1 /* $NetBSD: nd6.c,v 1.142 2012/03/22 20:34:41 drochner 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 1.142 2012/03/22 20:34:41 drochner 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 #include <net/net_osdep.h> 73 74 #define ND6_SLOWTIMER_INTERVAL (60 * 60) /* 1 hour */ 75 #define ND6_RECALC_REACHTM_INTERVAL (60 * 120) /* 2 hours */ 76 77 /* timer values */ 78 int nd6_prune = 1; /* walk list every 1 seconds */ 79 int nd6_delay = 5; /* delay first probe time 5 second */ 80 int nd6_umaxtries = 3; /* maximum unicast query */ 81 int nd6_mmaxtries = 3; /* maximum multicast query */ 82 int nd6_useloopback = 1; /* use loopback interface for local traffic */ 83 int nd6_gctimer = (60 * 60 * 24); /* 1 day: garbage collection timer */ 84 85 /* preventing too many loops in ND option parsing */ 86 int nd6_maxndopt = 10; /* max # of ND options allowed */ 87 88 int nd6_maxnudhint = 0; /* max # of subsequent upper layer hints */ 89 90 int nd6_maxqueuelen = 1; /* max # of packets cached in unresolved ND entries */ 91 92 #ifdef ND6_DEBUG 93 int nd6_debug = 1; 94 #else 95 int nd6_debug = 0; 96 #endif 97 98 /* for debugging? */ 99 static int nd6_inuse, nd6_allocated; 100 101 struct llinfo_nd6 llinfo_nd6 = { 102 .ln_prev = &llinfo_nd6, 103 .ln_next = &llinfo_nd6, 104 }; 105 struct nd_drhead nd_defrouter; 106 struct nd_prhead nd_prefix = { 0 }; 107 108 int nd6_recalc_reachtm_interval = ND6_RECALC_REACHTM_INTERVAL; 109 static const struct sockaddr_in6 all1_sa = { 110 .sin6_family = AF_INET6 111 , .sin6_len = sizeof(struct sockaddr_in6) 112 , .sin6_addr = {.s6_addr = {0xff, 0xff, 0xff, 0xff, 113 0xff, 0xff, 0xff, 0xff, 114 0xff, 0xff, 0xff, 0xff, 115 0xff, 0xff, 0xff, 0xff}} 116 }; 117 118 static void nd6_setmtu0(struct ifnet *, struct nd_ifinfo *); 119 static void nd6_slowtimo(void *); 120 static int regen_tmpaddr(struct in6_ifaddr *); 121 static struct llinfo_nd6 *nd6_free(struct rtentry *, int); 122 static void nd6_llinfo_timer(void *); 123 static void clear_llinfo_pqueue(struct llinfo_nd6 *); 124 125 callout_t nd6_slowtimo_ch; 126 callout_t nd6_timer_ch; 127 extern callout_t in6_tmpaddrtimer_ch; 128 129 static int fill_drlist(void *, size_t *, size_t); 130 static int fill_prlist(void *, size_t *, size_t); 131 132 MALLOC_DEFINE(M_IP6NDP, "NDP", "IPv6 Neighbour Discovery"); 133 134 void 135 nd6_init(void) 136 { 137 static int nd6_init_done = 0; 138 139 if (nd6_init_done) { 140 log(LOG_NOTICE, "nd6_init called more than once(ignored)\n"); 141 return; 142 } 143 144 /* initialization of the default router list */ 145 TAILQ_INIT(&nd_defrouter); 146 147 nd6_init_done = 1; 148 149 callout_init(&nd6_slowtimo_ch, CALLOUT_MPSAFE); 150 callout_init(&nd6_timer_ch, CALLOUT_MPSAFE); 151 152 /* start timer */ 153 callout_reset(&nd6_slowtimo_ch, ND6_SLOWTIMER_INTERVAL * hz, 154 nd6_slowtimo, NULL); 155 } 156 157 struct nd_ifinfo * 158 nd6_ifattach(struct ifnet *ifp) 159 { 160 struct nd_ifinfo *nd; 161 162 nd = (struct nd_ifinfo *)malloc(sizeof(*nd), M_IP6NDP, M_WAITOK|M_ZERO); 163 164 nd->initialized = 1; 165 166 nd->chlim = IPV6_DEFHLIM; 167 nd->basereachable = REACHABLE_TIME; 168 nd->reachable = ND_COMPUTE_RTIME(nd->basereachable); 169 nd->retrans = RETRANS_TIMER; 170 /* 171 * Note that the default value of ip6_accept_rtadv is 0. 172 * Because we do not set ND6_IFF_OVERRIDE_RTADV here, we won't 173 * accept RAs by default. 174 */ 175 nd->flags = ND6_IFF_PERFORMNUD | ND6_IFF_ACCEPT_RTADV; 176 177 /* XXX: we cannot call nd6_setmtu since ifp is not fully initialized */ 178 nd6_setmtu0(ifp, nd); 179 180 return nd; 181 } 182 183 void 184 nd6_ifdetach(struct nd_ifinfo *nd) 185 { 186 187 free(nd, M_IP6NDP); 188 } 189 190 void 191 nd6_setmtu(struct ifnet *ifp) 192 { 193 nd6_setmtu0(ifp, ND_IFINFO(ifp)); 194 } 195 196 void 197 nd6_setmtu0(struct ifnet *ifp, struct nd_ifinfo *ndi) 198 { 199 u_int32_t omaxmtu; 200 201 omaxmtu = ndi->maxmtu; 202 203 switch (ifp->if_type) { 204 case IFT_ARCNET: 205 ndi->maxmtu = MIN(ARC_PHDS_MAXMTU, ifp->if_mtu); /* RFC2497 */ 206 break; 207 case IFT_FDDI: 208 ndi->maxmtu = MIN(FDDIIPMTU, ifp->if_mtu); 209 break; 210 default: 211 ndi->maxmtu = ifp->if_mtu; 212 break; 213 } 214 215 /* 216 * Decreasing the interface MTU under IPV6 minimum MTU may cause 217 * undesirable situation. We thus notify the operator of the change 218 * explicitly. The check for omaxmtu is necessary to restrict the 219 * log to the case of changing the MTU, not initializing it. 220 */ 221 if (omaxmtu >= IPV6_MMTU && ndi->maxmtu < IPV6_MMTU) { 222 log(LOG_NOTICE, "nd6_setmtu0: new link MTU on %s (%lu) is too" 223 " small for IPv6 which needs %lu\n", 224 if_name(ifp), (unsigned long)ndi->maxmtu, (unsigned long) 225 IPV6_MMTU); 226 } 227 228 if (ndi->maxmtu > in6_maxmtu) 229 in6_setmaxmtu(); /* check all interfaces just in case */ 230 } 231 232 void 233 nd6_option_init(void *opt, int icmp6len, union nd_opts *ndopts) 234 { 235 236 memset(ndopts, 0, sizeof(*ndopts)); 237 ndopts->nd_opts_search = (struct nd_opt_hdr *)opt; 238 ndopts->nd_opts_last 239 = (struct nd_opt_hdr *)(((u_char *)opt) + icmp6len); 240 241 if (icmp6len == 0) { 242 ndopts->nd_opts_done = 1; 243 ndopts->nd_opts_search = NULL; 244 } 245 } 246 247 /* 248 * Take one ND option. 249 */ 250 struct nd_opt_hdr * 251 nd6_option(union nd_opts *ndopts) 252 { 253 struct nd_opt_hdr *nd_opt; 254 int olen; 255 256 if (ndopts == NULL) 257 panic("ndopts == NULL in nd6_option"); 258 if (ndopts->nd_opts_last == NULL) 259 panic("uninitialized ndopts in nd6_option"); 260 if (ndopts->nd_opts_search == NULL) 261 return NULL; 262 if (ndopts->nd_opts_done) 263 return NULL; 264 265 nd_opt = ndopts->nd_opts_search; 266 267 /* make sure nd_opt_len is inside the buffer */ 268 if ((void *)&nd_opt->nd_opt_len >= (void *)ndopts->nd_opts_last) { 269 memset(ndopts, 0, sizeof(*ndopts)); 270 return NULL; 271 } 272 273 olen = nd_opt->nd_opt_len << 3; 274 if (olen == 0) { 275 /* 276 * Message validation requires that all included 277 * options have a length that is greater than zero. 278 */ 279 memset(ndopts, 0, sizeof(*ndopts)); 280 return NULL; 281 } 282 283 ndopts->nd_opts_search = (struct nd_opt_hdr *)((char *)nd_opt + olen); 284 if (ndopts->nd_opts_search > ndopts->nd_opts_last) { 285 /* option overruns the end of buffer, invalid */ 286 memset(ndopts, 0, sizeof(*ndopts)); 287 return NULL; 288 } else if (ndopts->nd_opts_search == ndopts->nd_opts_last) { 289 /* reached the end of options chain */ 290 ndopts->nd_opts_done = 1; 291 ndopts->nd_opts_search = NULL; 292 } 293 return nd_opt; 294 } 295 296 /* 297 * Parse multiple ND options. 298 * This function is much easier to use, for ND routines that do not need 299 * multiple options of the same type. 300 */ 301 int 302 nd6_options(union nd_opts *ndopts) 303 { 304 struct nd_opt_hdr *nd_opt; 305 int i = 0; 306 307 if (ndopts == NULL) 308 panic("ndopts == NULL in nd6_options"); 309 if (ndopts->nd_opts_last == NULL) 310 panic("uninitialized ndopts in nd6_options"); 311 if (ndopts->nd_opts_search == NULL) 312 return 0; 313 314 while (1) { 315 nd_opt = nd6_option(ndopts); 316 if (nd_opt == NULL && ndopts->nd_opts_last == NULL) { 317 /* 318 * Message validation requires that all included 319 * options have a length that is greater than zero. 320 */ 321 ICMP6_STATINC(ICMP6_STAT_ND_BADOPT); 322 memset(ndopts, 0, sizeof(*ndopts)); 323 return -1; 324 } 325 326 if (nd_opt == NULL) 327 goto skip1; 328 329 switch (nd_opt->nd_opt_type) { 330 case ND_OPT_SOURCE_LINKADDR: 331 case ND_OPT_TARGET_LINKADDR: 332 case ND_OPT_MTU: 333 case ND_OPT_REDIRECTED_HEADER: 334 if (ndopts->nd_opt_array[nd_opt->nd_opt_type]) { 335 nd6log((LOG_INFO, 336 "duplicated ND6 option found (type=%d)\n", 337 nd_opt->nd_opt_type)); 338 /* XXX bark? */ 339 } else { 340 ndopts->nd_opt_array[nd_opt->nd_opt_type] 341 = nd_opt; 342 } 343 break; 344 case ND_OPT_PREFIX_INFORMATION: 345 if (ndopts->nd_opt_array[nd_opt->nd_opt_type] == 0) { 346 ndopts->nd_opt_array[nd_opt->nd_opt_type] 347 = nd_opt; 348 } 349 ndopts->nd_opts_pi_end = 350 (struct nd_opt_prefix_info *)nd_opt; 351 break; 352 default: 353 /* 354 * Unknown options must be silently ignored, 355 * to accommodate future extension to the protocol. 