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 #include <netinet6/send.h> 76 77 #include <sys/limits.h> 78 79 #include <security/mac/mac_framework.h> 80 81 #define ND6_SLOWTIMER_INTERVAL (60 * 60) /* 1 hour */ 82 #define ND6_RECALC_REACHTM_INTERVAL (60 * 120) /* 2 hours */ 83 84 #define SIN6(s) ((struct sockaddr_in6 *)s) 85 86 /* timer values */ 87 VNET_DEFINE(int, nd6_prune) = 1; /* walk list every 1 seconds */ 88 VNET_DEFINE(int, nd6_delay) = 5; /* delay first probe time 5 second */ 89 VNET_DEFINE(int, nd6_umaxtries) = 3; /* maximum unicast query */ 90 VNET_DEFINE(int, nd6_mmaxtries) = 3; /* maximum multicast query */ 91 VNET_DEFINE(int, nd6_useloopback) = 1; /* use loopback interface for 92 * local traffic */ 93 VNET_DEFINE(int, nd6_gctimer) = (60 * 60 * 24); /* 1 day: garbage 94 * collection timer */ 95 96 /* preventing too many loops in ND option parsing */ 97 static VNET_DEFINE(int, nd6_maxndopt) = 10; /* max # of ND options allowed */ 98 99 VNET_DEFINE(int, nd6_maxnudhint) = 0; /* max # of subsequent upper 100 * layer hints */ 101 static VNET_DEFINE(int, nd6_maxqueuelen) = 1; /* max pkts cached in unresolved 102 * ND entries */ 103 #define V_nd6_maxndopt VNET(nd6_maxndopt) 104 #define V_nd6_maxqueuelen VNET(nd6_maxqueuelen) 105 106 #ifdef ND6_DEBUG 107 VNET_DEFINE(int, nd6_debug) = 1; 108 #else 109 VNET_DEFINE(int, nd6_debug) = 0; 110 #endif 111 112 /* for debugging? */ 113 #if 0 114 static int nd6_inuse, nd6_allocated; 115 #endif 116 117 VNET_DEFINE(struct nd_drhead, nd_defrouter); 118 VNET_DEFINE(struct nd_prhead, nd_prefix); 119 120 VNET_DEFINE(int, nd6_recalc_reachtm_interval) = ND6_RECALC_REACHTM_INTERVAL; 121 #define V_nd6_recalc_reachtm_interval VNET(nd6_recalc_reachtm_interval) 122 123 static struct sockaddr_in6 all1_sa; 124 125 int (*send_sendso_input_hook)(struct mbuf *, struct ifnet *, int, int); 126 127 static int nd6_is_new_addr_neighbor __P((struct sockaddr_in6 *, 128 struct ifnet *)); 129 static void nd6_setmtu0(struct ifnet *, struct nd_ifinfo *); 130 static void nd6_slowtimo(void *); 131 static int regen_tmpaddr(struct in6_ifaddr *); 132 static struct llentry *nd6_free(struct llentry *, int); 133 static void nd6_llinfo_timer(void *); 134 static void clear_llinfo_pqueue(struct llentry *); 135 136 static VNET_DEFINE(struct callout, nd6_slowtimo_ch); 137 #define V_nd6_slowtimo_ch VNET(nd6_slowtimo_ch) 138 139 VNET_DEFINE(struct callout, nd6_timer_ch); 140 141 void 142 nd6_init(void) 143 { 144 int i; 145 146 LIST_INIT(&V_nd_prefix); 147 148 all1_sa.sin6_family = AF_INET6; 149 all1_sa.sin6_len = sizeof(struct sockaddr_in6); 150 for (i = 0; i < sizeof(all1_sa.sin6_addr); i++) 151 all1_sa.sin6_addr.s6_addr[i] = 0xff; 152 153 /* initialization of the default router list */ 154 TAILQ_INIT(&V_nd_defrouter); 155 156 /* start timer */ 157 callout_init(&V_nd6_slowtimo_ch, 0); 158 callout_reset(&V_nd6_slowtimo_ch, ND6_SLOWTIMER_INTERVAL * hz, 159 nd6_slowtimo, curvnet); 160 } 161 162 #ifdef VIMAGE 163 void 164 nd6_destroy() 165 { 166 167 callout_drain(&V_nd6_slowtimo_ch); 168 callout_drain(&V_nd6_timer_ch); 169 } 170 #endif 171 172 struct nd_ifinfo * 173 nd6_ifattach(struct ifnet *ifp) 174 { 175 struct nd_ifinfo *nd; 176 177 nd = (struct nd_ifinfo *)malloc(sizeof(*nd), M_IP6NDP, M_WAITOK); 178 bzero(nd, sizeof(*nd)); 179 180 nd->initialized = 1; 181 182 nd->chlim = IPV6_DEFHLIM; 183 nd->basereachable = REACHABLE_TIME; 184 nd->reachable = ND_COMPUTE_RTIME(nd->basereachable); 185 nd->retrans = RETRANS_TIMER; 186 187 nd->flags = ND6_IFF_PERFORMNUD; 188 189 /* A loopback interface always has ND6_IFF_AUTO_LINKLOCAL. */ 190 if (V_ip6_auto_linklocal || (ifp->if_flags & IFF_LOOPBACK)) 191 nd->flags |= ND6_IFF_AUTO_LINKLOCAL; 192 193 /* A loopback interface does not need to accept RTADV. */ 194 if (V_ip6_accept_rtadv && !(ifp->if_flags & IFF_LOOPBACK)) 195 nd->flags |= ND6_IFF_ACCEPT_RTADV; 196 if (V_ip6_no_radr && !(ifp->if_flags & IFF_LOOPBACK)) 197 nd->flags |= ND6_IFF_NO_RADR; 198 199 /* XXX: we cannot call nd6_setmtu since ifp is not fully initialized */ 200 nd6_setmtu0(ifp, nd); 201 202 return nd; 203 } 204 205 void 206 nd6_ifdetach(struct nd_ifinfo *nd) 207 { 208 209 free(nd, M_IP6NDP); 210 } 211 212 /* 213 * Reset ND level link MTU. This function is called when the physical MTU 214 * changes, which means we might have to adjust the ND level MTU. 215 */ 216 void 217 nd6_setmtu(struct ifnet *ifp) 218 { 219 220 nd6_setmtu0(ifp, ND_IFINFO(ifp)); 221 } 222 223 /* XXX todo: do not maintain copy of ifp->if_mtu in ndi->maxmtu */ 224 void 225 nd6_setmtu0(struct ifnet *ifp, struct nd_ifinfo *ndi) 226 { 227 u_int32_t omaxmtu; 228 229 omaxmtu = ndi->maxmtu; 230 231 switch (ifp->if_type) { 232 case IFT_ARCNET: 233 ndi->maxmtu = MIN(ARC_PHDS_MAXMTU, ifp->if_mtu); /* RFC2497 */ 234 break; 235 case IFT_FDDI: 236 ndi->maxmtu = MIN(FDDIIPMTU, ifp->if_mtu); /* RFC2467 */ 237 break; 238 case IFT_ISO88025: 239 ndi->maxmtu = MIN(ISO88025_MAX_MTU, ifp->if_mtu); 240 break; 241 default: 242 ndi->maxmtu = ifp->if_mtu; 243 break; 244 } 245 246 /* 247 * Decreasing the interface MTU under IPV6 minimum MTU may cause 248 * undesirable situation. We thus notify the operator of the change 249 * explicitly. The check for omaxmtu is necessary to restrict the 250 * log to the case of changing the MTU, not initializing it. 251 */ 252 if (omaxmtu >= IPV6_MMTU && ndi->maxmtu < IPV6_MMTU) { 253 log(LOG_NOTICE, "nd6_setmtu0: " 254 "new link MTU on %s (%lu) is too small for IPv6\n", 255 if_name(ifp), (unsigned long)ndi->maxmtu); 256 } 257 258 if (ndi->maxmtu > V_in6_maxmtu) 259 in6_setmaxmtu(); /* check all interfaces just in case */ 260 261 } 262 263 void 264 nd6_option_init(void *opt, int icmp6len, union nd_opts *ndopts) 265 { 266 267 bzero(ndopts, sizeof(*ndopts)); 268 ndopts->nd_opts_search = (struct nd_opt_hdr *)opt; 269 ndopts->nd_opts_last 270 = (struct nd_opt_hdr *)(((u_char *)opt) + icmp6len); 271 272 if (icmp6len == 0) { 273 ndopts->nd_opts_done = 1; 274 ndopts->nd_opts_search = NULL; 275 } 276 } 277 278 /* 279 * Take one ND option. 280 */ 281 struct nd_opt_hdr * 282 nd6_option(union nd_opts *ndopts) 283 { 284 struct nd_opt_hdr *nd_opt; 285 int olen; 286 287 if (ndopts == NULL) 288 panic("ndopts == NULL in nd6_option"); 289 if (ndopts->nd_opts_last == NULL) 290 panic("uninitialized ndopts in nd6_option"); 291 if (ndopts->nd_opts_search == NULL) 292 return NULL; 293 if (ndopts->nd_opts_done) 294 return NULL; 295 296 nd_opt = ndopts->nd_opts_search; 297 298 /* make sure nd_opt_len is inside the buffer */ 299 if ((caddr_t)&nd_opt->nd_opt_len >= (caddr_t)ndopts->nd_opts_last) { 300 bzero(ndopts, sizeof(*ndopts)); 301 return NULL; 302 } 303 304 olen = nd_opt->nd_opt_len << 3; 305 if (olen == 0) { 306 /* 307 * Message validation requires that all included 308 * options have a length that is greater than zero. 309 */ 310 bzero(ndopts, sizeof(*ndopts)); 311 return NULL; 312 } 313 314 ndopts->nd_opts_search = (struct nd_opt_hdr *)((caddr_t)nd_opt + olen); 315 if (ndopts->nd_opts_search > ndopts->nd_opts_last) { 316 /* option overruns the end of buffer, invalid */ 317 bzero(ndopts, sizeof(*ndopts)); 318 return NULL; 319 } else if (ndopts->nd_opts_search == ndopts->nd_opts_last) { 320 /* reached the end of options chain */ 321 ndopts->nd_opts_done = 1; 322 ndopts->nd_opts_search = NULL; 323 } 324 return nd_opt; 325 } 326 327 /* 328 * Parse multiple ND options. 