1 /* $NetBSD: nd6.c,v 1.77 2002/10/09 20:22:16 itojun Exp $ */ 2 /* $KAME: nd6.c,v 1.279 2002/06/08 11:16:51 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 <sys/cdefs.h> 34 __KERNEL_RCSID(0, "$NetBSD: nd6.c,v 1.77 2002/10/09 20:22:16 itojun Exp $"); 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/ioctl.h> 48 #include <sys/syslog.h> 49 #include <sys/queue.h> 50 51 #include <net/if.h> 52 #include <net/if_dl.h> 53 #include <net/if_types.h> 54 #include <net/route.h> 55 #include <net/if_ether.h> 56 #include <net/if_fddi.h> 57 #include <net/if_arc.h> 58 59 #include <netinet/in.h> 60 #include <netinet6/in6_var.h> 61 #include <netinet/ip6.h> 62 #include <netinet6/ip6_var.h> 63 #include <netinet6/nd6.h> 64 #include <netinet/icmp6.h> 65 66 #include "loop.h" 67 extern struct ifnet loif[NLOOP]; 68 69 #include <net/net_osdep.h> 70 71 #define ND6_SLOWTIMER_INTERVAL (60 * 60) /* 1 hour */ 72 #define ND6_RECALC_REACHTM_INTERVAL (60 * 120) /* 2 hours */ 73 74 #define SIN6(s) ((struct sockaddr_in6 *)s) 75 #define SDL(s) ((struct sockaddr_dl *)s) 76 77 /* timer values */ 78 int nd6_prune = 1; /* walk list every 1 seconds */ 79 int nd6_delay = 5; /* delay first probe time 5 second */ 80 int nd6_umaxtries = 3; /* maximum unicast query */ 81 int nd6_mmaxtries = 3; /* maximum multicast query */ 82 int nd6_useloopback = 1; /* use loopback interface for local traffic */ 83 int nd6_gctimer = (60 * 60 * 24); /* 1 day: garbage collection timer */ 84 85 /* preventing too many loops in ND option parsing */ 86 int nd6_maxndopt = 10; /* max # of ND options allowed */ 87 88 int nd6_maxnudhint = 0; /* max # of subsequent upper layer hints */ 89 90 #ifdef ND6_DEBUG 91 int nd6_debug = 1; 92 #else 93 int nd6_debug = 0; 94 #endif 95 96 /* for debugging? */ 97 static int nd6_inuse, nd6_allocated; 98 99 struct llinfo_nd6 llinfo_nd6 = {&llinfo_nd6, &llinfo_nd6}; 100 struct nd_drhead nd_defrouter; 101 struct nd_prhead nd_prefix = { 0 }; 102 103 int nd6_recalc_reachtm_interval = ND6_RECALC_REACHTM_INTERVAL; 104 static struct sockaddr_in6 all1_sa; 105 106 static void nd6_setmtu0 __P((struct ifnet *, struct nd_ifinfo *)); 107 static void nd6_slowtimo __P((void *)); 108 static struct llinfo_nd6 *nd6_free __P((struct rtentry *, int)); 109 110 struct callout nd6_slowtimo_ch; 111 struct callout nd6_timer_ch; 112 113 static int fill_drlist __P((void *, size_t *, size_t)); 114 static int fill_prlist __P((void *, size_t *, size_t)); 115 116 void 117 nd6_init() 118 { 119 static int nd6_init_done = 0; 120 int i; 121 122 if (nd6_init_done) { 123 log(LOG_NOTICE, "nd6_init called more than once(ignored)\n"); 124 return; 125 } 126 127 all1_sa.sin6_family = AF_INET6; 128 all1_sa.sin6_len = sizeof(struct sockaddr_in6); 129 for (i = 0; i < sizeof(all1_sa.sin6_addr); i++) 130 all1_sa.sin6_addr.s6_addr[i] = 0xff; 131 132 /* initialization of the default router list */ 133 TAILQ_INIT(&nd_defrouter); 134 135 nd6_init_done = 1; 136 137 /* start timer */ 138 callout_reset(&nd6_slowtimo_ch, ND6_SLOWTIMER_INTERVAL * hz, 139 nd6_slowtimo, NULL); 140 } 141 142 struct nd_ifinfo * 143 nd6_ifattach(ifp) 144 struct ifnet *ifp; 145 { 146 struct nd_ifinfo *nd; 147 148 nd = (struct nd_ifinfo *)malloc(sizeof(*nd), M_IP6NDP, M_WAITOK); 149 bzero(nd, sizeof(*nd)); 150 151 nd->initialized = 1; 152 153 nd->chlim = IPV6_DEFHLIM; 154 nd->basereachable = REACHABLE_TIME; 155 nd->reachable = ND_COMPUTE_RTIME(nd->basereachable); 156 nd->retrans = RETRANS_TIMER; 157 /* 158 * Note that the default value of ip6_accept_rtadv is 0, which means 159 * we won't accept RAs by default even if we set ND6_IFF_ACCEPT_RTADV 160 * here. 161 */ 162 nd->flags = (ND6_IFF_PERFORMNUD | ND6_IFF_ACCEPT_RTADV); 163 164 /* XXX: we cannot call nd6_setmtu since ifp is not fully initialized */ 165 nd6_setmtu0(ifp, nd); 166 167 return nd; 168 } 169 170 void 171 nd6_ifdetach(nd) 172 struct nd_ifinfo *nd; 173 { 174 175 free(nd, M_IP6NDP); 176 } 177 178 void 179 nd6_setmtu(ifp) 180 struct ifnet *ifp; 181 { 182 nd6_setmtu0(ifp, ND_IFINFO(ifp)); 183 } 184 185 void 186 nd6_setmtu0(ifp, ndi) 187 struct ifnet *ifp; 188 struct nd_ifinfo *ndi; 189 { 190 u_int32_t omaxmtu; 191 192 omaxmtu = ndi->maxmtu; 193 194 switch (ifp->if_type) { 195 case IFT_ARCNET: 196 ndi->maxmtu = MIN(ARC_PHDS_MAXMTU, ifp->if_mtu); /* RFC2497 */ 197 break; 198 case IFT_FDDI: 199 ndi->maxmtu = MIN(FDDIIPMTU, ifp->if_mtu); 200 break; 201 default: 202 ndi->maxmtu = ifp->if_mtu; 203 break; 204 } 205 206 /* 207 * Decreasing the interface MTU under IPV6 minimum MTU may cause 208 * undesirable situation. We thus notify the operator of the change 209 * explicitly. The check for omaxmtu is necessary to restrict the 210 * log to the case of changing the MTU, not initializing it. 211 */ 212 if (omaxmtu >= IPV6_MMTU && ndi->maxmtu < IPV6_MMTU) { 213 log(LOG_NOTICE, "nd6_setmtu0: " 214 "new link MTU on %s (%lu) is too small for IPv6\n", 215 if_name(ifp), (unsigned long)ndi->maxmtu); 216 } 217 218 if (ndi->maxmtu > in6_maxmtu) 219 in6_setmaxmtu(); /* check all interfaces just in case */ 220 } 221 222 void 223 nd6_option_init(opt, icmp6len, ndopts) 224 void *opt; 225 int icmp6len; 226 union nd_opts *ndopts; 227 { 228 229 bzero(ndopts, sizeof(*ndopts)); 230 ndopts->nd_opts_search = (struct nd_opt_hdr *)opt; 231 ndopts->nd_opts_last 232 = (struct nd_opt_hdr *)(((u_char *)opt) + icmp6len); 233 234 if (icmp6len == 0) { 235 ndopts->nd_opts_done = 1; 236 ndopts->nd_opts_search = NULL; 237 } 238 } 239 240 /* 241 * Take one ND option. 242 */ 243 struct nd_opt_hdr * 244 nd6_option(ndopts) 245 union nd_opts *ndopts; 246 { 247 struct nd_opt_hdr *nd_opt; 248 int olen; 249 250 if (!ndopts) 251 panic("ndopts == NULL in nd6_option"); 252 if (!ndopts->nd_opts_last) 253 panic("uninitialized ndopts in nd6_option"); 254 if (!ndopts->nd_opts_search) 255 return NULL; 256 if (ndopts->nd_opts_done) 257 return NULL; 258 259 nd_opt = ndopts->nd_opts_search; 260 261 /* make sure nd_opt_len is inside the buffer */ 262 if ((caddr_t)&nd_opt->nd_opt_len >= (caddr_t)ndopts->nd_opts_last) { 263 bzero(ndopts, sizeof(*ndopts)); 264 return NULL; 265 } 266 267 olen = nd_opt->nd_opt_len << 3; 268 if (olen == 0) { 269 /* 270 * Message validation requires that all included 271 * options have a length that is greater than zero. 272 */ 273 bzero(ndopts, sizeof(*ndopts)); 274 return NULL; 275 } 276 277 ndopts->nd_opts_search = (struct nd_opt_hdr *)((caddr_t)nd_opt + olen); 278 if (ndopts->nd_opts_search > ndopts->nd_opts_last) { 279 /* option overruns the end of buffer, invalid */ 280 bzero(ndopts, sizeof(*ndopts)); 281 return NULL; 282 } else if (ndopts->nd_opts_search == ndopts->nd_opts_last) { 283 /* reached the end of options chain */ 284 ndopts->nd_opts_done = 1; 285 ndopts->nd_opts_search = NULL; 286 } 287 return nd_opt; 288 } 289 290 /* 291 * Parse multiple ND options. 