nd6.c revision 195727
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: head/sys/netinet6/nd6.c 195727 2009-07-16 21:13:04Z rwatson $"); 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#include <sys/vimage.h> 55 56#include <net/if.h> 57#include <net/if_arc.h> 58#include <net/if_dl.h> 59#include <net/if_types.h> 60#include <net/iso88025.h> 61#include <net/fddi.h> 62#include <net/route.h> 63#include <net/vnet.h> 64 65#include <netinet/in.h> 66#include <net/if_llatbl.h> 67#define L3_ADDR_SIN6(le) ((struct sockaddr_in6 *) L3_ADDR(le)) 68#include <netinet/if_ether.h> 69#include <netinet6/in6_var.h> 70#include <netinet/ip6.h> 71#include <netinet6/ip6_var.h> 72#include <netinet6/scope6_var.h> 73#include <netinet6/nd6.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 85VNET_DEFINE(int, nd6_prune); 86VNET_DEFINE(int, nd6_delay); 87VNET_DEFINE(int, nd6_umaxtries); 88VNET_DEFINE(int, nd6_mmaxtries); 89VNET_DEFINE(int, nd6_useloopback); 90VNET_DEFINE(int, nd6_gctimer); 91 92/* preventing too many loops in ND option parsing */ 93static VNET_DEFINE(int, nd6_maxndopt); 94VNET_DEFINE(int, nd6_maxnudhint); 95static VNET_DEFINE(int, nd6_maxqueuelen); 96#define V_nd6_maxndopt VNET(nd6_maxndopt) 97#define V_nd6_maxqueuelen VNET(nd6_maxqueuelen) 98 99VNET_DEFINE(int, nd6_debug); 100 101/* for debugging? */ 102#if 0 103static int nd6_inuse, nd6_allocated; 104#endif 105 106VNET_DEFINE(struct nd_drhead, nd_defrouter); 107VNET_DEFINE(struct nd_prhead, nd_prefix); 108 109VNET_DEFINE(int, nd6_recalc_reachtm_interval); 110#define V_nd6_recalc_reachtm_interval VNET(nd6_recalc_reachtm_interval) 111 112static struct sockaddr_in6 all1_sa; 113 114static int nd6_is_new_addr_neighbor __P((struct sockaddr_in6 *, 115 struct ifnet *)); 116static void nd6_setmtu0(struct ifnet *, struct nd_ifinfo *); 117static void nd6_slowtimo(void *); 118static int regen_tmpaddr(struct in6_ifaddr *); 119static struct llentry *nd6_free(struct llentry *, int); 120static void nd6_llinfo_timer(void *); 121static void clear_llinfo_pqueue(struct llentry *); 122 123static VNET_DEFINE(struct callout, nd6_slowtimo_ch); 124#define V_nd6_slowtimo_ch VNET(nd6_slowtimo_ch) 125 126VNET_DEFINE(struct callout, nd6_timer_ch); 127 128VNET_DECLARE(int, dad_ignore_ns); 129VNET_DECLARE(int, dad_maxtry); 130#define V_dad_ignore_ns VNET(dad_ignore_ns) 131#define V_dad_maxtry VNET(dad_maxtry) 132 133void 134nd6_init(void) 135{ 136 int i; 137 138 V_nd6_prune = 1; /* walk list every 1 seconds */ 139 V_nd6_delay = 5; /* delay first probe time 5 second */ 140 V_nd6_umaxtries = 3; /* maximum unicast query */ 141 V_nd6_mmaxtries = 3; /* maximum multicast query */ 142 V_nd6_useloopback = 1; /* use loopback interface for local traffic */ 143 V_nd6_gctimer = (60 * 60 * 24); /* 1 day: garbage collection timer */ 144 145 /* preventing too many loops in ND option parsing */ 146 V_nd6_maxndopt = 10; /* max # of ND options allowed */ 147 148 V_nd6_maxnudhint = 0; /* max # of subsequent upper layer hints */ 149 V_nd6_maxqueuelen = 1; /* max pkts cached in unresolved ND entries */ 150 151#ifdef ND6_DEBUG 152 V_nd6_debug = 1; 153#else 154 V_nd6_debug = 0; 155#endif 156 157 V_nd6_recalc_reachtm_interval = ND6_RECALC_REACHTM_INTERVAL; 158 159 V_dad_ignore_ns = 0; /* ignore NS in DAD - specwise incorrect*/ 160 V_dad_maxtry = 15; /* max # of *tries* to transmit DAD packet */ 161 162 /* 163 * XXX just to get this to compile KMM 164 */ 165#ifdef notyet 166 V_llinfo_nd6.ln_next = &V_llinfo_nd6; 167 V_llinfo_nd6.ln_prev = &V_llinfo_nd6; 168#endif 169 LIST_INIT(&V_nd_prefix); 170 171 V_ip6_use_tempaddr = 0; 172 V_ip6_temp_preferred_lifetime = DEF_TEMP_PREFERRED_LIFETIME; 173 V_ip6_temp_valid_lifetime = DEF_TEMP_VALID_LIFETIME; 174 V_ip6_temp_regen_advance = TEMPADDR_REGEN_ADVANCE; 175 176 V_ip6_desync_factor = 0; 177 178 all1_sa.sin6_family = AF_INET6; 179 all1_sa.sin6_len = sizeof(struct sockaddr_in6); 180 for (i = 0; i < sizeof(all1_sa.sin6_addr); i++) 181 all1_sa.sin6_addr.s6_addr[i] = 0xff; 182 183 /* initialization of the default router list */ 184 TAILQ_INIT(&V_nd_defrouter); 185 /* start timer */ 186 callout_init(&V_nd6_slowtimo_ch, 0); 187 callout_reset(&V_nd6_slowtimo_ch, ND6_SLOWTIMER_INTERVAL * hz, 188 nd6_slowtimo, curvnet); 189} 190 191 192#ifdef VIMAGE 193void 194nd6_destroy() 195{ 196 197 callout_drain(&V_nd6_slowtimo_ch); 198 callout_drain(&V_nd6_timer_ch); 199} 200#endif 201 202struct nd_ifinfo * 203nd6_ifattach(struct ifnet *ifp) 204{ 205 struct nd_ifinfo *nd; 206 207 nd = (struct nd_ifinfo *)malloc(sizeof(*nd), M_IP6NDP, M_WAITOK); 208 bzero(nd, sizeof(*nd)); 209 210 nd->initialized = 1; 211 212 nd->chlim = IPV6_DEFHLIM; 213 nd->basereachable = REACHABLE_TIME; 214 nd->reachable = ND_COMPUTE_RTIME(nd->basereachable); 215 nd->retrans = RETRANS_TIMER; 216 /* 217 * Note that the default value of ip6_accept_rtadv is 0, which means 218 * we won't accept RAs by default even if we set ND6_IFF_ACCEPT_RTADV 219 * here. 220 */ 221 nd->flags = (ND6_IFF_PERFORMNUD | ND6_IFF_ACCEPT_RTADV); 222 223 /* XXX: we cannot call nd6_setmtu since ifp is not fully initialized */ 224 nd6_setmtu0(ifp, nd); 225 226 return nd; 227} 228 229void 230nd6_ifdetach(struct nd_ifinfo *nd) 231{ 232 233 free(nd, M_IP6NDP); 234} 235 236/* 237 * Reset ND level link MTU. This function is called when the physical MTU 238 * changes, which means we might have to adjust the ND level MTU. 239 */ 240void 241nd6_setmtu(struct ifnet *ifp) 242{ 243 244 nd6_setmtu0(ifp, ND_IFINFO(ifp)); 245} 246 247/* XXX todo: do not maintain copy of ifp->if_mtu in ndi->maxmtu */ 248void 249nd6_setmtu0(struct ifnet *ifp, struct nd_ifinfo *ndi) 250{ 251 u_int32_t omaxmtu; 252 253 omaxmtu = ndi->maxmtu; 254 255 switch (ifp->if_type) { 256 case IFT_ARCNET: 257 ndi->maxmtu = MIN(ARC_PHDS_MAXMTU, ifp->if_mtu); /* RFC2497 */ 258 break; 259 case IFT_FDDI: 260 ndi->maxmtu = MIN(FDDIIPMTU, ifp->if_mtu); /* RFC2467 */ 261 break; 262 case IFT_ISO88025: 263 ndi->maxmtu = MIN(ISO88025_MAX_MTU, ifp->if_mtu); 264 break; 265 default: 266 ndi->maxmtu = ifp->if_mtu; 267 break; 268 } 269 270 /* 271 * Decreasing the interface MTU under IPV6 minimum MTU may cause 272 * undesirable situation. We thus notify the operator of the change 273 * explicitly. The check for omaxmtu is necessary to restrict the 274 * log to the case of changing the MTU, not initializing it. 275 */ 276 if (omaxmtu >= IPV6_MMTU && ndi->maxmtu < IPV6_MMTU) { 277 log(LOG_NOTICE, "nd6_setmtu0: " 278 "new link MTU on %s (%lu) is too small for IPv6\n", 279 if_name(ifp), (unsigned long)ndi->maxmtu); 280 } 281 282 if (ndi->maxmtu > V_in6_maxmtu) 283 in6_setmaxmtu(); /* check all interfaces just in case */ 284 285} 286 287void 288nd6_option_init(void *opt, int icmp6len, union nd_opts *ndopts) 289{ 290 291 bzero(ndopts, sizeof(*ndopts)); 292 ndopts->nd_opts_search = (struct nd_opt_hdr *)opt; 293 ndopts->nd_opts_last 294 = (struct nd_opt_hdr *)(((u_char *)opt) + icmp6len); 295 296 if (icmp6len == 0) { 297 ndopts->nd_opts_done = 1; 298 ndopts->nd_opts_search = NULL; 299 } 300} 301 302/* 303 * Take one ND option. 