1 /* $NetBSD: nd6.c,v 1.135 2009/11/06 20:41:22 dyoung Exp $ */ 2 /* $KAME: nd6.c,v 1.279 2002/06/08 11:16:51 itojun Exp $ */ 3 4 /* 5 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project. 6 * All rights reserved. 7 * 8 * Redistribution and use in source and binary forms, with or without 9 * modification, are permitted provided that the following conditions 10 * are met: 11 * 1. Redistributions of source code must retain the above copyright 12 * notice, this list of conditions and the following disclaimer. 13 * 2. Redistributions in binary form must reproduce the above copyright 14 * notice, this list of conditions and the following disclaimer in the 15 * documentation and/or other materials provided with the distribution. 16 * 3. Neither the name of the project nor the names of its contributors 17 * may be used to endorse or promote products derived from this software 18 * without specific prior written permission. 19 * 20 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND 21 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 22 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 23 * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE 24 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 25 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 26 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 27 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 28 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 29 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 30 * SUCH DAMAGE. 31 */ 32 33 #include <sys/cdefs.h> 34 __KERNEL_RCSID(0, "$NetBSD: nd6.c,v 1.135 2009/11/06 20:41:22 dyoung Exp $"); 35 36 #include "opt_ipsec.h" 37 38 #include <sys/param.h> 39 #include <sys/systm.h> 40 #include <sys/callout.h> 41 #include <sys/malloc.h> 42 #include <sys/mbuf.h> 43 #include <sys/socket.h> 44 #include <sys/socketvar.h> 45 #include <sys/sockio.h> 46 #include <sys/time.h> 47 #include <sys/kernel.h> 48 #include <sys/protosw.h> 49 #include <sys/errno.h> 50 #include <sys/ioctl.h> 51 #include <sys/syslog.h> 52 #include <sys/queue.h> 53 54 #include <net/if.h> 55 #include <net/if_dl.h> 56 #include <net/if_types.h> 57 #include <net/route.h> 58 #include <net/if_ether.h> 59 #include <net/if_fddi.h> 60 #include <net/if_arc.h> 61 62 #include <netinet/in.h> 63 #include <netinet6/in6_var.h> 64 #include <netinet/ip6.h> 65 #include <netinet6/ip6_var.h> 66 #include <netinet6/scope6_var.h> 67 #include <netinet6/nd6.h> 68 #include <netinet/icmp6.h> 69 #include <netinet6/icmp6_private.h> 70 71 #ifdef IPSEC 72 #include <netinet6/ipsec.h> 73 #endif 74 75 #include <net/net_osdep.h> 76 77 #define ND6_SLOWTIMER_INTERVAL (60 * 60) /* 1 hour */ 78 #define ND6_RECALC_REACHTM_INTERVAL (60 * 120) /* 2 hours */ 79 80 /* timer values */ 81 int nd6_prune = 1; /* walk list every 1 seconds */ 82 int nd6_delay = 5; /* delay first probe time 5 second */ 83 int nd6_umaxtries = 3; /* maximum unicast query */ 84 int nd6_mmaxtries = 3; /* maximum multicast query */ 85 int nd6_useloopback = 1; /* use loopback interface for local traffic */ 86 int nd6_gctimer = (60 * 60 * 24); /* 1 day: garbage collection timer */ 87 88 /* preventing too many loops in ND option parsing */ 89 int nd6_maxndopt = 10; /* max # of ND options allowed */ 90 91 int nd6_maxnudhint = 0; /* max # of subsequent upper layer hints */ 92 93 int nd6_maxqueuelen = 1; /* max # of packets cached in unresolved ND entries */ 94 95 #ifdef ND6_DEBUG 96 int nd6_debug = 1; 97 #else 98 int nd6_debug = 0; 99 #endif 100 101 /* for debugging? */ 102 static int nd6_inuse, nd6_allocated; 103 104 struct llinfo_nd6 llinfo_nd6 = { 105 .ln_prev = &llinfo_nd6, 106 .ln_next = &llinfo_nd6, 107 }; 108 struct nd_drhead nd_defrouter; 109 struct nd_prhead nd_prefix = { 0 }; 110 111 int nd6_recalc_reachtm_interval = ND6_RECALC_REACHTM_INTERVAL; 112 static const struct sockaddr_in6 all1_sa = { 113 .sin6_family = AF_INET6 114 , .sin6_len = sizeof(struct sockaddr_in6) 115 , .sin6_addr = {.s6_addr = {0xff, 0xff, 0xff, 0xff, 116 0xff, 0xff, 0xff, 0xff, 117 0xff, 0xff, 0xff, 0xff, 118 0xff, 0xff, 0xff, 0xff}} 119 }; 120 121 static void nd6_setmtu0(struct ifnet *, struct nd_ifinfo *); 122 static void nd6_slowtimo(void *); 123 static int regen_tmpaddr(struct in6_ifaddr *); 124 static struct llinfo_nd6 *nd6_free(struct rtentry *, int); 125 static void nd6_llinfo_timer(void *); 126 static void clear_llinfo_pqueue(struct llinfo_nd6 *); 127 128 callout_t nd6_slowtimo_ch; 129 callout_t nd6_timer_ch; 130 extern callout_t in6_tmpaddrtimer_ch; 131 132 static int fill_drlist(void *, size_t *, size_t); 133 static int fill_prlist(void *, size_t *, size_t); 134 135 MALLOC_DEFINE(M_IP6NDP, "NDP", "IPv6 Neighbour Discovery"); 136 137 void 138 nd6_init(void) 139 { 140 static int nd6_init_done = 0; 141 142 if (nd6_init_done) { 143 log(LOG_NOTICE, "nd6_init called more than once(ignored)\n"); 144 return; 145 } 146 147 /* initialization of the default router list */ 148 TAILQ_INIT(&nd_defrouter); 149 150 nd6_init_done = 1; 151 152 callout_init(&nd6_slowtimo_ch, CALLOUT_MPSAFE); 153 callout_init(&nd6_timer_ch, CALLOUT_MPSAFE); 154 155 /* start timer */ 156 callout_reset(&nd6_slowtimo_ch, ND6_SLOWTIMER_INTERVAL * hz, 157 nd6_slowtimo, NULL); 158 } 159 160 struct nd_ifinfo * 161 nd6_ifattach(struct ifnet *ifp) 162 { 163 struct nd_ifinfo *nd; 164 165 nd = (struct nd_ifinfo *)malloc(sizeof(*nd), M_IP6NDP, M_WAITOK); 166 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. 176 * Because we do not set ND6_IFF_OVERRIDE_RTADV here, we won't 177 * accept RAs by default. 178 */ 179 nd->flags = ND6_IFF_PERFORMNUD | ND6_IFF_ACCEPT_RTADV; 180 181 /* XXX: we cannot call nd6_setmtu since ifp is not fully initialized */ 182 nd6_setmtu0(ifp, nd); 183 184 return nd; 185 } 186 187 void 188 nd6_ifdetach(struct nd_ifinfo *nd) 189 { 190 191 free(nd, M_IP6NDP); 192 } 193 194 void 195 nd6_setmtu(struct ifnet *ifp) 196 { 197 nd6_setmtu0(ifp, ND_IFINFO(ifp)); 198 } 199 200 void 201 nd6_setmtu0(struct ifnet *ifp, struct nd_ifinfo *ndi) 202 { 203 u_int32_t omaxmtu; 204 205 omaxmtu = ndi->maxmtu; 206 207 switch (ifp->if_type) { 208 case IFT_ARCNET: 209 ndi->maxmtu = MIN(ARC_PHDS_MAXMTU, ifp->if_mtu); /* RFC2497 */ 210 break; 211 case IFT_FDDI: 212 ndi->maxmtu = MIN(FDDIIPMTU, ifp->if_mtu); 213 break; 214 default: 215 ndi->maxmtu = ifp->if_mtu; 216 break; 217 } 218 219 /* 220 * Decreasing the interface MTU under IPV6 minimum MTU may cause 221 * undesirable situation. We thus notify the operator of the change 222 * explicitly. The check for omaxmtu is necessary to restrict the 223 * log to the case of changing the MTU, not initializing it. 224 */ 225 if (omaxmtu >= IPV6_MMTU && ndi->maxmtu < IPV6_MMTU) { 226 log(LOG_NOTICE, "nd6_setmtu0: new link MTU on %s (%lu) is too" 227 " small for IPv6 which needs %lu\n", 228 if_name(ifp), (unsigned long)ndi->maxmtu, (unsigned long) 229 IPV6_MMTU); 230 } 231 232 if (ndi->maxmtu > in6_maxmtu) 233 in6_setmaxmtu(); /* check all interfaces just in case */ 234 } 235 236 void 237 nd6_option_init(void *opt, int icmp6len, union nd_opts *ndopts) 238 { 239 240 memset(ndopts, 0, sizeof(*ndopts)); 241 ndopts->nd_opts_search = (struct nd_opt_hdr *)opt; 242 ndopts->nd_opts_last 243 = (struct nd_opt_hdr *)(((u_char *)opt) + icmp6len); 244 245 if (icmp6len == 0) { 246 ndopts->nd_opts_done = 1; 247 ndopts->nd_opts_search = NULL; 248 } 249 } 250 251 /* 252 * Take one ND option. 