1/*- 2 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project. 3 * All rights reserved. 4 * 5 * Redistribution and use in source and binary forms, with or without 6 * modification, are permitted provided that the following conditions 7 * are met: 8 * 1. Redistributions of source code must retain the above copyright 9 * notice, this list of conditions and the following disclaimer. 10 * 2. Redistributions in binary form must reproduce the above copyright 11 * notice, this list of conditions and the following disclaimer in the 12 * documentation and/or other materials provided with the distribution. 13 * 3. Neither the name of the project nor the names of its contributors 14 * may be used to endorse or promote products derived from this software 15 * without specific prior written permission. 16 * 17 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND 18 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 19 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 20 * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE 21 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 22 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 23 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 24 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 25 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 26 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 27 * SUCH DAMAGE. 28 * 29 * $KAME: nd6_rtr.c,v 1.111 2001/04/27 01:37:15 jinmei Exp $ 30 */ 31 32#include <sys/cdefs.h> 33__FBSDID("$FreeBSD: releng/10.3/sys/netinet6/nd6_rtr.c 281230 2015-04-07 20:20:24Z delphij $"); 34 35#include "opt_inet.h" 36#include "opt_inet6.h" 37 38#include <sys/param.h> 39#include <sys/systm.h> 40#include <sys/malloc.h> 41#include <sys/mbuf.h> 42#include <sys/socket.h> 43#include <sys/sockio.h> 44#include <sys/time.h> 45#include <sys/kernel.h> 46#include <sys/lock.h> 47#include <sys/errno.h> 48#include <sys/rwlock.h> 49#include <sys/syslog.h> 50#include <sys/queue.h> 51 52#include <net/if.h> 53#include <net/if_types.h> 54#include <net/if_dl.h> 55#include <net/route.h> 56#include <net/radix.h> 57#include <net/vnet.h> 58 59#include <netinet/in.h> 60#include <net/if_llatbl.h> 61#include <netinet6/in6_var.h> 62#include <netinet6/in6_ifattach.h> 63#include <netinet/ip6.h> 64#include <netinet6/ip6_var.h> 65#include <netinet6/nd6.h> 66#include <netinet/icmp6.h> 67#include <netinet6/scope6_var.h> 68 69static int rtpref(struct nd_defrouter *); 70static struct nd_defrouter *defrtrlist_update(struct nd_defrouter *); 71static int prelist_update(struct nd_prefixctl *, struct nd_defrouter *, 72 struct mbuf *, int); 73static struct in6_ifaddr *in6_ifadd(struct nd_prefixctl *, int); 74static struct nd_pfxrouter *pfxrtr_lookup(struct nd_prefix *, 75 struct nd_defrouter *); 76static void pfxrtr_add(struct nd_prefix *, struct nd_defrouter *); 77static void pfxrtr_del(struct nd_pfxrouter *); 78static struct nd_pfxrouter *find_pfxlist_reachable_router 79(struct nd_prefix *); 80static void defrouter_delreq(struct nd_defrouter *); 81static void nd6_rtmsg(int, struct rtentry *); 82 83static int in6_init_prefix_ltimes(struct nd_prefix *); 84static void in6_init_address_ltimes(struct nd_prefix *, 85 struct in6_addrlifetime *); 86 87static int nd6_prefix_onlink(struct nd_prefix *); 88static int nd6_prefix_offlink(struct nd_prefix *); 89 90static int rt6_deleteroute(struct radix_node *, void *); 91 92VNET_DECLARE(int, nd6_recalc_reachtm_interval); 93#define V_nd6_recalc_reachtm_interval VNET(nd6_recalc_reachtm_interval) 94 95static VNET_DEFINE(struct ifnet *, nd6_defifp); 96VNET_DEFINE(int, nd6_defifindex); 97#define V_nd6_defifp VNET(nd6_defifp) 98 99VNET_DEFINE(int, ip6_use_tempaddr) = 0; 100 101VNET_DEFINE(int, ip6_desync_factor); 102VNET_DEFINE(u_int32_t, ip6_temp_preferred_lifetime) = DEF_TEMP_PREFERRED_LIFETIME; 103VNET_DEFINE(u_int32_t, ip6_temp_valid_lifetime) = DEF_TEMP_VALID_LIFETIME; 104 105VNET_DEFINE(int, ip6_temp_regen_advance) = TEMPADDR_REGEN_ADVANCE; 106 107/* RTPREF_MEDIUM has to be 0! */ 108#define RTPREF_HIGH 1 109#define RTPREF_MEDIUM 0 110#define RTPREF_LOW (-1) 111#define RTPREF_RESERVED (-2) 112#define RTPREF_INVALID (-3) /* internal */ 113 114/* 115 * Receive Router Solicitation Message - just for routers. 116 * Router solicitation/advertisement is mostly managed by userland program 117 * (rtadvd) so here we have no function like nd6_ra_output(). 118 * 119 * Based on RFC 2461 120 */ 121void 122nd6_rs_input(struct mbuf *m, int off, int icmp6len) 123{ 124 struct ifnet *ifp = m->m_pkthdr.rcvif; 125 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *); 126 struct nd_router_solicit *nd_rs; 127 struct in6_addr saddr6 = ip6->ip6_src; 128 char *lladdr = NULL; 129 int lladdrlen = 0; 130 union nd_opts ndopts; 131 char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN]; 132 133 /* 134 * Accept RS only when V_ip6_forwarding=1 and the interface has 135 * no ND6_IFF_ACCEPT_RTADV. 136 */ 137 if (!V_ip6_forwarding || ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV) 138 goto freeit; 139 140 /* Sanity checks */ 141 if (ip6->ip6_hlim != 255) { 142 nd6log((LOG_ERR, 143 "nd6_rs_input: invalid hlim (%d) from %s to %s on %s\n", 144 ip6->ip6_hlim, ip6_sprintf(ip6bufs, &ip6->ip6_src), 145 ip6_sprintf(ip6bufd, &ip6->ip6_dst), if_name(ifp))); 146 goto bad; 147 } 148 149 /* 150 * Don't update the neighbor cache, if src = ::. 151 * This indicates that the src has no IP address assigned yet. 152 */ 153 if (IN6_IS_ADDR_UNSPECIFIED(&saddr6)) 154 goto freeit; 155 156#ifndef PULLDOWN_TEST 157 IP6_EXTHDR_CHECK(m, off, icmp6len,); 158 nd_rs = (struct nd_router_solicit *)((caddr_t)ip6 + off); 159#else 160 IP6_EXTHDR_GET(nd_rs, struct nd_router_solicit *, m, off, icmp6len); 161 if (nd_rs == NULL) { 162 ICMP6STAT_INC(icp6s_tooshort); 163 return; 164 } 165#endif 166 167 icmp6len -= sizeof(*nd_rs); 168 nd6_option_init(nd_rs + 1, icmp6len, &ndopts); 169 if (nd6_options(&ndopts) < 0) { 170 nd6log((LOG_INFO, 171 "nd6_rs_input: invalid ND option, ignored\n")); 172 /* nd6_options have incremented stats */ 173 goto freeit; 174 } 175 176 if (ndopts.nd_opts_src_lladdr) { 177 lladdr = (char *)(ndopts.nd_opts_src_lladdr + 1); 178 lladdrlen = ndopts.nd_opts_src_lladdr->nd_opt_len << 3; 179 } 180 181 if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) { 182 nd6log((LOG_INFO, 183 "nd6_rs_input: lladdrlen mismatch for %s " 184 "(if %d, RS packet %d)\n", 185 ip6_sprintf(ip6bufs, &saddr6), 186 ifp->if_addrlen, lladdrlen - 2)); 187 goto bad; 188 } 189 190 nd6_cache_lladdr(ifp, &saddr6, lladdr, lladdrlen, ND_ROUTER_SOLICIT, 0); 191 192 freeit: 193 m_freem(m); 194 return; 195 196 bad: 197 ICMP6STAT_INC(icp6s_badrs); 198 m_freem(m); 199} 200 201/* 202 * Receive Router Advertisement Message. 203 * 204 * Based on RFC 2461 205 * TODO: on-link bit on prefix information 206 * TODO: ND_RA_FLAG_{OTHER,MANAGED} processing 207 */ 208void 209nd6_ra_input(struct mbuf *m, int off, int icmp6len) 210{ 211 struct ifnet *ifp = m->m_pkthdr.rcvif; 212 struct nd_ifinfo *ndi = ND_IFINFO(ifp); 213 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *); 214 struct nd_router_advert *nd_ra; 215 struct in6_addr saddr6 = ip6->ip6_src; 216 int mcast = 0; 217 union nd_opts ndopts; 218 struct nd_defrouter *dr; 219 char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN]; 220 221 /* 222 * We only accept RAs only when the per-interface flag 223 * ND6_IFF_ACCEPT_RTADV is on the receiving interface. 224 */ 225 if (!(ndi->flags & ND6_IFF_ACCEPT_RTADV)) 226 goto freeit; 227 228 if (ip6->ip6_hlim != 255) { 229 nd6log((LOG_ERR, 230 "nd6_ra_input: invalid hlim (%d) from %s to %s on %s\n", 231 ip6->ip6_hlim, ip6_sprintf(ip6bufs, &ip6->ip6_src), 232 ip6_sprintf(ip6bufd, &ip6->ip6_dst), if_name(ifp))); 233 goto bad; 234 } 235 236 if (!IN6_IS_ADDR_LINKLOCAL(&saddr6)) { 237 nd6log((LOG_ERR, 238 "nd6_ra_input: src %s is not link-local\n", 239 ip6_sprintf(ip6bufs, &saddr6))); 240 goto bad; 241 } 242 243#ifndef PULLDOWN_TEST 244 IP6_EXTHDR_CHECK(m, off, icmp6len,); 245 nd_ra = (struct nd_router_advert *)((caddr_t)ip6 + off); 246#else 247 IP6_EXTHDR_GET(nd_ra, struct nd_router_advert *, m, off, icmp6len); 248 if (nd_ra == NULL) { 249 ICMP6STAT_INC(icp6s_tooshort); 250 return; 251 } 252#endif 253 254 icmp6len -= sizeof(*nd_ra); 255 nd6_option_init(nd_ra + 1, icmp6len, &ndopts); 256 if (nd6_options(&ndopts) < 0) { 257 nd6log((LOG_INFO, 258 "nd6_ra_input: invalid ND option, ignored\n")); 259 /* nd6_options have incremented stats */ 260 goto freeit; 261 } 262 263 { 264 struct nd_defrouter dr0; 265 u_int32_t advreachable = nd_ra->nd_ra_reachable; 266 267 /* remember if this is a multicasted advertisement */ 268 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) 269 mcast = 1; 270 271 bzero(&dr0, sizeof(dr0)); 272 dr0.rtaddr = saddr6; 273 dr0.flags = nd_ra->nd_ra_flags_reserved; 274 /* 275 * Effectively-disable routes from RA messages when 276 * ND6_IFF_NO_RADR enabled on the receiving interface or 277 * (ip6.forwarding == 1 && ip6.rfc6204w3 != 1). 