nd6.c revision 198976
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: head/sys/netinet6/nd6.c 198976 2009-11-06 10:07:38Z ume $"); 34 35#include "opt_inet.h" 36#include "opt_inet6.h" 37 38#include <sys/param.h> 39#include <sys/systm.h> 40#include <sys/callout.h> 41#include <sys/malloc.h> 42#include <sys/mbuf.h> 43#include <sys/socket.h> 44#include <sys/sockio.h> 45#include <sys/time.h> 46#include <sys/kernel.h> 47#include <sys/protosw.h> 48#include <sys/errno.h> 49#include <sys/syslog.h> 50#include <sys/lock.h> 51#include <sys/rwlock.h> 52#include <sys/queue.h> 53#include <sys/sysctl.h> 54 55#include <net/if.h> 56#include <net/if_arc.h> 57#include <net/if_dl.h> 58#include <net/if_types.h> 59#include <net/iso88025.h> 60#include <net/fddi.h> 61#include <net/route.h> 62#include <net/vnet.h> 63 64#include <netinet/in.h> 65#include <net/if_llatbl.h> 66#define L3_ADDR_SIN6(le) ((struct sockaddr_in6 *) L3_ADDR(le)) 67#include <netinet/if_ether.h> 68#include <netinet6/in6_var.h> 69#include <netinet/ip6.h> 70#include <netinet6/ip6_var.h> 71#include <netinet6/scope6_var.h> 72#include <netinet6/nd6.h> 73#include <netinet6/in6_ifattach.h> 74#include <netinet/icmp6.h> 75 76#include <sys/limits.h> 77 78#include <security/mac/mac_framework.h> 79 80#define ND6_SLOWTIMER_INTERVAL (60 * 60) /* 1 hour */ 81#define ND6_RECALC_REACHTM_INTERVAL (60 * 120) /* 2 hours */ 82 83#define SIN6(s) ((struct sockaddr_in6 *)s) 84 85VNET_DEFINE(int, nd6_prune); 86VNET_DEFINE(int, nd6_delay); 87VNET_DEFINE(int, nd6_umaxtries); 88VNET_DEFINE(int, nd6_mmaxtries); 89VNET_DEFINE(int, nd6_useloopback); 90VNET_DEFINE(int, nd6_gctimer); 91 92/* preventing too many loops in ND option parsing */ 93static VNET_DEFINE(int, nd6_maxndopt); 94VNET_DEFINE(int, nd6_maxnudhint); 95static VNET_DEFINE(int, nd6_maxqueuelen); 96#define V_nd6_maxndopt VNET(nd6_maxndopt) 97#define V_nd6_maxqueuelen VNET(nd6_maxqueuelen) 98 99VNET_DEFINE(int, nd6_debug); 100 101/* for debugging? */ 102#if 0 103static int nd6_inuse, nd6_allocated; 104#endif 105 106VNET_DEFINE(struct nd_drhead, nd_defrouter); 107VNET_DEFINE(struct nd_prhead, nd_prefix); 108 109VNET_DEFINE(int, nd6_recalc_reachtm_interval); 110#define V_nd6_recalc_reachtm_interval VNET(nd6_recalc_reachtm_interval) 111 112static struct sockaddr_in6 all1_sa; 113 114static int nd6_is_new_addr_neighbor __P((struct sockaddr_in6 *, 115 struct ifnet *)); 116static void nd6_setmtu0(struct ifnet *, struct nd_ifinfo *); 117static void nd6_slowtimo(void *); 118static int regen_tmpaddr(struct in6_ifaddr *); 119static struct llentry *nd6_free(struct llentry *, int); 120static void nd6_llinfo_timer(void *); 121static void clear_llinfo_pqueue(struct llentry *); 122 123static VNET_DEFINE(struct callout, nd6_slowtimo_ch); 124#define V_nd6_slowtimo_ch VNET(nd6_slowtimo_ch) 125 126VNET_DEFINE(struct callout, nd6_timer_ch); 127 128VNET_DECLARE(int, dad_ignore_ns); 129VNET_DECLARE(int, dad_maxtry); 130#define V_dad_ignore_ns VNET(dad_ignore_ns) 131#define V_dad_maxtry VNET(dad_maxtry) 132 133void 134nd6_init(void) 135{ 136 int i; 137 138 V_nd6_prune = 1; /* walk list every 1 seconds */ 139 V_nd6_delay = 5; /* delay first probe time 5 second */ 140 V_nd6_umaxtries = 3; /* maximum unicast query */ 141 V_nd6_mmaxtries = 3; /* maximum multicast query */ 142 V_nd6_useloopback = 1; /* use loopback interface for local traffic */ 143 V_nd6_gctimer = (60 * 60 * 24); /* 1 day: garbage collection timer */ 144 145 /* preventing too many loops in ND option parsing */ 146 V_nd6_maxndopt = 10; /* max # of ND options allowed */ 147 148 V_nd6_maxnudhint = 0; /* max # of subsequent upper layer hints */ 149 V_nd6_maxqueuelen = 1; /* max pkts cached in unresolved ND entries */ 150 151#ifdef ND6_DEBUG 152 V_nd6_debug = 1; 153#else 154 V_nd6_debug = 0; 155#endif 156 157 V_nd6_recalc_reachtm_interval = ND6_RECALC_REACHTM_INTERVAL; 158 159 V_dad_ignore_ns = 0; /* ignore NS in DAD - specwise incorrect*/ 160 V_dad_maxtry = 15; /* max # of *tries* to transmit DAD packet */ 161 162 /* 163 * XXX just to get this to compile KMM 164 */ 165#ifdef notyet 166 V_llinfo_nd6.ln_next = &V_llinfo_nd6; 167 V_llinfo_nd6.ln_prev = &V_llinfo_nd6; 168#endif 169 LIST_INIT(&V_nd_prefix); 170 171 V_ip6_use_tempaddr = 0; 172 V_ip6_temp_preferred_lifetime = DEF_TEMP_PREFERRED_LIFETIME; 173 V_ip6_temp_valid_lifetime = DEF_TEMP_VALID_LIFETIME; 174 V_ip6_temp_regen_advance = TEMPADDR_REGEN_ADVANCE; 175 176 V_ip6_desync_factor = 0; 177 178 all1_sa.sin6_family = AF_INET6; 179 all1_sa.sin6_len = sizeof(struct sockaddr_in6); 180 for (i = 0; i < sizeof(all1_sa.sin6_addr); i++) 181 all1_sa.sin6_addr.s6_addr[i] = 0xff; 182 183 /* initialization of the default router list */ 184 TAILQ_INIT(&V_nd_defrouter); 185 /* start timer */ 186 callout_init(&V_nd6_slowtimo_ch, 0); 187 callout_reset(&V_nd6_slowtimo_ch, ND6_SLOWTIMER_INTERVAL * hz, 188 nd6_slowtimo, curvnet); 189} 190 191 192#ifdef VIMAGE 193void 194nd6_destroy() 195{ 196 197 callout_drain(&V_nd6_slowtimo_ch); 198 callout_drain(&V_nd6_timer_ch); 199} 200#endif 201 202struct nd_ifinfo * 203nd6_ifattach(struct ifnet *ifp) 204{ 205 struct nd_ifinfo *nd; 206 207 nd = (struct nd_ifinfo *)malloc(sizeof(*nd), M_IP6NDP, M_WAITOK); 208 bzero(nd, sizeof(*nd)); 209 210 nd->initialized = 1; 211 212 nd->chlim = IPV6_DEFHLIM; 213 nd->basereachable = REACHABLE_TIME; 214 nd->reachable = ND_COMPUTE_RTIME(nd->basereachable); 215 nd->retrans = RETRANS_TIMER; 216 217 nd->flags = ND6_IFF_PERFORMNUD; 218 219 /* A loopback interface always has ND6_IFF_AUTO_LINKLOCAL. */ 220 if (V_ip6_auto_linklocal || (ifp->if_flags & IFF_LOOPBACK)) 221 nd->flags |= ND6_IFF_AUTO_LINKLOCAL; 222 223 /* A loopback interface does not need to accept RTADV. */ 224 if (V_ip6_accept_rtadv && !(ifp->if_flags & IFF_LOOPBACK)) 225 nd->flags |= ND6_IFF_ACCEPT_RTADV; 226 227 /* XXX: we cannot call nd6_setmtu since ifp is not fully initialized */ 228 nd6_setmtu0(ifp, nd); 229 230 return nd; 231} 232 233void 234nd6_ifdetach(struct nd_ifinfo *nd) 235{ 236 237 free(nd, M_IP6NDP); 238} 239 240/* 241 * Reset ND level link MTU. This function is called when the physical MTU 242 * changes, which means we might have to adjust the ND level MTU. 243 */ 244void 245nd6_setmtu(struct ifnet *ifp) 246{ 247 248 nd6_setmtu0(ifp, ND_IFINFO(ifp)); 249} 250 251/* XXX todo: do not maintain copy of ifp->if_mtu in ndi->maxmtu */ 252void 253nd6_setmtu0(struct ifnet *ifp, struct nd_ifinfo *ndi) 254{ 255 u_int32_t omaxmtu; 256 257 omaxmtu = ndi->maxmtu; 258 259 switch (ifp->if_type) { 260 case IFT_ARCNET: 261 ndi->maxmtu = MIN(ARC_PHDS_MAXMTU, ifp->if_mtu); /* RFC2497 */ 262 break; 263 case IFT_FDDI: 264 ndi->maxmtu = MIN(FDDIIPMTU, ifp->if_mtu); /* RFC2467 */ 265 break; 266 case IFT_ISO88025: 267 ndi->maxmtu = MIN(ISO88025_MAX_MTU, ifp->if_mtu); 268 break; 269 default: 270 ndi->maxmtu = ifp->if_mtu; 271 break; 272 } 273 274 /* 275 * Decreasing the interface MTU under IPV6 minimum MTU may cause 276 * undesirable situation. We thus notify the operator of the change 277 * explicitly. The check for omaxmtu is necessary to restrict the 278 * log to the case of changing the MTU, not initializing it. 279 */ 280 if (omaxmtu >= IPV6_MMTU && ndi->maxmtu < IPV6_MMTU) { 281 log(LOG_NOTICE, "nd6_setmtu0: " 282 "new link MTU on %s (%lu) is too small for IPv6\n", 283 if_name(ifp), (unsigned long)ndi->maxmtu); 284 } 285 286 if (ndi->maxmtu > V_in6_maxmtu) 287 in6_setmaxmtu(); /* check all interfaces just in case */ 288 289} 290 291void 292nd6_option_init(void *opt, int icmp6len, union nd_opts *ndopts) 293{ 294 295 bzero(ndopts, sizeof(*ndopts)); 296 ndopts->nd_opts_search = (struct nd_opt_hdr *)opt; 297 ndopts->nd_opts_last 298 = (struct nd_opt_hdr *)(((u_char *)opt) + icmp6len); 299 300 if (icmp6len == 0) { 301 ndopts->nd_opts_done = 1; 302 ndopts->nd_opts_search = NULL; 303 } 304} 305 306/* 307 * Take one ND option. 