1/* $NetBSD: nd6.c,v 1.141.8.2 2013/12/17 20:47:49 bouyer Exp $ */ 2/* $KAME: nd6.c,v 1.279 2002/06/08 11:16:51 itojun Exp $ */ 3 4/* 5 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project. 6 * All rights reserved. 7 * 8 * Redistribution and use in source and binary forms, with or without 9 * modification, are permitted provided that the following conditions 10 * are met: 11 * 1. Redistributions of source code must retain the above copyright 12 * notice, this list of conditions and the following disclaimer. 13 * 2. Redistributions in binary form must reproduce the above copyright 14 * notice, this list of conditions and the following disclaimer in the 15 * documentation and/or other materials provided with the distribution. 16 * 3. Neither the name of the project nor the names of its contributors 17 * may be used to endorse or promote products derived from this software 18 * without specific prior written permission. 19 * 20 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND 21 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 22 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 23 * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE 24 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 25 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 26 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 27 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 28 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 29 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 30 * SUCH DAMAGE. 31 */ 32 33#include <sys/cdefs.h> 34__KERNEL_RCSID(0, "$NetBSD: nd6.c,v 1.141.8.2 2013/12/17 20:47:49 bouyer Exp $"); 35 36#include "opt_ipsec.h" 37 38#include <sys/param.h> 39#include <sys/systm.h> 40#include <sys/callout.h> 41#include <sys/malloc.h> 42#include <sys/mbuf.h> 43#include <sys/socket.h> 44#include <sys/socketvar.h> 45#include <sys/sockio.h> 46#include <sys/time.h> 47#include <sys/kernel.h> 48#include <sys/protosw.h> 49#include <sys/errno.h> 50#include <sys/ioctl.h> 51#include <sys/syslog.h> 52#include <sys/queue.h> 53#include <sys/cprng.h> 54 55#include <net/if.h> 56#include <net/if_dl.h> 57#include <net/if_types.h> 58#include <net/route.h> 59#include <net/if_ether.h> 60#include <net/if_fddi.h> 61#include <net/if_arc.h> 62 63#include <netinet/in.h> 64#include <netinet6/in6_var.h> 65#include <netinet/ip6.h> 66#include <netinet6/ip6_var.h> 67#include <netinet6/scope6_var.h> 68#include <netinet6/nd6.h> 69#include <netinet/icmp6.h> 70#include <netinet6/icmp6_private.h> 71 72#ifdef KAME_IPSEC 73#include <netinet6/ipsec.h> 74#endif 75 76#include <net/net_osdep.h> 77 78#define ND6_SLOWTIMER_INTERVAL (60 * 60) /* 1 hour */ 79#define ND6_RECALC_REACHTM_INTERVAL (60 * 120) /* 2 hours */ 80 81/* timer values */ 82int nd6_prune = 1; /* walk list every 1 seconds */ 83int nd6_delay = 5; /* delay first probe time 5 second */ 84int nd6_umaxtries = 3; /* maximum unicast query */ 85int nd6_mmaxtries = 3; /* maximum multicast query */ 86int nd6_useloopback = 1; /* use loopback interface for local traffic */ 87int nd6_gctimer = (60 * 60 * 24); /* 1 day: garbage collection timer */ 88 89/* preventing too many loops in ND option parsing */ 90int nd6_maxndopt = 10; /* max # of ND options allowed */ 91 92int nd6_maxnudhint = 0; /* max # of subsequent upper layer hints */ 93 94int nd6_maxqueuelen = 1; /* max # of packets cached in unresolved ND entries */ 95 96#ifdef ND6_DEBUG 97int nd6_debug = 1; 98#else 99int nd6_debug = 0; 100#endif 101 102/* for debugging? */ 103static int nd6_inuse, nd6_allocated; 104 105struct llinfo_nd6 llinfo_nd6 = { 106 .ln_prev = &llinfo_nd6, 107 .ln_next = &llinfo_nd6, 108}; 109struct nd_drhead nd_defrouter; 110struct nd_prhead nd_prefix = { 0 }; 111 112int nd6_recalc_reachtm_interval = ND6_RECALC_REACHTM_INTERVAL; 113static const struct sockaddr_in6 all1_sa = { 114 .sin6_family = AF_INET6 115 , .sin6_len = sizeof(struct sockaddr_in6) 116 , .sin6_addr = {.s6_addr = {0xff, 0xff, 0xff, 0xff, 117 0xff, 0xff, 0xff, 0xff, 118 0xff, 0xff, 0xff, 0xff, 119 0xff, 0xff, 0xff, 0xff}} 120}; 121 122static void nd6_setmtu0(struct ifnet *, struct nd_ifinfo *); 123static void nd6_slowtimo(void *); 124static int regen_tmpaddr(struct in6_ifaddr *); 125static struct llinfo_nd6 *nd6_free(struct rtentry *, int); 126static void nd6_llinfo_timer(void *); 127static void clear_llinfo_pqueue(struct llinfo_nd6 *); 128 129callout_t nd6_slowtimo_ch; 130callout_t nd6_timer_ch; 131extern callout_t in6_tmpaddrtimer_ch; 132 133static int fill_drlist(void *, size_t *, size_t); 134static int fill_prlist(void *, size_t *, size_t); 135 136MALLOC_DEFINE(M_IP6NDP, "NDP", "IPv6 Neighbour Discovery"); 137 138#define LN_DEQUEUE(ln) do { \ 139 (ln)->ln_next->ln_prev = (ln)->ln_prev; \ 140 (ln)->ln_prev->ln_next = (ln)->ln_next; \ 141 } while (/*CONSTCOND*/0) 142#define LN_INSERTHEAD(ln) do { \ 143 (ln)->ln_next = llinfo_nd6.ln_next; \ 144 llinfo_nd6.ln_next = (ln); \ 145 (ln)->ln_prev = &llinfo_nd6; \ 146 (ln)->ln_next->ln_prev = (ln); \ 147 } while (/*CONSTCOND*/0) 148 149void 150nd6_init(void) 151{ 152 static int nd6_init_done = 0; 153 154 if (nd6_init_done) { 155 log(LOG_NOTICE, "nd6_init called more than once(ignored)\n"); 156 return; 157 } 158 159 /* initialization of the default router list */ 160 TAILQ_INIT(&nd_defrouter); 161 162 nd6_init_done = 1; 163 164 callout_init(&nd6_slowtimo_ch, CALLOUT_MPSAFE); 165 callout_init(&nd6_timer_ch, CALLOUT_MPSAFE); 166 167 /* start timer */ 168 callout_reset(&nd6_slowtimo_ch, ND6_SLOWTIMER_INTERVAL * hz, 169 nd6_slowtimo, NULL); 170} 171 172struct nd_ifinfo * 173nd6_ifattach(struct ifnet *ifp) 174{ 175 struct nd_ifinfo *nd; 176 177 nd = (struct nd_ifinfo *)malloc(sizeof(*nd), M_IP6NDP, M_WAITOK|M_ZERO); 178 179 nd->initialized = 1; 180 181 nd->chlim = IPV6_DEFHLIM; 182 nd->basereachable = REACHABLE_TIME; 183 nd->reachable = ND_COMPUTE_RTIME(nd->basereachable); 184 nd->retrans = RETRANS_TIMER; 185 /* 186 * Note that the default value of ip6_accept_rtadv is 0. 187 * Because we do not set ND6_IFF_OVERRIDE_RTADV here, we won't 188 * accept RAs by default. 189 */ 190 nd->flags = ND6_IFF_PERFORMNUD | ND6_IFF_ACCEPT_RTADV; 191 192 /* XXX: we cannot call nd6_setmtu since ifp is not fully initialized */ 193 nd6_setmtu0(ifp, nd); 194 195 return nd; 196} 197 198void 199nd6_ifdetach(struct nd_ifinfo *nd) 200{ 201 202 free(nd, M_IP6NDP); 203} 204 205void 206nd6_setmtu(struct ifnet *ifp) 207{ 208 nd6_setmtu0(ifp, ND_IFINFO(ifp)); 209} 210 211void 212nd6_setmtu0(struct ifnet *ifp, struct nd_ifinfo *ndi) 213{ 214 u_int32_t omaxmtu; 215 216 omaxmtu = ndi->maxmtu; 217 218 switch (ifp->if_type) { 219 case IFT_ARCNET: 220 ndi->maxmtu = MIN(ARC_PHDS_MAXMTU, ifp->if_mtu); /* RFC2497 */ 221 break; 222 case IFT_FDDI: 223 ndi->maxmtu = MIN(FDDIIPMTU, ifp->if_mtu); 224 break; 225 default: 226 ndi->maxmtu = ifp->if_mtu; 227 break; 228 } 229 230 /* 231 * Decreasing the interface MTU under IPV6 minimum MTU may cause 232 * undesirable situation. We thus notify the operator of the change 233 * explicitly. The check for omaxmtu is necessary to restrict the 234 * log to the case of changing the MTU, not initializing it. 235 */ 236 if (omaxmtu >= IPV6_MMTU && ndi->maxmtu < IPV6_MMTU) { 237 log(LOG_NOTICE, "nd6_setmtu0: new link MTU on %s (%lu) is too" 238 " small for IPv6 which needs %lu\n", 239 if_name(ifp), (unsigned long)ndi->maxmtu, (unsigned long) 240 IPV6_MMTU); 241 } 242 243 if (ndi->maxmtu > in6_maxmtu) 244 in6_setmaxmtu(); /* check all interfaces just in case */ 245} 246 247void 248nd6_option_init(void *opt, int icmp6len, union nd_opts *ndopts) 249{ 250 251 memset(ndopts, 0, sizeof(*ndopts)); 252 ndopts->nd_opts_search = (struct nd_opt_hdr *)opt; 253 ndopts->nd_opts_last 254 = (struct nd_opt_hdr *)(((u_char *)opt) + icmp6len); 255 256 if (icmp6len == 0) { 257 ndopts->nd_opts_done = 1; 258 ndopts->nd_opts_search = NULL; 259 } 260} 261 262/* 263 * Take one ND option. 264 */ 265struct nd_opt_hdr * 266nd6_option(union nd_opts *ndopts) 267{ 268 struct nd_opt_hdr *nd_opt; 269 int olen; 270 271 if (ndopts == NULL) 272 panic("ndopts == NULL in nd6_option"); 273 if (ndopts->nd_opts_last == NULL) 274 panic("uninitialized ndopts in nd6_option"); 275 if (ndopts->nd_opts_search == NULL) 276 return NULL; 277 if (ndopts->nd_opts_done) 278 return NULL; 279 280 nd_opt = ndopts->nd_opts_search; 281 282 /* make sure nd_opt_len is inside the buffer */ 283 if ((void *)&nd_opt->nd_opt_len >= (void *)ndopts->nd_opts_last) { 284 memset(ndopts, 0, sizeof(*ndopts)); 285 return NULL; 286 } 287 288 olen = nd_opt->nd_opt_len << 3; 289 if (olen == 0) { 290 /* 291 * Message validation requires that all included 292 * options have a length that is greater than zero. 293 */ 294 memset(ndopts, 0, sizeof(*ndopts)); 295 return NULL; 296 } 297 298 ndopts->nd_opts_search = (struct nd_opt_hdr *)((char *)nd_opt + olen); 299 if (ndopts->nd_opts_search > ndopts->nd_opts_last) { 300 /* option overruns the end of buffer, invalid */ 301 memset(ndopts, 0, sizeof(*ndopts)); 302 return NULL; 303 } else if (ndopts->nd_opts_search == ndopts->nd_opts_last) { 304 /* reached the end of options chain */ 305 ndopts->nd_opts_done = 1; 306 ndopts->nd_opts_search = NULL; 307 } 308 return nd_opt; 309} 310 311/* 312 * Parse multiple ND options. 