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_INCOMPLETE: 479 if (ln->ln_asked < nd6_mmaxtries) { 480 ln->ln_asked++; 481 ln->ln_expire = time_uptime + 482 ND_IFINFO(ifp)->retrans / 1000; 483 nd6_ns_output(ifp, NULL, &dst->sin6_addr, 484 ln, 0); 485 } else { 486 struct mbuf *m = ln->ln_hold; 487 if (m) { 488 if (rt->rt_ifp) { 489 /* 490 * Fake rcvif to make ICMP error 491 * more helpful in diagnosing 492 * for the receiver. 493 * XXX: should we consider 494 * older rcvif? 495 */ 496 m->m_pkthdr.rcvif = rt->rt_ifp; 497 } 498 icmp6_error(m, ICMP6_DST_UNREACH, 499 ICMP6_DST_UNREACH_ADDR, 0); 500 ln->ln_hold = NULL; 501 } 502 next = nd6_free(rt); 503 } 504 break; 505 case ND6_LLINFO_REACHABLE: 506 if (ln->ln_expire) { 507 ln->ln_state = ND6_LLINFO_STALE; 508 ln->ln_expire = time_uptime + nd6_gctimer; 509 } 510 break; 511 512 case ND6_LLINFO_STALE: 513 /* Garbage Collection(RFC 2461 5.3) */ 514 if (ln->ln_expire) 515 next = nd6_free(rt); 516 break; 517 518 case ND6_LLINFO_DELAY: 519 if (ndi && (ndi->flags & ND6_IFF_PERFORMNUD)) { 520 /* We need NUD */ 521 ln->ln_asked = 1; 522 ln->ln_state = ND6_LLINFO_PROBE; 523 ln->ln_expire = time_uptime + 524 ndi->retrans / 1000; 525 nd6_ns_output(ifp, &dst->sin6_addr, 526 &dst->sin6_addr, 527 ln, 0); 528 } else { 529 ln->ln_state = ND6_LLINFO_STALE; /* XXX */ 530 ln->ln_expire = time_uptime + nd6_gctimer; 531 } 532 break; 533 case ND6_LLINFO_PROBE: 534 if (ln->ln_asked < nd6_umaxtries) { 535 ln->ln_asked++; 536 ln->ln_expire = time_uptime + 537 ND_IFINFO(ifp)->retrans / 1000; 538 nd6_ns_output(ifp, &dst->sin6_addr, 539 &dst->sin6_addr, ln, 0); 540 } else { 541 next = nd6_free(rt); 542 } 543 break; 544 } 545 ln = next; 546 } 547 548 /* expire default router list */ 549 dr = TAILQ_FIRST(&nd_defrouter); 550 while (dr) { 551 if (dr->expire && dr->expire < time_uptime) { 552 struct nd_defrouter *t; 553 t = TAILQ_NEXT(dr, dr_entry); 554 defrtrlist_del(dr); 555 dr = t; 556 } else { 557 dr = TAILQ_NEXT(dr, dr_entry); 558 } 559 } 560 561 /* 562 * expire interface addresses. 563 * in the past the loop was inside prefix expiry processing. 564 * However, from a stricter speci-confrmance standpoint, we should 565 * rather separate address lifetimes and prefix lifetimes. 566 */ 567 addrloop: 568 for (ia6 = in6_ifaddr; ia6; ia6 = nia6) { 569 nia6 = ia6->ia_next; 570 /* check address lifetime */ 571 if (IFA6_IS_INVALID(ia6)) { 572 int regen = 0; 573 574 /* 575 * If the expiring address is temporary, try 576 * regenerating a new one. This would be useful when 577 * we suspended a laptop PC, then turned it on after a 578 * period that could invalidate all temporary 579 * addresses. Although we may have to restart the 580 * loop (see below), it must be after purging the 581 * address. Otherwise, we'd see an infinite loop of 582 * regeneration. 583 */ 584 if (ip6_use_tempaddr && 585 (ia6->ia6_flags & IN6_IFF_TEMPORARY)) { 586 if (regen_tmpaddr(ia6) == 0) 587 regen = 1; 588 } 589 590 in6_purgeaddr(&ia6->ia_ifa); 591 592 if (regen) 593 goto addrloop; /* XXX: see below */ 594 } 595 if (IFA6_IS_DEPRECATED(ia6)) { 596 int oldflags = ia6->ia6_flags; 597 598 ia6->ia6_flags |= IN6_IFF_DEPRECATED; 599 600 /* 601 * If a temporary address has just become deprecated, 602 * regenerate a new one if possible. 603 */ 604 if (ip6_use_tempaddr && 605 (ia6->ia6_flags & IN6_IFF_TEMPORARY) && 606 !(oldflags & IN6_IFF_DEPRECATED)) { 607 608 if (regen_tmpaddr(ia6) == 0) { 609 /* 610 * A new temporary address is 611 * generated. 612 * XXX: this means the address chain 613 * has changed while we are still in 614 * the loop. Although the change 615 * would not cause disaster (because 616 * it's not a deletion, but an 617 * addition,) we'd rather restart the 618 * loop just for safety. Or does this 619 * significantly reduce performance?? 620 */ 621 goto addrloop; 622 } 623 } 624 } else { 625 /* 626 * A new RA might have made a deprecated address 627 * preferred. 628 */ 629 ia6->ia6_flags &= ~IN6_IFF_DEPRECATED; 630 } 631 } 632 633 /* expire prefix list */ 634 pr = nd_prefix.lh_first; 635 while (pr) { 636 /* 637 * check prefix lifetime. 638 * since pltime is just for autoconf, pltime processing for 639 * prefix is not necessary. 640 */ 641 if (pr->ndpr_expire && pr->ndpr_expire < time_uptime) { 642 struct nd_prefix *t; 643 t = pr->ndpr_next; 644 645 /* 646 * address expiration and prefix expiration are 647 * separate. NEVER perform in6_purgeaddr here. 648 */ 649 650 prelist_remove(pr); 651 pr = t; 652 } else 653 pr = pr->ndpr_next; 654 } 655 656 mtx_unlock(&nd6_mtx); 657 658 callout_reset(&nd6_timer_ch, nd6_prune * hz, nd6_timer, NULL); 659 } 660 661 static void 662 nd6_timer(void *arg __unused) 663 { 664 struct lwkt_msg *lmsg = &nd6_timer_netmsg.lmsg; 665 666 KASSERT(mycpuid == 0, ("not on cpu0")); 667 crit_enter(); 668 if (lmsg->ms_flags & MSGF_DONE) 669 lwkt_sendmsg_oncpu(netisr_cpuport(0), lmsg); 670 crit_exit(); 671 } 672 673 void 674 nd6_timer_init(void) 675 { 676 callout_init_mp(&nd6_timer_ch); 677 netmsg_init(&nd6_timer_netmsg, NULL, &netisr_adone_rport, 678 MSGF_PRIORITY, nd6_timer_dispatch); 679 callout_reset_bycpu(&nd6_timer_ch, hz, nd6_timer, NULL, 0); 680 } 681 682 static int 683 regen_tmpaddr(struct in6_ifaddr *ia6) /* deprecated/invalidated temporary 684 address */ 685 { 686 struct ifaddr_container *ifac; 687 struct ifnet *ifp; 688 struct in6_ifaddr *public_ifa6 = NULL; 689 690 ifp = ia6->ia_ifa.ifa_ifp; 691 TAILQ_FOREACH(ifac, &ifp->if_addrheads[mycpuid], ifa_link) { 692 struct ifaddr *ifa = ifac->ifa; 693 struct in6_ifaddr *it6; 694 695 if (ifa->ifa_addr->sa_family != AF_INET6) 696 continue; 697 698 it6 = (struct in6_ifaddr *)ifa; 699 700 /* ignore no autoconf addresses. */ 701 if (!(it6->ia6_flags & IN6_IFF_AUTOCONF)) 702 continue; 703 704 /* ignore autoconf addresses with different prefixes. */ 705 if (it6->ia6_ndpr == NULL || it6->ia6_ndpr != ia6->ia6_ndpr) 706 continue; 707 708 /* 709 * Now we are looking at an autoconf address with the same 710 * prefix as ours. If the address is temporary and is still 711 * preferred, do not create another one. It would be rare, but 712 * could happen, for example, when we resume a laptop PC after 713 * a long period. 