308 */ 309struct nd_opt_hdr * 310nd6_option(union nd_opts *ndopts) 311{ 312 struct nd_opt_hdr *nd_opt; 313 int olen; 314 315 if (ndopts == NULL) 316 panic("ndopts == NULL in nd6_option"); 317 if (ndopts->nd_opts_last == NULL) 318 panic("uninitialized ndopts in nd6_option"); 319 if (ndopts->nd_opts_search == NULL) 320 return NULL; 321 if (ndopts->nd_opts_done) 322 return NULL; 323 324 nd_opt = ndopts->nd_opts_search; 325 326 /* make sure nd_opt_len is inside the buffer */ 327 if ((caddr_t)&nd_opt->nd_opt_len >= (caddr_t)ndopts->nd_opts_last) { 328 bzero(ndopts, sizeof(*ndopts)); 329 return NULL; 330 } 331 332 olen = nd_opt->nd_opt_len << 3; 333 if (olen == 0) { 334 /* 335 * Message validation requires that all included 336 * options have a length that is greater than zero. 337 */ 338 bzero(ndopts, sizeof(*ndopts)); 339 return NULL; 340 } 341 342 ndopts->nd_opts_search = (struct nd_opt_hdr *)((caddr_t)nd_opt + olen); 343 if (ndopts->nd_opts_search > ndopts->nd_opts_last) { 344 /* option overruns the end of buffer, invalid */ 345 bzero(ndopts, sizeof(*ndopts)); 346 return NULL; 347 } else if (ndopts->nd_opts_search == ndopts->nd_opts_last) { 348 /* reached the end of options chain */ 349 ndopts->nd_opts_done = 1; 350 ndopts->nd_opts_search = NULL; 351 } 352 return nd_opt; 353} 354 355/* 356 * Parse multiple ND options. 357 * This function is much easier to use, for ND routines that do not need 358 * multiple options of the same type. 359 */ 360int 361nd6_options(union nd_opts *ndopts) 362{ 363 struct nd_opt_hdr *nd_opt; 364 int i = 0; 365 366 if (ndopts == NULL) 367 panic("ndopts == NULL in nd6_options"); 368 if (ndopts->nd_opts_last == NULL) 369 panic("uninitialized ndopts in nd6_options"); 370 if (ndopts->nd_opts_search == NULL) 371 return 0; 372 373 while (1) { 374 nd_opt = nd6_option(ndopts); 375 if (nd_opt == NULL && ndopts->nd_opts_last == NULL) { 376 /* 377 * Message validation requires that all included 378 * options have a length that is greater than zero. 379 */ 380 ICMP6STAT_INC(icp6s_nd_badopt); 381 bzero(ndopts, sizeof(*ndopts)); 382 return -1; 383 } 384 385 if (nd_opt == NULL) 386 goto skip1; 387 388 switch (nd_opt->nd_opt_type) { 389 case ND_OPT_SOURCE_LINKADDR: 390 case ND_OPT_TARGET_LINKADDR: 391 case ND_OPT_MTU: 392 case ND_OPT_REDIRECTED_HEADER: 393 if (ndopts->nd_opt_array[nd_opt->nd_opt_type]) { 394 nd6log((LOG_INFO, 395 "duplicated ND6 option found (type=%d)\n", 396 nd_opt->nd_opt_type)); 397 /* XXX bark? */ 398 } else { 399 ndopts->nd_opt_array[nd_opt->nd_opt_type] 400 = nd_opt; 401 } 402 break; 403 case ND_OPT_PREFIX_INFORMATION: 404 if (ndopts->nd_opt_array[nd_opt->nd_opt_type] == 0) { 405 ndopts->nd_opt_array[nd_opt->nd_opt_type] 406 = nd_opt; 407 } 408 ndopts->nd_opts_pi_end = 409 (struct nd_opt_prefix_info *)nd_opt; 410 break; 411 default: 412 /* 413 * Unknown options must be silently ignored, 414 * to accomodate future extension to the protocol. 415 */ 416 nd6log((LOG_DEBUG, 417 "nd6_options: unsupported option %d - " 418 "option ignored\n", nd_opt->nd_opt_type)); 419 } 420 421skip1: 422 i++; 423 if (i > V_nd6_maxndopt) { 424 ICMP6STAT_INC(icp6s_nd_toomanyopt); 425 nd6log((LOG_INFO, "too many loop in nd opt\n")); 426 break; 427 } 428 429 if (ndopts->nd_opts_done) 430 break; 431 } 432 433 return 0; 434} 435 436/* 437 * ND6 timer routine to handle ND6 entries 438 */ 439void 440nd6_llinfo_settimer_locked(struct llentry *ln, long tick) 441{ 442 if (tick < 0) { 443 ln->la_expire = 0; 444 ln->ln_ntick = 0; 445 callout_stop(&ln->ln_timer_ch); 446 /* 447 * XXX - do we know that there is 448 * callout installed? i.e. are we 449 * guaranteed that we're not dropping 450 * a reference that we did not add? 451 * KMM 452 */ 453 LLE_REMREF(ln); 454 } else { 455 ln->la_expire = time_second + tick / hz; 456 LLE_ADDREF(ln); 457 if (tick > INT_MAX) { 458 ln->ln_ntick = tick - INT_MAX; 459 callout_reset(&ln->ln_timer_ch, INT_MAX, 460 nd6_llinfo_timer, ln); 461 } else { 462 ln->ln_ntick = 0; 463 callout_reset(&ln->ln_timer_ch, tick, 464 nd6_llinfo_timer, ln); 465 } 466 } 467} 468 469void 470nd6_llinfo_settimer(struct llentry *ln, long tick) 471{ 472 473 LLE_WLOCK(ln); 474 nd6_llinfo_settimer_locked(ln, tick); 475 LLE_WUNLOCK(ln); 476} 477 478static void 479nd6_llinfo_timer(void *arg) 480{ 481 struct llentry *ln; 482 struct in6_addr *dst; 483 struct ifnet *ifp; 484 struct nd_ifinfo *ndi = NULL; 485 486 ln = (struct llentry *)arg; 487 if (ln == NULL) { 488 panic("%s: NULL entry!\n", __func__); 489 return; 490 } 491 492 if ((ifp = ((ln->lle_tbl != NULL) ? ln->lle_tbl->llt_ifp : NULL)) == NULL) 493 panic("ln ifp == NULL"); 494 495 CURVNET_SET(ifp->if_vnet); 496 497 if (ln->ln_ntick > 0) { 498 if (ln->ln_ntick > INT_MAX) { 499 ln->ln_ntick -= INT_MAX; 500 nd6_llinfo_settimer(ln, INT_MAX); 501 } else { 502 ln->ln_ntick = 0; 503 nd6_llinfo_settimer(ln, ln->ln_ntick); 504 } 505 goto done; 506 } 507 508 ndi = ND_IFINFO(ifp); 509 dst = &L3_ADDR_SIN6(ln)->sin6_addr; 510 if ((ln->la_flags & LLE_STATIC) || (ln->la_expire > time_second)) { 511 goto done; 512 } 513 514 if (ln->la_flags & LLE_DELETED) { 515 (void)nd6_free(ln, 0); 516 ln = NULL; 517 goto done; 518 } 519 520 switch (ln->ln_state) { 521 case ND6_LLINFO_INCOMPLETE: 522 if (ln->la_asked < V_nd6_mmaxtries) { 523 ln->la_asked++; 524 nd6_llinfo_settimer(ln, (long)ndi->retrans * hz / 1000); 525 nd6_ns_output(ifp, NULL, dst, ln, 0); 526 } else { 527 struct mbuf *m = ln->la_hold; 528 if (m) { 529 struct mbuf *m0; 530 531 /* 532 * assuming every packet in la_hold has the 533 * same IP header 534 */ 535 m0 = m->m_nextpkt; 536 m->m_nextpkt = NULL; 537 icmp6_error2(m, ICMP6_DST_UNREACH, 538 ICMP6_DST_UNREACH_ADDR, 0, ifp); 539 540 ln->la_hold = m0; 541 clear_llinfo_pqueue(ln); 542 } 543 (void)nd6_free(ln, 0); 544 ln = NULL; 545 } 546 break; 547 case ND6_LLINFO_REACHABLE: 548 if (!ND6_LLINFO_PERMANENT(ln)) { 549 ln->ln_state = ND6_LLINFO_STALE; 550 nd6_llinfo_settimer(ln, (long)V_nd6_gctimer * hz); 551 } 552 break; 553 554 case ND6_LLINFO_STALE: 555 /* Garbage Collection(RFC 2461 5.3) */ 556 if (!ND6_LLINFO_PERMANENT(ln)) { 557 (void)nd6_free(ln, 1); 558 ln = NULL; 559 } 560 break; 561 562 case ND6_LLINFO_DELAY: 563 if (ndi && (ndi->flags & ND6_IFF_PERFORMNUD) != 0) { 564 /* We need NUD */ 565 ln->la_asked = 1; 566 ln->ln_state = ND6_LLINFO_PROBE; 567 nd6_llinfo_settimer(ln, (long)ndi->retrans * hz / 1000); 568 nd6_ns_output(ifp, dst, dst, ln, 0); 569 } else { 570 ln->ln_state = ND6_LLINFO_STALE; /* XXX */ 571 nd6_llinfo_settimer(ln, (long)V_nd6_gctimer * hz); 572 } 573 break; 574 case ND6_LLINFO_PROBE: 575 if (ln->la_asked < V_nd6_umaxtries) { 576 ln->la_asked++; 577 nd6_llinfo_settimer(ln, (long)ndi->retrans * hz / 1000); 578 nd6_ns_output(ifp, dst, dst, ln, 0); 579 } else { 580 (void)nd6_free(ln, 0); 581 ln = NULL; 582 } 583 break; 584 } 585 CURVNET_RESTORE(); 586done: 587 if (ln != NULL) 588 LLE_FREE(ln); 589} 590 591 592/* 593 * ND6 timer routine to expire default route list and prefix list 594 */ 595void 596nd6_timer(void *arg) 597{ 598 CURVNET_SET((struct vnet *) arg); 599 int s; 600 struct nd_defrouter *dr; 601 struct nd_prefix *pr; 602 struct in6_ifaddr *ia6, *nia6; 603 struct in6_addrlifetime *lt6; 604 605 callout_reset(&V_nd6_timer_ch, V_nd6_prune * hz, 606 nd6_timer, curvnet); 607 608 /* expire default router list */ 609 s = splnet(); 610 dr = TAILQ_FIRST(&V_nd_defrouter); 611 while (dr) { 612 if (dr->expire && dr->expire < time_second) { 613 struct nd_defrouter *t; 614 t = TAILQ_NEXT(dr, dr_entry); 615 defrtrlist_del(dr); 616 dr = t; 617 } else { 618 dr = TAILQ_NEXT(dr, dr_entry); 619 } 620 } 621 622 /* 623 * expire interface addresses. 