1 /*- 2 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project. 3 * All rights reserved. 4 * 5 * Redistribution and use in source and binary forms, with or without 6 * modification, are permitted provided that the following conditions 7 * are met: 8 * 1. Redistributions of source code must retain the above copyright 9 * notice, this list of conditions and the following disclaimer. 10 * 2. Redistributions in binary form must reproduce the above copyright 11 * notice, this list of conditions and the following disclaimer in the 12 * documentation and/or other materials provided with the distribution. 13 * 3. Neither the name of the project nor the names of its contributors 14 * may be used to endorse or promote products derived from this software 15 * without specific prior written permission. 16 * 17 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND 18 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 19 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 20 * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE 21 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 22 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 23 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 24 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 25 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 26 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 27 * SUCH DAMAGE. 28 * 29 * $KAME: nd6.c,v 1.144 2001/05/24 07:44:00 itojun Exp $ 30 */ 31 32 #include <sys/cdefs.h> 33 __FBSDID("$FreeBSD$"); 34 35 #include "opt_inet.h" 36 #include "opt_inet6.h" 37 #include "opt_mac.h" 38 39 #include <sys/param.h> 40 #include <sys/systm.h> 41 #include <sys/callout.h> 42 #include <sys/malloc.h> 43 #include <sys/mbuf.h> 44 #include <sys/socket.h> 45 #include <sys/sockio.h> 46 #include <sys/time.h> 47 #include <sys/kernel.h> 48 #include <sys/protosw.h> 49 #include <sys/errno.h> 50 #include <sys/syslog.h> 51 #include <sys/queue.h> 52 #include <sys/sysctl.h> 53 54 #include <net/if.h> 55 #include <net/if_arc.h> 56 #include <net/if_dl.h> 57 #include <net/if_types.h> 58 #include <net/iso88025.h> 59 #include <net/fddi.h> 60 #include <net/route.h> 61 #include <net/vnet.h> 62 63 #include <netinet/in.h> 64 #include <netinet/if_ether.h> 65 #include <netinet6/in6_var.h> 66 #include <netinet/ip6.h> 67 #include <netinet6/ip6_var.h> 68 #include <netinet6/scope6_var.h> 69 #include <netinet6/nd6.h> 70 #include <netinet/icmp6.h> 71 #include <netinet6/vinet6.h> 72 73 #include <sys/limits.h> 74 #include <sys/vimage.h> 75 76 #include <security/mac/mac_framework.h> 77 78 #define ND6_SLOWTIMER_INTERVAL (60 * 60) /* 1 hour */ 79 #define ND6_RECALC_REACHTM_INTERVAL (60 * 120) /* 2 hours */ 80 81 #define SIN6(s) ((struct sockaddr_in6 *)s) 82 #define SDL(s) ((struct sockaddr_dl *)s) 83 84 #ifdef VIMAGE_GLOBALS 85 int nd6_prune; 86 int nd6_delay; 87 int nd6_umaxtries; 88 int nd6_mmaxtries; 89 int nd6_useloopback; 90 int nd6_gctimer; 91 92 /* preventing too many loops in ND option parsing */ 93 int nd6_maxndopt; 94 95 int nd6_maxnudhint; 96 int nd6_maxqueuelen; 97 98 int nd6_debug; 99 100 /* for debugging? */ 101 static int nd6_inuse, nd6_allocated; 102 struct llinfo_nd6 llinfo_nd6; 103 104 struct nd_drhead nd_defrouter; 105 struct nd_prhead nd_prefix; 106 107 int nd6_recalc_reachtm_interval; 108 #endif /* VIMAGE_GLOBALS */ 109 110 static struct sockaddr_in6 all1_sa; 111 112 static int nd6_is_new_addr_neighbor __P((struct sockaddr_in6 *, 113 struct ifnet *)); 114 static void nd6_setmtu0(struct ifnet *, struct nd_ifinfo *); 115 static void nd6_slowtimo(void *); 116 static int regen_tmpaddr(struct in6_ifaddr *); 117 static struct llinfo_nd6 *nd6_free(struct rtentry *, int); 118 static void nd6_llinfo_timer(void *); 119 static void clear_llinfo_pqueue(struct llinfo_nd6 *); 120 121 #ifdef VIMAGE_GLOBALS 122 struct callout nd6_slowtimo_ch; 123 struct callout nd6_timer_ch; 124 extern struct callout in6_tmpaddrtimer_ch; 125 extern int dad_ignore_ns; 126 extern int dad_maxtry; 127 #endif 128 129 void 130 nd6_init(void) 131 { 132 INIT_VNET_INET6(curvnet); 133 static int nd6_init_done = 0; 134 int i; 135 136 if (nd6_init_done) { 137 log(LOG_NOTICE, "nd6_init called more than once(ignored)\n"); 138 return; 139 } 140 141 V_nd6_prune = 1; /* walk list every 1 seconds */ 142 V_nd6_delay = 5; /* delay first probe time 5 second */ 143 V_nd6_umaxtries = 3; /* maximum unicast query */ 144 V_nd6_mmaxtries = 3; /* maximum multicast query */ 145 V_nd6_useloopback = 1; /* use loopback interface for local traffic */ 146 V_nd6_gctimer = (60 * 60 * 24); /* 1 day: garbage collection timer */ 147 148 /* preventing too many loops in ND option parsing */ 149 V_nd6_maxndopt = 10; /* max # of ND options allowed */ 150 151 V_nd6_maxnudhint = 0; /* max # of subsequent upper layer hints */ 152 V_nd6_maxqueuelen = 1; /* max pkts cached in unresolved ND entries */ 153 154 #ifdef ND6_DEBUG 155 V_nd6_debug = 1; 156 #else 157 V_nd6_debug = 0; 158 #endif 159 160 V_nd6_recalc_reachtm_interval = ND6_RECALC_REACHTM_INTERVAL; 161 162 V_dad_ignore_ns = 0; /* ignore NS in DAD - specwise incorrect*/ 163 V_dad_maxtry = 15; /* max # of *tries* to transmit DAD packet */ 164 165 V_llinfo_nd6.ln_next = &V_llinfo_nd6; 166 V_llinfo_nd6.ln_prev = &V_llinfo_nd6; 167 LIST_INIT(&V_nd_prefix); 168 169 ip6_use_tempaddr = 0; 170 ip6_temp_preferred_lifetime = DEF_TEMP_PREFERRED_LIFETIME; 171 ip6_temp_valid_lifetime = DEF_TEMP_VALID_LIFETIME; 172 ip6_temp_regen_advance = TEMPADDR_REGEN_ADVANCE; 173 174 all1_sa.sin6_family = AF_INET6; 175 all1_sa.sin6_len = sizeof(struct sockaddr_in6); 176 for (i = 0; i < sizeof(all1_sa.sin6_addr); i++) 177 all1_sa.sin6_addr.s6_addr[i] = 0xff; 178 179 /* initialization of the default router list */ 180 TAILQ_INIT(&V_nd_defrouter); 181 /* start timer */ 182 callout_init(&V_nd6_slowtimo_ch, 0); 183 callout_reset(&V_nd6_slowtimo_ch, ND6_SLOWTIMER_INTERVAL * hz, 184 nd6_slowtimo, NULL); 185 186 nd6_init_done = 1; 187 188 } 189 190 struct nd_ifinfo * 191 nd6_ifattach(struct ifnet *ifp) 192 { 193 struct nd_ifinfo *nd; 194 195 nd = (struct nd_ifinfo *)malloc(sizeof(*nd), M_IP6NDP, M_WAITOK); 196 bzero(nd, sizeof(*nd)); 197 198 nd->initialized = 1; 199 200 nd->chlim = IPV6_DEFHLIM; 201 nd->basereachable = REACHABLE_TIME; 202 nd->reachable = ND_COMPUTE_RTIME(nd->basereachable); 203 nd->retrans = RETRANS_TIMER; 204 /* 205 * Note that the default value of ip6_accept_rtadv is 0, which means 206 * we won't accept RAs by default even if we set ND6_IFF_ACCEPT_RTADV 207 * here. 208 */ 209 nd->flags = (ND6_IFF_PERFORMNUD | ND6_IFF_ACCEPT_RTADV); 210 211 /* XXX: we cannot call nd6_setmtu since ifp is not fully initialized */ 212 nd6_setmtu0(ifp, nd); 213 214 return nd; 215 } 216 217 void 218 nd6_ifdetach(struct nd_ifinfo *nd) 219 { 220 221 free(nd, M_IP6NDP); 222 } 223 224 /* 225 * Reset ND level link MTU. This function is called when the physical MTU 226 * changes, which means we might have to adjust the ND level MTU. 227 */ 228 void 229 nd6_setmtu(struct ifnet *ifp) 230 { 231 232 nd6_setmtu0(ifp, ND_IFINFO(ifp)); 233 } 234 235 /* XXX todo: do not maintain copy of ifp->if_mtu in ndi->maxmtu */ 236 void 237 nd6_setmtu0(struct ifnet *ifp, struct nd_ifinfo *ndi) 238 { 239 INIT_VNET_INET6(ifp->if_vnet); 240 u_int32_t omaxmtu; 241 242 omaxmtu = ndi->maxmtu; 243 244 switch (ifp->if_type) { 245 case IFT_ARCNET: 246 ndi->maxmtu = MIN(ARC_PHDS_MAXMTU, ifp->if_mtu); /* RFC2497 */ 247 break; 248 case IFT_FDDI: 249 ndi->maxmtu = MIN(FDDIIPMTU, ifp->if_mtu); /* RFC2467 */ 250 break; 251 case IFT_ISO88025: 252 ndi->maxmtu = MIN(ISO88025_MAX_MTU, ifp->if_mtu); 253 break; 254 default: 255 ndi->maxmtu = ifp->if_mtu; 256 break; 257 } 258 259 /* 260 * Decreasing the interface MTU under IPV6 minimum MTU may cause 261 * undesirable situation. We thus notify the operator of the change 262 * explicitly. The check for omaxmtu is necessary to restrict the 263 * log to the case of changing the MTU, not initializing it. 264 */ 265 if (omaxmtu >= IPV6_MMTU && ndi->maxmtu < IPV6_MMTU) { 266 log(LOG_NOTICE, "nd6_setmtu0: " 267 "new link MTU on %s (%lu) is too small for IPv6\n", 268 if_name(ifp), (unsigned long)ndi->maxmtu); 269 } 270 271 if (ndi->maxmtu > V_in6_maxmtu) 272 in6_setmaxmtu(); /* check all interfaces just in case */ 273 274 #undef MIN 275 } 276 277 void 278 nd6_option_init(void *opt, int icmp6len, union nd_opts *ndopts) 279 { 280 281 bzero(ndopts, sizeof(*ndopts)); 282 ndopts->nd_opts_search = (struct nd_opt_hdr *)opt; 283 ndopts->nd_opts_last 284 = (struct nd_opt_hdr *)(((u_char *)opt) + icmp6len); 285 286 if (icmp6len == 0) { 287 ndopts->nd_opts_done = 1; 288 ndopts->nd_opts_search = NULL; 289 } 290 } 291 292 /* 293 * Take one ND option. 