2 * bgp_updgrp_packet.c: BGP update group packet handling routines
4 * @copyright Copyright (C) 2014 Cumulus Networks, Inc.
6 * @author Avneesh Sachdev <avneesh@sproute.net>
7 * @author Rajesh Varadarajan <rajesh@sproute.net>
8 * @author Pradosh Mohapatra <pradosh@sproute.net>
10 * This file is part of GNU Zebra.
12 * GNU Zebra is free software; you can redistribute it and/or modify it
13 * under the terms of the GNU General Public License as published by the
14 * Free Software Foundation; either version 2, or (at your option) any
17 * GNU Zebra is distributed in the hope that it will be useful, but
18 * WITHOUT ANY WARRANTY; without even the implied warranty of
19 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
20 * General Public License for more details.
22 * You should have received a copy of the GNU General Public License along
23 * with this program; see the file COPYING; if not, write to the Free Software
24 * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
34 #include "sockunion.h"
42 #include "workqueue.h"
47 #include "bgpd/bgpd.h"
48 #include "bgpd/bgp_debug.h"
49 #include "bgpd/bgp_errors.h"
50 #include "bgpd/bgp_fsm.h"
51 #include "bgpd/bgp_route.h"
52 #include "bgpd/bgp_packet.h"
53 #include "bgpd/bgp_advertise.h"
54 #include "bgpd/bgp_updgrp.h"
55 #include "bgpd/bgp_nexthop.h"
56 #include "bgpd/bgp_nht.h"
57 #include "bgpd/bgp_mplsvpn.h"
58 #include "bgpd/bgp_label.h"
67 struct bpacket
*bpacket_alloc()
71 pkt
= (struct bpacket
*)XCALLOC(MTYPE_BGP_PACKET
,
72 sizeof(struct bpacket
));
77 void bpacket_free(struct bpacket
*pkt
)
80 stream_free(pkt
->buffer
);
82 XFREE(MTYPE_BGP_PACKET
, pkt
);
85 void bpacket_queue_init(struct bpacket_queue
*q
)
87 TAILQ_INIT(&(q
->pkts
));
91 * bpacket_queue_sanity_check
93 void bpacket_queue_sanity_check(struct bpacket_queue
__attribute__((__unused__
))
99 pkt
= bpacket_queue_last (q
);
101 assert (!pkt
->buffer
);
104 * Make sure the count of packets is correct.
108 pkt
= bpacket_queue_first (q
);
113 if (num_pkts
> q
->curr_count
)
116 pkt
= TAILQ_NEXT (pkt
, pkt_train
);
119 assert (num_pkts
== q
->curr_count
);
124 * bpacket_queue_add_packet
126 * Internal function of bpacket_queue - and adds a
127 * packet entry to the end of the list.
129 * Users of bpacket_queue should use bpacket_queue_add instead.
131 static void bpacket_queue_add_packet(struct bpacket_queue
*q
,
134 struct bpacket
*last_pkt
;
136 if (TAILQ_EMPTY(&(q
->pkts
)))
137 TAILQ_INSERT_TAIL(&(q
->pkts
), pkt
, pkt_train
);
139 last_pkt
= bpacket_queue_last(q
);
140 TAILQ_INSERT_AFTER(&(q
->pkts
), last_pkt
, pkt
, pkt_train
);
143 if (q
->hwm_count
< q
->curr_count
)
144 q
->hwm_count
= q
->curr_count
;
148 * Adds a packet to the bpacket_queue.
150 * The stream passed is consumed by this function. So, the caller should
151 * not free or use the stream after
152 * invoking this function.
154 struct bpacket
*bpacket_queue_add(struct bpacket_queue
*q
, struct stream
*s
,
155 struct bpacket_attr_vec_arr
*vecarrp
)
158 struct bpacket
*last_pkt
;
161 pkt
= bpacket_alloc();
162 if (TAILQ_EMPTY(&(q
->pkts
))) {
166 memcpy(&pkt
->arr
, vecarrp
,
167 sizeof(struct bpacket_attr_vec_arr
));
169 bpacket_attr_vec_arr_reset(&pkt
->arr
);
170 bpacket_queue_add_packet(q
, pkt
);
171 bpacket_queue_sanity_check(q
);
176 * Fill in the new information into the current sentinel and create a
179 bpacket_queue_sanity_check(q
);
180 last_pkt
= bpacket_queue_last(q
);
181 assert(last_pkt
->buffer
== NULL
);
182 last_pkt
->buffer
= s
;
184 memcpy(&last_pkt
->arr
, vecarrp
,
185 sizeof(struct bpacket_attr_vec_arr
));
187 bpacket_attr_vec_arr_reset(&last_pkt
->arr
);
189 pkt
->ver
= last_pkt
->ver
;
191 bpacket_queue_add_packet(q
, pkt
);
193 bpacket_queue_sanity_check(q
);
197 struct bpacket
*bpacket_queue_first(struct bpacket_queue
*q
)
199 return (TAILQ_FIRST(&(q
->pkts
)));
202 struct bpacket
*bpacket_queue_last(struct bpacket_queue
*q
)
204 return TAILQ_LAST(&(q
->pkts
), pkt_queue
);
207 struct bpacket
*bpacket_queue_remove(struct bpacket_queue
*q
)
209 struct bpacket
*first
;
211 first
= bpacket_queue_first(q
);
213 TAILQ_REMOVE(&(q
->pkts
), first
, pkt_train
);
219 unsigned int bpacket_queue_length(struct bpacket_queue
*q
)
221 return q
->curr_count
- 1;
224 unsigned int bpacket_queue_hwm_length(struct bpacket_queue
*q
)
226 return q
->hwm_count
- 1;
229 int bpacket_queue_is_full(struct bgp
*bgp
, struct bpacket_queue
*q
)
231 if (q
->curr_count
>= bgp
->default_subgroup_pkt_queue_max
)
236 void bpacket_add_peer(struct bpacket
*pkt
, struct peer_af
*paf
)
241 LIST_INSERT_HEAD(&(pkt
->peers
), paf
, pkt_train
);
242 paf
->next_pkt_to_send
= pkt
;
246 * bpacket_queue_cleanup
248 void bpacket_queue_cleanup(struct bpacket_queue
*q
)
252 while ((pkt
= bpacket_queue_remove(q
))) {
258 * bpacket_queue_compact
260 * Delete packets that do not need to be transmitted to any peer from
263 * @return the number of packets deleted.