624 * in the past the loop was inside prefix expiry processing. 625 * However, from a stricter speci-confrmance standpoint, we should 626 * rather separate address lifetimes and prefix lifetimes. 627 * 628 * XXXRW: in6_ifaddrhead locking. 629 */ 630 addrloop: 631 TAILQ_FOREACH_SAFE(ia6, &V_in6_ifaddrhead, ia_link, nia6) { 632 /* check address lifetime */ 633 lt6 = &ia6->ia6_lifetime; 634 if (IFA6_IS_INVALID(ia6)) { 635 int regen = 0; 636 637 /* 638 * If the expiring address is temporary, try 639 * regenerating a new one. This would be useful when 640 * we suspended a laptop PC, then turned it on after a 641 * period that could invalidate all temporary 642 * addresses. Although we may have to restart the 643 * loop (see below), it must be after purging the 644 * address. Otherwise, we'd see an infinite loop of 645 * regeneration. 646 */ 647 if (V_ip6_use_tempaddr && 648 (ia6->ia6_flags & IN6_IFF_TEMPORARY) != 0) { 649 if (regen_tmpaddr(ia6) == 0) 650 regen = 1; 651 } 652 653 in6_purgeaddr(&ia6->ia_ifa); 654 655 if (regen) 656 goto addrloop; /* XXX: see below */ 657 } else if (IFA6_IS_DEPRECATED(ia6)) { 658 int oldflags = ia6->ia6_flags; 659 660 ia6->ia6_flags |= IN6_IFF_DEPRECATED; 661 662 /* 663 * If a temporary address has just become deprecated, 664 * regenerate a new one if possible. 665 */ 666 if (V_ip6_use_tempaddr && 667 (ia6->ia6_flags & IN6_IFF_TEMPORARY) != 0 && 668 (oldflags & IN6_IFF_DEPRECATED) == 0) { 669 670 if (regen_tmpaddr(ia6) == 0) { 671 /* 672 * A new temporary address is 673 * generated. 674 * XXX: this means the address chain 675 * has changed while we are still in 676 * the loop. Although the change 677 * would not cause disaster (because 678 * it's not a deletion, but an 679 * addition,) we'd rather restart the 680 * loop just for safety. Or does this 681 * significantly reduce performance?? 682 */ 683 goto addrloop; 684 } 685 } 686 } else { 687 /* 688 * A new RA might have made a deprecated address 689 * preferred. 690 */ 691 ia6->ia6_flags &= ~IN6_IFF_DEPRECATED; 692 } 693 } 694 695 /* expire prefix list */ 696 pr = V_nd_prefix.lh_first; 697 while (pr) { 698 /* 699 * check prefix lifetime. 700 * since pltime is just for autoconf, pltime processing for 701 * prefix is not necessary. 702 */ 703 if (pr->ndpr_vltime != ND6_INFINITE_LIFETIME && 704 time_second - pr->ndpr_lastupdate > pr->ndpr_vltime) { 705 struct nd_prefix *t; 706 t = pr->ndpr_next; 707 708 /* 709 * address expiration and prefix expiration are 710 * separate. NEVER perform in6_purgeaddr here. 711 */ 712 713 prelist_remove(pr); 714 pr = t; 715 } else 716 pr = pr->ndpr_next; 717 } 718 splx(s); 719 CURVNET_RESTORE(); 720} 721 722/* 723 * ia6 - deprecated/invalidated temporary address 724 */ 725static int 726regen_tmpaddr(struct in6_ifaddr *ia6) 727{ 728 struct ifaddr *ifa; 729 struct ifnet *ifp; 730 struct in6_ifaddr *public_ifa6 = NULL; 731 732 ifp = ia6->ia_ifa.ifa_ifp; 733 IF_ADDR_LOCK(ifp); 734 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 735 struct in6_ifaddr *it6; 736 737 if (ifa->ifa_addr->sa_family != AF_INET6) 738 continue; 739 740 it6 = (struct in6_ifaddr *)ifa; 741 742 /* ignore no autoconf addresses. */ 743 if ((it6->ia6_flags & IN6_IFF_AUTOCONF) == 0) 744 continue; 745 746 /* ignore autoconf addresses with different prefixes. */ 747 if (it6->ia6_ndpr == NULL || it6->ia6_ndpr != ia6->ia6_ndpr) 748 continue; 749 750 /* 751 * Now we are looking at an autoconf address with the same 752 * prefix as ours. If the address is temporary and is still 753 * preferred, do not create another one. It would be rare, but 754 * could happen, for example, when we resume a laptop PC after 755 * a long period. 756 */ 757 if ((it6->ia6_flags & IN6_IFF_TEMPORARY) != 0 && 758 !IFA6_IS_DEPRECATED(it6)) { 759 public_ifa6 = NULL; 760 break; 761 } 762 763 /* 764 * This is a public autoconf address that has the same prefix 765 * as ours. If it is preferred, keep it. We can't break the 766 * loop here, because there may be a still-preferred temporary 767 * address with the prefix. 768 */ 769 if (!IFA6_IS_DEPRECATED(it6)) 770 public_ifa6 = it6; 771 } 772 773 if (public_ifa6 != NULL) { 774 int e; 775 776 if ((e = in6_tmpifadd(public_ifa6, 0, 0)) != 0) { 777 IF_ADDR_UNLOCK(ifp); 778 log(LOG_NOTICE, "regen_tmpaddr: failed to create a new" 779 " tmp addr,errno=%d\n", e); 780 return (-1); 781 } 782 IF_ADDR_UNLOCK(ifp); 783 return (0); 784 } 785 786 IF_ADDR_UNLOCK(ifp); 787 return (-1); 788} 789 790/* 791 * Nuke neighbor cache/prefix/default router management table, right before 792 * ifp goes away. 793 */ 794void 795nd6_purge(struct ifnet *ifp) 796{ 797 struct nd_defrouter *dr, *ndr; 798 struct nd_prefix *pr, *npr; 799 800 /* 801 * Nuke default router list entries toward ifp. 802 * We defer removal of default router list entries that is installed 803 * in the routing table, in order to keep additional side effects as 804 * small as possible. 805 */ 806 for (dr = TAILQ_FIRST(&V_nd_defrouter); dr; dr = ndr) { 807 ndr = TAILQ_NEXT(dr, dr_entry); 808 if (dr->installed) 809 continue; 810 811 if (dr->ifp == ifp) 812 defrtrlist_del(dr); 813 } 814 815 for (dr = TAILQ_FIRST(&V_nd_defrouter); dr; dr = ndr) { 816 ndr = TAILQ_NEXT(dr, dr_entry); 817 if (!dr->installed) 818 continue; 819 820 if (dr->ifp == ifp) 821 defrtrlist_del(dr); 822 } 823 824 /* Nuke prefix list entries toward ifp */ 825 for (pr = V_nd_prefix.lh_first; pr; pr = npr) { 826 npr = pr->ndpr_next; 827 if (pr->ndpr_ifp == ifp) { 828 /* 829 * Because if_detach() does *not* release prefixes 830 * while purging addresses the reference count will 831 * still be above zero. We therefore reset it to 832 * make sure that the prefix really gets purged. 833 */ 834 pr->ndpr_refcnt = 0; 835 836 /* 837 * Previously, pr->ndpr_addr is removed as well, 838 * but I strongly believe we don't have to do it. 839 * nd6_purge() is only called from in6_ifdetach(), 840 * which removes all the associated interface addresses 841 * by itself. 842 * (jinmei@kame.net 20010129) 843 */ 844 prelist_remove(pr); 845 } 846 } 847 848 /* cancel default outgoing interface setting */ 849 if (V_nd6_defifindex == ifp->if_index) 850 nd6_setdefaultiface(0); 851 852 if (!V_ip6_forwarding && ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV) { 853 /* Refresh default router list. */ 854 defrouter_select(); 855 } 856 857 /* XXXXX 858 * We do not nuke the neighbor cache entries here any more 859 * because the neighbor cache is kept in if_afdata[AF_INET6]. 860 * nd6_purge() is invoked by in6_ifdetach() which is called 861 * from if_detach() where everything gets purged. So let 862 * in6_domifdetach() do the actual L2 table purging work. 863 */ 864} 865 866/* 867 * the caller acquires and releases the lock on the lltbls 868 * Returns the llentry locked 869 */ 870struct llentry * 871nd6_lookup(struct in6_addr *addr6, int flags, struct ifnet *ifp) 872{ 873 struct sockaddr_in6 sin6; 874 struct llentry *ln; 875 int llflags = 0; 876 877 bzero(&sin6, sizeof(sin6)); 878 sin6.sin6_len = sizeof(struct sockaddr_in6); 879 sin6.sin6_family = AF_INET6; 880 sin6.sin6_addr = *addr6; 881 882 IF_AFDATA_LOCK_ASSERT(ifp); 883 884 if (flags & ND6_CREATE) 885 llflags |= LLE_CREATE; 886 if (flags & ND6_EXCLUSIVE) 887 llflags |= LLE_EXCLUSIVE; 888 889 ln = lla_lookup(LLTABLE6(ifp), llflags, (struct sockaddr *)&sin6); 890 if ((ln != NULL) && (flags & LLE_CREATE)) { 891 ln->ln_state = ND6_LLINFO_NOSTATE; 892 callout_init(&ln->ln_timer_ch, 0); 893 } 894 895 return (ln); 896} 897 898/* 899 * Test whether a given IPv6 address is a neighbor or not, ignoring 900 * the actual neighbor cache. The neighbor cache is ignored in order 901 * to not reenter the routing code from within itself. 