nd6.c revision 228244
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 228244 2011-12-03 22:12:57Z hrs $"); 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 default: 384 /* 385 * Unknown options must be silently ignored, 386 * to accomodate future extension to the protocol. 387 */ 388 nd6log((LOG_DEBUG, 389 "nd6_options: unsupported option %d - " 390 "option ignored\n", nd_opt->nd_opt_type)); 391 } 392 393skip1: 394 i++; 395 if (i > V_nd6_maxndopt) { 396 ICMP6STAT_INC(icp6s_nd_toomanyopt); 397 nd6log((LOG_INFO, "too many loop in nd opt\n")); 398 break; 399 } 400 401 if (ndopts->nd_opts_done) 402 break; 403 } 404 405 return 0; 406} 407 408/* 409 * ND6 timer routine to handle ND6 entries 410 */ 411void 412nd6_llinfo_settimer_locked(struct llentry *ln, long tick) 413{ 414 int canceled; 415 416 LLE_WLOCK_ASSERT(ln); 417 418 if (tick < 0) { 419 ln->la_expire = 0; 420 ln->ln_ntick = 0; 421 canceled = callout_stop(&ln->ln_timer_ch); 422 } else { 423 ln->la_expire = time_second + tick / hz; 424 LLE_ADDREF(ln); 425 if (tick > INT_MAX) { 426 ln->ln_ntick = tick - INT_MAX; 427 canceled = callout_reset(&ln->ln_timer_ch, INT_MAX, 428 nd6_llinfo_timer, ln); 429 } else { 430 ln->ln_ntick = 0; 431 canceled = callout_reset(&ln->ln_timer_ch, tick, 432 nd6_llinfo_timer, ln); 433 } 434 } 435 if (canceled) 436 LLE_REMREF(ln); 437} 438 439void 440nd6_llinfo_settimer(struct llentry *ln, long tick) 441{ 442 443 LLE_WLOCK(ln); 444 nd6_llinfo_settimer_locked(ln, tick); 445 LLE_WUNLOCK(ln); 446} 447 448static void 449nd6_llinfo_timer(void *arg) 450{ 451 struct llentry *ln; 452 struct in6_addr *dst; 453 struct ifnet *ifp; 454 struct nd_ifinfo *ndi = NULL; 455 456 KASSERT(arg != NULL, ("%s: arg NULL", __func__)); 457 ln = (struct llentry *)arg; 458 LLE_WLOCK_ASSERT(ln); 459 ifp = ln->lle_tbl->llt_ifp; 460 461 CURVNET_SET(ifp->if_vnet); 462 463 if (ln->ln_ntick > 0) { 464 if (ln->ln_ntick > INT_MAX) { 465 ln->ln_ntick -= INT_MAX; 466 nd6_llinfo_settimer_locked(ln, INT_MAX); 467 } else { 468 ln->ln_ntick = 0; 469 nd6_llinfo_settimer_locked(ln, ln->ln_ntick); 470 } 471 goto done; 472 } 473 474 ndi = ND_IFINFO(ifp); 475 dst = &L3_ADDR_SIN6(ln)->sin6_addr; 476 if (ln->la_flags & LLE_STATIC) { 477 goto done; 478 } 479 480 if (ln->la_flags & LLE_DELETED) { 481 (void)nd6_free(ln, 0); 482 ln = NULL; 483 goto done; 484 } 485 486 switch (ln->ln_state) { 487 case ND6_LLINFO_INCOMPLETE: 488 if (ln->la_asked < V_nd6_mmaxtries) { 489 ln->la_asked++; 490 nd6_llinfo_settimer_locked(ln, (long)ndi->retrans * hz / 1000); 491 LLE_WUNLOCK(ln); 492 nd6_ns_output(ifp, NULL, dst, ln, 0); 493 LLE_WLOCK(ln); 494 } else { 495 struct mbuf *m = ln->la_hold; 496 if (m) { 497 struct mbuf *m0; 498 499 /* 500 * assuming every packet in la_hold has the 501 * same IP header. Send error after unlock. 502 */ 503 m0 = m->m_nextpkt; 504 m->m_nextpkt = NULL; 505 ln->la_hold = m0; 506 clear_llinfo_pqueue(ln); 507 } 508 (void)nd6_free(ln, 0); 509 ln = NULL; 510 if (m != NULL) 511 icmp6_error2(m, ICMP6_DST_UNREACH, 512 ICMP6_DST_UNREACH_ADDR, 0, ifp); 513 } 514 break; 515 case ND6_LLINFO_REACHABLE: 516 if (!ND6_LLINFO_PERMANENT(ln)) { 517 ln->ln_state = ND6_LLINFO_STALE; 518 nd6_llinfo_settimer_locked(ln, (long)V_nd6_gctimer * hz); 519 } 520 break; 521 522 case ND6_LLINFO_STALE: 523 /* Garbage Collection(RFC 2461 5.3) */ 524 if (!ND6_LLINFO_PERMANENT(ln)) { 525 (void)nd6_free(ln, 1); 526 ln = NULL; 527 } 528 break; 529 530 case ND6_LLINFO_DELAY: 531 if (ndi && (ndi->flags & ND6_IFF_PERFORMNUD) != 0) { 532 /* We need NUD */ 533 ln->la_asked = 1; 534 ln->ln_state = ND6_LLINFO_PROBE; 535 nd6_llinfo_settimer_locked(ln, (long)ndi->retrans * hz / 1000); 536 LLE_WUNLOCK(ln); 537 nd6_ns_output(ifp, dst, dst, ln, 0); 538 LLE_WLOCK(ln); 539 } else { 540 ln->ln_state = ND6_LLINFO_STALE; /* XXX */ 541 nd6_llinfo_settimer_locked(ln, (long)V_nd6_gctimer * hz); 542 } 543 break; 544 case ND6_LLINFO_PROBE: 545 if (ln->la_asked < V_nd6_umaxtries) { 546 ln->la_asked++; 547 nd6_llinfo_settimer_locked(ln, (long)ndi->retrans * hz / 1000); 548 LLE_WUNLOCK(ln); 549 nd6_ns_output(ifp, dst, dst, ln, 0); 550 LLE_WLOCK(ln); 551 } else { 552 (void)nd6_free(ln, 0); 553 ln = NULL; 554 } 555 break; 556 default: 557 panic("%s: paths in a dark night can be confusing: %d", 558 __func__, ln->ln_state); 559 } 560done: 561 if (ln != NULL) 562 LLE_FREE_LOCKED(ln); 563 CURVNET_RESTORE(); 564} 565 566 567/* 568 * ND6 timer routine to expire default route list and prefix list 569 */ 570void 571nd6_timer(void *arg) 572{ 573 CURVNET_SET((struct vnet *) arg); 574 int s; 575 struct nd_defrouter *dr; 576 struct nd_prefix *pr; 577 struct in6_ifaddr *ia6, *nia6; 578 struct in6_addrlifetime *lt6; 579 580 callout_reset(&V_nd6_timer_ch, V_nd6_prune * hz, 581 nd6_timer, curvnet); 582 583 /* expire default router list */ 584 s = splnet(); 585 dr = TAILQ_FIRST(&V_nd_defrouter); 586 while (dr) { 587 if (dr->expire && dr->expire < time_second) { 588 struct nd_defrouter *t; 589 t = TAILQ_NEXT(dr, dr_entry); 590 defrtrlist_del(dr); 591 dr = t; 592 } else { 593 dr = TAILQ_NEXT(dr, dr_entry); 594 } 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 lt6 = &ia6->ia6_lifetime; 609 if (IFA6_IS_INVALID(ia6)) { 610 int regen = 0; 611 612 /* 613 * If the expiring address is temporary, try 614 * regenerating a new one. This would be useful when 615 * we suspended a laptop PC, then turned it on after a 616 * period that could invalidate all temporary 617 * addresses. Although we may have to restart the 618 * loop (see below), it must be after purging the 619 * address. Otherwise, we'd see an infinite loop of 620 * regeneration. 621 */ 622 if (V_ip6_use_tempaddr && 623 (ia6->ia6_flags & IN6_IFF_TEMPORARY) != 0) { 624 if (regen_tmpaddr(ia6) == 0) 625 regen = 1; 626 } 627 628 in6_purgeaddr(&ia6->ia_ifa); 629 630 if (regen) 631 goto addrloop; /* XXX: see below */ 632 } else if (IFA6_IS_DEPRECATED(ia6)) { 633 int oldflags = ia6->ia6_flags; 634 635 ia6->ia6_flags |= IN6_IFF_DEPRECATED; 636 637 /* 638 * If a temporary address has just become deprecated, 639 * regenerate a new one if possible. 640 */ 641 if (V_ip6_use_tempaddr && 642 (ia6->ia6_flags & IN6_IFF_TEMPORARY) != 0 && 643 (oldflags & IN6_IFF_DEPRECATED) == 0) { 644 645 if (regen_tmpaddr(ia6) == 0) { 646 /* 647 * A new temporary address is 648 * generated. 649 * XXX: this means the address chain 650 * has changed while we are still in 651 * the loop. Although the change 652 * would not cause disaster (because 653 * it's not a deletion, but an 654 * addition,) we'd rather restart the 655 * loop just for safety. Or does this 656 * significantly reduce performance?? 