nd6_rtr.c revision 299145
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: stable/10/sys/netinet6/nd6_rtr.c 299145 2016-05-05 23:06:39Z markj $"); 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.raflags = 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} 505 506struct nd_defrouter * 507defrouter_lookup(struct in6_addr *addr, struct ifnet *ifp) 508{ 509 struct nd_defrouter *dr; 510 511 TAILQ_FOREACH(dr, &V_nd_defrouter, dr_entry) { 512 if (dr->ifp == ifp && IN6_ARE_ADDR_EQUAL(addr, &dr->rtaddr)) 513 return (dr); 514 } 515 516 return (NULL); /* search failed */ 517} 518 519/* 520 * Remove the default route for a given router. 521 * This is just a subroutine function for defrouter_select(), and should 522 * not be called from anywhere else. 523 */ 524static void 525defrouter_delreq(struct nd_defrouter *dr) 526{ 527 struct sockaddr_in6 def, mask, gate; 528 struct rtentry *oldrt = NULL; 529 530 bzero(&def, sizeof(def)); 531 bzero(&mask, sizeof(mask)); 532 bzero(&gate, sizeof(gate)); 533 534 def.sin6_len = mask.sin6_len = gate.sin6_len = 535 sizeof(struct sockaddr_in6); 536 def.sin6_family = gate.sin6_family = AF_INET6; 537 gate.sin6_addr = dr->rtaddr; 538 539 in6_rtrequest(RTM_DELETE, (struct sockaddr *)&def, 540 (struct sockaddr *)&gate, 541 (struct sockaddr *)&mask, RTF_GATEWAY, &oldrt, RT_DEFAULT_FIB); 542 if (oldrt) { 543 nd6_rtmsg(RTM_DELETE, oldrt); 544 RTFREE(oldrt); 545 } 546 547 dr->installed = 0; 548} 549 550/* 551 * remove all default routes from default router list 552 */ 553void 554defrouter_reset(void) 555{ 556 struct nd_defrouter *dr; 557 558 TAILQ_FOREACH(dr, &V_nd_defrouter, dr_entry) 559 defrouter_delreq(dr); 560 561 /* 562 * XXX should we also nuke any default routers in the kernel, by 563 * going through them by rtalloc1()? 564 */ 565} 566 567void 568defrtrlist_del(struct nd_defrouter *dr) 569{ 570 struct nd_defrouter *deldr = NULL; 571 struct nd_prefix *pr; 572 573 /* 574 * Flush all the routing table entries that use the router 575 * as a next hop. 576 */ 577 if (ND_IFINFO(dr->ifp)->flags & ND6_IFF_ACCEPT_RTADV) 578 rt6_flush(&dr->rtaddr, dr->ifp); 579 580 if (dr->installed) { 581 deldr = dr; 582 defrouter_delreq(dr); 583 } 584 TAILQ_REMOVE(&V_nd_defrouter, dr, dr_entry); 585 586 /* 587 * Also delete all the pointers to the router in each prefix lists. 588 */ 589 LIST_FOREACH(pr, &V_nd_prefix, ndpr_entry) { 590 struct nd_pfxrouter *pfxrtr; 591 if ((pfxrtr = pfxrtr_lookup(pr, dr)) != NULL) 592 pfxrtr_del(pfxrtr); 593 } 594 pfxlist_onlink_check(); 595 596 /* 597 * If the router is the primary one, choose a new one. 598 * Note that defrouter_select() will remove the current gateway 599 * from the routing table. 600 */ 601 if (deldr) 602 defrouter_select(); 603 604 free(dr, M_IP6NDP); 605} 606 607/* 608 * Default Router Selection according to Section 6.3.6 of RFC 2461 and 609 * draft-ietf-ipngwg-router-selection: 610 * 1) Routers that are reachable or probably reachable should be preferred. 611 * If we have more than one (probably) reachable router, prefer ones 612 * with the highest router preference. 613 * 2) When no routers on the list are known to be reachable or 614 * probably reachable, routers SHOULD be selected in a round-robin 615 * fashion, regardless of router preference values. 616 * 3) If the Default Router List is empty, assume that all 617 * destinations are on-link. 618 * 619 * We assume nd_defrouter is sorted by router preference value. 620 * Since the code below covers both with and without router preference cases, 621 * we do not need to classify the cases by ifdef. 622 * 623 * At this moment, we do not try to install more than one default router, 624 * even when the multipath routing is available, because we're not sure about 625 * the benefits for stub hosts comparing to the risk of making the code 626 * complicated and the possibility of introducing bugs. 627 */ 628void 629defrouter_select(void) 630{ 631 struct nd_defrouter *dr, *selected_dr = NULL, *installed_dr = NULL; 632 struct llentry *ln = NULL; 633 634 /* 635 * Let's handle easy case (3) first: 636 * If default router list is empty, there's nothing to be done. 637 */ 638 if (TAILQ_EMPTY(&V_nd_defrouter)) 639 return; 640 641 /* 642 * Search for a (probably) reachable router from the list. 643 * We just pick up the first reachable one (if any), assuming that 644 * the ordering rule of the list described in defrtrlist_update(). 645 */ 646 TAILQ_FOREACH(dr, &V_nd_defrouter, dr_entry) { 647 IF_AFDATA_RLOCK(dr->ifp); 648 if (selected_dr == NULL && 649 (ln = nd6_lookup(&dr->rtaddr, 0, dr->ifp)) && 650 ND6_IS_LLINFO_PROBREACH(ln)) { 651 selected_dr = dr; 652 } 653 IF_AFDATA_RUNLOCK(dr->ifp); 654 if (ln != NULL) { 655 LLE_RUNLOCK(ln); 656 ln = NULL; 657 } 658 659 if (dr->installed && installed_dr == NULL) 660 installed_dr = dr; 661 else if (dr->installed && installed_dr) { 662 /* this should not happen. warn for diagnosis. */ 663 log(LOG_ERR, "defrouter_select: more than one router" 664 " is installed\n"); 665 } 666 } 667 /* 668 * If none of the default routers was found to be reachable, 669 * round-robin the list regardless of preference. 670 * Otherwise, if we have an installed router, check if the selected 671 * (reachable) router should really be preferred to the installed one. 672 * We only prefer the new router when the old one is not reachable 673 * or when the new one has a really higher preference value. 674 */ 675 if (selected_dr == NULL) { 676 if (installed_dr == NULL || !TAILQ_NEXT(installed_dr, dr_entry)) 677 selected_dr = TAILQ_FIRST(&V_nd_defrouter); 678 else 679 selected_dr = TAILQ_NEXT(installed_dr, dr_entry); 680 } else if (installed_dr) { 681 IF_AFDATA_RLOCK(installed_dr->ifp); 682 if ((ln = nd6_lookup(&installed_dr->rtaddr, 0, installed_dr->ifp)) && 683 ND6_IS_LLINFO_PROBREACH(ln) && 684 rtpref(selected_dr) <= rtpref(installed_dr)) { 685 selected_dr = installed_dr; 686 } 687 IF_AFDATA_RUNLOCK(installed_dr->ifp); 688 if (ln != NULL) 689 LLE_RUNLOCK(ln); 690 } 691 692 /* 693 * If the selected router is different than the installed one, 694 * remove the installed router and install the selected one. 695 * Note that the selected router is never NULL here. 696 */ 697 if (installed_dr != selected_dr) { 698 if (installed_dr) 699 defrouter_delreq(installed_dr); 700 defrouter_addreq(selected_dr); 701 } 702} 703 704/* 705 * for default router selection 706 * regards router-preference field as a 2-bit signed integer 707 */ 708static int 709rtpref(struct nd_defrouter *dr) 710{ 711 switch (dr->raflags & ND_RA_FLAG_RTPREF_MASK) { 712 case ND_RA_FLAG_RTPREF_HIGH: 713 return (RTPREF_HIGH); 714 case ND_RA_FLAG_RTPREF_MEDIUM: 715 case ND_RA_FLAG_RTPREF_RSV: 716 return (RTPREF_MEDIUM); 717 case ND_RA_FLAG_RTPREF_LOW: 718 return (RTPREF_LOW); 719 default: 720 /* 721 * This case should never happen. If it did, it would mean a 722 * serious bug of kernel internal. We thus always bark here. 723 * Or, can we even panic? 