1 /*- 2 * SPDX-License-Identifier: BSD-3-Clause 3 * 4 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project. 5 * All rights reserved. 6 * 7 * Redistribution and use in source and binary forms, with or without 8 * modification, are permitted provided that the following conditions 9 * are met: 10 * 1. Redistributions of source code must retain the above copyright 11 * notice, this list of conditions and the following disclaimer. 12 * 2. Redistributions in binary form must reproduce the above copyright 13 * notice, this list of conditions and the following disclaimer in the 14 * documentation and/or other materials provided with the distribution. 15 * 3. Neither the name of the project nor the names of its contributors 16 * may be used to endorse or promote products derived from this software 17 * without specific prior written permission. 18 * 19 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND 20 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 21 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 22 * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE 23 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 24 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 25 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 26 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 27 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 28 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 29 * SUCH DAMAGE. 30 * 31 * $KAME: nd6_rtr.c,v 1.111 2001/04/27 01:37:15 jinmei Exp $ 32 */ 33 34 #include <sys/cdefs.h> 35 __FBSDID("$FreeBSD$"); 36 37 #include "opt_inet.h" 38 #include "opt_inet6.h" 39 40 #include <sys/param.h> 41 #include <sys/systm.h> 42 #include <sys/malloc.h> 43 #include <sys/mbuf.h> 44 #include <sys/refcount.h> 45 #include <sys/socket.h> 46 #include <sys/sockio.h> 47 #include <sys/time.h> 48 #include <sys/kernel.h> 49 #include <sys/lock.h> 50 #include <sys/errno.h> 51 #include <sys/rmlock.h> 52 #include <sys/rwlock.h> 53 #include <sys/sysctl.h> 54 #include <sys/syslog.h> 55 #include <sys/queue.h> 56 57 #include <net/if.h> 58 #include <net/if_var.h> 59 #include <net/if_types.h> 60 #include <net/if_dl.h> 61 #include <net/route.h> 62 #include <net/route/nhop.h> 63 #include <net/route/route_ctl.h> 64 #include <net/radix.h> 65 #include <net/vnet.h> 66 67 #include <netinet/in.h> 68 #include <net/if_llatbl.h> 69 #include <netinet6/in6_var.h> 70 #include <netinet6/in6_ifattach.h> 71 #include <netinet/ip6.h> 72 #include <netinet6/ip6_var.h> 73 #include <netinet6/nd6.h> 74 #include <netinet/icmp6.h> 75 #include <netinet6/scope6_var.h> 76 77 static struct nd_defrouter *defrtrlist_update(struct nd_defrouter *); 78 static int prelist_update(struct nd_prefixctl *, struct nd_defrouter *, 79 struct mbuf *, int); 80 static int nd6_prefix_onlink(struct nd_prefix *); 81 82 TAILQ_HEAD(nd6_drhead, nd_defrouter); 83 VNET_DEFINE_STATIC(struct nd6_drhead, nd6_defrouter); 84 #define V_nd6_defrouter VNET(nd6_defrouter) 85 86 VNET_DECLARE(int, nd6_recalc_reachtm_interval); 87 #define V_nd6_recalc_reachtm_interval VNET(nd6_recalc_reachtm_interval) 88 89 VNET_DEFINE_STATIC(struct ifnet *, nd6_defifp); 90 VNET_DEFINE(int, nd6_defifindex); 91 #define V_nd6_defifp VNET(nd6_defifp) 92 93 VNET_DEFINE(int, ip6_use_tempaddr) = 0; 94 95 VNET_DEFINE(int, ip6_desync_factor); 96 VNET_DEFINE(u_int32_t, ip6_temp_preferred_lifetime) = DEF_TEMP_PREFERRED_LIFETIME; 97 VNET_DEFINE(u_int32_t, ip6_temp_valid_lifetime) = DEF_TEMP_VALID_LIFETIME; 98 99 VNET_DEFINE(int, ip6_temp_regen_advance) = TEMPADDR_REGEN_ADVANCE; 100 101 #ifdef EXPERIMENTAL 102 VNET_DEFINE(int, nd6_ignore_ipv6_only_ra) = 1; 103 #endif 104 105 SYSCTL_DECL(_net_inet6_icmp6); 106 107 /* RTPREF_MEDIUM has to be 0! */ 108 #define RTPREF_HIGH 1 109 #define RTPREF_MEDIUM 0 110 #define RTPREF_LOW (-1) 111 #define RTPREF_RESERVED (-2) 112 #define RTPREF_INVALID (-3) /* internal */ 113 114 static void 115 defrouter_ref(struct nd_defrouter *dr) 116 { 117 118 refcount_acquire(&dr->refcnt); 119 } 120 121 void 122 defrouter_rele(struct nd_defrouter *dr) 123 { 124 125 if (refcount_release(&dr->refcnt)) 126 free(dr, M_IP6NDP); 127 } 128 129 /* 130 * Remove a router from the global list and optionally stash it in a 131 * caller-supplied queue. 132 */ 133 static void 134 defrouter_unlink(struct nd_defrouter *dr, struct nd6_drhead *drq) 135 { 136 137 ND6_WLOCK_ASSERT(); 138 139 TAILQ_REMOVE(&V_nd6_defrouter, dr, dr_entry); 140 V_nd6_list_genid++; 141 if (drq != NULL) 142 TAILQ_INSERT_TAIL(drq, dr, dr_entry); 143 } 144 145 /* 146 * Receive Router Solicitation Message - just for routers. 147 * Router solicitation/advertisement is mostly managed by userland program 148 * (rtadvd) so here we have no function like nd6_ra_output(). 149 * 150 * Based on RFC 2461 151 */ 152 void 153 nd6_rs_input(struct mbuf *m, int off, int icmp6len) 154 { 155 struct ifnet *ifp; 156 struct ip6_hdr *ip6; 157 struct nd_router_solicit *nd_rs; 158 struct in6_addr saddr6; 159 union nd_opts ndopts; 160 char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN]; 161 char *lladdr; 162 int lladdrlen; 163 164 ifp = m->m_pkthdr.rcvif; 165 166 /* 167 * Accept RS only when V_ip6_forwarding=1 and the interface has 168 * no ND6_IFF_ACCEPT_RTADV. 169 */ 170 if (!V_ip6_forwarding || ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV) 171 goto freeit; 172 173 /* RFC 6980: Nodes MUST silently ignore fragments */ 174 if(m->m_flags & M_FRAGMENTED) 175 goto freeit; 176 177 /* Sanity checks */ 178 ip6 = mtod(m, struct ip6_hdr *); 179 if (__predict_false(ip6->ip6_hlim != 255)) { 180 ICMP6STAT_INC(icp6s_invlhlim); 181 nd6log((LOG_ERR, 182 "%s: invalid hlim (%d) from %s to %s on %s\n", __func__, 183 ip6->ip6_hlim, ip6_sprintf(ip6bufs, &ip6->ip6_src), 184 ip6_sprintf(ip6bufd, &ip6->ip6_dst), if_name(ifp))); 185 goto bad; 186 } 187 188 /* 189 * Don't update the neighbor cache, if src = ::. 190 * This indicates that the src has no IP address assigned yet. 191 */ 192 saddr6 = ip6->ip6_src; 193 if (IN6_IS_ADDR_UNSPECIFIED(&saddr6)) 194 goto freeit; 195 196 if (m->m_len < off + icmp6len) { 197 m = m_pullup(m, off + icmp6len); 198 if (m == NULL) { 199 IP6STAT_INC(ip6s_exthdrtoolong); 200 return; 201 } 202 } 203 ip6 = mtod(m, struct ip6_hdr *); 204 nd_rs = (struct nd_router_solicit *)((caddr_t)ip6 + off); 205 206 icmp6len -= sizeof(*nd_rs); 207 nd6_option_init(nd_rs + 1, icmp6len, &ndopts); 208 if (nd6_options(&ndopts) < 0) { 209 nd6log((LOG_INFO, 210 "%s: invalid ND option, ignored\n", __func__)); 211 /* nd6_options have incremented stats */ 212 goto freeit; 213 } 214 215 lladdr = NULL; 216 lladdrlen = 0; 217 if (ndopts.nd_opts_src_lladdr) { 218 lladdr = (char *)(ndopts.nd_opts_src_lladdr + 1); 219 lladdrlen = ndopts.nd_opts_src_lladdr->nd_opt_len << 3; 220 } 221 222 if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) { 223 nd6log((LOG_INFO, 224 "%s: lladdrlen mismatch for %s (if %d, RS packet %d)\n", 225 __func__, ip6_sprintf(ip6bufs, &saddr6), 226 ifp->if_addrlen, lladdrlen - 2)); 227 goto bad; 228 } 229 230 nd6_cache_lladdr(ifp, &saddr6, lladdr, lladdrlen, ND_ROUTER_SOLICIT, 0); 231 232 freeit: 233 m_freem(m); 234 return; 235 236 bad: 237 ICMP6STAT_INC(icp6s_badrs); 238 m_freem(m); 239 } 240 241 #ifdef EXPERIMENTAL 242 /* 243 * An initial update routine for draft-ietf-6man-ipv6only-flag. 244 * We need to iterate over all default routers for the given 245 * interface to see whether they are all advertising the "S" 246 * (IPv6-Only) flag. If they do set, otherwise unset, the 247 * interface flag we later use to filter on. 248 */ 249 static void 250 defrtr_ipv6_only_ifp(struct ifnet *ifp) 251 { 252 struct nd_defrouter *dr; 253 bool ipv6_only, ipv6_only_old; 254 #ifdef INET 255 struct epoch_tracker et; 256 struct ifaddr *ifa; 257 bool has_ipv4_addr; 258 #endif 259 260 if (V_nd6_ignore_ipv6_only_ra != 0) 261 return; 262 263 ipv6_only = true; 264 ND6_RLOCK(); 265 TAILQ_FOREACH(dr, &V_nd6_defrouter, dr_entry) 266 if (dr->ifp == ifp && 267 (dr->raflags & ND_RA_FLAG_IPV6_ONLY) == 0) 268 ipv6_only = false; 269 ND6_RUNLOCK(); 270 271 IF_AFDATA_WLOCK(ifp); 272 ipv6_only_old = ND_IFINFO(ifp)->flags & ND6_IFF_IPV6_ONLY; 273 IF_AFDATA_WUNLOCK(ifp); 274 275 /* If nothing changed, we have an early exit. */ 276 if (ipv6_only == ipv6_only_old) 277 return; 278 279 #ifdef INET 280 /* 281 * Should we want to set the IPV6-ONLY flag, check if the 282 * interface has a non-0/0 and non-link-local IPv4 address 283 * configured on it. If it has we will assume working 284 * IPv4 operations and will clear the interface flag. 285 */ 286 has_ipv4_addr = false; 287 if (ipv6_only) { 288 NET_EPOCH_ENTER(et); 289 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 290 if (ifa->ifa_addr->sa_family != AF_INET) 291 continue; 292 if (in_canforward( 293 satosin(ifa->ifa_addr)->sin_addr)) { 294 has_ipv4_addr = true; 295 break; 296 } 297 } 298 NET_EPOCH_EXIT(et); 299 } 300 if (ipv6_only && has_ipv4_addr) { 301 log(LOG_NOTICE, "%s rcvd RA w/ IPv6-Only flag set but has IPv4 " 302 "configured, ignoring IPv6-Only flag.