nd6.c revision 121092
1/* $FreeBSD: head/sys/netinet6/nd6.c 121092 2003-10-14 18:49:08Z sam $ */ 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/mac.h> 41#include <sys/malloc.h> 42#include <sys/mbuf.h> 43#include <sys/socket.h> 44#include <sys/sockio.h> 45#include <sys/time.h> 46#include <sys/kernel.h> 47#include <sys/protosw.h> 48#include <sys/errno.h> 49#include <sys/syslog.h> 50#include <sys/queue.h> 51#include <sys/sysctl.h> 52 53#include <net/if.h> 54#include <net/if_dl.h> 55#include <net/if_types.h> 56#include <net/if_atm.h> 57#include <net/iso88025.h> 58#include <net/fddi.h> 59#include <net/route.h> 60 61#include <netinet/in.h> 62#include <netinet/if_ether.h> 63#include <netinet6/in6_var.h> 64#include <netinet/ip6.h> 65#include <netinet6/ip6_var.h> 66#include <netinet6/nd6.h> 67#include <netinet6/in6_prefix.h> 68#include <netinet/icmp6.h> 69 70#include <net/net_osdep.h> 71 72#define ND6_SLOWTIMER_INTERVAL (60 * 60) /* 1 hour */ 73#define ND6_RECALC_REACHTM_INTERVAL (60 * 120) /* 2 hours */ 74 75#define SIN6(s) ((struct sockaddr_in6 *)s) 76#define SDL(s) ((struct sockaddr_dl *)s) 77 78/* timer values */ 79int nd6_prune = 1; /* walk list every 1 seconds */ 80int nd6_delay = 5; /* delay first probe time 5 second */ 81int nd6_umaxtries = 3; /* maximum unicast query */ 82int nd6_mmaxtries = 3; /* maximum multicast query */ 83int nd6_useloopback = 1; /* use loopback interface for local traffic */ 84int nd6_gctimer = (60 * 60 * 24); /* 1 day: garbage collection timer */ 85 86/* preventing too many loops in ND option parsing */ 87int nd6_maxndopt = 10; /* max # of ND options allowed */ 88 89int nd6_maxnudhint = 0; /* max # of subsequent upper layer hints */ 90 91#ifdef ND6_DEBUG 92int nd6_debug = 1; 93#else 94int nd6_debug = 0; 95#endif 96 97/* for debugging? */ 98static int nd6_inuse, nd6_allocated; 99 100struct llinfo_nd6 llinfo_nd6 = {&llinfo_nd6, &llinfo_nd6}; 101static size_t nd_ifinfo_indexlim = 8; 102struct nd_ifinfo *nd_ifinfo = NULL; 103struct nd_drhead nd_defrouter; 104struct nd_prhead nd_prefix = { 0 }; 105 106int nd6_recalc_reachtm_interval = ND6_RECALC_REACHTM_INTERVAL; 107static struct sockaddr_in6 all1_sa; 108 109static void nd6_slowtimo __P((void *)); 110static int regen_tmpaddr __P((struct in6_ifaddr *)); 111 112struct callout nd6_slowtimo_ch; 113struct callout nd6_timer_ch; 114extern struct callout in6_tmpaddrtimer_ch; 115 116void 117nd6_init() 118{ 119 static int nd6_init_done = 0; 120 int i; 121 122 if (nd6_init_done) { 123 log(LOG_NOTICE, "nd6_init called more than once(ignored)\n"); 124 return; 125 } 126 127 all1_sa.sin6_family = AF_INET6; 128 all1_sa.sin6_len = sizeof(struct sockaddr_in6); 129 for (i = 0; i < sizeof(all1_sa.sin6_addr); i++) 130 all1_sa.sin6_addr.s6_addr[i] = 0xff; 131 132 /* initialization of the default router list */ 133 TAILQ_INIT(&nd_defrouter); 134 135 nd6_init_done = 1; 136 137 /* start timer */ 138 callout_reset(&nd6_slowtimo_ch, ND6_SLOWTIMER_INTERVAL * hz, 139 nd6_slowtimo, NULL); 140} 141 142void 143nd6_ifattach(ifp) 144 struct ifnet *ifp; 145{ 146 147 /* 148 * We have some arrays that should be indexed by if_index. 149 * since if_index will grow dynamically, they should grow too. 150 */ 151 if (nd_ifinfo == NULL || if_index >= nd_ifinfo_indexlim) { 152 size_t n; 153 caddr_t q; 154 155 while (if_index >= nd_ifinfo_indexlim) 156 nd_ifinfo_indexlim <<= 1; 157 158 /* grow nd_ifinfo */ 159 n = nd_ifinfo_indexlim * sizeof(struct nd_ifinfo); 160 q = (caddr_t)malloc(n, M_IP6NDP, M_WAITOK); 161 bzero(q, n); 162 if (nd_ifinfo) { 163 bcopy((caddr_t)nd_ifinfo, q, n/2); 164 free((caddr_t)nd_ifinfo, M_IP6NDP); 165 } 166 nd_ifinfo = (struct nd_ifinfo *)q; 167 } 168 169#define ND nd_ifinfo[ifp->if_index] 170 171 /* 172 * Don't initialize if called twice. 173 * XXX: to detect this, we should choose a member that is never set 174 * before initialization of the ND structure itself. We formaly used 175 * the linkmtu member, which was not suitable because it could be 176 * initialized via "ifconfig mtu". 177 */ 178 if (ND.basereachable) 179 return; 180 181 ND.linkmtu = ifnet_byindex(ifp->if_index)->if_mtu; 182 ND.chlim = IPV6_DEFHLIM; 183 ND.basereachable = REACHABLE_TIME; 184 ND.reachable = ND_COMPUTE_RTIME(ND.basereachable); 185 ND.retrans = RETRANS_TIMER; 186 ND.receivedra = 0; 187 /* 188 * Note that the default value of ip6_accept_rtadv is 0, which means 189 * we won't accept RAs by default even if we set ND6_IFF_ACCEPT_RTADV 190 * here. 191 */ 192 ND.flags = (ND6_IFF_PERFORMNUD | ND6_IFF_ACCEPT_RTADV); 193 nd6_setmtu(ifp); 194#undef ND 195} 196 197/* 198 * Reset ND level link MTU. This function is called when the physical MTU 199 * changes, which means we might have to adjust the ND level MTU. 200 */ 201void 202nd6_setmtu(ifp) 203 struct ifnet *ifp; 204{ 205 struct nd_ifinfo *ndi = &nd_ifinfo[ifp->if_index]; 206 u_long oldmaxmtu = ndi->maxmtu; 207 u_long oldlinkmtu = ndi->linkmtu; 208 209 switch (ifp->if_type) { 210 case IFT_ARCNET: /* XXX MTU handling needs more work */ 211 ndi->maxmtu = MIN(60480, ifp->if_mtu); 212 break; 213 case IFT_ETHER: 214 ndi->maxmtu = MIN(ETHERMTU, ifp->if_mtu); 215 break; 216 case IFT_FDDI: 217 ndi->maxmtu = MIN(FDDIIPMTU, ifp->if_mtu); 218 break; 219 case IFT_ATM: 220 ndi->maxmtu = MIN(ATMMTU, ifp->if_mtu); 221 break; 222 case IFT_IEEE1394: /* XXX should be IEEE1394MTU(1500) */ 223 ndi->maxmtu = MIN(ETHERMTU, ifp->if_mtu); 224 break; 225#ifdef IFT_IEEE80211 226 case IFT_IEEE80211: /* XXX should be IEEE80211MTU(1500) */ 227 ndi->maxmtu = MIN(ETHERMTU, ifp->if_mtu); 228 break; 229#endif 230 case IFT_ISO88025: 231 ndi->maxmtu = MIN(ISO88025_MAX_MTU, ifp->if_mtu); 232 break; 233 default: 234 ndi->maxmtu = ifp->if_mtu; 235 break; 236 } 237 238 if (oldmaxmtu != ndi->maxmtu) { 239 /* 240 * If the ND level MTU is not set yet, or if the maxmtu 241 * is reset to a smaller value than the ND level MTU, 242 * also reset the ND level MTU. 243 */ 244 if (ndi->linkmtu == 0 || 245 ndi->maxmtu < ndi->linkmtu) { 246 ndi->linkmtu = ndi->maxmtu; 247 /* also adjust in6_maxmtu if necessary. */ 248 if (oldlinkmtu == 0) { 249 /* 250 * XXX: the case analysis is grotty, but 251 * it is not efficient to call in6_setmaxmtu() 252 * here when we are during the initialization 253 * procedure. 254 */ 255 if (in6_maxmtu < ndi->linkmtu) 256 in6_maxmtu = ndi->linkmtu; 257 } else 258 in6_setmaxmtu(); 259 } 260 } 261#undef MIN 262} 263 264void 265nd6_option_init(opt, icmp6len, ndopts) 266 void *opt; 267 int icmp6len; 268 union nd_opts *ndopts; 269{ 270 271 bzero(ndopts, sizeof(*ndopts)); 272 ndopts->nd_opts_search = (struct nd_opt_hdr *)opt; 273 ndopts->nd_opts_last 274 = (struct nd_opt_hdr *)(((u_char *)opt) + icmp6len); 275 276 if (icmp6len == 0) { 277 ndopts->nd_opts_done = 1; 278 ndopts->nd_opts_search = NULL; 279 } 280} 281 282/* 283 * Take one ND option. 284 */ 285struct nd_opt_hdr * 286nd6_option(ndopts) 287 union nd_opts *ndopts; 288{ 289 struct nd_opt_hdr *nd_opt; 290 int olen; 291 292 if (!ndopts) 293 panic("ndopts == NULL in nd6_option"); 294 if (!ndopts->nd_opts_last) 295 panic("uninitialized ndopts in nd6_option"); 296 if (!ndopts->nd_opts_search) 297 return NULL; 298 if (ndopts->nd_opts_done) 299 return NULL; 300 301 nd_opt = ndopts->nd_opts_search; 302 303 /* make sure nd_opt_len is inside the buffer */ 304 if ((caddr_t)&nd_opt->nd_opt_len >= (caddr_t)ndopts->nd_opts_last) { 305 bzero(ndopts, sizeof(*ndopts)); 306 return NULL; 307 } 308 309 olen = nd_opt->nd_opt_len << 3; 310 if (olen == 0) { 311 /* 312 * Message validation requires that all included 313 * options have a length that is greater than zero. 314 */ 315 bzero(ndopts, sizeof(*ndopts)); 316 return NULL; 317 } 318 319 ndopts->nd_opts_search = (struct nd_opt_hdr *)((caddr_t)nd_opt + olen); 320 if (ndopts->nd_opts_search > ndopts->nd_opts_last) { 321 /* option overruns the end of buffer, invalid */ 322 bzero(ndopts, sizeof(*ndopts)); 323 return NULL; 324 } else if (ndopts->nd_opts_search == ndopts->nd_opts_last) { 325 /* reached the end of options chain */ 326 ndopts->nd_opts_done = 1; 327 ndopts->nd_opts_search = NULL; 328 } 329 return nd_opt; 330} 331 332/* 333 * Parse multiple ND options. 