]> git.proxmox.com Git - mirror_frr.git/blame - bgpd/bgp_network.c
Merge pull request #3370 from pguibert6WIND/default_vrf_initialization
[mirror_frr.git] / bgpd / bgp_network.c
CommitLineData
718e3744 1/* BGP network related fucntions
896014f4
DL
2 * Copyright (C) 1999 Kunihiro Ishiguro
3 *
4 * This file is part of GNU Zebra.
5 *
6 * GNU Zebra is free software; you can redistribute it and/or modify it
7 * under the terms of the GNU General Public License as published by the
8 * Free Software Foundation; either version 2, or (at your option) any
9 * later version.
10 *
11 * GNU Zebra is distributed in the hope that it will be useful, but
12 * WITHOUT ANY WARRANTY; without even the implied warranty of
13 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
14 * General Public License for more details.
15 *
16 * You should have received a copy of the GNU General Public License along
17 * with this program; see the file COPYING; if not, write to the Free Software
18 * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
19 */
718e3744 20
21#include <zebra.h>
22
23#include "thread.h"
24#include "sockunion.h"
0df7c91f 25#include "sockopt.h"
718e3744 26#include "memory.h"
27#include "log.h"
28#include "if.h"
29#include "prefix.h"
30#include "command.h"
edd7c245 31#include "privs.h"
0df7c91f 32#include "linklist.h"
fdbc8e77 33#include "network.h"
3f9c7369 34#include "queue.h"
66e373ae 35#include "hash.h"
039f3a34 36#include "filter.h"
61cf4b37 37#include "ns.h"
174482ef 38#include "lib_errors.h"
17cdd31e 39#include "nexthop.h"
718e3744 40
41#include "bgpd/bgpd.h"
1ff9a340 42#include "bgpd/bgp_open.h"
718e3744 43#include "bgpd/bgp_fsm.h"
44#include "bgpd/bgp_attr.h"
45#include "bgpd/bgp_debug.h"
14454c9f 46#include "bgpd/bgp_errors.h"
718e3744 47#include "bgpd/bgp_network.h"
17cdd31e 48#include "bgpd/bgp_zebra.h"
edd7c245 49
50extern struct zebra_privs_t bgpd_privs;
51
97896a91 52static char *bgp_get_bound_name(struct peer *peer);
f2345335 53
d023aec4 54/* BGP listening socket. */
d62a17ae 55struct bgp_listener {
56 int fd;
57 union sockunion su;
58 struct thread *thread;
61cf4b37 59 struct bgp *bgp;
d023aec4 60};
6b0655a2 61
0df7c91f
PJ
62/*
63 * Set MD5 key for the socket, for the given IPv4 peer address.
64 * If the password is NULL or zero-length, the option will be disabled.
65 */
d62a17ae 66static int bgp_md5_set_socket(int socket, union sockunion *su,
67 const char *password)
0df7c91f 68{
d62a17ae 69 int ret = -1;
70 int en = ENOSYS;
ea8b7c71 71#if HAVE_DECL_TCP_MD5SIG
d62a17ae 72 union sockunion su2;
ea8b7c71 73#endif /* HAVE_TCP_MD5SIG */
d62a17ae 74
75 assert(socket >= 0);
76
77#if HAVE_DECL_TCP_MD5SIG
78 /* Ensure there is no extraneous port information. */
79 memcpy(&su2, su, sizeof(union sockunion));
80 if (su2.sa.sa_family == AF_INET)
81 su2.sin.sin_port = 0;
82 else
83 su2.sin6.sin6_port = 0;
84 ret = sockopt_tcp_signature(socket, &su2, password);
85 en = errno;
0df7c91f 86#endif /* HAVE_TCP_MD5SIG */
0df7c91f 87
d62a17ae 88 if (ret < 0)
e50f7cfd 89 flog_warn(EC_BGP_NO_TCP_MD5,
28f22e26 90 "can't set TCP_MD5SIG option on socket %d: %s",
d62a17ae 91 socket, safe_strerror(en));
92
93 return ret;
0df7c91f
PJ
94}
95
96/* Helper for bgp_connect */
d62a17ae 97static int bgp_md5_set_connect(int socket, union sockunion *su,
98 const char *password)
0df7c91f 99{
d62a17ae 100 int ret = -1;
101
102#if HAVE_DECL_TCP_MD5SIG
6bb30c2c
DL
103 frr_elevate_privs(&bgpd_privs) {
104 ret = bgp_md5_set_socket(socket, su, password);
d62a17ae 105 }
0df7c91f 106#endif /* HAVE_TCP_MD5SIG */
d62a17ae 107
108 return ret;
0df7c91f
PJ
109}
110
d62a17ae 111static int bgp_md5_set_password(struct peer *peer, const char *password)
0df7c91f 112{
d62a17ae 113 struct listnode *node;
114 int ret = 0;
115 struct bgp_listener *listener;
116
6bb30c2c 117 frr_elevate_privs(&bgpd_privs) {
d62a17ae 118 /* Set or unset the password on the listen socket(s). Outbound
6bb30c2c 119 * connections are taken care of in bgp_connect() below.