313 * This function is much easier to use, for ND routines that do not need 314 * multiple options of the same type. 315 */ 316int 317nd6_options(union nd_opts *ndopts) 318{ 319 struct nd_opt_hdr *nd_opt; 320 int i = 0; 321 322 if (ndopts == NULL) 323 panic("ndopts == NULL in nd6_options"); 324 if (ndopts->nd_opts_last == NULL) 325 panic("uninitialized ndopts in nd6_options"); 326 if (ndopts->nd_opts_search == NULL) 327 return 0; 328 329 while (1) { 330 nd_opt = nd6_option(ndopts); 331 if (nd_opt == NULL && ndopts->nd_opts_last == NULL) { 332 /* 333 * Message validation requires that all included 334 * options have a length that is greater than zero. 335 */ 336 ICMP6_STATINC(ICMP6_STAT_ND_BADOPT); 337 memset(ndopts, 0, sizeof(*ndopts)); 338 return -1; 339 } 340 341 if (nd_opt == NULL) 342 goto skip1; 343 344 switch (nd_opt->nd_opt_type) { 345 case ND_OPT_SOURCE_LINKADDR: 346 case ND_OPT_TARGET_LINKADDR: 347 case ND_OPT_MTU: 348 case ND_OPT_REDIRECTED_HEADER: 349 if (ndopts->nd_opt_array[nd_opt->nd_opt_type]) { 350 nd6log((LOG_INFO, 351 "duplicated ND6 option found (type=%d)\n", 352 nd_opt->nd_opt_type)); 353 /* XXX bark? */ 354 } else { 355 ndopts->nd_opt_array[nd_opt->nd_opt_type] 356 = nd_opt; 357 } 358 break; 359 case ND_OPT_PREFIX_INFORMATION: 360 if (ndopts->nd_opt_array[nd_opt->nd_opt_type] == 0) { 361 ndopts->nd_opt_array[nd_opt->nd_opt_type] 362 = nd_opt; 363 } 364 ndopts->nd_opts_pi_end = 365 (struct nd_opt_prefix_info *)nd_opt; 366 break; 367 default: 368 /* 369 * Unknown options must be silently ignored, 370 * to accommodate future extension to the protocol. 371 */ 372 nd6log((LOG_DEBUG, 373 "nd6_options: unsupported option %d - " 374 "option ignored\n", nd_opt->nd_opt_type)); 375 } 376 377skip1: 378 i++; 379 if (i > nd6_maxndopt) { 380 ICMP6_STATINC(ICMP6_STAT_ND_TOOMANYOPT); 381 nd6log((LOG_INFO, "too many loop in nd opt\n")); 382 break; 383 } 384 385 if (ndopts->nd_opts_done) 386 break; 387 } 388 389 return 0; 390} 391 392/* 393 * ND6 timer routine to handle ND6 entries 394 */ 395void 396nd6_llinfo_settimer(struct llinfo_nd6 *ln, long xtick) 397{ 398 int s; 399 400 s = splsoftnet(); 401 402 if (xtick < 0) { 403 ln->ln_expire = 0; 404 ln->ln_ntick = 0; 405 callout_stop(&ln->ln_timer_ch); 406 } else { 407 ln->ln_expire = time_second + xtick / hz; 408 if (xtick > INT_MAX) { 409 ln->ln_ntick = xtick - INT_MAX; 410 callout_reset(&ln->ln_timer_ch, INT_MAX, 411 nd6_llinfo_timer, ln); 412 } else { 413 ln->ln_ntick = 0; 414 callout_reset(&ln->ln_timer_ch, xtick, 415 nd6_llinfo_timer, ln); 416 } 417 } 418 419 splx(s); 420} 421 422static void 423nd6_llinfo_timer(void *arg) 424{ 425 struct llinfo_nd6 *ln; 426 struct rtentry *rt; 427 const struct sockaddr_in6 *dst; 428 struct ifnet *ifp; 429 struct nd_ifinfo *ndi = NULL; 430 431 mutex_enter(softnet_lock); 432 KERNEL_LOCK(1, NULL); 433 434 ln = (struct llinfo_nd6 *)arg; 435 436 if (ln->ln_ntick > 0) { 437 nd6_llinfo_settimer(ln, ln->ln_ntick); 438 KERNEL_UNLOCK_ONE(NULL); 439 mutex_exit(softnet_lock); 440 return; 441 } 442 443 if ((rt = ln->ln_rt) == NULL) 444 panic("ln->ln_rt == NULL"); 445 if ((ifp = rt->rt_ifp) == NULL) 446 panic("ln->ln_rt->rt_ifp == NULL"); 447 ndi = ND_IFINFO(ifp); 448 dst = satocsin6(rt_getkey(rt)); 449 450 /* sanity check */ 451 if (rt->rt_llinfo && (struct llinfo_nd6 *)rt->rt_llinfo != ln) 452 panic("rt_llinfo(%p) is not equal to ln(%p)", 453 rt->rt_llinfo, ln); 454 if (!dst) 455 panic("dst=0 in nd6_timer(ln=%p)", ln); 456 457 switch (ln->ln_state) { 458 case ND6_LLINFO_INCOMPLETE: 459 if (ln->ln_asked < nd6_mmaxtries) { 460 ln->ln_asked++; 461 nd6_llinfo_settimer(ln, (long)ndi->retrans * hz / 1000); 462 nd6_ns_output(ifp, NULL, &dst->sin6_addr, ln, 0); 463 } else { 464 struct mbuf *m = ln->ln_hold; 465 if (m) { 466 struct mbuf *m0; 467 468 /* 469 * assuming every packet in ln_hold has 470 * the same IP header 471 */ 472 m0 = m->m_nextpkt; 473 m->m_nextpkt = NULL; 474 icmp6_error2(m, ICMP6_DST_UNREACH, 475 ICMP6_DST_UNREACH_ADDR, 0, rt->rt_ifp); 476 477 ln->ln_hold = m0; 478 clear_llinfo_pqueue(ln); 479 } 480 (void)nd6_free(rt, 0); 481 ln = NULL; 482 } 483 break; 484 case ND6_LLINFO_REACHABLE: 485 if (!ND6_LLINFO_PERMANENT(ln)) { 486 ln->ln_state = ND6_LLINFO_STALE; 487 nd6_llinfo_settimer(ln, (long)nd6_gctimer * hz); 488 } 489 break; 490 case ND6_LLINFO_PURGE: 491 case ND6_LLINFO_STALE: 492 /* Garbage Collection(RFC 2461 5.3) */ 493 if (!ND6_LLINFO_PERMANENT(ln)) { 494 (void)nd6_free(rt, 1); 495 ln = NULL; 496 } 497 break; 498 499 case ND6_LLINFO_DELAY: 500 if (ndi && (ndi->flags & ND6_IFF_PERFORMNUD) != 0) { 501 /* We need NUD */ 502 ln->ln_asked = 1; 503 ln->ln_state = ND6_LLINFO_PROBE; 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 ln->ln_state = ND6_LLINFO_STALE; /* XXX */ 509 nd6_llinfo_settimer(ln, (long)nd6_gctimer * hz); 510 } 511 break; 512 case ND6_LLINFO_PROBE: 513 if (ln->ln_asked < nd6_umaxtries) { 514 ln->ln_asked++; 515 nd6_llinfo_settimer(ln, (long)ndi->retrans * hz / 1000); 516 nd6_ns_output(ifp, &dst->sin6_addr, 517 &dst->sin6_addr, ln, 0); 518 } else { 519 (void)nd6_free(rt, 0); 520 ln = NULL; 521 } 522 break; 523 } 524 525 KERNEL_UNLOCK_ONE(NULL); 526 mutex_exit(softnet_lock); 527} 528 529/* 530 * ND6 timer routine to expire default route list and prefix list 531 */ 532void 533nd6_timer(void *ignored_arg) 534{ 535 struct nd_defrouter *next_dr, *dr; 536 struct nd_prefix *next_pr, *pr; 537 struct in6_ifaddr *ia6, *nia6; 538 539 callout_reset(&nd6_timer_ch, nd6_prune * hz, 540 nd6_timer, NULL); 541 542 mutex_enter(softnet_lock); 543 KERNEL_LOCK(1, NULL); 544 545 /* expire default router list */ 546 547 TAILQ_FOREACH_SAFE(dr, &nd_defrouter, dr_entry, next_dr) { 548 if (dr->expire && dr->expire < time_second) { 549 defrtrlist_del(dr); 550 } 551 } 552 553 /* 554 * expire interface addresses. 555 * in the past the loop was inside prefix expiry processing. 556 * However, from a stricter speci-confrmance standpoint, we should 557 * rather separate address lifetimes and prefix lifetimes. 558 */ 559 addrloop: 560 for (ia6 = in6_ifaddr; ia6; ia6 = nia6) { 561 nia6 = ia6->ia_next; 562 /* check address lifetime */ 563 if (IFA6_IS_INVALID(ia6)) { 564 int regen = 0; 565 566 /* 567 * If the expiring address is temporary, try 568 * regenerating a new one. This would be useful when 569 * we suspended a laptop PC, then turned it on after a 570 * period that could invalidate all temporary 571 * addresses. Although we may have to restart the 572 * loop (see below), it must be after purging the 573 * address. Otherwise, we'd see an infinite loop of 574 * regeneration. 575 */ 576 if (ip6_use_tempaddr && 577 (ia6->ia6_flags & IN6_IFF_TEMPORARY) != 0) { 578 if (regen_tmpaddr(ia6) == 0) 579 regen = 1; 580 } 581 582 in6_purgeaddr(&ia6->ia_ifa); 583 584 if (regen) 585 goto addrloop; /* XXX: see below */ 586 } else if (IFA6_IS_DEPRECATED(ia6)) { 587 int oldflags = ia6->ia6_flags; 588 589 ia6->ia6_flags |= IN6_IFF_DEPRECATED; 590 591 /* 592 * If a temporary address has just become deprecated, 593 * regenerate a new one if possible. 594 */ 595 if (ip6_use_tempaddr && 596 (ia6->ia6_flags & IN6_IFF_TEMPORARY) != 0 && 597 (oldflags & IN6_IFF_DEPRECATED) == 0) { 598 599 if (regen_tmpaddr(ia6) == 0) { 600 /* 601 * A new temporary address is 602 * generated. 603 * XXX: this means the address chain 604 * has changed while we are still in 605 * the loop. Although the change 606 * would not cause disaster (because 607 * it's not a deletion, but an 608 * addition,) we'd rather restart the 609 * loop just for safety. Or does this 610 * significantly reduce performance?? 611 */ 612 goto addrloop; 613 } 614 } 615 } else { 616 /* 617 * A new RA might have made a deprecated address 618 * preferred. 619 */ 620 ia6->ia6_flags &= ~IN6_IFF_DEPRECATED; 621 } 622 } 623 624 /* expire prefix list */ 625 LIST_FOREACH_SAFE(pr, &nd_prefix, ndpr_entry, next_pr) { 626 /* 627 * check prefix lifetime. 628 * since pltime is just for autoconf, pltime processing for 629 * prefix is not necessary. 