294 */ 295 struct nd_opt_hdr * 296 nd6_option(union nd_opts *ndopts) 297 { 298 struct nd_opt_hdr *nd_opt; 299 int olen; 300 301 if (ndopts == NULL) 302 panic("ndopts == NULL in nd6_option"); 303 if (ndopts->nd_opts_last == NULL) 304 panic("uninitialized ndopts in nd6_option"); 305 if (ndopts->nd_opts_search == NULL) 306 return NULL; 307 if (ndopts->nd_opts_done) 308 return NULL; 309 310 nd_opt = ndopts->nd_opts_search; 311 312 /* make sure nd_opt_len is inside the buffer */ 313 if ((caddr_t)&nd_opt->nd_opt_len >= (caddr_t)ndopts->nd_opts_last) { 314 bzero(ndopts, sizeof(*ndopts)); 315 return NULL; 316 } 317 318 olen = nd_opt->nd_opt_len << 3; 319 if (olen == 0) { 320 /* 321 * Message validation requires that all included 322 * options have a length that is greater than zero. 323 */ 324 bzero(ndopts, sizeof(*ndopts)); 325 return NULL; 326 } 327 328 ndopts->nd_opts_search = (struct nd_opt_hdr *)((caddr_t)nd_opt + olen); 329 if (ndopts->nd_opts_search > ndopts->nd_opts_last) { 330 /* option overruns the end of buffer, invalid */ 331 bzero(ndopts, sizeof(*ndopts)); 332 return NULL; 333 } else if (ndopts->nd_opts_search == ndopts->nd_opts_last) { 334 /* reached the end of options chain */ 335 ndopts->nd_opts_done = 1; 336 ndopts->nd_opts_search = NULL; 337 } 338 return nd_opt; 339 } 340 341 /* 342 * Parse multiple ND options. 343 * This function is much easier to use, for ND routines that do not need 344 * multiple options of the same type. 345 */ 346 int 347 nd6_options(union nd_opts *ndopts) 348 { 349 INIT_VNET_INET6(curvnet); 350 struct nd_opt_hdr *nd_opt; 351 int i = 0; 352 353 if (ndopts == NULL) 354 panic("ndopts == NULL in nd6_options"); 355 if (ndopts->nd_opts_last == NULL) 356 panic("uninitialized ndopts in nd6_options"); 357 if (ndopts->nd_opts_search == NULL) 358 return 0; 359 360 while (1) { 361 nd_opt = nd6_option(ndopts); 362 if (nd_opt == NULL && ndopts->nd_opts_last == NULL) { 363 /* 364 * Message validation requires that all included 365 * options have a length that is greater than zero. 366 */ 367 V_icmp6stat.icp6s_nd_badopt++; 368 bzero(ndopts, sizeof(*ndopts)); 369 return -1; 370 } 371 372 if (nd_opt == NULL) 373 goto skip1; 374 375 switch (nd_opt->nd_opt_type) { 376 case ND_OPT_SOURCE_LINKADDR: 377 case ND_OPT_TARGET_LINKADDR: 378 case ND_OPT_MTU: 379 case ND_OPT_REDIRECTED_HEADER: 380 if (ndopts->nd_opt_array[nd_opt->nd_opt_type]) { 381 nd6log((LOG_INFO, 382 "duplicated ND6 option found (type=%d)\n", 383 nd_opt->nd_opt_type)); 384 /* XXX bark? */ 385 } else { 386 ndopts->nd_opt_array[nd_opt->nd_opt_type] 387 = nd_opt; 388 } 389 break; 390 case ND_OPT_PREFIX_INFORMATION: 391 if (ndopts->nd_opt_array[nd_opt->nd_opt_type] == 0) { 392 ndopts->nd_opt_array[nd_opt->nd_opt_type] 393 = nd_opt; 394 } 395 ndopts->nd_opts_pi_end = 396 (struct nd_opt_prefix_info *)nd_opt; 397 break; 398 default: 399 /* 400 * Unknown options must be silently ignored, 401 * to accomodate future extension to the protocol. 402 */ 403 nd6log((LOG_DEBUG, 404 "nd6_options: unsupported option %d - " 405 "option ignored\n", nd_opt->nd_opt_type)); 406 } 407 408 skip1: 409 i++; 410 if (i > V_nd6_maxndopt) { 411 V_icmp6stat.icp6s_nd_toomanyopt++; 412 nd6log((LOG_INFO, "too many loop in nd opt\n")); 413 break; 414 } 415 416 if (ndopts->nd_opts_done) 417 break; 418 } 419 420 return 0; 421 } 422 423 /* 424 * ND6 timer routine to handle ND6 entries 425 */ 426 void 427 nd6_llinfo_settimer(struct llinfo_nd6 *ln, long tick) 428 { 429 if (tick < 0) { 430 ln->ln_expire = 0; 431 ln->ln_ntick = 0; 432 callout_stop(&ln->ln_timer_ch); 433 } else { 434 ln->ln_expire = time_second + tick / hz; 435 if (tick > INT_MAX) { 436 ln->ln_ntick = tick - INT_MAX; 437 callout_reset(&ln->ln_timer_ch, INT_MAX, 438 nd6_llinfo_timer, ln); 439 } else { 440 ln->ln_ntick = 0; 441 callout_reset(&ln->ln_timer_ch, tick, 442 nd6_llinfo_timer, ln); 443 } 444 } 445 } 446 447 static void 448 nd6_llinfo_timer(void *arg) 449 { 450 struct llinfo_nd6 *ln; 451 struct rtentry *rt; 452 struct in6_addr *dst; 453 struct ifnet *ifp; 454 struct nd_ifinfo *ndi = NULL; 455 456 ln = (struct llinfo_nd6 *)arg; 457 458 if (ln->ln_ntick > 0) { 459 if (ln->ln_ntick > INT_MAX) { 460 ln->ln_ntick -= INT_MAX; 461 nd6_llinfo_settimer(ln, INT_MAX); 462 } else { 463 ln->ln_ntick = 0; 464 nd6_llinfo_settimer(ln, ln->ln_ntick); 465 } 466 return; 467 } 468 469 if ((rt = ln->ln_rt) == NULL) 470 panic("ln->ln_rt == NULL"); 471 if ((ifp = rt->rt_ifp) == NULL) 472 panic("ln->ln_rt->rt_ifp == NULL"); 473 ndi = ND_IFINFO(ifp); 474 475 CURVNET_SET(ifp->if_vnet); 476 INIT_VNET_INET6(curvnet); 477 478 /* sanity check */ 479 if (rt->rt_llinfo && (struct llinfo_nd6 *)rt->rt_llinfo != ln) 480 panic("rt_llinfo(%p) is not equal to ln(%p)", 481 rt->rt_llinfo, ln); 482 if (rt_key(rt) == NULL) 483 panic("rt key is NULL in nd6_timer(ln=%p)", ln); 484 485 dst = &((struct sockaddr_in6 *)rt_key(rt))->sin6_addr; 486 487 switch (ln->ln_state) { 488 case ND6_LLINFO_INCOMPLETE: 489 if (ln->ln_asked < V_nd6_mmaxtries) { 490 ln->ln_asked++; 491 nd6_llinfo_settimer(ln, (long)ndi->retrans * hz / 1000); 492 nd6_ns_output(ifp, NULL, dst, ln, 0); 493 } else { 494 struct mbuf *m = ln->ln_hold; 495 if (m) { 496 struct mbuf *m0; 497 498 /* 499 * assuming every packet in ln_hold has the 500 * same IP header 501 */ 502 m0 = m->m_nextpkt; 503 m->m_nextpkt = NULL; 504 icmp6_error2(m, ICMP6_DST_UNREACH, 505 ICMP6_DST_UNREACH_ADDR, 0, rt->rt_ifp); 506 507 ln->ln_hold = m0; 508 clear_llinfo_pqueue(ln); 509 } 510 if (rt && rt->rt_llinfo) 511 (void)nd6_free(rt, 0); 512 ln = NULL; 513 } 514 break; 515 case ND6_LLINFO_REACHABLE: 516 if (!ND6_LLINFO_PERMANENT(ln)) { 517 ln->ln_state = ND6_LLINFO_STALE; 518 nd6_llinfo_settimer(ln, (long)V_nd6_gctimer * hz); 519 } 520 break; 521 522 case ND6_LLINFO_STALE: 523 /* Garbage Collection(RFC 2461 5.3) */ 524 if (!ND6_LLINFO_PERMANENT(ln)) { 525 if (rt && rt->rt_llinfo) 526 (void)nd6_free(rt, 1); 527 ln = NULL; 528 } 529 break; 530 531 case ND6_LLINFO_DELAY: 532 if (ndi && (ndi->flags & ND6_IFF_PERFORMNUD) != 0) { 533 /* We need NUD */ 534 ln->ln_asked = 1; 535 ln->ln_state = ND6_LLINFO_PROBE; 536 nd6_llinfo_settimer(ln, (long)ndi->retrans * hz / 1000); 537 nd6_ns_output(ifp, dst, dst, ln, 0); 538 } else { 539 ln->ln_state = ND6_LLINFO_STALE; /* XXX */ 540 nd6_llinfo_settimer(ln, (long)V_nd6_gctimer * hz); 541 } 542 break; 543 case ND6_LLINFO_PROBE: 544 if (ln->ln_asked < V_nd6_umaxtries) { 545 ln->ln_asked++; 546 nd6_llinfo_settimer(ln, (long)ndi->retrans * hz / 1000); 547 nd6_ns_output(ifp, dst, dst, ln, 0); 548 } else if (rt->rt_ifa != NULL && 549 rt->rt_ifa->ifa_addr->sa_family == AF_INET6 && 550 (((struct in6_ifaddr *)rt->rt_ifa)->ia_flags & IFA_ROUTE)) { 551 /* 552 * This is an unreachable neighbor whose address is 553 * specified as the destination of a p2p interface 554 * (see in6_ifinit()). We should not free the entry 555 * since this is sort of a "static" entry generated 556 * via interface address configuration. 557 */ 558 ln->ln_asked = 0; 559 ln->ln_expire = 0; /* make it permanent */ 560 ln->ln_state = ND6_LLINFO_STALE; 561 } else { 562 if (rt && rt->rt_llinfo) 563 (void)nd6_free(rt, 0); 564 ln = NULL; 565 } 566 break; 567 } 568 CURVNET_RESTORE(); 569 } 570 571 572 /* 573 * ND6 timer routine to expire default route list and prefix list 574 */ 575 void 576 nd6_timer(void *arg) 577 { 578 CURVNET_SET_QUIET((struct vnet *) arg); 579 INIT_VNET_INET6((struct vnet *) arg); 580 int s; 581 struct nd_defrouter *dr; 582 struct nd_prefix *pr; 583 struct in6_ifaddr *ia6, *nia6; 584 struct in6_addrlifetime *lt6; 585 586 callout_reset(&V_nd6_timer_ch, V_nd6_prune * hz, 587 nd6_timer, NULL); 588 589 /* expire default router list */ 590 s = splnet(); 591 dr = TAILQ_FIRST(&V_nd_defrouter); 592 while (dr) { 593 if (dr->expire && dr->expire < time_second) { 594 struct nd_defrouter *t; 595 t = TAILQ_NEXT(dr, dr_entry); 596 defrtrlist_del(dr); 597 dr = t; 598 } else { 599 dr = TAILQ_NEXT(dr, dr_entry); 600 } 601 } 602 603 /* 604 * expire interface addresses. 605 * in the past the loop was inside prefix expiry processing. 606 * However, from a stricter speci-confrmance standpoint, we should 607 * rather separate address lifetimes and prefix lifetimes. 608 */ 609 addrloop: 610 for (ia6 = V_in6_ifaddr; ia6; ia6 = nia6) { 611 nia6 = ia6->ia_next; 612 /* check address lifetime */ 613 lt6 = &ia6->ia6_lifetime; 614 if (IFA6_IS_INVALID(ia6)) { 615 int regen = 0; 616 617 /* 618 * If the expiring address is temporary, try 619 * regenerating a new one. This would be useful when 620 * we suspended a laptop PC, then turned it on after a 621 * period that could invalidate all temporary 622 * addresses. Although we may have to restart the 623 * loop (see below), it must be after purging the 624 * address. Otherwise, we'd see an infinite loop of 625 * regeneration. 