265 static int bpacket_queue_compact(struct bpacket_queue
*q
)
268 struct bpacket
*pkt
, *removed_pkt
;
273 pkt
= bpacket_queue_first(q
);
278 * Don't delete the sentinel.
283 if (!LIST_EMPTY(&(pkt
->peers
)))
286 removed_pkt
= bpacket_queue_remove(q
);
287 assert(pkt
== removed_pkt
);
288 bpacket_free(removed_pkt
);
293 bpacket_queue_sanity_check(q
);
297 void bpacket_queue_advance_peer(struct peer_af
*paf
)
300 struct bpacket
*old_pkt
;
302 old_pkt
= paf
->next_pkt_to_send
;
303 if (old_pkt
->buffer
== NULL
)
304 /* Already at end of list */
307 LIST_REMOVE(paf
, pkt_train
);
308 pkt
= TAILQ_NEXT(old_pkt
, pkt_train
);
309 bpacket_add_peer(pkt
, paf
);
311 if (!bpacket_queue_compact(PAF_PKTQ(paf
)))
315 * Deleted one or more packets. Check if we can now merge this
316 * peer's subgroup into another subgroup.
318 update_subgroup_check_merge(paf
->subgroup
, "advanced peer in queue");
322 * bpacket_queue_remove_peer
324 * Remove the peer from the packet queue of the subgroup it belongs
327 void bpacket_queue_remove_peer(struct peer_af
*paf
)
329 struct bpacket_queue
*q
;
336 LIST_REMOVE(paf
, pkt_train
);
337 paf
->next_pkt_to_send
= NULL
;
339 bpacket_queue_compact(q
);
342 unsigned int bpacket_queue_virtual_length(struct peer_af
*paf
)
345 struct bpacket
*last
;
346 struct bpacket_queue
*q
;
348 pkt
= paf
->next_pkt_to_send
;
349 if (!pkt
|| (pkt
->buffer
== NULL
))
350 /* Already at end of list */
354 if (TAILQ_EMPTY(&(q
->pkts
)))
357 last
= TAILQ_LAST(&(q
->pkts
), pkt_queue
);
358 if (last
->ver
>= pkt
->ver
)
359 return last
->ver
- pkt
->ver
;
361 /* sequence # rolled over */
362 return (UINT_MAX
- pkt
->ver
+ 1) + last
->ver
;
366 * Dump the bpacket queue
368 void bpacket_queue_show_vty(struct bpacket_queue
*q
, struct vty
*vty
)
373 pkt
= bpacket_queue_first(q
);
375 vty_out(vty
, " Packet %p ver %u buffer %p\n", pkt
, pkt
->ver
,
378 LIST_FOREACH (paf
, &(pkt
->peers
), pkt_train
) {
379 vty_out(vty
, " - %s\n", paf
->peer
->host
);
381 pkt
= bpacket_next(pkt
);
386 struct stream
*bpacket_reformat_for_peer(struct bpacket
*pkt
,
389 struct stream
*s
= NULL
;
390 bpacket_attr_vec
*vec
;
395 s
= stream_dup(pkt
->buffer
);
396 peer
= PAF_PEER(paf
);
398 vec
= &pkt
->arr
.entries
[BGP_ATTR_VEC_NH
];
399 if (CHECK_FLAG(vec
->flags
, BPKT_ATTRVEC_FLAGS_UPDATED
)) {
402 int route_map_sets_nh
;
403 nhlen
= stream_getc_from(s
, vec
->offset
);
404 if (peer_cap_enhe(peer
, paf
->afi
, paf
->safi
))
407 nhafi
= BGP_NEXTHOP_AFI_FROM_NHLEN(nhlen
);
409 if (nhafi
== AFI_IP
) {
410 struct in_addr v4nh
, *mod_v4nh
;
412 size_t offset_nh
= vec
->offset
+ 1;
417 BPKT_ATTRVEC_FLAGS_RMAP_IPV4_NH_CHANGED
)
420 BPKT_ATTRVEC_FLAGS_RMAP_NH_PEER_ADDRESS
));
423 case BGP_ATTR_NHLEN_IPV4
:
425 case BGP_ATTR_NHLEN_VPNV4
:
429 /* TODO: handle IPv6 nexthops */
431 EC_BGP_INVALID_NEXTHOP_LENGTH
,
432 "%s: %s: invalid MP nexthop length (AFI IP): %u",
433 __func__
, peer
->host
, nhlen
);
438 stream_get_from(&v4nh
, s
, offset_nh
, IPV4_MAX_BYTELEN
);
442 * If route-map has set the nexthop, that is always
444 * specified as peer-address, the peering address is
446 * Otherwise, if NH is unavailable from attribute, the
448 * is picked up; the "NH unavailable" case also covers
450 * and some other scenarios -- see
451 * subgroup_announce_check(). In
452 * all other cases, use the nexthop carried in the
454 * it is EBGP non-multiaccess and there is no
455 * next-hop-unchanged setting.
