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 /* What about ND_OPT_ROUTE_INFO? RFC 4191 */ 384 case ND_OPT_RDNSS: /* RFC 6106 */ 385 case ND_OPT_DNSSL: /* RFC 6106 */ 386 /* 387 * Silently ignore options we know and do not care about 388 * in the kernel. 389 */ 390 break; 391 default: 392 /* 393 * Unknown options must be silently ignored, 394 * to accomodate future extension to the protocol. 395 */ 396 nd6log((LOG_DEBUG, 397 "nd6_options: unsupported option %d - " 398 "option ignored\n", nd_opt->nd_opt_type)); 399 } 400 401 skip1: 402 i++; 403 if (i > V_nd6_maxndopt) { 404 ICMP6STAT_INC(icp6s_nd_toomanyopt); 405 nd6log((LOG_INFO, "too many loop in nd opt\n")); 406 break; 407 } 408 409 if (ndopts->nd_opts_done) 410 break; 411 } 412 413 return 0; 414 } 415 416 /* 417 * ND6 timer routine to handle ND6 entries 418 */ 419 void 420 nd6_llinfo_settimer_locked(struct llentry *ln, long tick) 421 { 422 int canceled; 423 424 LLE_WLOCK_ASSERT(ln); 425 426 if (tick < 0) { 427 ln->la_expire = 0; 428 ln->ln_ntick = 0; 429 canceled = callout_stop(&ln->ln_timer_ch); 430 } else { 431 ln->la_expire = time_second + tick / hz; 432 LLE_ADDREF(ln); 433 if (tick > INT_MAX) { 434 ln->ln_ntick = tick - INT_MAX; 435 canceled = callout_reset(&ln->ln_timer_ch, INT_MAX, 436 nd6_llinfo_timer, ln); 437 } else { 438 ln->ln_ntick = 0; 439 canceled = callout_reset(&ln->ln_timer_ch, tick, 440 nd6_llinfo_timer, ln); 441 } 442 } 443 if (canceled) 444 LLE_REMREF(ln); 445 } 446 447 void 448 nd6_llinfo_settimer(struct llentry *ln, long tick) 449 { 450 451 LLE_WLOCK(ln); 452 nd6_llinfo_settimer_locked(ln, tick); 453 LLE_WUNLOCK(ln); 454 } 455 456 static void 457 nd6_llinfo_timer(void *arg) 458 { 459 struct llentry *ln; 460 struct in6_addr *dst; 461 struct ifnet *ifp; 462 struct nd_ifinfo *ndi = NULL; 463 464 KASSERT(arg != NULL, ("%s: arg NULL", __func__)); 465 ln = (struct llentry *)arg; 466 LLE_WLOCK_ASSERT(ln); 467 ifp = ln->lle_tbl->llt_ifp; 468 469 CURVNET_SET(ifp->if_vnet); 470 471 if (ln->ln_ntick > 0) { 472 if (ln->ln_ntick > INT_MAX) { 473 ln->ln_ntick -= INT_MAX; 474 nd6_llinfo_settimer_locked(ln, INT_MAX); 475 } else { 476 ln->ln_ntick = 0; 477 nd6_llinfo_settimer_locked(ln, ln->ln_ntick); 478 } 479 goto done; 480 } 481 482 ndi = ND_IFINFO(ifp); 483 dst = &L3_ADDR_SIN6(ln)->sin6_addr; 484 if (ln->la_flags & LLE_STATIC) { 485 goto done; 486 } 487 488 if (ln->la_flags & LLE_DELETED) { 489 (void)nd6_free(ln, 0); 490 ln = NULL; 491 goto done; 492 } 493 494 switch (ln->ln_state) { 495 case ND6_LLINFO_INCOMPLETE: 496 if (ln->la_asked < V_nd6_mmaxtries) { 497 ln->la_asked++; 498 nd6_llinfo_settimer_locked(ln, (long)ndi->retrans * hz / 1000); 499 LLE_WUNLOCK(ln); 500 nd6_ns_output(ifp, NULL, dst, ln, 0); 501 LLE_WLOCK(ln); 502 } else { 503 struct mbuf *m = ln->la_hold; 504 if (m) { 505 struct mbuf *m0; 506 507 /* 508 * assuming every packet in la_hold has the 509 * same IP header. Send error after unlock. 510 */ 511 m0 = m->m_nextpkt; 512 m->m_nextpkt = NULL; 513 ln->la_hold = m0; 514 clear_llinfo_pqueue(ln); 515 } 516 (void)nd6_free(ln, 0); 517 ln = NULL; 518 if (m != NULL) 519 icmp6_error2(m, ICMP6_DST_UNREACH, 520 ICMP6_DST_UNREACH_ADDR, 0, ifp); 521 } 522 break; 523 case ND6_LLINFO_REACHABLE: 524 if (!ND6_LLINFO_PERMANENT(ln)) { 525 ln->ln_state = ND6_LLINFO_STALE; 526 nd6_llinfo_settimer_locked(ln, (long)V_nd6_gctimer * hz); 527 } 528 break; 529 530 case ND6_LLINFO_STALE: 531 /* Garbage Collection(RFC 2461 5.3) */ 532 if (!ND6_LLINFO_PERMANENT(ln)) { 533 (void)nd6_free(ln, 1); 534 ln = NULL; 535 } 536 break; 537 538 case ND6_LLINFO_DELAY: 539 if (ndi && (ndi->flags & ND6_IFF_PERFORMNUD) != 0) { 540 /* We need NUD */ 541 ln->la_asked = 1; 542 ln->ln_state = ND6_LLINFO_PROBE; 543 nd6_llinfo_settimer_locked(ln, (long)ndi->retrans * hz / 1000); 544 LLE_WUNLOCK(ln); 545 nd6_ns_output(ifp, dst, dst, ln, 0); 546 LLE_WLOCK(ln); 547 } else { 548 ln->ln_state = ND6_LLINFO_STALE; /* XXX */ 549 nd6_llinfo_settimer_locked(ln, (long)V_nd6_gctimer * hz); 550 } 551 break; 552 case ND6_LLINFO_PROBE: 553 if (ln->la_asked < V_nd6_umaxtries) { 554 ln->la_asked++; 555 nd6_llinfo_settimer_locked(ln, (long)ndi->retrans * hz / 1000); 556 LLE_WUNLOCK(ln); 557 nd6_ns_output(ifp, dst, dst, ln, 0); 558 LLE_WLOCK(ln); 559 } else { 560 (void)nd6_free(ln, 0); 561 ln = NULL; 562 } 563 break; 564 default: 565 panic("%s: paths in a dark night can be confusing: %d", 566 __func__, ln->ln_state); 567 } 568 done: 569 if (ln != NULL) 570 LLE_FREE_LOCKED(ln); 571 CURVNET_RESTORE(); 572 } 573 574 575 /* 576 * ND6 timer routine to expire default route list and prefix list 577 */ 578 void 579 nd6_timer(void *arg) 580 { 581 CURVNET_SET((struct vnet *) arg); 582 int s; 583 struct nd_defrouter *dr, *ndr; 584 struct nd_prefix *pr, *npr; 585 struct in6_ifaddr *ia6, *nia6; 586 587 callout_reset(&V_nd6_timer_ch, V_nd6_prune * hz, 588 nd6_timer, curvnet); 589 590 /* expire default router list */ 591 s = splnet(); 592 TAILQ_FOREACH_SAFE(dr, &V_nd_defrouter, dr_entry, ndr) { 593 if (dr->expire && dr->expire < time_second) 594 defrtrlist_del(dr); 595 } 596 597 /* 598 * expire interface addresses. 599 * in the past the loop was inside prefix expiry processing. 600 * However, from a stricter speci-confrmance standpoint, we should 601 * rather separate address lifetimes and prefix lifetimes. 602 * 603 * XXXRW: in6_ifaddrhead locking. 604 */ 605 addrloop: 606 TAILQ_FOREACH_SAFE(ia6, &V_in6_ifaddrhead, ia_link, nia6) { 607 /* check address lifetime */ 608 if (IFA6_IS_INVALID(ia6)) { 609 int regen = 0; 610 611 /* 612 * If the expiring address is temporary, try 613 * regenerating a new one. This would be useful when 614 * we suspended a laptop PC, then turned it on after a 615 * period that could invalidate all temporary 616 * addresses. Although we may have to restart the 617 * loop (see below), it must be after purging the 618 * address. Otherwise, we'd see an infinite loop of 619 * regeneration. 620 */ 621 if (V_ip6_use_tempaddr && 622 (ia6->ia6_flags & IN6_IFF_TEMPORARY) != 0) { 623 if (regen_tmpaddr(ia6) == 0) 624 regen = 1; 625 } 626 627 in6_purgeaddr(&ia6->ia_ifa); 628 629 if (regen) 630 goto addrloop; /* XXX: see below */ 631 } else if (IFA6_IS_DEPRECATED(ia6)) { 632 int oldflags = ia6->ia6_flags; 633 634 ia6->ia6_flags |= IN6_IFF_DEPRECATED; 635 636 /* 637 * If a temporary address has just become deprecated, 638 * regenerate a new one if possible. 639 */ 640 if (V_ip6_use_tempaddr && 641 (ia6->ia6_flags & IN6_IFF_TEMPORARY) != 0 && 642 (oldflags & IN6_IFF_DEPRECATED) == 0) { 643 644 if (regen_tmpaddr(ia6) == 0) { 645 /* 646 * A new temporary address is 647 * generated. 648 * XXX: this means the address chain 649 * has changed while we are still in 650 * the loop. Although the change 651 * would not cause disaster (because 652 * it's not a deletion, but an 653 * addition,) we'd rather restart the 654 * loop just for safety. Or does this 655 * significantly reduce performance?? 