1 /* $NetBSD: nd6.c,v 1.128 2008/05/15 01:33:28 dyoung Exp $ */ 2 /* $KAME: nd6.c,v 1.279 2002/06/08 11:16:51 itojun Exp $ */ 3 4 /* 5 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project. 6 * All rights reserved. 7 * 8 * Redistribution and use in source and binary forms, with or without 9 * modification, are permitted provided that the following conditions 10 * are met: 11 * 1. Redistributions of source code must retain the above copyright 12 * notice, this list of conditions and the following disclaimer. 13 * 2. Redistributions in binary form must reproduce the above copyright 14 * notice, this list of conditions and the following disclaimer in the 15 * documentation and/or other materials provided with the distribution. 16 * 3. Neither the name of the project nor the names of its contributors 17 * may be used to endorse or promote products derived from this software 18 * without specific prior written permission. 19 * 20 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND 21 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 22 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 23 * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE 24 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 25 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 26 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 27 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 28 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 29 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 30 * SUCH DAMAGE. 31 */ 32 33 #include <sys/cdefs.h> 34 __KERNEL_RCSID(0, "$NetBSD: nd6.c,v 1.128 2008/05/15 01:33:28 dyoung Exp $"); 35 36 #include "opt_ipsec.h" 37 38 #include <sys/param.h> 39 #include <sys/systm.h> 40 #include <sys/callout.h> 41 #include <sys/malloc.h> 42 #include <sys/mbuf.h> 43 #include <sys/socket.h> 44 #include <sys/socketvar.h> 45 #include <sys/sockio.h> 46 #include <sys/time.h> 47 #include <sys/kernel.h> 48 #include <sys/protosw.h> 49 #include <sys/errno.h> 50 #include <sys/ioctl.h> 51 #include <sys/syslog.h> 52 #include <sys/queue.h> 53 54 #include <net/if.h> 55 #include <net/if_dl.h> 56 #include <net/if_types.h> 57 #include <net/route.h> 58 #include <net/if_ether.h> 59 #include <net/if_fddi.h> 60 #include <net/if_arc.h> 61 62 #include <netinet/in.h> 63 #include <netinet6/in6_var.h> 64 #include <netinet/ip6.h> 65 #include <netinet6/ip6_var.h> 66 #include <netinet6/scope6_var.h> 67 #include <netinet6/nd6.h> 68 #include <netinet/icmp6.h> 69 #include <netinet6/icmp6_private.h> 70 71 #ifdef IPSEC 72 #include <netinet6/ipsec.h> 73 #endif 74 75 #include <net/net_osdep.h> 76 77 #define ND6_SLOWTIMER_INTERVAL (60 * 60) /* 1 hour */ 78 #define ND6_RECALC_REACHTM_INTERVAL (60 * 120) /* 2 hours */ 79 80 /* timer values */ 81 int nd6_prune = 1; /* walk list every 1 seconds */ 82 int nd6_delay = 5; /* delay first probe time 5 second */ 83 int nd6_umaxtries = 3; /* maximum unicast query */ 84 int nd6_mmaxtries = 3; /* maximum multicast query */ 85 int nd6_useloopback = 1; /* use loopback interface for local traffic */ 86 int nd6_gctimer = (60 * 60 * 24); /* 1 day: garbage collection timer */ 87 88 /* preventing too many loops in ND option parsing */ 89 int nd6_maxndopt = 10; /* max # of ND options allowed */ 90 91 int nd6_maxnudhint = 0; /* max # of subsequent upper layer hints */ 92 93 int nd6_maxqueuelen = 1; /* max # of packets cached in unresolved ND entries */ 94 95 #ifdef ND6_DEBUG 96 int nd6_debug = 1; 97 #else 98 int nd6_debug = 0; 99 #endif 100 101 /* for debugging? */ 102 static int nd6_inuse, nd6_allocated; 103 104 struct llinfo_nd6 llinfo_nd6 = { 105 .ln_prev = &llinfo_nd6, 106 .ln_next = &llinfo_nd6, 107 }; 108 struct nd_drhead nd_defrouter; 109 struct nd_prhead nd_prefix = { 0 }; 110 111 int nd6_recalc_reachtm_interval = ND6_RECALC_REACHTM_INTERVAL; 112 static const struct sockaddr_in6 all1_sa = { 113 .sin6_family = AF_INET6 114 , .sin6_len = sizeof(struct sockaddr_in6) 115 , .sin6_addr = {.s6_addr = {0xff, 0xff, 0xff, 0xff, 116 0xff, 0xff, 0xff, 0xff, 117 0xff, 0xff, 0xff, 0xff, 118 0xff, 0xff, 0xff, 0xff}} 119 }; 120 121 static void nd6_setmtu0(struct ifnet *, struct nd_ifinfo *); 122 static void nd6_slowtimo(void *); 123 static int regen_tmpaddr(struct in6_ifaddr *); 124 static struct llinfo_nd6 *nd6_free(struct rtentry *, int); 125 static void nd6_llinfo_timer(void *); 126 static void clear_llinfo_pqueue(struct llinfo_nd6 *); 127 128 callout_t nd6_slowtimo_ch; 129 callout_t nd6_timer_ch; 130 extern callout_t in6_tmpaddrtimer_ch; 131 132 static int fill_drlist(void *, size_t *, size_t); 133 static int fill_prlist(void *, size_t *, size_t); 134 135 MALLOC_DEFINE(M_IP6NDP, "NDP", "IPv6 Neighbour Discovery"); 136 137 void 138 nd6_init(void) 139 { 140 static int nd6_init_done = 0; 141 142 if (nd6_init_done) { 143 log(LOG_NOTICE, "nd6_init called more than once(ignored)\n"); 144 return; 145 } 146 147 /* initialization of the default router list */ 148 TAILQ_INIT(&nd_defrouter); 149 150 nd6_init_done = 1; 151 152 callout_init(&nd6_slowtimo_ch, CALLOUT_MPSAFE); 153 callout_init(&nd6_timer_ch, CALLOUT_MPSAFE); 154 155 /* start timer */ 156 callout_reset(&nd6_slowtimo_ch, ND6_SLOWTIMER_INTERVAL * hz, 157 nd6_slowtimo, NULL); 158 } 159 160 struct nd_ifinfo * 161 nd6_ifattach(struct ifnet *ifp) 162 { 163 struct nd_ifinfo *nd; 164 165 nd = (struct nd_ifinfo *)malloc(sizeof(*nd), M_IP6NDP, M_WAITOK); 166 bzero(nd, sizeof(*nd)); 167 168 nd->initialized = 1; 169 170 nd->chlim = IPV6_DEFHLIM; 171 nd->basereachable = REACHABLE_TIME; 172 nd->reachable = ND_COMPUTE_RTIME(nd->basereachable); 173 nd->retrans = RETRANS_TIMER; 174 /* 175 * Note that the default value of ip6_accept_rtadv is 0, which means 176 * we won't accept RAs by default even if we set ND6_IFF_ACCEPT_RTADV 177 * here. 178 */ 179 nd->flags = (ND6_IFF_PERFORMNUD | ND6_IFF_ACCEPT_RTADV); 180 181 /* XXX: we cannot call nd6_setmtu since ifp is not fully initialized */ 182 nd6_setmtu0(ifp, nd); 183 184 return nd; 185 } 186 187 void 188 nd6_ifdetach(struct nd_ifinfo *nd) 189 { 190 191 free(nd, M_IP6NDP); 192 } 193 194 void 195 nd6_setmtu(struct ifnet *ifp) 196 { 197 nd6_setmtu0(ifp, ND_IFINFO(ifp)); 198 } 199 200 void 201 nd6_setmtu0(struct ifnet *ifp, struct nd_ifinfo *ndi) 202 { 203 u_int32_t omaxmtu; 204 205 omaxmtu = ndi->maxmtu; 206 207 switch (ifp->if_type) { 208 case IFT_ARCNET: 209 ndi->maxmtu = MIN(ARC_PHDS_MAXMTU, ifp->if_mtu); /* RFC2497 */ 210 break; 211 case IFT_FDDI: 212 ndi->maxmtu = MIN(FDDIIPMTU, ifp->if_mtu); 213 break; 214 default: 215 ndi->maxmtu = ifp->if_mtu; 216 break; 217 } 218 219 /* 220 * Decreasing the interface MTU under IPV6 minimum MTU may cause 221 * undesirable situation. We thus notify the operator of the change 222 * explicitly. The check for omaxmtu is necessary to restrict the 223 * log to the case of changing the MTU, not initializing it. 224 */ 225 if (omaxmtu >= IPV6_MMTU && ndi->maxmtu < IPV6_MMTU) { 226 log(LOG_NOTICE, "nd6_setmtu0: new link MTU on %s (%lu) is too" 227 " small for IPv6 which needs %lu\n", 228 if_name(ifp), (unsigned long)ndi->maxmtu, (unsigned long) 229 IPV6_MMTU); 230 } 231 232 if (ndi->maxmtu > in6_maxmtu) 233 in6_setmaxmtu(); /* check all interfaces just in case */ 234 } 235 236 void 237 nd6_option_init(void *opt, int icmp6len, union nd_opts *ndopts) 238 { 239 240 bzero(ndopts, sizeof(*ndopts)); 241 ndopts->nd_opts_search = (struct nd_opt_hdr *)opt; 242 ndopts->nd_opts_last 243 = (struct nd_opt_hdr *)(((u_char *)opt) + icmp6len); 244 245 if (icmp6len == 0) { 246 ndopts->nd_opts_done = 1; 247 ndopts->nd_opts_search = NULL; 248 } 249 } 250 251 /* 252 * Take one ND option. 253 */ 254 struct nd_opt_hdr * 255 nd6_option(union nd_opts *ndopts) 256 { 257 struct nd_opt_hdr *nd_opt; 258 int olen; 259 260 if (ndopts == NULL) 261 panic("ndopts == NULL in nd6_option"); 262 if (ndopts->nd_opts_last == NULL) 263 panic("uninitialized ndopts in nd6_option"); 264 if (ndopts->nd_opts_search == NULL) 265 return NULL; 266 if (ndopts->nd_opts_done) 267 return NULL; 268 269 nd_opt = ndopts->nd_opts_search; 270 271 /* make sure nd_opt_len is inside the buffer */ 272 if ((void *)&nd_opt->nd_opt_len >= (void *)ndopts->nd_opts_last) { 273 bzero(ndopts, sizeof(*ndopts)); 274 return NULL; 275 } 276 277 olen = nd_opt->nd_opt_len << 3; 278 if (olen == 0) { 279 /* 280 * Message validation requires that all included 281 * options have a length that is greater than zero. 282 */ 283 bzero(ndopts, sizeof(*ndopts)); 284 return NULL; 285 } 286 287 ndopts->nd_opts_search = (struct nd_opt_hdr *)((char *)nd_opt + olen); 288 if (ndopts->nd_opts_search > ndopts->nd_opts_last) { 289 /* option overruns the end of buffer, invalid */ 290 bzero(ndopts, sizeof(*ndopts)); 291 return NULL; 292 } else if (ndopts->nd_opts_search == ndopts->nd_opts_last) { 293 /* reached the end of options chain */ 294 ndopts->nd_opts_done = 1; 295 ndopts->nd_opts_search = NULL; 296 } 297 return nd_opt; 298 } 299 300 /* 301 * Parse multiple ND options. 