nd6.c revision 299145
1/*- 2 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project. 3 * All rights reserved. 4 * 5 * Redistribution and use in source and binary forms, with or without 6 * modification, are permitted provided that the following conditions 7 * are met: 8 * 1. Redistributions of source code must retain the above copyright 9 * notice, this list of conditions and the following disclaimer. 10 * 2. Redistributions in binary form must reproduce the above copyright 11 * notice, this list of conditions and the following disclaimer in the 12 * documentation and/or other materials provided with the distribution. 13 * 3. Neither the name of the project nor the names of its contributors 14 * may be used to endorse or promote products derived from this software 15 * without specific prior written permission. 16 * 17 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND 18 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 19 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 20 * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE 21 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 22 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 23 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 24 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 25 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 26 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 27 * SUCH DAMAGE. 28 * 29 * $KAME: nd6.c,v 1.144 2001/05/24 07:44:00 itojun Exp $ 30 */ 31 32#include <sys/cdefs.h> 33__FBSDID("$FreeBSD: stable/10/sys/netinet6/nd6.c 299145 2016-05-05 23:06:39Z markj $"); 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 115VNET_DEFINE(struct nd_drhead, nd_defrouter); 116VNET_DEFINE(struct nd_prhead, nd_prefix); 117 118VNET_DEFINE(int, nd6_recalc_reachtm_interval) = ND6_RECALC_REACHTM_INTERVAL; 119#define V_nd6_recalc_reachtm_interval VNET(nd6_recalc_reachtm_interval) 120 121int (*send_sendso_input_hook)(struct mbuf *, struct ifnet *, int, int); 122 123static int nd6_is_new_addr_neighbor(struct sockaddr_in6 *, 124 struct ifnet *); 125static void nd6_setmtu0(struct ifnet *, struct nd_ifinfo *); 126static void nd6_slowtimo(void *); 127static int regen_tmpaddr(struct in6_ifaddr *); 128static struct llentry *nd6_free(struct llentry *, int); 129static void nd6_llinfo_timer(void *); 130static void clear_llinfo_pqueue(struct llentry *); 131static int nd6_output_lle(struct ifnet *, struct ifnet *, struct mbuf *, 132 struct sockaddr_in6 *); 133static int nd6_output_ifp(struct ifnet *, struct ifnet *, struct mbuf *, 134 struct sockaddr_in6 *); 135 136static VNET_DEFINE(struct callout, nd6_slowtimo_ch); 137#define V_nd6_slowtimo_ch VNET(nd6_slowtimo_ch) 138 139VNET_DEFINE(struct callout, nd6_timer_ch); 140 141void 142nd6_init(void) 143{ 144 145 LIST_INIT(&V_nd_prefix); 146 147 /* initialization of the default router list */ 148 TAILQ_INIT(&V_nd_defrouter); 149 150 /* start timer */ 151 callout_init(&V_nd6_slowtimo_ch, 0); 152 callout_reset(&V_nd6_slowtimo_ch, ND6_SLOWTIMER_INTERVAL * hz, 153 nd6_slowtimo, curvnet); 154 155 nd6_dad_init(); 156} 157 158#ifdef VIMAGE 159void 160nd6_destroy() 161{ 162 163 callout_drain(&V_nd6_slowtimo_ch); 164 callout_drain(&V_nd6_timer_ch); 165} 166#endif 167 168struct nd_ifinfo * 169nd6_ifattach(struct ifnet *ifp) 170{ 171 struct nd_ifinfo *nd; 172 173 nd = malloc(sizeof(*nd), M_IP6NDP, M_WAITOK | M_ZERO); 174 nd->initialized = 1; 175 176 nd->chlim = IPV6_DEFHLIM; 177 nd->basereachable = REACHABLE_TIME; 178 nd->reachable = ND_COMPUTE_RTIME(nd->basereachable); 179 nd->retrans = RETRANS_TIMER; 180 181 nd->flags = ND6_IFF_PERFORMNUD; 182 183 /* A loopback interface always has ND6_IFF_AUTO_LINKLOCAL. 184 * XXXHRS: Clear ND6_IFF_AUTO_LINKLOCAL on an IFT_BRIDGE interface by 185 * default regardless of the V_ip6_auto_linklocal configuration to 186 * give a reasonable default behavior. 187 */ 188 if ((V_ip6_auto_linklocal && ifp->if_type != IFT_BRIDGE) || 189 (ifp->if_flags & IFF_LOOPBACK)) 190 nd->flags |= ND6_IFF_AUTO_LINKLOCAL; 191 /* 192 * A loopback interface does not need to accept RTADV. 193 * XXXHRS: Clear ND6_IFF_ACCEPT_RTADV on an IFT_BRIDGE interface by 194 * default regardless of the V_ip6_accept_rtadv configuration to 195 * prevent the interface from accepting RA messages arrived 196 * on one of the member interfaces with ND6_IFF_ACCEPT_RTADV. 197 */ 198 if (V_ip6_accept_rtadv && 199 !(ifp->if_flags & IFF_LOOPBACK) && 200 (ifp->if_type != IFT_BRIDGE)) 201 nd->flags |= ND6_IFF_ACCEPT_RTADV; 202 if (V_ip6_no_radr && !(ifp->if_flags & IFF_LOOPBACK)) 203 nd->flags |= ND6_IFF_NO_RADR; 204 205 /* XXX: we cannot call nd6_setmtu since ifp is not fully initialized */ 206 nd6_setmtu0(ifp, nd); 207 208 return nd; 209} 210 211void 212nd6_ifdetach(struct nd_ifinfo *nd) 213{ 214 215 free(nd, M_IP6NDP); 216} 217 218/* 219 * Reset ND level link MTU. This function is called when the physical MTU 220 * changes, which means we might have to adjust the ND level MTU. 221 */ 222void 223nd6_setmtu(struct ifnet *ifp) 224{ 225 if (ifp->if_afdata[AF_INET6] == NULL) 226 return; 227 228 nd6_setmtu0(ifp, ND_IFINFO(ifp)); 229} 230 231/* XXX todo: do not maintain copy of ifp->if_mtu in ndi->maxmtu */ 232void 233nd6_setmtu0(struct ifnet *ifp, struct nd_ifinfo *ndi) 234{ 235 u_int32_t omaxmtu; 236 237 omaxmtu = ndi->maxmtu; 238 239 switch (ifp->if_type) { 240 case IFT_ARCNET: 241 ndi->maxmtu = MIN(ARC_PHDS_MAXMTU, ifp->if_mtu); /* RFC2497 */ 242 break; 243 case IFT_FDDI: 244 ndi->maxmtu = MIN(FDDIIPMTU, ifp->if_mtu); /* RFC2467 */ 245 break; 246 case IFT_ISO88025: 247 ndi->maxmtu = MIN(ISO88025_MAX_MTU, ifp->if_mtu); 248 break; 249 default: 250 ndi->maxmtu = ifp->if_mtu; 251 break; 252 } 253 254 /* 255 * Decreasing the interface MTU under IPV6 minimum MTU may cause 256 * undesirable situation. We thus notify the operator of the change 257 * explicitly. The check for omaxmtu is necessary to restrict the 258 * log to the case of changing the MTU, not initializing it. 259 */ 260 if (omaxmtu >= IPV6_MMTU && ndi->maxmtu < IPV6_MMTU) { 261 log(LOG_NOTICE, "nd6_setmtu0: " 262 "new link MTU on %s (%lu) is too small for IPv6\n", 263 if_name(ifp), (unsigned long)ndi->maxmtu); 264 } 265 266 if (ndi->maxmtu > V_in6_maxmtu) 267 in6_setmaxmtu(); /* check all interfaces just in case */ 268 269} 270 271void 272nd6_option_init(void *opt, int icmp6len, union nd_opts *ndopts) 273{ 274 275 bzero(ndopts, sizeof(*ndopts)); 276 ndopts->nd_opts_search = (struct nd_opt_hdr *)opt; 277 ndopts->nd_opts_last 278 = (struct nd_opt_hdr *)(((u_char *)opt) + icmp6len); 279 280 if (icmp6len == 0) { 281 ndopts->nd_opts_done = 1; 282 ndopts->nd_opts_search = NULL; 283 } 284} 285 286/* 287 * Take one ND option. 288 */ 289struct nd_opt_hdr * 290nd6_option(union nd_opts *ndopts) 291{ 292 struct nd_opt_hdr *nd_opt; 293 int olen; 294 295 KASSERT(ndopts != NULL, ("%s: ndopts == NULL", __func__)); 296 KASSERT(ndopts->nd_opts_last != NULL, ("%s: uninitialized ndopts", 297 __func__)); 298 if (ndopts->nd_opts_search == NULL) 299 return NULL; 300 if (ndopts->nd_opts_done) 301 return NULL; 302 303 nd_opt = ndopts->nd_opts_search; 304 305 /* make sure nd_opt_len is inside the buffer */ 306 if ((caddr_t)&nd_opt->nd_opt_len >= (caddr_t)ndopts->nd_opts_last) { 307 bzero(ndopts, sizeof(*ndopts)); 308 return NULL; 309 } 310 311 olen = nd_opt->nd_opt_len << 3; 312 if (olen == 0) { 313 /* 314 * Message validation requires that all included 315 * options have a length that is greater than zero. 316 */ 317 bzero(ndopts, sizeof(*ndopts)); 318 return NULL; 319 } 320 321 ndopts->nd_opts_search = (struct nd_opt_hdr *)((caddr_t)nd_opt + olen); 322 if (ndopts->nd_opts_search > ndopts->nd_opts_last) { 323 /* option overruns the end of buffer, invalid */ 324 bzero(ndopts, sizeof(*ndopts)); 325 return NULL; 326 } else if (ndopts->nd_opts_search == ndopts->nd_opts_last) { 327 /* reached the end of options chain */ 328 ndopts->nd_opts_done = 1; 329 ndopts->nd_opts_search = NULL; 330 } 331 return nd_opt; 332} 333 334/* 335 * Parse multiple ND options. 