1 /* $FreeBSD: src/sys/netinet6/nd6.c,v 1.2.2.15 2003/05/06 06:46:58 suz Exp $ */ 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 /* 34 * XXX 35 * KAME 970409 note: 36 * BSD/OS version heavily modifies this code, related to llinfo. 37 * Since we don't have BSD/OS version of net/route.c in our hand, 38 * I left the code mostly as it was in 970310. -- itojun 39 */ 40 41 #include "opt_inet.h" 42 #include "opt_inet6.h" 43 44 #include <sys/param.h> 45 #include <sys/systm.h> 46 #include <sys/callout.h> 47 #include <sys/malloc.h> 48 #include <sys/mbuf.h> 49 #include <sys/socket.h> 50 #include <sys/sockio.h> 51 #include <sys/time.h> 52 #include <sys/kernel.h> 53 #include <sys/protosw.h> 54 #include <sys/errno.h> 55 #include <sys/syslog.h> 56 #include <sys/queue.h> 57 #include <sys/sysctl.h> 58 #include <sys/mutex.h> 59 60 #include <sys/thread2.h> 61 #include <sys/mutex2.h> 62 63 #include <net/if.h> 64 #include <net/if_dl.h> 65 #include <net/if_types.h> 66 #include <net/if_atm.h> 67 #include <net/route.h> 68 69 #include <netinet/in.h> 70 #include <netinet/if_ether.h> 71 #include <netinet6/in6_var.h> 72 #include <netinet/ip6.h> 73 #include <netinet6/ip6_var.h> 74 #include <netinet6/nd6.h> 75 #include <netinet6/in6_prefix.h> 76 #include <netinet/icmp6.h> 77 78 #include <net/net_osdep.h> 79 80 #define ND6_SLOWTIMER_INTERVAL (60 * 60) /* 1 hour */ 81 #define ND6_RECALC_REACHTM_INTERVAL (60 * 120) /* 2 hours */ 82 83 #define SIN6(s) ((struct sockaddr_in6 *)s) 84 #define SDL(s) ((struct sockaddr_dl *)s) 85 86 /* timer values */ 87 int nd6_prune = 1; /* walk list every 1 seconds */ 88 int nd6_delay = 5; /* delay first probe time 5 second */ 89 int nd6_umaxtries = 3; /* maximum unicast query */ 90 int nd6_mmaxtries = 3; /* maximum multicast query */ 91 int nd6_useloopback = 1; /* use loopback interface for local traffic */ 92 int nd6_gctimer = (60 * 60 * 24); /* 1 day: garbage collection timer */ 93 94 /* preventing too many loops in ND option parsing */ 95 int nd6_maxndopt = 10; /* max # of ND options allowed */ 96 97 int nd6_maxnudhint = 0; /* max # of subsequent upper layer hints */ 98 99 #ifdef ND6_DEBUG 100 int nd6_debug = 1; 101 #else 102 int nd6_debug = 0; 103 #endif 104 105 /* for debugging? */ 106 static int nd6_inuse, nd6_allocated; 107 108 struct llinfo_nd6 llinfo_nd6 = {&llinfo_nd6, &llinfo_nd6}; 109 struct nd_drhead nd_defrouter; 110 struct nd_prhead nd_prefix = { 0 }; 111 struct mtx nd6_mtx = MTX_INITIALIZER; 112 113 int nd6_recalc_reachtm_interval = ND6_RECALC_REACHTM_INTERVAL; 114 static struct sockaddr_in6 all1_sa; 115 116 static void nd6_setmtu0 (struct ifnet *, struct nd_ifinfo *); 117 static void nd6_slowtimo (void *); 118 static int regen_tmpaddr (struct in6_ifaddr *); 119 120 struct callout nd6_slowtimo_ch; 121 struct callout nd6_timer_ch; 122 extern struct callout in6_tmpaddrtimer_ch; 123 124 void 125 nd6_init(void) 126 { 127 static int nd6_init_done = 0; 128 int i; 129 130 if (nd6_init_done) { 131 log(LOG_NOTICE, "nd6_init called more than once(ignored)\n"); 132 return; 133 } 134 135 all1_sa.sin6_family = AF_INET6; 136 all1_sa.sin6_len = sizeof(struct sockaddr_in6); 137 for (i = 0; i < sizeof(all1_sa.sin6_addr); i++) 138 all1_sa.sin6_addr.s6_addr[i] = 0xff; 139 140 /* initialization of the default router list */ 141 TAILQ_INIT(&nd_defrouter); 142 143 nd6_init_done = 1; 144 145 /* start timer */ 146 callout_init(&nd6_slowtimo_ch); 147 callout_reset(&nd6_slowtimo_ch, ND6_SLOWTIMER_INTERVAL * hz, 148 nd6_slowtimo, NULL); 149 } 150 151 struct nd_ifinfo * 152 nd6_ifattach(struct ifnet *ifp) 153 { 154 struct nd_ifinfo *nd; 155 156 nd = (struct nd_ifinfo *)kmalloc(sizeof(*nd), M_IP6NDP, 157 M_WAITOK | M_ZERO); 158 159 nd->initialized = 1; 160 161 nd->linkmtu = ifindex2ifnet[ifp->if_index]->if_mtu; 162 nd->chlim = IPV6_DEFHLIM; 163 nd->basereachable = REACHABLE_TIME; 164 nd->reachable = ND_COMPUTE_RTIME(nd->basereachable); 165 nd->retrans = RETRANS_TIMER; 166 nd->receivedra = 0; 167 168 /* 169 * Note that the default value of ip6_accept_rtadv is 0, which means 170 * we won't accept RAs by default even if we set ND6_IFF_ACCEPT_RTADV 171 * here. 172 */ 173 nd->flags = (ND6_IFF_PERFORMNUD | ND6_IFF_ACCEPT_RTADV); 174 175 /* XXX: we cannot call nd6_setmtu since ifp is not fully initialized */ 176 nd6_setmtu0(ifp, nd); 177 return nd; 178 } 179 180 void 181 nd6_ifdetach(struct nd_ifinfo *nd) 182 { 183 kfree(nd, M_IP6NDP); 184 } 185 186 /* 187 * Reset ND level link MTU. This function is called when the physical MTU 188 * changes, which means we might have to adjust the ND level MTU. 189 */ 190 void 191 nd6_setmtu(struct ifnet *ifp) 192 { 193 nd6_setmtu0(ifp, ND_IFINFO(ifp)); 194 } 195 196 /* XXX todo: do not maintain copy of ifp->if_mtu in ndi->maxmtu */ 197 void 198 nd6_setmtu0(struct ifnet *ifp, struct nd_ifinfo *ndi) 199 { 200 u_long oldmaxmtu; 201 u_long oldlinkmtu; 202 203 oldmaxmtu = ndi->maxmtu; 204 oldlinkmtu = ndi->linkmtu; 205 206 switch (ifp->if_type) { 207 case IFT_ETHER: 208 ndi->maxmtu = MIN(ETHERMTU, ifp->if_mtu); 209 break; 210 case IFT_ATM: 211 ndi->maxmtu = MIN(ATMMTU, ifp->if_mtu); 212 break; 213 case IFT_IEEE1394: /* XXX should be IEEE1394MTU(1500) */ 214 ndi->maxmtu = MIN(ETHERMTU, ifp->if_mtu); 215 break; 216 #ifdef IFT_IEEE80211 217 case IFT_IEEE80211: /* XXX should be IEEE80211MTU(1500) */ 218 ndi->maxmtu = MIN(ETHERMTU, ifp->if_mtu); 219 break; 220 #endif 221 default: 222 ndi->maxmtu = ifp->if_mtu; 223 break; 224 } 225 226 if (oldmaxmtu != ndi->maxmtu) { 227 /* 228 * If the ND level MTU is not set yet, or if the maxmtu 229 * is reset to a smaller value than the ND level MTU, 230 * also reset the ND level MTU. 231 */ 232 if (ndi->linkmtu == 0 || 233 ndi->maxmtu < ndi->linkmtu) { 234 ndi->linkmtu = ndi->maxmtu; 235 /* also adjust in6_maxmtu if necessary. */ 236 if (oldlinkmtu == 0) { 237 /* 238 * XXX: the case analysis is grotty, but 239 * it is not efficient to call in6_setmaxmtu() 240 * here when we are during the initialization 241 * procedure. 242 */ 243 if (in6_maxmtu < ndi->linkmtu) 244 in6_maxmtu = ndi->linkmtu; 245 } else 246 in6_setmaxmtu(); 247 } 248 } 249 #undef MIN 250 } 251 252 void 253 nd6_option_init(void *opt, int icmp6len, union nd_opts *ndopts) 254 { 255 bzero(ndopts, sizeof(*ndopts)); 256 ndopts->nd_opts_search = (struct nd_opt_hdr *)opt; 257 ndopts->nd_opts_last 258 = (struct nd_opt_hdr *)(((u_char *)opt) + icmp6len); 259 260 if (icmp6len == 0) { 261 ndopts->nd_opts_done = 1; 262 ndopts->nd_opts_search = NULL; 263 } 264 } 265 266 /* 267 * Take one ND option. 268 */ 269 struct nd_opt_hdr * 270 nd6_option(union nd_opts *ndopts) 271 { 272 struct nd_opt_hdr *nd_opt; 273 int olen; 274 275 if (!ndopts) 276 panic("ndopts == NULL in nd6_option"); 277 if (!ndopts->nd_opts_last) 278 panic("uninitialized ndopts in nd6_option"); 279 if (!ndopts->nd_opts_search) 280 return NULL; 281 if (ndopts->nd_opts_done) 282 return NULL; 283 284 nd_opt = ndopts->nd_opts_search; 285 286 /* make sure nd_opt_len is inside the buffer */ 287 if ((caddr_t)&nd_opt->nd_opt_len >= (caddr_t)ndopts->nd_opts_last) { 288 bzero(ndopts, sizeof(*ndopts)); 289 return NULL; 290 } 291 292 olen = nd_opt->nd_opt_len << 3; 293 if (olen == 0) { 294 /* 295 * Message validation requires that all included 296 * options have a length that is greater than zero. 297 */ 298 bzero(ndopts, sizeof(*ndopts)); 299 return NULL; 300 } 301 302 ndopts->nd_opts_search = (struct nd_opt_hdr *)((caddr_t)nd_opt + olen); 303 if (ndopts->nd_opts_search > ndopts->nd_opts_last) { 304 /* option overruns the end of buffer, invalid */ 305 bzero(ndopts, sizeof(*ndopts)); 306 return NULL; 307 } else if (ndopts->nd_opts_search == ndopts->nd_opts_last) { 308 /* reached the end of options chain */ 309 ndopts->nd_opts_done = 1; 310 ndopts->nd_opts_search = NULL; 311 } 312 return nd_opt; 313 } 314 315 /* 316 * Parse multiple ND options. 