1 /* $NetBSD: nd6_rtr.c,v 1.147 2019/12/27 10:17:56 msaitoh Exp $ */ 2 /* $KAME: nd6_rtr.c,v 1.95 2001/02/07 08:09:47 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_rtr.c,v 1.147 2019/12/27 10:17:56 msaitoh Exp $"); 35 36 #ifdef _KERNEL_OPT 37 #include "opt_net_mpsafe.h" 38 #endif 39 40 #include <sys/param.h> 41 #include <sys/systm.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/errno.h> 49 #include <sys/ioctl.h> 50 #include <sys/syslog.h> 51 #include <sys/cprng.h> 52 53 #include <net/if.h> 54 #include <net/if_types.h> 55 #include <net/if_dl.h> 56 57 #include <netinet/in.h> 58 #include <netinet6/in6_var.h> 59 #include <netinet6/in6_ifattach.h> 60 #include <netinet/ip6.h> 61 #include <netinet6/ip6_var.h> 62 #include <netinet6/nd6.h> 63 #include <netinet/icmp6.h> 64 #include <netinet6/icmp6_private.h> 65 #include <netinet6/scope6_var.h> 66 67 static int rtpref(struct nd_defrouter *); 68 static struct nd_defrouter *defrtrlist_update(struct nd_defrouter *); 69 static int prelist_update(struct nd_prefixctl *, struct nd_defrouter *, 70 struct mbuf *, int); 71 static struct in6_ifaddr *in6_ifadd(struct nd_prefixctl *, int, struct psref *); 72 static struct nd_pfxrouter *pfxrtr_lookup(struct nd_prefix *, 73 struct nd_defrouter *); 74 static void pfxrtr_add(struct nd_prefix *, struct nd_defrouter *); 75 static void pfxrtr_del(struct nd_pfxrouter *); 76 static struct nd_pfxrouter *find_pfxlist_reachable_router 77 (struct nd_prefix *); 78 79 static void defrouter_addreq(struct nd_defrouter *); 80 static void defrouter_delreq(struct nd_defrouter *); 81 82 static int in6_init_prefix_ltimes(struct nd_prefix *); 83 static void in6_init_address_ltimes(struct nd_prefix *, 84 struct in6_addrlifetime *); 85 static void purge_detached(struct ifnet *); 86 87 static int rt6_deleteroute_matcher(struct rtentry *, void *); 88 89 static int nd6_prelist_add(struct nd_prefixctl *, struct nd_defrouter *, 90 struct nd_prefix **); 91 static int nd6_prefix_onlink(struct nd_prefix *); 92 static int nd6_prefix_offlink(struct nd_prefix *); 93 static struct nd_prefix *nd6_prefix_lookup(struct nd_prefixctl *); 94 95 extern int nd6_recalc_reachtm_interval; 96 97 int ip6_use_tempaddr = 0; 98 99 int ip6_desync_factor; 100 u_int32_t ip6_temp_preferred_lifetime = DEF_TEMP_PREFERRED_LIFETIME; 101 u_int32_t ip6_temp_valid_lifetime = DEF_TEMP_VALID_LIFETIME; 102 int ip6_temp_regen_advance = TEMPADDR_REGEN_ADVANCE; 103 104 int nd6_numroutes = 0; 105 106 /* RTPREF_MEDIUM has to be 0! */ 107 #define RTPREF_HIGH 1 108 #define RTPREF_MEDIUM 0 109 #define RTPREF_LOW (-1) 110 #define RTPREF_RESERVED (-2) 111 #define RTPREF_INVALID (-3) /* internal */ 112 113 static inline bool 114 nd6_is_llinfo_probreach(struct nd_defrouter *dr) 115 { 116 struct llentry *ln = NULL; 117 118 ln = nd6_lookup(&dr->rtaddr, dr->ifp, false); 119 if (ln == NULL) 120 return false; 121 122 if (!ND6_IS_LLINFO_PROBREACH(ln)) { 123 LLE_RUNLOCK(ln); 124 return false; 125 } 126 127 LLE_RUNLOCK(ln); 128 return true; 129 } 130 131 /* 132 * Receive Router Solicitation Message - just for routers. 133 * Router solicitation/advertisement is mostly managed by a userland program 134 * (rtadvd) so here we have no function like nd6_ra_output(). 135 * 136 * Based on RFC 2461 137 */ 138 void 139 nd6_rs_input(struct mbuf *m, int off, int icmp6len) 140 { 141 struct ifnet *ifp; 142 struct nd_ifinfo *ndi; 143 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *); 144 struct nd_router_solicit *nd_rs; 145 struct in6_addr saddr6 = ip6->ip6_src; 146 char *lladdr = NULL; 147 int lladdrlen = 0; 148 union nd_opts ndopts; 149 struct psref psref; 150 char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN]; 151 152 ifp = m_get_rcvif_psref(m, &psref); 153 if (ifp == NULL) 154 goto freeit; 155 156 ndi = ND_IFINFO(ifp); 157 158 /* If I'm not a router, ignore it. */ 159 if (nd6_accepts_rtadv(ndi) || !ip6_forwarding) 160 goto freeit; 161 162 /* Sanity checks */ 163 if (ip6->ip6_hlim != 255) { 164 nd6log(LOG_ERR, "invalid hlim (%d) from %s to %s on %s\n", 165 ip6->ip6_hlim, IN6_PRINT(ip6bufs, &ip6->ip6_src), 166 IN6_PRINT(ip6bufd, &ip6->ip6_dst), if_name(ifp)); 167 goto bad; 168 } 169 170 /* 171 * Don't update the neighbor cache, if src = ::. 172 * This indicates that the src has no IP address assigned yet. 173 */ 174 if (IN6_IS_ADDR_UNSPECIFIED(&saddr6)) 175 goto freeit; 176 177 IP6_EXTHDR_GET(nd_rs, struct nd_router_solicit *, m, off, icmp6len); 178 if (nd_rs == NULL) { 179 ICMP6_STATINC(ICMP6_STAT_TOOSHORT); 180 m_put_rcvif_psref(ifp, &psref); 181 return; 182 } 183 184 icmp6len -= sizeof(*nd_rs); 185 nd6_option_init(nd_rs + 1, icmp6len, &ndopts); 186 if (nd6_options(&ndopts) < 0) { 187 nd6log(LOG_INFO, "invalid ND option, ignored\n"); 188 /* nd6_options have incremented stats */ 189 goto freeit; 190 } 191 192 if (ndopts.nd_opts_src_lladdr) { 193 lladdr = (char *)(ndopts.nd_opts_src_lladdr + 1); 194 lladdrlen = ndopts.nd_opts_src_lladdr->nd_opt_len << 3; 195 } 196 197 if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) { 198 nd6log(LOG_INFO, "lladdrlen mismatch for %s " 199 "(if %d, RS packet %d)\n", 200 IN6_PRINT(ip6bufs, &saddr6), 201 ifp->if_addrlen, lladdrlen - 2); 202 goto bad; 203 } 204 205 nd6_cache_lladdr(ifp, &saddr6, lladdr, lladdrlen, ND_ROUTER_SOLICIT, 0); 206 207 freeit: 208 m_put_rcvif_psref(ifp, &psref); 209 m_freem(m); 210 return; 211 212 bad: 213 ICMP6_STATINC(ICMP6_STAT_BADRS); 214 m_put_rcvif_psref(ifp, &psref); 215 m_freem(m); 216 } 217 218 /* 219 * Receive Router Advertisement Message. 220 * 221 * Based on RFC 2461 222 * TODO: on-link bit on prefix information 223 * TODO: ND_RA_FLAG_{OTHER,MANAGED} processing 224 */ 225 void 226 nd6_ra_input(struct mbuf *m, int off, int icmp6len) 227 { 228 struct ifnet *ifp; 229 struct nd_ifinfo *ndi; 230 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *); 231 struct nd_router_advert *nd_ra; 232 struct in6_addr saddr6 = ip6->ip6_src; 233 int mcast = 0; 234 union nd_opts ndopts; 235 struct nd_defrouter *dr; 236 struct psref psref; 237 char ip6buf[INET6_ADDRSTRLEN], ip6buf2[INET6_ADDRSTRLEN]; 238 239 ifp = m_get_rcvif_psref(m, &psref); 240 if (ifp == NULL) 241 goto freeit; 242 243 ndi = ND_IFINFO(ifp); 244 /* 245 * We only accept RAs when the system-wide variable allows the 246 * acceptance, and the per-interface variable allows RAs on the 247 * receiving interface. 248 */ 249 if (!nd6_accepts_rtadv(ndi)) 250 goto freeit; 251 252 if (ip6->ip6_hlim != 255) { 253 nd6log(LOG_ERR, "invalid hlim (%d) from %s to %s on %s\n", 254 ip6->ip6_hlim, IN6_PRINT(ip6buf, &ip6->ip6_src), 255 IN6_PRINT(ip6buf2, &ip6->ip6_dst), if_name(ifp)); 256 goto bad; 257 } 258 259 if (!IN6_IS_ADDR_LINKLOCAL(&saddr6)) { 260 nd6log(LOG_ERR, "src %s is not link-local\n", 261 IN6_PRINT(ip6buf, &saddr6)); 262 goto bad; 263 } 264 265 IP6_EXTHDR_GET(nd_ra, struct nd_router_advert *, m, off, icmp6len); 266 if (nd_ra == NULL) { 267 ICMP6_STATINC(ICMP6_STAT_TOOSHORT); 268 m_put_rcvif_psref(ifp, &psref); 269 return; 270 } 271 272 icmp6len -= sizeof(*nd_ra); 273 nd6_option_init(nd_ra + 1, icmp6len, &ndopts); 274 if (nd6_options(&ndopts) < 0) { 275 nd6log(LOG_INFO, "invalid ND option, ignored\n"); 276 /* nd6_options have incremented stats */ 277 goto freeit; 278 } 279 280 { 281 struct nd_defrouter drtr; 282 u_int32_t advreachable = nd_ra->nd_ra_reachable; 283 284 /* remember if this is a multicasted advertisement */ 285 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) 286 mcast = 1; 287 288 memset(&drtr, 0, sizeof(drtr)); 289 drtr.rtaddr = saddr6; 290 drtr.flags = nd_ra->nd_ra_flags_reserved; 291 drtr.rtlifetime = ntohs(nd_ra->nd_ra_router_lifetime); 292 drtr.expire = time_uptime + drtr.rtlifetime; 293 drtr.ifp = ifp; 294 /* unspecified or not? (RFC 2461 6.3.4) */ 295 if (advreachable) { 296 NTOHL(advreachable); 297 if (advreachable <= MAX_REACHABLE_TIME && 298 ndi->basereachable != advreachable) { 299 ndi->basereachable = advreachable; 300 ndi->reachable = ND_COMPUTE_RTIME(ndi->basereachable); 301 ndi->recalctm = nd6_recalc_reachtm_interval; /* reset */ 302 } 303 } 304 if (nd_ra->nd_ra_retransmit) 305 ndi->retrans = ntohl(nd_ra->nd_ra_retransmit); 306 if (nd_ra->nd_ra_curhoplimit) { 307 if (ndi->chlim < nd_ra->nd_ra_curhoplimit) 308 ndi->chlim = nd_ra->nd_ra_curhoplimit; 309 else if (ndi->chlim != nd_ra->nd_ra_curhoplimit) 310 log(LOG_ERR, "nd_ra_input: lower CurHopLimit sent from " 311 "%s on %s (current=%d, received=%d), ignored\n", 312 IN6_PRINT(ip6buf, &ip6->ip6_src), 313 if_name(ifp), ndi->chlim, nd_ra->nd_ra_curhoplimit); 314 } 315 IFNET_LOCK(ifp); 316 ND6_WLOCK(); 317 dr = defrtrlist_update(&drtr); 318 } 319 320 /* 321 * prefix 322 */ 323 if (ndopts.nd_opts_pi) { 324 struct nd_opt_hdr *pt; 325 struct nd_opt_prefix_info *pi = NULL; 326 struct nd_prefixctl prc; 327 328 for (pt = (struct nd_opt_hdr *)ndopts.nd_opts_pi; 329 pt <= (struct nd_opt_hdr *)ndopts.nd_opts_pi_end; 330 pt = (struct nd_opt_hdr *)((char *)pt + 331 (pt->nd_opt_len << 3))) { 332 if (pt->nd_opt_type != ND_OPT_PREFIX_INFORMATION) 333 continue; 334 pi = (struct nd_opt_prefix_info *)pt; 335 336 if (pi->nd_opt_pi_len != 4) { 337 nd6log(LOG_INFO, "invalid option " 338 "len %d for prefix information option, " 339 "ignored\n", pi->nd_opt_pi_len); 340 continue; 341 } 342 343 if (128 < pi->nd_opt_pi_prefix_len) { 344 nd6log(LOG_INFO, "invalid prefix " 345 "len %d for prefix information option, " 346 "ignored\n", pi->nd_opt_pi_prefix_len); 347 continue; 348 } 349 350 if (IN6_IS_ADDR_MULTICAST(&pi->nd_opt_pi_prefix) 351 || IN6_IS_ADDR_LINKLOCAL(&pi->nd_opt_pi_prefix)) { 352 nd6log(LOG_INFO, 353 "invalid prefix %s, ignored\n", 354 IN6_PRINT(ip6buf, &pi->nd_opt_pi_prefix)); 355 continue; 356 } 357 358 memset(&prc, 0, sizeof(prc)); 359 sockaddr_in6_init(&prc.ndprc_prefix, 360 &pi->nd_opt_pi_prefix, 0, 0, 0); 361 prc.ndprc_ifp = ifp; 362 363 prc.ndprc_raf_onlink = (pi->nd_opt_pi_flags_reserved & 364 ND_OPT_PI_FLAG_ONLINK) ? 1 : 0; 365 prc.ndprc_raf_auto = (pi->nd_opt_pi_flags_reserved & 366 ND_OPT_PI_FLAG_AUTO) ? 1 : 0; 367 prc.ndprc_plen = pi->nd_opt_pi_prefix_len; 368 prc.ndprc_vltime = ntohl(pi->nd_opt_pi_valid_time); 369 prc.ndprc_pltime = ntohl(pi->nd_opt_pi_preferred_time); 370 371 (void)prelist_update(&prc, dr, m, mcast); 372 } 373 } 374 ND6_UNLOCK(); 375 IFNET_UNLOCK(ifp); 376 377 /* 378 * MTU 379 */ 380 if (ndopts.nd_opts_mtu && ndopts.nd_opts_mtu->nd_opt_mtu_len == 1) { 381 u_long mtu; 382 u_long maxmtu; 383 384 mtu = ntohl(ndopts.nd_opts_mtu->nd_opt_mtu_mtu); 385 386 /* lower bound */ 387 if (mtu < IPV6_MMTU) { 388 nd6log(LOG_INFO, "bogus mtu option " 389 "mtu=%lu sent from %s, ignoring\n", 390 mtu, IN6_PRINT(ip6buf, &ip6->ip6_src)); 391 goto skip; 392 } 393 394 /* upper bound */ 395 maxmtu = (ndi->maxmtu && ndi->maxmtu < ifp->if_mtu) 396 ? ndi->maxmtu : ifp->if_mtu; 397 if (mtu <= maxmtu) { 398 int change = (ndi->linkmtu != mtu); 399 400 ndi->linkmtu = mtu; 401 if (change) /* in6_maxmtu may change */ 402 in6_setmaxmtu(); 403 } else { 404 nd6log(LOG_INFO, 405 "bogus mtu mtu=%lu sent from %s; " 406 "exceeds maxmtu %lu, ignoring\n", 407 mtu, IN6_PRINT(ip6buf, &ip6->ip6_src), maxmtu); 408 } 409 } 410 411 skip: 412 413 /* 414 * Source link layer address 415 */ 416 { 417 char *lladdr = NULL; 418 int lladdrlen = 0; 419 420 if (ndopts.nd_opts_src_lladdr) { 421 lladdr = (char *)(ndopts.nd_opts_src_lladdr + 1); 422 lladdrlen = ndopts.nd_opts_src_lladdr->nd_opt_len << 3; 423 } 424 425 if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) { 426 nd6log(LOG_INFO, "lladdrlen mismatch for %s " 427 "(if %d, RA packet %d)\n", IN6_PRINT(ip6buf, &saddr6), 428 ifp->if_addrlen, lladdrlen - 2); 429 goto bad; 430 } 431 432 nd6_cache_lladdr(ifp, &saddr6, lladdr, lladdrlen, ND_ROUTER_ADVERT, 0); 433 434 /* 435 * Installing a link-layer address might change the state of the 436 * router's neighbor cache, which might also affect our on-link 437 * detection of adveritsed prefixes. 438 */ 439 ND6_WLOCK(); 440 nd6_pfxlist_onlink_check(); 441 ND6_UNLOCK(); 442 } 443 444 freeit: 445 m_put_rcvif_psref(ifp, &psref); 446 m_freem(m); 447 return; 448 449 bad: 450 ICMP6_STATINC(ICMP6_STAT_BADRA); 451 m_put_rcvif_psref(ifp, &psref); 452 m_freem(m); 453 } 454 455 /* 456 * default router list processing sub routines 457 */ 458 static void 459 defrouter_addreq(struct nd_defrouter *newdr) 460 { 461 union { 462 struct sockaddr_in6 sin6; 463 struct sockaddr sa; 464 } def, mask, gate; 465 #ifndef NET_MPSAFE 466 int s; 467 #endif 468 int error; 469 470 memset(&def, 0, sizeof(def)); 471 memset(&mask, 0, sizeof(mask)); 472 memset(&gate, 0,sizeof(gate)); /* for safety */ 473 474 def.sin6.sin6_len = mask.sin6.sin6_len = gate.sin6.sin6_len = 475 sizeof(struct sockaddr_in6); 476 def.sin6.sin6_family = mask.sin6.sin6_family = gate.sin6.sin6_family = AF_INET6; 477 gate.sin6.sin6_addr = newdr->rtaddr; 478 #ifndef SCOPEDROUTING 479 gate.sin6.sin6_scope_id = 0; /* XXX */ 480 #endif 481 482 #ifndef NET_MPSAFE 483 s = splsoftnet(); 484 #endif 485 error = rtrequest_newmsg(RTM_ADD, &def.sa, &gate.sa, &mask.sa, 486 RTF_GATEWAY); 487 if (error == 0) { 488 nd6_numroutes++; 489 newdr->installed = 1; 490 } 491 #ifndef NET_MPSAFE 492 splx(s); 493 #endif 494 return; 495 } 496 497 struct nd_defrouter * 498 nd6_defrouter_lookup(const struct in6_addr *addr, struct ifnet *ifp) 499 { 500 struct nd_defrouter *dr; 501 502 ND6_ASSERT_LOCK(); 503 504 ND_DEFROUTER_LIST_FOREACH(dr) { 505 if (dr->ifp == ifp && IN6_ARE_ADDR_EQUAL(addr, &dr->rtaddr)) 506 break; 507 } 508 509 return dr; /* search failed */ 510 } 511 512 void 513 nd6_defrtrlist_del(struct nd_defrouter *dr, struct in6_ifextra *ext) 514 { 515 struct nd_defrouter *deldr = NULL; 516 struct nd_prefix *pr; 517 struct nd_ifinfo *ndi; 518 519 ND6_ASSERT_WLOCK(); 520 521 if (ext == NULL) 522 ext = dr->ifp->if_afdata[AF_INET6]; 523 524 /* detach already in progress, can not do anything */ 525 if (ext == NULL) 526 return; 527 528 ndi = ext->nd_ifinfo; 529 530 /* 531 * Flush all the routing table entries that use the router 532 * as a next hop. 533 */ 534 /* XXX: better condition? */ 535 if (!ip6_forwarding && nd6_accepts_rtadv(ndi)) 536 nd6_rt_flush(&dr->rtaddr, dr->ifp); 537 538 if (dr->installed) { 539 deldr = dr; 540 defrouter_delreq(dr); 541 } 542 ND_DEFROUTER_LIST_REMOVE(dr); 543 544 /* 545 * Also delete all the pointers to the router in each prefix lists. 546 */ 547 ND_PREFIX_LIST_FOREACH(pr) { 548 struct nd_pfxrouter *pfxrtr; 549 if ((pfxrtr = pfxrtr_lookup(pr, dr)) != NULL) 550 pfxrtr_del(pfxrtr); 551 } 552 nd6_pfxlist_onlink_check(); 553 554 /* 555 * If the router is the primary one, choose a new one. 556 * Note that nd6_defrouter_select() will remove the current gateway 557 * from the routing table. 558 */ 559 if (deldr) 560 nd6_defrouter_select(); 561 562 ext->ndefrouters--; 563 if (ext->ndefrouters < 0) { 564 log(LOG_WARNING, "nd6_defrtrlist_del: negative count on %s\n", 565 dr->ifp->if_xname); 566 } 567 568 free(dr, M_IP6NDP); 569 } 570 571 /* 572 * Remove the default route for a given router. 573 * This is just a subroutine function for nd6_defrouter_select(), and should 574 * not be called from anywhere else. 575 */ 576 static void 577 defrouter_delreq(struct nd_defrouter *dr) 578 { 579 union { 580 struct sockaddr_in6 sin6; 581 struct sockaddr sa; 582 } def, mask, gw; 583 int error; 584 585 memset(&def, 0, sizeof(def)); 586 memset(&mask, 0, sizeof(mask)); 587 memset(&gw, 0, sizeof(gw)); /* for safety */ 588 589 def.sin6.sin6_len = mask.sin6.sin6_len = gw.sin6.sin6_len = 590 sizeof(struct sockaddr_in6); 591 def.sin6.sin6_family = mask.sin6.sin6_family = gw.sin6.sin6_family = AF_INET6; 592 gw.sin6.sin6_addr = dr->rtaddr; 593 #ifndef SCOPEDROUTING 594 gw.sin6.sin6_scope_id = 0; /* XXX */ 595 #endif 596 597 error = rtrequest_newmsg(RTM_DELETE, &def.sa, &gw.sa, &mask.sa, 598 RTF_GATEWAY); 599 if (error == 0) 600 nd6_numroutes--; 601 602 dr->installed = 0; 603 } 604 605 /* 606 * remove all default routes from default router list 607 */ 608 void 609 nd6_defrouter_reset(void) 610 { 611 struct nd_defrouter *dr; 612 613 ND6_ASSERT_WLOCK(); 614 615 ND_DEFROUTER_LIST_FOREACH(dr) 616 defrouter_delreq(dr); 617 618 /* 619 * XXX should we also nuke any default routers in the kernel, by 620 * going through them by rtalloc1()? 621 */ 622 } 623 624 /* 625 * Default Router Selection according to Section 6.3.6 of RFC 2461 and 626 * draft-ietf-ipngwg-router-selection: 627 * 1) Routers that are reachable or probably reachable should be preferred. 628 * If we have more than one (probably) reachable router, prefer ones 629 * with the highest router preference. 