714 */ 715 if ((it6->ia6_flags & IN6_IFF_TEMPORARY) && 716 !IFA6_IS_DEPRECATED(it6)) { 717 public_ifa6 = NULL; 718 break; 719 } 720 721 /* 722 * This is a public autoconf address that has the same prefix 723 * as ours. If it is preferred, keep it. We can't break the 724 * loop here, because there may be a still-preferred temporary 725 * address with the prefix. 726 */ 727 if (!IFA6_IS_DEPRECATED(it6)) 728 public_ifa6 = it6; 729 } 730 731 if (public_ifa6 != NULL) { 732 int e; 733 734 if ((e = in6_tmpifadd(public_ifa6, 0)) != 0) { 735 log(LOG_NOTICE, "regen_tmpaddr: failed to create a new" 736 " tmp addr,errno=%d\n", e); 737 return (-1); 738 } 739 return (0); 740 } 741 742 return (-1); 743 } 744 745 /* 746 * Nuke neighbor cache/prefix/default router management table, right before 747 * ifp goes away. 748 */ 749 void 750 nd6_purge(struct ifnet *ifp) 751 { 752 struct llinfo_nd6 *ln, *nln; 753 struct nd_defrouter *dr, *ndr, drany; 754 struct nd_prefix *pr, *npr; 755 756 /* Nuke default router list entries toward ifp */ 757 if ((dr = TAILQ_FIRST(&nd_defrouter)) != NULL) { 758 /* 759 * The first entry of the list may be stored in 760 * the routing table, so we'll delete it later. 761 */ 762 for (dr = TAILQ_NEXT(dr, dr_entry); dr; dr = ndr) { 763 ndr = TAILQ_NEXT(dr, dr_entry); 764 if (dr->ifp == ifp) 765 defrtrlist_del(dr); 766 } 767 dr = TAILQ_FIRST(&nd_defrouter); 768 if (dr->ifp == ifp) 769 defrtrlist_del(dr); 770 } 771 772 /* Nuke prefix list entries toward ifp */ 773 for (pr = nd_prefix.lh_first; pr; pr = npr) { 774 npr = pr->ndpr_next; 775 if (pr->ndpr_ifp == ifp) { 776 /* 777 * Previously, pr->ndpr_addr is removed as well, 778 * but I strongly believe we don't have to do it. 779 * nd6_purge() is only called from in6_ifdetach(), 780 * which removes all the associated interface addresses 781 * by itself. 782 * (jinmei@kame.net 20010129) 783 */ 784 prelist_remove(pr); 785 } 786 } 787 788 /* cancel default outgoing interface setting */ 789 if (nd6_defifindex == ifp->if_index) 790 nd6_setdefaultiface(0); 791 792 if (!ip6_forwarding && ip6_accept_rtadv) { /* XXX: too restrictive? */ 793 /* refresh default router list */ 794 bzero(&drany, sizeof(drany)); 795 defrouter_delreq(&drany, 0); 796 defrouter_select(); 797 } 798 799 /* 800 * Nuke neighbor cache entries for the ifp. 801 * Note that rt->rt_ifp may not be the same as ifp, 802 * due to KAME goto ours hack. See RTM_RESOLVE case in 803 * nd6_rtrequest(), and ip6_input(). 804 */ 805 ln = llinfo_nd6.ln_next; 806 while (ln && ln != &llinfo_nd6) { 807 struct rtentry *rt; 808 struct sockaddr_dl *sdl; 809 810 nln = ln->ln_next; 811 rt = ln->ln_rt; 812 if (rt && rt->rt_gateway && 813 rt->rt_gateway->sa_family == AF_LINK) { 814 sdl = (struct sockaddr_dl *)rt->rt_gateway; 815 if (sdl->sdl_index == ifp->if_index) 816 nln = nd6_free(rt); 817 } 818 ln = nln; 819 } 820 } 821 822 struct rtentry * 823 nd6_lookup(struct in6_addr *addr6, int create, struct ifnet *ifp) 824 { 825 struct rtentry *rt; 826 struct sockaddr_in6 sin6; 827 828 bzero(&sin6, sizeof(sin6)); 829 sin6.sin6_len = sizeof(struct sockaddr_in6); 830 sin6.sin6_family = AF_INET6; 831 sin6.sin6_addr = *addr6; 832 833 if (create) 834 rt = rtlookup((struct sockaddr *)&sin6); 835 else 836 rt = rtpurelookup((struct sockaddr *)&sin6); 837 if (rt && !(rt->rt_flags & RTF_LLINFO)) { 838 /* 839 * This is the case for the default route. 840 * If we want to create a neighbor cache for the address, we 841 * should free the route for the destination and allocate an 842 * interface route. 843 */ 844 if (create) { 845 --rt->rt_refcnt; 846 rt = NULL; 847 } 848 } 849 if (!rt) { 850 if (create && ifp) { 851 int e; 852 853 /* 854 * If no route is available and create is set, 855 * we allocate a host route for the destination 856 * and treat it like an interface route. 857 * This hack is necessary for a neighbor which can't 858 * be covered by our own prefix. 859 */ 860 struct ifaddr *ifa; 861 862 ifa = ifaof_ifpforaddr((struct sockaddr *)&sin6, ifp); 863 if (ifa == NULL) 864 return (NULL); 865 866 /* 867 * Create a new route. RTF_LLINFO is necessary 868 * to create a Neighbor Cache entry for the 869 * destination in nd6_rtrequest which will be 870 * called in rtrequest via ifa->ifa_rtrequest. 871 */ 872 if ((e = rtrequest(RTM_ADD, (struct sockaddr *)&sin6, 873 ifa->ifa_addr, (struct sockaddr *)&all1_sa, 874 (ifa->ifa_flags | RTF_HOST | RTF_LLINFO) & 875 ~RTF_CLONING, &rt)) != 0) { 876 log(LOG_ERR, 877 "nd6_lookup: failed to add route for a " 878 "neighbor(%s), errno=%d\n", 879 ip6_sprintf(addr6), e); 880 } 881 if (rt == NULL) 882 return (NULL); 883 if (rt->rt_llinfo) { 884 struct llinfo_nd6 *ln = 885 (struct llinfo_nd6 *)rt->rt_llinfo; 886 887 ln->ln_state = ND6_LLINFO_NOSTATE; 888 } 889 } else 890 return (NULL); 891 } 892 rt->rt_refcnt--; 893 894 if (!ND6_RTENTRY_IS_NEIGHBOR(rt, ifp)) { 895 if (create) { 896 log(LOG_DEBUG, 897 "nd6_lookup: failed to lookup %s (if = %s)\n", 898 ip6_sprintf(addr6), ifp ? if_name(ifp) : "unspec"); 899 /* xxx more logs... kazu */ 900 } 901 return (NULL); 902 } 903 return (rt); 904 } 905 906 static struct rtentry * 907 nd6_neighbor_lookup(struct in6_addr *addr6, struct ifnet *ifp) 908 { 909 struct rtentry *rt; 910 struct sockaddr_in6 sin6; 911 912 bzero(&sin6, sizeof(sin6)); 913 sin6.sin6_len = sizeof(struct sockaddr_in6); 914 sin6.sin6_family = AF_INET6; 915 sin6.sin6_addr = *addr6; 916 917 rt = rtpurelookup((struct sockaddr *)&sin6); 918 if (rt == NULL) 919 return (NULL); 920 rt->rt_refcnt--; 921 922 if (!ND6_RTENTRY_IS_NEIGHBOR(rt, ifp)) { 923 if (nd6_onlink_ns_rfc4861 && 924 (ND6_RTENTRY_IS_LLCLONING(rt) || /* not cloned yet */ 925 (rt->rt_parent != NULL && /* cloning */ 926 ND6_RTENTRY_IS_LLCLONING(rt->rt_parent)))) { 927 /* 928 * If cloning ever happened or is happening, 929 * rtentry for addr6 would or will become a 930 * neighbor cache. 931 */ 932 } else { 933 rt = NULL; 934 } 935 } 936 return (rt); 937 } 938 939 /* 940 * Detect if a given IPv6 address identifies a neighbor on a given link. 941 * XXX: should take care of the destination of a p2p link? 942 */ 943 int 944 nd6_is_addr_neighbor(struct sockaddr_in6 *addr, struct ifnet *ifp) 945 { 946 struct ifaddr_container *ifac; 947 int i; 948 949 #define IFADDR6(a) ((((struct in6_ifaddr *)(a))->ia_addr).sin6_addr) 950 #define IFMASK6(a) ((((struct in6_ifaddr *)(a))->ia_prefixmask).sin6_addr) 951 952 /* 953 * A link-local address is always a neighbor. 954 * XXX: we should use the sin6_scope_id field rather than the embedded 955 * interface index. 