each (bgp_nexthop_cache
, &bgp
->nexthop_cache_table
[afi
],
838 if (!prefix_same(&bnc_nhc
->prefix
, &bnc_iter
->prefix
) ||
839 bnc_iter
->srte_color
== 0 ||
840 CHECK_FLAG(bnc_iter
->flags
, BGP_NEXTHOP_VALID
))
843 bgp_process_nexthop_update(bnc_iter
, &nhr
, false);
849 * Cleanup nexthop registration and status information for BGP nexthops
850 * pertaining to this VRF. This is invoked upon VRF deletion.
852 void bgp_cleanup_nexthops(struct bgp
*bgp
)
854 for (afi_t afi
= AFI_IP
; afi
< AFI_MAX
; afi
++) {
855 struct bgp_nexthop_cache
*bnc
;
857 frr_each (bgp_nexthop_cache
, &bgp
->nexthop_cache_table
[afi
],
859 /* Clear relevant flags. */
860 UNSET_FLAG(bnc
->flags
, BGP_NEXTHOP_VALID
);
861 UNSET_FLAG(bnc
->flags
, BGP_NEXTHOP_REGISTERED
);
862 UNSET_FLAG(bnc
->flags
, BGP_NEXTHOP_PEER_NOTIFIED
);
863 UNSET_FLAG(bnc
->flags
, BGP_NEXTHOP_EVPN_INCOMPLETE
);
869 * make_prefix - make a prefix structure from the path (essentially
872 static int make_prefix(int afi
, struct bgp_path_info
*pi
, struct prefix
*p
)
875 int is_bgp_static
= ((pi
->type
== ZEBRA_ROUTE_BGP
)
876 && (pi
->sub_type
== BGP_ROUTE_STATIC
))
879 struct bgp_dest
*net
= pi
->net
;
880 const struct prefix
*p_orig
= bgp_dest_get_prefix(net
);
883 if (p_orig
->family
== AF_FLOWSPEC
) {
886 return bgp_flowspec_get_first_nh(pi
->peer
->bgp
,
889 memset(p
, 0, sizeof(struct prefix
));
894 p
->u
.prefix4
= p_orig
->u
.prefix4
;
895 p
->prefixlen
= p_orig
->prefixlen
;
897 if (IS_MAPPED_IPV6(&pi
->attr
->mp_nexthop_global
)) {
898 ipv4_mapped_ipv6_to_ipv4(
899 &pi
->attr
->mp_nexthop_global
, &ipv4
);
901 p
->prefixlen
= IPV4_MAX_BITLEN
;
903 if (p_orig
->family
== AF_EVPN
)
905 pi
->attr
->mp_nexthop_global_in
;
907 p
->u
.prefix4
= pi
->attr
->nexthop
;
908 p
->prefixlen
= IPV4_MAX_BITLEN
;
913 p
->family
= AF_INET6
;
916 p
->u
.prefix6
= p_orig
->u
.prefix6
;
917 p
->prefixlen
= p_orig
->prefixlen
;
919 /* If we receive MP_REACH nexthop with ::(LL)
920 * or LL(LL), use LL address as nexthop cache.
922 if (pi
->attr
->mp_nexthop_len
923 == BGP_ATTR_NHLEN_IPV6_GLOBAL_AND_LL
924 && (IN6_IS_ADDR_UNSPECIFIED(
925 &pi
->attr
->mp_nexthop_global
)
926 || IN6_IS_ADDR_LINKLOCAL(
927 &pi
->attr
->mp_nexthop_global
)))
928 p
->u
.prefix6
= pi
->attr
->mp_nexthop_local
;
929 /* If we receive MR_REACH with (GA)::(LL)
930 * then check for route-map to choose GA or LL
932 else if (pi
->attr
->mp_nexthop_len
933 == BGP_ATTR_NHLEN_IPV6_GLOBAL_AND_LL
) {
934 if (pi
->attr
->mp_nexthop_prefer_global
)
936 pi
->attr
->mp_nexthop_global
;
939 pi
->attr
->mp_nexthop_local
;
941 p
->u
.prefix6
= pi
->attr
->mp_nexthop_global
;
942 p
->prefixlen
= IPV6_MAX_BITLEN
;
946 if (BGP_DEBUG(nht
, NHT
)) {
948 "%s: Attempting to make prefix with unknown AFI %d (not %d or %d)",
949 __func__
, afi
, AFI_IP
, AFI_IP6
);
957 * sendmsg_zebra_rnh -- Format and send a nexthop register/Unregister
960 * struct bgp_nexthop_cache *bnc -- the nexthop structure.