\n", ifp->if_xname); 303 ipv6_only = false; 304 } 305 #endif 306 307 IF_AFDATA_WLOCK(ifp); 308 if (ipv6_only) 309 ND_IFINFO(ifp)->flags |= ND6_IFF_IPV6_ONLY; 310 else 311 ND_IFINFO(ifp)->flags &= ~ND6_IFF_IPV6_ONLY; 312 IF_AFDATA_WUNLOCK(ifp); 313 314 #ifdef notyet 315 /* Send notification of flag change. */ 316 #endif 317 } 318 319 static void 320 defrtr_ipv6_only_ipf_down(struct ifnet *ifp) 321 { 322 323 IF_AFDATA_WLOCK(ifp); 324 ND_IFINFO(ifp)->flags &= ~ND6_IFF_IPV6_ONLY; 325 IF_AFDATA_WUNLOCK(ifp); 326 } 327 #endif /* EXPERIMENTAL */ 328 329 void 330 nd6_ifnet_link_event(void *arg __unused, struct ifnet *ifp, int linkstate) 331 { 332 333 /* 334 * XXX-BZ we might want to trigger re-evaluation of our default router 335 * availability. E.g., on link down the default router might be 336 * unreachable but a different interface might still have connectivity. 337 */ 338 339 #ifdef EXPERIMENTAL 340 if (linkstate == LINK_STATE_DOWN) 341 defrtr_ipv6_only_ipf_down(ifp); 342 #endif 343 } 344 345 /* 346 * Receive Router Advertisement Message. 347 * 348 * Based on RFC 2461 349 * TODO: on-link bit on prefix information 350 * TODO: ND_RA_FLAG_{OTHER,MANAGED} processing 351 */ 352 void 353 nd6_ra_input(struct mbuf *m, int off, int icmp6len) 354 { 355 struct ifnet *ifp; 356 struct nd_ifinfo *ndi; 357 struct ip6_hdr *ip6; 358 struct nd_router_advert *nd_ra; 359 struct in6_addr saddr6; 360 struct nd_defrouter *dr; 361 union nd_opts ndopts; 362 char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN]; 363 int mcast; 364 365 /* 366 * We only accept RAs only when the per-interface flag 367 * ND6_IFF_ACCEPT_RTADV is on the receiving interface. 368 */ 369 ifp = m->m_pkthdr.rcvif; 370 ndi = ND_IFINFO(ifp); 371 if (!(ndi->flags & ND6_IFF_ACCEPT_RTADV)) 372 goto freeit; 373 374 /* RFC 6980: Nodes MUST silently ignore fragments */ 375 if(m->m_flags & M_FRAGMENTED) 376 goto freeit; 377 378 ip6 = mtod(m, struct ip6_hdr *); 379 if (__predict_false(ip6->ip6_hlim != 255)) { 380 ICMP6STAT_INC(icp6s_invlhlim); 381 nd6log((LOG_ERR, 382 "%s: invalid hlim (%d) from %s to %s on %s\n", __func__, 383 ip6->ip6_hlim, ip6_sprintf(ip6bufs, &ip6->ip6_src), 384 ip6_sprintf(ip6bufd, &ip6->ip6_dst), if_name(ifp))); 385 goto bad; 386 } 387 388 saddr6 = ip6->ip6_src; 389 if (!IN6_IS_ADDR_LINKLOCAL(&saddr6)) { 390 nd6log((LOG_ERR, 391 "%s: src %s is not link-local\n", __func__, 392 ip6_sprintf(ip6bufs, &saddr6))); 393 goto bad; 394 } 395 396 if (m->m_len < off + icmp6len) { 397 m = m_pullup(m, off + icmp6len); 398 if (m == NULL) { 399 IP6STAT_INC(ip6s_exthdrtoolong); 400 return; 401 } 402 } 403 ip6 = mtod(m, struct ip6_hdr *); 404 nd_ra = (struct nd_router_advert *)((caddr_t)ip6 + off); 405 406 icmp6len -= sizeof(*nd_ra); 407 nd6_option_init(nd_ra + 1, icmp6len, &ndopts); 408 if (nd6_options(&ndopts) < 0) { 409 nd6log((LOG_INFO, 410 "%s: invalid ND option, ignored\n", __func__)); 411 /* nd6_options have incremented stats */ 412 goto freeit; 413 } 414 415 mcast = 0; 416 dr = NULL; 417 { 418 struct nd_defrouter dr0; 419 u_int32_t advreachable = nd_ra->nd_ra_reachable; 420 421 /* remember if this is a multicasted advertisement */ 422 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) 423 mcast = 1; 424 425 bzero(&dr0, sizeof(dr0)); 426 dr0.rtaddr = saddr6; 427 dr0.raflags = nd_ra->nd_ra_flags_reserved; 428 /* 429 * Effectively-disable routes from RA messages when 430 * ND6_IFF_NO_RADR enabled on the receiving interface or 431 * (ip6.forwarding == 1 && ip6.rfc6204w3 != 1). 432 */ 433 if (ndi->flags & ND6_IFF_NO_RADR) 434 dr0.rtlifetime = 0; 435 else if (V_ip6_forwarding && !V_ip6_rfc6204w3) 436 dr0.rtlifetime = 0; 437 else 438 dr0.rtlifetime = ntohs(nd_ra->nd_ra_router_lifetime); 439 dr0.expire = time_uptime + dr0.rtlifetime; 440 dr0.ifp = ifp; 441 /* unspecified or not? (RFC 2461 6.3.4) */ 442 if (advreachable) { 443 advreachable = ntohl(advreachable); 444 if (advreachable <= MAX_REACHABLE_TIME && 445 ndi->basereachable != advreachable) { 446 ndi->basereachable = advreachable; 447 ndi->reachable = ND_COMPUTE_RTIME(ndi->basereachable); 448 ndi->recalctm = V_nd6_recalc_reachtm_interval; /* reset */ 449 } 450 } 451 if (nd_ra->nd_ra_retransmit) 452 ndi->retrans = ntohl(nd_ra->nd_ra_retransmit); 453 if (nd_ra->nd_ra_curhoplimit) { 454 if (ndi->chlim < nd_ra->nd_ra_curhoplimit) 455 ndi->chlim = nd_ra->nd_ra_curhoplimit; 456 else if (ndi->chlim != nd_ra->nd_ra_curhoplimit) { 457 log(LOG_ERR, "RA with a lower CurHopLimit sent from " 458 "%s on %s (current = %d, received = %d). " 459 "Ignored.\n", ip6_sprintf(ip6bufs, &ip6->ip6_src), 460 if_name(ifp), ndi->chlim, nd_ra->nd_ra_curhoplimit); 461 } 462 } 463 dr = defrtrlist_update(&dr0); 464 #ifdef EXPERIMENTAL 465 defrtr_ipv6_only_ifp(ifp); 466 #endif 467 } 468 469 /* 470 * prefix 471 */ 472 if (ndopts.nd_opts_pi) { 473 struct nd_opt_hdr *pt; 474 struct nd_opt_prefix_info *pi = NULL; 475 struct nd_prefixctl pr; 476 477 for (pt = (struct nd_opt_hdr *)ndopts.nd_opts_pi; 478 pt <= (struct nd_opt_hdr *)ndopts.nd_opts_pi_end; 479 pt = (struct nd_opt_hdr *)((caddr_t)pt + 480 (pt->nd_opt_len << 3))) { 481 if (pt->nd_opt_type != ND_OPT_PREFIX_INFORMATION) 482 continue; 483 pi = (struct nd_opt_prefix_info *)pt; 484 485 if (pi->nd_opt_pi_len != 4) { 486 nd6log((LOG_INFO, 487 "%s: invalid option len %d for prefix " 488 "information option, ignored\n", __func__, 489 pi->nd_opt_pi_len)); 490 continue; 491 } 492 493 if (128 < pi->nd_opt_pi_prefix_len) { 494 nd6log((LOG_INFO, 495 "%s: invalid prefix len %d for prefix " 496 "information option, ignored\n", __func__, 497 pi->nd_opt_pi_prefix_len)); 498 continue; 499 } 500 501 if (IN6_IS_ADDR_MULTICAST(&pi->nd_opt_pi_prefix) 502 || IN6_IS_ADDR_LINKLOCAL(&pi->nd_opt_pi_prefix)) { 503 nd6log((LOG_INFO, 504 "%s: invalid prefix %s, ignored\n", 505 __func__, ip6_sprintf(ip6bufs, 506 &pi->nd_opt_pi_prefix))); 507 continue; 508 } 509 510 bzero(&pr, sizeof(pr)); 511 pr.ndpr_prefix.sin6_family = AF_INET6; 512 pr.ndpr_prefix.sin6_len = sizeof(pr.ndpr_prefix); 513 pr.ndpr_prefix.sin6_addr = pi->nd_opt_pi_prefix; 514 pr.ndpr_ifp = (struct ifnet *)m->m_pkthdr.rcvif; 515 516 pr.ndpr_raf_onlink = (pi->nd_opt_pi_flags_reserved & 517 ND_OPT_PI_FLAG_ONLINK) ? 1 : 0; 518 pr.ndpr_raf_auto = (pi->nd_opt_pi_flags_reserved & 519 ND_OPT_PI_FLAG_AUTO) ? 1 : 0; 520 pr.ndpr_plen = pi->nd_opt_pi_prefix_len; 521 pr.ndpr_vltime = ntohl(pi->nd_opt_pi_valid_time); 522 pr.ndpr_pltime = ntohl(pi->nd_opt_pi_preferred_time); 523 (void)prelist_update(&pr, dr, m, mcast); 524 } 525 } 526 if (dr != NULL) { 527 defrouter_rele(dr); 528 dr = NULL; 529 } 530 531 /* 532 * MTU 533 */ 534 if (ndopts.nd_opts_mtu && ndopts.nd_opts_mtu->nd_opt_mtu_len == 1) { 535 u_long mtu; 536 u_long maxmtu; 537 538 mtu = (u_long)ntohl(ndopts.nd_opts_mtu->nd_opt_mtu_mtu); 539 540 /* lower bound */ 541 if (mtu < IPV6_MMTU) { 542 nd6log((LOG_INFO, "%s: bogus mtu option mtu=%lu sent " 543 "from %s, ignoring\n", __func__, 544 mtu, ip6_sprintf(ip6bufs, &ip6->ip6_src))); 545 goto skip; 546 } 547 548 /* upper bound */ 549 maxmtu = (ndi->maxmtu && ndi->maxmtu < ifp->if_mtu) 550 ? ndi->maxmtu : ifp->if_mtu; 551 if (mtu <= maxmtu) { 552 int change = (ndi->linkmtu != mtu); 553 554 ndi->linkmtu = mtu; 555 if (change) { 556 /* in6_maxmtu may change */ 557 in6_setmaxmtu(); 558 rt_updatemtu(ifp); 559 } 560 } else { 561 nd6log((LOG_INFO, "%s: bogus mtu=%lu sent from %s; " 562 "exceeds maxmtu %lu, ignoring\n", __func__, 563 mtu, ip6_sprintf(ip6bufs, &ip6->ip6_src), maxmtu)); 564 } 565 } 566 567 skip: 568 569 /* 570 * Source link layer address 571 */ 572 { 573 char *lladdr = NULL; 574 int lladdrlen = 0; 575 576 if (ndopts.nd_opts_src_lladdr) { 577 lladdr = (char *)(ndopts.nd_opts_src_lladdr + 1); 578 lladdrlen = ndopts.nd_opts_src_lladdr->nd_opt_len << 3; 579 } 580 581 if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) { 582 nd6log((LOG_INFO, 583 "%s: lladdrlen mismatch for %s (if %d, RA packet %d)\n", 584 __func__, ip6_sprintf(ip6bufs, &saddr6), 585 ifp->if_addrlen, lladdrlen - 2)); 586 goto bad; 587 } 588 589 nd6_cache_lladdr(ifp, &saddr6, lladdr, 590 lladdrlen, ND_ROUTER_ADVERT, 0); 591 592 /* 593 * Installing a link-layer address might change the state of the 594 * router's neighbor cache, which might also affect our on-link 595 * detection of adveritsed prefixes. 596 */ 597 pfxlist_onlink_check(); 598 } 599 600 freeit: 601 m_freem(m); 602 return; 603 604 bad: 605 ICMP6STAT_INC(icp6s_badra); 606 m_freem(m); 607 } 608 609 /* PFXRTR */ 610 static struct nd_pfxrouter * 611 pfxrtr_lookup(struct nd_prefix *pr, struct nd_defrouter *dr) 612 { 613 struct nd_pfxrouter *search; 614 615 ND6_LOCK_ASSERT(); 616 617 LIST_FOREACH(search, &pr->ndpr_advrtrs, pfr_entry) { 618 if (search->router == dr) 619 break; 620 } 621 return (search); 622 } 623 624 static void 625 pfxrtr_add(struct nd_prefix *pr, struct nd_defrouter *dr) 626 { 627 struct nd_pfxrouter *new; 628 bool update; 629 630 ND6_UNLOCK_ASSERT(); 631 632 ND6_RLOCK(); 633 if (pfxrtr_lookup(pr, dr) != NULL) { 634 ND6_RUNLOCK(); 635 return; 636 } 637 ND6_RUNLOCK(); 638 639 new = malloc(sizeof(*new), M_IP6NDP, M_NOWAIT | M_ZERO); 640 if (new == NULL) 641 return; 642 defrouter_ref(dr); 643 new->router = dr; 644 645 ND6_WLOCK(); 646 if (pfxrtr_lookup(pr, dr) == NULL) { 647 LIST_INSERT_HEAD(&pr->ndpr_advrtrs, new, pfr_entry); 648 update = true; 649 } else { 650 /* We lost a race to add the reference. */ 651 defrouter_rele(dr); 652 free(new, M_IP6NDP); 653 update = false; 654 } 655 ND6_WUNLOCK(); 656 657 if (update) 658 pfxlist_onlink_check(); 659 } 660 661 static void 662 pfxrtr_del(struct nd_pfxrouter *pfr) 663 { 664 665 ND6_WLOCK_ASSERT(); 666 667 LIST_REMOVE(pfr, pfr_entry); 668 defrouter_rele(pfr->router); 669 free(pfr, M_IP6NDP); 670 } 671 672 /* Default router list processing sub routines. */ 673 static void 674 defrouter_addreq(struct nd_defrouter *new) 675 { 676 struct sockaddr_in6 def, mask, gate; 677 struct rt_addrinfo info; 678 struct rib_cmd_info rc; 679 unsigned int fibnum; 680 int error; 681 682 bzero(&def, sizeof(def)); 683 bzero(&mask, sizeof(mask)); 684 bzero(&gate, sizeof(gate)); 685 686 def.sin6_len = mask.sin6_len = gate.sin6_len = 687 sizeof(struct sockaddr_in6); 688 def.sin6_family = gate.sin6_family = AF_INET6; 689 gate.sin6_addr = new->rtaddr; 690 fibnum = new->ifp->if_fib; 691 692 bzero((caddr_t)&info, sizeof(info)); 693 info.rti_flags = RTF_GATEWAY; 694 info.rti_info[RTAX_DST] = (struct sockaddr *)&def; 695 info.