1 /* $NetBSD: nd6.c,v 1.141 2012/02/03 03:32:45 christos 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 1.141 2012/02/03 03:32:45 christos 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 #include <sys/cprng.h> 54 55 #include <net/if.h> 56 #include <net/if_dl.h> 57 #include <net/if_types.h> 58 #include <net/route.h> 59 #include <net/if_ether.h> 60 #include <net/if_fddi.h> 61 #include <net/if_arc.h> 62 63 #include <netinet/in.h> 64 #include <netinet6/in6_var.h> 65 #include <netinet/ip6.h> 66 #include <netinet6/ip6_var.h> 67 #include <netinet6/scope6_var.h> 68 #include <netinet6/nd6.h> 69 #include <netinet/icmp6.h> 70 #include <netinet6/icmp6_private.h> 71 72 #ifdef KAME_IPSEC 73 #include <netinet6/ipsec.h> 74 #endif 75 76 #include <net/net_osdep.h> 77 78 #define ND6_SLOWTIMER_INTERVAL (60 * 60) /* 1 hour */ 79 #define ND6_RECALC_REACHTM_INTERVAL (60 * 120) /* 2 hours */ 80 81 /* timer values */ 82 int nd6_prune = 1; /* walk list every 1 seconds */ 83 int nd6_delay = 5; /* delay first probe time 5 second */ 84 int nd6_umaxtries = 3; /* maximum unicast query */ 85 int nd6_mmaxtries = 3; /* maximum multicast query */ 86 int nd6_useloopback = 1; /* use loopback interface for local traffic */ 87 int nd6_gctimer = (60 * 60 * 24); /* 1 day: garbage collection timer */ 88 89 /* preventing too many loops in ND option parsing */ 90 int nd6_maxndopt = 10; /* max # of ND options allowed */ 91 92 int nd6_maxnudhint = 0; /* max # of subsequent upper layer hints */ 93 94 int nd6_maxqueuelen = 1; /* max # of packets cached in unresolved ND entries */ 95 96 #ifdef ND6_DEBUG 97 int nd6_debug = 1; 98 #else 99 int nd6_debug = 0; 100 #endif 101 102 /* for debugging? */ 103 static int nd6_inuse, nd6_allocated; 104 105 struct llinfo_nd6 llinfo_nd6 = { 106 .ln_prev = &llinfo_nd6, 107 .ln_next = &llinfo_nd6, 108 }; 109 struct nd_drhead nd_defrouter; 110 struct nd_prhead nd_prefix = { 0 }; 111 112 int nd6_recalc_reachtm_interval = ND6_RECALC_REACHTM_INTERVAL; 113 static const struct sockaddr_in6 all1_sa = { 114 .sin6_family = AF_INET6 115 , .sin6_len = sizeof(struct sockaddr_in6) 116 , .sin6_addr = {.s6_addr = {0xff, 0xff, 0xff, 0xff, 117 0xff, 0xff, 0xff, 0xff, 118 0xff, 0xff, 0xff, 0xff, 119 0xff, 0xff, 0xff, 0xff}} 120 }; 121 122 static void nd6_setmtu0(struct ifnet *, struct nd_ifinfo *); 123 static void nd6_slowtimo(void *); 124 static int regen_tmpaddr(struct in6_ifaddr *); 125 static struct llinfo_nd6 *nd6_free(struct rtentry *, int); 126 static void nd6_llinfo_timer(void *); 127 static void clear_llinfo_pqueue(struct llinfo_nd6 *); 128 129 callout_t nd6_slowtimo_ch; 130 callout_t nd6_timer_ch; 131 extern callout_t in6_tmpaddrtimer_ch; 132 133 static int fill_drlist(void *, size_t *, size_t); 134 static int fill_prlist(void *, size_t *, size_t); 135 136 MALLOC_DEFINE(M_IP6NDP, "NDP", "IPv6 Neighbour Discovery"); 137 138 void 139 nd6_init(void) 140 { 141 static int nd6_init_done = 0; 142 143 if (nd6_init_done) { 144 log(LOG_NOTICE, "nd6_init called more than once(ignored)\n"); 145 return; 146 } 147 148 /* initialization of the default router list */ 149 TAILQ_INIT(&nd_defrouter); 150 151 nd6_init_done = 1; 152 153 callout_init(&nd6_slowtimo_ch, CALLOUT_MPSAFE); 154 callout_init(&nd6_timer_ch, CALLOUT_MPSAFE); 155 156 /* start timer */ 157 callout_reset(&nd6_slowtimo_ch, ND6_SLOWTIMER_INTERVAL * hz, 158 nd6_slowtimo, NULL); 159 } 160 161 struct nd_ifinfo * 162 nd6_ifattach(struct ifnet *ifp) 163 { 164 struct nd_ifinfo *nd; 165 166 nd = (struct nd_ifinfo *)malloc(sizeof(*nd), M_IP6NDP, M_WAITOK|M_ZERO); 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. 176 * Because we do not set ND6_IFF_OVERRIDE_RTADV here, we won't 177 * accept RAs by default. 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 memset(ndopts, 0, 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 memset(ndopts, 0, 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 memset(ndopts, 0, 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 memset(ndopts, 0, 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 memset(ndopts, 0, 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; 527 528 callout_reset(&nd6_timer_ch, nd6_prune * hz, 529 nd6_timer, NULL); 530 531 mutex_enter(softnet_lock); 532 KERNEL_LOCK(1, NULL); 533 534 /* expire default router list */ 535 536 TAILQ_FOREACH_SAFE(dr, &nd_defrouter, dr_entry, next_dr) { 537 if (dr->expire && dr->expire < time_second) { 538 defrtrlist_del(dr); 539 } 540 } 541 542 /* 543 * expire interface addresses. 544 * in the past the loop was inside prefix expiry processing. 545 * However, from a stricter speci-confrmance standpoint, we should 546 * rather separate address lifetimes and prefix lifetimes. 547 */ 548 addrloop: 549 for (ia6 = in6_ifaddr; ia6; ia6 = nia6) { 550 nia6 = ia6->ia_next; 551 /* check address lifetime */ 552 if (IFA6_IS_INVALID(ia6)) { 553 int regen = 0; 554 555 /* 556 * If the expiring address is temporary, try 557 * regenerating a new one. This would be useful when 558 * we suspended a laptop PC, then turned it on after a 559 * period that could invalidate all temporary 560 * addresses. Although we may have to restart the 561 * loop (see below), it must be after purging the 562 * address. Otherwise, we'd see an infinite loop of 563 * regeneration. 564 */ 565 if (ip6_use_tempaddr && 566 (ia6->ia6_flags & IN6_IFF_TEMPORARY) != 0) { 567 if (regen_tmpaddr(ia6) == 0) 568 regen = 1; 569 } 570 571 in6_purgeaddr(&ia6->ia_ifa); 572 573 if (regen) 574 goto addrloop; /* XXX: see below */ 575 } else if (IFA6_IS_DEPRECATED(ia6)) { 576 int oldflags = ia6->ia6_flags; 577 578 ia6->ia6_flags |= IN6_IFF_DEPRECATED; 579 580 /* 581 * If a temporary address has just become deprecated, 582 * regenerate a new one if possible. 583 */ 584 if (ip6_use_tempaddr && 585 (ia6->ia6_flags & IN6_IFF_TEMPORARY) != 0 && 586 (oldflags & IN6_IFF_DEPRECATED) == 0) { 587 588 if (regen_tmpaddr(ia6) == 0) { 589 /* 590 * A new temporary address is 591 * generated. 592 * XXX: this means the address chain 593 * has changed while we are still in 594 * the loop. Although the change 595 * would not cause disaster (because 596 * it's not a deletion, but an 597 * addition,) we'd rather restart the 598 * loop just for safety. Or does this 599 * significantly reduce performance?? 600 */ 601 goto addrloop; 602 } 603 } 604 } else { 605 /* 606 * A new RA might have made a deprecated address 607 * preferred. 608 */ 609 ia6->ia6_flags &= ~IN6_IFF_DEPRECATED; 610 } 611 } 612 613 /* expire prefix list */ 614 LIST_FOREACH_SAFE(pr, &nd_prefix, ndpr_entry, next_pr) { 615 /* 616 * check prefix lifetime. 617 * since pltime is just for autoconf, pltime processing for 618 * prefix is not necessary. 619 */ 620 if (pr->ndpr_vltime != ND6_INFINITE_LIFETIME && 621 time_second - pr->ndpr_lastupdate > pr->ndpr_vltime) { 622 623 /* 624 * address expiration and prefix expiration are 625 * separate. NEVER perform in6_purgeaddr here. 626 */ 627 628 prelist_remove(pr); 629 } 630 } 631 632 KERNEL_UNLOCK_ONE(NULL); 633 mutex_exit(softnet_lock); 634 } 635 636 /* ia6: deprecated/invalidated temporary address */ 637 static int 638 regen_tmpaddr(struct in6_ifaddr *ia6) 639 { 640 struct ifaddr *ifa; 641 struct ifnet *ifp; 642 struct in6_ifaddr *public_ifa6 = NULL; 643 644 ifp = ia6->ia_ifa.ifa_ifp; 645 IFADDR_FOREACH(ifa, ifp) { 646 struct in6_ifaddr *it6; 647 648 if (ifa->ifa_addr->sa_family != AF_INET6) 649 continue; 650 651 it6 = (struct in6_ifaddr *)ifa; 652 653 /* ignore no autoconf addresses. */ 654 if ((it6->ia6_flags & IN6_IFF_AUTOCONF) == 0) 655 continue; 656 657 /* ignore autoconf addresses with different prefixes. */ 658 if (it6->ia6_ndpr == NULL || it6->ia6_ndpr != ia6->ia6_ndpr) 659 continue; 660 661 /* 662 * Now we are looking at an autoconf address with the same 663 * prefix as ours. If the address is temporary and is still 664 * preferred, do not create another one. It would be rare, but 665 * could happen, for example, when we resume a laptop PC after 666 * a long period. 