961 * int command -- command to send to zebra
965 static void sendmsg_zebra_rnh(struct bgp_nexthop_cache
*bnc
, int command
)
967 bool exact_match
= false;
968 bool resolve_via_default
= false;
974 /* Don't try to register if Zebra doesn't know of this instance. */
975 if (!IS_BGP_INST_KNOWN_TO_ZEBRA(bnc
->bgp
)) {
976 if (BGP_DEBUG(zebra
, ZEBRA
))
978 "%s: No zebra instance to talk to, not installing NHT entry",
983 if (!bgp_zebra_num_connects()) {
984 if (BGP_DEBUG(zebra
, ZEBRA
))
986 "%s: We have not connected yet, cannot send nexthops",
989 if (command
== ZEBRA_NEXTHOP_REGISTER
) {
990 if (CHECK_FLAG(bnc
->flags
, BGP_NEXTHOP_CONNECTED
))
992 if (CHECK_FLAG(bnc
->flags
, BGP_STATIC_ROUTE_EXACT_MATCH
))
993 resolve_via_default
= true;
996 if (BGP_DEBUG(zebra
, ZEBRA
))
997 zlog_debug("%s: sending cmd %s for %pFX (vrf %s)", __func__
,
998 zserv_command_string(command
), &bnc
->prefix
,
999 bnc
->bgp
->name_pretty
);
1001 ret
= zclient_send_rnh(zclient
, command
, &bnc
->prefix
, SAFI_UNICAST
,
1002 exact_match
, resolve_via_default
,
1004 if (ret
== ZCLIENT_SEND_FAILURE
) {
1005 flog_warn(EC_BGP_ZEBRA_SEND
,
1006 "sendmsg_nexthop: zclient_send_message() failed");
1010 if (command
== ZEBRA_NEXTHOP_REGISTER
)
1011 SET_FLAG(bnc
->flags
, BGP_NEXTHOP_REGISTERED
);
1012 else if (command
== ZEBRA_NEXTHOP_UNREGISTER
)
1013 UNSET_FLAG(bnc
->flags
, BGP_NEXTHOP_REGISTERED
);
1018 * register_zebra_rnh - register a NH/route with Zebra for notification
1019 * when the route or the route to the nexthop changes.
1021 * struct bgp_nexthop_cache *bnc
1025 static void register_zebra_rnh(struct bgp_nexthop_cache
*bnc
)
1027 /* Check if we have already registered */
1028 if (bnc
->flags
& BGP_NEXTHOP_REGISTERED
)
1032 SET_FLAG(bnc
->flags
, BGP_NEXTHOP_REGISTERED
);
1036 sendmsg_zebra_rnh(bnc
, ZEBRA_NEXTHOP_REGISTER
);
1040 * unregister_zebra_rnh -- Unregister the route/nexthop from Zebra.
1042 * struct bgp_nexthop_cache *bnc
1046 static void unregister_zebra_rnh(struct bgp_nexthop_cache
*bnc
)
1048 /* Check if we have already registered */
1049 if (!CHECK_FLAG(bnc
->flags
, BGP_NEXTHOP_REGISTERED
))
1053 UNSET_FLAG(bnc
->flags
, BGP_NEXTHOP_REGISTERED
);
1057 sendmsg_zebra_rnh(bnc
, ZEBRA_NEXTHOP_UNREGISTER
);
1061 * evaluate_paths - Evaluate the paths/nets associated with a nexthop.
1063 * struct bgp_nexthop_cache *bnc -- the nexthop structure.
