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_fsm.h"
50 #include "bgpd/bgp_route.h"
51 #include "bgpd/bgp_packet.h"
52 #include "bgpd/bgp_advertise.h"
53 #include "bgpd/bgp_updgrp.h"
54 #include "bgpd/bgp_nexthop.h"
55 #include "bgpd/bgp_nht.h"
56 #include "bgpd/bgp_mplsvpn.h"
57 #include "bgpd/bgp_label.h"
66 struct bpacket
*bpacket_alloc()
70 pkt
= (struct bpacket
*)XCALLOC(MTYPE_BGP_PACKET
,
71 sizeof(struct bpacket
));
76 void bpacket_free(struct bpacket
*pkt
)
79 stream_free(pkt
->buffer
);
81 XFREE(MTYPE_BGP_PACKET
, pkt
);
84 void bpacket_queue_init(struct bpacket_queue
*q
)
86 TAILQ_INIT(&(q
->pkts
));
90 * bpacket_queue_sanity_check
92 void bpacket_queue_sanity_check(struct bpacket_queue
__attribute__((__unused__
))
98 pkt
= bpacket_queue_last (q
);
100 assert (!pkt
->buffer
);
103 * Make sure the count of packets is correct.
107 pkt
= bpacket_queue_first (q
);
112 if (num_pkts
> q
->curr_count
)
115 pkt
= TAILQ_NEXT (pkt
, pkt_train
);
118 assert (num_pkts
== q
->curr_count
);
123 * bpacket_queue_add_packet
125 * Internal function of bpacket_queue - and adds a
126 * packet entry to the end of the list.
128 * Users of bpacket_queue should use bpacket_queue_add instead.
130 static void bpacket_queue_add_packet(struct bpacket_queue
*q
,
133 struct bpacket
*last_pkt
;
135 if (TAILQ_EMPTY(&(q
->pkts
)))
136 TAILQ_INSERT_TAIL(&(q
->pkts
), pkt
, pkt_train
);
138 last_pkt
= bpacket_queue_last(q
);
139 TAILQ_INSERT_AFTER(&(q
->pkts
), last_pkt
, pkt
, pkt_train
);
142 if (q
->hwm_count
< q
->curr_count
)
143 q
->hwm_count
= q
->curr_count
;
147 * Adds a packet to the bpacket_queue.
149 * The stream passed is consumed by this function. So, the caller should
150 * not free or use the stream after
151 * invoking this function.
153 struct bpacket
*bpacket_queue_add(struct bpacket_queue
*q
, struct stream
*s
,
154 struct bpacket_attr_vec_arr
*vecarrp
)
157 struct bpacket
*last_pkt
;
160 pkt
= bpacket_alloc();
161 if (TAILQ_EMPTY(&(q
->pkts
))) {
165 memcpy(&pkt
->arr
, vecarrp
,
166 sizeof(struct bpacket_attr_vec_arr
));
168 bpacket_attr_vec_arr_reset(&pkt
->arr
);
169 bpacket_queue_add_packet(q
, pkt
);
170 bpacket_queue_sanity_check(q
);
175 * Fill in the new information into the current sentinel and create a
178 bpacket_queue_sanity_check(q
);
179 last_pkt
= bpacket_queue_last(q
);
180 assert(last_pkt
->buffer
== NULL
);
181 last_pkt
->buffer
= s
;
183 memcpy(&last_pkt
->arr
, vecarrp
,
184 sizeof(struct bpacket_attr_vec_arr
));
186 bpacket_attr_vec_arr_reset(&last_pkt
->arr
);
188 pkt
->ver
= last_pkt
->ver
;
190 bpacket_queue_add_packet(q
, pkt
);
192 bpacket_queue_sanity_check(q
);
196 struct bpacket
*bpacket_queue_first(struct bpacket_queue
*q
)
198 return (TAILQ_FIRST(&(q
->pkts
)));
201 struct bpacket
*bpacket_queue_last(struct bpacket_queue
*q
)
203 return TAILQ_LAST(&(q
->pkts
), pkt_queue
);
206 struct bpacket
*bpacket_queue_remove(struct bpacket_queue
*q
)
208 struct bpacket
*first
;
210 first
= bpacket_queue_first(q
);
212 TAILQ_REMOVE(&(q
->pkts
), first
, pkt_train
);
218 unsigned int bpacket_queue_length(struct bpacket_queue
*q
)
220 return q
->curr_count
- 1;
223 unsigned int bpacket_queue_hwm_length(struct bpacket_queue
*q
)
225 return q
->hwm_count
- 1;
228 int bpacket_queue_is_full(struct bgp
*bgp
, struct bpacket_queue
*q
)
230 if (q
->curr_count
>= bgp
->default_subgroup_pkt_queue_max
)
235 void bpacket_add_peer(struct bpacket
*pkt
, struct peer_af
*paf
)
240 LIST_INSERT_HEAD(&(pkt
->peers
), paf
, pkt_train
);
241 paf
->next_pkt_to_send
= pkt
;
245 * bpacket_queue_cleanup
247 void bpacket_queue_cleanup(struct bpacket_queue
*q
)
251 while ((pkt
= bpacket_queue_remove(q
))) {
257 * bpacket_queue_compact
259 * Delete packets that do not need to be transmitted to any peer from
262 * @return the number of packets deleted.
