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: releng/10.3/sys/netinet6/nd6.c 292566 2015-12-21 20:29:55Z kp $"); 34 35#include "opt_inet.h" 36#include "opt_inet6.h" 37#include "opt_kdtrace.h" 38 39#include <sys/param.h> 40#include <sys/systm.h> 41#include <sys/callout.h> 42#include <sys/malloc.h> 43#include <sys/mbuf.h> 44#include <sys/socket.h> 45#include <sys/sockio.h> 46#include <sys/time.h> 47#include <sys/kernel.h> 48#include <sys/protosw.h> 49#include <sys/errno.h> 50#include <sys/syslog.h> 51#include <sys/lock.h> 52#include <sys/rwlock.h> 53#include <sys/queue.h> 54#include <sys/sdt.h> 55#include <sys/sysctl.h> 56 57#include <net/if.h> 58#include <net/if_arc.h> 59#include <net/if_dl.h> 60#include <net/if_types.h> 61#include <net/iso88025.h> 62#include <net/fddi.h> 63#include <net/route.h> 64#include <net/vnet.h> 65 66#include <netinet/in.h> 67#include <netinet/in_kdtrace.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 <netinet6/in6_ifattach.h> 77#include <netinet/icmp6.h> 78#include <netinet6/send.h> 79 80#include <sys/limits.h> 81 82#include <security/mac/mac_framework.h> 83 84#define ND6_SLOWTIMER_INTERVAL (60 * 60) /* 1 hour */ 85#define ND6_RECALC_REACHTM_INTERVAL (60 * 120) /* 2 hours */ 86 87#define SIN6(s) ((const struct sockaddr_in6 *)(s)) 88 89/* timer values */ 90VNET_DEFINE(int, nd6_prune) = 1; /* walk list every 1 seconds */ 91VNET_DEFINE(int, nd6_delay) = 5; /* delay first probe time 5 second */ 92VNET_DEFINE(int, nd6_umaxtries) = 3; /* maximum unicast query */ 93VNET_DEFINE(int, nd6_mmaxtries) = 3; /* maximum multicast query */ 94VNET_DEFINE(int, nd6_useloopback) = 1; /* use loopback interface for 95 * local traffic */ 96VNET_DEFINE(int, nd6_gctimer) = (60 * 60 * 24); /* 1 day: garbage 97 * collection timer */ 98 99/* preventing too many loops in ND option parsing */ 100static VNET_DEFINE(int, nd6_maxndopt) = 10; /* max # of ND options allowed */ 101 102VNET_DEFINE(int, nd6_maxnudhint) = 0; /* max # of subsequent upper 103 * layer hints */ 104static VNET_DEFINE(int, nd6_maxqueuelen) = 1; /* max pkts cached in unresolved 105 * ND entries */ 106#define V_nd6_maxndopt VNET(nd6_maxndopt) 107#define V_nd6_maxqueuelen VNET(nd6_maxqueuelen) 108 109#ifdef ND6_DEBUG 110VNET_DEFINE(int, nd6_debug) = 1; 111#else 112VNET_DEFINE(int, nd6_debug) = 0; 113#endif 114 115/* for debugging? */ 116#if 0 117static int nd6_inuse, nd6_allocated; 118#endif 119 120VNET_DEFINE(struct nd_drhead, nd_defrouter); 121VNET_DEFINE(struct nd_prhead, nd_prefix); 122 123VNET_DEFINE(int, nd6_recalc_reachtm_interval) = ND6_RECALC_REACHTM_INTERVAL; 124#define V_nd6_recalc_reachtm_interval VNET(nd6_recalc_reachtm_interval) 125 126int (*send_sendso_input_hook)(struct mbuf *, struct ifnet *, int, int); 127 128static int nd6_is_new_addr_neighbor(struct sockaddr_in6 *, 129 struct ifnet *); 130static void nd6_setmtu0(struct ifnet *, struct nd_ifinfo *); 131static void nd6_slowtimo(void *); 132static int regen_tmpaddr(struct in6_ifaddr *); 133static struct llentry *nd6_free(struct llentry *, int); 134static void nd6_llinfo_timer(void *); 135static void clear_llinfo_pqueue(struct llentry *); 136static int nd6_output_lle(struct ifnet *, struct ifnet *, struct mbuf *, 137 struct sockaddr_in6 *); 138static int nd6_output_ifp(struct ifnet *, struct ifnet *, struct mbuf *, 139 struct sockaddr_in6 *); 140 141static VNET_DEFINE(struct callout, nd6_slowtimo_ch); 142#define V_nd6_slowtimo_ch VNET(nd6_slowtimo_ch) 143 144VNET_DEFINE(struct callout, nd6_timer_ch); 145 146void 147nd6_init(void) 148{ 149 150 LIST_INIT(&V_nd_prefix); 151 152 /* initialization of the default router list */ 153 TAILQ_INIT(&V_nd_defrouter); 154 155 /* start timer */ 156 callout_init(&V_nd6_slowtimo_ch, 0); 157 callout_reset(&V_nd6_slowtimo_ch, ND6_SLOWTIMER_INTERVAL * hz, 158 nd6_slowtimo, curvnet); 159 160 nd6_dad_init(); 161} 162 163#ifdef VIMAGE 164void 165nd6_destroy() 166{ 167 168 callout_drain(&V_nd6_slowtimo_ch); 169 callout_drain(&V_nd6_timer_ch); 170} 171#endif 172 173struct nd_ifinfo * 174nd6_ifattach(struct ifnet *ifp) 175{ 176 struct nd_ifinfo *nd; 177 178 nd = (struct nd_ifinfo *)malloc(sizeof(*nd), M_IP6NDP, M_WAITOK|M_ZERO); 179 nd->initialized = 1; 180 181 nd->chlim = IPV6_DEFHLIM; 182 nd->basereachable = REACHABLE_TIME; 183 nd->reachable = ND_COMPUTE_RTIME(nd->basereachable); 184 nd->retrans = RETRANS_TIMER; 185 186 nd->flags = ND6_IFF_PERFORMNUD; 187 188 /* A loopback interface always has ND6_IFF_AUTO_LINKLOCAL. 189 * XXXHRS: Clear ND6_IFF_AUTO_LINKLOCAL on an IFT_BRIDGE interface by 190 * default regardless of the V_ip6_auto_linklocal configuration to 191 * give a reasonable default behavior. 192 */ 193 if ((V_ip6_auto_linklocal && ifp->if_type != IFT_BRIDGE) || 194 (ifp->if_flags & IFF_LOOPBACK)) 195 nd->flags |= ND6_IFF_AUTO_LINKLOCAL; 196 /* 197 * A loopback interface does not need to accept RTADV. 198 * XXXHRS: Clear ND6_IFF_ACCEPT_RTADV on an IFT_BRIDGE interface by 199 * default regardless of the V_ip6_accept_rtadv configuration to 200 * prevent the interface from accepting RA messages arrived 201 * on one of the member interfaces with ND6_IFF_ACCEPT_RTADV. 202 */ 203 if (V_ip6_accept_rtadv && 204 !(ifp->if_flags & IFF_LOOPBACK) && 205 (ifp->if_type != IFT_BRIDGE)) 206 nd->flags |= ND6_IFF_ACCEPT_RTADV; 207 if (V_ip6_no_radr && !(ifp->if_flags & IFF_LOOPBACK)) 208 nd->flags |= ND6_IFF_NO_RADR; 209 210 /* XXX: we cannot call nd6_setmtu since ifp is not fully initialized */ 211 nd6_setmtu0(ifp, nd); 212 213 return nd; 214} 215 216void 217nd6_ifdetach(struct nd_ifinfo *nd) 218{ 219 220 free(nd, M_IP6NDP); 221} 222 223/* 224 * Reset ND level link MTU. This function is called when the physical MTU 225 * changes, which means we might have to adjust the ND level MTU. 226 */ 227void 228nd6_setmtu(struct ifnet *ifp) 229{ 230 if (ifp->if_afdata[AF_INET6] == NULL) 231 return; 232 233 nd6_setmtu0(ifp, ND_IFINFO(ifp)); 234} 235 236/* XXX todo: do not maintain copy of ifp->if_mtu in ndi->maxmtu */ 237void 238nd6_setmtu0(struct ifnet *ifp, struct nd_ifinfo *ndi) 239{ 240 u_int32_t omaxmtu; 241 242 omaxmtu = ndi->maxmtu; 243 244 switch (ifp->if_type) { 245 case IFT_ARCNET: 246 ndi->maxmtu = MIN(ARC_PHDS_MAXMTU, ifp->if_mtu); /* RFC2497 */ 247 break; 248 case IFT_FDDI: 249 ndi->maxmtu = MIN(FDDIIPMTU, ifp->if_mtu); /* RFC2467 */ 250 break; 251 case IFT_ISO88025: 252 ndi->maxmtu = MIN(ISO88025_MAX_MTU, ifp->if_mtu); 253 break; 254 default: 255 ndi->maxmtu = ifp->if_mtu; 256 break; 257 } 258 259 /* 260 * Decreasing the interface MTU under IPV6 minimum MTU may cause 261 * undesirable situation. We thus notify the operator of the change 262 * explicitly. The check for omaxmtu is necessary to restrict the 263 * log to the case of changing the MTU, not initializing it. 264 */ 265 if (omaxmtu >= IPV6_MMTU && ndi->maxmtu < IPV6_MMTU) { 266 log(LOG_NOTICE, "nd6_setmtu0: " 267 "new link MTU on %s (%lu) is too small for IPv6\n", 268 if_name(ifp), (unsigned long)ndi->maxmtu); 269 } 270 271 if (ndi->maxmtu > V_in6_maxmtu) 272 in6_setmaxmtu(); /* check all interfaces just in case */ 273 274} 275 276void 277nd6_option_init(void *opt, int icmp6len, union nd_opts *ndopts) 278{ 279 280 bzero(ndopts, sizeof(*ndopts)); 281 ndopts->nd_opts_search = (struct nd_opt_hdr *)opt; 282 ndopts->nd_opts_last 283 = (struct nd_opt_hdr *)(((u_char *)opt) + icmp6len); 284 285 if (icmp6len == 0) { 286 ndopts->nd_opts_done = 1; 287 ndopts->nd_opts_search = NULL; 288 } 289} 290 291/* 292 * Take one ND option. 