356 */ 357 nd6log((LOG_DEBUG, 358 "nd6_options: unsupported option %d - " 359 "option ignored\n", nd_opt->nd_opt_type)); 360 } 361 362 skip1: 363 i++; 364 if (i > nd6_maxndopt) { 365 ICMP6_STATINC(ICMP6_STAT_ND_TOOMANYOPT); 366 nd6log((LOG_INFO, "too many loop in nd opt\n")); 367 break; 368 } 369 370 if (ndopts->nd_opts_done) 371 break; 372 } 373 374 return 0; 375 } 376 377 /* 378 * ND6 timer routine to handle ND6 entries 379 */ 380 void 381 nd6_llinfo_settimer(struct llinfo_nd6 *ln, long xtick) 382 { 383 int s; 384 385 s = splsoftnet(); 386 387 if (xtick < 0) { 388 ln->ln_expire = 0; 389 ln->ln_ntick = 0; 390 callout_stop(&ln->ln_timer_ch); 391 } else { 392 ln->ln_expire = time_second + xtick / hz; 393 if (xtick > INT_MAX) { 394 ln->ln_ntick = xtick - INT_MAX; 395 callout_reset(&ln->ln_timer_ch, INT_MAX, 396 nd6_llinfo_timer, ln); 397 } else { 398 ln->ln_ntick = 0; 399 callout_reset(&ln->ln_timer_ch, xtick, 400 nd6_llinfo_timer, ln); 401 } 402 } 403 404 splx(s); 405 } 406 407 static void 408 nd6_llinfo_timer(void *arg) 409 { 410 struct llinfo_nd6 *ln; 411 struct rtentry *rt; 412 const struct sockaddr_in6 *dst; 413 struct ifnet *ifp; 414 struct nd_ifinfo *ndi = NULL; 415 416 mutex_enter(softnet_lock); 417 KERNEL_LOCK(1, NULL); 418 419 ln = (struct llinfo_nd6 *)arg; 420 421 if (ln->ln_ntick > 0) { 422 nd6_llinfo_settimer(ln, ln->ln_ntick); 423 KERNEL_UNLOCK_ONE(NULL); 424 mutex_exit(softnet_lock); 425 return; 426 } 427 428 if ((rt = ln->ln_rt) == NULL) 429 panic("ln->ln_rt == NULL"); 430 if ((ifp = rt->rt_ifp) == NULL) 431 panic("ln->ln_rt->rt_ifp == NULL"); 432 ndi = ND_IFINFO(ifp); 433 dst = satocsin6(rt_getkey(rt)); 434 435 /* sanity check */ 436 if (rt->rt_llinfo && (struct llinfo_nd6 *)rt->rt_llinfo != ln) 437 panic("rt_llinfo(%p) is not equal to ln(%p)", 438 rt->rt_llinfo, ln); 439 if (!dst) 440 panic("dst=0 in nd6_timer(ln=%p)", ln); 441 442 switch (ln->ln_state) { 443 case ND6_LLINFO_INCOMPLETE: 444 if (ln->ln_asked < nd6_mmaxtries) { 445 ln->ln_asked++; 446 nd6_llinfo_settimer(ln, (long)ndi->retrans * hz / 1000); 447 nd6_ns_output(ifp, NULL, &dst->sin6_addr, ln, 0); 448 } else { 449 struct mbuf *m = ln->ln_hold; 450 if (m) { 451 struct mbuf *m0; 452 453 /* 454 * assuming every packet in ln_hold has 455 * the same IP header 456 */ 457 m0 = m->m_nextpkt; 458 m->m_nextpkt = NULL; 459 icmp6_error2(m, ICMP6_DST_UNREACH, 460 ICMP6_DST_UNREACH_ADDR, 0, rt->rt_ifp); 461 462 ln->ln_hold = m0; 463 clear_llinfo_pqueue(ln); 464 } 465 (void)nd6_free(rt, 0); 466 ln = NULL; 467 } 468 break; 469 case ND6_LLINFO_REACHABLE: 470 if (!ND6_LLINFO_PERMANENT(ln)) { 471 ln->ln_state = ND6_LLINFO_STALE; 472 nd6_llinfo_settimer(ln, (long)nd6_gctimer * hz); 473 } 474 break; 475 476 case ND6_LLINFO_STALE: 477 /* Garbage Collection(RFC 2461 5.3) */ 478 if (!ND6_LLINFO_PERMANENT(ln)) { 479 (void)nd6_free(rt, 1); 480 ln = NULL; 481 } 482 break; 483 484 case ND6_LLINFO_DELAY: 485 if (ndi && (ndi->flags & ND6_IFF_PERFORMNUD) != 0) { 486 /* We need NUD */ 487 ln->ln_asked = 1; 488 ln->ln_state = ND6_LLINFO_PROBE; 489 nd6_llinfo_settimer(ln, (long)ndi->retrans * hz / 1000); 490 nd6_ns_output(ifp, &dst->sin6_addr, 491 &dst->sin6_addr, ln, 0); 492 } else { 493 ln->ln_state = ND6_LLINFO_STALE; /* XXX */ 494 nd6_llinfo_settimer(ln, (long)nd6_gctimer * hz); 495 } 496 break; 497 case ND6_LLINFO_PROBE: 498 if (ln->ln_asked < nd6_umaxtries) { 499 ln->ln_asked++; 500 nd6_llinfo_settimer(ln, (long)ndi->retrans * hz / 1000); 501 nd6_ns_output(ifp, &dst->sin6_addr, 502 &dst->sin6_addr, ln, 0); 503 } else { 504 (void)nd6_free(rt, 0); 505 ln = NULL; 506 } 507 break; 508 } 509 510 KERNEL_UNLOCK_ONE(NULL); 511 mutex_exit(softnet_lock); 512 } 513 514 /* 515 * ND6 timer routine to expire default route list and prefix list 516 */ 517 void 518 nd6_timer(void *ignored_arg) 519 { 520 struct nd_defrouter *next_dr, *dr; 521 struct nd_prefix *next_pr, *pr; 522 struct in6_ifaddr *ia6, *nia6; 523 524 callout_reset(&nd6_timer_ch, nd6_prune * hz, 525 nd6_timer, NULL); 526 527 mutex_enter(softnet_lock); 528 KERNEL_LOCK(1, NULL); 529 530 /* expire default router list */ 531 532 TAILQ_FOREACH_SAFE(dr, &nd_defrouter, dr_entry, next_dr) { 533 if (dr->expire && dr->expire < time_second) { 534 defrtrlist_del(dr); 535 } 536 } 537 538 /* 539 * expire interface addresses. 540 * in the past the loop was inside prefix expiry processing. 541 * However, from a stricter speci-confrmance standpoint, we should 542 * rather separate address lifetimes and prefix lifetimes. 543 */ 544 addrloop: 545 for (ia6 = in6_ifaddr; ia6; ia6 = nia6) { 546 nia6 = ia6->ia_next; 547 /* check address lifetime */ 548 if (IFA6_IS_INVALID(ia6)) { 549 int regen = 0; 550 551 /* 552 * If the expiring address is temporary, try 553 * regenerating a new one. This would be useful when 554 * we suspended a laptop PC, then turned it on after a 555 * period that could invalidate all temporary 556 * addresses. Although we may have to restart the 557 * loop (see below), it must be after purging the 558 * address. Otherwise, we'd see an infinite loop of 559 * regeneration. 560 */ 561 if (ip6_use_tempaddr && 562 (ia6->ia6_flags & IN6_IFF_TEMPORARY) != 0) { 563 if (regen_tmpaddr(ia6) == 0) 564 regen = 1; 565 } 566 567 in6_purgeaddr(&ia6->ia_ifa); 568 569 if (regen) 570 goto addrloop; /* XXX: see below */ 571 } else if (IFA6_IS_DEPRECATED(ia6)) { 572 int oldflags = ia6->ia6_flags; 573 574 ia6->ia6_flags |= IN6_IFF_DEPRECATED; 575 576 /* 577 * If a temporary address has just become deprecated, 578 * regenerate a new one if possible. 579 */ 580 if (ip6_use_tempaddr && 581 (ia6->ia6_flags & IN6_IFF_TEMPORARY) != 0 && 582 (oldflags & IN6_IFF_DEPRECATED) == 0) { 583 584 if (regen_tmpaddr(ia6) == 0) { 585 /* 586 * A new temporary address is 587 * generated. 588 * XXX: this means the address chain 589 * has changed while we are still in 590 * the loop. Although the change 591 * would not cause disaster (because 592 * it's not a deletion, but an 593 * addition,) we'd rather restart the 594 * loop just for safety. Or does this 595 * significantly reduce performance?? 596 */ 597 goto addrloop; 598 } 599 } 600 } else { 601 /* 602 * A new RA might have made a deprecated address 603 * preferred. 604 */ 605 ia6->ia6_flags &= ~IN6_IFF_DEPRECATED; 606 } 607 } 608 609 /* expire prefix list */ 610 LIST_FOREACH_SAFE(pr, &nd_prefix, ndpr_entry, next_pr) { 611 /* 612 * check prefix lifetime. 613 * since pltime is just for autoconf, pltime processing for 614 * prefix is not necessary. 615 */ 616 if (pr->ndpr_vltime != ND6_INFINITE_LIFETIME && 617 time_second - pr->ndpr_lastupdate > pr->ndpr_vltime) { 618 619 /* 620 * address expiration and prefix expiration are 621 * separate. NEVER perform in6_purgeaddr here. 622 */ 623 624 prelist_remove(pr); 625 } 626 } 627 628 KERNEL_UNLOCK_ONE(NULL); 629 mutex_exit(softnet_lock); 630 } 631 632 /* ia6: deprecated/invalidated temporary address */ 633 static int 634 regen_tmpaddr(struct in6_ifaddr *ia6) 635 { 636 struct ifaddr *ifa; 637 struct ifnet *ifp; 638 struct in6_ifaddr *public_ifa6 = NULL; 639 640 ifp = ia6->ia_ifa.ifa_ifp; 641 IFADDR_FOREACH(ifa, ifp) { 642 struct in6_ifaddr *it6; 643 644 if (ifa->ifa_addr->sa_family != AF_INET6) 645 continue; 646 647 it6 = (struct in6_ifaddr *)ifa; 648 649 /* ignore no autoconf addresses. */ 650 if ((it6->ia6_flags & IN6_IFF_AUTOCONF) == 0) 651 continue; 652 653 /* ignore autoconf addresses with different prefixes. */ 654 if (it6->ia6_ndpr == NULL || it6->ia6_ndpr != ia6->ia6_ndpr) 655 continue; 656 657 /* 658 * Now we are looking at an autoconf address with the same 659 * prefix as ours. If the address is temporary and is still 660 * preferred, do not create another one. It would be rare, but 661 * could happen, for example, when we resume a laptop PC after 662 * a long period. 663 */ 664 if ((it6->ia6_flags & IN6_IFF_TEMPORARY) != 0 && 665 !IFA6_IS_DEPRECATED(it6)) { 666 public_ifa6 = NULL; 667 break; 668 } 669 670 /* 671 * This is a public autoconf address that has the same prefix 672 * as ours. If it is preferred, keep it. We can't break the 673 * loop here, because there may be a still-preferred temporary 674 * address with the prefix. 675 */ 676 if (!IFA6_IS_DEPRECATED(it6)) 677 public_ifa6 = it6; 678 } 679 680 if (public_ifa6 != NULL) { 681 int e; 682 683 /* 684 * Random factor is introduced in the preferred lifetime, so 685 * we do not need additional delay (3rd arg to in6_tmpifadd). 