1 /* $FreeBSD: src/sys/netinet6/nd6.c,v 1.2.2.15 2003/05/06 06:46:58 suz Exp $ */ 2 /* $KAME: nd6.c,v 1.144 2001/05/24 07:44:00 itojun Exp $ */ 3 4 /* 5 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project. 6 * All rights reserved. 7 * 8 * Redistribution and use in source and binary forms, with or without 9 * modification, are permitted provided that the following conditions 10 * are met: 11 * 1. Redistributions of source code must retain the above copyright 12 * notice, this list of conditions and the following disclaimer. 13 * 2. Redistributions in binary form must reproduce the above copyright 14 * notice, this list of conditions and the following disclaimer in the 15 * documentation and/or other materials provided with the distribution. 16 * 3. Neither the name of the project nor the names of its contributors 17 * may be used to endorse or promote products derived from this software 18 * without specific prior written permission. 19 * 20 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND 21 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 22 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 23 * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE 24 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 25 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 26 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 27 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 28 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 29 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 30 * SUCH DAMAGE. 31 */ 32 33 #include "opt_inet.h" 34 #include "opt_inet6.h" 35 36 #include <sys/param.h> 37 #include <sys/systm.h> 38 #include <sys/callout.h> 39 #include <sys/malloc.h> 40 #include <sys/mbuf.h> 41 #include <sys/socket.h> 42 #include <sys/sockio.h> 43 #include <sys/time.h> 44 #include <sys/kernel.h> 45 #include <sys/protosw.h> 46 #include <sys/errno.h> 47 #include <sys/syslog.h> 48 #include <sys/queue.h> 49 #include <sys/sysctl.h> 50 #include <sys/mutex.h> 51 52 #include <sys/thread2.h> 53 #include <sys/mutex2.h> 54 55 #include <net/if.h> 56 #include <net/if_dl.h> 57 #include <net/if_types.h> 58 #include <net/route.h> 59 #include <net/netisr2.h> 60 #include <net/netmsg2.h> 61 62 #include <netinet/in.h> 63 #include <netinet/if_ether.h> 64 #include <netinet6/in6_var.h> 65 #include <netinet/ip6.h> 66 #include <netinet6/ip6_var.h> 67 #include <netinet6/nd6.h> 68 #include <netinet/icmp6.h> 69 70 #include <net/net_osdep.h> 71 72 #define ND6_SLOWTIMER_INTERVAL (60 * 60) /* 1 hour */ 73 #define ND6_RECALC_REACHTM_INTERVAL (60 * 120) /* 2 hours */ 74 75 #define SIN6(s) ((struct sockaddr_in6 *)s) 76 #define SDL(s) ((struct sockaddr_dl *)s) 77 78 /* 79 * Note that the check for rt_llinfo is necessary because a cloned 80 * route from a parent route that has the L flag (e.g. the default 81 * route to a p2p interface) may have the flag, too, while the 82 * destination is not actually a neighbor. 83 * XXX: we can't use rt->rt_ifp to check for the interface, since 84 * it might be the loopback interface if the entry is for our 85 * own address on a non-loopback interface. Instead, we should 86 * use rt->rt_ifa->ifa_ifp, which would specify the REAL 87 * interface. 88 */ 89 #define ND6_RTENTRY_IS_NEIGHBOR(rt, ifp) \ 90 !(((rt)->rt_flags & RTF_GATEWAY) || \ 91 ((rt)->rt_flags & RTF_LLINFO) == 0 || \ 92 (rt)->rt_gateway->sa_family != AF_LINK || \ 93 (rt)->rt_llinfo == NULL || \ 94 ((ifp) != NULL && (rt)->rt_ifa->ifa_ifp != (ifp))) 95 96 #define ND6_RTENTRY_IS_LLCLONING(rt) \ 97 (((rt)->rt_flags & (RTF_PRCLONING | RTF_LLINFO)) == \ 98 (RTF_PRCLONING | RTF_LLINFO) || \ 99 ((rt)->rt_flags & RTF_CLONING)) 100 101 /* timer values */ 102 int nd6_prune = 1; /* walk list every 1 seconds */ 103 int nd6_delay = 5; /* delay first probe time 5 second */ 104 int nd6_umaxtries = 3; /* maximum unicast query */ 105 int nd6_mmaxtries = 3; /* maximum multicast query */ 106 int nd6_useloopback = 1; /* use loopback interface for local traffic */ 107 int nd6_gctimer = (60 * 60 * 24); /* 1 day: garbage collection timer */ 108 109 /* preventing too many loops in ND option parsing */ 110 int nd6_maxndopt = 10; /* max # of ND options allowed */ 111 112 int nd6_maxnudhint = 0; /* max # of subsequent upper layer hints */ 113 114 #ifdef ND6_DEBUG 115 int nd6_debug = 1; 116 #else 117 int nd6_debug = 0; 118 #endif 119 120 /* for debugging? */ 121 static int nd6_inuse, nd6_allocated; 122 123 struct llinfo_nd6 llinfo_nd6 = {&llinfo_nd6, &llinfo_nd6}; 124 struct nd_drhead nd_defrouter; 125 struct nd_prhead nd_prefix = { 0 }; 126 struct mtx nd6_mtx = MTX_INITIALIZER("nd6"); 127 128 int nd6_recalc_reachtm_interval = ND6_RECALC_REACHTM_INTERVAL; 129 static struct sockaddr_in6 all1_sa; 130 131 static void nd6_setmtu0 (struct ifnet *, struct nd_ifinfo *); 132 static int regen_tmpaddr (struct in6_ifaddr *); 133 static void nd6_slowtimo(void *); 134 static void nd6_slowtimo_dispatch(netmsg_t); 135 static void nd6_timer(void *); 136 static void nd6_timer_dispatch(netmsg_t); 137 138 static struct callout nd6_slowtimo_ch; 139 static struct netmsg_base nd6_slowtimo_netmsg; 140 141 static struct callout nd6_timer_ch; 142 static struct netmsg_base nd6_timer_netmsg; 143 144 void 145 nd6_init(void) 146 { 147 static int nd6_init_done = 0; 148 int i; 149 150 if (nd6_init_done) { 151 log(LOG_NOTICE, "nd6_init called more than once(ignored)\n"); 152 return; 153 } 154 155 all1_sa.sin6_family = AF_INET6; 156 all1_sa.sin6_len = sizeof(struct sockaddr_in6); 157 for (i = 0; i < sizeof(all1_sa.sin6_addr); i++) 158 all1_sa.sin6_addr.s6_addr[i] = 0xff; 159 160 /* initialization of the default router list */ 161 TAILQ_INIT(&nd_defrouter); 162 163 nd6_init_done = 1; 164 165 /* start timer */ 166 callout_init_mp(&nd6_slowtimo_ch); 167 netmsg_init(&nd6_slowtimo_netmsg, NULL, &netisr_adone_rport, 168 MSGF_PRIORITY, nd6_slowtimo_dispatch); 169 callout_reset_bycpu(&nd6_slowtimo_ch, ND6_SLOWTIMER_INTERVAL * hz, 170 nd6_slowtimo, NULL, 0); 171 } 172 173 struct nd_ifinfo * 174 nd6_ifattach(struct ifnet *ifp) 175 { 176 struct nd_ifinfo *nd; 177 178 nd = (struct nd_ifinfo *)kmalloc(sizeof(*nd), M_IP6NDP, 179 M_WAITOK | M_ZERO); 180 181 nd->initialized = 1; 182 183 nd->chlim = IPV6_DEFHLIM; 184 nd->basereachable = REACHABLE_TIME; 185 nd->reachable = ND_COMPUTE_RTIME(nd->basereachable); 186 nd->retrans = RETRANS_TIMER; 187 188 /* 189 * Note that the default value of ip6_accept_rtadv is 0, which means 190 * we won't accept RAs by default even if we set ND6_IFF_ACCEPT_RTADV 191 * here. 192 */ 193 nd->flags = (ND6_IFF_PERFORMNUD | ND6_IFF_ACCEPT_RTADV); 194 195 /* XXX: we cannot call nd6_setmtu since ifp is not fully initialized */ 196 nd6_setmtu0(ifp, nd); 197 return nd; 198 } 199 200 void 201 nd6_ifdetach(struct nd_ifinfo *nd) 202 { 203 kfree(nd, M_IP6NDP); 204 } 205 206 /* 207 * Reset ND level link MTU. This function is called when the physical MTU 208 * changes, which means we might have to adjust the ND level MTU. 209 */ 210 void 211 nd6_setmtu(struct ifnet *ifp) 212 { 213 nd6_setmtu0(ifp, ND_IFINFO(ifp)); 214 } 215 216 struct netmsg_nd6setmtu { 217 struct netmsg_base nmsg; 218 struct ifnet *ifp; 219 struct nd_ifinfo *ndi; 220 }; 221 222 /* XXX todo: do not maintain copy of ifp->if_mtu in ndi->maxmtu */ 223 static void 224 nd6_setmtu0_dispatch(netmsg_t msg) 225 { 226 struct netmsg_nd6setmtu *nmsg = (struct netmsg_nd6setmtu *)msg; 227 struct ifnet *ifp = nmsg->ifp; 228 struct nd_ifinfo *ndi = nmsg->ndi; 229 uint32_t omaxmtu; 230 231 omaxmtu = ndi->maxmtu; 232 233 switch (ifp->if_type) { 234 case IFT_ETHER: 235 ndi->maxmtu = MIN(ETHERMTU, ifp->if_mtu); 236 break; 237 case IFT_IEEE1394: /* XXX should be IEEE1394MTU(1500) */ 238 ndi->maxmtu = MIN(ETHERMTU, ifp->if_mtu); 239 break; 240 #ifdef IFT_IEEE80211 241 case IFT_IEEE80211: /* XXX should be IEEE80211MTU(1500) */ 242 ndi->maxmtu = MIN(ETHERMTU, ifp->if_mtu); 243 break; 244 #endif 245 default: 246 ndi->maxmtu = ifp->if_mtu; 247 break; 248 } 249 250 /* 251 * Decreasing the interface MTU under IPV6 minimum MTU may cause 252 * undesirable situation. We thus notify the operator of the change 253 * explicitly. The check for omaxmtu is necessary to restrict the 254 * log to the case of changing the MTU, not initializing it. 255 */ 256 if (omaxmtu >= IPV6_MMTU && ndi->maxmtu < IPV6_MMTU) { 257 log(LOG_NOTICE, "nd6_setmtu0: " 258 "new link MTU on %s (%lu) is too small for IPv6\n", 259 if_name(ifp), (unsigned long)ndi->maxmtu); 260 } 261 262 if (ndi->maxmtu > in6_maxmtu) 263 in6_setmaxmtu(); /* check all interfaces just in case */ 264 265 lwkt_replymsg(&nmsg->nmsg.lmsg, 0); 266 } 267 268 void 269 nd6_setmtu0(struct ifnet *ifp, struct nd_ifinfo *ndi) 270 { 271 struct netmsg_nd6setmtu nmsg; 272 273 netmsg_init(&nmsg.nmsg, NULL, &curthread->td_msgport, 0, 274 nd6_setmtu0_dispatch); 275 nmsg.ifp = ifp; 276 nmsg.ndi = ndi; 277 lwkt_domsg(netisr_cpuport(0), &nmsg.nmsg.lmsg, 0); 278 } 279 280 void 281 nd6_option_init(void *opt, int icmp6len, union nd_opts *ndopts) 282 { 283 bzero(ndopts, sizeof(*ndopts)); 284 ndopts->nd_opts_search = (struct nd_opt_hdr *)opt; 285 ndopts->nd_opts_last 286 = (struct nd_opt_hdr *)(((u_char *)opt) + icmp6len); 287 288 if (icmp6len == 0) { 289 ndopts->nd_opts_done = 1; 290 ndopts->nd_opts_search = NULL; 291 } 292 } 293 294 /* 295 * Take one ND option. 296 */ 297 struct nd_opt_hdr * 298 nd6_option(union nd_opts *ndopts) 299 { 300 struct nd_opt_hdr *nd_opt; 301 int olen; 302 303 if (!ndopts) 304 panic("ndopts == NULL in nd6_option"); 305 if (!ndopts->nd_opts_last) 306 panic("uninitialized ndopts in nd6_option"); 307 if (!ndopts->nd_opts_search) 308 return NULL; 309 if (ndopts->nd_opts_done) 310 return NULL; 311 312 nd_opt = ndopts->nd_opts_search; 313 314 /* make sure nd_opt_len is inside the buffer */ 315 if ((caddr_t)&nd_opt->nd_opt_len >= (caddr_t)ndopts->nd_opts_last) { 316 bzero(ndopts, sizeof(*ndopts)); 317 return NULL; 318 } 319 320 olen = nd_opt->nd_opt_len << 3; 321 if (olen == 0) { 322 /* 323 * Message validation requires that all included 324 * options have a length that is greater than zero. 325 */ 326 bzero(ndopts, sizeof(*ndopts)); 327 return NULL; 328 } 329 330 ndopts->nd_opts_search = (struct nd_opt_hdr *)((caddr_t)nd_opt + olen); 331 if (ndopts->nd_opts_search > ndopts->nd_opts_last) { 332 /* option overruns the end of buffer, invalid */ 333 bzero(ndopts, sizeof(*ndopts)); 334 return NULL; 335 } else if (ndopts->nd_opts_search == ndopts->nd_opts_last) { 336 /* reached the end of options chain */ 337 ndopts->nd_opts_done = 1; 338 ndopts->nd_opts_search = NULL; 339 } 340 return nd_opt; 341 } 342 343 /* 344 * Parse multiple ND options. 345 * This function is much easier to use, for ND routines that do not need 346 * multiple options of the same type. 347 */ 348 int 349 nd6_options(union nd_opts *ndopts) 350 { 351 struct nd_opt_hdr *nd_opt; 352 int i = 0; 353 354 if (!ndopts) 355 panic("ndopts == NULL in nd6_options"); 356 if (!ndopts->nd_opts_last) 357 panic("uninitialized ndopts in nd6_options"); 358 if (!ndopts->nd_opts_search) 359 return 0; 360 361 while (1) { 362 nd_opt = nd6_option(ndopts); 363 if (!nd_opt && !ndopts->nd_opts_last) { 364 /* 365 * Message validation requires that all included 366 * options have a length that is greater than zero. 367 */ 368 icmp6stat.icp6s_nd_badopt++; 369 bzero(ndopts, sizeof(*ndopts)); 370 return -1; 371 } 372 373 if (!nd_opt) 374 goto skip1; 375 376 switch (nd_opt->nd_opt_type) { 377 case ND_OPT_SOURCE_LINKADDR: 378 case ND_OPT_TARGET_LINKADDR: 379 case ND_OPT_MTU: 380 case ND_OPT_REDIRECTED_HEADER: 381 if (ndopts->nd_opt_array[nd_opt->nd_opt_type]) { 382 nd6log((LOG_INFO, 383 "duplicated ND6 option found (type=%d)\n", 384 nd_opt->nd_opt_type)); 385 /* XXX bark? */ 386 } else { 387 ndopts->nd_opt_array[nd_opt->nd_opt_type] 388 = nd_opt; 389 } 390 break; 391 case ND_OPT_PREFIX_INFORMATION: 392 if (ndopts->nd_opt_array[nd_opt->nd_opt_type] == 0) { 393 ndopts->nd_opt_array[nd_opt->nd_opt_type] 394 = nd_opt; 395 } 396 ndopts->nd_opts_pi_end = 397 (struct nd_opt_prefix_info *)nd_opt; 398 break; 399 default: 400 /* 401 * Unknown options must be silently ignored, 402 * to accomodate future extension to the protocol. 403 */ 404 nd6log((LOG_DEBUG, 405 "nd6_options: unsupported option %d - " 406 "option ignored\n", nd_opt->nd_opt_type)); 407 } 408 409 skip1: 410 i++; 411 if (i > nd6_maxndopt) { 412 icmp6stat.icp6s_nd_toomanyopt++; 413 nd6log((LOG_INFO, "too many loop in nd opt\n")); 414 break; 415 } 416 417 if (ndopts->nd_opts_done) 418 break; 419 } 420 421 return 0; 422 } 423 424 /* 425 * ND6 timer routine to expire default route list and prefix list 426 */ 427 static void 428 nd6_timer_dispatch(netmsg_t nmsg) 429 { 430 struct llinfo_nd6 *ln; 431 struct nd_defrouter *dr; 432 struct nd_prefix *pr; 433 struct ifnet *ifp; 434 struct in6_ifaddr *ia6, *nia6; 435 436 ASSERT_NETISR0; 437 438 crit_enter(); 439 lwkt_replymsg(&nmsg->lmsg, 0); /* reply ASAP */ 440 crit_exit(); 441 442 mtx_lock(&nd6_mtx); 443 444 ln = llinfo_nd6.ln_next; 445 while (ln && ln != &llinfo_nd6) { 446 struct rtentry *rt; 447 struct sockaddr_in6 *dst; 448 struct llinfo_nd6 *next = ln->ln_next; 449 /* XXX: used for the DELAY case only: */ 450 struct nd_ifinfo *ndi = NULL; 451 452 if ((rt = ln->ln_rt) == NULL) { 453 ln = next; 454 continue; 455 } 456 if ((ifp = rt->rt_ifp) == NULL) { 457 ln = next; 458 continue; 459 } 460 ndi = ND_IFINFO(ifp); 461 dst = (struct sockaddr_in6 *)rt_key(rt); 462 463 if (ln->ln_expire > time_uptime) { 464 ln = next; 465 continue; 466 } 467 468 /* sanity check */ 469 if (!rt) 470 panic("rt=0 in nd6_timer(ln=%p)", ln); 471 if (rt->rt_llinfo && (struct llinfo_nd6 *)rt->rt_llinfo != ln) 472 panic("rt_llinfo(%p) is not equal to ln(%p)", 473 rt->rt_llinfo, ln); 474 if (!dst) 475 panic("dst=0 in nd6_timer(ln=%p)", ln); 476 477 switch (ln->ln_state) { 478 case ND6_LLINFO_WAITDELETE: 479 next = nd6_free(rt); 480 break; 481 case ND6_LLINFO_INCOMPLETE: 482 if (ln->ln_asked++ >= nd6_mmaxtries) { 483 struct mbuf *m = ln->ln_hold; 484 if (m) { 485 if (rt->rt_ifp) { 486 /* 487 * Fake rcvif to make ICMP error 488 * more helpful in diagnosing 489 * for the receiver. 490 * XXX: should we consider 491 * older rcvif? 492 */ 493 m->m_pkthdr.rcvif = rt->rt_ifp; 494 } 495 /* XXX This will log a diagnostic 496 * that it's not an IPv6 header. */ 497 icmp6_error(m, ICMP6_DST_UNREACH, 498 ICMP6_DST_UNREACH_ADDR, 0); 499 ln->ln_hold = NULL; 500 } 501 ln->ln_state = ND6_LLINFO_WAITDELETE; 502 rt_rtmsg(RTM_MISS, rt, rt->rt_ifp, 0); 503 } 504 ln->ln_expire = time_uptime + 505 ND_IFINFO(ifp)->retrans / 1000; 506 nd6_ns_output(ifp, NULL, &dst->sin6_addr, 507 ln, 0); 508 break; 509 case ND6_LLINFO_REACHABLE: 510 if (ln->ln_expire) { 511 ln->ln_state = ND6_LLINFO_STALE; 512 ln->ln_expire = time_uptime + nd6_gctimer; 513 } 514 break; 515 516 case ND6_LLINFO_STALE: 517 /* Garbage Collection(RFC 2461 5.3) */ 518 if (ln->ln_expire) 519 next = nd6_free(rt); 520 break; 521 522 case ND6_LLINFO_DELAY: 523 if (ndi && (ndi->flags & ND6_IFF_PERFORMNUD)) { 524 /* We need NUD */ 525 ln->ln_asked = 1; 526 ln->ln_state = ND6_LLINFO_PROBE; 527 ln->ln_expire = time_uptime + 528 ndi->retrans / 1000; 529 nd6_ns_output(ifp, &dst->sin6_addr, 530 &dst->sin6_addr, 531 ln, 0); 532 } else { 533 ln->ln_state = ND6_LLINFO_STALE; /* XXX */ 534 ln->ln_expire = time_uptime + nd6_gctimer; 535 } 536 break; 537 case ND6_LLINFO_PROBE: 538 if (ln->ln_asked < nd6_umaxtries) { 539 ln->ln_asked++; 540 ln->ln_expire = time_uptime + 541 ND_IFINFO(ifp)->retrans / 1000; 542 nd6_ns_output(ifp, &dst->sin6_addr, 543 &dst->sin6_addr, ln, 0); 544 } else { 545 rt_rtmsg(RTM_MISS, rt, rt->rt_ifp, 0); 546 next = nd6_free(rt); 547 } 548 break; 549 } 550 ln = next; 551 } 552 553 /* expire default router list */ 554 dr = TAILQ_FIRST(&nd_defrouter); 555 while (dr) { 556 if (dr->expire && dr->expire < time_uptime) { 557 struct nd_defrouter *t; 558 t = TAILQ_NEXT(dr, dr_entry); 559 defrtrlist_del(dr); 560 dr = t; 561 } else { 562 dr = TAILQ_NEXT(dr, dr_entry); 563 } 564 } 565 566 /* 567 * expire interface addresses. 