302 * This function is much easier to use, for ND routines that do not need 303 * multiple options of the same type. 304 */ 305 int 306 nd6_options(union nd_opts *ndopts) 307 { 308 struct nd_opt_hdr *nd_opt; 309 int i = 0; 310 311 if (ndopts == NULL) 312 panic("ndopts == NULL in nd6_options"); 313 if (ndopts->nd_opts_last == NULL) 314 panic("uninitialized ndopts in nd6_options"); 315 if (ndopts->nd_opts_search == NULL) 316 return 0; 317 318 while (1) { 319 nd_opt = nd6_option(ndopts); 320 if (nd_opt == NULL && ndopts->nd_opts_last == NULL) { 321 /* 322 * Message validation requires that all included 323 * options have a length that is greater than zero. 324 */ 325 ICMP6_STATINC(ICMP6_STAT_ND_BADOPT); 326 bzero(ndopts, sizeof(*ndopts)); 327 return -1; 328 } 329 330 if (nd_opt == NULL) 331 goto skip1; 332 333 switch (nd_opt->nd_opt_type) { 334 case ND_OPT_SOURCE_LINKADDR: 335 case ND_OPT_TARGET_LINKADDR: 336 case ND_OPT_MTU: 337 case ND_OPT_REDIRECTED_HEADER: 338 if (ndopts->nd_opt_array[nd_opt->nd_opt_type]) { 339 nd6log((LOG_INFO, 340 "duplicated ND6 option found (type=%d)\n", 341 nd_opt->nd_opt_type)); 342 /* XXX bark? */ 343 } else { 344 ndopts->nd_opt_array[nd_opt->nd_opt_type] 345 = nd_opt; 346 } 347 break; 348 case ND_OPT_PREFIX_INFORMATION: 349 if (ndopts->nd_opt_array[nd_opt->nd_opt_type] == 0) { 350 ndopts->nd_opt_array[nd_opt->nd_opt_type] 351 = nd_opt; 352 } 353 ndopts->nd_opts_pi_end = 354 (struct nd_opt_prefix_info *)nd_opt; 355 break; 356 default: 357 /* 358 * Unknown options must be silently ignored, 359 * to accommodate future extension to the protocol. 360 */ 361 nd6log((LOG_DEBUG, 362 "nd6_options: unsupported option %d - " 363 "option ignored\n", nd_opt->nd_opt_type)); 364 } 365 366 skip1: 367 i++; 368 if (i > nd6_maxndopt) { 369 ICMP6_STATINC(ICMP6_STAT_ND_TOOMANYOPT); 370 nd6log((LOG_INFO, "too many loop in nd opt\n")); 371 break; 372 } 373 374 if (ndopts->nd_opts_done) 375 break; 376 } 377 378 return 0; 379 } 380 381 /* 382 * ND6 timer routine to handle ND6 entries 383 */ 384 void 385 nd6_llinfo_settimer(struct llinfo_nd6 *ln, long xtick) 386 { 387 int s; 388 389 s = splsoftnet(); 390 391 if (xtick < 0) { 392 ln->ln_expire = 0; 393 ln->ln_ntick = 0; 394 callout_stop(&ln->ln_timer_ch); 395 } else { 396 ln->ln_expire = time_second + xtick / hz; 397 if (xtick > INT_MAX) { 398 ln->ln_ntick = xtick - INT_MAX; 399 callout_reset(&ln->ln_timer_ch, INT_MAX, 400 nd6_llinfo_timer, ln); 401 } else { 402 ln->ln_ntick = 0; 403 callout_reset(&ln->ln_timer_ch, xtick, 404 nd6_llinfo_timer, ln); 405 } 406 } 407 408 splx(s); 409 } 410 411 static void 412 nd6_llinfo_timer(void *arg) 413 { 414 struct llinfo_nd6 *ln; 415 struct rtentry *rt; 416 const struct sockaddr_in6 *dst; 417 struct ifnet *ifp; 418 struct nd_ifinfo *ndi = NULL; 419 420 mutex_enter(softnet_lock); 421 KERNEL_LOCK(1, NULL); 422 423 ln = (struct llinfo_nd6 *)arg; 424 425 if (ln->ln_ntick > 0) { 426 nd6_llinfo_settimer(ln, ln->ln_ntick); 427 KERNEL_UNLOCK_ONE(NULL); 428 mutex_exit(softnet_lock); 429 return; 430 } 431 432 if ((rt = ln->ln_rt) == NULL) 433 panic("ln->ln_rt == NULL"); 434 if ((ifp = rt->rt_ifp) == NULL) 435 panic("ln->ln_rt->rt_ifp == NULL"); 436 ndi = ND_IFINFO(ifp); 437 dst = satocsin6(rt_getkey(rt)); 438 439 /* sanity check */ 440 if (rt->rt_llinfo && (struct llinfo_nd6 *)rt->rt_llinfo != ln) 441 panic("rt_llinfo(%p) is not equal to ln(%p)", 442 rt->rt_llinfo, ln); 443 if (!dst) 444 panic("dst=0 in nd6_timer(ln=%p)", ln); 445 446 switch (ln->ln_state) { 447 case ND6_LLINFO_INCOMPLETE: 448 if (ln->ln_asked < nd6_mmaxtries) { 449 ln->ln_asked++; 450 nd6_llinfo_settimer(ln, (long)ndi->retrans * hz / 1000); 451 nd6_ns_output(ifp, NULL, &dst->sin6_addr, ln, 0); 452 } else { 453 struct mbuf *m = ln->ln_hold; 454 if (m) { 455 struct mbuf *m0; 456 457 /* 458 * assuming every packet in ln_hold has 459 * the same IP header 460 */ 461 m0 = m->m_nextpkt; 462 m->m_nextpkt = NULL; 463 icmp6_error2(m, ICMP6_DST_UNREACH, 464 ICMP6_DST_UNREACH_ADDR, 0, rt->rt_ifp); 465 466 ln->ln_hold = m0; 467 clear_llinfo_pqueue(ln); 468 } 469 (void)nd6_free(rt, 0); 470 ln = NULL; 471 } 472 break; 473 case ND6_LLINFO_REACHABLE: 474 if (!ND6_LLINFO_PERMANENT(ln)) { 475 ln->ln_state = ND6_LLINFO_STALE; 476 nd6_llinfo_settimer(ln, (long)nd6_gctimer * hz); 477 } 478 break; 479 480 case ND6_LLINFO_STALE: 481 /* Garbage Collection(RFC 2461 5.3) */ 482 if (!ND6_LLINFO_PERMANENT(ln)) { 483 (void)nd6_free(rt, 1); 484 ln = NULL; 485 } 486 break; 487 488 case ND6_LLINFO_DELAY: 489 if (ndi && (ndi->flags & ND6_IFF_PERFORMNUD) != 0) { 490 /* We need NUD */ 491 ln->ln_asked = 1; 492 ln->ln_state = ND6_LLINFO_PROBE; 493 nd6_llinfo_settimer(ln, (long)ndi->retrans * hz / 1000); 494 nd6_ns_output(ifp, &dst->sin6_addr, 495 &dst->sin6_addr, ln, 0); 496 } else { 497 ln->ln_state = ND6_LLINFO_STALE; /* XXX */ 498 nd6_llinfo_settimer(ln, (long)nd6_gctimer * hz); 499 } 500 break; 501 case ND6_LLINFO_PROBE: 502 if (ln->ln_asked < nd6_umaxtries) { 503 ln->ln_asked++; 504 nd6_llinfo_settimer(ln, (long)ndi->retrans * hz / 1000); 505 nd6_ns_output(ifp, &dst->sin6_addr, 506 &dst->sin6_addr, ln, 0); 507 } else { 508 (void)nd6_free(rt, 0); 509 ln = NULL; 510 } 511 break; 512 } 513 514 KERNEL_UNLOCK_ONE(NULL); 515 mutex_exit(softnet_lock); 516 } 517 518 /* 519 * ND6 timer routine to expire default route list and prefix list 520 */ 521 void 522 nd6_timer(void *ignored_arg) 523 { 524 struct nd_defrouter *next_dr, *dr; 525 struct nd_prefix *next_pr, *pr; 526 struct in6_ifaddr *ia6, *nia6; 527 struct in6_addrlifetime *lt6; 528 529 callout_reset(&nd6_timer_ch, nd6_prune * hz, 530 nd6_timer, NULL); 531 532 mutex_enter(softnet_lock); 533 KERNEL_LOCK(1, NULL); 534 535 /* expire default router list */ 536 537 for (dr = TAILQ_FIRST(&nd_defrouter); dr != NULL; dr = next_dr) { 538 next_dr = TAILQ_NEXT(dr, dr_entry); 539 if (dr->expire && dr->expire < time_second) { 540 defrtrlist_del(dr); 541 } 542 } 543 544 /* 545 * expire interface addresses. 546 * in the past the loop was inside prefix expiry processing. 547 * However, from a stricter speci-confrmance standpoint, we should 548 * rather separate address lifetimes and prefix lifetimes. 549 */ 550 addrloop: 551 for (ia6 = in6_ifaddr; ia6; ia6 = nia6) { 552 nia6 = ia6->ia_next; 553 /* check address lifetime */ 554 lt6 = &ia6->ia6_lifetime; 555 if (IFA6_IS_INVALID(ia6)) { 556 int regen = 0; 557 558 /* 559 * If the expiring address is temporary, try 560 * regenerating a new one. This would be useful when 561 * we suspended a laptop PC, then turned it on after a 562 * period that could invalidate all temporary 563 * addresses. Although we may have to restart the 564 * loop (see below), it must be after purging the 565 * address. Otherwise, we'd see an infinite loop of 566 * regeneration. 567 */ 568 if (ip6_use_tempaddr && 569 (ia6->ia6_flags & IN6_IFF_TEMPORARY) != 0) { 570 if (regen_tmpaddr(ia6) == 0) 571 regen = 1; 572 } 573 574 in6_purgeaddr(&ia6->ia_ifa); 575 576 if (regen) 577 goto addrloop; /* XXX: see below */ 578 } else if (IFA6_IS_DEPRECATED(ia6)) { 579 int oldflags = ia6->ia6_flags; 580 581 ia6->ia6_flags |= IN6_IFF_DEPRECATED; 582 583 /* 584 * If a temporary address has just become deprecated, 585 * regenerate a new one if possible. 586 */ 587 if (ip6_use_tempaddr && 588 (ia6->ia6_flags & IN6_IFF_TEMPORARY) != 0 && 589 (oldflags & IN6_IFF_DEPRECATED) == 0) { 590 591 if (regen_tmpaddr(ia6) == 0) { 592 /* 593 * A new temporary address is 594 * generated. 595 * XXX: this means the address chain 596 * has changed while we are still in 597 * the loop. Although the change 598 * would not cause disaster (because 599 * it's not a deletion, but an 600 * addition,) we'd rather restart the 601 * loop just for safety. Or does this 602 * significantly reduce performance?? 603 */ 604 goto addrloop; 605 } 606 } 607 } else { 608 /* 609 * A new RA might have made a deprecated address 610 * preferred. 611 */ 612 ia6->ia6_flags &= ~IN6_IFF_DEPRECATED; 613 } 614 } 615 616 /* expire prefix list */ 617 for (pr = LIST_FIRST(&nd_prefix); pr != NULL; pr = next_pr) { 618 next_pr = LIST_NEXT(pr, ndpr_entry); 619 /* 620 * check prefix lifetime. 