336 * This function is much easier to use, for ND routines that do not need 337 * multiple options of the same type. 338 */ 339int 340nd6_options(union nd_opts *ndopts) 341{ 342 struct nd_opt_hdr *nd_opt; 343 int i = 0; 344 345 KASSERT(ndopts != NULL, ("%s: ndopts == NULL", __func__)); 346 KASSERT(ndopts->nd_opts_last != NULL, ("%s: uninitialized ndopts", 347 __func__)); 348 if (ndopts->nd_opts_search == NULL) 349 return 0; 350 351 while (1) { 352 nd_opt = nd6_option(ndopts); 353 if (nd_opt == NULL && ndopts->nd_opts_last == NULL) { 354 /* 355 * Message validation requires that all included 356 * options have a length that is greater than zero. 357 */ 358 ICMP6STAT_INC(icp6s_nd_badopt); 359 bzero(ndopts, sizeof(*ndopts)); 360 return -1; 361 } 362 363 if (nd_opt == NULL) 364 goto skip1; 365 366 switch (nd_opt->nd_opt_type) { 367 case ND_OPT_SOURCE_LINKADDR: 368 case ND_OPT_TARGET_LINKADDR: 369 case ND_OPT_MTU: 370 case ND_OPT_REDIRECTED_HEADER: 371 case ND_OPT_NONCE: 372 if (ndopts->nd_opt_array[nd_opt->nd_opt_type]) { 373 nd6log((LOG_INFO, 374 "duplicated ND6 option found (type=%d)\n", 375 nd_opt->nd_opt_type)); 376 /* XXX bark? */ 377 } else { 378 ndopts->nd_opt_array[nd_opt->nd_opt_type] 379 = nd_opt; 380 } 381 break; 382 case ND_OPT_PREFIX_INFORMATION: 383 if (ndopts->nd_opt_array[nd_opt->nd_opt_type] == 0) { 384 ndopts->nd_opt_array[nd_opt->nd_opt_type] 385 = nd_opt; 386 } 387 ndopts->nd_opts_pi_end = 388 (struct nd_opt_prefix_info *)nd_opt; 389 break; 390 /* What about ND_OPT_ROUTE_INFO? RFC 4191 */ 391 case ND_OPT_RDNSS: /* RFC 6106 */ 392 case ND_OPT_DNSSL: /* RFC 6106 */ 393 /* 394 * Silently ignore options we know and do not care about 395 * in the kernel. 396 */ 397 break; 398 default: 399 /* 400 * Unknown options must be silently ignored, 401 * to accomodate future extension to the protocol. 402 */ 403 nd6log((LOG_DEBUG, 404 "nd6_options: unsupported option %d - " 405 "option ignored\n", nd_opt->nd_opt_type)); 406 } 407 408skip1: 409 i++; 410 if (i > V_nd6_maxndopt) { 411 ICMP6STAT_INC(icp6s_nd_toomanyopt); 412 nd6log((LOG_INFO, "too many loop in nd opt\n")); 413 break; 414 } 415 416 if (ndopts->nd_opts_done) 417 break; 418 } 419 420 return 0; 421} 422 423/* 424 * ND6 timer routine to handle ND6 entries 425 */ 426void 427nd6_llinfo_settimer_locked(struct llentry *ln, long tick) 428{ 429 int canceled; 430 431 LLE_WLOCK_ASSERT(ln); 432 433 if (tick < 0) { 434 ln->la_expire = 0; 435 ln->ln_ntick = 0; 436 canceled = callout_stop(&ln->ln_timer_ch); 437 } else { 438 ln->la_expire = time_uptime + tick / hz; 439 LLE_ADDREF(ln); 440 if (tick > INT_MAX) { 441 ln->ln_ntick = tick - INT_MAX; 442 canceled = callout_reset(&ln->ln_timer_ch, INT_MAX, 443 nd6_llinfo_timer, ln); 444 } else { 445 ln->ln_ntick = 0; 446 canceled = callout_reset(&ln->ln_timer_ch, tick, 447 nd6_llinfo_timer, ln); 448 } 449 } 450 if (canceled) 451 LLE_REMREF(ln); 452} 453 454void 455nd6_llinfo_settimer(struct llentry *ln, long tick) 456{ 457 458 LLE_WLOCK(ln); 459 nd6_llinfo_settimer_locked(ln, tick); 460 LLE_WUNLOCK(ln); 461} 462 463static void 464nd6_llinfo_timer(void *arg) 465{ 466 struct llentry *ln; 467 struct in6_addr *dst; 468 struct ifnet *ifp; 469 struct nd_ifinfo *ndi = NULL; 470 471 KASSERT(arg != NULL, ("%s: arg NULL", __func__)); 472 ln = (struct llentry *)arg; 473 LLE_WLOCK(ln); 474 if (callout_pending(&ln->la_timer)) { 475 /* 476 * Here we are a bit odd here in the treatment of 477 * active/pending. If the pending bit is set, it got 478 * rescheduled before I ran. The active 479 * bit we ignore, since if it was stopped 480 * in ll_tablefree() and was currently running 481 * it would have return 0 so the code would 482 * not have deleted it since the callout could 483 * not be stopped so we want to go through 484 * with the delete here now. If the callout 485 * was restarted, the pending bit will be back on and 486 * we just want to bail since the callout_reset would 487 * return 1 and our reference would have been removed 488 * by nd6_llinfo_settimer_locked above since canceled 489 * would have been 1. 490 */ 491 LLE_WUNLOCK(ln); 492 return; 493 } 494 ifp = ln->lle_tbl->llt_ifp; 495 CURVNET_SET(ifp->if_vnet); 496 497 if (ln->ln_ntick > 0) { 498 if (ln->ln_ntick > INT_MAX) { 499 ln->ln_ntick -= INT_MAX; 500 nd6_llinfo_settimer_locked(ln, INT_MAX); 501 } else { 502 ln->ln_ntick = 0; 503 nd6_llinfo_settimer_locked(ln, ln->ln_ntick); 504 } 505 goto done; 506 } 507 508 ndi = ND_IFINFO(ifp); 509 dst = &L3_ADDR_SIN6(ln)->sin6_addr; 510 if (ln->la_flags & LLE_STATIC) { 511 goto done; 512 } 513 514 if (ln->la_flags & LLE_DELETED) { 515 (void)nd6_free(ln, 0); 516 ln = NULL; 517 goto done; 518 } 519 520 switch (ln->ln_state) { 521 case ND6_LLINFO_INCOMPLETE: 522 if (ln->la_asked < V_nd6_mmaxtries) { 523 ln->la_asked++; 524 nd6_llinfo_settimer_locked(ln, (long)ndi->retrans * hz / 1000); 525 LLE_WUNLOCK(ln); 526 nd6_ns_output(ifp, NULL, dst, ln, NULL); 527 LLE_WLOCK(ln); 528 } else { 529 struct mbuf *m = ln->la_hold; 530 if (m) { 531 struct mbuf *m0; 532 533 /* 534 * assuming every packet in la_hold has the 535 * same IP header. Send error after unlock. 536 */ 537 m0 = m->m_nextpkt; 538 m->m_nextpkt = NULL; 539 ln->la_hold = m0; 540 clear_llinfo_pqueue(ln); 541 } 542 EVENTHANDLER_INVOKE(lle_event, ln, LLENTRY_TIMEDOUT); 543 (void)nd6_free(ln, 0); 544 ln = NULL; 545 if (m != NULL) 546 icmp6_error2(m, ICMP6_DST_UNREACH, 547 ICMP6_DST_UNREACH_ADDR, 0, ifp); 548 } 549 break; 550 case ND6_LLINFO_REACHABLE: 551 if (!ND6_LLINFO_PERMANENT(ln)) { 552 ln->ln_state = ND6_LLINFO_STALE; 553 nd6_llinfo_settimer_locked(ln, (long)V_nd6_gctimer * hz); 554 } 555 break; 556 557 case ND6_LLINFO_STALE: 558 /* Garbage Collection(RFC 2461 5.3) */ 559 if (!ND6_LLINFO_PERMANENT(ln)) { 560 EVENTHANDLER_INVOKE(lle_event, ln, LLENTRY_EXPIRED); 561 (void)nd6_free(ln, 1); 562 ln = NULL; 563 } 564 break; 565 566 case ND6_LLINFO_DELAY: 567 if (ndi && (ndi->flags & ND6_IFF_PERFORMNUD) != 0) { 568 /* We need NUD */ 569 ln->la_asked = 1; 570 ln->ln_state = ND6_LLINFO_PROBE; 571 nd6_llinfo_settimer_locked(ln, (long)ndi->retrans * hz / 1000); 572 LLE_WUNLOCK(ln); 573 nd6_ns_output(ifp, dst, dst, ln, NULL); 574 LLE_WLOCK(ln); 575 } else { 576 ln->ln_state = ND6_LLINFO_STALE; /* XXX */ 577 nd6_llinfo_settimer_locked(ln, (long)V_nd6_gctimer * hz); 578 } 579 break; 580 case ND6_LLINFO_PROBE: 581 if (ln->la_asked < V_nd6_umaxtries) { 582 ln->la_asked++; 583 nd6_llinfo_settimer_locked(ln, (long)ndi->retrans * hz / 1000); 584 LLE_WUNLOCK(ln); 585 nd6_ns_output(ifp, dst, dst, ln, NULL); 586 LLE_WLOCK(ln); 587 } else { 588 EVENTHANDLER_INVOKE(lle_event, ln, LLENTRY_EXPIRED); 589 (void)nd6_free(ln, 0); 590 ln = NULL; 591 } 592 break; 593 default: 594 panic("%s: paths in a dark night can be confusing: %d", 595 __func__, ln->ln_state); 596 } 597done: 598 if (ln != NULL) 599 LLE_FREE_LOCKED(ln); 600 CURVNET_RESTORE(); 601} 602 603 604/* 605 * ND6 timer routine to expire default route list and prefix list 606 */ 607void 608nd6_timer(void *arg) 609{ 610 CURVNET_SET((struct vnet *) arg); 611 struct nd_defrouter *dr, *ndr; 612 struct nd_prefix *pr, *npr; 613 struct in6_ifaddr *ia6, *nia6; 614 615 callout_reset(&V_nd6_timer_ch, V_nd6_prune * hz, 616 nd6_timer, curvnet); 617 618 /* expire default router list */ 619 TAILQ_FOREACH_SAFE(dr, &V_nd_defrouter, dr_entry, ndr) { 620 if (dr->expire && dr->expire < time_uptime) 621 defrtrlist_del(dr); 622 } 623 624 /* 625 * expire interface addresses. 626 * in the past the loop was inside prefix expiry processing. 