293 */ 294struct nd_opt_hdr * 295nd6_option(union nd_opts *ndopts) 296{ 297 struct nd_opt_hdr *nd_opt; 298 int olen; 299 300 KASSERT(ndopts != NULL, ("%s: ndopts == NULL", __func__)); 301 KASSERT(ndopts->nd_opts_last != NULL, ("%s: uninitialized ndopts", 302 __func__)); 303 if (ndopts->nd_opts_search == NULL) 304 return NULL; 305 if (ndopts->nd_opts_done) 306 return NULL; 307 308 nd_opt = ndopts->nd_opts_search; 309 310 /* make sure nd_opt_len is inside the buffer */ 311 if ((caddr_t)&nd_opt->nd_opt_len >= (caddr_t)ndopts->nd_opts_last) { 312 bzero(ndopts, sizeof(*ndopts)); 313 return NULL; 314 } 315 316 olen = nd_opt->nd_opt_len << 3; 317 if (olen == 0) { 318 /* 319 * Message validation requires that all included 320 * options have a length that is greater than zero. 321 */ 322 bzero(ndopts, sizeof(*ndopts)); 323 return NULL; 324 } 325 326 ndopts->nd_opts_search = (struct nd_opt_hdr *)((caddr_t)nd_opt + olen); 327 if (ndopts->nd_opts_search > ndopts->nd_opts_last) { 328 /* option overruns the end of buffer, invalid */ 329 bzero(ndopts, sizeof(*ndopts)); 330 return NULL; 331 } else if (ndopts->nd_opts_search == ndopts->nd_opts_last) { 332 /* reached the end of options chain */ 333 ndopts->nd_opts_done = 1; 334 ndopts->nd_opts_search = NULL; 335 } 336 return nd_opt; 337} 338 339/* 340 * Parse multiple ND options. 341 * This function is much easier to use, for ND routines that do not need 342 * multiple options of the same type. 343 */ 344int 345nd6_options(union nd_opts *ndopts) 346{ 347 struct nd_opt_hdr *nd_opt; 348 int i = 0; 349 350 KASSERT(ndopts != NULL, ("%s: ndopts == NULL", __func__)); 351 KASSERT(ndopts->nd_opts_last != NULL, ("%s: uninitialized ndopts", 352 __func__)); 353 if (ndopts->nd_opts_search == NULL) 354 return 0; 355 356 while (1) { 357 nd_opt = nd6_option(ndopts); 358 if (nd_opt == NULL && ndopts->nd_opts_last == NULL) { 359 /* 360 * Message validation requires that all included 361 * options have a length that is greater than zero. 362 */ 363 ICMP6STAT_INC(icp6s_nd_badopt); 364 bzero(ndopts, sizeof(*ndopts)); 365 return -1; 366 } 367 368 if (nd_opt == NULL) 369 goto skip1; 370 371 switch (nd_opt->nd_opt_type) { 372 case ND_OPT_SOURCE_LINKADDR: 373 case ND_OPT_TARGET_LINKADDR: 374 case ND_OPT_MTU: 375 case ND_OPT_REDIRECTED_HEADER: 376 case ND_OPT_NONCE: 377 if (ndopts->nd_opt_array[nd_opt->nd_opt_type]) { 378 nd6log((LOG_INFO, 379 "duplicated ND6 option found (type=%d)\n", 380 nd_opt->nd_opt_type)); 381 /* XXX bark? */ 382 } else { 383 ndopts->nd_opt_array[nd_opt->nd_opt_type] 384 = nd_opt; 385 } 386 break; 387 case ND_OPT_PREFIX_INFORMATION: 388 if (ndopts->nd_opt_array[nd_opt->nd_opt_type] == 0) { 389 ndopts->nd_opt_array[nd_opt->nd_opt_type] 390 = nd_opt; 391 } 392 ndopts->nd_opts_pi_end = 393 (struct nd_opt_prefix_info *)nd_opt; 394 break; 395 /* What about ND_OPT_ROUTE_INFO? RFC 4191 */ 396 case ND_OPT_RDNSS: /* RFC 6106 */ 397 case ND_OPT_DNSSL: /* RFC 6106 */ 398 /* 399 * Silently ignore options we know and do not care about 400 * in the kernel. 401 */ 402 break; 403 default: 404 /* 405 * Unknown options must be silently ignored, 406 * to accomodate future extension to the protocol. 407 */ 408 nd6log((LOG_DEBUG, 409 "nd6_options: unsupported option %d - " 410 "option ignored\n", nd_opt->nd_opt_type)); 411 } 412 413skip1: 414 i++; 415 if (i > V_nd6_maxndopt) { 416 ICMP6STAT_INC(icp6s_nd_toomanyopt); 417 nd6log((LOG_INFO, "too many loop in nd opt\n")); 418 break; 419 } 420 421 if (ndopts->nd_opts_done) 422 break; 423 } 424 425 return 0; 426} 427 428/* 429 * ND6 timer routine to handle ND6 entries 430 */ 431void 432nd6_llinfo_settimer_locked(struct llentry *ln, long tick) 433{ 434 int canceled; 435 436 LLE_WLOCK_ASSERT(ln); 437 438 if (tick < 0) { 439 ln->la_expire = 0; 440 ln->ln_ntick = 0; 441 canceled = callout_stop(&ln->ln_timer_ch); 442 } else { 443 ln->la_expire = time_uptime + tick / hz; 444 LLE_ADDREF(ln); 445 if (tick > INT_MAX) { 446 ln->ln_ntick = tick - INT_MAX; 447 canceled = callout_reset(&ln->ln_timer_ch, INT_MAX, 448 nd6_llinfo_timer, ln); 449 } else { 450 ln->ln_ntick = 0; 451 canceled = callout_reset(&ln->ln_timer_ch, tick, 452 nd6_llinfo_timer, ln); 453 } 454 } 455 if (canceled) 456 LLE_REMREF(ln); 457} 458 459void 460nd6_llinfo_settimer(struct llentry *ln, long tick) 461{ 462 463 LLE_WLOCK(ln); 464 nd6_llinfo_settimer_locked(ln, tick); 465 LLE_WUNLOCK(ln); 466} 467 468static void 469nd6_llinfo_timer(void *arg) 470{ 471 struct llentry *ln; 472 struct in6_addr *dst; 473 struct ifnet *ifp; 474 struct nd_ifinfo *ndi = NULL; 475 476 KASSERT(arg != NULL, ("%s: arg NULL", __func__)); 477 ln = (struct llentry *)arg; 478 LLE_WLOCK(ln); 479 if (callout_pending(&ln->la_timer)) { 480 /* 481 * Here we are a bit odd here in the treatment of 482 * active/pending. If the pending bit is set, it got 483 * rescheduled before I ran. The active 484 * bit we ignore, since if it was stopped 485 * in ll_tablefree() and was currently running 486 * it would have return 0 so the code would 487 * not have deleted it since the callout could 488 * not be stopped so we want to go through 489 * with the delete here now. If the callout 490 * was restarted, the pending bit will be back on and 491 * we just want to bail since the callout_reset would 492 * return 1 and our reference would have been removed 493 * by nd6_llinfo_settimer_locked above since canceled 494 * would have been 1. 495 */ 496 LLE_WUNLOCK(ln); 497 return; 498 } 499 ifp = ln->lle_tbl->llt_ifp; 500 CURVNET_SET(ifp->if_vnet); 501 502 if (ln->ln_ntick > 0) { 503 if (ln->ln_ntick > INT_MAX) { 504 ln->ln_ntick -= INT_MAX; 505 nd6_llinfo_settimer_locked(ln, INT_MAX); 506 } else { 507 ln->ln_ntick = 0; 508 nd6_llinfo_settimer_locked(ln, ln->ln_ntick); 509 } 510 goto done; 511 } 512 513 ndi = ND_IFINFO(ifp); 514 dst = &L3_ADDR_SIN6(ln)->sin6_addr; 515 if (ln->la_flags & LLE_STATIC) { 516 goto done; 517 } 518 519 if (ln->la_flags & LLE_DELETED) { 520 (void)nd6_free(ln, 0); 521 ln = NULL; 522 goto done; 523 } 524 525 switch (ln->ln_state) { 526 case ND6_LLINFO_INCOMPLETE: 527 if (ln->la_asked < V_nd6_mmaxtries) { 528 ln->la_asked++; 529 nd6_llinfo_settimer_locked(ln, (long)ndi->retrans * hz / 1000); 530 LLE_WUNLOCK(ln); 531 nd6_ns_output(ifp, NULL, dst, ln, NULL); 532 LLE_WLOCK(ln); 533 } else { 534 struct mbuf *m = ln->la_hold; 535 if (m) { 536 struct mbuf *m0; 537 538 /* 539 * assuming every packet in la_hold has the 540 * same IP header. Send error after unlock. 541 */ 542 m0 = m->m_nextpkt; 543 m->m_nextpkt = NULL; 544 ln->la_hold = m0; 545 clear_llinfo_pqueue(ln); 546 } 547 EVENTHANDLER_INVOKE(lle_event, ln, LLENTRY_TIMEDOUT); 548 (void)nd6_free(ln, 0); 549 ln = NULL; 550 if (m != NULL) 551 icmp6_error2(m, ICMP6_DST_UNREACH, 552 ICMP6_DST_UNREACH_ADDR, 0, ifp); 553 } 554 break; 555 case ND6_LLINFO_REACHABLE: 556 if (!ND6_LLINFO_PERMANENT(ln)) { 557 ln->ln_state = ND6_LLINFO_STALE; 558 nd6_llinfo_settimer_locked(ln, (long)V_nd6_gctimer * hz); 559 } 560 break; 561 562 case ND6_LLINFO_STALE: 563 /* Garbage Collection(RFC 2461 5.3) */ 564 if (!ND6_LLINFO_PERMANENT(ln)) { 565 EVENTHANDLER_INVOKE(lle_event, ln, LLENTRY_EXPIRED); 566 (void)nd6_free(ln, 1); 567 ln = NULL; 568 } 569 break; 570 571 case ND6_LLINFO_DELAY: 572 if (ndi && (ndi->flags & ND6_IFF_PERFORMNUD) != 0) { 573 /* We need NUD */ 574 ln->la_asked = 1; 575 ln->ln_state = ND6_LLINFO_PROBE; 576 nd6_llinfo_settimer_locked(ln, (long)ndi->retrans * hz / 1000); 577 LLE_WUNLOCK(ln); 578 nd6_ns_output(ifp, dst, dst, ln, NULL); 579 LLE_WLOCK(ln); 580 } else { 581 ln->ln_state = ND6_LLINFO_STALE; /* XXX */ 582 nd6_llinfo_settimer_locked(ln, (long)V_nd6_gctimer * hz); 583 } 584 break; 585 case ND6_LLINFO_PROBE: 586 if (ln->la_asked < V_nd6_umaxtries) { 587 ln->la_asked++; 588 nd6_llinfo_settimer_locked(ln, (long)ndi->retrans * hz / 1000); 589 LLE_WUNLOCK(ln); 590 nd6_ns_output(ifp, dst, dst, ln, NULL); 591 LLE_WLOCK(ln); 592 } else { 593 EVENTHANDLER_INVOKE(lle_event, ln, LLENTRY_EXPIRED); 594 (void)nd6_free(ln, 0); 595 ln = NULL; 596 } 597 break; 598 default: 599 panic("%s: paths in a dark night can be confusing: %d", 600 __func__, ln->ln_state); 601 } 602done: 603 if (ln != NULL) 604 LLE_FREE_LOCKED(ln); 605 CURVNET_RESTORE(); 606} 607 608 609/* 610 * ND6 timer routine to expire default route list and prefix list 611 */ 612void 613nd6_timer(void *arg) 614{ 615 CURVNET_SET((struct vnet *) arg); 616 struct nd_defrouter *dr, *ndr; 617 struct nd_prefix *pr, *npr; 618 struct in6_ifaddr *ia6, *nia6; 619 620 callout_reset(&V_nd6_timer_ch, V_nd6_prune * hz, 621 nd6_timer, curvnet); 622 623 /* expire default router list */ 624 TAILQ_FOREACH_SAFE(dr, &V_nd_defrouter, dr_entry, ndr) { 625 if (dr->expire && dr->expire < time_uptime) 626 defrtrlist_del(dr); 627 } 628 629 /* 630 * expire interface addresses. 