304 */ 305struct nd_opt_hdr * 306nd6_option(union nd_opts *ndopts) 307{ 308 struct nd_opt_hdr *nd_opt; 309 int olen; 310 311 if (ndopts == NULL) 312 panic("ndopts == NULL in nd6_option"); 313 if (ndopts->nd_opts_last == NULL) 314 panic("uninitialized ndopts in nd6_option"); 315 if (ndopts->nd_opts_search == NULL) 316 return NULL; 317 if (ndopts->nd_opts_done) 318 return NULL; 319 320 nd_opt = ndopts->nd_opts_search; 321 322 /* make sure nd_opt_len is inside the buffer */ 323 if ((caddr_t)&nd_opt->nd_opt_len >= (caddr_t)ndopts->nd_opts_last) { 324 bzero(ndopts, sizeof(*ndopts)); 325 return NULL; 326 } 327 328 olen = nd_opt->nd_opt_len << 3; 329 if (olen == 0) { 330 /* 331 * Message validation requires that all included 332 * options have a length that is greater than zero. 333 */ 334 bzero(ndopts, sizeof(*ndopts)); 335 return NULL; 336 } 337 338 ndopts->nd_opts_search = (struct nd_opt_hdr *)((caddr_t)nd_opt + olen); 339 if (ndopts->nd_opts_search > ndopts->nd_opts_last) { 340 /* option overruns the end of buffer, invalid */ 341 bzero(ndopts, sizeof(*ndopts)); 342 return NULL; 343 } else if (ndopts->nd_opts_search == ndopts->nd_opts_last) { 344 /* reached the end of options chain */ 345 ndopts->nd_opts_done = 1; 346 ndopts->nd_opts_search = NULL; 347 } 348 return nd_opt; 349} 350 351/* 352 * Parse multiple ND options. 353 * This function is much easier to use, for ND routines that do not need 354 * multiple options of the same type. 355 */ 356int 357nd6_options(union nd_opts *ndopts) 358{ 359 struct nd_opt_hdr *nd_opt; 360 int i = 0; 361 362 if (ndopts == NULL) 363 panic("ndopts == NULL in nd6_options"); 364 if (ndopts->nd_opts_last == NULL) 365 panic("uninitialized ndopts in nd6_options"); 366 if (ndopts->nd_opts_search == NULL) 367 return 0; 368 369 while (1) { 370 nd_opt = nd6_option(ndopts); 371 if (nd_opt == NULL && ndopts->nd_opts_last == NULL) { 372 /* 373 * Message validation requires that all included 374 * options have a length that is greater than zero. 375 */ 376 ICMP6STAT_INC(icp6s_nd_badopt); 377 bzero(ndopts, sizeof(*ndopts)); 378 return -1; 379 } 380 381 if (nd_opt == NULL) 382 goto skip1; 383 384 switch (nd_opt->nd_opt_type) { 385 case ND_OPT_SOURCE_LINKADDR: 386 case ND_OPT_TARGET_LINKADDR: 387 case ND_OPT_MTU: 388 case ND_OPT_REDIRECTED_HEADER: 389 if (ndopts->nd_opt_array[nd_opt->nd_opt_type]) { 390 nd6log((LOG_INFO, 391 "duplicated ND6 option found (type=%d)\n", 392 nd_opt->nd_opt_type)); 393 /* XXX bark? */ 394 } else { 395 ndopts->nd_opt_array[nd_opt->nd_opt_type] 396 = nd_opt; 397 } 398 break; 399 case ND_OPT_PREFIX_INFORMATION: 400 if (ndopts->nd_opt_array[nd_opt->nd_opt_type] == 0) { 401 ndopts->nd_opt_array[nd_opt->nd_opt_type] 402 = nd_opt; 403 } 404 ndopts->nd_opts_pi_end = 405 (struct nd_opt_prefix_info *)nd_opt; 406 break; 407 default: 408 /* 409 * Unknown options must be silently ignored, 410 * to accomodate future extension to the protocol. 411 */ 412 nd6log((LOG_DEBUG, 413 "nd6_options: unsupported option %d - " 414 "option ignored\n", nd_opt->nd_opt_type)); 415 } 416 417skip1: 418 i++; 419 if (i > V_nd6_maxndopt) { 420 ICMP6STAT_INC(icp6s_nd_toomanyopt); 421 nd6log((LOG_INFO, "too many loop in nd opt\n")); 422 break; 423 } 424 425 if (ndopts->nd_opts_done) 426 break; 427 } 428 429 return 0; 430} 431 432/* 433 * ND6 timer routine to handle ND6 entries 434 */ 435void 436nd6_llinfo_settimer_locked(struct llentry *ln, long tick) 437{ 438 if (tick < 0) { 439 ln->la_expire = 0; 440 ln->ln_ntick = 0; 441 callout_stop(&ln->ln_timer_ch); 442 /* 443 * XXX - do we know that there is 444 * callout installed? i.e. are we 445 * guaranteed that we're not dropping 446 * a reference that we did not add? 447 * KMM 448 */ 449 LLE_REMREF(ln); 450 } else { 451 ln->la_expire = time_second + tick / hz; 452 LLE_ADDREF(ln); 453 if (tick > INT_MAX) { 454 ln->ln_ntick = tick - INT_MAX; 455 callout_reset(&ln->ln_timer_ch, INT_MAX, 456 nd6_llinfo_timer, ln); 457 } else { 458 ln->ln_ntick = 0; 459 callout_reset(&ln->ln_timer_ch, tick, 460 nd6_llinfo_timer, ln); 461 } 462 } 463} 464 465void 466nd6_llinfo_settimer(struct llentry *ln, long tick) 467{ 468 469 LLE_WLOCK(ln); 470 nd6_llinfo_settimer_locked(ln, tick); 471 LLE_WUNLOCK(ln); 472} 473 474static void 475nd6_llinfo_timer(void *arg) 476{ 477 struct llentry *ln; 478 struct in6_addr *dst; 479 struct ifnet *ifp; 480 struct nd_ifinfo *ndi = NULL; 481 482 ln = (struct llentry *)arg; 483 if (ln == NULL) { 484 panic("%s: NULL entry!\n", __func__); 485 return; 486 } 487 488 if ((ifp = ((ln->lle_tbl != NULL) ? ln->lle_tbl->llt_ifp : NULL)) == NULL) 489 panic("ln ifp == NULL"); 490 491 CURVNET_SET(ifp->if_vnet); 492 493 if (ln->ln_ntick > 0) { 494 if (ln->ln_ntick > INT_MAX) { 495 ln->ln_ntick -= INT_MAX; 496 nd6_llinfo_settimer(ln, INT_MAX); 497 } else { 498 ln->ln_ntick = 0; 499 nd6_llinfo_settimer(ln, ln->ln_ntick); 500 } 501 goto done; 502 } 503 504 ndi = ND_IFINFO(ifp); 505 dst = &L3_ADDR_SIN6(ln)->sin6_addr; 506 if ((ln->la_flags & LLE_STATIC) || (ln->la_expire > time_second)) { 507 goto done; 508 } 509 510 if (ln->la_flags & LLE_DELETED) { 511 (void)nd6_free(ln, 0); 512 goto done; 513 } 514 515 switch (ln->ln_state) { 516 case ND6_LLINFO_INCOMPLETE: 517 if (ln->la_asked < V_nd6_mmaxtries) { 518 ln->la_asked++; 519 nd6_llinfo_settimer(ln, (long)ndi->retrans * hz / 1000); 520 nd6_ns_output(ifp, NULL, dst, ln, 0); 521 } else { 522 struct mbuf *m = ln->la_hold; 523 if (m) { 524 struct mbuf *m0; 525 526 /* 527 * assuming every packet in la_hold has the 528 * same IP header 529 */ 530 m0 = m->m_nextpkt; 531 m->m_nextpkt = NULL; 532 icmp6_error2(m, ICMP6_DST_UNREACH, 533 ICMP6_DST_UNREACH_ADDR, 0, ifp); 534 535 ln->la_hold = m0; 536 clear_llinfo_pqueue(ln); 537 } 538 (void)nd6_free(ln, 0); 539 ln = NULL; 540 } 541 break; 542 case ND6_LLINFO_REACHABLE: 543 if (!ND6_LLINFO_PERMANENT(ln)) { 544 ln->ln_state = ND6_LLINFO_STALE; 545 nd6_llinfo_settimer(ln, (long)V_nd6_gctimer * hz); 546 } 547 break; 548 549 case ND6_LLINFO_STALE: 550 /* Garbage Collection(RFC 2461 5.3) */ 551 if (!ND6_LLINFO_PERMANENT(ln)) { 552 (void)nd6_free(ln, 1); 553 ln = NULL; 554 } 555 break; 556 557 case ND6_LLINFO_DELAY: 558 if (ndi && (ndi->flags & ND6_IFF_PERFORMNUD) != 0) { 559 /* We need NUD */ 560 ln->la_asked = 1; 561 ln->ln_state = ND6_LLINFO_PROBE; 562 nd6_llinfo_settimer(ln, (long)ndi->retrans * hz / 1000); 563 nd6_ns_output(ifp, dst, dst, ln, 0); 564 } else { 565 ln->ln_state = ND6_LLINFO_STALE; /* XXX */ 566 nd6_llinfo_settimer(ln, (long)V_nd6_gctimer * hz); 567 } 568 break; 569 case ND6_LLINFO_PROBE: 570 if (ln->la_asked < V_nd6_umaxtries) { 571 ln->la_asked++; 572 nd6_llinfo_settimer(ln, (long)ndi->retrans * hz / 1000); 573 nd6_ns_output(ifp, dst, dst, ln, 0); 574 } else { 575 (void)nd6_free(ln, 0); 576 ln = NULL; 577 } 578 break; 579 } 580 CURVNET_RESTORE(); 581done: 582 if (ln != NULL) 583 LLE_FREE(ln); 584} 585 586 587/* 588 * ND6 timer routine to expire default route list and prefix list 589 */ 590void 591nd6_timer(void *arg) 592{ 593 CURVNET_SET((struct vnet *) arg); 594 int s; 595 struct nd_defrouter *dr; 596 struct nd_prefix *pr; 597 struct in6_ifaddr *ia6, *nia6; 598 struct in6_addrlifetime *lt6; 599 600 callout_reset(&V_nd6_timer_ch, V_nd6_prune * hz, 601 nd6_timer, curvnet); 602 603 /* expire default router list */ 604 s = splnet(); 605 dr = TAILQ_FIRST(&V_nd_defrouter); 606 while (dr) { 607 if (dr->expire && dr->expire < time_second) { 608 struct nd_defrouter *t; 609 t = TAILQ_NEXT(dr, dr_entry); 610 defrtrlist_del(dr); 611 dr = t; 612 } else { 613 dr = TAILQ_NEXT(dr, dr_entry); 614 } 615 } 616 617 /* 618 * expire interface addresses. 