1 /*- 2 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project. 3 * All rights reserved. 4 * 5 * Redistribution and use in source and binary forms, with or without 6 * modification, are permitted provided that the following conditions 7 * are met: 8 * 1. Redistributions of source code must retain the above copyright 9 * notice, this list of conditions and the following disclaimer. 10 * 2. Redistributions in binary form must reproduce the above copyright 11 * notice, this list of conditions and the following disclaimer in the 12 * documentation and/or other materials provided with the distribution. 13 * 3. Neither the name of the project nor the names of its contributors 14 * may be used to endorse or promote products derived from this software 15 * without specific prior written permission. 16 * 17 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND 18 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 19 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 20 * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE 21 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 22 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 23 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 24 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 25 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 26 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 27 * SUCH DAMAGE. 28 * 29 * $KAME: nd6.c,v 1.144 2001/05/24 07:44:00 itojun Exp $ 30 */ 31 32 #include <sys/cdefs.h> 33 __FBSDID("$FreeBSD$"); 34 35 #include "opt_inet.h" 36 #include "opt_inet6.h" 37 38 #include <sys/param.h> 39 #include <sys/systm.h> 40 #include <sys/callout.h> 41 #include <sys/malloc.h> 42 #include <sys/mbuf.h> 43 #include <sys/socket.h> 44 #include <sys/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/lock.h> 51 #include <sys/rwlock.h> 52 #include <sys/queue.h> 53 #include <sys/sysctl.h> 54 55 #include <net/if.h> 56 #include <net/if_arc.h> 57 #include <net/if_dl.h> 58 #include <net/if_types.h> 59 #include <net/iso88025.h> 60 #include <net/fddi.h> 61 #include <net/route.h> 62 #include <net/vnet.h> 63 64 #include <netinet/in.h> 65 #include <net/if_llatbl.h> 66 #define L3_ADDR_SIN6(le) ((struct sockaddr_in6 *) L3_ADDR(le)) 67 #include <netinet/if_ether.h> 68 #include <netinet6/in6_var.h> 69 #include <netinet/ip6.h> 70 #include <netinet6/ip6_var.h> 71 #include <netinet6/scope6_var.h> 72 #include <netinet6/nd6.h> 73 #include <netinet6/in6_ifattach.h> 74 #include <netinet/icmp6.h> 75 76 #include <sys/limits.h> 77 78 #include <security/mac/mac_framework.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 85 /* timer values */ 86 VNET_DEFINE(int, nd6_prune) = 1; /* walk list every 1 seconds */ 87 VNET_DEFINE(int, nd6_delay) = 5; /* delay first probe time 5 second */ 88 VNET_DEFINE(int, nd6_umaxtries) = 3; /* maximum unicast query */ 89 VNET_DEFINE(int, nd6_mmaxtries) = 3; /* maximum multicast query */ 90 VNET_DEFINE(int, nd6_useloopback) = 1; /* use loopback interface for 91 * local traffic */ 92 VNET_DEFINE(int, nd6_gctimer) = (60 * 60 * 24); /* 1 day: garbage 93 * collection timer */ 94 95 /* preventing too many loops in ND option parsing */ 96 static VNET_DEFINE(int, nd6_maxndopt) = 10; /* max # of ND options allowed */ 97 98 VNET_DEFINE(int, nd6_maxnudhint) = 0; /* max # of subsequent upper 99 * layer hints */ 100 static VNET_DEFINE(int, nd6_maxqueuelen) = 1; /* max pkts cached in unresolved 101 * ND entries */ 102 #define V_nd6_maxndopt VNET(nd6_maxndopt) 103 #define V_nd6_maxqueuelen VNET(nd6_maxqueuelen) 104 105 #ifdef ND6_DEBUG 106 VNET_DEFINE(int, nd6_debug) = 1; 107 #else 108 VNET_DEFINE(int, nd6_debug) = 0; 109 #endif 110 111 /* for debugging? */ 112 #if 0 113 static int nd6_inuse, nd6_allocated; 114 #endif 115 116 VNET_DEFINE(struct nd_drhead, nd_defrouter); 117 VNET_DEFINE(struct nd_prhead, nd_prefix); 118 119 VNET_DEFINE(int, nd6_recalc_reachtm_interval) = ND6_RECALC_REACHTM_INTERVAL; 120 #define V_nd6_recalc_reachtm_interval VNET(nd6_recalc_reachtm_interval) 121 122 static struct sockaddr_in6 all1_sa; 123 124 static int nd6_is_new_addr_neighbor __P((struct sockaddr_in6 *, 125 struct ifnet *)); 126 static void nd6_setmtu0(struct ifnet *, struct nd_ifinfo *); 127 static void nd6_slowtimo(void *); 128 static int regen_tmpaddr(struct in6_ifaddr *); 129 static struct llentry *nd6_free(struct llentry *, int); 130 static void nd6_llinfo_timer(void *); 131 static void clear_llinfo_pqueue(struct llentry *); 132 133 static VNET_DEFINE(struct callout, nd6_slowtimo_ch); 134 #define V_nd6_slowtimo_ch VNET(nd6_slowtimo_ch) 135 136 VNET_DEFINE(struct callout, nd6_timer_ch); 137 138 void 139 nd6_init(void) 140 { 141 int i; 142 143 LIST_INIT(&V_nd_prefix); 144 145 all1_sa.sin6_family = AF_INET6; 146 all1_sa.sin6_len = sizeof(struct sockaddr_in6); 147 for (i = 0; i < sizeof(all1_sa.sin6_addr); i++) 148 all1_sa.sin6_addr.s6_addr[i] = 0xff; 149 150 /* initialization of the default router list */ 151 TAILQ_INIT(&V_nd_defrouter); 152 153 /* start timer */ 154 callout_init(&V_nd6_slowtimo_ch, 0); 155 callout_reset(&V_nd6_slowtimo_ch, ND6_SLOWTIMER_INTERVAL * hz, 156 nd6_slowtimo, curvnet); 157 } 158 159 #ifdef VIMAGE 160 void 161 nd6_destroy() 162 { 163 164 callout_drain(&V_nd6_slowtimo_ch); 165 callout_drain(&V_nd6_timer_ch); 166 } 167 #endif 168 169 struct nd_ifinfo * 170 nd6_ifattach(struct ifnet *ifp) 171 { 172 struct nd_ifinfo *nd; 173 174 nd = (struct nd_ifinfo *)malloc(sizeof(*nd), M_IP6NDP, M_WAITOK); 175 bzero(nd, sizeof(*nd)); 176 177 nd->initialized = 1; 178 179 nd->chlim = IPV6_DEFHLIM; 180 nd->basereachable = REACHABLE_TIME; 181 nd->reachable = ND_COMPUTE_RTIME(nd->basereachable); 182 nd->retrans = RETRANS_TIMER; 183 184 nd->flags = ND6_IFF_PERFORMNUD; 185 186 /* A loopback interface always has ND6_IFF_AUTO_LINKLOCAL. */ 187 if (V_ip6_auto_linklocal || (ifp->if_flags & IFF_LOOPBACK)) 188 nd->flags |= ND6_IFF_AUTO_LINKLOCAL; 189 190 /* A loopback interface does not need to accept RTADV. */ 191 if (V_ip6_accept_rtadv && !(ifp->if_flags & IFF_LOOPBACK)) 192 nd->flags |= ND6_IFF_ACCEPT_RTADV; 193 194 /* XXX: we cannot call nd6_setmtu since ifp is not fully initialized */ 195 nd6_setmtu0(ifp, nd); 196 197 return nd; 198 } 199 200 void 201 nd6_ifdetach(struct nd_ifinfo *nd) 202 { 203 204 free(nd, M_IP6NDP); 205 } 206 207 /* 208 * Reset ND level link MTU. This function is called when the physical MTU 209 * changes, which means we might have to adjust the ND level MTU. 210 */ 211 void 212 nd6_setmtu(struct ifnet *ifp) 213 { 214 215 nd6_setmtu0(ifp, ND_IFINFO(ifp)); 216 } 217 218 /* XXX todo: do not maintain copy of ifp->if_mtu in ndi->maxmtu */ 219 void 220 nd6_setmtu0(struct ifnet *ifp, struct nd_ifinfo *ndi) 221 { 222 u_int32_t omaxmtu; 223 224 omaxmtu = ndi->maxmtu; 225 226 switch (ifp->if_type) { 227 case IFT_ARCNET: 228 ndi->maxmtu = MIN(ARC_PHDS_MAXMTU, ifp->if_mtu); /* RFC2497 */ 229 break; 230 case IFT_FDDI: 231 ndi->maxmtu = MIN(FDDIIPMTU, ifp->if_mtu); /* RFC2467 */ 232 break; 233 case IFT_ISO88025: 234 ndi->maxmtu = MIN(ISO88025_MAX_MTU, ifp->if_mtu); 235 break; 236 default: 237 ndi->maxmtu = ifp->if_mtu; 238 break; 239 } 240 241 /* 242 * Decreasing the interface MTU under IPV6 minimum MTU may cause 243 * undesirable situation. We thus notify the operator of the change 244 * explicitly. The check for omaxmtu is necessary to restrict the 245 * log to the case of changing the MTU, not initializing it. 246 */ 247 if (omaxmtu >= IPV6_MMTU && ndi->maxmtu < IPV6_MMTU) { 248 log(LOG_NOTICE, "nd6_setmtu0: " 249 "new link MTU on %s (%lu) is too small for IPv6\n", 250 if_name(ifp), (unsigned long)ndi->maxmtu); 251 } 252 253 if (ndi->maxmtu > V_in6_maxmtu) 254 in6_setmaxmtu(); /* check all interfaces just in case */ 255 256 } 257 258 void 259 nd6_option_init(void *opt, int icmp6len, union nd_opts *ndopts) 260 { 261 262 bzero(ndopts, sizeof(*ndopts)); 263 ndopts->nd_opts_search = (struct nd_opt_hdr *)opt; 264 ndopts->nd_opts_last 265 = (struct nd_opt_hdr *)(((u_char *)opt) + icmp6len); 266 267 if (icmp6len == 0) { 268 ndopts->nd_opts_done = 1; 269 ndopts->nd_opts_search = NULL; 270 } 271 } 272 273 /* 274 * Take one ND option. 