656 */ 657 goto addrloop; 658 } 659 } 660 } else { 661 /* 662 * A new RA might have made a deprecated address 663 * preferred. 664 */ 665 ia6->ia6_flags &= ~IN6_IFF_DEPRECATED; 666 } 667 } 668 669 /* expire prefix list */ 670 LIST_FOREACH_SAFE(pr, &V_nd_prefix, ndpr_entry, npr) { 671 /* 672 * check prefix lifetime. 673 * since pltime is just for autoconf, pltime processing for 674 * prefix is not necessary. 675 */ 676 if (pr->ndpr_vltime != ND6_INFINITE_LIFETIME && 677 time_second - pr->ndpr_lastupdate > pr->ndpr_vltime) { 678 679 /* 680 * address expiration and prefix expiration are 681 * separate. NEVER perform in6_purgeaddr here. 682 */ 683 prelist_remove(pr); 684 } 685 } 686 splx(s); 687 CURVNET_RESTORE(); 688 } 689 690 /* 691 * ia6 - deprecated/invalidated temporary address 692 */ 693 static int 694 regen_tmpaddr(struct in6_ifaddr *ia6) 695 { 696 struct ifaddr *ifa; 697 struct ifnet *ifp; 698 struct in6_ifaddr *public_ifa6 = NULL; 699 700 ifp = ia6->ia_ifa.ifa_ifp; 701 IF_ADDR_RLOCK(ifp); 702 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 703 struct in6_ifaddr *it6; 704 705 if (ifa->ifa_addr->sa_family != AF_INET6) 706 continue; 707 708 it6 = (struct in6_ifaddr *)ifa; 709 710 /* ignore no autoconf addresses. */ 711 if ((it6->ia6_flags & IN6_IFF_AUTOCONF) == 0) 712 continue; 713 714 /* ignore autoconf addresses with different prefixes. */ 715 if (it6->ia6_ndpr == NULL || it6->ia6_ndpr != ia6->ia6_ndpr) 716 continue; 717 718 /* 719 * Now we are looking at an autoconf address with the same 720 * prefix as ours. If the address is temporary and is still 721 * preferred, do not create another one. It would be rare, but 722 * could happen, for example, when we resume a laptop PC after 723 * a long period. 724 */ 725 if ((it6->ia6_flags & IN6_IFF_TEMPORARY) != 0 && 726 !IFA6_IS_DEPRECATED(it6)) { 727 public_ifa6 = NULL; 728 break; 729 } 730 731 /* 732 * This is a public autoconf address that has the same prefix 733 * as ours. If it is preferred, keep it. We can't break the 734 * loop here, because there may be a still-preferred temporary 735 * address with the prefix. 736 */ 737 if (!IFA6_IS_DEPRECATED(it6)) 738 public_ifa6 = it6; 739 740 if (public_ifa6 != NULL) 741 ifa_ref(&public_ifa6->ia_ifa); 742 } 743 IF_ADDR_RUNLOCK(ifp); 744 745 if (public_ifa6 != NULL) { 746 int e; 747 748 if ((e = in6_tmpifadd(public_ifa6, 0, 0)) != 0) { 749 ifa_free(&public_ifa6->ia_ifa); 750 log(LOG_NOTICE, "regen_tmpaddr: failed to create a new" 751 " tmp addr,errno=%d\n", e); 752 return (-1); 753 } 754 ifa_free(&public_ifa6->ia_ifa); 755 return (0); 756 } 757 758 return (-1); 759 } 760 761 /* 762 * Nuke neighbor cache/prefix/default router management table, right before 763 * ifp goes away. 764 */ 765 void 766 nd6_purge(struct ifnet *ifp) 767 { 768 struct nd_defrouter *dr, *ndr; 769 struct nd_prefix *pr, *npr; 770 771 /* 772 * Nuke default router list entries toward ifp. 773 * We defer removal of default router list entries that is installed 774 * in the routing table, in order to keep additional side effects as 775 * small as possible. 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 TAILQ_FOREACH_SAFE(dr, &V_nd_defrouter, dr_entry, ndr) { 786 if (!dr->installed) 787 continue; 788 789 if (dr->ifp == ifp) 790 defrtrlist_del(dr); 791 } 792 793 /* Nuke prefix list entries toward ifp */ 794 LIST_FOREACH_SAFE(pr, &V_nd_prefix, ndpr_entry, npr) { 795 if (pr->ndpr_ifp == ifp) { 796 /* 797 * Because if_detach() does *not* release prefixes 798 * while purging addresses the reference count will 799 * still be above zero. We therefore reset it to 800 * make sure that the prefix really gets purged. 801 */ 802 pr->ndpr_refcnt = 0; 803 804 /* 805 * Previously, pr->ndpr_addr is removed as well, 806 * but I strongly believe we don't have to do it. 807 * nd6_purge() is only called from in6_ifdetach(), 808 * which removes all the associated interface addresses 809 * by itself. 810 * (jinmei@kame.net 20010129) 811 */ 812 prelist_remove(pr); 813 } 814 } 815 816 /* cancel default outgoing interface setting */ 817 if (V_nd6_defifindex == ifp->if_index) 818 nd6_setdefaultiface(0); 819 820 if (ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV) { 821 /* Refresh default router list. */ 822 defrouter_select(); 823 } 824 825 /* XXXXX 826 * We do not nuke the neighbor cache entries here any more 827 * because the neighbor cache is kept in if_afdata[AF_INET6]. 828 * nd6_purge() is invoked by in6_ifdetach() which is called 829 * from if_detach() where everything gets purged. So let 830 * in6_domifdetach() do the actual L2 table purging work. 831 */ 832 } 833 834 /* 835 * the caller acquires and releases the lock on the lltbls 836 * Returns the llentry locked 837 */ 838 struct llentry * 839 nd6_lookup(struct in6_addr *addr6, int flags, struct ifnet *ifp) 840 { 841 struct sockaddr_in6 sin6; 842 struct llentry *ln; 843 int llflags; 844 845 bzero(&sin6, sizeof(sin6)); 846 sin6.sin6_len = sizeof(struct sockaddr_in6); 847 sin6.sin6_family = AF_INET6; 848 sin6.sin6_addr = *addr6; 849 850 IF_AFDATA_LOCK_ASSERT(ifp); 851 852 llflags = 0; 853 if (flags & ND6_CREATE) 854 llflags |= LLE_CREATE; 855 if (flags & ND6_EXCLUSIVE) 856 llflags |= LLE_EXCLUSIVE; 857 858 ln = lla_lookup(LLTABLE6(ifp), llflags, (struct sockaddr *)&sin6); 859 if ((ln != NULL) && (llflags & LLE_CREATE)) 860 ln->ln_state = ND6_LLINFO_NOSTATE; 861 862 return (ln); 863 } 864 865 /* 866 * Test whether a given IPv6 address is a neighbor or not, ignoring 867 * the actual neighbor cache. The neighbor cache is ignored in order 868 * to not reenter the routing code from within itself. 869 */ 870 static int 871 nd6_is_new_addr_neighbor(struct sockaddr_in6 *addr, struct ifnet *ifp) 872 { 873 struct nd_prefix *pr; 874 struct ifaddr *dstaddr; 875 876 /* 877 * A link-local address is always a neighbor. 878 * XXX: a link does not necessarily specify a single interface. 879 */ 880 if (IN6_IS_ADDR_LINKLOCAL(&addr->sin6_addr)) { 881 struct sockaddr_in6 sin6_copy; 882 u_int32_t zone; 883 884 /* 885 * We need sin6_copy since sa6_recoverscope() may modify the 886 * content (XXX). 887 */ 888 sin6_copy = *addr; 889 if (sa6_recoverscope(&sin6_copy)) 890 return (0); /* XXX: should be impossible */ 891 if (in6_setscope(&sin6_copy.sin6_addr, ifp, &zone)) 892 return (0); 893 if (sin6_copy.sin6_scope_id == zone) 894 return (1); 895 else 896 return (0); 897 } 898 899 /* 900 * If the address matches one of our addresses, 901 * it should be a neighbor. 902 * If the address matches one of our on-link prefixes, it should be a 903 * neighbor. 904 */ 905 LIST_FOREACH(pr, &V_nd_prefix, ndpr_entry) { 906 if (pr->ndpr_ifp != ifp) 907 continue; 908 909 if (!