3 * Copyright (C) 2017 Cumulus Networks, Inc.
6 * This program is free software; you can redistribute it and/or modify
7 * it under the terms of the GNU General Public License as published by
8 * the Free Software Foundation; either version 2 of the License, or
9 * (at your option) any later version.
11 * This program 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
29 #include "lib/printfrr.h"
32 #include "pimd/pim_nht.h"
33 #include "pim_instance.h"
37 #include "pim_ifchannel.h"
38 #include "pim_mroute.h"
39 #include "pim_zebra.h"
40 #include "pim_upstream.h"
42 #include "pim_jp_agg.h"
43 #include "pim_zebra.h"
44 #include "pim_zlookup.h"
49 * pim_sendmsg_zebra_rnh -- Format and send a nexthop register/Unregister
52 void pim_sendmsg_zebra_rnh(struct pim_instance
*pim
, struct zclient
*zclient
,
53 struct pim_nexthop_cache
*pnc
, int command
)
58 pim_addr_to_prefix(&p
, pnc
->rpf
.rpf_addr
);
59 ret
= zclient_send_rnh(zclient
, command
, &p
, SAFI_UNICAST
, false, false,
61 if (ret
== ZCLIENT_SEND_FAILURE
)
62 zlog_warn("sendmsg_nexthop: zclient_send_message() failed");
64 if (PIM_DEBUG_PIM_NHT
)
66 "%s: NHT %sregistered addr %pFX(%s) with Zebra ret:%d ",
68 (command
== ZEBRA_NEXTHOP_REGISTER
) ? " " : "de", &p
,
74 struct pim_nexthop_cache
*pim_nexthop_cache_find(struct pim_instance
*pim
,
77 struct pim_nexthop_cache
*pnc
= NULL
;
78 struct pim_nexthop_cache lookup
;
80 lookup
.rpf
.rpf_addr
= rpf
->rpf_addr
;
81 pnc
= hash_lookup(pim
->rpf_hash
, &lookup
);
86 static struct pim_nexthop_cache
*pim_nexthop_cache_add(struct pim_instance
*pim
,
87 struct pim_rpf
*rpf_addr
)
89 struct pim_nexthop_cache
*pnc
;
92 pnc
= XCALLOC(MTYPE_PIM_NEXTHOP_CACHE
,
93 sizeof(struct pim_nexthop_cache
));
94 pnc
->rpf
.rpf_addr
= rpf_addr
->rpf_addr
;
96 pnc
= hash_get(pim
->rpf_hash
, pnc
, hash_alloc_intern
);
98 pnc
->rp_list
= list_new();
99 pnc
->rp_list
->cmp
= pim_rp_list_cmp
;
101 snprintfrr(hash_name
, sizeof(hash_name
), "PNC %pPA(%s) Upstream Hash",
102 &pnc
->rpf
.rpf_addr
, pim
->vrf
->name
);
103 pnc
->upstream_hash
= hash_create_size(8192, pim_upstream_hash_key
,
104 pim_upstream_equal
, hash_name
);
109 static struct pim_nexthop_cache
*pim_nht_get(struct pim_instance
*pim
,
112 struct pim_nexthop_cache
*pnc
= NULL
;
114 struct zclient
*zclient
= NULL
;
116 zclient
= pim_zebra_zclient_get();
117 memset(&rpf
, 0, sizeof(rpf
));
120 pnc
= pim_nexthop_cache_find(pim
, &rpf
);
122 pnc
= pim_nexthop_cache_add(pim
, &rpf
);
123 pim_sendmsg_zebra_rnh(pim
, zclient
, pnc
,
124 ZEBRA_NEXTHOP_REGISTER
);
125 if (PIM_DEBUG_PIM_NHT_DETAIL
)
127 "%s: NHT cache and zebra notification added for %pPA(%s)",
128 __func__
, &addr
, pim
->vrf
->name
);
134 /* TBD: this does several distinct things and should probably be split up.
