nd6.c revision 191337
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 191337 2009-04-20 21:37:46Z rwatson $"); 34 35#include "opt_inet.h" 36#include "opt_inet6.h" 37#include "opt_mac.h" 38#include "opt_route.h" 39 40#include <sys/param.h> 41#include <sys/systm.h> 42#include <sys/callout.h> 43#include <sys/malloc.h> 44#include <sys/mbuf.h> 45#include <sys/socket.h> 46#include <sys/sockio.h> 47#include <sys/time.h> 48#include <sys/kernel.h> 49#include <sys/protosw.h> 50#include <sys/errno.h> 51#include <sys/syslog.h> 52#include <sys/lock.h> 53#include <sys/rwlock.h> 54#include <sys/queue.h> 55#include <sys/sysctl.h> 56#include <sys/vimage.h> 57 58#include <net/if.h> 59#include <net/if_arc.h> 60#include <net/if_dl.h> 61#include <net/if_types.h> 62#include <net/iso88025.h> 63#include <net/fddi.h> 64#include <net/route.h> 65#include <net/vnet.h> 66 67#include <netinet/in.h> 68#include <net/if_llatbl.h> 69#define L3_ADDR_SIN6(le) ((struct sockaddr_in6 *) L3_ADDR(le)) 70#include <netinet/if_ether.h> 71#include <netinet6/in6_var.h> 72#include <netinet/ip6.h> 73#include <netinet6/ip6_var.h> 74#include <netinet6/scope6_var.h> 75#include <netinet6/nd6.h> 76#include <netinet/icmp6.h> 77#include <netinet6/vinet6.h> 78 79#include <sys/limits.h> 80 81#include <security/mac/mac_framework.h> 82 83#define ND6_SLOWTIMER_INTERVAL (60 * 60) /* 1 hour */ 84#define ND6_RECALC_REACHTM_INTERVAL (60 * 120) /* 2 hours */ 85 86#define SIN6(s) ((struct sockaddr_in6 *)s) 87 88#ifdef VIMAGE_GLOBALS 89int nd6_prune; 90int nd6_delay; 91int nd6_umaxtries; 92int nd6_mmaxtries; 93int nd6_useloopback; 94int nd6_gctimer; 95 96/* preventing too many loops in ND option parsing */ 97int nd6_maxndopt; 98 99int nd6_maxnudhint; 100int nd6_maxqueuelen; 101 102int nd6_debug; 103 104/* for debugging? */ 105#if 0 106static int nd6_inuse, nd6_allocated; 107#endif 108 109struct nd_drhead nd_defrouter; 110struct nd_prhead nd_prefix; 111 112int nd6_recalc_reachtm_interval; 113#endif /* VIMAGE_GLOBALS */ 114 115static struct sockaddr_in6 all1_sa; 116 117static int nd6_is_new_addr_neighbor __P((struct sockaddr_in6 *, 118 struct ifnet *)); 119static void nd6_setmtu0(struct ifnet *, struct nd_ifinfo *); 120static void nd6_slowtimo(void *); 121static int regen_tmpaddr(struct in6_ifaddr *); 122static struct llentry *nd6_free(struct llentry *, int); 123static void nd6_llinfo_timer(void *); 124static void clear_llinfo_pqueue(struct llentry *); 125 126#ifdef VIMAGE_GLOBALS 127struct callout nd6_slowtimo_ch; 128struct callout nd6_timer_ch; 129extern struct callout in6_tmpaddrtimer_ch; 130extern int dad_ignore_ns; 131extern int dad_maxtry; 132#endif 133 134void 135nd6_init(void) 136{ 137 INIT_VNET_INET6(curvnet); 138 static int nd6_init_done = 0; 139 int i; 140 141 if (nd6_init_done) { 142 log(LOG_NOTICE, "nd6_init called more than once(ignored)\n"); 143 return; 144 } 145 146 V_nd6_prune = 1; /* walk list every 1 seconds */ 147 V_nd6_delay = 5; /* delay first probe time 5 second */ 148 V_nd6_umaxtries = 3; /* maximum unicast query */ 149 V_nd6_mmaxtries = 3; /* maximum multicast query */ 150 V_nd6_useloopback = 1; /* use loopback interface for local traffic */ 151 V_nd6_gctimer = (60 * 60 * 24); /* 1 day: garbage collection timer */ 152 153 /* preventing too many loops in ND option parsing */ 154 V_nd6_maxndopt = 10; /* max # of ND options allowed */ 155 156 V_nd6_maxnudhint = 0; /* max # of subsequent upper layer hints */ 157 V_nd6_maxqueuelen = 1; /* max pkts cached in unresolved ND entries */ 158 159#ifdef ND6_DEBUG 160 V_nd6_debug = 1; 161#else 162 V_nd6_debug = 0; 163#endif 164 165 V_nd6_recalc_reachtm_interval = ND6_RECALC_REACHTM_INTERVAL; 166 167 V_dad_ignore_ns = 0; /* ignore NS in DAD - specwise incorrect*/ 168 V_dad_maxtry = 15; /* max # of *tries* to transmit DAD packet */ 169 170 /* 171 * XXX just to get this to compile KMM 172 */ 173#ifdef notyet 174 V_llinfo_nd6.ln_next = &V_llinfo_nd6; 175 V_llinfo_nd6.ln_prev = &V_llinfo_nd6; 176#endif 177 LIST_INIT(&V_nd_prefix); 178 179 V_ip6_use_tempaddr = 0; 180 V_ip6_temp_preferred_lifetime = DEF_TEMP_PREFERRED_LIFETIME; 181 V_ip6_temp_valid_lifetime = DEF_TEMP_VALID_LIFETIME; 182 V_ip6_temp_regen_advance = TEMPADDR_REGEN_ADVANCE; 183 184 all1_sa.sin6_family = AF_INET6; 185 all1_sa.sin6_len = sizeof(struct sockaddr_in6); 186 for (i = 0; i < sizeof(all1_sa.sin6_addr); i++) 187 all1_sa.sin6_addr.s6_addr[i] = 0xff; 188 189 /* initialization of the default router list */ 190 TAILQ_INIT(&V_nd_defrouter); 191 /* start timer */ 192 callout_init(&V_nd6_slowtimo_ch, 0); 193 callout_reset(&V_nd6_slowtimo_ch, ND6_SLOWTIMER_INTERVAL * hz, 194 nd6_slowtimo, NULL); 195 196 nd6_init_done = 1; 197 198} 199 200struct nd_ifinfo * 201nd6_ifattach(struct ifnet *ifp) 202{ 203 struct nd_ifinfo *nd; 204 205 nd = (struct nd_ifinfo *)malloc(sizeof(*nd), M_IP6NDP, M_WAITOK); 206 bzero(nd, sizeof(*nd)); 207 208 nd->initialized = 1; 209 210 nd->chlim = IPV6_DEFHLIM; 211 nd->basereachable = REACHABLE_TIME; 212 nd->reachable = ND_COMPUTE_RTIME(nd->basereachable); 213 nd->retrans = RETRANS_TIMER; 214 /* 215 * Note that the default value of ip6_accept_rtadv is 0, which means 216 * we won't accept RAs by default even if we set ND6_IFF_ACCEPT_RTADV 217 * here. 218 */ 219 nd->flags = (ND6_IFF_PERFORMNUD | ND6_IFF_ACCEPT_RTADV); 220 221 /* XXX: we cannot call nd6_setmtu since ifp is not fully initialized */ 222 nd6_setmtu0(ifp, nd); 223 224 return nd; 225} 226 227void 228nd6_ifdetach(struct nd_ifinfo *nd) 229{ 230 231 free(nd, M_IP6NDP); 232} 233 234/* 235 * Reset ND level link MTU. This function is called when the physical MTU 236 * changes, which means we might have to adjust the ND level MTU. 237 */ 238void 239nd6_setmtu(struct ifnet *ifp) 240{ 241 242 nd6_setmtu0(ifp, ND_IFINFO(ifp)); 243} 244 245/* XXX todo: do not maintain copy of ifp->if_mtu in ndi->maxmtu */ 246void 247nd6_setmtu0(struct ifnet *ifp, struct nd_ifinfo *ndi) 248{ 249 INIT_VNET_INET6(ifp->if_vnet); 250 u_int32_t omaxmtu; 251 252 omaxmtu = ndi->maxmtu; 253 254 switch (ifp->if_type) { 255 case IFT_ARCNET: 256 ndi->maxmtu = MIN(ARC_PHDS_MAXMTU, ifp->if_mtu); /* RFC2497 */ 257 break; 258 case IFT_FDDI: 259 ndi->maxmtu = MIN(FDDIIPMTU, ifp->if_mtu); /* RFC2467 */ 260 break; 261 case IFT_ISO88025: 262 ndi->maxmtu = MIN(ISO88025_MAX_MTU, ifp->if_mtu); 263 break; 264 default: 265 ndi->maxmtu = ifp->if_mtu; 266 break; 267 } 268 269 /* 270 * Decreasing the interface MTU under IPV6 minimum MTU may cause 271 * undesirable situation. We thus notify the operator of the change 272 * explicitly. The check for omaxmtu is necessary to restrict the 273 * log to the case of changing the MTU, not initializing it. 274 */ 275 if (omaxmtu >= IPV6_MMTU && ndi->maxmtu < IPV6_MMTU) { 276 log(LOG_NOTICE, "nd6_setmtu0: " 277 "new link MTU on %s (%lu) is too small for IPv6\n", 278 if_name(ifp), (unsigned long)ndi->maxmtu); 279 } 280 281 if (ndi->maxmtu > V_in6_maxmtu) 282 in6_setmaxmtu(); /* check all interfaces just in case */ 283 284#undef MIN 285} 286 287void 288nd6_option_init(void *opt, int icmp6len, union nd_opts *ndopts) 289{ 290 291 bzero(ndopts, sizeof(*ndopts)); 292 ndopts->nd_opts_search = (struct nd_opt_hdr *)opt; 293 ndopts->nd_opts_last 294 = (struct nd_opt_hdr *)(((u_char *)opt) + icmp6len); 295 296 if (icmp6len == 0) { 297 ndopts->nd_opts_done = 1; 298 ndopts->nd_opts_search = NULL; 299 } 300} 301 302/* 303 * Take one ND option. 