568 * in the past the loop was inside prefix expiry processing. 569 * However, from a stricter speci-confrmance standpoint, we should 570 * rather separate address lifetimes and prefix lifetimes. 571 */ 572 addrloop: 573 for (ia6 = in6_ifaddr; ia6; ia6 = nia6) { 574 nia6 = ia6->ia_next; 575 /* check address lifetime */ 576 if (IFA6_IS_INVALID(ia6)) { 577 int regen = 0; 578 579 /* 580 * If the expiring address is temporary, try 581 * regenerating a new one. This would be useful when 582 * we suspended a laptop PC, then turned it on after a 583 * period that could invalidate all temporary 584 * addresses. Although we may have to restart the 585 * loop (see below), it must be after purging the 586 * address. Otherwise, we'd see an infinite loop of 587 * regeneration. 588 */ 589 if (ip6_use_tempaddr && 590 (ia6->ia6_flags & IN6_IFF_TEMPORARY)) { 591 if (regen_tmpaddr(ia6) == 0) 592 regen = 1; 593 } 594 595 in6_purgeaddr(&ia6->ia_ifa); 596 597 if (regen) 598 goto addrloop; /* XXX: see below */ 599 /* ia6 is no longer good, continue on to next */ 600 continue; 601 } 602 if (IFA6_IS_DEPRECATED(ia6)) { 603 int oldflags = ia6->ia6_flags; 604 605 if ((oldflags & IN6_IFF_DEPRECATED) == 0) { 606 ia6->ia6_flags |= IN6_IFF_DEPRECATED; 607 in6_newaddrmsg((struct ifaddr *)ia6); 608 } 609 610 /* 611 * If a temporary address has just become deprecated, 612 * regenerate a new one if possible. 613 */ 614 if (ip6_use_tempaddr && 615 (ia6->ia6_flags & IN6_IFF_TEMPORARY) && 616 !(oldflags & IN6_IFF_DEPRECATED)) { 617 618 if (regen_tmpaddr(ia6) == 0) { 619 /* 620 * A new temporary address is 621 * generated. 622 * XXX: this means the address chain 623 * has changed while we are still in 624 * the loop. Although the change 625 * would not cause disaster (because 626 * it's not a deletion, but an 627 * addition,) we'd rather restart the 628 * loop just for safety. Or does this 629 * significantly reduce performance?? 630 */ 631 goto addrloop; 632 } 633 } 634 } else { 635 /* 636 * A new RA might have made a deprecated address 637 * preferred. 638 */ 639 if (ia6->ia6_flags & IN6_IFF_DEPRECATED) { 640 ia6->ia6_flags &= ~IN6_IFF_DEPRECATED; 641 in6_newaddrmsg((struct ifaddr *)ia6); 642 } 643 } 644 } 645 646 /* expire prefix list */ 647 pr = nd_prefix.lh_first; 648 while (pr) { 649 /* 650 * check prefix lifetime. 651 * since pltime is just for autoconf, pltime processing for 652 * prefix is not necessary. 653 */ 654 if (pr->ndpr_expire && pr->ndpr_expire < time_uptime) { 655 struct nd_prefix *t; 656 t = pr->ndpr_next; 657 658 /* 659 * address expiration and prefix expiration are 660 * separate. NEVER perform in6_purgeaddr here. 661 */ 662 663 prelist_remove(pr); 664 pr = t; 665 } else 666 pr = pr->ndpr_next; 667 } 668 669 mtx_unlock(&nd6_mtx); 670 671 callout_reset(&nd6_timer_ch, nd6_prune * hz, nd6_timer, NULL); 672 } 673 674 static void 675 nd6_timer(void *arg __unused) 676 { 677 struct lwkt_msg *lmsg = &nd6_timer_netmsg.lmsg; 678 679 KASSERT(mycpuid == 0, ("not on cpu0")); 680 crit_enter(); 681 if (lmsg->ms_flags & MSGF_DONE) 682 lwkt_sendmsg_oncpu(netisr_cpuport(0), lmsg); 683 crit_exit(); 684 } 685 686 void 687 nd6_timer_init(void) 688 { 689 callout_init_mp(&nd6_timer_ch); 690 netmsg_init(&nd6_timer_netmsg, NULL, &netisr_adone_rport, 691 MSGF_PRIORITY, nd6_timer_dispatch); 692 callout_reset_bycpu(&nd6_timer_ch, hz, nd6_timer, NULL, 0); 693 } 694 695 static int 696 regen_tmpaddr(struct in6_ifaddr *ia6) /* deprecated/invalidated temporary 697 address */ 698 { 699 struct ifaddr_container *ifac; 700 struct ifnet *ifp; 701 struct in6_ifaddr *public_ifa6 = NULL; 702 703 ifp = ia6->ia_ifa.ifa_ifp; 704 TAILQ_FOREACH(ifac, &ifp->if_addrheads[mycpuid], ifa_link) { 705 struct ifaddr *ifa = ifac->ifa; 706 struct in6_ifaddr *it6; 707 708 if (ifa->ifa_addr->sa_family != AF_INET6) 709 continue; 710 711 it6 = (struct in6_ifaddr *)ifa; 712 713 /* ignore no autoconf addresses. */ 714 if (!(it6->ia6_flags & IN6_IFF_AUTOCONF)) 715 continue; 716 717 /* ignore autoconf addresses with different prefixes. */ 718 if (it6->ia6_ndpr == NULL || it6->ia6_ndpr != ia6->ia6_ndpr) 719 continue; 720 721 /* 722 * Now we are looking at an autoconf address with the same 723 * prefix as ours. If the address is temporary and is still 724 * preferred, do not create another one. It would be rare, but 725 * could happen, for example, when we resume a laptop PC after 726 * a long period. 727 */ 728 if ((it6->ia6_flags & IN6_IFF_TEMPORARY) && 729 !IFA6_IS_DEPRECATED(it6)) { 730 public_ifa6 = NULL; 731 break; 732 } 733 734 /* 735 * This is a public autoconf address that has the same prefix 736 * as ours. If it is preferred, keep it. We can't break the 737 * loop here, because there may be a still-preferred temporary 738 * address with the prefix. 739 */ 740 if (!IFA6_IS_DEPRECATED(it6)) 741 public_ifa6 = it6; 742 } 743 744 if (public_ifa6 != NULL) { 745 int e; 746 747 if ((e = in6_tmpifadd(public_ifa6, 0)) != 0) { 748 log(LOG_NOTICE, "regen_tmpaddr: failed to create a new" 749 " tmp addr,errno=%d\n", e); 750 return (-1); 751 } 752 return (0); 753 } 754 755 return (-1); 756 } 757 758 /* 759 * Nuke neighbor cache/prefix/default router management table, right before 760 * ifp goes away. 761 */ 762 void 763 nd6_purge(struct ifnet *ifp) 764 { 765 struct llinfo_nd6 *ln, *nln; 766 struct nd_defrouter *dr, *ndr, drany; 767 struct nd_prefix *pr, *npr; 768 769 /* Nuke default router list entries toward ifp */ 770 if ((dr = TAILQ_FIRST(&nd_defrouter)) != NULL) { 771 /* 772 * The first entry of the list may be stored in 773 * the routing table, so we'll delete it later. 774 */ 775 for (dr = TAILQ_NEXT(dr, dr_entry); dr; dr = ndr) { 776 ndr = TAILQ_NEXT(dr, dr_entry); 777 if (dr->ifp == ifp) 778 defrtrlist_del(dr); 779 } 780 dr = TAILQ_FIRST(&nd_defrouter); 781 if (dr->ifp == ifp) 782 defrtrlist_del(dr); 783 } 784 785 /* Nuke prefix list entries toward ifp */ 786 for (pr = nd_prefix.lh_first; pr; pr = npr) { 787 npr = pr->ndpr_next; 788 if (pr->ndpr_ifp == ifp) { 789 /* 790 * Previously, pr->ndpr_addr is removed as well, 791 * but I strongly believe we don't have to do it. 792 * nd6_purge() is only called from in6_ifdetach(), 793 * which removes all the associated interface addresses 794 * by itself. 795 * (jinmei@kame.net 20010129) 796 */ 797 prelist_remove(pr); 798 } 799 } 800 801 /* cancel default outgoing interface setting */ 802 if (nd6_defifindex == ifp->if_index) 803 nd6_setdefaultiface(0); 804 805 if (!ip6_forwarding && ip6_accept_rtadv) { /* XXX: too restrictive? */ 806 /* refresh default router list */ 807 bzero(&drany, sizeof(drany)); 808 defrouter_delreq(&drany, 0); 809 defrouter_select(); 810 } 811 812 /* 813 * Nuke neighbor cache entries for the ifp. 814 * Note that rt->rt_ifp may not be the same as ifp, 815 * due to KAME goto ours hack. See RTM_RESOLVE case in 816 * nd6_rtrequest(), and ip6_input(). 817 */ 818 ln = llinfo_nd6.ln_next; 819 while (ln && ln != &llinfo_nd6) { 820 struct rtentry *rt; 821 struct sockaddr_dl *sdl; 822 823 nln = ln->ln_next; 824 rt = ln->ln_rt; 825 if (rt && rt->rt_gateway && 826 rt->rt_gateway->sa_family == AF_LINK) { 827 sdl = (struct sockaddr_dl *)rt->rt_gateway; 828 if (sdl->sdl_index == ifp->if_index) 829 nln = nd6_free(rt); 830 } 831 ln = nln; 832 } 833 } 834 835 struct rtentry * 836 nd6_lookup(struct in6_addr *addr6, int create, struct ifnet *ifp) 837 { 838 struct rtentry *rt; 839 struct sockaddr_in6 sin6; 840 841 bzero(&sin6, sizeof(sin6)); 842 sin6.sin6_len = sizeof(struct sockaddr_in6); 843 sin6.sin6_family = AF_INET6; 844 sin6.sin6_addr = *addr6; 845 846 if (create) 847 rt = rtlookup((struct sockaddr *)&sin6); 848 else 849 rt = rtpurelookup((struct sockaddr *)&sin6); 850 if (rt && !(rt->rt_flags & RTF_LLINFO)) { 851 /* 852 * This is the case for the default route. 853 * If we want to create a neighbor cache for the address, we 854 * should free the route for the destination and allocate an 855 * interface route. 