275 */ 276 struct nd_opt_hdr * 277 nd6_option(union nd_opts *ndopts) 278 { 279 struct nd_opt_hdr *nd_opt; 280 int olen; 281 282 if (ndopts == NULL) 283 panic("ndopts == NULL in nd6_option"); 284 if (ndopts->nd_opts_last == NULL) 285 panic("uninitialized ndopts in nd6_option"); 286 if (ndopts->nd_opts_search == NULL) 287 return NULL; 288 if (ndopts->nd_opts_done) 289 return NULL; 290 291 nd_opt = ndopts->nd_opts_search; 292 293 /* make sure nd_opt_len is inside the buffer */ 294 if ((caddr_t)&nd_opt->nd_opt_len >= (caddr_t)ndopts->nd_opts_last) { 295 bzero(ndopts, sizeof(*ndopts)); 296 return NULL; 297 } 298 299 olen = nd_opt->nd_opt_len << 3; 300 if (olen == 0) { 301 /* 302 * Message validation requires that all included 303 * options have a length that is greater than zero. 304 */ 305 bzero(ndopts, sizeof(*ndopts)); 306 return NULL; 307 } 308 309 ndopts->nd_opts_search = (struct nd_opt_hdr *)((caddr_t)nd_opt + olen); 310 if (ndopts->nd_opts_search > ndopts->nd_opts_last) { 311 /* option overruns the end of buffer, invalid */ 312 bzero(ndopts, sizeof(*ndopts)); 313 return NULL; 314 } else if (ndopts->nd_opts_search == ndopts->nd_opts_last) { 315 /* reached the end of options chain */ 316 ndopts->nd_opts_done = 1; 317 ndopts->nd_opts_search = NULL; 318 } 319 return nd_opt; 320 } 321 322 /* 323 * Parse multiple ND options. 324 * This function is much easier to use, for ND routines that do not need 325 * multiple options of the same type. 326 */ 327 int 328 nd6_options(union nd_opts *ndopts) 329 { 330 struct nd_opt_hdr *nd_opt; 331 int i = 0; 332 333 if (ndopts == NULL) 334 panic("ndopts == NULL in nd6_options"); 335 if (ndopts->nd_opts_last == NULL) 336 panic("uninitialized ndopts in nd6_options"); 337 if (ndopts->nd_opts_search == NULL) 338 return 0; 339 340 while (1) { 341 nd_opt = nd6_option(ndopts); 342 if (nd_opt == NULL && ndopts->nd_opts_last == NULL) { 343 /* 344 * Message validation requires that all included 345 * options have a length that is greater than zero. 346 */ 347 ICMP6STAT_INC(icp6s_nd_badopt); 348 bzero(ndopts, sizeof(*ndopts)); 349 return -1; 350 } 351 352 if (nd_opt == NULL) 353 goto skip1; 354 355 switch (nd_opt->nd_opt_type) { 356 case ND_OPT_SOURCE_LINKADDR: 357 case ND_OPT_TARGET_LINKADDR: 358 case ND_OPT_MTU: 359 case ND_OPT_REDIRECTED_HEADER: 360 if (ndopts->nd_opt_array[nd_opt->nd_opt_type]) { 361 nd6log((LOG_INFO, 362 "duplicated ND6 option found (type=%d)\n", 363 nd_opt->nd_opt_type)); 364 /* XXX bark? */ 365 } else { 366 ndopts->nd_opt_array[nd_opt->nd_opt_type] 367 = nd_opt; 368 } 369 break; 370 case ND_OPT_PREFIX_INFORMATION: 371 if (ndopts->nd_opt_array[nd_opt->nd_opt_type] == 0) { 372 ndopts->nd_opt_array[nd_opt->nd_opt_type] 373 = nd_opt; 374 } 375 ndopts->nd_opts_pi_end = 376 (struct nd_opt_prefix_info *)nd_opt; 377 break; 378 default: 379 /* 380 * Unknown options must be silently ignored, 381 * to accomodate future extension to the protocol. 382 */ 383 nd6log((LOG_DEBUG, 384 "nd6_options: unsupported option %d - " 385 "option ignored\n", nd_opt->nd_opt_type)); 386 } 387 388 skip1: 389 i++; 390 if (i > V_nd6_maxndopt) { 391 ICMP6STAT_INC(icp6s_nd_toomanyopt); 392 nd6log((LOG_INFO, "too many loop in nd opt\n")); 393 break; 394 } 395 396 if (ndopts->nd_opts_done) 397 break; 398 } 399 400 return 0; 401 } 402 403 /* 404 * ND6 timer routine to handle ND6 entries 405 */ 406 void 407 nd6_llinfo_settimer_locked(struct llentry *ln, long tick) 408 { 409 int canceled; 410 411 if (tick < 0) { 412 ln->la_expire = 0; 413 ln->ln_ntick = 0; 414 canceled = callout_stop(&ln->ln_timer_ch); 415 } else { 416 ln->la_expire = time_second + tick / hz; 417 LLE_ADDREF(ln); 418 if (tick > INT_MAX) { 419 ln->ln_ntick = tick - INT_MAX; 420 canceled = callout_reset(&ln->ln_timer_ch, INT_MAX, 421 nd6_llinfo_timer, ln); 422 } else { 423 ln->ln_ntick = 0; 424 canceled = callout_reset(&ln->ln_timer_ch, tick, 425 nd6_llinfo_timer, ln); 426 } 427 } 428 if (canceled) 429 LLE_REMREF(ln); 430 } 431 432 void 433 nd6_llinfo_settimer(struct llentry *ln, long tick) 434 { 435 436 LLE_WLOCK(ln); 437 nd6_llinfo_settimer_locked(ln, tick); 438 LLE_WUNLOCK(ln); 439 } 440 441 static void 442 nd6_llinfo_timer(void *arg) 443 { 444 struct llentry *ln; 445 struct in6_addr *dst; 446 struct ifnet *ifp; 447 struct nd_ifinfo *ndi = NULL; 448 449 ln = (struct llentry *)arg; 450 if (ln == NULL) { 451 panic("%s: NULL entry!\n", __func__); 452 return; 453 } 454 455 if ((ifp = ((ln->lle_tbl != NULL) ? ln->lle_tbl->llt_ifp : NULL)) == NULL) 456 panic("ln ifp == NULL"); 457 458 CURVNET_SET(ifp->if_vnet); 459 460 if (ln->ln_ntick > 0) { 461 if (ln->ln_ntick > INT_MAX) { 462 ln->ln_ntick -= INT_MAX; 463 nd6_llinfo_settimer(ln, INT_MAX); 464 } else { 465 ln->ln_ntick = 0; 466 nd6_llinfo_settimer(ln, ln->ln_ntick); 467 } 468 goto done; 469 } 470 471 ndi = ND_IFINFO(ifp); 472 dst = &L3_ADDR_SIN6(ln)->sin6_addr; 473 if (ln->la_flags & LLE_STATIC) { 474 goto done; 475 } 476 477 if (ln->la_flags & LLE_DELETED) { 478 (void)nd6_free(ln, 0); 479 ln = NULL; 480 goto done; 481 } 482 483 switch (ln->ln_state) { 484 case ND6_LLINFO_INCOMPLETE: 485 if (ln->la_asked < V_nd6_mmaxtries) { 486 ln->la_asked++; 487 nd6_llinfo_settimer(ln, (long)ndi->retrans * hz / 1000); 488 nd6_ns_output(ifp, NULL, dst, ln, 0); 489 } else { 490 struct mbuf *m = ln->la_hold; 491 if (m) { 492 struct mbuf *m0; 493 494 /* 495 * assuming every packet in la_hold has the 496 * same IP header 497 */ 498 m0 = m->m_nextpkt; 499 m->m_nextpkt = NULL; 500 icmp6_error2(m, ICMP6_DST_UNREACH, 501 ICMP6_DST_UNREACH_ADDR, 0, ifp); 502 503 ln->la_hold = m0; 504 clear_llinfo_pqueue(ln); 505 } 506 (void)nd6_free(ln, 0); 507 ln = NULL; 508 } 509 break; 510 case ND6_LLINFO_REACHABLE: 511 if (!ND6_LLINFO_PERMANENT(ln)) { 512 ln->ln_state = ND6_LLINFO_STALE; 513 nd6_llinfo_settimer(ln, (long)V_nd6_gctimer * hz); 514 } 515 break; 516 517 case ND6_LLINFO_STALE: 518 /* Garbage Collection(RFC 2461 5.3) */ 519 if (!ND6_LLINFO_PERMANENT(ln)) { 520 (void)nd6_free(ln, 1); 521 ln = NULL; 522 } 523 break; 524 525 case ND6_LLINFO_DELAY: 526 if (ndi && (ndi->flags & ND6_IFF_PERFORMNUD) != 0) { 527 /* We need NUD */ 528 ln->la_asked = 1; 529 ln->ln_state = ND6_LLINFO_PROBE; 530 nd6_llinfo_settimer(ln, (long)ndi->retrans * hz / 1000); 531 nd6_ns_output(ifp, dst, dst, ln, 0); 532 } else { 533 ln->ln_state = ND6_LLINFO_STALE; /* XXX */ 534 nd6_llinfo_settimer(ln, (long)V_nd6_gctimer * hz); 535 } 536 break; 537 case ND6_LLINFO_PROBE: 538 if (ln->la_asked < V_nd6_umaxtries) { 539 ln->la_asked++; 540 nd6_llinfo_settimer(ln, (long)ndi->retrans * hz / 1000); 541 nd6_ns_output(ifp, dst, dst, ln, 0); 542 } else { 543 (void)nd6_free(ln, 0); 544 ln = NULL; 545 } 546 break; 547 } 548 done: 549 if (ln != NULL) 550 LLE_FREE(ln); 551 CURVNET_RESTORE(); 552 } 553 554 555 /* 556 * ND6 timer routine to expire default route list and prefix list 557 */ 558 void 559 nd6_timer(void *arg) 560 { 561 CURVNET_SET((struct vnet *) arg); 562 int s; 563 struct nd_defrouter *dr; 564 struct nd_prefix *pr; 565 struct in6_ifaddr *ia6, *nia6; 566 struct in6_addrlifetime *lt6; 567 568 callout_reset(&V_nd6_timer_ch, V_nd6_prune * hz, 569 nd6_timer, curvnet); 570 571 /* expire default router list */ 572 s = splnet(); 573 dr = TAILQ_FIRST(&V_nd_defrouter); 574 while (dr) { 575 if (dr->expire && dr->expire < time_second) { 576 struct nd_defrouter *t; 577 t = TAILQ_NEXT(dr, dr_entry); 578 defrtrlist_del(dr); 579 dr = t; 580 } else { 581 dr = TAILQ_NEXT(dr, dr_entry); 582 } 583 } 584 585 /* 586 * expire interface addresses. 