304 */ 305struct nd_opt_hdr * 306nd6_option(union nd_opts *ndopts) 307{ 308 struct nd_opt_hdr *nd_opt; 309 int olen; 310 311 if (ndopts == NULL) 312 panic("ndopts == NULL in nd6_option"); 313 if (ndopts->nd_opts_last == NULL) 314 panic("uninitialized ndopts in nd6_option"); 315 if (ndopts->nd_opts_search == NULL) 316 return NULL; 317 if (ndopts->nd_opts_done) 318 return NULL; 319 320 nd_opt = ndopts->nd_opts_search; 321 322 /* make sure nd_opt_len is inside the buffer */ 323 if ((caddr_t)&nd_opt->nd_opt_len >= (caddr_t)ndopts->nd_opts_last) { 324 bzero(ndopts, sizeof(*ndopts)); 325 return NULL; 326 } 327 328 olen = nd_opt->nd_opt_len << 3; 329 if (olen == 0) { 330 /* 331 * Message validation requires that all included 332 * options have a length that is greater than zero. 333 */ 334 bzero(ndopts, sizeof(*ndopts)); 335 return NULL; 336 } 337 338 ndopts->nd_opts_search = (struct nd_opt_hdr *)((caddr_t)nd_opt + olen); 339 if (ndopts->nd_opts_search > ndopts->nd_opts_last) { 340 /* option overruns the end of buffer, invalid */ 341 bzero(ndopts, sizeof(*ndopts)); 342 return NULL; 343 } else if (ndopts->nd_opts_search == ndopts->nd_opts_last) { 344 /* reached the end of options chain */ 345 ndopts->nd_opts_done = 1; 346 ndopts->nd_opts_search = NULL; 347 } 348 return nd_opt; 349} 350 351/* 352 * Parse multiple ND options. 353 * This function is much easier to use, for ND routines that do not need 354 * multiple options of the same type. 355 */ 356int 357nd6_options(union nd_opts *ndopts) 358{ 359 INIT_VNET_INET6(curvnet); 360 struct nd_opt_hdr *nd_opt; 361 int i = 0; 362 363 if (ndopts == NULL) 364 panic("ndopts == NULL in nd6_options"); 365 if (ndopts->nd_opts_last == NULL) 366 panic("uninitialized ndopts in nd6_options"); 367 if (ndopts->nd_opts_search == NULL) 368 return 0; 369 370 while (1) { 371 nd_opt = nd6_option(ndopts); 372 if (nd_opt == NULL && ndopts->nd_opts_last == NULL) { 373 /* 374 * Message validation requires that all included 375 * options have a length that is greater than zero. 376 */ 377 ICMP6STAT_INC(icp6s_nd_badopt); 378 bzero(ndopts, sizeof(*ndopts)); 379 return -1; 380 } 381 382 if (nd_opt == NULL) 383 goto skip1; 384 385 switch (nd_opt->nd_opt_type) { 386 case ND_OPT_SOURCE_LINKADDR: 387 case ND_OPT_TARGET_LINKADDR: 388 case ND_OPT_MTU: 389 case ND_OPT_REDIRECTED_HEADER: 390 if (ndopts->nd_opt_array[nd_opt->nd_opt_type]) { 391 nd6log((LOG_INFO, 392 "duplicated ND6 option found (type=%d)\n", 393 nd_opt->nd_opt_type)); 394 /* XXX bark? */ 395 } else { 396 ndopts->nd_opt_array[nd_opt->nd_opt_type] 397 = nd_opt; 398 } 399 break; 400 case ND_OPT_PREFIX_INFORMATION: 401 if (ndopts->nd_opt_array[nd_opt->nd_opt_type] == 0) { 402 ndopts->nd_opt_array[nd_opt->nd_opt_type] 403 = nd_opt; 404 } 405 ndopts->nd_opts_pi_end = 406 (struct nd_opt_prefix_info *)nd_opt; 407 break; 408 default: 409 /* 410 * Unknown options must be silently ignored, 411 * to accomodate future extension to the protocol. 412 */ 413 nd6log((LOG_DEBUG, 414 "nd6_options: unsupported option %d - " 415 "option ignored\n", nd_opt->nd_opt_type)); 416 } 417 418skip1: 419 i++; 420 if (i > V_nd6_maxndopt) { 421 ICMP6STAT_INC(icp6s_nd_toomanyopt); 422 nd6log((LOG_INFO, "too many loop in nd opt\n")); 423 break; 424 } 425 426 if (ndopts->nd_opts_done) 427 break; 428 } 429 430 return 0; 431} 432 433/* 434 * ND6 timer routine to handle ND6 entries 435 */ 436void 437nd6_llinfo_settimer_locked(struct llentry *ln, long tick) 438{ 439 if (tick < 0) { 440 ln->la_expire = 0; 441 ln->ln_ntick = 0; 442 callout_stop(&ln->ln_timer_ch); 443 /* 444 * XXX - do we know that there is 445 * callout installed? i.e. are we 446 * guaranteed that we're not dropping 447 * a reference that we did not add? 448 * KMM 449 */ 450 LLE_REMREF(ln); 451 } else { 452 ln->la_expire = time_second + tick / hz; 453 LLE_ADDREF(ln); 454 if (tick > INT_MAX) { 455 ln->ln_ntick = tick - INT_MAX; 456 callout_reset(&ln->ln_timer_ch, INT_MAX, 457 nd6_llinfo_timer, ln); 458 } else { 459 ln->ln_ntick = 0; 460 callout_reset(&ln->ln_timer_ch, tick, 461 nd6_llinfo_timer, ln); 462 } 463 } 464} 465 466void 467nd6_llinfo_settimer(struct llentry *ln, long tick) 468{ 469 470 LLE_WLOCK(ln); 471 nd6_llinfo_settimer_locked(ln, tick); 472 LLE_WUNLOCK(ln); 473} 474 475static void 476nd6_llinfo_timer(void *arg) 477{ 478 struct llentry *ln; 479 struct in6_addr *dst; 480 struct ifnet *ifp; 481 struct nd_ifinfo *ndi = NULL; 482 483 ln = (struct llentry *)arg; 484 if (ln == NULL) { 485 panic("%s: NULL entry!\n", __func__); 486 return; 487 } 488 489 if ((ifp = ((ln->lle_tbl != NULL) ? ln->lle_tbl->llt_ifp : NULL)) == NULL) 490 panic("ln ifp == NULL"); 491 492 CURVNET_SET(ifp->if_vnet); 493 INIT_VNET_INET6(curvnet); 494 495 if (ln->ln_ntick > 0) { 496 if (ln->ln_ntick > INT_MAX) { 497 ln->ln_ntick -= INT_MAX; 498 nd6_llinfo_settimer(ln, INT_MAX); 499 } else { 500 ln->ln_ntick = 0; 501 nd6_llinfo_settimer(ln, ln->ln_ntick); 502 } 503 goto done; 504 } 505 506 ndi = ND_IFINFO(ifp); 507 dst = &L3_ADDR_SIN6(ln)->sin6_addr; 508 if ((ln->la_flags & LLE_STATIC) || (ln->la_expire > time_second)) { 509 goto done; 510 } 511 512 if (ln->la_flags & LLE_DELETED) { 513 (void)nd6_free(ln, 0); 514 goto done; 515 } 516 517 switch (ln->ln_state) { 518 case ND6_LLINFO_INCOMPLETE: 519 if (ln->la_asked < V_nd6_mmaxtries) { 520 ln->la_asked++; 521 nd6_llinfo_settimer(ln, (long)ndi->retrans * hz / 1000); 522 nd6_ns_output(ifp, NULL, dst, ln, 0); 523 } else { 524 struct mbuf *m = ln->la_hold; 525 if (m) { 526 struct mbuf *m0; 527 528 /* 529 * assuming every packet in la_hold has the 530 * same IP header 531 */ 532 m0 = m->m_nextpkt; 533 m->m_nextpkt = NULL; 534 icmp6_error2(m, ICMP6_DST_UNREACH, 535 ICMP6_DST_UNREACH_ADDR, 0, ifp); 536 537 ln->la_hold = m0; 538 clear_llinfo_pqueue(ln); 539 } 540 (void)nd6_free(ln, 0); 541 ln = NULL; 542 } 543 break; 544 case ND6_LLINFO_REACHABLE: 545 if (!ND6_LLINFO_PERMANENT(ln)) { 546 ln->ln_state = ND6_LLINFO_STALE; 547 nd6_llinfo_settimer(ln, (long)V_nd6_gctimer * hz); 548 } 549 break; 550 551 case ND6_LLINFO_STALE: 552 /* Garbage Collection(RFC 2461 5.3) */ 553 if (!ND6_LLINFO_PERMANENT(ln)) { 554 (void)nd6_free(ln, 1); 555 ln = NULL; 556 } 557 break; 558 559 case ND6_LLINFO_DELAY: 560 if (ndi && (ndi->flags & ND6_IFF_PERFORMNUD) != 0) { 561 /* We need NUD */ 562 ln->la_asked = 1; 563 ln->ln_state = ND6_LLINFO_PROBE; 564 nd6_llinfo_settimer(ln, (long)ndi->retrans * hz / 1000); 565 nd6_ns_output(ifp, dst, dst, ln, 0); 566 } else { 567 ln->ln_state = ND6_LLINFO_STALE; /* XXX */ 568 nd6_llinfo_settimer(ln, (long)V_nd6_gctimer * hz); 569 } 570 break; 571 case ND6_LLINFO_PROBE: 572 if (ln->la_asked < V_nd6_umaxtries) { 573 ln->la_asked++; 574 nd6_llinfo_settimer(ln, (long)ndi->retrans * hz / 1000); 575 nd6_ns_output(ifp, dst, dst, ln, 0); 576 } else { 577 (void)nd6_free(ln, 0); 578 ln = NULL; 579 } 580 break; 581 } 582 CURVNET_RESTORE(); 583done: 584 if (ln != NULL) 585 LLE_FREE(ln); 586} 587 588 589/* 590 * ND6 timer routine to expire default route list and prefix list 591 */ 592void 593nd6_timer(void *arg) 594{ 595 CURVNET_SET_QUIET((struct vnet *) arg); 596 INIT_VNET_INET6((struct vnet *) arg); 597 int s; 598 struct nd_defrouter *dr; 599 struct nd_prefix *pr; 600 struct in6_ifaddr *ia6, *nia6; 601 struct in6_addrlifetime *lt6; 602 603 callout_reset(&V_nd6_timer_ch, V_nd6_prune * hz, 604 nd6_timer, NULL); 605 606 /* expire default router list */ 607 s = splnet(); 608 dr = TAILQ_FIRST(&V_nd_defrouter); 609 while (dr) { 610 if (dr->expire && dr->expire < time_second) { 611 struct nd_defrouter *t; 612 t = TAILQ_NEXT(dr, dr_entry); 613 defrtrlist_del(dr); 614 dr = t; 615 } else { 616 dr = TAILQ_NEXT(dr, dr_entry); 617 } 618 } 619 620 /* 621 * expire interface addresses. 