d62a17ae 120 */
6bb30c2c
DL
121 for (ALL_LIST_ELEMENTS_RO(bm->listen_sockets, node, listener))
122 if (listener->su.sa.sa_family
123 == peer->su.sa.sa_family) {
124 ret = bgp_md5_set_socket(listener->fd,
125 &peer->su, password);
126 break;
127 }
128 }
d62a17ae 129 return ret;
0df7c91f 130}
3374bef0 131
d62a17ae 132int bgp_md5_set(struct peer *peer)
89ca90fa 133{
d62a17ae 134 /* Set the password from listen socket. */
135 return bgp_md5_set_password(peer, peer->password);
89ca90fa 136}
137
d62a17ae 138int bgp_md5_unset(struct peer *peer)
89ca90fa 139{
d62a17ae 140 /* Unset the password from listen socket. */
141 return bgp_md5_set_password(peer, NULL);
89ca90fa 142}
143
d62a17ae 144int bgp_set_socket_ttl(struct peer *peer, int bgp_sock)
ef0b0c3e 145{
d62a17ae 146 char buf[INET_ADDRSTRLEN];
147 int ret = 0;
148
149 /* In case of peer is EBGP, we should set TTL for this connection. */
150 if (!peer->gtsm_hops && (peer_sort(peer) == BGP_PEER_EBGP)) {
151 ret = sockopt_ttl(peer->su.sa.sa_family, bgp_sock, peer->ttl);
152 if (ret) {
af4c2728 153 flog_err(
450971aa 154 EC_LIB_SOCKET,
d62a17ae 155 "%s: Can't set TxTTL on peer (rtrid %s) socket, err = %d",
14454c9f
DS
156 __func__,
157 inet_ntop(AF_INET, &peer->remote_id, buf,
158 sizeof(buf)),
d62a17ae 159 errno);
160 return ret;
161 }
162 } else if (peer->gtsm_hops) {
163 /* On Linux, setting minttl without setting ttl seems to mess
164 with the
165 outgoing ttl. Therefore setting both.
166 */
167 ret = sockopt_ttl(peer->su.sa.sa_family, bgp_sock, MAXTTL);
168 if (ret) {
af4c2728 169 flog_err(
450971aa 170 EC_LIB_SOCKET,
d62a17ae 171 "%s: Can't set TxTTL on peer (rtrid %s) socket, err = %d",
14454c9f
DS
172 __func__,
173 inet_ntop(AF_INET, &peer->remote_id, buf,
174 sizeof(buf)),
d62a17ae 175 errno);
176 return ret;
177 }
178 ret = sockopt_minttl(peer->su.sa.sa_family, bgp_sock,
179 MAXTTL + 1 - peer->gtsm_hops);
180 if (ret) {
af4c2728 181 flog_err(
450971aa 182 EC_LIB_SOCKET,
d62a17ae 183 "%s: Can't set MinTTL on peer (rtrid %s) socket, err = %d",
14454c9f
DS
184 __func__,
185 inet_ntop(AF_INET, &peer->remote_id, buf,
186 sizeof(buf)),
d62a17ae 187 errno);
188 return ret;
189 }
190 }
191
192 return ret;
ef0b0c3e
DL
193}
194
8dee0396 195/*
196 * Obtain the BGP instance that the incoming connection should be processed
197 * against. This is important because more than one VRF could be using the
198 * same IP address space. The instance is got by obtaining the device to
199 * which the incoming connection is bound to. This could either be a VRF
200 * or it could be an interface, which in turn determines the VRF.