627 * However, from a stricter speci-confrmance standpoint, we should 628 * rather separate address lifetimes and prefix lifetimes. 629 * 630 * XXXRW: in6_ifaddrhead locking. 631 */ 632 addrloop: 633 TAILQ_FOREACH_SAFE(ia6, &V_in6_ifaddrhead, ia_link, nia6) { 634 /* check address lifetime */ 635 if (IFA6_IS_INVALID(ia6)) { 636 int regen = 0; 637 638 /* 639 * If the expiring address is temporary, try 640 * regenerating a new one. This would be useful when 641 * we suspended a laptop PC, then turned it on after a 642 * period that could invalidate all temporary 643 * addresses. Although we may have to restart the 644 * loop (see below), it must be after purging the 645 * address. Otherwise, we'd see an infinite loop of 646 * regeneration. 647 */ 648 if (V_ip6_use_tempaddr && 649 (ia6->ia6_flags & IN6_IFF_TEMPORARY) != 0) { 650 if (regen_tmpaddr(ia6) == 0) 651 regen = 1; 652 } 653 654 in6_purgeaddr(&ia6->ia_ifa); 655 656 if (regen) 657 goto addrloop; /* XXX: see below */ 658 } else if (IFA6_IS_DEPRECATED(ia6)) { 659 int oldflags = ia6->ia6_flags; 660 661 ia6->ia6_flags |= IN6_IFF_DEPRECATED; 662 663 /* 664 * If a temporary address has just become deprecated, 665 * regenerate a new one if possible. 666 */ 667 if (V_ip6_use_tempaddr && 668 (ia6->ia6_flags & IN6_IFF_TEMPORARY) != 0 && 669 (oldflags & IN6_IFF_DEPRECATED) == 0) { 670 671 if (regen_tmpaddr(ia6) == 0) { 672 /* 673 * A new temporary address is 674 * generated. 675 * XXX: this means the address chain 676 * has changed while we are still in 677 * the loop. Although the change 678 * would not cause disaster (because 679 * it's not a deletion, but an 680 * addition,) we'd rather restart the 681 * loop just for safety. Or does this 682 * significantly reduce performance?? 683 */ 684 goto addrloop; 685 } 686 } 687 } else if ((ia6->ia6_flags & IN6_IFF_TENTATIVE) != 0) { 688 /* 689 * Schedule DAD for a tentative address. This happens 690 * if the interface was down or not running 691 * when the address was configured. 692 */ 693 int delay; 694 695 delay = arc4random() % 696 (MAX_RTR_SOLICITATION_DELAY * hz); 697 nd6_dad_start((struct ifaddr *)ia6, delay); 698 } else { 699 /* 700 * Check status of the interface. If it is down, 701 * mark the address as tentative for future DAD. 702 */ 703 if ((ia6->ia_ifp->if_flags & IFF_UP) == 0 || 704 (ia6->ia_ifp->if_drv_flags & IFF_DRV_RUNNING) 705 == 0 || 706 (ND_IFINFO(ia6->ia_ifp)->flags & 707 ND6_IFF_IFDISABLED) != 0) { 708 ia6->ia6_flags &= ~IN6_IFF_DUPLICATED; 709 ia6->ia6_flags |= IN6_IFF_TENTATIVE; 710 } 711 /* 712 * A new RA might have made a deprecated address 713 * preferred. 714 */ 715 ia6->ia6_flags &= ~IN6_IFF_DEPRECATED; 716 } 717 } 718 719 /* expire prefix list */ 720 LIST_FOREACH_SAFE(pr, &V_nd_prefix, ndpr_entry, npr) { 721 /* 722 * check prefix lifetime. 723 * since pltime is just for autoconf, pltime processing for 724 * prefix is not necessary. 725 */ 726 if (pr->ndpr_vltime != ND6_INFINITE_LIFETIME && 727 time_uptime - pr->ndpr_lastupdate > pr->ndpr_vltime) { 728 729 /* 730 * address expiration and prefix expiration are 731 * separate. NEVER perform in6_purgeaddr here. 732 */ 733 prelist_remove(pr); 734 } 735 } 736 CURVNET_RESTORE(); 737} 738 739/* 740 * ia6 - deprecated/invalidated temporary address 741 */ 742static int 743regen_tmpaddr(struct in6_ifaddr *ia6) 744{ 745 struct ifaddr *ifa; 746 struct ifnet *ifp; 747 struct in6_ifaddr *public_ifa6 = NULL; 748 749 ifp = ia6->ia_ifa.ifa_ifp; 750 IF_ADDR_RLOCK(ifp); 751 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 752 struct in6_ifaddr *it6; 753 754 if (ifa->ifa_addr->sa_family != AF_INET6) 755 continue; 756 757 it6 = (struct in6_ifaddr *)ifa; 758 759 /* ignore no autoconf addresses. */ 760 if ((it6->ia6_flags & IN6_IFF_AUTOCONF) == 0) 761 continue; 762 763 /* ignore autoconf addresses with different prefixes. */ 764 if (it6->ia6_ndpr == NULL || it6->ia6_ndpr != ia6->ia6_ndpr) 765 continue; 766 767 /* 768 * Now we are looking at an autoconf address with the same 769 * prefix as ours. If the address is temporary and is still 770 * preferred, do not create another one. It would be rare, but 771 * could happen, for example, when we resume a laptop PC after 772 * a long period. 773 */ 774 if ((it6->ia6_flags & IN6_IFF_TEMPORARY) != 0 && 775 !IFA6_IS_DEPRECATED(it6)) { 776 public_ifa6 = NULL; 777 break; 778 } 779 780 /* 781 * This is a public autoconf address that has the same prefix 782 * as ours. If it is preferred, keep it. We can't break the 783 * loop here, because there may be a still-preferred temporary 784 * address with the prefix. 785 */ 786 if (!IFA6_IS_DEPRECATED(it6)) 787 public_ifa6 = it6; 788 } 789 if (public_ifa6 != NULL) 790 ifa_ref(&public_ifa6->ia_ifa); 791 IF_ADDR_RUNLOCK(ifp); 792 793 if (public_ifa6 != NULL) { 794 int e; 795 796 if ((e = in6_tmpifadd(public_ifa6, 0, 0)) != 0) { 797 ifa_free(&public_ifa6->ia_ifa); 798 log(LOG_NOTICE, "regen_tmpaddr: failed to create a new" 799 " tmp addr,errno=%d\n", e); 800 return (-1); 801 } 802 ifa_free(&public_ifa6->ia_ifa); 803 return (0); 804 } 805 806 return (-1); 807} 808 809/* 810 * Nuke neighbor cache/prefix/default router management table, right before 811 * ifp goes away. 812 */ 813void 814nd6_purge(struct ifnet *ifp) 815{ 816 struct nd_defrouter *dr, *ndr; 817 struct nd_prefix *pr, *npr; 818 819 /* 820 * Nuke default router list entries toward ifp. 821 * We defer removal of default router list entries that is installed 822 * in the routing table, in order to keep additional side effects as 823 * small as possible. 824 */ 825 TAILQ_FOREACH_SAFE(dr, &V_nd_defrouter, dr_entry, ndr) { 826 if (dr->installed) 827 continue; 828 829 if (dr->ifp == ifp) 830 defrtrlist_del(dr); 831 } 832 833 TAILQ_FOREACH_SAFE(dr, &V_nd_defrouter, dr_entry, ndr) { 834 if (!dr->installed) 835 continue; 836 837 if (dr->ifp == ifp) 838 defrtrlist_del(dr); 839 } 840 841 /* Nuke prefix list entries toward ifp */ 842 LIST_FOREACH_SAFE(pr, &V_nd_prefix, ndpr_entry, npr) { 843 if (pr->ndpr_ifp == ifp) { 844 /* 845 * Because if_detach() does *not* release prefixes 846 * while purging addresses the reference count will 847 * still be above zero. We therefore reset it to 848 * make sure that the prefix really gets purged. 849 */ 850 pr->ndpr_refcnt = 0; 851 852 /* 853 * Previously, pr->ndpr_addr is removed as well, 854 * but I strongly believe we don't have to do it. 855 * nd6_purge() is only called from in6_ifdetach(), 856 * which removes all the associated interface addresses 857 * by itself. 858 * (jinmei@kame.net 20010129) 859 */ 860 prelist_remove(pr); 861 } 862 } 863 864 /* cancel default outgoing interface setting */ 865 if (V_nd6_defifindex == ifp->if_index) 866 nd6_setdefaultiface(0); 867 868 if (ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV) { 869 /* Refresh default router list. */ 870 defrouter_select(); 871 } 872 873 /* XXXXX 874 * We do not nuke the neighbor cache entries here any more 875 * because the neighbor cache is kept in if_afdata[AF_INET6]. 876 * nd6_purge() is invoked by in6_ifdetach() which is called 877 * from if_detach() where everything gets purged. So let 878 * in6_domifdetach() do the actual L2 table purging work. 879 */ 880} 881 882/* 883 * the caller acquires and releases the lock on the lltbls 884 * Returns the llentry locked 885 */ 886struct llentry * 887nd6_lookup(struct in6_addr *addr6, int flags, struct ifnet *ifp) 888{ 889 struct sockaddr_in6 sin6; 890 struct llentry *ln; 891 int llflags; 892 893 bzero(&sin6, sizeof(sin6)); 894 sin6.sin6_len = sizeof(struct sockaddr_in6); 895 sin6.sin6_family = AF_INET6; 896 sin6.sin6_addr = *addr6; 897 898 IF_AFDATA_LOCK_ASSERT(ifp); 899 900 llflags = 0; 901 if (flags & ND6_CREATE) 902 llflags |= LLE_CREATE; 903 if (flags & ND6_EXCLUSIVE) 904 llflags |= LLE_EXCLUSIVE; 905 906 ln = lla_lookup(LLTABLE6(ifp), llflags, (struct sockaddr *)&sin6); 907 if ((ln != NULL) && (llflags & LLE_CREATE)) 908 ln->ln_state = ND6_LLINFO_NOSTATE; 909 910 return (ln); 911} 912 913/* 914 * Test whether a given IPv6 address is a neighbor or not, ignoring 915 * the actual neighbor cache. The neighbor cache is ignored in order 916 * to not reenter the routing code from within itself. 