724 */ 725 log(LOG_ERR, "rtpref: impossible RA flag %x\n", dr->raflags); 726 return (RTPREF_INVALID); 727 } 728 /* NOTREACHED */ 729} 730 731static struct nd_defrouter * 732defrtrlist_update(struct nd_defrouter *new) 733{ 734 struct nd_defrouter *dr, *n; 735 int oldpref; 736 737 if ((dr = defrouter_lookup(&new->rtaddr, new->ifp)) != NULL) { 738 /* entry exists */ 739 if (new->rtlifetime == 0) { 740 defrtrlist_del(dr); 741 return (NULL); 742 } 743 744 oldpref = rtpref(dr); 745 746 /* override */ 747 dr->raflags = new->raflags; /* XXX flag check */ 748 dr->rtlifetime = new->rtlifetime; 749 dr->expire = new->expire; 750 751 /* 752 * If the preference does not change, there's no need 753 * to sort the entries. Also make sure the selected 754 * router is still installed in the kernel. 755 */ 756 if (dr->installed && rtpref(new) == oldpref) 757 return (dr); 758 759 /* 760 * The preferred router may have changed, so relocate this 761 * router. 762 */ 763 TAILQ_REMOVE(&V_nd_defrouter, dr, dr_entry); 764 n = dr; 765 goto insert; 766 } 767 768 /* entry does not exist */ 769 if (new->rtlifetime == 0) 770 return (NULL); 771 772 n = malloc(sizeof(*n), M_IP6NDP, M_NOWAIT | M_ZERO); 773 if (n == NULL) 774 return (NULL); 775 memcpy(n, new, sizeof(*n)); 776 777insert: 778 /* 779 * Insert the new router in the Default Router List; 780 * The Default Router List should be in the descending order 781 * of router-preferece. Routers with the same preference are 782 * sorted in the arriving time order. 783 */ 784 785 /* insert at the end of the group */ 786 TAILQ_FOREACH(dr, &V_nd_defrouter, dr_entry) { 787 if (rtpref(n) > rtpref(dr)) 788 break; 789 } 790 if (dr) 791 TAILQ_INSERT_BEFORE(dr, n, dr_entry); 792 else 793 TAILQ_INSERT_TAIL(&V_nd_defrouter, n, dr_entry); 794 795 defrouter_select(); 796 797 return (n); 798} 799 800static struct nd_pfxrouter * 801pfxrtr_lookup(struct nd_prefix *pr, struct nd_defrouter *dr) 802{ 803 struct nd_pfxrouter *search; 804 805 LIST_FOREACH(search, &pr->ndpr_advrtrs, pfr_entry) { 806 if (search->router == dr) 807 break; 808 } 809 810 return (search); 811} 812 813static void 814pfxrtr_add(struct nd_prefix *pr, struct nd_defrouter *dr) 815{ 816 struct nd_pfxrouter *new; 817 818 new = malloc(sizeof(*new), M_IP6NDP, M_NOWAIT | M_ZERO); 819 if (new == NULL) 820 return; 821 new->router = dr; 822 823 LIST_INSERT_HEAD(&pr->ndpr_advrtrs, new, pfr_entry); 824 825 pfxlist_onlink_check(); 826} 827 828static void 829pfxrtr_del(struct nd_pfxrouter *pfr) 830{ 831 LIST_REMOVE(pfr, pfr_entry); 832 free(pfr, M_IP6NDP); 833} 834 835struct nd_prefix * 836nd6_prefix_lookup(struct nd_prefixctl *key) 837{ 838 struct nd_prefix *search; 839 840 LIST_FOREACH(search, &V_nd_prefix, ndpr_entry) { 841 if (key->ndpr_ifp == search->ndpr_ifp && 842 key->ndpr_plen == search->ndpr_plen && 843 in6_are_prefix_equal(&key->ndpr_prefix.sin6_addr, 844 &search->ndpr_prefix.sin6_addr, key->ndpr_plen)) { 845 break; 846 } 847 } 848 849 return (search); 850} 851 852int 853nd6_prelist_add(struct nd_prefixctl *pr, struct nd_defrouter *dr, 854 struct nd_prefix **newp) 855{ 856 struct nd_prefix *new = NULL; 857 int error = 0; 858 int i; 859 char ip6buf[INET6_ADDRSTRLEN]; 860 861 new = malloc(sizeof(*new), M_IP6NDP, M_NOWAIT | M_ZERO); 862 if (new == NULL) 863 return (ENOMEM); 864 new->ndpr_ifp = pr->ndpr_ifp; 865 new->ndpr_prefix = pr->ndpr_prefix; 866 new->ndpr_plen = pr->ndpr_plen; 867 new->ndpr_vltime = pr->ndpr_vltime; 868 new->ndpr_pltime = pr->ndpr_pltime; 869 new->ndpr_flags = pr->ndpr_flags; 870 if ((error = in6_init_prefix_ltimes(new)) != 0) { 871 free(new, M_IP6NDP); 872 return(error); 873 } 874 new->ndpr_lastupdate = time_uptime; 875 if (newp != NULL) 876 *newp = new; 877 878 /* initialization */ 879 LIST_INIT(&new->ndpr_advrtrs); 880 in6_prefixlen2mask(&new->ndpr_mask, new->ndpr_plen); 881 /* make prefix in the canonical form */ 882 for (i = 0; i < 4; i++) 883 new->ndpr_prefix.sin6_addr.s6_addr32[i] &= 884 new->ndpr_mask.s6_addr32[i]; 885 886 /* link ndpr_entry to nd_prefix list */ 887 LIST_INSERT_HEAD(&V_nd_prefix, new, ndpr_entry); 888 889 /* ND_OPT_PI_FLAG_ONLINK processing */ 890 if (new->ndpr_raf_onlink) { 891 int e; 892 893 if ((e = nd6_prefix_onlink(new)) != 0) { 894 nd6log((LOG_ERR, "nd6_prelist_add: failed to make " 895 "the prefix %s/%d on-link on %s (errno=%d)\n", 896 ip6_sprintf(ip6buf, &pr->ndpr_prefix.sin6_addr), 897 pr->ndpr_plen, if_name(pr->ndpr_ifp), e)); 898 /* proceed anyway. XXX: is it correct? */ 899 } 900 } 901 902 if (dr) 903 pfxrtr_add(new, dr); 904 905 return 0; 906} 907 908void 909prelist_remove(struct nd_prefix *pr) 910{ 911 struct nd_pfxrouter *pfr, *next; 912 int e; 913 char ip6buf[INET6_ADDRSTRLEN]; 914 915 /* make sure to invalidate the prefix until it is really freed. */ 916 pr->ndpr_vltime = 0; 917 pr->ndpr_pltime = 0; 918 919 /* 920 * Though these flags are now meaningless, we'd rather keep the value 921 * of pr->ndpr_raf_onlink and pr->ndpr_raf_auto not to confuse users 922 * when executing "ndp -p". 923 */ 924 925 if ((pr->ndpr_stateflags & NDPRF_ONLINK) != 0 && 926 (e = nd6_prefix_offlink(pr)) != 0) { 927 nd6log((LOG_ERR, "prelist_remove: failed to make %s/%d offlink " 928 "on %s, errno=%d\n", 929 ip6_sprintf(ip6buf, &pr->ndpr_prefix.sin6_addr), 930 pr->ndpr_plen, if_name(pr->ndpr_ifp), e)); 931 /* what should we do? */ 932 } 933 934 if (pr->ndpr_refcnt > 0) 935 return; /* notice here? */ 936 937 /* unlink ndpr_entry from nd_prefix list */ 938 LIST_REMOVE(pr, ndpr_entry); 939 940 /* free list of routers that advertised the prefix */ 941 LIST_FOREACH_SAFE(pfr, &pr->ndpr_advrtrs, pfr_entry, next) { 942 pfxrtr_del(pfr); 943 } 944 free(pr, M_IP6NDP); 945 946 pfxlist_onlink_check(); 947} 948 949/* 950 * dr - may be NULL 951 */ 952 953static int 954prelist_update(struct nd_prefixctl *new, struct nd_defrouter *dr, 955 struct mbuf *m, int mcast) 956{ 957 struct in6_ifaddr *ia6 = NULL, *ia6_match = NULL; 958 struct ifaddr *ifa; 959 struct ifnet *ifp = new->ndpr_ifp; 960 struct nd_prefix *pr; 961 int error = 0; 962 int newprefix = 0; 963 int auth; 964 struct in6_addrlifetime lt6_tmp; 965 char ip6buf[INET6_ADDRSTRLEN]; 966 967 auth = 0; 968 if (m) { 969 /* 970 * Authenticity for NA consists authentication for 971 * both IP header and IP datagrams, doesn't it ? 