334 * This function is much easier to use, for ND routines that do not need 335 * multiple options of the same type. 336 */ 337int 338nd6_options(ndopts) 339 union nd_opts *ndopts; 340{ 341 struct nd_opt_hdr *nd_opt; 342 int i = 0; 343 344 if (!ndopts) 345 panic("ndopts == NULL in nd6_options"); 346 if (!ndopts->nd_opts_last) 347 panic("uninitialized ndopts in nd6_options"); 348 if (!ndopts->nd_opts_search) 349 return 0; 350 351 while (1) { 352 nd_opt = nd6_option(ndopts); 353 if (!nd_opt && !ndopts->nd_opts_last) { 354 /* 355 * Message validation requires that all included 356 * options have a length that is greater than zero. 357 */ 358 icmp6stat.icp6s_nd_badopt++; 359 bzero(ndopts, sizeof(*ndopts)); 360 return -1; 361 } 362 363 if (!nd_opt) 364 goto skip1; 365 366 switch (nd_opt->nd_opt_type) { 367 case ND_OPT_SOURCE_LINKADDR: 368 case ND_OPT_TARGET_LINKADDR: 369 case ND_OPT_MTU: 370 case ND_OPT_REDIRECTED_HEADER: 371 if (ndopts->nd_opt_array[nd_opt->nd_opt_type]) { 372 nd6log((LOG_INFO, 373 "duplicated ND6 option found (type=%d)\n", 374 nd_opt->nd_opt_type)); 375 /* XXX bark? */ 376 } else { 377 ndopts->nd_opt_array[nd_opt->nd_opt_type] 378 = nd_opt; 379 } 380 break; 381 case ND_OPT_PREFIX_INFORMATION: 382 if (ndopts->nd_opt_array[nd_opt->nd_opt_type] == 0) { 383 ndopts->nd_opt_array[nd_opt->nd_opt_type] 384 = nd_opt; 385 } 386 ndopts->nd_opts_pi_end = 387 (struct nd_opt_prefix_info *)nd_opt; 388 break; 389 default: 390 /* 391 * Unknown options must be silently ignored, 392 * to accomodate future extension to the protocol. 393 */ 394 nd6log((LOG_DEBUG, 395 "nd6_options: unsupported option %d - " 396 "option ignored\n", nd_opt->nd_opt_type)); 397 } 398 399skip1: 400 i++; 401 if (i > nd6_maxndopt) { 402 icmp6stat.icp6s_nd_toomanyopt++; 403 nd6log((LOG_INFO, "too many loop in nd opt\n")); 404 break; 405 } 406 407 if (ndopts->nd_opts_done) 408 break; 409 } 410 411 return 0; 412} 413 414/* 415 * ND6 timer routine to expire default route list and prefix list 416 */ 417void 418nd6_timer(ignored_arg) 419 void *ignored_arg; 420{ 421 int s; 422 struct llinfo_nd6 *ln; 423 struct nd_defrouter *dr; 424 struct nd_prefix *pr; 425 struct ifnet *ifp; 426 struct in6_ifaddr *ia6, *nia6; 427 struct in6_addrlifetime *lt6; 428 429 s = splnet(); 430 callout_reset(&nd6_timer_ch, nd6_prune * hz, 431 nd6_timer, NULL); 432 433 ln = llinfo_nd6.ln_next; 434 while (ln && ln != &llinfo_nd6) { 435 struct rtentry *rt; 436 struct sockaddr_in6 *dst; 437 struct llinfo_nd6 *next = ln->ln_next; 438 /* XXX: used for the DELAY case only: */ 439 struct nd_ifinfo *ndi = NULL; 440 441 if ((rt = ln->ln_rt) == NULL) { 442 ln = next; 443 continue; 444 } 445 if ((ifp = rt->rt_ifp) == NULL) { 446 ln = next; 447 continue; 448 } 449 ndi = &nd_ifinfo[ifp->if_index]; 450 dst = (struct sockaddr_in6 *)rt_key(rt); 451 452 if (ln->ln_expire > time_second) { 453 ln = next; 454 continue; 455 } 456 457 /* sanity check */ 458 if (!rt) 459 panic("rt=0 in nd6_timer(ln=%p)", ln); 460 if (rt->rt_llinfo && (struct llinfo_nd6 *)rt->rt_llinfo != ln) 461 panic("rt_llinfo(%p) is not equal to ln(%p)", 462 rt->rt_llinfo, ln); 463 if (!dst) 464 panic("dst=0 in nd6_timer(ln=%p)", ln); 465 466 switch (ln->ln_state) { 467 case ND6_LLINFO_INCOMPLETE: 468 if (ln->ln_asked < nd6_mmaxtries) { 469 ln->ln_asked++; 470 ln->ln_expire = time_second + 471 nd_ifinfo[ifp->if_index].retrans / 1000; 472 nd6_ns_output(ifp, NULL, &dst->sin6_addr, 473 ln, 0); 474 } else { 475 struct mbuf *m = ln->ln_hold; 476 if (m) { 477 if (rt->rt_ifp) { 478 /* 479 * Fake rcvif to make ICMP error 480 * more helpful in diagnosing 481 * for the receiver. 482 * XXX: should we consider 483 * older rcvif? 484 */ 485 m->m_pkthdr.rcvif = rt->rt_ifp; 486 } 487 icmp6_error(m, ICMP6_DST_UNREACH, 488 ICMP6_DST_UNREACH_ADDR, 0); 489 ln->ln_hold = NULL; 490 } 491 next = nd6_free(rt); 492 } 493 break; 494 case ND6_LLINFO_REACHABLE: 495 if (ln->ln_expire) { 496 ln->ln_state = ND6_LLINFO_STALE; 497 ln->ln_expire = time_second + nd6_gctimer; 498 } 499 break; 500 501 case ND6_LLINFO_STALE: 502 /* Garbage Collection(RFC 2461 5.3) */ 503 if (ln->ln_expire) 504 next = nd6_free(rt); 505 break; 506 507 case ND6_LLINFO_DELAY: 508 if (ndi && (ndi->flags & ND6_IFF_PERFORMNUD) != 0) { 509 /* We need NUD */ 510 ln->ln_asked = 1; 511 ln->ln_state = ND6_LLINFO_PROBE; 512 ln->ln_expire = time_second + 513 ndi->retrans / 1000; 514 nd6_ns_output(ifp, &dst->sin6_addr, 515 &dst->sin6_addr, 516 ln, 0); 517 } else { 518 ln->ln_state = ND6_LLINFO_STALE; /* XXX */ 519 ln->ln_expire = time_second + nd6_gctimer; 520 } 521 break; 522 case ND6_LLINFO_PROBE: 523 if (ln->ln_asked < nd6_umaxtries) { 524 ln->ln_asked++; 525 ln->ln_expire = time_second + 526 nd_ifinfo[ifp->if_index].retrans / 1000; 527 nd6_ns_output(ifp, &dst->sin6_addr, 528 &dst->sin6_addr, ln, 0); 529 } else { 530 next = nd6_free(rt); 531 } 532 break; 533 } 534 ln = next; 535 } 536 537 /* expire default router list */ 538 dr = TAILQ_FIRST(&nd_defrouter); 539 while (dr) { 540 if (dr->expire && dr->expire < time_second) { 541 struct nd_defrouter *t; 542 t = TAILQ_NEXT(dr, dr_entry); 543 defrtrlist_del(dr); 544 dr = t; 545 } else { 546 dr = TAILQ_NEXT(dr, dr_entry); 547 } 548 } 549 550 /* 551 * expire interface addresses. 552 * in the past the loop was inside prefix expiry processing. 553 * However, from a stricter speci-confrmance standpoint, we should 554 * rather separate address lifetimes and prefix lifetimes. 555 */ 556 addrloop: 557 for (ia6 = in6_ifaddr; ia6; ia6 = nia6) { 558 nia6 = ia6->ia_next; 559 /* check address lifetime */ 560 lt6 = &ia6->ia6_lifetime; 561 if (IFA6_IS_INVALID(ia6)) { 562 int regen = 0; 563 564 /* 565 * If the expiring address is temporary, try 566 * regenerating a new one. This would be useful when 567 * we suspended a laptop PC, then turned it on after a 568 * period that could invalidate all temporary 569 * addresses. Although we may have to restart the 570 * loop (see below), it must be after purging the 571 * address. Otherwise, we'd see an infinite loop of 572 * regeneration. 573 */ 574 if (ip6_use_tempaddr && 575 (ia6->ia6_flags & IN6_IFF_TEMPORARY) != 0) { 576 if (regen_tmpaddr(ia6) == 0) 577 regen = 1; 578 } 579 580 in6_purgeaddr(&ia6->ia_ifa); 581 582 if (regen) 583 goto addrloop; /* XXX: see below */ 584 } 585 if (IFA6_IS_DEPRECATED(ia6)) { 586 int oldflags = ia6->ia6_flags; 587 588 ia6->ia6_flags |= IN6_IFF_DEPRECATED; 589 590 /* 591 * If a temporary address has just become deprecated, 592 * regenerate a new one if possible. 593 */ 594 if (ip6_use_tempaddr && 595 (ia6->ia6_flags & IN6_IFF_TEMPORARY) != 0 && 596 (oldflags & IN6_IFF_DEPRECATED) == 0) { 597 598 if (regen_tmpaddr(ia6) == 0) { 599 /* 600 * A new temporary address is 601 * generated. 602 * XXX: this means the address chain 603 * has changed while we are still in 604 * the loop. Although the change 605 * would not cause disaster (because 606 * it's not a deletion, but an 607 * addition,) we'd rather restart the 608 * loop just for safety. Or does this 609 * significantly reduce performance?? 610 */ 611 goto addrloop; 612 } 613 } 614 } else { 615 /* 616 * A new RA might have made a deprecated address 617 * preferred. 618 */ 619 ia6->ia6_flags &= ~IN6_IFF_DEPRECATED; 620 } 621 } 622 623 /* expire prefix list */ 624 pr = nd_prefix.lh_first; 625 while (pr) { 626 /* 627 * check prefix lifetime. 