bgpd/bgp_network.c

   1 /* BGP network related fucntions
   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  */
  20
  21 #include <zebra.h>
  22
  23 #include "thread.h"
  24 #include "sockunion.h"
  25 #include "sockopt.h"
  26 #include "memory.h"
  27 #include "log.h"
  28 #include "if.h"
  29 #include "prefix.h"
  30 #include "command.h"
  31 #include "privs.h"
  32 #include "linklist.h"
  33 #include "network.h"
  34 #include "queue.h"
  35 #include "hash.h"
  36 #include "filter.h"
  37 #include "ns.h"
  38 #include "lib_errors.h"
  39 #include "nexthop.h"
  40
  41 #include "bgpd/bgpd.h"
  42 #include "bgpd/bgp_open.h"
  43 #include "bgpd/bgp_fsm.h"
  44 #include "bgpd/bgp_attr.h"
  45 #include "bgpd/bgp_debug.h"
  46 #include "bgpd/bgp_errors.h"
  47 #include "bgpd/bgp_network.h"
  48 #include "bgpd/bgp_zebra.h"
  49
  50 extern struct zebra_privs_t bgpd_privs;
  51
  52 static char *bgp_get_bound_name(struct peer *peer);
  53
  54 /* BGP listening socket. */
  55 struct bgp_listener {
  56         int fd;
  57         union sockunion su;
  58         struct thread *thread;
  59         struct bgp *bgp;
  60 };
  61
  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  */
  66 static int bgp_md5_set_socket(int socket, union sockunion *su,
  67                               const char *password)
  68 {
  69         int ret = -1;
  70         int en = ENOSYS;
  71 #if HAVE_DECL_TCP_MD5SIG
  72         union sockunion su2;
  73 #endif /* HAVE_TCP_MD5SIG */
  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;
  86 #endif /* HAVE_TCP_MD5SIG */
  87
  88         if (ret < 0)
  89                 flog_warn(EC_BGP_NO_TCP_MD5,
  90                           "can't set TCP_MD5SIG option on socket %d: %s",
  91                           socket, safe_strerror(en));
  92
  93         return ret;
  94 }
  95
  96 /* Helper for bgp_connect */
  97 static int bgp_md5_set_connect(int socket, union sockunion *su,
  98                                const char *password)
  99 {
 100         int ret = -1;
 101
 102 #if HAVE_DECL_TCP_MD5SIG
 103         frr_elevate_privs(&bgpd_privs) {
 104                 ret = bgp_md5_set_socket(socket, su, password);
 105         }
 106 #endif /* HAVE_TCP_MD5SIG */
 107
 108         return ret;
 109 }
 110
 111 static int bgp_md5_set_password(struct peer *peer, const char *password)
 112 {
 113         struct listnode *node;
 114         int ret = 0;
 115         struct bgp_listener *listener;
 116
 117         frr_elevate_privs(&bgpd_privs) {
 118         /* Set or unset the password on the listen socket(s). Outbound
 119          * connections are taken care of in bgp_connect() below.