456 * Note: It is assumed route-map cannot set the nexthop
460 if (route_map_sets_nh
) {
463 BPKT_ATTRVEC_FLAGS_RMAP_NH_PEER_ADDRESS
)) {
464 mod_v4nh
= &peer
->nexthop
.v4
;
467 } else if (!v4nh
.s_addr
) {
468 mod_v4nh
= &peer
->nexthop
.v4
;
471 peer
->sort
== BGP_PEER_EBGP
472 && (bgp_multiaccess_check_v4(v4nh
, peer
) == 0)
475 BPKT_ATTRVEC_FLAGS_RMAP_NH_UNCHANGED
)
476 && !peer_af_flag_check(
477 peer
, paf
->afi
, paf
->safi
,
478 PEER_FLAG_NEXTHOP_UNCHANGED
)) {
479 /* NOTE: not handling case where NH has new AFI
481 mod_v4nh
= &peer
->nexthop
.v4
;
485 if (nh_modified
) /* allow for VPN RD */
486 stream_put_in_addr_at(s
, offset_nh
, mod_v4nh
);
488 if (bgp_debug_update(peer
, NULL
, NULL
, 0))
489 zlog_debug("u%" PRIu64
":s%" PRIu64
490 " %s send UPDATE w/ nexthop %s%s",
491 PAF_SUBGRP(paf
)->update_group
->id
,
492 PAF_SUBGRP(paf
)->id
, peer
->host
,
493 inet_ntoa(*mod_v4nh
),
494 (nhlen
== 12 ? " and RD" : ""));
495 } else if (nhafi
== AFI_IP6
) {
496 struct in6_addr v6nhglobal
, *mod_v6nhg
;
497 struct in6_addr v6nhlocal
, *mod_v6nhl
;
498 int gnh_modified
, lnh_modified
;
499 size_t offset_nhglobal
= vec
->offset
+ 1;
500 size_t offset_nhlocal
= vec
->offset
+ 1;
502 gnh_modified
= lnh_modified
= 0;
503 mod_v6nhg
= &v6nhglobal
;
504 mod_v6nhl
= &v6nhlocal
;
509 BPKT_ATTRVEC_FLAGS_RMAP_IPV6_GNH_CHANGED
)
512 BPKT_ATTRVEC_FLAGS_RMAP_NH_PEER_ADDRESS
));
515 * The logic here is rather similar to that for IPv4,
517 * additional work being to handle 1 or 2 nexthops.
519 * party nexthop is not propagated for EBGP right now.
522 case BGP_ATTR_NHLEN_IPV6_GLOBAL
:
524 case BGP_ATTR_NHLEN_IPV6_GLOBAL_AND_LL
:
525 offset_nhlocal
+= IPV6_MAX_BYTELEN
;
527 case BGP_ATTR_NHLEN_VPNV6_GLOBAL
:
528 offset_nhglobal
+= 8;
530 case BGP_ATTR_NHLEN_VPNV6_GLOBAL_AND_LL
:
531 offset_nhglobal
+= 8;
532 offset_nhlocal
+= 8 * 2 + IPV6_MAX_BYTELEN
;
535 /* TODO: handle IPv4 nexthops */
537 EC_BGP_INVALID_NEXTHOP_LENGTH
,
538 "%s: %s: invalid MP nexthop length (AFI IP6): %u",
539 __func__
, peer
->host
, nhlen
);
544 stream_get_from(&v6nhglobal
, s
, offset_nhglobal
,
546 if (route_map_sets_nh
) {
549 BPKT_ATTRVEC_FLAGS_RMAP_NH_PEER_ADDRESS
)) {
550 mod_v6nhg
= &peer
->nexthop
.v6_global
;
553 } else if (IN6_IS_ADDR_UNSPECIFIED(&v6nhglobal
)) {
554 mod_v6nhg
= &peer
->nexthop
.v6_global
;
557 peer
->sort
== BGP_PEER_EBGP
560 BPKT_ATTRVEC_FLAGS_RMAP_NH_UNCHANGED
)
561 && !peer_af_flag_check(
562 peer
, nhafi
, paf
->safi
,
563 PEER_FLAG_NEXTHOP_UNCHANGED
)) {
564 /* NOTE: not handling case where NH has new AFI
566 mod_v6nhg
= &peer
->nexthop
.v6_global
;
571 if (nhlen
== BGP_ATTR_NHLEN_IPV6_GLOBAL_AND_LL
572 || nhlen
== BGP_ATTR_NHLEN_VPNV6_GLOBAL_AND_LL
) {
573 stream_get_from(&v6nhlocal
, s
, offset_nhlocal
,
575 if (IN6_IS_ADDR_UNSPECIFIED(&v6nhlocal
)) {
576 mod_v6nhl
= &peer
->nexthop
.v6_local
;
582 stream_put_in6_addr_at(s
, offset_nhglobal
,
585 stream_put_in6_addr_at(s
, offset_nhlocal
,
588 if (bgp_debug_update(peer
, NULL
, NULL
, 0)) {
589 if (nhlen
== 32 || nhlen
== 48)
591 "u%" PRIu64
":s%" PRIu64
592 " %s send UPDATE w/ mp_nexthops %s, %s%s",
595 PAF_SUBGRP(paf
)->id