956 */ 957 if (IN6_IS_ADDR_LINKLOCAL(&addr->sin6_addr) && 958 ntohs(*(u_int16_t *)&addr->sin6_addr.s6_addr[2]) == ifp->if_index) 959 return (1); 960 961 /* 962 * If the address matches one of our addresses, 963 * it should be a neighbor. 964 */ 965 TAILQ_FOREACH(ifac, &ifp->if_addrheads[mycpuid], ifa_link) { 966 struct ifaddr *ifa = ifac->ifa; 967 968 if (ifa->ifa_addr->sa_family != AF_INET6) 969 next: continue; 970 971 for (i = 0; i < 4; i++) { 972 if ((IFADDR6(ifa).s6_addr32[i] ^ 973 addr->sin6_addr.s6_addr32[i]) & 974 IFMASK6(ifa).s6_addr32[i]) 975 goto next; 976 } 977 return (1); 978 } 979 980 /* 981 * Even if the address matches none of our addresses, it might be 982 * in the neighbor cache. 983 */ 984 if (nd6_neighbor_lookup(&addr->sin6_addr, ifp) != NULL) 985 return (1); 986 987 return (0); 988 #undef IFADDR6 989 #undef IFMASK6 990 } 991 992 /* 993 * Free an nd6 llinfo entry. 994 */ 995 struct llinfo_nd6 * 996 nd6_free(struct rtentry *rt) 997 { 998 struct llinfo_nd6 *ln = (struct llinfo_nd6 *)rt->rt_llinfo, *next; 999 struct in6_addr in6 = ((struct sockaddr_in6 *)rt_key(rt))->sin6_addr; 1000 struct nd_defrouter *dr; 1001 1002 /* 1003 * we used to have kpfctlinput(PRC_HOSTDEAD) here. 1004 * even though it is not harmful, it was not really necessary. 1005 */ 1006 1007 if (!ip6_forwarding && ip6_accept_rtadv) { /* XXX: too restrictive? */ 1008 mtx_lock(&nd6_mtx); 1009 dr = defrouter_lookup( 1010 &((struct sockaddr_in6 *)rt_key(rt))->sin6_addr, 1011 rt->rt_ifp); 1012 1013 if (ln->ln_router || dr) { 1014 /* 1015 * rt6_flush must be called whether or not the neighbor 1016 * is in the Default Router List. 1017 * See a corresponding comment in nd6_na_input(). 1018 */ 1019 rt6_flush(&in6, rt->rt_ifp); 1020 } 1021 1022 if (dr) { 1023 /* 1024 * Unreachablity of a router might affect the default 1025 * router selection and on-link detection of advertised 1026 * prefixes. 1027 */ 1028 1029 /* 1030 * Temporarily fake the state to choose a new default 1031 * router and to perform on-link determination of 1032 * prefixes correctly. 1033 * Below the state will be set correctly, 1034 * or the entry itself will be deleted. 1035 */ 1036 ln->ln_state = ND6_LLINFO_INCOMPLETE; 1037 1038 /* 1039 * Since defrouter_select() does not affect the 1040 * on-link determination and MIP6 needs the check 1041 * before the default router selection, we perform 1042 * the check now. 1043 */ 1044 pfxlist_onlink_check(); 1045 1046 if (dr == TAILQ_FIRST(&nd_defrouter)) { 1047 /* 1048 * It is used as the current default router, 1049 * so we have to move it to the end of the 1050 * list and choose a new one. 1051 * XXX: it is not very efficient if this is 1052 * the only router. 1053 */ 1054 TAILQ_REMOVE(&nd_defrouter, dr, dr_entry); 1055 TAILQ_INSERT_TAIL(&nd_defrouter, dr, dr_entry); 1056 1057 defrouter_select(); 1058 } 1059 } 1060 mtx_unlock(&nd6_mtx); 1061 } 1062 1063 /* 1064 * Before deleting the entry, remember the next entry as the 1065 * return value. We need this because pfxlist_onlink_check() above 1066 * might have freed other entries (particularly the old next entry) as 1067 * a side effect (XXX). 1068 */ 1069 next = ln->ln_next; 1070 1071 /* 1072 * Detach the route from the routing tree and the list of neighbor 1073 * caches, and disable the route entry not to be used in already 1074 * cached routes. 1075 */ 1076 rtrequest(RTM_DELETE, rt_key(rt), NULL, rt_mask(rt), 0, NULL); 1077 1078 return (next); 1079 } 1080 1081 /* 1082 * Upper-layer reachability hint for Neighbor Unreachability Detection. 1083 * 1084 * XXX cost-effective metods? 1085 */ 1086 void 1087 nd6_nud_hint(struct rtentry *rt, struct in6_addr *dst6, int force) 1088 { 1089 struct llinfo_nd6 *ln; 1090 1091 /* 1092 * If the caller specified "rt", use that. Otherwise, resolve the 1093 * routing table by supplied "dst6". 1094 */ 1095 if (!rt) { 1096 if (!dst6) 1097 return; 1098 if (!(rt = nd6_lookup(dst6, 0, NULL))) 1099 return; 1100 } 1101 1102 if ((rt->rt_flags & RTF_GATEWAY) || 1103 !(rt->rt_flags & RTF_LLINFO) || 1104 rt->rt_llinfo == NULL || rt->rt_gateway == NULL || 1105 rt->rt_gateway->sa_family != AF_LINK) { 1106 /* This is not a host route. */ 1107 return; 1108 } 1109 1110 ln = (struct llinfo_nd6 *)rt->rt_llinfo; 1111 if (ln->ln_state < ND6_LLINFO_REACHABLE) 1112 return; 1113 1114 /* 1115 * if we get upper-layer reachability confirmation many times, 1116 * it is possible we have false information. 1117 */ 1118 if (!force) { 1119 ln->ln_byhint++; 1120 if (ln->ln_byhint > nd6_maxnudhint) 1121 return; 1122 } 1123 1124 ln->ln_state = ND6_LLINFO_REACHABLE; 1125 if (ln->ln_expire) 1126 ln->ln_expire = time_uptime + 1127 ND_IFINFO(rt->rt_ifp)->reachable; 1128 } 1129 1130 void 1131 nd6_rtrequest(int req, struct rtentry *rt) 1132 { 1133 struct sockaddr *gate = rt->rt_gateway; 1134 struct llinfo_nd6 *ln = (struct llinfo_nd6 *)rt->rt_llinfo; 1135 static struct sockaddr_dl null_sdl = {sizeof(null_sdl), AF_LINK}; 1136 struct ifnet *ifp = rt->rt_ifp; 1137 struct ifaddr *ifa; 1138 1139 if ((rt->rt_flags & RTF_GATEWAY)) 1140 return; 1141 1142 if (nd6_need_cache(ifp) == 0 && !(rt->rt_flags & RTF_HOST)) { 1143 /* 1144 * This is probably an interface direct route for a link 1145 * which does not need neighbor caches (e.g. fe80::%lo0/64). 1146 * We do not need special treatment below for such a route. 1147 * Moreover, the RTF_LLINFO flag which would be set below 1148 * would annoy the ndp(8) command. 1149 */ 1150 return; 1151 } 1152 1153 if (req == RTM_RESOLVE && 1154 (nd6_need_cache(ifp) == 0 || /* stf case */ 1155 !nd6_is_addr_neighbor((struct sockaddr_in6 *)rt_key(rt), ifp))) { 1156 /* 1157 * FreeBSD and BSD/OS often make a cloned host route based 1158 * on a less-specific route (e.g. the default route). 1159 * If the less specific route does not have a "gateway" 1160 * (this is the case when the route just goes to a p2p or an 1161 * stf interface), we'll mistakenly make a neighbor cache for 1162 * the host route, and will see strange neighbor solicitation 1163 * for the corresponding destination. In order to avoid the 1164 * confusion, we check if the destination of the route is 1165 * a neighbor in terms of neighbor discovery, and stop the 1166 * process if not. Additionally, we remove the LLINFO flag 1167 * so that ndp(8) will not try to get the neighbor information 1168 * of the destination. 