587 * in the past the loop was inside prefix expiry processing. 588 * However, from a stricter speci-confrmance standpoint, we should 589 * rather separate address lifetimes and prefix lifetimes. 590 * 591 * XXXRW: in6_ifaddrhead locking. 592 */ 593 addrloop: 594 TAILQ_FOREACH_SAFE(ia6, &V_in6_ifaddrhead, ia_link, nia6) { 595 /* check address lifetime */ 596 lt6 = &ia6->ia6_lifetime; 597 if (IFA6_IS_INVALID(ia6)) { 598 int regen = 0; 599 600 /* 601 * If the expiring address is temporary, try 602 * regenerating a new one. This would be useful when 603 * we suspended a laptop PC, then turned it on after a 604 * period that could invalidate all temporary 605 * addresses. Although we may have to restart the 606 * loop (see below), it must be after purging the 607 * address. Otherwise, we'd see an infinite loop of 608 * regeneration. 609 */ 610 if (V_ip6_use_tempaddr && 611 (ia6->ia6_flags & IN6_IFF_TEMPORARY) != 0) { 612 if (regen_tmpaddr(ia6) == 0) 613 regen = 1; 614 } 615 616 in6_purgeaddr(&ia6->ia_ifa); 617 618 if (regen) 619 goto addrloop; /* XXX: see below */ 620 } else if (IFA6_IS_DEPRECATED(ia6)) { 621 int oldflags = ia6->ia6_flags; 622 623 ia6->ia6_flags |= IN6_IFF_DEPRECATED; 624 625 /* 626 * If a temporary address has just become deprecated, 627 * regenerate a new one if possible. 628 */ 629 if (V_ip6_use_tempaddr && 630 (ia6->ia6_flags & IN6_IFF_TEMPORARY) != 0 && 631 (oldflags & IN6_IFF_DEPRECATED) == 0) { 632 633 if (regen_tmpaddr(ia6) == 0) { 634 /* 635 * A new temporary address is 636 * generated. 637 * XXX: this means the address chain 638 * has changed while we are still in 639 * the loop. Although the change 640 * would not cause disaster (because 641 * it's not a deletion, but an 642 * addition,) we'd rather restart the 643 * loop just for safety. Or does this 644 * significantly reduce performance?? 645 */ 646 goto addrloop; 647 } 648 } 649 } else { 650 /* 651 * A new RA might have made a deprecated address 652 * preferred. 653 */ 654 ia6->ia6_flags &= ~IN6_IFF_DEPRECATED; 655 } 656 } 657 658 /* expire prefix list */ 659 pr = V_nd_prefix.lh_first; 660 while (pr) { 661 /* 662 * check prefix lifetime. 663 * since pltime is just for autoconf, pltime processing for 664 * prefix is not necessary. 665 */ 666 if (pr->ndpr_vltime != ND6_INFINITE_LIFETIME && 667 time_second - pr->ndpr_lastupdate > pr->ndpr_vltime) { 668 struct nd_prefix *t; 669 t = pr->ndpr_next; 670 671 /* 672 * address expiration and prefix expiration are 673 * separate. NEVER perform in6_purgeaddr here. 674 */ 675 676 prelist_remove(pr); 677 pr = t; 678 } else 679 pr = pr->ndpr_next; 680 } 681 splx(s); 682 CURVNET_RESTORE(); 683 } 684 685 /* 686 * ia6 - deprecated/invalidated temporary address 687 */ 688 static int 689 regen_tmpaddr(struct in6_ifaddr *ia6) 690 { 691 struct ifaddr *ifa; 692 struct ifnet *ifp; 693 struct in6_ifaddr *public_ifa6 = NULL; 694 695 ifp = ia6->ia_ifa.ifa_ifp; 696 IF_ADDR_LOCK(ifp); 697 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 698 struct in6_ifaddr *it6; 699 700 if (ifa->ifa_addr->sa_family != AF_INET6) 701 continue; 702 703 it6 = (struct in6_ifaddr *)ifa; 704 705 /* ignore no autoconf addresses. */ 706 if ((it6->ia6_flags & IN6_IFF_AUTOCONF) == 0) 707 continue; 708 709 /* ignore autoconf addresses with different prefixes. */ 710 if (it6->ia6_ndpr == NULL || it6->ia6_ndpr != ia6->ia6_ndpr) 711 continue; 712 713 /* 714 * Now we are looking at an autoconf address with the same 715 * prefix as ours. If the address is temporary and is still 716 * preferred, do not create another one. It would be rare, but 717 * could happen, for example, when we resume a laptop PC after 718 * a long period. 719 */ 720 if ((it6->ia6_flags & IN6_IFF_TEMPORARY) != 0 && 721 !IFA6_IS_DEPRECATED(it6)) { 722 public_ifa6 = NULL; 723 break; 724 } 725 726 /* 727 * This is a public autoconf address that has the same prefix 728 * as ours. If it is preferred, keep it. We can't break the 729 * loop here, because there may be a still-preferred temporary 730 * address with the prefix. 731 */ 732 if (!IFA6_IS_DEPRECATED(it6)) 733 public_ifa6 = it6; 734 735 if (public_ifa6 != NULL) 736 ifa_ref(&public_ifa6->ia_ifa); 737 } 738 IF_ADDR_UNLOCK(ifp); 739 740 if (public_ifa6 != NULL) { 741 int e; 742 743 if ((e = in6_tmpifadd(public_ifa6, 0, 0)) != 0) { 744 ifa_free(&public_ifa6->ia_ifa); 745 log(LOG_NOTICE, "regen_tmpaddr: failed to create a new" 746 " tmp addr,errno=%d\n", e); 747 return (-1); 748 } 749 ifa_free(&public_ifa6->ia_ifa); 750 return (0); 751 } 752 753 return (-1); 754 } 755 756 /* 757 * Nuke neighbor cache/prefix/default router management table, right before 758 * ifp goes away. 759 */ 760 void 761 nd6_purge(struct ifnet *ifp) 762 { 763 struct nd_defrouter *dr, *ndr; 764 struct nd_prefix *pr, *npr; 765 766 /* 767 * Nuke default router list entries toward ifp. 768 * We defer removal of default router list entries that is installed 769 * in the routing table, in order to keep additional side effects as 770 * small as possible. 771 */ 772 for (dr = TAILQ_FIRST(&V_nd_defrouter); dr; dr = ndr) { 773 ndr = TAILQ_NEXT(dr, dr_entry); 774 if (dr->installed) 775 continue; 776 777 if (dr->ifp == ifp) 778 defrtrlist_del(dr); 779 } 780 781 for (dr = TAILQ_FIRST(&V_nd_defrouter); dr; dr = ndr) { 782 ndr = TAILQ_NEXT(dr, dr_entry); 783 if (!dr->installed) 784 continue; 785 786 if (dr->ifp == ifp) 787 defrtrlist_del(dr); 788 } 789 790 /* Nuke prefix list entries toward ifp */ 791 for (pr = V_nd_prefix.lh_first; pr; pr = npr) { 792 npr = pr->ndpr_next; 793 if (pr->ndpr_ifp == ifp) { 794 /* 795 * Because if_detach() does *not* release prefixes 796 * while purging addresses the reference count will 797 * still be above zero. We therefore reset it to 798 * make sure that the prefix really gets purged. 799 */ 800 pr->ndpr_refcnt = 0; 801 802 /* 803 * Previously, pr->ndpr_addr is removed as well, 804 * but I strongly believe we don't have to do it. 805 * nd6_purge() is only called from in6_ifdetach(), 806 * which removes all the associated interface addresses 807 * by itself. 808 * (jinmei@kame.net 20010129) 809 */ 810 prelist_remove(pr); 811 } 812 } 813 814 /* cancel default outgoing interface setting */ 815 if (V_nd6_defifindex == ifp->if_index) 816 nd6_setdefaultiface(0); 817 818 if (!V_ip6_forwarding && ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV) { 819 /* Refresh default router list. */ 820 defrouter_select(); 821 } 822 823 /* XXXXX 824 * We do not nuke the neighbor cache entries here any more 825 * because the neighbor cache is kept in if_afdata[AF_INET6]. 826 * nd6_purge() is invoked by in6_ifdetach() which is called 827 * from if_detach() where everything gets purged. So let 828 * in6_domifdetach() do the actual L2 table purging work. 829 */ 830 } 831 832 /* 833 * the caller acquires and releases the lock on the lltbls 834 * Returns the llentry locked 835 */ 836 struct llentry * 837 nd6_lookup(struct in6_addr *addr6, int flags, struct ifnet *ifp) 838 { 839 struct sockaddr_in6 sin6; 840 struct llentry *ln; 841 int llflags = 0; 842 843 bzero(&sin6, sizeof(sin6)); 844 sin6.sin6_len = sizeof(struct sockaddr_in6); 845 sin6.sin6_family = AF_INET6; 846 sin6.sin6_addr = *addr6; 847 848 IF_AFDATA_LOCK_ASSERT(ifp); 849 850 if (flags & ND6_CREATE) 851 llflags |= LLE_CREATE; 852 if (flags & ND6_EXCLUSIVE) 853 llflags |= LLE_EXCLUSIVE; 854 855 ln = lla_lookup(LLTABLE6(ifp), llflags, (struct sockaddr *)&sin6); 856 if ((ln != NULL) && (flags & LLE_CREATE)) { 857 ln->ln_state = ND6_LLINFO_NOSTATE; 858 callout_init(&ln->ln_timer_ch, 0); 859 } 860 861 return (ln); 862 } 863 864 /* 865 * Test whether a given IPv6 address is a neighbor or not, ignoring 866 * the actual neighbor cache. The neighbor cache is ignored in order 867 * to not reenter the routing code from within itself. 