, peer
->host
,
596 inet_ntop(AF_INET6
, mod_v6nhg
,
598 inet_ntop(AF_INET6
, mod_v6nhl
,
600 (nhlen
== 48 ? " and RD" : ""));
603 "u%" PRIu64
":s%" PRIu64
604 " %s send UPDATE w/ mp_nexthop %s%s",
607 PAF_SUBGRP(paf
)->id
, peer
->host
,
608 inet_ntop(AF_INET6
, mod_v6nhg
,
610 (nhlen
== 24 ? " and RD" : ""));
612 } else if (paf
->afi
== AFI_L2VPN
) {
613 struct in_addr v4nh
, *mod_v4nh
;
616 stream_get_from(&v4nh
, s
, vec
->offset
+ 1, 4);
619 /* No route-map changes allowed for EVPN nexthops. */
621 mod_v4nh
= &peer
->nexthop
.v4
;
626 stream_put_in_addr_at(s
, vec
->offset
+ 1,
629 if (bgp_debug_update(peer
, NULL
, NULL
, 0))
630 zlog_debug("u%" PRIu64
":s%" PRIu64
631 " %s send UPDATE w/ nexthop %s",
632 PAF_SUBGRP(paf
)->update_group
->id
,
633 PAF_SUBGRP(paf
)->id
, peer
->host
,
634 inet_ntoa(*mod_v4nh
));
642 * Update the vecarr offsets to go beyond 'pos' bytes, i.e. add 'pos'
645 static void bpacket_attr_vec_arr_update(struct bpacket_attr_vec_arr
*vecarr
,
653 for (i
= 0; i
< BGP_ATTR_VEC_MAX
; i
++)
654 vecarr
->entries
[i
].offset
+= pos
;
658 * Return if there are packets to build for this subgroup.
660 int subgroup_packets_to_build(struct update_subgroup
*subgrp
)
662 struct bgp_advertise
*adv
;
667 adv
= BGP_ADV_FIFO_HEAD(&subgrp
->sync
->withdraw
);
671 adv
= BGP_ADV_FIFO_HEAD(&subgrp
->sync
->update
);
678 /* Make BGP update packet. */
679 struct bpacket
*subgroup_update_packet(struct update_subgroup
*subgrp
)
681 struct bpacket_attr_vec_arr vecarr
;
685 struct stream
*snlri
;
686 struct stream
*packet
;
687 struct bgp_adj_out
*adj
;
688 struct bgp_advertise
*adv
;
689 struct bgp_node
*rn
= NULL
;
690 struct bgp_info
*binfo
= NULL
;
691 bgp_size_t total_attr_len
= 0;
692 unsigned long attrlen_pos
= 0;
693 size_t mpattrlen_pos
= 0;
694 size_t mpattr_pos
= 0;
697 int space_remaining
= 0;
698 int space_needed
= 0;
699 char send_attr_str
[BUFSIZ
];
700 int send_attr_printed
= 0;
702 int addpath_encode
= 0;
703 int addpath_overhead
= 0;
704 uint32_t addpath_tx_id
= 0;
705 struct prefix_rd
*prd
= NULL
;
706 mpls_label_t label
= MPLS_INVALID_LABEL
, *label_pnt
= NULL
;
707 uint32_t num_labels
= 0;
712 if (bpacket_queue_is_full(SUBGRP_INST(subgrp
), SUBGRP_PKTQ(subgrp
)))
715 peer
= SUBGRP_PEER(subgrp
);
716 afi
= SUBGRP_AFI(subgrp
);
717 safi
= SUBGRP_SAFI(subgrp
);
720 snlri
= subgrp
->scratch
;
723 bpacket_attr_vec_arr_reset(&vecarr
);
725 addpath_encode
= bgp_addpath_encode_tx(peer
, afi
, safi
);
726 addpath_overhead
= addpath_encode
? BGP_ADDPATH_ID_LEN
: 0;
728 adv
= BGP_ADV_FIFO_HEAD(&subgrp
->sync
->update
);
733 addpath_tx_id
= adj
->addpath_tx_id
;
736 space_remaining
= STREAM_CONCAT_REMAIN(s
, snlri
, STREAM_SIZE(s
))
737 - BGP_MAX_PACKET_SIZE_OVERFLOW
;
739 BGP_NLRI_LENGTH
+ addpath_overhead
740 + bgp_packet_mpattr_prefix_size(afi
, safi
, &rn
->p
);
742 /* When remaining space can't include NLRI and it's length. */
743 if (space_remaining
< space_needed
)
746 /* If packet is empty, set attribute. */
747 if (stream_empty(s
)) {
748 struct peer
*from
= NULL
;
753 /* 1: Write the BGP message header - 16 bytes marker, 2
755 * one byte message type.
