1 /*- 2 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project. 3 * All rights reserved. 4 * 5 * Redistribution and use in source and binary forms, with or without 6 * modification, are permitted provided that the following conditions 7 * are met: 8 * 1. Redistributions of source code must retain the above copyright 9 * notice, this list of conditions and the following disclaimer. 10 * 2. Redistributions in binary form must reproduce the above copyright 11 * notice, this list of conditions and the following disclaimer in the 12 * documentation and/or other materials provided with the distribution. 13 * 3. Neither the name of the project nor the names of its contributors 14 * may be used to endorse or promote products derived from this software 15 * without specific prior written permission. 16 * 17 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND 18 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 19 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 20 * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE 21 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 22 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 23 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 24 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 25 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 26 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 27 * SUCH DAMAGE. 28 * 29 * $KAME: nd6.c,v 1.144 2001/05/24 07:44:00 itojun Exp $ 30 */ 31 32 #include <sys/cdefs.h> 33 __FBSDID("$FreeBSD$"); 34 35 #include "opt_inet.h" 36 #include "opt_inet6.h" 37 38 #include <sys/param.h> 39 #include <sys/systm.h> 40 #include <sys/callout.h> 41 #include <sys/malloc.h> 42 #include <sys/mbuf.h> 43 #include <sys/socket.h> 44 #include <sys/sockio.h> 45 #include <sys/time.h> 46 #include <sys/kernel.h> 47 #include <sys/protosw.h> 48 #include <sys/errno.h> 49 #include <sys/syslog.h> 50 #include <sys/lock.h> 51 #include <sys/rwlock.h> 52 #include <sys/queue.h> 53 #include <sys/sysctl.h> 54 55 #include <net/if.h> 56 #include <net/if_arc.h> 57 #include <net/if_dl.h> 58 #include <net/if_types.h> 59 #include <net/iso88025.h> 60 #include <net/fddi.h> 61 #include <net/route.h> 62 #include <net/vnet.h> 63 64 #include <netinet/in.h> 65 #include <net/if_llatbl.h> 66 #define L3_ADDR_SIN6(le) ((struct sockaddr_in6 *) L3_ADDR(le)) 67 #include <netinet/if_ether.h> 68 #include <netinet6/in6_var.h> 69 #include <netinet/ip6.h> 70 #include <netinet6/ip6_var.h> 71 #include <netinet6/scope6_var.h> 72 #include <netinet6/nd6.h> 73 #include <netinet6/in6_ifattach.h> 74 #include <netinet/icmp6.h> 75 76 #include <sys/limits.h> 77 78 #include <security/mac/mac_framework.h> 79 80 #define ND6_SLOWTIMER_INTERVAL (60 * 60) /* 1 hour */ 81 #define ND6_RECALC_REACHTM_INTERVAL (60 * 120) /* 2 hours */ 82 83 #define SIN6(s) ((struct sockaddr_in6 *)s) 84 85 VNET_DEFINE(int, nd6_prune); 86 VNET_DEFINE(int, nd6_delay); 87 VNET_DEFINE(int, nd6_umaxtries); 88 VNET_DEFINE(int, nd6_mmaxtries); 89 VNET_DEFINE(int, nd6_useloopback); 90 VNET_DEFINE(int, nd6_gctimer); 91 92 /* preventing too many loops in ND option parsing */ 93 static VNET_DEFINE(int, nd6_maxndopt); 94 VNET_DEFINE(int, nd6_maxnudhint); 95 static VNET_DEFINE(int, nd6_maxqueuelen); 96 #define V_nd6_maxndopt VNET(nd6_maxndopt) 97 #define V_nd6_maxqueuelen VNET(nd6_maxqueuelen) 98 99 VNET_DEFINE(int, nd6_debug); 100 101 /* for debugging? */ 102 #if 0 103 static int nd6_inuse, nd6_allocated; 104 #endif 105 106 VNET_DEFINE(struct nd_drhead, nd_defrouter); 107 VNET_DEFINE(struct nd_prhead, nd_prefix); 108 109 VNET_DEFINE(int, nd6_recalc_reachtm_interval); 110 #define V_nd6_recalc_reachtm_interval VNET(nd6_recalc_reachtm_interval) 111 112 static struct sockaddr_in6 all1_sa; 113 114 static int nd6_is_new_addr_neighbor __P((struct sockaddr_in6 *, 115 struct ifnet *)); 116 static void nd6_setmtu0(struct ifnet *, struct nd_ifinfo *); 117 static void nd6_slowtimo(void *); 118 static int regen_tmpaddr(struct in6_ifaddr *); 119 static struct llentry *nd6_free(struct llentry *, int); 120 static void nd6_llinfo_timer(void *); 121 static void clear_llinfo_pqueue(struct llentry *); 122 123 static VNET_DEFINE(struct callout, nd6_slowtimo_ch); 124 #define V_nd6_slowtimo_ch VNET(nd6_slowtimo_ch) 125 126 VNET_DEFINE(struct callout, nd6_timer_ch); 127 128 VNET_DECLARE(int, dad_ignore_ns); 129 VNET_DECLARE(int, dad_maxtry); 130 #define V_dad_ignore_ns VNET(dad_ignore_ns) 131 #define V_dad_maxtry VNET(dad_maxtry) 132 133 void 134 nd6_init(void) 135 { 136 int i; 137 138 V_nd6_prune = 1; /* walk list every 1 seconds */ 139 V_nd6_delay = 5; /* delay first probe time 5 second */ 140 V_nd6_umaxtries = 3; /* maximum unicast query */ 141 V_nd6_mmaxtries = 3; /* maximum multicast query */ 142 V_nd6_useloopback = 1; /* use loopback interface for local traffic */ 143 V_nd6_gctimer = (60 * 60 * 24); /* 1 day: garbage collection timer */ 144 145 /* preventing too many loops in ND option parsing */ 146 V_nd6_maxndopt = 10; /* max # of ND options allowed */ 147 148 V_nd6_maxnudhint = 0; /* max # of subsequent upper layer hints */ 149 V_nd6_maxqueuelen = 1; /* max pkts cached in unresolved ND entries */ 150 151 #ifdef ND6_DEBUG 152 V_nd6_debug = 1; 153 #else 154 V_nd6_debug = 0; 155 #endif 156 157 V_nd6_recalc_reachtm_interval = ND6_RECALC_REACHTM_INTERVAL; 158 159 V_dad_ignore_ns = 0; /* ignore NS in DAD - specwise incorrect*/ 160 V_dad_maxtry = 15; /* max # of *tries* to transmit DAD packet */ 161 162 /* 163 * XXX just to get this to compile KMM 164 */ 165 #ifdef notyet 166 V_llinfo_nd6.ln_next = &V_llinfo_nd6; 167 V_llinfo_nd6.ln_prev = &V_llinfo_nd6; 168 #endif 169 LIST_INIT(&V_nd_prefix); 170 171 V_ip6_use_tempaddr = 0; 172 V_ip6_temp_preferred_lifetime = DEF_TEMP_PREFERRED_LIFETIME; 173 V_ip6_temp_valid_lifetime = DEF_TEMP_VALID_LIFETIME; 174 V_ip6_temp_regen_advance = TEMPADDR_REGEN_ADVANCE; 175 176 V_ip6_desync_factor = 0; 177 178 all1_sa.sin6_family = AF_INET6; 179 all1_sa.sin6_len = sizeof(struct sockaddr_in6); 180 for (i = 0; i < sizeof(all1_sa.sin6_addr); i++) 181 all1_sa.sin6_addr.s6_addr[i] = 0xff; 182 183 /* initialization of the default router list */ 184 TAILQ_INIT(&V_nd_defrouter); 185 /* start timer */ 186 callout_init(&V_nd6_slowtimo_ch, 0); 187 callout_reset(&V_nd6_slowtimo_ch, ND6_SLOWTIMER_INTERVAL * hz, 188 nd6_slowtimo, curvnet); 189 } 190 191 192 #ifdef VIMAGE 193 void 194 nd6_destroy() 195 { 196 197 callout_drain(&V_nd6_slowtimo_ch); 198 callout_drain(&V_nd6_timer_ch); 199 } 200 #endif 201 202 struct nd_ifinfo * 203 nd6_ifattach(struct ifnet *ifp) 204 { 205 struct nd_ifinfo *nd; 206 207 nd = (struct nd_ifinfo *)malloc(sizeof(*nd), M_IP6NDP, M_WAITOK); 208 bzero(nd, sizeof(*nd)); 209 210 nd->initialized = 1; 211 212 nd->chlim = IPV6_DEFHLIM; 213 nd->basereachable = REACHABLE_TIME; 214 nd->reachable = ND_COMPUTE_RTIME(nd->basereachable); 215 nd->retrans = RETRANS_TIMER; 216 217 nd->flags = ND6_IFF_PERFORMNUD; 218 219 /* A loopback interface always has ND6_IFF_AUTO_LINKLOCAL. */ 220 if (V_ip6_auto_linklocal || (ifp->if_flags & IFF_LOOPBACK)) 221 nd->flags |= ND6_IFF_AUTO_LINKLOCAL; 222 223 /* A loopback interface does not need to accept RTADV. */ 224 if (V_ip6_accept_rtadv && !