622 * in the past the loop was inside prefix expiry processing. 623 * However, from a stricter speci-confrmance standpoint, we should 624 * rather separate address lifetimes and prefix lifetimes. 625 */ 626 addrloop: 627 for (ia6 = V_in6_ifaddr; ia6; ia6 = nia6) { 628 nia6 = ia6->ia_next; 629 /* check address lifetime */ 630 lt6 = &ia6->ia6_lifetime; 631 if (IFA6_IS_INVALID(ia6)) { 632 int regen = 0; 633 634 /* 635 * If the expiring address is temporary, try 636 * regenerating a new one. This would be useful when 637 * we suspended a laptop PC, then turned it on after a 638 * period that could invalidate all temporary 639 * addresses. Although we may have to restart the 640 * loop (see below), it must be after purging the 641 * address. Otherwise, we'd see an infinite loop of 642 * regeneration. 643 */ 644 if (V_ip6_use_tempaddr && 645 (ia6->ia6_flags & IN6_IFF_TEMPORARY) != 0) { 646 if (regen_tmpaddr(ia6) == 0) 647 regen = 1; 648 } 649 650 in6_purgeaddr(&ia6->ia_ifa); 651 652 if (regen) 653 goto addrloop; /* XXX: see below */ 654 } else if (IFA6_IS_DEPRECATED(ia6)) { 655 int oldflags = ia6->ia6_flags; 656 657 ia6->ia6_flags |= IN6_IFF_DEPRECATED; 658 659 /* 660 * If a temporary address has just become deprecated, 661 * regenerate a new one if possible. 662 */ 663 if (V_ip6_use_tempaddr && 664 (ia6->ia6_flags & IN6_IFF_TEMPORARY) != 0 && 665 (oldflags & IN6_IFF_DEPRECATED) == 0) { 666 667 if (regen_tmpaddr(ia6) == 0) { 668 /* 669 * A new temporary address is 670 * generated. 671 * XXX: this means the address chain 672 * has changed while we are still in 673 * the loop. Although the change 674 * would not cause disaster (because 675 * it's not a deletion, but an 676 * addition,) we'd rather restart the 677 * loop just for safety. Or does this 678 * significantly reduce performance?? 679 */ 680 goto addrloop; 681 } 682 } 683 } else { 684 /* 685 * A new RA might have made a deprecated address 686 * preferred. 687 */ 688 ia6->ia6_flags &= ~IN6_IFF_DEPRECATED; 689 } 690 } 691 692 /* expire prefix list */ 693 pr = V_nd_prefix.lh_first; 694 while (pr) { 695 /* 696 * check prefix lifetime. 697 * since pltime is just for autoconf, pltime processing for 698 * prefix is not necessary. 699 */ 700 if (pr->ndpr_vltime != ND6_INFINITE_LIFETIME && 701 time_second - pr->ndpr_lastupdate > pr->ndpr_vltime) { 702 struct nd_prefix *t; 703 t = pr->ndpr_next; 704 705 /* 706 * address expiration and prefix expiration are 707 * separate. NEVER perform in6_purgeaddr here. 708 */ 709 710 prelist_remove(pr); 711 pr = t; 712 } else 713 pr = pr->ndpr_next; 714 } 715 splx(s); 716 CURVNET_RESTORE(); 717} 718 719/* 720 * ia6 - deprecated/invalidated temporary address 721 */ 722static int 723regen_tmpaddr(struct in6_ifaddr *ia6) 724{ 725 struct ifaddr *ifa; 726 struct ifnet *ifp; 727 struct in6_ifaddr *public_ifa6 = NULL; 728 729 ifp = ia6->ia_ifa.ifa_ifp; 730 IF_ADDR_LOCK(ifp); 731 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_list) { 732 struct in6_ifaddr *it6; 733 734 if (ifa->ifa_addr->sa_family != AF_INET6) 735 continue; 736 737 it6 = (struct in6_ifaddr *)ifa; 738 739 /* ignore no autoconf addresses. */ 740 if ((it6->ia6_flags & IN6_IFF_AUTOCONF) == 0) 741 continue; 742 743 /* ignore autoconf addresses with different prefixes. */ 744 if (it6->ia6_ndpr == NULL || it6->ia6_ndpr != ia6->ia6_ndpr) 745 continue; 746 747 /* 748 * Now we are looking at an autoconf address with the same 749 * prefix as ours. If the address is temporary and is still 750 * preferred, do not create another one. It would be rare, but 751 * could happen, for example, when we resume a laptop PC after 752 * a long period. 753 */ 754 if ((it6->ia6_flags & IN6_IFF_TEMPORARY) != 0 && 755 !IFA6_IS_DEPRECATED(it6)) { 756 public_ifa6 = NULL; 757 break; 758 } 759 760 /* 761 * This is a public autoconf address that has the same prefix 762 * as ours. If it is preferred, keep it. We can't break the 763 * loop here, because there may be a still-preferred temporary 764 * address with the prefix. 765 */ 766 if (!IFA6_IS_DEPRECATED(it6)) 767 public_ifa6 = it6; 768 } 769 770 if (public_ifa6 != NULL) { 771 int e; 772 773 if ((e = in6_tmpifadd(public_ifa6, 0, 0)) != 0) { 774 IF_ADDR_UNLOCK(ifp); 775 log(LOG_NOTICE, "regen_tmpaddr: failed to create a new" 776 " tmp addr,errno=%d\n", e); 777 return (-1); 778 } 779 IF_ADDR_UNLOCK(ifp); 780 return (0); 781 } 782 783 IF_ADDR_UNLOCK(ifp); 784 return (-1); 785} 786 787/* 788 * Nuke neighbor cache/prefix/default router management table, right before 789 * ifp goes away. 790 */ 791void 792nd6_purge(struct ifnet *ifp) 793{ 794 INIT_VNET_INET6(ifp->if_vnet); 795 struct nd_defrouter *dr, *ndr; 796 struct nd_prefix *pr, *npr; 797 798 /* 799 * Nuke default router list entries toward ifp. 800 * We defer removal of default router list entries that is installed 801 * in the routing table, in order to keep additional side effects as 802 * small as possible. 803 */ 804 for (dr = TAILQ_FIRST(&V_nd_defrouter); dr; dr = ndr) { 805 ndr = TAILQ_NEXT(dr, dr_entry); 806 if (dr->installed) 807 continue; 808 809 if (dr->ifp == ifp) 810 defrtrlist_del(dr); 811 } 812 813 for (dr = TAILQ_FIRST(&V_nd_defrouter); dr; dr = ndr) { 814 ndr = TAILQ_NEXT(dr, dr_entry); 815 if (!dr->installed) 816 continue; 817 818 if (dr->ifp == ifp) 819 defrtrlist_del(dr); 820 } 821 822 /* Nuke prefix list entries toward ifp */ 823 for (pr = V_nd_prefix.lh_first; pr; pr = npr) { 824 npr = pr->ndpr_next; 825 if (pr->ndpr_ifp == ifp) { 826 /* 827 * Because if_detach() does *not* release prefixes 828 * while purging addresses the reference count will 829 * still be above zero. We therefore reset it to 830 * make sure that the prefix really gets purged. 831 */ 832 pr->ndpr_refcnt = 0; 833 834 /* 835 * Previously, pr->ndpr_addr is removed as well, 836 * but I strongly believe we don't have to do it. 837 * nd6_purge() is only called from in6_ifdetach(), 838 * which removes all the associated interface addresses 839 * by itself. 