621 * since pltime is just for autoconf, pltime processing for 622 * prefix is not necessary. 623 */ 624 if (pr->ndpr_vltime != ND6_INFINITE_LIFETIME && 625 time_second - pr->ndpr_lastupdate > pr->ndpr_vltime) { 626 627 /* 628 * address expiration and prefix expiration are 629 * separate. NEVER perform in6_purgeaddr here. 630 */ 631 632 prelist_remove(pr); 633 } 634 } 635 636 KERNEL_UNLOCK_ONE(NULL); 637 mutex_exit(softnet_lock); 638 } 639 640 /* ia6: deprecated/invalidated temporary address */ 641 static int 642 regen_tmpaddr(struct in6_ifaddr *ia6) 643 { 644 struct ifaddr *ifa; 645 struct ifnet *ifp; 646 struct in6_ifaddr *public_ifa6 = NULL; 647 648 ifp = ia6->ia_ifa.ifa_ifp; 649 IFADDR_FOREACH(ifa, ifp) { 650 struct in6_ifaddr *it6; 651 652 if (ifa->ifa_addr->sa_family != AF_INET6) 653 continue; 654 655 it6 = (struct in6_ifaddr *)ifa; 656 657 /* ignore no autoconf addresses. */ 658 if ((it6->ia6_flags & IN6_IFF_AUTOCONF) == 0) 659 continue; 660 661 /* ignore autoconf addresses with different prefixes. */ 662 if (it6->ia6_ndpr == NULL || it6->ia6_ndpr != ia6->ia6_ndpr) 663 continue; 664 665 /* 666 * Now we are looking at an autoconf address with the same 667 * prefix as ours. If the address is temporary and is still 668 * preferred, do not create another one. It would be rare, but 669 * could happen, for example, when we resume a laptop PC after 670 * a long period. 671 */ 672 if ((it6->ia6_flags & IN6_IFF_TEMPORARY) != 0 && 673 !IFA6_IS_DEPRECATED(it6)) { 674 public_ifa6 = NULL; 675 break; 676 } 677 678 /* 679 * This is a public autoconf address that has the same prefix 680 * as ours. If it is preferred, keep it. We can't break the 681 * loop here, because there may be a still-preferred temporary 682 * address with the prefix. 683 */ 684 if (!IFA6_IS_DEPRECATED(it6)) 685 public_ifa6 = it6; 686 } 687 688 if (public_ifa6 != NULL) { 689 int e; 690 691 /* 692 * Random factor is introduced in the preferred lifetime, so 693 * we do not need additional delay (3rd arg to in6_tmpifadd). 694 */ 695 if ((e = in6_tmpifadd(public_ifa6, 0, 0)) != 0) { 696 log(LOG_NOTICE, "regen_tmpaddr: failed to create a new" 697 " tmp addr, errno=%d\n", e); 698 return -1; 699 } 700 return 0; 701 } 702 703 return -1; 704 } 705 706 /* 707 * Nuke neighbor cache/prefix/default router management table, right before 708 * ifp goes away. 709 */ 710 void 711 nd6_purge(struct ifnet *ifp) 712 { 713 struct llinfo_nd6 *ln, *nln; 714 struct nd_defrouter *dr, *ndr; 715 struct nd_prefix *pr, *npr; 716 717 /* 718 * Nuke default router list entries toward ifp. 719 * We defer removal of default router list entries that is installed 720 * in the routing table, in order to keep additional side effects as 721 * small as possible. 722 */ 723 for (dr = TAILQ_FIRST(&nd_defrouter); dr; dr = ndr) { 724 ndr = TAILQ_NEXT(dr, dr_entry); 725 if (dr->installed) 726 continue; 727 728 if (dr->ifp == ifp) 729 defrtrlist_del(dr); 730 } 731 for (dr = TAILQ_FIRST(&nd_defrouter); dr; dr = ndr) { 732 ndr = TAILQ_NEXT(dr, dr_entry); 733 if (!dr->installed) 734 continue; 735 736 if (dr->ifp == ifp) 737 defrtrlist_del(dr); 738 } 739 740 /* Nuke prefix list entries toward ifp */ 741 for (pr = LIST_FIRST(&nd_prefix); pr != NULL; pr = npr) { 742 npr = LIST_NEXT(pr, ndpr_entry); 743 if (pr->ndpr_ifp == ifp) { 744 /* 745 * Because if_detach() does *not* release prefixes 746 * while purging addresses the reference count will 747 * still be above zero. We therefore reset it to 748 * make sure that the prefix really gets purged. 749 */ 750 pr->ndpr_refcnt = 0; 751 /* 752 * Previously, pr->ndpr_addr is removed as well, 753 * but I strongly believe we don't have to do it. 754 * nd6_purge() is only called from in6_ifdetach(), 755 * which removes all the associated interface addresses 756 * by itself. 757 * (jinmei@kame.net 20010129) 758 */ 759 prelist_remove(pr); 760 } 761 } 762 763 /* cancel default outgoing interface setting */ 764 if (nd6_defifindex == ifp->if_index) 765 nd6_setdefaultiface(0); 766 767 if (!ip6_forwarding && ip6_accept_rtadv) { /* XXX: too restrictive? */ 768 /* refresh default router list */ 769 defrouter_select(); 770 } 771 772 /* 773 * Nuke neighbor cache entries for the ifp. 774 * Note that rt->rt_ifp may not be the same as ifp, 775 * due to KAME goto ours hack. See RTM_RESOLVE case in 776 * nd6_rtrequest(), and ip6_input(). 777 */ 778 ln = llinfo_nd6.ln_next; 779 while (ln && ln != &llinfo_nd6) { 780 struct rtentry *rt; 781 const struct sockaddr_dl *sdl; 782 783 nln = ln->ln_next; 784 rt = ln->ln_rt; 785 if (rt && rt->rt_gateway && 786 rt->rt_gateway->sa_family == AF_LINK) { 787 sdl = satocsdl(rt->rt_gateway); 788 if (sdl->sdl_index == ifp->if_index) 789 nln = nd6_free(rt, 0); 790 } 791 ln = nln; 792 } 793 } 794 795 struct rtentry * 796 nd6_lookup(const struct in6_addr *addr6, int create, struct ifnet *ifp) 797 { 798 struct rtentry *rt; 799 struct sockaddr_in6 sin6; 800 801 sockaddr_in6_init(&sin6, addr6, 0, 0, 0); 802 rt = rtalloc1((struct sockaddr *)&sin6, create); 803 if (rt && (rt->rt_flags & RTF_LLINFO) == 0) { 804 /* 805 * This is the case for the default route. 806 * If we want to create a neighbor cache for the address, we 807 * should free the route for the destination and allocate an 808 * interface route. 809 */ 810 if (create) { 811 RTFREE(rt); 812 rt = NULL; 813 } 814 } 815 if (rt != NULL) 816 ; 817 else if (create && ifp) { 818 int e; 819 820 /* 821 * If no route is available and create is set, 822 * we allocate a host route for the destination 823 * and treat it like an interface route. 824 * This hack is necessary for a neighbor which can't 825 * be covered by our own prefix. 826 */ 827 struct ifaddr *ifa = 828 ifaof_ifpforaddr((struct sockaddr *)&sin6, ifp); 829 if (ifa == NULL) 830 return NULL; 831 832 /* 833 * Create a new route. RTF_LLINFO is necessary 834 * to create a Neighbor Cache entry for the 835 * destination in nd6_rtrequest which will be 836 * called in rtrequest via ifa->ifa_rtrequest. 837 */ 838 if ((e = rtrequest(RTM_ADD, (const struct sockaddr *)&sin6, 839 ifa->ifa_addr, (const struct sockaddr *)&all1_sa, 840 (ifa->ifa_flags | RTF_HOST | RTF_LLINFO) & 841 ~RTF_CLONING, &rt)) != 0) { 842 #if 0 843 log(LOG_ERR, 844 "nd6_lookup: failed to add route for a " 845 "neighbor(%s), errno=%d\n", 846 ip6_sprintf(addr6), e); 847 #endif 848 return NULL; 849 } 850 if (rt == NULL) 851 return NULL; 852 if (rt->rt_llinfo) { 853 struct llinfo_nd6 *ln = 854 (struct llinfo_nd6 *)rt->rt_llinfo; 855 ln->ln_state = ND6_LLINFO_NOSTATE; 856 } 857 } else 858 return NULL; 859 rt->rt_refcnt--; 860 /* 861 * Validation for the entry. 862 * Note that the check for rt_llinfo is necessary because a cloned 863 * route from a parent route that has the L flag (e.g. the default 864 * route to a p2p interface) may have the flag, too, while the 865 * destination is not actually a neighbor. 866 * XXX: we can't use rt->rt_ifp to check for the interface, since 867 * it might be the loopback interface if the entry is for our 868 * own address on a non-loopback interface. Instead, we should 869 * use rt->rt_ifa->ifa_ifp, which would specify the REAL 870 * interface. 871 * Note also that ifa_ifp and ifp may differ when we connect two 872 * interfaces to a same link, install a link prefix to an interface, 873 * and try to install a neighbor cache on an interface that does not 874 * have a route to the prefix. 875 */ 876 if ((rt->rt_flags & RTF_GATEWAY) || (rt->rt_flags & RTF_LLINFO) == 0 || 877 rt->rt_gateway->sa_family != AF_LINK || rt->rt_llinfo == NULL || 878 (ifp && rt->rt_ifa->ifa_ifp != ifp)) { 879 if (create) { 880 nd6log((LOG_DEBUG, 881 "nd6_lookup: failed to lookup %s (if = %s)\n", 882 ip6_sprintf(addr6), 883 ifp ? if_name(ifp) : "unspec")); 884 } 885 return NULL; 886 } 887 return rt; 888 } 889 890 /* 891 * Detect if a given IPv6 address identifies a neighbor on a given link. 892 * XXX: should take care of the destination of a p2p link? 893 */ 894 int 895 nd6_is_addr_neighbor(const struct sockaddr_in6 *addr, struct ifnet *ifp) 896 { 897 struct nd_prefix *pr; 898 899 /* 900 * A link-local address is always a neighbor. 901 * XXX: a link does not necessarily specify a single interface. 902 */ 903 if (IN6_IS_ADDR_LINKLOCAL(&addr->sin6_addr)) { 904 struct sockaddr_in6 sin6_copy; 905 u_int32_t zone; 906 907 /* 908 * We need sin6_copy since sa6_recoverscope() may modify the 909 * content (XXX). 