972 */ 973#if defined(M_AUTHIPHDR) && defined(M_AUTHIPDGM) 974 auth = ((m->m_flags & M_AUTHIPHDR) && 975 (m->m_flags & M_AUTHIPDGM)); 976#endif 977 } 978 979 if ((pr = nd6_prefix_lookup(new)) != NULL) { 980 /* 981 * nd6_prefix_lookup() ensures that pr and new have the same 982 * prefix on a same interface. 983 */ 984 985 /* 986 * Update prefix information. Note that the on-link (L) bit 987 * and the autonomous (A) bit should NOT be changed from 1 988 * to 0. 989 */ 990 if (new->ndpr_raf_onlink == 1) 991 pr->ndpr_raf_onlink = 1; 992 if (new->ndpr_raf_auto == 1) 993 pr->ndpr_raf_auto = 1; 994 if (new->ndpr_raf_onlink) { 995 pr->ndpr_vltime = new->ndpr_vltime; 996 pr->ndpr_pltime = new->ndpr_pltime; 997 (void)in6_init_prefix_ltimes(pr); /* XXX error case? */ 998 pr->ndpr_lastupdate = time_uptime; 999 } 1000 1001 if (new->ndpr_raf_onlink && 1002 (pr->ndpr_stateflags & NDPRF_ONLINK) == 0) { 1003 int e; 1004 1005 if ((e = nd6_prefix_onlink(pr)) != 0) { 1006 nd6log((LOG_ERR, 1007 "prelist_update: failed to make " 1008 "the prefix %s/%d on-link on %s " 1009 "(errno=%d)\n", 1010 ip6_sprintf(ip6buf, 1011 &pr->ndpr_prefix.sin6_addr), 1012 pr->ndpr_plen, if_name(pr->ndpr_ifp), e)); 1013 /* proceed anyway. XXX: is it correct? */ 1014 } 1015 } 1016 1017 if (dr && pfxrtr_lookup(pr, dr) == NULL) 1018 pfxrtr_add(pr, dr); 1019 } else { 1020 struct nd_prefix *newpr = NULL; 1021 1022 newprefix = 1; 1023 1024 if (new->ndpr_vltime == 0) 1025 goto end; 1026 if (new->ndpr_raf_onlink == 0 && new->ndpr_raf_auto == 0) 1027 goto end; 1028 1029 error = nd6_prelist_add(new, dr, &newpr); 1030 if (error != 0 || newpr == NULL) { 1031 nd6log((LOG_NOTICE, "prelist_update: " 1032 "nd6_prelist_add failed for %s/%d on %s " 1033 "errno=%d, returnpr=%p\n", 1034 ip6_sprintf(ip6buf, &new->ndpr_prefix.sin6_addr), 1035 new->ndpr_plen, if_name(new->ndpr_ifp), 1036 error, newpr)); 1037 goto end; /* we should just give up in this case. */ 1038 } 1039 1040 /* 1041 * XXX: from the ND point of view, we can ignore a prefix 1042 * with the on-link bit being zero. However, we need a 1043 * prefix structure for references from autoconfigured 1044 * addresses. Thus, we explicitly make sure that the prefix 1045 * itself expires now. 1046 */ 1047 if (newpr->ndpr_raf_onlink == 0) { 1048 newpr->ndpr_vltime = 0; 1049 newpr->ndpr_pltime = 0; 1050 in6_init_prefix_ltimes(newpr); 1051 } 1052 1053 pr = newpr; 1054 } 1055 1056 /* 1057 * Address autoconfiguration based on Section 5.5.3 of RFC 2462. 1058 * Note that pr must be non NULL at this point. 1059 */ 1060 1061 /* 5.5.3 (a). Ignore the prefix without the A bit set. */ 1062 if (!new->ndpr_raf_auto) 1063 goto end; 1064 1065 /* 1066 * 5.5.3 (b). the link-local prefix should have been ignored in 1067 * nd6_ra_input. 1068 */ 1069 1070 /* 5.5.3 (c). Consistency check on lifetimes: pltime <= vltime. */ 1071 if (new->ndpr_pltime > new->ndpr_vltime) { 1072 error = EINVAL; /* XXX: won't be used */ 1073 goto end; 1074 } 1075 1076 /* 1077 * 5.5.3 (d). If the prefix advertised is not equal to the prefix of 1078 * an address configured by stateless autoconfiguration already in the 1079 * list of addresses associated with the interface, and the Valid 1080 * Lifetime is not 0, form an address. We first check if we have 1081 * a matching prefix. 1082 * Note: we apply a clarification in rfc2462bis-02 here. We only 1083 * consider autoconfigured addresses while RFC2462 simply said 1084 * "address". 1085 */ 1086 IF_ADDR_RLOCK(ifp); 1087 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1088 struct in6_ifaddr *ifa6; 1089 u_int32_t remaininglifetime; 1090 1091 if (ifa->ifa_addr->sa_family != AF_INET6) 1092 continue; 1093 1094 ifa6 = (struct in6_ifaddr *)ifa; 1095 1096 /* 1097 * We only consider autoconfigured addresses as per rfc2462bis. 1098 */ 1099 if (!(ifa6->ia6_flags & IN6_IFF_AUTOCONF)) 1100 continue; 1101 1102 /* 1103 * Spec is not clear here, but I believe we should concentrate 1104 * on unicast (i.e. not anycast) addresses. 1105 * XXX: other ia6_flags? detached or duplicated? 1106 */ 1107 if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0) 1108 continue; 1109 1110 /* 1111 * Ignore the address if it is not associated with a prefix 1112 * or is associated with a prefix that is different from this 1113 * one. (pr is never NULL here) 1114 */ 1115 if (ifa6->ia6_ndpr != pr) 1116 continue; 1117 1118 if (ia6_match == NULL) /* remember the first one */ 1119 ia6_match = ifa6; 1120 1121 /* 1122 * An already autoconfigured address matched. Now that we 1123 * are sure there is at least one matched address, we can 1124 * proceed to 5.5.3. (e): update the lifetimes according to the 1125 * "two hours" rule and the privacy extension. 1126 * We apply some clarifications in rfc2462bis: 1127 * - use remaininglifetime instead of storedlifetime as a 1128 * variable name 1129 * - remove the dead code in the "two-hour" rule 1130 */ 1131#define TWOHOUR (120*60) 1132 lt6_tmp = ifa6->ia6_lifetime; 1133 1134 if (lt6_tmp.ia6t_vltime == ND6_INFINITE_LIFETIME) 1135 remaininglifetime = ND6_INFINITE_LIFETIME; 1136 else if (time_uptime - ifa6->ia6_updatetime > 1137 lt6_tmp.ia6t_vltime) { 1138 /* 1139 * The case of "invalid" address. We should usually 1140 * not see this case. 1141 */ 1142 remaininglifetime = 0; 1143 } else 1144 remaininglifetime = lt6_tmp.ia6t_vltime - 1145 (time_uptime - ifa6->ia6_updatetime); 1146 1147 /* when not updating, keep the current stored lifetime. */ 1148 lt6_tmp.ia6t_vltime = remaininglifetime; 1149 1150 if (TWOHOUR < new->ndpr_vltime || 1151 remaininglifetime < new->ndpr_vltime) { 1152 lt6_tmp.ia6t_vltime = new->ndpr_vltime; 1153 } else if (remaininglifetime <= TWOHOUR) { 1154 if (auth) { 1155 lt6_tmp.ia6t_vltime = new->ndpr_vltime; 1156 } 1157 } else { 1158 /* 1159 * new->ndpr_vltime <= TWOHOUR && 1160 * TWOHOUR < remaininglifetime 1161 */ 1162 lt6_tmp.ia6t_vltime = TWOHOUR; 1163 } 1164 1165 /* The 2 hour rule is not imposed for preferred lifetime. */ 1166 lt6_tmp.ia6t_pltime = new->ndpr_pltime; 1167 1168 in6_init_address_ltimes(pr, <6_tmp); 1169 1170 /* 1171 * We need to treat lifetimes for temporary addresses 1172 * differently, according to 1173 * draft-ietf-ipv6-privacy-addrs-v2-01.txt 3.3 (1); 1174 * we only update the lifetimes when they are in the maximum 1175 * intervals. 