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: in6_src.c,v 1.132 2003/08/26 04:42:27 keiichi Exp $ 30 */ 31 32 /*- 33 * Copyright (c) 1982, 1986, 1991, 1993 34 * The Regents of the University of California. All rights reserved. 35 * 36 * Redistribution and use in source and binary forms, with or without 37 * modification, are permitted provided that the following conditions 38 * are met: 39 * 1. Redistributions of source code must retain the above copyright 40 * notice, this list of conditions and the following disclaimer. 41 * 2. Redistributions in binary form must reproduce the above copyright 42 * notice, this list of conditions and the following disclaimer in the 43 * documentation and/or other materials provided with the distribution. 44 * 4. Neither the name of the University nor the names of its contributors 45 * may be used to endorse or promote products derived from this software 46 * without specific prior written permission. 47 * 48 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 49 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 50 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 51 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 52 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 53 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 54 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 55 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 56 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 57 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 58 * SUCH DAMAGE. 59 * 60 * @(#)in_pcb.c 8.2 (Berkeley) 1/4/94 61 */ 62 63 #include <sys/cdefs.h> 64 __FBSDID("$FreeBSD$"); 65 66 #include "opt_inet.h" 67 #include "opt_inet6.h" 68 #include "opt_mpath.h" 69 70 #include <sys/param.h> 71 #include <sys/systm.h> 72 #include <sys/lock.h> 73 #include <sys/malloc.h> 74 #include <sys/mbuf.h> 75 #include <sys/priv.h> 76 #include <sys/protosw.h> 77 #include <sys/socket.h> 78 #include <sys/socketvar.h> 79 #include <sys/sockio.h> 80 #include <sys/sysctl.h> 81 #include <sys/errno.h> 82 #include <sys/time.h> 83 #include <sys/jail.h> 84 #include <sys/kernel.h> 85 #include <sys/sx.h> 86 87 #include <net/if.h> 88 #include <net/if_dl.h> 89 #include <net/route.h> 90 #include <net/if_llatbl.h> 91 #ifdef RADIX_MPATH 92 #include <net/radix_mpath.h> 93 #endif 94 95 #include <netinet/in.h> 96 #include <netinet/in_var.h> 97 #include <netinet/in_systm.h> 98 #include <netinet/ip.h> 99 #include <netinet/in_pcb.h> 100 #include <netinet/ip_var.h> 101 #include <netinet/udp.h> 102 #include <netinet/udp_var.h> 103 104 #include <netinet6/in6_var.h> 105 #include <netinet/ip6.h> 106 #include <netinet6/in6_pcb.h> 107 #include <netinet6/ip6_var.h> 108 #include <netinet6/scope6_var.h> 109 #include <netinet6/nd6.h> 110 111 static struct mtx addrsel_lock; 112 #define ADDRSEL_LOCK_INIT() mtx_init(&addrsel_lock, "addrsel_lock", NULL, MTX_DEF) 113 #define ADDRSEL_LOCK() mtx_lock(&addrsel_lock) 114 #define ADDRSEL_UNLOCK() mtx_unlock(&addrsel_lock) 115 #define