686 */ 687 if ((e = in6_tmpifadd(public_ifa6, 0, 0)) != 0) { 688 log(LOG_NOTICE, "regen_tmpaddr: failed to create a new" 689 " tmp addr, errno=%d\n", e); 690 return -1; 691 } 692 return 0; 693 } 694 695 return -1; 696 } 697 698 bool 699 nd6_accepts_rtadv(const struct nd_ifinfo *ndi) 700 { 701 switch (ndi->flags & (ND6_IFF_ACCEPT_RTADV|ND6_IFF_OVERRIDE_RTADV)) { 702 case ND6_IFF_OVERRIDE_RTADV|ND6_IFF_ACCEPT_RTADV: 703 return true; 704 case ND6_IFF_ACCEPT_RTADV: 705 return ip6_accept_rtadv != 0; 706 case ND6_IFF_OVERRIDE_RTADV: 707 case 0: 708 default: 709 return false; 710 } 711 } 712 713 /* 714 * Nuke neighbor cache/prefix/default router management table, right before 715 * ifp goes away. 716 */ 717 void 718 nd6_purge(struct ifnet *ifp) 719 { 720 struct llinfo_nd6 *ln, *nln; 721 struct nd_defrouter *dr, *ndr; 722 struct nd_prefix *pr, *npr; 723 724 /* 725 * Nuke default router list entries toward ifp. 726 * We defer removal of default router list entries that is installed 727 * in the routing table, in order to keep additional side effects as 728 * small as possible. 729 */ 730 TAILQ_FOREACH_SAFE(dr, &nd_defrouter, dr_entry, ndr) { 731 if (dr->installed) 732 continue; 733 734 if (dr->ifp == ifp) 735 defrtrlist_del(dr); 736 } 737 738 TAILQ_FOREACH_SAFE(dr, &nd_defrouter, dr_entry, ndr) { 739 if (!dr->installed) 740 continue; 741 742 if (dr->ifp == ifp) 743 defrtrlist_del(dr); 744 } 745 746 /* Nuke prefix list entries toward ifp */ 747 LIST_FOREACH_SAFE(pr, &nd_prefix, ndpr_entry, npr) { 748 if (pr->ndpr_ifp == ifp) { 749 /* 750 * Because if_detach() does *not* release prefixes 751 * while purging addresses the reference count will 752 * still be above zero. We therefore reset it to 753 * make sure that the prefix really gets purged. 754 */ 755 pr->ndpr_refcnt = 0; 756 /* 757 * Previously, pr->ndpr_addr is removed as well, 758 * but I strongly believe we don't have to do it. 759 * nd6_purge() is only called from in6_ifdetach(), 760 * which removes all the associated interface addresses 761 * by itself. 762 * (jinmei@kame.net 20010129) 763 */ 764 prelist_remove(pr); 765 } 766 } 767 768 /* cancel default outgoing interface setting */ 769 if (nd6_defifindex == ifp->if_index) 770 nd6_setdefaultiface(0); 771 772 /* XXX: too restrictive? */ 773 if (!ip6_forwarding && ifp->if_afdata[AF_INET6]) { 774 struct nd_ifinfo *ndi = ND_IFINFO(ifp); 775 if (ndi && nd6_accepts_rtadv(ndi)) { 776 /* refresh default router list */ 777 defrouter_select(); 778 } 779 } 780 781 /* 782 * Nuke neighbor cache entries for the ifp. 783 * Note that rt->rt_ifp may not be the same as ifp, 784 * due to KAME goto ours hack. See RTM_RESOLVE case in 785 * nd6_rtrequest(), and ip6_input(). 786 */ 787 ln = llinfo_nd6.ln_next; 788 while (ln != NULL && ln != &llinfo_nd6) { 789 struct rtentry *rt; 790 const struct sockaddr_dl *sdl; 791 792 nln = ln->ln_next; 793 rt = ln->ln_rt; 794 if (rt && rt->rt_gateway && 795 rt->rt_gateway->sa_family == AF_LINK) { 796 sdl = satocsdl(rt->rt_gateway); 797 if (sdl->sdl_index == ifp->if_index) 798 nln = nd6_free(rt, 0); 799 } 800 ln = nln; 801 } 802 } 803 804 struct rtentry * 805 nd6_lookup(const struct in6_addr *addr6, int create, struct ifnet *ifp) 806 { 807 struct rtentry *rt; 808 struct sockaddr_in6 sin6; 809 810 sockaddr_in6_init(&sin6, addr6, 0, 0, 0); 811 rt = rtalloc1((struct sockaddr *)&sin6, create); 812 if (rt != NULL && (rt->rt_flags & RTF_LLINFO) == 0) { 813 /* 814 * This is the case for the default route. 815 * If we want to create a neighbor cache for the address, we 816 * should free the route for the destination and allocate an 817 * interface route. 818 */ 819 if (create) { 820 RTFREE(rt); 821 rt = NULL; 822 } 823 } 824 if (rt != NULL) 825 ; 826 else if (create && ifp) { 827 int e; 828 829 /* 830 * If no route is available and create is set, 831 * we allocate a host route for the destination 832 * and treat it like an interface route. 833 * This hack is necessary for a neighbor which can't 834 * be covered by our own prefix. 835 */ 836 struct ifaddr *ifa = 837 ifaof_ifpforaddr((struct sockaddr *)&sin6, ifp); 838 if (ifa == NULL) 839 return NULL; 840 841 /* 842 * Create a new route. RTF_LLINFO is necessary 843 * to create a Neighbor Cache entry for the 844 * destination in nd6_rtrequest which will be 845 * called in rtrequest via ifa->ifa_rtrequest. 846 */ 847 if ((e = rtrequest(RTM_ADD, (const struct sockaddr *)&sin6, 848 ifa->ifa_addr, (const struct sockaddr *)&all1_sa, 849 (ifa->ifa_flags | RTF_HOST | RTF_LLINFO) & 850 ~RTF_CLONING, &rt)) != 0) { 851 #if 0 852 log(LOG_ERR, 853 "nd6_lookup: failed to add route for a " 854 "neighbor(%s), errno=%d\n", 855 ip6_sprintf(addr6), e); 856 #endif 857 return NULL; 858 } 859 if (rt == NULL) 860 return NULL; 861 if (rt->rt_llinfo) { 862 struct llinfo_nd6 *ln = 863 (struct llinfo_nd6 *)rt->rt_llinfo; 864 ln->ln_state = ND6_LLINFO_NOSTATE; 865 } 866 } else 867 return NULL; 868 rt->rt_refcnt--; 869 /* 870 * Validation for the entry. 871 * Note that the check for rt_llinfo is necessary because a cloned 872 * route from a parent route that has the L flag (e.g. the default 873 * route to a p2p interface) may have the flag, too, while the 874 * destination is not actually a neighbor. 875 * XXX: we can't use rt->rt_ifp to check for the interface, since 876 * it might be the loopback interface if the entry is for our 877 * own address on a non-loopback interface. Instead, we should 878 * use rt->rt_ifa->ifa_ifp, which would specify the REAL 879 * interface. 880 * Note also that ifa_ifp and ifp may differ when we connect two 881 * interfaces to a same link, install a link prefix to an interface, 882 * and try to install a neighbor cache on an interface that does not 883 * have a route to the prefix. 884 */ 885 if ((rt->rt_flags & RTF_GATEWAY) || (rt->rt_flags & RTF_LLINFO) == 0 || 886 rt->rt_gateway->sa_family != AF_LINK || rt->rt_llinfo == NULL || 887 (ifp && rt->rt_ifa->ifa_ifp != ifp)) { 888 if (create) { 889 nd6log((LOG_DEBUG, 890 "nd6_lookup: failed to lookup %s (if = %s)\n", 891 ip6_sprintf(addr6), 892 ifp ? if_name(ifp) : "unspec")); 893 } 894 return NULL; 895 } 896 return rt; 897 } 898 899 /* 900 * Detect if a given IPv6 address identifies a neighbor on a given link. 901 * XXX: should take care of the destination of a p2p link? 902 */ 903 int 904 nd6_is_addr_neighbor(const struct sockaddr_in6 *addr, struct ifnet *ifp) 905 { 906 struct nd_prefix *pr; 907 908 /* 909 * A link-local address is always a neighbor. 910 * XXX: a link does not necessarily specify a single interface. 911 */ 912 if (IN6_IS_ADDR_LINKLOCAL(&addr->sin6_addr)) { 913 struct sockaddr_in6 sin6_copy; 914 u_int32_t zone; 915 916 /* 917 * We need sin6_copy since sa6_recoverscope() may modify the 918 * content (XXX). 919 */ 920 sin6_copy = *addr; 921 if (sa6_recoverscope(&sin6_copy)) 922 return 0; /* XXX: should be impossible */ 923 if (in6_setscope(&sin6_copy.sin6_addr, ifp, &zone)) 924 return 0; 925 if (sin6_copy.sin6_scope_id == zone) 926 return 1; 927 else 928 return 0; 929 } 930 931 /* 932 * If the address matches one of our on-link prefixes, it should be a 933 * neighbor. 934 */ 935 LIST_FOREACH(pr, &nd_prefix, ndpr_entry) { 936 if (pr->ndpr_ifp != ifp) 937 continue; 938 939 if (!(pr->ndpr_stateflags & NDPRF_ONLINK)) 940 continue; 941 942 if (IN6_ARE_MASKED_ADDR_EQUAL(&pr->ndpr_prefix.sin6_addr, 943 &addr->sin6_addr, &pr->ndpr_mask)) 944 return 1; 945 } 946 947 /* 948 * If the default router list is empty, all addresses are regarded 949 * as on-link, and thus, as a neighbor. 950 * XXX: we restrict the condition to hosts, because routers usually do 951 * not have the "default router list". 952 */ 953 if (!ip6_forwarding && TAILQ_FIRST(&nd_defrouter) == NULL && 954 nd6_defifindex == ifp->if_index) { 955 return 1; 956 } 957 958 /* 959 * Even if the address matches none of our addresses, it might be 960 * in the neighbor cache. 961 */ 962 if (nd6_lookup(&addr->sin6_addr, 0, ifp) != NULL) 963 return 1; 964 965 return 0; 966 } 967 968 /* 969 * Free an nd6 llinfo entry. 970 * Since the function would cause significant changes in the kernel, DO NOT 971 * make it global, unless you have a strong reason for the change, and are sure 972 * that the change is safe. 973 */ 974 static struct llinfo_nd6 * 975 nd6_free(struct rtentry *rt, int gc) 976 { 977 struct llinfo_nd6 *ln = (struct llinfo_nd6 *)rt->rt_llinfo, *next; 978 struct in6_addr in6 = satocsin6(rt_getkey(rt))->sin6_addr; 979 struct nd_defrouter *dr; 980 981 /* 982 * we used to have pfctlinput(PRC_HOSTDEAD) here. 983 * even though it is not harmful, it was not really necessary. 