856 */ 857 if (create) { 858 --rt->rt_refcnt; 859 rt = NULL; 860 } 861 } 862 if (!rt) { 863 if (create && ifp) { 864 int e; 865 866 /* 867 * If no route is available and create is set, 868 * we allocate a host route for the destination 869 * and treat it like an interface route. 870 * This hack is necessary for a neighbor which can't 871 * be covered by our own prefix. 872 */ 873 struct ifaddr *ifa; 874 875 ifa = ifaof_ifpforaddr((struct sockaddr *)&sin6, ifp); 876 if (ifa == NULL) 877 return (NULL); 878 879 /* 880 * Create a new route. RTF_LLINFO is necessary 881 * to create a Neighbor Cache entry for the 882 * destination in nd6_rtrequest which will be 883 * called in rtrequest via ifa->ifa_rtrequest. 884 */ 885 if ((e = rtrequest(RTM_ADD, (struct sockaddr *)&sin6, 886 ifa->ifa_addr, (struct sockaddr *)&all1_sa, 887 (ifa->ifa_flags | RTF_HOST | RTF_LLINFO) & 888 ~RTF_CLONING, &rt)) != 0) { 889 log(LOG_ERR, 890 "nd6_lookup: failed to add route for a " 891 "neighbor(%s), errno=%d\n", 892 ip6_sprintf(addr6), e); 893 } 894 if (rt == NULL) 895 return (NULL); 896 if (rt->rt_llinfo) { 897 struct llinfo_nd6 *ln = 898 (struct llinfo_nd6 *)rt->rt_llinfo; 899 900 ln->ln_state = ND6_LLINFO_NOSTATE; 901 } 902 } else 903 return (NULL); 904 } 905 rt->rt_refcnt--; 906 907 if (!ND6_RTENTRY_IS_NEIGHBOR(rt, ifp)) { 908 if (create) { 909 log(LOG_DEBUG, 910 "nd6_lookup: failed to lookup %s (if = %s)\n", 911 ip6_sprintf(addr6), ifp ? if_name(ifp) : "unspec"); 912 /* xxx more logs... kazu */ 913 } 914 return (NULL); 915 } 916 return (rt); 917 } 918 919 static struct rtentry * 920 nd6_neighbor_lookup(struct in6_addr *addr6, struct ifnet *ifp) 921 { 922 struct rtentry *rt; 923 struct sockaddr_in6 sin6; 924 925 bzero(&sin6, sizeof(sin6)); 926 sin6.sin6_len = sizeof(struct sockaddr_in6); 927 sin6.sin6_family = AF_INET6; 928 sin6.sin6_addr = *addr6; 929 930 rt = rtpurelookup((struct sockaddr *)&sin6); 931 if (rt == NULL) 932 return (NULL); 933 rt->rt_refcnt--; 934 935 if (!ND6_RTENTRY_IS_NEIGHBOR(rt, ifp)) { 936 if (nd6_onlink_ns_rfc4861 && 937 (ND6_RTENTRY_IS_LLCLONING(rt) || /* not cloned yet */ 938 (rt->rt_parent != NULL && /* cloning */ 939 ND6_RTENTRY_IS_LLCLONING(rt->rt_parent)))) { 940 /* 941 * If cloning ever happened or is happening, 942 * rtentry for addr6 would or will become a 943 * neighbor cache. 944 */ 945 } else { 946 rt = NULL; 947 } 948 } 949 return (rt); 950 } 951 952 /* 953 * Detect if a given IPv6 address identifies a neighbor on a given link. 954 * XXX: should take care of the destination of a p2p link? 955 */ 956 int 957 nd6_is_addr_neighbor(struct sockaddr_in6 *addr, struct ifnet *ifp) 958 { 959 struct ifaddr_container *ifac; 960 int i; 961 962 #define IFADDR6(a) ((((struct in6_ifaddr *)(a))->ia_addr).sin6_addr) 963 #define IFMASK6(a) ((((struct in6_ifaddr *)(a))->ia_prefixmask).sin6_addr) 964 965 /* 966 * A link-local address is always a neighbor. 967 * XXX: we should use the sin6_scope_id field rather than the embedded 968 * interface index. 969 */ 970 if (IN6_IS_ADDR_LINKLOCAL(&addr->sin6_addr) && 971 ntohs(*(u_int16_t *)&addr->sin6_addr.s6_addr[2]) == ifp->if_index) 972 return (1); 973 974 /* 975 * If the address matches one of our addresses, 976 * it should be a neighbor. 977 */ 978 TAILQ_FOREACH(ifac, &ifp->if_addrheads[mycpuid], ifa_link) { 979 struct ifaddr *ifa = ifac->ifa; 980 981 if (ifa->ifa_addr->sa_family != AF_INET6) 982 next: continue; 983 984 for (i = 0; i < 4; i++) { 985 if ((IFADDR6(ifa).s6_addr32[i] ^ 986 addr->sin6_addr.s6_addr32[i]) & 987 IFMASK6(ifa).s6_addr32[i]) 988 goto next; 989 } 990 return (1); 991 } 992 993 /* 994 * Even if the address matches none of our addresses, it might be 995 * in the neighbor cache. 996 */ 997 if (nd6_neighbor_lookup(&addr->sin6_addr, ifp) != NULL) 998 return (1); 999 1000 return (0); 1001 #undef IFADDR6 1002 #undef IFMASK6 1003 } 1004 1005 /* 1006 * Free an nd6 llinfo entry. 1007 */ 1008 struct llinfo_nd6 * 1009 nd6_free(struct rtentry *rt) 1010 { 1011 struct llinfo_nd6 *ln = (struct llinfo_nd6 *)rt->rt_llinfo, *next; 1012 struct in6_addr in6 = ((struct sockaddr_in6 *)rt_key(rt))->sin6_addr; 1013 struct nd_defrouter *dr; 1014 1015 /* 1016 * we used to have kpfctlinput(PRC_HOSTDEAD) here. 1017 * even though it is not harmful, it was not really necessary. 1018 */ 1019 1020 if (!ip6_forwarding && ip6_accept_rtadv) { /* XXX: too restrictive? */ 1021 mtx_lock(&nd6_mtx); 1022 dr = defrouter_lookup( 1023 &((struct sockaddr_in6 *)rt_key(rt))->sin6_addr, 1024 rt->rt_ifp); 1025 1026 if (ln->ln_router || dr) { 1027 /* 1028 * rt6_flush must be called whether or not the neighbor 1029 * is in the Default Router List. 1030 * See a corresponding comment in nd6_na_input(). 1031 */ 1032 rt6_flush(&in6, rt->rt_ifp); 1033 } 1034 1035 if (dr) { 1036 /* 1037 * Unreachablity of a router might affect the default 1038 * router selection and on-link detection of advertised 1039 * prefixes. 1040 */ 1041 1042 /* 1043 * Temporarily fake the state to choose a new default 1044 * router and to perform on-link determination of 1045 * prefixes correctly. 1046 * Below the state will be set correctly, 1047 * or the entry itself will be deleted. 1048 */ 1049 ln->ln_state = ND6_LLINFO_INCOMPLETE; 1050 1051 /* 1052 * Since defrouter_select() does not affect the 1053 * on-link determination and MIP6 needs the check 1054 * before the default router selection, we perform 1055 * the check now. 1056 */ 1057 pfxlist_onlink_check(); 1058 1059 if (dr == TAILQ_FIRST(&nd_defrouter)) { 1060 /* 1061 * It is used as the current default router, 1062 * so we have to move it to the end of the 1063 * list and choose a new one. 1064 * XXX: it is not very efficient if this is 1065 * the only router. 1066 */ 1067 TAILQ_REMOVE(&nd_defrouter, dr, dr_entry); 1068 TAILQ_INSERT_TAIL(&nd_defrouter, dr, dr_entry); 1069 1070 defrouter_select(); 1071 } 1072 } 1073 mtx_unlock(&nd6_mtx); 1074 } 1075 1076 /* 1077 * Before deleting the entry, remember the next entry as the 1078 * return value. We need this because pfxlist_onlink_check() above 1079 * might have freed other entries (particularly the old next entry) as 1080 * a side effect (XXX). 1081 */ 1082 next = ln->ln_next; 1083 1084 /* 1085 * Detach the route from the routing tree and the list of neighbor 1086 * caches, and disable the route entry not to be used in already 1087 * cached routes. 1088 * 1089 * ND expiry happens under one big timer. 1090 * To avoid overflowing the route socket, don't report this. 1091 * Now that RTM_MISS is reported when an address is unresolvable 1092 * the benefit of reporting this deletion is questionable. 1093 */ 1094 rtrequest(RTM_DELETE, rt_key(rt), NULL, rt_mask(rt), 0, NULL); 1095 1096 return (next); 1097 } 1098 1099 /* 1100 * Upper-layer reachability hint for Neighbor Unreachability Detection. 1101 * 1102 * XXX cost-effective metods? 1103 */ 1104 void 1105 nd6_nud_hint(struct rtentry *rt, struct in6_addr *dst6, int force) 1106 { 1107 struct llinfo_nd6 *ln; 1108 1109 /* 1110 * If the caller specified "rt", use that. Otherwise, resolve the 1111 * routing table by supplied "dst6". 1112 */ 1113 if (!rt) { 1114 if (!dst6) 1115 return; 1116 if (!(rt = nd6_lookup(dst6, 0, NULL))) 1117 return; 1118 } 1119 1120 if ((rt->rt_flags & RTF_GATEWAY) || 1121 !(rt->rt_flags & RTF_LLINFO) || 1122 rt->rt_llinfo == NULL || rt->rt_gateway == NULL || 1123 rt->rt_gateway->sa_family != AF_LINK) { 1124 /* This is not a host route. */ 1125 return; 1126 } 1127 1128 ln = (struct llinfo_nd6 *)rt->rt_llinfo; 1129 if (ln->ln_state < ND6_LLINFO_REACHABLE) 1130 return; 1131 1132 /* 1133 * if we get upper-layer reachability confirmation many times, 1134 * it is possible we have false information. 