1 /*- 2 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project. 3 * All rights reserved. 4 * 5 * Redistribution and use in source and binary forms, with or without 6 * modification, are permitted provided that the following conditions 7 * are met: 8 * 1. Redistributions of source code must retain the above copyright 9 * notice, this list of conditions and the following disclaimer. 10 * 2. Redistributions in binary form must reproduce the above copyright 11 * notice, this list of conditions and the following disclaimer in the 12 * documentation and/or other materials provided with the distribution. 13 * 3. Neither the name of the project nor the names of its contributors 14 * may be used to endorse or promote products derived from this software 15 * without specific prior written permission. 16 * 17 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND 18 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 19 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 20 * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE 21 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 22 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 23 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 24 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 25 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 26 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 27 * SUCH DAMAGE. 28 * 29 * $KAME: nd6_rtr.c,v 1.111 2001/04/27 01:37:15 jinmei Exp $ 30 */ 31 32 #include <sys/cdefs.h> 33 __FBSDID("$FreeBSD$"); 34 35 #include "opt_inet.h" 36 #include "opt_inet6.h" 37 38 #include <sys/param.h> 39 #include <sys/systm.h> 40 #include <sys/malloc.h> 41 #include <sys/mbuf.h> 42 #include <sys/refcount.h> 43 #include <sys/socket.h> 44 #include <sys/sockio.h> 45 #include <sys/time.h> 46 #include <sys/kernel.h> 47 #include <sys/lock.h> 48 #include <sys/errno.h> 49 #include <sys/rmlock.h> 50 #include <sys/rwlock.h> 51 #include <sys/syslog.h> 52 #include <sys/queue.h> 53 54 #include <net/if.h> 55 #include <net/if_var.h> 56 #include <net/if_types.h> 57 #include <net/if_dl.h> 58 #include <net/route.h> 59 #include <net/route_var.h> 60 #include <net/radix.h> 61 #include <net/vnet.h> 62 63 #include <netinet/in.h> 64 #include <net/if_llatbl.h> 65 #include <netinet6/in6_var.h> 66 #include <netinet6/in6_ifattach.h> 67 #include <netinet/ip6.h> 68 #include <netinet6/ip6_var.h> 69 #include <netinet6/nd6.h> 70 #include <netinet/icmp6.h> 71 #include <netinet6/scope6_var.h> 72 73 static int rtpref(struct nd_defrouter *); 74 static struct nd_defrouter *defrtrlist_update(struct nd_defrouter *); 75 static int prelist_update(struct nd_prefixctl *, struct nd_defrouter *, 76 struct mbuf *, int); 77 static struct in6_ifaddr *in6_ifadd(struct nd_prefixctl *, int); 78 static struct nd_pfxrouter *pfxrtr_lookup(struct nd_prefix *, 79 struct nd_defrouter *); 80 static void pfxrtr_add(struct nd_prefix *, struct nd_defrouter *); 81 static void pfxrtr_del(struct nd_pfxrouter *); 82 static struct nd_pfxrouter *find_pfxlist_reachable_router(struct nd_prefix *); 83 static void defrouter_delreq(struct nd_defrouter *); 84 static void nd6_rtmsg(int, struct rtentry *); 85 86 static int in6_init_prefix_ltimes(struct nd_prefix *); 87 static void in6_init_address_ltimes(struct nd_prefix *, 88 struct in6_addrlifetime *); 89 90 static int rt6_deleteroute(const struct rtentry *, void *); 91 92 VNET_DECLARE(int, nd6_recalc_reachtm_interval); 93 #define V_nd6_recalc_reachtm_interval VNET(nd6_recalc_reachtm_interval) 94 95 static VNET_DEFINE(struct ifnet *, nd6_defifp); 96 VNET_DEFINE(int, nd6_defifindex); 97 #define V_nd6_defifp VNET(nd6_defifp) 98 99 VNET_DEFINE(int, ip6_use_tempaddr) = 0; 100 101 VNET_DEFINE(int, ip6_desync_factor); 102 VNET_DEFINE(u_int32_t, ip6_temp_preferred_lifetime) = DEF_TEMP_PREFERRED_LIFETIME; 103 VNET_DEFINE(u_int32_t, ip6_temp_valid_lifetime) = DEF_TEMP_VALID_LIFETIME; 104 105 VNET_DEFINE(int, ip6_temp_regen_advance) = TEMPADDR_REGEN_ADVANCE; 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 /* 115 * Receive Router Solicitation Message - just for routers. 116 * Router solicitation/advertisement is mostly managed by userland program 117 * (rtadvd) so here we have no function like nd6_ra_output(). 118 * 119 * Based on RFC 2461 120 */ 121 void 122 nd6_rs_input(struct mbuf *m, int off, int icmp6len) 123 { 124 struct ifnet *ifp = m->m_pkthdr.rcvif; 125 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *); 126 struct nd_router_solicit *nd_rs; 127 struct in6_addr saddr6 = ip6->ip6_src; 128 char *lladdr = NULL; 129 int lladdrlen = 0; 130 union nd_opts ndopts; 131 char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN]; 132 133 /* 134 * Accept RS only when V_ip6_forwarding=1 and the interface has 135 * no ND6_IFF_ACCEPT_RTADV. 136 */ 137 if (!V_ip6_forwarding || ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV) 138 goto freeit; 139 140 /* Sanity checks */ 141 if (ip6->ip6_hlim != 255) { 142 nd6log((LOG_ERR, 143 "nd6_rs_input: invalid hlim (%d) from %s to %s on %s\n", 144 ip6->ip6_hlim, ip6_sprintf(ip6bufs, &ip6->ip6_src), 145 ip6_sprintf(ip6bufd, &ip6->ip6_dst), if_name(ifp))); 146 goto bad; 147 } 148 149 /* 150 * Don't update the neighbor cache, if src = ::. 151 * This indicates that the src has no IP address assigned yet. 152 */ 153 if (IN6_IS_ADDR_UNSPECIFIED(&saddr6)) 154 goto freeit; 155 156 #ifndef PULLDOWN_TEST 157 IP6_EXTHDR_CHECK(m, off, icmp6len,); 158 nd_rs = (struct nd_router_solicit *)((caddr_t)ip6 + off); 159 #else 160 IP6_EXTHDR_GET(nd_rs, struct nd_router_solicit *, m, off, icmp6len); 161 if (nd_rs == NULL) { 162 ICMP6STAT_INC(icp6s_tooshort); 163 return; 164 } 165 #endif 166 167 icmp6len -= sizeof(*nd_rs); 168 nd6_option_init(nd_rs + 1, icmp6len, &ndopts); 169 if (nd6_options(&ndopts) < 0) { 170 nd6log((LOG_INFO, 171 "nd6_rs_input: invalid ND option, ignored\n")); 172 /* nd6_options have incremented stats */ 173 goto freeit; 174 } 175 176 if (ndopts.nd_opts_src_lladdr) { 177 lladdr = (char *)(ndopts.nd_opts_src_lladdr + 1); 178 lladdrlen = ndopts.nd_opts_src_lladdr->nd_opt_len << 3; 179 } 180 181 if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) { 182 nd6log((LOG_INFO, 183 "nd6_rs_input: lladdrlen mismatch for %s " 184 "(if %d, RS packet %d)\n", 185 ip6_sprintf(ip6bufs, &saddr6), 186 ifp->if_addrlen, lladdrlen - 2)); 187 goto bad; 188 } 189 190 nd6_cache_lladdr(ifp, &saddr6, lladdr, lladdrlen, ND_ROUTER_SOLICIT, 0); 191 192 freeit: 193 m_freem(m); 194 return; 195 196 bad: 197 ICMP6STAT_INC(icp6s_badrs); 198 m_freem(m); 199 } 200 201 /* 202 * Receive Router Advertisement Message. 203 * 204 * Based on RFC 2461 205 * TODO: on-link bit on prefix information 206 * TODO: ND_RA_FLAG_{OTHER,MANAGED} processing 207 */ 208 void 209 nd6_ra_input(struct mbuf *m, int off, int icmp6len) 210 { 211 struct ifnet *ifp = m->m_pkthdr.rcvif; 212 struct nd_ifinfo *ndi = ND_IFINFO(ifp); 213 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *); 214 struct nd_router_advert *nd_ra; 215 struct in6_addr saddr6 = ip6->ip6_src; 216 int mcast = 0; 217 union nd_opts ndopts; 218 struct nd_defrouter *dr; 219 char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN]; 220 221 dr = NULL; 222 223 /* 224 * We only accept RAs only when the per-interface flag 225 * ND6_IFF_ACCEPT_RTADV is on the receiving interface. 226 */ 227 if (!(ndi->flags & ND6_IFF_ACCEPT_RTADV)) 228 goto freeit; 229 230 if (ip6->ip6_hlim != 255) { 231 nd6log((LOG_ERR, 232 "nd6_ra_input: invalid hlim (%d) from %s to %s on %s\n", 233 ip6->ip6_hlim, ip6_sprintf(ip6bufs, &ip6->ip6_src), 234 ip6_sprintf(ip6bufd, &ip6->ip6_dst), if_name(ifp))); 235 goto bad; 236 } 237 238 if (!IN6_IS_ADDR_LINKLOCAL(&saddr6)) { 239 nd6log((LOG_ERR, 240 "nd6_ra_input: src %s is not link-local\n", 241 ip6_sprintf(ip6bufs, &saddr6))); 242 goto bad; 243 } 244 245 #ifndef PULLDOWN_TEST 246 IP6_EXTHDR_CHECK(m, off, icmp6len,); 247 nd_ra = (struct nd_router_advert *)((caddr_t)ip6 + off); 248 #else 249 IP6_EXTHDR_GET(nd_ra, struct nd_router_advert *, m, off, icmp6len); 250 if (nd_ra == NULL) { 251 ICMP6STAT_INC(icp6s_tooshort); 252 return; 253 } 254 #endif 255 256 icmp6len -= sizeof(*nd_ra); 257 nd6_option_init(nd_ra + 1, icmp6len, &ndopts); 258 if (nd6_options(&ndopts) < 0) { 259 nd6log((LOG_INFO, 260 "nd6_ra_input: invalid ND option, ignored\n")); 261 /* nd6_options have incremented stats */ 262 goto freeit; 263 } 264 265 { 266 struct nd_defrouter dr0; 267 u_int32_t advreachable = nd_ra->nd_ra_reachable; 268 269 /* remember if this is a multicasted advertisement */ 270 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) 271 mcast = 1; 272 273 bzero(&dr0, sizeof(dr0)); 274 dr0.rtaddr = saddr6; 275 dr0.raflags = nd_ra->nd_ra_flags_reserved; 276 /* 277 * Effectively-disable routes from RA messages when 278 * ND6_IFF_NO_RADR enabled on the receiving interface or 279 * (ip6.forwarding == 1 && ip6.rfc6204w3 != 1). 