1067 void evaluate_paths(struct bgp_nexthop_cache
*bnc
)
1069 struct bgp_dest
*dest
;
1070 struct bgp_path_info
*path
;
1072 struct peer
*peer
= (struct peer
*)bnc
->nht_info
;
1073 struct bgp_table
*table
;
1075 struct bgp
*bgp_path
;
1076 const struct prefix
*p
;
1078 if (BGP_DEBUG(nht
, NHT
)) {
1079 char bnc_buf
[BNC_FLAG_DUMP_SIZE
];
1080 char chg_buf
[BNC_FLAG_DUMP_SIZE
];
1083 "NH update for %pFX(%d)(%u)(%s) - flags %s chgflags %s- evaluate paths",
1084 &bnc
->prefix
, bnc
->ifindex
, bnc
->srte_color
,
1085 bnc
->bgp
->name_pretty
,
1086 bgp_nexthop_dump_bnc_flags(bnc
, bnc_buf
,
1088 bgp_nexthop_dump_bnc_change_flags(bnc
, chg_buf
,
1092 LIST_FOREACH (path
, &(bnc
->paths
), nh_thread
) {
1093 if (!(path
->type
== ZEBRA_ROUTE_BGP
1094 && ((path
->sub_type
== BGP_ROUTE_NORMAL
)
1095 || (path
->sub_type
== BGP_ROUTE_STATIC
)
1096 || (path
->sub_type
== BGP_ROUTE_IMPORTED
))))
1100 assert(dest
&& bgp_dest_table(dest
));
1101 p
= bgp_dest_get_prefix(dest
);
1102 afi
= family2afi(p
->family
);
1103 table
= bgp_dest_table(dest
);
1107 * handle routes from other VRFs (they can have a
1108 * nexthop in THIS VRF). bgp_path is the bgp instance
1109 * that owns the route referencing this nexthop.
1111 bgp_path
= table
->bgp
;
1114 * Path becomes valid/invalid depending on whether the nexthop
1115 * reachable/unreachable.
1117 * In case of unicast routes that were imported from vpn
1118 * and that have labels, they are valid only if there are
1119 * nexthops with labels
1121 * If the nexthop is EVPN gateway-IP,
1122 * do not check for a valid label.
1125 bool bnc_is_valid_nexthop
= false;
1126 bool path_valid
= false;
1128 if (safi
== SAFI_UNICAST
&& path
->sub_type
== BGP_ROUTE_IMPORTED
1129 && path
->extra
&& path
->extra
->num_labels
1130 && (path
->attr
->evpn_overlay
.type
1131 != OVERLAY_INDEX_GATEWAY_IP
)) {
1132 bnc_is_valid_nexthop
=
1133 bgp_isvalid_nexthop_for_mpls(bnc
, path
) ? true
1136 if (bgp_update_martian_nexthop(
1137 bnc
->bgp
, afi
, safi
, path
->type
,
1138 path
->sub_type
, path
->attr
, dest
)) {
1139 if (BGP_DEBUG(nht
, NHT
))
1141 "%s: prefix %pBD (vrf %s), ignoring path due to martian or self-next-hop",
1142 __func__
, dest
, bgp_path
->name
);
1144 bnc_is_valid_nexthop
=
1145 bgp_isvalid_nexthop(bnc
) ? true : false;
1148 if (BGP_DEBUG(nht
, NHT
)) {
1151 "... eval path %d/%d %pBD RD %pRD %s flags 0x%x",
1153 (struct prefix_rd
*)bgp_dest_get_prefix(
1155 bgp_path
->name_pretty
, path
->flags
);
1158 "... eval path %d/%d %pBD %s flags 0x%x",
1159 afi
, safi
, dest
, bgp_path
->name_pretty
,
1163 /* Skip paths marked for removal or as history. */
1164 if (CHECK_FLAG(path
->flags
, BGP_PATH_REMOVED
)
1165 || CHECK_FLAG(path
->flags
, BGP_PATH_HISTORY
))
1168 /* Copy the metric to the path. Will be used for bestpath
1170 if (bgp_isvalid_nexthop(bnc
) && bnc
->metric
)
1171 (bgp_path_info_extra_get(path
))->igpmetric
=
1173 else if (path
->extra
)
1174 path
->extra
->igpmetric
= 0;
1176 if (CHECK_FLAG(bnc
->change_flags
, BGP_NEXTHOP_METRIC_CHANGED
)
1177 || CHECK_FLAG(bnc
->change_flags
, BGP_NEXTHOP_CHANGED
)
1178 || path
->attr
->srte_color
!= 0)
1179 SET_FLAG(path
->flags
, BGP_PATH_IGP_CHANGED
);
1181 path_valid
= CHECK_FLAG(path
->flags
, BGP_PATH_VALID
);
1182 if (path_valid
!= bnc_is_valid_nexthop
) {
1184 /* No longer valid, clear flag; also for EVPN
1185 * routes, unimport from VRFs if needed.
1187 bgp_aggregate_decrement(bgp_path
, p
, path
, afi
,
1189 bgp_path_info_unset_flag(dest
, path
,
1191 if (safi
== SAFI_EVPN
&&
1192 bgp_evpn_is_prefix_nht_supported(bgp_dest_get_prefix(dest
)))
1193 bgp_evpn_unimport_route(bgp_path
,
1194 afi
, safi
, bgp_dest_get_prefix(dest
), path
);
1196 /* Path becomes valid, set flag; also for EVPN
1197 * routes, import from VRFs if needed.
