1 /* $FreeBSD$ */ 2 /* $KAME: in6_src.c,v 1.132 2003/08/26 04:42:27 keiichi Exp $ */ 3 4 /*- 5 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project. 6 * All rights reserved. 7 * 8 * Redistribution and use in source and binary forms, with or without 9 * modification, are permitted provided that the following conditions 10 * are met: 11 * 1. Redistributions of source code must retain the above copyright 12 * notice, this list of conditions and the following disclaimer. 13 * 2. Redistributions in binary form must reproduce the above copyright 14 * notice, this list of conditions and the following disclaimer in the 15 * documentation and/or other materials provided with the distribution. 16 * 3. Neither the name of the project nor the names of its contributors 17 * may be used to endorse or promote products derived from this software 18 * without specific prior written permission. 19 * 20 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND 21 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 22 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 23 * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE 24 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 25 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 26 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 27 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 28 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 29 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 30 * SUCH DAMAGE. 31 */ 32 33 /*- 34 * Copyright (c) 1982, 1986, 1991, 1993 35 * The Regents of the University of California. All rights reserved. 36 * 37 * Redistribution and use in source and binary forms, with or without 38 * modification, are permitted provided that the following conditions 39 * are met: 40 * 1. Redistributions of source code must retain the above copyright 41 * notice, this list of conditions and the following disclaimer. 42 * 2. Redistributions in binary form must reproduce the above copyright 43 * notice, this list of conditions and the following disclaimer in the 44 * documentation and/or other materials provided with the distribution. 45 * 4. Neither the name of the University nor the names of its contributors 46 * may be used to endorse or promote products derived from this software 47 * without specific prior written permission. 48 * 49 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 50 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 51 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 52 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 53 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 54 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 55 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 56 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 57 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 58 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 59 * SUCH DAMAGE. 60 * 61 * @(#)in_pcb.c 8.2 (Berkeley) 1/4/94 62 */ 63 64 #include "opt_inet.h" 65 #include "opt_inet6.h" 66 67 #include <sys/param.h> 68 #include <sys/systm.h> 69 #include <sys/lock.h> 70 #include <sys/malloc.h> 71 #include <sys/mbuf.h> 72 #include <sys/priv.h> 73 #include <sys/protosw.h> 74 #include <sys/socket.h> 75 #include <sys/socketvar.h> 76 #include <sys/sockio.h> 77 #include <sys/sysctl.h> 78 #include <sys/errno.h> 79 #include <sys/time.h> 80 #include <sys/kernel.h> 81 #include <sys/sx.h> 82 83 #include <net/if.h> 84 #include <net/route.h> 85 86 #include <netinet/in.h> 87 #include <netinet/in_var.h> 88 #include <netinet/in_systm.h> 89 #include <netinet/ip.h> 90 #include <netinet/in_pcb.h> 91 #include <netinet6/in6_var.h> 92 #include <netinet/ip6.h> 93 #include <netinet6/in6_pcb.h> 94 #include <netinet6/ip6_var.h> 95 #include <netinet6/scope6_var.h> 96 #include <netinet6/nd6.h> 97 98 static struct mtx addrsel_lock; 99 #define ADDRSEL_LOCK_INIT() mtx_init(&addrsel_lock, "addrsel_lock", NULL, MTX_DEF) 100 #define ADDRSEL_LOCK() mtx_lock(&addrsel_lock) 101 #define ADDRSEL_UNLOCK() mtx_unlock(&addrsel_lock) 102 #define