910 */ 911 sin6_copy = *addr; 912 if (sa6_recoverscope(&sin6_copy)) 913 return 0; /* XXX: should be impossible */ 914 if (in6_setscope(&sin6_copy.sin6_addr, ifp, &zone)) 915 return 0; 916 if (sin6_copy.sin6_scope_id == zone) 917 return 1; 918 else 919 return 0; 920 } 921 922 /* 923 * If the address matches one of our on-link prefixes, it should be a 924 * neighbor. 925 */ 926 LIST_FOREACH(pr, &nd_prefix, ndpr_entry) { 927 if (pr->ndpr_ifp != ifp) 928 continue; 929 930 if (!(pr->ndpr_stateflags & NDPRF_ONLINK)) 931 continue; 932 933 if (IN6_ARE_MASKED_ADDR_EQUAL(&pr->ndpr_prefix.sin6_addr, 934 &addr->sin6_addr, &pr->ndpr_mask)) 935 return 1; 936 } 937 938 /* 939 * If the default router list is empty, all addresses are regarded 940 * as on-link, and thus, as a neighbor. 941 * XXX: we restrict the condition to hosts, because routers usually do 942 * not have the "default router list". 943 */ 944 if (!ip6_forwarding && TAILQ_FIRST(&nd_defrouter) == NULL && 945 nd6_defifindex == ifp->if_index) { 946 return 1; 947 } 948 949 /* 950 * Even if the address matches none of our addresses, it might be 951 * in the neighbor cache. 952 */ 953 if (nd6_lookup(&addr->sin6_addr, 0, ifp) != NULL) 954 return 1; 955 956 return 0; 957 } 958 959 /* 960 * Free an nd6 llinfo entry. 961 * Since the function would cause significant changes in the kernel, DO NOT 962 * make it global, unless you have a strong reason for the change, and are sure 963 * that the change is safe. 964 */ 965 static struct llinfo_nd6 * 966 nd6_free(struct rtentry *rt, int gc) 967 { 968 struct llinfo_nd6 *ln = (struct llinfo_nd6 *)rt->rt_llinfo, *next; 969 struct in6_addr in6 = satocsin6(rt_getkey(rt))->sin6_addr; 970 struct nd_defrouter *dr; 971 972 /* 973 * we used to have pfctlinput(PRC_HOSTDEAD) here. 974 * even though it is not harmful, it was not really necessary. 975 */ 976 977 /* cancel timer */ 978 nd6_llinfo_settimer(ln, -1); 979 980 if (!ip6_forwarding) { 981 int s; 982 s = splsoftnet(); 983 dr = defrouter_lookup(&satocsin6(rt_getkey(rt))->sin6_addr, 984 rt->rt_ifp); 985 986 if (dr != NULL && dr->expire && 987 ln->ln_state == ND6_LLINFO_STALE && gc) { 988 /* 989 * If the reason for the deletion is just garbage 990 * collection, and the neighbor is an active default 991 * router, do not delete it. Instead, reset the GC 992 * timer using the router's lifetime. 993 * Simply deleting the entry would affect default 994 * router selection, which is not necessarily a good 995 * thing, especially when we're using router preference 996 * values. 997 * XXX: the check for ln_state would be redundant, 998 * but we intentionally keep it just in case. 999 */ 1000 if (dr->expire > time_second) 1001 nd6_llinfo_settimer(ln, 1002 (dr->expire - time_second) * hz); 1003 else 1004 nd6_llinfo_settimer(ln, (long)nd6_gctimer * hz); 1005 splx(s); 1006 return ln->ln_next; 1007 } 1008 1009 if (ln->ln_router || dr) { 1010 /* 1011 * rt6_flush must be called whether or not the neighbor 1012 * is in the Default Router List. 1013 * See a corresponding comment in nd6_na_input(). 1014 */ 1015 rt6_flush(&in6, rt->rt_ifp); 1016 } 1017 1018 if (dr) { 1019 /* 1020 * Unreachablity of a router might affect the default 1021 * router selection and on-link detection of advertised 1022 * prefixes. 1023 */ 1024 1025 /* 1026 * Temporarily fake the state to choose a new default 1027 * router and to perform on-link determination of 1028 * prefixes correctly. 1029 * Below the state will be set correctly, 1030 * or the entry itself will be deleted. 1031 */ 1032 ln->ln_state = ND6_LLINFO_INCOMPLETE; 1033 1034 /* 1035 * Since defrouter_select() does not affect the 1036 * on-link determination and MIP6 needs the check 1037 * before the default router selection, we perform 1038 * the check now. 1039 */ 1040 pfxlist_onlink_check(); 1041 1042 /* 1043 * refresh default router list 1044 */ 1045 defrouter_select(); 1046 } 1047 splx(s); 1048 } 1049 1050 /* 1051 * Before deleting the entry, remember the next entry as the 1052 * return value. We need this because pfxlist_onlink_check() above 1053 * might have freed other entries (particularly the old next entry) as 1054 * a side effect (XXX). 1055 */ 1056 next = ln->ln_next; 1057 1058 /* 1059 * Detach the route from the routing tree and the list of neighbor 1060 * caches, and disable the route entry not to be used in already 1061 * cached routes. 1062 */ 1063 rtrequest(RTM_DELETE, rt_getkey(rt), (struct sockaddr *)0, 1064 rt_mask(rt), 0, (struct rtentry **)0); 1065 1066 return next; 1067 } 1068 1069 /* 1070 * Upper-layer reachability hint for Neighbor Unreachability Detection. 1071 * 1072 * XXX cost-effective methods? 1073 */ 1074 void 1075 nd6_nud_hint(struct rtentry *rt, struct in6_addr *dst6, int force) 1076 { 1077 struct llinfo_nd6 *ln; 1078 1079 /* 1080 * If the caller specified "rt", use that. Otherwise, resolve the 1081 * routing table by supplied "dst6". 1082 */ 1083 if (rt == NULL) { 1084 if (dst6 == NULL) 1085 return; 1086 if ((rt = nd6_lookup(dst6, 0, NULL)) == NULL) 1087 return; 1088 } 1089 1090 if ((rt->rt_flags & RTF_GATEWAY) != 0 || 1091 (rt->rt_flags & RTF_LLINFO) == 0 || 1092 !rt->rt_llinfo || !rt->rt_gateway || 1093 rt->rt_gateway->sa_family != AF_LINK) { 1094 /* This is not a host route. */ 1095 return; 1096 } 1097 1098 ln = (struct llinfo_nd6 *)rt->rt_llinfo; 1099 if (ln->ln_state < ND6_LLINFO_REACHABLE) 1100 return; 1101 1102 /* 1103 * if we get upper-layer reachability confirmation many times, 1104 * it is possible we have false information. 1105 */ 1106 if (!force) { 1107 ln->ln_byhint++; 1108 if (ln->ln_byhint > nd6_maxnudhint) 1109 return; 1110 } 1111 1112 ln->ln_state = ND6_LLINFO_REACHABLE; 1113 if (!ND6_LLINFO_PERMANENT(ln)) { 1114 nd6_llinfo_settimer(ln, 1115 (long)ND_IFINFO(rt->rt_ifp)->reachable * hz); 1116 } 1117 } 1118 1119 void 1120 nd6_rtrequest(int req, struct rtentry *rt, struct rt_addrinfo *info) 1121 { 1122 struct sockaddr *gate = rt->rt_gateway; 1123 struct llinfo_nd6 *ln = (struct llinfo_nd6 *)rt->rt_llinfo; 1124 struct ifnet *ifp = rt->rt_ifp; 1125 uint8_t namelen = strlen(ifp->if_xname), addrlen = ifp->if_addrlen; 1126 struct ifaddr *ifa; 1127 1128 RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key); 1129 1130 if ((rt->rt_flags & RTF_GATEWAY) != 0) 1131 return; 1132 1133 if (nd6_need_cache(ifp) == 0 && (rt->rt_flags & RTF_HOST) == 0) { 1134 RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key); 1135 /* 1136 * This is probably an interface direct route for a link 1137 * which does not need neighbor caches (e.g. fe80::%lo0/64). 1138 * We do not need special treatment below for such a route. 1139 * Moreover, the RTF_LLINFO flag which would be set below 1140 * would annoy the ndp(8) command. 1141 */ 1142 return; 1143 } 1144 1145 if (req == RTM_RESOLVE && 1146 (nd6_need_cache(ifp) == 0 || /* stf case */ 1147 !nd6_is_addr_neighbor(satocsin6(rt_getkey(rt)), ifp))) { 1148 RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key); 1149 /* 1150 * FreeBSD and BSD/OS often make a cloned host route based 1151 * on a less-specific route (e.g. the default route). 1152 * If the less specific route does not have a "gateway" 1153 * (this is the case when the route just goes to a p2p or an 1154 * stf interface), we'll mistakenly make a neighbor cache for 1155 * the host route, and will see strange neighbor solicitation 1156 * for the corresponding destination. In order to avoid the 1157 * confusion, we check if the destination of the route is 1158 * a neighbor in terms of neighbor discovery, and stop the 1159 * process if not. Additionally, we remove the LLINFO flag 1160 * so that ndp(8) will not try to get the neighbor information 1161 * of the destination. 1162 */ 1163 rt->rt_flags &= ~RTF_LLINFO; 1164 return; 1165 } 1166 1167 switch (req) { 1168 case RTM_ADD: 1169 RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key); 1170 /* 1171 * There is no backward compatibility :) 1172 * 1173 * if ((rt->rt_flags & RTF_HOST) == 0 && 1174 * SIN(rt_mask(rt))->sin_addr.s_addr != 0xffffffff) 1175 * rt->rt_flags |= RTF_CLONING; 1176 */ 1177 if ((rt->rt_flags & RTF_CLONING) || 1178 ((rt->rt_flags & RTF_LLINFO) && !ln)) { 1179 union { 1180 struct sockaddr sa; 1181 struct sockaddr_dl sdl; 1182 struct sockaddr_storage ss; 1183 } u; 1184 /* 1185 * Case 1: This route should come from a route to 1186 * interface (RTF_CLONING case) or the route should be 1187 * treated as on-link but is currently not 1188 * (RTF_LLINFO && !ln case). 