657 */ 658 goto addrloop; 659 } 660 } 661 } else { 662 /* 663 * A new RA might have made a deprecated address 664 * preferred. 665 */ 666 ia6->ia6_flags &= ~IN6_IFF_DEPRECATED; 667 } 668 } 669 670 /* expire prefix list */ 671 pr = V_nd_prefix.lh_first; 672 while (pr) { 673 /* 674 * check prefix lifetime. 675 * since pltime is just for autoconf, pltime processing for 676 * prefix is not necessary. 677 */ 678 if (pr->ndpr_vltime != ND6_INFINITE_LIFETIME && 679 time_second - pr->ndpr_lastupdate > pr->ndpr_vltime) { 680 struct nd_prefix *t; 681 t = pr->ndpr_next; 682 683 /* 684 * address expiration and prefix expiration are 685 * separate. NEVER perform in6_purgeaddr here. 686 */ 687 688 prelist_remove(pr); 689 pr = t; 690 } else 691 pr = pr->ndpr_next; 692 } 693 splx(s); 694 CURVNET_RESTORE(); 695} 696 697/* 698 * ia6 - deprecated/invalidated temporary address 699 */ 700static int 701regen_tmpaddr(struct in6_ifaddr *ia6) 702{ 703 struct ifaddr *ifa; 704 struct ifnet *ifp; 705 struct in6_ifaddr *public_ifa6 = NULL; 706 707 ifp = ia6->ia_ifa.ifa_ifp; 708 IF_ADDR_LOCK(ifp); 709 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 710 struct in6_ifaddr *it6; 711 712 if (ifa->ifa_addr->sa_family != AF_INET6) 713 continue; 714 715 it6 = (struct in6_ifaddr *)ifa; 716 717 /* ignore no autoconf addresses. */ 718 if ((it6->ia6_flags & IN6_IFF_AUTOCONF) == 0) 719 continue; 720 721 /* ignore autoconf addresses with different prefixes. */ 722 if (it6->ia6_ndpr == NULL || it6->ia6_ndpr != ia6->ia6_ndpr) 723 continue; 724 725 /* 726 * Now we are looking at an autoconf address with the same 727 * prefix as ours. If the address is temporary and is still 728 * preferred, do not create another one. It would be rare, but 729 * could happen, for example, when we resume a laptop PC after 730 * a long period. 731 */ 732 if ((it6->ia6_flags & IN6_IFF_TEMPORARY) != 0 && 733 !IFA6_IS_DEPRECATED(it6)) { 734 public_ifa6 = NULL; 735 break; 736 } 737 738 /* 739 * This is a public autoconf address that has the same prefix 740 * as ours. If it is preferred, keep it. We can't break the 741 * loop here, because there may be a still-preferred temporary 742 * address with the prefix. 743 */ 744 if (!IFA6_IS_DEPRECATED(it6)) 745 public_ifa6 = it6; 746 747 if (public_ifa6 != NULL) 748 ifa_ref(&public_ifa6->ia_ifa); 749 } 750 IF_ADDR_UNLOCK(ifp); 751 752 if (public_ifa6 != NULL) { 753 int e; 754 755 if ((e = in6_tmpifadd(public_ifa6, 0, 0)) != 0) { 756 ifa_free(&public_ifa6->ia_ifa); 757 log(LOG_NOTICE, "regen_tmpaddr: failed to create a new" 758 " tmp addr,errno=%d\n", e); 759 return (-1); 760 } 761 ifa_free(&public_ifa6->ia_ifa); 762 return (0); 763 } 764 765 return (-1); 766} 767 768/* 769 * Nuke neighbor cache/prefix/default router management table, right before 770 * ifp goes away. 771 */ 772void 773nd6_purge(struct ifnet *ifp) 774{ 775 struct nd_defrouter *dr, *ndr; 776 struct nd_prefix *pr, *npr; 777 778 /* 779 * Nuke default router list entries toward ifp. 780 * We defer removal of default router list entries that is installed 781 * in the routing table, in order to keep additional side effects as 782 * small as possible. 783 */ 784 for (dr = TAILQ_FIRST(&V_nd_defrouter); dr; dr = ndr) { 785 ndr = TAILQ_NEXT(dr, dr_entry); 786 if (dr->installed) 787 continue; 788 789 if (dr->ifp == ifp) 790 defrtrlist_del(dr); 791 } 792 793 for (dr = TAILQ_FIRST(&V_nd_defrouter); dr; dr = ndr) { 794 ndr = TAILQ_NEXT(dr, dr_entry); 795 if (!dr->installed) 796 continue; 797 798 if (dr->ifp == ifp) 799 defrtrlist_del(dr); 800 } 801 802 /* Nuke prefix list entries toward ifp */ 803 for (pr = V_nd_prefix.lh_first; pr; pr = npr) { 804 npr = pr->ndpr_next; 805 if (pr->ndpr_ifp == ifp) { 806 /* 807 * Because if_detach() does *not* release prefixes 808 * while purging addresses the reference count will 809 * still be above zero. We therefore reset it to 810 * make sure that the prefix really gets purged. 811 */ 812 pr->ndpr_refcnt = 0; 813 814 /* 815 * Previously, pr->ndpr_addr is removed as well, 816 * but I strongly believe we don't have to do it. 817 * nd6_purge() is only called from in6_ifdetach(), 818 * which removes all the associated interface addresses 819 * by itself. 820 * (jinmei@kame.net 20010129) 821 */ 822 prelist_remove(pr); 823 } 824 } 825 826 /* cancel default outgoing interface setting */ 827 if (V_nd6_defifindex == ifp->if_index) 828 nd6_setdefaultiface(0); 829 830 if (ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV) { 831 /* Refresh default router list. */ 832 defrouter_select(); 833 } 834 835 /* XXXXX 836 * We do not nuke the neighbor cache entries here any more 837 * because the neighbor cache is kept in if_afdata[AF_INET6]. 838 * nd6_purge() is invoked by in6_ifdetach() which is called 839 * from if_detach() where everything gets purged. So let 840 * in6_domifdetach() do the actual L2 table purging work. 841 */ 842} 843 844/* 845 * the caller acquires and releases the lock on the lltbls 846 * Returns the llentry locked 847 */ 848struct llentry * 849nd6_lookup(struct in6_addr *addr6, int flags, struct ifnet *ifp) 850{ 851 struct sockaddr_in6 sin6; 852 struct llentry *ln; 853 int llflags; 854 855 bzero(&sin6, sizeof(sin6)); 856 sin6.sin6_len = sizeof(struct sockaddr_in6); 857 sin6.sin6_family = AF_INET6; 858 sin6.sin6_addr = *addr6; 859 860 IF_AFDATA_LOCK_ASSERT(ifp); 861 862 llflags = 0; 863 if (flags & ND6_CREATE) 864 llflags |= LLE_CREATE; 865 if (flags & ND6_EXCLUSIVE) 866 llflags |= LLE_EXCLUSIVE; 867 868 ln = lla_lookup(LLTABLE6(ifp), llflags, (struct sockaddr *)&sin6); 869 if ((ln != NULL) && (llflags & LLE_CREATE)) 870 ln->ln_state = ND6_LLINFO_NOSTATE; 871 872 return (ln); 873} 874 875/* 876 * Test whether a given IPv6 address is a neighbor or not, ignoring 877 * the actual neighbor cache. The neighbor cache is ignored in order 878 * to not reenter the routing code from within itself. 879 */ 880static int 881nd6_is_new_addr_neighbor(struct sockaddr_in6 *addr, struct ifnet *ifp) 882{ 883 struct nd_prefix *pr; 884 struct ifaddr *dstaddr; 885 886 /* 887 * A link-local address is always a neighbor. 888 * XXX: a link does not necessarily specify a single interface. 889 */ 890 if (IN6_IS_ADDR_LINKLOCAL(&addr->sin6_addr)) { 891 struct sockaddr_in6 sin6_copy; 892 u_int32_t zone; 893 894 /* 895 * We need sin6_copy since sa6_recoverscope() may modify the 896 * content (XXX). 897 */ 898 sin6_copy = *addr; 899 if (sa6_recoverscope(&sin6_copy)) 900 return (0); /* XXX: should be impossible */ 901 if (in6_setscope(&sin6_copy.sin6_addr, ifp, &zone)) 902 return (0); 903 if (sin6_copy.sin6_scope_id == zone) 904 return (1); 905 else 906 return (0); 907 } 908 909 /* 910 * If the address matches one of our addresses, 911 * it should be a neighbor. 