619 * in the past the loop was inside prefix expiry processing. 620 * However, from a stricter speci-confrmance standpoint, we should 621 * rather separate address lifetimes and prefix lifetimes. 622 * 623 * XXXRW: in6_ifaddrhead locking. 624 */ 625 addrloop: 626 TAILQ_FOREACH_SAFE(ia6, &V_in6_ifaddrhead, ia_link, nia6) { 627 /* check address lifetime */ 628 lt6 = &ia6->ia6_lifetime; 629 if (IFA6_IS_INVALID(ia6)) { 630 int regen = 0; 631 632 /* 633 * If the expiring address is temporary, try 634 * regenerating a new one. This would be useful when 635 * we suspended a laptop PC, then turned it on after a 636 * period that could invalidate all temporary 637 * addresses. Although we may have to restart the 638 * loop (see below), it must be after purging the 639 * address. Otherwise, we'd see an infinite loop of 640 * regeneration. 641 */ 642 if (V_ip6_use_tempaddr && 643 (ia6->ia6_flags & IN6_IFF_TEMPORARY) != 0) { 644 if (regen_tmpaddr(ia6) == 0) 645 regen = 1; 646 } 647 648 in6_purgeaddr(&ia6->ia_ifa); 649 650 if (regen) 651 goto addrloop; /* XXX: see below */ 652 } else if (IFA6_IS_DEPRECATED(ia6)) { 653 int oldflags = ia6->ia6_flags; 654 655 ia6->ia6_flags |= IN6_IFF_DEPRECATED; 656 657 /* 658 * If a temporary address has just become deprecated, 659 * regenerate a new one if possible. 660 */ 661 if (V_ip6_use_tempaddr && 662 (ia6->ia6_flags & IN6_IFF_TEMPORARY) != 0 && 663 (oldflags & IN6_IFF_DEPRECATED) == 0) { 664 665 if (regen_tmpaddr(ia6) == 0) { 666 /* 667 * A new temporary address is 668 * generated. 669 * XXX: this means the address chain 670 * has changed while we are still in 671 * the loop. Although the change 672 * would not cause disaster (because 673 * it's not a deletion, but an 674 * addition,) we'd rather restart the 675 * loop just for safety. Or does this 676 * significantly reduce performance?? 677 */ 678 goto addrloop; 679 } 680 } 681 } else { 682 /* 683 * A new RA might have made a deprecated address 684 * preferred. 685 */ 686 ia6->ia6_flags &= ~IN6_IFF_DEPRECATED; 687 } 688 } 689 690 /* expire prefix list */ 691 pr = V_nd_prefix.lh_first; 692 while (pr) { 693 /* 694 * check prefix lifetime. 695 * since pltime is just for autoconf, pltime processing for 696 * prefix is not necessary. 697 */ 698 if (pr->ndpr_vltime != ND6_INFINITE_LIFETIME && 699 time_second - pr->ndpr_lastupdate > pr->ndpr_vltime) { 700 struct nd_prefix *t; 701 t = pr->ndpr_next; 702 703 /* 704 * address expiration and prefix expiration are 705 * separate. NEVER perform in6_purgeaddr here. 706 */ 707 708 prelist_remove(pr); 709 pr = t; 710 } else 711 pr = pr->ndpr_next; 712 } 713 splx(s); 714 CURVNET_RESTORE(); 715} 716 717/* 718 * ia6 - deprecated/invalidated temporary address 719 */ 720static int 721regen_tmpaddr(struct in6_ifaddr *ia6) 722{ 723 struct ifaddr *ifa; 724 struct ifnet *ifp; 725 struct in6_ifaddr *public_ifa6 = NULL; 726 727 ifp = ia6->ia_ifa.ifa_ifp; 728 IF_ADDR_LOCK(ifp); 729 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 730 struct in6_ifaddr *it6; 731 732 if (ifa->ifa_addr->sa_family != AF_INET6) 733 continue; 734 735 it6 = (struct in6_ifaddr *)ifa; 736 737 /* ignore no autoconf addresses. */ 738 if ((it6->ia6_flags & IN6_IFF_AUTOCONF) == 0) 739 continue; 740 741 /* ignore autoconf addresses with different prefixes. */ 742 if (it6->ia6_ndpr == NULL || it6->ia6_ndpr != ia6->ia6_ndpr) 743 continue; 744 745 /* 746 * Now we are looking at an autoconf address with the same 747 * prefix as ours. If the address is temporary and is still 748 * preferred, do not create another one. It would be rare, but 749 * could happen, for example, when we resume a laptop PC after 750 * a long period. 751 */ 752 if ((it6->ia6_flags & IN6_IFF_TEMPORARY) != 0 && 753 !IFA6_IS_DEPRECATED(it6)) { 754 public_ifa6 = NULL; 755 break; 756 } 757 758 /* 759 * This is a public autoconf address that has the same prefix 760 * as ours. If it is preferred, keep it. We can't break the 761 * loop here, because there may be a still-preferred temporary 762 * address with the prefix. 763 */ 764 if (!IFA6_IS_DEPRECATED(it6)) 765 public_ifa6 = it6; 766 } 767 768 if (public_ifa6 != NULL) { 769 int e; 770 771 if ((e = in6_tmpifadd(public_ifa6, 0, 0)) != 0) { 772 IF_ADDR_UNLOCK(ifp); 773 log(LOG_NOTICE, "regen_tmpaddr: failed to create a new" 774 " tmp addr,errno=%d\n", e); 775 return (-1); 776 } 777 IF_ADDR_UNLOCK(ifp); 778 return (0); 779 } 780 781 IF_ADDR_UNLOCK(ifp); 782 return (-1); 783} 784 785/* 786 * Nuke neighbor cache/prefix/default router management table, right before 787 * ifp goes away. 788 */ 789void 790nd6_purge(struct ifnet *ifp) 791{ 792 struct nd_defrouter *dr, *ndr; 793 struct nd_prefix *pr, *npr; 794 795 /* 796 * Nuke default router list entries toward ifp. 797 * We defer removal of default router list entries that is installed 798 * in the routing table, in order to keep additional side effects as 799 * small as possible. 800 */ 801 for (dr = TAILQ_FIRST(&V_nd_defrouter); dr; dr = ndr) { 802 ndr = TAILQ_NEXT(dr, dr_entry); 803 if (dr->installed) 804 continue; 805 806 if (dr->ifp == ifp) 807 defrtrlist_del(dr); 808 } 809 810 for (dr = TAILQ_FIRST(&V_nd_defrouter); dr; dr = ndr) { 811 ndr = TAILQ_NEXT(dr, dr_entry); 812 if (!dr->installed) 813 continue; 814 815 if (dr->ifp == ifp) 816 defrtrlist_del(dr); 817 } 818 819 /* Nuke prefix list entries toward ifp */ 820 for (pr = V_nd_prefix.lh_first; pr; pr = npr) { 821 npr = pr->ndpr_next; 822 if (pr->ndpr_ifp == ifp) { 823 /* 824 * Because if_detach() does *not* release prefixes 825 * while purging addresses the reference count will 826 * still be above zero. We therefore reset it to 827 * make sure that the prefix really gets purged. 828 */ 829 pr->ndpr_refcnt = 0; 830 831 /* 832 * Previously, pr->ndpr_addr is removed as well, 833 * but I strongly believe we don't have to do it. 834 * nd6_purge() is only called from in6_ifdetach(), 835 * which removes all the associated interface addresses 836 * by itself. 837 * (jinmei@kame.net 20010129) 838 */ 839 prelist_remove(pr); 840 } 841 } 842 843 /* cancel default outgoing interface setting */ 844 if (V_nd6_defifindex == ifp->if_index) 845 nd6_setdefaultiface(0); 846 847 if (!V_ip6_forwarding && V_ip6_accept_rtadv) { /* XXX: too restrictive? */ 848 /* refresh default router list 849 * 850 * 851 */ 852 defrouter_select(); 853 854 } 855 856 /* XXXXX 857 * We do not nuke the neighbor cache entries here any more 858 * because the neighbor cache is kept in if_afdata[AF_INET6]. 859 * nd6_purge() is invoked by in6_ifdetach() which is called 860 * from if_detach() where everything gets purged. So let 861 * in6_domifdetach() do the actual L2 table purging work. 862 */ 863} 864 865/* 866 * the caller acquires and releases the lock on the lltbls 867 * Returns the llentry locked 868 */ 869struct llentry * 870nd6_lookup(struct in6_addr *addr6, int flags, struct ifnet *ifp) 871{ 872 struct sockaddr_in6 sin6; 873 struct llentry *ln; 874 int llflags = 0; 875 876 bzero(&sin6, sizeof(sin6)); 877 sin6.sin6_len = sizeof(struct sockaddr_in6); 878 sin6.sin6_family = AF_INET6; 879 sin6.sin6_addr = *addr6; 880 881 IF_AFDATA_LOCK_ASSERT(ifp); 882 883 if (flags & ND6_CREATE) 884 llflags |= LLE_CREATE; 885 if (flags & ND6_EXCLUSIVE) 886 llflags |= LLE_EXCLUSIVE; 887 888 ln = lla_lookup(LLTABLE6(ifp), llflags, (struct sockaddr *)&sin6); 889 if ((ln != NULL) && (flags & LLE_CREATE)) { 890 ln->ln_state = ND6_LLINFO_NOSTATE; 891 callout_init(&ln->ln_timer_ch, 0); 892 } 893 894 return (ln); 895} 896 897/* 898 * Test whether a given IPv6 address is a neighbor or not, ignoring 899 * the actual neighbor cache. The neighbor cache is ignored in order 900 * to not reenter the routing code from within itself. 