(ifp->if_flags & IFF_LOOPBACK)) 225 nd->flags |= ND6_IFF_ACCEPT_RTADV; 226 227 /* XXX: we cannot call nd6_setmtu since ifp is not fully initialized */ 228 nd6_setmtu0(ifp, nd); 229 230 return nd; 231 } 232 233 void 234 nd6_ifdetach(struct nd_ifinfo *nd) 235 { 236 237 free(nd, M_IP6NDP); 238 } 239 240 /* 241 * Reset ND level link MTU. This function is called when the physical MTU 242 * changes, which means we might have to adjust the ND level MTU. 243 */ 244 void 245 nd6_setmtu(struct ifnet *ifp) 246 { 247 248 nd6_setmtu0(ifp, ND_IFINFO(ifp)); 249 } 250 251 /* XXX todo: do not maintain copy of ifp->if_mtu in ndi->maxmtu */ 252 void 253 nd6_setmtu0(struct ifnet *ifp, struct nd_ifinfo *ndi) 254 { 255 u_int32_t omaxmtu; 256 257 omaxmtu = ndi->maxmtu; 258 259 switch (ifp->if_type) { 260 case IFT_ARCNET: 261 ndi->maxmtu = MIN(ARC_PHDS_MAXMTU, ifp->if_mtu); /* RFC2497 */ 262 break; 263 case IFT_FDDI: 264 ndi->maxmtu = MIN(FDDIIPMTU, ifp->if_mtu); /* RFC2467 */ 265 break; 266 case IFT_ISO88025: 267 ndi->maxmtu = MIN(ISO88025_MAX_MTU, ifp->if_mtu); 268 break; 269 default: 270 ndi->maxmtu = ifp->if_mtu; 271 break; 272 } 273 274 /* 275 * Decreasing the interface MTU under IPV6 minimum MTU may cause 276 * undesirable situation. We thus notify the operator of the change 277 * explicitly. The check for omaxmtu is necessary to restrict the 278 * log to the case of changing the MTU, not initializing it. 279 */ 280 if (omaxmtu >= IPV6_MMTU && ndi->maxmtu < IPV6_MMTU) { 281 log(LOG_NOTICE, "nd6_setmtu0: " 282 "new link MTU on %s (%lu) is too small for IPv6\n", 283 if_name(ifp), (unsigned long)ndi->maxmtu); 284 } 285 286 if (ndi->maxmtu > V_in6_maxmtu) 287 in6_setmaxmtu(); /* check all interfaces just in case */ 288 289 } 290 291 void 292 nd6_option_init(void *opt, int icmp6len, union nd_opts *ndopts) 293 { 294 295 bzero(ndopts, sizeof(*ndopts)); 296 ndopts->nd_opts_search = (struct nd_opt_hdr *)opt; 297 ndopts->nd_opts_last 298 = (struct nd_opt_hdr *)(((u_char *)opt) + icmp6len); 299 300 if (icmp6len == 0) { 301 ndopts->nd_opts_done = 1; 302 ndopts->nd_opts_search = NULL; 303 } 304 } 305 306 /* 307 * Take one ND option. 308 */ 309 struct nd_opt_hdr * 310 nd6_option(union nd_opts *ndopts) 311 { 312 struct nd_opt_hdr *nd_opt; 313 int olen; 314 315 if (ndopts == NULL) 316 panic("ndopts == NULL in nd6_option"); 317 if (ndopts->nd_opts_last == NULL) 318 panic("uninitialized ndopts in nd6_option"); 319 if (ndopts->nd_opts_search == NULL) 320 return NULL; 321 if (ndopts->nd_opts_done) 322 return NULL; 323 324 nd_opt = ndopts->nd_opts_search; 325 326 /* make sure nd_opt_len is inside the buffer */ 327 if ((caddr_t)&nd_opt->nd_opt_len >= (caddr_t)ndopts->nd_opts_last) { 328 bzero(ndopts, sizeof(*ndopts)); 329 return NULL; 330 } 331 332 olen = nd_opt->nd_opt_len << 3; 333 if (olen == 0) { 334 /* 335 * Message validation requires that all included 336 * options have a length that is greater than zero. 337 */ 338 bzero(ndopts, sizeof(*ndopts)); 339 return NULL; 340 } 341 342 ndopts->nd_opts_search = (struct nd_opt_hdr *)((caddr_t)nd_opt + olen); 343 if (ndopts->nd_opts_search > ndopts->nd_opts_last) { 344 /* option overruns the end of buffer, invalid */ 345 bzero(ndopts, sizeof(*ndopts)); 346 return NULL; 347 } else if (ndopts->nd_opts_search == ndopts->nd_opts_last) { 348 /* reached the end of options chain */ 349 ndopts->nd_opts_done = 1; 350 ndopts->nd_opts_search = NULL; 351 } 352 return nd_opt; 353 } 354 355 /* 356 * Parse multiple ND options. 357 * This function is much easier to use, for ND routines that do not need 358 * multiple options of the same type. 359 */ 360 int 361 nd6_options(union nd_opts *ndopts) 362 { 363 struct nd_opt_hdr *nd_opt; 364 int i = 0; 365 366 if (ndopts == NULL) 367 panic("ndopts == NULL in nd6_options"); 368 if (ndopts->nd_opts_last == NULL) 369 panic("uninitialized ndopts in nd6_options"); 370 if (ndopts->nd_opts_search == NULL) 371 return 0; 372 373 while (1) { 374 nd_opt = nd6_option(ndopts); 375 if (nd_opt == NULL && ndopts->nd_opts_last == NULL) { 376 /* 377 * Message validation requires that all included 378 * options have a length that is greater than zero. 379 */ 380 ICMP6STAT_INC(icp6s_nd_badopt); 381 bzero(ndopts, sizeof(*ndopts)); 382 return -1; 383 } 384 385 if (nd_opt == NULL) 386 goto skip1; 387 388 switch (nd_opt->nd_opt_type) { 389 case ND_OPT_SOURCE_LINKADDR: 390 case ND_OPT_TARGET_LINKADDR: 391 case ND_OPT_MTU: 392 case ND_OPT_REDIRECTED_HEADER: 393 if (ndopts->nd_opt_array[nd_opt->nd_opt_type]) { 394 nd6log((LOG_INFO, 395 "duplicated ND6 option found (type=%d)\n", 396 nd_opt->nd_opt_type)); 397 /* XXX bark? */ 398 } else { 399 ndopts->nd_opt_array[nd_opt->nd_opt_type] 400 = nd_opt; 401 } 402 break; 403 case ND_OPT_PREFIX_INFORMATION: 404 if (ndopts->nd_opt_array[nd_opt->nd_opt_type] == 0) { 405 ndopts->nd_opt_array[nd_opt->nd_opt_type] 406 = nd_opt; 407 } 408 ndopts->nd_opts_pi_end = 409 (struct nd_opt_prefix_info *)nd_opt; 410 break; 411 default: 412 /* 413 * Unknown options must be silently ignored, 414 * to accomodate future extension to the protocol. 415 */ 416 nd6log((LOG_DEBUG, 417 "nd6_options: unsupported option %d - " 418 "option ignored\n", nd_opt->nd_opt_type)); 419 } 420 421 skip1: 422 i++; 423 if (i > V_nd6_maxndopt) { 424 ICMP6STAT_INC(icp6s_nd_toomanyopt); 425 nd6log((LOG_INFO, "too many loop in nd opt\n")); 426 break; 427 } 428 429 if (ndopts->nd_opts_done) 430 break; 431 } 432 433 return 0; 434 } 435 436 /* 437 * ND6 timer routine to handle ND6 entries 438 */ 439 void 440 nd6_llinfo_settimer_locked(struct llentry *ln, long tick) 441 { 442 int canceled; 443 444 if (tick < 0) { 445 ln->la_expire = 0; 446 ln->ln_ntick = 0; 447 canceled = callout_stop(&ln->ln_timer_ch); 448 } else { 449 ln->la_expire = time_second + tick / hz; 450 LLE_ADDREF(ln); 451 if (tick > INT_MAX) { 452 ln->ln_ntick = tick - INT_MAX; 453 canceled = callout_reset(&ln->ln_timer_ch, INT_MAX, 454 nd6_llinfo_timer, ln); 455 } else { 456 ln->ln_ntick = 0; 457 canceled = callout_reset(&ln->ln_timer_ch, tick, 458 nd6_llinfo_timer, ln); 459 } 460 } 461 if (canceled) 462 LLE_REMREF(ln); 463 } 464 465 void 466 nd6_llinfo_settimer(struct llentry *ln, long tick) 467 { 468 469 LLE_WLOCK(ln); 470 nd6_llinfo_settimer_locked(ln, tick); 471 LLE_WUNLOCK(ln); 472 } 473 474 static void 475 nd6_llinfo_timer(void *arg) 476 { 477 struct llentry *ln; 478 struct in6_addr *dst; 479 struct ifnet *ifp; 480 struct nd_ifinfo *ndi = NULL; 481 482 ln = (struct llentry *)arg; 483 if (ln == NULL) { 484 panic("%s: NULL entry!\n", __func__); 485 return; 486 } 487 488 if ((ifp = ((ln->lle_tbl != NULL) ? ln->lle_tbl->llt_ifp : NULL)) == NULL) 489 panic("ln ifp == NULL"); 490 491 CURVNET_SET(ifp->if_vnet); 492 493 if (ln->ln_ntick > 0) { 494 if (ln->ln_ntick > INT_MAX) { 495 ln->ln_ntick -= INT_MAX; 496 nd6_llinfo_settimer(ln, INT_MAX); 497 } else { 498 ln->ln_ntick = 0; 499 nd6_llinfo_settimer(ln, ln->ln_ntick); 500 } 501 goto done; 502 } 503 504 ndi = ND_IFINFO(ifp); 505 dst = &L3_ADDR_SIN6(ln)->sin6_addr; 506 if (ln->la_flags & LLE_STATIC) { 507 goto done; 508 } 509 510 if (ln->la_flags & LLE_DELETED) { 511 (void)nd6_free(ln, 0); 512 ln = NULL; 513 goto done; 514 } 515 516 switch (ln->ln_state) { 517 case ND6_LLINFO_INCOMPLETE: 518 if (ln->la_asked < V_nd6_mmaxtries) { 519 ln->la_asked++; 520 nd6_llinfo_settimer(ln, (long)ndi->retrans * hz / 1000); 521 nd6_ns_output(ifp, NULL, dst, ln, 0); 522 } else { 523 struct mbuf *m = ln->la_hold; 524 if (m) { 525 struct mbuf *m0; 526 527 /* 528 * assuming every packet in la_hold has the 529 * same IP header 530 */ 531 m0 = m->m_nextpkt; 532 m->m_nextpkt = NULL; 533 icmp6_error2(m, ICMP6_DST_UNREACH, 534 ICMP6_DST_UNREACH_ADDR, 0, ifp); 535 536 ln->la_hold = m0; 537 clear_llinfo_pqueue(ln); 538 } 539 (void)nd6_free(ln, 0); 540 ln = NULL; 541 } 542 break; 543 case ND6_LLINFO_REACHABLE: 544 if (!ND6_LLINFO_PERMANENT(ln)) { 545 ln->ln_state = ND6_LLINFO_STALE; 546 nd6_llinfo_settimer(ln, (long)V_nd6_gctimer * hz); 547 } 548 break; 549 550 case ND6_LLINFO_STALE: 551 /* Garbage Collection(RFC 2461 5.3) */ 552 if (!ND6_LLINFO_PERMANENT(ln)) { 553 (void)nd6_free(ln, 1); 554 ln = NULL; 555 } 556 break; 557 558 case ND6_LLINFO_DELAY: 559 if (ndi && (ndi->flags & ND6_IFF_PERFORMNUD) != 0) { 560 /* We need NUD */ 561 ln->la_asked = 1; 562 ln->ln_state = ND6_LLINFO_PROBE; 563 nd6_llinfo_settimer(ln, (long)ndi->retrans * hz / 1000); 564 nd6_ns_output(ifp, dst, dst, ln, 0); 565 } else { 566 ln->ln_state = ND6_LLINFO_STALE; /* XXX */ 567 nd6_llinfo_settimer(ln, (long)V_nd6_gctimer * hz); 568 } 569 break; 570 case ND6_LLINFO_PROBE: 571 if (ln->la_asked < V_nd6_umaxtries) { 572 ln->la_asked++; 573 nd6_llinfo_settimer(ln, (long)ndi->retrans * hz / 1000); 574 nd6_ns_output(ifp, dst, dst, ln, 0); 575 } else { 576 (void)nd6_free(ln, 0); 577 ln = NULL; 578 } 579 break; 580 } 581 done: 582 if (ln != NULL) 583 LLE_FREE(ln); 584 CURVNET_RESTORE(); 585 } 586 587 588 /* 589 * ND6 timer routine to expire default route list and prefix list 590 */ 591 void 592 nd6_timer(void *arg) 593 { 594 CURVNET_SET((struct vnet *) arg); 595 int s; 596 struct nd_defrouter *dr; 597 struct nd_prefix *pr; 598 struct in6_ifaddr *ia6, *nia6; 599 struct in6_addrlifetime *lt6; 600 601 callout_reset(&V_nd6_timer_ch, V_nd6_prune * hz, 602 nd6_timer, curvnet); 603 604 /* expire default router list */ 605 s = splnet(); 606 dr = TAILQ_FIRST(&V_nd_defrouter); 607 while (dr) { 608 if (dr->expire && dr->expire < time_second) { 609 struct nd_defrouter *t; 610 t = TAILQ_NEXT(dr, dr_entry); 611 defrtrlist_del(dr); 612 dr = t; 613 } else { 614 dr = TAILQ_NEXT(dr, dr_entry); 615 } 616 } 617 618 /* 619 * expire interface addresses. 