630 * 2) When no routers on the list are known to be reachable or 631 * probably reachable, routers SHOULD be selected in a round-robin 632 * fashion, regardless of router preference values. 633 * 3) If the Default Router List is empty, assume that all 634 * destinations are on-link. 635 * 636 * We assume nd_defrouter is sorted by router preference value. 637 * Since the code below covers both with and without router preference cases, 638 * we do not need to classify the cases by ifdef. 639 * 640 * At this moment, we do not try to install more than one default router, 641 * even when the multipath routing is available, because we're not sure about 642 * the benefits for stub hosts comparing to the risk of making the code 643 * complicated and the possibility of introducing bugs. 644 */ 645 void 646 nd6_defrouter_select(void) 647 { 648 struct nd_ifinfo *ndi; 649 struct nd_defrouter *dr, *selected_dr = NULL, *installed_dr = NULL; 650 651 ND6_ASSERT_WLOCK(); 652 653 /* 654 * This function should be called only when acting as an autoconfigured 655 * host. Although the remaining part of this function is not effective 656 * if the node is not an autoconfigured host, we explicitly exclude 657 * such cases here for safety. 658 */ 659 if (ip6_forwarding) { 660 nd6log(LOG_WARNING, "called unexpectedly (forwarding=%d, " 661 "accept_rtadv=%d)\n", ip6_forwarding, ip6_accept_rtadv); 662 return; 663 } 664 665 /* 666 * Let's handle easy case (3) first: 667 * If default router list is empty, there's nothing to be done. 668 */ 669 if (ND_DEFROUTER_LIST_EMPTY()) 670 return; 671 672 /* 673 * Search for a (probably) reachable router from the list. 674 * We just pick up the first reachable one (if any), assuming that 675 * the ordering rule of the list described in defrtrlist_update(). 676 */ 677 ND_DEFROUTER_LIST_FOREACH(dr) { 678 ndi = ND_IFINFO(dr->ifp); 679 if (nd6_accepts_rtadv(ndi)) 680 continue; 681 682 if (selected_dr == NULL && 683 nd6_is_llinfo_probreach(dr)) 684 selected_dr = dr; 685 686 if (dr->installed && !installed_dr) 687 installed_dr = dr; 688 else if (dr->installed && installed_dr) { 689 /* this should not happen. warn for diagnosis. */ 690 log(LOG_ERR, "nd6_defrouter_select: more than one router" 691 " is installed\n"); 692 } 693 } 694 /* 695 * If none of the default routers was found to be reachable, 696 * round-robin the list regardless of preference. 697 * Otherwise, if we have an installed router, check if the selected 698 * (reachable) router should really be preferred to the installed one. 699 * We only prefer the new router when the old one is not reachable 700 * or when the new one has a really higher preference value. 701 */ 702 if (selected_dr == NULL) { 703 if (installed_dr == NULL || 704 ND_DEFROUTER_LIST_NEXT(installed_dr) == NULL) 705 selected_dr = ND_DEFROUTER_LIST_FIRST(); 706 else 707 selected_dr = ND_DEFROUTER_LIST_NEXT(installed_dr); 708 } else if (installed_dr && 709 nd6_is_llinfo_probreach(installed_dr) && 710 rtpref(selected_dr) <= rtpref(installed_dr)) { 711 selected_dr = installed_dr; 712 } 713 714 /* 715 * If the selected router is different than the installed one, 716 * remove the installed router and install the selected one. 717 * Note that the selected router is never NULL here. 718 */ 719 if (installed_dr != selected_dr) { 720 if (installed_dr) 721 defrouter_delreq(installed_dr); 722 defrouter_addreq(selected_dr); 723 } 724 725 return; 726 } 727 728 /* 729 * for default router selection 730 * regards router-preference field as a 2-bit signed integer 731 */ 732 static int 733 rtpref(struct nd_defrouter *dr) 734 { 735 switch (dr->flags & ND_RA_FLAG_RTPREF_MASK) { 736 case ND_RA_FLAG_RTPREF_HIGH: 737 return (RTPREF_HIGH); 738 case ND_RA_FLAG_RTPREF_MEDIUM: 739 case ND_RA_FLAG_RTPREF_RSV: 740 return (RTPREF_MEDIUM); 741 case ND_RA_FLAG_RTPREF_LOW: 742 return (RTPREF_LOW); 743 default: 744 /* 745 * This case should never happen. If it did, it would mean a 746 * serious bug of kernel internal. We thus always bark here. 747 * Or, can we even panic? 748 */ 749 log(LOG_ERR, "rtpref: impossible RA flag %x\n", dr->flags); 750 return (RTPREF_INVALID); 751 } 752 /* NOTREACHED */ 753 } 754 755 static struct nd_defrouter * 756 defrtrlist_update(struct nd_defrouter *newdr) 757 { 758 struct nd_defrouter *dr, *n, *ret = NULL; 759 struct in6_ifextra *ext = newdr->ifp->if_afdata[AF_INET6]; 760 761 ND6_ASSERT_WLOCK(); 762 763 if ((dr = nd6_defrouter_lookup(&newdr->rtaddr, newdr->ifp)) != NULL) { 764 /* entry exists */ 765 if (newdr->rtlifetime == 0) { 766 nd6_defrtrlist_del(dr, ext); 767 dr = NULL; 768 } else { 769 int oldpref = rtpref(dr); 770 771 /* override */ 772 dr->flags = newdr->flags; /* xxx flag check */ 773 dr->rtlifetime = newdr->rtlifetime; 774 dr->expire = newdr->expire; 775 776 /* 777 * If the preference does not change, there's no need 778 * to sort the entries. 779 */ 780 if (rtpref(newdr) == oldpref) { 781 ret = dr; 782 goto out; 783 } 784 785 /* 786 * preferred router may be changed, so relocate 787 * this router. 788 * XXX: calling TAILQ_REMOVE directly is a bad manner. 789 * However, since nd6_defrtrlist_del() has many side 790 * effects, we intentionally do so here. 791 * nd6_defrouter_select() below will handle routing 792 * changes later. 793 */ 794 ND_DEFROUTER_LIST_REMOVE(dr); 795 n = dr; 796 goto insert; 797 } 798 ret = dr; 799 goto out; 800 } 801 802 if (ip6_maxifdefrouters >= 0 && ext->ndefrouters >= ip6_maxifdefrouters) 803 goto out; 804 805 /* entry does not exist */ 806 if (newdr->rtlifetime == 0) 807 goto out; 808 809 if (ip6_rtadv_maxroutes <= nd6_numroutes) { 810 ICMP6_STATINC(ICMP6_STAT_DROPPED_RAROUTE); 811 goto out; 812 } 813 814 n = (struct nd_defrouter *)malloc(sizeof(*n), M_IP6NDP, M_NOWAIT); 815 if (n == NULL) 816 goto out; 817 memset(n, 0, sizeof(*n)); 818 *n = *newdr; 819 820 insert: 821 /* 822 * Insert the new router in the Default Router List; 823 * The Default Router List should be in the descending order 824 * of router-preference. Routers with the same preference are 825 * sorted in the arriving time order. 826 */ 827 828 /* insert at the end of the group */ 829 ND_DEFROUTER_LIST_FOREACH(dr) { 830 if (rtpref(n) > rtpref(dr)) 831 break; 832 } 833 if (dr) 834 ND_DEFROUTER_LIST_INSERT_BEFORE(dr, n); 835 else 836 ND_DEFROUTER_LIST_INSERT_TAIL(n); 837 838 nd6_defrouter_select(); 839 840 ext->ndefrouters++; 841 842 ret = n; 843 out: 844 return ret; 845 } 846 847 static struct nd_pfxrouter * 848 pfxrtr_lookup(struct nd_prefix *pr, struct nd_defrouter *dr) 849 { 850 struct nd_pfxrouter *search; 851 852 ND6_ASSERT_LOCK(); 853 854 LIST_FOREACH(search, &pr->ndpr_advrtrs, pfr_entry) { 855 if (search->router == dr) 856 break; 857 } 858 859 return (search); 860 } 861 862 static void 863 pfxrtr_add(struct nd_prefix *pr, struct nd_defrouter *dr) 864 { 865 struct nd_pfxrouter *newpfr; 866 867 ND6_ASSERT_WLOCK(); 868 869 newpfr = malloc(sizeof(*newpfr), M_IP6NDP, M_NOWAIT|M_ZERO); 870 if (newpfr == NULL) 871 return; 872 newpfr->router = dr; 873 874 LIST_INSERT_HEAD(&pr->ndpr_advrtrs, newpfr, pfr_entry); 875 876 nd6_pfxlist_onlink_check(); 877 } 878 879 static void 880 pfxrtr_del(struct nd_pfxrouter *pfr) 881 { 882 LIST_REMOVE(pfr, pfr_entry); 883 free(pfr, M_IP6NDP); 884 } 885 886 static struct nd_prefix * 887 nd6_prefix_lookup(struct nd_prefixctl *key) 888 { 889 struct nd_prefix *search; 890 891 ND_PREFIX_LIST_FOREACH(search) { 892 if (key->ndprc_ifp == search->ndpr_ifp && 893 key->ndprc_plen == search->ndpr_plen && 894 in6_are_prefix_equal(&key->ndprc_prefix.sin6_addr, 895 &search->ndpr_prefix.sin6_addr, key->ndprc_plen)) { 896 break; 897 } 898 } 899 900 return (search); 901 } 902 903 static void 904 purge_detached(struct ifnet *ifp) 905 { 906 struct nd_prefix *pr, *pr_next; 907 struct in6_ifaddr *ia; 908 struct ifaddr *ifa, *ifa_next; 909 910 restart: 911 ND6_ASSERT_WLOCK(); 912 913 ND_PREFIX_LIST_FOREACH_SAFE(pr, pr_next) { 914 int s; 915 916 /* 917 * This function is called when we need to make more room for 918 * new prefixes rather than keeping old, possibly stale ones. 919 * Detached prefixes would be a good candidate; if all routers 920 * that advertised the prefix expired, the prefix is also 921 * probably stale. 