201 */
d62a17ae 202static int bgp_get_instance_for_inc_conn(int sock, struct bgp **bgp_inst)
8dee0396 203{
120c6587 204#ifndef SO_BINDTODEVICE
d62a17ae 205 /* only Linux has SO_BINDTODEVICE, but we're in Linux-specific code here
206 * anyway since the assumption is that the interface name returned by
207 * getsockopt() is useful in identifying the VRF, particularly with
208 * Linux's
209 * VRF l3master device. The whole mechanism is specific to Linux, so...
210 * when other platforms add VRF support, this will need handling here as
211 * well. (or, some restructuring) */
212 *bgp_inst = bgp_get_default();
213 return !*bgp_inst;
120c6587
DL
214
215#else
d62a17ae 216 char name[VRF_NAMSIZ + 1];
217 socklen_t name_len = VRF_NAMSIZ;
218 struct bgp *bgp;
219 int rc;
220 struct listnode *node, *nnode;
221
222 *bgp_inst = NULL;
223 name[0] = '\0';
224 rc = getsockopt(sock, SOL_SOCKET, SO_BINDTODEVICE, name, &name_len);
225 if (rc != 0) {
226#if defined(HAVE_CUMULUS)
1c50c1c0
QY
227 flog_err(EC_LIB_SOCKET,
228 "[Error] BGP SO_BINDTODEVICE get failed (%s), sock %d",
229 safe_strerror(errno), sock);
d62a17ae 230 return -1;
b5826a12 231#endif
d62a17ae 232 }
233
234 if (!strlen(name)) {
235 *bgp_inst = bgp_get_default();
236 return 0; /* default instance. */
237 }
238
239 /* First try match to instance; if that fails, check for interfaces. */
240 bgp = bgp_lookup_by_name(name);
241 if (bgp) {
242 if (!bgp->vrf_id) // unexpected
243 return -1;
244 *bgp_inst = bgp;
245 return 0;
246 }
247
248 /* TODO - This will be optimized once interfaces move into the NS */
249 for (ALL_LIST_ELEMENTS(bm->bgp, node, nnode, bgp)) {
250 struct interface *ifp;
251
252 if (bgp->inst_type == BGP_INSTANCE_TYPE_VIEW)
253 continue;
254
255 ifp = if_lookup_by_name(name, bgp->vrf_id);
256 if (ifp) {
257 *bgp_inst = bgp;
258 return 0;
259 }
260 }
261
262 /* We didn't match to either an instance or an interface. */
263 return -1;
120c6587 264#endif
8dee0396 265}
266
718e3744 267/* Accept bgp connection. */
d62a17ae 268static int bgp_accept(struct thread *thread)
718e3744 269{
d62a17ae 270 int bgp_sock;
271 int accept_sock;
272 union sockunion su;
273 struct bgp_listener *listener = THREAD_ARG(thread);
274 struct peer *peer;
275 struct peer *peer1;
276 char buf[SU_ADDRSTRLEN];
277 struct bgp *bgp = NULL;
278
279 sockunion_init(&su);
280
281 /* Register accept thread. */
282 accept_sock = THREAD_FD(thread);
283 if (accept_sock < 0) {
450971aa 284 flog_err_sys(EC_LIB_SOCKET, "accept_sock is nevative value %d",
09c866e3 285 accept_sock);
d62a17ae 286 return -1;
287 }
288 listener->thread = NULL;
61cf4b37 289
d62a17ae 290 thread_add_read(bm->master, bgp_accept, listener, accept_sock,
291 &listener->thread);
292
293 /* Accept client connection. */
294 bgp_sock = sockunion_accept(accept_sock, &su);
295 if (bgp_sock < 0) {
450971aa 296 flog_err_sys(EC_LIB_SOCKET,
09c866e3
QY
297 "[Error] BGP socket accept failed (%s)",
298 safe_strerror(errno));
d62a17ae 299 return -1;
300 }
301 set_nonblocking(bgp_sock);
302
61cf4b37
PG
303 /* Obtain BGP instance this connection is meant for.