1135 */ 1136 if (!force) { 1137 ln->ln_byhint++; 1138 if (ln->ln_byhint > nd6_maxnudhint) 1139 return; 1140 } 1141 1142 ln->ln_state = ND6_LLINFO_REACHABLE; 1143 if (ln->ln_expire) 1144 ln->ln_expire = time_uptime + 1145 ND_IFINFO(rt->rt_ifp)->reachable; 1146 } 1147 1148 void 1149 nd6_rtrequest(int req, struct rtentry *rt) 1150 { 1151 struct sockaddr *gate = rt->rt_gateway; 1152 struct llinfo_nd6 *ln = (struct llinfo_nd6 *)rt->rt_llinfo; 1153 static struct sockaddr_dl null_sdl = {sizeof(null_sdl), AF_LINK}; 1154 struct ifnet *ifp = rt->rt_ifp; 1155 struct ifaddr *ifa; 1156 1157 if ((rt->rt_flags & RTF_GATEWAY)) 1158 return; 1159 1160 if (nd6_need_cache(ifp) == 0 && !(rt->rt_flags & RTF_HOST)) { 1161 /* 1162 * This is probably an interface direct route for a link 1163 * which does not need neighbor caches (e.g. fe80::%lo0/64). 1164 * We do not need special treatment below for such a route. 1165 * Moreover, the RTF_LLINFO flag which would be set below 1166 * would annoy the ndp(8) command. 1167 */ 1168 return; 1169 } 1170 1171 if (req == RTM_RESOLVE && 1172 (nd6_need_cache(ifp) == 0 || /* stf case */ 1173 !nd6_is_addr_neighbor((struct sockaddr_in6 *)rt_key(rt), ifp))) { 1174 /* 1175 * FreeBSD and BSD/OS often make a cloned host route based 1176 * on a less-specific route (e.g. the default route). 1177 * If the less specific route does not have a "gateway" 1178 * (this is the case when the route just goes to a p2p or an 1179 * stf interface), we'll mistakenly make a neighbor cache for 1180 * the host route, and will see strange neighbor solicitation 1181 * for the corresponding destination. In order to avoid the 1182 * confusion, we check if the destination of the route is 1183 * a neighbor in terms of neighbor discovery, and stop the 1184 * process if not. Additionally, we remove the LLINFO flag 1185 * so that ndp(8) will not try to get the neighbor information 1186 * of the destination. 1187 */ 1188 rt->rt_flags &= ~RTF_LLINFO; 1189 return; 1190 } 1191 1192 switch (req) { 1193 case RTM_ADD: 1194 /* 1195 * There is no backward compatibility :) 1196 * 1197 * if (!(rt->rt_flags & RTF_HOST) && 1198 * SIN(rt_mask(rt))->sin_addr.s_addr != 0xffffffff) 1199 * rt->rt_flags |= RTF_CLONING; 1200 */ 1201 if (rt->rt_flags & (RTF_CLONING | RTF_LLINFO)) { 1202 /* 1203 * Case 1: This route should come from 1204 * a route to interface. RTF_LLINFO flag is set 1205 * for a host route whose destination should be 1206 * treated as on-link. 1207 */ 1208 rt_setgate(rt, rt_key(rt), 1209 (struct sockaddr *)&null_sdl); 1210 gate = rt->rt_gateway; 1211 SDL(gate)->sdl_type = ifp->if_type; 1212 SDL(gate)->sdl_index = ifp->if_index; 1213 if (ln) 1214 ln->ln_expire = time_uptime; 1215 if (ln && ln->ln_expire == 0) { 1216 /* kludge for desktops */ 1217 ln->ln_expire = 1; 1218 } 1219 if ((rt->rt_flags & RTF_CLONING)) 1220 break; 1221 } 1222 /* 1223 * In IPv4 code, we try to annonuce new RTF_ANNOUNCE entry here. 1224 * We don't do that here since llinfo is not ready yet. 1225 * 1226 * There are also couple of other things to be discussed: 1227 * - unsolicited NA code needs improvement beforehand 1228 * - RFC2461 says we MAY send multicast unsolicited NA 1229 * (7.2.6 paragraph 4), however, it also says that we 1230 * SHOULD provide a mechanism to prevent multicast NA storm. 1231 * we don't have anything like it right now. 1232 * note that the mechanism needs a mutual agreement 1233 * between proxies, which means that we need to implement 1234 * a new protocol, or a new kludge. 1235 * - from RFC2461 6.2.4, host MUST NOT send an unsolicited NA. 1236 * we need to check ip6forwarding before sending it. 1237 * (or should we allow proxy ND configuration only for 1238 * routers? there's no mention about proxy ND from hosts) 1239 */ 1240 #if 0 1241 /* XXX it does not work */ 1242 if ((rt->rt_flags & RTF_ANNOUNCE) && mycpuid == 0) { 1243 nd6_na_output(ifp, 1244 &SIN6(rt_key(rt))->sin6_addr, 1245 &SIN6(rt_key(rt))->sin6_addr, 1246 ip6_forwarding ? ND_NA_FLAG_ROUTER : 0, 1247 1, NULL); 1248 } 1249 #endif 1250 /* FALLTHROUGH */ 1251 case RTM_RESOLVE: 1252 if ((ifp->if_flags & (IFF_POINTOPOINT | IFF_LOOPBACK)) == 0) { 1253 /* 1254 * Address resolution isn't necessary for a point to 1255 * point link, so we can skip this test for a p2p link. 1256 */ 1257 if (gate->sa_family != AF_LINK || 1258 gate->sa_len < sizeof(null_sdl)) { 1259 log(LOG_DEBUG, 1260 "nd6_rtrequest: bad gateway value: %s\n", 1261 if_name(ifp)); 1262 break; 1263 } 1264 SDL(gate)->sdl_type = ifp->if_type; 1265 SDL(gate)->sdl_index = ifp->if_index; 1266 } 1267 if (ln != NULL) 1268 break; /* This happens on a route change */ 1269 /* 1270 * Case 2: This route may come from cloning, or a manual route 1271 * add with a LL address. 1272 */ 1273 R_Malloc(ln, struct llinfo_nd6 *, sizeof(*ln)); 1274 rt->rt_llinfo = (caddr_t)ln; 1275 if (!ln) { 1276 log(LOG_DEBUG, "nd6_rtrequest: malloc failed\n"); 1277 break; 1278 } 1279 nd6_inuse++; 1280 nd6_allocated++; 1281 bzero(ln, sizeof(*ln)); 1282 ln->ln_rt = rt; 1283 /* this is required for "ndp" command. - shin */ 1284 if (req == RTM_ADD) { 1285 /* 1286 * gate should have some valid AF_LINK entry, 1287 * and ln->ln_expire should have some lifetime 1288 * which is specified by ndp command. 1289 */ 1290 ln->ln_state = ND6_LLINFO_REACHABLE; 1291 ln->ln_byhint = 0; 1292 } else { 1293 /* 1294 * When req == RTM_RESOLVE, rt is created and 1295 * initialized in rtrequest(), so rt_expire is 0. 1296 */ 1297 ln->ln_state = ND6_LLINFO_NOSTATE; 1298 ln->ln_expire = time_uptime; 1299 } 1300 rt->rt_flags |= RTF_LLINFO; 1301 ln->ln_next = llinfo_nd6.ln_next; 1302 llinfo_nd6.ln_next = ln; 1303 ln->ln_prev = &llinfo_nd6; 1304 ln->ln_next->ln_prev = ln; 1305 1306 /* 1307 * check if rt_key(rt) is one of my address assigned 1308 * to the interface. 1309 */ 1310 ifa = (struct ifaddr *)in6ifa_ifpwithaddr(rt->rt_ifp, 1311 &SIN6(rt_key(rt))->sin6_addr); 1312 if (ifa) { 1313 caddr_t macp = nd6_ifptomac(ifp); 1314 ln->ln_expire = 0; 1315 ln->ln_state = ND6_LLINFO_REACHABLE; 1316 ln->ln_byhint = 0; 1317 if (macp) { 1318 bcopy(macp, LLADDR(SDL(gate)), ifp->if_addrlen); 1319 SDL(gate)->sdl_alen = ifp->if_addrlen; 1320 } 1321 if (nd6_useloopback) { 1322 rt->rt_ifp = loif; /* XXX */ 1323 /* 1324 * Make sure rt_ifa be equal to the ifaddr 1325 * corresponding to the address. 1326 * We need this because when we refer 1327 * rt_ifa->ia6_flags in ip6_input, we assume 1328 * that the rt_ifa points to the address instead 1329 * of the loopback address. 1330 */ 1331 if (ifa != rt->rt_ifa) { 1332 IFAFREE(rt->rt_ifa); 1333 IFAREF(ifa); 1334 rt->rt_ifa = ifa; 1335 } 1336 } 1337 } else if (rt->rt_flags & RTF_ANNOUNCE) { 1338 ln->ln_expire = 0; 1339 ln->ln_state = ND6_LLINFO_REACHABLE; 1340 ln->ln_byhint = 0; 1341 1342 /* 1343 * Join solicited node multicast for proxy ND, and only 1344 * join it once on cpu0. 1345 */ 1346 if ((ifp->if_flags & IFF_MULTICAST) && mycpuid == 0) { 1347 struct in6_addr llsol; 1348 int error; 1349 1350 llsol = SIN6(rt_key(rt))->sin6_addr; 1351 llsol.s6_addr16[0] = htons(0xff02); 1352 llsol.s6_addr16[1] = htons(ifp->if_index); 1353 llsol.s6_addr32[1] = 0; 1354 llsol.s6_addr32[2] = htonl(1); 1355 llsol.s6_addr8[12] = 0xff; 1356 1357 if (!in6_addmulti(&llsol, ifp, &error)) { 1358 nd6log((LOG_ERR, "%s: failed to join " 1359 "%s (errno=%d)\n", if_name(ifp), 1360 ip6_sprintf(&llsol), error)); 1361 } 1362 } 1363 } 1364 break; 1365 1366 case RTM_DELETE: 1367 if (!ln) 1368 break; 1369 /* 1370 * Leave from solicited node multicast for proxy ND, and only 1371 * leave it once on cpu0 (since we joined it once on cpu0). 