620 * in the past the loop was inside prefix expiry processing. 621 * However, from a stricter speci-confrmance standpoint, we should 622 * rather separate address lifetimes and prefix lifetimes. 623 * 624 * XXXRW: in6_ifaddrhead locking. 625 */ 626 addrloop: 627 TAILQ_FOREACH_SAFE(ia6, &V_in6_ifaddrhead, ia_link, nia6) { 628 /* check address lifetime */ 629 lt6 = &ia6->ia6_lifetime; 630 if (IFA6_IS_INVALID(ia6)) { 631 int regen = 0; 632 633 /* 634 * If the expiring address is temporary, try 635 * regenerating a new one. This would be useful when 636 * we suspended a laptop PC, then turned it on after a 637 * period that could invalidate all temporary 638 * addresses. Although we may have to restart the 639 * loop (see below), it must be after purging the 640 * address. Otherwise, we'd see an infinite loop of 641 * regeneration. 642 */ 643 if (V_ip6_use_tempaddr && 644 (ia6->ia6_flags & IN6_IFF_TEMPORARY) != 0) { 645 if (regen_tmpaddr(ia6) == 0) 646 regen = 1; 647 } 648 649 in6_purgeaddr(&ia6->ia_ifa); 650 651 if (regen) 652 goto addrloop; /* XXX: see below */ 653 } else if (IFA6_IS_DEPRECATED(ia6)) { 654 int oldflags = ia6->ia6_flags; 655 656 ia6->ia6_flags |= IN6_IFF_DEPRECATED; 657 658 /* 659 * If a temporary address has just become deprecated, 660 * regenerate a new one if possible. 661 */ 662 if (V_ip6_use_tempaddr && 663 (ia6->ia6_flags & IN6_IFF_TEMPORARY) != 0 && 664 (oldflags & IN6_IFF_DEPRECATED) == 0) { 665 666 if (regen_tmpaddr(ia6) == 0) { 667 /* 668 * A new temporary address is 669 * generated. 670 * XXX: this means the address chain 671 * has changed while we are still in 672 * the loop. Although the change 673 * would not cause disaster (because 674 * it's not a deletion, but an 675 * addition,) we'd rather restart the 676 * loop just for safety. Or does this 677 * significantly reduce performance?? 678 */ 679 goto addrloop; 680 } 681 } 682 } else { 683 /* 684 * A new RA might have made a deprecated address 685 * preferred. 686 */ 687 ia6->ia6_flags &= ~IN6_IFF_DEPRECATED; 688 } 689 } 690 691 /* expire prefix list */ 692 pr = V_nd_prefix.lh_first; 693 while (pr) { 694 /* 695 * check prefix lifetime. 696 * since pltime is just for autoconf, pltime processing for 697 * prefix is not necessary. 698 */ 699 if (pr->ndpr_vltime != ND6_INFINITE_LIFETIME && 700 time_second - pr->ndpr_lastupdate > pr->ndpr_vltime) { 701 struct nd_prefix *t; 702 t = pr->ndpr_next; 703 704 /* 705 * address expiration and prefix expiration are 706 * separate. NEVER perform in6_purgeaddr here. 707 */ 708 709 prelist_remove(pr); 710 pr = t; 711 } else 712 pr = pr->ndpr_next; 713 } 714 splx(s); 715 CURVNET_RESTORE(); 716 } 717 718 /* 719 * ia6 - deprecated/invalidated temporary address 720 */ 721 static int 722 regen_tmpaddr(struct in6_ifaddr *ia6) 723 { 724 struct ifaddr *ifa; 725 struct ifnet *ifp; 726 struct in6_ifaddr *public_ifa6 = NULL; 727 728 ifp = ia6->ia_ifa.ifa_ifp; 729 IF_ADDR_LOCK(ifp); 730 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 731 struct in6_ifaddr *it6; 732 733 if (ifa->ifa_addr->sa_family != AF_INET6) 734 continue; 735 736 it6 = (struct in6_ifaddr *)ifa; 737 738 /* ignore no autoconf addresses. */ 739 if ((it6->ia6_flags & IN6_IFF_AUTOCONF) == 0) 740 continue; 741 742 /* ignore autoconf addresses with different prefixes. */ 743 if (it6->ia6_ndpr == NULL || it6->ia6_ndpr != ia6->ia6_ndpr) 744 continue; 745 746 /* 747 * Now we are looking at an autoconf address with the same 748 * prefix as ours. If the address is temporary and is still 749 * preferred, do not create another one. It would be rare, but 750 * could happen, for example, when we resume a laptop PC after 751 * a long period. 752 */ 753 if ((it6->ia6_flags & IN6_IFF_TEMPORARY) != 0 && 754 !IFA6_IS_DEPRECATED(it6)) { 755 public_ifa6 = NULL; 756 break; 757 } 758 759 /* 760 * This is a public autoconf address that has the same prefix 761 * as ours. If it is preferred, keep it. We can't break the 762 * loop here, because there may be a still-preferred temporary 763 * address with the prefix. 764 */ 765 if (!IFA6_IS_DEPRECATED(it6)) 766 public_ifa6 = it6; 767 768 if (public_ifa6 != NULL) 769 ifa_ref(&public_ifa6->ia_ifa); 770 } 771 IF_ADDR_UNLOCK(ifp); 772 773 if (public_ifa6 != NULL) { 774 int e; 775 776 if ((e = in6_tmpifadd(public_ifa6, 0, 0)) != 0) { 777 ifa_free(&public_ifa6->ia_ifa); 778 log(LOG_NOTICE, "regen_tmpaddr: failed to create a new" 779 " tmp addr,errno=%d\n", e); 780 return (-1); 781 } 782 ifa_free(&public_ifa6->ia_ifa); 783 return (0); 784 } 785 786 return (-1); 787 } 788 789 /* 790 * Nuke neighbor cache/prefix/default router management table, right before 791 * ifp goes away. 792 */ 793 void 794 nd6_purge(struct ifnet *ifp) 795 { 796 struct nd_defrouter *dr, *ndr; 797 struct nd_prefix *pr, *npr; 798 799 /* 800 * Nuke default router list entries toward ifp. 801 * We defer removal of default router list entries that is installed 802 * in the routing table, in order to keep additional side effects as 803 * small as possible. 804 */ 805 for (dr = TAILQ_FIRST(&V_nd_defrouter); dr; dr = ndr) { 806 ndr = TAILQ_NEXT(dr, dr_entry); 807 if (dr->installed) 808 continue; 809 810 if (dr->ifp == ifp) 811 defrtrlist_del(dr); 812 } 813 814 for (dr = TAILQ_FIRST(&V_nd_defrouter); dr; dr = ndr) { 815 ndr = TAILQ_NEXT(dr, dr_entry); 816 if (!dr->installed) 817 continue; 818 819 if (dr->ifp == ifp) 820 defrtrlist_del(dr); 821 } 822 823 /* Nuke prefix list entries toward ifp */ 824 for (pr = V_nd_prefix.lh_first; pr; pr = npr) { 825 npr = pr->ndpr_next; 826 if (pr->ndpr_ifp == ifp) { 827 /* 828 * Because if_detach() does *not* release prefixes 829 * while purging addresses the reference count will 830 * still be above zero. We therefore reset it to 831 * make sure that the prefix really gets purged. 832 */ 833 pr->ndpr_refcnt = 0; 834 835 /* 836 * Previously, pr->ndpr_addr is removed as well, 837 * but I strongly believe we don't have to do it. 838 * nd6_purge() is only called from in6_ifdetach(), 839 * which removes all the associated interface addresses 840 * by itself. 841 * (jinmei@kame.net 20010129) 842 */ 843 prelist_remove(pr); 844 } 845 } 846 847 /* cancel default outgoing interface setting */ 848 if (V_nd6_defifindex == ifp->if_index) 849 nd6_setdefaultiface(0); 850 851 if (!V_ip6_forwarding && ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV) { 852 /* Refresh default router list. */ 853 defrouter_select(); 854 } 855 856 /* XXXXX 857 * We do not nuke the neighbor cache entries here any more 858 * because the neighbor cache is kept in if_afdata[AF_INET6]. 