nd6.c revision 236021
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: stable/9/sys/netinet6/nd6.c 236021 2012-05-25 19:56:40Z marius $"); 34 35#include "opt_inet.h" 36#include "opt_inet6.h" 37 38#include <sys/param.h> 39#include <sys/systm.h> 40#include <sys/callout.h> 41#include <sys/malloc.h> 42#include <sys/mbuf.h> 43#include <sys/socket.h> 44#include <sys/sockio.h> 45#include <sys/time.h> 46#include <sys/kernel.h> 47#include <sys/protosw.h> 48#include <sys/errno.h> 49#include <sys/syslog.h> 50#include <sys/lock.h> 51#include <sys/rwlock.h> 52#include <sys/queue.h> 53#include <sys/sysctl.h> 54 55#include <net/if.h> 56#include <net/if_arc.h> 57#include <net/if_dl.h> 58#include <net/if_types.h> 59#include <net/iso88025.h> 60#include <net/fddi.h> 61#include <net/route.h> 62#include <net/vnet.h> 63 64#include <netinet/in.h> 65#include <net/if_llatbl.h> 66#define L3_ADDR_SIN6(le) ((struct sockaddr_in6 *) L3_ADDR(le)) 67#include <netinet/if_ether.h> 68#include <netinet6/in6_var.h> 69#include <netinet/ip6.h> 70#include <netinet6/ip6_var.h> 71#include <netinet6/scope6_var.h> 72#include <netinet6/nd6.h> 73#include <netinet6/in6_ifattach.h> 74#include <netinet/icmp6.h> 75#include <netinet6/send.h> 76 77#include <sys/limits.h> 78 79#include <security/mac/mac_framework.h> 80 81#define ND6_SLOWTIMER_INTERVAL (60 * 60) /* 1 hour */ 82#define ND6_RECALC_REACHTM_INTERVAL (60 * 120) /* 2 hours */ 83 84#define SIN6(s) ((struct sockaddr_in6 *)s) 85 86/* timer values */ 87VNET_DEFINE(int, nd6_prune) = 1; /* walk list every 1 seconds */ 88VNET_DEFINE(int, nd6_delay) = 5; /* delay first probe time 5 second */ 89VNET_DEFINE(int, nd6_umaxtries) = 3; /* maximum unicast query */ 90VNET_DEFINE(int, nd6_mmaxtries) = 3; /* maximum multicast query */ 91VNET_DEFINE(int, nd6_useloopback) = 1; /* use loopback interface for 92 * local traffic */ 93VNET_DEFINE(int, nd6_gctimer) = (60 * 60 * 24); /* 1 day: garbage 94 * collection timer */ 95 96/* preventing too many loops in ND option parsing */ 97static VNET_DEFINE(int, nd6_maxndopt) = 10; /* max # of ND options allowed */ 98 99VNET_DEFINE(int, nd6_maxnudhint) = 0; /* max # of subsequent upper 100 * layer hints */ 101static VNET_DEFINE(int, nd6_maxqueuelen) = 1; /* max pkts cached in unresolved 102 * ND entries */ 103#define V_nd6_maxndopt VNET(nd6_maxndopt) 104#define V_nd6_maxqueuelen VNET(nd6_maxqueuelen) 105 106#ifdef ND6_DEBUG 107VNET_DEFINE(int, nd6_debug) = 1; 108#else 109VNET_DEFINE(int, nd6_debug) = 0; 110#endif 111 112/* for debugging? */ 113#if 0 114static int nd6_inuse, nd6_allocated; 115#endif 116 117VNET_DEFINE(struct nd_drhead, nd_defrouter); 118VNET_DEFINE(struct nd_prhead, nd_prefix); 119 120VNET_DEFINE(int, nd6_recalc_reachtm_interval) = ND6_RECALC_REACHTM_INTERVAL; 121#define V_nd6_recalc_reachtm_interval VNET(nd6_recalc_reachtm_interval) 122 123static struct sockaddr_in6 all1_sa; 124 125int (*send_sendso_input_hook)(struct mbuf *, struct ifnet *, int, int); 126 127static int nd6_is_new_addr_neighbor __P((struct sockaddr_in6 *, 128 struct ifnet *)); 129static void nd6_setmtu0(struct ifnet *, struct nd_ifinfo *); 130static void nd6_slowtimo(void *); 131static int regen_tmpaddr(struct in6_ifaddr *); 132static struct llentry *nd6_free(struct llentry *, int); 133static void nd6_llinfo_timer(void *); 134static void clear_llinfo_pqueue(struct llentry *); 135 136static VNET_DEFINE(struct callout, nd6_slowtimo_ch); 137#define V_nd6_slowtimo_ch VNET(nd6_slowtimo_ch) 138 139VNET_DEFINE(struct callout, nd6_timer_ch); 140 141void 142nd6_init(void) 143{ 144 int i; 145 146 LIST_INIT(&V_nd_prefix); 147 148 all1_sa.sin6_family = AF_INET6; 149 all1_sa.sin6_len = sizeof(struct sockaddr_in6); 150 for (i = 0; i < sizeof(all1_sa.sin6_addr); i++) 151 all1_sa.sin6_addr.s6_addr[i] = 0xff; 152 153 /* initialization of the default router list */ 154 TAILQ_INIT(&V_nd_defrouter); 155 156 /* start timer */ 157 callout_init(&V_nd6_slowtimo_ch, 0); 158 callout_reset(&V_nd6_slowtimo_ch, ND6_SLOWTIMER_INTERVAL * hz, 159 nd6_slowtimo, curvnet); 160} 161 162#ifdef VIMAGE 163void 164nd6_destroy() 165{ 166 167 callout_drain(&V_nd6_slowtimo_ch); 168 callout_drain(&V_nd6_timer_ch); 169} 170#endif 171 172struct nd_ifinfo * 173nd6_ifattach(struct ifnet *ifp) 174{ 175 struct nd_ifinfo *nd; 176 177 nd = (struct nd_ifinfo *)malloc(sizeof(*nd), M_IP6NDP, M_WAITOK); 178 bzero(nd, sizeof(*nd)); 179 180 nd->initialized = 1; 181 182 nd->chlim = IPV6_DEFHLIM; 183 nd->basereachable = REACHABLE_TIME; 184 nd->reachable = ND_COMPUTE_RTIME(nd->basereachable); 185 nd->retrans = RETRANS_TIMER; 186 187 nd->flags = ND6_IFF_PERFORMNUD; 188 189 /* A loopback interface always has ND6_IFF_AUTO_LINKLOCAL. */ 190 if (V_ip6_auto_linklocal || (ifp->if_flags & IFF_LOOPBACK)) 191 nd->flags |= ND6_IFF_AUTO_LINKLOCAL; 192 193 /* A loopback interface does not need to accept RTADV. */ 194 if (V_ip6_accept_rtadv && !(ifp->if_flags & IFF_LOOPBACK)) 195 nd->flags |= ND6_IFF_ACCEPT_RTADV; 196 if (V_ip6_no_radr && !(ifp->if_flags & IFF_LOOPBACK)) 197 nd->flags |= ND6_IFF_NO_RADR; 198 199 /* XXX: we cannot call nd6_setmtu since ifp is not fully initialized */ 200 nd6_setmtu0(ifp, nd); 201 202 return nd; 203} 204 205void 206nd6_ifdetach(struct nd_ifinfo *nd) 207{ 208 209 free(nd, M_IP6NDP); 210} 211 212/* 213 * Reset ND level link MTU. This function is called when the physical MTU 214 * changes, which means we might have to adjust the ND level MTU. 215 */ 216void 217nd6_setmtu(struct ifnet *ifp) 218{ 219 220 nd6_setmtu0(ifp, ND_IFINFO(ifp)); 221} 222 223/* XXX todo: do not maintain copy of ifp->if_mtu in ndi->maxmtu */ 224void 225nd6_setmtu0(struct ifnet *ifp, struct nd_ifinfo *ndi) 226{ 227 u_int32_t omaxmtu; 228 229 omaxmtu = ndi->maxmtu; 230 231 switch (ifp->if_type) { 232 case IFT_ARCNET: 233 ndi->maxmtu = MIN(ARC_PHDS_MAXMTU, ifp->if_mtu); /* RFC2497 */ 234 break; 235 case IFT_FDDI: 236 ndi->maxmtu = MIN(FDDIIPMTU, ifp->if_mtu); /* RFC2467 */ 237 break; 238 case IFT_ISO88025: 239 ndi->maxmtu = MIN(ISO88025_MAX_MTU, ifp->if_mtu); 240 break; 241 default: 242 ndi->maxmtu = ifp->if_mtu; 243 break; 244 } 245 246 /* 247 * Decreasing the interface MTU under IPV6 minimum MTU may cause 248 * undesirable situation. We thus notify the operator of the change 249 * explicitly. The check for omaxmtu is necessary to restrict the 250 * log to the case of changing the MTU, not initializing it. 251 */ 252 if (omaxmtu >= IPV6_MMTU && ndi->maxmtu < IPV6_MMTU) { 253 log(LOG_NOTICE, "nd6_setmtu0: " 254 "new link MTU on %s (%lu) is too small for IPv6\n", 255 if_name(ifp), (unsigned long)ndi->maxmtu); 256 } 257 258 if (ndi->maxmtu > V_in6_maxmtu) 259 in6_setmaxmtu(); /* check all interfaces just in case */ 260 261} 262 263void 264nd6_option_init(void *opt, int icmp6len, union nd_opts *ndopts) 265{ 266 267 bzero(ndopts, sizeof(*ndopts)); 268 ndopts->nd_opts_search = (struct nd_opt_hdr *)opt; 269 ndopts->nd_opts_last 270 = (struct nd_opt_hdr *)(((u_char *)opt) + icmp6len); 271 272 if (icmp6len == 0) { 273 ndopts->nd_opts_done = 1; 274 ndopts->nd_opts_search = NULL; 275 } 276} 277 278/* 279 * Take one ND option. 280 */ 281struct nd_opt_hdr * 282nd6_option(union nd_opts *ndopts) 283{ 284 struct nd_opt_hdr *nd_opt; 285 int olen; 286 287 if (ndopts == NULL) 288 panic("ndopts == NULL in nd6_option"); 289 if (ndopts->nd_opts_last == NULL) 290 panic("uninitialized ndopts in nd6_option"); 291 if (ndopts->nd_opts_search == NULL) 292 return NULL; 293 if (ndopts->nd_opts_done) 294 return NULL; 295 296 nd_opt = ndopts->nd_opts_search; 297 298 /* make sure nd_opt_len is inside the buffer */ 299 if ((caddr_t)&nd_opt->nd_opt_len >= (caddr_t)ndopts->nd_opts_last) { 300 bzero(ndopts, sizeof(*ndopts)); 301 return NULL; 302 } 303 304 olen = nd_opt->nd_opt_len << 3; 305 if (olen == 0) { 306 /* 307 * Message validation requires that all included 308 * options have a length that is greater than zero. 309 */ 310 bzero(ndopts, sizeof(*ndopts)); 311 return NULL; 312 } 313 314 ndopts->nd_opts_search = (struct nd_opt_hdr *)((caddr_t)nd_opt + olen); 315 if (ndopts->nd_opts_search > ndopts->nd_opts_last) { 316 /* option overruns the end of buffer, invalid */ 317 bzero(ndopts, sizeof(*ndopts)); 318 return NULL; 319 } else if (ndopts->nd_opts_search == ndopts->nd_opts_last) { 320 /* reached the end of options chain */ 321 ndopts->nd_opts_done = 1; 322 ndopts->nd_opts_search = NULL; 323 } 324 return nd_opt; 325} 326 327/* 328 * Parse multiple ND options. 