278 */ 279 if (ndi->flags & ND6_IFF_NO_RADR) 280 dr0.rtlifetime = 0; 281 else if (V_ip6_forwarding && !V_ip6_rfc6204w3) 282 dr0.rtlifetime = 0; 283 else 284 dr0.rtlifetime = ntohs(nd_ra->nd_ra_router_lifetime); 285 dr0.expire = time_uptime + dr0.rtlifetime; 286 dr0.ifp = ifp; 287 /* unspecified or not? (RFC 2461 6.3.4) */ 288 if (advreachable) { 289 advreachable = ntohl(advreachable); 290 if (advreachable <= MAX_REACHABLE_TIME && 291 ndi->basereachable != advreachable) { 292 ndi->basereachable = advreachable; 293 ndi->reachable = ND_COMPUTE_RTIME(ndi->basereachable); 294 ndi->recalctm = V_nd6_recalc_reachtm_interval; /* reset */ 295 } 296 } 297 if (nd_ra->nd_ra_retransmit) 298 ndi->retrans = ntohl(nd_ra->nd_ra_retransmit); 299 if (nd_ra->nd_ra_curhoplimit) { 300 if (ndi->chlim < nd_ra->nd_ra_curhoplimit) 301 ndi->chlim = nd_ra->nd_ra_curhoplimit; 302 else if (ndi->chlim != nd_ra->nd_ra_curhoplimit) { 303 log(LOG_ERR, "RA with a lower CurHopLimit sent from " 304 "%s on %s (current = %d, received = %d). " 305 "Ignored.\n", ip6_sprintf(ip6bufs, &ip6->ip6_src), 306 if_name(ifp), ndi->chlim, nd_ra->nd_ra_curhoplimit); 307 } 308 } 309 dr = defrtrlist_update(&dr0); 310 } 311 312 /* 313 * prefix 314 */ 315 if (ndopts.nd_opts_pi) { 316 struct nd_opt_hdr *pt; 317 struct nd_opt_prefix_info *pi = NULL; 318 struct nd_prefixctl pr; 319 320 for (pt = (struct nd_opt_hdr *)ndopts.nd_opts_pi; 321 pt <= (struct nd_opt_hdr *)ndopts.nd_opts_pi_end; 322 pt = (struct nd_opt_hdr *)((caddr_t)pt + 323 (pt->nd_opt_len << 3))) { 324 if (pt->nd_opt_type != ND_OPT_PREFIX_INFORMATION) 325 continue; 326 pi = (struct nd_opt_prefix_info *)pt; 327 328 if (pi->nd_opt_pi_len != 4) { 329 nd6log((LOG_INFO, 330 "nd6_ra_input: invalid option " 331 "len %d for prefix information option, " 332 "ignored\n", pi->nd_opt_pi_len)); 333 continue; 334 } 335 336 if (128 < pi->nd_opt_pi_prefix_len) { 337 nd6log((LOG_INFO, 338 "nd6_ra_input: invalid prefix " 339 "len %d for prefix information option, " 340 "ignored\n", pi->nd_opt_pi_prefix_len)); 341 continue; 342 } 343 344 if (IN6_IS_ADDR_MULTICAST(&pi->nd_opt_pi_prefix) 345 || IN6_IS_ADDR_LINKLOCAL(&pi->nd_opt_pi_prefix)) { 346 nd6log((LOG_INFO, 347 "nd6_ra_input: invalid prefix " 348 "%s, ignored\n", 349 ip6_sprintf(ip6bufs, 350 &pi->nd_opt_pi_prefix))); 351 continue; 352 } 353 354 bzero(&pr, sizeof(pr)); 355 pr.ndpr_prefix.sin6_family = AF_INET6; 356 pr.ndpr_prefix.sin6_len = sizeof(pr.ndpr_prefix); 357 pr.ndpr_prefix.sin6_addr = pi->nd_opt_pi_prefix; 358 pr.ndpr_ifp = (struct ifnet *)m->m_pkthdr.rcvif; 359 360 pr.ndpr_raf_onlink = (pi->nd_opt_pi_flags_reserved & 361 ND_OPT_PI_FLAG_ONLINK) ? 1 : 0; 362 pr.ndpr_raf_auto = (pi->nd_opt_pi_flags_reserved & 363 ND_OPT_PI_FLAG_AUTO) ? 1 : 0; 364 pr.ndpr_plen = pi->nd_opt_pi_prefix_len; 365 pr.ndpr_vltime = ntohl(pi->nd_opt_pi_valid_time); 366 pr.ndpr_pltime = ntohl(pi->nd_opt_pi_preferred_time); 367 (void)prelist_update(&pr, dr, m, mcast); 368 } 369 } 370 371 /* 372 * MTU 373 */ 374 if (ndopts.nd_opts_mtu && ndopts.nd_opts_mtu->nd_opt_mtu_len == 1) { 375 u_long mtu; 376 u_long maxmtu; 377 378 mtu = (u_long)ntohl(ndopts.nd_opts_mtu->nd_opt_mtu_mtu); 379 380 /* lower bound */ 381 if (mtu < IPV6_MMTU) { 382 nd6log((LOG_INFO, "nd6_ra_input: bogus mtu option " 383 "mtu=%lu sent from %s, ignoring\n", 384 mtu, ip6_sprintf(ip6bufs, &ip6->ip6_src))); 385 goto skip; 386 } 387 388 /* upper bound */ 389 maxmtu = (ndi->maxmtu && ndi->maxmtu < ifp->if_mtu) 390 ? ndi->maxmtu : ifp->if_mtu; 391 if (mtu <= maxmtu) { 392 int change = (ndi->linkmtu != mtu); 393 394 ndi->linkmtu = mtu; 395 if (change) /* in6_maxmtu may change */ 396 in6_setmaxmtu(); 397 } else { 398 nd6log((LOG_INFO, "nd6_ra_input: bogus mtu " 399 "mtu=%lu sent from %s; " 400 "exceeds maxmtu %lu, ignoring\n", 401 mtu, ip6_sprintf(ip6bufs, &ip6->ip6_src), maxmtu)); 402 } 403 } 404 405 skip: 406 407 /* 408 * Source link layer address 409 */ 410 { 411 char *lladdr = NULL; 412 int lladdrlen = 0; 413 414 if (ndopts.nd_opts_src_lladdr) { 415 lladdr = (char *)(ndopts.nd_opts_src_lladdr + 1); 416 lladdrlen = ndopts.nd_opts_src_lladdr->nd_opt_len << 3; 417 } 418 419 if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) { 420 nd6log((LOG_INFO, 421 "nd6_ra_input: lladdrlen mismatch for %s " 422 "(if %d, RA packet %d)\n", ip6_sprintf(ip6bufs, &saddr6), 423 ifp->if_addrlen, lladdrlen - 2)); 424 goto bad; 425 } 426 427 nd6_cache_lladdr(ifp, &saddr6, lladdr, 428 lladdrlen, ND_ROUTER_ADVERT, 0); 429 430 /* 431 * Installing a link-layer address might change the state of the 432 * router's neighbor cache, which might also affect our on-link 433 * detection of adveritsed prefixes. 434 */ 435 pfxlist_onlink_check(); 436 } 437 438 freeit: 439 m_freem(m); 440 return; 441 442 bad: 443 ICMP6STAT_INC(icp6s_badra); 444 m_freem(m); 445} 446 447/* 448 * default router list proccessing sub routines 449 */ 450 451/* tell the change to user processes watching the routing socket. */ 452static void 453nd6_rtmsg(int cmd, struct rtentry *rt) 454{ 455 struct rt_addrinfo info; 456 struct ifnet *ifp; 457 struct ifaddr *ifa; 458 459 bzero((caddr_t)&info, sizeof(info)); 460 info.rti_info[RTAX_DST] = rt_key(rt); 461 info.rti_info[RTAX_GATEWAY] = rt->rt_gateway; 462 info.rti_info[RTAX_NETMASK] = rt_mask(rt); 463 ifp = rt->rt_ifp; 464 if (ifp != NULL) { 465 IF_ADDR_RLOCK(ifp); 466 ifa = TAILQ_FIRST(&ifp->if_addrhead); 467 info.rti_info[RTAX_IFP] = ifa->ifa_addr; 468 ifa_ref(ifa); 469 IF_ADDR_RUNLOCK(ifp); 470 info.rti_info[RTAX_IFA] = rt->rt_ifa->ifa_addr; 471 } else 472 ifa = NULL; 473 474 rt_missmsg_fib(cmd, &info, rt->rt_flags, 0, rt->rt_fibnum); 475 if (ifa != NULL) 476 ifa_free(ifa); 477} 478 479static void 480defrouter_addreq(struct nd_defrouter *new) 481{ 482 struct sockaddr_in6 def, mask, gate; 483 struct rtentry *newrt = NULL; 484 int error; 485 486 bzero(&def, sizeof(def)); 487 bzero(&mask, sizeof(mask)); 488 bzero(&gate, sizeof(gate)); 489 490 def.sin6_len = mask.sin6_len = gate.sin6_len = 491 sizeof(struct sockaddr_in6); 492 def.sin6_family = gate.sin6_family = AF_INET6; 493 gate.sin6_addr = new->rtaddr; 494 495 error = in6_rtrequest(RTM_ADD, (struct sockaddr *)&def, 496 (struct sockaddr *)&gate, (struct sockaddr *)&mask, 497 RTF_GATEWAY, &newrt, RT_DEFAULT_FIB); 498 if (newrt) { 499 nd6_rtmsg(RTM_ADD, newrt); /* tell user process */ 500 RTFREE(newrt); 501 } 502 if (error == 0) 503 new->installed = 1; 504 return; 505} 506 507struct nd_defrouter * 508defrouter_lookup(struct in6_addr *addr, struct ifnet *ifp) 509{ 510 struct nd_defrouter *dr; 511 512 TAILQ_FOREACH(dr, &V_nd_defrouter, dr_entry) { 513 if (dr->ifp == ifp && IN6_ARE_ADDR_EQUAL(addr, &dr->rtaddr)) 514 return (dr); 515 } 516 517 return (NULL); /* search failed */ 518} 519 520/* 521 * Remove the default route for a given router. 522 * This is just a subroutine function for defrouter_select(), and should 523 * not be called from anywhere else. 524 */ 525static void 526defrouter_delreq(struct nd_defrouter *dr) 527{ 528 struct sockaddr_in6 def, mask, gate; 529 struct rtentry *oldrt = NULL; 530 531 bzero(&def, sizeof(def)); 532 bzero(&mask, sizeof(mask)); 533 bzero(&gate, sizeof(gate)); 534 535 def.sin6_len = mask.sin6_len = gate.sin6_len = 536 sizeof(struct sockaddr_in6); 537 def.sin6_family = gate.sin6_family = AF_INET6; 538 gate.sin6_addr = dr->rtaddr; 539 540 in6_rtrequest(RTM_DELETE, (struct sockaddr *)&def, 541 (struct sockaddr *)&gate, 542 (struct sockaddr *)&mask, RTF_GATEWAY, &oldrt, RT_DEFAULT_FIB); 543 if (oldrt) { 544 nd6_rtmsg(RTM_DELETE, oldrt); 545 RTFREE(oldrt); 546 } 547 548 dr->installed = 0; 549} 550 551/* 552 * remove all default routes from default router list 553 */ 554void 555defrouter_reset(void) 556{ 557 struct nd_defrouter *dr; 558 559 TAILQ_FOREACH(dr, &V_nd_defrouter, dr_entry) 560 defrouter_delreq(dr); 561 562 /* 563 * XXX should we also nuke any default routers in the kernel, by 564 * going through them by rtalloc1()? 