631 * in the past the loop was inside prefix expiry processing. 632 * However, from a stricter speci-confrmance standpoint, we should 633 * rather separate address lifetimes and prefix lifetimes. 634 * 635 * XXXRW: in6_ifaddrhead locking. 636 */ 637 addrloop: 638 TAILQ_FOREACH_SAFE(ia6, &V_in6_ifaddrhead, ia_link, nia6) { 639 /* check address lifetime */ 640 if (IFA6_IS_INVALID(ia6)) { 641 int regen = 0; 642 643 /* 644 * If the expiring address is temporary, try 645 * regenerating a new one. This would be useful when 646 * we suspended a laptop PC, then turned it on after a 647 * period that could invalidate all temporary 648 * addresses. Although we may have to restart the 649 * loop (see below), it must be after purging the 650 * address. Otherwise, we'd see an infinite loop of 651 * regeneration. 652 */ 653 if (V_ip6_use_tempaddr && 654 (ia6->ia6_flags & IN6_IFF_TEMPORARY) != 0) { 655 if (regen_tmpaddr(ia6) == 0) 656 regen = 1; 657 } 658 659 in6_purgeaddr(&ia6->ia_ifa); 660 661 if (regen) 662 goto addrloop; /* XXX: see below */ 663 } else if (IFA6_IS_DEPRECATED(ia6)) { 664 int oldflags = ia6->ia6_flags; 665 666 ia6->ia6_flags |= IN6_IFF_DEPRECATED; 667 668 /* 669 * If a temporary address has just become deprecated, 670 * regenerate a new one if possible. 671 */ 672 if (V_ip6_use_tempaddr && 673 (ia6->ia6_flags & IN6_IFF_TEMPORARY) != 0 && 674 (oldflags & IN6_IFF_DEPRECATED) == 0) { 675 676 if (regen_tmpaddr(ia6) == 0) { 677 /* 678 * A new temporary address is 679 * generated. 680 * XXX: this means the address chain 681 * has changed while we are still in 682 * the loop. Although the change 683 * would not cause disaster (because 684 * it's not a deletion, but an 685 * addition,) we'd rather restart the 686 * loop just for safety. Or does this 687 * significantly reduce performance?? 688 */ 689 goto addrloop; 690 } 691 } 692 } else if ((ia6->ia6_flags & IN6_IFF_TENTATIVE) != 0) { 693 /* 694 * Schedule DAD for a tentative address. This happens 695 * if the interface was down or not running 696 * when the address was configured. 697 */ 698 int delay; 699 700 delay = arc4random() % 701 (MAX_RTR_SOLICITATION_DELAY * hz); 702 nd6_dad_start((struct ifaddr *)ia6, delay); 703 } else { 704 /* 705 * Check status of the interface. If it is down, 706 * mark the address as tentative for future DAD. 707 */ 708 if ((ia6->ia_ifp->if_flags & IFF_UP) == 0 || 709 (ia6->ia_ifp->if_drv_flags & IFF_DRV_RUNNING) 710 == 0 || 711 (ND_IFINFO(ia6->ia_ifp)->flags & 712 ND6_IFF_IFDISABLED) != 0) { 713 ia6->ia6_flags &= ~IN6_IFF_DUPLICATED; 714 ia6->ia6_flags |= IN6_IFF_TENTATIVE; 715 } 716 /* 717 * A new RA might have made a deprecated address 718 * preferred. 719 */ 720 ia6->ia6_flags &= ~IN6_IFF_DEPRECATED; 721 } 722 } 723 724 /* expire prefix list */ 725 LIST_FOREACH_SAFE(pr, &V_nd_prefix, ndpr_entry, npr) { 726 /* 727 * check prefix lifetime. 728 * since pltime is just for autoconf, pltime processing for 729 * prefix is not necessary. 730 */ 731 if (pr->ndpr_vltime != ND6_INFINITE_LIFETIME && 732 time_uptime - pr->ndpr_lastupdate > pr->ndpr_vltime) { 733 734 /* 735 * address expiration and prefix expiration are 736 * separate. NEVER perform in6_purgeaddr here. 737 */ 738 prelist_remove(pr); 739 } 740 } 741 CURVNET_RESTORE(); 742} 743 744/* 745 * ia6 - deprecated/invalidated temporary address 746 */ 747static int 748regen_tmpaddr(struct in6_ifaddr *ia6) 749{ 750 struct ifaddr *ifa; 751 struct ifnet *ifp; 752 struct in6_ifaddr *public_ifa6 = NULL; 753 754 ifp = ia6->ia_ifa.ifa_ifp; 755 IF_ADDR_RLOCK(ifp); 756 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 757 struct in6_ifaddr *it6; 758 759 if (ifa->ifa_addr->sa_family != AF_INET6) 760 continue; 761 762 it6 = (struct in6_ifaddr *)ifa; 763 764 /* ignore no autoconf addresses. */ 765 if ((it6->ia6_flags & IN6_IFF_AUTOCONF) == 0) 766 continue; 767 768 /* ignore autoconf addresses with different prefixes. */ 769 if (it6->ia6_ndpr == NULL || it6->ia6_ndpr != ia6->ia6_ndpr) 770 continue; 771 772 /* 773 * Now we are looking at an autoconf address with the same 774 * prefix as ours. If the address is temporary and is still 775 * preferred, do not create another one. It would be rare, but 776 * could happen, for example, when we resume a laptop PC after 777 * a long period. 778 */ 779 if ((it6->ia6_flags & IN6_IFF_TEMPORARY) != 0 && 780 !IFA6_IS_DEPRECATED(it6)) { 781 public_ifa6 = NULL; 782 break; 783 } 784 785 /* 786 * This is a public autoconf address that has the same prefix 787 * as ours. If it is preferred, keep it. We can't break the 788 * loop here, because there may be a still-preferred temporary 789 * address with the prefix. 790 */ 791 if (!IFA6_IS_DEPRECATED(it6)) 792 public_ifa6 = it6; 793 } 794 if (public_ifa6 != NULL) 795 ifa_ref(&public_ifa6->ia_ifa); 796 IF_ADDR_RUNLOCK(ifp); 797 798 if (public_ifa6 != NULL) { 799 int e; 800 801 if ((e = in6_tmpifadd(public_ifa6, 0, 0)) != 0) { 802 ifa_free(&public_ifa6->ia_ifa); 803 log(LOG_NOTICE, "regen_tmpaddr: failed to create a new" 804 " tmp addr,errno=%d\n", e); 805 return (-1); 806 } 807 ifa_free(&public_ifa6->ia_ifa); 808 return (0); 809 } 810 811 return (-1); 812} 813 814/* 815 * Nuke neighbor cache/prefix/default router management table, right before 816 * ifp goes away. 817 */ 818void 819nd6_purge(struct ifnet *ifp) 820{ 821 struct nd_defrouter *dr, *ndr; 822 struct nd_prefix *pr, *npr; 823 824 /* 825 * Nuke default router list entries toward ifp. 826 * We defer removal of default router list entries that is installed 827 * in the routing table, in order to keep additional side effects as 828 * small as possible. 829 */ 830 TAILQ_FOREACH_SAFE(dr, &V_nd_defrouter, dr_entry, ndr) { 831 if (dr->installed) 832 continue; 833 834 if (dr->ifp == ifp) 835 defrtrlist_del(dr); 836 } 837 838 TAILQ_FOREACH_SAFE(dr, &V_nd_defrouter, dr_entry, ndr) { 839 if (!dr->installed) 840 continue; 841 842 if (dr->ifp == ifp) 843 defrtrlist_del(dr); 844 } 845 846 /* Nuke prefix list entries toward ifp */ 847 LIST_FOREACH_SAFE(pr, &V_nd_prefix, ndpr_entry, npr) { 848 if (pr->ndpr_ifp == ifp) { 849 /* 850 * Because if_detach() does *not* release prefixes 851 * while purging addresses the reference count will 852 * still be above zero. We therefore reset it to 853 * make sure that the prefix really gets purged. 854 */ 855 pr->ndpr_refcnt = 0; 856 857 /* 858 * Previously, pr->ndpr_addr is removed as well, 859 * but I strongly believe we don't have to do it. 860 * nd6_purge() is only called from in6_ifdetach(), 861 * which removes all the associated interface addresses 862 * by itself. 863 * (jinmei@kame.net 20010129) 864 */ 865 prelist_remove(pr); 866 } 867 } 868 869 /* cancel default outgoing interface setting */ 870 if (V_nd6_defifindex == ifp->if_index) 871 nd6_setdefaultiface(0); 872 873 if (ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV) { 874 /* Refresh default router list. */ 875 defrouter_select(); 876 } 877 878 /* XXXXX 879 * We do not nuke the neighbor cache entries here any more 880 * because the neighbor cache is kept in if_afdata[AF_INET6]. 