135 * (checking state vs. returning pnc vs. adding upstream vs. adding rp)
137 int pim_find_or_track_nexthop(struct pim_instance
*pim
, pim_addr addr
,
138 struct pim_upstream
*up
, struct rp_info
*rp
,
139 struct pim_nexthop_cache
*out_pnc
)
141 struct pim_nexthop_cache
*pnc
;
142 struct listnode
*ch_node
= NULL
;
144 pnc
= pim_nht_get(pim
, addr
);
146 assertf(up
|| rp
, "addr=%pPA", &addr
);
149 ch_node
= listnode_lookup(pnc
->rp_list
, rp
);
151 listnode_add_sort(pnc
->rp_list
, rp
);
155 (void)hash_get(pnc
->upstream_hash
, up
, hash_alloc_intern
);
157 if (CHECK_FLAG(pnc
->flags
, PIM_NEXTHOP_VALID
)) {
159 memcpy(out_pnc
, pnc
, sizeof(struct pim_nexthop_cache
));
167 void pim_nht_bsr_add(struct pim_instance
*pim
, struct in_addr addr
)
169 struct pim_nexthop_cache
*pnc
;
171 pnc
= pim_nht_get(pim
, addr
);
175 #endif /* PIM_IPV == 4 */
177 static void pim_nht_drop_maybe(struct pim_instance
*pim
,
178 struct pim_nexthop_cache
*pnc
)
180 if (PIM_DEBUG_PIM_NHT
)
182 "%s: NHT %pPA(%s) rp_list count:%d upstream count:%ld BSR count:%u",
183 __func__
, &pnc
->rpf
.rpf_addr
, pim
->vrf
->name
,
184 pnc
->rp_list
->count
, pnc
->upstream_hash
->count
,
187 if (pnc
->rp_list
->count
== 0 && pnc
->upstream_hash
->count
== 0
188 && pnc
->bsr_count
== 0) {
189 struct zclient
*zclient
= pim_zebra_zclient_get();
191 pim_sendmsg_zebra_rnh(pim
, zclient
, pnc
,
192 ZEBRA_NEXTHOP_UNREGISTER
);
194 list_delete(&pnc
->rp_list
);
195 hash_free(pnc
->upstream_hash
);
197 hash_release(pim
->rpf_hash
, pnc
);
199 nexthops_free(pnc
->nexthop
);
200 XFREE(MTYPE_PIM_NEXTHOP_CACHE
, pnc
);
204 void pim_delete_tracked_nexthop(struct pim_instance
*pim
, pim_addr addr
,
205 struct pim_upstream
*up
, struct rp_info
*rp
)
207 struct pim_nexthop_cache
*pnc
= NULL
;
208 struct pim_nexthop_cache lookup
;
209 struct pim_upstream
*upstream
= NULL
;
211 /* Remove from RPF hash if it is the last entry */
212 lookup
.rpf
.rpf_addr
= addr
;
213 pnc
= hash_lookup(pim
->rpf_hash
, &lookup
);
215 zlog_warn("attempting to delete nonexistent NHT entry %pPA",
221 /* Release the (*, G)upstream from pnc->upstream_hash,
222 * whose Group belongs to the RP getting deleted
224 frr_each (rb_pim_upstream
, &pim
->upstream_head
, upstream
) {
226 struct rp_info
*trp_info
;
228 if (!pim_addr_is_any(upstream
->sg
.src
))
231 pim_addr_to_prefix(&grp
, upstream
->sg
.grp
);
232 trp_info
= pim_rp_find_match_group(pim
, &grp
);
234 hash_release(pnc
->upstream_hash
, upstream
);
236 listnode_delete(pnc
->rp_list
, rp
);
240 hash_release(pnc
->upstream_hash
, up
);
242 pim_nht_drop_maybe(pim
, pnc
);
246 void pim_nht_bsr_del(struct pim_instance
*pim
, struct in_addr addr
)
248 struct pim_nexthop_cache
*pnc
= NULL
;
249 struct pim_nexthop_cache lookup
;
252 * Nothing to do here if the address to unregister
253 * is 0.0.0.0 as that the BSR has not been registered
256 if (addr
.s_addr
== INADDR_ANY
)
259 lookup
.rpf
.rpf_addr
= addr
;
261 pnc
= hash_lookup(pim
->rpf_hash
, &lookup
);
264 zlog_warn("attempting to delete nonexistent NHT BSR entry %pI4",
269 assertf(pnc
->bsr_count
> 0, "addr=%pI4", &addr
);
272 pim_nht_drop_maybe(pim
, pnc
);
275 bool pim_nht_bsr_rpf_check(struct pim_instance
*pim
, struct in_addr bsr_addr
,