292 * This function is much easier to use, for ND routines that do not need 293 * multiple options of the same type. 294 */ 295 int 296 nd6_options(ndopts) 297 union nd_opts *ndopts; 298 { 299 struct nd_opt_hdr *nd_opt; 300 int i = 0; 301 302 if (!ndopts) 303 panic("ndopts == NULL in nd6_options"); 304 if (!ndopts->nd_opts_last) 305 panic("uninitialized ndopts in nd6_options"); 306 if (!ndopts->nd_opts_search) 307 return 0; 308 309 while (1) { 310 nd_opt = nd6_option(ndopts); 311 if (!nd_opt && !ndopts->nd_opts_last) { 312 /* 313 * Message validation requires that all included 314 * options have a length that is greater than zero. 315 */ 316 icmp6stat.icp6s_nd_badopt++; 317 bzero(ndopts, sizeof(*ndopts)); 318 return -1; 319 } 320 321 if (!nd_opt) 322 goto skip1; 323 324 switch (nd_opt->nd_opt_type) { 325 case ND_OPT_SOURCE_LINKADDR: 326 case ND_OPT_TARGET_LINKADDR: 327 case ND_OPT_MTU: 328 case ND_OPT_REDIRECTED_HEADER: 329 if (ndopts->nd_opt_array[nd_opt->nd_opt_type]) { 330 nd6log((LOG_INFO, 331 "duplicated ND6 option found (type=%d)\n", 332 nd_opt->nd_opt_type)); 333 /* XXX bark? */ 334 } else { 335 ndopts->nd_opt_array[nd_opt->nd_opt_type] 336 = nd_opt; 337 } 338 break; 339 case ND_OPT_PREFIX_INFORMATION: 340 if (ndopts->nd_opt_array[nd_opt->nd_opt_type] == 0) { 341 ndopts->nd_opt_array[nd_opt->nd_opt_type] 342 = nd_opt; 343 } 344 ndopts->nd_opts_pi_end = 345 (struct nd_opt_prefix_info *)nd_opt; 346 break; 347 default: 348 /* 349 * Unknown options must be silently ignored, 350 * to accomodate future extension to the protocol. 351 */ 352 nd6log((LOG_DEBUG, 353 "nd6_options: unsupported option %d - " 354 "option ignored\n", nd_opt->nd_opt_type)); 355 } 356 357 skip1: 358 i++; 359 if (i > nd6_maxndopt) { 360 icmp6stat.icp6s_nd_toomanyopt++; 361 nd6log((LOG_INFO, "too many loop in nd opt\n")); 362 break; 363 } 364 365 if (ndopts->nd_opts_done) 366 break; 367 } 368 369 return 0; 370 } 371 372 /* 373 * ND6 timer routine to expire default route list and prefix list 374 */ 375 void 376 nd6_timer(ignored_arg) 377 void *ignored_arg; 378 { 379 int s; 380 struct llinfo_nd6 *ln; 381 struct nd_defrouter *dr; 382 struct nd_prefix *pr; 383 long time_second = time.tv_sec; 384 struct ifnet *ifp; 385 struct in6_ifaddr *ia6, *nia6; 386 struct in6_addrlifetime *lt6; 387 388 s = splsoftnet(); 389 callout_reset(&nd6_timer_ch, nd6_prune * hz, 390 nd6_timer, NULL); 391 392 ln = llinfo_nd6.ln_next; 393 while (ln && ln != &llinfo_nd6) { 394 struct rtentry *rt; 395 struct sockaddr_in6 *dst; 396 struct llinfo_nd6 *next = ln->ln_next; 397 /* XXX: used for the DELAY case only: */ 398 struct nd_ifinfo *ndi = NULL; 399 400 if ((rt = ln->ln_rt) == NULL) { 401 ln = next; 402 continue; 403 } 404 if ((ifp = rt->rt_ifp) == NULL) { 405 ln = next; 406 continue; 407 } 408 ndi = ND_IFINFO(ifp); 409 dst = (struct sockaddr_in6 *)rt_key(rt); 410 411 if (ln->ln_expire > time_second) { 412 ln = next; 413 continue; 414 } 415 416 /* sanity check */ 417 if (!rt) 418 panic("rt=0 in nd6_timer(ln=%p)", ln); 419 if (rt->rt_llinfo && (struct llinfo_nd6 *)rt->rt_llinfo != ln) 420 panic("rt_llinfo(%p) is not equal to ln(%p)", 421 rt->rt_llinfo, ln); 422 if (!dst) 423 panic("dst=0 in nd6_timer(ln=%p)", ln); 424 425 switch (ln->ln_state) { 426 case ND6_LLINFO_INCOMPLETE: 427 if (ln->ln_asked < nd6_mmaxtries) { 428 ln->ln_asked++; 429 ln->ln_expire = time_second + 430 ND6_RETRANS_SEC(ND_IFINFO(ifp)->retrans); 431 nd6_ns_output(ifp, NULL, &dst->sin6_addr, 432 ln, 0); 433 } else { 434 struct mbuf *m = ln->ln_hold; 435 if (m) { 436 /* 437 * Fake rcvif to make the ICMP error 438 * more helpful in diagnosing for the 439 * receiver. 440 * XXX: should we consider 441 * older rcvif? 442 */ 443 m->m_pkthdr.rcvif = rt->rt_ifp; 444 445 icmp6_error(m, ICMP6_DST_UNREACH, 446 ICMP6_DST_UNREACH_ADDR, 0); 447 ln->ln_hold = NULL; 448 } 449 next = nd6_free(rt, 0); 450 } 451 break; 452 case ND6_LLINFO_REACHABLE: 453 if (ln->ln_expire) { 454 ln->ln_state = ND6_LLINFO_STALE; 455 ln->ln_expire = time_second + nd6_gctimer; 456 } 457 break; 458 459 case ND6_LLINFO_STALE: 460 /* Garbage Collection(RFC 2461 5.3) */ 461 if (ln->ln_expire) 462 next = nd6_free(rt, 1); 463 break; 464 465 case ND6_LLINFO_DELAY: 466 if (ndi && (ndi->flags & ND6_IFF_PERFORMNUD) != 0) { 467 /* We need NUD */ 468 ln->ln_asked = 1; 469 ln->ln_state = ND6_LLINFO_PROBE; 470 ln->ln_expire = time_second + 471 ND6_RETRANS_SEC(ndi->retrans); 472 nd6_ns_output(ifp, &dst->sin6_addr, 473 &dst->sin6_addr, ln, 0); 474 } else { 475 ln->ln_state = ND6_LLINFO_STALE; /* XXX */ 476 ln->ln_expire = time_second + nd6_gctimer; 477 } 478 break; 479 case ND6_LLINFO_PROBE: 480 if (ln->ln_asked < nd6_umaxtries) { 481 ln->ln_asked++; 482 ln->ln_expire = time_second + 483 ND6_RETRANS_SEC(ND_IFINFO(ifp)->retrans); 484 nd6_ns_output(ifp, &dst->sin6_addr, 485 &dst->sin6_addr, ln, 0); 486 } else { 487 next = nd6_free(rt, 0); 488 } 489 break; 490 } 491 ln = next; 492 } 493 494 /* expire default router list */ 495 dr = TAILQ_FIRST(&nd_defrouter); 496 while (dr) { 497 if (dr->expire && dr->expire < time_second) { 498 struct nd_defrouter *t; 499 t = TAILQ_NEXT(dr, dr_entry); 500 defrtrlist_del(dr); 501 dr = t; 502 } else { 503 dr = TAILQ_NEXT(dr, dr_entry); 504 } 505 } 506 507 /* 508 * expire interface addresses. 509 * in the past the loop was inside prefix expiry processing. 510 * However, from a stricter speci-confrmance standpoint, we should 511 * rather separate address lifetimes and prefix lifetimes. 512 */ 513 for (ia6 = in6_ifaddr; ia6; ia6 = nia6) { 514 nia6 = ia6->ia_next; 515 /* check address lifetime */ 516 lt6 = &ia6->ia6_lifetime; 517 if (IFA6_IS_INVALID(ia6)) { 518 in6_purgeaddr(&ia6->ia_ifa); 519 } 520 if (IFA6_IS_DEPRECATED(ia6)) { 521 ia6->ia6_flags |= IN6_IFF_DEPRECATED; 522 } else { 523 /* 524 * A new RA might have made a deprecated address 525 * preferred. 526 */ 527 ia6->ia6_flags &= ~IN6_IFF_DEPRECATED; 528 } 529 } 530 531 /* expire prefix list */ 532 pr = nd_prefix.lh_first; 533 while (pr) { 534 /* 535 * check prefix lifetime. 536 * since pltime is just for autoconf, pltime processing for 537 * prefix is not necessary. 538 */ 539 if (pr->ndpr_vltime != ND6_INFINITE_LIFETIME && 540 time_second - pr->ndpr_lastupdate > pr->ndpr_vltime) { 541 struct nd_prefix *t; 542 t = pr->ndpr_next; 543 544 /* 545 * address expiration and prefix expiration are 546 * separate. NEVER perform in6_purgeaddr here. 547 */ 548 549 prelist_remove(pr); 550 pr = t; 551 } else 552 pr = pr->ndpr_next; 553 } 554 splx(s); 555 } 556 557 /* 558 * Nuke neighbor cache/prefix/default router management table, right before 559 * ifp goes away. 560 */ 561 void 562 nd6_purge(ifp) 563 struct ifnet *ifp; 564 { 565 struct llinfo_nd6 *ln, *nln; 566 struct nd_defrouter *dr, *ndr; 567 struct nd_prefix *pr, *npr; 568 569 /* 570 * Nuke default router list entries toward ifp. 571 * We defer removal of default router list entries that is installed 572 * in the routing table, in order to keep additional side effects as 573 * small as possible. 574 */ 575 for (dr = TAILQ_FIRST(&nd_defrouter); dr; dr = ndr) { 576 ndr = TAILQ_NEXT(dr, dr_entry); 577 if (dr->installed) 578 continue; 579 580 if (dr->ifp == ifp) 581 defrtrlist_del(dr); 582 } 583 for (dr = TAILQ_FIRST(&nd_defrouter); dr; dr = ndr) { 584 ndr = TAILQ_NEXT(dr, dr_entry); 585 if (!dr->installed) 586 continue; 587 588 if (dr->ifp == ifp) 589 defrtrlist_del(dr); 590 } 591 592 /* Nuke prefix list entries toward ifp */ 593 for (pr = nd_prefix.lh_first; pr; pr = npr) { 594 npr = pr->ndpr_next; 595 if (pr->ndpr_ifp == ifp) { 596 /* 597 * Previously, pr->ndpr_addr is removed as well, 598 * but I strongly believe we don't have to do it. 599 * nd6_purge() is only called from in6_ifdetach(), 600 * which removes all the associated interface addresses 601 * by itself. 