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&gate; 696 info.rti_info[RTAX_NETMASK] = (struct sockaddr *)&mask; 697 698 NET_EPOCH_ASSERT(); 699 error = rib_action(fibnum, RTM_ADD, &info, &rc); 700 if (error == 0) { 701 struct nhop_object *nh = nhop_select_func(rc.rc_nh_new, 0); 702 rt_routemsg(RTM_ADD, rc.rc_rt, nh, fibnum); 703 new->installed = 1; 704 } 705 } 706 707 /* 708 * Remove the default route for a given router. 709 * This is just a subroutine function for defrouter_select_fib(), and 710 * should not be called from anywhere else. 711 */ 712 static void 713 defrouter_delreq(struct nd_defrouter *dr) 714 { 715 struct sockaddr_in6 def, mask, gate; 716 struct rt_addrinfo info; 717 struct rib_cmd_info rc; 718 struct epoch_tracker et; 719 unsigned int fibnum; 720 int error; 721 722 bzero(&def, sizeof(def)); 723 bzero(&mask, sizeof(mask)); 724 bzero(&gate, sizeof(gate)); 725 726 def.sin6_len = mask.sin6_len = gate.sin6_len = 727 sizeof(struct sockaddr_in6); 728 def.sin6_family = gate.sin6_family = AF_INET6; 729 gate.sin6_addr = dr->rtaddr; 730 fibnum = dr->ifp->if_fib; 731 732 bzero((caddr_t)&info, sizeof(info)); 733 info.rti_flags = RTF_GATEWAY; 734 info.rti_info[RTAX_DST] = (struct sockaddr *)&def; 735 info.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&gate; 736 info.rti_info[RTAX_NETMASK] = (struct sockaddr *)&mask; 737 738 NET_EPOCH_ENTER(et); 739 error = rib_action(fibnum, RTM_DELETE, &info, &rc); 740 if (error == 0) { 741 struct nhop_object *nh = nhop_select_func(rc.rc_nh_old, 0); 742 rt_routemsg(RTM_DELETE, rc.rc_rt, nh, fibnum); 743 } 744 NET_EPOCH_EXIT(et); 745 746 dr->installed = 0; 747 } 748 749 static void 750 defrouter_del(struct nd_defrouter *dr) 751 { 752 struct nd_defrouter *deldr = NULL; 753 struct nd_prefix *pr; 754 struct nd_pfxrouter *pfxrtr; 755 756 ND6_UNLOCK_ASSERT(); 757 758 /* 759 * Flush all the routing table entries that use the router 760 * as a next hop. 761 */ 762 if (ND_IFINFO(dr->ifp)->flags & ND6_IFF_ACCEPT_RTADV) 763 rt6_flush(&dr->rtaddr, dr->ifp); 764 765 #ifdef EXPERIMENTAL 766 defrtr_ipv6_only_ifp(dr->ifp); 767 #endif 768 769 if (dr->installed) { 770 deldr = dr; 771 defrouter_delreq(dr); 772 } 773 774 /* 775 * Also delete all the pointers to the router in each prefix lists. 776 */ 777 ND6_WLOCK(); 778 LIST_FOREACH(pr, &V_nd_prefix, ndpr_entry) { 779 if ((pfxrtr = pfxrtr_lookup(pr, dr)) != NULL) 780 pfxrtr_del(pfxrtr); 781 } 782 ND6_WUNLOCK(); 783 784 pfxlist_onlink_check(); 785 786 /* 787 * If the router is the primary one, choose a new one. 788 * Note that defrouter_select_fib() will remove the current 789 * gateway from the routing table. 790 */ 791 if (deldr) 792 defrouter_select_fib(deldr->ifp->if_fib); 793 794 /* 795 * Release the list reference. 796 */ 797 defrouter_rele(dr); 798 } 799 800 struct nd_defrouter * 801 defrouter_lookup_locked(const struct in6_addr *addr, struct ifnet *ifp) 802 { 803 struct nd_defrouter *dr; 804 805 ND6_LOCK_ASSERT(); 806 TAILQ_FOREACH(dr, &V_nd6_defrouter, dr_entry) 807 if (dr->ifp == ifp && IN6_ARE_ADDR_EQUAL(addr, &dr->rtaddr)) { 808 defrouter_ref(dr); 809 return (dr); 810 } 811 return (NULL); 812 } 813 814 struct nd_defrouter * 815 defrouter_lookup(const struct in6_addr *addr, struct ifnet *ifp) 816 { 817 struct nd_defrouter *dr; 818 819 ND6_RLOCK(); 820 dr = defrouter_lookup_locked(addr, ifp); 821 ND6_RUNLOCK(); 822 return (dr); 823 } 824 825 /* 826 * Remove all default routes from default router list. 827 */ 828 void 829 defrouter_reset(void) 830 { 831 struct nd_defrouter *dr, **dra; 832 int count, i; 833 834 count = i = 0; 835 836 /* 837 * We can't delete routes with the ND lock held, so make a copy of the 838 * current default router list and use that when deleting routes. 839 */ 840 ND6_RLOCK(); 841 TAILQ_FOREACH(dr, &V_nd6_defrouter, dr_entry) 842 count++; 843 ND6_RUNLOCK(); 844 845 dra = malloc(count * sizeof(*dra), M_TEMP, M_WAITOK | M_ZERO); 846 847 ND6_RLOCK(); 848 TAILQ_FOREACH(dr, &V_nd6_defrouter, dr_entry) { 849 if (i == count) 850 break; 851 defrouter_ref(dr); 852 dra[i++] = dr; 853 } 854 ND6_RUNLOCK(); 855 856 for (i = 0; i < count && dra[i] != NULL; i++) { 857 defrouter_delreq(dra[i]); 858 defrouter_rele(dra[i]); 859 } 860 free(dra, M_TEMP); 861 862 /* 863 * XXX should we also nuke any default routers in the kernel, by 864 * going through them by rtalloc1()? 865 */ 866 } 867 868 /* 869 * Look up a matching default router list entry and remove it. Returns true if a 870 * matching entry was found, false otherwise. 871 */ 872 bool 873 defrouter_remove(struct in6_addr *addr, struct ifnet *ifp) 874 { 875 struct nd_defrouter *dr; 876 877 ND6_WLOCK(); 878 dr = defrouter_lookup_locked(addr, ifp); 879 if (dr == NULL) { 880 ND6_WUNLOCK(); 881 return (false); 882 } 883 884 defrouter_unlink(dr, NULL); 885 ND6_WUNLOCK(); 886 defrouter_del(dr); 887 defrouter_rele(dr); 888 return (true); 889 } 890 891 /* 892 * for default router selection 893 * regards router-preference field as a 2-bit signed integer 894 */ 895 static int 896 rtpref(struct nd_defrouter *dr) 897 { 898 switch (dr->raflags & ND_RA_FLAG_RTPREF_MASK) { 899 case ND_RA_FLAG_RTPREF_HIGH: 900 return (RTPREF_HIGH); 901 case ND_RA_FLAG_RTPREF_MEDIUM: 902 case ND_RA_FLAG_RTPREF_RSV: 903 return (RTPREF_MEDIUM); 904 case ND_RA_FLAG_RTPREF_LOW: 905 return (RTPREF_LOW); 906 default: 907 /* 908 * This case should never happen. If it did, it would mean a 909 * serious bug of kernel internal. We thus always bark here. 910 * Or, can we even panic? 911 */ 912 log(LOG_ERR, "rtpref: impossible RA flag %x\n", dr->raflags); 913 return (RTPREF_INVALID); 914 } 915 /* NOTREACHED */ 916 } 917 918 static bool 919 is_dr_reachable(const struct nd_defrouter *dr) { 920 struct llentry *ln = NULL; 921 922 ln = nd6_lookup(&dr->rtaddr, LLE_SF(AF_INET6, 0), dr->ifp); 923 if (ln == NULL) 924 return (false); 925 bool reachable = ND6_IS_LLINFO_PROBREACH(ln); 926 LLE_RUNLOCK(ln); 927 return reachable; 928 } 929 930 /* 931 * Default Router Selection according to Section 6.3.6 of RFC 2461 and 932 * draft-ietf-ipngwg-router-selection: 933 * 1) Routers that are reachable or probably reachable should be preferred. 934 * If we have more than one (probably) reachable router, prefer ones 935 * with the highest router preference. 936 * 2) When no routers on the list are known to be reachable or 937 * probably reachable, routers SHOULD be selected in a round-robin 938 * fashion, regardless of router preference values. 939 * 3) If the Default Router List is empty, assume that all 940 * destinations are on-link. 941 * 942 * We assume nd_defrouter is sorted by router preference value. 943 * Since the code below covers both with and without router preference cases, 944 * we do not need to classify the cases by ifdef. 945 * 946 * At this moment, we do not try to install more than one default router, 947 * even when the multipath routing is available, because we're not sure about 948 * the benefits for stub hosts comparing to the risk of making the code 949 * complicated and the possibility of introducing bugs. 950 * 951 * We maintain a single list of routers for multiple FIBs, only considering one 952 * at a time based on the receiving interface's FIB. If @fibnum is RT_ALL_FIBS, 953 * we do the whole thing multiple times. 954 */ 955 void 956 defrouter_select_fib(int fibnum) 957 { 958 struct epoch_tracker et; 959 struct nd_defrouter *dr, *selected_dr, *installed_dr; 960 961 if (fibnum == RT_ALL_FIBS) { 962 for (fibnum = 0; fibnum < rt_numfibs; fibnum++) { 963 defrouter_select_fib(fibnum); 964 } 965 return; 966 } 967 968 ND6_RLOCK(); 969 /* 970 * Let's handle easy case (3) first: 971 * If default router list is empty, there's nothing to be done. 972 */ 973 if (TAILQ_EMPTY(&V_nd6_defrouter)) { 974 ND6_RUNLOCK(); 975 return; 976 } 977 978 /* 979 * Search for a (probably) reachable router from the list. 980 * We just pick up the first reachable one (if any), assuming that 981 * the ordering rule of the list described in defrtrlist_update(). 982 */ 983 selected_dr = installed_dr = NULL; 984 NET_EPOCH_ENTER(et); 985 TAILQ_FOREACH(dr, &V_nd6_defrouter, dr_entry) { 986 if (dr->ifp->if_fib != fibnum) 987 continue; 988 989 if (selected_dr == NULL && is_dr_reachable(dr)) { 990 selected_dr = dr; 991 defrouter_ref(selected_dr); 992 } 993 994 if (dr->installed) { 995 if (installed_dr == NULL) { 996 installed_dr = dr; 997 defrouter_ref(installed_dr); 998 } else { 999 /* 1000 * this should not happen. 1001 * warn for diagnosis. 1002 */ 1003 log(LOG_ERR, "defrouter_select_fib: more than " 1004 "one router is installed\n"); 1005 } 1006 } 1007 } 1008 1009 /* 1010 * If none of the default routers was found to be reachable, 1011 * round-robin the list regardless of preference. 