1176 */ 1177 if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) != 0) { 1178 u_int32_t maxvltime, maxpltime; 1179 1180 if (V_ip6_temp_valid_lifetime > 1181 (u_int32_t)((time_uptime - ifa6->ia6_createtime) + 1182 V_ip6_desync_factor)) { 1183 maxvltime = V_ip6_temp_valid_lifetime - 1184 (time_uptime - ifa6->ia6_createtime) - 1185 V_ip6_desync_factor; 1186 } else 1187 maxvltime = 0; 1188 if (V_ip6_temp_preferred_lifetime > 1189 (u_int32_t)((time_uptime - ifa6->ia6_createtime) + 1190 V_ip6_desync_factor)) { 1191 maxpltime = V_ip6_temp_preferred_lifetime - 1192 (time_uptime - ifa6->ia6_createtime) - 1193 V_ip6_desync_factor; 1194 } else 1195 maxpltime = 0; 1196 1197 if (lt6_tmp.ia6t_vltime == ND6_INFINITE_LIFETIME || 1198 lt6_tmp.ia6t_vltime > maxvltime) { 1199 lt6_tmp.ia6t_vltime = maxvltime; 1200 } 1201 if (lt6_tmp.ia6t_pltime == ND6_INFINITE_LIFETIME || 1202 lt6_tmp.ia6t_pltime > maxpltime) { 1203 lt6_tmp.ia6t_pltime = maxpltime; 1204 } 1205 } 1206 ifa6->ia6_lifetime = lt6_tmp; 1207 ifa6->ia6_updatetime = time_uptime; 1208 } 1209 IF_ADDR_RUNLOCK(ifp); 1210 if (ia6_match == NULL && new->ndpr_vltime) { 1211 int ifidlen; 1212 1213 /* 1214 * 5.5.3 (d) (continued) 1215 * No address matched and the valid lifetime is non-zero. 1216 * Create a new address. 1217 */ 1218 1219 /* 1220 * Prefix Length check: 1221 * If the sum of the prefix length and interface identifier 1222 * length does not equal 128 bits, the Prefix Information 1223 * option MUST be ignored. The length of the interface 1224 * identifier is defined in a separate link-type specific 1225 * document. 1226 */ 1227 ifidlen = in6_if2idlen(ifp); 1228 if (ifidlen < 0) { 1229 /* this should not happen, so we always log it. */ 1230 log(LOG_ERR, "prelist_update: IFID undefined (%s)\n", 1231 if_name(ifp)); 1232 goto end; 1233 } 1234 if (ifidlen + pr->ndpr_plen != 128) { 1235 nd6log((LOG_INFO, 1236 "prelist_update: invalid prefixlen " 1237 "%d for %s, ignored\n", 1238 pr->ndpr_plen, if_name(ifp))); 1239 goto end; 1240 } 1241 1242 if ((ia6 = in6_ifadd(new, mcast)) != NULL) { 1243 /* 1244 * note that we should use pr (not new) for reference. 1245 */ 1246 pr->ndpr_refcnt++; 1247 ia6->ia6_ndpr = pr; 1248 1249 /* 1250 * RFC 3041 3.3 (2). 1251 * When a new public address is created as described 1252 * in RFC2462, also create a new temporary address. 1253 * 1254 * RFC 3041 3.5. 1255 * When an interface connects to a new link, a new 1256 * randomized interface identifier should be generated 1257 * immediately together with a new set of temporary 1258 * addresses. Thus, we specifiy 1 as the 2nd arg of 1259 * in6_tmpifadd(). 1260 */ 1261 if (V_ip6_use_tempaddr) { 1262 int e; 1263 if ((e = in6_tmpifadd(ia6, 1, 1)) != 0) { 1264 nd6log((LOG_NOTICE, "prelist_update: " 1265 "failed to create a temporary " 1266 "address, errno=%d\n", 1267 e)); 1268 } 1269 } 1270 ifa_free(&ia6->ia_ifa); 1271 1272 /* 1273 * A newly added address might affect the status 1274 * of other addresses, so we check and update it. 1275 * XXX: what if address duplication happens? 1276 */ 1277 pfxlist_onlink_check(); 1278 } else { 1279 /* just set an error. do not bark here. */ 1280 error = EADDRNOTAVAIL; /* XXX: might be unused. */ 1281 } 1282 } 1283 1284 end: 1285 return error; 1286} 1287 1288/* 1289 * A supplement function used in the on-link detection below; 1290 * detect if a given prefix has a (probably) reachable advertising router. 1291 * XXX: lengthy function name... 1292 */ 1293static struct nd_pfxrouter * 1294find_pfxlist_reachable_router(struct nd_prefix *pr) 1295{ 1296 struct nd_pfxrouter *pfxrtr; 1297 struct llentry *ln; 1298 int canreach; 1299 1300 LIST_FOREACH(pfxrtr, &pr->ndpr_advrtrs, pfr_entry) { 1301 IF_AFDATA_RLOCK(pfxrtr->router->ifp); 1302 ln = nd6_lookup(&pfxrtr->router->rtaddr, 0, pfxrtr->router->ifp); 1303 IF_AFDATA_RUNLOCK(pfxrtr->router->ifp); 1304 if (ln == NULL) 1305 continue; 1306 canreach = ND6_IS_LLINFO_PROBREACH(ln); 1307 LLE_RUNLOCK(ln); 1308 if (canreach) 1309 break; 1310 } 1311 return (pfxrtr); 1312} 1313 1314/* 1315 * Check if each prefix in the prefix list has at least one available router 1316 * that advertised the prefix (a router is "available" if its neighbor cache 1317 * entry is reachable or probably reachable). 1318 * If the check fails, the prefix may be off-link, because, for example, 1319 * we have moved from the network but the lifetime of the prefix has not 1320 * expired yet. So we should not use the prefix if there is another prefix 1321 * that has an available router. 1322 * But, if there is no prefix that has an available router, we still regards 1323 * all the prefixes as on-link. This is because we can't tell if all the 1324 * routers are simply dead or if we really moved from the network and there 1325 * is no router around us. 1326 */ 1327void 1328pfxlist_onlink_check() 1329{ 1330 struct nd_prefix *pr; 1331 struct in6_ifaddr *ifa; 1332 struct nd_defrouter *dr; 1333 struct nd_pfxrouter *pfxrtr = NULL; 1334 1335 /* 1336 * Check if there is a prefix that has a reachable advertising 1337 * router. 1338 */ 1339 LIST_FOREACH(pr, &V_nd_prefix, ndpr_entry) { 1340 if (pr->ndpr_raf_onlink && find_pfxlist_reachable_router(pr)) 1341 break; 1342 } 1343 1344 /* 1345 * If we have no such prefix, check whether we still have a router 1346 * that does not advertise any prefixes. 1347 */ 1348 if (pr == NULL) { 1349 TAILQ_FOREACH(dr, &V_nd_defrouter, dr_entry) { 1350 struct nd_prefix *pr0; 1351 1352 LIST_FOREACH(pr0, &V_nd_prefix, ndpr_entry) { 1353 if ((pfxrtr = pfxrtr_lookup(pr0, dr)) != NULL) 1354 break; 1355 } 1356 if (pfxrtr != NULL) 1357 break; 1358 } 1359 } 1360 if (pr != NULL || (!TAILQ_EMPTY(&V_nd_defrouter) && pfxrtr == NULL)) { 1361 /* 1362 * There is at least one prefix that has a reachable router, 1363 * or at least a router which probably does not advertise 1364 * any prefixes. The latter would be the case when we move 1365 * to a new link where we have a router that does not provide 1366 * prefixes and we configure an address by hand. 1367 * Detach prefixes which have no reachable advertising 1368 * router, and attach other prefixes. 