917 */ 918static int 919nd6_is_new_addr_neighbor(struct sockaddr_in6 *addr, struct ifnet *ifp) 920{ 921 struct nd_prefix *pr; 922 struct ifaddr *dstaddr; 923 924 /* 925 * A link-local address is always a neighbor. 926 * XXX: a link does not necessarily specify a single interface. 927 */ 928 if (IN6_IS_ADDR_LINKLOCAL(&addr->sin6_addr)) { 929 struct sockaddr_in6 sin6_copy; 930 u_int32_t zone; 931 932 /* 933 * We need sin6_copy since sa6_recoverscope() may modify the 934 * content (XXX). 935 */ 936 sin6_copy = *addr; 937 if (sa6_recoverscope(&sin6_copy)) 938 return (0); /* XXX: should be impossible */ 939 if (in6_setscope(&sin6_copy.sin6_addr, ifp, &zone)) 940 return (0); 941 if (sin6_copy.sin6_scope_id == zone) 942 return (1); 943 else 944 return (0); 945 } 946 947 /* 948 * If the address matches one of our addresses, 949 * it should be a neighbor. 950 * If the address matches one of our on-link prefixes, it should be a 951 * neighbor. 952 */ 953 LIST_FOREACH(pr, &V_nd_prefix, ndpr_entry) { 954 if (pr->ndpr_ifp != ifp) 955 continue; 956 957 if (!(pr->ndpr_stateflags & NDPRF_ONLINK)) { 958 struct rtentry *rt; 959 960 /* Always use the default FIB here. */ 961 rt = in6_rtalloc1((struct sockaddr *)&pr->ndpr_prefix, 962 0, 0, RT_DEFAULT_FIB); 963 if (rt == NULL) 964 continue; 965 /* 966 * This is the case where multiple interfaces 967 * have the same prefix, but only one is installed 968 * into the routing table and that prefix entry 969 * is not the one being examined here. In the case 970 * where RADIX_MPATH is enabled, multiple route 971 * entries (of the same rt_key value) will be 972 * installed because the interface addresses all 973 * differ. 974 */ 975 if (!IN6_ARE_ADDR_EQUAL(&pr->ndpr_prefix.sin6_addr, 976 &((struct sockaddr_in6 *)rt_key(rt))->sin6_addr)) { 977 RTFREE_LOCKED(rt); 978 continue; 979 } 980 RTFREE_LOCKED(rt); 981 } 982 983 if (IN6_ARE_MASKED_ADDR_EQUAL(&pr->ndpr_prefix.sin6_addr, 984 &addr->sin6_addr, &pr->ndpr_mask)) 985 return (1); 986 } 987 988 /* 989 * If the address is assigned on the node of the other side of 990 * a p2p interface, the address should be a neighbor. 991 */ 992 dstaddr = ifa_ifwithdstaddr((struct sockaddr *)addr); 993 if (dstaddr != NULL) { 994 if (dstaddr->ifa_ifp == ifp) { 995 ifa_free(dstaddr); 996 return (1); 997 } 998 ifa_free(dstaddr); 999 } 1000 1001 /* 1002 * If the default router list is empty, all addresses are regarded 1003 * as on-link, and thus, as a neighbor. 1004 */ 1005 if (ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV && 1006 TAILQ_EMPTY(&V_nd_defrouter) && 1007 V_nd6_defifindex == ifp->if_index) { 1008 return (1); 1009 } 1010 1011 return (0); 1012} 1013 1014 1015/* 1016 * Detect if a given IPv6 address identifies a neighbor on a given link. 1017 * XXX: should take care of the destination of a p2p link? 1018 */ 1019int 1020nd6_is_addr_neighbor(struct sockaddr_in6 *addr, struct ifnet *ifp) 1021{ 1022 struct llentry *lle; 1023 int rc = 0; 1024 1025 IF_AFDATA_UNLOCK_ASSERT(ifp); 1026 if (nd6_is_new_addr_neighbor(addr, ifp)) 1027 return (1); 1028 1029 /* 1030 * Even if the address matches none of our addresses, it might be 1031 * in the neighbor cache. 1032 */ 1033 IF_AFDATA_RLOCK(ifp); 1034 if ((lle = nd6_lookup(&addr->sin6_addr, 0, ifp)) != NULL) { 1035 LLE_RUNLOCK(lle); 1036 rc = 1; 1037 } 1038 IF_AFDATA_RUNLOCK(ifp); 1039 return (rc); 1040} 1041 1042/* 1043 * Free an nd6 llinfo entry. 1044 * Since the function would cause significant changes in the kernel, DO NOT 1045 * make it global, unless you have a strong reason for the change, and are sure 1046 * that the change is safe. 1047 */ 1048static struct llentry * 1049nd6_free(struct llentry *ln, int gc) 1050{ 1051 struct llentry *next; 1052 struct nd_defrouter *dr; 1053 struct ifnet *ifp; 1054 1055 LLE_WLOCK_ASSERT(ln); 1056 1057 /* 1058 * we used to have pfctlinput(PRC_HOSTDEAD) here. 1059 * even though it is not harmful, it was not really necessary. 1060 */ 1061 1062 /* cancel timer */ 1063 nd6_llinfo_settimer_locked(ln, -1); 1064 1065 ifp = ln->lle_tbl->llt_ifp; 1066 1067 if (ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV) { 1068 dr = defrouter_lookup(&L3_ADDR_SIN6(ln)->sin6_addr, ifp); 1069 1070 if (dr != NULL && dr->expire && 1071 ln->ln_state == ND6_LLINFO_STALE && gc) { 1072 /* 1073 * If the reason for the deletion is just garbage 1074 * collection, and the neighbor is an active default 1075 * router, do not delete it. Instead, reset the GC 1076 * timer using the router's lifetime. 1077 * Simply deleting the entry would affect default 1078 * router selection, which is not necessarily a good 1079 * thing, especially when we're using router preference 1080 * values. 1081 * XXX: the check for ln_state would be redundant, 1082 * but we intentionally keep it just in case. 1083 */ 1084 if (dr->expire > time_uptime) 1085 nd6_llinfo_settimer_locked(ln, 1086 (dr->expire - time_uptime) * hz); 1087 else 1088 nd6_llinfo_settimer_locked(ln, 1089 (long)V_nd6_gctimer * hz); 1090 1091 next = LIST_NEXT(ln, lle_next); 1092 LLE_REMREF(ln); 1093 LLE_WUNLOCK(ln); 1094 return (next); 1095 } 1096 1097 if (dr) { 1098 /* 1099 * Unreachablity of a router might affect the default 1100 * router selection and on-link detection of advertised 1101 * prefixes. 1102 */ 1103 1104 /* 1105 * Temporarily fake the state to choose a new default 1106 * router and to perform on-link determination of 1107 * prefixes correctly. 1108 * Below the state will be set correctly, 1109 * or the entry itself will be deleted. 1110 */ 1111 ln->ln_state = ND6_LLINFO_INCOMPLETE; 1112 } 1113 1114 if (ln->ln_router || dr) { 1115 1116 /* 1117 * We need to unlock to avoid a LOR with rt6_flush() with the 1118 * rnh and for the calls to pfxlist_onlink_check() and 1119 * defrouter_select() in the block further down for calls 1120 * into nd6_lookup(). We still hold a ref. 1121 */ 1122 LLE_WUNLOCK(ln); 1123 1124 /* 1125 * rt6_flush must be called whether or not the neighbor 1126 * is in the Default Router List. 1127 * See a corresponding comment in nd6_na_input(). 1128 */ 1129 rt6_flush(&L3_ADDR_SIN6(ln)->sin6_addr, ifp); 1130 } 1131 1132 if (dr) { 1133 /* 1134 * Since defrouter_select() does not affect the 1135 * on-link determination and MIP6 needs the check 1136 * before the default router selection, we perform 1137 * the check now. 1138 */ 1139 pfxlist_onlink_check(); 1140 1141 /* 1142 * Refresh default router list. 1143 */ 1144 defrouter_select(); 1145 } 1146 1147 if (ln->ln_router || dr) 1148 LLE_WLOCK(ln); 1149 } 1150 1151 /* 1152 * Before deleting the entry, remember the next entry as the 1153 * return value. We need this because pfxlist_onlink_check() above 1154 * might have freed other entries (particularly the old next entry) as 1155 * a side effect (XXX). 1156 */ 1157 next = LIST_NEXT(ln, lle_next); 1158 1159 /* 1160 * Save to unlock. We still hold an extra reference and will not 1161 * free(9) in llentry_free() if someone else holds one as well. 1162 */ 1163 LLE_WUNLOCK(ln); 1164 IF_AFDATA_LOCK(ifp); 1165 LLE_WLOCK(ln); 1166 1167 /* Guard against race with other llentry_free(). */ 1168 if (ln->la_flags & LLE_LINKED) { 1169 LLE_REMREF(ln); 1170 llentry_free(ln); 1171 } else 1172 LLE_FREE_LOCKED(ln); 1173 1174 IF_AFDATA_UNLOCK(ifp); 1175 1176 return (next); 1177} 1178 1179/* 1180 * Upper-layer reachability hint for Neighbor Unreachability Detection. 1181 * 1182 * XXX cost-effective methods? 