329 * This function is much easier to use, for ND routines that do not need 330 * multiple options of the same type. 331 */ 332 int 333 nd6_options(union nd_opts *ndopts) 334 { 335 struct nd_opt_hdr *nd_opt; 336 int i = 0; 337 338 if (ndopts == NULL) 339 panic("ndopts == NULL in nd6_options"); 340 if (ndopts->nd_opts_last == NULL) 341 panic("uninitialized ndopts in nd6_options"); 342 if (ndopts->nd_opts_search == NULL) 343 return 0; 344 345 while (1) { 346 nd_opt = nd6_option(ndopts); 347 if (nd_opt == NULL && ndopts->nd_opts_last == NULL) { 348 /* 349 * Message validation requires that all included 350 * options have a length that is greater than zero. 351 */ 352 ICMP6STAT_INC(icp6s_nd_badopt); 353 bzero(ndopts, sizeof(*ndopts)); 354 return -1; 355 } 356 357 if (nd_opt == NULL) 358 goto skip1; 359 360 switch (nd_opt->nd_opt_type) { 361 case ND_OPT_SOURCE_LINKADDR: 362 case ND_OPT_TARGET_LINKADDR: 363 case ND_OPT_MTU: 364 case ND_OPT_REDIRECTED_HEADER: 365 if (ndopts->nd_opt_array[nd_opt->nd_opt_type]) { 366 nd6log((LOG_INFO, 367 "duplicated ND6 option found (type=%d)\n", 368 nd_opt->nd_opt_type)); 369 /* XXX bark? */ 370 } else { 371 ndopts->nd_opt_array[nd_opt->nd_opt_type] 372 = nd_opt; 373 } 374 break; 375 case ND_OPT_PREFIX_INFORMATION: 376 if (ndopts->nd_opt_array[nd_opt->nd_opt_type] == 0) { 377 ndopts->nd_opt_array[nd_opt->nd_opt_type] 378 = nd_opt; 379 } 380 ndopts->nd_opts_pi_end = 381 (struct nd_opt_prefix_info *)nd_opt; 382 break; 383 default: 384 /* 385 * Unknown options must be silently ignored, 386 * to accomodate future extension to the protocol. 387 */ 388 nd6log((LOG_DEBUG, 389 "nd6_options: unsupported option %d - " 390 "option ignored\n", nd_opt->nd_opt_type)); 391 } 392 393 skip1: 394 i++; 395 if (i > V_nd6_maxndopt) { 396 ICMP6STAT_INC(icp6s_nd_toomanyopt); 397 nd6log((LOG_INFO, "too many loop in nd opt\n")); 398 break; 399 } 400 401 if (ndopts->nd_opts_done) 402 break; 403 } 404 405 return 0; 406 } 407 408 /* 409 * ND6 timer routine to handle ND6 entries 410 */ 411 void 412 nd6_llinfo_settimer_locked(struct llentry *ln, long tick) 413 { 414 int canceled; 415 416 LLE_WLOCK_ASSERT(ln); 417 418 if (tick < 0) { 419 ln->la_expire = 0; 420 ln->ln_ntick = 0; 421 canceled = callout_stop(&ln->ln_timer_ch); 422 } else { 423 ln->la_expire = time_second + tick / hz; 424 LLE_ADDREF(ln); 425 if (tick > INT_MAX) { 426 ln->ln_ntick = tick - INT_MAX; 427 canceled = callout_reset(&ln->ln_timer_ch, INT_MAX, 428 nd6_llinfo_timer, ln); 429 } else { 430 ln->ln_ntick = 0; 431 canceled = callout_reset(&ln->ln_timer_ch, tick, 432 nd6_llinfo_timer, ln); 433 } 434 } 435 if (canceled) 436 LLE_REMREF(ln); 437 } 438 439 void 440 nd6_llinfo_settimer(struct llentry *ln, long tick) 441 { 442 443 LLE_WLOCK(ln); 444 nd6_llinfo_settimer_locked(ln, tick); 445 LLE_WUNLOCK(ln); 446 } 447 448 static void 449 nd6_llinfo_timer(void *arg) 450 { 451 struct llentry *ln; 452 struct in6_addr *dst; 453 struct ifnet *ifp; 454 struct nd_ifinfo *ndi = NULL; 455 456 KASSERT(arg != NULL, ("%s: arg NULL", __func__)); 457 ln = (struct llentry *)arg; 458 LLE_WLOCK_ASSERT(ln); 459 ifp = ln->lle_tbl->llt_ifp; 460 461 CURVNET_SET(ifp->if_vnet); 462 463 if (ln->ln_ntick > 0) { 464 if (ln->ln_ntick > INT_MAX) { 465 ln->ln_ntick -= INT_MAX; 466 nd6_llinfo_settimer_locked(ln, INT_MAX); 467 } else { 468 ln->ln_ntick = 0; 469 nd6_llinfo_settimer_locked(ln, ln->ln_ntick); 470 } 471 goto done; 472 } 473 474 ndi = ND_IFINFO(ifp); 475 dst = &L3_ADDR_SIN6(ln)->sin6_addr; 476 if (ln->la_flags & LLE_STATIC) { 477 goto done; 478 } 479 480 if (ln->la_flags & LLE_DELETED) { 481 (void)nd6_free(ln, 0); 482 ln = NULL; 483 goto done; 484 } 485 486 switch (ln->ln_state) { 487 case ND6_LLINFO_INCOMPLETE: 488 if (ln->la_asked < V_nd6_mmaxtries) { 489 ln->la_asked++; 490 nd6_llinfo_settimer_locked(ln, (long)ndi->retrans * hz / 1000); 491 LLE_WUNLOCK(ln); 492 nd6_ns_output(ifp, NULL, dst, ln, 0); 493 LLE_WLOCK(ln); 494 } else { 495 struct mbuf *m = ln->la_hold; 496 if (m) { 497 struct mbuf *m0; 498 499 /* 500 * assuming every packet in la_hold has the 501 * same IP header. Send error after unlock. 502 */ 503 m0 = m->m_nextpkt; 504 m->m_nextpkt = NULL; 505 ln->la_hold = m0; 506 clear_llinfo_pqueue(ln); 507 } 508 (void)nd6_free(ln, 0); 509 ln = NULL; 510 if (m != NULL) 511 icmp6_error2(m, ICMP6_DST_UNREACH, 512 ICMP6_DST_UNREACH_ADDR, 0, ifp); 513 } 514 break; 515 case ND6_LLINFO_REACHABLE: 516 if (!ND6_LLINFO_PERMANENT(ln)) { 517 ln->ln_state = ND6_LLINFO_STALE; 518 nd6_llinfo_settimer_locked(ln, (long)V_nd6_gctimer * hz); 519 } 520 break; 521 522 case ND6_LLINFO_STALE: 523 /* Garbage Collection(RFC 2461 5.3) */ 524 if (!ND6_LLINFO_PERMANENT(ln)) { 525 (void)nd6_free(ln, 1); 526 ln = NULL; 527 } 528 break; 529 530 case ND6_LLINFO_DELAY: 531 if (ndi && (ndi->flags & ND6_IFF_PERFORMNUD) != 0) { 532 /* We need NUD */ 533 ln->la_asked = 1; 534 ln->ln_state = ND6_LLINFO_PROBE; 535 nd6_llinfo_settimer_locked(ln, (long)ndi->retrans * hz / 1000); 536 LLE_WUNLOCK(ln); 537 nd6_ns_output(ifp, dst, dst, ln, 0); 538 LLE_WLOCK(ln); 539 } else { 540 ln->ln_state = ND6_LLINFO_STALE; /* XXX */ 541 nd6_llinfo_settimer_locked(ln, (long)V_nd6_gctimer * hz); 542 } 543 break; 544 case ND6_LLINFO_PROBE: 545 if (ln->la_asked < V_nd6_umaxtries) { 546 ln->la_asked++; 547 nd6_llinfo_settimer_locked(ln, (long)ndi->retrans * hz / 1000); 548 LLE_WUNLOCK(ln); 549 nd6_ns_output(ifp, dst, dst, ln, 0); 550 LLE_WLOCK(ln); 551 } else { 552 (void)nd6_free(ln, 0); 553 ln = NULL; 554 } 555 break; 556 default: 557 panic("%s: paths in a dark night can be confusing: %d", 558 __func__, ln->ln_state); 559 } 560 done: 561 if (ln != NULL) 562 LLE_FREE_LOCKED(ln); 563 CURVNET_RESTORE(); 564 } 565 566 567 /* 568 * ND6 timer routine to expire default route list and prefix list 569 */ 570 void 571 nd6_timer(void *arg) 572 { 573 CURVNET_SET((struct vnet *) arg); 574 int s; 575 struct nd_defrouter *dr, *ndr; 576 struct nd_prefix *pr, *npr; 577 struct in6_ifaddr *ia6, *nia6; 578 579 callout_reset(&V_nd6_timer_ch, V_nd6_prune * hz, 580 nd6_timer, curvnet); 581 582 /* expire default router list */ 583 s = splnet(); 584 TAILQ_FOREACH_SAFE(dr, &V_nd_defrouter, dr_entry, ndr) { 585 if (dr->expire && dr->expire < time_second) 586 defrtrlist_del(dr); 587 } 588 589 /* 590 * expire interface addresses. 591 * in the past the loop was inside prefix expiry processing. 592 * However, from a stricter speci-confrmance standpoint, we should 593 * rather separate address lifetimes and prefix lifetimes. 594 * 595 * XXXRW: in6_ifaddrhead locking. 596 */ 597 addrloop: 598 TAILQ_FOREACH_SAFE(ia6, &V_in6_ifaddrhead, ia_link, nia6) { 599 /* check address lifetime */ 600 if (IFA6_IS_INVALID(ia6)) { 601 int regen = 0; 602 603 /* 604 * If the expiring address is temporary, try 605 * regenerating a new one. This would be useful when 606 * we suspended a laptop PC, then turned it on after a 607 * period that could invalidate all temporary 608 * addresses. Although we may have to restart the 609 * loop (see below), it must be after purging the 610 * address. Otherwise, we'd see an infinite loop of 611 * regeneration. 612 */ 613 if (V_ip6_use_tempaddr && 614 (ia6->ia6_flags & IN6_IFF_TEMPORARY) != 0) { 615 if (regen_tmpaddr(ia6) == 0) 616 regen = 1; 617 } 618 619 in6_purgeaddr(&ia6->ia_ifa); 620 621 if (regen) 622 goto addrloop; /* XXX: see below */ 623 } else if (IFA6_IS_DEPRECATED(ia6)) { 624 int oldflags = ia6->ia6_flags; 625 626 ia6->ia6_flags |= IN6_IFF_DEPRECATED; 627 628 /* 629 * If a temporary address has just become deprecated, 630 * regenerate a new one if possible. 631 */ 632 if (V_ip6_use_tempaddr && 633 (ia6->ia6_flags & IN6_IFF_TEMPORARY) != 0 && 634 (oldflags & IN6_IFF_DEPRECATED) == 0) { 635 636 if (regen_tmpaddr(ia6) == 0) { 637 /* 638 * A new temporary address is 639 * generated. 640 * XXX: this means the address chain 641 * has changed while we are still in 642 * the loop. Although the change 643 * would not cause disaster (because 644 * it's not a deletion, but an 645 * addition,) we'd rather restart the 646 * loop just for safety. Or does this 647 * significantly reduce performance?? 648 */ 649 goto addrloop; 650 } 651 } 652 } else { 653 /* 654 * A new RA might have made a deprecated address 655 * preferred. 656 */ 657 ia6->ia6_flags &= ~IN6_IFF_DEPRECATED; 658 } 659 } 660 661 /* expire prefix list */ 662 LIST_FOREACH_SAFE(pr, &V_nd_prefix, ndpr_entry, npr) { 663 /* 664 * check prefix lifetime. 