1 /* BGP-4 Finite State Machine
2 * From RFC1771 [A Border Gateway Protocol 4 (BGP-4)]
3 * Copyright (C) 1996, 97, 98 Kunihiro Ishiguro
5 * This file is part of GNU Zebra.
7 * GNU Zebra is free software; you can redistribute it and/or modify it
8 * under the terms of the GNU General Public License as published by the
9 * Free Software Foundation; either version 2, or (at your option) any
12 * GNU Zebra is distributed in the hope that it will be useful, but
13 * WITHOUT ANY WARRANTY; without even the implied warranty of
14 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
15 * General Public License for more details.
17 * You should have received a copy of the GNU General Public License along
18 * with this program; see the file COPYING; if not, write to the Free Software
19 * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
26 #include "sockunion.h"
33 #include "workqueue.h"
37 #include "lib_errors.h"
40 #include "bgpd/bgpd.h"
41 #include "bgpd/bgp_attr.h"
42 #include "bgpd/bgp_debug.h"
43 #include "bgpd/bgp_errors.h"
44 #include "bgpd/bgp_fsm.h"
45 #include "bgpd/bgp_packet.h"
46 #include "bgpd/bgp_network.h"
47 #include "bgpd/bgp_route.h"
48 #include "bgpd/bgp_dump.h"
49 #include "bgpd/bgp_open.h"
50 #include "bgpd/bgp_advertise.h"
51 #include "bgpd/bgp_updgrp.h"
52 #include "bgpd/bgp_nht.h"
53 #include "bgpd/bgp_bfd.h"
54 #include "bgpd/bgp_memory.h"
55 #include "bgpd/bgp_keepalives.h"
56 #include "bgpd/bgp_io.h"
57 #include "bgpd/bgp_zebra.h"
59 DEFINE_HOOK(peer_backward_transition
, (struct peer
* peer
), (peer
))
60 DEFINE_HOOK(peer_status_changed
, (struct peer
* peer
), (peer
))
61 extern const char *get_afi_safi_str(afi_t afi
, safi_t safi
, bool for_json
);
62 /* Definition of display strings corresponding to FSM events. This should be
63 * kept consistent with the events defined in bgpd.h
65 static const char *const bgp_event_str
[] = {
69 "TCP_connection_open",
70 "TCP_connection_closed",
71 "TCP_connection_open_failed",
73 "ConnectRetry_timer_expired",
75 "KeepAlive_timer_expired",
76 "Receive_OPEN_message",
77 "Receive_KEEPALIVE_message",
78 "Receive_UPDATE_message",
79 "Receive_NOTIFICATION_message",
83 /* BGP FSM (finite state machine) has three types of functions. Type
84 one is thread functions. Type two is event functions. Type three
85 is FSM functions. Timer functions are set by bgp_timer_set
88 /* BGP event function. */
89 int bgp_event(struct thread
*);
91 /* BGP thread functions. */
92 static int bgp_start_timer(struct thread
*);
93 static int bgp_connect_timer(struct thread
*);
94 static int bgp_holdtime_timer(struct thread
*);
96 /* BGP FSM functions. */
97 static int bgp_start(struct peer
*);
99 /* Register peer with NHT */
100 static int bgp_peer_reg_with_nht(struct peer
*peer
)
104 if (peer
->sort
== BGP_PEER_EBGP
&& peer
->ttl
== BGP_DEFAULT_TTL
105 && !CHECK_FLAG(peer
->flags
, PEER_FLAG_DISABLE_CONNECTED_CHECK
)
106 && !CHECK_FLAG(peer
->bgp
->flags
, BGP_FLAG_DISABLE_NH_CONNECTED_CHK
))
109 return bgp_find_or_add_nexthop(
110 peer
->bgp
, peer
->bgp
, family2afi(peer
->su
.sa
.sa_family
),
111 NULL
, peer
, connected
);
114 static void peer_xfer_stats(struct peer
*peer_dst
, struct peer
*peer_src
)
116 /* Copy stats over. These are only the pre-established state stats */
117 peer_dst
->open_in
+= peer_src
->open_in
;
118 peer_dst
->open_out
+= peer_src
->open_out
;
119 peer_dst
->keepalive_in
+= peer_src
->keepalive_in
;
120 peer_dst
->keepalive_out
+= peer_src
->keepalive_out
;
121 peer_dst
->notify_in
+= peer_src
->notify_in
;
122 peer_dst
->notify_out
+= peer_src
->notify_out
;
123 peer_dst
->dynamic_cap_in
+= peer_src
->dynamic_cap_in
;
124 peer_dst
->dynamic_cap_out
+= peer_src
->dynamic_cap_out
;
127 static struct peer
*peer_xfer_conn(struct peer
*from_peer
)
134 unsigned char last_evt
, last_maj_evt
;
136 assert(from_peer
!= NULL
);
138 peer
= from_peer
->doppelganger
;
140 if (!peer
|| !CHECK_FLAG(peer
->flags
, PEER_FLAG_CONFIG_NODE
))
144 * Let's check that we are not going to loose known configuration
145 * state based upon doppelganger rules.
147 FOREACH_AFI_SAFI (afi
, safi
) {
148 if (from_peer
->afc
[afi
][safi
] != peer
->afc
[afi
][safi
]) {
150 EC_BGP_DOPPELGANGER_CONFIG
,
151 "from_peer->afc[%d][%d] is not the same as what we are overwriting",
157 if (bgp_debug_neighbor_events(peer
))
158 zlog_debug("%s: peer transfer %p fd %d -> %p fd %d)",
159 from_peer
->host
, from_peer
, from_peer
->fd
, peer
,
162 bgp_writes_off(peer
);
164 bgp_writes_off(from_peer
);
165 bgp_reads_off(from_peer
);
168 * Before exchanging FD remove doppelganger from
169 * keepalive peer hash. It could be possible conf peer
170 * fd is set to -1. If blocked on lock then keepalive
171 * thread can access peer pointer with fd -1.
173 bgp_keepalives_off(from_peer
);
175 BGP_TIMER_OFF(peer
->t_routeadv
);
176 BGP_TIMER_OFF(peer
->t_connect
);
177 BGP_TIMER_OFF(peer
->t_connect_check_r
);
178 BGP_TIMER_OFF(peer
->t_connect_check_w
);
179 BGP_TIMER_OFF(from_peer
->t_routeadv
);
180 BGP_TIMER_OFF(from_peer
->t_connect
);
181 BGP_TIMER_OFF(from_peer
->t_connect_check_r
);
182 BGP_TIMER_OFF(from_peer
->t_connect_check_w
);
183 BGP_TIMER_OFF(from_peer
->t_process_packet
);
186 * At this point in time, it is possible that there are packets pending
187 * on various buffers. Those need to be transferred or dropped,
188 * otherwise we'll get spurious failures during session establishment.
190 frr_with_mutex(&peer
->io_mtx
, &from_peer
->io_mtx
) {
192 peer
->fd
= from_peer
->fd
;
195 stream_fifo_clean(peer
->ibuf
);
196 stream_fifo_clean(peer
->obuf
);
199 * this should never happen, since bgp_process_packet() is the
200 * only task that sets and unsets the current packet and it
201 * runs in our pthread.
206 "[%s] Dropping pending packet on connection transfer:",
208 /* there used to be a bgp_packet_dump call here, but
209 * that's extremely confusing since there's no way to
210 * identify the packet in MRT dumps or BMP as dropped
211 * due to connection transfer.
213 stream_free(peer
->curr
);
217 // copy each packet from old peer's output queue to new peer
218 while (from_peer
->obuf
->head
)
219 stream_fifo_push(peer
->obuf
,
220 stream_fifo_pop(from_peer
->obuf
));
222 // copy each packet from old peer's input queue to new peer
223 while (from_peer
->ibuf
->head
)
224 stream_fifo_push(peer
->ibuf
,
225 stream_fifo_pop(from_peer
->ibuf
));
227 ringbuf_wipe(peer
->ibuf_work
);
228 ringbuf_copy(peer
->ibuf_work
, from_peer
->ibuf_work
,
229 ringbuf_remain(from_peer
->ibuf_work
));
232 peer
->as
= from_peer
->as
;
233 peer
->v_holdtime
= from_peer
->v_holdtime
;
234 peer
->v_keepalive
= from_peer
->v_keepalive
;
235 peer
->v_routeadv
= from_peer
->v_routeadv
;
236 peer
->v_gr_restart
= from_peer
->v_gr_restart
;
237 peer
->cap
= from_peer
->cap
;
238 status
= peer
->status
;
239 pstatus
= peer
->ostatus
;
240 last_evt
= peer
->last_event
;
241 last_maj_evt
= peer
->last_major_event
;
242 peer
->status
= from_peer
->status
;
243 peer
->ostatus
= from_peer
->ostatus
;
244 peer
->last_event
= from_peer
->last_event
;
245 peer
->last_major_event
= from_peer
->last_major_event
;
246 from_peer
->status
= status
;
247 from_peer
->ostatus
= pstatus
;
248 from_peer
->last_event
= last_evt
;
249 from_peer
->last_major_event
= last_maj_evt
;
250 peer
->remote_id
= from_peer
->remote_id
;
251 peer
->last_reset
= from_peer
->last_reset
;
253 peer
->peer_gr_present_state
= from_peer
->peer_gr_present_state
;
254 peer
->peer_gr_new_status_flag
= from_peer
->peer_gr_new_status_flag
;
255 bgp_peer_gr_flags_update(peer
);
257 BGP_GR_ROUTER_DETECT_AND_SEND_CAPABILITY_TO_ZEBRA(peer
->bgp
,
260 if (bgp_peer_gr_mode_get(peer
) == PEER_DISABLE
) {
262 UNSET_FLAG(peer
->sflags
, PEER_STATUS_NSF_MODE
);
264 if (CHECK_FLAG(peer
->sflags
, PEER_STATUS_NSF_WAIT
)) {
269 if (from_peer
->hostname
!= NULL
) {
270 if (peer
->hostname
) {
271 XFREE(MTYPE_BGP_PEER_HOST
, peer
->hostname
);
272 peer
->hostname
= NULL
;
275 peer
->hostname
= from_peer
->hostname
;
276 from_peer
->hostname
= NULL
;
279 if (from_peer
->domainname
!= NULL
) {
280 if (peer
->domainname
) {
281 XFREE(MTYPE_BGP_PEER_HOST
, peer
->domainname
);
282 peer
->domainname
= NULL
;
285 peer
->domainname
= from_peer
->domainname
;
286 from_peer
->domainname
= NULL
;
289 FOREACH_AFI_SAFI (afi
, safi
) {
290 peer
->af_flags
[afi
][safi
] = from_peer
->af_flags
[afi
][safi
];
291 peer
->af_sflags
[afi
][safi
] = from_peer
->af_sflags
[afi
][safi
];
292 peer
->af_cap
[afi
][safi
] = from_peer
->af_cap
[afi
][safi
];
293 peer
->afc_nego
[afi
][safi
] = from_peer
->afc_nego
[afi
][safi
];
294 peer
->afc_adv
[afi
][safi
] = from_peer
->afc_adv
[afi
][safi
];
295 peer
->afc_recv
[afi
][safi
] = from_peer
->afc_recv
[afi
][safi
];
296 peer
->orf_plist
[afi
][safi
] = from_peer
->orf_plist
[afi
][safi
];
299 if (bgp_getsockname(peer
) < 0) {
302 "%%bgp_getsockname() failed for %s peer %s fd %d (from_peer fd %d)",
303 (CHECK_FLAG(peer
->sflags
, PEER_STATUS_ACCEPT_PEER
)
306 peer
->host
, peer
->fd
, from_peer
->fd
);
311 if (from_peer
->status
> Active
) {
312 if (bgp_getsockname(from_peer
) < 0) {
315 "%%bgp_getsockname() failed for %s from_peer %s fd %d (peer fd %d)",
317 (CHECK_FLAG(from_peer
->sflags
,
318 PEER_STATUS_ACCEPT_PEER
)
321 from_peer
->host
, from_peer
->fd
, peer
->fd
);
328 // Note: peer_xfer_stats() must be called with I/O turned OFF
330 peer_xfer_stats(peer
, from_peer
);
332 /* Register peer for NHT. This is to allow RAs to be enabled when
333 * needed, even on a passive connection.