565 */ 566} 567 568void 569defrtrlist_del(struct nd_defrouter *dr) 570{ 571 struct nd_defrouter *deldr = NULL; 572 struct nd_prefix *pr; 573 574 /* 575 * Flush all the routing table entries that use the router 576 * as a next hop. 577 */ 578 if (ND_IFINFO(dr->ifp)->flags & ND6_IFF_ACCEPT_RTADV) 579 rt6_flush(&dr->rtaddr, dr->ifp); 580 581 if (dr->installed) { 582 deldr = dr; 583 defrouter_delreq(dr); 584 } 585 TAILQ_REMOVE(&V_nd_defrouter, dr, dr_entry); 586 587 /* 588 * Also delete all the pointers to the router in each prefix lists. 589 */ 590 LIST_FOREACH(pr, &V_nd_prefix, ndpr_entry) { 591 struct nd_pfxrouter *pfxrtr; 592 if ((pfxrtr = pfxrtr_lookup(pr, dr)) != NULL) 593 pfxrtr_del(pfxrtr); 594 } 595 pfxlist_onlink_check(); 596 597 /* 598 * If the router is the primary one, choose a new one. 599 * Note that defrouter_select() will remove the current gateway 600 * from the routing table. 601 */ 602 if (deldr) 603 defrouter_select(); 604 605 free(dr, M_IP6NDP); 606} 607 608/* 609 * Default Router Selection according to Section 6.3.6 of RFC 2461 and 610 * draft-ietf-ipngwg-router-selection: 611 * 1) Routers that are reachable or probably reachable should be preferred. 612 * If we have more than one (probably) reachable router, prefer ones 613 * with the highest router preference. 614 * 2) When no routers on the list are known to be reachable or 615 * probably reachable, routers SHOULD be selected in a round-robin 616 * fashion, regardless of router preference values. 617 * 3) If the Default Router List is empty, assume that all 618 * destinations are on-link. 619 * 620 * We assume nd_defrouter is sorted by router preference value. 621 * Since the code below covers both with and without router preference cases, 622 * we do not need to classify the cases by ifdef. 623 * 624 * At this moment, we do not try to install more than one default router, 625 * even when the multipath routing is available, because we're not sure about 626 * the benefits for stub hosts comparing to the risk of making the code 627 * complicated and the possibility of introducing bugs. 628 */ 629void 630defrouter_select(void) 631{ 632 struct nd_defrouter *dr, *selected_dr = NULL, *installed_dr = NULL; 633 struct llentry *ln = NULL; 634 635 /* 636 * Let's handle easy case (3) first: 637 * If default router list is empty, there's nothing to be done. 638 */ 639 if (TAILQ_EMPTY(&V_nd_defrouter)) 640 return; 641 642 /* 643 * Search for a (probably) reachable router from the list. 644 * We just pick up the first reachable one (if any), assuming that 645 * the ordering rule of the list described in defrtrlist_update(). 646 */ 647 TAILQ_FOREACH(dr, &V_nd_defrouter, dr_entry) { 648 IF_AFDATA_RLOCK(dr->ifp); 649 if (selected_dr == NULL && 650 (ln = nd6_lookup(&dr->rtaddr, 0, dr->ifp)) && 651 ND6_IS_LLINFO_PROBREACH(ln)) { 652 selected_dr = dr; 653 } 654 IF_AFDATA_RUNLOCK(dr->ifp); 655 if (ln != NULL) { 656 LLE_RUNLOCK(ln); 657 ln = NULL; 658 } 659 660 if (dr->installed && installed_dr == NULL) 661 installed_dr = dr; 662 else if (dr->installed && installed_dr) { 663 /* this should not happen. warn for diagnosis. */ 664 log(LOG_ERR, "defrouter_select: more than one router" 665 " is installed\n"); 666 } 667 } 668 /* 669 * If none of the default routers was found to be reachable, 670 * round-robin the list regardless of preference. 671 * Otherwise, if we have an installed router, check if the selected 672 * (reachable) router should really be preferred to the installed one. 673 * We only prefer the new router when the old one is not reachable 674 * or when the new one has a really higher preference value. 675 */ 676 if (selected_dr == NULL) { 677 if (installed_dr == NULL || !TAILQ_NEXT(installed_dr, dr_entry)) 678 selected_dr = TAILQ_FIRST(&V_nd_defrouter); 679 else 680 selected_dr = TAILQ_NEXT(installed_dr, dr_entry); 681 } else if (installed_dr) { 682 IF_AFDATA_RLOCK(installed_dr->ifp); 683 if ((ln = nd6_lookup(&installed_dr->rtaddr, 0, installed_dr->ifp)) && 684 ND6_IS_LLINFO_PROBREACH(ln) && 685 rtpref(selected_dr) <= rtpref(installed_dr)) { 686 selected_dr = installed_dr; 687 } 688 IF_AFDATA_RUNLOCK(installed_dr->ifp); 689 if (ln != NULL) 690 LLE_RUNLOCK(ln); 691 } 692 693 /* 694 * If the selected router is different than the installed one, 695 * remove the installed router and install the selected one. 696 * Note that the selected router is never NULL here. 697 */ 698 if (installed_dr != selected_dr) { 699 if (installed_dr) 700 defrouter_delreq(installed_dr); 701 defrouter_addreq(selected_dr); 702 } 703 704 return; 705} 706 707/* 708 * for default router selection 709 * regards router-preference field as a 2-bit signed integer 710 */ 711static int 712rtpref(struct nd_defrouter *dr) 713{ 714 switch (dr->flags & ND_RA_FLAG_RTPREF_MASK) { 715 case ND_RA_FLAG_RTPREF_HIGH: 716 return (RTPREF_HIGH); 717 case ND_RA_FLAG_RTPREF_MEDIUM: 718 case ND_RA_FLAG_RTPREF_RSV: 719 return (RTPREF_MEDIUM); 720 case ND_RA_FLAG_RTPREF_LOW: 721 return (RTPREF_LOW); 722 default: 723 /* 724 * This case should never happen. If it did, it would mean a 725 * serious bug of kernel internal. We thus always bark here. 726 * Or, can we even panic? 727 */ 728 log(LOG_ERR, "rtpref: impossible RA flag %x\n", dr->flags); 729 return (RTPREF_INVALID); 730 } 731 /* NOTREACHED */ 732} 733 734static struct nd_defrouter * 735defrtrlist_update(struct nd_defrouter *new) 736{ 737 struct nd_defrouter *dr, *n; 738 739 if ((dr = defrouter_lookup(&new->rtaddr, new->ifp)) != NULL) { 740 /* entry exists */ 741 if (new->rtlifetime == 0) { 742 defrtrlist_del(dr); 743 dr = NULL; 744 } else { 745 int oldpref = rtpref(dr); 746 747 /* override */ 748 dr->flags = new->flags; /* xxx flag check */ 749 dr->rtlifetime = new->rtlifetime; 750 dr->expire = new->expire; 751 752 /* 753 * If the preference does not change, there's no need 754 * to sort the entries. Also make sure the selected 755 * router is still installed in the kernel. 756 */ 757 if (dr->installed && rtpref(new) == oldpref) 758 return (dr); 759 760 /* 761 * preferred router may be changed, so relocate 762 * this router. 763 * XXX: calling TAILQ_REMOVE directly is a bad manner. 764 * However, since defrtrlist_del() has many side 765 * effects, we intentionally do so here. 766 * defrouter_select() below will handle routing 767 * changes later. 768 */ 769 TAILQ_REMOVE(&V_nd_defrouter, dr, dr_entry); 770 n = dr; 771 goto insert; 772 } 773 return (dr); 774 } 775 776 /* entry does not exist */ 777 if (new->rtlifetime == 0) 778 return (NULL); 779 780 n = (struct nd_defrouter *)malloc(sizeof(*n), M_IP6NDP, M_NOWAIT); 781 if (n == NULL) 782 return (NULL); 783 bzero(n, sizeof(*n)); 784 *n = *new; 785 786insert: 787 /* 788 * Insert the new router in the Default Router List; 789 * The Default Router List should be in the descending order 790 * of router-preferece. Routers with the same preference are 791 * sorted in the arriving time order. 