1189 */ 1190 sockaddr_dl_init(&u.sdl, sizeof(u.ss), 1191 ifp->if_index, ifp->if_type, 1192 NULL, namelen, NULL, addrlen); 1193 rt_setgate(rt, &u.sa); 1194 gate = rt->rt_gateway; 1195 RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key); 1196 if (ln) 1197 nd6_llinfo_settimer(ln, 0); 1198 RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key); 1199 if ((rt->rt_flags & RTF_CLONING) != 0) 1200 break; 1201 } 1202 RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key); 1203 /* 1204 * In IPv4 code, we try to annonuce new RTF_ANNOUNCE entry here. 1205 * We don't do that here since llinfo is not ready yet. 1206 * 1207 * There are also couple of other things to be discussed: 1208 * - unsolicited NA code needs improvement beforehand 1209 * - RFC2461 says we MAY send multicast unsolicited NA 1210 * (7.2.6 paragraph 4), however, it also says that we 1211 * SHOULD provide a mechanism to prevent multicast NA storm. 1212 * we don't have anything like it right now. 1213 * note that the mechanism needs a mutual agreement 1214 * between proxies, which means that we need to implement 1215 * a new protocol, or a new kludge. 1216 * - from RFC2461 6.2.4, host MUST NOT send an unsolicited NA. 1217 * we need to check ip6forwarding before sending it. 1218 * (or should we allow proxy ND configuration only for 1219 * routers? there's no mention about proxy ND from hosts) 1220 */ 1221 #if 0 1222 /* XXX it does not work */ 1223 if (rt->rt_flags & RTF_ANNOUNCE) 1224 nd6_na_output(ifp, 1225 &satocsin6(rt_getkey(rt))->sin6_addr, 1226 &satocsin6(rt_getkey(rt))->sin6_addr, 1227 ip6_forwarding ? ND_NA_FLAG_ROUTER : 0, 1228 1, NULL); 1229 #endif 1230 /* FALLTHROUGH */ 1231 case RTM_RESOLVE: 1232 if ((ifp->if_flags & (IFF_POINTOPOINT | IFF_LOOPBACK)) == 0) { 1233 RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key); 1234 /* 1235 * Address resolution isn't necessary for a point to 1236 * point link, so we can skip this test for a p2p link. 1237 */ 1238 if (gate->sa_family != AF_LINK || 1239 gate->sa_len < 1240 sockaddr_dl_measure(namelen, addrlen)) { 1241 log(LOG_DEBUG, 1242 "nd6_rtrequest: bad gateway value: %s\n", 1243 if_name(ifp)); 1244 break; 1245 } 1246 satosdl(gate)->sdl_type = ifp->if_type; 1247 satosdl(gate)->sdl_index = ifp->if_index; 1248 RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key); 1249 } 1250 if (ln != NULL) 1251 break; /* This happens on a route change */ 1252 RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key); 1253 /* 1254 * Case 2: This route may come from cloning, or a manual route 1255 * add with a LL address. 1256 */ 1257 R_Malloc(ln, struct llinfo_nd6 *, sizeof(*ln)); 1258 rt->rt_llinfo = (void *)ln; 1259 RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key); 1260 if (ln == NULL) { 1261 log(LOG_DEBUG, "nd6_rtrequest: malloc failed\n"); 1262 break; 1263 } 1264 RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key); 1265 nd6_inuse++; 1266 nd6_allocated++; 1267 bzero(ln, sizeof(*ln)); 1268 ln->ln_rt = rt; 1269 callout_init(&ln->ln_timer_ch, CALLOUT_MPSAFE); 1270 /* this is required for "ndp" command. - shin */ 1271 if (req == RTM_ADD) { 1272 /* 1273 * gate should have some valid AF_LINK entry, 1274 * and ln->ln_expire should have some lifetime 1275 * which is specified by ndp command. 1276 */ 1277 ln->ln_state = ND6_LLINFO_REACHABLE; 1278 ln->ln_byhint = 0; 1279 } else { 1280 /* 1281 * When req == RTM_RESOLVE, rt is created and 1282 * initialized in rtrequest(), so rt_expire is 0. 1283 */ 1284 ln->ln_state = ND6_LLINFO_NOSTATE; 1285 nd6_llinfo_settimer(ln, 0); 1286 } 1287 RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key); 1288 rt->rt_flags |= RTF_LLINFO; 1289 ln->ln_next = llinfo_nd6.ln_next; 1290 llinfo_nd6.ln_next = ln; 1291 ln->ln_prev = &llinfo_nd6; 1292 ln->ln_next->ln_prev = ln; 1293 1294 RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key); 1295 /* 1296 * check if rt_getkey(rt) is an address assigned 1297 * to the interface. 1298 */ 1299 ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp, 1300 &satocsin6(rt_getkey(rt))->sin6_addr); 1301 RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key); 1302 if (ifa) { 1303 const void *mac; 1304 nd6_llinfo_settimer(ln, -1); 1305 ln->ln_state = ND6_LLINFO_REACHABLE; 1306 ln->ln_byhint = 0; 1307 if ((mac = nd6_ifptomac(ifp)) != NULL) { 1308 /* XXX check for error */ 1309 (void)sockaddr_dl_setaddr(satosdl(gate), 1310 gate->sa_len, mac, ifp->if_addrlen); 1311 } 1312 if (nd6_useloopback) { 1313 ifp = rt->rt_ifp = lo0ifp; /* XXX */ 1314 /* 1315 * Make sure rt_ifa be equal to the ifaddr 1316 * corresponding to the address. 1317 * We need this because when we refer 1318 * rt_ifa->ia6_flags in ip6_input, we assume 1319 * that the rt_ifa points to the address instead 1320 * of the loopback address. 1321 */ 1322 if (ifa != rt->rt_ifa) 1323 rt_replace_ifa(rt, ifa); 1324 rt->rt_flags &= ~RTF_CLONED; 1325 } 1326 } else if (rt->rt_flags & RTF_ANNOUNCE) { 1327 nd6_llinfo_settimer(ln, -1); 1328 ln->ln_state = ND6_LLINFO_REACHABLE; 1329 ln->ln_byhint = 0; 1330 1331 /* join solicited node multicast for proxy ND */ 1332 if (ifp->if_flags & IFF_MULTICAST) { 1333 struct in6_addr llsol; 1334 int error; 1335 1336 llsol = satocsin6(rt_getkey(rt))->sin6_addr; 1337 llsol.s6_addr32[0] = htonl(0xff020000); 1338 llsol.s6_addr32[1] = 0; 1339 llsol.s6_addr32[2] = htonl(1); 1340 llsol.s6_addr8[12] = 0xff; 1341 if (in6_setscope(&llsol, ifp, NULL)) 1342 break; 1343 if (!in6_addmulti(&llsol, ifp, &error, 0)) { 1344 nd6log((LOG_ERR, "%s: failed to join " 1345 "%s (errno=%d)\n", if_name(ifp), 1346 ip6_sprintf(&llsol), error)); 1347 } 1348 } 1349 } 1350 break; 1351 1352 case RTM_DELETE: 1353 if (ln == NULL) 1354 break; 1355 /* leave from solicited node multicast for proxy ND */ 1356 if ((rt->rt_flags & RTF_ANNOUNCE) != 0 && 1357 (ifp->if_flags & IFF_MULTICAST) != 0) { 1358 struct in6_addr llsol; 1359 struct in6_multi *in6m; 1360 1361 llsol = satocsin6(rt_getkey(rt))->sin6_addr; 1362 llsol.s6_addr32[0] = htonl(0xff020000); 1363 llsol.s6_addr32[1] = 0; 1364 llsol.s6_addr32[2] = htonl(1); 1365 llsol.s6_addr8[12] = 0xff; 1366 if (in6_setscope(&llsol, ifp, NULL) == 0) { 1367 IN6_LOOKUP_MULTI(llsol, ifp, in6m); 1368 if (in6m) 1369 in6_delmulti(in6m); 1370 } 1371 } 1372 nd6_inuse--; 1373 ln->ln_next->ln_prev = ln->ln_prev; 1374 ln->ln_prev->ln_next = ln->ln_next; 1375 ln->ln_prev = NULL; 1376 nd6_llinfo_settimer(ln, -1); 1377 rt->rt_llinfo = 0; 1378 rt->rt_flags &= ~RTF_LLINFO; 1379 clear_llinfo_pqueue(ln); 1380 Free((void *)ln); 1381 } 1382 } 1383 1384 int 1385 nd6_ioctl(u_long cmd, void *data, struct ifnet *ifp) 1386 { 1387 struct in6_drlist *drl = (struct in6_drlist *)data; 1388 struct in6_oprlist *oprl = (struct in6_oprlist *)data; 1389 struct in6_ndireq *ndi = (struct in6_ndireq *)data; 1390 struct in6_nbrinfo *nbi = (struct in6_nbrinfo *)data; 1391 struct in6_ndifreq *ndif = (struct in6_ndifreq *)data; 1392 struct nd_defrouter *dr; 1393 struct nd_prefix *pr; 1394 struct rtentry *rt; 1395 int i = 0, error = 0; 1396 int s; 1397 1398 switch (cmd) { 1399 case SIOCGDRLST_IN6: 1400 /* 1401 * obsolete API, use sysctl under net.inet6.icmp6 1402 */ 1403 bzero(drl, sizeof(*drl)); 1404 s = splsoftnet(); 1405 TAILQ_FOREACH(dr, &nd_defrouter, dr_entry) { 1406 if (i >= DRLSTSIZ) 1407 break; 1408 drl->defrouter[i].rtaddr = dr->rtaddr; 1409 in6_clearscope(&drl->defrouter[i].rtaddr); 1410 1411 drl->defrouter[i].flags = dr->flags; 1412 drl->defrouter[i].rtlifetime = dr->rtlifetime; 1413 drl->defrouter[i].expire = dr->expire; 1414 drl->defrouter[i].if_index = dr->ifp->if_index; 1415 i++; 1416 } 1417 splx(s); 1418 break; 1419 case SIOCGPRLST_IN6: 1420 /* 1421 * obsolete API, use sysctl under net.inet6.icmp6 1422 * 1423 * XXX the structure in6_prlist was changed in backward- 1424 * incompatible manner. in6_oprlist is used for SIOCGPRLST_IN6, 1425 * in6_prlist is used for nd6_sysctl() - fill_prlist(). 