1169 */ 1170 rt->rt_flags &= ~RTF_LLINFO; 1171 return; 1172 } 1173 1174 switch (req) { 1175 case RTM_ADD: 1176 /* 1177 * There is no backward compatibility :) 1178 * 1179 * if (!(rt->rt_flags & RTF_HOST) && 1180 * SIN(rt_mask(rt))->sin_addr.s_addr != 0xffffffff) 1181 * rt->rt_flags |= RTF_CLONING; 1182 */ 1183 if (rt->rt_flags & (RTF_CLONING | RTF_LLINFO)) { 1184 /* 1185 * Case 1: This route should come from 1186 * a route to interface. RTF_LLINFO flag is set 1187 * for a host route whose destination should be 1188 * treated as on-link. 1189 */ 1190 rt_setgate(rt, rt_key(rt), 1191 (struct sockaddr *)&null_sdl, 1192 RTL_DONTREPORT); 1193 gate = rt->rt_gateway; 1194 SDL(gate)->sdl_type = ifp->if_type; 1195 SDL(gate)->sdl_index = ifp->if_index; 1196 if (ln) 1197 ln->ln_expire = time_uptime; 1198 if (ln && ln->ln_expire == 0) { 1199 /* kludge for desktops */ 1200 ln->ln_expire = 1; 1201 } 1202 if ((rt->rt_flags & RTF_CLONING)) 1203 break; 1204 } 1205 /* 1206 * In IPv4 code, we try to annonuce new RTF_ANNOUNCE entry here. 1207 * We don't do that here since llinfo is not ready yet. 1208 * 1209 * There are also couple of other things to be discussed: 1210 * - unsolicited NA code needs improvement beforehand 1211 * - RFC2461 says we MAY send multicast unsolicited NA 1212 * (7.2.6 paragraph 4), however, it also says that we 1213 * SHOULD provide a mechanism to prevent multicast NA storm. 1214 * we don't have anything like it right now. 1215 * note that the mechanism needs a mutual agreement 1216 * between proxies, which means that we need to implement 1217 * a new protocol, or a new kludge. 1218 * - from RFC2461 6.2.4, host MUST NOT send an unsolicited NA. 1219 * we need to check ip6forwarding before sending it. 1220 * (or should we allow proxy ND configuration only for 1221 * routers? there's no mention about proxy ND from hosts) 1222 */ 1223 #if 0 1224 /* XXX it does not work */ 1225 if ((rt->rt_flags & RTF_ANNOUNCE) && mycpuid == 0) { 1226 nd6_na_output(ifp, 1227 &SIN6(rt_key(rt))->sin6_addr, 1228 &SIN6(rt_key(rt))->sin6_addr, 1229 ip6_forwarding ? ND_NA_FLAG_ROUTER : 0, 1230 1, NULL); 1231 } 1232 #endif 1233 /* FALLTHROUGH */ 1234 case RTM_RESOLVE: 1235 if ((ifp->if_flags & (IFF_POINTOPOINT | IFF_LOOPBACK)) == 0) { 1236 /* 1237 * Address resolution isn't necessary for a point to 1238 * point link, so we can skip this test for a p2p link. 1239 */ 1240 if (gate->sa_family != AF_LINK || 1241 gate->sa_len < sizeof(null_sdl)) { 1242 log(LOG_DEBUG, 1243 "nd6_rtrequest: bad gateway value: %s\n", 1244 if_name(ifp)); 1245 break; 1246 } 1247 SDL(gate)->sdl_type = ifp->if_type; 1248 SDL(gate)->sdl_index = ifp->if_index; 1249 } 1250 if (ln != NULL) 1251 break; /* This happens on a route change */ 1252 /* 1253 * Case 2: This route may come from cloning, or a manual route 1254 * add with a LL address. 1255 */ 1256 R_Malloc(ln, struct llinfo_nd6 *, sizeof(*ln)); 1257 rt->rt_llinfo = (caddr_t)ln; 1258 if (!ln) { 1259 log(LOG_DEBUG, "nd6_rtrequest: malloc failed\n"); 1260 break; 1261 } 1262 nd6_inuse++; 1263 nd6_allocated++; 1264 bzero(ln, sizeof(*ln)); 1265 ln->ln_rt = rt; 1266 /* this is required for "ndp" command. - shin */ 1267 if (req == RTM_ADD) { 1268 /* 1269 * gate should have some valid AF_LINK entry, 1270 * and ln->ln_expire should have some lifetime 1271 * which is specified by ndp command. 1272 */ 1273 ln->ln_state = ND6_LLINFO_REACHABLE; 1274 ln->ln_byhint = 0; 1275 } else { 1276 /* 1277 * When req == RTM_RESOLVE, rt is created and 1278 * initialized in rtrequest(), so rt_expire is 0. 1279 */ 1280 ln->ln_state = ND6_LLINFO_NOSTATE; 1281 ln->ln_expire = time_uptime; 1282 } 1283 rt->rt_flags |= RTF_LLINFO; 1284 ln->ln_next = llinfo_nd6.ln_next; 1285 llinfo_nd6.ln_next = ln; 1286 ln->ln_prev = &llinfo_nd6; 1287 ln->ln_next->ln_prev = ln; 1288 1289 /* 1290 * check if rt_key(rt) is one of my address assigned 1291 * to the interface. 1292 */ 1293 ifa = (struct ifaddr *)in6ifa_ifpwithaddr(rt->rt_ifp, 1294 &SIN6(rt_key(rt))->sin6_addr); 1295 if (ifa) { 1296 caddr_t macp = nd6_ifptomac(ifp); 1297 ln->ln_expire = 0; 1298 ln->ln_state = ND6_LLINFO_REACHABLE; 1299 ln->ln_byhint = 0; 1300 if (macp) { 1301 bcopy(macp, LLADDR(SDL(gate)), ifp->if_addrlen); 1302 SDL(gate)->sdl_alen = ifp->if_addrlen; 1303 } 1304 if (nd6_useloopback) { 1305 rt->rt_ifp = loif; /* XXX */ 1306 /* 1307 * Make sure rt_ifa be equal to the ifaddr 1308 * corresponding to the address. 1309 * We need this because when we refer 1310 * rt_ifa->ia6_flags in ip6_input, we assume 1311 * that the rt_ifa points to the address instead 1312 * of the loopback address. 1313 */ 1314 if (ifa != rt->rt_ifa) { 1315 IFAFREE(rt->rt_ifa); 1316 IFAREF(ifa); 1317 rt->rt_ifa = ifa; 1318 } 1319 } 1320 } else if (rt->rt_flags & RTF_ANNOUNCE) { 1321 ln->ln_expire = 0; 1322 ln->ln_state = ND6_LLINFO_REACHABLE; 1323 ln->ln_byhint = 0; 1324 1325 /* 1326 * Join solicited node multicast for proxy ND, and only 1327 * join it once on cpu0. 1328 */ 1329 if ((ifp->if_flags & IFF_MULTICAST) && mycpuid == 0) { 1330 struct in6_addr llsol; 1331 int error; 1332 1333 llsol = SIN6(rt_key(rt))->sin6_addr; 1334 llsol.s6_addr16[0] = htons(0xff02); 1335 llsol.s6_addr16[1] = htons(ifp->if_index); 1336 llsol.s6_addr32[1] = 0; 1337 llsol.s6_addr32[2] = htonl(1); 1338 llsol.s6_addr8[12] = 0xff; 1339 1340 if (!in6_addmulti(&llsol, ifp, &error)) { 1341 nd6log((LOG_ERR, "%s: failed to join " 1342 "%s (errno=%d)\n", if_name(ifp), 1343 ip6_sprintf(&llsol), error)); 1344 } 1345 } 1346 } 1347 break; 1348 1349 case RTM_DELETE: 1350 if (!ln) 1351 break; 1352 /* 1353 * Leave from solicited node multicast for proxy ND, and only 1354 * leave it once on cpu0 (since we joined it once on cpu0). 