902 */ 903static int 904nd6_is_new_addr_neighbor(struct sockaddr_in6 *addr, struct ifnet *ifp) 905{ 906 struct nd_prefix *pr; 907 struct ifaddr *dstaddr; 908 909 /* 910 * A link-local address is always a neighbor. 911 * XXX: a link does not necessarily specify a single interface. 912 */ 913 if (IN6_IS_ADDR_LINKLOCAL(&addr->sin6_addr)) { 914 struct sockaddr_in6 sin6_copy; 915 u_int32_t zone; 916 917 /* 918 * We need sin6_copy since sa6_recoverscope() may modify the 919 * content (XXX). 920 */ 921 sin6_copy = *addr; 922 if (sa6_recoverscope(&sin6_copy)) 923 return (0); /* XXX: should be impossible */ 924 if (in6_setscope(&sin6_copy.sin6_addr, ifp, &zone)) 925 return (0); 926 if (sin6_copy.sin6_scope_id == zone) 927 return (1); 928 else 929 return (0); 930 } 931 932 /* 933 * If the address matches one of our addresses, 934 * it should be a neighbor. 935 * If the address matches one of our on-link prefixes, it should be a 936 * neighbor. 937 */ 938 for (pr = V_nd_prefix.lh_first; pr; pr = pr->ndpr_next) { 939 if (pr->ndpr_ifp != ifp) 940 continue; 941 942 if (!(pr->ndpr_stateflags & NDPRF_ONLINK)) 943 continue; 944 945 if (IN6_ARE_MASKED_ADDR_EQUAL(&pr->ndpr_prefix.sin6_addr, 946 &addr->sin6_addr, &pr->ndpr_mask)) 947 return (1); 948 } 949 950 /* 951 * If the address is assigned on the node of the other side of 952 * a p2p interface, the address should be a neighbor. 953 */ 954 dstaddr = ifa_ifwithdstaddr((struct sockaddr *)addr); 955 if (dstaddr != NULL) { 956 if (dstaddr->ifa_ifp == ifp) { 957 ifa_free(dstaddr); 958 return (1); 959 } 960 ifa_free(dstaddr); 961 } 962 963 /* 964 * If the default router list is empty, all addresses are regarded 965 * as on-link, and thus, as a neighbor. 966 * XXX: we restrict the condition to hosts, because routers usually do 967 * not have the "default router list". 968 */ 969 if (!V_ip6_forwarding && TAILQ_FIRST(&V_nd_defrouter) == NULL && 970 V_nd6_defifindex == ifp->if_index) { 971 return (1); 972 } 973 974 return (0); 975} 976 977 978/* 979 * Detect if a given IPv6 address identifies a neighbor on a given link. 980 * XXX: should take care of the destination of a p2p link? 981 */ 982int 983nd6_is_addr_neighbor(struct sockaddr_in6 *addr, struct ifnet *ifp) 984{ 985 struct llentry *lle; 986 int rc = 0; 987 988 IF_AFDATA_UNLOCK_ASSERT(ifp); 989 if (nd6_is_new_addr_neighbor(addr, ifp)) 990 return (1); 991 992 /* 993 * Even if the address matches none of our addresses, it might be 994 * in the neighbor cache. 995 */ 996 IF_AFDATA_LOCK(ifp); 997 if ((lle = nd6_lookup(&addr->sin6_addr, 0, ifp)) != NULL) { 998 LLE_RUNLOCK(lle); 999 rc = 1; 1000 } 1001 IF_AFDATA_UNLOCK(ifp); 1002 return (rc); 1003} 1004 1005/* 1006 * Free an nd6 llinfo entry. 1007 * Since the function would cause significant changes in the kernel, DO NOT 1008 * make it global, unless you have a strong reason for the change, and are sure 1009 * that the change is safe. 1010 */ 1011static struct llentry * 1012nd6_free(struct llentry *ln, int gc) 1013{ 1014 struct llentry *next; 1015 struct nd_defrouter *dr; 1016 struct ifnet *ifp=NULL; 1017 1018 /* 1019 * we used to have pfctlinput(PRC_HOSTDEAD) here. 1020 * even though it is not harmful, it was not really necessary. 1021 */ 1022 1023 /* cancel timer */ 1024 nd6_llinfo_settimer(ln, -1); 1025 1026 if (!V_ip6_forwarding) { 1027 int s; 1028 s = splnet(); 1029 dr = defrouter_lookup(&L3_ADDR_SIN6(ln)->sin6_addr, ln->lle_tbl->llt_ifp); 1030 1031 if (dr != NULL && dr->expire && 1032 ln->ln_state == ND6_LLINFO_STALE && gc) { 1033 /* 1034 * If the reason for the deletion is just garbage 1035 * collection, and the neighbor is an active default 1036 * router, do not delete it. Instead, reset the GC 1037 * timer using the router's lifetime. 1038 * Simply deleting the entry would affect default 1039 * router selection, which is not necessarily a good 1040 * thing, especially when we're using router preference 1041 * values. 1042 * XXX: the check for ln_state would be redundant, 1043 * but we intentionally keep it just in case. 1044 */ 1045 if (dr->expire > time_second) 1046 nd6_llinfo_settimer(ln, 1047 (dr->expire - time_second) * hz); 1048 else 1049 nd6_llinfo_settimer(ln, (long)V_nd6_gctimer * hz); 1050 splx(s); 1051 return (LIST_NEXT(ln, lle_next)); 1052 } 1053 1054 if (ln->ln_router || dr) { 1055 /* 1056 * rt6_flush must be called whether or not the neighbor 1057 * is in the Default Router List. 1058 * See a corresponding comment in nd6_na_input(). 1059 */ 1060 rt6_flush(&L3_ADDR_SIN6(ln)->sin6_addr, ln->lle_tbl->llt_ifp); 1061 } 1062 1063 if (dr) { 1064 /* 1065 * Unreachablity of a router might affect the default 1066 * router selection and on-link detection of advertised 1067 * prefixes. 1068 */ 1069 1070 /* 1071 * Temporarily fake the state to choose a new default 1072 * router and to perform on-link determination of 1073 * prefixes correctly. 1074 * Below the state will be set correctly, 1075 * or the entry itself will be deleted. 1076 */ 1077 ln->ln_state = ND6_LLINFO_INCOMPLETE; 1078 1079 /* 1080 * Since defrouter_select() does not affect the 1081 * on-link determination and MIP6 needs the check 1082 * before the default router selection, we perform 1083 * the check now. 1084 */ 1085 pfxlist_onlink_check(); 1086 1087 /* 1088 * refresh default router list 1089 */ 1090 defrouter_select(); 1091 } 1092 splx(s); 1093 } 1094 1095 /* 1096 * Before deleting the entry, remember the next entry as the 1097 * return value. We need this because pfxlist_onlink_check() above 1098 * might have freed other entries (particularly the old next entry) as 1099 * a side effect (XXX). 1100 */ 1101 next = LIST_NEXT(ln, lle_next); 1102 1103 ifp = ln->lle_tbl->llt_ifp; 1104 IF_AFDATA_LOCK(ifp); 1105 LLE_WLOCK(ln); 1106 llentry_free(ln); 1107 IF_AFDATA_UNLOCK(ifp); 1108 1109 return (next); 1110} 1111 1112/* 1113 * Upper-layer reachability hint for Neighbor Unreachability Detection. 1114 * 1115 * XXX cost-effective methods? 1116 */ 1117void 1118nd6_nud_hint(struct rtentry *rt, struct in6_addr *dst6, int force) 1119{ 1120 struct llentry *ln; 1121 struct ifnet *ifp; 1122 1123 if ((dst6 == NULL) || (rt == NULL)) 1124 return; 1125 1126 ifp = rt->rt_ifp; 1127 IF_AFDATA_LOCK(ifp); 1128 ln = nd6_lookup(dst6, ND6_EXCLUSIVE, NULL); 1129 IF_AFDATA_UNLOCK(ifp); 1130 if (ln == NULL) 1131 return; 1132 1133 if (ln->ln_state < ND6_LLINFO_REACHABLE) 1134 goto done; 1135 1136 /* 1137 * if we get upper-layer reachability confirmation many times, 1138 * it is possible we have false information. 1139 */ 1140 if (!force) { 1141 ln->ln_byhint++; 1142 if (ln->ln_byhint > V_nd6_maxnudhint) { 1143 goto done; 1144 } 1145 } 1146 1147 ln->ln_state = ND6_LLINFO_REACHABLE; 1148 if (!ND6_LLINFO_PERMANENT(ln)) { 1149 nd6_llinfo_settimer(ln, 1150 (long)ND_IFINFO(rt->rt_ifp)->reachable * hz); 1151 } 1152done: 1153 LLE_WUNLOCK(ln); 1154} 1155 1156 1157int 1158nd6_ioctl(u_long cmd, caddr_t data, struct ifnet *ifp) 1159{ 1160 struct in6_drlist *drl = (struct in6_drlist *)data; 1161 struct in6_oprlist *oprl = (struct in6_oprlist *)data; 1162 struct in6_ndireq *ndi = (struct in6_ndireq *)data; 1163 struct in6_nbrinfo *nbi = (struct in6_nbrinfo *)data; 1164 struct in6_ndifreq *ndif = (struct in6_ndifreq *)data; 1165 struct nd_defrouter *dr; 1166 struct nd_prefix *pr; 1167 int i = 0, error = 0; 1168 int s; 1169 1170 switch (cmd) { 1171 case SIOCGDRLST_IN6: 1172 /* 1173 * obsolete API, use sysctl under net.inet6.icmp6 1174 */ 1175 bzero(drl, sizeof(*drl)); 1176 s = splnet(); 1177 dr = TAILQ_FIRST(&V_nd_defrouter); 1178 while (dr && i < DRLSTSIZ) { 1179 drl->defrouter[i].rtaddr = dr->rtaddr; 1180 in6_clearscope(&drl->defrouter[i].rtaddr); 1181 1182 drl->defrouter[i].flags = dr->flags; 1183 drl->defrouter[i].rtlifetime = dr->rtlifetime; 1184 drl->defrouter[i].expire = dr->expire; 1185 drl->defrouter[i].if_index = dr->ifp->if_index; 1186 i++; 1187 dr = TAILQ_NEXT(dr, dr_entry); 1188 } 1189 splx(s); 1190 break; 1191 case SIOCGPRLST_IN6: 1192 /* 1193 * obsolete API, use sysctl under net.inet6.icmp6 1194 * 1195 * XXX the structure in6_prlist was changed in backward- 1196 * incompatible manner. in6_oprlist is used for SIOCGPRLST_IN6, 1197 * in6_prlist is used for nd6_sysctl() - fill_prlist(). 