1 /* $OpenBSD: session.c,v 1.399 2020/02/12 10:33:56 claudio Exp $ */ 2 3 /* 4 * Copyright (c) 2003, 2004, 2005 Henning Brauer <henning@openbsd.org> 5 * Copyright (c) 2017 Peter van Dijk <peter.van.dijk@powerdns.com> 6 * 7 * Permission to use, copy, modify, and distribute this software for any 8 * purpose with or without fee is hereby granted, provided that the above 9 * copyright notice and this permission notice appear in all copies. 10 * 11 * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES 12 * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF 13 * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR 14 * ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES 15 * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN 16 * ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF 17 * OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE. 18 */ 19 20 #include <sys/types.h> 21 22 #include <sys/mman.h> 23 #include <sys/socket.h> 24 #include <sys/time.h> 25 #include <sys/resource.h> 26 #include <sys/un.h> 27 #include <netinet/in.h> 28 #include <netinet/ip.h> 29 #include <netinet/tcp.h> 30 #include <arpa/inet.h> 31 #include <limits.h> 32 33 #include <err.h> 34 #include <errno.h> 35 #include <fcntl.h> 36 #include <ifaddrs.h> 37 #include <poll.h> 38 #include <pwd.h> 39 #include <signal.h> 40 #include <stdio.h> 41 #include <stdlib.h> 42 #include <string.h> 43 #include <syslog.h> 44 #include <unistd.h> 45 46 #include "bgpd.h" 47 #include "session.h" 48 #include "log.h" 49 50 #define PFD_PIPE_MAIN 0 51 #define PFD_PIPE_ROUTE 1 52 #define PFD_PIPE_ROUTE_CTL 2 53 #define PFD_SOCK_CTL 3 54 #define PFD_SOCK_RCTL 4 55 #define PFD_LISTENERS_START 5 56 57 void session_sighdlr(int); 58 int setup_listeners(u_int *); 59 void init_peer(struct peer *); 60 void start_timer_holdtime(struct peer *); 61 void start_timer_keepalive(struct peer *); 62 void session_close_connection(struct peer *); 63 void change_state(struct peer *, enum session_state, enum session_events); 64 int session_setup_socket(struct peer *); 65 void session_accept(int); 66 int session_connect(struct peer *); 67 void session_tcp_established(struct peer *); 68 void session_capa_ann_none(struct peer *); 69 int session_capa_add(struct ibuf *, u_int8_t, u_int8_t); 70 int session_capa_add_mp(struct ibuf *, u_int8_t); 71 int session_capa_add_gr(struct peer *, struct ibuf *, u_int8_t); 72 struct bgp_msg *session_newmsg(enum msg_type, u_int16_t); 73 int session_sendmsg(struct bgp_msg *, struct peer *); 74 void session_open(struct peer *); 75 void session_keepalive(struct peer *); 76 void session_update(u_int32_t, void *, size_t); 77 void session_notification(struct peer *, u_int8_t, u_int8_t, void *, 78 ssize_t); 79 void session_rrefresh(struct peer *, u_int8_t); 80 int session_graceful_restart(struct peer *); 81 int session_graceful_stop(struct peer *); 82 int session_dispatch_msg(struct pollfd *, struct peer *); 83 void session_process_msg(struct peer *); 84 int parse_header(struct peer *, u_char *, u_int16_t *, u_int8_t *); 85 int parse_open(struct peer *); 86 int parse_update(struct peer *); 87 int parse_refresh(struct peer *); 88 int parse_notification(struct peer *); 89 int parse_capabilities(struct peer *, u_char *, u_int16_t, u_int32_t *); 90 int capa_neg_calc(struct peer *); 91 void session_dispatch_imsg(struct imsgbuf *, int, u_int *); 92 void session_up(struct peer *); 93 void session_down(struct peer *); 94 int imsg_rde(int, u_int32_t, void *, u_int16_t); 95 void session_demote(struct peer *, int); 96 void merge_peers(struct bgpd_config *, struct bgpd_config *); 97 98 int la_cmp(struct listen_addr *, struct listen_addr *); 99 void session_template_clone(struct peer *, struct sockaddr *, 100 u_int32_t, u_int32_t); 101 int session_match_mask(struct peer *, struct bgpd_addr *); 102 103 struct bgpd_config *conf, *nconf; 104 struct bgpd_sysdep sysdep; 105 volatile sig_atomic_t session_quit; 106 int pending_reconf; 107 int csock = -1, rcsock = -1; 108 u_int peer_cnt; 109 struct imsgbuf *ibuf_rde; 110 struct imsgbuf *ibuf_rde_ctl; 111 struct imsgbuf *ibuf_main; 112 113 struct mrt_head mrthead; 114 time_t pauseaccept; 115 116 static inline int 117 peer_compare(const struct peer *a, const struct peer *b) 118 { 119 return a->conf.id - b->conf.id; 120 } 121 122 RB_GENERATE(peer_head, peer, entry, peer_compare); 123 124 void 125 session_sighdlr(int sig) 126 { 127 switch (sig) { 128 case SIGINT: 129 case SIGTERM: 130 session_quit = 1; 131 break; 132 } 133 } 134 135 int 136 setup_listeners(u_int *la_cnt) 137 { 138 int ttl = 255; 139 struct listen_addr *la; 140 u_int cnt = 0; 141 142 TAILQ_FOREACH(la, conf->listen_addrs, entry) { 143 la->reconf = RECONF_NONE; 144 cnt++; 145 146 if (la->flags & LISTENER_LISTENING) 147 continue; 148 149 if (la->fd == -1) { 150 log_warn("cannot establish listener on %s: invalid fd", 151 log_sockaddr((struct sockaddr *)&la->sa, 152 la->sa_len)); 153 continue; 154 } 155 156 if (tcp_md5_prep_listener(la, &conf->peers) == -1) 157 fatal("tcp_md5_prep_listener"); 158 159 /* set ttl to 255 so that ttl-security works */ 160 if (la->sa.ss_family == AF_INET && setsockopt(la->fd, 161 IPPROTO_IP, IP_TTL, &ttl, sizeof(ttl)) == -1) { 162 log_warn("setup_listeners setsockopt TTL"); 163 continue; 164 } 165 if (la->sa.ss_family == AF_INET6 && setsockopt(la->fd, 166 IPPROTO_IPV6, IPV6_UNICAST_HOPS, &ttl, sizeof(ttl)) == -1) { 167 log_warn("setup_listeners setsockopt hoplimit"); 168 continue; 169 } 170 171 if (listen(la->fd, MAX_BACKLOG)) { 172 close(la->fd); 173 fatal("listen"); 174 } 175 176 la->flags |= LISTENER_LISTENING; 177 178 log_info("listening on %s", 179 log_sockaddr((struct sockaddr *)&la->sa, la->sa_len)); 180 } 181 182 *la_cnt = cnt; 183 184 return (0); 185 } 186 187 void 188 session_main(int debug, int verbose) 189 { 190 int timeout; 191 unsigned int i, j, idx_peers, idx_listeners, idx_mrts; 192 u_int pfd_elms = 0, peer_l_elms = 0, mrt_l_elms = 0; 193 u_int listener_cnt, ctl_cnt, mrt_cnt; 194 u_int new_cnt; 195 struct passwd *pw; 196 struct peer *p, **peer_l = NULL, *next; 197 struct mrt *m, *xm, **mrt_l = NULL; 198 struct pollfd *pfd = NULL; 199 struct ctl_conn *ctl_conn; 200 struct listen_addr *la; 201 void *newp; 202 time_t now; 203 short events; 204 205 log_init(debug, LOG_DAEMON); 206 log_setverbose(verbose); 207 208 bgpd_process = PROC_SE; 209 log_procinit(log_procnames[bgpd_process]); 210 211 if ((pw = getpwnam(BGPD_USER)) == NULL) 212 fatal(NULL); 213 214 if (chroot(pw->pw_dir) == -1) 215 fatal("chroot"); 216 if (chdir("/") == -1) 217 fatal("chdir(\"/\")"); 218 219 setproctitle("session engine"); 220 221 if (setgroups(1, &pw->pw_gid) || 222 setresgid(pw->pw_gid, pw->pw_gid, pw->pw_gid) || 223 setresuid(pw->pw_uid, pw->pw_uid, pw->pw_uid)) 224 fatal("can't drop privileges"); 225 226 if (pledge("stdio inet recvfd", NULL) == -1) 227 fatal("pledge"); 228 229 signal(SIGTERM, session_sighdlr); 230 signal(SIGINT, session_sighdlr); 231 signal(SIGPIPE, SIG_IGN); 232 signal(SIGHUP, SIG_IGN); 233 signal(SIGALRM, SIG_IGN); 234 signal(SIGUSR1, SIG_IGN); 235 236 if ((ibuf_main = malloc(sizeof(struct imsgbuf))) == NULL) 237 fatal(NULL); 238 imsg_init(ibuf_main, 3); 239 240 TAILQ_INIT(&ctl_conns); 241 LIST_INIT(&mrthead); 242 listener_cnt = 0; 243 peer_cnt = 0; 244 ctl_cnt = 0; 245 246 conf = new_config(); 247 log_info("session engine ready"); 248 249 while (session_quit == 0) { 250 /* check for peers to be initialized or deleted */ 251 if (!pending_reconf) { 252 RB_FOREACH_SAFE(p, peer_head, &conf->peers, next) { 253 /* cloned peer that idled out? */ 254 if (p->template && (p->state == STATE_IDLE || 255 p->state == STATE_ACTIVE) && 256 getmonotime() - p->stats.last_updown >= 257 INTERVAL_HOLD_CLONED) 258 p->reconf_action = RECONF_DELETE; 259 260 /* new peer that needs init? */ 261 if (p->state == STATE_NONE) 262 init_peer(p); 263 264 /* reinit due? */ 265 if (p->reconf_action == RECONF_REINIT) { 266 session_stop(p, ERR_CEASE_ADMIN_RESET); 267 if (!p->conf.down) 268 timer_set(p, Timer_IdleHold, 0); 269 } 270 271 /* deletion due? */ 272 if (p->reconf_action == RECONF_DELETE) { 273 if (p->demoted) 274 session_demote(p, -1); 275 p->conf.demote_group[0] = 0; 276 session_stop(p, ERR_CEASE_PEER_UNCONF); 277 timer_remove_all(p); 278 tcp_md5_del_listener(conf, p); 279 log_peer_warnx(&p->conf, "removed"); 280 RB_REMOVE(peer_head, &conf->peers, p); 281 free(p); 282 peer_cnt--; 283 continue; 284 } 285 p->reconf_action = RECONF_NONE; 286 } 287 } 288 289 if (peer_cnt > peer_l_elms) { 290 if ((newp = reallocarray(peer_l, peer_cnt, 291 sizeof(struct peer *))) == NULL) { 292 /* panic for now */ 293 log_warn("could not resize peer_l from %u -> %u" 294 " entries", peer_l_elms, peer_cnt); 295 fatalx("exiting"); 296 } 297 peer_l = newp; 298 peer_l_elms = peer_cnt; 299 } 300 301 mrt_cnt = 0; 302 for (m = LIST_FIRST(&mrthead); m != NULL; m = xm) { 303 xm = LIST_NEXT(m, entry); 304 if (m->state == MRT_STATE_REMOVE) { 305 mrt_clean(m); 306 LIST_REMOVE(m, entry); 307 free(m); 308 continue; 309 } 310 if (m->wbuf.queued) 311 mrt_cnt++; 312 } 313 314 if (mrt_cnt > mrt_l_elms) { 315 if ((newp = reallocarray(mrt_l, mrt_cnt, 316 sizeof(struct mrt *))) == NULL) { 317 /* panic for now */ 318 log_warn("could not resize mrt_l from %u -> %u" 319 " entries", mrt_l_elms, mrt_cnt); 320 fatalx("exiting"); 321 } 322 mrt_l = newp; 323 mrt_l_elms = mrt_cnt; 324 } 325 326 new_cnt = PFD_LISTENERS_START + listener_cnt + peer_cnt + 327 ctl_cnt + mrt_cnt; 328 if (new_cnt > pfd_elms) { 329 if ((newp = reallocarray(pfd, new_cnt, 330 sizeof(struct pollfd))) == NULL) { 331 /* panic for now */ 332 log_warn("could not resize pfd from %u -> %u" 333 " entries", pfd_elms, new_cnt); 334 fatalx("exiting"); 335 } 336 pfd = newp; 337 pfd_elms = new_cnt; 338 } 339 340 bzero(pfd, sizeof(struct pollfd) * pfd_elms); 341 342 set_pollfd(&pfd[PFD_PIPE_MAIN], ibuf_main); 343 set_pollfd(&pfd[PFD_PIPE_ROUTE], ibuf_rde); 344 set_pollfd(&pfd[PFD_PIPE_ROUTE_CTL], ibuf_rde_ctl); 345 346 if (pauseaccept == 0) { 347 pfd[PFD_SOCK_CTL].fd = csock; 348 pfd[PFD_SOCK_CTL].events = POLLIN; 349 pfd[PFD_SOCK_RCTL].fd = rcsock; 350 pfd[PFD_SOCK_RCTL].events = POLLIN; 351 } else { 352 pfd[PFD_SOCK_CTL].fd = -1; 353 pfd[PFD_SOCK_RCTL].fd = -1; 354 } 355 356 i = PFD_LISTENERS_START; 357 TAILQ_FOREACH(la, conf->listen_addrs, entry) { 358 if (pauseaccept == 0) { 359 pfd[i].fd = la->fd; 360 pfd[i].events = POLLIN; 361 } else 362 pfd[i].fd = -1; 363 i++; 364 } 365 idx_listeners = i; 366 timeout = 240; /* loop every 240s at least */ 367 368 now = getmonotime(); 369 RB_FOREACH(p, peer_head, &conf->peers) { 370 time_t nextaction; 371 struct peer_timer *pt; 372 373 /* check timers */ 374 if ((pt = timer_nextisdue(p, now)) != NULL) { 375 switch (pt->type) { 376 case Timer_Hold: 377 bgp_fsm(p, EVNT_TIMER_HOLDTIME); 378 break; 379 case Timer_ConnectRetry: 380 bgp_fsm(p, EVNT_TIMER_CONNRETRY); 381 break; 382 case Timer_Keepalive: 383 bgp_fsm(p, EVNT_TIMER_KEEPALIVE); 384 break; 385 case Timer_IdleHold: 386 bgp_fsm(p, EVNT_START); 387 break; 388 case Timer_IdleHoldReset: 389 p->IdleHoldTime = 390 INTERVAL_IDLE_HOLD_INITIAL; 391 p->errcnt = 0; 392 timer_stop(p, Timer_IdleHoldReset); 393 break; 394 case Timer_CarpUndemote: 395 timer_stop(p, Timer_CarpUndemote); 396 if (p->demoted && 397 p->state == STATE_ESTABLISHED) 398 session_demote(p, -1); 399 break; 400 case Timer_RestartTimeout: 401 timer_stop(p, Timer_RestartTimeout); 