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 } 768 769 if (public_ifa6 != NULL) { 770 int e; 771 772 if ((e = in6_tmpifadd(public_ifa6, 0, 0)) != 0) { 773 IF_ADDR_UNLOCK(ifp); 774 log(LOG_NOTICE, "regen_tmpaddr: failed to create a new" 775 " tmp addr,errno=%d\n", e); 776 return (-1); 777 } 778 IF_ADDR_UNLOCK(ifp); 779 return (0); 780 } 781 782 IF_ADDR_UNLOCK(ifp); 783 return (-1); 784 } 785 786 /* 787 * Nuke neighbor cache/prefix/default router management table, right before 788 * ifp goes away. 789 */ 790 void 791 nd6_purge(struct ifnet *ifp) 792 { 793 struct nd_defrouter *dr, *ndr; 794 struct nd_prefix *pr, *npr; 795 796 /* 797 * Nuke default router list entries toward ifp. 798 * We defer removal of default router list entries that is installed 799 * in the routing table, in order to keep additional side effects as 800 * small as possible. 801 */ 802 for (dr = TAILQ_FIRST(&V_nd_defrouter); dr; dr = ndr) { 803 ndr = TAILQ_NEXT(dr, dr_entry); 804 if (dr->installed) 805 continue; 806 807 if (dr->ifp == ifp) 808 defrtrlist_del(dr); 809 } 810 811 for (dr = TAILQ_FIRST(&V_nd_defrouter); dr; dr = ndr) { 812 ndr = TAILQ_NEXT(dr, dr_entry); 813 if (!dr->installed) 814 continue; 815 816 if (dr->ifp == ifp) 817 defrtrlist_del(dr); 818 } 819 820 /* Nuke prefix list entries toward ifp */ 821 for (pr = V_nd_prefix.lh_first; pr; pr = npr) { 822 npr = pr->ndpr_next; 823 if (pr->ndpr_ifp == ifp) { 824 /* 825 * Because if_detach() does *not* release prefixes 826 * while purging addresses the reference count will 827 * still be above zero. We therefore reset it to 828 * make sure that the prefix really gets purged. 829 */ 830 pr->ndpr_refcnt = 0; 831 832 /* 833 * Previously, pr->ndpr_addr is removed as well, 834 * but I strongly believe we don't have to do it. 835 * nd6_purge() is only called from in6_ifdetach(), 836 * which removes all the associated interface addresses 837 * by itself. 838 * (jinmei@kame.net 20010129) 839 */ 840 prelist_remove(pr); 841 } 842 } 843 844 /* cancel default outgoing interface setting */ 845 if (V_nd6_defifindex == ifp->if_index) 846 nd6_setdefaultiface(0); 847 848 if (!V_ip6_forwarding && ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV) { 849 /* Refresh default router list. */ 850 defrouter_select(); 851 } 852 853 /* XXXXX 854 * We do not nuke the neighbor cache entries here any more 855 * because the neighbor cache is kept in if_afdata[AF_INET6]. 856 * nd6_purge() is invoked by in6_ifdetach() which is called 857 * from if_detach() where everything gets purged. So let 858 * in6_domifdetach() do the actual L2 table purging work. 859 */ 860 } 861 862 /* 863 * the caller acquires and releases the lock on the lltbls 864 * Returns the llentry locked 865 */ 866 struct llentry * 867 nd6_lookup(struct in6_addr *addr6, int flags, struct ifnet *ifp) 868 { 869 struct sockaddr_in6 sin6; 870 struct llentry *ln; 871 int llflags = 0; 872 873 bzero(&sin6, sizeof(sin6)); 874 sin6.sin6_len = sizeof(struct sockaddr_in6); 875 sin6.sin6_family = AF_INET6; 876 sin6.sin6_addr = *addr6; 877 878 IF_AFDATA_LOCK_ASSERT(ifp); 879 880 if (flags & ND6_CREATE) 881 llflags |= LLE_CREATE; 882 if (flags & ND6_EXCLUSIVE) 883 llflags |= LLE_EXCLUSIVE; 884 885 ln = lla_lookup(LLTABLE6(ifp), llflags, (struct sockaddr *)&sin6); 886 if ((ln != NULL) && (flags & LLE_CREATE)) { 887 ln->ln_state = ND6_LLINFO_NOSTATE; 888 callout_init(&ln->ln_timer_ch, 0); 889 } 890 891 return (ln); 892 } 893 894 /* 895 * Test whether a given IPv6 address is a neighbor or not, ignoring 896 * the actual neighbor cache. The neighbor cache is ignored in order 897 * to not reenter the routing code from within itself. 898 */ 899 static int 900 nd6_is_new_addr_neighbor(struct sockaddr_in6 *addr, struct ifnet *ifp) 901 { 902 struct nd_prefix *pr; 903 struct ifaddr *dstaddr; 904 905 /* 906 * A link-local address is always a neighbor. 907 * XXX: a link does not necessarily specify a single interface. 908 */ 909 if (IN6_IS_ADDR_LINKLOCAL(&addr->sin6_addr)) { 910 struct sockaddr_in6 sin6_copy; 911 u_int32_t zone; 912 913 /* 914 * We need sin6_copy since sa6_recoverscope() may modify the 915 * content (XXX). 916 */ 917 sin6_copy = *addr; 918 if (sa6_recoverscope(&sin6_copy)) 919 return (0); /* XXX: should be impossible */ 920 if (in6_setscope(&sin6_copy.sin6_addr, ifp, &zone)) 921 return (0); 922 if (sin6_copy.sin6_scope_id == zone) 923 return (1); 924 else 925 return (0); 926 } 927 928 /* 929 * If the address matches one of our addresses, 930 * it should be a neighbor. 931 * If the address matches one of our on-link prefixes, it should be a 932 * neighbor. 933 */ 934 for (pr = V_nd_prefix.lh_first; pr; pr = pr->ndpr_next) { 935 if (pr->ndpr_ifp != ifp) 936 continue; 937 938 if (!(pr->ndpr_stateflags & NDPRF_ONLINK)) 939 continue; 940 941 if (IN6_ARE_MASKED_ADDR_EQUAL(&pr->ndpr_prefix.sin6_addr, 942 &addr->sin6_addr, &pr->ndpr_mask)) 943 return (1); 944 } 945 946 /* 947 * If the address is assigned on the node of the other side of 948 * a p2p interface, the address should be a neighbor. 949 */ 950 dstaddr = ifa_ifwithdstaddr((struct sockaddr *)addr); 951 if (dstaddr != NULL) { 952 if (dstaddr->ifa_ifp == ifp) { 953 ifa_free(dstaddr); 954 return (1); 955 } 956 ifa_free(dstaddr); 957 } 958 959 /* 960 * If the default router list is empty, all addresses are regarded 961 * as on-link, and thus, as a neighbor. 