881 * nd6_purge() is invoked by in6_ifdetach() which is called 882 * from if_detach() where everything gets purged. So let 883 * in6_domifdetach() do the actual L2 table purging work. 884 */ 885} 886 887/* 888 * the caller acquires and releases the lock on the lltbls 889 * Returns the llentry locked 890 */ 891struct llentry * 892nd6_lookup(struct in6_addr *addr6, int flags, struct ifnet *ifp) 893{ 894 struct sockaddr_in6 sin6; 895 struct llentry *ln; 896 int llflags; 897 898 bzero(&sin6, sizeof(sin6)); 899 sin6.sin6_len = sizeof(struct sockaddr_in6); 900 sin6.sin6_family = AF_INET6; 901 sin6.sin6_addr = *addr6; 902 903 IF_AFDATA_LOCK_ASSERT(ifp); 904 905 llflags = 0; 906 if (flags & ND6_CREATE) 907 llflags |= LLE_CREATE; 908 if (flags & ND6_EXCLUSIVE) 909 llflags |= LLE_EXCLUSIVE; 910 911 ln = lla_lookup(LLTABLE6(ifp), llflags, (struct sockaddr *)&sin6); 912 if ((ln != NULL) && (llflags & LLE_CREATE)) 913 ln->ln_state = ND6_LLINFO_NOSTATE; 914 915 return (ln); 916} 917 918/* 919 * Test whether a given IPv6 address is a neighbor or not, ignoring 920 * the actual neighbor cache. The neighbor cache is ignored in order 921 * to not reenter the routing code from within itself. 922 */ 923static int 924nd6_is_new_addr_neighbor(struct sockaddr_in6 *addr, struct ifnet *ifp) 925{ 926 struct nd_prefix *pr; 927 struct ifaddr *dstaddr; 928 929 /* 930 * A link-local address is always a neighbor. 931 * XXX: a link does not necessarily specify a single interface. 932 */ 933 if (IN6_IS_ADDR_LINKLOCAL(&addr->sin6_addr)) { 934 struct sockaddr_in6 sin6_copy; 935 u_int32_t zone; 936 937 /* 938 * We need sin6_copy since sa6_recoverscope() may modify the 939 * content (XXX). 940 */ 941 sin6_copy = *addr; 942 if (sa6_recoverscope(&sin6_copy)) 943 return (0); /* XXX: should be impossible */ 944 if (in6_setscope(&sin6_copy.sin6_addr, ifp, &zone)) 945 return (0); 946 if (sin6_copy.sin6_scope_id == zone) 947 return (1); 948 else 949 return (0); 950 } 951 952 /* 953 * If the address matches one of our addresses, 954 * it should be a neighbor. 955 * If the address matches one of our on-link prefixes, it should be a 956 * neighbor. 957 */ 958 LIST_FOREACH(pr, &V_nd_prefix, ndpr_entry) { 959 if (pr->ndpr_ifp != ifp) 960 continue; 961 962 if (!(pr->ndpr_stateflags & NDPRF_ONLINK)) { 963 struct rtentry *rt; 964 965 /* Always use the default FIB here. */ 966 rt = in6_rtalloc1((struct sockaddr *)&pr->ndpr_prefix, 967 0, 0, RT_DEFAULT_FIB); 968 if (rt == NULL) 969 continue; 970 /* 971 * This is the case where multiple interfaces 972 * have the same prefix, but only one is installed 973 * into the routing table and that prefix entry 974 * is not the one being examined here. In the case 975 * where RADIX_MPATH is enabled, multiple route 976 * entries (of the same rt_key value) will be 977 * installed because the interface addresses all 978 * differ. 979 */ 980 if (!IN6_ARE_ADDR_EQUAL(&pr->ndpr_prefix.sin6_addr, 981 &((struct sockaddr_in6 *)rt_key(rt))->sin6_addr)) { 982 RTFREE_LOCKED(rt); 983 continue; 984 } 985 RTFREE_LOCKED(rt); 986 } 987 988 if (IN6_ARE_MASKED_ADDR_EQUAL(&pr->ndpr_prefix.sin6_addr, 989 &addr->sin6_addr, &pr->ndpr_mask)) 990 return (1); 991 } 992 993 /* 994 * If the address is assigned on the node of the other side of 995 * a p2p interface, the address should be a neighbor. 996 */ 997 dstaddr = ifa_ifwithdstaddr((struct sockaddr *)addr); 998 if (dstaddr != NULL) { 999 if (dstaddr->ifa_ifp == ifp) { 1000 ifa_free(dstaddr); 1001 return (1); 1002 } 1003 ifa_free(dstaddr); 1004 } 1005 1006 /* 1007 * If the default router list is empty, all addresses are regarded 1008 * as on-link, and thus, as a neighbor. 1009 */ 1010 if (ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV && 1011 TAILQ_EMPTY(&V_nd_defrouter) && 1012 V_nd6_defifindex == ifp->if_index) { 1013 return (1); 1014 } 1015 1016 return (0); 1017} 1018 1019 1020/* 1021 * Detect if a given IPv6 address identifies a neighbor on a given link. 1022 * XXX: should take care of the destination of a p2p link? 1023 */ 1024int 1025nd6_is_addr_neighbor(struct sockaddr_in6 *addr, struct ifnet *ifp) 1026{ 1027 struct llentry *lle; 1028 int rc = 0; 1029 1030 IF_AFDATA_UNLOCK_ASSERT(ifp); 1031 if (nd6_is_new_addr_neighbor(addr, ifp)) 1032 return (1); 1033 1034 /* 1035 * Even if the address matches none of our addresses, it might be 1036 * in the neighbor cache. 1037 */ 1038 IF_AFDATA_RLOCK(ifp); 1039 if ((lle = nd6_lookup(&addr->sin6_addr, 0, ifp)) != NULL) { 1040 LLE_RUNLOCK(lle); 1041 rc = 1; 1042 } 1043 IF_AFDATA_RUNLOCK(ifp); 1044 return (rc); 1045} 1046 1047/* 1048 * Free an nd6 llinfo entry. 1049 * Since the function would cause significant changes in the kernel, DO NOT 1050 * make it global, unless you have a strong reason for the change, and are sure 1051 * that the change is safe. 1052 */ 1053static struct llentry * 1054nd6_free(struct llentry *ln, int gc) 1055{ 1056 struct llentry *next; 1057 struct nd_defrouter *dr; 1058 struct ifnet *ifp; 1059 1060 LLE_WLOCK_ASSERT(ln); 1061 1062 /* 1063 * we used to have pfctlinput(PRC_HOSTDEAD) here. 1064 * even though it is not harmful, it was not really necessary. 1065 */ 1066 1067 /* cancel timer */ 1068 nd6_llinfo_settimer_locked(ln, -1); 1069 1070 ifp = ln->lle_tbl->llt_ifp; 1071 1072 if (ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV) { 1073 dr = defrouter_lookup(&L3_ADDR_SIN6(ln)->sin6_addr, ifp); 1074 1075 if (dr != NULL && dr->expire && 1076 ln->ln_state == ND6_LLINFO_STALE && gc) { 1077 /* 1078 * If the reason for the deletion is just garbage 1079 * collection, and the neighbor is an active default 1080 * router, do not delete it. Instead, reset the GC 1081 * timer using the router's lifetime. 1082 * Simply deleting the entry would affect default 1083 * router selection, which is not necessarily a good 1084 * thing, especially when we're using router preference 1085 * values. 1086 * XXX: the check for ln_state would be redundant, 1087 * but we intentionally keep it just in case. 1088 */ 1089 if (dr->expire > time_uptime) 1090 nd6_llinfo_settimer_locked(ln, 1091 (dr->expire - time_uptime) * hz); 1092 else 1093 nd6_llinfo_settimer_locked(ln, 1094 (long)V_nd6_gctimer * hz); 1095 1096 next = LIST_NEXT(ln, lle_next); 1097 LLE_REMREF(ln); 1098 LLE_WUNLOCK(ln); 1099 return (next); 1100 } 1101 1102 if (dr) { 1103 /* 1104 * Unreachablity of a router might affect the default 1105 * router selection and on-link detection of advertised 1106 * prefixes. 1107 */ 1108 1109 /* 1110 * Temporarily fake the state to choose a new default 1111 * router and to perform on-link determination of 1112 * prefixes correctly. 1113 * Below the state will be set correctly, 1114 * or the entry itself will be deleted. 1115 */ 1116 ln->ln_state = ND6_LLINFO_INCOMPLETE; 1117 } 1118 1119 if (ln->ln_router || dr) { 1120 1121 /* 1122 * We need to unlock to avoid a LOR with rt6_flush() with the 1123 * rnh and for the calls to pfxlist_onlink_check() and 1124 * defrouter_select() in the block further down for calls 1125 * into nd6_lookup(). We still hold a ref. 1126 */ 1127 LLE_WUNLOCK(ln); 1128 1129 /* 1130 * rt6_flush must be called whether or not the neighbor 1131 * is in the Default Router List. 1132 * See a corresponding comment in nd6_na_input(). 1133 */ 1134 rt6_flush(&L3_ADDR_SIN6(ln)->sin6_addr, ifp); 1135 } 1136 1137 if (dr) { 1138 /* 1139 * Since defrouter_select() does not affect the 1140 * on-link determination and MIP6 needs the check 1141 * before the default router selection, we perform 1142 * the check now. 1143 */ 1144 pfxlist_onlink_check(); 1145 1146 /* 1147 * Refresh default router list. 1148 */ 1149 defrouter_select(); 1150 } 1151 1152 if (ln->ln_router || dr) 1153 LLE_WLOCK(ln); 1154 } 1155 1156 /* 1157 * Before deleting the entry, remember the next entry as the 1158 * return value. We need this because pfxlist_onlink_check() above 1159 * might have freed other entries (particularly the old next entry) as 1160 * a side effect (XXX). 