665 * since pltime is just for autoconf, pltime processing for 666 * prefix is not necessary. 667 */ 668 if (pr->ndpr_vltime != ND6_INFINITE_LIFETIME && 669 time_second - pr->ndpr_lastupdate > pr->ndpr_vltime) { 670 671 /* 672 * address expiration and prefix expiration are 673 * separate. NEVER perform in6_purgeaddr here. 674 */ 675 prelist_remove(pr); 676 } 677 } 678 splx(s); 679 CURVNET_RESTORE(); 680 } 681 682 /* 683 * ia6 - deprecated/invalidated temporary address 684 */ 685 static int 686 regen_tmpaddr(struct in6_ifaddr *ia6) 687 { 688 struct ifaddr *ifa; 689 struct ifnet *ifp; 690 struct in6_ifaddr *public_ifa6 = NULL; 691 692 ifp = ia6->ia_ifa.ifa_ifp; 693 IF_ADDR_RLOCK(ifp); 694 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 695 struct in6_ifaddr *it6; 696 697 if (ifa->ifa_addr->sa_family != AF_INET6) 698 continue; 699 700 it6 = (struct in6_ifaddr *)ifa; 701 702 /* ignore no autoconf addresses. */ 703 if ((it6->ia6_flags & IN6_IFF_AUTOCONF) == 0) 704 continue; 705 706 /* ignore autoconf addresses with different prefixes. */ 707 if (it6->ia6_ndpr == NULL || it6->ia6_ndpr != ia6->ia6_ndpr) 708 continue; 709 710 /* 711 * Now we are looking at an autoconf address with the same 712 * prefix as ours. If the address is temporary and is still 713 * preferred, do not create another one. It would be rare, but 714 * could happen, for example, when we resume a laptop PC after 715 * a long period. 716 */ 717 if ((it6->ia6_flags & IN6_IFF_TEMPORARY) != 0 && 718 !IFA6_IS_DEPRECATED(it6)) { 719 public_ifa6 = NULL; 720 break; 721 } 722 723 /* 724 * This is a public autoconf address that has the same prefix 725 * as ours. If it is preferred, keep it. We can't break the 726 * loop here, because there may be a still-preferred temporary 727 * address with the prefix. 728 */ 729 if (!IFA6_IS_DEPRECATED(it6)) 730 public_ifa6 = it6; 731 732 if (public_ifa6 != NULL) 733 ifa_ref(&public_ifa6->ia_ifa); 734 } 735 IF_ADDR_RUNLOCK(ifp); 736 737 if (public_ifa6 != NULL) { 738 int e; 739 740 if ((e = in6_tmpifadd(public_ifa6, 0, 0)) != 0) { 741 ifa_free(&public_ifa6->ia_ifa); 742 log(LOG_NOTICE, "regen_tmpaddr: failed to create a new" 743 " tmp addr,errno=%d\n", e); 744 return (-1); 745 } 746 ifa_free(&public_ifa6->ia_ifa); 747 return (0); 748 } 749 750 return (-1); 751 } 752 753 /* 754 * Nuke neighbor cache/prefix/default router management table, right before 755 * ifp goes away. 756 */ 757 void 758 nd6_purge(struct ifnet *ifp) 759 { 760 struct nd_defrouter *dr, *ndr; 761 struct nd_prefix *pr, *npr; 762 763 /* 764 * Nuke default router list entries toward ifp. 765 * We defer removal of default router list entries that is installed 766 * in the routing table, in order to keep additional side effects as 767 * small as possible. 768 */ 769 TAILQ_FOREACH_SAFE(dr, &V_nd_defrouter, dr_entry, ndr) { 770 if (dr->installed) 771 continue; 772 773 if (dr->ifp == ifp) 774 defrtrlist_del(dr); 775 } 776 777 TAILQ_FOREACH_SAFE(dr, &V_nd_defrouter, dr_entry, ndr) { 778 if (!dr->installed) 779 continue; 780 781 if (dr->ifp == ifp) 782 defrtrlist_del(dr); 783 } 784 785 /* Nuke prefix list entries toward ifp */ 786 LIST_FOREACH_SAFE(pr, &V_nd_prefix, ndpr_entry, npr) { 787 if (pr->ndpr_ifp == ifp) { 788 /* 789 * Because if_detach() does *not* release prefixes 790 * while purging addresses the reference count will 791 * still be above zero. We therefore reset it to 792 * make sure that the prefix really gets purged. 793 */ 794 pr->ndpr_refcnt = 0; 795 796 /* 797 * Previously, pr->ndpr_addr is removed as well, 798 * but I strongly believe we don't have to do it. 799 * nd6_purge() is only called from in6_ifdetach(), 800 * which removes all the associated interface addresses 801 * by itself. 802 * (jinmei@kame.net 20010129) 803 */ 804 prelist_remove(pr); 805 } 806 } 807 808 /* cancel default outgoing interface setting */ 809 if (V_nd6_defifindex == ifp->if_index) 810 nd6_setdefaultiface(0); 811 812 if (ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV) { 813 /* Refresh default router list. */ 814 defrouter_select(); 815 } 816 817 /* XXXXX 818 * We do not nuke the neighbor cache entries here any more 819 * because the neighbor cache is kept in if_afdata[AF_INET6]. 820 * nd6_purge() is invoked by in6_ifdetach() which is called 821 * from if_detach() where everything gets purged. So let 822 * in6_domifdetach() do the actual L2 table purging work. 823 */ 824 } 825 826 /* 827 * the caller acquires and releases the lock on the lltbls 828 * Returns the llentry locked 829 */ 830 struct llentry * 831 nd6_lookup(struct in6_addr *addr6, int flags, struct ifnet *ifp) 832 { 833 struct sockaddr_in6 sin6; 834 struct llentry *ln; 835 int llflags; 836 837 bzero(&sin6, sizeof(sin6)); 838 sin6.sin6_len = sizeof(struct sockaddr_in6); 839 sin6.sin6_family = AF_INET6; 840 sin6.sin6_addr = *addr6; 841 842 IF_AFDATA_LOCK_ASSERT(ifp); 843 844 llflags = 0; 845 if (flags & ND6_CREATE) 846 llflags |= LLE_CREATE; 847 if (flags & ND6_EXCLUSIVE) 848 llflags |= LLE_EXCLUSIVE; 849 850 ln = lla_lookup(LLTABLE6(ifp), llflags, (struct sockaddr *)&sin6); 851 if ((ln != NULL) && (llflags & LLE_CREATE)) 852 ln->ln_state = ND6_LLINFO_NOSTATE; 853 854 return (ln); 855 } 856 857 /* 858 * Test whether a given IPv6 address is a neighbor or not, ignoring 859 * the actual neighbor cache. The neighbor cache is ignored in order 860 * to not reenter the routing code from within itself. 861 */ 862 static int 863 nd6_is_new_addr_neighbor(struct sockaddr_in6 *addr, struct ifnet *ifp) 864 { 865 struct nd_prefix *pr; 866 struct ifaddr *dstaddr; 867 868 /* 869 * A link-local address is always a neighbor. 870 * XXX: a link does not necessarily specify a single interface. 871 */ 872 if (IN6_IS_ADDR_LINKLOCAL(&addr->sin6_addr)) { 873 struct sockaddr_in6 sin6_copy; 874 u_int32_t zone; 875 876 /* 877 * We need sin6_copy since sa6_recoverscope() may modify the 878 * content (XXX). 879 */ 880 sin6_copy = *addr; 881 if (sa6_recoverscope(&sin6_copy)) 882 return (0); /* XXX: should be impossible */ 883 if (in6_setscope(&sin6_copy.sin6_addr, ifp, &zone)) 884 return (0); 885 if (sin6_copy.sin6_scope_id == zone) 886 return (1); 887 else 888 return (0); 889 } 890 891 /* 892 * If the address matches one of our addresses, 893 * it should be a neighbor. 894 * If the address matches one of our on-link prefixes, it should be a 895 * neighbor. 896 */ 897 LIST_FOREACH(pr, &V_nd_prefix, ndpr_entry) { 898 if (pr->ndpr_ifp != ifp) 899 continue; 900 901 if (!(pr->ndpr_stateflags & NDPRF_ONLINK)) { 902 struct rtentry *rt; 903 904 /* Always use the default FIB here. */ 905 rt = in6_rtalloc1((struct sockaddr *)&pr->ndpr_prefix, 906 0, 0, RT_DEFAULT_FIB); 907 if (rt == NULL) 908 continue; 909 /* 910 * This is the case where multiple interfaces 911 * have the same prefix, but only one is installed 912 * into the routing table and that prefix entry 913 * is not the one being examined here. In the case 914 * where RADIX_MPATH is enabled, multiple route 915 * entries (of the same rt_key value) will be 916 * installed because the interface addresses all 917 * differ. 918 */ 919 if (!IN6_ARE_ADDR_EQUAL(&pr->ndpr_prefix.sin6_addr, 920 &((struct sockaddr_in6 *)rt_key(rt))->sin6_addr)) { 921 RTFREE_LOCKED(rt); 922 continue; 923 } 924 RTFREE_LOCKED(rt); 925 } 926 927 if (IN6_ARE_MASKED_ADDR_EQUAL(&pr->ndpr_prefix.sin6_addr, 928 &addr->sin6_addr, &pr->ndpr_mask)) 929 return (1); 930 } 931 932 /* 933 * If the address is assigned on the node of the other side of 934 * a p2p interface, the address should be a neighbor. 935 */ 936 dstaddr = ifa_ifwithdstaddr((struct sockaddr *)addr); 937 if (dstaddr != NULL) { 938 if (dstaddr->ifa_ifp == ifp) { 939 ifa_free(dstaddr); 940 return (1); 941 } 942 ifa_free(dstaddr); 943 } 944 945 /* 946 * If the default router list is empty, all addresses are regarded 947 * as on-link, and thus, as a neighbor. 