1426 */ 1427 /* 1428 * XXX meaning of fields, especialy "raflags", is very 1429 * differnet between RA prefix list and RR/static prefix list. 1430 * how about separating ioctls into two? 1431 */ 1432 bzero(oprl, sizeof(*oprl)); 1433 s = splsoftnet(); 1434 LIST_FOREACH(pr, &nd_prefix, ndpr_entry) { 1435 struct nd_pfxrouter *pfr; 1436 int j; 1437 1438 if (i >= PRLSTSIZ) 1439 break; 1440 oprl->prefix[i].prefix = pr->ndpr_prefix.sin6_addr; 1441 oprl->prefix[i].raflags = pr->ndpr_raf; 1442 oprl->prefix[i].prefixlen = pr->ndpr_plen; 1443 oprl->prefix[i].vltime = pr->ndpr_vltime; 1444 oprl->prefix[i].pltime = pr->ndpr_pltime; 1445 oprl->prefix[i].if_index = pr->ndpr_ifp->if_index; 1446 if (pr->ndpr_vltime == ND6_INFINITE_LIFETIME) 1447 oprl->prefix[i].expire = 0; 1448 else { 1449 time_t maxexpire; 1450 1451 /* XXX: we assume time_t is signed. */ 1452 maxexpire = (-1) & 1453 ~((time_t)1 << 1454 ((sizeof(maxexpire) * 8) - 1)); 1455 if (pr->ndpr_vltime < 1456 maxexpire - pr->ndpr_lastupdate) { 1457 oprl->prefix[i].expire = 1458 pr->ndpr_lastupdate + 1459 pr->ndpr_vltime; 1460 } else 1461 oprl->prefix[i].expire = maxexpire; 1462 } 1463 1464 j = 0; 1465 LIST_FOREACH(pfr, &pr->ndpr_advrtrs, pfr_entry) { 1466 if (j < DRLSTSIZ) { 1467 #define RTRADDR oprl->prefix[i].advrtr[j] 1468 RTRADDR = pfr->router->rtaddr; 1469 in6_clearscope(&RTRADDR); 1470 #undef RTRADDR 1471 } 1472 j++; 1473 } 1474 oprl->prefix[i].advrtrs = j; 1475 oprl->prefix[i].origin = PR_ORIG_RA; 1476 1477 i++; 1478 } 1479 splx(s); 1480 1481 break; 1482 case OSIOCGIFINFO_IN6: 1483 #define ND ndi->ndi 1484 /* XXX: old ndp(8) assumes a positive value for linkmtu. */ 1485 memset(&ND, 0, sizeof(ND)); 1486 ND.linkmtu = IN6_LINKMTU(ifp); 1487 ND.maxmtu = ND_IFINFO(ifp)->maxmtu; 1488 ND.basereachable = ND_IFINFO(ifp)->basereachable; 1489 ND.reachable = ND_IFINFO(ifp)->reachable; 1490 ND.retrans = ND_IFINFO(ifp)->retrans; 1491 ND.flags = ND_IFINFO(ifp)->flags; 1492 ND.recalctm = ND_IFINFO(ifp)->recalctm; 1493 ND.chlim = ND_IFINFO(ifp)->chlim; 1494 break; 1495 case SIOCGIFINFO_IN6: 1496 ND = *ND_IFINFO(ifp); 1497 break; 1498 case SIOCSIFINFO_IN6: 1499 /* 1500 * used to change host variables from userland. 1501 * intented for a use on router to reflect RA configurations. 1502 */ 1503 /* 0 means 'unspecified' */ 1504 if (ND.linkmtu != 0) { 1505 if (ND.linkmtu < IPV6_MMTU || 1506 ND.linkmtu > IN6_LINKMTU(ifp)) { 1507 error = EINVAL; 1508 break; 1509 } 1510 ND_IFINFO(ifp)->linkmtu = ND.linkmtu; 1511 } 1512 1513 if (ND.basereachable != 0) { 1514 int obasereachable = ND_IFINFO(ifp)->basereachable; 1515 1516 ND_IFINFO(ifp)->basereachable = ND.basereachable; 1517 if (ND.basereachable != obasereachable) 1518 ND_IFINFO(ifp)->reachable = 1519 ND_COMPUTE_RTIME(ND.basereachable); 1520 } 1521 if (ND.retrans != 0) 1522 ND_IFINFO(ifp)->retrans = ND.retrans; 1523 if (ND.chlim != 0) 1524 ND_IFINFO(ifp)->chlim = ND.chlim; 1525 /* FALLTHROUGH */ 1526 case SIOCSIFINFO_FLAGS: 1527 ND_IFINFO(ifp)->flags = ND.flags; 1528 break; 1529 #undef ND 1530 case SIOCSNDFLUSH_IN6: /* XXX: the ioctl name is confusing... */ 1531 /* sync kernel routing table with the default router list */ 1532 defrouter_reset(); 1533 defrouter_select(); 1534 break; 1535 case SIOCSPFXFLUSH_IN6: 1536 { 1537 /* flush all the prefix advertised by routers */ 1538 struct nd_prefix *pfx, *next; 1539 1540 s = splsoftnet(); 1541 for (pfx = LIST_FIRST(&nd_prefix); pfx; pfx = next) { 1542 struct in6_ifaddr *ia, *ia_next; 1543 1544 next = LIST_NEXT(pfx, ndpr_entry); 1545 1546 if (IN6_IS_ADDR_LINKLOCAL(&pfx->ndpr_prefix.sin6_addr)) 1547 continue; /* XXX */ 1548 1549 /* do we really have to remove addresses as well? */ 1550 for (ia = in6_ifaddr; ia; ia = ia_next) { 1551 /* ia might be removed. keep the next ptr. */ 1552 ia_next = ia->ia_next; 1553 1554 if ((ia->ia6_flags & IN6_IFF_AUTOCONF) == 0) 1555 continue; 1556 1557 if (ia->ia6_ndpr == pfx) 1558 in6_purgeaddr(&ia->ia_ifa); 1559 } 1560 prelist_remove(pfx); 1561 } 1562 splx(s); 1563 break; 1564 } 1565 case SIOCSRTRFLUSH_IN6: 1566 { 1567 /* flush all the default routers */ 1568 struct nd_defrouter *drtr, *next; 1569 1570 s = splsoftnet(); 1571 defrouter_reset(); 1572 for (drtr = TAILQ_FIRST(&nd_defrouter); drtr; drtr = next) { 1573 next = TAILQ_NEXT(drtr, dr_entry); 1574 defrtrlist_del(drtr); 1575 } 1576 defrouter_select(); 1577 splx(s); 1578 break; 1579 } 1580 case SIOCGNBRINFO_IN6: 1581 { 1582 struct llinfo_nd6 *ln; 1583 struct in6_addr nb_addr = nbi->addr; /* make local for safety */ 1584 1585 if ((error = in6_setscope(&nb_addr, ifp, NULL)) != 0) 1586 return error; 1587 1588 s = splsoftnet(); 1589 if ((rt = nd6_lookup(&nb_addr, 0, ifp)) == NULL || 1590 (ln = (struct llinfo_nd6 *)rt->rt_llinfo) == NULL) { 1591 error = EINVAL; 1592 splx(s); 1593 break; 1594 } 1595 nbi->state = ln->ln_state; 1596 nbi->asked = ln->ln_asked; 1597 nbi->isrouter = ln->ln_router; 1598 nbi->expire = ln->ln_expire; 1599 splx(s); 1600 1601 break; 1602 } 1603 case SIOCGDEFIFACE_IN6: /* XXX: should be implemented as a sysctl? */ 1604 ndif->ifindex = nd6_defifindex; 1605 break; 1606 case SIOCSDEFIFACE_IN6: /* XXX: should be implemented as a sysctl? */ 1607 return nd6_setdefaultiface(ndif->ifindex); 1608 } 1609 return error; 1610 } 1611 1612 void 1613 nd6_llinfo_release_pkts(struct llinfo_nd6 *ln, struct ifnet *ifp, 1614 struct rtentry *rt) 1615 { 1616 struct mbuf *m_hold, *m_hold_next; 1617 1618 for (m_hold = ln->ln_hold, ln->ln_hold = NULL; 1619 m_hold != NULL; 1620 m_hold = m_hold_next) { 1621 m_hold_next = m_hold->m_nextpkt; 1622 m_hold->m_nextpkt = NULL; 1623 1624 /* 1625 * we assume ifp is not a p2p here, so 1626 * just set the 2nd argument as the 1627 * 1st one. 1628 */ 1629 nd6_output(ifp, ifp, m_hold, satocsin6(rt_getkey(rt)), rt); 1630 } 1631 } 1632 1633 /* 1634 * Create neighbor cache entry and cache link-layer address, 1635 * on reception of inbound ND6 packets. (RS/RA/NS/redirect) 1636 */ 1637 struct rtentry * 1638 nd6_cache_lladdr( 1639 struct ifnet *ifp, 1640 struct in6_addr *from, 1641 char *lladdr, 1642 int lladdrlen, 1643 int type, /* ICMP6 type */ 1644 int code /* type dependent information */ 1645 ) 1646 { 1647 struct rtentry *rt = NULL; 1648 struct llinfo_nd6 *ln = NULL; 1649 int is_newentry; 1650 struct sockaddr_dl *sdl = NULL; 1651 int do_update; 1652 int olladdr; 1653 int llchange; 1654 int newstate = 0; 1655 1656 if (ifp == NULL) 1657 panic("ifp == NULL in nd6_cache_lladdr"); 1658 if (from == NULL) 1659 panic("from == NULL in nd6_cache_lladdr"); 1660 1661 /* nothing must be updated for unspecified address */ 1662 if (IN6_IS_ADDR_UNSPECIFIED(from)) 1663 return NULL; 1664 1665 /* 1666 * Validation about ifp->if_addrlen and lladdrlen must be done in 1667 * the caller. 1668 * 1669 * XXX If the link does not have link-layer adderss, what should 1670 * we do? (ifp->if_addrlen == 0) 1671 * Spec says nothing in sections for RA, RS and NA. There's small 1672 * description on it in NS section (RFC 2461 7.2.3). 1673 */ 1674 1675 rt = nd6_lookup(from, 0, ifp); 1676 if (rt == NULL) { 1677 #if 0 1678 /* nothing must be done if there's no lladdr */ 1679 if (!lladdr || !lladdrlen) 1680 return NULL; 1681 #endif 1682 1683 rt = nd6_lookup(from, 1, ifp); 1684 is_newentry = 1; 1685 } else { 1686 /* do nothing if static ndp is set */ 1687 if (rt->rt_flags & RTF_STATIC) 1688 return NULL; 1689 is_newentry = 0; 1690 } 1691 1692 if (rt == NULL) 1693 return NULL; 1694 if ((rt->rt_flags & (RTF_GATEWAY | RTF_LLINFO)) != RTF_LLINFO) { 1695 fail: 1696 (void)nd6_free(rt, 0); 1697 return NULL; 1698 } 1699 ln = (struct llinfo_nd6 *)rt->rt_llinfo; 1700 if (ln == NULL) 1701 goto fail; 1702 if (rt->rt_gateway == NULL) 1703 goto fail; 1704 if (rt->rt_gateway->sa_family != AF_LINK) 1705 goto fail; 1706 sdl = satosdl(rt->rt_gateway); 1707 1708 olladdr = (sdl->sdl_alen) ? 1 : 0; 1709 if (olladdr && lladdr) { 1710 if (memcmp(lladdr, CLLADDR(sdl), ifp->if_addrlen)) 1711 llchange = 1; 1712 else 1713 llchange = 0; 1714 } else 1715 llchange = 0; 1716 1717 /* 1718 * newentry olladdr lladdr llchange (*=record) 1719 * 0 n n -- (1) 1720 * 0 y n -- (2) 1721 * 0 n y -- (3) * STALE 1722 * 0 y y n (4) * 1723 * 0 y y y (5) * STALE 1724 * 1 -- n -- (6) NOSTATE(= PASSIVE) 1725 * 1 -- y -- (7) * STALE 1726 */ 1727 1728 if (lladdr) { /* (3-5) and (7) */ 1729 /* 1730 * Record source link-layer address 1731 * XXX is it dependent to ifp->if_type? 1732 */ 1733 /* XXX check for error */ 1734 (void)sockaddr_dl_setaddr(sdl, sdl->sdl_len, lladdr, 1735 ifp->if_addrlen); 1736 } 1737 1738 if (!is_newentry) { 1739 if ((!olladdr && lladdr) || /* (3) */ 1740 (olladdr && lladdr && llchange)) { /* (5) */ 1741 do_update = 1; 1742 newstate = ND6_LLINFO_STALE; 1743 } else /* (1-2,4) */ 1744 do_update = 0; 1745 } else { 1746 do_update = 1; 1747 if (lladdr == NULL) /* (6) */ 1748 newstate = ND6_LLINFO_NOSTATE; 1749 else /* (7) */ 1750 newstate = ND6_LLINFO_STALE; 1751 } 1752 1753 if (do_update) { 1754 /* 1755 * Update the state of the neighbor cache. 1756 */ 1757 ln->ln_state = newstate; 1758 1759 if (ln->ln_state == ND6_LLINFO_STALE) { 1760 /* 1761 * XXX: since nd6_output() below will cause 1762 * state tansition to DELAY and reset the timer, 1763 * we must set the timer now, although it is actually 1764 * meaningless. 