1012 * Otherwise, if we have an installed router, check if the selected 1013 * (reachable) router should really be preferred to the installed one. 1014 * We only prefer the new router when the old one is not reachable 1015 * or when the new one has a really higher preference value. 1016 */ 1017 if (selected_dr == NULL) { 1018 if (installed_dr == NULL || 1019 TAILQ_NEXT(installed_dr, dr_entry) == NULL) 1020 dr = TAILQ_FIRST(&V_nd6_defrouter); 1021 else 1022 dr = TAILQ_NEXT(installed_dr, dr_entry); 1023 1024 /* Ensure we select a router for this FIB. */ 1025 TAILQ_FOREACH_FROM(dr, &V_nd6_defrouter, dr_entry) { 1026 if (dr->ifp->if_fib == fibnum) { 1027 selected_dr = dr; 1028 defrouter_ref(selected_dr); 1029 break; 1030 } 1031 } 1032 } else if (installed_dr != NULL) { 1033 if (is_dr_reachable(installed_dr) && 1034 rtpref(selected_dr) <= rtpref(installed_dr)) { 1035 defrouter_rele(selected_dr); 1036 selected_dr = installed_dr; 1037 } 1038 } 1039 ND6_RUNLOCK(); 1040 1041 /* 1042 * If we selected a router for this FIB and it's different 1043 * than the installed one, remove the installed router and 1044 * install the selected one in its place. 1045 */ 1046 if (installed_dr != selected_dr) { 1047 if (installed_dr != NULL) { 1048 defrouter_delreq(installed_dr); 1049 defrouter_rele(installed_dr); 1050 } 1051 if (selected_dr != NULL) 1052 defrouter_addreq(selected_dr); 1053 } 1054 if (selected_dr != NULL) 1055 defrouter_rele(selected_dr); 1056 NET_EPOCH_EXIT(et); 1057 } 1058 1059 static struct nd_defrouter * 1060 defrtrlist_update(struct nd_defrouter *new) 1061 { 1062 struct nd_defrouter *dr, *n; 1063 uint64_t genid; 1064 int oldpref; 1065 bool writelocked; 1066 1067 if (new->rtlifetime == 0) { 1068 defrouter_remove(&new->rtaddr, new->ifp); 1069 return (NULL); 1070 } 1071 1072 ND6_RLOCK(); 1073 writelocked = false; 1074 restart: 1075 dr = defrouter_lookup_locked(&new->rtaddr, new->ifp); 1076 if (dr != NULL) { 1077 oldpref = rtpref(dr); 1078 1079 /* override */ 1080 dr->raflags = new->raflags; /* XXX flag check */ 1081 dr->rtlifetime = new->rtlifetime; 1082 dr->expire = new->expire; 1083 1084 /* 1085 * If the preference does not change, there's no need 1086 * to sort the entries. Also make sure the selected 1087 * router is still installed in the kernel. 1088 */ 1089 if (dr->installed && rtpref(new) == oldpref) { 1090 if (writelocked) 1091 ND6_WUNLOCK(); 1092 else 1093 ND6_RUNLOCK(); 1094 return (dr); 1095 } 1096 } 1097 1098 /* 1099 * The router needs to be reinserted into the default router 1100 * list, so upgrade to a write lock. If that fails and the list 1101 * has potentially changed while the lock was dropped, we'll 1102 * redo the lookup with the write lock held. 1103 */ 1104 if (!writelocked) { 1105 writelocked = true; 1106 if (!ND6_TRY_UPGRADE()) { 1107 genid = V_nd6_list_genid; 1108 ND6_RUNLOCK(); 1109 ND6_WLOCK(); 1110 if (genid != V_nd6_list_genid) 1111 goto restart; 1112 } 1113 } 1114 1115 if (dr != NULL) { 1116 /* 1117 * The preferred router may have changed, so relocate this 1118 * router. 1119 */ 1120 TAILQ_REMOVE(&V_nd6_defrouter, dr, dr_entry); 1121 n = dr; 1122 } else { 1123 n = malloc(sizeof(*n), M_IP6NDP, M_NOWAIT | M_ZERO); 1124 if (n == NULL) { 1125 ND6_WUNLOCK(); 1126 return (NULL); 1127 } 1128 memcpy(n, new, sizeof(*n)); 1129 /* Initialize with an extra reference for the caller. */ 1130 refcount_init(&n->refcnt, 2); 1131 } 1132 1133 /* 1134 * Insert the new router in the Default Router List; 1135 * The Default Router List should be in the descending order 1136 * of router-preferece. Routers with the same preference are 1137 * sorted in the arriving time order. 1138 */ 1139 1140 /* insert at the end of the group */ 1141 TAILQ_FOREACH(dr, &V_nd6_defrouter, dr_entry) { 1142 if (rtpref(n) > rtpref(dr)) 1143 break; 1144 } 1145 if (dr != NULL) 1146 TAILQ_INSERT_BEFORE(dr, n, dr_entry); 1147 else 1148 TAILQ_INSERT_TAIL(&V_nd6_defrouter, n, dr_entry); 1149 V_nd6_list_genid++; 1150 ND6_WUNLOCK(); 1151 1152 defrouter_select_fib(new->ifp->if_fib); 1153 1154 return (n); 1155 } 1156 1157 static int 1158 in6_init_prefix_ltimes(struct nd_prefix *ndpr) 1159 { 1160 if (ndpr->ndpr_pltime == ND6_INFINITE_LIFETIME) 1161 ndpr->ndpr_preferred = 0; 1162 else 1163 ndpr->ndpr_preferred = time_uptime + ndpr->ndpr_pltime; 1164 if (ndpr->ndpr_vltime == ND6_INFINITE_LIFETIME) 1165 ndpr->ndpr_expire = 0; 1166 else 1167 ndpr->ndpr_expire = time_uptime + ndpr->ndpr_vltime; 1168 1169 return 0; 1170 } 1171 1172 static void 1173 in6_init_address_ltimes(struct nd_prefix *new, struct in6_addrlifetime *lt6) 1174 { 1175 /* init ia6t_expire */ 1176 if (lt6->ia6t_vltime == ND6_INFINITE_LIFETIME) 1177 lt6->ia6t_expire = 0; 1178 else { 1179 lt6->ia6t_expire = time_uptime; 1180 lt6->ia6t_expire += lt6->ia6t_vltime; 1181 } 1182 1183 /* init ia6t_preferred */ 1184 if (lt6->ia6t_pltime == ND6_INFINITE_LIFETIME) 1185 lt6->ia6t_preferred = 0; 1186 else { 1187 lt6->ia6t_preferred = time_uptime; 1188 lt6->ia6t_preferred += lt6->ia6t_pltime; 1189 } 1190 } 1191 1192 static struct in6_ifaddr * 1193 in6_ifadd(struct nd_prefixctl *pr, int mcast) 1194 { 1195 struct ifnet *ifp = pr->ndpr_ifp; 1196 struct ifaddr *ifa; 1197 struct in6_aliasreq ifra; 1198 struct in6_ifaddr *ia, *ib; 1199 int error, plen0; 1200 struct in6_addr mask; 1201 int prefixlen = pr->ndpr_plen; 1202 int updateflags; 1203 char ip6buf[INET6_ADDRSTRLEN]; 1204 1205 in6_prefixlen2mask(&mask, prefixlen); 1206 1207 /* 1208 * find a link-local address (will be interface ID). 1209 * Is it really mandatory? Theoretically, a global or a site-local 1210 * address can be configured without a link-local address, if we 1211 * have a unique interface identifier... 1212 * 1213 * it is not mandatory to have a link-local address, we can generate 1214 * interface identifier on the fly. we do this because: 1215 * (1) it should be the easiest way to find interface identifier. 1216 * (2) RFC2462 5.4 suggesting the use of the same interface identifier 1217 * for multiple addresses on a single interface, and possible shortcut 1218 * of DAD. we omitted DAD for this reason in the past. 1219 * (3) a user can prevent autoconfiguration of global address 1220 * by removing link-local address by hand (this is partly because we 1221 * don't have other way to control the use of IPv6 on an interface. 1222 * this has been our design choice - cf. NRL's "ifconfig auto"). 1223 * (4) it is easier to manage when an interface has addresses 1224 * with the same interface identifier, than to have multiple addresses 1225 * with different interface identifiers. 1226 */ 1227 ifa = (struct ifaddr *)in6ifa_ifpforlinklocal(ifp, 0); /* 0 is OK? */ 1228 if (ifa) 1229 ib = (struct in6_ifaddr *)ifa; 1230 else 1231 return NULL; 1232 1233 /* prefixlen + ifidlen must be equal to 128 */ 1234 plen0 = in6_mask2len(&ib->ia_prefixmask.sin6_addr, NULL); 1235 if (prefixlen != plen0) { 1236 ifa_free(ifa); 1237 nd6log((LOG_INFO, 1238 "%s: wrong prefixlen for %s (prefix=%d ifid=%d)\n", 1239 __func__, if_name(ifp), prefixlen, 128 - plen0)); 1240 return NULL; 1241 } 1242 1243 /* make ifaddr */ 1244 in6_prepare_ifra(&ifra, &pr->ndpr_prefix.sin6_addr, &mask); 1245 1246 IN6_MASK_ADDR(&ifra.ifra_addr.sin6_addr, &mask); 1247 /* interface ID */ 1248 ifra.ifra_addr.sin6_addr.s6_addr32[0] |= 1249 (ib->ia_addr.sin6_addr.s6_addr32[0] & ~mask.s6_addr32[0]); 1250 ifra.ifra_addr.sin6_addr.s6_addr32[1] |= 1251 (ib->ia_addr.sin6_addr.s6_addr32[1] & ~mask.s6_addr32[1]); 1252 ifra.ifra_addr.sin6_addr.s6_addr32[2] |= 1253 (ib->ia_addr.sin6_addr.s6_addr32[2] & ~mask.s6_addr32[2]); 1254 ifra.ifra_addr.sin6_addr.s6_addr32[3] |= 1255 (ib->ia_addr.sin6_addr.s6_addr32[3] & ~mask.s6_addr32[3]); 1256 ifa_free(ifa); 1257 1258 /* lifetimes. */ 1259 ifra.ifra_lifetime.ia6t_vltime = pr->ndpr_vltime; 1260 ifra.ifra_lifetime.ia6t_pltime = pr->ndpr_pltime; 1261 1262 /* XXX: scope zone ID? */ 1263 1264 ifra.ifra_flags |= IN6_IFF_AUTOCONF; /* obey autoconf */ 1265 1266 /* 1267 * Make sure that we do not have this address already. This should 1268 * usually not happen, but we can still see this case, e.g., if we 1269 * have manually configured the exact address to be configured. 1270 */ 1271 ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp, 1272 &ifra.ifra_addr.sin6_addr); 1273 if (ifa != NULL) { 1274 ifa_free(ifa); 1275 /* this should be rare enough to make an explicit log */ 1276 log(LOG_INFO, "in6_ifadd: %s is already configured\n", 1277 ip6_sprintf(ip6buf, &ifra.ifra_addr.sin6_addr)); 1278 return (NULL); 1279 } 1280 1281 /* 1282 * Allocate ifaddr structure, link into chain, etc. 1283 * If we are going to create a new address upon receiving a multicasted 1284 * RA, we need to impose a random delay before starting DAD. 1285 * [draft-ietf-ipv6-rfc2462bis-02.txt, Section 5.4.2] 1286 */ 1287 updateflags = 0; 1288 if (mcast) 1289 updateflags |= IN6_IFAUPDATE_DADDELAY; 1290 if ((error = in6_update_ifa(ifp, &ifra, NULL, updateflags)) != 0) { 1291 nd6log((LOG_ERR, 1292 "%s: failed to make ifaddr %s on %s (errno=%d)\n", __func__, 1293 ip6_sprintf(ip6buf, &ifra.ifra_addr.sin6_addr), 1294 if_name(ifp), error)); 1295 return (NULL); /* ifaddr must not have been allocated. */ 1296 } 1297 1298 ia = in6ifa_ifpwithaddr(ifp, &ifra.ifra_addr.sin6_addr); 1299 /* 1300 * XXXRW: Assumption of non-NULLness here might not be true with 1301 * fine-grained locking -- should we validate it? Or just return 1302 * earlier ifa rather than looking it up again? 