1 /* $NetBSD: nd6.c,v 1.132 2009/07/25 23:12:09 tonnerre 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 1.132 2009/07/25 23:12:09 tonnerre Exp $"); 35 36 #include "opt_ipsec.h" 37 38 #include <sys/param.h> 39 #include <sys/systm.h> 40 #include <sys/callout.h> 41 #include <sys/malloc.h> 42 #include <sys/mbuf.h> 43 #include <sys/socket.h> 44 #include <sys/socketvar.h> 45 #include <sys/sockio.h> 46 #include <sys/time.h> 47 #include <sys/kernel.h> 48 #include <sys/protosw.h> 49 #include <sys/errno.h> 50 #include <sys/ioctl.h> 51 #include <sys/syslog.h> 52 #include <sys/queue.h> 53 54 #include <net/if.h> 55 #include <net/if_dl.h> 56 #include <net/if_types.h> 57 #include <net/route.h> 58 #include <net/if_ether.h> 59 #include <net/if_fddi.h> 60 #include <net/if_arc.h> 61 62 #include <netinet/in.h> 63 #include <netinet6/in6_var.h> 64 #include <netinet/ip6.h> 65 #include <netinet6/ip6_var.h> 66 #include <netinet6/scope6_var.h> 67 #include <netinet6/nd6.h> 68 #include <netinet/icmp6.h> 69 #include <netinet6/icmp6_private.h> 70 71 #ifdef IPSEC 72 #include <netinet6/ipsec.h> 73 #endif 74 75 #include <net/net_osdep.h> 76 77 #define ND6_SLOWTIMER_INTERVAL (60 * 60) /* 1 hour */ 78 #define ND6_RECALC_REACHTM_INTERVAL (60 * 120) /* 2 hours */ 79 80 /* timer values */ 81 int nd6_prune = 1; /* walk list every 1 seconds */ 82 int nd6_delay = 5; /* delay first probe time 5 second */ 83 int nd6_umaxtries = 3; /* maximum unicast query */ 84 int nd6_mmaxtries = 3; /* maximum multicast query */ 85 int nd6_useloopback = 1; /* use loopback interface for local traffic */ 86 int nd6_gctimer = (60 * 60 * 24); /* 1 day: garbage collection timer */ 87 88 /* preventing too many loops in ND option parsing */ 89 int nd6_maxndopt = 10; /* max # of ND options allowed */ 90 91 int nd6_maxnudhint = 0; /* max # of subsequent upper layer hints */ 92 93 int nd6_maxqueuelen = 1; /* max # of packets cached in unresolved ND entries */ 94 95 #ifdef ND6_DEBUG 96 int nd6_debug = 1; 97 #else 98 int nd6_debug = 0; 99 #endif 100 101 /* for debugging? */ 102 static int nd6_inuse, nd6_allocated; 103 104 struct llinfo_nd6 llinfo_nd6 = { 105 .ln_prev = &llinfo_nd6, 106 .ln_next = &llinfo_nd6, 107 }; 108 struct nd_drhead nd_defrouter; 109 struct nd_prhead nd_prefix = { 0 }; 110 111 int nd6_recalc_reachtm_interval = ND6_RECALC_REACHTM_INTERVAL; 112 static const struct sockaddr_in6 all1_sa = { 113 .sin6_family = AF_INET6 114 , .sin6_len = sizeof(struct sockaddr_in6) 115 , .sin6_addr = {.s6_addr = {0xff, 0xff, 0xff, 0xff, 116 0xff, 0xff, 0xff, 0xff, 117 0xff, 0xff, 0xff, 0xff, 118 0xff, 0xff, 0xff, 0xff}} 119 }; 120 121 static void nd6_setmtu0(struct ifnet *, struct nd_ifinfo *); 122 static void nd6_slowtimo(void *); 123 static int regen_tmpaddr(struct in6_ifaddr *); 124 static struct llinfo_nd6 *nd6_free(struct rtentry *, int); 125 static void nd6_llinfo_timer(void *); 126 static void clear_llinfo_pqueue(struct llinfo_nd6 *); 127 128 callout_t nd6_slowtimo_ch; 129 callout_t nd6_timer_ch; 130 extern callout_t in6_tmpaddrtimer_ch; 131 132 static int fill_drlist(void *, size_t *, size_t); 133 static int fill_prlist(void *, size_t *, size_t); 134 135 MALLOC_DEFINE(M_IP6NDP, "NDP", "IPv6 Neighbour Discovery"); 136 137 void 138 nd6_init(void) 139 { 140 static int nd6_init_done = 0; 141 142 if (nd6_init_done) { 143 log(LOG_NOTICE, "nd6_init called more than once(ignored)\n"); 144 return; 145 } 146 147 /* initialization of the default router list */ 148 TAILQ_INIT(&nd_defrouter); 149 150 nd6_init_done = 1; 151 152 callout_init(&nd6_slowtimo_ch, CALLOUT_MPSAFE); 153 callout_init(&nd6_timer_ch, CALLOUT_MPSAFE); 154 155 /* start timer */ 156 callout_reset(&nd6_slowtimo_ch, ND6_SLOWTIMER_INTERVAL * hz, 157 nd6_slowtimo, NULL); 158 } 159 160 struct nd_ifinfo * 161 nd6_ifattach(struct ifnet *ifp) 162 { 163 struct nd_ifinfo *nd; 164 165 nd = (struct nd_ifinfo *)malloc(sizeof(*nd), M_IP6NDP, M_WAITOK); 166 memset(nd, 0, sizeof(*nd)); 167 168 nd->initialized = 1; 169 170 nd->chlim = IPV6_DEFHLIM; 171 nd->basereachable = REACHABLE_TIME; 172 nd->reachable = ND_COMPUTE_RTIME(nd->basereachable); 173 nd->retrans = RETRANS_TIMER; 174 /* 175 * Note that the default value of ip6_accept_rtadv is 0, which means 176 * we won't accept RAs by default even if we set ND6_IFF_ACCEPT_RTADV 177 * here. 178 */ 179 nd->flags = (ND6_IFF_PERFORMNUD | 180 ip6_accept_rtadv ? ND6_IFF_ACCEPT_RTADV : 0); 181 182 /* XXX: we cannot call nd6_setmtu since ifp is not fully initialized */ 183 nd6_setmtu0(ifp, nd); 184 185 return nd; 186 } 187 188 void 189 nd6_ifdetach(struct nd_ifinfo *nd) 190 { 191 192 free(nd, M_IP6NDP); 193 } 194 195 void 196 nd6_setmtu(struct ifnet *ifp) 197 { 198 nd6_setmtu0(ifp, ND_IFINFO(ifp)); 199 } 200 201 void 202 nd6_setmtu0(struct ifnet *ifp, struct nd_ifinfo *ndi) 203 { 204 u_int32_t omaxmtu; 205 206 omaxmtu = ndi->maxmtu; 207 208 switch (ifp->if_type) { 209 case IFT_ARCNET: 210 ndi->maxmtu = MIN(ARC_PHDS_MAXMTU, ifp->if_mtu); /* RFC2497 */ 211 break; 212 case IFT_FDDI: 213 ndi->maxmtu = MIN(FDDIIPMTU, ifp->if_mtu); 214 break; 215 default: 216 ndi->maxmtu = ifp->if_mtu; 217 break; 218 } 219 220 /* 221 * Decreasing the interface MTU under IPV6 minimum MTU may cause 222 * undesirable situation. We thus notify the operator of the change 223 * explicitly. The check for omaxmtu is necessary to restrict the 224 * log to the case of changing the MTU, not initializing it. 225 */ 226 if (omaxmtu >= IPV6_MMTU && ndi->maxmtu < IPV6_MMTU) { 227 log(LOG_NOTICE, "nd6_setmtu0: new link MTU on %s (%lu) is too" 228 " small for IPv6 which needs %lu\n", 229 if_name(ifp), (unsigned long)ndi->maxmtu, (unsigned long) 230 IPV6_MMTU); 231 } 232 233 if (ndi->maxmtu > in6_maxmtu) 234 in6_setmaxmtu(); /* check all interfaces just in case */ 235 } 236 237 void 238 nd6_option_init(void *opt, int icmp6len, union nd_opts *ndopts) 239 { 240 241 memset(ndopts, 0, sizeof(*ndopts)); 242 ndopts->nd_opts_search = (struct nd_opt_hdr *)opt; 243 ndopts->nd_opts_last 244 = (struct nd_opt_hdr *)(((u_char *)opt) + icmp6len); 245 246 if (icmp6len == 0) { 247 ndopts->nd_opts_done = 1; 248 ndopts->nd_opts_search = NULL; 249 } 250 } 251 252 /* 253 * Take one ND option. 254 */ 255 struct nd_opt_hdr * 256 nd6_option(union nd_opts *ndopts) 257 { 258 struct nd_opt_hdr *nd_opt; 259 int olen; 260 261 if (ndopts == NULL) 262 panic("ndopts == NULL in nd6_option"); 263 if (ndopts->nd_opts_last == NULL) 264 panic("uninitialized ndopts in nd6_option"); 265 if (ndopts->nd_opts_search == NULL) 266 return NULL; 267 if (ndopts->nd_opts_done) 268 return NULL; 269 270 nd_opt = ndopts->nd_opts_search; 271 272 /* make sure nd_opt_len is inside the buffer */ 273 if ((void *)&nd_opt->nd_opt_len >= (void *)ndopts->nd_opts_last) { 274 memset(ndopts, 0, sizeof(*ndopts)); 275 return NULL; 276 } 277 278 olen = nd_opt->nd_opt_len << 3; 279 if (olen == 0) { 280 /* 281 * Message validation requires that all included 282 * options have a length that is greater than zero. 283 */ 284 memset(ndopts, 0, sizeof(*ndopts)); 285 return NULL; 286 } 287 288 ndopts->nd_opts_search = (struct nd_opt_hdr *)((char *)nd_opt + olen); 289 if (ndopts->nd_opts_search > ndopts->nd_opts_last) { 290 /* option overruns the end of buffer, invalid */ 291 memset(ndopts, 0, sizeof(*ndopts)); 292 return NULL; 293 } else if (ndopts->nd_opts_search == ndopts->nd_opts_last) { 294 /* reached the end of options chain */ 295 ndopts->nd_opts_done = 1; 296 ndopts->nd_opts_search = NULL; 297 } 298 return nd_opt; 299 } 300 301 /* 302 * Parse multiple ND options. 303 * This function is much easier to use, for ND routines that do not need 304 * multiple options of the same type. 305 */ 306 int 307 nd6_options(union nd_opts *ndopts) 308 { 309 struct nd_opt_hdr *nd_opt; 310 int i = 0; 311 312 if (ndopts == NULL) 313 panic("ndopts == NULL in nd6_options"); 314 if (ndopts->nd_opts_last == NULL) 315 panic("uninitialized ndopts in nd6_options"); 316 if (ndopts->nd_opts_search == NULL) 317 return 0; 318 319 while (1) { 320 nd_opt = nd6_option(ndopts); 321 if (nd_opt == NULL && ndopts->nd_opts_last == NULL) { 322 /* 323 * Message validation requires that all included 324 * options have a length that is greater than zero. 325 */ 326 ICMP6_STATINC(ICMP6_STAT_ND_BADOPT); 327 memset(ndopts, 0, sizeof(*ndopts)); 328 return -1; 329 } 330 331 if (nd_opt == NULL) 332 goto skip1; 333 334 switch (nd_opt->nd_opt_type) { 335 case ND_OPT_SOURCE_LINKADDR: 336 case ND_OPT_TARGET_LINKADDR: 337 case ND_OPT_MTU: 338 case ND_OPT_REDIRECTED_HEADER: 339 if (ndopts->nd_opt_array[nd_opt->nd_opt_type]) { 340 nd6log((LOG_INFO, 341 "duplicated ND6 option found (type=%d)\n", 342 nd_opt->nd_opt_type)); 343 /* XXX bark? */ 344 } else { 345 ndopts->nd_opt_array[nd_opt->nd_opt_type] 346 = nd_opt; 347 } 348 break; 349 case ND_OPT_PREFIX_INFORMATION: 350 if (ndopts->nd_opt_array[nd_opt->nd_opt_type] == 0) { 351 ndopts->nd_opt_array[nd_opt->nd_opt_type] 352 = nd_opt; 353 } 354 ndopts->nd_opts_pi_end = 355 (struct nd_opt_prefix_info *)nd_opt; 356 break; 357 default: 358 /* 359 * Unknown options must be silently ignored, 360 * to accommodate future extension to the protocol. 