922 */ 923 if (pr->ndpr_ifp != ifp || 924 IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr) || 925 ((pr->ndpr_stateflags & NDPRF_DETACHED) == 0 && 926 !LIST_EMPTY(&pr->ndpr_advrtrs))) 927 continue; 928 929 s = pserialize_read_enter(); 930 for (ifa = IFADDR_READER_FIRST(ifp); ifa; ifa = ifa_next) { 931 ifa_next = IFADDR_READER_NEXT(ifa); 932 if (ifa->ifa_addr->sa_family != AF_INET6) 933 continue; 934 ia = (struct in6_ifaddr *)ifa; 935 if ((ia->ia6_flags & IN6_IFF_AUTOCONF) == 936 IN6_IFF_AUTOCONF && ia->ia6_ndpr == pr) { 937 pserialize_read_exit(s); 938 ND6_UNLOCK(); 939 940 /* in6_purgeaddr may destroy pr. */ 941 in6_purgeaddr(ifa); 942 943 ND6_WLOCK(); 944 goto restart; 945 } 946 } 947 pserialize_read_exit(s); 948 949 KASSERTMSG(pr->ndpr_refcnt == 0, "ndpr_refcnt=%d", 950 pr->ndpr_refcnt); 951 nd6_prelist_remove(pr); 952 } 953 } 954 955 static int 956 nd6_prelist_add(struct nd_prefixctl *prc, struct nd_defrouter *dr, 957 struct nd_prefix **newp) 958 { 959 struct nd_prefix *newpr = NULL; 960 int i; 961 int error; 962 struct in6_ifextra *ext = prc->ndprc_ifp->if_afdata[AF_INET6]; 963 964 ND6_ASSERT_WLOCK(); 965 966 if (ip6_maxifprefixes >= 0) { 967 if (ext->nprefixes >= ip6_maxifprefixes / 2) 968 purge_detached(prc->ndprc_ifp); 969 if (ext->nprefixes >= ip6_maxifprefixes) 970 return ENOMEM; 971 } 972 973 error = 0; 974 newpr = malloc(sizeof(*newpr), M_IP6NDP, M_NOWAIT|M_ZERO); 975 if (newpr == NULL) 976 return ENOMEM; 977 newpr->ndpr_ifp = prc->ndprc_ifp; 978 newpr->ndpr_prefix = prc->ndprc_prefix; 979 newpr->ndpr_plen = prc->ndprc_plen; 980 newpr->ndpr_vltime = prc->ndprc_vltime; 981 newpr->ndpr_pltime = prc->ndprc_pltime; 982 newpr->ndpr_flags = prc->ndprc_flags; 983 if ((error = in6_init_prefix_ltimes(newpr)) != 0) { 984 free(newpr, M_IP6NDP); 985 return(error); 986 } 987 newpr->ndpr_lastupdate = time_uptime; 988 if (newp != NULL) 989 *newp = newpr; 990 991 /* initialization */ 992 LIST_INIT(&newpr->ndpr_advrtrs); 993 in6_prefixlen2mask(&newpr->ndpr_mask, newpr->ndpr_plen); 994 /* make prefix in the canonical form */ 995 for (i = 0; i < 4; i++) { 996 newpr->ndpr_prefix.sin6_addr.s6_addr32[i] &= 997 newpr->ndpr_mask.s6_addr32[i]; 998 } 999 1000 /* link ndpr_entry to nd_prefix list */ 1001 ND_PREFIX_LIST_INSERT_HEAD(newpr); 1002 1003 /* ND_OPT_PI_FLAG_ONLINK processing */ 1004 if (newpr->ndpr_raf_onlink) { 1005 int e; 1006 1007 if ((e = nd6_prefix_onlink(newpr)) != 0) { 1008 char ip6buf[INET6_ADDRSTRLEN]; 1009 nd6log(LOG_ERR, "failed to make " 1010 "the prefix %s/%d on-link on %s (errno=%d)\n", 1011 IN6_PRINT(ip6buf, &prc->ndprc_prefix.sin6_addr), 1012 prc->ndprc_plen, if_name(prc->ndprc_ifp), e); 1013 /* proceed anyway. XXX: is it correct? */ 1014 } 1015 } 1016 1017 if (dr) 1018 pfxrtr_add(newpr, dr); 1019 1020 ext->nprefixes++; 1021 1022 return 0; 1023 } 1024 1025 void 1026 nd6_prefix_unref(struct nd_prefix *pr) 1027 { 1028 1029 ND6_WLOCK(); 1030 pr->ndpr_refcnt--; 1031 if (pr->ndpr_refcnt == 0) 1032 nd6_prelist_remove(pr); 1033 ND6_UNLOCK(); 1034 } 1035 1036 void 1037 nd6_invalidate_prefix(struct nd_prefix *pr) 1038 { 1039 int e; 1040 1041 ND6_ASSERT_WLOCK(); 1042 1043 /* make sure to invalidate the prefix until it is really freed. */ 1044 pr->ndpr_vltime = 0; 1045 pr->ndpr_pltime = 0; 1046 #if 0 1047 /* 1048 * Though these flags are now meaningless, we'd rather keep the value 1049 * not to confuse users when executing "ndp -p". 1050 */ 1051 pr->ndpr_raf_onlink = 0; 1052 pr->ndpr_raf_auto = 0; 1053 #endif 1054 if ((pr->ndpr_stateflags & NDPRF_ONLINK) != 0 && 1055 (e = nd6_prefix_offlink(pr)) != 0) { 1056 char ip6buf[INET6_ADDRSTRLEN]; 1057 nd6log(LOG_ERR, 1058 "failed to make %s/%d offlink on %s, errno=%d\n", 1059 IN6_PRINT(ip6buf, &pr->ndpr_prefix.sin6_addr), 1060 pr->ndpr_plen, if_name(pr->ndpr_ifp), e); 1061 /* what should we do? */ 1062 } 1063 } 1064 1065 void 1066 nd6_prelist_remove(struct nd_prefix *pr) 1067 { 1068 struct nd_pfxrouter *pfr, *next; 1069 struct in6_ifextra *ext = pr->ndpr_ifp->if_afdata[AF_INET6]; 1070 1071 ND6_ASSERT_WLOCK(); 1072 KASSERTMSG(pr->ndpr_refcnt == 0, "ndpr_refcnt=%d", pr->ndpr_refcnt); 1073 1074 nd6_invalidate_prefix(pr); 1075 1076 /* unlink ndpr_entry from nd_prefix list */ 1077 ND_PREFIX_LIST_REMOVE(pr); 1078 1079 /* free list of routers that adversed the prefix */ 1080 for (pfr = LIST_FIRST(&pr->ndpr_advrtrs); pfr != NULL; pfr = next) { 1081 next = LIST_NEXT(pfr, pfr_entry); 1082 1083 free(pfr, M_IP6NDP); 1084 } 1085 1086 if (ext) { 1087 ext->nprefixes--; 1088 if (ext->nprefixes < 0) { 1089 log(LOG_WARNING, "nd6_prelist_remove: negative count on " 1090 "%s\n", pr->ndpr_ifp->if_xname); 1091 } 1092 } 1093 1094 free(pr, M_IP6NDP); 1095 1096 nd6_pfxlist_onlink_check(); 1097 } 1098 1099 static int 1100 prelist_update(struct nd_prefixctl *newprc, 1101 struct nd_defrouter *dr, /* may be NULL */ 1102 struct mbuf *m, 1103 int mcast) 1104 { 1105 struct in6_ifaddr *ia6_match = NULL; 1106 struct ifaddr *ifa; 1107 struct ifnet *ifp = newprc->ndprc_ifp; 1108 struct nd_prefix *pr; 1109 int error = 0; 1110 int auth; 1111 struct in6_addrlifetime lt6_tmp; 1112 int ss; 1113 char ip6buf[INET6_ADDRSTRLEN]; 1114 1115 KASSERT(m != NULL); 1116 ND6_ASSERT_WLOCK(); 1117 1118 auth = (m->m_flags & M_AUTHIPHDR) ? 1 : 0; 1119 1120 if ((pr = nd6_prefix_lookup(newprc)) != NULL) { 1121 /* 1122 * nd6_prefix_lookup() ensures that pr and newprc have the same 1123 * prefix on a same interface. 1124 */ 1125 1126 /* 1127 * Update prefix information. Note that the on-link (L) bit 1128 * and the autonomous (A) bit should NOT be changed from 1 1129 * to 0. 1130 */ 1131 if (newprc->ndprc_raf_onlink == 1) 1132 pr->ndpr_raf_onlink = 1; 1133 if (newprc->ndprc_raf_auto == 1) 1134 pr->ndpr_raf_auto = 1; 1135 if (newprc->ndprc_raf_onlink) { 1136 pr->ndpr_vltime = newprc->ndprc_vltime; 1137 pr->ndpr_pltime = newprc->ndprc_pltime; 1138 (void)in6_init_prefix_ltimes(pr); /* XXX error case? */ 1139 pr->ndpr_lastupdate = time_uptime; 1140 } 1141 1142 if (newprc->ndprc_raf_onlink && 1143 (pr->ndpr_stateflags & NDPRF_ONLINK) == 0) { 1144 int e; 1145 1146 if ((e = nd6_prefix_onlink(pr)) != 0) { 1147 nd6log(LOG_ERR, 1148 "failed to make " 1149 "the prefix %s/%d on-link on %s " 1150 "(errno=%d)\n", 1151 IN6_PRINT(ip6buf, 1152 &pr->ndpr_prefix.sin6_addr), 1153 pr->ndpr_plen, if_name(pr->ndpr_ifp), e); 1154 /* proceed anyway. XXX: is it correct? */ 1155 } 1156 } 1157 1158 if (dr && pfxrtr_lookup(pr, dr) == NULL) 1159 pfxrtr_add(pr, dr); 1160 } else { 1161 struct nd_prefix *newpr = NULL; 1162 1163 if (newprc->ndprc_vltime == 0) 1164 goto end; 1165 if (newprc->ndprc_raf_onlink == 0 && newprc->ndprc_raf_auto == 0) 1166 goto end; 1167 1168 if (ip6_rtadv_maxroutes <= nd6_numroutes) { 1169 ICMP6_STATINC(ICMP6_STAT_DROPPED_RAROUTE); 1170 goto end; 1171 } 1172 1173 error = nd6_prelist_add(newprc, dr, &newpr); 1174 if (error != 0 || newpr == NULL) { 1175 nd6log(LOG_NOTICE, 1176 "nd6_prelist_add failed for %s/%d on %s " 1177 "errno=%d, returnpr=%p\n", 1178 IN6_PRINT(ip6buf, &newprc->ndprc_prefix.sin6_addr), 1179 newprc->ndprc_plen, if_name(newprc->ndprc_ifp), 1180 error, newpr); 1181 goto end; /* we should just give up in this case. */ 1182 } 1183 1184 /* 1185 * XXX: from the ND point of view, we can ignore a prefix 1186 * with the on-link bit being zero. However, we need a 1187 * prefix structure for references from autoconfigured 1188 * addresses. Thus, we explicitly make sure that the prefix 1189 * itself expires now. 1190 */ 1191 if (newpr->ndpr_raf_onlink == 0) { 1192 newpr->ndpr_vltime = 0; 1193 newpr->ndpr_pltime = 0; 1194 in6_init_prefix_ltimes(newpr); 1195 } 1196 1197 pr = newpr; 1198 } 1199 1200 /* 1201 * Address autoconfiguration based on Section 5.5.3 of RFC 2462. 1202 * Note that pr must be non NULL at this point. 1203 */ 1204 1205 /* 5.5.3 (a). Ignore the prefix without the A bit set. */ 1206 if (!newprc->ndprc_raf_auto) 1207 goto end; 1208 1209 /* 1210 * 5.5.3 (b). the link-local prefix should have been ignored in 1211 * nd6_ra_input. 1212 */ 1213 1214 /* 5.5.3 (c). Consistency check on lifetimes: pltime <= vltime. */ 1215 if (newprc->ndprc_pltime > newprc->ndprc_vltime) { 1216 error = EINVAL; /* XXX: won't be used */ 1217 goto end; 1218 } 1219 1220 /* 1221 * 5.5.3 (d). If the prefix advertised is not equal to the prefix of 1222 * an address configured by stateless autoconfiguration already in the 1223 * list of addresses associated with the interface, and the Valid 1224 * Lifetime is not 0, form an address. We first check if we have 1225 * a matching prefix. 