667 */ 668 if ((it6->ia6_flags & IN6_IFF_TEMPORARY) != 0 && 669 !IFA6_IS_DEPRECATED(it6)) { 670 public_ifa6 = NULL; 671 break; 672 } 673 674 /* 675 * This is a public autoconf address that has the same prefix 676 * as ours. If it is preferred, keep it. We can't break the 677 * loop here, because there may be a still-preferred temporary 678 * address with the prefix. 679 */ 680 if (!IFA6_IS_DEPRECATED(it6)) 681 public_ifa6 = it6; 682 } 683 684 if (public_ifa6 != NULL) { 685 int e; 686 687 /* 688 * Random factor is introduced in the preferred lifetime, so 689 * we do not need additional delay (3rd arg to in6_tmpifadd). 690 */ 691 if ((e = in6_tmpifadd(public_ifa6, 0, 0)) != 0) { 692 log(LOG_NOTICE, "regen_tmpaddr: failed to create a new" 693 " tmp addr, errno=%d\n", e); 694 return -1; 695 } 696 return 0; 697 } 698 699 return -1; 700 } 701 702 bool 703 nd6_accepts_rtadv(const struct nd_ifinfo *ndi) 704 { 705 switch (ndi->flags & (ND6_IFF_ACCEPT_RTADV|ND6_IFF_OVERRIDE_RTADV)) { 706 case ND6_IFF_OVERRIDE_RTADV|ND6_IFF_ACCEPT_RTADV: 707 return true; 708 case ND6_IFF_ACCEPT_RTADV: 709 return ip6_accept_rtadv != 0; 710 case ND6_IFF_OVERRIDE_RTADV: 711 case 0: 712 default: 713 return false; 714 } 715 } 716 717 /* 718 * Nuke neighbor cache/prefix/default router management table, right before 719 * ifp goes away. 720 */ 721 void 722 nd6_purge(struct ifnet *ifp) 723 { 724 struct llinfo_nd6 *ln, *nln; 725 struct nd_defrouter *dr, *ndr; 726 struct nd_prefix *pr, *npr; 727 728 /* 729 * Nuke default router list entries toward ifp. 730 * We defer removal of default router list entries that is installed 731 * in the routing table, in order to keep additional side effects as 732 * small as possible. 733 */ 734 TAILQ_FOREACH_SAFE(dr, &nd_defrouter, dr_entry, ndr) { 735 if (dr->installed) 736 continue; 737 738 if (dr->ifp == ifp) 739 defrtrlist_del(dr); 740 } 741 742 TAILQ_FOREACH_SAFE(dr, &nd_defrouter, dr_entry, ndr) { 743 if (!dr->installed) 744 continue; 745 746 if (dr->ifp == ifp) 747 defrtrlist_del(dr); 748 } 749 750 /* Nuke prefix list entries toward ifp */ 751 LIST_FOREACH_SAFE(pr, &nd_prefix, ndpr_entry, npr) { 752 if (pr->ndpr_ifp == ifp) { 753 /* 754 * Because if_detach() does *not* release prefixes 755 * while purging addresses the reference count will 756 * still be above zero. We therefore reset it to 757 * make sure that the prefix really gets purged. 758 */ 759 pr->ndpr_refcnt = 0; 760 /* 761 * Previously, pr->ndpr_addr is removed as well, 762 * but I strongly believe we don't have to do it. 763 * nd6_purge() is only called from in6_ifdetach(), 764 * which removes all the associated interface addresses 765 * by itself. 766 * (jinmei@kame.net 20010129) 767 */ 768 prelist_remove(pr); 769 } 770 } 771 772 /* cancel default outgoing interface setting */ 773 if (nd6_defifindex == ifp->if_index) 774 nd6_setdefaultiface(0); 775 776 /* XXX: too restrictive? */ 777 if (!ip6_forwarding && ifp->if_afdata[AF_INET6]) { 778 struct nd_ifinfo *ndi = ND_IFINFO(ifp); 779 if (ndi && nd6_accepts_rtadv(ndi)) { 780 /* refresh default router list */ 781 defrouter_select(); 782 } 783 } 784 785 /* 786 * Nuke neighbor cache entries for the ifp. 787 * Note that rt->rt_ifp may not be the same as ifp, 788 * due to KAME goto ours hack. See RTM_RESOLVE case in 789 * nd6_rtrequest(), and ip6_input(). 790 */ 791 ln = llinfo_nd6.ln_next; 792 while (ln != NULL && ln != &llinfo_nd6) { 793 struct rtentry *rt; 794 const struct sockaddr_dl *sdl; 795 796 nln = ln->ln_next; 797 rt = ln->ln_rt; 798 if (rt && rt->rt_gateway && 799 rt->rt_gateway->sa_family == AF_LINK) { 800 sdl = satocsdl(rt->rt_gateway); 801 if (sdl->sdl_index == ifp->if_index) 802 nln = nd6_free(rt, 0); 803 } 804 ln = nln; 805 } 806 } 807 808 struct rtentry * 809 nd6_lookup(const struct in6_addr *addr6, int create, struct ifnet *ifp) 810 { 811 struct rtentry *rt; 812 struct sockaddr_in6 sin6; 813 814 sockaddr_in6_init(&sin6, addr6, 0, 0, 0); 815 rt = rtalloc1((struct sockaddr *)&sin6, create); 816 if (rt != NULL && (rt->rt_flags & RTF_LLINFO) == 0) { 817 /* 818 * This is the case for the default route. 819 * If we want to create a neighbor cache for the address, we 820 * should free the route for the destination and allocate an 821 * interface route. 822 */ 823 if (create) { 824 RTFREE(rt); 825 rt = NULL; 826 } 827 } 828 if (rt != NULL) 829 ; 830 else if (create && ifp) { 831 int e; 832 833 /* 834 * If no route is available and create is set, 835 * we allocate a host route for the destination 836 * and treat it like an interface route. 837 * This hack is necessary for a neighbor which can't 838 * be covered by our own prefix. 839 */ 840 struct ifaddr *ifa = 841 ifaof_ifpforaddr((struct sockaddr *)&sin6, ifp); 842 if (ifa == NULL) 843 return NULL; 844 845 /* 846 * Create a new route. RTF_LLINFO is necessary 847 * to create a Neighbor Cache entry for the 848 * destination in nd6_rtrequest which will be 849 * called in rtrequest via ifa->ifa_rtrequest. 850 */ 851 if ((e = rtrequest(RTM_ADD, (const struct sockaddr *)&sin6, 852 ifa->ifa_addr, (const struct sockaddr *)&all1_sa, 853 (ifa->ifa_flags | RTF_HOST | RTF_LLINFO) & 854 ~RTF_CLONING, &rt)) != 0) { 855 #if 0 856 log(LOG_ERR, 857 "nd6_lookup: failed to add route for a " 858 "neighbor(%s), errno=%d\n", 859 ip6_sprintf(addr6), e); 860 #endif 861 return NULL; 862 } 863 if (rt == NULL) 864 return NULL; 865 if (rt->rt_llinfo) { 866 struct llinfo_nd6 *ln = 867 (struct llinfo_nd6 *)rt->rt_llinfo; 868 ln->ln_state = ND6_LLINFO_NOSTATE; 869 } 870 } else 871 return NULL; 872 rt->rt_refcnt--; 873 /* 874 * Validation for the entry. 875 * Note that the check for rt_llinfo is necessary because a cloned 876 * route from a parent route that has the L flag (e.g. the default 877 * route to a p2p interface) may have the flag, too, while the 878 * destination is not actually a neighbor. 879 * XXX: we can't use rt->rt_ifp to check for the interface, since 880 * it might be the loopback interface if the entry is for our 881 * own address on a non-loopback interface. Instead, we should 882 * use rt->rt_ifa->ifa_ifp, which would specify the REAL 883 * interface. 884 * Note also that ifa_ifp and ifp may differ when we connect two 885 * interfaces to a same link, install a link prefix to an interface, 886 * and try to install a neighbor cache on an interface that does not 887 * have a route to the prefix. 888 */ 889 if ((rt->rt_flags & RTF_GATEWAY) || (rt->rt_flags & RTF_LLINFO) == 0 || 890 rt->rt_gateway->sa_family != AF_LINK || rt->rt_llinfo == NULL || 891 (ifp && rt->rt_ifa->ifa_ifp != ifp)) { 892 if (create) { 893 nd6log((LOG_DEBUG, 894 "nd6_lookup: failed to lookup %s (if = %s)\n", 895 ip6_sprintf(addr6), 896 ifp ? if_name(ifp) : "unspec")); 897 } 898 return NULL; 899 } 900 return rt; 901 } 902 903 /* 904 * Detect if a given IPv6 address identifies a neighbor on a given link. 905 * XXX: should take care of the destination of a p2p link? 906 */ 907 int 908 nd6_is_addr_neighbor(const struct sockaddr_in6 *addr, struct ifnet *ifp) 909 { 910 struct nd_prefix *pr; 911 912 /* 913 * A link-local address is always a neighbor. 