361 */ 362 nd6log((LOG_DEBUG, 363 "nd6_options: unsupported option %d - " 364 "option ignored\n", nd_opt->nd_opt_type)); 365 } 366 367 skip1: 368 i++; 369 if (i > nd6_maxndopt) { 370 ICMP6_STATINC(ICMP6_STAT_ND_TOOMANYOPT); 371 nd6log((LOG_INFO, "too many loop in nd opt\n")); 372 break; 373 } 374 375 if (ndopts->nd_opts_done) 376 break; 377 } 378 379 return 0; 380 } 381 382 /* 383 * ND6 timer routine to handle ND6 entries 384 */ 385 void 386 nd6_llinfo_settimer(struct llinfo_nd6 *ln, long xtick) 387 { 388 int s; 389 390 s = splsoftnet(); 391 392 if (xtick < 0) { 393 ln->ln_expire = 0; 394 ln->ln_ntick = 0; 395 callout_stop(&ln->ln_timer_ch); 396 } else { 397 ln->ln_expire = time_second + xtick / hz; 398 if (xtick > INT_MAX) { 399 ln->ln_ntick = xtick - INT_MAX; 400 callout_reset(&ln->ln_timer_ch, INT_MAX, 401 nd6_llinfo_timer, ln); 402 } else { 403 ln->ln_ntick = 0; 404 callout_reset(&ln->ln_timer_ch, xtick, 405 nd6_llinfo_timer, ln); 406 } 407 } 408 409 splx(s); 410 } 411 412 static void 413 nd6_llinfo_timer(void *arg) 414 { 415 struct llinfo_nd6 *ln; 416 struct rtentry *rt; 417 const struct sockaddr_in6 *dst; 418 struct ifnet *ifp; 419 struct nd_ifinfo *ndi = NULL; 420 421 mutex_enter(softnet_lock); 422 KERNEL_LOCK(1, NULL); 423 424 ln = (struct llinfo_nd6 *)arg; 425 426 if (ln->ln_ntick > 0) { 427 nd6_llinfo_settimer(ln, ln->ln_ntick); 428 KERNEL_UNLOCK_ONE(NULL); 429 mutex_exit(softnet_lock); 430 return; 431 } 432 433 if ((rt = ln->ln_rt) == NULL) 434 panic("ln->ln_rt == NULL"); 435 if ((ifp = rt->rt_ifp) == NULL) 436 panic("ln->ln_rt->rt_ifp == NULL"); 437 ndi = ND_IFINFO(ifp); 438 dst = satocsin6(rt_getkey(rt)); 439 440 /* sanity check */ 441 if (rt->rt_llinfo && (struct llinfo_nd6 *)rt->rt_llinfo != ln) 442 panic("rt_llinfo(%p) is not equal to ln(%p)", 443 rt->rt_llinfo, ln); 444 if (!dst) 445 panic("dst=0 in nd6_timer(ln=%p)", ln); 446 447 switch (ln->ln_state) { 448 case ND6_LLINFO_INCOMPLETE: 449 if (ln->ln_asked < nd6_mmaxtries) { 450 ln->ln_asked++; 451 nd6_llinfo_settimer(ln, (long)ndi->retrans * hz / 1000); 452 nd6_ns_output(ifp, NULL, &dst->sin6_addr, ln, 0); 453 } else { 454 struct mbuf *m = ln->ln_hold; 455 if (m) { 456 struct mbuf *m0; 457 458 /* 459 * assuming every packet in ln_hold has 460 * the same IP header 461 */ 462 m0 = m->m_nextpkt; 463 m->m_nextpkt = NULL; 464 icmp6_error2(m, ICMP6_DST_UNREACH, 465 ICMP6_DST_UNREACH_ADDR, 0, rt->rt_ifp); 466 467 ln->ln_hold = m0; 468 clear_llinfo_pqueue(ln); 469 } 470 (void)nd6_free(rt, 0); 471 ln = NULL; 472 } 473 break; 474 case ND6_LLINFO_REACHABLE: 475 if (!ND6_LLINFO_PERMANENT(ln)) { 476 ln->ln_state = ND6_LLINFO_STALE; 477 nd6_llinfo_settimer(ln, (long)nd6_gctimer * hz); 478 } 479 break; 480 481 case ND6_LLINFO_STALE: 482 /* Garbage Collection(RFC 2461 5.3) */ 483 if (!ND6_LLINFO_PERMANENT(ln)) { 484 (void)nd6_free(rt, 1); 485 ln = NULL; 486 } 487 break; 488 489 case ND6_LLINFO_DELAY: 490 if (ndi && (ndi->flags & ND6_IFF_PERFORMNUD) != 0) { 491 /* We need NUD */ 492 ln->ln_asked = 1; 493 ln->ln_state = ND6_LLINFO_PROBE; 494 nd6_llinfo_settimer(ln, (long)ndi->retrans * hz / 1000); 495 nd6_ns_output(ifp, &dst->sin6_addr, 496 &dst->sin6_addr, ln, 0); 497 } else { 498 ln->ln_state = ND6_LLINFO_STALE; /* XXX */ 499 nd6_llinfo_settimer(ln, (long)nd6_gctimer * hz); 500 } 501 break; 502 case ND6_LLINFO_PROBE: 503 if (ln->ln_asked < nd6_umaxtries) { 504 ln->ln_asked++; 505 nd6_llinfo_settimer(ln, (long)ndi->retrans * hz / 1000); 506 nd6_ns_output(ifp, &dst->sin6_addr, 507 &dst->sin6_addr, ln, 0); 508 } else { 509 (void)nd6_free(rt, 0); 510 ln = NULL; 511 } 512 break; 513 } 514 515 KERNEL_UNLOCK_ONE(NULL); 516 mutex_exit(softnet_lock); 517 } 518 519 /* 520 * ND6 timer routine to expire default route list and prefix list 521 */ 522 void 523 nd6_timer(void *ignored_arg) 524 { 525 struct nd_defrouter *next_dr, *dr; 526 struct nd_prefix *next_pr, *pr; 527 struct in6_ifaddr *ia6, *nia6; 528 struct in6_addrlifetime *lt6; 529 530 callout_reset(&nd6_timer_ch, nd6_prune * hz, 531 nd6_timer, NULL); 532 533 mutex_enter(softnet_lock); 534 KERNEL_LOCK(1, NULL); 535 536 /* expire default router list */ 537 538 for (dr = TAILQ_FIRST(&nd_defrouter); dr != NULL; dr = next_dr) { 539 next_dr = TAILQ_NEXT(dr, dr_entry); 540 if (dr->expire && dr->expire < time_second) { 541 defrtrlist_del(dr); 542 } 543 } 544 545 /* 546 * expire interface addresses. 547 * in the past the loop was inside prefix expiry processing. 548 * However, from a stricter speci-confrmance standpoint, we should 549 * rather separate address lifetimes and prefix lifetimes. 550 */ 551 addrloop: 552 for (ia6 = in6_ifaddr; ia6; ia6 = nia6) { 553 nia6 = ia6->ia_next; 554 /* check address lifetime */ 555 lt6 = &ia6->ia6_lifetime; 556 if (IFA6_IS_INVALID(ia6)) { 557 int regen = 0; 558 559 /* 560 * If the expiring address is temporary, try 561 * regenerating a new one. This would be useful when 562 * we suspended a laptop PC, then turned it on after a 563 * period that could invalidate all temporary 564 * addresses. Although we may have to restart the 565 * loop (see below), it must be after purging the 566 * address. Otherwise, we'd see an infinite loop of 567 * regeneration. 568 */ 569 if (ip6_use_tempaddr && 570 (ia6->ia6_flags & IN6_IFF_TEMPORARY) != 0) { 571 if (regen_tmpaddr(ia6) == 0) 572 regen = 1; 573 } 574 575 in6_purgeaddr(&ia6->ia_ifa); 576 577 if (regen) 578 goto addrloop; /* XXX: see below */ 579 } else if (IFA6_IS_DEPRECATED(ia6)) { 580 int oldflags = ia6->ia6_flags; 581 582 ia6->ia6_flags |= IN6_IFF_DEPRECATED; 583 584 /* 585 * If a temporary address has just become deprecated, 586 * regenerate a new one if possible. 587 */ 588 if (ip6_use_tempaddr && 589 (ia6->ia6_flags & IN6_IFF_TEMPORARY) != 0 && 590 (oldflags & IN6_IFF_DEPRECATED) == 0) { 591 592 if (regen_tmpaddr(ia6) == 0) { 593 /* 594 * A new temporary address is 595 * generated. 596 * XXX: this means the address chain 597 * has changed while we are still in 598 * the loop. Although the change 599 * would not cause disaster (because 600 * it's not a deletion, but an 601 * addition,) we'd rather restart the 602 * loop just for safety. Or does this 603 * significantly reduce performance?? 604 */ 605 goto addrloop; 606 } 607 } 608 } else { 609 /* 610 * A new RA might have made a deprecated address 611 * preferred. 612 */ 613 ia6->ia6_flags &= ~IN6_IFF_DEPRECATED; 614 } 615 } 616 617 /* expire prefix list */ 618 for (pr = LIST_FIRST(&nd_prefix); pr != NULL; pr = next_pr) { 619 next_pr = LIST_NEXT(pr, ndpr_entry); 620 /* 621 * check prefix lifetime. 622 * since pltime is just for autoconf, pltime processing for 623 * prefix is not necessary. 624 */ 625 if (pr->ndpr_vltime != ND6_INFINITE_LIFETIME && 626 time_second - pr->ndpr_lastupdate > pr->ndpr_vltime) { 627 628 /* 629 * address expiration and prefix expiration are 630 * separate. NEVER perform in6_purgeaddr here. 631 */ 632 633 prelist_remove(pr); 634 } 635 } 636 637 KERNEL_UNLOCK_ONE(NULL); 638 mutex_exit(softnet_lock); 639 } 640 641 /* ia6: deprecated/invalidated temporary address */ 642 static int 643 regen_tmpaddr(struct in6_ifaddr *ia6) 644 { 645 struct ifaddr *ifa; 646 struct ifnet *ifp; 647 struct in6_ifaddr *public_ifa6 = NULL; 648 649 ifp = ia6->ia_ifa.ifa_ifp; 650 IFADDR_FOREACH(ifa, ifp) { 651 struct in6_ifaddr *it6; 652 653 if (ifa->ifa_addr->sa_family != AF_INET6) 654 continue; 655 656 it6 = (struct in6_ifaddr *)ifa; 657 658 /* ignore no autoconf addresses. */ 659 if ((it6->ia6_flags & IN6_IFF_AUTOCONF) == 0) 660 continue; 661 662 /* ignore autoconf addresses with different prefixes. */ 663 if (it6->ia6_ndpr == NULL || it6->ia6_ndpr != ia6->ia6_ndpr) 664 continue; 665 666 /* 667 * Now we are looking at an autoconf address with the same 668 * prefix as ours. If the address is temporary and is still 669 * preferred, do not create another one. It would be rare, but 670 * could happen, for example, when we resume a laptop PC after 671 * a long period. 