317 * This function is much easier to use, for ND routines that do not need 318 * multiple options of the same type. 319 */ 320 int 321 nd6_options(union nd_opts *ndopts) 322 { 323 struct nd_opt_hdr *nd_opt; 324 int i = 0; 325 326 if (!ndopts) 327 panic("ndopts == NULL in nd6_options"); 328 if (!ndopts->nd_opts_last) 329 panic("uninitialized ndopts in nd6_options"); 330 if (!ndopts->nd_opts_search) 331 return 0; 332 333 while (1) { 334 nd_opt = nd6_option(ndopts); 335 if (!nd_opt && !ndopts->nd_opts_last) { 336 /* 337 * Message validation requires that all included 338 * options have a length that is greater than zero. 339 */ 340 icmp6stat.icp6s_nd_badopt++; 341 bzero(ndopts, sizeof(*ndopts)); 342 return -1; 343 } 344 345 if (!nd_opt) 346 goto skip1; 347 348 switch (nd_opt->nd_opt_type) { 349 case ND_OPT_SOURCE_LINKADDR: 350 case ND_OPT_TARGET_LINKADDR: 351 case ND_OPT_MTU: 352 case ND_OPT_REDIRECTED_HEADER: 353 if (ndopts->nd_opt_array[nd_opt->nd_opt_type]) { 354 nd6log((LOG_INFO, 355 "duplicated ND6 option found (type=%d)\n", 356 nd_opt->nd_opt_type)); 357 /* XXX bark? */ 358 } else { 359 ndopts->nd_opt_array[nd_opt->nd_opt_type] 360 = nd_opt; 361 } 362 break; 363 case ND_OPT_PREFIX_INFORMATION: 364 if (ndopts->nd_opt_array[nd_opt->nd_opt_type] == 0) { 365 ndopts->nd_opt_array[nd_opt->nd_opt_type] 366 = nd_opt; 367 } 368 ndopts->nd_opts_pi_end = 369 (struct nd_opt_prefix_info *)nd_opt; 370 break; 371 default: 372 /* 373 * Unknown options must be silently ignored, 374 * to accomodate future extension to the protocol. 375 */ 376 nd6log((LOG_DEBUG, 377 "nd6_options: unsupported option %d - " 378 "option ignored\n", nd_opt->nd_opt_type)); 379 } 380 381 skip1: 382 i++; 383 if (i > nd6_maxndopt) { 384 icmp6stat.icp6s_nd_toomanyopt++; 385 nd6log((LOG_INFO, "too many loop in nd opt\n")); 386 break; 387 } 388 389 if (ndopts->nd_opts_done) 390 break; 391 } 392 393 return 0; 394 } 395 396 /* 397 * ND6 timer routine to expire default route list and prefix list 398 */ 399 void 400 nd6_timer(void *ignored_arg) 401 { 402 struct llinfo_nd6 *ln; 403 struct nd_defrouter *dr; 404 struct nd_prefix *pr; 405 struct ifnet *ifp; 406 struct in6_ifaddr *ia6, *nia6; 407 struct in6_addrlifetime *lt6; 408 409 mtx_lock(&nd6_mtx); 410 callout_reset(&nd6_timer_ch, nd6_prune * hz, 411 nd6_timer, NULL); 412 413 ln = llinfo_nd6.ln_next; 414 while (ln && ln != &llinfo_nd6) { 415 struct rtentry *rt; 416 struct sockaddr_in6 *dst; 417 struct llinfo_nd6 *next = ln->ln_next; 418 /* XXX: used for the DELAY case only: */ 419 struct nd_ifinfo *ndi = NULL; 420 421 if ((rt = ln->ln_rt) == NULL) { 422 ln = next; 423 continue; 424 } 425 if ((ifp = rt->rt_ifp) == NULL) { 426 ln = next; 427 continue; 428 } 429 ndi = ND_IFINFO(ifp); 430 dst = (struct sockaddr_in6 *)rt_key(rt); 431 432 if (ln->ln_expire > time_second) { 433 ln = next; 434 continue; 435 } 436 437 /* sanity check */ 438 if (!rt) 439 panic("rt=0 in nd6_timer(ln=%p)", ln); 440 if (rt->rt_llinfo && (struct llinfo_nd6 *)rt->rt_llinfo != ln) 441 panic("rt_llinfo(%p) is not equal to ln(%p)", 442 rt->rt_llinfo, ln); 443 if (!dst) 444 panic("dst=0 in nd6_timer(ln=%p)", ln); 445 446 switch (ln->ln_state) { 447 case ND6_LLINFO_INCOMPLETE: 448 if (ln->ln_asked < nd6_mmaxtries) { 449 ln->ln_asked++; 450 ln->ln_expire = time_second + 451 ND_IFINFO(ifp)->retrans / 1000; 452 nd6_ns_output(ifp, NULL, &dst->sin6_addr, 453 ln, 0); 454 } else { 455 struct mbuf *m = ln->ln_hold; 456 if (m) { 457 if (rt->rt_ifp) { 458 /* 459 * Fake rcvif to make ICMP error 460 * more helpful in diagnosing 461 * for the receiver. 462 * XXX: should we consider 463 * older rcvif? 464 */ 465 m->m_pkthdr.rcvif = rt->rt_ifp; 466 } 467 icmp6_error(m, ICMP6_DST_UNREACH, 468 ICMP6_DST_UNREACH_ADDR, 0); 469 ln->ln_hold = NULL; 470 } 471 next = nd6_free(rt); 472 } 473 break; 474 case ND6_LLINFO_REACHABLE: 475 if (ln->ln_expire) { 476 ln->ln_state = ND6_LLINFO_STALE; 477 ln->ln_expire = time_second + nd6_gctimer; 478 } 479 break; 480 481 case ND6_LLINFO_STALE: 482 /* Garbage Collection(RFC 2461 5.3) */ 483 if (ln->ln_expire) 484 next = nd6_free(rt); 485 break; 486 487 case ND6_LLINFO_DELAY: 488 if (ndi && (ndi->flags & ND6_IFF_PERFORMNUD)) { 489 /* We need NUD */ 490 ln->ln_asked = 1; 491 ln->ln_state = ND6_LLINFO_PROBE; 492 ln->ln_expire = time_second + 493 ndi->retrans / 1000; 494 nd6_ns_output(ifp, &dst->sin6_addr, 495 &dst->sin6_addr, 496 ln, 0); 497 } else { 498 ln->ln_state = ND6_LLINFO_STALE; /* XXX */ 499 ln->ln_expire = time_second + nd6_gctimer; 500 } 501 break; 502 case ND6_LLINFO_PROBE: 503 if (ln->ln_asked < nd6_umaxtries) { 504 ln->ln_asked++; 505 ln->ln_expire = time_second + 506 ND_IFINFO(ifp)->retrans / 1000; 507 nd6_ns_output(ifp, &dst->sin6_addr, 508 &dst->sin6_addr, ln, 0); 509 } else { 510 next = nd6_free(rt); 511 } 512 break; 513 } 514 ln = next; 515 } 516 517 /* expire default router list */ 518 dr = TAILQ_FIRST(&nd_defrouter); 519 while (dr) { 520 if (dr->expire && dr->expire < time_second) { 521 struct nd_defrouter *t; 522 t = TAILQ_NEXT(dr, dr_entry); 523 defrtrlist_del(dr); 524 dr = t; 525 } else { 526 dr = TAILQ_NEXT(dr, dr_entry); 527 } 528 } 529 530 /* 531 * expire interface addresses. 532 * in the past the loop was inside prefix expiry processing. 533 * However, from a stricter speci-confrmance standpoint, we should 534 * rather separate address lifetimes and prefix lifetimes. 535 */ 536 addrloop: 537 for (ia6 = in6_ifaddr; ia6; ia6 = nia6) { 538 nia6 = ia6->ia_next; 539 /* check address lifetime */ 540 lt6 = &ia6->ia6_lifetime; 541 if (IFA6_IS_INVALID(ia6)) { 542 int regen = 0; 543 544 /* 545 * If the expiring address is temporary, try 546 * regenerating a new one. This would be useful when 547 * we suspended a laptop PC, then turned it on after a 548 * period that could invalidate all temporary 549 * addresses. Although we may have to restart the 550 * loop (see below), it must be after purging the 551 * address. Otherwise, we'd see an infinite loop of 552 * regeneration. 553 */ 554 if (ip6_use_tempaddr && 555 (ia6->ia6_flags & IN6_IFF_TEMPORARY)) { 556 if (regen_tmpaddr(ia6) == 0) 557 regen = 1; 558 } 559 560 in6_purgeaddr(&ia6->ia_ifa); 561 562 if (regen) 563 goto addrloop; /* XXX: see below */ 564 } 565 if (IFA6_IS_DEPRECATED(ia6)) { 566 int oldflags = ia6->ia6_flags; 567 568 ia6->ia6_flags |= IN6_IFF_DEPRECATED; 569 570 /* 571 * If a temporary address has just become deprecated, 572 * regenerate a new one if possible. 573 */ 574 if (ip6_use_tempaddr && 575 (ia6->ia6_flags & IN6_IFF_TEMPORARY) && 576 !(oldflags & IN6_IFF_DEPRECATED)) { 577 578 if (regen_tmpaddr(ia6) == 0) { 579 /* 580 * A new temporary address is 581 * generated. 582 * XXX: this means the address chain 583 * has changed while we are still in 584 * the loop. Although the change 585 * would not cause disaster (because 586 * it's not a deletion, but an 587 * addition,) we'd rather restart the 588 * loop just for safety. Or does this 589 * significantly reduce performance?? 590 */ 591 goto addrloop; 592 } 593 } 594 } else { 595 /* 596 * A new RA might have made a deprecated address 597 * preferred. 598 */ 599 ia6->ia6_flags &= ~IN6_IFF_DEPRECATED; 600 } 601 } 602 603 /* expire prefix list */ 604 pr = nd_prefix.lh_first; 605 while (pr) { 606 /* 607 * check prefix lifetime. 