(pr->ndpr_stateflags & NDPRF_ONLINK)) { 910 struct rtentry *rt; 911 912 /* Always use the default FIB here. */ 913 rt = in6_rtalloc1((struct sockaddr *)&pr->ndpr_prefix, 914 0, 0, RT_DEFAULT_FIB); 915 if (rt == NULL) 916 continue; 917 /* 918 * This is the case where multiple interfaces 919 * have the same prefix, but only one is installed 920 * into the routing table and that prefix entry 921 * is not the one being examined here. In the case 922 * where RADIX_MPATH is enabled, multiple route 923 * entries (of the same rt_key value) will be 924 * installed because the interface addresses all 925 * differ. 926 */ 927 if (!IN6_ARE_ADDR_EQUAL(&pr->ndpr_prefix.sin6_addr, 928 &((struct sockaddr_in6 *)rt_key(rt))->sin6_addr)) { 929 RTFREE_LOCKED(rt); 930 continue; 931 } 932 RTFREE_LOCKED(rt); 933 } 934 935 if (IN6_ARE_MASKED_ADDR_EQUAL(&pr->ndpr_prefix.sin6_addr, 936 &addr->sin6_addr, &pr->ndpr_mask)) 937 return (1); 938 } 939 940 /* 941 * If the address is assigned on the node of the other side of 942 * a p2p interface, the address should be a neighbor. 943 */ 944 dstaddr = ifa_ifwithdstaddr((struct sockaddr *)addr); 945 if (dstaddr != NULL) { 946 if (dstaddr->ifa_ifp == ifp) { 947 ifa_free(dstaddr); 948 return (1); 949 } 950 ifa_free(dstaddr); 951 } 952 953 /* 954 * If the default router list is empty, all addresses are regarded 955 * as on-link, and thus, as a neighbor. 956 */ 957 if (ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV && 958 TAILQ_EMPTY(&V_nd_defrouter) && 959 V_nd6_defifindex == ifp->if_index) { 960 return (1); 961 } 962 963 return (0); 964 } 965 966 967 /* 968 * Detect if a given IPv6 address identifies a neighbor on a given link. 969 * XXX: should take care of the destination of a p2p link? 970 */ 971 int 972 nd6_is_addr_neighbor(struct sockaddr_in6 *addr, struct ifnet *ifp) 973 { 974 struct llentry *lle; 975 int rc = 0; 976 977 IF_AFDATA_UNLOCK_ASSERT(ifp); 978 if (nd6_is_new_addr_neighbor(addr, ifp)) 979 return (1); 980 981 /* 982 * Even if the address matches none of our addresses, it might be 983 * in the neighbor cache. 984 */ 985 IF_AFDATA_LOCK(ifp); 986 if ((lle = nd6_lookup(&addr->sin6_addr, 0, ifp)) != NULL) { 987 LLE_RUNLOCK(lle); 988 rc = 1; 989 } 990 IF_AFDATA_UNLOCK(ifp); 991 return (rc); 992 } 993 994 /* 995 * Free an nd6 llinfo entry. 996 * Since the function would cause significant changes in the kernel, DO NOT 997 * make it global, unless you have a strong reason for the change, and are sure 998 * that the change is safe. 999 */ 1000 static struct llentry * 1001 nd6_free(struct llentry *ln, int gc) 1002 { 1003 struct llentry *next; 1004 struct nd_defrouter *dr; 1005 struct ifnet *ifp; 1006 1007 LLE_WLOCK_ASSERT(ln); 1008 1009 /* 1010 * we used to have pfctlinput(PRC_HOSTDEAD) here. 1011 * even though it is not harmful, it was not really necessary. 1012 */ 1013 1014 /* cancel timer */ 1015 nd6_llinfo_settimer_locked(ln, -1); 1016 1017 ifp = ln->lle_tbl->llt_ifp; 1018 1019 if (ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV) { 1020 dr = defrouter_lookup(&L3_ADDR_SIN6(ln)->sin6_addr, ifp); 1021 1022 if (dr != NULL && dr->expire && 1023 ln->ln_state == ND6_LLINFO_STALE && gc) { 1024 /* 1025 * If the reason for the deletion is just garbage 1026 * collection, and the neighbor is an active default 1027 * router, do not delete it. Instead, reset the GC 1028 * timer using the router's lifetime. 1029 * Simply deleting the entry would affect default 1030 * router selection, which is not necessarily a good 1031 * thing, especially when we're using router preference 1032 * values. 1033 * XXX: the check for ln_state would be redundant, 1034 * but we intentionally keep it just in case. 1035 */ 1036 if (dr->expire > time_second) 1037 nd6_llinfo_settimer_locked(ln, 1038 (dr->expire - time_second) * hz); 1039 else 1040 nd6_llinfo_settimer_locked(ln, 1041 (long)V_nd6_gctimer * hz); 1042 1043 next = LIST_NEXT(ln, lle_next); 1044 LLE_REMREF(ln); 1045 LLE_WUNLOCK(ln); 1046 return (next); 1047 } 1048 1049 if (dr) { 1050 /* 1051 * Unreachablity of a router might affect the default 1052 * router selection and on-link detection of advertised 1053 * prefixes. 1054 */ 1055 1056 /* 1057 * Temporarily fake the state to choose a new default 1058 * router and to perform on-link determination of 1059 * prefixes correctly. 1060 * Below the state will be set correctly, 1061 * or the entry itself will be deleted. 1062 */ 1063 ln->ln_state = ND6_LLINFO_INCOMPLETE; 1064 } 1065 1066 if (ln->ln_router || dr) { 1067 1068 /* 1069 * We need to unlock to avoid a LOR with rt6_flush() with the 1070 * rnh and for the calls to pfxlist_onlink_check() and 1071 * defrouter_select() in the block further down for calls 1072 * into nd6_lookup(). We still hold a ref. 1073 */ 1074 LLE_WUNLOCK(ln); 1075 1076 /* 1077 * rt6_flush must be called whether or not the neighbor 1078 * is in the Default Router List. 1079 * See a corresponding comment in nd6_na_input(). 1080 */ 1081 rt6_flush(&L3_ADDR_SIN6(ln)->sin6_addr, ifp); 1082 } 1083 1084 if (dr) { 1085 /* 1086 * Since defrouter_select() does not affect the 1087 * on-link determination and MIP6 needs the check 1088 * before the default router selection, we perform 1089 * the check now. 1090 */ 1091 pfxlist_onlink_check(); 1092 1093 /* 1094 * Refresh default router list. 1095 */ 1096 defrouter_select(); 1097 } 1098 1099 if (ln->ln_router || dr) 1100 LLE_WLOCK(ln); 1101 } 1102 1103 /* 1104 * Before deleting the entry, remember the next entry as the 1105 * return value. We need this because pfxlist_onlink_check() above 1106 * might have freed other entries (particularly the old next entry) as 1107 * a side effect (XXX). 1108 */ 1109 next = LIST_NEXT(ln, lle_next); 1110 1111 /* 1112 * Save to unlock. We still hold an extra reference and will not 1113 * free(9) in llentry_free() if someone else holds one as well. 1114 */ 1115 LLE_WUNLOCK(ln); 1116 IF_AFDATA_LOCK(ifp); 1117 LLE_WLOCK(ln); 1118 LLE_REMREF(ln); 1119 llentry_free(ln); 1120 IF_AFDATA_UNLOCK(ifp); 1121 1122 return (next); 1123 } 1124 1125 /* 1126 * Upper-layer reachability hint for Neighbor Unreachability Detection. 1127 * 1128 * XXX cost-effective methods? 1129 */ 1130 void 1131 nd6_nud_hint(struct rtentry *rt, struct in6_addr *dst6, int force) 1132 { 1133 struct llentry *ln; 1134 struct ifnet *ifp; 1135 1136 if ((dst6 == NULL) || (rt == NULL)) 1137 return; 1138 1139 ifp = rt->rt_ifp; 1140 IF_AFDATA_LOCK(ifp); 1141 ln = nd6_lookup(dst6, ND6_EXCLUSIVE, NULL); 1142 IF_AFDATA_UNLOCK(ifp); 1143 if (ln == NULL) 1144 return; 1145 1146 if (ln->ln_state < ND6_LLINFO_REACHABLE) 1147 goto done; 1148 1149 /* 1150 * if we get upper-layer reachability confirmation many times, 1151 * it is possible we have false information. 