672 */ 673 if ((it6->ia6_flags & IN6_IFF_TEMPORARY) != 0 && 674 !IFA6_IS_DEPRECATED(it6)) { 675 public_ifa6 = NULL; 676 break; 677 } 678 679 /* 680 * This is a public autoconf address that has the same prefix 681 * as ours. If it is preferred, keep it. We can't break the 682 * loop here, because there may be a still-preferred temporary 683 * address with the prefix. 684 */ 685 if (!IFA6_IS_DEPRECATED(it6)) 686 public_ifa6 = it6; 687 } 688 689 if (public_ifa6 != NULL) { 690 int e; 691 692 /* 693 * Random factor is introduced in the preferred lifetime, so 694 * we do not need additional delay (3rd arg to in6_tmpifadd). 695 */ 696 if ((e = in6_tmpifadd(public_ifa6, 0, 0)) != 0) { 697 log(LOG_NOTICE, "regen_tmpaddr: failed to create a new" 698 " tmp addr, errno=%d\n", e); 699 return -1; 700 } 701 return 0; 702 } 703 704 return -1; 705 } 706 707 /* 708 * Nuke neighbor cache/prefix/default router management table, right before 709 * ifp goes away. 710 */ 711 void 712 nd6_purge(struct ifnet *ifp) 713 { 714 struct nd_ifinfo *ndi = ND_IFINFO(ifp); 715 struct llinfo_nd6 *ln, *nln; 716 struct nd_defrouter *dr, *ndr; 717 struct nd_prefix *pr, *npr; 718 719 /* 720 * Nuke default router list entries toward ifp. 721 * We defer removal of default router list entries that is installed 722 * in the routing table, in order to keep additional side effects as 723 * small as possible. 724 */ 725 for (dr = TAILQ_FIRST(&nd_defrouter); dr != NULL; dr = ndr) { 726 ndr = TAILQ_NEXT(dr, dr_entry); 727 if (dr->installed) 728 continue; 729 730 if (dr->ifp == ifp) 731 defrtrlist_del(dr); 732 } 733 for (dr = TAILQ_FIRST(&nd_defrouter); dr != NULL; dr = ndr) { 734 ndr = TAILQ_NEXT(dr, dr_entry); 735 if (!dr->installed) 736 continue; 737 738 if (dr->ifp == ifp) 739 defrtrlist_del(dr); 740 } 741 742 /* Nuke prefix list entries toward ifp */ 743 for (pr = LIST_FIRST(&nd_prefix); pr != NULL; pr = npr) { 744 npr = LIST_NEXT(pr, ndpr_entry); 745 if (pr->ndpr_ifp == ifp) { 746 /* 747 * Because if_detach() does *not* release prefixes 748 * while purging addresses the reference count will 749 * still be above zero. We therefore reset it to 750 * make sure that the prefix really gets purged. 751 */ 752 pr->ndpr_refcnt = 0; 753 /* 754 * Previously, pr->ndpr_addr is removed as well, 755 * but I strongly believe we don't have to do it. 756 * nd6_purge() is only called from in6_ifdetach(), 757 * which removes all the associated interface addresses 758 * by itself. 759 * (jinmei@kame.net 20010129) 760 */ 761 prelist_remove(pr); 762 } 763 } 764 765 /* cancel default outgoing interface setting */ 766 if (nd6_defifindex == ifp->if_index) 767 nd6_setdefaultiface(0); 768 769 /* XXX: too restrictive? */ 770 if (!ip6_forwarding && (ndi->flags & ND6_IFF_ACCEPT_RTADV)) { 771 /* refresh default router list */ 772 defrouter_select(); 773 } 774 775 /* 776 * Nuke neighbor cache entries for the ifp. 777 * Note that rt->rt_ifp may not be the same as ifp, 778 * due to KAME goto ours hack. See RTM_RESOLVE case in 779 * nd6_rtrequest(), and ip6_input(). 780 */ 781 ln = llinfo_nd6.ln_next; 782 while (ln != NULL && ln != &llinfo_nd6) { 783 struct rtentry *rt; 784 const struct sockaddr_dl *sdl; 785 786 nln = ln->ln_next; 787 rt = ln->ln_rt; 788 if (rt && rt->rt_gateway && 789 rt->rt_gateway->sa_family == AF_LINK) { 790 sdl = satocsdl(rt->rt_gateway); 791 if (sdl->sdl_index == ifp->if_index) 792 nln = nd6_free(rt, 0); 793 } 794 ln = nln; 795 } 796 } 797 798 struct rtentry * 799 nd6_lookup(const struct in6_addr *addr6, int create, struct ifnet *ifp) 800 { 801 struct rtentry *rt; 802 struct sockaddr_in6 sin6; 803 804 sockaddr_in6_init(&sin6, addr6, 0, 0, 0); 805 rt = rtalloc1((struct sockaddr *)&sin6, create); 806 if (rt != NULL && (rt->rt_flags & RTF_LLINFO) == 0) { 807 /* 808 * This is the case for the default route. 809 * If we want to create a neighbor cache for the address, we 810 * should free the route for the destination and allocate an 811 * interface route. 812 */ 813 if (create) { 814 RTFREE(rt); 815 rt = NULL; 816 } 817 } 818 if (rt != NULL) 819 ; 820 else if (create && ifp) { 821 int e; 822 823 /* 824 * If no route is available and create is set, 825 * we allocate a host route for the destination 826 * and treat it like an interface route. 827 * This hack is necessary for a neighbor which can't 828 * be covered by our own prefix. 829 */ 830 struct ifaddr *ifa = 831 ifaof_ifpforaddr((struct sockaddr *)&sin6, ifp); 832 if (ifa == NULL) 833 return NULL; 834 835 /* 836 * Create a new route. RTF_LLINFO is necessary 837 * to create a Neighbor Cache entry for the 838 * destination in nd6_rtrequest which will be 839 * called in rtrequest via ifa->ifa_rtrequest. 840 */ 841 if ((e = rtrequest(RTM_ADD, (const struct sockaddr *)&sin6, 842 ifa->ifa_addr, (const struct sockaddr *)&all1_sa, 843 (ifa->ifa_flags | RTF_HOST | RTF_LLINFO) & 844 ~RTF_CLONING, &rt)) != 0) { 845 #if 0 846 log(LOG_ERR, 847 "nd6_lookup: failed to add route for a " 848 "neighbor(%s), errno=%d\n", 849 ip6_sprintf(addr6), e); 850 #endif 851 return NULL; 852 } 853 if (rt == NULL) 854 return NULL; 855 if (rt->rt_llinfo) { 856 struct llinfo_nd6 *ln = 857 (struct llinfo_nd6 *)rt->rt_llinfo; 858 ln->ln_state = ND6_LLINFO_NOSTATE; 859 } 860 } else 861 return NULL; 862 rt->rt_refcnt--; 863 /* 864 * Validation for the entry. 865 * Note that the check for rt_llinfo is necessary because a cloned 866 * route from a parent route that has the L flag (e.g. the default 867 * route to a p2p interface) may have the flag, too, while the 868 * destination is not actually a neighbor. 869 * XXX: we can't use rt->rt_ifp to check for the interface, since 870 * it might be the loopback interface if the entry is for our 871 * own address on a non-loopback interface. Instead, we should 872 * use rt->rt_ifa->ifa_ifp, which would specify the REAL 873 * interface. 874 * Note also that ifa_ifp and ifp may differ when we connect two 875 * interfaces to a same link, install a link prefix to an interface, 876 * and try to install a neighbor cache on an interface that does not 877 * have a route to the prefix. 878 */ 879 if ((rt->rt_flags & RTF_GATEWAY) || (rt->rt_flags & RTF_LLINFO) == 0 || 880 rt->rt_gateway->sa_family != AF_LINK || rt->rt_llinfo == NULL || 881 (ifp && rt->rt_ifa->ifa_ifp != ifp)) { 882 if (create) { 883 nd6log((LOG_DEBUG, 884 "nd6_lookup: failed to lookup %s (if = %s)\n", 885 ip6_sprintf(addr6), 886 ifp ? if_name(ifp) : "unspec")); 887 } 888 return NULL; 889 } 890 return rt; 891 } 892 893 /* 894 * Detect if a given IPv6 address identifies a neighbor on a given link. 895 * XXX: should take care of the destination of a p2p link? 896 */ 897 int 898 nd6_is_addr_neighbor(const struct sockaddr_in6 *addr, struct ifnet *ifp) 899 { 900 struct nd_prefix *pr; 901 902 /* 903 * A link-local address is always a neighbor. 904 * XXX: a link does not necessarily specify a single interface. 905 */ 906 if (IN6_IS_ADDR_LINKLOCAL(&addr->sin6_addr)) { 907 struct sockaddr_in6 sin6_copy; 908 u_int32_t zone; 909 910 /* 911 * We need sin6_copy since sa6_recoverscope() may modify the 912 * content (XXX). 913 */ 914 sin6_copy = *addr; 915 if (sa6_recoverscope(&sin6_copy)) 916 return 0; /* XXX: should be impossible */ 917 if (in6_setscope(&sin6_copy.sin6_addr, ifp, &zone)) 918 return 0; 919 if (sin6_copy.sin6_scope_id == zone) 920 return 1; 921 else 922 return 0; 923 } 924 925 /* 926 * If the address matches one of our on-link prefixes, it should be a 927 * neighbor. 928 */ 929 LIST_FOREACH(pr, &nd_prefix, ndpr_entry) { 930 if (pr->ndpr_ifp != ifp) 931 continue; 932 933 if (!(pr->ndpr_stateflags & NDPRF_ONLINK)) 934 continue; 935 936 if (IN6_ARE_MASKED_ADDR_EQUAL(&pr->ndpr_prefix.sin6_addr, 937 &addr->sin6_addr, &pr->ndpr_mask)) 938 return 1; 939 } 940 941 /* 942 * If the default router list is empty, all addresses are regarded 943 * as on-link, and thus, as a neighbor. 944 * XXX: we restrict the condition to hosts, because routers usually do 945 * not have the "default router list". 946 */ 947 if (!ip6_forwarding && TAILQ_FIRST(&nd_defrouter) == NULL && 948 nd6_defifindex == ifp->if_index) { 949 return 1; 950 } 951 952 /* 953 * Even if the address matches none of our addresses, it might be 954 * in the neighbor cache. 