276 struct interface
*src_ifp
, pim_addr src_ip
)
278 struct pim_nexthop_cache
*pnc
= NULL
;
279 struct pim_nexthop_cache lookup
;
280 struct pim_neighbor
*nbr
= NULL
;
282 struct interface
*ifp
;
284 lookup
.rpf
.rpf_addr
= bsr_addr
;
286 pnc
= hash_lookup(pim
->rpf_hash
, &lookup
);
287 if (!pnc
|| !CHECK_FLAG(pnc
->flags
, PIM_NEXTHOP_ANSWER_RECEIVED
)) {
288 /* BSM from a new freshly registered BSR - do a synchronous
289 * zebra query since otherwise we'd drop the first packet,
290 * leading to additional delay in picking up BSM data
293 /* FIXME: this should really be moved into a generic NHT
294 * function that does "add and get immediate result" or maybe
295 * "check cache or get immediate result." But until that can
296 * be worked in, here's a copy of the code below :(
298 struct pim_zlookup_nexthop nexthop_tab
[router
->multipath
];
300 struct interface
*ifp
= NULL
;
303 memset(nexthop_tab
, 0, sizeof(nexthop_tab
));
304 num_ifindex
= zclient_lookup_nexthop(
305 pim
, nexthop_tab
, router
->multipath
, bsr_addr
,
306 PIM_NEXTHOP_LOOKUP_MAX
);
308 if (num_ifindex
<= 0)
311 for (i
= 0; i
< num_ifindex
; i
++) {
312 struct pim_zlookup_nexthop
*znh
= &nexthop_tab
[i
];
314 /* pim_zlookup_nexthop has no ->type */
316 /* 1:1 match code below with znh instead of nh */
317 ifp
= if_lookup_by_index(znh
->ifindex
,
320 if (!ifp
|| !ifp
->info
)
323 if (if_is_loopback(ifp
) && if_is_loopback(src_ifp
))
326 nbr
= pim_neighbor_find(ifp
, znh
->nexthop_addr
);
330 return znh
->ifindex
== src_ifp
->ifindex
&&
331 (!pim_addr_cmp(znh
->nexthop_addr
, src_ip
));
336 if (!CHECK_FLAG(pnc
->flags
, PIM_NEXTHOP_VALID
))
339 /* if we accept BSMs from more than one ECMP nexthop, this will cause
340 * BSM message "multiplication" for each ECMP hop. i.e. if you have
341 * 4-way ECMP and 4 hops you end up with 256 copies of each BSM
344 * so... only accept the first (IPv4) valid nexthop as source.
347 for (nh
= pnc
->nexthop
; nh
; nh
= nh
->next
) {
352 case NEXTHOP_TYPE_IPV4
:
353 if (nh
->ifindex
== IFINDEX_INTERNAL
)
357 case NEXTHOP_TYPE_IPV4_IFINDEX
:
358 nhaddr
= nh
->gate
.ipv4
;
361 case NEXTHOP_TYPE_IPV6
:
362 if (nh
->ifindex
== IFINDEX_INTERNAL
)
366 case NEXTHOP_TYPE_IPV6_IFINDEX
:
367 nhaddr
= nh
->gate
.ipv6
;
370 case NEXTHOP_TYPE_IFINDEX
:
378 ifp
= if_lookup_by_index(nh
->ifindex
, pim
->vrf
->vrf_id
);
379 if (!ifp
|| !ifp
->info
)
382 if (if_is_loopback(ifp
) && if_is_loopback(src_ifp
))
385 /* MRIB (IGP) may be pointing at a router where PIM is down */
386 nbr
= pim_neighbor_find(ifp
, nhaddr
);
390 return nh
->ifindex
== src_ifp
->ifindex
&&
391 (!pim_addr_cmp(nhaddr
, src_ip
));
395 #endif /* PIM_IPV == 4 */
397 void pim_rp_nexthop_del(struct rp_info
*rp_info
)
399 rp_info
->rp
.source_nexthop
.interface
= NULL
;
400 rp_info
->rp
.source_nexthop
.mrib_nexthop_addr
= PIMADDR_ANY
;
401 rp_info
->rp
.source_nexthop
.mrib_metric_preference
=
402 router
->infinite_assert_metric
.metric_preference
;
403 rp_info
->rp
.source_nexthop
.mrib_route_metric
=
404 router
->infinite_assert_metric
.route_metric
;
407 /* Update RP nexthop info based on Nexthop update received from Zebra.*/
408 static void pim_update_rp_nh(struct pim_instance
*pim
,
409 struct pim_nexthop_cache
*pnc
)