329 * This function is much easier to use, for ND routines that do not need 330 * multiple options of the same type. 331 */ 332int 333nd6_options(union nd_opts *ndopts) 334{ 335 struct nd_opt_hdr *nd_opt; 336 int i = 0; 337 338 if (ndopts == NULL) 339 panic("ndopts == NULL in nd6_options"); 340 if (ndopts->nd_opts_last == NULL) 341 panic("uninitialized ndopts in nd6_options"); 342 if (ndopts->nd_opts_search == NULL) 343 return 0; 344 345 while (1) { 346 nd_opt = nd6_option(ndopts); 347 if (nd_opt == NULL && ndopts->nd_opts_last == NULL) { 348 /* 349 * Message validation requires that all included 350 * options have a length that is greater than zero. 351 */ 352 ICMP6STAT_INC(icp6s_nd_badopt); 353 bzero(ndopts, sizeof(*ndopts)); 354 return -1; 355 } 356 357 if (nd_opt == NULL) 358 goto skip1; 359 360 switch (nd_opt->nd_opt_type) { 361 case ND_OPT_SOURCE_LINKADDR: 362 case ND_OPT_TARGET_LINKADDR: 363 case ND_OPT_MTU: 364 case ND_OPT_REDIRECTED_HEADER: 365 if (ndopts->nd_opt_array[nd_opt->nd_opt_type]) { 366 nd6log((LOG_INFO, 367 "duplicated ND6 option found (type=%d)\n", 368 nd_opt->nd_opt_type)); 369 /* XXX bark? */ 370 } else { 371 ndopts->nd_opt_array[nd_opt->nd_opt_type] 372 = nd_opt; 373 } 374 break; 375 case ND_OPT_PREFIX_INFORMATION: 376 if (ndopts->nd_opt_array[nd_opt->nd_opt_type] == 0) { 377 ndopts->nd_opt_array[nd_opt->nd_opt_type] 378 = nd_opt; 379 } 380 ndopts->nd_opts_pi_end = 381 (struct nd_opt_prefix_info *)nd_opt; 382 break; 383 /* What about ND_OPT_ROUTE_INFO? RFC 4191 */ 384 case ND_OPT_RDNSS: /* RFC 6106 */ 385 case ND_OPT_DNSSL: /* RFC 6106 */ 386 /* 387 * Silently ignore options we know and do not care about 388 * in the kernel. 389 */ 390 break; 391 default: 392 /* 393 * Unknown options must be silently ignored, 394 * to accomodate future extension to the protocol. 395 */ 396 nd6log((LOG_DEBUG, 397 "nd6_options: unsupported option %d - " 398 "option ignored\n", nd_opt->nd_opt_type)); 399 } 400 401skip1: 402 i++; 403 if (i > V_nd6_maxndopt) { 404 ICMP6STAT_INC(icp6s_nd_toomanyopt); 405 nd6log((LOG_INFO, "too many loop in nd opt\n")); 406 break; 407 } 408 409 if (ndopts->nd_opts_done) 410 break; 411 } 412 413 return 0; 414} 415 416/* 417 * ND6 timer routine to handle ND6 entries 418 */ 419void 420nd6_llinfo_settimer_locked(struct llentry *ln, long tick) 421{ 422 int canceled; 423 424 LLE_WLOCK_ASSERT(ln); 425 426 if (tick < 0) { 427 ln->la_expire = 0; 428 ln->ln_ntick = 0; 429 canceled = callout_stop(&ln->ln_timer_ch); 430 } else { 431 ln->la_expire = time_second + tick / hz; 432 LLE_ADDREF(ln); 433 if (tick > INT_MAX) { 434 ln->ln_ntick = tick - INT_MAX; 435 canceled = callout_reset(&ln->ln_timer_ch, INT_MAX, 436 nd6_llinfo_timer, ln); 437 } else { 438 ln->ln_ntick = 0; 439 canceled = callout_reset(&ln->ln_timer_ch, tick, 440 nd6_llinfo_timer, ln); 441 } 442 } 443 if (canceled) 444 LLE_REMREF(ln); 445} 446 447void 448nd6_llinfo_settimer(struct llentry *ln, long tick) 449{ 450 451 LLE_WLOCK(ln); 452 nd6_llinfo_settimer_locked(ln, tick); 453 LLE_WUNLOCK(ln); 454} 455 456static void 457nd6_llinfo_timer(void *arg) 458{ 459 struct llentry *ln; 460 struct in6_addr *dst; 461 struct ifnet *ifp; 462 struct nd_ifinfo *ndi = NULL; 463 464 KASSERT(arg != NULL, ("%s: arg NULL", __func__)); 465 ln = (struct llentry *)arg; 466 LLE_WLOCK_ASSERT(ln); 467 ifp = ln->lle_tbl->llt_ifp; 468 469 CURVNET_SET(ifp->if_vnet); 470 471 if (ln->ln_ntick > 0) { 472 if (ln->ln_ntick > INT_MAX) { 473 ln->ln_ntick -= INT_MAX; 474 nd6_llinfo_settimer_locked(ln, INT_MAX); 475 } else { 476 ln->ln_ntick = 0; 477 nd6_llinfo_settimer_locked(ln, ln->ln_ntick); 478 } 479 goto done; 480 } 481 482 ndi = ND_IFINFO(ifp); 483 dst = &L3_ADDR_SIN6(ln)->sin6_addr; 484 if (ln->la_flags & LLE_STATIC) { 485 goto done; 486 } 487 488 if (ln->la_flags & LLE_DELETED) { 489 (void)nd6_free(ln, 0); 490 ln = NULL; 491 goto done; 492 } 493 494 switch (ln->ln_state) { 495 case ND6_LLINFO_INCOMPLETE: 496 if (ln->la_asked < V_nd6_mmaxtries) { 497 ln->la_asked++; 498 nd6_llinfo_settimer_locked(ln, (long)ndi->retrans * hz / 1000); 499 LLE_WUNLOCK(ln); 500 nd6_ns_output(ifp, NULL, dst, ln, 0); 501 LLE_WLOCK(ln); 502 } else { 503 struct mbuf *m = ln->la_hold; 504 if (m) { 505 struct mbuf *m0; 506 507 /* 508 * assuming every packet in la_hold has the 509 * same IP header. Send error after unlock. 510 */ 511 m0 = m->m_nextpkt; 512 m->m_nextpkt = NULL; 513 ln->la_hold = m0; 514 clear_llinfo_pqueue(ln); 515 } 516 (void)nd6_free(ln, 0); 517 ln = NULL; 518 if (m != NULL) 519 icmp6_error2(m, ICMP6_DST_UNREACH, 520 ICMP6_DST_UNREACH_ADDR, 0, ifp); 521 } 522 break; 523 case ND6_LLINFO_REACHABLE: 524 if (!ND6_LLINFO_PERMANENT(ln)) { 525 ln->ln_state = ND6_LLINFO_STALE; 526 nd6_llinfo_settimer_locked(ln, (long)V_nd6_gctimer * hz); 527 } 528 break; 529 530 case ND6_LLINFO_STALE: 531 /* Garbage Collection(RFC 2461 5.3) */ 532 if (!ND6_LLINFO_PERMANENT(ln)) { 533 (void)nd6_free(ln, 1); 534 ln = NULL; 535 } 536 break; 537 538 case ND6_LLINFO_DELAY: 539 if (ndi && (ndi->flags & ND6_IFF_PERFORMNUD) != 0) { 540 /* We need NUD */ 541 ln->la_asked = 1; 542 ln->ln_state = ND6_LLINFO_PROBE; 543 nd6_llinfo_settimer_locked(ln, (long)ndi->retrans * hz / 1000); 544 LLE_WUNLOCK(ln); 545 nd6_ns_output(ifp, dst, dst, ln, 0); 546 LLE_WLOCK(ln); 547 } else { 548 ln->ln_state = ND6_LLINFO_STALE; /* XXX */ 549 nd6_llinfo_settimer_locked(ln, (long)V_nd6_gctimer * hz); 550 } 551 break; 552 case ND6_LLINFO_PROBE: 553 if (ln->la_asked < V_nd6_umaxtries) { 554 ln->la_asked++; 555 nd6_llinfo_settimer_locked(ln, (long)ndi->retrans * hz / 1000); 556 LLE_WUNLOCK(ln); 557 nd6_ns_output(ifp, dst, dst, ln, 0); 558 LLE_WLOCK(ln); 559 } else { 560 (void)nd6_free(ln, 0); 561 ln = NULL; 562 } 563 break; 564 default: 565 panic("%s: paths in a dark night can be confusing: %d", 566 __func__, ln->ln_state); 567 } 568done: 569 if (ln != NULL) 570 LLE_FREE_LOCKED(ln); 571 CURVNET_RESTORE(); 572} 573 574 575/* 576 * ND6 timer routine to expire default route list and prefix list 577 */ 578void 579nd6_timer(void *arg) 580{ 581 CURVNET_SET((struct vnet *) arg); 582 int s; 583 struct nd_defrouter *dr, *ndr; 584 struct nd_prefix *pr, *npr; 585 struct in6_ifaddr *ia6, *nia6; 586 587 callout_reset(&V_nd6_timer_ch, V_nd6_prune * hz, 588 nd6_timer, curvnet); 589 590 /* expire default router list */ 591 s = splnet(); 592 TAILQ_FOREACH_SAFE(dr, &V_nd_defrouter, dr_entry, ndr) { 593 if (dr->expire && dr->expire < time_second) 594 defrtrlist_del(dr); 595 } 596 597 /* 598 * expire interface addresses. 599 * in the past the loop was inside prefix expiry processing. 600 * However, from a stricter speci-confrmance standpoint, we should 601 * rather separate address lifetimes and prefix lifetimes. 602 * 603 * XXXRW: in6_ifaddrhead locking. 604 */ 605 addrloop: 606 TAILQ_FOREACH_SAFE(ia6, &V_in6_ifaddrhead, ia_link, nia6) { 607 /* check address lifetime */ 608 if (IFA6_IS_INVALID(ia6)) { 609 int regen = 0; 610 611 /* 612 * If the expiring address is temporary, try 613 * regenerating a new one. This would be useful when 614 * we suspended a laptop PC, then turned it on after a 615 * period that could invalidate all temporary 616 * addresses. Although we may have to restart the 617 * loop (see below), it must be after purging the 618 * address. Otherwise, we'd see an infinite loop of 619 * regeneration. 620 */ 621 if (V_ip6_use_tempaddr && 622 (ia6->ia6_flags & IN6_IFF_TEMPORARY) != 0) { 623 if (regen_tmpaddr(ia6) == 0) 624 regen = 1; 625 } 626 627 in6_purgeaddr(&ia6->ia_ifa); 628 629 if (regen) 630 goto addrloop; /* XXX: see below */ 631 } else if (IFA6_IS_DEPRECATED(ia6)) { 632 int oldflags = ia6->ia6_flags; 633 634 ia6->ia6_flags |= IN6_IFF_DEPRECATED; 635 636 /* 637 * If a temporary address has just become deprecated, 638 * regenerate a new one if possible. 