984 */ 985 986 /* cancel timer */ 987 nd6_llinfo_settimer(ln, -1); 988 989 if (!ip6_forwarding) { 990 int s; 991 s = splsoftnet(); 992 dr = defrouter_lookup(&satocsin6(rt_getkey(rt))->sin6_addr, 993 rt->rt_ifp); 994 995 if (dr != NULL && dr->expire && 996 ln->ln_state == ND6_LLINFO_STALE && gc) { 997 /* 998 * If the reason for the deletion is just garbage 999 * collection, and the neighbor is an active default 1000 * router, do not delete it. Instead, reset the GC 1001 * timer using the router's lifetime. 1002 * Simply deleting the entry would affect default 1003 * router selection, which is not necessarily a good 1004 * thing, especially when we're using router preference 1005 * values. 1006 * XXX: the check for ln_state would be redundant, 1007 * but we intentionally keep it just in case. 1008 */ 1009 if (dr->expire > time_second) 1010 nd6_llinfo_settimer(ln, 1011 (dr->expire - time_second) * hz); 1012 else 1013 nd6_llinfo_settimer(ln, (long)nd6_gctimer * hz); 1014 splx(s); 1015 return ln->ln_next; 1016 } 1017 1018 if (ln->ln_router || dr) { 1019 /* 1020 * rt6_flush must be called whether or not the neighbor 1021 * is in the Default Router List. 1022 * See a corresponding comment in nd6_na_input(). 1023 */ 1024 rt6_flush(&in6, rt->rt_ifp); 1025 } 1026 1027 if (dr) { 1028 /* 1029 * Unreachablity of a router might affect the default 1030 * router selection and on-link detection of advertised 1031 * prefixes. 1032 */ 1033 1034 /* 1035 * Temporarily fake the state to choose a new default 1036 * router and to perform on-link determination of 1037 * prefixes correctly. 1038 * Below the state will be set correctly, 1039 * or the entry itself will be deleted. 1040 */ 1041 ln->ln_state = ND6_LLINFO_INCOMPLETE; 1042 1043 /* 1044 * Since defrouter_select() does not affect the 1045 * on-link determination and MIP6 needs the check 1046 * before the default router selection, we perform 1047 * the check now. 1048 */ 1049 pfxlist_onlink_check(); 1050 1051 /* 1052 * refresh default router list 1053 */ 1054 defrouter_select(); 1055 } 1056 splx(s); 1057 } 1058 1059 /* 1060 * Before deleting the entry, remember the next entry as the 1061 * return value. We need this because pfxlist_onlink_check() above 1062 * might have freed other entries (particularly the old next entry) as 1063 * a side effect (XXX). 1064 */ 1065 next = ln->ln_next; 1066 1067 /* 1068 * Detach the route from the routing tree and the list of neighbor 1069 * caches, and disable the route entry not to be used in already 1070 * cached routes. 1071 */ 1072 rtrequest(RTM_DELETE, rt_getkey(rt), NULL, rt_mask(rt), 0, NULL); 1073 1074 return next; 1075 } 1076 1077 /* 1078 * Upper-layer reachability hint for Neighbor Unreachability Detection. 1079 * 1080 * XXX cost-effective methods? 1081 */ 1082 void 1083 nd6_nud_hint(struct rtentry *rt, struct in6_addr *dst6, int force) 1084 { 1085 struct llinfo_nd6 *ln; 1086 1087 /* 1088 * If the caller specified "rt", use that. Otherwise, resolve the 1089 * routing table by supplied "dst6". 1090 */ 1091 if (rt == NULL) { 1092 if (dst6 == NULL) 1093 return; 1094 if ((rt = nd6_lookup(dst6, 0, NULL)) == NULL) 1095 return; 1096 } 1097 1098 if ((rt->rt_flags & RTF_GATEWAY) != 0 || 1099 (rt->rt_flags & RTF_LLINFO) == 0 || 1100 !rt->rt_llinfo || !rt->rt_gateway || 1101 rt->rt_gateway->sa_family != AF_LINK) { 1102 /* This is not a host route. */ 1103 return; 1104 } 1105 1106 ln = (struct llinfo_nd6 *)rt->rt_llinfo; 1107 if (ln->ln_state < ND6_LLINFO_REACHABLE) 1108 return; 1109 1110 /* 1111 * if we get upper-layer reachability confirmation many times, 1112 * it is possible we have false information. 1113 */ 1114 if (!force) { 1115 ln->ln_byhint++; 1116 if (ln->ln_byhint > nd6_maxnudhint) 1117 return; 1118 } 1119 1120 ln->ln_state = ND6_LLINFO_REACHABLE; 1121 if (!ND6_LLINFO_PERMANENT(ln)) { 1122 nd6_llinfo_settimer(ln, 1123 (long)ND_IFINFO(rt->rt_ifp)->reachable * hz); 1124 } 1125 } 1126 1127 void 1128 nd6_rtrequest(int req, struct rtentry *rt, const struct rt_addrinfo *info) 1129 { 1130 struct sockaddr *gate = rt->rt_gateway; 1131 struct llinfo_nd6 *ln = (struct llinfo_nd6 *)rt->rt_llinfo; 1132 struct ifnet *ifp = rt->rt_ifp; 1133 uint8_t namelen = strlen(ifp->if_xname), addrlen = ifp->if_addrlen; 1134 struct ifaddr *ifa; 1135 1136 RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key); 1137 1138 if (req == RTM_LLINFO_UPD) { 1139 int rc; 1140 struct in6_addr *in6; 1141 struct in6_addr in6_all; 1142 int anycast; 1143 1144 if ((ifa = info->rti_ifa) == NULL) 1145 return; 1146 1147 in6 = &ifatoia6(ifa)->ia_addr.sin6_addr; 1148 anycast = ifatoia6(ifa)->ia6_flags & IN6_IFF_ANYCAST; 1149 1150 in6_all = in6addr_linklocal_allnodes; 1151 if ((rc = in6_setscope(&in6_all, ifa->ifa_ifp, NULL)) != 0) { 1152 log(LOG_ERR, "%s: failed to set scope %s " 1153 "(errno=%d)\n", __func__, if_name(ifp), rc); 1154 return; 1155 } 1156 1157 /* XXX don't set Override for proxy addresses */ 1158 nd6_na_output(ifa->ifa_ifp, &in6_all, in6, 1159 (anycast ? 0 : ND_NA_FLAG_OVERRIDE) 1160 #if 0 1161 | (ip6_forwarding ? ND_NA_FLAG_ROUTER : 0) 1162 #endif 1163 , 1, NULL); 1164 return; 1165 } 1166 1167 if ((rt->rt_flags & RTF_GATEWAY) != 0) 1168 return; 1169 1170 if (nd6_need_cache(ifp) == 0 && (rt->rt_flags & RTF_HOST) == 0) { 1171 RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key); 1172 /* 1173 * This is probably an interface direct route for a link 1174 * which does not need neighbor caches (e.g. fe80::%lo0/64). 1175 * We do not need special treatment below for such a route. 1176 * Moreover, the RTF_LLINFO flag which would be set below 1177 * would annoy the ndp(8) command. 1178 */ 1179 return; 1180 } 1181 1182 if (req == RTM_RESOLVE && 1183 (nd6_need_cache(ifp) == 0 || /* stf case */ 1184 !nd6_is_addr_neighbor(satocsin6(rt_getkey(rt)), ifp))) { 1185 RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key); 1186 /* 1187 * FreeBSD and BSD/OS often make a cloned host route based 1188 * on a less-specific route (e.g. the default route). 1189 * If the less specific route does not have a "gateway" 1190 * (this is the case when the route just goes to a p2p or an 1191 * stf interface), we'll mistakenly make a neighbor cache for 1192 * the host route, and will see strange neighbor solicitation 1193 * for the corresponding destination. In order to avoid the 1194 * confusion, we check if the destination of the route is 1195 * a neighbor in terms of neighbor discovery, and stop the 1196 * process if not. Additionally, we remove the LLINFO flag 1197 * so that ndp(8) will not try to get the neighbor information 1198 * of the destination. 1199 */ 1200 rt->rt_flags &= ~RTF_LLINFO; 1201 return; 1202 } 1203 1204 switch (req) { 1205 case RTM_ADD: 1206 RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key); 1207 /* 1208 * There is no backward compatibility :) 1209 * 1210 * if ((rt->rt_flags & RTF_HOST) == 0 && 1211 * SIN(rt_mask(rt))->sin_addr.s_addr != 0xffffffff) 1212 * rt->rt_flags |= RTF_CLONING; 1213 */ 1214 if ((rt->rt_flags & RTF_CLONING) || 1215 ((rt->rt_flags & RTF_LLINFO) && ln == NULL)) { 1216 union { 1217 struct sockaddr sa; 1218 struct sockaddr_dl sdl; 1219 struct sockaddr_storage ss; 1220 } u; 1221 /* 1222 * Case 1: This route should come from a route to 1223 * interface (RTF_CLONING case) or the route should be 1224 * treated as on-link but is currently not 1225 * (RTF_LLINFO && ln == NULL case). 1226 */ 1227 if (sockaddr_dl_init(&u.sdl, sizeof(u.ss), 1228 ifp->if_index, ifp->if_type, 1229 NULL, namelen, NULL, addrlen) == NULL) { 1230 printf("%s.%d: sockaddr_dl_init(, %zu, ) " 1231 "failed on %s\n", __func__, __LINE__, 1232 sizeof(u.ss), if_name(ifp)); 1233 } 1234 rt_setgate(rt, &u.sa); 1235 gate = rt->rt_gateway; 1236 RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key); 1237 if (ln != NULL) 1238 nd6_llinfo_settimer(ln, 0); 1239 RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key); 1240 if ((rt->rt_flags & RTF_CLONING) != 0) 1241 break; 1242 } 1243 RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key); 1244 /* 1245 * In IPv4 code, we try to annonuce new RTF_ANNOUNCE entry here. 1246 * We don't do that here since llinfo is not ready yet. 1247 * 1248 * There are also couple of other things to be discussed: 1249 * - unsolicited NA code needs improvement beforehand 1250 * - RFC2461 says we MAY send multicast unsolicited NA 1251 * (7.2.6 paragraph 4), however, it also says that we 1252 * SHOULD provide a mechanism to prevent multicast NA storm. 