280 */ 281 if (ndi->flags & ND6_IFF_NO_RADR) 282 dr0.rtlifetime = 0; 283 else if (V_ip6_forwarding && !V_ip6_rfc6204w3) 284 dr0.rtlifetime = 0; 285 else 286 dr0.rtlifetime = ntohs(nd_ra->nd_ra_router_lifetime); 287 dr0.expire = time_uptime + dr0.rtlifetime; 288 dr0.ifp = ifp; 289 /* unspecified or not? (RFC 2461 6.3.4) */ 290 if (advreachable) { 291 advreachable = ntohl(advreachable); 292 if (advreachable <= MAX_REACHABLE_TIME && 293 ndi->basereachable != advreachable) { 294 ndi->basereachable = advreachable; 295 ndi->reachable = ND_COMPUTE_RTIME(ndi->basereachable); 296 ndi->recalctm = V_nd6_recalc_reachtm_interval; /* reset */ 297 } 298 } 299 if (nd_ra->nd_ra_retransmit) 300 ndi->retrans = ntohl(nd_ra->nd_ra_retransmit); 301 if (nd_ra->nd_ra_curhoplimit) { 302 if (ndi->chlim < nd_ra->nd_ra_curhoplimit) 303 ndi->chlim = nd_ra->nd_ra_curhoplimit; 304 else if (ndi->chlim != nd_ra->nd_ra_curhoplimit) { 305 log(LOG_ERR, "RA with a lower CurHopLimit sent from " 306 "%s on %s (current = %d, received = %d). " 307 "Ignored.\n", ip6_sprintf(ip6bufs, &ip6->ip6_src), 308 if_name(ifp), ndi->chlim, nd_ra->nd_ra_curhoplimit); 309 } 310 } 311 dr = defrtrlist_update(&dr0); 312 } 313 314 /* 315 * prefix 316 */ 317 if (ndopts.nd_opts_pi) { 318 struct nd_opt_hdr *pt; 319 struct nd_opt_prefix_info *pi = NULL; 320 struct nd_prefixctl pr; 321 322 for (pt = (struct nd_opt_hdr *)ndopts.nd_opts_pi; 323 pt <= (struct nd_opt_hdr *)ndopts.nd_opts_pi_end; 324 pt = (struct nd_opt_hdr *)((caddr_t)pt + 325 (pt->nd_opt_len << 3))) { 326 if (pt->nd_opt_type != ND_OPT_PREFIX_INFORMATION) 327 continue; 328 pi = (struct nd_opt_prefix_info *)pt; 329 330 if (pi->nd_opt_pi_len != 4) { 331 nd6log((LOG_INFO, 332 "nd6_ra_input: invalid option " 333 "len %d for prefix information option, " 334 "ignored\n", pi->nd_opt_pi_len)); 335 continue; 336 } 337 338 if (128 < pi->nd_opt_pi_prefix_len) { 339 nd6log((LOG_INFO, 340 "nd6_ra_input: invalid prefix " 341 "len %d for prefix information option, " 342 "ignored\n", pi->nd_opt_pi_prefix_len)); 343 continue; 344 } 345 346 if (IN6_IS_ADDR_MULTICAST(&pi->nd_opt_pi_prefix) 347 || IN6_IS_ADDR_LINKLOCAL(&pi->nd_opt_pi_prefix)) { 348 nd6log((LOG_INFO, 349 "nd6_ra_input: invalid prefix " 350 "%s, ignored\n", 351 ip6_sprintf(ip6bufs, 352 &pi->nd_opt_pi_prefix))); 353 continue; 354 } 355 356 bzero(&pr, sizeof(pr)); 357 pr.ndpr_prefix.sin6_family = AF_INET6; 358 pr.ndpr_prefix.sin6_len = sizeof(pr.ndpr_prefix); 359 pr.ndpr_prefix.sin6_addr = pi->nd_opt_pi_prefix; 360 pr.ndpr_ifp = (struct ifnet *)m->m_pkthdr.rcvif; 361 362 pr.ndpr_raf_onlink = (pi->nd_opt_pi_flags_reserved & 363 ND_OPT_PI_FLAG_ONLINK) ? 1 : 0; 364 pr.ndpr_raf_auto = (pi->nd_opt_pi_flags_reserved & 365 ND_OPT_PI_FLAG_AUTO) ? 1 : 0; 366 pr.ndpr_plen = pi->nd_opt_pi_prefix_len; 367 pr.ndpr_vltime = ntohl(pi->nd_opt_pi_valid_time); 368 pr.ndpr_pltime = ntohl(pi->nd_opt_pi_preferred_time); 369 (void)prelist_update(&pr, dr, m, mcast); 370 } 371 } 372 if (dr != NULL) { 373 defrouter_rele(dr); 374 dr = NULL; 375 } 376 377 /* 378 * MTU 379 */ 380 if (ndopts.nd_opts_mtu && ndopts.nd_opts_mtu->nd_opt_mtu_len == 1) { 381 u_long mtu; 382 u_long maxmtu; 383 384 mtu = (u_long)ntohl(ndopts.nd_opts_mtu->nd_opt_mtu_mtu); 385 386 /* lower bound */ 387 if (mtu < IPV6_MMTU) { 388 nd6log((LOG_INFO, "nd6_ra_input: bogus mtu option " 389 "mtu=%lu sent from %s, ignoring\n", 390 mtu, ip6_sprintf(ip6bufs, &ip6->ip6_src))); 391 goto skip; 392 } 393 394 /* upper bound */ 395 maxmtu = (ndi->maxmtu && ndi->maxmtu < ifp->if_mtu) 396 ? ndi->maxmtu : ifp->if_mtu; 397 if (mtu <= maxmtu) { 398 int change = (ndi->linkmtu != mtu); 399 400 ndi->linkmtu = mtu; 401 if (change) /* in6_maxmtu may change */ 402 in6_setmaxmtu(); 403 } else { 404 nd6log((LOG_INFO, "nd6_ra_input: bogus mtu " 405 "mtu=%lu sent from %s; " 406 "exceeds maxmtu %lu, ignoring\n", 407 mtu, ip6_sprintf(ip6bufs, &ip6->ip6_src), maxmtu)); 408 } 409 } 410 411 skip: 412 413 /* 414 * Source link layer address 415 */ 416 { 417 char *lladdr = NULL; 418 int lladdrlen = 0; 419 420 if (ndopts.nd_opts_src_lladdr) { 421 lladdr = (char *)(ndopts.nd_opts_src_lladdr + 1); 422 lladdrlen = ndopts.nd_opts_src_lladdr->nd_opt_len << 3; 423 } 424 425 if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) { 426 nd6log((LOG_INFO, 427 "nd6_ra_input: lladdrlen mismatch for %s " 428 "(if %d, RA packet %d)\n", ip6_sprintf(ip6bufs, &saddr6), 429 ifp->if_addrlen, lladdrlen - 2)); 430 goto bad; 431 } 432 433 nd6_cache_lladdr(ifp, &saddr6, lladdr, 434 lladdrlen, ND_ROUTER_ADVERT, 0); 435 436 /* 437 * Installing a link-layer address might change the state of the 438 * router's neighbor cache, which might also affect our on-link 439 * detection of adveritsed prefixes. 440 */ 441 pfxlist_onlink_check(); 442 } 443 444 freeit: 445 m_freem(m); 446 return; 447 448 bad: 449 ICMP6STAT_INC(icp6s_badra); 450 m_freem(m); 451 } 452 453 /* tell the change to user processes watching the routing socket. */ 454 static void 455 nd6_rtmsg(int cmd, struct rtentry *rt) 456 { 457 struct rt_addrinfo info; 458 struct ifnet *ifp; 459 struct ifaddr *ifa; 460 461 bzero((caddr_t)&info, sizeof(info)); 462 info.rti_info[RTAX_DST] = rt_key(rt); 463 info.rti_info[RTAX_GATEWAY] = rt->rt_gateway; 464 info.rti_info[RTAX_NETMASK] = rt_mask(rt); 465 ifp = rt->rt_ifp; 466 if (ifp != NULL) { 467 IF_ADDR_RLOCK(ifp); 468 ifa = TAILQ_FIRST(&ifp->if_addrhead); 469 info.rti_info[RTAX_IFP] = ifa->ifa_addr; 470 ifa_ref(ifa); 471 IF_ADDR_RUNLOCK(ifp); 472 info.rti_info[RTAX_IFA] = rt->rt_ifa->ifa_addr; 473 } else 474 ifa = NULL; 475 476 rt_missmsg_fib(cmd, &info, rt->rt_flags, 0, rt->rt_fibnum); 477 if (ifa != NULL) 478 ifa_free(ifa); 479 } 480 481 /* 482 * default router list processing sub routines 483 */ 484 485 static void 486 defrouter_addreq(struct nd_defrouter *new) 487 { 488 struct sockaddr_in6 def, mask, gate; 489 struct rtentry *newrt = NULL; 490 int error; 491 492 bzero(&def, sizeof(def)); 493 bzero(&mask, sizeof(mask)); 494 bzero(&gate, sizeof(gate)); 495 496 def.sin6_len = mask.sin6_len = gate.sin6_len = 497 sizeof(struct sockaddr_in6); 498 def.sin6_family = gate.sin6_family = AF_INET6; 499 gate.sin6_addr = new->rtaddr; 500 501 error = in6_rtrequest(RTM_ADD, (struct sockaddr *)&def, 502 (struct sockaddr *)&gate, (struct sockaddr *)&mask, 503 RTF_GATEWAY, &newrt, new->ifp->if_fib); 504 if (newrt) { 505 nd6_rtmsg(RTM_ADD, newrt); /* tell user process */ 506 RTFREE(newrt); 507 } 508 if (error == 0) 509 new->installed = 1; 510 } 511 512 struct nd_defrouter * 513 defrouter_lookup_locked(struct in6_addr *addr, struct ifnet *ifp) 514 { 515 struct nd_defrouter *dr; 516 517 ND6_LOCK_ASSERT(); 518 TAILQ_FOREACH(dr, &V_nd_defrouter, dr_entry) 519 if (dr->ifp == ifp && IN6_ARE_ADDR_EQUAL(addr, &dr->rtaddr)) { 520 defrouter_ref(dr); 521 return (dr); 522 } 523 return (NULL); 524 } 525 526 struct nd_defrouter * 527 defrouter_lookup(struct in6_addr *addr, struct ifnet *ifp) 528 { 529 struct nd_defrouter *dr; 530 531 ND6_RLOCK(); 532 dr = defrouter_lookup_locked(addr, ifp); 533 ND6_RUNLOCK(); 534 return (dr); 535 } 536 537 void 538 defrouter_ref(struct nd_defrouter *dr) 539 { 540 541 refcount_acquire(&dr->refcnt); 542 } 543 544 void 545 defrouter_rele(struct nd_defrouter *dr) 546 { 547 548 if (refcount_release(&dr->refcnt)) 549 free(dr, M_IP6NDP); 550 } 551 552 /* 553 * Remove the default route for a given router. 554 * This is just a subroutine function for defrouter_select_fib(), and 555 * should not be called from anywhere else. 556 */ 557 static void 558 defrouter_delreq(struct nd_defrouter *dr) 559 { 560 struct sockaddr_in6 def, mask, gate; 561 struct rtentry *oldrt = NULL; 562 563 bzero(&def, sizeof(def)); 564 bzero(&mask, sizeof(mask)); 565 bzero(&gate, sizeof(gate)); 566 567 def.sin6_len = mask.sin6_len = gate.sin6_len = 568 sizeof(struct sockaddr_in6); 569 def.sin6_family = gate.sin6_family = AF_INET6; 570 gate.sin6_addr = dr->rtaddr; 571 572 in6_rtrequest(RTM_DELETE, (struct sockaddr *)&def, 573 (struct sockaddr *)&gate, 574 (struct sockaddr *)&mask, RTF_GATEWAY, &oldrt, dr->ifp->if_fib); 575 if (oldrt) { 576 nd6_rtmsg(RTM_DELETE, oldrt); 577 RTFREE(oldrt); 578 } 579 580 dr->installed = 0; 581 } 582 583 /* 584 * Remove all default routes from default router list. 585 */ 586 void 587 defrouter_reset(void) 588 { 589 struct nd_defrouter *dr, **dra; 590 int count, i; 591 592 count = i = 0; 593 594 /* 595 * We can't delete routes with the ND lock held, so make a copy of the 596 * current default router list and use that when deleting routes. 597 */ 598 ND6_RLOCK(); 599 TAILQ_FOREACH(dr, &V_nd_defrouter, dr_entry) 600 count++; 601 ND6_RUNLOCK(); 602 603 dra = malloc(count * sizeof(*dra), M_TEMP, M_WAITOK | M_ZERO); 604 605 ND6_RLOCK(); 606 TAILQ_FOREACH(dr, &V_nd_defrouter, dr_entry) { 607 if (i == count) 608 break; 609 defrouter_ref(dr); 610 dra[i++] = dr; 611 } 612 ND6_RUNLOCK(); 613 614 for (i = 0; i < count && dra[i] != NULL; i++) { 615 defrouter_delreq(dra[i]); 616 defrouter_rele(dra[i]); 617 } 618 free(dra, M_TEMP); 619 620 /* 621 * XXX should we also nuke any default routers in the kernel, by 622 * going through them by rtalloc1()? 