840 * (jinmei@kame.net 20010129) 841 */ 842 prelist_remove(pr); 843 } 844 } 845 846 /* cancel default outgoing interface setting */ 847 if (V_nd6_defifindex == ifp->if_index) 848 nd6_setdefaultiface(0); 849 850 if (!V_ip6_forwarding && V_ip6_accept_rtadv) { /* XXX: too restrictive? */ 851 /* refresh default router list 852 * 853 * 854 */ 855 defrouter_select(); 856 857 } 858 859 /* XXXXX 860 * We do not nuke the neighbor cache entries here any more 861 * because the neighbor cache is kept in if_afdata[AF_INET6]. 862 * nd6_purge() is invoked by in6_ifdetach() which is called 863 * from if_detach() where everything gets purged. So let 864 * in6_domifdetach() do the actual L2 table purging work. 865 */ 866} 867 868/* 869 * the caller acquires and releases the lock on the lltbls 870 * Returns the llentry locked 871 */ 872struct llentry * 873nd6_lookup(struct in6_addr *addr6, int flags, struct ifnet *ifp) 874{ 875 INIT_VNET_INET6(curvnet); 876 struct sockaddr_in6 sin6; 877 struct llentry *ln; 878 int llflags = 0; 879 880 bzero(&sin6, sizeof(sin6)); 881 sin6.sin6_len = sizeof(struct sockaddr_in6); 882 sin6.sin6_family = AF_INET6; 883 sin6.sin6_addr = *addr6; 884 885 IF_AFDATA_LOCK_ASSERT(ifp); 886 887 if (flags & ND6_CREATE) 888 llflags |= LLE_CREATE; 889 if (flags & ND6_EXCLUSIVE) 890 llflags |= LLE_EXCLUSIVE; 891 892 ln = lla_lookup(LLTABLE6(ifp), llflags, (struct sockaddr *)&sin6); 893 if ((ln != NULL) && (flags & LLE_CREATE)) { 894 ln->ln_state = ND6_LLINFO_NOSTATE; 895 callout_init(&ln->ln_timer_ch, 0); 896 } 897 898 return (ln); 899} 900 901/* 902 * Test whether a given IPv6 address is a neighbor or not, ignoring 903 * the actual neighbor cache. The neighbor cache is ignored in order 904 * to not reenter the routing code from within itself. 905 */ 906static int 907nd6_is_new_addr_neighbor(struct sockaddr_in6 *addr, struct ifnet *ifp) 908{ 909 INIT_VNET_INET6(ifp->if_vnet); 910 struct nd_prefix *pr; 911 struct ifaddr *dstaddr; 912 913 /* 914 * A link-local address is always a neighbor. 915 * XXX: a link does not necessarily specify a single interface. 916 */ 917 if (IN6_IS_ADDR_LINKLOCAL(&addr->sin6_addr)) { 918 struct sockaddr_in6 sin6_copy; 919 u_int32_t zone; 920 921 /* 922 * We need sin6_copy since sa6_recoverscope() may modify the 923 * content (XXX). 924 */ 925 sin6_copy = *addr; 926 if (sa6_recoverscope(&sin6_copy)) 927 return (0); /* XXX: should be impossible */ 928 if (in6_setscope(&sin6_copy.sin6_addr, ifp, &zone)) 929 return (0); 930 if (sin6_copy.sin6_scope_id == zone) 931 return (1); 932 else 933 return (0); 934 } 935 936 /* 937 * If the address matches one of our addresses, 938 * it should be a neighbor. 939 * If the address matches one of our on-link prefixes, it should be a 940 * neighbor. 941 */ 942 for (pr = V_nd_prefix.lh_first; pr; pr = pr->ndpr_next) { 943 if (pr->ndpr_ifp != ifp) 944 continue; 945 946 if (!(pr->ndpr_stateflags & NDPRF_ONLINK)) 947 continue; 948 949 if (IN6_ARE_MASKED_ADDR_EQUAL(&pr->ndpr_prefix.sin6_addr, 950 &addr->sin6_addr, &pr->ndpr_mask)) 951 return (1); 952 } 953 954 /* 955 * If the address is assigned on the node of the other side of 956 * a p2p interface, the address should be a neighbor. 957 */ 958 dstaddr = ifa_ifwithdstaddr((struct sockaddr *)addr); 959 if ((dstaddr != NULL) && (dstaddr->ifa_ifp == ifp)) 960 return (1); 961 962 /* 963 * If the default router list is empty, all addresses are regarded 964 * as on-link, and thus, as a neighbor. 965 * XXX: we restrict the condition to hosts, because routers usually do 966 * not have the "default router list". 967 */ 968 if (!V_ip6_forwarding && TAILQ_FIRST(&V_nd_defrouter) == NULL && 969 V_nd6_defifindex == ifp->if_index) { 970 return (1); 971 } 972 973 return (0); 974} 975 976 977/* 978 * Detect if a given IPv6 address identifies a neighbor on a given link. 979 * XXX: should take care of the destination of a p2p link? 980 */ 981int 982nd6_is_addr_neighbor(struct sockaddr_in6 *addr, struct ifnet *ifp) 983{ 984 struct llentry *lle; 985 int rc = 0; 986 987 IF_AFDATA_UNLOCK_ASSERT(ifp); 988 if (nd6_is_new_addr_neighbor(addr, ifp)) 989 return (1); 990 991 /* 992 * Even if the address matches none of our addresses, it might be 993 * in the neighbor cache. 994 */ 995 IF_AFDATA_LOCK(ifp); 996 if ((lle = nd6_lookup(&addr->sin6_addr, 0, ifp)) != NULL) { 997 LLE_RUNLOCK(lle); 998 rc = 1; 999 } 1000 IF_AFDATA_UNLOCK(ifp); 1001 return (rc); 1002} 1003 1004/* 1005 * Free an nd6 llinfo entry. 1006 * Since the function would cause significant changes in the kernel, DO NOT 1007 * make it global, unless you have a strong reason for the change, and are sure 1008 * that the change is safe. 1009 */ 1010static struct llentry * 1011nd6_free(struct llentry *ln, int gc) 1012{ 1013 INIT_VNET_INET6(curvnet); 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 INIT_VNET_INET6(curvnet); 1121 struct llentry *ln; 1122 struct ifnet *ifp; 1123 1124 if ((dst6 == NULL) || (rt == NULL)) 1125 return; 1126 1127 ifp = rt->rt_ifp; 1128 IF_AFDATA_LOCK(ifp); 1129 ln = nd6_lookup(dst6, ND6_EXCLUSIVE, NULL); 1130 IF_AFDATA_UNLOCK(ifp); 1131 if (ln == NULL) 1132 return; 1133 1134 if (ln->ln_state < ND6_LLINFO_REACHABLE) 1135 goto done; 1136 1137 /* 1138 * if we get upper-layer reachability confirmation many times, 1139 * it is possible we have false information. 1140 */ 1141 if (!force) { 1142 ln->ln_byhint++; 1143 if (ln->ln_byhint > V_nd6_maxnudhint) { 1144 goto done; 1145 } 1146 } 1147 1148 ln->ln_state = ND6_LLINFO_REACHABLE; 1149 if (!ND6_LLINFO_PERMANENT(ln)) { 1150 nd6_llinfo_settimer(ln, 1151 (long)ND_IFINFO(rt->rt_ifp)->reachable * hz); 1152 } 1153done: 1154 LLE_WUNLOCK(ln); 1155} 1156 1157 1158int 1159nd6_ioctl(u_long cmd, caddr_t data, struct ifnet *ifp) 1160{ 1161 INIT_VNET_INET6(ifp->if_vnet); 1162 struct in6_drlist *drl = (struct in6_drlist *)data; 1163 struct in6_oprlist *oprl = (struct in6_oprlist *)data; 1164 struct in6_ndireq *ndi = (struct in6_ndireq *)data; 1165 struct in6_nbrinfo *nbi = (struct in6_nbrinfo *)data; 1166 struct in6_ndifreq *ndif = (struct in6_ndifreq *)data; 1167 struct nd_defrouter *dr; 1168 struct nd_prefix *pr; 1169 int i = 0, error = 0; 1170 int s; 1171 1172 switch (cmd) { 1173 case SIOCGDRLST_IN6: 1174 /* 1175 * obsolete API, use sysctl under net.inet6.icmp6 1176 */ 1177 bzero(drl, sizeof(*drl)); 1178 s = splnet(); 1179 dr = TAILQ_FIRST(&V_nd_defrouter); 1180 while (dr && i < DRLSTSIZ) { 1181 drl->defrouter[i].rtaddr = dr->rtaddr; 1182 in6_clearscope(&drl->defrouter[i].rtaddr); 1183 1184 drl->defrouter[i].flags = dr->flags; 1185 drl->defrouter[i].rtlifetime = dr->rtlifetime; 1186 drl->defrouter[i].expire = dr->expire; 1187 drl->defrouter[i].if_index = dr->ifp->if_index; 1188 i++; 1189 dr = TAILQ_NEXT(dr, dr_entry); 1190 } 1191 splx(s); 1192 break; 1193 case SIOCGPRLST_IN6: 1194 /* 1195 * obsolete API, use sysctl under net.inet6.icmp6 1196 * 1197 * XXX the structure in6_prlist was changed in backward- 1198 * incompatible manner. in6_oprlist is used for SIOCGPRLST_IN6, 1199 * in6_prlist is used for nd6_sysctl() - fill_prlist(). 