411 struct listnode
*node
= NULL
;
412 struct rp_info
*rp_info
= NULL
;
414 /*Traverse RP list and update each RP Nexthop info */
415 for (ALL_LIST_ELEMENTS_RO(pnc
->rp_list
, node
, rp_info
)) {
416 if (pim_rpf_addr_is_inaddr_any(&rp_info
->rp
))
419 // Compute PIM RPF using cached nexthop
420 if (!pim_ecmp_nexthop_lookup(pim
, &rp_info
->rp
.source_nexthop
,
421 rp_info
->rp
.rpf_addr
,
423 pim_rp_nexthop_del(rp_info
);
427 /* Update Upstream nexthop info based on Nexthop update received from Zebra.*/
428 static int pim_update_upstream_nh_helper(struct hash_bucket
*bucket
, void *arg
)
430 struct pim_instance
*pim
= (struct pim_instance
*)arg
;
431 struct pim_upstream
*up
= (struct pim_upstream
*)bucket
->data
;
433 enum pim_rpf_result rpf_result
;
436 old
.source_nexthop
.interface
= up
->rpf
.source_nexthop
.interface
;
437 rpf_result
= pim_rpf_update(pim
, up
, &old
, __func__
);
439 /* update kernel multicast forwarding cache (MFC); if the
440 * RPF nbr is now unreachable the MFC has already been updated
443 if (rpf_result
!= PIM_RPF_FAILURE
)
444 pim_upstream_mroute_iif_update(up
->channel_oil
, __func__
);
446 if (rpf_result
== PIM_RPF_CHANGED
||
447 (rpf_result
== PIM_RPF_FAILURE
&& old
.source_nexthop
.interface
))
448 pim_zebra_upstream_rpf_changed(pim
, up
, &old
);
451 if (PIM_DEBUG_PIM_NHT
) {
453 "%s: NHT upstream %s(%s) old ifp %s new ifp %s",
454 __func__
, up
->sg_str
, pim
->vrf
->name
,
455 old
.source_nexthop
.interface
? old
.source_nexthop
458 up
->rpf
.source_nexthop
.interface
? up
->rpf
.source_nexthop
463 return HASHWALK_CONTINUE
;
466 static int pim_update_upstream_nh(struct pim_instance
*pim
,
467 struct pim_nexthop_cache
*pnc
)
469 hash_walk(pnc
->upstream_hash
, pim_update_upstream_nh_helper
, pim
);
471 pim_zebra_update_all_interfaces(pim
);
476 uint32_t pim_compute_ecmp_hash(struct prefix
*src
, struct prefix
*grp
)
483 hash_val
= prefix_hash_key(src
);
485 hash_val
^= prefix_hash_key(grp
);
489 static int pim_ecmp_nexthop_search(struct pim_instance
*pim
,
490 struct pim_nexthop_cache
*pnc
,
491 struct pim_nexthop
*nexthop
, pim_addr src
,
492 struct prefix
*grp
, int neighbor_needed
)
494 struct pim_neighbor
*nbrs
[router
->multipath
], *nbr
= NULL
;
495 struct interface
*ifps
[router
->multipath
];
496 struct nexthop
*nh_node
= NULL
;
497 ifindex_t first_ifindex
;
498 struct interface
*ifp
= NULL
;
499 uint32_t hash_val
= 0, mod_val
= 0;
500 uint8_t nh_iter
= 0, found
= 0;
501 uint32_t i
, num_nbrs
= 0;
502 pim_addr nh_addr
= nexthop
->mrib_nexthop_addr
;
503 pim_addr grp_addr
= pim_addr_from_prefix(grp
);
505 if (!pnc
|| !pnc
->nexthop_num
|| !nexthop
)
508 memset(&nbrs
, 0, sizeof(nbrs
));
509 memset(&ifps
, 0, sizeof(ifps
));
512 // Current Nexthop is VALID, check to stay on the current path.
513 if (nexthop
->interface
&& nexthop
->interface
->info
&&
514 (!pim_addr_is_any(nh_addr
))) {
515 /* User configured knob to explicitly switch
516 to new path is disabled or current path
517 metric is less than nexthop update.
520 if (pim
->ecmp_rebalance_enable
== 0) {
521 uint8_t curr_route_valid
= 0;
522 // Check if current nexthop is present in new updated
524 // If the current nexthop is not valid, candidate to
525 // choose new Nexthop.