868 */ 869 static int 870 nd6_is_new_addr_neighbor(struct sockaddr_in6 *addr, struct ifnet *ifp) 871 { 872 struct nd_prefix *pr; 873 struct ifaddr *dstaddr; 874 875 /* 876 * A link-local address is always a neighbor. 877 * XXX: a link does not necessarily specify a single interface. 878 */ 879 if (IN6_IS_ADDR_LINKLOCAL(&addr->sin6_addr)) { 880 struct sockaddr_in6 sin6_copy; 881 u_int32_t zone; 882 883 /* 884 * We need sin6_copy since sa6_recoverscope() may modify the 885 * content (XXX). 886 */ 887 sin6_copy = *addr; 888 if (sa6_recoverscope(&sin6_copy)) 889 return (0); /* XXX: should be impossible */ 890 if (in6_setscope(&sin6_copy.sin6_addr, ifp, &zone)) 891 return (0); 892 if (sin6_copy.sin6_scope_id == zone) 893 return (1); 894 else 895 return (0); 896 } 897 898 /* 899 * If the address matches one of our addresses, 900 * it should be a neighbor. 901 * If the address matches one of our on-link prefixes, it should be a 902 * neighbor. 903 */ 904 for (pr = V_nd_prefix.lh_first; pr; pr = pr->ndpr_next) { 905 if (pr->ndpr_ifp != ifp) 906 continue; 907 908 if (!(pr->ndpr_stateflags & NDPRF_ONLINK)) { 909 struct rtentry *rt; 910 rt = rtalloc1((struct sockaddr *)&pr->ndpr_prefix, 0, 0); 911 if (rt == NULL) 912 continue; 913 /* 914 * This is the case where multiple interfaces 915 * have the same prefix, but only one is installed 916 * into the routing table and that prefix entry 917 * is not the one being examined here. In the case 918 * where RADIX_MPATH is enabled, multiple route 919 * entries (of the same rt_key value) will be 920 * installed because the interface addresses all 921 * differ. 922 */ 923 if (!IN6_ARE_ADDR_EQUAL(&pr->ndpr_prefix.sin6_addr, 924 &((struct sockaddr_in6 *)rt_key(rt))->sin6_addr)) { 925 RTFREE_LOCKED(rt); 926 continue; 927 } 928 RTFREE_LOCKED(rt); 929 } 930 931 if (IN6_ARE_MASKED_ADDR_EQUAL(&pr->ndpr_prefix.sin6_addr, 932 &addr->sin6_addr, &pr->ndpr_mask)) 933 return (1); 934 } 935 936 /* 937 * If the address is assigned on the node of the other side of 938 * a p2p interface, the address should be a neighbor. 939 */ 940 dstaddr = ifa_ifwithdstaddr((struct sockaddr *)addr); 941 if (dstaddr != NULL) { 942 if (dstaddr->ifa_ifp == ifp) { 943 ifa_free(dstaddr); 944 return (1); 945 } 946 ifa_free(dstaddr); 947 } 948 949 /* 950 * If the default router list is empty, all addresses are regarded 951 * as on-link, and thus, as a neighbor. 952 * XXX: we restrict the condition to hosts, because routers usually do 953 * not have the "default router list". 954 */ 955 if (!V_ip6_forwarding && TAILQ_FIRST(&V_nd_defrouter) == NULL && 956 V_nd6_defifindex == ifp->if_index) { 957 return (1); 958 } 959 960 return (0); 961 } 962 963 964 /* 965 * Detect if a given IPv6 address identifies a neighbor on a given link. 966 * XXX: should take care of the destination of a p2p link? 967 */ 968 int 969 nd6_is_addr_neighbor(struct sockaddr_in6 *addr, struct ifnet *ifp) 970 { 971 struct llentry *lle; 972 int rc = 0; 973 974 IF_AFDATA_UNLOCK_ASSERT(ifp); 975 if (nd6_is_new_addr_neighbor(addr, ifp)) 976 return (1); 977 978 /* 979 * Even if the address matches none of our addresses, it might be 980 * in the neighbor cache. 981 */ 982 IF_AFDATA_LOCK(ifp); 983 if ((lle = nd6_lookup(&addr->sin6_addr, 0, ifp)) != NULL) { 984 LLE_RUNLOCK(lle); 985 rc = 1; 986 } 987 IF_AFDATA_UNLOCK(ifp); 988 return (rc); 989 } 990 991 /* 992 * Free an nd6 llinfo entry. 993 * Since the function would cause significant changes in the kernel, DO NOT 994 * make it global, unless you have a strong reason for the change, and are sure 995 * that the change is safe. 996 */ 997 static struct llentry * 998 nd6_free(struct llentry *ln, int gc) 999 { 1000 struct llentry *next; 1001 struct nd_defrouter *dr; 1002 struct ifnet *ifp=NULL; 1003 1004 /* 1005 * we used to have pfctlinput(PRC_HOSTDEAD) here. 1006 * even though it is not harmful, it was not really necessary. 1007 */ 1008 1009 /* cancel timer */ 1010 nd6_llinfo_settimer(ln, -1); 1011 1012 if (!V_ip6_forwarding) { 1013 int s; 1014 s = splnet(); 1015 dr = defrouter_lookup(&L3_ADDR_SIN6(ln)->sin6_addr, ln->lle_tbl->llt_ifp); 1016 1017 if (dr != NULL && dr->expire && 1018 ln->ln_state == ND6_LLINFO_STALE && gc) { 1019 /* 1020 * If the reason for the deletion is just garbage 1021 * collection, and the neighbor is an active default 1022 * router, do not delete it. Instead, reset the GC 1023 * timer using the router's lifetime. 1024 * Simply deleting the entry would affect default 1025 * router selection, which is not necessarily a good 1026 * thing, especially when we're using router preference 1027 * values. 1028 * XXX: the check for ln_state would be redundant, 1029 * but we intentionally keep it just in case. 1030 */ 1031 if (dr->expire > time_second) 1032 nd6_llinfo_settimer(ln, 1033 (dr->expire - time_second) * hz); 1034 else 1035 nd6_llinfo_settimer(ln, (long)V_nd6_gctimer * hz); 1036 splx(s); 1037 LLE_WLOCK(ln); 1038 LLE_REMREF(ln); 1039 LLE_WUNLOCK(ln); 1040 return (LIST_NEXT(ln, lle_next)); 1041 } 1042 1043 if (ln->ln_router || dr) { 1044 /* 1045 * rt6_flush must be called whether or not the neighbor 1046 * is in the Default Router List. 1047 * See a corresponding comment in nd6_na_input(). 1048 */ 1049 rt6_flush(&L3_ADDR_SIN6(ln)->sin6_addr, ln->lle_tbl->llt_ifp); 1050 } 1051 1052 if (dr) { 1053 /* 1054 * Unreachablity of a router might affect the default 1055 * router selection and on-link detection of advertised 1056 * prefixes. 1057 */ 1058 1059 /* 1060 * Temporarily fake the state to choose a new default 1061 * router and to perform on-link determination of 1062 * prefixes correctly. 1063 * Below the state will be set correctly, 1064 * or the entry itself will be deleted. 1065 */ 1066 ln->ln_state = ND6_LLINFO_INCOMPLETE; 1067 1068 /* 1069 * Since defrouter_select() does not affect the 1070 * on-link determination and MIP6 needs the check 1071 * before the default router selection, we perform 1072 * the check now. 1073 */ 1074 pfxlist_onlink_check(); 1075 1076 /* 1077 * refresh default router list 1078 */ 1079 defrouter_select(); 1080 } 1081 splx(s); 1082 } 1083 1084 /* 1085 * Before deleting the entry, remember the next entry as the 1086 * return value. We need this because pfxlist_onlink_check() above 1087 * might have freed other entries (particularly the old next entry) as 1088 * a side effect (XXX). 1089 */ 1090 next = LIST_NEXT(ln, lle_next); 1091 1092 ifp = ln->lle_tbl->llt_ifp; 1093 IF_AFDATA_LOCK(ifp); 1094 LLE_WLOCK(ln); 1095 LLE_REMREF(ln); 1096 llentry_free(ln); 1097 IF_AFDATA_UNLOCK(ifp); 1098 1099 return (next); 1100 } 1101 1102 /* 1103 * Upper-layer reachability hint for Neighbor Unreachability Detection. 1104 * 1105 * XXX cost-effective methods? 1106 */ 1107 void 1108 nd6_nud_hint(struct rtentry *rt, struct in6_addr *dst6, int force) 1109 { 1110 struct llentry *ln; 1111 struct ifnet *ifp; 1112 1113 if ((dst6 == NULL) || (rt == NULL)) 1114 return; 1115 1116 ifp = rt->rt_ifp; 1117 IF_AFDATA_LOCK(ifp); 1118 ln = nd6_lookup(dst6, ND6_EXCLUSIVE, NULL); 1119 IF_AFDATA_UNLOCK(ifp); 1120 if (ln == NULL) 1121 return; 1122 1123 if (ln->ln_state < ND6_LLINFO_REACHABLE) 1124 goto done; 1125 1126 /* 1127 * if we get upper-layer reachability confirmation many times, 1128 * it is possible we have false information. 