1 /* $FreeBSD$ */ 2 /* $KAME: nd6.c,v 1.144 2001/05/24 07:44:00 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 "opt_inet.h" 34 #include "opt_inet6.h" 35 #include "opt_mac.h" 36 37 #include <sys/param.h> 38 #include <sys/systm.h> 39 #include <sys/callout.h> 40 #include <sys/malloc.h> 41 #include <sys/mbuf.h> 42 #include <sys/socket.h> 43 #include <sys/sockio.h> 44 #include <sys/time.h> 45 #include <sys/kernel.h> 46 #include <sys/protosw.h> 47 #include <sys/errno.h> 48 #include <sys/syslog.h> 49 #include <sys/queue.h> 50 #include <sys/sysctl.h> 51 52 #include <net/if.h> 53 #include <net/if_arc.h> 54 #include <net/if_dl.h> 55 #include <net/if_types.h> 56 #include <net/iso88025.h> 57 #include <net/fddi.h> 58 #include <net/route.h> 59 60 #include <netinet/in.h> 61 #include <netinet/if_ether.h> 62 #include <netinet6/in6_var.h> 63 #include <netinet/ip6.h> 64 #include <netinet6/ip6_var.h> 65 #include <netinet6/scope6_var.h> 66 #include <netinet6/nd6.h> 67 #include <netinet/icmp6.h> 68 69 #include <sys/limits.h> 70 71 #include <security/mac/mac_framework.h> 72 73 #define ND6_SLOWTIMER_INTERVAL (60 * 60) /* 1 hour */ 74 #define ND6_RECALC_REACHTM_INTERVAL (60 * 120) /* 2 hours */ 75 76 #define SIN6(s) ((struct sockaddr_in6 *)s) 77 #define SDL(s) ((struct sockaddr_dl *)s) 78 79 /* timer values */ 80 int nd6_prune = 1; /* walk list every 1 seconds */ 81 int nd6_delay = 5; /* delay first probe time 5 second */ 82 int nd6_umaxtries = 3; /* maximum unicast query */ 83 int nd6_mmaxtries = 3; /* maximum multicast query */ 84 int nd6_useloopback = 1; /* use loopback interface for local traffic */ 85 int nd6_gctimer = (60 * 60 * 24); /* 1 day: garbage collection timer */ 86 87 /* preventing too many loops in ND option parsing */ 88 int nd6_maxndopt = 10; /* max # of ND options allowed */ 89 90 int nd6_maxnudhint = 0; /* max # of subsequent upper layer hints */ 91 int nd6_maxqueuelen = 1; /* max # of packets cached in unresolved ND entries */ 92 93 #ifdef ND6_DEBUG 94 int nd6_debug = 1; 95 #else 96 int nd6_debug = 0; 97 #endif 98 99 /* for debugging? */ 100 static int nd6_inuse, nd6_allocated; 101 102 struct llinfo_nd6 llinfo_nd6 = {&llinfo_nd6, &llinfo_nd6}; 103 struct nd_drhead nd_defrouter; 104 struct nd_prhead nd_prefix = { 0 }; 105 106 int nd6_recalc_reachtm_interval = ND6_RECALC_REACHTM_INTERVAL; 107 static struct sockaddr_in6 all1_sa; 108 109 static int nd6_is_new_addr_neighbor __P((struct sockaddr_in6 *, 110 struct ifnet *)); 111 static void nd6_setmtu0 __P((struct ifnet *, struct nd_ifinfo *)); 112 static void nd6_slowtimo __P((void *)); 113 static int regen_tmpaddr __P((struct in6_ifaddr *)); 114 static struct llinfo_nd6 *nd6_free __P((struct rtentry *, int)); 115 static void nd6_llinfo_timer __P((void *)); 116 static void clear_llinfo_pqueue __P((struct llinfo_nd6 *)); 117 118 struct callout nd6_slowtimo_ch; 119 struct callout nd6_timer_ch; 120 extern struct callout in6_tmpaddrtimer_ch; 121 122 void 123 nd6_init() 124 { 125 static int nd6_init_done = 0; 126 int i; 127 128 if (nd6_init_done) { 129 log(LOG_NOTICE, "nd6_init called more than once(ignored)\n"); 130 return; 131 } 132 133 all1_sa.sin6_family = AF_INET6; 134 all1_sa.sin6_len = sizeof(struct sockaddr_in6); 135 for (i = 0; i < sizeof(all1_sa.sin6_addr); i++) 136 all1_sa.sin6_addr.s6_addr[i] = 0xff; 137 138 /* initialization of the default router list */ 139 TAILQ_INIT(&nd_defrouter); 140 141 nd6_init_done = 1; 142 143 /* start timer */ 144 callout_init(&nd6_slowtimo_ch, 0); 145 callout_reset(&nd6_slowtimo_ch, ND6_SLOWTIMER_INTERVAL * hz, 146 nd6_slowtimo, NULL); 147 } 148 149 struct nd_ifinfo * 150 nd6_ifattach(ifp) 151 struct ifnet *ifp; 152 { 153 struct nd_ifinfo *nd; 154 155 nd = (struct nd_ifinfo *)malloc(sizeof(*nd), M_IP6NDP, M_WAITOK); 156 bzero(nd, sizeof(*nd)); 157 158 nd->initialized = 1; 159 160 nd->chlim = IPV6_DEFHLIM; 161 nd->basereachable = REACHABLE_TIME; 162 nd->reachable = ND_COMPUTE_RTIME(nd->basereachable); 163 nd->retrans = RETRANS_TIMER; 164 /* 165 * Note that the default value of ip6_accept_rtadv is 0, which means 166 * we won't accept RAs by default even if we set ND6_IFF_ACCEPT_RTADV 167 * here. 168 */ 169 nd->flags = (ND6_IFF_PERFORMNUD | ND6_IFF_ACCEPT_RTADV); 170 171 /* XXX: we cannot call nd6_setmtu since ifp is not fully initialized */ 172 nd6_setmtu0(ifp, nd); 173 174 return nd; 175 } 176 177 void 178 nd6_ifdetach(nd) 179 struct nd_ifinfo *nd; 180 { 181 182 free(nd, M_IP6NDP); 183 } 184 185 /* 186 * Reset ND level link MTU. This function is called when the physical MTU 187 * changes, which means we might have to adjust the ND level MTU. 188 */ 189 void 190 nd6_setmtu(ifp) 191 struct ifnet *ifp; 192 { 193 194 nd6_setmtu0(ifp, ND_IFINFO(ifp)); 195 } 196 197 /* XXX todo: do not maintain copy of ifp->if_mtu in ndi->maxmtu */ 198 void 199 nd6_setmtu0(ifp, ndi) 200 struct ifnet *ifp; 201 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); /* RFC2467 */ 213 break; 214 case IFT_ISO88025: 215 ndi->maxmtu = MIN(ISO88025_MAX_MTU, ifp->if_mtu); 216 break; 217 default: 218 ndi->maxmtu = ifp->if_mtu; 219 break; 220 } 221 222 /* 223 * Decreasing the interface MTU under IPV6 minimum MTU may cause 224 * undesirable situation. We thus notify the operator of the change 225 * explicitly. The check for omaxmtu is necessary to restrict the 226 * log to the case of changing the MTU, not initializing it. 227 */ 228 if (omaxmtu >= IPV6_MMTU && ndi->maxmtu < IPV6_MMTU) { 229 log(LOG_NOTICE, "nd6_setmtu0: " 230 "new link MTU on %s (%lu) is too small for IPv6\n", 231 if_name(ifp), (unsigned long)ndi->maxmtu); 232 } 233 234 if (ndi->maxmtu > in6_maxmtu) 235 in6_setmaxmtu(); /* check all interfaces just in case */ 236 237 #undef MIN 238 } 239 240 void 241 nd6_option_init(opt, icmp6len, ndopts) 242 void *opt; 243 int icmp6len; 244 union nd_opts *ndopts; 245 { 246 247 bzero(ndopts, sizeof(*ndopts)); 248 ndopts->nd_opts_search = (struct nd_opt_hdr *)opt; 249 ndopts->nd_opts_last 250 = (struct nd_opt_hdr *)(((u_char *)opt) + icmp6len); 251 252 if (icmp6len == 0) { 253 ndopts->nd_opts_done = 1; 254 ndopts->nd_opts_search = NULL; 255 } 256 } 257 258 /* 259 * Take one ND option. 260 */ 261 struct nd_opt_hdr * 262 nd6_option(ndopts) 263 union nd_opts *ndopts; 264 { 265 struct nd_opt_hdr *nd_opt; 266 int olen; 267 268 if (ndopts == NULL) 269 panic("ndopts == NULL in nd6_option"); 270 if (ndopts->nd_opts_last == NULL) 271 panic("uninitialized ndopts in nd6_option"); 272 if (ndopts->nd_opts_search == NULL) 273 return NULL; 274 if (ndopts->nd_opts_done) 275 return NULL; 276 277 nd_opt = ndopts->nd_opts_search; 278 279 /* make sure nd_opt_len is inside the buffer */ 280 if ((caddr_t)&nd_opt->nd_opt_len >= (caddr_t)ndopts->nd_opts_last) { 281 bzero(ndopts, sizeof(*ndopts)); 282 return NULL; 283 } 284 285 olen = nd_opt->nd_opt_len << 3; 286 if (olen == 0) { 287 /* 288 * Message validation requires that all included 289 * options have a length that is greater than zero. 290 */ 291 bzero(ndopts, sizeof(*ndopts)); 292 return NULL; 293 } 294 295 ndopts->nd_opts_search = (struct nd_opt_hdr *)((caddr_t)nd_opt + olen); 296 if (ndopts->nd_opts_search > ndopts->nd_opts_last) { 297 /* option overruns the end of buffer, invalid */ 298 bzero(ndopts, sizeof(*ndopts)); 299 return NULL; 300 } else if (ndopts->nd_opts_search == ndopts->nd_opts_last) { 301 /* reached the end of options chain */ 302 ndopts->nd_opts_done = 1; 303 ndopts->nd_opts_search = NULL; 304 } 305 return nd_opt; 306 } 307 308 /* 309 * Parse multiple ND options. 310 * This function is much easier to use, for ND routines that do not need 311 * multiple options of the same type. 312 */ 313 int 314 nd6_options(ndopts) 315 union nd_opts *ndopts; 316 { 317 struct nd_opt_hdr *nd_opt; 318 int i = 0; 319 320 if (ndopts == NULL) 321 panic("ndopts == NULL in nd6_options"); 322 if (ndopts->nd_opts_last == NULL) 323 panic("uninitialized ndopts in nd6_options"); 324 if (ndopts->nd_opts_search == NULL) 325 return 0; 326 327 while (1) { 328 nd_opt = nd6_option(ndopts); 329 if (nd_opt == NULL && ndopts->nd_opts_last == NULL) { 330 /* 331 * Message validation requires that all included 332 * options have a length that is greater than zero. 333 */ 334 icmp6stat.icp6s_nd_badopt++; 335 bzero(ndopts, sizeof(*ndopts)); 336 return -1; 337 } 338 339 if (nd_opt == NULL) 340 goto skip1; 341 342 switch (nd_opt->nd_opt_type) { 343 case ND_OPT_SOURCE_LINKADDR: 344 case ND_OPT_TARGET_LINKADDR: 345 case ND_OPT_MTU: 346 case ND_OPT_REDIRECTED_HEADER: 347 if (ndopts->nd_opt_array[nd_opt->nd_opt_type]) { 348 nd6log((LOG_INFO, 349 "duplicated ND6 option found (type=%d)\n", 350 nd_opt->nd_opt_type)); 351 /* XXX bark? */ 352 } else { 353 ndopts->nd_opt_array[nd_opt->nd_opt_type] 354 = nd_opt; 355 } 356 break; 357 case ND_OPT_PREFIX_INFORMATION: 358 if (ndopts->nd_opt_array[nd_opt->nd_opt_type] == 0) { 359 ndopts->nd_opt_array[nd_opt->nd_opt_type] 360 = nd_opt; 361 } 362 ndopts->nd_opts_pi_end = 363 (struct nd_opt_prefix_info *)nd_opt; 364 break; 365 default: 366 /* 367 * Unknown options must be silently ignored, 368 * to accomodate future extension to the protocol. 