ADDRSEL_LOCK_ASSERT() mtx_assert(&addrsel_lock, MA_OWNED) 103 104 static struct sx addrsel_sxlock; 105 #define ADDRSEL_SXLOCK_INIT() sx_init(&addrsel_sxlock, "addrsel_sxlock") 106 #define ADDRSEL_SLOCK() sx_slock(&addrsel_sxlock) 107 #define ADDRSEL_SUNLOCK() sx_sunlock(&addrsel_sxlock) 108 #define ADDRSEL_XLOCK() sx_xlock(&addrsel_sxlock) 109 #define ADDRSEL_XUNLOCK() sx_xunlock(&addrsel_sxlock) 110 111 #define ADDR_LABEL_NOTAPP (-1) 112 struct in6_addrpolicy defaultaddrpolicy; 113 114 int ip6_prefer_tempaddr = 0; 115 116 static int selectroute __P((struct sockaddr_in6 *, struct ip6_pktopts *, 117 struct ip6_moptions *, struct route_in6 *, struct ifnet **, 118 struct rtentry **, int, int)); 119 static int in6_selectif __P((struct sockaddr_in6 *, struct ip6_pktopts *, 120 struct ip6_moptions *, struct route_in6 *ro, struct ifnet **)); 121 122 static struct in6_addrpolicy *lookup_addrsel_policy __P((struct sockaddr_in6 *)); 123 124 static void init_policy_queue __P((void)); 125 static int add_addrsel_policyent __P((struct in6_addrpolicy *)); 126 static int delete_addrsel_policyent __P((struct in6_addrpolicy *)); 127 static int walk_addrsel_policy __P((int (*)(struct in6_addrpolicy *, void *), 128 void *)); 129 static int dump_addrsel_policyent __P((struct in6_addrpolicy *, void *)); 130 static struct in6_addrpolicy *match_addrsel_policy __P((struct sockaddr_in6 *)); 131 132 /* 133 * Return an IPv6 address, which is the most appropriate for a given 134 * destination and user specified options. 135 * If necessary, this function lookups the routing table and returns 136 * an entry to the caller for later use. 137 */ 138 #define REPLACE(r) do {\ 139 if ((r) < sizeof(ip6stat.ip6s_sources_rule) / \ 140 sizeof(ip6stat.ip6s_sources_rule[0])) /* check for safety */ \ 141 ip6stat.ip6s_sources_rule[(r)]++; \ 142 /* printf("in6_selectsrc: replace %s with %s by %d\n", ia_best ? ip6_sprintf(&ia_best->ia_addr.sin6_addr) : "none", ip6_sprintf(&ia->ia_addr.sin6_addr), (r)); */ \ 143 goto replace; \ 144 } while(0) 145 #define NEXT(r) do {\ 146 if ((r) < sizeof(ip6stat.ip6s_sources_rule) / \ 147 sizeof(ip6stat.ip6s_sources_rule[0])) /* check for safety */ \ 148 ip6stat.ip6s_sources_rule[(r)]++; \ 149 /* printf("in6_selectsrc: keep %s against %s by %d\n", ia_best ? ip6_sprintf(&ia_best->ia_addr.sin6_addr) : "none", ip6_sprintf(&ia->ia_addr.sin6_addr), (r)); */ \ 150 goto next; /* XXX: we can't use 'continue' here */ \ 151 } while(0) 152 #define BREAK(r) do { \ 153 if ((r) < sizeof(ip6stat.ip6s_sources_rule) / \ 154 sizeof(ip6stat.ip6s_sources_rule[0])) /* check for safety */ \ 155 ip6stat.ip6s_sources_rule[(r)]++; \ 156 goto out; /* XXX: we can't use 'break' here */ \ 157 } while(0) 158 159 struct in6_addr * 160 in6_selectsrc(dstsock, opts, mopts, ro, laddr, ifpp, errorp) 161 struct sockaddr_in6 *dstsock; 162 struct ip6_pktopts *opts; 163 struct ip6_moptions *mopts; 164 struct route_in6 *ro; 165 struct in6_addr *laddr; 166 struct ifnet **ifpp; 167 int *errorp; 168 { 169 struct in6_addr dst; 170 struct ifnet *ifp = NULL; 171 struct in6_ifaddr *ia = NULL, *ia_best = NULL; 172 struct in6_pktinfo *pi = NULL; 173 int dst_scope = -1, best_scope = -1, best_matchlen = -1; 174 struct in6_addrpolicy *dst_policy = NULL, *best_policy = NULL; 175 u_int32_t odstzone; 176 int prefer_tempaddr; 177 178 dst = dstsock->sin6_addr; /* make a copy for local operation */ 179 *errorp = 0; 180 if (ifpp) 181 *ifpp = NULL; 182 183 /* 184 * If the source address is explicitly specified by the caller, 185 * check if the requested source address is indeed a unicast address 186 * assigned to the node, and can be used as the packet's source 187 * address. If everything is okay, use the address as source. 188 */ 189 if (opts && (pi = opts->ip6po_pktinfo) && 190 !IN6_IS_ADDR_UNSPECIFIED(&pi->ipi6_addr)) { 191 struct sockaddr_in6 srcsock; 192 struct in6_ifaddr *ia6; 193 194 /* get the outgoing interface */ 195 if ((*errorp = in6_selectif(dstsock, opts, mopts, ro, &ifp)) 196 != 0) { 197 return (NULL); 198 } 199 200 /* 201 * determine the appropriate zone id of the source based on 202 * the zone of the destination and the outgoing interface. 