901 */ 902static int 903nd6_is_new_addr_neighbor(struct sockaddr_in6 *addr, struct ifnet *ifp) 904{ 905 struct nd_prefix *pr; 906 struct ifaddr *dstaddr; 907 908 /* 909 * A link-local address is always a neighbor. 910 * XXX: a link does not necessarily specify a single interface. 911 */ 912 if (IN6_IS_ADDR_LINKLOCAL(&addr->sin6_addr)) { 913 struct sockaddr_in6 sin6_copy; 914 u_int32_t zone; 915 916 /* 917 * We need sin6_copy since sa6_recoverscope() may modify the 918 * content (XXX). 919 */ 920 sin6_copy = *addr; 921 if (sa6_recoverscope(&sin6_copy)) 922 return (0); /* XXX: should be impossible */ 923 if (in6_setscope(&sin6_copy.sin6_addr, ifp, &zone)) 924 return (0); 925 if (sin6_copy.sin6_scope_id == zone) 926 return (1); 927 else 928 return (0); 929 } 930 931 /* 932 * If the address matches one of our addresses, 933 * it should be a neighbor. 934 * If the address matches one of our on-link prefixes, it should be a 935 * neighbor. 936 */ 937 for (pr = V_nd_prefix.lh_first; pr; pr = pr->ndpr_next) { 938 if (pr->ndpr_ifp != ifp) 939 continue; 940 941 if (!(pr->ndpr_stateflags & NDPRF_ONLINK)) 942 continue; 943 944 if (IN6_ARE_MASKED_ADDR_EQUAL(&pr->ndpr_prefix.sin6_addr, 945 &addr->sin6_addr, &pr->ndpr_mask)) 946 return (1); 947 } 948 949 /* 950 * If the address is assigned on the node of the other side of 951 * a p2p interface, the address should be a neighbor. 952 */ 953 dstaddr = ifa_ifwithdstaddr((struct sockaddr *)addr); 954 if (dstaddr != NULL) { 955 if (dstaddr->ifa_ifp == ifp) { 956 ifa_free(dstaddr); 957 return (1); 958 } 959 ifa_free(dstaddr); 960 } 961 962 /* 963 * If the default router list is empty, all addresses are regarded 964 * as on-link, and thus, as a neighbor. 965 * XXX: we restrict the condition to hosts, because routers usually do 966 * not have the "default router list". 967 */ 968 if (!V_ip6_forwarding && TAILQ_FIRST(&V_nd_defrouter) == NULL && 969 V_nd6_defifindex == ifp->if_index) { 970 return (1); 971 } 972 973 return (0); 974} 975 976 977/* 978 * Detect if a given IPv6 address identifies a neighbor on a given link. 979 * XXX: should take care of the destination of a p2p link? 980 */ 981int 982nd6_is_addr_neighbor(struct sockaddr_in6 *addr, struct ifnet *ifp) 983{ 984 struct llentry *lle; 985 int rc = 0; 986 987 IF_AFDATA_UNLOCK_ASSERT(ifp); 988 if (nd6_is_new_addr_neighbor(addr, ifp)) 989 return (1); 990 991 /* 992 * Even if the address matches none of our addresses, it might be 993 * in the neighbor cache. 994 */ 995 IF_AFDATA_LOCK(ifp); 996 if ((lle = nd6_lookup(&addr->sin6_addr, 0, ifp)) != NULL) { 997 LLE_RUNLOCK(lle); 998 rc = 1; 999 } 1000 IF_AFDATA_UNLOCK(ifp); 1001 return (rc); 1002} 1003 1004/* 1005 * Free an nd6 llinfo entry. 1006 * Since the function would cause significant changes in the kernel, DO NOT 1007 * make it global, unless you have a strong reason for the change, and are sure 1008 * that the change is safe. 1009 */ 1010static struct llentry * 1011nd6_free(struct llentry *ln, int gc) 1012{ 1013 struct llentry *next; 1014 struct nd_defrouter *dr; 1015 struct ifnet *ifp=NULL; 1016 1017 /* 1018 * we used to have pfctlinput(PRC_HOSTDEAD) here. 1019 * even though it is not harmful, it was not really necessary. 1020 */ 1021 1022 /* cancel timer */ 1023 nd6_llinfo_settimer(ln, -1); 1024 1025 if (!V_ip6_forwarding) { 1026 int s; 1027 s = splnet(); 1028 dr = defrouter_lookup(&L3_ADDR_SIN6(ln)->sin6_addr, ln->lle_tbl->llt_ifp); 1029 1030 if (dr != NULL && dr->expire && 1031 ln->ln_state == ND6_LLINFO_STALE && gc) { 1032 /* 1033 * If the reason for the deletion is just garbage 1034 * collection, and the neighbor is an active default 1035 * router, do not delete it. Instead, reset the GC 1036 * timer using the router's lifetime. 1037 * Simply deleting the entry would affect default 1038 * router selection, which is not necessarily a good 1039 * thing, especially when we're using router preference 1040 * values. 1041 * XXX: the check for ln_state would be redundant, 1042 * but we intentionally keep it just in case. 1043 */ 1044 if (dr->expire > time_second) 1045 nd6_llinfo_settimer(ln, 1046 (dr->expire - time_second) * hz); 1047 else 1048 nd6_llinfo_settimer(ln, (long)V_nd6_gctimer * hz); 1049 splx(s); 1050 return (LIST_NEXT(ln, lle_next)); 1051 } 1052 1053 if (ln->ln_router || dr) { 1054 /* 1055 * rt6_flush must be called whether or not the neighbor 1056 * is in the Default Router List. 1057 * See a corresponding comment in nd6_na_input(). 1058 */ 1059 rt6_flush(&L3_ADDR_SIN6(ln)->sin6_addr, ln->lle_tbl->llt_ifp); 1060 } 1061 1062 if (dr) { 1063 /* 1064 * Unreachablity of a router might affect the default 1065 * router selection and on-link detection of advertised 1066 * prefixes. 1067 */ 1068 1069 /* 1070 * Temporarily fake the state to choose a new default 1071 * router and to perform on-link determination of 1072 * prefixes correctly. 1073 * Below the state will be set correctly, 1074 * or the entry itself will be deleted. 1075 */ 1076 ln->ln_state = ND6_LLINFO_INCOMPLETE; 1077 1078 /* 1079 * Since defrouter_select() does not affect the 1080 * on-link determination and MIP6 needs the check 1081 * before the default router selection, we perform 1082 * the check now. 1083 */ 1084 pfxlist_onlink_check(); 1085 1086 /* 1087 * refresh default router list 1088 */ 1089 defrouter_select(); 1090 } 1091 splx(s); 1092 } 1093 1094 /* 1095 * Before deleting the entry, remember the next entry as the 1096 * return value. We need this because pfxlist_onlink_check() above 1097 * might have freed other entries (particularly the old next entry) as 1098 * a side effect (XXX). 1099 */ 1100 next = LIST_NEXT(ln, lle_next); 1101 1102 ifp = ln->lle_tbl->llt_ifp; 1103 IF_AFDATA_LOCK(ifp); 1104 LLE_WLOCK(ln); 1105 llentry_free(ln); 1106 IF_AFDATA_UNLOCK(ifp); 1107 1108 return (next); 1109} 1110 1111/* 1112 * Upper-layer reachability hint for Neighbor Unreachability Detection. 1113 * 1114 * XXX cost-effective methods? 1115 */ 1116void 1117nd6_nud_hint(struct rtentry *rt, struct in6_addr *dst6, int force) 1118{ 1119 struct llentry *ln; 1120 struct ifnet *ifp; 1121 1122 if ((dst6 == NULL) || (rt == NULL)) 1123 return; 1124 1125 ifp = rt->rt_ifp; 1126 IF_AFDATA_LOCK(ifp); 1127 ln = nd6_lookup(dst6, ND6_EXCLUSIVE, NULL); 1128 IF_AFDATA_UNLOCK(ifp); 1129 if (ln == NULL) 1130 return; 1131 1132 if (ln->ln_state < ND6_LLINFO_REACHABLE) 1133 goto done; 1134 1135 /* 1136 * if we get upper-layer reachability confirmation many times, 1137 * it is possible we have false information. 