369 */ 370 nd6log((LOG_DEBUG, 371 "nd6_options: unsupported option %d - " 372 "option ignored\n", nd_opt->nd_opt_type)); 373 } 374 375 skip1: 376 i++; 377 if (i > nd6_maxndopt) { 378 icmp6stat.icp6s_nd_toomanyopt++; 379 nd6log((LOG_INFO, "too many loop in nd opt\n")); 380 break; 381 } 382 383 if (ndopts->nd_opts_done) 384 break; 385 } 386 387 return 0; 388 } 389 390 /* 391 * ND6 timer routine to handle ND6 entries 392 */ 393 void 394 nd6_llinfo_settimer(ln, tick) 395 struct llinfo_nd6 *ln; 396 long tick; 397 { 398 if (tick < 0) { 399 ln->ln_expire = 0; 400 ln->ln_ntick = 0; 401 callout_stop(&ln->ln_timer_ch); 402 } else { 403 ln->ln_expire = time_second + tick / hz; 404 if (tick > INT_MAX) { 405 ln->ln_ntick = tick - INT_MAX; 406 callout_reset(&ln->ln_timer_ch, INT_MAX, 407 nd6_llinfo_timer, ln); 408 } else { 409 ln->ln_ntick = 0; 410 callout_reset(&ln->ln_timer_ch, tick, 411 nd6_llinfo_timer, ln); 412 } 413 } 414 } 415 416 static void 417 nd6_llinfo_timer(arg) 418 void *arg; 419 { 420 struct llinfo_nd6 *ln; 421 struct rtentry *rt; 422 struct in6_addr *dst; 423 struct ifnet *ifp; 424 struct nd_ifinfo *ndi = NULL; 425 426 ln = (struct llinfo_nd6 *)arg; 427 428 if (ln->ln_ntick > 0) { 429 if (ln->ln_ntick > INT_MAX) { 430 ln->ln_ntick -= INT_MAX; 431 nd6_llinfo_settimer(ln, INT_MAX); 432 } else { 433 ln->ln_ntick = 0; 434 nd6_llinfo_settimer(ln, ln->ln_ntick); 435 } 436 return; 437 } 438 439 if ((rt = ln->ln_rt) == NULL) 440 panic("ln->ln_rt == NULL"); 441 if ((ifp = rt->rt_ifp) == NULL) 442 panic("ln->ln_rt->rt_ifp == NULL"); 443 ndi = ND_IFINFO(ifp); 444 445 /* sanity check */ 446 if (rt->rt_llinfo && (struct llinfo_nd6 *)rt->rt_llinfo != ln) 447 panic("rt_llinfo(%p) is not equal to ln(%p)", 448 rt->rt_llinfo, ln); 449 if (rt_key(rt) == NULL) 450 panic("rt key is NULL in nd6_timer(ln=%p)", ln); 451 452 dst = &((struct sockaddr_in6 *)rt_key(rt))->sin6_addr; 453 454 switch (ln->ln_state) { 455 case ND6_LLINFO_INCOMPLETE: 456 if (ln->ln_asked < nd6_mmaxtries) { 457 ln->ln_asked++; 458 nd6_llinfo_settimer(ln, (long)ndi->retrans * hz / 1000); 459 nd6_ns_output(ifp, NULL, dst, ln, 0); 460 } else { 461 struct mbuf *m = ln->ln_hold; 462 if (m) { 463 struct mbuf *m0; 464 465 /* 466 * assuming every packet in ln_hold has the 467 * same IP header 468 */ 469 m0 = m->m_nextpkt; 470 m->m_nextpkt = NULL; 471 icmp6_error2(m, ICMP6_DST_UNREACH, 472 ICMP6_DST_UNREACH_ADDR, 0, rt->rt_ifp); 473 474 ln->ln_hold = m0; 475 clear_llinfo_pqueue(ln); 476 } 477 if (rt) 478 (void)nd6_free(rt, 0); 479 ln = NULL; 480 } 481 break; 482 case ND6_LLINFO_REACHABLE: 483 if (!ND6_LLINFO_PERMANENT(ln)) { 484 ln->ln_state = ND6_LLINFO_STALE; 485 nd6_llinfo_settimer(ln, (long)nd6_gctimer * hz); 486 } 487 break; 488 489 case ND6_LLINFO_STALE: 490 /* Garbage Collection(RFC 2461 5.3) */ 491 if (!ND6_LLINFO_PERMANENT(ln)) { 492 (void)nd6_free(rt, 1); 493 ln = NULL; 494 } 495 break; 496 497 case ND6_LLINFO_DELAY: 498 if (ndi && (ndi->flags & ND6_IFF_PERFORMNUD) != 0) { 499 /* We need NUD */ 500 ln->ln_asked = 1; 501 ln->ln_state = ND6_LLINFO_PROBE; 502 nd6_llinfo_settimer(ln, (long)ndi->retrans * hz / 1000); 503 nd6_ns_output(ifp, dst, dst, ln, 0); 504 } else { 505 ln->ln_state = ND6_LLINFO_STALE; /* XXX */ 506 nd6_llinfo_settimer(ln, (long)nd6_gctimer * hz); 507 } 508 break; 509 case ND6_LLINFO_PROBE: 510 if (ln->ln_asked < nd6_umaxtries) { 511 ln->ln_asked++; 512 nd6_llinfo_settimer(ln, (long)ndi->retrans * hz / 1000); 513 nd6_ns_output(ifp, dst, dst, ln, 0); 514 } else if (rt->rt_ifa != NULL && 515 rt->rt_ifa->ifa_addr->sa_family == AF_INET6 && 516 (((struct in6_ifaddr *)rt->rt_ifa)->ia_flags & IFA_ROUTE)) { 517 /* 518 * This is an unreachable neighbor whose address is 519 * specified as the destination of a p2p interface 520 * (see in6_ifinit()). We should not free the entry 521 * since this is sort of a "static" entry generated 522 * via interface address configuration. 523 */ 524 ln->ln_asked = 0; 525 ln->ln_expire = 0; /* make it permanent */ 526 ln->ln_state = ND6_LLINFO_STALE; 527 } else { 528 (void)nd6_free(rt, 0); 529 ln = NULL; 530 } 531 break; 532 } 533 } 534 535 536 /* 537 * ND6 timer routine to expire default route list and prefix list 538 */ 539 void 540 nd6_timer(ignored_arg) 541 void *ignored_arg; 542 { 543 int s; 544 struct nd_defrouter *dr; 545 struct nd_prefix *pr; 546 struct in6_ifaddr *ia6, *nia6; 547 struct in6_addrlifetime *lt6; 548 549 callout_reset(&nd6_timer_ch, nd6_prune * hz, 550 nd6_timer, NULL); 551 552 /* expire default router list */ 553 s = splnet(); 554 dr = TAILQ_FIRST(&nd_defrouter); 555 while (dr) { 556 if (dr->expire && dr->expire < time_second) { 557 struct nd_defrouter *t; 558 t = TAILQ_NEXT(dr, dr_entry); 559 defrtrlist_del(dr); 560 dr = t; 561 } else { 562 dr = TAILQ_NEXT(dr, dr_entry); 563 } 564 } 565 566 /* 567 * expire interface addresses. 568 * in the past the loop was inside prefix expiry processing. 569 * However, from a stricter speci-confrmance standpoint, we should 570 * rather separate address lifetimes and prefix lifetimes. 571 */ 572 addrloop: 573 for (ia6 = in6_ifaddr; ia6; ia6 = nia6) { 574 nia6 = ia6->ia_next; 575 /* check address lifetime */ 576 lt6 = &ia6->ia6_lifetime; 577 if (IFA6_IS_INVALID(ia6)) { 578 int regen = 0; 579 580 /* 581 * If the expiring address is temporary, try 582 * regenerating a new one. This would be useful when 583 * we suspended a laptop PC, then turned it on after a 584 * period that could invalidate all temporary 585 * addresses. Although we may have to restart the 586 * loop (see below), it must be after purging the 587 * address. Otherwise, we'd see an infinite loop of 588 * regeneration. 589 */ 590 if (ip6_use_tempaddr && 591 (ia6->ia6_flags & IN6_IFF_TEMPORARY) != 0) { 592 if (regen_tmpaddr(ia6) == 0) 593 regen = 1; 594 } 595 596 in6_purgeaddr(&ia6->ia_ifa); 597 598 if (regen) 599 goto addrloop; /* XXX: see below */ 600 } else if (IFA6_IS_DEPRECATED(ia6)) { 601 int oldflags = ia6->ia6_flags; 602 603 ia6->ia6_flags |= IN6_IFF_DEPRECATED; 604 605 /* 606 * If a temporary address has just become deprecated, 607 * regenerate a new one if possible. 608 */ 609 if (ip6_use_tempaddr && 610 (ia6->ia6_flags & IN6_IFF_TEMPORARY) != 0 && 611 (oldflags & IN6_IFF_DEPRECATED) == 0) { 612 613 if (regen_tmpaddr(ia6) == 0) { 614 /* 615 * A new temporary address is 616 * generated. 617 * XXX: this means the address chain 618 * has changed while we are still in 619 * the loop. Although the change 620 * would not cause disaster (because 621 * it's not a deletion, but an 622 * addition,) we'd rather restart the 623 * loop just for safety. Or does this 624 * significantly reduce performance?? 625 */ 626 goto addrloop; 627 } 628 } 629 } else { 630 /* 631 * A new RA might have made a deprecated address 632 * preferred. 633 */ 634 ia6->ia6_flags &= ~IN6_IFF_DEPRECATED; 635 } 636 } 637 638 /* expire prefix list */ 639 pr = nd_prefix.lh_first; 640 while (pr) { 641 /* 642 * check prefix lifetime. 643 * since pltime is just for autoconf, pltime processing for 644 * prefix is not necessary. 