628 * since pltime is just for autoconf, pltime processing for 629 * prefix is not necessary. 630 */ 631 if (pr->ndpr_expire && pr->ndpr_expire < time_second) { 632 struct nd_prefix *t; 633 t = pr->ndpr_next; 634 635 /* 636 * address expiration and prefix expiration are 637 * separate. NEVER perform in6_purgeaddr here. 638 */ 639 640 prelist_remove(pr); 641 pr = t; 642 } else 643 pr = pr->ndpr_next; 644 } 645 splx(s); 646} 647 648static int 649regen_tmpaddr(ia6) 650 struct in6_ifaddr *ia6; /* deprecated/invalidated temporary address */ 651{ 652 struct ifaddr *ifa; 653 struct ifnet *ifp; 654 struct in6_ifaddr *public_ifa6 = NULL; 655 656 ifp = ia6->ia_ifa.ifa_ifp; 657 for (ifa = ifp->if_addrlist.tqh_first; ifa; 658 ifa = ifa->ifa_list.tqe_next) { 659 struct in6_ifaddr *it6; 660 661 if (ifa->ifa_addr->sa_family != AF_INET6) 662 continue; 663 664 it6 = (struct in6_ifaddr *)ifa; 665 666 /* ignore no autoconf addresses. */ 667 if ((it6->ia6_flags & IN6_IFF_AUTOCONF) == 0) 668 continue; 669 670 /* ignore autoconf addresses with different prefixes. */ 671 if (it6->ia6_ndpr == NULL || it6->ia6_ndpr != ia6->ia6_ndpr) 672 continue; 673 674 /* 675 * Now we are looking at an autoconf address with the same 676 * prefix as ours. If the address is temporary and is still 677 * preferred, do not create another one. It would be rare, but 678 * could happen, for example, when we resume a laptop PC after 679 * a long period. 680 */ 681 if ((it6->ia6_flags & IN6_IFF_TEMPORARY) != 0 && 682 !IFA6_IS_DEPRECATED(it6)) { 683 public_ifa6 = NULL; 684 break; 685 } 686 687 /* 688 * This is a public autoconf address that has the same prefix 689 * as ours. If it is preferred, keep it. We can't break the 690 * loop here, because there may be a still-preferred temporary 691 * address with the prefix. 692 */ 693 if (!IFA6_IS_DEPRECATED(it6)) 694 public_ifa6 = it6; 695 } 696 697 if (public_ifa6 != NULL) { 698 int e; 699 700 if ((e = in6_tmpifadd(public_ifa6, 0)) != 0) { 701 log(LOG_NOTICE, "regen_tmpaddr: failed to create a new" 702 " tmp addr,errno=%d\n", e); 703 return (-1); 704 } 705 return (0); 706 } 707 708 return (-1); 709} 710 711/* 712 * Nuke neighbor cache/prefix/default router management table, right before 713 * ifp goes away. 714 */ 715void 716nd6_purge(ifp) 717 struct ifnet *ifp; 718{ 719 struct llinfo_nd6 *ln, *nln; 720 struct nd_defrouter *dr, *ndr, drany; 721 struct nd_prefix *pr, *npr; 722 723 /* Nuke default router list entries toward ifp */ 724 if ((dr = TAILQ_FIRST(&nd_defrouter)) != NULL) { 725 /* 726 * The first entry of the list may be stored in 727 * the routing table, so we'll delete it later. 728 */ 729 for (dr = TAILQ_NEXT(dr, dr_entry); dr; dr = ndr) { 730 ndr = TAILQ_NEXT(dr, dr_entry); 731 if (dr->ifp == ifp) 732 defrtrlist_del(dr); 733 } 734 dr = TAILQ_FIRST(&nd_defrouter); 735 if (dr->ifp == ifp) 736 defrtrlist_del(dr); 737 } 738 739 /* Nuke prefix list entries toward ifp */ 740 for (pr = nd_prefix.lh_first; pr; pr = npr) { 741 npr = pr->ndpr_next; 742 if (pr->ndpr_ifp == ifp) { 743 /* 744 * Previously, pr->ndpr_addr is removed as well, 745 * but I strongly believe we don't have to do it. 746 * nd6_purge() is only called from in6_ifdetach(), 747 * which removes all the associated interface addresses 748 * by itself. 749 * (jinmei@kame.net 20010129) 750 */ 751 prelist_remove(pr); 752 } 753 } 754 755 /* cancel default outgoing interface setting */ 756 if (nd6_defifindex == ifp->if_index) 757 nd6_setdefaultiface(0); 758 759 if (!ip6_forwarding && ip6_accept_rtadv) { /* XXX: too restrictive? */ 760 /* refresh default router list */ 761 bzero(&drany, sizeof(drany)); 762 defrouter_delreq(&drany, 0); 763 defrouter_select(); 764 } 765 766 /* 767 * Nuke neighbor cache entries for the ifp. 768 * Note that rt->rt_ifp may not be the same as ifp, 769 * due to KAME goto ours hack. See RTM_RESOLVE case in 770 * nd6_rtrequest(), and ip6_input(). 771 */ 772 ln = llinfo_nd6.ln_next; 773 while (ln && ln != &llinfo_nd6) { 774 struct rtentry *rt; 775 struct sockaddr_dl *sdl; 776 777 nln = ln->ln_next; 778 rt = ln->ln_rt; 779 if (rt && rt->rt_gateway && 780 rt->rt_gateway->sa_family == AF_LINK) { 781 sdl = (struct sockaddr_dl *)rt->rt_gateway; 782 if (sdl->sdl_index == ifp->if_index) 783 nln = nd6_free(rt); 784 } 785 ln = nln; 786 } 787} 788 789struct rtentry * 790nd6_lookup(addr6, create, ifp) 791 struct in6_addr *addr6; 792 int create; 793 struct ifnet *ifp; 794{ 795 struct rtentry *rt; 796 struct sockaddr_in6 sin6; 797 798 bzero(&sin6, sizeof(sin6)); 799 sin6.sin6_len = sizeof(struct sockaddr_in6); 800 sin6.sin6_family = AF_INET6; 801 sin6.sin6_addr = *addr6; 802 rt = rtalloc1((struct sockaddr *)&sin6, create, 0UL); 803 if (rt) { 804 if ((rt->rt_flags & RTF_LLINFO) == 0 && create) { 805 /* 806 * This is the case for the default route. 807 * If we want to create a neighbor cache for the 808 * address, we should free the route for the 809 * destination and allocate an interface route. 810 */ 811 RTFREE_LOCKED(rt); 812 rt = 0; 813 } 814 } 815 if (!rt) { 816 if (create && ifp) { 817 int e; 818 819 /* 820 * If no route is available and create is set, 821 * we allocate a host route for the destination 822 * and treat it like an interface route. 823 * This hack is necessary for a neighbor which can't 824 * be covered by our own prefix. 825 */ 826 struct ifaddr *ifa = 827 ifaof_ifpforaddr((struct sockaddr *)&sin6, ifp); 828 if (ifa == NULL) 829 return (NULL); 830 831 /* 832 * Create a new route. RTF_LLINFO is necessary 833 * to create a Neighbor Cache entry for the 834 * destination in nd6_rtrequest which will be 835 * called in rtrequest via ifa->ifa_rtrequest. 836 */ 837 if ((e = rtrequest(RTM_ADD, (struct sockaddr *)&sin6, 838 ifa->ifa_addr, (struct sockaddr *)&all1_sa, 839 (ifa->ifa_flags | RTF_HOST | RTF_LLINFO) & 840 ~RTF_CLONING, &rt)) != 0) { 841 log(LOG_ERR, 842 "nd6_lookup: failed to add route for a " 843 "neighbor(%s), errno=%d\n", 844 ip6_sprintf(addr6), e); 845 } 846 if (rt == NULL) 847 return (NULL); 848 RT_LOCK(rt); 849 if (rt->rt_llinfo) { 850 struct llinfo_nd6 *ln = 851 (struct llinfo_nd6 *)rt->rt_llinfo; 852 ln->ln_state = ND6_LLINFO_NOSTATE; 853 } 854 } else 855 return (NULL); 856 } 857 RT_LOCK_ASSERT(rt); 858 rt->rt_refcnt--; 859 /* 860 * Validation for the entry. 861 * Note that the check for rt_llinfo is necessary because a cloned 862 * route from a parent route that has the L flag (e.g. the default 863 * route to a p2p interface) may have the flag, too, while the 864 * destination is not actually a neighbor. 865 * XXX: we can't use rt->rt_ifp to check for the interface, since 866 * it might be the loopback interface if the entry is for our 867 * own address on a non-loopback interface. Instead, we should 868 * use rt->rt_ifa->ifa_ifp, which would specify the REAL 869 * interface. 870 */ 871 if ((rt->rt_flags & RTF_GATEWAY) || (rt->rt_flags & RTF_LLINFO) == 0 || 872 rt->rt_gateway->sa_family != AF_LINK || rt->rt_llinfo == NULL || 873 (ifp && rt->rt_ifa->ifa_ifp != ifp)) { 874 if (create) { 875 log(LOG_DEBUG, 876 "nd6_lookup: failed to lookup %s (if = %s)\n", 877 ip6_sprintf(addr6), 878 ifp ? if_name(ifp) : "unspec"); 879 /* xxx more logs... kazu */ 880 } 881 RT_UNLOCK(rt); 882 return (NULL); 883 } 884 RT_UNLOCK(rt); /* XXX not ready to return rt locked */ 885 return (rt); 886} 887 888/* 889 * Detect if a given IPv6 address identifies a neighbor on a given link. 890 * XXX: should take care of the destination of a p2p link? 891 */ 892int 893nd6_is_addr_neighbor(addr, ifp) 894 struct sockaddr_in6 *addr; 895 struct ifnet *ifp; 896{ 897 struct ifaddr *ifa; 898 int i; 899 900#define IFADDR6(a) ((((struct in6_ifaddr *)(a))->ia_addr).sin6_addr) 901#define IFMASK6(a) ((((struct in6_ifaddr *)(a))->ia_prefixmask).sin6_addr) 902 903 /* 904 * A link-local address is always a neighbor. 