1369 */ 1370 LIST_FOREACH(pr, &V_nd_prefix, ndpr_entry) { 1371 /* XXX: a link-local prefix should never be detached */ 1372 if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr)) 1373 continue; 1374 1375 /* 1376 * we aren't interested in prefixes without the L bit 1377 * set. 1378 */ 1379 if (pr->ndpr_raf_onlink == 0) 1380 continue; 1381 1382 if (pr->ndpr_raf_auto == 0) 1383 continue; 1384 1385 if ((pr->ndpr_stateflags & NDPRF_DETACHED) == 0 && 1386 find_pfxlist_reachable_router(pr) == NULL) 1387 pr->ndpr_stateflags |= NDPRF_DETACHED; 1388 if ((pr->ndpr_stateflags & NDPRF_DETACHED) != 0 && 1389 find_pfxlist_reachable_router(pr) != 0) 1390 pr->ndpr_stateflags &= ~NDPRF_DETACHED; 1391 } 1392 } else { 1393 /* there is no prefix that has a reachable router */ 1394 LIST_FOREACH(pr, &V_nd_prefix, ndpr_entry) { 1395 if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr)) 1396 continue; 1397 1398 if (pr->ndpr_raf_onlink == 0) 1399 continue; 1400 1401 if (pr->ndpr_raf_auto == 0) 1402 continue; 1403 1404 if ((pr->ndpr_stateflags & NDPRF_DETACHED) != 0) 1405 pr->ndpr_stateflags &= ~NDPRF_DETACHED; 1406 } 1407 } 1408 1409 /* 1410 * Remove each interface route associated with a (just) detached 1411 * prefix, and reinstall the interface route for a (just) attached 1412 * prefix. Note that all attempt of reinstallation does not 1413 * necessarily success, when a same prefix is shared among multiple 1414 * interfaces. Such cases will be handled in nd6_prefix_onlink, 1415 * so we don't have to care about them. 1416 */ 1417 LIST_FOREACH(pr, &V_nd_prefix, ndpr_entry) { 1418 int e; 1419 char ip6buf[INET6_ADDRSTRLEN]; 1420 1421 if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr)) 1422 continue; 1423 1424 if (pr->ndpr_raf_onlink == 0) 1425 continue; 1426 1427 if (pr->ndpr_raf_auto == 0) 1428 continue; 1429 1430 if ((pr->ndpr_stateflags & NDPRF_DETACHED) != 0 && 1431 (pr->ndpr_stateflags & NDPRF_ONLINK) != 0) { 1432 if ((e = nd6_prefix_offlink(pr)) != 0) { 1433 nd6log((LOG_ERR, 1434 "pfxlist_onlink_check: failed to " 1435 "make %s/%d offlink, errno=%d\n", 1436 ip6_sprintf(ip6buf, 1437 &pr->ndpr_prefix.sin6_addr), 1438 pr->ndpr_plen, e)); 1439 } 1440 } 1441 if ((pr->ndpr_stateflags & NDPRF_DETACHED) == 0 && 1442 (pr->ndpr_stateflags & NDPRF_ONLINK) == 0 && 1443 pr->ndpr_raf_onlink) { 1444 if ((e = nd6_prefix_onlink(pr)) != 0) { 1445 nd6log((LOG_ERR, 1446 "pfxlist_onlink_check: failed to " 1447 "make %s/%d onlink, errno=%d\n", 1448 ip6_sprintf(ip6buf, 1449 &pr->ndpr_prefix.sin6_addr), 1450 pr->ndpr_plen, e)); 1451 } 1452 } 1453 } 1454 1455 /* 1456 * Changes on the prefix status might affect address status as well. 1457 * Make sure that all addresses derived from an attached prefix are 1458 * attached, and that all addresses derived from a detached prefix are 1459 * detached. Note, however, that a manually configured address should 1460 * always be attached. 1461 * The precise detection logic is same as the one for prefixes. 1462 * 1463 * XXXRW: in6_ifaddrhead locking. 1464 */ 1465 TAILQ_FOREACH(ifa, &V_in6_ifaddrhead, ia_link) { 1466 if (!(ifa->ia6_flags & IN6_IFF_AUTOCONF)) 1467 continue; 1468 1469 if (ifa->ia6_ndpr == NULL) { 1470 /* 1471 * This can happen when we first configure the address 1472 * (i.e. the address exists, but the prefix does not). 1473 * XXX: complicated relationships... 1474 */ 1475 continue; 1476 } 1477 1478 if (find_pfxlist_reachable_router(ifa->ia6_ndpr)) 1479 break; 1480 } 1481 if (ifa) { 1482 TAILQ_FOREACH(ifa, &V_in6_ifaddrhead, ia_link) { 1483 if ((ifa->ia6_flags & IN6_IFF_AUTOCONF) == 0) 1484 continue; 1485 1486 if (ifa->ia6_ndpr == NULL) /* XXX: see above. */ 1487 continue; 1488 1489 if (find_pfxlist_reachable_router(ifa->ia6_ndpr)) { 1490 if (ifa->ia6_flags & IN6_IFF_DETACHED) { 1491 ifa->ia6_flags &= ~IN6_IFF_DETACHED; 1492 ifa->ia6_flags |= IN6_IFF_TENTATIVE; 1493 nd6_dad_start((struct ifaddr *)ifa, 0); 1494 } 1495 } else { 1496 ifa->ia6_flags |= IN6_IFF_DETACHED; 1497 } 1498 } 1499 } 1500 else { 1501 TAILQ_FOREACH(ifa, &V_in6_ifaddrhead, ia_link) { 1502 if ((ifa->ia6_flags & IN6_IFF_AUTOCONF) == 0) 1503 continue; 1504 1505 if (ifa->ia6_flags & IN6_IFF_DETACHED) { 1506 ifa->ia6_flags &= ~IN6_IFF_DETACHED; 1507 ifa->ia6_flags |= IN6_IFF_TENTATIVE; 1508 /* Do we need a delay in this case? */ 1509 nd6_dad_start((struct ifaddr *)ifa, 0); 1510 } 1511 } 1512 } 1513} 1514 1515static int 1516nd6_prefix_onlink_rtrequest(struct nd_prefix *pr, struct ifaddr *ifa) 1517{ 1518 static struct sockaddr_dl null_sdl = {sizeof(null_sdl), AF_LINK}; 1519 struct radix_node_head *rnh; 1520 struct rtentry *rt; 1521 struct sockaddr_in6 mask6; 1522 u_long rtflags; 1523 int error, a_failure, fibnum; 1524 1525 /* 1526 * in6_ifinit() sets nd6_rtrequest to ifa_rtrequest for all ifaddrs. 1527 * ifa->ifa_rtrequest = nd6_rtrequest; 1528 */ 1529 bzero(&mask6, sizeof(mask6)); 1530 mask6.sin6_len = sizeof(mask6); 1531 mask6.sin6_addr = pr->ndpr_mask; 1532 rtflags = (ifa->ifa_flags & ~IFA_RTSELF) | RTF_UP; 1533 1534 a_failure = 0; 1535 for (fibnum = 0; fibnum < rt_numfibs; fibnum++) { 1536 1537 rt = NULL; 1538 error = in6_rtrequest(RTM_ADD, 1539 (struct sockaddr *)&pr->ndpr_prefix, ifa->ifa_addr, 1540 (struct sockaddr *)&mask6, rtflags, &rt, fibnum); 1541 if (error == 0) { 1542 KASSERT(rt != NULL, ("%s: in6_rtrequest return no " 1543 "error(%d) but rt is NULL, pr=%p, ifa=%p", __func__, 1544 error, pr, ifa)); 1545 1546 rnh = rt_tables_get_rnh(rt->rt_fibnum, AF_INET6); 1547 /* XXX what if rhn == NULL? */ 1548 RADIX_NODE_HEAD_LOCK(rnh); 1549 RT_LOCK(rt); 1550 if (rt_setgate(rt, rt_key(rt), 1551 (struct sockaddr *)&null_sdl) == 0) { 1552 struct sockaddr_dl *dl; 1553 1554 dl = (struct sockaddr_dl *)rt->rt_gateway; 1555 dl->sdl_type = rt->rt_ifp->if_type; 1556 dl->sdl_index = rt->rt_ifp->if_index; 1557 } 1558 RADIX_NODE_HEAD_UNLOCK(rnh); 1559 nd6_rtmsg(RTM_ADD, rt); 1560 RT_UNLOCK(rt); 1561 pr->ndpr_stateflags |= NDPRF_ONLINK; 1562 } else { 1563 char ip6buf[INET6_ADDRSTRLEN]; 1564 char ip6bufg[INET6_ADDRSTRLEN]; 1565 char ip6bufm[INET6_ADDRSTRLEN]; 1566 struct sockaddr_in6 *sin6; 1567 1568 sin6 = (struct sockaddr_in6 *)ifa->ifa_addr; 1569 nd6log((LOG_ERR, "nd6_prefix_onlink: failed to add " 1570 "route for a prefix (%s/%d) on %s, gw=%s, mask=%s, " 1571 "flags=%lx errno = %d\n", 1572 ip6_sprintf(ip6buf, &pr->ndpr_prefix.sin6_addr), 1573 pr->ndpr_plen, if_name(pr->ndpr_ifp), 1574 ip6_sprintf(ip6bufg, &sin6->sin6_addr), 