1152 */ 1153 if (!force) { 1154 ln->ln_byhint++; 1155 if (ln->ln_byhint > V_nd6_maxnudhint) { 1156 goto done; 1157 } 1158 } 1159 1160 ln->ln_state = ND6_LLINFO_REACHABLE; 1161 if (!ND6_LLINFO_PERMANENT(ln)) { 1162 nd6_llinfo_settimer_locked(ln, 1163 (long)ND_IFINFO(rt->rt_ifp)->reachable * hz); 1164 } 1165 done: 1166 LLE_WUNLOCK(ln); 1167 } 1168 1169 1170 /* 1171 * Rejuvenate this function for routing operations related 1172 * processing. 1173 */ 1174 void 1175 nd6_rtrequest(int req, struct rtentry *rt, struct rt_addrinfo *info) 1176 { 1177 struct sockaddr_in6 *gateway = (struct sockaddr_in6 *)rt->rt_gateway; 1178 struct nd_defrouter *dr; 1179 struct ifnet *ifp = rt->rt_ifp; 1180 1181 RT_LOCK_ASSERT(rt); 1182 1183 switch (req) { 1184 case RTM_ADD: 1185 break; 1186 1187 case RTM_DELETE: 1188 if (!ifp) 1189 return; 1190 /* 1191 * Only indirect routes are interesting. 1192 */ 1193 if ((rt->rt_flags & RTF_GATEWAY) == 0) 1194 return; 1195 /* 1196 * check for default route 1197 */ 1198 if (IN6_ARE_ADDR_EQUAL(&in6addr_any, 1199 &SIN6(rt_key(rt))->sin6_addr)) { 1200 1201 dr = defrouter_lookup(&gateway->sin6_addr, ifp); 1202 if (dr != NULL) 1203 dr->installed = 0; 1204 } 1205 break; 1206 } 1207 } 1208 1209 1210 int 1211 nd6_ioctl(u_long cmd, caddr_t data, struct ifnet *ifp) 1212 { 1213 struct in6_drlist *drl = (struct in6_drlist *)data; 1214 struct in6_oprlist *oprl = (struct in6_oprlist *)data; 1215 struct in6_ndireq *ndi = (struct in6_ndireq *)data; 1216 struct in6_nbrinfo *nbi = (struct in6_nbrinfo *)data; 1217 struct in6_ndifreq *ndif = (struct in6_ndifreq *)data; 1218 struct nd_defrouter *dr; 1219 struct nd_prefix *pr; 1220 int i = 0, error = 0; 1221 int s; 1222 1223 switch (cmd) { 1224 case SIOCGDRLST_IN6: 1225 /* 1226 * obsolete API, use sysctl under net.inet6.icmp6 1227 */ 1228 bzero(drl, sizeof(*drl)); 1229 s = splnet(); 1230 TAILQ_FOREACH(dr, &V_nd_defrouter, dr_entry) { 1231 if (i >= DRLSTSIZ) 1232 break; 1233 drl->defrouter[i].rtaddr = dr->rtaddr; 1234 in6_clearscope(&drl->defrouter[i].rtaddr); 1235 1236 drl->defrouter[i].flags = dr->flags; 1237 drl->defrouter[i].rtlifetime = dr->rtlifetime; 1238 drl->defrouter[i].expire = dr->expire; 1239 drl->defrouter[i].if_index = dr->ifp->if_index; 1240 i++; 1241 } 1242 splx(s); 1243 break; 1244 case SIOCGPRLST_IN6: 1245 /* 1246 * obsolete API, use sysctl under net.inet6.icmp6 1247 * 1248 * XXX the structure in6_prlist was changed in backward- 1249 * incompatible manner. in6_oprlist is used for SIOCGPRLST_IN6, 1250 * in6_prlist is used for nd6_sysctl() - fill_prlist(). 1251 */ 1252 /* 1253 * XXX meaning of fields, especialy "raflags", is very 1254 * differnet between RA prefix list and RR/static prefix list. 1255 * how about separating ioctls into two? 1256 */ 1257 bzero(oprl, sizeof(*oprl)); 1258 s = splnet(); 1259 LIST_FOREACH(pr, &V_nd_prefix, ndpr_entry) { 1260 struct nd_pfxrouter *pfr; 1261 int j; 1262 1263 if (i >= PRLSTSIZ) 1264 break; 1265 oprl->prefix[i].prefix = pr->ndpr_prefix.sin6_addr; 1266 oprl->prefix[i].raflags = pr->ndpr_raf; 1267 oprl->prefix[i].prefixlen = pr->ndpr_plen; 1268 oprl->prefix[i].vltime = pr->ndpr_vltime; 1269 oprl->prefix[i].pltime = pr->ndpr_pltime; 1270 oprl->prefix[i].if_index = pr->ndpr_ifp->if_index; 1271 if (pr->ndpr_vltime == ND6_INFINITE_LIFETIME) 1272 oprl->prefix[i].expire = 0; 1273 else { 1274 time_t maxexpire; 1275 1276 /* XXX: we assume time_t is signed. */ 1277 maxexpire = (-1) & 1278 ~((time_t)1 << 1279 ((sizeof(maxexpire) * 8) - 1)); 1280 if (pr->ndpr_vltime < 1281 maxexpire - pr->ndpr_lastupdate) { 1282 oprl->prefix[i].expire = 1283 pr->ndpr_lastupdate + 1284 pr->ndpr_vltime; 1285 } else 1286 oprl->prefix[i].expire = maxexpire; 1287 } 1288 1289 j = 0; 1290 LIST_FOREACH(pfr, &pr->ndpr_advrtrs, pfr_entry) { 1291 if (j < DRLSTSIZ) { 1292 #define RTRADDR oprl->prefix[i].advrtr[j] 1293 RTRADDR = pfr->router->rtaddr; 1294 in6_clearscope(&RTRADDR); 1295 #undef RTRADDR 1296 } 1297 j++; 1298 } 1299 oprl->prefix[i].advrtrs = j; 1300 oprl->prefix[i].origin = PR_ORIG_RA; 1301 1302 i++; 1303 } 1304 splx(s); 1305 1306 break; 1307 case OSIOCGIFINFO_IN6: 1308 #define ND ndi->ndi 1309 /* XXX: old ndp(8) assumes a positive value for linkmtu. */ 1310 bzero(&ND, sizeof(ND)); 1311 ND.linkmtu = IN6_LINKMTU(ifp); 1312 ND.maxmtu = ND_IFINFO(ifp)->maxmtu; 1313 ND.basereachable = ND_IFINFO(ifp)->basereachable; 1314 ND.reachable = ND_IFINFO(ifp)->reachable; 1315 ND.retrans = ND_IFINFO(ifp)->retrans; 1316 ND.flags = ND_IFINFO(ifp)->flags; 1317 ND.recalctm = ND_IFINFO(ifp)->recalctm; 1318 ND.chlim = ND_IFINFO(ifp)->chlim; 1319 break; 1320 case SIOCGIFINFO_IN6: 1321 ND = *ND_IFINFO(ifp); 1322 break; 1323 case SIOCSIFINFO_IN6: 1324 /* 1325 * used to change host variables from userland. 1326 * intented for a use on router to reflect RA configurations. 1327 */ 1328 /* 0 means 'unspecified' */ 1329 if (ND.linkmtu != 0) { 1330 if (ND.linkmtu < IPV6_MMTU || 1331 ND.linkmtu > IN6_LINKMTU(ifp)) { 1332 error = EINVAL; 1333 break; 1334 } 1335 ND_IFINFO(ifp)->linkmtu = ND.linkmtu; 1336 } 1337 1338 if (ND.basereachable != 0) { 1339 int obasereachable = ND_IFINFO(ifp)->basereachable; 1340 1341 ND_IFINFO(ifp)->basereachable = ND.basereachable; 1342 if (ND.basereachable != obasereachable) 1343 ND_IFINFO(ifp)->reachable = 1344 ND_COMPUTE_RTIME(ND.basereachable); 1345 } 1346 if (ND.retrans != 0) 1347 ND_IFINFO(ifp)->retrans = ND.retrans; 1348 if (ND.chlim != 0) 1349 ND_IFINFO(ifp)->chlim = ND.chlim; 1350 /* FALLTHROUGH */ 1351 case SIOCSIFINFO_FLAGS: 1352 { 1353 struct ifaddr *ifa; 1354 struct in6_ifaddr *ia; 1355 1356 if ((ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) && 1357 !(ND.flags & ND6_IFF_IFDISABLED)) { 1358 /* ifdisabled 1->0 transision */ 1359 1360 /* 1361 * If the interface is marked as ND6_IFF_IFDISABLED and 1362 * has an link-local address with IN6_IFF_DUPLICATED, 1363 * do not clear ND6_IFF_IFDISABLED. 1364 * See RFC 4862, Section 5.4.5. 1365 */ 1366 int duplicated_linklocal = 0; 1367 1368 IF_ADDR_RLOCK(ifp); 1369 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1370 if (ifa->ifa_addr->sa_family != AF_INET6) 1371 continue; 1372 ia = (struct in6_ifaddr *)ifa; 1373 if ((ia->ia6_flags & IN6_IFF_DUPLICATED) && 1374 IN6_IS_ADDR_LINKLOCAL(IA6_IN6(ia))) { 1375 duplicated_linklocal = 1; 1376 break; 1377 } 1378 } 1379 IF_ADDR_RUNLOCK(ifp); 1380 1381 if (duplicated_linklocal) { 1382 ND.flags |= ND6_IFF_IFDISABLED; 1383 log(LOG_ERR, "Cannot enable an interface" 1384 " with a link-local address marked" 1385 " duplicate.\n"); 1386 } else { 1387 ND_IFINFO(ifp)->flags &= ~ND6_IFF_IFDISABLED; 1388 if (ifp->if_flags & IFF_UP) 1389 in6_if_up(ifp); 1390 } 1391 } else if (!(ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) && 1392 (ND.flags & ND6_IFF_IFDISABLED)) { 1393 /* ifdisabled 0->1 transision */ 1394 /* Mark all IPv6 address as tentative. */ 1395 1396 ND_IFINFO(ifp)->flags |= ND6_IFF_IFDISABLED; 1397 IF_ADDR_RLOCK(ifp); 1398 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1399 if (ifa->ifa_addr->sa_family != AF_INET6) 1400 continue; 1401 ia = (struct in6_ifaddr *)ifa; 1402 ia->ia6_flags |= IN6_IFF_TENTATIVE; 1403 } 1404 IF_ADDR_RUNLOCK(ifp); 1405 } 1406 1407 if (ND.flags & ND6_IFF_AUTO_LINKLOCAL) { 1408 if (!(ND_IFINFO(ifp)->flags & ND6_IFF_AUTO_LINKLOCAL)) { 1409 /* auto_linklocal 0->1 transision */ 1410 1411 /* If no link-local address on ifp, configure */ 1412 ND_IFINFO(ifp)->flags |= ND6_IFF_AUTO_LINKLOCAL; 1413 in6_ifattach(ifp, NULL); 1414 } else if (!(ND.flags & ND6_IFF_IFDISABLED) && 1415 ifp->if_flags & IFF_UP) { 1416 /* 1417 * When the IF already has 1418 * ND6_IFF_AUTO_LINKLOCAL, no link-local 1419 * address is assigned, and IFF_UP, try to 1420 * assign one. 1421 */ 1422 int haslinklocal = 0; 1423 1424 IF_ADDR_RLOCK(ifp); 1425 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1426 if (ifa->ifa_addr->sa_family != AF_INET6) 1427 continue; 1428 ia = (struct in6_ifaddr *)ifa; 1429 if (IN6_IS_ADDR_LINKLOCAL(IA6_IN6(ia))) { 1430 haslinklocal = 1; 1431 break; 1432 } 1433 } 1434 IF_ADDR_RUNLOCK(ifp); 1435 if (!haslinklocal) 1436 in6_ifattach(ifp, NULL); 1437 } 1438 } 1439 } 1440 ND_IFINFO(ifp)->flags = ND.flags; 1441 break; 1442 #undef ND 1443 case SIOCSNDFLUSH_IN6: /* XXX: the ioctl name is confusing... */ 1444 /* sync kernel routing table with the default router list */ 1445 defrouter_reset(); 1446 defrouter_select(); 1447 break; 1448 case SIOCSPFXFLUSH_IN6: 1449 { 1450 /* flush all the prefix advertised by routers */ 1451 struct nd_prefix *pr, *next; 1452 1453 s = splnet(); 1454 LIST_FOREACH_SAFE(pr, &V_nd_prefix, ndpr_entry, next) { 1455 struct in6_ifaddr *ia, *ia_next; 1456 1457 if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr)) 1458 continue; /* XXX */ 1459 1460 /* do we really have to remove addresses as well? */ 1461 /* XXXRW: in6_ifaddrhead locking. */ 1462 TAILQ_FOREACH_SAFE(ia, &V_in6_ifaddrhead, ia_link, 1463 ia_next) { 1464 if ((ia->ia6_flags & IN6_IFF_AUTOCONF) == 0) 1465 continue; 1466 1467 if (ia->ia6_ndpr == pr) 1468 in6_purgeaddr(&ia->ia_ifa); 1469 } 1470 prelist_remove(pr); 1471 } 1472 splx(s); 1473 break; 1474 } 1475 case SIOCSRTRFLUSH_IN6: 1476 { 1477 /* flush all the default routers */ 1478 struct nd_defrouter *dr, *next; 1479 1480 s = splnet(); 1481 defrouter_reset(); 1482 TAILQ_FOREACH_SAFE(dr, &V_nd_defrouter, dr_entry, next) { 1483 defrtrlist_del(dr); 1484 } 1485 defrouter_select(); 1486 splx(s); 1487 break; 1488 } 1489 case SIOCGNBRINFO_IN6: 1490 { 1491 struct llentry *ln; 1492 struct in6_addr nb_addr = nbi->addr; /* make local for safety */ 1493 1494 if ((error = in6_setscope(&nb_addr, ifp, NULL)) != 0) 1495 return (error); 1496 1497 IF_AFDATA_LOCK(ifp); 1498 ln = nd6_lookup(&nb_addr, 0, ifp); 1499 IF_AFDATA_UNLOCK(ifp); 1500 1501 if (ln == NULL) { 1502 error = EINVAL; 1503 break; 1504 } 1505 nbi->state = ln->ln_state; 1506 nbi->asked = ln->la_asked; 1507 nbi->isrouter = ln->ln_router; 1508 nbi->expire = ln->la_expire; 1509 LLE_RUNLOCK(ln); 1510 break; 1511 } 1512 case SIOCGDEFIFACE_IN6: /* XXX: should be implemented as a sysctl? */ 1513 ndif->ifindex = V_nd6_defifindex; 1514 break; 1515 case SIOCSDEFIFACE_IN6: /* XXX: should be implemented as a sysctl? */ 1516 return (nd6_setdefaultiface(ndif->ifindex)); 1517 } 1518 return (error); 1519 } 1520 1521 /* 1522 * Create neighbor cache entry and cache link-layer address, 1523 * on reception of inbound ND6 packets. (RS/RA/NS/redirect) 1524 * 1525 * type - ICMP6 type 1526 * code - type dependent information 1527 * 1528 * XXXXX 1529 * The caller of this function already acquired the ndp 1530 * cache table lock because the cache entry is returned. 1531 */ 1532 struct llentry * 1533 nd6_cache_lladdr(struct ifnet *ifp, struct in6_addr *from, char *lladdr, 1534 int lladdrlen, int type, int code) 1535 { 1536 struct llentry *ln = NULL; 1537 int is_newentry; 1538 int do_update; 1539 int olladdr; 1540 int llchange; 1541 int flags; 1542 int newstate = 0; 1543 uint16_t router = 0; 1544 struct sockaddr_in6 sin6; 1545 struct mbuf *chain = NULL; 1546 int static_route = 0; 1547 1548 IF_AFDATA_UNLOCK_ASSERT(ifp); 1549 1550 if (ifp == NULL) 1551 panic("ifp == NULL in nd6_cache_lladdr"); 1552 if (from == NULL) 1553 panic("from == NULL in nd6_cache_lladdr"); 1554 1555 /* nothing must be updated for unspecified address */ 1556 if (IN6_IS_ADDR_UNSPECIFIED(from)) 1557 return NULL; 1558 1559 /* 1560 * Validation about ifp->if_addrlen and lladdrlen must be done in 1561 * the caller. 1562 * 1563 * XXX If the link does not have link-layer adderss, what should 1564 * we do? (ifp->if_addrlen == 0) 1565 * Spec says nothing in sections for RA, RS and NA. There's small 1566 * description on it in NS section (RFC 2461 7.2.3). 1567 */ 1568 flags = lladdr ? ND6_EXCLUSIVE : 0; 1569 IF_AFDATA_LOCK(ifp); 1570 ln = nd6_lookup(from, flags, ifp); 1571 1572 if (ln == NULL) { 1573 flags |= ND6_EXCLUSIVE; 1574 ln = nd6_lookup(from, flags | ND6_CREATE, ifp); 1575 IF_AFDATA_UNLOCK(ifp); 1576 is_newentry = 1; 1577 } else { 1578 IF_AFDATA_UNLOCK(ifp); 1579 /* do nothing if static ndp is set */ 1580 if (ln->la_flags & LLE_STATIC) { 1581 static_route = 1; 1582 goto done; 1583 } 1584 is_newentry = 0; 1585 } 1586 if (ln == NULL) 1587 return (NULL); 1588 1589 olladdr = (ln->la_flags & LLE_VALID) ? 1 : 0; 1590 if (olladdr && lladdr) { 1591 llchange = bcmp(lladdr, &ln->ll_addr, 1592 ifp->if_addrlen); 1593 } else 1594 llchange = 0; 1595 1596 /* 1597 * newentry olladdr lladdr llchange (*=record) 1598 * 0 n n -- (1) 1599 * 0 y n -- (2) 1600 * 0 n y -- (3) * STALE 1601 * 0 y y n (4) * 1602 * 0 y y y (5) * STALE 1603 * 1 -- n -- (6) NOSTATE(= PASSIVE) 1604 * 1 -- y -- (7) * STALE 1605 */ 1606 1607 if (lladdr) { /* (3-5) and (7) */ 1608 /* 1609 * Record source link-layer address 1610 * XXX is it dependent to ifp->if_type? 1611 */ 1612 bcopy(lladdr, &ln->ll_addr, ifp->if_addrlen); 1613 ln->la_flags |= LLE_VALID; 1614 } 1615 1616 if (!is_newentry) { 1617 if ((!olladdr && lladdr != NULL) || /* (3) */ 1618 (olladdr && lladdr != NULL && llchange)) { /* (5) */ 1619 do_update = 1; 1620 newstate = ND6_LLINFO_STALE; 1621 } else /* (1-2,4) */ 1622 do_update = 0; 1623 } else { 1624 do_update = 1; 1625 if (lladdr == NULL) /* (6) */ 1626 newstate = ND6_LLINFO_NOSTATE; 1627 else /* (7) */ 1628 newstate = ND6_LLINFO_STALE; 1629 } 1630 1631 if (do_update) { 1632 /* 1633 * Update the state of the neighbor cache. 1634 */ 1635 ln->ln_state = newstate; 1636 1637 if (ln->ln_state == ND6_LLINFO_STALE) { 1638 /* 1639 * XXX: since nd6_output() below will cause 1640 * state tansition to DELAY and reset the timer, 1641 * we must set the timer now, although it is actually 1642 * meaningless. 