602 * (jinmei@kame.net 20010129) 603 */ 604 prelist_remove(pr); 605 } 606 } 607 608 /* cancel default outgoing interface setting */ 609 if (nd6_defifindex == ifp->if_index) 610 nd6_setdefaultiface(0); 611 612 if (!ip6_forwarding && ip6_accept_rtadv) { /* XXX: too restrictive? */ 613 /* refresh default router list */ 614 defrouter_select(); 615 } 616 617 /* 618 * Nuke neighbor cache entries for the ifp. 619 * Note that rt->rt_ifp may not be the same as ifp, 620 * due to KAME goto ours hack. See RTM_RESOLVE case in 621 * nd6_rtrequest(), and ip6_input(). 622 */ 623 ln = llinfo_nd6.ln_next; 624 while (ln && ln != &llinfo_nd6) { 625 struct rtentry *rt; 626 struct sockaddr_dl *sdl; 627 628 nln = ln->ln_next; 629 rt = ln->ln_rt; 630 if (rt && rt->rt_gateway && 631 rt->rt_gateway->sa_family == AF_LINK) { 632 sdl = (struct sockaddr_dl *)rt->rt_gateway; 633 if (sdl->sdl_index == ifp->if_index) 634 nln = nd6_free(rt, 0); 635 } 636 ln = nln; 637 } 638 } 639 640 struct rtentry * 641 nd6_lookup(addr6, create, ifp) 642 struct in6_addr *addr6; 643 int create; 644 struct ifnet *ifp; 645 { 646 struct rtentry *rt; 647 struct sockaddr_in6 sin6; 648 649 bzero(&sin6, sizeof(sin6)); 650 sin6.sin6_len = sizeof(struct sockaddr_in6); 651 sin6.sin6_family = AF_INET6; 652 sin6.sin6_addr = *addr6; 653 rt = rtalloc1((struct sockaddr *)&sin6, create); 654 if (rt && (rt->rt_flags & RTF_LLINFO) == 0) { 655 /* 656 * This is the case for the default route. 657 * If we want to create a neighbor cache for the address, we 658 * should free the route for the destination and allocate an 659 * interface route. 660 */ 661 if (create) { 662 RTFREE(rt); 663 rt = 0; 664 } 665 } 666 if (!rt) { 667 if (create && ifp) { 668 int e; 669 670 /* 671 * If no route is available and create is set, 672 * we allocate a host route for the destination 673 * and treat it like an interface route. 674 * This hack is necessary for a neighbor which can't 675 * be covered by our own prefix. 676 */ 677 struct ifaddr *ifa = 678 ifaof_ifpforaddr((struct sockaddr *)&sin6, ifp); 679 if (ifa == NULL) 680 return (NULL); 681 682 /* 683 * Create a new route. RTF_LLINFO is necessary 684 * to create a Neighbor Cache entry for the 685 * destination in nd6_rtrequest which will be 686 * called in rtrequest via ifa->ifa_rtrequest. 687 */ 688 if ((e = rtrequest(RTM_ADD, (struct sockaddr *)&sin6, 689 ifa->ifa_addr, (struct sockaddr *)&all1_sa, 690 (ifa->ifa_flags | RTF_HOST | RTF_LLINFO) & 691 ~RTF_CLONING, &rt)) != 0) { 692 #if 0 693 log(LOG_ERR, 694 "nd6_lookup: failed to add route for a " 695 "neighbor(%s), errno=%d\n", 696 ip6_sprintf(addr6), e); 697 #endif 698 return (NULL); 699 } 700 if (rt == NULL) 701 return (NULL); 702 if (rt->rt_llinfo) { 703 struct llinfo_nd6 *ln = 704 (struct llinfo_nd6 *)rt->rt_llinfo; 705 ln->ln_state = ND6_LLINFO_NOSTATE; 706 } 707 } else 708 return (NULL); 709 } 710 rt->rt_refcnt--; 711 /* 712 * Validation for the entry. 713 * Note that the check for rt_llinfo is necessary because a cloned 714 * route from a parent route that has the L flag (e.g. the default 715 * route to a p2p interface) may have the flag, too, while the 716 * destination is not actually a neighbor. 717 * XXX: we can't use rt->rt_ifp to check for the interface, since 718 * it might be the loopback interface if the entry is for our 719 * own address on a non-loopback interface. Instead, we should 720 * use rt->rt_ifa->ifa_ifp, which would specify the REAL 721 * interface. 722 */ 723 if ((rt->rt_flags & RTF_GATEWAY) || (rt->rt_flags & RTF_LLINFO) == 0 || 724 rt->rt_gateway->sa_family != AF_LINK || rt->rt_llinfo == NULL || 725 (ifp && rt->rt_ifa->ifa_ifp != ifp)) { 726 if (create) { 727 nd6log((LOG_DEBUG, 728 "nd6_lookup: failed to lookup %s (if = %s)\n", 729 ip6_sprintf(addr6), 730 ifp ? if_name(ifp) : "unspec")); 731 } 732 return (NULL); 733 } 734 return (rt); 735 } 736 737 /* 738 * Detect if a given IPv6 address identifies a neighbor on a given link. 739 * XXX: should take care of the destination of a p2p link? 740 */ 741 int 742 nd6_is_addr_neighbor(addr, ifp) 743 struct sockaddr_in6 *addr; 744 struct ifnet *ifp; 745 { 746 struct nd_prefix *pr; 747 struct rtentry *rt; 748 749 /* 750 * A link-local address is always a neighbor. 751 * XXX: we should use the sin6_scope_id field rather than the embedded 752 * interface index. 753 * XXX: a link does not necessarily specify a single interface. 754 */ 755 if (IN6_IS_ADDR_LINKLOCAL(&addr->sin6_addr) && 756 ntohs(*(u_int16_t *)&addr->sin6_addr.s6_addr[2]) == ifp->if_index) 757 return (1); 758 759 /* 760 * If the address matches one of our on-link prefixes, it should be a 761 * neighbor. 762 */ 763 for (pr = nd_prefix.lh_first; pr; pr = pr->ndpr_next) { 764 if (pr->ndpr_ifp != ifp) 765 continue; 766 767 if (!(pr->ndpr_stateflags & NDPRF_ONLINK)) 768 continue; 769 770 if (IN6_ARE_MASKED_ADDR_EQUAL(&pr->ndpr_prefix.sin6_addr, 771 &addr->sin6_addr, &pr->ndpr_mask)) 772 return (1); 773 } 774 775 /* 776 * If the default router list is empty, all addresses are regarded 777 * as on-link, and thus, as a neighbor. 778 * XXX: we restrict the condition to hosts, because routers usually do 779 * not have the "default router list". 780 */ 781 if (!ip6_forwarding && TAILQ_FIRST(&nd_defrouter) == NULL && 782 nd6_defifindex == ifp->if_index) { 783 return (1); 784 } 785 786 /* 787 * Even if the address matches none of our addresses, it might be 788 * in the neighbor cache. 789 */ 790 if ((rt = nd6_lookup(&addr->sin6_addr, 0, ifp)) != NULL) 791 return (1); 792 793 return (0); 794 } 795 796 /* 797 * Free an nd6 llinfo entry. 798 * Since the function would cause significant changes in the kernel, DO NOT 799 * make it global, unless you have a strong reason for the change, and are sure 800 * that the change is safe. 801 */ 802 static struct llinfo_nd6 * 803 nd6_free(rt, gc) 804 struct rtentry *rt; 805 int gc; 806 { 807 struct llinfo_nd6 *ln = (struct llinfo_nd6 *)rt->rt_llinfo, *next; 808 struct in6_addr in6 = ((struct sockaddr_in6 *)rt_key(rt))->sin6_addr; 809 struct nd_defrouter *dr; 810 811 /* 812 * we used to have pfctlinput(PRC_HOSTDEAD) here. 813 * even though it is not harmful, it was not really necessary. 814 */ 815 816 if (!ip6_forwarding && ip6_accept_rtadv) { /* XXX: too restrictive? */ 817 int s; 818 s = splsoftnet(); 819 dr = defrouter_lookup(&((struct sockaddr_in6 *)rt_key(rt))->sin6_addr, 820 rt->rt_ifp); 821 822 if (dr != NULL && dr->expire && 823 ln->ln_state == ND6_LLINFO_STALE && gc) { 824 /* 825 * If the reason for the deletion is just garbage 826 * collection, and the neighbor is an active default 827 * router, do not delete it. Instead, reset the GC 828 * timer using the router's lifetime. 829 * Simply deleting the entry would affect default 830 * router selection, which is not necessarily a good 831 * thing, especially when we're using router preference 832 * values. 833 * XXX: the check for ln_state would be redundant, 834 * but we intentionally keep it just in case. 835 */ 836 ln->ln_expire = dr->expire; 837 splx(s); 838 return (ln->ln_next); 839 } 840 841 if (ln->ln_router || dr) { 842 /* 843 * rt6_flush must be called whether or not the neighbor 844 * is in the Default Router List. 845 * See a corresponding comment in nd6_na_input(). 846 */ 847 rt6_flush(&in6, rt->rt_ifp); 848 } 849 850 if (dr) { 851 /* 852 * Unreachablity of a router might affect the default 853 * router selection and on-link detection of advertised 854 * prefixes. 