335 bgp_peer_reg_with_nht(peer
);
339 thread_add_timer_msec(bm
->master
, bgp_process_packet
, peer
, 0,
340 &peer
->t_process_packet
);
345 /* Hook function called after bgp event is occered. And vty's
346 neighbor command invoke this function after making neighbor
348 void bgp_timer_set(struct peer
*peer
)
350 switch (peer
->status
) {
352 /* First entry point of peer's finite state machine. In Idle
353 status start timer is on unless peer is shutdown or peer is
354 inactive. All other timer must be turned off */
355 if (BGP_PEER_START_SUPPRESSED(peer
) || !peer_active(peer
)
356 || (peer
->bgp
->inst_type
!= BGP_INSTANCE_TYPE_VIEW
&&
357 peer
->bgp
->vrf_id
== VRF_UNKNOWN
)) {
358 BGP_TIMER_OFF(peer
->t_start
);
360 BGP_TIMER_ON(peer
->t_start
, bgp_start_timer
,
363 BGP_TIMER_OFF(peer
->t_connect
);
364 BGP_TIMER_OFF(peer
->t_holdtime
);
365 bgp_keepalives_off(peer
);
366 BGP_TIMER_OFF(peer
->t_routeadv
);
370 /* After start timer is expired, the peer moves to Connect
371 status. Make sure start timer is off and connect timer is
373 BGP_TIMER_OFF(peer
->t_start
);
374 BGP_TIMER_ON(peer
->t_connect
, bgp_connect_timer
,
376 BGP_TIMER_OFF(peer
->t_holdtime
);
377 bgp_keepalives_off(peer
);
378 BGP_TIMER_OFF(peer
->t_routeadv
);
382 /* Active is waiting connection from remote peer. And if
383 connect timer is expired, change status to Connect. */
384 BGP_TIMER_OFF(peer
->t_start
);
385 /* If peer is passive mode, do not set connect timer. */
386 if (CHECK_FLAG(peer
->flags
, PEER_FLAG_PASSIVE
)
387 || CHECK_FLAG(peer
->sflags
, PEER_STATUS_NSF_WAIT
)) {
388 BGP_TIMER_OFF(peer
->t_connect
);
390 BGP_TIMER_ON(peer
->t_connect
, bgp_connect_timer
,
393 BGP_TIMER_OFF(peer
->t_holdtime
);
394 bgp_keepalives_off(peer
);
395 BGP_TIMER_OFF(peer
->t_routeadv
);
399 /* OpenSent status. */
400 BGP_TIMER_OFF(peer
->t_start
);
401 BGP_TIMER_OFF(peer
->t_connect
);
402 if (peer
->v_holdtime
!= 0) {
403 BGP_TIMER_ON(peer
->t_holdtime
, bgp_holdtime_timer
,
406 BGP_TIMER_OFF(peer
->t_holdtime
);
408 bgp_keepalives_off(peer
);
409 BGP_TIMER_OFF(peer
->t_routeadv
);
413 /* OpenConfirm status. */
414 BGP_TIMER_OFF(peer
->t_start
);
415 BGP_TIMER_OFF(peer
->t_connect
);
417 /* If the negotiated Hold Time value is zero, then the Hold Time
418 timer and KeepAlive timers are not started. */
419 if (peer
->v_holdtime
== 0) {
420 BGP_TIMER_OFF(peer
->t_holdtime
);
421 bgp_keepalives_off(peer
);
423 BGP_TIMER_ON(peer
->t_holdtime
, bgp_holdtime_timer
,
425 bgp_keepalives_on(peer
);
427 BGP_TIMER_OFF(peer
->t_routeadv
);
431 /* In Established status start and connect timer is turned
433 BGP_TIMER_OFF(peer
->t_start
);
434 BGP_TIMER_OFF(peer
->t_connect
);
436 /* Same as OpenConfirm, if holdtime is zero then both holdtime
437 and keepalive must be turned off. */
438 if (peer
->v_holdtime
== 0) {
439 BGP_TIMER_OFF(peer
->t_holdtime
);
440 bgp_keepalives_off(peer
);
442 BGP_TIMER_ON(peer
->t_holdtime
, bgp_holdtime_timer
,
444 bgp_keepalives_on(peer
);
448 BGP_TIMER_OFF(peer
->t_gr_restart
);
449 BGP_TIMER_OFF(peer
->t_gr_stale
);
450 BGP_TIMER_OFF(peer
->t_pmax_restart
);
453 BGP_TIMER_OFF(peer
->t_start
);
454 BGP_TIMER_OFF(peer
->t_connect
);
455 BGP_TIMER_OFF(peer
->t_holdtime
);
456 bgp_keepalives_off(peer
);
457 BGP_TIMER_OFF(peer
->t_routeadv
);
462 /* BGP start timer. This function set BGP_Start event to thread value
463 and process event. */
464 static int bgp_start_timer(struct thread
*thread
)
468 peer
= THREAD_ARG(thread
);
469 peer
->t_start
= NULL
;
471 if (bgp_debug_neighbor_events(peer
))
472 zlog_debug("%s [FSM] Timer (start timer expire).", peer
->host
);
474 THREAD_VAL(thread
) = BGP_Start
;
475 bgp_event(thread
); /* bgp_event unlocks peer */
480 /* BGP connect retry timer. */
481 static int bgp_connect_timer(struct thread
*thread
)
486 peer
= THREAD_ARG(thread
);
488 assert(!peer
->t_write
);
489 assert(!peer
->t_read
);
491 peer
->t_connect
= NULL
;
493 if (bgp_debug_neighbor_events(peer
))
494 zlog_debug("%s [FSM] Timer (connect timer expire)", peer
->host
);
496 if (CHECK_FLAG(peer
->sflags
, PEER_STATUS_ACCEPT_PEER
)) {
500 THREAD_VAL(thread
) = ConnectRetry_timer_expired
;
501 bgp_event(thread
); /* bgp_event unlocks peer */
508 /* BGP holdtime timer. */
509 static int bgp_holdtime_timer(struct thread
*thread
)
513 peer
= THREAD_ARG(thread
);
514 peer
->t_holdtime
= NULL
;
516 if (bgp_debug_neighbor_events(peer
))
517 zlog_debug("%s [FSM] Timer (holdtime timer expire)",
520 THREAD_VAL(thread
) = Hold_Timer_expired
;
521 bgp_event(thread
); /* bgp_event unlocks peer */
526 int bgp_routeadv_timer(struct thread
*thread
)
530 peer
= THREAD_ARG(thread
);
531 peer
->t_routeadv
= NULL
;
533 if (bgp_debug_neighbor_events(peer
))
534 zlog_debug("%s [FSM] Timer (routeadv timer expire)",
537 peer
->synctime
= bgp_clock();
539 thread_add_timer_msec(bm
->master
, bgp_generate_updgrp_packets
, peer
, 0,
540 &peer
->t_generate_updgrp_packets
);
542 /* MRAI timer will be started again when FIFO is built, no need to
548 /* BGP Peer Down Cause */
549 const char *const peer_down_str
[] = {"",
553 "Cluster ID changed",
554 "Confederation identifier changed",
555 "Confederation peer changed",
556 "RR client config change",
557 "RS client config change",
558 "Update source change",
559 "Address family activated",
562 "BGP Notification received",
563 "BGP Notification send",
564 "Peer closed the session",
566 "Peer-group add member",
567 "Peer-group delete member",
568 "Capability changed",
569 "Passive config change",
570 "Multihop config change",
571 "NSF peer closed the session",
572 "Intf peering v6only config change",
575 "Neighbor address lost",
577 "Waiting for Peer IPv6 LLA",
578 "Waiting for VRF to be initialized",
579 "No AFI/SAFI activated for peer",
580 "AS Set config change",
581 "Waiting for peer OPEN"};
583 static int bgp_graceful_restart_timer_expire(struct thread
*thread
)
589 peer
= THREAD_ARG(thread
);
590 peer
->t_gr_restart
= NULL
;
592 /* NSF delete stale route */
593 for (afi
= AFI_IP
; afi
< AFI_MAX
; afi
++)
594 for (safi
= SAFI_UNICAST
; safi
<= SAFI_MPLS_VPN
; safi
++)
595 if (peer
->nsf
[afi
][safi
])
596 bgp_clear_stale_route(peer
, afi
, safi
);
598 UNSET_FLAG(peer
->sflags
, PEER_STATUS_NSF_WAIT
);
599 BGP_TIMER_OFF(peer
->t_gr_stale
);
601 if (bgp_debug_neighbor_events(peer
)) {
602 zlog_debug("%s graceful restart timer expired", peer
->host
);
603 zlog_debug("%s graceful restart stalepath timer stopped",
612 static int bgp_graceful_stale_timer_expire(struct thread
*thread
)
618 peer
= THREAD_ARG(thread
);
619 peer
->t_gr_stale
= NULL
;
621 if (bgp_debug_neighbor_events(peer
))
622 zlog_debug("%s graceful restart stalepath timer expired",
625 /* NSF delete stale route */
626 for (afi
= AFI_IP
; afi
< AFI_MAX
; afi
++)
627 for (safi
= SAFI_UNICAST
; safi
<= SAFI_MPLS_VPN
; safi
++)
628 if (peer
->nsf
[afi
][safi
])
629 bgp_clear_stale_route(peer
, afi
, safi
);
634 /* Selection deferral timer processing function */
635 static int bgp_graceful_deferral_timer_expire(struct thread
*thread
)
637 struct afi_safi_info
*info
;
642 info
= THREAD_ARG(thread
);
647 if (BGP_DEBUG(update
, UPDATE_OUT
))
649 "afi %d, safi %d : graceful restart deferral timer expired",
652 bgp
->gr_info
[afi
][safi
].t_select_deferral
= NULL
;
654 bgp
->gr_info
[afi
][safi
].eor_required
= 0;
655 bgp
->gr_info
[afi
][safi
].eor_received
= 0;
656 XFREE(MTYPE_TMP
, info
);
658 /* Best path selection */
659 return bgp_best_path_select_defer(bgp
, afi
, safi
);
662 static int bgp_update_delay_applicable(struct bgp
*bgp
)
664 /* update_delay_over flag should be reset (set to 0) for any new
665 applicability of the update-delay during BGP process lifetime.
666 And it should be set after an occurence of the update-delay is
668 if (!bgp
->update_delay_over
)
674 int bgp_update_delay_active(struct bgp
*bgp
)
676 if (bgp
->t_update_delay
)
682 int bgp_update_delay_configured(struct bgp
*bgp
)
684 if (bgp
->v_update_delay
)
690 /* Do the post-processing needed when bgp comes out of the read-only mode
691 on ending the update delay. */
692 void bgp_update_delay_end(struct bgp
*bgp
)
694 THREAD_TIMER_OFF(bgp
->t_update_delay
);
695 THREAD_TIMER_OFF(bgp
->t_establish_wait
);
697 /* Reset update-delay related state */
698 bgp
->update_delay_over
= 1;
699 bgp
->established
= 0;
700 bgp
->restarted_peers
= 0;
701 bgp
->implicit_eors
= 0;
702 bgp
->explicit_eors
= 0;
704 quagga_timestamp(3, bgp
->update_delay_end_time
,
705 sizeof(bgp
->update_delay_end_time
));
708 * Add an end-of-initial-update marker to the main process queues so
710 * the route advertisement timer for the peers can be started. Also set
711 * the zebra and peer update hold flags. These flags are used to achieve
712 * three stages in the update-delay post processing:
713 * 1. Finish best-path selection for all the prefixes held on the
715 * (routes in BGP are updated, and peers sync queues are populated
717 * 2. As the eoiu mark is reached in the bgp process routine, ship all
719 * routes to zebra. With that zebra should see updates from BGP
722 * 3. Unblock the peer update writes. With that peer update packing
724 * the prefixes should be at its maximum.