623 */ 624 } 625 626 /* 627 * Look up a matching default router list entry and remove it. Returns true if a 628 * matching entry was found, false otherwise. 629 */ 630 bool 631 defrouter_remove(struct in6_addr *addr, struct ifnet *ifp) 632 { 633 struct nd_defrouter *dr; 634 635 ND6_WLOCK(); 636 dr = defrouter_lookup_locked(addr, ifp); 637 if (dr == NULL) { 638 ND6_WUNLOCK(); 639 return (false); 640 } 641 642 defrouter_unlink(dr, NULL); 643 ND6_WUNLOCK(); 644 defrouter_del(dr); 645 defrouter_rele(dr); 646 return (true); 647 } 648 649 /* 650 * Remove a router from the global list and optionally stash it in a 651 * caller-supplied queue. 652 * 653 * The ND lock must be held. 654 */ 655 void 656 defrouter_unlink(struct nd_defrouter *dr, struct nd_drhead *drq) 657 { 658 659 ND6_WLOCK_ASSERT(); 660 TAILQ_REMOVE(&V_nd_defrouter, dr, dr_entry); 661 V_nd6_list_genid++; 662 if (drq != NULL) 663 TAILQ_INSERT_TAIL(drq, dr, dr_entry); 664 } 665 666 void 667 defrouter_del(struct nd_defrouter *dr) 668 { 669 struct nd_defrouter *deldr = NULL; 670 struct nd_prefix *pr; 671 struct nd_pfxrouter *pfxrtr; 672 673 ND6_UNLOCK_ASSERT(); 674 675 /* 676 * Flush all the routing table entries that use the router 677 * as a next hop. 678 */ 679 if (ND_IFINFO(dr->ifp)->flags & ND6_IFF_ACCEPT_RTADV) 680 rt6_flush(&dr->rtaddr, dr->ifp); 681 682 if (dr->installed) { 683 deldr = dr; 684 defrouter_delreq(dr); 685 } 686 687 /* 688 * Also delete all the pointers to the router in each prefix lists. 689 */ 690 ND6_WLOCK(); 691 LIST_FOREACH(pr, &V_nd_prefix, ndpr_entry) { 692 if ((pfxrtr = pfxrtr_lookup(pr, dr)) != NULL) 693 pfxrtr_del(pfxrtr); 694 } 695 ND6_WUNLOCK(); 696 697 pfxlist_onlink_check(); 698 699 /* 700 * If the router is the primary one, choose a new one. 701 * Note that defrouter_select_fib() will remove the current 702 * gateway from the routing table. 703 */ 704 if (deldr) 705 defrouter_select_fib(deldr->ifp->if_fib); 706 707 /* 708 * Release the list reference. 709 */ 710 defrouter_rele(dr); 711 } 712 713 /* 714 * Default Router Selection according to Section 6.3.6 of RFC 2461 and 715 * draft-ietf-ipngwg-router-selection: 716 * 1) Routers that are reachable or probably reachable should be preferred. 717 * If we have more than one (probably) reachable router, prefer ones 718 * with the highest router preference. 719 * 2) When no routers on the list are known to be reachable or 720 * probably reachable, routers SHOULD be selected in a round-robin 721 * fashion, regardless of router preference values. 722 * 3) If the Default Router List is empty, assume that all 723 * destinations are on-link. 724 * 725 * We assume nd_defrouter is sorted by router preference value. 726 * Since the code below covers both with and without router preference cases, 727 * we do not need to classify the cases by ifdef. 728 * 729 * At this moment, we do not try to install more than one default router, 730 * even when the multipath routing is available, because we're not sure about 731 * the benefits for stub hosts comparing to the risk of making the code 732 * complicated and the possibility of introducing bugs. 733 * 734 * We maintain a single list of routers for multiple FIBs, only considering one 735 * at a time based on the receiving interface's FIB. If @fibnum is RT_ALL_FIBS, 736 * we do the whole thing multiple times. 737 */ 738 void 739 defrouter_select_fib(int fibnum) 740 { 741 struct nd_defrouter *dr, *selected_dr, *installed_dr; 742 struct llentry *ln = NULL; 743 744 if (fibnum == RT_ALL_FIBS) { 745 for (fibnum = 0; fibnum < rt_numfibs; fibnum++) { 746 defrouter_select_fib(fibnum); 747 } 748 } 749 750 ND6_RLOCK(); 751 /* 752 * Let's handle easy case (3) first: 753 * If default router list is empty, there's nothing to be done. 754 */ 755 if (TAILQ_EMPTY(&V_nd_defrouter)) { 756 ND6_RUNLOCK(); 757 return; 758 } 759 760 /* 761 * Search for a (probably) reachable router from the list. 762 * We just pick up the first reachable one (if any), assuming that 763 * the ordering rule of the list described in defrtrlist_update(). 764 */ 765 selected_dr = installed_dr = NULL; 766 TAILQ_FOREACH(dr, &V_nd_defrouter, dr_entry) { 767 IF_AFDATA_RLOCK(dr->ifp); 768 if (selected_dr == NULL && dr->ifp->if_fib == fibnum && 769 (ln = nd6_lookup(&dr->rtaddr, 0, dr->ifp)) && 770 ND6_IS_LLINFO_PROBREACH(ln)) { 771 selected_dr = dr; 772 defrouter_ref(selected_dr); 773 } 774 IF_AFDATA_RUNLOCK(dr->ifp); 775 if (ln != NULL) { 776 LLE_RUNLOCK(ln); 777 ln = NULL; 778 } 779 780 if (dr->installed && dr->ifp->if_fib == fibnum) { 781 if (installed_dr == NULL) { 782 installed_dr = dr; 783 defrouter_ref(installed_dr); 784 } else { 785 /* 786 * this should not happen. 787 * warn for diagnosis. 788 */ 789 log(LOG_ERR, "defrouter_select_fib: more than " 790 "one router is installed\n"); 791 } 792 } 793 } 794 /* 795 * If none of the default routers was found to be reachable, 796 * round-robin the list regardless of preference. 797 * Otherwise, if we have an installed router, check if the selected 798 * (reachable) router should really be preferred to the installed one. 799 * We only prefer the new router when the old one is not reachable 800 * or when the new one has a really higher preference value. 801 */ 802 if (selected_dr == NULL) { 803 if (installed_dr == NULL || 804 TAILQ_NEXT(installed_dr, dr_entry) == NULL) 805 dr = TAILQ_FIRST(&V_nd_defrouter); 806 else 807 dr = TAILQ_NEXT(installed_dr, dr_entry); 808 809 /* Ensure we select a router for this FIB. */ 810 TAILQ_FOREACH_FROM(dr, &V_nd_defrouter, dr_entry) { 811 if (dr->ifp->if_fib == fibnum) { 812 selected_dr = dr; 813 defrouter_ref(selected_dr); 814 break; 815 } 816 } 817 } else if (installed_dr != NULL) { 818 IF_AFDATA_RLOCK(installed_dr->ifp); 819 if ((ln = nd6_lookup(&installed_dr->rtaddr, 0, 820 installed_dr->ifp)) && 821 ND6_IS_LLINFO_PROBREACH(ln) && 822 installed_dr->ifp->if_fib == fibnum && 823 rtpref(selected_dr) <= rtpref(installed_dr)) { 824 defrouter_rele(selected_dr); 825 selected_dr = installed_dr; 826 } 827 IF_AFDATA_RUNLOCK(installed_dr->ifp); 828 if (ln != NULL) 829 LLE_RUNLOCK(ln); 830 } 831 ND6_RUNLOCK(); 832 833 /* 834 * If we selected a router for this FIB and it's different 835 * than the installed one, remove the installed router and 836 * install the selected one in its place. 837 */ 838 if (installed_dr != selected_dr) { 839 if (installed_dr != NULL) { 840 defrouter_delreq(installed_dr); 841 defrouter_rele(installed_dr); 842 } 843 if (selected_dr != NULL) 844 defrouter_addreq(selected_dr); 845 } 846 if (selected_dr != NULL) 847 defrouter_rele(selected_dr); 848 } 849 850 /* 851 * Maintain old KPI for default router selection. 852 * If unspecified, we can re-select routers for all FIBs. 853 */ 854 void 855 defrouter_select(void) 856 { 857 defrouter_select_fib(RT_ALL_FIBS); 858 } 859 860 /* 861 * for default router selection 862 * regards router-preference field as a 2-bit signed integer 863 */ 864 static int 865 rtpref(struct nd_defrouter *dr) 866 { 867 switch (dr->raflags & ND_RA_FLAG_RTPREF_MASK) { 868 case ND_RA_FLAG_RTPREF_HIGH: 869 return (RTPREF_HIGH); 870 case ND_RA_FLAG_RTPREF_MEDIUM: 871 case ND_RA_FLAG_RTPREF_RSV: 872 return (RTPREF_MEDIUM); 873 case ND_RA_FLAG_RTPREF_LOW: 874 return (RTPREF_LOW); 875 default: 876 /* 877 * This case should never happen. If it did, it would mean a 878 * serious bug of kernel internal. We thus always bark here. 879 * Or, can we even panic? 880 */ 881 log(LOG_ERR, "rtpref: impossible RA flag %x\n", dr->raflags); 882 return (RTPREF_INVALID); 883 } 884 /* NOTREACHED */ 885 } 886 887 static struct nd_defrouter * 888 defrtrlist_update(struct nd_defrouter *new) 889 { 890 struct nd_defrouter *dr, *n; 891 uint64_t genid; 892 int oldpref; 893 bool writelocked; 894 895 if (new->rtlifetime == 0) { 896 defrouter_remove(&new->rtaddr, new->ifp); 897 return (NULL); 898 } 899 900 ND6_RLOCK(); 901 writelocked = false; 902 restart: 903 dr = defrouter_lookup_locked(&new->rtaddr, new->ifp); 904 if (dr != NULL) { 905 oldpref = rtpref(dr); 906 907 /* override */ 908 dr->raflags = new->raflags; /* XXX flag check */ 909 dr->rtlifetime = new->rtlifetime; 910 dr->expire = new->expire; 911 912 /* 913 * If the preference does not change, there's no need 914 * to sort the entries. Also make sure the selected 915 * router is still installed in the kernel. 