792 */ 793 794 /* insert at the end of the group */ 795 TAILQ_FOREACH(dr, &V_nd_defrouter, dr_entry) { 796 if (rtpref(n) > rtpref(dr)) 797 break; 798 } 799 if (dr) 800 TAILQ_INSERT_BEFORE(dr, n, dr_entry); 801 else 802 TAILQ_INSERT_TAIL(&V_nd_defrouter, n, dr_entry); 803 804 defrouter_select(); 805 806 return (n); 807} 808 809static struct nd_pfxrouter * 810pfxrtr_lookup(struct nd_prefix *pr, struct nd_defrouter *dr) 811{ 812 struct nd_pfxrouter *search; 813 814 LIST_FOREACH(search, &pr->ndpr_advrtrs, pfr_entry) { 815 if (search->router == dr) 816 break; 817 } 818 819 return (search); 820} 821 822static void 823pfxrtr_add(struct nd_prefix *pr, struct nd_defrouter *dr) 824{ 825 struct nd_pfxrouter *new; 826 827 new = (struct nd_pfxrouter *)malloc(sizeof(*new), M_IP6NDP, M_NOWAIT); 828 if (new == NULL) 829 return; 830 bzero(new, sizeof(*new)); 831 new->router = dr; 832 833 LIST_INSERT_HEAD(&pr->ndpr_advrtrs, new, pfr_entry); 834 835 pfxlist_onlink_check(); 836} 837 838static void 839pfxrtr_del(struct nd_pfxrouter *pfr) 840{ 841 LIST_REMOVE(pfr, pfr_entry); 842 free(pfr, M_IP6NDP); 843} 844 845struct nd_prefix * 846nd6_prefix_lookup(struct nd_prefixctl *key) 847{ 848 struct nd_prefix *search; 849 850 LIST_FOREACH(search, &V_nd_prefix, ndpr_entry) { 851 if (key->ndpr_ifp == search->ndpr_ifp && 852 key->ndpr_plen == search->ndpr_plen && 853 in6_are_prefix_equal(&key->ndpr_prefix.sin6_addr, 854 &search->ndpr_prefix.sin6_addr, key->ndpr_plen)) { 855 break; 856 } 857 } 858 859 return (search); 860} 861 862int 863nd6_prelist_add(struct nd_prefixctl *pr, struct nd_defrouter *dr, 864 struct nd_prefix **newp) 865{ 866 struct nd_prefix *new = NULL; 867 int error = 0; 868 int i; 869 char ip6buf[INET6_ADDRSTRLEN]; 870 871 new = (struct nd_prefix *)malloc(sizeof(*new), M_IP6NDP, M_NOWAIT); 872 if (new == NULL) 873 return(ENOMEM); 874 bzero(new, sizeof(*new)); 875 new->ndpr_ifp = pr->ndpr_ifp; 876 new->ndpr_prefix = pr->ndpr_prefix; 877 new->ndpr_plen = pr->ndpr_plen; 878 new->ndpr_vltime = pr->ndpr_vltime; 879 new->ndpr_pltime = pr->ndpr_pltime; 880 new->ndpr_flags = pr->ndpr_flags; 881 if ((error = in6_init_prefix_ltimes(new)) != 0) { 882 free(new, M_IP6NDP); 883 return(error); 884 } 885 new->ndpr_lastupdate = time_uptime; 886 if (newp != NULL) 887 *newp = new; 888 889 /* initialization */ 890 LIST_INIT(&new->ndpr_advrtrs); 891 in6_prefixlen2mask(&new->ndpr_mask, new->ndpr_plen); 892 /* make prefix in the canonical form */ 893 for (i = 0; i < 4; i++) 894 new->ndpr_prefix.sin6_addr.s6_addr32[i] &= 895 new->ndpr_mask.s6_addr32[i]; 896 897 /* link ndpr_entry to nd_prefix list */ 898 LIST_INSERT_HEAD(&V_nd_prefix, new, ndpr_entry); 899 900 /* ND_OPT_PI_FLAG_ONLINK processing */ 901 if (new->ndpr_raf_onlink) { 902 int e; 903 904 if ((e = nd6_prefix_onlink(new)) != 0) { 905 nd6log((LOG_ERR, "nd6_prelist_add: failed to make " 906 "the prefix %s/%d on-link on %s (errno=%d)\n", 907 ip6_sprintf(ip6buf, &pr->ndpr_prefix.sin6_addr), 908 pr->ndpr_plen, if_name(pr->ndpr_ifp), e)); 909 /* proceed anyway. XXX: is it correct? */ 910 } 911 } 912 913 if (dr) 914 pfxrtr_add(new, dr); 915 916 return 0; 917} 918 919void 920prelist_remove(struct nd_prefix *pr) 921{ 922 struct nd_pfxrouter *pfr, *next; 923 int e; 924 char ip6buf[INET6_ADDRSTRLEN]; 925 926 /* make sure to invalidate the prefix until it is really freed. */ 927 pr->ndpr_vltime = 0; 928 pr->ndpr_pltime = 0; 929 930 /* 931 * Though these flags are now meaningless, we'd rather keep the value 932 * of pr->ndpr_raf_onlink and pr->ndpr_raf_auto not to confuse users 933 * when executing "ndp -p". 934 */ 935 936 if ((pr->ndpr_stateflags & NDPRF_ONLINK) != 0 && 937 (e = nd6_prefix_offlink(pr)) != 0) { 938 nd6log((LOG_ERR, "prelist_remove: failed to make %s/%d offlink " 939 "on %s, errno=%d\n", 940 ip6_sprintf(ip6buf, &pr->ndpr_prefix.sin6_addr), 941 pr->ndpr_plen, if_name(pr->ndpr_ifp), e)); 942 /* what should we do? */ 943 } 944 945 if (pr->ndpr_refcnt > 0) 946 return; /* notice here? */ 947 948 /* unlink ndpr_entry from nd_prefix list */ 949 LIST_REMOVE(pr, ndpr_entry); 950 951 /* free list of routers that adversed the prefix */ 952 LIST_FOREACH_SAFE(pfr, &pr->ndpr_advrtrs, pfr_entry, next) { 953 free(pfr, M_IP6NDP); 954 } 955 free(pr, M_IP6NDP); 956 957 pfxlist_onlink_check(); 958} 959 960/* 961 * dr - may be NULL 962 */ 963 964static int 965prelist_update(struct nd_prefixctl *new, struct nd_defrouter *dr, 966 struct mbuf *m, int mcast) 967{ 968 struct in6_ifaddr *ia6 = NULL, *ia6_match = NULL; 969 struct ifaddr *ifa; 970 struct ifnet *ifp = new->ndpr_ifp; 971 struct nd_prefix *pr; 972 int error = 0; 973 int newprefix = 0; 974 int auth; 975 struct in6_addrlifetime lt6_tmp; 976 char ip6buf[INET6_ADDRSTRLEN]; 977 978 auth = 0; 979 if (m) { 980 /* 981 * Authenticity for NA consists authentication for 982 * both IP header and IP datagrams, doesn't it ? 983 */ 984#if defined(M_AUTHIPHDR) && defined(M_AUTHIPDGM) 985 auth = ((m->m_flags & M_AUTHIPHDR) && 986 (m->m_flags & M_AUTHIPDGM)); 987#endif 988 } 989 990 if ((pr = nd6_prefix_lookup(new)) != NULL) { 991 /* 992 * nd6_prefix_lookup() ensures that pr and new have the same 993 * prefix on a same interface. 994 */ 995 996 /* 997 * Update prefix information. Note that the on-link (L) bit 998 * and the autonomous (A) bit should NOT be changed from 1 999 * to 0. 1000 */ 1001 if (new->ndpr_raf_onlink == 1) 1002 pr->ndpr_raf_onlink = 1; 1003 if (new->ndpr_raf_auto == 1) 1004 pr->ndpr_raf_auto = 1; 1005 if (new->ndpr_raf_onlink) { 1006 pr->ndpr_vltime = new->ndpr_vltime; 1007 pr->ndpr_pltime = new->ndpr_pltime; 1008 (void)in6_init_prefix_ltimes(pr); /* XXX error case? */ 1009 pr->ndpr_lastupdate = time_uptime; 1010 } 1011 1012 if (new->ndpr_raf_onlink && 1013 (pr->ndpr_stateflags & NDPRF_ONLINK) == 0) { 1014 int e; 1015 1016 if ((e = nd6_prefix_onlink(pr)) != 0) { 1017 nd6log((LOG_ERR, 1018 "prelist_update: failed to make " 1019 "the prefix %s/%d on-link on %s " 1020 "(errno=%d)\n", 1021 ip6_sprintf(ip6buf, 1022 &pr->ndpr_prefix.sin6_addr), 1023 pr->ndpr_plen, if_name(pr->ndpr_ifp), e)); 1024 /* proceed anyway. XXX: is it correct? */ 1025 } 1026 } 1027 1028 if (dr && pfxrtr_lookup(pr, dr) == NULL) 1029 pfxrtr_add(pr, dr); 1030 } else { 1031 struct nd_prefix *newpr = NULL; 1032 1033 newprefix = 1; 1034 1035 if (new->ndpr_vltime == 0) 1036 goto end; 1037 if (new->ndpr_raf_onlink == 0 && new->ndpr_raf_auto == 0) 1038 goto end; 1039 1040 error = nd6_prelist_add(new, dr, &newpr); 1041 if (error != 0 || newpr == NULL) { 1042 nd6log((LOG_NOTICE, "prelist_update: " 1043 "nd6_prelist_add failed for %s/%d on %s " 1044 "errno=%d, returnpr=%p\n", 1045 ip6_sprintf(ip6buf, &new->ndpr_prefix.sin6_addr), 1046 new->ndpr_plen, if_name(new->ndpr_ifp), 1047 error, newpr)); 1048 goto end; /* we should just give up in this case. */ 1049 } 1050 1051 /* 1052 * XXX: from the ND point of view, we can ignore a prefix 1053 * with the on-link bit being zero. However, we need a 1054 * prefix structure for references from autoconfigured 1055 * addresses. Thus, we explicitly make sure that the prefix 1056 * itself expires now. 1057 */ 1058 if (newpr->ndpr_raf_onlink == 0) { 1059 newpr->ndpr_vltime = 0; 1060 newpr->ndpr_pltime = 0; 1061 in6_init_prefix_ltimes(newpr); 1062 } 1063 1064 pr = newpr; 1065 } 1066 1067 /* 1068 * Address autoconfiguration based on Section 5.5.3 of RFC 2462. 1069 * Note that pr must be non NULL at this point. 1070 */ 1071 1072 /* 5.5.3 (a). Ignore the prefix without the A bit set. */ 1073 if (!new->ndpr_raf_auto) 1074 goto end; 1075 1076 /* 1077 * 5.5.3 (b). the link-local prefix should have been ignored in 1078 * nd6_ra_input. 1079 */ 1080 1081 /* 5.5.3 (c). Consistency check on lifetimes: pltime <= vltime. */ 1082 if (new->ndpr_pltime > new->ndpr_vltime) { 1083 error = EINVAL; /* XXX: won't be used */ 1084 goto end; 1085 } 1086 1087 /* 1088 * 5.5.3 (d). If the prefix advertised is not equal to the prefix of 1089 * an address configured by stateless autoconfiguration already in the 1090 * list of addresses associated with the interface, and the Valid 1091 * Lifetime is not 0, form an address. We first check if we have 1092 * a matching prefix. 1093 * Note: we apply a clarification in rfc2462bis-02 here. We only 1094 * consider autoconfigured addresses while RFC2462 simply said 1095 * "address". 1096 */ 1097 IF_ADDR_RLOCK(ifp); 1098 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1099 struct in6_ifaddr *ifa6; 1100 u_int32_t remaininglifetime; 1101 1102 if (ifa->ifa_addr->sa_family != AF_INET6) 1103 continue; 1104 1105 ifa6 = (struct in6_ifaddr *)ifa; 1106 1107 /* 1108 * We only consider autoconfigured addresses as per rfc2462bis. 1109 */ 1110 if (!(ifa6->ia6_flags & IN6_IFF_AUTOCONF)) 1111 continue; 1112 1113 /* 1114 * Spec is not clear here, but I believe we should concentrate 1115 * on unicast (i.e. not anycast) addresses. 1116 * XXX: other ia6_flags? detached or duplicated? 1117 */ 1118 if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0) 1119 continue; 1120 1121 /* 1122 * Ignore the address if it is not associated with a prefix 1123 * or is associated with a prefix that is different from this 1124 * one. (pr is never NULL here) 1125 */ 1126 if (ifa6->ia6_ndpr != pr) 1127 continue; 1128 1129 if (ia6_match == NULL) /* remember the first one */ 1130 ia6_match = ifa6; 1131 1132 /* 1133 * An already autoconfigured address matched. Now that we 1134 * are sure there is at least one matched address, we can 1135 * proceed to 5.5.3. (e): update the lifetimes according to the 1136 * "two hours" rule and the privacy extension. 