1161 */ 1162 next = LIST_NEXT(ln, lle_next); 1163 1164 /* 1165 * Save to unlock. We still hold an extra reference and will not 1166 * free(9) in llentry_free() if someone else holds one as well. 1167 */ 1168 LLE_WUNLOCK(ln); 1169 IF_AFDATA_LOCK(ifp); 1170 LLE_WLOCK(ln); 1171 1172 /* Guard against race with other llentry_free(). */ 1173 if (ln->la_flags & LLE_LINKED) { 1174 LLE_REMREF(ln); 1175 llentry_free(ln); 1176 } else 1177 LLE_FREE_LOCKED(ln); 1178 1179 IF_AFDATA_UNLOCK(ifp); 1180 1181 return (next); 1182} 1183 1184/* 1185 * Upper-layer reachability hint for Neighbor Unreachability Detection. 1186 * 1187 * XXX cost-effective methods? 1188 */ 1189void 1190nd6_nud_hint(struct rtentry *rt, struct in6_addr *dst6, int force) 1191{ 1192 struct llentry *ln; 1193 struct ifnet *ifp; 1194 1195 if ((dst6 == NULL) || (rt == NULL)) 1196 return; 1197 1198 ifp = rt->rt_ifp; 1199 IF_AFDATA_RLOCK(ifp); 1200 ln = nd6_lookup(dst6, ND6_EXCLUSIVE, NULL); 1201 IF_AFDATA_RUNLOCK(ifp); 1202 if (ln == NULL) 1203 return; 1204 1205 if (ln->ln_state < ND6_LLINFO_REACHABLE) 1206 goto done; 1207 1208 /* 1209 * if we get upper-layer reachability confirmation many times, 1210 * it is possible we have false information. 1211 */ 1212 if (!force) { 1213 ln->ln_byhint++; 1214 if (ln->ln_byhint > V_nd6_maxnudhint) { 1215 goto done; 1216 } 1217 } 1218 1219 ln->ln_state = ND6_LLINFO_REACHABLE; 1220 if (!ND6_LLINFO_PERMANENT(ln)) { 1221 nd6_llinfo_settimer_locked(ln, 1222 (long)ND_IFINFO(rt->rt_ifp)->reachable * hz); 1223 } 1224done: 1225 LLE_WUNLOCK(ln); 1226} 1227 1228 1229/* 1230 * Rejuvenate this function for routing operations related 1231 * processing. 1232 */ 1233void 1234nd6_rtrequest(int req, struct rtentry *rt, struct rt_addrinfo *info) 1235{ 1236 struct sockaddr_in6 *gateway; 1237 struct nd_defrouter *dr; 1238 struct ifnet *ifp; 1239 1240 RT_LOCK_ASSERT(rt); 1241 gateway = (struct sockaddr_in6 *)rt->rt_gateway; 1242 ifp = rt->rt_ifp; 1243 1244 switch (req) { 1245 case RTM_ADD: 1246 break; 1247 1248 case RTM_DELETE: 1249 if (!ifp) 1250 return; 1251 /* 1252 * Only indirect routes are interesting. 1253 */ 1254 if ((rt->rt_flags & RTF_GATEWAY) == 0) 1255 return; 1256 /* 1257 * check for default route 1258 */ 1259 if (IN6_ARE_ADDR_EQUAL(&in6addr_any, 1260 &SIN6(rt_key(rt))->sin6_addr)) { 1261 1262 dr = defrouter_lookup(&gateway->sin6_addr, ifp); 1263 if (dr != NULL) 1264 dr->installed = 0; 1265 } 1266 break; 1267 } 1268} 1269 1270 1271int 1272nd6_ioctl(u_long cmd, caddr_t data, struct ifnet *ifp) 1273{ 1274 struct in6_ndireq *ndi = (struct in6_ndireq *)data; 1275 struct in6_nbrinfo *nbi = (struct in6_nbrinfo *)data; 1276 struct in6_ndifreq *ndif = (struct in6_ndifreq *)data; 1277 int error = 0; 1278 1279 if (ifp->if_afdata[AF_INET6] == NULL) 1280 return (EPFNOSUPPORT); 1281 switch (cmd) { 1282 case OSIOCGIFINFO_IN6: 1283#define ND ndi->ndi 1284 /* XXX: old ndp(8) assumes a positive value for linkmtu. */ 1285 bzero(&ND, sizeof(ND)); 1286 ND.linkmtu = IN6_LINKMTU(ifp); 1287 ND.maxmtu = ND_IFINFO(ifp)->maxmtu; 1288 ND.basereachable = ND_IFINFO(ifp)->basereachable; 1289 ND.reachable = ND_IFINFO(ifp)->reachable; 1290 ND.retrans = ND_IFINFO(ifp)->retrans; 1291 ND.flags = ND_IFINFO(ifp)->flags; 1292 ND.recalctm = ND_IFINFO(ifp)->recalctm; 1293 ND.chlim = ND_IFINFO(ifp)->chlim; 1294 break; 1295 case SIOCGIFINFO_IN6: 1296 ND = *ND_IFINFO(ifp); 1297 break; 1298 case SIOCSIFINFO_IN6: 1299 /* 1300 * used to change host variables from userland. 1301 * intented for a use on router to reflect RA configurations. 1302 */ 1303 /* 0 means 'unspecified' */ 1304 if (ND.linkmtu != 0) { 1305 if (ND.linkmtu < IPV6_MMTU || 1306 ND.linkmtu > IN6_LINKMTU(ifp)) { 1307 error = EINVAL; 1308 break; 1309 } 1310 ND_IFINFO(ifp)->linkmtu = ND.linkmtu; 1311 } 1312 1313 if (ND.basereachable != 0) { 1314 int obasereachable = ND_IFINFO(ifp)->basereachable; 1315 1316 ND_IFINFO(ifp)->basereachable = ND.basereachable; 1317 if (ND.basereachable != obasereachable) 1318 ND_IFINFO(ifp)->reachable = 1319 ND_COMPUTE_RTIME(ND.basereachable); 1320 } 1321 if (ND.retrans != 0) 1322 ND_IFINFO(ifp)->retrans = ND.retrans; 1323 if (ND.chlim != 0) 1324 ND_IFINFO(ifp)->chlim = ND.chlim; 1325 /* FALLTHROUGH */ 1326 case SIOCSIFINFO_FLAGS: 1327 { 1328 struct ifaddr *ifa; 1329 struct in6_ifaddr *ia; 1330 1331 if ((ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) && 1332 !(ND.flags & ND6_IFF_IFDISABLED)) { 1333 /* ifdisabled 1->0 transision */ 1334 1335 /* 1336 * If the interface is marked as ND6_IFF_IFDISABLED and 1337 * has an link-local address with IN6_IFF_DUPLICATED, 1338 * do not clear ND6_IFF_IFDISABLED. 1339 * See RFC 4862, Section 5.4.5. 1340 */ 1341 IF_ADDR_RLOCK(ifp); 1342 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1343 if (ifa->ifa_addr->sa_family != AF_INET6) 1344 continue; 1345 ia = (struct in6_ifaddr *)ifa; 1346 if ((ia->ia6_flags & IN6_IFF_DUPLICATED) && 1347 IN6_IS_ADDR_LINKLOCAL(IA6_IN6(ia))) 1348 break; 1349 } 1350 IF_ADDR_RUNLOCK(ifp); 1351 1352 if (ifa != NULL) { 1353 /* LLA is duplicated. */ 1354 ND.flags |= ND6_IFF_IFDISABLED; 1355 log(LOG_ERR, "Cannot enable an interface" 1356 " with a link-local address marked" 1357 " duplicate.\n"); 1358 } else { 1359 ND_IFINFO(ifp)->flags &= ~ND6_IFF_IFDISABLED; 1360 if (ifp->if_flags & IFF_UP) 1361 in6_if_up(ifp); 1362 } 1363 } else if (!(ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) && 1364 (ND.flags & ND6_IFF_IFDISABLED)) { 1365 /* ifdisabled 0->1 transision */ 1366 /* Mark all IPv6 address as tentative. */ 1367 1368 ND_IFINFO(ifp)->flags |= ND6_IFF_IFDISABLED; 1369 if (V_ip6_dad_count > 0 && 1370 (ND_IFINFO(ifp)->flags & ND6_IFF_NO_DAD) == 0) { 1371 IF_ADDR_RLOCK(ifp); 1372 TAILQ_FOREACH(ifa, &ifp->if_addrhead, 1373 ifa_link) { 1374 if (ifa->ifa_addr->sa_family != 1375 AF_INET6) 1376 continue; 1377 ia = (struct in6_ifaddr *)ifa; 1378 ia->ia6_flags |= IN6_IFF_TENTATIVE; 1379 } 1380 IF_ADDR_RUNLOCK(ifp); 1381 } 1382 } 1383 1384 if (ND.flags & ND6_IFF_AUTO_LINKLOCAL) { 1385 if (!(ND_IFINFO(ifp)->flags & ND6_IFF_AUTO_LINKLOCAL)) { 1386 /* auto_linklocal 0->1 transision */ 1387 1388 /* If no link-local address on ifp, configure */ 1389 ND_IFINFO(ifp)->flags |= ND6_IFF_AUTO_LINKLOCAL; 1390 in6_ifattach(ifp, NULL); 1391 } else if (!(ND.flags & ND6_IFF_IFDISABLED) && 1392 ifp->if_flags & IFF_UP) { 1393 /* 1394 * When the IF already has 1395 * ND6_IFF_AUTO_LINKLOCAL, no link-local 1396 * address is assigned, and IFF_UP, try to 1397 * assign one. 1398 */ 1399 IF_ADDR_RLOCK(ifp); 1400 TAILQ_FOREACH(ifa, &ifp->if_addrhead, 1401 ifa_link) { 1402 if (ifa->ifa_addr->sa_family != 1403 AF_INET6) 1404 continue; 1405 ia = (struct in6_ifaddr *)ifa; 1406 if (IN6_IS_ADDR_LINKLOCAL(IA6_IN6(ia))) 1407 break; 1408 } 1409 IF_ADDR_RUNLOCK(ifp); 1410 if (ifa != NULL) 1411 /* No LLA is configured. */ 1412 in6_ifattach(ifp, NULL); 1413 } 1414 } 1415 } 1416 ND_IFINFO(ifp)->flags = ND.flags; 1417 break; 1418#undef ND 1419 case SIOCSNDFLUSH_IN6: /* XXX: the ioctl name is confusing... */ 1420 /* sync kernel routing table with the default router list */ 1421 defrouter_reset(); 1422 defrouter_select(); 1423 break; 1424 case SIOCSPFXFLUSH_IN6: 1425 { 1426 /* flush all the prefix advertised by routers */ 1427 struct nd_prefix *pr, *next; 1428 1429 LIST_FOREACH_SAFE(pr, &V_nd_prefix, ndpr_entry, next) { 1430 struct in6_ifaddr *ia, *ia_next; 1431 1432 if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr)) 1433 continue; /* XXX */ 1434 1435 /* do we really have to remove addresses as well? */ 1436 /* XXXRW: in6_ifaddrhead locking. */ 1437 TAILQ_FOREACH_SAFE(ia, &V_in6_ifaddrhead, ia_link, 1438 ia_next) { 1439 if ((ia->ia6_flags & IN6_IFF_AUTOCONF) == 0) 1440 continue; 1441 1442 if (ia->ia6_ndpr == pr) 1443 in6_purgeaddr(&ia->ia_ifa); 1444 } 1445 prelist_remove(pr); 1446 } 1447 break; 1448 } 1449 case SIOCSRTRFLUSH_IN6: 1450 { 1451 /* flush all the default routers */ 1452 struct nd_defrouter *dr, *next; 1453 1454 defrouter_reset(); 1455 TAILQ_FOREACH_SAFE(dr, &V_nd_defrouter, dr_entry, next) { 1456 defrtrlist_del(dr); 1457 } 1458 defrouter_select(); 1459 break; 1460 } 1461 case SIOCGNBRINFO_IN6: 1462 { 1463 struct llentry *ln; 1464 struct in6_addr nb_addr = nbi->addr; /* make local for safety */ 1465 1466 if ((error = in6_setscope(&nb_addr, ifp, NULL)) != 0) 1467 return (error); 1468 1469 IF_AFDATA_RLOCK(ifp); 1470 ln = nd6_lookup(&nb_addr, 0, ifp); 1471 IF_AFDATA_RUNLOCK(ifp); 1472 1473 if (ln == NULL) { 1474 error = EINVAL; 1475 break; 1476 } 1477 nbi->state = ln->ln_state; 1478 nbi->asked = ln->la_asked; 1479 nbi->isrouter = ln->ln_router; 1480 if (ln->la_expire == 0) 1481 nbi->expire = 0; 1482 else 1483 nbi->expire = ln->la_expire + 1484 (time_second - time_uptime); 1485 LLE_RUNLOCK(ln); 1486 break; 1487 } 1488 case SIOCGDEFIFACE_IN6: /* XXX: should be implemented as a sysctl? */ 1489 ndif->ifindex = V_nd6_defifindex; 1490 break; 1491 case SIOCSDEFIFACE_IN6: /* XXX: should be implemented as a sysctl? */ 1492 return (nd6_setdefaultiface(ndif->ifindex)); 1493 } 1494 return (error); 1495} 1496 1497/* 1498 * Create neighbor cache entry and cache link-layer address, 1499 * on reception of inbound ND6 packets. (RS/RA/NS/redirect) 1500 * 1501 * type - ICMP6 type 1502 * code - type dependent information 1503 * 1504 * XXXXX 1505 * The caller of this function already acquired the ndp 1506 * cache table lock because the cache entry is returned. 1507 */ 1508struct llentry * 1509nd6_cache_lladdr(struct ifnet *ifp, struct in6_addr *from, char *lladdr, 1510 int lladdrlen, int type, int code) 1511{ 1512 struct llentry *ln = NULL; 1513 int is_newentry; 1514 int do_update; 1515 int olladdr; 1516 int llchange; 1517 int flags; 1518 int newstate = 0; 1519 uint16_t router = 0; 1520 struct sockaddr_in6 sin6; 1521 struct mbuf *chain = NULL; 1522 int static_route = 0; 1523 1524 IF_AFDATA_UNLOCK_ASSERT(ifp); 1525 1526 KASSERT(ifp != NULL, ("%s: ifp == NULL", __func__)); 1527 KASSERT(from != NULL, ("%s: from == NULL", __func__)); 1528 1529 /* nothing must be updated for unspecified address */ 1530 if (IN6_IS_ADDR_UNSPECIFIED(from)) 1531 return NULL; 1532 1533 /* 1534 * Validation about ifp->if_addrlen and lladdrlen must be done in 1535 * the caller. 1536 * 1537 * XXX If the link does not have link-layer adderss, what should 1538 * we do? (ifp->if_addrlen == 0) 1539 * Spec says nothing in sections for RA, RS and NA. There's small 1540 * description on it in NS section (RFC 2461 7.2.3). 1541 */ 1542 flags = lladdr ? ND6_EXCLUSIVE : 0; 1543 IF_AFDATA_RLOCK(ifp); 1544 ln = nd6_lookup(from, flags, ifp); 1545 IF_AFDATA_RUNLOCK(ifp); 1546 if (ln == NULL) { 1547 flags |= ND6_EXCLUSIVE; 1548 IF_AFDATA_LOCK(ifp); 1549 ln = nd6_lookup(from, flags | ND6_CREATE, ifp); 1550 IF_AFDATA_UNLOCK(ifp); 1551 is_newentry = 1; 1552 } else { 1553 /* do nothing if static ndp is set */ 1554 if (ln->la_flags & LLE_STATIC) { 1555 static_route = 1; 1556 goto done; 1557 } 1558 is_newentry = 0; 1559 } 1560 if (ln == NULL) 1561 return (NULL); 1562 1563 olladdr = (ln->la_flags & LLE_VALID) ? 1 : 0; 1564 if (olladdr && lladdr) { 1565 llchange = bcmp(lladdr, &ln->ll_addr, 1566 ifp->if_addrlen); 1567 } else 1568 llchange = 0; 1569 1570 /* 1571 * newentry olladdr lladdr llchange (*=record) 1572 * 0 n n -- (1) 1573 * 0 y n -- (2) 1574 * 0 n y -- (3) * STALE 1575 * 0 y y n (4) * 1576 * 0 y y y (5) * STALE 1577 * 1 -- n -- (6) NOSTATE(= PASSIVE) 1578 * 1 -- y -- (7) * STALE 1579 */ 1580 1581 if (lladdr) { /* (3-5) and (7) */ 1582 /* 1583 * Record source link-layer address 1584 * XXX is it dependent to ifp->if_type? 1585 */ 1586 bcopy(lladdr, &ln->ll_addr, ifp->if_addrlen); 1587 ln->la_flags |= LLE_VALID; 1588 EVENTHANDLER_INVOKE(lle_event, ln, LLENTRY_RESOLVED); 1589 } 1590 1591 if (!is_newentry) { 1592 if ((!olladdr && lladdr != NULL) || /* (3) */ 1593 (olladdr && lladdr != NULL && llchange)) { /* (5) */ 1594 do_update = 1; 1595 newstate = ND6_LLINFO_STALE; 1596 } else /* (1-2,4) */ 1597 do_update = 0; 1598 } else { 1599 do_update = 1; 1600 if (lladdr == NULL) /* (6) */ 1601 newstate = ND6_LLINFO_NOSTATE; 1602 else /* (7) */ 1603 newstate = ND6_LLINFO_STALE; 1604 } 1605 1606 if (do_update) { 1607 /* 1608 * Update the state of the neighbor cache. 1609 */ 1610 ln->ln_state = newstate; 1611 1612 if (ln->ln_state == ND6_LLINFO_STALE) { 1613 if (ln->la_hold != NULL) 1614 nd6_grab_holdchain(ln, &chain, &sin6); 1615 } else if (ln->ln_state == ND6_LLINFO_INCOMPLETE) { 1616 /* probe right away */ 1617 nd6_llinfo_settimer_locked((void *)ln, 0); 1618 } 1619 } 1620 1621 /* 1622 * ICMP6 type dependent behavior. 1623 * 1624 * NS: clear IsRouter if new entry 1625 * RS: clear IsRouter 1626 * RA: set IsRouter if there's lladdr 1627 * redir: clear IsRouter if new entry 1628 * 1629 * RA case, (1): 1630 * The spec says that we must set IsRouter in the following cases: 1631 * - If lladdr exist, set IsRouter. This means (1-5). 1632 * - If it is old entry (!newentry), set IsRouter. This means (7). 1633 * So, based on the spec, in (1-5) and (7) cases we must set IsRouter. 1634 * A quetion arises for (1) case. (1) case has no lladdr in the 1635 * neighbor cache, this is similar to (6). 1636 * This case is rare but we figured that we MUST NOT set IsRouter. 1637 * 1638 * newentry olladdr lladdr llchange NS RS RA redir 1639 * D R 1640 * 0 n n -- (1) c ? s 1641 * 0 y n -- (2) c s s 1642 * 0 n y -- (3) c s s 1643 * 0 y y n (4) c s s 1644 * 0 y y y (5) c s s 1645 * 1 -- n -- (6) c c c s 1646 * 1 -- y -- (7) c c s c s 1647 * 1648 * (c=clear s=set) 1649 */ 1650 switch (type & 0xff) { 1651 case ND_NEIGHBOR_SOLICIT: 1652 /* 1653 * New entry must have is_router flag cleared. 1654 */ 1655 if (is_newentry) /* (6-7) */ 1656 ln->ln_router = 0; 1657 break; 1658 case ND_REDIRECT: 1659 /* 1660 * If the icmp is a redirect to a better router, always set the 1661 * is_router flag. Otherwise, if the entry is newly created, 1662 * clear the flag. [RFC 2461, sec 8.3] 1663 */ 1664 if (code == ND_REDIRECT_ROUTER) 1665 ln->ln_router = 1; 1666 else if (is_newentry) /* (6-7) */ 1667 ln->ln_router = 0; 1668 break; 1669 case ND_ROUTER_SOLICIT: 1670 /* 1671 * is_router flag must always be cleared. 1672 */ 1673 ln->ln_router = 0; 1674 break; 1675 case ND_ROUTER_ADVERT: 1676 /* 1677 * Mark an entry with lladdr as a router. 1678 */ 1679 if ((!is_newentry && (olladdr || lladdr)) || /* (2-5) */ 1680 (is_newentry && lladdr)) { /* (7) */ 1681 ln->ln_router = 1; 1682 } 1683 break; 1684 } 1685 1686 if (ln != NULL) { 1687 static_route = (ln->la_flags & LLE_STATIC); 1688 router = ln->ln_router; 1689 1690 if (flags & ND6_EXCLUSIVE) 1691 LLE_WUNLOCK(ln); 1692 else 1693 LLE_RUNLOCK(ln); 1694 if (static_route) 1695 ln = NULL; 1696 } 1697 if (chain != NULL) 1698 nd6_flush_holdchain(ifp, ifp, chain, &sin6); 1699 1700 /* 1701 * When the link-layer address of a router changes, select the 1702 * best router again. In particular, when the neighbor entry is newly 1703 * created, it might affect the selection policy. 1704 * Question: can we restrict the first condition to the "is_newentry" 1705 * case? 1706 * XXX: when we hear an RA from a new router with the link-layer 1707 * address option, defrouter_select() is called twice, since 1708 * defrtrlist_update called the function as well. However, I believe 1709 * we can compromise the overhead, since it only happens the first 1710 * time. 1711 * XXX: although defrouter_select() should not have a bad effect 1712 * for those are not autoconfigured hosts, we explicitly avoid such 1713 * cases for safety. 1714 */ 1715 if (do_update && router && 1716 ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV) { 1717 /* 1718 * guaranteed recursion 1719 */ 1720 defrouter_select(); 1721 } 1722 1723 return (ln); 1724done: 1725 if (ln != NULL) { 1726 if (flags & ND6_EXCLUSIVE) 1727 LLE_WUNLOCK(ln); 1728 else 1729 LLE_RUNLOCK(ln); 1730 if (static_route) 1731 ln = NULL; 1732 } 1733 return (ln); 1734} 1735 1736static void 1737nd6_slowtimo(void *arg) 1738{ 1739 CURVNET_SET((struct vnet *) arg); 1740 struct nd_ifinfo *nd6if; 1741 struct ifnet *ifp; 1742 1743 callout_reset(&V_nd6_slowtimo_ch, ND6_SLOWTIMER_INTERVAL * hz, 1744 nd6_slowtimo, curvnet); 1745 IFNET_RLOCK_NOSLEEP(); 1746 TAILQ_FOREACH(ifp, &V_ifnet, if_list) { 1747 if (ifp->if_afdata[AF_INET6] == NULL) 1748 continue; 1749 nd6if = ND_IFINFO(ifp); 1750 if (nd6if->basereachable && /* already initialized */ 1751 (nd6if->recalctm -= ND6_SLOWTIMER_INTERVAL) <= 0) { 1752 /* 1753 * Since reachable time rarely changes by router 1754 * advertisements, we SHOULD insure that a new random 1755 * value gets recomputed at least once every few hours. 