912 * If the address matches one of our on-link prefixes, it should be a 913 * neighbor. 914 */ 915 for (pr = V_nd_prefix.lh_first; pr; pr = pr->ndpr_next) { 916 if (pr->ndpr_ifp != ifp) 917 continue; 918 919 if (!(pr->ndpr_stateflags & NDPRF_ONLINK)) { 920 struct rtentry *rt; 921 rt = rtalloc1((struct sockaddr *)&pr->ndpr_prefix, 0, 0); 922 if (rt == NULL) 923 continue; 924 /* 925 * This is the case where multiple interfaces 926 * have the same prefix, but only one is installed 927 * into the routing table and that prefix entry 928 * is not the one being examined here. In the case 929 * where RADIX_MPATH is enabled, multiple route 930 * entries (of the same rt_key value) will be 931 * installed because the interface addresses all 932 * differ. 933 */ 934 if (!IN6_ARE_ADDR_EQUAL(&pr->ndpr_prefix.sin6_addr, 935 &((struct sockaddr_in6 *)rt_key(rt))->sin6_addr)) { 936 RTFREE_LOCKED(rt); 937 continue; 938 } 939 RTFREE_LOCKED(rt); 940 } 941 942 if (IN6_ARE_MASKED_ADDR_EQUAL(&pr->ndpr_prefix.sin6_addr, 943 &addr->sin6_addr, &pr->ndpr_mask)) 944 return (1); 945 } 946 947 /* 948 * If the address is assigned on the node of the other side of 949 * a p2p interface, the address should be a neighbor. 950 */ 951 dstaddr = ifa_ifwithdstaddr((struct sockaddr *)addr); 952 if (dstaddr != NULL) { 953 if (dstaddr->ifa_ifp == ifp) { 954 ifa_free(dstaddr); 955 return (1); 956 } 957 ifa_free(dstaddr); 958 } 959 960 /* 961 * If the default router list is empty, all addresses are regarded 962 * as on-link, and thus, as a neighbor. 963 */ 964 if (ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV && 965 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 */ 978int 979nd6_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 */ 1007static struct llentry * 1008nd6_free(struct llentry *ln, int gc) 1009{ 1010 struct llentry *next; 1011 struct nd_defrouter *dr; 1012 struct ifnet *ifp; 1013 1014 LLE_WLOCK_ASSERT(ln); 1015 1016 /* 1017 * we used to have pfctlinput(PRC_HOSTDEAD) here. 1018 * even though it is not harmful, it was not really necessary. 1019 */ 1020 1021 /* cancel timer */ 1022 nd6_llinfo_settimer_locked(ln, -1); 1023 1024 ifp = ln->lle_tbl->llt_ifp; 1025 1026 if (ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV) { 1027 dr = defrouter_lookup(&L3_ADDR_SIN6(ln)->sin6_addr, ifp); 1028 1029 if (dr != NULL && dr->expire && 1030 ln->ln_state == ND6_LLINFO_STALE && gc) { 1031 /* 1032 * If the reason for the deletion is just garbage 1033 * collection, and the neighbor is an active default 1034 * router, do not delete it. Instead, reset the GC 1035 * timer using the router's lifetime. 1036 * Simply deleting the entry would affect default 1037 * router selection, which is not necessarily a good 1038 * thing, especially when we're using router preference 1039 * values. 1040 * XXX: the check for ln_state would be redundant, 1041 * but we intentionally keep it just in case. 1042 */ 1043 if (dr->expire > time_second) 1044 nd6_llinfo_settimer_locked(ln, 1045 (dr->expire - time_second) * hz); 1046 else 1047 nd6_llinfo_settimer_locked(ln, 1048 (long)V_nd6_gctimer * hz); 1049 1050 next = LIST_NEXT(ln, lle_next); 1051 LLE_REMREF(ln); 1052 LLE_WUNLOCK(ln); 1053 return (next); 1054 } 1055 1056 if (dr) { 1057 /* 1058 * Unreachablity of a router might affect the default 1059 * router selection and on-link detection of advertised 1060 * prefixes. 1061 */ 1062 1063 /* 1064 * Temporarily fake the state to choose a new default 1065 * router and to perform on-link determination of 1066 * prefixes correctly. 1067 * Below the state will be set correctly, 1068 * or the entry itself will be deleted. 1069 */ 1070 ln->ln_state = ND6_LLINFO_INCOMPLETE; 1071 } 1072 1073 if (ln->ln_router || dr) { 1074 1075 /* 1076 * We need to unlock to avoid a LOR with rt6_flush() with the 1077 * rnh and for the calls to pfxlist_onlink_check() and 1078 * defrouter_select() in the block further down for calls 1079 * into nd6_lookup(). We still hold a ref. 1080 */ 1081 LLE_WUNLOCK(ln); 1082 1083 /* 1084 * rt6_flush must be called whether or not the neighbor 1085 * is in the Default Router List. 1086 * See a corresponding comment in nd6_na_input(). 1087 */ 1088 rt6_flush(&L3_ADDR_SIN6(ln)->sin6_addr, ifp); 1089 } 1090 1091 if (dr) { 1092 /* 1093 * Since defrouter_select() does not affect the 1094 * on-link determination and MIP6 needs the check 1095 * before the default router selection, we perform 1096 * the check now. 1097 */ 1098 pfxlist_onlink_check(); 1099 1100 /* 1101 * Refresh default router list. 1102 */ 1103 defrouter_select(); 1104 } 1105 1106 if (ln->ln_router || dr) 1107 LLE_WLOCK(ln); 1108 } 1109 1110 /* 1111 * Before deleting the entry, remember the next entry as the 1112 * return value. We need this because pfxlist_onlink_check() above 1113 * might have freed other entries (particularly the old next entry) as 1114 * a side effect (XXX). 1115 */ 1116 next = LIST_NEXT(ln, lle_next); 1117 1118 /* 1119 * Save to unlock. We still hold an extra reference and will not 1120 * free(9) in llentry_free() if someone else holds one as well. 1121 */ 1122 LLE_WUNLOCK(ln); 1123 IF_AFDATA_LOCK(ifp); 1124 LLE_WLOCK(ln); 1125 LLE_REMREF(ln); 1126 llentry_free(ln); 1127 IF_AFDATA_UNLOCK(ifp); 1128 1129 return (next); 1130} 1131 1132/* 1133 * Upper-layer reachability hint for Neighbor Unreachability Detection. 1134 * 1135 * XXX cost-effective methods? 1136 */ 1137void 1138nd6_nud_hint(struct rtentry *rt, struct in6_addr *dst6, int force) 1139{ 1140 struct llentry *ln; 1141 struct ifnet *ifp; 1142 1143 if ((dst6 == NULL) || (rt == NULL)) 1144 return; 1145 1146 ifp = rt->rt_ifp; 1147 IF_AFDATA_LOCK(ifp); 1148 ln = nd6_lookup(dst6, ND6_EXCLUSIVE, NULL); 1149 IF_AFDATA_UNLOCK(ifp); 1150 if (ln == NULL) 1151 return; 1152 1153 if (ln->ln_state < ND6_LLINFO_REACHABLE) 1154 goto done; 1155 1156 /* 1157 * if we get upper-layer reachability confirmation many times, 1158 * it is possible we have false information. 