2 * bgp_updgrp.c: BGP update group structures
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
27 #ifndef _QUAGGA_BGP_UPDGRP_H
28 #define _QUAGGA_BGP_UPDGRP_H
30 #include "bgp_advertise.h"
33 * The following three heuristic constants determine how long advertisement to
34 * a subgroup will be delayed after it is created. The intent is to allow
35 * transient changes in peer state (primarily session establishment) to settle,
36 * so that more peers can be grouped together and benefit from sharing
37 * advertisement computations with the subgroup.
39 * These values have a very large impact on initial convergence time; any
40 * changes should be accompanied by careful performance testing at all scales.
42 * The coalesce time 'C' for a new subgroup within a particular BGP instance
43 * 'B' with total number of known peers 'P', established or not, is computed as
46 * C = MIN(BGP_MAX_SUBGROUP_COALESCE_TIME,
47 * BGP_DEFAULT_SUBGROUP_COALESCE_TIME +
48 * (P*BGP_PEER_ADJUST_SUBGROUP_COALESCE_TIME))
50 #define BGP_DEFAULT_SUBGROUP_COALESCE_TIME 1000
51 #define BGP_MAX_SUBGROUP_COALESCE_TIME 10000
52 #define BGP_PEER_ADJUST_SUBGROUP_COALESCE_TIME 50
54 #define PEER_UPDGRP_FLAGS \
55 (PEER_FLAG_LOCAL_AS_NO_PREPEND | PEER_FLAG_LOCAL_AS_REPLACE_AS)
57 #define PEER_UPDGRP_AF_FLAGS \
58 (PEER_FLAG_SEND_COMMUNITY | PEER_FLAG_SEND_EXT_COMMUNITY \
59 | PEER_FLAG_SEND_LARGE_COMMUNITY \
60 | PEER_FLAG_DEFAULT_ORIGINATE | PEER_FLAG_REFLECTOR_CLIENT \
61 | PEER_FLAG_RSERVER_CLIENT | PEER_FLAG_NEXTHOP_SELF \
62 | PEER_FLAG_NEXTHOP_UNCHANGED | PEER_FLAG_FORCE_NEXTHOP_SELF \
63 | PEER_FLAG_AS_PATH_UNCHANGED | PEER_FLAG_MED_UNCHANGED \
64 | PEER_FLAG_NEXTHOP_LOCAL_UNCHANGED | PEER_FLAG_REMOVE_PRIVATE_AS \
65 | PEER_FLAG_REMOVE_PRIVATE_AS_ALL \
66 | PEER_FLAG_REMOVE_PRIVATE_AS_REPLACE \
67 | PEER_FLAG_REMOVE_PRIVATE_AS_ALL_REPLACE \
68 | PEER_FLAG_AS_OVERRIDE)
70 #define PEER_UPDGRP_CAP_FLAGS (PEER_CAP_AS4_RCV)
72 #define PEER_UPDGRP_AF_CAP_FLAGS \
73 (PEER_CAP_ORF_PREFIX_SM_RCV | PEER_CAP_ORF_PREFIX_SM_OLD_RCV \
74 | PEER_CAP_ADDPATH_AF_TX_ADV | PEER_CAP_ADDPATH_AF_RX_RCV \
75 | PEER_CAP_ENHE_AF_NEGO)
77 typedef enum { BGP_ATTR_VEC_NH
= 0, BGP_ATTR_VEC_MAX
} bpacket_attr_vec_type
;
84 #define BPKT_ATTRVEC_FLAGS_UPDATED (1 << 0)
85 #define BPKT_ATTRVEC_FLAGS_RMAP_NH_PEER_ADDRESS (1 << 1)
86 #define BPKT_ATTRVEC_FLAGS_REFLECTED (1 << 2)
87 #define BPKT_ATTRVEC_FLAGS_RMAP_NH_UNCHANGED (1 << 3)
88 #define BPKT_ATTRVEC_FLAGS_RMAP_IPV4_NH_CHANGED (1 << 4)
89 #define BPKT_ATTRVEC_FLAGS_RMAP_IPV6_GNH_CHANGED (1 << 5)
90 #define BPKT_ATTRVEC_FLAGS_RMAP_IPV6_LNH_CHANGED (1 << 6)
92 typedef struct bpacket_attr_vec_arr
{
93 bpacket_attr_vec entries
[BGP_ATTR_VEC_MAX
];
94 } bpacket_attr_vec_arr
;
97 /* for being part of an update subgroup's message list */
98 TAILQ_ENTRY(bpacket
) pkt_train
;
100 /* list of peers (well, peer_afs) that the packet needs to be sent to */
101 LIST_HEAD(pkt_peer_list