948 */ 949 if (ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV && 950 TAILQ_EMPTY(&V_nd_defrouter) && 951 V_nd6_defifindex == ifp->if_index) { 952 return (1); 953 } 954 955 return (0); 956 } 957 958 959 /* 960 * Detect if a given IPv6 address identifies a neighbor on a given link. 961 * XXX: should take care of the destination of a p2p link? 962 */ 963 int 964 nd6_is_addr_neighbor(struct sockaddr_in6 *addr, struct ifnet *ifp) 965 { 966 struct llentry *lle; 967 int rc = 0; 968 969 IF_AFDATA_UNLOCK_ASSERT(ifp); 970 if (nd6_is_new_addr_neighbor(addr, ifp)) 971 return (1); 972 973 /* 974 * Even if the address matches none of our addresses, it might be 975 * in the neighbor cache. 976 */ 977 IF_AFDATA_LOCK(ifp); 978 if ((lle = nd6_lookup(&addr->sin6_addr, 0, ifp)) != NULL) { 979 LLE_RUNLOCK(lle); 980 rc = 1; 981 } 982 IF_AFDATA_UNLOCK(ifp); 983 return (rc); 984 } 985 986 /* 987 * Free an nd6 llinfo entry. 988 * Since the function would cause significant changes in the kernel, DO NOT 989 * make it global, unless you have a strong reason for the change, and are sure 990 * that the change is safe. 991 */ 992 static struct llentry * 993 nd6_free(struct llentry *ln, int gc) 994 { 995 struct llentry *next; 996 struct nd_defrouter *dr; 997 struct ifnet *ifp; 998 999 LLE_WLOCK_ASSERT(ln); 1000 1001 /* 1002 * we used to have pfctlinput(PRC_HOSTDEAD) here. 1003 * even though it is not harmful, it was not really necessary. 1004 */ 1005 1006 /* cancel timer */ 1007 nd6_llinfo_settimer_locked(ln, -1); 1008 1009 ifp = ln->lle_tbl->llt_ifp; 1010 1011 if (ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV) { 1012 dr = defrouter_lookup(&L3_ADDR_SIN6(ln)->sin6_addr, ifp); 1013 1014 if (dr != NULL && dr->expire && 1015 ln->ln_state == ND6_LLINFO_STALE && gc) { 1016 /* 1017 * If the reason for the deletion is just garbage 1018 * collection, and the neighbor is an active default 1019 * router, do not delete it. Instead, reset the GC 1020 * timer using the router's lifetime. 1021 * Simply deleting the entry would affect default 1022 * router selection, which is not necessarily a good 1023 * thing, especially when we're using router preference 1024 * values. 1025 * XXX: the check for ln_state would be redundant, 1026 * but we intentionally keep it just in case. 1027 */ 1028 if (dr->expire > time_second) 1029 nd6_llinfo_settimer_locked(ln, 1030 (dr->expire - time_second) * hz); 1031 else 1032 nd6_llinfo_settimer_locked(ln, 1033 (long)V_nd6_gctimer * hz); 1034 1035 next = LIST_NEXT(ln, lle_next); 1036 LLE_REMREF(ln); 1037 LLE_WUNLOCK(ln); 1038 return (next); 1039 } 1040 1041 if (dr) { 1042 /* 1043 * Unreachablity of a router might affect the default 1044 * router selection and on-link detection of advertised 1045 * prefixes. 1046 */ 1047 1048 /* 1049 * Temporarily fake the state to choose a new default 1050 * router and to perform on-link determination of 1051 * prefixes correctly. 1052 * Below the state will be set correctly, 1053 * or the entry itself will be deleted. 1054 */ 1055 ln->ln_state = ND6_LLINFO_INCOMPLETE; 1056 } 1057 1058 if (ln->ln_router || dr) { 1059 1060 /* 1061 * We need to unlock to avoid a LOR with rt6_flush() with the 1062 * rnh and for the calls to pfxlist_onlink_check() and 1063 * defrouter_select() in the block further down for calls 1064 * into nd6_lookup(). We still hold a ref. 1065 */ 1066 LLE_WUNLOCK(ln); 1067 1068 /* 1069 * rt6_flush must be called whether or not the neighbor 1070 * is in the Default Router List. 1071 * See a corresponding comment in nd6_na_input(). 1072 */ 1073 rt6_flush(&L3_ADDR_SIN6(ln)->sin6_addr, ifp); 1074 } 1075 1076 if (dr) { 1077 /* 1078 * Since defrouter_select() does not affect the 1079 * on-link determination and MIP6 needs the check 1080 * before the default router selection, we perform 1081 * the check now. 1082 */ 1083 pfxlist_onlink_check(); 1084 1085 /* 1086 * Refresh default router list. 1087 */ 1088 defrouter_select(); 1089 } 1090 1091 if (ln->ln_router || dr) 1092 LLE_WLOCK(ln); 1093 } 1094 1095 /* 1096 * Before deleting the entry, remember the next entry as the 1097 * return value. We need this because pfxlist_onlink_check() above 1098 * might have freed other entries (particularly the old next entry) as 1099 * a side effect (XXX). 1100 */ 1101 next = LIST_NEXT(ln, lle_next); 1102 1103 /* 1104 * Save to unlock. We still hold an extra reference and will not 1105 * free(9) in llentry_free() if someone else holds one as well. 1106 */ 1107 LLE_WUNLOCK(ln); 1108 IF_AFDATA_LOCK(ifp); 1109 LLE_WLOCK(ln); 1110 LLE_REMREF(ln); 1111 llentry_free(ln); 1112 IF_AFDATA_UNLOCK(ifp); 1113 1114 return (next); 1115 } 1116 1117 /* 1118 * Upper-layer reachability hint for Neighbor Unreachability Detection. 1119 * 1120 * XXX cost-effective methods? 1121 */ 1122 void 1123 nd6_nud_hint(struct rtentry *rt, struct in6_addr *dst6, int force) 1124 { 1125 struct llentry *ln; 1126 struct ifnet *ifp; 1127 1128 if ((dst6 == NULL) || (rt == NULL)) 1129 return; 1130 1131 ifp = rt->rt_ifp; 1132 IF_AFDATA_LOCK(ifp); 1133 ln = nd6_lookup(dst6, ND6_EXCLUSIVE, NULL); 1134 IF_AFDATA_UNLOCK(ifp); 1135 if (ln == NULL) 1136 return; 1137 1138 if (ln->ln_state < ND6_LLINFO_REACHABLE) 1139 goto done; 1140 1141 /* 1142 * if we get upper-layer reachability confirmation many times, 1143 * it is possible we have false information. 1144 */ 1145 if (!force) { 1146 ln->ln_byhint++; 1147 if (ln->ln_byhint > V_nd6_maxnudhint) { 1148 goto done; 1149 } 1150 } 1151 1152 ln->ln_state = ND6_LLINFO_REACHABLE; 1153 if (!ND6_LLINFO_PERMANENT(ln)) { 1154 nd6_llinfo_settimer_locked(ln, 1155 (long)ND_IFINFO(rt->rt_ifp)->reachable * hz); 1156 } 1157 done: 1158 LLE_WUNLOCK(ln); 1159 } 1160 1161 1162 /* 1163 * Rejuvenate this function for routing operations related 1164 * processing. 1165 */ 1166 void 1167 nd6_rtrequest(int req, struct rtentry *rt, struct rt_addrinfo *info) 1168 { 1169 struct sockaddr_in6 *gateway = (struct sockaddr_in6 *)rt->rt_gateway; 1170 struct nd_defrouter *dr; 1171 struct ifnet *ifp = rt->rt_ifp; 1172 1173 RT_LOCK_ASSERT(rt); 1174 1175 switch (req) { 1176 case RTM_ADD: 1177 break; 1178 1179 case RTM_DELETE: 1180 if (!ifp) 1181 return; 1182 /* 1183 * Only indirect routes are interesting. 1184 */ 1185 if ((rt->rt_flags & RTF_GATEWAY) == 0) 1186 return; 1187 /* 1188 * check for default route 1189 */ 1190 if (IN6_ARE_ADDR_EQUAL(&in6addr_any, 1191 &SIN6(rt_key(rt))->sin6_addr)) { 1192 1193 dr = defrouter_lookup(&gateway->sin6_addr, ifp); 1194 if (dr != NULL) 1195 dr->installed = 0; 1196 } 1197 break; 1198 } 1199 } 1200 1201 1202 int 1203 nd6_ioctl(u_long cmd, caddr_t data, struct ifnet *ifp) 1204 { 1205 struct in6_drlist *drl = (struct in6_drlist *)data; 1206 struct in6_oprlist *oprl = (struct in6_oprlist *)data; 1207 struct in6_ndireq *ndi = (struct in6_ndireq *)data; 1208 struct in6_nbrinfo *nbi = (struct in6_nbrinfo *)data; 1209 struct in6_ndifreq *ndif = (struct in6_ndifreq *)data; 1210 struct nd_defrouter *dr; 1211 struct nd_prefix *pr; 1212 int i = 0, error = 0; 1213 int s; 1214 1215 switch (cmd) { 1216 case SIOCGDRLST_IN6: 1217 /* 1218 * obsolete API, use sysctl under net.inet6.icmp6 1219 */ 1220 bzero(drl, sizeof(*drl)); 1221 s = splnet(); 1222 TAILQ_FOREACH(dr, &V_nd_defrouter, dr_entry) { 1223 if (i >= DRLSTSIZ) 1224 break; 1225 drl->defrouter[i].rtaddr = dr->rtaddr; 1226 in6_clearscope(&drl->defrouter[i].rtaddr); 1227 1228 drl->defrouter[i].flags = dr->flags; 1229 drl->defrouter[i].rtlifetime = dr->rtlifetime; 1230 drl->defrouter[i].expire = dr->expire; 1231 drl->defrouter[i].if_index = dr->ifp->if_index; 1232 i++; 1233 } 1234 splx(s); 1235 break; 1236 case SIOCGPRLST_IN6: 1237 /* 1238 * obsolete API, use sysctl under net.inet6.icmp6 1239 * 1240 * XXX the structure in6_prlist was changed in backward- 1241 * incompatible manner. in6_oprlist is used for SIOCGPRLST_IN6, 1242 * in6_prlist is used for nd6_sysctl() - fill_prlist(). 