855 */ 856 857 /* 858 * Temporarily fake the state to choose a new default 859 * router and to perform on-link determination of 860 * prefixes correctly. 861 * Below the state will be set correctly, 862 * or the entry itself will be deleted. 863 */ 864 ln->ln_state = ND6_LLINFO_INCOMPLETE; 865 866 /* 867 * Since defrouter_select() does not affect the 868 * on-link determination and MIP6 needs the check 869 * before the default router selection, we perform 870 * the check now. 871 */ 872 pfxlist_onlink_check(); 873 874 /* 875 * refresh default router list 876 */ 877 defrouter_select(); 878 } 879 splx(s); 880 } 881 882 /* 883 * Before deleting the entry, remember the next entry as the 884 * return value. We need this because pfxlist_onlink_check() above 885 * might have freed other entries (particularly the old next entry) as 886 * a side effect (XXX). 887 */ 888 next = ln->ln_next; 889 890 /* 891 * Detach the route from the routing tree and the list of neighbor 892 * caches, and disable the route entry not to be used in already 893 * cached routes. 894 */ 895 rtrequest(RTM_DELETE, rt_key(rt), (struct sockaddr *)0, 896 rt_mask(rt), 0, (struct rtentry **)0); 897 898 return (next); 899 } 900 901 /* 902 * Upper-layer reachability hint for Neighbor Unreachability Detection. 903 * 904 * XXX cost-effective metods? 905 */ 906 void 907 nd6_nud_hint(rt, dst6, force) 908 struct rtentry *rt; 909 struct in6_addr *dst6; 910 int force; 911 { 912 struct llinfo_nd6 *ln; 913 long time_second = time.tv_sec; 914 915 /* 916 * If the caller specified "rt", use that. Otherwise, resolve the 917 * routing table by supplied "dst6". 918 */ 919 if (!rt) { 920 if (!dst6) 921 return; 922 if (!(rt = nd6_lookup(dst6, 0, NULL))) 923 return; 924 } 925 926 if ((rt->rt_flags & RTF_GATEWAY) != 0 || 927 (rt->rt_flags & RTF_LLINFO) == 0 || 928 !rt->rt_llinfo || !rt->rt_gateway || 929 rt->rt_gateway->sa_family != AF_LINK) { 930 /* This is not a host route. */ 931 return; 932 } 933 934 ln = (struct llinfo_nd6 *)rt->rt_llinfo; 935 if (ln->ln_state < ND6_LLINFO_REACHABLE) 936 return; 937 938 /* 939 * if we get upper-layer reachability confirmation many times, 940 * it is possible we have false information. 941 */ 942 if (!force) { 943 ln->ln_byhint++; 944 if (ln->ln_byhint > nd6_maxnudhint) 945 return; 946 } 947 948 ln->ln_state = ND6_LLINFO_REACHABLE; 949 if (ln->ln_expire) 950 ln->ln_expire = time_second + ND_IFINFO(rt->rt_ifp)->reachable; 951 } 952 953 void 954 nd6_rtrequest(req, rt, info) 955 int req; 956 struct rtentry *rt; 957 struct rt_addrinfo *info; /* xxx unused */ 958 { 959 struct sockaddr *gate = rt->rt_gateway; 960 struct llinfo_nd6 *ln = (struct llinfo_nd6 *)rt->rt_llinfo; 961 static struct sockaddr_dl null_sdl = {sizeof(null_sdl), AF_LINK}; 962 struct ifnet *ifp = rt->rt_ifp; 963 struct ifaddr *ifa; 964 long time_second = time.tv_sec; 965 int mine = 0; 966 967 if ((rt->rt_flags & RTF_GATEWAY) != 0) 968 return; 969 970 if (nd6_need_cache(ifp) == 0 && (rt->rt_flags & RTF_HOST) == 0) { 971 /* 972 * This is probably an interface direct route for a link 973 * which does not need neighbor caches (e.g. fe80::%lo0/64). 974 * We do not need special treatment below for such a route. 975 * Moreover, the RTF_LLINFO flag which would be set below 976 * would annoy the ndp(8) command. 977 */ 978 return; 979 } 980 981 if (req == RTM_RESOLVE && 982 (nd6_need_cache(ifp) == 0 || /* stf case */ 983 !nd6_is_addr_neighbor((struct sockaddr_in6 *)rt_key(rt), ifp))) { 984 /* 985 * FreeBSD and BSD/OS often make a cloned host route based 986 * on a less-specific route (e.g. the default route). 987 * If the less specific route does not have a "gateway" 988 * (this is the case when the route just goes to a p2p or an 989 * stf interface), we'll mistakenly make a neighbor cache for 990 * the host route, and will see strange neighbor solicitation 991 * for the corresponding destination. In order to avoid the 992 * confusion, we check if the destination of the route is 993 * a neighbor in terms of neighbor discovery, and stop the 994 * process if not. Additionally, we remove the LLINFO flag 995 * so that ndp(8) will not try to get the neighbor information 996 * of the destination. 997 */ 998 rt->rt_flags &= ~RTF_LLINFO; 999 return; 1000 } 1001 1002 switch (req) { 1003 case RTM_ADD: 1004 /* 1005 * There is no backward compatibility :) 1006 * 1007 * if ((rt->rt_flags & RTF_HOST) == 0 && 1008 * SIN(rt_mask(rt))->sin_addr.s_addr != 0xffffffff) 1009 * rt->rt_flags |= RTF_CLONING; 1010 */ 1011 if (rt->rt_flags & (RTF_CLONING | RTF_LLINFO)) { 1012 /* 1013 * Case 1: This route should come from 1014 * a route to interface. RTF_LLINFO flag is set 1015 * for a host route whose destination should be 1016 * treated as on-link. 1017 */ 1018 rt_setgate(rt, rt_key(rt), 1019 (struct sockaddr *)&null_sdl); 1020 gate = rt->rt_gateway; 1021 SDL(gate)->sdl_type = ifp->if_type; 1022 SDL(gate)->sdl_index = ifp->if_index; 1023 if (ln) 1024 ln->ln_expire = time_second; 1025 #if 1 1026 if (ln && ln->ln_expire == 0) { 1027 /* kludge for desktops */ 1028 #if 0 1029 printf("nd6_rtrequest: time.tv_sec is zero; " 1030 "treat it as 1\n"); 1031 #endif 1032 ln->ln_expire = 1; 1033 } 1034 #endif 1035 if ((rt->rt_flags & RTF_CLONING) != 0) 1036 break; 1037 } 1038 /* 1039 * In IPv4 code, we try to annonuce new RTF_ANNOUNCE entry here. 1040 * We don't do that here since llinfo is not ready yet. 1041 * 1042 * There are also couple of other things to be discussed: 1043 * - unsolicited NA code needs improvement beforehand 1044 * - RFC2461 says we MAY send multicast unsolicited NA 1045 * (7.2.6 paragraph 4), however, it also says that we 1046 * SHOULD provide a mechanism to prevent multicast NA storm. 1047 * we don't have anything like it right now. 1048 * note that the mechanism needs a mutual agreement 1049 * between proxies, which means that we need to implement 1050 * a new protocol, or a new kludge. 1051 * - from RFC2461 6.2.4, host MUST NOT send an unsolicited NA. 1052 * we need to check ip6forwarding before sending it. 1053 * (or should we allow proxy ND configuration only for 1054 * routers? there's no mention about proxy ND from hosts) 1055 */ 1056 #if 0 1057 /* XXX it does not work */ 1058 if (rt->rt_flags & RTF_ANNOUNCE) 1059 nd6_na_output(ifp, 1060 &SIN6(rt_key(rt))->sin6_addr, 1061 &SIN6(rt_key(rt))->sin6_addr, 1062 ip6_forwarding ? ND_NA_FLAG_ROUTER : 0, 1063 1, NULL); 1064 #endif 1065 /* FALLTHROUGH */ 1066 case RTM_RESOLVE: 1067 if ((ifp->if_flags & (IFF_POINTOPOINT | IFF_LOOPBACK)) == 0) { 1068 /* 1069 * Address resolution isn't necessary for a point to 1070 * point link, so we can skip this test for a p2p link. 1071 */ 1072 if (gate->sa_family != AF_LINK || 1073 gate->sa_len < sizeof(null_sdl)) { 1074 log(LOG_DEBUG, 1075 "nd6_rtrequest: bad gateway value: %s\n", 1076 if_name(ifp)); 1077 break; 1078 } 1079 SDL(gate)->sdl_type = ifp->if_type; 1080 SDL(gate)->sdl_index = ifp->if_index; 1081 } 1082 if (ln != NULL) 1083 break; /* This happens on a route change */ 1084 /* 1085 * Case 2: This route may come from cloning, or a manual route 1086 * add with a LL address. 