645 */ 646 if (pr->ndpr_vltime != ND6_INFINITE_LIFETIME && 647 time_second - pr->ndpr_lastupdate > pr->ndpr_vltime) { 648 struct nd_prefix *t; 649 t = pr->ndpr_next; 650 651 /* 652 * address expiration and prefix expiration are 653 * separate. NEVER perform in6_purgeaddr here. 654 */ 655 656 prelist_remove(pr); 657 pr = t; 658 } else 659 pr = pr->ndpr_next; 660 } 661 splx(s); 662 } 663 664 static int 665 regen_tmpaddr(ia6) 666 struct in6_ifaddr *ia6; /* deprecated/invalidated temporary address */ 667 { 668 struct ifaddr *ifa; 669 struct ifnet *ifp; 670 struct in6_ifaddr *public_ifa6 = NULL; 671 672 ifp = ia6->ia_ifa.ifa_ifp; 673 for (ifa = ifp->if_addrlist.tqh_first; ifa; 674 ifa = ifa->ifa_list.tqe_next) { 675 struct in6_ifaddr *it6; 676 677 if (ifa->ifa_addr->sa_family != AF_INET6) 678 continue; 679 680 it6 = (struct in6_ifaddr *)ifa; 681 682 /* ignore no autoconf addresses. */ 683 if ((it6->ia6_flags & IN6_IFF_AUTOCONF) == 0) 684 continue; 685 686 /* ignore autoconf addresses with different prefixes. */ 687 if (it6->ia6_ndpr == NULL || it6->ia6_ndpr != ia6->ia6_ndpr) 688 continue; 689 690 /* 691 * Now we are looking at an autoconf address with the same 692 * prefix as ours. If the address is temporary and is still 693 * preferred, do not create another one. It would be rare, but 694 * could happen, for example, when we resume a laptop PC after 695 * a long period. 696 */ 697 if ((it6->ia6_flags & IN6_IFF_TEMPORARY) != 0 && 698 !IFA6_IS_DEPRECATED(it6)) { 699 public_ifa6 = NULL; 700 break; 701 } 702 703 /* 704 * This is a public autoconf address that has the same prefix 705 * as ours. If it is preferred, keep it. We can't break the 706 * loop here, because there may be a still-preferred temporary 707 * address with the prefix. 708 */ 709 if (!IFA6_IS_DEPRECATED(it6)) 710 public_ifa6 = it6; 711 } 712 713 if (public_ifa6 != NULL) { 714 int e; 715 716 if ((e = in6_tmpifadd(public_ifa6, 0, 0)) != 0) { 717 log(LOG_NOTICE, "regen_tmpaddr: failed to create a new" 718 " tmp addr,errno=%d\n", e); 719 return (-1); 720 } 721 return (0); 722 } 723 724 return (-1); 725 } 726 727 /* 728 * Nuke neighbor cache/prefix/default router management table, right before 729 * ifp goes away. 730 */ 731 void 732 nd6_purge(ifp) 733 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 for (dr = TAILQ_FIRST(&nd_defrouter); dr; dr = ndr) { 746 ndr = TAILQ_NEXT(dr, dr_entry); 747 if (dr->installed) 748 continue; 749 750 if (dr->ifp == ifp) 751 defrtrlist_del(dr); 752 } 753 754 for (dr = TAILQ_FIRST(&nd_defrouter); dr; dr = ndr) { 755 ndr = TAILQ_NEXT(dr, dr_entry); 756 if (!dr->installed) 757 continue; 758 759 if (dr->ifp == ifp) 760 defrtrlist_del(dr); 761 } 762 763 /* Nuke prefix list entries toward ifp */ 764 for (pr = nd_prefix.lh_first; pr; pr = npr) { 765 npr = pr->ndpr_next; 766 if (pr->ndpr_ifp == ifp) { 767 /* 768 * Because if_detach() does *not* release prefixes 769 * while purging addresses the reference count will 770 * still be above zero. We therefore reset it to 771 * make sure that the prefix really gets purged. 772 */ 773 pr->ndpr_refcnt = 0; 774 775 /* 776 * Previously, pr->ndpr_addr is removed as well, 777 * but I strongly believe we don't have to do it. 778 * nd6_purge() is only called from in6_ifdetach(), 779 * which removes all the associated interface addresses 780 * by itself. 781 * (jinmei@kame.net 20010129) 782 */ 783 prelist_remove(pr); 784 } 785 } 786 787 /* cancel default outgoing interface setting */ 788 if (nd6_defifindex == ifp->if_index) 789 nd6_setdefaultiface(0); 790 791 if (!ip6_forwarding && ip6_accept_rtadv) { /* XXX: too restrictive? */ 792 /* refresh default router list */ 793 defrouter_select(); 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 && ln != &llinfo_nd6) { 804 struct rtentry *rt; 805 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 = (struct sockaddr_dl *)rt->rt_gateway; 812 if (sdl->sdl_index == ifp->if_index) 813 nln = nd6_free(rt, 0); 814 } 815 ln = nln; 816 } 817 } 818 819 struct rtentry * 820 nd6_lookup(addr6, create, ifp) 821 struct in6_addr *addr6; 822 int create; 823 struct ifnet *ifp; 824 { 825 struct rtentry *rt; 826 struct sockaddr_in6 sin6; 827 char ip6buf[INET6_ADDRSTRLEN]; 828 829 bzero(&sin6, sizeof(sin6)); 830 sin6.sin6_len = sizeof(struct sockaddr_in6); 831 sin6.sin6_family = AF_INET6; 832 sin6.sin6_addr = *addr6; 833 rt = rtalloc1((struct sockaddr *)&sin6, create, 0UL); 834 if (rt) { 835 if ((rt->rt_flags & RTF_LLINFO) == 0 && create) { 836 /* 837 * This is the case for the default route. 838 * If we want to create a neighbor cache for the 839 * address, we should free the route for the 840 * destination and allocate an interface route. 841 */ 842 RTFREE_LOCKED(rt); 843 rt = NULL; 844 } 845 } 846 if (rt == NULL) { 847 if (create && ifp) { 848 int e; 849 850 /* 851 * If no route is available and create is set, 852 * we allocate a host route for the destination 853 * and treat it like an interface route. 854 * This hack is necessary for a neighbor which can't 855 * be covered by our own prefix. 856 */ 857 struct ifaddr *ifa = 858 ifaof_ifpforaddr((struct sockaddr *)&sin6, ifp); 859 if (ifa == NULL) 860 return (NULL); 861 862 /* 863 * Create a new route. RTF_LLINFO is necessary 864 * to create a Neighbor Cache entry for the 865 * destination in nd6_rtrequest which will be 866 * called in rtrequest via ifa->ifa_rtrequest. 867 */ 868 if ((e = rtrequest(RTM_ADD, (struct sockaddr *)&sin6, 869 ifa->ifa_addr, (struct sockaddr *)&all1_sa, 870 (ifa->ifa_flags | RTF_HOST | RTF_LLINFO) & 871 ~RTF_CLONING, &rt)) != 0) { 872 log(LOG_ERR, 873 "nd6_lookup: failed to add route for a " 874 "neighbor(%s), errno=%d\n", 875 ip6_sprintf(ip6buf, addr6), e); 876 } 877 if (rt == NULL) 878 return (NULL); 879 RT_LOCK(rt); 880 if (rt->rt_llinfo) { 881 struct llinfo_nd6 *ln = 882 (struct llinfo_nd6 *)rt->rt_llinfo; 883 ln->ln_state = ND6_LLINFO_NOSTATE; 884 } 885 } else 886 return (NULL); 887 } 888 RT_LOCK_ASSERT(rt); 889 RT_REMREF(rt); 890 /* 891 * Validation for the entry. 892 * Note that the check for rt_llinfo is necessary because a cloned 893 * route from a parent route that has the L flag (e.g. the default 894 * route to a p2p interface) may have the flag, too, while the 895 * destination is not actually a neighbor. 896 * XXX: we can't use rt->rt_ifp to check for the interface, since 897 * it might be the loopback interface if the entry is for our 898 * own address on a non-loopback interface. Instead, we should 899 * use rt->rt_ifa->ifa_ifp, which would specify the REAL 900 * interface. 901 * Note also that ifa_ifp and ifp may differ when we connect two 902 * interfaces to a same link, install a link prefix to an interface, 903 * and try to install a neighbor cache on an interface that does not 904 * have a route to the prefix. 905 */ 906 if ((rt->rt_flags & RTF_GATEWAY) || (rt->rt_flags & RTF_LLINFO) == 0 || 907 rt->rt_gateway->sa_family != AF_LINK || rt->rt_llinfo == NULL || 908 (ifp && rt->rt_ifa->ifa_ifp != ifp)) { 909 if (create) { 910 nd6log((LOG_DEBUG, 911 "nd6_lookup: failed to lookup %s (if = %s)\n", 912 ip6_sprintf(ip6buf, addr6), 913 ifp ? if_name(ifp) : "unspec")); 914 } 915 RT_UNLOCK(rt); 916 return (NULL); 917 } 918 RT_UNLOCK(rt); /* XXX not ready to return rt locked */ 919 return (rt); 920 } 921 922 /* 923 * Test whether a given IPv6 address is a neighbor or not, ignoring 924 * the actual neighbor cache. The neighbor cache is ignored in order 925 * to not reenter the routing code from within itself. 926 */ 927 static int 928 nd6_is_new_addr_neighbor(addr, ifp) 929 struct sockaddr_in6 *addr; 930 struct ifnet *ifp; 931 { 932 struct nd_prefix *pr; 933 struct ifaddr *dstaddr; 934 935 /* 936 * A link-local address is always a neighbor. 937 * XXX: a link does not necessarily specify a single interface. 938 */ 939 if (IN6_IS_ADDR_LINKLOCAL(&addr->sin6_addr)) { 940 struct sockaddr_in6 sin6_copy; 941 u_int32_t zone; 942 943 /* 944 * We need sin6_copy since sa6_recoverscope() may modify the 945 * content (XXX). 946 */ 947 sin6_copy = *addr; 948 if (sa6_recoverscope(&sin6_copy)) 949 return (0); /* XXX: should be impossible */ 950 if (in6_setscope(&sin6_copy.sin6_addr, ifp, &zone)) 951 return (0); 952 if (sin6_copy.sin6_scope_id == zone) 953 return (1); 954 else 955 return (0); 956 } 957 958 /* 959 * If the address matches one of our addresses, 960 * it should be a neighbor. 961 * If the address matches one of our on-link prefixes, it should be a 962 * neighbor. 963 */ 964 for (pr = nd_prefix.lh_first; pr; pr = pr->ndpr_next) { 965 if (pr->ndpr_ifp != ifp) 966 continue; 967 968 if (!(pr->ndpr_stateflags & NDPRF_ONLINK)) 969 continue; 970 971 if (IN6_ARE_MASKED_ADDR_EQUAL(&pr->ndpr_prefix.sin6_addr, 972 &addr->sin6_addr, &pr->ndpr_mask)) 973 return (1); 974 } 975 976 /* 977 * If the address is assigned on the node of the other side of 978 * a p2p interface, the address should be a neighbor. 