1253 * we don't have anything like it right now. 1254 * note that the mechanism needs a mutual agreement 1255 * between proxies, which means that we need to implement 1256 * a new protocol, or a new kludge. 1257 * - from RFC2461 6.2.4, host MUST NOT send an unsolicited NA. 1258 * we need to check ip6forwarding before sending it. 1259 * (or should we allow proxy ND configuration only for 1260 * routers? there's no mention about proxy ND from hosts) 1261 */ 1262 #if 0 1263 /* XXX it does not work */ 1264 if (rt->rt_flags & RTF_ANNOUNCE) 1265 nd6_na_output(ifp, 1266 &satocsin6(rt_getkey(rt))->sin6_addr, 1267 &satocsin6(rt_getkey(rt))->sin6_addr, 1268 ip6_forwarding ? ND_NA_FLAG_ROUTER : 0, 1269 1, NULL); 1270 #endif 1271 /* FALLTHROUGH */ 1272 case RTM_RESOLVE: 1273 if ((ifp->if_flags & (IFF_POINTOPOINT | IFF_LOOPBACK)) == 0) { 1274 RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key); 1275 /* 1276 * Address resolution isn't necessary for a point to 1277 * point link, so we can skip this test for a p2p link. 1278 */ 1279 if (gate->sa_family != AF_LINK || 1280 gate->sa_len < 1281 sockaddr_dl_measure(namelen, addrlen)) { 1282 log(LOG_DEBUG, 1283 "nd6_rtrequest: bad gateway value: %s\n", 1284 if_name(ifp)); 1285 break; 1286 } 1287 satosdl(gate)->sdl_type = ifp->if_type; 1288 satosdl(gate)->sdl_index = ifp->if_index; 1289 RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key); 1290 } 1291 if (ln != NULL) 1292 break; /* This happens on a route change */ 1293 RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key); 1294 /* 1295 * Case 2: This route may come from cloning, or a manual route 1296 * add with a LL address. 1297 */ 1298 R_Malloc(ln, struct llinfo_nd6 *, sizeof(*ln)); 1299 rt->rt_llinfo = ln; 1300 RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key); 1301 if (ln == NULL) { 1302 log(LOG_DEBUG, "nd6_rtrequest: malloc failed\n"); 1303 break; 1304 } 1305 RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key); 1306 nd6_inuse++; 1307 nd6_allocated++; 1308 memset(ln, 0, sizeof(*ln)); 1309 ln->ln_rt = rt; 1310 callout_init(&ln->ln_timer_ch, CALLOUT_MPSAFE); 1311 /* this is required for "ndp" command. - shin */ 1312 if (req == RTM_ADD) { 1313 /* 1314 * gate should have some valid AF_LINK entry, 1315 * and ln->ln_expire should have some lifetime 1316 * which is specified by ndp command. 1317 */ 1318 ln->ln_state = ND6_LLINFO_REACHABLE; 1319 ln->ln_byhint = 0; 1320 } else { 1321 /* 1322 * When req == RTM_RESOLVE, rt is created and 1323 * initialized in rtrequest(), so rt_expire is 0. 1324 */ 1325 ln->ln_state = ND6_LLINFO_NOSTATE; 1326 nd6_llinfo_settimer(ln, 0); 1327 } 1328 RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key); 1329 rt->rt_flags |= RTF_LLINFO; 1330 ln->ln_next = llinfo_nd6.ln_next; 1331 llinfo_nd6.ln_next = ln; 1332 ln->ln_prev = &llinfo_nd6; 1333 ln->ln_next->ln_prev = ln; 1334 1335 RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key); 1336 /* 1337 * check if rt_getkey(rt) is an address assigned 1338 * to the interface. 1339 */ 1340 ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp, 1341 &satocsin6(rt_getkey(rt))->sin6_addr); 1342 RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key); 1343 if (ifa != NULL) { 1344 const void *mac; 1345 nd6_llinfo_settimer(ln, -1); 1346 ln->ln_state = ND6_LLINFO_REACHABLE; 1347 ln->ln_byhint = 0; 1348 if ((mac = nd6_ifptomac(ifp)) != NULL) { 1349 /* XXX check for error */ 1350 if (sockaddr_dl_setaddr(satosdl(gate), 1351 gate->sa_len, mac, 1352 ifp->if_addrlen) == NULL) { 1353 printf("%s.%d: " 1354 "sockaddr_dl_setaddr(, %d, ) " 1355 "failed on %s\n", __func__, 1356 __LINE__, gate->sa_len, 1357 if_name(ifp)); 1358 } 1359 } 1360 if (nd6_useloopback) { 1361 ifp = rt->rt_ifp = lo0ifp; /* XXX */ 1362 /* 1363 * Make sure rt_ifa be equal to the ifaddr 1364 * corresponding to the address. 1365 * We need this because when we refer 1366 * rt_ifa->ia6_flags in ip6_input, we assume 1367 * that the rt_ifa points to the address instead 1368 * of the loopback address. 1369 */ 1370 if (ifa != rt->rt_ifa) 1371 rt_replace_ifa(rt, ifa); 1372 rt->rt_flags &= ~RTF_CLONED; 1373 } 1374 } else if (rt->rt_flags & RTF_ANNOUNCE) { 1375 nd6_llinfo_settimer(ln, -1); 1376 ln->ln_state = ND6_LLINFO_REACHABLE; 1377 ln->ln_byhint = 0; 1378 1379 /* join solicited node multicast for proxy ND */ 1380 if (ifp->if_flags & IFF_MULTICAST) { 1381 struct in6_addr llsol; 1382 int error; 1383 1384 llsol = satocsin6(rt_getkey(rt))->sin6_addr; 1385 llsol.s6_addr32[0] = htonl(0xff020000); 1386 llsol.s6_addr32[1] = 0; 1387 llsol.s6_addr32[2] = htonl(1); 1388 llsol.s6_addr8[12] = 0xff; 1389 if (in6_setscope(&llsol, ifp, NULL)) 1390 break; 1391 if (!in6_addmulti(&llsol, ifp, &error, 0)) { 1392 nd6log((LOG_ERR, "%s: failed to join " 1393 "%s (errno=%d)\n", if_name(ifp), 1394 ip6_sprintf(&llsol), error)); 1395 } 1396 } 1397 } 1398 break; 1399 1400 case RTM_DELETE: 1401 if (ln == NULL) 1402 break; 1403 /* leave from solicited node multicast for proxy ND */ 1404 if ((rt->rt_flags & RTF_ANNOUNCE) != 0 && 1405 (ifp->if_flags & IFF_MULTICAST) != 0) { 1406 struct in6_addr llsol; 1407 struct in6_multi *in6m; 1408 1409 llsol = satocsin6(rt_getkey(rt))->sin6_addr; 1410 llsol.s6_addr32[0] = htonl(0xff020000); 1411 llsol.s6_addr32[1] = 0; 1412 llsol.s6_addr32[2] = htonl(1); 1413 llsol.s6_addr8[12] = 0xff; 1414 if (in6_setscope(&llsol, ifp, NULL) == 0) { 1415 IN6_LOOKUP_MULTI(llsol, ifp, in6m); 1416 if (in6m) 1417 in6_delmulti(in6m); 1418 } 1419 } 1420 nd6_inuse--; 1421 ln->ln_next->ln_prev = ln->ln_prev; 1422 ln->ln_prev->ln_next = ln->ln_next; 1423 ln->ln_prev = NULL; 1424 nd6_llinfo_settimer(ln, -1); 1425 rt->rt_llinfo = 0; 1426 rt->rt_flags &= ~RTF_LLINFO; 1427 clear_llinfo_pqueue(ln); 1428 Free(ln); 1429 } 1430 } 1431 1432 int 1433 nd6_ioctl(u_long cmd, void *data, struct ifnet *ifp) 1434 { 1435 struct in6_drlist *drl = (struct in6_drlist *)data; 1436 struct in6_oprlist *oprl = (struct in6_oprlist *)data; 1437 struct in6_ndireq *ndi = (struct in6_ndireq *)data; 1438 struct in6_nbrinfo *nbi = (struct in6_nbrinfo *)data; 1439 struct in6_ndifreq *ndif = (struct in6_ndifreq *)data; 1440 struct nd_defrouter *dr; 1441 struct nd_prefix *pr; 1442 struct rtentry *rt; 1443 int i = 0, error = 0; 1444 int s; 1445 1446 switch (cmd) { 1447 case SIOCGDRLST_IN6: 1448 /* 1449 * obsolete API, use sysctl under net.inet6.icmp6 1450 */ 1451 memset(drl, 0, sizeof(*drl)); 1452 s = splsoftnet(); 1453 TAILQ_FOREACH(dr, &nd_defrouter, dr_entry) { 1454 if (i >= DRLSTSIZ) 1455 break; 1456 drl->defrouter[i].rtaddr = dr->rtaddr; 1457 in6_clearscope(&drl->defrouter[i].rtaddr); 1458 1459 drl->defrouter[i].flags = dr->flags; 1460 drl->defrouter[i].rtlifetime = dr->rtlifetime; 1461 drl->defrouter[i].expire = dr->expire; 1462 drl->defrouter[i].if_index = dr->ifp->if_index; 1463 i++; 1464 } 1465 splx(s); 1466 break; 1467 case SIOCGPRLST_IN6: 1468 /* 1469 * obsolete API, use sysctl under net.inet6.icmp6 1470 * 1471 * XXX the structure in6_prlist was changed in backward- 1472 * incompatible manner. in6_oprlist is used for SIOCGPRLST_IN6, 1473 * in6_prlist is used for nd6_sysctl() - fill_prlist(). 1474 */ 1475 /* 1476 * XXX meaning of fields, especialy "raflags", is very 1477 * differnet between RA prefix list and RR/static prefix list. 1478 * how about separating ioctls into two? 1479 */ 1480 memset(oprl, 0, sizeof(*oprl)); 1481 s = splsoftnet(); 1482 LIST_FOREACH(pr, &nd_prefix, ndpr_entry) { 1483 struct nd_pfxrouter *pfr; 1484 int j; 1485 1486 if (i >= PRLSTSIZ) 1487 break; 1488 oprl->prefix[i].prefix = pr->ndpr_prefix.sin6_addr; 1489 oprl->prefix[i].raflags = pr->ndpr_raf; 1490 oprl->prefix[i].prefixlen = pr->ndpr_plen; 1491 oprl->prefix[i].vltime = pr->ndpr_vltime; 1492 oprl->prefix[i].pltime = pr->ndpr_pltime; 1493 oprl->prefix[i].if_index = pr->ndpr_ifp->if_index; 1494 if (pr->ndpr_vltime == ND6_INFINITE_LIFETIME) 1495 oprl->prefix[i].expire = 0; 1496 else { 1497 time_t maxexpire; 1498 1499 /* XXX: we assume time_t is signed. */ 1500 maxexpire = (-1) & 1501 ~((time_t)1 << 1502 ((sizeof(maxexpire) * 8) - 1)); 1503 if (pr->ndpr_vltime < 1504 maxexpire - pr->ndpr_lastupdate) { 1505 oprl->prefix[i].expire = 1506 pr->ndpr_lastupdate + 1507 pr->ndpr_vltime; 1508 } else 1509 oprl->prefix[i].expire = maxexpire; 1510 } 1511 1512 j = 0; 1513 LIST_FOREACH(pfr, &pr->ndpr_advrtrs, pfr_entry) { 1514 if (j < DRLSTSIZ) { 1515 #define RTRADDR oprl->prefix[i].advrtr[j] 1516 RTRADDR = pfr->router->rtaddr; 1517 in6_clearscope(&RTRADDR); 1518 #undef RTRADDR 1519 } 1520 j++; 1521 } 1522 oprl->prefix[i].advrtrs = j; 1523 oprl->prefix[i].origin = PR_ORIG_RA; 1524 1525 i++; 1526 } 1527 splx(s); 1528 1529 break; 1530 case OSIOCGIFINFO_IN6: 1531 #define ND ndi->ndi 1532 /* XXX: old ndp(8) assumes a positive value for linkmtu. */ 1533 memset(&ND, 0, sizeof(ND)); 1534 ND.linkmtu = IN6_LINKMTU(ifp); 1535 ND.maxmtu = ND_IFINFO(ifp)->maxmtu; 1536 ND.basereachable = ND_IFINFO(ifp)->basereachable; 1537 ND.reachable = ND_IFINFO(ifp)->reachable; 1538 ND.retrans = ND_IFINFO(ifp)->retrans; 1539 ND.flags = ND_IFINFO(ifp)->flags; 1540 ND.recalctm = ND_IFINFO(ifp)->recalctm; 1541 ND.chlim = ND_IFINFO(ifp)->chlim; 1542 break; 1543 case SIOCGIFINFO_IN6: 1544 ND = *ND_IFINFO(ifp); 1545 break; 1546 case SIOCSIFINFO_IN6: 1547 /* 1548 * used to change host variables from userland. 