639 */ 640 if (V_ip6_use_tempaddr && 641 (ia6->ia6_flags & IN6_IFF_TEMPORARY) != 0 && 642 (oldflags & IN6_IFF_DEPRECATED) == 0) { 643 644 if (regen_tmpaddr(ia6) == 0) { 645 /* 646 * A new temporary address is 647 * generated. 648 * XXX: this means the address chain 649 * has changed while we are still in 650 * the loop. Although the change 651 * would not cause disaster (because 652 * it's not a deletion, but an 653 * addition,) we'd rather restart the 654 * loop just for safety. Or does this 655 * significantly reduce performance?? 656 */ 657 goto addrloop; 658 } 659 } 660 } else { 661 /* 662 * A new RA might have made a deprecated address 663 * preferred. 664 */ 665 ia6->ia6_flags &= ~IN6_IFF_DEPRECATED; 666 } 667 } 668 669 /* expire prefix list */ 670 LIST_FOREACH_SAFE(pr, &V_nd_prefix, ndpr_entry, npr) { 671 /* 672 * check prefix lifetime. 673 * since pltime is just for autoconf, pltime processing for 674 * prefix is not necessary. 675 */ 676 if (pr->ndpr_vltime != ND6_INFINITE_LIFETIME && 677 time_second - pr->ndpr_lastupdate > pr->ndpr_vltime) { 678 679 /* 680 * address expiration and prefix expiration are 681 * separate. NEVER perform in6_purgeaddr here. 682 */ 683 prelist_remove(pr); 684 } 685 } 686 splx(s); 687 CURVNET_RESTORE(); 688} 689 690/* 691 * ia6 - deprecated/invalidated temporary address 692 */ 693static int 694regen_tmpaddr(struct in6_ifaddr *ia6) 695{ 696 struct ifaddr *ifa; 697 struct ifnet *ifp; 698 struct in6_ifaddr *public_ifa6 = NULL; 699 700 ifp = ia6->ia_ifa.ifa_ifp; 701 IF_ADDR_RLOCK(ifp); 702 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 703 struct in6_ifaddr *it6; 704 705 if (ifa->ifa_addr->sa_family != AF_INET6) 706 continue; 707 708 it6 = (struct in6_ifaddr *)ifa; 709 710 /* ignore no autoconf addresses. */ 711 if ((it6->ia6_flags & IN6_IFF_AUTOCONF) == 0) 712 continue; 713 714 /* ignore autoconf addresses with different prefixes. */ 715 if (it6->ia6_ndpr == NULL || it6->ia6_ndpr != ia6->ia6_ndpr) 716 continue; 717 718 /* 719 * Now we are looking at an autoconf address with the same 720 * prefix as ours. If the address is temporary and is still 721 * preferred, do not create another one. It would be rare, but 722 * could happen, for example, when we resume a laptop PC after 723 * a long period. 724 */ 725 if ((it6->ia6_flags & IN6_IFF_TEMPORARY) != 0 && 726 !IFA6_IS_DEPRECATED(it6)) { 727 public_ifa6 = NULL; 728 break; 729 } 730 731 /* 732 * This is a public autoconf address that has the same prefix 733 * as ours. If it is preferred, keep it. We can't break the 734 * loop here, because there may be a still-preferred temporary 735 * address with the prefix. 736 */ 737 if (!IFA6_IS_DEPRECATED(it6)) 738 public_ifa6 = it6; 739 740 if (public_ifa6 != NULL) 741 ifa_ref(&public_ifa6->ia_ifa); 742 } 743 IF_ADDR_RUNLOCK(ifp); 744 745 if (public_ifa6 != NULL) { 746 int e; 747 748 if ((e = in6_tmpifadd(public_ifa6, 0, 0)) != 0) { 749 ifa_free(&public_ifa6->ia_ifa); 750 log(LOG_NOTICE, "regen_tmpaddr: failed to create a new" 751 " tmp addr,errno=%d\n", e); 752 return (-1); 753 } 754 ifa_free(&public_ifa6->ia_ifa); 755 return (0); 756 } 757 758 return (-1); 759} 760 761/* 762 * Nuke neighbor cache/prefix/default router management table, right before 763 * ifp goes away. 764 */ 765void 766nd6_purge(struct ifnet *ifp) 767{ 768 struct nd_defrouter *dr, *ndr; 769 struct nd_prefix *pr, *npr; 770 771 /* 772 * Nuke default router list entries toward ifp. 773 * We defer removal of default router list entries that is installed 774 * in the routing table, in order to keep additional side effects as 775 * small as possible. 776 */ 777 TAILQ_FOREACH_SAFE(dr, &V_nd_defrouter, dr_entry, ndr) { 778 if (dr->installed) 779 continue; 780 781 if (dr->ifp == ifp) 782 defrtrlist_del(dr); 783 } 784 785 TAILQ_FOREACH_SAFE(dr, &V_nd_defrouter, dr_entry, ndr) { 786 if (!dr->installed) 787 continue; 788 789 if (dr->ifp == ifp) 790 defrtrlist_del(dr); 791 } 792 793 /* Nuke prefix list entries toward ifp */ 794 LIST_FOREACH_SAFE(pr, &V_nd_prefix, ndpr_entry, npr) { 795 if (pr->ndpr_ifp == ifp) { 796 /* 797 * Because if_detach() does *not* release prefixes 798 * while purging addresses the reference count will 799 * still be above zero. We therefore reset it to 800 * make sure that the prefix really gets purged. 801 */ 802 pr->ndpr_refcnt = 0; 803 804 /* 805 * Previously, pr->ndpr_addr is removed as well, 806 * but I strongly believe we don't have to do it. 807 * nd6_purge() is only called from in6_ifdetach(), 808 * which removes all the associated interface addresses 809 * by itself. 810 * (jinmei@kame.net 20010129) 811 */ 812 prelist_remove(pr); 813 } 814 } 815 816 /* cancel default outgoing interface setting */ 817 if (V_nd6_defifindex == ifp->if_index) 818 nd6_setdefaultiface(0); 819 820 if (ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV) { 821 /* Refresh default router list. */ 822 defrouter_select(); 823 } 824 825 /* XXXXX 826 * We do not nuke the neighbor cache entries here any more 827 * because the neighbor cache is kept in if_afdata[AF_INET6]. 828 * nd6_purge() is invoked by in6_ifdetach() which is called 829 * from if_detach() where everything gets purged. So let 830 * in6_domifdetach() do the actual L2 table purging work. 831 */ 832} 833 834/* 835 * the caller acquires and releases the lock on the lltbls 836 * Returns the llentry locked 837 */ 838struct llentry * 839nd6_lookup(struct in6_addr *addr6, int flags, struct ifnet *ifp) 840{ 841 struct sockaddr_in6 sin6; 842 struct llentry *ln; 843 int llflags; 844 845 bzero(&sin6, sizeof(sin6)); 846 sin6.sin6_len = sizeof(struct sockaddr_in6); 847 sin6.sin6_family = AF_INET6; 848 sin6.sin6_addr = *addr6; 849 850 IF_AFDATA_LOCK_ASSERT(ifp); 851 852 llflags = 0; 853 if (flags & ND6_CREATE) 854 llflags |= LLE_CREATE; 855 if (flags & ND6_EXCLUSIVE) 856 llflags |= LLE_EXCLUSIVE; 857 858 ln = lla_lookup(LLTABLE6(ifp), llflags, (struct sockaddr *)&sin6); 859 if ((ln != NULL) && (llflags & LLE_CREATE)) 860 ln->ln_state = ND6_LLINFO_NOSTATE; 861 862 return (ln); 863} 864 865/* 866 * Test whether a given IPv6 address is a neighbor or not, ignoring 867 * the actual neighbor cache. The neighbor cache is ignored in order 868 * to not reenter the routing code from within itself. 869 */ 870static int 871nd6_is_new_addr_neighbor(struct sockaddr_in6 *addr, struct ifnet *ifp) 872{ 873 struct nd_prefix *pr; 874 struct ifaddr *dstaddr; 875 876 /* 877 * A link-local address is always a neighbor. 878 * XXX: a link does not necessarily specify a single interface. 879 */ 880 if (IN6_IS_ADDR_LINKLOCAL(&addr->sin6_addr)) { 881 struct sockaddr_in6 sin6_copy; 882 u_int32_t zone; 883 884 /* 885 * We need sin6_copy since sa6_recoverscope() may modify the 886 * content (XXX). 887 */ 888 sin6_copy = *addr; 889 if (sa6_recoverscope(&sin6_copy)) 890 return (0); /* XXX: should be impossible */ 891 if (in6_setscope(&sin6_copy.sin6_addr, ifp, &zone)) 892 return (0); 893 if (sin6_copy.sin6_scope_id == zone) 894 return (1); 895 else 896 return (0); 897 } 898 899 /* 900 * If the address matches one of our addresses, 901 * it should be a neighbor. 902 * If the address matches one of our on-link prefixes, it should be a 903 * neighbor. 904 */ 905 LIST_FOREACH(pr, &V_nd_prefix, ndpr_entry) { 906 if (pr->ndpr_ifp != ifp) 907 continue; 908 909 if (!(pr->ndpr_stateflags & NDPRF_ONLINK)) { 910 struct rtentry *rt; 911 912 /* Always use the default FIB here. */ 913 rt = in6_rtalloc1((struct sockaddr *)&pr->ndpr_prefix, 914 0, 0, RT_DEFAULT_FIB); 915 if (rt == NULL) 916 continue; 917 /* 918 * This is the case where multiple interfaces 919 * have the same prefix, but only one is installed 920 * into the routing table and that prefix entry 921 * is not the one being examined here. In the case 922 * where RADIX_MPATH is enabled, multiple route 923 * entries (of the same rt_key value) will be 924 * installed because the interface addresses all 925 * differ. 