914 * XXX: a link does not necessarily specify a single interface. 915 */ 916 if (IN6_IS_ADDR_LINKLOCAL(&addr->sin6_addr)) { 917 struct sockaddr_in6 sin6_copy; 918 u_int32_t zone; 919 920 /* 921 * We need sin6_copy since sa6_recoverscope() may modify the 922 * content (XXX). 923 */ 924 sin6_copy = *addr; 925 if (sa6_recoverscope(&sin6_copy)) 926 return 0; /* XXX: should be impossible */ 927 if (in6_setscope(&sin6_copy.sin6_addr, ifp, &zone)) 928 return 0; 929 if (sin6_copy.sin6_scope_id == zone) 930 return 1; 931 else 932 return 0; 933 } 934 935 /* 936 * If the address matches one of our on-link prefixes, it should be a 937 * neighbor. 938 */ 939 LIST_FOREACH(pr, &nd_prefix, ndpr_entry) { 940 if (pr->ndpr_ifp != ifp) 941 continue; 942 943 if (!(pr->ndpr_stateflags & NDPRF_ONLINK)) 944 continue; 945 946 if (IN6_ARE_MASKED_ADDR_EQUAL(&pr->ndpr_prefix.sin6_addr, 947 &addr->sin6_addr, &pr->ndpr_mask)) 948 return 1; 949 } 950 951 /* 952 * If the default router list is empty, all addresses are regarded 953 * as on-link, and thus, as a neighbor. 954 * XXX: we restrict the condition to hosts, because routers usually do 955 * not have the "default router list". 956 */ 957 if (!ip6_forwarding && TAILQ_FIRST(&nd_defrouter) == NULL && 958 nd6_defifindex == ifp->if_index) { 959 return 1; 960 } 961 962 /* 963 * Even if the address matches none of our addresses, it might be 964 * in the neighbor cache. 965 */ 966 if (nd6_lookup(&addr->sin6_addr, 0, ifp) != NULL) 967 return 1; 968 969 return 0; 970 } 971 972 /* 973 * Free an nd6 llinfo entry. 974 * Since the function would cause significant changes in the kernel, DO NOT 975 * make it global, unless you have a strong reason for the change, and are sure 976 * that the change is safe. 977 */ 978 static struct llinfo_nd6 * 979 nd6_free(struct rtentry *rt, int gc) 980 { 981 struct llinfo_nd6 *ln = (struct llinfo_nd6 *)rt->rt_llinfo, *next; 982 struct in6_addr in6 = satocsin6(rt_getkey(rt))->sin6_addr; 983 struct nd_defrouter *dr; 984 985 /* 986 * we used to have pfctlinput(PRC_HOSTDEAD) here. 987 * even though it is not harmful, it was not really necessary. 988 */ 989 990 /* cancel timer */ 991 nd6_llinfo_settimer(ln, -1); 992 993 if (!ip6_forwarding) { 994 int s; 995 s = splsoftnet(); 996 dr = defrouter_lookup(&satocsin6(rt_getkey(rt))->sin6_addr, 997 rt->rt_ifp); 998 999 if (dr != NULL && dr->expire && 1000 ln->ln_state == ND6_LLINFO_STALE && gc) { 1001 /* 1002 * If the reason for the deletion is just garbage 1003 * collection, and the neighbor is an active default 1004 * router, do not delete it. Instead, reset the GC 1005 * timer using the router's lifetime. 1006 * Simply deleting the entry would affect default 1007 * router selection, which is not necessarily a good 1008 * thing, especially when we're using router preference 1009 * values. 1010 * XXX: the check for ln_state would be redundant, 1011 * but we intentionally keep it just in case. 1012 */ 1013 if (dr->expire > time_second) 1014 nd6_llinfo_settimer(ln, 1015 (dr->expire - time_second) * hz); 1016 else 1017 nd6_llinfo_settimer(ln, (long)nd6_gctimer * hz); 1018 splx(s); 1019 return ln->ln_next; 1020 } 1021 1022 if (ln->ln_router || dr) { 1023 /* 1024 * rt6_flush must be called whether or not the neighbor 1025 * is in the Default Router List. 1026 * See a corresponding comment in nd6_na_input(). 1027 */ 1028 rt6_flush(&in6, rt->rt_ifp); 1029 } 1030 1031 if (dr) { 1032 /* 1033 * Unreachablity of a router might affect the default 1034 * router selection and on-link detection of advertised 1035 * prefixes. 1036 */ 1037 1038 /* 1039 * Temporarily fake the state to choose a new default 1040 * router and to perform on-link determination of 1041 * prefixes correctly. 1042 * Below the state will be set correctly, 1043 * or the entry itself will be deleted. 1044 */ 1045 ln->ln_state = ND6_LLINFO_INCOMPLETE; 1046 1047 /* 1048 * Since defrouter_select() does not affect the 1049 * on-link determination and MIP6 needs the check 1050 * before the default router selection, we perform 1051 * the check now. 1052 */ 1053 pfxlist_onlink_check(); 1054 1055 /* 1056 * refresh default router list 1057 */ 1058 defrouter_select(); 1059 } 1060 splx(s); 1061 } 1062 1063 /* 1064 * Before deleting the entry, remember the next entry as the 1065 * return value. We need this because pfxlist_onlink_check() above 1066 * might have freed other entries (particularly the old next entry) as 1067 * a side effect (XXX). 1068 */ 1069 next = ln->ln_next; 1070 1071 /* 1072 * Detach the route from the routing tree and the list of neighbor 1073 * caches, and disable the route entry not to be used in already 1074 * cached routes. 1075 */ 1076 rtrequest(RTM_DELETE, rt_getkey(rt), NULL, rt_mask(rt), 0, NULL); 1077 1078 return next; 1079 } 1080 1081 /* 1082 * Upper-layer reachability hint for Neighbor Unreachability Detection. 1083 * 1084 * XXX cost-effective methods? 1085 */ 1086 void 1087 nd6_nud_hint(struct rtentry *rt, struct in6_addr *dst6, int force) 1088 { 1089 struct llinfo_nd6 *ln; 1090 1091 /* 1092 * If the caller specified "rt", use that. Otherwise, resolve the 1093 * routing table by supplied "dst6". 1094 */ 1095 if (rt == NULL) { 1096 if (dst6 == NULL) 1097 return; 1098 if ((rt = nd6_lookup(dst6, 0, NULL)) == NULL) 1099 return; 1100 } 1101 1102 if ((rt->rt_flags & RTF_GATEWAY) != 0 || 1103 (rt->rt_flags & RTF_LLINFO) == 0 || 1104 !rt->rt_llinfo || !rt->rt_gateway || 1105 rt->rt_gateway->sa_family != AF_LINK) { 1106 /* This is not a host route. */ 1107 return; 1108 } 1109 1110 ln = (struct llinfo_nd6 *)rt->rt_llinfo; 1111 if (ln->ln_state < ND6_LLINFO_REACHABLE) 1112 return; 1113 1114 /* 1115 * if we get upper-layer reachability confirmation many times, 1116 * it is possible we have false information. 1117 */ 1118 if (!force) { 1119 ln->ln_byhint++; 1120 if (ln->ln_byhint > nd6_maxnudhint) 1121 return; 1122 } 1123 1124 ln->ln_state = ND6_LLINFO_REACHABLE; 1125 if (!ND6_LLINFO_PERMANENT(ln)) { 1126 nd6_llinfo_settimer(ln, 1127 (long)ND_IFINFO(rt->rt_ifp)->reachable * hz); 1128 } 1129 } 1130 1131 void 1132 nd6_rtrequest(int req, struct rtentry *rt, const struct rt_addrinfo *info) 1133 { 1134 struct sockaddr *gate = rt->rt_gateway; 1135 struct llinfo_nd6 *ln = (struct llinfo_nd6 *)rt->rt_llinfo; 1136 struct ifnet *ifp = rt->rt_ifp; 1137 uint8_t namelen = strlen(ifp->if_xname), addrlen = ifp->if_addrlen; 1138 struct ifaddr *ifa; 1139 1140 RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key); 1141 1142 if (req == RTM_LLINFO_UPD) { 1143 int rc; 1144 struct in6_addr *in6; 1145 struct in6_addr in6_all; 1146 int anycast; 1147 1148 if ((ifa = info->rti_ifa) == NULL) 1149 return; 1150 1151 in6 = &ifatoia6(ifa)->ia_addr.sin6_addr; 1152 anycast = ifatoia6(ifa)->ia6_flags & IN6_IFF_ANYCAST; 1153 1154 in6_all = in6addr_linklocal_allnodes; 1155 if ((rc = in6_setscope(&in6_all, ifa->ifa_ifp, NULL)) != 0) { 1156 log(LOG_ERR, "%s: failed to set scope %s " 1157 "(errno=%d)\n", __func__, if_name(ifp), rc); 1158 return; 1159 } 1160 1161 /* XXX don't set Override for proxy addresses */ 1162 nd6_na_output(ifa->ifa_ifp, &in6_all, in6, 1163 (anycast ? 