402 session_graceful_stop(p); 403 break; 404 default: 405 fatalx("King Bula lost in time"); 406 } 407 } 408 if ((nextaction = timer_nextduein(p, now)) != -1 && 409 nextaction < timeout) 410 timeout = nextaction; 411 412 /* are we waiting for a write? */ 413 events = POLLIN; 414 if (p->wbuf.queued > 0 || p->state == STATE_CONNECT) 415 events |= POLLOUT; 416 /* is there still work to do? */ 417 if (p->rpending && p->rbuf && p->rbuf->wpos) 418 timeout = 0; 419 420 /* poll events */ 421 if (p->fd != -1 && events != 0) { 422 pfd[i].fd = p->fd; 423 pfd[i].events = events; 424 peer_l[i - idx_listeners] = p; 425 i++; 426 } 427 } 428 429 idx_peers = i; 430 431 LIST_FOREACH(m, &mrthead, entry) 432 if (m->wbuf.queued) { 433 pfd[i].fd = m->wbuf.fd; 434 pfd[i].events = POLLOUT; 435 mrt_l[i - idx_peers] = m; 436 i++; 437 } 438 439 idx_mrts = i; 440 441 TAILQ_FOREACH(ctl_conn, &ctl_conns, entry) { 442 pfd[i].fd = ctl_conn->ibuf.fd; 443 pfd[i].events = POLLIN; 444 if (ctl_conn->ibuf.w.queued > 0) 445 pfd[i].events |= POLLOUT; 446 i++; 447 } 448 449 if (pauseaccept && timeout > 1) 450 timeout = 1; 451 if (timeout < 0) 452 timeout = 0; 453 if (poll(pfd, i, timeout * 1000) == -1) 454 if (errno != EINTR) 455 fatal("poll error"); 456 457 /* 458 * If we previously saw fd exhaustion, we stop accept() 459 * for 1 second to throttle the accept() loop. 460 */ 461 if (pauseaccept && getmonotime() > pauseaccept + 1) 462 pauseaccept = 0; 463 464 if (handle_pollfd(&pfd[PFD_PIPE_MAIN], ibuf_main) == -1) { 465 log_warnx("SE: Lost connection to parent"); 466 session_quit = 1; 467 continue; 468 } else 469 session_dispatch_imsg(ibuf_main, PFD_PIPE_MAIN, 470 &listener_cnt); 471 472 if (handle_pollfd(&pfd[PFD_PIPE_ROUTE], ibuf_rde) == -1) { 473 log_warnx("SE: Lost connection to RDE"); 474 msgbuf_clear(&ibuf_rde->w); 475 free(ibuf_rde); 476 ibuf_rde = NULL; 477 } else 478 session_dispatch_imsg(ibuf_rde, PFD_PIPE_ROUTE, 479 &listener_cnt); 480 481 if (handle_pollfd(&pfd[PFD_PIPE_ROUTE_CTL], ibuf_rde_ctl) == 482 -1) { 483 log_warnx("SE: Lost connection to RDE control"); 484 msgbuf_clear(&ibuf_rde_ctl->w); 485 free(ibuf_rde_ctl); 486 ibuf_rde_ctl = NULL; 487 } else 488 session_dispatch_imsg(ibuf_rde_ctl, PFD_PIPE_ROUTE_CTL, 489 &listener_cnt); 490 491 if (pfd[PFD_SOCK_CTL].revents & POLLIN) 492 ctl_cnt += control_accept(csock, 0); 493 494 if (pfd[PFD_SOCK_RCTL].revents & POLLIN) 495 ctl_cnt += control_accept(rcsock, 1); 496 497 for (j = PFD_LISTENERS_START; j < idx_listeners; j++) 498 if (pfd[j].revents & POLLIN) 499 session_accept(pfd[j].fd); 500 501 for (; j < idx_peers; j++) 502 session_dispatch_msg(&pfd[j], 503 peer_l[j - idx_listeners]); 504 505 RB_FOREACH(p, peer_head, &conf->peers) 506 if (p->rbuf && p->rbuf->wpos) 507 session_process_msg(p); 508 509 for (; j < idx_mrts; j++) 510 if (pfd[j].revents & POLLOUT) 511 mrt_write(mrt_l[j - idx_peers]); 512 513 for (; j < i; j++) 514 control_dispatch_msg(&pfd[j], &ctl_cnt, &conf->peers); 515 } 516 517 RB_FOREACH_SAFE(p, peer_head, &conf->peers, next) { 518 RB_REMOVE(peer_head, &conf->peers, p); 519 strlcpy(p->conf.shutcomm, 520 "bgpd shutting down", 521 sizeof(p->conf.shutcomm)); 522 session_stop(p, ERR_CEASE_ADMIN_DOWN); 523 timer_remove_all(p); 524 free(p); 525 } 526 527 while ((m = LIST_FIRST(&mrthead)) != NULL) { 528 mrt_clean(m); 529 LIST_REMOVE(m, entry); 530 free(m); 531 } 532 533 free_config(conf); 534 free(peer_l); 535 free(mrt_l); 536 free(pfd); 537 538 /* close pipes */ 539 if (ibuf_rde) { 540 msgbuf_write(&ibuf_rde->w); 541 msgbuf_clear(&ibuf_rde->w); 542 close(ibuf_rde->fd); 543 free(ibuf_rde); 544 } 545 if (ibuf_rde_ctl) { 546 msgbuf_clear(&ibuf_rde_ctl->w); 547 close(ibuf_rde_ctl->fd); 548 free(ibuf_rde_ctl); 549 } 550 msgbuf_write(&ibuf_main->w); 551 msgbuf_clear(&ibuf_main->w); 552 close(ibuf_main->fd); 553 free(ibuf_main); 554 555 control_shutdown(csock); 556 control_shutdown(rcsock); 557 log_info("session engine exiting"); 558 exit(0); 559 } 560 561 void 562 init_peer(struct peer *p) 563 { 564 TAILQ_INIT(&p->timers); 565 p->fd = p->wbuf.fd = -1; 566 567 if (p->conf.if_depend[0]) 568 imsg_compose(ibuf_main, IMSG_IFINFO, 0, 0, -1, 569 p->conf.if_depend, sizeof(p->conf.if_depend)); 570 else 571 p->depend_ok = 1; 572 573 peer_cnt++; 574 575 change_state(p, STATE_IDLE, EVNT_NONE); 576 if (p->conf.down) 577 timer_stop(p, Timer_IdleHold); /* no autostart */ 578 else 579 timer_set(p, Timer_IdleHold, 0); /* start ASAP */ 580 581 /* 582 * on startup, demote if requested. 583 * do not handle new peers. they must reach ESTABLISHED beforehands. 584 * peers added at runtime have reconf_action set to RECONF_REINIT. 585 */ 586 if (p->reconf_action != RECONF_REINIT && p->conf.demote_group[0]) 587 session_demote(p, +1); 588 } 589 590 void 591 bgp_fsm(struct peer *peer, enum session_events event) 592 { 593 switch (peer->state) { 594 case STATE_NONE: 595 /* nothing */ 596 break; 597 case STATE_IDLE: 598 switch (event) { 599 case EVNT_START: 600 timer_stop(peer, Timer_Hold); 601 timer_stop(peer, Timer_Keepalive); 602 timer_stop(peer, Timer_IdleHold); 603 604 /* allocate read buffer */ 605 peer->rbuf = calloc(1, sizeof(struct ibuf_read)); 606 if (peer->rbuf == NULL) 607 fatal(NULL); 608 609 /* init write buffer */ 610 msgbuf_init(&peer->wbuf); 611 612 peer->stats.last_sent_errcode = 0; 613 peer->stats.last_sent_suberr = 0; 614 peer->stats.last_rcvd_errcode = 0; 615 peer->stats.last_rcvd_suberr = 0; 616 617 if (!peer->depend_ok) 618 timer_stop(peer, Timer_ConnectRetry); 619 else if (peer->passive || peer->conf.passive || 620 peer->conf.template) { 621 change_state(peer, STATE_ACTIVE, event); 622 timer_stop(peer, Timer_ConnectRetry); 623 } else { 624 change_state(peer, STATE_CONNECT, event); 625 timer_set(peer, Timer_ConnectRetry, 626 conf->connectretry); 627 session_connect(peer); 628 } 629 peer->passive = 0; 630 break; 631 default: 632 /* ignore */ 633 break; 634 } 635 break; 636 case STATE_CONNECT: 637 switch (event) { 638 case EVNT_START: 639 /* ignore */ 640 break; 641 case EVNT_CON_OPEN: 642 session_tcp_established(peer); 643 session_open(peer); 644 timer_stop(peer, Timer_ConnectRetry); 645 peer->holdtime = INTERVAL_HOLD_INITIAL; 646 start_timer_holdtime(peer); 647 change_state(peer, STATE_OPENSENT, event); 648 break; 649 case EVNT_CON_OPENFAIL: 650 timer_set(peer, Timer_ConnectRetry, 651 conf->connectretry); 652 session_close_connection(peer); 653 change_state(peer, STATE_ACTIVE, event); 654 break; 655 case EVNT_TIMER_CONNRETRY: 656 timer_set(peer, Timer_ConnectRetry, 657 conf->connectretry); 658 session_connect(peer); 659 break; 660 default: 661 change_state(peer, STATE_IDLE, event); 662 break; 663 } 664 break; 665 case STATE_ACTIVE: 666 switch (event) { 667 case EVNT_START: 668 /* ignore */ 669 break; 670 case EVNT_CON_OPEN: 671 session_tcp_established(peer); 672 session_open(peer); 673 timer_stop(peer, Timer_ConnectRetry); 674 peer->holdtime = INTERVAL_HOLD_INITIAL; 675 start_timer_holdtime(peer); 676 change_state(peer, STATE_OPENSENT, event); 677 break; 678 case EVNT_CON_OPENFAIL: 679 timer_set(peer, Timer_ConnectRetry, 680 conf->connectretry); 681 session_close_connection(peer); 682 change_state(peer, STATE_ACTIVE, event); 683 break; 684 case EVNT_TIMER_CONNRETRY: 685 timer_set(peer, Timer_ConnectRetry, 686 peer->holdtime); 687 change_state(peer, STATE_CONNECT, event); 688 session_connect(peer); 689 break; 690 default: 691 change_state(peer, STATE_IDLE, event); 692 break; 693 } 694 break; 695 case STATE_OPENSENT: 696 switch (event) { 697 case EVNT_START: 698 /* ignore */ 699 break; 700 case EVNT_STOP: 701 change_state(peer, STATE_IDLE, event); 702 break; 703 case EVNT_CON_CLOSED: 704 session_close_connection(peer); 705 timer_set(peer, Timer_ConnectRetry, 706 conf->connectretry); 707 change_state(peer, STATE_ACTIVE, event); 708 break; 709 case EVNT_CON_FATAL: 710 change_state(peer, STATE_IDLE, event); 711 break; 712 case EVNT_TIMER_HOLDTIME: 713 session_notification(peer, ERR_HOLDTIMEREXPIRED, 714 0, NULL, 0); 715 change_state(peer, STATE_IDLE, event); 716 break; 717 case EVNT_RCVD_OPEN: 718 /* parse_open calls change_state itself on failure */ 719 if (parse_open(peer)) 720 break; 721 session_keepalive(peer); 722 change_state(peer, STATE_OPENCONFIRM, event); 723 break; 724 case EVNT_RCVD_NOTIFICATION: 725 if (parse_notification(peer)) { 726 change_state(peer, STATE_IDLE, event); 727 /* don't punish, capa negotiation */ 728 timer_set(peer, Timer_IdleHold, 0); 729 peer->IdleHoldTime /= 2; 730 } else 731 change_state(peer, STATE_IDLE, event); 732 break; 733 default: 734 session_notification(peer, 735 ERR_FSM, ERR_FSM_UNEX_OPENSENT, NULL, 0); 736 change_state(peer, STATE_IDLE, event); 737 break; 738 } 739 break; 740 case STATE_OPENCONFIRM: 741 switch (event) { 742 case EVNT_START: 743 /* ignore */ 744 break; 745 case EVNT_STOP: 746 change_state(peer, STATE_IDLE, event); 747 break; 748 case EVNT_CON_CLOSED: 749 case EVNT_CON_FATAL: 750 change_state(peer, STATE_IDLE, event); 751 break; 752 case EVNT_TIMER_HOLDTIME: 753 session_notification(peer, ERR_HOLDTIMEREXPIRED, 754 0, NULL, 0); 755 change_state(peer, STATE_IDLE, event); 756 break; 757 case EVNT_TIMER_KEEPALIVE: 758 session_keepalive(peer); 759 break; 760 case EVNT_RCVD_KEEPALIVE: 761 start_timer_holdtime(peer); 762 change_state(peer, STATE_ESTABLISHED, event); 763 break; 764 case EVNT_RCVD_NOTIFICATION: 765 parse_notification(peer); 766 change_state(peer, STATE_IDLE, event); 767 break; 768 default: 769 session_notification(peer, 770 ERR_FSM, ERR_FSM_UNEX_OPENCONFIRM, NULL, 0); 771 change_state(peer, STATE_IDLE, event); 772 break; 773 } 774 break; 775 case STATE_ESTABLISHED: 776 switch (event) { 777 case EVNT_START: 778 /* ignore */ 779 break; 780 case EVNT_STOP: 781 change_state(peer, STATE_IDLE, event); 782 break; 783 case EVNT_CON_CLOSED: 784 case EVNT_CON_FATAL: 785 change_state(peer, STATE_IDLE, event); 786 break; 787 case EVNT_TIMER_HOLDTIME: 788 session_notification(peer, ERR_HOLDTIMEREXPIRED, 789 0, NULL, 0); 790 change_state(peer, STATE_IDLE, event); 791 break; 792 case EVNT_TIMER_KEEPALIVE: 793 session_keepalive(peer); 794 break; 795 case EVNT_RCVD_KEEPALIVE: 796 start_timer_holdtime(peer); 797 break; 798 case EVNT_RCVD_UPDATE: 799 start_timer_holdtime(peer); 800 if (parse_update(peer)) 801 change_state(peer, STATE_IDLE, event); 802 else 803 start_timer_holdtime(peer); 804 break; 805 case EVNT_RCVD_NOTIFICATION: 806 parse_notification(peer); 807 change_state(peer, STATE_IDLE, event); 808 break; 809 default: 810 session_notification(peer, 811 ERR_FSM, ERR_FSM_UNEX_ESTABLISHED, NULL, 0); 812 change_state(peer, STATE_IDLE, event); 813 break; 814 } 815 break; 816 } 817 } 818 819 void 820 start_timer_holdtime(struct peer *peer) 821 { 822 if (peer->holdtime > 0) 823 timer_set(peer, Timer_Hold, peer->holdtime); 824 else 825 timer_stop(peer, Timer_Hold); 826 } 827 828 void 829 start_timer_keepalive(struct peer *peer) 830 { 831 if (peer->holdtime > 0) 832 timer_set(peer, Timer_Keepalive, peer->holdtime / 3); 833 else 834 timer_stop(peer, Timer_Keepalive); 835 } 836 837 void 838 session_close_connection(struct peer *peer) 839 { 840 if (peer->fd != -1) { 841 close(peer->fd); 842 pauseaccept = 0; 843 } 844 peer->fd = peer->wbuf.fd = -1; 845 } 846 847 void 848 change_state(struct peer *peer, enum session_state state, 849 enum session_events event) 850 { 851 struct mrt *mrt; 852 853 switch (state) { 854 case STATE_IDLE: 