, peer_af
) peers
;
103 struct stream
*buffer
;
104 bpacket_attr_vec_arr arr
;
109 struct bpacket_queue
{
110 TAILQ_HEAD(pkt_queue
, bpacket
) pkts
;
112 unsigned int conf_max_count
;
113 unsigned int curr_count
;
114 unsigned int hwm_count
;
115 unsigned int max_count_reached_count
;
118 struct update_group
{
119 /* back pointer to the BGP instance */
122 /* list of subgroups that belong to the update group */
123 LIST_HEAD(subgrp_list
, update_subgroup
) subgrps
;
125 /* lazy way to store configuration common to all peers
126 hash function will compute from this data */
136 uint32_t join_events
;
137 uint32_t prune_events
;
138 uint32_t merge_events
;
139 uint32_t updgrp_switch_events
;
140 uint32_t peer_refreshes_combined
;
142 uint32_t split_events
;
143 uint32_t merge_checks_triggered
;
145 uint32_t subgrps_created
;
146 uint32_t subgrps_deleted
;
148 uint32_t num_dbg_en_peers
;
152 * Shorthand for a global statistics counter.
154 #define UPDGRP_GLOBAL_STAT(updgrp, stat) \
155 ((updgrp)->bgp->update_group_stats.stat)
158 * Add the given value to a counter on an update group and the bgp
161 #define UPDGRP_INCR_STAT_BY(updgrp, stat, value) \
163 (updgrp)->stat += (value); \
164 UPDGRP_GLOBAL_STAT(updgrp, stat) += (value); \
168 * Increment a counter on a update group and its parent structures.
170 #define UPDGRP_INCR_STAT(subgrp, stat) UPDGRP_INCR_STAT_BY(subgrp, stat, 1)
172 struct update_subgroup
{
173 /* back pointer to the parent update group */
174 struct update_group
*update_group
;
176 /* list of peers that belong to the subgroup */
177 LIST_HEAD(peer_list
, peer_af
) peers
;
180 /* for being part of an update group's subgroup list */
181 LIST_ENTRY(update_subgroup
) updgrp_train
;
183 struct bpacket_queue pkt_queue
;
186 * List of adj-out structures for this subgroup.
187 * It essentially represents the snapshot of every prefix that
188 * has been advertised to the members of the subgroup
190 TAILQ_HEAD(adjout_queue
, bgp_adj_out
) adjq
;
192 /* packet buffer for update generation */
195 /* We use a separate stream to encode MP_REACH_NLRI for efficient
196 * NLRI packing. peer->obuf_work stores all the other attributes. The
197 * actual packet is then constructed by concatenating the two.
199 struct stream
*scratch
;
201 /* synchronization list and time */
202 struct bgp_synchronize
*sync
;
204 /* send prefix count */
207 /* send prefix count prior to packet update */
210 /* announcement attribute hash */
213 struct thread
*t_coalesce
;
216 struct thread
*t_merge_check
;
218 /* table version that the subgroup has caught up to. */
221 /* version maintained to record adj changes */
222 uint64_t adj_version
;
227 * Identifying information about the subgroup that this subgroup was
232 uint64_t update_group_id
;
233 uint64_t subgroup_id
;
236 uint32_t join_events
;
237 uint32_t prune_events
;
240 * This is bumped up when another subgroup merges into this one.