0 : ND_NA_FLAG_OVERRIDE) 1164 #if 0 1165 | (ip6_forwarding ? ND_NA_FLAG_ROUTER : 0) 1166 #endif 1167 , 1, NULL); 1168 return; 1169 } 1170 1171 if ((rt->rt_flags & RTF_GATEWAY) != 0) 1172 return; 1173 1174 if (nd6_need_cache(ifp) == 0 && (rt->rt_flags & RTF_HOST) == 0) { 1175 RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key); 1176 /* 1177 * This is probably an interface direct route for a link 1178 * which does not need neighbor caches (e.g. fe80::%lo0/64). 1179 * We do not need special treatment below for such a route. 1180 * Moreover, the RTF_LLINFO flag which would be set below 1181 * would annoy the ndp(8) command. 1182 */ 1183 return; 1184 } 1185 1186 if (req == RTM_RESOLVE && 1187 (nd6_need_cache(ifp) == 0 || /* stf case */ 1188 !nd6_is_addr_neighbor(satocsin6(rt_getkey(rt)), ifp))) { 1189 RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key); 1190 /* 1191 * FreeBSD and BSD/OS often make a cloned host route based 1192 * on a less-specific route (e.g. the default route). 1193 * If the less specific route does not have a "gateway" 1194 * (this is the case when the route just goes to a p2p or an 1195 * stf interface), we'll mistakenly make a neighbor cache for 1196 * the host route, and will see strange neighbor solicitation 1197 * for the corresponding destination. In order to avoid the 1198 * confusion, we check if the destination of the route is 1199 * a neighbor in terms of neighbor discovery, and stop the 1200 * process if not. Additionally, we remove the LLINFO flag 1201 * so that ndp(8) will not try to get the neighbor information 1202 * of the destination. 1203 */ 1204 rt->rt_flags &= ~RTF_LLINFO; 1205 return; 1206 } 1207 1208 switch (req) { 1209 case RTM_ADD: 1210 RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key); 1211 /* 1212 * There is no backward compatibility :) 1213 * 1214 * if ((rt->rt_flags & RTF_HOST) == 0 && 1215 * SIN(rt_mask(rt))->sin_addr.s_addr != 0xffffffff) 1216 * rt->rt_flags |= RTF_CLONING; 1217 */ 1218 if ((rt->rt_flags & RTF_CLONING) || 1219 ((rt->rt_flags & RTF_LLINFO) && ln == NULL)) { 1220 union { 1221 struct sockaddr sa; 1222 struct sockaddr_dl sdl; 1223 struct sockaddr_storage ss; 1224 } u; 1225 /* 1226 * Case 1: This route should come from a route to 1227 * interface (RTF_CLONING case) or the route should be 1228 * treated as on-link but is currently not 1229 * (RTF_LLINFO && ln == NULL case). 1230 */ 1231 if (sockaddr_dl_init(&u.sdl, sizeof(u.ss), 1232 ifp->if_index, ifp->if_type, 1233 NULL, namelen, NULL, addrlen) == NULL) { 1234 printf("%s.%d: sockaddr_dl_init(, %zu, ) " 1235 "failed on %s\n", __func__, __LINE__, 1236 sizeof(u.ss), if_name(ifp)); 1237 } 1238 rt_setgate(rt, &u.sa); 1239 gate = rt->rt_gateway; 1240 RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key); 1241 if (ln != NULL) 1242 nd6_llinfo_settimer(ln, 0); 1243 RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key); 1244 if ((rt->rt_flags & RTF_CLONING) != 0) 1245 break; 1246 } 1247 RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key); 1248 /* 1249 * In IPv4 code, we try to annonuce new RTF_ANNOUNCE entry here. 1250 * We don't do that here since llinfo is not ready yet. 1251 * 1252 * There are also couple of other things to be discussed: 1253 * - unsolicited NA code needs improvement beforehand 1254 * - RFC2461 says we MAY send multicast unsolicited NA 1255 * (7.2.6 paragraph 4), however, it also says that we 1256 * SHOULD provide a mechanism to prevent multicast NA storm. 1257 * we don't have anything like it right now. 1258 * note that the mechanism needs a mutual agreement 1259 * between proxies, which means that we need to implement 1260 * a new protocol, or a new kludge. 1261 * - from RFC2461 6.2.4, host MUST NOT send an unsolicited NA. 1262 * we need to check ip6forwarding before sending it. 1263 * (or should we allow proxy ND configuration only for 1264 * routers? there's no mention about proxy ND from hosts) 1265 */ 1266 #if 0 1267 /* XXX it does not work */ 1268 if (rt->rt_flags & RTF_ANNOUNCE) 1269 nd6_na_output(ifp, 1270 &satocsin6(rt_getkey(rt))->sin6_addr, 1271 &satocsin6(rt_getkey(rt))->sin6_addr, 1272 ip6_forwarding ? ND_NA_FLAG_ROUTER : 0, 1273 1, NULL); 1274 #endif 1275 /* FALLTHROUGH */ 1276 case RTM_RESOLVE: 1277 if ((ifp->if_flags & (IFF_POINTOPOINT | IFF_LOOPBACK)) == 0) { 1278 RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key); 1279 /* 1280 * Address resolution isn't necessary for a point to 1281 * point link, so we can skip this test for a p2p link. 1282 */ 1283 if (gate->sa_family != AF_LINK || 1284 gate->sa_len < 1285 sockaddr_dl_measure(namelen, addrlen)) { 1286 log(LOG_DEBUG, 1287 "nd6_rtrequest: bad gateway value: %s\n", 1288 if_name(ifp)); 1289 break; 1290 } 1291 satosdl(gate)->sdl_type = ifp->if_type; 1292 satosdl(gate)->sdl_index = ifp->if_index; 1293 RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key); 1294 } 1295 if (ln != NULL) 1296 break; /* This happens on a route change */ 1297 RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key); 1298 /* 1299 * Case 2: This route may come from cloning, or a manual route 1300 * add with a LL address. 1301 */ 1302 R_Malloc(ln, struct llinfo_nd6 *, sizeof(*ln)); 1303 rt->rt_llinfo = ln; 1304 RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key); 1305 if (ln == NULL) { 1306 log(LOG_DEBUG, "nd6_rtrequest: malloc failed\n"); 1307 break; 1308 } 1309 RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key); 1310 nd6_inuse++; 1311 nd6_allocated++; 1312 memset(ln, 0, sizeof(*ln)); 1313 ln->ln_rt = rt; 1314 callout_init(&ln->ln_timer_ch, CALLOUT_MPSAFE); 1315 /* this is required for "ndp" command. - shin */ 1316 if (req == RTM_ADD) { 1317 /* 1318 * gate should have some valid AF_LINK entry, 1319 * and ln->ln_expire should have some lifetime 1320 * which is specified by ndp command. 1321 */ 1322 ln->ln_state = ND6_LLINFO_REACHABLE; 1323 ln->ln_byhint = 0; 1324 } else { 1325 /* 1326 * When req == RTM_RESOLVE, rt is created and 1327 * initialized in rtrequest(), so rt_expire is 0. 1328 */ 1329 ln->ln_state = ND6_LLINFO_NOSTATE; 1330 nd6_llinfo_settimer(ln, 0); 1331 } 1332 RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key); 1333 rt->rt_flags |= RTF_LLINFO; 1334 ln->ln_next = llinfo_nd6.ln_next; 1335 llinfo_nd6.ln_next = ln; 1336 ln->ln_prev = &llinfo_nd6; 1337 ln->ln_next->ln_prev = ln; 1338 1339 RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key); 1340 /* 1341 * check if rt_getkey(rt) is an address assigned 1342 * to the interface. 1343 */ 1344 ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp, 1345 &satocsin6(rt_getkey(rt))->sin6_addr); 1346 RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key); 1347 if (ifa != NULL) { 1348 const void *mac; 1349 nd6_llinfo_settimer(ln, -1); 1350 ln->ln_state = ND6_LLINFO_REACHABLE; 1351 ln->ln_byhint = 0; 1352 if ((mac = nd6_ifptomac(ifp)) != NULL) { 1353 /* XXX check for error */ 1354 if (sockaddr_dl_setaddr(satosdl(gate), 1355 gate->sa_len, mac, 1356 ifp->if_addrlen) == NULL) { 1357 printf("%s.%d: " 1358 "sockaddr_dl_setaddr(, %d, ) " 1359 "failed on %s\n", __func__, 1360 __LINE__, gate->sa_len, 1361 if_name(ifp)); 1362 } 1363 } 1364 if (nd6_useloopback) { 1365 ifp = rt->rt_ifp = lo0ifp; /* XXX */ 1366 /* 1367 * Make sure rt_ifa be equal to the ifaddr 1368 * corresponding to the address. 