979 */ 980 dstaddr = ifa_ifwithdstaddr((struct sockaddr *)addr); 981 if ((dstaddr != NULL) && (dstaddr->ifa_ifp == ifp)) 982 return (1); 983 984 /* 985 * If the default router list is empty, all addresses are regarded 986 * as on-link, and thus, as a neighbor. 987 * XXX: we restrict the condition to hosts, because routers usually do 988 * not have the "default router list". 989 */ 990 if (!ip6_forwarding && TAILQ_FIRST(&nd_defrouter) == NULL && 991 nd6_defifindex == ifp->if_index) { 992 return (1); 993 } 994 995 return (0); 996 } 997 998 999 /* 1000 * Detect if a given IPv6 address identifies a neighbor on a given link. 1001 * XXX: should take care of the destination of a p2p link? 1002 */ 1003 int 1004 nd6_is_addr_neighbor(addr, ifp) 1005 struct sockaddr_in6 *addr; 1006 struct ifnet *ifp; 1007 { 1008 1009 if (nd6_is_new_addr_neighbor(addr, ifp)) 1010 return (1); 1011 1012 /* 1013 * Even if the address matches none of our addresses, it might be 1014 * in the neighbor cache. 1015 */ 1016 if (nd6_lookup(&addr->sin6_addr, 0, ifp) != NULL) 1017 return (1); 1018 1019 return (0); 1020 } 1021 1022 /* 1023 * Free an nd6 llinfo entry. 1024 * Since the function would cause significant changes in the kernel, DO NOT 1025 * make it global, unless you have a strong reason for the change, and are sure 1026 * that the change is safe. 1027 */ 1028 static struct llinfo_nd6 * 1029 nd6_free(rt, gc) 1030 struct rtentry *rt; 1031 int gc; 1032 { 1033 struct llinfo_nd6 *ln = (struct llinfo_nd6 *)rt->rt_llinfo, *next; 1034 struct in6_addr in6 = ((struct sockaddr_in6 *)rt_key(rt))->sin6_addr; 1035 struct nd_defrouter *dr; 1036 1037 /* 1038 * we used to have pfctlinput(PRC_HOSTDEAD) here. 1039 * even though it is not harmful, it was not really necessary. 1040 */ 1041 1042 /* cancel timer */ 1043 nd6_llinfo_settimer(ln, -1); 1044 1045 if (!ip6_forwarding) { 1046 int s; 1047 s = splnet(); 1048 dr = defrouter_lookup(&((struct sockaddr_in6 *)rt_key(rt))->sin6_addr, 1049 rt->rt_ifp); 1050 1051 if (dr != NULL && dr->expire && 1052 ln->ln_state == ND6_LLINFO_STALE && gc) { 1053 /* 1054 * If the reason for the deletion is just garbage 1055 * collection, and the neighbor is an active default 1056 * router, do not delete it. Instead, reset the GC 1057 * timer using the router's lifetime. 1058 * Simply deleting the entry would affect default 1059 * router selection, which is not necessarily a good 1060 * thing, especially when we're using router preference 1061 * values. 1062 * XXX: the check for ln_state would be redundant, 1063 * but we intentionally keep it just in case. 1064 */ 1065 if (dr->expire > time_second) 1066 nd6_llinfo_settimer(ln, 1067 (dr->expire - time_second) * hz); 1068 else 1069 nd6_llinfo_settimer(ln, (long)nd6_gctimer * hz); 1070 splx(s); 1071 return (ln->ln_next); 1072 } 1073 1074 if (ln->ln_router || dr) { 1075 /* 1076 * rt6_flush must be called whether or not the neighbor 1077 * is in the Default Router List. 1078 * See a corresponding comment in nd6_na_input(). 1079 */ 1080 rt6_flush(&in6, rt->rt_ifp); 1081 } 1082 1083 if (dr) { 1084 /* 1085 * Unreachablity of a router might affect the default 1086 * router selection and on-link detection of advertised 1087 * prefixes. 1088 */ 1089 1090 /* 1091 * Temporarily fake the state to choose a new default 1092 * router and to perform on-link determination of 1093 * prefixes correctly. 1094 * Below the state will be set correctly, 1095 * or the entry itself will be deleted. 1096 */ 1097 ln->ln_state = ND6_LLINFO_INCOMPLETE; 1098 1099 /* 1100 * Since defrouter_select() does not affect the 1101 * on-link determination and MIP6 needs the check 1102 * before the default router selection, we perform 1103 * the check now. 1104 */ 1105 pfxlist_onlink_check(); 1106 1107 /* 1108 * refresh default router list 1109 */ 1110 defrouter_select(); 1111 } 1112 splx(s); 1113 } 1114 1115 /* 1116 * Before deleting the entry, remember the next entry as the 1117 * return value. We need this because pfxlist_onlink_check() above 1118 * might have freed other entries (particularly the old next entry) as 1119 * a side effect (XXX). 1120 */ 1121 next = ln->ln_next; 1122 1123 /* 1124 * Detach the route from the routing tree and the list of neighbor 1125 * caches, and disable the route entry not to be used in already 1126 * cached routes. 1127 */ 1128 rtrequest(RTM_DELETE, rt_key(rt), (struct sockaddr *)0, 1129 rt_mask(rt), 0, (struct rtentry **)0); 1130 1131 return (next); 1132 } 1133 1134 /* 1135 * Upper-layer reachability hint for Neighbor Unreachability Detection. 1136 * 1137 * XXX cost-effective methods? 1138 */ 1139 void 1140 nd6_nud_hint(rt, dst6, force) 1141 struct rtentry *rt; 1142 struct in6_addr *dst6; 1143 int force; 1144 { 1145 struct llinfo_nd6 *ln; 1146 1147 /* 1148 * If the caller specified "rt", use that. Otherwise, resolve the 1149 * routing table by supplied "dst6". 1150 */ 1151 if (rt == NULL) { 1152 if (dst6 == NULL) 1153 return; 1154 if ((rt = nd6_lookup(dst6, 0, NULL)) == NULL) 1155 return; 1156 } 1157 1158 if ((rt->rt_flags & RTF_GATEWAY) != 0 || 1159 (rt->rt_flags & RTF_LLINFO) == 0 || 1160 rt->rt_llinfo == NULL || rt->rt_gateway == NULL || 1161 rt->rt_gateway->sa_family != AF_LINK) { 1162 /* This is not a host route. */ 1163 return; 1164 } 1165 1166 ln = (struct llinfo_nd6 *)rt->rt_llinfo; 1167 if (ln->ln_state < ND6_LLINFO_REACHABLE) 1168 return; 1169 1170 /* 1171 * if we get upper-layer reachability confirmation many times, 1172 * it is possible we have false information. 1173 */ 1174 if (!force) { 1175 ln->ln_byhint++; 1176 if (ln->ln_byhint > nd6_maxnudhint) 1177 return; 1178 } 1179 1180 ln->ln_state = ND6_LLINFO_REACHABLE; 1181 if (!ND6_LLINFO_PERMANENT(ln)) { 1182 nd6_llinfo_settimer(ln, 1183 (long)ND_IFINFO(rt->rt_ifp)->reachable * hz); 1184 } 1185 } 1186 1187 void 1188 nd6_rtrequest(req, rt, info) 1189 int req; 1190 struct rtentry *rt; 1191 struct rt_addrinfo *info; /* xxx unused */ 1192 { 1193 struct sockaddr *gate = rt->rt_gateway; 1194 struct llinfo_nd6 *ln = (struct llinfo_nd6 *)rt->rt_llinfo; 1195 static struct sockaddr_dl null_sdl = {sizeof(null_sdl), AF_LINK}; 1196 struct ifnet *ifp = rt->rt_ifp; 1197 struct ifaddr *ifa; 1198 1199 RT_LOCK_ASSERT(rt); 1200 1201 if ((rt->rt_flags & RTF_GATEWAY) != 0) 1202 return; 1203 1204 if (nd6_need_cache(ifp) == 0 && (rt->rt_flags & RTF_HOST) == 0) { 1205 /* 1206 * This is probably an interface direct route for a link 1207 * which does not need neighbor caches (e.g. fe80::%lo0/64). 1208 * We do not need special treatment below for such a route. 1209 * Moreover, the RTF_LLINFO flag which would be set below 1210 * would annoy the ndp(8) command. 1211 */ 1212 return; 1213 } 1214 1215 if (req == RTM_RESOLVE && 1216 (nd6_need_cache(ifp) == 0 || /* stf case */ 1217 !nd6_is_new_addr_neighbor((struct sockaddr_in6 *)rt_key(rt), 1218 ifp))) { 1219 /* 1220 * FreeBSD and BSD/OS often make a cloned host route based 1221 * on a less-specific route (e.g. the default route). 1222 * If the less specific route does not have a "gateway" 1223 * (this is the case when the route just goes to a p2p or an 1224 * stf interface), we'll mistakenly make a neighbor cache for 1225 * the host route, and will see strange neighbor solicitation 1226 * for the corresponding destination. In order to avoid the 1227 * confusion, we check if the destination of the route is 1228 * a neighbor in terms of neighbor discovery, and stop the 1229 * process if not. Additionally, we remove the LLINFO flag 1230 * so that ndp(8) will not try to get the neighbor information 1231 * of the destination. 1232 */ 1233 rt->rt_flags &= ~RTF_LLINFO; 1234 return; 1235 } 1236 1237 switch (req) { 1238 case RTM_ADD: 1239 /* 1240 * There is no backward compatibility :) 1241 * 1242 * if ((rt->rt_flags & RTF_HOST) == 0 && 1243 * SIN(rt_mask(rt))->sin_addr.s_addr != 0xffffffff) 1244 * rt->rt_flags |= RTF_CLONING; 1245 */ 1246 if ((rt->rt_flags & RTF_CLONING) || 1247 ((rt->rt_flags & RTF_LLINFO) && ln == NULL)) { 1248 /* 1249 * Case 1: This route should come from a route to 1250 * interface (RTF_CLONING case) or the route should be 1251 * treated as on-link but is currently not 1252 * (RTF_LLINFO && ln == NULL case). 