626 */ 627 if (V_ip6_use_tempaddr && 628 (ia6->ia6_flags & IN6_IFF_TEMPORARY) != 0) { 629 if (regen_tmpaddr(ia6) == 0) 630 regen = 1; 631 } 632 633 in6_purgeaddr(&ia6->ia_ifa); 634 635 if (regen) 636 goto addrloop; /* XXX: see below */ 637 } else if (IFA6_IS_DEPRECATED(ia6)) { 638 int oldflags = ia6->ia6_flags; 639 640 ia6->ia6_flags |= IN6_IFF_DEPRECATED; 641 642 /* 643 * If a temporary address has just become deprecated, 644 * regenerate a new one if possible. 645 */ 646 if (V_ip6_use_tempaddr && 647 (ia6->ia6_flags & IN6_IFF_TEMPORARY) != 0 && 648 (oldflags & IN6_IFF_DEPRECATED) == 0) { 649 650 if (regen_tmpaddr(ia6) == 0) { 651 /* 652 * A new temporary address is 653 * generated. 654 * XXX: this means the address chain 655 * has changed while we are still in 656 * the loop. Although the change 657 * would not cause disaster (because 658 * it's not a deletion, but an 659 * addition,) we'd rather restart the 660 * loop just for safety. Or does this 661 * significantly reduce performance?? 662 */ 663 goto addrloop; 664 } 665 } 666 } else { 667 /* 668 * A new RA might have made a deprecated address 669 * preferred. 670 */ 671 ia6->ia6_flags &= ~IN6_IFF_DEPRECATED; 672 } 673 } 674 675 /* expire prefix list */ 676 pr = V_nd_prefix.lh_first; 677 while (pr) { 678 /* 679 * check prefix lifetime. 680 * since pltime is just for autoconf, pltime processing for 681 * prefix is not necessary. 682 */ 683 if (pr->ndpr_vltime != ND6_INFINITE_LIFETIME && 684 time_second - pr->ndpr_lastupdate > pr->ndpr_vltime) { 685 struct nd_prefix *t; 686 t = pr->ndpr_next; 687 688 /* 689 * address expiration and prefix expiration are 690 * separate. NEVER perform in6_purgeaddr here. 691 */ 692 693 prelist_remove(pr); 694 pr = t; 695 } else 696 pr = pr->ndpr_next; 697 } 698 splx(s); 699 CURVNET_RESTORE(); 700 } 701 702 /* 703 * ia6 - deprecated/invalidated temporary address 704 */ 705 static int 706 regen_tmpaddr(struct in6_ifaddr *ia6) 707 { 708 struct ifaddr *ifa; 709 struct ifnet *ifp; 710 struct in6_ifaddr *public_ifa6 = NULL; 711 712 ifp = ia6->ia_ifa.ifa_ifp; 713 for (ifa = ifp->if_addrlist.tqh_first; ifa; 714 ifa = ifa->ifa_list.tqe_next) { 715 struct in6_ifaddr *it6; 716 717 if (ifa->ifa_addr->sa_family != AF_INET6) 718 continue; 719 720 it6 = (struct in6_ifaddr *)ifa; 721 722 /* ignore no autoconf addresses. */ 723 if ((it6->ia6_flags & IN6_IFF_AUTOCONF) == 0) 724 continue; 725 726 /* ignore autoconf addresses with different prefixes. */ 727 if (it6->ia6_ndpr == NULL || it6->ia6_ndpr != ia6->ia6_ndpr) 728 continue; 729 730 /* 731 * Now we are looking at an autoconf address with the same 732 * prefix as ours. If the address is temporary and is still 733 * preferred, do not create another one. It would be rare, but 734 * could happen, for example, when we resume a laptop PC after 735 * a long period. 736 */ 737 if ((it6->ia6_flags & IN6_IFF_TEMPORARY) != 0 && 738 !IFA6_IS_DEPRECATED(it6)) { 739 public_ifa6 = NULL; 740 break; 741 } 742 743 /* 744 * This is a public autoconf address that has the same prefix 745 * as ours. If it is preferred, keep it. We can't break the 746 * loop here, because there may be a still-preferred temporary 747 * address with the prefix. 748 */ 749 if (!IFA6_IS_DEPRECATED(it6)) 750 public_ifa6 = it6; 751 } 752 753 if (public_ifa6 != NULL) { 754 int e; 755 756 if ((e = in6_tmpifadd(public_ifa6, 0, 0)) != 0) { 757 log(LOG_NOTICE, "regen_tmpaddr: failed to create a new" 758 " tmp addr,errno=%d\n", e); 759 return (-1); 760 } 761 return (0); 762 } 763 764 return (-1); 765 } 766 767 /* 768 * Nuke neighbor cache/prefix/default router management table, right before 769 * ifp goes away. 770 */ 771 void 772 nd6_purge(struct ifnet *ifp) 773 { 774 INIT_VNET_INET6(ifp->if_vnet); 775 struct llinfo_nd6 *ln, *nln; 776 struct nd_defrouter *dr, *ndr; 777 struct nd_prefix *pr, *npr; 778 779 /* 780 * Nuke default router list entries toward ifp. 781 * We defer removal of default router list entries that is installed 782 * in the routing table, in order to keep additional side effects as 783 * small as possible. 784 */ 785 for (dr = TAILQ_FIRST(&V_nd_defrouter); dr; dr = ndr) { 786 ndr = TAILQ_NEXT(dr, dr_entry); 787 if (dr->installed) 788 continue; 789 790 if (dr->ifp == ifp) 791 defrtrlist_del(dr); 792 } 793 794 for (dr = TAILQ_FIRST(&V_nd_defrouter); dr; dr = ndr) { 795 ndr = TAILQ_NEXT(dr, dr_entry); 796 if (!dr->installed) 797 continue; 798 799 if (dr->ifp == ifp) 800 defrtrlist_del(dr); 801 } 802 803 /* Nuke prefix list entries toward ifp */ 804 for (pr = V_nd_prefix.lh_first; pr; pr = npr) { 805 npr = pr->ndpr_next; 806 if (pr->ndpr_ifp == ifp) { 807 /* 808 * Because if_detach() does *not* release prefixes 809 * while purging addresses the reference count will 810 * still be above zero. We therefore reset it to 811 * make sure that the prefix really gets purged. 812 */ 813 pr->ndpr_refcnt = 0; 814 815 /* 816 * Previously, pr->ndpr_addr is removed as well, 817 * but I strongly believe we don't have to do it. 818 * nd6_purge() is only called from in6_ifdetach(), 819 * which removes all the associated interface addresses 820 * by itself. 821 * (jinmei@kame.net 20010129) 822 */ 823 prelist_remove(pr); 824 } 825 } 826 827 /* cancel default outgoing interface setting */ 828 if (V_nd6_defifindex == ifp->if_index) 829 nd6_setdefaultiface(0); 830 831 if (!V_ip6_forwarding && V_ip6_accept_rtadv) { /* XXX: too restrictive? */ 832 /* refresh default router list */ 833 defrouter_select(); 834 } 835 836 /* 837 * Nuke neighbor cache entries for the ifp. 838 * Note that rt->rt_ifp may not be the same as ifp, 839 * due to KAME goto ours hack. See RTM_RESOLVE case in 840 * nd6_rtrequest(), and ip6_input(). 841 */ 842 ln = V_llinfo_nd6.ln_next; 843 while (ln && ln != &V_llinfo_nd6) { 844 struct rtentry *rt; 845 struct sockaddr_dl *sdl; 846 847 nln = ln->ln_next; 848 rt = ln->ln_rt; 849 if (rt && rt->rt_gateway && 850 rt->rt_gateway->sa_family == AF_LINK) { 851 sdl = (struct sockaddr_dl *)rt->rt_gateway; 852 if (sdl->sdl_index == ifp->if_index) 853 nln = nd6_free(rt, 0); 854 } 855 ln = nln; 856 } 857 } 858 859 struct rtentry * 860 nd6_lookup(struct in6_addr *addr6, int create, struct ifnet *ifp) 861 { 862 INIT_VNET_INET6(curvnet); 863 struct rtentry *rt; 864 struct sockaddr_in6 sin6; 865 char ip6buf[INET6_ADDRSTRLEN]; 866 867 bzero(&sin6, sizeof(sin6)); 868 sin6.sin6_len = sizeof(struct sockaddr_in6); 869 sin6.sin6_family = AF_INET6; 870 sin6.sin6_addr = *addr6; 871 rt = rtalloc1((struct sockaddr *)&sin6, create, 0UL); 872 if (rt) { 873 if ((rt->rt_flags & RTF_LLINFO) == 0 && create) { 874 /* 875 * This is the case for the default route. 876 * If we want to create a neighbor cache for the 877 * address, we should free the route for the 878 * destination and allocate an interface route. 879 */ 880 RTFREE_LOCKED(rt); 881 rt = NULL; 882 } 883 } 884 if (rt == NULL) { 885 if (create && ifp) { 886 int e; 887 888 /* 889 * If no route is available and create is set, 890 * we allocate a host route for the destination 891 * and treat it like an interface route. 892 * This hack is necessary for a neighbor which can't 893 * be covered by our own prefix. 894 */ 895 struct ifaddr *ifa = 896 ifaof_ifpforaddr((struct sockaddr *)&sin6, ifp); 897 if (ifa == NULL) 898 return (NULL); 899 900 /* 901 * Create a new route. RTF_LLINFO is necessary 902 * to create a Neighbor Cache entry for the 903 * destination in nd6_rtrequest which will be 904 * called in rtrequest via ifa->ifa_rtrequest. 905 */ 906 if ((e = rtrequest(RTM_ADD, (struct sockaddr *)&sin6, 907 ifa->ifa_addr, (struct sockaddr *)&all1_sa, 908 (ifa->ifa_flags | RTF_HOST | RTF_LLINFO) & 909 ~RTF_CLONING, &rt)) != 0) { 910 log(LOG_ERR, 911 "nd6_lookup: failed to add route for a " 912 "neighbor(%s), errno=%d\n", 913 ip6_sprintf(ip6buf, addr6), e); 914 } 915 if (rt == NULL) 916 return (NULL); 917 RT_LOCK(rt); 918 if (rt->rt_llinfo) { 919 struct llinfo_nd6 *ln = 920 (struct llinfo_nd6 *)rt->rt_llinfo; 921 ln->ln_state = ND6_LLINFO_NOSTATE; 922 } 923 } else 924 return (NULL); 925 } 926 RT_LOCK_ASSERT(rt); 927 RT_REMREF(rt); 928 /* 929 * Validation for the entry. 930 * Note that the check for rt_llinfo is necessary because a cloned 931 * route from a parent route that has the L flag (e.g. the default 932 * route to a p2p interface) may have the flag, too, while the 933 * destination is not actually a neighbor. 934 * XXX: we can't use rt->rt_ifp to check for the interface, since 935 * it might be the loopback interface if the entry is for our 936 * own address on a non-loopback interface. Instead, we should 937 * use rt->rt_ifa->ifa_ifp, which would specify the REAL 938 * interface. 939 * Note also that ifa_ifp and ifp may differ when we connect two 940 * interfaces to a same link, install a link prefix to an interface, 941 * and try to install a neighbor cache on an interface that does not 942 * have a route to the prefix. 