304 * - if it is a VRF netns sock, then BGP is in listener structure
305 * - otherwise, the bgp instance need to be demultiplexed
306 */
307 if (listener->bgp)
308 bgp = listener->bgp;
309 else if (bgp_get_instance_for_inc_conn(bgp_sock, &bgp)) {
d62a17ae 310 if (bgp_debug_neighbor_events(NULL))
311 zlog_debug(
312 "[Event] Could not get instance for incoming conn from %s",
313 inet_sutop(&su, buf));
314 close(bgp_sock);
315 return -1;
316 }
317
318 /* Set socket send buffer size */
319 setsockopt_so_sendbuf(bgp_sock, BGP_SOCKET_SNDBUF_SIZE);
320
321 /* Check remote IP address */
322 peer1 = peer_lookup(bgp, &su);
323
324 if (!peer1) {
325 peer1 = peer_lookup_dynamic_neighbor(bgp, &su);
326 if (peer1) {
327 /* Dynamic neighbor has been created, let it proceed */
328 peer1->fd = bgp_sock;
329 bgp_fsm_change_status(peer1, Active);
330 BGP_TIMER_OFF(
331 peer1->t_start); /* created in peer_create() */
332
333 if (peer_active(peer1))
334 BGP_EVENT_ADD(peer1, TCP_connection_open);
335
336 return 0;
337 }
338 }
339
340 if (!peer1) {
341 if (bgp_debug_neighbor_events(NULL)) {
342 zlog_debug(
343 "[Event] %s connection rejected - not configured"
344 " and not valid for dynamic",
345 inet_sutop(&su, buf));
346 }
347 close(bgp_sock);
348 return -1;
349 }
350
351 if (CHECK_FLAG(peer1->flags, PEER_FLAG_SHUTDOWN)) {
352 if (bgp_debug_neighbor_events(peer1))
353 zlog_debug(
354 "[Event] connection from %s rejected due to admin shutdown",
355 inet_sutop(&su, buf));
356 close(bgp_sock);
357 return -1;
358 }
359
360 /*
361 * Do not accept incoming connections in Clearing state. This can result
362 * in incorect state transitions - e.g., the connection goes back to
363 * Established and then the Clearing_Completed event is generated. Also,
364 * block incoming connection in Deleted state.
365 */
366 if (peer1->status == Clearing || peer1->status == Deleted) {
367 if (bgp_debug_neighbor_events(peer1))
368 zlog_debug(
369 "[Event] Closing incoming conn for %s (%p) state %d",
370 peer1->host, peer1, peer1->status);
371 close(bgp_sock);
372 return -1;
373 }
374
375 /* Check that at least one AF is activated for the peer. */
376 if (!peer_active(peer1)) {
377 if (bgp_debug_neighbor_events(peer1))
378 zlog_debug(
379 "%s - incoming conn rejected - no AF activated for peer",
380 peer1->host);
381 close(bgp_sock);
382 return -1;
383 }
384
385 if (bgp_debug_neighbor_events(peer1))
386 zlog_debug("[Event] BGP connection from host %s fd %d",
387 inet_sutop(&su, buf), bgp_sock);
388
389 if (peer1->doppelganger) {
390 /* We have an existing connection. Kill the existing one and run
391 with this one.