962 * XXX: we restrict the condition to hosts, because routers usually do 963 * not have the "default router list". 964 */ 965 if (!V_ip6_forwarding && TAILQ_FIRST(&V_nd_defrouter) == NULL && 966 V_nd6_defifindex == ifp->if_index) { 967 return (1); 968 } 969 970 return (0); 971 } 972 973 974 /* 975 * Detect if a given IPv6 address identifies a neighbor on a given link. 976 * XXX: should take care of the destination of a p2p link? 977 */ 978 int 979 nd6_is_addr_neighbor(struct sockaddr_in6 *addr, struct ifnet *ifp) 980 { 981 struct llentry *lle; 982 int rc = 0; 983 984 IF_AFDATA_UNLOCK_ASSERT(ifp); 985 if (nd6_is_new_addr_neighbor(addr, ifp)) 986 return (1); 987 988 /* 989 * Even if the address matches none of our addresses, it might be 990 * in the neighbor cache. 991 */ 992 IF_AFDATA_LOCK(ifp); 993 if ((lle = nd6_lookup(&addr->sin6_addr, 0, ifp)) != NULL) { 994 LLE_RUNLOCK(lle); 995 rc = 1; 996 } 997 IF_AFDATA_UNLOCK(ifp); 998 return (rc); 999 } 1000 1001 /* 1002 * Free an nd6 llinfo entry. 1003 * Since the function would cause significant changes in the kernel, DO NOT 1004 * make it global, unless you have a strong reason for the change, and are sure 1005 * that the change is safe. 1006 */ 1007 static struct llentry * 1008 nd6_free(struct llentry *ln, int gc) 1009 { 1010 struct llentry *next; 1011 struct nd_defrouter *dr; 1012 struct ifnet *ifp=NULL; 1013 1014 /* 1015 * we used to have pfctlinput(PRC_HOSTDEAD) here. 1016 * even though it is not harmful, it was not really necessary. 1017 */ 1018 1019 /* cancel timer */ 1020 nd6_llinfo_settimer(ln, -1); 1021 1022 if (!V_ip6_forwarding) { 1023 int s; 1024 s = splnet(); 1025 dr = defrouter_lookup(&L3_ADDR_SIN6(ln)->sin6_addr, ln->lle_tbl->llt_ifp); 1026 1027 if (dr != NULL && dr->expire && 1028 ln->ln_state == ND6_LLINFO_STALE && gc) { 1029 /* 1030 * If the reason for the deletion is just garbage 1031 * collection, and the neighbor is an active default 1032 * router, do not delete it. Instead, reset the GC 1033 * timer using the router's lifetime. 1034 * Simply deleting the entry would affect default 1035 * router selection, which is not necessarily a good 1036 * thing, especially when we're using router preference 1037 * values. 1038 * XXX: the check for ln_state would be redundant, 1039 * but we intentionally keep it just in case. 1040 */ 1041 if (dr->expire > time_second) 1042 nd6_llinfo_settimer(ln, 1043 (dr->expire - time_second) * hz); 1044 else 1045 nd6_llinfo_settimer(ln, (long)V_nd6_gctimer * hz); 1046 splx(s); 1047 LLE_WLOCK(ln); 1048 LLE_REMREF(ln); 1049 LLE_WUNLOCK(ln); 1050 return (LIST_NEXT(ln, lle_next)); 1051 } 1052 1053 if (ln->ln_router || dr) { 1054 /* 1055 * rt6_flush must be called whether or not the neighbor 1056 * is in the Default Router List. 1057 * See a corresponding comment in nd6_na_input(). 1058 */ 1059 rt6_flush(&L3_ADDR_SIN6(ln)->sin6_addr, ln->lle_tbl->llt_ifp); 1060 } 1061 1062 if (dr) { 1063 /* 1064 * Unreachablity of a router might affect the default 1065 * router selection and on-link detection of advertised 1066 * prefixes. 1067 */ 1068 1069 /* 1070 * Temporarily fake the state to choose a new default 1071 * router and to perform on-link determination of 1072 * prefixes correctly. 1073 * Below the state will be set correctly, 1074 * or the entry itself will be deleted. 1075 */ 1076 ln->ln_state = ND6_LLINFO_INCOMPLETE; 1077 1078 /* 1079 * Since defrouter_select() does not affect the 1080 * on-link determination and MIP6 needs the check 1081 * before the default router selection, we perform 1082 * the check now. 1083 */ 1084 pfxlist_onlink_check(); 1085 1086 /* 1087 * refresh default router list 1088 */ 1089 defrouter_select(); 1090 } 1091 splx(s); 1092 } 1093 1094 /* 1095 * Before deleting the entry, remember the next entry as the 1096 * return value. We need this because pfxlist_onlink_check() above 1097 * might have freed other entries (particularly the old next entry) as 1098 * a side effect (XXX). 1099 */ 1100 next = LIST_NEXT(ln, lle_next); 1101 1102 ifp = ln->lle_tbl->llt_ifp; 1103 IF_AFDATA_LOCK(ifp); 1104 LLE_WLOCK(ln); 1105 llentry_free(ln); 1106 IF_AFDATA_UNLOCK(ifp); 1107 1108 return (next); 1109 } 1110 1111 /* 1112 * Upper-layer reachability hint for Neighbor Unreachability Detection. 1113 * 1114 * XXX cost-effective methods? 1115 */ 1116 void 1117 nd6_nud_hint(struct rtentry *rt, struct in6_addr *dst6, int force) 1118 { 1119 struct llentry *ln; 1120 struct ifnet *ifp; 1121 1122 if ((dst6 == NULL) || (rt == NULL)) 1123 return; 1124 1125 ifp = rt->rt_ifp; 1126 IF_AFDATA_LOCK(ifp); 1127 ln = nd6_lookup(dst6, ND6_EXCLUSIVE, NULL); 1128 IF_AFDATA_UNLOCK(ifp); 1129 if (ln == NULL) 1130 return; 1131 1132 if (ln->ln_state < ND6_LLINFO_REACHABLE) 1133 goto done; 1134 1135 /* 1136 * if we get upper-layer reachability confirmation many times, 1137 * it is possible we have false information. 