1198 */ 1199 /* 1200 * XXX meaning of fields, especialy "raflags", is very 1201 * differnet between RA prefix list and RR/static prefix list. 1202 * how about separating ioctls into two? 1203 */ 1204 bzero(oprl, sizeof(*oprl)); 1205 s = splnet(); 1206 pr = V_nd_prefix.lh_first; 1207 while (pr && i < PRLSTSIZ) { 1208 struct nd_pfxrouter *pfr; 1209 int j; 1210 1211 oprl->prefix[i].prefix = pr->ndpr_prefix.sin6_addr; 1212 oprl->prefix[i].raflags = pr->ndpr_raf; 1213 oprl->prefix[i].prefixlen = pr->ndpr_plen; 1214 oprl->prefix[i].vltime = pr->ndpr_vltime; 1215 oprl->prefix[i].pltime = pr->ndpr_pltime; 1216 oprl->prefix[i].if_index = pr->ndpr_ifp->if_index; 1217 if (pr->ndpr_vltime == ND6_INFINITE_LIFETIME) 1218 oprl->prefix[i].expire = 0; 1219 else { 1220 time_t maxexpire; 1221 1222 /* XXX: we assume time_t is signed. */ 1223 maxexpire = (-1) & 1224 ~((time_t)1 << 1225 ((sizeof(maxexpire) * 8) - 1)); 1226 if (pr->ndpr_vltime < 1227 maxexpire - pr->ndpr_lastupdate) { 1228 oprl->prefix[i].expire = 1229 pr->ndpr_lastupdate + 1230 pr->ndpr_vltime; 1231 } else 1232 oprl->prefix[i].expire = maxexpire; 1233 } 1234 1235 pfr = pr->ndpr_advrtrs.lh_first; 1236 j = 0; 1237 while (pfr) { 1238 if (j < DRLSTSIZ) { 1239#define RTRADDR oprl->prefix[i].advrtr[j] 1240 RTRADDR = pfr->router->rtaddr; 1241 in6_clearscope(&RTRADDR); 1242#undef RTRADDR 1243 } 1244 j++; 1245 pfr = pfr->pfr_next; 1246 } 1247 oprl->prefix[i].advrtrs = j; 1248 oprl->prefix[i].origin = PR_ORIG_RA; 1249 1250 i++; 1251 pr = pr->ndpr_next; 1252 } 1253 splx(s); 1254 1255 break; 1256 case OSIOCGIFINFO_IN6: 1257#define ND ndi->ndi 1258 /* XXX: old ndp(8) assumes a positive value for linkmtu. */ 1259 bzero(&ND, sizeof(ND)); 1260 ND.linkmtu = IN6_LINKMTU(ifp); 1261 ND.maxmtu = ND_IFINFO(ifp)->maxmtu; 1262 ND.basereachable = ND_IFINFO(ifp)->basereachable; 1263 ND.reachable = ND_IFINFO(ifp)->reachable; 1264 ND.retrans = ND_IFINFO(ifp)->retrans; 1265 ND.flags = ND_IFINFO(ifp)->flags; 1266 ND.recalctm = ND_IFINFO(ifp)->recalctm; 1267 ND.chlim = ND_IFINFO(ifp)->chlim; 1268 break; 1269 case SIOCGIFINFO_IN6: 1270 ND = *ND_IFINFO(ifp); 1271 break; 1272 case SIOCSIFINFO_IN6: 1273 /* 1274 * used to change host variables from userland. 1275 * intented for a use on router to reflect RA configurations. 1276 */ 1277 /* 0 means 'unspecified' */ 1278 if (ND.linkmtu != 0) { 1279 if (ND.linkmtu < IPV6_MMTU || 1280 ND.linkmtu > IN6_LINKMTU(ifp)) { 1281 error = EINVAL; 1282 break; 1283 } 1284 ND_IFINFO(ifp)->linkmtu = ND.linkmtu; 1285 } 1286 1287 if (ND.basereachable != 0) { 1288 int obasereachable = ND_IFINFO(ifp)->basereachable; 1289 1290 ND_IFINFO(ifp)->basereachable = ND.basereachable; 1291 if (ND.basereachable != obasereachable) 1292 ND_IFINFO(ifp)->reachable = 1293 ND_COMPUTE_RTIME(ND.basereachable); 1294 } 1295 if (ND.retrans != 0) 1296 ND_IFINFO(ifp)->retrans = ND.retrans; 1297 if (ND.chlim != 0) 1298 ND_IFINFO(ifp)->chlim = ND.chlim; 1299 /* FALLTHROUGH */ 1300 case SIOCSIFINFO_FLAGS: 1301 { 1302 struct ifaddr *ifa; 1303 struct in6_ifaddr *ia; 1304 1305 if ((ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) && 1306 !(ND.flags & ND6_IFF_IFDISABLED)) { 1307 /* ifdisabled 1->0 transision */ 1308 1309 /* 1310 * If the interface is marked as ND6_IFF_IFDISABLED and 1311 * has an link-local address with IN6_IFF_DUPLICATED, 1312 * do not clear ND6_IFF_IFDISABLED. 1313 * See RFC 4862, Section 5.4.5. 1314 */ 1315 int duplicated_linklocal = 0; 1316 1317 IF_ADDR_LOCK(ifp); 1318 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1319 if (ifa->ifa_addr->sa_family != AF_INET6) 1320 continue; 1321 ia = (struct in6_ifaddr *)ifa; 1322 if ((ia->ia6_flags & IN6_IFF_DUPLICATED) && 1323 IN6_IS_ADDR_LINKLOCAL(&ia->ia_addr.sin6_addr)) { 1324 duplicated_linklocal = 1; 1325 break; 1326 } 1327 } 1328 IF_ADDR_UNLOCK(ifp); 1329 1330 if (duplicated_linklocal) { 1331 ND.flags |= ND6_IFF_IFDISABLED; 1332 log(LOG_ERR, "Cannot enable an interface" 1333 " with a link-local address marked" 1334 " duplicate.\n"); 1335 } else { 1336 ND_IFINFO(ifp)->flags &= ~ND6_IFF_IFDISABLED; 1337 in6_if_up(ifp); 1338 } 1339 } else if (!(ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) && 1340 (ND.flags & ND6_IFF_IFDISABLED)) { 1341 /* ifdisabled 0->1 transision */ 1342 /* Mark all IPv6 address as tentative. */ 1343 1344 ND_IFINFO(ifp)->flags |= ND6_IFF_IFDISABLED; 1345 IF_ADDR_LOCK(ifp); 1346 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1347 if (ifa->ifa_addr->sa_family != AF_INET6) 1348 continue; 1349 ia = (struct in6_ifaddr *)ifa; 1350 ia->ia6_flags |= IN6_IFF_TENTATIVE; 1351 } 1352 IF_ADDR_UNLOCK(ifp); 1353 } 1354 1355 if (!(ND_IFINFO(ifp)->flags & ND6_IFF_AUTO_LINKLOCAL) && 1356 (ND.flags & ND6_IFF_AUTO_LINKLOCAL)) { 1357 /* auto_linklocal 0->1 transision */ 1358 1359 /* If no link-local address on ifp, configure */ 1360 ND_IFINFO(ifp)->flags |= ND6_IFF_AUTO_LINKLOCAL; 1361 in6_ifattach(ifp, NULL); 1362 } 1363 } 1364 ND_IFINFO(ifp)->flags = ND.flags; 1365 break; 1366#undef ND 1367 case SIOCSNDFLUSH_IN6: /* XXX: the ioctl name is confusing... */ 1368 /* sync kernel routing table with the default router list */ 1369 defrouter_reset(); 1370 defrouter_select(); 1371 break; 1372 case SIOCSPFXFLUSH_IN6: 1373 { 1374 /* flush all the prefix advertised by routers */ 1375 struct nd_prefix *pr, *next; 1376 1377 s = splnet(); 1378 for (pr = V_nd_prefix.lh_first; pr; pr = next) { 1379 struct in6_ifaddr *ia, *ia_next; 1380 1381 next = pr->ndpr_next; 1382 1383 if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr)) 1384 continue; /* XXX */ 1385 1386 /* do we really have to remove addresses as well? */ 1387 /* XXXRW: in6_ifaddrhead locking. */ 1388 TAILQ_FOREACH_SAFE(ia, &V_in6_ifaddrhead, ia_link, 1389 ia_next) { 1390 if ((ia->ia6_flags & IN6_IFF_AUTOCONF) == 0) 1391 continue; 1392 1393 if (ia->ia6_ndpr == pr) 1394 in6_purgeaddr(&ia->ia_ifa); 1395 } 1396 prelist_remove(pr); 1397 } 1398 splx(s); 1399 break; 1400 } 1401 case SIOCSRTRFLUSH_IN6: 1402 { 1403 /* flush all the default routers */ 1404 struct nd_defrouter *dr, *next; 1405 1406 s = splnet(); 1407 defrouter_reset(); 1408 for (dr = TAILQ_FIRST(&V_nd_defrouter); dr; dr = next) { 1409 next = TAILQ_NEXT(dr, dr_entry); 1410 defrtrlist_del(dr); 1411 } 1412 defrouter_select(); 1413 splx(s); 1414 break; 1415 } 1416 case SIOCGNBRINFO_IN6: 1417 { 1418 struct llentry *ln; 1419 struct in6_addr nb_addr = nbi->addr; /* make local for safety */ 1420 1421 if ((error = in6_setscope(&nb_addr, ifp, NULL)) != 0) 1422 return (error); 1423 1424 IF_AFDATA_LOCK(ifp); 1425 ln = nd6_lookup(&nb_addr, 0, ifp); 1426 IF_AFDATA_UNLOCK(ifp); 1427 1428 if (ln == NULL) { 1429 error = EINVAL; 1430 break; 1431 } 1432 nbi->state = ln->ln_state; 1433 nbi->asked = ln->la_asked; 1434 nbi->isrouter = ln->ln_router; 1435 nbi->expire = ln->la_expire; 1436 LLE_RUNLOCK(ln); 1437 break; 1438 } 1439 case SIOCGDEFIFACE_IN6: /* XXX: should be implemented as a sysctl? */ 1440 ndif->ifindex = V_nd6_defifindex; 1441 break; 1442 case SIOCSDEFIFACE_IN6: /* XXX: should be implemented as a sysctl? */ 1443 return (nd6_setdefaultiface(ndif->ifindex)); 1444 } 1445 return (error); 1446} 1447 1448/* 1449 * Create neighbor cache entry and cache link-layer address, 1450 * on reception of inbound ND6 packets. (RS/RA/NS/redirect) 1451 * 1452 * type - ICMP6 type 1453 * code - type dependent information 1454 * 1455 * XXXXX 1456 * The caller of this function already acquired the ndp 1457 * cache table lock because the cache entry is returned. 