ADDRSEL_LOCK_ASSERT() mtx_assert(&addrsel_lock, MA_OWNED) 116 117 static struct sx addrsel_sxlock; 118 #define ADDRSEL_SXLOCK_INIT() sx_init(&addrsel_sxlock, "addrsel_sxlock") 119 #define ADDRSEL_SLOCK() sx_slock(&addrsel_sxlock) 120 #define ADDRSEL_SUNLOCK() sx_sunlock(&addrsel_sxlock) 121 #define ADDRSEL_XLOCK() sx_xlock(&addrsel_sxlock) 122 #define ADDRSEL_XUNLOCK() sx_xunlock(&addrsel_sxlock) 123 124 #define ADDR_LABEL_NOTAPP (-1) 125 static VNET_DEFINE(struct in6_addrpolicy, defaultaddrpolicy); 126 #define V_defaultaddrpolicy VNET(defaultaddrpolicy) 127 128 VNET_DEFINE(int, ip6_prefer_tempaddr) = 0; 129 130 static int selectroute(struct sockaddr_in6 *, struct ip6_pktopts *, 131 struct ip6_moptions *, struct route_in6 *, struct ifnet **, 132 struct rtentry **, int, u_int); 133 static int in6_selectif(struct sockaddr_in6 *, struct ip6_pktopts *, 134 struct ip6_moptions *, struct route_in6 *ro, struct ifnet **, 135 struct ifnet *, u_int); 136 137 static struct in6_addrpolicy *lookup_addrsel_policy(struct sockaddr_in6 *); 138 139 static void init_policy_queue(void); 140 static int add_addrsel_policyent(struct in6_addrpolicy *); 141 static int delete_addrsel_policyent(struct in6_addrpolicy *); 142 static int walk_addrsel_policy(int (*)(struct in6_addrpolicy *, void *), 143 void *); 144 static int dump_addrsel_policyent(struct in6_addrpolicy *, void *); 145 static struct in6_addrpolicy *match_addrsel_policy(struct sockaddr_in6 *); 146 147 /* 148 * Return an IPv6 address, which is the most appropriate for a given 149 * destination and user specified options. 150 * If necessary, this function lookups the routing table and returns 151 * an entry to the caller for later use. 152 */ 153 #define REPLACE(r) do {\ 154 IP6STAT_INC(ip6s_sources_rule[(r)]); \ 155 rule = (r); \ 156 /* { \ 157 char ip6buf[INET6_ADDRSTRLEN], ip6b[INET6_ADDRSTRLEN]; \ 158 printf("in6_selectsrc: replace %s with %s by %d\n", ia_best ? ip6_sprintf(ip6buf, &ia_best->ia_addr.sin6_addr) : "none", ip6_sprintf(ip6b, &ia->ia_addr.sin6_addr), (r)); \ 159 } */ \ 160 goto replace; \ 161 } while(0) 162 #define NEXT(r) do {\ 163 /* { \ 164 char ip6buf[INET6_ADDRSTRLEN], ip6b[INET6_ADDRSTRLEN]; \ 165 printf("in6_selectsrc: keep %s against %s by %d\n", ia_best ? ip6_sprintf(ip6buf, &ia_best->ia_addr.sin6_addr) : "none", ip6_sprintf(ip6b, &ia->ia_addr.sin6_addr), (r)); \ 166 } */ \ 167 goto next; /* XXX: we can't use 'continue' here */ \ 168 } while(0) 169 #define BREAK(r) do { \ 170 IP6STAT_INC(ip6s_sources_rule[(r)]); \ 171 rule = (r); \ 172 goto out; /* XXX: we can't use 'break' here */ \ 173 } while(0) 174 175 int 176 in6_selectsrc(struct sockaddr_in6 *dstsock, struct ip6_pktopts *opts, 177 struct inpcb *inp, struct route_in6 *ro, struct ucred *cred, 178 struct ifnet **ifpp, struct in6_addr *srcp) 179 { 180 struct in6_addr dst, tmp; 181 struct ifnet *ifp = NULL, *oifp = NULL; 182 struct in6_ifaddr *ia = NULL, *ia_best = NULL; 183 struct in6_pktinfo *pi = NULL; 184 int dst_scope = -1, best_scope = -1, best_matchlen = -1; 185 struct