1183 */ 1184void 1185nd6_nud_hint(struct rtentry *rt, struct in6_addr *dst6, int force) 1186{ 1187 struct llentry *ln; 1188 struct ifnet *ifp; 1189 1190 if ((dst6 == NULL) || (rt == NULL)) 1191 return; 1192 1193 ifp = rt->rt_ifp; 1194 IF_AFDATA_RLOCK(ifp); 1195 ln = nd6_lookup(dst6, ND6_EXCLUSIVE, NULL); 1196 IF_AFDATA_RUNLOCK(ifp); 1197 if (ln == NULL) 1198 return; 1199 1200 if (ln->ln_state < ND6_LLINFO_REACHABLE) 1201 goto done; 1202 1203 /* 1204 * if we get upper-layer reachability confirmation many times, 1205 * it is possible we have false information. 1206 */ 1207 if (!force) { 1208 ln->ln_byhint++; 1209 if (ln->ln_byhint > V_nd6_maxnudhint) { 1210 goto done; 1211 } 1212 } 1213 1214 ln->ln_state = ND6_LLINFO_REACHABLE; 1215 if (!ND6_LLINFO_PERMANENT(ln)) { 1216 nd6_llinfo_settimer_locked(ln, 1217 (long)ND_IFINFO(rt->rt_ifp)->reachable * hz); 1218 } 1219done: 1220 LLE_WUNLOCK(ln); 1221} 1222 1223 1224/* 1225 * Rejuvenate this function for routing operations related 1226 * processing. 1227 */ 1228void 1229nd6_rtrequest(int req, struct rtentry *rt, struct rt_addrinfo *info) 1230{ 1231 struct sockaddr_in6 *gateway; 1232 struct nd_defrouter *dr; 1233 struct ifnet *ifp; 1234 1235 RT_LOCK_ASSERT(rt); 1236 gateway = (struct sockaddr_in6 *)rt->rt_gateway; 1237 ifp = rt->rt_ifp; 1238 1239 switch (req) { 1240 case RTM_ADD: 1241 break; 1242 1243 case RTM_DELETE: 1244 if (!ifp) 1245 return; 1246 /* 1247 * Only indirect routes are interesting. 1248 */ 1249 if ((rt->rt_flags & RTF_GATEWAY) == 0) 1250 return; 1251 /* 1252 * check for default route 1253 */ 1254 if (IN6_ARE_ADDR_EQUAL(&in6addr_any, 1255 &SIN6(rt_key(rt))->sin6_addr)) { 1256 1257 dr = defrouter_lookup(&gateway->sin6_addr, ifp); 1258 if (dr != NULL) 1259 dr->installed = 0; 1260 } 1261 break; 1262 } 1263} 1264 1265 1266int 1267nd6_ioctl(u_long cmd, caddr_t data, struct ifnet *ifp) 1268{ 1269 struct in6_ndireq *ndi = (struct in6_ndireq *)data; 1270 struct in6_nbrinfo *nbi = (struct in6_nbrinfo *)data; 1271 struct in6_ndifreq *ndif = (struct in6_ndifreq *)data; 1272 int error = 0; 1273 1274 if (ifp->if_afdata[AF_INET6] == NULL) 1275 return (EPFNOSUPPORT); 1276 switch (cmd) { 1277 case OSIOCGIFINFO_IN6: 1278#define ND ndi->ndi 1279 /* XXX: old ndp(8) assumes a positive value for linkmtu. */ 1280 bzero(&ND, sizeof(ND)); 1281 ND.linkmtu = IN6_LINKMTU(ifp); 1282 ND.maxmtu = ND_IFINFO(ifp)->maxmtu; 1283 ND.basereachable = ND_IFINFO(ifp)->basereachable; 1284 ND.reachable = ND_IFINFO(ifp)->reachable; 1285 ND.retrans = ND_IFINFO(ifp)->retrans; 1286 ND.flags = ND_IFINFO(ifp)->flags; 1287 ND.recalctm = ND_IFINFO(ifp)->recalctm; 1288 ND.chlim = ND_IFINFO(ifp)->chlim; 1289 break; 1290 case SIOCGIFINFO_IN6: 1291 ND = *ND_IFINFO(ifp); 1292 break; 1293 case SIOCSIFINFO_IN6: 1294 /* 1295 * used to change host variables from userland. 1296 * intented for a use on router to reflect RA configurations. 1297 */ 1298 /* 0 means 'unspecified' */ 1299 if (ND.linkmtu != 0) { 1300 if (ND.linkmtu < IPV6_MMTU || 1301 ND.linkmtu > IN6_LINKMTU(ifp)) { 1302 error = EINVAL; 1303 break; 1304 } 1305 ND_IFINFO(ifp)->linkmtu = ND.linkmtu; 1306 } 1307 1308 if (ND.basereachable != 0) { 1309 int obasereachable = ND_IFINFO(ifp)->basereachable; 1310 1311 ND_IFINFO(ifp)->basereachable = ND.basereachable; 1312 if (ND.basereachable != obasereachable) 1313 ND_IFINFO(ifp)->reachable = 1314 ND_COMPUTE_RTIME(ND.basereachable); 1315 } 1316 if (ND.retrans != 0) 1317 ND_IFINFO(ifp)->retrans = ND.retrans; 1318 if (ND.chlim != 0) 1319 ND_IFINFO(ifp)->chlim = ND.chlim; 1320 /* FALLTHROUGH */ 1321 case SIOCSIFINFO_FLAGS: 1322 { 1323 struct ifaddr *ifa; 1324 struct in6_ifaddr *ia; 1325 1326 if ((ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) && 1327 !(ND.flags & ND6_IFF_IFDISABLED)) { 1328 /* ifdisabled 1->0 transision */ 1329 1330 /* 1331 * If the interface is marked as ND6_IFF_IFDISABLED and 1332 * has an link-local address with IN6_IFF_DUPLICATED, 1333 * do not clear ND6_IFF_IFDISABLED. 1334 * See RFC 4862, Section 5.4.5. 1335 */ 1336 IF_ADDR_RLOCK(ifp); 1337 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1338 if (ifa->ifa_addr->sa_family != AF_INET6) 1339 continue; 1340 ia = (struct in6_ifaddr *)ifa; 1341 if ((ia->ia6_flags & IN6_IFF_DUPLICATED) && 1342 IN6_IS_ADDR_LINKLOCAL(IA6_IN6(ia))) 1343 break; 1344 } 1345 IF_ADDR_RUNLOCK(ifp); 1346 1347 if (ifa != NULL) { 1348 /* LLA is duplicated. */ 1349 ND.flags |= ND6_IFF_IFDISABLED; 1350 log(LOG_ERR, "Cannot enable an interface" 1351 " with a link-local address marked" 1352 " duplicate.\n"); 1353 } else { 1354 ND_IFINFO(ifp)->flags &= ~ND6_IFF_IFDISABLED; 1355 if (ifp->if_flags & IFF_UP) 1356 in6_if_up(ifp); 1357 } 1358 } else if (!(ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) && 1359 (ND.flags & ND6_IFF_IFDISABLED)) { 1360 /* ifdisabled 0->1 transision */ 1361 /* Mark all IPv6 address as tentative. */ 1362 1363 ND_IFINFO(ifp)->flags |= ND6_IFF_IFDISABLED; 1364 if (V_ip6_dad_count > 0 && 1365 (ND_IFINFO(ifp)->flags & ND6_IFF_NO_DAD) == 0) { 1366 IF_ADDR_RLOCK(ifp); 1367 TAILQ_FOREACH(ifa, &ifp->if_addrhead, 1368 ifa_link) { 1369 if (ifa->ifa_addr->sa_family != 1370 AF_INET6) 1371 continue; 1372 ia = (struct in6_ifaddr *)ifa; 1373 ia->ia6_flags |= IN6_IFF_TENTATIVE; 1374 } 1375 IF_ADDR_RUNLOCK(ifp); 1376 } 1377 } 1378 1379 if (ND.flags & ND6_IFF_AUTO_LINKLOCAL) { 1380 if (!(ND_IFINFO(ifp)->flags & ND6_IFF_AUTO_LINKLOCAL)) { 1381 /* auto_linklocal 0->1 transision */ 1382 1383 /* If no link-local address on ifp, configure */ 1384 ND_IFINFO(ifp)->flags |= ND6_IFF_AUTO_LINKLOCAL; 1385 in6_ifattach(ifp, NULL); 1386 } else if (!(ND.flags & ND6_IFF_IFDISABLED) && 1387 ifp->if_flags & IFF_UP) { 1388 /* 1389 * When the IF already has 1390 * ND6_IFF_AUTO_LINKLOCAL, no link-local 1391 * address is assigned, and IFF_UP, try to 1392 * assign one. 1393 */ 1394 IF_ADDR_RLOCK(ifp); 1395 TAILQ_FOREACH(ifa, &ifp->if_addrhead, 1396 ifa_link) { 1397 if (ifa->ifa_addr->sa_family != 1398 AF_INET6) 1399 continue; 1400 ia = (struct in6_ifaddr *)ifa; 1401 if (IN6_IS_ADDR_LINKLOCAL(IA6_IN6(ia))) 1402 break; 1403 } 1404 IF_ADDR_RUNLOCK(ifp); 1405 if (ifa != NULL) 1406 /* No LLA is configured. */ 1407 in6_ifattach(ifp, NULL); 1408 } 1409 } 1410 } 1411 ND_IFINFO(ifp)->flags = ND.flags; 1412 break; 1413#undef ND 1414 case SIOCSNDFLUSH_IN6: /* XXX: the ioctl name is confusing... */ 1415 /* sync kernel routing table with the default router list */ 1416 defrouter_reset(); 1417 defrouter_select(); 1418 break; 1419 case SIOCSPFXFLUSH_IN6: 1420 { 1421 /* flush all the prefix advertised by routers */ 1422 struct nd_prefix *pr, *next; 1423 1424 LIST_FOREACH_SAFE(pr, &V_nd_prefix, ndpr_entry, next) { 1425 struct in6_ifaddr *ia, *ia_next; 1426 1427 if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr)) 1428 continue; /* XXX */ 1429 1430 /* do we really have to remove addresses as well? */ 1431 /* XXXRW: in6_ifaddrhead locking. */ 1432 TAILQ_FOREACH_SAFE(ia, &V_in6_ifaddrhead, ia_link, 1433 ia_next) { 1434 if ((ia->ia6_flags & IN6_IFF_AUTOCONF) == 0) 1435 continue; 1436 1437 if (ia->ia6_ndpr == pr) 1438 in6_purgeaddr(&ia->ia_ifa); 1439 } 1440 prelist_remove(pr); 1441 } 1442 break; 1443 } 1444 case SIOCSRTRFLUSH_IN6: 1445 { 1446 /* flush all the default routers */ 1447 struct nd_defrouter *dr, *next; 1448 1449 defrouter_reset(); 1450 TAILQ_FOREACH_SAFE(dr, &V_nd_defrouter, dr_entry, next) { 1451 defrtrlist_del(dr); 1452 } 1453 defrouter_select(); 1454 break; 1455 } 1456 case SIOCGNBRINFO_IN6: 1457 { 1458 struct llentry *ln; 1459 struct in6_addr nb_addr = nbi->addr; /* make local for safety */ 1460 1461 if ((error = in6_setscope(&nb_addr, ifp, NULL)) != 0) 1462 return (error); 1463 1464 IF_AFDATA_RLOCK(ifp); 1465 ln = nd6_lookup(&nb_addr, 0, ifp); 1466 IF_AFDATA_RUNLOCK(ifp); 1467 1468 if (ln == NULL) { 1469 error = EINVAL; 1470 break; 1471 } 1472 nbi->state = ln->ln_state; 1473 nbi->asked = ln->la_asked; 1474 nbi->isrouter = ln->ln_router; 1475 if (ln->la_expire == 0) 1476 nbi->expire = 0; 1477 else 1478 nbi->expire = ln->la_expire + 1479 (time_second - time_uptime); 1480 LLE_RUNLOCK(ln); 1481 break; 1482 } 1483 case SIOCGDEFIFACE_IN6: /* XXX: should be implemented as a sysctl? */ 1484 ndif->ifindex = V_nd6_defifindex; 1485 break; 1486 case SIOCSDEFIFACE_IN6: /* XXX: should be implemented as a sysctl? */ 1487 return (nd6_setdefaultiface(ndif->ifindex)); 1488 } 1489 return (error); 1490} 1491 1492/* 1493 * Create neighbor cache entry and cache link-layer address, 1494 * on reception of inbound ND6 packets. (RS/RA/NS/redirect) 1495 * 1496 * type - ICMP6 type 1497 * code - type dependent information 1498 * 1499 * XXXXX 1500 * The caller of this function already acquired the ndp 1501 * cache table lock because the cache entry is returned. 