526 for (nh_node
= pnc
->nexthop
; nh_node
;
527 nh_node
= nh_node
->next
) {
528 curr_route_valid
= (nexthop
->interface
->ifindex
529 == nh_node
->ifindex
);
530 if (curr_route_valid
)
534 if (curr_route_valid
&&
535 !pim_if_connected_to_source(nexthop
->interface
,
537 nbr
= pim_neighbor_find(
539 nexthop
->mrib_nexthop_addr
);
541 && !if_is_loopback(nexthop
->interface
)) {
542 if (PIM_DEBUG_PIM_NHT
)
544 "%s: current nexthop does not have nbr ",
547 /* update metric even if the upstream
548 * neighbor stays unchanged
550 nexthop
->mrib_metric_preference
=
552 nexthop
->mrib_route_metric
=
554 if (PIM_DEBUG_PIM_NHT
)
556 "%s: (%pPA,%pPA)(%s) current nexthop %s is valid, skipping new path selection",
560 nexthop
->interface
->name
);
568 * Look up all interfaces and neighbors,
569 * store for later usage
571 for (nh_node
= pnc
->nexthop
, i
= 0; nh_node
;
572 nh_node
= nh_node
->next
, i
++) {
574 if_lookup_by_index(nh_node
->ifindex
, pim
->vrf
->vrf_id
);
577 pim_addr nhaddr
= nh_node
->gate
.ipv4
;
579 pim_addr nhaddr
= nh_node
->gate
.ipv6
;
581 nbrs
[i
] = pim_neighbor_find(ifps
[i
], nhaddr
);
582 if (nbrs
[i
] || pim_if_connected_to_source(ifps
[i
], src
))
586 if (pim
->ecmp_enable
) {
587 struct prefix src_pfx
;
588 uint32_t consider
= pnc
->nexthop_num
;
590 if (neighbor_needed
&& num_nbrs
< consider
)
596 // PIM ECMP flag is enable then choose ECMP path.
597 pim_addr_to_prefix(&src_pfx
, src
);
598 hash_val
= pim_compute_ecmp_hash(&src_pfx
, grp
);
599 mod_val
= hash_val
% consider
;
602 for (nh_node
= pnc
->nexthop
; nh_node
&& (found
== 0);
603 nh_node
= nh_node
->next
) {
604 first_ifindex
= nh_node
->ifindex
;
607 if (PIM_DEBUG_PIM_NHT
)
609 "%s %s: could not find interface for ifindex %d (address %pPA(%s))",
610 __FILE__
, __func__
, first_ifindex
, &src
,
612 if (nh_iter
== mod_val
)
613 mod_val
++; // Select nexthpath
618 if (PIM_DEBUG_PIM_NHT
)
620 "%s: multicast not enabled on input interface %s(%s) (ifindex=%d, RPF for source %pPA)",
621 __func__
, ifp
->name
, pim
->vrf
->name
,
622 first_ifindex
, &src
);
623 if (nh_iter
== mod_val
)
624 mod_val
++; // Select nexthpath
629 if (neighbor_needed
&& !pim_if_connected_to_source(ifp
, src
)) {
631 if (!nbr
&& !if_is_loopback(ifp
)) {
632 if (PIM_DEBUG_PIM_NHT
)
634 "%s: pim nbr not found on input interface %s(%s)",
637 if (nh_iter
== mod_val
)
638 mod_val
++; // Select nexthpath
644 if (nh_iter
== mod_val
) {
645 nexthop
->interface
= ifp
;
647 nexthop
->mrib_nexthop_addr
= nh_node
->gate
.ipv4
;
649 nexthop
->mrib_nexthop_addr
= nh_node
->gate
.ipv6
;
651 nexthop
->mrib_metric_preference
= pnc
->distance
;
652 nexthop
->mrib_route_metric
= pnc
->metric
;
653 nexthop
->last_lookup
= src
;
654 nexthop
->last_lookup_time
= pim_time_monotonic_usec();
657 if (PIM_DEBUG_PIM_NHT
)
659 "%s: (%pPA,%pPA)(%s) selected nhop interface %s addr %pPAs mod_val %u iter %d ecmp %d",
660 __func__
, &src
, &grp_addr
,
661 pim
->vrf
->name
, ifp
->name
, &nh_addr
,
662 mod_val
, nh_iter
, pim
->ecmp_enable
);
673 /* This API is used to parse Registered address nexthop update coming from Zebra
675 int pim_parse_nexthop_update(ZAPI_CALLBACK_ARGS
)
677 struct nexthop
*nexthop
;
678 struct nexthop
*nhlist_head
= NULL
;
679 struct nexthop
*nhlist_tail
= NULL
;
682 struct pim_nexthop_cache