955 */ 956 if (nd6_lookup(&addr->sin6_addr, 0, ifp) != NULL) 957 return 1; 958 959 return 0; 960 } 961 962 /* 963 * Free an nd6 llinfo entry. 964 * Since the function would cause significant changes in the kernel, DO NOT 965 * make it global, unless you have a strong reason for the change, and are sure 966 * that the change is safe. 967 */ 968 static struct llinfo_nd6 * 969 nd6_free(struct rtentry *rt, int gc) 970 { 971 struct llinfo_nd6 *ln = (struct llinfo_nd6 *)rt->rt_llinfo, *next; 972 struct in6_addr in6 = satocsin6(rt_getkey(rt))->sin6_addr; 973 struct nd_defrouter *dr; 974 975 /* 976 * we used to have pfctlinput(PRC_HOSTDEAD) here. 977 * even though it is not harmful, it was not really necessary. 978 */ 979 980 /* cancel timer */ 981 nd6_llinfo_settimer(ln, -1); 982 983 if (!ip6_forwarding) { 984 int s; 985 s = splsoftnet(); 986 dr = defrouter_lookup(&satocsin6(rt_getkey(rt))->sin6_addr, 987 rt->rt_ifp); 988 989 if (dr != NULL && dr->expire && 990 ln->ln_state == ND6_LLINFO_STALE && gc) { 991 /* 992 * If the reason for the deletion is just garbage 993 * collection, and the neighbor is an active default 994 * router, do not delete it. Instead, reset the GC 995 * timer using the router's lifetime. 996 * Simply deleting the entry would affect default 997 * router selection, which is not necessarily a good 998 * thing, especially when we're using router preference 999 * values. 1000 * XXX: the check for ln_state would be redundant, 1001 * but we intentionally keep it just in case. 1002 */ 1003 if (dr->expire > time_second) 1004 nd6_llinfo_settimer(ln, 1005 (dr->expire - time_second) * hz); 1006 else 1007 nd6_llinfo_settimer(ln, (long)nd6_gctimer * hz); 1008 splx(s); 1009 return ln->ln_next; 1010 } 1011 1012 if (ln->ln_router || dr) { 1013 /* 1014 * rt6_flush must be called whether or not the neighbor 1015 * is in the Default Router List. 1016 * See a corresponding comment in nd6_na_input(). 1017 */ 1018 rt6_flush(&in6, rt->rt_ifp); 1019 } 1020 1021 if (dr) { 1022 /* 1023 * Unreachablity of a router might affect the default 1024 * router selection and on-link detection of advertised 1025 * prefixes. 1026 */ 1027 1028 /* 1029 * Temporarily fake the state to choose a new default 1030 * router and to perform on-link determination of 1031 * prefixes correctly. 1032 * Below the state will be set correctly, 1033 * or the entry itself will be deleted. 1034 */ 1035 ln->ln_state = ND6_LLINFO_INCOMPLETE; 1036 1037 /* 1038 * Since defrouter_select() does not affect the 1039 * on-link determination and MIP6 needs the check 1040 * before the default router selection, we perform 1041 * the check now. 1042 */ 1043 pfxlist_onlink_check(); 1044 1045 /* 1046 * refresh default router list 1047 */ 1048 defrouter_select(); 1049 } 1050 splx(s); 1051 } 1052 1053 /* 1054 * Before deleting the entry, remember the next entry as the 1055 * return value. We need this because pfxlist_onlink_check() above 1056 * might have freed other entries (particularly the old next entry) as 1057 * a side effect (XXX). 1058 */ 1059 next = ln->ln_next; 1060 1061 /* 1062 * Detach the route from the routing tree and the list of neighbor 1063 * caches, and disable the route entry not to be used in already 1064 * cached routes. 1065 */ 1066 rtrequest(RTM_DELETE, rt_getkey(rt), NULL, rt_mask(rt), 0, NULL); 1067 1068 return next; 1069 } 1070 1071 /* 1072 * Upper-layer reachability hint for Neighbor Unreachability Detection. 1073 * 1074 * XXX cost-effective methods? 1075 */ 1076 void 1077 nd6_nud_hint(struct rtentry *rt, struct in6_addr *dst6, int force) 1078 { 1079 struct llinfo_nd6 *ln; 1080 1081 /* 1082 * If the caller specified "rt", use that. Otherwise, resolve the 1083 * routing table by supplied "dst6". 1084 */ 1085 if (rt == NULL) { 1086 if (dst6 == NULL) 1087 return; 1088 if ((rt = nd6_lookup(dst6, 0, NULL)) == NULL) 1089 return; 1090 } 1091 1092 if ((rt->rt_flags & RTF_GATEWAY) != 0 || 1093 (rt->rt_flags & RTF_LLINFO) == 0 || 1094 !rt->rt_llinfo || !rt->rt_gateway || 1095 rt->rt_gateway->sa_family != AF_LINK) { 1096 /* This is not a host route. */ 1097 return; 1098 } 1099 1100 ln = (struct llinfo_nd6 *)rt->rt_llinfo; 1101 if (ln->ln_state < ND6_LLINFO_REACHABLE) 1102 return; 1103 1104 /* 1105 * if we get upper-layer reachability confirmation many times, 1106 * it is possible we have false information. 1107 */ 1108 if (!force) { 1109 ln->ln_byhint++; 1110 if (ln->ln_byhint > nd6_maxnudhint) 1111 return; 1112 } 1113 1114 ln->ln_state = ND6_LLINFO_REACHABLE; 1115 if (!ND6_LLINFO_PERMANENT(ln)) { 1116 nd6_llinfo_settimer(ln, 1117 (long)ND_IFINFO(rt->rt_ifp)->reachable * hz); 1118 } 1119 } 1120 1121 void 1122 nd6_rtrequest(int req, struct rtentry *rt, const struct rt_addrinfo *info) 1123 { 1124 struct sockaddr *gate = rt->rt_gateway; 1125 struct llinfo_nd6 *ln = (struct llinfo_nd6 *)rt->rt_llinfo; 1126 struct ifnet *ifp = rt->rt_ifp; 1127 uint8_t namelen = strlen(ifp->if_xname), addrlen = ifp->if_addrlen; 1128 struct ifaddr *ifa; 1129 1130 RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key); 1131 1132 if (req == RTM_LLINFO_UPD) { 1133 int rc; 1134 struct in6_addr *in6; 1135 struct in6_addr in6_all; 1136 int anycast; 1137 1138 if ((ifa = info->rti_ifa) == NULL) 1139 return; 1140 1141 in6 = &ifatoia6(ifa)->ia_addr.sin6_addr; 1142 anycast = ifatoia6(ifa)->ia6_flags & IN6_IFF_ANYCAST; 1143 1144 in6_all = in6addr_linklocal_allnodes; 1145 if ((rc = in6_setscope(&in6_all, ifa->ifa_ifp, NULL)) != 0) { 1146 log(LOG_ERR, "%s: failed to set scope %s " 1147 "(errno=%d)\n", __func__, if_name(ifp), rc); 1148 return; 1149 } 1150 1151 /* XXX don't set Override for proxy addresses */ 1152 nd6_na_output(ifa->ifa_ifp, &in6_all, in6, 1153 (anycast ? 0 : ND_NA_FLAG_OVERRIDE) 1154 #if 0 1155 | (ip6_forwarding ? ND_NA_FLAG_ROUTER : 0) 1156 #endif 1157 , 1, NULL); 1158 return; 1159 } 1160 1161 if ((rt->rt_flags & RTF_GATEWAY) != 0) 1162 return; 1163 1164 if (nd6_need_cache(ifp) == 0 && (rt->rt_flags & RTF_HOST) == 0) { 1165 RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key); 1166 /* 1167 * This is probably an interface direct route for a link 1168 * which does not need neighbor caches (e.g. fe80::%lo0/64). 1169 * We do not need special treatment below for such a route. 1170 * Moreover, the RTF_LLINFO flag which would be set below 1171 * would annoy the ndp(8) command. 1172 */ 1173 return; 1174 } 1175 1176 if (req == RTM_RESOLVE && 1177 (nd6_need_cache(ifp) == 0 || /* stf case */ 1178 !nd6_is_addr_neighbor(satocsin6(rt_getkey(rt)), ifp))) { 1179 RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key); 1180 /* 1181 * FreeBSD and BSD/OS often make a cloned host route based 1182 * on a less-specific route (e.g. the default route). 1183 * If the less specific route does not have a "gateway" 1184 * (this is the case when the route just goes to a p2p or an 1185 * stf interface), we'll mistakenly make a neighbor cache for 1186 * the host route, and will see strange neighbor solicitation 1187 * for the corresponding destination. In order to avoid the 1188 * confusion, we check if the destination of the route is 1189 * a neighbor in terms of neighbor discovery, and stop the 1190 * process if not. Additionally, we remove the LLINFO flag 1191 * so that ndp(8) will not try to get the neighbor information 1192 * of the destination. 1193 */ 1194 rt->rt_flags &= ~RTF_LLINFO; 1195 return; 1196 } 1197 1198 switch (req) { 1199 case RTM_ADD: 1200 RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key); 1201 /* 1202 * There is no backward compatibility :) 1203 * 1204 * if ((rt->rt_flags & RTF_HOST) == 0 && 1205 * SIN(rt_mask(rt))->sin_addr.s_addr != 0xffffffff) 1206 * rt->rt_flags |= RTF_CLONING; 1207 */ 1208 if ((rt->rt_flags & RTF_CLONING) || 1209 ((rt->rt_flags & RTF_LLINFO) && ln == NULL)) { 1210 union { 1211 struct sockaddr sa; 1212 struct sockaddr_dl sdl; 1213 struct sockaddr_storage ss; 1214 } u; 1215 /* 1216 * Case 1: This route should come from a route to 1217 * interface (RTF_CLONING case) or the route should be 1218 * treated as on-link but is currently not 1219 * (RTF_LLINFO && ln == NULL case). 