1372 */ 1373 if ((rt->rt_flags & RTF_ANNOUNCE) && 1374 (ifp->if_flags & IFF_MULTICAST) && mycpuid == 0) { 1375 struct in6_addr llsol; 1376 struct in6_multi *in6m; 1377 1378 llsol = SIN6(rt_key(rt))->sin6_addr; 1379 llsol.s6_addr16[0] = htons(0xff02); 1380 llsol.s6_addr16[1] = htons(ifp->if_index); 1381 llsol.s6_addr32[1] = 0; 1382 llsol.s6_addr32[2] = htonl(1); 1383 llsol.s6_addr8[12] = 0xff; 1384 1385 in6m = IN6_LOOKUP_MULTI(&llsol, ifp); 1386 if (in6m) 1387 in6_delmulti(in6m); 1388 } 1389 nd6_inuse--; 1390 ln->ln_next->ln_prev = ln->ln_prev; 1391 ln->ln_prev->ln_next = ln->ln_next; 1392 ln->ln_prev = NULL; 1393 rt->rt_llinfo = 0; 1394 rt->rt_flags &= ~RTF_LLINFO; 1395 if (ln->ln_hold) 1396 m_freem(ln->ln_hold); 1397 Free((caddr_t)ln); 1398 } 1399 } 1400 1401 int 1402 nd6_ioctl(u_long cmd, caddr_t data, struct ifnet *ifp) 1403 { 1404 struct in6_drlist *drl = (struct in6_drlist *)data; 1405 struct in6_prlist *prl = (struct in6_prlist *)data; 1406 struct in6_ndireq *ndi = (struct in6_ndireq *)data; 1407 struct in6_nbrinfo *nbi = (struct in6_nbrinfo *)data; 1408 struct in6_ndifreq *ndif = (struct in6_ndifreq *)data; 1409 struct nd_defrouter *dr, any; 1410 struct nd_prefix *pr; 1411 struct rtentry *rt; 1412 int i = 0, error = 0; 1413 1414 switch (cmd) { 1415 case SIOCGDRLST_IN6: 1416 /* 1417 * obsolete API, use sysctl under net.inet6.icmp6 1418 */ 1419 bzero(drl, sizeof(*drl)); 1420 mtx_lock(&nd6_mtx); 1421 dr = TAILQ_FIRST(&nd_defrouter); 1422 while (dr && i < DRLSTSIZ) { 1423 drl->defrouter[i].rtaddr = dr->rtaddr; 1424 if (IN6_IS_ADDR_LINKLOCAL(&drl->defrouter[i].rtaddr)) { 1425 /* XXX: need to this hack for KAME stack */ 1426 drl->defrouter[i].rtaddr.s6_addr16[1] = 0; 1427 } else 1428 log(LOG_ERR, 1429 "default router list contains a " 1430 "non-linklocal address(%s)\n", 1431 ip6_sprintf(&drl->defrouter[i].rtaddr)); 1432 1433 drl->defrouter[i].flags = dr->flags; 1434 drl->defrouter[i].rtlifetime = dr->rtlifetime; 1435 drl->defrouter[i].expire = dr->expire; 1436 drl->defrouter[i].if_index = dr->ifp->if_index; 1437 i++; 1438 dr = TAILQ_NEXT(dr, dr_entry); 1439 } 1440 mtx_unlock(&nd6_mtx); 1441 break; 1442 case SIOCGPRLST_IN6: 1443 /* 1444 * obsolete API, use sysctl under net.inet6.icmp6 1445 */ 1446 /* 1447 * XXX meaning of fields, especialy "raflags", is very 1448 * differnet between RA prefix list and RR/static prefix list. 1449 * how about separating ioctls into two? 1450 */ 1451 bzero(prl, sizeof(*prl)); 1452 mtx_lock(&nd6_mtx); 1453 pr = nd_prefix.lh_first; 1454 while (pr && i < PRLSTSIZ) { 1455 struct nd_pfxrouter *pfr; 1456 int j; 1457 1458 in6_embedscope(&prl->prefix[i].prefix, 1459 &pr->ndpr_prefix, NULL, NULL); 1460 prl->prefix[i].raflags = pr->ndpr_raf; 1461 prl->prefix[i].prefixlen = pr->ndpr_plen; 1462 prl->prefix[i].vltime = pr->ndpr_vltime; 1463 prl->prefix[i].pltime = pr->ndpr_pltime; 1464 prl->prefix[i].if_index = pr->ndpr_ifp->if_index; 1465 prl->prefix[i].expire = pr->ndpr_expire; 1466 1467 pfr = pr->ndpr_advrtrs.lh_first; 1468 j = 0; 1469 while (pfr) { 1470 if (j < DRLSTSIZ) { 1471 #define RTRADDR prl->prefix[i].advrtr[j] 1472 RTRADDR = pfr->router->rtaddr; 1473 if (IN6_IS_ADDR_LINKLOCAL(&RTRADDR)) { 1474 /* XXX: hack for KAME */ 1475 RTRADDR.s6_addr16[1] = 0; 1476 } else 1477 log(LOG_ERR, 1478 "a router(%s) advertises " 1479 "a prefix with " 1480 "non-link local address\n", 1481 ip6_sprintf(&RTRADDR)); 1482 #undef RTRADDR 1483 } 1484 j++; 1485 pfr = pfr->pfr_next; 1486 } 1487 prl->prefix[i].advrtrs = j; 1488 prl->prefix[i].origin = PR_ORIG_RA; 1489 1490 i++; 1491 pr = pr->ndpr_next; 1492 } 1493 mtx_unlock(&nd6_mtx); 1494 1495 break; 1496 case OSIOCGIFINFO_IN6: 1497 /* XXX: old ndp(8) assumes a positive value for linkmtu. */ 1498 bzero(&ndi->ndi, sizeof(ndi->ndi)); 1499 ndi->ndi.linkmtu = IN6_LINKMTU(ifp); 1500 ndi->ndi.maxmtu = ND_IFINFO(ifp)->maxmtu; 1501 ndi->ndi.basereachable = ND_IFINFO(ifp)->basereachable; 1502 ndi->ndi.reachable = ND_IFINFO(ifp)->reachable; 1503 ndi->ndi.retrans = ND_IFINFO(ifp)->retrans; 1504 ndi->ndi.flags = ND_IFINFO(ifp)->flags; 1505 ndi->ndi.recalctm = ND_IFINFO(ifp)->recalctm; 1506 ndi->ndi.chlim = ND_IFINFO(ifp)->chlim; 1507 break; 1508 case SIOCGIFINFO_IN6: 1509 ndi->ndi = *ND_IFINFO(ifp); 1510 ndi->ndi.linkmtu = IN6_LINKMTU(ifp); 1511 break; 1512 case SIOCSIFINFO_IN6: 1513 /* 1514 * used to change host variables from userland. 1515 * intented for a use on router to reflect RA configurations. 1516 */ 1517 /* 0 means 'unspecified' */ 1518 if (ndi->ndi.linkmtu != 0) { 1519 if (ndi->ndi.linkmtu < IPV6_MMTU || 1520 ndi->ndi.linkmtu > IN6_LINKMTU(ifp)) { 1521 error = EINVAL; 1522 break; 1523 } 1524 ND_IFINFO(ifp)->linkmtu = ndi->ndi.linkmtu; 1525 } 1526 1527 if (ndi->ndi.basereachable != 0) { 1528 int obasereachable = ND_IFINFO(ifp)->basereachable; 1529 1530 ND_IFINFO(ifp)->basereachable = ndi->ndi.basereachable; 1531 if (ndi->ndi.basereachable != obasereachable) 1532 ND_IFINFO(ifp)->reachable = 1533 ND_COMPUTE_RTIME(ndi->ndi.basereachable); 1534 } 1535 if (ndi->ndi.retrans != 0) 1536 ND_IFINFO(ifp)->retrans = ndi->ndi.retrans; 1537 if (ndi->ndi.chlim != 0) 1538 ND_IFINFO(ifp)->chlim = ndi->ndi.chlim; 1539 /* FALLTHROUGH */ 1540 case SIOCSIFINFO_FLAGS: 1541 ND_IFINFO(ifp)->flags = ndi->ndi.flags; 1542 break; 1543 case SIOCSNDFLUSH_IN6: /* XXX: the ioctl name is confusing... */ 1544 /* flush default router list */ 1545 /* 1546 * xxx sumikawa: should not delete route if default 1547 * route equals to the top of default router list 1548 */ 1549 bzero(&any, sizeof(any)); 1550 defrouter_delreq(&any, 0); 1551 defrouter_select(); 1552 /* xxx sumikawa: flush prefix list */ 1553 break; 1554 case SIOCSPFXFLUSH_IN6: 1555 { 1556 /* flush all the prefix advertised by routers */ 1557 struct nd_prefix *pr, *next; 1558 1559 mtx_lock(&nd6_mtx); 1560 for (pr = nd_prefix.lh_first; pr; pr = next) { 1561 struct in6_ifaddr *ia, *ia_next; 1562 1563 next = pr->ndpr_next; 1564 1565 if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr)) 1566 continue; /* XXX */ 1567 1568 /* do we really have to remove addresses as well? */ 1569 for (ia = in6_ifaddr; ia; ia = ia_next) { 1570 /* ia might be removed. keep the next ptr. */ 1571 ia_next = ia->ia_next; 1572 1573 if (!(ia->ia6_flags & IN6_IFF_AUTOCONF)) 1574 continue; 1575 1576 if (ia->ia6_ndpr == pr) 1577 in6_purgeaddr(&ia->ia_ifa); 1578 } 1579 prelist_remove(pr); 1580 } 1581 mtx_unlock(&nd6_mtx); 1582 break; 1583 } 1584 case SIOCSRTRFLUSH_IN6: 1585 { 1586 /* flush all the default routers */ 1587 struct nd_defrouter *dr, *next; 1588 1589 mtx_lock(&nd6_mtx); 1590 if ((dr = TAILQ_FIRST(&nd_defrouter)) != NULL) { 1591 /* 1592 * The first entry of the list may be stored in 1593 * the routing table, so we'll delete it later. 1594 */ 1595 for (dr = TAILQ_NEXT(dr, dr_entry); dr; dr = next) { 1596 next = TAILQ_NEXT(dr, dr_entry); 1597 defrtrlist_del(dr); 1598 } 1599 defrtrlist_del(TAILQ_FIRST(&nd_defrouter)); 1600 } 1601 mtx_unlock(&nd6_mtx); 1602 break; 1603 } 1604 case SIOCGNBRINFO_IN6: 1605 { 1606 struct llinfo_nd6 *ln; 1607 struct in6_addr nb_addr = nbi->addr; /* make local for safety */ 1608 1609 /* 1610 * XXX: KAME specific hack for scoped addresses 1611 * XXXX: for other scopes than link-local? 1612 */ 1613 if (IN6_IS_ADDR_LINKLOCAL(&nbi->addr) || 1614 IN6_IS_ADDR_MC_LINKLOCAL(&nbi->addr)) { 1615 u_int16_t *idp = (u_int16_t *)&nb_addr.s6_addr[2]; 1616 1617 if (*idp == 0) 1618 *idp = htons(ifp->if_index); 1619 } 1620 1621 mtx_lock(&nd6_mtx); 1622 if ((rt = nd6_lookup(&nb_addr, 0, ifp)) == NULL) { 1623 error = EINVAL; 1624 mtx_unlock(&nd6_mtx); 1625 break; 1626 } 1627 ln = (struct llinfo_nd6 *)rt->rt_llinfo; 1628 nbi->state = ln->ln_state; 1629 nbi->asked = ln->ln_asked; 1630 nbi->isrouter = ln->ln_router; 1631 nbi->expire = ln->ln_expire; 1632 mtx_unlock(&nd6_mtx); 1633 1634 break; 1635 } 1636 case SIOCGDEFIFACE_IN6: /* XXX: should be implemented as a sysctl? */ 1637 ndif->ifindex = nd6_defifindex; 1638 break; 1639 case SIOCSDEFIFACE_IN6: /* XXX: should be implemented as a sysctl? */ 1640 return (nd6_setdefaultiface(ndif->ifindex)); 1641 } 1642 return (error); 1643 } 1644 1645 /* 1646 * Create neighbor cache entry and cache link-layer address, 1647 * on reception of inbound ND6 packets. (RS/RA/NS/redirect) 1648 */ 1649 struct rtentry * 1650 nd6_cache_lladdr(struct ifnet *ifp, struct in6_addr *from, char *lladdr, 1651 int lladdrlen, 1652 int type, /* ICMP6 type */ 1653 int code /* type dependent information */) 1654 { 1655 struct rtentry *rt = NULL; 1656 struct llinfo_nd6 *ln = NULL; 1657 int is_newentry; 1658 struct sockaddr_dl *sdl = NULL; 1659 int do_update; 1660 int olladdr; 1661 int llchange; 1662 int newstate = 0; 1663 1664 if (!ifp) 1665 panic("ifp == NULL in nd6_cache_lladdr"); 1666 if (!from) 1667 panic("from == NULL in nd6_cache_lladdr"); 1668 1669 /* nothing must be updated for unspecified address */ 1670 if (IN6_IS_ADDR_UNSPECIFIED(from)) 1671 return NULL; 1672 1673 /* 1674 * Validation about ifp->if_addrlen and lladdrlen must be done in 1675 * the caller. 