916 */ 917 if (dr->installed && rtpref(new) == oldpref) { 918 if (writelocked) 919 ND6_WUNLOCK(); 920 else 921 ND6_RUNLOCK(); 922 return (dr); 923 } 924 } 925 926 /* 927 * The router needs to be reinserted into the default router 928 * list, so upgrade to a write lock. If that fails and the list 929 * has potentially changed while the lock was dropped, we'll 930 * redo the lookup with the write lock held. 931 */ 932 if (!writelocked) { 933 writelocked = true; 934 if (!ND6_TRY_UPGRADE()) { 935 genid = V_nd6_list_genid; 936 ND6_RUNLOCK(); 937 ND6_WLOCK(); 938 if (genid != V_nd6_list_genid) 939 goto restart; 940 } 941 } 942 943 if (dr != NULL) { 944 /* 945 * The preferred router may have changed, so relocate this 946 * router. 947 */ 948 TAILQ_REMOVE(&V_nd_defrouter, dr, dr_entry); 949 n = dr; 950 } else { 951 n = malloc(sizeof(*n), M_IP6NDP, M_NOWAIT | M_ZERO); 952 if (n == NULL) { 953 ND6_WUNLOCK(); 954 return (NULL); 955 } 956 memcpy(n, new, sizeof(*n)); 957 /* Initialize with an extra reference for the caller. */ 958 refcount_init(&n->refcnt, 2); 959 } 960 961 /* 962 * Insert the new router in the Default Router List; 963 * The Default Router List should be in the descending order 964 * of router-preferece. Routers with the same preference are 965 * sorted in the arriving time order. 966 */ 967 968 /* insert at the end of the group */ 969 TAILQ_FOREACH(dr, &V_nd_defrouter, dr_entry) { 970 if (rtpref(n) > rtpref(dr)) 971 break; 972 } 973 if (dr != NULL) 974 TAILQ_INSERT_BEFORE(dr, n, dr_entry); 975 else 976 TAILQ_INSERT_TAIL(&V_nd_defrouter, n, dr_entry); 977 V_nd6_list_genid++; 978 ND6_WUNLOCK(); 979 980 defrouter_select_fib(new->ifp->if_fib); 981 982 return (n); 983 } 984 985 static struct nd_pfxrouter * 986 pfxrtr_lookup(struct nd_prefix *pr, struct nd_defrouter *dr) 987 { 988 struct nd_pfxrouter *search; 989 990 ND6_LOCK_ASSERT(); 991 992 LIST_FOREACH(search, &pr->ndpr_advrtrs, pfr_entry) { 993 if (search->router == dr) 994 break; 995 } 996 return (search); 997 } 998 999 static void 1000 pfxrtr_add(struct nd_prefix *pr, struct nd_defrouter *dr) 1001 { 1002 struct nd_pfxrouter *new; 1003 bool update; 1004 1005 ND6_UNLOCK_ASSERT(); 1006 1007 ND6_RLOCK(); 1008 if (pfxrtr_lookup(pr, dr) != NULL) { 1009 ND6_RUNLOCK(); 1010 return; 1011 } 1012 ND6_RUNLOCK(); 1013 1014 new = malloc(sizeof(*new), M_IP6NDP, M_NOWAIT | M_ZERO); 1015 if (new == NULL) 1016 return; 1017 defrouter_ref(dr); 1018 new->router = dr; 1019 1020 ND6_WLOCK(); 1021 if (pfxrtr_lookup(pr, dr) == NULL) { 1022 LIST_INSERT_HEAD(&pr->ndpr_advrtrs, new, pfr_entry); 1023 update = true; 1024 } else { 1025 /* We lost a race to add the reference. */ 1026 defrouter_rele(dr); 1027 free(new, M_IP6NDP); 1028 update = false; 1029 } 1030 ND6_WUNLOCK(); 1031 1032 if (update) 1033 pfxlist_onlink_check(); 1034 } 1035 1036 static void 1037 pfxrtr_del(struct nd_pfxrouter *pfr) 1038 { 1039 1040 ND6_WLOCK_ASSERT(); 1041 1042 LIST_REMOVE(pfr, pfr_entry); 1043 defrouter_rele(pfr->router); 1044 free(pfr, M_IP6NDP); 1045 } 1046 1047 static struct nd_prefix * 1048 nd6_prefix_lookup_locked(struct nd_prefixctl *key) 1049 { 1050 struct nd_prefix *search; 1051 1052 ND6_LOCK_ASSERT(); 1053 1054 LIST_FOREACH(search, &V_nd_prefix, ndpr_entry) { 1055 if (key->ndpr_ifp == search->ndpr_ifp && 1056 key->ndpr_plen == search->ndpr_plen && 1057 in6_are_prefix_equal(&key->ndpr_prefix.sin6_addr, 1058 &search->ndpr_prefix.sin6_addr, key->ndpr_plen)) { 1059 nd6_prefix_ref(search); 1060 break; 1061 } 1062 } 1063 return (search); 1064 } 1065 1066 struct nd_prefix * 1067 nd6_prefix_lookup(struct nd_prefixctl *key) 1068 { 1069 struct nd_prefix *search; 1070 1071 ND6_RLOCK(); 1072 search = nd6_prefix_lookup_locked(key); 1073 ND6_RUNLOCK(); 1074 return (search); 1075 } 1076 1077 void 1078 nd6_prefix_ref(struct nd_prefix *pr) 1079 { 1080 1081 refcount_acquire(&pr->ndpr_refcnt); 1082 } 1083 1084 void 1085 nd6_prefix_rele(struct nd_prefix *pr) 1086 { 1087 1088 if (refcount_release(&pr->ndpr_refcnt)) { 1089 KASSERT(LIST_EMPTY(&pr->ndpr_advrtrs), 1090 ("prefix %p has advertising routers", pr)); 1091 free(pr, M_IP6NDP); 1092 } 1093 } 1094 1095 int 1096 nd6_prelist_add(struct nd_prefixctl *pr, struct nd_defrouter *dr, 1097 struct nd_prefix **newp) 1098 { 1099 struct nd_prefix *new; 1100 char ip6buf[INET6_ADDRSTRLEN]; 1101 int error; 1102 1103 new = malloc(sizeof(*new), M_IP6NDP, M_NOWAIT | M_ZERO); 1104 if (new == NULL) 1105 return (ENOMEM); 1106 refcount_init(&new->ndpr_refcnt, newp != NULL ? 2 : 1); 1107 new->ndpr_ifp = pr->ndpr_ifp; 1108 new->ndpr_prefix = pr->ndpr_prefix; 1109 new->ndpr_plen = pr->ndpr_plen; 1110 new->ndpr_vltime = pr->ndpr_vltime; 1111 new->ndpr_pltime = pr->ndpr_pltime; 1112 new->ndpr_flags = pr->ndpr_flags; 1113 if ((error = in6_init_prefix_ltimes(new)) != 0) { 1114 free(new, M_IP6NDP); 1115 return (error); 1116 } 1117 new->ndpr_lastupdate = time_uptime; 1118 1119 /* initialization */ 1120 LIST_INIT(&new->ndpr_advrtrs); 1121 in6_prefixlen2mask(&new->ndpr_mask, new->ndpr_plen); 1122 /* make prefix in the canonical form */ 1123 IN6_MASK_ADDR(&new->ndpr_prefix.sin6_addr, &new->ndpr_mask); 1124 1125 ND6_WLOCK(); 1126 LIST_INSERT_HEAD(&V_nd_prefix, new, ndpr_entry); 1127 V_nd6_list_genid++; 1128 ND6_WUNLOCK(); 1129 1130 /* ND_OPT_PI_FLAG_ONLINK processing */ 1131 if (new->ndpr_raf_onlink) { 1132 ND6_ONLINK_LOCK(); 1133 if ((error = nd6_prefix_onlink(new)) != 0) { 1134 nd6log((LOG_ERR, "nd6_prelist_add: failed to make " 1135 "the prefix %s/%d on-link on %s (errno=%d)\n", 1136 ip6_sprintf(ip6buf, &pr->ndpr_prefix.sin6_addr), 1137 pr->ndpr_plen, if_name(pr->ndpr_ifp), error)); 1138 /* proceed anyway. XXX: is it correct? */ 1139 } 1140 ND6_ONLINK_UNLOCK(); 1141 } 1142 1143 if (dr != NULL) 1144 pfxrtr_add(new, dr); 1145 if (newp != NULL) 1146 *newp = new; 1147 return (0); 1148 } 1149 1150 /* 1151 * Remove a prefix from the prefix list and optionally stash it in a 1152 * caller-provided list. 1153 * 1154 * The ND6 lock must be held. 1155 */ 1156 void 1157 nd6_prefix_unlink(struct nd_prefix *pr, struct nd_prhead *list) 1158 { 1159 1160 ND6_WLOCK_ASSERT(); 1161 1162 LIST_REMOVE(pr, ndpr_entry); 1163 V_nd6_list_genid++; 1164 if (list != NULL) 1165 LIST_INSERT_HEAD(list, pr, ndpr_entry); 1166 } 1167 1168 /* 1169 * Free an unlinked prefix, first marking it off-link if necessary. 1170 */ 1171 void 1172 nd6_prefix_del(struct nd_prefix *pr) 1173 { 1174 struct nd_pfxrouter *pfr, *next; 1175 int e; 1176 char ip6buf[INET6_ADDRSTRLEN]; 1177 1178 KASSERT(pr->ndpr_addrcnt == 0, 1179 ("prefix %p has referencing addresses", pr)); 1180 ND6_UNLOCK_ASSERT(); 1181 1182 /* 1183 * Though these flags are now meaningless, we'd rather keep the value 1184 * of pr->ndpr_raf_onlink and pr->ndpr_raf_auto not to confuse users 1185 * when executing "ndp -p". 1186 */ 1187 if ((pr->ndpr_stateflags & NDPRF_ONLINK) != 0) { 1188 ND6_ONLINK_LOCK(); 1189 if ((e = nd6_prefix_offlink(pr)) != 0) { 1190 nd6log((LOG_ERR, 1191 "nd6_prefix_del: failed to make %s/%d offlink " 1192 "on %s, errno=%d\n", 1193 ip6_sprintf(ip6buf, &pr->ndpr_prefix.sin6_addr), 1194 pr->ndpr_plen, if_name(pr->ndpr_ifp), e)); 1195 /* what should we do? */ 1196 } 1197 ND6_ONLINK_UNLOCK(); 1198 } 1199 1200 /* Release references to routers that have advertised this prefix. */ 1201 ND6_WLOCK(); 1202 LIST_FOREACH_SAFE(pfr, &pr->ndpr_advrtrs, pfr_entry, next) 1203 pfxrtr_del(pfr); 1204 ND6_WUNLOCK(); 1205 1206 nd6_prefix_rele(pr); 1207 1208 pfxlist_onlink_check(); 1209 } 1210 1211 static int 1212 prelist_update(struct nd_prefixctl *new, struct nd_defrouter *dr, 1213 struct mbuf *m, int mcast) 1214 { 1215 struct in6_ifaddr *ia6 = NULL, *ia6_match = NULL; 1216 struct ifaddr *ifa; 1217 struct ifnet *ifp = new->ndpr_ifp; 1218 struct nd_prefix *pr; 1219 int error = 0; 1220 int auth; 1221 struct in6_addrlifetime lt6_tmp; 1222 char ip6buf[INET6_ADDRSTRLEN]; 1223 1224 auth = 0; 1225 if (m) { 1226 /* 1227 * Authenticity for NA consists authentication for 1228 * both IP header and IP datagrams, doesn't it ? 1229 */ 1230 #if defined(M_AUTHIPHDR) && defined(M_AUTHIPDGM) 1231 auth = ((m->m_flags & M_AUTHIPHDR) && 1232 (m->m_flags & M_AUTHIPDGM)); 1233 #endif 1234 } 1235 1236 if ((pr = nd6_prefix_lookup(new)) != NULL) { 1237 /* 1238 * nd6_prefix_lookup() ensures that pr and new have the same 1239 * prefix on a same interface. 1240 */ 1241 1242 /* 1243 * Update prefix information. Note that the on-link (L) bit 1244 * and the autonomous (A) bit should NOT be changed from 1 1245 * to 0. 