1220 */ 1221 sockaddr_dl_init(&u.sdl, sizeof(u.ss), 1222 ifp->if_index, ifp->if_type, 1223 NULL, namelen, NULL, addrlen); 1224 rt_setgate(rt, &u.sa); 1225 gate = rt->rt_gateway; 1226 RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key); 1227 if (ln != NULL) 1228 nd6_llinfo_settimer(ln, 0); 1229 RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key); 1230 if ((rt->rt_flags & RTF_CLONING) != 0) 1231 break; 1232 } 1233 RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key); 1234 /* 1235 * In IPv4 code, we try to annonuce new RTF_ANNOUNCE entry here. 1236 * We don't do that here since llinfo is not ready yet. 1237 * 1238 * There are also couple of other things to be discussed: 1239 * - unsolicited NA code needs improvement beforehand 1240 * - RFC2461 says we MAY send multicast unsolicited NA 1241 * (7.2.6 paragraph 4), however, it also says that we 1242 * SHOULD provide a mechanism to prevent multicast NA storm. 1243 * we don't have anything like it right now. 1244 * note that the mechanism needs a mutual agreement 1245 * between proxies, which means that we need to implement 1246 * a new protocol, or a new kludge. 1247 * - from RFC2461 6.2.4, host MUST NOT send an unsolicited NA. 1248 * we need to check ip6forwarding before sending it. 1249 * (or should we allow proxy ND configuration only for 1250 * routers? there's no mention about proxy ND from hosts) 1251 */ 1252 #if 0 1253 /* XXX it does not work */ 1254 if (rt->rt_flags & RTF_ANNOUNCE) 1255 nd6_na_output(ifp, 1256 &satocsin6(rt_getkey(rt))->sin6_addr, 1257 &satocsin6(rt_getkey(rt))->sin6_addr, 1258 ip6_forwarding ? ND_NA_FLAG_ROUTER : 0, 1259 1, NULL); 1260 #endif 1261 /* FALLTHROUGH */ 1262 case RTM_RESOLVE: 1263 if ((ifp->if_flags & (IFF_POINTOPOINT | IFF_LOOPBACK)) == 0) { 1264 RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key); 1265 /* 1266 * Address resolution isn't necessary for a point to 1267 * point link, so we can skip this test for a p2p link. 1268 */ 1269 if (gate->sa_family != AF_LINK || 1270 gate->sa_len < 1271 sockaddr_dl_measure(namelen, addrlen)) { 1272 log(LOG_DEBUG, 1273 "nd6_rtrequest: bad gateway value: %s\n", 1274 if_name(ifp)); 1275 break; 1276 } 1277 satosdl(gate)->sdl_type = ifp->if_type; 1278 satosdl(gate)->sdl_index = ifp->if_index; 1279 RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key); 1280 } 1281 if (ln != NULL) 1282 break; /* This happens on a route change */ 1283 RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key); 1284 /* 1285 * Case 2: This route may come from cloning, or a manual route 1286 * add with a LL address. 1287 */ 1288 R_Malloc(ln, struct llinfo_nd6 *, sizeof(*ln)); 1289 rt->rt_llinfo = ln; 1290 RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key); 1291 if (ln == NULL) { 1292 log(LOG_DEBUG, "nd6_rtrequest: malloc failed\n"); 1293 break; 1294 } 1295 RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key); 1296 nd6_inuse++; 1297 nd6_allocated++; 1298 memset(ln, 0, sizeof(*ln)); 1299 ln->ln_rt = rt; 1300 callout_init(&ln->ln_timer_ch, CALLOUT_MPSAFE); 1301 /* this is required for "ndp" command. - shin */ 1302 if (req == RTM_ADD) { 1303 /* 1304 * gate should have some valid AF_LINK entry, 1305 * and ln->ln_expire should have some lifetime 1306 * which is specified by ndp command. 1307 */ 1308 ln->ln_state = ND6_LLINFO_REACHABLE; 1309 ln->ln_byhint = 0; 1310 } else { 1311 /* 1312 * When req == RTM_RESOLVE, rt is created and 1313 * initialized in rtrequest(), so rt_expire is 0. 1314 */ 1315 ln->ln_state = ND6_LLINFO_NOSTATE; 1316 nd6_llinfo_settimer(ln, 0); 1317 } 1318 RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key); 1319 rt->rt_flags |= RTF_LLINFO; 1320 ln->ln_next = llinfo_nd6.ln_next; 1321 llinfo_nd6.ln_next = ln; 1322 ln->ln_prev = &llinfo_nd6; 1323 ln->ln_next->ln_prev = ln; 1324 1325 RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key); 1326 /* 1327 * check if rt_getkey(rt) is an address assigned 1328 * to the interface. 1329 */ 1330 ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp, 1331 &satocsin6(rt_getkey(rt))->sin6_addr); 1332 RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key); 1333 if (ifa != NULL) { 1334 const void *mac; 1335 nd6_llinfo_settimer(ln, -1); 1336 ln->ln_state = ND6_LLINFO_REACHABLE; 1337 ln->ln_byhint = 0; 1338 if ((mac = nd6_ifptomac(ifp)) != NULL) { 1339 /* XXX check for error */ 1340 (void)sockaddr_dl_setaddr(satosdl(gate), 1341 gate->sa_len, mac, ifp->if_addrlen); 1342 } 1343 if (nd6_useloopback) { 1344 ifp = rt->rt_ifp = lo0ifp; /* XXX */ 1345 /* 1346 * Make sure rt_ifa be equal to the ifaddr 1347 * corresponding to the address. 1348 * We need this because when we refer 1349 * rt_ifa->ia6_flags in ip6_input, we assume 1350 * that the rt_ifa points to the address instead 1351 * of the loopback address. 1352 */ 1353 if (ifa != rt->rt_ifa) 1354 rt_replace_ifa(rt, ifa); 1355 rt->rt_flags &= ~RTF_CLONED; 1356 } 1357 } else if (rt->rt_flags & RTF_ANNOUNCE) { 1358 nd6_llinfo_settimer(ln, -1); 1359 ln->ln_state = ND6_LLINFO_REACHABLE; 1360 ln->ln_byhint = 0; 1361 1362 /* join solicited node multicast for proxy ND */ 1363 if (ifp->if_flags & IFF_MULTICAST) { 1364 struct in6_addr llsol; 1365 int error; 1366 1367 llsol = satocsin6(rt_getkey(rt))->sin6_addr; 1368 llsol.s6_addr32[0] = htonl(0xff020000); 1369 llsol.s6_addr32[1] = 0; 1370 llsol.s6_addr32[2] = htonl(1); 1371 llsol.s6_addr8[12] = 0xff; 1372 if (in6_setscope(&llsol, ifp, NULL)) 1373 break; 1374 if (!in6_addmulti(&llsol, ifp, &error, 0)) { 1375 nd6log((LOG_ERR, "%s: failed to join " 1376 "%s (errno=%d)\n", if_name(ifp), 1377 ip6_sprintf(&llsol), error)); 1378 } 1379 } 1380 } 1381 break; 1382 1383 case RTM_DELETE: 1384 if (ln == NULL) 1385 break; 1386 /* leave from solicited node multicast for proxy ND */ 1387 if ((rt->rt_flags & RTF_ANNOUNCE) != 0 && 1388 (ifp->if_flags & IFF_MULTICAST) != 0) { 1389 struct in6_addr llsol; 1390 struct in6_multi *in6m; 1391 1392 llsol = satocsin6(rt_getkey(rt))->sin6_addr; 1393 llsol.s6_addr32[0] = htonl(0xff020000); 1394 llsol.s6_addr32[1] = 0; 1395 llsol.s6_addr32[2] = htonl(1); 1396 llsol.s6_addr8[12] = 0xff; 1397 if (in6_setscope(&llsol, ifp, NULL) == 0) { 1398 IN6_LOOKUP_MULTI(llsol, ifp, in6m); 1399 if (in6m) 1400 in6_delmulti(in6m); 1401 } 1402 } 1403 nd6_inuse--; 1404 ln->ln_next->ln_prev = ln->ln_prev; 1405 ln->ln_prev->ln_next = ln->ln_next; 1406 ln->ln_prev = NULL; 1407 nd6_llinfo_settimer(ln, -1); 1408 rt->rt_llinfo = 0; 1409 rt->rt_flags &= ~RTF_LLINFO; 1410 clear_llinfo_pqueue(ln); 1411 Free(ln); 1412 } 1413 } 1414 1415 int 1416 nd6_ioctl(u_long cmd, void *data, struct ifnet *ifp) 1417 { 1418 struct in6_drlist *drl = (struct in6_drlist *)data; 1419 struct in6_oprlist *oprl = (struct in6_oprlist *)data; 1420 struct in6_ndireq *ndi = (struct in6_ndireq *)data; 1421 struct in6_nbrinfo *nbi = (struct in6_nbrinfo *)data; 1422 struct in6_ndifreq *ndif = (struct in6_ndifreq *)data; 1423 struct nd_defrouter *dr; 1424 struct nd_prefix *pr; 1425 struct rtentry *rt; 1426 int i = 0, error = 0; 1427 int s; 1428 1429 switch (cmd) { 1430 case SIOCGDRLST_IN6: 1431 /* 1432 * obsolete API, use sysctl under net.inet6.icmp6 1433 */ 1434 memset(drl, 0, sizeof(*drl)); 1435 s = splsoftnet(); 1436 TAILQ_FOREACH(dr, &nd_defrouter, dr_entry) { 1437 if (i >= DRLSTSIZ) 1438 break; 1439 drl->defrouter[i].rtaddr = dr->rtaddr; 1440 in6_clearscope(&drl->defrouter[i].rtaddr); 1441 1442 drl->defrouter[i].flags = dr->flags; 1443 drl->defrouter[i].rtlifetime = dr->rtlifetime; 1444 drl->defrouter[i].expire = dr->expire; 1445 drl->defrouter[i].if_index = dr->ifp->if_index; 1446 i++; 1447 } 1448 splx(s); 1449 break; 1450 case SIOCGPRLST_IN6: 1451 /* 1452 * obsolete API, use sysctl under net.inet6.icmp6 1453 * 1454 * XXX the structure in6_prlist was changed in backward- 1455 * incompatible manner. in6_oprlist is used for SIOCGPRLST_IN6, 1456 * in6_prlist is used for nd6_sysctl() - fill_prlist(). 