1549 * intented for a use on router to reflect RA configurations. 1550 */ 1551 /* 0 means 'unspecified' */ 1552 if (ND.linkmtu != 0) { 1553 if (ND.linkmtu < IPV6_MMTU || 1554 ND.linkmtu > IN6_LINKMTU(ifp)) { 1555 error = EINVAL; 1556 break; 1557 } 1558 ND_IFINFO(ifp)->linkmtu = ND.linkmtu; 1559 } 1560 1561 if (ND.basereachable != 0) { 1562 int obasereachable = ND_IFINFO(ifp)->basereachable; 1563 1564 ND_IFINFO(ifp)->basereachable = ND.basereachable; 1565 if (ND.basereachable != obasereachable) 1566 ND_IFINFO(ifp)->reachable = 1567 ND_COMPUTE_RTIME(ND.basereachable); 1568 } 1569 if (ND.retrans != 0) 1570 ND_IFINFO(ifp)->retrans = ND.retrans; 1571 if (ND.chlim != 0) 1572 ND_IFINFO(ifp)->chlim = ND.chlim; 1573 /* FALLTHROUGH */ 1574 case SIOCSIFINFO_FLAGS: 1575 ND_IFINFO(ifp)->flags = ND.flags; 1576 break; 1577 #undef ND 1578 case SIOCSNDFLUSH_IN6: /* XXX: the ioctl name is confusing... */ 1579 /* sync kernel routing table with the default router list */ 1580 defrouter_reset(); 1581 defrouter_select(); 1582 break; 1583 case SIOCSPFXFLUSH_IN6: 1584 { 1585 /* flush all the prefix advertised by routers */ 1586 struct nd_prefix *pfx, *next; 1587 1588 s = splsoftnet(); 1589 LIST_FOREACH_SAFE(pfx, &nd_prefix, ndpr_entry, next) { 1590 struct in6_ifaddr *ia, *ia_next; 1591 1592 if (IN6_IS_ADDR_LINKLOCAL(&pfx->ndpr_prefix.sin6_addr)) 1593 continue; /* XXX */ 1594 1595 /* do we really have to remove addresses as well? */ 1596 for (ia = in6_ifaddr; ia; ia = ia_next) { 1597 /* ia might be removed. keep the next ptr. */ 1598 ia_next = ia->ia_next; 1599 1600 if ((ia->ia6_flags & IN6_IFF_AUTOCONF) == 0) 1601 continue; 1602 1603 if (ia->ia6_ndpr == pfx) 1604 in6_purgeaddr(&ia->ia_ifa); 1605 } 1606 prelist_remove(pfx); 1607 } 1608 splx(s); 1609 break; 1610 } 1611 case SIOCSRTRFLUSH_IN6: 1612 { 1613 /* flush all the default routers */ 1614 struct nd_defrouter *drtr, *next; 1615 1616 s = splsoftnet(); 1617 defrouter_reset(); 1618 TAILQ_FOREACH_SAFE(drtr, &nd_defrouter, dr_entry, next) { 1619 defrtrlist_del(drtr); 1620 } 1621 defrouter_select(); 1622 splx(s); 1623 break; 1624 } 1625 case SIOCGNBRINFO_IN6: 1626 { 1627 struct llinfo_nd6 *ln; 1628 struct in6_addr nb_addr = nbi->addr; /* make local for safety */ 1629 1630 if ((error = in6_setscope(&nb_addr, ifp, NULL)) != 0) 1631 return error; 1632 1633 s = splsoftnet(); 1634 if ((rt = nd6_lookup(&nb_addr, 0, ifp)) == NULL || 1635 (ln = (struct llinfo_nd6 *)rt->rt_llinfo) == NULL) { 1636 error = EINVAL; 1637 splx(s); 1638 break; 1639 } 1640 nbi->state = ln->ln_state; 1641 nbi->asked = ln->ln_asked; 1642 nbi->isrouter = ln->ln_router; 1643 nbi->expire = ln->ln_expire; 1644 splx(s); 1645 1646 break; 1647 } 1648 case SIOCGDEFIFACE_IN6: /* XXX: should be implemented as a sysctl? */ 1649 ndif->ifindex = nd6_defifindex; 1650 break; 1651 case SIOCSDEFIFACE_IN6: /* XXX: should be implemented as a sysctl? */ 1652 return nd6_setdefaultiface(ndif->ifindex); 1653 } 1654 return error; 1655 } 1656 1657 void 1658 nd6_llinfo_release_pkts(struct llinfo_nd6 *ln, struct ifnet *ifp, 1659 struct rtentry *rt) 1660 { 1661 struct mbuf *m_hold, *m_hold_next; 1662 1663 for (m_hold = ln->ln_hold, ln->ln_hold = NULL; 1664 m_hold != NULL; 1665 m_hold = m_hold_next) { 1666 m_hold_next = m_hold->m_nextpkt; 1667 m_hold->m_nextpkt = NULL; 1668 1669 /* 1670 * we assume ifp is not a p2p here, so 1671 * just set the 2nd argument as the 1672 * 1st one. 1673 */ 1674 nd6_output(ifp, ifp, m_hold, satocsin6(rt_getkey(rt)), rt); 1675 } 1676 } 1677 1678 /* 1679 * Create neighbor cache entry and cache link-layer address, 1680 * on reception of inbound ND6 packets. (RS/RA/NS/redirect) 1681 */ 1682 struct rtentry * 1683 nd6_cache_lladdr( 1684 struct ifnet *ifp, 1685 struct in6_addr *from, 1686 char *lladdr, 1687 int lladdrlen, 1688 int type, /* ICMP6 type */ 1689 int code /* type dependent information */ 1690 ) 1691 { 1692 struct nd_ifinfo *ndi = ND_IFINFO(ifp); 1693 struct rtentry *rt = NULL; 1694 struct llinfo_nd6 *ln = NULL; 1695 int is_newentry; 1696 struct sockaddr_dl *sdl = NULL; 1697 int do_update; 1698 int olladdr; 1699 int llchange; 1700 int newstate = 0; 1701 1702 if (ifp == NULL) 1703 panic("ifp == NULL in nd6_cache_lladdr"); 1704 if (from == NULL) 1705 panic("from == NULL in nd6_cache_lladdr"); 1706 1707 /* nothing must be updated for unspecified address */ 1708 if (IN6_IS_ADDR_UNSPECIFIED(from)) 1709 return NULL; 1710 1711 /* 1712 * Validation about ifp->if_addrlen and lladdrlen must be done in 1713 * the caller. 1714 * 1715 * XXX If the link does not have link-layer adderss, what should 1716 * we do? (ifp->if_addrlen == 0) 1717 * Spec says nothing in sections for RA, RS and NA. There's small 1718 * description on it in NS section (RFC 2461 7.2.3). 1719 */ 1720 1721 rt = nd6_lookup(from, 0, ifp); 1722 if (rt == NULL) { 1723 #if 0 1724 /* nothing must be done if there's no lladdr */ 1725 if (!lladdr || !lladdrlen) 1726 return NULL; 1727 #endif 1728 1729 rt = nd6_lookup(from, 1, ifp); 1730 is_newentry = 1; 1731 } else { 1732 /* do nothing if static ndp is set */ 1733 if (rt->rt_flags & RTF_STATIC) 1734 return NULL; 1735 is_newentry = 0; 1736 } 1737 1738 if (rt == NULL) 1739 return NULL; 1740 if ((rt->rt_flags & (RTF_GATEWAY | RTF_LLINFO)) != RTF_LLINFO) { 1741 fail: 1742 (void)nd6_free(rt, 0); 1743 return NULL; 1744 } 1745 ln = (struct llinfo_nd6 *)rt->rt_llinfo; 1746 if (ln == NULL) 1747 goto fail; 1748 if (rt->rt_gateway == NULL) 1749 goto fail; 1750 if (rt->rt_gateway->sa_family != AF_LINK) 1751 goto fail; 1752 sdl = satosdl(rt->rt_gateway); 1753 1754 olladdr = (sdl->sdl_alen) ? 1 : 0; 1755 if (olladdr && lladdr) { 1756 if (memcmp(lladdr, CLLADDR(sdl), ifp->if_addrlen)) 1757 llchange = 1; 1758 else 1759 llchange = 0; 1760 } else 1761 llchange = 0; 1762 1763 /* 1764 * newentry olladdr lladdr llchange (*=record) 1765 * 0 n n -- (1) 1766 * 0 y n -- (2) 1767 * 0 n y -- (3) * STALE 1768 * 0 y y n (4) * 1769 * 0 y y y (5) * STALE 1770 * 1 -- n -- (6) NOSTATE(= PASSIVE) 1771 * 1 -- y -- (7) * STALE 1772 */ 1773 1774 if (lladdr) { /* (3-5) and (7) */ 1775 /* 1776 * Record source link-layer address 1777 * XXX is it dependent to ifp->if_type? 1778 */ 1779 /* XXX check for error */ 1780 if (sockaddr_dl_setaddr(sdl, sdl->sdl_len, lladdr, 1781 ifp->if_addrlen) == NULL) { 1782 printf("%s.%d: sockaddr_dl_setaddr(, %d, ) " 1783 "failed on %s\n", __func__, __LINE__, 1784 sdl->sdl_len, if_name(ifp)); 1785 } 1786 } 1787 1788 if (!is_newentry) { 1789 if ((!olladdr && lladdr) || /* (3) */ 1790 (olladdr && lladdr && llchange)) { /* (5) */ 1791 do_update = 1; 1792 newstate = ND6_LLINFO_STALE; 1793 } else /* (1-2,4) */ 1794 do_update = 0; 1795 } else { 1796 do_update = 1; 1797 if (lladdr == NULL) /* (6) */ 1798 newstate = ND6_LLINFO_NOSTATE; 1799 else /* (7) */ 1800 newstate = ND6_LLINFO_STALE; 1801 } 1802 1803 if (do_update) { 1804 /* 1805 * Update the state of the neighbor cache. 1806 */ 1807 ln->ln_state = newstate; 1808 1809 if (ln->ln_state == ND6_LLINFO_STALE) { 1810 /* 1811 * XXX: since nd6_output() below will cause 1812 * state tansition to DELAY and reset the timer, 1813 * we must set the timer now, although it is actually 1814 * meaningless. 