264 static int bpacket_queue_compact(struct bpacket_queue
*q
)
267 struct bpacket
*pkt
, *removed_pkt
;
272 pkt
= bpacket_queue_first(q
);
277 * Don't delete the sentinel.
282 if (!LIST_EMPTY(&(pkt
->peers
)))
285 removed_pkt
= bpacket_queue_remove(q
);
286 assert(pkt
== removed_pkt
);
287 bpacket_free(removed_pkt
);
292 bpacket_queue_sanity_check(q
);
296 void bpacket_queue_advance_peer(struct peer_af
*paf
)
299 struct bpacket
*old_pkt
;
301 old_pkt
= paf
->next_pkt_to_send
;
302 if (old_pkt
->buffer
== NULL
)
303 /* Already at end of list */
306 LIST_REMOVE(paf
, pkt_train
);
307 pkt
= TAILQ_NEXT(old_pkt
, pkt_train
);
308 bpacket_add_peer(pkt
, paf
);
310 if (!bpacket_queue_compact(PAF_PKTQ(paf
)))
314 * Deleted one or more packets. Check if we can now merge this
315 * peer's subgroup into another subgroup.
317 update_subgroup_check_merge(paf
->subgroup
, "advanced peer in queue");
321 * bpacket_queue_remove_peer
323 * Remove the peer from the packet queue of the subgroup it belongs
326 void bpacket_queue_remove_peer(struct peer_af
*paf
)
328 struct bpacket_queue
*q
;
335 LIST_REMOVE(paf
, pkt_train
);
336 paf
->next_pkt_to_send
= NULL
;
338 bpacket_queue_compact(q
);
341 unsigned int bpacket_queue_virtual_length(struct peer_af
*paf
)
344 struct bpacket
*last
;
345 struct bpacket_queue
*q
;
347 pkt
= paf
->next_pkt_to_send
;
348 if (!pkt
|| (pkt
->buffer
== NULL
))
349 /* Already at end of list */
353 if (TAILQ_EMPTY(&(q
->pkts
)))
356 last
= TAILQ_LAST(&(q
->pkts
), pkt_queue
);
357 if (last
->ver
>= pkt
->ver
)
358 return last
->ver
- pkt
->ver
;
360 /* sequence # rolled over */
361 return (UINT_MAX
- pkt
->ver
+ 1) + last
->ver
;
365 * Dump the bpacket queue
367 void bpacket_queue_show_vty(struct bpacket_queue
*q
, struct vty
*vty
)
372 pkt
= bpacket_queue_first(q
);
374 vty_out(vty
, " Packet %p ver %u buffer %p\n", pkt
, pkt
->ver
,
377 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
)) {
401 afi_t nhafi
= AFI_MAX
; /* NH AFI is based on nhlen! */
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 "%s: %s: invalid MP nexthop length (AFI IP): %u",
432 __func__
, peer
->host
, nhlen
);
437 stream_get_from(&v4nh
, s
, offset_nh
, IPV4_MAX_BYTELEN
);
441 * If route-map has set the nexthop, that is always
443 * specified as peer-address, the peering address is
445 * Otherwise, if NH is unavailable from attribute, the
447 * is picked up; the "NH unavailable" case also covers
449 * and some other scenarios -- see
450 * subgroup_announce_check(). In
451 * all other cases, use the nexthop carried in the
453 * it is EBGP non-multiaccess and there is no
454 * next-hop-unchanged setting.