1159 */ 1160 if (!force) { 1161 ln->ln_byhint++; 1162 if (ln->ln_byhint > V_nd6_maxnudhint) { 1163 goto done; 1164 } 1165 } 1166 1167 ln->ln_state = ND6_LLINFO_REACHABLE; 1168 if (!ND6_LLINFO_PERMANENT(ln)) { 1169 nd6_llinfo_settimer_locked(ln, 1170 (long)ND_IFINFO(rt->rt_ifp)->reachable * hz); 1171 } 1172done: 1173 LLE_WUNLOCK(ln); 1174} 1175 1176 1177int 1178nd6_ioctl(u_long cmd, caddr_t data, struct ifnet *ifp) 1179{ 1180 struct in6_drlist *drl = (struct in6_drlist *)data; 1181 struct in6_oprlist *oprl = (struct in6_oprlist *)data; 1182 struct in6_ndireq *ndi = (struct in6_ndireq *)data; 1183 struct in6_nbrinfo *nbi = (struct in6_nbrinfo *)data; 1184 struct in6_ndifreq *ndif = (struct in6_ndifreq *)data; 1185 struct nd_defrouter *dr; 1186 struct nd_prefix *pr; 1187 int i = 0, error = 0; 1188 int s; 1189 1190 switch (cmd) { 1191 case SIOCGDRLST_IN6: 1192 /* 1193 * obsolete API, use sysctl under net.inet6.icmp6 1194 */ 1195 bzero(drl, sizeof(*drl)); 1196 s = splnet(); 1197 dr = TAILQ_FIRST(&V_nd_defrouter); 1198 while (dr && i < DRLSTSIZ) { 1199 drl->defrouter[i].rtaddr = dr->rtaddr; 1200 in6_clearscope(&drl->defrouter[i].rtaddr); 1201 1202 drl->defrouter[i].flags = dr->flags; 1203 drl->defrouter[i].rtlifetime = dr->rtlifetime; 1204 drl->defrouter[i].expire = dr->expire; 1205 drl->defrouter[i].if_index = dr->ifp->if_index; 1206 i++; 1207 dr = TAILQ_NEXT(dr, dr_entry); 1208 } 1209 splx(s); 1210 break; 1211 case SIOCGPRLST_IN6: 1212 /* 1213 * obsolete API, use sysctl under net.inet6.icmp6 1214 * 1215 * XXX the structure in6_prlist was changed in backward- 1216 * incompatible manner. in6_oprlist is used for SIOCGPRLST_IN6, 1217 * in6_prlist is used for nd6_sysctl() - fill_prlist(). 1218 */ 1219 /* 1220 * XXX meaning of fields, especialy "raflags", is very 1221 * differnet between RA prefix list and RR/static prefix list. 1222 * how about separating ioctls into two? 1223 */ 1224 bzero(oprl, sizeof(*oprl)); 1225 s = splnet(); 1226 pr = V_nd_prefix.lh_first; 1227 while (pr && i < PRLSTSIZ) { 1228 struct nd_pfxrouter *pfr; 1229 int j; 1230 1231 oprl->prefix[i].prefix = pr->ndpr_prefix.sin6_addr; 1232 oprl->prefix[i].raflags = pr->ndpr_raf; 1233 oprl->prefix[i].prefixlen = pr->ndpr_plen; 1234 oprl->prefix[i].vltime = pr->ndpr_vltime; 1235 oprl->prefix[i].pltime = pr->ndpr_pltime; 1236 oprl->prefix[i].if_index = pr->ndpr_ifp->if_index; 1237 if (pr->ndpr_vltime == ND6_INFINITE_LIFETIME) 1238 oprl->prefix[i].expire = 0; 1239 else { 1240 time_t maxexpire; 1241 1242 /* XXX: we assume time_t is signed. */ 1243 maxexpire = (-1) & 1244 ~((time_t)1 << 1245 ((sizeof(maxexpire) * 8) - 1)); 1246 if (pr->ndpr_vltime < 1247 maxexpire - pr->ndpr_lastupdate) { 1248 oprl->prefix[i].expire = 1249 pr->ndpr_lastupdate + 1250 pr->ndpr_vltime; 1251 } else 1252 oprl->prefix[i].expire = maxexpire; 1253 } 1254 1255 pfr = pr->ndpr_advrtrs.lh_first; 1256 j = 0; 1257 while (pfr) { 1258 if (j < DRLSTSIZ) { 1259#define RTRADDR oprl->prefix[i].advrtr[j] 1260 RTRADDR = pfr->router->rtaddr; 1261 in6_clearscope(&RTRADDR); 1262#undef RTRADDR 1263 } 1264 j++; 1265 pfr = pfr->pfr_next; 1266 } 1267 oprl->prefix[i].advrtrs = j; 1268 oprl->prefix[i].origin = PR_ORIG_RA; 1269 1270 i++; 1271 pr = pr->ndpr_next; 1272 } 1273 splx(s); 1274 1275 break; 1276 case OSIOCGIFINFO_IN6: 1277#define ND ndi->ndi 1278 /* XXX: old ndp(8) assumes a positive value for linkmtu. */ 1279 bzero(&ND, sizeof(ND)); 1280 ND.linkmtu = IN6_LINKMTU(ifp); 1281 ND.maxmtu = ND_IFINFO(ifp)->maxmtu; 1282 ND.basereachable = ND_IFINFO(ifp)->basereachable; 1283 ND.reachable = ND_IFINFO(ifp)->reachable; 1284 ND.retrans = ND_IFINFO(ifp)->retrans; 1285 ND.flags = ND_IFINFO(ifp)->flags; 1286 ND.recalctm = ND_IFINFO(ifp)->recalctm; 1287 ND.chlim = ND_IFINFO(ifp)->chlim; 1288 break; 1289 case SIOCGIFINFO_IN6: 1290 ND = *ND_IFINFO(ifp); 1291 break; 1292 case SIOCSIFINFO_IN6: 1293 /* 1294 * used to change host variables from userland. 1295 * intented for a use on router to reflect RA configurations. 1296 */ 1297 /* 0 means 'unspecified' */ 1298 if (ND.linkmtu != 0) { 1299 if (ND.linkmtu < IPV6_MMTU || 1300 ND.linkmtu > IN6_LINKMTU(ifp)) { 1301 error = EINVAL; 1302 break; 1303 } 1304 ND_IFINFO(ifp)->linkmtu = ND.linkmtu; 1305 } 1306 1307 if (ND.basereachable != 0) { 1308 int obasereachable = ND_IFINFO(ifp)->basereachable; 1309 1310 ND_IFINFO(ifp)->basereachable = ND.basereachable; 1311 if (ND.basereachable != obasereachable) 1312 ND_IFINFO(ifp)->reachable = 1313 ND_COMPUTE_RTIME(ND.basereachable); 1314 } 1315 if (ND.retrans != 0) 1316 ND_IFINFO(ifp)->retrans = ND.retrans; 1317 if (ND.chlim != 0) 1318 ND_IFINFO(ifp)->chlim = ND.chlim; 1319 /* FALLTHROUGH */ 1320 case SIOCSIFINFO_FLAGS: 1321 { 1322 struct ifaddr *ifa; 1323 struct in6_ifaddr *ia; 1324 1325 /* 1326 * Try to clear ifdisabled flag when enabling 1327 * accept_rtadv or auto_linklocal. 1328 */ 1329 if ((ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) && 1330 !(ND.flags & ND6_IFF_IFDISABLED) && 1331 (ND.flags & (ND6_IFF_ACCEPT_RTADV | 1332 ND6_IFF_AUTO_LINKLOCAL))) 1333 ND.flags &= ~ND6_IFF_IFDISABLED; 1334 1335 if ((ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) && 1336 !(ND.flags & ND6_IFF_IFDISABLED)) { 1337 /* ifdisabled 1->0 transision */ 1338 1339 /* 1340 * If the interface is marked as ND6_IFF_IFDISABLED and 1341 * has an link-local address with IN6_IFF_DUPLICATED, 1342 * do not clear ND6_IFF_IFDISABLED. 1343 * See RFC 4862, Section 5.4.5. 1344 */ 1345 int duplicated_linklocal = 0; 1346 1347 IF_ADDR_LOCK(ifp); 1348 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1349 if (ifa->ifa_addr->sa_family != AF_INET6) 1350 continue; 1351 ia = (struct in6_ifaddr *)ifa; 1352 if ((ia->ia6_flags & IN6_IFF_DUPLICATED) && 1353 IN6_IS_ADDR_LINKLOCAL(IA6_IN6(ia))) { 1354 duplicated_linklocal = 1; 1355 break; 1356 } 1357 } 1358 IF_ADDR_UNLOCK(ifp); 1359 1360 if (duplicated_linklocal) { 1361 ND.flags |= ND6_IFF_IFDISABLED; 1362 log(LOG_ERR, "Cannot enable an interface" 1363 " with a link-local address marked" 1364 " duplicate.\n"); 1365 } else { 1366 ND_IFINFO(ifp)->flags &= ~ND6_IFF_IFDISABLED; 1367 if (ifp->if_flags & IFF_UP) 1368 in6_if_up(ifp); 1369 } 1370 } else if (!(ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) && 1371 (ND.flags & ND6_IFF_IFDISABLED)) { 1372 /* ifdisabled 0->1 transision */ 1373 /* Mark all IPv6 address as tentative. */ 1374 1375 ND_IFINFO(ifp)->flags |= ND6_IFF_IFDISABLED; 1376 IF_ADDR_LOCK(ifp); 1377 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1378 if (ifa->ifa_addr->sa_family != AF_INET6) 1379 continue; 1380 ia = (struct in6_ifaddr *)ifa; 1381 ia->ia6_flags |= IN6_IFF_TENTATIVE; 1382 } 1383 IF_ADDR_UNLOCK(ifp); 1384 } 1385 1386 if (ND.flags & ND6_IFF_AUTO_LINKLOCAL) { 1387 if (!(ND_IFINFO(ifp)->flags & ND6_IFF_AUTO_LINKLOCAL)) { 1388 /* auto_linklocal 0->1 transision */ 1389 1390 /* If no link-local address on ifp, configure */ 1391 ND_IFINFO(ifp)->flags |= ND6_IFF_AUTO_LINKLOCAL; 1392 in6_ifattach(ifp, NULL); 1393 } else if (!