943 */ 944 if ((rt->rt_flags & RTF_GATEWAY) || (rt->rt_flags & RTF_LLINFO) == 0 || 945 rt->rt_gateway->sa_family != AF_LINK || rt->rt_llinfo == NULL || 946 (ifp && rt->rt_ifa->ifa_ifp != ifp)) { 947 if (create) { 948 nd6log((LOG_DEBUG, 949 "nd6_lookup: failed to lookup %s (if = %s)\n", 950 ip6_sprintf(ip6buf, addr6), 951 ifp ? if_name(ifp) : "unspec")); 952 } 953 RT_UNLOCK(rt); 954 return (NULL); 955 } 956 RT_UNLOCK(rt); /* XXX not ready to return rt locked */ 957 return (rt); 958 } 959 960 /* 961 * Test whether a given IPv6 address is a neighbor or not, ignoring 962 * the actual neighbor cache. The neighbor cache is ignored in order 963 * to not reenter the routing code from within itself. 964 */ 965 static int 966 nd6_is_new_addr_neighbor(struct sockaddr_in6 *addr, struct ifnet *ifp) 967 { 968 INIT_VNET_INET6(ifp->if_vnet); 969 struct nd_prefix *pr; 970 struct ifaddr *dstaddr; 971 972 /* 973 * A link-local address is always a neighbor. 974 * XXX: a link does not necessarily specify a single interface. 975 */ 976 if (IN6_IS_ADDR_LINKLOCAL(&addr->sin6_addr)) { 977 struct sockaddr_in6 sin6_copy; 978 u_int32_t zone; 979 980 /* 981 * We need sin6_copy since sa6_recoverscope() may modify the 982 * content (XXX). 983 */ 984 sin6_copy = *addr; 985 if (sa6_recoverscope(&sin6_copy)) 986 return (0); /* XXX: should be impossible */ 987 if (in6_setscope(&sin6_copy.sin6_addr, ifp, &zone)) 988 return (0); 989 if (sin6_copy.sin6_scope_id == zone) 990 return (1); 991 else 992 return (0); 993 } 994 995 /* 996 * If the address matches one of our addresses, 997 * it should be a neighbor. 998 * If the address matches one of our on-link prefixes, it should be a 999 * neighbor. 1000 */ 1001 for (pr = V_nd_prefix.lh_first; pr; pr = pr->ndpr_next) { 1002 if (pr->ndpr_ifp != ifp) 1003 continue; 1004 1005 if (!(pr->ndpr_stateflags & NDPRF_ONLINK)) 1006 continue; 1007 1008 if (IN6_ARE_MASKED_ADDR_EQUAL(&pr->ndpr_prefix.sin6_addr, 1009 &addr->sin6_addr, &pr->ndpr_mask)) 1010 return (1); 1011 } 1012 1013 /* 1014 * If the address is assigned on the node of the other side of 1015 * a p2p interface, the address should be a neighbor. 1016 */ 1017 dstaddr = ifa_ifwithdstaddr((struct sockaddr *)addr); 1018 if ((dstaddr != NULL) && (dstaddr->ifa_ifp == ifp)) 1019 return (1); 1020 1021 /* 1022 * If the default router list is empty, all addresses are regarded 1023 * as on-link, and thus, as a neighbor. 1024 * XXX: we restrict the condition to hosts, because routers usually do 1025 * not have the "default router list". 1026 */ 1027 if (!V_ip6_forwarding && TAILQ_FIRST(&V_nd_defrouter) == NULL && 1028 V_nd6_defifindex == ifp->if_index) { 1029 return (1); 1030 } 1031 1032 return (0); 1033 } 1034 1035 1036 /* 1037 * Detect if a given IPv6 address identifies a neighbor on a given link. 1038 * XXX: should take care of the destination of a p2p link? 1039 */ 1040 int 1041 nd6_is_addr_neighbor(struct sockaddr_in6 *addr, struct ifnet *ifp) 1042 { 1043 1044 if (nd6_is_new_addr_neighbor(addr, ifp)) 1045 return (1); 1046 1047 /* 1048 * Even if the address matches none of our addresses, it might be 1049 * in the neighbor cache. 1050 */ 1051 if (nd6_lookup(&addr->sin6_addr, 0, ifp) != NULL) 1052 return (1); 1053 1054 return (0); 1055 } 1056 1057 /* 1058 * Free an nd6 llinfo entry. 1059 * Since the function would cause significant changes in the kernel, DO NOT 1060 * make it global, unless you have a strong reason for the change, and are sure 1061 * that the change is safe. 1062 */ 1063 static struct llinfo_nd6 * 1064 nd6_free(struct rtentry *rt, int gc) 1065 { 1066 INIT_VNET_INET6(curvnet); 1067 struct llinfo_nd6 *ln = (struct llinfo_nd6 *)rt->rt_llinfo, *next; 1068 struct in6_addr in6 = ((struct sockaddr_in6 *)rt_key(rt))->sin6_addr; 1069 struct nd_defrouter *dr; 1070 1071 /* 1072 * we used to have pfctlinput(PRC_HOSTDEAD) here. 1073 * even though it is not harmful, it was not really necessary. 1074 */ 1075 1076 /* cancel timer */ 1077 nd6_llinfo_settimer(ln, -1); 1078 1079 if (!V_ip6_forwarding) { 1080 int s; 1081 s = splnet(); 1082 dr = defrouter_lookup(&((struct sockaddr_in6 *)rt_key(rt))->sin6_addr, 1083 rt->rt_ifp); 1084 1085 if (dr != NULL && dr->expire && 1086 ln->ln_state == ND6_LLINFO_STALE && gc) { 1087 /* 1088 * If the reason for the deletion is just garbage 1089 * collection, and the neighbor is an active default 1090 * router, do not delete it. Instead, reset the GC 1091 * timer using the router's lifetime. 1092 * Simply deleting the entry would affect default 1093 * router selection, which is not necessarily a good 1094 * thing, especially when we're using router preference 1095 * values. 1096 * XXX: the check for ln_state would be redundant, 1097 * but we intentionally keep it just in case. 1098 */ 1099 if (dr->expire > time_second) 1100 nd6_llinfo_settimer(ln, 1101 (dr->expire - time_second) * hz); 1102 else 1103 nd6_llinfo_settimer(ln, (long)V_nd6_gctimer * hz); 1104 splx(s); 1105 return (ln->ln_next); 1106 } 1107 1108 if (ln->ln_router || dr) { 1109 /* 1110 * rt6_flush must be called whether or not the neighbor 1111 * is in the Default Router List. 1112 * See a corresponding comment in nd6_na_input(). 1113 */ 1114 rt6_flush(&in6, rt->rt_ifp); 1115 } 1116 1117 if (dr) { 1118 /* 1119 * Unreachablity of a router might affect the default 1120 * router selection and on-link detection of advertised 1121 * prefixes. 1122 */ 1123 1124 /* 1125 * Temporarily fake the state to choose a new default 1126 * router and to perform on-link determination of 1127 * prefixes correctly. 1128 * Below the state will be set correctly, 1129 * or the entry itself will be deleted. 1130 */ 1131 ln->ln_state = ND6_LLINFO_INCOMPLETE; 1132 1133 /* 1134 * Since defrouter_select() does not affect the 1135 * on-link determination and MIP6 needs the check 1136 * before the default router selection, we perform 1137 * the check now. 1138 */ 1139 pfxlist_onlink_check(); 1140 1141 /* 1142 * refresh default router list 1143 */ 1144 defrouter_select(); 1145 } 1146 splx(s); 1147 } 1148 1149 /* 1150 * Before deleting the entry, remember the next entry as the 1151 * return value. We need this because pfxlist_onlink_check() above 1152 * might have freed other entries (particularly the old next entry) as 1153 * a side effect (XXX). 1154 */ 1155 next = ln->ln_next; 1156 1157 /* 1158 * Detach the route from the routing tree and the list of neighbor 1159 * caches, and disable the route entry not to be used in already 1160 * cached routes. 1161 */ 1162 rtrequest(RTM_DELETE, rt_key(rt), (struct sockaddr *)0, 1163 rt_mask(rt), 0, (struct rtentry **)0); 1164 1165 return (next); 1166 } 1167 1168 /* 1169 * Upper-layer reachability hint for Neighbor Unreachability Detection. 1170 * 1171 * XXX cost-effective methods? 1172 */ 1173 void 1174 nd6_nud_hint(struct rtentry *rt, struct in6_addr *dst6, int force) 1175 { 1176 INIT_VNET_INET6(curvnet); 1177 struct llinfo_nd6 *ln; 1178 1179 /* 1180 * If the caller specified "rt", use that. Otherwise, resolve the 1181 * routing table by supplied "dst6". 1182 */ 1183 if (rt == NULL) { 1184 if (dst6 == NULL) 1185 return; 1186 if ((rt = nd6_lookup(dst6, 0, NULL)) == NULL) 1187 return; 1188 } 1189 1190 if ((rt->rt_flags & RTF_GATEWAY) != 0 || 1191 (rt->rt_flags & RTF_LLINFO) == 0 || 1192 rt->rt_llinfo == NULL || rt->rt_gateway == NULL || 1193 rt->rt_gateway->sa_family != AF_LINK) { 1194 /* This is not a host route. */ 1195 return; 1196 } 1197 1198 ln = (struct llinfo_nd6 *)rt->rt_llinfo; 1199 if (ln->ln_state < ND6_LLINFO_REACHABLE) 1200 return; 1201 1202 /* 1203 * if we get upper-layer reachability confirmation many times, 1204 * it is possible we have false information. 1205 */ 1206 if (!force) { 1207 ln->ln_byhint++; 1208 if (ln->ln_byhint > V_nd6_maxnudhint) 1209 return; 1210 } 1211 1212 ln->ln_state = ND6_LLINFO_REACHABLE; 1213 if (!ND6_LLINFO_PERMANENT(ln)) { 1214 nd6_llinfo_settimer(ln, 1215 (long)ND_IFINFO(rt->rt_ifp)->reachable * hz); 1216 } 1217 } 1218 1219 /* 1220 * info - XXX unused 1221 */ 1222 void 1223 nd6_rtrequest(int req, struct rtentry *rt, struct rt_addrinfo *info) 1224 { 1225 struct sockaddr *gate = rt->rt_gateway; 1226 struct llinfo_nd6 *ln = (struct llinfo_nd6 *)rt->rt_llinfo; 1227 static struct sockaddr_dl null_sdl = {sizeof(null_sdl), AF_LINK}; 1228 struct ifnet *ifp = rt->rt_ifp; 1229 struct ifaddr *ifa; 1230 INIT_VNET_NET(ifp->if_vnet); 1231 INIT_VNET_INET6(ifp->if_vnet); 1232 1233 RT_LOCK_ASSERT(rt); 1234 1235 if ((rt->rt_flags & RTF_GATEWAY) != 0) 1236 return; 1237 1238 if (nd6_need_cache(ifp) == 0 && (rt->rt_flags & RTF_HOST) == 0) { 1239 /* 1240 * This is probably an interface direct route for a link 1241 * which does not need neighbor caches (e.g. fe80::%lo0/64). 1242 * We do not need special treatment below for such a route. 