392 */
393 if (bgp_debug_neighbor_events(peer1))
394 zlog_debug(
395 "[Event] New active connection from peer %s, Killing"
396 " previous active connection",
397 peer1->host);
398 peer_delete(peer1->doppelganger);
399 }
400
401 if (bgp_set_socket_ttl(peer1, bgp_sock) < 0)
402 if (bgp_debug_neighbor_events(peer1))
403 zlog_debug(
404 "[Event] Unable to set min/max TTL on peer %s, Continuing",
405 peer1->host);
406
407 peer = peer_create(&su, peer1->conf_if, peer1->bgp, peer1->local_as,
408 peer1->as, peer1->as_type, 0, 0, NULL);
d62a17ae 409 hash_release(peer->bgp->peerhash, peer);
410 hash_get(peer->bgp->peerhash, peer, hash_alloc_intern);
411
412 peer_xfer_config(peer, peer1);
413 UNSET_FLAG(peer->flags, PEER_FLAG_CONFIG_NODE);
414
415 peer->doppelganger = peer1;
416 peer1->doppelganger = peer;
417 peer->fd = bgp_sock;
97896a91 418 vrf_bind(peer->bgp->vrf_id, bgp_sock, bgp_get_bound_name(peer));
d62a17ae 419 bgp_fsm_change_status(peer, Active);
420 BGP_TIMER_OFF(peer->t_start); /* created in peer_create() */
421
422 SET_FLAG(peer->sflags, PEER_STATUS_ACCEPT_PEER);
423
424 /* Make dummy peer until read Open packet. */
425 if (peer1->status == Established
426 && CHECK_FLAG(peer1->sflags, PEER_STATUS_NSF_MODE)) {
427 /* If we have an existing established connection with graceful
428 * restart
429 * capability announced with one or more address families, then
430 * drop
431 * existing established connection and move state to connect.
432 */
433 peer1->last_reset = PEER_DOWN_NSF_CLOSE_SESSION;
434 SET_FLAG(peer1->sflags, PEER_STATUS_NSF_WAIT);
435 bgp_event_update(peer1, TCP_connection_closed);
436 }
437
438 if (peer_active(peer)) {
439 BGP_EVENT_ADD(peer, TCP_connection_open);
440 }
441
442 return 0;
718e3744 443}
444
445/* BGP socket bind. */
97896a91 446static char *bgp_get_bound_name(struct peer *peer)
718e3744 447{
d62a17ae 448 char *name = NULL;
449
a2bce1c8
DS
450 if (!peer)
451 return NULL;
452
996c9314
LB
453 if ((peer->bgp->vrf_id == VRF_DEFAULT) && !peer->ifname
454 && !peer->conf_if)
97896a91
PG
455 return NULL;
456
d62a17ae 457 if (peer->su.sa.sa_family != AF_INET
458 && peer->su.sa.sa_family != AF_INET6)
97896a91 459 return NULL; // unexpected
d62a17ae 460
461 /* For IPv6 peering, interface (unnumbered or link-local with interface)
462 * takes precedence over VRF. For IPv4 peering, explicit interface or
463 * VRF are the situations to bind.
464 */
465 if (peer->su.sa.sa_family == AF_INET6)
466 name = (peer->conf_if ? peer->conf_if
467 : (peer->ifname ? peer->ifname
468 : peer->bgp->name));
469 else
470 name = peer->ifname ? peer->ifname : peer->bgp->name;
471
97896a91 472 return name;
718e3744 473}
474
d62a17ae 475static int bgp_update_address(struct interface *ifp, const union sockunion *dst,
476 union sockunion *addr)
718e3744 477{
d62a17ae 478 struct prefix *p, *sel, d;
479 struct connected *connected;
480 struct listnode *node;
481 int common;
482
483 sockunion2hostprefix(dst, &d);
484 sel = NULL;
485 common = -1;
486
487 for (ALL_LIST_ELEMENTS_RO(ifp->connected, node, connected)) {
488 p = connected->address;
489 if (p->family != d.family)
490 continue;