630 */ 631 if (pr->ndpr_vltime != ND6_INFINITE_LIFETIME && 632 time_second - pr->ndpr_lastupdate > pr->ndpr_vltime) { 633 634 /* 635 * address expiration and prefix expiration are 636 * separate. NEVER perform in6_purgeaddr here. 637 */ 638 639 prelist_remove(pr); 640 } 641 } 642 643 KERNEL_UNLOCK_ONE(NULL); 644 mutex_exit(softnet_lock); 645} 646 647/* ia6: deprecated/invalidated temporary address */ 648static int 649regen_tmpaddr(struct in6_ifaddr *ia6) 650{ 651 struct ifaddr *ifa; 652 struct ifnet *ifp; 653 struct in6_ifaddr *public_ifa6 = NULL; 654 655 ifp = ia6->ia_ifa.ifa_ifp; 656 IFADDR_FOREACH(ifa, ifp) { 657 struct in6_ifaddr *it6; 658 659 if (ifa->ifa_addr->sa_family != AF_INET6) 660 continue; 661 662 it6 = (struct in6_ifaddr *)ifa; 663 664 /* ignore no autoconf addresses. */ 665 if ((it6->ia6_flags & IN6_IFF_AUTOCONF) == 0) 666 continue; 667 668 /* ignore autoconf addresses with different prefixes. */ 669 if (it6->ia6_ndpr == NULL || it6->ia6_ndpr != ia6->ia6_ndpr) 670 continue; 671 672 /* 673 * Now we are looking at an autoconf address with the same 674 * prefix as ours. If the address is temporary and is still 675 * preferred, do not create another one. It would be rare, but 676 * could happen, for example, when we resume a laptop PC after 677 * a long period. 678 */ 679 if ((it6->ia6_flags & IN6_IFF_TEMPORARY) != 0 && 680 !IFA6_IS_DEPRECATED(it6)) { 681 public_ifa6 = NULL; 682 break; 683 } 684 685 /* 686 * This is a public autoconf address that has the same prefix 687 * as ours. If it is preferred, keep it. We can't break the 688 * loop here, because there may be a still-preferred temporary 689 * address with the prefix. 690 */ 691 if (!IFA6_IS_DEPRECATED(it6)) 692 public_ifa6 = it6; 693 } 694 695 if (public_ifa6 != NULL) { 696 int e; 697 698 /* 699 * Random factor is introduced in the preferred lifetime, so 700 * we do not need additional delay (3rd arg to in6_tmpifadd). 701 */ 702 if ((e = in6_tmpifadd(public_ifa6, 0, 0)) != 0) { 703 log(LOG_NOTICE, "regen_tmpaddr: failed to create a new" 704 " tmp addr, errno=%d\n", e); 705 return -1; 706 } 707 return 0; 708 } 709 710 return -1; 711} 712 713bool 714nd6_accepts_rtadv(const struct nd_ifinfo *ndi) 715{ 716 switch (ndi->flags & (ND6_IFF_ACCEPT_RTADV|ND6_IFF_OVERRIDE_RTADV)) { 717 case ND6_IFF_OVERRIDE_RTADV|ND6_IFF_ACCEPT_RTADV: 718 return true; 719 case ND6_IFF_ACCEPT_RTADV: 720 return ip6_accept_rtadv != 0; 721 case ND6_IFF_OVERRIDE_RTADV: 722 case 0: 723 default: 724 return false; 725 } 726} 727 728/* 729 * Nuke neighbor cache/prefix/default router management table, right before 730 * ifp goes away. 731 */ 732void 733nd6_purge(struct ifnet *ifp) 734{ 735 struct llinfo_nd6 *ln, *nln; 736 struct nd_defrouter *dr, *ndr; 737 struct nd_prefix *pr, *npr; 738 739 /* 740 * Nuke default router list entries toward ifp. 741 * We defer removal of default router list entries that is installed 742 * in the routing table, in order to keep additional side effects as 743 * small as possible. 744 */ 745 TAILQ_FOREACH_SAFE(dr, &nd_defrouter, dr_entry, ndr) { 746 if (dr->installed) 747 continue; 748 749 if (dr->ifp == ifp) 750 defrtrlist_del(dr); 751 } 752 753 TAILQ_FOREACH_SAFE(dr, &nd_defrouter, dr_entry, ndr) { 754 if (!dr->installed) 755 continue; 756 757 if (dr->ifp == ifp) 758 defrtrlist_del(dr); 759 } 760 761 /* Nuke prefix list entries toward ifp */ 762 LIST_FOREACH_SAFE(pr, &nd_prefix, ndpr_entry, npr) { 763 if (pr->ndpr_ifp == ifp) { 764 /* 765 * Because if_detach() does *not* release prefixes 766 * while purging addresses the reference count will 767 * still be above zero. We therefore reset it to 768 * make sure that the prefix really gets purged. 769 */ 770 pr->ndpr_refcnt = 0; 771 /* 772 * Previously, pr->ndpr_addr is removed as well, 773 * but I strongly believe we don't have to do it. 774 * nd6_purge() is only called from in6_ifdetach(), 775 * which removes all the associated interface addresses 776 * by itself. 777 * (jinmei@kame.net 20010129) 778 */ 779 prelist_remove(pr); 780 } 781 } 782 783 /* cancel default outgoing interface setting */ 784 if (nd6_defifindex == ifp->if_index) 785 nd6_setdefaultiface(0); 786 787 /* XXX: too restrictive? */ 788 if (!ip6_forwarding && ifp->if_afdata[AF_INET6]) { 789 struct nd_ifinfo *ndi = ND_IFINFO(ifp); 790 if (ndi && nd6_accepts_rtadv(ndi)) { 791 /* refresh default router list */ 792 defrouter_select(); 793 } 794 } 795 796 /* 797 * Nuke neighbor cache entries for the ifp. 798 * Note that rt->rt_ifp may not be the same as ifp, 799 * due to KAME goto ours hack. See RTM_RESOLVE case in 800 * nd6_rtrequest(), and ip6_input(). 801 */ 802 ln = llinfo_nd6.ln_next; 803 while (ln != NULL && ln != &llinfo_nd6) { 804 struct rtentry *rt; 805 const struct sockaddr_dl *sdl; 806 807 nln = ln->ln_next; 808 rt = ln->ln_rt; 809 if (rt && rt->rt_gateway && 810 rt->rt_gateway->sa_family == AF_LINK) { 811 sdl = satocsdl(rt->rt_gateway); 812 if (sdl->sdl_index == ifp->if_index) 813 nln = nd6_free(rt, 0); 814 } 815 ln = nln; 816 } 817} 818 819struct rtentry * 820nd6_lookup(const struct in6_addr *addr6, int create, struct ifnet *ifp) 821{ 822 struct rtentry *rt; 823 struct sockaddr_in6 sin6; 824 825 sockaddr_in6_init(&sin6, addr6, 0, 0, 0); 826 rt = rtalloc1((struct sockaddr *)&sin6, create); 827 if (rt != NULL && (rt->rt_flags & RTF_LLINFO) == 0) { 828 /* 829 * This is the case for the default route. 830 * If we want to create a neighbor cache for the address, we 831 * should free the route for the destination and allocate an 832 * interface route. 833 */ 834 if (create) { 835 RTFREE(rt); 836 rt = NULL; 837 } 838 } 839 if (rt != NULL) 840 ; 841 else if (create && ifp) { 842 int e; 843 844 /* 845 * If no route is available and create is set, 846 * we allocate a host route for the destination 847 * and treat it like an interface route. 848 * This hack is necessary for a neighbor which can't 849 * be covered by our own prefix. 850 */ 851 struct ifaddr *ifa = 852 ifaof_ifpforaddr((struct sockaddr *)&sin6, ifp); 853 if (ifa == NULL) 854 return NULL; 855 856 /* 857 * Create a new route. RTF_LLINFO is necessary 858 * to create a Neighbor Cache entry for the 859 * destination in nd6_rtrequest which will be 860 * called in rtrequest via ifa->ifa_rtrequest. 861 */ 862 if ((e = rtrequest(RTM_ADD, (const struct sockaddr *)&sin6, 863 ifa->ifa_addr, (const struct sockaddr *)&all1_sa, 864 (ifa->ifa_flags | RTF_HOST | RTF_LLINFO) & 865 ~RTF_CLONING, &rt)) != 0) { 866#if 0 867 log(LOG_ERR, 868 "nd6_lookup: failed to add route for a " 869 "neighbor(%s), errno=%d\n", 870 ip6_sprintf(addr6), e); 871#endif 872 return NULL; 873 } 874 if (rt == NULL) 875 return NULL; 876 if (rt->rt_llinfo) { 877 struct llinfo_nd6 *ln = 878 (struct llinfo_nd6 *)rt->rt_llinfo; 879 ln->ln_state = ND6_LLINFO_NOSTATE; 880 } 881 } else 882 return NULL; 883 rt->rt_refcnt--; 884 /* 885 * Validation for the entry. 886 * Note that the check for rt_llinfo is necessary because a cloned 887 * route from a parent route that has the L flag (e.g. the default 888 * route to a p2p interface) may have the flag, too, while the 889 * destination is not actually a neighbor. 890 * XXX: we can't use rt->rt_ifp to check for the interface, since 891 * it might be the loopback interface if the entry is for our 892 * own address on a non-loopback interface. Instead, we should 893 * use rt->rt_ifa->ifa_ifp, which would specify the REAL 894 * interface. 895 * Note also that ifa_ifp and ifp may differ when we connect two 896 * interfaces to a same link, install a link prefix to an interface, 897 * and try to install a neighbor cache on an interface that does not 898 * have a route to the prefix. 899 */ 900 if ((rt->rt_flags & RTF_GATEWAY) || (rt->rt_flags & RTF_LLINFO) == 0 || 901 rt->rt_gateway->sa_family != AF_LINK || rt->rt_llinfo == NULL || 902 (ifp && rt->rt_ifa->ifa_ifp != ifp)) { 903 if (create) { 904 nd6log((LOG_DEBUG, 905 "nd6_lookup: failed to lookup %s (if = %s)\n", 906 ip6_sprintf(addr6), 907 ifp ? if_name(ifp) : "unspec")); 908 } 909 return NULL; 910 } 911 return rt; 912} 913 914/* 915 * Detect if a given IPv6 address identifies a neighbor on a given link. 916 * XXX: should take care of the destination of a p2p link? 917 */ 918int 919nd6_is_addr_neighbor(const struct sockaddr_in6 *addr, struct ifnet *ifp) 920{ 921 struct nd_prefix *pr; 922 923 /* 924 * A link-local address is always a neighbor. 925 * XXX: a link does not necessarily specify a single interface. 926 */ 927 if (IN6_IS_ADDR_LINKLOCAL(&addr->sin6_addr)) { 928 struct sockaddr_in6 sin6_copy; 929 u_int32_t zone; 930 931 /* 932 * We need sin6_copy since sa6_recoverscope() may modify the 933 * content (XXX). 934 */ 935 sin6_copy = *addr; 936 if (sa6_recoverscope(&sin6_copy)) 937 return 0; /* XXX: should be impossible */ 938 if (in6_setscope(&sin6_copy.sin6_addr, ifp, &zone)) 939 return 0; 940 if (sin6_copy.sin6_scope_id == zone) 941 return 1; 942 else 943 return 0; 944 } 945 946 /* 947 * If the address matches one of our on-link prefixes, it should be a 948 * neighbor. 949 */ 950 LIST_FOREACH(pr, &nd_prefix, ndpr_entry) { 951 if (pr->ndpr_ifp != ifp) 952 continue; 953 954 if (!(pr->ndpr_stateflags & NDPRF_ONLINK)) 955 continue; 956 957 if (IN6_ARE_MASKED_ADDR_EQUAL(&pr->ndpr_prefix.sin6_addr, 958 &addr->sin6_addr, &pr->ndpr_mask)) 959 return 1; 960 } 961 962 /* 963 * If the default router list is empty, all addresses are regarded 964 * as on-link, and thus, as a neighbor. 