1355 */ 1356 if ((rt->rt_flags & RTF_ANNOUNCE) && 1357 (ifp->if_flags & IFF_MULTICAST) && mycpuid == 0) { 1358 struct in6_addr llsol; 1359 struct in6_multi *in6m; 1360 1361 llsol = SIN6(rt_key(rt))->sin6_addr; 1362 llsol.s6_addr16[0] = htons(0xff02); 1363 llsol.s6_addr16[1] = htons(ifp->if_index); 1364 llsol.s6_addr32[1] = 0; 1365 llsol.s6_addr32[2] = htonl(1); 1366 llsol.s6_addr8[12] = 0xff; 1367 1368 in6m = IN6_LOOKUP_MULTI(&llsol, ifp); 1369 if (in6m) 1370 in6_delmulti(in6m); 1371 } 1372 nd6_inuse--; 1373 ln->ln_next->ln_prev = ln->ln_prev; 1374 ln->ln_prev->ln_next = ln->ln_next; 1375 ln->ln_prev = NULL; 1376 rt->rt_llinfo = 0; 1377 rt->rt_flags &= ~RTF_LLINFO; 1378 if (ln->ln_hold) 1379 m_freem(ln->ln_hold); 1380 Free((caddr_t)ln); 1381 } 1382 } 1383 1384 int 1385 nd6_ioctl(u_long cmd, caddr_t data, struct ifnet *ifp) 1386 { 1387 struct in6_drlist *drl = (struct in6_drlist *)data; 1388 struct in6_prlist *prl = (struct in6_prlist *)data; 1389 struct in6_ndireq *ndi = (struct in6_ndireq *)data; 1390 struct in6_nbrinfo *nbi = (struct in6_nbrinfo *)data; 1391 struct in6_ndifreq *ndif = (struct in6_ndifreq *)data; 1392 struct nd_defrouter *dr, any; 1393 struct nd_prefix *pr; 1394 struct rtentry *rt; 1395 int i = 0, error = 0; 1396 1397 switch (cmd) { 1398 case SIOCGDRLST_IN6: 1399 /* 1400 * obsolete API, use sysctl under net.inet6.icmp6 1401 */ 1402 bzero(drl, sizeof(*drl)); 1403 mtx_lock(&nd6_mtx); 1404 dr = TAILQ_FIRST(&nd_defrouter); 1405 while (dr && i < DRLSTSIZ) { 1406 drl->defrouter[i].rtaddr = dr->rtaddr; 1407 if (IN6_IS_ADDR_LINKLOCAL(&drl->defrouter[i].rtaddr)) { 1408 /* XXX: need to this hack for KAME stack */ 1409 drl->defrouter[i].rtaddr.s6_addr16[1] = 0; 1410 } else 1411 log(LOG_ERR, 1412 "default router list contains a " 1413 "non-linklocal address(%s)\n", 1414 ip6_sprintf(&drl->defrouter[i].rtaddr)); 1415 1416 drl->defrouter[i].flags = dr->flags; 1417 drl->defrouter[i].rtlifetime = dr->rtlifetime; 1418 drl->defrouter[i].expire = dr->expire; 1419 drl->defrouter[i].if_index = dr->ifp->if_index; 1420 i++; 1421 dr = TAILQ_NEXT(dr, dr_entry); 1422 } 1423 mtx_unlock(&nd6_mtx); 1424 break; 1425 case SIOCGPRLST_IN6: 1426 /* 1427 * obsolete API, use sysctl under net.inet6.icmp6 1428 */ 1429 /* 1430 * XXX meaning of fields, especialy "raflags", is very 1431 * differnet between RA prefix list and RR/static prefix list. 1432 * how about separating ioctls into two? 1433 */ 1434 bzero(prl, sizeof(*prl)); 1435 mtx_lock(&nd6_mtx); 1436 pr = nd_prefix.lh_first; 1437 while (pr && i < PRLSTSIZ) { 1438 struct nd_pfxrouter *pfr; 1439 int j; 1440 1441 in6_embedscope(&prl->prefix[i].prefix, 1442 &pr->ndpr_prefix, NULL, NULL); 1443 prl->prefix[i].raflags = pr->ndpr_raf; 1444 prl->prefix[i].prefixlen = pr->ndpr_plen; 1445 prl->prefix[i].vltime = pr->ndpr_vltime; 1446 prl->prefix[i].pltime = pr->ndpr_pltime; 1447 prl->prefix[i].if_index = pr->ndpr_ifp->if_index; 1448 prl->prefix[i].expire = pr->ndpr_expire; 1449 1450 pfr = pr->ndpr_advrtrs.lh_first; 1451 j = 0; 1452 while (pfr) { 1453 if (j < DRLSTSIZ) { 1454 #define RTRADDR prl->prefix[i].advrtr[j] 1455 RTRADDR = pfr->router->rtaddr; 1456 if (IN6_IS_ADDR_LINKLOCAL(&RTRADDR)) { 1457 /* XXX: hack for KAME */ 1458 RTRADDR.s6_addr16[1] = 0; 1459 } else 1460 log(LOG_ERR, 1461 "a router(%s) advertises " 1462 "a prefix with " 1463 "non-link local address\n", 1464 ip6_sprintf(&RTRADDR)); 1465 #undef RTRADDR 1466 } 1467 j++; 1468 pfr = pfr->pfr_next; 1469 } 1470 prl->prefix[i].advrtrs = j; 1471 prl->prefix[i].origin = PR_ORIG_RA; 1472 1473 i++; 1474 pr = pr->ndpr_next; 1475 } 1476 mtx_unlock(&nd6_mtx); 1477 1478 break; 1479 case OSIOCGIFINFO_IN6: 1480 /* XXX: old ndp(8) assumes a positive value for linkmtu. */ 1481 bzero(&ndi->ndi, sizeof(ndi->ndi)); 1482 ndi->ndi.linkmtu = IN6_LINKMTU(ifp); 1483 ndi->ndi.maxmtu = ND_IFINFO(ifp)->maxmtu; 1484 ndi->ndi.basereachable = ND_IFINFO(ifp)->basereachable; 1485 ndi->ndi.reachable = ND_IFINFO(ifp)->reachable; 1486 ndi->ndi.retrans = ND_IFINFO(ifp)->retrans; 1487 ndi->ndi.flags = ND_IFINFO(ifp)->flags; 1488 ndi->ndi.recalctm = ND_IFINFO(ifp)->recalctm; 1489 ndi->ndi.chlim = ND_IFINFO(ifp)->chlim; 1490 break; 1491 case SIOCGIFINFO_IN6: 1492 ndi->ndi = *ND_IFINFO(ifp); 1493 ndi->ndi.linkmtu = IN6_LINKMTU(ifp); 1494 break; 1495 case SIOCSIFINFO_FLAGS: 1496 ND_IFINFO(ifp)->flags = ndi->ndi.flags; 1497 break; 1498 case SIOCSNDFLUSH_IN6: /* XXX: the ioctl name is confusing... */ 1499 /* flush default router list */ 1500 /* 1501 * xxx sumikawa: should not delete route if default 1502 * route equals to the top of default router list 1503 */ 1504 bzero(&any, sizeof(any)); 1505 defrouter_delreq(&any, 0); 1506 defrouter_select(); 1507 /* xxx sumikawa: flush prefix list */ 1508 break; 1509 case SIOCSPFXFLUSH_IN6: 1510 { 1511 /* flush all the prefix advertised by routers */ 1512 struct nd_prefix *pr, *next; 1513 1514 mtx_lock(&nd6_mtx); 1515 for (pr = nd_prefix.lh_first; pr; pr = next) { 1516 struct in6_ifaddr *ia, *ia_next; 1517 1518 next = pr->ndpr_next; 1519 1520 if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr)) 1521 continue; /* XXX */ 1522 1523 /* do we really have to remove addresses as well? */ 1524 for (ia = in6_ifaddr; ia; ia = ia_next) { 1525 /* ia might be removed. keep the next ptr. */ 1526 ia_next = ia->ia_next; 1527 1528 if (!(ia->ia6_flags & IN6_IFF_AUTOCONF)) 1529 continue; 1530 1531 if (ia->ia6_ndpr == pr) 1532 in6_purgeaddr(&ia->ia_ifa); 1533 } 1534 prelist_remove(pr); 1535 } 1536 mtx_unlock(&nd6_mtx); 1537 break; 1538 } 1539 case SIOCSRTRFLUSH_IN6: 1540 { 1541 /* flush all the default routers */ 1542 struct nd_defrouter *dr, *next; 1543 1544 mtx_lock(&nd6_mtx); 1545 if ((dr = TAILQ_FIRST(&nd_defrouter)) != NULL) { 1546 /* 1547 * The first entry of the list may be stored in 1548 * the routing table, so we'll delete it later. 1549 */ 1550 for (dr = TAILQ_NEXT(dr, dr_entry); dr; dr = next) { 1551 next = TAILQ_NEXT(dr, dr_entry); 1552 defrtrlist_del(dr); 1553 } 1554 defrtrlist_del(TAILQ_FIRST(&nd_defrouter)); 1555 } 1556 mtx_unlock(&nd6_mtx); 1557 break; 1558 } 1559 case SIOCGNBRINFO_IN6: 1560 { 1561 struct llinfo_nd6 *ln; 1562 struct in6_addr nb_addr = nbi->addr; /* make local for safety */ 1563 1564 /* 1565 * XXX: KAME specific hack for scoped addresses 1566 * XXXX: for other scopes than link-local? 