1137 * We apply some clarifications in rfc2462bis: 1138 * - use remaininglifetime instead of storedlifetime as a 1139 * variable name 1140 * - remove the dead code in the "two-hour" rule 1141 */ 1142#define TWOHOUR (120*60) 1143 lt6_tmp = ifa6->ia6_lifetime; 1144 1145 if (lt6_tmp.ia6t_vltime == ND6_INFINITE_LIFETIME) 1146 remaininglifetime = ND6_INFINITE_LIFETIME; 1147 else if (time_uptime - ifa6->ia6_updatetime > 1148 lt6_tmp.ia6t_vltime) { 1149 /* 1150 * The case of "invalid" address. We should usually 1151 * not see this case. 1152 */ 1153 remaininglifetime = 0; 1154 } else 1155 remaininglifetime = lt6_tmp.ia6t_vltime - 1156 (time_uptime - ifa6->ia6_updatetime); 1157 1158 /* when not updating, keep the current stored lifetime. */ 1159 lt6_tmp.ia6t_vltime = remaininglifetime; 1160 1161 if (TWOHOUR < new->ndpr_vltime || 1162 remaininglifetime < new->ndpr_vltime) { 1163 lt6_tmp.ia6t_vltime = new->ndpr_vltime; 1164 } else if (remaininglifetime <= TWOHOUR) { 1165 if (auth) { 1166 lt6_tmp.ia6t_vltime = new->ndpr_vltime; 1167 } 1168 } else { 1169 /* 1170 * new->ndpr_vltime <= TWOHOUR && 1171 * TWOHOUR < remaininglifetime 1172 */ 1173 lt6_tmp.ia6t_vltime = TWOHOUR; 1174 } 1175 1176 /* The 2 hour rule is not imposed for preferred lifetime. */ 1177 lt6_tmp.ia6t_pltime = new->ndpr_pltime; 1178 1179 in6_init_address_ltimes(pr, <6_tmp); 1180 1181 /* 1182 * We need to treat lifetimes for temporary addresses 1183 * differently, according to 1184 * draft-ietf-ipv6-privacy-addrs-v2-01.txt 3.3 (1); 1185 * we only update the lifetimes when they are in the maximum 1186 * intervals. 1187 */ 1188 if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) != 0) { 1189 u_int32_t maxvltime, maxpltime; 1190 1191 if (V_ip6_temp_valid_lifetime > 1192 (u_int32_t)((time_uptime - ifa6->ia6_createtime) + 1193 V_ip6_desync_factor)) { 1194 maxvltime = V_ip6_temp_valid_lifetime - 1195 (time_uptime - ifa6->ia6_createtime) - 1196 V_ip6_desync_factor; 1197 } else 1198 maxvltime = 0; 1199 if (V_ip6_temp_preferred_lifetime > 1200 (u_int32_t)((time_uptime - ifa6->ia6_createtime) + 1201 V_ip6_desync_factor)) { 1202 maxpltime = V_ip6_temp_preferred_lifetime - 1203 (time_uptime - ifa6->ia6_createtime) - 1204 V_ip6_desync_factor; 1205 } else 1206 maxpltime = 0; 1207 1208 if (lt6_tmp.ia6t_vltime == ND6_INFINITE_LIFETIME || 1209 lt6_tmp.ia6t_vltime > maxvltime) { 1210 lt6_tmp.ia6t_vltime = maxvltime; 1211 } 1212 if (lt6_tmp.ia6t_pltime == ND6_INFINITE_LIFETIME || 1213 lt6_tmp.ia6t_pltime > maxpltime) { 1214 lt6_tmp.ia6t_pltime = maxpltime; 1215 } 1216 } 1217 ifa6->ia6_lifetime = lt6_tmp; 1218 ifa6->ia6_updatetime = time_uptime; 1219 } 1220 IF_ADDR_RUNLOCK(ifp); 1221 if (ia6_match == NULL && new->ndpr_vltime) { 1222 int ifidlen; 1223 1224 /* 1225 * 5.5.3 (d) (continued) 1226 * No address matched and the valid lifetime is non-zero. 1227 * Create a new address. 1228 */ 1229 1230 /* 1231 * Prefix Length check: 1232 * If the sum of the prefix length and interface identifier 1233 * length does not equal 128 bits, the Prefix Information 1234 * option MUST be ignored. The length of the interface 1235 * identifier is defined in a separate link-type specific 1236 * document. 1237 */ 1238 ifidlen = in6_if2idlen(ifp); 1239 if (ifidlen < 0) { 1240 /* this should not happen, so we always log it. */ 1241 log(LOG_ERR, "prelist_update: IFID undefined (%s)\n", 1242 if_name(ifp)); 1243 goto end; 1244 } 1245 if (ifidlen + pr->ndpr_plen != 128) { 1246 nd6log((LOG_INFO, 1247 "prelist_update: invalid prefixlen " 1248 "%d for %s, ignored\n", 1249 pr->ndpr_plen, if_name(ifp))); 1250 goto end; 1251 } 1252 1253 if ((ia6 = in6_ifadd(new, mcast)) != NULL) { 1254 /* 1255 * note that we should use pr (not new) for reference. 1256 */ 1257 pr->ndpr_refcnt++; 1258 ia6->ia6_ndpr = pr; 1259 1260 /* 1261 * RFC 3041 3.3 (2). 1262 * When a new public address is created as described 1263 * in RFC2462, also create a new temporary address. 1264 * 1265 * RFC 3041 3.5. 1266 * When an interface connects to a new link, a new 1267 * randomized interface identifier should be generated 1268 * immediately together with a new set of temporary 1269 * addresses. Thus, we specifiy 1 as the 2nd arg of 1270 * in6_tmpifadd(). 1271 */ 1272 if (V_ip6_use_tempaddr) { 1273 int e; 1274 if ((e = in6_tmpifadd(ia6, 1, 1)) != 0) { 1275 nd6log((LOG_NOTICE, "prelist_update: " 1276 "failed to create a temporary " 1277 "address, errno=%d\n", 1278 e)); 1279 } 1280 } 1281 ifa_free(&ia6->ia_ifa); 1282 1283 /* 1284 * A newly added address might affect the status 1285 * of other addresses, so we check and update it. 1286 * XXX: what if address duplication happens? 1287 */ 1288 pfxlist_onlink_check(); 1289 } else { 1290 /* just set an error. do not bark here. */ 1291 error = EADDRNOTAVAIL; /* XXX: might be unused. */ 1292 } 1293 } 1294 1295 end: 1296 return error; 1297} 1298 1299/* 1300 * A supplement function used in the on-link detection below; 1301 * detect if a given prefix has a (probably) reachable advertising router. 1302 * XXX: lengthy function name... 1303 */ 1304static struct nd_pfxrouter * 1305find_pfxlist_reachable_router(struct nd_prefix *pr) 1306{ 1307 struct nd_pfxrouter *pfxrtr; 1308 struct llentry *ln; 1309 int canreach; 1310 1311 LIST_FOREACH(pfxrtr, &pr->ndpr_advrtrs, pfr_entry) { 1312 IF_AFDATA_RLOCK(pfxrtr->router->ifp); 1313 ln = nd6_lookup(&pfxrtr->router->rtaddr, 0, pfxrtr->router->ifp); 1314 IF_AFDATA_RUNLOCK(pfxrtr->router->ifp); 1315 if (ln == NULL) 1316 continue; 1317 canreach = ND6_IS_LLINFO_PROBREACH(ln); 1318 LLE_RUNLOCK(ln); 1319 if (canreach) 1320 break; 1321 } 1322 return (pfxrtr); 1323} 1324 1325/* 1326 * Check if each prefix in the prefix list has at least one available router 1327 * that advertised the prefix (a router is "available" if its neighbor cache 1328 * entry is reachable or probably reachable). 1329 * If the check fails, the prefix may be off-link, because, for example, 1330 * we have moved from the network but the lifetime of the prefix has not 1331 * expired yet. So we should not use the prefix if there is another prefix 1332 * that has an available router. 1333 * But, if there is no prefix that has an available router, we still regards 1334 * all the prefixes as on-link. This is because we can't tell if all the 1335 * routers are simply dead or if we really moved from the network and there 1336 * is no router around us. 1337 */ 1338void 1339pfxlist_onlink_check() 1340{ 1341 struct nd_prefix *pr; 1342 struct in6_ifaddr *ifa; 1343 struct nd_defrouter *dr; 1344 struct nd_pfxrouter *pfxrtr = NULL; 1345 1346 /* 1347 * Check if there is a prefix that has a reachable advertising 1348 * router. 1349 */ 1350 LIST_FOREACH(pr, &V_nd_prefix, ndpr_entry) { 1351 if (pr->ndpr_raf_onlink && find_pfxlist_reachable_router(pr)) 1352 break; 1353 } 1354 1355 /* 1356 * If we have no such prefix, check whether we still have a router 1357 * that does not advertise any prefixes. 1358 */ 1359 if (pr == NULL) { 1360 TAILQ_FOREACH(dr, &V_nd_defrouter, dr_entry) { 1361 struct nd_prefix *pr0; 1362 1363 LIST_FOREACH(pr0, &V_nd_prefix, ndpr_entry) { 1364 if ((pfxrtr = pfxrtr_lookup(pr0, dr)) != NULL) 1365 break; 1366 } 1367 if (pfxrtr != NULL) 1368 break; 1369 } 1370 } 1371 if (pr != NULL || (!TAILQ_EMPTY(&V_nd_defrouter) && pfxrtr == NULL)) { 1372 /* 1373 * There is at least one prefix that has a reachable router, 1374 * or at least a router which probably does not advertise 1375 * any prefixes. The latter would be the case when we move 1376 * to a new link where we have a router that does not provide 1377 * prefixes and we configure an address by hand. 1378 * Detach prefixes which have no reachable advertising 1379 * router, and attach other prefixes. 