855 /* carp demotion first. new peers handled in init_peer */ 856 if (peer->state == STATE_ESTABLISHED && 857 peer->conf.demote_group[0] && !peer->demoted) 858 session_demote(peer, +1); 859 860 /* 861 * try to write out what's buffered (maybe a notification), 862 * don't bother if it fails 863 */ 864 if (peer->state >= STATE_OPENSENT && peer->wbuf.queued) 865 msgbuf_write(&peer->wbuf); 866 867 /* 868 * we must start the timer for the next EVNT_START 869 * if we are coming here due to an error and the 870 * session was not established successfully before, the 871 * starttimerinterval needs to be exponentially increased 872 */ 873 if (peer->IdleHoldTime == 0) 874 peer->IdleHoldTime = INTERVAL_IDLE_HOLD_INITIAL; 875 peer->holdtime = INTERVAL_HOLD_INITIAL; 876 timer_stop(peer, Timer_ConnectRetry); 877 timer_stop(peer, Timer_Keepalive); 878 timer_stop(peer, Timer_Hold); 879 timer_stop(peer, Timer_IdleHold); 880 timer_stop(peer, Timer_IdleHoldReset); 881 session_close_connection(peer); 882 msgbuf_clear(&peer->wbuf); 883 free(peer->rbuf); 884 peer->rbuf = NULL; 885 peer->rpending = 0; 886 bzero(&peer->capa.peer, sizeof(peer->capa.peer)); 887 if (!peer->template) 888 imsg_compose(ibuf_main, IMSG_PFKEY_RELOAD, 889 peer->conf.id, 0, -1, NULL, 0); 890 891 if (event != EVNT_STOP) { 892 timer_set(peer, Timer_IdleHold, peer->IdleHoldTime); 893 if (event != EVNT_NONE && 894 peer->IdleHoldTime < MAX_IDLE_HOLD/2) 895 peer->IdleHoldTime *= 2; 896 } 897 if (peer->state == STATE_ESTABLISHED) { 898 if (peer->capa.neg.grestart.restart == 2 && 899 (event == EVNT_CON_CLOSED || 900 event == EVNT_CON_FATAL)) { 901 /* don't punish graceful restart */ 902 timer_set(peer, Timer_IdleHold, 0); 903 peer->IdleHoldTime /= 2; 904 session_graceful_restart(peer); 905 } else 906 session_down(peer); 907 } 908 if (peer->state == STATE_NONE || 909 peer->state == STATE_ESTABLISHED) { 910 /* initialize capability negotiation structures */ 911 memcpy(&peer->capa.ann, &peer->conf.capabilities, 912 sizeof(peer->capa.ann)); 913 if (!peer->conf.announce_capa) 914 session_capa_ann_none(peer); 915 } 916 break; 917 case STATE_CONNECT: 918 if (peer->state == STATE_ESTABLISHED && 919 peer->capa.neg.grestart.restart == 2) { 920 /* do the graceful restart dance */ 921 session_graceful_restart(peer); 922 peer->holdtime = INTERVAL_HOLD_INITIAL; 923 timer_stop(peer, Timer_ConnectRetry); 924 timer_stop(peer, Timer_Keepalive); 925 timer_stop(peer, Timer_Hold); 926 timer_stop(peer, Timer_IdleHold); 927 timer_stop(peer, Timer_IdleHoldReset); 928 session_close_connection(peer); 929 msgbuf_clear(&peer->wbuf); 930 bzero(&peer->capa.peer, sizeof(peer->capa.peer)); 931 } 932 break; 933 case STATE_ACTIVE: 934 if (!peer->template) 935 imsg_compose(ibuf_main, IMSG_PFKEY_RELOAD, 936 peer->conf.id, 0, -1, NULL, 0); 937 break; 938 case STATE_OPENSENT: 939 break; 940 case STATE_OPENCONFIRM: 941 break; 942 case STATE_ESTABLISHED: 943 timer_set(peer, Timer_IdleHoldReset, peer->IdleHoldTime); 944 if (peer->demoted) 945 timer_set(peer, Timer_CarpUndemote, 946 INTERVAL_HOLD_DEMOTED); 947 session_up(peer); 948 break; 949 default: /* something seriously fucked */ 950 break; 951 } 952 953 log_statechange(peer, state, event); 954 LIST_FOREACH(mrt, &mrthead, entry) { 955 if (!(mrt->type == MRT_ALL_IN || mrt->type == MRT_ALL_OUT)) 956 continue; 957 if ((mrt->peer_id == 0 && mrt->group_id == 0) || 958 mrt->peer_id == peer->conf.id || (mrt->group_id != 0 && 959 mrt->group_id == peer->conf.groupid)) 960 mrt_dump_state(mrt, peer->state, state, peer); 961 } 962 peer->prev_state = peer->state; 963 peer->state = state; 964 } 965 966 void 967 session_accept(int listenfd) 968 { 969 int connfd; 970 socklen_t len; 971 struct sockaddr_storage cliaddr; 972 struct peer *p = NULL; 973 974 len = sizeof(cliaddr); 975 if ((connfd = accept4(listenfd, 976 (struct sockaddr *)&cliaddr, &len, 977 SOCK_CLOEXEC | SOCK_NONBLOCK)) == -1) { 978 if (errno == ENFILE || errno == EMFILE) 979 pauseaccept = getmonotime(); 980 else if (errno != EWOULDBLOCK && errno != EINTR && 981 errno != ECONNABORTED) 982 log_warn("accept"); 983 return; 984 } 985 986 p = getpeerbyip(conf, (struct sockaddr *)&cliaddr); 987 988 if (p != NULL && p->state == STATE_IDLE && p->errcnt < 2) { 989 if (timer_running(p, Timer_IdleHold, NULL)) { 990 /* fast reconnect after clear */ 991 p->passive = 1; 992 bgp_fsm(p, EVNT_START); 993 } 994 } 995 996 if (p != NULL && 997 (p->state == STATE_CONNECT || p->state == STATE_ACTIVE)) { 998 if (p->fd != -1) { 999 if (p->state == STATE_CONNECT) 1000 session_close_connection(p); 1001 else { 1002 close(connfd); 1003 return; 1004 } 1005 } 1006 1007 open: 1008 if (p->conf.auth.method != AUTH_NONE && sysdep.no_pfkey) { 1009 log_peer_warnx(&p->conf, 1010 "ipsec or md5sig configured but not available"); 1011 close(connfd); 1012 return; 1013 } 1014 1015 if (tcp_md5_check(connfd, p) == -1) { 1016 close(connfd); 1017 return; 1018 } 1019 p->fd = p->wbuf.fd = connfd; 1020 if (session_setup_socket(p)) { 1021 close(connfd); 1022 return; 1023 } 1024 bgp_fsm(p, EVNT_CON_OPEN); 1025 return; 1026 } else if (p != NULL && p->state == STATE_ESTABLISHED && 1027 p->capa.neg.grestart.restart == 2) { 1028 /* first do the graceful restart dance */ 1029 change_state(p, STATE_CONNECT, EVNT_CON_CLOSED); 1030 /* then do part of the open dance */ 1031 goto open; 1032 } else { 1033 log_conn_attempt(p, (struct sockaddr *)&cliaddr, len); 1034 close(connfd); 1035 } 1036 } 1037 1038 int 1039 session_connect(struct peer *peer) 1040 { 1041 struct sockaddr *sa; 1042 socklen_t sa_len; 1043 1044 /* 1045 * we do not need the overcomplicated collision detection RFC 1771 1046 * describes; we simply make sure there is only ever one concurrent 1047 * tcp connection per peer. 1048 */ 1049 if (peer->fd != -1) 1050 return (-1); 1051 1052 if ((peer->fd = socket(aid2af(peer->conf.remote_addr.aid), 1053 SOCK_STREAM | SOCK_CLOEXEC | SOCK_NONBLOCK, IPPROTO_TCP)) == -1) { 1054 log_peer_warn(&peer->conf, "session_connect socket"); 1055 bgp_fsm(peer, EVNT_CON_OPENFAIL); 1056 return (-1); 1057 } 1058 1059 if (peer->conf.auth.method != AUTH_NONE && sysdep.no_pfkey) { 1060 log_peer_warnx(&peer->conf, 1061 "ipsec or md5sig configured but not available"); 1062 bgp_fsm(peer, EVNT_CON_OPENFAIL); 1063 return (-1); 1064 } 1065 1066 tcp_md5_set(peer->fd, peer); 1067 peer->wbuf.fd = peer->fd; 1068 1069 /* if update source is set we need to bind() */ 1070 if ((sa = addr2sa(&peer->conf.local_addr, 0, &sa_len)) != NULL) { 1071 if (bind(peer->fd, sa, sa_len) == -1) { 1072 log_peer_warn(&peer->conf, "session_connect bind"); 1073 bgp_fsm(peer, EVNT_CON_OPENFAIL); 1074 return (-1); 1075 } 1076 } 1077 1078 if (session_setup_socket(peer)) { 1079 bgp_fsm(peer, EVNT_CON_OPENFAIL); 1080 return (-1); 1081 } 1082 1083 sa = addr2sa(&peer->conf.remote_addr, BGP_PORT, &sa_len); 1084 if (connect(peer->fd, sa, sa_len) == -1) { 1085 if (errno != EINPROGRESS) { 1086 if (errno != peer->lasterr) 1087 log_peer_warn(&peer->conf, "connect"); 1088 peer->lasterr = errno; 1089 bgp_fsm(peer, EVNT_CON_OPENFAIL); 1090 return (-1); 1091 } 1092 } else 1093 bgp_fsm(peer, EVNT_CON_OPEN); 1094 1095 return (0); 1096 } 1097 1098 int 1099 session_setup_socket(struct peer *p) 1100 { 1101 int ttl = p->conf.distance; 1102 int pre = IPTOS_PREC_INTERNETCONTROL; 1103 int nodelay = 1; 1104 int bsize; 1105 1106 switch (p->conf.remote_addr.aid) { 1107 case AID_INET: 1108 /* set precedence, see RFC 1771 appendix 5 */ 1109 if (setsockopt(p->fd, IPPROTO_IP, IP_TOS, &pre, sizeof(pre)) == 1110 -1) { 1111 log_peer_warn(&p->conf, 1112 "session_setup_socket setsockopt TOS"); 1113 return (-1); 1114 } 1115 1116 if (p->conf.ebgp) { 1117 /* 1118 * set TTL to foreign router's distance 1119 * 1=direct n=multihop with ttlsec, we always use 255 1120 */ 1121 if (p->conf.ttlsec) { 1122 ttl = 256 - p->conf.distance; 1123 if (setsockopt(p->fd, IPPROTO_IP, IP_MINTTL, 1124 &ttl, sizeof(ttl)) == -1) { 1125 log_peer_warn(&p->conf, 1126 "session_setup_socket: " 1127 "setsockopt MINTTL"); 1128 return (-1); 1129 } 1130 ttl = 255; 1131 } 1132 1133 if (setsockopt(p->fd, IPPROTO_IP, IP_TTL, &ttl, 1134 sizeof(ttl)) == -1) { 1135 log_peer_warn(&p->conf, 1136 "session_setup_socket setsockopt TTL"); 1137 return (-1); 1138 } 1139 } 1140 break; 1141 case AID_INET6: 1142 if (p->conf.ebgp) { 1143 /* 1144 * set hoplimit to foreign router's distance 1145 * 1=direct n=multihop with ttlsec, we always use 255 1146 */ 1147 if (p->conf.ttlsec) { 1148 ttl = 256 - p->conf.distance; 1149 if (setsockopt(p->fd, IPPROTO_IPV6, 1150 IPV6_MINHOPCOUNT, &ttl, sizeof(ttl)) 1151 == -1) { 1152 log_peer_warn(&p->conf, 1153 "session_setup_socket: " 1154 "setsockopt MINHOPCOUNT"); 1155 return (-1); 1156 } 1157 ttl = 255; 1158 } 1159 if (setsockopt(p->fd, IPPROTO_IPV6, IPV6_UNICAST_HOPS, 1160 &ttl, sizeof(ttl)) == -1) { 1161 log_peer_warn(&p->conf, 1162 "session_setup_socket setsockopt hoplimit"); 1163 return (-1); 1164 } 1165 } 1166 break; 1167 } 1168 1169 /* set TCP_NODELAY */ 1170 if (setsockopt(p->fd, IPPROTO_TCP, TCP_NODELAY, &nodelay, 1171 sizeof(nodelay)) == -1) { 1172 log_peer_warn(&p->conf, 1173 "session_setup_socket setsockopt TCP_NODELAY"); 1174 return (-1); 1175 } 1176 1177 /* only increase bufsize (and thus window) if md5 or ipsec is in use */ 1178 if (p->conf.auth.method != AUTH_NONE) { 1179 /* try to increase bufsize. no biggie if it fails */ 1180 bsize = 65535; 1181 while (bsize > 8192 && 1182 setsockopt(p->fd, SOL_SOCKET, SO_RCVBUF, &bsize, 1183 sizeof(bsize)) == -1 && errno != EINVAL) 1184 bsize /= 2; 1185 bsize = 65535; 1186 while (bsize > 8192 && 1187 setsockopt(p->fd, SOL_SOCKET, SO_SNDBUF, &bsize, 1188 sizeof(bsize)) == -1 && errno != EINVAL) 1189 bsize /= 2; 1190 } 1191 1192 return (0); 1193 } 1194 1195 /* compare two sockaddrs by converting them into bgpd_addr */ 1196 static int 1197 sa_cmp(struct sockaddr *a, struct sockaddr *b) 1198 { 1199 struct bgpd_addr ba, bb; 1200 1201 sa2addr(a, &ba, NULL); 1202 sa2addr(b, &bb, NULL); 1203 1204 return (memcmp(&ba, &bb, sizeof(ba)) == 0); 1205 } 1206 1207 static void 1208 get_alternate_addr(struct sockaddr *sa, struct bgpd_addr *alt) 1209 { 1210 struct ifaddrs *ifap, *ifa, *match; 1211 1212 if (getifaddrs(&ifap) == -1) 1213 fatal("getifaddrs"); 1214 1215 for (match = ifap; match != NULL; match = match->ifa_next) 1216 if (sa_cmp(sa, match->ifa_addr) == 0) 1217 break; 1218 1219 if (match == NULL) { 1220 log_warnx("%s: local address not found", __func__); 1221 return; 1222 } 1223 1224 switch (sa->sa_family) { 1225 case AF_INET6: 1226 for (ifa = ifap; ifa != NULL; ifa = ifa->ifa_next) { 1227 if (ifa->ifa_addr->sa_family == AF_INET && 1228 strcmp(ifa->ifa_name, match->ifa_name) == 0) { 1229 sa2addr(ifa->ifa_addr, alt, NULL); 1230 break; 1231 } 1232 } 1233 break; 1234 case AF_INET: 1235 for (ifa = ifap; ifa != NULL; ifa = ifa->ifa_next) { 1236 struct sockaddr_in6 *s = 1237 (struct sockaddr_in6 *)ifa->ifa_addr; 1238 if (ifa->ifa_addr->sa_family == AF_INET6 && 1239 strcmp(ifa->ifa_name, match->ifa_name) == 0) { 1240 /* only accept global scope addresses */ 1241 if (IN6_IS_ADDR_LINKLOCAL(&s->sin6_addr) || 1242 IN6_IS_ADDR_SITELOCAL(&s->sin6_addr)) 1243 continue; 1244 sa2addr(ifa->ifa_addr, alt, NULL); 1245 break; 1246 } 1247 } 1248 break; 1249 default: 1250 log_warnx("%s: unsupported address family %d", __func__, 1251 sa->sa_family); 1252 break; 1253 } 1254 1255 freeifaddrs(ifap); 1256 } 1257 1258 void 1259 