(ND.flags & ND6_IFF_IFDISABLED) && 1394 ifp->if_flags & IFF_UP) { 1395 /* 1396 * When the IF already has 1397 * ND6_IFF_AUTO_LINKLOCAL, no link-local 1398 * address is assigned, and IFF_UP, try to 1399 * assign one. 1400 */ 1401 int haslinklocal = 0; 1402 1403 IF_ADDR_LOCK(ifp); 1404 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1405 if (ifa->ifa_addr->sa_family != AF_INET6) 1406 continue; 1407 ia = (struct in6_ifaddr *)ifa; 1408 if (IN6_IS_ADDR_LINKLOCAL(IA6_IN6(ia))) { 1409 haslinklocal = 1; 1410 break; 1411 } 1412 } 1413 IF_ADDR_UNLOCK(ifp); 1414 if (!haslinklocal) 1415 in6_ifattach(ifp, NULL); 1416 } 1417 } 1418 } 1419 ND_IFINFO(ifp)->flags = ND.flags; 1420 break; 1421#undef ND 1422 case SIOCSNDFLUSH_IN6: /* XXX: the ioctl name is confusing... */ 1423 /* sync kernel routing table with the default router list */ 1424 defrouter_reset(); 1425 defrouter_select(); 1426 break; 1427 case SIOCSPFXFLUSH_IN6: 1428 { 1429 /* flush all the prefix advertised by routers */ 1430 struct nd_prefix *pr, *next; 1431 1432 s = splnet(); 1433 for (pr = V_nd_prefix.lh_first; pr; pr = next) { 1434 struct in6_ifaddr *ia, *ia_next; 1435 1436 next = pr->ndpr_next; 1437 1438 if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr)) 1439 continue; /* XXX */ 1440 1441 /* do we really have to remove addresses as well? */ 1442 /* XXXRW: in6_ifaddrhead locking. */ 1443 TAILQ_FOREACH_SAFE(ia, &V_in6_ifaddrhead, ia_link, 1444 ia_next) { 1445 if ((ia->ia6_flags & IN6_IFF_AUTOCONF) == 0) 1446 continue; 1447 1448 if (ia->ia6_ndpr == pr) 1449 in6_purgeaddr(&ia->ia_ifa); 1450 } 1451 prelist_remove(pr); 1452 } 1453 splx(s); 1454 break; 1455 } 1456 case SIOCSRTRFLUSH_IN6: 1457 { 1458 /* flush all the default routers */ 1459 struct nd_defrouter *dr, *next; 1460 1461 s = splnet(); 1462 defrouter_reset(); 1463 for (dr = TAILQ_FIRST(&V_nd_defrouter); dr; dr = next) { 1464 next = TAILQ_NEXT(dr, dr_entry); 1465 defrtrlist_del(dr); 1466 } 1467 defrouter_select(); 1468 splx(s); 1469 break; 1470 } 1471 case SIOCGNBRINFO_IN6: 1472 { 1473 struct llentry *ln; 1474 struct in6_addr nb_addr = nbi->addr; /* make local for safety */ 1475 1476 if ((error = in6_setscope(&nb_addr, ifp, NULL)) != 0) 1477 return (error); 1478 1479 IF_AFDATA_LOCK(ifp); 1480 ln = nd6_lookup(&nb_addr, 0, ifp); 1481 IF_AFDATA_UNLOCK(ifp); 1482 1483 if (ln == NULL) { 1484 error = EINVAL; 1485 break; 1486 } 1487 nbi->state = ln->ln_state; 1488 nbi->asked = ln->la_asked; 1489 nbi->isrouter = ln->ln_router; 1490 nbi->expire = ln->la_expire; 1491 LLE_RUNLOCK(ln); 1492 break; 1493 } 1494 case SIOCGDEFIFACE_IN6: /* XXX: should be implemented as a sysctl? */ 1495 ndif->ifindex = V_nd6_defifindex; 1496 break; 1497 case SIOCSDEFIFACE_IN6: /* XXX: should be implemented as a sysctl? */ 1498 return (nd6_setdefaultiface(ndif->ifindex)); 1499 } 1500 return (error); 1501} 1502 1503/* 1504 * Create neighbor cache entry and cache link-layer address, 1505 * on reception of inbound ND6 packets. (RS/RA/NS/redirect) 1506 * 1507 * type - ICMP6 type 1508 * code - type dependent information 1509 * 1510 * XXXXX 1511 * The caller of this function already acquired the ndp 1512 * cache table lock because the cache entry is returned. 1513 */ 1514struct llentry * 1515nd6_cache_lladdr(struct ifnet *ifp, struct in6_addr *from, char *lladdr, 1516 int lladdrlen, int type, int code) 1517{ 1518 struct llentry *ln = NULL; 1519 int is_newentry; 1520 int do_update; 1521 int olladdr; 1522 int llchange; 1523 int flags; 1524 int newstate = 0; 1525 uint16_t router = 0; 1526 struct sockaddr_in6 sin6; 1527 struct mbuf *chain = NULL; 1528 int static_route = 0; 1529 1530 IF_AFDATA_UNLOCK_ASSERT(ifp); 1531 1532 if (ifp == NULL) 1533 panic("ifp == NULL in nd6_cache_lladdr"); 1534 if (from == NULL) 1535 panic("from == NULL in nd6_cache_lladdr"); 1536 1537 /* nothing must be updated for unspecified address */ 1538 if (IN6_IS_ADDR_UNSPECIFIED(from)) 1539 return NULL; 1540 1541 /* 1542 * Validation about ifp->if_addrlen and lladdrlen must be done in 1543 * the caller. 1544 * 1545 * XXX If the link does not have link-layer adderss, what should 1546 * we do? (ifp->if_addrlen == 0) 1547 * Spec says nothing in sections for RA, RS and NA. There's small 1548 * description on it in NS section (RFC 2461 7.2.3). 1549 */ 1550 flags = lladdr ? ND6_EXCLUSIVE : 0; 1551 IF_AFDATA_LOCK(ifp); 1552 ln = nd6_lookup(from, flags, ifp); 1553 1554 if (ln == NULL) { 1555 flags |= ND6_EXCLUSIVE; 1556 ln = nd6_lookup(from, flags | ND6_CREATE, ifp); 1557 IF_AFDATA_UNLOCK(ifp); 1558 is_newentry = 1; 1559 } else { 1560 IF_AFDATA_UNLOCK(ifp); 1561 /* do nothing if static ndp is set */ 1562 if (ln->la_flags & LLE_STATIC) { 1563 static_route = 1; 1564 goto done; 1565 } 1566 is_newentry = 0; 1567 } 1568 if (ln == NULL) 1569 return (NULL); 1570 1571 olladdr = (ln->la_flags & LLE_VALID) ? 1 : 0; 1572 if (olladdr && lladdr) { 1573 llchange = bcmp(lladdr, &ln->ll_addr, 1574 ifp->if_addrlen); 1575 } else 1576 llchange = 0; 1577 1578 /* 1579 * newentry olladdr lladdr llchange (*=record) 1580 * 0 n n -- (1) 1581 * 0 y n -- (2) 1582 * 0 n y -- (3) * STALE 1583 * 0 y y n (4) * 1584 * 0 y y y (5) * STALE 1585 * 1 -- n -- (6) NOSTATE(= PASSIVE) 1586 * 1 -- y -- (7) * STALE 1587 */ 1588 1589 if (lladdr) { /* (3-5) and (7) */ 1590 /* 1591 * Record source link-layer address 1592 * XXX is it dependent to ifp->if_type? 1593 */ 1594 bcopy(lladdr, &ln->ll_addr, ifp->if_addrlen); 1595 ln->la_flags |= LLE_VALID; 1596 } 1597 1598 if (!is_newentry) { 1599 if ((!olladdr && lladdr != NULL) || /* (3) */ 1600 (olladdr && lladdr != NULL && llchange)) { /* (5) */ 1601 do_update = 1; 1602 newstate = ND6_LLINFO_STALE; 1603 } else /* (1-2,4) */ 1604 do_update = 0; 1605 } else { 1606 do_update = 1; 1607 if (lladdr == NULL) /* (6) */ 1608 newstate = ND6_LLINFO_NOSTATE; 1609 else /* (7) */ 1610 newstate = ND6_LLINFO_STALE; 1611 } 1612 1613 if (do_update) { 1614 /* 1615 * Update the state of the neighbor cache. 1616 */ 1617 ln->ln_state = newstate; 1618 1619 if (ln->ln_state == ND6_LLINFO_STALE) { 1620 /* 1621 * XXX: since nd6_output() below will cause 1622 * state tansition to DELAY and reset the timer, 1623 * we must set the timer now, although it is actually 1624 * meaningless. 1625 */ 1626 nd6_llinfo_settimer_locked(ln, (long)V_nd6_gctimer * hz); 1627 1628 if (ln->la_hold) { 1629 struct mbuf *m_hold, *m_hold_next; 1630 1631 /* 1632 * reset the la_hold in advance, to explicitly 1633 * prevent a la_hold lookup in nd6_output() 1634 * (wouldn't happen, though...) 1635 */ 1636 for (m_hold = ln->la_hold, ln->la_hold = NULL; 1637 m_hold; m_hold = m_hold_next) { 1638 m_hold_next = m_hold->m_nextpkt; 1639 m_hold->m_nextpkt = NULL; 1640 1641 /* 1642 * we assume ifp is not a p2p here, so 1643 * just set the 2nd argument as the 1644 * 1st one. 1645 */ 1646 nd6_output_lle(ifp, ifp, m_hold, L3_ADDR_SIN6(ln), NULL, ln, &chain); 1647 } 1648 /* 1649 * If we have mbufs in the chain we need to do 1650 * deferred transmit. Copy the address from the 1651 * llentry before dropping the lock down below. 