455 * Note: It is assumed route-map cannot set the nexthop
459 if (route_map_sets_nh
) {
462 BPKT_ATTRVEC_FLAGS_RMAP_NH_PEER_ADDRESS
)) {
463 mod_v4nh
= &peer
->nexthop
.v4
;
466 } else if (!v4nh
.s_addr
) {
467 mod_v4nh
= &peer
->nexthop
.v4
;
470 peer
->sort
== BGP_PEER_EBGP
471 && paf
->safi
!= SAFI_EVPN
472 && (bgp_multiaccess_check_v4(v4nh
, peer
) == 0)
475 BPKT_ATTRVEC_FLAGS_RMAP_NH_UNCHANGED
)
476 && !peer_af_flag_check(
477 peer
, nhafi
, 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 "%s: %s: invalid MP nexthop length (AFI IP6): %u",
538 __func__
, peer
->host
, nhlen
);
543 stream_get_from(&v6nhglobal
, s
, offset_nhglobal
,
545 if (route_map_sets_nh
) {
548 BPKT_ATTRVEC_FLAGS_RMAP_NH_PEER_ADDRESS
)) {
549 mod_v6nhg
= &peer
->nexthop
.v6_global
;
552 } else if (IN6_IS_ADDR_UNSPECIFIED(&v6nhglobal
)) {
553 mod_v6nhg
= &peer
->nexthop
.v6_global
;
556 peer
->sort
== BGP_PEER_EBGP
559 BPKT_ATTRVEC_FLAGS_RMAP_NH_UNCHANGED
)
560 && !peer_af_flag_check(
561 peer
, nhafi
, paf
->safi
,
562 PEER_FLAG_NEXTHOP_UNCHANGED
)) {
563 /* NOTE: not handling case where NH has new AFI
565 mod_v6nhg
= &peer
->nexthop
.v6_global
;
570 if (nhlen
== BGP_ATTR_NHLEN_IPV6_GLOBAL_AND_LL
571 || nhlen
== BGP_ATTR_NHLEN_VPNV6_GLOBAL_AND_LL
) {
572 stream_get_from(&v6nhlocal
, s
, offset_nhlocal
,
574 if (IN6_IS_ADDR_UNSPECIFIED(&v6nhlocal
)) {
575 mod_v6nhl
= &peer
->nexthop
.v6_local
;
581 stream_put_in6_addr_at(s
, offset_nhglobal
,
584 stream_put_in6_addr_at(s
, offset_nhlocal
,
587 if (bgp_debug_update(peer
, NULL
, NULL
, 0)) {
588 if (nhlen
== 32 || nhlen
== 48)
590 "u%" PRIu64
":s%" PRIu64
591 " %s send UPDATE w/ mp_nexthops %s, %s%s",
594 PAF_SUBGRP(paf
)->id
, peer
->host
,
595 inet_ntop(AF_INET6
, mod_v6nhg
,
597 inet_ntop(AF_INET6
, mod_v6nhl
,
599 (nhlen
== 48 ? " and RD" : ""));
602 "u%" PRIu64
":s%" PRIu64
603 " %s send UPDATE w/ mp_nexthop %s%s",
606 PAF_SUBGRP(paf
)->id
, peer
->host
,
607 inet_ntop(AF_INET6
, mod_v6nhg
,
609 (nhlen
== 24 ? " and RD" : ""));
611 } else if (paf
->afi
== AFI_L2VPN
) {
612 struct in_addr v4nh
, *mod_v4nh
;
615 stream_get_from(&v4nh
, s
, vec
->offset
+ 1, 4);
618 /* No route-map changes allowed for EVPN nexthops. */
620 mod_v4nh
= &peer
->nexthop
.v4
;
625 stream_put_in_addr_at(s
, vec
->offset
+ 1,
628 if (bgp_debug_update(peer
, NULL
, NULL
, 0))
629 zlog_debug("u%" PRIu64
":s%" PRIu64
630 " %s send UPDATE w/ nexthop %s",
631 PAF_SUBGRP(paf
)->update_group
->id
,
632 PAF_SUBGRP(paf
)->id
, peer
->host
,
633 inet_ntoa(*mod_v4nh
));
637 bgp_packet_add(peer
, s
);
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 u_int32_t addpath_tx_id
= 0;
705 struct prefix_rd
*prd
= NULL
;
706 mpls_label_t label
= MPLS_INVALID_LABEL
;
711 if (bpacket_queue_is_full(SUBGRP_INST(subgrp
), SUBGRP_PKTQ(subgrp
)))
714 peer
= SUBGRP_PEER(subgrp
);
715 afi
= SUBGRP_AFI(subgrp
);
716 safi
= SUBGRP_SAFI(subgrp
);
719 snlri
= subgrp
->scratch
;
722 bpacket_attr_vec_arr_reset(&vecarr
);
724 addpath_encode
= bgp_addpath_encode_tx(peer
, afi
, safi
);
725 addpath_overhead
= addpath_encode
? BGP_ADDPATH_ID_LEN
: 0;
727 adv
= BGP_ADV_FIFO_HEAD(&subgrp
->sync
->update
);
732 addpath_tx_id
= adj
->addpath_tx_id
;
735 space_remaining
= STREAM_CONCAT_REMAIN(s
, snlri
, STREAM_SIZE(s
))
736 - BGP_MAX_PACKET_SIZE_OVERFLOW
;
737 space_needed
= BGP_NLRI_LENGTH
+ addpath_overhead
+
738 bgp_packet_mpattr_prefix_size(afi
, safi
, &rn
->p
);
740 /* When remaining space can't include NLRI and it's length. */
741 if (space_remaining
< space_needed
)
744 /* If packet is empty, set attribute. */
745 if (stream_empty(s
)) {
746 struct peer
*from
= NULL
;
751 /* 1: Write the BGP message header - 16 bytes marker, 2
753 * one byte message type.