242 uint32_t merge_events
;
243 uint32_t updgrp_switch_events
;
244 uint32_t peer_refreshes_combined
;
246 uint32_t split_events
;
247 uint32_t merge_checks_triggered
;
252 #define SUBGRP_STATUS_DEFAULT_ORIGINATE (1 << 0)
253 #define SUBGRP_STATUS_FORCE_UPDATES (1 << 1)
254 #define SUBGRP_STATUS_TABLE_REPARSING (1 << 2)
257 #define SUBGRP_FLAG_NEEDS_REFRESH (1 << 0)
261 * Add the given value to the specified counter on a subgroup and its
264 #define SUBGRP_INCR_STAT_BY(subgrp, stat, value) \
266 (subgrp)->stat += (value); \
267 if ((subgrp)->update_group) \
268 UPDGRP_INCR_STAT_BY((subgrp)->update_group, stat, \
273 * Increment a counter on a subgroup and its parent structures.
275 #define SUBGRP_INCR_STAT(subgrp, stat) SUBGRP_INCR_STAT_BY(subgrp, stat, 1)
278 * Decrement a counter on a subgroup and its parent structures.
280 #define SUBGRP_DECR_STAT(subgrp, stat) SUBGRP_INCR_STAT_BY(subgrp, stat, -1)
282 typedef int (*updgrp_walkcb
)(struct update_group
*updgrp
, void *ctx
);
284 /* really a private structure */
285 struct updwalk_context
{
287 struct bgp_dest
*dest
;
288 struct bgp_path_info
*pi
;
291 bgp_policy_type_e policy_type
;
292 const char *policy_name
;
293 int policy_event_start_flag
;
294 int policy_route_update
;
299 #define UPDWALK_FLAGS_ADVQUEUE (1 << 0)
300 #define UPDWALK_FLAGS_ADVERTISED (1 << 1)
303 #define UPDWALK_CONTINUE HASHWALK_CONTINUE
304 #define UPDWALK_ABORT HASHWALK_ABORT
306 #define PAF_PEER(p) ((p)->peer)
307 #define PAF_SUBGRP(p) ((p)->subgroup)
308 #define PAF_UPDGRP(p) ((p)->subgroup->update_group)
309 #define PAF_PKTQ(f) SUBGRP_PKTQ((f)->subgroup)
311 #define UPDGRP_PEER(u) ((u)->conf)
312 #define UPDGRP_AFI(u) ((u)->afi)
313 #define UPDGRP_SAFI(u) ((u)->safi)
314 #define UPDGRP_INST(u) ((u)->bgp)
315 #define UPDGRP_AFFLAGS(u) ((u)->conf->af_flags[UPDGRP_AFI(u)][UPDGRP_SAFI(u)])
316 #define UPDGRP_DBG_ON(u) ((u)->num_dbg_en_peers)
317 #define UPDGRP_PEER_DBG_EN(u) (((u)->num_dbg_en_peers)++)
318 #define UPDGRP_PEER_DBG_DIS(u) (((u)->num_dbg_en_peers)--)
319 #define UPDGRP_PEER_DBG_OFF(u) (u)->num_dbg_en_peers = 0
321 #define SUBGRP_AFI(s) UPDGRP_AFI((s)->update_group)
322 #define SUBGRP_SAFI(s) UPDGRP_SAFI((s)->update_group)
323 #define SUBGRP_PEER(s) UPDGRP_PEER((s)->update_group)
324 #define SUBGRP_PCOUNT(s) ((s)->peer_count)
325 #define SUBGRP_PFIRST(s) LIST_FIRST(&((s)->peers))
326 #define SUBGRP_PKTQ(s) &((s)->pkt_queue)
327 #define SUBGRP_INST(s) UPDGRP_INST((s)->update_group)
328 #define SUBGRP_AFFLAGS(s) UPDGRP_AFFLAGS((s)->update_group)
329 #define SUBGRP_UPDGRP(s) ((s)->update_group)
332 * Walk all subgroups in an update group.