1199 bgp_path_info_set_flag(dest
, path
,
1201 bgp_aggregate_increment(bgp_path
, p
, path
, afi
,
1203 if (safi
== SAFI_EVPN
&&
1204 bgp_evpn_is_prefix_nht_supported(bgp_dest_get_prefix(dest
)))
1205 bgp_evpn_import_route(bgp_path
,
1206 afi
, safi
, bgp_dest_get_prefix(dest
), path
);
1210 bgp_process(bgp_path
, dest
, afi
, safi
);
1214 int valid_nexthops
= bgp_isvalid_nexthop(bnc
);
1216 if (valid_nexthops
) {
1218 * Peering cannot occur across a blackhole nexthop
1220 if (bnc
->nexthop_num
== 1 && bnc
->nexthop
1221 && bnc
->nexthop
->type
== NEXTHOP_TYPE_BLACKHOLE
) {
1222 peer
->last_reset
= PEER_DOWN_WAITING_NHT
;
1225 peer
->last_reset
= PEER_DOWN_WAITING_OPEN
;
1227 peer
->last_reset
= PEER_DOWN_WAITING_NHT
;
1229 if (!CHECK_FLAG(bnc
->flags
, BGP_NEXTHOP_PEER_NOTIFIED
)) {
1230 if (BGP_DEBUG(nht
, NHT
))
1232 "%s: Updating peer (%s(%s)) status with NHT nexthops %d",
1233 __func__
, peer
->host
,
1234 peer
->bgp
->name_pretty
,
1236 bgp_fsm_nht_update(peer
, !!valid_nexthops
);
1237 SET_FLAG(bnc
->flags
, BGP_NEXTHOP_PEER_NOTIFIED
);
1241 RESET_FLAG(bnc
->change_flags
);
1245 * path_nh_map - make or break path-to-nexthop association.
1247 * path - pointer to the path structure
1248 * bnc - pointer to the nexthop structure
1249 * make - if set, make the association. if unset, just break the existing
1252 void path_nh_map(struct bgp_path_info
*path
, struct bgp_nexthop_cache
*bnc
,
1255 if (path
->nexthop
) {
1256 LIST_REMOVE(path
, nh_thread
);
1257 path
->nexthop
->path_count
--;
1258 path
->nexthop
= NULL
;
1261 LIST_INSERT_HEAD(&(bnc
->paths
), path
, nh_thread
);
1262 path
->nexthop
= bnc
;
1263 path
->nexthop
->path_count
++;
1268 * This function is called to register nexthops to zebra
1269 * as that we may have tried to install the nexthops
1270 * before we actually have a zebra connection
1272 void bgp_nht_register_nexthops(struct bgp
*bgp
)
1274 for (afi_t afi
= AFI_IP
; afi
< AFI_MAX
; afi
++) {
1275 struct bgp_nexthop_cache
*bnc
;
1277 frr_each (bgp_nexthop_cache
, &bgp
->nexthop_cache_table
[afi
],
1279 register_zebra_rnh(bnc
);
1284 void bgp_nht_reg_enhe_cap_intfs(struct peer
*peer
)
1287 struct bgp_nexthop_cache
*bnc
;
1288 struct nexthop
*nhop
;
1289 struct interface
*ifp
;
1291 ifindex_t ifindex
= 0;
1297 if (!sockunion2hostprefix(&peer
->su
, &p
)) {
1298 zlog_warn("%s: Unable to convert sockunion to prefix for %s",
1299 __func__
, peer
->host
);
1303 if (p
.family
!= AF_INET6
)
1306 * Gather the ifindex for if up/down events to be
1307 * tagged into this fun
1309 if (peer
->conf_if
&& IN6_IS_ADDR_LINKLOCAL(&peer
->su
.sin6
.sin6_addr
))
1310 ifindex
= peer
->su
.sin6
.sin6_scope_id
;
1312 bnc
= bnc_find(&bgp
->nexthop_cache_table
[AFI_IP6
], &p
, 0, ifindex
);
1316 if (peer
!= bnc
->nht_info
)
1319 for (nhop
= bnc
->nexthop
; nhop
; nhop
= nhop
->next
) {
1320 ifp
= if_lookup_by_index(nhop
->ifindex
, nhop
->vrf_id
);
1325 zclient_send_interface_radv_req(zclient
,
1328 BGP_UNNUM_DEFAULT_RA_INTERVAL