1246 */ 1247 if (new->ndpr_raf_onlink == 1) 1248 pr->ndpr_raf_onlink = 1; 1249 if (new->ndpr_raf_auto == 1) 1250 pr->ndpr_raf_auto = 1; 1251 if (new->ndpr_raf_onlink) { 1252 pr->ndpr_vltime = new->ndpr_vltime; 1253 pr->ndpr_pltime = new->ndpr_pltime; 1254 (void)in6_init_prefix_ltimes(pr); /* XXX error case? */ 1255 pr->ndpr_lastupdate = time_uptime; 1256 } 1257 1258 if (new->ndpr_raf_onlink && 1259 (pr->ndpr_stateflags & NDPRF_ONLINK) == 0) { 1260 ND6_ONLINK_LOCK(); 1261 if ((error = nd6_prefix_onlink(pr)) != 0) { 1262 nd6log((LOG_ERR, 1263 "prelist_update: failed to make " 1264 "the prefix %s/%d on-link on %s " 1265 "(errno=%d)\n", 1266 ip6_sprintf(ip6buf, 1267 &pr->ndpr_prefix.sin6_addr), 1268 pr->ndpr_plen, if_name(pr->ndpr_ifp), 1269 error)); 1270 /* proceed anyway. XXX: is it correct? */ 1271 } 1272 ND6_ONLINK_UNLOCK(); 1273 } 1274 1275 if (dr != NULL) 1276 pfxrtr_add(pr, dr); 1277 } else { 1278 if (new->ndpr_vltime == 0) 1279 goto end; 1280 if (new->ndpr_raf_onlink == 0 && new->ndpr_raf_auto == 0) 1281 goto end; 1282 1283 error = nd6_prelist_add(new, dr, &pr); 1284 if (error != 0) { 1285 nd6log((LOG_NOTICE, "prelist_update: " 1286 "nd6_prelist_add failed for %s/%d on %s errno=%d\n", 1287 ip6_sprintf(ip6buf, &new->ndpr_prefix.sin6_addr), 1288 new->ndpr_plen, if_name(new->ndpr_ifp), error)); 1289 goto end; /* we should just give up in this case. */ 1290 } 1291 1292 /* 1293 * XXX: from the ND point of view, we can ignore a prefix 1294 * with the on-link bit being zero. However, we need a 1295 * prefix structure for references from autoconfigured 1296 * addresses. Thus, we explicitly make sure that the prefix 1297 * itself expires now. 1298 */ 1299 if (pr->ndpr_raf_onlink == 0) { 1300 pr->ndpr_vltime = 0; 1301 pr->ndpr_pltime = 0; 1302 in6_init_prefix_ltimes(pr); 1303 } 1304 } 1305 1306 /* 1307 * Address autoconfiguration based on Section 5.5.3 of RFC 2462. 1308 * Note that pr must be non NULL at this point. 1309 */ 1310 1311 /* 5.5.3 (a). Ignore the prefix without the A bit set. */ 1312 if (!new->ndpr_raf_auto) 1313 goto end; 1314 1315 /* 1316 * 5.5.3 (b). the link-local prefix should have been ignored in 1317 * nd6_ra_input. 1318 */ 1319 1320 /* 5.5.3 (c). Consistency check on lifetimes: pltime <= vltime. */ 1321 if (new->ndpr_pltime > new->ndpr_vltime) { 1322 error = EINVAL; /* XXX: won't be used */ 1323 goto end; 1324 } 1325 1326 /* 1327 * 5.5.3 (d). If the prefix advertised is not equal to the prefix of 1328 * an address configured by stateless autoconfiguration already in the 1329 * list of addresses associated with the interface, and the Valid 1330 * Lifetime is not 0, form an address. We first check if we have 1331 * a matching prefix. 1332 * Note: we apply a clarification in rfc2462bis-02 here. We only 1333 * consider autoconfigured addresses while RFC2462 simply said 1334 * "address". 1335 */ 1336 IF_ADDR_RLOCK(ifp); 1337 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1338 struct in6_ifaddr *ifa6; 1339 u_int32_t remaininglifetime; 1340 1341 if (ifa->ifa_addr->sa_family != AF_INET6) 1342 continue; 1343 1344 ifa6 = (struct in6_ifaddr *)ifa; 1345 1346 /* 1347 * We only consider autoconfigured addresses as per rfc2462bis. 1348 */ 1349 if (!(ifa6->ia6_flags & IN6_IFF_AUTOCONF)) 1350 continue; 1351 1352 /* 1353 * Spec is not clear here, but I believe we should concentrate 1354 * on unicast (i.e. not anycast) addresses. 1355 * XXX: other ia6_flags? detached or duplicated? 1356 */ 1357 if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0) 1358 continue; 1359 1360 /* 1361 * Ignore the address if it is not associated with a prefix 1362 * or is associated with a prefix that is different from this 1363 * one. (pr is never NULL here) 1364 */ 1365 if (ifa6->ia6_ndpr != pr) 1366 continue; 1367 1368 if (ia6_match == NULL) /* remember the first one */ 1369 ia6_match = ifa6; 1370 1371 /* 1372 * An already autoconfigured address matched. Now that we 1373 * are sure there is at least one matched address, we can 1374 * proceed to 5.5.3. (e): update the lifetimes according to the 1375 * "two hours" rule and the privacy extension. 1376 * We apply some clarifications in rfc2462bis: 1377 * - use remaininglifetime instead of storedlifetime as a 1378 * variable name 1379 * - remove the dead code in the "two-hour" rule 1380 */ 1381 #define TWOHOUR (120*60) 1382 lt6_tmp = ifa6->ia6_lifetime; 1383 1384 if (lt6_tmp.ia6t_vltime == ND6_INFINITE_LIFETIME) 1385 remaininglifetime = ND6_INFINITE_LIFETIME; 1386 else if (time_uptime - ifa6->ia6_updatetime > 1387 lt6_tmp.ia6t_vltime) { 1388 /* 1389 * The case of "invalid" address. We should usually 1390 * not see this case. 1391 */ 1392 remaininglifetime = 0; 1393 } else 1394 remaininglifetime = lt6_tmp.ia6t_vltime - 1395 (time_uptime - ifa6->ia6_updatetime); 1396 1397 /* when not updating, keep the current stored lifetime. */ 1398 lt6_tmp.ia6t_vltime = remaininglifetime; 1399 1400 if (TWOHOUR < new->ndpr_vltime || 1401 remaininglifetime < new->ndpr_vltime) { 1402 lt6_tmp.ia6t_vltime = new->ndpr_vltime; 1403 } else if (remaininglifetime <= TWOHOUR) { 1404 if (auth) { 1405 lt6_tmp.ia6t_vltime = new->ndpr_vltime; 1406 } 1407 } else { 1408 /* 1409 * new->ndpr_vltime <= TWOHOUR && 1410 * TWOHOUR < remaininglifetime 1411 */ 1412 lt6_tmp.ia6t_vltime = TWOHOUR; 1413 } 1414 1415 /* The 2 hour rule is not imposed for preferred lifetime. */ 1416 lt6_tmp.ia6t_pltime = new->ndpr_pltime; 1417 1418 in6_init_address_ltimes(pr, <6_tmp); 1419 1420 /* 1421 * We need to treat lifetimes for temporary addresses 1422 * differently, according to 1423 * draft-ietf-ipv6-privacy-addrs-v2-01.txt 3.3 (1); 1424 * we only update the lifetimes when they are in the maximum 1425 * intervals. 1426 */ 1427 if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) != 0) { 1428 u_int32_t maxvltime, maxpltime; 1429 1430 if (V_ip6_temp_valid_lifetime > 1431 (u_int32_t)((time_uptime - ifa6->ia6_createtime) + 1432 V_ip6_desync_factor)) { 1433 maxvltime = V_ip6_temp_valid_lifetime - 1434 (time_uptime - ifa6->ia6_createtime) - 1435 V_ip6_desync_factor; 1436 } else 1437 maxvltime = 0; 1438 if (V_ip6_temp_preferred_lifetime > 1439 (u_int32_t)((time_uptime - ifa6->ia6_createtime) + 1440 V_ip6_desync_factor)) { 1441 maxpltime = V_ip6_temp_preferred_lifetime - 1442 (time_uptime - ifa6->ia6_createtime) - 1443 V_ip6_desync_factor; 1444 } else 1445 maxpltime = 0; 1446 1447 if (lt6_tmp.ia6t_vltime == ND6_INFINITE_LIFETIME || 1448 lt6_tmp.ia6t_vltime > maxvltime) { 1449 lt6_tmp.ia6t_vltime = maxvltime; 1450 } 1451 if (lt6_tmp.ia6t_pltime == ND6_INFINITE_LIFETIME || 1452 lt6_tmp.ia6t_pltime > maxpltime) { 1453 lt6_tmp.ia6t_pltime = maxpltime; 1454 } 1455 } 1456 ifa6->ia6_lifetime = lt6_tmp; 1457 ifa6->ia6_updatetime = time_uptime; 1458 } 1459 IF_ADDR_RUNLOCK(ifp); 1460 if (ia6_match == NULL && new->ndpr_vltime) { 1461 int ifidlen; 1462 1463 /* 1464 * 5.5.3 (d) (continued) 1465 * No address matched and the valid lifetime is non-zero. 1466 * Create a new address. 1467 */ 1468 1469 /* 1470 * Prefix Length check: 1471 * If the sum of the prefix length and interface identifier 1472 * length does not equal 128 bits, the Prefix Information 1473 * option MUST be ignored. The length of the interface 1474 * identifier is defined in a separate link-type specific 1475 * document. 1476 */ 1477 ifidlen = in6_if2idlen(ifp); 1478 if (ifidlen < 0) { 1479 /* this should not happen, so we always log it. */ 1480 log(LOG_ERR, "prelist_update: IFID undefined (%s)\n", 1481 if_name(ifp)); 1482 goto end; 1483 } 1484 if (ifidlen + pr->ndpr_plen != 128) { 1485 nd6log((LOG_INFO, 1486 "prelist_update: invalid prefixlen " 1487 "%d for %s, ignored\n", 1488 pr->ndpr_plen, if_name(ifp))); 1489 goto end; 1490 } 1491 1492 if ((ia6 = in6_ifadd(new, mcast)) != NULL) { 1493 /* 1494 * note that we should use pr (not new) for reference. 1495 */ 1496 pr->ndpr_addrcnt++; 1497 ia6->ia6_ndpr = pr; 1498 1499 /* 1500 * RFC 3041 3.3 (2). 1501 * When a new public address is created as described 1502 * in RFC2462, also create a new temporary address. 1503 * 1504 * RFC 3041 3.5. 1505 * When an interface connects to a new link, a new 1506 * randomized interface identifier should be generated 1507 * immediately together with a new set of temporary 1508 * addresses. Thus, we specifiy 1 as the 2nd arg of 1509 * in6_tmpifadd(). 1510 */ 1511 if (V_ip6_use_tempaddr) { 1512 int e; 1513 if ((e = in6_tmpifadd(ia6, 1, 1)) != 0) { 1514 nd6log((LOG_NOTICE, "prelist_update: " 1515 "failed to create a temporary " 1516 "address, errno=%d\n", 1517 e)); 1518 } 1519 } 1520 ifa_free(&ia6->ia_ifa); 1521 1522 /* 1523 * A newly added address might affect the status 1524 * of other addresses, so we check and update it. 1525 * XXX: what if address duplication happens? 1526 */ 1527 pfxlist_onlink_check(); 1528 } else { 1529 /* just set an error. do not bark here. */ 1530 error = EADDRNOTAVAIL; /* XXX: might be unused. */ 1531 } 1532 } 1533 1534 end: 1535 if (pr != NULL) 1536 nd6_prefix_rele(pr); 1537 return (error); 1538 } 1539 1540 /* 1541 * A supplement function used in the on-link detection below; 1542 * detect if a given prefix has a (probably) reachable advertising router. 1543 * XXX: lengthy function name... 1544 */ 1545 static struct nd_pfxrouter * 1546 find_pfxlist_reachable_router(struct nd_prefix *pr) 1547 { 1548 struct nd_pfxrouter *pfxrtr; 1549 struct llentry *ln; 1550 int canreach; 1551 1552 ND6_LOCK_ASSERT(); 1553 1554 LIST_FOREACH(pfxrtr, &pr->ndpr_advrtrs, pfr_entry) { 1555 IF_AFDATA_RLOCK(pfxrtr->router->ifp); 1556 ln = nd6_lookup(&pfxrtr->router->rtaddr, 0, pfxrtr->router->ifp); 1557 IF_AFDATA_RUNLOCK(pfxrtr->router->ifp); 1558 if (ln == NULL) 1559 continue; 1560 canreach = ND6_IS_LLINFO_PROBREACH(ln); 1561 LLE_RUNLOCK(ln); 1562 if (canreach) 1563 break; 1564 } 1565 return (pfxrtr); 1566 } 1567 1568 /* 1569 * Check if each prefix in the prefix list has at least one available router 1570 * that advertised the prefix (a router is "available" if its neighbor cache 1571 * entry is reachable or probably reachable). 