253 */ 254 struct nd_opt_hdr * 255 nd6_option(union nd_opts *ndopts) 256 { 257 struct nd_opt_hdr *nd_opt; 258 int olen; 259 260 if (ndopts == NULL) 261 panic("ndopts == NULL in nd6_option"); 262 if (ndopts->nd_opts_last == NULL) 263 panic("uninitialized ndopts in nd6_option"); 264 if (ndopts->nd_opts_search == NULL) 265 return NULL; 266 if (ndopts->nd_opts_done) 267 return NULL; 268 269 nd_opt = ndopts->nd_opts_search; 270 271 /* make sure nd_opt_len is inside the buffer */ 272 if ((void *)&nd_opt->nd_opt_len >= (void *)ndopts->nd_opts_last) { 273 memset(ndopts, 0, sizeof(*ndopts)); 274 return NULL; 275 } 276 277 olen = nd_opt->nd_opt_len << 3; 278 if (olen == 0) { 279 /* 280 * Message validation requires that all included 281 * options have a length that is greater than zero. 282 */ 283 memset(ndopts, 0, sizeof(*ndopts)); 284 return NULL; 285 } 286 287 ndopts->nd_opts_search = (struct nd_opt_hdr *)((char *)nd_opt + olen); 288 if (ndopts->nd_opts_search > ndopts->nd_opts_last) { 289 /* option overruns the end of buffer, invalid */ 290 memset(ndopts, 0, sizeof(*ndopts)); 291 return NULL; 292 } else if (ndopts->nd_opts_search == ndopts->nd_opts_last) { 293 /* reached the end of options chain */ 294 ndopts->nd_opts_done = 1; 295 ndopts->nd_opts_search = NULL; 296 } 297 return nd_opt; 298 } 299 300 /* 301 * Parse multiple ND options. 302 * This function is much easier to use, for ND routines that do not need 303 * multiple options of the same type. 304 */ 305 int 306 nd6_options(union nd_opts *ndopts) 307 { 308 struct nd_opt_hdr *nd_opt; 309 int i = 0; 310 311 if (ndopts == NULL) 312 panic("ndopts == NULL in nd6_options"); 313 if (ndopts->nd_opts_last == NULL) 314 panic("uninitialized ndopts in nd6_options"); 315 if (ndopts->nd_opts_search == NULL) 316 return 0; 317 318 while (1) { 319 nd_opt = nd6_option(ndopts); 320 if (nd_opt == NULL && ndopts->nd_opts_last == NULL) { 321 /* 322 * Message validation requires that all included 323 * options have a length that is greater than zero. 324 */ 325 ICMP6_STATINC(ICMP6_STAT_ND_BADOPT); 326 memset(ndopts, 0, sizeof(*ndopts)); 327 return -1; 328 } 329 330 if (nd_opt == NULL) 331 goto skip1; 332 333 switch (nd_opt->nd_opt_type) { 334 case ND_OPT_SOURCE_LINKADDR: 335 case ND_OPT_TARGET_LINKADDR: 336 case ND_OPT_MTU: 337 case ND_OPT_REDIRECTED_HEADER: 338 if (ndopts->nd_opt_array[nd_opt->nd_opt_type]) { 339 nd6log((LOG_INFO, 340 "duplicated ND6 option found (type=%d)\n", 341 nd_opt->nd_opt_type)); 342 /* XXX bark? */ 343 } else { 344 ndopts->nd_opt_array[nd_opt->nd_opt_type] 345 = nd_opt; 346 } 347 break; 348 case ND_OPT_PREFIX_INFORMATION: 349 if (ndopts->nd_opt_array[nd_opt->nd_opt_type] == 0) { 350 ndopts->nd_opt_array[nd_opt->nd_opt_type] 351 = nd_opt; 352 } 353 ndopts->nd_opts_pi_end = 354 (struct nd_opt_prefix_info *)nd_opt; 355 break; 356 default: 357 /* 358 * Unknown options must be silently ignored, 359 * to accommodate future extension to the protocol. 360 */ 361 nd6log((LOG_DEBUG, 362 "nd6_options: unsupported option %d - " 363 "option ignored\n", nd_opt->nd_opt_type)); 364 } 365 366 skip1: 367 i++; 368 if (i > nd6_maxndopt) { 369 ICMP6_STATINC(ICMP6_STAT_ND_TOOMANYOPT); 370 nd6log((LOG_INFO, "too many loop in nd opt\n")); 371 break; 372 } 373 374 if (ndopts->nd_opts_done) 375 break; 376 } 377 378 return 0; 379 } 380 381 /* 382 * ND6 timer routine to handle ND6 entries 383 */ 384 void 385 nd6_llinfo_settimer(struct llinfo_nd6 *ln, long xtick) 386 { 387 int s; 388 389 s = splsoftnet(); 390 391 if (xtick < 0) { 392 ln->ln_expire = 0; 393 ln->ln_ntick = 0; 394 callout_stop(&ln->ln_timer_ch); 395 } else { 396 ln->ln_expire = time_second + xtick / hz; 397 if (xtick > INT_MAX) { 398 ln->ln_ntick = xtick - INT_MAX; 399 callout_reset(&ln->ln_timer_ch, INT_MAX, 400 nd6_llinfo_timer, ln); 401 } else { 402 ln->ln_ntick = 0; 403 callout_reset(&ln->ln_timer_ch, xtick, 404 nd6_llinfo_timer, ln); 405 } 406 } 407 408 splx(s); 409 } 410 411 static void 412 nd6_llinfo_timer(void *arg) 413 { 414 struct llinfo_nd6 *ln; 415 struct rtentry *rt; 416 const struct sockaddr_in6 *dst; 417 struct ifnet *ifp; 418 struct nd_ifinfo *ndi = NULL; 419 420 mutex_enter(softnet_lock); 421 KERNEL_LOCK(1, NULL); 422 423 ln = (struct llinfo_nd6 *)arg; 424 425 if (ln->ln_ntick > 0) { 426 nd6_llinfo_settimer(ln, ln->ln_ntick); 427 KERNEL_UNLOCK_ONE(NULL); 428 mutex_exit(softnet_lock); 429 return; 430 } 431 432 if ((rt = ln->ln_rt) == NULL) 433 panic("ln->ln_rt == NULL"); 434 if ((ifp = rt->rt_ifp) == NULL) 435 panic("ln->ln_rt->rt_ifp == NULL"); 436 ndi = ND_IFINFO(ifp); 437 dst = satocsin6(rt_getkey(rt)); 438 439 /* sanity check */ 440 if (rt->rt_llinfo && (struct llinfo_nd6 *)rt->rt_llinfo != ln) 441 panic("rt_llinfo(%p) is not equal to ln(%p)", 442 rt->rt_llinfo, ln); 443 if (!dst) 444 panic("dst=0 in nd6_timer(ln=%p)", ln); 445 446 switch (ln->ln_state) { 447 case ND6_LLINFO_INCOMPLETE: 448 if (ln->ln_asked < nd6_mmaxtries) { 449 ln->ln_asked++; 450 nd6_llinfo_settimer(ln, (long)ndi->retrans * hz / 1000); 451 nd6_ns_output(ifp, NULL, &dst->sin6_addr, ln, 0); 452 } else { 453 struct mbuf *m = ln->ln_hold; 454 if (m) { 455 struct mbuf *m0; 456 457 /* 458 * assuming every packet in ln_hold has 459 * the same IP header 460 */ 461 m0 = m->m_nextpkt; 462 m->m_nextpkt = NULL; 463 icmp6_error2(m, ICMP6_DST_UNREACH, 464 ICMP6_DST_UNREACH_ADDR, 0, rt->rt_ifp); 465 466 ln->ln_hold = m0; 467 clear_llinfo_pqueue(ln); 468 } 469 (void)nd6_free(rt, 0); 470 ln = NULL; 471 } 472 break; 473 case ND6_LLINFO_REACHABLE: 474 if (!ND6_LLINFO_PERMANENT(ln)) { 475 ln->ln_state = ND6_LLINFO_STALE; 476 nd6_llinfo_settimer(ln, (long)nd6_gctimer * hz); 477 } 478 break; 479 480 case ND6_LLINFO_STALE: 481 /* Garbage Collection(RFC 2461 5.3) */ 482 if (!ND6_LLINFO_PERMANENT(ln)) { 483 (void)nd6_free(rt, 1); 484 ln = NULL; 485 } 486 break; 487 488 case ND6_LLINFO_DELAY: 489 if (ndi && (ndi->flags & ND6_IFF_PERFORMNUD) != 0) { 490 /* We need NUD */ 491 ln->ln_asked = 1; 492 ln->ln_state = ND6_LLINFO_PROBE; 493 nd6_llinfo_settimer(ln, (long)ndi->retrans * hz / 1000); 494 nd6_ns_output(ifp, &dst->sin6_addr, 495 &dst->sin6_addr, ln, 0); 496 } else { 497 ln->ln_state = ND6_LLINFO_STALE; /* XXX */ 498 nd6_llinfo_settimer(ln, (long)nd6_gctimer * hz); 499 } 500 break; 501 case ND6_LLINFO_PROBE: 502 if (ln->ln_asked < nd6_umaxtries) { 503 ln->ln_asked++; 504 nd6_llinfo_settimer(ln, (long)ndi->retrans * hz / 1000); 505 nd6_ns_output(ifp, &dst->sin6_addr, 506 &dst->sin6_addr, ln, 0); 507 } else { 508 (void)nd6_free(rt, 0); 509 ln = NULL; 510 } 511 break; 512 } 513 514 KERNEL_UNLOCK_ONE(NULL); 515 mutex_exit(softnet_lock); 516 } 517 518 /* 519 * ND6 timer routine to expire default route list and prefix list 520 */ 521 void 522 nd6_timer(void *ignored_arg) 523 { 524 struct nd_defrouter *next_dr, *dr; 525 struct nd_prefix *next_pr, *pr; 526 struct in6_ifaddr *ia6, *nia6; 527 struct in6_addrlifetime *lt6; 528 529 callout_reset(&nd6_timer_ch, nd6_prune * hz, 530 nd6_timer, NULL); 531 532 mutex_enter(softnet_lock); 533 KERNEL_LOCK(1, NULL); 534 535 /* expire default router list */ 536 537 for (dr = TAILQ_FIRST(&nd_defrouter); dr != NULL; dr = next_dr) { 538 next_dr = TAILQ_NEXT(dr, dr_entry); 539 if (dr->expire && dr->expire < time_second) { 540 defrtrlist_del(dr); 541 } 542 } 543 544 /* 545 * expire interface addresses. 546 * in the past the loop was inside prefix expiry processing. 547 * However, from a stricter speci-confrmance standpoint, we should 548 * rather separate address lifetimes and prefix lifetimes. 549 */ 550 addrloop: 551 for (ia6 = in6_ifaddr; ia6; ia6 = nia6) { 552 nia6 = ia6->ia_next; 553 /* check address lifetime */ 554 lt6 = &ia6->ia6_lifetime; 555 if (IFA6_IS_INVALID(ia6)) { 556 int regen = 0; 557 558 /* 559 * If the expiring address is temporary, try 560 * regenerating a new one. This would be useful when 561 * we suspended a laptop PC, then turned it on after a 562 * period that could invalidate all temporary 563 * addresses. Although we may have to restart the 564 * loop (see below), it must be after purging the 565 * address. Otherwise, we'd see an infinite loop of 566 * regeneration. 567 */ 568 if (ip6_use_tempaddr && 569 (ia6->ia6_flags & IN6_IFF_TEMPORARY) != 0) { 570 if (regen_tmpaddr(ia6) == 0) 571 regen = 1; 572 } 573 574 in6_purgeaddr(&ia6->ia_ifa); 575 576 if (regen) 577 goto addrloop; /* XXX: see below */ 578 } else if (IFA6_IS_DEPRECATED(ia6)) { 579 int oldflags = ia6->ia6_flags; 580 581 ia6->ia6_flags |= IN6_IFF_DEPRECATED; 582 583 /* 584 * If a temporary address has just become deprecated, 585 * regenerate a new one if possible. 