608 * since pltime is just for autoconf, pltime processing for 609 * prefix is not necessary. 610 */ 611 if (pr->ndpr_expire && pr->ndpr_expire < time_second) { 612 struct nd_prefix *t; 613 t = pr->ndpr_next; 614 615 /* 616 * address expiration and prefix expiration are 617 * separate. NEVER perform in6_purgeaddr here. 618 */ 619 620 prelist_remove(pr); 621 pr = t; 622 } else 623 pr = pr->ndpr_next; 624 } 625 mtx_unlock(&nd6_mtx); 626 } 627 628 static int 629 regen_tmpaddr(struct in6_ifaddr *ia6) /* deprecated/invalidated temporary 630 address */ 631 { 632 struct ifaddr_container *ifac; 633 struct ifnet *ifp; 634 struct in6_ifaddr *public_ifa6 = NULL; 635 636 ifp = ia6->ia_ifa.ifa_ifp; 637 TAILQ_FOREACH(ifac, &ifp->if_addrheads[mycpuid], ifa_link) { 638 struct ifaddr *ifa = ifac->ifa; 639 struct in6_ifaddr *it6; 640 641 if (ifa->ifa_addr->sa_family != AF_INET6) 642 continue; 643 644 it6 = (struct in6_ifaddr *)ifa; 645 646 /* ignore no autoconf addresses. */ 647 if (!(it6->ia6_flags & IN6_IFF_AUTOCONF)) 648 continue; 649 650 /* ignore autoconf addresses with different prefixes. */ 651 if (it6->ia6_ndpr == NULL || it6->ia6_ndpr != ia6->ia6_ndpr) 652 continue; 653 654 /* 655 * Now we are looking at an autoconf address with the same 656 * prefix as ours. If the address is temporary and is still 657 * preferred, do not create another one. It would be rare, but 658 * could happen, for example, when we resume a laptop PC after 659 * a long period. 660 */ 661 if ((it6->ia6_flags & IN6_IFF_TEMPORARY) && 662 !IFA6_IS_DEPRECATED(it6)) { 663 public_ifa6 = NULL; 664 break; 665 } 666 667 /* 668 * This is a public autoconf address that has the same prefix 669 * as ours. If it is preferred, keep it. We can't break the 670 * loop here, because there may be a still-preferred temporary 671 * address with the prefix. 672 */ 673 if (!IFA6_IS_DEPRECATED(it6)) 674 public_ifa6 = it6; 675 } 676 677 if (public_ifa6 != NULL) { 678 int e; 679 680 if ((e = in6_tmpifadd(public_ifa6, 0)) != 0) { 681 log(LOG_NOTICE, "regen_tmpaddr: failed to create a new" 682 " tmp addr,errno=%d\n", e); 683 return (-1); 684 } 685 return (0); 686 } 687 688 return (-1); 689 } 690 691 /* 692 * Nuke neighbor cache/prefix/default router management table, right before 693 * ifp goes away. 694 */ 695 void 696 nd6_purge(struct ifnet *ifp) 697 { 698 struct llinfo_nd6 *ln, *nln; 699 struct nd_defrouter *dr, *ndr, drany; 700 struct nd_prefix *pr, *npr; 701 702 /* Nuke default router list entries toward ifp */ 703 if ((dr = TAILQ_FIRST(&nd_defrouter)) != NULL) { 704 /* 705 * The first entry of the list may be stored in 706 * the routing table, so we'll delete it later. 707 */ 708 for (dr = TAILQ_NEXT(dr, dr_entry); dr; dr = ndr) { 709 ndr = TAILQ_NEXT(dr, dr_entry); 710 if (dr->ifp == ifp) 711 defrtrlist_del(dr); 712 } 713 dr = TAILQ_FIRST(&nd_defrouter); 714 if (dr->ifp == ifp) 715 defrtrlist_del(dr); 716 } 717 718 /* Nuke prefix list entries toward ifp */ 719 for (pr = nd_prefix.lh_first; pr; pr = npr) { 720 npr = pr->ndpr_next; 721 if (pr->ndpr_ifp == ifp) { 722 /* 723 * Previously, pr->ndpr_addr is removed as well, 724 * but I strongly believe we don't have to do it. 725 * nd6_purge() is only called from in6_ifdetach(), 726 * which removes all the associated interface addresses 727 * by itself. 728 * (jinmei@kame.net 20010129) 729 */ 730 prelist_remove(pr); 731 } 732 } 733 734 /* cancel default outgoing interface setting */ 735 if (nd6_defifindex == ifp->if_index) 736 nd6_setdefaultiface(0); 737 738 if (!ip6_forwarding && ip6_accept_rtadv) { /* XXX: too restrictive? */ 739 /* refresh default router list */ 740 bzero(&drany, sizeof(drany)); 741 defrouter_delreq(&drany, 0); 742 defrouter_select(); 743 } 744 745 /* 746 * Nuke neighbor cache entries for the ifp. 747 * Note that rt->rt_ifp may not be the same as ifp, 748 * due to KAME goto ours hack. See RTM_RESOLVE case in 749 * nd6_rtrequest(), and ip6_input(). 750 */ 751 ln = llinfo_nd6.ln_next; 752 while (ln && ln != &llinfo_nd6) { 753 struct rtentry *rt; 754 struct sockaddr_dl *sdl; 755 756 nln = ln->ln_next; 757 rt = ln->ln_rt; 758 if (rt && rt->rt_gateway && 759 rt->rt_gateway->sa_family == AF_LINK) { 760 sdl = (struct sockaddr_dl *)rt->rt_gateway; 761 if (sdl->sdl_index == ifp->if_index) 762 nln = nd6_free(rt); 763 } 764 ln = nln; 765 } 766 } 767 768 struct rtentry * 769 nd6_lookup(struct in6_addr *addr6, int create, struct ifnet *ifp) 770 { 771 struct rtentry *rt; 772 struct sockaddr_in6 sin6; 773 774 bzero(&sin6, sizeof(sin6)); 775 sin6.sin6_len = sizeof(struct sockaddr_in6); 776 sin6.sin6_family = AF_INET6; 777 sin6.sin6_addr = *addr6; 778 779 if (create) 780 rt = rtlookup((struct sockaddr *)&sin6); 781 else 782 rt = rtpurelookup((struct sockaddr *)&sin6); 783 if (rt && !(rt->rt_flags & RTF_LLINFO)) { 784 /* 785 * This is the case for the default route. 786 * If we want to create a neighbor cache for the address, we 787 * should free the route for the destination and allocate an 788 * interface route. 789 */ 790 if (create) { 791 --rt->rt_refcnt; 792 rt = NULL; 793 } 794 } 795 if (!rt) { 796 if (create && ifp) { 797 int e; 798 799 /* 800 * If no route is available and create is set, 801 * we allocate a host route for the destination 802 * and treat it like an interface route. 803 * This hack is necessary for a neighbor which can't 804 * be covered by our own prefix. 805 */ 806 struct ifaddr *ifa = 807 ifaof_ifpforaddr((struct sockaddr *)&sin6, ifp); 808 if (ifa == NULL) 809 return (NULL); 810 811 /* 812 * Create a new route. RTF_LLINFO is necessary 813 * to create a Neighbor Cache entry for the 814 * destination in nd6_rtrequest which will be 815 * called in rtrequest via ifa->ifa_rtrequest. 816 */ 817 if ((e = rtrequest(RTM_ADD, (struct sockaddr *)&sin6, 818 ifa->ifa_addr, 819 (struct sockaddr *)&all1_sa, 820 (ifa->ifa_flags | 821 RTF_HOST | RTF_LLINFO) & 822 ~RTF_CLONING, 823 &rt)) != 0) 824 log(LOG_ERR, 825 "nd6_lookup: failed to add route for a " 826 "neighbor(%s), errno=%d\n", 827 ip6_sprintf(addr6), e); 828 if (rt == NULL) 829 return (NULL); 830 if (rt->rt_llinfo) { 831 struct llinfo_nd6 *ln = 832 (struct llinfo_nd6 *)rt->rt_llinfo; 833 ln->ln_state = ND6_LLINFO_NOSTATE; 834 } 835 } else 836 return (NULL); 837 } 838 rt->rt_refcnt--; 839 /* 840 * Validation for the entry. 841 * Note that the check for rt_llinfo is necessary because a cloned 842 * route from a parent route that has the L flag (e.g. the default 843 * route to a p2p interface) may have the flag, too, while the 844 * destination is not actually a neighbor. 845 * XXX: we can't use rt->rt_ifp to check for the interface, since 846 * it might be the loopback interface if the entry is for our 847 * own address on a non-loopback interface. Instead, we should 848 * use rt->rt_ifa->ifa_ifp, which would specify the REAL 849 * interface. 850 */ 851 if ((rt->rt_flags & RTF_GATEWAY) || !(rt->rt_flags & RTF_LLINFO) || 852 rt->rt_gateway->sa_family != AF_LINK || rt->rt_llinfo == NULL || 853 (ifp && rt->rt_ifa->ifa_ifp != ifp)) { 854 if (create) { 855 log(LOG_DEBUG, "nd6_lookup: failed to lookup %s (if = %s)\n", 856 ip6_sprintf(addr6), ifp ? if_name(ifp) : "unspec"); 857 /* xxx more logs... kazu */ 858 } 859 return (NULL); 860 } 861 return (rt); 862 } 863 864 /* 865 * Detect if a given IPv6 address identifies a neighbor on a given link. 866 * XXX: should take care of the destination of a p2p link? 867 */ 868 int 869 nd6_is_addr_neighbor(struct sockaddr_in6 *addr, struct ifnet *ifp) 870 { 871 struct ifaddr_container *ifac; 872 int i; 873 874 #define IFADDR6(a) ((((struct in6_ifaddr *)(a))->ia_addr).sin6_addr) 875 #define IFMASK6(a) ((((struct in6_ifaddr *)(a))->ia_prefixmask).sin6_addr) 876 877 /* 878 * A link-local address is always a neighbor. 