*pnc
= NULL
;
683 struct interface
*ifp
= NULL
;
684 struct vrf
*vrf
= vrf_lookup_by_id(vrf_id
);
685 struct pim_instance
*pim
;
686 struct zapi_route nhr
;
693 if (!zapi_nexthop_update_decode(zclient
->ibuf
, &match
, &nhr
)) {
694 zlog_err("%s: Decode of nexthop update from zebra failed",
699 if (cmd
== ZEBRA_NEXTHOP_UPDATE
) {
700 rpf
.rpf_addr
= pim_addr_from_prefix(&match
);
701 pnc
= pim_nexthop_cache_find(pim
, &rpf
);
703 if (PIM_DEBUG_PIM_NHT
)
705 "%s: Skipping NHT update, addr %pPA is not in local cached DB.",
706 __func__
, &rpf
.rpf_addr
);
711 * We do not currently handle ZEBRA_IMPORT_CHECK_UPDATE
716 pnc
->last_update
= pim_time_monotonic_usec();
718 if (nhr
.nexthop_num
) {
719 pnc
->nexthop_num
= 0; // Only increment for pim enabled rpf.
721 for (i
= 0; i
< nhr
.nexthop_num
; i
++) {
722 nexthop
= nexthop_from_zapi_nexthop(&nhr
.nexthops
[i
]);
723 switch (nexthop
->type
) {
724 case NEXTHOP_TYPE_IFINDEX
:
726 * Connected route (i.e. no nexthop), use
727 * RPF address from nexthop cache (i.e.
728 * destination) as PIM nexthop.
731 nexthop
->type
= NEXTHOP_TYPE_IPV4_IFINDEX
;
732 nexthop
->gate
.ipv4
= pnc
->rpf
.rpf_addr
;
734 nexthop
->type
= NEXTHOP_TYPE_IPV6_IFINDEX
;
735 nexthop
->gate
.ipv6
= pnc
->rpf
.rpf_addr
;
739 /* RFC5549 IPv4-over-IPv6 nexthop handling:
740 * if we get an IPv6 nexthop in IPv4 PIM, hunt down a
741 * PIM neighbor and use that instead.
743 case NEXTHOP_TYPE_IPV6_IFINDEX
: {
744 struct interface
*ifp1
= NULL
;
745 struct pim_neighbor
*nbr
= NULL
;
747 ifp1
= if_lookup_by_index(nexthop
->ifindex
,
753 /* FIXME: should really use nbr's
754 * secondary address list here
756 nbr
= pim_neighbor_find_if(ifp1
);
758 /* Overwrite with Nbr address as NH addr */
760 nexthop
->gate
.ipv4
= nbr
->source_addr
;
762 // Mark nexthop address to 0 until PIM
764 nexthop
->gate
.ipv4
= PIMADDR_ANY
;
769 case NEXTHOP_TYPE_IPV6_IFINDEX
:
771 case NEXTHOP_TYPE_IPV6
:
772 case NEXTHOP_TYPE_IPV4
:
773 case NEXTHOP_TYPE_IPV4_IFINDEX
:
774 case NEXTHOP_TYPE_BLACKHOLE
:
775 /* nothing to do for the other nexthop types */
779 ifp
= if_lookup_by_index(nexthop
->ifindex
,
782 if (PIM_DEBUG_PIM_NHT
) {
783 char buf
[NEXTHOP_STRLEN
];
785 "%s: could not find interface for ifindex %d(%s) (addr %s)",
786 __func__
, nexthop
->ifindex
,
788 nexthop2str(nexthop
, buf
,
791 nexthop_free(nexthop
);
795 if (PIM_DEBUG_PIM_NHT
) {
797 pim_addr nhaddr
= nexthop
->gate
.ipv4
;
799 pim_addr nhaddr
= nexthop
->gate
.ipv6
;
802 "%s: NHT addr %pFX(%s) %d-nhop via %pPA(%s) type %d distance:%u metric:%u ",
803 __func__
, &match
, pim
->vrf
->name
, i
+ 1,
804 &nhaddr
, ifp
->name
, nexthop
->type
,
805 nhr
.distance
, nhr
.metric
);
810 * Though Multicast is not enabled on this
811 * Interface store it in database otheriwse we
812 * may miss this update and this will not cause
813 * any issue, because while choosing the path we
814 * are ommitting the Interfaces which are not
817 if (PIM_DEBUG_PIM_NHT
) {
818 char buf
[NEXTHOP_STRLEN
];
821 "%s: multicast not enabled on input interface %s(%s) (ifindex=%d, addr %s)",
825 nexthop2str(nexthop
, buf
,
831 nhlist_tail
->next
= nexthop
;
832 nhlist_tail
= nexthop
;
834 nhlist_tail
= nexthop
;
835 nhlist_head
= nexthop
;
837 // Only keep track of nexthops which are PIM enabled.