1129 */ 1130 if (!force) { 1131 ln->ln_byhint++; 1132 if (ln->ln_byhint > V_nd6_maxnudhint) { 1133 goto done; 1134 } 1135 } 1136 1137 ln->ln_state = ND6_LLINFO_REACHABLE; 1138 if (!ND6_LLINFO_PERMANENT(ln)) { 1139 nd6_llinfo_settimer_locked(ln, 1140 (long)ND_IFINFO(rt->rt_ifp)->reachable * hz); 1141 } 1142 done: 1143 LLE_WUNLOCK(ln); 1144 } 1145 1146 1147 int 1148 nd6_ioctl(u_long cmd, caddr_t data, struct ifnet *ifp) 1149 { 1150 struct in6_drlist *drl = (struct in6_drlist *)data; 1151 struct in6_oprlist *oprl = (struct in6_oprlist *)data; 1152 struct in6_ndireq *ndi = (struct in6_ndireq *)data; 1153 struct in6_nbrinfo *nbi = (struct in6_nbrinfo *)data; 1154 struct in6_ndifreq *ndif = (struct in6_ndifreq *)data; 1155 struct nd_defrouter *dr; 1156 struct nd_prefix *pr; 1157 int i = 0, error = 0; 1158 int s; 1159 1160 switch (cmd) { 1161 case SIOCGDRLST_IN6: 1162 /* 1163 * obsolete API, use sysctl under net.inet6.icmp6 1164 */ 1165 bzero(drl, sizeof(*drl)); 1166 s = splnet(); 1167 dr = TAILQ_FIRST(&V_nd_defrouter); 1168 while (dr && i < DRLSTSIZ) { 1169 drl->defrouter[i].rtaddr = dr->rtaddr; 1170 in6_clearscope(&drl->defrouter[i].rtaddr); 1171 1172 drl->defrouter[i].flags = dr->flags; 1173 drl->defrouter[i].rtlifetime = dr->rtlifetime; 1174 drl->defrouter[i].expire = dr->expire; 1175 drl->defrouter[i].if_index = dr->ifp->if_index; 1176 i++; 1177 dr = TAILQ_NEXT(dr, dr_entry); 1178 } 1179 splx(s); 1180 break; 1181 case SIOCGPRLST_IN6: 1182 /* 1183 * obsolete API, use sysctl under net.inet6.icmp6 1184 * 1185 * XXX the structure in6_prlist was changed in backward- 1186 * incompatible manner. in6_oprlist is used for SIOCGPRLST_IN6, 1187 * in6_prlist is used for nd6_sysctl() - fill_prlist(). 1188 */ 1189 /* 1190 * XXX meaning of fields, especialy "raflags", is very 1191 * differnet between RA prefix list and RR/static prefix list. 1192 * how about separating ioctls into two? 1193 */ 1194 bzero(oprl, sizeof(*oprl)); 1195 s = splnet(); 1196 pr = V_nd_prefix.lh_first; 1197 while (pr && i < PRLSTSIZ) { 1198 struct nd_pfxrouter *pfr; 1199 int j; 1200 1201 oprl->prefix[i].prefix = pr->ndpr_prefix.sin6_addr; 1202 oprl->prefix[i].raflags = pr->ndpr_raf; 1203 oprl->prefix[i].prefixlen = pr->ndpr_plen; 1204 oprl->prefix[i].vltime = pr->ndpr_vltime; 1205 oprl->prefix[i].pltime = pr->ndpr_pltime; 1206 oprl->prefix[i].if_index = pr->ndpr_ifp->if_index; 1207 if (pr->ndpr_vltime == ND6_INFINITE_LIFETIME) 1208 oprl->prefix[i].expire = 0; 1209 else { 1210 time_t maxexpire; 1211 1212 /* XXX: we assume time_t is signed. */ 1213 maxexpire = (-1) & 1214 ~((time_t)1 << 1215 ((sizeof(maxexpire) * 8) - 1)); 1216 if (pr->ndpr_vltime < 1217 maxexpire - pr->ndpr_lastupdate) { 1218 oprl->prefix[i].expire = 1219 pr->ndpr_lastupdate + 1220 pr->ndpr_vltime; 1221 } else 1222 oprl->prefix[i].expire = maxexpire; 1223 } 1224 1225 pfr = pr->ndpr_advrtrs.lh_first; 1226 j = 0; 1227 while (pfr) { 1228 if (j < DRLSTSIZ) { 1229 #define RTRADDR oprl->prefix[i].advrtr[j] 1230 RTRADDR = pfr->router->rtaddr; 1231 in6_clearscope(&RTRADDR); 1232 #undef RTRADDR 1233 } 1234 j++; 1235 pfr = pfr->pfr_next; 1236 } 1237 oprl->prefix[i].advrtrs = j; 1238 oprl->prefix[i].origin = PR_ORIG_RA; 1239 1240 i++; 1241 pr = pr->ndpr_next; 1242 } 1243 splx(s); 1244 1245 break; 1246 case OSIOCGIFINFO_IN6: 1247 #define ND ndi->ndi 1248 /* XXX: old ndp(8) assumes a positive value for linkmtu. */ 1249 bzero(&ND, sizeof(ND)); 1250 ND.linkmtu = IN6_LINKMTU(ifp); 1251 ND.maxmtu = ND_IFINFO(ifp)->maxmtu; 1252 ND.basereachable = ND_IFINFO(ifp)->basereachable; 1253 ND.reachable = ND_IFINFO(ifp)->reachable; 1254 ND.retrans = ND_IFINFO(ifp)->retrans; 1255 ND.flags = ND_IFINFO(ifp)->flags; 1256 ND.recalctm = ND_IFINFO(ifp)->recalctm; 1257 ND.chlim = ND_IFINFO(ifp)->chlim; 1258 break; 1259 case SIOCGIFINFO_IN6: 1260 ND = *ND_IFINFO(ifp); 1261 break; 1262 case SIOCSIFINFO_IN6: 1263 /* 1264 * used to change host variables from userland. 1265 * intented for a use on router to reflect RA configurations. 1266 */ 1267 /* 0 means 'unspecified' */ 1268 if (ND.linkmtu != 0) { 1269 if (ND.linkmtu < IPV6_MMTU || 1270 ND.linkmtu > IN6_LINKMTU(ifp)) { 1271 error = EINVAL; 1272 break; 1273 } 1274 ND_IFINFO(ifp)->linkmtu = ND.linkmtu; 1275 } 1276 1277 if (ND.basereachable != 0) { 1278 int obasereachable = ND_IFINFO(ifp)->basereachable; 1279 1280 ND_IFINFO(ifp)->basereachable = ND.basereachable; 1281 if (ND.basereachable != obasereachable) 1282 ND_IFINFO(ifp)->reachable = 1283 ND_COMPUTE_RTIME(ND.basereachable); 1284 } 1285 if (ND.retrans != 0) 1286 ND_IFINFO(ifp)->retrans = ND.retrans; 1287 if (ND.chlim != 0) 1288 ND_IFINFO(ifp)->chlim = ND.chlim; 1289 /* FALLTHROUGH */ 1290 case SIOCSIFINFO_FLAGS: 1291 { 1292 struct ifaddr *ifa; 1293 struct in6_ifaddr *ia; 1294 1295 if ((ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) && 1296 !(ND.flags & ND6_IFF_IFDISABLED)) { 1297 /* ifdisabled 1->0 transision */ 1298 1299 /* 1300 * If the interface is marked as ND6_IFF_IFDISABLED and 1301 * has an link-local address with IN6_IFF_DUPLICATED, 1302 * do not clear ND6_IFF_IFDISABLED. 1303 * See RFC 4862, Section 5.4.5. 1304 */ 1305 int duplicated_linklocal = 0; 1306 1307 IF_ADDR_LOCK(ifp); 1308 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1309 if (ifa->ifa_addr->sa_family != AF_INET6) 1310 continue; 1311 ia = (struct in6_ifaddr *)ifa; 1312 if ((ia->ia6_flags & IN6_IFF_DUPLICATED) && 1313 IN6_IS_ADDR_LINKLOCAL(&ia->ia_addr.sin6_addr)) { 1314 duplicated_linklocal = 1; 1315 break; 1316 } 1317 } 1318 IF_ADDR_UNLOCK(ifp); 1319 1320 if (duplicated_linklocal) { 1321 ND.flags |= ND6_IFF_IFDISABLED; 1322 log(LOG_ERR, "Cannot enable an interface" 1323 " with a link-local address marked" 1324 " duplicate.\n"); 1325 } else { 1326 ND_IFINFO(ifp)->flags &= ~ND6_IFF_IFDISABLED; 1327 in6_if_up(ifp); 1328 } 1329 } else if (!(ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) && 1330 (ND.flags & ND6_IFF_IFDISABLED)) { 1331 /* ifdisabled 0->1 transision */ 1332 /* Mark all IPv6 address as tentative. */ 1333 1334 ND_IFINFO(ifp)->flags |= ND6_IFF_IFDISABLED; 1335 IF_ADDR_LOCK(ifp); 1336 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1337 if (ifa->ifa_addr->sa_family != AF_INET6) 1338 continue; 1339 ia = (struct in6_ifaddr *)ifa; 1340 ia->ia6_flags |= IN6_IFF_TENTATIVE; 1341 } 1342 IF_ADDR_UNLOCK(ifp); 1343 } 1344 1345 if (!(ND_IFINFO(ifp)->flags & ND6_IFF_AUTO_LINKLOCAL) && 1346 (ND.flags & ND6_IFF_AUTO_LINKLOCAL)) { 1347 /* auto_linklocal 0->1 transision */ 1348 1349 /* If no link-local address on ifp, configure */ 1350 ND_IFINFO(ifp)->flags |= ND6_IFF_AUTO_LINKLOCAL; 1351 in6_ifattach(ifp, NULL); 1352 } 1353 } 1354 ND_IFINFO(ifp)->flags = ND.flags; 1355 break; 1356 #undef ND 1357 case SIOCSNDFLUSH_IN6: /* XXX: the ioctl name is confusing... */ 1358 /* sync kernel routing table with the default router list */ 1359 defrouter_reset(); 1360 defrouter_select(); 1361 break; 1362 case SIOCSPFXFLUSH_IN6: 1363 { 1364 /* flush all the prefix advertised by routers */ 1365 struct nd_prefix *pr, *next; 1366 1367 s = splnet(); 1368 for (pr = V_nd_prefix.lh_first; pr; pr = next) { 1369 struct in6_ifaddr *ia, *ia_next; 1370 1371 next = pr->ndpr_next; 1372 1373 if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr)) 1374 continue; /* XXX */ 1375 1376 /* do we really have to remove addresses as well? */ 1377 /* XXXRW: in6_ifaddrhead locking. */ 1378 TAILQ_FOREACH_SAFE(ia, &V_in6_ifaddrhead, ia_link, 1379 ia_next) { 1380 if ((ia->ia6_flags & IN6_IFF_AUTOCONF) == 0) 1381 continue; 1382 1383 if (ia->ia6_ndpr == pr) 1384 in6_purgeaddr(&ia->ia_ifa); 1385 } 1386 prelist_remove(pr); 1387 } 1388 splx(s); 1389 break; 1390 } 1391 case SIOCSRTRFLUSH_IN6: 1392 { 1393 /* flush all the default routers */ 1394 struct nd_defrouter *dr, *next; 1395 1396 s = splnet(); 1397 defrouter_reset(); 1398 for (dr = TAILQ_FIRST(&V_nd_defrouter); dr; dr = next) { 1399 next = TAILQ_NEXT(dr, dr_entry); 1400 defrtrlist_del(dr); 1401 } 1402 defrouter_select(); 1403 splx(s); 1404 break; 1405 } 1406 case SIOCGNBRINFO_IN6: 1407 { 1408 struct llentry *ln; 1409 struct in6_addr nb_addr = nbi->addr; /* make local for safety */ 1410 1411 if ((error = in6_setscope(&nb_addr, ifp, NULL)) != 0) 1412 return (error); 1413 1414 IF_AFDATA_LOCK(ifp); 1415 ln = nd6_lookup(&nb_addr, 0, ifp); 1416 IF_AFDATA_UNLOCK(ifp); 1417 1418 if (ln == NULL) { 1419 error = EINVAL; 1420 break; 1421 } 1422 nbi->state = ln->ln_state; 1423 nbi->asked = ln->la_asked; 1424 nbi->isrouter = ln->ln_router; 1425 nbi->expire = ln->la_expire; 1426 LLE_RUNLOCK(ln); 1427 break; 1428 } 1429 case SIOCGDEFIFACE_IN6: /* XXX: should be implemented as a sysctl? */ 1430 ndif->ifindex = V_nd6_defifindex; 1431 break; 1432 case SIOCSDEFIFACE_IN6: /* XXX: should be implemented as a sysctl? */ 1433 return (nd6_setdefaultiface(ndif->ifindex)); 1434 } 1435 return (error); 1436 } 1437 1438 /* 1439 * Create neighbor cache entry and cache link-layer address, 1440 * on reception of inbound ND6 packets. (RS/RA/NS/redirect) 1441 * 1442 * type - ICMP6 type 1443 * code - type dependent information 1444 * 1445 * XXXXX 1446 * The caller of this function already acquired the ndp 1447 * cache table lock because the cache entry is returned. 