1457 */ 1458 /* 1459 * XXX meaning of fields, especialy "raflags", is very 1460 * differnet between RA prefix list and RR/static prefix list. 1461 * how about separating ioctls into two? 1462 */ 1463 memset(oprl, 0, sizeof(*oprl)); 1464 s = splsoftnet(); 1465 LIST_FOREACH(pr, &nd_prefix, ndpr_entry) { 1466 struct nd_pfxrouter *pfr; 1467 int j; 1468 1469 if (i >= PRLSTSIZ) 1470 break; 1471 oprl->prefix[i].prefix = pr->ndpr_prefix.sin6_addr; 1472 oprl->prefix[i].raflags = pr->ndpr_raf; 1473 oprl->prefix[i].prefixlen = pr->ndpr_plen; 1474 oprl->prefix[i].vltime = pr->ndpr_vltime; 1475 oprl->prefix[i].pltime = pr->ndpr_pltime; 1476 oprl->prefix[i].if_index = pr->ndpr_ifp->if_index; 1477 if (pr->ndpr_vltime == ND6_INFINITE_LIFETIME) 1478 oprl->prefix[i].expire = 0; 1479 else { 1480 time_t maxexpire; 1481 1482 /* XXX: we assume time_t is signed. */ 1483 maxexpire = (-1) & 1484 ~((time_t)1 << 1485 ((sizeof(maxexpire) * 8) - 1)); 1486 if (pr->ndpr_vltime < 1487 maxexpire - pr->ndpr_lastupdate) { 1488 oprl->prefix[i].expire = 1489 pr->ndpr_lastupdate + 1490 pr->ndpr_vltime; 1491 } else 1492 oprl->prefix[i].expire = maxexpire; 1493 } 1494 1495 j = 0; 1496 LIST_FOREACH(pfr, &pr->ndpr_advrtrs, pfr_entry) { 1497 if (j < DRLSTSIZ) { 1498 #define RTRADDR oprl->prefix[i].advrtr[j] 1499 RTRADDR = pfr->router->rtaddr; 1500 in6_clearscope(&RTRADDR); 1501 #undef RTRADDR 1502 } 1503 j++; 1504 } 1505 oprl->prefix[i].advrtrs = j; 1506 oprl->prefix[i].origin = PR_ORIG_RA; 1507 1508 i++; 1509 } 1510 splx(s); 1511 1512 break; 1513 case OSIOCGIFINFO_IN6: 1514 #define ND ndi->ndi 1515 /* XXX: old ndp(8) assumes a positive value for linkmtu. */ 1516 memset(&ND, 0, sizeof(ND)); 1517 ND.linkmtu = IN6_LINKMTU(ifp); 1518 ND.maxmtu = ND_IFINFO(ifp)->maxmtu; 1519 ND.basereachable = ND_IFINFO(ifp)->basereachable; 1520 ND.reachable = ND_IFINFO(ifp)->reachable; 1521 ND.retrans = ND_IFINFO(ifp)->retrans; 1522 ND.flags = ND_IFINFO(ifp)->flags; 1523 ND.recalctm = ND_IFINFO(ifp)->recalctm; 1524 ND.chlim = ND_IFINFO(ifp)->chlim; 1525 break; 1526 case SIOCGIFINFO_IN6: 1527 ND = *ND_IFINFO(ifp); 1528 break; 1529 case SIOCSIFINFO_IN6: 1530 /* 1531 * used to change host variables from userland. 1532 * intented for a use on router to reflect RA configurations. 1533 */ 1534 /* 0 means 'unspecified' */ 1535 if (ND.linkmtu != 0) { 1536 if (ND.linkmtu < IPV6_MMTU || 1537 ND.linkmtu > IN6_LINKMTU(ifp)) { 1538 error = EINVAL; 1539 break; 1540 } 1541 ND_IFINFO(ifp)->linkmtu = ND.linkmtu; 1542 } 1543 1544 if (ND.basereachable != 0) { 1545 int obasereachable = ND_IFINFO(ifp)->basereachable; 1546 1547 ND_IFINFO(ifp)->basereachable = ND.basereachable; 1548 if (ND.basereachable != obasereachable) 1549 ND_IFINFO(ifp)->reachable = 1550 ND_COMPUTE_RTIME(ND.basereachable); 1551 } 1552 if (ND.retrans != 0) 1553 ND_IFINFO(ifp)->retrans = ND.retrans; 1554 if (ND.chlim != 0) 1555 ND_IFINFO(ifp)->chlim = ND.chlim; 1556 /* FALLTHROUGH */ 1557 case SIOCSIFINFO_FLAGS: 1558 ND_IFINFO(ifp)->flags = ND.flags; 1559 break; 1560 #undef ND 1561 case SIOCSNDFLUSH_IN6: /* XXX: the ioctl name is confusing... */ 1562 /* sync kernel routing table with the default router list */ 1563 defrouter_reset(); 1564 defrouter_select(); 1565 break; 1566 case SIOCSPFXFLUSH_IN6: 1567 { 1568 /* flush all the prefix advertised by routers */ 1569 struct nd_prefix *pfx, *next; 1570 1571 s = splsoftnet(); 1572 for (pfx = LIST_FIRST(&nd_prefix); pfx; pfx = next) { 1573 struct in6_ifaddr *ia, *ia_next; 1574 1575 next = LIST_NEXT(pfx, ndpr_entry); 1576 1577 if (IN6_IS_ADDR_LINKLOCAL(&pfx->ndpr_prefix.sin6_addr)) 1578 continue; /* XXX */ 1579 1580 /* do we really have to remove addresses as well? */ 1581 for (ia = in6_ifaddr; ia; ia = ia_next) { 1582 /* ia might be removed. keep the next ptr. */ 1583 ia_next = ia->ia_next; 1584 1585 if ((ia->ia6_flags & IN6_IFF_AUTOCONF) == 0) 1586 continue; 1587 1588 if (ia->ia6_ndpr == pfx) 1589 in6_purgeaddr(&ia->ia_ifa); 1590 } 1591 prelist_remove(pfx); 1592 } 1593 splx(s); 1594 break; 1595 } 1596 case SIOCSRTRFLUSH_IN6: 1597 { 1598 /* flush all the default routers */ 1599 struct nd_defrouter *drtr, *next; 1600 1601 s = splsoftnet(); 1602 defrouter_reset(); 1603 for (drtr = TAILQ_FIRST(&nd_defrouter); drtr; drtr = next) { 1604 next = TAILQ_NEXT(drtr, dr_entry); 1605 defrtrlist_del(drtr); 1606 } 1607 defrouter_select(); 1608 splx(s); 1609 break; 1610 } 1611 case SIOCGNBRINFO_IN6: 1612 { 1613 struct llinfo_nd6 *ln; 1614 struct in6_addr nb_addr = nbi->addr; /* make local for safety */ 1615 1616 if ((error = in6_setscope(&nb_addr, ifp, NULL)) != 0) 1617 return error; 1618 1619 s = splsoftnet(); 1620 if ((rt = nd6_lookup(&nb_addr, 0, ifp)) == NULL || 1621 (ln = (struct llinfo_nd6 *)rt->rt_llinfo) == NULL) { 1622 error = EINVAL; 1623 splx(s); 1624 break; 1625 } 1626 nbi->state = ln->ln_state; 1627 nbi->asked = ln->ln_asked; 1628 nbi->isrouter = ln->ln_router; 1629 nbi->expire = ln->ln_expire; 1630 splx(s); 1631 1632 break; 1633 } 1634 case SIOCGDEFIFACE_IN6: /* XXX: should be implemented as a sysctl? */ 1635 ndif->ifindex = nd6_defifindex; 1636 break; 1637 case SIOCSDEFIFACE_IN6: /* XXX: should be implemented as a sysctl? */ 1638 return nd6_setdefaultiface(ndif->ifindex); 1639 } 1640 return error; 1641 } 1642 1643 void 1644 nd6_llinfo_release_pkts(struct llinfo_nd6 *ln, struct ifnet *ifp, 1645 struct rtentry *rt) 1646 { 1647 struct mbuf *m_hold, *m_hold_next; 1648 1649 for (m_hold = ln->ln_hold, ln->ln_hold = NULL; 1650 m_hold != NULL; 1651 m_hold = m_hold_next) { 1652 m_hold_next = m_hold->m_nextpkt; 1653 m_hold->m_nextpkt = NULL; 1654 1655 /* 1656 * we assume ifp is not a p2p here, so 1657 * just set the 2nd argument as the 1658 * 1st one. 1659 */ 1660 nd6_output(ifp, ifp, m_hold, satocsin6(rt_getkey(rt)), rt); 1661 } 1662 } 1663 1664 /* 1665 * Create neighbor cache entry and cache link-layer address, 1666 * on reception of inbound ND6 packets. (RS/RA/NS/redirect) 1667 */ 1668 struct rtentry * 1669 nd6_cache_lladdr( 1670 struct ifnet *ifp, 1671 struct in6_addr *from, 1672 char *lladdr, 1673 int lladdrlen, 1674 int type, /* ICMP6 type */ 1675 int code /* type dependent information */ 1676 ) 1677 { 1678 struct nd_ifinfo *ndi = ND_IFINFO(ifp); 1679 struct rtentry *rt = NULL; 1680 struct llinfo_nd6 *ln = NULL; 1681 int is_newentry; 1682 struct sockaddr_dl *sdl = NULL; 1683 int do_update; 1684 int olladdr; 1685 int llchange; 1686 int newstate = 0; 1687 1688 if (ifp == NULL) 1689 panic("ifp == NULL in nd6_cache_lladdr"); 1690 if (from == NULL) 1691 panic("from == NULL in nd6_cache_lladdr"); 1692 1693 /* nothing must be updated for unspecified address */ 1694 if (IN6_IS_ADDR_UNSPECIFIED(from)) 1695 return NULL; 1696 1697 /* 1698 * Validation about ifp->if_addrlen and lladdrlen must be done in 1699 * the caller. 1700 * 1701 * XXX If the link does not have link-layer adderss, what should 1702 * we do? (ifp->if_addrlen == 0) 1703 * Spec says nothing in sections for RA, RS and NA. There's small 1704 * description on it in NS section (RFC 2461 7.2.3). 1705 */ 1706 1707 rt = nd6_lookup(from, 0, ifp); 1708 if (rt == NULL) { 1709 #if 0 1710 /* nothing must be done if there's no lladdr */ 1711 if (!lladdr || !lladdrlen) 1712 return NULL; 1713 #endif 1714 1715 rt = nd6_lookup(from, 1, ifp); 1716 is_newentry = 1; 1717 } else { 1718 /* do nothing if static ndp is set */ 1719 if (rt->rt_flags & RTF_STATIC) 1720 return NULL; 1721 is_newentry = 0; 1722 } 1723 1724 if (rt == NULL) 1725 return NULL; 1726 if ((rt->rt_flags & (RTF_GATEWAY | RTF_LLINFO)) != RTF_LLINFO) { 1727 fail: 1728 (void)nd6_free(rt, 0); 1729 return NULL; 1730 } 1731 ln = (struct llinfo_nd6 *)rt->rt_llinfo; 1732 if (ln == NULL) 1733 goto fail; 1734 if (rt->rt_gateway == NULL) 1735 goto fail; 1736 if (rt->rt_gateway->sa_family != AF_LINK) 1737 goto fail; 1738 sdl = satosdl(rt->rt_gateway); 1739 1740 olladdr = (sdl->sdl_alen) ? 1 : 0; 1741 if (olladdr && lladdr) { 1742 if (memcmp(lladdr, CLLADDR(sdl), ifp->if_addrlen)) 1743 llchange = 1; 1744 else 1745 llchange = 0; 1746 } else 1747 llchange = 0; 1748 1749 /* 1750 * newentry olladdr lladdr llchange (*=record) 1751 * 0 n n -- (1) 1752 * 0 y n -- (2) 1753 * 0 n y -- (3) * STALE 1754 * 0 y y n (4) * 1755 * 0 y y y (5) * STALE 1756 * 1 -- n -- (6) NOSTATE(= PASSIVE) 1757 * 1 -- y -- (7) * STALE 1758 */ 1759 1760 if (lladdr) { /* (3-5) and (7) */ 1761 /* 1762 * Record source link-layer address 1763 * XXX is it dependent to ifp->if_type? 1764 */ 1765 /* XXX check for error */ 1766 (void)sockaddr_dl_setaddr(sdl, sdl->sdl_len, lladdr, 1767 ifp->if_addrlen); 1768 } 1769 1770 if (!is_newentry) { 1771 if ((!olladdr && lladdr) || /* (3) */ 1772 (olladdr && lladdr && llchange)) { /* (5) */ 1773 do_update = 1; 1774 newstate = ND6_LLINFO_STALE; 1775 } else /* (1-2,4) */ 1776 do_update = 0; 1777 } else { 1778 do_update = 1; 1779 if (lladdr == NULL) /* (6) */ 1780 newstate = ND6_LLINFO_NOSTATE; 1781 else /* (7) */ 1782 newstate = ND6_LLINFO_STALE; 1783 } 1784 1785 if (do_update) { 1786 /* 1787 * Update the state of the neighbor cache. 1788 */ 1789 ln->ln_state = newstate; 1790 1791 if (ln->ln_state == ND6_LLINFO_STALE) { 1792 /* 1793 * XXX: since nd6_output() below will cause 1794 * state tansition to DELAY and reset the timer, 1795 * we must set the timer now, although it is actually 1796 * meaningless. 