840 /* Reset existing pnc->nexthop before assigning new list */
841 nexthops_free(pnc
->nexthop
);
842 pnc
->nexthop
= nhlist_head
;
843 if (pnc
->nexthop_num
) {
844 pnc
->flags
|= PIM_NEXTHOP_VALID
;
845 pnc
->distance
= nhr
.distance
;
846 pnc
->metric
= nhr
.metric
;
849 pnc
->flags
&= ~PIM_NEXTHOP_VALID
;
850 pnc
->nexthop_num
= nhr
.nexthop_num
;
851 nexthops_free(pnc
->nexthop
);
854 SET_FLAG(pnc
->flags
, PIM_NEXTHOP_ANSWER_RECEIVED
);
856 if (PIM_DEBUG_PIM_NHT
)
858 "%s: NHT Update for %pFX(%s) num_nh %d num_pim_nh %d vrf:%u up %ld rp %d",
859 __func__
, &match
, pim
->vrf
->name
, nhr
.nexthop_num
,
860 pnc
->nexthop_num
, vrf_id
, pnc
->upstream_hash
->count
,
861 listcount(pnc
->rp_list
));
863 pim_rpf_set_refresh_time(pim
);
865 if (listcount(pnc
->rp_list
))
866 pim_update_rp_nh(pim
, pnc
);
867 if (pnc
->upstream_hash
->count
)
868 pim_update_upstream_nh(pim
, pnc
);
873 int pim_ecmp_nexthop_lookup(struct pim_instance
*pim
,
874 struct pim_nexthop
*nexthop
, pim_addr src
,
875 struct prefix
*grp
, int neighbor_needed
)
877 struct pim_nexthop_cache
*pnc
;
878 struct pim_zlookup_nexthop nexthop_tab
[router
->multipath
];
879 struct pim_neighbor
*nbrs
[router
->multipath
], *nbr
= NULL
;
882 struct interface
*ifps
[router
->multipath
], *ifp
;
886 uint32_t hash_val
= 0, mod_val
= 0;
887 uint32_t num_nbrs
= 0;
889 if (PIM_DEBUG_PIM_NHT_DETAIL
)
890 zlog_debug("%s: Looking up: %pPA(%s), last lookup time: %lld",
891 __func__
, &src
, pim
->vrf
->name
,
892 nexthop
->last_lookup_time
);
896 pnc
= pim_nexthop_cache_find(pim
, &rpf
);
898 if (CHECK_FLAG(pnc
->flags
, PIM_NEXTHOP_ANSWER_RECEIVED
))
899 return pim_ecmp_nexthop_search(pim
, pnc
, nexthop
, src
, grp
,
903 memset(nexthop_tab
, 0,
904 sizeof(struct pim_zlookup_nexthop
) * router
->multipath
);
906 zclient_lookup_nexthop(pim
, nexthop_tab
, router
->multipath
, src
,
907 PIM_NEXTHOP_LOOKUP_MAX
);
908 if (num_ifindex
< 1) {
909 if (PIM_DEBUG_PIM_NHT
)
911 "%s: could not find nexthop ifindex for address %pPA(%s)",
912 __func__
, &src
, pim
->vrf
->name
);
916 memset(&nbrs
, 0, sizeof(nbrs
));
917 memset(&ifps
, 0, sizeof(ifps
));
920 * Look up all interfaces and neighbors,
921 * store for later usage
923 for (i
= 0; i
< num_ifindex
; i
++) {
924 ifps
[i
] = if_lookup_by_index(nexthop_tab
[i
].ifindex
,
927 nbrs
[i
] = pim_neighbor_find(
928 ifps
[i
], nexthop_tab
[i
].nexthop_addr
);
929 if (nbrs
[i
] || pim_if_connected_to_source(ifps
[i
], src
))
934 // If PIM ECMP enable then choose ECMP path.