1087 */ 1088 R_Malloc(ln, struct llinfo_nd6 *, sizeof(*ln)); 1089 rt->rt_llinfo = (caddr_t)ln; 1090 if (!ln) { 1091 log(LOG_DEBUG, "nd6_rtrequest: malloc failed\n"); 1092 break; 1093 } 1094 nd6_inuse++; 1095 nd6_allocated++; 1096 Bzero(ln, sizeof(*ln)); 1097 ln->ln_rt = rt; 1098 /* this is required for "ndp" command. - shin */ 1099 if (req == RTM_ADD) { 1100 /* 1101 * gate should have some valid AF_LINK entry, 1102 * and ln->ln_expire should have some lifetime 1103 * which is specified by ndp command. 1104 */ 1105 ln->ln_state = ND6_LLINFO_REACHABLE; 1106 ln->ln_byhint = 0; 1107 } else { 1108 /* 1109 * When req == RTM_RESOLVE, rt is created and 1110 * initialized in rtrequest(), so rt_expire is 0. 1111 */ 1112 ln->ln_state = ND6_LLINFO_NOSTATE; 1113 ln->ln_expire = time_second; 1114 } 1115 rt->rt_flags |= RTF_LLINFO; 1116 ln->ln_next = llinfo_nd6.ln_next; 1117 llinfo_nd6.ln_next = ln; 1118 ln->ln_prev = &llinfo_nd6; 1119 ln->ln_next->ln_prev = ln; 1120 1121 /* 1122 * check if rt_key(rt) is one of my address assigned 1123 * to the interface. 1124 */ 1125 ifa = (struct ifaddr *)in6ifa_ifpwithaddr(rt->rt_ifp, 1126 &SIN6(rt_key(rt))->sin6_addr); 1127 if (ifa) { 1128 caddr_t macp = nd6_ifptomac(ifp); 1129 ln->ln_expire = 0; 1130 ln->ln_state = ND6_LLINFO_REACHABLE; 1131 ln->ln_byhint = 0; 1132 mine = 1; 1133 if (macp) { 1134 Bcopy(macp, LLADDR(SDL(gate)), ifp->if_addrlen); 1135 SDL(gate)->sdl_alen = ifp->if_addrlen; 1136 } 1137 if (nd6_useloopback) { 1138 rt->rt_ifp = &loif[0]; /* XXX */ 1139 /* 1140 * Make sure rt_ifa be equal to the ifaddr 1141 * corresponding to the address. 1142 * We need this because when we refer 1143 * rt_ifa->ia6_flags in ip6_input, we assume 1144 * that the rt_ifa points to the address instead 1145 * of the loopback address. 1146 */ 1147 if (ifa != rt->rt_ifa) { 1148 IFAFREE(rt->rt_ifa); 1149 IFAREF(ifa); 1150 rt->rt_ifa = ifa; 1151 } 1152 } 1153 } else if (rt->rt_flags & RTF_ANNOUNCE) { 1154 ln->ln_expire = 0; 1155 ln->ln_state = ND6_LLINFO_REACHABLE; 1156 ln->ln_byhint = 0; 1157 1158 /* join solicited node multicast for proxy ND */ 1159 if (ifp->if_flags & IFF_MULTICAST) { 1160 struct in6_addr llsol; 1161 int error; 1162 1163 llsol = SIN6(rt_key(rt))->sin6_addr; 1164 llsol.s6_addr16[0] = htons(0xff02); 1165 llsol.s6_addr16[1] = htons(ifp->if_index); 1166 llsol.s6_addr32[1] = 0; 1167 llsol.s6_addr32[2] = htonl(1); 1168 llsol.s6_addr8[12] = 0xff; 1169 1170 if (!in6_addmulti(&llsol, ifp, &error)) { 1171 nd6log((LOG_ERR, "%s: failed to join " 1172 "%s (errno=%d)\n", if_name(ifp), 1173 ip6_sprintf(&llsol), error)); 1174 } 1175 } 1176 } 1177 break; 1178 1179 case RTM_DELETE: 1180 if (!ln) 1181 break; 1182 /* leave from solicited node multicast for proxy ND */ 1183 if ((rt->rt_flags & RTF_ANNOUNCE) != 0 && 1184 (ifp->if_flags & IFF_MULTICAST) != 0) { 1185 struct in6_addr llsol; 1186 struct in6_multi *in6m; 1187 1188 llsol = SIN6(rt_key(rt))->sin6_addr; 1189 llsol.s6_addr16[0] = htons(0xff02); 1190 llsol.s6_addr16[1] = htons(ifp->if_index); 1191 llsol.s6_addr32[1] = 0; 1192 llsol.s6_addr32[2] = htonl(1); 1193 llsol.s6_addr8[12] = 0xff; 1194 1195 IN6_LOOKUP_MULTI(llsol, ifp, in6m); 1196 if (in6m) 1197 in6_delmulti(in6m); 1198 } 1199 nd6_inuse--; 1200 ln->ln_next->ln_prev = ln->ln_prev; 1201 ln->ln_prev->ln_next = ln->ln_next; 1202 ln->ln_prev = NULL; 1203 rt->rt_llinfo = 0; 1204 rt->rt_flags &= ~RTF_LLINFO; 1205 if (ln->ln_hold) 1206 m_freem(ln->ln_hold); 1207 Free((caddr_t)ln); 1208 } 1209 } 1210 1211 int 1212 nd6_ioctl(cmd, data, ifp) 1213 u_long cmd; 1214 caddr_t data; 1215 struct ifnet *ifp; 1216 { 1217 struct in6_drlist *drl = (struct in6_drlist *)data; 1218 struct in6_oprlist *oprl = (struct in6_oprlist *)data; 1219 struct in6_ndireq *ndi = (struct in6_ndireq *)data; 1220 struct in6_nbrinfo *nbi = (struct in6_nbrinfo *)data; 1221 struct in6_ndifreq *ndif = (struct in6_ndifreq *)data; 1222 struct nd_defrouter *dr; 1223 struct nd_prefix *pr; 1224 struct rtentry *rt; 1225 int i = 0, error = 0; 1226 int s; 1227 1228 switch (cmd) { 1229 case SIOCGDRLST_IN6: 1230 /* 1231 * obsolete API, use sysctl under net.inet6.icmp6 1232 */ 1233 bzero(drl, sizeof(*drl)); 1234 s = splsoftnet(); 1235 dr = TAILQ_FIRST(&nd_defrouter); 1236 while (dr && i < DRLSTSIZ) { 1237 drl->defrouter[i].rtaddr = dr->rtaddr; 1238 if (IN6_IS_ADDR_LINKLOCAL(&drl->defrouter[i].rtaddr)) { 1239 /* XXX: need to this hack for KAME stack */ 1240 drl->defrouter[i].rtaddr.s6_addr16[1] = 0; 1241 } else 1242 log(LOG_ERR, 1243 "default router list contains a " 1244 "non-linklocal address(%s)\n", 1245 ip6_sprintf(&drl->defrouter[i].rtaddr)); 1246 1247 drl->defrouter[i].flags = dr->flags; 1248 drl->defrouter[i].rtlifetime = dr->rtlifetime; 1249 drl->defrouter[i].expire = dr->expire; 1250 drl->defrouter[i].if_index = dr->ifp->if_index; 1251 i++; 1252 dr = TAILQ_NEXT(dr, dr_entry); 1253 } 1254 splx(s); 1255 break; 1256 case SIOCGPRLST_IN6: 1257 /* 1258 * obsolete API, use sysctl under net.inet6.icmp6 1259 * 1260 * XXX the structure in6_prlist was changed in backward- 1261 * incompatible manner. in6_oprlist is used for SIOCGPRLST_IN6, 1262 * in6_prlist is used for nd6_sysctl() - fill_prlist(). 1263 */ 1264 /* 1265 * XXX meaning of fields, especialy "raflags", is very 1266 * differnet between RA prefix list and RR/static prefix list. 1267 * how about separating ioctls into two? 1268 */ 1269 bzero(oprl, sizeof(*oprl)); 1270 s = splsoftnet(); 1271 pr = nd_prefix.lh_first; 1272 while (pr && i < PRLSTSIZ) { 1273 struct nd_pfxrouter *pfr; 1274 int j; 1275 1276 oprl->prefix[i].prefix = pr->ndpr_prefix.sin6_addr; 1277 oprl->prefix[i].raflags = pr->ndpr_raf; 1278 oprl->prefix[i].prefixlen = pr->ndpr_plen; 1279 oprl->prefix[i].vltime = pr->ndpr_vltime; 1280 oprl->prefix[i].pltime = pr->ndpr_pltime; 1281 oprl->prefix[i].if_index = pr->ndpr_ifp->if_index; 1282 oprl->prefix[i].expire = pr->ndpr_expire; 1283 1284 pfr = pr->ndpr_advrtrs.lh_first; 1285 j = 0; 1286 while (pfr) { 1287 if (j < DRLSTSIZ) { 1288 #define RTRADDR oprl->prefix[i].advrtr[j] 1289 RTRADDR = pfr->router->rtaddr; 1290 if (IN6_IS_ADDR_LINKLOCAL(&RTRADDR)) { 1291 /* XXX: hack for KAME */ 1292 RTRADDR.s6_addr16[1] = 0; 1293 } else 1294 log(LOG_ERR, 1295 "a router(%s) advertises " 1296 "a prefix with " 1297 "non-link local address\n", 1298 ip6_sprintf(&RTRADDR)); 1299 #undef RTRADDR 1300 } 1301 j++; 1302 pfr = pfr->pfr_next; 1303 } 1304 oprl->prefix[i].advrtrs = j; 1305 oprl->prefix[i].origin = PR_ORIG_RA; 1306 1307 i++; 1308 pr = pr->ndpr_next; 1309 } 1310 splx(s); 1311 1312 break; 1313 case OSIOCGIFINFO_IN6: 1314 /* XXX: old ndp(8) assumes a positive value for linkmtu. */ 1315 bzero(&ndi->ndi, sizeof(ndi->ndi)); 1316 ndi->ndi.linkmtu = IN6_LINKMTU(ifp); 1317 ndi->ndi.maxmtu = ND_IFINFO(ifp)->maxmtu; 1318 ndi->ndi.basereachable = ND_IFINFO(ifp)->basereachable; 1319 ndi->ndi.reachable = ND_IFINFO(ifp)->reachable; 1320 ndi->ndi.retrans = ND_IFINFO(ifp)->retrans; 1321 ndi->ndi.flags = ND_IFINFO(ifp)->flags; 1322 ndi->ndi.recalctm = ND_IFINFO(ifp)->recalctm; 1323 ndi->ndi.chlim = ND_IFINFO(ifp)->chlim; 1324 break; 1325 case SIOCGIFINFO_IN6: 1326 ndi->ndi = *ND_IFINFO(ifp); 1327 break; 1328 case SIOCSIFINFO_FLAGS: 1329 ND_IFINFO(ifp)->flags = ndi->ndi.flags; 1330 break; 1331 case SIOCSNDFLUSH_IN6: /* XXX: the ioctl name is confusing... */ 1332 /* sync kernel routing table with the default router list */ 1333 defrouter_reset(); 1334 defrouter_select(); 1335 break; 1336 case SIOCSPFXFLUSH_IN6: 1337 { 1338 /* flush all the prefix advertised by routers */ 1339 struct nd_prefix *pr, *next; 1340 1341 s = splsoftnet(); 1342 for (pr = nd_prefix.lh_first; pr; pr = next) { 1343 struct in6_ifaddr *ia, *ia_next; 1344 1345 next = pr->ndpr_next; 1346 1347 if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr)) 1348 continue; /* XXX */ 1349 1350 /* do we really have to remove addresses as well? */ 1351 for (ia = in6_ifaddr; ia; ia = ia_next) { 1352 /* ia might be removed. keep the next ptr. */ 1353 ia_next = ia->ia_next; 1354 1355 if ((ia->ia6_flags & IN6_IFF_AUTOCONF) == 0) 1356 continue; 1357 1358 if (ia->ia6_ndpr == pr) 1359 in6_purgeaddr(&ia->ia_ifa); 1360 } 1361 prelist_remove(pr); 1362 } 1363 splx(s); 1364 break; 1365 } 1366 case SIOCSRTRFLUSH_IN6: 1367 { 1368 /* flush all the default routers */ 1369 struct nd_defrouter *dr, *next; 1370 1371 s = splsoftnet(); 1372 defrouter_reset(); 1373 for (dr = TAILQ_FIRST(&nd_defrouter); dr; dr = next) { 1374 next = TAILQ_NEXT(dr, dr_entry); 1375 defrtrlist_del(dr); 1376 } 1377 defrouter_select(); 1378 splx(s); 1379 break; 1380 } 1381 case SIOCGNBRINFO_IN6: 1382 { 1383 struct llinfo_nd6 *ln; 1384 struct in6_addr nb_addr = nbi->addr; /* make local for safety */ 1385 1386 /* 1387 * XXX: KAME specific hack for scoped addresses 1388 * XXXX: for other scopes than link-local? 1389 */ 1390 if (IN6_IS_ADDR_LINKLOCAL(&nbi->addr) || 1391 IN6_IS_ADDR_MC_LINKLOCAL(&nbi->addr)) { 1392 u_int16_t *idp = (u_int16_t *)&nb_addr.s6_addr[2]; 1393 1394 if (*idp == 0) 1395 *idp = htons(ifp->if_index); 1396 } 1397 1398 s = splsoftnet(); 1399 if ((rt = nd6_lookup(&nb_addr, 0, ifp)) == NULL || 1400 (ln = (struct llinfo_nd6 *)rt->rt_llinfo) == NULL) { 1401 error = EINVAL; 1402 splx(s); 1403 break; 1404 } 1405 nbi->state = ln->ln_state; 1406 nbi->asked = ln->ln_asked; 1407 nbi->isrouter = ln->ln_router; 1408 nbi->expire = ln->ln_expire; 1409 splx(s); 1410 1411 break; 1412 } 1413 case SIOCGDEFIFACE_IN6: /* XXX: should be implemented as a sysctl? */ 1414 ndif->ifindex = nd6_defifindex; 1415 break; 1416 case SIOCSDEFIFACE_IN6: /* XXX: should be implemented as a sysctl? */ 1417 return (nd6_setdefaultiface(ndif->ifindex)); 1418 } 1419 return (error); 1420 } 1421 1422 /* 1423 * Create neighbor cache entry and cache link-layer address, 1424 * on reception of inbound ND6 packets. (RS/RA/NS/redirect) 1425 */ 1426 struct rtentry * 1427 nd6_cache_lladdr(ifp, from, lladdr, lladdrlen, type, code) 1428 struct ifnet *ifp; 1429 struct in6_addr *from; 1430 char *lladdr; 1431 int lladdrlen; 1432 int type; /* ICMP6 type */ 1433 int code; /* type dependent information */ 1434 { 1435 struct rtentry *rt = NULL; 1436 struct llinfo_nd6 *ln = NULL; 1437 int is_newentry; 1438 struct sockaddr_dl *sdl = NULL; 1439 int do_update; 1440 int olladdr; 1441 int llchange; 1442 int newstate = 0; 1443 long time_second = time.tv_sec; 1444 1445 if (!ifp) 1446 panic("ifp == NULL in nd6_cache_lladdr"); 1447 if (!from) 1448 panic("from == NULL in nd6_cache_lladdr"); 1449 1450 /* nothing must be updated for unspecified address */ 1451 if (IN6_IS_ADDR_UNSPECIFIED(from)) 1452 return NULL; 1453 1454 /* 1455 * Validation about ifp->if_addrlen and lladdrlen must be done in 1456 * the caller. 1457 * 1458 * XXX If the link does not have link-layer adderss, what should 1459 * we do? (ifp->if_addrlen == 0) 1460 * Spec says nothing in sections for RA, RS and NA. There's small 1461 * description on it in NS section (RFC 2461 7.2.3). 1462 */ 1463 1464 rt = nd6_lookup(from, 0, ifp); 1465 if (!rt) { 1466 #if 0 1467 /* nothing must be done if there's no lladdr */ 1468 if (!lladdr || !lladdrlen) 1469 return NULL; 1470 #endif 1471 1472 rt = nd6_lookup(from, 1, ifp); 1473 is_newentry = 1; 1474 } else { 1475 /* do nothing if static ndp is set */ 1476 if (rt->rt_flags & RTF_STATIC) 1477 return NULL; 1478 is_newentry = 0; 1479 } 1480 1481 if (!rt) 1482 return NULL; 1483 if ((rt->rt_flags & (RTF_GATEWAY | RTF_LLINFO)) != RTF_LLINFO) { 1484 fail: 1485 (void)nd6_free(rt, 0); 1486 return NULL; 1487 } 1488 ln = (struct llinfo_nd6 *)rt->rt_llinfo; 1489 if (!ln) 1490 goto fail; 1491 if (!rt->rt_gateway) 1492 goto fail; 1493 if (rt->rt_gateway->sa_family != AF_LINK) 1494 goto fail; 1495 sdl = SDL(rt->rt_gateway); 1496 1497 olladdr = (sdl->sdl_alen) ? 1 : 0; 1498 if (olladdr && lladdr) { 1499 if (bcmp(lladdr, LLADDR(sdl), ifp->if_addrlen)) 1500 llchange = 1; 1501 else 1502 llchange = 0; 1503 } else 1504 llchange = 0; 1505 1506 /* 1507 * newentry olladdr lladdr llchange (*=record) 1508 * 0 n n -- (1) 1509 * 0 y n -- (2) 1510 * 0 n y -- (3) * STALE 1511 * 0 y y n (4) * 1512 * 0 y y y (5) * STALE 1513 * 1 -- n -- (6) NOSTATE(= PASSIVE) 1514 * 1 -- y -- (7) * STALE 1515 */ 1516 1517 if (lladdr) { /* (3-5) and (7) */ 1518 /* 1519 * Record source link-layer address 1520 * XXX is it dependent to ifp->if_type? 1521 */ 1522 sdl->sdl_alen = ifp->if_addrlen; 1523 bcopy(lladdr, LLADDR(sdl), ifp->if_addrlen); 1524 } 1525 1526 if (!is_newentry) { 1527 if ((!olladdr && lladdr) || /* (3) */ 1528 (olladdr && lladdr && llchange)) { /* (5) */ 1529 do_update = 1; 1530 newstate = ND6_LLINFO_STALE; 1531 } else /* (1-2,4) */ 1532 do_update = 0; 1533 } else { 1534 do_update = 1; 1535 if (!lladdr) /* (6) */ 1536 newstate = ND6_LLINFO_NOSTATE; 1537 else /* (7) */ 1538 newstate = ND6_LLINFO_STALE; 1539 } 1540 1541 if (do_update) { 1542 /* 1543 * Update the state of the neighbor cache. 1544 */ 1545 ln->ln_state = newstate; 1546 1547 if (ln->ln_state == ND6_LLINFO_STALE) { 1548 /* 1549 * XXX: since nd6_output() below will cause 1550 * state tansition to DELAY and reset the timer, 1551 * we must set the timer now, although it is actually 1552 * meaningless. 1553 */ 1554 ln->ln_expire = time_second + nd6_gctimer; 1555 1556 if (ln->ln_hold) { 1557 /* 1558 * we assume ifp is not a p2p here, so just 1559 * set the 2nd argument as the 1st one. 1560 */ 1561 nd6_output(ifp, ifp, ln->ln_hold, 1562 (struct sockaddr_in6 *)rt_key(rt), rt); 1563 ln->ln_hold = NULL; 1564 } 1565 } else if (ln->ln_state == ND6_LLINFO_INCOMPLETE) { 1566 /* probe right away */ 1567 ln->ln_expire = time_second; 1568 } 1569 } 1570 1571 /* 1572 * ICMP6 type dependent behavior. 1573 * 1574 * NS: clear IsRouter if new entry 1575 * RS: clear IsRouter 1576 * RA: set IsRouter if there's lladdr 1577 * redir: clear IsRouter if new entry 1578 * 1579 * RA case, (1): 1580 * The spec says that we must set IsRouter in the following cases: 1581 * - If lladdr exist, set IsRouter. This means (1-5). 1582 * - If it is old entry (!newentry), set IsRouter. This means (7). 1583 * So, based on the spec, in (1-5) and (7) cases we must set IsRouter. 1584 * A quetion arises for (1) case. (1) case has no lladdr in the 1585 * neighbor cache, this is similar to (6). 1586 * This case is rare but we figured that we MUST NOT set IsRouter. 1587 * 1588 * newentry olladdr lladdr llchange NS RS RA redir 1589 * D R 1590 * 0 n n -- (1) c ? s 1591 * 0 y n -- (2) c s s 1592 * 0 n y -- (3) c s s 1593 * 0 y y n (4) c s s 1594 * 0 y y y (5) c s s 1595 * 1 -- n -- (6) c c c s 1596 * 1 -- y -- (7) c c s c s 1597 * 1598 * (c=clear s=set) 1599 */ 1600 switch (type & 0xff) { 1601 case ND_NEIGHBOR_SOLICIT: 1602 /* 1603 * New entry must have is_router flag cleared. 