1303 */ 1304 return (ia); /* this is always non-NULL and referenced. */ 1305 } 1306 1307 static struct nd_prefix * 1308 nd6_prefix_lookup_locked(struct nd_prefixctl *key) 1309 { 1310 struct nd_prefix *search; 1311 1312 ND6_LOCK_ASSERT(); 1313 1314 LIST_FOREACH(search, &V_nd_prefix, ndpr_entry) { 1315 if (key->ndpr_ifp == search->ndpr_ifp && 1316 key->ndpr_plen == search->ndpr_plen && 1317 in6_are_prefix_equal(&key->ndpr_prefix.sin6_addr, 1318 &search->ndpr_prefix.sin6_addr, key->ndpr_plen)) { 1319 nd6_prefix_ref(search); 1320 break; 1321 } 1322 } 1323 return (search); 1324 } 1325 1326 struct nd_prefix * 1327 nd6_prefix_lookup(struct nd_prefixctl *key) 1328 { 1329 struct nd_prefix *search; 1330 1331 ND6_RLOCK(); 1332 search = nd6_prefix_lookup_locked(key); 1333 ND6_RUNLOCK(); 1334 return (search); 1335 } 1336 1337 void 1338 nd6_prefix_ref(struct nd_prefix *pr) 1339 { 1340 1341 refcount_acquire(&pr->ndpr_refcnt); 1342 } 1343 1344 void 1345 nd6_prefix_rele(struct nd_prefix *pr) 1346 { 1347 1348 if (refcount_release(&pr->ndpr_refcnt)) { 1349 KASSERT(LIST_EMPTY(&pr->ndpr_advrtrs), 1350 ("prefix %p has advertising routers", pr)); 1351 free(pr, M_IP6NDP); 1352 } 1353 } 1354 1355 int 1356 nd6_prelist_add(struct nd_prefixctl *pr, struct nd_defrouter *dr, 1357 struct nd_prefix **newp) 1358 { 1359 struct nd_prefix *new; 1360 char ip6buf[INET6_ADDRSTRLEN]; 1361 int error; 1362 1363 new = malloc(sizeof(*new), M_IP6NDP, M_NOWAIT | M_ZERO); 1364 if (new == NULL) 1365 return (ENOMEM); 1366 refcount_init(&new->ndpr_refcnt, newp != NULL ? 2 : 1); 1367 new->ndpr_ifp = pr->ndpr_ifp; 1368 new->ndpr_prefix = pr->ndpr_prefix; 1369 new->ndpr_plen = pr->ndpr_plen; 1370 new->ndpr_vltime = pr->ndpr_vltime; 1371 new->ndpr_pltime = pr->ndpr_pltime; 1372 new->ndpr_flags = pr->ndpr_flags; 1373 if ((error = in6_init_prefix_ltimes(new)) != 0) { 1374 free(new, M_IP6NDP); 1375 return (error); 1376 } 1377 new->ndpr_lastupdate = time_uptime; 1378 1379 /* initialization */ 1380 LIST_INIT(&new->ndpr_advrtrs); 1381 in6_prefixlen2mask(&new->ndpr_mask, new->ndpr_plen); 1382 /* make prefix in the canonical form */ 1383 IN6_MASK_ADDR(&new->ndpr_prefix.sin6_addr, &new->ndpr_mask); 1384 1385 ND6_WLOCK(); 1386 LIST_INSERT_HEAD(&V_nd_prefix, new, ndpr_entry); 1387 V_nd6_list_genid++; 1388 ND6_WUNLOCK(); 1389 1390 /* ND_OPT_PI_FLAG_ONLINK processing */ 1391 if (new->ndpr_raf_onlink) { 1392 struct epoch_tracker et; 1393 1394 ND6_ONLINK_LOCK(); 1395 NET_EPOCH_ENTER(et); 1396 if ((error = nd6_prefix_onlink(new)) != 0) { 1397 nd6log((LOG_ERR, "%s: failed to make the prefix %s/%d " 1398 "on-link on %s (errno=%d)\n", __func__, 1399 ip6_sprintf(ip6buf, &pr->ndpr_prefix.sin6_addr), 1400 pr->ndpr_plen, if_name(pr->ndpr_ifp), error)); 1401 /* proceed anyway. XXX: is it correct? */ 1402 } 1403 NET_EPOCH_EXIT(et); 1404 ND6_ONLINK_UNLOCK(); 1405 } 1406 1407 if (dr != NULL) 1408 pfxrtr_add(new, dr); 1409 if (newp != NULL) 1410 *newp = new; 1411 return (0); 1412 } 1413 1414 /* 1415 * Remove a prefix from the prefix list and optionally stash it in a 1416 * caller-provided list. 1417 * 1418 * The ND6 lock must be held. 1419 */ 1420 void 1421 nd6_prefix_unlink(struct nd_prefix *pr, struct nd_prhead *list) 1422 { 1423 1424 ND6_WLOCK_ASSERT(); 1425 1426 LIST_REMOVE(pr, ndpr_entry); 1427 V_nd6_list_genid++; 1428 if (list != NULL) 1429 LIST_INSERT_HEAD(list, pr, ndpr_entry); 1430 } 1431 1432 /* 1433 * Free an unlinked prefix, first marking it off-link if necessary. 1434 */ 1435 void 1436 nd6_prefix_del(struct nd_prefix *pr) 1437 { 1438 struct nd_pfxrouter *pfr, *next; 1439 int e; 1440 char ip6buf[INET6_ADDRSTRLEN]; 1441 1442 KASSERT(pr->ndpr_addrcnt == 0, 1443 ("prefix %p has referencing addresses", pr)); 1444 ND6_UNLOCK_ASSERT(); 1445 1446 /* 1447 * Though these flags are now meaningless, we'd rather keep the value 1448 * of pr->ndpr_raf_onlink and pr->ndpr_raf_auto not to confuse users 1449 * when executing "ndp -p". 1450 */ 1451 if ((pr->ndpr_stateflags & NDPRF_ONLINK) != 0) { 1452 ND6_ONLINK_LOCK(); 1453 if ((e = nd6_prefix_offlink(pr)) != 0) { 1454 nd6log((LOG_ERR, 1455 "%s: failed to make the prefix %s/%d offlink on %s " 1456 "(errno=%d)\n", __func__, 1457 ip6_sprintf(ip6buf, &pr->ndpr_prefix.sin6_addr), 1458 pr->ndpr_plen, if_name(pr->ndpr_ifp), e)); 1459 /* what should we do? */ 1460 } 1461 ND6_ONLINK_UNLOCK(); 1462 } 1463 1464 /* Release references to routers that have advertised this prefix. */ 1465 ND6_WLOCK(); 1466 LIST_FOREACH_SAFE(pfr, &pr->ndpr_advrtrs, pfr_entry, next) 1467 pfxrtr_del(pfr); 1468 ND6_WUNLOCK(); 1469 1470 nd6_prefix_rele(pr); 1471 1472 pfxlist_onlink_check(); 1473 } 1474 1475 static int 1476 prelist_update(struct nd_prefixctl *new, struct nd_defrouter *dr, 1477 struct mbuf *m, int mcast) 1478 { 1479 struct in6_ifaddr *ia6 = NULL, *ia6_match = NULL; 1480 struct ifaddr *ifa; 1481 struct ifnet *ifp = new->ndpr_ifp; 1482 struct nd_prefix *pr; 1483 int error = 0; 1484 int auth; 1485 struct in6_addrlifetime lt6_tmp; 1486 char ip6buf[INET6_ADDRSTRLEN]; 1487 1488 NET_EPOCH_ASSERT(); 1489 1490 auth = 0; 1491 if (m) { 1492 /* 1493 * Authenticity for NA consists authentication for 1494 * both IP header and IP datagrams, doesn't it ? 1495 */ 1496 #if defined(M_AUTHIPHDR) && defined(M_AUTHIPDGM) 1497 auth = ((m->m_flags & M_AUTHIPHDR) && 1498 (m->m_flags & M_AUTHIPDGM)); 1499 #endif 1500 } 1501 1502 if ((pr = nd6_prefix_lookup(new)) != NULL) { 1503 /* 1504 * nd6_prefix_lookup() ensures that pr and new have the same 1505 * prefix on a same interface. 1506 */ 1507 1508 /* 1509 * Update prefix information. Note that the on-link (L) bit 1510 * and the autonomous (A) bit should NOT be changed from 1 1511 * to 0. 1512 */ 1513 if (new->ndpr_raf_onlink == 1) 1514 pr->ndpr_raf_onlink = 1; 1515 if (new->ndpr_raf_auto == 1) 1516 pr->ndpr_raf_auto = 1; 1517 if (new->ndpr_raf_onlink) { 1518 pr->ndpr_vltime = new->ndpr_vltime; 1519 pr->ndpr_pltime = new->ndpr_pltime; 1520 (void)in6_init_prefix_ltimes(pr); /* XXX error case? */ 1521 pr->ndpr_lastupdate = time_uptime; 1522 } 1523 1524 if (new->ndpr_raf_onlink && 1525 (pr->ndpr_stateflags & NDPRF_ONLINK) == 0) { 1526 ND6_ONLINK_LOCK(); 1527 if ((error = nd6_prefix_onlink(pr)) != 0) { 1528 nd6log((LOG_ERR, 1529 "%s: failed to make the prefix %s/%d " 1530 "on-link on %s (errno=%d)\n", __func__, 1531 ip6_sprintf(ip6buf, 1532 &pr->ndpr_prefix.sin6_addr), 1533 pr->ndpr_plen, if_name(pr->ndpr_ifp), 1534 error)); 1535 /* proceed anyway. XXX: is it correct? */ 1536 } 1537 ND6_ONLINK_UNLOCK(); 1538 } 1539 1540 if (dr != NULL) 1541 pfxrtr_add(pr, dr); 1542 } else { 1543 if (new->ndpr_vltime == 0) 1544 goto end; 1545 if (new->ndpr_raf_onlink == 0 && new->ndpr_raf_auto == 0) 1546 goto end; 1547 1548 error = nd6_prelist_add(new, dr, &pr); 1549 if (error != 0) { 1550 nd6log((LOG_NOTICE, "%s: nd6_prelist_add() failed for " 1551 "the prefix %s/%d on %s (errno=%d)\n", __func__, 1552 ip6_sprintf(ip6buf, &new->ndpr_prefix.sin6_addr), 1553 new->ndpr_plen, if_name(new->ndpr_ifp), error)); 1554 goto end; /* we should just give up in this case. */ 1555 } 1556 1557 /* 1558 * XXX: from the ND point of view, we can ignore a prefix 1559 * with the on-link bit being zero. However, we need a 1560 * prefix structure for references from autoconfigured 1561 * addresses. Thus, we explicitly make sure that the prefix 1562 * itself expires now. 1563 */ 1564 if (pr->ndpr_raf_onlink == 0) { 1565 pr->ndpr_vltime = 0; 1566 pr->ndpr_pltime = 0; 1567 in6_init_prefix_ltimes(pr); 1568 } 1569 } 1570 1571 /* 1572 * Address autoconfiguration based on Section 5.5.3 of RFC 2462. 1573 * Note that pr must be non NULL at this point. 1574 */ 1575 1576 /* 5.5.3 (a). Ignore the prefix without the A bit set. */ 1577 if (!new->ndpr_raf_auto) 1578 goto end; 1579 1580 /* 1581 * 5.5.3 (b). the link-local prefix should have been ignored in 1582 * nd6_ra_input. 1583 */ 1584 1585 /* 5.5.3 (c). Consistency check on lifetimes: pltime <= vltime. */ 1586 if (new->ndpr_pltime > new->ndpr_vltime) { 1587 error = EINVAL; /* XXX: won't be used */ 1588 goto end; 1589 } 1590 1591 /* 1592 * 5.5.3 (d). If the prefix advertised is not equal to the prefix of 1593 * an address configured by stateless autoconfiguration already in the 1594 * list of addresses associated with the interface, and the Valid 1595 * Lifetime is not 0, form an address. We first check if we have 1596 * a matching prefix. 1597 * Note: we apply a clarification in rfc2462bis-02 here. We only 1598 * consider autoconfigured addresses while RFC2462 simply said 1599 * "address". 1600 */ 1601 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1602 struct in6_ifaddr *ifa6; 1603 u_int32_t remaininglifetime; 1604 1605 if (ifa->ifa_addr->sa_family != AF_INET6) 1606 continue; 1607 1608 ifa6 = (struct in6_ifaddr *)ifa; 1609 1610 /* 1611 * We only consider autoconfigured addresses as per rfc2462bis. 1612 */ 1613 if (!(ifa6->ia6_flags & IN6_IFF_AUTOCONF)) 1614 continue; 1615 1616 /* 1617 * Spec is not clear here, but I believe we should concentrate 1618 * on unicast (i.e. not anycast) addresses. 1619 * XXX: other ia6_flags? detached or duplicated? 1620 */ 1621 if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0) 1622 continue; 1623 1624 /* 1625 * Ignore the address if it is not associated with a prefix 1626 * or is associated with a prefix that is different from this 1627 * one. (pr is never NULL here) 1628 */ 1629 if (ifa6->ia6_ndpr != pr) 1630 continue; 1631 1632 if (ia6_match == NULL) /* remember the first one */ 1633 ia6_match = ifa6; 1634 1635 /* 1636 * An already autoconfigured address matched. Now that we 1637 * are sure there is at least one matched address, we can 1638 * proceed to 5.5.3. (e): update the lifetimes according to the 1639 * "two hours" rule and the privacy extension. 