 120          */
 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         }
 129         return ret;
 130 }
 131
 132 int bgp_md5_set(struct peer *peer)
 133 {
 134         /* Set the password from listen socket. */
 135         return bgp_md5_set_password(peer, peer->password);
 136 }
 137
 138 int bgp_md5_unset(struct peer *peer)
 139 {
 140         /* Unset the password from listen socket. */
 141         return bgp_md5_set_password(peer, NULL);
 142 }
 143
 144 int bgp_set_socket_ttl(struct peer *peer, int bgp_sock)
 145 {
 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) {
 153                         flog_err(
 154                                 EC_LIB_SOCKET,
 155                                 "%s: Can't set TxTTL on peer (rtrid %s) socket, err = %d",
 156                                 __func__,
 157                                 inet_ntop(AF_INET, &peer->remote_id, buf,
 158                                           sizeof(buf)),
 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) {
 169                         flog_err(
 170                                 EC_LIB_SOCKET,
 171                                 "%s: Can't set TxTTL on peer (rtrid %s) socket, err = %d",
 172                                 __func__,
 173                                 inet_ntop(AF_INET, &peer->remote_id, buf,
 174                                           sizeof(buf)),
 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) {
 181                         flog_err(
 182                                 EC_LIB_SOCKET,
 183                                 "%s: Can't set MinTTL on peer (rtrid %s) socket, err = %d",
 184                                 __func__,
 185                                 inet_ntop(AF_INET, &peer->remote_id, buf,
 186                                           sizeof(buf)),
 187                                 errno);
 188                         return ret;
 189                 }
 190         }
 191
 192         return ret;
 193 }
 194
 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  */
 202 static int bgp_get_instance_for_inc_conn(int sock, struct bgp **bgp_inst)
 203 {
 204 #ifndef SO_BINDTODEVICE
 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;
 214
 215 #else
 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)
 227                 flog_err(EC_LIB_SOCKET,
 228                          "[Error] BGP SO_BINDTODEVICE get failed (%s), sock %d",
 229                          safe_strerror(errno), sock);
 230                 return -1;
 231 #endif
 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;
 264 #endif
 265 }
 266
 267 /* Accept bgp connection. */
 268 static int bgp_accept(struct thread *thread)
 269 {
 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) {
 284                 flog_err_sys(EC_LIB_SOCKET, "accept_sock is nevative value %d",
 285                              accept_sock);
 286                 return -1;
 287         }
 288         listener->thread = NULL;
 289
 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) {
 296                 flog_err_sys(EC_LIB_SOCKET,
 297                              "[Error] BGP socket accept failed (%s)",
 298                              safe_strerror(errno));
 299                 return -1;
 300         }
 301         set_nonblocking(bgp_sock);
 302
 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)) {
 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);
 409         peer->su = su;
 410         hash_release(peer->bgp->peerhash, peer);
 411         hash_get(peer->bgp->peerhash, peer, hash_alloc_intern);
 412
 413         peer_xfer_config(peer, peer1);
 414         UNSET_FLAG(peer->flags, PEER_FLAG_CONFIG_NODE);
 415
 416         peer->doppelganger = peer1;
 417         peer1->doppelganger = peer;
 418         peer->fd = bgp_sock;
 419         vrf_bind(peer->bgp->vrf_id, bgp_sock, bgp_get_bound_name(peer));
 420         bgp_fsm_change_status(peer, Active);
 421         BGP_TIMER_OFF(peer->t_start); /* created in peer_create() */
 422
 423         SET_FLAG(peer->sflags, PEER_STATUS_ACCEPT_PEER);
 424
 425         /* Make dummy peer until read Open packet. */
 426         if (peer1->status == Established
 427             && CHECK_FLAG(peer1->sflags, PEER_STATUS_NSF_MODE)) {
 428                 /* If we have an existing established connection with graceful
 429                  * restart
 430                  * capability announced with one or more address families, then
 431                  * drop
 432                  * existing established connection and move state to connect.
 433                  */
 434                 peer1->last_reset = PEER_DOWN_NSF_CLOSE_SESSION;
 435                 SET_FLAG(peer1->sflags, PEER_STATUS_NSF_WAIT);
 436                 bgp_event_update(peer1, TCP_connection_closed);
 437         }
 438
 439         if (peer_active(peer)) {
 440                 BGP_EVENT_ADD(peer, TCP_connection_open);
 441         }
 442
 443         return 0;
 444 }
 445
 446 /* BGP socket bind. */
 447 static char *bgp_get_bound_name(struct peer *peer)
 448 {
 449         char *name = NULL;
 450
 451         if (!peer)
 452                 return NULL;
 453
 454         if ((peer->bgp->vrf_id == VRF_DEFAULT) && !peer->ifname
 455             && !peer->conf_if)
 456                 return NULL;
 457
 458         if (peer->su.sa.sa_family != AF_INET
 459             && peer->su.sa.sa_family != AF_INET6)
 460                 return NULL; // unexpected
 461
 462         /* For IPv6 peering, interface (unnumbered or link-local with interface)
 463          * takes precedence over VRF. For IPv4 peering, explicit interface or
 464          * VRF are the situations to bind.