965 * XXX: we restrict the condition to hosts, because routers usually do 966 * not have the "default router list". 967 */ 968 if (!ip6_forwarding && TAILQ_FIRST(&nd_defrouter) == NULL && 969 nd6_defifindex == ifp->if_index) { 970 return 1; 971 } 972 973 /* 974 * Even if the address matches none of our addresses, it might be 975 * in the neighbor cache. 976 */ 977 if (nd6_lookup(&addr->sin6_addr, 0, ifp) != NULL) 978 return 1; 979 980 return 0; 981} 982 983/* 984 * Free an nd6 llinfo entry. 985 * Since the function would cause significant changes in the kernel, DO NOT 986 * make it global, unless you have a strong reason for the change, and are sure 987 * that the change is safe. 988 */ 989static struct llinfo_nd6 * 990nd6_free(struct rtentry *rt, int gc) 991{ 992 struct llinfo_nd6 *ln = (struct llinfo_nd6 *)rt->rt_llinfo, *next; 993 struct in6_addr in6 = satocsin6(rt_getkey(rt))->sin6_addr; 994 struct nd_defrouter *dr; 995 996 /* 997 * we used to have pfctlinput(PRC_HOSTDEAD) here. 998 * even though it is not harmful, it was not really necessary. 999 */ 1000 1001 /* cancel timer */ 1002 nd6_llinfo_settimer(ln, -1); 1003 1004 if (!ip6_forwarding) { 1005 int s; 1006 s = splsoftnet(); 1007 dr = defrouter_lookup(&satocsin6(rt_getkey(rt))->sin6_addr, 1008 rt->rt_ifp); 1009 1010 if (dr != NULL && dr->expire && 1011 ln->ln_state == ND6_LLINFO_STALE && gc) { 1012 /* 1013 * If the reason for the deletion is just garbage 1014 * collection, and the neighbor is an active default 1015 * router, do not delete it. Instead, reset the GC 1016 * timer using the router's lifetime. 1017 * Simply deleting the entry would affect default 1018 * router selection, which is not necessarily a good 1019 * thing, especially when we're using router preference 1020 * values. 1021 * XXX: the check for ln_state would be redundant, 1022 * but we intentionally keep it just in case. 1023 */ 1024 if (dr->expire > time_second) 1025 nd6_llinfo_settimer(ln, 1026 (dr->expire - time_second) * hz); 1027 else 1028 nd6_llinfo_settimer(ln, (long)nd6_gctimer * hz); 1029 splx(s); 1030 return ln->ln_next; 1031 } 1032 1033 if (ln->ln_router || dr) { 1034 /* 1035 * rt6_flush must be called whether or not the neighbor 1036 * is in the Default Router List. 1037 * See a corresponding comment in nd6_na_input(). 1038 */ 1039 rt6_flush(&in6, rt->rt_ifp); 1040 } 1041 1042 if (dr) { 1043 /* 1044 * Unreachablity of a router might affect the default 1045 * router selection and on-link detection of advertised 1046 * prefixes. 1047 */ 1048 1049 /* 1050 * Temporarily fake the state to choose a new default 1051 * router and to perform on-link determination of 1052 * prefixes correctly. 1053 * Below the state will be set correctly, 1054 * or the entry itself will be deleted. 1055 */ 1056 ln->ln_state = ND6_LLINFO_INCOMPLETE; 1057 1058 /* 1059 * Since defrouter_select() does not affect the 1060 * on-link determination and MIP6 needs the check 1061 * before the default router selection, we perform 1062 * the check now. 1063 */ 1064 pfxlist_onlink_check(); 1065 1066 /* 1067 * refresh default router list 1068 */ 1069 defrouter_select(); 1070 } 1071 splx(s); 1072 } 1073 1074 /* 1075 * Before deleting the entry, remember the next entry as the 1076 * return value. We need this because pfxlist_onlink_check() above 1077 * might have freed other entries (particularly the old next entry) as 1078 * a side effect (XXX). 1079 */ 1080 next = ln->ln_next; 1081 1082 /* 1083 * Detach the route from the routing tree and the list of neighbor 1084 * caches, and disable the route entry not to be used in already 1085 * cached routes. 1086 */ 1087 rtrequest(RTM_DELETE, rt_getkey(rt), NULL, rt_mask(rt), 0, NULL); 1088 1089 return next; 1090} 1091 1092/* 1093 * Upper-layer reachability hint for Neighbor Unreachability Detection. 1094 * 1095 * XXX cost-effective methods? 1096 */ 1097void 1098nd6_nud_hint(struct rtentry *rt, struct in6_addr *dst6, int force) 1099{ 1100 struct llinfo_nd6 *ln; 1101 1102 /* 1103 * If the caller specified "rt", use that. Otherwise, resolve the 1104 * routing table by supplied "dst6". 1105 */ 1106 if (rt == NULL) { 1107 if (dst6 == NULL) 1108 return; 1109 if ((rt = nd6_lookup(dst6, 0, NULL)) == NULL) 1110 return; 1111 } 1112 1113 if ((rt->rt_flags & RTF_GATEWAY) != 0 || 1114 (rt->rt_flags & RTF_LLINFO) == 0 || 1115 !rt->rt_llinfo || !rt->rt_gateway || 1116 rt->rt_gateway->sa_family != AF_LINK) { 1117 /* This is not a host route. */ 1118 return; 1119 } 1120 1121 ln = (struct llinfo_nd6 *)rt->rt_llinfo; 1122 if (ln->ln_state < ND6_LLINFO_REACHABLE) 1123 return; 1124 1125 /* 1126 * if we get upper-layer reachability confirmation many times, 1127 * it is possible we have false information. 1128 */ 1129 if (!force) { 1130 ln->ln_byhint++; 1131 if (ln->ln_byhint > nd6_maxnudhint) 1132 return; 1133 } 1134 1135 ln->ln_state = ND6_LLINFO_REACHABLE; 1136 if (!ND6_LLINFO_PERMANENT(ln)) { 1137 nd6_llinfo_settimer(ln, 1138 (long)ND_IFINFO(rt->rt_ifp)->reachable * hz); 1139 } 1140} 1141 1142void 1143nd6_rtrequest(int req, struct rtentry *rt, const struct rt_addrinfo *info) 1144{ 1145 struct sockaddr *gate = rt->rt_gateway; 1146 struct llinfo_nd6 *ln = (struct llinfo_nd6 *)rt->rt_llinfo; 1147 struct ifnet *ifp = rt->rt_ifp; 1148 uint8_t namelen = strlen(ifp->if_xname), addrlen = ifp->if_addrlen; 1149 struct ifaddr *ifa; 1150 1151 RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key); 1152 1153 if (req == RTM_LLINFO_UPD) { 1154 int rc; 1155 struct in6_addr *in6; 1156 struct in6_addr in6_all; 1157 int anycast; 1158 1159 if ((ifa = info->rti_ifa) == NULL) 1160 return; 1161 1162 in6 = &ifatoia6(ifa)->ia_addr.sin6_addr; 1163 anycast = ifatoia6(ifa)->ia6_flags & IN6_IFF_ANYCAST; 1164 1165 in6_all = in6addr_linklocal_allnodes; 1166 if ((rc = in6_setscope(&in6_all, ifa->ifa_ifp, NULL)) != 0) { 1167 log(LOG_ERR, "%s: failed to set scope %s " 1168 "(errno=%d)\n", __func__, if_name(ifp), rc); 1169 return; 1170 } 1171 1172 /* XXX don't set Override for proxy addresses */ 1173 nd6_na_output(ifa->ifa_ifp, &in6_all, in6, 1174 (anycast ? 0 : ND_NA_FLAG_OVERRIDE) 1175#if 0 1176 | (ip6_forwarding ? ND_NA_FLAG_ROUTER : 0) 1177#endif 1178 , 1, NULL); 1179 return; 1180 } 1181 1182 if ((rt->rt_flags & RTF_GATEWAY) != 0) 1183 return; 1184 1185 if (nd6_need_cache(ifp) == 0 && (rt->rt_flags & RTF_HOST) == 0) { 1186 RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key); 1187 /* 1188 * This is probably an interface direct route for a link 1189 * which does not need neighbor caches (e.g. fe80::%lo0/64). 1190 * We do not need special treatment below for such a route. 1191 * Moreover, the RTF_LLINFO flag which would be set below 1192 * would annoy the ndp(8) command. 1193 */ 1194 return; 1195 } 1196 1197 if (req == RTM_RESOLVE && 1198 (nd6_need_cache(ifp) == 0 || /* stf case */ 1199 !nd6_is_addr_neighbor(satocsin6(rt_getkey(rt)), ifp))) { 1200 RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key); 1201 /* 1202 * FreeBSD and BSD/OS often make a cloned host route based 1203 * on a less-specific route (e.g. the default route). 1204 * If the less specific route does not have a "gateway" 1205 * (this is the case when the route just goes to a p2p or an 1206 * stf interface), we'll mistakenly make a neighbor cache for 1207 * the host route, and will see strange neighbor solicitation 1208 * for the corresponding destination. In order to avoid the 1209 * confusion, we check if the destination of the route is 1210 * a neighbor in terms of neighbor discovery, and stop the 1211 * process if not. Additionally, we remove the LLINFO flag 1212 * so that ndp(8) will not try to get the neighbor information 1213 * of the destination. 1214 */ 1215 rt->rt_flags &= ~RTF_LLINFO; 1216 return; 1217 } 1218 1219 switch (req) { 1220 case RTM_ADD: 1221 RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key); 1222 /* 1223 * There is no backward compatibility :) 1224 * 1225 * if ((rt->rt_flags & RTF_HOST) == 0 && 1226 * SIN(rt_mask(rt))->sin_addr.s_addr != 0xffffffff) 1227 * rt->rt_flags |= RTF_CLONING; 1228 */ 1229 if ((rt->rt_flags & RTF_CLONING) || 1230 ((rt->rt_flags & RTF_LLINFO) && ln == NULL)) { 1231 union { 1232 struct sockaddr sa; 1233 struct sockaddr_dl sdl; 1234 struct sockaddr_storage ss; 1235 } u; 1236 /* 1237 * Case 1: This route should come from a route to 1238 * interface (RTF_CLONING case) or the route should be 1239 * treated as on-link but is currently not 1240 * (RTF_LLINFO && ln == NULL case). 1241 */ 1242 if (sockaddr_dl_init(&u.sdl, sizeof(u.ss), 1243 ifp->if_index, ifp->if_type, 1244 NULL, namelen, NULL, addrlen) == NULL) { 1245 printf("%s.