586 */ 587 if (ip6_use_tempaddr && 588 (ia6->ia6_flags & IN6_IFF_TEMPORARY) != 0 && 589 (oldflags & IN6_IFF_DEPRECATED) == 0) { 590 591 if (regen_tmpaddr(ia6) == 0) { 592 /* 593 * A new temporary address is 594 * generated. 595 * XXX: this means the address chain 596 * has changed while we are still in 597 * the loop. Although the change 598 * would not cause disaster (because 599 * it's not a deletion, but an 600 * addition,) we'd rather restart the 601 * loop just for safety. Or does this 602 * significantly reduce performance?? 603 */ 604 goto addrloop; 605 } 606 } 607 } else { 608 /* 609 * A new RA might have made a deprecated address 610 * preferred. 611 */ 612 ia6->ia6_flags &= ~IN6_IFF_DEPRECATED; 613 } 614 } 615 616 /* expire prefix list */ 617 for (pr = LIST_FIRST(&nd_prefix); pr != NULL; pr = next_pr) { 618 next_pr = LIST_NEXT(pr, ndpr_entry); 619 /* 620 * check prefix lifetime. 621 * since pltime is just for autoconf, pltime processing for 622 * prefix is not necessary. 623 */ 624 if (pr->ndpr_vltime != ND6_INFINITE_LIFETIME && 625 time_second - pr->ndpr_lastupdate > pr->ndpr_vltime) { 626 627 /* 628 * address expiration and prefix expiration are 629 * separate. NEVER perform in6_purgeaddr here. 630 */ 631 632 prelist_remove(pr); 633 } 634 } 635 636 KERNEL_UNLOCK_ONE(NULL); 637 mutex_exit(softnet_lock); 638 } 639 640 /* ia6: deprecated/invalidated temporary address */ 641 static int 642 regen_tmpaddr(struct in6_ifaddr *ia6) 643 { 644 struct ifaddr *ifa; 645 struct ifnet *ifp; 646 struct in6_ifaddr *public_ifa6 = NULL; 647 648 ifp = ia6->ia_ifa.ifa_ifp; 649 IFADDR_FOREACH(ifa, ifp) { 650 struct in6_ifaddr *it6; 651 652 if (ifa->ifa_addr->sa_family != AF_INET6) 653 continue; 654 655 it6 = (struct in6_ifaddr *)ifa; 656 657 /* ignore no autoconf addresses. */ 658 if ((it6->ia6_flags & IN6_IFF_AUTOCONF) == 0) 659 continue; 660 661 /* ignore autoconf addresses with different prefixes. */ 662 if (it6->ia6_ndpr == NULL || it6->ia6_ndpr != ia6->ia6_ndpr) 663 continue; 664 665 /* 666 * Now we are looking at an autoconf address with the same 667 * prefix as ours. If the address is temporary and is still 668 * preferred, do not create another one. It would be rare, but 669 * could happen, for example, when we resume a laptop PC after 670 * a long period. 671 */ 672 if ((it6->ia6_flags & IN6_IFF_TEMPORARY) != 0 && 673 !IFA6_IS_DEPRECATED(it6)) { 674 public_ifa6 = NULL; 675 break; 676 } 677 678 /* 679 * This is a public autoconf address that has the same prefix 680 * as ours. If it is preferred, keep it. We can't break the 681 * loop here, because there may be a still-preferred temporary 682 * address with the prefix. 683 */ 684 if (!IFA6_IS_DEPRECATED(it6)) 685 public_ifa6 = it6; 686 } 687 688 if (public_ifa6 != NULL) { 689 int e; 690 691 /* 692 * Random factor is introduced in the preferred lifetime, so 693 * we do not need additional delay (3rd arg to in6_tmpifadd). 694 */ 695 if ((e = in6_tmpifadd(public_ifa6, 0, 0)) != 0) { 696 log(LOG_NOTICE, "regen_tmpaddr: failed to create a new" 697 " tmp addr, errno=%d\n", e); 698 return -1; 699 } 700 return 0; 701 } 702 703 return -1; 704 } 705 706 bool 707 nd6_accepts_rtadv(const struct nd_ifinfo *ndi) 708 { 709 switch (ndi->flags & (ND6_IFF_ACCEPT_RTADV|ND6_IFF_OVERRIDE_RTADV)) { 710 case ND6_IFF_OVERRIDE_RTADV|ND6_IFF_ACCEPT_RTADV: 711 return true; 712 case ND6_IFF_ACCEPT_RTADV: 713 return ip6_accept_rtadv != 0; 714 case ND6_IFF_OVERRIDE_RTADV: 715 case 0: 716 default: 717 return false; 718 } 719 } 720 721 /* 722 * Nuke neighbor cache/prefix/default router management table, right before 723 * ifp goes away. 724 */ 725 void 726 nd6_purge(struct ifnet *ifp) 727 { 728 struct nd_ifinfo *ndi = ND_IFINFO(ifp); 729 struct llinfo_nd6 *ln, *nln; 730 struct nd_defrouter *dr, *ndr; 731 struct nd_prefix *pr, *npr; 732 733 /* 734 * Nuke default router list entries toward ifp. 735 * We defer removal of default router list entries that is installed 736 * in the routing table, in order to keep additional side effects as 737 * small as possible. 738 */ 739 for (dr = TAILQ_FIRST(&nd_defrouter); dr != NULL; dr = ndr) { 740 ndr = TAILQ_NEXT(dr, dr_entry); 741 if (dr->installed) 742 continue; 743 744 if (dr->ifp == ifp) 745 defrtrlist_del(dr); 746 } 747 for (dr = TAILQ_FIRST(&nd_defrouter); dr != NULL; dr = ndr) { 748 ndr = TAILQ_NEXT(dr, dr_entry); 749 if (!dr->installed) 750 continue; 751 752 if (dr->ifp == ifp) 753 defrtrlist_del(dr); 754 } 755 756 /* Nuke prefix list entries toward ifp */ 757 for (pr = LIST_FIRST(&nd_prefix); pr != NULL; pr = npr) { 758 npr = LIST_NEXT(pr, ndpr_entry); 759 if (pr->ndpr_ifp == ifp) { 760 /* 761 * Because if_detach() does *not* release prefixes 762 * while purging addresses the reference count will 763 * still be above zero. We therefore reset it to 764 * make sure that the prefix really gets purged. 765 */ 766 pr->ndpr_refcnt = 0; 767 /* 768 * Previously, pr->ndpr_addr is removed as well, 769 * but I strongly believe we don't have to do it. 770 * nd6_purge() is only called from in6_ifdetach(), 771 * which removes all the associated interface addresses 772 * by itself. 773 * (jinmei@kame.net 20010129) 774 */ 775 prelist_remove(pr); 776 } 777 } 778 779 /* cancel default outgoing interface setting */ 780 if (nd6_defifindex == ifp->if_index) 781 nd6_setdefaultiface(0); 782 783 /* XXX: too restrictive? */ 784 if (!ip6_forwarding && ndi && nd6_accepts_rtadv(ndi)) { 785 /* refresh default router list */ 786 defrouter_select(); 787 } 788 789 /* 790 * Nuke neighbor cache entries for the ifp. 791 * Note that rt->rt_ifp may not be the same as ifp, 792 * due to KAME goto ours hack. See RTM_RESOLVE case in 793 * nd6_rtrequest(), and ip6_input(). 794 */ 795 ln = llinfo_nd6.ln_next; 796 while (ln != NULL && ln != &llinfo_nd6) { 797 struct rtentry *rt; 798 const struct sockaddr_dl *sdl; 799 800 nln = ln->ln_next; 801 rt = ln->ln_rt; 802 if (rt && rt->rt_gateway && 803 rt->rt_gateway->sa_family == AF_LINK) { 804 sdl = satocsdl(rt->rt_gateway); 805 if (sdl->sdl_index == ifp->if_index) 806 nln = nd6_free(rt, 0); 807 } 808 ln = nln; 809 } 810 } 811 812 struct rtentry * 813 nd6_lookup(const struct in6_addr *addr6, int create, struct ifnet *ifp) 814 { 815 struct rtentry *rt; 816 struct sockaddr_in6 sin6; 817 818 sockaddr_in6_init(&sin6, addr6, 0, 0, 0); 819 rt = rtalloc1((struct sockaddr *)&sin6, create); 820 if (rt != NULL && (rt->rt_flags & RTF_LLINFO) == 0) { 821 /* 822 * This is the case for the default route. 823 * If we want to create a neighbor cache for the address, we 824 * should free the route for the destination and allocate an 825 * interface route. 826 */ 827 if (create) { 828 RTFREE(rt); 829 rt = NULL; 830 } 831 } 832 if (rt != NULL) 833 ; 834 else if (create && ifp) { 835 int e; 836 837 /* 838 * If no route is available and create is set, 839 * we allocate a host route for the destination 840 * and treat it like an interface route. 841 * This hack is necessary for a neighbor which can't 842 * be covered by our own prefix. 843 */ 844 struct ifaddr *ifa = 845 ifaof_ifpforaddr((struct sockaddr *)&sin6, ifp); 846 if (ifa == NULL) 847 return NULL; 848 849 /* 850 * Create a new route. RTF_LLINFO is necessary 851 * to create a Neighbor Cache entry for the 852 * destination in nd6_rtrequest which will be 853 * called in rtrequest via ifa->ifa_rtrequest. 854 */ 855 if ((e = rtrequest(RTM_ADD, (const struct sockaddr *)&sin6, 856 ifa->ifa_addr, (const struct sockaddr *)&all1_sa, 857 (ifa->ifa_flags | RTF_HOST | RTF_LLINFO) & 858 ~RTF_CLONING, &rt)) != 0) { 859 #if 0 860 log(LOG_ERR, 861 "nd6_lookup: failed to add route for a " 862 "neighbor(%s), errno=%d\n", 863 ip6_sprintf(addr6), e); 864 #endif 865 return NULL; 866 } 867 if (rt == NULL) 868 return NULL; 869 if (rt->rt_llinfo) { 870 struct llinfo_nd6 *ln = 871 (struct llinfo_nd6 *)rt->rt_llinfo; 872 ln->ln_state = ND6_LLINFO_NOSTATE; 873 } 874 } else 875 return NULL; 876 rt->rt_refcnt--; 877 /* 878 * Validation for the entry. 