 465          */
 466         if (peer->su.sa.sa_family == AF_INET6)
 467                 name = (peer->conf_if ? peer->conf_if
 468                                       : (peer->ifname ? peer->ifname
 469                                                       : peer->bgp->name));
 470         else
 471                 name = peer->ifname ? peer->ifname : peer->bgp->name;
 472
 473         return name;
 474 }
 475
 476 static int bgp_update_address(struct interface *ifp, const union sockunion *dst,
 477                               union sockunion *addr)
 478 {
 479         struct prefix *p, *sel, d;
 480         struct connected *connected;
 481         struct listnode *node;
 482         int common;
 483
 484         sockunion2hostprefix(dst, &d);
 485         sel = NULL;
 486         common = -1;
 487
 488         for (ALL_LIST_ELEMENTS_RO(ifp->connected, node, connected)) {
 489                 p = connected->address;
 490                 if (p->family != d.family)
 491                         continue;
 492                 if (prefix_common_bits(p, &d) > common) {
 493                         sel = p;
 494                         common = prefix_common_bits(sel, &d);
 495                 }
 496         }
 497
 498         if (!sel)
 499                 return 1;
 500
 501         prefix2sockunion(sel, addr);
 502         return 0;
 503 }
 504
 505 /* Update source selection.  */
 506 static int bgp_update_source(struct peer *peer)
 507 {
 508         struct interface *ifp;
 509         union sockunion addr;
 510         int ret = 0;
 511
 512         sockunion_init(&addr);
 513
 514         /* Source is specified with interface name.  */
 515         if (peer->update_if) {
 516                 ifp = if_lookup_by_name(peer->update_if, peer->bgp->vrf_id);
 517                 if (!ifp)
 518                         return -1;
 519
 520                 if (bgp_update_address(ifp, &peer->su, &addr))
 521                         return -1;
 522
 523                 ret = sockunion_bind(peer->fd, &addr, 0, &addr);
 524         }
 525
 526         /* Source is specified with IP address.  */
 527         if (peer->update_source)
 528                 ret = sockunion_bind(peer->fd, peer->update_source, 0,
 529                                      peer->update_source);
 530
 531         return ret;
 532 }
 533
 534 #define DATAPLANE_MARK 254      /* main table ID */
 535
 536 /* BGP try to connect to the peer.  */
 537 int bgp_connect(struct peer *peer)
 538 {
 539         assert(!CHECK_FLAG(peer->thread_flags, PEER_THREAD_WRITES_ON));
 540         assert(!CHECK_FLAG(peer->thread_flags, PEER_THREAD_READS_ON));
 541         ifindex_t ifindex = 0;
 542
 543         if (peer->conf_if && BGP_PEER_SU_UNSPEC(peer)) {
 544                 zlog_debug("Peer address not learnt: Returning from connect");
 545                 return 0;
 546         }
 547         frr_elevate_privs(&bgpd_privs) {
 548         /* Make socket for the peer. */
 549                 peer->fd = vrf_sockunion_socket(&peer->su, peer->bgp->vrf_id,
 550                                                 bgp_get_bound_name(peer));
 551         }
 552         if (peer->fd < 0)
 553                 return -1;
 554
 555         set_nonblocking(peer->fd);
 556
 557         /* Set socket send buffer size */
 558         setsockopt_so_sendbuf(peer->fd, BGP_SOCKET_SNDBUF_SIZE);
 559
 560         if (bgp_set_socket_ttl(peer, peer->fd) < 0)
 561                 return -1;
 562
 563         sockopt_reuseaddr(peer->fd);
 564         sockopt_reuseport(peer->fd);
 565         if (sockopt_mark_default(peer->fd, DATAPLANE_MARK, &bgpd_privs) < 0)
 566                 flog_warn(EC_BGP_NO_SOCKOPT_MARK,
 567                           "Unable to set mark on FD for peer %s, err=%s",
 568                           peer->host, safe_strerror(errno));
 569
 570 #ifdef IPTOS_PREC_INTERNETCONTROL
 571         frr_elevate_privs(&bgpd_privs) {
 572                 if (sockunion_family(&peer->su) == AF_INET)
 573                         setsockopt_ipv4_tos(peer->fd,
 574                                             IPTOS_PREC_INTERNETCONTROL);
 575                 else if (sockunion_family(&peer->su) == AF_INET6)
 576                         setsockopt_ipv6_tclass(peer->fd,