755 bgp_packet_set_marker(s
, BGP_MSG_UPDATE
);
757 /* 2: withdrawn routes length */
760 /* 3: total attributes length - attrlen_pos stores the
762 attrlen_pos
= stream_get_endp(s
);
765 /* 4: if there is MP_REACH_NLRI attribute, that should
767 * attribute, according to
768 * draft-ietf-idr-error-handling. Save the
771 mpattr_pos
= stream_get_endp(s
);
773 /* 5: Encode all the attributes, except MP_REACH_NLRI
775 total_attr_len
= bgp_packet_attribute(
776 NULL
, peer
, s
, adv
->baa
->attr
, &vecarr
, NULL
,
777 afi
, safi
, from
, NULL
, NULL
, 0, 0);
780 STREAM_CONCAT_REMAIN(s
, snlri
, STREAM_SIZE(s
))
781 - BGP_MAX_PACKET_SIZE_OVERFLOW
;
782 space_needed
= BGP_NLRI_LENGTH
+ addpath_overhead
+
783 bgp_packet_mpattr_prefix_size(afi
, safi
,
786 /* If the attributes alone do not leave any room for
789 if (space_remaining
< space_needed
) {
791 "u%" PRIu64
":s%" PRIu64
792 " attributes too long, cannot send UPDATE",
793 subgrp
->update_group
->id
, subgrp
->id
);
795 /* Flush the FIFO update queue */
797 adv
= bgp_advertise_clean_subgroup(
802 if (BGP_DEBUG(update
, UPDATE_OUT
)
803 || BGP_DEBUG(update
, UPDATE_PREFIX
)) {
804 memset(send_attr_str
, 0, BUFSIZ
);
805 send_attr_printed
= 0;
806 bgp_dump_attr(adv
->baa
->attr
, send_attr_str
,
811 if ((afi
== AFI_IP
&& safi
== SAFI_UNICAST
)
812 && !peer_cap_enhe(peer
, afi
, safi
))
813 stream_put_prefix_addpath(s
, &rn
->p
, addpath_encode
,
816 /* Encode the prefix in MP_REACH_NLRI attribute */
818 prd
= (struct prefix_rd
*)&rn
->prn
->p
;
820 if (safi
== SAFI_LABELED_UNICAST
)
821 label
= bgp_adv_label(rn
, binfo
, peer
, afi
,
823 else if (binfo
&& binfo
->extra
)
824 label
= binfo
->extra
->label
;
826 if (stream_empty(snlri
))
827 mpattrlen_pos
= bgp_packet_mpattr_start(
828 snlri
, peer
, afi
, safi
, &vecarr
,
831 bgp_packet_mpattr_prefix(snlri
, afi
, safi
, &rn
->p
, prd
,
832 &label
, addpath_encode
,
833 addpath_tx_id
, adv
->baa
->attr
);
838 if (bgp_debug_update(NULL
, &rn
->p
, subgrp
->update_group
, 0)) {
839 char pfx_buf
[BGP_PRD_PATH_STRLEN
];
841 if (!send_attr_printed
) {
842 zlog_debug("u%" PRIu64
":s%" PRIu64
843 " send UPDATE w/ attr: %s",
844 subgrp
->update_group
->id
, subgrp
->id
,
846 if (!stream_empty(snlri
)) {
850 pkt_afi
= afi_int2iana(afi
);
851 pkt_safi
= safi_int2iana(safi
);
853 "u%" PRIu64
":s%" PRIu64
854 " send MP_REACH for afi/safi %d/%d",
855 subgrp
->update_group
->id
,
856 subgrp
->id
, pkt_afi
, pkt_safi
);
859 send_attr_printed
= 1;
862 bgp_debug_rdpfxpath2str(afi
, safi
, prd
, &rn
->p
, &label
,
863 addpath_encode
, addpath_tx_id
,
864 pfx_buf
, sizeof(pfx_buf
));
865 zlog_debug("u%" PRIu64
":s%" PRIu64
" send UPDATE %s",
866 subgrp
->update_group
->id
, subgrp
->id
,
870 /* Synchnorize attribute. */
872 bgp_attr_unintern(&adj
->attr