session_tcp_established(struct peer *peer) 1260 { 1261 struct sockaddr_storage ss; 1262 socklen_t len; 1263 1264 len = sizeof(ss); 1265 if (getsockname(peer->fd, (struct sockaddr *)&ss, &len) == -1) 1266 log_warn("getsockname"); 1267 sa2addr((struct sockaddr *)&ss, &peer->local, &peer->local_port); 1268 get_alternate_addr((struct sockaddr *)&ss, &peer->local_alt); 1269 len = sizeof(ss); 1270 if (getpeername(peer->fd, (struct sockaddr *)&ss, &len) == -1) 1271 log_warn("getpeername"); 1272 sa2addr((struct sockaddr *)&ss, &peer->remote, &peer->remote_port); 1273 } 1274 1275 void 1276 session_capa_ann_none(struct peer *peer) 1277 { 1278 bzero(&peer->capa.ann, sizeof(peer->capa.ann)); 1279 } 1280 1281 int 1282 session_capa_add(struct ibuf *opb, u_int8_t capa_code, u_int8_t capa_len) 1283 { 1284 int errs = 0; 1285 1286 errs += ibuf_add(opb, &capa_code, sizeof(capa_code)); 1287 errs += ibuf_add(opb, &capa_len, sizeof(capa_len)); 1288 return (errs); 1289 } 1290 1291 int 1292 session_capa_add_mp(struct ibuf *buf, u_int8_t aid) 1293 { 1294 u_int8_t safi, pad = 0; 1295 u_int16_t afi; 1296 int errs = 0; 1297 1298 if (aid2afi(aid, &afi, &safi) == -1) 1299 fatalx("session_capa_add_mp: bad afi/safi pair"); 1300 afi = htons(afi); 1301 errs += ibuf_add(buf, &afi, sizeof(afi)); 1302 errs += ibuf_add(buf, &pad, sizeof(pad)); 1303 errs += ibuf_add(buf, &safi, sizeof(safi)); 1304 1305 return (errs); 1306 } 1307 1308 int 1309 session_capa_add_gr(struct peer *p, struct ibuf *b, u_int8_t aid) 1310 { 1311 u_int errs = 0; 1312 u_int16_t afi; 1313 u_int8_t flags, safi; 1314 1315 if (aid2afi(aid, &afi, &safi)) { 1316 log_warn("session_capa_add_gr: bad AID"); 1317 return (1); 1318 } 1319 if (p->capa.neg.grestart.flags[aid] & CAPA_GR_RESTARTING) 1320 flags = CAPA_GR_F_FLAG; 1321 else 1322 flags = 0; 1323 1324 afi = htons(afi); 1325 errs += ibuf_add(b, &afi, sizeof(afi)); 1326 errs += ibuf_add(b, &safi, sizeof(safi)); 1327 errs += ibuf_add(b, &flags, sizeof(flags)); 1328 1329 return (errs); 1330 } 1331 1332 struct bgp_msg * 1333 session_newmsg(enum msg_type msgtype, u_int16_t len) 1334 { 1335 struct bgp_msg *msg; 1336 struct msg_header hdr; 1337 struct ibuf *buf; 1338 int errs = 0; 1339 1340 memset(&hdr.marker, 0xff, sizeof(hdr.marker)); 1341 hdr.len = htons(len); 1342 hdr.type = msgtype; 1343 1344 if ((buf = ibuf_open(len)) == NULL) 1345 return (NULL); 1346 1347 errs += ibuf_add(buf, &hdr.marker, sizeof(hdr.marker)); 1348 errs += ibuf_add(buf, &hdr.len, sizeof(hdr.len)); 1349 errs += ibuf_add(buf, &hdr.type, sizeof(hdr.type)); 1350 1351 if (errs || (msg = calloc(1, sizeof(*msg))) == NULL) { 1352 ibuf_free(buf); 1353 return (NULL); 1354 } 1355 1356 msg->buf = buf; 1357 msg->type = msgtype; 1358 msg->len = len; 1359 1360 return (msg); 1361 } 1362 1363 int 1364 session_sendmsg(struct bgp_msg *msg, struct peer *p) 1365 { 1366 struct mrt *mrt; 1367 1368 LIST_FOREACH(mrt, &mrthead, entry) { 1369 if (!(mrt->type == MRT_ALL_OUT || (msg->type == UPDATE && 1370 mrt->type == MRT_UPDATE_OUT))) 1371 continue; 1372 if ((mrt->peer_id == 0 && mrt->group_id == 0) || 1373 mrt->peer_id == p->conf.id || (mrt->group_id != 0 && 1374 mrt->group_id == p->conf.groupid)) 1375 mrt_dump_bgp_msg(mrt, msg->buf->buf, msg->len, p); 1376 } 1377 1378 ibuf_close(&p->wbuf, msg->buf); 1379 if (!p->throttled && p->wbuf.queued > SESS_MSG_HIGH_MARK) { 1380 if (imsg_rde(IMSG_XOFF, p->conf.id, NULL, 0) == -1) 1381 log_peer_warn(&p->conf, "imsg_compose XOFF"); 1382 else 1383 p->throttled = 1; 1384 } 1385 1386 free(msg); 1387 return (0); 1388 } 1389 1390 void 1391 session_open(struct peer *p) 1392 { 1393 struct bgp_msg *buf; 1394 struct ibuf *opb; 1395 struct msg_open msg; 1396 u_int16_t len; 1397 u_int8_t i, op_type, optparamlen = 0; 1398 int errs = 0; 1399 int mpcapa = 0; 1400 1401 1402 if ((opb = ibuf_dynamic(0, UCHAR_MAX - sizeof(op_type) - 1403 sizeof(optparamlen))) == NULL) { 1404 bgp_fsm(p, EVNT_CON_FATAL); 1405 return; 1406 } 1407 1408 /* multiprotocol extensions, RFC 4760 */ 1409 for (i = 0; i < AID_MAX; i++) 1410 if (p->capa.ann.mp[i]) { /* 4 bytes data */ 1411 errs += session_capa_add(opb, CAPA_MP, 4); 1412 errs += session_capa_add_mp(opb, i); 1413 mpcapa++; 1414 } 1415 1416 /* route refresh, RFC 2918 */ 1417 if (p->capa.ann.refresh) /* no data */ 1418 errs += session_capa_add(opb, CAPA_REFRESH, 0); 1419 1420 /* graceful restart and End-of-RIB marker, RFC 4724 */ 1421 if (p->capa.ann.grestart.restart) { 1422 int rst = 0; 1423 u_int16_t hdr; 1424 u_int8_t grlen; 1425 1426 if (mpcapa) { 1427 grlen = 2 + 4 * mpcapa; 1428 for (i = 0; i < AID_MAX; i++) { 1429 if (p->capa.neg.grestart.flags[i] & 1430 CAPA_GR_RESTARTING) 1431 rst++; 1432 } 1433 } else { /* AID_INET */ 1434 grlen = 2 + 4; 1435 if (p->capa.neg.grestart.flags[AID_INET] & 1436 CAPA_GR_RESTARTING) 1437 rst++; 1438 } 1439 1440 hdr = conf->holdtime; /* default timeout */ 1441 /* if client does graceful restart don't set R flag */ 1442 if (!rst) 1443 hdr |= CAPA_GR_R_FLAG; 1444 hdr = htons(hdr); 1445 1446 errs += session_capa_add(opb, CAPA_RESTART, grlen); 1447 errs += ibuf_add(opb, &hdr, sizeof(hdr)); 1448 1449 if (mpcapa) { 1450 for (i = 0; i < AID_MAX; i++) { 1451 if (p->capa.ann.mp[i]) { 1452 errs += session_capa_add_gr(p, opb, i); 1453 } 1454 } 1455 } else { /* AID_INET */ 1456 errs += session_capa_add_gr(p, opb, AID_INET); 1457 } 1458 } 1459 1460 /* 4-bytes AS numbers, draft-ietf-idr-as4bytes-13 */ 1461 if (p->capa.ann.as4byte) { /* 4 bytes data */ 1462 u_int32_t nas; 1463 1464 nas = htonl(p->conf.local_as); 1465 errs += session_capa_add(opb, CAPA_AS4BYTE, sizeof(nas)); 1466 errs += ibuf_add(opb, &nas, sizeof(nas)); 1467 } 1468 1469 if (ibuf_size(opb)) 1470 optparamlen = ibuf_size(opb) + sizeof(op_type) + 1471 sizeof(optparamlen); 1472 1473 len = MSGSIZE_OPEN_MIN + optparamlen; 1474 if (errs || (buf = session_newmsg(OPEN, len)) == NULL) { 1475 ibuf_free(opb); 1476 bgp_fsm(p, EVNT_CON_FATAL); 1477 return; 1478 } 1479 1480 msg.version = 4; 1481 msg.myas = htons(p->conf.local_short_as); 1482 if (p->conf.holdtime) 1483 msg.holdtime = htons(p->conf.holdtime); 1484 else 1485 msg.holdtime = htons(conf->holdtime); 1486 msg.bgpid = conf->bgpid; /* is already in network byte order */ 1487 msg.optparamlen = optparamlen; 1488 1489 errs += ibuf_add(buf->buf, &msg.version, sizeof(msg.version)); 1490 errs += ibuf_add(buf->buf, &msg.myas, sizeof(msg.myas)); 1491 errs += ibuf_add(buf->buf, &msg.holdtime, sizeof(msg.holdtime)); 1492 errs += ibuf_add(buf->buf, &msg.bgpid, sizeof(msg.bgpid)); 1493 errs += ibuf_add(buf->buf, &msg.optparamlen, sizeof(msg.optparamlen)); 1494 1495 if (optparamlen) { 1496 op_type = OPT_PARAM_CAPABILITIES; 1497 optparamlen = ibuf_size(opb); 1498 errs += ibuf_add(buf->buf, &op_type, sizeof(op_type)); 1499 errs += ibuf_add(buf->buf, &optparamlen, sizeof(optparamlen)); 1500 errs += ibuf_add(buf->buf, opb->buf, ibuf_size(opb)); 1501 } 1502 1503 ibuf_free(opb); 1504 1505 if (errs) { 1506 ibuf_free(buf->buf); 1507 free(buf); 1508 bgp_fsm(p, EVNT_CON_FATAL); 1509 return; 1510 } 1511 1512 if (session_sendmsg(buf, p) == -1) { 1513 bgp_fsm(p, EVNT_CON_FATAL); 1514 return; 1515 } 1516 1517 p->stats.msg_sent_open++; 1518 } 1519 1520 void 1521 session_keepalive(struct peer *p) 1522 { 1523 struct bgp_msg *buf; 1524 1525 if ((buf = session_newmsg(KEEPALIVE, MSGSIZE_KEEPALIVE)) == NULL || 1526 session_sendmsg(buf, p) == -1) { 1527 bgp_fsm(p, EVNT_CON_FATAL); 1528 return; 1529 } 1530 1531 start_timer_keepalive(p); 1532 p->stats.msg_sent_keepalive++; 1533 } 1534 1535 void 1536 session_update(u_int32_t peerid, void *data, size_t datalen) 1537 { 1538 struct peer *p; 1539 struct bgp_msg *buf; 1540 1541 if ((p = getpeerbyid(conf, peerid)) == NULL) { 1542 log_warnx("no such peer: id=%u", peerid); 1543 return; 1544 } 1545 1546 if (p->state != STATE_ESTABLISHED) 1547 return; 1548 1549 if ((buf = session_newmsg(UPDATE, MSGSIZE_HEADER + datalen)) == NULL) { 1550 bgp_fsm(p, EVNT_CON_FATAL); 1551 return; 1552 } 1553 1554 if (ibuf_add(buf->buf, data, datalen)) { 1555 ibuf_free(buf->buf); 1556 free(buf); 1557 bgp_fsm(p, EVNT_CON_FATAL); 1558 return; 1559 } 1560 1561 if (session_sendmsg(buf, p) == -1) { 1562 bgp_fsm(p, EVNT_CON_FATAL); 1563 return; 1564 } 1565 1566 start_timer_keepalive(p); 1567 p->stats.msg_sent_update++; 1568 } 1569 1570 void 1571 session_notification(struct peer *p, u_int8_t errcode, u_int8_t subcode, 1572 void *data, ssize_t datalen) 1573 { 1574 struct bgp_msg *buf; 1575 int errs = 0; 1576 1577 if (p->stats.last_sent_errcode) /* some notification already sent */ 1578 return; 1579 1580 log_notification(p, errcode, subcode, data, datalen, "sending"); 1581 1582 if ((buf = session_newmsg(NOTIFICATION, 1583 MSGSIZE_NOTIFICATION_MIN + datalen)) == NULL) { 1584 bgp_fsm(p, EVNT_CON_FATAL); 1585 return; 1586 } 1587 1588 errs += ibuf_add(buf->buf, &errcode, sizeof(errcode)); 1589 errs += ibuf_add(buf->buf, &subcode, sizeof(subcode)); 1590 1591 if (datalen > 0) 1592 errs += ibuf_add(buf->buf, data, datalen); 1593 1594 if (errs) { 1595 ibuf_free(buf->buf); 1596 free(buf); 1597 bgp_fsm(p, EVNT_CON_FATAL); 1598 return; 1599 } 1600 1601 if (session_sendmsg(buf, p) == -1) { 1602 bgp_fsm(p, EVNT_CON_FATAL); 1603 return; 1604 } 1605 1606 p->stats.msg_sent_notification++; 1607 p->stats.last_sent_errcode = errcode; 1608 p->stats.last_sent_suberr = subcode; 1609 } 1610 1611 int 1612 session_neighbor_rrefresh(struct peer *p) 1613 { 1614 u_int8_t i; 1615 1616 if (!p->capa.peer.refresh) 1617 return (-1); 1618 1619 for (i = 0; i < AID_MAX; i++) { 1620 if (p->capa.peer.mp[i] != 0) 1621 session_rrefresh(p, i); 1622 } 1623 1624 return (0); 1625 } 1626 1627 void 1628 session_rrefresh(struct peer *p, u_int8_t aid) 1629 { 1630 struct bgp_msg *buf; 1631 int errs = 0; 1632 u_int16_t afi; 1633 u_int8_t safi, null8 = 0; 1634 1635 if (aid2afi(aid, &afi, &safi) == -1) 1636 fatalx("session_rrefresh: bad afi/safi pair"); 1637 1638 if ((buf = session_newmsg(RREFRESH, MSGSIZE_RREFRESH)) == NULL) { 1639 bgp_fsm(p, EVNT_CON_FATAL); 1640 return; 1641 } 1642 1643 afi = htons(afi); 1644 errs += ibuf_add(buf->buf, &afi, sizeof(afi)); 1645 errs += ibuf_add(buf->buf, &null8, sizeof(null8)); 1646 errs += ibuf_add(buf->buf, &safi, sizeof(safi)); 1647 1648 if (errs) { 1649 ibuf_free(buf->buf); 1650 free(buf); 1651 bgp_fsm(p, EVNT_CON_FATAL); 1652 return; 1653 } 1654 1655 if (session_sendmsg(buf, p) == -1) { 1656 bgp_fsm(p, EVNT_CON_FATAL); 1657 return; 1658 } 1659 1660 p->stats.msg_sent_rrefresh++; 1661 } 1662 1663 int 1664 session_graceful_restart(struct peer *p) 1665 { 1666 u_int8_t i; 1667 1668 timer_set(p, Timer_RestartTimeout, p->capa.neg.grestart.timeout); 1669 1670 for (i = 0; i < AID_MAX; i++) { 1671 if (p->capa.neg.grestart.flags[i] & CAPA_GR_PRESENT) { 1672 if (imsg_rde(IMSG_SESSION_STALE, p->conf.id, 1673 &i, sizeof(i)) == -1) 1674 return (-1); 1675 log_peer_warnx(&p->conf, 1676 "graceful restart of %s, keeping routes", 1677 aid2str(i)); 1678 p->capa.neg.grestart.flags[i] |= CAPA_GR_RESTARTING; 1679 } else if (p->capa.neg.mp[i]) { 1680 if (imsg_rde(IMSG_SESSION_FLUSH, p->conf.id, 1681 &i, sizeof(i)) == -1) 1682 return (-1); 1683 log_peer_warnx(&p->conf, 1684 "graceful restart of %s, flushing routes", 1685 aid2str(i)); 1686 } 1687 } 1688 return (0); 1689 } 1690 1691 int 1692 session_graceful_stop(struct peer *p) 1693 { 1694 u_int8_t i; 1695 1696 for (i = 0; i < AID_MAX; i++) { 1697 /* 1698 * Only flush if the peer is restarting and the timeout fired. 