1138 */ 1139 if (!force) { 1140 ln->ln_byhint++; 1141 if (ln->ln_byhint > V_nd6_maxnudhint) { 1142 goto done; 1143 } 1144 } 1145 1146 ln->ln_state = ND6_LLINFO_REACHABLE; 1147 if (!ND6_LLINFO_PERMANENT(ln)) { 1148 nd6_llinfo_settimer(ln, 1149 (long)ND_IFINFO(rt->rt_ifp)->reachable * hz); 1150 } 1151done: 1152 LLE_WUNLOCK(ln); 1153} 1154 1155 1156int 1157nd6_ioctl(u_long cmd, caddr_t data, struct ifnet *ifp) 1158{ 1159 struct in6_drlist *drl = (struct in6_drlist *)data; 1160 struct in6_oprlist *oprl = (struct in6_oprlist *)data; 1161 struct in6_ndireq *ndi = (struct in6_ndireq *)data; 1162 struct in6_nbrinfo *nbi = (struct in6_nbrinfo *)data; 1163 struct in6_ndifreq *ndif = (struct in6_ndifreq *)data; 1164 struct nd_defrouter *dr; 1165 struct nd_prefix *pr; 1166 int i = 0, error = 0; 1167 int s; 1168 1169 switch (cmd) { 1170 case SIOCGDRLST_IN6: 1171 /* 1172 * obsolete API, use sysctl under net.inet6.icmp6 1173 */ 1174 bzero(drl, sizeof(*drl)); 1175 s = splnet(); 1176 dr = TAILQ_FIRST(&V_nd_defrouter); 1177 while (dr && i < DRLSTSIZ) { 1178 drl->defrouter[i].rtaddr = dr->rtaddr; 1179 in6_clearscope(&drl->defrouter[i].rtaddr); 1180 1181 drl->defrouter[i].flags = dr->flags; 1182 drl->defrouter[i].rtlifetime = dr->rtlifetime; 1183 drl->defrouter[i].expire = dr->expire; 1184 drl->defrouter[i].if_index = dr->ifp->if_index; 1185 i++; 1186 dr = TAILQ_NEXT(dr, dr_entry); 1187 } 1188 splx(s); 1189 break; 1190 case SIOCGPRLST_IN6: 1191 /* 1192 * obsolete API, use sysctl under net.inet6.icmp6 1193 * 1194 * XXX the structure in6_prlist was changed in backward- 1195 * incompatible manner. in6_oprlist is used for SIOCGPRLST_IN6, 1196 * in6_prlist is used for nd6_sysctl() - fill_prlist(). 1197 */ 1198 /* 1199 * XXX meaning of fields, especialy "raflags", is very 1200 * differnet between RA prefix list and RR/static prefix list. 1201 * how about separating ioctls into two? 1202 */ 1203 bzero(oprl, sizeof(*oprl)); 1204 s = splnet(); 1205 pr = V_nd_prefix.lh_first; 1206 while (pr && i < PRLSTSIZ) { 1207 struct nd_pfxrouter *pfr; 1208 int j; 1209 1210 oprl->prefix[i].prefix = pr->ndpr_prefix.sin6_addr; 1211 oprl->prefix[i].raflags = pr->ndpr_raf; 1212 oprl->prefix[i].prefixlen = pr->ndpr_plen; 1213 oprl->prefix[i].vltime = pr->ndpr_vltime; 1214 oprl->prefix[i].pltime = pr->ndpr_pltime; 1215 oprl->prefix[i].if_index = pr->ndpr_ifp->if_index; 1216 if (pr->ndpr_vltime == ND6_INFINITE_LIFETIME) 1217 oprl->prefix[i].expire = 0; 1218 else { 1219 time_t maxexpire; 1220 1221 /* XXX: we assume time_t is signed. */ 1222 maxexpire = (-1) & 1223 ~((time_t)1 << 1224 ((sizeof(maxexpire) * 8) - 1)); 1225 if (pr->ndpr_vltime < 1226 maxexpire - pr->ndpr_lastupdate) { 1227 oprl->prefix[i].expire = 1228 pr->ndpr_lastupdate + 1229 pr->ndpr_vltime; 1230 } else 1231 oprl->prefix[i].expire = maxexpire; 1232 } 1233 1234 pfr = pr->ndpr_advrtrs.lh_first; 1235 j = 0; 1236 while (pfr) { 1237 if (j < DRLSTSIZ) { 1238#define RTRADDR oprl->prefix[i].advrtr[j] 1239 RTRADDR = pfr->router->rtaddr; 1240 in6_clearscope(&RTRADDR); 1241#undef RTRADDR 1242 } 1243 j++; 1244 pfr = pfr->pfr_next; 1245 } 1246 oprl->prefix[i].advrtrs = j; 1247 oprl->prefix[i].origin = PR_ORIG_RA; 1248 1249 i++; 1250 pr = pr->ndpr_next; 1251 } 1252 splx(s); 1253 1254 break; 1255 case OSIOCGIFINFO_IN6: 1256#define ND ndi->ndi 1257 /* XXX: old ndp(8) assumes a positive value for linkmtu. */ 1258 bzero(&ND, sizeof(ND)); 1259 ND.linkmtu = IN6_LINKMTU(ifp); 1260 ND.maxmtu = ND_IFINFO(ifp)->maxmtu; 1261 ND.basereachable = ND_IFINFO(ifp)->basereachable; 1262 ND.reachable = ND_IFINFO(ifp)->reachable; 1263 ND.retrans = ND_IFINFO(ifp)->retrans; 1264 ND.flags = ND_IFINFO(ifp)->flags; 1265 ND.recalctm = ND_IFINFO(ifp)->recalctm; 1266 ND.chlim = ND_IFINFO(ifp)->chlim; 1267 break; 1268 case SIOCGIFINFO_IN6: 1269 ND = *ND_IFINFO(ifp); 1270 break; 1271 case SIOCSIFINFO_IN6: 1272 /* 1273 * used to change host variables from userland. 1274 * intented for a use on router to reflect RA configurations. 1275 */ 1276 /* 0 means 'unspecified' */ 1277 if (ND.linkmtu != 0) { 1278 if (ND.linkmtu < IPV6_MMTU || 1279 ND.linkmtu > IN6_LINKMTU(ifp)) { 1280 error = EINVAL; 1281 break; 1282 } 1283 ND_IFINFO(ifp)->linkmtu = ND.linkmtu; 1284 } 1285 1286 if (ND.basereachable != 0) { 1287 int obasereachable = ND_IFINFO(ifp)->basereachable; 1288 1289 ND_IFINFO(ifp)->basereachable = ND.basereachable; 1290 if (ND.basereachable != obasereachable) 1291 ND_IFINFO(ifp)->reachable = 1292 ND_COMPUTE_RTIME(ND.basereachable); 1293 } 1294 if (ND.retrans != 0) 1295 ND_IFINFO(ifp)->retrans = ND.retrans; 1296 if (ND.chlim != 0) 1297 ND_IFINFO(ifp)->chlim = ND.chlim; 1298 /* FALLTHROUGH */ 1299 case SIOCSIFINFO_FLAGS: 1300 ND_IFINFO(ifp)->flags = ND.flags; 1301 break; 1302#undef ND 1303 case SIOCSNDFLUSH_IN6: /* XXX: the ioctl name is confusing... */ 1304 /* sync kernel routing table with the default router list */ 1305 defrouter_reset(); 1306 defrouter_select(); 1307 break; 1308 case SIOCSPFXFLUSH_IN6: 1309 { 1310 /* flush all the prefix advertised by routers */ 1311 struct nd_prefix *pr, *next; 1312 1313 s = splnet(); 1314 for (pr = V_nd_prefix.lh_first; pr; pr = next) { 1315 struct in6_ifaddr *ia, *ia_next; 1316 1317 next = pr->ndpr_next; 1318 1319 if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr)) 1320 continue; /* XXX */ 1321 1322 /* do we really have to remove addresses as well? */ 1323 /* XXXRW: in6_ifaddrhead locking. */ 1324 TAILQ_FOREACH_SAFE(ia, &V_in6_ifaddrhead, ia_link, 1325 ia_next) { 1326 if ((ia->ia6_flags & IN6_IFF_AUTOCONF) == 0) 1327 continue; 1328 1329 if (ia->ia6_ndpr == pr) 1330 in6_purgeaddr(&ia->ia_ifa); 1331 } 1332 prelist_remove(pr); 1333 } 1334 splx(s); 1335 break; 1336 } 1337 case SIOCSRTRFLUSH_IN6: 1338 { 1339 /* flush all the default routers */ 1340 struct nd_defrouter *dr, *next; 1341 1342 s = splnet(); 1343 defrouter_reset(); 1344 for (dr = TAILQ_FIRST(&V_nd_defrouter); dr; dr = next) { 1345 next = TAILQ_NEXT(dr, dr_entry); 1346 defrtrlist_del(dr); 1347 } 1348 defrouter_select(); 1349 splx(s); 1350 break; 1351 } 1352 case SIOCGNBRINFO_IN6: 1353 { 1354 struct llentry *ln; 1355 struct in6_addr nb_addr = nbi->addr; /* make local for safety */ 1356 1357 if ((error = in6_setscope(&nb_addr, ifp, NULL)) != 0) 1358 return (error); 1359 1360 IF_AFDATA_LOCK(ifp); 1361 ln = nd6_lookup(&nb_addr, 0, ifp); 1362 IF_AFDATA_UNLOCK(ifp); 1363 1364 if (ln == NULL) { 1365 error = EINVAL; 1366 break; 1367 } 1368 nbi->state = ln->ln_state; 1369 nbi->asked = ln->la_asked; 1370 nbi->isrouter = ln->ln_router; 1371 nbi->expire = ln->la_expire; 1372 LLE_RUNLOCK(ln); 1373 break; 1374 } 1375 case SIOCGDEFIFACE_IN6: /* XXX: should be implemented as a sysctl? */ 1376 ndif->ifindex = V_nd6_defifindex; 1377 break; 1378 case SIOCSDEFIFACE_IN6: /* XXX: should be implemented as a sysctl? */ 1379 return (nd6_setdefaultiface(ndif->ifindex)); 1380 } 1381 return (error); 1382} 1383 1384/* 1385 * Create neighbor cache entry and cache link-layer address, 1386 * on reception of inbound ND6 packets. (RS/RA/NS/redirect) 1387 * 1388 * type - ICMP6 type 1389 * code - type dependent information 1390 * 1391 * XXXXX 1392 * The caller of this function already acquired the ndp 1393 * cache table lock because the cache entry is returned. 1394 */ 1395struct llentry * 1396nd6_cache_lladdr(struct ifnet *ifp, struct in6_addr *from, char *lladdr, 1397 int lladdrlen, int type, int code) 1398{ 1399 struct llentry *ln = NULL; 1400 int is_newentry; 1401 int do_update; 1402 int olladdr; 1403 int llchange; 1404 int flags = 0; 1405 int newstate = 0; 1406 uint16_t router = 0; 1407 struct sockaddr_in6 sin6; 1408 struct mbuf *chain = NULL; 1409 int static_route = 0; 1410 1411 IF_AFDATA_UNLOCK_ASSERT(ifp); 1412 1413 if (ifp == NULL) 1414 panic("ifp == NULL in nd6_cache_lladdr"); 1415 if (from == NULL) 1416 panic("from == NULL in nd6_cache_lladdr"); 1417 1418 /* nothing must be updated for unspecified address */ 1419 if (IN6_IS_ADDR_UNSPECIFIED(from)) 1420 return NULL; 1421 1422 /* 1423 * Validation about ifp->if_addrlen and lladdrlen must be done in 1424 * the caller. 