926 */ 927 if (!IN6_ARE_ADDR_EQUAL(&pr->ndpr_prefix.sin6_addr, 928 &((struct sockaddr_in6 *)rt_key(rt))->sin6_addr)) { 929 RTFREE_LOCKED(rt); 930 continue; 931 } 932 RTFREE_LOCKED(rt); 933 } 934 935 if (IN6_ARE_MASKED_ADDR_EQUAL(&pr->ndpr_prefix.sin6_addr, 936 &addr->sin6_addr, &pr->ndpr_mask)) 937 return (1); 938 } 939 940 /* 941 * If the address is assigned on the node of the other side of 942 * a p2p interface, the address should be a neighbor. 943 */ 944 dstaddr = ifa_ifwithdstaddr((struct sockaddr *)addr); 945 if (dstaddr != NULL) { 946 if (dstaddr->ifa_ifp == ifp) { 947 ifa_free(dstaddr); 948 return (1); 949 } 950 ifa_free(dstaddr); 951 } 952 953 /* 954 * If the default router list is empty, all addresses are regarded 955 * as on-link, and thus, as a neighbor. 956 */ 957 if (ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV && 958 TAILQ_EMPTY(&V_nd_defrouter) && 959 V_nd6_defifindex == ifp->if_index) { 960 return (1); 961 } 962 963 return (0); 964} 965 966 967/* 968 * Detect if a given IPv6 address identifies a neighbor on a given link. 969 * XXX: should take care of the destination of a p2p link? 970 */ 971int 972nd6_is_addr_neighbor(struct sockaddr_in6 *addr, struct ifnet *ifp) 973{ 974 struct llentry *lle; 975 int rc = 0; 976 977 IF_AFDATA_UNLOCK_ASSERT(ifp); 978 if (nd6_is_new_addr_neighbor(addr, ifp)) 979 return (1); 980 981 /* 982 * Even if the address matches none of our addresses, it might be 983 * in the neighbor cache. 984 */ 985 IF_AFDATA_LOCK(ifp); 986 if ((lle = nd6_lookup(&addr->sin6_addr, 0, ifp)) != NULL) { 987 LLE_RUNLOCK(lle); 988 rc = 1; 989 } 990 IF_AFDATA_UNLOCK(ifp); 991 return (rc); 992} 993 994/* 995 * Free an nd6 llinfo entry. 996 * Since the function would cause significant changes in the kernel, DO NOT 997 * make it global, unless you have a strong reason for the change, and are sure 998 * that the change is safe. 999 */ 1000static struct llentry * 1001nd6_free(struct llentry *ln, int gc) 1002{ 1003 struct llentry *next; 1004 struct nd_defrouter *dr; 1005 struct ifnet *ifp; 1006 1007 LLE_WLOCK_ASSERT(ln); 1008 1009 /* 1010 * we used to have pfctlinput(PRC_HOSTDEAD) here. 1011 * even though it is not harmful, it was not really necessary. 1012 */ 1013 1014 /* cancel timer */ 1015 nd6_llinfo_settimer_locked(ln, -1); 1016 1017 ifp = ln->lle_tbl->llt_ifp; 1018 1019 if (ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV) { 1020 dr = defrouter_lookup(&L3_ADDR_SIN6(ln)->sin6_addr, ifp); 1021 1022 if (dr != NULL && dr->expire && 1023 ln->ln_state == ND6_LLINFO_STALE && gc) { 1024 /* 1025 * If the reason for the deletion is just garbage 1026 * collection, and the neighbor is an active default 1027 * router, do not delete it. Instead, reset the GC 1028 * timer using the router's lifetime. 1029 * Simply deleting the entry would affect default 1030 * router selection, which is not necessarily a good 1031 * thing, especially when we're using router preference 1032 * values. 1033 * XXX: the check for ln_state would be redundant, 1034 * but we intentionally keep it just in case. 1035 */ 1036 if (dr->expire > time_second) 1037 nd6_llinfo_settimer_locked(ln, 1038 (dr->expire - time_second) * hz); 1039 else 1040 nd6_llinfo_settimer_locked(ln, 1041 (long)V_nd6_gctimer * hz); 1042 1043 next = LIST_NEXT(ln, lle_next); 1044 LLE_REMREF(ln); 1045 LLE_WUNLOCK(ln); 1046 return (next); 1047 } 1048 1049 if (dr) { 1050 /* 1051 * Unreachablity of a router might affect the default 1052 * router selection and on-link detection of advertised 1053 * prefixes. 1054 */ 1055 1056 /* 1057 * Temporarily fake the state to choose a new default 1058 * router and to perform on-link determination of 1059 * prefixes correctly. 1060 * Below the state will be set correctly, 1061 * or the entry itself will be deleted. 1062 */ 1063 ln->ln_state = ND6_LLINFO_INCOMPLETE; 1064 } 1065 1066 if (ln->ln_router || dr) { 1067 1068 /* 1069 * We need to unlock to avoid a LOR with rt6_flush() with the 1070 * rnh and for the calls to pfxlist_onlink_check() and 1071 * defrouter_select() in the block further down for calls 1072 * into nd6_lookup(). We still hold a ref. 1073 */ 1074 LLE_WUNLOCK(ln); 1075 1076 /* 1077 * rt6_flush must be called whether or not the neighbor 1078 * is in the Default Router List. 1079 * See a corresponding comment in nd6_na_input(). 1080 */ 1081 rt6_flush(&L3_ADDR_SIN6(ln)->sin6_addr, ifp); 1082 } 1083 1084 if (dr) { 1085 /* 1086 * Since defrouter_select() does not affect the 1087 * on-link determination and MIP6 needs the check 1088 * before the default router selection, we perform 1089 * the check now. 1090 */ 1091 pfxlist_onlink_check(); 1092 1093 /* 1094 * Refresh default router list. 1095 */ 1096 defrouter_select(); 1097 } 1098 1099 if (ln->ln_router || dr) 1100 LLE_WLOCK(ln); 1101 } 1102 1103 /* 1104 * Before deleting the entry, remember the next entry as the 1105 * return value. We need this because pfxlist_onlink_check() above 1106 * might have freed other entries (particularly the old next entry) as 1107 * a side effect (XXX). 1108 */ 1109 next = LIST_NEXT(ln, lle_next); 1110 1111 /* 1112 * Save to unlock. We still hold an extra reference and will not 1113 * free(9) in llentry_free() if someone else holds one as well. 1114 */ 1115 LLE_WUNLOCK(ln); 1116 IF_AFDATA_LOCK(ifp); 1117 LLE_WLOCK(ln); 1118 LLE_REMREF(ln); 1119 llentry_free(ln); 1120 IF_AFDATA_UNLOCK(ifp); 1121 1122 return (next); 1123} 1124 1125/* 1126 * Upper-layer reachability hint for Neighbor Unreachability Detection. 1127 * 1128 * XXX cost-effective methods? 1129 */ 1130void 1131nd6_nud_hint(struct rtentry *rt, struct in6_addr *dst6, int force) 1132{ 1133 struct llentry *ln; 1134 struct ifnet *ifp; 1135 1136 if ((dst6 == NULL) || (rt == NULL)) 1137 return; 1138 1139 ifp = rt->rt_ifp; 1140 IF_AFDATA_LOCK(ifp); 1141 ln = nd6_lookup(dst6, ND6_EXCLUSIVE, NULL); 1142 IF_AFDATA_UNLOCK(ifp); 1143 if (ln == NULL) 1144 return; 1145 1146 if (ln->ln_state < ND6_LLINFO_REACHABLE) 1147 goto done; 1148 1149 /* 1150 * if we get upper-layer reachability confirmation many times, 1151 * it is possible we have false information. 1152 */ 1153 if (!force) { 1154 ln->ln_byhint++; 1155 if (ln->ln_byhint > V_nd6_maxnudhint) { 1156 goto done; 1157 } 1158 } 1159 1160 ln->ln_state = ND6_LLINFO_REACHABLE; 1161 if (!ND6_LLINFO_PERMANENT(ln)) { 1162 nd6_llinfo_settimer_locked(ln, 1163 (long)ND_IFINFO(rt->rt_ifp)->reachable * hz); 1164 } 1165done: 1166 LLE_WUNLOCK(ln); 1167} 1168 1169 1170/* 1171 * Rejuvenate this function for routing operations related 1172 * processing. 1173 */ 1174void 1175nd6_rtrequest(int req, struct rtentry *rt, struct rt_addrinfo *info) 1176{ 1177 struct sockaddr_in6 *gateway = (struct sockaddr_in6 *)rt->rt_gateway; 1178 struct nd_defrouter *dr; 1179 struct ifnet *ifp = rt->rt_ifp; 1180 1181 RT_LOCK_ASSERT(rt); 1182 1183 switch (req) { 1184 case RTM_ADD: 1185 break; 1186 1187 case RTM_DELETE: 1188 if (!ifp) 1189 return; 1190 /* 1191 * Only indirect routes are interesting. 1192 */ 1193 if ((rt->rt_flags & RTF_GATEWAY) == 0) 1194 return; 1195 /* 1196 * check for default route 1197 */ 1198 if (IN6_ARE_ADDR_EQUAL(&in6addr_any, 1199 &SIN6(rt_key(rt))->sin6_addr)) { 1200 1201 dr = defrouter_lookup(&gateway->sin6_addr, ifp); 1202 if (dr != NULL) 1203 dr->installed = 0; 1204 } 1205 break; 1206 } 1207} 1208 1209 1210int 1211nd6_ioctl(u_long cmd, caddr_t data, struct ifnet *ifp) 1212{ 1213 struct in6_drlist *drl = (struct in6_drlist *)data; 1214 struct in6_oprlist *oprl = (struct in6_oprlist *)data; 1215 struct in6_ndireq *ndi = (struct in6_ndireq *)data; 1216 struct in6_nbrinfo *nbi = (struct in6_nbrinfo *)data; 1217 struct in6_ndifreq *ndif = (struct in6_ndifreq *)data; 1218 struct nd_defrouter *dr; 1219 struct nd_prefix *pr; 1220 int i = 0, error = 0; 1221 int s; 1222 1223 switch (cmd) { 1224 case SIOCGDRLST_IN6: 1225 /* 1226 * obsolete API, use sysctl under net.inet6.icmp6 1227 */ 1228 bzero(drl, sizeof(*drl)); 1229 s = splnet(); 1230 TAILQ_FOREACH(dr, &V_nd_defrouter, dr_entry) { 1231 if (i >= DRLSTSIZ) 1232 break; 1233 drl->defrouter[i].rtaddr = dr->rtaddr; 1234 in6_clearscope(&drl->defrouter[i].rtaddr); 1235 1236 drl->defrouter[i].flags = dr->flags; 1237 drl->defrouter[i].rtlifetime = dr->rtlifetime; 1238 drl->defrouter[i].expire = dr->expire; 1239 drl->defrouter[i].if_index = dr->ifp->if_index; 1240 i++; 1241 } 1242 splx(s); 1243 break; 1244 case SIOCGPRLST_IN6: 1245 /* 1246 * obsolete API, use sysctl under net.inet6.icmp6 1247 * 1248 * XXX the structure in6_prlist was changed in backward- 1249 * incompatible manner. in6_oprlist is used for SIOCGPRLST_IN6, 1250 * in6_prlist is used for nd6_sysctl() - fill_prlist(). 