1200 */ 1201 /* 1202 * XXX meaning of fields, especialy "raflags", is very 1203 * differnet between RA prefix list and RR/static prefix list. 1204 * how about separating ioctls into two? 1205 */ 1206 bzero(oprl, sizeof(*oprl)); 1207 s = splnet(); 1208 pr = V_nd_prefix.lh_first; 1209 while (pr && i < PRLSTSIZ) { 1210 struct nd_pfxrouter *pfr; 1211 int j; 1212 1213 oprl->prefix[i].prefix = pr->ndpr_prefix.sin6_addr; 1214 oprl->prefix[i].raflags = pr->ndpr_raf; 1215 oprl->prefix[i].prefixlen = pr->ndpr_plen; 1216 oprl->prefix[i].vltime = pr->ndpr_vltime; 1217 oprl->prefix[i].pltime = pr->ndpr_pltime; 1218 oprl->prefix[i].if_index = pr->ndpr_ifp->if_index; 1219 if (pr->ndpr_vltime == ND6_INFINITE_LIFETIME) 1220 oprl->prefix[i].expire = 0; 1221 else { 1222 time_t maxexpire; 1223 1224 /* XXX: we assume time_t is signed. */ 1225 maxexpire = (-1) & 1226 ~((time_t)1 << 1227 ((sizeof(maxexpire) * 8) - 1)); 1228 if (pr->ndpr_vltime < 1229 maxexpire - pr->ndpr_lastupdate) { 1230 oprl->prefix[i].expire = 1231 pr->ndpr_lastupdate + 1232 pr->ndpr_vltime; 1233 } else 1234 oprl->prefix[i].expire = maxexpire; 1235 } 1236 1237 pfr = pr->ndpr_advrtrs.lh_first; 1238 j = 0; 1239 while (pfr) { 1240 if (j < DRLSTSIZ) { 1241#define RTRADDR oprl->prefix[i].advrtr[j] 1242 RTRADDR = pfr->router->rtaddr; 1243 in6_clearscope(&RTRADDR); 1244#undef RTRADDR 1245 } 1246 j++; 1247 pfr = pfr->pfr_next; 1248 } 1249 oprl->prefix[i].advrtrs = j; 1250 oprl->prefix[i].origin = PR_ORIG_RA; 1251 1252 i++; 1253 pr = pr->ndpr_next; 1254 } 1255 splx(s); 1256 1257 break; 1258 case OSIOCGIFINFO_IN6: 1259#define ND ndi->ndi 1260 /* XXX: old ndp(8) assumes a positive value for linkmtu. */ 1261 bzero(&ND, sizeof(ND)); 1262 ND.linkmtu = IN6_LINKMTU(ifp); 1263 ND.maxmtu = ND_IFINFO(ifp)->maxmtu; 1264 ND.basereachable = ND_IFINFO(ifp)->basereachable; 1265 ND.reachable = ND_IFINFO(ifp)->reachable; 1266 ND.retrans = ND_IFINFO(ifp)->retrans; 1267 ND.flags = ND_IFINFO(ifp)->flags; 1268 ND.recalctm = ND_IFINFO(ifp)->recalctm; 1269 ND.chlim = ND_IFINFO(ifp)->chlim; 1270 break; 1271 case SIOCGIFINFO_IN6: 1272 ND = *ND_IFINFO(ifp); 1273 break; 1274 case SIOCSIFINFO_IN6: 1275 /* 1276 * used to change host variables from userland. 1277 * intented for a use on router to reflect RA configurations. 1278 */ 1279 /* 0 means 'unspecified' */ 1280 if (ND.linkmtu != 0) { 1281 if (ND.linkmtu < IPV6_MMTU || 1282 ND.linkmtu > IN6_LINKMTU(ifp)) { 1283 error = EINVAL; 1284 break; 1285 } 1286 ND_IFINFO(ifp)->linkmtu = ND.linkmtu; 1287 } 1288 1289 if (ND.basereachable != 0) { 1290 int obasereachable = ND_IFINFO(ifp)->basereachable; 1291 1292 ND_IFINFO(ifp)->basereachable = ND.basereachable; 1293 if (ND.basereachable != obasereachable) 1294 ND_IFINFO(ifp)->reachable = 1295 ND_COMPUTE_RTIME(ND.basereachable); 1296 } 1297 if (ND.retrans != 0) 1298 ND_IFINFO(ifp)->retrans = ND.retrans; 1299 if (ND.chlim != 0) 1300 ND_IFINFO(ifp)->chlim = ND.chlim; 1301 /* FALLTHROUGH */ 1302 case SIOCSIFINFO_FLAGS: 1303 ND_IFINFO(ifp)->flags = ND.flags; 1304 break; 1305#undef ND 1306 case SIOCSNDFLUSH_IN6: /* XXX: the ioctl name is confusing... */ 1307 /* sync kernel routing table with the default router list */ 1308 defrouter_reset(); 1309 defrouter_select(); 1310 break; 1311 case SIOCSPFXFLUSH_IN6: 1312 { 1313 /* flush all the prefix advertised by routers */ 1314 struct nd_prefix *pr, *next; 1315 1316 s = splnet(); 1317 for (pr = V_nd_prefix.lh_first; pr; pr = next) { 1318 struct in6_ifaddr *ia, *ia_next; 1319 1320 next = pr->ndpr_next; 1321 1322 if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr)) 1323 continue; /* XXX */ 1324 1325 /* do we really have to remove addresses as well? */ 1326 for (ia = V_in6_ifaddr; ia; ia = ia_next) { 1327 /* ia might be removed. keep the next ptr. */ 1328 ia_next = ia->ia_next; 1329 1330 if ((ia->ia6_flags & IN6_IFF_AUTOCONF) == 0) 1331 continue; 1332 1333 if (ia->ia6_ndpr == pr) 1334 in6_purgeaddr(&ia->ia_ifa); 1335 } 1336 prelist_remove(pr); 1337 } 1338 splx(s); 1339 break; 1340 } 1341 case SIOCSRTRFLUSH_IN6: 1342 { 1343 /* flush all the default routers */ 1344 struct nd_defrouter *dr, *next; 1345 1346 s = splnet(); 1347 defrouter_reset(); 1348 for (dr = TAILQ_FIRST(&V_nd_defrouter); dr; dr = next) { 1349 next = TAILQ_NEXT(dr, dr_entry); 1350 defrtrlist_del(dr); 1351 } 1352 defrouter_select(); 1353 splx(s); 1354 break; 1355 } 1356 case SIOCGNBRINFO_IN6: 1357 { 1358 struct llentry *ln; 1359 struct in6_addr nb_addr = nbi->addr; /* make local for safety */ 1360 1361 if ((error = in6_setscope(&nb_addr, ifp, NULL)) != 0) 1362 return (error); 1363 1364 IF_AFDATA_LOCK(ifp); 1365 ln = nd6_lookup(&nb_addr, 0, ifp); 1366 IF_AFDATA_UNLOCK(ifp); 1367 1368 if (ln == NULL) { 1369 error = EINVAL; 1370 break; 1371 } 1372 nbi->state = ln->ln_state; 1373 nbi->asked = ln->la_asked; 1374 nbi->isrouter = ln->ln_router; 1375 nbi->expire = ln->la_expire; 1376 LLE_RUNLOCK(ln); 1377 break; 1378 } 1379 case SIOCGDEFIFACE_IN6: /* XXX: should be implemented as a sysctl? */ 1380 ndif->ifindex = V_nd6_defifindex; 1381 break; 1382 case SIOCSDEFIFACE_IN6: /* XXX: should be implemented as a sysctl? */ 1383 return (nd6_setdefaultiface(ndif->ifindex)); 1384 } 1385 return (error); 1386} 1387 1388/* 1389 * Create neighbor cache entry and cache link-layer address, 1390 * on reception of inbound ND6 packets. (RS/RA/NS/redirect) 1391 * 1392 * type - ICMP6 type 1393 * code - type dependent information 1394 * 1395 * XXXXX 1396 * The caller of this function already acquired the ndp 1397 * cache table lock because the cache entry is returned. 1398 */ 1399struct llentry * 1400nd6_cache_lladdr(struct ifnet *ifp, struct in6_addr *from, char *lladdr, 1401 int lladdrlen, int type, int code) 1402{ 1403 INIT_VNET_INET6(curvnet); 1404 struct llentry *ln = NULL; 1405 int is_newentry; 1406 int do_update; 1407 int olladdr; 1408 int llchange; 1409 int flags = 0; 1410 int newstate = 0; 1411 uint16_t router = 0; 1412 struct sockaddr_in6 sin6; 1413 struct mbuf *chain = NULL; 1414 int static_route = 0; 1415 1416 IF_AFDATA_UNLOCK_ASSERT(ifp); 1417 1418 if (ifp == NULL) 1419 panic("ifp == NULL in nd6_cache_lladdr"); 1420 if (from == NULL) 1421 panic("from == NULL in nd6_cache_lladdr"); 1422 1423 /* nothing must be updated for unspecified address */ 1424 if (IN6_IS_ADDR_UNSPECIFIED(from)) 1425 return NULL; 1426 1427 /* 1428 * Validation about ifp->if_addrlen and lladdrlen must be done in 1429 * the caller. 1430 * 1431 * XXX If the link does not have link-layer adderss, what should 1432 * we do? (ifp->if_addrlen == 0) 1433 * Spec says nothing in sections for RA, RS and NA. There's small 1434 * description on it in NS section (RFC 2461 7.2.3). 1435 */ 1436 flags |= lladdr ? ND6_EXCLUSIVE : 0; 1437 IF_AFDATA_LOCK(ifp); 1438 ln = nd6_lookup(from, flags, ifp); 1439 1440 if (ln == NULL) { 1441 flags |= LLE_EXCLUSIVE; 1442 ln = nd6_lookup(from, flags |ND6_CREATE, ifp); 1443 IF_AFDATA_UNLOCK(ifp); 1444 is_newentry = 1; 1445 } else { 1446 IF_AFDATA_UNLOCK(ifp); 1447 /* do nothing if static ndp is set */ 1448 if (ln->la_flags & LLE_STATIC) { 1449 static_route = 1; 1450 goto done; 1451 } 1452 is_newentry = 0; 1453 } 1454 if (ln == NULL) 1455 return (NULL); 1456 1457 olladdr = (ln->la_flags & LLE_VALID) ? 