757 bgp_packet_set_marker(s
, BGP_MSG_UPDATE
);
759 /* 2: withdrawn routes length */
762 /* 3: total attributes length - attrlen_pos stores the
764 attrlen_pos
= stream_get_endp(s
);
767 /* 4: if there is MP_REACH_NLRI attribute, that should
769 * attribute, according to
770 * draft-ietf-idr-error-handling. Save the
773 mpattr_pos
= stream_get_endp(s
);
775 /* 5: Encode all the attributes, except MP_REACH_NLRI
777 total_attr_len
= bgp_packet_attribute(
778 NULL
, peer
, s
, adv
->baa
->attr
, &vecarr
, NULL
,
779 afi
, safi
, from
, NULL
, NULL
, 0, 0, 0);
782 STREAM_CONCAT_REMAIN(s
, snlri
, STREAM_SIZE(s
))
783 - BGP_MAX_PACKET_SIZE_OVERFLOW
;
784 space_needed
= BGP_NLRI_LENGTH
+ addpath_overhead
785 + bgp_packet_mpattr_prefix_size(
788 /* If the attributes alone do not leave any room for
791 if (space_remaining
< space_needed
) {
793 EC_BGP_UPDGRP_ATTR_LEN
,
794 "u%" PRIu64
":s%" PRIu64
795 " attributes too long, cannot send UPDATE",
796 subgrp
->update_group
->id
, subgrp
->id
);
798 /* Flush the FIFO update queue */
800 adv
= bgp_advertise_clean_subgroup(
805 if (BGP_DEBUG(update
, UPDATE_OUT
)
806 || BGP_DEBUG(update
, UPDATE_PREFIX
)) {
807 memset(send_attr_str
, 0, BUFSIZ
);
808 send_attr_printed
= 0;
809 bgp_dump_attr(adv
->baa
->attr
, send_attr_str
,
814 if ((afi
== AFI_IP
&& safi
== SAFI_UNICAST
)
815 && !peer_cap_enhe(peer
, afi
, safi
))
816 stream_put_prefix_addpath(s
, &rn
->p
, addpath_encode
,
819 /* Encode the prefix in MP_REACH_NLRI attribute */
821 prd
= (struct prefix_rd
*)&rn
->prn
->p
;
823 if (safi
== SAFI_LABELED_UNICAST
) {
824 label
= bgp_adv_label(rn
, binfo
, peer
, afi
,
828 } else if (binfo
&& binfo
->extra
) {
829 label_pnt
= &binfo
->extra
->label
[0];
830 num_labels
= binfo
->extra
->num_labels
;
833 if (stream_empty(snlri
))
834 mpattrlen_pos
= bgp_packet_mpattr_start(
835 snlri
, peer
, afi
, safi
, &vecarr
,
838 bgp_packet_mpattr_prefix(snlri
, afi
, safi
, &rn
->p
, prd
,
839 label_pnt
, num_labels
,
840 addpath_encode
, addpath_tx_id
,
846 if (bgp_debug_update(NULL
, &rn
->p
, subgrp
->update_group
, 0)) {
847 char pfx_buf
[BGP_PRD_PATH_STRLEN
];
849 if (!send_attr_printed
) {
850 zlog_debug("u%" PRIu64
":s%" PRIu64
851 " send UPDATE w/ attr: %s",
852 subgrp
->update_group
->id
, subgrp
->id
,
854 if (!stream_empty(snlri
)) {
856 iana_safi_t pkt_safi
;
858 pkt_afi
= afi_int2iana(afi
);
859 pkt_safi
= safi_int2iana(safi
);
861 "u%" PRIu64
":s%" PRIu64
862 " send MP_REACH for afi/safi %d/%d",
863 subgrp
->update_group
->id
,
864 subgrp
->id
, pkt_afi
, pkt_safi
);
867 send_attr_printed
= 1;
870 bgp_debug_rdpfxpath2str(afi
, safi
, prd
, &rn
->p
,
871 label_pnt
, num_labels
,
872 addpath_encode
, addpath_tx_id
,
873 pfx_buf
, sizeof(pfx_buf
));
874 zlog_debug("u%" PRIu64
":s%" PRIu64
" send UPDATE %s",
875 subgrp
->update_group
->id
, subgrp
->id
,
879 /* Synchnorize attribute. */
881 bgp_attr_unintern(&adj
->attr
);
885 adj
->attr
= bgp_attr_intern(adv
->baa
->attr
);
887 adv
= bgp_advertise_clean_subgroup(subgrp
, adj
);
890 if (!stream_empty(s
)) {
891 if (!stream_empty(snlri
)) {
892 bgp_packet_mpattr_end(snlri
, mpattrlen_pos
);
893 total_attr_len
+= stream_get_endp(snlri
);
896 /* set the total attribute length correctly */
897 stream_putw_at(s
, attrlen_pos
, total_attr_len
);
899 if (!stream_empty(snlri
)) {
900 packet
= stream_dupcat(s
, snlri
, mpattr_pos
);
901 bpacket_attr_vec_arr_update(&vecarr
, mpattr_pos
);
903 packet
= stream_dup(s
);
904 bgp_packet_set_size(packet
);
905 if (bgp_debug_update(NULL