1226 * Note: we apply a clarification in rfc2462bis-02 here. We only 1227 * consider autoconfigured addresses while RFC2462 simply said 1228 * "address". 1229 */ 1230 ss = pserialize_read_enter(); 1231 IFADDR_READER_FOREACH(ifa, ifp) { 1232 struct in6_ifaddr *ia6; 1233 u_int32_t remaininglifetime; 1234 1235 if (ifa->ifa_addr->sa_family != AF_INET6) 1236 continue; 1237 1238 ia6 = (struct in6_ifaddr *)ifa; 1239 1240 /* 1241 * We only consider autoconfigured addresses as per rfc2462bis. 1242 */ 1243 if (!(ia6->ia6_flags & IN6_IFF_AUTOCONF)) 1244 continue; 1245 1246 /* 1247 * Spec is not clear here, but I believe we should concentrate 1248 * on unicast (i.e. not anycast) addresses. 1249 * XXX: other ia6_flags? detached or duplicated? 1250 */ 1251 if ((ia6->ia6_flags & IN6_IFF_ANYCAST) != 0) 1252 continue; 1253 1254 /* 1255 * Ignore the address if it is not associated with a prefix 1256 * or is associated with a prefix that is different from this 1257 * one. (pr is never NULL here) 1258 */ 1259 if (ia6->ia6_ndpr != pr) 1260 continue; 1261 1262 if (ia6_match == NULL) /* remember the first one */ 1263 ia6_match = ia6; 1264 1265 /* 1266 * An already autoconfigured address matched. Now that we 1267 * are sure there is at least one matched address, we can 1268 * proceed to 5.5.3. (e): update the lifetimes according to the 1269 * "two hours" rule and the privacy extension. 1270 * We apply some clarifications in rfc2462bis: 1271 * - use remaininglifetime instead of storedlifetime as a 1272 * variable name 1273 * - remove the dead code in the "two-hour" rule 1274 */ 1275 #define TWOHOUR (120*60) 1276 lt6_tmp = ia6->ia6_lifetime; 1277 if (lt6_tmp.ia6t_vltime == ND6_INFINITE_LIFETIME) 1278 remaininglifetime = ND6_INFINITE_LIFETIME; 1279 else if (time_uptime - ia6->ia6_updatetime > 1280 lt6_tmp.ia6t_vltime) { 1281 /* 1282 * The case of "invalid" address. We should usually 1283 * not see this case. 1284 */ 1285 remaininglifetime = 0; 1286 } else 1287 remaininglifetime = lt6_tmp.ia6t_vltime - 1288 (time_uptime - ia6->ia6_updatetime); 1289 1290 /* when not updating, keep the current stored lifetime. */ 1291 lt6_tmp.ia6t_vltime = remaininglifetime; 1292 1293 if (TWOHOUR < newprc->ndprc_vltime || 1294 remaininglifetime < newprc->ndprc_vltime) { 1295 lt6_tmp.ia6t_vltime = newprc->ndprc_vltime; 1296 } else if (remaininglifetime <= TWOHOUR) { 1297 if (auth) 1298 lt6_tmp.ia6t_vltime = newprc->ndprc_vltime; 1299 } else { 1300 /* 1301 * newprc->ndprc_vltime <= TWOHOUR && 1302 * TWOHOUR < remaininglifetime 1303 */ 1304 lt6_tmp.ia6t_vltime = TWOHOUR; 1305 } 1306 1307 /* The 2 hour rule is not imposed for preferred lifetime. */ 1308 lt6_tmp.ia6t_pltime = newprc->ndprc_pltime; 1309 1310 in6_init_address_ltimes(pr, <6_tmp); 1311 1312 /* 1313 * We need to treat lifetimes for temporary addresses 1314 * differently, according to 1315 * draft-ietf-ipv6-privacy-addrs-v2-01.txt 3.3 (1); 1316 * we only update the lifetimes when they are in the maximum 1317 * intervals. 1318 */ 1319 if ((ia6->ia6_flags & IN6_IFF_TEMPORARY) != 0) { 1320 u_int32_t maxvltime, maxpltime; 1321 1322 if (ip6_temp_valid_lifetime > 1323 (u_int32_t)((time_uptime - ia6->ia6_createtime) + 1324 ip6_desync_factor)) { 1325 maxvltime = ip6_temp_valid_lifetime - 1326 (time_uptime - ia6->ia6_createtime) - 1327 ip6_desync_factor; 1328 } else 1329 maxvltime = 0; 1330 if (ip6_temp_preferred_lifetime > 1331 (u_int32_t)((time_uptime - ia6->ia6_createtime) + 1332 ip6_desync_factor)) { 1333 maxpltime = ip6_temp_preferred_lifetime - 1334 (time_uptime - ia6->ia6_createtime) - 1335 ip6_desync_factor; 1336 } else 1337 maxpltime = 0; 1338 1339 if (lt6_tmp.ia6t_vltime == ND6_INFINITE_LIFETIME || 1340 lt6_tmp.ia6t_vltime > maxvltime) { 1341 lt6_tmp.ia6t_vltime = maxvltime; 1342 } 1343 if (lt6_tmp.ia6t_pltime == ND6_INFINITE_LIFETIME || 1344 lt6_tmp.ia6t_pltime > maxpltime) { 1345 lt6_tmp.ia6t_pltime = maxpltime; 1346 } 1347 } 1348 1349 ia6->ia6_lifetime = lt6_tmp; 1350 ia6->ia6_updatetime = time_uptime; 1351 } 1352 pserialize_read_exit(ss); 1353 1354 if (ia6_match == NULL && newprc->ndprc_vltime) { 1355 int ifidlen; 1356 struct in6_ifaddr *ia6; 1357 struct psref psref; 1358 1359 /* 1360 * 5.5.3 (d) (continued) 1361 * No address matched and the valid lifetime is non-zero. 1362 * Create a new address. 1363 */ 1364 1365 /* 1366 * Prefix Length check: 1367 * If the sum of the prefix length and interface identifier 1368 * length does not equal 128 bits, the Prefix Information 1369 * option MUST be ignored. The length of the interface 1370 * identifier is defined in a separate link-type specific 1371 * document. 1372 */ 1373 ifidlen = in6_if2idlen(ifp); 1374 if (ifidlen < 0) { 1375 /* this should not happen, so we always log it. */ 1376 log(LOG_ERR, "%s: IFID undefined (%s)\n", 1377 __func__, if_name(ifp)); 1378 goto end; 1379 } 1380 if (ifidlen + pr->ndpr_plen != 128) { 1381 nd6log(LOG_INFO, 1382 "invalid prefixlen %d for %s, ignored\n", 1383 pr->ndpr_plen, if_name(ifp)); 1384 goto end; 1385 } 1386 1387 if ((ia6 = in6_ifadd(newprc, mcast, &psref)) != NULL) { 1388 /* 1389 * note that we should use pr (not newprc) for reference. 1390 */ 1391 pr->ndpr_refcnt++; 1392 ia6->ia6_ndpr = pr; 1393 1394 /* toggle onlink state if the address was assigned 1395 * a prefix route. */ 1396 if (ia6->ia_flags & IFA_ROUTE) 1397 pr->ndpr_stateflags |= NDPRF_ONLINK; 1398 1399 /* 1400 * draft-ietf-ipngwg-temp-addresses-v2-00 3.3 (2). 1401 * When a new public address is created as described 1402 * in RFC2462, also create a new temporary address. 1403 * 1404 * draft-ietf-ipngwg-temp-addresses-v2-00 3.5. 1405 * When an interface connects to a new link, a new 1406 * randomized interface identifier should be generated 1407 * immediately together with a new set of temporary 1408 * addresses. Thus, we specifiy 1 as the 2nd arg of 1409 * in6_tmpifadd(). 1410 */ 1411 if (ip6_use_tempaddr) { 1412 int e; 1413 if ((e = in6_tmpifadd(ia6, 1, 1)) != 0) { 1414 nd6log(LOG_NOTICE, 1415 "failed to create a temporary " 1416 "address, errno=%d\n", e); 1417 } 1418 } 1419 ia6_release(ia6, &psref); 1420 1421 /* 1422 * A newly added address might affect the status 1423 * of other addresses, so we check and update it. 1424 * XXX: what if address duplication happens? 1425 */ 1426 nd6_pfxlist_onlink_check(); 1427 } else { 1428 /* just set an error. do not bark here. */ 1429 error = EADDRNOTAVAIL; /* XXX: might be unused. */ 1430 } 1431 } 1432 1433 end: 1434 return error; 1435 } 1436 1437 /* 1438 * A supplement function used in the on-link detection below; 1439 * detect if a given prefix has a (probably) reachable advertising router. 1440 * XXX: lengthy function name... 1441 */ 1442 static struct nd_pfxrouter * 1443 find_pfxlist_reachable_router(struct nd_prefix *pr) 1444 { 1445 struct nd_pfxrouter *pfxrtr; 1446 1447 for (pfxrtr = LIST_FIRST(&pr->ndpr_advrtrs); pfxrtr; 1448 pfxrtr = LIST_NEXT(pfxrtr, pfr_entry)) { 1449 if (pfxrtr->router->ifp->if_flags & IFF_UP && 1450 pfxrtr->router->ifp->if_link_state != LINK_STATE_DOWN && 1451 nd6_is_llinfo_probreach(pfxrtr->router)) 1452 break; /* found */ 1453 } 1454 1455 return (pfxrtr); 1456 } 1457 1458 /* 1459 * Check if each prefix in the prefix list has at least one available router 1460 * that advertised the prefix (a router is "available" if its neighbor cache 1461 * entry is reachable or probably reachable). 1462 * If the check fails, the prefix may be off-link, because, for example, 1463 * we have moved from the network but the lifetime of the prefix has not 1464 * expired yet. So we should not use the prefix if there is another prefix 1465 * that has an available router. 1466 * But, if there is no prefix that has an available router, we still regards 1467 * all the prefixes as on-link. This is because we can't tell if all the 1468 * routers are simply dead or if we really moved from the network and there 1469 * is no router around us. 1470 */ 1471 void 1472 nd6_pfxlist_onlink_check(void) 1473 { 1474 struct nd_prefix *pr; 1475 struct in6_ifaddr *ia; 1476 struct nd_defrouter *dr; 1477 struct nd_pfxrouter *pfxrtr = NULL; 1478 int s; 1479 char ip6buf[INET6_ADDRSTRLEN]; 1480 1481 ND6_ASSERT_WLOCK(); 1482 1483 /* 1484 * Check if there is a prefix that has a reachable advertising 1485 * router. 