1448 */ 1449 struct llentry * 1450 nd6_cache_lladdr(struct ifnet *ifp, struct in6_addr *from, char *lladdr, 1451 int lladdrlen, int type, int code) 1452 { 1453 struct llentry *ln = NULL; 1454 int is_newentry; 1455 int do_update; 1456 int olladdr; 1457 int llchange; 1458 int flags = 0; 1459 int newstate = 0; 1460 uint16_t router = 0; 1461 struct sockaddr_in6 sin6; 1462 struct mbuf *chain = NULL; 1463 int static_route = 0; 1464 1465 IF_AFDATA_UNLOCK_ASSERT(ifp); 1466 1467 if (ifp == NULL) 1468 panic("ifp == NULL in nd6_cache_lladdr"); 1469 if (from == NULL) 1470 panic("from == NULL in nd6_cache_lladdr"); 1471 1472 /* nothing must be updated for unspecified address */ 1473 if (IN6_IS_ADDR_UNSPECIFIED(from)) 1474 return NULL; 1475 1476 /* 1477 * Validation about ifp->if_addrlen and lladdrlen must be done in 1478 * the caller. 1479 * 1480 * XXX If the link does not have link-layer adderss, what should 1481 * we do? (ifp->if_addrlen == 0) 1482 * Spec says nothing in sections for RA, RS and NA. There's small 1483 * description on it in NS section (RFC 2461 7.2.3). 1484 */ 1485 flags |= lladdr ? ND6_EXCLUSIVE : 0; 1486 IF_AFDATA_LOCK(ifp); 1487 ln = nd6_lookup(from, flags, ifp); 1488 1489 if (ln == NULL) { 1490 flags |= LLE_EXCLUSIVE; 1491 ln = nd6_lookup(from, flags |ND6_CREATE, ifp); 1492 IF_AFDATA_UNLOCK(ifp); 1493 is_newentry = 1; 1494 } else { 1495 IF_AFDATA_UNLOCK(ifp); 1496 /* do nothing if static ndp is set */ 1497 if (ln->la_flags & LLE_STATIC) { 1498 static_route = 1; 1499 goto done; 1500 } 1501 is_newentry = 0; 1502 } 1503 if (ln == NULL) 1504 return (NULL); 1505 1506 olladdr = (ln->la_flags & LLE_VALID) ? 1 : 0; 1507 if (olladdr && lladdr) { 1508 llchange = bcmp(lladdr, &ln->ll_addr, 1509 ifp->if_addrlen); 1510 } else 1511 llchange = 0; 1512 1513 /* 1514 * newentry olladdr lladdr llchange (*=record) 1515 * 0 n n -- (1) 1516 * 0 y n -- (2) 1517 * 0 n y -- (3) * STALE 1518 * 0 y y n (4) * 1519 * 0 y y y (5) * STALE 1520 * 1 -- n -- (6) NOSTATE(= PASSIVE) 1521 * 1 -- y -- (7) * STALE 1522 */ 1523 1524 if (lladdr) { /* (3-5) and (7) */ 1525 /* 1526 * Record source link-layer address 1527 * XXX is it dependent to ifp->if_type? 1528 */ 1529 bcopy(lladdr, &ln->ll_addr, ifp->if_addrlen); 1530 ln->la_flags |= LLE_VALID; 1531 } 1532 1533 if (!is_newentry) { 1534 if ((!olladdr && lladdr != NULL) || /* (3) */ 1535 (olladdr && lladdr != NULL && llchange)) { /* (5) */ 1536 do_update = 1; 1537 newstate = ND6_LLINFO_STALE; 1538 } else /* (1-2,4) */ 1539 do_update = 0; 1540 } else { 1541 do_update = 1; 1542 if (lladdr == NULL) /* (6) */ 1543 newstate = ND6_LLINFO_NOSTATE; 1544 else /* (7) */ 1545 newstate = ND6_LLINFO_STALE; 1546 } 1547 1548 if (do_update) { 1549 /* 1550 * Update the state of the neighbor cache. 1551 */ 1552 ln->ln_state = newstate; 1553 1554 if (ln->ln_state == ND6_LLINFO_STALE) { 1555 /* 1556 * XXX: since nd6_output() below will cause 1557 * state tansition to DELAY and reset the timer, 1558 * we must set the timer now, although it is actually 1559 * meaningless. 1560 */ 1561 nd6_llinfo_settimer_locked(ln, (long)V_nd6_gctimer * hz); 1562 1563 if (ln->la_hold) { 1564 struct mbuf *m_hold, *m_hold_next; 1565 1566 /* 1567 * reset the la_hold in advance, to explicitly 1568 * prevent a la_hold lookup in nd6_output() 1569 * (wouldn't happen, though...) 1570 */ 1571 for (m_hold = ln->la_hold, ln->la_hold = NULL; 1572 m_hold; m_hold = m_hold_next) { 1573 m_hold_next = m_hold->m_nextpkt; 1574 m_hold->m_nextpkt = NULL; 1575 1576 /* 1577 * we assume ifp is not a p2p here, so 1578 * just set the 2nd argument as the 1579 * 1st one. 1580 */ 1581 nd6_output_lle(ifp, ifp, m_hold, L3_ADDR_SIN6(ln), NULL, ln, &chain); 1582 } 1583 /* 1584 * If we have mbufs in the chain we need to do 1585 * deferred transmit. Copy the address from the 1586 * llentry before dropping the lock down below. 1587 */ 1588 if (chain != NULL) 1589 memcpy(&sin6, L3_ADDR_SIN6(ln), sizeof(sin6)); 1590 } 1591 } else if (ln->ln_state == ND6_LLINFO_INCOMPLETE) { 1592 /* probe right away */ 1593 nd6_llinfo_settimer_locked((void *)ln, 0); 1594 } 1595 } 1596 1597 /* 1598 * ICMP6 type dependent behavior. 1599 * 1600 * NS: clear IsRouter if new entry 1601 * RS: clear IsRouter 1602 * RA: set IsRouter if there's lladdr 1603 * redir: clear IsRouter if new entry 1604 * 1605 * RA case, (1): 1606 * The spec says that we must set IsRouter in the following cases: 1607 * - If lladdr exist, set IsRouter. This means (1-5). 1608 * - If it is old entry (!newentry), set IsRouter. This means (7). 1609 * So, based on the spec, in (1-5) and (7) cases we must set IsRouter. 1610 * A quetion arises for (1) case. (1) case has no lladdr in the 1611 * neighbor cache, this is similar to (6). 1612 * This case is rare but we figured that we MUST NOT set IsRouter. 1613 * 1614 * newentry olladdr lladdr llchange NS RS RA redir 1615 * D R 1616 * 0 n n -- (1) c ? s 1617 * 0 y n -- (2) c s s 1618 * 0 n y -- (3) c s s 1619 * 0 y y n (4) c s s 1620 * 0 y y y (5) c s s 1621 * 1 -- n -- (6) c c c s 1622 * 1 -- y -- (7) c c s c s 1623 * 1624 * (c=clear s=set) 1625 */ 1626 switch (type & 0xff) { 1627 case ND_NEIGHBOR_SOLICIT: 1628 /* 1629 * New entry must have is_router flag cleared. 1630 */ 1631 if (is_newentry) /* (6-7) */ 1632 ln->ln_router = 0; 1633 break; 1634 case ND_REDIRECT: 1635 /* 1636 * If the icmp is a redirect to a better router, always set the 1637 * is_router flag. Otherwise, if the entry is newly created, 1638 * clear the flag. [RFC 2461, sec 8.3] 1639 */ 1640 if (code == ND_REDIRECT_ROUTER) 1641 ln->ln_router = 1; 1642 else if (is_newentry) /* (6-7) */ 1643 ln->ln_router = 0; 1644 break; 1645 case ND_ROUTER_SOLICIT: 1646 /* 1647 * is_router flag must always be cleared. 1648 */ 1649 ln->ln_router = 0; 1650 break; 1651 case ND_ROUTER_ADVERT: 1652 /* 1653 * Mark an entry with lladdr as a router. 1654 */ 1655 if ((!is_newentry && (olladdr || lladdr)) || /* (2-5) */ 1656 (is_newentry && lladdr)) { /* (7) */ 1657 ln->ln_router = 1; 1658 } 1659 break; 1660 } 1661 1662 if (ln != NULL) { 1663 static_route = (ln->la_flags & LLE_STATIC); 1664 router = ln->ln_router; 1665 1666 if (flags & ND6_EXCLUSIVE) 1667 LLE_WUNLOCK(ln); 1668 else 1669 LLE_RUNLOCK(ln); 1670 if (static_route) 1671 ln = NULL; 1672 } 1673 if (chain) 1674 nd6_output_flush(ifp, ifp, chain, &sin6, NULL); 1675 1676 /* 1677 * When the link-layer address of a router changes, select the 1678 * best router again. In particular, when the neighbor entry is newly 1679 * created, it might affect the selection policy. 1680 * Question: can we restrict the first condition to the "is_newentry" 1681 * case? 1682 * XXX: when we hear an RA from a new router with the link-layer 1683 * address option, defrouter_select() is called twice, since 1684 * defrtrlist_update called the function as well. However, I believe 1685 * we can compromise the overhead, since it only happens the first 1686 * time. 1687 * XXX: although defrouter_select() should not have a bad effect 1688 * for those are not autoconfigured hosts, we explicitly avoid such 1689 * cases for safety. 