);
1332 void bgp_nht_dereg_enhe_cap_intfs(struct peer
*peer
)
1335 struct bgp_nexthop_cache
*bnc
;
1336 struct nexthop
*nhop
;
1337 struct interface
*ifp
;
1339 ifindex_t ifindex
= 0;
1346 if (!sockunion2hostprefix(&peer
->su
, &p
)) {
1347 zlog_warn("%s: Unable to convert sockunion to prefix for %s",
1348 __func__
, peer
->host
);
1352 if (p
.family
!= AF_INET6
)
1355 * Gather the ifindex for if up/down events to be
1356 * tagged into this fun
1358 if (peer
->conf_if
&& IN6_IS_ADDR_LINKLOCAL(&peer
->su
.sin6
.sin6_addr
))
1359 ifindex
= peer
->su
.sin6
.sin6_scope_id
;
1361 bnc
= bnc_find(&bgp
->nexthop_cache_table
[AFI_IP6
], &p
, 0, ifindex
);
1365 if (peer
!= bnc
->nht_info
)
1368 for (nhop
= bnc
->nexthop
; nhop
; nhop
= nhop
->next
) {
1369 ifp
= if_lookup_by_index(nhop
->ifindex
, nhop
->vrf_id
);
1374 zclient_send_interface_radv_req(zclient
, nhop
->vrf_id
, ifp
, 0,
1379 /****************************************************************************
1380 * L3 NHGs are used for fast failover of nexthops in the dplane. These are
1381 * the APIs for allocating L3 NHG ids. Management of the L3 NHG itself is
1382 * left to the application using it.
1383 * PS: Currently EVPN host routes is the only app using L3 NHG for fast
1384 * failover of remote ES links.
1385 ***************************************************************************/
1386 static bitfield_t bgp_nh_id_bitmap
;
1387 static uint32_t bgp_l3nhg_start
;
1389 /* XXX - currently we do nothing on the callbacks */
1390 static void bgp_l3nhg_add_cb(const char *name
)
1393 static void bgp_l3nhg_add_nexthop_cb(const struct nexthop_group_cmd
*nhgc
,
1394 const struct nexthop
*nhop
)
1397 static void bgp_l3nhg_del_nexthop_cb(const struct nexthop_group_cmd
*nhgc
,
1398 const struct nexthop
*nhop
)
1401 static void bgp_l3nhg_del_cb(const char *name
)
1405 static void bgp_l3nhg_zebra_init(void)
1407 static bool bgp_l3nhg_zebra_inited
;
1408 if (bgp_l3nhg_zebra_inited
)
1411 bgp_l3nhg_zebra_inited
= true;
1412 bgp_l3nhg_start
= zclient_get_nhg_start(ZEBRA_ROUTE_BGP
);
1413 nexthop_group_init(bgp_l3nhg_add_cb
, bgp_l3nhg_add_nexthop_cb
,
1414 bgp_l3nhg_del_nexthop_cb
, bgp_l3nhg_del_cb
);
1418 void bgp_l3nhg_init(void)
1422 id_max
= MIN(ZEBRA_NHG_PROTO_SPACING
- 1, 16 * 1024);
1423 bf_init(bgp_nh_id_bitmap
, id_max
);
1424 bf_assign_zero_index(bgp_nh_id_bitmap
);
1426 if (BGP_DEBUG(nht
, NHT
) || BGP_DEBUG(evpn_mh
, EVPN_MH_ES
))
1427 zlog_debug("bgp l3_nhg range %u - %u", bgp_l3nhg_start
+ 1,
1428 bgp_l3nhg_start
+ id_max
);
1431 void bgp_l3nhg_finish(void)
1433 bf_free(bgp_nh_id_bitmap
);
1436 uint32_t bgp_l3nhg_id_alloc(void)
1438 uint32_t nhg_id
= 0;
1440 bgp_l3nhg_zebra_init();
1441 bf_assign_index(bgp_nh_id_bitmap
, nhg_id
);
1443 nhg_id
+= bgp_l3nhg_start
;
1448 void bgp_l3nhg_id_free(uint32_t nhg_id
)
1450 if (!nhg_id
|| (nhg_id
<= bgp_l3nhg_start
))
1453 nhg_id
-= bgp_l3nhg_start
;
1455 bf_release_index(bgp_nh_id_bitmap
, nhg_id
);