1502 */ 1503struct llentry * 1504nd6_cache_lladdr(struct ifnet *ifp, struct in6_addr *from, char *lladdr, 1505 int lladdrlen, int type, int code) 1506{ 1507 struct llentry *ln = NULL; 1508 int is_newentry; 1509 int do_update; 1510 int olladdr; 1511 int llchange; 1512 int flags; 1513 int newstate = 0; 1514 uint16_t router = 0; 1515 struct sockaddr_in6 sin6; 1516 struct mbuf *chain = NULL; 1517 int static_route = 0; 1518 1519 IF_AFDATA_UNLOCK_ASSERT(ifp); 1520 1521 KASSERT(ifp != NULL, ("%s: ifp == NULL", __func__)); 1522 KASSERT(from != NULL, ("%s: from == NULL", __func__)); 1523 1524 /* nothing must be updated for unspecified address */ 1525 if (IN6_IS_ADDR_UNSPECIFIED(from)) 1526 return NULL; 1527 1528 /* 1529 * Validation about ifp->if_addrlen and lladdrlen must be done in 1530 * the caller. 1531 * 1532 * XXX If the link does not have link-layer adderss, what should 1533 * we do? (ifp->if_addrlen == 0) 1534 * Spec says nothing in sections for RA, RS and NA. There's small 1535 * description on it in NS section (RFC 2461 7.2.3). 1536 */ 1537 flags = lladdr ? ND6_EXCLUSIVE : 0; 1538 IF_AFDATA_RLOCK(ifp); 1539 ln = nd6_lookup(from, flags, ifp); 1540 IF_AFDATA_RUNLOCK(ifp); 1541 if (ln == NULL) { 1542 flags |= ND6_EXCLUSIVE; 1543 IF_AFDATA_LOCK(ifp); 1544 ln = nd6_lookup(from, flags | ND6_CREATE, ifp); 1545 IF_AFDATA_UNLOCK(ifp); 1546 is_newentry = 1; 1547 } else { 1548 /* do nothing if static ndp is set */ 1549 if (ln->la_flags & LLE_STATIC) { 1550 static_route = 1; 1551 goto done; 1552 } 1553 is_newentry = 0; 1554 } 1555 if (ln == NULL) 1556 return (NULL); 1557 1558 olladdr = (ln->la_flags & LLE_VALID) ? 1 : 0; 1559 if (olladdr && lladdr) { 1560 llchange = bcmp(lladdr, &ln->ll_addr, 1561 ifp->if_addrlen); 1562 } else 1563 llchange = 0; 1564 1565 /* 1566 * newentry olladdr lladdr llchange (*=record) 1567 * 0 n n -- (1) 1568 * 0 y n -- (2) 1569 * 0 n y -- (3) * STALE 1570 * 0 y y n (4) * 1571 * 0 y y y (5) * STALE 1572 * 1 -- n -- (6) NOSTATE(= PASSIVE) 1573 * 1 -- y -- (7) * STALE 1574 */ 1575 1576 if (lladdr) { /* (3-5) and (7) */ 1577 /* 1578 * Record source link-layer address 1579 * XXX is it dependent to ifp->if_type? 1580 */ 1581 bcopy(lladdr, &ln->ll_addr, ifp->if_addrlen); 1582 ln->la_flags |= LLE_VALID; 1583 EVENTHANDLER_INVOKE(lle_event, ln, LLENTRY_RESOLVED); 1584 } 1585 1586 if (!is_newentry) { 1587 if ((!olladdr && lladdr != NULL) || /* (3) */ 1588 (olladdr && lladdr != NULL && llchange)) { /* (5) */ 1589 do_update = 1; 1590 newstate = ND6_LLINFO_STALE; 1591 } else /* (1-2,4) */ 1592 do_update = 0; 1593 } else { 1594 do_update = 1; 1595 if (lladdr == NULL) /* (6) */ 1596 newstate = ND6_LLINFO_NOSTATE; 1597 else /* (7) */ 1598 newstate = ND6_LLINFO_STALE; 1599 } 1600 1601 if (do_update) { 1602 /* 1603 * Update the state of the neighbor cache. 1604 */ 1605 ln->ln_state = newstate; 1606 1607 if (ln->ln_state == ND6_LLINFO_STALE) { 1608 if (ln->la_hold != NULL) 1609 nd6_grab_holdchain(ln, &chain, &sin6); 1610 } else if (ln->ln_state == ND6_LLINFO_INCOMPLETE) { 1611 /* probe right away */ 1612 nd6_llinfo_settimer_locked((void *)ln, 0); 1613 } 1614 } 1615 1616 /* 1617 * ICMP6 type dependent behavior. 1618 * 1619 * NS: clear IsRouter if new entry 1620 * RS: clear IsRouter 1621 * RA: set IsRouter if there's lladdr 1622 * redir: clear IsRouter if new entry 1623 * 1624 * RA case, (1): 1625 * The spec says that we must set IsRouter in the following cases: 1626 * - If lladdr exist, set IsRouter. This means (1-5). 1627 * - If it is old entry (!newentry), set IsRouter. This means (7). 1628 * So, based on the spec, in (1-5) and (7) cases we must set IsRouter. 1629 * A quetion arises for (1) case. (1) case has no lladdr in the 1630 * neighbor cache, this is similar to (6). 1631 * This case is rare but we figured that we MUST NOT set IsRouter. 1632 * 1633 * newentry olladdr lladdr llchange NS RS RA redir 1634 * D R 1635 * 0 n n -- (1) c ? s 1636 * 0 y n -- (2) c s s 1637 * 0 n y -- (3) c s s 1638 * 0 y y n (4) c s s 1639 * 0 y y y (5) c s s 1640 * 1 -- n -- (6) c c c s 1641 * 1 -- y -- (7) c c s c s 1642 * 1643 * (c=clear s=set) 1644 */ 1645 switch (type & 0xff) { 1646 case ND_NEIGHBOR_SOLICIT: 1647 /* 1648 * New entry must have is_router flag cleared. 1649 */ 1650 if (is_newentry) /* (6-7) */ 1651 ln->ln_router = 0; 1652 break; 1653 case ND_REDIRECT: 1654 /* 1655 * If the icmp is a redirect to a better router, always set the 1656 * is_router flag. Otherwise, if the entry is newly created, 1657 * clear the flag. [RFC 2461, sec 8.3] 1658 */ 1659 if (code == ND_REDIRECT_ROUTER) 1660 ln->ln_router = 1; 1661 else if (is_newentry) /* (6-7) */ 1662 ln->ln_router = 0; 1663 break; 1664 case ND_ROUTER_SOLICIT: 1665 /* 1666 * is_router flag must always be cleared. 1667 */ 1668 ln->ln_router = 0; 1669 break; 1670 case ND_ROUTER_ADVERT: 1671 /* 1672 * Mark an entry with lladdr as a router. 1673 */ 1674 if ((!is_newentry && (olladdr || lladdr)) || /* (2-5) */ 1675 (is_newentry && lladdr)) { /* (7) */ 1676 ln->ln_router = 1; 1677 } 1678 break; 1679 } 1680 1681 if (ln != NULL) { 1682 static_route = (ln->la_flags & LLE_STATIC); 1683 router = ln->ln_router; 1684 1685 if (flags & ND6_EXCLUSIVE) 1686 LLE_WUNLOCK(ln); 1687 else 1688 LLE_RUNLOCK(ln); 1689 if (static_route) 1690 ln = NULL; 1691 } 1692 if (chain != NULL) 1693 nd6_flush_holdchain(ifp, ifp, chain, &sin6); 1694 1695 /* 1696 * When the link-layer address of a router changes, select the 1697 * best router again. In particular, when the neighbor entry is newly 1698 * created, it might affect the selection policy. 1699 * Question: can we restrict the first condition to the "is_newentry" 1700 * case? 1701 * XXX: when we hear an RA from a new router with the link-layer 1702 * address option, defrouter_select() is called twice, since 1703 * defrtrlist_update called the function as well. However, I believe 1704 * we can compromise the overhead, since it only happens the first 1705 * time. 1706 * XXX: although defrouter_select() should not have a bad effect 1707 * for those are not autoconfigured hosts, we explicitly avoid such 1708 * cases for safety. 1709 */ 1710 if (do_update && router && 1711 ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV) { 1712 /* 1713 * guaranteed recursion 1714 */ 1715 defrouter_select(); 1716 } 1717 1718 return (ln); 1719done: 1720 if (ln != NULL) { 1721 if (flags & ND6_EXCLUSIVE) 1722 LLE_WUNLOCK(ln); 1723 else 1724 LLE_RUNLOCK(ln); 1725 if (static_route) 1726 ln = NULL; 1727 } 1728 return (ln); 1729} 1730 1731static void 1732nd6_slowtimo(void *arg) 1733{ 1734 CURVNET_SET((struct vnet *) arg); 1735 struct nd_ifinfo *nd6if; 1736 struct ifnet *ifp; 1737 1738 callout_reset(&V_nd6_slowtimo_ch, ND6_SLOWTIMER_INTERVAL * hz, 1739 nd6_slowtimo, curvnet); 1740 IFNET_RLOCK_NOSLEEP(); 1741 TAILQ_FOREACH(ifp, &V_ifnet, if_list) { 1742 if (ifp->if_afdata[AF_INET6] == NULL) 1743 continue; 1744 nd6if = ND_IFINFO(ifp); 1745 if (nd6if->basereachable && /* already initialized */ 1746 (nd6if->recalctm -= ND6_SLOWTIMER_INTERVAL) <= 0) { 1747 /* 1748 * Since reachable time rarely changes by router 1749 * advertisements, we SHOULD insure that a new random 1750 * value gets recomputed at least once every few hours. 