726 bgp_add_eoiu_mark(bgp
);
727 bgp
->main_zebra_update_hold
= 1;
728 bgp
->main_peers_update_hold
= 1;
730 /* Resume the queue processing. This should trigger the event that would
732 care of processing any work that was queued during the read-only
734 work_queue_unplug(bm
->process_main_queue
);
740 void bgp_start_routeadv(struct bgp
*bgp
)
742 struct listnode
*node
, *nnode
;
745 zlog_info("bgp_start_routeadv(), update hold status %d",
746 bgp
->main_peers_update_hold
);
748 if (bgp
->main_peers_update_hold
)
751 quagga_timestamp(3, bgp
->update_delay_peers_resume_time
,
752 sizeof(bgp
->update_delay_peers_resume_time
));
754 for (ALL_LIST_ELEMENTS(bgp
->peer
, node
, nnode
, peer
)) {
755 if (peer
->status
!= Established
)
757 BGP_TIMER_OFF(peer
->t_routeadv
);
758 BGP_TIMER_ON(peer
->t_routeadv
, bgp_routeadv_timer
, 0);
765 void bgp_adjust_routeadv(struct peer
*peer
)
767 time_t nowtime
= bgp_clock();
769 unsigned long remain
;
771 /* Bypass checks for special case of MRAI being 0 */
772 if (peer
->v_routeadv
== 0) {
773 /* Stop existing timer, just in case it is running for a
775 * duration and schedule write thread immediately.
777 if (peer
->t_routeadv
)
778 BGP_TIMER_OFF(peer
->t_routeadv
);
780 peer
->synctime
= bgp_clock();
781 thread_add_timer_msec(bm
->master
, bgp_generate_updgrp_packets
,
783 &peer
->t_generate_updgrp_packets
);
790 * If the last update was written more than MRAI back, expire the timer
791 * instantly so that we can send the update out sooner.
793 * <------- MRAI --------->
794 * |-----------------|-----------------------|
795 * <------------- m ------------>
804 diff
= difftime(nowtime
, peer
->last_update
);
805 if (diff
> (double)peer
->v_routeadv
) {
806 BGP_TIMER_OFF(peer
->t_routeadv
);
807 BGP_TIMER_ON(peer
->t_routeadv
, bgp_routeadv_timer
, 0);
813 * - Find when to expire the MRAI timer.
814 * If MRAI timer is not active, assume we can start it now.
816 * <------- MRAI --------->
817 * |------------|-----------------------|
818 * <-------- m ----------><----- r ----->
827 if (peer
->t_routeadv
)
828 remain
= thread_timer_remain_second(peer
->t_routeadv
);
830 remain
= peer
->v_routeadv
;
831 diff
= peer
->v_routeadv
- diff
;
832 if (diff
<= (double)remain
) {
833 BGP_TIMER_OFF(peer
->t_routeadv
);
834 BGP_TIMER_ON(peer
->t_routeadv
, bgp_routeadv_timer
, diff
);
838 static int bgp_maxmed_onstartup_applicable(struct bgp
*bgp
)
840 if (!bgp
->maxmed_onstartup_over
)
846 int bgp_maxmed_onstartup_configured(struct bgp
*bgp
)
848 if (bgp
->v_maxmed_onstartup
!= BGP_MAXMED_ONSTARTUP_UNCONFIGURED
)
854 int bgp_maxmed_onstartup_active(struct bgp
*bgp
)
856 if (bgp
->t_maxmed_onstartup
)
862 void bgp_maxmed_update(struct bgp
*bgp
)
864 uint8_t maxmed_active
;
865 uint32_t maxmed_value
;
867 if (bgp
->v_maxmed_admin
) {
869 maxmed_value
= bgp
->maxmed_admin_value
;
870 } else if (bgp
->t_maxmed_onstartup
) {
872 maxmed_value
= bgp
->maxmed_onstartup_value
;
875 maxmed_value
= BGP_MAXMED_VALUE_DEFAULT
;
878 if (bgp
->maxmed_active
!= maxmed_active
879 || bgp
->maxmed_value
!= maxmed_value
) {
880 bgp
->maxmed_active
= maxmed_active
;
881 bgp
->maxmed_value
= maxmed_value
;
883 update_group_announce(bgp
);
887 /* The maxmed onstartup timer expiry callback. */
888 static int bgp_maxmed_onstartup_timer(struct thread
*thread
)
892 zlog_info("Max med on startup ended - timer expired.");
894 bgp
= THREAD_ARG(thread
);
895 THREAD_TIMER_OFF(bgp
->t_maxmed_onstartup
);
896 bgp
->maxmed_onstartup_over
= 1;
898 bgp_maxmed_update(bgp
);
903 static void bgp_maxmed_onstartup_begin(struct bgp
*bgp
)
905 /* Applicable only once in the process lifetime on the startup */
906 if (bgp
->maxmed_onstartup_over
)
909 zlog_info("Begin maxmed onstartup mode - timer %d seconds",
910 bgp
->v_maxmed_onstartup
);
912 thread_add_timer(bm
->master
, bgp_maxmed_onstartup_timer
, bgp
,
913 bgp
->v_maxmed_onstartup
, &bgp
->t_maxmed_onstartup
);
915 if (!bgp
->v_maxmed_admin
) {
916 bgp
->maxmed_active
= 1;
917 bgp
->maxmed_value
= bgp
->maxmed_onstartup_value
;
920 /* Route announce to all peers should happen after this in
924 static void bgp_maxmed_onstartup_process_status_change(struct peer
*peer
)
926 if (peer
->status
== Established
&& !peer
->bgp
->established
) {
927 bgp_maxmed_onstartup_begin(peer
->bgp
);
931 /* The update delay timer expiry callback. */
932 static int bgp_update_delay_timer(struct thread
*thread
)
936 zlog_info("Update delay ended - timer expired.");
938 bgp
= THREAD_ARG(thread
);
939 THREAD_TIMER_OFF(bgp
->t_update_delay
);
940 bgp_update_delay_end(bgp
);
945 /* The establish wait timer expiry callback. */
946 static int bgp_establish_wait_timer(struct thread
*thread
)
950 zlog_info("Establish wait - timer expired.");
952 bgp
= THREAD_ARG(thread
);
953 THREAD_TIMER_OFF(bgp
->t_establish_wait
);
954 bgp_check_update_delay(bgp
);
959 /* Steps to begin the update delay:
960 - initialize queues if needed
961 - stop the queue processing
963 static void bgp_update_delay_begin(struct bgp
*bgp
)
965 struct listnode
*node
, *nnode
;
968 /* Stop the processing of queued work. Enqueue shall continue */
969 work_queue_plug(bm
->process_main_queue
);
971 for (ALL_LIST_ELEMENTS(bgp
->peer
, node
, nnode
, peer
))
972 peer
->update_delay_over
= 0;
974 /* Start the update-delay timer */
975 thread_add_timer(bm
->master
, bgp_update_delay_timer
, bgp
,
976 bgp
->v_update_delay
, &bgp
->t_update_delay
);
978 if (bgp
->v_establish_wait
!= bgp
->v_update_delay
)
979 thread_add_timer(bm
->master
, bgp_establish_wait_timer
, bgp
,
980 bgp
->v_establish_wait
, &bgp
->t_establish_wait
);
982 quagga_timestamp(3, bgp
->update_delay_begin_time
,
983 sizeof(bgp
->update_delay_begin_time
));
986 static void bgp_update_delay_process_status_change(struct peer
*peer
)
988 if (peer
->status
== Established
) {
989 if (!peer
->bgp
->established
++) {
990 bgp_update_delay_begin(peer
->bgp
);
992 "Begin read-only mode - update-delay timer %d seconds",
993 peer
->bgp
->v_update_delay
);
995 if (CHECK_FLAG(peer
->cap
, PEER_CAP_RESTART_BIT_RCV
))
996 bgp_update_restarted_peers(peer
);
998 if (peer
->ostatus
== Established
999 && bgp_update_delay_active(peer
->bgp
)) {
1000 /* Adjust the update-delay state to account for this flap.
1001 NOTE: Intentionally skipping adjusting implicit_eors or
1003 counters. Extra sanity check in bgp_check_update_delay()
1005 be enough to take care of any additive discrepancy in bgp eor
1007 peer
->bgp
->established
--;
1008 peer
->update_delay_over
= 0;
1012 /* Called after event occurred, this function change status and reset
1013 read/write and timer thread. */
1014 void bgp_fsm_change_status(struct peer
*peer
, int status
)
1017 uint32_t peer_count
;
1020 peer_count
= bgp
->established_peers
;
1022 if (status
== Established
)
1023 bgp
->established_peers
++;
1024 else if ((peer
->status
== Established
) && (status
!= Established
))
1025 bgp
->established_peers
--;
1027 if (bgp_debug_neighbor_events(peer
)) {
1028 struct vrf
*vrf
= vrf_lookup_by_id(bgp
->vrf_id
);
1030 zlog_debug("%s : vrf %s(%u), Status: %s established_peers %u", __func__
,
1031 vrf
? vrf
->name
: "Unknown", bgp
->vrf_id
,
1032 lookup_msg(bgp_status_msg
, status
, NULL
),
1033 bgp
->established_peers
);
1036 /* Set to router ID to the value provided by RIB if there are no peers
1037 * in the established state and peer count did not change
1039 if ((peer_count
!= bgp
->established_peers
) &&
1040 (bgp
->established_peers
== 0))
1041 bgp_router_id_zebra_bump(bgp
->vrf_id
, NULL
);
1043 /* Transition into Clearing or Deleted must /always/ clear all routes..
1044 * (and must do so before actually changing into Deleted..