1756 * (RFC 2461, 6.3.4) 1757 */ 1758 nd6if->recalctm = V_nd6_recalc_reachtm_interval; 1759 nd6if->reachable = ND_COMPUTE_RTIME(nd6if->basereachable); 1760 } 1761 } 1762 IFNET_RUNLOCK_NOSLEEP(); 1763 CURVNET_RESTORE(); 1764} 1765 1766void 1767nd6_grab_holdchain(struct llentry *ln, struct mbuf **chain, 1768 struct sockaddr_in6 *sin6) 1769{ 1770 1771 LLE_WLOCK_ASSERT(ln); 1772 1773 *chain = ln->la_hold; 1774 ln->la_hold = NULL; 1775 memcpy(sin6, L3_ADDR_SIN6(ln), sizeof(*sin6)); 1776 1777 if (ln->ln_state == ND6_LLINFO_STALE) { 1778 1779 /* 1780 * The first time we send a packet to a 1781 * neighbor whose entry is STALE, we have 1782 * to change the state to DELAY and a sets 1783 * a timer to expire in DELAY_FIRST_PROBE_TIME 1784 * seconds to ensure do neighbor unreachability 1785 * detection on expiration. 1786 * (RFC 2461 7.3.3) 1787 */ 1788 ln->la_asked = 0; 1789 ln->ln_state = ND6_LLINFO_DELAY; 1790 nd6_llinfo_settimer_locked(ln, (long)V_nd6_delay * hz); 1791 } 1792} 1793 1794static int 1795nd6_output_ifp(struct ifnet *ifp, struct ifnet *origifp, struct mbuf *m, 1796 struct sockaddr_in6 *dst) 1797{ 1798 int error; 1799 int ip6len; 1800 struct ip6_hdr *ip6; 1801 struct m_tag *mtag; 1802 1803#ifdef MAC 1804 mac_netinet6_nd6_send(ifp, m); 1805#endif 1806 1807 /* 1808 * If called from nd6_ns_output() (NS), nd6_na_output() (NA), 1809 * icmp6_redirect_output() (REDIRECT) or from rip6_output() (RS, RA 1810 * as handled by rtsol and rtadvd), mbufs will be tagged for SeND 1811 * to be diverted to user space. When re-injected into the kernel, 1812 * send_output() will directly dispatch them to the outgoing interface. 1813 */ 1814 if (send_sendso_input_hook != NULL) { 1815 mtag = m_tag_find(m, PACKET_TAG_ND_OUTGOING, NULL); 1816 if (mtag != NULL) { 1817 ip6 = mtod(m, struct ip6_hdr *); 1818 ip6len = sizeof(struct ip6_hdr) + ntohs(ip6->ip6_plen); 1819 /* Use the SEND socket */ 1820 error = send_sendso_input_hook(m, ifp, SND_OUT, 1821 ip6len); 1822 /* -1 == no app on SEND socket */ 1823 if (error == 0 || error != -1) 1824 return (error); 1825 } 1826 } 1827 1828 m_clrprotoflags(m); /* Avoid confusing lower layers. */ 1829 IP_PROBE(send, NULL, NULL, mtod(m, struct ip6_hdr *), ifp, NULL, 1830 mtod(m, struct ip6_hdr *)); 1831 1832 if ((ifp->if_flags & IFF_LOOPBACK) == 0) 1833 origifp = ifp; 1834 1835 error = (*ifp->if_output)(origifp, m, (struct sockaddr *)dst, NULL); 1836 return (error); 1837} 1838 1839/* 1840 * IPv6 packet output - light version. 1841 * Checks if destination LLE exists and is in proper state 1842 * (e.g no modification required). If not true, fall back to 1843 * "heavy" version. 1844 */ 1845int 1846nd6_output(struct ifnet *ifp, struct ifnet *origifp, struct mbuf *m, 1847 struct sockaddr_in6 *dst, struct rtentry *rt0) 1848{ 1849 struct llentry *ln = NULL; 1850 1851 /* discard the packet if IPv6 operation is disabled on the interface */ 1852 if ((ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED)) { 1853 m_freem(m); 1854 return (ENETDOWN); /* better error? */ 1855 } 1856 1857 if (IN6_IS_ADDR_MULTICAST(&dst->sin6_addr)) 1858 goto sendpkt; 1859 1860 if (nd6_need_cache(ifp) == 0) 1861 goto sendpkt; 1862 1863 IF_AFDATA_RLOCK(ifp); 1864 ln = nd6_lookup(&dst->sin6_addr, 0, ifp); 1865 IF_AFDATA_RUNLOCK(ifp); 1866 1867 /* 1868 * Perform fast path for the following cases: 1869 * 1) lle state is REACHABLE 1870 * 2) lle state is DELAY (NS message sentNS message sent) 1871 * 1872 * Every other case involves lle modification, so we handle 1873 * them separately. 1874 */ 1875 if (ln == NULL || (ln->ln_state != ND6_LLINFO_REACHABLE && 1876 ln->ln_state != ND6_LLINFO_DELAY)) { 1877 /* Fall back to slow processing path */ 1878 if (ln != NULL) 1879 LLE_RUNLOCK(ln); 1880 return (nd6_output_lle(ifp, origifp, m, dst)); 1881 } 1882 1883sendpkt: 1884 if (ln != NULL) 1885 LLE_RUNLOCK(ln); 1886 1887 return (nd6_output_ifp(ifp, origifp, m, dst)); 1888} 1889 1890 1891/* 1892 * Output IPv6 packet - heavy version. 1893 * Function assume that either 1894 * 1) destination LLE does not exist, is invalid or stale, so 1895 * ND6_EXCLUSIVE lock needs to be acquired 1896 * 2) destination lle is provided (with ND6_EXCLUSIVE lock), 1897 * in that case packets are queued in &chain. 1898 * 1899 */ 1900static int 1901nd6_output_lle(struct ifnet *ifp, struct ifnet *origifp, struct mbuf *m, 1902 struct sockaddr_in6 *dst) 1903{ 1904 struct llentry *lle = NULL; 1905 int flags = 0; 1906 1907 KASSERT(m != NULL, ("NULL mbuf, nothing to send")); 1908 /* discard the packet if IPv6 operation is disabled on the interface */ 1909 if ((ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED)) { 1910 m_freem(m); 1911 return (ENETDOWN); /* better error? */ 1912 } 1913 1914 if (IN6_IS_ADDR_MULTICAST(&dst->sin6_addr)) 1915 goto sendpkt; 1916 1917 if (nd6_need_cache(ifp) == 0) 1918 goto sendpkt; 1919 1920 /* 1921 * Address resolution or Neighbor Unreachability Detection 1922 * for the next hop. 1923 * At this point, the destination of the packet must be a unicast 1924 * or an anycast address(i.e. not a multicast). 1925 */ 1926 if (lle == NULL) { 1927 IF_AFDATA_RLOCK(ifp); 1928 lle = nd6_lookup(&dst->sin6_addr, ND6_EXCLUSIVE, ifp); 1929 IF_AFDATA_RUNLOCK(ifp); 1930 if ((lle == NULL) && nd6_is_addr_neighbor(dst, ifp)) { 1931 /* 1932 * Since nd6_is_addr_neighbor() internally calls nd6_lookup(), 1933 * the condition below is not very efficient. But we believe 1934 * it is tolerable, because this should be a rare case. 1935 */ 1936 flags = ND6_CREATE | ND6_EXCLUSIVE; 1937 IF_AFDATA_LOCK(ifp); 1938 lle = nd6_lookup(&dst->sin6_addr, flags, ifp); 1939 IF_AFDATA_UNLOCK(ifp); 1940 } 1941 } 1942 if (lle == NULL) { 1943 if ((ifp->if_flags & IFF_POINTOPOINT) == 0 && 1944 !(ND_IFINFO(ifp)->flags & ND6_IFF_PERFORMNUD)) { 1945 char ip6buf[INET6_ADDRSTRLEN]; 1946 log(LOG_DEBUG, 1947 "nd6_output: can't allocate llinfo for %s " 1948 "(ln=%p)\n", 1949 ip6_sprintf(ip6buf, &dst->sin6_addr), lle); 1950 m_freem(m); 1951 return (ENOBUFS); 1952 } 1953 goto sendpkt; /* send anyway */ 1954 } 1955 1956 LLE_WLOCK_ASSERT(lle); 1957 1958 /* We don't have to do link-layer address resolution on a p2p link. */ 1959 if ((ifp->if_flags & IFF_POINTOPOINT) != 0 && 1960 lle->ln_state < ND6_LLINFO_REACHABLE) { 1961 lle->ln_state = ND6_LLINFO_STALE; 1962 nd6_llinfo_settimer_locked(lle, (long)V_nd6_gctimer * hz); 1963 } 1964 1965 /* 1966 * The first time we send a packet to a neighbor whose entry is 1967 * STALE, we have to change the state to DELAY and a sets a timer to 1968 * expire in DELAY_FIRST_PROBE_TIME seconds to ensure do 1969 * neighbor unreachability detection on expiration. 1970 * (RFC 2461 7.3.3) 1971 */ 1972 if (lle->ln_state == ND6_LLINFO_STALE) { 1973 lle->la_asked = 0; 1974 lle->ln_state = ND6_LLINFO_DELAY; 1975 nd6_llinfo_settimer_locked(lle, (long)V_nd6_delay * hz); 1976 } 1977 1978 /* 1979 * If the neighbor cache entry has a state other than INCOMPLETE 1980 * (i.e. its link-layer address is already resolved), just 1981 * send the packet. 1982 */ 1983 if (lle->ln_state > ND6_LLINFO_INCOMPLETE) 1984 goto sendpkt; 1985 1986 /* 1987 * There is a neighbor cache entry, but no ethernet address 1988 * response yet. Append this latest packet to the end of the 1989 * packet queue in the mbuf, unless the number of the packet 1990 * does not exceed nd6_maxqueuelen. When it exceeds nd6_maxqueuelen, 1991 * the oldest packet in the queue will be removed. 