905 * XXX: we should use the sin6_scope_id field rather than the embedded 906 * interface index. 907 */ 908 if (IN6_IS_ADDR_LINKLOCAL(&addr->sin6_addr) && 909 ntohs(*(u_int16_t *)&addr->sin6_addr.s6_addr[2]) == ifp->if_index) 910 return (1); 911 912 /* 913 * If the address matches one of our addresses, 914 * it should be a neighbor. 915 */ 916 for (ifa = ifp->if_addrlist.tqh_first; ifa; 917 ifa = ifa->ifa_list.tqe_next) { 918 if (ifa->ifa_addr->sa_family != AF_INET6) 919 next: continue; 920 921 for (i = 0; i < 4; i++) { 922 if ((IFADDR6(ifa).s6_addr32[i] ^ 923 addr->sin6_addr.s6_addr32[i]) & 924 IFMASK6(ifa).s6_addr32[i]) 925 goto next; 926 } 927 return (1); 928 } 929 930 /* 931 * Even if the address matches none of our addresses, it might be 932 * in the neighbor cache. 933 */ 934 if (nd6_lookup(&addr->sin6_addr, 0, ifp) != NULL) 935 return (1); 936 937 return (0); 938#undef IFADDR6 939#undef IFMASK6 940} 941 942/* 943 * Free an nd6 llinfo entry. 944 */ 945struct llinfo_nd6 * 946nd6_free(rt) 947 struct rtentry *rt; 948{ 949 struct llinfo_nd6 *ln = (struct llinfo_nd6 *)rt->rt_llinfo, *next; 950 struct in6_addr in6 = ((struct sockaddr_in6 *)rt_key(rt))->sin6_addr; 951 struct nd_defrouter *dr; 952 953 /* 954 * we used to have pfctlinput(PRC_HOSTDEAD) here. 955 * even though it is not harmful, it was not really necessary. 956 */ 957 958 if (!ip6_forwarding && ip6_accept_rtadv) { /* XXX: too restrictive? */ 959 int s; 960 s = splnet(); 961 dr = defrouter_lookup(&((struct sockaddr_in6 *)rt_key(rt))->sin6_addr, 962 rt->rt_ifp); 963 964 if (ln->ln_router || dr) { 965 /* 966 * rt6_flush must be called whether or not the neighbor 967 * is in the Default Router List. 968 * See a corresponding comment in nd6_na_input(). 969 */ 970 rt6_flush(&in6, rt->rt_ifp); 971 } 972 973 if (dr) { 974 /* 975 * Unreachablity of a router might affect the default 976 * router selection and on-link detection of advertised 977 * prefixes. 978 */ 979 980 /* 981 * Temporarily fake the state to choose a new default 982 * router and to perform on-link determination of 983 * prefixes correctly. 984 * Below the state will be set correctly, 985 * or the entry itself will be deleted. 986 */ 987 ln->ln_state = ND6_LLINFO_INCOMPLETE; 988 989 /* 990 * Since defrouter_select() does not affect the 991 * on-link determination and MIP6 needs the check 992 * before the default router selection, we perform 993 * the check now. 994 */ 995 pfxlist_onlink_check(); 996 997 if (dr == TAILQ_FIRST(&nd_defrouter)) { 998 /* 999 * It is used as the current default router, 1000 * so we have to move it to the end of the 1001 * list and choose a new one. 1002 * XXX: it is not very efficient if this is 1003 * the only router. 1004 */ 1005 TAILQ_REMOVE(&nd_defrouter, dr, dr_entry); 1006 TAILQ_INSERT_TAIL(&nd_defrouter, dr, dr_entry); 1007 1008 defrouter_select(); 1009 } 1010 } 1011 splx(s); 1012 } 1013 1014 /* 1015 * Before deleting the entry, remember the next entry as the 1016 * return value. We need this because pfxlist_onlink_check() above 1017 * might have freed other entries (particularly the old next entry) as 1018 * a side effect (XXX). 1019 */ 1020 next = ln->ln_next; 1021 1022 /* 1023 * Detach the route from the routing tree and the list of neighbor 1024 * caches, and disable the route entry not to be used in already 1025 * cached routes. 1026 */ 1027 rtrequest(RTM_DELETE, rt_key(rt), (struct sockaddr *)0, 1028 rt_mask(rt), 0, (struct rtentry **)0); 1029 1030 return (next); 1031} 1032 1033/* 1034 * Upper-layer reachability hint for Neighbor Unreachability Detection. 1035 * 1036 * XXX cost-effective metods? 1037 */ 1038void 1039nd6_nud_hint(rt, dst6, force) 1040 struct rtentry *rt; 1041 struct in6_addr *dst6; 1042 int force; 1043{ 1044 struct llinfo_nd6 *ln; 1045 1046 /* 1047 * If the caller specified "rt", use that. Otherwise, resolve the 1048 * routing table by supplied "dst6". 1049 */ 1050 if (!rt) { 1051 if (!dst6) 1052 return; 1053 if (!(rt = nd6_lookup(dst6, 0, NULL))) 1054 return; 1055 } 1056 1057 if ((rt->rt_flags & RTF_GATEWAY) != 0 || 1058 (rt->rt_flags & RTF_LLINFO) == 0 || 1059 !rt->rt_llinfo || !rt->rt_gateway || 1060 rt->rt_gateway->sa_family != AF_LINK) { 1061 /* This is not a host route. */ 1062 return; 1063 } 1064 1065 ln = (struct llinfo_nd6 *)rt->rt_llinfo; 1066 if (ln->ln_state < ND6_LLINFO_REACHABLE) 1067 return; 1068 1069 /* 1070 * if we get upper-layer reachability confirmation many times, 1071 * it is possible we have false information. 1072 */ 1073 if (!force) { 1074 ln->ln_byhint++; 1075 if (ln->ln_byhint > nd6_maxnudhint) 1076 return; 1077 } 1078 1079 ln->ln_state = ND6_LLINFO_REACHABLE; 1080 if (ln->ln_expire) 1081 ln->ln_expire = time_second + 1082 nd_ifinfo[rt->rt_ifp->if_index].reachable; 1083} 1084 1085void 1086nd6_rtrequest(req, rt, info) 1087 int req; 1088 struct rtentry *rt; 1089 struct rt_addrinfo *info; /* xxx unused */ 1090{ 1091 struct sockaddr *gate = rt->rt_gateway; 1092 struct llinfo_nd6 *ln = (struct llinfo_nd6 *)rt->rt_llinfo; 1093 static struct sockaddr_dl null_sdl = {sizeof(null_sdl), AF_LINK}; 1094 struct ifnet *ifp = rt->rt_ifp; 1095 struct ifaddr *ifa; 1096 1097 RT_LOCK_ASSERT(rt); 1098 1099 if ((rt->rt_flags & RTF_GATEWAY) != 0) 1100 return; 1101 1102 if (nd6_need_cache(ifp) == 0 && (rt->rt_flags & RTF_HOST) == 0) { 1103 /* 1104 * This is probably an interface direct route for a link 1105 * which does not need neighbor caches (e.g. fe80::%lo0/64). 1106 * We do not need special treatment below for such a route. 1107 * Moreover, the RTF_LLINFO flag which would be set below 1108 * would annoy the ndp(8) command. 1109 */ 1110 return; 1111 } 1112 1113 if (req == RTM_RESOLVE && 1114 (nd6_need_cache(ifp) == 0 || /* stf case */ 1115 !nd6_is_addr_neighbor((struct sockaddr_in6 *)rt_key(rt), ifp))) { 1116 /* 1117 * FreeBSD and BSD/OS often make a cloned host route based 1118 * on a less-specific route (e.g. the default route). 1119 * If the less specific route does not have a "gateway" 1120 * (this is the case when the route just goes to a p2p or an 1121 * stf interface), we'll mistakenly make a neighbor cache for 1122 * the host route, and will see strange neighbor solicitation 1123 * for the corresponding destination. In order to avoid the 1124 * confusion, we check if the destination of the route is 1125 * a neighbor in terms of neighbor discovery, and stop the 1126 * process if not. Additionally, we remove the LLINFO flag 1127 * so that ndp(8) will not try to get the neighbor information 1128 * of the destination. 1129 */ 1130 rt->rt_flags &= ~RTF_LLINFO; 1131 return; 1132 } 1133 1134 switch (req) { 1135 case RTM_ADD: 1136 /* 1137 * There is no backward compatibility :) 1138 * 1139 * if ((rt->rt_flags & RTF_HOST) == 0 && 1140 * SIN(rt_mask(rt))->sin_addr.s_addr != 0xffffffff) 1141 * rt->rt_flags |= RTF_CLONING; 1142 */ 1143 if (rt->rt_flags & (RTF_CLONING | RTF_LLINFO)) { 1144 /* 1145 * Case 1: This route should come from 1146 * a route to interface. RTF_LLINFO flag is set 1147 * for a host route whose destination should be 1148 * treated as on-link. 1149 */ 1150 rt_setgate(rt, rt_key(rt), 1151 (struct sockaddr *)&null_sdl); 1152 gate = rt->rt_gateway; 1153 SDL(gate)->sdl_type = ifp->if_type; 1154 SDL(gate)->sdl_index = ifp->if_index; 1155 if (ln) 1156 ln->ln_expire = time_second; 1157 if (ln && ln->ln_expire == 0) { 1158 /* kludge for desktops */ 1159 ln->ln_expire = 1; 1160 } 1161 if ((rt->rt_flags & RTF_CLONING) != 0) 1162 break; 1163 } 1164 /* 1165 * In IPv4 code, we try to annonuce new RTF_ANNOUNCE entry here. 1166 * We don't do that here since llinfo is not ready yet. 1167 * 1168 * There are also couple of other things to be discussed: 1169 * - unsolicited NA code needs improvement beforehand 1170 * - RFC2461 says we MAY send multicast unsolicited NA 1171 * (7.2.6 paragraph 4), however, it also says that we 1172 * SHOULD provide a mechanism to prevent multicast NA storm. 