879 * XXX: we should use the sin6_scope_id field rather than the embedded 880 * interface index. 881 */ 882 if (IN6_IS_ADDR_LINKLOCAL(&addr->sin6_addr) && 883 ntohs(*(u_int16_t *)&addr->sin6_addr.s6_addr[2]) == ifp->if_index) 884 return (1); 885 886 /* 887 * If the address matches one of our addresses, 888 * it should be a neighbor. 889 */ 890 TAILQ_FOREACH(ifac, &ifp->if_addrheads[mycpuid], ifa_link) { 891 struct ifaddr *ifa = ifac->ifa; 892 893 if (ifa->ifa_addr->sa_family != AF_INET6) 894 next: continue; 895 896 for (i = 0; i < 4; i++) { 897 if ((IFADDR6(ifa).s6_addr32[i] ^ 898 addr->sin6_addr.s6_addr32[i]) & 899 IFMASK6(ifa).s6_addr32[i]) 900 goto next; 901 } 902 return (1); 903 } 904 905 /* 906 * Even if the address matches none of our addresses, it might be 907 * in the neighbor cache. 908 */ 909 if (nd6_lookup(&addr->sin6_addr, 0, ifp) != NULL) 910 return (1); 911 912 return (0); 913 #undef IFADDR6 914 #undef IFMASK6 915 } 916 917 /* 918 * Free an nd6 llinfo entry. 919 */ 920 struct llinfo_nd6 * 921 nd6_free(struct rtentry *rt) 922 { 923 struct llinfo_nd6 *ln = (struct llinfo_nd6 *)rt->rt_llinfo, *next; 924 struct in6_addr in6 = ((struct sockaddr_in6 *)rt_key(rt))->sin6_addr; 925 struct nd_defrouter *dr; 926 927 /* 928 * we used to have kpfctlinput(PRC_HOSTDEAD) here. 929 * even though it is not harmful, it was not really necessary. 930 */ 931 932 if (!ip6_forwarding && ip6_accept_rtadv) { /* XXX: too restrictive? */ 933 mtx_lock(&nd6_mtx); 934 dr = defrouter_lookup(&((struct sockaddr_in6 *)rt_key(rt))->sin6_addr, 935 rt->rt_ifp); 936 937 if (ln->ln_router || dr) { 938 /* 939 * rt6_flush must be called whether or not the neighbor 940 * is in the Default Router List. 941 * See a corresponding comment in nd6_na_input(). 942 */ 943 rt6_flush(&in6, rt->rt_ifp); 944 } 945 946 if (dr) { 947 /* 948 * Unreachablity of a router might affect the default 949 * router selection and on-link detection of advertised 950 * prefixes. 951 */ 952 953 /* 954 * Temporarily fake the state to choose a new default 955 * router and to perform on-link determination of 956 * prefixes correctly. 957 * Below the state will be set correctly, 958 * or the entry itself will be deleted. 959 */ 960 ln->ln_state = ND6_LLINFO_INCOMPLETE; 961 962 /* 963 * Since defrouter_select() does not affect the 964 * on-link determination and MIP6 needs the check 965 * before the default router selection, we perform 966 * the check now. 967 */ 968 pfxlist_onlink_check(); 969 970 if (dr == TAILQ_FIRST(&nd_defrouter)) { 971 /* 972 * It is used as the current default router, 973 * so we have to move it to the end of the 974 * list and choose a new one. 975 * XXX: it is not very efficient if this is 976 * the only router. 977 */ 978 TAILQ_REMOVE(&nd_defrouter, dr, dr_entry); 979 TAILQ_INSERT_TAIL(&nd_defrouter, dr, dr_entry); 980 981 defrouter_select(); 982 } 983 } 984 mtx_unlock(&nd6_mtx); 985 } 986 987 /* 988 * Before deleting the entry, remember the next entry as the 989 * return value. We need this because pfxlist_onlink_check() above 990 * might have freed other entries (particularly the old next entry) as 991 * a side effect (XXX). 992 */ 993 next = ln->ln_next; 994 995 /* 996 * Detach the route from the routing tree and the list of neighbor 997 * caches, and disable the route entry not to be used in already 998 * cached routes. 999 */ 1000 rtrequest(RTM_DELETE, rt_key(rt), NULL, rt_mask(rt), 0, NULL); 1001 1002 return (next); 1003 } 1004 1005 /* 1006 * Upper-layer reachability hint for Neighbor Unreachability Detection. 1007 * 1008 * XXX cost-effective metods? 1009 */ 1010 void 1011 nd6_nud_hint(struct rtentry *rt, struct in6_addr *dst6, int force) 1012 { 1013 struct llinfo_nd6 *ln; 1014 1015 /* 1016 * If the caller specified "rt", use that. Otherwise, resolve the 1017 * routing table by supplied "dst6". 1018 */ 1019 if (!rt) { 1020 if (!dst6) 1021 return; 1022 if (!(rt = nd6_lookup(dst6, 0, NULL))) 1023 return; 1024 } 1025 1026 if ((rt->rt_flags & RTF_GATEWAY) || 1027 !(rt->rt_flags & RTF_LLINFO) || 1028 rt->rt_llinfo == NULL || rt->rt_gateway == NULL || 1029 rt->rt_gateway->sa_family != AF_LINK) { 1030 /* This is not a host route. */ 1031 return; 1032 } 1033 1034 ln = (struct llinfo_nd6 *)rt->rt_llinfo; 1035 if (ln->ln_state < ND6_LLINFO_REACHABLE) 1036 return; 1037 1038 /* 1039 * if we get upper-layer reachability confirmation many times, 1040 * it is possible we have false information. 1041 */ 1042 if (!force) { 1043 ln->ln_byhint++; 1044 if (ln->ln_byhint > nd6_maxnudhint) 1045 return; 1046 } 1047 1048 ln->ln_state = ND6_LLINFO_REACHABLE; 1049 if (ln->ln_expire) 1050 ln->ln_expire = time_second + 1051 ND_IFINFO(rt->rt_ifp)->reachable; 1052 } 1053 1054 void 1055 nd6_rtrequest(int req, struct rtentry *rt, 1056 struct rt_addrinfo *info) /* xxx unused */ 1057 { 1058 struct sockaddr *gate = rt->rt_gateway; 1059 struct llinfo_nd6 *ln = (struct llinfo_nd6 *)rt->rt_llinfo; 1060 static struct sockaddr_dl null_sdl = {sizeof(null_sdl), AF_LINK}; 1061 struct ifnet *ifp = rt->rt_ifp; 1062 struct ifaddr *ifa; 1063 1064 if ((rt->rt_flags & RTF_GATEWAY)) 1065 return; 1066 1067 if (nd6_need_cache(ifp) == 0 && !(rt->rt_flags & RTF_HOST)) { 1068 /* 1069 * This is probably an interface direct route for a link 1070 * which does not need neighbor caches (e.g. fe80::%lo0/64). 1071 * We do not need special treatment below for such a route. 1072 * Moreover, the RTF_LLINFO flag which would be set below 1073 * would annoy the ndp(8) command. 1074 */ 1075 return; 1076 } 1077 1078 if (req == RTM_RESOLVE && 1079 (nd6_need_cache(ifp) == 0 || /* stf case */ 1080 !nd6_is_addr_neighbor((struct sockaddr_in6 *)rt_key(rt), ifp))) { 1081 /* 1082 * FreeBSD and BSD/OS often make a cloned host route based 1083 * on a less-specific route (e.g. the default route). 1084 * If the less specific route does not have a "gateway" 1085 * (this is the case when the route just goes to a p2p or an 1086 * stf interface), we'll mistakenly make a neighbor cache for 1087 * the host route, and will see strange neighbor solicitation 1088 * for the corresponding destination. In order to avoid the 1089 * confusion, we check if the destination of the route is 1090 * a neighbor in terms of neighbor discovery, and stop the 1091 * process if not. Additionally, we remove the LLINFO flag 1092 * so that ndp(8) will not try to get the neighbor information 1093 * of the destination. 1094 */ 1095 rt->rt_flags &= ~RTF_LLINFO; 1096 return; 1097 } 1098 1099 switch (req) { 1100 case RTM_ADD: 1101 /* 1102 * There is no backward compatibility :) 1103 * 1104 * if (!(rt->rt_flags & RTF_HOST) && 1105 * SIN(rt_mask(rt))->sin_addr.s_addr != 0xffffffff) 1106 * rt->rt_flags |= RTF_CLONING; 1107 */ 1108 if (rt->rt_flags & (RTF_CLONING | RTF_LLINFO)) { 1109 /* 1110 * Case 1: This route should come from 1111 * a route to interface. RTF_LLINFO flag is set 1112 * for a host route whose destination should be 1113 * treated as on-link. 1114 */ 1115 rt_setgate(rt, rt_key(rt), 1116 (struct sockaddr *)&null_sdl, 1117 RTL_DONTREPORT); 1118 gate = rt->rt_gateway; 1119 SDL(gate)->sdl_type = ifp->if_type; 1120 SDL(gate)->sdl_index = ifp->if_index; 1121 if (ln) 1122 ln->ln_expire = time_second; 1123 #if 1 1124 if (ln && ln->ln_expire == 0) { 1125 /* kludge for desktops */ 1126 #if 0 1127 kprintf("nd6_rtequest: time.tv_sec is zero; " 1128 "treat it as 1\n"); 1129 #endif 1130 ln->ln_expire = 1; 1131 } 1132 #endif 1133 if ((rt->rt_flags & RTF_CLONING)) 1134 break; 1135 } 1136 /* 1137 * In IPv4 code, we try to annonuce new RTF_ANNOUNCE entry here. 1138 * We don't do that here since llinfo is not ready yet. 1139 * 1140 * There are also couple of other things to be discussed: 1141 * - unsolicited NA code needs improvement beforehand 1142 * - RFC2461 says we MAY send multicast unsolicited NA 1143 * (7.2.6 paragraph 4), however, it also says that we 1144 * SHOULD provide a mechanism to prevent multicast NA storm. 