1243 * Moreover, the RTF_LLINFO flag which would be set below 1244 * would annoy the ndp(8) command. 1245 */ 1246 return; 1247 } 1248 1249 if (req == RTM_RESOLVE && 1250 (nd6_need_cache(ifp) == 0 || /* stf case */ 1251 !nd6_is_new_addr_neighbor((struct sockaddr_in6 *)rt_key(rt), 1252 ifp))) { 1253 /* 1254 * FreeBSD and BSD/OS often make a cloned host route based 1255 * on a less-specific route (e.g. the default route). 1256 * If the less specific route does not have a "gateway" 1257 * (this is the case when the route just goes to a p2p or an 1258 * stf interface), we'll mistakenly make a neighbor cache for 1259 * the host route, and will see strange neighbor solicitation 1260 * for the corresponding destination. In order to avoid the 1261 * confusion, we check if the destination of the route is 1262 * a neighbor in terms of neighbor discovery, and stop the 1263 * process if not. Additionally, we remove the LLINFO flag 1264 * so that ndp(8) will not try to get the neighbor information 1265 * of the destination. 1266 */ 1267 rt->rt_flags &= ~RTF_LLINFO; 1268 return; 1269 } 1270 1271 switch (req) { 1272 case RTM_ADD: 1273 /* 1274 * There is no backward compatibility :) 1275 * 1276 * if ((rt->rt_flags & RTF_HOST) == 0 && 1277 * SIN(rt_mask(rt))->sin_addr.s_addr != 0xffffffff) 1278 * rt->rt_flags |= RTF_CLONING; 1279 */ 1280 if ((rt->rt_flags & RTF_CLONING) || 1281 ((rt->rt_flags & RTF_LLINFO) && ln == NULL)) { 1282 /* 1283 * Case 1: This route should come from a route to 1284 * interface (RTF_CLONING case) or the route should be 1285 * treated as on-link but is currently not 1286 * (RTF_LLINFO && ln == NULL case). 1287 */ 1288 rt_setgate(rt, rt_key(rt), 1289 (struct sockaddr *)&null_sdl); 1290 gate = rt->rt_gateway; 1291 SDL(gate)->sdl_type = ifp->if_type; 1292 SDL(gate)->sdl_index = ifp->if_index; 1293 if (ln) 1294 nd6_llinfo_settimer(ln, 0); 1295 if ((rt->rt_flags & RTF_CLONING) != 0) 1296 break; 1297 } 1298 /* 1299 * In IPv4 code, we try to annonuce new RTF_ANNOUNCE entry here. 1300 * We don't do that here since llinfo is not ready yet. 1301 * 1302 * There are also couple of other things to be discussed: 1303 * - unsolicited NA code needs improvement beforehand 1304 * - RFC2461 says we MAY send multicast unsolicited NA 1305 * (7.2.6 paragraph 4), however, it also says that we 1306 * SHOULD provide a mechanism to prevent multicast NA storm. 1307 * we don't have anything like it right now. 1308 * note that the mechanism needs a mutual agreement 1309 * between proxies, which means that we need to implement 1310 * a new protocol, or a new kludge. 1311 * - from RFC2461 6.2.4, host MUST NOT send an unsolicited NA. 1312 * we need to check ip6forwarding before sending it. 1313 * (or should we allow proxy ND configuration only for 1314 * routers? there's no mention about proxy ND from hosts) 1315 */ 1316 /* FALLTHROUGH */ 1317 case RTM_RESOLVE: 1318 if ((ifp->if_flags & (IFF_POINTOPOINT | IFF_LOOPBACK)) == 0) { 1319 /* 1320 * Address resolution isn't necessary for a point to 1321 * point link, so we can skip this test for a p2p link. 1322 */ 1323 if (gate->sa_family != AF_LINK || 1324 gate->sa_len < sizeof(null_sdl)) { 1325 log(LOG_DEBUG, 1326 "nd6_rtrequest: bad gateway value: %s\n", 1327 if_name(ifp)); 1328 break; 1329 } 1330 SDL(gate)->sdl_type = ifp->if_type; 1331 SDL(gate)->sdl_index = ifp->if_index; 1332 } 1333 if (ln != NULL) 1334 break; /* This happens on a route change */ 1335 /* 1336 * Case 2: This route may come from cloning, or a manual route 1337 * add with a LL address. 1338 */ 1339 R_Malloc(ln, struct llinfo_nd6 *, sizeof(*ln)); 1340 rt->rt_llinfo = (caddr_t)ln; 1341 if (ln == NULL) { 1342 log(LOG_DEBUG, "nd6_rtrequest: malloc failed\n"); 1343 break; 1344 } 1345 V_nd6_inuse++; 1346 V_nd6_allocated++; 1347 bzero(ln, sizeof(*ln)); 1348 RT_ADDREF(rt); 1349 ln->ln_rt = rt; 1350 callout_init(&ln->ln_timer_ch, 0); 1351 1352 /* this is required for "ndp" command. - shin */ 1353 if (req == RTM_ADD) { 1354 /* 1355 * gate should have some valid AF_LINK entry, 1356 * and ln->ln_expire should have some lifetime 1357 * which is specified by ndp command. 1358 */ 1359 ln->ln_state = ND6_LLINFO_REACHABLE; 1360 ln->ln_byhint = 0; 1361 } else { 1362 /* 1363 * When req == RTM_RESOLVE, rt is created and 1364 * initialized in rtrequest(), so rt_expire is 0. 1365 */ 1366 ln->ln_state = ND6_LLINFO_NOSTATE; 1367 nd6_llinfo_settimer(ln, 0); 1368 } 1369 rt->rt_flags |= RTF_LLINFO; 1370 ln->ln_next = V_llinfo_nd6.ln_next; 1371 V_llinfo_nd6.ln_next = ln; 1372 ln->ln_prev = &V_llinfo_nd6; 1373 ln->ln_next->ln_prev = ln; 1374 1375 /* 1376 * check if rt_key(rt) is one of my address assigned 1377 * to the interface. 1378 */ 1379 ifa = (struct ifaddr *)in6ifa_ifpwithaddr(rt->rt_ifp, 1380 &SIN6(rt_key(rt))->sin6_addr); 1381 if (ifa) { 1382 caddr_t macp = nd6_ifptomac(ifp); 1383 nd6_llinfo_settimer(ln, -1); 1384 ln->ln_state = ND6_LLINFO_REACHABLE; 1385 ln->ln_byhint = 0; 1386 if (macp) { 1387 bcopy(macp, LLADDR(SDL(gate)), ifp->if_addrlen); 1388 SDL(gate)->sdl_alen = ifp->if_addrlen; 1389 } 1390 if (V_nd6_useloopback) { 1391 rt->rt_ifp = &V_loif[0]; /* XXX */ 1392 /* 1393 * Make sure rt_ifa be equal to the ifaddr 1394 * corresponding to the address. 1395 * We need this because when we refer 1396 * rt_ifa->ia6_flags in ip6_input, we assume 1397 * that the rt_ifa points to the address instead 1398 * of the loopback address. 1399 */ 1400 if (ifa != rt->rt_ifa) { 1401 IFAFREE(rt->rt_ifa); 1402 IFAREF(ifa); 1403 rt->rt_ifa = ifa; 1404 } 1405 } 1406 } else if (rt->rt_flags & RTF_ANNOUNCE) { 1407 nd6_llinfo_settimer(ln, -1); 1408 ln->ln_state = ND6_LLINFO_REACHABLE; 1409 ln->ln_byhint = 0; 1410 1411 /* join solicited node multicast for proxy ND */ 1412 if (ifp->if_flags & IFF_MULTICAST) { 1413 struct in6_addr llsol; 1414 int error; 1415 1416 llsol = SIN6(rt_key(rt))->sin6_addr; 1417 llsol.s6_addr32[0] = IPV6_ADDR_INT32_MLL; 1418 llsol.s6_addr32[1] = 0; 1419 llsol.s6_addr32[2] = htonl(1); 1420 llsol.s6_addr8[12] = 0xff; 1421 if (in6_setscope(&llsol, ifp, NULL)) 1422 break; 1423 if (in6_addmulti(&llsol, ifp, 1424 &error, 0) == NULL) { 1425 char ip6buf[INET6_ADDRSTRLEN]; 1426 nd6log((LOG_ERR, "%s: failed to join " 1427 "%s (errno=%d)\n", if_name(ifp), 1428 ip6_sprintf(ip6buf, &llsol), 1429 error)); 1430 } 1431 } 1432 } 1433 break; 1434 1435 case RTM_DELETE: 1436 if (ln == NULL) 1437 break; 1438 /* leave from solicited node multicast for proxy ND */ 1439 if ((rt->rt_flags & RTF_ANNOUNCE) != 0 && 1440 (ifp->if_flags & IFF_MULTICAST) != 0) { 1441 struct in6_addr llsol; 1442 struct in6_multi *in6m; 1443 1444 llsol = SIN6(rt_key(rt))->sin6_addr; 1445 llsol.s6_addr32[0] = IPV6_ADDR_INT32_MLL; 1446 llsol.s6_addr32[1] = 0; 1447 llsol.s6_addr32[2] = htonl(1); 1448 llsol.s6_addr8[12] = 0xff; 1449 if (in6_setscope(&llsol, ifp, NULL) == 0) { 1450 IN6_LOOKUP_MULTI(llsol, ifp, in6m); 1451 if (in6m) 1452 in6_delmulti(in6m); 1453 } else 1454 ; /* XXX: should not happen. bark here? */ 1455 } 1456 V_nd6_inuse--; 1457 ln->ln_next->ln_prev = ln->ln_prev; 1458 ln->ln_prev->ln_next = ln->ln_next; 1459 ln->ln_prev = NULL; 1460 nd6_llinfo_settimer(ln, -1); 1461 RT_REMREF(rt); 1462 rt->rt_llinfo = 0; 1463 rt->rt_flags &= ~RTF_LLINFO; 1464 clear_llinfo_pqueue(ln); 1465 Free((caddr_t)ln); 1466 } 1467 } 1468 1469 int 1470 nd6_ioctl(u_long cmd, caddr_t data, struct ifnet *ifp) 1471 { 1472 INIT_VNET_INET6(ifp->if_vnet); 1473 struct in6_drlist *drl = (struct in6_drlist *)data; 1474 struct in6_oprlist *oprl = (struct in6_oprlist *)data; 1475 struct in6_ndireq *ndi = (struct in6_ndireq *)data; 1476 struct in6_nbrinfo *nbi = (struct in6_nbrinfo *)data; 1477 struct in6_ndifreq *ndif = (struct in6_ndifreq *)data; 1478 struct nd_defrouter *dr; 1479 struct nd_prefix *pr; 1480 struct rtentry *rt; 1481 int i = 0, error = 0; 1482 int s; 1483 1484 switch (cmd) { 1485 case SIOCGDRLST_IN6: 1486 /* 1487 * obsolete API, use sysctl under net.inet6.icmp6 1488 */ 1489 bzero(drl, sizeof(*drl)); 1490 s = splnet(); 1491 dr = TAILQ_FIRST(&V_nd_defrouter); 1492 while (dr && i < DRLSTSIZ) { 1493 drl->defrouter[i].rtaddr = dr->rtaddr; 1494 in6_clearscope(&drl->defrouter[i].rtaddr); 1495 1496 drl->defrouter[i].flags = dr->flags; 1497 drl->defrouter[i].rtlifetime = dr->rtlifetime; 1498 drl->defrouter[i].expire = dr->expire; 1499 drl->defrouter[i].if_index = dr->ifp->if_index; 1500 i++; 1501 dr = TAILQ_NEXT(dr, dr_entry); 1502 } 1503 splx(s); 1504 break; 1505 case SIOCGPRLST_IN6: 1506 /* 1507 * obsolete API, use sysctl under net.inet6.icmp6 1508 * 1509 * XXX the structure in6_prlist was changed in backward- 1510 * incompatible manner. in6_oprlist is used for SIOCGPRLST_IN6, 1511 * in6_prlist is used for nd6_sysctl() - fill_prlist(). 