1173 * we don't have anything like it right now. 1174 * note that the mechanism needs a mutual agreement 1175 * between proxies, which means that we need to implement 1176 * a new protocol, or a new kludge. 1177 * - from RFC2461 6.2.4, host MUST NOT send an unsolicited NA. 1178 * we need to check ip6forwarding before sending it. 1179 * (or should we allow proxy ND configuration only for 1180 * routers? there's no mention about proxy ND from hosts) 1181 */ 1182#if 0 1183 /* XXX it does not work */ 1184 if (rt->rt_flags & RTF_ANNOUNCE) 1185 nd6_na_output(ifp, 1186 &SIN6(rt_key(rt))->sin6_addr, 1187 &SIN6(rt_key(rt))->sin6_addr, 1188 ip6_forwarding ? ND_NA_FLAG_ROUTER : 0, 1189 1, NULL); 1190#endif 1191 /* FALLTHROUGH */ 1192 case RTM_RESOLVE: 1193 if ((ifp->if_flags & (IFF_POINTOPOINT | IFF_LOOPBACK)) == 0) { 1194 /* 1195 * Address resolution isn't necessary for a point to 1196 * point link, so we can skip this test for a p2p link. 1197 */ 1198 if (gate->sa_family != AF_LINK || 1199 gate->sa_len < sizeof(null_sdl)) { 1200 log(LOG_DEBUG, 1201 "nd6_rtrequest: bad gateway value: %s\n", 1202 if_name(ifp)); 1203 break; 1204 } 1205 SDL(gate)->sdl_type = ifp->if_type; 1206 SDL(gate)->sdl_index = ifp->if_index; 1207 } 1208 if (ln != NULL) 1209 break; /* This happens on a route change */ 1210 /* 1211 * Case 2: This route may come from cloning, or a manual route 1212 * add with a LL address. 1213 */ 1214 R_Malloc(ln, struct llinfo_nd6 *, sizeof(*ln)); 1215 rt->rt_llinfo = (caddr_t)ln; 1216 if (!ln) { 1217 log(LOG_DEBUG, "nd6_rtrequest: malloc failed\n"); 1218 break; 1219 } 1220 nd6_inuse++; 1221 nd6_allocated++; 1222 Bzero(ln, sizeof(*ln)); 1223 ln->ln_rt = rt; 1224 /* this is required for "ndp" command. - shin */ 1225 if (req == RTM_ADD) { 1226 /* 1227 * gate should have some valid AF_LINK entry, 1228 * and ln->ln_expire should have some lifetime 1229 * which is specified by ndp command. 1230 */ 1231 ln->ln_state = ND6_LLINFO_REACHABLE; 1232 ln->ln_byhint = 0; 1233 } else { 1234 /* 1235 * When req == RTM_RESOLVE, rt is created and 1236 * initialized in rtrequest(), so rt_expire is 0. 1237 */ 1238 ln->ln_state = ND6_LLINFO_NOSTATE; 1239 ln->ln_expire = time_second; 1240 } 1241 rt->rt_flags |= RTF_LLINFO; 1242 ln->ln_next = llinfo_nd6.ln_next; 1243 llinfo_nd6.ln_next = ln; 1244 ln->ln_prev = &llinfo_nd6; 1245 ln->ln_next->ln_prev = ln; 1246 1247 /* 1248 * check if rt_key(rt) is one of my address assigned 1249 * to the interface. 1250 */ 1251 ifa = (struct ifaddr *)in6ifa_ifpwithaddr(rt->rt_ifp, 1252 &SIN6(rt_key(rt))->sin6_addr); 1253 if (ifa) { 1254 caddr_t macp = nd6_ifptomac(ifp); 1255 ln->ln_expire = 0; 1256 ln->ln_state = ND6_LLINFO_REACHABLE; 1257 ln->ln_byhint = 0; 1258 if (macp) { 1259 Bcopy(macp, LLADDR(SDL(gate)), ifp->if_addrlen); 1260 SDL(gate)->sdl_alen = ifp->if_addrlen; 1261 } 1262 if (nd6_useloopback) { 1263 rt->rt_ifp = &loif[0]; /* XXX */ 1264 /* 1265 * Make sure rt_ifa be equal to the ifaddr 1266 * corresponding to the address. 1267 * We need this because when we refer 1268 * rt_ifa->ia6_flags in ip6_input, we assume 1269 * that the rt_ifa points to the address instead 1270 * of the loopback address. 1271 */ 1272 if (ifa != rt->rt_ifa) { 1273 IFAFREE(rt->rt_ifa); 1274 IFAREF(ifa); 1275 rt->rt_ifa = ifa; 1276 } 1277 } 1278 } else if (rt->rt_flags & RTF_ANNOUNCE) { 1279 ln->ln_expire = 0; 1280 ln->ln_state = ND6_LLINFO_REACHABLE; 1281 ln->ln_byhint = 0; 1282 1283 /* join solicited node multicast for proxy ND */ 1284 if (ifp->if_flags & IFF_MULTICAST) { 1285 struct in6_addr llsol; 1286 int error; 1287 1288 llsol = SIN6(rt_key(rt))->sin6_addr; 1289 llsol.s6_addr16[0] = htons(0xff02); 1290 llsol.s6_addr16[1] = htons(ifp->if_index); 1291 llsol.s6_addr32[1] = 0; 1292 llsol.s6_addr32[2] = htonl(1); 1293 llsol.s6_addr8[12] = 0xff; 1294 1295 if (!in6_addmulti(&llsol, ifp, &error)) { 1296 nd6log((LOG_ERR, "%s: failed to join " 1297 "%s (errno=%d)\n", if_name(ifp), 1298 ip6_sprintf(&llsol), error)); 1299 } 1300 } 1301 } 1302 break; 1303 1304 case RTM_DELETE: 1305 if (!ln) 1306 break; 1307 /* leave from solicited node multicast for proxy ND */ 1308 if ((rt->rt_flags & RTF_ANNOUNCE) != 0 && 1309 (ifp->if_flags & IFF_MULTICAST) != 0) { 1310 struct in6_addr llsol; 1311 struct in6_multi *in6m; 1312 1313 llsol = SIN6(rt_key(rt))->sin6_addr; 1314 llsol.s6_addr16[0] = htons(0xff02); 1315 llsol.s6_addr16[1] = htons(ifp->if_index); 1316 llsol.s6_addr32[1] = 0; 1317 llsol.s6_addr32[2] = htonl(1); 1318 llsol.s6_addr8[12] = 0xff; 1319 1320 IN6_LOOKUP_MULTI(llsol, ifp, in6m); 1321 if (in6m) 1322 in6_delmulti(in6m); 1323 } 1324 nd6_inuse--; 1325 ln->ln_next->ln_prev = ln->ln_prev; 1326 ln->ln_prev->ln_next = ln->ln_next; 1327 ln->ln_prev = NULL; 1328 rt->rt_llinfo = 0; 1329 rt->rt_flags &= ~RTF_LLINFO; 1330 if (ln->ln_hold) 1331 m_freem(ln->ln_hold); 1332 Free((caddr_t)ln); 1333 } 1334} 1335 1336int 1337nd6_ioctl(cmd, data, ifp) 1338 u_long cmd; 1339 caddr_t data; 1340 struct ifnet *ifp; 1341{ 1342 struct in6_drlist *drl = (struct in6_drlist *)data; 1343 struct in6_prlist *prl = (struct in6_prlist *)data; 1344 struct in6_ndireq *ndi = (struct in6_ndireq *)data; 1345 struct in6_nbrinfo *nbi = (struct in6_nbrinfo *)data; 1346 struct in6_ndifreq *ndif = (struct in6_ndifreq *)data; 1347 struct nd_defrouter *dr, any; 1348 struct nd_prefix *pr; 1349 struct rtentry *rt; 1350 int i = 0, error = 0; 1351 int s; 1352 1353 switch (cmd) { 1354 case SIOCGDRLST_IN6: 1355 /* 1356 * obsolete API, use sysctl under net.inet6.icmp6 1357 */ 1358 bzero(drl, sizeof(*drl)); 1359 s = splnet(); 1360 dr = TAILQ_FIRST(&nd_defrouter); 1361 while (dr && i < DRLSTSIZ) { 1362 drl->defrouter[i].rtaddr = dr->rtaddr; 1363 if (IN6_IS_ADDR_LINKLOCAL(&drl->defrouter[i].rtaddr)) { 1364 /* XXX: need to this hack for KAME stack */ 1365 drl->defrouter[i].rtaddr.s6_addr16[1] = 0; 1366 } else 1367 log(LOG_ERR, 1368 "default router list contains a " 1369 "non-linklocal address(%s)\n", 1370 ip6_sprintf(&drl->defrouter[i].rtaddr)); 1371 1372 drl->defrouter[i].flags = dr->flags; 1373 drl->defrouter[i].rtlifetime = dr->rtlifetime; 1374 drl->defrouter[i].expire = dr->expire; 1375 drl->defrouter[i].if_index = dr->ifp->if_index; 1376 i++; 1377 dr = TAILQ_NEXT(dr, dr_entry); 1378 } 1379 splx(s); 1380 break; 1381 case SIOCGPRLST_IN6: 1382 /* 1383 * obsolete API, use sysctl under net.inet6.icmp6 1384 */ 1385 /* 1386 * XXX meaning of fields, especialy "raflags", is very 1387 * differnet between RA prefix list and RR/static prefix list. 1388 * how about separating ioctls into two? 1389 */ 1390 bzero(prl, sizeof(*prl)); 1391 s = splnet(); 1392 pr = nd_prefix.lh_first; 1393 while (pr && i < PRLSTSIZ) { 1394 struct nd_pfxrouter *pfr; 1395 int j; 1396 1397 (void)in6_embedscope(&prl->prefix[i].prefix, 1398 &pr->ndpr_prefix, NULL, NULL); 1399 prl->prefix[i].raflags = pr->ndpr_raf; 1400 prl->prefix[i].prefixlen = pr->ndpr_plen; 1401 prl->prefix[i].vltime = pr->ndpr_vltime; 1402 prl->prefix[i].pltime = pr->ndpr_pltime; 1403 prl->prefix[i].if_index = pr->ndpr_ifp->if_index; 1404 prl->prefix[i].expire = pr->ndpr_expire; 1405 1406 pfr = pr->ndpr_advrtrs.lh_first; 1407 j = 0; 1408 while (pfr) { 1409 if (j < DRLSTSIZ) { 1410#define RTRADDR prl->prefix[i].advrtr[j] 1411 RTRADDR = pfr->router->rtaddr; 1412 if (IN6_IS_ADDR_LINKLOCAL(&RTRADDR)) { 1413 /* XXX: hack for KAME */ 1414 RTRADDR.s6_addr16[1] = 