1751 * (RFC 2461, 6.3.4) 1752 */ 1753 nd6if->recalctm = V_nd6_recalc_reachtm_interval; 1754 nd6if->reachable = ND_COMPUTE_RTIME(nd6if->basereachable); 1755 } 1756 } 1757 IFNET_RUNLOCK_NOSLEEP(); 1758 CURVNET_RESTORE(); 1759} 1760 1761void 1762nd6_grab_holdchain(struct llentry *ln, struct mbuf **chain, 1763 struct sockaddr_in6 *sin6) 1764{ 1765 1766 LLE_WLOCK_ASSERT(ln); 1767 1768 *chain = ln->la_hold; 1769 ln->la_hold = NULL; 1770 memcpy(sin6, L3_ADDR_SIN6(ln), sizeof(*sin6)); 1771 1772 if (ln->ln_state == ND6_LLINFO_STALE) { 1773 1774 /* 1775 * The first time we send a packet to a 1776 * neighbor whose entry is STALE, we have 1777 * to change the state to DELAY and a sets 1778 * a timer to expire in DELAY_FIRST_PROBE_TIME 1779 * seconds to ensure do neighbor unreachability 1780 * detection on expiration. 1781 * (RFC 2461 7.3.3) 1782 */ 1783 ln->la_asked = 0; 1784 ln->ln_state = ND6_LLINFO_DELAY; 1785 nd6_llinfo_settimer_locked(ln, (long)V_nd6_delay * hz); 1786 } 1787} 1788 1789static int 1790nd6_output_ifp(struct ifnet *ifp, struct ifnet *origifp, struct mbuf *m, 1791 struct sockaddr_in6 *dst) 1792{ 1793 int error; 1794 int ip6len; 1795 struct ip6_hdr *ip6; 1796 struct m_tag *mtag; 1797 1798#ifdef MAC 1799 mac_netinet6_nd6_send(ifp, m); 1800#endif 1801 1802 /* 1803 * If called from nd6_ns_output() (NS), nd6_na_output() (NA), 1804 * icmp6_redirect_output() (REDIRECT) or from rip6_output() (RS, RA 1805 * as handled by rtsol and rtadvd), mbufs will be tagged for SeND 1806 * to be diverted to user space. When re-injected into the kernel, 1807 * send_output() will directly dispatch them to the outgoing interface. 1808 */ 1809 if (send_sendso_input_hook != NULL) { 1810 mtag = m_tag_find(m, PACKET_TAG_ND_OUTGOING, NULL); 1811 if (mtag != NULL) { 1812 ip6 = mtod(m, struct ip6_hdr *); 1813 ip6len = sizeof(struct ip6_hdr) + ntohs(ip6->ip6_plen); 1814 /* Use the SEND socket */ 1815 error = send_sendso_input_hook(m, ifp, SND_OUT, 1816 ip6len); 1817 /* -1 == no app on SEND socket */ 1818 if (error == 0 || error != -1) 1819 return (error); 1820 } 1821 } 1822 1823 m_clrprotoflags(m); /* Avoid confusing lower layers. */ 1824 IP_PROBE(send, NULL, NULL, mtod(m, struct ip6_hdr *), ifp, NULL, 1825 mtod(m, struct ip6_hdr *)); 1826 1827 if ((ifp->if_flags & IFF_LOOPBACK) == 0) 1828 origifp = ifp; 1829 1830 error = (*ifp->if_output)(origifp, m, (struct sockaddr *)dst, NULL); 1831 return (error); 1832} 1833 1834/* 1835 * IPv6 packet output - light version. 1836 * Checks if destination LLE exists and is in proper state 1837 * (e.g no modification required). If not true, fall back to 1838 * "heavy" version. 1839 */ 1840int 1841nd6_output(struct ifnet *ifp, struct ifnet *origifp, struct mbuf *m, 1842 struct sockaddr_in6 *dst, struct rtentry *rt0) 1843{ 1844 struct llentry *ln = NULL; 1845 1846 /* discard the packet if IPv6 operation is disabled on the interface */ 1847 if ((ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED)) { 1848 m_freem(m); 1849 return (ENETDOWN); /* better error? */ 1850 } 1851 1852 if (IN6_IS_ADDR_MULTICAST(&dst->sin6_addr)) 1853 goto sendpkt; 1854 1855 if (nd6_need_cache(ifp) == 0) 1856 goto sendpkt; 1857 1858 IF_AFDATA_RLOCK(ifp); 1859 ln = nd6_lookup(&dst->sin6_addr, 0, ifp); 1860 IF_AFDATA_RUNLOCK(ifp); 1861 1862 /* 1863 * Perform fast path for the following cases: 1864 * 1) lle state is REACHABLE 1865 * 2) lle state is DELAY (NS message sentNS message sent) 1866 * 1867 * Every other case involves lle modification, so we handle 1868 * them separately. 1869 */ 1870 if (ln == NULL || (ln->ln_state != ND6_LLINFO_REACHABLE && 1871 ln->ln_state != ND6_LLINFO_DELAY)) { 1872 /* Fall back to slow processing path */ 1873 if (ln != NULL) 1874 LLE_RUNLOCK(ln); 1875 return (nd6_output_lle(ifp, origifp, m, dst)); 1876 } 1877 1878sendpkt: 1879 if (ln != NULL) 1880 LLE_RUNLOCK(ln); 1881 1882 return (nd6_output_ifp(ifp, origifp, m, dst)); 1883} 1884 1885 1886/* 1887 * Output IPv6 packet - heavy version. 1888 * Function assume that either 1889 * 1) destination LLE does not exist, is invalid or stale, so 1890 * ND6_EXCLUSIVE lock needs to be acquired 1891 * 2) destination lle is provided (with ND6_EXCLUSIVE lock), 1892 * in that case packets are queued in &chain. 1893 * 1894 */ 1895static int 1896nd6_output_lle(struct ifnet *ifp, struct ifnet *origifp, struct mbuf *m, 1897 struct sockaddr_in6 *dst) 1898{ 1899 struct llentry *lle = NULL; 1900 int flags = 0; 1901 1902 KASSERT(m != NULL, ("NULL mbuf, nothing to send")); 1903 /* discard the packet if IPv6 operation is disabled on the interface */ 1904 if ((ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED)) { 1905 m_freem(m); 1906 return (ENETDOWN); /* better error? */ 1907 } 1908 1909 if (IN6_IS_ADDR_MULTICAST(&dst->sin6_addr)) 1910 goto sendpkt; 1911 1912 if (nd6_need_cache(ifp) == 0) 1913 goto sendpkt; 1914 1915 /* 1916 * Address resolution or Neighbor Unreachability Detection 1917 * for the next hop. 1918 * At this point, the destination of the packet must be a unicast 1919 * or an anycast address(i.e. not a multicast). 1920 */ 1921 if (lle == NULL) { 1922 IF_AFDATA_RLOCK(ifp); 1923 lle = nd6_lookup(&dst->sin6_addr, ND6_EXCLUSIVE, ifp); 1924 IF_AFDATA_RUNLOCK(ifp); 1925 if ((lle == NULL) && nd6_is_addr_neighbor(dst, ifp)) { 1926 /* 1927 * Since nd6_is_addr_neighbor() internally calls nd6_lookup(), 1928 * the condition below is not very efficient. But we believe 1929 * it is tolerable, because this should be a rare case. 1930 */ 1931 flags = ND6_CREATE | ND6_EXCLUSIVE; 1932 IF_AFDATA_LOCK(ifp); 1933 lle = nd6_lookup(&dst->sin6_addr, flags, ifp); 1934 IF_AFDATA_UNLOCK(ifp); 1935 } 1936 } 1937 if (lle == NULL) { 1938 if ((ifp->if_flags & IFF_POINTOPOINT) == 0 && 1939 !(ND_IFINFO(ifp)->flags & ND6_IFF_PERFORMNUD)) { 1940 char ip6buf[INET6_ADDRSTRLEN]; 1941 log(LOG_DEBUG, 1942 "nd6_output: can't allocate llinfo for %s " 1943 "(ln=%p)\n", 1944 ip6_sprintf(ip6buf, &dst->sin6_addr), lle); 1945 m_freem(m); 1946 return (ENOBUFS); 1947 } 1948 goto sendpkt; /* send anyway */ 1949 } 1950 1951 LLE_WLOCK_ASSERT(lle); 1952 1953 /* We don't have to do link-layer address resolution on a p2p link. */ 1954 if ((ifp->if_flags & IFF_POINTOPOINT) != 0 && 1955 lle->ln_state < ND6_LLINFO_REACHABLE) { 1956 lle->ln_state = ND6_LLINFO_STALE; 1957 nd6_llinfo_settimer_locked(lle, (long)V_nd6_gctimer * hz); 1958 } 1959 1960 /* 1961 * The first time we send a packet to a neighbor whose entry is 1962 * STALE, we have to change the state to DELAY and a sets a timer to 1963 * expire in DELAY_FIRST_PROBE_TIME seconds to ensure do 1964 * neighbor unreachability detection on expiration. 1965 * (RFC 2461 7.3.3) 1966 */ 1967 if (lle->ln_state == ND6_LLINFO_STALE) { 1968 lle->la_asked = 0; 1969 lle->ln_state = ND6_LLINFO_DELAY; 1970 nd6_llinfo_settimer_locked(lle, (long)V_nd6_delay * hz); 1971 } 1972 1973 /* 1974 * If the neighbor cache entry has a state other than INCOMPLETE 1975 * (i.e. its link-layer address is already resolved), just 1976 * send the packet. 1977 */ 1978 if (lle->ln_state > ND6_LLINFO_INCOMPLETE) 1979 goto sendpkt; 1980 1981 /* 1982 * There is a neighbor cache entry, but no ethernet address 1983 * response yet. Append this latest packet to the end of the 1984 * packet queue in the mbuf, unless the number of the packet 1985 * does not exceed nd6_maxqueuelen. When it exceeds nd6_maxqueuelen, 1986 * the oldest packet in the queue will be removed. 