in6_addrpolicy *dst_policy = NULL, *best_policy = NULL; 186 u_int32_t odstzone; 187 int prefer_tempaddr; 188 int error, rule; 189 struct ip6_moptions *mopts; 190 191 KASSERT(srcp != NULL, ("%s: srcp is NULL", __func__)); 192 193 dst = dstsock->sin6_addr; /* make a copy for local operation */ 194 if (ifpp) { 195 /* 196 * Save a possibly passed in ifp for in6_selectsrc. Only 197 * neighbor discovery code should use this feature, where 198 * we may know the interface but not the FIB number holding 199 * the connected subnet in case someone deleted it from the 200 * default FIB and we need to check the interface. 201 */ 202 if (*ifpp != NULL) 203 oifp = *ifpp; 204 *ifpp = NULL; 205 } 206 207 if (inp != NULL) { 208 INP_LOCK_ASSERT(inp); 209 mopts = inp->in6p_moptions; 210 } else { 211 mopts = NULL; 212 } 213 214 /* 215 * If the source address is explicitly specified by the caller, 216 * check if the requested source address is indeed a unicast address 217 * assigned to the node, and can be used as the packet's source 218 * address. If everything is okay, use the address as source. 219 */ 220 if (opts && (pi = opts->ip6po_pktinfo) && 221 !IN6_IS_ADDR_UNSPECIFIED(&pi->ipi6_addr)) { 222 struct sockaddr_in6 srcsock; 223 struct in6_ifaddr *ia6; 224 225 /* get the outgoing interface */ 226 if ((error = in6_selectif(dstsock, opts, mopts, ro, &ifp, oifp, 227 (inp != NULL) ? inp->inp_inc.inc_fibnum : RT_DEFAULT_FIB)) 228 != 0) 229 return (error); 230 231 /* 232 * determine the appropriate zone id of the source based on 233 * the zone of the destination and the outgoing interface. 234 * If the specified address is ambiguous wrt the scope zone, 235 * the interface must be specified; otherwise, ifa_ifwithaddr() 236 * will fail matching the address. 237 */ 238 bzero(&srcsock, sizeof(srcsock)); 239 srcsock.sin6_family = AF_INET6; 240 srcsock.sin6_len = sizeof(srcsock); 241 srcsock.sin6_addr = pi->ipi6_addr; 242 if (ifp) { 243 error = in6_setscope(&srcsock.sin6_addr, ifp, NULL); 244 if (error) 245 return (error); 246 } 247 if (cred != NULL && (error = prison_local_ip6(cred, 248 &srcsock.sin6_addr, (inp != NULL && 249 (inp->inp_flags & IN6P_IPV6_V6ONLY) != 0))) != 0) 250 return (error); 251 252 ia6 = (struct in6_ifaddr *)ifa_ifwithaddr( 253 (struct sockaddr *)&srcsock); 254 if (ia6 == NULL || 255 (ia6->ia6_flags & (IN6_IFF_ANYCAST | IN6_IFF_NOTREADY))) { 256 if (ia6 != NULL) 257 ifa_free(&ia6->ia_ifa); 258 return (EADDRNOTAVAIL); 259 } 260 pi->ipi6_addr = srcsock.sin6_addr; /* XXX: this overrides pi */ 261 if (ifpp) 262 *ifpp = ifp; 263 bcopy(&ia6->ia_addr.sin6_addr, srcp, sizeof(*srcp)); 264 ifa_free(&ia6->ia_ifa); 265 return (0); 266 } 267 268 /* 269 * Otherwise, if the socket has already bound the source, just use it. 270 */ 271 if (inp != NULL && !IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) { 272 if (cred != NULL && 273 (error = prison_local_ip6(cred, &inp->in6p_laddr, 274 ((inp->inp_flags & IN6P_IPV6_V6ONLY) != 0))) != 0) 275 return (error); 276 bcopy(&inp->in6p_laddr, srcp, sizeof(*srcp)); 277 return (0); 278 } 279 280 /* 281 * Bypass source address selection and use the primary jail IP 282 * if requested. 