1253 */ 1254 rt_setgate(rt, rt_key(rt), 1255 (struct sockaddr *)&null_sdl); 1256 gate = rt->rt_gateway; 1257 SDL(gate)->sdl_type = ifp->if_type; 1258 SDL(gate)->sdl_index = ifp->if_index; 1259 if (ln) 1260 nd6_llinfo_settimer(ln, 0); 1261 if ((rt->rt_flags & RTF_CLONING) != 0) 1262 break; 1263 } 1264 /* 1265 * In IPv4 code, we try to annonuce new RTF_ANNOUNCE entry here. 1266 * We don't do that here since llinfo is not ready yet. 1267 * 1268 * There are also couple of other things to be discussed: 1269 * - unsolicited NA code needs improvement beforehand 1270 * - RFC2461 says we MAY send multicast unsolicited NA 1271 * (7.2.6 paragraph 4), however, it also says that we 1272 * SHOULD provide a mechanism to prevent multicast NA storm. 1273 * we don't have anything like it right now. 1274 * note that the mechanism needs a mutual agreement 1275 * between proxies, which means that we need to implement 1276 * a new protocol, or a new kludge. 1277 * - from RFC2461 6.2.4, host MUST NOT send an unsolicited NA. 1278 * we need to check ip6forwarding before sending it. 1279 * (or should we allow proxy ND configuration only for 1280 * routers? there's no mention about proxy ND from hosts) 1281 */ 1282 /* FALLTHROUGH */ 1283 case RTM_RESOLVE: 1284 if ((ifp->if_flags & (IFF_POINTOPOINT | IFF_LOOPBACK)) == 0) { 1285 /* 1286 * Address resolution isn't necessary for a point to 1287 * point link, so we can skip this test for a p2p link. 1288 */ 1289 if (gate->sa_family != AF_LINK || 1290 gate->sa_len < sizeof(null_sdl)) { 1291 log(LOG_DEBUG, 1292 "nd6_rtrequest: bad gateway value: %s\n", 1293 if_name(ifp)); 1294 break; 1295 } 1296 SDL(gate)->sdl_type = ifp->if_type; 1297 SDL(gate)->sdl_index = ifp->if_index; 1298 } 1299 if (ln != NULL) 1300 break; /* This happens on a route change */ 1301 /* 1302 * Case 2: This route may come from cloning, or a manual route 1303 * add with a LL address. 1304 */ 1305 R_Malloc(ln, struct llinfo_nd6 *, sizeof(*ln)); 1306 rt->rt_llinfo = (caddr_t)ln; 1307 if (ln == NULL) { 1308 log(LOG_DEBUG, "nd6_rtrequest: malloc failed\n"); 1309 break; 1310 } 1311 nd6_inuse++; 1312 nd6_allocated++; 1313 bzero(ln, sizeof(*ln)); 1314 RT_ADDREF(rt); 1315 ln->ln_rt = rt; 1316 callout_init(&ln->ln_timer_ch, 0); 1317 1318 /* this is required for "ndp" command. - shin */ 1319 if (req == RTM_ADD) { 1320 /* 1321 * gate should have some valid AF_LINK entry, 1322 * and ln->ln_expire should have some lifetime 1323 * which is specified by ndp command. 1324 */ 1325 ln->ln_state = ND6_LLINFO_REACHABLE; 1326 ln->ln_byhint = 0; 1327 } else { 1328 /* 1329 * When req == RTM_RESOLVE, rt is created and 1330 * initialized in rtrequest(), so rt_expire is 0. 1331 */ 1332 ln->ln_state = ND6_LLINFO_NOSTATE; 1333 nd6_llinfo_settimer(ln, 0); 1334 } 1335 rt->rt_flags |= RTF_LLINFO; 1336 ln->ln_next = llinfo_nd6.ln_next; 1337 llinfo_nd6.ln_next = ln; 1338 ln->ln_prev = &llinfo_nd6; 1339 ln->ln_next->ln_prev = ln; 1340 1341 /* 1342 * check if rt_key(rt) is one of my address assigned 1343 * to the interface. 1344 */ 1345 ifa = (struct ifaddr *)in6ifa_ifpwithaddr(rt->rt_ifp, 1346 &SIN6(rt_key(rt))->sin6_addr); 1347 if (ifa) { 1348 caddr_t macp = nd6_ifptomac(ifp); 1349 nd6_llinfo_settimer(ln, -1); 1350 ln->ln_state = ND6_LLINFO_REACHABLE; 1351 ln->ln_byhint = 0; 1352 if (macp) { 1353 bcopy(macp, LLADDR(SDL(gate)), ifp->if_addrlen); 1354 SDL(gate)->sdl_alen = ifp->if_addrlen; 1355 } 1356 if (nd6_useloopback) { 1357 rt->rt_ifp = &loif[0]; /* XXX */ 1358 /* 1359 * Make sure rt_ifa be equal to the ifaddr 1360 * corresponding to the address. 1361 * We need this because when we refer 1362 * rt_ifa->ia6_flags in ip6_input, we assume 1363 * that the rt_ifa points to the address instead 1364 * of the loopback address. 1365 */ 1366 if (ifa != rt->rt_ifa) { 1367 IFAFREE(rt->rt_ifa); 1368 IFAREF(ifa); 1369 rt->rt_ifa = ifa; 1370 } 1371 } 1372 } else if (rt->rt_flags & RTF_ANNOUNCE) { 1373 nd6_llinfo_settimer(ln, -1); 1374 ln->ln_state = ND6_LLINFO_REACHABLE; 1375 ln->ln_byhint = 0; 1376 1377 /* join solicited node multicast for proxy ND */ 1378 if (ifp->if_flags & IFF_MULTICAST) { 1379 struct in6_addr llsol; 1380 int error; 1381 1382 llsol = SIN6(rt_key(rt))->sin6_addr; 1383 llsol.s6_addr32[0] = IPV6_ADDR_INT32_MLL; 1384 llsol.s6_addr32[1] = 0; 1385 llsol.s6_addr32[2] = htonl(1); 1386 llsol.s6_addr8[12] = 0xff; 1387 if (in6_setscope(&llsol, ifp, NULL)) 1388 break; 1389 if (in6_addmulti(&llsol, ifp, 1390 &error, 0) == NULL) { 1391 char ip6buf[INET6_ADDRSTRLEN]; 1392 nd6log((LOG_ERR, "%s: failed to join " 1393 "%s (errno=%d)\n", if_name(ifp), 1394 ip6_sprintf(ip6buf, &llsol), 1395 error)); 1396 } 1397 } 1398 } 1399 break; 1400 1401 case RTM_DELETE: 1402 if (ln == NULL) 1403 break; 1404 /* leave from solicited node multicast for proxy ND */ 1405 if ((rt->rt_flags & RTF_ANNOUNCE) != 0 && 1406 (ifp->if_flags & IFF_MULTICAST) != 0) { 1407 struct in6_addr llsol; 1408 struct in6_multi *in6m; 1409 1410 llsol = SIN6(rt_key(rt))->sin6_addr; 1411 llsol.s6_addr32[0] = IPV6_ADDR_INT32_MLL; 1412 llsol.s6_addr32[1] = 0; 1413 llsol.s6_addr32[2] = htonl(1); 1414 llsol.s6_addr8[12] = 0xff; 1415 if (in6_setscope(&llsol, ifp, NULL) == 0) { 1416 IN6_LOOKUP_MULTI(llsol, ifp, in6m); 1417 if (in6m) 1418 in6_delmulti(in6m); 1419 } else 1420 ; /* XXX: should not happen. bark here? */ 1421 } 1422 nd6_inuse--; 1423 ln->ln_next->ln_prev = ln->ln_prev; 1424 ln->ln_prev->ln_next = ln->ln_next; 1425 ln->ln_prev = NULL; 1426 nd6_llinfo_settimer(ln, -1); 1427 RT_REMREF(rt); 1428 rt->rt_llinfo = 0; 1429 rt->rt_flags &= ~RTF_LLINFO; 1430 clear_llinfo_pqueue(ln); 1431 Free((caddr_t)ln); 1432 } 1433 } 1434 1435 int 1436 nd6_ioctl(cmd, data, ifp) 1437 u_long cmd; 1438 caddr_t data; 1439 struct ifnet *ifp; 1440 { 1441 struct in6_drlist *drl = (struct in6_drlist *)data; 1442 struct in6_oprlist *oprl = (struct in6_oprlist *)data; 1443 struct in6_ndireq *ndi = (struct in6_ndireq *)data; 1444 struct in6_nbrinfo *nbi = (struct in6_nbrinfo *)data; 1445 struct in6_ndifreq *ndif = (struct in6_ndifreq *)data; 1446 struct nd_defrouter *dr; 1447 struct nd_prefix *pr; 1448 struct rtentry *rt; 1449 int i = 0, error = 0; 1450 int s; 1451 1452 switch (cmd) { 1453 case SIOCGDRLST_IN6: 1454 /* 1455 * obsolete API, use sysctl under net.inet6.icmp6 1456 */ 1457 bzero(drl, sizeof(*drl)); 1458 s = splnet(); 1459 dr = TAILQ_FIRST(&nd_defrouter); 1460 while (dr && i < DRLSTSIZ) { 1461 drl->defrouter[i].rtaddr = dr->rtaddr; 1462 in6_clearscope(&drl->defrouter[i].rtaddr); 1463 1464 drl->defrouter[i].flags = dr->flags; 1465 drl->defrouter[i].rtlifetime = dr->rtlifetime; 1466 drl->defrouter[i].expire = dr->expire; 1467 drl->defrouter[i].if_index = dr->ifp->if_index; 1468 i++; 1469 dr = TAILQ_NEXT(dr, dr_entry); 1470 } 1471 splx(s); 1472 break; 1473 case SIOCGPRLST_IN6: 1474 /* 1475 * obsolete API, use sysctl under net.inet6.icmp6 1476 * 1477 * XXX the structure in6_prlist was changed in backward- 1478 * incompatible manner. in6_oprlist is used for SIOCGPRLST_IN6, 1479 * in6_prlist is used for nd6_sysctl() - fill_prlist(). 1480 */ 1481 /* 1482 * XXX meaning of fields, especialy "raflags", is very 1483 * differnet between RA prefix list and RR/static prefix list. 1484 * how about separating ioctls into two? 1485 */ 1486 bzero(oprl, sizeof(*oprl)); 1487 s = splnet(); 1488 pr = nd_prefix.lh_first; 1489 while (pr && i < PRLSTSIZ) { 1490 struct nd_pfxrouter *pfr; 1491 int j; 1492 1493 oprl->prefix[i].prefix = pr->ndpr_prefix.sin6_addr; 1494 oprl->prefix[i].raflags = pr->ndpr_raf; 1495 oprl->prefix[i].prefixlen = pr->ndpr_plen; 1496 oprl->prefix[i].vltime = pr->ndpr_vltime; 1497 oprl->prefix[i].pltime = pr->ndpr_pltime; 1498 oprl->prefix[i].if_index = pr->ndpr_ifp->if_index; 1499 if (pr->ndpr_vltime == ND6_INFINITE_LIFETIME) 1500 oprl->prefix[i].expire = 0; 1501 else { 1502 time_t maxexpire; 1503 1504 /* XXX: we assume time_t is signed. */ 1505 maxexpire = (-1) & 1506 ~((time_t)1 << 1507 ((sizeof(maxexpire) * 8) - 1)); 1508 if (pr->ndpr_vltime < 1509 maxexpire - pr->ndpr_lastupdate) { 1510 oprl->prefix[i].expire = 1511 pr->ndpr_lastupdate + 1512 pr->ndpr_vltime; 1513 } else 1514 oprl->prefix[i].expire = maxexpire; 1515 } 1516 1517 pfr = pr->ndpr_advrtrs.lh_first; 1518 j = 0; 1519 while (pfr) { 1520 if (j < DRLSTSIZ) { 1521 #define RTRADDR oprl->prefix[i].advrtr[j] 1522 RTRADDR = pfr->router->rtaddr; 1523 in6_clearscope(&RTRADDR); 1524 #undef RTRADDR 1525 } 1526 j++; 1527 pfr = pfr->pfr_next; 1528 } 1529 oprl->prefix[i].advrtrs = j; 1530 oprl->prefix[i].origin = PR_ORIG_RA; 1531 1532 i++; 1533 pr = pr->ndpr_next; 1534 } 1535 splx(s); 1536 1537 break; 1538 case OSIOCGIFINFO_IN6: 1539 #define ND ndi->ndi 1540 /* XXX: old ndp(8) assumes a positive value for linkmtu. */ 1541 bzero(&ND, sizeof(ND)); 1542 ND.linkmtu = IN6_LINKMTU(ifp); 1543 ND.maxmtu = ND_IFINFO(ifp)->maxmtu; 1544 ND.basereachable = ND_IFINFO(ifp)->basereachable; 1545 ND.reachable = ND_IFINFO(ifp)->reachable; 1546 ND.retrans = ND_IFINFO(ifp)->retrans; 1547 ND.flags = ND_IFINFO(ifp)->flags; 1548 ND.recalctm = ND_IFINFO(ifp)->recalctm; 1549 ND.chlim = ND_IFINFO(ifp)->chlim; 1550 break; 1551 case SIOCGIFINFO_IN6: 1552 ND = *ND_IFINFO(ifp); 1553 break; 1554 case SIOCSIFINFO_IN6: 1555 /* 1556 * used to change host variables from userland. 