1369 * We need this because when we refer 1370 * rt_ifa->ia6_flags in ip6_input, we assume 1371 * that the rt_ifa points to the address instead 1372 * of the loopback address. 1373 */ 1374 if (ifa != rt->rt_ifa) 1375 rt_replace_ifa(rt, ifa); 1376 rt->rt_flags &= ~RTF_CLONED; 1377 } 1378 } else if (rt->rt_flags & RTF_ANNOUNCE) { 1379 nd6_llinfo_settimer(ln, -1); 1380 ln->ln_state = ND6_LLINFO_REACHABLE; 1381 ln->ln_byhint = 0; 1382 1383 /* join solicited node multicast for proxy ND */ 1384 if (ifp->if_flags & IFF_MULTICAST) { 1385 struct in6_addr llsol; 1386 int error; 1387 1388 llsol = satocsin6(rt_getkey(rt))->sin6_addr; 1389 llsol.s6_addr32[0] = htonl(0xff020000); 1390 llsol.s6_addr32[1] = 0; 1391 llsol.s6_addr32[2] = htonl(1); 1392 llsol.s6_addr8[12] = 0xff; 1393 if (in6_setscope(&llsol, ifp, NULL)) 1394 break; 1395 if (!in6_addmulti(&llsol, ifp, &error, 0)) { 1396 nd6log((LOG_ERR, "%s: failed to join " 1397 "%s (errno=%d)\n", if_name(ifp), 1398 ip6_sprintf(&llsol), error)); 1399 } 1400 } 1401 } 1402 break; 1403 1404 case RTM_DELETE: 1405 if (ln == NULL) 1406 break; 1407 /* leave from solicited node multicast for proxy ND */ 1408 if ((rt->rt_flags & RTF_ANNOUNCE) != 0 && 1409 (ifp->if_flags & IFF_MULTICAST) != 0) { 1410 struct in6_addr llsol; 1411 struct in6_multi *in6m; 1412 1413 llsol = satocsin6(rt_getkey(rt))->sin6_addr; 1414 llsol.s6_addr32[0] = htonl(0xff020000); 1415 llsol.s6_addr32[1] = 0; 1416 llsol.s6_addr32[2] = htonl(1); 1417 llsol.s6_addr8[12] = 0xff; 1418 if (in6_setscope(&llsol, ifp, NULL) == 0) { 1419 IN6_LOOKUP_MULTI(llsol, ifp, in6m); 1420 if (in6m) 1421 in6_delmulti(in6m); 1422 } 1423 } 1424 nd6_inuse--; 1425 ln->ln_next->ln_prev = ln->ln_prev; 1426 ln->ln_prev->ln_next = ln->ln_next; 1427 ln->ln_prev = NULL; 1428 nd6_llinfo_settimer(ln, -1); 1429 rt->rt_llinfo = 0; 1430 rt->rt_flags &= ~RTF_LLINFO; 1431 clear_llinfo_pqueue(ln); 1432 Free(ln); 1433 } 1434 } 1435 1436 int 1437 nd6_ioctl(u_long cmd, void *data, struct ifnet *ifp) 1438 { 1439 struct in6_drlist *drl = (struct in6_drlist *)data; 1440 struct in6_oprlist *oprl = (struct in6_oprlist *)data; 1441 struct in6_ndireq *ndi = (struct in6_ndireq *)data; 1442 struct in6_nbrinfo *nbi = (struct in6_nbrinfo *)data; 1443 struct in6_ndifreq *ndif = (struct in6_ndifreq *)data; 1444 struct nd_defrouter *dr; 1445 struct nd_prefix *pr; 1446 struct rtentry *rt; 1447 int i = 0, error = 0; 1448 int s; 1449 1450 switch (cmd) { 1451 case SIOCGDRLST_IN6: 1452 /* 1453 * obsolete API, use sysctl under net.inet6.icmp6 1454 */ 1455 memset(drl, 0, sizeof(*drl)); 1456 s = splsoftnet(); 1457 TAILQ_FOREACH(dr, &nd_defrouter, dr_entry) { 1458 if (i >= DRLSTSIZ) 1459 break; 1460 drl->defrouter[i].rtaddr = dr->rtaddr; 1461 in6_clearscope(&drl->defrouter[i].rtaddr); 1462 1463 drl->defrouter[i].flags = dr->flags; 1464 drl->defrouter[i].rtlifetime = dr->rtlifetime; 1465 drl->defrouter[i].expire = dr->expire; 1466 drl->defrouter[i].if_index = dr->ifp->if_index; 1467 i++; 1468 } 1469 splx(s); 1470 break; 1471 case SIOCGPRLST_IN6: 1472 /* 1473 * obsolete API, use sysctl under net.inet6.icmp6 1474 * 1475 * XXX the structure in6_prlist was changed in backward- 1476 * incompatible manner. in6_oprlist is used for SIOCGPRLST_IN6, 1477 * in6_prlist is used for nd6_sysctl() - fill_prlist(). 1478 */ 1479 /* 1480 * XXX meaning of fields, especialy "raflags", is very 1481 * differnet between RA prefix list and RR/static prefix list. 1482 * how about separating ioctls into two? 1483 */ 1484 memset(oprl, 0, sizeof(*oprl)); 1485 s = splsoftnet(); 1486 LIST_FOREACH(pr, &nd_prefix, ndpr_entry) { 1487 struct nd_pfxrouter *pfr; 1488 int j; 1489 1490 if (i >= PRLSTSIZ) 1491 break; 1492 oprl->prefix[i].prefix = pr->ndpr_prefix.sin6_addr; 1493 oprl->prefix[i].raflags = pr->ndpr_raf; 1494 oprl->prefix[i].prefixlen = pr->ndpr_plen; 1495 oprl->prefix[i].vltime = pr->ndpr_vltime; 1496 oprl->prefix[i].pltime = pr->ndpr_pltime; 1497 oprl->prefix[i].if_index = pr->ndpr_ifp->if_index; 1498 if (pr->ndpr_vltime == ND6_INFINITE_LIFETIME) 1499 oprl->prefix[i].expire = 0; 1500 else { 1501 time_t maxexpire; 1502 1503 /* XXX: we assume time_t is signed. */ 1504 maxexpire = (-1) & 1505 ~((time_t)1 << 1506 ((sizeof(maxexpire) * 8) - 1)); 1507 if (pr->ndpr_vltime < 1508 maxexpire - pr->ndpr_lastupdate) { 1509 oprl->prefix[i].expire = 1510 pr->ndpr_lastupdate + 1511 pr->ndpr_vltime; 1512 } else 1513 oprl->prefix[i].expire = maxexpire; 1514 } 1515 1516 j = 0; 1517 LIST_FOREACH(pfr, &pr->ndpr_advrtrs, pfr_entry) { 1518 if (j < DRLSTSIZ) { 1519 #define RTRADDR oprl->prefix[i].advrtr[j] 1520 RTRADDR = pfr->router->rtaddr; 1521 in6_clearscope(&RTRADDR); 1522 #undef RTRADDR 1523 } 1524 j++; 1525 } 1526 oprl->prefix[i].advrtrs = j; 1527 oprl->prefix[i].origin = PR_ORIG_RA; 1528 1529 i++; 1530 } 1531 splx(s); 1532 1533 break; 1534 case OSIOCGIFINFO_IN6: 1535 #define ND ndi->ndi 1536 /* XXX: old ndp(8) assumes a positive value for linkmtu. */ 1537 memset(&ND, 0, sizeof(ND)); 1538 ND.linkmtu = IN6_LINKMTU(ifp); 1539 ND.maxmtu = ND_IFINFO(ifp)->maxmtu; 1540 ND.basereachable = ND_IFINFO(ifp)->basereachable; 1541 ND.reachable = ND_IFINFO(ifp)->reachable; 1542 ND.retrans = ND_IFINFO(ifp)->retrans; 1543 ND.flags = ND_IFINFO(ifp)->flags; 1544 ND.recalctm = ND_IFINFO(ifp)->recalctm; 1545 ND.chlim = ND_IFINFO(ifp)->chlim; 1546 break; 1547 case SIOCGIFINFO_IN6: 1548 ND = *ND_IFINFO(ifp); 1549 break; 1550 case SIOCSIFINFO_IN6: 1551 /* 1552 * used to change host variables from userland. 1553 * intented for a use on router to reflect RA configurations. 1554 */ 1555 /* 0 means 'unspecified' */ 1556 if (ND.linkmtu != 0) { 1557 if (ND.linkmtu < IPV6_MMTU || 1558 ND.linkmtu > IN6_LINKMTU(ifp)) { 1559 error = EINVAL; 1560 break; 1561 } 1562 ND_IFINFO(ifp)->linkmtu = ND.linkmtu; 1563 } 1564 1565 if (ND.basereachable != 0) { 1566 int obasereachable = ND_IFINFO(ifp)->basereachable; 1567 1568 ND_IFINFO(ifp)->basereachable = ND.basereachable; 1569 if (ND.basereachable != obasereachable) 1570 ND_IFINFO(ifp)->reachable = 1571 ND_COMPUTE_RTIME(ND.basereachable); 1572 } 1573 if (ND.retrans != 0) 1574 ND_IFINFO(ifp)->retrans = ND.retrans; 1575 if (ND.chlim != 0) 1576 ND_IFINFO(ifp)->chlim = ND.chlim; 1577 /* FALLTHROUGH */ 1578 case SIOCSIFINFO_FLAGS: 1579 ND_IFINFO(ifp)->flags = ND.flags; 1580 break; 1581 #undef ND 1582 case SIOCSNDFLUSH_IN6: /* XXX: the ioctl name is confusing... */ 1583 /* sync kernel routing table with the default router list */ 1584 defrouter_reset(); 1585 defrouter_select(); 1586 break; 1587 case SIOCSPFXFLUSH_IN6: 1588 { 1589 /* flush all the prefix advertised by routers */ 1590 struct nd_prefix *pfx, *next; 1591 1592 s = splsoftnet(); 1593 LIST_FOREACH_SAFE(pfx, &nd_prefix, ndpr_entry, next) { 1594 struct in6_ifaddr *ia, *ia_next; 1595 1596 if (IN6_IS_ADDR_LINKLOCAL(&pfx->ndpr_prefix.sin6_addr)) 1597 continue; /* XXX */ 1598 1599 /* do we really have to remove addresses as well? */ 1600 for (ia = in6_ifaddr; ia; ia = ia_next) { 1601 /* ia might be removed. keep the next ptr. */ 1602 ia_next = ia->ia_next; 1603 1604 if ((ia->ia6_flags & IN6_IFF_AUTOCONF) == 0) 1605 continue; 1606 1607 if (ia->ia6_ndpr == pfx) 1608 in6_purgeaddr(&ia->ia_ifa); 1609 } 1610 prelist_remove(pfx); 1611 } 1612 splx(s); 1613 break; 1614 } 1615 case SIOCSRTRFLUSH_IN6: 1616 { 1617 /* flush all the default routers */ 1618 struct nd_defrouter *drtr, *next; 1619 1620 s = splsoftnet(); 1621 defrouter_reset(); 1622 