1458 */ 1459struct llentry * 1460nd6_cache_lladdr(struct ifnet *ifp, struct in6_addr *from, char *lladdr, 1461 int lladdrlen, int type, int code) 1462{ 1463 struct llentry *ln = NULL; 1464 int is_newentry; 1465 int do_update; 1466 int olladdr; 1467 int llchange; 1468 int flags = 0; 1469 int newstate = 0; 1470 uint16_t router = 0; 1471 struct sockaddr_in6 sin6; 1472 struct mbuf *chain = NULL; 1473 int static_route = 0; 1474 1475 IF_AFDATA_UNLOCK_ASSERT(ifp); 1476 1477 if (ifp == NULL) 1478 panic("ifp == NULL in nd6_cache_lladdr"); 1479 if (from == NULL) 1480 panic("from == NULL in nd6_cache_lladdr"); 1481 1482 /* nothing must be updated for unspecified address */ 1483 if (IN6_IS_ADDR_UNSPECIFIED(from)) 1484 return NULL; 1485 1486 /* 1487 * Validation about ifp->if_addrlen and lladdrlen must be done in 1488 * the caller. 1489 * 1490 * XXX If the link does not have link-layer adderss, what should 1491 * we do? (ifp->if_addrlen == 0) 1492 * Spec says nothing in sections for RA, RS and NA. There's small 1493 * description on it in NS section (RFC 2461 7.2.3). 1494 */ 1495 flags |= lladdr ? ND6_EXCLUSIVE : 0; 1496 IF_AFDATA_LOCK(ifp); 1497 ln = nd6_lookup(from, flags, ifp); 1498 1499 if (ln == NULL) { 1500 flags |= LLE_EXCLUSIVE; 1501 ln = nd6_lookup(from, flags |ND6_CREATE, ifp); 1502 IF_AFDATA_UNLOCK(ifp); 1503 is_newentry = 1; 1504 } else { 1505 IF_AFDATA_UNLOCK(ifp); 1506 /* do nothing if static ndp is set */ 1507 if (ln->la_flags & LLE_STATIC) { 1508 static_route = 1; 1509 goto done; 1510 } 1511 is_newentry = 0; 1512 } 1513 if (ln == NULL) 1514 return (NULL); 1515 1516 olladdr = (ln->la_flags & LLE_VALID) ? 1 : 0; 1517 if (olladdr && lladdr) { 1518 llchange = bcmp(lladdr, &ln->ll_addr, 1519 ifp->if_addrlen); 1520 } else 1521 llchange = 0; 1522 1523 /* 1524 * newentry olladdr lladdr llchange (*=record) 1525 * 0 n n -- (1) 1526 * 0 y n -- (2) 1527 * 0 n y -- (3) * STALE 1528 * 0 y y n (4) * 1529 * 0 y y y (5) * STALE 1530 * 1 -- n -- (6) NOSTATE(= PASSIVE) 1531 * 1 -- y -- (7) * STALE 1532 */ 1533 1534 if (lladdr) { /* (3-5) and (7) */ 1535 /* 1536 * Record source link-layer address 1537 * XXX is it dependent to ifp->if_type? 1538 */ 1539 bcopy(lladdr, &ln->ll_addr, ifp->if_addrlen); 1540 ln->la_flags |= LLE_VALID; 1541 } 1542 1543 if (!is_newentry) { 1544 if ((!olladdr && lladdr != NULL) || /* (3) */ 1545 (olladdr && lladdr != NULL && llchange)) { /* (5) */ 1546 do_update = 1; 1547 newstate = ND6_LLINFO_STALE; 1548 } else /* (1-2,4) */ 1549 do_update = 0; 1550 } else { 1551 do_update = 1; 1552 if (lladdr == NULL) /* (6) */ 1553 newstate = ND6_LLINFO_NOSTATE; 1554 else /* (7) */ 1555 newstate = ND6_LLINFO_STALE; 1556 } 1557 1558 if (do_update) { 1559 /* 1560 * Update the state of the neighbor cache. 1561 */ 1562 ln->ln_state = newstate; 1563 1564 if (ln->ln_state == ND6_LLINFO_STALE) { 1565 /* 1566 * XXX: since nd6_output() below will cause 1567 * state tansition to DELAY and reset the timer, 1568 * we must set the timer now, although it is actually 1569 * meaningless. 1570 */ 1571 nd6_llinfo_settimer_locked(ln, (long)V_nd6_gctimer * hz); 1572 1573 if (ln->la_hold) { 1574 struct mbuf *m_hold, *m_hold_next; 1575 1576 /* 1577 * reset the la_hold in advance, to explicitly 1578 * prevent a la_hold lookup in nd6_output() 1579 * (wouldn't happen, though...) 1580 */ 1581 for (m_hold = ln->la_hold, ln->la_hold = NULL; 1582 m_hold; m_hold = m_hold_next) { 1583 m_hold_next = m_hold->m_nextpkt; 1584 m_hold->m_nextpkt = NULL; 1585 1586 /* 1587 * we assume ifp is not a p2p here, so 1588 * just set the 2nd argument as the 1589 * 1st one. 1590 */ 1591 nd6_output_lle(ifp, ifp, m_hold, L3_ADDR_SIN6(ln), NULL, ln, &chain); 1592 } 1593 /* 1594 * If we have mbufs in the chain we need to do 1595 * deferred transmit. Copy the address from the 1596 * llentry before dropping the lock down below. 1597 */ 1598 if (chain != NULL) 1599 memcpy(&sin6, L3_ADDR_SIN6(ln), sizeof(sin6)); 1600 } 1601 } else if (ln->ln_state == ND6_LLINFO_INCOMPLETE) { 1602 /* probe right away */ 1603 nd6_llinfo_settimer_locked((void *)ln, 0); 1604 } 1605 } 1606 1607 /* 1608 * ICMP6 type dependent behavior. 1609 * 1610 * NS: clear IsRouter if new entry 1611 * RS: clear IsRouter 1612 * RA: set IsRouter if there's lladdr 1613 * redir: clear IsRouter if new entry 1614 * 1615 * RA case, (1): 1616 * The spec says that we must set IsRouter in the following cases: 1617 * - If lladdr exist, set IsRouter. This means (1-5). 1618 * - If it is old entry (!newentry), set IsRouter. This means (7). 1619 * So, based on the spec, in (1-5) and (7) cases we must set IsRouter. 1620 * A quetion arises for (1) case. (1) case has no lladdr in the 1621 * neighbor cache, this is similar to (6). 1622 * This case is rare but we figured that we MUST NOT set IsRouter. 1623 * 1624 * newentry olladdr lladdr llchange NS RS RA redir 1625 * D R 1626 * 0 n n -- (1) c ? s 1627 * 0 y n -- (2) c s s 1628 * 0 n y -- (3) c s s 1629 * 0 y y n (4) c s s 1630 * 0 y y y (5) c s s 1631 * 1 -- n -- (6) c c c s 1632 * 1 -- y -- (7) c c s c s 1633 * 1634 * (c=clear s=set) 1635 */ 1636 switch (type & 0xff) { 1637 case ND_NEIGHBOR_SOLICIT: 1638 /* 1639 * New entry must have is_router flag cleared. 1640 */ 1641 if (is_newentry) /* (6-7) */ 1642 ln->ln_router = 0; 1643 break; 1644 case ND_REDIRECT: 1645 /* 1646 * If the icmp is a redirect to a better router, always set the 1647 * is_router flag. Otherwise, if the entry is newly created, 1648 * clear the flag. [RFC 2461, sec 8.3] 1649 */ 1650 if (code == ND_REDIRECT_ROUTER) 1651 ln->ln_router = 1; 1652 else if (is_newentry) /* (6-7) */ 1653 ln->ln_router = 0; 1654 break; 1655 case ND_ROUTER_SOLICIT: 1656 /* 1657 * is_router flag must always be cleared. 1658 */ 1659 ln->ln_router = 0; 1660 break; 1661 case ND_ROUTER_ADVERT: 1662 /* 1663 * Mark an entry with lladdr as a router. 1664 */ 1665 if ((!is_newentry && (olladdr || lladdr)) || /* (2-5) */ 1666 (is_newentry && lladdr)) { /* (7) */ 1667 ln->ln_router = 1; 1668 } 1669 break; 1670 } 1671 1672 if (ln != NULL) { 1673 static_route = (ln->la_flags & LLE_STATIC); 1674 router = ln->ln_router; 1675 1676 if (flags & ND6_EXCLUSIVE) 1677 LLE_WUNLOCK(ln); 1678 else 1679 LLE_RUNLOCK(ln); 1680 if (static_route) 1681 ln = NULL; 1682 } 1683 if (chain) 1684 nd6_output_flush(ifp, ifp, chain, &sin6, NULL); 1685 1686 /* 1687 * When the link-layer address of a router changes, select the 1688 * best router again. In particular, when the neighbor entry is newly 1689 * created, it might affect the selection policy. 1690 * Question: can we restrict the first condition to the "is_newentry" 1691 * case? 1692 * XXX: when we hear an RA from a new router with the link-layer 1693 * address option, defrouter_select() is called twice, since 1694 * defrtrlist_update called the function as well. However, I believe 1695 * we can compromise the overhead, since it only happens the first 1696 * time. 1697 * XXX: although defrouter_select() should not have a bad effect 1698 * for those are not autoconfigured hosts, we explicitly avoid such 1699 * cases for safety. 