1699 * In all other cases the session was already flushed when the 1700 * session went down or when the new open message was parsed. 1701 */ 1702 if (p->capa.neg.grestart.flags[i] & CAPA_GR_RESTARTING) { 1703 log_peer_warnx(&p->conf, "graceful restart of %s, " 1704 "time-out, flushing", aid2str(i)); 1705 if (imsg_rde(IMSG_SESSION_FLUSH, p->conf.id, 1706 &i, sizeof(i)) == -1) 1707 return (-1); 1708 } 1709 p->capa.neg.grestart.flags[i] &= ~CAPA_GR_RESTARTING; 1710 } 1711 return (0); 1712 } 1713 1714 int 1715 session_dispatch_msg(struct pollfd *pfd, struct peer *p) 1716 { 1717 ssize_t n; 1718 socklen_t len; 1719 int error; 1720 1721 if (p->state == STATE_CONNECT) { 1722 if (pfd->revents & POLLOUT) { 1723 if (pfd->revents & POLLIN) { 1724 /* error occurred */ 1725 len = sizeof(error); 1726 if (getsockopt(pfd->fd, SOL_SOCKET, SO_ERROR, 1727 &error, &len) == -1 || error) { 1728 if (error) 1729 errno = error; 1730 if (errno != p->lasterr) { 1731 log_peer_warn(&p->conf, 1732 "socket error"); 1733 p->lasterr = errno; 1734 } 1735 bgp_fsm(p, EVNT_CON_OPENFAIL); 1736 return (1); 1737 } 1738 } 1739 bgp_fsm(p, EVNT_CON_OPEN); 1740 return (1); 1741 } 1742 if (pfd->revents & POLLHUP) { 1743 bgp_fsm(p, EVNT_CON_OPENFAIL); 1744 return (1); 1745 } 1746 if (pfd->revents & (POLLERR|POLLNVAL)) { 1747 bgp_fsm(p, EVNT_CON_FATAL); 1748 return (1); 1749 } 1750 return (0); 1751 } 1752 1753 if (pfd->revents & POLLHUP) { 1754 bgp_fsm(p, EVNT_CON_CLOSED); 1755 return (1); 1756 } 1757 if (pfd->revents & (POLLERR|POLLNVAL)) { 1758 bgp_fsm(p, EVNT_CON_FATAL); 1759 return (1); 1760 } 1761 1762 if (pfd->revents & POLLOUT && p->wbuf.queued) { 1763 if ((error = msgbuf_write(&p->wbuf)) <= 0 && errno != EAGAIN) { 1764 if (error == 0) 1765 log_peer_warnx(&p->conf, "Connection closed"); 1766 else if (error == -1) 1767 log_peer_warn(&p->conf, "write error"); 1768 bgp_fsm(p, EVNT_CON_FATAL); 1769 return (1); 1770 } 1771 p->stats.last_write = getmonotime(); 1772 if (p->throttled && p->wbuf.queued < SESS_MSG_LOW_MARK) { 1773 if (imsg_rde(IMSG_XON, p->conf.id, NULL, 0) == -1) 1774 log_peer_warn(&p->conf, "imsg_compose XON"); 1775 else 1776 p->throttled = 0; 1777 } 1778 if (!(pfd->revents & POLLIN)) 1779 return (1); 1780 } 1781 1782 if (p->rbuf && pfd->revents & POLLIN) { 1783 if ((n = read(p->fd, p->rbuf->buf + p->rbuf->wpos, 1784 sizeof(p->rbuf->buf) - p->rbuf->wpos)) == -1) { 1785 if (errno != EINTR && errno != EAGAIN) { 1786 log_peer_warn(&p->conf, "read error"); 1787 bgp_fsm(p, EVNT_CON_FATAL); 1788 } 1789 return (1); 1790 } 1791 if (n == 0) { /* connection closed */ 1792 bgp_fsm(p, EVNT_CON_CLOSED); 1793 return (1); 1794 } 1795 1796 p->rbuf->wpos += n; 1797 p->stats.last_read = getmonotime(); 1798 return (1); 1799 } 1800 return (0); 1801 } 1802 1803 void 1804 session_process_msg(struct peer *p) 1805 { 1806 struct mrt *mrt; 1807 ssize_t rpos, av, left; 1808 int processed = 0; 1809 u_int16_t msglen; 1810 u_int8_t msgtype; 1811 1812 rpos = 0; 1813 av = p->rbuf->wpos; 1814 p->rpending = 0; 1815 1816 /* 1817 * session might drop to IDLE -> buffers deallocated 1818 * we MUST check rbuf != NULL before use 1819 */ 1820 for (;;) { 1821 if (p->rbuf == NULL) 1822 return; 1823 if (rpos + MSGSIZE_HEADER > av) 1824 break; 1825 if (parse_header(p, p->rbuf->buf + rpos, &msglen, 1826 &msgtype) == -1) 1827 return; 1828 if (rpos + msglen > av) 1829 break; 1830 p->rbuf->rptr = p->rbuf->buf + rpos; 1831 1832 /* dump to MRT as soon as we have a full packet */ 1833 LIST_FOREACH(mrt, &mrthead, entry) { 1834 if (!(mrt->type == MRT_ALL_IN || (msgtype == UPDATE && 1835 mrt->type == MRT_UPDATE_IN))) 1836 continue; 1837 if ((mrt->peer_id == 0 && mrt->group_id == 0) || 1838 mrt->peer_id == p->conf.id || (mrt->group_id != 0 && 1839 mrt->group_id == p->conf.groupid)) 1840 mrt_dump_bgp_msg(mrt, p->rbuf->rptr, msglen, p); 1841 } 1842 1843 switch (msgtype) { 1844 case OPEN: 1845 bgp_fsm(p, EVNT_RCVD_OPEN); 1846 p->stats.msg_rcvd_open++; 1847 break; 1848 case UPDATE: 1849 bgp_fsm(p, EVNT_RCVD_UPDATE); 1850 p->stats.msg_rcvd_update++; 1851 break; 1852 case NOTIFICATION: 1853 bgp_fsm(p, EVNT_RCVD_NOTIFICATION); 1854 p->stats.msg_rcvd_notification++; 1855 break; 1856 case KEEPALIVE: 1857 bgp_fsm(p, EVNT_RCVD_KEEPALIVE); 1858 p->stats.msg_rcvd_keepalive++; 1859 break; 1860 case RREFRESH: 1861 parse_refresh(p); 1862 p->stats.msg_rcvd_rrefresh++; 1863 break; 1864 default: /* cannot happen */ 1865 session_notification(p, ERR_HEADER, ERR_HDR_TYPE, 1866 &msgtype, 1); 1867 log_warnx("received message with unknown type %u", 1868 msgtype); 1869 bgp_fsm(p, EVNT_CON_FATAL); 1870 } 1871 rpos += msglen; 1872 if (++processed > MSG_PROCESS_LIMIT) { 1873 p->rpending = 1; 1874 break; 1875 } 1876 } 1877 1878 if (rpos < av) { 1879 left = av - rpos; 1880 memmove(&p->rbuf->buf, p->rbuf->buf + rpos, left); 1881 p->rbuf->wpos = left; 1882 } else 1883 p->rbuf->wpos = 0; 1884 } 1885 1886 int 1887 parse_header(struct peer *peer, u_char *data, u_int16_t *len, u_int8_t *type) 1888 { 1889 u_char *p; 1890 u_int16_t olen; 1891 static const u_int8_t marker[MSGSIZE_HEADER_MARKER] = { 0xff, 0xff, 1892 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 1893 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff }; 1894 1895 /* caller MUST make sure we are getting 19 bytes! */ 1896 p = data; 1897 if (memcmp(p, marker, sizeof(marker))) { 1898 log_peer_warnx(&peer->conf, "sync error"); 1899 session_notification(peer, ERR_HEADER, ERR_HDR_SYNC, NULL, 0); 1900 bgp_fsm(peer, EVNT_CON_FATAL); 1901 return (-1); 1902 } 1903 p += MSGSIZE_HEADER_MARKER; 1904 1905 memcpy(&olen, p, 2); 1906 *len = ntohs(olen); 1907 p += 2; 1908 memcpy(type, p, 1); 1909 1910 if (*len < MSGSIZE_HEADER || *len > MAX_PKTSIZE) { 1911 log_peer_warnx(&peer->conf, 1912 "received message: illegal length: %u byte", *len); 1913 session_notification(peer, ERR_HEADER, ERR_HDR_LEN, 1914 &olen, sizeof(olen)); 1915 bgp_fsm(peer, EVNT_CON_FATAL); 1916 return (-1); 1917 } 1918 1919 switch (*type) { 1920 case OPEN: 1921 if (*len < MSGSIZE_OPEN_MIN) { 1922 log_peer_warnx(&peer->conf, 1923 "received OPEN: illegal len: %u byte", *len); 1924 session_notification(peer, ERR_HEADER, ERR_HDR_LEN, 1925 &olen, sizeof(olen)); 1926 bgp_fsm(peer, EVNT_CON_FATAL); 1927 return (-1); 1928 } 1929 break; 1930 case NOTIFICATION: 1931 if (*len < MSGSIZE_NOTIFICATION_MIN) { 1932 log_peer_warnx(&peer->conf, 1933 "received NOTIFICATION: illegal len: %u byte", 1934 *len); 1935 session_notification(peer, ERR_HEADER, ERR_HDR_LEN, 1936 &olen, sizeof(olen)); 1937 bgp_fsm(peer, EVNT_CON_FATAL); 1938 return (-1); 1939 } 1940 break; 1941 case UPDATE: 1942 if (*len < MSGSIZE_UPDATE_MIN) { 1943 log_peer_warnx(&peer->conf, 1944 "received UPDATE: illegal len: %u byte", *len); 1945 session_notification(peer, ERR_HEADER, ERR_HDR_LEN, 1946 &olen, sizeof(olen)); 1947 bgp_fsm(peer, EVNT_CON_FATAL); 1948 return (-1); 1949 } 1950 break; 1951 case KEEPALIVE: 1952 if (*len != MSGSIZE_KEEPALIVE) { 1953 log_peer_warnx(&peer->conf, 1954 "received KEEPALIVE: illegal len: %u byte", *len); 1955 session_notification(peer, ERR_HEADER, ERR_HDR_LEN, 1956 &olen, sizeof(olen)); 1957 bgp_fsm(peer, EVNT_CON_FATAL); 1958 return (-1); 1959 } 1960 break; 1961 case RREFRESH: 1962 if (*len != MSGSIZE_RREFRESH) { 1963 log_peer_warnx(&peer->conf, 1964 "received RREFRESH: illegal len: %u byte", *len); 1965 session_notification(peer, ERR_HEADER, ERR_HDR_LEN, 1966 &olen, sizeof(olen)); 1967 bgp_fsm(peer, EVNT_CON_FATAL); 1968 return (-1); 1969 } 1970 break; 1971 default: 1972 log_peer_warnx(&peer->conf, 1973 "received msg with unknown type %u", *type); 1974 session_notification(peer, ERR_HEADER, ERR_HDR_TYPE, 1975 type, 1); 1976 bgp_fsm(peer, EVNT_CON_FATAL); 1977 return (-1); 1978 } 1979 return (0); 1980 } 1981 1982 int 1983 parse_open(struct peer *peer) 1984 { 1985 u_char *p, *op_val; 1986 u_int8_t version, rversion; 1987 u_int16_t short_as, msglen; 1988 u_int16_t holdtime, oholdtime, myholdtime; 1989 u_int32_t as, bgpid; 1990 u_int8_t optparamlen, plen; 1991 u_int8_t op_type, op_len; 1992 1993 p = peer->rbuf->rptr; 1994 p += MSGSIZE_HEADER_MARKER; 1995 memcpy(&msglen, p, sizeof(msglen)); 1996 msglen = ntohs(msglen); 1997 1998 p = peer->rbuf->rptr; 1999 p += MSGSIZE_HEADER; /* header is already checked */ 2000 2001 memcpy(&version, p, sizeof(version)); 2002 p += sizeof(version); 2003 2004 if (version != BGP_VERSION) { 2005 log_peer_warnx(&peer->conf, 2006 "peer wants unrecognized version %u", version); 2007 if (version > BGP_VERSION) 2008 rversion = version - BGP_VERSION; 2009 else 2010 rversion = BGP_VERSION; 2011 session_notification(peer, ERR_OPEN, ERR_OPEN_VERSION, 2012 &rversion, sizeof(rversion)); 2013 change_state(peer, STATE_IDLE, EVNT_RCVD_OPEN); 2014 return (-1); 2015 } 2016 2017 memcpy(&short_as, p, sizeof(short_as)); 2018 p += sizeof(short_as); 2019 as = peer->short_as = ntohs(short_as); 2020 if (as == 0) { 2021 log_peer_warnx(&peer->conf, 2022 "peer requests unacceptable AS %u", as); 2023 session_notification(peer, ERR_OPEN, ERR_OPEN_AS, 2024 NULL, 0); 2025 change_state(peer, STATE_IDLE, EVNT_RCVD_OPEN); 2026 return (-1); 2027 } 2028 2029 memcpy(&oholdtime, p, sizeof(oholdtime)); 2030 p += sizeof(oholdtime); 2031 2032 holdtime = ntohs(oholdtime); 2033 if (holdtime && holdtime < peer->conf.min_holdtime) { 2034 log_peer_warnx(&peer->conf, 2035 "peer requests unacceptable holdtime %u", holdtime); 2036 session_notification(peer, ERR_OPEN, ERR_OPEN_HOLDTIME, 2037 NULL, 0); 2038 change_state(peer, STATE_IDLE, EVNT_RCVD_OPEN); 2039 return (-1); 2040 } 2041 2042 myholdtime = peer->conf.holdtime; 2043 if (!myholdtime) 2044 myholdtime = conf->holdtime; 2045 if (holdtime < myholdtime) 2046 peer->holdtime = holdtime; 2047 else 2048 peer->holdtime = myholdtime; 2049 2050 memcpy(&bgpid, p, sizeof(bgpid)); 2051 p += sizeof(bgpid); 2052 2053 /* check bgpid for validity - just disallow 0 */ 2054 if (ntohl(bgpid) == 0) { 2055 log_peer_warnx(&peer->conf, "peer BGPID %u unacceptable", 2056 ntohl(bgpid)); 2057 session_notification(peer, ERR_OPEN, ERR_OPEN_BGPID, 2058 NULL, 0); 2059 change_state(peer, STATE_IDLE, EVNT_RCVD_OPEN); 2060 return (-1); 2061 } 2062 peer->remote_bgpid = bgpid; 2063 2064 memcpy(&optparamlen, p, sizeof(optparamlen)); 2065 p += sizeof(optparamlen); 2066 2067 if (optparamlen != msglen - MSGSIZE_OPEN_MIN) { 2068 log_peer_warnx(&peer->conf, 2069 "corrupt OPEN message received: length mismatch"); 2070 session_notification(peer, ERR_OPEN, 0, NULL, 0); 2071 change_state(peer, STATE_IDLE, EVNT_RCVD_OPEN); 2072 return (-1); 2073 } 2074 2075 plen = optparamlen; 2076 while (plen > 0) { 2077 if (plen < 2) { 2078 log_peer_warnx(&peer->conf, 2079 "corrupt OPEN message received, len wrong"); 2080 session_notification(peer, ERR_OPEN, 0, NULL, 0); 2081 change_state(peer, STATE_IDLE, EVNT_RCVD_OPEN); 2082 return (-1); 2083 } 2084 memcpy(&op_type, p, sizeof(op_type)); 2085 p += sizeof(op_type); 2086 plen -= sizeof(op_type); 2087 memcpy(&op_len, p, sizeof(op_len)); 2088 p += sizeof(op_len); 2089 plen -= sizeof(op_len); 2090 if (op_len > 0) { 2091 if (plen < op_len) { 2092 log_peer_warnx(&peer->conf, 2093 "corrupt OPEN message received, len wrong"); 2094 session_notification(peer, ERR_OPEN, 0, 2095 NULL, 0); 2096 change_state(peer, STATE_IDLE, EVNT_RCVD_OPEN); 2097 return (-1); 2098 } 2099 op_val = p; 2100 p += op_len; 2101 plen -= op_len; 2102 } else 2103 op_val = NULL; 2104 2105 switch (op_type) { 2106 case OPT_PARAM_CAPABILITIES: /* RFC 3392 */ 2107 if (parse_capabilities(peer, op_val, op_len, 2108 &as) == -1) { 2109 session_notification(peer, ERR_OPEN, 0, 2110 NULL, 0); 2111 change_state(peer, STATE_IDLE, EVNT_RCVD_OPEN); 2112 return (-1); 2113 } 2114 break; 2115 case OPT_PARAM_AUTH: /* deprecated */ 2116 default: 2117 /* 2118 * unsupported type 2119 * the RFCs tell us to leave the data section empty 2120 * and notify the peer with ERR_OPEN, ERR_OPEN_OPT. 2121 * How the peer should know _which_ optional parameter 2122 * we don't support is beyond me. 2123 */ 2124 log_peer_warnx(&peer->conf, 2125 "received OPEN message with unsupported optional " 2126 "parameter: type %u", op_type); 2127 session_notification(peer, ERR_OPEN, ERR_OPEN_OPT, 2128 NULL, 0); 2129 change_state(peer, STATE_IDLE, EVNT_RCVD_OPEN); 2130 timer_set(peer, Timer_IdleHold, 0); /* no punish */ 2131 peer->IdleHoldTime /= 2; 2132 return (-1); 2133 } 2134 } 2135 2136 /* if remote-as is zero and it's a cloned neighbor, accept any */ 2137 if (peer->template && !peer->conf.remote_as && as != AS_TRANS) { 2138 peer->conf.remote_as = as; 2139 peer->conf.ebgp = (peer->conf.remote_as != peer->conf.local_as); 2140 if (!peer->conf.ebgp) 2141 /* force enforce_as off for iBGP sessions */ 2142 peer->conf.enforce_as = ENFORCE_AS_OFF; 2143 } 2144 2145 if (peer->conf.remote_as != as) { 2146 log_peer_warnx(&peer->conf, "peer sent wrong AS %s", 2147 log_as(as)); 2148 session_notification(peer, ERR_OPEN, ERR_OPEN_AS, NULL, 0); 2149 change_state(peer, STATE_IDLE, EVNT_RCVD_OPEN); 2150 return (-1); 2151 } 2152 2153 if (capa_neg_calc(peer) == -1) { 2154 log_peer_warnx(&peer->conf, 2155 "capability negotiation calculation failed"); 2156 session_notification(peer, ERR_OPEN, 0, NULL, 0); 2157 change_state(peer, STATE_IDLE, EVNT_RCVD_OPEN); 2158 return (-1); 2159 } 2160 2161 return (0); 2162 } 2163 2164 int 2165 parse_update(struct peer *peer) 2166 { 2167 u_char *p; 2168 u_int16_t datalen; 2169 2170 /* 2171 * we pass the message verbatim to the rde. 2172 * in case of errors the whole session is reset with a 2173 * notification anyway, we only need to know the peer 2174 */ 2175 p = peer->rbuf->rptr; 2176 p += MSGSIZE_HEADER_MARKER; 2177 memcpy(&datalen, p, sizeof(datalen)); 2178 datalen = ntohs(datalen); 2179 2180 p = peer->rbuf->rptr; 2181 p += MSGSIZE_HEADER; /* header is already checked */ 2182 datalen -= MSGSIZE_HEADER; 2183 2184 if (imsg_rde(IMSG_UPDATE, peer->conf.id, p, datalen) == -1) 2185 return (-1); 2186 2187 return (0); 2188 } 2189 2190 int 2191 parse_refresh(struct peer *peer) 2192 { 2193 u_char *p; 2194 u_int16_t afi; 2195 u_int8_t aid, safi; 2196 2197 p = peer->rbuf->rptr; 2198 p += MSGSIZE_HEADER; /* header is already checked */ 2199 2200 /* 2201 * We could check if we actually announced the capability but 2202 * as long as the message is correctly encoded we don't care. 2203 */ 2204 2205 /* afi, 2 byte */ 2206 memcpy(&afi, p, sizeof(afi)); 2207 afi = ntohs(afi); 2208 p += 2; 2209 /* reserved, 1 byte */ 2210 p += 1; 2211 /* safi, 1 byte */ 2212 memcpy(&safi, p, sizeof(safi)); 2213 2214 /* afi/safi unchecked - unrecognized values will be ignored anyway */ 2215 if (afi2aid(afi, safi, &aid) == -1) { 2216 log_peer_warnx(&peer->conf, "peer sent bad refresh, " 2217 "invalid afi/safi pair"); 2218 return (0); 2219 } 2220 2221 if (imsg_rde(IMSG_REFRESH, peer->conf.id, &aid, sizeof(aid)) == -1) 2222 return (-1); 2223 2224 return (0); 2225 } 2226 2227 int 2228 parse_notification(struct peer *peer) 2229 { 2230 u_char *p; 2231 u_int16_t datalen; 2232 u_int8_t errcode; 2233 u_int8_t subcode; 2234 u_int8_t capa_code; 2235 u_int8_t capa_len; 2236 size_t shutcomm_len; 2237 u_int8_t i; 2238 2239 /* just log */ 2240 p = peer->rbuf->rptr; 2241 p += MSGSIZE_HEADER_MARKER; 2242 memcpy(&datalen, p, sizeof(datalen)); 2243 datalen = ntohs(datalen); 2244 2245 p = peer->rbuf->rptr; 2246 p += MSGSIZE_HEADER; /* header is already checked */ 2247 datalen -= MSGSIZE_HEADER; 2248 2249 memcpy(&errcode, p, sizeof(errcode)); 2250 p += sizeof(errcode); 2251 datalen -= sizeof(errcode); 2252 2253 memcpy(&subcode, p, sizeof(subcode)); 2254 p += sizeof(subcode); 2255 datalen -= sizeof(subcode); 2256 2257 log_notification(peer, errcode, subcode, p, datalen, "received"); 2258 peer->errcnt++; 2259 peer->stats.last_rcvd_errcode = errcode; 2260 peer->stats.last_rcvd_suberr = subcode; 2261 2262 if (errcode == ERR_OPEN && subcode == ERR_OPEN_CAPA) { 2263 if (datalen == 0) { /* zebra likes to send those.. humbug */ 2264 log_peer_warnx(&peer->conf, "received \"unsupported " 2265 "capability\" notification without data part, " 2266 "disabling capability announcements altogether"); 2267 session_capa_ann_none(peer); 2268 } 2269 2270 while (datalen > 0) { 2271 if (datalen < 2) { 2272 log_peer_warnx(&peer->conf, 2273 "parse_notification: " 2274 "expect len >= 2, len is %u", datalen); 2275 return (-1); 2276 } 2277 memcpy(&capa_code, p, sizeof(capa_code)); 2278 p += sizeof(capa_code); 2279 datalen -= sizeof(capa_code); 2280 memcpy(&capa_len, p, sizeof(capa_len)); 2281 p += sizeof(capa_len); 2282 datalen -= sizeof(capa_len); 2283 if (datalen < capa_len) { 2284 log_peer_warnx(&peer->conf, 2285 "parse_notification: capa_len %u exceeds " 2286 "remaining msg length %u", capa_len, 2287 datalen); 2288 return (-1); 2289 } 2290 p += capa_len; 2291 datalen -= capa_len; 2292 switch (capa_code) { 2293 case CAPA_MP: 2294 for (i = 0; i < AID_MAX; i++) 2295 peer->capa.ann.mp[i] = 0; 2296 log_peer_warnx(&peer->conf, 2297 "disabling multiprotocol capability"); 2298 break; 2299 case CAPA_REFRESH: 2300 peer->capa.ann.refresh = 0; 2301 log_peer_warnx(&peer->conf, 2302 "disabling route refresh capability"); 2303 break; 2304 case CAPA_RESTART: 2305 peer->capa.ann.grestart.restart = 0; 2306 log_peer_warnx(&peer->conf, 2307 "disabling restart capability"); 2308 break; 2309 case CAPA_AS4BYTE: 2310 peer->capa.ann.as4byte = 0; 2311 log_peer_warnx(&peer->conf, 2312 "disabling 4-byte AS num capability"); 2313 break; 2314 default: /* should not happen... */ 2315 log_peer_warnx(&peer->conf, "received " 2316 "\"unsupported capability\" notification " 2317 "for unknown capability %u, disabling " 2318 "capability announcements altogether", 2319 capa_code); 2320 session_capa_ann_none(peer); 2321 break; 2322 } 2323 } 2324 2325 return (1); 2326 } 2327 2328 if (errcode == ERR_OPEN && subcode == ERR_OPEN_OPT) { 2329 session_capa_ann_none(peer); 2330 return (1); 2331 } 2332 2333 if (errcode == ERR_CEASE && 2334 (subcode == ERR_CEASE_ADMIN_DOWN || 2335 subcode == ERR_CEASE_ADMIN_RESET)) { 2336 if (datalen > 1) { 2337 shutcomm_len = *p++; 2338 datalen--; 2339 if (datalen < shutcomm_len) { 2340 log_peer_warnx(&peer->conf, 2341 "received truncated shutdown reason"); 2342 return (0); 2343 } 2344 if (shutcomm_len > SHUT_COMM_LEN - 1) { 2345 log_peer_warnx(&peer->conf, 2346 "received overly long shutdown reason"); 2347 return (0); 2348 } 2349 memcpy(peer->stats.last_shutcomm, p, shutcomm_len); 2350 peer->stats.last_shutcomm[shutcomm_len] = '\0'; 2351 log_peer_warnx(&peer->conf, 2352 "received shutdown reason: \"%s\"", 2353 log_shutcomm(peer->stats.last_shutcomm)); 2354 p += shutcomm_len; 2355 datalen -= shutcomm_len; 2356 } 2357 } 2358 2359 return (0); 2360 } 2361 2362 int 2363 parse_capabilities(struct peer *peer, u_char *d, u_int16_t dlen, u_int32_t *as) 2364 { 2365 u_char *capa_val; 2366 u_int32_t remote_as; 2367 u_int16_t len; 2368 u_int16_t afi; 2369 u_int16_t gr_header; 2370 u_int8_t safi; 2371 u_int8_t aid; 2372 u_int8_t gr_flags; 2373 u_int8_t capa_code; 2374 u_int8_t capa_len; 2375 u_int8_t i; 2376 2377 len = dlen; 2378 while (len > 0) { 2379 if (len < 2) { 2380 log_peer_warnx(&peer->conf, "Bad capabilities attr " 2381 "length: %u, too short", len); 2382 return (-1); 2383 } 2384 memcpy(&capa_code, d, sizeof(capa_code)); 2385 d += sizeof(capa_code); 2386 len -= sizeof(capa_code); 2387 memcpy(&capa_len, d, sizeof(capa_len)); 2388 d += sizeof(capa_len); 2389 len -= sizeof(capa_len); 2390 if (capa_len > 0) { 2391 if (len < capa_len) { 2392 log_peer_warnx(&peer->conf, 2393 "Bad capabilities attr length: " 2394 "len %u smaller than capa_len %u", 2395 len, capa_len); 2396 return (-1); 2397 } 2398 capa_val = d; 2399 d += capa_len; 2400 len -= capa_len; 2401 } else 2402 capa_val = NULL; 2403 2404 switch (capa_code) { 2405 case CAPA_MP: /* RFC 4760 */ 2406 if (capa_len != 4) { 2407 log_peer_warnx(&peer->conf, 2408 "Bad multi protocol capability length: " 2409 "%u", capa_len); 2410 break; 2411 } 2412 memcpy(&afi, capa_val, sizeof(afi)); 2413 afi = ntohs(afi); 2414 memcpy(&safi, capa_val + 3, sizeof(safi)); 2415 if (afi2aid(afi, safi, &aid) == -1) { 2416 log_peer_warnx(&peer->conf, 2417 "Received multi protocol capability: " 2418 " unknown AFI %u, safi %u pair", 2419 afi, safi); 2420 break; 2421 } 2422 peer->capa.peer.mp[aid] = 1; 2423 break; 2424 case CAPA_REFRESH: 2425 peer->capa.peer.refresh = 1; 2426 break; 2427 case CAPA_RESTART: 2428 if (capa_len == 2) { 2429 /* peer only supports EoR marker */ 2430 peer->capa.peer.grestart.restart = 1; 2431 peer->capa.peer.grestart.timeout = 0; 2432 break; 2433 } else if (capa_len % 4 != 2) { 2434 log_peer_warnx(&peer->conf, 2435 "Bad graceful restart capability length: " 2436 "%u", capa_len); 2437 peer->capa.peer.grestart.restart = 0; 2438 peer->capa.peer.grestart.timeout = 0; 2439 break; 2440 } 2441 2442 memcpy(&gr_header, capa_val, sizeof(gr_header)); 2443 gr_header = ntohs(gr_header); 2444 peer->capa.peer.grestart.timeout = 2445 gr_header & CAPA_GR_TIMEMASK; 2446 if (peer->capa.peer.grestart.timeout == 0) { 2447 log_peer_warnx(&peer->conf, "Received " 2448 "graceful restart timeout is zero"); 2449 peer->capa.peer.grestart.restart = 0; 2450 break; 2451 } 2452 2453 for (i = 2; i <= capa_len - 4; i += 4) { 2454 memcpy(&afi, capa_val + i, sizeof(afi)); 2455 afi = ntohs(afi); 2456 memcpy(&safi, capa_val + i + 2, sizeof(safi)); 2457 if (afi2aid(afi, safi, &aid) == -1) { 2458 log_peer_warnx(&peer->conf, 2459 "Received graceful restart capa: " 2460 " unknown AFI %u, safi %u pair", 2461 afi, safi); 2462 continue; 2463 } 2464 memcpy(&gr_flags, capa_val + i + 3, 2465 sizeof(gr_flags)); 2466 peer->capa.peer.grestart.flags[aid] |= 2467 CAPA_GR_PRESENT; 2468 if (gr_flags & CAPA_GR_F_FLAG) 2469 peer->capa.peer.grestart.flags[aid] |= 2470 CAPA_GR_FORWARD; 2471 if (gr_header & CAPA_GR_R_FLAG) 2472 peer->capa.peer.grestart.flags[aid] |= 2473 CAPA_GR_RESTART; 2474 peer->capa.peer.grestart.restart = 2; 2475 } 2476 break; 2477 case CAPA_AS4BYTE: 2478 if (capa_len != 4) { 2479 log_peer_warnx(&peer->conf, 2480 "Bad AS4BYTE capability length: " 2481 "%u", capa_len); 2482 peer->capa.peer.as4byte = 0; 2483 break; 2484 } 2485 memcpy(&remote_as, capa_val, sizeof(remote_as)); 2486 *as = ntohl(remote_as); 2487 if (*as == 0) { 2488 log_peer_warnx(&peer->conf, 2489 "peer requests unacceptable AS %u", *as); 2490 session_notification(peer, ERR_OPEN, 2491 ERR_OPEN_AS, NULL, 0); 2492 change_state(peer, STATE_IDLE, EVNT_RCVD_OPEN); 2493 return (-1); 2494 } 2495 peer->capa.peer.as4byte = 1; 2496 break; 2497 default: 2498 break; 2499 } 2500 } 2501 2502 return (0); 2503 } 2504 2505 int 2506 capa_neg_calc(struct peer *p) 2507 { 2508 u_int8_t i, hasmp = 0; 2509 2510 /* refresh: does not realy matter here, use peer setting */ 2511 p->capa.neg.refresh = p->capa.peer.refresh; 2512 2513 /* as4byte: both side must announce capability */ 2514 if (p->capa.ann.as4byte && p->capa.peer.as4byte) 2515 p->capa.neg.as4byte = 1; 2516 else 2517 p->capa.neg.as4byte = 0; 2518 2519 /* MP: both side must announce capability */ 2520 for (i = 0; i < AID_MAX; i++) { 2521 if (p->capa.ann.mp[i] && p->capa.peer.mp[i]) { 2522 p->capa.neg.mp[i] = 1; 2523 hasmp = 1; 2524 } else 2525 p->capa.neg.mp[i] = 0; 2526 } 2527 /* if no MP capability present default to IPv4 unicast mode */ 2528 if (!hasmp) 2529 p->capa.neg.mp[AID_INET] = 1; 2530 2531 /* 2532 * graceful restart: only the peer capabilities are of interest here. 2533 * It is necessary to compare the new values with the previous ones 2534 * and act acordingly. AFI/SAFI that are not part in the MP capability 2535 * are treated as not being present. 