1604 */ 1605 if (is_newentry) /* (6-7) */ 1606 ln->ln_router = 0; 1607 break; 1608 case ND_REDIRECT: 1609 /* 1610 * If the icmp is a redirect to a better router, always set the 1611 * is_router flag. Otherwise, if the entry is newly created, 1612 * clear the flag. [RFC 2461, sec 8.3] 1613 */ 1614 if (code == ND_REDIRECT_ROUTER) 1615 ln->ln_router = 1; 1616 else if (is_newentry) /* (6-7) */ 1617 ln->ln_router = 0; 1618 break; 1619 case ND_ROUTER_SOLICIT: 1620 /* 1621 * is_router flag must always be cleared. 1622 */ 1623 ln->ln_router = 0; 1624 break; 1625 case ND_ROUTER_ADVERT: 1626 /* 1627 * Mark an entry with lladdr as a router. 1628 */ 1629 if ((!is_newentry && (olladdr || lladdr)) || /* (2-5) */ 1630 (is_newentry && lladdr)) { /* (7) */ 1631 ln->ln_router = 1; 1632 } 1633 break; 1634 } 1635 1636 /* 1637 * When the link-layer address of a router changes, select the 1638 * best router again. In particular, when the neighbor entry is newly 1639 * created, it might affect the selection policy. 1640 * Question: can we restrict the first condition to the "is_newentry" 1641 * case? 1642 * XXX: when we hear an RA from a new router with the link-layer 1643 * address option, defrouter_select() is called twice, since 1644 * defrtrlist_update called the function as well. However, I believe 1645 * we can compromise the overhead, since it only happens the first 1646 * time. 1647 * XXX: although defrouter_select() should not have a bad effect 1648 * for those are not autoconfigured hosts, we explicitly avoid such 1649 * cases for safety. 1650 */ 1651 if (do_update && ln->ln_router && !ip6_forwarding && ip6_accept_rtadv) 1652 defrouter_select(); 1653 1654 return rt; 1655 } 1656 1657 static void 1658 nd6_slowtimo(ignored_arg) 1659 void *ignored_arg; 1660 { 1661 int s = splsoftnet(); 1662 struct nd_ifinfo *nd6if; 1663 struct ifnet *ifp; 1664 1665 callout_reset(&nd6_slowtimo_ch, ND6_SLOWTIMER_INTERVAL * hz, 1666 nd6_slowtimo, NULL); 1667 for (ifp = TAILQ_FIRST(&ifnet); ifp; ifp = TAILQ_NEXT(ifp, if_list)) 1668 { 1669 nd6if = ND_IFINFO(ifp); 1670 if (nd6if->basereachable && /* already initialized */ 1671 (nd6if->recalctm -= ND6_SLOWTIMER_INTERVAL) <= 0) { 1672 /* 1673 * Since reachable time rarely changes by router 1674 * advertisements, we SHOULD insure that a new random 1675 * value gets recomputed at least once every few hours. 1676 * (RFC 2461, 6.3.4) 1677 */ 1678 nd6if->recalctm = nd6_recalc_reachtm_interval; 1679 nd6if->reachable = ND_COMPUTE_RTIME(nd6if->basereachable); 1680 } 1681 } 1682 splx(s); 1683 } 1684 1685 #define senderr(e) { error = (e); goto bad;} 1686 int 1687 nd6_output(ifp, origifp, m0, dst, rt0) 1688 struct ifnet *ifp; 1689 struct ifnet *origifp; 1690 struct mbuf *m0; 1691 struct sockaddr_in6 *dst; 1692 struct rtentry *rt0; 1693 { 1694 struct mbuf *m = m0; 1695 struct rtentry *rt = rt0; 1696 struct sockaddr_in6 *gw6 = NULL; 1697 struct llinfo_nd6 *ln = NULL; 1698 int error = 0; 1699 long time_second = time.tv_sec; 1700 1701 if (IN6_IS_ADDR_MULTICAST(&dst->sin6_addr)) 1702 goto sendpkt; 1703 1704 if (nd6_need_cache(ifp) == 0) 1705 goto sendpkt; 1706 1707 /* 1708 * next hop determination. This routine is derived from ether_outpout. 1709 */ 1710 if (rt) { 1711 if ((rt->rt_flags & RTF_UP) == 0) { 1712 if ((rt0 = rt = rtalloc1((struct sockaddr *)dst, 1713 1)) != NULL) 1714 { 1715 rt->rt_refcnt--; 1716 if (rt->rt_ifp != ifp) { 1717 /* XXX: loop care? */ 1718 return nd6_output(ifp, origifp, m0, 1719 dst, rt); 1720 } 1721 } else 1722 senderr(EHOSTUNREACH); 1723 } 1724 1725 if (rt->rt_flags & RTF_GATEWAY) { 1726 gw6 = (struct sockaddr_in6 *)rt->rt_gateway; 1727 1728 /* 1729 * We skip link-layer address resolution and NUD 1730 * if the gateway is not a neighbor from ND point 1731 * of view, regardless of the value of nd_ifinfo.flags. 1732 * The second condition is a bit tricky; we skip 1733 * if the gateway is our own address, which is 1734 * sometimes used to install a route to a p2p link. 1735 */ 1736 if (!nd6_is_addr_neighbor(gw6, ifp) || 1737 in6ifa_ifpwithaddr(ifp, &gw6->sin6_addr)) { 1738 /* 1739 * We allow this kind of tricky route only 1740 * when the outgoing interface is p2p. 1741 * XXX: we may need a more generic rule here. 1742 */ 1743 if ((ifp->if_flags & IFF_POINTOPOINT) == 0) 1744 senderr(EHOSTUNREACH); 1745 1746 goto sendpkt; 1747 } 1748 1749 if (rt->rt_gwroute == 0) 1750 goto lookup; 1751 if (((rt = rt->rt_gwroute)->rt_flags & RTF_UP) == 0) { 1752 rtfree(rt); rt = rt0; 1753 lookup: 1754 rt->rt_gwroute = rtalloc1(rt->rt_gateway, 1); 1755 if ((rt = rt->rt_gwroute) == 0) 1756 senderr(EHOSTUNREACH); 1757 /* the "G" test below also prevents rt == rt0 */ 1758 if ((rt->rt_flags & RTF_GATEWAY) || 1759 (rt->rt_ifp != ifp)) { 1760 rt->rt_refcnt--; 1761 rt0->rt_gwroute = 0; 1762 senderr(EHOSTUNREACH); 1763 } 1764 } 1765 } 1766 } 1767 1768 /* 1769 * Address resolution or Neighbor Unreachability Detection 1770 * for the next hop. 1771 * At this point, the destination of the packet must be a unicast 1772 * or an anycast address(i.e. not a multicast). 1773 */ 1774 1775 /* Look up the neighbor cache for the nexthop */ 1776 if (rt && (rt->rt_flags & RTF_LLINFO) != 0) 1777 ln = (struct llinfo_nd6 *)rt->rt_llinfo; 1778 else { 1779 /* 1780 * Since nd6_is_addr_neighbor() internally calls nd6_lookup(), 1781 * the condition below is not very efficient. But we believe 1782 * it is tolerable, because this should be a rare case. 1783 */ 1784 if (nd6_is_addr_neighbor(dst, ifp) && 1785 (rt = nd6_lookup(&dst->sin6_addr, 1, ifp)) != NULL) 1786 ln = (struct llinfo_nd6 *)rt->rt_llinfo; 1787 } 1788 if (!ln || !rt) { 1789 if ((ifp->if_flags & IFF_POINTOPOINT) == 0 && 1790 !(ND_IFINFO(ifp)->flags & ND6_IFF_PERFORMNUD)) { 1791 log(LOG_DEBUG, 1792 "nd6_output: can't allocate llinfo for %s " 1793 "(ln=%p, rt=%p)\n", 1794 ip6_sprintf(&dst->sin6_addr), ln, rt); 1795 senderr(EIO); /* XXX: good error? */ 1796 } 1797 1798 goto sendpkt; /* send anyway */ 1799 } 1800 1801 /* We don't have to do link-layer address resolution on a p2p link. */ 1802 if ((ifp->if_flags & IFF_POINTOPOINT) != 0 && 1803 ln->ln_state < ND6_LLINFO_REACHABLE) { 1804 ln->ln_state = ND6_LLINFO_STALE; 1805 ln->ln_expire = time_second + nd6_gctimer; 1806 } 1807 1808 /* 1809 * The first time we send a packet to a neighbor whose entry is 1810 * STALE, we have to change the state to DELAY and a sets a timer to 1811 * expire in DELAY_FIRST_PROBE_TIME seconds to ensure do 1812 * neighbor unreachability detection on expiration. 1813 * (RFC 2461 7.3.3) 1814 */ 1815 if (ln->ln_state == ND6_LLINFO_STALE) { 1816 ln->ln_asked = 0; 1817 ln->ln_state = ND6_LLINFO_DELAY; 1818 ln->ln_expire = time_second + nd6_delay; 1819 } 1820 1821 /* 1822 * If the neighbor cache entry has a state other than INCOMPLETE 1823 * (i.e. its link-layer address is already resolved), just 1824 * send the packet. 1825 */ 1826 if (ln->ln_state > ND6_LLINFO_INCOMPLETE) 1827 goto sendpkt; 1828 1829 /* 1830 * There is a neighbor cache entry, but no ethernet address 1831 * response yet. Replace the held mbuf (if any) with this 1832 * latest one. 1833 */ 1834 if (ln->ln_state == ND6_LLINFO_NOSTATE) 1835 ln->ln_state = ND6_LLINFO_INCOMPLETE; 1836 if (ln->ln_hold) 1837 m_freem(ln->ln_hold); 1838 ln->ln_hold = m; 1839 /* 1840 * If there has been no NS for the neighbor after entering the 1841 * INCOMPLETE state, send the first solicitation. 