1251 */ 1252 /* 1253 * XXX meaning of fields, especialy "raflags", is very 1254 * differnet between RA prefix list and RR/static prefix list. 1255 * how about separating ioctls into two? 1256 */ 1257 bzero(oprl, sizeof(*oprl)); 1258 s = splnet(); 1259 LIST_FOREACH(pr, &V_nd_prefix, ndpr_entry) { 1260 struct nd_pfxrouter *pfr; 1261 int j; 1262 1263 if (i >= PRLSTSIZ) 1264 break; 1265 oprl->prefix[i].prefix = pr->ndpr_prefix.sin6_addr; 1266 oprl->prefix[i].raflags = pr->ndpr_raf; 1267 oprl->prefix[i].prefixlen = pr->ndpr_plen; 1268 oprl->prefix[i].vltime = pr->ndpr_vltime; 1269 oprl->prefix[i].pltime = pr->ndpr_pltime; 1270 oprl->prefix[i].if_index = pr->ndpr_ifp->if_index; 1271 if (pr->ndpr_vltime == ND6_INFINITE_LIFETIME) 1272 oprl->prefix[i].expire = 0; 1273 else { 1274 time_t maxexpire; 1275 1276 /* XXX: we assume time_t is signed. */ 1277 maxexpire = (-1) & 1278 ~((time_t)1 << 1279 ((sizeof(maxexpire) * 8) - 1)); 1280 if (pr->ndpr_vltime < 1281 maxexpire - pr->ndpr_lastupdate) { 1282 oprl->prefix[i].expire = 1283 pr->ndpr_lastupdate + 1284 pr->ndpr_vltime; 1285 } else 1286 oprl->prefix[i].expire = maxexpire; 1287 } 1288 1289 j = 0; 1290 LIST_FOREACH(pfr, &pr->ndpr_advrtrs, pfr_entry) { 1291 if (j < DRLSTSIZ) { 1292#define RTRADDR oprl->prefix[i].advrtr[j] 1293 RTRADDR = pfr->router->rtaddr; 1294 in6_clearscope(&RTRADDR); 1295#undef RTRADDR 1296 } 1297 j++; 1298 } 1299 oprl->prefix[i].advrtrs = j; 1300 oprl->prefix[i].origin = PR_ORIG_RA; 1301 1302 i++; 1303 } 1304 splx(s); 1305 1306 break; 1307 case OSIOCGIFINFO_IN6: 1308#define ND ndi->ndi 1309 /* XXX: old ndp(8) assumes a positive value for linkmtu. */ 1310 bzero(&ND, sizeof(ND)); 1311 ND.linkmtu = IN6_LINKMTU(ifp); 1312 ND.maxmtu = ND_IFINFO(ifp)->maxmtu; 1313 ND.basereachable = ND_IFINFO(ifp)->basereachable; 1314 ND.reachable = ND_IFINFO(ifp)->reachable; 1315 ND.retrans = ND_IFINFO(ifp)->retrans; 1316 ND.flags = ND_IFINFO(ifp)->flags; 1317 ND.recalctm = ND_IFINFO(ifp)->recalctm; 1318 ND.chlim = ND_IFINFO(ifp)->chlim; 1319 break; 1320 case SIOCGIFINFO_IN6: 1321 ND = *ND_IFINFO(ifp); 1322 break; 1323 case SIOCSIFINFO_IN6: 1324 /* 1325 * used to change host variables from userland. 1326 * intented for a use on router to reflect RA configurations. 1327 */ 1328 /* 0 means 'unspecified' */ 1329 if (ND.linkmtu != 0) { 1330 if (ND.linkmtu < IPV6_MMTU || 1331 ND.linkmtu > IN6_LINKMTU(ifp)) { 1332 error = EINVAL; 1333 break; 1334 } 1335 ND_IFINFO(ifp)->linkmtu = ND.linkmtu; 1336 } 1337 1338 if (ND.basereachable != 0) { 1339 int obasereachable = ND_IFINFO(ifp)->basereachable; 1340 1341 ND_IFINFO(ifp)->basereachable = ND.basereachable; 1342 if (ND.basereachable != obasereachable) 1343 ND_IFINFO(ifp)->reachable = 1344 ND_COMPUTE_RTIME(ND.basereachable); 1345 } 1346 if (ND.retrans != 0) 1347 ND_IFINFO(ifp)->retrans = ND.retrans; 1348 if (ND.chlim != 0) 1349 ND_IFINFO(ifp)->chlim = ND.chlim; 1350 /* FALLTHROUGH */ 1351 case SIOCSIFINFO_FLAGS: 1352 { 1353 struct ifaddr *ifa; 1354 struct in6_ifaddr *ia; 1355 1356 /* 1357 * Try to clear ifdisabled flag when enabling 1358 * accept_rtadv or auto_linklocal. 1359 */ 1360 if ((ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) && 1361 !(ND.flags & ND6_IFF_IFDISABLED) && 1362 (ND.flags & (ND6_IFF_ACCEPT_RTADV | 1363 ND6_IFF_AUTO_LINKLOCAL))) 1364 ND.flags &= ~ND6_IFF_IFDISABLED; 1365 1366 if ((ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) && 1367 !(ND.flags & ND6_IFF_IFDISABLED)) { 1368 /* ifdisabled 1->0 transision */ 1369 1370 /* 1371 * If the interface is marked as ND6_IFF_IFDISABLED and 1372 * has an link-local address with IN6_IFF_DUPLICATED, 1373 * do not clear ND6_IFF_IFDISABLED. 1374 * See RFC 4862, Section 5.4.5. 1375 */ 1376 int duplicated_linklocal = 0; 1377 1378 IF_ADDR_RLOCK(ifp); 1379 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1380 if (ifa->ifa_addr->sa_family != AF_INET6) 1381 continue; 1382 ia = (struct in6_ifaddr *)ifa; 1383 if ((ia->ia6_flags & IN6_IFF_DUPLICATED) && 1384 IN6_IS_ADDR_LINKLOCAL(IA6_IN6(ia))) { 1385 duplicated_linklocal = 1; 1386 break; 1387 } 1388 } 1389 IF_ADDR_RUNLOCK(ifp); 1390 1391 if (duplicated_linklocal) { 1392 ND.flags |= ND6_IFF_IFDISABLED; 1393 log(LOG_ERR, "Cannot enable an interface" 1394 " with a link-local address marked" 1395 " duplicate.\n"); 1396 } else { 1397 ND_IFINFO(ifp)->flags &= ~ND6_IFF_IFDISABLED; 1398 if (ifp->if_flags & IFF_UP) 1399 in6_if_up(ifp); 1400 } 1401 } else if (!(ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) && 1402 (ND.flags & ND6_IFF_IFDISABLED)) { 1403 /* ifdisabled 0->1 transision */ 1404 /* Mark all IPv6 address as tentative. */ 1405 1406 ND_IFINFO(ifp)->flags |= ND6_IFF_IFDISABLED; 1407 IF_ADDR_RLOCK(ifp); 1408 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1409 if (ifa->ifa_addr->sa_family != AF_INET6) 1410 continue; 1411 ia = (struct in6_ifaddr *)ifa; 1412 ia->ia6_flags |= IN6_IFF_TENTATIVE; 1413 } 1414 IF_ADDR_RUNLOCK(ifp); 1415 } 1416 1417 if (ND.flags & ND6_IFF_AUTO_LINKLOCAL) { 1418 if (!(ND_IFINFO(ifp)->flags & ND6_IFF_AUTO_LINKLOCAL)) { 1419 /* auto_linklocal 0->1 transision */ 1420 1421 /* If no link-local address on ifp, configure */ 1422 ND_IFINFO(ifp)->flags |= ND6_IFF_AUTO_LINKLOCAL; 1423 in6_ifattach(ifp, NULL); 1424 } else if (!(ND.flags & ND6_IFF_IFDISABLED) && 1425 ifp->if_flags & IFF_UP) { 1426 /* 1427 * When the IF already has 1428 * ND6_IFF_AUTO_LINKLOCAL, no link-local 1429 * address is assigned, and IFF_UP, try to 1430 * assign one. 1431 */ 1432 int haslinklocal = 0; 1433 1434 IF_ADDR_RLOCK(ifp); 1435 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1436 if (ifa->ifa_addr->sa_family != AF_INET6) 1437 continue; 1438 ia = (struct in6_ifaddr *)ifa; 1439 if (IN6_IS_ADDR_LINKLOCAL(IA6_IN6(ia))) { 1440 haslinklocal = 1; 1441 break; 1442 } 1443 } 1444 IF_ADDR_RUNLOCK(ifp); 1445 if (!haslinklocal) 1446 in6_ifattach(ifp, NULL); 1447 } 1448 } 1449 } 1450 ND_IFINFO(ifp)->flags = ND.flags; 1451 break; 1452#undef ND 1453 case SIOCSNDFLUSH_IN6: /* XXX: the ioctl name is confusing... */ 1454 /* sync kernel routing table with the default router list */ 1455 defrouter_reset(); 1456 defrouter_select(); 1457 break; 1458 case SIOCSPFXFLUSH_IN6: 1459 { 1460 /* flush all the prefix advertised by routers */ 1461 struct nd_prefix *pr, *next; 1462 1463 s = splnet(); 1464 LIST_FOREACH_SAFE(pr, &V_nd_prefix, ndpr_entry, next) { 1465 struct in6_ifaddr *ia, *ia_next; 1466 1467 if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr)) 1468 continue; /* XXX */ 1469 1470 /* do we really have to remove addresses as well? */ 1471 /* XXXRW: in6_ifaddrhead locking. */ 1472 TAILQ_FOREACH_SAFE(ia, &V_in6_ifaddrhead, ia_link, 1473 ia_next) { 1474 if ((ia->ia6_flags & IN6_IFF_AUTOCONF) == 0) 1475 continue; 1476 1477 if (ia->ia6_ndpr == pr) 1478 in6_purgeaddr(&ia->ia_ifa); 1479 } 1480 prelist_remove(pr); 1481 } 1482 splx(s); 1483 break; 1484 } 1485 case SIOCSRTRFLUSH_IN6: 1486 { 1487 /* flush all the default routers */ 1488 struct nd_defrouter *dr, *next; 1489 1490 s = splnet(); 1491 defrouter_reset(); 1492 TAILQ_FOREACH_SAFE(dr, &V_nd_defrouter, dr_entry, next) { 1493 defrtrlist_del(dr); 1494 } 1495 defrouter_select(); 1496 splx(s); 1497 break; 1498 } 1499 case SIOCGNBRINFO_IN6: 1500 { 1501 struct llentry *ln; 1502 struct in6_addr nb_addr = nbi->addr; /* make local for safety */ 1503 1504 if ((error = in6_setscope(&nb_addr, ifp, NULL)) != 0) 1505 return (error); 1506 1507 IF_AFDATA_LOCK(ifp); 1508 ln = nd6_lookup(&nb_addr, 0, ifp); 1509 IF_AFDATA_UNLOCK(ifp); 1510 1511 if (ln == NULL) { 1512 error = EINVAL; 1513 break; 1514 } 1515 nbi->state = ln->ln_state; 1516 nbi->asked = ln->la_asked; 1517 nbi->isrouter = ln->ln_router; 1518 nbi->expire = ln->la_expire; 1519 LLE_RUNLOCK(ln); 1520 break; 1521 } 1522 case SIOCGDEFIFACE_IN6: /* XXX: should be implemented as a sysctl? */ 1523 ndif->ifindex = V_nd6_defifindex; 1524 break; 1525 case SIOCSDEFIFACE_IN6: /* XXX: should be implemented as a sysctl? */ 1526 return (nd6_setdefaultiface(ndif->ifindex)); 1527 } 1528 return (error); 1529} 1530 1531/* 1532 * Create neighbor cache entry and cache link-layer address, 1533 * on reception of inbound ND6 packets. (RS/RA/NS/redirect) 1534 * 1535 * type - ICMP6 type 1536 * code - type dependent information 1537 * 1538 * XXXXX 1539 * The caller of this function already acquired the ndp 1540 * cache table lock because the cache entry is returned. 