TAILQ_FOREACH_SAFE(drtr, &nd_defrouter, dr_entry, next) { 1623 defrtrlist_del(drtr); 1624 } 1625 defrouter_select(); 1626 splx(s); 1627 break; 1628 } 1629 case SIOCGNBRINFO_IN6: 1630 { 1631 struct llinfo_nd6 *ln; 1632 struct in6_addr nb_addr = nbi->addr; /* make local for safety */ 1633 1634 if ((error = in6_setscope(&nb_addr, ifp, NULL)) != 0) 1635 return error; 1636 1637 s = splsoftnet(); 1638 if ((rt = nd6_lookup(&nb_addr, 0, ifp)) == NULL || 1639 (ln = (struct llinfo_nd6 *)rt->rt_llinfo) == NULL) { 1640 error = EINVAL; 1641 splx(s); 1642 break; 1643 } 1644 nbi->state = ln->ln_state; 1645 nbi->asked = ln->ln_asked; 1646 nbi->isrouter = ln->ln_router; 1647 nbi->expire = ln->ln_expire; 1648 splx(s); 1649 1650 break; 1651 } 1652 case SIOCGDEFIFACE_IN6: /* XXX: should be implemented as a sysctl? */ 1653 ndif->ifindex = nd6_defifindex; 1654 break; 1655 case SIOCSDEFIFACE_IN6: /* XXX: should be implemented as a sysctl? */ 1656 return nd6_setdefaultiface(ndif->ifindex); 1657 } 1658 return error; 1659 } 1660 1661 void 1662 nd6_llinfo_release_pkts(struct llinfo_nd6 *ln, struct ifnet *ifp, 1663 struct rtentry *rt) 1664 { 1665 struct mbuf *m_hold, *m_hold_next; 1666 1667 for (m_hold = ln->ln_hold, ln->ln_hold = NULL; 1668 m_hold != NULL; 1669 m_hold = m_hold_next) { 1670 m_hold_next = m_hold->m_nextpkt; 1671 m_hold->m_nextpkt = NULL; 1672 1673 /* 1674 * we assume ifp is not a p2p here, so 1675 * just set the 2nd argument as the 1676 * 1st one. 1677 */ 1678 nd6_output(ifp, ifp, m_hold, satocsin6(rt_getkey(rt)), rt); 1679 } 1680 } 1681 1682 /* 1683 * Create neighbor cache entry and cache link-layer address, 1684 * on reception of inbound ND6 packets. (RS/RA/NS/redirect) 1685 */ 1686 struct rtentry * 1687 nd6_cache_lladdr( 1688 struct ifnet *ifp, 1689 struct in6_addr *from, 1690 char *lladdr, 1691 int lladdrlen, 1692 int type, /* ICMP6 type */ 1693 int code /* type dependent information */ 1694 ) 1695 { 1696 struct nd_ifinfo *ndi = ND_IFINFO(ifp); 1697 struct rtentry *rt = NULL; 1698 struct llinfo_nd6 *ln = NULL; 1699 int is_newentry; 1700 struct sockaddr_dl *sdl = NULL; 1701 int do_update; 1702 int olladdr; 1703 int llchange; 1704 int newstate = 0; 1705 1706 if (ifp == NULL) 1707 panic("ifp == NULL in nd6_cache_lladdr"); 1708 if (from == NULL) 1709 panic("from == NULL in nd6_cache_lladdr"); 1710 1711 /* nothing must be updated for unspecified address */ 1712 if (IN6_IS_ADDR_UNSPECIFIED(from)) 1713 return NULL; 1714 1715 /* 1716 * Validation about ifp->if_addrlen and lladdrlen must be done in 1717 * the caller. 1718 * 1719 * XXX If the link does not have link-layer adderss, what should 1720 * we do? (ifp->if_addrlen == 0) 1721 * Spec says nothing in sections for RA, RS and NA. There's small 1722 * description on it in NS section (RFC 2461 7.2.3). 1723 */ 1724 1725 rt = nd6_lookup(from, 0, ifp); 1726 if (rt == NULL) { 1727 #if 0 1728 /* nothing must be done if there's no lladdr */ 1729 if (!lladdr || !lladdrlen) 1730 return NULL; 1731 #endif 1732 1733 rt = nd6_lookup(from, 1, ifp); 1734 is_newentry = 1; 1735 } else { 1736 /* do nothing if static ndp is set */ 1737 if (rt->rt_flags & RTF_STATIC) 1738 return NULL; 1739 is_newentry = 0; 1740 } 1741 1742 if (rt == NULL) 1743 return NULL; 1744 if ((rt->rt_flags & (RTF_GATEWAY | RTF_LLINFO)) != RTF_LLINFO) { 1745 fail: 1746 (void)nd6_free(rt, 0); 1747 return NULL; 1748 } 1749 ln = (struct llinfo_nd6 *)rt->rt_llinfo; 1750 if (ln == NULL) 1751 goto fail; 1752 if (rt->rt_gateway == NULL) 1753 goto fail; 1754 if (rt->rt_gateway->sa_family != AF_LINK) 1755 goto fail; 1756 sdl = satosdl(rt->rt_gateway); 1757 1758 olladdr = (sdl->sdl_alen) ? 1 : 0; 1759 if (olladdr && lladdr) { 1760 if (memcmp(lladdr, CLLADDR(sdl), ifp->if_addrlen)) 1761 llchange = 1; 1762 else 1763 llchange = 0; 1764 } else 1765 llchange = 0; 1766 1767 /* 1768 * newentry olladdr lladdr llchange (*=record) 1769 * 0 n n -- (1) 1770 * 0 y n -- (2) 1771 * 0 n y -- (3) * STALE 1772 * 0 y y n (4) * 1773 * 0 y y y (5) * STALE 1774 * 1 -- n -- (6) NOSTATE(= PASSIVE) 1775 * 1 -- y -- (7) * STALE 1776 */ 1777 1778 if (lladdr) { /* (3-5) and (7) */ 1779 /* 1780 * Record source link-layer address 1781 * XXX is it dependent to ifp->if_type? 1782 */ 1783 /* XXX check for error */ 1784 if (sockaddr_dl_setaddr(sdl, sdl->sdl_len, lladdr, 1785 ifp->if_addrlen) == NULL) { 1786 printf("%s.%d: sockaddr_dl_setaddr(, %d, ) " 1787 "failed on %s\n", __func__, __LINE__, 1788 sdl->sdl_len, if_name(ifp)); 1789 } 1790 } 1791 1792 if (!is_newentry) { 1793 if ((!olladdr && lladdr) || /* (3) */ 1794 (olladdr && lladdr && llchange)) { /* (5) */ 1795 do_update = 1; 1796 newstate = ND6_LLINFO_STALE; 1797 } else /* (1-2,4) */ 1798 do_update = 0; 1799 } else { 1800 do_update = 1; 1801 if (lladdr == NULL) /* (6) */ 1802 newstate = ND6_LLINFO_NOSTATE; 1803 else /* (7) */ 1804 newstate = ND6_LLINFO_STALE; 1805 } 1806 1807 if (do_update) { 1808 /* 1809 * Update the state of the neighbor cache. 1810 */ 1811 ln->ln_state = newstate; 1812 1813 if (ln->ln_state == ND6_LLINFO_STALE) { 1814 /* 1815 * XXX: since nd6_output() below will cause 1816 * state tansition to DELAY and reset the timer, 1817 * we must set the timer now, although it is actually 1818 * meaningless. 1819 */ 1820 nd6_llinfo_settimer(ln, (long)nd6_gctimer * hz); 1821 1822 nd6_llinfo_release_pkts(ln, ifp, rt); 1823 } else if (ln->ln_state == ND6_LLINFO_INCOMPLETE) { 1824 /* probe right away */ 1825 nd6_llinfo_settimer((void *)ln, 0); 1826 } 1827 } 1828 1829 /* 1830 * ICMP6 type dependent behavior. 1831 * 1832 * NS: clear IsRouter if new entry 1833 * RS: clear IsRouter 1834 * RA: set IsRouter if there's lladdr 1835 * redir: clear IsRouter if new entry 1836 * 1837 * RA case, (1): 1838 * The spec says that we must set IsRouter in the following cases: 1839 * - If lladdr exist, set IsRouter. This means (1-5). 1840 * - If it is old entry (!newentry), set IsRouter. This means (7). 1841 * So, based on the spec, in (1-5) and (7) cases we must set IsRouter. 1842 * A quetion arises for (1) case. (1) case has no lladdr in the 1843 * neighbor cache, this is similar to (6). 1844 * This case is rare but we figured that we MUST NOT set IsRouter. 1845 * 1846 * newentry olladdr lladdr llchange NS RS RA redir 1847 * D R 1848 * 0 n n -- (1) c ? s 1849 * 0 y n -- (2) c s s 1850 * 0 n y -- (3) c s s 1851 * 0 y y n (4) c s s 1852 * 0 y y y (5) c s s 1853 * 1 -- n -- (6) c c c s 1854 * 1 -- y -- (7) c c s c s 1855 * 1856 * (c=clear s=set) 1857 */ 1858 switch (type & 0xff) { 1859 case ND_NEIGHBOR_SOLICIT: 1860 /* 1861 * New entry must have is_router flag cleared. 1862 */ 1863 if (is_newentry) /* (6-7) */ 1864 ln->ln_router = 0; 1865 break; 1866 case ND_REDIRECT: 1867 /* 1868 * If the icmp is a redirect to a better router, always set the 1869 * is_router flag. Otherwise, if the entry is newly created, 1870 * clear the flag. [RFC 2461, sec 8.3] 1871 */ 1872 if (code == ND_REDIRECT_ROUTER) 1873 ln->ln_router = 1; 1874 else if (is_newentry) /* (6-7) */ 1875 ln->ln_router = 0; 1876 break; 1877 case ND_ROUTER_SOLICIT: 1878 /* 1879 * is_router flag must always be cleared. 1880 */ 1881 ln->ln_router = 0; 1882 break; 1883 case ND_ROUTER_ADVERT: 1884 /* 1885 * Mark an entry with lladdr as a router. 1886 */ 1887 if ((!is_newentry && (olladdr || lladdr)) || /* (2-5) */ 1888 (is_newentry && lladdr)) { /* (7) */ 1889 ln->ln_router = 1; 1890 } 1891 break; 1892 } 1893 1894 /* 1895 * When the link-layer address of a router changes, select the 1896 * best router again. In particular, when the neighbor entry is newly 1897 * created, it might affect the selection policy. 