1046 if (status
>= Clearing
) {
1047 bgp_clear_route_all(peer
);
1049 /* If no route was queued for the clear-node processing,
1051 * completion event here. This is needed because if there are no
1053 * to trigger the background clear-node thread, the event won't
1055 * generated and the peer would be stuck in Clearing. Note that
1057 * event is for the peer and helps the peer transition out of
1059 * state; it should not be generated per (AFI,SAFI). The event
1061 * directly posted here without calling clear_node_complete() as
1063 * shouldn't do an extra unlock. This event will get processed
1065 * the state change that happens below, so peer will be in
1069 if (!work_queue_is_scheduled(peer
->clear_node_queue
))
1070 BGP_EVENT_ADD(peer
, Clearing_Completed
);
1073 /* Preserve old status and change into new status. */
1074 peer
->ostatus
= peer
->status
;
1075 peer
->status
= status
;
1077 /* Save event that caused status change. */
1078 peer
->last_major_event
= peer
->cur_event
;
1080 /* Operations after status change */
1081 hook_call(peer_status_changed
, peer
);
1083 if (status
== Established
)
1084 UNSET_FLAG(peer
->sflags
, PEER_STATUS_ACCEPT_PEER
);
1086 /* If max-med processing is applicable, do the necessary. */
1087 if (status
== Established
) {
1088 if (bgp_maxmed_onstartup_configured(peer
->bgp
)
1089 && bgp_maxmed_onstartup_applicable(peer
->bgp
))
1090 bgp_maxmed_onstartup_process_status_change(peer
);
1092 peer
->bgp
->maxmed_onstartup_over
= 1;
1095 /* If update-delay processing is applicable, do the necessary. */
1096 if (bgp_update_delay_configured(peer
->bgp
)
1097 && bgp_update_delay_applicable(peer
->bgp
))
1098 bgp_update_delay_process_status_change(peer
);
1100 if (bgp_debug_neighbor_events(peer
))
1101 zlog_debug("%s went from %s to %s", peer
->host
,
1102 lookup_msg(bgp_status_msg
, peer
->ostatus
, NULL
),
1103 lookup_msg(bgp_status_msg
, peer
->status
, NULL
));
1106 /* Flush the event queue and ensure the peer is shut down */
1107 static int bgp_clearing_completed(struct peer
*peer
)
1109 int rc
= bgp_stop(peer
);
1112 BGP_EVENT_FLUSH(peer
);
1117 /* Administrative BGP peer stop event. */
1118 /* May be called multiple times for the same peer */
1119 int bgp_stop(struct peer
*peer
)
1123 char orf_name
[BUFSIZ
];
1125 struct bgp
*bgp
= peer
->bgp
;
1126 struct graceful_restart_info
*gr_info
= NULL
;
1128 peer
->nsf_af_count
= 0;
1130 if (peer_dynamic_neighbor(peer
)
1131 && !(CHECK_FLAG(peer
->flags
, PEER_FLAG_DELETE
))) {
1132 if (bgp_debug_neighbor_events(peer
))
1133 zlog_debug("%s (dynamic neighbor) deleted", peer
->host
);
1138 /* Can't do this in Clearing; events are used for state transitions */
1139 if (peer
->status
!= Clearing
) {
1140 /* Delete all existing events of the peer */
1141 BGP_EVENT_FLUSH(peer
);
1144 /* Increment Dropped count. */
1145 if (peer
->status
== Established
) {
1148 /* bgp log-neighbor-changes of neighbor Down */
1149 if (CHECK_FLAG(peer
->bgp
->flags
,
1150 BGP_FLAG_LOG_NEIGHBOR_CHANGES
)) {
1151 struct vrf
*vrf
= vrf_lookup_by_id(peer
->bgp
->vrf_id
);
1154 "%%ADJCHANGE: neighbor %s(%s) in vrf %s Down %s",
1156 (peer
->hostname
) ? peer
->hostname
: "Unknown",
1157 vrf
? ((vrf
->vrf_id
!= VRF_DEFAULT
)
1161 peer_down_str
[(int)peer
->last_reset
]);
1164 /* graceful restart */
1165 if (peer
->t_gr_stale
) {
1166 BGP_TIMER_OFF(peer
->t_gr_stale
);
1167 if (bgp_debug_neighbor_events(peer
))
1169 "%s graceful restart stalepath timer stopped",
1172 if (CHECK_FLAG(peer
->sflags
, PEER_STATUS_NSF_WAIT
)) {
1173 if (bgp_debug_neighbor_events(peer
)) {
1175 "%s graceful restart timer started for %d sec",
1176 peer
->host
, peer
->v_gr_restart
);
1178 "%s graceful restart stalepath timer started for %d sec",
1179 peer
->host
, peer
->bgp
->stalepath_time
);
1181 BGP_TIMER_ON(peer
->t_gr_restart
,
1182 bgp_graceful_restart_timer_expire
,
1183 peer
->v_gr_restart
);
1184 BGP_TIMER_ON(peer
->t_gr_stale
,
1185 bgp_graceful_stale_timer_expire
,
1186 peer
->bgp
->stalepath_time
);
1188 UNSET_FLAG(peer
->sflags
, PEER_STATUS_NSF_MODE
);
1190 for (afi
= AFI_IP
; afi
< AFI_MAX
; afi
++)
1191 for (safi
= SAFI_UNICAST
; safi
<= SAFI_MPLS_VPN
;
1193 peer
->nsf
[afi
][safi
] = 0;
1196 /* If peer reset before receiving EOR, decrement EOR count and
1197 * cancel the selection deferral timer if there are no
1198 * pending EOR messages to be received
1200 if (BGP_PEER_GRACEFUL_RESTART_CAPABLE(peer
)) {
1201 FOREACH_AFI_SAFI (afi
, safi
) {
1202 if (!peer
->afc_nego
[afi
][safi
]
1203 || CHECK_FLAG(peer
->af_sflags
[afi
][safi
],
1204 PEER_STATUS_EOR_RECEIVED
))
1207 gr_info
= &bgp
->gr_info
[afi
][safi
];
1211 if (gr_info
->eor_required
)
1212 gr_info
->eor_required
--;
1214 if (BGP_DEBUG(update
, UPDATE_OUT
))
1215 zlog_debug("peer %s, EOR_required %d",
1217 gr_info
->eor_required
);
1219 /* There is no pending EOR message */
1220 if (gr_info
->eor_required
== 0) {
1222 gr_info
->t_select_deferral
);
1223 gr_info
->eor_received
= 0;
1228 /* set last reset time */
1229 peer
->resettime
= peer
->uptime
= bgp_clock();
1231 if (BGP_DEBUG(update_groups
, UPDATE_GROUPS
))
1232 zlog_debug("%s remove from all update group",
1234 update_group_remove_peer_afs(peer
);
1236 hook_call(peer_backward_transition
, peer
);
1238 /* Reset peer synctime */
1242 /* stop keepalives */
1243 bgp_keepalives_off(peer
);
1245 /* Stop read and write threads. */
1246 bgp_writes_off(peer
);
1247 bgp_reads_off(peer
);
1249 THREAD_OFF(peer
->t_connect_check_r
);
1250 THREAD_OFF(peer
->t_connect_check_w
);
1252 /* Stop all timers. */
1253 BGP_TIMER_OFF(peer
->t_start
);
1254 BGP_TIMER_OFF(peer
->t_connect
);
1255 BGP_TIMER_OFF(peer
->t_holdtime
);
1256 BGP_TIMER_OFF(peer
->t_routeadv
);
1258 /* Clear input and output buffer. */
1259 frr_with_mutex(&peer
->io_mtx
) {
1261 stream_fifo_clean(peer
->ibuf
);
1263 stream_fifo_clean(peer
->obuf
);
1265 if (peer
->ibuf_work
)
1266 ringbuf_wipe(peer
->ibuf_work
);
1267 if (peer
->obuf_work
)
1268 stream_reset(peer
->obuf_work
);
1271 stream_free(peer
->curr
);
1276 /* Close of file descriptor. */
1277 if (peer
->fd
>= 0) {
1282 FOREACH_AFI_SAFI (afi
, safi
) {
1283 /* Reset all negotiated variables */
1284 peer
->afc_nego
[afi
][safi
] = 0;
1285 peer
->afc_adv
[afi
][safi
] = 0;
1286 peer
->afc_recv
[afi
][safi
] = 0;
1288 /* peer address family capability flags*/
1289 peer
->af_cap
[afi
][safi
] = 0;
1291 /* peer address family status flags*/
1292 peer
->af_sflags
[afi
][safi
] = 0;
1294 /* Received ORF prefix-filter */
1295 peer
->orf_plist
[afi
][safi
] = NULL
;
1297 if ((peer
->status
== OpenConfirm
)
1298 || (peer
->status
== Established
)) {
1299 /* ORF received prefix-filter pnt */
1300 sprintf(orf_name
, "%s.%d.%d", peer
->host
, afi
, safi
);
1301 prefix_bgp_orf_remove_all(afi
, orf_name
);
1305 /* Reset keepalive and holdtime */
1306 if (CHECK_FLAG(peer
->flags
, PEER_FLAG_TIMER
)) {
1307 peer
->v_keepalive
= peer
->keepalive
;
1308 peer
->v_holdtime
= peer
->holdtime
;
1310 peer
->v_keepalive
= peer
->bgp
->default_keepalive
;
1311 peer
->v_holdtime
= peer
->bgp
->default_holdtime
;
1314 peer
->update_time
= 0;
1316 /* Until we are sure that there is no problem about prefix count
1317 this should be commented out.*/
1319 /* Reset prefix count */
1320 peer
->pcount
[AFI_IP
][SAFI_UNICAST
] = 0;
1321 peer
->pcount
[AFI_IP
][SAFI_MULTICAST
] = 0;
1322 peer
->pcount
[AFI_IP
][SAFI_LABELED_UNICAST
] = 0;
1323 peer
->pcount
[AFI_IP
][SAFI_MPLS_VPN
] = 0;
1324 peer
->pcount
[AFI_IP6
][SAFI_UNICAST
] = 0;
1325 peer
->pcount
[AFI_IP6
][SAFI_MULTICAST
] = 0;
1326 peer
->pcount
[AFI_IP6
][SAFI_LABELED_UNICAST
] = 0;
1329 if (!CHECK_FLAG(peer
->flags
, PEER_FLAG_CONFIG_NODE
)
1330 && !(CHECK_FLAG(peer
->flags
, PEER_FLAG_DELETE
))) {
1334 bgp_peer_conf_if_to_su_update(peer
);
1339 /* BGP peer is stoped by the error. */
1340 static int bgp_stop_with_error(struct peer
*peer
)
1342 /* Double start timer. */
1345 /* Overflow check. */
1346 if (peer
->v_start
>= (60 * 2))
1347 peer
->v_start
= (60 * 2);
1349 if (peer_dynamic_neighbor(peer
)) {
1350 if (bgp_debug_neighbor_events(peer
))
1351 zlog_debug("%s (dynamic neighbor) deleted", peer
->host
);
1356 return (bgp_stop(peer
));
1360 /* something went wrong, send notify and tear down */
1361 static int bgp_stop_with_notify(struct peer
*peer
, uint8_t code
,
1364 /* Send notify to remote peer */
1365 bgp_notify_send(peer
, code
, sub_code
);
1367 if (peer_dynamic_neighbor(peer
)) {
1368 if (bgp_debug_neighbor_events(peer
))
1369 zlog_debug("%s (dynamic neighbor) deleted", peer
->host
);
1374 /* Clear start timer value to default. */
1375 peer
->v_start
= BGP_INIT_START_TIMER
;
1377 return (bgp_stop(peer
));
1381 * Determines whether a TCP session has successfully established for a peer and
1382 * events as appropriate.
1384 * This function is called when setting up a new session. After connect() is
1385 * called on the peer's socket (in bgp_start()), the fd is passed to poll()
1386 * to wait for connection success or failure. When poll() returns, this
1387 * function is called to evaluate the result.
1389 * Due to differences in behavior of poll() on Linux and BSD - specifically,
1390 * the value of .revents in the case of a closed connection - this function is
1391 * scheduled both for a read and a write event. The write event is triggered
1392 * when the connection is established. A read event is triggered when the
1393 * connection is closed. Thus we need to cancel whichever one did not occur.
1395 static int bgp_connect_check(struct thread
*thread
)
1402 peer
= THREAD_ARG(thread
);
1403 assert(!CHECK_FLAG(peer
->thread_flags
, PEER_THREAD_READS_ON
));
1404 assert(!CHECK_FLAG(peer
->thread_flags
, PEER_THREAD_WRITES_ON
));
1405 assert(!peer
->t_read
);
1406 assert(!peer
->t_write
);
1408 THREAD_OFF(peer
->t_connect_check_r
);
1409 THREAD_OFF(peer
->t_connect_check_w
);
1411 /* Check file descriptor. */
1412 slen
= sizeof(status
);
1413 ret
= getsockopt(peer
->fd
, SOL_SOCKET
, SO_ERROR
, (void *)&status
,
1416 /* If getsockopt is fail, this is fatal error. */
1418 zlog_err("can't get sockopt for nonblocking connect: %d(%s)",
1419 errno
, safe_strerror(errno
));
1420 BGP_EVENT_ADD(peer
, TCP_fatal_error
);
1424 /* When status is 0 then TCP connection is established. */
1426 BGP_EVENT_ADD(peer
, TCP_connection_open
);
1429 if (bgp_debug_neighbor_events(peer
))
1430 zlog_debug("%s [Event] Connect failed %d(%s)",
1431 peer
->host
, status
, safe_strerror(status
));
1432 BGP_EVENT_ADD(peer
, TCP_connection_open_failed
);
1437 /* TCP connection open. Next we send open message to remote peer. And
1438 add read thread for reading open message. */
1439 static int bgp_connect_success(struct peer
*peer
)
1442 flog_err(EC_BGP_CONNECT
,
1443 "bgp_connect_success peer's fd is negative value %d",
1449 if (bgp_getsockname(peer
) < 0) {
1450 flog_err_sys(EC_LIB_SOCKET
,
1451 "%s: bgp_getsockname(): failed for peer %s, fd %d",
1452 __FUNCTION__
, peer
->host
, peer
->fd
);
1454 peer
, BGP_NOTIFY_FSM_ERR
,
1455 BGP_NOTIFY_SUBCODE_UNSPECIFIC
); /* internal error */
1456 bgp_writes_on(peer
);
1462 if (bgp_debug_neighbor_events(peer
)) {
1463 char buf1
[SU_ADDRSTRLEN
];
1465 if (!CHECK_FLAG(peer
->sflags
, PEER_STATUS_ACCEPT_PEER
))
1466 zlog_debug("%s open active, local address %s",
1468 sockunion2str(peer
->su_local
, buf1
,
1471 zlog_debug("%s passive open", peer
->host
);
1474 bgp_open_send(peer
);
1479 /* TCP connect fail */
1480 static int bgp_connect_fail(struct peer
*peer
)
1482 if (peer_dynamic_neighbor(peer
)) {
1483 if (bgp_debug_neighbor_events(peer
))
1484 zlog_debug("%s (dynamic neighbor) deleted", peer
->host
);
1489 return (bgp_stop(peer
));
1492 /* This function is the first starting point of all BGP connection. It
1493 try to connect to remote peer with non-blocking IO. */
1494 int bgp_start(struct peer
*peer
)
1498 bgp_peer_conf_if_to_su_update(peer
);
1500 if (peer
->su
.sa
.sa_family
== AF_UNSPEC
) {
1501 if (bgp_debug_neighbor_events(peer
))
1503 "%s [FSM] Unable to get neighbor's IP address, waiting...",
1505 peer
->last_reset
= PEER_DOWN_NBR_ADDR
;
1509 if (BGP_PEER_START_SUPPRESSED(peer
)) {
1510 if (bgp_debug_neighbor_events(peer
))
1511 flog_err(EC_BGP_FSM
,
1512 "%s [FSM] Trying to start suppressed peer"
1513 " - this is never supposed to happen!",
1518 /* Scrub some information that might be left over from a previous,
1521 /* Connection information. */
1522 if (peer
->su_local
) {
1523 sockunion_free(peer
->su_local
);
1524 peer
->su_local
= NULL
;
1527 if (peer
->su_remote
) {
1528 sockunion_free(peer
->su_remote
);
1529 peer
->su_remote
= NULL
;
1532 /* Clear remote router-id. */
1533 peer
->remote_id
.s_addr
= INADDR_ANY
;
1535 /* Clear peer capability flag. */
1538 /* If the peer is passive mode, force to move to Active mode. */
1539 if (CHECK_FLAG(peer
->flags
, PEER_FLAG_PASSIVE
)) {
1540 BGP_EVENT_ADD(peer
, TCP_connection_open_failed
);
1544 if (peer
->bgp
->inst_type
!= BGP_INSTANCE_TYPE_VIEW
&&
1545 peer
->bgp
->vrf_id
== VRF_UNKNOWN
) {
1546 if (bgp_debug_neighbor_events(peer
))
1549 "%s [FSM] In a VRF that is not initialised yet",
1551 peer
->last_reset
= PEER_DOWN_VRF_UNINIT
;
1555 /* Register peer for NHT. If next hop is already resolved, proceed
1556 * with connection setup, else wait.