1987 */ 1988 if (lle->ln_state == ND6_LLINFO_NOSTATE) 1989 lle->ln_state = ND6_LLINFO_INCOMPLETE; 1990 1991 if (lle->la_hold != NULL) { 1992 struct mbuf *m_hold; 1993 int i; 1994 1995 i = 0; 1996 for (m_hold = lle->la_hold; m_hold; m_hold = m_hold->m_nextpkt){ 1997 i++; 1998 if (m_hold->m_nextpkt == NULL) { 1999 m_hold->m_nextpkt = m; 2000 break; 2001 } 2002 } 2003 while (i >= V_nd6_maxqueuelen) { 2004 m_hold = lle->la_hold; 2005 lle->la_hold = lle->la_hold->m_nextpkt; 2006 m_freem(m_hold); 2007 i--; 2008 } 2009 } else { 2010 lle->la_hold = m; 2011 } 2012 2013 /* 2014 * If there has been no NS for the neighbor after entering the 2015 * INCOMPLETE state, send the first solicitation. 2016 */ 2017 if (!ND6_LLINFO_PERMANENT(lle) && lle->la_asked == 0) { 2018 lle->la_asked++; 2019 2020 nd6_llinfo_settimer_locked(lle, 2021 (long)ND_IFINFO(ifp)->retrans * hz / 1000); 2022 LLE_WUNLOCK(lle); 2023 nd6_ns_output(ifp, NULL, &dst->sin6_addr, lle, NULL); 2024 } else { 2025 /* We did the lookup so we need to do the unlock here. */ 2026 LLE_WUNLOCK(lle); 2027 } 2028 2029 return (0); 2030 2031 sendpkt: 2032 if (lle != NULL) 2033 LLE_WUNLOCK(lle); 2034 2035 return (nd6_output_ifp(ifp, origifp, m, dst)); 2036} 2037 2038 2039int 2040nd6_flush_holdchain(struct ifnet *ifp, struct ifnet *origifp, struct mbuf *chain, 2041 struct sockaddr_in6 *dst) 2042{ 2043 struct mbuf *m, *m_head; 2044 struct ifnet *outifp; 2045 int error = 0; 2046 2047 m_head = chain; 2048 if ((ifp->if_flags & IFF_LOOPBACK) != 0) 2049 outifp = origifp; 2050 else 2051 outifp = ifp; 2052 2053 while (m_head) { 2054 m = m_head; 2055 m_head = m_head->m_nextpkt; 2056 error = nd6_output_ifp(ifp, origifp, m, dst); 2057 } 2058 2059 /* 2060 * XXX 2061 * note that intermediate errors are blindly ignored - but this is 2062 * the same convention as used with nd6_output when called by 2063 * nd6_cache_lladdr 2064 */ 2065 return (error); 2066} 2067 2068 2069int 2070nd6_need_cache(struct ifnet *ifp) 2071{ 2072 /* 2073 * XXX: we currently do not make neighbor cache on any interface 2074 * other than ARCnet, Ethernet, FDDI and GIF. 2075 * 2076 * RFC2893 says: 2077 * - unidirectional tunnels needs no ND 2078 */ 2079 switch (ifp->if_type) { 2080 case IFT_ARCNET: 2081 case IFT_ETHER: 2082 case IFT_FDDI: 2083 case IFT_IEEE1394: 2084#ifdef IFT_L2VLAN 2085 case IFT_L2VLAN: 2086#endif 2087#ifdef IFT_IEEE80211 2088 case IFT_IEEE80211: 2089#endif 2090 case IFT_INFINIBAND: 2091 case IFT_BRIDGE: 2092 case IFT_PROPVIRTUAL: 2093 return (1); 2094 default: 2095 return (0); 2096 } 2097} 2098 2099/* 2100 * the callers of this function need to be re-worked to drop 2101 * the lle lock, drop here for now 2102 */ 2103int 2104nd6_storelladdr(struct ifnet *ifp, struct mbuf *m, 2105 const struct sockaddr *dst, u_char *desten, struct llentry **lle) 2106{ 2107 struct llentry *ln; 2108 2109 *lle = NULL; 2110 IF_AFDATA_UNLOCK_ASSERT(ifp); 2111 if (m != NULL && m->m_flags & M_MCAST) { 2112 int i; 2113 2114 switch (ifp->if_type) { 2115 case IFT_ETHER: 2116 case IFT_FDDI: 2117#ifdef IFT_L2VLAN 2118 case IFT_L2VLAN: 2119#endif 2120#ifdef IFT_IEEE80211 2121 case IFT_IEEE80211: 2122#endif 2123 case IFT_BRIDGE: 2124 case IFT_ISO88025: 2125 ETHER_MAP_IPV6_MULTICAST(&SIN6(dst)->sin6_addr, 2126 desten); 2127 return (0); 2128 case IFT_IEEE1394: 2129 /* 2130 * netbsd can use if_broadcastaddr, but we don't do so 2131 * to reduce # of ifdef. 2132 */ 2133 for (i = 0; i < ifp->if_addrlen; i++) 2134 desten[i] = ~0; 2135 return (0); 2136 case IFT_ARCNET: 2137 *desten = 0; 2138 return (0); 2139 default: 2140 m_freem(m); 2141 return (EAFNOSUPPORT); 2142 } 2143 } 2144 2145 2146 /* 2147 * the entry should have been created in nd6_store_lladdr 2148 */ 2149 IF_AFDATA_RLOCK(ifp); 2150 ln = lla_lookup(LLTABLE6(ifp), 0, dst); 2151 IF_AFDATA_RUNLOCK(ifp); 2152 if ((ln == NULL) || !(ln->la_flags & LLE_VALID)) { 2153 if (ln != NULL) 2154 LLE_RUNLOCK(ln); 2155 /* this could happen, if we could not allocate memory */ 2156 m_freem(m); 2157 return (1); 2158 } 2159 2160 bcopy(&ln->ll_addr, desten, ifp->if_addrlen); 2161 *lle = ln; 2162 LLE_RUNLOCK(ln); 2163 /* 2164 * A *small* use after free race exists here 2165 */ 2166 return (0); 2167} 2168 2169static void 2170clear_llinfo_pqueue(struct llentry *ln) 2171{ 2172 struct mbuf *m_hold, *m_hold_next; 2173 2174 for (m_hold = ln->la_hold; m_hold; m_hold = m_hold_next) { 2175 m_hold_next = m_hold->m_nextpkt; 2176 m_freem(m_hold); 2177 } 2178 2179 ln->la_hold = NULL; 2180} 2181 2182static int nd6_sysctl_drlist(SYSCTL_HANDLER_ARGS); 2183static int nd6_sysctl_prlist(SYSCTL_HANDLER_ARGS); 2184#ifdef SYSCTL_DECL 2185SYSCTL_DECL(_net_inet6_icmp6); 2186#endif 2187SYSCTL_NODE(_net_inet6_icmp6, ICMPV6CTL_ND6_DRLIST, nd6_drlist, 2188 CTLFLAG_RD, nd6_sysctl_drlist, ""); 2189SYSCTL_NODE(_net_inet6_icmp6, ICMPV6CTL_ND6_PRLIST, nd6_prlist, 2190 CTLFLAG_RD, nd6_sysctl_prlist, ""); 2191SYSCTL_VNET_INT(_net_inet6_icmp6, ICMPV6CTL_ND6_MAXQLEN, nd6_maxqueuelen, 2192 CTLFLAG_RW, &VNET_NAME(nd6_maxqueuelen), 1, ""); 2193SYSCTL_VNET_INT(_net_inet6_icmp6, OID_AUTO, nd6_gctimer, 2194 CTLFLAG_RW, &VNET_NAME(nd6_gctimer), (60 * 60 * 24), ""); 2195 2196static int 2197nd6_sysctl_drlist(SYSCTL_HANDLER_ARGS) 2198{ 2199 struct in6_defrouter d; 2200 struct nd_defrouter *dr; 2201 int error; 2202 2203 if (req->newptr) 2204 return (EPERM); 2205 2206 bzero(&d, sizeof(d)); 2207 d.rtaddr.sin6_family = AF_INET6; 2208 d.rtaddr.sin6_len = sizeof(d.rtaddr); 2209 2210 /* 2211 * XXX locking 2212 */ 2213 TAILQ_FOREACH(dr, &V_nd_defrouter, dr_entry) { 2214 d.rtaddr.sin6_addr = dr->rtaddr; 2215 error = sa6_recoverscope(&d.rtaddr); 2216 if (error != 0) 2217 return (error); 2218 d.flags = dr->raflags; 2219 d.rtlifetime = dr->rtlifetime; 2220 d.expire = dr->expire + (time_second - time_uptime); 2221 d.if_index = dr->ifp->if_index; 2222 error = SYSCTL_OUT(req, &d, sizeof(d)); 2223 if (error != 0) 2224 return (error); 2225 } 2226 return (0); 2227} 2228 2229static int 2230nd6_sysctl_prlist(SYSCTL_HANDLER_ARGS) 2231{ 2232 struct in6_prefix p; 2233 struct sockaddr_in6 s6; 2234 struct nd_prefix *pr; 2235 struct nd_pfxrouter *pfr; 2236 time_t maxexpire; 2237 int error; 2238 char ip6buf[INET6_ADDRSTRLEN]; 2239 2240 if (req->newptr) 2241 return (EPERM); 2242 2243 bzero(&p, sizeof(p)); 2244 p.origin = PR_ORIG_RA; 2245 bzero(&s6, sizeof(s6)); 2246 s6.sin6_family = AF_INET6; 2247 s6.sin6_len = sizeof(s6); 2248 2249 /* 2250 * XXX locking 2251 */ 2252 LIST_FOREACH(pr, &V_nd_prefix, ndpr_entry) { 2253 p.prefix = pr->ndpr_prefix; 2254 if (sa6_recoverscope(&p.prefix)) { 2255 log(LOG_ERR, "scope error in prefix list (%s)\n", 2256 ip6_sprintf(ip6buf, &p.prefix.sin6_addr)); 2257 /* XXX: press on... */ 2258 } 2259 p.raflags = pr->ndpr_raf; 2260 p.prefixlen = pr->ndpr_plen; 2261 p.vltime = pr->ndpr_vltime; 2262 p.pltime = pr->ndpr_pltime; 2263 p.if_index = pr->ndpr_ifp->if_index; 2264 if (pr->ndpr_vltime == ND6_INFINITE_LIFETIME) 2265 p.expire = 0; 2266 else { 2267 /* XXX: we assume time_t is signed. */ 2268 maxexpire = (-1) & 2269 ~((time_t)1 << ((sizeof(maxexpire) * 8) - 1)); 2270 if (pr->ndpr_vltime < maxexpire - pr->ndpr_lastupdate) 2271 p.expire = pr->ndpr_lastupdate + 2272 pr->ndpr_vltime + 2273 (time_second - time_uptime); 2274 else 2275 p.expire = maxexpire; 2276 } 2277 p.refcnt = pr->ndpr_refcnt; 2278 p.flags = pr->ndpr_stateflags; 2279 p.advrtrs = 0; 2280 LIST_FOREACH(pfr, &pr->ndpr_advrtrs, pfr_entry) 2281 p.advrtrs++; 2282 error = SYSCTL_OUT(req, &p, sizeof(p)); 2283 if (error != 0) 2284 return (error); 2285 LIST_FOREACH(pfr, &pr->ndpr_advrtrs, pfr_entry) { 2286 s6.sin6_addr = pfr->router->rtaddr; 2287 if (sa6_recoverscope(&s6)) 2288 log(LOG_ERR, 2289 "scope error in prefix list (%s)\n", 2290 ip6_sprintf(ip6buf, &pfr->router->rtaddr)); 2291 error = SYSCTL_OUT(req, &s6, sizeof(s6)); 2292 if (error != 0) 2293 return (error); 2294 } 2295 } 2296 return (0); 2297} 2298