859 * nd6_purge() is invoked by in6_ifdetach() which is called 860 * from if_detach() where everything gets purged. So let 861 * in6_domifdetach() do the actual L2 table purging work. 862 */ 863 } 864 865 /* 866 * the caller acquires and releases the lock on the lltbls 867 * Returns the llentry locked 868 */ 869 struct llentry * 870 nd6_lookup(struct in6_addr *addr6, int flags, struct ifnet *ifp) 871 { 872 struct sockaddr_in6 sin6; 873 struct llentry *ln; 874 int llflags = 0; 875 876 bzero(&sin6, sizeof(sin6)); 877 sin6.sin6_len = sizeof(struct sockaddr_in6); 878 sin6.sin6_family = AF_INET6; 879 sin6.sin6_addr = *addr6; 880 881 IF_AFDATA_LOCK_ASSERT(ifp); 882 883 if (flags & ND6_CREATE) 884 llflags |= LLE_CREATE; 885 if (flags & ND6_EXCLUSIVE) 886 llflags |= LLE_EXCLUSIVE; 887 888 ln = lla_lookup(LLTABLE6(ifp), llflags, (struct sockaddr *)&sin6); 889 if ((ln != NULL) && (flags & LLE_CREATE)) { 890 ln->ln_state = ND6_LLINFO_NOSTATE; 891 callout_init(&ln->ln_timer_ch, 0); 892 } 893 894 return (ln); 895 } 896 897 /* 898 * Test whether a given IPv6 address is a neighbor or not, ignoring 899 * the actual neighbor cache. The neighbor cache is ignored in order 900 * to not reenter the routing code from within itself. 901 */ 902 static int 903 nd6_is_new_addr_neighbor(struct sockaddr_in6 *addr, struct ifnet *ifp) 904 { 905 struct nd_prefix *pr; 906 struct ifaddr *dstaddr; 907 908 /* 909 * A link-local address is always a neighbor. 910 * XXX: a link does not necessarily specify a single interface. 911 */ 912 if (IN6_IS_ADDR_LINKLOCAL(&addr->sin6_addr)) { 913 struct sockaddr_in6 sin6_copy; 914 u_int32_t zone; 915 916 /* 917 * We need sin6_copy since sa6_recoverscope() may modify the 918 * content (XXX). 919 */ 920 sin6_copy = *addr; 921 if (sa6_recoverscope(&sin6_copy)) 922 return (0); /* XXX: should be impossible */ 923 if (in6_setscope(&sin6_copy.sin6_addr, ifp, &zone)) 924 return (0); 925 if (sin6_copy.sin6_scope_id == zone) 926 return (1); 927 else 928 return (0); 929 } 930 931 /* 932 * If the address matches one of our addresses, 933 * it should be a neighbor. 934 * If the address matches one of our on-link prefixes, it should be a 935 * neighbor. 936 */ 937 for (pr = V_nd_prefix.lh_first; pr; pr = pr->ndpr_next) { 938 if (pr->ndpr_ifp != ifp) 939 continue; 940 941 if (!(pr->ndpr_stateflags & NDPRF_ONLINK)) { 942 struct rtentry *rt; 943 rt = rtalloc1((struct sockaddr *)&pr->ndpr_prefix, 0, 0); 944 if (rt == NULL) 945 continue; 946 /* 947 * This is the case where multiple interfaces 948 * have the same prefix, but only one is installed 949 * into the routing table and that prefix entry 950 * is not the one being examined here. In the case 951 * where RADIX_MPATH is enabled, multiple route 952 * entries (of the same rt_key value) will be 953 * installed because the interface addresses all 954 * differ. 955 */ 956 if (!IN6_ARE_ADDR_EQUAL(&pr->ndpr_prefix.sin6_addr, 957 &((struct sockaddr_in6 *)rt_key(rt))->sin6_addr)) { 958 RTFREE_LOCKED(rt); 959 continue; 960 } 961 RTFREE_LOCKED(rt); 962 } 963 964 if (IN6_ARE_MASKED_ADDR_EQUAL(&pr->ndpr_prefix.sin6_addr, 965 &addr->sin6_addr, &pr->ndpr_mask)) 966 return (1); 967 } 968 969 /* 970 * If the address is assigned on the node of the other side of 971 * a p2p interface, the address should be a neighbor. 972 */ 973 dstaddr = ifa_ifwithdstaddr((struct sockaddr *)addr); 974 if (dstaddr != NULL) { 975 if (dstaddr->ifa_ifp == ifp) { 976 ifa_free(dstaddr); 977 return (1); 978 } 979 ifa_free(dstaddr); 980 } 981 982 /* 983 * If the default router list is empty, all addresses are regarded 984 * as on-link, and thus, as a neighbor. 985 * XXX: we restrict the condition to hosts, because routers usually do 986 * not have the "default router list". 987 */ 988 if (!V_ip6_forwarding && TAILQ_FIRST(&V_nd_defrouter) == NULL && 989 V_nd6_defifindex == ifp->if_index) { 990 return (1); 991 } 992 993 return (0); 994 } 995 996 997 /* 998 * Detect if a given IPv6 address identifies a neighbor on a given link. 999 * XXX: should take care of the destination of a p2p link? 1000 */ 1001 int 1002 nd6_is_addr_neighbor(struct sockaddr_in6 *addr, struct ifnet *ifp) 1003 { 1004 struct llentry *lle; 1005 int rc = 0; 1006 1007 IF_AFDATA_UNLOCK_ASSERT(ifp); 1008 if (nd6_is_new_addr_neighbor(addr, ifp)) 1009 return (1); 1010 1011 /* 1012 * Even if the address matches none of our addresses, it might be 1013 * in the neighbor cache. 1014 */ 1015 IF_AFDATA_LOCK(ifp); 1016 if ((lle = nd6_lookup(&addr->sin6_addr, 0, ifp)) != NULL) { 1017 LLE_RUNLOCK(lle); 1018 rc = 1; 1019 } 1020 IF_AFDATA_UNLOCK(ifp); 1021 return (rc); 1022 } 1023 1024 /* 1025 * Free an nd6 llinfo entry. 1026 * Since the function would cause significant changes in the kernel, DO NOT 1027 * make it global, unless you have a strong reason for the change, and are sure 1028 * that the change is safe. 1029 */ 1030 static struct llentry * 1031 nd6_free(struct llentry *ln, int gc) 1032 { 1033 struct llentry *next; 1034 struct nd_defrouter *dr; 1035 struct ifnet *ifp=NULL; 1036 1037 /* 1038 * we used to have pfctlinput(PRC_HOSTDEAD) here. 1039 * even though it is not harmful, it was not really necessary. 1040 */ 1041 1042 /* cancel timer */ 1043 nd6_llinfo_settimer(ln, -1); 1044 1045 if (!V_ip6_forwarding) { 1046 int s; 1047 s = splnet(); 1048 dr = defrouter_lookup(&L3_ADDR_SIN6(ln)->sin6_addr, ln->lle_tbl->llt_ifp); 1049 1050 if (dr != NULL && dr->expire && 1051 ln->ln_state == ND6_LLINFO_STALE && gc) { 1052 /* 1053 * If the reason for the deletion is just garbage 1054 * collection, and the neighbor is an active default 1055 * router, do not delete it. Instead, reset the GC 1056 * timer using the router's lifetime. 1057 * Simply deleting the entry would affect default 1058 * router selection, which is not necessarily a good 1059 * thing, especially when we're using router preference 1060 * values. 1061 * XXX: the check for ln_state would be redundant, 1062 * but we intentionally keep it just in case. 1063 */ 1064 if (dr->expire > time_second) 1065 nd6_llinfo_settimer(ln, 1066 (dr->expire - time_second) * hz); 1067 else 1068 nd6_llinfo_settimer(ln, (long)V_nd6_gctimer * hz); 1069 splx(s); 1070 LLE_WLOCK(ln); 1071 LLE_REMREF(ln); 1072 LLE_WUNLOCK(ln); 1073 return (LIST_NEXT(ln, lle_next)); 1074 } 1075 1076 if (ln->ln_router || dr) { 1077 /* 1078 * rt6_flush must be called whether or not the neighbor 1079 * is in the Default Router List. 1080 * See a corresponding comment in nd6_na_input(). 1081 */ 1082 rt6_flush(&L3_ADDR_SIN6(ln)->sin6_addr, ln->lle_tbl->llt_ifp); 1083 } 1084 1085 if (dr) { 1086 /* 1087 * Unreachablity of a router might affect the default 1088 * router selection and on-link detection of advertised 1089 * prefixes. 1090 */ 1091 1092 /* 1093 * Temporarily fake the state to choose a new default 1094 * router and to perform on-link determination of 1095 * prefixes correctly. 1096 * Below the state will be set correctly, 1097 * or the entry itself will be deleted. 1098 */ 1099 ln->ln_state = ND6_LLINFO_INCOMPLETE; 1100 1101 /* 1102 * Since defrouter_select() does not affect the 1103 * on-link determination and MIP6 needs the check 1104 * before the default router selection, we perform 1105 * the check now. 1106 */ 1107 pfxlist_onlink_check(); 1108 1109 /* 1110 * refresh default router list 1111 */ 1112 defrouter_select(); 1113 } 1114 splx(s); 1115 } 1116 1117 /* 1118 * Before deleting the entry, remember the next entry as the 1119 * return value. We need this because pfxlist_onlink_check() above 1120 * might have freed other entries (particularly the old next entry) as 1121 * a side effect (XXX). 1122 */ 1123 next = LIST_NEXT(ln, lle_next); 1124 1125 ifp = ln->lle_tbl->llt_ifp; 1126 IF_AFDATA_LOCK(ifp); 1127 LLE_WLOCK(ln); 1128 LLE_REMREF(ln); 1129 llentry_free(ln); 1130 IF_AFDATA_UNLOCK(ifp); 1131 1132 return (next); 1133 } 1134 1135 /* 1136 * Upper-layer reachability hint for Neighbor Unreachability Detection. 1137 * 1138 * XXX cost-effective methods? 1139 */ 1140 void 1141 nd6_nud_hint(struct rtentry *rt, struct in6_addr *dst6, int force) 1142 { 1143 struct llentry *ln; 1144 struct ifnet *ifp; 1145 1146 if ((dst6 == NULL) || (rt == NULL)) 1147 return; 1148 1149 ifp = rt->rt_ifp; 1150 IF_AFDATA_LOCK(ifp); 1151 ln = nd6_lookup(dst6, ND6_EXCLUSIVE, NULL); 1152 IF_AFDATA_UNLOCK(ifp); 1153 if (ln == NULL) 1154 return; 1155 1156 if (ln->ln_state < ND6_LLINFO_REACHABLE) 1157 goto done; 1158 1159 /* 1160 * if we get upper-layer reachability confirmation many times, 1161 * it is possible we have false information. 