1541 */ 1542struct llentry * 1543nd6_cache_lladdr(struct ifnet *ifp, struct in6_addr *from, char *lladdr, 1544 int lladdrlen, int type, int code) 1545{ 1546 struct llentry *ln = NULL; 1547 int is_newentry; 1548 int do_update; 1549 int olladdr; 1550 int llchange; 1551 int flags; 1552 int newstate = 0; 1553 uint16_t router = 0; 1554 struct sockaddr_in6 sin6; 1555 struct mbuf *chain = NULL; 1556 int static_route = 0; 1557 1558 IF_AFDATA_UNLOCK_ASSERT(ifp); 1559 1560 if (ifp == NULL) 1561 panic("ifp == NULL in nd6_cache_lladdr"); 1562 if (from == NULL) 1563 panic("from == NULL in nd6_cache_lladdr"); 1564 1565 /* nothing must be updated for unspecified address */ 1566 if (IN6_IS_ADDR_UNSPECIFIED(from)) 1567 return NULL; 1568 1569 /* 1570 * Validation about ifp->if_addrlen and lladdrlen must be done in 1571 * the caller. 1572 * 1573 * XXX If the link does not have link-layer adderss, what should 1574 * we do? (ifp->if_addrlen == 0) 1575 * Spec says nothing in sections for RA, RS and NA. There's small 1576 * description on it in NS section (RFC 2461 7.2.3). 1577 */ 1578 flags = lladdr ? ND6_EXCLUSIVE : 0; 1579 IF_AFDATA_LOCK(ifp); 1580 ln = nd6_lookup(from, flags, ifp); 1581 1582 if (ln == NULL) { 1583 flags |= ND6_EXCLUSIVE; 1584 ln = nd6_lookup(from, flags | ND6_CREATE, ifp); 1585 IF_AFDATA_UNLOCK(ifp); 1586 is_newentry = 1; 1587 } else { 1588 IF_AFDATA_UNLOCK(ifp); 1589 /* do nothing if static ndp is set */ 1590 if (ln->la_flags & LLE_STATIC) { 1591 static_route = 1; 1592 goto done; 1593 } 1594 is_newentry = 0; 1595 } 1596 if (ln == NULL) 1597 return (NULL); 1598 1599 olladdr = (ln->la_flags & LLE_VALID) ? 1 : 0; 1600 if (olladdr && lladdr) { 1601 llchange = bcmp(lladdr, &ln->ll_addr, 1602 ifp->if_addrlen); 1603 } else 1604 llchange = 0; 1605 1606 /* 1607 * newentry olladdr lladdr llchange (*=record) 1608 * 0 n n -- (1) 1609 * 0 y n -- (2) 1610 * 0 n y -- (3) * STALE 1611 * 0 y y n (4) * 1612 * 0 y y y (5) * STALE 1613 * 1 -- n -- (6) NOSTATE(= PASSIVE) 1614 * 1 -- y -- (7) * STALE 1615 */ 1616 1617 if (lladdr) { /* (3-5) and (7) */ 1618 /* 1619 * Record source link-layer address 1620 * XXX is it dependent to ifp->if_type? 1621 */ 1622 bcopy(lladdr, &ln->ll_addr, ifp->if_addrlen); 1623 ln->la_flags |= LLE_VALID; 1624 } 1625 1626 if (!is_newentry) { 1627 if ((!olladdr && lladdr != NULL) || /* (3) */ 1628 (olladdr && lladdr != NULL && llchange)) { /* (5) */ 1629 do_update = 1; 1630 newstate = ND6_LLINFO_STALE; 1631 } else /* (1-2,4) */ 1632 do_update = 0; 1633 } else { 1634 do_update = 1; 1635 if (lladdr == NULL) /* (6) */ 1636 newstate = ND6_LLINFO_NOSTATE; 1637 else /* (7) */ 1638 newstate = ND6_LLINFO_STALE; 1639 } 1640 1641 if (do_update) { 1642 /* 1643 * Update the state of the neighbor cache. 1644 */ 1645 ln->ln_state = newstate; 1646 1647 if (ln->ln_state == ND6_LLINFO_STALE) { 1648 /* 1649 * XXX: since nd6_output() below will cause 1650 * state tansition to DELAY and reset the timer, 1651 * we must set the timer now, although it is actually 1652 * meaningless. 1653 */ 1654 nd6_llinfo_settimer_locked(ln, (long)V_nd6_gctimer * hz); 1655 1656 if (ln->la_hold) { 1657 struct mbuf *m_hold, *m_hold_next; 1658 1659 /* 1660 * reset the la_hold in advance, to explicitly 1661 * prevent a la_hold lookup in nd6_output() 1662 * (wouldn't happen, though...) 1663 */ 1664 for (m_hold = ln->la_hold, ln->la_hold = NULL; 1665 m_hold; m_hold = m_hold_next) { 1666 m_hold_next = m_hold->m_nextpkt; 1667 m_hold->m_nextpkt = NULL; 1668 1669 /* 1670 * we assume ifp is not a p2p here, so 1671 * just set the 2nd argument as the 1672 * 1st one. 1673 */ 1674 nd6_output_lle(ifp, ifp, m_hold, L3_ADDR_SIN6(ln), NULL, ln, &chain); 1675 } 1676 /* 1677 * If we have mbufs in the chain we need to do 1678 * deferred transmit. Copy the address from the 1679 * llentry before dropping the lock down below. 1680 */ 1681 if (chain != NULL) 1682 memcpy(&sin6, L3_ADDR_SIN6(ln), sizeof(sin6)); 1683 } 1684 } else if (ln->ln_state == ND6_LLINFO_INCOMPLETE) { 1685 /* probe right away */ 1686 nd6_llinfo_settimer_locked((void *)ln, 0); 1687 } 1688 } 1689 1690 /* 1691 * ICMP6 type dependent behavior. 1692 * 1693 * NS: clear IsRouter if new entry 1694 * RS: clear IsRouter 1695 * RA: set IsRouter if there's lladdr 1696 * redir: clear IsRouter if new entry 1697 * 1698 * RA case, (1): 1699 * The spec says that we must set IsRouter in the following cases: 1700 * - If lladdr exist, set IsRouter. This means (1-5). 1701 * - If it is old entry (!newentry), set IsRouter. This means (7). 1702 * So, based on the spec, in (1-5) and (7) cases we must set IsRouter. 1703 * A quetion arises for (1) case. (1) case has no lladdr in the 1704 * neighbor cache, this is similar to (6). 1705 * This case is rare but we figured that we MUST NOT set IsRouter. 1706 * 1707 * newentry olladdr lladdr llchange NS RS RA redir 1708 * D R 1709 * 0 n n -- (1) c ? s 1710 * 0 y n -- (2) c s s 1711 * 0 n y -- (3) c s s 1712 * 0 y y n (4) c s s 1713 * 0 y y y (5) c s s 1714 * 1 -- n -- (6) c c c s 1715 * 1 -- y -- (7) c c s c s 1716 * 1717 * (c=clear s=set) 1718 */ 1719 switch (type & 0xff) { 1720 case ND_NEIGHBOR_SOLICIT: 1721 /* 1722 * New entry must have is_router flag cleared. 1723 */ 1724 if (is_newentry) /* (6-7) */ 1725 ln->ln_router = 0; 1726 break; 1727 case ND_REDIRECT: 1728 /* 1729 * If the icmp is a redirect to a better router, always set the 1730 * is_router flag. Otherwise, if the entry is newly created, 1731 * clear the flag. [RFC 2461, sec 8.3] 1732 */ 1733 if (code == ND_REDIRECT_ROUTER) 1734 ln->ln_router = 1; 1735 else if (is_newentry) /* (6-7) */ 1736 ln->ln_router = 0; 1737 break; 1738 case ND_ROUTER_SOLICIT: 1739 /* 1740 * is_router flag must always be cleared. 1741 */ 1742 ln->ln_router = 0; 1743 break; 1744 case ND_ROUTER_ADVERT: 1745 /* 1746 * Mark an entry with lladdr as a router. 1747 */ 1748 if ((!is_newentry && (olladdr || lladdr)) || /* (2-5) */ 1749 (is_newentry && lladdr)) { /* (7) */ 1750 ln->ln_router = 1; 1751 } 1752 break; 1753 } 1754 1755 if (ln != NULL) { 1756 static_route = (ln->la_flags & LLE_STATIC); 1757 router = ln->ln_router; 1758 1759 if (flags & ND6_EXCLUSIVE) 1760 LLE_WUNLOCK(ln); 1761 else 1762 LLE_RUNLOCK(ln); 1763 if (static_route) 1764 ln = NULL; 1765 } 1766 if (chain) 1767 nd6_output_flush(ifp, ifp, chain, &sin6, NULL); 1768 1769 /* 1770 * When the link-layer address of a router changes, select the 1771 * best router again. In particular, when the neighbor entry is newly 1772 * created, it might affect the selection policy. 1773 * Question: can we restrict the first condition to the "is_newentry" 1774 * case? 1775 * XXX: when we hear an RA from a new router with the link-layer 1776 * address option, defrouter_select() is called twice, since 1777 * defrtrlist_update called the function as well. However, I believe 1778 * we can compromise the overhead, since it only happens the first 1779 * time. 1780 * XXX: although defrouter_select() should not have a bad effect 1781 * for those are not autoconfigured hosts, we explicitly avoid such 1782 * cases for safety. 