1567 */ 1568 if (IN6_IS_ADDR_LINKLOCAL(&nbi->addr) || 1569 IN6_IS_ADDR_MC_LINKLOCAL(&nbi->addr)) { 1570 u_int16_t *idp = (u_int16_t *)&nb_addr.s6_addr[2]; 1571 1572 if (*idp == 0) 1573 *idp = htons(ifp->if_index); 1574 } 1575 1576 mtx_lock(&nd6_mtx); 1577 if ((rt = nd6_lookup(&nb_addr, 0, ifp)) == NULL) { 1578 error = EINVAL; 1579 mtx_unlock(&nd6_mtx); 1580 break; 1581 } 1582 ln = (struct llinfo_nd6 *)rt->rt_llinfo; 1583 nbi->state = ln->ln_state; 1584 nbi->asked = ln->ln_asked; 1585 nbi->isrouter = ln->ln_router; 1586 nbi->expire = ln->ln_expire; 1587 mtx_unlock(&nd6_mtx); 1588 1589 break; 1590 } 1591 case SIOCGDEFIFACE_IN6: /* XXX: should be implemented as a sysctl? */ 1592 ndif->ifindex = nd6_defifindex; 1593 break; 1594 case SIOCSDEFIFACE_IN6: /* XXX: should be implemented as a sysctl? */ 1595 return (nd6_setdefaultiface(ndif->ifindex)); 1596 } 1597 return (error); 1598 } 1599 1600 /* 1601 * Create neighbor cache entry and cache link-layer address, 1602 * on reception of inbound ND6 packets. (RS/RA/NS/redirect) 1603 */ 1604 struct rtentry * 1605 nd6_cache_lladdr(struct ifnet *ifp, struct in6_addr *from, char *lladdr, 1606 int lladdrlen, 1607 int type, /* ICMP6 type */ 1608 int code /* type dependent information */) 1609 { 1610 struct rtentry *rt = NULL; 1611 struct llinfo_nd6 *ln = NULL; 1612 int is_newentry; 1613 struct sockaddr_dl *sdl = NULL; 1614 int do_update; 1615 int olladdr; 1616 int llchange; 1617 int newstate = 0; 1618 1619 if (!ifp) 1620 panic("ifp == NULL in nd6_cache_lladdr"); 1621 if (!from) 1622 panic("from == NULL in nd6_cache_lladdr"); 1623 1624 /* nothing must be updated for unspecified address */ 1625 if (IN6_IS_ADDR_UNSPECIFIED(from)) 1626 return NULL; 1627 1628 /* 1629 * Validation about ifp->if_addrlen and lladdrlen must be done in 1630 * the caller. 1631 * 1632 * XXX If the link does not have link-layer adderss, what should 1633 * we do? (ifp->if_addrlen == 0) 1634 * Spec says nothing in sections for RA, RS and NA. There's small 1635 * description on it in NS section (RFC 2461 7.2.3). 1636 */ 1637 1638 rt = nd6_lookup(from, 0, ifp); 1639 if (!rt) { 1640 #if 0 1641 /* nothing must be done if there's no lladdr */ 1642 if (!lladdr || !lladdrlen) 1643 return NULL; 1644 #endif 1645 1646 rt = nd6_lookup(from, 1, ifp); 1647 is_newentry = 1; 1648 } else { 1649 /* do nothing if static ndp is set */ 1650 if (rt->rt_flags & RTF_STATIC) 1651 return NULL; 1652 is_newentry = 0; 1653 } 1654 1655 if (!rt) 1656 return NULL; 1657 if ((rt->rt_flags & (RTF_GATEWAY | RTF_LLINFO)) != RTF_LLINFO) { 1658 fail: 1659 nd6_free(rt); 1660 return NULL; 1661 } 1662 ln = (struct llinfo_nd6 *)rt->rt_llinfo; 1663 if (!ln) 1664 goto fail; 1665 if (!rt->rt_gateway) 1666 goto fail; 1667 if (rt->rt_gateway->sa_family != AF_LINK) 1668 goto fail; 1669 sdl = SDL(rt->rt_gateway); 1670 1671 olladdr = (sdl->sdl_alen) ? 1 : 0; 1672 if (olladdr && lladdr) { 1673 if (bcmp(lladdr, LLADDR(sdl), ifp->if_addrlen)) 1674 llchange = 1; 1675 else 1676 llchange = 0; 1677 } else 1678 llchange = 0; 1679 1680 /* 1681 * newentry olladdr lladdr llchange (*=record) 1682 * 0 n n -- (1) 1683 * 0 y n -- (2) 1684 * 0 n y -- (3) * STALE 1685 * 0 y y n (4) * 1686 * 0 y y y (5) * STALE 1687 * 1 -- n -- (6) NOSTATE(= PASSIVE) 1688 * 1 -- y -- (7) * STALE 1689 */ 1690 1691 if (lladdr) { /* (3-5) and (7) */ 1692 /* 1693 * Record source link-layer address 1694 * XXX is it dependent to ifp->if_type? 1695 */ 1696 sdl->sdl_alen = ifp->if_addrlen; 1697 bcopy(lladdr, LLADDR(sdl), ifp->if_addrlen); 1698 } 1699 1700 if (!is_newentry) { 1701 if ((!olladdr && lladdr) || /* (3) */ 1702 (olladdr && lladdr && llchange)) { /* (5) */ 1703 do_update = 1; 1704 newstate = ND6_LLINFO_STALE; 1705 } else { /* (1-2,4) */ 1706 do_update = 0; 1707 } 1708 } else { 1709 do_update = 1; 1710 if (!lladdr) /* (6) */ 1711 newstate = ND6_LLINFO_NOSTATE; 1712 else /* (7) */ 1713 newstate = ND6_LLINFO_STALE; 1714 } 1715 1716 if (do_update) { 1717 /* 1718 * Update the state of the neighbor cache. 1719 */ 1720 ln->ln_state = newstate; 1721 1722 if (ln->ln_state == ND6_LLINFO_STALE) { 1723 /* 1724 * XXX: since nd6_output() below will cause 1725 * state tansition to DELAY and reset the timer, 1726 * we must set the timer now, although it is actually 1727 * meaningless. 1728 */ 1729 ln->ln_expire = time_uptime + nd6_gctimer; 1730 1731 if (ln->ln_hold) { 1732 /* 1733 * we assume ifp is not a p2p here, so just 1734 * set the 2nd argument as the 1st one. 1735 */ 1736 nd6_output(ifp, ifp, ln->ln_hold, 1737 (struct sockaddr_in6 *)rt_key(rt), rt); 1738 ln->ln_hold = NULL; 1739 } 1740 } else if (ln->ln_state == ND6_LLINFO_INCOMPLETE) { 1741 /* probe right away */ 1742 ln->ln_expire = time_uptime; 1743 } 1744 } 1745 1746 /* 1747 * ICMP6 type dependent behavior. 1748 * 1749 * NS: clear IsRouter if new entry 1750 * RS: clear IsRouter 1751 * RA: set IsRouter if there's lladdr 1752 * redir: clear IsRouter if new entry 1753 * 1754 * RA case, (1): 1755 * The spec says that we must set IsRouter in the following cases: 1756 * - If lladdr exist, set IsRouter. This means (1-5). 1757 * - If it is old entry (!newentry), set IsRouter. This means (7). 1758 * So, based on the spec, in (1-5) and (7) cases we must set IsRouter. 1759 * A quetion arises for (1) case. (1) case has no lladdr in the 1760 * neighbor cache, this is similar to (6). 1761 * This case is rare but we figured that we MUST NOT set IsRouter. 1762 * 1763 * newentry olladdr lladdr llchange NS RS RA redir 1764 * D R 1765 * 0 n n -- (1) c ? s 1766 * 0 y n -- (2) c s s 1767 * 0 n y -- (3) c s s 1768 * 0 y y n (4) c s s 1769 * 0 y y y (5) c s s 1770 * 1 -- n -- (6) c c c s 1771 * 1 -- y -- (7) c c s c s 1772 * 1773 * (c=clear s=set) 1774 */ 1775 switch (type & 0xff) { 1776 case ND_NEIGHBOR_SOLICIT: 1777 /* 1778 * New entry must have is_router flag cleared. 1779 */ 1780 if (is_newentry) /* (6-7) */ 1781 ln->ln_router = 0; 1782 break; 1783 case ND_REDIRECT: 1784 /* 1785 * If the icmp is a redirect to a better router, always set the 1786 * is_router flag. Otherwise, if the entry is newly created, 1787 * clear the flag. [RFC 2461, sec 8.3] 1788 */ 1789 if (code == ND_REDIRECT_ROUTER) 1790 ln->ln_router = 1; 1791 else if (is_newentry) /* (6-7) */ 1792 ln->ln_router = 0; 1793 break; 1794 case ND_ROUTER_SOLICIT: 1795 /* 1796 * is_router flag must always be cleared. 1797 */ 1798 ln->ln_router = 0; 1799 break; 1800 case ND_ROUTER_ADVERT: 1801 /* 1802 * Mark an entry with lladdr as a router. 1803 */ 1804 if ((!is_newentry && (olladdr || lladdr)) || /* (2-5) */ 1805 (is_newentry && lladdr)) { /* (7) */ 1806 ln->ln_router = 1; 1807 } 1808 break; 1809 } 1810 1811 /* 1812 * When the link-layer address of a router changes, select the 1813 * best router again. In particular, when the neighbor entry is newly 1814 * created, it might affect the selection policy. 