1145 * we don't have anything like it right now. 1146 * note that the mechanism needs a mutual agreement 1147 * between proxies, which means that we need to implement 1148 * a new protocol, or a new kludge. 1149 * - from RFC2461 6.2.4, host MUST NOT send an unsolicited NA. 1150 * we need to check ip6forwarding before sending it. 1151 * (or should we allow proxy ND configuration only for 1152 * routers? there's no mention about proxy ND from hosts) 1153 */ 1154 #if 0 1155 /* XXX it does not work */ 1156 if (rt->rt_flags & RTF_ANNOUNCE) 1157 nd6_na_output(ifp, 1158 &SIN6(rt_key(rt))->sin6_addr, 1159 &SIN6(rt_key(rt))->sin6_addr, 1160 ip6_forwarding ? ND_NA_FLAG_ROUTER : 0, 1161 1, NULL); 1162 #endif 1163 /* FALLTHROUGH */ 1164 case RTM_RESOLVE: 1165 if ((ifp->if_flags & (IFF_POINTOPOINT | IFF_LOOPBACK)) == 0) { 1166 /* 1167 * Address resolution isn't necessary for a point to 1168 * point link, so we can skip this test for a p2p link. 1169 */ 1170 if (gate->sa_family != AF_LINK || 1171 gate->sa_len < sizeof(null_sdl)) { 1172 log(LOG_DEBUG, 1173 "nd6_rtrequest: bad gateway value: %s\n", 1174 if_name(ifp)); 1175 break; 1176 } 1177 SDL(gate)->sdl_type = ifp->if_type; 1178 SDL(gate)->sdl_index = ifp->if_index; 1179 } 1180 if (ln != NULL) 1181 break; /* This happens on a route change */ 1182 /* 1183 * Case 2: This route may come from cloning, or a manual route 1184 * add with a LL address. 1185 */ 1186 R_Malloc(ln, struct llinfo_nd6 *, sizeof(*ln)); 1187 rt->rt_llinfo = (caddr_t)ln; 1188 if (!ln) { 1189 log(LOG_DEBUG, "nd6_rtrequest: malloc failed\n"); 1190 break; 1191 } 1192 nd6_inuse++; 1193 nd6_allocated++; 1194 bzero(ln, sizeof(*ln)); 1195 ln->ln_rt = rt; 1196 /* this is required for "ndp" command. - shin */ 1197 if (req == RTM_ADD) { 1198 /* 1199 * gate should have some valid AF_LINK entry, 1200 * and ln->ln_expire should have some lifetime 1201 * which is specified by ndp command. 1202 */ 1203 ln->ln_state = ND6_LLINFO_REACHABLE; 1204 ln->ln_byhint = 0; 1205 } else { 1206 /* 1207 * When req == RTM_RESOLVE, rt is created and 1208 * initialized in rtrequest(), so rt_expire is 0. 1209 */ 1210 ln->ln_state = ND6_LLINFO_NOSTATE; 1211 ln->ln_expire = time_second; 1212 } 1213 rt->rt_flags |= RTF_LLINFO; 1214 ln->ln_next = llinfo_nd6.ln_next; 1215 llinfo_nd6.ln_next = ln; 1216 ln->ln_prev = &llinfo_nd6; 1217 ln->ln_next->ln_prev = ln; 1218 1219 /* 1220 * check if rt_key(rt) is one of my address assigned 1221 * to the interface. 1222 */ 1223 ifa = (struct ifaddr *)in6ifa_ifpwithaddr(rt->rt_ifp, 1224 &SIN6(rt_key(rt))->sin6_addr); 1225 if (ifa) { 1226 caddr_t macp = nd6_ifptomac(ifp); 1227 ln->ln_expire = 0; 1228 ln->ln_state = ND6_LLINFO_REACHABLE; 1229 ln->ln_byhint = 0; 1230 if (macp) { 1231 bcopy(macp, LLADDR(SDL(gate)), ifp->if_addrlen); 1232 SDL(gate)->sdl_alen = ifp->if_addrlen; 1233 } 1234 if (nd6_useloopback) { 1235 rt->rt_ifp = &loif[0]; /* XXX */ 1236 /* 1237 * Make sure rt_ifa be equal to the ifaddr 1238 * corresponding to the address. 1239 * We need this because when we refer 1240 * rt_ifa->ia6_flags in ip6_input, we assume 1241 * that the rt_ifa points to the address instead 1242 * of the loopback address. 1243 */ 1244 if (ifa != rt->rt_ifa) { 1245 IFAFREE(rt->rt_ifa); 1246 IFAREF(ifa); 1247 rt->rt_ifa = ifa; 1248 } 1249 } 1250 } else if (rt->rt_flags & RTF_ANNOUNCE) { 1251 ln->ln_expire = 0; 1252 ln->ln_state = ND6_LLINFO_REACHABLE; 1253 ln->ln_byhint = 0; 1254 1255 /* join solicited node multicast for proxy ND */ 1256 if (ifp->if_flags & IFF_MULTICAST) { 1257 struct in6_addr llsol; 1258 int error; 1259 1260 llsol = SIN6(rt_key(rt))->sin6_addr; 1261 llsol.s6_addr16[0] = htons(0xff02); 1262 llsol.s6_addr16[1] = htons(ifp->if_index); 1263 llsol.s6_addr32[1] = 0; 1264 llsol.s6_addr32[2] = htonl(1); 1265 llsol.s6_addr8[12] = 0xff; 1266 1267 if (!in6_addmulti(&llsol, ifp, &error)) { 1268 nd6log((LOG_ERR, "%s: failed to join " 1269 "%s (errno=%d)\n", if_name(ifp), 1270 ip6_sprintf(&llsol), error)); 1271 } 1272 } 1273 } 1274 break; 1275 1276 case RTM_DELETE: 1277 if (!ln) 1278 break; 1279 /* leave from solicited node multicast for proxy ND */ 1280 if ((rt->rt_flags & RTF_ANNOUNCE) && 1281 (ifp->if_flags & IFF_MULTICAST)) { 1282 struct in6_addr llsol; 1283 struct in6_multi *in6m; 1284 1285 llsol = SIN6(rt_key(rt))->sin6_addr; 1286 llsol.s6_addr16[0] = htons(0xff02); 1287 llsol.s6_addr16[1] = htons(ifp->if_index); 1288 llsol.s6_addr32[1] = 0; 1289 llsol.s6_addr32[2] = htonl(1); 1290 llsol.s6_addr8[12] = 0xff; 1291 1292 IN6_LOOKUP_MULTI(llsol, ifp, in6m); 1293 if (in6m) 1294 in6_delmulti(in6m); 1295 } 1296 nd6_inuse--; 1297 ln->ln_next->ln_prev = ln->ln_prev; 1298 ln->ln_prev->ln_next = ln->ln_next; 1299 ln->ln_prev = NULL; 1300 rt->rt_llinfo = 0; 1301 rt->rt_flags &= ~RTF_LLINFO; 1302 if (ln->ln_hold) 1303 m_freem(ln->ln_hold); 1304 Free((caddr_t)ln); 1305 } 1306 } 1307 1308 int 1309 nd6_ioctl(u_long cmd, caddr_t data, struct ifnet *ifp) 1310 { 1311 struct in6_drlist *drl = (struct in6_drlist *)data; 1312 struct in6_prlist *prl = (struct in6_prlist *)data; 1313 struct in6_ndireq *ndi = (struct in6_ndireq *)data; 1314 struct in6_nbrinfo *nbi = (struct in6_nbrinfo *)data; 1315 struct in6_ndifreq *ndif = (struct in6_ndifreq *)data; 1316 struct nd_defrouter *dr, any; 1317 struct nd_prefix *pr; 1318 struct rtentry *rt; 1319 int i = 0, error = 0; 1320 1321 switch (cmd) { 1322 case SIOCGDRLST_IN6: 1323 /* 1324 * obsolete API, use sysctl under net.inet6.icmp6 1325 */ 1326 bzero(drl, sizeof(*drl)); 1327 mtx_lock(&nd6_mtx); 1328 dr = TAILQ_FIRST(&nd_defrouter); 1329 while (dr && i < DRLSTSIZ) { 1330 drl->defrouter[i].rtaddr = dr->rtaddr; 1331 if (IN6_IS_ADDR_LINKLOCAL(&drl->defrouter[i].rtaddr)) { 1332 /* XXX: need to this hack for KAME stack */ 1333 drl->defrouter[i].rtaddr.s6_addr16[1] = 0; 1334 } else 1335 log(LOG_ERR, 1336 "default router list contains a " 1337 "non-linklocal address(%s)\n", 1338 ip6_sprintf(&drl->defrouter[i].rtaddr)); 1339 1340 drl->defrouter[i].flags = dr->flags; 1341 drl->defrouter[i].rtlifetime = dr->rtlifetime; 1342 drl->defrouter[i].expire = dr->expire; 1343 drl->defrouter[i].if_index = dr->ifp->if_index; 1344 i++; 1345 dr = TAILQ_NEXT(dr, dr_entry); 1346 } 1347 mtx_unlock(&nd6_mtx); 1348 break; 1349 case SIOCGPRLST_IN6: 1350 /* 1351 * obsolete API, use sysctl under net.inet6.icmp6 1352 */ 1353 /* 1354 * XXX meaning of fields, especialy "raflags", is very 1355 * differnet between RA prefix list and RR/static prefix list. 1356 * how about separating ioctls into two? 1357 */ 1358 bzero(prl, sizeof(*prl)); 1359 mtx_lock(&nd6_mtx); 1360 pr = nd_prefix.lh_first; 1361 while (pr && i < PRLSTSIZ) { 1362 struct nd_pfxrouter *pfr; 1363 int j; 1364 1365 in6_embedscope(&prl->prefix[i].prefix, 1366 &pr->ndpr_prefix, NULL, NULL); 1367 prl->prefix[i].raflags = pr->ndpr_raf; 1368 prl->prefix[i].prefixlen = pr->ndpr_plen; 1369 prl->prefix[i].vltime = pr->ndpr_vltime; 1370 prl->prefix[i].pltime = pr->ndpr_pltime; 1371 prl->prefix[i].if_index = pr->ndpr_ifp->if_index; 1372 prl->prefix[i].expire = pr->ndpr_expire; 1373 1374 pfr = pr->ndpr_advrtrs.lh_first; 1375 j = 