1640 * We apply some clarifications in rfc2462bis: 1641 * - use remaininglifetime instead of storedlifetime as a 1642 * variable name 1643 * - remove the dead code in the "two-hour" rule 1644 */ 1645 #define TWOHOUR (120*60) 1646 lt6_tmp = ifa6->ia6_lifetime; 1647 1648 if (lt6_tmp.ia6t_vltime == ND6_INFINITE_LIFETIME) 1649 remaininglifetime = ND6_INFINITE_LIFETIME; 1650 else if (time_uptime - ifa6->ia6_updatetime > 1651 lt6_tmp.ia6t_vltime) { 1652 /* 1653 * The case of "invalid" address. We should usually 1654 * not see this case. 1655 */ 1656 remaininglifetime = 0; 1657 } else 1658 remaininglifetime = lt6_tmp.ia6t_vltime - 1659 (time_uptime - ifa6->ia6_updatetime); 1660 1661 /* when not updating, keep the current stored lifetime. */ 1662 lt6_tmp.ia6t_vltime = remaininglifetime; 1663 1664 if (TWOHOUR < new->ndpr_vltime || 1665 remaininglifetime < new->ndpr_vltime) { 1666 lt6_tmp.ia6t_vltime = new->ndpr_vltime; 1667 } else if (remaininglifetime <= TWOHOUR) { 1668 if (auth) { 1669 lt6_tmp.ia6t_vltime = new->ndpr_vltime; 1670 } 1671 } else { 1672 /* 1673 * new->ndpr_vltime <= TWOHOUR && 1674 * TWOHOUR < remaininglifetime 1675 */ 1676 lt6_tmp.ia6t_vltime = TWOHOUR; 1677 } 1678 1679 /* The 2 hour rule is not imposed for preferred lifetime. */ 1680 lt6_tmp.ia6t_pltime = new->ndpr_pltime; 1681 1682 in6_init_address_ltimes(pr, <6_tmp); 1683 1684 /* 1685 * We need to treat lifetimes for temporary addresses 1686 * differently, according to 1687 * draft-ietf-ipv6-privacy-addrs-v2-01.txt 3.3 (1); 1688 * we only update the lifetimes when they are in the maximum 1689 * intervals. 1690 */ 1691 if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) != 0) { 1692 u_int32_t maxvltime, maxpltime; 1693 1694 if (V_ip6_temp_valid_lifetime > 1695 (u_int32_t)((time_uptime - ifa6->ia6_createtime) + 1696 V_ip6_desync_factor)) { 1697 maxvltime = V_ip6_temp_valid_lifetime - 1698 (time_uptime - ifa6->ia6_createtime) - 1699 V_ip6_desync_factor; 1700 } else 1701 maxvltime = 0; 1702 if (V_ip6_temp_preferred_lifetime > 1703 (u_int32_t)((time_uptime - ifa6->ia6_createtime) + 1704 V_ip6_desync_factor)) { 1705 maxpltime = V_ip6_temp_preferred_lifetime - 1706 (time_uptime - ifa6->ia6_createtime) - 1707 V_ip6_desync_factor; 1708 } else 1709 maxpltime = 0; 1710 1711 if (lt6_tmp.ia6t_vltime == ND6_INFINITE_LIFETIME || 1712 lt6_tmp.ia6t_vltime > maxvltime) { 1713 lt6_tmp.ia6t_vltime = maxvltime; 1714 } 1715 if (lt6_tmp.ia6t_pltime == ND6_INFINITE_LIFETIME || 1716 lt6_tmp.ia6t_pltime > maxpltime) { 1717 lt6_tmp.ia6t_pltime = maxpltime; 1718 } 1719 } 1720 ifa6->ia6_lifetime = lt6_tmp; 1721 ifa6->ia6_updatetime = time_uptime; 1722 } 1723 if (ia6_match == NULL && new->ndpr_vltime) { 1724 int ifidlen; 1725 1726 /* 1727 * 5.5.3 (d) (continued) 1728 * No address matched and the valid lifetime is non-zero. 1729 * Create a new address. 1730 */ 1731 1732 /* 1733 * Prefix Length check: 1734 * If the sum of the prefix length and interface identifier 1735 * length does not equal 128 bits, the Prefix Information 1736 * option MUST be ignored. The length of the interface 1737 * identifier is defined in a separate link-type specific 1738 * document. 1739 */ 1740 ifidlen = in6_if2idlen(ifp); 1741 if (ifidlen < 0) { 1742 /* this should not happen, so we always log it. */ 1743 log(LOG_ERR, "prelist_update: IFID undefined (%s)\n", 1744 if_name(ifp)); 1745 goto end; 1746 } 1747 if (ifidlen + pr->ndpr_plen != 128) { 1748 nd6log((LOG_INFO, 1749 "%s: invalid prefixlen %d for %s, ignored\n", 1750 __func__, pr->ndpr_plen, if_name(ifp))); 1751 goto end; 1752 } 1753 1754 if ((ia6 = in6_ifadd(new, mcast)) != NULL) { 1755 /* 1756 * note that we should use pr (not new) for reference. 1757 */ 1758 pr->ndpr_addrcnt++; 1759 ia6->ia6_ndpr = pr; 1760 1761 /* 1762 * RFC 3041 3.3 (2). 1763 * When a new public address is created as described 1764 * in RFC2462, also create a new temporary address. 1765 * 1766 * RFC 3041 3.5. 1767 * When an interface connects to a new link, a new 1768 * randomized interface identifier should be generated 1769 * immediately together with a new set of temporary 1770 * addresses. Thus, we specifiy 1 as the 2nd arg of 1771 * in6_tmpifadd(). 1772 */ 1773 if (V_ip6_use_tempaddr) { 1774 int e; 1775 if ((e = in6_tmpifadd(ia6, 1, 1)) != 0) { 1776 nd6log((LOG_NOTICE, "%s: failed to " 1777 "create a temporary address " 1778 "(errno=%d)\n", __func__, e)); 1779 } 1780 } 1781 ifa_free(&ia6->ia_ifa); 1782 1783 /* 1784 * A newly added address might affect the status 1785 * of other addresses, so we check and update it. 1786 * XXX: what if address duplication happens? 1787 */ 1788 pfxlist_onlink_check(); 1789 } else { 1790 /* just set an error. do not bark here. */ 1791 error = EADDRNOTAVAIL; /* XXX: might be unused. */ 1792 } 1793 } 1794 1795 end: 1796 if (pr != NULL) 1797 nd6_prefix_rele(pr); 1798 return (error); 1799 } 1800 1801 /* 1802 * A supplement function used in the on-link detection below; 1803 * detect if a given prefix has a (probably) reachable advertising router. 1804 * XXX: lengthy function name... 1805 */ 1806 static struct nd_pfxrouter * 1807 find_pfxlist_reachable_router(struct nd_prefix *pr) 1808 { 1809 struct epoch_tracker et; 1810 struct nd_pfxrouter *pfxrtr; 1811 1812 ND6_LOCK_ASSERT(); 1813 1814 NET_EPOCH_ENTER(et); 1815 LIST_FOREACH(pfxrtr, &pr->ndpr_advrtrs, pfr_entry) { 1816 if (is_dr_reachable(pfxrtr->router)) 1817 break; 1818 } 1819 NET_EPOCH_EXIT(et); 1820 return (pfxrtr); 1821 } 1822 1823 /* 1824 * Check if each prefix in the prefix list has at least one available router 1825 * that advertised the prefix (a router is "available" if its neighbor cache 1826 * entry is reachable or probably reachable). 1827 * If the check fails, the prefix may be off-link, because, for example, 1828 * we have moved from the network but the lifetime of the prefix has not 1829 * expired yet. So we should not use the prefix if there is another prefix 1830 * that has an available router. 1831 * But, if there is no prefix that has an available router, we still regard 1832 * all the prefixes as on-link. This is because we can't tell if all the 1833 * routers are simply dead or if we really moved from the network and there 1834 * is no router around us. 1835 */ 1836 void 1837 pfxlist_onlink_check(void) 1838 { 1839 struct nd_prefix *pr; 1840 struct in6_ifaddr *ifa; 1841 struct nd_defrouter *dr; 1842 struct nd_pfxrouter *pfxrtr = NULL; 1843 struct rm_priotracker in6_ifa_tracker; 1844 uint64_t genid; 1845 uint32_t flags; 1846 1847 ND6_ONLINK_LOCK(); 1848 ND6_RLOCK(); 1849 1850 /* 1851 * Check if there is a prefix that has a reachable advertising 1852 * router. 1853 */ 1854 LIST_FOREACH(pr, &V_nd_prefix, ndpr_entry) { 1855 if (pr->ndpr_raf_onlink && find_pfxlist_reachable_router(pr)) 1856 break; 1857 } 1858 1859 /* 1860 * If we have no such prefix, check whether we still have a router 1861 * that does not advertise any prefixes. 1862 */ 1863 if (pr == NULL) { 1864 TAILQ_FOREACH(dr, &V_nd6_defrouter, dr_entry) { 1865 struct nd_prefix *pr0; 1866 1867 LIST_FOREACH(pr0, &V_nd_prefix, ndpr_entry) { 1868 if ((pfxrtr = pfxrtr_lookup(pr0, dr)) != NULL) 1869 break; 1870 } 1871 if (pfxrtr != NULL) 1872 break; 1873 } 1874 } 1875 if (pr != NULL || (!TAILQ_EMPTY(&V_nd6_defrouter) && pfxrtr == NULL)) { 1876 /* 1877 * There is at least one prefix that has a reachable router, 1878 * or at least a router which probably does not advertise 1879 * any prefixes. The latter would be the case when we move 1880 * to a new link where we have a router that does not provide 1881 * prefixes and we configure an address by hand. 1882 * Detach prefixes which have no reachable advertising 1883 * router, and attach other prefixes. 1884 */ 1885 LIST_FOREACH(pr, &V_nd_prefix, ndpr_entry) { 1886 /* XXX: a link-local prefix should never be detached */ 1887 if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr) || 1888 pr->ndpr_raf_onlink == 0 || 1889 pr->ndpr_raf_auto == 0) 1890 continue; 1891 1892 if ((pr->ndpr_stateflags & NDPRF_DETACHED) == 0 && 1893 find_pfxlist_reachable_router(pr) == NULL) 1894 pr->ndpr_stateflags |= NDPRF_DETACHED; 1895 else if ((pr->ndpr_stateflags & NDPRF_DETACHED) != 0 && 1896 find_pfxlist_reachable_router(pr) != NULL) 1897 pr->ndpr_stateflags &= ~NDPRF_DETACHED; 1898 } 1899 } else { 1900 /* there is no prefix that has a reachable router */ 1901 LIST_FOREACH(pr, &V_nd_prefix, ndpr_entry) { 1902 if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr) || 1903 pr->ndpr_raf_onlink == 0 || 1904 pr->ndpr_raf_auto == 0) 1905 continue; 1906 pr->ndpr_stateflags &= ~NDPRF_DETACHED; 1907 } 1908 } 1909 1910 /* 1911 * Remove each interface route associated with a (just) detached 1912 * prefix, and reinstall the interface route for a (just) attached 1913 * prefix. Note that all attempt of reinstallation does not 1914 * necessarily success, when a same prefix is shared among multiple 1915 * interfaces. Such cases will be handled in nd6_prefix_onlink, 1916 * so we don't have to care about them. 1917 */ 1918 restart: 1919 LIST_FOREACH(pr, &V_nd_prefix, ndpr_entry) { 1920 char ip6buf[INET6_ADDRSTRLEN]; 1921 int e; 1922 1923 if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr) || 1924 pr->ndpr_raf_onlink == 0 || 1925 pr->ndpr_raf_auto == 0) 1926 continue; 1927 1928 flags = pr->ndpr_stateflags & (NDPRF_DETACHED | NDPRF_ONLINK); 1929 if (flags == 0 || flags == (NDPRF_DETACHED | NDPRF_ONLINK)) { 1930 genid = V_nd6_list_genid; 1931 ND6_RUNLOCK(); 1932 if ((flags & NDPRF_ONLINK) != 0 && 1933 (e = nd6_prefix_offlink(pr)) != 0) { 1934 nd6log((LOG_ERR, 1935 "%s: failed to make %s/%d offlink " 1936 "(errno=%d)\n", __func__, 1937 ip6_sprintf(ip6buf, 1938 &pr->ndpr_prefix.sin6_addr), 1939 pr->ndpr_plen, e)); 1940 } else if ((flags & NDPRF_ONLINK) == 0 && 1941 (e = nd6_prefix_onlink(pr)) != 0) { 1942 nd6log((LOG_ERR, 1943 "%s: failed to make %s/%d onlink " 1944 "(errno=%d)\n", __func__, 1945 ip6_sprintf(ip6buf, 1946 &pr->ndpr_prefix.sin6_addr), 1947 pr->ndpr_plen, e)); 1948 } 1949 ND6_RLOCK(); 1950 if (genid != V_nd6_list_genid) 1951 goto restart; 1952 } 1953 } 1954 1955 /* 1956 * Changes on the prefix status might affect address status as well. 1957 * Make sure that all addresses derived from an attached prefix are 1958 * attached, and that all addresses derived from a detached prefix are 1959 * detached. Note, however, that a manually configured address should 1960 * always be attached. 