1138 */ 1139 if (!force) { 1140 ln->ln_byhint++; 1141 if (ln->ln_byhint > V_nd6_maxnudhint) { 1142 goto done; 1143 } 1144 } 1145 1146 ln->ln_state = ND6_LLINFO_REACHABLE; 1147 if (!ND6_LLINFO_PERMANENT(ln)) { 1148 nd6_llinfo_settimer(ln, 1149 (long)ND_IFINFO(rt->rt_ifp)->reachable * hz); 1150 } 1151 done: 1152 LLE_WUNLOCK(ln); 1153 } 1154 1155 1156 int 1157 nd6_ioctl(u_long cmd, caddr_t data, struct ifnet *ifp) 1158 { 1159 struct in6_drlist *drl = (struct in6_drlist *)data; 1160 struct in6_oprlist *oprl = (struct in6_oprlist *)data; 1161 struct in6_ndireq *ndi = (struct in6_ndireq *)data; 1162 struct in6_nbrinfo *nbi = (struct in6_nbrinfo *)data; 1163 struct in6_ndifreq *ndif = (struct in6_ndifreq *)data; 1164 struct nd_defrouter *dr; 1165 struct nd_prefix *pr; 1166 int i = 0, error = 0; 1167 int s; 1168 1169 switch (cmd) { 1170 case SIOCGDRLST_IN6: 1171 /* 1172 * obsolete API, use sysctl under net.inet6.icmp6 1173 */ 1174 bzero(drl, sizeof(*drl)); 1175 s = splnet(); 1176 dr = TAILQ_FIRST(&V_nd_defrouter); 1177 while (dr && i < DRLSTSIZ) { 1178 drl->defrouter[i].rtaddr = dr->rtaddr; 1179 in6_clearscope(&drl->defrouter[i].rtaddr); 1180 1181 drl->defrouter[i].flags = dr->flags; 1182 drl->defrouter[i].rtlifetime = dr->rtlifetime; 1183 drl->defrouter[i].expire = dr->expire; 1184 drl->defrouter[i].if_index = dr->ifp->if_index; 1185 i++; 1186 dr = TAILQ_NEXT(dr, dr_entry); 1187 } 1188 splx(s); 1189 break; 1190 case SIOCGPRLST_IN6: 1191 /* 1192 * obsolete API, use sysctl under net.inet6.icmp6 1193 * 1194 * XXX the structure in6_prlist was changed in backward- 1195 * incompatible manner. in6_oprlist is used for SIOCGPRLST_IN6, 1196 * in6_prlist is used for nd6_sysctl() - fill_prlist(). 1197 */ 1198 /* 1199 * XXX meaning of fields, especialy "raflags", is very 1200 * differnet between RA prefix list and RR/static prefix list. 1201 * how about separating ioctls into two? 1202 */ 1203 bzero(oprl, sizeof(*oprl)); 1204 s = splnet(); 1205 pr = V_nd_prefix.lh_first; 1206 while (pr && i < PRLSTSIZ) { 1207 struct nd_pfxrouter *pfr; 1208 int j; 1209 1210 oprl->prefix[i].prefix = pr->ndpr_prefix.sin6_addr; 1211 oprl->prefix[i].raflags = pr->ndpr_raf; 1212 oprl->prefix[i].prefixlen = pr->ndpr_plen; 1213 oprl->prefix[i].vltime = pr->ndpr_vltime; 1214 oprl->prefix[i].pltime = pr->ndpr_pltime; 1215 oprl->prefix[i].if_index = pr->ndpr_ifp->if_index; 1216 if (pr->ndpr_vltime == ND6_INFINITE_LIFETIME) 1217 oprl->prefix[i].expire = 0; 1218 else { 1219 time_t maxexpire; 1220 1221 /* XXX: we assume time_t is signed. */ 1222 maxexpire = (-1) & 1223 ~((time_t)1 << 1224 ((sizeof(maxexpire) * 8) - 1)); 1225 if (pr->ndpr_vltime < 1226 maxexpire - pr->ndpr_lastupdate) { 1227 oprl->prefix[i].expire = 1228 pr->ndpr_lastupdate + 1229 pr->ndpr_vltime; 1230 } else 1231 oprl->prefix[i].expire = maxexpire; 1232 } 1233 1234 pfr = pr->ndpr_advrtrs.lh_first; 1235 j = 0; 1236 while (pfr) { 1237 if (j < DRLSTSIZ) { 1238 #define RTRADDR oprl->prefix[i].advrtr[j] 1239 RTRADDR = pfr->router->rtaddr; 1240 in6_clearscope(&RTRADDR); 1241 #undef RTRADDR 1242 } 1243 j++; 1244 pfr = pfr->pfr_next; 1245 } 1246 oprl->prefix[i].advrtrs = j; 1247 oprl->prefix[i].origin = PR_ORIG_RA; 1248 1249 i++; 1250 pr = pr->ndpr_next; 1251 } 1252 splx(s); 1253 1254 break; 1255 case OSIOCGIFINFO_IN6: 1256 #define ND ndi->ndi 1257 /* XXX: old ndp(8) assumes a positive value for linkmtu. */ 1258 bzero(&ND, sizeof(ND)); 1259 ND.linkmtu = IN6_LINKMTU(ifp); 1260 ND.maxmtu = ND_IFINFO(ifp)->maxmtu; 1261 ND.basereachable = ND_IFINFO(ifp)->basereachable; 1262 ND.reachable = ND_IFINFO(ifp)->reachable; 1263 ND.retrans = ND_IFINFO(ifp)->retrans; 1264 ND.flags = ND_IFINFO(ifp)->flags; 1265 ND.recalctm = ND_IFINFO(ifp)->recalctm; 1266 ND.chlim = ND_IFINFO(ifp)->chlim; 1267 break; 1268 case SIOCGIFINFO_IN6: 1269 ND = *ND_IFINFO(ifp); 1270 break; 1271 case SIOCSIFINFO_IN6: 1272 /* 1273 * used to change host variables from userland. 1274 * intented for a use on router to reflect RA configurations. 1275 */ 1276 /* 0 means 'unspecified' */ 1277 if (ND.linkmtu != 0) { 1278 if (ND.linkmtu < IPV6_MMTU || 1279 ND.linkmtu > IN6_LINKMTU(ifp)) { 1280 error = EINVAL; 1281 break; 1282 } 1283 ND_IFINFO(ifp)->linkmtu = ND.linkmtu; 1284 } 1285 1286 if (ND.basereachable != 0) { 1287 int obasereachable = ND_IFINFO(ifp)->basereachable; 1288 1289 ND_IFINFO(ifp)->basereachable = ND.basereachable; 1290 if (ND.basereachable != obasereachable) 1291 ND_IFINFO(ifp)->reachable = 1292 ND_COMPUTE_RTIME(ND.basereachable); 1293 } 1294 if (ND.retrans != 0) 1295 ND_IFINFO(ifp)->retrans = ND.retrans; 1296 if (ND.chlim != 0) 1297 ND_IFINFO(ifp)->chlim = ND.chlim; 1298 /* FALLTHROUGH */ 1299 case SIOCSIFINFO_FLAGS: 1300 { 1301 struct ifaddr *ifa; 1302 struct in6_ifaddr *ia; 1303 1304 if ((ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) && 1305 !(ND.flags & ND6_IFF_IFDISABLED)) { 1306 /* ifdisabled 1->0 transision */ 1307 1308 /* 1309 * If the interface is marked as ND6_IFF_IFDISABLED and 1310 * has an link-local address with IN6_IFF_DUPLICATED, 1311 * do not clear ND6_IFF_IFDISABLED. 1312 * See RFC 4862, Section 5.4.5. 1313 */ 1314 int duplicated_linklocal = 0; 1315 1316 IF_ADDR_LOCK(ifp); 1317 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1318 if (ifa->ifa_addr->sa_family != AF_INET6) 1319 continue; 1320 ia = (struct in6_ifaddr *)ifa; 1321 if ((ia->ia6_flags & IN6_IFF_DUPLICATED) && 1322 IN6_IS_ADDR_LINKLOCAL(&ia->ia_addr.sin6_addr)) { 1323 duplicated_linklocal = 1; 1324 break; 1325 } 1326 } 1327 IF_ADDR_UNLOCK(ifp); 1328 1329 if (duplicated_linklocal) { 1330 ND.flags |= ND6_IFF_IFDISABLED; 1331 log(LOG_ERR, "Cannot enable an interface" 1332 " with a link-local address marked" 1333 " duplicate.