1992 */ 1993 if (lle->ln_state == ND6_LLINFO_NOSTATE) 1994 lle->ln_state = ND6_LLINFO_INCOMPLETE; 1995 1996 if (lle->la_hold != NULL) { 1997 struct mbuf *m_hold; 1998 int i; 1999 2000 i = 0; 2001 for (m_hold = lle->la_hold; m_hold; m_hold = m_hold->m_nextpkt){ 2002 i++; 2003 if (m_hold->m_nextpkt == NULL) { 2004 m_hold->m_nextpkt = m; 2005 break; 2006 } 2007 } 2008 while (i >= V_nd6_maxqueuelen) { 2009 m_hold = lle->la_hold; 2010 lle->la_hold = lle->la_hold->m_nextpkt; 2011 m_freem(m_hold); 2012 i--; 2013 } 2014 } else { 2015 lle->la_hold = m; 2016 } 2017 2018 /* 2019 * If there has been no NS for the neighbor after entering the 2020 * INCOMPLETE state, send the first solicitation. 2021 */ 2022 if (!ND6_LLINFO_PERMANENT(lle) && lle->la_asked == 0) { 2023 lle->la_asked++; 2024 2025 nd6_llinfo_settimer_locked(lle, 2026 (long)ND_IFINFO(ifp)->retrans * hz / 1000); 2027 LLE_WUNLOCK(lle); 2028 nd6_ns_output(ifp, NULL, &dst->sin6_addr, lle, NULL); 2029 } else { 2030 /* We did the lookup so we need to do the unlock here. */ 2031 LLE_WUNLOCK(lle); 2032 } 2033 2034 return (0); 2035 2036 sendpkt: 2037 if (lle != NULL) 2038 LLE_WUNLOCK(lle); 2039 2040 return (nd6_output_ifp(ifp, origifp, m, dst)); 2041} 2042 2043 2044int 2045nd6_flush_holdchain(struct ifnet *ifp, struct ifnet *origifp, struct mbuf *chain, 2046 struct sockaddr_in6 *dst) 2047{ 2048 struct mbuf *m, *m_head; 2049 struct ifnet *outifp; 2050 int error = 0; 2051 2052 m_head = chain; 2053 if ((ifp->if_flags & IFF_LOOPBACK) != 0) 2054 outifp = origifp; 2055 else 2056 outifp = ifp; 2057 2058 while (m_head) { 2059 m = m_head; 2060 m_head = m_head->m_nextpkt; 2061 error = nd6_output_ifp(ifp, origifp, m, dst); 2062 } 2063 2064 /* 2065 * XXX 2066 * note that intermediate errors are blindly ignored - but this is 2067 * the same convention as used with nd6_output when called by 2068 * nd6_cache_lladdr 2069 */ 2070 return (error); 2071} 2072 2073 2074int 2075nd6_need_cache(struct ifnet *ifp) 2076{ 2077 /* 2078 * XXX: we currently do not make neighbor cache on any interface 2079 * other than ARCnet, Ethernet, FDDI and GIF. 2080 * 2081 * RFC2893 says: 2082 * - unidirectional tunnels needs no ND 2083 */ 2084 switch (ifp->if_type) { 2085 case IFT_ARCNET: 2086 case IFT_ETHER: 2087 case IFT_FDDI: 2088 case IFT_IEEE1394: 2089#ifdef IFT_L2VLAN 2090 case IFT_L2VLAN: 2091#endif 2092#ifdef IFT_IEEE80211 2093 case IFT_IEEE80211: 2094#endif 2095 case IFT_INFINIBAND: 2096 case IFT_BRIDGE: 2097 case IFT_PROPVIRTUAL: 2098 return (1); 2099 default: 2100 return (0); 2101 } 2102} 2103 2104/* 2105 * the callers of this function need to be re-worked to drop 2106 * the lle lock, drop here for now 2107 */ 2108int 2109nd6_storelladdr(struct ifnet *ifp, struct mbuf *m, 2110 const struct sockaddr *dst, u_char *desten, struct llentry **lle) 2111{ 2112 struct llentry *ln; 2113 2114 *lle = NULL; 2115 IF_AFDATA_UNLOCK_ASSERT(ifp); 2116 if (m != NULL && m->m_flags & M_MCAST) { 2117 int i; 2118 2119 switch (ifp->if_type) { 2120 case IFT_ETHER: 2121 case IFT_FDDI: 2122#ifdef IFT_L2VLAN 2123 case IFT_L2VLAN: 2124#endif 2125#ifdef IFT_IEEE80211 2126 case IFT_IEEE80211: 2127#endif 2128 case IFT_BRIDGE: 2129 case IFT_ISO88025: 2130 ETHER_MAP_IPV6_MULTICAST(&SIN6(dst)->sin6_addr, 2131 desten); 2132 return (0); 2133 case IFT_IEEE1394: 2134 /* 2135 * netbsd can use if_broadcastaddr, but we don't do so 2136 * to reduce # of ifdef. 2137 */ 2138 for (i = 0; i < ifp->if_addrlen; i++) 2139 desten[i] = ~0; 2140 return (0); 2141 case IFT_ARCNET: 2142 *desten = 0; 2143 return (0); 2144 default: 2145 m_freem(m); 2146 return (EAFNOSUPPORT); 2147 } 2148 } 2149 2150 2151 /* 2152 * the entry should have been created in nd6_store_lladdr 2153 */ 2154 IF_AFDATA_RLOCK(ifp); 2155 ln = lla_lookup(LLTABLE6(ifp), 0, dst); 2156 IF_AFDATA_RUNLOCK(ifp); 2157 if ((ln == NULL) || !(ln->la_flags & LLE_VALID)) { 2158 if (ln != NULL) 2159 LLE_RUNLOCK(ln); 2160 /* this could happen, if we could not allocate memory */ 2161 m_freem(m); 2162 return (1); 2163 } 2164 2165 bcopy(&ln->ll_addr, desten, ifp->if_addrlen); 2166 *lle = ln; 2167 LLE_RUNLOCK(ln); 2168 /* 2169 * A *small* use after free race exists here 2170 */ 2171 return (0); 2172} 2173 2174static void 2175clear_llinfo_pqueue(struct llentry *ln) 2176{ 2177 struct mbuf *m_hold, *m_hold_next; 2178 2179 for (m_hold = ln->la_hold; m_hold; m_hold = m_hold_next) { 2180 m_hold_next = m_hold->m_nextpkt; 2181 m_freem(m_hold); 2182 } 2183 2184 ln->la_hold = NULL; 2185 return; 2186} 2187 2188static int nd6_sysctl_drlist(SYSCTL_HANDLER_ARGS); 2189static int nd6_sysctl_prlist(SYSCTL_HANDLER_ARGS); 2190#ifdef SYSCTL_DECL 2191SYSCTL_DECL(_net_inet6_icmp6); 2192#endif 2193SYSCTL_NODE(_net_inet6_icmp6, ICMPV6CTL_ND6_DRLIST, nd6_drlist, 2194 CTLFLAG_RD, nd6_sysctl_drlist, ""); 2195SYSCTL_NODE(_net_inet6_icmp6, ICMPV6CTL_ND6_PRLIST, nd6_prlist, 2196 CTLFLAG_RD, nd6_sysctl_prlist, ""); 2197SYSCTL_VNET_INT(_net_inet6_icmp6, ICMPV6CTL_ND6_MAXQLEN, nd6_maxqueuelen, 2198 CTLFLAG_RW, &VNET_NAME(nd6_maxqueuelen), 1, ""); 2199SYSCTL_VNET_INT(_net_inet6_icmp6, OID_AUTO, nd6_gctimer, 2200 CTLFLAG_RW, &VNET_NAME(nd6_gctimer), (60 * 60 * 24), ""); 2201 2202static int 2203nd6_sysctl_drlist(SYSCTL_HANDLER_ARGS) 2204{ 2205 struct in6_defrouter d; 2206 struct nd_defrouter *dr; 2207 int error; 2208 2209 if (req->newptr) 2210 return (EPERM); 2211 2212 bzero(&d, sizeof(d)); 2213 d.rtaddr.sin6_family = AF_INET6; 2214 d.rtaddr.sin6_len = sizeof(d.rtaddr); 2215 2216 /* 2217 * XXX locking 2218 */ 2219 TAILQ_FOREACH(dr, &V_nd_defrouter, dr_entry) { 2220 d.rtaddr.sin6_addr = dr->rtaddr; 2221 error = sa6_recoverscope(&d.rtaddr); 2222 if (error != 0) 2223 return (error); 2224 d.flags = dr->flags; 2225 d.rtlifetime = dr->rtlifetime; 2226 d.expire = dr->expire + (time_second - time_uptime); 2227 d.if_index = dr->ifp->if_index; 2228 error = SYSCTL_OUT(req, &d, sizeof(d)); 2229 if (error != 0) 2230 return (error); 2231 } 2232 return (0); 2233} 2234 2235static int 2236nd6_sysctl_prlist(SYSCTL_HANDLER_ARGS) 2237{ 2238 struct in6_prefix p; 2239 struct sockaddr_in6 s6; 2240 struct nd_prefix *pr; 2241 struct nd_pfxrouter *pfr; 2242 time_t maxexpire; 2243 int error; 2244 char ip6buf[INET6_ADDRSTRLEN]; 2245 2246 if (req->newptr) 2247 return (EPERM); 2248 2249 bzero(&p, sizeof(p)); 2250 p.origin = PR_ORIG_RA; 2251 bzero(&s6, sizeof(s6)); 2252 s6.sin6_family = AF_INET6; 2253 s6.sin6_len = sizeof(s6); 2254 2255 /* 2256 * XXX locking 2257 */ 2258 LIST_FOREACH(pr, &V_nd_prefix, ndpr_entry) { 2259 p.prefix = pr->ndpr_prefix; 2260 if (sa6_recoverscope(&p.prefix)) { 2261 log(LOG_ERR, "scope error in prefix list (%s)\n", 2262 ip6_sprintf(ip6buf, &p.prefix.sin6_addr)); 2263 /* XXX: press on... */ 2264 } 2265 p.raflags = pr->ndpr_raf; 2266 p.prefixlen = pr->ndpr_plen; 2267 p.vltime = pr->ndpr_vltime; 2268 p.pltime = pr->ndpr_pltime; 2269 p.if_index = pr->ndpr_ifp->if_index; 2270 if (pr->ndpr_vltime == ND6_INFINITE_LIFETIME) 2271 p.expire = 0; 2272 else { 2273 /* XXX: we assume time_t is signed. */ 2274 maxexpire = (-1) & 2275 ~((time_t)1 << ((sizeof(maxexpire) * 8) - 1)); 2276 if (pr->ndpr_vltime < maxexpire - pr->ndpr_lastupdate) 2277 p.expire = pr->ndpr_lastupdate + 2278 pr->ndpr_vltime + 2279 (time_second - time_uptime); 2280 else 2281 p.expire = maxexpire; 2282 } 2283 p.refcnt = pr->ndpr_refcnt; 2284 p.flags = pr->ndpr_stateflags; 2285 p.advrtrs = 0; 2286 LIST_FOREACH(pfr, &pr->ndpr_advrtrs, pfr_entry) 2287 p.advrtrs++; 2288 error = SYSCTL_OUT(req, &p, sizeof(p)); 2289 if (error != 0) 2290 return (error); 2291 LIST_FOREACH(pfr, &pr->ndpr_advrtrs, pfr_entry) { 2292 s6.sin6_addr = pfr->router->rtaddr; 2293 if (sa6_recoverscope(&s6)) 2294 log(LOG_ERR, 2295 "scope error in prefix list (%s)\n", 2296 ip6_sprintf(ip6buf, &pfr->router->rtaddr)); 2297 error = SYSCTL_OUT(req, &s6, sizeof(s6)); 2298 if (error != 0) 2299 return (error); 2300 } 2301 } 2302 return (0); 2303} 2304