1512 */ 1513 /* 1514 * XXX meaning of fields, especialy "raflags", is very 1515 * differnet between RA prefix list and RR/static prefix list. 1516 * how about separating ioctls into two? 1517 */ 1518 bzero(oprl, sizeof(*oprl)); 1519 s = splnet(); 1520 pr = V_nd_prefix.lh_first; 1521 while (pr && i < PRLSTSIZ) { 1522 struct nd_pfxrouter *pfr; 1523 int j; 1524 1525 oprl->prefix[i].prefix = pr->ndpr_prefix.sin6_addr; 1526 oprl->prefix[i].raflags = pr->ndpr_raf; 1527 oprl->prefix[i].prefixlen = pr->ndpr_plen; 1528 oprl->prefix[i].vltime = pr->ndpr_vltime; 1529 oprl->prefix[i].pltime = pr->ndpr_pltime; 1530 oprl->prefix[i].if_index = pr->ndpr_ifp->if_index; 1531 if (pr->ndpr_vltime == ND6_INFINITE_LIFETIME) 1532 oprl->prefix[i].expire = 0; 1533 else { 1534 time_t maxexpire; 1535 1536 /* XXX: we assume time_t is signed. */ 1537 maxexpire = (-1) & 1538 ~((time_t)1 << 1539 ((sizeof(maxexpire) * 8) - 1)); 1540 if (pr->ndpr_vltime < 1541 maxexpire - pr->ndpr_lastupdate) { 1542 oprl->prefix[i].expire = 1543 pr->ndpr_lastupdate + 1544 pr->ndpr_vltime; 1545 } else 1546 oprl->prefix[i].expire = maxexpire; 1547 } 1548 1549 pfr = pr->ndpr_advrtrs.lh_first; 1550 j = 0; 1551 while (pfr) { 1552 if (j < DRLSTSIZ) { 1553 #define RTRADDR oprl->prefix[i].advrtr[j] 1554 RTRADDR = pfr->router->rtaddr; 1555 in6_clearscope(&RTRADDR); 1556 #undef RTRADDR 1557 } 1558 j++; 1559 pfr = pfr->pfr_next; 1560 } 1561 oprl->prefix[i].advrtrs = j; 1562 oprl->prefix[i].origin = PR_ORIG_RA; 1563 1564 i++; 1565 pr = pr->ndpr_next; 1566 } 1567 splx(s); 1568 1569 break; 1570 case OSIOCGIFINFO_IN6: 1571 #define ND ndi->ndi 1572 /* XXX: old ndp(8) assumes a positive value for linkmtu. */ 1573 bzero(&ND, sizeof(ND)); 1574 ND.linkmtu = IN6_LINKMTU(ifp); 1575 ND.maxmtu = ND_IFINFO(ifp)->maxmtu; 1576 ND.basereachable = ND_IFINFO(ifp)->basereachable; 1577 ND.reachable = ND_IFINFO(ifp)->reachable; 1578 ND.retrans = ND_IFINFO(ifp)->retrans; 1579 ND.flags = ND_IFINFO(ifp)->flags; 1580 ND.recalctm = ND_IFINFO(ifp)->recalctm; 1581 ND.chlim = ND_IFINFO(ifp)->chlim; 1582 break; 1583 case SIOCGIFINFO_IN6: 1584 ND = *ND_IFINFO(ifp); 1585 break; 1586 case SIOCSIFINFO_IN6: 1587 /* 1588 * used to change host variables from userland. 1589 * intented for a use on router to reflect RA configurations. 1590 */ 1591 /* 0 means 'unspecified' */ 1592 if (ND.linkmtu != 0) { 1593 if (ND.linkmtu < IPV6_MMTU || 1594 ND.linkmtu > IN6_LINKMTU(ifp)) { 1595 error = EINVAL; 1596 break; 1597 } 1598 ND_IFINFO(ifp)->linkmtu = ND.linkmtu; 1599 } 1600 1601 if (ND.basereachable != 0) { 1602 int obasereachable = ND_IFINFO(ifp)->basereachable; 1603 1604 ND_IFINFO(ifp)->basereachable = ND.basereachable; 1605 if (ND.basereachable != obasereachable) 1606 ND_IFINFO(ifp)->reachable = 1607 ND_COMPUTE_RTIME(ND.basereachable); 1608 } 1609 if (ND.retrans != 0) 1610 ND_IFINFO(ifp)->retrans = ND.retrans; 1611 if (ND.chlim != 0) 1612 ND_IFINFO(ifp)->chlim = ND.chlim; 1613 /* FALLTHROUGH */ 1614 case SIOCSIFINFO_FLAGS: 1615 ND_IFINFO(ifp)->flags = ND.flags; 1616 break; 1617 #undef ND 1618 case SIOCSNDFLUSH_IN6: /* XXX: the ioctl name is confusing... */ 1619 /* sync kernel routing table with the default router list */ 1620 defrouter_reset(); 1621 defrouter_select(); 1622 break; 1623 case SIOCSPFXFLUSH_IN6: 1624 { 1625 /* flush all the prefix advertised by routers */ 1626 struct nd_prefix *pr, *next; 1627 1628 s = splnet(); 1629 for (pr = V_nd_prefix.lh_first; pr; pr = next) { 1630 struct in6_ifaddr *ia, *ia_next; 1631 1632 next = pr->ndpr_next; 1633 1634 if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr)) 1635 continue; /* XXX */ 1636 1637 /* do we really have to remove addresses as well? */ 1638 for (ia = V_in6_ifaddr; ia; ia = ia_next) { 1639 /* ia might be removed. keep the next ptr. */ 1640 ia_next = ia->ia_next; 1641 1642 if ((ia->ia6_flags & IN6_IFF_AUTOCONF) == 0) 1643 continue; 1644 1645 if (ia->ia6_ndpr == pr) 1646 in6_purgeaddr(&ia->ia_ifa); 1647 } 1648 prelist_remove(pr); 1649 } 1650 splx(s); 1651 break; 1652 } 1653 case SIOCSRTRFLUSH_IN6: 1654 { 1655 /* flush all the default routers */ 1656 struct nd_defrouter *dr, *next; 1657 1658 s = splnet(); 1659 defrouter_reset(); 1660 for (dr = TAILQ_FIRST(&V_nd_defrouter); dr; dr = next) { 1661 next = TAILQ_NEXT(dr, dr_entry); 1662 defrtrlist_del(dr); 1663 } 1664 defrouter_select(); 1665 splx(s); 1666 break; 1667 } 1668 case SIOCGNBRINFO_IN6: 1669 { 1670 struct llinfo_nd6 *ln; 1671 struct in6_addr nb_addr = nbi->addr; /* make local for safety */ 1672 1673 if ((error = in6_setscope(&nb_addr, ifp, NULL)) != 0) 1674 return (error); 1675 1676 s = splnet(); 1677 if ((rt = nd6_lookup(&nb_addr, 0, ifp)) == NULL) { 1678 error = EINVAL; 1679 splx(s); 1680 break; 1681 } 1682 ln = (struct llinfo_nd6 *)rt->rt_llinfo; 1683 nbi->state = ln->ln_state; 1684 nbi->asked = ln->ln_asked; 1685 nbi->isrouter = ln->ln_router; 1686 nbi->expire = ln->ln_expire; 1687 splx(s); 1688 1689 break; 1690 } 1691 case SIOCGDEFIFACE_IN6: /* XXX: should be implemented as a sysctl? */ 1692 ndif->ifindex = V_nd6_defifindex; 1693 break; 1694 case SIOCSDEFIFACE_IN6: /* XXX: should be implemented as a sysctl? */ 1695 return (nd6_setdefaultiface(ndif->ifindex)); 1696 } 1697 return (error); 1698 } 1699 1700 /* 1701 * Create neighbor cache entry and cache link-layer address, 1702 * on reception of inbound ND6 packets. (RS/RA/NS/redirect) 1703 * 1704 * type - ICMP6 type 1705 * code - type dependent information 1706 */ 1707 struct rtentry * 1708 nd6_cache_lladdr(struct ifnet *ifp, struct in6_addr *from, char *lladdr, 1709 int lladdrlen, int type, int code) 1710 { 1711 INIT_VNET_INET6(curvnet); 1712 struct rtentry *rt = NULL; 1713 struct llinfo_nd6 *ln = NULL; 1714 int is_newentry; 1715 struct sockaddr_dl *sdl = NULL; 1716 int do_update; 1717 int olladdr; 1718 int llchange; 1719 int newstate = 0; 1720 1721 if (ifp == NULL) 1722 panic("ifp == NULL in nd6_cache_lladdr"); 1723 if (from == NULL) 1724 panic("from == NULL in nd6_cache_lladdr"); 1725 1726 /* nothing must be updated for unspecified address */ 1727 if (IN6_IS_ADDR_UNSPECIFIED(from)) 1728 return NULL; 1729 1730 /* 1731 * Validation about ifp->if_addrlen and lladdrlen must be done in 1732 * the caller. 1733 * 1734 * XXX If the link does not have link-layer adderss, what should 1735 * we do? (ifp->if_addrlen == 0) 1736 * Spec says nothing in sections for RA, RS and NA. There's small 1737 * description on it in NS section (RFC 2461 7.2.3). 1738 */ 1739 1740 rt = nd6_lookup(from, 0, ifp); 1741 if (rt == NULL) { 1742 rt = nd6_lookup(from, 1, ifp); 1743 is_newentry = 1; 1744 } else { 1745 /* do nothing if static ndp is set */ 1746 if (rt->rt_flags & RTF_STATIC) 1747 return NULL; 1748 is_newentry = 0; 1749 } 1750 1751 if (rt == NULL) 1752 return NULL; 1753 if ((rt->rt_flags & (RTF_GATEWAY | RTF_LLINFO)) != RTF_LLINFO) { 1754 fail: 1755 (void)nd6_free(rt, 0); 1756 return NULL; 1757 } 1758 ln = (struct llinfo_nd6 *)rt->rt_llinfo; 1759 if (ln == NULL) 1760 goto fail; 1761 if (rt->rt_gateway == NULL) 1762 goto fail; 1763 if (rt->rt_gateway->sa_family != AF_LINK) 1764 goto fail; 1765 sdl = SDL(rt->rt_gateway); 1766 1767 olladdr = (sdl->sdl_alen) ? 1 : 0; 1768 if (olladdr && lladdr) { 1769 if (bcmp(lladdr, LLADDR(sdl), ifp->if_addrlen)) 1770 llchange = 1; 1771 else 1772 llchange = 0; 1773 } else 1774 llchange = 0; 1775 1776 /* 1777 * newentry olladdr lladdr llchange (*=record) 1778 * 0 n n -- (1) 1779 * 0 y n -- (2) 1780 * 0 n y -- (3) * STALE 1781 * 0 y y n (4) * 1782 * 0 y y y (5) * STALE 1783 * 1 -- n -- (6) NOSTATE(= PASSIVE) 1784 * 1 -- y -- (7) * STALE 1785 */ 1786 1787 if (lladdr) { /* (3-5) and (7) */ 1788 /* 1789 * Record source link-layer address 1790 * XXX is it dependent to ifp->if_type? 1791 */ 1792 sdl->sdl_alen = ifp->if_addrlen; 1793 bcopy(lladdr, LLADDR(sdl), ifp->if_addrlen); 1794 } 1795 1796 if (!is_newentry) { 1797 if ((!olladdr && lladdr != NULL) || /* (3) */ 1798 (olladdr && lladdr != NULL && llchange)) { /* (5) */ 1799 do_update = 1; 1800 newstate = ND6_LLINFO_STALE; 1801 } else /* (1-2,4) */ 1802 do_update = 0; 1803 } else { 1804 do_update = 1; 1805 if (lladdr == NULL) /* (6) */ 1806 newstate = ND6_LLINFO_NOSTATE; 1807 else /* (7) */ 1808 newstate = ND6_LLINFO_STALE; 1809 } 1810 1811 if (do_update) { 1812 /* 1813 * Update the state of the neighbor cache. 1814 */ 1815 ln->ln_state = newstate; 1816 1817 if (ln->ln_state == ND6_LLINFO_STALE) { 1818 /* 1819 * XXX: since nd6_output() below will cause 1820 * state tansition to DELAY and reset the timer, 1821 * we must set the timer now, although it is actually 1822 * meaningless. 1823 */ 1824 nd6_llinfo_settimer(ln, (long)V_nd6_gctimer * hz); 1825 1826 if (ln->ln_hold) { 1827 struct mbuf *m_hold, *m_hold_next; 1828 1829 /* 1830 * reset the ln_hold in advance, to explicitly 1831 * prevent a ln_hold lookup in nd6_output() 1832 * (wouldn't happen, though...) 