);
876 adj
->attr
= bgp_attr_intern(adv
->baa
->attr
);
878 adv
= bgp_advertise_clean_subgroup(subgrp
, adj
);
881 if (!stream_empty(s
)) {
882 if (!stream_empty(snlri
)) {
883 bgp_packet_mpattr_end(snlri
, mpattrlen_pos
);
884 total_attr_len
+= stream_get_endp(snlri
);
887 /* set the total attribute length correctly */
888 stream_putw_at(s
, attrlen_pos
, total_attr_len
);
890 if (!stream_empty(snlri
)) {
891 packet
= stream_dupcat(s
, snlri
, mpattr_pos
);
892 bpacket_attr_vec_arr_update(&vecarr
, mpattr_pos
);
894 packet
= stream_dup(s
);
895 bgp_packet_set_size(packet
);
896 if (bgp_debug_update(NULL
, NULL
, subgrp
->update_group
, 0))
897 zlog_debug("u%" PRIu64
":s%" PRIu64
898 " send UPDATE len %zd numpfx %d",
899 subgrp
->update_group
->id
, subgrp
->id
,
900 (stream_get_endp(packet
)
901 - stream_get_getp(packet
)),
903 pkt
= bpacket_queue_add(SUBGRP_PKTQ(subgrp
), packet
, &vecarr
);
911 /* Make BGP withdraw packet. */
913 16-octet marker | 2-octet length | 1-octet type |
914 2-octet withdrawn route length | withdrawn prefixes | 2-octet attrlen (=0)
916 /* For other afi/safis:
917 16-octet marker | 2-octet length | 1-octet type |
918 2-octet withdrawn route length (=0) | 2-octet attrlen |
919 mp_unreach attr type | attr len | afi | safi | withdrawn prefixes
921 struct bpacket
*subgroup_withdraw_packet(struct update_subgroup
*subgrp
)
925 struct bgp_adj_out
*adj
;
926 struct bgp_advertise
*adv
;
929 bgp_size_t unfeasible_len
;
930 bgp_size_t total_attr_len
;
932 size_t attrlen_pos
= 0;
933 size_t mplen_pos
= 0;
934 u_char first_time
= 1;
937 int space_remaining
= 0;
938 int space_needed
= 0;
940 int addpath_encode
= 0;
941 int addpath_overhead
= 0;
942 u_int32_t addpath_tx_id
= 0;
943 struct prefix_rd
*prd
= NULL
;
949 if (bpacket_queue_is_full(SUBGRP_INST(subgrp
), SUBGRP_PKTQ(subgrp
)))
952 peer
= SUBGRP_PEER(subgrp
);
953 afi
= SUBGRP_AFI(subgrp
);
954 safi
= SUBGRP_SAFI(subgrp
);
957 addpath_encode
= bgp_addpath_encode_tx(peer
, afi
, safi
);
958 addpath_overhead
= addpath_encode
? BGP_ADDPATH_ID_LEN
: 0;
960 while ((adv
= BGP_ADV_FIFO_HEAD(&subgrp
->sync
->withdraw
)) != NULL
) {
964 addpath_tx_id
= adj
->addpath_tx_id
;
967 STREAM_REMAIN(s
) - BGP_MAX_PACKET_SIZE_OVERFLOW
;
969 BGP_NLRI_LENGTH
+ addpath_overhead
+ BGP_TOTAL_ATTR_LEN
970 + bgp_packet_mpattr_prefix_size(afi
, safi
, &rn
->p
);
972 if (space_remaining
< space_needed
)
975 if (stream_empty(s
)) {
976 bgp_packet_set_marker(s
, BGP_MSG_UPDATE
);
977 stream_putw(s
, 0); /* unfeasible routes length */
981 if (afi
== AFI_IP
&& safi
== SAFI_UNICAST
982 && !peer_cap_enhe(peer
, afi
, safi
))
983 stream_put_prefix_addpath(s
, &rn
->p
, addpath_encode
,
987 prd
= (struct prefix_rd
*)&rn
->prn
->p
;
989 /* If first time, format the MP_UNREACH header */
994 pkt_afi