1765 */ 1766 nd6_llinfo_settimer(ln, (long)nd6_gctimer * hz); 1767 1768 nd6_llinfo_release_pkts(ln, ifp, rt); 1769 } else if (ln->ln_state == ND6_LLINFO_INCOMPLETE) { 1770 /* probe right away */ 1771 nd6_llinfo_settimer((void *)ln, 0); 1772 } 1773 } 1774 1775 /* 1776 * ICMP6 type dependent behavior. 1777 * 1778 * NS: clear IsRouter if new entry 1779 * RS: clear IsRouter 1780 * RA: set IsRouter if there's lladdr 1781 * redir: clear IsRouter if new entry 1782 * 1783 * RA case, (1): 1784 * The spec says that we must set IsRouter in the following cases: 1785 * - If lladdr exist, set IsRouter. This means (1-5). 1786 * - If it is old entry (!newentry), set IsRouter. This means (7). 1787 * So, based on the spec, in (1-5) and (7) cases we must set IsRouter. 1788 * A quetion arises for (1) case. (1) case has no lladdr in the 1789 * neighbor cache, this is similar to (6). 1790 * This case is rare but we figured that we MUST NOT set IsRouter. 1791 * 1792 * newentry olladdr lladdr llchange NS RS RA redir 1793 * D R 1794 * 0 n n -- (1) c ? s 1795 * 0 y n -- (2) c s s 1796 * 0 n y -- (3) c s s 1797 * 0 y y n (4) c s s 1798 * 0 y y y (5) c s s 1799 * 1 -- n -- (6) c c c s 1800 * 1 -- y -- (7) c c s c s 1801 * 1802 * (c=clear s=set) 1803 */ 1804 switch (type & 0xff) { 1805 case ND_NEIGHBOR_SOLICIT: 1806 /* 1807 * New entry must have is_router flag cleared. 1808 */ 1809 if (is_newentry) /* (6-7) */ 1810 ln->ln_router = 0; 1811 break; 1812 case ND_REDIRECT: 1813 /* 1814 * If the icmp is a redirect to a better router, always set the 1815 * is_router flag. Otherwise, if the entry is newly created, 1816 * clear the flag. [RFC 2461, sec 8.3] 1817 */ 1818 if (code == ND_REDIRECT_ROUTER) 1819 ln->ln_router = 1; 1820 else if (is_newentry) /* (6-7) */ 1821 ln->ln_router = 0; 1822 break; 1823 case ND_ROUTER_SOLICIT: 1824 /* 1825 * is_router flag must always be cleared. 1826 */ 1827 ln->ln_router = 0; 1828 break; 1829 case ND_ROUTER_ADVERT: 1830 /* 1831 * Mark an entry with lladdr as a router. 1832 */ 1833 if ((!is_newentry && (olladdr || lladdr)) || /* (2-5) */ 1834 (is_newentry && lladdr)) { /* (7) */ 1835 ln->ln_router = 1; 1836 } 1837 break; 1838 } 1839 1840 /* 1841 * When the link-layer address of a router changes, select the 1842 * best router again. In particular, when the neighbor entry is newly 1843 * created, it might affect the selection policy. 1844 * Question: can we restrict the first condition to the "is_newentry" 1845 * case? 1846 * XXX: when we hear an RA from a new router with the link-layer 1847 * address option, defrouter_select() is called twice, since 1848 * defrtrlist_update called the function as well. However, I believe 1849 * we can compromise the overhead, since it only happens the first 1850 * time. 1851 * XXX: although defrouter_select() should not have a bad effect 1852 * for those are not autoconfigured hosts, we explicitly avoid such 1853 * cases for safety. 1854 */ 1855 if (do_update && ln->ln_router && !ip6_forwarding && ip6_accept_rtadv) 1856 defrouter_select(); 1857 1858 return rt; 1859 } 1860 1861 static void 1862 nd6_slowtimo(void *ignored_arg) 1863 { 1864 struct nd_ifinfo *nd6if; 1865 struct ifnet *ifp; 1866 1867 mutex_enter(softnet_lock); 1868 KERNEL_LOCK(1, NULL); 1869 callout_reset(&nd6_slowtimo_ch, ND6_SLOWTIMER_INTERVAL * hz, 1870 nd6_slowtimo, NULL); 1871 TAILQ_FOREACH(ifp, &ifnet, if_list) { 1872 nd6if = ND_IFINFO(ifp); 1873 if (nd6if->basereachable && /* already initialized */ 1874 (nd6if->recalctm -= ND6_SLOWTIMER_INTERVAL) <= 0) { 1875 /* 1876 * Since reachable time rarely changes by router 1877 * advertisements, we SHOULD insure that a new random 1878 * value gets recomputed at least once every few hours. 1879 * (RFC 2461, 6.3.4) 1880 */ 1881 nd6if->recalctm = nd6_recalc_reachtm_interval; 1882 nd6if->reachable = ND_COMPUTE_RTIME(nd6if->basereachable); 1883 } 1884 } 1885 KERNEL_UNLOCK_ONE(NULL); 1886 mutex_exit(softnet_lock); 1887 } 1888 1889 #define senderr(e) { error = (e); goto bad;} 1890 int 1891 nd6_output(struct ifnet *ifp, struct ifnet *origifp, struct mbuf *m0, 1892 const struct sockaddr_in6 *dst, struct rtentry *rt0) 1893 { 1894 struct mbuf *m = m0; 1895 struct rtentry *rt = rt0; 1896 struct sockaddr_in6 *gw6 = NULL; 1897 struct llinfo_nd6 *ln = NULL; 1898 int error = 0; 1899 1900 if (IN6_IS_ADDR_MULTICAST(&dst->sin6_addr)) 1901 goto sendpkt; 1902 1903 if (nd6_need_cache(ifp) == 0) 1904 goto sendpkt; 1905 1906 /* 1907 * next hop determination. This routine is derived from ether_output. 1908 */ 1909 if (rt) { 1910 if ((rt->rt_flags & RTF_UP) == 0) { 1911 if ((rt0 = rt = rtalloc1(sin6tocsa(dst), 1)) != NULL) { 1912 rt->rt_refcnt--; 1913 if (rt->rt_ifp != ifp) 1914 senderr(EHOSTUNREACH); 1915 } else 1916 senderr(EHOSTUNREACH); 1917 } 1918 1919 if (rt->rt_flags & RTF_GATEWAY) { 1920 gw6 = (struct sockaddr_in6 *)rt->rt_gateway; 1921 1922 /* 1923 * We skip link-layer address resolution and NUD 1924 * if the gateway is not a neighbor from ND point 1925 * of view, regardless of the value of nd_ifinfo.flags. 1926 * The second condition is a bit tricky; we skip 1927 * if the gateway is our own address, which is 1928 * sometimes used to install a route to a p2p link. 1929 */ 1930 if (!nd6_is_addr_neighbor(gw6, ifp) || 1931 in6ifa_ifpwithaddr(ifp, &gw6->sin6_addr)) { 1932 /* 1933 * We allow this kind of tricky route only 1934 * when the outgoing interface is p2p. 1935 * XXX: we may need a more generic rule here. 1936 */ 1937 if ((ifp->if_flags & IFF_POINTOPOINT) == 0) 1938 senderr(EHOSTUNREACH); 1939 1940 goto sendpkt; 1941 } 1942 1943 if (rt->rt_gwroute == NULL) 1944 goto lookup; 1945 if (((rt = rt->rt_gwroute)->rt_flags & RTF_UP) == 0) { 1946 rtfree(rt); rt = rt0; 1947 lookup: 1948 rt->rt_gwroute = rtalloc1(rt->rt_gateway, 1); 1949 if ((rt = rt->rt_gwroute) == NULL) 1950 senderr(EHOSTUNREACH); 1951 /* the "G" test below also prevents rt == rt0 */ 1952 if ((rt->rt_flags & RTF_GATEWAY) || 1953 (rt->rt_ifp != ifp)) { 1954 rt->rt_refcnt--; 1955 rt0->rt_gwroute = NULL; 1956 senderr(EHOSTUNREACH); 1957 } 1958 } 1959 } 1960 } 1961 1962 /* 1963 * Address resolution or Neighbor Unreachability Detection 1964 * for the next hop. 1965 * At this point, the destination of the packet must be a unicast 1966 * or an anycast address(i.e. not a multicast). 1967 */ 1968 1969 /* Look up the neighbor cache for the nexthop */ 1970 if (rt != NULL && (rt->rt_flags & RTF_LLINFO) != 0) 1971 ln = (struct llinfo_nd6 *)rt->rt_llinfo; 1972 else { 1973 /* 1974 * Since nd6_is_addr_neighbor() internally calls nd6_lookup(), 1975 * the condition below is not very efficient. But we believe 1976 * it is tolerable, because this should be a rare case. 1977 */ 1978 if (nd6_is_addr_neighbor(dst, ifp) && 1979 (rt = nd6_lookup(&dst->sin6_addr, 1, ifp)) != NULL) 1980 ln = (struct llinfo_nd6 *)rt->rt_llinfo; 1981 } 1982 if (ln == NULL || rt == NULL) { 1983 if ((ifp->if_flags & IFF_POINTOPOINT) == 0 && 1984 !(ND_IFINFO(ifp)->flags & ND6_IFF_PERFORMNUD)) { 1985 log(LOG_DEBUG, 1986 "nd6_output: can't allocate llinfo for %s " 1987 "(ln=%p, rt=%p)\n", 1988 ip6_sprintf(&dst->sin6_addr), ln, rt); 1989 senderr(EIO); /* XXX: good error? */ 1990 } 1991 1992 goto sendpkt; /* send anyway */ 1993 } 1994 1995 /* We don't have to do link-layer address resolution on a p2p link. */ 1996 if ((ifp->if_flags & IFF_POINTOPOINT) != 0 && 1997 ln->ln_state < ND6_LLINFO_REACHABLE) { 1998 ln->ln_state = ND6_LLINFO_STALE; 1999 nd6_llinfo_settimer(ln, (long)nd6_gctimer * hz); 2000 } 2001 2002 /* 2003 * The first time we send a packet to a neighbor whose entry is 2004 * STALE, we have to change the state to DELAY and a sets a timer to 2005 * expire in DELAY_FIRST_PROBE_TIME seconds to ensure do 2006 * neighbor unreachability detection on expiration. 2007 * (RFC 2461 7.3.3) 2008 */ 2009 if (ln->ln_state == ND6_LLINFO_STALE) { 2010 ln->ln_asked = 0; 2011 ln->ln_state = ND6_LLINFO_DELAY; 2012 nd6_llinfo_settimer(ln, (long)nd6_delay * hz); 2013 } 2014 2015 /* 2016 * If the neighbor cache entry has a state other than INCOMPLETE 2017 * (i.e. its link-layer address is already resolved), just 2018 * send the packet. 2019 */ 2020 if (ln->ln_state > ND6_LLINFO_INCOMPLETE) 2021 goto sendpkt; 2022 2023 /* 2024 * There is a neighbor cache entry, but no ethernet address 2025 * response yet. Append this latest packet to the end of the 2026 * packet queue in the mbuf, unless the number of the packet 2027 * does not exceed nd6_maxqueuelen. When it exceeds nd6_maxqueuelen, 2028 * the oldest packet in the queue will be removed. 