1676 * 1677 * XXX If the link does not have link-layer adderss, what should 1678 * we do? (ifp->if_addrlen == 0) 1679 * Spec says nothing in sections for RA, RS and NA. There's small 1680 * description on it in NS section (RFC 2461 7.2.3). 1681 */ 1682 1683 rt = nd6_lookup(from, 0, ifp); 1684 if (!rt) { 1685 #if 0 1686 /* nothing must be done if there's no lladdr */ 1687 if (!lladdr || !lladdrlen) 1688 return NULL; 1689 #endif 1690 1691 rt = nd6_lookup(from, 1, ifp); 1692 is_newentry = 1; 1693 } else { 1694 /* do nothing if static ndp is set */ 1695 if (rt->rt_flags & RTF_STATIC) 1696 return NULL; 1697 is_newentry = 0; 1698 } 1699 1700 if (!rt) 1701 return NULL; 1702 if ((rt->rt_flags & (RTF_GATEWAY | RTF_LLINFO)) != RTF_LLINFO) { 1703 fail: 1704 nd6_free(rt); 1705 return NULL; 1706 } 1707 ln = (struct llinfo_nd6 *)rt->rt_llinfo; 1708 if (!ln) 1709 goto fail; 1710 if (!rt->rt_gateway) 1711 goto fail; 1712 if (rt->rt_gateway->sa_family != AF_LINK) 1713 goto fail; 1714 sdl = SDL(rt->rt_gateway); 1715 1716 olladdr = (sdl->sdl_alen) ? 1 : 0; 1717 if (olladdr && lladdr) { 1718 if (bcmp(lladdr, LLADDR(sdl), ifp->if_addrlen)) 1719 llchange = 1; 1720 else 1721 llchange = 0; 1722 } else 1723 llchange = 0; 1724 1725 /* 1726 * newentry olladdr lladdr llchange (*=record) 1727 * 0 n n -- (1) 1728 * 0 y n -- (2) 1729 * 0 n y -- (3) * STALE 1730 * 0 y y n (4) * 1731 * 0 y y y (5) * STALE 1732 * 1 -- n -- (6) NOSTATE(= PASSIVE) 1733 * 1 -- y -- (7) * STALE 1734 */ 1735 1736 if (lladdr) { /* (3-5) and (7) */ 1737 /* 1738 * Record source link-layer address 1739 * XXX is it dependent to ifp->if_type? 1740 */ 1741 sdl->sdl_alen = ifp->if_addrlen; 1742 bcopy(lladdr, LLADDR(sdl), ifp->if_addrlen); 1743 } 1744 1745 if (!is_newentry) { 1746 if ((!olladdr && lladdr) || /* (3) */ 1747 (olladdr && lladdr && llchange)) { /* (5) */ 1748 do_update = 1; 1749 newstate = ND6_LLINFO_STALE; 1750 } else { /* (1-2,4) */ 1751 do_update = 0; 1752 } 1753 } else { 1754 do_update = 1; 1755 if (!lladdr) /* (6) */ 1756 newstate = ND6_LLINFO_NOSTATE; 1757 else /* (7) */ 1758 newstate = ND6_LLINFO_STALE; 1759 } 1760 1761 if (do_update) { 1762 /* 1763 * Update the state of the neighbor cache. 1764 */ 1765 ln->ln_state = newstate; 1766 1767 if (ln->ln_state == ND6_LLINFO_STALE) { 1768 /* 1769 * XXX: since nd6_output() below will cause 1770 * state tansition to DELAY and reset the timer, 1771 * we must set the timer now, although it is actually 1772 * meaningless. 1773 */ 1774 ln->ln_expire = time_uptime + nd6_gctimer; 1775 1776 if (ln->ln_hold) { 1777 /* 1778 * we assume ifp is not a p2p here, so just 1779 * set the 2nd argument as the 1st one. 1780 */ 1781 nd6_output(ifp, ifp, ln->ln_hold, 1782 (struct sockaddr_in6 *)rt_key(rt), rt); 1783 ln->ln_hold = NULL; 1784 } 1785 } else if (ln->ln_state == ND6_LLINFO_INCOMPLETE) { 1786 /* probe right away */ 1787 ln->ln_expire = time_uptime; 1788 } 1789 } 1790 1791 /* 1792 * ICMP6 type dependent behavior. 1793 * 1794 * NS: clear IsRouter if new entry 1795 * RS: clear IsRouter 1796 * RA: set IsRouter if there's lladdr 1797 * redir: clear IsRouter if new entry 1798 * 1799 * RA case, (1): 1800 * The spec says that we must set IsRouter in the following cases: 1801 * - If lladdr exist, set IsRouter. This means (1-5). 1802 * - If it is old entry (!newentry), set IsRouter. This means (7). 1803 * So, based on the spec, in (1-5) and (7) cases we must set IsRouter. 1804 * A quetion arises for (1) case. (1) case has no lladdr in the 1805 * neighbor cache, this is similar to (6). 1806 * This case is rare but we figured that we MUST NOT set IsRouter. 1807 * 1808 * newentry olladdr lladdr llchange NS RS RA redir 1809 * D R 1810 * 0 n n -- (1) c ? s 1811 * 0 y n -- (2) c s s 1812 * 0 n y -- (3) c s s 1813 * 0 y y n (4) c s s 1814 * 0 y y y (5) c s s 1815 * 1 -- n -- (6) c c c s 1816 * 1 -- y -- (7) c c s c s 1817 * 1818 * (c=clear s=set) 1819 */ 1820 switch (type & 0xff) { 1821 case ND_NEIGHBOR_SOLICIT: 1822 /* 1823 * New entry must have is_router flag cleared. 1824 */ 1825 if (is_newentry) /* (6-7) */ 1826 ln->ln_router = 0; 1827 break; 1828 case ND_REDIRECT: 1829 /* 1830 * If the icmp is a redirect to a better router, always set the 1831 * is_router flag. Otherwise, if the entry is newly created, 1832 * clear the flag. [RFC 2461, sec 8.3] 1833 */ 1834 if (code == ND_REDIRECT_ROUTER) 1835 ln->ln_router = 1; 1836 else if (is_newentry) /* (6-7) */ 1837 ln->ln_router = 0; 1838 break; 1839 case ND_ROUTER_SOLICIT: 1840 /* 1841 * is_router flag must always be cleared. 1842 */ 1843 ln->ln_router = 0; 1844 break; 1845 case ND_ROUTER_ADVERT: 1846 /* 1847 * Mark an entry with lladdr as a router. 1848 */ 1849 if ((!is_newentry && (olladdr || lladdr)) || /* (2-5) */ 1850 (is_newentry && lladdr)) { /* (7) */ 1851 ln->ln_router = 1; 1852 } 1853 break; 1854 } 1855 1856 if (llchange || lladdr) 1857 rt_rtmsg(llchange ? RTM_CHANGE : RTM_ADD, rt, rt->rt_ifp, 0); 1858 1859 /* 1860 * When the link-layer address of a router changes, select the 1861 * best router again. In particular, when the neighbor entry is newly 1862 * created, it might affect the selection policy. 1863 * Question: can we restrict the first condition to the "is_newentry" 1864 * case? 1865 * XXX: when we hear an RA from a new router with the link-layer 1866 * address option, defrouter_select() is called twice, since 1867 * defrtrlist_update called the function as well. However, I believe 1868 * we can compromise the overhead, since it only happens the first 1869 * time. 1870 * XXX: although defrouter_select() should not have a bad effect 1871 * for those are not autoconfigured hosts, we explicitly avoid such 1872 * cases for safety. 1873 */ 1874 if (do_update && ln->ln_router && !ip6_forwarding && ip6_accept_rtadv) 1875 defrouter_select(); 1876 1877 return rt; 1878 } 1879 1880 static void 1881 nd6_slowtimo(void *arg __unused) 1882 { 1883 struct lwkt_msg *lmsg = &nd6_slowtimo_netmsg.lmsg; 1884 1885 KASSERT(mycpuid == 0, ("not on cpu0")); 1886 crit_enter(); 1887 if (lmsg->ms_flags & MSGF_DONE) 1888 lwkt_sendmsg_oncpu(netisr_cpuport(0), lmsg); 1889 crit_exit(); 1890 } 1891 1892 static void 1893 nd6_slowtimo_dispatch(netmsg_t nmsg) 1894 { 1895 const struct ifnet_array *arr; 1896 struct nd_ifinfo *nd6if; 1897 int i; 1898 1899 ASSERT_NETISR0; 1900 1901 crit_enter(); 1902 lwkt_replymsg(&nmsg->lmsg, 0); /* reply ASAP */ 1903 crit_exit(); 1904 1905 arr = ifnet_array_get(); 1906 1907 mtx_lock(&nd6_mtx); 1908 for (i = 0; i < arr->ifnet_count; ++i) { 1909 struct ifnet *ifp = arr->ifnet_arr[i]; 1910 1911 if (ifp->if_afdata[AF_INET6] == NULL) 1912 continue; 1913 nd6if = ND_IFINFO(ifp); 1914 if (nd6if->basereachable && /* already initialized */ 1915 (nd6if->recalctm -= ND6_SLOWTIMER_INTERVAL) <= 0) { 1916 /* 1917 * Since reachable time rarely changes by router 1918 * advertisements, we SHOULD insure that a new random 1919 * value gets recomputed at least once every few hours. 