0; 1376 while (pfr) { 1377 if (j < DRLSTSIZ) { 1378 #define RTRADDR prl->prefix[i].advrtr[j] 1379 RTRADDR = pfr->router->rtaddr; 1380 if (IN6_IS_ADDR_LINKLOCAL(&RTRADDR)) { 1381 /* XXX: hack for KAME */ 1382 RTRADDR.s6_addr16[1] = 0; 1383 } else 1384 log(LOG_ERR, 1385 "a router(%s) advertises " 1386 "a prefix with " 1387 "non-link local address\n", 1388 ip6_sprintf(&RTRADDR)); 1389 #undef RTRADDR 1390 } 1391 j++; 1392 pfr = pfr->pfr_next; 1393 } 1394 prl->prefix[i].advrtrs = j; 1395 prl->prefix[i].origin = PR_ORIG_RA; 1396 1397 i++; 1398 pr = pr->ndpr_next; 1399 } 1400 { 1401 struct rr_prefix *rpp; 1402 1403 for (rpp = LIST_FIRST(&rr_prefix); rpp; 1404 rpp = LIST_NEXT(rpp, rp_entry)) { 1405 if (i >= PRLSTSIZ) 1406 break; 1407 in6_embedscope(&prl->prefix[i].prefix, 1408 &pr->ndpr_prefix, NULL, NULL); 1409 prl->prefix[i].raflags = rpp->rp_raf; 1410 prl->prefix[i].prefixlen = rpp->rp_plen; 1411 prl->prefix[i].vltime = rpp->rp_vltime; 1412 prl->prefix[i].pltime = rpp->rp_pltime; 1413 prl->prefix[i].if_index = rpp->rp_ifp->if_index; 1414 prl->prefix[i].expire = rpp->rp_expire; 1415 prl->prefix[i].advrtrs = 0; 1416 prl->prefix[i].origin = rpp->rp_origin; 1417 i++; 1418 } 1419 } 1420 mtx_unlock(&nd6_mtx); 1421 1422 break; 1423 case OSIOCGIFINFO_IN6: 1424 /* XXX: old ndp(8) assumes a positive value for linkmtu. */ 1425 bzero(&ndi->ndi, sizeof(ndi->ndi)); 1426 ndi->ndi.linkmtu = ND_IFINFO(ifp)->linkmtu; 1427 ndi->ndi.maxmtu = ND_IFINFO(ifp)->maxmtu; 1428 ndi->ndi.basereachable = ND_IFINFO(ifp)->basereachable; 1429 ndi->ndi.reachable = ND_IFINFO(ifp)->reachable; 1430 ndi->ndi.retrans = ND_IFINFO(ifp)->retrans; 1431 ndi->ndi.flags = ND_IFINFO(ifp)->flags; 1432 ndi->ndi.recalctm = ND_IFINFO(ifp)->recalctm; 1433 ndi->ndi.chlim = ND_IFINFO(ifp)->chlim; 1434 ndi->ndi.receivedra = ND_IFINFO(ifp)->receivedra; 1435 break; 1436 case SIOCGIFINFO_IN6: 1437 ndi->ndi = *ND_IFINFO(ifp); 1438 break; 1439 case SIOCSIFINFO_FLAGS: 1440 ND_IFINFO(ifp)->flags = ndi->ndi.flags; 1441 break; 1442 case SIOCSNDFLUSH_IN6: /* XXX: the ioctl name is confusing... */ 1443 /* flush default router list */ 1444 /* 1445 * xxx sumikawa: should not delete route if default 1446 * route equals to the top of default router list 1447 */ 1448 bzero(&any, sizeof(any)); 1449 defrouter_delreq(&any, 0); 1450 defrouter_select(); 1451 /* xxx sumikawa: flush prefix list */ 1452 break; 1453 case SIOCSPFXFLUSH_IN6: 1454 { 1455 /* flush all the prefix advertised by routers */ 1456 struct nd_prefix *pr, *next; 1457 1458 mtx_lock(&nd6_mtx); 1459 for (pr = nd_prefix.lh_first; pr; pr = next) { 1460 struct in6_ifaddr *ia, *ia_next; 1461 1462 next = pr->ndpr_next; 1463 1464 if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr)) 1465 continue; /* XXX */ 1466 1467 /* do we really have to remove addresses as well? */ 1468 for (ia = in6_ifaddr; ia; ia = ia_next) { 1469 /* ia might be removed. keep the next ptr. */ 1470 ia_next = ia->ia_next; 1471 1472 if (!(ia->ia6_flags & IN6_IFF_AUTOCONF)) 1473 continue; 1474 1475 if (ia->ia6_ndpr == pr) 1476 in6_purgeaddr(&ia->ia_ifa); 1477 } 1478 prelist_remove(pr); 1479 } 1480 mtx_unlock(&nd6_mtx); 1481 break; 1482 } 1483 case SIOCSRTRFLUSH_IN6: 1484 { 1485 /* flush all the default routers */ 1486 struct nd_defrouter *dr, *next; 1487 1488 mtx_lock(&nd6_mtx); 1489 if ((dr = TAILQ_FIRST(&nd_defrouter)) != NULL) { 1490 /* 1491 * The first entry of the list may be stored in 1492 * the routing table, so we'll delete it later. 1493 */ 1494 for (dr = TAILQ_NEXT(dr, dr_entry); dr; dr = next) { 1495 next = TAILQ_NEXT(dr, dr_entry); 1496 defrtrlist_del(dr); 1497 } 1498 defrtrlist_del(TAILQ_FIRST(&nd_defrouter)); 1499 } 1500 mtx_unlock(&nd6_mtx); 1501 break; 1502 } 1503 case SIOCGNBRINFO_IN6: 1504 { 1505 struct llinfo_nd6 *ln; 1506 struct in6_addr nb_addr = nbi->addr; /* make local for safety */ 1507 1508 /* 1509 * XXX: KAME specific hack for scoped addresses 1510 * XXXX: for other scopes than link-local? 1511 */ 1512 if (IN6_IS_ADDR_LINKLOCAL(&nbi->addr) || 1513 IN6_IS_ADDR_MC_LINKLOCAL(&nbi->addr)) { 1514 u_int16_t *idp = (u_int16_t *)&nb_addr.s6_addr[2]; 1515 1516 if (*idp == 0) 1517 *idp = htons(ifp->if_index); 1518 } 1519 1520 mtx_lock(&nd6_mtx); 1521 if ((rt = nd6_lookup(&nb_addr, 0, ifp)) == NULL) { 1522 error = EINVAL; 1523 mtx_unlock(&nd6_mtx); 1524 break; 1525 } 1526 ln = (struct llinfo_nd6 *)rt->rt_llinfo; 1527 nbi->state = ln->ln_state; 1528 nbi->asked = ln->ln_asked; 1529 nbi->isrouter = ln->ln_router; 1530 nbi->expire = ln->ln_expire; 1531 mtx_unlock(&nd6_mtx); 1532 1533 break; 1534 } 1535 case SIOCGDEFIFACE_IN6: /* XXX: should be implemented as a sysctl? */ 1536 ndif->ifindex = nd6_defifindex; 1537 break; 1538 case SIOCSDEFIFACE_IN6: /* XXX: should be implemented as a sysctl? */ 1539 return (nd6_setdefaultiface(ndif->ifindex)); 1540 break; 1541 } 1542 return (error); 1543 } 1544 1545 /* 1546 * Create neighbor cache entry and cache link-layer address, 1547 * on reception of inbound ND6 packets. (RS/RA/NS/redirect) 1548 */ 1549 struct rtentry * 1550 nd6_cache_lladdr(struct ifnet *ifp, struct in6_addr *from, char *lladdr, 1551 int lladdrlen, 1552 int type, /* ICMP6 type */ 1553 int code /* type dependent information */) 1554 { 1555 struct rtentry *rt = NULL; 1556 struct llinfo_nd6 *ln = NULL; 1557 int is_newentry; 1558 struct sockaddr_dl *sdl = NULL; 1559 int do_update; 1560 int olladdr; 1561 int llchange; 1562 int newstate = 0; 1563 1564 if (!ifp) 1565 panic("ifp == NULL in nd6_cache_lladdr"); 1566 if (!from) 1567 panic("from == NULL in nd6_cache_lladdr"); 1568 1569 /* nothing must be updated for unspecified address */ 1570 if (IN6_IS_ADDR_UNSPECIFIED(from)) 1571 return NULL; 1572 1573 /* 1574 * Validation about ifp->if_addrlen and lladdrlen must be done in 1575 * the caller. 1576 * 1577 * XXX If the link does not have link-layer adderss, what should 1578 * we do? (ifp->if_addrlen == 0) 1579 * Spec says nothing in sections for RA, RS and NA. There's small 1580 * description on it in NS section (RFC 2461 7.2.3). 1581 */ 1582 1583 rt = nd6_lookup(from, 0, ifp); 1584 if (!rt) { 1585 #if 0 1586 /* nothing must be done if there's no lladdr */ 1587 if (!lladdr || !lladdrlen) 1588 return NULL; 1589 #endif 1590 1591 rt = nd6_lookup(from, 1, ifp); 1592 is_newentry = 1; 1593 } else { 1594 /* do nothing if static ndp is set */ 1595 if (rt->rt_flags & RTF_STATIC) 1596 return NULL; 1597 is_newentry = 0; 1598 } 1599 1600 if (!rt) 1601 return NULL; 1602 if ((rt->rt_flags & (RTF_GATEWAY | RTF_LLINFO)) != RTF_LLINFO) { 1603 fail: 1604 nd6_free(rt); 1605 return NULL; 1606 } 1607 ln = (struct llinfo_nd6 *)rt->rt_llinfo; 1608 if (!ln) 1609 goto fail; 1610 if (!rt->rt_gateway) 1611 goto fail; 1612 if (rt->rt_gateway->sa_family != AF_LINK) 1613 goto fail; 1614 sdl = SDL(rt->rt_gateway); 1615 1616 olladdr = (sdl->sdl_alen) ? 1 : 0; 1617 if (olladdr && lladdr) { 1618 if (bcmp(lladdr, LLADDR(sdl), ifp->if_addrlen)) 1619 llchange = 1; 1620 else 1621 llchange = 0; 1622 } else 1623 llchange = 0; 1624 1625 /* 1626 * newentry olladdr lladdr llchange (*=record) 1627 * 0 n n -- (1) 1628 * 0 y n -- (2) 1629 * 0 n y -- (3) * STALE 1630 * 0 y y n (4) * 1631 * 0 y y y (5) * STALE 1632 * 1 -- n -- (6) NOSTATE(= PASSIVE) 1633 * 1 -- y -- (7) * STALE 1634 */ 1635 1636 if (lladdr) { /* (3-5) and (7) */ 1637 /* 1638 * Record source link-layer address 1639 * XXX is it dependent to ifp->if_type? 1640 */ 1641 sdl->sdl_alen = ifp->if_addrlen; 1642 bcopy(lladdr, LLADDR(sdl), ifp->if_addrlen); 1643 } 1644 1645 if (!is_newentry) { 1646 if ((!olladdr && lladdr) /* (3) */ 1647 || (olladdr && lladdr && llchange)) { /* (5) */ 1648 do_update = 1; 1649 newstate = ND6_LLINFO_STALE; 1650 } else /* (1-2,4) */ 1651 do_update = 0; 1652 } else { 1653 do_update = 1; 1654 if (!lladdr) /* (6) */ 1655 newstate = ND6_LLINFO_NOSTATE; 1656 else /* (7) */ 1657 newstate = ND6_LLINFO_STALE; 1658 } 1659 1660 if (do_update) { 1661 /* 1662 * Update the state of the neighbor cache. 