0; 1415 } else 1416 log(LOG_ERR, 1417 "a router(%s) advertises " 1418 "a prefix with " 1419 "non-link local address\n", 1420 ip6_sprintf(&RTRADDR)); 1421#undef RTRADDR 1422 } 1423 j++; 1424 pfr = pfr->pfr_next; 1425 } 1426 prl->prefix[i].advrtrs = j; 1427 prl->prefix[i].origin = PR_ORIG_RA; 1428 1429 i++; 1430 pr = pr->ndpr_next; 1431 } 1432 { 1433 struct rr_prefix *rpp; 1434 1435 for (rpp = LIST_FIRST(&rr_prefix); rpp; 1436 rpp = LIST_NEXT(rpp, rp_entry)) { 1437 if (i >= PRLSTSIZ) 1438 break; 1439 (void)in6_embedscope(&prl->prefix[i].prefix, 1440 &pr->ndpr_prefix, NULL, NULL); 1441 prl->prefix[i].raflags = rpp->rp_raf; 1442 prl->prefix[i].prefixlen = rpp->rp_plen; 1443 prl->prefix[i].vltime = rpp->rp_vltime; 1444 prl->prefix[i].pltime = rpp->rp_pltime; 1445 prl->prefix[i].if_index = rpp->rp_ifp->if_index; 1446 prl->prefix[i].expire = rpp->rp_expire; 1447 prl->prefix[i].advrtrs = 0; 1448 prl->prefix[i].origin = rpp->rp_origin; 1449 i++; 1450 } 1451 } 1452 splx(s); 1453 1454 break; 1455 case OSIOCGIFINFO_IN6: 1456 if (!nd_ifinfo || i >= nd_ifinfo_indexlim) { 1457 error = EINVAL; 1458 break; 1459 } 1460 ndi->ndi.linkmtu = nd_ifinfo[ifp->if_index].linkmtu; 1461 ndi->ndi.maxmtu = nd_ifinfo[ifp->if_index].maxmtu; 1462 ndi->ndi.basereachable = 1463 nd_ifinfo[ifp->if_index].basereachable; 1464 ndi->ndi.reachable = nd_ifinfo[ifp->if_index].reachable; 1465 ndi->ndi.retrans = nd_ifinfo[ifp->if_index].retrans; 1466 ndi->ndi.flags = nd_ifinfo[ifp->if_index].flags; 1467 ndi->ndi.recalctm = nd_ifinfo[ifp->if_index].recalctm; 1468 ndi->ndi.chlim = nd_ifinfo[ifp->if_index].chlim; 1469 ndi->ndi.receivedra = nd_ifinfo[ifp->if_index].receivedra; 1470 break; 1471 case SIOCGIFINFO_IN6: 1472 if (!nd_ifinfo || i >= nd_ifinfo_indexlim) { 1473 error = EINVAL; 1474 break; 1475 } 1476 ndi->ndi = nd_ifinfo[ifp->if_index]; 1477 break; 1478 case SIOCSIFINFO_FLAGS: 1479 /* XXX: almost all other fields of ndi->ndi is unused */ 1480 if (!nd_ifinfo || i >= nd_ifinfo_indexlim) { 1481 error = EINVAL; 1482 break; 1483 } 1484 nd_ifinfo[ifp->if_index].flags = ndi->ndi.flags; 1485 break; 1486 case SIOCSNDFLUSH_IN6: /* XXX: the ioctl name is confusing... */ 1487 /* flush default router list */ 1488 /* 1489 * xxx sumikawa: should not delete route if default 1490 * route equals to the top of default router list 1491 */ 1492 bzero(&any, sizeof(any)); 1493 defrouter_delreq(&any, 0); 1494 defrouter_select(); 1495 /* xxx sumikawa: flush prefix list */ 1496 break; 1497 case SIOCSPFXFLUSH_IN6: 1498 { 1499 /* flush all the prefix advertised by routers */ 1500 struct nd_prefix *pr, *next; 1501 1502 s = splnet(); 1503 for (pr = nd_prefix.lh_first; pr; pr = next) { 1504 struct in6_ifaddr *ia, *ia_next; 1505 1506 next = pr->ndpr_next; 1507 1508 if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr)) 1509 continue; /* XXX */ 1510 1511 /* do we really have to remove addresses as well? */ 1512 for (ia = in6_ifaddr; ia; ia = ia_next) { 1513 /* ia might be removed. keep the next ptr. */ 1514 ia_next = ia->ia_next; 1515 1516 if ((ia->ia6_flags & IN6_IFF_AUTOCONF) == 0) 1517 continue; 1518 1519 if (ia->ia6_ndpr == pr) 1520 in6_purgeaddr(&ia->ia_ifa); 1521 } 1522 prelist_remove(pr); 1523 } 1524 splx(s); 1525 break; 1526 } 1527 case SIOCSRTRFLUSH_IN6: 1528 { 1529 /* flush all the default routers */ 1530 struct nd_defrouter *dr, *next; 1531 1532 s = splnet(); 1533 if ((dr = TAILQ_FIRST(&nd_defrouter)) != NULL) { 1534 /* 1535 * The first entry of the list may be stored in 1536 * the routing table, so we'll delete it later. 1537 */ 1538 for (dr = TAILQ_NEXT(dr, dr_entry); dr; dr = next) { 1539 next = TAILQ_NEXT(dr, dr_entry); 1540 defrtrlist_del(dr); 1541 } 1542 defrtrlist_del(TAILQ_FIRST(&nd_defrouter)); 1543 } 1544 splx(s); 1545 break; 1546 } 1547 case SIOCGNBRINFO_IN6: 1548 { 1549 struct llinfo_nd6 *ln; 1550 struct in6_addr nb_addr = nbi->addr; /* make local for safety */ 1551 1552 /* 1553 * XXX: KAME specific hack for scoped addresses 1554 * XXXX: for other scopes than link-local? 1555 */ 1556 if (IN6_IS_ADDR_LINKLOCAL(&nbi->addr) || 1557 IN6_IS_ADDR_MC_LINKLOCAL(&nbi->addr)) { 1558 u_int16_t *idp = (u_int16_t *)&nb_addr.s6_addr[2]; 1559 1560 if (*idp == 0) 1561 *idp = htons(ifp->if_index); 1562 } 1563 1564 s = splnet(); 1565 if ((rt = nd6_lookup(&nb_addr, 0, ifp)) == NULL) { 1566 error = EINVAL; 1567 splx(s); 1568 break; 1569 } 1570 ln = (struct llinfo_nd6 *)rt->rt_llinfo; 1571 nbi->state = ln->ln_state; 1572 nbi->asked = ln->ln_asked; 1573 nbi->isrouter = ln->ln_router; 1574 nbi->expire = ln->ln_expire; 1575 splx(s); 1576 1577 break; 1578 } 1579 case SIOCGDEFIFACE_IN6: /* XXX: should be implemented as a sysctl? */ 1580 ndif->ifindex = nd6_defifindex; 1581 break; 1582 case SIOCSDEFIFACE_IN6: /* XXX: should be implemented as a sysctl? */ 1583 return (nd6_setdefaultiface(ndif->ifindex)); 1584 } 1585 return (error); 1586} 1587 1588/* 1589 * Create neighbor cache entry and cache link-layer address, 1590 * on reception of inbound ND6 packets. (RS/RA/NS/redirect) 1591 */ 1592struct rtentry * 1593nd6_cache_lladdr(ifp, from, lladdr, lladdrlen, type, code) 1594 struct ifnet *ifp; 1595 struct in6_addr *from; 1596 char *lladdr; 1597 int lladdrlen; 1598 int type; /* ICMP6 type */ 1599 int code; /* type dependent information */ 1600{ 1601 struct rtentry *rt = NULL; 1602 struct llinfo_nd6 *ln = NULL; 1603 int is_newentry; 1604 struct sockaddr_dl *sdl = NULL; 1605 int do_update; 1606 int olladdr; 1607 int llchange; 1608 int newstate = 0; 1609 1610 if (!ifp) 1611 panic("ifp == NULL in nd6_cache_lladdr"); 1612 if (!from) 1613 panic("from == NULL in nd6_cache_lladdr"); 1614 1615 /* nothing must be updated for unspecified address */ 1616 if (IN6_IS_ADDR_UNSPECIFIED(from)) 1617 return NULL; 1618 1619 /* 1620 * Validation about ifp->if_addrlen and lladdrlen must be done in 1621 * the caller. 1622 * 1623 * XXX If the link does not have link-layer adderss, what should 1624 * we do? (ifp->if_addrlen == 0) 1625 * Spec says nothing in sections for RA, RS and NA. There's small 1626 * description on it in NS section (RFC 2461 7.2.3). 1627 */ 1628 1629 rt = nd6_lookup(from, 0, ifp); 1630 if (!rt) { 1631#if 0 1632 /* nothing must be done if there's no lladdr */ 1633 if (!lladdr || !lladdrlen) 1634 return NULL; 1635#endif 1636 1637 rt = nd6_lookup(from, 1, ifp); 1638 is_newentry = 1; 1639 } else { 1640 /* do nothing if static ndp is set */ 1641 if (rt->rt_flags & RTF_STATIC) 1642 return NULL; 1643 is_newentry = 0; 1644 } 1645 1646 if (!rt) 1647 return NULL; 1648 if ((rt->rt_flags & (RTF_GATEWAY | RTF_LLINFO)) != RTF_LLINFO) { 1649fail: 1650 (void)nd6_free(rt); 1651 return NULL; 1652 } 1653 ln = (struct llinfo_nd6 *)rt->rt_llinfo; 1654 if (!ln) 1655 goto fail; 1656 if (!rt->rt_gateway) 1657 goto fail; 1658 if (rt->rt_gateway->sa_family != AF_LINK) 1659 goto fail; 1660 sdl = SDL(rt->rt_gateway); 1661 1662 olladdr = (sdl->sdl_alen) ? 1 : 0; 1663 if (olladdr && lladdr) { 1664 if (bcmp(lladdr, LLADDR(sdl), ifp->if_addrlen)) 1665 llchange = 1; 1666 else 1667 llchange = 0; 1668 } else 1669 llchange = 0; 1670 1671 /* 1672 * newentry olladdr lladdr llchange (*=record) 1673 * 0 n n -- (1) 1674 * 0 y n -- (2) 1675 * 0 n y -- (3) * STALE 1676 * 0 y y n (4) * 1677 * 0 y y y (5) * STALE 1678 * 1 -- n -- (6) NOSTATE(= PASSIVE) 1679 * 1 -- y -- (7) * STALE 1680 */ 1681 1682 if (lladdr) { /* (3-5) and (7) */ 1683 /* 1684 * Record source link-layer address 1685 * XXX is it dependent to ifp->if_type? 1686 */ 1687 sdl->sdl_alen = ifp->if_addrlen; 1688 bcopy(lladdr, LLADDR(sdl), ifp->if_addrlen); 1689 } 1690 1691 if (!is_newentry) { 1692 if ((!olladdr && lladdr) || /* (3) */ 1693 (olladdr && lladdr && llchange)) { /* (5) */ 1694 do_update = 1; 1695 newstate = ND6_LLINFO_STALE; 1696 } else /* (1-2,4) */ 1697 do_update = 0; 1698 } else { 1699 do_update = 1; 1700 if (!lladdr) /* (6) */ 1701 newstate = ND6_LLINFO_NOSTATE; 1702 else /* (7) */ 1703 newstate = ND6_LLINFO_STALE; 1704 } 1705 1706 if (do_update) { 1707 /* 1708 * Update the state of the neighbor cache. 