1572 * If the check fails, the prefix may be off-link, because, for example, 1573 * we have moved from the network but the lifetime of the prefix has not 1574 * expired yet. So we should not use the prefix if there is another prefix 1575 * that has an available router. 1576 * But, if there is no prefix that has an available router, we still regard 1577 * all the prefixes as on-link. This is because we can't tell if all the 1578 * routers are simply dead or if we really moved from the network and there 1579 * is no router around us. 1580 */ 1581 void 1582 pfxlist_onlink_check(void) 1583 { 1584 struct nd_prefix *pr; 1585 struct in6_ifaddr *ifa; 1586 struct nd_defrouter *dr; 1587 struct nd_pfxrouter *pfxrtr = NULL; 1588 struct rm_priotracker in6_ifa_tracker; 1589 uint64_t genid; 1590 uint32_t flags; 1591 1592 ND6_ONLINK_LOCK(); 1593 ND6_RLOCK(); 1594 1595 /* 1596 * Check if there is a prefix that has a reachable advertising 1597 * router. 1598 */ 1599 LIST_FOREACH(pr, &V_nd_prefix, ndpr_entry) { 1600 if (pr->ndpr_raf_onlink && find_pfxlist_reachable_router(pr)) 1601 break; 1602 } 1603 1604 /* 1605 * If we have no such prefix, check whether we still have a router 1606 * that does not advertise any prefixes. 1607 */ 1608 if (pr == NULL) { 1609 TAILQ_FOREACH(dr, &V_nd_defrouter, dr_entry) { 1610 struct nd_prefix *pr0; 1611 1612 LIST_FOREACH(pr0, &V_nd_prefix, ndpr_entry) { 1613 if ((pfxrtr = pfxrtr_lookup(pr0, dr)) != NULL) 1614 break; 1615 } 1616 if (pfxrtr != NULL) 1617 break; 1618 } 1619 } 1620 if (pr != NULL || (!TAILQ_EMPTY(&V_nd_defrouter) && pfxrtr == NULL)) { 1621 /* 1622 * There is at least one prefix that has a reachable router, 1623 * or at least a router which probably does not advertise 1624 * any prefixes. The latter would be the case when we move 1625 * to a new link where we have a router that does not provide 1626 * prefixes and we configure an address by hand. 1627 * Detach prefixes which have no reachable advertising 1628 * router, and attach other prefixes. 1629 */ 1630 LIST_FOREACH(pr, &V_nd_prefix, ndpr_entry) { 1631 /* XXX: a link-local prefix should never be detached */ 1632 if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr) || 1633 pr->ndpr_raf_onlink == 0 || 1634 pr->ndpr_raf_auto == 0) 1635 continue; 1636 1637 if ((pr->ndpr_stateflags & NDPRF_DETACHED) == 0 && 1638 find_pfxlist_reachable_router(pr) == NULL) 1639 pr->ndpr_stateflags |= NDPRF_DETACHED; 1640 else if ((pr->ndpr_stateflags & NDPRF_DETACHED) != 0 && 1641 find_pfxlist_reachable_router(pr) != NULL) 1642 pr->ndpr_stateflags &= ~NDPRF_DETACHED; 1643 } 1644 } else { 1645 /* there is no prefix that has a reachable router */ 1646 LIST_FOREACH(pr, &V_nd_prefix, ndpr_entry) { 1647 if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr) || 1648 pr->ndpr_raf_onlink == 0 || 1649 pr->ndpr_raf_auto == 0) 1650 continue; 1651 pr->ndpr_stateflags &= ~NDPRF_DETACHED; 1652 } 1653 } 1654 1655 /* 1656 * Remove each interface route associated with a (just) detached 1657 * prefix, and reinstall the interface route for a (just) attached 1658 * prefix. Note that all attempt of reinstallation does not 1659 * necessarily success, when a same prefix is shared among multiple 1660 * interfaces. Such cases will be handled in nd6_prefix_onlink, 1661 * so we don't have to care about them. 1662 */ 1663 restart: 1664 LIST_FOREACH(pr, &V_nd_prefix, ndpr_entry) { 1665 char ip6buf[INET6_ADDRSTRLEN]; 1666 int e; 1667 1668 if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr) || 1669 pr->ndpr_raf_onlink == 0 || 1670 pr->ndpr_raf_auto == 0) 1671 continue; 1672 1673 flags = pr->ndpr_stateflags & (NDPRF_DETACHED | NDPRF_ONLINK); 1674 if (flags == 0 || flags == (NDPRF_DETACHED | NDPRF_ONLINK)) { 1675 genid = V_nd6_list_genid; 1676 ND6_RUNLOCK(); 1677 if ((flags & NDPRF_ONLINK) != 0 && 1678 (e = nd6_prefix_offlink(pr)) != 0) { 1679 nd6log((LOG_ERR, 1680 "pfxlist_onlink_check: failed to " 1681 "make %s/%d offlink, errno=%d\n", 1682 ip6_sprintf(ip6buf, 1683 &pr->ndpr_prefix.sin6_addr), 1684 pr->ndpr_plen, e)); 1685 } else if ((flags & NDPRF_ONLINK) == 0 && 1686 (e = nd6_prefix_onlink(pr)) != 0) { 1687 nd6log((LOG_ERR, 1688 "pfxlist_onlink_check: failed to " 1689 "make %s/%d onlink, errno=%d\n", 1690 ip6_sprintf(ip6buf, 1691 &pr->ndpr_prefix.sin6_addr), 1692 pr->ndpr_plen, e)); 1693 } 1694 ND6_RLOCK(); 1695 if (genid != V_nd6_list_genid) 1696 goto restart; 1697 } 1698 } 1699 1700 /* 1701 * Changes on the prefix status might affect address status as well. 1702 * Make sure that all addresses derived from an attached prefix are 1703 * attached, and that all addresses derived from a detached prefix are 1704 * detached. Note, however, that a manually configured address should 1705 * always be attached. 1706 * The precise detection logic is same as the one for prefixes. 1707 */ 1708 IN6_IFADDR_RLOCK(&in6_ifa_tracker); 1709 TAILQ_FOREACH(ifa, &V_in6_ifaddrhead, ia_link) { 1710 if (!(ifa->ia6_flags & IN6_IFF_AUTOCONF)) 1711 continue; 1712 1713 if (ifa->ia6_ndpr == NULL) { 1714 /* 1715 * This can happen when we first configure the address 1716 * (i.e. the address exists, but the prefix does not). 1717 * XXX: complicated relationships... 1718 */ 1719 continue; 1720 } 1721 1722 if (find_pfxlist_reachable_router(ifa->ia6_ndpr)) 1723 break; 1724 } 1725 if (ifa) { 1726 TAILQ_FOREACH(ifa, &V_in6_ifaddrhead, ia_link) { 1727 if ((ifa->ia6_flags & IN6_IFF_AUTOCONF) == 0) 1728 continue; 1729 1730 if (ifa->ia6_ndpr == NULL) /* XXX: see above. */ 1731 continue; 1732 1733 if (find_pfxlist_reachable_router(ifa->ia6_ndpr)) { 1734 if (ifa->ia6_flags & IN6_IFF_DETACHED) { 1735 ifa->ia6_flags &= ~IN6_IFF_DETACHED; 1736 ifa->ia6_flags |= IN6_IFF_TENTATIVE; 1737 nd6_dad_start((struct ifaddr *)ifa, 0); 1738 } 1739 } else { 1740 ifa->ia6_flags |= IN6_IFF_DETACHED; 1741 } 1742 } 1743 } else { 1744 TAILQ_FOREACH(ifa, &V_in6_ifaddrhead, ia_link) { 1745 if ((ifa->ia6_flags & IN6_IFF_AUTOCONF) == 0) 1746 continue; 1747 1748 if (ifa->ia6_flags & IN6_IFF_DETACHED) { 1749 ifa->ia6_flags &= ~IN6_IFF_DETACHED; 1750 ifa->ia6_flags |= IN6_IFF_TENTATIVE; 1751 /* Do we need a delay in this case? */ 1752 nd6_dad_start((struct ifaddr *)ifa, 0); 1753 } 1754 } 1755 } 1756 IN6_IFADDR_RUNLOCK(&in6_ifa_tracker); 1757 ND6_RUNLOCK(); 1758 ND6_ONLINK_UNLOCK(); 1759 } 1760 1761 static int 1762 nd6_prefix_onlink_rtrequest(struct nd_prefix *pr, struct ifaddr *ifa) 1763 { 1764 static struct sockaddr_dl null_sdl = {sizeof(null_sdl), AF_LINK}; 1765 struct rib_head *rnh; 1766 struct rtentry *rt; 1767 struct sockaddr_in6 mask6; 1768 u_long rtflags; 1769 int error, a_failure, fibnum, maxfib; 1770 1771 /* 1772 * in6_ifinit() sets nd6_rtrequest to ifa_rtrequest for all ifaddrs. 1773 * ifa->ifa_rtrequest = nd6_rtrequest; 1774 */ 1775 bzero(&mask6, sizeof(mask6)); 1776 mask6.sin6_len = sizeof(mask6); 1777 mask6.sin6_addr = pr->ndpr_mask; 1778 rtflags = (ifa->ifa_flags & ~IFA_RTSELF) | RTF_UP; 1779 1780 if(V_rt_add_addr_allfibs) { 1781 fibnum = 0; 1782 maxfib = rt_numfibs; 1783 } else { 1784 fibnum = ifa->ifa_ifp->if_fib; 1785 maxfib = fibnum + 1; 1786 } 1787 a_failure = 0; 1788 for (; fibnum < maxfib; fibnum++) { 1789 1790 rt = NULL; 1791 error = in6_rtrequest(RTM_ADD, 1792 (struct sockaddr *)&pr->ndpr_prefix, ifa->ifa_addr, 1793 (struct sockaddr *)&mask6, rtflags, &rt, fibnum); 1794 if (error == 0) { 1795 KASSERT(rt != NULL, ("%s: in6_rtrequest return no " 1796 "error(%d) but rt is NULL, pr=%p, ifa=%p", __func__, 1797 error, pr, ifa)); 1798 1799 rnh = rt_tables_get_rnh(rt->rt_fibnum, AF_INET6); 1800 /* XXX what if rhn == NULL? */ 1801 RIB_WLOCK(rnh); 1802 RT_LOCK(rt); 1803 if (rt_setgate(rt, rt_key(rt), 1804 (struct sockaddr *)&null_sdl) == 0) { 1805 struct sockaddr_dl *dl; 1806 1807 dl = (struct sockaddr_dl *)rt->rt_gateway; 1808 dl->sdl_type = rt->rt_ifp->if_type; 1809 dl->sdl_index = rt->rt_ifp->if_index; 1810 } 1811 RIB_WUNLOCK(rnh); 1812 nd6_rtmsg(RTM_ADD, rt); 1813 RT_UNLOCK(rt); 1814 pr->ndpr_stateflags |= NDPRF_ONLINK; 1815 } else { 1816 char ip6buf[INET6_ADDRSTRLEN]; 1817 char ip6bufg[INET6_ADDRSTRLEN]; 1818 char ip6bufm[INET6_ADDRSTRLEN]; 1819 struct sockaddr_in6 *sin6; 1820 1821 sin6 = (struct sockaddr_in6 *)ifa->ifa_addr; 1822 nd6log((LOG_ERR, "nd6_prefix_onlink: failed to add " 1823 "route for a prefix (%s/%d) on %s, gw=%s, mask=%s, " 1824 "flags=%lx errno = %d\n", 1825 ip6_sprintf(ip6buf, &pr->ndpr_prefix.sin6_addr), 1826 pr->ndpr_plen, if_name(pr->ndpr_ifp), 1827 ip6_sprintf(ip6bufg, &sin6->sin6_addr), 1828 ip6_sprintf(ip6bufm, &mask6.sin6_addr), 1829 rtflags, error)); 1830 1831 /* Save last error to return, see rtinit(). */ 1832 a_failure = error; 1833 } 1834 1835 if (rt != NULL) { 1836 RT_LOCK(rt); 1837 RT_REMREF(rt); 1838 RT_UNLOCK(rt); 1839 } 1840 } 1841 1842 /* Return the last error we got. */ 1843 return (a_failure); 1844 } 1845 1846 int 1847 nd6_prefix_onlink(struct nd_prefix *pr) 1848 { 1849 struct ifaddr *ifa; 1850 struct ifnet *ifp = pr->ndpr_ifp; 1851 struct nd_prefix *opr; 1852 char ip6buf[INET6_ADDRSTRLEN]; 1853 int error; 1854 1855 ND6_ONLINK_LOCK_ASSERT(); 1856 ND6_UNLOCK_ASSERT(); 1857 1858 if ((pr->ndpr_stateflags & NDPRF_ONLINK) != 0) 1859 return (EEXIST); 1860 1861 /* 1862 * Add the interface route associated with the prefix. Before 1863 * installing the route, check if there's the same prefix on another 1864 * interface, and the prefix has already installed the interface route. 