%d: sockaddr_dl_init(, %zu, ) " 1246 "failed on %s\n", __func__, __LINE__, 1247 sizeof(u.ss), if_name(ifp)); 1248 } 1249 rt_setgate(rt, &u.sa); 1250 gate = rt->rt_gateway; 1251 RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key); 1252 if (ln != NULL) 1253 nd6_llinfo_settimer(ln, 0); 1254 RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key); 1255 if ((rt->rt_flags & RTF_CLONING) != 0) 1256 break; 1257 } 1258 RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key); 1259 /* 1260 * In IPv4 code, we try to annonuce new RTF_ANNOUNCE entry here. 1261 * We don't do that here since llinfo is not ready yet. 1262 * 1263 * There are also couple of other things to be discussed: 1264 * - unsolicited NA code needs improvement beforehand 1265 * - RFC2461 says we MAY send multicast unsolicited NA 1266 * (7.2.6 paragraph 4), however, it also says that we 1267 * SHOULD provide a mechanism to prevent multicast NA storm. 1268 * we don't have anything like it right now. 1269 * note that the mechanism needs a mutual agreement 1270 * between proxies, which means that we need to implement 1271 * a new protocol, or a new kludge. 1272 * - from RFC2461 6.2.4, host MUST NOT send an unsolicited NA. 1273 * we need to check ip6forwarding before sending it. 1274 * (or should we allow proxy ND configuration only for 1275 * routers? there's no mention about proxy ND from hosts) 1276 */ 1277#if 0 1278 /* XXX it does not work */ 1279 if (rt->rt_flags & RTF_ANNOUNCE) 1280 nd6_na_output(ifp, 1281 &satocsin6(rt_getkey(rt))->sin6_addr, 1282 &satocsin6(rt_getkey(rt))->sin6_addr, 1283 ip6_forwarding ? ND_NA_FLAG_ROUTER : 0, 1284 1, NULL); 1285#endif 1286 /* FALLTHROUGH */ 1287 case RTM_RESOLVE: 1288 if ((ifp->if_flags & (IFF_POINTOPOINT | IFF_LOOPBACK)) == 0) { 1289 RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key); 1290 /* 1291 * Address resolution isn't necessary for a point to 1292 * point link, so we can skip this test for a p2p link. 1293 */ 1294 if (gate->sa_family != AF_LINK || 1295 gate->sa_len < 1296 sockaddr_dl_measure(namelen, addrlen)) { 1297 log(LOG_DEBUG, 1298 "nd6_rtrequest: bad gateway value: %s\n", 1299 if_name(ifp)); 1300 break; 1301 } 1302 satosdl(gate)->sdl_type = ifp->if_type; 1303 satosdl(gate)->sdl_index = ifp->if_index; 1304 RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key); 1305 } 1306 if (ln != NULL) 1307 break; /* This happens on a route change */ 1308 RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key); 1309 /* 1310 * Case 2: This route may come from cloning, or a manual route 1311 * add with a LL address. 1312 */ 1313 R_Malloc(ln, struct llinfo_nd6 *, sizeof(*ln)); 1314 rt->rt_llinfo = ln; 1315 RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key); 1316 if (ln == NULL) { 1317 log(LOG_DEBUG, "nd6_rtrequest: malloc failed\n"); 1318 break; 1319 } 1320 RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key); 1321 nd6_inuse++; 1322 nd6_allocated++; 1323 memset(ln, 0, sizeof(*ln)); 1324 ln->ln_rt = rt; 1325 callout_init(&ln->ln_timer_ch, CALLOUT_MPSAFE); 1326 /* this is required for "ndp" command. - shin */ 1327 if (req == RTM_ADD) { 1328 /* 1329 * gate should have some valid AF_LINK entry, 1330 * and ln->ln_expire should have some lifetime 1331 * which is specified by ndp command. 1332 */ 1333 ln->ln_state = ND6_LLINFO_REACHABLE; 1334 ln->ln_byhint = 0; 1335 } else { 1336 /* 1337 * When req == RTM_RESOLVE, rt is created and 1338 * initialized in rtrequest(), so rt_expire is 0. 1339 */ 1340 ln->ln_state = ND6_LLINFO_NOSTATE; 1341 nd6_llinfo_settimer(ln, 0); 1342 } 1343 RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key); 1344 rt->rt_flags |= RTF_LLINFO; 1345 ln->ln_next = llinfo_nd6.ln_next; 1346 llinfo_nd6.ln_next = ln; 1347 ln->ln_prev = &llinfo_nd6; 1348 ln->ln_next->ln_prev = ln; 1349 1350 /* 1351 * If we have too many cache entries, initiate immediate 1352 * purging for some "less recently used" entries. Note that 1353 * we cannot directly call nd6_free() here because it would 1354 * cause re-entering rtable related routines triggering an LOR 1355 * problem for FreeBSD. 1356 */ 1357 if (ip6_neighborgcthresh >= 0 && 1358 nd6_inuse >= ip6_neighborgcthresh) { 1359 int i; 1360 1361 for (i = 0; i < 10 && llinfo_nd6.ln_prev != ln; i++) { 1362 struct llinfo_nd6 *ln_end = llinfo_nd6.ln_prev; 1363 1364 /* Move this entry to the head */ 1365 LN_DEQUEUE(ln_end); 1366 LN_INSERTHEAD(ln_end); 1367 1368 if (ND6_LLINFO_PERMANENT(ln_end)) 1369 continue; 1370 1371 if (ln_end->ln_state > ND6_LLINFO_INCOMPLETE) 1372 ln_end->ln_state = ND6_LLINFO_STALE; 1373 else 1374 ln_end->ln_state = ND6_LLINFO_PURGE; 1375 nd6_llinfo_settimer(ln_end, 0); 1376 } 1377 } 1378 1379 RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key); 1380 /* 1381 * check if rt_getkey(rt) is an address assigned 1382 * to the interface. 1383 */ 1384 ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp, 1385 &satocsin6(rt_getkey(rt))->sin6_addr); 1386 RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key); 1387 if (ifa != NULL) { 1388 const void *mac; 1389 nd6_llinfo_settimer(ln, -1); 1390 ln->ln_state = ND6_LLINFO_REACHABLE; 1391 ln->ln_byhint = 0; 1392 if ((mac = nd6_ifptomac(ifp)) != NULL) { 1393 /* XXX check for error */ 1394 if (sockaddr_dl_setaddr(satosdl(gate), 1395 gate->sa_len, mac, 1396 ifp->if_addrlen) == NULL) { 1397 printf("%s.%d: " 1398 "sockaddr_dl_setaddr(, %d, ) " 1399 "failed on %s\n", __func__, 1400 __LINE__, gate->sa_len, 1401 if_name(ifp)); 1402 } 1403 } 1404 if (nd6_useloopback) { 1405 ifp = rt->rt_ifp = lo0ifp; /* XXX */ 1406 /* 1407 * Make sure rt_ifa be equal to the ifaddr 1408 * corresponding to the address. 1409 * We need this because when we refer 1410 * rt_ifa->ia6_flags in ip6_input, we assume 1411 * that the rt_ifa points to the address instead 1412 * of the loopback address. 1413 */ 1414 if (ifa != rt->rt_ifa) 1415 rt_replace_ifa(rt, ifa); 1416 rt->rt_flags &= ~RTF_CLONED; 1417 } 1418 } else if (rt->rt_flags & RTF_ANNOUNCE) { 1419 nd6_llinfo_settimer(ln, -1); 1420 ln->ln_state = ND6_LLINFO_REACHABLE; 1421 ln->ln_byhint = 0; 1422 1423 /* join solicited node multicast for proxy ND */ 1424 if (ifp->if_flags & IFF_MULTICAST) { 1425 struct in6_addr llsol; 1426 int error; 1427 1428 llsol = satocsin6(rt_getkey(rt))->sin6_addr; 1429 llsol.s6_addr32[0] = htonl(0xff020000); 1430 llsol.s6_addr32[1] = 0; 1431 llsol.s6_addr32[2] = htonl(1); 1432 llsol.s6_addr8[12] = 0xff; 1433 if (in6_setscope(&llsol, ifp, NULL)) 1434 break; 1435 if (!in6_addmulti(&llsol, ifp, &error, 0)) { 1436 nd6log((LOG_ERR, "%s: failed to join " 1437 "%s (errno=%d)\n", if_name(ifp), 1438 ip6_sprintf(&llsol), error)); 1439 } 1440 } 1441 } 1442 break; 1443 1444 case RTM_DELETE: 1445 if (ln == NULL) 1446 break; 1447 /* leave from solicited node multicast for proxy ND */ 1448 if ((rt->rt_flags & RTF_ANNOUNCE) != 0 && 1449 (ifp->if_flags & IFF_MULTICAST) != 0) { 1450 struct in6_addr llsol; 1451 struct in6_multi *in6m; 1452 1453 llsol = satocsin6(rt_getkey(rt))->sin6_addr; 1454 llsol.s6_addr32[0] = htonl(0xff020000); 1455 llsol.s6_addr32[1] = 0; 1456 llsol.s6_addr32[2] = htonl(1); 1457 llsol.s6_addr8[12] = 0xff; 1458 if (in6_setscope(&llsol, ifp, NULL) == 0) { 1459 IN6_LOOKUP_MULTI(llsol, ifp, in6m); 1460 if (in6m) 1461 in6_delmulti(in6m); 1462 } 1463 } 1464 nd6_inuse--; 1465 ln->ln_next->ln_prev = ln->ln_prev; 1466 ln->ln_prev->ln_next = ln->ln_next; 1467 ln->ln_prev = NULL; 1468 nd6_llinfo_settimer(ln, -1); 1469 rt->rt_llinfo = 0; 1470 rt->rt_flags &= ~RTF_LLINFO; 1471 clear_llinfo_pqueue(ln); 1472 Free(ln); 1473 } 1474} 1475 1476int 1477nd6_ioctl(u_long cmd, void *data, struct ifnet *ifp) 1478{ 1479 struct in6_drlist *drl = (struct in6_drlist *)data; 1480 struct in6_oprlist *oprl = (struct in6_oprlist *)data; 1481 struct in6_ndireq *ndi = (struct in6_ndireq *)data; 1482 struct in6_nbrinfo *nbi = (struct in6_nbrinfo *)data; 1483 struct in6_ndifreq *ndif = (struct in6_ndifreq *)data; 1484 struct nd_defrouter *dr; 1485 struct nd_prefix *pr; 1486 struct rtentry *rt; 1487 int i = 0, error = 0; 1488 int s; 1489 1490 switch (cmd) { 1491 case SIOCGDRLST_IN6: 1492 /* 1493 * obsolete API, use sysctl under net.inet6.icmp6 1494 */ 1495 memset(drl, 0, sizeof(*drl)); 1496 s = splsoftnet(); 1497 TAILQ_FOREACH(dr, &nd_defrouter, dr_entry) { 1498 if (i >= DRLSTSIZ) 1499 break; 1500 drl->defrouter[i].rtaddr = dr->rtaddr; 1501 in6_clearscope(&drl->defrouter[i].rtaddr); 1502 1503 drl->defrouter[i].flags = dr->flags; 1504 drl->defrouter[i].rtlifetime = dr->rtlifetime; 1505 drl->defrouter[i].expire = dr->expire; 1506 drl->defrouter[i].if_index = dr->ifp->if_index; 1507 i++; 1508 } 1509 splx(s); 1510 break; 1511 case SIOCGPRLST_IN6: 1512 /* 1513 * obsolete API, use sysctl under net.inet6.icmp6 1514 * 1515 * XXX the structure in6_prlist was changed in backward- 1516 * incompatible manner. in6_oprlist is used for SIOCGPRLST_IN6, 1517 * in6_prlist is used for nd6_sysctl() - fill_prlist(). 