\n"); 1334 } else { 1335 ND_IFINFO(ifp)->flags &= ~ND6_IFF_IFDISABLED; 1336 in6_if_up(ifp); 1337 } 1338 } else if (!(ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) && 1339 (ND.flags & ND6_IFF_IFDISABLED)) { 1340 /* ifdisabled 0->1 transision */ 1341 /* Mark all IPv6 address as tentative. */ 1342 1343 ND_IFINFO(ifp)->flags |= ND6_IFF_IFDISABLED; 1344 IF_ADDR_LOCK(ifp); 1345 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1346 if (ifa->ifa_addr->sa_family != AF_INET6) 1347 continue; 1348 ia = (struct in6_ifaddr *)ifa; 1349 ia->ia6_flags |= IN6_IFF_TENTATIVE; 1350 } 1351 IF_ADDR_UNLOCK(ifp); 1352 } 1353 1354 if (!(ND_IFINFO(ifp)->flags & ND6_IFF_AUTO_LINKLOCAL) && 1355 (ND.flags & ND6_IFF_AUTO_LINKLOCAL)) { 1356 /* auto_linklocal 0->1 transision */ 1357 1358 /* If no link-local address on ifp, configure */ 1359 ND_IFINFO(ifp)->flags |= ND6_IFF_AUTO_LINKLOCAL; 1360 in6_ifattach(ifp, NULL); 1361 } 1362 } 1363 ND_IFINFO(ifp)->flags = ND.flags; 1364 break; 1365 #undef ND 1366 case SIOCSNDFLUSH_IN6: /* XXX: the ioctl name is confusing... */ 1367 /* sync kernel routing table with the default router list */ 1368 defrouter_reset(); 1369 defrouter_select(); 1370 break; 1371 case SIOCSPFXFLUSH_IN6: 1372 { 1373 /* flush all the prefix advertised by routers */ 1374 struct nd_prefix *pr, *next; 1375 1376 s = splnet(); 1377 for (pr = V_nd_prefix.lh_first; pr; pr = next) { 1378 struct in6_ifaddr *ia, *ia_next; 1379 1380 next = pr->ndpr_next; 1381 1382 if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr)) 1383 continue; /* XXX */ 1384 1385 /* do we really have to remove addresses as well? */ 1386 /* XXXRW: in6_ifaddrhead locking. */ 1387 TAILQ_FOREACH_SAFE(ia, &V_in6_ifaddrhead, ia_link, 1388 ia_next) { 1389 if ((ia->ia6_flags & IN6_IFF_AUTOCONF) == 0) 1390 continue; 1391 1392 if (ia->ia6_ndpr == pr) 1393 in6_purgeaddr(&ia->ia_ifa); 1394 } 1395 prelist_remove(pr); 1396 } 1397 splx(s); 1398 break; 1399 } 1400 case SIOCSRTRFLUSH_IN6: 1401 { 1402 /* flush all the default routers */ 1403 struct nd_defrouter *dr, *next; 1404 1405 s = splnet(); 1406 defrouter_reset(); 1407 for (dr = TAILQ_FIRST(&V_nd_defrouter); dr; dr = next) { 1408 next = TAILQ_NEXT(dr, dr_entry); 1409 defrtrlist_del(dr); 1410 } 1411 defrouter_select(); 1412 splx(s); 1413 break; 1414 } 1415 case SIOCGNBRINFO_IN6: 1416 { 1417 struct llentry *ln; 1418 struct in6_addr nb_addr = nbi->addr; /* make local for safety */ 1419 1420 if ((error = in6_setscope(&nb_addr, ifp, NULL)) != 0) 1421 return (error); 1422 1423 IF_AFDATA_LOCK(ifp); 1424 ln = nd6_lookup(&nb_addr, 0, ifp); 1425 IF_AFDATA_UNLOCK(ifp); 1426 1427 if (ln == NULL) { 1428 error = EINVAL; 1429 break; 1430 } 1431 nbi->state = ln->ln_state; 1432 nbi->asked = ln->la_asked; 1433 nbi->isrouter = ln->ln_router; 1434 nbi->expire = ln->la_expire; 1435 LLE_RUNLOCK(ln); 1436 break; 1437 } 1438 case SIOCGDEFIFACE_IN6: /* XXX: should be implemented as a sysctl? */ 1439 ndif->ifindex = V_nd6_defifindex; 1440 break; 1441 case SIOCSDEFIFACE_IN6: /* XXX: should be implemented as a sysctl? */ 1442 return (nd6_setdefaultiface(ndif->ifindex)); 1443 } 1444 return (error); 1445 } 1446 1447 /* 1448 * Create neighbor cache entry and cache link-layer address, 1449 * on reception of inbound ND6 packets. (RS/RA/NS/redirect) 1450 * 1451 * type - ICMP6 type 1452 * code - type dependent information 1453 * 1454 * XXXXX 1455 * The caller of this function already acquired the ndp 1456 * cache table lock because the cache entry is returned. 1457 */ 1458 struct llentry * 1459 nd6_cache_lladdr(struct ifnet *ifp, struct in6_addr *from, char *lladdr, 1460 int lladdrlen, int type, int code) 1461 { 1462 struct llentry *ln = NULL; 1463 int is_newentry; 1464 int do_update; 1465 int olladdr; 1466 int llchange; 1467 int flags = 0; 1468 int newstate = 0; 1469 uint16_t router = 0; 1470 struct sockaddr_in6 sin6; 1471 struct mbuf *chain = NULL; 1472 int static_route = 0; 1473 1474 IF_AFDATA_UNLOCK_ASSERT(ifp); 1475 1476 if (ifp == NULL) 1477 panic("ifp == NULL in nd6_cache_lladdr"); 1478 if (from == NULL) 1479 panic("from == NULL in nd6_cache_lladdr"); 1480 1481 /* nothing must be updated for unspecified address */ 1482 if (IN6_IS_ADDR_UNSPECIFIED(from)) 1483 return NULL; 1484 1485 /* 1486 * Validation about ifp->if_addrlen and lladdrlen must be done in 1487 * the caller. 1488 * 1489 * XXX If the link does not have link-layer adderss, what should 1490 * we do? (ifp->if_addrlen == 0) 1491 * Spec says nothing in sections for RA, RS and NA. There's small 1492 * description on it in NS section (RFC 2461 7.2.3). 1493 */ 1494 flags |= lladdr ? ND6_EXCLUSIVE : 0; 1495 IF_AFDATA_LOCK(ifp); 1496 ln = nd6_lookup(from, flags, ifp); 1497 1498 if (ln == NULL) { 1499 flags |= LLE_EXCLUSIVE; 1500 ln = nd6_lookup(from, flags |ND6_CREATE, ifp); 1501 IF_AFDATA_UNLOCK(ifp); 1502 is_newentry = 1; 1503 } else { 1504 IF_AFDATA_UNLOCK(ifp); 1505 /* do nothing if static ndp is set */ 1506 if (ln->la_flags & LLE_STATIC) { 1507 static_route = 1; 1508 goto done; 1509 } 1510 is_newentry = 0; 1511 } 1512 if (ln == NULL) 1513 return (NULL); 1514 1515 olladdr = (ln->la_flags & LLE_VALID) ? 