203 * If the specified address is ambiguous wrt the scope zone, 204 * the interface must be specified; otherwise, ifa_ifwithaddr() 205 * will fail matching the address. 206 */ 207 bzero(&srcsock, sizeof(srcsock)); 208 srcsock.sin6_family = AF_INET6; 209 srcsock.sin6_len = sizeof(srcsock); 210 srcsock.sin6_addr = pi->ipi6_addr; 211 if (ifp) { 212 *errorp = in6_setscope(&srcsock.sin6_addr, ifp, NULL); 213 if (*errorp != 0) 214 return (NULL); 215 } 216 217 ia6 = (struct in6_ifaddr *)ifa_ifwithaddr((struct sockaddr *)(&srcsock)); 218 if (ia6 == NULL || 219 (ia6->ia6_flags & (IN6_IFF_ANYCAST | IN6_IFF_NOTREADY))) { 220 *errorp = EADDRNOTAVAIL; 221 return (NULL); 222 } 223 pi->ipi6_addr = srcsock.sin6_addr; /* XXX: this overrides pi */ 224 if (ifpp) 225 *ifpp = ifp; 226 return (&ia6->ia_addr.sin6_addr); 227 } 228 229 /* 230 * Otherwise, if the socket has already bound the source, just use it. 231 */ 232 if (laddr && !IN6_IS_ADDR_UNSPECIFIED(laddr)) 233 return (laddr); 234 235 /* 236 * If the address is not specified, choose the best one based on 237 * the outgoing interface and the destination address. 238 */ 239 /* get the outgoing interface */ 240 if ((*errorp = in6_selectif(dstsock, opts, mopts, ro, &ifp)) != 0) 241 return (NULL); 242 243 #ifdef DIAGNOSTIC 244 if (ifp == NULL) /* this should not happen */ 245 panic("in6_selectsrc: NULL ifp"); 246 #endif 247 *errorp = in6_setscope(&dst, ifp, &odstzone); 248 if (*errorp != 0) 249 return (NULL); 250 251 for (ia = in6_ifaddr; ia; ia = ia->ia_next) { 252 int new_scope = -1, new_matchlen = -1; 253 struct in6_addrpolicy *new_policy = NULL; 254 u_int32_t srczone, osrczone, dstzone; 255 struct in6_addr src; 256 struct ifnet *ifp1 = ia->ia_ifp; 257 258 /* 259 * We'll never take an address that breaks the scope zone 260 * of the destination. We also skip an address if its zone 261 * does not contain the outgoing interface. 262 * XXX: we should probably use sin6_scope_id here. 263 */ 264 if (in6_setscope(&dst, ifp1, &dstzone) || 265 odstzone != dstzone) { 266 continue; 267 } 268 src = ia->ia_addr.sin6_addr; 269 if (in6_setscope(&src, ifp, &osrczone) || 270 in6_setscope(&src, ifp1, &srczone) || 271 osrczone != srczone) { 272 continue; 273 } 274 275 /* avoid unusable addresses */ 276 if ((ia->ia6_flags & 277 (IN6_IFF_NOTREADY | IN6_IFF_ANYCAST | IN6_IFF_DETACHED))) { 278 continue; 279 } 280 if (!ip6_use_deprecated && IFA6_IS_DEPRECATED(ia)) 281 continue; 282 283 /* Rule 1: Prefer same address */ 284 if (IN6_ARE_ADDR_EQUAL(&dst, &ia->ia_addr.sin6_addr)) { 285 ia_best = ia; 286 BREAK(1); /* there should be no better candidate */ 287 } 288 289 if (ia_best == NULL) 290 REPLACE(0); 291 292 /* Rule 2: Prefer appropriate scope */ 293 if (dst_scope < 0) 294 dst_scope = in6_addrscope(&dst); 295 new_scope = in6_addrscope(&ia->ia_addr.sin6_addr); 296 if (IN6_ARE_SCOPE_CMP(best_scope, new_scope) < 0) { 297 if (IN6_ARE_SCOPE_CMP(best_scope, dst_scope) < 0) 298 REPLACE(2); 299 NEXT(2); 300 } else if (IN6_ARE_SCOPE_CMP(new_scope, best_scope) < 0) { 301 if (IN6_ARE_SCOPE_CMP(new_scope, dst_scope) < 0) 302 NEXT(2); 303 REPLACE(2); 304 } 305 306 /* 307 * Rule 3: Avoid deprecated addresses. Note that the case of 308 * !ip6_use_deprecated is already rejected above. 309 */ 310 if (!IFA6_IS_DEPRECATED(ia_best) && IFA6_IS_DEPRECATED(ia)) 311 NEXT(3); 312 if (IFA6_IS_DEPRECATED(ia_best) && !IFA6_IS_DEPRECATED(ia)) 313 REPLACE(3); 314 315 /* Rule 4: Prefer home addresses */ 316 /* 317 * XXX: This is a TODO. We should probably merge the MIP6 318 * case above. 319 */ 320 321 /* Rule 5: Prefer outgoing interface */ 322 if (ia_best->ia_ifp == ifp && ia->ia_ifp != ifp) 323 NEXT(5); 324 if (ia_best->ia_ifp != ifp && ia->ia_ifp == ifp) 325 REPLACE(5); 326 327 /* 328 * Rule 6: Prefer matching label 329 * Note that best_policy should be non-NULL here. 330 */ 331 if (dst_policy == NULL) 332 dst_policy = lookup_addrsel_policy(dstsock); 333 if (dst_policy->label != ADDR_LABEL_NOTAPP) { 334 new_policy = lookup_addrsel_policy(&ia->ia_addr); 335 if (dst_policy->label == best_policy->label && 336 dst_policy->label != new_policy->label) 337 NEXT(6); 338 if (dst_policy->label != best_policy->label && 339 dst_policy->label == new_policy->label) 340 REPLACE(6); 341 } 342 343 /* 344 * Rule 7: Prefer public addresses. 