1815 * Question: can we restrict the first condition to the "is_newentry" 1816 * case? 1817 * XXX: when we hear an RA from a new router with the link-layer 1818 * address option, defrouter_select() is called twice, since 1819 * defrtrlist_update called the function as well. However, I believe 1820 * we can compromise the overhead, since it only happens the first 1821 * time. 1822 * XXX: although defrouter_select() should not have a bad effect 1823 * for those are not autoconfigured hosts, we explicitly avoid such 1824 * cases for safety. 1825 */ 1826 if (do_update && ln->ln_router && !ip6_forwarding && ip6_accept_rtadv) 1827 defrouter_select(); 1828 1829 return rt; 1830 } 1831 1832 static void 1833 nd6_slowtimo(void *arg __unused) 1834 { 1835 struct lwkt_msg *lmsg = &nd6_slowtimo_netmsg.lmsg; 1836 1837 KASSERT(mycpuid == 0, ("not on cpu0")); 1838 crit_enter(); 1839 if (lmsg->ms_flags & MSGF_DONE) 1840 lwkt_sendmsg_oncpu(netisr_cpuport(0), lmsg); 1841 crit_exit(); 1842 } 1843 1844 static void 1845 nd6_slowtimo_dispatch(netmsg_t nmsg) 1846 { 1847 const struct ifnet_array *arr; 1848 struct nd_ifinfo *nd6if; 1849 int i; 1850 1851 ASSERT_NETISR0; 1852 1853 crit_enter(); 1854 lwkt_replymsg(&nmsg->lmsg, 0); /* reply ASAP */ 1855 crit_exit(); 1856 1857 arr = ifnet_array_get(); 1858 1859 mtx_lock(&nd6_mtx); 1860 for (i = 0; i < arr->ifnet_count; ++i) { 1861 struct ifnet *ifp = arr->ifnet_arr[i]; 1862 1863 if (ifp->if_afdata[AF_INET6] == NULL) 1864 continue; 1865 nd6if = ND_IFINFO(ifp); 1866 if (nd6if->basereachable && /* already initialized */ 1867 (nd6if->recalctm -= ND6_SLOWTIMER_INTERVAL) <= 0) { 1868 /* 1869 * Since reachable time rarely changes by router 1870 * advertisements, we SHOULD insure that a new random 1871 * value gets recomputed at least once every few hours. 1872 * (RFC 2461, 6.3.4) 1873 */ 1874 nd6if->recalctm = nd6_recalc_reachtm_interval; 1875 nd6if->reachable = ND_COMPUTE_RTIME(nd6if->basereachable); 1876 } 1877 } 1878 mtx_unlock(&nd6_mtx); 1879 1880 callout_reset(&nd6_slowtimo_ch, ND6_SLOWTIMER_INTERVAL * hz, 1881 nd6_slowtimo, NULL); 1882 } 1883 1884 #define gotoerr(e) { error = (e); goto bad;} 1885 1886 int 1887 nd6_output(struct ifnet *ifp, struct ifnet *origifp, struct mbuf *m, 1888 struct sockaddr_in6 *dst, struct rtentry *rt) 1889 { 1890 struct llinfo_nd6 *ln = NULL; 1891 int error = 0; 1892 1893 if (IN6_IS_ADDR_MULTICAST(&dst->sin6_addr)) 1894 goto sendpkt; 1895 1896 if (nd6_need_cache(ifp) == 0) 1897 goto sendpkt; 1898 1899 /* 1900 * Next hop determination. This routine is derived from rt_llroute. 1901 */ 1902 if (rt != NULL) { 1903 if (!(rt->rt_flags & RTF_UP)) { 1904 rt = rtlookup((struct sockaddr *)dst); 1905 if (rt == NULL) 1906 gotoerr(EHOSTUNREACH); 1907 rt->rt_refcnt--; 1908 if (rt->rt_ifp != ifp) { 1909 /* XXX: loop care? */ 1910 return nd6_output(ifp, origifp, m, dst, rt); 1911 } 1912 } 1913 if (rt->rt_flags & RTF_GATEWAY) { 1914 struct sockaddr_in6 *gw6; 1915 1916 /* 1917 * We skip link-layer address resolution and NUD 1918 * if the gateway is not a neighbor from ND point 1919 * of view, regardless of the value of nd_ifinfo.flags. 1920 * The second condition is a bit tricky; we skip 1921 * if the gateway is our own address, which is 1922 * sometimes used to install a route to a p2p link. 1923 */ 1924 gw6 = (struct sockaddr_in6 *)rt->rt_gateway; 1925 if (!nd6_is_addr_neighbor(gw6, ifp) || 1926 in6ifa_ifpwithaddr(ifp, &gw6->sin6_addr)) { 1927 /* 1928 * We allow this kind of tricky route only 1929 * when the outgoing interface is p2p. 1930 * XXX: we may need a more generic rule here. 1931 */ 1932 if (!(ifp->if_flags & IFF_POINTOPOINT)) 1933 gotoerr(EHOSTUNREACH); 1934 1935 goto sendpkt; 1936 } 1937 1938 if (rt->rt_gwroute == NULL) { 1939 rt->rt_gwroute = rtlookup(rt->rt_gateway); 1940 if (rt->rt_gwroute == NULL) 1941 gotoerr(EHOSTUNREACH); 1942 } else if (!(rt->rt_gwroute->rt_flags & RTF_UP)) { 1943 rtfree(rt->rt_gwroute); 1944 rt->rt_gwroute = rtlookup(rt->rt_gateway); 1945 if (rt->rt_gwroute == NULL) 1946 gotoerr(EHOSTUNREACH); 1947 } 1948 rt = rt->rt_gwroute; 1949 } 1950 } 1951 1952 /* 1953 * Address resolution or Neighbor Unreachability Detection 1954 * for the next hop. 1955 * At this point, the destination of the packet must be a unicast 1956 * or an anycast address(i.e. not a multicast). 1957 */ 1958 1959 /* Look up the neighbor cache for the nexthop */ 1960 if (rt && (rt->rt_flags & RTF_LLINFO)) 1961 ln = (struct llinfo_nd6 *)rt->rt_llinfo; 1962 else { 1963 /* 1964 * Since nd6_is_addr_neighbor() internally calls nd6_lookup(), 1965 * the condition below is not very efficient. But we believe 1966 * it is tolerable, because this should be a rare case. 1967 */ 1968 if (nd6_is_addr_neighbor(dst, ifp) && 1969 (rt = nd6_lookup(&dst->sin6_addr, 1, ifp)) != NULL) 1970 ln = (struct llinfo_nd6 *)rt->rt_llinfo; 1971 } 1972 if (!ln || !rt) { 1973 if (!(ifp->if_flags & IFF_POINTOPOINT) && 1974 !(ND_IFINFO(ifp)->flags & ND6_IFF_PERFORMNUD)) { 1975 log(LOG_DEBUG, 1976 "nd6_output: can't allocate llinfo for %s " 1977 "(ln=%p, rt=%p)\n", 1978 ip6_sprintf(&dst->sin6_addr), ln, rt); 1979 gotoerr(EIO); /* XXX: good error? */ 1980 } 1981 1982 goto sendpkt; /* send anyway */ 1983 } 1984 1985 /* We don't have to do link-layer address resolution on a p2p link. */ 1986 if ((ifp->if_flags & IFF_POINTOPOINT) && 1987 ln->ln_state < ND6_LLINFO_REACHABLE) { 1988 ln->ln_state = ND6_LLINFO_STALE; 1989 ln->ln_expire = time_uptime + nd6_gctimer; 1990 } 1991 1992 /* 1993 * The first time we send a packet to a neighbor whose entry is 1994 * STALE, we have to change the state to DELAY and a sets a timer to 1995 * expire in DELAY_FIRST_PROBE_TIME seconds to ensure do 1996 * neighbor unreachability detection on expiration. 1997 * (RFC 2461 7.3.3) 1998 */ 1999 if (ln->ln_state == ND6_LLINFO_STALE) { 2000 ln->ln_asked = 0; 2001 ln->ln_state = ND6_LLINFO_DELAY; 2002 ln->ln_expire = time_uptime + nd6_delay; 2003 } 2004 2005 /* 2006 * If the neighbor cache entry has a state other than INCOMPLETE 2007 * (i.e. its link-layer address is already resolved), just 2008 * send the packet. 2009 */ 2010 if (ln->ln_state > ND6_LLINFO_INCOMPLETE) 2011 goto sendpkt; 2012 2013 /* 2014 * There is a neighbor cache entry, but no ethernet address 2015 * response yet. Replace the held mbuf (if any) with this 2016 * latest one. 2017 * 2018 * This code conforms to the rate-limiting rule described in Section 2019 * 7.2.2 of RFC 2461, because the timer is set correctly after sending 2020 * an NS below. 