1486 */ 1487 ND_PREFIX_LIST_FOREACH(pr) { 1488 if (pr->ndpr_raf_onlink && find_pfxlist_reachable_router(pr)) 1489 break; 1490 } 1491 /* 1492 * If we have no such prefix, check whether we still have a router 1493 * that does not advertise any prefixes. 1494 */ 1495 if (pr == NULL) { 1496 ND_DEFROUTER_LIST_FOREACH(dr) { 1497 struct nd_prefix *pr0; 1498 1499 ND_PREFIX_LIST_FOREACH(pr0) { 1500 if ((pfxrtr = pfxrtr_lookup(pr0, dr)) != NULL) 1501 break; 1502 } 1503 if (pfxrtr) 1504 break; 1505 } 1506 } 1507 if (pr != NULL || (!ND_DEFROUTER_LIST_EMPTY() && !pfxrtr)) { 1508 /* 1509 * There is at least one prefix that has a reachable router, 1510 * or at least a router which probably does not advertise 1511 * any prefixes. The latter would be the case when we move 1512 * to a new link where we have a router that does not provide 1513 * prefixes and we configure an address by hand. 1514 * Detach prefixes which have no reachable advertising 1515 * router, and attach other prefixes. 1516 */ 1517 ND_PREFIX_LIST_FOREACH(pr) { 1518 /* XXX: a link-local prefix should never be detached */ 1519 if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr)) 1520 continue; 1521 1522 /* 1523 * we aren't interested in prefixes without the L bit 1524 * set. 1525 */ 1526 if (pr->ndpr_raf_onlink == 0) 1527 continue; 1528 1529 if ((pr->ndpr_stateflags & NDPRF_DETACHED) == 0 && 1530 find_pfxlist_reachable_router(pr) == NULL) 1531 pr->ndpr_stateflags |= NDPRF_DETACHED; 1532 if ((pr->ndpr_stateflags & NDPRF_DETACHED) != 0 && 1533 find_pfxlist_reachable_router(pr) != 0) 1534 pr->ndpr_stateflags &= ~NDPRF_DETACHED; 1535 } 1536 } else { 1537 /* there is no prefix that has a reachable router */ 1538 ND_PREFIX_LIST_FOREACH(pr) { 1539 if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr)) 1540 continue; 1541 1542 if (pr->ndpr_raf_onlink == 0) 1543 continue; 1544 1545 if ((pr->ndpr_stateflags & NDPRF_DETACHED) != 0) 1546 pr->ndpr_stateflags &= ~NDPRF_DETACHED; 1547 } 1548 } 1549 1550 /* 1551 * Remove each interface route associated with a (just) detached 1552 * prefix, and reinstall the interface route for a (just) attached 1553 * prefix. Note that all attempt of reinstallation does not 1554 * necessarily success, when a same prefix is shared among multiple 1555 * interfaces. Such cases will be handled in nd6_prefix_onlink, 1556 * so we don't have to care about them. 1557 */ 1558 ND_PREFIX_LIST_FOREACH(pr) { 1559 int e; 1560 1561 if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr)) 1562 continue; 1563 1564 if (pr->ndpr_raf_onlink == 0) 1565 continue; 1566 1567 if ((pr->ndpr_stateflags & NDPRF_DETACHED) != 0 && 1568 (pr->ndpr_stateflags & NDPRF_ONLINK) != 0) { 1569 if ((e = nd6_prefix_offlink(pr)) != 0) { 1570 nd6log(LOG_ERR, 1571 "failed to make %s/%d offlink, errno=%d\n", 1572 IN6_PRINT(ip6buf, 1573 &pr->ndpr_prefix.sin6_addr), 1574 pr->ndpr_plen, e); 1575 } 1576 } 1577 if ((pr->ndpr_stateflags & NDPRF_DETACHED) == 0 && 1578 (pr->ndpr_stateflags & NDPRF_ONLINK) == 0 && 1579 pr->ndpr_raf_onlink) { 1580 if ((e = nd6_prefix_onlink(pr)) != 0) { 1581 nd6log(LOG_ERR, 1582 "failed to make %s/%d onlink, errno=%d\n", 1583 IN6_PRINT(ip6buf, 1584 &pr->ndpr_prefix.sin6_addr), 1585 pr->ndpr_plen, e); 1586 } 1587 } 1588 } 1589 1590 int bound = curlwp_bind(); 1591 /* 1592 * Changes on the prefix status might affect address status as well. 1593 * Make sure that all addresses derived from an attached prefix are 1594 * attached, and that all addresses derived from a detached prefix are 1595 * detached. Note, however, that a manually configured address should 1596 * always be attached. 1597 * The precise detection logic is same as the one for prefixes. 1598 */ 1599 s = pserialize_read_enter(); 1600 IN6_ADDRLIST_READER_FOREACH(ia) { 1601 struct psref psref; 1602 bool found; 1603 1604 if (!(ia->ia6_flags & IN6_IFF_AUTOCONF)) 1605 continue; 1606 1607 if (ia->ia6_ndpr == NULL) { 1608 /* 1609 * This can happen when we first configure the address 1610 * (i.e. the address exists, but the prefix does not). 1611 * XXX: complicated relationships... 1612 */ 1613 continue; 1614 } 1615 1616 ia6_acquire(ia, &psref); 1617 pserialize_read_exit(s); 1618 1619 found = find_pfxlist_reachable_router(ia->ia6_ndpr) != NULL; 1620 1621 s = pserialize_read_enter(); 1622 ia6_release(ia, &psref); 1623 if (found) 1624 break; 1625 } 1626 pserialize_read_exit(s); 1627 1628 if (ia) { 1629 s = pserialize_read_enter(); 1630 IN6_ADDRLIST_READER_FOREACH(ia) { 1631 struct ifaddr *ifa = (struct ifaddr *)ia; 1632 struct psref psref; 1633 1634 if ((ia->ia6_flags & IN6_IFF_AUTOCONF) == 0) 1635 continue; 1636 1637 if (ia->ia6_ndpr == NULL) /* XXX: see above. */ 1638 continue; 1639 1640 ia6_acquire(ia, &psref); 1641 pserialize_read_exit(s); 1642 1643 if (find_pfxlist_reachable_router(ia->ia6_ndpr)) { 1644 if (ia->ia6_flags & IN6_IFF_DETACHED) { 1645 ia->ia6_flags &= ~IN6_IFF_DETACHED; 1646 ia->ia6_flags |= IN6_IFF_TENTATIVE; 1647 nd6_dad_start(ifa, 1648 0); 1649 /* We will notify the routing socket 1650 * of the DAD result, so no need to 1651 * here */ 1652 } 1653 } else { 1654 if ((ia->ia6_flags & IN6_IFF_DETACHED) == 0) { 1655 ia->ia6_flags |= IN6_IFF_DETACHED; 1656 rt_addrmsg(RTM_NEWADDR, ifa); 1657 } 1658 } 1659 1660 s = pserialize_read_enter(); 1661 ia6_release(ia, &psref); 1662 } 1663 pserialize_read_exit(s); 1664 } 1665 else { 1666 s = pserialize_read_enter(); 1667 IN6_ADDRLIST_READER_FOREACH(ia) { 1668 if ((ia->ia6_flags & IN6_IFF_AUTOCONF) == 0) 1669 continue; 1670 1671 if (ia->ia6_flags & IN6_IFF_DETACHED) { 1672 struct ifaddr *ifa = (struct ifaddr *)ia; 1673 struct psref psref; 1674 1675 ia->ia6_flags &= ~IN6_IFF_DETACHED; 1676 ia->ia6_flags |= IN6_IFF_TENTATIVE; 1677 1678 ia6_acquire(ia, &psref); 1679 pserialize_read_exit(s); 1680 1681 /* Do we need a delay in this case? */ 1682 nd6_dad_start(ifa, 0); 1683 1684 s = pserialize_read_enter(); 1685 ia6_release(ia, &psref); 1686 } 1687 } 1688 pserialize_read_exit(s); 1689 } 1690 1691 curlwp_bindx(bound); 1692 } 1693 1694 static int 1695 nd6_prefix_onlink(struct nd_prefix *pr) 1696 { 1697 struct ifaddr *ifa; 1698 struct ifnet *ifp = pr->ndpr_ifp; 1699 struct sockaddr_in6 mask6; 1700 struct nd_prefix *opr; 1701 u_long rtflags; 1702 int error = 0; 1703 struct psref psref; 1704 int bound; 1705 char ip6buf[INET6_ADDRSTRLEN]; 1706 char ip6bufp[INET6_ADDRSTRLEN], ip6bufm[INET6_ADDRSTRLEN]; 1707 1708 ND6_ASSERT_WLOCK(); 1709 1710 /* sanity check */ 1711 if ((pr->ndpr_stateflags & NDPRF_ONLINK) != 0) { 1712 nd6log(LOG_ERR, "%s/%d is already on-link\n", 1713 IN6_PRINT(ip6buf, &pr->ndpr_prefix.sin6_addr), 1714 pr->ndpr_plen); 1715 return (EEXIST); 1716 } 1717 1718 /* 1719 * Add the interface route associated with the prefix. Before 1720 * installing the route, check if there's the same prefix on another 1721 * interface, and the prefix has already installed the interface route. 1722 * Although such a configuration is expected to be rare, we explicitly 1723 * allow it. 1724 */ 1725 ND_PREFIX_LIST_FOREACH(opr) { 1726 if (opr == pr) 1727 continue; 1728 1729 if ((opr->ndpr_stateflags & NDPRF_ONLINK) == 0) 1730 continue; 1731 1732 if (opr->ndpr_plen == pr->ndpr_plen && 1733 in6_are_prefix_equal(&pr->ndpr_prefix.sin6_addr, 1734 &opr->ndpr_prefix.sin6_addr, pr->ndpr_plen)) 1735 return (0); 1736 } 1737 1738 /* 1739 * We prefer link-local addresses as the associated interface address. 1740 */ 1741 /* search for a link-local addr */ 1742 bound = curlwp_bind(); 1743 ifa = (struct ifaddr *)in6ifa_ifpforlinklocal_psref(ifp, 1744 IN6_IFF_NOTREADY | IN6_IFF_ANYCAST, &psref); 1745 if (ifa == NULL) { 1746 int s = pserialize_read_enter(); 1747 IFADDR_READER_FOREACH(ifa, ifp) { 1748 if (ifa->ifa_addr->sa_family == AF_INET6) 1749 break; 1750 } 1751 if (ifa != NULL) 1752 ifa_acquire(ifa, &psref); 1753 pserialize_read_exit(s); 1754 /* should we care about ia6_flags? */ 1755 } 1756 if (ifa == NULL) { 1757 /* 1758 * This can still happen, when, for example, we receive an RA 1759 * containing a prefix with the L bit set and the A bit clear, 1760 * after removing all IPv6 addresses on the receiving 1761 * interface. This should, of course, be rare though. 