345 * We allow users to reverse the logic by configuring 346 * a sysctl variable, so that privacy conscious users can 347 * always prefer temporary addresses. 348 */ 349 if (opts == NULL || 350 opts->ip6po_prefer_tempaddr == IP6PO_TEMPADDR_SYSTEM) { 351 prefer_tempaddr = ip6_prefer_tempaddr; 352 } else if (opts->ip6po_prefer_tempaddr == 353 IP6PO_TEMPADDR_NOTPREFER) { 354 prefer_tempaddr = 0; 355 } else 356 prefer_tempaddr = 1; 357 if (!(ia_best->ia6_flags & IN6_IFF_TEMPORARY) && 358 (ia->ia6_flags & IN6_IFF_TEMPORARY)) { 359 if (prefer_tempaddr) 360 REPLACE(7); 361 else 362 NEXT(7); 363 } 364 if ((ia_best->ia6_flags & IN6_IFF_TEMPORARY) && 365 !(ia->ia6_flags & IN6_IFF_TEMPORARY)) { 366 if (prefer_tempaddr) 367 NEXT(7); 368 else 369 REPLACE(7); 370 } 371 372 /* 373 * Rule 8: prefer addresses on alive interfaces. 374 * This is a KAME specific rule. 375 */ 376 if ((ia_best->ia_ifp->if_flags & IFF_UP) && 377 !(ia->ia_ifp->if_flags & IFF_UP)) 378 NEXT(8); 379 if (!(ia_best->ia_ifp->if_flags & IFF_UP) && 380 (ia->ia_ifp->if_flags & IFF_UP)) 381 REPLACE(8); 382 383 /* 384 * Rule 14: Use longest matching prefix. 385 * Note: in the address selection draft, this rule is 386 * documented as "Rule 8". However, since it is also 387 * documented that this rule can be overridden, we assign 388 * a large number so that it is easy to assign smaller numbers 389 * to more preferred rules. 390 */ 391 new_matchlen = in6_matchlen(&ia->ia_addr.sin6_addr, &dst); 392 if (best_matchlen < new_matchlen) 393 REPLACE(14); 394 if (new_matchlen < best_matchlen) 395 NEXT(14); 396 397 /* Rule 15 is reserved. */ 398 399 /* 400 * Last resort: just keep the current candidate. 401 * Or, do we need more rules? 402 */ 403 continue; 404 405 replace: 406 ia_best = ia; 407 best_scope = (new_scope >= 0 ? new_scope : 408 in6_addrscope(&ia_best->ia_addr.sin6_addr)); 409 best_policy = (new_policy ? new_policy : 410 lookup_addrsel_policy(&ia_best->ia_addr)); 411 best_matchlen = (new_matchlen >= 0 ? new_matchlen : 412 in6_matchlen(&ia_best->ia_addr.sin6_addr, 413 &dst)); 414 415 next: 416 continue; 417 418 out: 419 break; 420 } 421 422 if ((ia = ia_best) == NULL) { 423 *errorp = EADDRNOTAVAIL; 424 return (NULL); 425 } 426 427 if (ifpp) 428 *ifpp = ifp; 429 430 return (&ia->ia_addr.sin6_addr); 431 } 432 433 static int 434 selectroute(dstsock, opts, mopts, ro, retifp, retrt, clone, norouteok) 435 struct sockaddr_in6 *dstsock; 436 struct ip6_pktopts *opts; 437 struct ip6_moptions *mopts; 438 struct route_in6 *ro; 439 struct ifnet **retifp; 440 struct rtentry **retrt; 441 int clone; /* meaningful only for bsdi and freebsd. */ 442 int norouteok; 443 { 444 int error = 0; 445 struct ifnet *ifp = NULL; 446 struct rtentry *rt = NULL; 447 struct sockaddr_in6 *sin6_next; 448 struct in6_pktinfo *pi = NULL; 449 struct in6_addr *dst = &dstsock->sin6_addr; 450 #if 0 451 char ip6buf[INET6_ADDRSTRLEN]; 452 453 if (dstsock->sin6_addr.s6_addr32[0] == 0 && 454 dstsock->sin6_addr.s6_addr32[1] == 0 && 455 !IN6_IS_ADDR_LOOPBACK(&dstsock->sin6_addr)) { 456 printf("in6_selectroute: strange destination %s\n", 457 ip6_sprintf(ip6buf, &dstsock->sin6_addr)); 458 } else { 459 printf("in6_selectroute: destination = %s%%%d\n", 460 ip6_sprintf(ip6buf, &dstsock->sin6_addr), 461 dstsock->sin6_scope_id); /* for debug */ 462 } 463 #endif 464 465 /* If the caller specify the outgoing interface explicitly, use it. */ 466 if (opts && (pi = opts->ip6po_pktinfo) != NULL && pi->ipi6_ifindex) { 467 /* XXX boundary check is assumed to be already done. */ 468 ifp = ifnet_byindex(pi->ipi6_ifindex); 469 if (ifp != NULL && 470 (norouteok || retrt == NULL || 471 IN6_IS_ADDR_MULTICAST(dst))) { 472 /* 473 * we do not have to check or get the route for 474 * multicast. 