879 * Note that the check for rt_llinfo is necessary because a cloned 880 * route from a parent route that has the L flag (e.g. the default 881 * route to a p2p interface) may have the flag, too, while the 882 * destination is not actually a neighbor. 883 * XXX: we can't use rt->rt_ifp to check for the interface, since 884 * it might be the loopback interface if the entry is for our 885 * own address on a non-loopback interface. Instead, we should 886 * use rt->rt_ifa->ifa_ifp, which would specify the REAL 887 * interface. 888 * Note also that ifa_ifp and ifp may differ when we connect two 889 * interfaces to a same link, install a link prefix to an interface, 890 * and try to install a neighbor cache on an interface that does not 891 * have a route to the prefix. 892 */ 893 if ((rt->rt_flags & RTF_GATEWAY) || (rt->rt_flags & RTF_LLINFO) == 0 || 894 rt->rt_gateway->sa_family != AF_LINK || rt->rt_llinfo == NULL || 895 (ifp && rt->rt_ifa->ifa_ifp != ifp)) { 896 if (create) { 897 nd6log((LOG_DEBUG, 898 "nd6_lookup: failed to lookup %s (if = %s)\n", 899 ip6_sprintf(addr6), 900 ifp ? if_name(ifp) : "unspec")); 901 } 902 return NULL; 903 } 904 return rt; 905 } 906 907 /* 908 * Detect if a given IPv6 address identifies a neighbor on a given link. 909 * XXX: should take care of the destination of a p2p link? 910 */ 911 int 912 nd6_is_addr_neighbor(const struct sockaddr_in6 *addr, struct ifnet *ifp) 913 { 914 struct nd_prefix *pr; 915 916 /* 917 * A link-local address is always a neighbor. 918 * XXX: a link does not necessarily specify a single interface. 919 */ 920 if (IN6_IS_ADDR_LINKLOCAL(&addr->sin6_addr)) { 921 struct sockaddr_in6 sin6_copy; 922 u_int32_t zone; 923 924 /* 925 * We need sin6_copy since sa6_recoverscope() may modify the 926 * content (XXX). 927 */ 928 sin6_copy = *addr; 929 if (sa6_recoverscope(&sin6_copy)) 930 return 0; /* XXX: should be impossible */ 931 if (in6_setscope(&sin6_copy.sin6_addr, ifp, &zone)) 932 return 0; 933 if (sin6_copy.sin6_scope_id == zone) 934 return 1; 935 else 936 return 0; 937 } 938 939 /* 940 * If the address matches one of our on-link prefixes, it should be a 941 * neighbor. 942 */ 943 LIST_FOREACH(pr, &nd_prefix, ndpr_entry) { 944 if (pr->ndpr_ifp != ifp) 945 continue; 946 947 if (!(pr->ndpr_stateflags & NDPRF_ONLINK)) 948 continue; 949 950 if (IN6_ARE_MASKED_ADDR_EQUAL(&pr->ndpr_prefix.sin6_addr, 951 &addr->sin6_addr, &pr->ndpr_mask)) 952 return 1; 953 } 954 955 /* 956 * If the default router list is empty, all addresses are regarded 957 * as on-link, and thus, as a neighbor. 958 * XXX: we restrict the condition to hosts, because routers usually do 959 * not have the "default router list". 960 */ 961 if (!ip6_forwarding && TAILQ_FIRST(&nd_defrouter) == NULL && 962 nd6_defifindex == ifp->if_index) { 963 return 1; 964 } 965 966 /* 967 * Even if the address matches none of our addresses, it might be 968 * in the neighbor cache. 969 */ 970 if (nd6_lookup(&addr->sin6_addr, 0, ifp) != NULL) 971 return 1; 972 973 return 0; 974 } 975 976 /* 977 * Free an nd6 llinfo entry. 978 * Since the function would cause significant changes in the kernel, DO NOT 979 * make it global, unless you have a strong reason for the change, and are sure 980 * that the change is safe. 981 */ 982 static struct llinfo_nd6 * 983 nd6_free(struct rtentry *rt, int gc) 984 { 985 struct llinfo_nd6 *ln = (struct llinfo_nd6 *)rt->rt_llinfo, *next; 986 struct in6_addr in6 = satocsin6(rt_getkey(rt))->sin6_addr; 987 struct nd_defrouter *dr; 988 989 /* 990 * we used to have pfctlinput(PRC_HOSTDEAD) here. 991 * even though it is not harmful, it was not really necessary. 992 */ 993 994 /* cancel timer */ 995 nd6_llinfo_settimer(ln, -1); 996 997 if (!ip6_forwarding) { 998 int s; 999 s = splsoftnet(); 1000 dr = defrouter_lookup(&satocsin6(rt_getkey(rt))->sin6_addr, 1001 rt->rt_ifp); 1002 1003 if (dr != NULL && dr->expire && 1004 ln->ln_state == ND6_LLINFO_STALE && gc) { 1005 /* 1006 * If the reason for the deletion is just garbage 1007 * collection, and the neighbor is an active default 1008 * router, do not delete it. Instead, reset the GC 1009 * timer using the router's lifetime. 1010 * Simply deleting the entry would affect default 1011 * router selection, which is not necessarily a good 1012 * thing, especially when we're using router preference 1013 * values. 1014 * XXX: the check for ln_state would be redundant, 1015 * but we intentionally keep it just in case. 1016 */ 1017 if (dr->expire > time_second) 1018 nd6_llinfo_settimer(ln, 1019 (dr->expire - time_second) * hz); 1020 else 1021 nd6_llinfo_settimer(ln, (long)nd6_gctimer * hz); 1022 splx(s); 1023 return ln->ln_next; 1024 } 1025 1026 if (ln->ln_router || dr) { 1027 /* 1028 * rt6_flush must be called whether or not the neighbor 1029 * is in the Default Router List. 1030 * See a corresponding comment in nd6_na_input(). 1031 */ 1032 rt6_flush(&in6, rt->rt_ifp); 1033 } 1034 1035 if (dr) { 1036 /* 1037 * Unreachablity of a router might affect the default 1038 * router selection and on-link detection of advertised 1039 * prefixes. 1040 */ 1041 1042 /* 1043 * Temporarily fake the state to choose a new default 1044 * router and to perform on-link determination of 1045 * prefixes correctly. 1046 * Below the state will be set correctly, 1047 * or the entry itself will be deleted. 1048 */ 1049 ln->ln_state = ND6_LLINFO_INCOMPLETE; 1050 1051 /* 1052 * Since defrouter_select() does not affect the 1053 * on-link determination and MIP6 needs the check 1054 * before the default router selection, we perform 1055 * the check now. 1056 */ 1057 pfxlist_onlink_check(); 1058 1059 /* 1060 * refresh default router list 1061 */ 1062 defrouter_select(); 1063 } 1064 splx(s); 1065 } 1066 1067 /* 1068 * Before deleting the entry, remember the next entry as the 1069 * return value. We need this because pfxlist_onlink_check() above 1070 * might have freed other entries (particularly the old next entry) as 1071 * a side effect (XXX). 1072 */ 1073 next = ln->ln_next; 1074 1075 /* 1076 * Detach the route from the routing tree and the list of neighbor 1077 * caches, and disable the route entry not to be used in already 1078 * cached routes. 1079 */ 1080 rtrequest(RTM_DELETE, rt_getkey(rt), NULL, rt_mask(rt), 0, NULL); 1081 1082 return next; 1083 } 1084 1085 /* 1086 * Upper-layer reachability hint for Neighbor Unreachability Detection. 1087 * 1088 * XXX cost-effective methods? 1089 */ 1090 void 1091 nd6_nud_hint(struct rtentry *rt, struct in6_addr *dst6, int force) 1092 { 1093 struct llinfo_nd6 *ln; 1094 1095 /* 1096 * If the caller specified "rt", use that. Otherwise, resolve the 1097 * routing table by supplied "dst6". 1098 */ 1099 if (rt == NULL) { 1100 if (dst6 == NULL) 1101 return; 1102 if ((rt = nd6_lookup(dst6, 0, NULL)) == NULL) 1103 return; 1104 } 1105 1106 if ((rt->rt_flags & RTF_GATEWAY) != 0 || 1107 (rt->rt_flags & RTF_LLINFO) == 0 || 1108 !rt->rt_llinfo || !rt->rt_gateway || 1109 rt->rt_gateway->sa_family != AF_LINK) { 1110 /* This is not a host route. */ 1111 return; 1112 } 1113 1114 ln = (struct llinfo_nd6 *)rt->rt_llinfo; 1115 if (ln->ln_state < ND6_LLINFO_REACHABLE) 1116 return; 1117 1118 /* 1119 * if we get upper-layer reachability confirmation many times, 1120 * it is possible we have false information. 