1833 */ 1834 for (m_hold = ln->ln_hold, ln->ln_hold = NULL; 1835 m_hold; m_hold = m_hold_next) { 1836 m_hold_next = m_hold->m_nextpkt; 1837 m_hold->m_nextpkt = NULL; 1838 1839 /* 1840 * we assume ifp is not a p2p here, so 1841 * just set the 2nd argument as the 1842 * 1st one. 1843 */ 1844 nd6_output(ifp, ifp, m_hold, 1845 (struct sockaddr_in6 *)rt_key(rt), 1846 rt); 1847 } 1848 } 1849 } else if (ln->ln_state == ND6_LLINFO_INCOMPLETE) { 1850 /* probe right away */ 1851 nd6_llinfo_settimer((void *)ln, 0); 1852 } 1853 } 1854 1855 /* 1856 * ICMP6 type dependent behavior. 1857 * 1858 * NS: clear IsRouter if new entry 1859 * RS: clear IsRouter 1860 * RA: set IsRouter if there's lladdr 1861 * redir: clear IsRouter if new entry 1862 * 1863 * RA case, (1): 1864 * The spec says that we must set IsRouter in the following cases: 1865 * - If lladdr exist, set IsRouter. This means (1-5). 1866 * - If it is old entry (!newentry), set IsRouter. This means (7). 1867 * So, based on the spec, in (1-5) and (7) cases we must set IsRouter. 1868 * A quetion arises for (1) case. (1) case has no lladdr in the 1869 * neighbor cache, this is similar to (6). 1870 * This case is rare but we figured that we MUST NOT set IsRouter. 1871 * 1872 * newentry olladdr lladdr llchange NS RS RA redir 1873 * D R 1874 * 0 n n -- (1) c ? s 1875 * 0 y n -- (2) c s s 1876 * 0 n y -- (3) c s s 1877 * 0 y y n (4) c s s 1878 * 0 y y y (5) c s s 1879 * 1 -- n -- (6) c c c s 1880 * 1 -- y -- (7) c c s c s 1881 * 1882 * (c=clear s=set) 1883 */ 1884 switch (type & 0xff) { 1885 case ND_NEIGHBOR_SOLICIT: 1886 /* 1887 * New entry must have is_router flag cleared. 1888 */ 1889 if (is_newentry) /* (6-7) */ 1890 ln->ln_router = 0; 1891 break; 1892 case ND_REDIRECT: 1893 /* 1894 * If the icmp is a redirect to a better router, always set the 1895 * is_router flag. Otherwise, if the entry is newly created, 1896 * clear the flag. [RFC 2461, sec 8.3] 1897 */ 1898 if (code == ND_REDIRECT_ROUTER) 1899 ln->ln_router = 1; 1900 else if (is_newentry) /* (6-7) */ 1901 ln->ln_router = 0; 1902 break; 1903 case ND_ROUTER_SOLICIT: 1904 /* 1905 * is_router flag must always be cleared. 1906 */ 1907 ln->ln_router = 0; 1908 break; 1909 case ND_ROUTER_ADVERT: 1910 /* 1911 * Mark an entry with lladdr as a router. 1912 */ 1913 if ((!is_newentry && (olladdr || lladdr)) || /* (2-5) */ 1914 (is_newentry && lladdr)) { /* (7) */ 1915 ln->ln_router = 1; 1916 } 1917 break; 1918 } 1919 1920 /* 1921 * When the link-layer address of a router changes, select the 1922 * best router again. In particular, when the neighbor entry is newly 1923 * created, it might affect the selection policy. 1924 * Question: can we restrict the first condition to the "is_newentry" 1925 * case? 1926 * XXX: when we hear an RA from a new router with the link-layer 1927 * address option, defrouter_select() is called twice, since 1928 * defrtrlist_update called the function as well. However, I believe 1929 * we can compromise the overhead, since it only happens the first 1930 * time. 1931 * XXX: although defrouter_select() should not have a bad effect 1932 * for those are not autoconfigured hosts, we explicitly avoid such 1933 * cases for safety. 1934 */ 1935 if (do_update && ln->ln_router && !V_ip6_forwarding && V_ip6_accept_rtadv) 1936 defrouter_select(); 1937 1938 return rt; 1939 } 1940 1941 static void 1942 nd6_slowtimo(void *arg) 1943 { 1944 CURVNET_SET((struct vnet *) arg); 1945 INIT_VNET_NET((struct vnet *) arg); 1946 INIT_VNET_INET6((struct vnet *) arg); 1947 struct nd_ifinfo *nd6if; 1948 struct ifnet *ifp; 1949 1950 callout_reset(&V_nd6_slowtimo_ch, ND6_SLOWTIMER_INTERVAL * hz, 1951 nd6_slowtimo, NULL); 1952 IFNET_RLOCK(); 1953 for (ifp = TAILQ_FIRST(&V_ifnet); ifp; 1954 ifp = TAILQ_NEXT(ifp, if_list)) { 1955 nd6if = ND_IFINFO(ifp); 1956 if (nd6if->basereachable && /* already initialized */ 1957 (nd6if->recalctm -= ND6_SLOWTIMER_INTERVAL) <= 0) { 1958 /* 1959 * Since reachable time rarely changes by router 1960 * advertisements, we SHOULD insure that a new random 1961 * value gets recomputed at least once every few hours. 1962 * (RFC 2461, 6.3.4) 1963 */ 1964 nd6if->recalctm = V_nd6_recalc_reachtm_interval; 1965 nd6if->reachable = ND_COMPUTE_RTIME(nd6if->basereachable); 1966 } 1967 } 1968 IFNET_RUNLOCK(); 1969 CURVNET_RESTORE(); 1970 } 1971 1972 #define senderr(e) { error = (e); goto bad;} 1973 int 1974 nd6_output(struct ifnet *ifp, struct ifnet *origifp, struct mbuf *m0, 1975 struct sockaddr_in6 *dst, struct rtentry *rt0) 1976 { 1977 INIT_VNET_INET6(curvnet); 1978 struct mbuf *m = m0; 1979 struct rtentry *rt = rt0; 1980 struct sockaddr_in6 *gw6 = NULL; 1981 struct llinfo_nd6 *ln = NULL; 1982 int error = 0; 1983 1984 if (IN6_IS_ADDR_MULTICAST(&dst->sin6_addr)) 1985 goto sendpkt; 1986 1987 if (nd6_need_cache(ifp) == 0) 1988 goto sendpkt; 1989 1990 /* 1991 * next hop determination. This routine is derived from ether_output. 1992 */ 1993 /* NB: the locking here is tortuous... */ 1994 if (rt != NULL) 1995 RT_LOCK(rt); 1996 again: 1997 if (rt != NULL) { 1998 if ((rt->rt_flags & RTF_UP) == 0) { 1999 RT_UNLOCK(rt); 2000 rt0 = rt = rtalloc1((struct sockaddr *)dst, 1, 0UL); 2001 if (rt != NULL) { 2002 RT_REMREF(rt); 2003 if (rt->rt_ifp != ifp) 2004 /* 2005 * XXX maybe we should update ifp too, 2006 * but the original code didn't and I 2007 * don't know what is correct here. 2008 */ 2009 goto again; 2010 } else 2011 senderr(EHOSTUNREACH); 2012 } 2013 2014 if (rt->rt_flags & RTF_GATEWAY) { 2015 gw6 = (struct sockaddr_in6 *)rt->rt_gateway; 2016 2017 /* 2018 * We skip link-layer address resolution and NUD 2019 * if the gateway is not a neighbor from ND point 2020 * of view, regardless of the value of nd_ifinfo.flags. 2021 * The second condition is a bit tricky; we skip 2022 * if the gateway is our own address, which is 2023 * sometimes used to install a route to a p2p link. 2024 */ 2025 if (!nd6_is_addr_neighbor(gw6, ifp) || 2026 in6ifa_ifpwithaddr(ifp, &gw6->sin6_addr)) { 2027 RT_UNLOCK(rt); 2028 /* 2029 * We allow this kind of tricky route only 2030 * when the outgoing interface is p2p. 2031 * XXX: we may need a more generic rule here. 2032 */ 2033 if ((ifp->if_flags & IFF_POINTOPOINT) == 0) 2034 senderr(EHOSTUNREACH); 2035 2036 goto sendpkt; 2037 } 2038 2039 if (rt->rt_gwroute == NULL) 2040 goto lookup; 2041 rt = rt->rt_gwroute; 2042 RT_LOCK(rt); /* NB: gwroute */ 2043 if ((rt->rt_flags & RTF_UP) == 0) { 2044 RTFREE_LOCKED(rt); /* unlock gwroute */ 2045 rt = rt0; 2046 rt0->rt_gwroute = NULL; 2047 lookup: 2048 RT_UNLOCK(rt0); 2049 rt = rtalloc1(rt->rt_gateway, 1, 0UL); 2050 if (rt == rt0) { 2051 RT_REMREF(rt0); 2052 RT_UNLOCK(rt0); 2053 senderr(EHOSTUNREACH); 2054 } 2055 RT_LOCK(rt0); 2056 if (rt0->rt_gwroute != NULL) 2057 RTFREE(rt0->rt_gwroute); 2058 rt0->rt_gwroute = rt; 2059 if (rt == NULL) { 2060 RT_UNLOCK(rt0); 2061 senderr(EHOSTUNREACH); 2062 } 2063 } 2064 RT_UNLOCK(rt0); 2065 } 2066 RT_UNLOCK(rt); 2067 } 2068 2069 /* 2070 * Address resolution or Neighbor Unreachability Detection 2071 * for the next hop. 2072 * At this point, the destination of the packet must be a unicast 2073 * or an anycast address(i.e. not a multicast). 2074 */ 2075 2076 /* Look up the neighbor cache for the nexthop */ 2077 if (rt && (rt->rt_flags & RTF_LLINFO) != 0) 2078 ln = (struct llinfo_nd6 *)rt->rt_llinfo; 2079 else { 2080 /* 2081 * Since nd6_is_addr_neighbor() internally calls nd6_lookup(), 2082 * the condition below is not very efficient. But we believe 2083 * it is tolerable, because this should be a rare case. 2084 */ 2085 if (nd6_is_addr_neighbor(dst, ifp) && 2086 (rt = nd6_lookup(&dst->sin6_addr, 1, ifp)) != NULL) 2087 ln = (struct llinfo_nd6 *)rt->rt_llinfo; 2088 } 2089 if (ln == NULL || rt == NULL) { 2090 if ((ifp->if_flags & IFF_POINTOPOINT) == 0 && 2091 !(ND_IFINFO(ifp)->flags & ND6_IFF_PERFORMNUD)) { 2092 char ip6buf[INET6_ADDRSTRLEN]; 2093 log(LOG_DEBUG, 2094 "nd6_output: can't allocate llinfo for %s " 2095 "(ln=%p, rt=%p)\n", 2096 ip6_sprintf(ip6buf, &dst->sin6_addr), ln, rt); 2097 senderr(EIO); /* XXX: good error? */ 2098 } 2099 2100 goto sendpkt; /* send anyway */ 2101 } 2102 2103 /* We don't have to do link-layer address resolution on a p2p link. */ 2104 if ((ifp->if_flags & IFF_POINTOPOINT) != 0 && 2105 ln->ln_state < ND6_LLINFO_REACHABLE) { 2106 ln->ln_state = ND6_LLINFO_STALE; 2107 nd6_llinfo_settimer(ln, (long)V_nd6_gctimer * hz); 2108 } 2109 2110 /* 2111 * The first time we send a packet to a neighbor whose entry is 2112 * STALE, we have to change the state to DELAY and a sets a timer to 2113 * expire in DELAY_FIRST_PROBE_TIME seconds to ensure do 2114 * neighbor unreachability detection on expiration. 