1380 */ 1381 LIST_FOREACH(pr, &V_nd_prefix, ndpr_entry) { 1382 /* XXX: a link-local prefix should never be detached */ 1383 if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr)) 1384 continue; 1385 1386 /* 1387 * we aren't interested in prefixes without the L bit 1388 * set. 1389 */ 1390 if (pr->ndpr_raf_onlink == 0) 1391 continue; 1392 1393 if (pr->ndpr_raf_auto == 0) 1394 continue; 1395 1396 if ((pr->ndpr_stateflags & NDPRF_DETACHED) == 0 && 1397 find_pfxlist_reachable_router(pr) == NULL) 1398 pr->ndpr_stateflags |= NDPRF_DETACHED; 1399 if ((pr->ndpr_stateflags & NDPRF_DETACHED) != 0 && 1400 find_pfxlist_reachable_router(pr) != 0) 1401 pr->ndpr_stateflags &= ~NDPRF_DETACHED; 1402 } 1403 } else { 1404 /* there is no prefix that has a reachable router */ 1405 LIST_FOREACH(pr, &V_nd_prefix, ndpr_entry) { 1406 if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr)) 1407 continue; 1408 1409 if (pr->ndpr_raf_onlink == 0) 1410 continue; 1411 1412 if (pr->ndpr_raf_auto == 0) 1413 continue; 1414 1415 if ((pr->ndpr_stateflags & NDPRF_DETACHED) != 0) 1416 pr->ndpr_stateflags &= ~NDPRF_DETACHED; 1417 } 1418 } 1419 1420 /* 1421 * Remove each interface route associated with a (just) detached 1422 * prefix, and reinstall the interface route for a (just) attached 1423 * prefix. Note that all attempt of reinstallation does not 1424 * necessarily success, when a same prefix is shared among multiple 1425 * interfaces. Such cases will be handled in nd6_prefix_onlink, 1426 * so we don't have to care about them. 1427 */ 1428 LIST_FOREACH(pr, &V_nd_prefix, ndpr_entry) { 1429 int e; 1430 char ip6buf[INET6_ADDRSTRLEN]; 1431 1432 if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr)) 1433 continue; 1434 1435 if (pr->ndpr_raf_onlink == 0) 1436 continue; 1437 1438 if (pr->ndpr_raf_auto == 0) 1439 continue; 1440 1441 if ((pr->ndpr_stateflags & NDPRF_DETACHED) != 0 && 1442 (pr->ndpr_stateflags & NDPRF_ONLINK) != 0) { 1443 if ((e = nd6_prefix_offlink(pr)) != 0) { 1444 nd6log((LOG_ERR, 1445 "pfxlist_onlink_check: failed to " 1446 "make %s/%d offlink, errno=%d\n", 1447 ip6_sprintf(ip6buf, 1448 &pr->ndpr_prefix.sin6_addr), 1449 pr->ndpr_plen, e)); 1450 } 1451 } 1452 if ((pr->ndpr_stateflags & NDPRF_DETACHED) == 0 && 1453 (pr->ndpr_stateflags & NDPRF_ONLINK) == 0 && 1454 pr->ndpr_raf_onlink) { 1455 if ((e = nd6_prefix_onlink(pr)) != 0) { 1456 nd6log((LOG_ERR, 1457 "pfxlist_onlink_check: failed to " 1458 "make %s/%d onlink, errno=%d\n", 1459 ip6_sprintf(ip6buf, 1460 &pr->ndpr_prefix.sin6_addr), 1461 pr->ndpr_plen, e)); 1462 } 1463 } 1464 } 1465 1466 /* 1467 * Changes on the prefix status might affect address status as well. 1468 * Make sure that all addresses derived from an attached prefix are 1469 * attached, and that all addresses derived from a detached prefix are 1470 * detached. Note, however, that a manually configured address should 1471 * always be attached. 1472 * The precise detection logic is same as the one for prefixes. 1473 * 1474 * XXXRW: in6_ifaddrhead locking. 1475 */ 1476 TAILQ_FOREACH(ifa, &V_in6_ifaddrhead, ia_link) { 1477 if (!(ifa->ia6_flags & IN6_IFF_AUTOCONF)) 1478 continue; 1479 1480 if (ifa->ia6_ndpr == NULL) { 1481 /* 1482 * This can happen when we first configure the address 1483 * (i.e. the address exists, but the prefix does not). 1484 * XXX: complicated relationships... 1485 */ 1486 continue; 1487 } 1488 1489 if (find_pfxlist_reachable_router(ifa->ia6_ndpr)) 1490 break; 1491 } 1492 if (ifa) { 1493 TAILQ_FOREACH(ifa, &V_in6_ifaddrhead, ia_link) { 1494 if ((ifa->ia6_flags & IN6_IFF_AUTOCONF) == 0) 1495 continue; 1496 1497 if (ifa->ia6_ndpr == NULL) /* XXX: see above. */ 1498 continue; 1499 1500 if (find_pfxlist_reachable_router(ifa->ia6_ndpr)) { 1501 if (ifa->ia6_flags & IN6_IFF_DETACHED) { 1502 ifa->ia6_flags &= ~IN6_IFF_DETACHED; 1503 ifa->ia6_flags |= IN6_IFF_TENTATIVE; 1504 nd6_dad_start((struct ifaddr *)ifa, 0); 1505 } 1506 } else { 1507 ifa->ia6_flags |= IN6_IFF_DETACHED; 1508 } 1509 } 1510 } 1511 else { 1512 TAILQ_FOREACH(ifa, &V_in6_ifaddrhead, ia_link) { 1513 if ((ifa->ia6_flags & IN6_IFF_AUTOCONF) == 0) 1514 continue; 1515 1516 if (ifa->ia6_flags & IN6_IFF_DETACHED) { 1517 ifa->ia6_flags &= ~IN6_IFF_DETACHED; 1518 ifa->ia6_flags |= IN6_IFF_TENTATIVE; 1519 /* Do we need a delay in this case? */ 1520 nd6_dad_start((struct ifaddr *)ifa, 0); 1521 } 1522 } 1523 } 1524} 1525 1526static int 1527nd6_prefix_onlink_rtrequest(struct nd_prefix *pr, struct ifaddr *ifa) 1528{ 1529 static struct sockaddr_dl null_sdl = {sizeof(null_sdl), AF_LINK}; 1530 struct radix_node_head *rnh; 1531 struct rtentry *rt; 1532 struct sockaddr_in6 mask6; 1533 u_long rtflags; 1534 int error, a_failure, fibnum; 1535 1536 /* 1537 * in6_ifinit() sets nd6_rtrequest to ifa_rtrequest for all ifaddrs. 1538 * ifa->ifa_rtrequest = nd6_rtrequest; 1539 */ 1540 bzero(&mask6, sizeof(mask6)); 1541 mask6.sin6_len = sizeof(mask6); 1542 mask6.sin6_addr = pr->ndpr_mask; 1543 rtflags = (ifa->ifa_flags & ~IFA_RTSELF) | RTF_UP; 1544 1545 a_failure = 0; 1546 for (fibnum = 0; fibnum < rt_numfibs; fibnum++) { 1547 1548 rt = NULL; 1549 error = in6_rtrequest(RTM_ADD, 1550 (struct sockaddr *)&pr->ndpr_prefix, ifa->ifa_addr, 1551 (struct sockaddr *)&mask6, rtflags, &rt, fibnum); 1552 if (error == 0) { 1553 KASSERT(rt != NULL, ("%s: in6_rtrequest return no " 1554 "error(%d) but rt is NULL, pr=%p, ifa=%p", __func__, 1555 error, pr, ifa)); 1556 1557 rnh = rt_tables_get_rnh(rt->rt_fibnum, AF_INET6); 1558 /* XXX what if rhn == NULL? */ 1559 RADIX_NODE_HEAD_LOCK(rnh); 1560 RT_LOCK(rt); 1561 if (rt_setgate(rt, rt_key(rt), 1562 (struct sockaddr *)&null_sdl) == 0) { 1563 struct sockaddr_dl *dl; 1564 1565 dl = (struct sockaddr_dl *)rt->rt_gateway; 1566 dl->sdl_type = rt->rt_ifp->if_type; 1567 dl->sdl_index = rt->rt_ifp->if_index; 1568 } 1569 RADIX_NODE_HEAD_UNLOCK(rnh); 1570 nd6_rtmsg(RTM_ADD, rt); 1571 RT_UNLOCK(rt); 1572 pr->ndpr_stateflags |= NDPRF_ONLINK; 1573 } else { 1574 char ip6buf[INET6_ADDRSTRLEN]; 1575 char ip6bufg[INET6_ADDRSTRLEN]; 1576 char ip6bufm[INET6_ADDRSTRLEN]; 1577 struct sockaddr_in6 *sin6; 1578 1579 sin6 = (struct sockaddr_in6 *)ifa->ifa_addr; 1580 nd6log((LOG_ERR, "nd6_prefix_onlink: failed to add " 1581 "route for a prefix (%s/%d) on %s, gw=%s, mask=%s, " 1582 "flags=%lx errno = %d\n", 1583 ip6_sprintf(ip6buf, &pr->ndpr_prefix.sin6_addr), 1584 pr->ndpr_plen, if_name(pr->ndpr_ifp), 1585 ip6_sprintf(ip6bufg, &sin6->sin6_addr), 1586 ip6_sprintf(ip6bufm, &mask6.sin6_addr), 1587 rtflags, error)); 1588 1589 /* Save last error to return, see rtinit(). */ 1590 a_failure = error; 1591 } 1592 1593 if (rt != NULL) { 1594 RT_LOCK(rt); 1595 RT_REMREF(rt); 1596 RT_UNLOCK(rt); 1597 } 1598 } 1599 1600 /* Return the last error we got. */ 1601 return (a_failure); 1602} 1603 1604static int 1605nd6_prefix_onlink(struct nd_prefix *pr) 1606{ 1607 struct ifaddr *ifa; 1608 struct ifnet *ifp = pr->ndpr_ifp; 1609 struct nd_prefix *opr; 1610 int error = 0; 1611 char ip6buf[INET6_ADDRSTRLEN]; 1612 1613 /* sanity check */ 1614 if ((pr->ndpr_stateflags & NDPRF_ONLINK) != 0) { 1615 nd6log((LOG_ERR, 1616 "nd6_prefix_onlink: %s/%d is already on-link\n", 1617 ip6_sprintf(ip6buf, &pr->ndpr_prefix.sin6_addr), 1618 pr->ndpr_plen)); 1619 return (EEXIST); 1620 } 1621 1622 /* 1623 * Add the interface route associated with the prefix. Before 1624 * installing the route, check if there's the same prefix on another 1625 * interface, and the prefix has already installed the interface route. 1626 * Although such a configuration is expected to be rare, we explicitly 1627 * allow it. 1628 */ 1629 LIST_FOREACH(opr, &V_nd_prefix, ndpr_entry) { 1630 if (opr == pr) 1631 continue; 1632 1633 if ((opr->ndpr_stateflags & NDPRF_ONLINK) == 0) 1634 continue; 1635 1636 if (opr->ndpr_plen == pr->ndpr_plen && 1637 in6_are_prefix_equal(&pr->ndpr_prefix.sin6_addr, 1638 &opr->ndpr_prefix.sin6_addr, pr->ndpr_plen)) 1639 return (0); 1640 } 1641 1642 /* 1643 * We prefer link-local addresses as the associated interface address. 1644 */ 1645 /* search for a link-local addr */ 1646 ifa = (struct ifaddr *)in6ifa_ifpforlinklocal(ifp, 1647 IN6_IFF_NOTREADY | IN6_IFF_ANYCAST); 1648 if (ifa == NULL) { 1649 /* XXX: freebsd does not have ifa_ifwithaf */ 1650 IF_ADDR_RLOCK(ifp); 1651 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1652 if (ifa->ifa_addr->sa_family == AF_INET6) 1653 break; 1654 } 1655 if (ifa != NULL) 1656 ifa_ref(ifa); 1657 IF_ADDR_RUNLOCK(ifp); 1658 /* should we care about ia6_flags? */ 1659 } 1660 if (ifa == NULL) { 1661 /* 1662 * This can still happen, when, for example, we receive an RA 1663 * containing a prefix with the L bit set and the A bit clear, 1664 * after removing all IPv6 addresses on the receiving 1665 * interface. This should, of course, be rare though. 