, NULL
, subgrp
->update_group
, 0))
906 zlog_debug("u%" PRIu64
":s%" PRIu64
907 " send UPDATE len %zd numpfx %d",
908 subgrp
->update_group
->id
, subgrp
->id
,
909 (stream_get_endp(packet
)
910 - stream_get_getp(packet
)),
912 pkt
= bpacket_queue_add(SUBGRP_PKTQ(subgrp
), packet
, &vecarr
);
920 /* Make BGP withdraw packet. */
922 16-octet marker | 2-octet length | 1-octet type |
923 2-octet withdrawn route length | withdrawn prefixes | 2-octet attrlen (=0)
925 /* For other afi/safis:
926 16-octet marker | 2-octet length | 1-octet type |
927 2-octet withdrawn route length (=0) | 2-octet attrlen |
928 mp_unreach attr type | attr len | afi | safi | withdrawn prefixes
930 struct bpacket
*subgroup_withdraw_packet(struct update_subgroup
*subgrp
)
934 struct bgp_adj_out
*adj
;
935 struct bgp_advertise
*adv
;
938 bgp_size_t unfeasible_len
;
939 bgp_size_t total_attr_len
;
941 size_t attrlen_pos
= 0;
942 size_t mplen_pos
= 0;
943 uint8_t first_time
= 1;
946 int space_remaining
= 0;
947 int space_needed
= 0;
949 int addpath_encode
= 0;
950 int addpath_overhead
= 0;
951 uint32_t addpath_tx_id
= 0;
952 struct prefix_rd
*prd
= NULL
;
958 if (bpacket_queue_is_full(SUBGRP_INST(subgrp
), SUBGRP_PKTQ(subgrp
)))
961 peer
= SUBGRP_PEER(subgrp
);
962 afi
= SUBGRP_AFI(subgrp
);
963 safi
= SUBGRP_SAFI(subgrp
);
966 addpath_encode
= bgp_addpath_encode_tx(peer
, afi
, safi
);
967 addpath_overhead
= addpath_encode
? BGP_ADDPATH_ID_LEN
: 0;
969 while ((adv
= BGP_ADV_FIFO_HEAD(&subgrp
->sync
->withdraw
)) != NULL
) {
973 addpath_tx_id
= adj
->addpath_tx_id
;
976 STREAM_WRITEABLE(s
) - BGP_MAX_PACKET_SIZE_OVERFLOW
;
978 BGP_NLRI_LENGTH
+ addpath_overhead
+ BGP_TOTAL_ATTR_LEN
979 + bgp_packet_mpattr_prefix_size(afi
, safi
, &rn
->p
);
981 if (space_remaining
< space_needed
)
984 if (stream_empty(s
)) {
985 bgp_packet_set_marker(s
, BGP_MSG_UPDATE
);
986 stream_putw(s
, 0); /* unfeasible routes length */
990 if (afi
== AFI_IP
&& safi
== SAFI_UNICAST
991 && !peer_cap_enhe(peer
, afi
, safi
))
992 stream_put_prefix_addpath(s
, &rn
->p
, addpath_encode
,
996 prd
= (struct prefix_rd
*)&rn
->prn
->p
;
998 /* If first time, format the MP_UNREACH header */
1001 iana_safi_t pkt_safi
;
1003 pkt_afi
= afi_int2iana(afi
);
1004 pkt_safi
= safi_int2iana(safi
);
1006 attrlen_pos
= stream_get_endp(s
);
1007 /* total attr length = 0 for now. reevaluate
1010 mp_start
= stream_get_endp(s
);
1011 mplen_pos
= bgp_packet_mpunreach_start(s
, afi
,
1013 if (bgp_debug_update(NULL
, NULL
,
1014 subgrp
->update_group
, 0))
1016 "u%" PRIu64
":s%" PRIu64
1017 " send MP_UNREACH for afi/safi %d/%d",
1018 subgrp
->update_group
->id
,
1019 subgrp
->id
, pkt_afi
, pkt_safi
);
1022 bgp_packet_mpunreach_prefix(s
, &rn
->p
, afi
, safi
, prd
,
1023 NULL
, 0, addpath_encode
,
1024 addpath_tx_id
, NULL
);
1029 if (bgp_debug_update(NULL
, &rn
->p
, subgrp
->update_group
, 0)) {
1030 char pfx_buf
[BGP_PRD_PATH_STRLEN
];
1032 bgp_debug_rdpfxpath2str(afi
, safi
, prd
, &rn
->p
, NULL
, 0,
1033 addpath_encode
, addpath_tx_id
,
1034 pfx_buf
, sizeof(pfx_buf
));
1035 zlog_debug("u%" PRIu64
":s%" PRIu64
1036 " send UPDATE %s -- unreachable",
1037 subgrp
->update_group
->id
, subgrp
->id
,
1043 bgp_adj_out_remove_subgroup(rn
, adj
, subgrp
);
1044 bgp_unlock_node(rn
);
1047 if (!stream_empty(s
)) {
1048 if (afi
== AFI_IP
&& safi
== SAFI_UNICAST
1049 && !peer_cap_enhe(peer
, afi
, safi
)) {
1050 unfeasible_len
= stream_get_endp(s
) - BGP_HEADER_SIZE
1051 - BGP_UNFEASIBLE_LEN
;
1052 stream_putw_at(s
, BGP_HEADER_SIZE
, unfeasible_len