475 */ 476 goto done; 477 } else 478 goto getroute; 479 } 480 481 /* 482 * If the destination address is a multicast address and the outgoing 483 * interface for the address is specified by the caller, use it. 484 */ 485 if (IN6_IS_ADDR_MULTICAST(dst) && 486 mopts != NULL && (ifp = mopts->im6o_multicast_ifp) != NULL) { 487 goto done; /* we do not need a route for multicast. */ 488 } 489 490 getroute: 491 /* 492 * If the next hop address for the packet is specified by the caller, 493 * use it as the gateway. 494 */ 495 if (opts && opts->ip6po_nexthop) { 496 struct route_in6 *ron; 497 498 sin6_next = satosin6(opts->ip6po_nexthop); 499 500 /* at this moment, we only support AF_INET6 next hops */ 501 if (sin6_next->sin6_family != AF_INET6) { 502 error = EAFNOSUPPORT; /* or should we proceed? */ 503 goto done; 504 } 505 506 /* 507 * If the next hop is an IPv6 address, then the node identified 508 * by that address must be a neighbor of the sending host. 509 */ 510 ron = &opts->ip6po_nextroute; 511 if ((ron->ro_rt && 512 (ron->ro_rt->rt_flags & (RTF_UP | RTF_LLINFO)) != 513 (RTF_UP | RTF_LLINFO)) || 514 !IN6_ARE_ADDR_EQUAL(&satosin6(&ron->ro_dst)->sin6_addr, 515 &sin6_next->sin6_addr)) { 516 if (ron->ro_rt) { 517 RTFREE(ron->ro_rt); 518 ron->ro_rt = NULL; 519 } 520 *satosin6(&ron->ro_dst) = *sin6_next; 521 } 522 if (ron->ro_rt == NULL) { 523 rtalloc((struct route *)ron); /* multi path case? */ 524 if (ron->ro_rt == NULL || 525 !(ron->ro_rt->rt_flags & RTF_LLINFO)) { 526 if (ron->ro_rt) { 527 RTFREE(ron->ro_rt); 528 ron->ro_rt = NULL; 529 } 530 error = EHOSTUNREACH; 531 goto done; 532 } 533 } 534 rt = ron->ro_rt; 535 ifp = rt->rt_ifp; 536 537 /* 538 * When cloning is required, try to allocate a route to the 539 * destination so that the caller can store path MTU 540 * information. 541 */ 542 if (!clone) 543 goto done; 544 } 545 546 /* 547 * Use a cached route if it exists and is valid, else try to allocate 548 * a new one. Note that we should check the address family of the 549 * cached destination, in case of sharing the cache with IPv4. 550 */ 551 if (ro) { 552 if (ro->ro_rt && 553 (!(ro->ro_rt->rt_flags & RTF_UP) || 554 ((struct sockaddr *)(&ro->ro_dst))->sa_family != AF_INET6 || 555 !IN6_ARE_ADDR_EQUAL(&satosin6(&ro->ro_dst)->sin6_addr, 556 dst))) { 557 RTFREE(ro->ro_rt); 558 ro->ro_rt = (struct rtentry *)NULL; 559 } 560 if (ro->ro_rt == (struct rtentry *)NULL) { 561 struct sockaddr_in6 *sa6; 562 563 /* No route yet, so try to acquire one */ 564 bzero(&ro->ro_dst, sizeof(struct sockaddr_in6)); 565 sa6 = (struct sockaddr_in6 *)&ro->ro_dst; 566 *sa6 = *dstsock; 567 sa6->sin6_scope_id = 0; 568 569 if (clone) { 570 rtalloc((struct route *)ro); 571 } else { 572 ro->ro_rt = rtalloc1(&((struct route *)ro) 573 ->ro_dst, 0, 0UL); 574 if (ro->ro_rt) 575 RT_UNLOCK(ro->ro_rt); 576 } 577 } 578 579 /* 580 * do not care about the result if we have the nexthop 581 * explicitly specified. 582 */ 583 if (opts && opts->ip6po_nexthop) 584 goto done; 585 586 if (ro->ro_rt) { 587 ifp = ro->ro_rt->rt_ifp; 588 589 if (ifp == NULL) { /* can this really happen? */ 590 RTFREE(ro->ro_rt); 591 ro->ro_rt = NULL; 592 } 593 } 594 if (ro->ro_rt == NULL) 595 error = EHOSTUNREACH; 596 rt = ro->ro_rt; 597 598 /* 599 * Check if the outgoing interface conflicts with 600 * the interface specified by ipi6_ifindex (if specified). 601 * Note that loopback interface is always okay. 602 * (this may happen when we are sending a packet to one of 603 * our own addresses.) 604 */ 605 if (ifp && opts && opts->ip6po_pktinfo && 606 opts->ip6po_pktinfo->ipi6_ifindex) { 607 if (!(ifp->if_flags & IFF_LOOPBACK) && 608 ifp->if_index != 609 opts->ip6po_pktinfo->ipi6_ifindex) { 610 error = EHOSTUNREACH; 611 goto done; 612 } 613 } 614 } 615 616 done: 617 if (ifp == NULL && rt == NULL) { 618 /* 619 * This can happen if the caller did not pass a cached route 620 * nor any other hints. We treat this case an error. 