2115 * (RFC 2461 7.3.3) 2116 */ 2117 if (ln->ln_state == ND6_LLINFO_STALE) { 2118 ln->ln_asked = 0; 2119 ln->ln_state = ND6_LLINFO_DELAY; 2120 nd6_llinfo_settimer(ln, (long)V_nd6_delay * hz); 2121 } 2122 2123 /* 2124 * If the neighbor cache entry has a state other than INCOMPLETE 2125 * (i.e. its link-layer address is already resolved), just 2126 * send the packet. 2127 */ 2128 if (ln->ln_state > ND6_LLINFO_INCOMPLETE) 2129 goto sendpkt; 2130 2131 /* 2132 * There is a neighbor cache entry, but no ethernet address 2133 * response yet. Append this latest packet to the end of the 2134 * packet queue in the mbuf, unless the number of the packet 2135 * does not exceed nd6_maxqueuelen. When it exceeds nd6_maxqueuelen, 2136 * the oldest packet in the queue will be removed. 2137 */ 2138 if (ln->ln_state == ND6_LLINFO_NOSTATE) 2139 ln->ln_state = ND6_LLINFO_INCOMPLETE; 2140 if (ln->ln_hold) { 2141 struct mbuf *m_hold; 2142 int i; 2143 2144 i = 0; 2145 for (m_hold = ln->ln_hold; m_hold; m_hold = m_hold->m_nextpkt) { 2146 i++; 2147 if (m_hold->m_nextpkt == NULL) { 2148 m_hold->m_nextpkt = m; 2149 break; 2150 } 2151 } 2152 while (i >= V_nd6_maxqueuelen) { 2153 m_hold = ln->ln_hold; 2154 ln->ln_hold = ln->ln_hold->m_nextpkt; 2155 m_freem(m_hold); 2156 i--; 2157 } 2158 } else { 2159 ln->ln_hold = m; 2160 } 2161 2162 /* 2163 * If there has been no NS for the neighbor after entering the 2164 * INCOMPLETE state, send the first solicitation. 2165 */ 2166 if (!ND6_LLINFO_PERMANENT(ln) && ln->ln_asked == 0) { 2167 ln->ln_asked++; 2168 nd6_llinfo_settimer(ln, 2169 (long)ND_IFINFO(ifp)->retrans * hz / 1000); 2170 nd6_ns_output(ifp, NULL, &dst->sin6_addr, ln, 0); 2171 } 2172 return (0); 2173 2174 sendpkt: 2175 /* discard the packet if IPv6 operation is disabled on the interface */ 2176 if ((ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED)) { 2177 error = ENETDOWN; /* better error? */ 2178 goto bad; 2179 } 2180 2181 #ifdef MAC 2182 mac_netinet6_nd6_send(ifp, m); 2183 #endif 2184 if ((ifp->if_flags & IFF_LOOPBACK) != 0) { 2185 return ((*ifp->if_output)(origifp, m, (struct sockaddr *)dst, 2186 rt)); 2187 } 2188 return ((*ifp->if_output)(ifp, m, (struct sockaddr *)dst, rt)); 2189 2190 bad: 2191 if (m) 2192 m_freem(m); 2193 return (error); 2194 } 2195 #undef senderr 2196 2197 int 2198 nd6_need_cache(struct ifnet *ifp) 2199 { 2200 /* 2201 * XXX: we currently do not make neighbor cache on any interface 2202 * other than ARCnet, Ethernet, FDDI and GIF. 2203 * 2204 * RFC2893 says: 2205 * - unidirectional tunnels needs no ND 2206 */ 2207 switch (ifp->if_type) { 2208 case IFT_ARCNET: 2209 case IFT_ETHER: 2210 case IFT_FDDI: 2211 case IFT_IEEE1394: 2212 #ifdef IFT_L2VLAN 2213 case IFT_L2VLAN: 2214 #endif 2215 #ifdef IFT_IEEE80211 2216 case IFT_IEEE80211: 2217 #endif 2218 #ifdef IFT_CARP 2219 case IFT_CARP: 2220 #endif 2221 case IFT_GIF: /* XXX need more cases? */ 2222 case IFT_PPP: 2223 case IFT_TUNNEL: 2224 case IFT_BRIDGE: 2225 case IFT_PROPVIRTUAL: 2226 return (1); 2227 default: 2228 return (0); 2229 } 2230 } 2231 2232 int 2233 nd6_storelladdr(struct ifnet *ifp, struct rtentry *rt0, struct mbuf *m, 2234 struct sockaddr *dst, u_char *desten) 2235 { 2236 struct sockaddr_dl *sdl; 2237 struct rtentry *rt; 2238 int error; 2239 2240 if (m->m_flags & M_MCAST) { 2241 int i; 2242 2243 switch (ifp->if_type) { 2244 case IFT_ETHER: 2245 case IFT_FDDI: 2246 #ifdef IFT_L2VLAN 2247 case IFT_L2VLAN: 2248 #endif 2249 #ifdef IFT_IEEE80211 2250 case IFT_IEEE80211: 2251 #endif 2252 case IFT_BRIDGE: 2253 case IFT_ISO88025: 2254 ETHER_MAP_IPV6_MULTICAST(&SIN6(dst)->sin6_addr, 2255 desten); 2256 return (0); 2257 case IFT_IEEE1394: 2258 /* 2259 * netbsd can use if_broadcastaddr, but we don't do so 2260 * to reduce # of ifdef. 2261 */ 2262 for (i = 0; i < ifp->if_addrlen; i++) 2263 desten[i] = ~0; 2264 return (0); 2265 case IFT_ARCNET: 2266 *desten = 0; 2267 return (0); 2268 default: 2269 m_freem(m); 2270 return (EAFNOSUPPORT); 2271 } 2272 } 2273 2274 if (rt0 == NULL) { 2275 /* this could happen, if we could not allocate memory */ 2276 m_freem(m); 2277 return (ENOMEM); 2278 } 2279 2280 error = rt_check(&rt, &rt0, dst); 2281 if (error) { 2282 m_freem(m); 2283 return (error); 2284 } 2285 RT_UNLOCK(rt); 2286 2287 if (rt->rt_gateway->sa_family != AF_LINK) { 2288 printf("nd6_storelladdr: something odd happens\n"); 2289 m_freem(m); 2290 return (EINVAL); 2291 } 2292 sdl = SDL(rt->rt_gateway); 2293 if (sdl->sdl_alen == 0) { 2294 /* this should be impossible, but we bark here for debugging */ 2295 printf("nd6_storelladdr: sdl_alen == 0\n"); 2296 m_freem(m); 2297 return (EINVAL); 2298 } 2299 2300 bcopy(LLADDR(sdl), desten, sdl->sdl_alen); 2301 return (0); 2302 } 2303 2304 static void 2305 clear_llinfo_pqueue(struct llinfo_nd6 *ln) 2306 { 2307 struct mbuf *m_hold, *m_hold_next; 2308 2309 for (m_hold = ln->ln_hold; m_hold; m_hold = m_hold_next) { 2310 m_hold_next = m_hold->m_nextpkt; 2311 m_hold->m_nextpkt = NULL; 2312 m_freem(m_hold); 2313 } 2314 2315 ln->ln_hold = NULL; 2316 return; 2317 } 2318 2319 static int nd6_sysctl_drlist(SYSCTL_HANDLER_ARGS); 2320 static int nd6_sysctl_prlist(SYSCTL_HANDLER_ARGS); 2321 #ifdef SYSCTL_DECL 2322 SYSCTL_DECL(_net_inet6_icmp6); 2323 #endif 2324 SYSCTL_NODE(_net_inet6_icmp6, ICMPV6CTL_ND6_DRLIST, nd6_drlist, 2325 CTLFLAG_RD, nd6_sysctl_drlist, ""); 2326 SYSCTL_NODE(_net_inet6_icmp6, ICMPV6CTL_ND6_PRLIST, nd6_prlist, 2327 CTLFLAG_RD, nd6_sysctl_prlist, ""); 2328 SYSCTL_V_INT(V_NET, vnet_inet6, _net_inet6_icmp6, ICMPV6CTL_ND6_MAXQLEN, 2329 nd6_maxqueuelen, CTLFLAG_RW, nd6_maxqueuelen, 1, ""); 2330 2331 static int 2332 nd6_sysctl_drlist(SYSCTL_HANDLER_ARGS) 2333 { 2334 INIT_VNET_INET6(curvnet); 2335 int error; 2336 char buf[1024] __aligned(4); 2337 struct in6_defrouter *d, *de; 2338 struct nd_defrouter *dr; 2339 2340 if (req->newptr) 2341 return EPERM; 2342 error = 0; 2343 2344 for (dr = TAILQ_FIRST(&V_nd_defrouter); dr; 2345 dr = TAILQ_NEXT(dr, dr_entry)) { 2346 d = (struct in6_defrouter *)buf; 2347 de = (struct in6_defrouter *)(buf + sizeof(buf)); 2348 2349 if (d + 1 <= de) { 2350 bzero(d, sizeof(*d)); 2351 d->rtaddr.sin6_family = AF_INET6; 2352 d->rtaddr.sin6_len = sizeof(d->rtaddr); 2353 d->rtaddr.sin6_addr = dr->rtaddr; 2354 error = sa6_recoverscope(&d->rtaddr); 2355 if (error != 0) 2356 return (error); 2357 d->flags = dr->flags; 2358 d->rtlifetime = dr->rtlifetime; 2359 d->expire = dr->expire; 2360 d->if_index = dr->ifp->if_index; 2361 } else 2362 panic("buffer too short"); 2363 2364 error = SYSCTL_OUT(req, buf, sizeof(*d)); 2365 if (error) 2366 break; 2367 } 2368 2369 return (error); 2370 } 2371 2372 static int 2373 nd6_sysctl_prlist(SYSCTL_HANDLER_ARGS) 2374 { 2375 INIT_VNET_INET6(curvnet); 2376 int error; 2377 char buf[1024] __aligned(4); 2378 struct in6_prefix *p, *pe; 2379 struct nd_prefix *pr; 2380 char ip6buf[INET6_ADDRSTRLEN]; 2381 2382 if (req->newptr) 2383 return EPERM; 2384 error = 0; 2385 2386 for (pr = V_nd_prefix.lh_first; pr; pr = pr->ndpr_next) { 2387 u_short advrtrs; 2388 size_t advance; 2389 struct sockaddr_in6 *sin6, *s6; 2390 struct nd_pfxrouter *pfr; 2391 2392 p = (struct in6_prefix *)buf; 2393 pe = (struct in6_prefix *)(buf + sizeof(buf)); 2394 2395 if (p + 1 <= pe) { 2396 bzero(p, sizeof(*p)); 2397 sin6 = (struct sockaddr_in6 *)(p + 1); 2398 2399 p->prefix = pr->ndpr_prefix; 2400 if (sa6_recoverscope(&p->prefix)) { 2401 log(LOG_ERR, 2402 "scope error in prefix list (%s)\n", 2403 ip6_sprintf(ip6buf, &p->prefix.sin6_addr)); 2404 /* XXX: press on... */ 2405 } 2406 p->raflags = pr->ndpr_raf; 2407 p->prefixlen = pr->ndpr_plen; 2408 p->vltime = pr->ndpr_vltime; 2409 p->pltime = pr->ndpr_pltime; 2410 p->if_index = pr->ndpr_ifp->if_index; 2411 if (pr->ndpr_vltime == ND6_INFINITE_LIFETIME) 2412 p->expire = 0; 2413 else { 2414 time_t maxexpire; 2415 2416 /* XXX: we assume time_t is signed. */ 2417 maxexpire = (-1) & 2418 ~((time_t)1 << 2419 ((sizeof(maxexpire) * 8) - 1)); 2420 if (pr->ndpr_vltime < 2421 maxexpire - pr->ndpr_lastupdate) { 2422 p->expire = pr->ndpr_lastupdate + 2423 pr->ndpr_vltime; 2424 } else 2425 p->expire = maxexpire; 2426 } 2427 p->refcnt = pr->ndpr_refcnt; 2428 p->flags = pr->ndpr_stateflags; 2429 p->origin = PR_ORIG_RA; 2430 advrtrs = 0; 2431 for (pfr = pr->ndpr_advrtrs.lh_first; pfr; 2432 pfr = pfr->pfr_next) { 2433 if ((void *)&sin6[advrtrs + 1] > (void *)pe) { 2434 advrtrs++; 2435 continue; 2436 } 2437 s6 = &sin6[advrtrs]; 2438 bzero(s6, sizeof(*s6)); 2439 s6->sin6_family = AF_INET6; 2440 s6->sin6_len = sizeof(*sin6); 2441 s6->sin6_addr = pfr->router->rtaddr; 2442 if (sa6_recoverscope(s6)) { 2443 log(LOG_ERR, 2444 "scope error in " 2445 "prefix list (%s)\n", 2446 ip6_sprintf(ip6buf, 2447 &pfr->router->rtaddr)); 2448 } 2449 advrtrs++; 2450 } 2451 p->advrtrs = advrtrs; 2452 } else 2453 panic("buffer too short"); 2454 2455 advance = sizeof(*p) + sizeof(*sin6) * advrtrs; 2456 error = SYSCTL_OUT(req, buf, advance); 2457 if (error) 2458 break; 2459 } 2460 2461 return (error); 2462 } 2463