);
1055 /* Set the mp_unreach attr's length */
1056 bgp_packet_mpunreach_end(s
, mplen_pos
);
1058 /* Set total path attribute length. */
1059 total_attr_len
= stream_get_endp(s
) - mp_start
;
1060 stream_putw_at(s
, attrlen_pos
, total_attr_len
);
1062 bgp_packet_set_size(s
);
1063 if (bgp_debug_update(NULL
, NULL
, subgrp
->update_group
, 0))
1064 zlog_debug("u%" PRIu64
":s%" PRIu64
1065 " send UPDATE (withdraw) len %zd numpfx %d",
1066 subgrp
->update_group
->id
, subgrp
->id
,
1067 (stream_get_endp(s
) - stream_get_getp(s
)),
1069 pkt
= bpacket_queue_add(SUBGRP_PKTQ(subgrp
), stream_dup(s
),
1078 void subgroup_default_update_packet(struct update_subgroup
*subgrp
,
1079 struct attr
*attr
, struct peer
*from
)
1085 bgp_size_t total_attr_len
;
1088 struct bpacket_attr_vec_arr vecarr
;
1089 int addpath_encode
= 0;
1091 if (DISABLE_BGP_ANNOUNCE
)
1097 peer
= SUBGRP_PEER(subgrp
);
1098 afi
= SUBGRP_AFI(subgrp
);
1099 safi
= SUBGRP_SAFI(subgrp
);
1100 bpacket_attr_vec_arr_reset(&vecarr
);
1101 addpath_encode
= bgp_addpath_encode_tx(peer
, afi
, safi
);
1103 memset(&p
, 0, sizeof(p
));
1104 p
.family
= afi2family(afi
);
1107 /* Logging the attribute. */
1108 if (bgp_debug_update(NULL
, &p
, subgrp
->update_group
, 0)) {
1109 char attrstr
[BUFSIZ
];
1110 char buf
[PREFIX_STRLEN
];
1111 /* ' with addpath ID ' 17
1112 * max strlen of uint32 + 10
1113 * +/- (just in case) + 1
1114 * null terminator + 1
1115 * ============================ 29 */
1120 bgp_dump_attr(attr
, attrstr
, BUFSIZ
);
1123 snprintf(tx_id_buf
, sizeof(tx_id_buf
),
1124 " with addpath ID %u",
1125 BGP_ADDPATH_TX_ID_FOR_DEFAULT_ORIGINATE
);
1127 zlog_debug("u%" PRIu64
":s%" PRIu64
" send UPDATE %s%s %s",
1128 (SUBGRP_UPDGRP(subgrp
))->id
, subgrp
->id
,
1129 prefix2str(&p
, buf
, sizeof(buf
)), tx_id_buf
,
1133 s
= stream_new(BGP_MAX_PACKET_SIZE
);
1135 /* Make BGP update packet. */
1136 bgp_packet_set_marker(s
, BGP_MSG_UPDATE
);
1138 /* Unfeasible Routes Length. */
1141 /* Make place for total attribute length. */
1142 pos
= stream_get_endp(s
);
1144 total_attr_len
= bgp_packet_attribute(
1145 NULL
, peer
, s
, attr
, &vecarr
, &p
, afi
, safi
, from
, NULL
, NULL
,
1146 0, addpath_encode
, BGP_ADDPATH_TX_ID_FOR_DEFAULT_ORIGINATE
);
1148 /* Set Total Path Attribute Length. */
1149 stream_putw_at(s
, pos
, total_attr_len
);
1152 if (p
.family
== AF_INET
&& safi
== SAFI_UNICAST
1153 && !peer_cap_enhe(peer
, afi
, safi
))
1154 stream_put_prefix_addpath(
1155 s
, &p
, addpath_encode
,
1156 BGP_ADDPATH_TX_ID_FOR_DEFAULT_ORIGINATE
);
1159 bgp_packet_set_size(s
);
1161 (void)bpacket_queue_add(SUBGRP_PKTQ(subgrp
), s
, &vecarr
);
1162 subgroup_trigger_write(subgrp
);
1165 void subgroup_default_withdraw_packet(struct update_subgroup
*subgrp
)
1170 unsigned long attrlen_pos
= 0;
1172 bgp_size_t unfeasible_len
;
1173 bgp_size_t total_attr_len
= 0;
1174 size_t mp_start
= 0;
1175 size_t mplen_pos
= 0;
1178 int addpath_encode
= 0;
1180 if (DISABLE_BGP_ANNOUNCE
)
1183 peer
= SUBGRP_PEER(subgrp
);
1184 afi
= SUBGRP_AFI(subgrp
);
1185 safi
= SUBGRP_SAFI(subgrp
);
1186 addpath_encode
= bgp_addpath_encode_tx(peer
, afi
, safi
);
1188 memset(&p
, 0, sizeof(p
));
1189 p
.family
= afi2family(afi
);
1192 if (bgp_debug_update(NULL
, &p
, subgrp
->update_group
, 0)) {
1193 char buf
[PREFIX_STRLEN
];
1194 /* ' with addpath ID ' 17
1195 * max strlen of uint32 + 10
1196 * +/- (just in case) + 1
1197 * null terminator + 1
1198 * ============================ 29 */
1202 snprintf(tx_id_buf
, sizeof(tx_id_buf
),