1575 ip6_sprintf(ip6bufm, &mask6.sin6_addr), 1576 rtflags, error)); 1577 1578 /* Save last error to return, see rtinit(). */ 1579 a_failure = error; 1580 } 1581 1582 if (rt != NULL) { 1583 RT_LOCK(rt); 1584 RT_REMREF(rt); 1585 RT_UNLOCK(rt); 1586 } 1587 } 1588 1589 /* Return the last error we got. */ 1590 return (a_failure); 1591} 1592 1593static int 1594nd6_prefix_onlink(struct nd_prefix *pr) 1595{ 1596 struct ifaddr *ifa; 1597 struct ifnet *ifp = pr->ndpr_ifp; 1598 struct nd_prefix *opr; 1599 int error = 0; 1600 char ip6buf[INET6_ADDRSTRLEN]; 1601 1602 /* sanity check */ 1603 if ((pr->ndpr_stateflags & NDPRF_ONLINK) != 0) { 1604 nd6log((LOG_ERR, 1605 "nd6_prefix_onlink: %s/%d is already on-link\n", 1606 ip6_sprintf(ip6buf, &pr->ndpr_prefix.sin6_addr), 1607 pr->ndpr_plen)); 1608 return (EEXIST); 1609 } 1610 1611 /* 1612 * Add the interface route associated with the prefix. Before 1613 * installing the route, check if there's the same prefix on another 1614 * interface, and the prefix has already installed the interface route. 1615 * Although such a configuration is expected to be rare, we explicitly 1616 * allow it. 1617 */ 1618 LIST_FOREACH(opr, &V_nd_prefix, ndpr_entry) { 1619 if (opr == pr) 1620 continue; 1621 1622 if ((opr->ndpr_stateflags & NDPRF_ONLINK) == 0) 1623 continue; 1624 1625 if (opr->ndpr_plen == pr->ndpr_plen && 1626 in6_are_prefix_equal(&pr->ndpr_prefix.sin6_addr, 1627 &opr->ndpr_prefix.sin6_addr, pr->ndpr_plen)) 1628 return (0); 1629 } 1630 1631 /* 1632 * We prefer link-local addresses as the associated interface address. 1633 */ 1634 /* search for a link-local addr */ 1635 ifa = (struct ifaddr *)in6ifa_ifpforlinklocal(ifp, 1636 IN6_IFF_NOTREADY | IN6_IFF_ANYCAST); 1637 if (ifa == NULL) { 1638 /* XXX: freebsd does not have ifa_ifwithaf */ 1639 IF_ADDR_RLOCK(ifp); 1640 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1641 if (ifa->ifa_addr->sa_family == AF_INET6) 1642 break; 1643 } 1644 if (ifa != NULL) 1645 ifa_ref(ifa); 1646 IF_ADDR_RUNLOCK(ifp); 1647 /* should we care about ia6_flags? */ 1648 } 1649 if (ifa == NULL) { 1650 /* 1651 * This can still happen, when, for example, we receive an RA 1652 * containing a prefix with the L bit set and the A bit clear, 1653 * after removing all IPv6 addresses on the receiving 1654 * interface. This should, of course, be rare though. 1655 */ 1656 nd6log((LOG_NOTICE, 1657 "nd6_prefix_onlink: failed to find any ifaddr" 1658 " to add route for a prefix(%s/%d) on %s\n", 1659 ip6_sprintf(ip6buf, &pr->ndpr_prefix.sin6_addr), 1660 pr->ndpr_plen, if_name(ifp))); 1661 return (0); 1662 } 1663 1664 error = nd6_prefix_onlink_rtrequest(pr, ifa); 1665 1666 if (ifa != NULL) 1667 ifa_free(ifa); 1668 1669 return (error); 1670} 1671 1672static int 1673nd6_prefix_offlink(struct nd_prefix *pr) 1674{ 1675 int error = 0; 1676 struct ifnet *ifp = pr->ndpr_ifp; 1677 struct nd_prefix *opr; 1678 struct sockaddr_in6 sa6, mask6; 1679 struct rtentry *rt; 1680 char ip6buf[INET6_ADDRSTRLEN]; 1681 int fibnum, a_failure; 1682 1683 /* sanity check */ 1684 if ((pr->ndpr_stateflags & NDPRF_ONLINK) == 0) { 1685 nd6log((LOG_ERR, 1686 "nd6_prefix_offlink: %s/%d is already off-link\n", 1687 ip6_sprintf(ip6buf, &pr->ndpr_prefix.sin6_addr), 1688 pr->ndpr_plen)); 1689 return (EEXIST); 1690 } 1691 1692 bzero(&sa6, sizeof(sa6)); 1693 sa6.sin6_family = AF_INET6; 1694 sa6.sin6_len = sizeof(sa6); 1695 bcopy(&pr->ndpr_prefix.sin6_addr, &sa6.sin6_addr, 1696 sizeof(struct in6_addr)); 1697 bzero(&mask6, sizeof(mask6)); 1698 mask6.sin6_family = AF_INET6; 1699 mask6.sin6_len = sizeof(sa6); 1700 bcopy(&pr->ndpr_mask, &mask6.sin6_addr, sizeof(struct in6_addr)); 1701 1702 a_failure = 0; 1703 for (fibnum = 0; fibnum < rt_numfibs; fibnum++) { 1704 rt = NULL; 1705 error = in6_rtrequest(RTM_DELETE, (struct sockaddr *)&sa6, NULL, 1706 (struct sockaddr *)&mask6, 0, &rt, fibnum); 1707 if (error == 0) { 1708 /* report the route deletion to the routing socket. */ 1709 if (rt != NULL) 1710 nd6_rtmsg(RTM_DELETE, rt); 1711 } else { 1712 /* Save last error to return, see rtinit(). */ 1713 a_failure = error; 1714 } 1715 if (rt != NULL) { 1716 RTFREE(rt); 1717 } 1718 } 1719 error = a_failure; 1720 a_failure = 1; 1721 if (error == 0) { 1722 pr->ndpr_stateflags &= ~NDPRF_ONLINK; 1723 1724 /* 1725 * There might be the same prefix on another interface, 1726 * the prefix which could not be on-link just because we have 1727 * the interface route (see comments in nd6_prefix_onlink). 1728 * If there's one, try to make the prefix on-link on the 1729 * interface. 1730 */ 1731 LIST_FOREACH(opr, &V_nd_prefix, ndpr_entry) { 1732 if (opr == pr) 1733 continue; 1734 1735 if ((opr->ndpr_stateflags & NDPRF_ONLINK) != 0) 1736 continue; 1737 1738 /* 1739 * KAME specific: detached prefixes should not be 1740 * on-link. 1741 */ 1742 if ((opr->ndpr_stateflags & NDPRF_DETACHED) != 0) 1743 continue; 1744 1745 if (opr->ndpr_plen == pr->ndpr_plen && 1746 in6_are_prefix_equal(&pr->ndpr_prefix.sin6_addr, 1747 &opr->ndpr_prefix.sin6_addr, pr->ndpr_plen)) { 1748 int e; 1749 1750 if ((e = nd6_prefix_onlink(opr)) != 0) { 1751 nd6log((LOG_ERR, 1752 "nd6_prefix_offlink: failed to " 1753 "recover a prefix %s/%d from %s " 1754 "to %s (errno = %d)\n", 1755 ip6_sprintf(ip6buf, 1756 &opr->ndpr_prefix.sin6_addr), 1757 opr->ndpr_plen, if_name(ifp), 1758 if_name(opr->ndpr_ifp), e)); 1759 } else 1760 a_failure = 0; 1761 } 1762 } 1763 } else { 1764 /* XXX: can we still set the NDPRF_ONLINK flag? */ 1765 nd6log((LOG_ERR, 1766 "nd6_prefix_offlink: failed to delete route: " 1767 "%s/%d on %s (errno = %d)\n", 1768 ip6_sprintf(ip6buf, &sa6.sin6_addr), pr->ndpr_plen, 1769 if_name(ifp), error)); 1770 } 1771 1772 if (a_failure) 1773 lltable_prefix_free(AF_INET6, (struct sockaddr *)&sa6, 1774 (struct sockaddr *)&mask6, LLE_STATIC); 1775 1776 return (error); 1777} 1778 1779static struct in6_ifaddr * 1780in6_ifadd(struct nd_prefixctl *pr, int mcast) 1781{ 1782 struct ifnet *ifp = pr->ndpr_ifp; 1783 struct ifaddr *ifa; 1784 struct in6_aliasreq ifra; 1785 struct in6_ifaddr *ia, *ib; 1786 int error, plen0; 1787 struct in6_addr mask; 1788 int prefixlen = pr->ndpr_plen; 1789 int updateflags; 1790 char ip6buf[INET6_ADDRSTRLEN]; 1791 1792 in6_prefixlen2mask(&mask, prefixlen); 1793 1794 /* 1795 * find a link-local address (will be interface ID). 1796 * Is it really mandatory? Theoretically, a global or a site-local 1797 * address can be configured without a link-local address, if we 1798 * have a unique interface identifier... 