 577                                                IPTOS_PREC_INTERNETCONTROL);
 578         }
 579 #endif
 580
 581         if (peer->password)
 582                 bgp_md5_set_connect(peer->fd, &peer->su, peer->password);
 583
 584         /* Update source bind. */
 585         if (bgp_update_source(peer) < 0) {
 586                 return connect_error;
 587         }
 588
 589         if (peer->conf_if || peer->ifname)
 590                 ifindex = ifname2ifindex(peer->conf_if ? peer->conf_if
 591                                                        : peer->ifname,
 592                                          peer->bgp->vrf_id);
 593
 594         if (bgp_debug_neighbor_events(peer))
 595                 zlog_debug("%s [Event] Connect start to %s fd %d", peer->host,
 596                            peer->host, peer->fd);
 597
 598         /* Connect to the remote peer. */
 599         return sockunion_connect(peer->fd, &peer->su, htons(peer->port),
 600                                  ifindex);
 601 }
 602
 603 /* After TCP connection is established.  Get local address and port. */
 604 int bgp_getsockname(struct peer *peer)
 605 {
 606         if (peer->su_local) {
 607                 sockunion_free(peer->su_local);
 608                 peer->su_local = NULL;
 609         }
 610
 611         if (peer->su_remote) {
 612                 sockunion_free(peer->su_remote);
 613                 peer->su_remote = NULL;
 614         }
 615
 616         peer->su_local = sockunion_getsockname(peer->fd);
 617         if (!peer->su_local)
 618                 return -1;
 619         peer->su_remote = sockunion_getpeername(peer->fd);
 620         if (!peer->su_remote)
 621                 return -1;
 622
 623         if (!bgp_zebra_nexthop_set(peer->su_local, peer->su_remote,
 624                                    &peer->nexthop, peer)) {
 625                 flog_err(EC_BGP_NH_UPD,
 626                          "%s: nexthop_set failed, resetting connection - intf %p",
 627                          peer->host, peer->nexthop.ifp);
 628                 return -1;
 629         }
 630         return 0;
 631 }
 632
 633
 634 static int bgp_listener(int sock, struct sockaddr *sa, socklen_t salen,
 635                         struct bgp *bgp)
 636 {
 637         struct bgp_listener *listener;
 638         int ret, en;
 639
 640         sockopt_reuseaddr(sock);
 641         sockopt_reuseport(sock);
 642
 643         frr_elevate_privs(&bgpd_privs) {
 644
 645 #ifdef IPTOS_PREC_INTERNETCONTROL
 646                 if (sa->sa_family == AF_INET)
 647                         setsockopt_ipv4_tos(sock, IPTOS_PREC_INTERNETCONTROL);
 648                 else if (sa->sa_family == AF_INET6)
 649                         setsockopt_ipv6_tclass(sock,
 650                                                IPTOS_PREC_INTERNETCONTROL);
 651 #endif
 652
 653                 sockopt_v6only(sa->sa_family, sock);
 654
 655                 ret = bind(sock, sa, salen);
 656                 en = errno;
 657         }
 658
 659         if (ret < 0) {
 660                 flog_err_sys(EC_LIB_SOCKET, "bind: %s", safe_strerror(en));
 661                 return ret;
 662         }
 663
 664         ret = listen(sock, SOMAXCONN);
 665         if (ret < 0) {
 666                 flog_err_sys(EC_LIB_SOCKET, "listen: %s", safe_strerror(errno));
 667                 return ret;
 668         }
 669
 670         listener = XCALLOC(MTYPE_BGP_LISTENER, sizeof(*listener));
 671         listener->fd = sock;
 672
 673         /* this socket needs a change of ns. record bgp back pointer */
 674         if (bgp->vrf_id != VRF_DEFAULT && vrf_is_mapped_on_netns(
 675                                                 vrf_lookup_by_id(bgp->vrf_id)))
 676                 listener->bgp = bgp;
 677
 678         memcpy(&listener->su, sa, salen);
 679         listener->thread = NULL;
 680         thread_add_read(bm->master, bgp_accept, listener, sock,
 681                         &listener->thread);
 682         listnode_add(bm->listen_sockets, listener);
 683
 684         return 0;
 685 }
 686
 687 /* IPv6 supported version of BGP server socket setup.  */
 688 int bgp_socket(struct bgp *bgp, unsigned short port, const char *address)
 689 {
 690         struct addrinfo *ainfo;
 691         struct addrinfo *ainfo_save;