1797 */ 1798 nd6_llinfo_settimer(ln, (long)nd6_gctimer * hz); 1799 1800 nd6_llinfo_release_pkts(ln, ifp, rt); 1801 } else if (ln->ln_state == ND6_LLINFO_INCOMPLETE) { 1802 /* probe right away */ 1803 nd6_llinfo_settimer((void *)ln, 0); 1804 } 1805 } 1806 1807 /* 1808 * ICMP6 type dependent behavior. 1809 * 1810 * NS: clear IsRouter if new entry 1811 * RS: clear IsRouter 1812 * RA: set IsRouter if there's lladdr 1813 * redir: clear IsRouter if new entry 1814 * 1815 * RA case, (1): 1816 * The spec says that we must set IsRouter in the following cases: 1817 * - If lladdr exist, set IsRouter. This means (1-5). 1818 * - If it is old entry (!newentry), set IsRouter. This means (7). 1819 * So, based on the spec, in (1-5) and (7) cases we must set IsRouter. 1820 * A quetion arises for (1) case. (1) case has no lladdr in the 1821 * neighbor cache, this is similar to (6). 1822 * This case is rare but we figured that we MUST NOT set IsRouter. 1823 * 1824 * newentry olladdr lladdr llchange NS RS RA redir 1825 * D R 1826 * 0 n n -- (1) c ? s 1827 * 0 y n -- (2) c s s 1828 * 0 n y -- (3) c s s 1829 * 0 y y n (4) c s s 1830 * 0 y y y (5) c s s 1831 * 1 -- n -- (6) c c c s 1832 * 1 -- y -- (7) c c s c s 1833 * 1834 * (c=clear s=set) 1835 */ 1836 switch (type & 0xff) { 1837 case ND_NEIGHBOR_SOLICIT: 1838 /* 1839 * New entry must have is_router flag cleared. 1840 */ 1841 if (is_newentry) /* (6-7) */ 1842 ln->ln_router = 0; 1843 break; 1844 case ND_REDIRECT: 1845 /* 1846 * If the icmp is a redirect to a better router, always set the 1847 * is_router flag. Otherwise, if the entry is newly created, 1848 * clear the flag. [RFC 2461, sec 8.3] 1849 */ 1850 if (code == ND_REDIRECT_ROUTER) 1851 ln->ln_router = 1; 1852 else if (is_newentry) /* (6-7) */ 1853 ln->ln_router = 0; 1854 break; 1855 case ND_ROUTER_SOLICIT: 1856 /* 1857 * is_router flag must always be cleared. 1858 */ 1859 ln->ln_router = 0; 1860 break; 1861 case ND_ROUTER_ADVERT: 1862 /* 1863 * Mark an entry with lladdr as a router. 1864 */ 1865 if ((!is_newentry && (olladdr || lladdr)) || /* (2-5) */ 1866 (is_newentry && lladdr)) { /* (7) */ 1867 ln->ln_router = 1; 1868 } 1869 break; 1870 } 1871 1872 /* 1873 * When the link-layer address of a router changes, select the 1874 * best router again. In particular, when the neighbor entry is newly 1875 * created, it might affect the selection policy. 1876 * Question: can we restrict the first condition to the "is_newentry" 1877 * case? 1878 * XXX: when we hear an RA from a new router with the link-layer 1879 * address option, defrouter_select() is called twice, since 1880 * defrtrlist_update called the function as well. However, I believe 1881 * we can compromise the overhead, since it only happens the first 1882 * time. 1883 * XXX: although defrouter_select() should not have a bad effect 1884 * for those are not autoconfigured hosts, we explicitly avoid such 1885 * cases for safety. 1886 */ 1887 if (do_update && ln->ln_router && !ip6_forwarding && 1888 (ndi->flags & ND6_IFF_ACCEPT_RTADV)) 1889 defrouter_select(); 1890 1891 return rt; 1892 } 1893 1894 static void 1895 nd6_slowtimo(void *ignored_arg) 1896 { 1897 struct nd_ifinfo *nd6if; 1898 struct ifnet *ifp; 1899 1900 mutex_enter(softnet_lock); 1901 KERNEL_LOCK(1, NULL); 1902 callout_reset(&nd6_slowtimo_ch, ND6_SLOWTIMER_INTERVAL * hz, 1903 nd6_slowtimo, NULL); 1904 TAILQ_FOREACH(ifp, &ifnet, if_list) { 1905 nd6if = ND_IFINFO(ifp); 1906 if (nd6if->basereachable && /* already initialized */ 1907 (nd6if->recalctm -= ND6_SLOWTIMER_INTERVAL) <= 0) { 1908 /* 1909 * Since reachable time rarely changes by router 1910 * advertisements, we SHOULD insure that a new random 1911 * value gets recomputed at least once every few hours. 1912 * (RFC 2461, 6.3.4) 1913 */ 1914 nd6if->recalctm = nd6_recalc_reachtm_interval; 1915 nd6if->reachable = ND_COMPUTE_RTIME(nd6if->basereachable); 1916 } 1917 } 1918 KERNEL_UNLOCK_ONE(NULL); 1919 mutex_exit(softnet_lock); 1920 } 1921 1922 #define senderr(e) { error = (e); goto bad;} 1923 int 1924 nd6_output(struct ifnet *ifp, struct ifnet *origifp, struct mbuf *m0, 1925 const struct sockaddr_in6 *dst, struct rtentry *rt0) 1926 { 1927 struct mbuf *m = m0; 1928 struct rtentry *rt = rt0; 1929 struct sockaddr_in6 *gw6 = NULL; 1930 struct llinfo_nd6 *ln = NULL; 1931 int error = 0; 1932 1933 if (IN6_IS_ADDR_MULTICAST(&dst->sin6_addr)) 1934 goto sendpkt; 1935 1936 if (nd6_need_cache(ifp) == 0) 1937 goto sendpkt; 1938 1939 /* 1940 * next hop determination. This routine is derived from ether_output. 1941 */ 1942 if (rt) { 1943 if ((rt->rt_flags & RTF_UP) == 0) { 1944 if ((rt0 = rt = rtalloc1(sin6tocsa(dst), 1)) != NULL) { 1945 rt->rt_refcnt--; 1946 if (rt->rt_ifp != ifp) 1947 senderr(EHOSTUNREACH); 1948 } else 1949 senderr(EHOSTUNREACH); 1950 } 1951 1952 if (rt->rt_flags & RTF_GATEWAY) { 1953 gw6 = (struct sockaddr_in6 *)rt->rt_gateway; 1954 1955 /* 1956 * We skip link-layer address resolution and NUD 1957 * if the gateway is not a neighbor from ND point 1958 * of view, regardless of the value of nd_ifinfo.flags. 1959 * The second condition is a bit tricky; we skip 1960 * if the gateway is our own address, which is 1961 * sometimes used to install a route to a p2p link. 1962 */ 1963 if (!nd6_is_addr_neighbor(gw6, ifp) || 1964 in6ifa_ifpwithaddr(ifp, &gw6->sin6_addr)) { 1965 /* 1966 * We allow this kind of tricky route only 1967 * when the outgoing interface is p2p. 1968 * XXX: we may need a more generic rule here. 1969 */ 1970 if ((ifp->if_flags & IFF_POINTOPOINT) == 0) 1971 senderr(EHOSTUNREACH); 1972 1973 goto sendpkt; 1974 } 1975 1976 if (rt->rt_gwroute == NULL) 1977 goto lookup; 1978 if (((rt = rt->rt_gwroute)->rt_flags & RTF_UP) == 0) { 1979 rtfree(rt); rt = rt0; 1980 lookup: 1981 rt->rt_gwroute = rtalloc1(rt->rt_gateway, 1); 1982 if ((rt = rt->rt_gwroute) == NULL) 1983 senderr(EHOSTUNREACH); 1984 /* the "G" test below also prevents rt == rt0 */ 1985 if ((rt->rt_flags & RTF_GATEWAY) || 1986 (rt->rt_ifp != ifp)) { 1987 rt->rt_refcnt--; 1988 rt0->rt_gwroute = NULL; 1989 senderr(EHOSTUNREACH); 1990 } 1991 } 1992 } 1993 } 1994 1995 /* 1996 * Address resolution or Neighbor Unreachability Detection 1997 * for the next hop. 1998 * At this point, the destination of the packet must be a unicast 1999 * or an anycast address(i.e. not a multicast). 2000 */ 2001 2002 /* Look up the neighbor cache for the nexthop */ 2003 if (rt != NULL && (rt->rt_flags & RTF_LLINFO) != 0) 2004 ln = (struct llinfo_nd6 *)rt->rt_llinfo; 2005 else { 2006 /* 2007 * Since nd6_is_addr_neighbor() internally calls nd6_lookup(), 2008 * the condition below is not very efficient. But we believe 2009 * it is tolerable, because this should be a rare case. 2010 */ 2011 if (nd6_is_addr_neighbor(dst, ifp) && 2012 (rt = nd6_lookup(&dst->sin6_addr, 1, ifp)) != NULL) 2013 ln = (struct llinfo_nd6 *)rt->rt_llinfo; 2014 } 2015 if (ln == NULL || rt == NULL) { 2016 if ((ifp->if_flags & IFF_POINTOPOINT) == 0 && 2017 !(ND_IFINFO(ifp)->flags & ND6_IFF_PERFORMNUD)) { 2018 log(LOG_DEBUG, 2019 "nd6_output: can't allocate llinfo for %s " 2020 "(ln=%p, rt=%p)\n", 2021 ip6_sprintf(&dst->sin6_addr), ln, rt); 2022 senderr(EIO); /* XXX: good error? */ 2023 } 2024 2025 goto sendpkt; /* send anyway */ 2026 } 2027 2028 /* We don't have to do link-layer address resolution on a p2p link. */ 2029 if ((ifp->if_flags & IFF_POINTOPOINT) != 0 && 2030 ln->ln_state < ND6_LLINFO_REACHABLE) { 2031 ln->ln_state = ND6_LLINFO_STALE; 2032 nd6_llinfo_settimer(ln, (long)nd6_gctimer * hz); 2033 } 2034 2035 /* 2036 * The first time we send a packet to a neighbor whose entry is 2037 * STALE, we have to change the state to DELAY and a sets a timer to 2038 * expire in DELAY_FIRST_PROBE_TIME seconds to ensure do 2039 * neighbor unreachability detection on expiration. 2040 * (RFC 2461 7.3.3) 2041 */ 2042 if (ln->ln_state == ND6_LLINFO_STALE) { 2043 ln->ln_asked = 0; 2044 ln->ln_state = ND6_LLINFO_DELAY; 2045 nd6_llinfo_settimer(ln, (long)nd6_delay * hz); 2046 } 2047 2048 /* 2049 * If the neighbor cache entry has a state other than INCOMPLETE 2050 * (i.e. its link-layer address is already resolved), just 2051 * send the packet. 2052 */ 2053 if (ln->ln_state > ND6_LLINFO_INCOMPLETE) 2054 goto sendpkt; 2055 2056 /* 2057 * There is a neighbor cache entry, but no ethernet address 2058 * response yet. Append this latest packet to the end of the 2059 * packet queue in the mbuf, unless the number of the packet 2060 * does not exceed nd6_maxqueuelen. When it exceeds nd6_maxqueuelen, 2061 * the oldest packet in the queue will be removed. 