1162 */ 1163 if (!force) { 1164 ln->ln_byhint++; 1165 if (ln->ln_byhint > V_nd6_maxnudhint) { 1166 goto done; 1167 } 1168 } 1169 1170 ln->ln_state = ND6_LLINFO_REACHABLE; 1171 if (!ND6_LLINFO_PERMANENT(ln)) { 1172 nd6_llinfo_settimer_locked(ln, 1173 (long)ND_IFINFO(rt->rt_ifp)->reachable * hz); 1174 } 1175 done: 1176 LLE_WUNLOCK(ln); 1177 } 1178 1179 1180 int 1181 nd6_ioctl(u_long cmd, caddr_t data, struct ifnet *ifp) 1182 { 1183 struct in6_drlist *drl = (struct in6_drlist *)data; 1184 struct in6_oprlist *oprl = (struct in6_oprlist *)data; 1185 struct in6_ndireq *ndi = (struct in6_ndireq *)data; 1186 struct in6_nbrinfo *nbi = (struct in6_nbrinfo *)data; 1187 struct in6_ndifreq *ndif = (struct in6_ndifreq *)data; 1188 struct nd_defrouter *dr; 1189 struct nd_prefix *pr; 1190 int i = 0, error = 0; 1191 int s; 1192 1193 switch (cmd) { 1194 case SIOCGDRLST_IN6: 1195 /* 1196 * obsolete API, use sysctl under net.inet6.icmp6 1197 */ 1198 bzero(drl, sizeof(*drl)); 1199 s = splnet(); 1200 dr = TAILQ_FIRST(&V_nd_defrouter); 1201 while (dr && i < DRLSTSIZ) { 1202 drl->defrouter[i].rtaddr = dr->rtaddr; 1203 in6_clearscope(&drl->defrouter[i].rtaddr); 1204 1205 drl->defrouter[i].flags = dr->flags; 1206 drl->defrouter[i].rtlifetime = dr->rtlifetime; 1207 drl->defrouter[i].expire = dr->expire; 1208 drl->defrouter[i].if_index = dr->ifp->if_index; 1209 i++; 1210 dr = TAILQ_NEXT(dr, dr_entry); 1211 } 1212 splx(s); 1213 break; 1214 case SIOCGPRLST_IN6: 1215 /* 1216 * obsolete API, use sysctl under net.inet6.icmp6 1217 * 1218 * XXX the structure in6_prlist was changed in backward- 1219 * incompatible manner. in6_oprlist is used for SIOCGPRLST_IN6, 1220 * in6_prlist is used for nd6_sysctl() - fill_prlist(). 1221 */ 1222 /* 1223 * XXX meaning of fields, especialy "raflags", is very 1224 * differnet between RA prefix list and RR/static prefix list. 1225 * how about separating ioctls into two? 1226 */ 1227 bzero(oprl, sizeof(*oprl)); 1228 s = splnet(); 1229 pr = V_nd_prefix.lh_first; 1230 while (pr && i < PRLSTSIZ) { 1231 struct nd_pfxrouter *pfr; 1232 int j; 1233 1234 oprl->prefix[i].prefix = pr->ndpr_prefix.sin6_addr; 1235 oprl->prefix[i].raflags = pr->ndpr_raf; 1236 oprl->prefix[i].prefixlen = pr->ndpr_plen; 1237 oprl->prefix[i].vltime = pr->ndpr_vltime; 1238 oprl->prefix[i].pltime = pr->ndpr_pltime; 1239 oprl->prefix[i].if_index = pr->ndpr_ifp->if_index; 1240 if (pr->ndpr_vltime == ND6_INFINITE_LIFETIME) 1241 oprl->prefix[i].expire = 0; 1242 else { 1243 time_t maxexpire; 1244 1245 /* XXX: we assume time_t is signed. */ 1246 maxexpire = (-1) & 1247 ~((time_t)1 << 1248 ((sizeof(maxexpire) * 8) - 1)); 1249 if (pr->ndpr_vltime < 1250 maxexpire - pr->ndpr_lastupdate) { 1251 oprl->prefix[i].expire = 1252 pr->ndpr_lastupdate + 1253 pr->ndpr_vltime; 1254 } else 1255 oprl->prefix[i].expire = maxexpire; 1256 } 1257 1258 pfr = pr->ndpr_advrtrs.lh_first; 1259 j = 0; 1260 while (pfr) { 1261 if (j < DRLSTSIZ) { 1262 #define RTRADDR oprl->prefix[i].advrtr[j] 1263 RTRADDR = pfr->router->rtaddr; 1264 in6_clearscope(&RTRADDR); 1265 #undef RTRADDR 1266 } 1267 j++; 1268 pfr = pfr->pfr_next; 1269 } 1270 oprl->prefix[i].advrtrs = j; 1271 oprl->prefix[i].origin = PR_ORIG_RA; 1272 1273 i++; 1274 pr = pr->ndpr_next; 1275 } 1276 splx(s); 1277 1278 break; 1279 case OSIOCGIFINFO_IN6: 1280 #define ND ndi->ndi 1281 /* XXX: old ndp(8) assumes a positive value for linkmtu. */ 1282 bzero(&ND, sizeof(ND)); 1283 ND.linkmtu = IN6_LINKMTU(ifp); 1284 ND.maxmtu = ND_IFINFO(ifp)->maxmtu; 1285 ND.basereachable = ND_IFINFO(ifp)->basereachable; 1286 ND.reachable = ND_IFINFO(ifp)->reachable; 1287 ND.retrans = ND_IFINFO(ifp)->retrans; 1288 ND.flags = ND_IFINFO(ifp)->flags; 1289 ND.recalctm = ND_IFINFO(ifp)->recalctm; 1290 ND.chlim = ND_IFINFO(ifp)->chlim; 1291 break; 1292 case SIOCGIFINFO_IN6: 1293 ND = *ND_IFINFO(ifp); 1294 break; 1295 case SIOCSIFINFO_IN6: 1296 /* 1297 * used to change host variables from userland. 1298 * intented for a use on router to reflect RA configurations. 1299 */ 1300 /* 0 means 'unspecified' */ 1301 if (ND.linkmtu != 0) { 1302 if (ND.linkmtu < IPV6_MMTU || 1303 ND.linkmtu > IN6_LINKMTU(ifp)) { 1304 error = EINVAL; 1305 break; 1306 } 1307 ND_IFINFO(ifp)->linkmtu = ND.linkmtu; 1308 } 1309 1310 if (ND.basereachable != 0) { 1311 int obasereachable = ND_IFINFO(ifp)->basereachable; 1312 1313 ND_IFINFO(ifp)->basereachable = ND.basereachable; 1314 if (ND.basereachable != obasereachable) 1315 ND_IFINFO(ifp)->reachable = 1316 ND_COMPUTE_RTIME(ND.basereachable); 1317 } 1318 if (ND.retrans != 0) 1319 ND_IFINFO(ifp)->retrans = ND.retrans; 1320 if (ND.chlim != 0) 1321 ND_IFINFO(ifp)->chlim = ND.chlim; 1322 /* FALLTHROUGH */ 1323 case SIOCSIFINFO_FLAGS: 1324 { 1325 struct ifaddr *ifa; 1326 struct in6_ifaddr *ia; 1327 1328 if ((ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) && 1329 !(ND.flags & ND6_IFF_IFDISABLED)) { 1330 /* ifdisabled 1->0 transision */ 1331 1332 /* 1333 * If the interface is marked as ND6_IFF_IFDISABLED and 1334 * has an link-local address with IN6_IFF_DUPLICATED, 1335 * do not clear ND6_IFF_IFDISABLED. 1336 * See RFC 4862, Section 5.4.5. 1337 */ 1338 int duplicated_linklocal = 0; 1339 1340 IF_ADDR_LOCK(ifp); 1341 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1342 if (ifa->ifa_addr->sa_family != AF_INET6) 1343 continue; 1344 ia = (struct in6_ifaddr *)ifa; 1345 if ((ia->ia6_flags & IN6_IFF_DUPLICATED) && 1346 IN6_IS_ADDR_LINKLOCAL(&ia->ia_addr.sin6_addr)) { 1347 duplicated_linklocal = 1; 1348 break; 1349 } 1350 } 1351 IF_ADDR_UNLOCK(ifp); 1352 1353 if (duplicated_linklocal) { 1354 ND.flags |= ND6_IFF_IFDISABLED; 1355 log(LOG_ERR, "Cannot enable an interface" 1356 " with a link-local address marked" 1357 " duplicate.\n"); 1358 } else { 1359 ND_IFINFO(ifp)->flags &= ~ND6_IFF_IFDISABLED; 1360 in6_if_up(ifp); 1361 } 1362 } else if (!(ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) && 1363 (ND.flags & ND6_IFF_IFDISABLED)) { 1364 /* ifdisabled 0->1 transision */ 1365 /* Mark all IPv6 address as tentative. */ 1366 1367 ND_IFINFO(ifp)->flags |= ND6_IFF_IFDISABLED; 1368 IF_ADDR_LOCK(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 ia->ia6_flags |= IN6_IFF_TENTATIVE; 1374 } 1375 IF_ADDR_UNLOCK(ifp); 1376 } 1377 1378 if (!(ND_IFINFO(ifp)->flags & ND6_IFF_AUTO_LINKLOCAL) && 1379 (ND.flags & ND6_IFF_AUTO_LINKLOCAL)) { 1380 /* auto_linklocal 0->1 transision */ 1381 1382 /* If no link-local address on ifp, configure */ 1383 ND_IFINFO(ifp)->flags |= ND6_IFF_AUTO_LINKLOCAL; 1384 in6_ifattach(ifp, NULL); 1385 } 1386 } 1387 ND_IFINFO(ifp)->flags = ND.flags; 1388 break; 1389 #undef ND 1390 case SIOCSNDFLUSH_IN6: /* XXX: the ioctl name is confusing... */ 1391 /* sync kernel routing table with the default router list */ 1392 defrouter_reset(); 1393 defrouter_select(); 1394 break; 1395 case SIOCSPFXFLUSH_IN6: 1396 { 1397 /* flush all the prefix advertised by routers */ 1398 struct nd_prefix *pr, *next; 1399 1400 s = splnet(); 1401 for (pr = V_nd_prefix.lh_first; pr; pr = next) { 1402 struct in6_ifaddr *ia, *ia_next; 1403 1404 next = pr->ndpr_next; 1405 1406 if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr)) 1407 continue; /* XXX */ 1408 1409 /* do we really have to remove addresses as well? */ 1410 /* XXXRW: in6_ifaddrhead locking. */ 1411 TAILQ_FOREACH_SAFE(ia, &V_in6_ifaddrhead, ia_link, 1412 ia_next) { 1413 if ((ia->ia6_flags & IN6_IFF_AUTOCONF) == 0) 1414 continue; 1415 1416 if (ia->ia6_ndpr == pr) 1417 in6_purgeaddr(&ia->ia_ifa); 1418 } 1419 prelist_remove(pr); 1420 } 1421 splx(s); 1422 break; 1423 } 1424 case SIOCSRTRFLUSH_IN6: 1425 { 1426 /* flush all the default routers */ 1427 struct nd_defrouter *dr, *next; 1428 1429 s = splnet(); 1430 defrouter_reset(); 1431 for (dr = TAILQ_FIRST(&V_nd_defrouter); dr; dr = next) { 1432 next = TAILQ_NEXT(dr, dr_entry); 1433 defrtrlist_del(dr); 1434 } 1435 defrouter_select(); 1436 