1783 */ 1784 if (do_update && router && 1785 ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV) { 1786 /* 1787 * guaranteed recursion 1788 */ 1789 defrouter_select(); 1790 } 1791 1792 return (ln); 1793done: 1794 if (ln != NULL) { 1795 if (flags & ND6_EXCLUSIVE) 1796 LLE_WUNLOCK(ln); 1797 else 1798 LLE_RUNLOCK(ln); 1799 if (static_route) 1800 ln = NULL; 1801 } 1802 return (ln); 1803} 1804 1805static void 1806nd6_slowtimo(void *arg) 1807{ 1808 CURVNET_SET((struct vnet *) arg); 1809 struct nd_ifinfo *nd6if; 1810 struct ifnet *ifp; 1811 1812 callout_reset(&V_nd6_slowtimo_ch, ND6_SLOWTIMER_INTERVAL * hz, 1813 nd6_slowtimo, curvnet); 1814 IFNET_RLOCK_NOSLEEP(); 1815 TAILQ_FOREACH(ifp, &V_ifnet, if_list) { 1816 nd6if = ND_IFINFO(ifp); 1817 if (nd6if->basereachable && /* already initialized */ 1818 (nd6if->recalctm -= ND6_SLOWTIMER_INTERVAL) <= 0) { 1819 /* 1820 * Since reachable time rarely changes by router 1821 * advertisements, we SHOULD insure that a new random 1822 * value gets recomputed at least once every few hours. 1823 * (RFC 2461, 6.3.4) 1824 */ 1825 nd6if->recalctm = V_nd6_recalc_reachtm_interval; 1826 nd6if->reachable = ND_COMPUTE_RTIME(nd6if->basereachable); 1827 } 1828 } 1829 IFNET_RUNLOCK_NOSLEEP(); 1830 CURVNET_RESTORE(); 1831} 1832 1833int 1834nd6_output(struct ifnet *ifp, struct ifnet *origifp, struct mbuf *m0, 1835 struct sockaddr_in6 *dst, struct rtentry *rt0) 1836{ 1837 1838 return (nd6_output_lle(ifp, origifp, m0, dst, rt0, NULL, NULL)); 1839} 1840 1841 1842/* 1843 * Note that I'm not enforcing any global serialization 1844 * lle state or asked changes here as the logic is too 1845 * complicated to avoid having to always acquire an exclusive 1846 * lock 1847 * KMM 1848 * 1849 */ 1850#define senderr(e) { error = (e); goto bad;} 1851 1852int 1853nd6_output_lle(struct ifnet *ifp, struct ifnet *origifp, struct mbuf *m0, 1854 struct sockaddr_in6 *dst, struct rtentry *rt0, struct llentry *lle, 1855 struct mbuf **chain) 1856{ 1857 struct mbuf *m = m0; 1858 struct m_tag *mtag; 1859 struct llentry *ln = lle; 1860 struct ip6_hdr *ip6; 1861 int error = 0; 1862 int flags = 0; 1863 int ip6len; 1864 1865#ifdef INVARIANTS 1866 if (lle != NULL) { 1867 1868 LLE_WLOCK_ASSERT(lle); 1869 1870 KASSERT(chain != NULL, (" lle locked but no mbuf chain pointer passed")); 1871 } 1872#endif 1873 if (IN6_IS_ADDR_MULTICAST(&dst->sin6_addr)) 1874 goto sendpkt; 1875 1876 if (nd6_need_cache(ifp) == 0) 1877 goto sendpkt; 1878 1879 /* 1880 * next hop determination. This routine is derived from ether_output. 1881 */ 1882 1883 /* 1884 * Address resolution or Neighbor Unreachability Detection 1885 * for the next hop. 1886 * At this point, the destination of the packet must be a unicast 1887 * or an anycast address(i.e. not a multicast). 1888 */ 1889 1890 flags = ((m != NULL) || (lle != NULL)) ? LLE_EXCLUSIVE : 0; 1891 if (ln == NULL) { 1892 retry: 1893 IF_AFDATA_LOCK(ifp); 1894 ln = lla_lookup(LLTABLE6(ifp), flags, (struct sockaddr *)dst); 1895 IF_AFDATA_UNLOCK(ifp); 1896 if ((ln == NULL) && nd6_is_addr_neighbor(dst, ifp)) { 1897 /* 1898 * Since nd6_is_addr_neighbor() internally calls nd6_lookup(), 1899 * the condition below is not very efficient. But we believe 1900 * it is tolerable, because this should be a rare case. 1901 */ 1902 flags = ND6_CREATE | (m ? ND6_EXCLUSIVE : 0); 1903 IF_AFDATA_LOCK(ifp); 1904 ln = nd6_lookup(&dst->sin6_addr, flags, ifp); 1905 IF_AFDATA_UNLOCK(ifp); 1906 } 1907 } 1908 if (ln == NULL) { 1909 if ((ifp->if_flags & IFF_POINTOPOINT) == 0 && 1910 !(ND_IFINFO(ifp)->flags & ND6_IFF_PERFORMNUD)) { 1911 char ip6buf[INET6_ADDRSTRLEN]; 1912 log(LOG_DEBUG, 1913 "nd6_output: can't allocate llinfo for %s " 1914 "(ln=%p)\n", 1915 ip6_sprintf(ip6buf, &dst->sin6_addr), ln); 1916 senderr(EIO); /* XXX: good error? */ 1917 } 1918 goto sendpkt; /* send anyway */ 1919 } 1920 1921 /* We don't have to do link-layer address resolution on a p2p link. */ 1922 if ((ifp->if_flags & IFF_POINTOPOINT) != 0 && 1923 ln->ln_state < ND6_LLINFO_REACHABLE) { 1924 if ((flags & LLE_EXCLUSIVE) == 0) { 1925 flags |= LLE_EXCLUSIVE; 1926 goto retry; 1927 } 1928 ln->ln_state = ND6_LLINFO_STALE; 1929 nd6_llinfo_settimer_locked(ln, (long)V_nd6_gctimer * hz); 1930 } 1931 1932 /* 1933 * The first time we send a packet to a neighbor whose entry is 1934 * STALE, we have to change the state to DELAY and a sets a timer to 1935 * expire in DELAY_FIRST_PROBE_TIME seconds to ensure do 1936 * neighbor unreachability detection on expiration. 1937 * (RFC 2461 7.3.3) 1938 */ 1939 if (ln->ln_state == ND6_LLINFO_STALE) { 1940 if ((flags & LLE_EXCLUSIVE) == 0) { 1941 flags |= LLE_EXCLUSIVE; 1942 LLE_RUNLOCK(ln); 1943 goto retry; 1944 } 1945 ln->la_asked = 0; 1946 ln->ln_state = ND6_LLINFO_DELAY; 1947 nd6_llinfo_settimer_locked(ln, (long)V_nd6_delay * hz); 1948 } 1949 1950 /* 1951 * If the neighbor cache entry has a state other than INCOMPLETE 1952 * (i.e. its link-layer address is already resolved), just 1953 * send the packet. 1954 */ 1955 if (ln->ln_state > ND6_LLINFO_INCOMPLETE) 1956 goto sendpkt; 1957 1958 /* 1959 * There is a neighbor cache entry, but no ethernet address 1960 * response yet. Append this latest packet to the end of the 1961 * packet queue in the mbuf, unless the number of the packet 1962 * does not exceed nd6_maxqueuelen. When it exceeds nd6_maxqueuelen, 1963 * the oldest packet in the queue will be removed. 1964 */ 1965 if (ln->ln_state == ND6_LLINFO_NOSTATE) 1966 ln->ln_state = ND6_LLINFO_INCOMPLETE; 1967 1968 if ((flags & LLE_EXCLUSIVE) == 0) { 1969 flags |= LLE_EXCLUSIVE; 1970 LLE_RUNLOCK(ln); 1971 goto retry; 1972 } 1973 1974 LLE_WLOCK_ASSERT(ln); 1975 1976 if (ln->la_hold) { 1977 struct mbuf *m_hold; 1978 int i; 1979 1980 i = 0; 1981 for (m_hold = ln->la_hold; m_hold; m_hold = m_hold->m_nextpkt) { 1982 i++; 1983 if (m_hold->m_nextpkt == NULL) { 1984 m_hold->m_nextpkt = m; 1985 break; 1986 } 1987 } 1988 while (i >= V_nd6_maxqueuelen) { 1989 m_hold = ln->la_hold; 1990 ln->la_hold = ln->la_hold->m_nextpkt; 1991 m_freem(m_hold); 1992 i--; 1993 } 1994 } else { 1995 ln->la_hold = m; 1996 } 1997 1998 /* 1999 * If there has been no NS for the neighbor after entering the 2000 * INCOMPLETE state, send the first solicitation. 2001 */ 2002 if (!ND6_LLINFO_PERMANENT(ln) && ln->la_asked == 0) { 2003 ln->la_asked++; 2004 2005 nd6_llinfo_settimer_locked(ln, 2006 (long)ND_IFINFO(ifp)->retrans * hz / 1000); 2007 LLE_WUNLOCK(ln); 2008 nd6_ns_output(ifp, NULL, &dst->sin6_addr, ln, 0); 2009 if (lle != NULL && ln == lle) 2010 LLE_WLOCK(lle); 2011 2012 } else if (lle == NULL || ln != lle) { 2013 /* 2014 * We did the lookup (no lle arg) so we 2015 * need to do the unlock here. 2016 */ 2017 LLE_WUNLOCK(ln); 2018 } 2019 2020 return (0); 2021 2022 sendpkt: 2023 /* discard the packet if IPv6 operation is disabled on the interface */ 2024 if ((ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED)) { 2025 error = ENETDOWN; /* better error? */ 2026 goto bad; 2027 } 2028 /* 2029 * ln is valid and the caller did not pass in 2030 * an llentry 2031 */ 2032 if ((ln != NULL) && (lle == NULL)) { 2033 if (flags & LLE_EXCLUSIVE) 2034 LLE_WUNLOCK(ln); 2035 else 2036 LLE_RUNLOCK(ln); 2037 } 2038 2039#ifdef MAC 2040 mac_netinet6_nd6_send(ifp, m); 2041#endif 2042 2043 /* 2044 * If called from nd6_ns_output() (NS), nd6_na_output() (NA), 2045 * icmp6_redirect_output() (REDIRECT) or from rip6_output() (RS, RA 2046 * as handled by rtsol and rtadvd), mbufs will be tagged for SeND 2047 * to be diverted to user space. When re-injected into the kernel, 2048 * send_output() will directly dispatch them to the outgoing interface. 