= afi_int2iana(afi
);
995 pkt_safi
= safi_int2iana(safi
);
997 attrlen_pos
= stream_get_endp(s
);
998 /* total attr length = 0 for now. reevaluate
1001 mp_start
= stream_get_endp(s
);
1002 mplen_pos
= bgp_packet_mpunreach_start(s
, afi
,
1004 if (bgp_debug_update(NULL
, NULL
,
1005 subgrp
->update_group
, 0))
1007 "u%" PRIu64
":s%" PRIu64
1008 " send MP_UNREACH for afi/safi %d/%d",
1009 subgrp
->update_group
->id
,
1010 subgrp
->id
, pkt_afi
, pkt_safi
);
1013 bgp_packet_mpunreach_prefix(s
, &rn
->p
, afi
, safi
, prd
,
1014 NULL
, addpath_encode
,
1015 addpath_tx_id
, NULL
);
1020 if (bgp_debug_update(NULL
, &rn
->p
, subgrp
->update_group
, 0)) {
1021 char pfx_buf
[BGP_PRD_PATH_STRLEN
];
1023 bgp_debug_rdpfxpath2str(afi
, safi
, prd
, &rn
->p
, NULL
,
1024 addpath_encode
, addpath_tx_id
,
1025 pfx_buf
, sizeof(pfx_buf
));
1026 zlog_debug("u%" PRIu64
":s%" PRIu64
1027 " send UPDATE %s -- unreachable",
1028 subgrp
->update_group
->id
, subgrp
->id
,
1034 bgp_adj_out_remove_subgroup(rn
, adj
, subgrp
);
1035 bgp_unlock_node(rn
);
1038 if (!stream_empty(s
)) {
1039 if (afi
== AFI_IP
&& safi
== SAFI_UNICAST
1040 && !peer_cap_enhe(peer
, afi
, safi
)) {
1041 unfeasible_len
= stream_get_endp(s
) - BGP_HEADER_SIZE
1042 - BGP_UNFEASIBLE_LEN
;
1043 stream_putw_at(s
, BGP_HEADER_SIZE
, unfeasible_len
);
1046 /* Set the mp_unreach attr's length */
1047 bgp_packet_mpunreach_end(s
, mplen_pos
);
1049 /* Set total path attribute length. */
1050 total_attr_len
= stream_get_endp(s
) - mp_start
;
1051 stream_putw_at(s
, attrlen_pos
, total_attr_len
);
1053 bgp_packet_set_size(s
);
1054 if (bgp_debug_update(NULL
, NULL
, subgrp
->update_group
, 0))
1055 zlog_debug("u%" PRIu64
":s%" PRIu64
1056 " send UPDATE (withdraw) len %zd numpfx %d",
1057 subgrp
->update_group
->id
, subgrp
->id
,
1058 (stream_get_endp(s
) - stream_get_getp(s
)),
1060 pkt
= bpacket_queue_add(SUBGRP_PKTQ(subgrp
), stream_dup(s
),
1069 void subgroup_default_update_packet(struct update_subgroup
*subgrp
,
1070 struct attr
*attr
, struct peer
*from
)
1076 bgp_size_t total_attr_len
;
1079 struct bpacket_attr_vec_arr vecarr
;
1080 int addpath_encode
= 0;
1082 if (DISABLE_BGP_ANNOUNCE
)
1088 peer
= SUBGRP_PEER(subgrp
);
1089 afi
= SUBGRP_AFI(subgrp
);
1090 safi
= SUBGRP_SAFI(subgrp
);
1091 bpacket_attr_vec_arr_reset(&vecarr
);
1092 addpath_encode
= bgp_addpath_encode_tx(peer
, afi
, safi
);
1095 str2prefix("0.0.0.0/0", &p
);
1097 str2prefix("::/0", &p
);
1099 /* Logging the attribute. */
1100 if (bgp_debug_update(NULL
, &p
, subgrp
->update_group
, 0)) {
1101 char attrstr
[BUFSIZ
];
1102 char buf
[PREFIX_STRLEN
];
1103 /* ' with addpath ID ' 17
1104 * max strlen of uint32 + 10
1105 * +/- (just in case) + 1
1106 * null terminator + 1
1107 * ============================ 29 */