1961 * The precise detection logic is same as the one for prefixes. 1962 */ 1963 IN6_IFADDR_RLOCK(&in6_ifa_tracker); 1964 CK_STAILQ_FOREACH(ifa, &V_in6_ifaddrhead, ia_link) { 1965 if (!(ifa->ia6_flags & IN6_IFF_AUTOCONF)) 1966 continue; 1967 1968 if (ifa->ia6_ndpr == NULL) { 1969 /* 1970 * This can happen when we first configure the address 1971 * (i.e. the address exists, but the prefix does not). 1972 * XXX: complicated relationships... 1973 */ 1974 continue; 1975 } 1976 1977 if (find_pfxlist_reachable_router(ifa->ia6_ndpr)) 1978 break; 1979 } 1980 if (ifa) { 1981 CK_STAILQ_FOREACH(ifa, &V_in6_ifaddrhead, ia_link) { 1982 if ((ifa->ia6_flags & IN6_IFF_AUTOCONF) == 0) 1983 continue; 1984 1985 if (ifa->ia6_ndpr == NULL) /* XXX: see above. */ 1986 continue; 1987 1988 if (find_pfxlist_reachable_router(ifa->ia6_ndpr)) { 1989 if (ifa->ia6_flags & IN6_IFF_DETACHED) { 1990 ifa->ia6_flags &= ~IN6_IFF_DETACHED; 1991 ifa->ia6_flags |= IN6_IFF_TENTATIVE; 1992 nd6_dad_start((struct ifaddr *)ifa, 0); 1993 } 1994 } else { 1995 ifa->ia6_flags |= IN6_IFF_DETACHED; 1996 } 1997 } 1998 } else { 1999 CK_STAILQ_FOREACH(ifa, &V_in6_ifaddrhead, ia_link) { 2000 if ((ifa->ia6_flags & IN6_IFF_AUTOCONF) == 0) 2001 continue; 2002 2003 if (ifa->ia6_flags & IN6_IFF_DETACHED) { 2004 ifa->ia6_flags &= ~IN6_IFF_DETACHED; 2005 ifa->ia6_flags |= IN6_IFF_TENTATIVE; 2006 /* Do we need a delay in this case? */ 2007 nd6_dad_start((struct ifaddr *)ifa, 0); 2008 } 2009 } 2010 } 2011 IN6_IFADDR_RUNLOCK(&in6_ifa_tracker); 2012 ND6_RUNLOCK(); 2013 ND6_ONLINK_UNLOCK(); 2014 } 2015 2016 /* 2017 * Add or remove interface route specified by @dst, @netmask and @ifp. 2018 * ifa can be NULL. 2019 * Returns 0 on success 2020 */ 2021 static int 2022 nd6_prefix_rtrequest(uint32_t fibnum, int cmd, struct sockaddr_in6 *dst, 2023 struct sockaddr_in6 *netmask, struct ifnet *ifp, struct ifaddr *ifa) 2024 { 2025 struct epoch_tracker et; 2026 int error; 2027 2028 /* Prepare gateway */ 2029 struct sockaddr_dl_short sdl = { 2030 .sdl_family = AF_LINK, 2031 .sdl_len = sizeof(struct sockaddr_dl_short), 2032 .sdl_type = ifp->if_type, 2033 .sdl_index = ifp->if_index, 2034 }; 2035 2036 struct rt_addrinfo info = { 2037 .rti_ifa = ifa, 2038 .rti_ifp = ifp, 2039 .rti_flags = RTF_PINNED | ((netmask != NULL) ? 0 : RTF_HOST), 2040 .rti_info = { 2041 [RTAX_DST] = (struct sockaddr *)dst, 2042 [RTAX_NETMASK] = (struct sockaddr *)netmask, 2043 [RTAX_GATEWAY] = (struct sockaddr *)&sdl, 2044 }, 2045 }; 2046 /* Don't set additional per-gw filters on removal */ 2047 2048 NET_EPOCH_ENTER(et); 2049 error = rib_handle_ifaddr_info(fibnum, cmd, &info); 2050 NET_EPOCH_EXIT(et); 2051 return (error); 2052 } 2053 2054 static int 2055 nd6_prefix_onlink_rtrequest(struct nd_prefix *pr, struct ifaddr *ifa) 2056 { 2057 int error; 2058 2059 struct sockaddr_in6 mask6 = { 2060 .sin6_family = AF_INET6, 2061 .sin6_len = sizeof(struct sockaddr_in6), 2062 .sin6_addr = pr->ndpr_mask, 2063 }; 2064 struct sockaddr_in6 *pmask6 = (pr->ndpr_plen != 128) ? &mask6 : NULL; 2065 2066 error = nd6_prefix_rtrequest(pr->ndpr_ifp->if_fib, RTM_ADD, 2067 &pr->ndpr_prefix, pmask6, pr->ndpr_ifp, ifa); 2068 if (error == 0) 2069 pr->ndpr_stateflags |= NDPRF_ONLINK; 2070 2071 return (error); 2072 } 2073 2074 static int 2075 nd6_prefix_onlink(struct nd_prefix *pr) 2076 { 2077 struct epoch_tracker et; 2078 struct ifaddr *ifa; 2079 struct ifnet *ifp = pr->ndpr_ifp; 2080 struct nd_prefix *opr; 2081 char ip6buf[INET6_ADDRSTRLEN]; 2082 int error; 2083 2084 ND6_ONLINK_LOCK_ASSERT(); 2085 ND6_UNLOCK_ASSERT(); 2086 2087 if ((pr->ndpr_stateflags & NDPRF_ONLINK) != 0) 2088 return (EEXIST); 2089 2090 /* 2091 * Add the interface route associated with the prefix. Before 2092 * installing the route, check if there's the same prefix on another 2093 * interface, and the prefix has already installed the interface route. 2094 * Although such a configuration is expected to be rare, we explicitly 2095 * allow it. 2096 */ 2097 ND6_RLOCK(); 2098 LIST_FOREACH(opr, &V_nd_prefix, ndpr_entry) { 2099 if (opr == pr) 2100 continue; 2101 2102 if ((opr->ndpr_stateflags & NDPRF_ONLINK) == 0) 2103 continue; 2104 2105 if (!V_rt_add_addr_allfibs && 2106 opr->ndpr_ifp->if_fib != pr->ndpr_ifp->if_fib) 2107 continue; 2108 2109 if (opr->ndpr_plen == pr->ndpr_plen && 2110 in6_are_prefix_equal(&pr->ndpr_prefix.sin6_addr, 2111 &opr->ndpr_prefix.sin6_addr, pr->ndpr_plen)) { 2112 ND6_RUNLOCK(); 2113 return (0); 2114 } 2115 } 2116 ND6_RUNLOCK(); 2117 2118 /* 2119 * We prefer link-local addresses as the associated interface address. 2120 */ 2121 /* search for a link-local addr */ 2122 NET_EPOCH_ENTER(et); 2123 ifa = (struct ifaddr *)in6ifa_ifpforlinklocal(ifp, 2124 IN6_IFF_NOTREADY | IN6_IFF_ANYCAST); 2125 if (ifa == NULL) { 2126 /* XXX: freebsd does not have ifa_ifwithaf */ 2127 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 2128 if (ifa->ifa_addr->sa_family == AF_INET6) { 2129 ifa_ref(ifa); 2130 break; 2131 } 2132 } 2133 /* should we care about ia6_flags? */ 2134 } 2135 if (ifa == NULL) { 2136 /* 2137 * This can still happen, when, for example, we receive an RA 2138 * containing a prefix with the L bit set and the A bit clear, 2139 * after removing all IPv6 addresses on the receiving 2140 * interface. This should, of course, be rare though. 2141 */ 2142 nd6log((LOG_NOTICE, 2143 "%s: failed to find any ifaddr to add route for a " 2144 "prefix(%s/%d) on %s\n", __func__, 2145 ip6_sprintf(ip6buf, &pr->ndpr_prefix.sin6_addr), 2146 pr->ndpr_plen, if_name(ifp))); 2147 error = 0; 2148 } else { 2149 error = nd6_prefix_onlink_rtrequest(pr, ifa); 2150 ifa_free(ifa); 2151 } 2152 NET_EPOCH_EXIT(et); 2153 2154 return (error); 2155 } 2156 2157 int 2158 nd6_prefix_offlink(struct nd_prefix *pr) 2159 { 2160 int error = 0; 2161 struct ifnet *ifp = pr->ndpr_ifp; 2162 struct nd_prefix *opr; 2163 char ip6buf[INET6_ADDRSTRLEN]; 2164 uint64_t genid; 2165 int a_failure; 2166 2167 ND6_ONLINK_LOCK_ASSERT(); 2168 ND6_UNLOCK_ASSERT(); 2169 2170 if ((pr->ndpr_stateflags & NDPRF_ONLINK) == 0) 2171 return (EEXIST); 2172 2173 struct sockaddr_in6 mask6 = { 2174 .sin6_family = AF_INET6, 2175 .sin6_len = sizeof(struct sockaddr_in6), 2176 .sin6_addr = pr->ndpr_mask, 2177 }; 2178 struct sockaddr_in6 *pmask6 = (pr->ndpr_plen != 128) ? &mask6 : NULL; 2179 2180 error = nd6_prefix_rtrequest(ifp->if_fib, RTM_DELETE, 2181 &pr->ndpr_prefix, pmask6, ifp, NULL); 2182 2183 a_failure = 1; 2184 if (error == 0) { 2185 pr->ndpr_stateflags &= ~NDPRF_ONLINK; 2186 2187 /* 2188 * There might be the same prefix on another interface, 2189 * the prefix which could not be on-link just because we have 2190 * the interface route (see comments in nd6_prefix_onlink). 2191 * If there's one, try to make the prefix on-link on the 2192 * interface. 2193 */ 2194 ND6_RLOCK(); 2195 restart: 2196 LIST_FOREACH(opr, &V_nd_prefix, ndpr_entry) { 2197 /* 2198 * KAME specific: detached prefixes should not be 2199 * on-link. 2200 */ 2201 if (opr == pr || (opr->ndpr_stateflags & 2202 (NDPRF_ONLINK | NDPRF_DETACHED)) != 0) 2203 continue; 2204 2205 if (opr->ndpr_plen == pr->ndpr_plen && 2206 in6_are_prefix_equal(&pr->ndpr_prefix.sin6_addr, 2207 &opr->ndpr_prefix.sin6_addr, pr->ndpr_plen)) { 2208 int e; 2209 2210 genid = V_nd6_list_genid; 2211 ND6_RUNLOCK(); 2212 if ((e = nd6_prefix_onlink(opr)) != 0) { 2213 nd6log((LOG_ERR, 2214 "%s: failed to recover a prefix " 2215 "%s/%d from %s to %s (errno=%d)\n", 2216 __func__, ip6_sprintf(ip6buf, 2217 &opr->ndpr_prefix.sin6_addr), 2218 opr->ndpr_plen, if_name(ifp), 2219 if_name(opr->ndpr_ifp), e)); 2220 } else 2221 a_failure = 0; 2222 ND6_RLOCK(); 2223 if (genid != V_nd6_list_genid) 2224 goto restart; 2225 } 2226 } 2227 ND6_RUNLOCK(); 2228 } else { 2229 /* XXX: can we still set the NDPRF_ONLINK flag? */ 2230 nd6log((LOG_ERR, 2231 "%s: failed to delete route: %s/%d on %s (errno=%d)\n", 2232 __func__, ip6_sprintf(ip6buf, &pr->ndpr_prefix.sin6_addr), 2233 pr->ndpr_plen, if_name(ifp), error)); 2234 } 2235 2236 if (a_failure) 2237 lltable_prefix_free(AF_INET6, 2238 (struct sockaddr *)&pr->ndpr_prefix, 2239 (struct sockaddr *)&mask6, LLE_STATIC); 2240 2241 return (error); 2242 } 2243 2244 /* 2245 * ia0 - corresponding public address 2246 */ 2247 int 2248 in6_tmpifadd(const struct in6_ifaddr *ia0, int forcegen, int delay) 2249 { 2250 struct ifnet *ifp = ia0->ia_ifa.ifa_ifp; 2251 struct in6_ifaddr *newia; 2252 struct in6_aliasreq ifra; 2253 int error; 2254 int trylimit = 3; /* XXX: adhoc value */ 2255 int updateflags; 2256 u_int32_t randid[2]; 2257 time_t vltime0, pltime0; 2258 2259 in6_prepare_ifra(&ifra, &ia0->ia_addr.sin6_addr, 2260 &ia0->ia_prefixmask.sin6_addr); 2261 2262 ifra.ifra_addr = ia0->ia_addr; /* XXX: do we need this ? */ 2263 /* clear the old IFID */ 2264 IN6_MASK_ADDR(&ifra.ifra_addr.sin6_addr, 2265 &ifra.ifra_prefixmask.sin6_addr); 2266 2267 again: 2268 if (in6_get_tmpifid(ifp, (u_int8_t *)randid, 2269 (const u_int8_t *)&ia0->ia_addr.sin6_addr.s6_addr[8], forcegen)) { 2270 nd6log((LOG_NOTICE, "%s: failed to find a good random IFID\n", 2271 __func__)); 2272 return (EINVAL); 2273 } 2274 ifra.ifra_addr.sin6_addr.s6_addr32[2] |= 2275 (randid[0] & ~(ifra.ifra_prefixmask.sin6_addr.s6_addr32[2])); 2276 ifra.ifra_addr.sin6_addr.s6_addr32[3] |= 2277 (randid[1] & ~(ifra.ifra_prefixmask.sin6_addr.s6_addr32[3])); 2278 2279 /* 2280 * in6_get_tmpifid() quite likely provided a unique interface ID. 2281 * However, we may still have a chance to see collision, because 2282 * there may be a time lag between generation of the ID and generation 2283 * of the address. So, we'll do one more sanity check. 