621 */ 622 error = EHOSTUNREACH; 623 } 624 if (error == EHOSTUNREACH) 625 ip6stat.ip6s_noroute++; 626 627 if (retifp != NULL) 628 *retifp = ifp; 629 if (retrt != NULL) 630 *retrt = rt; /* rt may be NULL */ 631 632 return (error); 633 } 634 635 static int 636 in6_selectif(dstsock, opts, mopts, ro, retifp) 637 struct sockaddr_in6 *dstsock; 638 struct ip6_pktopts *opts; 639 struct ip6_moptions *mopts; 640 struct route_in6 *ro; 641 struct ifnet **retifp; 642 { 643 int error; 644 struct route_in6 sro; 645 struct rtentry *rt = NULL; 646 647 if (ro == NULL) { 648 bzero(&sro, sizeof(sro)); 649 ro = &sro; 650 } 651 652 if ((error = selectroute(dstsock, opts, mopts, ro, retifp, 653 &rt, 0, 1)) != 0) { 654 if (ro == &sro && rt && rt == sro.ro_rt) 655 RTFREE(rt); 656 return (error); 657 } 658 659 /* 660 * do not use a rejected or black hole route. 661 * XXX: this check should be done in the L2 output routine. 662 * However, if we skipped this check here, we'd see the following 663 * scenario: 664 * - install a rejected route for a scoped address prefix 665 * (like fe80::/10) 666 * - send a packet to a destination that matches the scoped prefix, 667 * with ambiguity about the scope zone. 668 * - pick the outgoing interface from the route, and disambiguate the 669 * scope zone with the interface. 670 * - ip6_output() would try to get another route with the "new" 671 * destination, which may be valid. 672 * - we'd see no error on output. 673 * Although this may not be very harmful, it should still be confusing. 674 * We thus reject the case here. 675 */ 676 if (rt && (rt->rt_flags & (RTF_REJECT | RTF_BLACKHOLE))) { 677 int flags = (rt->rt_flags & RTF_HOST ? EHOSTUNREACH : ENETUNREACH); 678 679 if (ro == &sro && rt && rt == sro.ro_rt) 680 RTFREE(rt); 681 return (flags); 682 } 683 684 /* 685 * Adjust the "outgoing" interface. If we're going to loop the packet 686 * back to ourselves, the ifp would be the loopback interface. 687 * However, we'd rather know the interface associated to the 688 * destination address (which should probably be one of our own 689 * addresses.) 690 */ 691 if (rt && rt->rt_ifa && rt->rt_ifa->ifa_ifp) 692 *retifp = rt->rt_ifa->ifa_ifp; 693 694 if (ro == &sro && rt && rt == sro.ro_rt) 695 RTFREE(rt); 696 return (0); 697 } 698 699 int 700 in6_selectroute(dstsock, opts, mopts, ro, retifp, retrt, clone) 701 struct sockaddr_in6 *dstsock; 702 struct ip6_pktopts *opts; 703 struct ip6_moptions *mopts; 704 struct route_in6 *ro; 705 struct ifnet **retifp; 706 struct rtentry **retrt; 707 int clone; /* meaningful only for bsdi and freebsd. */ 708 { 709 return (selectroute(dstsock, opts, mopts, ro, retifp, 710 retrt, clone, 0)); 711 } 712 713 /* 714 * Default hop limit selection. The precedence is as follows: 715 * 1. Hoplimit value specified via ioctl. 716 * 2. (If the outgoing interface is detected) the current 717 * hop limit of the interface specified by router advertisement. 718 * 3. The system default hoplimit. 719 */ 720 int 721 in6_selecthlim(in6p, ifp) 722 struct in6pcb *in6p; 723 struct ifnet *ifp; 724 { 725 if (in6p && in6p->in6p_hops >= 0) 726 return (in6p->in6p_hops); 727 else if (ifp) 728 return (ND_IFINFO(ifp)->chlim); 729 else if (in6p && !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr)) { 730 struct route_in6 ro6; 731 struct ifnet *lifp; 732 733 bzero(&ro6, sizeof(ro6)); 734 ro6.ro_dst.sin6_family = AF_INET6; 735 ro6.ro_dst.sin6_len = sizeof(struct sockaddr_in6); 736 ro6.ro_dst.sin6_addr = in6p->in6p_faddr; 737 rtalloc((struct route *)&ro6); 738 if (ro6.ro_rt) { 739 lifp = ro6.ro_rt->rt_ifp; 740 RTFREE(ro6.ro_rt); 741 if (lifp) 742 return (ND_IFINFO(lifp)->chlim); 743 } else 744 return (ip6_defhlim); 745 } 746 return (ip6_defhlim); 747 } 748 749 /* 750 * XXX: this is borrowed from in6_pcbbind(). If possible, we should 751 * share this function by all *bsd*... 