1243 */ 1244 /* 1245 * XXX meaning of fields, especialy "raflags", is very 1246 * differnet between RA prefix list and RR/static prefix list. 1247 * how about separating ioctls into two? 1248 */ 1249 bzero(oprl, sizeof(*oprl)); 1250 s = splnet(); 1251 LIST_FOREACH(pr, &V_nd_prefix, ndpr_entry) { 1252 struct nd_pfxrouter *pfr; 1253 int j; 1254 1255 if (i >= PRLSTSIZ) 1256 break; 1257 oprl->prefix[i].prefix = pr->ndpr_prefix.sin6_addr; 1258 oprl->prefix[i].raflags = pr->ndpr_raf; 1259 oprl->prefix[i].prefixlen = pr->ndpr_plen; 1260 oprl->prefix[i].vltime = pr->ndpr_vltime; 1261 oprl->prefix[i].pltime = pr->ndpr_pltime; 1262 oprl->prefix[i].if_index = pr->ndpr_ifp->if_index; 1263 if (pr->ndpr_vltime == ND6_INFINITE_LIFETIME) 1264 oprl->prefix[i].expire = 0; 1265 else { 1266 time_t maxexpire; 1267 1268 /* XXX: we assume time_t is signed. */ 1269 maxexpire = (-1) & 1270 ~((time_t)1 << 1271 ((sizeof(maxexpire) * 8) - 1)); 1272 if (pr->ndpr_vltime < 1273 maxexpire - pr->ndpr_lastupdate) { 1274 oprl->prefix[i].expire = 1275 pr->ndpr_lastupdate + 1276 pr->ndpr_vltime; 1277 } else 1278 oprl->prefix[i].expire = maxexpire; 1279 } 1280 1281 j = 0; 1282 LIST_FOREACH(pfr, &pr->ndpr_advrtrs, pfr_entry) { 1283 if (j < DRLSTSIZ) { 1284 #define RTRADDR oprl->prefix[i].advrtr[j] 1285 RTRADDR = pfr->router->rtaddr; 1286 in6_clearscope(&RTRADDR); 1287 #undef RTRADDR 1288 } 1289 j++; 1290 } 1291 oprl->prefix[i].advrtrs = j; 1292 oprl->prefix[i].origin = PR_ORIG_RA; 1293 1294 i++; 1295 } 1296 splx(s); 1297 1298 break; 1299 case OSIOCGIFINFO_IN6: 1300 #define ND ndi->ndi 1301 /* XXX: old ndp(8) assumes a positive value for linkmtu. */ 1302 bzero(&ND, sizeof(ND)); 1303 ND.linkmtu = IN6_LINKMTU(ifp); 1304 ND.maxmtu = ND_IFINFO(ifp)->maxmtu; 1305 ND.basereachable = ND_IFINFO(ifp)->basereachable; 1306 ND.reachable = ND_IFINFO(ifp)->reachable; 1307 ND.retrans = ND_IFINFO(ifp)->retrans; 1308 ND.flags = ND_IFINFO(ifp)->flags; 1309 ND.recalctm = ND_IFINFO(ifp)->recalctm; 1310 ND.chlim = ND_IFINFO(ifp)->chlim; 1311 break; 1312 case SIOCGIFINFO_IN6: 1313 ND = *ND_IFINFO(ifp); 1314 break; 1315 case SIOCSIFINFO_IN6: 1316 /* 1317 * used to change host variables from userland. 1318 * intented for a use on router to reflect RA configurations. 1319 */ 1320 /* 0 means 'unspecified' */ 1321 if (ND.linkmtu != 0) { 1322 if (ND.linkmtu < IPV6_MMTU || 1323 ND.linkmtu > IN6_LINKMTU(ifp)) { 1324 error = EINVAL; 1325 break; 1326 } 1327 ND_IFINFO(ifp)->linkmtu = ND.linkmtu; 1328 } 1329 1330 if (ND.basereachable != 0) { 1331 int obasereachable = ND_IFINFO(ifp)->basereachable; 1332 1333 ND_IFINFO(ifp)->basereachable = ND.basereachable; 1334 if (ND.basereachable != obasereachable) 1335 ND_IFINFO(ifp)->reachable = 1336 ND_COMPUTE_RTIME(ND.basereachable); 1337 } 1338 if (ND.retrans != 0) 1339 ND_IFINFO(ifp)->retrans = ND.retrans; 1340 if (ND.chlim != 0) 1341 ND_IFINFO(ifp)->chlim = ND.chlim; 1342 /* FALLTHROUGH */ 1343 case SIOCSIFINFO_FLAGS: 1344 { 1345 struct ifaddr *ifa; 1346 struct in6_ifaddr *ia; 1347 1348 /* 1349 * Try to clear ifdisabled flag when enabling 1350 * accept_rtadv or auto_linklocal. 1351 */ 1352 if ((ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) && 1353 !(ND.flags & ND6_IFF_IFDISABLED) && 1354 (ND.flags & (ND6_IFF_ACCEPT_RTADV | 1355 ND6_IFF_AUTO_LINKLOCAL))) 1356 ND.flags &= ~ND6_IFF_IFDISABLED; 1357 1358 if ((ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) && 1359 !(ND.flags & ND6_IFF_IFDISABLED)) { 1360 /* ifdisabled 1->0 transision */ 1361 1362 /* 1363 * If the interface is marked as ND6_IFF_IFDISABLED and 1364 * has an link-local address with IN6_IFF_DUPLICATED, 1365 * do not clear ND6_IFF_IFDISABLED. 1366 * See RFC 4862, Section 5.4.5. 1367 */ 1368 int duplicated_linklocal = 0; 1369 1370 IF_ADDR_RLOCK(ifp); 1371 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1372 if (ifa->ifa_addr->sa_family != AF_INET6) 1373 continue; 1374 ia = (struct in6_ifaddr *)ifa; 1375 if ((ia->ia6_flags & IN6_IFF_DUPLICATED) && 1376 IN6_IS_ADDR_LINKLOCAL(IA6_IN6(ia))) { 1377 duplicated_linklocal = 1; 1378 break; 1379 } 1380 } 1381 IF_ADDR_RUNLOCK(ifp); 1382 1383 if (duplicated_linklocal) { 1384 ND.flags |= ND6_IFF_IFDISABLED; 1385 log(LOG_ERR, "Cannot enable an interface" 1386 " with a link-local address marked" 1387 " duplicate.\n"); 1388 } else { 1389 ND_IFINFO(ifp)->flags &= ~ND6_IFF_IFDISABLED; 1390 if (ifp->if_flags & IFF_UP) 1391 in6_if_up(ifp); 1392 } 1393 } else if (!(ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) && 1394 (ND.flags & ND6_IFF_IFDISABLED)) { 1395 /* ifdisabled 0->1 transision */ 1396 /* Mark all IPv6 address as tentative. */ 1397 1398 ND_IFINFO(ifp)->flags |= ND6_IFF_IFDISABLED; 1399 IF_ADDR_RLOCK(ifp); 1400 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1401 if (ifa->ifa_addr->sa_family != AF_INET6) 1402 continue; 1403 ia = (struct in6_ifaddr *)ifa; 1404 ia->ia6_flags |= IN6_IFF_TENTATIVE; 1405 } 1406 IF_ADDR_RUNLOCK(ifp); 1407 } 1408 1409 if (ND.flags & ND6_IFF_AUTO_LINKLOCAL) { 1410 if (!(ND_IFINFO(ifp)->flags & ND6_IFF_AUTO_LINKLOCAL)) { 1411 /* auto_linklocal 0->1 transision */ 1412 1413 /* If no link-local address on ifp, configure */ 1414 ND_IFINFO(ifp)->flags |= ND6_IFF_AUTO_LINKLOCAL; 1415 in6_ifattach(ifp, NULL); 1416 } else if (!(ND.flags & ND6_IFF_IFDISABLED) && 1417 ifp->if_flags & IFF_UP) { 1418 /* 1419 * When the IF already has 1420 * ND6_IFF_AUTO_LINKLOCAL, no link-local 1421 * address is assigned, and IFF_UP, try to 1422 * assign one. 1423 */ 1424 int haslinklocal = 0; 1425 1426 IF_ADDR_RLOCK(ifp); 1427 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1428 if (ifa->ifa_addr->sa_family != AF_INET6) 1429 continue; 1430 ia = (struct in6_ifaddr *)ifa; 1431 if (IN6_IS_ADDR_LINKLOCAL(IA6_IN6(ia))) { 1432 haslinklocal = 1; 1433 break; 1434 } 1435 } 1436 IF_ADDR_RUNLOCK(ifp); 1437 if (!haslinklocal) 1438 in6_ifattach(ifp, NULL); 1439 } 1440 } 1441 } 1442 ND_IFINFO(ifp)->flags = ND.flags; 1443 break; 1444 #undef ND 1445 case SIOCSNDFLUSH_IN6: /* XXX: the ioctl name is confusing... */ 1446 /* sync kernel routing table with the default router list */ 1447 defrouter_reset(); 1448 defrouter_select(); 1449 break; 1450 case SIOCSPFXFLUSH_IN6: 1451 { 1452 /* flush all the prefix advertised by routers */ 1453 struct nd_prefix *pr, *next; 1454 1455 s = splnet(); 1456 LIST_FOREACH_SAFE(pr, &V_nd_prefix, ndpr_entry, next) { 1457 struct in6_ifaddr *ia, *ia_next; 1458 1459 if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr)) 1460 continue; /* XXX */ 1461 1462 /* do we really have to remove addresses as well? */ 1463 /* XXXRW: in6_ifaddrhead locking. */ 1464 TAILQ_FOREACH_SAFE(ia, &V_in6_ifaddrhead, ia_link, 1465 ia_next) { 1466 if ((ia->ia6_flags & IN6_IFF_AUTOCONF) == 0) 1467 continue; 1468 1469 if (ia->ia6_ndpr == pr) 1470 in6_purgeaddr(&ia->ia_ifa); 1471 } 1472 prelist_remove(pr); 1473 } 1474 splx(s); 1475 break; 1476 } 1477 case SIOCSRTRFLUSH_IN6: 1478 { 1479 /* flush all the default routers */ 1480 struct nd_defrouter *dr, *next; 1481 1482 s = splnet(); 1483 defrouter_reset(); 1484 TAILQ_FOREACH_SAFE(dr, &V_nd_defrouter, dr_entry, next) { 1485 defrtrlist_del(dr); 1486 } 1487 defrouter_select(); 1488 splx(s); 1489 break; 1490 } 1491 case SIOCGNBRINFO_IN6: 1492 { 1493 struct llentry *ln; 1494 struct in6_addr nb_addr = nbi->addr; /* make local for safety */ 1495 1496 if ((error = in6_setscope(&nb_addr, ifp, NULL)) != 0) 1497 return (error); 1498 1499 IF_AFDATA_LOCK(ifp); 1500 ln = nd6_lookup(&nb_addr, 0, ifp); 1501 IF_AFDATA_UNLOCK(ifp); 1502 1503 if (ln == NULL) { 1504 error = EINVAL; 1505 break; 1506 } 1507 nbi->state = ln->ln_state; 1508 nbi->asked = ln->la_asked; 1509 nbi->isrouter = ln->ln_router; 1510 nbi->expire = ln->la_expire; 1511 LLE_RUNLOCK(ln); 1512 break; 1513 } 1514 case SIOCGDEFIFACE_IN6: /* XXX: should be implemented as a sysctl? */ 1515 ndif->ifindex = V_nd6_defifindex; 1516 break; 1517 case SIOCSDEFIFACE_IN6: /* XXX: should be implemented as a sysctl? */ 1518 return (nd6_setdefaultiface(ndif->ifindex)); 1519 } 1520 return (error); 1521 } 1522 1523 /* 1524 * Create neighbor cache entry and cache link-layer address, 1525 * on reception of inbound ND6 packets. (RS/RA/NS/redirect) 1526 * 1527 * type - ICMP6 type 1528 * code - type dependent information 1529 * 1530 * XXXXX 1531 * The caller of this function already acquired the ndp 1532 * cache table lock because the cache entry is returned. 