1518 */ 1519 /* 1520 * XXX meaning of fields, especialy "raflags", is very 1521 * differnet between RA prefix list and RR/static prefix list. 1522 * how about separating ioctls into two? 1523 */ 1524 memset(oprl, 0, sizeof(*oprl)); 1525 s = splsoftnet(); 1526 LIST_FOREACH(pr, &nd_prefix, ndpr_entry) { 1527 struct nd_pfxrouter *pfr; 1528 int j; 1529 1530 if (i >= PRLSTSIZ) 1531 break; 1532 oprl->prefix[i].prefix = pr->ndpr_prefix.sin6_addr; 1533 oprl->prefix[i].raflags = pr->ndpr_raf; 1534 oprl->prefix[i].prefixlen = pr->ndpr_plen; 1535 oprl->prefix[i].vltime = pr->ndpr_vltime; 1536 oprl->prefix[i].pltime = pr->ndpr_pltime; 1537 oprl->prefix[i].if_index = pr->ndpr_ifp->if_index; 1538 if (pr->ndpr_vltime == ND6_INFINITE_LIFETIME) 1539 oprl->prefix[i].expire = 0; 1540 else { 1541 time_t maxexpire; 1542 1543 /* XXX: we assume time_t is signed. */ 1544 maxexpire = (-1) & 1545 ~((time_t)1 << 1546 ((sizeof(maxexpire) * 8) - 1)); 1547 if (pr->ndpr_vltime < 1548 maxexpire - pr->ndpr_lastupdate) { 1549 oprl->prefix[i].expire = 1550 pr->ndpr_lastupdate + 1551 pr->ndpr_vltime; 1552 } else 1553 oprl->prefix[i].expire = maxexpire; 1554 } 1555 1556 j = 0; 1557 LIST_FOREACH(pfr, &pr->ndpr_advrtrs, pfr_entry) { 1558 if (j < DRLSTSIZ) { 1559#define RTRADDR oprl->prefix[i].advrtr[j] 1560 RTRADDR = pfr->router->rtaddr; 1561 in6_clearscope(&RTRADDR); 1562#undef RTRADDR 1563 } 1564 j++; 1565 } 1566 oprl->prefix[i].advrtrs = j; 1567 oprl->prefix[i].origin = PR_ORIG_RA; 1568 1569 i++; 1570 } 1571 splx(s); 1572 1573 break; 1574 case OSIOCGIFINFO_IN6: 1575#define ND ndi->ndi 1576 /* XXX: old ndp(8) assumes a positive value for linkmtu. */ 1577 memset(&ND, 0, sizeof(ND)); 1578 ND.linkmtu = IN6_LINKMTU(ifp); 1579 ND.maxmtu = ND_IFINFO(ifp)->maxmtu; 1580 ND.basereachable = ND_IFINFO(ifp)->basereachable; 1581 ND.reachable = ND_IFINFO(ifp)->reachable; 1582 ND.retrans = ND_IFINFO(ifp)->retrans; 1583 ND.flags = ND_IFINFO(ifp)->flags; 1584 ND.recalctm = ND_IFINFO(ifp)->recalctm; 1585 ND.chlim = ND_IFINFO(ifp)->chlim; 1586 break; 1587 case SIOCGIFINFO_IN6: 1588 ND = *ND_IFINFO(ifp); 1589 break; 1590 case SIOCSIFINFO_IN6: 1591 /* 1592 * used to change host variables from userland. 1593 * intented for a use on router to reflect RA configurations. 1594 */ 1595 /* 0 means 'unspecified' */ 1596 if (ND.linkmtu != 0) { 1597 if (ND.linkmtu < IPV6_MMTU || 1598 ND.linkmtu > IN6_LINKMTU(ifp)) { 1599 error = EINVAL; 1600 break; 1601 } 1602 ND_IFINFO(ifp)->linkmtu = ND.linkmtu; 1603 } 1604 1605 if (ND.basereachable != 0) { 1606 int obasereachable = ND_IFINFO(ifp)->basereachable; 1607 1608 ND_IFINFO(ifp)->basereachable = ND.basereachable; 1609 if (ND.basereachable != obasereachable) 1610 ND_IFINFO(ifp)->reachable = 1611 ND_COMPUTE_RTIME(ND.basereachable); 1612 } 1613 if (ND.retrans != 0) 1614 ND_IFINFO(ifp)->retrans = ND.retrans; 1615 if (ND.chlim != 0) 1616 ND_IFINFO(ifp)->chlim = ND.chlim; 1617 /* FALLTHROUGH */ 1618 case SIOCSIFINFO_FLAGS: 1619 ND_IFINFO(ifp)->flags = ND.flags; 1620 break; 1621#undef ND 1622 case SIOCSNDFLUSH_IN6: /* XXX: the ioctl name is confusing... */ 1623 /* sync kernel routing table with the default router list */ 1624 defrouter_reset(); 1625 defrouter_select(); 1626 break; 1627 case SIOCSPFXFLUSH_IN6: 1628 { 1629 /* flush all the prefix advertised by routers */ 1630 struct nd_prefix *pfx, *next; 1631 1632 s = splsoftnet(); 1633 LIST_FOREACH_SAFE(pfx, &nd_prefix, ndpr_entry, next) { 1634 struct in6_ifaddr *ia, *ia_next; 1635 1636 if (IN6_IS_ADDR_LINKLOCAL(&pfx->ndpr_prefix.sin6_addr)) 1637 continue; /* XXX */ 1638 1639 /* do we really have to remove addresses as well? */ 1640 for (ia = in6_ifaddr; ia; ia = ia_next) { 1641 /* ia might be removed. keep the next ptr. */ 1642 ia_next = ia->ia_next; 1643 1644 if ((ia->ia6_flags & IN6_IFF_AUTOCONF) == 0) 1645 continue; 1646 1647 if (ia->ia6_ndpr == pfx) 1648 in6_purgeaddr(&ia->ia_ifa); 1649 } 1650 prelist_remove(pfx); 1651 } 1652 splx(s); 1653 break; 1654 } 1655 case SIOCSRTRFLUSH_IN6: 1656 { 1657 /* flush all the default routers */ 1658 struct nd_defrouter *drtr, *next; 1659 1660 s = splsoftnet(); 1661 defrouter_reset(); 1662 TAILQ_FOREACH_SAFE(drtr, &nd_defrouter, dr_entry, next) { 1663 defrtrlist_del(drtr); 1664 } 1665 defrouter_select(); 1666 splx(s); 1667 break; 1668 } 1669 case SIOCGNBRINFO_IN6: 1670 { 1671 struct llinfo_nd6 *ln; 1672 struct in6_addr nb_addr = nbi->addr; /* make local for safety */ 1673 1674 if ((error = in6_setscope(&nb_addr, ifp, NULL)) != 0) 1675 return error; 1676 1677 s = splsoftnet(); 1678 if ((rt = nd6_lookup(&nb_addr, 0, ifp)) == NULL || 1679 (ln = (struct llinfo_nd6 *)rt->rt_llinfo) == NULL) { 1680 error = EINVAL; 1681 splx(s); 1682 break; 1683 } 1684 nbi->state = ln->ln_state; 1685 nbi->asked = ln->ln_asked; 1686 nbi->isrouter = ln->ln_router; 1687 nbi->expire = ln->ln_expire; 1688 splx(s); 1689 1690 break; 1691 } 1692 case SIOCGDEFIFACE_IN6: /* XXX: should be implemented as a sysctl? */ 1693 ndif->ifindex = nd6_defifindex; 1694 break; 1695 case SIOCSDEFIFACE_IN6: /* XXX: should be implemented as a sysctl? */ 1696 return nd6_setdefaultiface(ndif->ifindex); 1697 } 1698 return error; 1699} 1700 1701void 1702nd6_llinfo_release_pkts(struct llinfo_nd6 *ln, struct ifnet *ifp, 1703 struct rtentry *rt) 1704{ 1705 struct mbuf *m_hold, *m_hold_next; 1706 1707 for (m_hold = ln->ln_hold, ln->ln_hold = NULL; 1708 m_hold != NULL; 1709 m_hold = m_hold_next) { 1710 m_hold_next = m_hold->m_nextpkt; 1711 m_hold->m_nextpkt = NULL; 1712 1713 /* 1714 * we assume ifp is not a p2p here, so 1715 * just set the 2nd argument as the 1716 * 1st one. 1717 */ 1718 nd6_output(ifp, ifp, m_hold, satocsin6(rt_getkey(rt)), rt); 1719 } 1720} 1721 1722/* 1723 * Create neighbor cache entry and cache link-layer address, 1724 * on reception of inbound ND6 packets. (RS/RA/NS/redirect) 1725 */ 1726struct rtentry * 1727nd6_cache_lladdr( 1728 struct ifnet *ifp, 1729 struct in6_addr *from, 1730 char *lladdr, 1731 int lladdrlen, 1732 int type, /* ICMP6 type */ 1733 int code /* type dependent information */ 1734) 1735{ 1736 struct nd_ifinfo *ndi = ND_IFINFO(ifp); 1737 struct rtentry *rt = NULL; 1738 struct llinfo_nd6 *ln = NULL; 1739 int is_newentry; 1740 struct sockaddr_dl *sdl = NULL; 1741 int do_update; 1742 int olladdr; 1743 int llchange; 1744 int newstate = 0; 1745 1746 if (ifp == NULL) 1747 panic("ifp == NULL in nd6_cache_lladdr"); 1748 if (from == NULL) 1749 panic("from == NULL in nd6_cache_lladdr"); 1750 1751 /* nothing must be updated for unspecified address */ 1752 if (IN6_IS_ADDR_UNSPECIFIED(from)) 1753 return NULL; 1754 1755 /* 1756 * Validation about ifp->if_addrlen and lladdrlen must be done in 1757 * the caller. 1758 * 1759 * XXX If the link does not have link-layer adderss, what should 1760 * we do? (ifp->if_addrlen == 0) 1761 * Spec says nothing in sections for RA, RS and NA. There's small 1762 * description on it in NS section (RFC 2461 7.2.3). 1763 */ 1764 1765 rt = nd6_lookup(from, 0, ifp); 1766 if (rt == NULL) { 1767#if 0 1768 /* nothing must be done if there's no lladdr */ 1769 if (!lladdr || !lladdrlen) 1770 return NULL; 1771#endif 1772 1773 rt = nd6_lookup(from, 1, ifp); 1774 is_newentry = 1; 1775 } else { 1776 /* do nothing if static ndp is set */ 1777 if (rt->rt_flags & RTF_STATIC) 1778 return NULL; 1779 is_newentry = 0; 1780 } 1781 1782 if (rt == NULL) 1783 return NULL; 1784 if ((rt->rt_flags & (RTF_GATEWAY | RTF_LLINFO)) != RTF_LLINFO) { 1785fail: 1786 (void)nd6_free(rt, 0); 1787 return NULL; 1788 } 1789 ln = (struct llinfo_nd6 *)rt->rt_llinfo; 1790 if (ln == NULL) 1791 goto fail; 1792 if (rt->rt_gateway == NULL) 1793 goto fail; 1794 if (rt->rt_gateway->sa_family != AF_LINK) 1795 goto fail; 1796 sdl = satosdl(rt->rt_gateway); 1797 1798 olladdr = (sdl->sdl_alen) ? 1 : 0; 1799 if (olladdr && lladdr) { 1800 if (memcmp(lladdr, CLLADDR(sdl), ifp->if_addrlen)) 1801 llchange = 1; 1802 else 1803 llchange = 0; 1804 } else 1805 llchange = 0; 1806 1807 /* 1808 * newentry olladdr lladdr llchange (*=record) 1809 * 0 n n -- (1) 1810 * 0 y n -- (2) 1811 * 0 n y -- (3) * STALE 1812 * 0 y y n (4) * 1813 * 0 y y y (5) * STALE 1814 * 1 -- n -- (6) NOSTATE(= PASSIVE) 1815 * 1 -- y -- (7) * STALE 1816 */ 1817 1818 if (lladdr) { /* (3-5) and (7) */ 1819 /* 1820 * Record source link-layer address 1821 * XXX is it dependent to ifp->if_type? 1822 */ 1823 /* XXX check for error */ 1824 if (sockaddr_dl_setaddr(sdl, sdl->sdl_len, lladdr, 1825 ifp->if_addrlen) == NULL) { 1826 printf("%s.%d: sockaddr_dl_setaddr(, %d, ) " 1827 "failed on %s\n", __func__, __LINE__, 1828 sdl->sdl_len, if_name(ifp)); 1829 } 1830 } 1831 1832 if (!is_newentry) { 1833 if ((!olladdr && lladdr) || /* (3) */ 1834 (olladdr && lladdr && llchange)) { /* (5) */ 1835 do_update = 1; 1836 newstate = ND6_LLINFO_STALE; 1837 } else /* (1-2,4) */ 1838 do_update = 0; 1839 } else { 1840 do_update = 1; 1841 if (lladdr == NULL) /* (6) */ 1842 newstate = ND6_LLINFO_NOSTATE; 1843 else /* (7) */ 1844 newstate = ND6_LLINFO_STALE; 1845 } 1846 1847 if (do_update) { 1848 /* 1849 * Update the state of the neighbor cache. 