1 : 0; 1458 if (olladdr && lladdr) { 1459 llchange = bcmp(lladdr, &ln->ll_addr, 1460 ifp->if_addrlen); 1461 } else 1462 llchange = 0; 1463 1464 /* 1465 * newentry olladdr lladdr llchange (*=record) 1466 * 0 n n -- (1) 1467 * 0 y n -- (2) 1468 * 0 n y -- (3) * STALE 1469 * 0 y y n (4) * 1470 * 0 y y y (5) * STALE 1471 * 1 -- n -- (6) NOSTATE(= PASSIVE) 1472 * 1 -- y -- (7) * STALE 1473 */ 1474 1475 if (lladdr) { /* (3-5) and (7) */ 1476 /* 1477 * Record source link-layer address 1478 * XXX is it dependent to ifp->if_type? 1479 */ 1480 bcopy(lladdr, &ln->ll_addr, ifp->if_addrlen); 1481 ln->la_flags |= LLE_VALID; 1482 } 1483 1484 if (!is_newentry) { 1485 if ((!olladdr && lladdr != NULL) || /* (3) */ 1486 (olladdr && lladdr != NULL && llchange)) { /* (5) */ 1487 do_update = 1; 1488 newstate = ND6_LLINFO_STALE; 1489 } else /* (1-2,4) */ 1490 do_update = 0; 1491 } else { 1492 do_update = 1; 1493 if (lladdr == NULL) /* (6) */ 1494 newstate = ND6_LLINFO_NOSTATE; 1495 else /* (7) */ 1496 newstate = ND6_LLINFO_STALE; 1497 } 1498 1499 if (do_update) { 1500 /* 1501 * Update the state of the neighbor cache. 1502 */ 1503 ln->ln_state = newstate; 1504 1505 if (ln->ln_state == ND6_LLINFO_STALE) { 1506 /* 1507 * XXX: since nd6_output() below will cause 1508 * state tansition to DELAY and reset the timer, 1509 * we must set the timer now, although it is actually 1510 * meaningless. 1511 */ 1512 nd6_llinfo_settimer_locked(ln, (long)V_nd6_gctimer * hz); 1513 1514 if (ln->la_hold) { 1515 struct mbuf *m_hold, *m_hold_next; 1516 1517 /* 1518 * reset the la_hold in advance, to explicitly 1519 * prevent a la_hold lookup in nd6_output() 1520 * (wouldn't happen, though...) 1521 */ 1522 for (m_hold = ln->la_hold, ln->la_hold = NULL; 1523 m_hold; m_hold = m_hold_next) { 1524 m_hold_next = m_hold->m_nextpkt; 1525 m_hold->m_nextpkt = NULL; 1526 1527 /* 1528 * we assume ifp is not a p2p here, so 1529 * just set the 2nd argument as the 1530 * 1st one. 1531 */ 1532 nd6_output_lle(ifp, ifp, m_hold, L3_ADDR_SIN6(ln), NULL, ln, &chain); 1533 } 1534 /* 1535 * If we have mbufs in the chain we need to do 1536 * deferred transmit. Copy the address from the 1537 * llentry before dropping the lock down below. 1538 */ 1539 if (chain != NULL) 1540 memcpy(&sin6, L3_ADDR_SIN6(ln), sizeof(sin6)); 1541 } 1542 } else if (ln->ln_state == ND6_LLINFO_INCOMPLETE) { 1543 /* probe right away */ 1544 nd6_llinfo_settimer_locked((void *)ln, 0); 1545 } 1546 } 1547 1548 /* 1549 * ICMP6 type dependent behavior. 1550 * 1551 * NS: clear IsRouter if new entry 1552 * RS: clear IsRouter 1553 * RA: set IsRouter if there's lladdr 1554 * redir: clear IsRouter if new entry 1555 * 1556 * RA case, (1): 1557 * The spec says that we must set IsRouter in the following cases: 1558 * - If lladdr exist, set IsRouter. This means (1-5). 1559 * - If it is old entry (!newentry), set IsRouter. This means (7). 1560 * So, based on the spec, in (1-5) and (7) cases we must set IsRouter. 1561 * A quetion arises for (1) case. (1) case has no lladdr in the 1562 * neighbor cache, this is similar to (6). 1563 * This case is rare but we figured that we MUST NOT set IsRouter. 1564 * 1565 * newentry olladdr lladdr llchange NS RS RA redir 1566 * D R 1567 * 0 n n -- (1) c ? s 1568 * 0 y n -- (2) c s s 1569 * 0 n y -- (3) c s s 1570 * 0 y y n (4) c s s 1571 * 0 y y y (5) c s s 1572 * 1 -- n -- (6) c c c s 1573 * 1 -- y -- (7) c c s c s 1574 * 1575 * (c=clear s=set) 1576 */ 1577 switch (type & 0xff) { 1578 case ND_NEIGHBOR_SOLICIT: 1579 /* 1580 * New entry must have is_router flag cleared. 1581 */ 1582 if (is_newentry) /* (6-7) */ 1583 ln->ln_router = 0; 1584 break; 1585 case ND_REDIRECT: 1586 /* 1587 * If the icmp is a redirect to a better router, always set the 1588 * is_router flag. Otherwise, if the entry is newly created, 1589 * clear the flag. [RFC 2461, sec 8.3] 1590 */ 1591 if (code == ND_REDIRECT_ROUTER) 1592 ln->ln_router = 1; 1593 else if (is_newentry) /* (6-7) */ 1594 ln->ln_router = 0; 1595 break; 1596 case ND_ROUTER_SOLICIT: 1597 /* 1598 * is_router flag must always be cleared. 1599 */ 1600 ln->ln_router = 0; 1601 break; 1602 case ND_ROUTER_ADVERT: 1603 /* 1604 * Mark an entry with lladdr as a router. 1605 */ 1606 if ((!is_newentry && (olladdr || lladdr)) || /* (2-5) */ 1607 (is_newentry && lladdr)) { /* (7) */ 1608 ln->ln_router = 1; 1609 } 1610 break; 1611 } 1612 1613 if (ln != NULL) { 1614 static_route = (ln->la_flags & LLE_STATIC); 1615 router = ln->ln_router; 1616 1617 if (flags & ND6_EXCLUSIVE) 1618 LLE_WUNLOCK(ln); 1619 else 1620 LLE_RUNLOCK(ln); 1621 if (static_route) 1622 ln = NULL; 1623 } 1624 if (chain) 1625 nd6_output_flush(ifp, ifp, chain, &sin6, NULL); 1626 1627 /* 1628 * When the link-layer address of a router changes, select the 1629 * best router again. In particular, when the neighbor entry is newly 1630 * created, it might affect the selection policy. 1631 * Question: can we restrict the first condition to the "is_newentry" 1632 * case? 1633 * XXX: when we hear an RA from a new router with the link-layer 1634 * address option, defrouter_select() is called twice, since 1635 * defrtrlist_update called the function as well. However, I believe 1636 * we can compromise the overhead, since it only happens the first 1637 * time. 1638 * XXX: although defrouter_select() should not have a bad effect 1639 * for those are not autoconfigured hosts, we explicitly avoid such 1640 * cases for safety. 1641 */ 1642 if (do_update && router && !V_ip6_forwarding && V_ip6_accept_rtadv) { 1643 /* 1644 * guaranteed recursion 1645 */ 1646 defrouter_select(); 1647 } 1648 1649 return (ln); 1650done: 1651 if (ln != NULL) { 1652 if (flags & ND6_EXCLUSIVE) 1653 LLE_WUNLOCK(ln); 1654 else 1655 LLE_RUNLOCK(ln); 1656 if (static_route) 1657 ln = NULL; 1658 } 1659 return (ln); 1660} 1661 1662static void 1663nd6_slowtimo(void *arg) 1664{ 1665 CURVNET_SET((struct vnet *) arg); 1666 INIT_VNET_NET((struct vnet *) arg); 1667 INIT_VNET_INET6((struct vnet *) arg); 1668 struct nd_ifinfo *nd6if; 1669 struct ifnet *ifp; 1670 1671 callout_reset(&V_nd6_slowtimo_ch, ND6_SLOWTIMER_INTERVAL * hz, 1672 nd6_slowtimo, NULL); 1673 IFNET_RLOCK(); 1674 for (ifp = TAILQ_FIRST(&V_ifnet); ifp; 1675 ifp = TAILQ_NEXT(ifp, if_list)) { 1676 nd6if = ND_IFINFO(ifp); 1677 if (nd6if->basereachable && /* already initialized */ 1678 (nd6if->recalctm -= ND6_SLOWTIMER_INTERVAL) <= 0) { 1679 /* 1680 * Since reachable time rarely changes by router 1681 * advertisements, we SHOULD insure that a new random 1682 * value gets recomputed at least once every few hours. 