334 #define UPDGRP_FOREACH_SUBGRP(updgrp, subgrp) \
335 LIST_FOREACH (subgrp, &((updgrp)->subgrps), updgrp_train)
337 #define UPDGRP_FOREACH_SUBGRP_SAFE(updgrp, subgrp, tmp_subgrp) \
338 LIST_FOREACH_SAFE (subgrp, &((updgrp)->subgrps), updgrp_train, \
341 #define SUBGRP_FOREACH_PEER(subgrp, paf) \
342 LIST_FOREACH (paf, &(subgrp->peers), subgrp_train)
344 #define SUBGRP_FOREACH_PEER_SAFE(subgrp, paf, temp_paf) \
345 LIST_FOREACH_SAFE (paf, &(subgrp->peers), subgrp_train, temp_paf)
347 #define SUBGRP_FOREACH_ADJ(subgrp, adj) \
348 TAILQ_FOREACH (adj, &(subgrp->adjq), subgrp_adj_train)
350 #define SUBGRP_FOREACH_ADJ_SAFE(subgrp, adj, adj_temp) \
351 TAILQ_FOREACH_SAFE (adj, &(subgrp->adjq), subgrp_adj_train, adj_temp)
355 extern void update_bgp_group_init(struct bgp
*);
356 extern void udpate_bgp_group_free(struct bgp
*);
358 extern void update_group_show(struct bgp
*bgp
, afi_t afi
, safi_t safi
,
359 struct vty
*vty
, uint64_t subgrp_id
);
360 extern void update_group_show_stats(struct bgp
*bgp
, struct vty
*vty
);
361 extern void update_group_adjust_peer(struct peer_af
*paf
);
362 extern int update_group_adjust_soloness(struct peer
*peer
, int set
);
364 extern void update_subgroup_remove_peer(struct update_subgroup
*,
366 extern struct bgp_table
*update_subgroup_rib(struct update_subgroup
*);
367 extern void update_subgroup_split_peer(struct peer_af
*, struct update_group
*);
368 extern bool update_subgroup_check_merge(struct update_subgroup
*, const char *);
369 extern bool update_subgroup_trigger_merge_check(struct update_subgroup
*,
371 extern void update_group_policy_update(struct bgp
*bgp
, bgp_policy_type_e ptype
,
372 const char *pname
, int route_update
,
374 extern void update_group_af_walk(struct bgp
*bgp
, afi_t afi
, safi_t safi
,
375 updgrp_walkcb cb
, void *ctx
);
376 extern void update_group_walk(struct bgp
*bgp
, updgrp_walkcb cb
, void *ctx
);
377 extern void update_group_periodic_merge(struct bgp
*bgp
);
379 update_group_refresh_default_originate_route_map(struct thread
*thread
);
380 extern void update_group_start_advtimer(struct bgp
*bgp
);
382 extern void update_subgroup_inherit_info(struct update_subgroup
*to
,
383 struct update_subgroup
*from
);
385 /* bgp_updgrp_packet.c */
386 extern struct bpacket
*bpacket_alloc(void);
387 extern void bpacket_free(struct bpacket
*pkt
);
388 extern void bpacket_queue_init(struct bpacket_queue
*q
);
389 extern void bpacket_queue_cleanup(struct bpacket_queue
*q
);
390 extern void bpacket_queue_sanity_check(struct bpacket_queue
*q
);
391 extern struct bpacket
*bpacket_queue_add(struct bpacket_queue
*q
,
393 struct bpacket_attr_vec_arr
*vecarr
);
394 struct bpacket
*bpacket_queue_remove(struct bpacket_queue
*q
);
395 extern struct bpacket
*bpacket_queue_first(struct bpacket_queue
*q
);
396 struct bpacket
*bpacket_queue_last(struct bpacket_queue
*q
);
397 unsigned int bpacket_queue_length(struct bpacket_queue
*q
);
398 unsigned int bpacket_queue_hwm_length(struct bpacket_queue
*q
);
399 bool bpacket_queue_is_full(struct bgp
*bgp
, struct bpacket_queue
*q
);
400 extern void bpacket_queue_advance_peer(struct peer_af
*paf
);
401 extern void bpacket_queue_remove_peer(struct peer_af
*paf
);
402 extern void bpacket_add_peer(struct bpacket
*pkt
, struct peer_af
*paf
);