1121 */ 1122 if (!force) { 1123 ln->ln_byhint++; 1124 if (ln->ln_byhint > nd6_maxnudhint) 1125 return; 1126 } 1127 1128 ln->ln_state = ND6_LLINFO_REACHABLE; 1129 if (!ND6_LLINFO_PERMANENT(ln)) { 1130 nd6_llinfo_settimer(ln, 1131 (long)ND_IFINFO(rt->rt_ifp)->reachable * hz); 1132 } 1133 } 1134 1135 void 1136 nd6_rtrequest(int req, struct rtentry *rt, const struct rt_addrinfo *info) 1137 { 1138 struct sockaddr *gate = rt->rt_gateway; 1139 struct llinfo_nd6 *ln = (struct llinfo_nd6 *)rt->rt_llinfo; 1140 struct ifnet *ifp = rt->rt_ifp; 1141 uint8_t namelen = strlen(ifp->if_xname), addrlen = ifp->if_addrlen; 1142 struct ifaddr *ifa; 1143 1144 RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key); 1145 1146 if (req == RTM_LLINFO_UPD) { 1147 int rc; 1148 struct in6_addr *in6; 1149 struct in6_addr in6_all; 1150 int anycast; 1151 1152 if ((ifa = info->rti_ifa) == NULL) 1153 return; 1154 1155 in6 = &ifatoia6(ifa)->ia_addr.sin6_addr; 1156 anycast = ifatoia6(ifa)->ia6_flags & IN6_IFF_ANYCAST; 1157 1158 in6_all = in6addr_linklocal_allnodes; 1159 if ((rc = in6_setscope(&in6_all, ifa->ifa_ifp, NULL)) != 0) { 1160 log(LOG_ERR, "%s: failed to set scope %s " 1161 "(errno=%d)\n", __func__, if_name(ifp), rc); 1162 return; 1163 } 1164 1165 /* XXX don't set Override for proxy addresses */ 1166 nd6_na_output(ifa->ifa_ifp, &in6_all, in6, 1167 (anycast ? 0 : ND_NA_FLAG_OVERRIDE) 1168 #if 0 1169 | (ip6_forwarding ? ND_NA_FLAG_ROUTER : 0) 1170 #endif 1171 , 1, NULL); 1172 return; 1173 } 1174 1175 if ((rt->rt_flags & RTF_GATEWAY) != 0) 1176 return; 1177 1178 if (nd6_need_cache(ifp) == 0 && (rt->rt_flags & RTF_HOST) == 0) { 1179 RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key); 1180 /* 1181 * This is probably an interface direct route for a link 1182 * which does not need neighbor caches (e.g. fe80::%lo0/64). 1183 * We do not need special treatment below for such a route. 1184 * Moreover, the RTF_LLINFO flag which would be set below 1185 * would annoy the ndp(8) command. 1186 */ 1187 return; 1188 } 1189 1190 if (req == RTM_RESOLVE && 1191 (nd6_need_cache(ifp) == 0 || /* stf case */ 1192 !nd6_is_addr_neighbor(satocsin6(rt_getkey(rt)), ifp))) { 1193 RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key); 1194 /* 1195 * FreeBSD and BSD/OS often make a cloned host route based 1196 * on a less-specific route (e.g. the default route). 1197 * If the less specific route does not have a "gateway" 1198 * (this is the case when the route just goes to a p2p or an 1199 * stf interface), we'll mistakenly make a neighbor cache for 1200 * the host route, and will see strange neighbor solicitation 1201 * for the corresponding destination. In order to avoid the 1202 * confusion, we check if the destination of the route is 1203 * a neighbor in terms of neighbor discovery, and stop the 1204 * process if not. Additionally, we remove the LLINFO flag 1205 * so that ndp(8) will not try to get the neighbor information 1206 * of the destination. 1207 */ 1208 rt->rt_flags &= ~RTF_LLINFO; 1209 return; 1210 } 1211 1212 switch (req) { 1213 case RTM_ADD: 1214 RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key); 1215 /* 1216 * There is no backward compatibility :) 1217 * 1218 * if ((rt->rt_flags & RTF_HOST) == 0 && 1219 * SIN(rt_mask(rt))->sin_addr.s_addr != 0xffffffff) 1220 * rt->rt_flags |= RTF_CLONING; 1221 */ 1222 if ((rt->rt_flags & RTF_CLONING) || 1223 ((rt->rt_flags & RTF_LLINFO) && ln == NULL)) { 1224 union { 1225 struct sockaddr sa; 1226 struct sockaddr_dl sdl; 1227 struct sockaddr_storage ss; 1228 } u; 1229 /* 1230 * Case 1: This route should come from a route to 1231 * interface (RTF_CLONING case) or the route should be 1232 * treated as on-link but is currently not 1233 * (RTF_LLINFO && ln == NULL case). 1234 */ 1235 sockaddr_dl_init(&u.sdl, sizeof(u.ss), 1236 ifp->if_index, ifp->if_type, 1237 NULL, namelen, NULL, addrlen); 1238 rt_setgate(rt, &u.sa); 1239 gate = rt->rt_gateway; 1240 RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key); 1241 if (ln != NULL) 1242 nd6_llinfo_settimer(ln, 0); 1243 RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key); 1244 if ((rt->rt_flags & RTF_CLONING) != 0) 1245 break; 1246 } 1247 RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key); 1248 /* 1249 * In IPv4 code, we try to annonuce new RTF_ANNOUNCE entry here. 1250 * We don't do that here since llinfo is not ready yet. 1251 * 1252 * There are also couple of other things to be discussed: 1253 * - unsolicited NA code needs improvement beforehand 1254 * - RFC2461 says we MAY send multicast unsolicited NA 1255 * (7.2.6 paragraph 4), however, it also says that we 1256 * SHOULD provide a mechanism to prevent multicast NA storm. 1257 * we don't have anything like it right now. 1258 * note that the mechanism needs a mutual agreement 1259 * between proxies, which means that we need to implement 1260 * a new protocol, or a new kludge. 1261 * - from RFC2461 6.2.4, host MUST NOT send an unsolicited NA. 1262 * we need to check ip6forwarding before sending it. 1263 * (or should we allow proxy ND configuration only for 1264 * routers? there's no mention about proxy ND from hosts) 1265 */ 1266 #if 0 1267 /* XXX it does not work */ 1268 if (rt->rt_flags & RTF_ANNOUNCE) 1269 nd6_na_output(ifp, 1270 &satocsin6(rt_getkey(rt))->sin6_addr, 1271 &satocsin6(rt_getkey(rt))->sin6_addr, 1272 ip6_forwarding ? ND_NA_FLAG_ROUTER : 0, 1273 1, NULL); 1274 #endif 1275 /* FALLTHROUGH */ 1276 case RTM_RESOLVE: 1277 if ((ifp->if_flags & (IFF_POINTOPOINT | IFF_LOOPBACK)) == 0) { 1278 RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key); 1279 /* 1280 * Address resolution isn't necessary for a point to 1281 * point link, so we can skip this test for a p2p link. 1282 */ 1283 if (gate->sa_family != AF_LINK || 1284 gate->sa_len < 1285 sockaddr_dl_measure(namelen, addrlen)) { 1286 log(LOG_DEBUG, 1287 "nd6_rtrequest: bad gateway value: %s\n", 1288 if_name(ifp)); 1289 break; 1290 } 1291 satosdl(gate)->sdl_type = ifp->if_type; 1292 satosdl(gate)->sdl_index = ifp->if_index; 1293 RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key); 1294 } 1295 if (ln != NULL) 1296 break; /* This happens on a route change */ 1297 RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key); 1298 /* 1299 * Case 2: This route may come from cloning, or a manual route 1300 * add with a LL address. 1301 */ 1302 R_Malloc(ln, struct llinfo_nd6 *, sizeof(*ln)); 1303 rt->rt_llinfo = ln; 1304 RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key); 1305 if (ln == NULL) { 1306 log(LOG_DEBUG, "nd6_rtrequest: malloc failed\n"); 1307 break; 1308 } 1309 RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key); 1310 nd6_inuse++; 1311 nd6_allocated++; 1312 memset(ln, 0, sizeof(*ln)); 1313 ln->ln_rt = rt; 1314 callout_init(&ln->ln_timer_ch, CALLOUT_MPSAFE); 1315 /* this is required for "ndp" command. - shin */ 1316 if (req == RTM_ADD) { 1317 /* 1318 * gate should have some valid AF_LINK entry, 1319 * and ln->ln_expire should have some lifetime 1320 * which is specified by ndp command. 1321 */ 1322 ln->ln_state = ND6_LLINFO_REACHABLE; 1323 ln->ln_byhint = 0; 1324 } else { 1325 /* 1326 * When req == RTM_RESOLVE, rt is created and 1327 * initialized in rtrequest(), so rt_expire is 0. 1328 */ 1329 ln->ln_state = ND6_LLINFO_NOSTATE; 1330 nd6_llinfo_settimer(ln, 0); 1331 } 1332 RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key); 1333 rt->rt_flags |= RTF_LLINFO; 1334 ln->ln_next = llinfo_nd6.ln_next; 1335 llinfo_nd6.ln_next = ln; 1336 ln->ln_prev = &llinfo_nd6; 1337 ln->ln_next->ln_prev = ln; 1338 1339 RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key); 1340 /* 1341 * check if rt_getkey(rt) is an address assigned 1342 * to the interface. 1343 */ 1344 ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp, 1345 &satocsin6(rt_getkey(rt))->sin6_addr); 1346 RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key); 1347 if (ifa != NULL) { 1348 const void *mac; 1349 nd6_llinfo_settimer(ln, -1); 1350 ln->ln_state = ND6_LLINFO_REACHABLE; 1351 ln->ln_byhint = 0; 1352 if ((mac = nd6_ifptomac(ifp)) != NULL) { 1353 /* XXX check for error */ 1354 (void)sockaddr_dl_setaddr(satosdl(gate), 1355 gate->sa_len, mac, ifp->if_addrlen); 1356 } 1357 if (nd6_useloopback) { 1358 ifp = rt->rt_ifp = lo0ifp; /* XXX */ 1359 /* 1360 * Make sure rt_ifa be equal to the ifaddr 1361 * corresponding to the address. 