1203 " with addpath ID %u",
1204 BGP_ADDPATH_TX_ID_FOR_DEFAULT_ORIGINATE
);
1206 zlog_debug("u%" PRIu64
":s%" PRIu64
1207 " send UPDATE %s%s -- unreachable",
1208 (SUBGRP_UPDGRP(subgrp
))->id
, subgrp
->id
,
1209 prefix2str(&p
, buf
, sizeof(buf
)), tx_id_buf
);
1212 s
= stream_new(BGP_MAX_PACKET_SIZE
);
1214 /* Make BGP update packet. */
1215 bgp_packet_set_marker(s
, BGP_MSG_UPDATE
);
1217 /* Unfeasible Routes Length. */;
1218 cp
= stream_get_endp(s
);
1221 /* Withdrawn Routes. */
1222 if (p
.family
== AF_INET
&& safi
== SAFI_UNICAST
1223 && !peer_cap_enhe(peer
, afi
, safi
)) {
1224 stream_put_prefix_addpath(
1225 s
, &p
, addpath_encode
,
1226 BGP_ADDPATH_TX_ID_FOR_DEFAULT_ORIGINATE
);
1228 unfeasible_len
= stream_get_endp(s
) - cp
- 2;
1230 /* Set unfeasible len. */
1231 stream_putw_at(s
, cp
, unfeasible_len
);
1233 /* Set total path attribute length. */
1236 attrlen_pos
= stream_get_endp(s
);
1238 mp_start
= stream_get_endp(s
);
1239 mplen_pos
= bgp_packet_mpunreach_start(s
, afi
, safi
);
1240 bgp_packet_mpunreach_prefix(
1241 s
, &p
, afi
, safi
, NULL
, NULL
, 0, addpath_encode
,
1242 BGP_ADDPATH_TX_ID_FOR_DEFAULT_ORIGINATE
, NULL
);
1244 /* Set the mp_unreach attr's length */
1245 bgp_packet_mpunreach_end(s
, mplen_pos
);
1247 /* Set total path attribute length. */
1248 total_attr_len
= stream_get_endp(s
) - mp_start
;
1249 stream_putw_at(s
, attrlen_pos
, total_attr_len
);
1252 bgp_packet_set_size(s
);
1254 (void)bpacket_queue_add(SUBGRP_PKTQ(subgrp
), s
, NULL
);
1255 subgroup_trigger_write(subgrp
);
1259 bpacket_vec_arr_inherit_attr_flags(struct bpacket_attr_vec_arr
*vecarr
,
1260 bpacket_attr_vec_type type
,
1263 if (CHECK_FLAG(attr
->rmap_change_flags
,
1264 BATTR_RMAP_NEXTHOP_PEER_ADDRESS
))
1265 SET_FLAG(vecarr
->entries
[BGP_ATTR_VEC_NH
].flags
,
1266 BPKT_ATTRVEC_FLAGS_RMAP_NH_PEER_ADDRESS
);
1268 if (CHECK_FLAG(attr
->rmap_change_flags
, BATTR_REFLECTED
))
1269 SET_FLAG(vecarr
->entries
[BGP_ATTR_VEC_NH
].flags
,
1270 BPKT_ATTRVEC_FLAGS_REFLECTED
);
1272 if (CHECK_FLAG(attr
->rmap_change_flags
, BATTR_RMAP_NEXTHOP_UNCHANGED
))
1273 SET_FLAG(vecarr
->entries
[BGP_ATTR_VEC_NH
].flags
,
1274 BPKT_ATTRVEC_FLAGS_RMAP_NH_UNCHANGED
);
1276 if (CHECK_FLAG(attr
->rmap_change_flags
, BATTR_RMAP_IPV4_NHOP_CHANGED
))
1277 SET_FLAG(vecarr
->entries
[BGP_ATTR_VEC_NH
].flags
,
1278 BPKT_ATTRVEC_FLAGS_RMAP_IPV4_NH_CHANGED
);
1280 if (CHECK_FLAG(attr
->rmap_change_flags
,
1281 BATTR_RMAP_IPV6_GLOBAL_NHOP_CHANGED
))
1282 SET_FLAG(vecarr
->entries
[BGP_ATTR_VEC_NH
].flags
,
1283 BPKT_ATTRVEC_FLAGS_RMAP_IPV6_GNH_CHANGED
);
1285 if (CHECK_FLAG(attr
->rmap_change_flags
,
1286 BATTR_RMAP_IPV6_LL_NHOP_CHANGED
))
1287 SET_FLAG(vecarr
->entries
[BGP_ATTR_VEC_NH
].flags
,
1288 BPKT_ATTRVEC_FLAGS_RMAP_IPV6_LNH_CHANGED
);
1291 /* Reset the Attributes vector array. The vector array is used to override
1292 * certain output parameters in the packet for a particular peer
1294 void bpacket_attr_vec_arr_reset(struct bpacket_attr_vec_arr
*vecarr
)
1302 while (i
< BGP_ATTR_VEC_MAX
) {
1303 vecarr
->entries
[i
].flags
= 0;
1304 vecarr
->entries
[i
].offset
= 0;
1309 /* Setup a particular node entry in the vecarr */
1310 void bpacket_attr_vec_arr_set_vec(struct bpacket_attr_vec_arr
*vecarr
,
1311 bpacket_attr_vec_type type
, struct stream
*s
,
1316 assert(type
< BGP_ATTR_VEC_MAX
);
1318 SET_FLAG(vecarr
->entries
[type
].flags
, BPKT_ATTRVEC_FLAGS_UPDATED
);
1319 vecarr
->entries
[type
].offset
= stream_get_endp(s
);
1321 bpacket_vec_arr_inherit_attr_flags(vecarr
, type
, attr
);