491 if (prefix_common_bits(p, &d) > common) {
492 sel = p;
493 common = prefix_common_bits(sel, &d);
494 }
495 }
496
497 if (!sel)
498 return 1;
499
500 prefix2sockunion(sel, addr);
501 return 0;
718e3744 502}
503
504/* Update source selection. */
d62a17ae 505static int bgp_update_source(struct peer *peer)
718e3744 506{
d62a17ae 507 struct interface *ifp;
508 union sockunion addr;
509 int ret = 0;
718e3744 510
d62a17ae 511 sockunion_init(&addr);
dd793e4a 512
d62a17ae 513 /* Source is specified with interface name. */
514 if (peer->update_if) {
515 ifp = if_lookup_by_name(peer->update_if, peer->bgp->vrf_id);
516 if (!ifp)
517 return -1;
718e3744 518
d62a17ae 519 if (bgp_update_address(ifp, &peer->su, &addr))
520 return -1;
718e3744 521
d62a17ae 522 ret = sockunion_bind(peer->fd, &addr, 0, &addr);
523 }
718e3744 524
d62a17ae 525 /* Source is specified with IP address. */
526 if (peer->update_source)
527 ret = sockunion_bind(peer->fd, peer->update_source, 0,
528 peer->update_source);
49067496 529
d62a17ae 530 return ret;
718e3744 531}
532
ed40466a
DS
533#define DATAPLANE_MARK 254 /* main table ID */
534
718e3744 535/* BGP try to connect to the peer. */
d62a17ae 536int bgp_connect(struct peer *peer)
718e3744 537{
b750b0ba
QY
538 assert(!CHECK_FLAG(peer->thread_flags, PEER_THREAD_WRITES_ON));
539 assert(!CHECK_FLAG(peer->thread_flags, PEER_THREAD_READS_ON));
d62a17ae 540 ifindex_t ifindex = 0;
541
542 if (peer->conf_if && BGP_PEER_SU_UNSPEC(peer)) {
543 zlog_debug("Peer address not learnt: Returning from connect");
544 return 0;
545 }
01b9e3fd 546 frr_elevate_privs(&bgpd_privs) {
d62a17ae 547 /* Make socket for the peer. */
01b9e3fd
DL
548 peer->fd = vrf_sockunion_socket(&peer->su, peer->bgp->vrf_id,
549 bgp_get_bound_name(peer));
550 }
d62a17ae 551 if (peer->fd < 0)
552 return -1;
553
554 set_nonblocking(peer->fd);
555
556 /* Set socket send buffer size */
557 setsockopt_so_sendbuf(peer->fd, BGP_SOCKET_SNDBUF_SIZE);
558
559 if (bgp_set_socket_ttl(peer, peer->fd) < 0)
560 return -1;
561
562 sockopt_reuseaddr(peer->fd);
563 sockopt_reuseport(peer->fd);
564 if (sockopt_mark_default(peer->fd, DATAPLANE_MARK, &bgpd_privs) < 0)
e50f7cfd 565 flog_warn(EC_BGP_NO_SOCKOPT_MARK,
28f22e26 566 "Unable to set mark on FD for peer %s, err=%s",
d62a17ae 567 peer->host, safe_strerror(errno));
568
1423c809 569#ifdef IPTOS_PREC_INTERNETCONTROL
01b9e3fd
DL
570 frr_elevate_privs(&bgpd_privs) {
571 if (sockunion_family(&peer->su) == AF_INET)
633fc9b1
DL
572 setsockopt_ipv4_tos(peer->fd,
573 IPTOS_PREC_INTERNETCONTROL);
01b9e3fd 574 else if (sockunion_family(&peer->su) == AF_INET6)
633fc9b1
DL
575 setsockopt_ipv6_tclass(peer->fd,
576 IPTOS_PREC_INTERNETCONTROL);
01b9e3fd 577 }
1423c809
SH
578#endif
579
d62a17ae 580 if (peer->password)
581 bgp_md5_set_connect(peer->fd, &peer->su, peer->password);
718e3744 582
d62a17ae 583 /* Update source bind. */
584 if (bgp_update_source(peer) < 0) {
585 return connect_error;
586 }
718e3744 587
d62a17ae 588 if (peer->conf_if || peer->ifname)
589 ifindex = ifname2ifindex(peer->conf_if ? peer->conf_if
590 : peer->ifname,
591 peer->bgp->vrf_id);
718e3744 592
d62a17ae 593 if (bgp_debug_neighbor_events(peer))