1799 * 1800 * it is not mandatory to have a link-local address, we can generate 1801 * interface identifier on the fly. we do this because: 1802 * (1) it should be the easiest way to find interface identifier. 1803 * (2) RFC2462 5.4 suggesting the use of the same interface identifier 1804 * for multiple addresses on a single interface, and possible shortcut 1805 * of DAD. we omitted DAD for this reason in the past. 1806 * (3) a user can prevent autoconfiguration of global address 1807 * by removing link-local address by hand (this is partly because we 1808 * don't have other way to control the use of IPv6 on an interface. 1809 * this has been our design choice - cf. NRL's "ifconfig auto"). 1810 * (4) it is easier to manage when an interface has addresses 1811 * with the same interface identifier, than to have multiple addresses 1812 * with different interface identifiers. 1813 */ 1814 ifa = (struct ifaddr *)in6ifa_ifpforlinklocal(ifp, 0); /* 0 is OK? */ 1815 if (ifa) 1816 ib = (struct in6_ifaddr *)ifa; 1817 else 1818 return NULL; 1819 1820 /* prefixlen + ifidlen must be equal to 128 */ 1821 plen0 = in6_mask2len(&ib->ia_prefixmask.sin6_addr, NULL); 1822 if (prefixlen != plen0) { 1823 ifa_free(ifa); 1824 nd6log((LOG_INFO, "in6_ifadd: wrong prefixlen for %s " 1825 "(prefix=%d ifid=%d)\n", 1826 if_name(ifp), prefixlen, 128 - plen0)); 1827 return NULL; 1828 } 1829 1830 /* make ifaddr */ 1831 1832 bzero(&ifra, sizeof(ifra)); 1833 /* 1834 * in6_update_ifa() does not use ifra_name, but we accurately set it 1835 * for safety. 1836 */ 1837 strncpy(ifra.ifra_name, if_name(ifp), sizeof(ifra.ifra_name)); 1838 ifra.ifra_addr.sin6_family = AF_INET6; 1839 ifra.ifra_addr.sin6_len = sizeof(struct sockaddr_in6); 1840 /* prefix */ 1841 ifra.ifra_addr.sin6_addr = pr->ndpr_prefix.sin6_addr; 1842 ifra.ifra_addr.sin6_addr.s6_addr32[0] &= mask.s6_addr32[0]; 1843 ifra.ifra_addr.sin6_addr.s6_addr32[1] &= mask.s6_addr32[1]; 1844 ifra.ifra_addr.sin6_addr.s6_addr32[2] &= mask.s6_addr32[2]; 1845 ifra.ifra_addr.sin6_addr.s6_addr32[3] &= mask.s6_addr32[3]; 1846 1847 /* interface ID */ 1848 ifra.ifra_addr.sin6_addr.s6_addr32[0] |= 1849 (ib->ia_addr.sin6_addr.s6_addr32[0] & ~mask.s6_addr32[0]); 1850 ifra.ifra_addr.sin6_addr.s6_addr32[1] |= 1851 (ib->ia_addr.sin6_addr.s6_addr32[1] & ~mask.s6_addr32[1]); 1852 ifra.ifra_addr.sin6_addr.s6_addr32[2] |= 1853 (ib->ia_addr.sin6_addr.s6_addr32[2] & ~mask.s6_addr32[2]); 1854 ifra.ifra_addr.sin6_addr.s6_addr32[3] |= 1855 (ib->ia_addr.sin6_addr.s6_addr32[3] & ~mask.s6_addr32[3]); 1856 ifa_free(ifa); 1857 1858 /* new prefix mask. */ 1859 ifra.ifra_prefixmask.sin6_len = sizeof(struct sockaddr_in6); 1860 ifra.ifra_prefixmask.sin6_family = AF_INET6; 1861 bcopy(&mask, &ifra.ifra_prefixmask.sin6_addr, 1862 sizeof(ifra.ifra_prefixmask.sin6_addr)); 1863 1864 /* lifetimes. */ 1865 ifra.ifra_lifetime.ia6t_vltime = pr->ndpr_vltime; 1866 ifra.ifra_lifetime.ia6t_pltime = pr->ndpr_pltime; 1867 1868 /* XXX: scope zone ID? */ 1869 1870 ifra.ifra_flags |= IN6_IFF_AUTOCONF; /* obey autoconf */ 1871 1872 /* 1873 * Make sure that we do not have this address already. This should 1874 * usually not happen, but we can still see this case, e.g., if we 1875 * have manually configured the exact address to be configured. 1876 */ 1877 ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp, 1878 &ifra.ifra_addr.sin6_addr); 1879 if (ifa != NULL) { 1880 ifa_free(ifa); 1881 /* this should be rare enough to make an explicit log */ 1882 log(LOG_INFO, "in6_ifadd: %s is already configured\n", 1883 ip6_sprintf(ip6buf, &ifra.ifra_addr.sin6_addr)); 1884 return (NULL); 1885 } 1886 1887 /* 1888 * Allocate ifaddr structure, link into chain, etc. 1889 * If we are going to create a new address upon receiving a multicasted 1890 * RA, we need to impose a random delay before starting DAD. 1891 * [draft-ietf-ipv6-rfc2462bis-02.txt, Section 5.4.2] 1892 */ 1893 updateflags = 0; 1894 if (mcast) 1895 updateflags |= IN6_IFAUPDATE_DADDELAY; 1896 if ((error = in6_update_ifa(ifp, &ifra, NULL, updateflags)) != 0) { 1897 nd6log((LOG_ERR, 1898 "in6_ifadd: failed to make ifaddr %s on %s (errno=%d)\n", 1899 ip6_sprintf(ip6buf, &ifra.ifra_addr.sin6_addr), 1900 if_name(ifp), error)); 1901 return (NULL); /* ifaddr must not have been allocated. */ 1902 } 1903 1904 ia = in6ifa_ifpwithaddr(ifp, &ifra.ifra_addr.sin6_addr); 1905 /* 1906 * XXXRW: Assumption of non-NULLness here might not be true with 1907 * fine-grained locking -- should we validate it? Or just return 1908 * earlier ifa rather than looking it up again? 1909 */ 1910 return (ia); /* this is always non-NULL and referenced. */ 1911} 1912 1913/* 1914 * ia0 - corresponding public address 1915 */ 1916int 1917in6_tmpifadd(const struct in6_ifaddr *ia0, int forcegen, int delay) 1918{ 1919 struct ifnet *ifp = ia0->ia_ifa.ifa_ifp; 1920 struct in6_ifaddr *newia, *ia; 1921 struct in6_aliasreq ifra; 1922 int i, error; 1923 int trylimit = 3; /* XXX: adhoc value */ 1924 int updateflags; 1925 u_int32_t randid[2]; 1926 time_t vltime0, pltime0; 1927 1928 bzero(&ifra, sizeof(ifra)); 1929 strncpy(ifra.ifra_name, if_name(ifp), sizeof(ifra.ifra_name)); 1930 ifra.ifra_addr = ia0->ia_addr; 1931 /* copy prefix mask */ 1932 ifra.ifra_prefixmask = ia0->ia_prefixmask; 1933 /* clear the old IFID */ 1934 for (i = 0; i < 4; i++) { 1935 ifra.ifra_addr.sin6_addr.s6_addr32[i] &= 1936 ifra.ifra_prefixmask.sin6_addr.s6_addr32[i]; 1937 } 1938 1939 again: 1940 if (in6_get_tmpifid(ifp, (u_int8_t *)randid, 1941 (const u_int8_t *)&ia0->ia_addr.sin6_addr.s6_addr[8], forcegen)) { 1942 nd6log((LOG_NOTICE, "in6_tmpifadd: failed to find a good " 1943 "random IFID\n")); 1944 return (EINVAL); 1945 } 1946 ifra.ifra_addr.sin6_addr.s6_addr32[2] |= 1947 (randid[0] & ~(ifra.ifra_prefixmask.sin6_addr.s6_addr32[2])); 1948 ifra.ifra_addr.sin6_addr.s6_addr32[3] |= 1949 (randid[1] & ~(ifra.ifra_prefixmask.sin6_addr.s6_addr32[3])); 1950 1951 /* 1952 * in6_get_tmpifid() quite likely provided a unique interface ID. 1953 * However, we may still have a chance to see collision, because 1954 * there may be a time lag between generation of the ID and generation 1955 * of the address. So, we'll do one more sanity check. 1956 */ 1957 IN6_IFADDR_RLOCK(); 1958 TAILQ_FOREACH(ia, &V_in6_ifaddrhead, ia_link) { 1959 if (IN6_ARE_ADDR_EQUAL(&ia->ia_addr.sin6_addr, 1960 &ifra.ifra_addr.sin6_addr)) { 1961 if (trylimit-- == 0) { 1962 IN6_IFADDR_RUNLOCK(); 1963 /* 1964 * Give up. Something strange should have 1965 * happened. 