1652 */ 1653 if (chain != NULL) 1654 memcpy(&sin6, L3_ADDR_SIN6(ln), sizeof(sin6)); 1655 } 1656 } else if (ln->ln_state == ND6_LLINFO_INCOMPLETE) { 1657 /* probe right away */ 1658 nd6_llinfo_settimer_locked((void *)ln, 0); 1659 } 1660 } 1661 1662 /* 1663 * ICMP6 type dependent behavior. 1664 * 1665 * NS: clear IsRouter if new entry 1666 * RS: clear IsRouter 1667 * RA: set IsRouter if there's lladdr 1668 * redir: clear IsRouter if new entry 1669 * 1670 * RA case, (1): 1671 * The spec says that we must set IsRouter in the following cases: 1672 * - If lladdr exist, set IsRouter. This means (1-5). 1673 * - If it is old entry (!newentry), set IsRouter. This means (7). 1674 * So, based on the spec, in (1-5) and (7) cases we must set IsRouter. 1675 * A quetion arises for (1) case. (1) case has no lladdr in the 1676 * neighbor cache, this is similar to (6). 1677 * This case is rare but we figured that we MUST NOT set IsRouter. 1678 * 1679 * newentry olladdr lladdr llchange NS RS RA redir 1680 * D R 1681 * 0 n n -- (1) c ? s 1682 * 0 y n -- (2) c s s 1683 * 0 n y -- (3) c s s 1684 * 0 y y n (4) c s s 1685 * 0 y y y (5) c s s 1686 * 1 -- n -- (6) c c c s 1687 * 1 -- y -- (7) c c s c s 1688 * 1689 * (c=clear s=set) 1690 */ 1691 switch (type & 0xff) { 1692 case ND_NEIGHBOR_SOLICIT: 1693 /* 1694 * New entry must have is_router flag cleared. 1695 */ 1696 if (is_newentry) /* (6-7) */ 1697 ln->ln_router = 0; 1698 break; 1699 case ND_REDIRECT: 1700 /* 1701 * If the icmp is a redirect to a better router, always set the 1702 * is_router flag. Otherwise, if the entry is newly created, 1703 * clear the flag. [RFC 2461, sec 8.3] 1704 */ 1705 if (code == ND_REDIRECT_ROUTER) 1706 ln->ln_router = 1; 1707 else if (is_newentry) /* (6-7) */ 1708 ln->ln_router = 0; 1709 break; 1710 case ND_ROUTER_SOLICIT: 1711 /* 1712 * is_router flag must always be cleared. 1713 */ 1714 ln->ln_router = 0; 1715 break; 1716 case ND_ROUTER_ADVERT: 1717 /* 1718 * Mark an entry with lladdr as a router. 1719 */ 1720 if ((!is_newentry && (olladdr || lladdr)) || /* (2-5) */ 1721 (is_newentry && lladdr)) { /* (7) */ 1722 ln->ln_router = 1; 1723 } 1724 break; 1725 } 1726 1727 if (ln != NULL) { 1728 static_route = (ln->la_flags & LLE_STATIC); 1729 router = ln->ln_router; 1730 1731 if (flags & ND6_EXCLUSIVE) 1732 LLE_WUNLOCK(ln); 1733 else 1734 LLE_RUNLOCK(ln); 1735 if (static_route) 1736 ln = NULL; 1737 } 1738 if (chain) 1739 nd6_output_flush(ifp, ifp, chain, &sin6, NULL); 1740 1741 /* 1742 * When the link-layer address of a router changes, select the 1743 * best router again. In particular, when the neighbor entry is newly 1744 * created, it might affect the selection policy. 1745 * Question: can we restrict the first condition to the "is_newentry" 1746 * case? 1747 * XXX: when we hear an RA from a new router with the link-layer 1748 * address option, defrouter_select() is called twice, since 1749 * defrtrlist_update called the function as well. However, I believe 1750 * we can compromise the overhead, since it only happens the first 1751 * time. 1752 * XXX: although defrouter_select() should not have a bad effect 1753 * for those are not autoconfigured hosts, we explicitly avoid such 1754 * cases for safety. 1755 */ 1756 if (do_update && router && 1757 ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV) { 1758 /* 1759 * guaranteed recursion 1760 */ 1761 defrouter_select(); 1762 } 1763 1764 return (ln); 1765done: 1766 if (ln != NULL) { 1767 if (flags & ND6_EXCLUSIVE) 1768 LLE_WUNLOCK(ln); 1769 else 1770 LLE_RUNLOCK(ln); 1771 if (static_route) 1772 ln = NULL; 1773 } 1774 return (ln); 1775} 1776 1777static void 1778nd6_slowtimo(void *arg) 1779{ 1780 CURVNET_SET((struct vnet *) arg); 1781 struct nd_ifinfo *nd6if; 1782 struct ifnet *ifp; 1783 1784 callout_reset(&V_nd6_slowtimo_ch, ND6_SLOWTIMER_INTERVAL * hz, 1785 nd6_slowtimo, curvnet); 1786 IFNET_RLOCK_NOSLEEP(); 1787 for (ifp = TAILQ_FIRST(&V_ifnet); ifp; 1788 ifp = TAILQ_NEXT(ifp, if_list)) { 1789 nd6if = ND_IFINFO(ifp); 1790 if (nd6if->basereachable && /* already initialized */ 1791 (nd6if->recalctm -= ND6_SLOWTIMER_INTERVAL) <= 0) { 1792 /* 1793 * Since reachable time rarely changes by router 1794 * advertisements, we SHOULD insure that a new random 1795 * value gets recomputed at least once every few hours. 1796 * (RFC 2461, 6.3.4) 1797 */ 1798 nd6if->recalctm = V_nd6_recalc_reachtm_interval; 1799 nd6if->reachable = ND_COMPUTE_RTIME(nd6if->basereachable); 1800 } 1801 } 1802 IFNET_RUNLOCK_NOSLEEP(); 1803 CURVNET_RESTORE(); 1804} 1805 1806int 1807nd6_output(struct ifnet *ifp, struct ifnet *origifp, struct mbuf *m0, 1808 struct sockaddr_in6 *dst, struct rtentry *rt0) 1809{ 1810 1811 return (nd6_output_lle(ifp, origifp, m0, dst, rt0, NULL, NULL)); 1812} 1813 1814 1815/* 1816 * Note that I'm not enforcing any global serialization 1817 * lle state or asked changes here as the logic is too 1818 * complicated to avoid having to always acquire an exclusive 1819 * lock 1820 * KMM 1821 * 1822 */ 1823#define senderr(e) { error = (e); goto bad;} 1824 1825int 1826nd6_output_lle(struct ifnet *ifp, struct ifnet *origifp, struct mbuf *m0, 1827 struct sockaddr_in6 *dst, struct rtentry *rt0, struct llentry *lle, 1828 struct mbuf **chain) 1829{ 1830 struct mbuf *m = m0; 1831 struct m_tag *mtag; 1832 struct llentry *ln = lle; 1833 struct ip6_hdr *ip6; 1834 int error = 0; 1835 int flags = 0; 1836 int ip6len; 1837 1838#ifdef INVARIANTS 1839 if (lle != NULL) { 1840 1841 LLE_WLOCK_ASSERT(lle); 1842 1843 KASSERT(chain != NULL, (" lle locked but no mbuf chain pointer passed")); 1844 } 1845#endif 1846 if (IN6_IS_ADDR_MULTICAST(&dst->sin6_addr)) 1847 goto sendpkt; 1848 1849 if (nd6_need_cache(ifp) == 0) 1850 goto sendpkt; 1851 1852 /* 1853 * next hop determination. This routine is derived from ether_output. 1854 */ 1855 1856 /* 1857 * Address resolution or Neighbor Unreachability Detection 1858 * for the next hop. 1859 * At this point, the destination of the packet must be a unicast 1860 * or an anycast address(i.e. not a multicast). 1861 */ 1862 1863 flags = ((m != NULL) || (lle != NULL)) ? LLE_EXCLUSIVE : 0; 1864 if (ln == NULL) { 1865 retry: 1866 IF_AFDATA_LOCK(ifp); 1867 ln = lla_lookup(LLTABLE6(ifp), flags, (struct sockaddr *)dst); 1868 IF_AFDATA_UNLOCK(ifp); 1869 if ((ln == NULL) && nd6_is_addr_neighbor(dst, ifp)) { 1870 /* 1871 * Since nd6_is_addr_neighbor() internally calls nd6_lookup(), 1872 * the condition below is not very efficient. But we believe 1873 * it is tolerable, because this should be a rare case. 1874 */ 1875 flags = ND6_CREATE | (m ? ND6_EXCLUSIVE : 0); 1876 IF_AFDATA_LOCK(ifp); 1877 ln = nd6_lookup(&dst->sin6_addr, flags, ifp); 1878 IF_AFDATA_UNLOCK(ifp); 1879 } 1880 } 1881 if (ln == NULL) { 1882 if ((ifp->if_flags & IFF_POINTOPOINT) == 0 && 1883 !