1850 */ 1851 ln->ln_state = newstate; 1852 1853 if (ln->ln_state == ND6_LLINFO_STALE) { 1854 /* 1855 * XXX: since nd6_output() below will cause 1856 * state tansition to DELAY and reset the timer, 1857 * we must set the timer now, although it is actually 1858 * meaningless. 1859 */ 1860 nd6_llinfo_settimer(ln, (long)nd6_gctimer * hz); 1861 1862 nd6_llinfo_release_pkts(ln, ifp, rt); 1863 } else if (ln->ln_state == ND6_LLINFO_INCOMPLETE) { 1864 /* probe right away */ 1865 nd6_llinfo_settimer((void *)ln, 0); 1866 } 1867 } 1868 1869 /* 1870 * ICMP6 type dependent behavior. 1871 * 1872 * NS: clear IsRouter if new entry 1873 * RS: clear IsRouter 1874 * RA: set IsRouter if there's lladdr 1875 * redir: clear IsRouter if new entry 1876 * 1877 * RA case, (1): 1878 * The spec says that we must set IsRouter in the following cases: 1879 * - If lladdr exist, set IsRouter. This means (1-5). 1880 * - If it is old entry (!newentry), set IsRouter. This means (7). 1881 * So, based on the spec, in (1-5) and (7) cases we must set IsRouter. 1882 * A quetion arises for (1) case. (1) case has no lladdr in the 1883 * neighbor cache, this is similar to (6). 1884 * This case is rare but we figured that we MUST NOT set IsRouter. 1885 * 1886 * newentry olladdr lladdr llchange NS RS RA redir 1887 * D R 1888 * 0 n n -- (1) c ? s 1889 * 0 y n -- (2) c s s 1890 * 0 n y -- (3) c s s 1891 * 0 y y n (4) c s s 1892 * 0 y y y (5) c s s 1893 * 1 -- n -- (6) c c c s 1894 * 1 -- y -- (7) c c s c s 1895 * 1896 * (c=clear s=set) 1897 */ 1898 switch (type & 0xff) { 1899 case ND_NEIGHBOR_SOLICIT: 1900 /* 1901 * New entry must have is_router flag cleared. 1902 */ 1903 if (is_newentry) /* (6-7) */ 1904 ln->ln_router = 0; 1905 break; 1906 case ND_REDIRECT: 1907 /* 1908 * If the icmp is a redirect to a better router, always set the 1909 * is_router flag. Otherwise, if the entry is newly created, 1910 * clear the flag. [RFC 2461, sec 8.3] 1911 */ 1912 if (code == ND_REDIRECT_ROUTER) 1913 ln->ln_router = 1; 1914 else if (is_newentry) /* (6-7) */ 1915 ln->ln_router = 0; 1916 break; 1917 case ND_ROUTER_SOLICIT: 1918 /* 1919 * is_router flag must always be cleared. 1920 */ 1921 ln->ln_router = 0; 1922 break; 1923 case ND_ROUTER_ADVERT: 1924 /* 1925 * Mark an entry with lladdr as a router. 1926 */ 1927 if ((!is_newentry && (olladdr || lladdr)) || /* (2-5) */ 1928 (is_newentry && lladdr)) { /* (7) */ 1929 ln->ln_router = 1; 1930 } 1931 break; 1932 } 1933 1934 /* 1935 * When the link-layer address of a router changes, select the 1936 * best router again. In particular, when the neighbor entry is newly 1937 * created, it might affect the selection policy. 1938 * Question: can we restrict the first condition to the "is_newentry" 1939 * case? 1940 * XXX: when we hear an RA from a new router with the link-layer 1941 * address option, defrouter_select() is called twice, since 1942 * defrtrlist_update called the function as well. However, I believe 1943 * we can compromise the overhead, since it only happens the first 1944 * time. 1945 * XXX: although defrouter_select() should not have a bad effect 1946 * for those are not autoconfigured hosts, we explicitly avoid such 1947 * cases for safety. 1948 */ 1949 if (do_update && ln->ln_router && !ip6_forwarding && 1950 nd6_accepts_rtadv(ndi)) 1951 defrouter_select(); 1952 1953 return rt; 1954} 1955 1956static void 1957nd6_slowtimo(void *ignored_arg) 1958{ 1959 struct nd_ifinfo *nd6if; 1960 struct ifnet *ifp; 1961 1962 mutex_enter(softnet_lock); 1963 KERNEL_LOCK(1, NULL); 1964 callout_reset(&nd6_slowtimo_ch, ND6_SLOWTIMER_INTERVAL * hz, 1965 nd6_slowtimo, NULL); 1966 TAILQ_FOREACH(ifp, &ifnet, if_list) { 1967 nd6if = ND_IFINFO(ifp); 1968 if (nd6if->basereachable && /* already initialized */ 1969 (nd6if->recalctm -= ND6_SLOWTIMER_INTERVAL) <= 0) { 1970 /* 1971 * Since reachable time rarely changes by router 1972 * advertisements, we SHOULD insure that a new random 1973 * value gets recomputed at least once every few hours. 1974 * (RFC 2461, 6.3.4) 1975 */ 1976 nd6if->recalctm = nd6_recalc_reachtm_interval; 1977 nd6if->reachable = ND_COMPUTE_RTIME(nd6if->basereachable); 1978 } 1979 } 1980 KERNEL_UNLOCK_ONE(NULL); 1981 mutex_exit(softnet_lock); 1982} 1983 1984#define senderr(e) { error = (e); goto bad;} 1985int 1986nd6_output(struct ifnet *ifp, struct ifnet *origifp, struct mbuf *m0, 1987 const struct sockaddr_in6 *dst, struct rtentry *rt0) 1988{ 1989 struct mbuf *m = m0; 1990 struct rtentry *rt = rt0; 1991 struct sockaddr_in6 *gw6 = NULL; 1992 struct llinfo_nd6 *ln = NULL; 1993 int error = 0; 1994 1995 if (IN6_IS_ADDR_MULTICAST(&dst->sin6_addr)) 1996 goto sendpkt; 1997 1998 if (nd6_need_cache(ifp) == 0) 1999 goto sendpkt; 2000 2001 /* 2002 * next hop determination. This routine is derived from ether_output. 2003 */ 2004 if (rt) { 2005 if ((rt->rt_flags & RTF_UP) == 0) { 2006 if ((rt0 = rt = rtalloc1(sin6tocsa(dst), 1)) != NULL) { 2007 rt->rt_refcnt--; 2008 if (rt->rt_ifp != ifp) 2009 senderr(EHOSTUNREACH); 2010 } else 2011 senderr(EHOSTUNREACH); 2012 } 2013 2014 if (rt->rt_flags & RTF_GATEWAY) { 2015 gw6 = (struct sockaddr_in6 *)rt->rt_gateway; 2016 2017 /* 2018 * We skip link-layer address resolution and NUD 2019 * if the gateway is not a neighbor from ND point 2020 * of view, regardless of the value of nd_ifinfo.flags. 2021 * The second condition is a bit tricky; we skip 2022 * if the gateway is our own address, which is 2023 * sometimes used to install a route to a p2p link. 2024 */ 2025 if (!nd6_is_addr_neighbor(gw6, ifp) || 2026 in6ifa_ifpwithaddr(ifp, &gw6->sin6_addr)) { 2027 /* 2028 * We allow this kind of tricky route only 2029 * when the outgoing interface is p2p. 2030 * XXX: we may need a more generic rule here. 2031 */ 2032 if ((ifp->if_flags & IFF_POINTOPOINT) == 0) 2033 senderr(EHOSTUNREACH); 2034 2035 goto sendpkt; 2036 } 2037 2038 if (rt->rt_gwroute == NULL) 2039 goto lookup; 2040 if (((rt = rt->rt_gwroute)->rt_flags & RTF_UP) == 0) { 2041 rtfree(rt); rt = rt0; 2042 lookup: 2043 rt->rt_gwroute = rtalloc1(rt->rt_gateway, 1); 2044 if ((rt = rt->rt_gwroute) == NULL) 2045 senderr(EHOSTUNREACH); 2046 /* the "G" test below also prevents rt == rt0 */ 2047 if ((rt->rt_flags & RTF_GATEWAY) || 2048 (rt->rt_ifp != ifp)) { 2049 rt->rt_refcnt--; 2050 rt0->rt_gwroute = NULL; 2051 senderr(EHOSTUNREACH); 2052 } 2053 } 2054 } 2055 } 2056 2057 /* 2058 * Address resolution or Neighbor Unreachability Detection 2059 * for the next hop. 2060 * At this point, the destination of the packet must be a unicast 2061 * or an anycast address(i.e. not a multicast). 2062 */ 2063 2064 /* Look up the neighbor cache for the nexthop */ 2065 if (rt != NULL && (rt->rt_flags & RTF_LLINFO) != 0) 2066 ln = (struct llinfo_nd6 *)rt->rt_llinfo; 2067 else { 2068 /* 2069 * Since nd6_is_addr_neighbor() internally calls nd6_lookup(), 2070 * the condition below is not very efficient. But we believe 2071 * it is tolerable, because this should be a rare case. 2072 */ 2073 if (nd6_is_addr_neighbor(dst, ifp) && 2074 (rt = nd6_lookup(&dst->sin6_addr, 1, ifp)) != NULL) 2075 ln = (struct llinfo_nd6 *)rt->rt_llinfo; 2076 } 2077 if (ln == NULL || rt == NULL) { 2078 if ((ifp->if_flags & IFF_POINTOPOINT) == 0 && 2079 !(ND_IFINFO(ifp)->flags & ND6_IFF_PERFORMNUD)) { 2080 log(LOG_DEBUG, 2081 "nd6_output: can't allocate llinfo for %s " 2082 "(ln=%p, rt=%p)\n", 2083 ip6_sprintf(&dst->sin6_addr), ln, rt); 2084 senderr(EIO); /* XXX: good error? */ 2085 } 2086 2087 goto sendpkt; /* send anyway */ 2088 } 2089 2090 /* 2091 * Move this entry to the head of the queue so that it is less likely 2092 * for this entry to be a target of forced garbage collection (see 2093 * nd6_rtrequest()). 2094 */ 2095 LN_DEQUEUE(ln); 2096 LN_INSERTHEAD(ln); 2097 2098 /* We don't have to do link-layer address resolution on a p2p link. */ 2099 if ((ifp->if_flags & IFF_POINTOPOINT) != 0 && 2100 ln->ln_state < ND6_LLINFO_REACHABLE) { 2101 ln->ln_state = ND6_LLINFO_STALE; 2102 nd6_llinfo_settimer(ln, (long)nd6_gctimer * hz); 2103 } 2104 2105 /* 2106 * The first time we send a packet to a neighbor whose entry is 2107 * STALE, we have to change the state to DELAY and a sets a timer to 2108 * expire in DELAY_FIRST_PROBE_TIME seconds to ensure do 2109 * neighbor unreachability detection on expiration. 2110 * (RFC 2461 7.3.3) 2111 */ 2112 if (ln->ln_state == ND6_LLINFO_STALE) { 2113 ln->ln_asked = 0; 2114 ln->ln_state = ND6_LLINFO_DELAY; 2115 nd6_llinfo_settimer(ln, (long)nd6_delay * hz); 2116 } 2117 2118 /* 2119 * If the neighbor cache entry has a state other than INCOMPLETE 2120 * (i.e. its link-layer address is already resolved), just 2121 * send the packet. 