2284 */ 2285 2286 if (in6_localip(&ifra.ifra_addr.sin6_addr) != 0) { 2287 if (trylimit-- > 0) { 2288 forcegen = 1; 2289 goto again; 2290 } 2291 2292 /* Give up. Something strange should have happened. */ 2293 nd6log((LOG_NOTICE, "%s: failed to find a unique random IFID\n", 2294 __func__)); 2295 return (EEXIST); 2296 } 2297 2298 /* 2299 * The Valid Lifetime is the lower of the Valid Lifetime of the 2300 * public address or TEMP_VALID_LIFETIME. 2301 * The Preferred Lifetime is the lower of the Preferred Lifetime 2302 * of the public address or TEMP_PREFERRED_LIFETIME - 2303 * DESYNC_FACTOR. 2304 */ 2305 if (ia0->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) { 2306 vltime0 = IFA6_IS_INVALID(ia0) ? 0 : 2307 (ia0->ia6_lifetime.ia6t_vltime - 2308 (time_uptime - ia0->ia6_updatetime)); 2309 if (vltime0 > V_ip6_temp_valid_lifetime) 2310 vltime0 = V_ip6_temp_valid_lifetime; 2311 } else 2312 vltime0 = V_ip6_temp_valid_lifetime; 2313 if (ia0->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) { 2314 pltime0 = IFA6_IS_DEPRECATED(ia0) ? 0 : 2315 (ia0->ia6_lifetime.ia6t_pltime - 2316 (time_uptime - ia0->ia6_updatetime)); 2317 if (pltime0 > V_ip6_temp_preferred_lifetime - V_ip6_desync_factor){ 2318 pltime0 = V_ip6_temp_preferred_lifetime - 2319 V_ip6_desync_factor; 2320 } 2321 } else 2322 pltime0 = V_ip6_temp_preferred_lifetime - V_ip6_desync_factor; 2323 ifra.ifra_lifetime.ia6t_vltime = vltime0; 2324 ifra.ifra_lifetime.ia6t_pltime = pltime0; 2325 2326 /* 2327 * A temporary address is created only if this calculated Preferred 2328 * Lifetime is greater than REGEN_ADVANCE time units. 2329 */ 2330 if (ifra.ifra_lifetime.ia6t_pltime <= V_ip6_temp_regen_advance) 2331 return (0); 2332 2333 /* XXX: scope zone ID? */ 2334 2335 ifra.ifra_flags |= (IN6_IFF_AUTOCONF|IN6_IFF_TEMPORARY); 2336 2337 /* allocate ifaddr structure, link into chain, etc. */ 2338 updateflags = 0; 2339 if (delay) 2340 updateflags |= IN6_IFAUPDATE_DADDELAY; 2341 if ((error = in6_update_ifa(ifp, &ifra, NULL, updateflags)) != 0) 2342 return (error); 2343 2344 newia = in6ifa_ifpwithaddr(ifp, &ifra.ifra_addr.sin6_addr); 2345 if (newia == NULL) { /* XXX: can it happen? */ 2346 nd6log((LOG_ERR, 2347 "%s: ifa update succeeded, but we got no ifaddr\n", 2348 __func__)); 2349 return (EINVAL); /* XXX */ 2350 } 2351 newia->ia6_ndpr = ia0->ia6_ndpr; 2352 newia->ia6_ndpr->ndpr_addrcnt++; 2353 ifa_free(&newia->ia_ifa); 2354 2355 /* 2356 * A newly added address might affect the status of other addresses. 2357 * XXX: when the temporary address is generated with a new public 2358 * address, the onlink check is redundant. However, it would be safe 2359 * to do the check explicitly everywhere a new address is generated, 2360 * and, in fact, we surely need the check when we create a new 2361 * temporary address due to deprecation of an old temporary address. 2362 */ 2363 pfxlist_onlink_check(); 2364 2365 return (0); 2366 } 2367 2368 static int 2369 rt6_deleteroute(const struct rtentry *rt, const struct nhop_object *nh, 2370 void *arg) 2371 { 2372 struct in6_addr *gate = (struct in6_addr *)arg; 2373 int nh_rt_flags; 2374 2375 if (nh->gw_sa.sa_family != AF_INET6) 2376 return (0); 2377 2378 if (!IN6_ARE_ADDR_EQUAL(gate, &nh->gw6_sa.sin6_addr)) { 2379 return (0); 2380 } 2381 2382 /* 2383 * Do not delete a static route. 2384 * XXX: this seems to be a bit ad-hoc. Should we consider the 2385 * 'cloned' bit instead? 2386 */ 2387 nh_rt_flags = nhop_get_rtflags(nh); 2388 if ((nh_rt_flags & RTF_STATIC) != 0) 2389 return (0); 2390 2391 /* 2392 * We delete only host route. This means, in particular, we don't 2393 * delete default route. 2394 */ 2395 if ((nh_rt_flags & RTF_HOST) == 0) 2396 return (0); 2397 2398 return (1); 2399 #undef SIN6 2400 } 2401 2402 /* 2403 * Delete all the routing table entries that use the specified gateway. 2404 * XXX: this function causes search through all entries of routing table, so 2405 * it shouldn't be called when acting as a router. 2406 */ 2407 void 2408 rt6_flush(struct in6_addr *gateway, struct ifnet *ifp) 2409 { 2410 2411 /* We'll care only link-local addresses */ 2412 if (!IN6_IS_ADDR_LINKLOCAL(gateway)) 2413 return; 2414 2415 /* XXX Do we really need to walk any but the default FIB? */ 2416 rib_foreach_table_walk_del(AF_INET6, rt6_deleteroute, (void *)gateway); 2417 } 2418 2419 int 2420 nd6_setdefaultiface(int ifindex) 2421 { 2422 2423 if (V_nd6_defifindex != ifindex) { 2424 V_nd6_defifindex = ifindex; 2425 if (V_nd6_defifindex != 0) { 2426 struct epoch_tracker et; 2427 2428 /* 2429 * XXXGL: this function should use ifnet_byindex_ref! 2430 */ 2431 NET_EPOCH_ENTER(et); 2432 V_nd6_defifp = ifnet_byindex(V_nd6_defifindex); 2433 NET_EPOCH_EXIT(et); 2434 if (V_nd6_defifp == NULL) 2435 return (EINVAL); 2436 } else 2437 V_nd6_defifp = NULL; 2438 2439 /* 2440 * Our current implementation assumes one-to-one mapping between 2441 * interfaces and links, so it would be natural to use the 2442 * default interface as the default link. 2443 */ 2444 scope6_setdefault(V_nd6_defifp); 2445 } 2446 2447 return (0); 2448 } 2449 2450 bool 2451 nd6_defrouter_list_empty(void) 2452 { 2453 2454 return (TAILQ_EMPTY(&V_nd6_defrouter)); 2455 } 2456 2457 void 2458 nd6_defrouter_timer(void) 2459 { 2460 struct nd_defrouter *dr, *ndr; 2461 struct nd6_drhead drq; 2462 2463 TAILQ_INIT(&drq); 2464 2465 ND6_WLOCK(); 2466 TAILQ_FOREACH_SAFE(dr, &V_nd6_defrouter, dr_entry, ndr) 2467 if (dr->expire && dr->expire < time_uptime) 2468 defrouter_unlink(dr, &drq); 2469 ND6_WUNLOCK(); 2470 2471 while ((dr = TAILQ_FIRST(&drq)) != NULL) { 2472 TAILQ_REMOVE(&drq, dr, dr_entry); 2473 defrouter_del(dr); 2474 } 2475 } 2476 2477 /* 2478 * Nuke default router list entries toward ifp. 2479 * We defer removal of default router list entries that is installed in the 2480 * routing table, in order to keep additional side effects as small as possible. 2481 */ 2482 void 2483 nd6_defrouter_purge(struct ifnet *ifp) 2484 { 2485 struct nd_defrouter *dr, *ndr; 2486 struct nd6_drhead drq; 2487 2488 TAILQ_INIT(&drq); 2489 2490 ND6_WLOCK(); 2491 TAILQ_FOREACH_SAFE(dr, &V_nd6_defrouter, dr_entry, ndr) { 2492 if (dr->installed) 2493 continue; 2494 if (dr->ifp == ifp) 2495 defrouter_unlink(dr, &drq); 2496 } 2497 TAILQ_FOREACH_SAFE(dr, &V_nd6_defrouter, dr_entry, ndr) { 2498 if (!dr->installed) 2499 continue; 2500 if (dr->ifp == ifp) 2501 defrouter_unlink(dr, &drq); 2502 } 2503 ND6_WUNLOCK(); 2504 2505 /* Delete the unlinked router objects. */ 2506 while ((dr = TAILQ_FIRST(&drq)) != NULL) { 2507 TAILQ_REMOVE(&drq, dr, dr_entry); 2508 defrouter_del(dr); 2509 } 2510 } 2511 2512 void 2513 nd6_defrouter_flush_all(void) 2514 { 2515 struct nd_defrouter *dr; 2516 struct nd6_drhead drq; 2517 2518 TAILQ_INIT(&drq); 2519 2520 ND6_WLOCK(); 2521 while ((dr = TAILQ_FIRST(&V_nd6_defrouter)) != NULL) 2522 defrouter_unlink(dr, &drq); 2523 ND6_WUNLOCK(); 2524 2525 while ((dr = TAILQ_FIRST(&drq)) != NULL) { 2526 TAILQ_REMOVE(&drq, dr, dr_entry); 2527 defrouter_del(dr); 2528 } 2529 } 2530 2531 void 2532 nd6_defrouter_init(void) 2533 { 2534 2535 TAILQ_INIT(&V_nd6_defrouter); 2536 } 2537 2538 static int 2539 nd6_sysctl_drlist(SYSCTL_HANDLER_ARGS) 2540 { 2541 struct in6_defrouter d; 2542 struct nd_defrouter *dr; 2543 int error; 2544 2545 if (req->newptr != NULL) 2546 return (EPERM); 2547 2548 error = sysctl_wire_old_buffer(req, 0); 2549 if (error != 0) 2550 return (error); 2551 2552 bzero(&d, sizeof(d)); 2553 d.rtaddr.sin6_family = AF_INET6; 2554 d.rtaddr.sin6_len = sizeof(d.rtaddr); 2555 2556 ND6_RLOCK(); 2557 TAILQ_FOREACH(dr, &V_nd6_defrouter, dr_entry) { 2558 d.rtaddr.sin6_addr = dr->rtaddr; 2559 error = sa6_recoverscope(&d.rtaddr); 2560 if (error != 0) 2561 break; 2562 d.flags = dr->raflags; 2563 d.rtlifetime = dr->rtlifetime; 2564 d.expire = dr->expire + (time_second - time_uptime); 2565 d.if_index = dr->ifp->if_index; 2566 error = SYSCTL_OUT(req, &d, sizeof(d)); 2567 if (error != 0) 2568 break; 2569 } 2570 ND6_RUNLOCK(); 2571 return (error); 2572 } 2573 SYSCTL_PROC(_net_inet6_icmp6, ICMPV6CTL_ND6_DRLIST, nd6_drlist, 2574 CTLTYPE_OPAQUE | CTLFLAG_RD | CTLFLAG_MPSAFE, 2575 NULL, 0, nd6_sysctl_drlist, "S,in6_defrouter", 2576 "NDP default router list"); 2577