1643 */ 1644 nd6_llinfo_settimer_locked(ln, (long)V_nd6_gctimer * hz); 1645 1646 if (ln->la_hold) { 1647 struct mbuf *m_hold, *m_hold_next; 1648 1649 /* 1650 * reset the la_hold in advance, to explicitly 1651 * prevent a la_hold lookup in nd6_output() 1652 * (wouldn't happen, though...) 1653 */ 1654 for (m_hold = ln->la_hold, ln->la_hold = NULL; 1655 m_hold; m_hold = m_hold_next) { 1656 m_hold_next = m_hold->m_nextpkt; 1657 m_hold->m_nextpkt = NULL; 1658 1659 /* 1660 * we assume ifp is not a p2p here, so 1661 * just set the 2nd argument as the 1662 * 1st one. 1663 */ 1664 nd6_output_lle(ifp, ifp, m_hold, L3_ADDR_SIN6(ln), NULL, ln, &chain); 1665 } 1666 /* 1667 * If we have mbufs in the chain we need to do 1668 * deferred transmit. Copy the address from the 1669 * llentry before dropping the lock down below. 1670 */ 1671 if (chain != NULL) 1672 memcpy(&sin6, L3_ADDR_SIN6(ln), sizeof(sin6)); 1673 } 1674 } else if (ln->ln_state == ND6_LLINFO_INCOMPLETE) { 1675 /* probe right away */ 1676 nd6_llinfo_settimer_locked((void *)ln, 0); 1677 } 1678 } 1679 1680 /* 1681 * ICMP6 type dependent behavior. 1682 * 1683 * NS: clear IsRouter if new entry 1684 * RS: clear IsRouter 1685 * RA: set IsRouter if there's lladdr 1686 * redir: clear IsRouter if new entry 1687 * 1688 * RA case, (1): 1689 * The spec says that we must set IsRouter in the following cases: 1690 * - If lladdr exist, set IsRouter. This means (1-5). 1691 * - If it is old entry (!newentry), set IsRouter. This means (7). 1692 * So, based on the spec, in (1-5) and (7) cases we must set IsRouter. 1693 * A quetion arises for (1) case. (1) case has no lladdr in the 1694 * neighbor cache, this is similar to (6). 1695 * This case is rare but we figured that we MUST NOT set IsRouter. 1696 * 1697 * newentry olladdr lladdr llchange NS RS RA redir 1698 * D R 1699 * 0 n n -- (1) c ? s 1700 * 0 y n -- (2) c s s 1701 * 0 n y -- (3) c s s 1702 * 0 y y n (4) c s s 1703 * 0 y y y (5) c s s 1704 * 1 -- n -- (6) c c c s 1705 * 1 -- y -- (7) c c s c s 1706 * 1707 * (c=clear s=set) 1708 */ 1709 switch (type & 0xff) { 1710 case ND_NEIGHBOR_SOLICIT: 1711 /* 1712 * New entry must have is_router flag cleared. 1713 */ 1714 if (is_newentry) /* (6-7) */ 1715 ln->ln_router = 0; 1716 break; 1717 case ND_REDIRECT: 1718 /* 1719 * If the icmp is a redirect to a better router, always set the 1720 * is_router flag. Otherwise, if the entry is newly created, 1721 * clear the flag. [RFC 2461, sec 8.3] 1722 */ 1723 if (code == ND_REDIRECT_ROUTER) 1724 ln->ln_router = 1; 1725 else if (is_newentry) /* (6-7) */ 1726 ln->ln_router = 0; 1727 break; 1728 case ND_ROUTER_SOLICIT: 1729 /* 1730 * is_router flag must always be cleared. 1731 */ 1732 ln->ln_router = 0; 1733 break; 1734 case ND_ROUTER_ADVERT: 1735 /* 1736 * Mark an entry with lladdr as a router. 1737 */ 1738 if ((!is_newentry && (olladdr || lladdr)) || /* (2-5) */ 1739 (is_newentry && lladdr)) { /* (7) */ 1740 ln->ln_router = 1; 1741 } 1742 break; 1743 } 1744 1745 if (ln != NULL) { 1746 static_route = (ln->la_flags & LLE_STATIC); 1747 router = ln->ln_router; 1748 1749 if (flags & ND6_EXCLUSIVE) 1750 LLE_WUNLOCK(ln); 1751 else 1752 LLE_RUNLOCK(ln); 1753 if (static_route) 1754 ln = NULL; 1755 } 1756 if (chain) 1757 nd6_output_flush(ifp, ifp, chain, &sin6, NULL); 1758 1759 /* 1760 * When the link-layer address of a router changes, select the 1761 * best router again. In particular, when the neighbor entry is newly 1762 * created, it might affect the selection policy. 1763 * Question: can we restrict the first condition to the "is_newentry" 1764 * case? 1765 * XXX: when we hear an RA from a new router with the link-layer 1766 * address option, defrouter_select() is called twice, since 1767 * defrtrlist_update called the function as well. However, I believe 1768 * we can compromise the overhead, since it only happens the first 1769 * time. 1770 * XXX: although defrouter_select() should not have a bad effect 1771 * for those are not autoconfigured hosts, we explicitly avoid such 1772 * cases for safety. 1773 */ 1774 if (do_update && router && 1775 ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV) { 1776 /* 1777 * guaranteed recursion 1778 */ 1779 defrouter_select(); 1780 } 1781 1782 return (ln); 1783 done: 1784 if (ln != NULL) { 1785 if (flags & ND6_EXCLUSIVE) 1786 LLE_WUNLOCK(ln); 1787 else 1788 LLE_RUNLOCK(ln); 1789 if (static_route) 1790 ln = NULL; 1791 } 1792 return (ln); 1793 } 1794 1795 static void 1796 nd6_slowtimo(void *arg) 1797 { 1798 CURVNET_SET((struct vnet *) arg); 1799 struct nd_ifinfo *nd6if; 1800 struct ifnet *ifp; 1801 1802 callout_reset(&V_nd6_slowtimo_ch, ND6_SLOWTIMER_INTERVAL * hz, 1803 nd6_slowtimo, curvnet); 1804 IFNET_RLOCK_NOSLEEP(); 1805 TAILQ_FOREACH(ifp, &V_ifnet, if_list) { 1806 nd6if = ND_IFINFO(ifp); 1807 if (nd6if->basereachable && /* already initialized */ 1808 (nd6if->recalctm -= ND6_SLOWTIMER_INTERVAL) <= 0) { 1809 /* 1810 * Since reachable time rarely changes by router 1811 * advertisements, we SHOULD insure that a new random 1812 * value gets recomputed at least once every few hours. 1813 * (RFC 2461, 6.3.4) 1814 */ 1815 nd6if->recalctm = V_nd6_recalc_reachtm_interval; 1816 nd6if->reachable = ND_COMPUTE_RTIME(nd6if->basereachable); 1817 } 1818 } 1819 IFNET_RUNLOCK_NOSLEEP(); 1820 CURVNET_RESTORE(); 1821 } 1822 1823 int 1824 nd6_output(struct ifnet *ifp, struct ifnet *origifp, struct mbuf *m0, 1825 struct sockaddr_in6 *dst, struct rtentry *rt0) 1826 { 1827 1828 return (nd6_output_lle(ifp, origifp, m0, dst, rt0, NULL, NULL)); 1829 } 1830 1831 1832 /* 1833 * Note that I'm not enforcing any global serialization 1834 * lle state or asked changes here as the logic is too 1835 * complicated to avoid having to always acquire an exclusive 1836 * lock 1837 * KMM 1838 * 1839 */ 1840 #define senderr(e) { error = (e); goto bad;} 1841 1842 int 1843 nd6_output_lle(struct ifnet *ifp, struct ifnet *origifp, struct mbuf *m0, 1844 struct sockaddr_in6 *dst, struct rtentry *rt0, struct llentry *lle, 1845 struct mbuf **chain) 1846 { 1847 struct mbuf *m = m0; 1848 struct m_tag *mtag; 1849 struct llentry *ln = lle; 1850 struct ip6_hdr *ip6; 1851 int error = 0; 1852 int flags = 0; 1853 int ip6len; 1854 1855 #ifdef INVARIANTS 1856 if (lle != NULL) { 1857 1858 LLE_WLOCK_ASSERT(lle); 1859 1860 KASSERT(chain != NULL, (" lle locked but no mbuf chain pointer passed")); 1861 } 1862 #endif 1863 if (IN6_IS_ADDR_MULTICAST(&dst->sin6_addr)) 1864 goto sendpkt; 1865 1866 if (nd6_need_cache(ifp) == 0) 1867 goto sendpkt; 1868 1869 /* 1870 * next hop determination. This routine is derived from ether_output. 1871 */ 1872 1873 /* 1874 * Address resolution or Neighbor Unreachability Detection 1875 * for the next hop. 1876 * At this point, the destination of the packet must be a unicast 1877 * or an anycast address(i.e. not a multicast). 