1533 */ 1534 struct llentry * 1535 nd6_cache_lladdr(struct ifnet *ifp, struct in6_addr *from, char *lladdr, 1536 int lladdrlen, int type, int code) 1537 { 1538 struct llentry *ln = NULL; 1539 int is_newentry; 1540 int do_update; 1541 int olladdr; 1542 int llchange; 1543 int flags; 1544 int newstate = 0; 1545 uint16_t router = 0; 1546 struct sockaddr_in6 sin6; 1547 struct mbuf *chain = NULL; 1548 int static_route = 0; 1549 1550 IF_AFDATA_UNLOCK_ASSERT(ifp); 1551 1552 if (ifp == NULL) 1553 panic("ifp == NULL in nd6_cache_lladdr"); 1554 if (from == NULL) 1555 panic("from == NULL in nd6_cache_lladdr"); 1556 1557 /* nothing must be updated for unspecified address */ 1558 if (IN6_IS_ADDR_UNSPECIFIED(from)) 1559 return NULL; 1560 1561 /* 1562 * Validation about ifp->if_addrlen and lladdrlen must be done in 1563 * the caller. 1564 * 1565 * XXX If the link does not have link-layer adderss, what should 1566 * we do? (ifp->if_addrlen == 0) 1567 * Spec says nothing in sections for RA, RS and NA. There's small 1568 * description on it in NS section (RFC 2461 7.2.3). 1569 */ 1570 flags = lladdr ? ND6_EXCLUSIVE : 0; 1571 IF_AFDATA_LOCK(ifp); 1572 ln = nd6_lookup(from, flags, ifp); 1573 1574 if (ln == NULL) { 1575 flags |= ND6_EXCLUSIVE; 1576 ln = nd6_lookup(from, flags | ND6_CREATE, ifp); 1577 IF_AFDATA_UNLOCK(ifp); 1578 is_newentry = 1; 1579 } else { 1580 IF_AFDATA_UNLOCK(ifp); 1581 /* do nothing if static ndp is set */ 1582 if (ln->la_flags & LLE_STATIC) { 1583 static_route = 1; 1584 goto done; 1585 } 1586 is_newentry = 0; 1587 } 1588 if (ln == NULL) 1589 return (NULL); 1590 1591 olladdr = (ln->la_flags & LLE_VALID) ? 1 : 0; 1592 if (olladdr && lladdr) { 1593 llchange = bcmp(lladdr, &ln->ll_addr, 1594 ifp->if_addrlen); 1595 } else 1596 llchange = 0; 1597 1598 /* 1599 * newentry olladdr lladdr llchange (*=record) 1600 * 0 n n -- (1) 1601 * 0 y n -- (2) 1602 * 0 n y -- (3) * STALE 1603 * 0 y y n (4) * 1604 * 0 y y y (5) * STALE 1605 * 1 -- n -- (6) NOSTATE(= PASSIVE) 1606 * 1 -- y -- (7) * STALE 1607 */ 1608 1609 if (lladdr) { /* (3-5) and (7) */ 1610 /* 1611 * Record source link-layer address 1612 * XXX is it dependent to ifp->if_type? 1613 */ 1614 bcopy(lladdr, &ln->ll_addr, ifp->if_addrlen); 1615 ln->la_flags |= LLE_VALID; 1616 } 1617 1618 if (!is_newentry) { 1619 if ((!olladdr && lladdr != NULL) || /* (3) */ 1620 (olladdr && lladdr != NULL && llchange)) { /* (5) */ 1621 do_update = 1; 1622 newstate = ND6_LLINFO_STALE; 1623 } else /* (1-2,4) */ 1624 do_update = 0; 1625 } else { 1626 do_update = 1; 1627 if (lladdr == NULL) /* (6) */ 1628 newstate = ND6_LLINFO_NOSTATE; 1629 else /* (7) */ 1630 newstate = ND6_LLINFO_STALE; 1631 } 1632 1633 if (do_update) { 1634 /* 1635 * Update the state of the neighbor cache. 1636 */ 1637 ln->ln_state = newstate; 1638 1639 if (ln->ln_state == ND6_LLINFO_STALE) { 1640 /* 1641 * XXX: since nd6_output() below will cause 1642 * state tansition to DELAY and reset the timer, 1643 * we must set the timer now, although it is actually 1644 * meaningless. 1645 */ 1646 nd6_llinfo_settimer_locked(ln, (long)V_nd6_gctimer * hz); 1647 1648 if (ln->la_hold) { 1649 struct mbuf *m_hold, *m_hold_next; 1650 1651 /* 1652 * reset the la_hold in advance, to explicitly 1653 * prevent a la_hold lookup in nd6_output() 1654 * (wouldn't happen, though...) 1655 */ 1656 for (m_hold = ln->la_hold, ln->la_hold = NULL; 1657 m_hold; m_hold = m_hold_next) { 1658 m_hold_next = m_hold->m_nextpkt; 1659 m_hold->m_nextpkt = NULL; 1660 1661 /* 1662 * we assume ifp is not a p2p here, so 1663 * just set the 2nd argument as the 1664 * 1st one. 1665 */ 1666 nd6_output_lle(ifp, ifp, m_hold, L3_ADDR_SIN6(ln), NULL, ln, &chain); 1667 } 1668 /* 1669 * If we have mbufs in the chain we need to do 1670 * deferred transmit. Copy the address from the 1671 * llentry before dropping the lock down below. 1672 */ 1673 if (chain != NULL) 1674 memcpy(&sin6, L3_ADDR_SIN6(ln), sizeof(sin6)); 1675 } 1676 } else if (ln->ln_state == ND6_LLINFO_INCOMPLETE) { 1677 /* probe right away */ 1678 nd6_llinfo_settimer_locked((void *)ln, 0); 1679 } 1680 } 1681 1682 /* 1683 * ICMP6 type dependent behavior. 1684 * 1685 * NS: clear IsRouter if new entry 1686 * RS: clear IsRouter 1687 * RA: set IsRouter if there's lladdr 1688 * redir: clear IsRouter if new entry 1689 * 1690 * RA case, (1): 1691 * The spec says that we must set IsRouter in the following cases: 1692 * - If lladdr exist, set IsRouter. This means (1-5). 1693 * - If it is old entry (!newentry), set IsRouter. This means (7). 1694 * So, based on the spec, in (1-5) and (7) cases we must set IsRouter. 1695 * A quetion arises for (1) case. (1) case has no lladdr in the 1696 * neighbor cache, this is similar to (6). 1697 * This case is rare but we figured that we MUST NOT set IsRouter. 1698 * 1699 * newentry olladdr lladdr llchange NS RS RA redir 1700 * D R 1701 * 0 n n -- (1) c ? s 1702 * 0 y n -- (2) c s s 1703 * 0 n y -- (3) c s s 1704 * 0 y y n (4) c s s 1705 * 0 y y y (5) c s s 1706 * 1 -- n -- (6) c c c s 1707 * 1 -- y -- (7) c c s c s 1708 * 1709 * (c=clear s=set) 1710 */ 1711 switch (type & 0xff) { 1712 case ND_NEIGHBOR_SOLICIT: 1713 /* 1714 * New entry must have is_router flag cleared. 1715 */ 1716 if (is_newentry) /* (6-7) */ 1717 ln->ln_router = 0; 1718 break; 1719 case ND_REDIRECT: 1720 /* 1721 * If the icmp is a redirect to a better router, always set the 1722 * is_router flag. Otherwise, if the entry is newly created, 1723 * clear the flag. [RFC 2461, sec 8.3] 1724 */ 1725 if (code == ND_REDIRECT_ROUTER) 1726 ln->ln_router = 1; 1727 else if (is_newentry) /* (6-7) */ 1728 ln->ln_router = 0; 1729 break; 1730 case ND_ROUTER_SOLICIT: 1731 /* 1732 * is_router flag must always be cleared. 1733 */ 1734 ln->ln_router = 0; 1735 break; 1736 case ND_ROUTER_ADVERT: 1737 /* 1738 * Mark an entry with lladdr as a router. 1739 */ 1740 if ((!is_newentry && (olladdr || lladdr)) || /* (2-5) */ 1741 (is_newentry && lladdr)) { /* (7) */ 1742 ln->ln_router = 1; 1743 } 1744 break; 1745 } 1746 1747 if (ln != NULL) { 1748 static_route = (ln->la_flags & LLE_STATIC); 1749 router = ln->ln_router; 1750 1751 if (flags & ND6_EXCLUSIVE) 1752 LLE_WUNLOCK(ln); 1753 else 1754 LLE_RUNLOCK(ln); 1755 if (static_route) 1756 ln = NULL; 1757 } 1758 if (chain) 1759 nd6_output_flush(ifp, ifp, chain, &sin6, NULL); 1760 1761 /* 1762 * When the link-layer address of a router changes, select the 1763 * best router again. In particular, when the neighbor entry is newly 1764 * created, it might affect the selection policy. 1765 * Question: can we restrict the first condition to the "is_newentry" 1766 * case? 1767 * XXX: when we hear an RA from a new router with the link-layer 1768 * address option, defrouter_select() is called twice, since 1769 * defrtrlist_update called the function as well. However, I believe 1770 * we can compromise the overhead, since it only happens the first 1771 * time. 1772 * XXX: although defrouter_select() should not have a bad effect 1773 * for those are not autoconfigured hosts, we explicitly avoid such 1774 * cases for safety. 1775 */ 1776 if (do_update && router && 1777 ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV) { 1778 /* 1779 * guaranteed recursion 1780 */ 1781 defrouter_select(); 1782 } 1783 1784 return (ln); 1785 done: 1786 if (ln != NULL) { 1787 if (flags & ND6_EXCLUSIVE) 1788 LLE_WUNLOCK(ln); 1789 else 1790 LLE_RUNLOCK(ln); 1791 if (static_route) 1792 ln = NULL; 1793 } 1794 return (ln); 1795 } 1796 1797 static void 1798 nd6_slowtimo(void *arg) 1799 { 1800 CURVNET_SET((struct vnet *) arg); 1801 struct nd_ifinfo *nd6if; 1802 struct ifnet *ifp; 1803 1804 callout_reset(&V_nd6_slowtimo_ch, ND6_SLOWTIMER_INTERVAL * hz, 1805 nd6_slowtimo, curvnet); 1806 IFNET_RLOCK_NOSLEEP(); 1807 TAILQ_FOREACH(ifp, &V_ifnet, if_list) { 1808 nd6if = ND_IFINFO(ifp); 1809 if (nd6if->basereachable && /* already initialized */ 1810 (nd6if->recalctm -= ND6_SLOWTIMER_INTERVAL) <= 0) { 1811 /* 1812 * Since reachable time rarely changes by router 1813 * advertisements, we SHOULD insure that a new random 1814 * value gets recomputed at least once every few hours. 