1 : 0; 1516 if (olladdr && lladdr) { 1517 llchange = bcmp(lladdr, &ln->ll_addr, 1518 ifp->if_addrlen); 1519 } else 1520 llchange = 0; 1521 1522 /* 1523 * newentry olladdr lladdr llchange (*=record) 1524 * 0 n n -- (1) 1525 * 0 y n -- (2) 1526 * 0 n y -- (3) * STALE 1527 * 0 y y n (4) * 1528 * 0 y y y (5) * STALE 1529 * 1 -- n -- (6) NOSTATE(= PASSIVE) 1530 * 1 -- y -- (7) * STALE 1531 */ 1532 1533 if (lladdr) { /* (3-5) and (7) */ 1534 /* 1535 * Record source link-layer address 1536 * XXX is it dependent to ifp->if_type? 1537 */ 1538 bcopy(lladdr, &ln->ll_addr, ifp->if_addrlen); 1539 ln->la_flags |= LLE_VALID; 1540 } 1541 1542 if (!is_newentry) { 1543 if ((!olladdr && lladdr != NULL) || /* (3) */ 1544 (olladdr && lladdr != NULL && llchange)) { /* (5) */ 1545 do_update = 1; 1546 newstate = ND6_LLINFO_STALE; 1547 } else /* (1-2,4) */ 1548 do_update = 0; 1549 } else { 1550 do_update = 1; 1551 if (lladdr == NULL) /* (6) */ 1552 newstate = ND6_LLINFO_NOSTATE; 1553 else /* (7) */ 1554 newstate = ND6_LLINFO_STALE; 1555 } 1556 1557 if (do_update) { 1558 /* 1559 * Update the state of the neighbor cache. 1560 */ 1561 ln->ln_state = newstate; 1562 1563 if (ln->ln_state == ND6_LLINFO_STALE) { 1564 /* 1565 * XXX: since nd6_output() below will cause 1566 * state tansition to DELAY and reset the timer, 1567 * we must set the timer now, although it is actually 1568 * meaningless. 1569 */ 1570 nd6_llinfo_settimer_locked(ln, (long)V_nd6_gctimer * hz); 1571 1572 if (ln->la_hold) { 1573 struct mbuf *m_hold, *m_hold_next; 1574 1575 /* 1576 * reset the la_hold in advance, to explicitly 1577 * prevent a la_hold lookup in nd6_output() 1578 * (wouldn't happen, though...) 1579 */ 1580 for (m_hold = ln->la_hold, ln->la_hold = NULL; 1581 m_hold; m_hold = m_hold_next) { 1582 m_hold_next = m_hold->m_nextpkt; 1583 m_hold->m_nextpkt = NULL; 1584 1585 /* 1586 * we assume ifp is not a p2p here, so 1587 * just set the 2nd argument as the 1588 * 1st one. 1589 */ 1590 nd6_output_lle(ifp, ifp, m_hold, L3_ADDR_SIN6(ln), NULL, ln, &chain); 1591 } 1592 /* 1593 * If we have mbufs in the chain we need to do 1594 * deferred transmit. Copy the address from the 1595 * llentry before dropping the lock down below. 1596 */ 1597 if (chain != NULL) 1598 memcpy(&sin6, L3_ADDR_SIN6(ln), sizeof(sin6)); 1599 } 1600 } else if (ln->ln_state == ND6_LLINFO_INCOMPLETE) { 1601 /* probe right away */ 1602 nd6_llinfo_settimer_locked((void *)ln, 0); 1603 } 1604 } 1605 1606 /* 1607 * ICMP6 type dependent behavior. 1608 * 1609 * NS: clear IsRouter if new entry 1610 * RS: clear IsRouter 1611 * RA: set IsRouter if there's lladdr 1612 * redir: clear IsRouter if new entry 1613 * 1614 * RA case, (1): 1615 * The spec says that we must set IsRouter in the following cases: 1616 * - If lladdr exist, set IsRouter. This means (1-5). 1617 * - If it is old entry (!newentry), set IsRouter. This means (7). 1618 * So, based on the spec, in (1-5) and (7) cases we must set IsRouter. 1619 * A quetion arises for (1) case. (1) case has no lladdr in the 1620 * neighbor cache, this is similar to (6). 1621 * This case is rare but we figured that we MUST NOT set IsRouter. 1622 * 1623 * newentry olladdr lladdr llchange NS RS RA redir 1624 * D R 1625 * 0 n n -- (1) c ? s 1626 * 0 y n -- (2) c s s 1627 * 0 n y -- (3) c s s 1628 * 0 y y n (4) c s s 1629 * 0 y y y (5) c s s 1630 * 1 -- n -- (6) c c c s 1631 * 1 -- y -- (7) c c s c s 1632 * 1633 * (c=clear s=set) 1634 */ 1635 switch (type & 0xff) { 1636 case ND_NEIGHBOR_SOLICIT: 1637 /* 1638 * New entry must have is_router flag cleared. 1639 */ 1640 if (is_newentry) /* (6-7) */ 1641 ln->ln_router = 0; 1642 break; 1643 case ND_REDIRECT: 1644 /* 1645 * If the icmp is a redirect to a better router, always set the 1646 * is_router flag. Otherwise, if the entry is newly created, 1647 * clear the flag. [RFC 2461, sec 8.3] 1648 */ 1649 if (code == ND_REDIRECT_ROUTER) 1650 ln->ln_router = 1; 1651 else if (is_newentry) /* (6-7) */ 1652 ln->ln_router = 0; 1653 break; 1654 case ND_ROUTER_SOLICIT: 1655 /* 1656 * is_router flag must always be cleared. 1657 */ 1658 ln->ln_router = 0; 1659 break; 1660 case ND_ROUTER_ADVERT: 1661 /* 1662 * Mark an entry with lladdr as a router. 1663 */ 1664 if ((!is_newentry && (olladdr || lladdr)) || /* (2-5) */ 1665 (is_newentry && lladdr)) { /* (7) */ 1666 ln->ln_router = 1; 1667 } 1668 break; 1669 } 1670 1671 if (ln != NULL) { 1672 static_route = (ln->la_flags & LLE_STATIC); 1673 router = ln->ln_router; 1674 1675 if (flags & ND6_EXCLUSIVE) 1676 LLE_WUNLOCK(ln); 1677 else 1678 LLE_RUNLOCK(ln); 1679 if (static_route) 1680 ln = NULL; 1681 } 1682 if (chain) 1683 nd6_output_flush(ifp, ifp, chain, &sin6, NULL); 1684 1685 /* 1686 * When the link-layer address of a router changes, select the 1687 * best router again. In particular, when the neighbor entry is newly 1688 * created, it might affect the selection policy. 1689 * Question: can we restrict the first condition to the "is_newentry" 1690 * case? 1691 * XXX: when we hear an RA from a new router with the link-layer 1692 * address option, defrouter_select() is called twice, since 1693 * defrtrlist_update called the function as well. However, I believe 1694 * we can compromise the overhead, since it only happens the first 1695 * time. 1696 * XXX: although defrouter_select() should not have a bad effect 1697 * for those are not autoconfigured hosts, we explicitly avoid such 1698 * cases for safety. 