2122 */ 2123 if (ln->ln_state > ND6_LLINFO_INCOMPLETE) 2124 goto sendpkt; 2125 2126 /* 2127 * There is a neighbor cache entry, but no ethernet address 2128 * response yet. Append this latest packet to the end of the 2129 * packet queue in the mbuf, unless the number of the packet 2130 * does not exceed nd6_maxqueuelen. When it exceeds nd6_maxqueuelen, 2131 * the oldest packet in the queue will be removed. 2132 */ 2133 if (ln->ln_state == ND6_LLINFO_NOSTATE) 2134 ln->ln_state = ND6_LLINFO_INCOMPLETE; 2135 if (ln->ln_hold) { 2136 struct mbuf *m_hold; 2137 int i; 2138 2139 i = 0; 2140 for (m_hold = ln->ln_hold; m_hold; m_hold = m_hold->m_nextpkt) { 2141 i++; 2142 if (m_hold->m_nextpkt == NULL) { 2143 m_hold->m_nextpkt = m; 2144 break; 2145 } 2146 } 2147 while (i >= nd6_maxqueuelen) { 2148 m_hold = ln->ln_hold; 2149 ln->ln_hold = ln->ln_hold->m_nextpkt; 2150 m_freem(m_hold); 2151 i--; 2152 } 2153 } else { 2154 ln->ln_hold = m; 2155 } 2156 2157 /* 2158 * If there has been no NS for the neighbor after entering the 2159 * INCOMPLETE state, send the first solicitation. 2160 */ 2161 if (!ND6_LLINFO_PERMANENT(ln) && ln->ln_asked == 0) { 2162 ln->ln_asked++; 2163 nd6_llinfo_settimer(ln, 2164 (long)ND_IFINFO(ifp)->retrans * hz / 1000); 2165 nd6_ns_output(ifp, NULL, &dst->sin6_addr, ln, 0); 2166 } 2167 return 0; 2168 2169 sendpkt: 2170 /* discard the packet if IPv6 operation is disabled on the interface */ 2171 if ((ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED)) { 2172 error = ENETDOWN; /* better error? */ 2173 goto bad; 2174 } 2175 2176#ifdef KAME_IPSEC 2177 /* clean ipsec history once it goes out of the node */ 2178 ipsec_delaux(m); 2179#endif 2180 KERNEL_LOCK(1, NULL); 2181 if ((ifp->if_flags & IFF_LOOPBACK) != 0) 2182 error = (*ifp->if_output)(origifp, m, sin6tocsa(dst), rt); 2183 else 2184 error = (*ifp->if_output)(ifp, m, sin6tocsa(dst), rt); 2185 KERNEL_UNLOCK_ONE(NULL); 2186 return error; 2187 2188 bad: 2189 if (m != NULL) 2190 m_freem(m); 2191 return error; 2192} 2193#undef senderr 2194 2195int 2196nd6_need_cache(struct ifnet *ifp) 2197{ 2198 /* 2199 * XXX: we currently do not make neighbor cache on any interface 2200 * other than ARCnet, Ethernet, FDDI and GIF. 2201 * 2202 * RFC2893 says: 2203 * - unidirectional tunnels needs no ND 2204 */ 2205 switch (ifp->if_type) { 2206 case IFT_ARCNET: 2207 case IFT_ETHER: 2208 case IFT_FDDI: 2209 case IFT_IEEE1394: 2210 case IFT_CARP: 2211 case IFT_GIF: /* XXX need more cases? */ 2212 case IFT_PPP: 2213 case IFT_TUNNEL: 2214 return 1; 2215 default: 2216 return 0; 2217 } 2218} 2219 2220int 2221nd6_storelladdr(const struct ifnet *ifp, const struct rtentry *rt, 2222 struct mbuf *m, const struct sockaddr *dst, uint8_t *lldst, 2223 size_t dstsize) 2224{ 2225 const struct sockaddr_dl *sdl; 2226 2227 if (m->m_flags & M_MCAST) { 2228 switch (ifp->if_type) { 2229 case IFT_ETHER: 2230 case IFT_FDDI: 2231 ETHER_MAP_IPV6_MULTICAST(&satocsin6(dst)->sin6_addr, 2232 lldst); 2233 return 1; 2234 case IFT_IEEE1394: 2235 memcpy(lldst, ifp->if_broadcastaddr, 2236 MIN(dstsize, ifp->if_addrlen)); 2237 return 1; 2238 case IFT_ARCNET: 2239 *lldst = 0; 2240 return 1; 2241 default: 2242 m_freem(m); 2243 return 0; 2244 } 2245 } 2246 2247 if (rt == NULL) { 2248 /* this could happen, if we could not allocate memory */ 2249 m_freem(m); 2250 return 0; 2251 } 2252 if (rt->rt_gateway->sa_family != AF_LINK) { 2253 printf("%s: something odd happens\n", __func__); 2254 m_freem(m); 2255 return 0; 2256 } 2257 sdl = satocsdl(rt->rt_gateway); 2258 if (sdl->sdl_alen == 0 || sdl->sdl_alen > dstsize) { 2259 /* this should be impossible, but we bark here for debugging */ 2260 printf("%s: sdl_alen == %" PRIu8 ", dst=%s, if=%s\n", __func__, 2261 sdl->sdl_alen, ip6_sprintf(&satocsin6(dst)->sin6_addr), 2262 if_name(ifp)); 2263 m_freem(m); 2264 return 0; 2265 } 2266 2267 memcpy(lldst, CLLADDR(sdl), MIN(dstsize, sdl->sdl_alen)); 2268 return 1; 2269} 2270 2271static void 2272clear_llinfo_pqueue(struct llinfo_nd6 *ln) 2273{ 2274 struct mbuf *m_hold, *m_hold_next; 2275 2276 for (m_hold = ln->ln_hold; m_hold; m_hold = m_hold_next) { 2277 m_hold_next = m_hold->m_nextpkt; 2278 m_hold->m_nextpkt = NULL; 2279 m_freem(m_hold); 2280 } 2281 2282 ln->ln_hold = NULL; 2283 return; 2284} 2285 2286int 2287nd6_sysctl( 2288 int name, 2289 void *oldp, /* syscall arg, need copyout */ 2290 size_t *oldlenp, 2291 void *newp, /* syscall arg, need copyin */ 2292 size_t newlen 2293) 2294{ 2295 void *p; 2296 size_t ol; 2297 int error; 2298 2299 error = 0; 2300 2301 if (newp) 2302 return EPERM; 2303 if (oldp && !oldlenp) 2304 return EINVAL; 2305 ol = oldlenp ? *oldlenp : 0; 2306 2307 if (oldp) { 2308 p = malloc(*oldlenp, M_TEMP, M_WAITOK); 2309 if (p == NULL) 2310 return ENOMEM; 2311 } else 2312 p = NULL; 2313 switch (name) { 2314 case ICMPV6CTL_ND6_DRLIST: 2315 error = fill_drlist(p, oldlenp, ol); 2316 if (!error && p != NULL && oldp != NULL) 2317 error = copyout(p, oldp, *oldlenp); 2318 break; 2319 2320 case ICMPV6CTL_ND6_PRLIST: 2321 error = fill_prlist(p, oldlenp, ol); 2322 if (!error && p != NULL && oldp != NULL) 2323 error = copyout(p, oldp, *oldlenp); 2324 break; 2325 2326 case ICMPV6CTL_ND6_MAXQLEN: 2327 break; 2328 2329 default: 2330 error = ENOPROTOOPT; 2331 break; 2332 } 2333 if (p) 2334 free(p, M_TEMP); 2335 2336 return error; 2337} 2338 2339static int 2340fill_drlist(void *oldp, size_t *oldlenp, size_t ol) 2341{ 2342 int error = 0, s; 2343 struct in6_defrouter *d = NULL, *de = NULL; 2344 struct nd_defrouter *dr; 2345 size_t l; 2346 2347 s = splsoftnet(); 2348 2349 if (oldp) { 2350 d = (struct in6_defrouter *)oldp; 2351 de = (struct in6_defrouter *)((char *)oldp + *oldlenp); 2352 } 2353 l = 0; 2354 2355 TAILQ_FOREACH(dr, &nd_defrouter, dr_entry) { 2356 2357 if (oldp && d + 1 <= de) { 2358 memset(d, 0, sizeof(*d)); 2359 sockaddr_in6_init(&d->rtaddr, &dr->rtaddr, 0, 0, 0); 2360 if (sa6_recoverscope(&d->rtaddr)) { 2361 log(LOG_ERR, 2362 "scope error in router list (%s)\n", 2363 ip6_sprintf(&d->rtaddr.sin6_addr)); 2364 /* XXX: press on... */ 2365 } 2366 d->flags = dr->flags; 2367 d->rtlifetime = dr->rtlifetime; 2368 d->expire = dr->expire; 2369 d->if_index = dr->ifp->if_index; 2370 } 2371 2372 l += sizeof(*d); 2373 if (d) 2374 d++; 2375 } 2376 2377 if (oldp) { 2378 if (l > ol) 2379 error = ENOMEM; 2380 } 2381 if (oldlenp) 2382 *oldlenp = l; /* (void *)d - (void *)oldp */ 2383 2384 splx(s); 2385 2386 return error; 2387} 2388 2389static int 2390fill_prlist(void *oldp, size_t *oldlenp, size_t ol) 2391{ 2392 int error = 0, s; 2393 struct nd_prefix *pr; 2394 uint8_t *p = NULL, *ps = NULL; 2395 uint8_t *pe = NULL; 2396 size_t l; 2397 2398 s = splsoftnet(); 2399 2400 if (oldp) { 2401 ps = p = (uint8_t*)oldp; 2402 pe = (uint8_t*)oldp + *oldlenp; 2403 } 2404 l = 0; 2405 2406 LIST_FOREACH(pr, &nd_prefix, ndpr_entry) { 2407 u_short advrtrs; 2408 struct sockaddr_in6 sin6; 2409 struct nd_pfxrouter *pfr; 2410 struct in6_prefix pfx; 2411 2412 if (oldp && p + sizeof(struct in6_prefix) <= pe) 2413 { 2414 memset(&pfx, 0, sizeof(pfx)); 2415 ps = p; 2416 pfx.prefix = pr->ndpr_prefix; 2417 2418 if (sa6_recoverscope(&pfx.prefix)) { 2419 log(LOG_ERR, 2420 "scope error in prefix list (%s)\n", 2421 ip6_sprintf(&pfx.prefix.sin6_addr)); 2422 /* XXX: press on... */ 2423 } 2424 pfx.raflags = pr->ndpr_raf; 2425 pfx.prefixlen = pr->ndpr_plen; 2426 pfx.vltime = pr->ndpr_vltime; 2427 pfx.pltime = pr->ndpr_pltime; 2428 pfx.if_index = pr->ndpr_ifp->if_index; 2429 if (pr->ndpr_vltime == ND6_INFINITE_LIFETIME) 2430 pfx.expire = 0; 2431 else { 2432 time_t maxexpire; 2433 2434 /* XXX: we assume time_t is signed. */ 2435 maxexpire = (-1) & 2436 ~((time_t)1 << 2437 ((sizeof(maxexpire) * 8) - 1)); 2438 if (pr->ndpr_vltime < 2439 maxexpire - pr->ndpr_lastupdate) { 2440 pfx.expire = pr->ndpr_lastupdate + 2441 pr->ndpr_vltime; 2442 } else 2443 pfx.expire = maxexpire; 2444 } 2445 pfx.refcnt = pr->ndpr_refcnt; 2446 pfx.flags = pr->ndpr_stateflags; 2447 pfx.origin = PR_ORIG_RA; 2448 2449 p += sizeof(pfx); l += sizeof(pfx); 2450 2451 advrtrs = 0; 2452 LIST_FOREACH(pfr, &pr->ndpr_advrtrs, pfr_entry) { 2453 if (p + sizeof(sin6) > pe) { 2454 advrtrs++; 2455 continue; 2456 } 2457 2458 sockaddr_in6_init(&sin6, &pfr->router->rtaddr, 2459 0, 0, 0); 2460 if (sa6_recoverscope(&sin6)) { 2461 log(LOG_ERR, 2462 "scope error in " 2463 "prefix list (%s)\n", 2464 ip6_sprintf(&pfr->router->rtaddr)); 2465 } 2466 advrtrs++; 2467 memcpy(p, &sin6, sizeof(sin6)); 2468 p += sizeof(sin6); 2469 l += sizeof(sin6); 2470 } 2471 pfx.advrtrs = advrtrs; 2472 memcpy(ps, &pfx, sizeof(pfx)); 2473 } 2474 else { 2475 l += sizeof(pfx); 2476 advrtrs = 0; 2477 LIST_FOREACH(pfr, &pr->ndpr_advrtrs, pfr_entry) { 2478 advrtrs++; 2479 l += sizeof(sin6); 2480 } 2481 } 2482 } 2483 2484 if (oldp) { 2485 *oldlenp = l; /* (void *)d - (void *)oldp */ 2486 if (l > ol) 2487 error = ENOMEM; 2488 } else 2489 *oldlenp = l; 2490 2491 splx(s); 2492 2493 return error; 2494} 2495