1558 if (!bgp_peer_reg_with_nht(peer
)) {
1559 if (bgp_zebra_num_connects()) {
1560 if (bgp_debug_neighbor_events(peer
))
1561 zlog_debug("%s [FSM] Waiting for NHT",
1563 peer
->last_reset
= PEER_DOWN_WAITING_NHT
;
1564 BGP_EVENT_ADD(peer
, TCP_connection_open_failed
);
1569 assert(!peer
->t_write
);
1570 assert(!peer
->t_read
);
1571 assert(!CHECK_FLAG(peer
->thread_flags
, PEER_THREAD_WRITES_ON
));
1572 assert(!CHECK_FLAG(peer
->thread_flags
, PEER_THREAD_READS_ON
));
1573 status
= bgp_connect(peer
);
1577 if (bgp_debug_neighbor_events(peer
))
1578 zlog_debug("%s [FSM] Connect error", peer
->host
);
1579 BGP_EVENT_ADD(peer
, TCP_connection_open_failed
);
1581 case connect_success
:
1582 if (bgp_debug_neighbor_events(peer
))
1584 "%s [FSM] Connect immediately success, fd %d",
1585 peer
->host
, peer
->fd
);
1586 BGP_EVENT_ADD(peer
, TCP_connection_open
);
1588 case connect_in_progress
:
1589 /* To check nonblocking connect, we wait until socket is
1590 readable or writable. */
1591 if (bgp_debug_neighbor_events(peer
))
1593 "%s [FSM] Non blocking connect waiting result, fd %d",
1594 peer
->host
, peer
->fd
);
1596 flog_err(EC_BGP_FSM
,
1597 "bgp_start peer's fd is negative value %d",
1602 * - when the socket becomes ready, poll() will signify POLLOUT
1603 * - if it fails to connect, poll() will signify POLLHUP
1604 * - POLLHUP is handled as a 'read' event by thread.c
1606 * therefore, we schedule both a read and a write event with
1607 * bgp_connect_check() as the handler for each and cancel the
1608 * unused event in that function.
1610 thread_add_read(bm
->master
, bgp_connect_check
, peer
, peer
->fd
,
1611 &peer
->t_connect_check_r
);
1612 thread_add_write(bm
->master
, bgp_connect_check
, peer
, peer
->fd
,
1613 &peer
->t_connect_check_w
);
1619 /* Connect retry timer is expired when the peer status is Connect. */
1620 static int bgp_reconnect(struct peer
*peer
)
1622 if (bgp_stop(peer
) < 0)
1625 /* Send graceful restart capabilty */
1626 BGP_GR_ROUTER_DETECT_AND_SEND_CAPABILITY_TO_ZEBRA(peer
->bgp
,
1633 static int bgp_fsm_open(struct peer
*peer
)
1635 /* Send keepalive and make keepalive timer */
1636 bgp_keepalive_send(peer
);
1638 /* Reset holdtimer value. */
1639 BGP_TIMER_OFF(peer
->t_holdtime
);
1644 /* FSM error, unexpected event. This is error of BGP connection. So cut the
1645 peer and change to Idle status. */
1646 static int bgp_fsm_event_error(struct peer
*peer
)
1648 flog_err(EC_BGP_FSM
, "%s [FSM] unexpected packet received in state %s",
1649 peer
->host
, lookup_msg(bgp_status_msg
, peer
->status
, NULL
));
1651 return bgp_stop_with_notify(peer
, BGP_NOTIFY_FSM_ERR
, 0);
1654 /* Hold timer expire. This is error of BGP connection. So cut the
1655 peer and change to Idle status. */
1656 static int bgp_fsm_holdtime_expire(struct peer
*peer
)
1658 if (bgp_debug_neighbor_events(peer
))
1659 zlog_debug("%s [FSM] Hold timer expire", peer
->host
);
1661 return bgp_stop_with_notify(peer
, BGP_NOTIFY_HOLD_ERR
, 0);
1664 /* Start the selection deferral timer thread for the specified AFI, SAFI */
1665 static int bgp_start_deferral_timer(struct bgp
*bgp
, afi_t afi
, safi_t safi
,
1666 struct graceful_restart_info
*gr_info
)
1668 struct afi_safi_info
*thread_info
;
1670 /* If the deferral timer is active, then increment eor count */
1671 if (gr_info
->t_select_deferral
) {
1672 gr_info
->eor_required
++;
1676 /* Start the deferral timer when the first peer enabled for the graceful
1677 * restart is established
1679 if (gr_info
->eor_required
== 0) {
1680 thread_info
= XMALLOC(MTYPE_TMP
, sizeof(struct afi_safi_info
));
1682 thread_info
->afi
= afi
;
1683 thread_info
->safi
= safi
;
1684 thread_info
->bgp
= bgp
;
1686 thread_add_timer(bm
->master
, bgp_graceful_deferral_timer_expire
,
1687 thread_info
, bgp
->select_defer_time
,
1688 &gr_info
->t_select_deferral
);
1690 gr_info
->eor_required
++;
1691 /* Send message to RIB indicating route update pending */
1692 if (gr_info
->af_enabled
[afi
][safi
] == false) {
1693 gr_info
->af_enabled
[afi
][safi
] = true;
1694 /* Send message to RIB */
1695 bgp_zebra_update(afi
, safi
, bgp
->vrf_id
,
1696 ZEBRA_CLIENT_ROUTE_UPDATE_PENDING
);
1698 if (BGP_DEBUG(update
, UPDATE_OUT
))
1699 zlog_debug("Started the deferral timer for %s eor_required %d",
1700 get_afi_safi_str(afi
, safi
, false),
1701 gr_info
->eor_required
);
1705 /* Update the graceful restart information for the specified AFI, SAFI */
1706 static int bgp_update_gr_info(struct peer
*peer
, afi_t afi
, safi_t safi
)
1708 struct graceful_restart_info
*gr_info
;
1709 struct bgp
*bgp
= peer
->bgp
;
1712 if ((afi
< AFI_IP
) || (afi
>= AFI_MAX
)) {
1713 if (BGP_DEBUG(update
, UPDATE_OUT
))
1714 zlog_debug("%s : invalid afi %d", __func__
, afi
);
1718 if ((safi
< SAFI_UNICAST
) || (safi
> SAFI_MPLS_VPN
)) {
1719 if (BGP_DEBUG(update
, UPDATE_OUT
))
1720 zlog_debug("%s : invalid safi %d", __func__
, safi
);
1724 /* Restarting router */
1725 if (BGP_PEER_GRACEFUL_RESTART_CAPABLE(peer
)
1726 && BGP_PEER_RESTARTING_MODE(peer
)) {
1727 /* Check if the forwarding state is preserved */
1728 if (CHECK_FLAG(bgp
->flags
, BGP_FLAG_GR_PRESERVE_FWD
)) {
1729 gr_info
= &(bgp
->gr_info
[afi
][safi
]);
1730 ret
= bgp_start_deferral_timer(bgp
, afi
, safi
, gr_info
);
1737 * Transition to Established state.
1739 * Convert peer from stub to full fledged peer, set some timers, and generate
1742 static int bgp_establish(struct peer
*peer
)
1746 int nsf_af_count
= 0;
1751 other
= peer
->doppelganger
;
1752 peer
= peer_xfer_conn(peer
);
1754 flog_err(EC_BGP_CONNECT
, "%%Neighbor failed in xfer_conn");
1759 ret
= 1; /* bgp_establish specific code when xfer_conn
1762 /* Reset capability open status flag. */
1763 if (!CHECK_FLAG(peer
->sflags
, PEER_STATUS_CAPABILITY_OPEN
))
1764 SET_FLAG(peer
->sflags
, PEER_STATUS_CAPABILITY_OPEN
);
1766 /* Clear start timer value to default. */
1767 peer
->v_start
= BGP_INIT_START_TIMER
;
1769 /* Increment established count. */
1770 peer
->established
++;
1771 bgp_fsm_change_status(peer
, Established
);
1773 /* bgp log-neighbor-changes of neighbor Up */
1774 if (CHECK_FLAG(peer
->bgp
->flags
, BGP_FLAG_LOG_NEIGHBOR_CHANGES
)) {
1775 struct vrf
*vrf
= vrf_lookup_by_id(peer
->bgp
->vrf_id
);
1777 "%%ADJCHANGE: neighbor %s(%s) in vrf %s Up", peer
->host
,
1778 (peer
->hostname
) ? peer
->hostname
: "Unknown",
1779 vrf
? ((vrf
->vrf_id
!= VRF_DEFAULT
) ? vrf
->name
1783 /* assign update-group/subgroup */
1784 update_group_adjust_peer_afs(peer
);
1786 /* graceful restart */
1787 UNSET_FLAG(peer
->sflags
, PEER_STATUS_NSF_WAIT
);
1788 if (bgp_debug_neighbor_events(peer
)) {
1789 if (BGP_PEER_RESTARTING_MODE(peer
))
1790 zlog_debug("peer %s BGP_RESTARTING_MODE", peer
->host
);
1791 else if (BGP_PEER_HELPER_MODE(peer
))
1792 zlog_debug("peer %s BGP_HELPER_MODE", peer
->host
);
1794 for (afi
= AFI_IP
; afi
< AFI_MAX
; afi
++)
1795 for (safi
= SAFI_UNICAST
; safi
<= SAFI_MPLS_VPN
; safi
++) {
1796 if (peer
->afc_nego
[afi
][safi
]
1797 && CHECK_FLAG(peer
->cap
, PEER_CAP_RESTART_ADV
)
1798 && CHECK_FLAG(peer
->af_cap
[afi
][safi
],
1799 PEER_CAP_RESTART_AF_RCV
)) {
1800 if (peer
->nsf
[afi
][safi
]
1802 peer
->af_cap
[afi
][safi
],
1803 PEER_CAP_RESTART_AF_PRESERVE_RCV
))
1804 bgp_clear_stale_route(peer
, afi
, safi
);
1806 peer
->nsf
[afi
][safi
] = 1;
1809 if (peer
->nsf
[afi
][safi
])
1810 bgp_clear_stale_route(peer
, afi
, safi
);
1811 peer
->nsf
[afi
][safi
] = 0;
1813 /* Update the graceful restart information */
1814 if (peer
->afc_nego
[afi
][safi
]) {
1815 if (!BGP_SELECT_DEFER_DISABLE(peer
->bgp
)) {
1816 status
= bgp_update_gr_info(peer
, afi
,
1820 "Error in updating graceful restart for %s",
1825 if (BGP_PEER_GRACEFUL_RESTART_CAPABLE(
1827 && BGP_PEER_RESTARTING_MODE(peer
)
1830 BGP_FLAG_GR_PRESERVE_FWD
))
1831 peer
->bgp
->gr_info
[afi
][safi
]
1837 peer
->nsf_af_count
= nsf_af_count
;
1840 SET_FLAG(peer
->sflags
, PEER_STATUS_NSF_MODE
);
1842 UNSET_FLAG(peer
->sflags
, PEER_STATUS_NSF_MODE
);
1843 if (peer
->t_gr_stale
) {
1844 BGP_TIMER_OFF(peer
->t_gr_stale
);
1845 if (bgp_debug_neighbor_events(peer
))
1847 "%s graceful restart stalepath timer stopped",
1852 if (peer
->t_gr_restart
) {
1853 BGP_TIMER_OFF(peer
->t_gr_restart
);
1854 if (bgp_debug_neighbor_events(peer
))
1855 zlog_debug("%s graceful restart timer stopped",
1859 /* Reset uptime, turn on keepalives, send current table. */
1860 if (!peer
->v_holdtime
)
1861 bgp_keepalives_on(peer
);
1863 peer
->uptime
= bgp_clock();
1865 /* Send route-refresh when ORF is enabled */
1866 FOREACH_AFI_SAFI (afi
, safi
) {
1867 if (CHECK_FLAG(peer
->af_cap
[afi
][safi
],
1868 PEER_CAP_ORF_PREFIX_SM_ADV
)) {
1869 if (CHECK_FLAG(peer
->af_cap
[afi
][safi
],
1870 PEER_CAP_ORF_PREFIX_RM_RCV
))
1871 bgp_route_refresh_send(peer
, afi
, safi
,
1873 REFRESH_IMMEDIATE
, 0);
1874 else if (CHECK_FLAG(peer
->af_cap
[afi
][safi
],
1875 PEER_CAP_ORF_PREFIX_RM_OLD_RCV
))
1876 bgp_route_refresh_send(peer
, afi
, safi
,
1877 ORF_TYPE_PREFIX_OLD
,
1878 REFRESH_IMMEDIATE
, 0);
1882 /* First update is deferred until ORF or ROUTE-REFRESH is received */
1883 FOREACH_AFI_SAFI (afi
, safi
) {
1884 if (CHECK_FLAG(peer
->af_cap
[afi
][safi
],
1885 PEER_CAP_ORF_PREFIX_RM_ADV
))
1886 if (CHECK_FLAG(peer
->af_cap
[afi
][safi
],
1887 PEER_CAP_ORF_PREFIX_SM_RCV
)
1888 || CHECK_FLAG(peer
->af_cap
[afi
][safi
],
1889 PEER_CAP_ORF_PREFIX_SM_OLD_RCV
))
1890 SET_FLAG(peer
->af_sflags
[afi
][safi
],
1891 PEER_STATUS_ORF_WAIT_REFRESH
);
1894 bgp_announce_peer(peer
);
1896 /* Start the route advertisement timer to send updates to the peer - if
1898 * is not in read-only mode. If it is, the timer will be started at the
1900 * of read-only mode.