752 */ 753 int 754 in6_pcbsetport(laddr, inp, cred) 755 struct in6_addr *laddr; 756 struct inpcb *inp; 757 struct ucred *cred; 758 { 759 struct socket *so = inp->inp_socket; 760 u_int16_t lport = 0, first, last, *lastport; 761 int count, error = 0, wild = 0; 762 struct inpcbinfo *pcbinfo = inp->inp_pcbinfo; 763 764 INP_INFO_WLOCK_ASSERT(pcbinfo); 765 INP_LOCK_ASSERT(inp); 766 767 /* XXX: this is redundant when called from in6_pcbbind */ 768 if ((so->so_options & (SO_REUSEADDR|SO_REUSEPORT)) == 0) 769 wild = INPLOOKUP_WILDCARD; 770 771 inp->inp_flags |= INP_ANONPORT; 772 773 if (inp->inp_flags & INP_HIGHPORT) { 774 first = ipport_hifirstauto; /* sysctl */ 775 last = ipport_hilastauto; 776 lastport = &pcbinfo->lasthi; 777 } else if (inp->inp_flags & INP_LOWPORT) { 778 error = priv_check_cred(cred, PRIV_NETINET_RESERVEDPORT, 779 SUSER_ALLOWJAIL); 780 if (error) 781 return error; 782 first = ipport_lowfirstauto; /* 1023 */ 783 last = ipport_lowlastauto; /* 600 */ 784 lastport = &pcbinfo->lastlow; 785 } else { 786 first = ipport_firstauto; /* sysctl */ 787 last = ipport_lastauto; 788 lastport = &pcbinfo->lastport; 789 } 790 /* 791 * Simple check to ensure all ports are not used up causing 792 * a deadlock here. 793 * 794 * We split the two cases (up and down) so that the direction 795 * is not being tested on each round of the loop. 796 */ 797 if (first > last) { 798 /* 799 * counting down 800 */ 801 count = first - last; 802 803 do { 804 if (count-- < 0) { /* completely used? */ 805 /* 806 * Undo any address bind that may have 807 * occurred above. 808 */ 809 inp->in6p_laddr = in6addr_any; 810 return (EAGAIN); 811 } 812 --*lastport; 813 if (*lastport > first || *lastport < last) 814 *lastport = first; 815 lport = htons(*lastport); 816 } while (in6_pcblookup_local(pcbinfo, &inp->in6p_laddr, 817 lport, wild)); 818 } else { 819 /* 820 * counting up 821 */ 822 count = last - first; 823 824 do { 825 if (count-- < 0) { /* completely used? */ 826 /* 827 * Undo any address bind that may have 828 * occurred above. 829 */ 830 inp->in6p_laddr = in6addr_any; 831 return (EAGAIN); 832 } 833 ++*lastport; 834 if (*lastport < first || *lastport > last) 835 *lastport = first; 836 lport = htons(*lastport); 837 } while (in6_pcblookup_local(pcbinfo, 838 &inp->in6p_laddr, lport, wild)); 839 } 840 841 inp->inp_lport = lport; 842 if (in_pcbinshash(inp) != 0) { 843 inp->in6p_laddr = in6addr_any; 844 inp->inp_lport = 0; 845 return (EAGAIN); 846 } 847 848 return (0); 849 } 850 851 void 852 addrsel_policy_init() 853 { 854 ADDRSEL_LOCK_INIT(); 855 ADDRSEL_SXLOCK_INIT(); 856 857 init_policy_queue(); 858 859 /* initialize the "last resort" policy */ 860 bzero(&defaultaddrpolicy, sizeof(defaultaddrpolicy)); 861 defaultaddrpolicy.label = ADDR_LABEL_NOTAPP; 862 } 863 864 static struct in6_addrpolicy * 865 lookup_addrsel_policy(key) 866 struct sockaddr_in6 *key; 867 { 868 struct in6_addrpolicy *match = NULL; 869 870 ADDRSEL_LOCK(); 871 match = match_addrsel_policy(key); 872 873 if (match == NULL) 874 match = &defaultaddrpolicy; 875 else 876 match->use++; 877 ADDRSEL_UNLOCK(); 878 879 return (match); 880 } 881 882 /* 883 * Subroutines to manage the address selection policy table via sysctl. 884 */ 885 struct walkarg { 886 struct sysctl_req *w_req; 887 }; 888 889 static int in6_src_sysctl(SYSCTL_HANDLER_ARGS); 890 SYSCTL_DECL(_net_inet6_ip6); 891 SYSCTL_NODE(_net_inet6_ip6, IPV6CTL_ADDRCTLPOLICY, addrctlpolicy, 892 CTLFLAG_RD, in6_src_sysctl, ""); 893 894 static int 895 in6_src_sysctl(SYSCTL_HANDLER_ARGS) 896 { 897 struct walkarg w; 898 899 if (req->newptr) 900 return EPERM; 901 902 bzero(&w, sizeof(w)); 903 w.w_req = req; 904 905 return (walk_addrsel_policy(dump_addrsel_policyent, &w)); 906 } 907 908 int 909 in6_src_ioctl(cmd, data) 910 u_long cmd; 911 caddr_t data; 912 { 913 int i; 914 struct in6_addrpolicy ent0; 915 916 if (cmd != SIOCAADDRCTL_POLICY && cmd != SIOCDADDRCTL_POLICY) 917 return (EOPNOTSUPP); /* check for safety */ 918 919 ent0 = *(struct in6_addrpolicy *)data; 920 921 if (ent0.label == ADDR_LABEL_NOTAPP) 922 return (EINVAL); 923 /* check if the prefix mask is consecutive. */ 924 if (in6_mask2len(&ent0.addrmask.sin6_addr, NULL) < 0) 925 return (EINVAL); 926 /* clear trailing garbages (if any) of the prefix address. */ 927 for (i = 0; i < 4; i++) { 928 ent0.addr.sin6_addr.s6_addr32[i] &= 929 ent0.addrmask.sin6_addr.s6_addr32[i]; 930 } 931 ent0.use = 0; 932 933 switch (cmd) { 934 case SIOCAADDRCTL_POLICY: 935 return (add_addrsel_policyent(&ent0)); 936 case SIOCDADDRCTL_POLICY: 937 return (delete_addrsel_policyent(&ent0)); 938 } 939 940 return (0); /* XXX: compromise compilers */ 941 } 942 943 /* 944 * The followings are implementation of the policy table using a 945 * simple tail queue. 