1699 */ 1700 if (do_update && router && !V_ip6_forwarding && 1701 ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV) { 1702 /* 1703 * guaranteed recursion 1704 */ 1705 defrouter_select(); 1706 } 1707 1708 return (ln); 1709 done: 1710 if (ln != NULL) { 1711 if (flags & ND6_EXCLUSIVE) 1712 LLE_WUNLOCK(ln); 1713 else 1714 LLE_RUNLOCK(ln); 1715 if (static_route) 1716 ln = NULL; 1717 } 1718 return (ln); 1719 } 1720 1721 static void 1722 nd6_slowtimo(void *arg) 1723 { 1724 CURVNET_SET((struct vnet *) arg); 1725 struct nd_ifinfo *nd6if; 1726 struct ifnet *ifp; 1727 1728 callout_reset(&V_nd6_slowtimo_ch, ND6_SLOWTIMER_INTERVAL * hz, 1729 nd6_slowtimo, curvnet); 1730 IFNET_RLOCK_NOSLEEP(); 1731 for (ifp = TAILQ_FIRST(&V_ifnet); ifp; 1732 ifp = TAILQ_NEXT(ifp, if_list)) { 1733 nd6if = ND_IFINFO(ifp); 1734 if (nd6if->basereachable && /* already initialized */ 1735 (nd6if->recalctm -= ND6_SLOWTIMER_INTERVAL) <= 0) { 1736 /* 1737 * Since reachable time rarely changes by router 1738 * advertisements, we SHOULD insure that a new random 1739 * value gets recomputed at least once every few hours. 1740 * (RFC 2461, 6.3.4) 1741 */ 1742 nd6if->recalctm = V_nd6_recalc_reachtm_interval; 1743 nd6if->reachable = ND_COMPUTE_RTIME(nd6if->basereachable); 1744 } 1745 } 1746 IFNET_RUNLOCK_NOSLEEP(); 1747 CURVNET_RESTORE(); 1748 } 1749 1750 int 1751 nd6_output(struct ifnet *ifp, struct ifnet *origifp, struct mbuf *m0, 1752 struct sockaddr_in6 *dst, struct rtentry *rt0) 1753 { 1754 1755 return (nd6_output_lle(ifp, origifp, m0, dst, rt0, NULL, NULL)); 1756 } 1757 1758 1759 /* 1760 * Note that I'm not enforcing any global serialization 1761 * lle state or asked changes here as the logic is too 1762 * complicated to avoid having to always acquire an exclusive 1763 * lock 1764 * KMM 1765 * 1766 */ 1767 #define senderr(e) { error = (e); goto bad;} 1768 1769 int 1770 nd6_output_lle(struct ifnet *ifp, struct ifnet *origifp, struct mbuf *m0, 1771 struct sockaddr_in6 *dst, struct rtentry *rt0, struct llentry *lle, 1772 struct mbuf **chain) 1773 { 1774 struct mbuf *m = m0; 1775 struct llentry *ln = lle; 1776 int error = 0; 1777 int flags = 0; 1778 1779 #ifdef INVARIANTS 1780 if (lle != NULL) { 1781 1782 LLE_WLOCK_ASSERT(lle); 1783 1784 KASSERT(chain != NULL, (" lle locked but no mbuf chain pointer passed")); 1785 } 1786 #endif 1787 if (IN6_IS_ADDR_MULTICAST(&dst->sin6_addr)) 1788 goto sendpkt; 1789 1790 if (nd6_need_cache(ifp) == 0) 1791 goto sendpkt; 1792 1793 /* 1794 * next hop determination. This routine is derived from ether_output. 1795 */ 1796 1797 /* 1798 * Address resolution or Neighbor Unreachability Detection 1799 * for the next hop. 1800 * At this point, the destination of the packet must be a unicast 1801 * or an anycast address(i.e. not a multicast). 1802 */ 1803 1804 flags = ((m != NULL) || (lle != NULL)) ? LLE_EXCLUSIVE : 0; 1805 if (ln == NULL) { 1806 retry: 1807 IF_AFDATA_LOCK(ifp); 1808 ln = lla_lookup(LLTABLE6(ifp), flags, (struct sockaddr *)dst); 1809 IF_AFDATA_UNLOCK(ifp); 1810 if ((ln == NULL) && nd6_is_addr_neighbor(dst, ifp)) { 1811 /* 1812 * Since nd6_is_addr_neighbor() internally calls nd6_lookup(), 1813 * the condition below is not very efficient. But we believe 1814 * it is tolerable, because this should be a rare case. 1815 */ 1816 flags = ND6_CREATE | (m ? ND6_EXCLUSIVE : 0); 1817 IF_AFDATA_LOCK(ifp); 1818 ln = nd6_lookup(&dst->sin6_addr, flags, ifp); 1819 IF_AFDATA_UNLOCK(ifp); 1820 } 1821 } 1822 if (ln == NULL) { 1823 if ((ifp->if_flags & IFF_POINTOPOINT) == 0 && 1824 !(ND_IFINFO(ifp)->flags & ND6_IFF_PERFORMNUD)) { 1825 char ip6buf[INET6_ADDRSTRLEN]; 1826 log(LOG_DEBUG, 1827 "nd6_output: can't allocate llinfo for %s " 1828 "(ln=%p)\n", 1829 ip6_sprintf(ip6buf, &dst->sin6_addr), ln); 1830 senderr(EIO); /* XXX: good error? */ 1831 } 1832 goto sendpkt; /* send anyway */ 1833 } 1834 1835 /* We don't have to do link-layer address resolution on a p2p link. */ 1836 if ((ifp->if_flags & IFF_POINTOPOINT) != 0 && 1837 ln->ln_state < ND6_LLINFO_REACHABLE) { 1838 if ((flags & LLE_EXCLUSIVE) == 0) { 1839 flags |= LLE_EXCLUSIVE; 1840 goto retry; 1841 } 1842 ln->ln_state = ND6_LLINFO_STALE; 1843 nd6_llinfo_settimer_locked(ln, (long)V_nd6_gctimer * hz); 1844 } 1845 1846 /* 1847 * The first time we send a packet to a neighbor whose entry is 1848 * STALE, we have to change the state to DELAY and a sets a timer to 1849 * expire in DELAY_FIRST_PROBE_TIME seconds to ensure do 1850 * neighbor unreachability detection on expiration. 1851 * (RFC 2461 7.3.3) 1852 */ 1853 if (ln->ln_state == ND6_LLINFO_STALE) { 1854 if ((flags & LLE_EXCLUSIVE) == 0) { 1855 flags |= LLE_EXCLUSIVE; 1856 LLE_RUNLOCK(ln); 1857 goto retry; 1858 } 1859 ln->la_asked = 0; 1860 ln->ln_state = ND6_LLINFO_DELAY; 1861 nd6_llinfo_settimer_locked(ln, (long)V_nd6_delay * hz); 1862 } 1863 1864 /* 1865 * If the neighbor cache entry has a state other than INCOMPLETE 1866 * (i.e. its link-layer address is already resolved), just 1867 * send the packet. 