1815 * (RFC 2461, 6.3.4) 1816 */ 1817 nd6if->recalctm = V_nd6_recalc_reachtm_interval; 1818 nd6if->reachable = ND_COMPUTE_RTIME(nd6if->basereachable); 1819 } 1820 } 1821 IFNET_RUNLOCK_NOSLEEP(); 1822 CURVNET_RESTORE(); 1823 } 1824 1825 int 1826 nd6_output(struct ifnet *ifp, struct ifnet *origifp, struct mbuf *m0, 1827 struct sockaddr_in6 *dst, struct rtentry *rt0) 1828 { 1829 1830 return (nd6_output_lle(ifp, origifp, m0, dst, rt0, NULL, NULL)); 1831 } 1832 1833 1834 /* 1835 * Note that I'm not enforcing any global serialization 1836 * lle state or asked changes here as the logic is too 1837 * complicated to avoid having to always acquire an exclusive 1838 * lock 1839 * KMM 1840 * 1841 */ 1842 #define senderr(e) { error = (e); goto bad;} 1843 1844 int 1845 nd6_output_lle(struct ifnet *ifp, struct ifnet *origifp, struct mbuf *m0, 1846 struct sockaddr_in6 *dst, struct rtentry *rt0, struct llentry *lle, 1847 struct mbuf **chain) 1848 { 1849 struct mbuf *m = m0; 1850 struct m_tag *mtag; 1851 struct llentry *ln = lle; 1852 struct ip6_hdr *ip6; 1853 int error = 0; 1854 int flags = 0; 1855 int ip6len; 1856 1857 #ifdef INVARIANTS 1858 if (lle != NULL) { 1859 1860 LLE_WLOCK_ASSERT(lle); 1861 1862 KASSERT(chain != NULL, (" lle locked but no mbuf chain pointer passed")); 1863 } 1864 #endif 1865 if (IN6_IS_ADDR_MULTICAST(&dst->sin6_addr)) 1866 goto sendpkt; 1867 1868 if (nd6_need_cache(ifp) == 0) 1869 goto sendpkt; 1870 1871 /* 1872 * next hop determination. This routine is derived from ether_output. 1873 */ 1874 1875 /* 1876 * Address resolution or Neighbor Unreachability Detection 1877 * for the next hop. 1878 * At this point, the destination of the packet must be a unicast 1879 * or an anycast address(i.e. not a multicast). 1880 */ 1881 1882 flags = ((m != NULL) || (lle != NULL)) ? LLE_EXCLUSIVE : 0; 1883 if (ln == NULL) { 1884 retry: 1885 IF_AFDATA_LOCK(ifp); 1886 ln = lla_lookup(LLTABLE6(ifp), flags, (struct sockaddr *)dst); 1887 IF_AFDATA_UNLOCK(ifp); 1888 if ((ln == NULL) && nd6_is_addr_neighbor(dst, ifp)) { 1889 /* 1890 * Since nd6_is_addr_neighbor() internally calls nd6_lookup(), 1891 * the condition below is not very efficient. But we believe 1892 * it is tolerable, because this should be a rare case. 1893 */ 1894 flags = ND6_CREATE | (m ? ND6_EXCLUSIVE : 0); 1895 IF_AFDATA_LOCK(ifp); 1896 ln = nd6_lookup(&dst->sin6_addr, flags, ifp); 1897 IF_AFDATA_UNLOCK(ifp); 1898 } 1899 } 1900 if (ln == NULL) { 1901 if ((ifp->if_flags & IFF_POINTOPOINT) == 0 && 1902 !(ND_IFINFO(ifp)->flags & ND6_IFF_PERFORMNUD)) { 1903 char ip6buf[INET6_ADDRSTRLEN]; 1904 log(LOG_DEBUG, 1905 "nd6_output: can't allocate llinfo for %s " 1906 "(ln=%p)\n", 1907 ip6_sprintf(ip6buf, &dst->sin6_addr), ln); 1908 senderr(EIO); /* XXX: good error? */ 1909 } 1910 goto sendpkt; /* send anyway */ 1911 } 1912 1913 /* We don't have to do link-layer address resolution on a p2p link. */ 1914 if ((ifp->if_flags & IFF_POINTOPOINT) != 0 && 1915 ln->ln_state < ND6_LLINFO_REACHABLE) { 1916 if ((flags & LLE_EXCLUSIVE) == 0) { 1917 flags |= LLE_EXCLUSIVE; 1918 goto retry; 1919 } 1920 ln->ln_state = ND6_LLINFO_STALE; 1921 nd6_llinfo_settimer_locked(ln, (long)V_nd6_gctimer * hz); 1922 } 1923 1924 /* 1925 * The first time we send a packet to a neighbor whose entry is 1926 * STALE, we have to change the state to DELAY and a sets a timer to 1927 * expire in DELAY_FIRST_PROBE_TIME seconds to ensure do 1928 * neighbor unreachability detection on expiration. 1929 * (RFC 2461 7.3.3) 1930 */ 1931 if (ln->ln_state == ND6_LLINFO_STALE) { 1932 if ((flags & LLE_EXCLUSIVE) == 0) { 1933 flags |= LLE_EXCLUSIVE; 1934 LLE_RUNLOCK(ln); 1935 goto retry; 1936 } 1937 ln->la_asked = 0; 1938 ln->ln_state = ND6_LLINFO_DELAY; 1939 nd6_llinfo_settimer_locked(ln, (long)V_nd6_delay * hz); 1940 } 1941 1942 /* 1943 * If the neighbor cache entry has a state other than INCOMPLETE 1944 * (i.e. its link-layer address is already resolved), just 1945 * send the packet. 1946 */ 1947 if (ln->ln_state > ND6_LLINFO_INCOMPLETE) 1948 goto sendpkt; 1949 1950 /* 1951 * There is a neighbor cache entry, but no ethernet address 1952 * response yet. Append this latest packet to the end of the 1953 * packet queue in the mbuf, unless the number of the packet 1954 * does not exceed nd6_maxqueuelen. When it exceeds nd6_maxqueuelen, 1955 * the oldest packet in the queue will be removed. 1956 */ 1957 if (ln->ln_state == ND6_LLINFO_NOSTATE) 1958 ln->ln_state = ND6_LLINFO_INCOMPLETE; 1959 1960 if ((flags & LLE_EXCLUSIVE) == 0) { 1961 flags |= LLE_EXCLUSIVE; 1962 LLE_RUNLOCK(ln); 1963 goto retry; 1964 } 1965 1966 LLE_WLOCK_ASSERT(ln); 1967 1968 if (ln->la_hold) { 1969 struct mbuf *m_hold; 1970 int i; 1971 1972 i = 0; 1973 for (m_hold = ln->la_hold; m_hold; m_hold = m_hold->m_nextpkt) { 1974 i++; 1975 if (m_hold->m_nextpkt == NULL) { 1976 m_hold->m_nextpkt = m; 1977 break; 1978 } 1979 } 1980 while (i >= V_nd6_maxqueuelen) { 1981 m_hold = ln->la_hold; 1982 ln->la_hold = ln->la_hold->m_nextpkt; 1983 m_freem(m_hold); 1984 i--; 1985 } 1986 } else { 1987 ln->la_hold = m; 1988 } 1989 1990 /* 1991 * If there has been no NS for the neighbor after entering the 1992 * INCOMPLETE state, send the first solicitation. 1993 */ 1994 if (!ND6_LLINFO_PERMANENT(ln) && ln->la_asked == 0) { 1995 ln->la_asked++; 1996 1997 nd6_llinfo_settimer_locked(ln, 1998 (long)ND_IFINFO(ifp)->retrans * hz / 1000); 1999 LLE_WUNLOCK(ln); 2000 nd6_ns_output(ifp, NULL, &dst->sin6_addr, ln, 0); 2001 if (lle != NULL && ln == lle) 2002 LLE_WLOCK(lle); 2003 2004 } else if (lle == NULL || ln != lle) { 2005 /* 2006 * We did the lookup (no lle arg) so we 2007 * need to do the unlock here. 2008 */ 2009 LLE_WUNLOCK(ln); 2010 } 2011 2012 return (0); 2013 2014 sendpkt: 2015 /* discard the packet if IPv6 operation is disabled on the interface */ 2016 if ((ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED)) { 2017 error = ENETDOWN; /* better error? */ 2018 goto bad; 2019 } 2020 /* 2021 * ln is valid and the caller did not pass in 2022 * an llentry 2023 */ 2024 if ((ln != NULL) && (lle == NULL)) { 2025 if (flags & LLE_EXCLUSIVE) 2026 LLE_WUNLOCK(ln); 2027 else 2028 LLE_RUNLOCK(ln); 2029 } 2030 2031 #ifdef MAC 2032 mac_netinet6_nd6_send(ifp, m); 2033 #endif 2034 2035 /* 2036 * If called from nd6_ns_output() (NS), nd6_na_output() (NA), 2037 * icmp6_redirect_output() (REDIRECT) or from rip6_output() (RS, RA 2038 * as handled by rtsol and rtadvd), mbufs will be tagged for SeND 2039 * to be diverted to user space. When re-injected into the kernel, 2040 * send_output() will directly dispatch them to the outgoing interface. 