1112 bgp_dump_attr(attr
, attrstr
, BUFSIZ
);
1115 snprintf(tx_id_buf
, sizeof(tx_id_buf
),
1116 " with addpath ID %u",
1117 BGP_ADDPATH_TX_ID_FOR_DEFAULT_ORIGINATE
);
1119 zlog_debug("u%" PRIu64
":s%" PRIu64
" send UPDATE %s%s %s",
1120 (SUBGRP_UPDGRP(subgrp
))->id
, subgrp
->id
,
1121 prefix2str(&p
, buf
, sizeof(buf
)), tx_id_buf
,
1125 s
= stream_new(BGP_MAX_PACKET_SIZE
);
1127 /* Make BGP update packet. */
1128 bgp_packet_set_marker(s
, BGP_MSG_UPDATE
);
1130 /* Unfeasible Routes Length. */
1133 /* Make place for total attribute length. */
1134 pos
= stream_get_endp(s
);
1136 total_attr_len
= bgp_packet_attribute(
1137 NULL
, peer
, s
, attr
, &vecarr
, &p
, afi
, safi
, from
, NULL
, NULL
,
1138 addpath_encode
, BGP_ADDPATH_TX_ID_FOR_DEFAULT_ORIGINATE
);
1140 /* Set Total Path Attribute Length. */
1141 stream_putw_at(s
, pos
, total_attr_len
);
1144 if (p
.family
== AF_INET
&& safi
== SAFI_UNICAST
1145 && !peer_cap_enhe(peer
, afi
, safi
))
1146 stream_put_prefix_addpath(
1147 s
, &p
, addpath_encode
,
1148 BGP_ADDPATH_TX_ID_FOR_DEFAULT_ORIGINATE
);
1151 bgp_packet_set_size(s
);
1153 (void)bpacket_queue_add(SUBGRP_PKTQ(subgrp
), s
, &vecarr
);
1154 subgroup_trigger_write(subgrp
);
1157 void subgroup_default_withdraw_packet(struct update_subgroup
*subgrp
)
1162 unsigned long attrlen_pos
= 0;
1164 bgp_size_t unfeasible_len
;
1165 bgp_size_t total_attr_len
= 0;
1166 size_t mp_start
= 0;
1167 size_t mplen_pos
= 0;
1170 int addpath_encode
= 0;
1172 if (DISABLE_BGP_ANNOUNCE
)
1175 peer
= SUBGRP_PEER(subgrp
);
1176 afi
= SUBGRP_AFI(subgrp
);
1177 safi
= SUBGRP_SAFI(subgrp
);
1178 addpath_encode
= bgp_addpath_encode_tx(peer
, afi
, safi
);
1181 str2prefix("0.0.0.0/0", &p
);
1183 str2prefix("::/0", &p
);
1185 if (bgp_debug_update(NULL
, &p
, subgrp
->update_group
, 0)) {
1186 char buf
[PREFIX_STRLEN
];
1187 /* ' with addpath ID ' 17
1188 * max strlen of uint32 + 10
1189 * +/- (just in case) + 1
1190 * null terminator + 1
1191 * ============================ 29 */
1195 snprintf(tx_id_buf
, sizeof(tx_id_buf
),
1196 " with addpath ID %u",
1197 BGP_ADDPATH_TX_ID_FOR_DEFAULT_ORIGINATE
);
1199 zlog_debug("u%" PRIu64
":s%" PRIu64
1200 " send UPDATE %s%s -- unreachable",
1201 (SUBGRP_UPDGRP(subgrp
))->id
, subgrp
->id
,
1202 prefix2str(&p
, buf
, sizeof(buf
)), tx_id_buf
);
1205 s
= stream_new(BGP_MAX_PACKET_SIZE
);
1207 /* Make BGP update packet. */
1208 bgp_packet_set_marker(s
, BGP_MSG_UPDATE
);
1210 /* Unfeasible Routes Length. */;
1211 cp
= stream_get_endp(s
);
1214 /* Withdrawn Routes. */
1215 if (p
.family
== AF_INET
&& safi
== SAFI_UNICAST
1216 && !peer_cap_enhe(peer
, afi
, safi
)) {
1217 stream_put_prefix_addpath(
1218 s
, &p
, addpath_encode
,
1219 BGP_ADDPATH_TX_ID_FOR_DEFAULT_ORIGINATE
);
1221 unfeasible_len
= stream_get_endp(s
) - cp