1902 if (!bgp_update_delay_active(peer
->bgp
)) {
1903 BGP_TIMER_OFF(peer
->t_routeadv
);
1904 BGP_TIMER_ON(peer
->t_routeadv
, bgp_routeadv_timer
, 0);
1907 if (peer
->doppelganger
&& (peer
->doppelganger
->status
!= Deleted
)) {
1908 if (bgp_debug_neighbor_events(peer
))
1910 "[Event] Deleting stub connection for peer %s",
1913 if (peer
->doppelganger
->status
> Active
)
1914 bgp_notify_send(peer
->doppelganger
, BGP_NOTIFY_CEASE
,
1915 BGP_NOTIFY_CEASE_COLLISION_RESOLUTION
);
1917 peer_delete(peer
->doppelganger
);
1921 * If we are replacing the old peer for a doppelganger
1922 * then switch it around in the bgp->peerhash
1923 * the doppelgangers su and this peer's su are the same
1924 * so the hash_release is the same for either.
1926 hash_release(peer
->bgp
->peerhash
, peer
);
1927 hash_get(peer
->bgp
->peerhash
, peer
, hash_alloc_intern
);
1929 bgp_bfd_register_peer(peer
);
1933 /* Keepalive packet is received. */
1934 static int bgp_fsm_keepalive(struct peer
*peer
)
1936 BGP_TIMER_OFF(peer
->t_holdtime
);
1940 /* Update packet is received. */
1941 static int bgp_fsm_update(struct peer
*peer
)
1943 BGP_TIMER_OFF(peer
->t_holdtime
);
1947 /* This is empty event. */
1948 static int bgp_ignore(struct peer
*peer
)
1952 "%s [FSM] Ignoring event %s in state %s, prior events %s, %s, fd %d",
1953 peer
->host
, bgp_event_str
[peer
->cur_event
],
1954 lookup_msg(bgp_status_msg
, peer
->status
, NULL
),
1955 bgp_event_str
[peer
->last_event
],
1956 bgp_event_str
[peer
->last_major_event
], peer
->fd
);
1960 /* This is to handle unexpected events.. */
1961 static int bgp_fsm_exeption(struct peer
*peer
)
1965 "%s [FSM] Unexpected event %s in state %s, prior events %s, %s, fd %d",
1966 peer
->host
, bgp_event_str
[peer
->cur_event
],
1967 lookup_msg(bgp_status_msg
, peer
->status
, NULL
),
1968 bgp_event_str
[peer
->last_event
],
1969 bgp_event_str
[peer
->last_major_event
], peer
->fd
);
1970 return (bgp_stop(peer
));
1973 void bgp_fsm_event_update(struct peer
*peer
, int valid
)
1978 switch (peer
->status
) {
1981 BGP_EVENT_ADD(peer
, BGP_Start
);
1985 BGP_TIMER_OFF(peer
->t_connect
);
1986 BGP_EVENT_ADD(peer
, TCP_fatal_error
);
1991 BGP_TIMER_OFF(peer
->t_connect
);
1992 BGP_EVENT_ADD(peer
, ConnectRetry_timer_expired
);
1998 if (!valid
&& (peer
->gtsm_hops
== 1))
1999 BGP_EVENT_ADD(peer
, TCP_fatal_error
);
2007 /* Finite State Machine structure */
2008 static const struct {
2009 int (*func
)(struct peer
*);
2011 } FSM
[BGP_STATUS_MAX
- 1][BGP_EVENTS_MAX
- 1] = {
2013 /* Idle state: In Idle state, all events other than BGP_Start is
2014 ignored. With BGP_Start event, finite state machine calls
2016 {bgp_start
, Connect
}, /* BGP_Start */
2017 {bgp_stop
, Idle
}, /* BGP_Stop */
2018 {bgp_stop
, Idle
}, /* TCP_connection_open */
2019 {bgp_stop
, Idle
}, /* TCP_connection_closed */
2020 {bgp_ignore
, Idle
}, /* TCP_connection_open_failed */
2021 {bgp_stop
, Idle
}, /* TCP_fatal_error */
2022 {bgp_ignore
, Idle
}, /* ConnectRetry_timer_expired */
2023 {bgp_ignore
, Idle
}, /* Hold_Timer_expired */
2024 {bgp_ignore
, Idle
}, /* KeepAlive_timer_expired */
2025 {bgp_ignore
, Idle
}, /* Receive_OPEN_message */
2026 {bgp_ignore
, Idle
}, /* Receive_KEEPALIVE_message */
2027 {bgp_ignore
, Idle
}, /* Receive_UPDATE_message */
2028 {bgp_ignore
, Idle
}, /* Receive_NOTIFICATION_message */
2029 {bgp_ignore
, Idle
}, /* Clearing_Completed */
2033 {bgp_ignore
, Connect
}, /* BGP_Start */
2034 {bgp_stop
, Idle
}, /* BGP_Stop */
2035 {bgp_connect_success
, OpenSent
}, /* TCP_connection_open */
2036 {bgp_stop
, Idle
}, /* TCP_connection_closed */
2037 {bgp_connect_fail
, Active
}, /* TCP_connection_open_failed */
2038 {bgp_connect_fail
, Idle
}, /* TCP_fatal_error */
2039 {bgp_reconnect
, Connect
}, /* ConnectRetry_timer_expired */
2040 {bgp_fsm_exeption
, Idle
}, /* Hold_Timer_expired */
2041 {bgp_fsm_exeption
, Idle
}, /* KeepAlive_timer_expired */
2042 {bgp_fsm_exeption
, Idle
}, /* Receive_OPEN_message */
2043 {bgp_fsm_exeption
, Idle
}, /* Receive_KEEPALIVE_message */
2044 {bgp_fsm_exeption
, Idle
}, /* Receive_UPDATE_message */
2045 {bgp_stop
, Idle
}, /* Receive_NOTIFICATION_message */
2046 {bgp_fsm_exeption
, Idle
}, /* Clearing_Completed */
2050 {bgp_ignore
, Active
}, /* BGP_Start */
2051 {bgp_stop
, Idle
}, /* BGP_Stop */
2052 {bgp_connect_success
, OpenSent
}, /* TCP_connection_open */
2053 {bgp_stop
, Idle
}, /* TCP_connection_closed */
2054 {bgp_ignore
, Active
}, /* TCP_connection_open_failed */
2055 {bgp_fsm_exeption
, Idle
}, /* TCP_fatal_error */
2056 {bgp_start
, Connect
}, /* ConnectRetry_timer_expired */
2057 {bgp_fsm_exeption
, Idle
}, /* Hold_Timer_expired */
2058 {bgp_fsm_exeption
, Idle
}, /* KeepAlive_timer_expired */
2059 {bgp_fsm_exeption
, Idle
}, /* Receive_OPEN_message */
2060 {bgp_fsm_exeption
, Idle
}, /* Receive_KEEPALIVE_message */
2061 {bgp_fsm_exeption
, Idle
}, /* Receive_UPDATE_message */
2062 {bgp_fsm_exeption
, Idle
}, /* Receive_NOTIFICATION_message */
2063 {bgp_fsm_exeption
, Idle
}, /* Clearing_Completed */
2067 {bgp_ignore
, OpenSent
}, /* BGP_Start */
2068 {bgp_stop
, Idle
}, /* BGP_Stop */
2069 {bgp_stop
, Active
}, /* TCP_connection_open */
2070 {bgp_stop
, Active
}, /* TCP_connection_closed */
2071 {bgp_stop
, Active
}, /* TCP_connection_open_failed */
2072 {bgp_stop
, Active
}, /* TCP_fatal_error */
2073 {bgp_fsm_exeption
, Idle
}, /* ConnectRetry_timer_expired */
2074 {bgp_fsm_holdtime_expire
, Idle
}, /* Hold_Timer_expired */
2075 {bgp_fsm_exeption
, Idle
}, /* KeepAlive_timer_expired */
2076 {bgp_fsm_open
, OpenConfirm
}, /* Receive_OPEN_message */
2077 {bgp_fsm_event_error
, Idle
}, /* Receive_KEEPALIVE_message */
2078 {bgp_fsm_event_error
, Idle
}, /* Receive_UPDATE_message */
2079 {bgp_fsm_event_error
, Idle
}, /* Receive_NOTIFICATION_message */
2080 {bgp_fsm_exeption
, Idle
}, /* Clearing_Completed */
2084 {bgp_ignore
, OpenConfirm
}, /* BGP_Start */
2085 {bgp_stop
, Idle
}, /* BGP_Stop */
2086 {bgp_stop
, Idle
}, /* TCP_connection_open */
2087 {bgp_stop
, Idle
}, /* TCP_connection_closed */
2088 {bgp_stop
, Idle
}, /* TCP_connection_open_failed */
2089 {bgp_stop
, Idle
}, /* TCP_fatal_error */
2090 {bgp_fsm_exeption
, Idle
}, /* ConnectRetry_timer_expired */
2091 {bgp_fsm_holdtime_expire
, Idle
}, /* Hold_Timer_expired */
2092 {bgp_ignore
, OpenConfirm
}, /* KeepAlive_timer_expired */
2093 {bgp_fsm_exeption
, Idle
}, /* Receive_OPEN_message */
2094 {bgp_establish
, Established
}, /* Receive_KEEPALIVE_message */
2095 {bgp_fsm_exeption
, Idle
}, /* Receive_UPDATE_message */
2096 {bgp_stop_with_error
, Idle
}, /* Receive_NOTIFICATION_message */
2097 {bgp_fsm_exeption
, Idle
}, /* Clearing_Completed */
2101 {bgp_ignore
, Established
}, /* BGP_Start */
2102 {bgp_stop
, Clearing
}, /* BGP_Stop */
2103 {bgp_stop
, Clearing
}, /* TCP_connection_open */
2104 {bgp_stop
, Clearing
}, /* TCP_connection_closed */
2105 {bgp_stop
, Clearing
}, /* TCP_connection_open_failed */
2106 {bgp_stop
, Clearing
}, /* TCP_fatal_error */
2107 {bgp_stop
, Clearing
}, /* ConnectRetry_timer_expired */
2108 {bgp_fsm_holdtime_expire
, Clearing
}, /* Hold_Timer_expired */
2109 {bgp_ignore
, Established
}, /* KeepAlive_timer_expired */
2110 {bgp_stop
, Clearing
}, /* Receive_OPEN_message */
2112 Established
}, /* Receive_KEEPALIVE_message */
2113 {bgp_fsm_update
, Established
}, /* Receive_UPDATE_message */
2114 {bgp_stop_with_error
,
2115 Clearing
}, /* Receive_NOTIFICATION_message */
2116 {bgp_fsm_exeption
, Idle
}, /* Clearing_Completed */
2120 {bgp_ignore
, Clearing
}, /* BGP_Start */
2121 {bgp_stop
, Clearing
}, /* BGP_Stop */
2122 {bgp_stop
, Clearing
}, /* TCP_connection_open */
2123 {bgp_stop
, Clearing
}, /* TCP_connection_closed */
2124 {bgp_stop
, Clearing
}, /* TCP_connection_open_failed */
2125 {bgp_stop
, Clearing
}, /* TCP_fatal_error */