403 unsigned int bpacket_queue_virtual_length(struct peer_af
*paf
);
404 extern void bpacket_queue_show_vty(struct bpacket_queue
*q
, struct vty
*vty
);
405 bool subgroup_packets_to_build(struct update_subgroup
*subgrp
);
406 extern struct bpacket
*subgroup_update_packet(struct update_subgroup
*s
);
407 extern struct bpacket
*subgroup_withdraw_packet(struct update_subgroup
*s
);
408 extern struct stream
*bpacket_reformat_for_peer(struct bpacket
*pkt
,
409 struct peer_af
*paf
);
410 extern void bpacket_attr_vec_arr_reset(struct bpacket_attr_vec_arr
*vecarr
);
411 extern void bpacket_attr_vec_arr_set_vec(struct bpacket_attr_vec_arr
*vecarr
,
412 bpacket_attr_vec_type type
,
413 struct stream
*s
, struct attr
*attr
);
414 extern void subgroup_default_update_packet(struct update_subgroup
*subgrp
,
417 extern void subgroup_default_withdraw_packet(struct update_subgroup
*subgrp
);
419 /* bgp_updgrp_adv.c */
420 extern struct bgp_advertise
*
421 bgp_advertise_clean_subgroup(struct update_subgroup
*subgrp
,
422 struct bgp_adj_out
*adj
);
423 extern void update_group_show_adj_queue(struct bgp
*bgp
, afi_t afi
, safi_t safi
,
424 struct vty
*vty
, uint64_t id
);
425 extern void update_group_show_advertised(struct bgp
*bgp
, afi_t afi
,
426 safi_t safi
, struct vty
*vty
,
428 extern void update_group_show_packet_queue(struct bgp
*bgp
, afi_t afi
,
429 safi_t safi
, struct vty
*vty
,
431 extern void subgroup_announce_route(struct update_subgroup
*subgrp
);
432 extern void subgroup_announce_all(struct update_subgroup
*subgrp
);
434 extern void subgroup_default_originate(struct update_subgroup
*subgrp
,
436 extern void group_announce_route(struct bgp
*bgp
, afi_t afi
, safi_t safi
,
437 struct bgp_dest
*dest
,
438 struct bgp_path_info
*pi
);
439 extern void subgroup_clear_table(struct update_subgroup
*subgrp
);
440 extern void update_group_announce(struct bgp
*bgp
);
441 extern void update_group_announce_rrclients(struct bgp
*bgp
);
442 extern void peer_af_announce_route(struct peer_af
*paf
, int combine
);
443 extern struct bgp_adj_out
*bgp_adj_out_alloc(struct update_subgroup
*subgrp
,
444 struct bgp_dest
*dest
,
445 uint32_t addpath_tx_id
);
446 extern void bgp_adj_out_remove_subgroup(struct bgp_dest
*dest
,
447 struct bgp_adj_out
*adj
,
448 struct update_subgroup
*subgrp
);
449 extern void bgp_adj_out_set_subgroup(struct bgp_dest
*dest
,
450 struct update_subgroup
*subgrp
,
452 struct bgp_path_info
*path
);
453 extern void bgp_adj_out_unset_subgroup(struct bgp_dest
*dest
,
454 struct update_subgroup
*subgrp
,
455 char withdraw
, uint32_t addpath_tx_id
);
456 void subgroup_announce_table(struct update_subgroup
*subgrp
,
457 struct bgp_table
*table
);
458 extern void subgroup_trigger_write(struct update_subgroup
*subgrp
);
460 extern int update_group_clear_update_dbg(struct update_group
*updgrp
,
463 extern void update_bgp_group_free(struct bgp
*bgp
);
464 extern bool bgp_addpath_encode_tx(struct peer
*peer
, afi_t afi
, safi_t safi
);
465 extern bool bgp_check_selected(struct bgp_path_info
*bpi
, struct peer
*peer
,
466 bool addpath_capable
, afi_t afi
, safi_t safi
);
473 * bpacket_queue_is_empty
475 static inline int bpacket_queue_is_empty(struct bpacket_queue
*queue
)
479 * The packet queue is empty if it only contains a sentinel.