594 zlog_debug("%s [Event] Connect start to %s fd %d", peer->host,
595 peer->host, peer->fd);
718e3744 596
d62a17ae 597 /* Connect to the remote peer. */
598 return sockunion_connect(peer->fd, &peer->su, htons(peer->port),
599 ifindex);
718e3744 600}
601
602/* After TCP connection is established. Get local address and port. */
d62a17ae 603int bgp_getsockname(struct peer *peer)
718e3744 604{
d62a17ae 605 if (peer->su_local) {
606 sockunion_free(peer->su_local);
607 peer->su_local = NULL;
608 }
609
610 if (peer->su_remote) {
611 sockunion_free(peer->su_remote);
612 peer->su_remote = NULL;
613 }
614
615 peer->su_local = sockunion_getsockname(peer->fd);
616 if (!peer->su_local)
617 return -1;
618 peer->su_remote = sockunion_getpeername(peer->fd);
619 if (!peer->su_remote)
620 return -1;
621
17cdd31e
DS
622 if (!bgp_zebra_nexthop_set(peer->su_local, peer->su_remote,
623 &peer->nexthop, peer)) {
2564f080 624 flog_err(EC_BGP_NH_UPD,
17cdd31e
DS
625 "%s: nexthop_set failed, resetting connection - intf %p",
626 peer->host, peer->nexthop.ifp);
d62a17ae 627 return -1;
d62a17ae 628 }
d62a17ae 629 return 0;
718e3744 630}
631
d023aec4 632
61cf4b37
PG
633static int bgp_listener(int sock, struct sockaddr *sa, socklen_t salen,
634 struct bgp *bgp)
d023aec4 635{
d62a17ae 636 struct bgp_listener *listener;
637 int ret, en;
d023aec4 638
d62a17ae 639 sockopt_reuseaddr(sock);
640 sockopt_reuseport(sock);
d023aec4 641
01b9e3fd 642 frr_elevate_privs(&bgpd_privs) {
5c88f19d 643
d023aec4 644#ifdef IPTOS_PREC_INTERNETCONTROL
01b9e3fd
DL
645 if (sa->sa_family == AF_INET)
646 setsockopt_ipv4_tos(sock, IPTOS_PREC_INTERNETCONTROL);
647 else if (sa->sa_family == AF_INET6)
633fc9b1
DL
648 setsockopt_ipv6_tclass(sock,
649 IPTOS_PREC_INTERNETCONTROL);
d023aec4
SH
650#endif
651
01b9e3fd 652 sockopt_v6only(sa->sa_family, sock);
d62a17ae 653
01b9e3fd
DL
654 ret = bind(sock, sa, salen);
655 en = errno;
656 }
d62a17ae 657
658 if (ret < 0) {
450971aa 659 flog_err_sys(EC_LIB_SOCKET, "bind: %s", safe_strerror(en));
d62a17ae 660 return ret;
661 }
662
48522088 663 ret = listen(sock, SOMAXCONN);
d62a17ae 664 if (ret < 0) {
1c50c1c0 665 flog_err_sys(EC_LIB_SOCKET, "listen: %s", safe_strerror(errno));
d62a17ae 666 return ret;
667 }
668
61cf4b37 669 listener = XCALLOC(MTYPE_BGP_LISTENER, sizeof(*listener));
d62a17ae 670 listener->fd = sock;
61cf4b37
PG
671
672 /* this socket needs a change of ns. record bgp back pointer */
3d4c0b49
PG
673 if (bgp->vrf_id != VRF_DEFAULT && vrf_is_mapped_on_netns(
674 vrf_lookup_by_id(bgp->vrf_id)))
61cf4b37
PG
675 listener->bgp = bgp;
676
d62a17ae 677 memcpy(&listener->su, sa, salen);
678 listener->thread = NULL;
679 thread_add_read(bm->master, bgp_accept, listener, sock,
680 &listener->thread);
681 listnode_add(bm->listen_sockets, listener);
682
683 return 0;
d023aec4
SH
684}
685
718e3744 686/* IPv6 supported version of BGP server socket setup. */
61cf4b37 687int bgp_socket(struct bgp *bgp, unsigned short port, const char *address)
718e3744 688{
d62a17ae 689 struct addrinfo *ainfo;
690 struct addrinfo *ainfo_save;
691 static const struct addrinfo req = {
692 .ai_family = AF_UNSPEC,