2126 {bgp_stop
, Clearing
}, /* ConnectRetry_timer_expired */
2127 {bgp_stop
, Clearing
}, /* Hold_Timer_expired */
2128 {bgp_stop
, Clearing
}, /* KeepAlive_timer_expired */
2129 {bgp_stop
, Clearing
}, /* Receive_OPEN_message */
2130 {bgp_stop
, Clearing
}, /* Receive_KEEPALIVE_message */
2131 {bgp_stop
, Clearing
}, /* Receive_UPDATE_message */
2132 {bgp_stop
, Clearing
}, /* Receive_NOTIFICATION_message */
2133 {bgp_clearing_completed
, Idle
}, /* Clearing_Completed */
2137 {bgp_ignore
, Deleted
}, /* BGP_Start */
2138 {bgp_ignore
, Deleted
}, /* BGP_Stop */
2139 {bgp_ignore
, Deleted
}, /* TCP_connection_open */
2140 {bgp_ignore
, Deleted
}, /* TCP_connection_closed */
2141 {bgp_ignore
, Deleted
}, /* TCP_connection_open_failed */
2142 {bgp_ignore
, Deleted
}, /* TCP_fatal_error */
2143 {bgp_ignore
, Deleted
}, /* ConnectRetry_timer_expired */
2144 {bgp_ignore
, Deleted
}, /* Hold_Timer_expired */
2145 {bgp_ignore
, Deleted
}, /* KeepAlive_timer_expired */
2146 {bgp_ignore
, Deleted
}, /* Receive_OPEN_message */
2147 {bgp_ignore
, Deleted
}, /* Receive_KEEPALIVE_message */
2148 {bgp_ignore
, Deleted
}, /* Receive_UPDATE_message */
2149 {bgp_ignore
, Deleted
}, /* Receive_NOTIFICATION_message */
2150 {bgp_ignore
, Deleted
}, /* Clearing_Completed */
2154 /* Execute event process. */
2155 int bgp_event(struct thread
*thread
)
2161 peer
= THREAD_ARG(thread
);
2162 event
= THREAD_VAL(thread
);
2164 ret
= bgp_event_update(peer
, event
);
2169 int bgp_event_update(struct peer
*peer
, int event
)
2174 int passive_conn
= 0;
2177 /* default return code */
2178 ret
= FSM_PEER_NOOP
;
2180 other
= peer
->doppelganger
;
2182 (CHECK_FLAG(peer
->sflags
, PEER_STATUS_ACCEPT_PEER
)) ? 1 : 0;
2183 dyn_nbr
= peer_dynamic_neighbor(peer
);
2185 /* Logging this event. */
2186 next
= FSM
[peer
->status
- 1][event
- 1].next_state
;
2188 if (bgp_debug_neighbor_events(peer
) && peer
->status
!= next
)
2189 zlog_debug("%s [FSM] %s (%s->%s), fd %d", peer
->host
,
2190 bgp_event_str
[event
],
2191 lookup_msg(bgp_status_msg
, peer
->status
, NULL
),
2192 lookup_msg(bgp_status_msg
, next
, NULL
), peer
->fd
);
2194 peer
->last_event
= peer
->cur_event
;
2195 peer
->cur_event
= event
;
2197 /* Call function. */
2198 if (FSM
[peer
->status
- 1][event
- 1].func
)
2199 ret
= (*(FSM
[peer
->status
- 1][event
- 1].func
))(peer
);
2202 if (ret
== 1 && next
== Established
) {
2203 /* The case when doppelganger swap accurred in
2205 Update the peer pointer accordingly */
2206 ret
= FSM_PEER_TRANSFERRED
;
2210 /* If status is changed. */
2211 if (next
!= peer
->status
) {
2212 bgp_fsm_change_status(peer
, next
);
2215 * If we're going to ESTABLISHED then we executed a
2216 * peer transfer. In this case we can either return
2217 * FSM_PEER_TRANSITIONED or FSM_PEER_TRANSFERRED.
2218 * Opting for TRANSFERRED since transfer implies
2219 * session establishment.
2221 if (ret
!= FSM_PEER_TRANSFERRED
)
2222 ret
= FSM_PEER_TRANSITIONED
;
2225 /* Make sure timer is set. */
2226 bgp_timer_set(peer
);
2230 * If we got a return value of -1, that means there was an
2231 * error, restart the FSM. Since bgp_stop() was called on the
2232 * peer. only a few fields are safe to access here. In any case
2233 * we need to indicate that the peer was stopped in the return
2236 if (!dyn_nbr
&& !passive_conn
&& peer
->bgp
) {
2239 "%s [FSM] Failure handling event %s in state %s, "
2240 "prior events %s, %s, fd %d",
2241 peer
->host
, bgp_event_str
[peer
->cur_event
],
2242 lookup_msg(bgp_status_msg
, peer
->status
, NULL
),
2243 bgp_event_str
[peer
->last_event
],
2244 bgp_event_str
[peer
->last_major_event
],
2247 bgp_fsm_change_status(peer
, Idle
);
2248 bgp_timer_set(peer
);
2250 ret
= FSM_PEER_STOPPED
;
2257 int bgp_gr_lookup_n_update_all_peer(struct bgp
*bgp
,
2258 enum global_mode global_new_state
,
2259 enum global_mode global_old_state
)
2261 struct peer
*peer
= {0};
2262 struct listnode
*node
= {0};
2263 struct listnode
*nnode
= {0};
2264 enum peer_mode peer_old_state
= PEER_INVALID
;
2266 for (ALL_LIST_ELEMENTS(bgp
->peer
, node
, nnode
, peer
)) {
2268 if (BGP_DEBUG(graceful_restart
, GRACEFUL_RESTART
))
2269 zlog_debug("%s [BGP_GR] Peer: (%s) :", __func__
,
2272 peer_old_state
= bgp_peer_gr_mode_get(peer
);
2274 if (peer_old_state
== PEER_GLOBAL_INHERIT
) {
2277 *Reset only these peers and send a
2278 *new open message with the change capabilities.
2279 *Considering the mode to be "global_new_state" and
2280 *do all operation accordingly
2283 switch (global_new_state
) {
2285 BGP_PEER_GR_HELPER_ENABLE(peer
);
2288 BGP_PEER_GR_ENABLE(peer
);
2290 case GLOBAL_DISABLE
:
2291 BGP_PEER_GR_DISABLE(peer
);
2293 case GLOBAL_INVALID
:
2294 zlog_debug("%s [BGP_GR] GLOBAL_INVALID",
2296 return BGP_ERR_GR_OPERATION_FAILED
;
2301 bgp
->global_gr_present_state
= global_new_state
;
2303 return BGP_GR_SUCCESS
;
2306 int bgp_gr_update_all(struct bgp
*bgp
, int global_gr_cmd
)
2308 enum global_mode global_new_state
= GLOBAL_INVALID
;
2309 enum global_mode global_old_state
= GLOBAL_INVALID
;
2311 if (BGP_DEBUG(graceful_restart
, GRACEFUL_RESTART
))
2312 zlog_debug("%s [BGP_GR]START: global_gr_cmd :%s:", __func__
,
2313 print_global_gr_cmd(global_gr_cmd
));
2315 global_old_state
= bgp_global_gr_mode_get(bgp
);
2317 if (BGP_DEBUG(graceful_restart
, GRACEFUL_RESTART
))
2318 zlog_debug("[BGP_GR] global_old_gr_state :%s:",
2319 print_global_gr_mode(global_old_state
));
2321 if (global_old_state
!= GLOBAL_INVALID
) {
2323 bgp
->GLOBAL_GR_FSM
[global_old_state
][global_gr_cmd
];
2325 if (BGP_DEBUG(graceful_restart
, GRACEFUL_RESTART
))
2326 zlog_debug("[BGP_GR] global_new_gr_state :%s:",
2327 print_global_gr_mode(global_new_state
));
2329 zlog_err("%s [BGP_GR] global_old_state == GLOBAL_INVALID",
2331 return BGP_ERR_GR_OPERATION_FAILED
;
2334 if (global_new_state
== GLOBAL_INVALID
) {
2335 zlog_err("%s [BGP_GR] global_new_state == GLOBAL_INVALID",
2337 return BGP_ERR_GR_INVALID_CMD
;
2339 if (global_new_state
== global_old_state
) {
2341 if (BGP_DEBUG(graceful_restart
, GRACEFUL_RESTART
))
2343 "%s [BGP_GR] global_new_state == global_old_state :%s",
2345 print_global_gr_mode(global_new_state
));
2346 return BGP_GR_NO_OPERATION
;
2349 return bgp_gr_lookup_n_update_all_peer(bgp
, global_new_state
,
2353 const char *print_peer_gr_mode(enum peer_mode pr_mode
)
2355 const char *peer_gr_mode
= NULL
;
2359 peer_gr_mode
= "PEER_HELPER";
2362 peer_gr_mode
= "PEER_GR";
2365 peer_gr_mode
= "PEER_DISABLE";
2368 peer_gr_mode
= "PEER_INVALID";
2370 case PEER_GLOBAL_INHERIT
:
2371 peer_gr_mode
= "PEER_GLOBAL_INHERIT";
2375 return peer_gr_mode
;
2378 const char *print_peer_gr_cmd(enum peer_gr_command pr_gr_cmd
)
2380 const char *peer_gr_cmd
= NULL
;
2382 switch (pr_gr_cmd
) {
2384 peer_gr_cmd
= "PEER_GR_CMD";
2386 case NO_PEER_GR_CMD
:
2387 peer_gr_cmd
= "NO_PEER_GR_CMD";
2389 case PEER_DISABLE_CMD
:
2390 peer_gr_cmd
= "PEER_GR_CMD";
2392 case NO_PEER_DISABLE_CMD
:
2393 peer_gr_cmd
= "NO_PEER_GR_CMD";
2395 case PEER_HELPER_CMD
:
2396 peer_gr_cmd
= "PEER_HELPER_CMD";
2398 case NO_PEER_HELPER_CMD
:
2399 peer_gr_cmd
= "NO_PEER_HELPER_CMD";
2406 const char *print_global_gr_mode(enum global_mode gl_mode
)
2408 const char *global_gr_mode
= NULL
;
2412 global_gr_mode
= "GLOBAL_HELPER";
2415 global_gr_mode
= "GLOBAL_GR";
2417 case GLOBAL_DISABLE
:
2418 global_gr_mode
= "GLOBAL_DISABLE";
2420 case GLOBAL_INVALID
:
2421 global_gr_mode
= "GLOBAL_INVALID";
2425 return global_gr_mode
;
2428 const char *print_global_gr_cmd(enum global_gr_command gl_gr_cmd
)
2430 const char *global_gr_cmd
= NULL
;
2432 switch (gl_gr_cmd
) {
2434 global_gr_cmd
= "GLOBAL_GR_CMD";
2436 case NO_GLOBAL_GR_CMD
:
2437 global_gr_cmd
= "NO_GLOBAL_GR_CMD";
2439 case GLOBAL_DISABLE_CMD
:
2440 global_gr_cmd
= "GLOBAL_DISABLE_CMD";
2442 case NO_GLOBAL_DISABLE_CMD
:
2443 global_gr_cmd
= "NO_GLOBAL_DISABLE_CMD";