1865 * Although such a configuration is expected to be rare, we explicitly 1866 * allow it. 1867 */ 1868 ND6_RLOCK(); 1869 LIST_FOREACH(opr, &V_nd_prefix, ndpr_entry) { 1870 if (opr == pr) 1871 continue; 1872 1873 if ((opr->ndpr_stateflags & NDPRF_ONLINK) == 0) 1874 continue; 1875 1876 if (!V_rt_add_addr_allfibs && 1877 opr->ndpr_ifp->if_fib != pr->ndpr_ifp->if_fib) 1878 continue; 1879 1880 if (opr->ndpr_plen == pr->ndpr_plen && 1881 in6_are_prefix_equal(&pr->ndpr_prefix.sin6_addr, 1882 &opr->ndpr_prefix.sin6_addr, pr->ndpr_plen)) { 1883 ND6_RUNLOCK(); 1884 return (0); 1885 } 1886 } 1887 ND6_RUNLOCK(); 1888 1889 /* 1890 * We prefer link-local addresses as the associated interface address. 1891 */ 1892 /* search for a link-local addr */ 1893 ifa = (struct ifaddr *)in6ifa_ifpforlinklocal(ifp, 1894 IN6_IFF_NOTREADY | IN6_IFF_ANYCAST); 1895 if (ifa == NULL) { 1896 /* XXX: freebsd does not have ifa_ifwithaf */ 1897 IF_ADDR_RLOCK(ifp); 1898 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1899 if (ifa->ifa_addr->sa_family == AF_INET6) { 1900 ifa_ref(ifa); 1901 break; 1902 } 1903 } 1904 IF_ADDR_RUNLOCK(ifp); 1905 /* should we care about ia6_flags? */ 1906 } 1907 if (ifa == NULL) { 1908 /* 1909 * This can still happen, when, for example, we receive an RA 1910 * containing a prefix with the L bit set and the A bit clear, 1911 * after removing all IPv6 addresses on the receiving 1912 * interface. This should, of course, be rare though. 1913 */ 1914 nd6log((LOG_NOTICE, 1915 "nd6_prefix_onlink: failed to find any ifaddr" 1916 " to add route for a prefix(%s/%d) on %s\n", 1917 ip6_sprintf(ip6buf, &pr->ndpr_prefix.sin6_addr), 1918 pr->ndpr_plen, if_name(ifp))); 1919 return (0); 1920 } 1921 1922 error = nd6_prefix_onlink_rtrequest(pr, ifa); 1923 1924 if (ifa != NULL) 1925 ifa_free(ifa); 1926 1927 return (error); 1928 } 1929 1930 int 1931 nd6_prefix_offlink(struct nd_prefix *pr) 1932 { 1933 int error = 0; 1934 struct ifnet *ifp = pr->ndpr_ifp; 1935 struct nd_prefix *opr; 1936 struct sockaddr_in6 sa6, mask6; 1937 struct rtentry *rt; 1938 char ip6buf[INET6_ADDRSTRLEN]; 1939 uint64_t genid; 1940 int fibnum, maxfib, a_failure; 1941 1942 ND6_ONLINK_LOCK_ASSERT(); 1943 ND6_UNLOCK_ASSERT(); 1944 1945 if ((pr->ndpr_stateflags & NDPRF_ONLINK) == 0) 1946 return (EEXIST); 1947 1948 bzero(&sa6, sizeof(sa6)); 1949 sa6.sin6_family = AF_INET6; 1950 sa6.sin6_len = sizeof(sa6); 1951 bcopy(&pr->ndpr_prefix.sin6_addr, &sa6.sin6_addr, 1952 sizeof(struct in6_addr)); 1953 bzero(&mask6, sizeof(mask6)); 1954 mask6.sin6_family = AF_INET6; 1955 mask6.sin6_len = sizeof(sa6); 1956 bcopy(&pr->ndpr_mask, &mask6.sin6_addr, sizeof(struct in6_addr)); 1957 1958 if (V_rt_add_addr_allfibs) { 1959 fibnum = 0; 1960 maxfib = rt_numfibs; 1961 } else { 1962 fibnum = ifp->if_fib; 1963 maxfib = fibnum + 1; 1964 } 1965 1966 a_failure = 0; 1967 for (; fibnum < maxfib; fibnum++) { 1968 rt = NULL; 1969 error = in6_rtrequest(RTM_DELETE, (struct sockaddr *)&sa6, NULL, 1970 (struct sockaddr *)&mask6, 0, &rt, fibnum); 1971 if (error == 0) { 1972 /* report the route deletion to the routing socket. */ 1973 if (rt != NULL) 1974 nd6_rtmsg(RTM_DELETE, rt); 1975 } else { 1976 /* Save last error to return, see rtinit(). */ 1977 a_failure = error; 1978 } 1979 if (rt != NULL) { 1980 RTFREE(rt); 1981 } 1982 } 1983 error = a_failure; 1984 a_failure = 1; 1985 if (error == 0) { 1986 pr->ndpr_stateflags &= ~NDPRF_ONLINK; 1987 1988 /* 1989 * There might be the same prefix on another interface, 1990 * the prefix which could not be on-link just because we have 1991 * the interface route (see comments in nd6_prefix_onlink). 1992 * If there's one, try to make the prefix on-link on the 1993 * interface. 1994 */ 1995 ND6_RLOCK(); 1996 restart: 1997 LIST_FOREACH(opr, &V_nd_prefix, ndpr_entry) { 1998 /* 1999 * KAME specific: detached prefixes should not be 2000 * on-link. 2001 */ 2002 if (opr == pr || (opr->ndpr_stateflags & 2003 (NDPRF_ONLINK | NDPRF_DETACHED)) != 0) 2004 continue; 2005 2006 if (opr->ndpr_plen == pr->ndpr_plen && 2007 in6_are_prefix_equal(&pr->ndpr_prefix.sin6_addr, 2008 &opr->ndpr_prefix.sin6_addr, pr->ndpr_plen)) { 2009 int e; 2010 2011 genid = V_nd6_list_genid; 2012 ND6_RUNLOCK(); 2013 if ((e = nd6_prefix_onlink(opr)) != 0) { 2014 nd6log((LOG_ERR, 2015 "nd6_prefix_offlink: failed to " 2016 "recover a prefix %s/%d from %s " 2017 "to %s (errno = %d)\n", 2018 ip6_sprintf(ip6buf, 2019 &opr->ndpr_prefix.sin6_addr), 2020 opr->ndpr_plen, if_name(ifp), 2021 if_name(opr->ndpr_ifp), e)); 2022 } else 2023 a_failure = 0; 2024 ND6_RLOCK(); 2025 if (genid != V_nd6_list_genid) 2026 goto restart; 2027 } 2028 } 2029 ND6_RUNLOCK(); 2030 } else { 2031 /* XXX: can we still set the NDPRF_ONLINK flag? */ 2032 nd6log((LOG_ERR, 2033 "nd6_prefix_offlink: failed to delete route: " 2034 "%s/%d on %s (errno = %d)\n", 2035 ip6_sprintf(ip6buf, &sa6.sin6_addr), pr->ndpr_plen, 2036 if_name(ifp), error)); 2037 } 2038 2039 if (a_failure) 2040 lltable_prefix_free(AF_INET6, (struct sockaddr *)&sa6, 2041 (struct sockaddr *)&mask6, LLE_STATIC); 2042 2043 return (error); 2044 } 2045 2046 static struct in6_ifaddr * 2047 in6_ifadd(struct nd_prefixctl *pr, int mcast) 2048 { 2049 struct ifnet *ifp = pr->ndpr_ifp; 2050 struct ifaddr *ifa; 2051 struct in6_aliasreq ifra; 2052 struct in6_ifaddr *ia, *ib; 2053 int error, plen0; 2054 struct in6_addr mask; 2055 int prefixlen = pr->ndpr_plen; 2056 int updateflags; 2057 char ip6buf[INET6_ADDRSTRLEN]; 2058 2059 in6_prefixlen2mask(&mask, prefixlen); 2060 2061 /* 2062 * find a link-local address (will be interface ID). 2063 * Is it really mandatory? Theoretically, a global or a site-local 2064 * address can be configured without a link-local address, if we 2065 * have a unique interface identifier... 2066 * 2067 * it is not mandatory to have a link-local address, we can generate 2068 * interface identifier on the fly. we do this because: 2069 * (1) it should be the easiest way to find interface identifier. 2070 * (2) RFC2462 5.4 suggesting the use of the same interface identifier 2071 * for multiple addresses on a single interface, and possible shortcut 2072 * of DAD. we omitted DAD for this reason in the past. 2073 * (3) a user can prevent autoconfiguration of global address 2074 * by removing link-local address by hand (this is partly because we 2075 * don't have other way to control the use of IPv6 on an interface. 2076 * this has been our design choice - cf. NRL's "ifconfig auto"). 2077 * (4) it is easier to manage when an interface has addresses 2078 * with the same interface identifier, than to have multiple addresses 2079 * with different interface identifiers. 2080 */ 2081 ifa = (struct ifaddr *)in6ifa_ifpforlinklocal(ifp, 0); /* 0 is OK? */ 2082 if (ifa) 2083 ib = (struct in6_ifaddr *)ifa; 2084 else 2085 return NULL; 2086 2087 /* prefixlen + ifidlen must be equal to 128 */ 2088 plen0 = in6_mask2len(&ib->ia_prefixmask.sin6_addr, NULL); 2089 if (prefixlen != plen0) { 2090 ifa_free(ifa); 2091 nd6log((LOG_INFO, "in6_ifadd: wrong prefixlen for %s " 2092 "(prefix=%d ifid=%d)\n", 2093 if_name(ifp), prefixlen, 128 - plen0)); 2094 return NULL; 2095 } 2096 2097 /* make ifaddr */ 2098 in6_prepare_ifra(&ifra, &pr->ndpr_prefix.sin6_addr, &mask); 2099 2100 IN6_MASK_ADDR(&ifra.ifra_addr.sin6_addr, &mask); 2101 /* interface ID */ 2102 ifra.ifra_addr.sin6_addr.s6_addr32[0] |= 2103 (ib->ia_addr.sin6_addr.s6_addr32[0] & ~mask.s6_addr32[0]); 2104 ifra.ifra_addr.sin6_addr.s6_addr32[1] |= 2105 (ib->ia_addr.sin6_addr.s6_addr32[1] & ~mask.s6_addr32[1]); 2106 ifra.ifra_addr.sin6_addr.s6_addr32[2] |= 2107 (ib->ia_addr.sin6_addr.s6_addr32[2] & ~mask.s6_addr32[2]); 2108 ifra.ifra_addr.sin6_addr.s6_addr32[3] |= 2109 (ib->ia_addr.sin6_addr.s6_addr32[3] & ~mask.s6_addr32[3]); 2110 ifa_free(ifa); 2111 2112 /* lifetimes. */ 2113 ifra.ifra_lifetime.ia6t_vltime = pr->ndpr_vltime; 2114 ifra.ifra_lifetime.ia6t_pltime = pr->ndpr_pltime; 2115 2116 /* XXX: scope zone ID? */ 2117 2118 ifra.ifra_flags |= IN6_IFF_AUTOCONF; /* obey autoconf */ 2119 2120 /* 2121 * Make sure that we do not have this address already. This should 2122 * usually not happen, but we can still see this case, e.g., if we 2123 * have manually configured the exact address to be configured. 