693 .ai_flags = AI_PASSIVE,
694 .ai_socktype = SOCK_STREAM,
695 };
696 int ret, count;
697 char port_str[BUFSIZ];
698
699 snprintf(port_str, sizeof(port_str), "%d", port);
700 port_str[sizeof(port_str) - 1] = '\0';
701
01b9e3fd
DL
702 frr_elevate_privs(&bgpd_privs) {
703 ret = vrf_getaddrinfo(address, port_str, &req, &ainfo_save,
704 bgp->vrf_id);
705 }
d62a17ae 706 if (ret != 0) {
450971aa 707 flog_err_sys(EC_LIB_SOCKET, "getaddrinfo: %s",
09c866e3 708 gai_strerror(ret));
d62a17ae 709 return -1;
710 }
0b014ea6 711 if (bgp_option_check(BGP_OPT_NO_ZEBRA) &&
449cff3e
A
712 bgp->vrf_id != VRF_DEFAULT) {
713 freeaddrinfo(ainfo_save);
0b014ea6 714 return -1;
449cff3e 715 }
d62a17ae 716 count = 0;
717 for (ainfo = ainfo_save; ainfo; ainfo = ainfo->ai_next) {
718 int sock;
719
720 if (ainfo->ai_family != AF_INET && ainfo->ai_family != AF_INET6)
721 continue;
722
01b9e3fd
DL
723 frr_elevate_privs(&bgpd_privs) {
724 sock = vrf_socket(ainfo->ai_family,
725 ainfo->ai_socktype,
726 ainfo->ai_protocol, bgp->vrf_id,
633fc9b1
DL
727 (bgp->inst_type
728 == BGP_INSTANCE_TYPE_VRF
729 ? bgp->name : NULL));
01b9e3fd 730 }
d62a17ae 731 if (sock < 0) {
450971aa 732 flog_err_sys(EC_LIB_SOCKET, "socket: %s",
09c866e3 733 safe_strerror(errno));
d62a17ae 734 continue;
735 }
736
737 /* if we intend to implement ttl-security, this socket needs
738 * ttl=255 */
739 sockopt_ttl(ainfo->ai_family, sock, MAXTTL);
740
996c9314
LB
741 ret = bgp_listener(sock, ainfo->ai_addr, ainfo->ai_addrlen,
742 bgp);
d62a17ae 743 if (ret == 0)
744 ++count;
745 else
746 close(sock);
747 }
748 freeaddrinfo(ainfo_save);
06969768 749 if (count == 0 && bgp->inst_type != BGP_INSTANCE_TYPE_VRF) {
af4c2728 750 flog_err(
450971aa 751 EC_LIB_SOCKET,
996c9314
LB
752 "%s: no usable addresses please check other programs usage of specified port %d",
753 __func__, port);
450971aa 754 flog_err_sys(EC_LIB_SOCKET, "%s: Program cannot continue",
09c866e3 755 __func__);
b3b78f6e 756 exit(-1);
d62a17ae 757 }
758
759 return 0;
718e3744 760}
d023aec4 761
e5619c28
PG
762/* this function closes vrf socket
763 * this should be called only for vrf socket with netns backend
764 */
765void bgp_close_vrf_socket(struct bgp *bgp)
766{
767 struct listnode *node, *next;
768 struct bgp_listener *listener;
769
770 if (!bgp)
771 return;
772
773 if (bm->listen_sockets == NULL)
774 return;
775
776 for (ALL_LIST_ELEMENTS(bm->listen_sockets, node, next, listener)) {
777 if (listener->bgp == bgp) {
778 thread_cancel(listener->thread);
779 close(listener->fd);
780 listnode_delete(bm->listen_sockets, listener);
781 XFREE(MTYPE_BGP_LISTENER, listener);
782 }
783 }
784}
785
786/* this function closes main socket
787 */
d62a17ae 788void bgp_close(void)
d023aec4 789{
d62a17ae 790 struct listnode *node, *next;
791 struct bgp_listener *listener;
792
793 if (bm->listen_sockets == NULL)
794 return;
795
796 for (ALL_LIST_ELEMENTS(bm->listen_sockets, node, next, listener)) {
e5619c28
PG
797 if (listener->bgp)
798 continue;
d62a17ae 799 thread_cancel(listener->thread);
800 close(listener->fd);
801 listnode_delete(bm->listen_sockets, listener);
802 XFREE(MTYPE_BGP_LISTENER, listener);
803 }
d023aec4 804}