1898 * Question: can we restrict the first condition to the "is_newentry" 1899 * case? 1900 * XXX: when we hear an RA from a new router with the link-layer 1901 * address option, defrouter_select() is called twice, since 1902 * defrtrlist_update called the function as well. However, I believe 1903 * we can compromise the overhead, since it only happens the first 1904 * time. 1905 * XXX: although defrouter_select() should not have a bad effect 1906 * for those are not autoconfigured hosts, we explicitly avoid such 1907 * cases for safety. 1908 */ 1909 if (do_update && ln->ln_router && !ip6_forwarding && 1910 nd6_accepts_rtadv(ndi)) 1911 defrouter_select(); 1912 1913 return rt; 1914 } 1915 1916 static void 1917 nd6_slowtimo(void *ignored_arg) 1918 { 1919 struct nd_ifinfo *nd6if; 1920 struct ifnet *ifp; 1921 1922 mutex_enter(softnet_lock); 1923 KERNEL_LOCK(1, NULL); 1924 callout_reset(&nd6_slowtimo_ch, ND6_SLOWTIMER_INTERVAL * hz, 1925 nd6_slowtimo, NULL); 1926 TAILQ_FOREACH(ifp, &ifnet, if_list) { 1927 nd6if = ND_IFINFO(ifp); 1928 if (nd6if->basereachable && /* already initialized */ 1929 (nd6if->recalctm -= ND6_SLOWTIMER_INTERVAL) <= 0) { 1930 /* 1931 * Since reachable time rarely changes by router 1932 * advertisements, we SHOULD insure that a new random 1933 * value gets recomputed at least once every few hours. 1934 * (RFC 2461, 6.3.4) 1935 */ 1936 nd6if->recalctm = nd6_recalc_reachtm_interval; 1937 nd6if->reachable = ND_COMPUTE_RTIME(nd6if->basereachable); 1938 } 1939 } 1940 KERNEL_UNLOCK_ONE(NULL); 1941 mutex_exit(softnet_lock); 1942 } 1943 1944 #define senderr(e) { error = (e); goto bad;} 1945 int 1946 nd6_output(struct ifnet *ifp, struct ifnet *origifp, struct mbuf *m0, 1947 const struct sockaddr_in6 *dst, struct rtentry *rt0) 1948 { 1949 struct mbuf *m = m0; 1950 struct rtentry *rt = rt0; 1951 struct sockaddr_in6 *gw6 = NULL; 1952 struct llinfo_nd6 *ln = NULL; 1953 int error = 0; 1954 1955 if (IN6_IS_ADDR_MULTICAST(&dst->sin6_addr)) 1956 goto sendpkt; 1957 1958 if (nd6_need_cache(ifp) == 0) 1959 goto sendpkt; 1960 1961 /* 1962 * next hop determination. This routine is derived from ether_output. 1963 */ 1964 if (rt) { 1965 if ((rt->rt_flags & RTF_UP) == 0) { 1966 if ((rt0 = rt = rtalloc1(sin6tocsa(dst), 1)) != NULL) { 1967 rt->rt_refcnt--; 1968 if (rt->rt_ifp != ifp) 1969 senderr(EHOSTUNREACH); 1970 } else 1971 senderr(EHOSTUNREACH); 1972 } 1973 1974 if (rt->rt_flags & RTF_GATEWAY) { 1975 gw6 = (struct sockaddr_in6 *)rt->rt_gateway; 1976 1977 /* 1978 * We skip link-layer address resolution and NUD 1979 * if the gateway is not a neighbor from ND point 1980 * of view, regardless of the value of nd_ifinfo.flags. 1981 * The second condition is a bit tricky; we skip 1982 * if the gateway is our own address, which is 1983 * sometimes used to install a route to a p2p link. 1984 */ 1985 if (!nd6_is_addr_neighbor(gw6, ifp) || 1986 in6ifa_ifpwithaddr(ifp, &gw6->sin6_addr)) { 1987 /* 1988 * We allow this kind of tricky route only 1989 * when the outgoing interface is p2p. 1990 * XXX: we may need a more generic rule here. 1991 */ 1992 if ((ifp->if_flags & IFF_POINTOPOINT) == 0) 1993 senderr(EHOSTUNREACH); 1994 1995 goto sendpkt; 1996 } 1997 1998 if (rt->rt_gwroute == NULL) 1999 goto lookup; 2000 if (((rt = rt->rt_gwroute)->rt_flags & RTF_UP) == 0) { 2001 rtfree(rt); rt = rt0; 2002 lookup: 2003 rt->rt_gwroute = rtalloc1(rt->rt_gateway, 1); 2004 if ((rt = rt->rt_gwroute) == NULL) 2005 senderr(EHOSTUNREACH); 2006 /* the "G" test below also prevents rt == rt0 */ 2007 if ((rt->rt_flags & RTF_GATEWAY) || 2008 (rt->rt_ifp != ifp)) { 2009 rt->rt_refcnt--; 2010 rt0->rt_gwroute = NULL; 2011 senderr(EHOSTUNREACH); 2012 } 2013 } 2014 } 2015 } 2016 2017 /* 2018 * Address resolution or Neighbor Unreachability Detection 2019 * for the next hop. 2020 * At this point, the destination of the packet must be a unicast 2021 * or an anycast address(i.e. not a multicast). 2022 */ 2023 2024 /* Look up the neighbor cache for the nexthop */ 2025 if (rt != NULL && (rt->rt_flags & RTF_LLINFO) != 0) 2026 ln = (struct llinfo_nd6 *)rt->rt_llinfo; 2027 else { 2028 /* 2029 * Since nd6_is_addr_neighbor() internally calls nd6_lookup(), 2030 * the condition below is not very efficient. But we believe 2031 * it is tolerable, because this should be a rare case. 2032 */ 2033 if (nd6_is_addr_neighbor(dst, ifp) && 2034 (rt = nd6_lookup(&dst->sin6_addr, 1, ifp)) != NULL) 2035 ln = (struct llinfo_nd6 *)rt->rt_llinfo; 2036 } 2037 if (ln == NULL || rt == NULL) { 2038 if ((ifp->if_flags & IFF_POINTOPOINT) == 0 && 2039 !(ND_IFINFO(ifp)->flags & ND6_IFF_PERFORMNUD)) { 2040 log(LOG_DEBUG, 2041 "nd6_output: can't allocate llinfo for %s " 2042 "(ln=%p, rt=%p)\n", 2043 ip6_sprintf(&dst->sin6_addr), ln, rt); 2044 senderr(EIO); /* XXX: good error? */ 2045 } 2046 2047 goto sendpkt; /* send anyway */ 2048 } 2049 2050 /* We don't have to do link-layer address resolution on a p2p link. */ 2051 if ((ifp->if_flags & IFF_POINTOPOINT) != 0 && 2052 ln->ln_state < ND6_LLINFO_REACHABLE) { 2053 ln->ln_state = ND6_LLINFO_STALE; 2054 nd6_llinfo_settimer(ln, (long)nd6_gctimer * hz); 2055 } 2056 2057 /* 2058 * The first time we send a packet to a neighbor whose entry is 2059 * STALE, we have to change the state to DELAY and a sets a timer to 2060 * expire in DELAY_FIRST_PROBE_TIME seconds to ensure do 2061 * neighbor unreachability detection on expiration. 2062 * (RFC 2461 7.3.3) 2063 */ 2064 if (ln->ln_state == ND6_LLINFO_STALE) { 2065 ln->ln_asked = 0; 2066 ln->ln_state = ND6_LLINFO_DELAY; 2067 nd6_llinfo_settimer(ln, (long)nd6_delay * hz); 2068 } 2069 2070 /* 2071 * If the neighbor cache entry has a state other than INCOMPLETE 2072 * (i.e. its link-layer address is already resolved), just 2073 * send the packet. 2074 */ 2075 if (ln->ln_state > ND6_LLINFO_INCOMPLETE) 2076 goto sendpkt; 2077 2078 /* 2079 * There is a neighbor cache entry, but no ethernet address 2080 * response yet. Append this latest packet to the end of the 2081 * packet queue in the mbuf, unless the number of the packet 2082 * does not exceed nd6_maxqueuelen. When it exceeds nd6_maxqueuelen, 2083 * the oldest packet in the queue will be removed. 2084 */ 2085 if (ln->ln_state == ND6_LLINFO_NOSTATE) 2086 ln->ln_state = ND6_LLINFO_INCOMPLETE; 2087 if (ln->ln_hold) { 2088 struct mbuf *m_hold; 2089 int i; 2090 2091 i = 0; 2092 for (m_hold = ln->ln_hold; m_hold; m_hold = m_hold->m_nextpkt) { 2093 i++; 2094 if (m_hold->m_nextpkt == NULL) { 2095 m_hold->m_nextpkt = m; 2096 break; 2097 } 2098 } 2099 while (i >= nd6_maxqueuelen) { 2100 m_hold = ln->ln_hold; 2101 ln->ln_hold = ln->ln_hold->m_nextpkt; 2102 m_freem(m_hold); 2103 i--; 2104 } 2105 } else { 2106 ln->ln_hold = m; 2107 } 2108 2109 /* 2110 * If there has been no NS for the neighbor after entering the 2111 * INCOMPLETE state, send the first solicitation. 