1362 * We need this because when we refer 1363 * rt_ifa->ia6_flags in ip6_input, we assume 1364 * that the rt_ifa points to the address instead 1365 * of the loopback address. 1366 */ 1367 if (ifa != rt->rt_ifa) 1368 rt_replace_ifa(rt, ifa); 1369 rt->rt_flags &= ~RTF_CLONED; 1370 } 1371 } else if (rt->rt_flags & RTF_ANNOUNCE) { 1372 nd6_llinfo_settimer(ln, -1); 1373 ln->ln_state = ND6_LLINFO_REACHABLE; 1374 ln->ln_byhint = 0; 1375 1376 /* join solicited node multicast for proxy ND */ 1377 if (ifp->if_flags & IFF_MULTICAST) { 1378 struct in6_addr llsol; 1379 int error; 1380 1381 llsol = satocsin6(rt_getkey(rt))->sin6_addr; 1382 llsol.s6_addr32[0] = htonl(0xff020000); 1383 llsol.s6_addr32[1] = 0; 1384 llsol.s6_addr32[2] = htonl(1); 1385 llsol.s6_addr8[12] = 0xff; 1386 if (in6_setscope(&llsol, ifp, NULL)) 1387 break; 1388 if (!in6_addmulti(&llsol, ifp, &error, 0)) { 1389 nd6log((LOG_ERR, "%s: failed to join " 1390 "%s (errno=%d)\n", if_name(ifp), 1391 ip6_sprintf(&llsol), error)); 1392 } 1393 } 1394 } 1395 break; 1396 1397 case RTM_DELETE: 1398 if (ln == NULL) 1399 break; 1400 /* leave from solicited node multicast for proxy ND */ 1401 if ((rt->rt_flags & RTF_ANNOUNCE) != 0 && 1402 (ifp->if_flags & IFF_MULTICAST) != 0) { 1403 struct in6_addr llsol; 1404 struct in6_multi *in6m; 1405 1406 llsol = satocsin6(rt_getkey(rt))->sin6_addr; 1407 llsol.s6_addr32[0] = htonl(0xff020000); 1408 llsol.s6_addr32[1] = 0; 1409 llsol.s6_addr32[2] = htonl(1); 1410 llsol.s6_addr8[12] = 0xff; 1411 if (in6_setscope(&llsol, ifp, NULL) == 0) { 1412 IN6_LOOKUP_MULTI(llsol, ifp, in6m); 1413 if (in6m) 1414 in6_delmulti(in6m); 1415 } 1416 } 1417 nd6_inuse--; 1418 ln->ln_next->ln_prev = ln->ln_prev; 1419 ln->ln_prev->ln_next = ln->ln_next; 1420 ln->ln_prev = NULL; 1421 nd6_llinfo_settimer(ln, -1); 1422 rt->rt_llinfo = 0; 1423 rt->rt_flags &= ~RTF_LLINFO; 1424 clear_llinfo_pqueue(ln); 1425 Free(ln); 1426 } 1427 } 1428 1429 int 1430 nd6_ioctl(u_long cmd, void *data, struct ifnet *ifp) 1431 { 1432 struct in6_drlist *drl = (struct in6_drlist *)data; 1433 struct in6_oprlist *oprl = (struct in6_oprlist *)data; 1434 struct in6_ndireq *ndi = (struct in6_ndireq *)data; 1435 struct in6_nbrinfo *nbi = (struct in6_nbrinfo *)data; 1436 struct in6_ndifreq *ndif = (struct in6_ndifreq *)data; 1437 struct nd_defrouter *dr; 1438 struct nd_prefix *pr; 1439 struct rtentry *rt; 1440 int i = 0, error = 0; 1441 int s; 1442 1443 switch (cmd) { 1444 case SIOCGDRLST_IN6: 1445 /* 1446 * obsolete API, use sysctl under net.inet6.icmp6 1447 */ 1448 memset(drl, 0, sizeof(*drl)); 1449 s = splsoftnet(); 1450 TAILQ_FOREACH(dr, &nd_defrouter, dr_entry) { 1451 if (i >= DRLSTSIZ) 1452 break; 1453 drl->defrouter[i].rtaddr = dr->rtaddr; 1454 in6_clearscope(&drl->defrouter[i].rtaddr); 1455 1456 drl->defrouter[i].flags = dr->flags; 1457 drl->defrouter[i].rtlifetime = dr->rtlifetime; 1458 drl->defrouter[i].expire = dr->expire; 1459 drl->defrouter[i].if_index = dr->ifp->if_index; 1460 i++; 1461 } 1462 splx(s); 1463 break; 1464 case SIOCGPRLST_IN6: 1465 /* 1466 * obsolete API, use sysctl under net.inet6.icmp6 1467 * 1468 * XXX the structure in6_prlist was changed in backward- 1469 * incompatible manner. in6_oprlist is used for SIOCGPRLST_IN6, 1470 * in6_prlist is used for nd6_sysctl() - fill_prlist(). 1471 */ 1472 /* 1473 * XXX meaning of fields, especialy "raflags", is very 1474 * differnet between RA prefix list and RR/static prefix list. 1475 * how about separating ioctls into two? 1476 */ 1477 memset(oprl, 0, sizeof(*oprl)); 1478 s = splsoftnet(); 1479 LIST_FOREACH(pr, &nd_prefix, ndpr_entry) { 1480 struct nd_pfxrouter *pfr; 1481 int j; 1482 1483 if (i >= PRLSTSIZ) 1484 break; 1485 oprl->prefix[i].prefix = pr->ndpr_prefix.sin6_addr; 1486 oprl->prefix[i].raflags = pr->ndpr_raf; 1487 oprl->prefix[i].prefixlen = pr->ndpr_plen; 1488 oprl->prefix[i].vltime = pr->ndpr_vltime; 1489 oprl->prefix[i].pltime = pr->ndpr_pltime; 1490 oprl->prefix[i].if_index = pr->ndpr_ifp->if_index; 1491 if (pr->ndpr_vltime == ND6_INFINITE_LIFETIME) 1492 oprl->prefix[i].expire = 0; 1493 else { 1494 time_t maxexpire; 1495 1496 /* XXX: we assume time_t is signed. */ 1497 maxexpire = (-1) & 1498 ~((time_t)1 << 1499 ((sizeof(maxexpire) * 8) - 1)); 1500 if (pr->ndpr_vltime < 1501 maxexpire - pr->ndpr_lastupdate) { 1502 oprl->prefix[i].expire = 1503 pr->ndpr_lastupdate + 1504 pr->ndpr_vltime; 1505 } else 1506 oprl->prefix[i].expire = maxexpire; 1507 } 1508 1509 j = 0; 1510 LIST_FOREACH(pfr, &pr->ndpr_advrtrs, pfr_entry) { 1511 if (j < DRLSTSIZ) { 1512 #define RTRADDR oprl->prefix[i].advrtr[j] 1513 RTRADDR = pfr->router->rtaddr; 1514 in6_clearscope(&RTRADDR); 1515 #undef RTRADDR 1516 } 1517 j++; 1518 } 1519 oprl->prefix[i].advrtrs = j; 1520 oprl->prefix[i].origin = PR_ORIG_RA; 1521 1522 i++; 1523 } 1524 splx(s); 1525 1526 break; 1527 case OSIOCGIFINFO_IN6: 1528 #define ND ndi->ndi 1529 /* XXX: old ndp(8) assumes a positive value for linkmtu. */ 1530 memset(&ND, 0, sizeof(ND)); 1531 ND.linkmtu = IN6_LINKMTU(ifp); 1532 ND.maxmtu = ND_IFINFO(ifp)->maxmtu; 1533 ND.basereachable = ND_IFINFO(ifp)->basereachable; 1534 ND.reachable = ND_IFINFO(ifp)->reachable; 1535 ND.retrans = ND_IFINFO(ifp)->retrans; 1536 ND.flags = ND_IFINFO(ifp)->flags; 1537 ND.recalctm = ND_IFINFO(ifp)->recalctm; 1538 ND.chlim = ND_IFINFO(ifp)->chlim; 1539 break; 1540 case SIOCGIFINFO_IN6: 1541 ND = *ND_IFINFO(ifp); 1542 break; 1543 case SIOCSIFINFO_IN6: 1544 /* 1545 * used to change host variables from userland. 1546 * intented for a use on router to reflect RA configurations. 1547 */ 1548 /* 0 means 'unspecified' */ 1549 if (ND.linkmtu != 0) { 1550 if (ND.linkmtu < IPV6_MMTU || 1551 ND.linkmtu > IN6_LINKMTU(ifp)) { 1552 error = EINVAL; 1553 break; 1554 } 1555 ND_IFINFO(ifp)->linkmtu = ND.linkmtu; 1556 } 1557 1558 if (ND.basereachable != 0) { 1559 int obasereachable = ND_IFINFO(ifp)->basereachable; 1560 1561 ND_IFINFO(ifp)->basereachable = ND.basereachable; 1562 if (ND.basereachable != obasereachable) 1563 ND_IFINFO(ifp)->reachable = 1564 ND_COMPUTE_RTIME(ND.basereachable); 1565 } 1566 if (ND.retrans != 0) 1567 ND_IFINFO(ifp)->retrans = ND.retrans; 1568 if (ND.chlim != 0) 1569 ND_IFINFO(ifp)->chlim = ND.chlim; 1570 /* FALLTHROUGH */ 1571 case SIOCSIFINFO_FLAGS: 1572 ND_IFINFO(ifp)->flags = ND.flags; 1573 break; 1574 #undef ND 1575 case SIOCSNDFLUSH_IN6: /* XXX: the ioctl name is confusing... */ 1576 /* sync kernel routing table with the default router list */ 1577 defrouter_reset(); 1578 defrouter_select(); 1579 break; 1580 case SIOCSPFXFLUSH_IN6: 1581 { 1582 /* flush all the prefix advertised by routers */ 1583 struct nd_prefix *pfx, *next; 1584 1585 s = splsoftnet(); 1586 for (pfx = LIST_FIRST(&nd_prefix); pfx; pfx = next) { 1587 struct in6_ifaddr *ia, *ia_next; 1588 1589 next = LIST_NEXT(pfx, ndpr_entry); 1590 1591 if (IN6_IS_ADDR_LINKLOCAL(&pfx->ndpr_prefix.sin6_addr)) 1592 continue; /* XXX */ 1593 1594 /* do we really have to remove addresses as well? */ 1595 for (ia = in6_ifaddr; ia; ia = ia_next) { 1596 /* ia might be removed. keep the next ptr. */ 1597 ia_next = ia->ia_next; 1598 1599 if ((ia->ia6_flags & IN6_IFF_AUTOCONF) == 0) 1600 continue; 1601 1602 if (ia->ia6_ndpr == pfx) 1603 in6_purgeaddr(&ia->ia_ifa); 1604 } 1605 prelist_remove(pfx); 1606 } 1607 splx(s); 1608 break; 1609 } 1610 case SIOCSRTRFLUSH_IN6: 1611 { 1612 /* flush all the default routers */ 1613 struct nd_defrouter *drtr, *next; 1614 1615 s = splsoftnet(); 1616 defrouter_reset(); 1617 for (drtr = TAILQ_FIRST(&nd_defrouter); drtr; drtr = next) { 1618 next = TAILQ_NEXT(drtr, dr_entry); 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 (void)sockaddr_dl_setaddr(sdl, sdl->sdl_len, lladdr, 1781 ifp->if_addrlen); 1782 } 1783 1784 if (!is_newentry) { 1785 if ((!olladdr && lladdr) || /* (3) */ 1786 (olladdr && lladdr && llchange)) { /* (5) */ 1787 do_update = 1; 1788 newstate = ND6_LLINFO_STALE; 1789 } else /* (1-2,4) */ 1790 do_update = 0; 1791 } else { 1792 do_update = 1; 1793 if (lladdr == NULL) /* (6) */ 1794 newstate = ND6_LLINFO_NOSTATE; 1795 else /* (7) */ 1796 newstate = ND6_LLINFO_STALE; 1797 } 1798 1799 if (do_update) { 1800 /* 1801 * Update the state of the neighbor cache. 