1663 */ 1664 ln->ln_state = newstate; 1665 1666 if (ln->ln_state == ND6_LLINFO_STALE) { 1667 /* 1668 * XXX: since nd6_output() below will cause 1669 * state tansition to DELAY and reset the timer, 1670 * we must set the timer now, although it is actually 1671 * meaningless. 1672 */ 1673 ln->ln_expire = time_second + nd6_gctimer; 1674 1675 if (ln->ln_hold) { 1676 /* 1677 * we assume ifp is not a p2p here, so just 1678 * set the 2nd argument as the 1st one. 1679 */ 1680 nd6_output(ifp, ifp, ln->ln_hold, 1681 (struct sockaddr_in6 *)rt_key(rt), 1682 rt); 1683 ln->ln_hold = NULL; 1684 } 1685 } else if (ln->ln_state == ND6_LLINFO_INCOMPLETE) { 1686 /* probe right away */ 1687 ln->ln_expire = time_second; 1688 } 1689 } 1690 1691 /* 1692 * ICMP6 type dependent behavior. 1693 * 1694 * NS: clear IsRouter if new entry 1695 * RS: clear IsRouter 1696 * RA: set IsRouter if there's lladdr 1697 * redir: clear IsRouter if new entry 1698 * 1699 * RA case, (1): 1700 * The spec says that we must set IsRouter in the following cases: 1701 * - If lladdr exist, set IsRouter. This means (1-5). 1702 * - If it is old entry (!newentry), set IsRouter. This means (7). 1703 * So, based on the spec, in (1-5) and (7) cases we must set IsRouter. 1704 * A quetion arises for (1) case. (1) case has no lladdr in the 1705 * neighbor cache, this is similar to (6). 1706 * This case is rare but we figured that we MUST NOT set IsRouter. 1707 * 1708 * newentry olladdr lladdr llchange NS RS RA redir 1709 * D R 1710 * 0 n n -- (1) c ? s 1711 * 0 y n -- (2) c s s 1712 * 0 n y -- (3) c s s 1713 * 0 y y n (4) c s s 1714 * 0 y y y (5) c s s 1715 * 1 -- n -- (6) c c c s 1716 * 1 -- y -- (7) c c s c s 1717 * 1718 * (c=clear s=set) 1719 */ 1720 switch (type & 0xff) { 1721 case ND_NEIGHBOR_SOLICIT: 1722 /* 1723 * New entry must have is_router flag cleared. 1724 */ 1725 if (is_newentry) /* (6-7) */ 1726 ln->ln_router = 0; 1727 break; 1728 case ND_REDIRECT: 1729 /* 1730 * If the icmp is a redirect to a better router, always set the 1731 * is_router flag. Otherwise, if the entry is newly created, 1732 * clear the flag. [RFC 2461, sec 8.3] 1733 */ 1734 if (code == ND_REDIRECT_ROUTER) 1735 ln->ln_router = 1; 1736 else if (is_newentry) /* (6-7) */ 1737 ln->ln_router = 0; 1738 break; 1739 case ND_ROUTER_SOLICIT: 1740 /* 1741 * is_router flag must always be cleared. 1742 */ 1743 ln->ln_router = 0; 1744 break; 1745 case ND_ROUTER_ADVERT: 1746 /* 1747 * Mark an entry with lladdr as a router. 1748 */ 1749 if ((!is_newentry && (olladdr || lladdr)) /* (2-5) */ 1750 || (is_newentry && lladdr)) { /* (7) */ 1751 ln->ln_router = 1; 1752 } 1753 break; 1754 } 1755 1756 /* 1757 * When the link-layer address of a router changes, select the 1758 * best router again. In particular, when the neighbor entry is newly 1759 * created, it might affect the selection policy. 1760 * Question: can we restrict the first condition to the "is_newentry" 1761 * case? 1762 * XXX: when we hear an RA from a new router with the link-layer 1763 * address option, defrouter_select() is called twice, since 1764 * defrtrlist_update called the function as well. However, I believe 1765 * we can compromise the overhead, since it only happens the first 1766 * time. 1767 * XXX: although defrouter_select() should not have a bad effect 1768 * for those are not autoconfigured hosts, we explicitly avoid such 1769 * cases for safety. 1770 */ 1771 if (do_update && ln->ln_router && !ip6_forwarding && ip6_accept_rtadv) 1772 defrouter_select(); 1773 1774 return rt; 1775 } 1776 1777 static void 1778 nd6_slowtimo(void *ignored_arg) 1779 { 1780 struct nd_ifinfo *nd6if; 1781 struct ifnet *ifp; 1782 1783 mtx_lock(&nd6_mtx); 1784 callout_reset(&nd6_slowtimo_ch, ND6_SLOWTIMER_INTERVAL * hz, 1785 nd6_slowtimo, NULL); 1786 for (ifp = TAILQ_FIRST(&ifnet); ifp; ifp = TAILQ_NEXT(ifp, if_list)) { 1787 nd6if = ND_IFINFO(ifp); 1788 if (nd6if->basereachable && /* already initialized */ 1789 (nd6if->recalctm -= ND6_SLOWTIMER_INTERVAL) <= 0) { 1790 /* 1791 * Since reachable time rarely changes by router 1792 * advertisements, we SHOULD insure that a new random 1793 * value gets recomputed at least once every few hours. 1794 * (RFC 2461, 6.3.4) 1795 */ 1796 nd6if->recalctm = nd6_recalc_reachtm_interval; 1797 nd6if->reachable = ND_COMPUTE_RTIME(nd6if->basereachable); 1798 } 1799 } 1800 mtx_unlock(&nd6_mtx); 1801 } 1802 1803 #define gotoerr(e) { error = (e); goto bad;} 1804 1805 int 1806 nd6_output(struct ifnet *ifp, struct ifnet *origifp, struct mbuf *m, 1807 struct sockaddr_in6 *dst, struct rtentry *rt) 1808 { 1809 struct llinfo_nd6 *ln = NULL; 1810 int error = 0; 1811 1812 if (IN6_IS_ADDR_MULTICAST(&dst->sin6_addr)) 1813 goto sendpkt; 1814 1815 if (nd6_need_cache(ifp) == 0) 1816 goto sendpkt; 1817 1818 /* 1819 * next hop determination. This routine is derived from ether_outpout. 1820 */ 1821 if (rt != NULL) { 1822 if (!(rt->rt_flags & RTF_UP)) { 1823 rt = rtlookup((struct sockaddr *)dst); 1824 if (rt == NULL) 1825 gotoerr(EHOSTUNREACH); 1826 rt->rt_refcnt--; 1827 if (rt->rt_ifp != ifp) { 1828 /* XXX: loop care? */ 1829 return nd6_output(ifp, origifp, m, dst, rt); 1830 } 1831 } 1832 if (rt->rt_flags & RTF_GATEWAY) { 1833 struct sockaddr_in6 *gw6; 1834 1835 /* 1836 * We skip link-layer address resolution and NUD 1837 * if the gateway is not a neighbor from ND point 1838 * of view, regardless of the value of nd_ifinfo.flags. 1839 * The second condition is a bit tricky; we skip 1840 * if the gateway is our own address, which is 1841 * sometimes used to install a route to a p2p link. 1842 */ 1843 gw6 = (struct sockaddr_in6 *)rt->rt_gateway; 1844 if (!nd6_is_addr_neighbor(gw6, ifp) || 1845 in6ifa_ifpwithaddr(ifp, &gw6->sin6_addr)) { 1846 /* 1847 * We allow this kind of tricky route only 1848 * when the outgoing interface is p2p. 1849 * XXX: we may need a more generic rule here. 1850 */ 1851 if (!(ifp->if_flags & IFF_POINTOPOINT)) 1852 gotoerr(EHOSTUNREACH); 1853 1854 goto sendpkt; 1855 } 1856 1857 if (rt->rt_gwroute == NULL) { 1858 rt->rt_gwroute = rtlookup(rt->rt_gateway); 1859 if (rt->rt_gwroute == NULL) 1860 gotoerr(EHOSTUNREACH); 1861 } else if (!(rt->rt_gwroute->rt_flags & RTF_UP)) { 1862 rtfree(rt->rt_gwroute); 1863 rt->rt_gwroute = rtlookup(rt->rt_gateway); 1864 if (rt->rt_gwroute == NULL) 1865 gotoerr(EHOSTUNREACH); 1866 } 1867 } 1868 } 1869 1870 /* 1871 * Address resolution or Neighbor Unreachability Detection 1872 * for the next hop. 1873 * At this point, the destination of the packet must be a unicast 1874 * or an anycast address(i.e. not a multicast). 1875 */ 1876 1877 /* Look up the neighbor cache for the nexthop */ 1878 if (rt && (rt->rt_flags & RTF_LLINFO)) 1879 ln = (struct llinfo_nd6 *)rt->rt_llinfo; 1880 else { 1881 /* 1882 * Since nd6_is_addr_neighbor() internally calls nd6_lookup(), 1883 * the condition below is not very efficient. But we believe 1884 * it is tolerable, because this should be a rare case. 1885 */ 1886 if (nd6_is_addr_neighbor(dst, ifp) && 1887 (rt = nd6_lookup(&dst->sin6_addr, 1, ifp)) != NULL) 1888 ln = (struct llinfo_nd6 *)rt->rt_llinfo; 1889 } 1890 if (!ln || !rt) { 1891 if (!(ifp->if_flags & IFF_POINTOPOINT) && 1892 !