481 if (queue
->curr_count
!= 1)
484 assert(bpacket_queue_first(queue
)->buffer
== NULL
);
491 * Returns the packet after the given packet in a bpacket queue.
493 static inline struct bpacket
*bpacket_next(struct bpacket
*pkt
)
495 return TAILQ_NEXT(pkt
, pkt_train
);
499 * update_group_adjust_peer_afs
501 * Adjust all peer_af structures for the given peer.
503 static inline void update_group_adjust_peer_afs(struct peer
*peer
)
508 for (afidx
= BGP_AF_START
; afidx
< BGP_AF_MAX
; afidx
++) {
509 paf
= peer
->peer_af_array
[afidx
];
511 update_group_adjust_peer(paf
);
516 * update_group_remove_peer_afs
518 * Remove all peer_af structures for the given peer from their subgroups.
520 static inline void update_group_remove_peer_afs(struct peer
*peer
)
525 for (afidx
= BGP_AF_START
; afidx
< BGP_AF_MAX
; afidx
++) {
526 paf
= peer
->peer_af_array
[afidx
];
528 update_subgroup_remove_peer(PAF_SUBGRP(paf
), paf
);
533 * update_subgroup_needs_refresh
536 update_subgroup_needs_refresh(const struct update_subgroup
*subgrp
)
538 if (CHECK_FLAG(subgrp
->flags
, SUBGRP_FLAG_NEEDS_REFRESH
))
545 * update_subgroup_set_needs_refresh
548 update_subgroup_set_needs_refresh(struct update_subgroup
*subgrp
, int value
)
551 SET_FLAG(subgrp
->flags
, SUBGRP_FLAG_NEEDS_REFRESH
);
553 UNSET_FLAG(subgrp
->flags
, SUBGRP_FLAG_NEEDS_REFRESH
);
556 static inline struct update_subgroup
*peer_subgroup(struct peer
*peer
,
557 afi_t afi
, safi_t safi
)
561 paf
= peer_af_find(peer
, afi
, safi
);
563 return PAF_SUBGRP(paf
);
568 * update_group_adjust_peer_afs
570 * Adjust all peer_af structures for the given peer.
572 static inline void bgp_announce_peer(struct peer
*peer
)
577 for (afidx
= BGP_AF_START
; afidx
< BGP_AF_MAX
; afidx
++) {
578 paf
= peer
->peer_af_array
[afidx
];
580 subgroup_announce_all(PAF_SUBGRP(paf
));
585 * advertise_list_is_empty
587 static inline int advertise_list_is_empty(struct update_subgroup
*subgrp
)
589 if (bgp_adv_fifo_count(&subgrp
->sync
->update
)
590 || bgp_adv_fifo_count(&subgrp
->sync
->withdraw
)
591 || bgp_adv_fifo_count(&subgrp
->sync
->withdraw_low
)) {
598 #endif /* _QUAGGA_BGP_UPDGRP_H */