1709 */ 1710 ln->ln_state = newstate; 1711 1712 if (ln->ln_state == ND6_LLINFO_STALE) { 1713 /* 1714 * XXX: since nd6_output() below will cause 1715 * state tansition to DELAY and reset the timer, 1716 * we must set the timer now, although it is actually 1717 * meaningless. 1718 */ 1719 ln->ln_expire = time_second + nd6_gctimer; 1720 1721 if (ln->ln_hold) { 1722 /* 1723 * we assume ifp is not a p2p here, so just 1724 * set the 2nd argument as the 1st one. 1725 */ 1726 nd6_output(ifp, ifp, ln->ln_hold, 1727 (struct sockaddr_in6 *)rt_key(rt), rt); 1728 ln->ln_hold = NULL; 1729 } 1730 } else if (ln->ln_state == ND6_LLINFO_INCOMPLETE) { 1731 /* probe right away */ 1732 ln->ln_expire = time_second; 1733 } 1734 } 1735 1736 /* 1737 * ICMP6 type dependent behavior. 1738 * 1739 * NS: clear IsRouter if new entry 1740 * RS: clear IsRouter 1741 * RA: set IsRouter if there's lladdr 1742 * redir: clear IsRouter if new entry 1743 * 1744 * RA case, (1): 1745 * The spec says that we must set IsRouter in the following cases: 1746 * - If lladdr exist, set IsRouter. This means (1-5). 1747 * - If it is old entry (!newentry), set IsRouter. This means (7). 1748 * So, based on the spec, in (1-5) and (7) cases we must set IsRouter. 1749 * A quetion arises for (1) case. (1) case has no lladdr in the 1750 * neighbor cache, this is similar to (6). 1751 * This case is rare but we figured that we MUST NOT set IsRouter. 1752 * 1753 * newentry olladdr lladdr llchange NS RS RA redir 1754 * D R 1755 * 0 n n -- (1) c ? s 1756 * 0 y n -- (2) c s s 1757 * 0 n y -- (3) c s s 1758 * 0 y y n (4) c s s 1759 * 0 y y y (5) c s s 1760 * 1 -- n -- (6) c c c s 1761 * 1 -- y -- (7) c c s c s 1762 * 1763 * (c=clear s=set) 1764 */ 1765 switch (type & 0xff) { 1766 case ND_NEIGHBOR_SOLICIT: 1767 /* 1768 * New entry must have is_router flag cleared. 1769 */ 1770 if (is_newentry) /* (6-7) */ 1771 ln->ln_router = 0; 1772 break; 1773 case ND_REDIRECT: 1774 /* 1775 * If the icmp is a redirect to a better router, always set the 1776 * is_router flag. Otherwise, if the entry is newly created, 1777 * clear the flag. [RFC 2461, sec 8.3] 1778 */ 1779 if (code == ND_REDIRECT_ROUTER) 1780 ln->ln_router = 1; 1781 else if (is_newentry) /* (6-7) */ 1782 ln->ln_router = 0; 1783 break; 1784 case ND_ROUTER_SOLICIT: 1785 /* 1786 * is_router flag must always be cleared. 1787 */ 1788 ln->ln_router = 0; 1789 break; 1790 case ND_ROUTER_ADVERT: 1791 /* 1792 * Mark an entry with lladdr as a router. 1793 */ 1794 if ((!is_newentry && (olladdr || lladdr)) || /* (2-5) */ 1795 (is_newentry && lladdr)) { /* (7) */ 1796 ln->ln_router = 1; 1797 } 1798 break; 1799 } 1800 1801 /* 1802 * When the link-layer address of a router changes, select the 1803 * best router again. In particular, when the neighbor entry is newly 1804 * created, it might affect the selection policy. 1805 * Question: can we restrict the first condition to the "is_newentry" 1806 * case? 1807 * XXX: when we hear an RA from a new router with the link-layer 1808 * address option, defrouter_select() is called twice, since 1809 * defrtrlist_update called the function as well. However, I believe 1810 * we can compromise the overhead, since it only happens the first 1811 * time. 1812 * XXX: although defrouter_select() should not have a bad effect 1813 * for those are not autoconfigured hosts, we explicitly avoid such 1814 * cases for safety. 1815 */ 1816 if (do_update && ln->ln_router && !ip6_forwarding && ip6_accept_rtadv) 1817 defrouter_select(); 1818 1819 return rt; 1820} 1821 1822static void 1823nd6_slowtimo(ignored_arg) 1824 void *ignored_arg; 1825{ 1826 int s = splnet(); 1827 int i; 1828 struct nd_ifinfo *nd6if; 1829 1830 callout_reset(&nd6_slowtimo_ch, ND6_SLOWTIMER_INTERVAL * hz, 1831 nd6_slowtimo, NULL); 1832 for (i = 1; i < if_index + 1; i++) { 1833 if (!nd_ifinfo || i >= nd_ifinfo_indexlim) 1834 continue; 1835 nd6if = &nd_ifinfo[i]; 1836 if (nd6if->basereachable && /* already initialized */ 1837 (nd6if->recalctm -= ND6_SLOWTIMER_INTERVAL) <= 0) { 1838 /* 1839 * Since reachable time rarely changes by router 1840 * advertisements, we SHOULD insure that a new random 1841 * value gets recomputed at least once every few hours. 1842 * (RFC 2461, 6.3.4) 1843 */ 1844 nd6if->recalctm = nd6_recalc_reachtm_interval; 1845 nd6if->reachable = ND_COMPUTE_RTIME(nd6if->basereachable); 1846 } 1847 } 1848 splx(s); 1849} 1850 1851#define senderr(e) { error = (e); goto bad;} 1852int 1853nd6_output(ifp, origifp, m0, dst, rt0) 1854 struct ifnet *ifp; 1855 struct ifnet *origifp; 1856 struct mbuf *m0; 1857 struct sockaddr_in6 *dst; 1858 struct rtentry *rt0; 1859{ 1860 struct mbuf *m = m0; 1861 struct rtentry *rt = rt0; 1862 struct sockaddr_in6 *gw6 = NULL; 1863 struct llinfo_nd6 *ln = NULL; 1864 int error = 0; 1865 1866 if (IN6_IS_ADDR_MULTICAST(&dst->sin6_addr)) 1867 goto sendpkt; 1868 1869 if (nd6_need_cache(ifp) == 0) 1870 goto sendpkt; 1871 1872 /* 1873 * next hop determination. This routine is derived from ether_outpout. 1874 */ 1875 if (rt) { 1876 if ((rt->rt_flags & RTF_UP) == 0) { 1877 rt0 = rt = rtalloc1((struct sockaddr *)dst, 1, 0UL); 1878 if (rt != NULL) { 1879 rt->rt_refcnt--; 1880 RT_UNLOCK(rt); 1881 if (rt->rt_ifp != ifp) { 1882 /* XXX: loop care? */ 1883 return nd6_output(ifp, origifp, m0, 1884 dst, rt); 1885 } 1886 } else 1887 senderr(EHOSTUNREACH); 1888 } 1889 1890 if (rt->rt_flags & RTF_GATEWAY) { 1891 gw6 = (struct sockaddr_in6 *)rt->rt_gateway; 1892 1893 /* 1894 * We skip link-layer address resolution and NUD 1895 * if the gateway is not a neighbor from ND point 1896 * of view, regardless of the value of nd_ifinfo.flags. 1897 * The second condition is a bit tricky; we skip 1898 * if the gateway is our own address, which is 1899 * sometimes used to install a route to a p2p link. 1900 */ 1901 if (!nd6_is_addr_neighbor(gw6, ifp) || 1902 in6ifa_ifpwithaddr(ifp, &gw6->sin6_addr)) { 1903 /* 1904 * We allow this kind of tricky route only 1905 * when the outgoing interface is p2p. 1906 * XXX: we may need a more generic rule here. 1907 */ 1908 if ((ifp->if_flags & IFF_POINTOPOINT) == 0) 1909 senderr(EHOSTUNREACH); 1910 1911 goto sendpkt; 1912 } 1913 1914 if (rt->rt_gwroute == 0) 1915 goto lookup; 1916 if (((rt = rt->rt_gwroute)->rt_flags & RTF_UP) == 0) { 1917 rtfree(rt); rt = rt0; 1918 lookup: 1919 rt->rt_gwroute = rtalloc1(rt->rt_gateway, 1, 0UL); 1920 if ((rt = rt->rt_gwroute) == 0) 1921 senderr(EHOSTUNREACH); 1922 RT_UNLOCK(rt); 1923 } 1924 } 1925 } 1926 1927 /* 1928 * Address resolution or Neighbor Unreachability Detection 1929 * for the next hop. 1930 * At this point, the destination of the packet must be a unicast 1931 * or an anycast address(i.e. not a multicast). 1932 */ 1933 1934 /* Look up the neighbor cache for the nexthop */ 1935 if (rt && (rt->rt_flags & RTF_LLINFO) != 0) 1936 ln = (struct llinfo_nd6 *)rt->rt_llinfo; 1937 else { 1938 /* 1939 * Since nd6_is_addr_neighbor() internally calls nd6_lookup(), 1940 * the condition below is not very efficient. But we believe 1941 * it is tolerable, because this should be a rare case. 1942 */ 1943 if (nd6_is_addr_neighbor(dst, ifp) && 1944 (rt = nd6_lookup(&dst->sin6_addr, 1, ifp)) != NULL) 1945 ln = (struct llinfo_nd6 *)rt->rt_llinfo; 1946 } 1947 if (!ln || !rt) { 1948 if ((ifp->if_flags & IFF_POINTOPOINT) == 0 && 1949 !