1802 */ 1803 ln->ln_state = newstate; 1804 1805 if (ln->ln_state == ND6_LLINFO_STALE) { 1806 /* 1807 * XXX: since nd6_output() below will cause 1808 * state tansition to DELAY and reset the timer, 1809 * we must set the timer now, although it is actually 1810 * meaningless. 1811 */ 1812 nd6_llinfo_settimer(ln, (long)nd6_gctimer * hz); 1813 1814 nd6_llinfo_release_pkts(ln, ifp, rt); 1815 } else if (ln->ln_state == ND6_LLINFO_INCOMPLETE) { 1816 /* probe right away */ 1817 nd6_llinfo_settimer((void *)ln, 0); 1818 } 1819 } 1820 1821 /* 1822 * ICMP6 type dependent behavior. 1823 * 1824 * NS: clear IsRouter if new entry 1825 * RS: clear IsRouter 1826 * RA: set IsRouter if there's lladdr 1827 * redir: clear IsRouter if new entry 1828 * 1829 * RA case, (1): 1830 * The spec says that we must set IsRouter in the following cases: 1831 * - If lladdr exist, set IsRouter. This means (1-5). 1832 * - If it is old entry (!newentry), set IsRouter. This means (7). 1833 * So, based on the spec, in (1-5) and (7) cases we must set IsRouter. 1834 * A quetion arises for (1) case. (1) case has no lladdr in the 1835 * neighbor cache, this is similar to (6). 1836 * This case is rare but we figured that we MUST NOT set IsRouter. 1837 * 1838 * newentry olladdr lladdr llchange NS RS RA redir 1839 * D R 1840 * 0 n n -- (1) c ? s 1841 * 0 y n -- (2) c s s 1842 * 0 n y -- (3) c s s 1843 * 0 y y n (4) c s s 1844 * 0 y y y (5) c s s 1845 * 1 -- n -- (6) c c c s 1846 * 1 -- y -- (7) c c s c s 1847 * 1848 * (c=clear s=set) 1849 */ 1850 switch (type & 0xff) { 1851 case ND_NEIGHBOR_SOLICIT: 1852 /* 1853 * New entry must have is_router flag cleared. 1854 */ 1855 if (is_newentry) /* (6-7) */ 1856 ln->ln_router = 0; 1857 break; 1858 case ND_REDIRECT: 1859 /* 1860 * If the icmp is a redirect to a better router, always set the 1861 * is_router flag. Otherwise, if the entry is newly created, 1862 * clear the flag. [RFC 2461, sec 8.3] 1863 */ 1864 if (code == ND_REDIRECT_ROUTER) 1865 ln->ln_router = 1; 1866 else if (is_newentry) /* (6-7) */ 1867 ln->ln_router = 0; 1868 break; 1869 case ND_ROUTER_SOLICIT: 1870 /* 1871 * is_router flag must always be cleared. 1872 */ 1873 ln->ln_router = 0; 1874 break; 1875 case ND_ROUTER_ADVERT: 1876 /* 1877 * Mark an entry with lladdr as a router. 1878 */ 1879 if ((!is_newentry && (olladdr || lladdr)) || /* (2-5) */ 1880 (is_newentry && lladdr)) { /* (7) */ 1881 ln->ln_router = 1; 1882 } 1883 break; 1884 } 1885 1886 /* 1887 * When the link-layer address of a router changes, select the 1888 * best router again. In particular, when the neighbor entry is newly 1889 * created, it might affect the selection policy. 1890 * Question: can we restrict the first condition to the "is_newentry" 1891 * case? 1892 * XXX: when we hear an RA from a new router with the link-layer 1893 * address option, defrouter_select() is called twice, since 1894 * defrtrlist_update called the function as well. However, I believe 1895 * we can compromise the overhead, since it only happens the first 1896 * time. 1897 * XXX: although defrouter_select() should not have a bad effect 1898 * for those are not autoconfigured hosts, we explicitly avoid such 1899 * cases for safety. 1900 */ 1901 if (do_update && ln->ln_router && !ip6_forwarding && 1902 nd6_accepts_rtadv(ndi)) 1903 defrouter_select(); 1904 1905 return rt; 1906 } 1907 1908 static void 1909 nd6_slowtimo(void *ignored_arg) 1910 { 1911 struct nd_ifinfo *nd6if; 1912 struct ifnet *ifp; 1913 1914 mutex_enter(softnet_lock); 1915 KERNEL_LOCK(1, NULL); 1916 callout_reset(&nd6_slowtimo_ch, ND6_SLOWTIMER_INTERVAL * hz, 1917 nd6_slowtimo, NULL); 1918 TAILQ_FOREACH(ifp, &ifnet, if_list) { 1919 nd6if = ND_IFINFO(ifp); 1920 if (nd6if->basereachable && /* already initialized */ 1921 (nd6if->recalctm -= ND6_SLOWTIMER_INTERVAL) <= 0) { 1922 /* 1923 * Since reachable time rarely changes by router 1924 * advertisements, we SHOULD insure that a new random 1925 * value gets recomputed at least once every few hours. 1926 * (RFC 2461, 6.3.4) 1927 */ 1928 nd6if->recalctm = nd6_recalc_reachtm_interval; 1929 nd6if->reachable = ND_COMPUTE_RTIME(nd6if->basereachable); 1930 } 1931 } 1932 KERNEL_UNLOCK_ONE(NULL); 1933 mutex_exit(softnet_lock); 1934 } 1935 1936 #define senderr(e) { error = (e); goto bad;} 1937 int 1938 nd6_output(struct ifnet *ifp, struct ifnet *origifp, struct mbuf *m0, 1939 const struct sockaddr_in6 *dst, struct rtentry *rt0) 1940 { 1941 struct mbuf *m = m0; 1942 struct rtentry *rt = rt0; 1943 struct sockaddr_in6 *gw6 = NULL; 1944 struct llinfo_nd6 *ln = NULL; 1945 int error = 0; 1946 1947 if (IN6_IS_ADDR_MULTICAST(&dst->sin6_addr)) 1948 goto sendpkt; 1949 1950 if (nd6_need_cache(ifp) == 0) 1951 goto sendpkt; 1952 1953 /* 1954 * next hop determination. This routine is derived from ether_output. 1955 */ 1956 if (rt) { 1957 if ((rt->rt_flags & RTF_UP) == 0) { 1958 if ((rt0 = rt = rtalloc1(sin6tocsa(dst), 1)) != NULL) { 1959 rt->rt_refcnt--; 1960 if (rt->rt_ifp != ifp) 1961 senderr(EHOSTUNREACH); 1962 } else 1963 senderr(EHOSTUNREACH); 1964 } 1965 1966 if (rt->rt_flags & RTF_GATEWAY) { 1967 gw6 = (struct sockaddr_in6 *)rt->rt_gateway; 1968 1969 /* 1970 * We skip link-layer address resolution and NUD 1971 * if the gateway is not a neighbor from ND point 1972 * of view, regardless of the value of nd_ifinfo.flags. 1973 * The second condition is a bit tricky; we skip 1974 * if the gateway is our own address, which is 1975 * sometimes used to install a route to a p2p link. 1976 */ 1977 if (!nd6_is_addr_neighbor(gw6, ifp) || 1978 in6ifa_ifpwithaddr(ifp, &gw6->sin6_addr)) { 1979 /* 1980 * We allow this kind of tricky route only 1981 * when the outgoing interface is p2p. 1982 * XXX: we may need a more generic rule here. 1983 */ 1984 if ((ifp->if_flags & IFF_POINTOPOINT) == 0) 1985 senderr(EHOSTUNREACH); 1986 1987 goto sendpkt; 1988 } 1989 1990 if (rt->rt_gwroute == NULL) 1991 goto lookup; 1992 if (((rt = rt->rt_gwroute)->rt_flags & RTF_UP) == 0) { 1993 rtfree(rt); rt = rt0; 1994 lookup: 1995 rt->rt_gwroute = rtalloc1(rt->rt_gateway, 1); 1996 if ((rt = rt->rt_gwroute) == NULL) 1997 senderr(EHOSTUNREACH); 1998 /* the "G" test below also prevents rt == rt0 */ 1999 if ((rt->rt_flags & RTF_GATEWAY) || 2000 (rt->rt_ifp != ifp)) { 2001 rt->rt_refcnt--; 2002 rt0->rt_gwroute = NULL; 2003 senderr(EHOSTUNREACH); 2004 } 2005 } 2006 } 2007 } 2008 2009 /* 2010 * Address resolution or Neighbor Unreachability Detection 2011 * for the next hop. 2012 * At this point, the destination of the packet must be a unicast 2013 * or an anycast address(i.e. not a multicast). 2014 */ 2015 2016 /* Look up the neighbor cache for the nexthop */ 2017 if (rt != NULL && (rt->rt_flags & RTF_LLINFO) != 0) 2018 ln = (struct llinfo_nd6 *)rt->rt_llinfo; 2019 else { 2020 /* 2021 * Since nd6_is_addr_neighbor() internally calls nd6_lookup(), 2022 * the condition below is not very efficient. But we believe 2023 * it is tolerable, because this should be a rare case. 2024 */ 2025 if (nd6_is_addr_neighbor(dst, ifp) && 2026 (rt = nd6_lookup(&dst->sin6_addr, 1, ifp)) != NULL) 2027 ln = (struct llinfo_nd6 *)rt->rt_llinfo; 2028 } 2029 if (ln == NULL || rt == NULL) { 2030 if ((ifp->if_flags & IFF_POINTOPOINT) == 0 && 2031 !