1868 */ 1869 if (ln->ln_state > ND6_LLINFO_INCOMPLETE) 1870 goto sendpkt; 1871 1872 /* 1873 * There is a neighbor cache entry, but no ethernet address 1874 * response yet. Append this latest packet to the end of the 1875 * packet queue in the mbuf, unless the number of the packet 1876 * does not exceed nd6_maxqueuelen. When it exceeds nd6_maxqueuelen, 1877 * the oldest packet in the queue will be removed. 1878 */ 1879 if (ln->ln_state == ND6_LLINFO_NOSTATE) 1880 ln->ln_state = ND6_LLINFO_INCOMPLETE; 1881 1882 if ((flags & LLE_EXCLUSIVE) == 0) { 1883 flags |= LLE_EXCLUSIVE; 1884 LLE_RUNLOCK(ln); 1885 goto retry; 1886 } 1887 if (ln->la_hold) { 1888 struct mbuf *m_hold; 1889 int i; 1890 1891 i = 0; 1892 for (m_hold = ln->la_hold; m_hold; m_hold = m_hold->m_nextpkt) { 1893 i++; 1894 if (m_hold->m_nextpkt == NULL) { 1895 m_hold->m_nextpkt = m; 1896 break; 1897 } 1898 } 1899 while (i >= V_nd6_maxqueuelen) { 1900 m_hold = ln->la_hold; 1901 ln->la_hold = ln->la_hold->m_nextpkt; 1902 m_freem(m_hold); 1903 i--; 1904 } 1905 } else { 1906 ln->la_hold = m; 1907 } 1908 /* 1909 * We did the lookup (no lle arg) so we 1910 * need to do the unlock here 1911 */ 1912 if (lle == NULL) { 1913 if (flags & LLE_EXCLUSIVE) 1914 LLE_WUNLOCK(ln); 1915 else 1916 LLE_RUNLOCK(ln); 1917 } 1918 1919 /* 1920 * If there has been no NS for the neighbor after entering the 1921 * INCOMPLETE state, send the first solicitation. 1922 */ 1923 if (!ND6_LLINFO_PERMANENT(ln) && ln->la_asked == 0) { 1924 ln->la_asked++; 1925 1926 nd6_llinfo_settimer(ln, 1927 (long)ND_IFINFO(ifp)->retrans * hz / 1000); 1928 nd6_ns_output(ifp, NULL, &dst->sin6_addr, ln, 0); 1929 } 1930 return (0); 1931 1932 sendpkt: 1933 /* discard the packet if IPv6 operation is disabled on the interface */ 1934 if ((ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED)) { 1935 error = ENETDOWN; /* better error? */ 1936 goto bad; 1937 } 1938 /* 1939 * ln is valid and the caller did not pass in 1940 * an llentry 1941 */ 1942 if ((ln != NULL) && (lle == NULL)) { 1943 if (flags & LLE_EXCLUSIVE) 1944 LLE_WUNLOCK(ln); 1945 else 1946 LLE_RUNLOCK(ln); 1947 } 1948 1949 #ifdef MAC 1950 mac_netinet6_nd6_send(ifp, m); 1951 #endif 1952 /* 1953 * We were passed in a pointer to an lle with the lock held 1954 * this means that we can't call if_output as we will 1955 * recurse on the lle lock - so what we do is we create 1956 * a list of mbufs to send and transmit them in the caller 1957 * after the lock is dropped 1958 */ 1959 if (lle != NULL) { 1960 if (*chain == NULL) 1961 *chain = m; 1962 else { 1963 struct mbuf *m = *chain; 1964 1965 /* 1966 * append mbuf to end of deferred chain 1967 */ 1968 while (m->m_nextpkt != NULL) 1969 m = m->m_nextpkt; 1970 m->m_nextpkt = m; 1971 } 1972 return (error); 1973 } 1974 if ((ifp->if_flags & IFF_LOOPBACK) != 0) { 1975 return ((*ifp->if_output)(origifp, m, (struct sockaddr *)dst, 1976 NULL)); 1977 } 1978 error = (*ifp->if_output)(ifp, m, (struct sockaddr *)dst, NULL); 1979 return (error); 1980 1981 bad: 1982 /* 1983 * ln is valid and the caller did not pass in 1984 * an llentry 1985 */ 1986 if ((ln != NULL) && (lle == NULL)) { 1987 if (flags & LLE_EXCLUSIVE) 1988 LLE_WUNLOCK(ln); 1989 else 1990 LLE_RUNLOCK(ln); 1991 } 1992 if (m) 1993 m_freem(m); 1994 return (error); 1995 } 1996 #undef senderr 1997 1998 1999 int 2000 nd6_output_flush(struct ifnet *ifp, struct ifnet *origifp, struct mbuf *chain, 2001 struct sockaddr_in6 *dst, struct route *ro) 2002 { 2003 struct mbuf *m, *m_head; 2004 struct ifnet *outifp; 2005 int error = 0; 2006 2007 m_head = chain; 2008 if ((ifp->if_flags & IFF_LOOPBACK) != 0) 2009 outifp = origifp; 2010 else 2011 outifp = ifp; 2012 2013 while (m_head) { 2014 m = m_head; 2015 m_head = m_head->m_nextpkt; 2016 error = (*ifp->if_output)(ifp, m, (struct sockaddr *)dst, ro); 2017 } 2018 2019 /* 2020 * XXX 2021 * note that intermediate errors are blindly ignored - but this is 2022 * the same convention as used with nd6_output when called by 2023 * nd6_cache_lladdr 2024 */ 2025 return (error); 2026 } 2027 2028 2029 int 2030 nd6_need_cache(struct ifnet *ifp) 2031 { 2032 /* 2033 * XXX: we currently do not make neighbor cache on any interface 2034 * other than ARCnet, Ethernet, FDDI and GIF. 2035 * 2036 * RFC2893 says: 2037 * - unidirectional tunnels needs no ND 2038 */ 2039 switch (ifp->if_type) { 2040 case IFT_ARCNET: 2041 case IFT_ETHER: 2042 case IFT_FDDI: 2043 case IFT_IEEE1394: 2044 #ifdef IFT_L2VLAN 2045 case IFT_L2VLAN: 2046 #endif 2047 #ifdef IFT_IEEE80211 2048 case IFT_IEEE80211: 2049 #endif 2050 #ifdef IFT_CARP 2051 case IFT_CARP: 2052 #endif 2053 case IFT_GIF: /* XXX need more cases? */ 2054 case IFT_PPP: 2055 case IFT_TUNNEL: 2056 case IFT_BRIDGE: 2057 case IFT_PROPVIRTUAL: 2058 return (1); 2059 default: 2060 return (0); 2061 } 2062 } 2063 2064 /* 2065 * the callers of this function need to be re-worked to drop 2066 * the lle lock, drop here for now 2067 */ 2068 int 2069 nd6_storelladdr(struct ifnet *ifp, struct mbuf *m, 2070 struct sockaddr *dst, u_char *desten, struct llentry **lle) 2071 { 2072 struct llentry *ln; 2073 2074 *lle = NULL; 2075 IF_AFDATA_UNLOCK_ASSERT(ifp); 2076 if (m->m_flags & M_MCAST) { 2077 int i; 2078 2079 switch (ifp->if_type) { 2080 case IFT_ETHER: 2081 case IFT_FDDI: 2082 #ifdef IFT_L2VLAN 2083 case IFT_L2VLAN: 2084 #endif 2085 #ifdef IFT_IEEE80211 2086 case IFT_IEEE80211: 2087 #endif 2088 case IFT_BRIDGE: 2089 case IFT_ISO88025: 2090 ETHER_MAP_IPV6_MULTICAST(&SIN6(dst)->sin6_addr, 2091 desten); 2092 return (0); 2093 case IFT_IEEE1394: 2094 /* 2095 * netbsd can use if_broadcastaddr, but we don't do so 2096 * to reduce # of ifdef. 