2447 return global_gr_cmd
;
2450 enum global_mode
bgp_global_gr_mode_get(struct bgp
*bgp
)
2452 return bgp
->global_gr_present_state
;
2455 enum peer_mode
bgp_peer_gr_mode_get(struct peer
*peer
)
2457 return peer
->peer_gr_present_state
;
2460 int bgp_neighbor_graceful_restart(struct peer
*peer
, int peer_gr_cmd
)
2462 enum peer_mode peer_new_state
= PEER_INVALID
;
2463 enum peer_mode peer_old_state
= PEER_INVALID
;
2464 struct bgp_peer_gr peer_state
;
2465 int result
= BGP_GR_FAILURE
;
2468 * fetch peer_old_state from peer structure also
2469 * fetch global_old_state from bgp structure,
2470 * peer had a back pointer to bgpo struct ;
2473 if (BGP_DEBUG(graceful_restart
, GRACEFUL_RESTART
))
2474 zlog_debug("%s [BGP_GR] START:Peer: (%s) : peer_gr_cmd :%s:",
2475 __func__
, peer
->host
,
2476 print_peer_gr_cmd(peer_gr_cmd
));
2478 peer_old_state
= bgp_peer_gr_mode_get(peer
);
2480 if (peer_old_state
== PEER_INVALID
) {
2481 zlog_debug("[BGP_GR] peer_old_state == Invalid state !");
2482 return BGP_ERR_GR_OPERATION_FAILED
;
2485 peer_state
= peer
->PEER_GR_FSM
[peer_old_state
][peer_gr_cmd
];
2486 peer_new_state
= peer_state
.next_state
;
2488 if (peer_new_state
== PEER_INVALID
) {
2490 "[BGP_GR] Invalid bgp graceful restart command used !");
2491 return BGP_ERR_GR_INVALID_CMD
;
2494 if (peer_new_state
!= peer_old_state
) {
2495 result
= peer_state
.action_fun(peer
, peer_old_state
,
2498 if (BGP_DEBUG(graceful_restart
, GRACEFUL_RESTART
))
2500 "[BGP_GR] peer_old_state == peer_new_state !");
2501 return BGP_GR_NO_OPERATION
;
2504 if (result
== BGP_GR_SUCCESS
) {
2506 /* Update the mode i.e peer_new_state into the peer structure */
2507 peer
->peer_gr_present_state
= peer_new_state
;
2508 if (BGP_DEBUG(graceful_restart
, GRACEFUL_RESTART
))
2510 "[BGP_GR] Succesfully change the state of the peer to : %s : !",
2511 print_peer_gr_mode(peer_new_state
));
2513 return BGP_GR_SUCCESS
;
2519 unsigned int bgp_peer_gr_action(struct peer
*peer
, int old_peer_state
,
2522 if (BGP_DEBUG(graceful_restart
, GRACEFUL_RESTART
))
2524 "%s [BGP_GR] Move peer from old_peer_state :%s: to new_peer_state :%s: !!!!",
2525 __func__
, print_peer_gr_mode(old_peer_state
),
2526 print_peer_gr_mode(new_peer_state
));
2528 int bgp_gr_global_mode
= GLOBAL_INVALID
;
2529 unsigned int ret
= BGP_GR_FAILURE
;
2531 if (old_peer_state
== new_peer_state
) {
2532 /* Nothing to do over here as the present and old state is the
2534 return BGP_GR_NO_OPERATION
;
2536 if ((old_peer_state
== PEER_INVALID
)
2537 || (new_peer_state
== PEER_INVALID
)) {
2538 /* something bad happend , print error message */
2539 return BGP_ERR_GR_INVALID_CMD
;
2542 bgp_gr_global_mode
= bgp_global_gr_mode_get(peer
->bgp
);
2544 if ((old_peer_state
== PEER_GLOBAL_INHERIT
)
2545 && (new_peer_state
!= PEER_GLOBAL_INHERIT
)) {
2547 /* fetch the Mode running in the Global state machine
2548 *from the bgp structure into a variable called
2552 /* Here we are checking if the
2553 *1. peer_new_state == global_mode == helper_mode
2554 *2. peer_new_state == global_mode == GR_mode
2555 *3. peer_new_state == global_mode == disabled_mode
2558 BGP_PEER_GR_GLOBAL_INHERIT_UNSET(peer
);
2560 if (new_peer_state
== bgp_gr_global_mode
) {
2561 /*This is incremental updates i.e no tear down
2562 *of the existing session
2563 *as the peer is already working in the same mode.
2565 ret
= BGP_GR_SUCCESS
;
2567 if (BGP_DEBUG(graceful_restart
, GRACEFUL_RESTART
))
2569 "[BGP_GR] Peer state changed from :%s ",
2570 print_peer_gr_mode(old_peer_state
));
2572 bgp_peer_move_to_gr_mode(peer
, new_peer_state
);
2574 ret
= BGP_GR_SUCCESS
;
2577 /* In the case below peer is going into Global inherit mode i.e.
2578 * the peer would work as the mode configured at the global level
2580 else if ((new_peer_state
== PEER_GLOBAL_INHERIT
)
2581 && (old_peer_state
!= PEER_GLOBAL_INHERIT
)) {
2582 /* Here in this case it would be destructive
2583 * in all the cases except one case when,
2584 * Global GR is configured Disabled
2585 * and present_peer_state is not disable
2588 BGP_PEER_GR_GLOBAL_INHERIT_SET(peer
);
2590 if (old_peer_state
== bgp_gr_global_mode
) {
2592 /* This is incremental updates
2593 *i.e no tear down of the existing session
2594 *as the peer is already working in the same mode.
2596 ret
= BGP_GR_SUCCESS
;
2598 /* Destructive always */
2599 /* Tear down the old session
2600 * and send the new capability
2601 * as per the bgp_gr_global_mode
2604 if (BGP_DEBUG(graceful_restart
, GRACEFUL_RESTART
))
2606 "[BGP_GR] Peer state changed from :%s",
2607 print_peer_gr_mode(old_peer_state
));
2609 bgp_peer_move_to_gr_mode(peer
, bgp_gr_global_mode
);
2611 ret
= BGP_GR_SUCCESS
;
2615 *This else case, it include all the cases except -->
2616 *(new_peer_state != Peer_Global) &&
2617 *( old_peer_state != Peer_Global )
2619 if (BGP_DEBUG(graceful_restart
, GRACEFUL_RESTART
))
2620 zlog_debug("[BGP_GR] Peer state changed from :%s",
2621 print_peer_gr_mode(old_peer_state
));
2623 bgp_peer_move_to_gr_mode(peer
, new_peer_state
);
2625 ret
= BGP_GR_SUCCESS
;
2631 inline void bgp_peer_move_to_gr_mode(struct peer
*peer
, int new_state
)
2634 int bgp_global_gr_mode
= bgp_global_gr_mode_get(peer
->bgp
);
2636 switch (new_state
) {
2638 BGP_PEER_GR_HELPER_ENABLE(peer
);
2641 BGP_PEER_GR_ENABLE(peer
);
2644 BGP_PEER_GR_DISABLE(peer
);
2646 case PEER_GLOBAL_INHERIT
:
2647 BGP_PEER_GR_GLOBAL_INHERIT_SET(peer
);
2649 if (bgp_global_gr_mode
== GLOBAL_HELPER
) {
2650 BGP_PEER_GR_HELPER_ENABLE(peer
);
2651 } else if (bgp_global_gr_mode
== GLOBAL_GR
) {
2652 BGP_PEER_GR_ENABLE(peer
);
2653 } else if (bgp_global_gr_mode
== GLOBAL_DISABLE
) {
2654 BGP_PEER_GR_DISABLE(peer
);
2657 "[BGP_GR] Default switch inherit mode ::: SOMETHING IS WRONG !!!");
2662 "[BGP_GR] Default switch mode ::: SOMETHING IS WRONG !!!");
2665 if (BGP_DEBUG(graceful_restart
, GRACEFUL_RESTART
))
2666 zlog_debug("[BGP_GR] Peer state changed --to--> : %d : !",
2670 void bgp_peer_gr_flags_update(struct peer
*peer
)
2672 if (BGP_DEBUG(graceful_restart
, GRACEFUL_RESTART
))
2673 zlog_debug("%s [BGP_GR] called !", __func__
);
2674 if (CHECK_FLAG(peer
->peer_gr_new_status_flag
,
2675 PEER_GRACEFUL_RESTART_NEW_STATE_HELPER
))
2676 SET_FLAG(peer
->flags
, PEER_FLAG_GRACEFUL_RESTART_HELPER
);
2678 UNSET_FLAG(peer
->flags
, PEER_FLAG_GRACEFUL_RESTART_HELPER
);
2679 if (BGP_DEBUG(graceful_restart
, GRACEFUL_RESTART
))
2681 "[BGP_GR] Peer %s Flag PEER_FLAG_GRACEFUL_RESTART_HELPER : %s : !",
2683 (CHECK_FLAG(peer
->flags
,
2684 PEER_FLAG_GRACEFUL_RESTART_HELPER
)
2687 if (CHECK_FLAG(peer
->peer_gr_new_status_flag
,
2688 PEER_GRACEFUL_RESTART_NEW_STATE_RESTART
))
2689 SET_FLAG(peer
->flags
, PEER_FLAG_GRACEFUL_RESTART
);
2691 UNSET_FLAG(peer
->flags
, PEER_FLAG_GRACEFUL_RESTART
);
2692 if (BGP_DEBUG(graceful_restart
, GRACEFUL_RESTART
))
2694 "[BGP_GR] Peer %s Flag PEER_FLAG_GRACEFUL_RESTART : %s : !",
2696 (CHECK_FLAG(peer
->flags
, PEER_FLAG_GRACEFUL_RESTART
)
2699 if (CHECK_FLAG(peer
->peer_gr_new_status_flag
,
2700 PEER_GRACEFUL_RESTART_NEW_STATE_INHERIT
))
2701 SET_FLAG(peer
->flags
,
2702 PEER_FLAG_GRACEFUL_RESTART_GLOBAL_INHERIT
);
2704 UNSET_FLAG(peer
->flags
,
2705 PEER_FLAG_GRACEFUL_RESTART_GLOBAL_INHERIT
);
2706 if (BGP_DEBUG(graceful_restart
, GRACEFUL_RESTART
))
2708 "[BGP_GR] Peer %s Flag PEER_FLAG_GRACEFUL_RESTART_GLOBAL_INHERIT : %s : !",
2710 (CHECK_FLAG(peer
->flags
,
2711 PEER_FLAG_GRACEFUL_RESTART_GLOBAL_INHERIT
)
2715 if (!CHECK_FLAG(peer
->flags
, PEER_FLAG_GRACEFUL_RESTART
)
2716 && !CHECK_FLAG(peer
->flags
, PEER_FLAG_GRACEFUL_RESTART_HELPER
)) {
2717 zlog_debug("[BGP_GR] Peer %s UNSET PEER_STATUS_NSF_MODE!",
2720 UNSET_FLAG(peer
->sflags
, PEER_STATUS_NSF_MODE
);
2722 if (CHECK_FLAG(peer
->sflags
, PEER_STATUS_NSF_WAIT
)) {
2724 peer_nsf_stop(peer
);
2726 "[BGP_GR] Peer %s UNSET PEER_STATUS_NSF_WAIT!",