1557 * intented for a use on router to reflect RA configurations. 1558 */ 1559 /* 0 means 'unspecified' */ 1560 if (ND.linkmtu != 0) { 1561 if (ND.linkmtu < IPV6_MMTU || 1562 ND.linkmtu > IN6_LINKMTU(ifp)) { 1563 error = EINVAL; 1564 break; 1565 } 1566 ND_IFINFO(ifp)->linkmtu = ND.linkmtu; 1567 } 1568 1569 if (ND.basereachable != 0) { 1570 int obasereachable = ND_IFINFO(ifp)->basereachable; 1571 1572 ND_IFINFO(ifp)->basereachable = ND.basereachable; 1573 if (ND.basereachable != obasereachable) 1574 ND_IFINFO(ifp)->reachable = 1575 ND_COMPUTE_RTIME(ND.basereachable); 1576 } 1577 if (ND.retrans != 0) 1578 ND_IFINFO(ifp)->retrans = ND.retrans; 1579 if (ND.chlim != 0) 1580 ND_IFINFO(ifp)->chlim = ND.chlim; 1581 /* FALLTHROUGH */ 1582 case SIOCSIFINFO_FLAGS: 1583 ND_IFINFO(ifp)->flags = ND.flags; 1584 break; 1585 #undef ND 1586 case SIOCSNDFLUSH_IN6: /* XXX: the ioctl name is confusing... */ 1587 /* sync kernel routing table with the default router list */ 1588 defrouter_reset(); 1589 defrouter_select(); 1590 break; 1591 case SIOCSPFXFLUSH_IN6: 1592 { 1593 /* flush all the prefix advertised by routers */ 1594 struct nd_prefix *pr, *next; 1595 1596 s = splnet(); 1597 for (pr = nd_prefix.lh_first; pr; pr = next) { 1598 struct in6_ifaddr *ia, *ia_next; 1599 1600 next = pr->ndpr_next; 1601 1602 if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr)) 1603 continue; /* XXX */ 1604 1605 /* do we really have to remove addresses as well? */ 1606 for (ia = in6_ifaddr; ia; ia = ia_next) { 1607 /* ia might be removed. keep the next ptr. */ 1608 ia_next = ia->ia_next; 1609 1610 if ((ia->ia6_flags & IN6_IFF_AUTOCONF) == 0) 1611 continue; 1612 1613 if (ia->ia6_ndpr == pr) 1614 in6_purgeaddr(&ia->ia_ifa); 1615 } 1616 prelist_remove(pr); 1617 } 1618 splx(s); 1619 break; 1620 } 1621 case SIOCSRTRFLUSH_IN6: 1622 { 1623 /* flush all the default routers */ 1624 struct nd_defrouter *dr, *next; 1625 1626 s = splnet(); 1627 defrouter_reset(); 1628 for (dr = TAILQ_FIRST(&nd_defrouter); dr; dr = next) { 1629 next = TAILQ_NEXT(dr, dr_entry); 1630 defrtrlist_del(dr); 1631 } 1632 defrouter_select(); 1633 splx(s); 1634 break; 1635 } 1636 case SIOCGNBRINFO_IN6: 1637 { 1638 struct llinfo_nd6 *ln; 1639 struct in6_addr nb_addr = nbi->addr; /* make local for safety */ 1640 1641 if ((error = in6_setscope(&nb_addr, ifp, NULL)) != 0) 1642 return (error); 1643 1644 s = splnet(); 1645 if ((rt = nd6_lookup(&nb_addr, 0, ifp)) == NULL) { 1646 error = EINVAL; 1647 splx(s); 1648 break; 1649 } 1650 ln = (struct llinfo_nd6 *)rt->rt_llinfo; 1651 nbi->state = ln->ln_state; 1652 nbi->asked = ln->ln_asked; 1653 nbi->isrouter = ln->ln_router; 1654 nbi->expire = ln->ln_expire; 1655 splx(s); 1656 1657 break; 1658 } 1659 case SIOCGDEFIFACE_IN6: /* XXX: should be implemented as a sysctl? */ 1660 ndif->ifindex = nd6_defifindex; 1661 break; 1662 case SIOCSDEFIFACE_IN6: /* XXX: should be implemented as a sysctl? */ 1663 return (nd6_setdefaultiface(ndif->ifindex)); 1664 } 1665 return (error); 1666 } 1667 1668 /* 1669 * Create neighbor cache entry and cache link-layer address, 1670 * on reception of inbound ND6 packets. (RS/RA/NS/redirect) 1671 */ 1672 struct rtentry * 1673 nd6_cache_lladdr(ifp, from, lladdr, lladdrlen, type, code) 1674 struct ifnet *ifp; 1675 struct in6_addr *from; 1676 char *lladdr; 1677 int lladdrlen; 1678 int type; /* ICMP6 type */ 1679 int code; /* type dependent information */ 1680 { 1681 struct rtentry *rt = NULL; 1682 struct llinfo_nd6 *ln = NULL; 1683 int is_newentry; 1684 struct sockaddr_dl *sdl = NULL; 1685 int do_update; 1686 int olladdr; 1687 int llchange; 1688 int newstate = 0; 1689 1690 if (ifp == NULL) 1691 panic("ifp == NULL in nd6_cache_lladdr"); 1692 if (from == NULL) 1693 panic("from == NULL in nd6_cache_lladdr"); 1694 1695 /* nothing must be updated for unspecified address */ 1696 if (IN6_IS_ADDR_UNSPECIFIED(from)) 1697 return NULL; 1698 1699 /* 1700 * Validation about ifp->if_addrlen and lladdrlen must be done in 1701 * the caller. 1702 * 1703 * XXX If the link does not have link-layer adderss, what should 1704 * we do? (ifp->if_addrlen == 0) 1705 * Spec says nothing in sections for RA, RS and NA. There's small 1706 * description on it in NS section (RFC 2461 7.2.3). 1707 */ 1708 1709 rt = nd6_lookup(from, 0, ifp); 1710 if (rt == NULL) { 1711 rt = nd6_lookup(from, 1, ifp); 1712 is_newentry = 1; 1713 } else { 1714 /* do nothing if static ndp is set */ 1715 if (rt->rt_flags & RTF_STATIC) 1716 return NULL; 1717 is_newentry = 0; 1718 } 1719 1720 if (rt == NULL) 1721 return NULL; 1722 if ((rt->rt_flags & (RTF_GATEWAY | RTF_LLINFO)) != RTF_LLINFO) { 1723 fail: 1724 (void)nd6_free(rt, 0); 1725 return NULL; 1726 } 1727 ln = (struct llinfo_nd6 *)rt->rt_llinfo; 1728 if (ln == NULL) 1729 goto fail; 1730 if (rt->rt_gateway == NULL) 1731 goto fail; 1732 if (rt->rt_gateway->sa_family != AF_LINK) 1733 goto fail; 1734 sdl = SDL(rt->rt_gateway); 1735 1736 olladdr = (sdl->sdl_alen) ? 1 : 0; 1737 if (olladdr && lladdr) { 1738 if (bcmp(lladdr, LLADDR(sdl), ifp->if_addrlen)) 1739 llchange = 1; 1740 else 1741 llchange = 0; 1742 } else 1743 llchange = 0; 1744 1745 /* 1746 * newentry olladdr lladdr llchange (*=record) 1747 * 0 n n -- (1) 1748 * 0 y n -- (2) 1749 * 0 n y -- (3) * STALE 1750 * 0 y y n (4) * 1751 * 0 y y y (5) * STALE 1752 * 1 -- n -- (6) NOSTATE(= PASSIVE) 1753 * 1 -- y -- (7) * STALE 1754 */ 1755 1756 if (lladdr) { /* (3-5) and (7) */ 1757 /* 1758 * Record source link-layer address 1759 * XXX is it dependent to ifp->if_type? 1760 */ 1761 sdl->sdl_alen = ifp->if_addrlen; 1762 bcopy(lladdr, LLADDR(sdl), ifp->if_addrlen); 1763 } 1764 1765 if (!is_newentry) { 1766 if ((!olladdr && lladdr != NULL) || /* (3) */ 1767 (olladdr && lladdr != NULL && llchange)) { /* (5) */ 1768 do_update = 1; 1769 newstate = ND6_LLINFO_STALE; 1770 } else /* (1-2,4) */ 1771 do_update = 0; 1772 } else { 1773 do_update = 1; 1774 if (lladdr == NULL) /* (6) */ 1775 newstate = ND6_LLINFO_NOSTATE; 1776 else /* (7) */ 1777 newstate = ND6_LLINFO_STALE; 1778 } 1779 1780 if (do_update) { 1781 /* 1782 * Update the state of the neighbor cache. 1783 */ 1784 ln->ln_state = newstate; 1785 1786 if (ln->ln_state == ND6_LLINFO_STALE) { 1787 /* 1788 * XXX: since nd6_output() below will cause 1789 * state tansition to DELAY and reset the timer, 1790 * we must set the timer now, although it is actually 1791 * meaningless. 1792 */ 1793 nd6_llinfo_settimer(ln, (long)nd6_gctimer * hz); 1794 1795 if (ln->ln_hold) { 1796 struct mbuf *m_hold, *m_hold_next; 1797 for (m_hold = ln->ln_hold; m_hold; 1798 m_hold = m_hold_next) { 1799 struct mbuf *mpkt = NULL; 1800 1801 m_hold_next = m_hold->m_nextpkt; 1802 mpkt = m_copym(m_hold, 0, M_COPYALL, M_DONTWAIT); 1803 if (mpkt == NULL) { 1804 m_freem(m_hold); 1805 break; 1806 } 1807 mpkt->m_nextpkt = NULL; 1808 1809 /* 1810 * we assume ifp is not a p2p here, so 1811 * just set the 2nd argument as the 1812 * 1st one. 1813 */ 1814 nd6_output(ifp, ifp, mpkt, 1815 (struct sockaddr_in6 *)rt_key(rt), 1816 rt); 1817 } 1818 ln->ln_hold = NULL; 1819 } 1820 } else if (ln->ln_state == ND6_LLINFO_INCOMPLETE) { 1821 /* probe right away */ 1822 nd6_llinfo_settimer((void *)ln, 0); 1823 } 1824 } 1825 1826 /* 1827 * ICMP6 type dependent behavior. 1828 * 1829 * NS: clear IsRouter if new entry 1830 * RS: clear IsRouter 1831 * RA: set IsRouter if there's lladdr 1832 * redir: clear IsRouter if new entry 1833 * 1834 * RA case, (1): 1835 * The spec says that we must set IsRouter in the following cases: 1836 * - If lladdr exist, set IsRouter. This means (1-5). 