1815 */ 1816 nd6_llinfo_settimer(ln, (long)nd6_gctimer * hz); 1817 1818 nd6_llinfo_release_pkts(ln, ifp, rt); 1819 } else if (ln->ln_state == ND6_LLINFO_INCOMPLETE) { 1820 /* probe right away */ 1821 nd6_llinfo_settimer((void *)ln, 0); 1822 } 1823 } 1824 1825 /* 1826 * ICMP6 type dependent behavior. 1827 * 1828 * NS: clear IsRouter if new entry 1829 * RS: clear IsRouter 1830 * RA: set IsRouter if there's lladdr 1831 * redir: clear IsRouter if new entry 1832 * 1833 * RA case, (1): 1834 * The spec says that we must set IsRouter in the following cases: 1835 * - If lladdr exist, set IsRouter. This means (1-5). 1836 * - If it is old entry (!newentry), set IsRouter. This means (7). 1837 * So, based on the spec, in (1-5) and (7) cases we must set IsRouter. 1838 * A quetion arises for (1) case. (1) case has no lladdr in the 1839 * neighbor cache, this is similar to (6). 1840 * This case is rare but we figured that we MUST NOT set IsRouter. 1841 * 1842 * newentry olladdr lladdr llchange NS RS RA redir 1843 * D R 1844 * 0 n n -- (1) c ? s 1845 * 0 y n -- (2) c s s 1846 * 0 n y -- (3) c s s 1847 * 0 y y n (4) c s s 1848 * 0 y y y (5) c s s 1849 * 1 -- n -- (6) c c c s 1850 * 1 -- y -- (7) c c s c s 1851 * 1852 * (c=clear s=set) 1853 */ 1854 switch (type & 0xff) { 1855 case ND_NEIGHBOR_SOLICIT: 1856 /* 1857 * New entry must have is_router flag cleared. 1858 */ 1859 if (is_newentry) /* (6-7) */ 1860 ln->ln_router = 0; 1861 break; 1862 case ND_REDIRECT: 1863 /* 1864 * If the icmp is a redirect to a better router, always set the 1865 * is_router flag. Otherwise, if the entry is newly created, 1866 * clear the flag. [RFC 2461, sec 8.3] 1867 */ 1868 if (code == ND_REDIRECT_ROUTER) 1869 ln->ln_router = 1; 1870 else if (is_newentry) /* (6-7) */ 1871 ln->ln_router = 0; 1872 break; 1873 case ND_ROUTER_SOLICIT: 1874 /* 1875 * is_router flag must always be cleared. 1876 */ 1877 ln->ln_router = 0; 1878 break; 1879 case ND_ROUTER_ADVERT: 1880 /* 1881 * Mark an entry with lladdr as a router. 1882 */ 1883 if ((!is_newentry && (olladdr || lladdr)) || /* (2-5) */ 1884 (is_newentry && lladdr)) { /* (7) */ 1885 ln->ln_router = 1; 1886 } 1887 break; 1888 } 1889 1890 /* 1891 * When the link-layer address of a router changes, select the 1892 * best router again. In particular, when the neighbor entry is newly 1893 * created, it might affect the selection policy. 1894 * Question: can we restrict the first condition to the "is_newentry" 1895 * case? 1896 * XXX: when we hear an RA from a new router with the link-layer 1897 * address option, defrouter_select() is called twice, since 1898 * defrtrlist_update called the function as well. However, I believe 1899 * we can compromise the overhead, since it only happens the first 1900 * time. 1901 * XXX: although defrouter_select() should not have a bad effect 1902 * for those are not autoconfigured hosts, we explicitly avoid such 1903 * cases for safety. 1904 */ 1905 if (do_update && ln->ln_router && !ip6_forwarding && 1906 nd6_accepts_rtadv(ndi)) 1907 defrouter_select(); 1908 1909 return rt; 1910 } 1911 1912 static void 1913 nd6_slowtimo(void *ignored_arg) 1914 { 1915 struct nd_ifinfo *nd6if; 1916 struct ifnet *ifp; 1917 1918 mutex_enter(softnet_lock); 1919 KERNEL_LOCK(1, NULL); 1920 callout_reset(&nd6_slowtimo_ch, ND6_SLOWTIMER_INTERVAL * hz, 1921 nd6_slowtimo, NULL); 1922 TAILQ_FOREACH(ifp, &ifnet, if_list) { 1923 nd6if = ND_IFINFO(ifp); 1924 if (nd6if->basereachable && /* already initialized */ 1925 (nd6if->recalctm -= ND6_SLOWTIMER_INTERVAL) <= 0) { 1926 /* 1927 * Since reachable time rarely changes by router 1928 * advertisements, we SHOULD insure that a new random 1929 * value gets recomputed at least once every few hours. 1930 * (RFC 2461, 6.3.4) 1931 */ 1932 nd6if->recalctm = nd6_recalc_reachtm_interval; 1933 nd6if->reachable = ND_COMPUTE_RTIME(nd6if->basereachable); 1934 } 1935 } 1936 KERNEL_UNLOCK_ONE(NULL); 1937 mutex_exit(softnet_lock); 1938 } 1939 1940 #define senderr(e) { error = (e); goto bad;} 1941 int 1942 nd6_output(struct ifnet *ifp, struct ifnet *origifp, struct mbuf *m0, 1943 const struct sockaddr_in6 *dst, struct rtentry *rt0) 1944 { 1945 struct mbuf *m = m0; 1946 struct rtentry *rt = rt0; 1947 struct sockaddr_in6 *gw6 = NULL; 1948 struct llinfo_nd6 *ln = NULL; 1949 int error = 0; 1950 1951 if (IN6_IS_ADDR_MULTICAST(&dst->sin6_addr)) 1952 goto sendpkt; 1953 1954 if (nd6_need_cache(ifp) == 0) 1955 goto sendpkt; 1956 1957 /* 1958 * next hop determination. This routine is derived from ether_output. 1959 */ 1960 if (rt) { 1961 if ((rt->rt_flags & RTF_UP) == 0) { 1962 if ((rt0 = rt = rtalloc1(sin6tocsa(dst), 1)) != NULL) { 1963 rt->rt_refcnt--; 1964 if (rt->rt_ifp != ifp) 1965 senderr(EHOSTUNREACH); 1966 } else 1967 senderr(EHOSTUNREACH); 1968 } 1969 1970 if (rt->rt_flags & RTF_GATEWAY) { 1971 gw6 = (struct sockaddr_in6 *)rt->rt_gateway; 1972 1973 /* 1974 * We skip link-layer address resolution and NUD 1975 * if the gateway is not a neighbor from ND point 1976 * of view, regardless of the value of nd_ifinfo.flags. 1977 * The second condition is a bit tricky; we skip 1978 * if the gateway is our own address, which is 1979 * sometimes used to install a route to a p2p link. 1980 */ 1981 if (!nd6_is_addr_neighbor(gw6, ifp) || 1982 in6ifa_ifpwithaddr(ifp, &gw6->sin6_addr)) { 1983 /* 1984 * We allow this kind of tricky route only 1985 * when the outgoing interface is p2p. 1986 * XXX: we may need a more generic rule here. 1987 */ 1988 if ((ifp->if_flags & IFF_POINTOPOINT) == 0) 1989 senderr(EHOSTUNREACH); 1990 1991 goto sendpkt; 1992 } 1993 1994 if (rt->rt_gwroute == NULL) 1995 goto lookup; 1996 if (((rt = rt->rt_gwroute)->rt_flags & RTF_UP) == 0) { 1997 rtfree(rt); rt = rt0; 1998 lookup: 1999 rt->rt_gwroute = rtalloc1(rt->rt_gateway, 1); 2000 if ((rt = rt->rt_gwroute) == NULL) 2001 senderr(EHOSTUNREACH); 2002 /* the "G" test below also prevents rt == rt0 */ 2003 if ((rt->rt_flags & RTF_GATEWAY) || 2004 (rt->rt_ifp != ifp)) { 2005 rt->rt_refcnt--; 2006 rt0->rt_gwroute = NULL; 2007 senderr(EHOSTUNREACH); 2008 } 2009 } 2010 } 2011 } 2012 2013 /* 2014 * Address resolution or Neighbor Unreachability Detection 2015 * for the next hop. 2016 * At this point, the destination of the packet must be a unicast 2017 * or an anycast address(i.e. not a multicast). 2018 */ 2019 2020 /* Look up the neighbor cache for the nexthop */ 2021 if (rt != NULL && (rt->rt_flags & RTF_LLINFO) != 0) 2022 ln = (struct llinfo_nd6 *)rt->rt_llinfo; 2023 else { 2024 /* 2025 * Since nd6_is_addr_neighbor() internally calls nd6_lookup(), 2026 * the condition below is not very efficient. But we believe 2027 * it is tolerable, because this should be a rare case. 2028 */ 2029 if (nd6_is_addr_neighbor(dst, ifp) && 2030 (rt = nd6_lookup(&dst->sin6_addr, 1, ifp)) != NULL) 2031 ln = (struct llinfo_nd6 *)rt->rt_llinfo; 2032 } 2033 if (ln == NULL || rt == NULL) { 2034 if ((ifp->if_flags & IFF_POINTOPOINT) == 0 && 2035 !(ND_IFINFO(ifp)->flags & ND6_IFF_PERFORMNUD)) { 2036 log(LOG_DEBUG, 2037 "nd6_output: can't allocate llinfo for %s " 2038 "(ln=%p, rt=%p)\n", 2039 ip6_sprintf(&dst->sin6_addr), ln, rt); 2040 senderr(EIO); /* XXX: good error? */ 2041 } 2042 2043 goto sendpkt; /* send anyway */ 2044 } 2045 2046 /* We don't have to do link-layer address resolution on a p2p link. */ 2047 if ((ifp->if_flags & IFF_POINTOPOINT) != 0 && 2048 ln->ln_state < ND6_LLINFO_REACHABLE) { 2049 ln->ln_state = ND6_LLINFO_STALE; 2050 nd6_llinfo_settimer(ln, (long)nd6_gctimer * hz); 2051 } 2052 2053 /* 2054 * The first time we send a packet to a neighbor whose entry is 2055 * STALE, we have to change the state to DELAY and a sets a timer to 2056 * expire in DELAY_FIRST_PROBE_TIME seconds to ensure do 2057 * neighbor unreachability detection on expiration. 2058 * (RFC 2461 7.3.3) 2059 */ 2060 if (ln->ln_state == ND6_LLINFO_STALE) { 2061 ln->ln_asked = 0; 2062 ln->ln_state = ND6_LLINFO_DELAY; 2063 nd6_llinfo_settimer(ln, (long)nd6_delay * hz); 2064 } 2065 2066 /* 2067 * If the neighbor cache entry has a state other than INCOMPLETE 2068 * (i.e. its link-layer address is already resolved), just 2069 * send the packet. 2070 */ 2071 if (ln->ln_state > ND6_LLINFO_INCOMPLETE) 2072 goto sendpkt; 2073 2074 /* 2075 * There is a neighbor cache entry, but no ethernet address 2076 * response yet. Append this latest packet to the end of the 2077 * packet queue in the mbuf, unless the number of the packet 2078 * does not exceed nd6_maxqueuelen. When it exceeds nd6_maxqueuelen, 2079 * the oldest packet in the queue will be removed. 