1700 */ 1701 if (do_update && router && !V_ip6_forwarding && 1702 ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV) { 1703 /* 1704 * guaranteed recursion 1705 */ 1706 defrouter_select(); 1707 } 1708 1709 return (ln); 1710done: 1711 if (ln != NULL) { 1712 if (flags & ND6_EXCLUSIVE) 1713 LLE_WUNLOCK(ln); 1714 else 1715 LLE_RUNLOCK(ln); 1716 if (static_route) 1717 ln = NULL; 1718 } 1719 return (ln); 1720} 1721 1722static void 1723nd6_slowtimo(void *arg) 1724{ 1725 CURVNET_SET((struct vnet *) arg); 1726 struct nd_ifinfo *nd6if; 1727 struct ifnet *ifp; 1728 1729 callout_reset(&V_nd6_slowtimo_ch, ND6_SLOWTIMER_INTERVAL * hz, 1730 nd6_slowtimo, curvnet); 1731 IFNET_RLOCK_NOSLEEP(); 1732 for (ifp = TAILQ_FIRST(&V_ifnet); ifp; 1733 ifp = TAILQ_NEXT(ifp, if_list)) { 1734 nd6if = ND_IFINFO(ifp); 1735 if (nd6if->basereachable && /* already initialized */ 1736 (nd6if->recalctm -= ND6_SLOWTIMER_INTERVAL) <= 0) { 1737 /* 1738 * Since reachable time rarely changes by router 1739 * advertisements, we SHOULD insure that a new random 1740 * value gets recomputed at least once every few hours. 1741 * (RFC 2461, 6.3.4) 1742 */ 1743 nd6if->recalctm = V_nd6_recalc_reachtm_interval; 1744 nd6if->reachable = ND_COMPUTE_RTIME(nd6if->basereachable); 1745 } 1746 } 1747 IFNET_RUNLOCK_NOSLEEP(); 1748 CURVNET_RESTORE(); 1749} 1750 1751int 1752nd6_output(struct ifnet *ifp, struct ifnet *origifp, struct mbuf *m0, 1753 struct sockaddr_in6 *dst, struct rtentry *rt0) 1754{ 1755 1756 return (nd6_output_lle(ifp, origifp, m0, dst, rt0, NULL, NULL)); 1757} 1758 1759 1760/* 1761 * Note that I'm not enforcing any global serialization 1762 * lle state or asked changes here as the logic is too 1763 * complicated to avoid having to always acquire an exclusive 1764 * lock 1765 * KMM 1766 * 1767 */ 1768#define senderr(e) { error = (e); goto bad;} 1769 1770int 1771nd6_output_lle(struct ifnet *ifp, struct ifnet *origifp, struct mbuf *m0, 1772 struct sockaddr_in6 *dst, struct rtentry *rt0, struct llentry *lle, 1773 struct mbuf **chain) 1774{ 1775 struct mbuf *m = m0; 1776 struct llentry *ln = lle; 1777 int error = 0; 1778 int flags = 0; 1779 1780#ifdef INVARIANTS 1781 if (lle != NULL) { 1782 1783 LLE_WLOCK_ASSERT(lle); 1784 1785 KASSERT(chain != NULL, (" lle locked but no mbuf chain pointer passed")); 1786 } 1787#endif 1788 if (IN6_IS_ADDR_MULTICAST(&dst->sin6_addr)) 1789 goto sendpkt; 1790 1791 if (nd6_need_cache(ifp) == 0) 1792 goto sendpkt; 1793 1794 /* 1795 * next hop determination. This routine is derived from ether_output. 1796 */ 1797 1798 /* 1799 * Address resolution or Neighbor Unreachability Detection 1800 * for the next hop. 1801 * At this point, the destination of the packet must be a unicast 1802 * or an anycast address(i.e. not a multicast). 1803 */ 1804 1805 flags = ((m != NULL) || (lle != NULL)) ? LLE_EXCLUSIVE : 0; 1806 if (ln == NULL) { 1807 retry: 1808 IF_AFDATA_LOCK(ifp); 1809 ln = lla_lookup(LLTABLE6(ifp), flags, (struct sockaddr *)dst); 1810 IF_AFDATA_UNLOCK(ifp); 1811 if ((ln == NULL) && nd6_is_addr_neighbor(dst, ifp)) { 1812 /* 1813 * Since nd6_is_addr_neighbor() internally calls nd6_lookup(), 1814 * the condition below is not very efficient. But we believe 1815 * it is tolerable, because this should be a rare case. 1816 */ 1817 flags = ND6_CREATE | (m ? ND6_EXCLUSIVE : 0); 1818 IF_AFDATA_LOCK(ifp); 1819 ln = nd6_lookup(&dst->sin6_addr, flags, ifp); 1820 IF_AFDATA_UNLOCK(ifp); 1821 } 1822 } 1823 if (ln == NULL) { 1824 if ((ifp->if_flags & IFF_POINTOPOINT) == 0 && 1825 !(ND_IFINFO(ifp)->flags & ND6_IFF_PERFORMNUD)) { 1826 char ip6buf[INET6_ADDRSTRLEN]; 1827 log(LOG_DEBUG, 1828 "nd6_output: can't allocate llinfo for %s " 1829 "(ln=%p)\n", 1830 ip6_sprintf(ip6buf, &dst->sin6_addr), ln); 1831 senderr(EIO); /* XXX: good error? */ 1832 } 1833 goto sendpkt; /* send anyway */ 1834 } 1835 1836 /* We don't have to do link-layer address resolution on a p2p link. */ 1837 if ((ifp->if_flags & IFF_POINTOPOINT) != 0 && 1838 ln->ln_state < ND6_LLINFO_REACHABLE) { 1839 if ((flags & LLE_EXCLUSIVE) == 0) { 1840 flags |= LLE_EXCLUSIVE; 1841 goto retry; 1842 } 1843 ln->ln_state = ND6_LLINFO_STALE; 1844 nd6_llinfo_settimer_locked(ln, (long)V_nd6_gctimer * hz); 1845 } 1846 1847 /* 1848 * The first time we send a packet to a neighbor whose entry is 1849 * STALE, we have to change the state to DELAY and a sets a timer to 1850 * expire in DELAY_FIRST_PROBE_TIME seconds to ensure do 1851 * neighbor unreachability detection on expiration. 1852 * (RFC 2461 7.3.3) 1853 */ 1854 if (ln->ln_state == ND6_LLINFO_STALE) { 1855 if ((flags & LLE_EXCLUSIVE) == 0) { 1856 flags |= LLE_EXCLUSIVE; 1857 LLE_RUNLOCK(ln); 1858 goto retry; 1859 } 1860 ln->la_asked = 0; 1861 ln->ln_state = ND6_LLINFO_DELAY; 1862 nd6_llinfo_settimer_locked(ln, (long)V_nd6_delay * hz); 1863 } 1864 1865 /* 1866 * If the neighbor cache entry has a state other than INCOMPLETE 1867 * (i.e. its link-layer address is already resolved), just 1868 * send the packet. 1869 */ 1870 if (ln->ln_state > ND6_LLINFO_INCOMPLETE) 1871 goto sendpkt; 1872 1873 /* 1874 * There is a neighbor cache entry, but no ethernet address 1875 * response yet. Append this latest packet to the end of the 1876 * packet queue in the mbuf, unless the number of the packet 1877 * does not exceed nd6_maxqueuelen. When it exceeds nd6_maxqueuelen, 1878 * the oldest packet in the queue will be removed. 1879 */ 1880 if (ln->ln_state == ND6_LLINFO_NOSTATE) 1881 ln->ln_state = ND6_LLINFO_INCOMPLETE; 1882 1883 if ((flags & LLE_EXCLUSIVE) == 0) { 1884 flags |= LLE_EXCLUSIVE; 1885 LLE_RUNLOCK(ln); 1886 goto retry; 1887 } 1888 if (ln->la_hold) { 1889 struct mbuf *m_hold; 1890 int i; 1891 1892 i = 0; 1893 for (m_hold = ln->la_hold; m_hold; m_hold = m_hold->m_nextpkt) { 1894 i++; 1895 if (m_hold->m_nextpkt == NULL) { 1896 m_hold->m_nextpkt = m; 1897 break; 1898 } 1899 } 1900 while (i >= V_nd6_maxqueuelen) { 1901 m_hold = ln->la_hold; 1902 ln->la_hold = ln->la_hold->m_nextpkt; 1903 m_freem(m_hold); 1904 i--; 1905 } 1906 } else { 1907 ln->la_hold = m; 1908 } 1909 /* 1910 * We did the lookup (no lle arg) so we 1911 * need to do the unlock here 1912 */ 1913 if (lle == NULL) { 1914 if (flags & LLE_EXCLUSIVE) 1915 LLE_WUNLOCK(ln); 1916 else 1917 LLE_RUNLOCK(ln); 1918 } 1919 1920 /* 1921 * If there has been no NS for the neighbor after entering the 1922 * INCOMPLETE state, send the first solicitation. 1923 */ 1924 if (!ND6_LLINFO_PERMANENT(ln) && ln->la_asked == 0) { 1925 ln->la_asked++; 1926 1927 nd6_llinfo_settimer(ln, 1928 (long)ND_IFINFO(ifp)->retrans * hz / 1000); 1929 nd6_ns_output(ifp, NULL, &dst->sin6_addr, ln, 0); 1930 } 1931 return (0); 1932 1933 sendpkt: 1934 /* discard the packet if IPv6 operation is disabled on the interface */ 1935 if ((ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED)) { 1936 error = ENETDOWN; /* better error? */ 1937 goto bad; 1938 } 1939 /* 1940 * ln is valid and the caller did not pass in 1941 * an llentry 1942 */ 1943 if ((ln != NULL) && (lle == NULL)) { 1944 if (flags & LLE_EXCLUSIVE) 1945 LLE_WUNLOCK(ln); 1946 else 1947 LLE_RUNLOCK(ln); 1948 } 1949 1950#ifdef MAC 1951 mac_netinet6_nd6_send(ifp, m); 1952#endif 1953 /* 1954 * We were passed in a pointer to an lle with the lock held 1955 * this means that we can't call if_output as we will 1956 * recurse on the lle lock - so what we do is we create 1957 * a list of mbufs to send and transmit them in the caller 1958 * after the lock is dropped 1959 */ 1960 if (lle != NULL) { 1961 if (*chain == NULL) 1962 *chain = m; 1963 else { 1964 struct mbuf *m = *chain; 1965 1966 /* 1967 * append mbuf to end of deferred chain 1968 */ 1969 while (m->m_nextpkt != NULL) 1970 m = m->m_nextpkt; 1971 m->m_nextpkt = m; 1972 } 1973 return (error); 1974 } 1975 if ((ifp->if_flags & IFF_LOOPBACK) != 0) { 1976 return ((*ifp->if_output)(origifp, m, (struct sockaddr *)dst, 1977 NULL)); 1978 } 1979 