2029 */ 2030 if (ln->ln_state == ND6_LLINFO_NOSTATE) 2031 ln->ln_state = ND6_LLINFO_INCOMPLETE; 2032 if (ln->ln_hold) { 2033 struct mbuf *m_hold; 2034 int i; 2035 2036 i = 0; 2037 for (m_hold = ln->ln_hold; m_hold; m_hold = m_hold->m_nextpkt) { 2038 i++; 2039 if (m_hold->m_nextpkt == NULL) { 2040 m_hold->m_nextpkt = m; 2041 break; 2042 } 2043 } 2044 while (i >= nd6_maxqueuelen) { 2045 m_hold = ln->ln_hold; 2046 ln->ln_hold = ln->ln_hold->m_nextpkt; 2047 m_freem(m_hold); 2048 i--; 2049 } 2050 } else { 2051 ln->ln_hold = m; 2052 } 2053 2054 /* 2055 * If there has been no NS for the neighbor after entering the 2056 * INCOMPLETE state, send the first solicitation. 2057 */ 2058 if (!ND6_LLINFO_PERMANENT(ln) && ln->ln_asked == 0) { 2059 ln->ln_asked++; 2060 nd6_llinfo_settimer(ln, 2061 (long)ND_IFINFO(ifp)->retrans * hz / 1000); 2062 nd6_ns_output(ifp, NULL, &dst->sin6_addr, ln, 0); 2063 } 2064 return 0; 2065 2066 sendpkt: 2067 /* discard the packet if IPv6 operation is disabled on the interface */ 2068 if ((ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED)) { 2069 error = ENETDOWN; /* better error? */ 2070 goto bad; 2071 } 2072 2073 #ifdef IPSEC 2074 /* clean ipsec history once it goes out of the node */ 2075 ipsec_delaux(m); 2076 #endif 2077 if ((ifp->if_flags & IFF_LOOPBACK) != 0) 2078 return (*ifp->if_output)(origifp, m, sin6tocsa(dst), rt); 2079 return (*ifp->if_output)(ifp, m, sin6tocsa(dst), rt); 2080 2081 bad: 2082 if (m != NULL) 2083 m_freem(m); 2084 return error; 2085 } 2086 #undef senderr 2087 2088 int 2089 nd6_need_cache(struct ifnet *ifp) 2090 { 2091 /* 2092 * XXX: we currently do not make neighbor cache on any interface 2093 * other than ARCnet, Ethernet, FDDI and GIF. 2094 * 2095 * RFC2893 says: 2096 * - unidirectional tunnels needs no ND 2097 */ 2098 switch (ifp->if_type) { 2099 case IFT_ARCNET: 2100 case IFT_ETHER: 2101 case IFT_FDDI: 2102 case IFT_IEEE1394: 2103 case IFT_CARP: 2104 case IFT_GIF: /* XXX need more cases? */ 2105 case IFT_PPP: 2106 case IFT_TUNNEL: 2107 return 1; 2108 default: 2109 return 0; 2110 } 2111 } 2112 2113 int 2114 nd6_storelladdr(const struct ifnet *ifp, const struct rtentry *rt, 2115 struct mbuf *m, const struct sockaddr *dst, uint8_t *lldst, 2116 size_t dstsize) 2117 { 2118 const struct sockaddr_dl *sdl; 2119 2120 if (m->m_flags & M_MCAST) { 2121 switch (ifp->if_type) { 2122 case IFT_ETHER: 2123 case IFT_FDDI: 2124 ETHER_MAP_IPV6_MULTICAST(&satocsin6(dst)->sin6_addr, 2125 lldst); 2126 return 1; 2127 case IFT_IEEE1394: 2128 memcpy(lldst, ifp->if_broadcastaddr, 2129 MIN(dstsize, ifp->if_addrlen)); 2130 return 1; 2131 case IFT_ARCNET: 2132 *lldst = 0; 2133 return 1; 2134 default: 2135 m_freem(m); 2136 return 0; 2137 } 2138 } 2139 2140 if (rt == NULL) { 2141 /* this could happen, if we could not allocate memory */ 2142 m_freem(m); 2143 return 0; 2144 } 2145 if (rt->rt_gateway->sa_family != AF_LINK) { 2146 printf("%s: something odd happens\n", __func__); 2147 m_freem(m); 2148 return 0; 2149 } 2150 sdl = satocsdl(rt->rt_gateway); 2151 if (sdl->sdl_alen == 0 || sdl->sdl_alen > dstsize) { 2152 /* this should be impossible, but we bark here for debugging */ 2153 printf("%s: sdl_alen == 0, dst=%s, if=%s\n", __func__, 2154 ip6_sprintf(&satocsin6(dst)->sin6_addr), if_name(ifp)); 2155 m_freem(m); 2156 return 0; 2157 } 2158 2159 memcpy(lldst, CLLADDR(sdl), MIN(dstsize, sdl->sdl_alen)); 2160 return 1; 2161 } 2162 2163 static void 2164 clear_llinfo_pqueue(struct llinfo_nd6 *ln) 2165 { 2166 struct mbuf *m_hold, *m_hold_next; 2167 2168 for (m_hold = ln->ln_hold; m_hold; m_hold = m_hold_next) { 2169 m_hold_next = m_hold->m_nextpkt; 2170 m_hold->m_nextpkt = NULL; 2171 m_freem(m_hold); 2172 } 2173 2174 ln->ln_hold = NULL; 2175 return; 2176 } 2177 2178 int 2179 nd6_sysctl( 2180 int name, 2181 void *oldp, /* syscall arg, need copyout */ 2182 size_t *oldlenp, 2183 void *newp, /* syscall arg, need copyin */ 2184 size_t newlen 2185 ) 2186 { 2187 void *p; 2188 size_t ol; 2189 int error; 2190 2191 error = 0; 2192 2193 if (newp) 2194 return EPERM; 2195 if (oldp && !oldlenp) 2196 return EINVAL; 2197 ol = oldlenp ? *oldlenp : 0; 2198 2199 if (oldp) { 2200 p = malloc(*oldlenp, M_TEMP, M_WAITOK); 2201 if (p == NULL) 2202 return ENOMEM; 2203 } else 2204 p = NULL; 2205 switch (name) { 2206 case ICMPV6CTL_ND6_DRLIST: 2207 error = fill_drlist(p, oldlenp, ol); 2208 if (!error && p != NULL && oldp != NULL) 2209 error = copyout(p, oldp, *oldlenp); 2210 break; 2211 2212 case ICMPV6CTL_ND6_PRLIST: 2213 error = fill_prlist(p, oldlenp, ol); 2214 if (!error && p != NULL && oldp != NULL) 2215 error = copyout(p, oldp, *oldlenp); 2216 break; 2217 2218 case ICMPV6CTL_ND6_MAXQLEN: 2219 break; 2220 2221 default: 2222 error = ENOPROTOOPT; 2223 break; 2224 } 2225 if (p) 2226 free(p, M_TEMP); 2227 2228 return error; 2229 } 2230 2231 static int 2232 fill_drlist(void *oldp, size_t *oldlenp, size_t ol) 2233 { 2234 int error = 0, s; 2235 struct in6_defrouter *d = NULL, *de = NULL; 2236 struct nd_defrouter *dr; 2237 size_t l; 2238 2239 s = splsoftnet(); 2240 2241 if (oldp) { 2242 d = (struct in6_defrouter *)oldp; 2243 de = (struct in6_defrouter *)((char *)oldp + *oldlenp); 2244 } 2245 l = 0; 2246 2247 TAILQ_FOREACH(dr, &nd_defrouter, dr_entry) { 2248 2249 if (oldp && d + 1 <= de) { 2250 memset(d, 0, sizeof(*d)); 2251 sockaddr_in6_init(&d->rtaddr, &dr->rtaddr, 0, 0, 0); 2252 if (sa6_recoverscope(&d->rtaddr)) { 2253 log(LOG_ERR, 2254 "scope error in router list (%s)\n", 2255 ip6_sprintf(&d->rtaddr.sin6_addr)); 2256 /* XXX: press on... */ 2257 } 2258 d->flags = dr->flags; 2259 d->rtlifetime = dr->rtlifetime; 2260 d->expire = dr->expire; 2261 d->if_index = dr->ifp->if_index; 2262 } 2263 2264 l += sizeof(*d); 2265 if (d) 2266 d++; 2267 } 2268 2269 if (oldp) { 2270 if (l > ol) 2271 error = ENOMEM; 2272 } 2273 if (oldlenp) 2274 *oldlenp = l; /* (void *)d - (void *)oldp */ 2275 2276 splx(s); 2277 2278 return error; 2279 } 2280 2281 static int 2282 fill_prlist(void *oldp, size_t *oldlenp, size_t ol) 2283 { 2284 int error = 0, s; 2285 struct nd_prefix *pr; 2286 struct in6_prefix *p = NULL; 2287 struct in6_prefix *pe = NULL; 2288 size_t l; 2289 2290 s = splsoftnet(); 2291 2292 if (oldp) { 2293 p = (struct in6_prefix *)oldp; 2294 pe = (struct in6_prefix *)((char *)oldp + *oldlenp); 2295 } 2296 l = 0; 2297 2298 LIST_FOREACH(pr, &nd_prefix, ndpr_entry) { 2299 u_short advrtrs; 2300 size_t advance; 2301 struct sockaddr_in6 *sin6; 2302 struct sockaddr_in6 *s6; 2303 struct nd_pfxrouter *pfr; 2304 2305 if (oldp && p + 1 <= pe) 2306 { 2307 bzero(p, sizeof(*p)); 2308 sin6 = (struct sockaddr_in6 *)(p + 1); 2309 2310 p->prefix = pr->ndpr_prefix; 2311 if (sa6_recoverscope(&p->prefix)) { 2312 log(LOG_ERR, 2313 "scope error in prefix list (%s)\n", 2314 ip6_sprintf(&p->prefix.sin6_addr)); 2315 /* XXX: press on... */ 2316 } 2317 p->raflags = pr->ndpr_raf; 2318 p->prefixlen = pr->ndpr_plen; 2319 p->vltime = pr->ndpr_vltime; 2320 p->pltime = pr->ndpr_pltime; 2321 p->if_index = pr->ndpr_ifp->if_index; 2322 if (pr->ndpr_vltime == ND6_INFINITE_LIFETIME) 2323 p->expire = 0; 2324 else { 2325 time_t maxexpire; 2326 2327 /* XXX: we assume time_t is signed. */ 2328 maxexpire = (-1) & 2329 ~((time_t)1 << 2330 ((sizeof(maxexpire) * 8) - 1)); 2331 if (pr->ndpr_vltime < 2332 maxexpire - pr->ndpr_lastupdate) { 2333 p->expire = pr->ndpr_lastupdate + 2334 pr->ndpr_vltime; 2335 } else 2336 p->expire = maxexpire; 2337 } 2338 p->refcnt = pr->ndpr_refcnt; 2339 p->flags = pr->ndpr_stateflags; 2340 p->origin = PR_ORIG_RA; 2341 advrtrs = 0; 2342 LIST_FOREACH(pfr, &pr->ndpr_advrtrs, pfr_entry) { 2343 if ((void *)&sin6[advrtrs + 1] > (void *)pe) { 2344 advrtrs++; 2345 continue; 2346 } 2347 s6 = &sin6[advrtrs]; 2348 sockaddr_in6_init(s6, &pfr->router->rtaddr, 2349 0, 0, 0); 2350 if (sa6_recoverscope(s6)) { 2351 log(LOG_ERR, 2352 "scope error in " 2353 "prefix list (%s)\n", 2354 ip6_sprintf(&pfr->router->rtaddr)); 2355 } 2356 advrtrs++; 2357 } 2358 p->advrtrs = advrtrs; 2359 } 2360 else { 2361 advrtrs = 0; 2362 LIST_FOREACH(pfr, &pr->ndpr_advrtrs, pfr_entry) 2363 advrtrs++; 2364 } 2365 2366 advance = sizeof(*p) + sizeof(*sin6) * advrtrs; 2367 l += advance; 2368 if (p) 2369 p = (struct in6_prefix *)((char *)p + advance); 2370 } 2371 2372 if (oldp) { 2373 *oldlenp = l; /* (void *)d - (void *)oldp */ 2374 if (l > ol) 2375 error = ENOMEM; 2376 } else 2377 *oldlenp = l; 2378 2379 splx(s); 2380 2381 return error; 2382 } 2383