2536 */ 2537 2538 for (i = 0; i < AID_MAX; i++) { 2539 int8_t negflags; 2540 2541 /* disable GR if the AFI/SAFI is not present */ 2542 if (p->capa.peer.grestart.flags[i] & CAPA_GR_PRESENT && 2543 p->capa.neg.mp[i] == 0) 2544 p->capa.peer.grestart.flags[i] = 0; /* disable */ 2545 /* look at current GR state and decide what to do */ 2546 negflags = p->capa.neg.grestart.flags[i]; 2547 p->capa.neg.grestart.flags[i] = p->capa.peer.grestart.flags[i]; 2548 if (negflags & CAPA_GR_RESTARTING) { 2549 if (!(p->capa.peer.grestart.flags[i] & 2550 CAPA_GR_FORWARD)) { 2551 if (imsg_rde(IMSG_SESSION_FLUSH, p->conf.id, 2552 &i, sizeof(i)) == -1) 2553 return (-1); 2554 log_peer_warnx(&p->conf, "graceful restart of " 2555 "%s, not restarted, flushing", aid2str(i)); 2556 } else 2557 p->capa.neg.grestart.flags[i] |= 2558 CAPA_GR_RESTARTING; 2559 } 2560 } 2561 p->capa.neg.grestart.timeout = p->capa.peer.grestart.timeout; 2562 p->capa.neg.grestart.restart = p->capa.peer.grestart.restart; 2563 2564 return (0); 2565 } 2566 2567 void 2568 session_dispatch_imsg(struct imsgbuf *ibuf, int idx, u_int *listener_cnt) 2569 { 2570 struct imsg imsg; 2571 struct mrt xmrt; 2572 struct mrt *mrt; 2573 struct imsgbuf *i; 2574 struct peer *p; 2575 struct listen_addr *la, *nla; 2576 struct kif *kif; 2577 u_char *data; 2578 int n, fd, depend_ok, restricted; 2579 u_int16_t t; 2580 u_int8_t aid, errcode, subcode; 2581 2582 while (ibuf) { 2583 if ((n = imsg_get(ibuf, &imsg)) == -1) 2584 fatal("session_dispatch_imsg: imsg_get error"); 2585 2586 if (n == 0) 2587 break; 2588 2589 switch (imsg.hdr.type) { 2590 case IMSG_SOCKET_CONN: 2591 case IMSG_SOCKET_CONN_CTL: 2592 if (idx != PFD_PIPE_MAIN) 2593 fatalx("reconf request not from parent"); 2594 if ((fd = imsg.fd) == -1) { 2595 log_warnx("expected to receive imsg fd to " 2596 "RDE but didn't receive any"); 2597 break; 2598 } 2599 if ((i = malloc(sizeof(struct imsgbuf))) == NULL) 2600 fatal(NULL); 2601 imsg_init(i, fd); 2602 if (imsg.hdr.type == IMSG_SOCKET_CONN) { 2603 if (ibuf_rde) { 2604 log_warnx("Unexpected imsg connection " 2605 "to RDE received"); 2606 msgbuf_clear(&ibuf_rde->w); 2607 free(ibuf_rde); 2608 } 2609 ibuf_rde = i; 2610 } else { 2611 if (ibuf_rde_ctl) { 2612 log_warnx("Unexpected imsg ctl " 2613 "connection to RDE received"); 2614 msgbuf_clear(&ibuf_rde_ctl->w); 2615 free(ibuf_rde_ctl); 2616 } 2617 ibuf_rde_ctl = i; 2618 } 2619 break; 2620 case IMSG_RECONF_CONF: 2621 if (idx != PFD_PIPE_MAIN) 2622 fatalx("reconf request not from parent"); 2623 nconf = new_config(); 2624 2625 copy_config(nconf, imsg.data); 2626 pending_reconf = 1; 2627 break; 2628 case IMSG_RECONF_PEER: 2629 if (idx != PFD_PIPE_MAIN) 2630 fatalx("reconf request not from parent"); 2631 if ((p = calloc(1, sizeof(struct peer))) == NULL) 2632 fatal("new_peer"); 2633 memcpy(&p->conf, imsg.data, sizeof(struct peer_config)); 2634 p->state = p->prev_state = STATE_NONE; 2635 p->reconf_action = RECONF_REINIT; 2636 if (RB_INSERT(peer_head, &nconf->peers, p) != NULL) 2637 fatalx("%s: peer tree is corrupt", __func__); 2638 break; 2639 case IMSG_RECONF_LISTENER: 2640 if (idx != PFD_PIPE_MAIN) 2641 fatalx("reconf request not from parent"); 2642 if (nconf == NULL) 2643 fatalx("IMSG_RECONF_LISTENER but no config"); 2644 nla = imsg.data; 2645 TAILQ_FOREACH(la, conf->listen_addrs, entry) 2646 if (!la_cmp(la, nla)) 2647 break; 2648 2649 if (la == NULL) { 2650 if (nla->reconf != RECONF_REINIT) 2651 fatalx("king bula sez: " 2652 "expected REINIT"); 2653 2654 if ((nla->fd = imsg.fd) == -1) 2655 log_warnx("expected to receive fd for " 2656 "%s but didn't receive any", 2657 log_sockaddr((struct sockaddr *) 2658 &nla->sa, nla->sa_len)); 2659 2660 la = calloc(1, sizeof(struct listen_addr)); 2661 if (la == NULL) 2662 fatal(NULL); 2663 memcpy(&la->sa, &nla->sa, sizeof(la->sa)); 2664 la->flags = nla->flags; 2665 la->fd = nla->fd; 2666 la->reconf = RECONF_REINIT; 2667 TAILQ_INSERT_TAIL(nconf->listen_addrs, la, 2668 entry); 2669 } else { 2670 if (nla->reconf != RECONF_KEEP) 2671 fatalx("king bula sez: expected KEEP"); 2672 la->reconf = RECONF_KEEP; 2673 } 2674 2675 break; 2676 case IMSG_RECONF_CTRL: 2677 if (idx != PFD_PIPE_MAIN) 2678 fatalx("reconf request not from parent"); 2679 if (imsg.hdr.len != IMSG_HEADER_SIZE + 2680 sizeof(restricted)) 2681 fatalx("IFINFO imsg with wrong len"); 2682 memcpy(&restricted, imsg.data, sizeof(restricted)); 2683 if (imsg.fd == -1) { 2684 log_warnx("expected to receive fd for control " 2685 "socket but didn't receive any"); 2686 break; 2687 } 2688 if (restricted) { 2689 control_shutdown(rcsock); 2690 rcsock = imsg.fd; 2691 } else { 2692 control_shutdown(csock); 2693 csock = imsg.fd; 2694 } 2695 break; 2696 case IMSG_RECONF_DRAIN: 2697 if (idx != PFD_PIPE_MAIN) 2698 fatalx("reconf request not from parent"); 2699 imsg_compose(ibuf_main, IMSG_RECONF_DRAIN, 0, 0, 2700 -1, NULL, 0); 2701 break; 2702 case IMSG_RECONF_DONE: 2703 if (idx != PFD_PIPE_MAIN) 2704 fatalx("reconf request not from parent"); 2705 if (nconf == NULL) 2706 fatalx("got IMSG_RECONF_DONE but no config"); 2707 copy_config(conf, nconf); 2708 merge_peers(conf, nconf); 2709 2710 /* delete old listeners */ 2711 for (la = TAILQ_FIRST(conf->listen_addrs); la != NULL; 2712 la = nla) { 2713 nla = TAILQ_NEXT(la, entry); 2714 if (la->reconf == RECONF_NONE) { 2715 log_info("not listening on %s any more", 2716 log_sockaddr((struct sockaddr *) 2717 &la->sa, la->sa_len)); 2718 TAILQ_REMOVE(conf->listen_addrs, la, 2719 entry); 2720 close(la->fd); 2721 free(la); 2722 } 2723 } 2724 2725 /* add new listeners */ 2726 while ((la = TAILQ_FIRST(nconf->listen_addrs)) != 2727 NULL) { 2728 TAILQ_REMOVE(nconf->listen_addrs, la, entry); 2729 TAILQ_INSERT_TAIL(conf->listen_addrs, la, 2730 entry); 2731 } 2732 2733 setup_listeners(listener_cnt); 2734 free_config(nconf); 2735 nconf = NULL; 2736 pending_reconf = 0; 2737 log_info("SE reconfigured"); 2738 imsg_compose(ibuf_main, IMSG_RECONF_DONE, 0, 0, 2739 -1, NULL, 0); 2740 break; 2741 case IMSG_IFINFO: 2742 if (idx != PFD_PIPE_MAIN) 2743 fatalx("IFINFO message not from parent"); 2744 if (imsg.hdr.len != IMSG_HEADER_SIZE + 2745 sizeof(struct kif)) 2746 fatalx("IFINFO imsg with wrong len"); 2747 kif = imsg.data; 2748 depend_ok = kif->depend_state; 2749 2750 RB_FOREACH(p, peer_head, &conf->peers) 2751 if (!strcmp(p->conf.if_depend, kif->ifname)) { 2752 if (depend_ok && !p->depend_ok) { 2753 p->depend_ok = depend_ok; 2754 bgp_fsm(p, EVNT_START); 2755 } else if (!depend_ok && p->depend_ok) { 2756 p->depend_ok = depend_ok; 2757 session_stop(p, 2758 ERR_CEASE_OTHER_CHANGE); 2759 } 2760 } 2761 break; 2762 case IMSG_MRT_OPEN: 2763 case IMSG_MRT_REOPEN: 2764 if (imsg.hdr.len > IMSG_HEADER_SIZE + 2765 sizeof(struct mrt)) { 2766 log_warnx("wrong imsg len"); 2767 break; 2768 } 2769 2770 memcpy(&xmrt, imsg.data, sizeof(struct mrt)); 2771 if ((xmrt.wbuf.fd = imsg.fd) == -1) 2772 log_warnx("expected to receive fd for mrt dump " 2773 "but didn't receive any"); 2774 2775 mrt = mrt_get(&mrthead, &xmrt); 2776 if (mrt == NULL) { 2777 /* new dump */ 2778 mrt = calloc(1, sizeof(struct mrt)); 2779 if (mrt == NULL) 2780 fatal("session_dispatch_imsg"); 2781 memcpy(mrt, &xmrt, sizeof(struct mrt)); 2782 TAILQ_INIT(&mrt->wbuf.bufs); 2783 LIST_INSERT_HEAD(&mrthead, mrt, entry); 2784 } else { 2785 /* old dump reopened */ 2786 close(mrt->wbuf.fd); 2787 mrt->wbuf.fd = xmrt.wbuf.fd; 2788 } 2789 break; 2790 case IMSG_MRT_CLOSE: 2791 if (imsg.hdr.len > IMSG_HEADER_SIZE + 2792 sizeof(struct mrt)) { 2793 log_warnx("wrong imsg len"); 2794 break; 2795 } 2796 2797 memcpy(&xmrt, imsg.data, sizeof(struct mrt)); 2798 mrt = mrt_get(&mrthead, &xmrt); 2799 if (mrt != NULL) 2800 mrt_done(mrt); 2801 break; 2802 case IMSG_CTL_KROUTE: 2803 case IMSG_CTL_KROUTE_ADDR: 2804 case IMSG_CTL_SHOW_NEXTHOP: 2805 case IMSG_CTL_SHOW_INTERFACE: 2806 case IMSG_CTL_SHOW_FIB_TABLES: 2807 if (idx != PFD_PIPE_MAIN) 2808 fatalx("ctl kroute request not from parent"); 2809 control_imsg_relay(&imsg); 2810 break; 2811 case IMSG_CTL_SHOW_RIB: 2812 case IMSG_CTL_SHOW_RIB_PREFIX: 2813 case IMSG_CTL_SHOW_RIB_COMMUNITIES: 2814 case IMSG_CTL_SHOW_RIB_ATTR: 2815 case IMSG_CTL_SHOW_RIB_MEM: 2816 case IMSG_CTL_SHOW_RIB_HASH: 2817 case IMSG_CTL_SHOW_NETWORK: 2818 case IMSG_CTL_SHOW_NEIGHBOR: 2819 if (idx != PFD_PIPE_ROUTE_CTL) 2820 fatalx("ctl rib request not from RDE"); 2821 control_imsg_relay(&imsg); 2822 break; 2823 case IMSG_CTL_END: 2824 case IMSG_CTL_RESULT: 2825 control_imsg_relay(&imsg); 2826 break; 2827 case IMSG_UPDATE: 2828 if (idx != PFD_PIPE_ROUTE) 2829 fatalx("update request not from RDE"); 2830 if (imsg.hdr.len > IMSG_HEADER_SIZE + 2831 MAX_PKTSIZE - MSGSIZE_HEADER || 2832 imsg.hdr.len < IMSG_HEADER_SIZE + 2833 MSGSIZE_UPDATE_MIN - MSGSIZE_HEADER) 2834 log_warnx("RDE sent invalid update"); 2835 else 2836 session_update(imsg.hdr.peerid, imsg.data, 2837 imsg.hdr.len - IMSG_HEADER_SIZE); 2838 break; 2839 case IMSG_UPDATE_ERR: 2840 if (idx != PFD_PIPE_ROUTE) 2841 fatalx("update request not from RDE"); 2842 if (imsg.hdr.len < IMSG_HEADER_SIZE + 2) { 2843 log_warnx("RDE sent invalid notification"); 2844 break; 2845 } 2846 if ((p = getpeerbyid(conf, imsg.hdr.peerid)) == NULL) { 2847 log_warnx("no such peer: id=%u", 2848 imsg.hdr.peerid); 2849 break; 2850 } 2851 data = imsg.data; 2852 errcode = *data++; 2853 subcode = *data++; 2854 2855 if (imsg.hdr.len == IMSG_HEADER_SIZE + 2) 2856 data = NULL; 2857 2858 session_notification(p, errcode, subcode, 2859 data, imsg.hdr.len - IMSG_HEADER_SIZE - 2); 2860 switch (errcode) { 2861 case ERR_CEASE: 2862 switch (subcode) { 2863 case ERR_CEASE_MAX_PREFIX: 2864 case ERR_CEASE_MAX_SENT_PREFIX: 2865 t = p->conf.max_out_prefix_restart; 2866 if (subcode == ERR_CEASE_MAX_PREFIX) 2867 t = p->conf.max_prefix_restart; 2868 2869 bgp_fsm(p, EVNT_STOP); 2870 if (t) 2871 timer_set(p, Timer_IdleHold, 2872 60 * t); 2873 break; 2874 default: 2875 bgp_fsm(p, EVNT_CON_FATAL); 2876 break; 2877 } 2878 break; 2879 default: 2880 bgp_fsm(p, EVNT_CON_FATAL); 2881 break; 2882 } 2883 break; 2884 case IMSG_SESSION_RESTARTED: 2885 if (idx != PFD_PIPE_ROUTE) 2886 fatalx("update request not from RDE"); 2887 if (imsg.hdr.len < IMSG_HEADER_SIZE + sizeof(aid)) { 2888 log_warnx("RDE sent invalid restart msg"); 2889 break; 2890 } 2891 if ((p = getpeerbyid(conf, imsg.hdr.peerid)) == NULL) { 2892 log_warnx("no such peer: id=%u", 2893 imsg.hdr.peerid); 2894 break; 2895 } 2896 memcpy(&aid, imsg.data, sizeof(aid)); 2897 if (aid >= AID_MAX) 2898 fatalx("IMSG_SESSION_RESTARTED: bad AID"); 2899 if (p->capa.neg.grestart.flags[aid] & 2900 CAPA_GR_RESTARTING) { 2901 log_peer_warnx(&p->conf, 2902 "graceful restart of %s finished", 2903 