2124 */ 2125 ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp, 2126 &ifra.ifra_addr.sin6_addr); 2127 if (ifa != NULL) { 2128 ifa_free(ifa); 2129 /* this should be rare enough to make an explicit log */ 2130 log(LOG_INFO, "in6_ifadd: %s is already configured\n", 2131 ip6_sprintf(ip6buf, &ifra.ifra_addr.sin6_addr)); 2132 return (NULL); 2133 } 2134 2135 /* 2136 * Allocate ifaddr structure, link into chain, etc. 2137 * If we are going to create a new address upon receiving a multicasted 2138 * RA, we need to impose a random delay before starting DAD. 2139 * [draft-ietf-ipv6-rfc2462bis-02.txt, Section 5.4.2] 2140 */ 2141 updateflags = 0; 2142 if (mcast) 2143 updateflags |= IN6_IFAUPDATE_DADDELAY; 2144 if ((error = in6_update_ifa(ifp, &ifra, NULL, updateflags)) != 0) { 2145 nd6log((LOG_ERR, 2146 "in6_ifadd: failed to make ifaddr %s on %s (errno=%d)\n", 2147 ip6_sprintf(ip6buf, &ifra.ifra_addr.sin6_addr), 2148 if_name(ifp), error)); 2149 return (NULL); /* ifaddr must not have been allocated. */ 2150 } 2151 2152 ia = in6ifa_ifpwithaddr(ifp, &ifra.ifra_addr.sin6_addr); 2153 /* 2154 * XXXRW: Assumption of non-NULLness here might not be true with 2155 * fine-grained locking -- should we validate it? Or just return 2156 * earlier ifa rather than looking it up again? 2157 */ 2158 return (ia); /* this is always non-NULL and referenced. */ 2159 } 2160 2161 /* 2162 * ia0 - corresponding public address 2163 */ 2164 int 2165 in6_tmpifadd(const struct in6_ifaddr *ia0, int forcegen, int delay) 2166 { 2167 struct ifnet *ifp = ia0->ia_ifa.ifa_ifp; 2168 struct in6_ifaddr *newia; 2169 struct in6_aliasreq ifra; 2170 int error; 2171 int trylimit = 3; /* XXX: adhoc value */ 2172 int updateflags; 2173 u_int32_t randid[2]; 2174 time_t vltime0, pltime0; 2175 2176 in6_prepare_ifra(&ifra, &ia0->ia_addr.sin6_addr, 2177 &ia0->ia_prefixmask.sin6_addr); 2178 2179 ifra.ifra_addr = ia0->ia_addr; /* XXX: do we need this ? */ 2180 /* clear the old IFID */ 2181 IN6_MASK_ADDR(&ifra.ifra_addr.sin6_addr, 2182 &ifra.ifra_prefixmask.sin6_addr); 2183 2184 again: 2185 if (in6_get_tmpifid(ifp, (u_int8_t *)randid, 2186 (const u_int8_t *)&ia0->ia_addr.sin6_addr.s6_addr[8], forcegen)) { 2187 nd6log((LOG_NOTICE, "in6_tmpifadd: failed to find a good " 2188 "random IFID\n")); 2189 return (EINVAL); 2190 } 2191 ifra.ifra_addr.sin6_addr.s6_addr32[2] |= 2192 (randid[0] & ~(ifra.ifra_prefixmask.sin6_addr.s6_addr32[2])); 2193 ifra.ifra_addr.sin6_addr.s6_addr32[3] |= 2194 (randid[1] & ~(ifra.ifra_prefixmask.sin6_addr.s6_addr32[3])); 2195 2196 /* 2197 * in6_get_tmpifid() quite likely provided a unique interface ID. 2198 * However, we may still have a chance to see collision, because 2199 * there may be a time lag between generation of the ID and generation 2200 * of the address. So, we'll do one more sanity check. 2201 */ 2202 2203 if (in6_localip(&ifra.ifra_addr.sin6_addr) != 0) { 2204 if (trylimit-- > 0) { 2205 forcegen = 1; 2206 goto again; 2207 } 2208 2209 /* Give up. Something strange should have happened. */ 2210 nd6log((LOG_NOTICE, "in6_tmpifadd: failed to " 2211 "find a unique random IFID\n")); 2212 return (EEXIST); 2213 } 2214 2215 /* 2216 * The Valid Lifetime is the lower of the Valid Lifetime of the 2217 * public address or TEMP_VALID_LIFETIME. 2218 * The Preferred Lifetime is the lower of the Preferred Lifetime 2219 * of the public address or TEMP_PREFERRED_LIFETIME - 2220 * DESYNC_FACTOR. 2221 */ 2222 if (ia0->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) { 2223 vltime0 = IFA6_IS_INVALID(ia0) ? 0 : 2224 (ia0->ia6_lifetime.ia6t_vltime - 2225 (time_uptime - ia0->ia6_updatetime)); 2226 if (vltime0 > V_ip6_temp_valid_lifetime) 2227 vltime0 = V_ip6_temp_valid_lifetime; 2228 } else 2229 vltime0 = V_ip6_temp_valid_lifetime; 2230 if (ia0->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) { 2231 pltime0 = IFA6_IS_DEPRECATED(ia0) ? 0 : 2232 (ia0->ia6_lifetime.ia6t_pltime - 2233 (time_uptime - ia0->ia6_updatetime)); 2234 if (pltime0 > V_ip6_temp_preferred_lifetime - V_ip6_desync_factor){ 2235 pltime0 = V_ip6_temp_preferred_lifetime - 2236 V_ip6_desync_factor; 2237 } 2238 } else 2239 pltime0 = V_ip6_temp_preferred_lifetime - V_ip6_desync_factor; 2240 ifra.ifra_lifetime.ia6t_vltime = vltime0; 2241 ifra.ifra_lifetime.ia6t_pltime = pltime0; 2242 2243 /* 2244 * A temporary address is created only if this calculated Preferred 2245 * Lifetime is greater than REGEN_ADVANCE time units. 2246 */ 2247 if (ifra.ifra_lifetime.ia6t_pltime <= V_ip6_temp_regen_advance) 2248 return (0); 2249 2250 /* XXX: scope zone ID? */ 2251 2252 ifra.ifra_flags |= (IN6_IFF_AUTOCONF|IN6_IFF_TEMPORARY); 2253 2254 /* allocate ifaddr structure, link into chain, etc. */ 2255 updateflags = 0; 2256 if (delay) 2257 updateflags |= IN6_IFAUPDATE_DADDELAY; 2258 if ((error = in6_update_ifa(ifp, &ifra, NULL, updateflags)) != 0) 2259 return (error); 2260 2261 newia = in6ifa_ifpwithaddr(ifp, &ifra.ifra_addr.sin6_addr); 2262 if (newia == NULL) { /* XXX: can it happen? */ 2263 nd6log((LOG_ERR, 2264 "in6_tmpifadd: ifa update succeeded, but we got " 2265 "no ifaddr\n")); 2266 return (EINVAL); /* XXX */ 2267 } 2268 newia->ia6_ndpr = ia0->ia6_ndpr; 2269 newia->ia6_ndpr->ndpr_addrcnt++; 2270 ifa_free(&newia->ia_ifa); 2271 2272 /* 2273 * A newly added address might affect the status of other addresses. 2274 * XXX: when the temporary address is generated with a new public 2275 * address, the onlink check is redundant. However, it would be safe 2276 * to do the check explicitly everywhere a new address is generated, 2277 * and, in fact, we surely need the check when we create a new 2278 * temporary address due to deprecation of an old temporary address. 2279 */ 2280 pfxlist_onlink_check(); 2281 2282 return (0); 2283 } 2284 2285 static int 2286 in6_init_prefix_ltimes(struct nd_prefix *ndpr) 2287 { 2288 if (ndpr->ndpr_pltime == ND6_INFINITE_LIFETIME) 2289 ndpr->ndpr_preferred = 0; 2290 else 2291 ndpr->ndpr_preferred = time_uptime + ndpr->ndpr_pltime; 2292 if (ndpr->ndpr_vltime == ND6_INFINITE_LIFETIME) 2293 ndpr->ndpr_expire = 0; 2294 else 2295 ndpr->ndpr_expire = time_uptime + ndpr->ndpr_vltime; 2296 2297 return 0; 2298 } 2299 2300 static void 2301 in6_init_address_ltimes(struct nd_prefix *new, struct in6_addrlifetime *lt6) 2302 { 2303 /* init ia6t_expire */ 2304 if (lt6->ia6t_vltime == ND6_INFINITE_LIFETIME) 2305 lt6->ia6t_expire = 0; 2306 else { 2307 lt6->ia6t_expire = time_uptime; 2308 lt6->ia6t_expire += lt6->ia6t_vltime; 2309 } 2310 2311 /* init ia6t_preferred */ 2312 if (lt6->ia6t_pltime == ND6_INFINITE_LIFETIME) 2313 lt6->ia6t_preferred = 0; 2314 else { 2315 lt6->ia6t_preferred = time_uptime; 2316 lt6->ia6t_preferred += lt6->ia6t_pltime; 2317 } 2318 } 2319 2320 /* 2321 * Delete all the routing table entries that use the specified gateway. 2322 * XXX: this function causes search through all entries of routing table, so 2323 * it shouldn't be called when acting as a router. 2324 */ 2325 void 2326 rt6_flush(struct in6_addr *gateway, struct ifnet *ifp) 2327 { 2328 2329 /* We'll care only link-local addresses */ 2330 if (!IN6_IS_ADDR_LINKLOCAL(gateway)) 2331 return; 2332 2333 /* XXX Do we really need to walk any but the default FIB? */ 2334 rt_foreach_fib_walk_del(AF_INET6, rt6_deleteroute, (void *)gateway); 2335 } 2336 2337 static int 2338 rt6_deleteroute(const struct rtentry *rt, void *arg) 2339 { 2340 #define SIN6(s) ((struct sockaddr_in6 *)s) 2341 struct in6_addr *gate = (struct in6_addr *)arg; 2342 2343 if (rt->rt_gateway == NULL || rt->rt_gateway->sa_family != AF_INET6) 2344 return (0); 2345 2346 if (!IN6_ARE_ADDR_EQUAL(gate, &SIN6(rt->rt_gateway)->sin6_addr)) { 2347 return (0); 2348 } 2349 2350 /* 2351 * Do not delete a static route. 2352 * XXX: this seems to be a bit ad-hoc. Should we consider the 2353 * 'cloned' bit instead? 2354 */ 2355 if ((rt->rt_flags & RTF_STATIC) != 0) 2356 return (0); 2357 2358 /* 2359 * We delete only host route. This means, in particular, we don't 2360 * delete default route. 2361 */ 2362 if ((rt->rt_flags & RTF_HOST) == 0) 2363 return (0); 2364 2365 return (1); 2366 #undef SIN6 2367 } 2368 2369 int 2370 nd6_setdefaultiface(int ifindex) 2371 { 2372 int error = 0; 2373 2374 if (ifindex < 0 || V_if_index < ifindex) 2375 return (EINVAL); 2376 if (ifindex != 0 && !ifnet_byindex(ifindex)) 2377 return (EINVAL); 2378 2379 if (V_nd6_defifindex != ifindex) { 2380 V_nd6_defifindex = ifindex; 2381 if (V_nd6_defifindex > 0) 2382 V_nd6_defifp = ifnet_byindex(V_nd6_defifindex); 2383 else 2384 V_nd6_defifp = NULL; 2385 2386 /* 2387 * Our current implementation assumes one-to-one maping between 2388 * interfaces and links, so it would be natural to use the 2389 * default interface as the default link. 2390 */ 2391 scope6_setdefault(V_nd6_defifp); 2392 } 2393 2394 return (error); 2395 } 2396