1425 * 1426 * XXX If the link does not have link-layer adderss, what should 1427 * we do? (ifp->if_addrlen == 0) 1428 * Spec says nothing in sections for RA, RS and NA. There's small 1429 * description on it in NS section (RFC 2461 7.2.3). 1430 */ 1431 flags |= lladdr ? ND6_EXCLUSIVE : 0; 1432 IF_AFDATA_LOCK(ifp); 1433 ln = nd6_lookup(from, flags, ifp); 1434 1435 if (ln == NULL) { 1436 flags |= LLE_EXCLUSIVE; 1437 ln = nd6_lookup(from, flags |ND6_CREATE, ifp); 1438 IF_AFDATA_UNLOCK(ifp); 1439 is_newentry = 1; 1440 } else { 1441 IF_AFDATA_UNLOCK(ifp); 1442 /* do nothing if static ndp is set */ 1443 if (ln->la_flags & LLE_STATIC) { 1444 static_route = 1; 1445 goto done; 1446 } 1447 is_newentry = 0; 1448 } 1449 if (ln == NULL) 1450 return (NULL); 1451 1452 olladdr = (ln->la_flags & LLE_VALID) ? 1 : 0; 1453 if (olladdr && lladdr) { 1454 llchange = bcmp(lladdr, &ln->ll_addr, 1455 ifp->if_addrlen); 1456 } else 1457 llchange = 0; 1458 1459 /* 1460 * newentry olladdr lladdr llchange (*=record) 1461 * 0 n n -- (1) 1462 * 0 y n -- (2) 1463 * 0 n y -- (3) * STALE 1464 * 0 y y n (4) * 1465 * 0 y y y (5) * STALE 1466 * 1 -- n -- (6) NOSTATE(= PASSIVE) 1467 * 1 -- y -- (7) * STALE 1468 */ 1469 1470 if (lladdr) { /* (3-5) and (7) */ 1471 /* 1472 * Record source link-layer address 1473 * XXX is it dependent to ifp->if_type? 1474 */ 1475 bcopy(lladdr, &ln->ll_addr, ifp->if_addrlen); 1476 ln->la_flags |= LLE_VALID; 1477 } 1478 1479 if (!is_newentry) { 1480 if ((!olladdr && lladdr != NULL) || /* (3) */ 1481 (olladdr && lladdr != NULL && llchange)) { /* (5) */ 1482 do_update = 1; 1483 newstate = ND6_LLINFO_STALE; 1484 } else /* (1-2,4) */ 1485 do_update = 0; 1486 } else { 1487 do_update = 1; 1488 if (lladdr == NULL) /* (6) */ 1489 newstate = ND6_LLINFO_NOSTATE; 1490 else /* (7) */ 1491 newstate = ND6_LLINFO_STALE; 1492 } 1493 1494 if (do_update) { 1495 /* 1496 * Update the state of the neighbor cache. 1497 */ 1498 ln->ln_state = newstate; 1499 1500 if (ln->ln_state == ND6_LLINFO_STALE) { 1501 /* 1502 * XXX: since nd6_output() below will cause 1503 * state tansition to DELAY and reset the timer, 1504 * we must set the timer now, although it is actually 1505 * meaningless. 1506 */ 1507 nd6_llinfo_settimer_locked(ln, (long)V_nd6_gctimer * hz); 1508 1509 if (ln->la_hold) { 1510 struct mbuf *m_hold, *m_hold_next; 1511 1512 /* 1513 * reset the la_hold in advance, to explicitly 1514 * prevent a la_hold lookup in nd6_output() 1515 * (wouldn't happen, though...) 1516 */ 1517 for (m_hold = ln->la_hold, ln->la_hold = NULL; 1518 m_hold; m_hold = m_hold_next) { 1519 m_hold_next = m_hold->m_nextpkt; 1520 m_hold->m_nextpkt = NULL; 1521 1522 /* 1523 * we assume ifp is not a p2p here, so 1524 * just set the 2nd argument as the 1525 * 1st one. 1526 */ 1527 nd6_output_lle(ifp, ifp, m_hold, L3_ADDR_SIN6(ln), NULL, ln, &chain); 1528 } 1529 /* 1530 * If we have mbufs in the chain we need to do 1531 * deferred transmit. Copy the address from the 1532 * llentry before dropping the lock down below. 1533 */ 1534 if (chain != NULL) 1535 memcpy(&sin6, L3_ADDR_SIN6(ln), sizeof(sin6)); 1536 } 1537 } else if (ln->ln_state == ND6_LLINFO_INCOMPLETE) { 1538 /* probe right away */ 1539 nd6_llinfo_settimer_locked((void *)ln, 0); 1540 } 1541 } 1542 1543 /* 1544 * ICMP6 type dependent behavior. 1545 * 1546 * NS: clear IsRouter if new entry 1547 * RS: clear IsRouter 1548 * RA: set IsRouter if there's lladdr 1549 * redir: clear IsRouter if new entry 1550 * 1551 * RA case, (1): 1552 * The spec says that we must set IsRouter in the following cases: 1553 * - If lladdr exist, set IsRouter. This means (1-5). 1554 * - If it is old entry (!newentry), set IsRouter. This means (7). 1555 * So, based on the spec, in (1-5) and (7) cases we must set IsRouter. 1556 * A quetion arises for (1) case. (1) case has no lladdr in the 1557 * neighbor cache, this is similar to (6). 1558 * This case is rare but we figured that we MUST NOT set IsRouter. 1559 * 1560 * newentry olladdr lladdr llchange NS RS RA redir 1561 * D R 1562 * 0 n n -- (1) c ? s 1563 * 0 y n -- (2) c s s 1564 * 0 n y -- (3) c s s 1565 * 0 y y n (4) c s s 1566 * 0 y y y (5) c s s 1567 * 1 -- n -- (6) c c c s 1568 * 1 -- y -- (7) c c s c s 1569 * 1570 * (c=clear s=set) 1571 */ 1572 switch (type & 0xff) { 1573 case ND_NEIGHBOR_SOLICIT: 1574 /* 1575 * New entry must have is_router flag cleared. 1576 */ 1577 if (is_newentry) /* (6-7) */ 1578 ln->ln_router = 0; 1579 break; 1580 case ND_REDIRECT: 1581 /* 1582 * If the icmp is a redirect to a better router, always set the 1583 * is_router flag. Otherwise, if the entry is newly created, 1584 * clear the flag. [RFC 2461, sec 8.3] 1585 */ 1586 if (code == ND_REDIRECT_ROUTER) 1587 ln->ln_router = 1; 1588 else if (is_newentry) /* (6-7) */ 1589 ln->ln_router = 0; 1590 break; 1591 case ND_ROUTER_SOLICIT: 1592 /* 1593 * is_router flag must always be cleared. 1594 */ 1595 ln->ln_router = 0; 1596 break; 1597 case ND_ROUTER_ADVERT: 1598 /* 1599 * Mark an entry with lladdr as a router. 1600 */ 1601 if ((!is_newentry && (olladdr || lladdr)) || /* (2-5) */ 1602 (is_newentry && lladdr)) { /* (7) */ 1603 ln->ln_router = 1; 1604 } 1605 break; 1606 } 1607 1608 if (ln != NULL) { 1609 static_route = (ln->la_flags & LLE_STATIC); 1610 router = ln->ln_router; 1611 1612 if (flags & ND6_EXCLUSIVE) 1613 LLE_WUNLOCK(ln); 1614 else 1615 LLE_RUNLOCK(ln); 1616 if (static_route) 1617 ln = NULL; 1618 } 1619 if (chain) 1620 nd6_output_flush(ifp, ifp, chain, &sin6, NULL); 1621 1622 /* 1623 * When the link-layer address of a router changes, select the 1624 * best router again. In particular, when the neighbor entry is newly 1625 * created, it might affect the selection policy. 1626 * Question: can we restrict the first condition to the "is_newentry" 1627 * case? 1628 * XXX: when we hear an RA from a new router with the link-layer 1629 * address option, defrouter_select() is called twice, since 1630 * defrtrlist_update called the function as well. However, I believe 1631 * we can compromise the overhead, since it only happens the first 1632 * time. 1633 * XXX: although defrouter_select() should not have a bad effect 1634 * for those are not autoconfigured hosts, we explicitly avoid such 1635 * cases for safety. 1636 */ 1637 if (do_update && router && !V_ip6_forwarding && V_ip6_accept_rtadv) { 1638 /* 1639 * guaranteed recursion 1640 */ 1641 defrouter_select(); 1642 } 1643 1644 return (ln); 1645done: 1646 if (ln != NULL) { 1647 if (flags & ND6_EXCLUSIVE) 1648 LLE_WUNLOCK(ln); 1649 else 1650 LLE_RUNLOCK(ln); 1651 if (static_route) 1652 ln = NULL; 1653 } 1654 return (ln); 1655} 1656 1657static void 1658nd6_slowtimo(void *arg) 1659{ 1660 CURVNET_SET((struct vnet *) arg); 1661 struct nd_ifinfo *nd6if; 1662 struct ifnet *ifp; 1663 1664 callout_reset(&V_nd6_slowtimo_ch, ND6_SLOWTIMER_INTERVAL * hz, 1665 nd6_slowtimo, curvnet); 1666 IFNET_RLOCK(); 1667 for (ifp = TAILQ_FIRST(&V_ifnet); ifp; 1668 ifp = TAILQ_NEXT(ifp, if_list)) { 1669 nd6if = ND_IFINFO(ifp); 1670 if (nd6if->basereachable && /* already initialized */ 1671 (nd6if->recalctm -= ND6_SLOWTIMER_INTERVAL) <= 0) { 1672 /* 1673 * Since reachable time rarely changes by router 1674 * advertisements, we SHOULD insure that a new random 1675 * value gets recomputed at least once every few hours. 