aid2str(aid)); 2904 p->capa.neg.grestart.flags[aid] &= 2905 ~CAPA_GR_RESTARTING; 2906 timer_stop(p, Timer_RestartTimeout); 2907 2908 /* signal back to RDE to cleanup stale routes */ 2909 if (imsg_rde(IMSG_SESSION_RESTARTED, 2910 imsg.hdr.peerid, &aid, sizeof(aid)) == -1) 2911 fatal("imsg_compose: " 2912 "IMSG_SESSION_RESTARTED"); 2913 } 2914 break; 2915 case IMSG_SESSION_DOWN: 2916 if (idx != PFD_PIPE_ROUTE) 2917 fatalx("update request not from RDE"); 2918 if ((p = getpeerbyid(conf, imsg.hdr.peerid)) == NULL) { 2919 log_warnx("no such peer: id=%u", 2920 imsg.hdr.peerid); 2921 break; 2922 } 2923 session_stop(p, ERR_CEASE_ADMIN_DOWN); 2924 break; 2925 default: 2926 break; 2927 } 2928 imsg_free(&imsg); 2929 } 2930 } 2931 2932 int 2933 la_cmp(struct listen_addr *a, struct listen_addr *b) 2934 { 2935 struct sockaddr_in *in_a, *in_b; 2936 struct sockaddr_in6 *in6_a, *in6_b; 2937 2938 if (a->sa.ss_family != b->sa.ss_family) 2939 return (1); 2940 2941 switch (a->sa.ss_family) { 2942 case AF_INET: 2943 in_a = (struct sockaddr_in *)&a->sa; 2944 in_b = (struct sockaddr_in *)&b->sa; 2945 if (in_a->sin_addr.s_addr != in_b->sin_addr.s_addr) 2946 return (1); 2947 if (in_a->sin_port != in_b->sin_port) 2948 return (1); 2949 break; 2950 case AF_INET6: 2951 in6_a = (struct sockaddr_in6 *)&a->sa; 2952 in6_b = (struct sockaddr_in6 *)&b->sa; 2953 if (bcmp(&in6_a->sin6_addr, &in6_b->sin6_addr, 2954 sizeof(struct in6_addr))) 2955 return (1); 2956 if (in6_a->sin6_port != in6_b->sin6_port) 2957 return (1); 2958 break; 2959 default: 2960 fatal("king bula sez: unknown address family"); 2961 /* NOTREACHED */ 2962 } 2963 2964 return (0); 2965 } 2966 2967 struct peer * 2968 getpeerbydesc(struct bgpd_config *c, const char *descr) 2969 { 2970 struct peer *p, *res = NULL; 2971 int match = 0; 2972 2973 RB_FOREACH(p, peer_head, &c->peers) 2974 if (!strcmp(p->conf.descr, descr)) { 2975 res = p; 2976 match++; 2977 } 2978 2979 if (match > 1) 2980 log_info("neighbor description \"%s\" not unique, request " 2981 "aborted", descr); 2982 2983 if (match == 1) 2984 return (res); 2985 else 2986 return (NULL); 2987 } 2988 2989 struct peer * 2990 getpeerbyip(struct bgpd_config *c, struct sockaddr *ip) 2991 { 2992 struct bgpd_addr addr; 2993 struct peer *p, *newpeer, *loose = NULL; 2994 u_int32_t id; 2995 2996 sa2addr(ip, &addr, NULL); 2997 2998 /* we might want a more effective way to find peers by IP */ 2999 RB_FOREACH(p, peer_head, &c->peers) 3000 if (!p->conf.template && 3001 !memcmp(&addr, &p->conf.remote_addr, sizeof(addr))) 3002 return (p); 3003 3004 /* try template matching */ 3005 RB_FOREACH(p, peer_head, &c->peers) 3006 if (p->conf.template && 3007 p->conf.remote_addr.aid == addr.aid && 3008 session_match_mask(p, &addr)) 3009 if (loose == NULL || loose->conf.remote_masklen < 3010 p->conf.remote_masklen) 3011 loose = p; 3012 3013 if (loose != NULL) { 3014 /* clone */ 3015 if ((newpeer = malloc(sizeof(struct peer))) == NULL) 3016 fatal(NULL); 3017 memcpy(newpeer, loose, sizeof(struct peer)); 3018 for (id = PEER_ID_DYN_MAX; id > PEER_ID_STATIC_MAX; id--) { 3019 if (getpeerbyid(c, id) == NULL) /* we found a free id */ 3020 break; 3021 } 3022 newpeer->template = loose; 3023 session_template_clone(newpeer, ip, id, 0); 3024 newpeer->state = newpeer->prev_state = STATE_NONE; 3025 newpeer->reconf_action = RECONF_KEEP; 3026 newpeer->rbuf = NULL; 3027 newpeer->rpending = 0; 3028 init_peer(newpeer); 3029 bgp_fsm(newpeer, EVNT_START); 3030 if (RB_INSERT(peer_head, &c->peers, newpeer) != NULL) 3031 fatalx("%s: peer tree is corrupt", __func__); 3032 return (newpeer); 3033 } 3034 3035 return (NULL); 3036 } 3037 3038 struct peer * 3039 getpeerbyid(struct bgpd_config *c, u_int32_t peerid) 3040 { 3041 static struct peer lookup; 3042 3043 lookup.conf.id = peerid; 3044 3045 return RB_FIND(peer_head, &c->peers, &lookup); 3046 } 3047 3048 int 3049 peer_matched(struct peer *p, struct ctl_neighbor *n) 3050 { 3051 char *s; 3052 3053 if (n && n->addr.aid) { 3054 if (memcmp(&p->conf.remote_addr, &n->addr, 3055 sizeof(p->conf.remote_addr))) 3056 return 0; 3057 } else if (n && n->descr[0]) { 3058 s = n->is_group ? p->conf.group : p->conf.descr; 3059 if (strcmp(s, n->descr)) 3060 return 0; 3061 } 3062 return 1; 3063 } 3064 3065 void 3066 session_template_clone(struct peer *p, struct sockaddr *ip, u_int32_t id, 3067 u_int32_t as) 3068 { 3069 struct bgpd_addr remote_addr; 3070 3071 if (ip) 3072 sa2addr(ip, &remote_addr, NULL); 3073 else 3074 memcpy(&remote_addr, &p->conf.remote_addr, sizeof(remote_addr)); 3075 3076 memcpy(&p->conf, &p->template->conf, sizeof(struct peer_config)); 3077 3078 p->conf.id = id; 3079 3080 if (as) { 3081 p->conf.remote_as = as; 3082 p->conf.ebgp = (p->conf.remote_as != p->conf.local_as); 3083 if (!p->conf.ebgp) 3084 /* force enforce_as off for iBGP sessions */ 3085 p->conf.enforce_as = ENFORCE_AS_OFF; 3086 } 3087 3088 memcpy(&p->conf.remote_addr, &remote_addr, sizeof(remote_addr)); 3089 switch (p->conf.remote_addr.aid) { 3090 case AID_INET: 3091 p->conf.remote_masklen = 32; 3092 break; 3093 case AID_INET6: 3094 p->conf.remote_masklen = 128; 3095 break; 3096 } 3097 p->conf.template = 0; 3098 } 3099 3100 int 3101 session_match_mask(struct peer *p, struct bgpd_addr *a) 3102 { 3103 struct in_addr v4masked; 3104 struct in6_addr v6masked; 3105 3106 switch (p->conf.remote_addr.aid) { 3107 case AID_INET: 3108 inet4applymask(&v4masked, &a->v4, p->conf.remote_masklen); 3109 if (p->conf.remote_addr.v4.s_addr == v4masked.s_addr) 3110 return (1); 3111 return (0); 3112 case AID_INET6: 3113 inet6applymask(&v6masked, &a->v6, p->conf.remote_masklen); 3114 3115 if (memcmp(&v6masked, &p->conf.remote_addr.v6, 3116 sizeof(v6masked)) == 0) 3117 return (1); 3118 return (0); 3119 } 3120 return (0); 3121 } 3122 3123 void 3124 session_down(struct peer *peer) 3125 { 3126 bzero(&peer->capa.neg, sizeof(peer->capa.neg)); 3127 peer->stats.last_updown = getmonotime(); 3128 /* 3129 * session_down is called in the exit code path so check 3130 * if the RDE is still around, if not there is no need to 3131 * send the message. 3132 */ 3133 if (ibuf_rde == NULL) 3134 return; 3135 if (imsg_rde(IMSG_SESSION_DOWN, peer->conf.id, NULL, 0) == -1) 3136 fatalx("imsg_compose error"); 3137 } 3138 3139 void 3140 session_up(struct peer *p) 3141 { 3142 struct session_up sup; 3143 3144 if (imsg_rde(IMSG_SESSION_ADD, p->conf.id, 3145 &p->conf, sizeof(p->conf)) == -1) 3146 fatalx("imsg_compose error"); 3147 3148 if (p->local.aid == AID_INET) { 3149 sup.local_v4_addr = p->local; 3150 sup.local_v6_addr = p->local_alt; 3151 } else { 3152 sup.local_v6_addr = p->local; 3153 sup.local_v4_addr = p->local_alt; 3154 } 3155 sup.remote_addr = p->remote; 3156 3157 sup.remote_bgpid = p->remote_bgpid; 3158 sup.short_as = p->short_as; 3159 memcpy(&sup.capa, &p->capa.neg, sizeof(sup.capa)); 3160 p->stats.last_updown = getmonotime(); 3161 if (imsg_rde(IMSG_SESSION_UP, p->conf.id, &sup, sizeof(sup)) == -1) 3162 fatalx("imsg_compose error"); 3163 } 3164 3165 int 3166 imsg_ctl_parent(int type, u_int32_t peerid, pid_t pid, void *data, 3167 u_int16_t datalen) 3168 { 3169 return (imsg_compose(ibuf_main, type, peerid, pid, -1, data, datalen)); 3170 } 3171 3172 int 3173 imsg_ctl_rde(int type, pid_t pid, void *data, u_int16_t datalen) 3174 { 3175 if (ibuf_rde_ctl == NULL) { 3176 log_warnx("Can't send message %u to RDE, ctl pipe closed", 3177 type); 3178 return (0); 3179 } 3180 /* 3181 * Use control socket to talk to RDE to bypass the queue of the 3182 * regular imsg socket. 3183 */ 3184 return (imsg_compose(ibuf_rde_ctl, type, 0, pid, -1, data, datalen)); 3185 } 3186 3187 int 3188 imsg_rde(int type, uint32_t peerid, void *data, u_int16_t datalen) 3189 { 3190 if (ibuf_rde == NULL) { 3191 log_warnx("Can't send message %u to RDE, pipe closed", type); 3192 return (0); 3193 } 3194 3195 return (imsg_compose(ibuf_rde, type, peerid, 0, -1, data, datalen)); 3196 } 3197 3198 void 3199 session_demote(struct peer *p, int level) 3200 { 3201 struct demote_msg msg; 3202 3203 strlcpy(msg.demote_group, p->conf.demote_group, 3204 sizeof(msg.demote_group)); 3205 msg.level = level; 3206 if (imsg_compose(ibuf_main, IMSG_DEMOTE, p->conf.id, 0, -1, 3207 &msg, sizeof(msg)) == -1) 3208 fatalx("imsg_compose error"); 3209 3210 p->demoted += level; 3211 } 3212 3213 void 3214 session_stop(struct peer *peer, u_int8_t subcode) 3215 { 3216 char data[SHUT_COMM_LEN]; 3217 size_t datalen; 3218 size_t shutcomm_len; 3219 char *communication; 3220 3221 datalen = 0; 3222 communication = peer->conf.shutcomm; 3223 3224 if ((subcode == ERR_CEASE_ADMIN_DOWN || 3225 subcode == ERR_CEASE_ADMIN_RESET) 3226 && communication && *communication) { 3227 shutcomm_len = strlen(communication); 3228 if (shutcomm_len > SHUT_COMM_LEN - 1) { 3229 log_peer_warnx(&peer->conf, 3230 "trying to send overly long shutdown reason"); 3231 } else { 3232 data[0] = shutcomm_len; 3233 datalen = shutcomm_len + sizeof(data[0]); 3234 memcpy(data + 1, communication, shutcomm_len); 3235 } 3236 } 3237 switch (peer->state) { 3238 case STATE_OPENSENT: 3239 case STATE_OPENCONFIRM: 3240 case STATE_ESTABLISHED: 3241 session_notification(peer, ERR_CEASE, subcode, data, datalen); 3242 break; 3243 default: 3244 /* session not open, no need to send notification */ 3245 break; 3246 } 3247 bgp_fsm(peer, EVNT_STOP); 3248 } 3249 3250 void 3251 merge_peers(struct bgpd_config *c, struct bgpd_config *nc) 3252 { 3253 struct peer *p, *np, *next; 3254 3255 RB_FOREACH(p, peer_head, &c->peers) { 3256 /* templates are handled specially */ 3257 if (p->template != NULL) 3258 continue; 3259 np = getpeerbyid(nc, p->conf.id); 3260 if (np == NULL) { 3261 p->reconf_action = RECONF_DELETE; 3262 continue; 3263 } 3264 3265 /* peer no longer uses TCP MD5SIG so deconfigure */ 3266 if (p->conf.auth.method == AUTH_MD5SIG && 3267 np->conf.auth.method != AUTH_MD5SIG) 3268 tcp_md5_del_listener(c, p); 3269 else if (np->conf.auth.method == AUTH_MD5SIG) 3270 tcp_md5_add_listener(c, np); 3271 3272 memcpy(&p->conf, &np->conf, sizeof(p->conf)); 3273 RB_REMOVE(peer_head, &nc->peers, np); 3274 free(np); 3275 3276 p->reconf_action = RECONF_KEEP; 3277 3278 /* had demotion, is demoted, demote removed? */ 3279 if (p->demoted && !p->conf.demote_group[0]) 3280 session_demote(p, -1); 3281 3282 /* if session is not open then refresh pfkey data */ 3283 if (p->state < STATE_OPENSENT && !p->template) 3284 imsg_compose(ibuf_main, IMSG_PFKEY_RELOAD, 3285 p->conf.id, 0, -1, NULL, 0); 3286 3287 /* sync the RDE in case we keep the peer */ 3288 if (imsg_rde(IMSG_SESSION_ADD, p->conf.id, 3289 &p->conf, sizeof(struct peer_config)) == -1) 3290 fatalx("imsg_compose error"); 3291 3292 /* apply the config to all clones of a template */ 3293 if (p->conf.template) { 3294 struct peer *xp; 3295 RB_FOREACH(xp, peer_head, &c->peers) { 3296 if (xp->template != p) 3297 continue; 3298 session_template_clone(xp, NULL, xp->conf.id, 3299 xp->conf.remote_as); 3300 if (imsg_rde(IMSG_SESSION_ADD, xp->conf.id, 3301 &xp->conf, sizeof(xp->conf)) == -1) 3302 fatalx("imsg_compose error"); 3303 } 3304 } 3305 } 3306 3307 /* pfkeys of new peers already loaded by the parent process */ 3308 RB_FOREACH_SAFE(np, peer_head, &nc->peers, next) { 3309 RB_REMOVE(peer_head, &nc->peers, np); 3310 if (RB_INSERT(peer_head, &c->peers, np) != NULL) 3311 fatalx("%s: peer tree is corrupt", __func__); 3312 if (np->conf.auth.method == AUTH_MD5SIG) 3313 tcp_md5_add_listener(c, np); 3314 } 3315 } 3316