1666 */ 1667 nd6log((LOG_NOTICE, 1668 "nd6_prefix_onlink: failed to find any ifaddr" 1669 " to add route for a prefix(%s/%d) on %s\n", 1670 ip6_sprintf(ip6buf, &pr->ndpr_prefix.sin6_addr), 1671 pr->ndpr_plen, if_name(ifp))); 1672 return (0); 1673 } 1674 1675 error = nd6_prefix_onlink_rtrequest(pr, ifa); 1676 1677 if (ifa != NULL) 1678 ifa_free(ifa); 1679 1680 return (error); 1681} 1682 1683static int 1684nd6_prefix_offlink(struct nd_prefix *pr) 1685{ 1686 int error = 0; 1687 struct ifnet *ifp = pr->ndpr_ifp; 1688 struct nd_prefix *opr; 1689 struct sockaddr_in6 sa6, mask6; 1690 struct rtentry *rt; 1691 char ip6buf[INET6_ADDRSTRLEN]; 1692 int fibnum, a_failure; 1693 1694 /* sanity check */ 1695 if ((pr->ndpr_stateflags & NDPRF_ONLINK) == 0) { 1696 nd6log((LOG_ERR, 1697 "nd6_prefix_offlink: %s/%d is already off-link\n", 1698 ip6_sprintf(ip6buf, &pr->ndpr_prefix.sin6_addr), 1699 pr->ndpr_plen)); 1700 return (EEXIST); 1701 } 1702 1703 bzero(&sa6, sizeof(sa6)); 1704 sa6.sin6_family = AF_INET6; 1705 sa6.sin6_len = sizeof(sa6); 1706 bcopy(&pr->ndpr_prefix.sin6_addr, &sa6.sin6_addr, 1707 sizeof(struct in6_addr)); 1708 bzero(&mask6, sizeof(mask6)); 1709 mask6.sin6_family = AF_INET6; 1710 mask6.sin6_len = sizeof(sa6); 1711 bcopy(&pr->ndpr_mask, &mask6.sin6_addr, sizeof(struct in6_addr)); 1712 1713 a_failure = 0; 1714 for (fibnum = 0; fibnum < rt_numfibs; fibnum++) { 1715 rt = NULL; 1716 error = in6_rtrequest(RTM_DELETE, (struct sockaddr *)&sa6, NULL, 1717 (struct sockaddr *)&mask6, 0, &rt, fibnum); 1718 if (error == 0) { 1719 /* report the route deletion to the routing socket. */ 1720 if (rt != NULL) 1721 nd6_rtmsg(RTM_DELETE, rt); 1722 } else { 1723 /* Save last error to return, see rtinit(). */ 1724 a_failure = error; 1725 } 1726 if (rt != NULL) { 1727 RTFREE(rt); 1728 } 1729 } 1730 error = a_failure; 1731 a_failure = 1; 1732 if (error == 0) { 1733 pr->ndpr_stateflags &= ~NDPRF_ONLINK; 1734 1735 /* 1736 * There might be the same prefix on another interface, 1737 * the prefix which could not be on-link just because we have 1738 * the interface route (see comments in nd6_prefix_onlink). 1739 * If there's one, try to make the prefix on-link on the 1740 * interface. 1741 */ 1742 LIST_FOREACH(opr, &V_nd_prefix, ndpr_entry) { 1743 if (opr == pr) 1744 continue; 1745 1746 if ((opr->ndpr_stateflags & NDPRF_ONLINK) != 0) 1747 continue; 1748 1749 /* 1750 * KAME specific: detached prefixes should not be 1751 * on-link. 1752 */ 1753 if ((opr->ndpr_stateflags & NDPRF_DETACHED) != 0) 1754 continue; 1755 1756 if (opr->ndpr_plen == pr->ndpr_plen && 1757 in6_are_prefix_equal(&pr->ndpr_prefix.sin6_addr, 1758 &opr->ndpr_prefix.sin6_addr, pr->ndpr_plen)) { 1759 int e; 1760 1761 if ((e = nd6_prefix_onlink(opr)) != 0) { 1762 nd6log((LOG_ERR, 1763 "nd6_prefix_offlink: failed to " 1764 "recover a prefix %s/%d from %s " 1765 "to %s (errno = %d)\n", 1766 ip6_sprintf(ip6buf, 1767 &opr->ndpr_prefix.sin6_addr), 1768 opr->ndpr_plen, if_name(ifp), 1769 if_name(opr->ndpr_ifp), e)); 1770 } else 1771 a_failure = 0; 1772 } 1773 } 1774 } else { 1775 /* XXX: can we still set the NDPRF_ONLINK flag? */ 1776 nd6log((LOG_ERR, 1777 "nd6_prefix_offlink: failed to delete route: " 1778 "%s/%d on %s (errno = %d)\n", 1779 ip6_sprintf(ip6buf, &sa6.sin6_addr), pr->ndpr_plen, 1780 if_name(ifp), error)); 1781 } 1782 1783 if (a_failure) 1784 lltable_prefix_free(AF_INET6, (struct sockaddr *)&sa6, 1785 (struct sockaddr *)&mask6, LLE_STATIC); 1786 1787 return (error); 1788} 1789 1790static struct in6_ifaddr * 1791in6_ifadd(struct nd_prefixctl *pr, int mcast) 1792{ 1793 struct ifnet *ifp = pr->ndpr_ifp; 1794 struct ifaddr *ifa; 1795 struct in6_aliasreq ifra; 1796 struct in6_ifaddr *ia, *ib; 1797 int error, plen0; 1798 struct in6_addr mask; 1799 int prefixlen = pr->ndpr_plen; 1800 int updateflags; 1801 char ip6buf[INET6_ADDRSTRLEN]; 1802 1803 in6_prefixlen2mask(&mask, prefixlen); 1804 1805 /* 1806 * find a link-local address (will be interface ID). 1807 * Is it really mandatory? Theoretically, a global or a site-local 1808 * address can be configured without a link-local address, if we 1809 * have a unique interface identifier... 1810 * 1811 * it is not mandatory to have a link-local address, we can generate 1812 * interface identifier on the fly. we do this because: 1813 * (1) it should be the easiest way to find interface identifier. 1814 * (2) RFC2462 5.4 suggesting the use of the same interface identifier 1815 * for multiple addresses on a single interface, and possible shortcut 1816 * of DAD. we omitted DAD for this reason in the past. 1817 * (3) a user can prevent autoconfiguration of global address 1818 * by removing link-local address by hand (this is partly because we 1819 * don't have other way to control the use of IPv6 on an interface. 1820 * this has been our design choice - cf. NRL's "ifconfig auto"). 1821 * (4) it is easier to manage when an interface has addresses 1822 * with the same interface identifier, than to have multiple addresses 1823 * with different interface identifiers. 1824 */ 1825 ifa = (struct ifaddr *)in6ifa_ifpforlinklocal(ifp, 0); /* 0 is OK? */ 1826 if (ifa) 1827 ib = (struct in6_ifaddr *)ifa; 1828 else 1829 return NULL; 1830 1831 /* prefixlen + ifidlen must be equal to 128 */ 1832 plen0 = in6_mask2len(&ib->ia_prefixmask.sin6_addr, NULL); 1833 if (prefixlen != plen0) { 1834 ifa_free(ifa); 1835 nd6log((LOG_INFO, "in6_ifadd: wrong prefixlen for %s " 1836 "(prefix=%d ifid=%d)\n", 1837 if_name(ifp), prefixlen, 128 - plen0)); 1838 return NULL; 1839 } 1840 1841 /* make ifaddr */ 1842 1843 bzero(&ifra, sizeof(ifra)); 1844 /* 1845 * in6_update_ifa() does not use ifra_name, but we accurately set it 1846 * for safety. 1847 */ 1848 strncpy(ifra.ifra_name, if_name(ifp), sizeof(ifra.ifra_name)); 1849 ifra.ifra_addr.sin6_family = AF_INET6; 1850 ifra.ifra_addr.sin6_len = sizeof(struct sockaddr_in6); 1851 /* prefix */ 1852 ifra.ifra_addr.sin6_addr = pr->ndpr_prefix.sin6_addr; 1853 ifra.ifra_addr.sin6_addr.s6_addr32[0] &= mask.s6_addr32[0]; 1854 ifra.ifra_addr.sin6_addr.s6_addr32[1] &= mask.s6_addr32[1]; 1855 ifra.ifra_addr.sin6_addr.s6_addr32[2] &= mask.s6_addr32[2]; 1856 ifra.ifra_addr.sin6_addr.s6_addr32[3] &= mask.s6_addr32[3]; 1857 1858 /* interface ID */ 1859 ifra.ifra_addr.sin6_addr.s6_addr32[0] |= 1860 (ib->ia_addr.sin6_addr.s6_addr32[0] & ~mask.s6_addr32[0]); 1861 ifra.ifra_addr.sin6_addr.s6_addr32[1] |= 1862 (ib->ia_addr.sin6_addr.s6_addr32[1] & ~mask.s6_addr32[1]); 1863 ifra.ifra_addr.sin6_addr.s6_addr32[2] |= 1864 (ib->ia_addr.sin6_addr.s6_addr32[2] & ~mask.s6_addr32[2]); 1865 ifra.ifra_addr.sin6_addr.s6_addr32[3] |= 1866 (ib->ia_addr.sin6_addr.s6_addr32[3] & ~mask.s6_addr32[3]); 1867 ifa_free(ifa); 1868 1869 /* new prefix mask. */ 1870 ifra.ifra_prefixmask.sin6_len = sizeof(struct sockaddr_in6); 1871 ifra.ifra_prefixmask.sin6_family = AF_INET6; 1872 bcopy(&mask, &ifra.ifra_prefixmask.sin6_addr, 1873 sizeof(ifra.ifra_prefixmask.sin6_addr)); 1874 1875 /* lifetimes. */ 1876 ifra.ifra_lifetime.ia6t_vltime = pr->ndpr_vltime; 1877 ifra.ifra_lifetime.ia6t_pltime = pr->ndpr_pltime; 1878 1879 /* XXX: scope zone ID? */ 1880 1881 ifra.ifra_flags |= IN6_IFF_AUTOCONF; /* obey autoconf */ 1882 1883 /* 1884 * Make sure that we do not have this address already. This should 1885 * usually not happen, but we can still see this case, e.g., if we 1886 * have manually configured the exact address to be configured. 1887 */ 1888 ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp, 1889 &ifra.ifra_addr.sin6_addr); 1890 if (ifa != NULL) { 1891 ifa_free(ifa); 1892 /* this should be rare enough to make an explicit log */ 1893 log(LOG_INFO, "in6_ifadd: %s is already configured\n", 1894 ip6_sprintf(ip6buf, &ifra.ifra_addr.sin6_addr)); 1895 return (NULL); 1896 } 1897 1898 /* 1899 * Allocate ifaddr structure, link into chain, etc. 1900 * If we are going to create a new address upon receiving a multicasted 1901 * RA, we need to impose a random delay before starting DAD. 1902 * [draft-ietf-ipv6-rfc2462bis-02.txt, Section 5.4.2] 1903 */ 1904 updateflags = 0; 1905 if (mcast) 1906 updateflags |= IN6_IFAUPDATE_DADDELAY; 1907 if ((error = in6_update_ifa(ifp, &ifra, NULL, updateflags)) != 0) { 1908 nd6log((LOG_ERR, 1909 "in6_ifadd: failed to make ifaddr %s on %s (errno=%d)\n", 1910 ip6_sprintf(ip6buf, &ifra.ifra_addr.sin6_addr), 1911 if_name(ifp), error)); 1912 return (NULL); /* ifaddr must not have been allocated. */ 1913 } 1914 1915 ia = in6ifa_ifpwithaddr(ifp, &ifra.ifra_addr.sin6_addr); 1916 /* 1917 * XXXRW: Assumption of non-NULLness here might not be true with 1918 * fine-grained locking -- should we validate it? Or just return 1919 * earlier ifa rather than looking it up again? 