1676 * (RFC 2461, 6.3.4) 1677 */ 1678 nd6if->recalctm = V_nd6_recalc_reachtm_interval; 1679 nd6if->reachable = ND_COMPUTE_RTIME(nd6if->basereachable); 1680 } 1681 } 1682 IFNET_RUNLOCK(); 1683 CURVNET_RESTORE(); 1684} 1685 1686int 1687nd6_output(struct ifnet *ifp, struct ifnet *origifp, struct mbuf *m0, 1688 struct sockaddr_in6 *dst, struct rtentry *rt0) 1689{ 1690 1691 return (nd6_output_lle(ifp, origifp, m0, dst, rt0, NULL, NULL)); 1692} 1693 1694 1695/* 1696 * Note that I'm not enforcing any global serialization 1697 * lle state or asked changes here as the logic is too 1698 * complicated to avoid having to always acquire an exclusive 1699 * lock 1700 * KMM 1701 * 1702 */ 1703#define senderr(e) { error = (e); goto bad;} 1704 1705int 1706nd6_output_lle(struct ifnet *ifp, struct ifnet *origifp, struct mbuf *m0, 1707 struct sockaddr_in6 *dst, struct rtentry *rt0, struct llentry *lle, 1708 struct mbuf **chain) 1709{ 1710 struct mbuf *m = m0; 1711 struct llentry *ln = lle; 1712 int error = 0; 1713 int flags = 0; 1714 1715#ifdef INVARIANTS 1716 if (lle != NULL) { 1717 1718 LLE_WLOCK_ASSERT(lle); 1719 1720 KASSERT(chain != NULL, (" lle locked but no mbuf chain pointer passed")); 1721 } 1722#endif 1723 if (IN6_IS_ADDR_MULTICAST(&dst->sin6_addr)) 1724 goto sendpkt; 1725 1726 if (nd6_need_cache(ifp) == 0) 1727 goto sendpkt; 1728 1729 /* 1730 * next hop determination. This routine is derived from ether_output. 1731 */ 1732 1733 /* 1734 * Address resolution or Neighbor Unreachability Detection 1735 * for the next hop. 1736 * At this point, the destination of the packet must be a unicast 1737 * or an anycast address(i.e. not a multicast). 1738 */ 1739 1740 flags = ((m != NULL) || (lle != NULL)) ? LLE_EXCLUSIVE : 0; 1741 if (ln == NULL) { 1742 retry: 1743 IF_AFDATA_LOCK(ifp); 1744 ln = lla_lookup(LLTABLE6(ifp), flags, (struct sockaddr *)dst); 1745 IF_AFDATA_UNLOCK(ifp); 1746 if ((ln == NULL) && nd6_is_addr_neighbor(dst, ifp)) { 1747 /* 1748 * Since nd6_is_addr_neighbor() internally calls nd6_lookup(), 1749 * the condition below is not very efficient. But we believe 1750 * it is tolerable, because this should be a rare case. 1751 */ 1752 flags = ND6_CREATE | (m ? ND6_EXCLUSIVE : 0); 1753 IF_AFDATA_LOCK(ifp); 1754 ln = nd6_lookup(&dst->sin6_addr, flags, ifp); 1755 IF_AFDATA_UNLOCK(ifp); 1756 } 1757 } 1758 if (ln == NULL) { 1759 if ((ifp->if_flags & IFF_POINTOPOINT) == 0 && 1760 !(ND_IFINFO(ifp)->flags & ND6_IFF_PERFORMNUD)) { 1761 char ip6buf[INET6_ADDRSTRLEN]; 1762 log(LOG_DEBUG, 1763 "nd6_output: can't allocate llinfo for %s " 1764 "(ln=%p)\n", 1765 ip6_sprintf(ip6buf, &dst->sin6_addr), ln); 1766 senderr(EIO); /* XXX: good error? */ 1767 } 1768 goto sendpkt; /* send anyway */ 1769 } 1770 1771 /* We don't have to do link-layer address resolution on a p2p link. */ 1772 if ((ifp->if_flags & IFF_POINTOPOINT) != 0 && 1773 ln->ln_state < ND6_LLINFO_REACHABLE) { 1774 if ((flags & LLE_EXCLUSIVE) == 0) { 1775 flags |= LLE_EXCLUSIVE; 1776 goto retry; 1777 } 1778 ln->ln_state = ND6_LLINFO_STALE; 1779 nd6_llinfo_settimer_locked(ln, (long)V_nd6_gctimer * hz); 1780 } 1781 1782 /* 1783 * The first time we send a packet to a neighbor whose entry is 1784 * STALE, we have to change the state to DELAY and a sets a timer to 1785 * expire in DELAY_FIRST_PROBE_TIME seconds to ensure do 1786 * neighbor unreachability detection on expiration. 1787 * (RFC 2461 7.3.3) 1788 */ 1789 if (ln->ln_state == ND6_LLINFO_STALE) { 1790 if ((flags & LLE_EXCLUSIVE) == 0) { 1791 flags |= LLE_EXCLUSIVE; 1792 LLE_RUNLOCK(ln); 1793 goto retry; 1794 } 1795 ln->la_asked = 0; 1796 ln->ln_state = ND6_LLINFO_DELAY; 1797 nd6_llinfo_settimer_locked(ln, (long)V_nd6_delay * hz); 1798 } 1799 1800 /* 1801 * If the neighbor cache entry has a state other than INCOMPLETE 1802 * (i.e. its link-layer address is already resolved), just 1803 * send the packet. 1804 */ 1805 if (ln->ln_state > ND6_LLINFO_INCOMPLETE) 1806 goto sendpkt; 1807 1808 /* 1809 * There is a neighbor cache entry, but no ethernet address 1810 * response yet. Append this latest packet to the end of the 1811 * packet queue in the mbuf, unless the number of the packet 1812 * does not exceed nd6_maxqueuelen. When it exceeds nd6_maxqueuelen, 1813 * the oldest packet in the queue will be removed. 1814 */ 1815 if (ln->ln_state == ND6_LLINFO_NOSTATE) 1816 ln->ln_state = ND6_LLINFO_INCOMPLETE; 1817 1818 if ((flags & LLE_EXCLUSIVE) == 0) { 1819 flags |= LLE_EXCLUSIVE; 1820 LLE_RUNLOCK(ln); 1821 goto retry; 1822 } 1823 if (ln->la_hold) { 1824 struct mbuf *m_hold; 1825 int i; 1826 1827 i = 0; 1828 for (m_hold = ln->la_hold; m_hold; m_hold = m_hold->m_nextpkt) { 1829 i++; 1830 if (m_hold->m_nextpkt == NULL) { 1831 m_hold->m_nextpkt = m; 1832 break; 1833 } 1834 } 1835 while (i >= V_nd6_maxqueuelen) { 1836 m_hold = ln->la_hold; 1837 ln->la_hold = ln->la_hold->m_nextpkt; 1838 m_freem(m_hold); 1839 i--; 1840 } 1841 } else { 1842 ln->la_hold = m; 1843 } 1844 /* 1845 * We did the lookup (no lle arg) so we 1846 * need to do the unlock here 1847 */ 1848 if (lle == NULL) { 1849 if (flags & LLE_EXCLUSIVE) 1850 LLE_WUNLOCK(ln); 1851 else 1852 LLE_RUNLOCK(ln); 1853 } 1854 1855 /* 1856 * If there has been no NS for the neighbor after entering the 1857 * INCOMPLETE state, send the first solicitation. 1858 */ 1859 if (!ND6_LLINFO_PERMANENT(ln) && ln->la_asked == 0) { 1860 ln->la_asked++; 1861 1862 nd6_llinfo_settimer(ln, 1863 (long)ND_IFINFO(ifp)->retrans * hz / 1000); 1864 nd6_ns_output(ifp, NULL, &dst->sin6_addr, ln, 0); 1865 } 1866 return (0); 1867 1868 sendpkt: 1869 /* discard the packet if IPv6 operation is disabled on the interface */ 1870 if ((ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED)) { 1871 error = ENETDOWN; /* better error? */ 1872 goto bad; 1873 } 1874 /* 1875 * ln is valid and the caller did not pass in 1876 * an llentry 1877 */ 1878 if ((ln != NULL) && (lle == NULL)) { 1879 if (flags & LLE_EXCLUSIVE) 1880 LLE_WUNLOCK(ln); 1881 else 1882 LLE_RUNLOCK(ln); 1883 } 1884 1885#ifdef MAC 1886 mac_netinet6_nd6_send(ifp, m); 1887#endif 1888 /* 1889 * We were passed in a pointer to an lle with the lock held 1890 * this means that we can't call if_output as we will 1891 * recurse on the lle lock - so what we do is we create 1892 * a list of mbufs to send and transmit them in the caller 1893 * after the lock is dropped 1894 */ 1895 if (lle != NULL) { 1896 if (*chain == NULL) 1897 *chain = m; 1898 else { 1899 struct mbuf *m = *chain; 1900 1901 /* 1902 * append mbuf to end of deferred chain 1903 */ 1904 while (m->m_nextpkt != NULL) 1905 m = m->m_nextpkt; 1906 m->m_nextpkt = m; 1907 } 1908 return (error); 1909 } 1910 if ((ifp->if_flags & IFF_LOOPBACK) != 0) { 1911 return ((*ifp->if_output)(origifp, m, (struct sockaddr *)dst, 1912 NULL)); 1913 } 1914 error = (*ifp->if_output)(ifp, m, (struct sockaddr *)dst, NULL); 1915 return (error); 1916 1917 bad: 1918 /* 1919 * ln is valid and the caller did not pass in 1920 * an llentry 1921 */ 1922 if ((ln != NULL) && (lle == NULL)) { 1923 if (flags & LLE_EXCLUSIVE) 1924 LLE_WUNLOCK(ln); 1925 else 1926 LLE_RUNLOCK(ln); 1927 } 1928 if (m) 1929 m_freem(m); 