1878 */ 1879 1880 flags = ((m != NULL) || (lle != NULL)) ? LLE_EXCLUSIVE : 0; 1881 if (ln == NULL) { 1882 retry: 1883 IF_AFDATA_LOCK(ifp); 1884 ln = lla_lookup(LLTABLE6(ifp), flags, (struct sockaddr *)dst); 1885 IF_AFDATA_UNLOCK(ifp); 1886 if ((ln == NULL) && nd6_is_addr_neighbor(dst, ifp)) { 1887 /* 1888 * Since nd6_is_addr_neighbor() internally calls nd6_lookup(), 1889 * the condition below is not very efficient. But we believe 1890 * it is tolerable, because this should be a rare case. 1891 */ 1892 flags = ND6_CREATE | (m ? ND6_EXCLUSIVE : 0); 1893 IF_AFDATA_LOCK(ifp); 1894 ln = nd6_lookup(&dst->sin6_addr, flags, ifp); 1895 IF_AFDATA_UNLOCK(ifp); 1896 } 1897 } 1898 if (ln == NULL) { 1899 if ((ifp->if_flags & IFF_POINTOPOINT) == 0 && 1900 !(ND_IFINFO(ifp)->flags & ND6_IFF_PERFORMNUD)) { 1901 char ip6buf[INET6_ADDRSTRLEN]; 1902 log(LOG_DEBUG, 1903 "nd6_output: can't allocate llinfo for %s " 1904 "(ln=%p)\n", 1905 ip6_sprintf(ip6buf, &dst->sin6_addr), ln); 1906 senderr(EIO); /* XXX: good error? */ 1907 } 1908 goto sendpkt; /* send anyway */ 1909 } 1910 1911 /* We don't have to do link-layer address resolution on a p2p link. */ 1912 if ((ifp->if_flags & IFF_POINTOPOINT) != 0 && 1913 ln->ln_state < ND6_LLINFO_REACHABLE) { 1914 if ((flags & LLE_EXCLUSIVE) == 0) { 1915 flags |= LLE_EXCLUSIVE; 1916 goto retry; 1917 } 1918 ln->ln_state = ND6_LLINFO_STALE; 1919 nd6_llinfo_settimer_locked(ln, (long)V_nd6_gctimer * hz); 1920 } 1921 1922 /* 1923 * The first time we send a packet to a neighbor whose entry is 1924 * STALE, we have to change the state to DELAY and a sets a timer to 1925 * expire in DELAY_FIRST_PROBE_TIME seconds to ensure do 1926 * neighbor unreachability detection on expiration. 1927 * (RFC 2461 7.3.3) 1928 */ 1929 if (ln->ln_state == ND6_LLINFO_STALE) { 1930 if ((flags & LLE_EXCLUSIVE) == 0) { 1931 flags |= LLE_EXCLUSIVE; 1932 LLE_RUNLOCK(ln); 1933 goto retry; 1934 } 1935 ln->la_asked = 0; 1936 ln->ln_state = ND6_LLINFO_DELAY; 1937 nd6_llinfo_settimer_locked(ln, (long)V_nd6_delay * hz); 1938 } 1939 1940 /* 1941 * If the neighbor cache entry has a state other than INCOMPLETE 1942 * (i.e. its link-layer address is already resolved), just 1943 * send the packet. 1944 */ 1945 if (ln->ln_state > ND6_LLINFO_INCOMPLETE) 1946 goto sendpkt; 1947 1948 /* 1949 * There is a neighbor cache entry, but no ethernet address 1950 * response yet. Append this latest packet to the end of the 1951 * packet queue in the mbuf, unless the number of the packet 1952 * does not exceed nd6_maxqueuelen. When it exceeds nd6_maxqueuelen, 1953 * the oldest packet in the queue will be removed. 1954 */ 1955 if (ln->ln_state == ND6_LLINFO_NOSTATE) 1956 ln->ln_state = ND6_LLINFO_INCOMPLETE; 1957 1958 if ((flags & LLE_EXCLUSIVE) == 0) { 1959 flags |= LLE_EXCLUSIVE; 1960 LLE_RUNLOCK(ln); 1961 goto retry; 1962 } 1963 1964 LLE_WLOCK_ASSERT(ln); 1965 1966 if (ln->la_hold) { 1967 struct mbuf *m_hold; 1968 int i; 1969 1970 i = 0; 1971 for (m_hold = ln->la_hold; m_hold; m_hold = m_hold->m_nextpkt) { 1972 i++; 1973 if (m_hold->m_nextpkt == NULL) { 1974 m_hold->m_nextpkt = m; 1975 break; 1976 } 1977 } 1978 while (i >= V_nd6_maxqueuelen) { 1979 m_hold = ln->la_hold; 1980 ln->la_hold = ln->la_hold->m_nextpkt; 1981 m_freem(m_hold); 1982 i--; 1983 } 1984 } else { 1985 ln->la_hold = m; 1986 } 1987 1988 /* 1989 * If there has been no NS for the neighbor after entering the 1990 * INCOMPLETE state, send the first solicitation. 1991 */ 1992 if (!ND6_LLINFO_PERMANENT(ln) && ln->la_asked == 0) { 1993 ln->la_asked++; 1994 1995 nd6_llinfo_settimer_locked(ln, 1996 (long)ND_IFINFO(ifp)->retrans * hz / 1000); 1997 LLE_WUNLOCK(ln); 1998 nd6_ns_output(ifp, NULL, &dst->sin6_addr, ln, 0); 1999 if (lle != NULL && ln == lle) 2000 LLE_WLOCK(lle); 2001 2002 } else if (lle == NULL || ln != lle) { 2003 /* 2004 * We did the lookup (no lle arg) so we 2005 * need to do the unlock here. 2006 */ 2007 LLE_WUNLOCK(ln); 2008 } 2009 2010 return (0); 2011 2012 sendpkt: 2013 /* discard the packet if IPv6 operation is disabled on the interface */ 2014 if ((ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED)) { 2015 error = ENETDOWN; /* better error? */ 2016 goto bad; 2017 } 2018 /* 2019 * ln is valid and the caller did not pass in 2020 * an llentry 2021 */ 2022 if ((ln != NULL) && (lle == NULL)) { 2023 if (flags & LLE_EXCLUSIVE) 2024 LLE_WUNLOCK(ln); 2025 else 2026 LLE_RUNLOCK(ln); 2027 } 2028 2029 #ifdef MAC 2030 mac_netinet6_nd6_send(ifp, m); 2031 #endif 2032 2033 /* 2034 * If called from nd6_ns_output() (NS), nd6_na_output() (NA), 2035 * icmp6_redirect_output() (REDIRECT) or from rip6_output() (RS, RA 2036 * as handled by rtsol and rtadvd), mbufs will be tagged for SeND 2037 * to be diverted to user space. When re-injected into the kernel, 2038 * send_output() will directly dispatch them to the outgoing interface. 2039 */ 2040 if (send_sendso_input_hook != NULL) { 2041 mtag = m_tag_find(m, PACKET_TAG_ND_OUTGOING, NULL); 2042 if (mtag != NULL) { 2043 ip6 = mtod(m, struct ip6_hdr *); 2044 ip6len = sizeof(struct ip6_hdr) + ntohs(ip6->ip6_plen); 2045 /* Use the SEND socket */ 2046 error = send_sendso_input_hook(m, ifp, SND_OUT, 2047 ip6len); 2048 /* -1 == no app on SEND socket */ 2049 if (error == 0 || error != -1) 2050 return (error); 2051 } 2052 } 2053 2054 /* 2055 * We were passed in a pointer to an lle with the lock held 2056 * this means that we can't call if_output as we will 2057 * recurse on the lle lock - so what we do is we create 2058 * a list of mbufs to send and transmit them in the caller 2059 * after the lock is dropped 2060 */ 2061 if (lle != NULL) { 2062 if (*chain == NULL) 2063 *chain = m; 2064 else { 2065 struct mbuf *mb; 2066 2067 /* 2068 * append mbuf to end of deferred chain 2069 */ 2070 mb = *chain; 2071 while (mb->m_nextpkt != NULL) 2072 mb = mb->m_nextpkt; 2073 mb->m_nextpkt = m; 2074 } 2075 return (error); 2076 } 2077 if ((ifp->if_flags & IFF_LOOPBACK) != 0) { 2078 return ((*ifp->if_output)(origifp, m, (struct sockaddr *)dst, 2079 NULL)); 2080 } 2081 error = (*ifp->if_output)(ifp, m, (struct sockaddr *)dst, NULL); 2082 return (error); 2083 2084 bad: 2085 /* 2086 * ln is valid and the caller did not pass in 2087 * an llentry 2088 */ 2089 if ((ln != NULL) && (lle == NULL)) { 2090 if (flags & LLE_EXCLUSIVE) 2091 LLE_WUNLOCK(ln); 2092 else 2093 LLE_RUNLOCK(ln); 2094 } 2095 if (m) 2096 m_freem(m); 2097 return (error); 2098 } 2099 #undef senderr 2100 2101 2102 int 2103 nd6_output_flush(struct ifnet *ifp, struct ifnet *origifp, struct mbuf *chain, 2104 struct sockaddr_in6 *dst, struct route *ro) 2105 { 2106 struct mbuf *m, *m_head; 2107 struct ifnet *outifp; 2108 int error = 0; 2109 2110 m_head = chain; 2111 if ((ifp->if_flags & IFF_LOOPBACK) != 0) 2112 outifp = origifp; 2113 else 2114 outifp = ifp; 2115 2116 while (m_head) { 2117 m = m_head; 2118 m_head = m_head->m_nextpkt; 