1762 */ 1763 nd6log(LOG_NOTICE, "failed to find any ifaddr" 1764 " to add route for a prefix(%s/%d) on %s\n", 1765 IN6_PRINT(ip6buf, &pr->ndpr_prefix.sin6_addr), 1766 pr->ndpr_plen, if_name(ifp)); 1767 curlwp_bindx(bound); 1768 return (0); 1769 } 1770 1771 /* 1772 * in6_ifinit() sets nd6_rtrequest to ifa_rtrequest for all ifaddrs. 1773 * ifa->ifa_rtrequest = nd6_rtrequest; 1774 */ 1775 memset(&mask6, 0, sizeof(mask6)); 1776 mask6.sin6_family = AF_INET6; 1777 mask6.sin6_len = sizeof(mask6); 1778 mask6.sin6_addr = pr->ndpr_mask; 1779 /* rtrequest() will probably set RTF_UP, but we're not sure. */ 1780 rtflags = ifa->ifa_flags | RTF_UP; 1781 if (nd6_need_cache(ifp)) { 1782 /* explicitly set in case ifa_flags does not set the flag. */ 1783 rtflags |= RTF_CONNECTED; 1784 } else { 1785 /* 1786 * explicitly clear the cloning bit in case ifa_flags sets it. 1787 */ 1788 rtflags &= ~RTF_CONNECTED; 1789 } 1790 error = rtrequest_newmsg(RTM_ADD, sin6tosa(&pr->ndpr_prefix), 1791 ifa->ifa_addr, sin6tosa(&mask6), rtflags); 1792 if (error == 0) { 1793 nd6_numroutes++; 1794 pr->ndpr_stateflags |= NDPRF_ONLINK; 1795 } else { 1796 nd6log(LOG_ERR, "failed to add route for a" 1797 " prefix (%s/%d) on %s, gw=%s, mask=%s, flags=%lx " 1798 "errno = %d\n", 1799 IN6_PRINT(ip6bufp, &pr->ndpr_prefix.sin6_addr), 1800 pr->ndpr_plen, if_name(ifp), 1801 IN6_PRINT(ip6buf, 1802 &((struct sockaddr_in6 *)ifa->ifa_addr)->sin6_addr), 1803 IN6_PRINT(ip6bufm, &mask6.sin6_addr), rtflags, error); 1804 } 1805 ifa_release(ifa, &psref); 1806 curlwp_bindx(bound); 1807 1808 return (error); 1809 } 1810 1811 static int 1812 nd6_prefix_offlink(struct nd_prefix *pr) 1813 { 1814 int error = 0; 1815 struct ifnet *ifp = pr->ndpr_ifp; 1816 struct nd_prefix *opr; 1817 struct sockaddr_in6 sa6, mask6; 1818 char ip6buf[INET6_ADDRSTRLEN]; 1819 1820 ND6_ASSERT_WLOCK(); 1821 1822 /* sanity check */ 1823 if ((pr->ndpr_stateflags & NDPRF_ONLINK) == 0) { 1824 nd6log(LOG_ERR, "%s/%d is already off-link\n", 1825 IN6_PRINT(ip6buf, &pr->ndpr_prefix.sin6_addr), 1826 pr->ndpr_plen); 1827 return (EEXIST); 1828 } 1829 1830 sockaddr_in6_init(&sa6, &pr->ndpr_prefix.sin6_addr, 0, 0, 0); 1831 sockaddr_in6_init(&mask6, &pr->ndpr_mask, 0, 0, 0); 1832 error = rtrequest_newmsg(RTM_DELETE, sin6tosa(&sa6), NULL, 1833 sin6tosa(&mask6), 0); 1834 if (error == 0) { 1835 pr->ndpr_stateflags &= ~NDPRF_ONLINK; 1836 nd6_numroutes--; 1837 1838 /* 1839 * There might be the same prefix on another interface, 1840 * the prefix which could not be on-link just because we have 1841 * the interface route (see comments in nd6_prefix_onlink). 1842 * If there's one, try to make the prefix on-link on the 1843 * interface. 1844 */ 1845 ND_PREFIX_LIST_FOREACH(opr) { 1846 if (opr == pr) 1847 continue; 1848 1849 if ((opr->ndpr_stateflags & NDPRF_ONLINK) != 0) 1850 continue; 1851 1852 /* 1853 * KAME specific: detached prefixes should not be 1854 * on-link. 1855 */ 1856 if ((opr->ndpr_stateflags & NDPRF_DETACHED) != 0) 1857 continue; 1858 1859 if (opr->ndpr_plen == pr->ndpr_plen && 1860 in6_are_prefix_equal(&pr->ndpr_prefix.sin6_addr, 1861 &opr->ndpr_prefix.sin6_addr, pr->ndpr_plen)) { 1862 int e; 1863 1864 if ((e = nd6_prefix_onlink(opr)) != 0) { 1865 nd6log(LOG_ERR, "failed to " 1866 "recover a prefix %s/%d from %s " 1867 "to %s (errno = %d)\n", 1868 IN6_PRINT(ip6buf, 1869 &opr->ndpr_prefix.sin6_addr), 1870 opr->ndpr_plen, if_name(ifp), 1871 if_name(opr->ndpr_ifp), e); 1872 } 1873 } 1874 } 1875 } else { 1876 /* XXX: can we still set the NDPRF_ONLINK flag? */ 1877 nd6log(LOG_ERR, "failed to delete route: " 1878 "%s/%d on %s (errno = %d)\n", 1879 IN6_PRINT(ip6buf, &sa6.sin6_addr), pr->ndpr_plen, 1880 if_name(ifp), 1881 error); 1882 } 1883 1884 return error; 1885 } 1886 1887 static struct in6_ifaddr * 1888 in6_ifadd(struct nd_prefixctl *prc, int mcast, struct psref *psref) 1889 { 1890 struct ifnet *ifp = prc->ndprc_ifp; 1891 struct ifaddr *ifa; 1892 struct in6_aliasreq ifra; 1893 struct in6_ifaddr *ia, *ib; 1894 int error, plen0; 1895 struct in6_addr mask; 1896 int prefixlen = prc->ndprc_plen; 1897 int updateflags; 1898 int s; 1899 char ip6buf[INET6_ADDRSTRLEN]; 1900 1901 ND6_ASSERT_WLOCK(); 1902 1903 in6_prefixlen2mask(&mask, prefixlen); 1904 1905 /* 1906 * find a link-local address (will be interface ID). 1907 * Is it really mandatory? Theoretically, a global or a site-local 1908 * address can be configured without a link-local address, if we 1909 * have a unique interface identifier... 1910 * 1911 * it is not mandatory to have a link-local address, we can generate 1912 * interface identifier on the fly. we do this because: 1913 * (1) it should be the easiest way to find interface identifier. 1914 * (2) RFC2462 5.4 suggesting the use of the same interface identifier 1915 * for multiple addresses on a single interface, and possible shortcut 1916 * of DAD. we omitted DAD for this reason in the past. 1917 * (3) a user can prevent autoconfiguration of global address 1918 * by removing link-local address by hand (this is partly because we 1919 * don't have other way to control the use of IPv6 on an interface. 1920 * this has been our design choice - cf. NRL's "ifconfig auto"). 1921 * (4) it is easier to manage when an interface has addresses 1922 * with the same interface identifier, than to have multiple addresses 1923 * with different interface identifiers. 1924 */ 1925 s = pserialize_read_enter(); 1926 ifa = (struct ifaddr *)in6ifa_ifpforlinklocal(ifp, 0); /* 0 is OK? */ 1927 if (ifa) 1928 ib = (struct in6_ifaddr *)ifa; 1929 else { 1930 pserialize_read_exit(s); 1931 return NULL; 1932 } 1933 1934 #if 0 /* don't care link local addr state, and always do DAD */ 1935 /* if link-local address is not eligible, do not autoconfigure. */ 1936 if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_NOTREADY) { 1937 printf("in6_ifadd: link-local address not ready\n"); 1938 return NULL; 1939 } 1940 #endif 1941 1942 /* prefixlen + ifidlen must be equal to 128 */ 1943 plen0 = in6_mask2len(&ib->ia_prefixmask.sin6_addr, NULL); 1944 if (prefixlen != plen0) { 1945 nd6log(LOG_INFO, "wrong prefixlen for %s " 1946 "(prefix=%d ifid=%d)\n", 1947 if_name(ifp), prefixlen, 128 - plen0); 1948 pserialize_read_exit(s); 1949 return NULL; 1950 } 1951 1952 /* make ifaddr */ 1953 1954 memset(&ifra, 0, sizeof(ifra)); 1955 /* 1956 * in6_update_ifa() does not use ifra_name, but we accurately set it 1957 * for safety. 1958 */ 1959 strncpy(ifra.ifra_name, if_name(ifp), sizeof(ifra.ifra_name)); 1960 sockaddr_in6_init(&ifra.ifra_addr, &prc->ndprc_prefix.sin6_addr, 0, 0, 0); 1961 /* prefix */ 1962 ifra.ifra_addr.sin6_addr.s6_addr32[0] &= mask.s6_addr32[0]; 1963 ifra.ifra_addr.sin6_addr.s6_addr32[1] &= mask.s6_addr32[1]; 1964 ifra.ifra_addr.sin6_addr.s6_addr32[2] &= mask.s6_addr32[2]; 1965 ifra.ifra_addr.sin6_addr.s6_addr32[3] &= mask.s6_addr32[3]; 1966 1967 /* interface ID */ 1968 ifra.ifra_addr.sin6_addr.s6_addr32[0] |= 1969 (ib->ia_addr.sin6_addr.s6_addr32[0] & ~mask.s6_addr32[0]); 1970 ifra.ifra_addr.sin6_addr.s6_addr32[1] |= 1971 (ib->ia_addr.sin6_addr.s6_addr32[1] & ~mask.s6_addr32[1]); 1972 ifra.ifra_addr.sin6_addr.s6_addr32[2] |= 1973 (ib->ia_addr.sin6_addr.s6_addr32[2] & ~mask.s6_addr32[2]); 1974 ifra.ifra_addr.sin6_addr.s6_addr32[3] |= 1975 (ib->ia_addr.sin6_addr.s6_addr32[3] & ~mask.s6_addr32[3]); 1976 pserialize_read_exit(s); 1977 1978 /* new prefix mask. */ 1979 sockaddr_in6_init(&ifra.ifra_prefixmask, &mask, 0, 0, 0); 1980 1981 /* lifetimes */ 1982 ifra.ifra_lifetime.ia6t_vltime = prc->ndprc_vltime; 1983 ifra.ifra_lifetime.ia6t_pltime = prc->ndprc_pltime; 1984 1985 /* XXX: scope zone ID? */ 1986 1987 ifra.ifra_flags |= IN6_IFF_AUTOCONF; /* obey autoconf */ 1988 1989 /* 1990 * Make sure that we do not have this address already. This should 1991 * usually not happen, but we can still see this case, e.g., if we 1992 * have manually configured the exact address to be configured. 