1690 */ 1691 if (do_update && router && !V_ip6_forwarding && 1692 ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV) { 1693 /* 1694 * guaranteed recursion 1695 */ 1696 defrouter_select(); 1697 } 1698 1699 return (ln); 1700 done: 1701 if (ln != NULL) { 1702 if (flags & ND6_EXCLUSIVE) 1703 LLE_WUNLOCK(ln); 1704 else 1705 LLE_RUNLOCK(ln); 1706 if (static_route) 1707 ln = NULL; 1708 } 1709 return (ln); 1710 } 1711 1712 static void 1713 nd6_slowtimo(void *arg) 1714 { 1715 CURVNET_SET((struct vnet *) arg); 1716 struct nd_ifinfo *nd6if; 1717 struct ifnet *ifp; 1718 1719 callout_reset(&V_nd6_slowtimo_ch, ND6_SLOWTIMER_INTERVAL * hz, 1720 nd6_slowtimo, curvnet); 1721 IFNET_RLOCK_NOSLEEP(); 1722 for (ifp = TAILQ_FIRST(&V_ifnet); ifp; 1723 ifp = TAILQ_NEXT(ifp, if_list)) { 1724 nd6if = ND_IFINFO(ifp); 1725 if (nd6if->basereachable && /* already initialized */ 1726 (nd6if->recalctm -= ND6_SLOWTIMER_INTERVAL) <= 0) { 1727 /* 1728 * Since reachable time rarely changes by router 1729 * advertisements, we SHOULD insure that a new random 1730 * value gets recomputed at least once every few hours. 1731 * (RFC 2461, 6.3.4) 1732 */ 1733 nd6if->recalctm = V_nd6_recalc_reachtm_interval; 1734 nd6if->reachable = ND_COMPUTE_RTIME(nd6if->basereachable); 1735 } 1736 } 1737 IFNET_RUNLOCK_NOSLEEP(); 1738 CURVNET_RESTORE(); 1739 } 1740 1741 int 1742 nd6_output(struct ifnet *ifp, struct ifnet *origifp, struct mbuf *m0, 1743 struct sockaddr_in6 *dst, struct rtentry *rt0) 1744 { 1745 1746 return (nd6_output_lle(ifp, origifp, m0, dst, rt0, NULL, NULL)); 1747 } 1748 1749 1750 /* 1751 * Note that I'm not enforcing any global serialization 1752 * lle state or asked changes here as the logic is too 1753 * complicated to avoid having to always acquire an exclusive 1754 * lock 1755 * KMM 1756 * 1757 */ 1758 #define senderr(e) { error = (e); goto bad;} 1759 1760 int 1761 nd6_output_lle(struct ifnet *ifp, struct ifnet *origifp, struct mbuf *m0, 1762 struct sockaddr_in6 *dst, struct rtentry *rt0, struct llentry *lle, 1763 struct mbuf **chain) 1764 { 1765 struct mbuf *m = m0; 1766 struct llentry *ln = lle; 1767 int error = 0; 1768 int flags = 0; 1769 1770 #ifdef INVARIANTS 1771 if (lle != NULL) { 1772 1773 LLE_WLOCK_ASSERT(lle); 1774 1775 KASSERT(chain != NULL, (" lle locked but no mbuf chain pointer passed")); 1776 } 1777 #endif 1778 if (IN6_IS_ADDR_MULTICAST(&dst->sin6_addr)) 1779 goto sendpkt; 1780 1781 if (nd6_need_cache(ifp) == 0) 1782 goto sendpkt; 1783 1784 /* 1785 * next hop determination. This routine is derived from ether_output. 1786 */ 1787 1788 /* 1789 * Address resolution or Neighbor Unreachability Detection 1790 * for the next hop. 1791 * At this point, the destination of the packet must be a unicast 1792 * or an anycast address(i.e. not a multicast). 1793 */ 1794 1795 flags = ((m != NULL) || (lle != NULL)) ? LLE_EXCLUSIVE : 0; 1796 if (ln == NULL) { 1797 retry: 1798 IF_AFDATA_LOCK(ifp); 1799 ln = lla_lookup(LLTABLE6(ifp), flags, (struct sockaddr *)dst); 1800 IF_AFDATA_UNLOCK(ifp); 1801 if ((ln == NULL) && nd6_is_addr_neighbor(dst, ifp)) { 1802 /* 1803 * Since nd6_is_addr_neighbor() internally calls nd6_lookup(), 1804 * the condition below is not very efficient. But we believe 1805 * it is tolerable, because this should be a rare case. 1806 */ 1807 flags = ND6_CREATE | (m ? ND6_EXCLUSIVE : 0); 1808 IF_AFDATA_LOCK(ifp); 1809 ln = nd6_lookup(&dst->sin6_addr, flags, ifp); 1810 IF_AFDATA_UNLOCK(ifp); 1811 } 1812 } 1813 if (ln == NULL) { 1814 if ((ifp->if_flags & IFF_POINTOPOINT) == 0 && 1815 !(ND_IFINFO(ifp)->flags & ND6_IFF_PERFORMNUD)) { 1816 char ip6buf[INET6_ADDRSTRLEN]; 1817 log(LOG_DEBUG, 1818 "nd6_output: can't allocate llinfo for %s " 1819 "(ln=%p)\n", 1820 ip6_sprintf(ip6buf, &dst->sin6_addr), ln); 1821 senderr(EIO); /* XXX: good error? */ 1822 } 1823 goto sendpkt; /* send anyway */ 1824 } 1825 1826 /* We don't have to do link-layer address resolution on a p2p link. */ 1827 if ((ifp->if_flags & IFF_POINTOPOINT) != 0 && 1828 ln->ln_state < ND6_LLINFO_REACHABLE) { 1829 if ((flags & LLE_EXCLUSIVE) == 0) { 1830 flags |= LLE_EXCLUSIVE; 1831 goto retry; 1832 } 1833 ln->ln_state = ND6_LLINFO_STALE; 1834 nd6_llinfo_settimer_locked(ln, (long)V_nd6_gctimer * hz); 1835 } 1836 1837 /* 1838 * The first time we send a packet to a neighbor whose entry is 1839 * STALE, we have to change the state to DELAY and a sets a timer to 1840 * expire in DELAY_FIRST_PROBE_TIME seconds to ensure do 1841 * neighbor unreachability detection on expiration. 1842 * (RFC 2461 7.3.3) 1843 */ 1844 if (ln->ln_state == ND6_LLINFO_STALE) { 1845 if ((flags & LLE_EXCLUSIVE) == 0) { 1846 flags |= LLE_EXCLUSIVE; 1847 LLE_RUNLOCK(ln); 1848 goto retry; 1849 } 1850 ln->la_asked = 0; 1851 ln->ln_state = ND6_LLINFO_DELAY; 1852 nd6_llinfo_settimer_locked(ln, (long)V_nd6_delay * hz); 1853 } 1854 1855 /* 1856 * If the neighbor cache entry has a state other than INCOMPLETE 1857 * (i.e. its link-layer address is already resolved), just 1858 * send the packet. 1859 */ 1860 if (ln->ln_state > ND6_LLINFO_INCOMPLETE) 1861 goto sendpkt; 1862 1863 /* 1864 * There is a neighbor cache entry, but no ethernet address 1865 * response yet. Append this latest packet to the end of the 1866 * packet queue in the mbuf, unless the number of the packet 1867 * does not exceed nd6_maxqueuelen. When it exceeds nd6_maxqueuelen, 1868 * the oldest packet in the queue will be removed. 1869 */ 1870 if (ln->ln_state == ND6_LLINFO_NOSTATE) 1871 ln->ln_state = ND6_LLINFO_INCOMPLETE; 1872 1873 if ((flags & LLE_EXCLUSIVE) == 0) { 1874 flags |= LLE_EXCLUSIVE; 1875 LLE_RUNLOCK(ln); 1876 goto retry; 1877 } 1878 if (ln->la_hold) { 1879 struct mbuf *m_hold; 1880 int i; 1881 1882 i = 0; 1883 for (m_hold = ln->la_hold; m_hold; m_hold = m_hold->m_nextpkt) { 1884 i++; 1885 if (m_hold->m_nextpkt == NULL) { 1886 m_hold->m_nextpkt = m; 1887 break; 1888 } 1889 } 1890 while (i >= V_nd6_maxqueuelen) { 1891 m_hold = ln->la_hold; 1892 ln->la_hold = ln->la_hold->m_nextpkt; 1893 m_freem(m_hold); 1894 i--; 1895 } 1896 } else { 1897 ln->la_hold = m; 1898 } 1899 /* 1900 * We did the lookup (no lle arg) so we 1901 * need to do the unlock here 1902 */ 1903 if (lle == NULL) { 1904 if (flags & LLE_EXCLUSIVE) 1905 LLE_WUNLOCK(ln); 1906 else 1907 LLE_RUNLOCK(ln); 1908 } 1909 1910 /* 1911 * If there has been no NS for the neighbor after entering the 1912 * INCOMPLETE state, send the first solicitation. 1913 */ 1914 if (!ND6_LLINFO_PERMANENT(ln) && ln->la_asked == 0) { 1915 ln->la_asked++; 1916 1917 nd6_llinfo_settimer(ln, 1918 (long)ND_IFINFO(ifp)->retrans * hz / 1000); 1919 nd6_ns_output(ifp, NULL, &dst->sin6_addr, ln, 0); 1920 } 1921 return (0); 1922 1923 sendpkt: 1924 /* discard the packet if IPv6 operation is disabled on the interface */ 1925 if ((ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED)) { 1926 error = ENETDOWN; /* better error? */ 1927 goto bad; 1928 } 1929 /* 1930 * ln is valid and the caller did not pass in 1931 * an llentry 1932 */ 1933 if ((ln != NULL) && (lle == NULL)) { 1934 if (flags & LLE_EXCLUSIVE) 1935 LLE_WUNLOCK(ln); 1936 else 1937 LLE_RUNLOCK(ln); 1938 } 1939 1940 #ifdef MAC 1941 mac_netinet6_nd6_send(ifp, m); 1942 #endif 1943 /* 1944 * We were passed in a pointer to an lle with the lock held 1945 * this means that we can't call if_output as we will 1946 * recurse on the lle lock - so what we do is we create 1947 * a list of mbufs to send and transmit them in the caller 1948 * after the lock is dropped 1949 */ 1950 if (lle != NULL) { 1951 if (*chain == NULL) 1952 *chain = m; 1953 else { 1954 struct mbuf *m = *chain; 1955 1956 /* 1957 * append mbuf to end of deferred chain 1958 */ 1959 while (m->m_nextpkt != NULL) 1960 m = m->m_nextpkt; 1961 m->m_nextpkt = m; 1962 } 1963 return (error); 1964 } 1965 if ((ifp->if_flags & IFF_LOOPBACK) != 