2021 */ 2022 if (ln->ln_state == ND6_LLINFO_NOSTATE) { 2023 /* 2024 * This neighbor cache entry was just created; change its 2025 * state to INCOMPLETE and start its life cycle. 2026 * 2027 * We force an NS output below by setting ln_expire to 1 2028 * (nd6_rtrequest() could set it to the current time_uptime) 2029 * and zeroing out ln_asked (XXX this may not be necessary). 2030 */ 2031 ln->ln_state = ND6_LLINFO_INCOMPLETE; 2032 ln->ln_expire = 1; 2033 ln->ln_asked = 0; 2034 } 2035 if (ln->ln_hold) 2036 m_freem(ln->ln_hold); 2037 ln->ln_hold = m; 2038 if (ln->ln_expire) { 2039 if (ln->ln_asked < nd6_mmaxtries && 2040 ln->ln_expire < time_uptime) { 2041 ln->ln_asked++; 2042 ln->ln_expire = time_uptime + 2043 ND_IFINFO(ifp)->retrans / 1000; 2044 nd6_ns_output(ifp, NULL, &dst->sin6_addr, ln, 0); 2045 } 2046 } 2047 return (0); 2048 2049 sendpkt: 2050 if (ifp->if_flags & IFF_LOOPBACK) 2051 error = ifp->if_output(origifp, m, (struct sockaddr *)dst, rt); 2052 else 2053 error = ifp->if_output(ifp, m, (struct sockaddr *)dst, rt); 2054 return (error); 2055 2056 bad: 2057 m_freem(m); 2058 return (error); 2059 } 2060 #undef gotoerr 2061 2062 int 2063 nd6_need_cache(struct ifnet *ifp) 2064 { 2065 /* 2066 * XXX: we currently do not make neighbor cache on any interface 2067 * other than Ethernet and GIF. 2068 * 2069 * RFC2893 says: 2070 * - unidirectional tunnels needs no ND 2071 */ 2072 switch (ifp->if_type) { 2073 case IFT_ETHER: 2074 case IFT_IEEE1394: 2075 #ifdef IFT_L2VLAN 2076 case IFT_L2VLAN: 2077 #endif 2078 #ifdef IFT_IEEE80211 2079 case IFT_IEEE80211: 2080 #endif 2081 #ifdef IFT_CARP 2082 case IFT_CARP: 2083 #endif 2084 case IFT_GIF: /* XXX need more cases? */ 2085 return (1); 2086 default: 2087 return (0); 2088 } 2089 } 2090 2091 int 2092 nd6_storelladdr(struct ifnet *ifp, struct rtentry *rt0, struct mbuf *m, 2093 struct sockaddr *dst, u_char *desten) 2094 { 2095 struct sockaddr_dl *sdl; 2096 struct rtentry *rt; 2097 2098 2099 if (m->m_flags & M_MCAST) { 2100 switch (ifp->if_type) { 2101 case IFT_ETHER: 2102 #ifdef IFT_L2VLAN 2103 case IFT_L2VLAN: 2104 #endif 2105 #ifdef IFT_IEEE80211 2106 case IFT_IEEE80211: 2107 #endif 2108 ETHER_MAP_IPV6_MULTICAST(&SIN6(dst)->sin6_addr, 2109 desten); 2110 return (1); 2111 case IFT_IEEE1394: 2112 bcopy(ifp->if_broadcastaddr, desten, ifp->if_addrlen); 2113 return (1); 2114 default: 2115 m_freem(m); 2116 return (0); 2117 } 2118 } 2119 if (rt0 == NULL) { 2120 /* this could happen, if we could not allocate memory */ 2121 m_freem(m); 2122 return (0); 2123 } 2124 if (rt_llroute(dst, rt0, &rt) != 0) { 2125 m_freem(m); 2126 return (0); 2127 } 2128 if (rt->rt_gateway->sa_family != AF_LINK) { 2129 kprintf("nd6_storelladdr: something odd happens\n"); 2130 m_freem(m); 2131 return (0); 2132 } 2133 sdl = SDL(rt->rt_gateway); 2134 if (sdl->sdl_alen == 0) { 2135 /* this should be impossible, but we bark here for debugging */ 2136 kprintf("nd6_storelladdr: sdl_alen == 0\n"); 2137 m_freem(m); 2138 return (0); 2139 } 2140 2141 bcopy(LLADDR(sdl), desten, sdl->sdl_alen); 2142 return (1); 2143 } 2144 2145 static int nd6_sysctl_drlist(SYSCTL_HANDLER_ARGS); 2146 static int nd6_sysctl_prlist(SYSCTL_HANDLER_ARGS); 2147 #ifdef SYSCTL_DECL 2148 SYSCTL_DECL(_net_inet6_icmp6); 2149 #endif 2150 SYSCTL_NODE(_net_inet6_icmp6, ICMPV6CTL_ND6_DRLIST, nd6_drlist, 2151 CTLFLAG_RD, nd6_sysctl_drlist, "List default routers"); 2152 SYSCTL_NODE(_net_inet6_icmp6, ICMPV6CTL_ND6_PRLIST, nd6_prlist, 2153 CTLFLAG_RD, nd6_sysctl_prlist, "List prefixes"); 2154 2155 static int 2156 nd6_sysctl_drlist(SYSCTL_HANDLER_ARGS) 2157 { 2158 int error; 2159 char buf[1024]; 2160 struct in6_defrouter *d, *de; 2161 struct nd_defrouter *dr; 2162 2163 if (req->newptr) 2164 return EPERM; 2165 error = 0; 2166 2167 for (dr = TAILQ_FIRST(&nd_defrouter); dr; 2168 dr = TAILQ_NEXT(dr, dr_entry)) { 2169 d = (struct in6_defrouter *)buf; 2170 de = (struct in6_defrouter *)(buf + sizeof(buf)); 2171 2172 if (d + 1 <= de) { 2173 bzero(d, sizeof(*d)); 2174 d->rtaddr.sin6_family = AF_INET6; 2175 d->rtaddr.sin6_len = sizeof(d->rtaddr); 2176 if (in6_recoverscope(&d->rtaddr, &dr->rtaddr, 2177 dr->ifp) != 0) 2178 log(LOG_ERR, 2179 "scope error in " 2180 "default router list (%s)\n", 2181 ip6_sprintf(&dr->rtaddr)); 2182 d->flags = dr->flags; 2183 d->rtlifetime = dr->rtlifetime; 2184 d->expire = dr->expire; 2185 d->if_index = dr->ifp->if_index; 2186 } else 2187 panic("buffer too short"); 2188 2189 error = SYSCTL_OUT(req, buf, sizeof(*d)); 2190 if (error) 2191 break; 2192 } 2193 return error; 2194 } 2195 2196 static int 2197 nd6_sysctl_prlist(SYSCTL_HANDLER_ARGS) 2198 { 2199 int error; 2200 char buf[1024]; 2201 struct in6_prefix *p, *pe; 2202 struct nd_prefix *pr; 2203 2204 if (req->newptr) 2205 return EPERM; 2206 error = 0; 2207 2208 for (pr = nd_prefix.lh_first; pr; pr = pr->ndpr_next) { 2209 u_short advrtrs; 2210 size_t advance; 2211 struct sockaddr_in6 *sin6, *s6; 2212 struct nd_pfxrouter *pfr; 2213 2214 p = (struct in6_prefix *)buf; 2215 pe = (struct in6_prefix *)(buf + sizeof(buf)); 2216 2217 if (p + 1 <= pe) { 2218 bzero(p, sizeof(*p)); 2219 sin6 = (struct sockaddr_in6 *)(p + 1); 2220 2221 p->prefix = pr->ndpr_prefix; 2222 if (in6_recoverscope(&p->prefix, 2223 &p->prefix.sin6_addr, pr->ndpr_ifp) != 0) 2224 log(LOG_ERR, 2225 "scope error in prefix list (%s)\n", 2226 ip6_sprintf(&p->prefix.sin6_addr)); 2227 p->raflags = pr->ndpr_raf; 2228 p->prefixlen = pr->ndpr_plen; 2229 p->vltime = pr->ndpr_vltime; 2230 p->pltime = pr->ndpr_pltime; 2231 p->if_index = pr->ndpr_ifp->if_index; 2232 p->expire = pr->ndpr_expire; 2233 p->refcnt = pr->ndpr_refcnt; 2234 p->flags = pr->ndpr_stateflags; 2235 p->origin = PR_ORIG_RA; 2236 advrtrs = 0; 2237 for (pfr = pr->ndpr_advrtrs.lh_first; pfr; 2238 pfr = pfr->pfr_next) { 2239 if ((void *)&sin6[advrtrs + 1] > (void *)pe) { 2240 advrtrs++; 2241 continue; 2242 } 2243 s6 = &sin6[advrtrs]; 2244 bzero(s6, sizeof(*s6)); 2245 s6->sin6_family = AF_INET6; 2246 s6->sin6_len = sizeof(*sin6); 2247 if (in6_recoverscope(s6, &pfr->router->rtaddr, 2248 pfr->router->ifp) != 0) 2249 log(LOG_ERR, 2250 "scope error in " 2251 "prefix list (%s)\n", 2252 ip6_sprintf(&pfr->router->rtaddr)); 2253 advrtrs++; 2254 } 2255 p->advrtrs = advrtrs; 2256 } else { 2257 panic("buffer too short"); 2258 } 2259 2260 advance = sizeof(*p) + sizeof(*sin6) * advrtrs; 2261 error = SYSCTL_OUT(req, buf, advance); 2262 if (error) 2263 break; 2264 } 2265 return error; 2266 } 2267