1683 * (RFC 2461, 6.3.4) 1684 */ 1685 nd6if->recalctm = V_nd6_recalc_reachtm_interval; 1686 nd6if->reachable = ND_COMPUTE_RTIME(nd6if->basereachable); 1687 } 1688 } 1689 IFNET_RUNLOCK(); 1690 CURVNET_RESTORE(); 1691} 1692 1693int 1694nd6_output(struct ifnet *ifp, struct ifnet *origifp, struct mbuf *m0, 1695 struct sockaddr_in6 *dst, struct rtentry *rt0) 1696{ 1697 1698 return (nd6_output_lle(ifp, origifp, m0, dst, rt0, NULL, NULL)); 1699} 1700 1701 1702/* 1703 * Note that I'm not enforcing any global serialization 1704 * lle state or asked changes here as the logic is too 1705 * complicated to avoid having to always acquire an exclusive 1706 * lock 1707 * KMM 1708 * 1709 */ 1710#define senderr(e) { error = (e); goto bad;} 1711 1712int 1713nd6_output_lle(struct ifnet *ifp, struct ifnet *origifp, struct mbuf *m0, 1714 struct sockaddr_in6 *dst, struct rtentry *rt0, struct llentry *lle, 1715 struct mbuf **chain) 1716{ 1717 INIT_VNET_INET6(curvnet); 1718 struct mbuf *m = m0; 1719 struct llentry *ln = lle; 1720 int error = 0; 1721 int flags = 0; 1722 1723#ifdef INVARIANTS 1724 if (lle != NULL) { 1725 1726 LLE_WLOCK_ASSERT(lle); 1727 1728 KASSERT(chain != NULL, (" lle locked but no mbuf chain pointer passed")); 1729 } 1730#endif 1731 if (IN6_IS_ADDR_MULTICAST(&dst->sin6_addr)) 1732 goto sendpkt; 1733 1734 if (nd6_need_cache(ifp) == 0) 1735 goto sendpkt; 1736 1737 /* 1738 * next hop determination. This routine is derived from ether_output. 1739 */ 1740 1741 /* 1742 * Address resolution or Neighbor Unreachability Detection 1743 * for the next hop. 1744 * At this point, the destination of the packet must be a unicast 1745 * or an anycast address(i.e. not a multicast). 1746 */ 1747 1748 flags = ((m != NULL) || (lle != NULL)) ? LLE_EXCLUSIVE : 0; 1749 if (ln == NULL) { 1750 retry: 1751 IF_AFDATA_LOCK(ifp); 1752 ln = lla_lookup(LLTABLE6(ifp), flags, (struct sockaddr *)dst); 1753 IF_AFDATA_UNLOCK(ifp); 1754 if ((ln == NULL) && nd6_is_addr_neighbor(dst, ifp)) { 1755 /* 1756 * Since nd6_is_addr_neighbor() internally calls nd6_lookup(), 1757 * the condition below is not very efficient. But we believe 1758 * it is tolerable, because this should be a rare case. 1759 */ 1760 flags = ND6_CREATE | (m ? ND6_EXCLUSIVE : 0); 1761 IF_AFDATA_LOCK(ifp); 1762 ln = nd6_lookup(&dst->sin6_addr, flags, ifp); 1763 IF_AFDATA_UNLOCK(ifp); 1764 } 1765 } 1766 if (ln == NULL) { 1767 if ((ifp->if_flags & IFF_POINTOPOINT) == 0 && 1768 !(ND_IFINFO(ifp)->flags & ND6_IFF_PERFORMNUD)) { 1769 char ip6buf[INET6_ADDRSTRLEN]; 1770 log(LOG_DEBUG, 1771 "nd6_output: can't allocate llinfo for %s " 1772 "(ln=%p)\n", 1773 ip6_sprintf(ip6buf, &dst->sin6_addr), ln); 1774 senderr(EIO); /* XXX: good error? */ 1775 } 1776 goto sendpkt; /* send anyway */ 1777 } 1778 1779 /* We don't have to do link-layer address resolution on a p2p link. */ 1780 if ((ifp->if_flags & IFF_POINTOPOINT) != 0 && 1781 ln->ln_state < ND6_LLINFO_REACHABLE) { 1782 if ((flags & LLE_EXCLUSIVE) == 0) { 1783 flags |= LLE_EXCLUSIVE; 1784 goto retry; 1785 } 1786 ln->ln_state = ND6_LLINFO_STALE; 1787 nd6_llinfo_settimer_locked(ln, (long)V_nd6_gctimer * hz); 1788 } 1789 1790 /* 1791 * The first time we send a packet to a neighbor whose entry is 1792 * STALE, we have to change the state to DELAY and a sets a timer to 1793 * expire in DELAY_FIRST_PROBE_TIME seconds to ensure do 1794 * neighbor unreachability detection on expiration. 1795 * (RFC 2461 7.3.3) 1796 */ 1797 if (ln->ln_state == ND6_LLINFO_STALE) { 1798 if ((flags & LLE_EXCLUSIVE) == 0) { 1799 flags |= LLE_EXCLUSIVE; 1800 LLE_RUNLOCK(ln); 1801 goto retry; 1802 } 1803 ln->la_asked = 0; 1804 ln->ln_state = ND6_LLINFO_DELAY; 1805 nd6_llinfo_settimer_locked(ln, (long)V_nd6_delay * hz); 1806 } 1807 1808 /* 1809 * If the neighbor cache entry has a state other than INCOMPLETE 1810 * (i.e. its link-layer address is already resolved), just 1811 * send the packet. 1812 */ 1813 if (ln->ln_state > ND6_LLINFO_INCOMPLETE) 1814 goto sendpkt; 1815 1816 /* 1817 * There is a neighbor cache entry, but no ethernet address 1818 * response yet. Append this latest packet to the end of the 1819 * packet queue in the mbuf, unless the number of the packet 1820 * does not exceed nd6_maxqueuelen. When it exceeds nd6_maxqueuelen, 1821 * the oldest packet in the queue will be removed. 1822 */ 1823 if (ln->ln_state == ND6_LLINFO_NOSTATE) 1824 ln->ln_state = ND6_LLINFO_INCOMPLETE; 1825 1826 if ((flags & LLE_EXCLUSIVE) == 0) { 1827 flags |= LLE_EXCLUSIVE; 1828 LLE_RUNLOCK(ln); 1829 goto retry; 1830 } 1831 if (ln->la_hold) { 1832 struct mbuf *m_hold; 1833 int i; 1834 1835 i = 0; 1836 for (m_hold = ln->la_hold; m_hold; m_hold = m_hold->m_nextpkt) { 1837 i++; 1838 if (m_hold->m_nextpkt == NULL) { 1839 m_hold->m_nextpkt = m; 1840 break; 1841 } 1842 } 1843 while (i >= V_nd6_maxqueuelen) { 1844 m_hold = ln->la_hold; 1845 ln->la_hold = ln->la_hold->m_nextpkt; 1846 m_freem(m_hold); 1847 i--; 1848 } 1849 } else { 1850 ln->la_hold = m; 1851 } 1852 /* 1853 * We did the lookup (no lle arg) so we 1854 * need to do the unlock here 1855 */ 1856 if (lle == NULL) { 1857 if (flags & LLE_EXCLUSIVE) 1858 LLE_WUNLOCK(ln); 1859 else 1860 LLE_RUNLOCK(ln); 1861 } 1862 1863 /* 1864 * If there has been no NS for the neighbor after entering the 1865 * INCOMPLETE state, send the first solicitation. 1866 */ 1867 if (!ND6_LLINFO_PERMANENT(ln) && ln->la_asked == 0) { 1868 ln->la_asked++; 1869 1870 nd6_llinfo_settimer(ln, 1871 (long)ND_IFINFO(ifp)->retrans * hz / 1000); 1872 nd6_ns_output(ifp, NULL, &dst->sin6_addr, ln, 0); 1873 } 1874 return (0); 1875 1876 sendpkt: 1877 /* discard the packet if IPv6 operation is disabled on the interface */ 1878 if ((ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED)) { 1879 error = ENETDOWN; /* better error? */ 1880 goto bad; 1881 } 1882 /* 1883 * ln is valid and the caller did not pass in 1884 * an llentry 1885 */ 1886 if ((ln != NULL) && (lle == NULL)) { 1887 if (flags & LLE_EXCLUSIVE) 1888 LLE_WUNLOCK(ln); 1889 else 1890 LLE_RUNLOCK(ln); 1891 } 1892 1893#ifdef MAC 1894 mac_netinet6_nd6_send(ifp, m); 1895#endif 1896 /* 1897 * We were passed in a pointer to an lle with the lock held 1898 * this means that we can't call if_output as we will 1899 * recurse on the lle lock - so what we do is we create 1900 * a list of mbufs to send and transmit them in the caller 1901 * after the lock is dropped 1902 */ 1903 if (lle != NULL) { 1904 if (*chain == NULL) 1905 *chain = m; 1906 else { 1907 struct mbuf *m = *chain; 1908 1909 /* 1910 * append mbuf to end of deferred chain 1911 */ 1912 while (m->m_nextpkt != NULL) 1913 m = m->m_nextpkt; 1914 m->m_nextpkt = m; 1915 } 1916 return (error); 1917 } 1918 if ((ifp->if_flags & IFF_LOOPBACK) != 0) { 1919 return ((*ifp->if_output)(origifp, m, (struct sockaddr *)dst, 1920 NULL)); 1921 } 1922 error = (*ifp->if_output)(ifp, m, (struct sockaddr *)dst, NULL); 1923 return (error); 1924 1925 bad: 1926 /* 1927 * ln is valid and the caller did not pass in 1928 * an llentry 1929 */ 1930 if ((ln != NULL) && (lle == NULL)) { 1931 if (flags & LLE_EXCLUSIVE) 1932 LLE_WUNLOCK(ln); 1933 else 1934 LLE_RUNLOCK(ln); 1935 } 1936 if (m) 1937 m_freem(m); 1938 return (error); 1939} 1940#undef senderr 1941 1942 1943int 1944nd6_output_flush(struct ifnet *ifp, struct ifnet *origifp, struct mbuf *chain, 1945 struct sockaddr_in6 *dst, struct route *ro) 1946{ 1947 struct mbuf *m, *m_head; 1948 struct ifnet *outifp; 1949 int error = 0; 1950 1951 m_head = chain; 1952 if ((ifp->if_flags & IFF_LOOPBACK) != 