0) { 1966 return ((*ifp->if_output)(origifp, m, (struct sockaddr *)dst, 1967 NULL)); 1968 } 1969 error = (*ifp->if_output)(ifp, m, (struct sockaddr *)dst, NULL); 1970 return (error); 1971 1972 bad: 1973 /* 1974 * ln is valid and the caller did not pass in 1975 * an llentry 1976 */ 1977 if ((ln != NULL) && (lle == NULL)) { 1978 if (flags & LLE_EXCLUSIVE) 1979 LLE_WUNLOCK(ln); 1980 else 1981 LLE_RUNLOCK(ln); 1982 } 1983 if (m) 1984 m_freem(m); 1985 return (error); 1986 } 1987 #undef senderr 1988 1989 1990 int 1991 nd6_output_flush(struct ifnet *ifp, struct ifnet *origifp, struct mbuf *chain, 1992 struct sockaddr_in6 *dst, struct route *ro) 1993 { 1994 struct mbuf *m, *m_head; 1995 struct ifnet *outifp; 1996 int error = 0; 1997 1998 m_head = chain; 1999 if ((ifp->if_flags & IFF_LOOPBACK) != 0) 2000 outifp = origifp; 2001 else 2002 outifp = ifp; 2003 2004 while (m_head) { 2005 m = m_head; 2006 m_head = m_head->m_nextpkt; 2007 error = (*ifp->if_output)(ifp, m, (struct sockaddr *)dst, ro); 2008 } 2009 2010 /* 2011 * XXX 2012 * note that intermediate errors are blindly ignored - but this is 2013 * the same convention as used with nd6_output when called by 2014 * nd6_cache_lladdr 2015 */ 2016 return (error); 2017 } 2018 2019 2020 int 2021 nd6_need_cache(struct ifnet *ifp) 2022 { 2023 /* 2024 * XXX: we currently do not make neighbor cache on any interface 2025 * other than ARCnet, Ethernet, FDDI and GIF. 2026 * 2027 * RFC2893 says: 2028 * - unidirectional tunnels needs no ND 2029 */ 2030 switch (ifp->if_type) { 2031 case IFT_ARCNET: 2032 case IFT_ETHER: 2033 case IFT_FDDI: 2034 case IFT_IEEE1394: 2035 #ifdef IFT_L2VLAN 2036 case IFT_L2VLAN: 2037 #endif 2038 #ifdef IFT_IEEE80211 2039 case IFT_IEEE80211: 2040 #endif 2041 #ifdef IFT_CARP 2042 case IFT_CARP: 2043 #endif 2044 case IFT_GIF: /* XXX need more cases? */ 2045 case IFT_PPP: 2046 case IFT_TUNNEL: 2047 case IFT_BRIDGE: 2048 case IFT_PROPVIRTUAL: 2049 return (1); 2050 default: 2051 return (0); 2052 } 2053 } 2054 2055 /* 2056 * the callers of this function need to be re-worked to drop 2057 * the lle lock, drop here for now 2058 */ 2059 int 2060 nd6_storelladdr(struct ifnet *ifp, struct mbuf *m, 2061 struct sockaddr *dst, u_char *desten, struct llentry **lle) 2062 { 2063 struct llentry *ln; 2064 2065 *lle = NULL; 2066 IF_AFDATA_UNLOCK_ASSERT(ifp); 2067 if (m->m_flags & M_MCAST) { 2068 int i; 2069 2070 switch (ifp->if_type) { 2071 case IFT_ETHER: 2072 case IFT_FDDI: 2073 #ifdef IFT_L2VLAN 2074 case IFT_L2VLAN: 2075 #endif 2076 #ifdef IFT_IEEE80211 2077 case IFT_IEEE80211: 2078 #endif 2079 case IFT_BRIDGE: 2080 case IFT_ISO88025: 2081 ETHER_MAP_IPV6_MULTICAST(&SIN6(dst)->sin6_addr, 2082 desten); 2083 return (0); 2084 case IFT_IEEE1394: 2085 /* 2086 * netbsd can use if_broadcastaddr, but we don't do so 2087 * to reduce # of ifdef. 2088 */ 2089 for (i = 0; i < ifp->if_addrlen; i++) 2090 desten[i] = ~0; 2091 return (0); 2092 case IFT_ARCNET: 2093 *desten = 0; 2094 return (0); 2095 default: 2096 m_freem(m); 2097 return (EAFNOSUPPORT); 2098 } 2099 } 2100 2101 2102 /* 2103 * the entry should have been created in nd6_store_lladdr 2104 */ 2105 IF_AFDATA_LOCK(ifp); 2106 ln = lla_lookup(LLTABLE6(ifp), 0, dst); 2107 IF_AFDATA_UNLOCK(ifp); 2108 if ((ln == NULL) || !(ln->la_flags & LLE_VALID)) { 2109 if (ln != NULL) 2110 LLE_RUNLOCK(ln); 2111 /* this could happen, if we could not allocate memory */ 2112 m_freem(m); 2113 return (1); 2114 } 2115 2116 bcopy(&ln->ll_addr, desten, ifp->if_addrlen); 2117 *lle = ln; 2118 LLE_RUNLOCK(ln); 2119 /* 2120 * A *small* use after free race exists here 2121 */ 2122 return (0); 2123 } 2124 2125 static void 2126 clear_llinfo_pqueue(struct llentry *ln) 2127 { 2128 struct mbuf *m_hold, *m_hold_next; 2129 2130 for (m_hold = ln->la_hold; m_hold; m_hold = m_hold_next) { 2131 m_hold_next = m_hold->m_nextpkt; 2132 m_hold->m_nextpkt = NULL; 2133 m_freem(m_hold); 2134 } 2135 2136 ln->la_hold = NULL; 2137 return; 2138 } 2139 2140 static int nd6_sysctl_drlist(SYSCTL_HANDLER_ARGS); 2141 static int nd6_sysctl_prlist(SYSCTL_HANDLER_ARGS); 2142 #ifdef SYSCTL_DECL 2143 SYSCTL_DECL(_net_inet6_icmp6); 2144 #endif 2145 SYSCTL_NODE(_net_inet6_icmp6, ICMPV6CTL_ND6_DRLIST, nd6_drlist, 2146 CTLFLAG_RD, nd6_sysctl_drlist, ""); 2147 SYSCTL_NODE(_net_inet6_icmp6, ICMPV6CTL_ND6_PRLIST, nd6_prlist, 2148 CTLFLAG_RD, nd6_sysctl_prlist, ""); 2149 SYSCTL_VNET_INT(_net_inet6_icmp6, ICMPV6CTL_ND6_MAXQLEN, nd6_maxqueuelen, 2150 CTLFLAG_RW, &VNET_NAME(nd6_maxqueuelen), 1, ""); 2151 2152 static int 2153 nd6_sysctl_drlist(SYSCTL_HANDLER_ARGS) 2154 { 2155 int error; 2156 char buf[1024] __aligned(4); 2157 struct in6_defrouter *d, *de; 2158 struct nd_defrouter *dr; 2159 2160 if (req->newptr) 2161 return EPERM; 2162 error = 0; 2163 2164 for (dr = TAILQ_FIRST(&V_nd_defrouter); dr; 2165 dr = TAILQ_NEXT(dr, dr_entry)) { 2166 d = (struct in6_defrouter *)buf; 2167 de = (struct in6_defrouter *)(buf + sizeof(buf)); 2168 2169 if (d + 1 <= de) { 2170 bzero(d, sizeof(*d)); 2171 d->rtaddr.sin6_family = AF_INET6; 2172 d->rtaddr.sin6_len = sizeof(d->rtaddr); 2173 d->rtaddr.sin6_addr = dr->rtaddr; 2174 error = sa6_recoverscope(&d->rtaddr); 2175 if (error != 0) 2176 return (error); 2177 d->flags = dr->flags; 2178 d->rtlifetime = dr->rtlifetime; 2179 d->expire = dr->expire; 2180 d->if_index = dr->ifp->if_index; 2181 } else 2182 panic("buffer too short"); 2183 2184 error = SYSCTL_OUT(req, buf, sizeof(*d)); 2185 if (error) 2186 break; 2187 } 2188 2189 return (error); 2190 } 2191 2192 static int 2193 nd6_sysctl_prlist(SYSCTL_HANDLER_ARGS) 2194 { 2195 int error; 2196 char buf[1024] __aligned(4); 2197 struct in6_prefix *p, *pe; 2198 struct nd_prefix *pr; 2199 char ip6buf[INET6_ADDRSTRLEN]; 2200 2201 if (req->newptr) 2202 return EPERM; 2203 error = 0; 2204 2205 for (pr = V_nd_prefix.lh_first; pr; pr = pr->ndpr_next) { 2206 u_short advrtrs; 2207 size_t advance; 2208 struct sockaddr_in6 *sin6, *s6; 2209 struct nd_pfxrouter *pfr; 2210 2211 p = (struct in6_prefix *)buf; 2212 pe = (struct in6_prefix *)(buf + sizeof(buf)); 2213 2214 if (p + 1 <= pe) { 2215 bzero(p, sizeof(*p)); 2216 sin6 = (struct sockaddr_in6 *)(p + 1); 2217 2218 p->prefix = pr->ndpr_prefix; 2219 if (sa6_recoverscope(&p->prefix)) { 2220 log(LOG_ERR, 2221 "scope error in prefix list (%s)\n", 2222 ip6_sprintf(ip6buf, &p->prefix.sin6_addr)); 2223 /* XXX: press on... */ 2224 } 2225 p->raflags = pr->ndpr_raf; 2226 p->prefixlen = pr->ndpr_plen; 2227 p->vltime = pr->ndpr_vltime; 2228 p->pltime = pr->ndpr_pltime; 2229 p->if_index = pr->ndpr_ifp->if_index; 2230 if (pr->ndpr_vltime == ND6_INFINITE_LIFETIME) 2231 p->expire = 0; 2232 else { 2233 time_t maxexpire; 2234 2235 /* XXX: we assume time_t is signed. */ 2236 maxexpire = (-1) & 2237 ~((time_t)1 << 2238 ((sizeof(maxexpire) * 8) - 1)); 2239 if (pr->ndpr_vltime < 2240 maxexpire - pr->ndpr_lastupdate) { 2241 p->expire = pr->ndpr_lastupdate + 2242 pr->ndpr_vltime; 2243 } else 2244 p->expire = maxexpire; 2245 } 2246 p->refcnt = pr->ndpr_refcnt; 2247 p->flags = pr->ndpr_stateflags; 2248 p->origin = PR_ORIG_RA; 2249 advrtrs = 0; 2250 for (pfr = pr->ndpr_advrtrs.lh_first; pfr; 2251 pfr = pfr->pfr_next) { 2252 if ((void *)&sin6[advrtrs + 1] > (void *)pe) { 2253 advrtrs++; 2254 continue; 2255 } 2256 s6 = &sin6[advrtrs]; 2257 bzero(s6, sizeof(*s6)); 2258 s6->sin6_family = AF_INET6; 2259 s6->sin6_len = sizeof(*sin6); 2260 s6->sin6_addr = pfr->router->rtaddr; 2261 if (sa6_recoverscope(s6)) { 2262 log(LOG_ERR, 2263 "scope error in " 2264 "prefix list (%s)\n", 2265 ip6_sprintf(ip6buf, 2266 &pfr->router->rtaddr)); 2267 } 2268 advrtrs++; 2269 } 2270 p->advrtrs = advrtrs; 2271 } else 2272 panic("buffer too short"); 2273 2274 advance = sizeof(*p) + sizeof(*sin6) * advrtrs; 2275 error = SYSCTL_OUT(req, buf, advance); 2276 if (error) 2277 break; 2278 } 2279 2280 return (error); 2281 } 2282