splx(s); 1437 break; 1438 } 1439 case SIOCGNBRINFO_IN6: 1440 { 1441 struct llentry *ln; 1442 struct in6_addr nb_addr = nbi->addr; /* make local for safety */ 1443 1444 if ((error = in6_setscope(&nb_addr, ifp, NULL)) != 0) 1445 return (error); 1446 1447 IF_AFDATA_LOCK(ifp); 1448 ln = nd6_lookup(&nb_addr, 0, ifp); 1449 IF_AFDATA_UNLOCK(ifp); 1450 1451 if (ln == NULL) { 1452 error = EINVAL; 1453 break; 1454 } 1455 nbi->state = ln->ln_state; 1456 nbi->asked = ln->la_asked; 1457 nbi->isrouter = ln->ln_router; 1458 nbi->expire = ln->la_expire; 1459 LLE_RUNLOCK(ln); 1460 break; 1461 } 1462 case SIOCGDEFIFACE_IN6: /* XXX: should be implemented as a sysctl? */ 1463 ndif->ifindex = V_nd6_defifindex; 1464 break; 1465 case SIOCSDEFIFACE_IN6: /* XXX: should be implemented as a sysctl? */ 1466 return (nd6_setdefaultiface(ndif->ifindex)); 1467 } 1468 return (error); 1469 } 1470 1471 /* 1472 * Create neighbor cache entry and cache link-layer address, 1473 * on reception of inbound ND6 packets. (RS/RA/NS/redirect) 1474 * 1475 * type - ICMP6 type 1476 * code - type dependent information 1477 * 1478 * XXXXX 1479 * The caller of this function already acquired the ndp 1480 * cache table lock because the cache entry is returned. 1481 */ 1482 struct llentry * 1483 nd6_cache_lladdr(struct ifnet *ifp, struct in6_addr *from, char *lladdr, 1484 int lladdrlen, int type, int code) 1485 { 1486 struct llentry *ln = NULL; 1487 int is_newentry; 1488 int do_update; 1489 int olladdr; 1490 int llchange; 1491 int flags = 0; 1492 int newstate = 0; 1493 uint16_t router = 0; 1494 struct sockaddr_in6 sin6; 1495 struct mbuf *chain = NULL; 1496 int static_route = 0; 1497 1498 IF_AFDATA_UNLOCK_ASSERT(ifp); 1499 1500 if (ifp == NULL) 1501 panic("ifp == NULL in nd6_cache_lladdr"); 1502 if (from == NULL) 1503 panic("from == NULL in nd6_cache_lladdr"); 1504 1505 /* nothing must be updated for unspecified address */ 1506 if (IN6_IS_ADDR_UNSPECIFIED(from)) 1507 return NULL; 1508 1509 /* 1510 * Validation about ifp->if_addrlen and lladdrlen must be done in 1511 * the caller. 1512 * 1513 * XXX If the link does not have link-layer adderss, what should 1514 * we do? (ifp->if_addrlen == 0) 1515 * Spec says nothing in sections for RA, RS and NA. There's small 1516 * description on it in NS section (RFC 2461 7.2.3). 1517 */ 1518 flags |= lladdr ? ND6_EXCLUSIVE : 0; 1519 IF_AFDATA_LOCK(ifp); 1520 ln = nd6_lookup(from, flags, ifp); 1521 1522 if (ln == NULL) { 1523 flags |= LLE_EXCLUSIVE; 1524 ln = nd6_lookup(from, flags |ND6_CREATE, ifp); 1525 IF_AFDATA_UNLOCK(ifp); 1526 is_newentry = 1; 1527 } else { 1528 IF_AFDATA_UNLOCK(ifp); 1529 /* do nothing if static ndp is set */ 1530 if (ln->la_flags & LLE_STATIC) { 1531 static_route = 1; 1532 goto done; 1533 } 1534 is_newentry = 0; 1535 } 1536 if (ln == NULL) 1537 return (NULL); 1538 1539 olladdr = (ln->la_flags & LLE_VALID) ? 1 : 0; 1540 if (olladdr && lladdr) { 1541 llchange = bcmp(lladdr, &ln->ll_addr, 1542 ifp->if_addrlen); 1543 } else 1544 llchange = 0; 1545 1546 /* 1547 * newentry olladdr lladdr llchange (*=record) 1548 * 0 n n -- (1) 1549 * 0 y n -- (2) 1550 * 0 n y -- (3) * STALE 1551 * 0 y y n (4) * 1552 * 0 y y y (5) * STALE 1553 * 1 -- n -- (6) NOSTATE(= PASSIVE) 1554 * 1 -- y -- (7) * STALE 1555 */ 1556 1557 if (lladdr) { /* (3-5) and (7) */ 1558 /* 1559 * Record source link-layer address 1560 * XXX is it dependent to ifp->if_type? 1561 */ 1562 bcopy(lladdr, &ln->ll_addr, ifp->if_addrlen); 1563 ln->la_flags |= LLE_VALID; 1564 } 1565 1566 if (!is_newentry) { 1567 if ((!olladdr && lladdr != NULL) || /* (3) */ 1568 (olladdr && lladdr != NULL && llchange)) { /* (5) */ 1569 do_update = 1; 1570 newstate = ND6_LLINFO_STALE; 1571 } else /* (1-2,4) */ 1572 do_update = 0; 1573 } else { 1574 do_update = 1; 1575 if (lladdr == NULL) /* (6) */ 1576 newstate = ND6_LLINFO_NOSTATE; 1577 else /* (7) */ 1578 newstate = ND6_LLINFO_STALE; 1579 } 1580 1581 if (do_update) { 1582 /* 1583 * Update the state of the neighbor cache. 1584 */ 1585 ln->ln_state = newstate; 1586 1587 if (ln->ln_state == ND6_LLINFO_STALE) { 1588 /* 1589 * XXX: since nd6_output() below will cause 1590 * state tansition to DELAY and reset the timer, 1591 * we must set the timer now, although it is actually 1592 * meaningless. 1593 */ 1594 nd6_llinfo_settimer_locked(ln, (long)V_nd6_gctimer * hz); 1595 1596 if (ln->la_hold) { 1597 struct mbuf *m_hold, *m_hold_next; 1598 1599 /* 1600 * reset the la_hold in advance, to explicitly 1601 * prevent a la_hold lookup in nd6_output() 1602 * (wouldn't happen, though...) 1603 */ 1604 for (m_hold = ln->la_hold, ln->la_hold = NULL; 1605 m_hold; m_hold = m_hold_next) { 1606 m_hold_next = m_hold->m_nextpkt; 1607 m_hold->m_nextpkt = NULL; 1608 1609 /* 1610 * we assume ifp is not a p2p here, so 1611 * just set the 2nd argument as the 1612 * 1st one. 1613 */ 1614 nd6_output_lle(ifp, ifp, m_hold, L3_ADDR_SIN6(ln), NULL, ln, &chain); 1615 } 1616 /* 1617 * If we have mbufs in the chain we need to do 1618 * deferred transmit. Copy the address from the 1619 * llentry before dropping the lock down below. 1620 */ 1621 if (chain != NULL) 1622 memcpy(&sin6, L3_ADDR_SIN6(ln), sizeof(sin6)); 1623 } 1624 } else if (ln->ln_state == ND6_LLINFO_INCOMPLETE) { 1625 /* probe right away */ 1626 nd6_llinfo_settimer_locked((void *)ln, 0); 1627 } 1628 } 1629 1630 /* 1631 * ICMP6 type dependent behavior. 1632 * 1633 * NS: clear IsRouter if new entry 1634 * RS: clear IsRouter 1635 * RA: set IsRouter if there's lladdr 1636 * redir: clear IsRouter if new entry 1637 * 1638 * RA case, (1): 1639 * The spec says that we must set IsRouter in the following cases: 1640 * - If lladdr exist, set IsRouter. This means (1-5). 1641 * - If it is old entry (!newentry), set IsRouter. This means (7). 1642 * So, based on the spec, in (1-5) and (7) cases we must set IsRouter. 1643 * A quetion arises for (1) case. (1) case has no lladdr in the 1644 * neighbor cache, this is similar to (6). 1645 * This case is rare but we figured that we MUST NOT set IsRouter. 1646 * 1647 * newentry olladdr lladdr llchange NS RS RA redir 1648 * D R 1649 * 0 n n -- (1) c ? s 1650 * 0 y n -- (2) c s s 1651 * 0 n y -- (3) c s s 1652 * 0 y y n (4) c s s 1653 * 0 y y y (5) c s s 1654 * 1 -- n -- (6) c c c s 1655 * 1 -- y -- (7) c c s c s 1656 * 1657 * (c=clear s=set) 1658 */ 1659 switch (type & 0xff) { 1660 case ND_NEIGHBOR_SOLICIT: 1661 /* 1662 * New entry must have is_router flag cleared. 1663 */ 1664 if (is_newentry) /* (6-7) */ 1665 ln->ln_router = 0; 1666 break; 1667 case ND_REDIRECT: 1668 /* 1669 * If the icmp is a redirect to a better router, always set the 1670 * is_router flag. Otherwise, if the entry is newly created, 1671 * clear the flag. [RFC 2461, sec 8.3] 1672 */ 1673 if (code == ND_REDIRECT_ROUTER) 1674 ln->ln_router = 1; 1675 else if (is_newentry) /* (6-7) */ 1676 ln->ln_router = 0; 1677 break; 1678 case ND_ROUTER_SOLICIT: 1679 /* 1680 * is_router flag must always be cleared. 1681 */ 1682 ln->ln_router = 0; 1683 break; 1684 case ND_ROUTER_ADVERT: 1685 /* 1686 * Mark an entry with lladdr as a router. 1687 */ 1688 if ((!is_newentry && (olladdr || lladdr)) || /* (2-5) */ 1689 (is_newentry && lladdr)) { /* (7) */ 1690 ln->ln_router = 1; 1691 } 1692 break; 1693 } 1694 1695 if (ln != NULL) { 1696 static_route = (ln->la_flags & LLE_STATIC); 1697 router = ln->ln_router; 1698 1699 if (flags & ND6_EXCLUSIVE) 1700 LLE_WUNLOCK(ln); 1701 else 1702 LLE_RUNLOCK(ln); 1703 if (static_route) 1704 ln = NULL; 1705 } 1706 if (chain) 1707 nd6_output_flush(ifp, ifp, chain, &sin6, NULL); 1708 1709 /* 1710 * When the link-layer address of a router changes, select the 1711 * best router again. In particular, when the neighbor entry is newly 1712 * created, it might affect the selection policy. 