1993 */ 1994 s = pserialize_read_enter(); 1995 if (in6ifa_ifpwithaddr(ifp, &ifra.ifra_addr.sin6_addr) != NULL) { 1996 /* this should be rare enough to make an explicit log */ 1997 log(LOG_INFO, "in6_ifadd: %s is already configured\n", 1998 IN6_PRINT(ip6buf, &ifra.ifra_addr.sin6_addr)); 1999 pserialize_read_exit(s); 2000 return (NULL); 2001 } 2002 pserialize_read_exit(s); 2003 2004 /* 2005 * Allocate ifaddr structure, link into chain, etc. 2006 * If we are going to create a new address upon receiving a multicasted 2007 * RA, we need to impose a random delay before starting DAD. 2008 * [draft-ietf-ipv6-rfc2462bis-02.txt, Section 5.4.2] 2009 */ 2010 updateflags = 0; 2011 if (mcast) 2012 updateflags |= IN6_IFAUPDATE_DADDELAY; 2013 if ((error = in6_update_ifa(ifp, &ifra, updateflags)) != 0) { 2014 nd6log(LOG_ERR, "failed to make ifaddr %s on %s (errno=%d)\n", 2015 IN6_PRINT(ip6buf, &ifra.ifra_addr.sin6_addr), if_name(ifp), 2016 error); 2017 return (NULL); /* ifaddr must not have been allocated. */ 2018 } 2019 2020 ia = in6ifa_ifpwithaddr_psref(ifp, &ifra.ifra_addr.sin6_addr, psref); 2021 2022 return (ia); /* this is always non-NULL */ 2023 } 2024 2025 int 2026 in6_tmpifadd( 2027 const struct in6_ifaddr *ia0, /* corresponding public address */ 2028 int forcegen, 2029 int dad_delay) 2030 { 2031 struct ifnet *ifp = ia0->ia_ifa.ifa_ifp; 2032 struct in6_ifaddr *newia, *ia; 2033 struct in6_aliasreq ifra; 2034 int i, error; 2035 int trylimit = 3; /* XXX: adhoc value */ 2036 int updateflags; 2037 u_int32_t randid[2]; 2038 u_int32_t vltime0, pltime0; 2039 int s; 2040 2041 ND6_ASSERT_WLOCK(); 2042 2043 memset(&ifra, 0, sizeof(ifra)); 2044 strncpy(ifra.ifra_name, if_name(ifp), sizeof(ifra.ifra_name)); 2045 ifra.ifra_addr = ia0->ia_addr; 2046 /* copy prefix mask */ 2047 ifra.ifra_prefixmask = ia0->ia_prefixmask; 2048 /* clear the old IFID */ 2049 for (i = 0; i < 4; i++) { 2050 ifra.ifra_addr.sin6_addr.s6_addr32[i] &= 2051 ifra.ifra_prefixmask.sin6_addr.s6_addr32[i]; 2052 } 2053 2054 again: 2055 if (in6_get_tmpifid(ifp, (u_int8_t *)randid, 2056 (const u_int8_t *)&ia0->ia_addr.sin6_addr.s6_addr[8], forcegen)) { 2057 nd6log(LOG_NOTICE, "failed to find a good random IFID\n"); 2058 return (EINVAL); 2059 } 2060 ifra.ifra_addr.sin6_addr.s6_addr32[2] |= 2061 (randid[0] & ~(ifra.ifra_prefixmask.sin6_addr.s6_addr32[2])); 2062 ifra.ifra_addr.sin6_addr.s6_addr32[3] |= 2063 (randid[1] & ~(ifra.ifra_prefixmask.sin6_addr.s6_addr32[3])); 2064 2065 /* 2066 * in6_get_tmpifid() quite likely provided a unique interface ID. 2067 * However, we may still have a chance to see collision, because 2068 * there may be a time lag between generation of the ID and generation 2069 * of the address. So, we'll do one more sanity check. 2070 */ 2071 s = pserialize_read_enter(); 2072 IN6_ADDRLIST_READER_FOREACH(ia) { 2073 if (IN6_ARE_ADDR_EQUAL(&ia->ia_addr.sin6_addr, 2074 &ifra.ifra_addr.sin6_addr)) { 2075 pserialize_read_exit(s); 2076 if (trylimit-- == 0) { 2077 /* 2078 * Give up. Something strange should have 2079 * happened. 2080 */ 2081 nd6log(LOG_NOTICE, 2082 "failed to find a unique random IFID\n"); 2083 return (EEXIST); 2084 } 2085 forcegen = 1; 2086 goto again; 2087 } 2088 } 2089 pserialize_read_exit(s); 2090 2091 /* 2092 * The Valid Lifetime is the lower of the Valid Lifetime of the 2093 * public address or TEMP_VALID_LIFETIME. 2094 * The Preferred Lifetime is the lower of the Preferred Lifetime 2095 * of the public address or TEMP_PREFERRED_LIFETIME - 2096 * DESYNC_FACTOR. 2097 */ 2098 if (ia0->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) { 2099 vltime0 = IFA6_IS_INVALID(ia0) ? 0 : 2100 (ia0->ia6_lifetime.ia6t_vltime - 2101 (time_uptime - ia0->ia6_updatetime)); 2102 if (vltime0 > ip6_temp_valid_lifetime) 2103 vltime0 = ip6_temp_valid_lifetime; 2104 } else 2105 vltime0 = ip6_temp_valid_lifetime; 2106 if (ia0->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) { 2107 pltime0 = IFA6_IS_DEPRECATED(ia0) ? 0 : 2108 (ia0->ia6_lifetime.ia6t_pltime - 2109 (time_uptime - ia0->ia6_updatetime)); 2110 if (pltime0 > ip6_temp_preferred_lifetime - ip6_desync_factor){ 2111 pltime0 = ip6_temp_preferred_lifetime - 2112 ip6_desync_factor; 2113 } 2114 } else 2115 pltime0 = ip6_temp_preferred_lifetime - ip6_desync_factor; 2116 ifra.ifra_lifetime.ia6t_vltime = vltime0; 2117 ifra.ifra_lifetime.ia6t_pltime = pltime0; 2118 2119 /* 2120 * A temporary address is created only if this calculated Preferred 2121 * Lifetime is greater than REGEN_ADVANCE time units. 2122 */ 2123 if (ifra.ifra_lifetime.ia6t_pltime <= ip6_temp_regen_advance) 2124 return (0); 2125 2126 /* XXX: scope zone ID? */ 2127 2128 ifra.ifra_flags |= (IN6_IFF_AUTOCONF|IN6_IFF_TEMPORARY); 2129 2130 /* allocate ifaddr structure, link into chain, etc. */ 2131 updateflags = 0; 2132 if (dad_delay) 2133 updateflags |= IN6_IFAUPDATE_DADDELAY; 2134 if ((error = in6_update_ifa(ifp, &ifra, updateflags)) != 0) 2135 return (error); 2136 2137 s = pserialize_read_enter(); 2138 newia = in6ifa_ifpwithaddr(ifp, &ifra.ifra_addr.sin6_addr); 2139 if (newia == NULL) { /* XXX: can it happen? */ 2140 pserialize_read_exit(s); 2141 nd6log(LOG_ERR, 2142 "ifa update succeeded, but we got no ifaddr\n"); 2143 return (EINVAL); /* XXX */ 2144 } 2145 newia->ia6_ndpr = ia0->ia6_ndpr; 2146 newia->ia6_ndpr->ndpr_refcnt++; 2147 pserialize_read_exit(s); 2148 2149 /* 2150 * A newly added address might affect the status of other addresses. 2151 * XXX: when the temporary address is generated with a new public 2152 * address, the onlink check is redundant. However, it would be safe 2153 * to do the check explicitly everywhere a new address is generated, 2154 * and, in fact, we surely need the check when we create a new 2155 * temporary address due to deprecation of an old temporary address. 2156 */ 2157 nd6_pfxlist_onlink_check(); 2158 2159 return (0); 2160 } 2161 2162 static int 2163 in6_init_prefix_ltimes(struct nd_prefix *ndpr) 2164 { 2165 2166 ND6_ASSERT_WLOCK(); 2167 2168 /* check if preferred lifetime > valid lifetime. RFC2462 5.5.3 (c) */ 2169 if (ndpr->ndpr_pltime > ndpr->ndpr_vltime) { 2170 nd6log(LOG_INFO, "preferred lifetime" 2171 "(%d) is greater than valid lifetime(%d)\n", 2172 (u_int)ndpr->ndpr_pltime, (u_int)ndpr->ndpr_vltime); 2173 return (EINVAL); 2174 } 2175 if (ndpr->ndpr_pltime == ND6_INFINITE_LIFETIME) 2176 ndpr->ndpr_preferred = 0; 2177 else 2178 ndpr->ndpr_preferred = time_uptime + ndpr->ndpr_pltime; 2179 if (ndpr->ndpr_vltime == ND6_INFINITE_LIFETIME) 2180 ndpr->ndpr_expire = 0; 2181 else 2182 ndpr->ndpr_expire = time_uptime + ndpr->ndpr_vltime; 2183 2184 return 0; 2185 } 2186 2187 static void 2188 in6_init_address_ltimes(struct nd_prefix *newpr, 2189 struct in6_addrlifetime *lt6) 2190 { 2191 2192 /* Valid lifetime must not be updated unless explicitly specified. */ 2193 /* init ia6t_expire */ 2194 if (lt6->ia6t_vltime == ND6_INFINITE_LIFETIME) 2195 lt6->ia6t_expire = 0; 2196 else { 2197 lt6->ia6t_expire = time_uptime; 2198 lt6->ia6t_expire += lt6->ia6t_vltime; 2199 } 2200 2201 /* init ia6t_preferred */ 2202 if (lt6->ia6t_pltime == ND6_INFINITE_LIFETIME) 2203 lt6->ia6t_preferred = 0; 2204 else { 2205 lt6->ia6t_preferred = time_uptime; 2206 lt6->ia6t_preferred += lt6->ia6t_pltime; 2207 } 2208 } 2209 2210 /* 2211 * Delete all the routing table entries that use the specified gateway. 2212 * XXX: this function causes search through all entries of routing table, so 2213 * it shouldn't be called when acting as a router. 2214 */ 2215 void 2216 nd6_rt_flush(struct in6_addr *gateway, struct ifnet *ifp) 2217 { 2218 #ifndef NET_MPSAFE 2219 int s = splsoftnet(); 2220 #endif 2221 2222 /* We'll care only link-local addresses */ 2223 if (!IN6_IS_ADDR_LINKLOCAL(gateway)) 2224 goto out; 2225 2226 rt_delete_matched_entries(AF_INET6, rt6_deleteroute_matcher, gateway); 2227 2228 out: 2229 #ifndef NET_MPSAFE 2230 splx(s); 2231 #endif 2232 return; /* XXX gcc */ 2233 } 2234 2235 static int 2236 rt6_deleteroute_matcher(struct rtentry *rt, void *arg) 2237 { 2238 struct in6_addr *gate = (struct in6_addr *)arg; 2239 2240 if (rt->rt_gateway == NULL || rt->rt_gateway->sa_family != AF_INET6) 2241 return (0); 2242 2243 if (!IN6_ARE_ADDR_EQUAL(gate, &satosin6(rt->rt_gateway)->sin6_addr)) 2244 return (0); 2245 2246 /* 2247 * Do not delete a static route. 2248 * XXX: this seems to be a bit ad-hoc. Should we consider the 2249 * 'cloned' bit instead? 2250 */ 2251 if ((rt->rt_flags & RTF_STATIC) != 0) 2252 return (0); 2253 2254 /* 2255 * We delete only host route. This means, in particular, we don't 2256 * delete default route. 2257 */ 2258 if ((rt->rt_flags & RTF_HOST) == 0) 2259 return (0); 2260 2261 return 1; 2262 } 2263