(ND_IFINFO(ifp)->flags & ND6_IFF_PERFORMNUD)) { 2032 log(LOG_DEBUG, 2033 "nd6_output: can't allocate llinfo for %s " 2034 "(ln=%p, rt=%p)\n", 2035 ip6_sprintf(&dst->sin6_addr), ln, rt); 2036 senderr(EIO); /* XXX: good error? */ 2037 } 2038 2039 goto sendpkt; /* send anyway */ 2040 } 2041 2042 /* We don't have to do link-layer address resolution on a p2p link. */ 2043 if ((ifp->if_flags & IFF_POINTOPOINT) != 0 && 2044 ln->ln_state < ND6_LLINFO_REACHABLE) { 2045 ln->ln_state = ND6_LLINFO_STALE; 2046 nd6_llinfo_settimer(ln, (long)nd6_gctimer * hz); 2047 } 2048 2049 /* 2050 * The first time we send a packet to a neighbor whose entry is 2051 * STALE, we have to change the state to DELAY and a sets a timer to 2052 * expire in DELAY_FIRST_PROBE_TIME seconds to ensure do 2053 * neighbor unreachability detection on expiration. 2054 * (RFC 2461 7.3.3) 2055 */ 2056 if (ln->ln_state == ND6_LLINFO_STALE) { 2057 ln->ln_asked = 0; 2058 ln->ln_state = ND6_LLINFO_DELAY; 2059 nd6_llinfo_settimer(ln, (long)nd6_delay * hz); 2060 } 2061 2062 /* 2063 * If the neighbor cache entry has a state other than INCOMPLETE 2064 * (i.e. its link-layer address is already resolved), just 2065 * send the packet. 2066 */ 2067 if (ln->ln_state > ND6_LLINFO_INCOMPLETE) 2068 goto sendpkt; 2069 2070 /* 2071 * There is a neighbor cache entry, but no ethernet address 2072 * response yet. Append this latest packet to the end of the 2073 * packet queue in the mbuf, unless the number of the packet 2074 * does not exceed nd6_maxqueuelen. When it exceeds nd6_maxqueuelen, 2075 * the oldest packet in the queue will be removed. 2076 */ 2077 if (ln->ln_state == ND6_LLINFO_NOSTATE) 2078 ln->ln_state = ND6_LLINFO_INCOMPLETE; 2079 if (ln->ln_hold) { 2080 struct mbuf *m_hold; 2081 int i; 2082 2083 i = 0; 2084 for (m_hold = ln->ln_hold; m_hold; m_hold = m_hold->m_nextpkt) { 2085 i++; 2086 if (m_hold->m_nextpkt == NULL) { 2087 m_hold->m_nextpkt = m; 2088 break; 2089 } 2090 } 2091 while (i >= nd6_maxqueuelen) { 2092 m_hold = ln->ln_hold; 2093 ln->ln_hold = ln->ln_hold->m_nextpkt; 2094 m_freem(m_hold); 2095 i--; 2096 } 2097 } else { 2098 ln->ln_hold = m; 2099 } 2100 2101 /* 2102 * If there has been no NS for the neighbor after entering the 2103 * INCOMPLETE state, send the first solicitation. 2104 */ 2105 if (!ND6_LLINFO_PERMANENT(ln) && ln->ln_asked == 0) { 2106 ln->ln_asked++; 2107 nd6_llinfo_settimer(ln, 2108 (long)ND_IFINFO(ifp)->retrans * hz / 1000); 2109 nd6_ns_output(ifp, NULL, &dst->sin6_addr, ln, 0); 2110 } 2111 return 0; 2112 2113 sendpkt: 2114 /* discard the packet if IPv6 operation is disabled on the interface */ 2115 if ((ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED)) { 2116 error = ENETDOWN; /* better error? */ 2117 goto bad; 2118 } 2119 2120 #ifdef IPSEC 2121 /* clean ipsec history once it goes out of the node */ 2122 ipsec_delaux(m); 2123 #endif 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 == 0, dst=%s, if=%s\n", __func__, 2201 ip6_sprintf(&satocsin6(dst)->sin6_addr), if_name(ifp)); 2202 m_freem(m); 2203 return 0; 2204 } 2205 2206 memcpy(lldst, CLLADDR(sdl), MIN(dstsize, sdl->sdl_alen)); 2207 return 1; 2208 } 2209 2210 static void 2211 clear_llinfo_pqueue(struct llinfo_nd6 *ln) 2212 { 2213 struct mbuf *m_hold, *m_hold_next; 2214 2215 for (m_hold = ln->ln_hold; m_hold; m_hold = m_hold_next) { 2216 m_hold_next = m_hold->m_nextpkt; 2217 m_hold->m_nextpkt = NULL; 2218 m_freem(m_hold); 2219 } 2220 2221 ln->ln_hold = NULL; 2222 return; 2223 } 2224 2225 int 2226 nd6_sysctl( 2227 int name, 2228 void *oldp, /* syscall arg, need copyout */ 2229 size_t *oldlenp, 2230 void *newp, /* syscall arg, need copyin */ 2231 size_t newlen 2232 ) 2233 { 2234 void *p; 2235 size_t ol; 2236 int error; 2237 2238 error = 0; 2239 2240 if (newp) 2241 return EPERM; 2242 if (oldp && !oldlenp) 2243 return EINVAL; 2244 ol = oldlenp ? *oldlenp : 0; 2245 2246 if (oldp) { 2247 p = malloc(*oldlenp, M_TEMP, M_WAITOK); 2248 if (p == NULL) 2249 return ENOMEM; 2250 } else 2251 p = NULL; 2252 switch (name) { 2253 case ICMPV6CTL_ND6_DRLIST: 2254 error = fill_drlist(p, oldlenp, ol); 2255 if (!error && p != NULL && oldp != NULL) 2256 error = copyout(p, oldp, *oldlenp); 2257 break; 2258 2259 case ICMPV6CTL_ND6_PRLIST: 2260 error = fill_prlist(p, oldlenp, ol); 2261 if (!error && p != NULL && oldp != NULL) 2262 error = copyout(p, oldp, *oldlenp); 2263 break; 2264 2265 case ICMPV6CTL_ND6_MAXQLEN: 2266 break; 2267 2268 default: 2269 error = ENOPROTOOPT; 2270 break; 2271 } 2272 if (p) 2273 free(p, M_TEMP); 2274 2275 return error; 2276 } 2277 2278 static int 2279 fill_drlist(void *oldp, size_t *oldlenp, size_t ol) 2280 { 2281 int error = 0, s; 2282 struct in6_defrouter *d = NULL, *de = NULL; 2283 struct nd_defrouter *dr; 2284 size_t l; 2285 2286 s = splsoftnet(); 2287 2288 if (oldp) { 2289 d = (struct in6_defrouter *)oldp; 2290 de = (struct in6_defrouter *)((char *)oldp + *oldlenp); 2291 } 2292 l = 0; 2293 2294 TAILQ_FOREACH(dr, &nd_defrouter, dr_entry) { 2295 2296 if (oldp && d + 1 <= de) { 2297 memset(d, 0, sizeof(*d)); 2298 sockaddr_in6_init(&d->rtaddr, &dr->rtaddr, 0, 0, 0); 2299 if (sa6_recoverscope(&d->rtaddr)) { 2300 log(LOG_ERR, 2301 "scope error in router list (%s)\n", 2302 ip6_sprintf(&d->rtaddr.sin6_addr)); 2303 /* XXX: press on... */ 2304 } 2305 d->flags = dr->flags; 2306 d->rtlifetime = dr->rtlifetime; 2307 d->expire = dr->expire; 2308 d->if_index = dr->ifp->if_index; 2309 } 2310 2311 l += sizeof(*d); 2312 if (d) 2313 d++; 2314 } 2315 2316 if (oldp) { 2317 if (l > ol) 2318 error = ENOMEM; 2319 } 2320 if (oldlenp) 2321 *oldlenp = l; /* (void *)d - (void *)oldp */ 2322 2323 splx(s); 2324 2325 return error; 2326 } 2327 2328 static int 2329 fill_prlist(void *oldp, size_t *oldlenp, size_t ol) 2330 { 2331 int error = 0, s; 2332 struct nd_prefix *pr; 2333 struct in6_prefix *p = NULL; 2334 struct in6_prefix *pe = NULL; 2335 size_t l; 2336 2337 s = splsoftnet(); 2338 2339 if (oldp) { 2340 p = (struct in6_prefix *)oldp; 2341 pe = (struct in6_prefix *)((char *)oldp + *oldlenp); 2342 } 2343 l = 0; 2344 2345 LIST_FOREACH(pr, &nd_prefix, ndpr_entry) { 2346 u_short advrtrs; 2347 size_t advance; 2348 struct sockaddr_in6 *sin6; 2349 struct sockaddr_in6 *s6; 2350 struct nd_pfxrouter *pfr; 2351 2352 if (oldp && p + 1 <= pe) 2353 { 2354 memset(p, 0, sizeof(*p)); 2355 sin6 = (struct sockaddr_in6 *)(p + 1); 2356 2357 p->prefix = pr->ndpr_prefix; 2358 if (sa6_recoverscope(&p->prefix)) { 2359 log(LOG_ERR, 2360 "scope error in prefix list (%s)\n", 2361 ip6_sprintf(&p->prefix.sin6_addr)); 2362 /* XXX: press on... */ 2363 } 2364 p->raflags = pr->ndpr_raf; 2365 p->prefixlen = pr->ndpr_plen; 2366 p->vltime = pr->ndpr_vltime; 2367 p->pltime = pr->ndpr_pltime; 2368 p->if_index = pr->ndpr_ifp->if_index; 2369 if (pr->ndpr_vltime == ND6_INFINITE_LIFETIME) 2370 p->expire = 0; 2371 else { 2372 time_t maxexpire; 2373 2374 /* XXX: we assume time_t is signed. */ 2375 maxexpire = (-1) & 2376 ~((time_t)1 << 2377 ((sizeof(maxexpire) * 8) - 1)); 2378 if (pr->ndpr_vltime < 2379 maxexpire - pr->ndpr_lastupdate) { 2380 p->expire = pr->ndpr_lastupdate + 2381 pr->ndpr_vltime; 2382 } else 2383 p->expire = maxexpire; 2384 } 2385 p->refcnt = pr->ndpr_refcnt; 2386 p->flags = pr->ndpr_stateflags; 2387 p->origin = PR_ORIG_RA; 2388 advrtrs = 0; 2389 LIST_FOREACH(pfr, &pr->ndpr_advrtrs, pfr_entry) { 2390 if ((void *)&sin6[advrtrs + 1] > (void *)pe) { 2391 advrtrs++; 2392 continue; 2393 } 2394 s6 = &sin6[advrtrs]; 2395 sockaddr_in6_init(s6, &pfr->router->rtaddr, 2396 0, 0, 0); 2397 if (sa6_recoverscope(s6)) { 2398 log(LOG_ERR, 2399 "scope error in " 2400 "prefix list (%s)\n", 2401 ip6_sprintf(&pfr->router->rtaddr)); 2402 } 2403 advrtrs++; 2404 } 2405 p->advrtrs = advrtrs; 2406 } 2407 else { 2408 advrtrs = 0; 2409 LIST_FOREACH(pfr, &pr->ndpr_advrtrs, pfr_entry) 2410 advrtrs++; 2411 } 2412 2413 advance = sizeof(*p) + sizeof(*sin6) * advrtrs; 2414 l += advance; 2415 if (p) 2416 p = (struct in6_prefix *)((char *)p + advance); 2417 } 2418 2419 if (oldp) { 2420 *oldlenp = l; /* (void *)d - (void *)oldp */ 2421 if (l > ol) 2422 error = ENOMEM; 2423 } else 2424 *oldlenp = l; 2425 2426 splx(s); 2427 2428 return error; 2429 } 2430