 692         static const struct addrinfo req = {
 693                 .ai_family = AF_UNSPEC,
 694                 .ai_flags = AI_PASSIVE,
 695                 .ai_socktype = SOCK_STREAM,
 696         };
 697         int ret, count;
 698         char port_str[BUFSIZ];
 699
 700         snprintf(port_str, sizeof(port_str), "%d", port);
 701         port_str[sizeof(port_str) - 1] = '\0';
 702
 703         frr_elevate_privs(&bgpd_privs) {
 704                 ret = vrf_getaddrinfo(address, port_str, &req, &ainfo_save,
 705                                       bgp->vrf_id);
 706         }
 707         if (ret != 0) {
 708                 flog_err_sys(EC_LIB_SOCKET, "getaddrinfo: %s",
 709                              gai_strerror(ret));
 710                 return -1;
 711         }
 712
 713         count = 0;
 714         for (ainfo = ainfo_save; ainfo; ainfo = ainfo->ai_next) {
 715                 int sock;
 716
 717                 if (ainfo->ai_family != AF_INET && ainfo->ai_family != AF_INET6)
 718                         continue;
 719
 720                 frr_elevate_privs(&bgpd_privs) {
 721                         sock = vrf_socket(ainfo->ai_family,
 722                                           ainfo->ai_socktype,
 723                                           ainfo->ai_protocol, bgp->vrf_id,
 724                                           (bgp->inst_type
 725                                            == BGP_INSTANCE_TYPE_VRF
 726                                            ? bgp->name : NULL));
 727                 }
 728                 if (sock < 0) {
 729                         flog_err_sys(EC_LIB_SOCKET, "socket: %s",
 730                                      safe_strerror(errno));
 731                         continue;
 732                 }
 733
 734                 /* if we intend to implement ttl-security, this socket needs
 735                  * ttl=255 */
 736                 sockopt_ttl(ainfo->ai_family, sock, MAXTTL);
 737
 738                 ret = bgp_listener(sock, ainfo->ai_addr, ainfo->ai_addrlen,
 739                                    bgp);
 740                 if (ret == 0)
 741                         ++count;
 742                 else
 743                         close(sock);
 744         }
 745         freeaddrinfo(ainfo_save);
 746         if (count == 0 && bgp->inst_type != BGP_INSTANCE_TYPE_VRF) {
 747                 flog_err(
 748                         EC_LIB_SOCKET,
 749                         "%s: no usable addresses please check other programs usage of specified port %d",
 750                         __func__, port);
 751                 flog_err_sys(EC_LIB_SOCKET, "%s: Program cannot continue",
 752                              __func__);
 753                 exit(-1);
 754         }
 755
 756         return 0;
 757 }
 758
 759 /* this function closes vrf socket
 760  * this should be called only for vrf socket with netns backend
 761  */
 762 void bgp_close_vrf_socket(struct bgp *bgp)
 763 {
 764         struct listnode *node, *next;
 765         struct bgp_listener *listener;
 766
 767         if (!bgp)
 768                 return;
 769
 770         if (bm->listen_sockets == NULL)
 771                 return;
 772
 773         for (ALL_LIST_ELEMENTS(bm->listen_sockets, node, next, listener)) {
 774                 if (listener->bgp == bgp) {
 775                         thread_cancel(listener->thread);
 776                         close(listener->fd);
 777                         listnode_delete(bm->listen_sockets, listener);
 778                         XFREE(MTYPE_BGP_LISTENER, listener);
 779                 }
 780         }
 781 }
 782
 783 /* this function closes main socket
 784  */
 785 void bgp_close(void)
 786 {
 787         struct listnode *node, *next;
 788         struct bgp_listener *listener;
 789
 790         if (bm->listen_sockets == NULL)
 791                 return;
 792
 793         for (ALL_LIST_ELEMENTS(bm->listen_sockets, node, next, listener)) {
 794                 if (listener->bgp)
 795                         continue;
 796                 thread_cancel(listener->thread);
 797                 close(listener->fd);
 798                 listnode_delete(bm->listen_sockets, listener);
 799                 XFREE(MTYPE_BGP_LISTENER, listener);
 800         }
 801 }