(ND_IFINFO(ifp)->flags & ND6_IFF_PERFORMNUD)) { 1893 log(LOG_DEBUG, 1894 "nd6_output: can't allocate llinfo for %s " 1895 "(ln=%p, rt=%p)\n", 1896 ip6_sprintf(&dst->sin6_addr), ln, rt); 1897 gotoerr(EIO); /* XXX: good error? */ 1898 } 1899 1900 goto sendpkt; /* send anyway */ 1901 } 1902 1903 /* We don't have to do link-layer address resolution on a p2p link. */ 1904 if ((ifp->if_flags & IFF_POINTOPOINT) && 1905 ln->ln_state < ND6_LLINFO_REACHABLE) { 1906 ln->ln_state = ND6_LLINFO_STALE; 1907 ln->ln_expire = time_second + nd6_gctimer; 1908 } 1909 1910 /* 1911 * The first time we send a packet to a neighbor whose entry is 1912 * STALE, we have to change the state to DELAY and a sets a timer to 1913 * expire in DELAY_FIRST_PROBE_TIME seconds to ensure do 1914 * neighbor unreachability detection on expiration. 1915 * (RFC 2461 7.3.3) 1916 */ 1917 if (ln->ln_state == ND6_LLINFO_STALE) { 1918 ln->ln_asked = 0; 1919 ln->ln_state = ND6_LLINFO_DELAY; 1920 ln->ln_expire = time_second + nd6_delay; 1921 } 1922 1923 /* 1924 * If the neighbor cache entry has a state other than INCOMPLETE 1925 * (i.e. its link-layer address is already resolved), just 1926 * send the packet. 1927 */ 1928 if (ln->ln_state > ND6_LLINFO_INCOMPLETE) 1929 goto sendpkt; 1930 1931 /* 1932 * There is a neighbor cache entry, but no ethernet address 1933 * response yet. Replace the held mbuf (if any) with this 1934 * latest one. 1935 * 1936 * This code conforms to the rate-limiting rule described in Section 1937 * 7.2.2 of RFC 2461, because the timer is set correctly after sending 1938 * an NS below. 1939 */ 1940 if (ln->ln_state == ND6_LLINFO_NOSTATE) 1941 ln->ln_state = ND6_LLINFO_INCOMPLETE; 1942 if (ln->ln_hold) 1943 m_freem(ln->ln_hold); 1944 ln->ln_hold = m; 1945 if (ln->ln_expire) { 1946 if (ln->ln_asked < nd6_mmaxtries && 1947 ln->ln_expire < time_second) { 1948 ln->ln_asked++; 1949 ln->ln_expire = time_second + 1950 ND_IFINFO(ifp)->retrans / 1000; 1951 nd6_ns_output(ifp, NULL, &dst->sin6_addr, ln, 0); 1952 } 1953 } 1954 return (0); 1955 1956 sendpkt: 1957 if (ifp->if_flags & IFF_LOOPBACK) 1958 error = ifp->if_output(origifp, m, (struct sockaddr *)dst, rt); 1959 else 1960 error = ifp->if_output(ifp, m, (struct sockaddr *)dst, rt); 1961 return (error); 1962 1963 bad: 1964 m_freem(m); 1965 return (error); 1966 } 1967 #undef gotoerr 1968 1969 int 1970 nd6_need_cache(struct ifnet *ifp) 1971 { 1972 /* 1973 * XXX: we currently do not make neighbor cache on any interface 1974 * other than Ethernet and GIF. 1975 * 1976 * RFC2893 says: 1977 * - unidirectional tunnels needs no ND 1978 */ 1979 switch (ifp->if_type) { 1980 case IFT_ETHER: 1981 case IFT_IEEE1394: 1982 #ifdef IFT_L2VLAN 1983 case IFT_L2VLAN: 1984 #endif 1985 #ifdef IFT_IEEE80211 1986 case IFT_IEEE80211: 1987 #endif 1988 #ifdef IFT_CARP 1989 case IFT_CARP: 1990 #endif 1991 case IFT_GIF: /* XXX need more cases? */ 1992 return (1); 1993 default: 1994 return (0); 1995 } 1996 } 1997 1998 int 1999 nd6_storelladdr(struct ifnet *ifp, struct rtentry *rt0, struct mbuf *m, 2000 struct sockaddr *dst, u_char *desten) 2001 { 2002 struct sockaddr_dl *sdl; 2003 struct rtentry *rt; 2004 2005 2006 if (m->m_flags & M_MCAST) { 2007 switch (ifp->if_type) { 2008 case IFT_ETHER: 2009 #ifdef IFT_L2VLAN 2010 case IFT_L2VLAN: 2011 #endif 2012 #ifdef IFT_IEEE80211 2013 case IFT_IEEE80211: 2014 #endif 2015 ETHER_MAP_IPV6_MULTICAST(&SIN6(dst)->sin6_addr, 2016 desten); 2017 return (1); 2018 case IFT_IEEE1394: 2019 bcopy(ifp->if_broadcastaddr, desten, ifp->if_addrlen); 2020 return (1); 2021 default: 2022 m_freem(m); 2023 return (0); 2024 } 2025 } 2026 if (rt0 == NULL) { 2027 /* this could happen, if we could not allocate memory */ 2028 m_freem(m); 2029 return (0); 2030 } 2031 if (rt_llroute(dst, rt0, &rt) != 0) { 2032 m_freem(m); 2033 return (0); 2034 } 2035 if (rt->rt_gateway->sa_family != AF_LINK) { 2036 kprintf("nd6_storelladdr: something odd happens\n"); 2037 m_freem(m); 2038 return (0); 2039 } 2040 sdl = SDL(rt->rt_gateway); 2041 if (sdl->sdl_alen == 0) { 2042 /* this should be impossible, but we bark here for debugging */ 2043 kprintf("nd6_storelladdr: sdl_alen == 0\n"); 2044 m_freem(m); 2045 return (0); 2046 } 2047 2048 bcopy(LLADDR(sdl), desten, sdl->sdl_alen); 2049 return (1); 2050 } 2051 2052 static int nd6_sysctl_drlist(SYSCTL_HANDLER_ARGS); 2053 static int nd6_sysctl_prlist(SYSCTL_HANDLER_ARGS); 2054 #ifdef SYSCTL_DECL 2055 SYSCTL_DECL(_net_inet6_icmp6); 2056 #endif 2057 SYSCTL_NODE(_net_inet6_icmp6, ICMPV6CTL_ND6_DRLIST, nd6_drlist, 2058 CTLFLAG_RD, nd6_sysctl_drlist, "List default routers"); 2059 SYSCTL_NODE(_net_inet6_icmp6, ICMPV6CTL_ND6_PRLIST, nd6_prlist, 2060 CTLFLAG_RD, nd6_sysctl_prlist, "List prefixes"); 2061 2062 static int 2063 nd6_sysctl_drlist(SYSCTL_HANDLER_ARGS) 2064 { 2065 int error; 2066 char buf[1024]; 2067 struct in6_defrouter *d, *de; 2068 struct nd_defrouter *dr; 2069 2070 if (req->newptr) 2071 return EPERM; 2072 error = 0; 2073 2074 for (dr = TAILQ_FIRST(&nd_defrouter); 2075 dr; 2076 dr = TAILQ_NEXT(dr, dr_entry)) { 2077 d = (struct in6_defrouter *)buf; 2078 de = (struct in6_defrouter *)(buf + sizeof(buf)); 2079 2080 if (d + 1 <= de) { 2081 bzero(d, sizeof(*d)); 2082 d->rtaddr.sin6_family = AF_INET6; 2083 d->rtaddr.sin6_len = sizeof(d->rtaddr); 2084 if (in6_recoverscope(&d->rtaddr, &dr->rtaddr, 2085 dr->ifp) != 0) 2086 log(LOG_ERR, 2087 "scope error in " 2088 "default router list (%s)\n", 2089 ip6_sprintf(&dr->rtaddr)); 2090 d->flags = dr->flags; 2091 d->rtlifetime = dr->rtlifetime; 2092 d->expire = dr->expire; 2093 d->if_index = dr->ifp->if_index; 2094 } else 2095 panic("buffer too short"); 2096 2097 error = SYSCTL_OUT(req, buf, sizeof(*d)); 2098 if (error) 2099 break; 2100 } 2101 return error; 2102 } 2103 2104 static int 2105 nd6_sysctl_prlist(SYSCTL_HANDLER_ARGS) 2106 { 2107 int error; 2108 char buf[1024]; 2109 struct in6_prefix *p, *pe; 2110 struct nd_prefix *pr; 2111 2112 if (req->newptr) 2113 return EPERM; 2114 error = 0; 2115 2116 for (pr = nd_prefix.lh_first; pr; pr = pr->ndpr_next) { 2117 u_short advrtrs; 2118 size_t advance; 2119 struct sockaddr_in6 *sin6, *s6; 2120 struct nd_pfxrouter *pfr; 2121 2122 p = (struct in6_prefix *)buf; 2123 pe = (struct in6_prefix *)(buf + sizeof(buf)); 2124 2125 if (p + 1 <= pe) { 2126 bzero(p, sizeof(*p)); 2127 sin6 = (struct sockaddr_in6 *)(p + 1); 2128 2129 p->prefix = pr->ndpr_prefix; 2130 if (in6_recoverscope(&p->prefix, 2131 &p->prefix.sin6_addr, pr->ndpr_ifp) != 0) 2132 log(LOG_ERR, 2133 "scope error in prefix list (%s)\n", 2134 ip6_sprintf(&p->prefix.sin6_addr)); 2135 p->raflags = pr->ndpr_raf; 2136 p->prefixlen = pr->ndpr_plen; 2137 p->vltime = pr->ndpr_vltime; 2138 p->pltime = pr->ndpr_pltime; 2139 p->if_index = pr->ndpr_ifp->if_index; 2140 p->expire = pr->ndpr_expire; 2141 p->refcnt = pr->ndpr_refcnt; 2142 p->flags = pr->ndpr_stateflags; 2143 p->origin = PR_ORIG_RA; 2144 advrtrs = 0; 2145 for (pfr = pr->ndpr_advrtrs.lh_first; 2146 pfr; 2147 pfr = pfr->pfr_next) { 2148 if ((void *)&sin6[advrtrs + 1] > 2149 (void *)pe) { 2150 advrtrs++; 2151 continue; 2152 } 2153 s6 = &sin6[advrtrs]; 2154 bzero(s6, sizeof(*s6)); 2155 s6->sin6_family = AF_INET6; 2156 s6->sin6_len = sizeof(*sin6); 2157 if (in6_recoverscope(s6, &pfr->router->rtaddr, 2158 pfr->router->ifp) != 0) 2159 log(LOG_ERR, 2160 "scope error in " 2161 "prefix list (%s)\n", 2162 ip6_sprintf(&pfr->router->rtaddr)); 2163 advrtrs++; 2164 } 2165 p->advrtrs = advrtrs; 2166 } else 2167 panic("buffer too short"); 2168 2169 advance = sizeof(*p) + sizeof(*sin6) * advrtrs; 2170 error = SYSCTL_OUT(req, buf, advance); 2171 if (error) 2172 break; 2173 } 2174 return error; 2175 } 2176