- 2;
1223 /* Set unfeasible len. */
1224 stream_putw_at(s
, cp
, unfeasible_len
);
1226 /* Set total path attribute length. */
1229 attrlen_pos
= stream_get_endp(s
);
1231 mp_start
= stream_get_endp(s
);
1232 mplen_pos
= bgp_packet_mpunreach_start(s
, afi
, safi
);
1233 bgp_packet_mpunreach_prefix(
1234 s
, &p
, afi
, safi
, NULL
, NULL
, addpath_encode
,
1235 BGP_ADDPATH_TX_ID_FOR_DEFAULT_ORIGINATE
, NULL
);
1237 /* Set the mp_unreach attr's length */
1238 bgp_packet_mpunreach_end(s
, mplen_pos
);
1240 /* Set total path attribute length. */
1241 total_attr_len
= stream_get_endp(s
) - mp_start
;
1242 stream_putw_at(s
, attrlen_pos
, total_attr_len
);
1245 bgp_packet_set_size(s
);
1247 (void)bpacket_queue_add(SUBGRP_PKTQ(subgrp
), s
, NULL
);
1248 subgroup_trigger_write(subgrp
);
1252 bpacket_vec_arr_inherit_attr_flags(struct bpacket_attr_vec_arr
*vecarr
,
1253 bpacket_attr_vec_type type
,
1256 if (CHECK_FLAG(attr
->rmap_change_flags
,
1257 BATTR_RMAP_NEXTHOP_PEER_ADDRESS
))
1258 SET_FLAG(vecarr
->entries
[BGP_ATTR_VEC_NH
].flags
,
1259 BPKT_ATTRVEC_FLAGS_RMAP_NH_PEER_ADDRESS
);
1261 if (CHECK_FLAG(attr
->rmap_change_flags
, BATTR_REFLECTED
))
1262 SET_FLAG(vecarr
->entries
[BGP_ATTR_VEC_NH
].flags
,
1263 BPKT_ATTRVEC_FLAGS_REFLECTED
);
1265 if (CHECK_FLAG(attr
->rmap_change_flags
, BATTR_RMAP_NEXTHOP_UNCHANGED
))
1266 SET_FLAG(vecarr
->entries
[BGP_ATTR_VEC_NH
].flags
,
1267 BPKT_ATTRVEC_FLAGS_RMAP_NH_UNCHANGED
);
1269 if (CHECK_FLAG(attr
->rmap_change_flags
, BATTR_RMAP_IPV4_NHOP_CHANGED
))
1270 SET_FLAG(vecarr
->entries
[BGP_ATTR_VEC_NH
].flags
,
1271 BPKT_ATTRVEC_FLAGS_RMAP_IPV4_NH_CHANGED
);
1273 if (CHECK_FLAG(attr
->rmap_change_flags
,
1274 BATTR_RMAP_IPV6_GLOBAL_NHOP_CHANGED
))
1275 SET_FLAG(vecarr
->entries
[BGP_ATTR_VEC_NH
].flags
,
1276 BPKT_ATTRVEC_FLAGS_RMAP_IPV6_GNH_CHANGED
);
1278 if (CHECK_FLAG(attr
->rmap_change_flags
,
1279 BATTR_RMAP_IPV6_LL_NHOP_CHANGED
))
1280 SET_FLAG(vecarr
->entries
[BGP_ATTR_VEC_NH
].flags
,
1281 BPKT_ATTRVEC_FLAGS_RMAP_IPV6_LNH_CHANGED
);
1284 /* Reset the Attributes vector array. The vector array is used to override
1285 * certain output parameters in the packet for a particular peer
1287 void bpacket_attr_vec_arr_reset(struct bpacket_attr_vec_arr
*vecarr
)
1295 while (i
< BGP_ATTR_VEC_MAX
) {
1296 vecarr
->entries
[i
].flags
= 0;
1297 vecarr
->entries
[i
].offset
= 0;
1302 /* Setup a particular node entry in the vecarr */
1303 void bpacket_attr_vec_arr_set_vec(struct bpacket_attr_vec_arr
*vecarr
,
1304 bpacket_attr_vec_type type
, struct stream
*s
,
1309 assert(type
< BGP_ATTR_VEC_MAX
);
1311 SET_FLAG(vecarr
->entries
[type
].flags
, BPKT_ATTRVEC_FLAGS_UPDATED
);
1312 vecarr
->entries
[type
].offset
= stream_get_endp(s
);
1314 bpacket_vec_arr_inherit_attr_flags(vecarr
, type
, attr
);