2062 */ 2063 if (ln->ln_state == ND6_LLINFO_NOSTATE) 2064 ln->ln_state = ND6_LLINFO_INCOMPLETE; 2065 if (ln->ln_hold) { 2066 struct mbuf *m_hold; 2067 int i; 2068 2069 i = 0; 2070 for (m_hold = ln->ln_hold; m_hold; m_hold = m_hold->m_nextpkt) { 2071 i++; 2072 if (m_hold->m_nextpkt == NULL) { 2073 m_hold->m_nextpkt = m; 2074 break; 2075 } 2076 } 2077 while (i >= nd6_maxqueuelen) { 2078 m_hold = ln->ln_hold; 2079 ln->ln_hold = ln->ln_hold->m_nextpkt; 2080 m_freem(m_hold); 2081 i--; 2082 } 2083 } else { 2084 ln->ln_hold = m; 2085 } 2086 2087 /* 2088 * If there has been no NS for the neighbor after entering the 2089 * INCOMPLETE state, send the first solicitation. 2090 */ 2091 if (!ND6_LLINFO_PERMANENT(ln) && ln->ln_asked == 0) { 2092 ln->ln_asked++; 2093 nd6_llinfo_settimer(ln, 2094 (long)ND_IFINFO(ifp)->retrans * hz / 1000); 2095 nd6_ns_output(ifp, NULL, &dst->sin6_addr, ln, 0); 2096 } 2097 return 0; 2098 2099 sendpkt: 2100 /* discard the packet if IPv6 operation is disabled on the interface */ 2101 if ((ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED)) { 2102 error = ENETDOWN; /* better error? */ 2103 goto bad; 2104 } 2105 2106 #ifdef IPSEC 2107 /* clean ipsec history once it goes out of the node */ 2108 ipsec_delaux(m); 2109 #endif 2110 if ((ifp->if_flags & IFF_LOOPBACK) != 0) 2111 return (*ifp->if_output)(origifp, m, sin6tocsa(dst), rt); 2112 return (*ifp->if_output)(ifp, m, sin6tocsa(dst), rt); 2113 2114 bad: 2115 if (m != NULL) 2116 m_freem(m); 2117 return error; 2118 } 2119 #undef senderr 2120 2121 int 2122 nd6_need_cache(struct ifnet *ifp) 2123 { 2124 /* 2125 * XXX: we currently do not make neighbor cache on any interface 2126 * other than ARCnet, Ethernet, FDDI and GIF. 2127 * 2128 * RFC2893 says: 2129 * - unidirectional tunnels needs no ND 2130 */ 2131 switch (ifp->if_type) { 2132 case IFT_ARCNET: 2133 case IFT_ETHER: 2134 case IFT_FDDI: 2135 case IFT_IEEE1394: 2136 case IFT_CARP: 2137 case IFT_GIF: /* XXX need more cases? */ 2138 case IFT_PPP: 2139 case IFT_TUNNEL: 2140 return 1; 2141 default: 2142 return 0; 2143 } 2144 } 2145 2146 int 2147 nd6_storelladdr(const struct ifnet *ifp, const struct rtentry *rt, 2148 struct mbuf *m, const struct sockaddr *dst, uint8_t *lldst, 2149 size_t dstsize) 2150 { 2151 const struct sockaddr_dl *sdl; 2152 2153 if (m->m_flags & M_MCAST) { 2154 switch (ifp->if_type) { 2155 case IFT_ETHER: 2156 case IFT_FDDI: 2157 ETHER_MAP_IPV6_MULTICAST(&satocsin6(dst)->sin6_addr, 2158 lldst); 2159 return 1; 2160 case IFT_IEEE1394: 2161 memcpy(lldst, ifp->if_broadcastaddr, 2162 MIN(dstsize, ifp->if_addrlen)); 2163 return 1; 2164 case IFT_ARCNET: 2165 *lldst = 0; 2166 return 1; 2167 default: 2168 m_freem(m); 2169 return 0; 2170 } 2171 } 2172 2173 if (rt == NULL) { 2174 /* this could happen, if we could not allocate memory */ 2175 m_freem(m); 2176 return 0; 2177 } 2178 if (rt->rt_gateway->sa_family != AF_LINK) { 2179 printf("%s: something odd happens\n", __func__); 2180 m_freem(m); 2181 return 0; 2182 } 2183 sdl = satocsdl(rt->rt_gateway); 2184 if (sdl->sdl_alen == 0 || sdl->sdl_alen > dstsize) { 2185 /* this should be impossible, but we bark here for debugging */ 2186 printf("%s: sdl_alen == 0, dst=%s, if=%s\n", __func__, 2187 ip6_sprintf(&satocsin6(dst)->sin6_addr), if_name(ifp)); 2188 m_freem(m); 2189 return 0; 2190 } 2191 2192 memcpy(lldst, CLLADDR(sdl), MIN(dstsize, sdl->sdl_alen)); 2193 return 1; 2194 } 2195 2196 static void 2197 clear_llinfo_pqueue(struct llinfo_nd6 *ln) 2198 { 2199 struct mbuf *m_hold, *m_hold_next; 2200 2201 for (m_hold = ln->ln_hold; m_hold; m_hold = m_hold_next) { 2202 m_hold_next = m_hold->m_nextpkt; 2203 m_hold->m_nextpkt = NULL; 2204 m_freem(m_hold); 2205 } 2206 2207 ln->ln_hold = NULL; 2208 return; 2209 } 2210 2211 int 2212 nd6_sysctl( 2213 int name, 2214 void *oldp, /* syscall arg, need copyout */ 2215 size_t *oldlenp, 2216 void *newp, /* syscall arg, need copyin */ 2217 size_t newlen 2218 ) 2219 { 2220 void *p; 2221 size_t ol; 2222 int error; 2223 2224 error = 0; 2225 2226 if (newp) 2227 return EPERM; 2228 if (oldp && !oldlenp) 2229 return EINVAL; 2230 ol = oldlenp ? *oldlenp : 0; 2231 2232 if (oldp) { 2233 p = malloc(*oldlenp, M_TEMP, M_WAITOK); 2234 if (p == NULL) 2235 return ENOMEM; 2236 } else 2237 p = NULL; 2238 switch (name) { 2239 case ICMPV6CTL_ND6_DRLIST: 2240 error = fill_drlist(p, oldlenp, ol); 2241 if (!error && p != NULL && oldp != NULL) 2242 error = copyout(p, oldp, *oldlenp); 2243 break; 2244 2245 case ICMPV6CTL_ND6_PRLIST: 2246 error = fill_prlist(p, oldlenp, ol); 2247 if (!error && p != NULL && oldp != NULL) 2248 error = copyout(p, oldp, *oldlenp); 2249 break; 2250 2251 case ICMPV6CTL_ND6_MAXQLEN: 2252 break; 2253 2254 default: 2255 error = ENOPROTOOPT; 2256 break; 2257 } 2258 if (p) 2259 free(p, M_TEMP); 2260 2261 return error; 2262 } 2263 2264 static int 2265 fill_drlist(void *oldp, size_t *oldlenp, size_t ol) 2266 { 2267 int error = 0, s; 2268 struct in6_defrouter *d = NULL, *de = NULL; 2269 struct nd_defrouter *dr; 2270 size_t l; 2271 2272 s = splsoftnet(); 2273 2274 if (oldp) { 2275 d = (struct in6_defrouter *)oldp; 2276 de = (struct in6_defrouter *)((char *)oldp + *oldlenp); 2277 } 2278 l = 0; 2279 2280 TAILQ_FOREACH(dr, &nd_defrouter, dr_entry) { 2281 2282 if (oldp && d + 1 <= de) { 2283 memset(d, 0, sizeof(*d)); 2284 sockaddr_in6_init(&d->rtaddr, &dr->rtaddr, 0, 0, 0); 2285 if (sa6_recoverscope(&d->rtaddr)) { 2286 log(LOG_ERR, 2287 "scope error in router list (%s)\n", 2288 ip6_sprintf(&d->rtaddr.sin6_addr)); 2289 /* XXX: press on... */ 2290 } 2291 d->flags = dr->flags; 2292 d->rtlifetime = dr->rtlifetime; 2293 d->expire = dr->expire; 2294 d->if_index = dr->ifp->if_index; 2295 } 2296 2297 l += sizeof(*d); 2298 if (d) 2299 d++; 2300 } 2301 2302 if (oldp) { 2303 if (l > ol) 2304 error = ENOMEM; 2305 } 2306 if (oldlenp) 2307 *oldlenp = l; /* (void *)d - (void *)oldp */ 2308 2309 splx(s); 2310 2311 return error; 2312 } 2313 2314 static int 2315 fill_prlist(void *oldp, size_t *oldlenp, size_t ol) 2316 { 2317 int error = 0, s; 2318 struct nd_prefix *pr; 2319 struct in6_prefix *p = NULL; 2320 struct in6_prefix *pe = NULL; 2321 size_t l; 2322 2323 s = splsoftnet(); 2324 2325 if (oldp) { 2326 p = (struct in6_prefix *)oldp; 2327 pe = (struct in6_prefix *)((char *)oldp + *oldlenp); 2328 } 2329 l = 0; 2330 2331 LIST_FOREACH(pr, &nd_prefix, ndpr_entry) { 2332 u_short advrtrs; 2333 size_t advance; 2334 struct sockaddr_in6 *sin6; 2335 struct sockaddr_in6 *s6; 2336 struct nd_pfxrouter *pfr; 2337 2338 if (oldp && p + 1 <= pe) 2339 { 2340 memset(p, 0, sizeof(*p)); 2341 sin6 = (struct sockaddr_in6 *)(p + 1); 2342 2343 p->prefix = pr->ndpr_prefix; 2344 if (sa6_recoverscope(&p->prefix)) { 2345 log(LOG_ERR, 2346 "scope error in prefix list (%s)\n", 2347 ip6_sprintf(&p->prefix.sin6_addr)); 2348 /* XXX: press on... */ 2349 } 2350 p->raflags = pr->ndpr_raf; 2351 p->prefixlen = pr->ndpr_plen; 2352 p->vltime = pr->ndpr_vltime; 2353 p->pltime = pr->ndpr_pltime; 2354 p->if_index = pr->ndpr_ifp->if_index; 2355 if (pr->ndpr_vltime == ND6_INFINITE_LIFETIME) 2356 p->expire = 0; 2357 else { 2358 time_t maxexpire; 2359 2360 /* XXX: we assume time_t is signed. */ 2361 maxexpire = (-1) & 2362 ~((time_t)1 << 2363 ((sizeof(maxexpire) * 8) - 1)); 2364 if (pr->ndpr_vltime < 2365 maxexpire - pr->ndpr_lastupdate) { 2366 p->expire = pr->ndpr_lastupdate + 2367 pr->ndpr_vltime; 2368 } else 2369 p->expire = maxexpire; 2370 } 2371 p->refcnt = pr->ndpr_refcnt; 2372 p->flags = pr->ndpr_stateflags; 2373 p->origin = PR_ORIG_RA; 2374 advrtrs = 0; 2375 LIST_FOREACH(pfr, &pr->ndpr_advrtrs, pfr_entry) { 2376 if ((void *)&sin6[advrtrs + 1] > (void *)pe) { 2377 advrtrs++; 2378 continue; 2379 } 2380 s6 = &sin6[advrtrs]; 2381 sockaddr_in6_init(s6, &pfr->router->rtaddr, 2382 0, 0, 0); 2383 if (sa6_recoverscope(s6)) { 2384 log(LOG_ERR, 2385 "scope error in " 2386 "prefix list (%s)\n", 2387 ip6_sprintf(&pfr->router->rtaddr)); 2388 } 2389 advrtrs++; 2390 } 2391 p->advrtrs = advrtrs; 2392 } 2393 else { 2394 advrtrs = 0; 2395 LIST_FOREACH(pfr, &pr->ndpr_advrtrs, pfr_entry) 2396 advrtrs++; 2397 } 2398 2399 advance = sizeof(*p) + sizeof(*sin6) * advrtrs; 2400 l += advance; 2401 if (p) 2402 p = (struct in6_prefix *)((char *)p + advance); 2403 } 2404 2405 if (oldp) { 2406 *oldlenp = l; /* (void *)d - (void *)oldp */ 2407 if (l > ol) 2408 error = ENOMEM; 2409 } else 2410 *oldlenp = l; 2411 2412 splx(s); 2413 2414 return error; 2415 } 2416