935 if (pim
->ecmp_enable
) {
936 struct prefix src_pfx
;
937 uint32_t consider
= num_ifindex
;
939 if (neighbor_needed
&& num_nbrs
< consider
)
945 pim_addr_to_prefix(&src_pfx
, src
);
946 hash_val
= pim_compute_ecmp_hash(&src_pfx
, grp
);
947 mod_val
= hash_val
% consider
;
948 if (PIM_DEBUG_PIM_NHT_DETAIL
)
949 zlog_debug("%s: hash_val %u mod_val %u", __func__
,
954 while (!found
&& (i
< num_ifindex
)) {
955 first_ifindex
= nexthop_tab
[i
].ifindex
;
959 if (PIM_DEBUG_PIM_NHT
)
961 "%s %s: could not find interface for ifindex %d (address %pPA(%s))",
962 __FILE__
, __func__
, first_ifindex
, &src
,
971 if (PIM_DEBUG_PIM_NHT
)
973 "%s: multicast not enabled on input interface %s(%s) (ifindex=%d, RPF for source %pPA)",
974 __func__
, ifp
->name
, pim
->vrf
->name
,
975 first_ifindex
, &src
);
981 if (neighbor_needed
&& !pim_if_connected_to_source(ifp
, src
)) {
983 if (PIM_DEBUG_PIM_NHT_DETAIL
)
984 zlog_debug("ifp name: %s(%s), pim nbr: %p",
985 ifp
->name
, pim
->vrf
->name
, nbr
);
986 if (!nbr
&& !if_is_loopback(ifp
)) {
989 if (PIM_DEBUG_PIM_NHT
)
991 "%s: NBR (%pPA) not found on input interface %s(%s) (RPF for source %pPA)",
993 &nexthop_tab
[i
].nexthop_addr
,
994 ifp
->name
, pim
->vrf
->name
,
1002 if (PIM_DEBUG_PIM_NHT
)
1004 "%s: found nhop %pPA for addr %pPA interface %s(%s) metric %d dist %d",
1005 __func__
, &nexthop_tab
[i
].nexthop_addr
,
1006 &src
, ifp
->name
, pim
->vrf
->name
,
1007 nexthop_tab
[i
].route_metric
,
1008 nexthop_tab
[i
].protocol_distance
);
1009 /* update nexthop data */
1010 nexthop
->interface
= ifp
;
1011 nexthop
->mrib_nexthop_addr
=
1012 nexthop_tab
[i
].nexthop_addr
;
1013 nexthop
->mrib_metric_preference
=
1014 nexthop_tab
[i
].protocol_distance
;
1015 nexthop
->mrib_route_metric
=
1016 nexthop_tab
[i
].route_metric
;
1017 nexthop
->last_lookup
= src
;
1018 nexthop
->last_lookup_time
= pim_time_monotonic_usec();
1031 int pim_ecmp_fib_lookup_if_vif_index(struct pim_instance
*pim
, pim_addr src
,
1034 struct pim_nexthop nhop
;
1038 memset(&nhop
, 0, sizeof(nhop
));
1039 if (!pim_ecmp_nexthop_lookup(pim
, &nhop
, src
, grp
, 1)) {
1040 if (PIM_DEBUG_PIM_NHT
)
1042 "%s: could not find nexthop ifindex for address %pPA(%s)",
1043 __func__
, &src
, pim
->vrf
->name
);
1047 ifindex
= nhop
.interface
->ifindex
;
1048 if (PIM_DEBUG_PIM_NHT
)
1050 "%s: found nexthop ifindex=%d (interface %s(%s)) for address %pPA",
1052 ifindex2ifname(ifindex
, pim
->vrf
->vrf_id
),
1053 pim
->vrf
->name
, &src
);
1055 vif_index
= pim_if_find_vifindex_by_ifindex(pim
, ifindex
);
1057 if (vif_index
< 0) {
1058 if (PIM_DEBUG_PIM_NHT
) {
1060 "%s: low vif_index=%d(%s) < 1 nexthop for address %pPA",
1061 __func__
, vif_index
, pim
->vrf
->name
, &src
);