2041 */ 2042 if (send_sendso_input_hook != NULL) { 2043 mtag = m_tag_find(m, PACKET_TAG_ND_OUTGOING, NULL); 2044 if (mtag != NULL) { 2045 ip6 = mtod(m, struct ip6_hdr *); 2046 ip6len = sizeof(struct ip6_hdr) + ntohs(ip6->ip6_plen); 2047 /* Use the SEND socket */ 2048 error = send_sendso_input_hook(m, ifp, SND_OUT, 2049 ip6len); 2050 /* -1 == no app on SEND socket */ 2051 if (error == 0 || error != -1) 2052 return (error); 2053 } 2054 } 2055 2056 /* 2057 * We were passed in a pointer to an lle with the lock held 2058 * this means that we can't call if_output as we will 2059 * recurse on the lle lock - so what we do is we create 2060 * a list of mbufs to send and transmit them in the caller 2061 * after the lock is dropped 2062 */ 2063 if (lle != NULL) { 2064 if (*chain == NULL) 2065 *chain = m; 2066 else { 2067 struct mbuf *mb; 2068 2069 /* 2070 * append mbuf to end of deferred chain 2071 */ 2072 mb = *chain; 2073 while (mb->m_nextpkt != NULL) 2074 mb = mb->m_nextpkt; 2075 mb->m_nextpkt = m; 2076 } 2077 return (error); 2078 } 2079 if ((ifp->if_flags & IFF_LOOPBACK) != 0) { 2080 return ((*ifp->if_output)(origifp, m, (struct sockaddr *)dst, 2081 NULL)); 2082 } 2083 error = (*ifp->if_output)(ifp, m, (struct sockaddr *)dst, NULL); 2084 return (error); 2085 2086 bad: 2087 /* 2088 * ln is valid and the caller did not pass in 2089 * an llentry 2090 */ 2091 if ((ln != NULL) && (lle == NULL)) { 2092 if (flags & LLE_EXCLUSIVE) 2093 LLE_WUNLOCK(ln); 2094 else 2095 LLE_RUNLOCK(ln); 2096 } 2097 if (m) 2098 m_freem(m); 2099 return (error); 2100 } 2101 #undef senderr 2102 2103 2104 int 2105 nd6_output_flush(struct ifnet *ifp, struct ifnet *origifp, struct mbuf *chain, 2106 struct sockaddr_in6 *dst, struct route *ro) 2107 { 2108 struct mbuf *m, *m_head; 2109 struct ifnet *outifp; 2110 int error = 0; 2111 2112 m_head = chain; 2113 if ((ifp->if_flags & IFF_LOOPBACK) != 0) 2114 outifp = origifp; 2115 else 2116 outifp = ifp; 2117 2118 while (m_head) { 2119 m = m_head; 2120 m_head = m_head->m_nextpkt; 2121 error = (*ifp->if_output)(ifp, m, (struct sockaddr *)dst, ro); 2122 } 2123 2124 /* 2125 * XXX 2126 * note that intermediate errors are blindly ignored - but this is 2127 * the same convention as used with nd6_output when called by 2128 * nd6_cache_lladdr 2129 */ 2130 return (error); 2131 } 2132 2133 2134 int 2135 nd6_need_cache(struct ifnet *ifp) 2136 { 2137 /* 2138 * XXX: we currently do not make neighbor cache on any interface 2139 * other than ARCnet, Ethernet, FDDI and GIF. 2140 * 2141 * RFC2893 says: 2142 * - unidirectional tunnels needs no ND 2143 */ 2144 switch (ifp->if_type) { 2145 case IFT_ARCNET: 2146 case IFT_ETHER: 2147 case IFT_FDDI: 2148 case IFT_IEEE1394: 2149 #ifdef IFT_L2VLAN 2150 case IFT_L2VLAN: 2151 #endif 2152 #ifdef IFT_IEEE80211 2153 case IFT_IEEE80211: 2154 #endif 2155 case IFT_INFINIBAND: 2156 case IFT_GIF: /* XXX need more cases? */ 2157 case IFT_PPP: 2158 case IFT_TUNNEL: 2159 case IFT_BRIDGE: 2160 case IFT_PROPVIRTUAL: 2161 return (1); 2162 default: 2163 return (0); 2164 } 2165 } 2166 2167 /* 2168 * the callers of this function need to be re-worked to drop 2169 * the lle lock, drop here for now 2170 */ 2171 int 2172 nd6_storelladdr(struct ifnet *ifp, struct mbuf *m, 2173 struct sockaddr *dst, u_char *desten, struct llentry **lle) 2174 { 2175 struct llentry *ln; 2176 2177 *lle = NULL; 2178 IF_AFDATA_UNLOCK_ASSERT(ifp); 2179 if (m->m_flags & M_MCAST) { 2180 int i; 2181 2182 switch (ifp->if_type) { 2183 case IFT_ETHER: 2184 case IFT_FDDI: 2185 #ifdef IFT_L2VLAN 2186 case IFT_L2VLAN: 2187 #endif 2188 #ifdef IFT_IEEE80211 2189 case IFT_IEEE80211: 2190 #endif 2191 case IFT_BRIDGE: 2192 case IFT_ISO88025: 2193 ETHER_MAP_IPV6_MULTICAST(&SIN6(dst)->sin6_addr, 2194 desten); 2195 return (0); 2196 case IFT_IEEE1394: 2197 /* 2198 * netbsd can use if_broadcastaddr, but we don't do so 2199 * to reduce # of ifdef. 2200 */ 2201 for (i = 0; i < ifp->if_addrlen; i++) 2202 desten[i] = ~0; 2203 return (0); 2204 case IFT_ARCNET: 2205 *desten = 0; 2206 return (0); 2207 default: 2208 m_freem(m); 2209 return (EAFNOSUPPORT); 2210 } 2211 } 2212 2213 2214 /* 2215 * the entry should have been created in nd6_store_lladdr 2216 */ 2217 IF_AFDATA_LOCK(ifp); 2218 ln = lla_lookup(LLTABLE6(ifp), 0, dst); 2219 IF_AFDATA_UNLOCK(ifp); 2220 if ((ln == NULL) || !(ln->la_flags & LLE_VALID)) { 2221 if (ln != NULL) 2222 LLE_RUNLOCK(ln); 2223 /* this could happen, if we could not allocate memory */ 2224 m_freem(m); 2225 return (1); 2226 } 2227 2228 bcopy(&ln->ll_addr, desten, ifp->if_addrlen); 2229 *lle = ln; 2230 LLE_RUNLOCK(ln); 2231 /* 2232 * A *small* use after free race exists here 2233 */ 2234 return (0); 2235 } 2236 2237 static void 2238 clear_llinfo_pqueue(struct llentry *ln) 2239 { 2240 struct mbuf *m_hold, *m_hold_next; 2241 2242 for (m_hold = ln->la_hold; m_hold; m_hold = m_hold_next) { 2243 m_hold_next = m_hold->m_nextpkt; 2244 m_hold->m_nextpkt = NULL; 2245 m_freem(m_hold); 2246 } 2247 2248 ln->la_hold = NULL; 2249 return; 2250 } 2251 2252 static int nd6_sysctl_drlist(SYSCTL_HANDLER_ARGS); 2253 static int nd6_sysctl_prlist(SYSCTL_HANDLER_ARGS); 2254 #ifdef SYSCTL_DECL 2255 SYSCTL_DECL(_net_inet6_icmp6); 2256 #endif 2257 SYSCTL_NODE(_net_inet6_icmp6, ICMPV6CTL_ND6_DRLIST, nd6_drlist, 2258 CTLFLAG_RD, nd6_sysctl_drlist, ""); 2259 SYSCTL_NODE(_net_inet6_icmp6, ICMPV6CTL_ND6_PRLIST, nd6_prlist, 2260 CTLFLAG_RD, nd6_sysctl_prlist, ""); 2261 SYSCTL_VNET_INT(_net_inet6_icmp6, ICMPV6CTL_ND6_MAXQLEN, nd6_maxqueuelen, 2262 CTLFLAG_RW, &VNET_NAME(nd6_maxqueuelen), 1, ""); 2263 2264 static int 2265 nd6_sysctl_drlist(SYSCTL_HANDLER_ARGS) 2266 { 2267 int error; 2268 char buf[1024] __aligned(4); 2269 struct in6_defrouter *d, *de; 2270 struct nd_defrouter *dr; 2271 2272 if (req->newptr) 2273 return EPERM; 2274 error = 0; 2275 2276 TAILQ_FOREACH(dr, &V_nd_defrouter, dr_entry) { 2277 d = (struct in6_defrouter *)buf; 2278 de = (struct in6_defrouter *)(buf + sizeof(buf)); 2279 2280 if (d + 1 <= de) { 2281 bzero(d, sizeof(*d)); 2282 d->rtaddr.sin6_family = AF_INET6; 2283 d->rtaddr.sin6_len = sizeof(d->rtaddr); 2284 d->rtaddr.sin6_addr = dr->rtaddr; 2285 error = sa6_recoverscope(&d->rtaddr); 2286 if (error != 0) 2287 return (error); 2288 d->flags = dr->flags; 2289 d->rtlifetime = dr->rtlifetime; 2290 d->expire = dr->expire; 2291 d->if_index = dr->ifp->if_index; 2292 } else 2293 panic("buffer too short"); 2294 2295 error = SYSCTL_OUT(req, buf, sizeof(*d)); 2296 if (error) 2297 break; 2298 } 2299 2300 return (error); 2301 } 2302 2303 static int 2304 nd6_sysctl_prlist(SYSCTL_HANDLER_ARGS) 2305 { 2306 int error; 2307 char buf[1024] __aligned(4); 2308 struct in6_prefix *p, *pe; 2309 struct nd_prefix *pr; 2310 char ip6buf[INET6_ADDRSTRLEN]; 2311 2312 if (req->newptr) 2313 return EPERM; 2314 error = 0; 2315 2316 LIST_FOREACH(pr, &V_nd_prefix, ndpr_entry) { 2317 u_short advrtrs; 2318 size_t advance; 2319 struct sockaddr_in6 *sin6, *s6; 2320 struct nd_pfxrouter *pfr; 2321 2322 p = (struct in6_prefix *)buf; 2323 pe = (struct in6_prefix *)(buf + sizeof(buf)); 2324 2325 if (p + 1 <= pe) { 2326 bzero(p, sizeof(*p)); 2327 sin6 = (struct sockaddr_in6 *)(p + 1); 2328 2329 p->prefix = pr->ndpr_prefix; 2330 if (sa6_recoverscope(&p->prefix)) { 2331 log(LOG_ERR, 2332 "scope error in prefix list (%s)\n", 2333 ip6_sprintf(ip6buf, &p->prefix.sin6_addr)); 2334 /* XXX: press on... */ 2335 } 2336 p->raflags = pr->ndpr_raf; 2337 p->prefixlen = pr->ndpr_plen; 2338 p->vltime = pr->ndpr_vltime; 2339 p->pltime = pr->ndpr_pltime; 2340 p->if_index = pr->ndpr_ifp->if_index; 2341 if (pr->ndpr_vltime == ND6_INFINITE_LIFETIME) 2342 p->expire = 0; 2343 else { 2344 time_t maxexpire; 2345 2346 /* XXX: we assume time_t is signed. */ 2347 maxexpire = (-1) & 2348 ~((time_t)1 << 2349 ((sizeof(maxexpire) * 8) - 1)); 2350 if (pr->ndpr_vltime < 2351 maxexpire - pr->ndpr_lastupdate) { 2352 p->expire = pr->ndpr_lastupdate + 2353 pr->ndpr_vltime; 2354 } else 2355 p->expire = maxexpire; 2356 } 2357 p->refcnt = pr->ndpr_refcnt; 2358 p->flags = pr->ndpr_stateflags; 2359 p->origin = PR_ORIG_RA; 2360 advrtrs = 0; 2361 LIST_FOREACH(pfr, &pr->ndpr_advrtrs, pfr_entry) { 2362 if ((void *)&sin6[advrtrs + 1] > (void *)pe) { 2363 advrtrs++; 2364 continue; 2365 } 2366 s6 = &sin6[advrtrs]; 2367 bzero(s6, sizeof(*s6)); 2368 s6->sin6_family = AF_INET6; 2369 s6->sin6_len = sizeof(*sin6); 2370 s6->sin6_addr = pfr->router->rtaddr; 2371 if (sa6_recoverscope(s6)) { 2372 log(LOG_ERR, 2373 "scope error in " 2374 "prefix list (%s)\n", 2375 ip6_sprintf(ip6buf, 2376 &pfr->router->rtaddr)); 2377 } 2378 advrtrs++; 2379 } 2380 p->advrtrs = advrtrs; 2381 } else 2382 panic("buffer too short"); 2383 2384 advance = sizeof(*p) + sizeof(*sin6) * advrtrs; 2385 error = SYSCTL_OUT(req, buf, advance); 2386 if (error) 2387 break; 2388 } 2389 2390 return (error); 2391 } 2392