1920 */ 1921 return (ia); /* this is always non-NULL and referenced. */ 1922} 1923 1924/* 1925 * ia0 - corresponding public address 1926 */ 1927int 1928in6_tmpifadd(const struct in6_ifaddr *ia0, int forcegen, int delay) 1929{ 1930 struct ifnet *ifp = ia0->ia_ifa.ifa_ifp; 1931 struct in6_ifaddr *newia, *ia; 1932 struct in6_aliasreq ifra; 1933 int i, error; 1934 int trylimit = 3; /* XXX: adhoc value */ 1935 int updateflags; 1936 u_int32_t randid[2]; 1937 time_t vltime0, pltime0; 1938 1939 bzero(&ifra, sizeof(ifra)); 1940 strncpy(ifra.ifra_name, if_name(ifp), sizeof(ifra.ifra_name)); 1941 ifra.ifra_addr = ia0->ia_addr; 1942 /* copy prefix mask */ 1943 ifra.ifra_prefixmask = ia0->ia_prefixmask; 1944 /* clear the old IFID */ 1945 for (i = 0; i < 4; i++) { 1946 ifra.ifra_addr.sin6_addr.s6_addr32[i] &= 1947 ifra.ifra_prefixmask.sin6_addr.s6_addr32[i]; 1948 } 1949 1950 again: 1951 if (in6_get_tmpifid(ifp, (u_int8_t *)randid, 1952 (const u_int8_t *)&ia0->ia_addr.sin6_addr.s6_addr[8], forcegen)) { 1953 nd6log((LOG_NOTICE, "in6_tmpifadd: failed to find a good " 1954 "random IFID\n")); 1955 return (EINVAL); 1956 } 1957 ifra.ifra_addr.sin6_addr.s6_addr32[2] |= 1958 (randid[0] & ~(ifra.ifra_prefixmask.sin6_addr.s6_addr32[2])); 1959 ifra.ifra_addr.sin6_addr.s6_addr32[3] |= 1960 (randid[1] & ~(ifra.ifra_prefixmask.sin6_addr.s6_addr32[3])); 1961 1962 /* 1963 * in6_get_tmpifid() quite likely provided a unique interface ID. 1964 * However, we may still have a chance to see collision, because 1965 * there may be a time lag between generation of the ID and generation 1966 * of the address. So, we'll do one more sanity check. 1967 */ 1968 IN6_IFADDR_RLOCK(); 1969 TAILQ_FOREACH(ia, &V_in6_ifaddrhead, ia_link) { 1970 if (IN6_ARE_ADDR_EQUAL(&ia->ia_addr.sin6_addr, 1971 &ifra.ifra_addr.sin6_addr)) { 1972 if (trylimit-- == 0) { 1973 IN6_IFADDR_RUNLOCK(); 1974 /* 1975 * Give up. Something strange should have 1976 * happened. 1977 */ 1978 nd6log((LOG_NOTICE, "in6_tmpifadd: failed to " 1979 "find a unique random IFID\n")); 1980 return (EEXIST); 1981 } 1982 IN6_IFADDR_RUNLOCK(); 1983 forcegen = 1; 1984 goto again; 1985 } 1986 } 1987 IN6_IFADDR_RUNLOCK(); 1988 1989 /* 1990 * The Valid Lifetime is the lower of the Valid Lifetime of the 1991 * public address or TEMP_VALID_LIFETIME. 1992 * The Preferred Lifetime is the lower of the Preferred Lifetime 1993 * of the public address or TEMP_PREFERRED_LIFETIME - 1994 * DESYNC_FACTOR. 1995 */ 1996 if (ia0->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) { 1997 vltime0 = IFA6_IS_INVALID(ia0) ? 0 : 1998 (ia0->ia6_lifetime.ia6t_vltime - 1999 (time_uptime - ia0->ia6_updatetime)); 2000 if (vltime0 > V_ip6_temp_valid_lifetime) 2001 vltime0 = V_ip6_temp_valid_lifetime; 2002 } else 2003 vltime0 = V_ip6_temp_valid_lifetime; 2004 if (ia0->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) { 2005 pltime0 = IFA6_IS_DEPRECATED(ia0) ? 0 : 2006 (ia0->ia6_lifetime.ia6t_pltime - 2007 (time_uptime - ia0->ia6_updatetime)); 2008 if (pltime0 > V_ip6_temp_preferred_lifetime - V_ip6_desync_factor){ 2009 pltime0 = V_ip6_temp_preferred_lifetime - 2010 V_ip6_desync_factor; 2011 } 2012 } else 2013 pltime0 = V_ip6_temp_preferred_lifetime - V_ip6_desync_factor; 2014 ifra.ifra_lifetime.ia6t_vltime = vltime0; 2015 ifra.ifra_lifetime.ia6t_pltime = pltime0; 2016 2017 /* 2018 * A temporary address is created only if this calculated Preferred 2019 * Lifetime is greater than REGEN_ADVANCE time units. 2020 */ 2021 if (ifra.ifra_lifetime.ia6t_pltime <= V_ip6_temp_regen_advance) 2022 return (0); 2023 2024 /* XXX: scope zone ID? */ 2025 2026 ifra.ifra_flags |= (IN6_IFF_AUTOCONF|IN6_IFF_TEMPORARY); 2027 2028 /* allocate ifaddr structure, link into chain, etc. */ 2029 updateflags = 0; 2030 if (delay) 2031 updateflags |= IN6_IFAUPDATE_DADDELAY; 2032 if ((error = in6_update_ifa(ifp, &ifra, NULL, updateflags)) != 0) 2033 return (error); 2034 2035 newia = in6ifa_ifpwithaddr(ifp, &ifra.ifra_addr.sin6_addr); 2036 if (newia == NULL) { /* XXX: can it happen? */ 2037 nd6log((LOG_ERR, 2038 "in6_tmpifadd: ifa update succeeded, but we got " 2039 "no ifaddr\n")); 2040 return (EINVAL); /* XXX */ 2041 } 2042 newia->ia6_ndpr = ia0->ia6_ndpr; 2043 newia->ia6_ndpr->ndpr_refcnt++; 2044 ifa_free(&newia->ia_ifa); 2045 2046 /* 2047 * A newly added address might affect the status of other addresses. 2048 * XXX: when the temporary address is generated with a new public 2049 * address, the onlink check is redundant. However, it would be safe 2050 * to do the check explicitly everywhere a new address is generated, 2051 * and, in fact, we surely need the check when we create a new 2052 * temporary address due to deprecation of an old temporary address. 2053 */ 2054 pfxlist_onlink_check(); 2055 2056 return (0); 2057} 2058 2059static int 2060in6_init_prefix_ltimes(struct nd_prefix *ndpr) 2061{ 2062 if (ndpr->ndpr_pltime == ND6_INFINITE_LIFETIME) 2063 ndpr->ndpr_preferred = 0; 2064 else 2065 ndpr->ndpr_preferred = time_uptime + ndpr->ndpr_pltime; 2066 if (ndpr->ndpr_vltime == ND6_INFINITE_LIFETIME) 2067 ndpr->ndpr_expire = 0; 2068 else 2069 ndpr->ndpr_expire = time_uptime + ndpr->ndpr_vltime; 2070 2071 return 0; 2072} 2073 2074static void 2075in6_init_address_ltimes(struct nd_prefix *new, struct in6_addrlifetime *lt6) 2076{ 2077 /* init ia6t_expire */ 2078 if (lt6->ia6t_vltime == ND6_INFINITE_LIFETIME) 2079 lt6->ia6t_expire = 0; 2080 else { 2081 lt6->ia6t_expire = time_uptime; 2082 lt6->ia6t_expire += lt6->ia6t_vltime; 2083 } 2084 2085 /* init ia6t_preferred */ 2086 if (lt6->ia6t_pltime == ND6_INFINITE_LIFETIME) 2087 lt6->ia6t_preferred = 0; 2088 else { 2089 lt6->ia6t_preferred = time_uptime; 2090 lt6->ia6t_preferred += lt6->ia6t_pltime; 2091 } 2092} 2093 2094/* 2095 * Delete all the routing table entries that use the specified gateway. 2096 * XXX: this function causes search through all entries of routing table, so 2097 * it shouldn't be called when acting as a router. 2098 */ 2099void 2100rt6_flush(struct in6_addr *gateway, struct ifnet *ifp) 2101{ 2102 struct radix_node_head *rnh; 2103 u_int fibnum; 2104 2105 /* We'll care only link-local addresses */ 2106 if (!IN6_IS_ADDR_LINKLOCAL(gateway)) 2107 return; 2108 2109 /* XXX Do we really need to walk any but the default FIB? */ 2110 for (fibnum = 0; fibnum < rt_numfibs; fibnum++) { 2111 rnh = rt_tables_get_rnh(fibnum, AF_INET6); 2112 if (rnh == NULL) 2113 continue; 2114 2115 RADIX_NODE_HEAD_LOCK(rnh); 2116 rnh->rnh_walktree(rnh, rt6_deleteroute, (void *)gateway); 2117 RADIX_NODE_HEAD_UNLOCK(rnh); 2118 } 2119} 2120 2121static int 2122rt6_deleteroute(struct radix_node *rn, void *arg) 2123{ 2124#define SIN6(s) ((struct sockaddr_in6 *)s) 2125 struct rtentry *rt = (struct rtentry *)rn; 2126 struct in6_addr *gate = (struct in6_addr *)arg; 2127 2128 if (rt->rt_gateway == NULL || rt->rt_gateway->sa_family != AF_INET6) 2129 return (0); 2130 2131 if (!IN6_ARE_ADDR_EQUAL(gate, &SIN6(rt->rt_gateway)->sin6_addr)) { 2132 return (0); 2133 } 2134 2135 /* 2136 * Do not delete a static route. 2137 * XXX: this seems to be a bit ad-hoc. Should we consider the 2138 * 'cloned' bit instead? 2139 */ 2140 if ((rt->rt_flags & RTF_STATIC) != 0) 2141 return (0); 2142 2143 /* 2144 * We delete only host route. This means, in particular, we don't 2145 * delete default route. 2146 */ 2147 if ((rt->rt_flags & RTF_HOST) == 0) 2148 return (0); 2149 2150 return (in6_rtrequest(RTM_DELETE, rt_key(rt), rt->rt_gateway, 2151 rt_mask(rt), rt->rt_flags, NULL, rt->rt_fibnum)); 2152#undef SIN6 2153} 2154 2155int 2156nd6_setdefaultiface(int ifindex) 2157{ 2158 int error = 0; 2159 2160 if (ifindex < 0 || V_if_index < ifindex) 2161 return (EINVAL); 2162 if (ifindex != 0 && !ifnet_byindex(ifindex)) 2163 return (EINVAL); 2164 2165 if (V_nd6_defifindex != ifindex) { 2166 V_nd6_defifindex = ifindex; 2167 if (V_nd6_defifindex > 0) 2168 V_nd6_defifp = ifnet_byindex(V_nd6_defifindex); 2169 else 2170 V_nd6_defifp = NULL; 2171 2172 /* 2173 * Our current implementation assumes one-to-one maping between 2174 * interfaces and links, so it would be natural to use the 2175 * default interface as the default link. 2176 */ 2177 scope6_setdefault(V_nd6_defifp); 2178 } 2179 2180 return (error); 2181} 2182