1837 * - If it is old entry (!newentry), set IsRouter. This means (7). 1838 * So, based on the spec, in (1-5) and (7) cases we must set IsRouter. 1839 * A quetion arises for (1) case. (1) case has no lladdr in the 1840 * neighbor cache, this is similar to (6). 1841 * This case is rare but we figured that we MUST NOT set IsRouter. 1842 * 1843 * newentry olladdr lladdr llchange NS RS RA redir 1844 * D R 1845 * 0 n n -- (1) c ? s 1846 * 0 y n -- (2) c s s 1847 * 0 n y -- (3) c s s 1848 * 0 y y n (4) c s s 1849 * 0 y y y (5) c s s 1850 * 1 -- n -- (6) c c c s 1851 * 1 -- y -- (7) c c s c s 1852 * 1853 * (c=clear s=set) 1854 */ 1855 switch (type & 0xff) { 1856 case ND_NEIGHBOR_SOLICIT: 1857 /* 1858 * New entry must have is_router flag cleared. 1859 */ 1860 if (is_newentry) /* (6-7) */ 1861 ln->ln_router = 0; 1862 break; 1863 case ND_REDIRECT: 1864 /* 1865 * If the icmp is a redirect to a better router, always set the 1866 * is_router flag. Otherwise, if the entry is newly created, 1867 * clear the flag. [RFC 2461, sec 8.3] 1868 */ 1869 if (code == ND_REDIRECT_ROUTER) 1870 ln->ln_router = 1; 1871 else if (is_newentry) /* (6-7) */ 1872 ln->ln_router = 0; 1873 break; 1874 case ND_ROUTER_SOLICIT: 1875 /* 1876 * is_router flag must always be cleared. 1877 */ 1878 ln->ln_router = 0; 1879 break; 1880 case ND_ROUTER_ADVERT: 1881 /* 1882 * Mark an entry with lladdr as a router. 1883 */ 1884 if ((!is_newentry && (olladdr || lladdr)) || /* (2-5) */ 1885 (is_newentry && lladdr)) { /* (7) */ 1886 ln->ln_router = 1; 1887 } 1888 break; 1889 } 1890 1891 /* 1892 * When the link-layer address of a router changes, select the 1893 * best router again. In particular, when the neighbor entry is newly 1894 * created, it might affect the selection policy. 1895 * Question: can we restrict the first condition to the "is_newentry" 1896 * case? 1897 * XXX: when we hear an RA from a new router with the link-layer 1898 * address option, defrouter_select() is called twice, since 1899 * defrtrlist_update called the function as well. However, I believe 1900 * we can compromise the overhead, since it only happens the first 1901 * time. 1902 * XXX: although defrouter_select() should not have a bad effect 1903 * for those are not autoconfigured hosts, we explicitly avoid such 1904 * cases for safety. 1905 */ 1906 if (do_update && ln->ln_router && !ip6_forwarding && ip6_accept_rtadv) 1907 defrouter_select(); 1908 1909 return rt; 1910 } 1911 1912 static void 1913 nd6_slowtimo(ignored_arg) 1914 void *ignored_arg; 1915 { 1916 struct nd_ifinfo *nd6if; 1917 struct ifnet *ifp; 1918 1919 callout_reset(&nd6_slowtimo_ch, ND6_SLOWTIMER_INTERVAL * hz, 1920 nd6_slowtimo, NULL); 1921 IFNET_RLOCK(); 1922 for (ifp = TAILQ_FIRST(&ifnet); ifp; ifp = TAILQ_NEXT(ifp, if_list)) { 1923 nd6if = ND_IFINFO(ifp); 1924 if (nd6if->basereachable && /* already initialized */ 1925 (nd6if->recalctm -= ND6_SLOWTIMER_INTERVAL) <= 0) { 1926 /* 1927 * Since reachable time rarely changes by router 1928 * advertisements, we SHOULD insure that a new random 1929 * value gets recomputed at least once every few hours. 1930 * (RFC 2461, 6.3.4) 1931 */ 1932 nd6if->recalctm = nd6_recalc_reachtm_interval; 1933 nd6if->reachable = ND_COMPUTE_RTIME(nd6if->basereachable); 1934 } 1935 } 1936 IFNET_RUNLOCK(); 1937 } 1938 1939 #define senderr(e) { error = (e); goto bad;} 1940 int 1941 nd6_output(ifp, origifp, m0, dst, rt0) 1942 struct ifnet *ifp; 1943 struct ifnet *origifp; 1944 struct mbuf *m0; 1945 struct sockaddr_in6 *dst; 1946 struct rtentry *rt0; 1947 { 1948 struct mbuf *m = m0; 1949 struct rtentry *rt = rt0; 1950 struct sockaddr_in6 *gw6 = NULL; 1951 struct llinfo_nd6 *ln = NULL; 1952 int error = 0; 1953 1954 if (IN6_IS_ADDR_MULTICAST(&dst->sin6_addr)) 1955 goto sendpkt; 1956 1957 if (nd6_need_cache(ifp) == 0) 1958 goto sendpkt; 1959 1960 /* 1961 * next hop determination. This routine is derived from ether_output. 1962 */ 1963 /* NB: the locking here is tortuous... */ 1964 if (rt != NULL) 1965 RT_LOCK(rt); 1966 again: 1967 if (rt != NULL) { 1968 if ((rt->rt_flags & RTF_UP) == 0) { 1969 RT_UNLOCK(rt); 1970 rt0 = rt = rtalloc1((struct sockaddr *)dst, 1, 0UL); 1971 if (rt != NULL) { 1972 RT_REMREF(rt); 1973 if (rt->rt_ifp != ifp) 1974 /* 1975 * XXX maybe we should update ifp too, 1976 * but the original code didn't and I 1977 * don't know what is correct here. 1978 */ 1979 goto again; 1980 } else 1981 senderr(EHOSTUNREACH); 1982 } 1983 1984 if (rt->rt_flags & RTF_GATEWAY) { 1985 gw6 = (struct sockaddr_in6 *)rt->rt_gateway; 1986 1987 /* 1988 * We skip link-layer address resolution and NUD 1989 * if the gateway is not a neighbor from ND point 1990 * of view, regardless of the value of nd_ifinfo.flags. 1991 * The second condition is a bit tricky; we skip 1992 * if the gateway is our own address, which is 1993 * sometimes used to install a route to a p2p link. 1994 */ 1995 if (!nd6_is_addr_neighbor(gw6, ifp) || 1996 in6ifa_ifpwithaddr(ifp, &gw6->sin6_addr)) { 1997 RT_UNLOCK(rt); 1998 /* 1999 * We allow this kind of tricky route only 2000 * when the outgoing interface is p2p. 2001 * XXX: we may need a more generic rule here. 2002 */ 2003 if ((ifp->if_flags & IFF_POINTOPOINT) == 0) 2004 senderr(EHOSTUNREACH); 2005 2006 goto sendpkt; 2007 } 2008 2009 if (rt->rt_gwroute == NULL) 2010 goto lookup; 2011 rt = rt->rt_gwroute; 2012 RT_LOCK(rt); /* NB: gwroute */ 2013 if ((rt->rt_flags & RTF_UP) == 0) { 2014 rtfree(rt); /* unlock gwroute */ 2015 rt = rt0; 2016 lookup: 2017 RT_UNLOCK(rt0); 2018 rt = rtalloc1(rt->rt_gateway, 1, 0UL); 2019 if (rt == rt0) { 2020 rt0->rt_gwroute = NULL; 2021 RT_REMREF(rt0); 2022 RT_UNLOCK(rt0); 2023 senderr(EHOSTUNREACH); 2024 } 2025 RT_LOCK(rt0); 2026 rt0->rt_gwroute = rt; 2027 if (rt == NULL) { 2028 RT_UNLOCK(rt0); 2029 senderr(EHOSTUNREACH); 2030 } 2031 } 2032 RT_UNLOCK(rt0); 2033 } 2034 RT_UNLOCK(rt); 2035 } 2036 2037 /* 2038 * Address resolution or Neighbor Unreachability Detection 2039 * for the next hop. 2040 * At this point, the destination of the packet must be a unicast 2041 * or an anycast address(i.e. not a multicast). 2042 */ 2043 2044 /* Look up the neighbor cache for the nexthop */ 2045 if (rt && (rt->rt_flags & RTF_LLINFO) != 0) 2046 ln = (struct llinfo_nd6 *)rt->rt_llinfo; 2047 else { 2048 /* 2049 * Since nd6_is_addr_neighbor() internally calls nd6_lookup(), 2050 * the condition below is not very efficient. But we believe 2051 * it is tolerable, because this should be a rare case. 2052 */ 2053 if (nd6_is_addr_neighbor(dst, ifp) && 2054 (rt = nd6_lookup(&dst->sin6_addr, 1, ifp)) != NULL) 2055 ln = (struct llinfo_nd6 *)rt->rt_llinfo; 2056 } 2057 if (ln == NULL || rt == NULL) { 2058 if ((ifp->if_flags & IFF_POINTOPOINT) == 0 && 2059 !(ND_IFINFO(ifp)->flags & ND6_IFF_PERFORMNUD)) { 2060 char ip6buf[INET6_ADDRSTRLEN]; 2061 log(LOG_DEBUG, 2062 "nd6_output: can't allocate llinfo for %s " 2063 "(ln=%p, rt=%p)\n", 2064 ip6_sprintf(ip6buf, &dst->sin6_addr), ln, rt); 2065 senderr(EIO); /* XXX: good error? */ 2066 } 2067 2068 goto sendpkt; /* send anyway */ 2069 } 2070 2071 /* We don't have to do link-layer address resolution on a p2p link. */ 2072 if ((ifp->if_flags & IFF_POINTOPOINT) != 0 && 2073 ln->ln_state < ND6_LLINFO_REACHABLE) { 2074 ln->ln_state = ND6_LLINFO_STALE; 2075 nd6_llinfo_settimer(ln, (long)nd6_gctimer * hz); 2076 } 2077 2078 /* 2079 * The first time we send a packet to a neighbor whose entry is 2080 * STALE, we have to change the state to DELAY and a sets a timer to 2081 * expire in DELAY_FIRST_PROBE_TIME seconds to ensure do 2082 * neighbor unreachability detection on expiration. 2083 * (RFC 2461 7.3.3) 2084 */ 2085 if (ln->ln_state == ND6_LLINFO_STALE) { 2086 ln->ln_asked = 0; 2087 ln->ln_state = ND6_LLINFO_DELAY; 2088 nd6_llinfo_settimer(ln, (long)nd6_delay * hz); 2089 } 2090 2091 /* 2092 * If the neighbor cache entry has a state other than INCOMPLETE 2093 * (i.e. its link-layer address is already resolved), just 2094 * send the packet. 2095 */ 2096 if (ln->ln_state > ND6_LLINFO_INCOMPLETE) 2097 goto sendpkt; 2098 2099 /* 2100 * There is a neighbor cache entry, but no ethernet address 2101 * response yet. Append this latest packet to the end of the 2102 * packet queue in the mbuf, unless the number of the packet 2103 * does not exceed nd6_maxqueuelen. When it exceeds nd6_maxqueuelen, 2104 * the oldest packet in the queue will be removed. 