(nd_ifinfo[ifp->if_index].flags & ND6_IFF_PERFORMNUD)) { 1950 log(LOG_DEBUG, 1951 "nd6_output: can't allocate llinfo for %s " 1952 "(ln=%p, rt=%p)\n", 1953 ip6_sprintf(&dst->sin6_addr), ln, rt); 1954 senderr(EIO); /* XXX: good error? */ 1955 } 1956 1957 goto sendpkt; /* send anyway */ 1958 } 1959 1960 /* We don't have to do link-layer address resolution on a p2p link. */ 1961 if ((ifp->if_flags & IFF_POINTOPOINT) != 0 && 1962 ln->ln_state < ND6_LLINFO_REACHABLE) { 1963 ln->ln_state = ND6_LLINFO_STALE; 1964 ln->ln_expire = time_second + nd6_gctimer; 1965 } 1966 1967 /* 1968 * The first time we send a packet to a neighbor whose entry is 1969 * STALE, we have to change the state to DELAY and a sets a timer to 1970 * expire in DELAY_FIRST_PROBE_TIME seconds to ensure do 1971 * neighbor unreachability detection on expiration. 1972 * (RFC 2461 7.3.3) 1973 */ 1974 if (ln->ln_state == ND6_LLINFO_STALE) { 1975 ln->ln_asked = 0; 1976 ln->ln_state = ND6_LLINFO_DELAY; 1977 ln->ln_expire = time_second + nd6_delay; 1978 } 1979 1980 /* 1981 * If the neighbor cache entry has a state other than INCOMPLETE 1982 * (i.e. its link-layer address is already resolved), just 1983 * send the packet. 1984 */ 1985 if (ln->ln_state > ND6_LLINFO_INCOMPLETE) 1986 goto sendpkt; 1987 1988 /* 1989 * There is a neighbor cache entry, but no ethernet address 1990 * response yet. Replace the held mbuf (if any) with this 1991 * latest one. 1992 * 1993 * This code conforms to the rate-limiting rule described in Section 1994 * 7.2.2 of RFC 2461, because the timer is set correctly after sending 1995 * an NS below. 1996 */ 1997 if (ln->ln_state == ND6_LLINFO_NOSTATE) 1998 ln->ln_state = ND6_LLINFO_INCOMPLETE; 1999 if (ln->ln_hold) 2000 m_freem(ln->ln_hold); 2001 ln->ln_hold = m; 2002 if (ln->ln_expire) { 2003 if (ln->ln_asked < nd6_mmaxtries && 2004 ln->ln_expire < time_second) { 2005 ln->ln_asked++; 2006 ln->ln_expire = time_second + 2007 nd_ifinfo[ifp->if_index].retrans / 1000; 2008 nd6_ns_output(ifp, NULL, &dst->sin6_addr, ln, 0); 2009 } 2010 } 2011 return (0); 2012 2013 sendpkt: 2014 2015#ifdef MAC 2016 mac_create_mbuf_linklayer(ifp, m); 2017#endif 2018 if ((ifp->if_flags & IFF_LOOPBACK) != 0) { 2019 return ((*ifp->if_output)(origifp, m, (struct sockaddr *)dst, 2020 rt)); 2021 } 2022 return ((*ifp->if_output)(ifp, m, (struct sockaddr *)dst, rt)); 2023 2024 bad: 2025 if (m) 2026 m_freem(m); 2027 return (error); 2028} 2029#undef senderr 2030 2031int 2032nd6_need_cache(ifp) 2033 struct ifnet *ifp; 2034{ 2035 /* 2036 * XXX: we currently do not make neighbor cache on any interface 2037 * other than ARCnet, Ethernet, FDDI and GIF. 2038 * 2039 * RFC2893 says: 2040 * - unidirectional tunnels needs no ND 2041 */ 2042 switch (ifp->if_type) { 2043 case IFT_ARCNET: 2044 case IFT_ETHER: 2045 case IFT_FDDI: 2046 case IFT_IEEE1394: 2047#ifdef IFT_L2VLAN 2048 case IFT_L2VLAN: 2049#endif 2050#ifdef IFT_IEEE80211 2051 case IFT_IEEE80211: 2052#endif 2053 case IFT_GIF: /* XXX need more cases? */ 2054 return (1); 2055 default: 2056 return (0); 2057 } 2058} 2059 2060int 2061nd6_storelladdr(ifp, rt, m, dst, desten) 2062 struct ifnet *ifp; 2063 struct rtentry *rt; 2064 struct mbuf *m; 2065 struct sockaddr *dst; 2066 u_char *desten; 2067{ 2068 int i; 2069 struct sockaddr_dl *sdl; 2070 2071 if (m->m_flags & M_MCAST) { 2072 switch (ifp->if_type) { 2073 case IFT_ETHER: 2074 case IFT_FDDI: 2075#ifdef IFT_L2VLAN 2076 case IFT_L2VLAN: 2077#endif 2078#ifdef IFT_IEEE80211 2079 case IFT_IEEE80211: 2080#endif 2081 case IFT_ISO88025: 2082 ETHER_MAP_IPV6_MULTICAST(&SIN6(dst)->sin6_addr, 2083 desten); 2084 return (1); 2085 case IFT_IEEE1394: 2086 /* 2087 * netbsd can use if_broadcastaddr, but we don't do so 2088 * to reduce # of ifdef. 2089 */ 2090 for (i = 0; i < ifp->if_addrlen; i++) 2091 desten[i] = ~0; 2092 return (1); 2093 case IFT_ARCNET: 2094 *desten = 0; 2095 return (1); 2096 default: 2097 m_freem(m); 2098 return (0); 2099 } 2100 } 2101 2102 if (rt == NULL) { 2103 /* this could happen, if we could not allocate memory */ 2104 m_freem(m); 2105 return (0); 2106 } 2107 if (rt->rt_gateway->sa_family != AF_LINK) { 2108 printf("nd6_storelladdr: something odd happens\n"); 2109 m_freem(m); 2110 return (0); 2111 } 2112 sdl = SDL(rt->rt_gateway); 2113 if (sdl->sdl_alen == 0) { 2114 /* this should be impossible, but we bark here for debugging */ 2115 printf("nd6_storelladdr: sdl_alen == 0\n"); 2116 m_freem(m); 2117 return (0); 2118 } 2119 2120 bcopy(LLADDR(sdl), desten, sdl->sdl_alen); 2121 return (1); 2122} 2123 2124static int nd6_sysctl_drlist(SYSCTL_HANDLER_ARGS); 2125static int nd6_sysctl_prlist(SYSCTL_HANDLER_ARGS); 2126#ifdef SYSCTL_DECL 2127SYSCTL_DECL(_net_inet6_icmp6); 2128#endif 2129SYSCTL_NODE(_net_inet6_icmp6, ICMPV6CTL_ND6_DRLIST, nd6_drlist, 2130 CTLFLAG_RD, nd6_sysctl_drlist, ""); 2131SYSCTL_NODE(_net_inet6_icmp6, ICMPV6CTL_ND6_PRLIST, nd6_prlist, 2132 CTLFLAG_RD, nd6_sysctl_prlist, ""); 2133 2134static int 2135nd6_sysctl_drlist(SYSCTL_HANDLER_ARGS) 2136{ 2137 int error; 2138 char buf[1024]; 2139 struct in6_defrouter *d, *de; 2140 struct nd_defrouter *dr; 2141 2142 if (req->newptr) 2143 return EPERM; 2144 error = 0; 2145 2146 for (dr = TAILQ_FIRST(&nd_defrouter); dr; 2147 dr = TAILQ_NEXT(dr, dr_entry)) { 2148 d = (struct in6_defrouter *)buf; 2149 de = (struct in6_defrouter *)(buf + sizeof(buf)); 2150 2151 if (d + 1 <= de) { 2152 bzero(d, sizeof(*d)); 2153 d->rtaddr.sin6_family = AF_INET6; 2154 d->rtaddr.sin6_len = sizeof(d->rtaddr); 2155 if (in6_recoverscope(&d->rtaddr, &dr->rtaddr, 2156 dr->ifp) != 0) 2157 log(LOG_ERR, 2158 "scope error in " 2159 "default router list (%s)\n", 2160 ip6_sprintf(&dr->rtaddr)); 2161 d->flags = dr->flags; 2162 d->rtlifetime = dr->rtlifetime; 2163 d->expire = dr->expire; 2164 d->if_index = dr->ifp->if_index; 2165 } else 2166 panic("buffer too short"); 2167 2168 error = SYSCTL_OUT(req, buf, sizeof(*d)); 2169 if (error) 2170 break; 2171 } 2172 2173 return (error); 2174} 2175 2176static int 2177nd6_sysctl_prlist(SYSCTL_HANDLER_ARGS) 2178{ 2179 int error; 2180 char buf[1024]; 2181 struct in6_prefix *p, *pe; 2182 struct nd_prefix *pr; 2183 2184 if (req->newptr) 2185 return EPERM; 2186 error = 0; 2187 2188 for (pr = nd_prefix.lh_first; pr; pr = pr->ndpr_next) { 2189 u_short advrtrs; 2190 size_t advance; 2191 struct sockaddr_in6 *sin6, *s6; 2192 struct nd_pfxrouter *pfr; 2193 2194 p = (struct in6_prefix *)buf; 2195 pe = (struct in6_prefix *)(buf + sizeof(buf)); 2196 2197 if (p + 1 <= pe) { 2198 bzero(p, sizeof(*p)); 2199 sin6 = (struct sockaddr_in6 *)(p + 1); 2200 2201 p->prefix = pr->ndpr_prefix; 2202 if (in6_recoverscope(&p->prefix, 2203 &p->prefix.sin6_addr, pr->ndpr_ifp) != 0) 2204 log(LOG_ERR, 2205 "scope error in prefix list (%s)\n", 2206 ip6_sprintf(&p->prefix.sin6_addr)); 2207 p->raflags = pr->ndpr_raf; 2208 p->prefixlen = pr->ndpr_plen; 2209 p->vltime = pr->ndpr_vltime; 2210 p->pltime = pr->ndpr_pltime; 2211 p->if_index = pr->ndpr_ifp->if_index; 2212 p->expire = pr->ndpr_expire; 2213 p->refcnt = pr->ndpr_refcnt; 2214 p->flags = pr->ndpr_stateflags; 2215 p->origin = PR_ORIG_RA; 2216 advrtrs = 0; 2217 for (pfr = pr->ndpr_advrtrs.lh_first; pfr; 2218 pfr = pfr->pfr_next) { 2219 if ((void *)&sin6[advrtrs + 1] > (void *)pe) { 2220 advrtrs++; 2221 continue; 2222 } 2223 s6 = &sin6[advrtrs]; 2224 bzero(s6, sizeof(*s6)); 2225 s6->sin6_family = AF_INET6; 2226 s6->sin6_len = sizeof(*sin6); 2227 if (in6_recoverscope(s6, &pfr->router->rtaddr, 2228 pfr->router->ifp) != 0) 2229 log(LOG_ERR, 2230 "scope error in " 2231 "prefix list (%s)\n", 2232 ip6_sprintf(&pfr->router->rtaddr)); 2233 advrtrs++; 2234 } 2235 p->advrtrs = advrtrs; 2236 } else 2237 panic("buffer too short"); 2238 2239 advance = sizeof(*p) + sizeof(*sin6) * advrtrs; 2240 error = SYSCTL_OUT(req, buf, advance); 2241 if (error) 2242 break; 2243 } 2244 2245 return (error); 2246} 2247