1930 return (error); 1931} 1932#undef senderr 1933 1934 1935int 1936nd6_output_flush(struct ifnet *ifp, struct ifnet *origifp, struct mbuf *chain, 1937 struct sockaddr_in6 *dst, struct route *ro) 1938{ 1939 struct mbuf *m, *m_head; 1940 struct ifnet *outifp; 1941 int error = 0; 1942 1943 m_head = chain; 1944 if ((ifp->if_flags & IFF_LOOPBACK) != 0) 1945 outifp = origifp; 1946 else 1947 outifp = ifp; 1948 1949 while (m_head) { 1950 m = m_head; 1951 m_head = m_head->m_nextpkt; 1952 error = (*ifp->if_output)(ifp, m, (struct sockaddr *)dst, ro); 1953 } 1954 1955 /* 1956 * XXX 1957 * note that intermediate errors are blindly ignored - but this is 1958 * the same convention as used with nd6_output when called by 1959 * nd6_cache_lladdr 1960 */ 1961 return (error); 1962} 1963 1964 1965int 1966nd6_need_cache(struct ifnet *ifp) 1967{ 1968 /* 1969 * XXX: we currently do not make neighbor cache on any interface 1970 * other than ARCnet, Ethernet, FDDI and GIF. 1971 * 1972 * RFC2893 says: 1973 * - unidirectional tunnels needs no ND 1974 */ 1975 switch (ifp->if_type) { 1976 case IFT_ARCNET: 1977 case IFT_ETHER: 1978 case IFT_FDDI: 1979 case IFT_IEEE1394: 1980#ifdef IFT_L2VLAN 1981 case IFT_L2VLAN: 1982#endif 1983#ifdef IFT_IEEE80211 1984 case IFT_IEEE80211: 1985#endif 1986#ifdef IFT_CARP 1987 case IFT_CARP: 1988#endif 1989 case IFT_GIF: /* XXX need more cases? */ 1990 case IFT_PPP: 1991 case IFT_TUNNEL: 1992 case IFT_BRIDGE: 1993 case IFT_PROPVIRTUAL: 1994 return (1); 1995 default: 1996 return (0); 1997 } 1998} 1999 2000/* 2001 * the callers of this function need to be re-worked to drop 2002 * the lle lock, drop here for now 2003 */ 2004int 2005nd6_storelladdr(struct ifnet *ifp, struct mbuf *m, 2006 struct sockaddr *dst, u_char *desten, struct llentry **lle) 2007{ 2008 struct llentry *ln; 2009 2010 *lle = NULL; 2011 IF_AFDATA_UNLOCK_ASSERT(ifp); 2012 if (m->m_flags & M_MCAST) { 2013 int i; 2014 2015 switch (ifp->if_type) { 2016 case IFT_ETHER: 2017 case IFT_FDDI: 2018#ifdef IFT_L2VLAN 2019 case IFT_L2VLAN: 2020#endif 2021#ifdef IFT_IEEE80211 2022 case IFT_IEEE80211: 2023#endif 2024 case IFT_BRIDGE: 2025 case IFT_ISO88025: 2026 ETHER_MAP_IPV6_MULTICAST(&SIN6(dst)->sin6_addr, 2027 desten); 2028 return (0); 2029 case IFT_IEEE1394: 2030 /* 2031 * netbsd can use if_broadcastaddr, but we don't do so 2032 * to reduce # of ifdef. 2033 */ 2034 for (i = 0; i < ifp->if_addrlen; i++) 2035 desten[i] = ~0; 2036 return (0); 2037 case IFT_ARCNET: 2038 *desten = 0; 2039 return (0); 2040 default: 2041 m_freem(m); 2042 return (EAFNOSUPPORT); 2043 } 2044 } 2045 2046 2047 /* 2048 * the entry should have been created in nd6_store_lladdr 2049 */ 2050 IF_AFDATA_LOCK(ifp); 2051 ln = lla_lookup(LLTABLE6(ifp), 0, dst); 2052 IF_AFDATA_UNLOCK(ifp); 2053 if ((ln == NULL) || !(ln->la_flags & LLE_VALID)) { 2054 if (ln != NULL) 2055 LLE_RUNLOCK(ln); 2056 /* this could happen, if we could not allocate memory */ 2057 m_freem(m); 2058 return (1); 2059 } 2060 2061 bcopy(&ln->ll_addr, desten, ifp->if_addrlen); 2062 *lle = ln; 2063 LLE_RUNLOCK(ln); 2064 /* 2065 * A *small* use after free race exists here 2066 */ 2067 return (0); 2068} 2069 2070static void 2071clear_llinfo_pqueue(struct llentry *ln) 2072{ 2073 struct mbuf *m_hold, *m_hold_next; 2074 2075 for (m_hold = ln->la_hold; m_hold; m_hold = m_hold_next) { 2076 m_hold_next = m_hold->m_nextpkt; 2077 m_hold->m_nextpkt = NULL; 2078 m_freem(m_hold); 2079 } 2080 2081 ln->la_hold = NULL; 2082 return; 2083} 2084 2085static int nd6_sysctl_drlist(SYSCTL_HANDLER_ARGS); 2086static int nd6_sysctl_prlist(SYSCTL_HANDLER_ARGS); 2087#ifdef SYSCTL_DECL 2088SYSCTL_DECL(_net_inet6_icmp6); 2089#endif 2090SYSCTL_NODE(_net_inet6_icmp6, ICMPV6CTL_ND6_DRLIST, nd6_drlist, 2091 CTLFLAG_RD, nd6_sysctl_drlist, ""); 2092SYSCTL_NODE(_net_inet6_icmp6, ICMPV6CTL_ND6_PRLIST, nd6_prlist, 2093 CTLFLAG_RD, nd6_sysctl_prlist, ""); 2094SYSCTL_VNET_INT(_net_inet6_icmp6, ICMPV6CTL_ND6_MAXQLEN, nd6_maxqueuelen, 2095 CTLFLAG_RW, &VNET_NAME(nd6_maxqueuelen), 1, ""); 2096 2097static int 2098nd6_sysctl_drlist(SYSCTL_HANDLER_ARGS) 2099{ 2100 int error; 2101 char buf[1024] __aligned(4); 2102 struct in6_defrouter *d, *de; 2103 struct nd_defrouter *dr; 2104 2105 if (req->newptr) 2106 return EPERM; 2107 error = 0; 2108 2109 for (dr = TAILQ_FIRST(&V_nd_defrouter); dr; 2110 dr = TAILQ_NEXT(dr, dr_entry)) { 2111 d = (struct in6_defrouter *)buf; 2112 de = (struct in6_defrouter *)(buf + sizeof(buf)); 2113 2114 if (d + 1 <= de) { 2115 bzero(d, sizeof(*d)); 2116 d->rtaddr.sin6_family = AF_INET6; 2117 d->rtaddr.sin6_len = sizeof(d->rtaddr); 2118 d->rtaddr.sin6_addr = dr->rtaddr; 2119 error = sa6_recoverscope(&d->rtaddr); 2120 if (error != 0) 2121 return (error); 2122 d->flags = dr->flags; 2123 d->rtlifetime = dr->rtlifetime; 2124 d->expire = dr->expire; 2125 d->if_index = dr->ifp->if_index; 2126 } else 2127 panic("buffer too short"); 2128 2129 error = SYSCTL_OUT(req, buf, sizeof(*d)); 2130 if (error) 2131 break; 2132 } 2133 2134 return (error); 2135} 2136 2137static int 2138nd6_sysctl_prlist(SYSCTL_HANDLER_ARGS) 2139{ 2140 int error; 2141 char buf[1024] __aligned(4); 2142 struct in6_prefix *p, *pe; 2143 struct nd_prefix *pr; 2144 char ip6buf[INET6_ADDRSTRLEN]; 2145 2146 if (req->newptr) 2147 return EPERM; 2148 error = 0; 2149 2150 for (pr = V_nd_prefix.lh_first; pr; pr = pr->ndpr_next) { 2151 u_short advrtrs; 2152 size_t advance; 2153 struct sockaddr_in6 *sin6, *s6; 2154 struct nd_pfxrouter *pfr; 2155 2156 p = (struct in6_prefix *)buf; 2157 pe = (struct in6_prefix *)(buf + sizeof(buf)); 2158 2159 if (p + 1 <= pe) { 2160 bzero(p, sizeof(*p)); 2161 sin6 = (struct sockaddr_in6 *)(p + 1); 2162 2163 p->prefix = pr->ndpr_prefix; 2164 if (sa6_recoverscope(&p->prefix)) { 2165 log(LOG_ERR, 2166 "scope error in prefix list (%s)\n", 2167 ip6_sprintf(ip6buf, &p->prefix.sin6_addr)); 2168 /* XXX: press on... */ 2169 } 2170 p->raflags = pr->ndpr_raf; 2171 p->prefixlen = pr->ndpr_plen; 2172 p->vltime = pr->ndpr_vltime; 2173 p->pltime = pr->ndpr_pltime; 2174 p->if_index = pr->ndpr_ifp->if_index; 2175 if (pr->ndpr_vltime == ND6_INFINITE_LIFETIME) 2176 p->expire = 0; 2177 else { 2178 time_t maxexpire; 2179 2180 /* XXX: we assume time_t is signed. */ 2181 maxexpire = (-1) & 2182 ~((time_t)1 << 2183 ((sizeof(maxexpire) * 8) - 1)); 2184 if (pr->ndpr_vltime < 2185 maxexpire - pr->ndpr_lastupdate) { 2186 p->expire = pr->ndpr_lastupdate + 2187 pr->ndpr_vltime; 2188 } else 2189 p->expire = maxexpire; 2190 } 2191 p->refcnt = pr->ndpr_refcnt; 2192 p->flags = pr->ndpr_stateflags; 2193 p->origin = PR_ORIG_RA; 2194 advrtrs = 0; 2195 for (pfr = pr->ndpr_advrtrs.lh_first; pfr; 2196 pfr = pfr->pfr_next) { 2197 if ((void *)&sin6[advrtrs + 1] > (void *)pe) { 2198 advrtrs++; 2199 continue; 2200 } 2201 s6 = &sin6[advrtrs]; 2202 bzero(s6, sizeof(*s6)); 2203 s6->sin6_family = AF_INET6; 2204 s6->sin6_len = sizeof(*sin6); 2205 s6->sin6_addr = pfr->router->rtaddr; 2206 if (sa6_recoverscope(s6)) { 2207 log(LOG_ERR, 2208 "scope error in " 2209 "prefix list (%s)\n", 2210 ip6_sprintf(ip6buf, 2211 &pfr->router->rtaddr)); 2212 } 2213 advrtrs++; 2214 } 2215 p->advrtrs = advrtrs; 2216 } else 2217 panic("buffer too short"); 2218 2219 advance = sizeof(*p) + sizeof(*sin6) * advrtrs; 2220 error = SYSCTL_OUT(req, buf, advance); 2221 if (error) 2222 break; 2223 } 2224 2225 return (error); 2226} 2227