0) 1953 outifp = origifp; 1954 else 1955 outifp = ifp; 1956 1957 while (m_head) { 1958 m = m_head; 1959 m_head = m_head->m_nextpkt; 1960 error = (*ifp->if_output)(ifp, m, (struct sockaddr *)dst, ro); 1961 } 1962 1963 /* 1964 * XXX 1965 * note that intermediate errors are blindly ignored - but this is 1966 * the same convention as used with nd6_output when called by 1967 * nd6_cache_lladdr 1968 */ 1969 return (error); 1970} 1971 1972 1973int 1974nd6_need_cache(struct ifnet *ifp) 1975{ 1976 /* 1977 * XXX: we currently do not make neighbor cache on any interface 1978 * other than ARCnet, Ethernet, FDDI and GIF. 1979 * 1980 * RFC2893 says: 1981 * - unidirectional tunnels needs no ND 1982 */ 1983 switch (ifp->if_type) { 1984 case IFT_ARCNET: 1985 case IFT_ETHER: 1986 case IFT_FDDI: 1987 case IFT_IEEE1394: 1988#ifdef IFT_L2VLAN 1989 case IFT_L2VLAN: 1990#endif 1991#ifdef IFT_IEEE80211 1992 case IFT_IEEE80211: 1993#endif 1994#ifdef IFT_CARP 1995 case IFT_CARP: 1996#endif 1997 case IFT_GIF: /* XXX need more cases? */ 1998 case IFT_PPP: 1999 case IFT_TUNNEL: 2000 case IFT_BRIDGE: 2001 case IFT_PROPVIRTUAL: 2002 return (1); 2003 default: 2004 return (0); 2005 } 2006} 2007 2008/* 2009 * the callers of this function need to be re-worked to drop 2010 * the lle lock, drop here for now 2011 */ 2012int 2013nd6_storelladdr(struct ifnet *ifp, struct mbuf *m, 2014 struct sockaddr *dst, u_char *desten, struct llentry **lle) 2015{ 2016 struct llentry *ln; 2017 2018 *lle = NULL; 2019 IF_AFDATA_UNLOCK_ASSERT(ifp); 2020 if (m->m_flags & M_MCAST) { 2021 int i; 2022 2023 switch (ifp->if_type) { 2024 case IFT_ETHER: 2025 case IFT_FDDI: 2026#ifdef IFT_L2VLAN 2027 case IFT_L2VLAN: 2028#endif 2029#ifdef IFT_IEEE80211 2030 case IFT_IEEE80211: 2031#endif 2032 case IFT_BRIDGE: 2033 case IFT_ISO88025: 2034 ETHER_MAP_IPV6_MULTICAST(&SIN6(dst)->sin6_addr, 2035 desten); 2036 return (0); 2037 case IFT_IEEE1394: 2038 /* 2039 * netbsd can use if_broadcastaddr, but we don't do so 2040 * to reduce # of ifdef. 2041 */ 2042 for (i = 0; i < ifp->if_addrlen; i++) 2043 desten[i] = ~0; 2044 return (0); 2045 case IFT_ARCNET: 2046 *desten = 0; 2047 return (0); 2048 default: 2049 m_freem(m); 2050 return (EAFNOSUPPORT); 2051 } 2052 } 2053 2054 2055 /* 2056 * the entry should have been created in nd6_store_lladdr 2057 */ 2058 IF_AFDATA_LOCK(ifp); 2059 ln = lla_lookup(LLTABLE6(ifp), 0, dst); 2060 IF_AFDATA_UNLOCK(ifp); 2061 if ((ln == NULL) || !(ln->la_flags & LLE_VALID)) { 2062 if (ln != NULL) 2063 LLE_RUNLOCK(ln); 2064 /* this could happen, if we could not allocate memory */ 2065 m_freem(m); 2066 return (1); 2067 } 2068 2069 bcopy(&ln->ll_addr, desten, ifp->if_addrlen); 2070 *lle = ln; 2071 LLE_RUNLOCK(ln); 2072 /* 2073 * A *small* use after free race exists here 2074 */ 2075 return (0); 2076} 2077 2078static void 2079clear_llinfo_pqueue(struct llentry *ln) 2080{ 2081 struct mbuf *m_hold, *m_hold_next; 2082 2083 for (m_hold = ln->la_hold; m_hold; m_hold = m_hold_next) { 2084 m_hold_next = m_hold->m_nextpkt; 2085 m_hold->m_nextpkt = NULL; 2086 m_freem(m_hold); 2087 } 2088 2089 ln->la_hold = NULL; 2090 return; 2091} 2092 2093static int nd6_sysctl_drlist(SYSCTL_HANDLER_ARGS); 2094static int nd6_sysctl_prlist(SYSCTL_HANDLER_ARGS); 2095#ifdef SYSCTL_DECL 2096SYSCTL_DECL(_net_inet6_icmp6); 2097#endif 2098SYSCTL_NODE(_net_inet6_icmp6, ICMPV6CTL_ND6_DRLIST, nd6_drlist, 2099 CTLFLAG_RD, nd6_sysctl_drlist, ""); 2100SYSCTL_NODE(_net_inet6_icmp6, ICMPV6CTL_ND6_PRLIST, nd6_prlist, 2101 CTLFLAG_RD, nd6_sysctl_prlist, ""); 2102SYSCTL_V_INT(V_NET, vnet_inet6, _net_inet6_icmp6, ICMPV6CTL_ND6_MAXQLEN, 2103 nd6_maxqueuelen, CTLFLAG_RW, nd6_maxqueuelen, 1, ""); 2104 2105static int 2106nd6_sysctl_drlist(SYSCTL_HANDLER_ARGS) 2107{ 2108 INIT_VNET_INET6(curvnet); 2109 int error; 2110 char buf[1024] __aligned(4); 2111 struct in6_defrouter *d, *de; 2112 struct nd_defrouter *dr; 2113 2114 if (req->newptr) 2115 return EPERM; 2116 error = 0; 2117 2118 for (dr = TAILQ_FIRST(&V_nd_defrouter); dr; 2119 dr = TAILQ_NEXT(dr, dr_entry)) { 2120 d = (struct in6_defrouter *)buf; 2121 de = (struct in6_defrouter *)(buf + sizeof(buf)); 2122 2123 if (d + 1 <= de) { 2124 bzero(d, sizeof(*d)); 2125 d->rtaddr.sin6_family = AF_INET6; 2126 d->rtaddr.sin6_len = sizeof(d->rtaddr); 2127 d->rtaddr.sin6_addr = dr->rtaddr; 2128 error = sa6_recoverscope(&d->rtaddr); 2129 if (error != 0) 2130 return (error); 2131 d->flags = dr->flags; 2132 d->rtlifetime = dr->rtlifetime; 2133 d->expire = dr->expire; 2134 d->if_index = dr->ifp->if_index; 2135 } else 2136 panic("buffer too short"); 2137 2138 error = SYSCTL_OUT(req, buf, sizeof(*d)); 2139 if (error) 2140 break; 2141 } 2142 2143 return (error); 2144} 2145 2146static int 2147nd6_sysctl_prlist(SYSCTL_HANDLER_ARGS) 2148{ 2149 INIT_VNET_INET6(curvnet); 2150 int error; 2151 char buf[1024] __aligned(4); 2152 struct in6_prefix *p, *pe; 2153 struct nd_prefix *pr; 2154 char ip6buf[INET6_ADDRSTRLEN]; 2155 2156 if (req->newptr) 2157 return EPERM; 2158 error = 0; 2159 2160 for (pr = V_nd_prefix.lh_first; pr; pr = pr->ndpr_next) { 2161 u_short advrtrs; 2162 size_t advance; 2163 struct sockaddr_in6 *sin6, *s6; 2164 struct nd_pfxrouter *pfr; 2165 2166 p = (struct in6_prefix *)buf; 2167 pe = (struct in6_prefix *)(buf + sizeof(buf)); 2168 2169 if (p + 1 <= pe) { 2170 bzero(p, sizeof(*p)); 2171 sin6 = (struct sockaddr_in6 *)(p + 1); 2172 2173 p->prefix = pr->ndpr_prefix; 2174 if (sa6_recoverscope(&p->prefix)) { 2175 log(LOG_ERR, 2176 "scope error in prefix list (%s)\n", 2177 ip6_sprintf(ip6buf, &p->prefix.sin6_addr)); 2178 /* XXX: press on... */ 2179 } 2180 p->raflags = pr->ndpr_raf; 2181 p->prefixlen = pr->ndpr_plen; 2182 p->vltime = pr->ndpr_vltime; 2183 p->pltime = pr->ndpr_pltime; 2184 p->if_index = pr->ndpr_ifp->if_index; 2185 if (pr->ndpr_vltime == ND6_INFINITE_LIFETIME) 2186 p->expire = 0; 2187 else { 2188 time_t maxexpire; 2189 2190 /* XXX: we assume time_t is signed. */ 2191 maxexpire = (-1) & 2192 ~((time_t)1 << 2193 ((sizeof(maxexpire) * 8) - 1)); 2194 if (pr->ndpr_vltime < 2195 maxexpire - pr->ndpr_lastupdate) { 2196 p->expire = pr->ndpr_lastupdate + 2197 pr->ndpr_vltime; 2198 } else 2199 p->expire = maxexpire; 2200 } 2201 p->refcnt = pr->ndpr_refcnt; 2202 p->flags = pr->ndpr_stateflags; 2203 p->origin = PR_ORIG_RA; 2204 advrtrs = 0; 2205 for (pfr = pr->ndpr_advrtrs.lh_first; pfr; 2206 pfr = pfr->pfr_next) { 2207 if ((void *)&sin6[advrtrs + 1] > (void *)pe) { 2208 advrtrs++; 2209 continue; 2210 } 2211 s6 = &sin6[advrtrs]; 2212 bzero(s6, sizeof(*s6)); 2213 s6->sin6_family = AF_INET6; 2214 s6->sin6_len = sizeof(*sin6); 2215 s6->sin6_addr = pfr->router->rtaddr; 2216 if (sa6_recoverscope(s6)) { 2217 log(LOG_ERR, 2218 "scope error in " 2219 "prefix list (%s)\n", 2220 ip6_sprintf(ip6buf, 2221 &pfr->router->rtaddr)); 2222 } 2223 advrtrs++; 2224 } 2225 p->advrtrs = advrtrs; 2226 } else 2227 panic("buffer too short"); 2228 2229 advance = sizeof(*p) + sizeof(*sin6) * advrtrs; 2230 error = SYSCTL_OUT(req, buf, advance); 2231 if (error) 2232 break; 2233 } 2234 2235 return (error); 2236} 2237