1842 * Technically this can be against the rate-limiting rule described in 1843 * Section 7.2.2 of RFC 2461 because the interval to the next scheduled 1844 * solicitation issued in nd6_timer() may be less than the specified 1845 * retransmission time. This should not be a problem from a practical 1846 * point of view, because we'll typically see an immediate response 1847 * from the neighbor, which suppresses the succeeding solicitations. 1848 */ 1849 if (ln->ln_expire && ln->ln_asked == 0) { 1850 ln->ln_asked++; 1851 ln->ln_expire = time_second + 1852 ND6_RETRANS_SEC(ND_IFINFO(ifp)->retrans); 1853 nd6_ns_output(ifp, NULL, &dst->sin6_addr, ln, 0); 1854 } 1855 return (0); 1856 1857 sendpkt: 1858 1859 if ((ifp->if_flags & IFF_LOOPBACK) != 0) { 1860 return ((*ifp->if_output)(origifp, m, (struct sockaddr *)dst, 1861 rt)); 1862 } 1863 return ((*ifp->if_output)(ifp, m, (struct sockaddr *)dst, rt)); 1864 1865 bad: 1866 if (m) 1867 m_freem(m); 1868 return (error); 1869 } 1870 #undef senderr 1871 1872 int 1873 nd6_need_cache(ifp) 1874 struct ifnet *ifp; 1875 { 1876 /* 1877 * XXX: we currently do not make neighbor cache on any interface 1878 * other than ARCnet, Ethernet, FDDI and GIF. 1879 * 1880 * RFC2893 says: 1881 * - unidirectional tunnels needs no ND 1882 */ 1883 switch (ifp->if_type) { 1884 case IFT_ARCNET: 1885 case IFT_ETHER: 1886 case IFT_FDDI: 1887 case IFT_IEEE1394: 1888 case IFT_GIF: /* XXX need more cases? */ 1889 return (1); 1890 default: 1891 return (0); 1892 } 1893 } 1894 1895 int 1896 nd6_storelladdr(ifp, rt, m, dst, desten) 1897 struct ifnet *ifp; 1898 struct rtentry *rt; 1899 struct mbuf *m; 1900 struct sockaddr *dst; 1901 u_char *desten; 1902 { 1903 struct sockaddr_dl *sdl; 1904 1905 if (m->m_flags & M_MCAST) { 1906 switch (ifp->if_type) { 1907 case IFT_ETHER: 1908 case IFT_FDDI: 1909 ETHER_MAP_IPV6_MULTICAST(&SIN6(dst)->sin6_addr, 1910 desten); 1911 return (1); 1912 case IFT_IEEE1394: 1913 bcopy(ifp->if_broadcastaddr, desten, ifp->if_addrlen); 1914 return (1); 1915 case IFT_ARCNET: 1916 *desten = 0; 1917 return (1); 1918 default: 1919 m_freem(m); 1920 return (0); 1921 } 1922 } 1923 1924 if (rt == NULL) { 1925 /* this could happen, if we could not allocate memory */ 1926 m_freem(m); 1927 return (0); 1928 } 1929 if (rt->rt_gateway->sa_family != AF_LINK) { 1930 printf("nd6_storelladdr: something odd happens\n"); 1931 m_freem(m); 1932 return (0); 1933 } 1934 sdl = SDL(rt->rt_gateway); 1935 if (sdl->sdl_alen == 0) { 1936 /* this should be impossible, but we bark here for debugging */ 1937 printf("nd6_storelladdr: sdl_alen == 0, dst=%s, if=%s\n", 1938 ip6_sprintf(&SIN6(dst)->sin6_addr), if_name(ifp)); 1939 m_freem(m); 1940 return (0); 1941 } 1942 1943 bcopy(LLADDR(sdl), desten, sdl->sdl_alen); 1944 return (1); 1945 } 1946 1947 int 1948 nd6_sysctl(name, oldp, oldlenp, newp, newlen) 1949 int name; 1950 void *oldp; /* syscall arg, need copyout */ 1951 size_t *oldlenp; 1952 void *newp; /* syscall arg, need copyin */ 1953 size_t newlen; 1954 { 1955 void *p; 1956 size_t ol, l; 1957 int error; 1958 1959 error = 0; 1960 l = 0; 1961 1962 if (newp) 1963 return EPERM; 1964 if (oldp && !oldlenp) 1965 return EINVAL; 1966 ol = oldlenp ? *oldlenp : 0; 1967 1968 if (oldp) { 1969 p = malloc(*oldlenp, M_TEMP, M_WAITOK); 1970 if (!p) 1971 return ENOMEM; 1972 } else 1973 p = NULL; 1974 switch (name) { 1975 case ICMPV6CTL_ND6_DRLIST: 1976 error = fill_drlist(p, oldlenp, ol); 1977 if (!error && p && oldp) 1978 error = copyout(p, oldp, *oldlenp); 1979 break; 1980 1981 case ICMPV6CTL_ND6_PRLIST: 1982 error = fill_prlist(p, oldlenp, ol); 1983 if (!error && p && oldp) 1984 error = copyout(p, oldp, *oldlenp); 1985 break; 1986 1987 default: 1988 error = ENOPROTOOPT; 1989 break; 1990 } 1991 if (p) 1992 free(p, M_TEMP); 1993 1994 return (error); 1995 } 1996 1997 static int 1998 fill_drlist(oldp, oldlenp, ol) 1999 void *oldp; 2000 size_t *oldlenp, ol; 2001 { 2002 int error = 0, s; 2003 struct in6_defrouter *d = NULL, *de = NULL; 2004 struct nd_defrouter *dr; 2005 size_t l; 2006 2007 s = splsoftnet(); 2008 2009 if (oldp) { 2010 d = (struct in6_defrouter *)oldp; 2011 de = (struct in6_defrouter *)((caddr_t)oldp + *oldlenp); 2012 } 2013 l = 0; 2014 2015 for (dr = TAILQ_FIRST(&nd_defrouter); dr; 2016 dr = TAILQ_NEXT(dr, dr_entry)) { 2017 2018 if (oldp && d + 1 <= de) { 2019 bzero(d, sizeof(*d)); 2020 d->rtaddr.sin6_family = AF_INET6; 2021 d->rtaddr.sin6_len = sizeof(struct sockaddr_in6); 2022 d->rtaddr.sin6_addr = dr->rtaddr; 2023 in6_recoverscope(&d->rtaddr, &d->rtaddr.sin6_addr, 2024 dr->ifp); 2025 d->flags = dr->flags; 2026 d->rtlifetime = dr->rtlifetime; 2027 d->expire = dr->expire; 2028 d->if_index = dr->ifp->if_index; 2029 } 2030 2031 l += sizeof(*d); 2032 if (d) 2033 d++; 2034 } 2035 2036 if (oldp) { 2037 *oldlenp = l; /* (caddr_t)d - (caddr_t)oldp */ 2038 if (l > ol) 2039 error = ENOMEM; 2040 } else 2041 *oldlenp = l; 2042 2043 splx(s); 2044 2045 return (error); 2046 } 2047 2048 static int 2049 fill_prlist(oldp, oldlenp, ol) 2050 void *oldp; 2051 size_t *oldlenp, ol; 2052 { 2053 int error = 0, s; 2054 struct nd_prefix *pr; 2055 struct in6_prefix *p = NULL; 2056 struct in6_prefix *pe = NULL; 2057 size_t l; 2058 2059 s = splsoftnet(); 2060 2061 if (oldp) { 2062 p = (struct in6_prefix *)oldp; 2063 pe = (struct in6_prefix *)((caddr_t)oldp + *oldlenp); 2064 } 2065 l = 0; 2066 2067 for (pr = nd_prefix.lh_first; pr; pr = pr->ndpr_next) { 2068 u_short advrtrs; 2069 size_t advance; 2070 struct sockaddr_in6 *sin6; 2071 struct sockaddr_in6 *s6; 2072 struct nd_pfxrouter *pfr; 2073 2074 if (oldp && p + 1 <= pe) 2075 { 2076 bzero(p, sizeof(*p)); 2077 sin6 = (struct sockaddr_in6 *)(p + 1); 2078 2079 p->prefix = pr->ndpr_prefix; 2080 if (in6_recoverscope(&p->prefix, 2081 &p->prefix.sin6_addr, pr->ndpr_ifp) != 0) 2082 log(LOG_ERR, 2083 "scope error in prefix list (%s)\n", 2084 ip6_sprintf(&p->prefix.sin6_addr)); 2085 p->raflags = pr->ndpr_raf; 2086 p->prefixlen = pr->ndpr_plen; 2087 p->vltime = pr->ndpr_vltime; 2088 p->pltime = pr->ndpr_pltime; 2089 p->if_index = pr->ndpr_ifp->if_index; 2090 if (pr->ndpr_vltime == ND6_INFINITE_LIFETIME) 2091 p->expire = 0; 2092 else { 2093 time_t maxexpire; 2094 2095 /* XXX: we assume time_t is signed. */ 2096 maxexpire = (-1) & 2097 ~(1 << ((sizeof(maxexpire) * 8) - 1)); 2098 if (pr->ndpr_vltime < 2099 maxexpire - pr->ndpr_lastupdate) { 2100 p->expire = pr->ndpr_lastupdate + 2101 pr->ndpr_vltime; 2102 } else 2103 p->expire = maxexpire; 2104 } 2105 p->refcnt = pr->ndpr_refcnt; 2106 p->flags = pr->ndpr_stateflags; 2107 p->origin = PR_ORIG_RA; 2108 advrtrs = 0; 2109 for (pfr = pr->ndpr_advrtrs.lh_first; pfr; 2110 pfr = pfr->pfr_next) { 2111 if ((void *)&sin6[advrtrs + 1] > (void *)pe) { 2112 advrtrs++; 2113 continue; 2114 } 2115 s6 = &sin6[advrtrs]; 2116 s6->sin6_family = AF_INET6; 2117 s6->sin6_len = sizeof(struct sockaddr_in6); 2118 s6->sin6_addr = pfr->router->rtaddr; 2119 in6_recoverscope(s6, &s6->sin6_addr, 2120 pfr->router->ifp); 2121 advrtrs++; 2122 } 2123 p->advrtrs = advrtrs; 2124 } 2125 else { 2126 advrtrs = 0; 2127 for (pfr = pr->ndpr_advrtrs.lh_first; pfr; 2128 pfr = pfr->pfr_next) 2129 advrtrs++; 2130 } 2131 2132 advance = sizeof(*p) + sizeof(*sin6) * advrtrs; 2133 l += advance; 2134 if (p) 2135 p = (struct in6_prefix *)((caddr_t)p + advance); 2136 } 2137 2138 if (oldp) { 2139 *oldlenp = l; /* (caddr_t)d - (caddr_t)oldp */ 2140 if (l > ol) 2141 error = ENOMEM; 2142 } else 2143 *oldlenp = l; 2144 2145 splx(s); 2146 2147 return (error); 2148 } 2149