946 * XXX such details should be hidden. 947 * XXX implementation using binary tree should be more efficient. 948 */ 949 struct addrsel_policyent { 950 TAILQ_ENTRY(addrsel_policyent) ape_entry; 951 struct in6_addrpolicy ape_policy; 952 }; 953 954 TAILQ_HEAD(addrsel_policyhead, addrsel_policyent); 955 956 struct addrsel_policyhead addrsel_policytab; 957 958 static void 959 init_policy_queue() 960 { 961 TAILQ_INIT(&addrsel_policytab); 962 } 963 964 static int 965 add_addrsel_policyent(newpolicy) 966 struct in6_addrpolicy *newpolicy; 967 { 968 struct addrsel_policyent *new, *pol; 969 970 MALLOC(new, struct addrsel_policyent *, sizeof(*new), M_IFADDR, 971 M_WAITOK); 972 ADDRSEL_XLOCK(); 973 ADDRSEL_LOCK(); 974 975 /* duplication check */ 976 TAILQ_FOREACH(pol, &addrsel_policytab, ape_entry) { 977 if (IN6_ARE_ADDR_EQUAL(&newpolicy->addr.sin6_addr, 978 &pol->ape_policy.addr.sin6_addr) && 979 IN6_ARE_ADDR_EQUAL(&newpolicy->addrmask.sin6_addr, 980 &pol->ape_policy.addrmask.sin6_addr)) { 981 ADDRSEL_UNLOCK(); 982 ADDRSEL_XUNLOCK(); 983 FREE(new, M_IFADDR); 984 return (EEXIST); /* or override it? */ 985 } 986 } 987 988 bzero(new, sizeof(*new)); 989 990 /* XXX: should validate entry */ 991 new->ape_policy = *newpolicy; 992 993 TAILQ_INSERT_TAIL(&addrsel_policytab, new, ape_entry); 994 ADDRSEL_UNLOCK(); 995 ADDRSEL_XUNLOCK(); 996 997 return (0); 998 } 999 1000 static int 1001 delete_addrsel_policyent(key) 1002 struct in6_addrpolicy *key; 1003 { 1004 struct addrsel_policyent *pol; 1005 1006 ADDRSEL_XLOCK(); 1007 ADDRSEL_LOCK(); 1008 1009 /* search for the entry in the table */ 1010 TAILQ_FOREACH(pol, &addrsel_policytab, ape_entry) { 1011 if (IN6_ARE_ADDR_EQUAL(&key->addr.sin6_addr, 1012 &pol->ape_policy.addr.sin6_addr) && 1013 IN6_ARE_ADDR_EQUAL(&key->addrmask.sin6_addr, 1014 &pol->ape_policy.addrmask.sin6_addr)) { 1015 break; 1016 } 1017 } 1018 if (pol == NULL) { 1019 ADDRSEL_UNLOCK(); 1020 ADDRSEL_XUNLOCK(); 1021 return (ESRCH); 1022 } 1023 1024 TAILQ_REMOVE(&addrsel_policytab, pol, ape_entry); 1025 ADDRSEL_UNLOCK(); 1026 ADDRSEL_XUNLOCK(); 1027 1028 return (0); 1029 } 1030 1031 static int 1032 walk_addrsel_policy(callback, w) 1033 int (*callback) __P((struct in6_addrpolicy *, void *)); 1034 void *w; 1035 { 1036 struct addrsel_policyent *pol; 1037 int error = 0; 1038 1039 ADDRSEL_SLOCK(); 1040 TAILQ_FOREACH(pol, &addrsel_policytab, ape_entry) { 1041 if ((error = (*callback)(&pol->ape_policy, w)) != 0) { 1042 ADDRSEL_SUNLOCK(); 1043 return (error); 1044 } 1045 } 1046 ADDRSEL_SUNLOCK(); 1047 return (error); 1048 } 1049 1050 static int 1051 dump_addrsel_policyent(pol, arg) 1052 struct in6_addrpolicy *pol; 1053 void *arg; 1054 { 1055 int error = 0; 1056 struct walkarg *w = arg; 1057 1058 error = SYSCTL_OUT(w->w_req, pol, sizeof(*pol)); 1059 1060 return (error); 1061 } 1062 1063 static struct in6_addrpolicy * 1064 match_addrsel_policy(key) 1065 struct sockaddr_in6 *key; 1066 { 1067 struct addrsel_policyent *pent; 1068 struct in6_addrpolicy *bestpol = NULL, *pol; 1069 int matchlen, bestmatchlen = -1; 1070 u_char *mp, *ep, *k, *p, m; 1071 1072 TAILQ_FOREACH(pent, &addrsel_policytab, ape_entry) { 1073 matchlen = 0; 1074 1075 pol = &pent->ape_policy; 1076 mp = (u_char *)&pol->addrmask.sin6_addr; 1077 ep = mp + 16; /* XXX: scope field? */ 1078 k = (u_char *)&key->sin6_addr; 1079 p = (u_char *)&pol->addr.sin6_addr; 1080 for (; mp < ep && *mp; mp++, k++, p++) { 1081 m = *mp; 1082 if ((*k & m) != *p) 1083 goto next; /* not match */ 1084 if (m == 0xff) /* short cut for a typical case */ 1085 matchlen += 8; 1086 else { 1087 while (m >= 0x80) { 1088 matchlen++; 1089 m <<= 1; 1090 } 1091 } 1092 } 1093 1094 /* matched. check if this is better than the current best. */ 1095 if (bestpol == NULL || 1096 matchlen > bestmatchlen) { 1097 bestpol = pol; 1098 bestmatchlen = matchlen; 1099 } 1100 1101 next: 1102 continue; 1103 } 1104 1105 return (bestpol); 1106 } 1107