2049 */ 2050 if (send_sendso_input_hook != NULL) { 2051 mtag = m_tag_find(m, PACKET_TAG_ND_OUTGOING, NULL); 2052 if (mtag != NULL) { 2053 ip6 = mtod(m, struct ip6_hdr *); 2054 ip6len = sizeof(struct ip6_hdr) + ntohs(ip6->ip6_plen); 2055 /* Use the SEND socket */ 2056 error = send_sendso_input_hook(m, ifp, SND_OUT, 2057 ip6len); 2058 /* -1 == no app on SEND socket */ 2059 if (error == 0 || error != -1) 2060 return (error); 2061 } 2062 } 2063 2064 /* 2065 * We were passed in a pointer to an lle with the lock held 2066 * this means that we can't call if_output as we will 2067 * recurse on the lle lock - so what we do is we create 2068 * a list of mbufs to send and transmit them in the caller 2069 * after the lock is dropped 2070 */ 2071 if (lle != NULL) { 2072 if (*chain == NULL) 2073 *chain = m; 2074 else { 2075 struct mbuf *mb; 2076 2077 /* 2078 * append mbuf to end of deferred chain 2079 */ 2080 mb = *chain; 2081 while (mb->m_nextpkt != NULL) 2082 mb = mb->m_nextpkt; 2083 mb->m_nextpkt = m; 2084 } 2085 return (error); 2086 } 2087 if ((ifp->if_flags & IFF_LOOPBACK) != 0) { 2088 return ((*ifp->if_output)(origifp, m, (struct sockaddr *)dst, 2089 NULL)); 2090 } 2091 error = (*ifp->if_output)(ifp, m, (struct sockaddr *)dst, NULL); 2092 return (error); 2093 2094 bad: 2095 /* 2096 * ln is valid and the caller did not pass in 2097 * an llentry 2098 */ 2099 if ((ln != NULL) && (lle == NULL)) { 2100 if (flags & LLE_EXCLUSIVE) 2101 LLE_WUNLOCK(ln); 2102 else 2103 LLE_RUNLOCK(ln); 2104 } 2105 if (m) 2106 m_freem(m); 2107 return (error); 2108} 2109#undef senderr 2110 2111 2112int 2113nd6_output_flush(struct ifnet *ifp, struct ifnet *origifp, struct mbuf *chain, 2114 struct sockaddr_in6 *dst, struct route *ro) 2115{ 2116 struct mbuf *m, *m_head; 2117 struct ifnet *outifp; 2118 int error = 0; 2119 2120 m_head = chain; 2121 if ((ifp->if_flags & IFF_LOOPBACK) != 0) 2122 outifp = origifp; 2123 else 2124 outifp = ifp; 2125 2126 while (m_head) { 2127 m = m_head; 2128 m_head = m_head->m_nextpkt; 2129 error = (*ifp->if_output)(ifp, m, (struct sockaddr *)dst, ro); 2130 } 2131 2132 /* 2133 * XXX 2134 * note that intermediate errors are blindly ignored - but this is 2135 * the same convention as used with nd6_output when called by 2136 * nd6_cache_lladdr 2137 */ 2138 return (error); 2139} 2140 2141 2142int 2143nd6_need_cache(struct ifnet *ifp) 2144{ 2145 /* 2146 * XXX: we currently do not make neighbor cache on any interface 2147 * other than ARCnet, Ethernet, FDDI and GIF. 2148 * 2149 * RFC2893 says: 2150 * - unidirectional tunnels needs no ND 2151 */ 2152 switch (ifp->if_type) { 2153 case IFT_ARCNET: 2154 case IFT_ETHER: 2155 case IFT_FDDI: 2156 case IFT_IEEE1394: 2157#ifdef IFT_L2VLAN 2158 case IFT_L2VLAN: 2159#endif 2160#ifdef IFT_IEEE80211 2161 case IFT_IEEE80211: 2162#endif 2163#ifdef IFT_CARP 2164 case IFT_CARP: 2165#endif 2166 case IFT_INFINIBAND: 2167 case IFT_GIF: /* XXX need more cases? */ 2168 case IFT_PPP: 2169 case IFT_TUNNEL: 2170 case IFT_BRIDGE: 2171 case IFT_PROPVIRTUAL: 2172 return (1); 2173 default: 2174 return (0); 2175 } 2176} 2177 2178/* 2179 * the callers of this function need to be re-worked to drop 2180 * the lle lock, drop here for now 2181 */ 2182int 2183nd6_storelladdr(struct ifnet *ifp, struct mbuf *m, 2184 struct sockaddr *dst, u_char *desten, struct llentry **lle) 2185{ 2186 struct llentry *ln; 2187 2188 *lle = NULL; 2189 IF_AFDATA_UNLOCK_ASSERT(ifp); 2190 if (m->m_flags & M_MCAST) { 2191 int i; 2192 2193 switch (ifp->if_type) { 2194 case IFT_ETHER: 2195 case IFT_FDDI: 2196#ifdef IFT_L2VLAN 2197 case IFT_L2VLAN: 2198#endif 2199#ifdef IFT_IEEE80211 2200 case IFT_IEEE80211: 2201#endif 2202 case IFT_BRIDGE: 2203 case IFT_ISO88025: 2204 ETHER_MAP_IPV6_MULTICAST(&SIN6(dst)->sin6_addr, 2205 desten); 2206 return (0); 2207 case IFT_IEEE1394: 2208 /* 2209 * netbsd can use if_broadcastaddr, but we don't do so 2210 * to reduce # of ifdef. 2211 */ 2212 for (i = 0; i < ifp->if_addrlen; i++) 2213 desten[i] = ~0; 2214 return (0); 2215 case IFT_ARCNET: 2216 *desten = 0; 2217 return (0); 2218 default: 2219 m_freem(m); 2220 return (EAFNOSUPPORT); 2221 } 2222 } 2223 2224 2225 /* 2226 * the entry should have been created in nd6_store_lladdr 2227 */ 2228 IF_AFDATA_LOCK(ifp); 2229 ln = lla_lookup(LLTABLE6(ifp), 0, dst); 2230 IF_AFDATA_UNLOCK(ifp); 2231 if ((ln == NULL) || !(ln->la_flags & LLE_VALID)) { 2232 if (ln != NULL) 2233 LLE_RUNLOCK(ln); 2234 /* this could happen, if we could not allocate memory */ 2235 m_freem(m); 2236 return (1); 2237 } 2238 2239 bcopy(&ln->ll_addr, desten, ifp->if_addrlen); 2240 *lle = ln; 2241 LLE_RUNLOCK(ln); 2242 /* 2243 * A *small* use after free race exists here 2244 */ 2245 return (0); 2246} 2247 2248static void 2249clear_llinfo_pqueue(struct llentry *ln) 2250{ 2251 struct mbuf *m_hold, *m_hold_next; 2252 2253 for (m_hold = ln->la_hold; m_hold; m_hold = m_hold_next) { 2254 m_hold_next = m_hold->m_nextpkt; 2255 m_hold->m_nextpkt = NULL; 2256 m_freem(m_hold); 2257 } 2258 2259 ln->la_hold = NULL; 2260 return; 2261} 2262 2263static int nd6_sysctl_drlist(SYSCTL_HANDLER_ARGS); 2264static int nd6_sysctl_prlist(SYSCTL_HANDLER_ARGS); 2265#ifdef SYSCTL_DECL 2266SYSCTL_DECL(_net_inet6_icmp6); 2267#endif 2268SYSCTL_NODE(_net_inet6_icmp6, ICMPV6CTL_ND6_DRLIST, nd6_drlist, 2269 CTLFLAG_RD, nd6_sysctl_drlist, ""); 2270SYSCTL_NODE(_net_inet6_icmp6, ICMPV6CTL_ND6_PRLIST, nd6_prlist, 2271 CTLFLAG_RD, nd6_sysctl_prlist, ""); 2272SYSCTL_VNET_INT(_net_inet6_icmp6, ICMPV6CTL_ND6_MAXQLEN, nd6_maxqueuelen, 2273 CTLFLAG_RW, &VNET_NAME(nd6_maxqueuelen), 1, ""); 2274 2275static int 2276nd6_sysctl_drlist(SYSCTL_HANDLER_ARGS) 2277{ 2278 struct in6_defrouter d; 2279 struct nd_defrouter *dr; 2280 int error; 2281 2282 if (req->newptr) 2283 return (EPERM); 2284 2285 bzero(&d, sizeof(d)); 2286 d.rtaddr.sin6_family = AF_INET6; 2287 d.rtaddr.sin6_len = sizeof(d.rtaddr); 2288 2289 /* 2290 * XXX locking 2291 */ 2292 TAILQ_FOREACH(dr, &V_nd_defrouter, dr_entry) { 2293 d.rtaddr.sin6_addr = dr->rtaddr; 2294 error = sa6_recoverscope(&d.rtaddr); 2295 if (error != 0) 2296 return (error); 2297 d.flags = dr->flags; 2298 d.rtlifetime = dr->rtlifetime; 2299 d.expire = dr->expire; 2300 d.if_index = dr->ifp->if_index; 2301 error = SYSCTL_OUT(req, &d, sizeof(d)); 2302 if (error != 0) 2303 return (error); 2304 } 2305 return (0); 2306} 2307 2308static int 2309nd6_sysctl_prlist(SYSCTL_HANDLER_ARGS) 2310{ 2311 struct in6_prefix p; 2312 struct sockaddr_in6 s6; 2313 struct nd_prefix *pr; 2314 struct nd_pfxrouter *pfr; 2315 time_t maxexpire; 2316 int error; 2317 char ip6buf[INET6_ADDRSTRLEN]; 2318 2319 if (req->newptr) 2320 return (EPERM); 2321 2322 bzero(&p, sizeof(p)); 2323 p.origin = PR_ORIG_RA; 2324 bzero(&s6, sizeof(s6)); 2325 s6.sin6_family = AF_INET6; 2326 s6.sin6_len = sizeof(s6); 2327 2328 /* 2329 * XXX locking 2330 */ 2331 LIST_FOREACH(pr, &V_nd_prefix, ndpr_entry) { 2332 p.prefix = pr->ndpr_prefix; 2333 if (sa6_recoverscope(&p.prefix)) { 2334 log(LOG_ERR, "scope error in prefix list (%s)\n", 2335 ip6_sprintf(ip6buf, &p.prefix.sin6_addr)); 2336 /* XXX: press on... */ 2337 } 2338 p.raflags = pr->ndpr_raf; 2339 p.prefixlen = pr->ndpr_plen; 2340 p.vltime = pr->ndpr_vltime; 2341 p.pltime = pr->ndpr_pltime; 2342 p.if_index = pr->ndpr_ifp->if_index; 2343 if (pr->ndpr_vltime == ND6_INFINITE_LIFETIME) 2344 p.expire = 0; 2345 else { 2346 /* XXX: we assume time_t is signed. */ 2347 maxexpire = (-1) & 2348 ~((time_t)1 << ((sizeof(maxexpire) * 8) - 1)); 2349 if (pr->ndpr_vltime < maxexpire - pr->ndpr_lastupdate) 2350 p.expire = pr->ndpr_lastupdate + 2351 pr->ndpr_vltime; 2352 else 2353 p.expire = maxexpire; 2354 } 2355 p.refcnt = pr->ndpr_refcnt; 2356 p.flags = pr->ndpr_stateflags; 2357 p.advrtrs = 0; 2358 LIST_FOREACH(pfr, &pr->ndpr_advrtrs, pfr_entry) 2359 p.advrtrs++; 2360 error = SYSCTL_OUT(req, &p, sizeof(p)); 2361 if (error != 0) 2362 return (error); 2363 LIST_FOREACH(pfr, &pr->ndpr_advrtrs, pfr_entry) { 2364 s6.sin6_addr = pfr->router->rtaddr; 2365 if (sa6_recoverscope(&s6)) 2366 log(LOG_ERR, 2367 "scope error in prefix list (%s)\n", 2368 ip6_sprintf(ip6buf, &pfr->router->rtaddr)); 2369 error = SYSCTL_OUT(req, &s6, sizeof(s6)); 2370 if (error != 0) 2371 return (error); 2372 } 2373 } 2374 return (0); 2375} 2376