2080 */ 2081 if (ln->ln_state == ND6_LLINFO_NOSTATE) 2082 ln->ln_state = ND6_LLINFO_INCOMPLETE; 2083 if (ln->ln_hold) { 2084 struct mbuf *m_hold; 2085 int i; 2086 2087 i = 0; 2088 for (m_hold = ln->ln_hold; m_hold; m_hold = m_hold->m_nextpkt) { 2089 i++; 2090 if (m_hold->m_nextpkt == NULL) { 2091 m_hold->m_nextpkt = m; 2092 break; 2093 } 2094 } 2095 while (i >= nd6_maxqueuelen) { 2096 m_hold = ln->ln_hold; 2097 ln->ln_hold = ln->ln_hold->m_nextpkt; 2098 m_freem(m_hold); 2099 i--; 2100 } 2101 } else { 2102 ln->ln_hold = m; 2103 } 2104 2105 /* 2106 * If there has been no NS for the neighbor after entering the 2107 * INCOMPLETE state, send the first solicitation. 2108 */ 2109 if (!ND6_LLINFO_PERMANENT(ln) && ln->ln_asked == 0) { 2110 ln->ln_asked++; 2111 nd6_llinfo_settimer(ln, 2112 (long)ND_IFINFO(ifp)->retrans * hz / 1000); 2113 nd6_ns_output(ifp, NULL, &dst->sin6_addr, ln, 0); 2114 } 2115 return 0; 2116 2117 sendpkt: 2118 /* discard the packet if IPv6 operation is disabled on the interface */ 2119 if ((ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED)) { 2120 error = ENETDOWN; /* better error? */ 2121 goto bad; 2122 } 2123 2124 if ((ifp->if_flags & IFF_LOOPBACK) != 0) 2125 return (*ifp->if_output)(origifp, m, sin6tocsa(dst), rt); 2126 return (*ifp->if_output)(ifp, m, sin6tocsa(dst), rt); 2127 2128 bad: 2129 if (m != NULL) 2130 m_freem(m); 2131 return error; 2132 } 2133 #undef senderr 2134 2135 int 2136 nd6_need_cache(struct ifnet *ifp) 2137 { 2138 /* 2139 * XXX: we currently do not make neighbor cache on any interface 2140 * other than ARCnet, Ethernet, FDDI and GIF. 2141 * 2142 * RFC2893 says: 2143 * - unidirectional tunnels needs no ND 2144 */ 2145 switch (ifp->if_type) { 2146 case IFT_ARCNET: 2147 case IFT_ETHER: 2148 case IFT_FDDI: 2149 case IFT_IEEE1394: 2150 case IFT_CARP: 2151 case IFT_GIF: /* XXX need more cases? */ 2152 case IFT_PPP: 2153 case IFT_TUNNEL: 2154 return 1; 2155 default: 2156 return 0; 2157 } 2158 } 2159 2160 int 2161 nd6_storelladdr(const struct ifnet *ifp, const struct rtentry *rt, 2162 struct mbuf *m, const struct sockaddr *dst, uint8_t *lldst, 2163 size_t dstsize) 2164 { 2165 const struct sockaddr_dl *sdl; 2166 2167 if (m->m_flags & M_MCAST) { 2168 switch (ifp->if_type) { 2169 case IFT_ETHER: 2170 case IFT_FDDI: 2171 ETHER_MAP_IPV6_MULTICAST(&satocsin6(dst)->sin6_addr, 2172 lldst); 2173 return 1; 2174 case IFT_IEEE1394: 2175 memcpy(lldst, ifp->if_broadcastaddr, 2176 MIN(dstsize, ifp->if_addrlen)); 2177 return 1; 2178 case IFT_ARCNET: 2179 *lldst = 0; 2180 return 1; 2181 default: 2182 m_freem(m); 2183 return 0; 2184 } 2185 } 2186 2187 if (rt == NULL) { 2188 /* this could happen, if we could not allocate memory */ 2189 m_freem(m); 2190 return 0; 2191 } 2192 if (rt->rt_gateway->sa_family != AF_LINK) { 2193 printf("%s: something odd happens\n", __func__); 2194 m_freem(m); 2195 return 0; 2196 } 2197 sdl = satocsdl(rt->rt_gateway); 2198 if (sdl->sdl_alen == 0 || sdl->sdl_alen > dstsize) { 2199 /* this should be impossible, but we bark here for debugging */ 2200 printf("%s: sdl_alen == %" PRIu8 ", dst=%s, if=%s\n", __func__, 2201 sdl->sdl_alen, ip6_sprintf(&satocsin6(dst)->sin6_addr), 2202 if_name(ifp)); 2203 m_freem(m); 2204 return 0; 2205 } 2206 2207 memcpy(lldst, CLLADDR(sdl), MIN(dstsize, sdl->sdl_alen)); 2208 return 1; 2209 } 2210 2211 static void 2212 clear_llinfo_pqueue(struct llinfo_nd6 *ln) 2213 { 2214 struct mbuf *m_hold, *m_hold_next; 2215 2216 for (m_hold = ln->ln_hold; m_hold; m_hold = m_hold_next) { 2217 m_hold_next = m_hold->m_nextpkt; 2218 m_hold->m_nextpkt = NULL; 2219 m_freem(m_hold); 2220 } 2221 2222 ln->ln_hold = NULL; 2223 return; 2224 } 2225 2226 int 2227 nd6_sysctl( 2228 int name, 2229 void *oldp, /* syscall arg, need copyout */ 2230 size_t *oldlenp, 2231 void *newp, /* syscall arg, need copyin */ 2232 size_t newlen 2233 ) 2234 { 2235 void *p; 2236 size_t ol; 2237 int error; 2238 2239 error = 0; 2240 2241 if (newp) 2242 return EPERM; 2243 if (oldp && !oldlenp) 2244 return EINVAL; 2245 ol = oldlenp ? *oldlenp : 0; 2246 2247 if (oldp) { 2248 p = malloc(*oldlenp, M_TEMP, M_WAITOK); 2249 if (p == NULL) 2250 return ENOMEM; 2251 } else 2252 p = NULL; 2253 switch (name) { 2254 case ICMPV6CTL_ND6_DRLIST: 2255 error = fill_drlist(p, oldlenp, ol); 2256 if (!error && p != NULL && oldp != NULL) 2257 error = copyout(p, oldp, *oldlenp); 2258 break; 2259 2260 case ICMPV6CTL_ND6_PRLIST: 2261 error = fill_prlist(p, oldlenp, ol); 2262 if (!error && p != NULL && oldp != NULL) 2263 error = copyout(p, oldp, *oldlenp); 2264 break; 2265 2266 case ICMPV6CTL_ND6_MAXQLEN: 2267 break; 2268 2269 default: 2270 error = ENOPROTOOPT; 2271 break; 2272 } 2273 if (p) 2274 free(p, M_TEMP); 2275 2276 return error; 2277 } 2278 2279 static int 2280 fill_drlist(void *oldp, size_t *oldlenp, size_t ol) 2281 { 2282 int error = 0, s; 2283 struct in6_defrouter *d = NULL, *de = NULL; 2284 struct nd_defrouter *dr; 2285 size_t l; 2286 2287 s = splsoftnet(); 2288 2289 if (oldp) { 2290 d = (struct in6_defrouter *)oldp; 2291 de = (struct in6_defrouter *)((char *)oldp + *oldlenp); 2292 } 2293 l = 0; 2294 2295 TAILQ_FOREACH(dr, &nd_defrouter, dr_entry) { 2296 2297 if (oldp && d + 1 <= de) { 2298 memset(d, 0, sizeof(*d)); 2299 sockaddr_in6_init(&d->rtaddr, &dr->rtaddr, 0, 0, 0); 2300 if (sa6_recoverscope(&d->rtaddr)) { 2301 log(LOG_ERR, 2302 "scope error in router list (%s)\n", 2303 ip6_sprintf(&d->rtaddr.sin6_addr)); 2304 /* XXX: press on... */ 2305 } 2306 d->flags = dr->flags; 2307 d->rtlifetime = dr->rtlifetime; 2308 d->expire = dr->expire; 2309 d->if_index = dr->ifp->if_index; 2310 } 2311 2312 l += sizeof(*d); 2313 if (d) 2314 d++; 2315 } 2316 2317 if (oldp) { 2318 if (l > ol) 2319 error = ENOMEM; 2320 } 2321 if (oldlenp) 2322 *oldlenp = l; /* (void *)d - (void *)oldp */ 2323 2324 splx(s); 2325 2326 return error; 2327 } 2328 2329 static int 2330 fill_prlist(void *oldp, size_t *oldlenp, size_t ol) 2331 { 2332 int error = 0, s; 2333 struct nd_prefix *pr; 2334 struct in6_prefix *p = NULL; 2335 struct in6_prefix *pe = NULL; 2336 size_t l; 2337 2338 s = splsoftnet(); 2339 2340 if (oldp) { 2341 p = (struct in6_prefix *)oldp; 2342 pe = (struct in6_prefix *)((char *)oldp + *oldlenp); 2343 } 2344 l = 0; 2345 2346 LIST_FOREACH(pr, &nd_prefix, ndpr_entry) { 2347 u_short advrtrs; 2348 size_t advance; 2349 struct sockaddr_in6 *sin6; 2350 struct sockaddr_in6 *s6; 2351 struct nd_pfxrouter *pfr; 2352 2353 if (oldp && p + 1 <= pe) 2354 { 2355 memset(p, 0, sizeof(*p)); 2356 sin6 = (struct sockaddr_in6 *)(p + 1); 2357 2358 p->prefix = pr->ndpr_prefix; 2359 if (sa6_recoverscope(&p->prefix)) { 2360 log(LOG_ERR, 2361 "scope error in prefix list (%s)\n", 2362 ip6_sprintf(&p->prefix.sin6_addr)); 2363 /* XXX: press on... */ 2364 } 2365 p->raflags = pr->ndpr_raf; 2366 p->prefixlen = pr->ndpr_plen; 2367 p->vltime = pr->ndpr_vltime; 2368 p->pltime = pr->ndpr_pltime; 2369 p->if_index = pr->ndpr_ifp->if_index; 2370 if (pr->ndpr_vltime == ND6_INFINITE_LIFETIME) 2371 p->expire = 0; 2372 else { 2373 time_t maxexpire; 2374 2375 /* XXX: we assume time_t is signed. */ 2376 maxexpire = (-1) & 2377 ~((time_t)1 << 2378 ((sizeof(maxexpire) * 8) - 1)); 2379 if (pr->ndpr_vltime < 2380 maxexpire - pr->ndpr_lastupdate) { 2381 p->expire = pr->ndpr_lastupdate + 2382 pr->ndpr_vltime; 2383 } else 2384 p->expire = maxexpire; 2385 } 2386 p->refcnt = pr->ndpr_refcnt; 2387 p->flags = pr->ndpr_stateflags; 2388 p->origin = PR_ORIG_RA; 2389 advrtrs = 0; 2390 LIST_FOREACH(pfr, &pr->ndpr_advrtrs, pfr_entry) { 2391 if ((void *)&sin6[advrtrs + 1] > (void *)pe) { 2392 advrtrs++; 2393 continue; 2394 } 2395 s6 = &sin6[advrtrs]; 2396 sockaddr_in6_init(s6, &pfr->router->rtaddr, 2397 0, 0, 0); 2398 if (sa6_recoverscope(s6)) { 2399 log(LOG_ERR, 2400 "scope error in " 2401 "prefix list (%s)\n", 2402 ip6_sprintf(&pfr->router->rtaddr)); 2403 } 2404 advrtrs++; 2405 } 2406 p->advrtrs = advrtrs; 2407 } 2408 else { 2409 advrtrs = 0; 2410 LIST_FOREACH(pfr, &pr->ndpr_advrtrs, pfr_entry) 2411 advrtrs++; 2412 } 2413 2414 advance = sizeof(*p) + sizeof(*sin6) * advrtrs; 2415 l += advance; 2416 if (p) 2417 p = (struct in6_prefix *)((char *)p + advance); 2418 } 2419 2420 if (oldp) { 2421 *oldlenp = l; /* (void *)d - (void *)oldp */ 2422 if (l > ol) 2423 error = ENOMEM; 2424 } else 2425 *oldlenp = l; 2426 2427 splx(s); 2428 2429 return error; 2430 } 2431