2105 */ 2106 if (ln->ln_state == ND6_LLINFO_NOSTATE) 2107 ln->ln_state = ND6_LLINFO_INCOMPLETE; 2108 if (ln->ln_hold) { 2109 struct mbuf *m_hold; 2110 int i; 2111 2112 i = 0; 2113 for (m_hold = ln->ln_hold; m_hold; m_hold = m_hold->m_nextpkt) { 2114 i++; 2115 if (m_hold->m_nextpkt == NULL) { 2116 m_hold->m_nextpkt = m; 2117 break; 2118 } 2119 } 2120 while (i >= nd6_maxqueuelen) { 2121 m_hold = ln->ln_hold; 2122 ln->ln_hold = ln->ln_hold->m_nextpkt; 2123 m_freem(m_hold); 2124 i--; 2125 } 2126 } else { 2127 ln->ln_hold = m; 2128 } 2129 2130 /* 2131 * If there has been no NS for the neighbor after entering the 2132 * INCOMPLETE state, send the first solicitation. 2133 */ 2134 if (!ND6_LLINFO_PERMANENT(ln) && ln->ln_asked == 0) { 2135 ln->ln_asked++; 2136 nd6_llinfo_settimer(ln, 2137 (long)ND_IFINFO(ifp)->retrans * hz / 1000); 2138 nd6_ns_output(ifp, NULL, &dst->sin6_addr, ln, 0); 2139 } 2140 return (0); 2141 2142 sendpkt: 2143 /* discard the packet if IPv6 operation is disabled on the interface */ 2144 if ((ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED)) { 2145 error = ENETDOWN; /* better error? */ 2146 goto bad; 2147 } 2148 2149 #ifdef IPSEC 2150 /* clean ipsec history once it goes out of the node */ 2151 ipsec_delaux(m); 2152 #endif 2153 2154 #ifdef MAC 2155 mac_create_mbuf_linklayer(ifp, m); 2156 #endif 2157 if ((ifp->if_flags & IFF_LOOPBACK) != 0) { 2158 return ((*ifp->if_output)(origifp, m, (struct sockaddr *)dst, 2159 rt)); 2160 } 2161 return ((*ifp->if_output)(ifp, m, (struct sockaddr *)dst, rt)); 2162 2163 bad: 2164 if (m) 2165 m_freem(m); 2166 return (error); 2167 } 2168 #undef senderr 2169 2170 int 2171 nd6_need_cache(ifp) 2172 struct ifnet *ifp; 2173 { 2174 /* 2175 * XXX: we currently do not make neighbor cache on any interface 2176 * other than ARCnet, Ethernet, FDDI and GIF. 2177 * 2178 * RFC2893 says: 2179 * - unidirectional tunnels needs no ND 2180 */ 2181 switch (ifp->if_type) { 2182 case IFT_ARCNET: 2183 case IFT_ETHER: 2184 case IFT_FDDI: 2185 case IFT_IEEE1394: 2186 #ifdef IFT_L2VLAN 2187 case IFT_L2VLAN: 2188 #endif 2189 #ifdef IFT_IEEE80211 2190 case IFT_IEEE80211: 2191 #endif 2192 #ifdef IFT_CARP 2193 case IFT_CARP: 2194 #endif 2195 case IFT_GIF: /* XXX need more cases? */ 2196 case IFT_PPP: 2197 case IFT_TUNNEL: 2198 case IFT_BRIDGE: 2199 case IFT_PROPVIRTUAL: 2200 return (1); 2201 default: 2202 return (0); 2203 } 2204 } 2205 2206 int 2207 nd6_storelladdr(ifp, rt0, m, dst, desten) 2208 struct ifnet *ifp; 2209 struct rtentry *rt0; 2210 struct mbuf *m; 2211 struct sockaddr *dst; 2212 u_char *desten; 2213 { 2214 struct sockaddr_dl *sdl; 2215 struct rtentry *rt; 2216 int error; 2217 2218 if (m->m_flags & M_MCAST) { 2219 int i; 2220 2221 switch (ifp->if_type) { 2222 case IFT_ETHER: 2223 case IFT_FDDI: 2224 #ifdef IFT_L2VLAN 2225 case IFT_L2VLAN: 2226 #endif 2227 #ifdef IFT_IEEE80211 2228 case IFT_IEEE80211: 2229 #endif 2230 case IFT_BRIDGE: 2231 case IFT_ISO88025: 2232 ETHER_MAP_IPV6_MULTICAST(&SIN6(dst)->sin6_addr, 2233 desten); 2234 return (0); 2235 case IFT_IEEE1394: 2236 /* 2237 * netbsd can use if_broadcastaddr, but we don't do so 2238 * to reduce # of ifdef. 2239 */ 2240 for (i = 0; i < ifp->if_addrlen; i++) 2241 desten[i] = ~0; 2242 return (0); 2243 case IFT_ARCNET: 2244 *desten = 0; 2245 return (0); 2246 default: 2247 m_freem(m); 2248 return (EAFNOSUPPORT); 2249 } 2250 } 2251 2252 if (rt0 == NULL) { 2253 /* this could happen, if we could not allocate memory */ 2254 m_freem(m); 2255 return (ENOMEM); 2256 } 2257 2258 error = rt_check(&rt, &rt0, dst); 2259 if (error) { 2260 m_freem(m); 2261 return (error); 2262 } 2263 RT_UNLOCK(rt); 2264 2265 if (rt->rt_gateway->sa_family != AF_LINK) { 2266 printf("nd6_storelladdr: something odd happens\n"); 2267 m_freem(m); 2268 return (EINVAL); 2269 } 2270 sdl = SDL(rt->rt_gateway); 2271 if (sdl->sdl_alen == 0) { 2272 /* this should be impossible, but we bark here for debugging */ 2273 printf("nd6_storelladdr: sdl_alen == 0\n"); 2274 m_freem(m); 2275 return (EINVAL); 2276 } 2277 2278 bcopy(LLADDR(sdl), desten, sdl->sdl_alen); 2279 return (0); 2280 } 2281 2282 static void 2283 clear_llinfo_pqueue(ln) 2284 struct llinfo_nd6 *ln; 2285 { 2286 struct mbuf *m_hold, *m_hold_next; 2287 2288 for (m_hold = ln->ln_hold; m_hold; m_hold = m_hold_next) { 2289 m_hold_next = m_hold->m_nextpkt; 2290 m_hold->m_nextpkt = NULL; 2291 m_freem(m_hold); 2292 } 2293 2294 ln->ln_hold = NULL; 2295 return; 2296 } 2297 2298 static int nd6_sysctl_drlist(SYSCTL_HANDLER_ARGS); 2299 static int nd6_sysctl_prlist(SYSCTL_HANDLER_ARGS); 2300 #ifdef SYSCTL_DECL 2301 SYSCTL_DECL(_net_inet6_icmp6); 2302 #endif 2303 SYSCTL_NODE(_net_inet6_icmp6, ICMPV6CTL_ND6_DRLIST, nd6_drlist, 2304 CTLFLAG_RD, nd6_sysctl_drlist, ""); 2305 SYSCTL_NODE(_net_inet6_icmp6, ICMPV6CTL_ND6_PRLIST, nd6_prlist, 2306 CTLFLAG_RD, nd6_sysctl_prlist, ""); 2307 SYSCTL_INT(_net_inet6_icmp6, ICMPV6CTL_ND6_MAXQLEN, nd6_maxqueuelen, 2308 CTLFLAG_RW, &nd6_maxqueuelen, 1, ""); 2309 2310 static int 2311 nd6_sysctl_drlist(SYSCTL_HANDLER_ARGS) 2312 { 2313 int error; 2314 char buf[1024]; 2315 struct in6_defrouter *d, *de; 2316 struct nd_defrouter *dr; 2317 2318 if (req->newptr) 2319 return EPERM; 2320 error = 0; 2321 2322 for (dr = TAILQ_FIRST(&nd_defrouter); dr; 2323 dr = TAILQ_NEXT(dr, dr_entry)) { 2324 d = (struct in6_defrouter *)buf; 2325 de = (struct in6_defrouter *)(buf + sizeof(buf)); 2326 2327 if (d + 1 <= de) { 2328 bzero(d, sizeof(*d)); 2329 d->rtaddr.sin6_family = AF_INET6; 2330 d->rtaddr.sin6_len = sizeof(d->rtaddr); 2331 d->rtaddr.sin6_addr = dr->rtaddr; 2332 sa6_recoverscope(&d->rtaddr); 2333 d->flags = dr->flags; 2334 d->rtlifetime = dr->rtlifetime; 2335 d->expire = dr->expire; 2336 d->if_index = dr->ifp->if_index; 2337 } else 2338 panic("buffer too short"); 2339 2340 error = SYSCTL_OUT(req, buf, sizeof(*d)); 2341 if (error) 2342 break; 2343 } 2344 2345 return (error); 2346 } 2347 2348 static int 2349 nd6_sysctl_prlist(SYSCTL_HANDLER_ARGS) 2350 { 2351 int error; 2352 char buf[1024]; 2353 struct in6_prefix *p, *pe; 2354 struct nd_prefix *pr; 2355 char ip6buf[INET6_ADDRSTRLEN]; 2356 2357 if (req->newptr) 2358 return EPERM; 2359 error = 0; 2360 2361 for (pr = nd_prefix.lh_first; pr; pr = pr->ndpr_next) { 2362 u_short advrtrs; 2363 size_t advance; 2364 struct sockaddr_in6 *sin6, *s6; 2365 struct nd_pfxrouter *pfr; 2366 2367 p = (struct in6_prefix *)buf; 2368 pe = (struct in6_prefix *)(buf + sizeof(buf)); 2369 2370 if (p + 1 <= pe) { 2371 bzero(p, sizeof(*p)); 2372 sin6 = (struct sockaddr_in6 *)(p + 1); 2373 2374 p->prefix = pr->ndpr_prefix; 2375 if (sa6_recoverscope(&p->prefix)) { 2376 log(LOG_ERR, 2377 "scope error in prefix list (%s)\n", 2378 ip6_sprintf(ip6buf, &p->prefix.sin6_addr)); 2379 /* XXX: press on... */ 2380 } 2381 p->raflags = pr->ndpr_raf; 2382 p->prefixlen = pr->ndpr_plen; 2383 p->vltime = pr->ndpr_vltime; 2384 p->pltime = pr->ndpr_pltime; 2385 p->if_index = pr->ndpr_ifp->if_index; 2386 if (pr->ndpr_vltime == ND6_INFINITE_LIFETIME) 2387 p->expire = 0; 2388 else { 2389 time_t maxexpire; 2390 2391 /* XXX: we assume time_t is signed. */ 2392 maxexpire = (-1) & 2393 ~((time_t)1 << 2394 ((sizeof(maxexpire) * 8) - 1)); 2395 if (pr->ndpr_vltime < 2396 maxexpire - pr->ndpr_lastupdate) { 2397 p->expire = pr->ndpr_lastupdate + 2398 pr->ndpr_vltime; 2399 } else 2400 p->expire = maxexpire; 2401 } 2402 p->refcnt = pr->ndpr_refcnt; 2403 p->flags = pr->ndpr_stateflags; 2404 p->origin = PR_ORIG_RA; 2405 advrtrs = 0; 2406 for (pfr = pr->ndpr_advrtrs.lh_first; pfr; 2407 pfr = pfr->pfr_next) { 2408 if ((void *)&sin6[advrtrs + 1] > (void *)pe) { 2409 advrtrs++; 2410 continue; 2411 } 2412 s6 = &sin6[advrtrs]; 2413 bzero(s6, sizeof(*s6)); 2414 s6->sin6_family = AF_INET6; 2415 s6->sin6_len = sizeof(*sin6); 2416 s6->sin6_addr = pfr->router->rtaddr; 2417 if (sa6_recoverscope(s6)) { 2418 log(LOG_ERR, 2419 "scope error in " 2420 "prefix list (%s)\n", 2421 ip6_sprintf(ip6buf, 2422 &pfr->router->rtaddr)); 2423 } 2424 advrtrs++; 2425 } 2426 p->advrtrs = advrtrs; 2427 } else 2428 panic("buffer too short"); 2429 2430 advance = sizeof(*p) + sizeof(*sin6) * advrtrs; 2431 error = SYSCTL_OUT(req, buf, advance); 2432 if (error) 2433 break; 2434 } 2435 2436 return (error); 2437 } 2438