2112 */ 2113 if (!ND6_LLINFO_PERMANENT(ln) && ln->ln_asked == 0) { 2114 ln->ln_asked++; 2115 nd6_llinfo_settimer(ln, 2116 (long)ND_IFINFO(ifp)->retrans * hz / 1000); 2117 nd6_ns_output(ifp, NULL, &dst->sin6_addr, ln, 0); 2118 } 2119 return 0; 2120 2121 sendpkt: 2122 /* discard the packet if IPv6 operation is disabled on the interface */ 2123 if ((ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED)) { 2124 error = ENETDOWN; /* better error? */ 2125 goto bad; 2126 } 2127 2128 #ifdef KAME_IPSEC 2129 /* clean ipsec history once it goes out of the node */ 2130 ipsec_delaux(m); 2131 #endif 2132 if ((ifp->if_flags & IFF_LOOPBACK) != 0) 2133 return (*ifp->if_output)(origifp, m, sin6tocsa(dst), rt); 2134 return (*ifp->if_output)(ifp, m, sin6tocsa(dst), rt); 2135 2136 bad: 2137 if (m != NULL) 2138 m_freem(m); 2139 return error; 2140 } 2141 #undef senderr 2142 2143 int 2144 nd6_need_cache(struct ifnet *ifp) 2145 { 2146 /* 2147 * XXX: we currently do not make neighbor cache on any interface 2148 * other than ARCnet, Ethernet, FDDI and GIF. 2149 * 2150 * RFC2893 says: 2151 * - unidirectional tunnels needs no ND 2152 */ 2153 switch (ifp->if_type) { 2154 case IFT_ARCNET: 2155 case IFT_ETHER: 2156 case IFT_FDDI: 2157 case IFT_IEEE1394: 2158 case IFT_CARP: 2159 case IFT_GIF: /* XXX need more cases? */ 2160 case IFT_PPP: 2161 case IFT_TUNNEL: 2162 return 1; 2163 default: 2164 return 0; 2165 } 2166 } 2167 2168 int 2169 nd6_storelladdr(const struct ifnet *ifp, const struct rtentry *rt, 2170 struct mbuf *m, const struct sockaddr *dst, uint8_t *lldst, 2171 size_t dstsize) 2172 { 2173 const struct sockaddr_dl *sdl; 2174 2175 if (m->m_flags & M_MCAST) { 2176 switch (ifp->if_type) { 2177 case IFT_ETHER: 2178 case IFT_FDDI: 2179 ETHER_MAP_IPV6_MULTICAST(&satocsin6(dst)->sin6_addr, 2180 lldst); 2181 return 1; 2182 case IFT_IEEE1394: 2183 memcpy(lldst, ifp->if_broadcastaddr, 2184 MIN(dstsize, ifp->if_addrlen)); 2185 return 1; 2186 case IFT_ARCNET: 2187 *lldst = 0; 2188 return 1; 2189 default: 2190 m_freem(m); 2191 return 0; 2192 } 2193 } 2194 2195 if (rt == NULL) { 2196 /* this could happen, if we could not allocate memory */ 2197 m_freem(m); 2198 return 0; 2199 } 2200 if (rt->rt_gateway->sa_family != AF_LINK) { 2201 printf("%s: something odd happens\n", __func__); 2202 m_freem(m); 2203 return 0; 2204 } 2205 sdl = satocsdl(rt->rt_gateway); 2206 if (sdl->sdl_alen == 0 || sdl->sdl_alen > dstsize) { 2207 /* this should be impossible, but we bark here for debugging */ 2208 printf("%s: sdl_alen == %" PRIu8 ", dst=%s, if=%s\n", __func__, 2209 sdl->sdl_alen, ip6_sprintf(&satocsin6(dst)->sin6_addr), 2210 if_name(ifp)); 2211 m_freem(m); 2212 return 0; 2213 } 2214 2215 memcpy(lldst, CLLADDR(sdl), MIN(dstsize, sdl->sdl_alen)); 2216 return 1; 2217 } 2218 2219 static void 2220 clear_llinfo_pqueue(struct llinfo_nd6 *ln) 2221 { 2222 struct mbuf *m_hold, *m_hold_next; 2223 2224 for (m_hold = ln->ln_hold; m_hold; m_hold = m_hold_next) { 2225 m_hold_next = m_hold->m_nextpkt; 2226 m_hold->m_nextpkt = NULL; 2227 m_freem(m_hold); 2228 } 2229 2230 ln->ln_hold = NULL; 2231 return; 2232 } 2233 2234 int 2235 nd6_sysctl( 2236 int name, 2237 void *oldp, /* syscall arg, need copyout */ 2238 size_t *oldlenp, 2239 void *newp, /* syscall arg, need copyin */ 2240 size_t newlen 2241 ) 2242 { 2243 void *p; 2244 size_t ol; 2245 int error; 2246 2247 error = 0; 2248 2249 if (newp) 2250 return EPERM; 2251 if (oldp && !oldlenp) 2252 return EINVAL; 2253 ol = oldlenp ? *oldlenp : 0; 2254 2255 if (oldp) { 2256 p = malloc(*oldlenp, M_TEMP, M_WAITOK); 2257 if (p == NULL) 2258 return ENOMEM; 2259 } else 2260 p = NULL; 2261 switch (name) { 2262 case ICMPV6CTL_ND6_DRLIST: 2263 error = fill_drlist(p, oldlenp, ol); 2264 if (!error && p != NULL && oldp != NULL) 2265 error = copyout(p, oldp, *oldlenp); 2266 break; 2267 2268 case ICMPV6CTL_ND6_PRLIST: 2269 error = fill_prlist(p, oldlenp, ol); 2270 if (!error && p != NULL && oldp != NULL) 2271 error = copyout(p, oldp, *oldlenp); 2272 break; 2273 2274 case ICMPV6CTL_ND6_MAXQLEN: 2275 break; 2276 2277 default: 2278 error = ENOPROTOOPT; 2279 break; 2280 } 2281 if (p) 2282 free(p, M_TEMP); 2283 2284 return error; 2285 } 2286 2287 static int 2288 fill_drlist(void *oldp, size_t *oldlenp, size_t ol) 2289 { 2290 int error = 0, s; 2291 struct in6_defrouter *d = NULL, *de = NULL; 2292 struct nd_defrouter *dr; 2293 size_t l; 2294 2295 s = splsoftnet(); 2296 2297 if (oldp) { 2298 d = (struct in6_defrouter *)oldp; 2299 de = (struct in6_defrouter *)((char *)oldp + *oldlenp); 2300 } 2301 l = 0; 2302 2303 TAILQ_FOREACH(dr, &nd_defrouter, dr_entry) { 2304 2305 if (oldp && d + 1 <= de) { 2306 memset(d, 0, sizeof(*d)); 2307 sockaddr_in6_init(&d->rtaddr, &dr->rtaddr, 0, 0, 0); 2308 if (sa6_recoverscope(&d->rtaddr)) { 2309 log(LOG_ERR, 2310 "scope error in router list (%s)\n", 2311 ip6_sprintf(&d->rtaddr.sin6_addr)); 2312 /* XXX: press on... */ 2313 } 2314 d->flags = dr->flags; 2315 d->rtlifetime = dr->rtlifetime; 2316 d->expire = dr->expire; 2317 d->if_index = dr->ifp->if_index; 2318 } 2319 2320 l += sizeof(*d); 2321 if (d) 2322 d++; 2323 } 2324 2325 if (oldp) { 2326 if (l > ol) 2327 error = ENOMEM; 2328 } 2329 if (oldlenp) 2330 *oldlenp = l; /* (void *)d - (void *)oldp */ 2331 2332 splx(s); 2333 2334 return error; 2335 } 2336 2337 static int 2338 fill_prlist(void *oldp, size_t *oldlenp, size_t ol) 2339 { 2340 int error = 0, s; 2341 struct nd_prefix *pr; 2342 struct in6_prefix *p = NULL; 2343 struct in6_prefix *pe = NULL; 2344 size_t l; 2345 2346 s = splsoftnet(); 2347 2348 if (oldp) { 2349 p = (struct in6_prefix *)oldp; 2350 pe = (struct in6_prefix *)((char *)oldp + *oldlenp); 2351 } 2352 l = 0; 2353 2354 LIST_FOREACH(pr, &nd_prefix, ndpr_entry) { 2355 u_short advrtrs; 2356 size_t advance; 2357 struct sockaddr_in6 *sin6; 2358 struct sockaddr_in6 *s6; 2359 struct nd_pfxrouter *pfr; 2360 2361 if (oldp && p + 1 <= pe) 2362 { 2363 memset(p, 0, sizeof(*p)); 2364 sin6 = (struct sockaddr_in6 *)(p + 1); 2365 2366 p->prefix = pr->ndpr_prefix; 2367 if (sa6_recoverscope(&p->prefix)) { 2368 log(LOG_ERR, 2369 "scope error in prefix list (%s)\n", 2370 ip6_sprintf(&p->prefix.sin6_addr)); 2371 /* XXX: press on... */ 2372 } 2373 p->raflags = pr->ndpr_raf; 2374 p->prefixlen = pr->ndpr_plen; 2375 p->vltime = pr->ndpr_vltime; 2376 p->pltime = pr->ndpr_pltime; 2377 p->if_index = pr->ndpr_ifp->if_index; 2378 if (pr->ndpr_vltime == ND6_INFINITE_LIFETIME) 2379 p->expire = 0; 2380 else { 2381 time_t maxexpire; 2382 2383 /* XXX: we assume time_t is signed. */ 2384 maxexpire = (-1) & 2385 ~((time_t)1 << 2386 ((sizeof(maxexpire) * 8) - 1)); 2387 if (pr->ndpr_vltime < 2388 maxexpire - pr->ndpr_lastupdate) { 2389 p->expire = pr->ndpr_lastupdate + 2390 pr->ndpr_vltime; 2391 } else 2392 p->expire = maxexpire; 2393 } 2394 p->refcnt = pr->ndpr_refcnt; 2395 p->flags = pr->ndpr_stateflags; 2396 p->origin = PR_ORIG_RA; 2397 advrtrs = 0; 2398 LIST_FOREACH(pfr, &pr->ndpr_advrtrs, pfr_entry) { 2399 if ((void *)&sin6[advrtrs + 1] > (void *)pe) { 2400 advrtrs++; 2401 continue; 2402 } 2403 s6 = &sin6[advrtrs]; 2404 sockaddr_in6_init(s6, &pfr->router->rtaddr, 2405 0, 0, 0); 2406 if (sa6_recoverscope(s6)) { 2407 log(LOG_ERR, 2408 "scope error in " 2409 "prefix list (%s)\n", 2410 ip6_sprintf(&pfr->router->rtaddr)); 2411 } 2412 advrtrs++; 2413 } 2414 p->advrtrs = advrtrs; 2415 } 2416 else { 2417 advrtrs = 0; 2418 LIST_FOREACH(pfr, &pr->ndpr_advrtrs, pfr_entry) 2419 advrtrs++; 2420 } 2421 2422 advance = sizeof(*p) + sizeof(*sin6) * advrtrs; 2423 l += advance; 2424 if (p) 2425 p = (struct in6_prefix *)((char *)p + advance); 2426 } 2427 2428 if (oldp) { 2429 *oldlenp = l; /* (void *)d - (void *)oldp */ 2430 if (l > ol) 2431 error = ENOMEM; 2432 } else 2433 *oldlenp = l; 2434 2435 splx(s); 2436 2437 return error; 2438 } 2439