2097 */ 2098 for (i = 0; i < ifp->if_addrlen; i++) 2099 desten[i] = ~0; 2100 return (0); 2101 case IFT_ARCNET: 2102 *desten = 0; 2103 return (0); 2104 default: 2105 m_freem(m); 2106 return (EAFNOSUPPORT); 2107 } 2108 } 2109 2110 2111 /* 2112 * the entry should have been created in nd6_store_lladdr 2113 */ 2114 IF_AFDATA_LOCK(ifp); 2115 ln = lla_lookup(LLTABLE6(ifp), 0, dst); 2116 IF_AFDATA_UNLOCK(ifp); 2117 if ((ln == NULL) || !(ln->la_flags & LLE_VALID)) { 2118 if (ln != NULL) 2119 LLE_RUNLOCK(ln); 2120 /* this could happen, if we could not allocate memory */ 2121 m_freem(m); 2122 return (1); 2123 } 2124 2125 bcopy(&ln->ll_addr, desten, ifp->if_addrlen); 2126 *lle = ln; 2127 LLE_RUNLOCK(ln); 2128 /* 2129 * A *small* use after free race exists here 2130 */ 2131 return (0); 2132 } 2133 2134 static void 2135 clear_llinfo_pqueue(struct llentry *ln) 2136 { 2137 struct mbuf *m_hold, *m_hold_next; 2138 2139 for (m_hold = ln->la_hold; m_hold; m_hold = m_hold_next) { 2140 m_hold_next = m_hold->m_nextpkt; 2141 m_hold->m_nextpkt = NULL; 2142 m_freem(m_hold); 2143 } 2144 2145 ln->la_hold = NULL; 2146 return; 2147 } 2148 2149 static int nd6_sysctl_drlist(SYSCTL_HANDLER_ARGS); 2150 static int nd6_sysctl_prlist(SYSCTL_HANDLER_ARGS); 2151 #ifdef SYSCTL_DECL 2152 SYSCTL_DECL(_net_inet6_icmp6); 2153 #endif 2154 SYSCTL_NODE(_net_inet6_icmp6, ICMPV6CTL_ND6_DRLIST, nd6_drlist, 2155 CTLFLAG_RD, nd6_sysctl_drlist, ""); 2156 SYSCTL_NODE(_net_inet6_icmp6, ICMPV6CTL_ND6_PRLIST, nd6_prlist, 2157 CTLFLAG_RD, nd6_sysctl_prlist, ""); 2158 SYSCTL_VNET_INT(_net_inet6_icmp6, ICMPV6CTL_ND6_MAXQLEN, nd6_maxqueuelen, 2159 CTLFLAG_RW, &VNET_NAME(nd6_maxqueuelen), 1, ""); 2160 2161 static int 2162 nd6_sysctl_drlist(SYSCTL_HANDLER_ARGS) 2163 { 2164 int error; 2165 char buf[1024] __aligned(4); 2166 struct in6_defrouter *d, *de; 2167 struct nd_defrouter *dr; 2168 2169 if (req->newptr) 2170 return EPERM; 2171 error = 0; 2172 2173 for (dr = TAILQ_FIRST(&V_nd_defrouter); dr; 2174 dr = TAILQ_NEXT(dr, dr_entry)) { 2175 d = (struct in6_defrouter *)buf; 2176 de = (struct in6_defrouter *)(buf + sizeof(buf)); 2177 2178 if (d + 1 <= de) { 2179 bzero(d, sizeof(*d)); 2180 d->rtaddr.sin6_family = AF_INET6; 2181 d->rtaddr.sin6_len = sizeof(d->rtaddr); 2182 d->rtaddr.sin6_addr = dr->rtaddr; 2183 error = sa6_recoverscope(&d->rtaddr); 2184 if (error != 0) 2185 return (error); 2186 d->flags = dr->flags; 2187 d->rtlifetime = dr->rtlifetime; 2188 d->expire = dr->expire; 2189 d->if_index = dr->ifp->if_index; 2190 } else 2191 panic("buffer too short"); 2192 2193 error = SYSCTL_OUT(req, buf, sizeof(*d)); 2194 if (error) 2195 break; 2196 } 2197 2198 return (error); 2199 } 2200 2201 static int 2202 nd6_sysctl_prlist(SYSCTL_HANDLER_ARGS) 2203 { 2204 int error; 2205 char buf[1024] __aligned(4); 2206 struct in6_prefix *p, *pe; 2207 struct nd_prefix *pr; 2208 char ip6buf[INET6_ADDRSTRLEN]; 2209 2210 if (req->newptr) 2211 return EPERM; 2212 error = 0; 2213 2214 for (pr = V_nd_prefix.lh_first; pr; pr = pr->ndpr_next) { 2215 u_short advrtrs; 2216 size_t advance; 2217 struct sockaddr_in6 *sin6, *s6; 2218 struct nd_pfxrouter *pfr; 2219 2220 p = (struct in6_prefix *)buf; 2221 pe = (struct in6_prefix *)(buf + sizeof(buf)); 2222 2223 if (p + 1 <= pe) { 2224 bzero(p, sizeof(*p)); 2225 sin6 = (struct sockaddr_in6 *)(p + 1); 2226 2227 p->prefix = pr->ndpr_prefix; 2228 if (sa6_recoverscope(&p->prefix)) { 2229 log(LOG_ERR, 2230 "scope error in prefix list (%s)\n", 2231 ip6_sprintf(ip6buf, &p->prefix.sin6_addr)); 2232 /* XXX: press on... */ 2233 } 2234 p->raflags = pr->ndpr_raf; 2235 p->prefixlen = pr->ndpr_plen; 2236 p->vltime = pr->ndpr_vltime; 2237 p->pltime = pr->ndpr_pltime; 2238 p->if_index = pr->ndpr_ifp->if_index; 2239 if (pr->ndpr_vltime == ND6_INFINITE_LIFETIME) 2240 p->expire = 0; 2241 else { 2242 time_t maxexpire; 2243 2244 /* XXX: we assume time_t is signed. */ 2245 maxexpire = (-1) & 2246 ~((time_t)1 << 2247 ((sizeof(maxexpire) * 8) - 1)); 2248 if (pr->ndpr_vltime < 2249 maxexpire - pr->ndpr_lastupdate) { 2250 p->expire = pr->ndpr_lastupdate + 2251 pr->ndpr_vltime; 2252 } else 2253 p->expire = maxexpire; 2254 } 2255 p->refcnt = pr->ndpr_refcnt; 2256 p->flags = pr->ndpr_stateflags; 2257 p->origin = PR_ORIG_RA; 2258 advrtrs = 0; 2259 for (pfr = pr->ndpr_advrtrs.lh_first; pfr; 2260 pfr = pfr->pfr_next) { 2261 if ((void *)&sin6[advrtrs + 1] > (void *)pe) { 2262 advrtrs++; 2263 continue; 2264 } 2265 s6 = &sin6[advrtrs]; 2266 bzero(s6, sizeof(*s6)); 2267 s6->sin6_family = AF_INET6; 2268 s6->sin6_len = sizeof(*sin6); 2269 s6->sin6_addr = pfr->router->rtaddr; 2270 if (sa6_recoverscope(s6)) { 2271 log(LOG_ERR, 2272 "scope error in " 2273 "prefix list (%s)\n", 2274 ip6_sprintf(ip6buf, 2275 &pfr->router->rtaddr)); 2276 } 2277 advrtrs++; 2278 } 2279 p->advrtrs = advrtrs; 2280 } else 2281 panic("buffer too short"); 2282 2283 advance = sizeof(*p) + sizeof(*sin6) * advrtrs; 2284 error = SYSCTL_OUT(req, buf, advance); 2285 if (error) 2286 break; 2287 } 2288 2289 return (error); 2290 } 2291