(ND_IFINFO(ifp)->flags & ND6_IFF_PERFORMNUD)) { 1884 char ip6buf[INET6_ADDRSTRLEN]; 1885 log(LOG_DEBUG, 1886 "nd6_output: can't allocate llinfo for %s " 1887 "(ln=%p)\n", 1888 ip6_sprintf(ip6buf, &dst->sin6_addr), ln); 1889 senderr(EIO); /* XXX: good error? */ 1890 } 1891 goto sendpkt; /* send anyway */ 1892 } 1893 1894 /* We don't have to do link-layer address resolution on a p2p link. */ 1895 if ((ifp->if_flags & IFF_POINTOPOINT) != 0 && 1896 ln->ln_state < ND6_LLINFO_REACHABLE) { 1897 if ((flags & LLE_EXCLUSIVE) == 0) { 1898 flags |= LLE_EXCLUSIVE; 1899 goto retry; 1900 } 1901 ln->ln_state = ND6_LLINFO_STALE; 1902 nd6_llinfo_settimer_locked(ln, (long)V_nd6_gctimer * hz); 1903 } 1904 1905 /* 1906 * The first time we send a packet to a neighbor whose entry is 1907 * STALE, we have to change the state to DELAY and a sets a timer to 1908 * expire in DELAY_FIRST_PROBE_TIME seconds to ensure do 1909 * neighbor unreachability detection on expiration. 1910 * (RFC 2461 7.3.3) 1911 */ 1912 if (ln->ln_state == ND6_LLINFO_STALE) { 1913 if ((flags & LLE_EXCLUSIVE) == 0) { 1914 flags |= LLE_EXCLUSIVE; 1915 LLE_RUNLOCK(ln); 1916 goto retry; 1917 } 1918 ln->la_asked = 0; 1919 ln->ln_state = ND6_LLINFO_DELAY; 1920 nd6_llinfo_settimer_locked(ln, (long)V_nd6_delay * hz); 1921 } 1922 1923 /* 1924 * If the neighbor cache entry has a state other than INCOMPLETE 1925 * (i.e. its link-layer address is already resolved), just 1926 * send the packet. 1927 */ 1928 if (ln->ln_state > ND6_LLINFO_INCOMPLETE) 1929 goto sendpkt; 1930 1931 /* 1932 * There is a neighbor cache entry, but no ethernet address 1933 * response yet. Append this latest packet to the end of the 1934 * packet queue in the mbuf, unless the number of the packet 1935 * does not exceed nd6_maxqueuelen. When it exceeds nd6_maxqueuelen, 1936 * the oldest packet in the queue will be removed. 1937 */ 1938 if (ln->ln_state == ND6_LLINFO_NOSTATE) 1939 ln->ln_state = ND6_LLINFO_INCOMPLETE; 1940 1941 if ((flags & LLE_EXCLUSIVE) == 0) { 1942 flags |= LLE_EXCLUSIVE; 1943 LLE_RUNLOCK(ln); 1944 goto retry; 1945 } 1946 1947 LLE_WLOCK_ASSERT(ln); 1948 1949 if (ln->la_hold) { 1950 struct mbuf *m_hold; 1951 int i; 1952 1953 i = 0; 1954 for (m_hold = ln->la_hold; m_hold; m_hold = m_hold->m_nextpkt) { 1955 i++; 1956 if (m_hold->m_nextpkt == NULL) { 1957 m_hold->m_nextpkt = m; 1958 break; 1959 } 1960 } 1961 while (i >= V_nd6_maxqueuelen) { 1962 m_hold = ln->la_hold; 1963 ln->la_hold = ln->la_hold->m_nextpkt; 1964 m_freem(m_hold); 1965 i--; 1966 } 1967 } else { 1968 ln->la_hold = m; 1969 } 1970 1971 /* 1972 * If there has been no NS for the neighbor after entering the 1973 * INCOMPLETE state, send the first solicitation. 1974 */ 1975 if (!ND6_LLINFO_PERMANENT(ln) && ln->la_asked == 0) { 1976 ln->la_asked++; 1977 1978 nd6_llinfo_settimer_locked(ln, 1979 (long)ND_IFINFO(ifp)->retrans * hz / 1000); 1980 LLE_WUNLOCK(ln); 1981 nd6_ns_output(ifp, NULL, &dst->sin6_addr, ln, 0); 1982 if (lle != NULL && ln == lle) 1983 LLE_WLOCK(lle); 1984 1985 } else if (lle == NULL || ln != lle) { 1986 /* 1987 * We did the lookup (no lle arg) so we 1988 * need to do the unlock here. 1989 */ 1990 LLE_WUNLOCK(ln); 1991 } 1992 1993 return (0); 1994 1995 sendpkt: 1996 /* discard the packet if IPv6 operation is disabled on the interface */ 1997 if ((ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED)) { 1998 error = ENETDOWN; /* better error? */ 1999 goto bad; 2000 } 2001 /* 2002 * ln is valid and the caller did not pass in 2003 * an llentry 2004 */ 2005 if ((ln != NULL) && (lle == NULL)) { 2006 if (flags & LLE_EXCLUSIVE) 2007 LLE_WUNLOCK(ln); 2008 else 2009 LLE_RUNLOCK(ln); 2010 } 2011 2012#ifdef MAC 2013 mac_netinet6_nd6_send(ifp, m); 2014#endif 2015 2016 /* 2017 * If called from nd6_ns_output() (NS), nd6_na_output() (NA), 2018 * icmp6_redirect_output() (REDIRECT) or from rip6_output() (RS, RA 2019 * as handled by rtsol and rtadvd), mbufs will be tagged for SeND 2020 * to be diverted to user space. When re-injected into the kernel, 2021 * send_output() will directly dispatch them to the outgoing interface. 2022 */ 2023 if (send_sendso_input_hook != NULL) { 2024 mtag = m_tag_find(m, PACKET_TAG_ND_OUTGOING, NULL); 2025 if (mtag != NULL) { 2026 ip6 = mtod(m, struct ip6_hdr *); 2027 ip6len = sizeof(struct ip6_hdr) + ntohs(ip6->ip6_plen); 2028 /* Use the SEND socket */ 2029 error = send_sendso_input_hook(m, ifp, SND_OUT, 2030 ip6len); 2031 /* -1 == no app on SEND socket */ 2032 if (error == 0 || error != -1) 2033 return (error); 2034 } 2035 } 2036 2037 /* 2038 * We were passed in a pointer to an lle with the lock held 2039 * this means that we can't call if_output as we will 2040 * recurse on the lle lock - so what we do is we create 2041 * a list of mbufs to send and transmit them in the caller 2042 * after the lock is dropped 2043 */ 2044 if (lle != NULL) { 2045 if (*chain == NULL) 2046 *chain = m; 2047 else { 2048 struct mbuf *mb; 2049 2050 /* 2051 * append mbuf to end of deferred chain 2052 */ 2053 mb = *chain; 2054 while (mb->m_nextpkt != NULL) 2055 mb = mb->m_nextpkt; 2056 mb->m_nextpkt = m; 2057 } 2058 return (error); 2059 } 2060 if ((ifp->if_flags & IFF_LOOPBACK) != 0) { 2061 return ((*ifp->if_output)(origifp, m, (struct sockaddr *)dst, 2062 NULL)); 2063 } 2064 error = (*ifp->if_output)(ifp, m, (struct sockaddr *)dst, NULL); 2065 return (error); 2066 2067 bad: 2068 /* 2069 * ln is valid and the caller did not pass in 2070 * an llentry 2071 */ 2072 if ((ln != NULL) && (lle == NULL)) { 2073 if (flags & LLE_EXCLUSIVE) 2074 LLE_WUNLOCK(ln); 2075 else 2076 LLE_RUNLOCK(ln); 2077 } 2078 if (m) 2079 m_freem(m); 2080 return (error); 2081} 2082#undef senderr 2083 2084 2085int 2086nd6_output_flush(struct ifnet *ifp, struct ifnet *origifp, struct mbuf *chain, 2087 struct sockaddr_in6 *dst, struct route *ro) 2088{ 2089 struct mbuf *m, *m_head; 2090 struct ifnet *outifp; 2091 int error = 0; 2092 2093 m_head = chain; 2094 if ((ifp->if_flags & IFF_LOOPBACK) != 0) 2095 outifp = origifp; 2096 else 2097 outifp = ifp; 2098 2099 while (m_head) { 2100 m = m_head; 2101 m_head = m_head->m_nextpkt; 2102 error = (*ifp->if_output)(ifp, m, (struct sockaddr *)dst, ro); 2103 } 2104 2105 /* 2106 * XXX 2107 * note that intermediate errors are blindly ignored - but this is 2108 * the same convention as used with nd6_output when called by 2109 * nd6_cache_lladdr 2110 */ 2111 return (error); 2112} 2113 2114 2115int 2116nd6_need_cache(struct ifnet *ifp) 2117{ 2118 /* 2119 * XXX: we currently do not make neighbor cache on any interface 2120 * other than ARCnet, Ethernet, FDDI and GIF. 2121 * 2122 * RFC2893 says: 2123 * - unidirectional tunnels needs no ND 2124 */ 2125 switch (ifp->if_type) { 2126 case IFT_ARCNET: 2127 case IFT_ETHER: 2128 case IFT_FDDI: 2129 case IFT_IEEE1394: 2130#ifdef IFT_L2VLAN 2131 case IFT_L2VLAN: 2132#endif 2133#ifdef IFT_IEEE80211 2134 case IFT_IEEE80211: 2135#endif 2136#ifdef IFT_CARP 2137 case IFT_CARP: 2138#endif 2139 case IFT_INFINIBAND: 2140 case IFT_GIF: /* XXX need more cases? */ 2141 case IFT_PPP: 2142 case IFT_TUNNEL: 2143 case IFT_BRIDGE: 2144 case IFT_PROPVIRTUAL: 2145 return (1); 2146 default: 2147 return (0); 2148 } 2149} 2150 2151/* 2152 * the callers of this function need to be re-worked to drop 2153 * the lle lock, drop here for now 2154 */ 2155int 2156nd6_storelladdr(struct ifnet *ifp, struct mbuf *m, 2157 struct sockaddr *dst, u_char *desten, struct llentry **lle) 2158{ 2159 struct llentry *ln; 2160 2161 *lle = NULL; 2162 IF_AFDATA_UNLOCK_ASSERT(ifp); 2163 if (m->m_flags & M_MCAST) { 2164 int i; 2165 2166 switch (ifp->if_type) { 2167 case IFT_ETHER: 2168 case IFT_FDDI: 2169#ifdef IFT_L2VLAN 2170 case IFT_L2VLAN: 2171#endif 2172#ifdef IFT_IEEE80211 2173 case IFT_IEEE80211: 2174#endif 2175 case IFT_BRIDGE: 2176 case IFT_ISO88025: 2177 ETHER_MAP_IPV6_MULTICAST(&SIN6(dst)->sin6_addr, 2178 desten); 2179 return (0); 2180 case IFT_IEEE1394: 2181 /* 2182 * netbsd can use if_broadcastaddr, but we don't do so 2183 * to reduce # of ifdef. 2184 */ 2185 for (i = 0; i < ifp->if_addrlen; i++) 2186 desten[i] = ~0; 2187 return (0); 2188 case IFT_ARCNET: 2189 *desten = 0; 2190 return (0); 2191 default: 2192 m_freem(m); 2193 return (EAFNOSUPPORT); 2194 } 2195 } 2196 2197 2198 /* 2199 * the entry should have been created in nd6_store_lladdr 2200 */ 2201 IF_AFDATA_LOCK(ifp); 2202 ln = lla_lookup(LLTABLE6(ifp), 0, dst); 2203 IF_AFDATA_UNLOCK(ifp); 2204 if ((ln == NULL) || !(ln->la_flags & LLE_VALID)) { 2205 if (ln != NULL) 2206 LLE_RUNLOCK(ln); 2207 /* this could happen, if we could not allocate memory */ 2208 m_freem(m); 2209 return (1); 2210 } 2211 2212 bcopy(&ln->ll_addr, desten, ifp->if_addrlen); 2213 *lle = ln; 2214 LLE_RUNLOCK(ln); 2215 /* 2216 * A *small* use after free race exists here 2217 */ 2218 return (0); 2219} 2220 2221static void 2222clear_llinfo_pqueue(struct llentry *ln) 2223{ 2224 struct mbuf *m_hold, *m_hold_next; 2225 2226 for (m_hold = ln->la_hold; m_hold; m_hold = m_hold_next) { 2227 m_hold_next = m_hold->m_nextpkt; 2228 m_hold->m_nextpkt = NULL; 2229 m_freem(m_hold); 2230 } 2231 2232 ln->la_hold = NULL; 2233 return; 2234} 2235 2236static int nd6_sysctl_drlist(SYSCTL_HANDLER_ARGS); 2237static int nd6_sysctl_prlist(SYSCTL_HANDLER_ARGS); 2238#ifdef SYSCTL_DECL 2239SYSCTL_DECL(_net_inet6_icmp6); 2240#endif 2241SYSCTL_NODE(_net_inet6_icmp6, ICMPV6CTL_ND6_DRLIST, nd6_drlist, 2242 CTLFLAG_RD, nd6_sysctl_drlist, ""); 2243SYSCTL_NODE(_net_inet6_icmp6, ICMPV6CTL_ND6_PRLIST, nd6_prlist, 2244 CTLFLAG_RD, nd6_sysctl_prlist, ""); 2245SYSCTL_VNET_INT(_net_inet6_icmp6, ICMPV6CTL_ND6_MAXQLEN, nd6_maxqueuelen, 2246 CTLFLAG_RW, &VNET_NAME(nd6_maxqueuelen), 1, ""); 2247 2248static int 2249nd6_sysctl_drlist(SYSCTL_HANDLER_ARGS) 2250{ 2251 int error; 2252 char buf[1024] __aligned(4); 2253 struct in6_defrouter *d, *de; 2254 struct nd_defrouter *dr; 2255 2256 if (req->newptr) 2257 return EPERM; 2258 error = 0; 2259 2260 for (dr = TAILQ_FIRST(&V_nd_defrouter); dr; 2261 dr = TAILQ_NEXT(dr, dr_entry)) { 2262 d = (struct in6_defrouter *)buf; 2263 de = (struct in6_defrouter *)(buf + sizeof(buf)); 2264 2265 if (d + 1 <= de) { 2266 bzero(d, sizeof(*d)); 2267 d->rtaddr.sin6_family = AF_INET6; 2268 d->rtaddr.sin6_len = sizeof(d->rtaddr); 2269 d->rtaddr.sin6_addr = dr->rtaddr; 2270 error = sa6_recoverscope(&d->rtaddr); 2271 if (error != 0) 2272 return (error); 2273 d->flags = dr->flags; 2274 d->rtlifetime = dr->rtlifetime; 2275 d->expire = dr->expire; 2276 d->if_index = dr->ifp->if_index; 2277 } else 2278 panic("buffer too short"); 2279 2280 error = SYSCTL_OUT(req, buf, sizeof(*d)); 2281 if (error) 2282 break; 2283 } 2284 2285 return (error); 2286} 2287 2288static int 2289nd6_sysctl_prlist(SYSCTL_HANDLER_ARGS) 2290{ 2291 int error; 2292 char buf[1024] __aligned(4); 2293 struct in6_prefix *p, *pe; 2294 struct nd_prefix *pr; 2295 char ip6buf[INET6_ADDRSTRLEN]; 2296 2297 if (req->newptr) 2298 return EPERM; 2299 error = 0; 2300 2301 for (pr = V_nd_prefix.lh_first; pr; pr = pr->ndpr_next) { 2302 u_short advrtrs; 2303 size_t advance; 2304 struct sockaddr_in6 *sin6, *s6; 2305 struct nd_pfxrouter *pfr; 2306 2307 p = (struct in6_prefix *)buf; 2308 pe = (struct in6_prefix *)(buf + sizeof(buf)); 2309 2310 if (p + 1 <= pe) { 2311 bzero(p, sizeof(*p)); 2312 sin6 = (struct sockaddr_in6 *)(p + 1); 2313 2314 p->prefix = pr->ndpr_prefix; 2315 if (sa6_recoverscope(&p->prefix)) { 2316 log(LOG_ERR, 2317 "scope error in prefix list (%s)\n", 2318 ip6_sprintf(ip6buf, &p->prefix.sin6_addr)); 2319 /* XXX: press on... */ 2320 } 2321 p->raflags = pr->ndpr_raf; 2322 p->prefixlen = pr->ndpr_plen; 2323 p->vltime = pr->ndpr_vltime; 2324 p->pltime = pr->ndpr_pltime; 2325 p->if_index = pr->ndpr_ifp->if_index; 2326 if (pr->ndpr_vltime == ND6_INFINITE_LIFETIME) 2327 p->expire = 0; 2328 else { 2329 time_t maxexpire; 2330 2331 /* XXX: we assume time_t is signed. */ 2332 maxexpire = (-1) & 2333 ~((time_t)1 << 2334 ((sizeof(maxexpire) * 8) - 1)); 2335 if (pr->ndpr_vltime < 2336 maxexpire - pr->ndpr_lastupdate) { 2337 p->expire = pr->ndpr_lastupdate + 2338 pr->ndpr_vltime; 2339 } else 2340 p->expire = maxexpire; 2341 } 2342 p->refcnt = pr->ndpr_refcnt; 2343 p->flags = pr->ndpr_stateflags; 2344 p->origin = PR_ORIG_RA; 2345 advrtrs = 0; 2346 for (pfr = pr->ndpr_advrtrs.lh_first; pfr; 2347 pfr = pfr->pfr_next) { 2348 if ((void *)&sin6[advrtrs + 1] > (void *)pe) { 2349 advrtrs++; 2350 continue; 2351 } 2352 s6 = &sin6[advrtrs]; 2353 bzero(s6, sizeof(*s6)); 2354 s6->sin6_family = AF_INET6; 2355 s6->sin6_len = sizeof(*sin6); 2356 s6->sin6_addr = pfr->router->rtaddr; 2357 if (sa6_recoverscope(s6)) { 2358 log(LOG_ERR, 2359 "scope error in " 2360 "prefix list (%s)\n", 2361 ip6_sprintf(ip6buf, 2362 &pfr->router->rtaddr)); 2363 } 2364 advrtrs++; 2365 } 2366 p->advrtrs = advrtrs; 2367 } else 2368 panic("buffer too short"); 2369 2370 advance = sizeof(*p) + sizeof(*sin6) * advrtrs; 2371 error = SYSCTL_OUT(req, buf, advance); 2372 if (error) 2373 break; 2374 } 2375 2376 return (error); 2377} 2378