1920 * (RFC 2461, 6.3.4) 1921 */ 1922 nd6if->recalctm = nd6_recalc_reachtm_interval; 1923 nd6if->reachable = ND_COMPUTE_RTIME(nd6if->basereachable); 1924 } 1925 } 1926 mtx_unlock(&nd6_mtx); 1927 1928 callout_reset(&nd6_slowtimo_ch, ND6_SLOWTIMER_INTERVAL * hz, 1929 nd6_slowtimo, NULL); 1930 } 1931 1932 int 1933 nd6_output(struct ifnet *ifp, struct ifnet *origifp, struct mbuf *m, 1934 struct sockaddr_in6 *dst, struct rtentry *rt) 1935 { 1936 int error; 1937 1938 if (ifp->if_flags & IFF_LOOPBACK) 1939 error = ifp->if_output(origifp, m, (struct sockaddr *)dst, rt); 1940 else 1941 error = ifp->if_output(ifp, m, (struct sockaddr *)dst, rt); 1942 return error; 1943 } 1944 1945 int 1946 nd6_resolve(struct ifnet *ifp, struct rtentry *rt0, struct mbuf *m, 1947 struct sockaddr *dst0, u_char *desten) 1948 { 1949 struct sockaddr_in6 *dst = SIN6(dst0); 1950 struct rtentry *rt = NULL; 1951 struct llinfo_nd6 *ln = NULL; 1952 int error; 1953 1954 if (m->m_flags & M_MCAST) { 1955 switch (ifp->if_type) { 1956 case IFT_ETHER: 1957 #ifdef IFT_L2VLAN 1958 case IFT_L2VLAN: 1959 #endif 1960 #ifdef IFT_IEEE80211 1961 case IFT_IEEE80211: 1962 #endif 1963 ETHER_MAP_IPV6_MULTICAST(&dst->sin6_addr, 1964 desten); 1965 return 0; 1966 case IFT_IEEE1394: 1967 bcopy(ifp->if_broadcastaddr, desten, ifp->if_addrlen); 1968 return 0; 1969 default: 1970 error = EAFNOSUPPORT; 1971 goto bad; 1972 } 1973 } 1974 1975 if (rt0 != NULL) { 1976 error = rt_llroute(dst0, rt0, &rt); 1977 if (error != 0) 1978 goto bad; 1979 ln = rt->rt_llinfo; 1980 } 1981 1982 /* 1983 * Since nd6_is_addr_neighbor() internally calls nd6_lookup(), 1984 * the condition below is not very efficient. But we believe 1985 * it is tolerable, because this should be a rare case. 1986 */ 1987 if (ln == NULL && nd6_is_addr_neighbor(dst, ifp)) { 1988 rt = nd6_lookup(&dst->sin6_addr, 1, ifp); 1989 if (rt != NULL) 1990 ln = rt->rt_llinfo; 1991 } 1992 1993 if (ln == NULL || rt == NULL) { 1994 if (!(ifp->if_flags & IFF_POINTOPOINT) && 1995 !(ND_IFINFO(ifp)->flags & ND6_IFF_PERFORMNUD)) { 1996 log(LOG_DEBUG, 1997 "nd6_output: can't allocate llinfo for %s " 1998 "(ln=%p, rt=%p)\n", 1999 ip6_sprintf(&dst->sin6_addr), ln, rt); 2000 error = ENOBUFS; 2001 goto bad; 2002 } 2003 return 0; 2004 } 2005 2006 /* We don't have to do link-layer address resolution on a p2p link. */ 2007 if ((ifp->if_flags & IFF_POINTOPOINT) && 2008 ln->ln_state < ND6_LLINFO_REACHABLE) { 2009 ln->ln_state = ND6_LLINFO_STALE; 2010 ln->ln_expire = time_uptime + nd6_gctimer; 2011 } 2012 2013 /* 2014 * The first time we send a packet to a neighbor whose entry is 2015 * STALE, we have to change the state to DELAY and a sets a timer to 2016 * expire in DELAY_FIRST_PROBE_TIME seconds to ensure do 2017 * neighbor unreachability detection on expiration. 2018 * (RFC 2461 7.3.3) 2019 */ 2020 if (ln->ln_state == ND6_LLINFO_STALE) { 2021 ln->ln_asked = 0; 2022 ln->ln_state = ND6_LLINFO_DELAY; 2023 ln->ln_expire = time_uptime + nd6_delay; 2024 } 2025 2026 /* 2027 * If the neighbor cache entry has a state other than INCOMPLETE 2028 * (i.e. its link-layer address is already resolved), return it. 2029 */ 2030 if (ln->ln_state > ND6_LLINFO_INCOMPLETE) { 2031 struct sockaddr_dl *sdl = SDL(rt->rt_gateway); 2032 2033 KKASSERT(sdl->sdl_family == AF_LINK && sdl->sdl_alen != 0); 2034 bcopy(LLADDR(sdl), desten, sdl->sdl_alen); 2035 return 0; 2036 } 2037 2038 /* 2039 * There is a neighbor cache entry, but no ethernet address 2040 * response yet. Replace the held mbuf (if any) with this 2041 * latest one. 2042 */ 2043 if (ln->ln_hold) 2044 m_freem(ln->ln_hold); 2045 ln->ln_hold = m; 2046 2047 /* 2048 * This code conforms to the rate-limiting rule described in Section 2049 * 7.2.2 of RFC 2461, because the timer is set correctly after sending 2050 * an NS below. 2051 */ 2052 if (ln->ln_state == ND6_LLINFO_NOSTATE || 2053 ln->ln_state == ND6_LLINFO_WAITDELETE) { 2054 /* 2055 * This neighbor cache entry was just created; change its 2056 * state to INCOMPLETE and start its life cycle. 2057 * 2058 * We force an NS output below by setting ln_expire to 1 2059 * (nd6_rtrequest() could set it to the current time_uptime) 2060 * and zeroing out ln_asked (XXX this may not be necessary). 2061 */ 2062 ln->ln_state = ND6_LLINFO_INCOMPLETE; 2063 ln->ln_expire = 1; 2064 ln->ln_asked = 0; 2065 } 2066 if (ln->ln_expire && ln->ln_expire < time_uptime && ln->ln_asked == 0) { 2067 ln->ln_asked++; 2068 ln->ln_expire = time_uptime + ND_IFINFO(ifp)->retrans / 1000; 2069 nd6_ns_output(ifp, NULL, &dst->sin6_addr, ln, 0); 2070 } 2071 2072 if (ln->ln_asked >= nd6_mmaxtries) 2073 return (rt != NULL && rt->rt_flags & RTF_GATEWAY) ? 2074 EHOSTUNREACH : EHOSTDOWN; 2075 return EWOULDBLOCK; 2076 2077 bad: 2078 m_freem(m); 2079 return error; 2080 } 2081 2082 int 2083 nd6_need_cache(struct ifnet *ifp) 2084 { 2085 /* 2086 * XXX: we currently do not make neighbor cache on any interface 2087 * other than Ethernet and GIF. 2088 * 2089 * RFC2893 says: 2090 * - unidirectional tunnels needs no ND 2091 */ 2092 switch (ifp->if_type) { 2093 case IFT_ETHER: 2094 case IFT_IEEE1394: 2095 #ifdef IFT_L2VLAN 2096 case IFT_L2VLAN: 2097 #endif 2098 #ifdef IFT_IEEE80211 2099 case IFT_IEEE80211: 2100 #endif 2101 #ifdef IFT_CARP 2102 case IFT_CARP: 2103 #endif 2104 case IFT_GIF: /* XXX need more cases? */ 2105 return (1); 2106 default: 2107 return (0); 2108 } 2109 } 2110 2111 static int nd6_sysctl_drlist(SYSCTL_HANDLER_ARGS); 2112 static int nd6_sysctl_prlist(SYSCTL_HANDLER_ARGS); 2113 #ifdef SYSCTL_DECL 2114 SYSCTL_DECL(_net_inet6_icmp6); 2115 #endif 2116 SYSCTL_NODE(_net_inet6_icmp6, ICMPV6CTL_ND6_DRLIST, nd6_drlist, 2117 CTLFLAG_RD, nd6_sysctl_drlist, "List default routers"); 2118 SYSCTL_NODE(_net_inet6_icmp6, ICMPV6CTL_ND6_PRLIST, nd6_prlist, 2119 CTLFLAG_RD, nd6_sysctl_prlist, "List prefixes"); 2120 2121 static int 2122 nd6_sysctl_drlist(SYSCTL_HANDLER_ARGS) 2123 { 2124 int error; 2125 char buf[1024]; 2126 struct in6_defrouter *d, *de; 2127 struct nd_defrouter *dr; 2128 2129 if (req->newptr) 2130 return EPERM; 2131 error = 0; 2132 2133 for (dr = TAILQ_FIRST(&nd_defrouter); dr; 2134 dr = TAILQ_NEXT(dr, dr_entry)) { 2135 d = (struct in6_defrouter *)buf; 2136 de = (struct in6_defrouter *)(buf + sizeof(buf)); 2137 2138 if (d + 1 <= de) { 2139 bzero(d, sizeof(*d)); 2140 d->rtaddr.sin6_family = AF_INET6; 2141 d->rtaddr.sin6_len = sizeof(d->rtaddr); 2142 if (in6_recoverscope(&d->rtaddr, &dr->rtaddr, 2143 dr->ifp) != 0) 2144 log(LOG_ERR, 2145 "scope error in " 2146 "default router list (%s)\n", 2147 ip6_sprintf(&dr->rtaddr)); 2148 d->flags = dr->flags; 2149 d->rtlifetime = dr->rtlifetime; 2150 d->expire = dr->expire; 2151 d->if_index = dr->ifp->if_index; 2152 } else 2153 panic("buffer too short"); 2154 2155 error = SYSCTL_OUT(req, buf, sizeof(*d)); 2156 if (error) 2157 break; 2158 } 2159 return error; 2160 } 2161 2162 static int 2163 nd6_sysctl_prlist(SYSCTL_HANDLER_ARGS) 2164 { 2165 int error; 2166 char buf[1024]; 2167 struct in6_prefix *p, *pe; 2168 struct nd_prefix *pr; 2169 2170 if (req->newptr) 2171 return EPERM; 2172 error = 0; 2173 2174 for (pr = nd_prefix.lh_first; pr; pr = pr->ndpr_next) { 2175 u_short advrtrs; 2176 size_t advance; 2177 struct sockaddr_in6 *sin6, *s6; 2178 struct nd_pfxrouter *pfr; 2179 2180 p = (struct in6_prefix *)buf; 2181 pe = (struct in6_prefix *)(buf + sizeof(buf)); 2182 2183 if (p + 1 <= pe) { 2184 bzero(p, sizeof(*p)); 2185 sin6 = (struct sockaddr_in6 *)(p + 1); 2186 2187 p->prefix = pr->ndpr_prefix; 2188 if (in6_recoverscope(&p->prefix, 2189 &p->prefix.sin6_addr, pr->ndpr_ifp) != 0) 2190 log(LOG_ERR, 2191 "scope error in prefix list (%s)\n", 2192 ip6_sprintf(&p->prefix.sin6_addr)); 2193 p->raflags = pr->ndpr_raf; 2194 p->prefixlen = pr->ndpr_plen; 2195 p->vltime = pr->ndpr_vltime; 2196 p->pltime = pr->ndpr_pltime; 2197 p->if_index = pr->ndpr_ifp->if_index; 2198 p->expire = pr->ndpr_expire; 2199 p->refcnt = pr->ndpr_refcnt; 2200 p->flags = pr->ndpr_stateflags; 2201 p->origin = PR_ORIG_RA; 2202 advrtrs = 0; 2203 for (pfr = pr->ndpr_advrtrs.lh_first; pfr; 2204 pfr = pfr->pfr_next) { 2205 if ((void *)&sin6[advrtrs + 1] > (void *)pe) { 2206 advrtrs++; 2207 continue; 2208 } 2209 s6 = &sin6[advrtrs]; 2210 bzero(s6, sizeof(*s6)); 2211 s6->sin6_family = AF_INET6; 2212 s6->sin6_len = sizeof(*sin6); 2213 if (in6_recoverscope(s6, &pfr->router->rtaddr, 2214 pfr->router->ifp) != 0) 2215 log(LOG_ERR, 2216 "scope error in " 2217 "prefix list (%s)\n", 2218 ip6_sprintf(&pfr->router->rtaddr)); 2219 advrtrs++; 2220 } 2221 p->advrtrs = advrtrs; 2222 } else { 2223 panic("buffer too short"); 2224 } 2225 2226 advance = sizeof(*p) + sizeof(*sin6) * advrtrs; 2227 error = SYSCTL_OUT(req, buf, advance); 2228 if (error) 2229 break; 2230 } 2231 return error; 2232 } 2233