2119 error = (*ifp->if_output)(ifp, m, (struct sockaddr *)dst, ro); 2120 } 2121 2122 /* 2123 * XXX 2124 * note that intermediate errors are blindly ignored - but this is 2125 * the same convention as used with nd6_output when called by 2126 * nd6_cache_lladdr 2127 */ 2128 return (error); 2129 } 2130 2131 2132 int 2133 nd6_need_cache(struct ifnet *ifp) 2134 { 2135 /* 2136 * XXX: we currently do not make neighbor cache on any interface 2137 * other than ARCnet, Ethernet, FDDI and GIF. 2138 * 2139 * RFC2893 says: 2140 * - unidirectional tunnels needs no ND 2141 */ 2142 switch (ifp->if_type) { 2143 case IFT_ARCNET: 2144 case IFT_ETHER: 2145 case IFT_FDDI: 2146 case IFT_IEEE1394: 2147 #ifdef IFT_L2VLAN 2148 case IFT_L2VLAN: 2149 #endif 2150 #ifdef IFT_IEEE80211 2151 case IFT_IEEE80211: 2152 #endif 2153 case IFT_INFINIBAND: 2154 case IFT_GIF: /* XXX need more cases? */ 2155 case IFT_PPP: 2156 case IFT_TUNNEL: 2157 case IFT_BRIDGE: 2158 case IFT_PROPVIRTUAL: 2159 return (1); 2160 default: 2161 return (0); 2162 } 2163 } 2164 2165 /* 2166 * the callers of this function need to be re-worked to drop 2167 * the lle lock, drop here for now 2168 */ 2169 int 2170 nd6_storelladdr(struct ifnet *ifp, struct mbuf *m, 2171 struct sockaddr *dst, u_char *desten, struct llentry **lle) 2172 { 2173 struct llentry *ln; 2174 2175 *lle = NULL; 2176 IF_AFDATA_UNLOCK_ASSERT(ifp); 2177 if (m->m_flags & M_MCAST) { 2178 int i; 2179 2180 switch (ifp->if_type) { 2181 case IFT_ETHER: 2182 case IFT_FDDI: 2183 #ifdef IFT_L2VLAN 2184 case IFT_L2VLAN: 2185 #endif 2186 #ifdef IFT_IEEE80211 2187 case IFT_IEEE80211: 2188 #endif 2189 case IFT_BRIDGE: 2190 case IFT_ISO88025: 2191 ETHER_MAP_IPV6_MULTICAST(&SIN6(dst)->sin6_addr, 2192 desten); 2193 return (0); 2194 case IFT_IEEE1394: 2195 /* 2196 * netbsd can use if_broadcastaddr, but we don't do so 2197 * to reduce # of ifdef. 2198 */ 2199 for (i = 0; i < ifp->if_addrlen; i++) 2200 desten[i] = ~0; 2201 return (0); 2202 case IFT_ARCNET: 2203 *desten = 0; 2204 return (0); 2205 default: 2206 m_freem(m); 2207 return (EAFNOSUPPORT); 2208 } 2209 } 2210 2211 2212 /* 2213 * the entry should have been created in nd6_store_lladdr 2214 */ 2215 IF_AFDATA_LOCK(ifp); 2216 ln = lla_lookup(LLTABLE6(ifp), 0, dst); 2217 IF_AFDATA_UNLOCK(ifp); 2218 if ((ln == NULL) || !(ln->la_flags & LLE_VALID)) { 2219 if (ln != NULL) 2220 LLE_RUNLOCK(ln); 2221 /* this could happen, if we could not allocate memory */ 2222 m_freem(m); 2223 return (1); 2224 } 2225 2226 bcopy(&ln->ll_addr, desten, ifp->if_addrlen); 2227 *lle = ln; 2228 LLE_RUNLOCK(ln); 2229 /* 2230 * A *small* use after free race exists here 2231 */ 2232 return (0); 2233 } 2234 2235 static void 2236 clear_llinfo_pqueue(struct llentry *ln) 2237 { 2238 struct mbuf *m_hold, *m_hold_next; 2239 2240 for (m_hold = ln->la_hold; m_hold; m_hold = m_hold_next) { 2241 m_hold_next = m_hold->m_nextpkt; 2242 m_hold->m_nextpkt = NULL; 2243 m_freem(m_hold); 2244 } 2245 2246 ln->la_hold = NULL; 2247 return; 2248 } 2249 2250 static int nd6_sysctl_drlist(SYSCTL_HANDLER_ARGS); 2251 static int nd6_sysctl_prlist(SYSCTL_HANDLER_ARGS); 2252 #ifdef SYSCTL_DECL 2253 SYSCTL_DECL(_net_inet6_icmp6); 2254 #endif 2255 SYSCTL_NODE(_net_inet6_icmp6, ICMPV6CTL_ND6_DRLIST, nd6_drlist, 2256 CTLFLAG_RD, nd6_sysctl_drlist, ""); 2257 SYSCTL_NODE(_net_inet6_icmp6, ICMPV6CTL_ND6_PRLIST, nd6_prlist, 2258 CTLFLAG_RD, nd6_sysctl_prlist, ""); 2259 SYSCTL_VNET_INT(_net_inet6_icmp6, ICMPV6CTL_ND6_MAXQLEN, nd6_maxqueuelen, 2260 CTLFLAG_RW, &VNET_NAME(nd6_maxqueuelen), 1, ""); 2261 2262 static int 2263 nd6_sysctl_drlist(SYSCTL_HANDLER_ARGS) 2264 { 2265 int error; 2266 char buf[1024] __aligned(4); 2267 struct in6_defrouter *d, *de; 2268 struct nd_defrouter *dr; 2269 2270 if (req->newptr) 2271 return EPERM; 2272 error = 0; 2273 2274 TAILQ_FOREACH(dr, &V_nd_defrouter, dr_entry) { 2275 d = (struct in6_defrouter *)buf; 2276 de = (struct in6_defrouter *)(buf + sizeof(buf)); 2277 2278 if (d + 1 <= de) { 2279 bzero(d, sizeof(*d)); 2280 d->rtaddr.sin6_family = AF_INET6; 2281 d->rtaddr.sin6_len = sizeof(d->rtaddr); 2282 d->rtaddr.sin6_addr = dr->rtaddr; 2283 error = sa6_recoverscope(&d->rtaddr); 2284 if (error != 0) 2285 return (error); 2286 d->flags = dr->flags; 2287 d->rtlifetime = dr->rtlifetime; 2288 d->expire = dr->expire; 2289 d->if_index = dr->ifp->if_index; 2290 } else 2291 panic("buffer too short"); 2292 2293 error = SYSCTL_OUT(req, buf, sizeof(*d)); 2294 if (error) 2295 break; 2296 } 2297 2298 return (error); 2299 } 2300 2301 static int 2302 nd6_sysctl_prlist(SYSCTL_HANDLER_ARGS) 2303 { 2304 int error; 2305 char buf[1024] __aligned(4); 2306 struct in6_prefix *p, *pe; 2307 struct nd_prefix *pr; 2308 char ip6buf[INET6_ADDRSTRLEN]; 2309 2310 if (req->newptr) 2311 return EPERM; 2312 error = 0; 2313 2314 LIST_FOREACH(pr, &V_nd_prefix, ndpr_entry) { 2315 u_short advrtrs; 2316 size_t advance; 2317 struct sockaddr_in6 *sin6, *s6; 2318 struct nd_pfxrouter *pfr; 2319 2320 p = (struct in6_prefix *)buf; 2321 pe = (struct in6_prefix *)(buf + sizeof(buf)); 2322 2323 if (p + 1 <= pe) { 2324 bzero(p, sizeof(*p)); 2325 sin6 = (struct sockaddr_in6 *)(p + 1); 2326 2327 p->prefix = pr->ndpr_prefix; 2328 if (sa6_recoverscope(&p->prefix)) { 2329 log(LOG_ERR, 2330 "scope error in prefix list (%s)\n", 2331 ip6_sprintf(ip6buf, &p->prefix.sin6_addr)); 2332 /* XXX: press on... */ 2333 } 2334 p->raflags = pr->ndpr_raf; 2335 p->prefixlen = pr->ndpr_plen; 2336 p->vltime = pr->ndpr_vltime; 2337 p->pltime = pr->ndpr_pltime; 2338 p->if_index = pr->ndpr_ifp->if_index; 2339 if (pr->ndpr_vltime == ND6_INFINITE_LIFETIME) 2340 p->expire = 0; 2341 else { 2342 time_t maxexpire; 2343 2344 /* XXX: we assume time_t is signed. */ 2345 maxexpire = (-1) & 2346 ~((time_t)1 << 2347 ((sizeof(maxexpire) * 8) - 1)); 2348 if (pr->ndpr_vltime < 2349 maxexpire - pr->ndpr_lastupdate) { 2350 p->expire = pr->ndpr_lastupdate + 2351 pr->ndpr_vltime; 2352 } else 2353 p->expire = maxexpire; 2354 } 2355 p->refcnt = pr->ndpr_refcnt; 2356 p->flags = pr->ndpr_stateflags; 2357 p->origin = PR_ORIG_RA; 2358 advrtrs = 0; 2359 LIST_FOREACH(pfr, &pr->ndpr_advrtrs, pfr_entry) { 2360 if ((void *)&sin6[advrtrs + 1] > (void *)pe) { 2361 advrtrs++; 2362 continue; 2363 } 2364 s6 = &sin6[advrtrs]; 2365 bzero(s6, sizeof(*s6)); 2366 s6->sin6_family = AF_INET6; 2367 s6->sin6_len = sizeof(*sin6); 2368 s6->sin6_addr = pfr->router->rtaddr; 2369 if (sa6_recoverscope(s6)) { 2370 log(LOG_ERR, 2371 "scope error in " 2372 "prefix list (%s)\n", 2373 ip6_sprintf(ip6buf, 2374 &pfr->router->rtaddr)); 2375 } 2376 advrtrs++; 2377 } 2378 p->advrtrs = advrtrs; 2379 } else 2380 panic("buffer too short"); 2381 2382 advance = sizeof(*p) + sizeof(*sin6) * advrtrs; 2383 error = SYSCTL_OUT(req, buf, advance); 2384 if (error) 2385 break; 2386 } 2387 2388 return (error); 2389 } 2390