error = (*ifp->if_output)(ifp, m, (struct sockaddr *)dst, NULL); 1980 return (error); 1981 1982 bad: 1983 /* 1984 * ln is valid and the caller did not pass in 1985 * an llentry 1986 */ 1987 if ((ln != NULL) && (lle == NULL)) { 1988 if (flags & LLE_EXCLUSIVE) 1989 LLE_WUNLOCK(ln); 1990 else 1991 LLE_RUNLOCK(ln); 1992 } 1993 if (m) 1994 m_freem(m); 1995 return (error); 1996} 1997#undef senderr 1998 1999 2000int 2001nd6_output_flush(struct ifnet *ifp, struct ifnet *origifp, struct mbuf *chain, 2002 struct sockaddr_in6 *dst, struct route *ro) 2003{ 2004 struct mbuf *m, *m_head; 2005 struct ifnet *outifp; 2006 int error = 0; 2007 2008 m_head = chain; 2009 if ((ifp->if_flags & IFF_LOOPBACK) != 0) 2010 outifp = origifp; 2011 else 2012 outifp = ifp; 2013 2014 while (m_head) { 2015 m = m_head; 2016 m_head = m_head->m_nextpkt; 2017 error = (*ifp->if_output)(ifp, m, (struct sockaddr *)dst, ro); 2018 } 2019 2020 /* 2021 * XXX 2022 * note that intermediate errors are blindly ignored - but this is 2023 * the same convention as used with nd6_output when called by 2024 * nd6_cache_lladdr 2025 */ 2026 return (error); 2027} 2028 2029 2030int 2031nd6_need_cache(struct ifnet *ifp) 2032{ 2033 /* 2034 * XXX: we currently do not make neighbor cache on any interface 2035 * other than ARCnet, Ethernet, FDDI and GIF. 2036 * 2037 * RFC2893 says: 2038 * - unidirectional tunnels needs no ND 2039 */ 2040 switch (ifp->if_type) { 2041 case IFT_ARCNET: 2042 case IFT_ETHER: 2043 case IFT_FDDI: 2044 case IFT_IEEE1394: 2045#ifdef IFT_L2VLAN 2046 case IFT_L2VLAN: 2047#endif 2048#ifdef IFT_IEEE80211 2049 case IFT_IEEE80211: 2050#endif 2051#ifdef IFT_CARP 2052 case IFT_CARP: 2053#endif 2054 case IFT_GIF: /* XXX need more cases? */ 2055 case IFT_PPP: 2056 case IFT_TUNNEL: 2057 case IFT_BRIDGE: 2058 case IFT_PROPVIRTUAL: 2059 return (1); 2060 default: 2061 return (0); 2062 } 2063} 2064 2065/* 2066 * the callers of this function need to be re-worked to drop 2067 * the lle lock, drop here for now 2068 */ 2069int 2070nd6_storelladdr(struct ifnet *ifp, struct mbuf *m, 2071 struct sockaddr *dst, u_char *desten, struct llentry **lle) 2072{ 2073 struct llentry *ln; 2074 2075 *lle = NULL; 2076 IF_AFDATA_UNLOCK_ASSERT(ifp); 2077 if (m->m_flags & M_MCAST) { 2078 int i; 2079 2080 switch (ifp->if_type) { 2081 case IFT_ETHER: 2082 case IFT_FDDI: 2083#ifdef IFT_L2VLAN 2084 case IFT_L2VLAN: 2085#endif 2086#ifdef IFT_IEEE80211 2087 case IFT_IEEE80211: 2088#endif 2089 case IFT_BRIDGE: 2090 case IFT_ISO88025: 2091 ETHER_MAP_IPV6_MULTICAST(&SIN6(dst)->sin6_addr, 2092 desten); 2093 return (0); 2094 case IFT_IEEE1394: 2095 /* 2096 * netbsd can use if_broadcastaddr, but we don't do so 2097 * to reduce # of ifdef. 2098 */ 2099 for (i = 0; i < ifp->if_addrlen; i++) 2100 desten[i] = ~0; 2101 return (0); 2102 case IFT_ARCNET: 2103 *desten = 0; 2104 return (0); 2105 default: 2106 m_freem(m); 2107 return (EAFNOSUPPORT); 2108 } 2109 } 2110 2111 2112 /* 2113 * the entry should have been created in nd6_store_lladdr 2114 */ 2115 IF_AFDATA_LOCK(ifp); 2116 ln = lla_lookup(LLTABLE6(ifp), 0, dst); 2117 IF_AFDATA_UNLOCK(ifp); 2118 if ((ln == NULL) || !(ln->la_flags & LLE_VALID)) { 2119 if (ln != NULL) 2120 LLE_RUNLOCK(ln); 2121 /* this could happen, if we could not allocate memory */ 2122 m_freem(m); 2123 return (1); 2124 } 2125 2126 bcopy(&ln->ll_addr, desten, ifp->if_addrlen); 2127 *lle = ln; 2128 LLE_RUNLOCK(ln); 2129 /* 2130 * A *small* use after free race exists here 2131 */ 2132 return (0); 2133} 2134 2135static void 2136clear_llinfo_pqueue(struct llentry *ln) 2137{ 2138 struct mbuf *m_hold, *m_hold_next; 2139 2140 for (m_hold = ln->la_hold; m_hold; m_hold = m_hold_next) { 2141 m_hold_next = m_hold->m_nextpkt; 2142 m_hold->m_nextpkt = NULL; 2143 m_freem(m_hold); 2144 } 2145 2146 ln->la_hold = NULL; 2147 return; 2148} 2149 2150static int nd6_sysctl_drlist(SYSCTL_HANDLER_ARGS); 2151static int nd6_sysctl_prlist(SYSCTL_HANDLER_ARGS); 2152#ifdef SYSCTL_DECL 2153SYSCTL_DECL(_net_inet6_icmp6); 2154#endif 2155SYSCTL_NODE(_net_inet6_icmp6, ICMPV6CTL_ND6_DRLIST, nd6_drlist, 2156 CTLFLAG_RD, nd6_sysctl_drlist, ""); 2157SYSCTL_NODE(_net_inet6_icmp6, ICMPV6CTL_ND6_PRLIST, nd6_prlist, 2158 CTLFLAG_RD, nd6_sysctl_prlist, ""); 2159SYSCTL_VNET_INT(_net_inet6_icmp6, ICMPV6CTL_ND6_MAXQLEN, nd6_maxqueuelen, 2160 CTLFLAG_RW, &VNET_NAME(nd6_maxqueuelen), 1, ""); 2161 2162static int 2163nd6_sysctl_drlist(SYSCTL_HANDLER_ARGS) 2164{ 2165 int error; 2166 char buf[1024] __aligned(4); 2167 struct in6_defrouter *d, *de; 2168 struct nd_defrouter *dr; 2169 2170 if (req->newptr) 2171 return EPERM; 2172 error = 0; 2173 2174 for (dr = TAILQ_FIRST(&V_nd_defrouter); dr; 2175 dr = TAILQ_NEXT(dr, dr_entry)) { 2176 d = (struct in6_defrouter *)buf; 2177 de = (struct in6_defrouter *)(buf + sizeof(buf)); 2178 2179 if (d + 1 <= de) { 2180 bzero(d, sizeof(*d)); 2181 d->rtaddr.sin6_family = AF_INET6; 2182 d->rtaddr.sin6_len = sizeof(d->rtaddr); 2183 d->rtaddr.sin6_addr = dr->rtaddr; 2184 error = sa6_recoverscope(&d->rtaddr); 2185 if (error != 0) 2186 return (error); 2187 d->flags = dr->flags; 2188 d->rtlifetime = dr->rtlifetime; 2189 d->expire = dr->expire; 2190 d->if_index = dr->ifp->if_index; 2191 } else 2192 panic("buffer too short"); 2193 2194 error = SYSCTL_OUT(req, buf, sizeof(*d)); 2195 if (error) 2196 break; 2197 } 2198 2199 return (error); 2200} 2201 2202static int 2203nd6_sysctl_prlist(SYSCTL_HANDLER_ARGS) 2204{ 2205 int error; 2206 char buf[1024] __aligned(4); 2207 struct in6_prefix *p, *pe; 2208 struct nd_prefix *pr; 2209 char ip6buf[INET6_ADDRSTRLEN]; 2210 2211 if (req->newptr) 2212 return EPERM; 2213 error = 0; 2214 2215 for (pr = V_nd_prefix.lh_first; pr; pr = pr->ndpr_next) { 2216 u_short advrtrs; 2217 size_t advance; 2218 struct sockaddr_in6 *sin6, *s6; 2219 struct nd_pfxrouter *pfr; 2220 2221 p = (struct in6_prefix *)buf; 2222 pe = (struct in6_prefix *)(buf + sizeof(buf)); 2223 2224 if (p + 1 <= pe) { 2225 bzero(p, sizeof(*p)); 2226 sin6 = (struct sockaddr_in6 *)(p + 1); 2227 2228 p->prefix = pr->ndpr_prefix; 2229 if (sa6_recoverscope(&p->prefix)) { 2230 log(LOG_ERR, 2231 "scope error in prefix list (%s)\n", 2232 ip6_sprintf(ip6buf, &p->prefix.sin6_addr)); 2233 /* XXX: press on... */ 2234 } 2235 p->raflags = pr->ndpr_raf; 2236 p->prefixlen = pr->ndpr_plen; 2237 p->vltime = pr->ndpr_vltime; 2238 p->pltime = pr->ndpr_pltime; 2239 p->if_index = pr->ndpr_ifp->if_index; 2240 if (pr->ndpr_vltime == ND6_INFINITE_LIFETIME) 2241 p->expire = 0; 2242 else { 2243 time_t maxexpire; 2244 2245 /* XXX: we assume time_t is signed. */ 2246 maxexpire = (-1) & 2247 ~((time_t)1 << 2248 ((sizeof(maxexpire) * 8) - 1)); 2249 if (pr->ndpr_vltime < 2250 maxexpire - pr->ndpr_lastupdate) { 2251 p->expire = pr->ndpr_lastupdate + 2252 pr->ndpr_vltime; 2253 } else 2254 p->expire = maxexpire; 2255 } 2256 p->refcnt = pr->ndpr_refcnt; 2257 p->flags = pr->ndpr_stateflags; 2258 p->origin = PR_ORIG_RA; 2259 advrtrs = 0; 2260 for (pfr = pr->ndpr_advrtrs.lh_first; pfr; 2261 pfr = pfr->pfr_next) { 2262 if ((void *)&sin6[advrtrs + 1] > (void *)pe) { 2263 advrtrs++; 2264 continue; 2265 } 2266 s6 = &sin6[advrtrs]; 2267 bzero(s6, sizeof(*s6)); 2268 s6->sin6_family = AF_INET6; 2269 s6->sin6_len = sizeof(*sin6); 2270 s6->sin6_addr = pfr->router->rtaddr; 2271 if (sa6_recoverscope(s6)) { 2272 log(LOG_ERR, 2273 "scope error in " 2274 "prefix list (%s)\n", 2275 ip6_sprintf(ip6buf, 2276 &pfr->router->rtaddr)); 2277 } 2278 advrtrs++; 2279 } 2280 p->advrtrs = advrtrs; 2281 } else 2282 panic("buffer too short"); 2283 2284 advance = sizeof(*p) + sizeof(*sin6) * advrtrs; 2285 error = SYSCTL_OUT(req, buf, advance); 2286 if (error) 2287 break; 2288 } 2289 2290 return (error); 2291} 2292