1713 * Question: can we restrict the first condition to the "is_newentry" 1714 * case? 1715 * XXX: when we hear an RA from a new router with the link-layer 1716 * address option, defrouter_select() is called twice, since 1717 * defrtrlist_update called the function as well. However, I believe 1718 * we can compromise the overhead, since it only happens the first 1719 * time. 1720 * XXX: although defrouter_select() should not have a bad effect 1721 * for those are not autoconfigured hosts, we explicitly avoid such 1722 * cases for safety. 1723 */ 1724 if (do_update && router && !V_ip6_forwarding && 1725 ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV) { 1726 /* 1727 * guaranteed recursion 1728 */ 1729 defrouter_select(); 1730 } 1731 1732 return (ln); 1733 done: 1734 if (ln != NULL) { 1735 if (flags & ND6_EXCLUSIVE) 1736 LLE_WUNLOCK(ln); 1737 else 1738 LLE_RUNLOCK(ln); 1739 if (static_route) 1740 ln = NULL; 1741 } 1742 return (ln); 1743 } 1744 1745 static void 1746 nd6_slowtimo(void *arg) 1747 { 1748 CURVNET_SET((struct vnet *) arg); 1749 struct nd_ifinfo *nd6if; 1750 struct ifnet *ifp; 1751 1752 callout_reset(&V_nd6_slowtimo_ch, ND6_SLOWTIMER_INTERVAL * hz, 1753 nd6_slowtimo, curvnet); 1754 IFNET_RLOCK_NOSLEEP(); 1755 for (ifp = TAILQ_FIRST(&V_ifnet); ifp; 1756 ifp = TAILQ_NEXT(ifp, if_list)) { 1757 nd6if = ND_IFINFO(ifp); 1758 if (nd6if->basereachable && /* already initialized */ 1759 (nd6if->recalctm -= ND6_SLOWTIMER_INTERVAL) <= 0) { 1760 /* 1761 * Since reachable time rarely changes by router 1762 * advertisements, we SHOULD insure that a new random 1763 * value gets recomputed at least once every few hours. 1764 * (RFC 2461, 6.3.4) 1765 */ 1766 nd6if->recalctm = V_nd6_recalc_reachtm_interval; 1767 nd6if->reachable = ND_COMPUTE_RTIME(nd6if->basereachable); 1768 } 1769 } 1770 IFNET_RUNLOCK_NOSLEEP(); 1771 CURVNET_RESTORE(); 1772 } 1773 1774 int 1775 nd6_output(struct ifnet *ifp, struct ifnet *origifp, struct mbuf *m0, 1776 struct sockaddr_in6 *dst, struct rtentry *rt0) 1777 { 1778 1779 return (nd6_output_lle(ifp, origifp, m0, dst, rt0, NULL, NULL)); 1780 } 1781 1782 1783 /* 1784 * Note that I'm not enforcing any global serialization 1785 * lle state or asked changes here as the logic is too 1786 * complicated to avoid having to always acquire an exclusive 1787 * lock 1788 * KMM 1789 * 1790 */ 1791 #define senderr(e) { error = (e); goto bad;} 1792 1793 int 1794 nd6_output_lle(struct ifnet *ifp, struct ifnet *origifp, struct mbuf *m0, 1795 struct sockaddr_in6 *dst, struct rtentry *rt0, struct llentry *lle, 1796 struct mbuf **chain) 1797 { 1798 struct mbuf *m = m0; 1799 struct llentry *ln = lle; 1800 int error = 0; 1801 int flags = 0; 1802 1803 #ifdef INVARIANTS 1804 if (lle != NULL) { 1805 1806 LLE_WLOCK_ASSERT(lle); 1807 1808 KASSERT(chain != NULL, (" lle locked but no mbuf chain pointer passed")); 1809 } 1810 #endif 1811 if (IN6_IS_ADDR_MULTICAST(&dst->sin6_addr)) 1812 goto sendpkt; 1813 1814 if (nd6_need_cache(ifp) == 0) 1815 goto sendpkt; 1816 1817 /* 1818 * next hop determination. This routine is derived from ether_output. 1819 */ 1820 1821 /* 1822 * Address resolution or Neighbor Unreachability Detection 1823 * for the next hop. 1824 * At this point, the destination of the packet must be a unicast 1825 * or an anycast address(i.e. not a multicast). 1826 */ 1827 1828 flags = ((m != NULL) || (lle != NULL)) ? LLE_EXCLUSIVE : 0; 1829 if (ln == NULL) { 1830 retry: 1831 IF_AFDATA_LOCK(ifp); 1832 ln = lla_lookup(LLTABLE6(ifp), flags, (struct sockaddr *)dst); 1833 IF_AFDATA_UNLOCK(ifp); 1834 if ((ln == NULL) && nd6_is_addr_neighbor(dst, ifp)) { 1835 /* 1836 * Since nd6_is_addr_neighbor() internally calls nd6_lookup(), 1837 * the condition below is not very efficient. But we believe 1838 * it is tolerable, because this should be a rare case. 1839 */ 1840 flags = ND6_CREATE | (m ? ND6_EXCLUSIVE : 0); 1841 IF_AFDATA_LOCK(ifp); 1842 ln = nd6_lookup(&dst->sin6_addr, flags, ifp); 1843 IF_AFDATA_UNLOCK(ifp); 1844 } 1845 } 1846 if (ln == NULL) { 1847 if ((ifp->if_flags & IFF_POINTOPOINT) == 0 && 1848 !(ND_IFINFO(ifp)->flags & ND6_IFF_PERFORMNUD)) { 1849 char ip6buf[INET6_ADDRSTRLEN]; 1850 log(LOG_DEBUG, 1851 "nd6_output: can't allocate llinfo for %s " 1852 "(ln=%p)\n", 1853 ip6_sprintf(ip6buf, &dst->sin6_addr), ln); 1854 senderr(EIO); /* XXX: good error? */ 1855 } 1856 goto sendpkt; /* send anyway */ 1857 } 1858 1859 /* We don't have to do link-layer address resolution on a p2p link. */ 1860 if ((ifp->if_flags & IFF_POINTOPOINT) != 0 && 1861 ln->ln_state < ND6_LLINFO_REACHABLE) { 1862 if ((flags & LLE_EXCLUSIVE) == 0) { 1863 flags |= LLE_EXCLUSIVE; 1864 goto retry; 1865 } 1866 ln->ln_state = ND6_LLINFO_STALE; 1867 nd6_llinfo_settimer_locked(ln, (long)V_nd6_gctimer * hz); 1868 } 1869 1870 /* 1871 * The first time we send a packet to a neighbor whose entry is 1872 * STALE, we have to change the state to DELAY and a sets a timer to 1873 * expire in DELAY_FIRST_PROBE_TIME seconds to ensure do 1874 * neighbor unreachability detection on expiration. 1875 * (RFC 2461 7.3.3) 1876 */ 1877 if (ln->ln_state == ND6_LLINFO_STALE) { 1878 if ((flags & LLE_EXCLUSIVE) == 0) { 1879 flags |= LLE_EXCLUSIVE; 1880 LLE_RUNLOCK(ln); 1881 goto retry; 1882 } 1883 ln->la_asked = 0; 1884 ln->ln_state = ND6_LLINFO_DELAY; 1885 nd6_llinfo_settimer_locked(ln, (long)V_nd6_delay * hz); 1886 } 1887 1888 /* 1889 * If the neighbor cache entry has a state other than INCOMPLETE 1890 * (i.e. its link-layer address is already resolved), just 1891 * send the packet. 1892 */ 1893 if (ln->ln_state > ND6_LLINFO_INCOMPLETE) 1894 goto sendpkt; 1895 1896 /* 1897 * There is a neighbor cache entry, but no ethernet address 1898 * response yet. Append this latest packet to the end of the 1899 * packet queue in the mbuf, unless the number of the packet 1900 * does not exceed nd6_maxqueuelen. When it exceeds nd6_maxqueuelen, 1901 * the oldest packet in the queue will be removed. 1902 */ 1903 if (ln->ln_state == ND6_LLINFO_NOSTATE) 1904 ln->ln_state = ND6_LLINFO_INCOMPLETE; 1905 1906 if ((flags & LLE_EXCLUSIVE) == 0) { 1907 flags |= LLE_EXCLUSIVE; 1908 LLE_RUNLOCK(ln); 1909 goto retry; 1910 } 1911 if (ln->la_hold) { 1912 struct mbuf *m_hold; 1913 int i; 1914 1915 i = 0; 1916 for (m_hold = ln->la_hold; m_hold; m_hold = m_hold->m_nextpkt) { 1917 i++; 1918 if (m_hold->m_nextpkt == NULL) { 1919 m_hold->m_nextpkt = m; 1920 break; 1921 } 1922 } 1923 while (i >= V_nd6_maxqueuelen) { 1924 m_hold = ln->la_hold; 1925 ln->la_hold = ln->la_hold->m_nextpkt; 1926 m_freem(m_hold); 1927 i--; 1928 } 1929 } else { 1930 ln->la_hold = m; 1931 } 1932 /* 1933 * We did the lookup (no lle arg) so we 1934 * need to do the unlock here 1935 */ 1936 if (lle == NULL) { 1937 if (flags & LLE_EXCLUSIVE) 1938 LLE_WUNLOCK(ln); 1939 else 1940 LLE_RUNLOCK(ln); 1941 } 1942 1943 /* 1944 * If there has been no NS for the neighbor after entering the 1945 * INCOMPLETE state, send the first solicitation. 1946 */ 1947 if (!ND6_LLINFO_PERMANENT(ln) && ln->la_asked == 0) { 1948 ln->la_asked++; 1949 1950 nd6_llinfo_settimer(ln, 1951 (long)ND_IFINFO(ifp)->retrans * hz / 1000); 1952 nd6_ns_output(ifp, NULL, &dst->sin6_addr, ln, 0); 1953 } 1954 return (0); 1955 1956 sendpkt: 1957 /* discard the packet if IPv6 operation is disabled on the interface */ 1958 if ((ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED)) { 1959 error = ENETDOWN; /* better error? */ 1960 goto bad; 1961 } 1962 /* 1963 * ln is valid and the caller did not pass in 1964 * an llentry 1965 */ 1966 if ((ln != NULL) && (lle == NULL)) { 1967 if (flags & LLE_EXCLUSIVE) 1968 LLE_WUNLOCK(ln); 1969 else 1970 LLE_RUNLOCK(ln); 1971 } 1972 1973 #ifdef MAC 1974 mac_netinet6_nd6_send(ifp, m); 1975 #endif 1976 /* 1977 * We were passed in a pointer to an lle with the lock held 1978 * this means that we can't call if_output as we will 1979 * recurse on the lle lock - so what we do is we create 1980 * a list of mbufs to send and transmit them in the caller 1981 * after the lock is dropped 1982 */ 1983 if (lle != NULL) { 1984 if (*chain == NULL) 1985 *chain = m; 1986 else { 1987 struct mbuf *m = *chain; 1988 1989 /* 1990 * append mbuf to end of deferred chain 1991 */ 1992 while (m->m_nextpkt != NULL) 1993 m = m->m_nextpkt; 1994 m->m_nextpkt = m; 1995 } 1996 return (error); 1997 } 1998 if ((ifp->if_flags & IFF_LOOPBACK) != 0) { 1999 return ((*ifp->if_output)(origifp, m, (struct sockaddr *)dst, 2000 NULL)); 2001 } 2002 error = (*ifp->if_output)(ifp, m, (struct sockaddr *)dst, NULL); 2003 return (error); 2004 2005 bad: 2006 /* 2007 * ln is valid and the caller did not pass in 2008 * an llentry 2009 */ 2010 if ((ln != NULL) && (lle == NULL)) { 2011 if (flags & LLE_EXCLUSIVE) 2012 LLE_WUNLOCK(ln); 2013 else 2014 LLE_RUNLOCK(ln); 2015 } 2016 if (m) 2017 m_freem(m); 2018 return (error); 2019 } 2020 #undef senderr 2021 2022 2023 int 2024 nd6_output_flush(struct ifnet *ifp, struct ifnet *origifp, struct mbuf *chain, 2025 struct sockaddr_in6 *dst, struct route *ro) 2026 { 2027 struct mbuf *m, *m_head; 2028 struct ifnet *outifp; 2029 int error = 0; 2030 2031 m_head = chain; 2032 if ((ifp->if_flags & IFF_LOOPBACK) != 0) 2033 outifp = origifp; 2034 else 2035 outifp = ifp; 2036 2037 while (m_head) { 2038 m = m_head; 2039 m_head = m_head->m_nextpkt; 2040 error = (*ifp->if_output)(ifp, m, (struct sockaddr *)dst, ro); 2041 } 2042 2043 /* 2044 * XXX 2045 * note that intermediate errors are blindly ignored - but this is 2046 * the same convention as used with nd6_output when called by 2047 * nd6_cache_lladdr 2048 */ 2049 return (error); 2050 } 2051 2052 2053 int 2054 nd6_need_cache(struct ifnet *ifp) 2055 { 2056 /* 2057 * XXX: we currently do not make neighbor cache on any interface 2058 * other than ARCnet, Ethernet, FDDI and GIF. 2059 * 2060 * RFC2893 says: 2061 * - unidirectional tunnels needs no ND 2062 */ 2063 switch (ifp->if_type) { 2064 case IFT_ARCNET: 2065 case IFT_ETHER: 2066 case IFT_FDDI: 2067 case IFT_IEEE1394: 2068 #ifdef IFT_L2VLAN 2069 case IFT_L2VLAN: 2070 #endif 2071 #ifdef IFT_IEEE80211 2072 case IFT_IEEE80211: 2073 #endif 2074 #ifdef IFT_CARP 2075 case IFT_CARP: 2076 #endif 2077 case IFT_GIF: /* XXX need more cases? */ 2078 case IFT_PPP: 2079 case IFT_TUNNEL: 2080 case IFT_BRIDGE: 2081 case IFT_PROPVIRTUAL: 2082 return (1); 2083 default: 2084 return (0); 2085 } 2086 } 2087 2088 /* 2089 * the callers of this function need to be re-worked to drop 2090 * the lle lock, drop here for now 2091 */ 2092 int 2093 nd6_storelladdr(struct ifnet *ifp, struct mbuf *m, 2094 struct sockaddr *dst, u_char *desten, struct llentry **lle) 2095 { 2096 struct llentry *ln; 2097 2098 *lle = NULL; 2099 IF_AFDATA_UNLOCK_ASSERT(ifp); 2100 if (m->m_flags & M_MCAST) { 2101 int i; 2102 2103 switch (ifp->if_type) { 2104 case IFT_ETHER: 2105 case IFT_FDDI: 2106 #ifdef IFT_L2VLAN 2107 case IFT_L2VLAN: 2108 #endif 2109 #ifdef IFT_IEEE80211 2110 case IFT_IEEE80211: 2111 #endif 2112 case IFT_BRIDGE: 2113 case IFT_ISO88025: 2114 ETHER_MAP_IPV6_MULTICAST(&SIN6(dst)->sin6_addr, 2115 desten); 2116 return (0); 2117 case IFT_IEEE1394: 2118 /* 2119 * netbsd can use if_broadcastaddr, but we don't do so 2120 * to reduce # of ifdef. 2121 */ 2122 for (i = 0; i < ifp->if_addrlen; i++) 2123 desten[i] = ~0; 2124 return (0); 2125 case IFT_ARCNET: 2126 *desten = 0; 2127 return (0); 2128 default: 2129 m_freem(m); 2130 return (EAFNOSUPPORT); 2131 } 2132 } 2133 2134 2135 /* 2136 * the entry should have been created in nd6_store_lladdr 2137 */ 2138 IF_AFDATA_LOCK(ifp); 2139 ln = lla_lookup(LLTABLE6(ifp), 0, dst); 2140 IF_AFDATA_UNLOCK(ifp); 2141 if ((ln == NULL) || !(ln->la_flags & LLE_VALID)) { 2142 if (ln != NULL) 2143 LLE_RUNLOCK(ln); 2144 /* this could happen, if we could not allocate memory */ 2145 m_freem(m); 2146 return (1); 2147 } 2148 2149 bcopy(&ln->ll_addr, desten, ifp->if_addrlen); 2150 *lle = ln; 2151 LLE_RUNLOCK(ln); 2152 /* 2153 * A *small* use after free race exists here 2154 */ 2155 return (0); 2156 } 2157 2158 static void 2159 clear_llinfo_pqueue(struct llentry *ln) 2160 { 2161 struct mbuf *m_hold, *m_hold_next; 2162 2163 for (m_hold = ln->la_hold; m_hold; m_hold = m_hold_next) { 2164 m_hold_next = m_hold->m_nextpkt; 2165 m_hold->m_nextpkt = NULL; 2166 m_freem(m_hold); 2167 } 2168 2169 ln->la_hold = NULL; 2170 return; 2171 } 2172 2173 static int nd6_sysctl_drlist(SYSCTL_HANDLER_ARGS); 2174 static int nd6_sysctl_prlist(SYSCTL_HANDLER_ARGS); 2175 #ifdef SYSCTL_DECL 2176 SYSCTL_DECL(_net_inet6_icmp6); 2177 #endif 2178 SYSCTL_NODE(_net_inet6_icmp6, ICMPV6CTL_ND6_DRLIST, nd6_drlist, 2179 CTLFLAG_RD, nd6_sysctl_drlist, ""); 2180 SYSCTL_NODE(_net_inet6_icmp6, ICMPV6CTL_ND6_PRLIST, nd6_prlist, 2181 CTLFLAG_RD, nd6_sysctl_prlist, ""); 2182 SYSCTL_VNET_INT(_net_inet6_icmp6, ICMPV6CTL_ND6_MAXQLEN, nd6_maxqueuelen, 2183 CTLFLAG_RW, &VNET_NAME(nd6_maxqueuelen), 1, ""); 2184 2185 static int 2186 nd6_sysctl_drlist(SYSCTL_HANDLER_ARGS) 2187 { 2188 int error; 2189 char buf[1024] __aligned(4); 2190 struct in6_defrouter *d, *de; 2191 struct nd_defrouter *dr; 2192 2193 if (req->newptr) 2194 return EPERM; 2195 error = 0; 2196 2197 for (dr = TAILQ_FIRST(&V_nd_defrouter); dr; 2198 dr = TAILQ_NEXT(dr, dr_entry)) { 2199 d = (struct in6_defrouter *)buf; 2200 de = (struct in6_defrouter *)(buf + sizeof(buf)); 2201 2202 if (d + 1 <= de) { 2203 bzero(d, sizeof(*d)); 2204 d->rtaddr.sin6_family = AF_INET6; 2205 d->rtaddr.sin6_len = sizeof(d->rtaddr); 2206 d->rtaddr.sin6_addr = dr->rtaddr; 2207 error = sa6_recoverscope(&d->rtaddr); 2208 if (error != 0) 2209 return (error); 2210 d->flags = dr->flags; 2211 d->rtlifetime = dr->rtlifetime; 2212 d->expire = dr->expire; 2213 d->if_index = dr->ifp->if_index; 2214 } else 2215 panic("buffer too short"); 2216 2217 error = SYSCTL_OUT(req, buf, sizeof(*d)); 2218 if (error) 2219 break; 2220 } 2221 2222 return (error); 2223 } 2224 2225 static int 2226 nd6_sysctl_prlist(SYSCTL_HANDLER_ARGS) 2227 { 2228 int error; 2229 char buf[1024] __aligned(4); 2230 struct in6_prefix *p, *pe; 2231 struct nd_prefix *pr; 2232 char ip6buf[INET6_ADDRSTRLEN]; 2233 2234 if (req->newptr) 2235 return EPERM; 2236 error = 0; 2237 2238 for (pr = V_nd_prefix.lh_first; pr; pr = pr->ndpr_next) { 2239 u_short advrtrs; 2240 size_t advance; 2241 struct sockaddr_in6 *sin6, *s6; 2242 struct nd_pfxrouter *pfr; 2243 2244 p = (struct in6_prefix *)buf; 2245 pe = (struct in6_prefix *)(buf + sizeof(buf)); 2246 2247 if (p + 1 <= pe) { 2248 bzero(p, sizeof(*p)); 2249 sin6 = (struct sockaddr_in6 *)(p + 1); 2250 2251 p->prefix = pr->ndpr_prefix; 2252 if (sa6_recoverscope(&p->prefix)) { 2253 log(LOG_ERR, 2254 "scope error in prefix list (%s)\n", 2255 ip6_sprintf(ip6buf, &p->prefix.sin6_addr)); 2256 /* XXX: press on... */ 2257 } 2258 p->raflags = pr->ndpr_raf; 2259 p->prefixlen = pr->ndpr_plen; 2260 p->vltime = pr->ndpr_vltime; 2261 p->pltime = pr->ndpr_pltime; 2262 p->if_index = pr->ndpr_ifp->if_index; 2263 if (pr->ndpr_vltime == ND6_INFINITE_LIFETIME) 2264 p->expire = 0; 2265 else { 2266 time_t maxexpire; 2267 2268 /* XXX: we assume time_t is signed. */ 2269 maxexpire = (-1) & 2270 ~((time_t)1 << 2271 ((sizeof(maxexpire) * 8) - 1)); 2272 if (pr->ndpr_vltime < 2273 maxexpire - pr->ndpr_lastupdate) { 2274 p->expire = pr->ndpr_lastupdate + 2275 pr->ndpr_vltime; 2276 } else 2277 p->expire = maxexpire; 2278 } 2279 p->refcnt = pr->ndpr_refcnt; 2280 p->flags = pr->ndpr_stateflags; 2281 p->origin = PR_ORIG_RA; 2282 advrtrs = 0; 2283 for (pfr = pr->ndpr_advrtrs.lh_first; pfr; 2284 pfr = pfr->pfr_next) { 2285 if ((void *)&sin6[advrtrs + 1] > (void *)pe) { 2286 advrtrs++; 2287 continue; 2288 } 2289 s6 = &sin6[advrtrs]; 2290 bzero(s6, sizeof(*s6)); 2291 s6->sin6_family = AF_INET6; 2292 s6->sin6_len = sizeof(*sin6); 2293 s6->sin6_addr = pfr->router->rtaddr; 2294 if (sa6_recoverscope(s6)) { 2295 log(LOG_ERR, 2296 "scope error in " 2297 "prefix list (%s)\n", 2298 ip6_sprintf(ip6buf, 2299 &pfr->router->rtaddr)); 2300 } 2301 advrtrs++; 2302 } 2303 p->advrtrs = advrtrs; 2304 } else 2305 panic("buffer too short"); 2306 2307 advance = sizeof(*p) + sizeof(*sin6) * advrtrs; 2308 error = SYSCTL_OUT(req, buf, advance); 2309 if (error) 2310 break; 2311 } 2312 2313 return (error); 2314 } 2315