1966 */ 1967 nd6log((LOG_NOTICE, "in6_tmpifadd: failed to " 1968 "find a unique random IFID\n")); 1969 return (EEXIST); 1970 } 1971 IN6_IFADDR_RUNLOCK(); 1972 forcegen = 1; 1973 goto again; 1974 } 1975 } 1976 IN6_IFADDR_RUNLOCK(); 1977 1978 /* 1979 * The Valid Lifetime is the lower of the Valid Lifetime of the 1980 * public address or TEMP_VALID_LIFETIME. 1981 * The Preferred Lifetime is the lower of the Preferred Lifetime 1982 * of the public address or TEMP_PREFERRED_LIFETIME - 1983 * DESYNC_FACTOR. 1984 */ 1985 if (ia0->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) { 1986 vltime0 = IFA6_IS_INVALID(ia0) ? 0 : 1987 (ia0->ia6_lifetime.ia6t_vltime - 1988 (time_uptime - ia0->ia6_updatetime)); 1989 if (vltime0 > V_ip6_temp_valid_lifetime) 1990 vltime0 = V_ip6_temp_valid_lifetime; 1991 } else 1992 vltime0 = V_ip6_temp_valid_lifetime; 1993 if (ia0->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) { 1994 pltime0 = IFA6_IS_DEPRECATED(ia0) ? 0 : 1995 (ia0->ia6_lifetime.ia6t_pltime - 1996 (time_uptime - ia0->ia6_updatetime)); 1997 if (pltime0 > V_ip6_temp_preferred_lifetime - V_ip6_desync_factor){ 1998 pltime0 = V_ip6_temp_preferred_lifetime - 1999 V_ip6_desync_factor; 2000 } 2001 } else 2002 pltime0 = V_ip6_temp_preferred_lifetime - V_ip6_desync_factor; 2003 ifra.ifra_lifetime.ia6t_vltime = vltime0; 2004 ifra.ifra_lifetime.ia6t_pltime = pltime0; 2005 2006 /* 2007 * A temporary address is created only if this calculated Preferred 2008 * Lifetime is greater than REGEN_ADVANCE time units. 2009 */ 2010 if (ifra.ifra_lifetime.ia6t_pltime <= V_ip6_temp_regen_advance) 2011 return (0); 2012 2013 /* XXX: scope zone ID? */ 2014 2015 ifra.ifra_flags |= (IN6_IFF_AUTOCONF|IN6_IFF_TEMPORARY); 2016 2017 /* allocate ifaddr structure, link into chain, etc. */ 2018 updateflags = 0; 2019 if (delay) 2020 updateflags |= IN6_IFAUPDATE_DADDELAY; 2021 if ((error = in6_update_ifa(ifp, &ifra, NULL, updateflags)) != 0) 2022 return (error); 2023 2024 newia = in6ifa_ifpwithaddr(ifp, &ifra.ifra_addr.sin6_addr); 2025 if (newia == NULL) { /* XXX: can it happen? */ 2026 nd6log((LOG_ERR, 2027 "in6_tmpifadd: ifa update succeeded, but we got " 2028 "no ifaddr\n")); 2029 return (EINVAL); /* XXX */ 2030 } 2031 newia->ia6_ndpr = ia0->ia6_ndpr; 2032 newia->ia6_ndpr->ndpr_refcnt++; 2033 ifa_free(&newia->ia_ifa); 2034 2035 /* 2036 * A newly added address might affect the status of other addresses. 2037 * XXX: when the temporary address is generated with a new public 2038 * address, the onlink check is redundant. However, it would be safe 2039 * to do the check explicitly everywhere a new address is generated, 2040 * and, in fact, we surely need the check when we create a new 2041 * temporary address due to deprecation of an old temporary address. 2042 */ 2043 pfxlist_onlink_check(); 2044 2045 return (0); 2046} 2047 2048static int 2049in6_init_prefix_ltimes(struct nd_prefix *ndpr) 2050{ 2051 if (ndpr->ndpr_pltime == ND6_INFINITE_LIFETIME) 2052 ndpr->ndpr_preferred = 0; 2053 else 2054 ndpr->ndpr_preferred = time_uptime + ndpr->ndpr_pltime; 2055 if (ndpr->ndpr_vltime == ND6_INFINITE_LIFETIME) 2056 ndpr->ndpr_expire = 0; 2057 else 2058 ndpr->ndpr_expire = time_uptime + ndpr->ndpr_vltime; 2059 2060 return 0; 2061} 2062 2063static void 2064in6_init_address_ltimes(struct nd_prefix *new, struct in6_addrlifetime *lt6) 2065{ 2066 /* init ia6t_expire */ 2067 if (lt6->ia6t_vltime == ND6_INFINITE_LIFETIME) 2068 lt6->ia6t_expire = 0; 2069 else { 2070 lt6->ia6t_expire = time_uptime; 2071 lt6->ia6t_expire += lt6->ia6t_vltime; 2072 } 2073 2074 /* init ia6t_preferred */ 2075 if (lt6->ia6t_pltime == ND6_INFINITE_LIFETIME) 2076 lt6->ia6t_preferred = 0; 2077 else { 2078 lt6->ia6t_preferred = time_uptime; 2079 lt6->ia6t_preferred += lt6->ia6t_pltime; 2080 } 2081} 2082 2083/* 2084 * Delete all the routing table entries that use the specified gateway. 2085 * XXX: this function causes search through all entries of routing table, so 2086 * it shouldn't be called when acting as a router. 2087 */ 2088void 2089rt6_flush(struct in6_addr *gateway, struct ifnet *ifp) 2090{ 2091 struct radix_node_head *rnh; 2092 u_int fibnum; 2093 2094 /* We'll care only link-local addresses */ 2095 if (!IN6_IS_ADDR_LINKLOCAL(gateway)) 2096 return; 2097 2098 /* XXX Do we really need to walk any but the default FIB? */ 2099 for (fibnum = 0; fibnum < rt_numfibs; fibnum++) { 2100 rnh = rt_tables_get_rnh(fibnum, AF_INET6); 2101 if (rnh == NULL) 2102 continue; 2103 2104 RADIX_NODE_HEAD_LOCK(rnh); 2105 rnh->rnh_walktree(rnh, rt6_deleteroute, (void *)gateway); 2106 RADIX_NODE_HEAD_UNLOCK(rnh); 2107 } 2108} 2109 2110static int 2111rt6_deleteroute(struct radix_node *rn, void *arg) 2112{ 2113#define SIN6(s) ((struct sockaddr_in6 *)s) 2114 struct rtentry *rt = (struct rtentry *)rn; 2115 struct in6_addr *gate = (struct in6_addr *)arg; 2116 2117 if (rt->rt_gateway == NULL || rt->rt_gateway->sa_family != AF_INET6) 2118 return (0); 2119 2120 if (!IN6_ARE_ADDR_EQUAL(gate, &SIN6(rt->rt_gateway)->sin6_addr)) { 2121 return (0); 2122 } 2123 2124 /* 2125 * Do not delete a static route. 2126 * XXX: this seems to be a bit ad-hoc. Should we consider the 2127 * 'cloned' bit instead? 2128 */ 2129 if ((rt->rt_flags & RTF_STATIC) != 0) 2130 return (0); 2131 2132 /* 2133 * We delete only host route. This means, in particular, we don't 2134 * delete default route. 2135 */ 2136 if ((rt->rt_flags & RTF_HOST) == 0) 2137 return (0); 2138 2139 return (in6_rtrequest(RTM_DELETE, rt_key(rt), rt->rt_gateway, 2140 rt_mask(rt), rt->rt_flags, NULL, rt->rt_fibnum)); 2141#undef SIN6 2142} 2143 2144int 2145nd6_setdefaultiface(int ifindex) 2146{ 2147 int error = 0; 2148 2149 if (ifindex < 0 || V_if_index < ifindex) 2150 return (EINVAL); 2151 if (ifindex != 0 && !ifnet_byindex(ifindex)) 2152 return (EINVAL); 2153 2154 if (V_nd6_defifindex != ifindex) { 2155 V_nd6_defifindex = ifindex; 2156 if (V_nd6_defifindex > 0) 2157 V_nd6_defifp = ifnet_byindex(V_nd6_defifindex); 2158 else 2159 V_nd6_defifp = NULL; 2160 2161 /* 2162 * Our current implementation assumes one-to-one maping between 2163 * interfaces and links, so it would be natural to use the 2164 * default interface as the default link. 2165 */ 2166 scope6_setdefault(V_nd6_defifp); 2167 } 2168 2169 return (error); 2170} 2171