283 */ 284 if (cred != NULL && !prison_saddrsel_ip6(cred, srcp)) 285 return (0); 286 287 /* 288 * If the address is not specified, choose the best one based on 289 * the outgoing interface and the destination address. 290 */ 291 /* get the outgoing interface */ 292 if ((error = in6_selectif(dstsock, opts, mopts, ro, &ifp, oifp, 293 (inp != NULL) ? inp->inp_inc.inc_fibnum : RT_DEFAULT_FIB)) != 0) 294 return (error); 295 296 #ifdef DIAGNOSTIC 297 if (ifp == NULL) /* this should not happen */ 298 panic("in6_selectsrc: NULL ifp"); 299 #endif 300 error = in6_setscope(&dst, ifp, &odstzone); 301 if (error) 302 return (error); 303 304 rule = 0; 305 IN6_IFADDR_RLOCK(); 306 TAILQ_FOREACH(ia, &V_in6_ifaddrhead, ia_link) { 307 int new_scope = -1, new_matchlen = -1; 308 struct in6_addrpolicy *new_policy = NULL; 309 u_int32_t srczone, osrczone, dstzone; 310 struct in6_addr src; 311 struct ifnet *ifp1 = ia->ia_ifp; 312 313 /* 314 * We'll never take an address that breaks the scope zone 315 * of the destination. We also skip an address if its zone 316 * does not contain the outgoing interface. 317 * XXX: we should probably use sin6_scope_id here. 318 */ 319 if (in6_setscope(&dst, ifp1, &dstzone) || 320 odstzone != dstzone) { 321 continue; 322 } 323 src = ia->ia_addr.sin6_addr; 324 if (in6_setscope(&src, ifp, &osrczone) || 325 in6_setscope(&src, ifp1, &srczone) || 326 osrczone != srczone) { 327 continue; 328 } 329 330 /* avoid unusable addresses */ 331 if ((ia->ia6_flags & 332 (IN6_IFF_NOTREADY | IN6_IFF_ANYCAST | IN6_IFF_DETACHED))) { 333 continue; 334 } 335 if (!V_ip6_use_deprecated && IFA6_IS_DEPRECATED(ia)) 336 continue; 337 338 /* If jailed only take addresses of the jail into account. */ 339 if (cred != NULL && 340 prison_check_ip6(cred, &ia->ia_addr.sin6_addr) != 0) 341 continue; 342 343 /* Rule 1: Prefer same address */ 344 if (IN6_ARE_ADDR_EQUAL(&dst, &ia->ia_addr.sin6_addr)) { 345 ia_best = ia; 346 BREAK(1); /* there should be no better candidate */ 347 } 348 349 if (ia_best == NULL) 350 REPLACE(0); 351 352 /* Rule 2: Prefer appropriate scope */ 353 if (dst_scope < 0) 354 dst_scope = in6_addrscope(&dst); 355 new_scope = in6_addrscope(&ia->ia_addr.sin6_addr); 356 if (IN6_ARE_SCOPE_CMP(best_scope, new_scope) < 0) { 357 if (IN6_ARE_SCOPE_CMP(best_scope, dst_scope) < 0) 358 REPLACE(2); 359 NEXT(2); 360 } else if (IN6_ARE_SCOPE_CMP(new_scope, best_scope) < 0) { 361 if (IN6_ARE_SCOPE_CMP(new_scope, dst_scope) < 0) 362 NEXT(2); 363 REPLACE(2); 364 } 365 366 /* 367 * Rule 3: Avoid deprecated addresses. Note that the case of 368 * !ip6_use_deprecated is already rejected above. 369 */ 370 if (!IFA6_IS_DEPRECATED(ia_best) && IFA6_IS_DEPRECATED(ia)) 371 NEXT(3); 372 if (IFA6_IS_DEPRECATED(ia_best) && !IFA6_IS_DEPRECATED(ia)) 373 REPLACE(3); 374 375 /* Rule 4: Prefer home addresses */ 376 /* 377 * XXX: This is a TODO. We should probably merge the MIP6 378 * case above. 379 */ 380 381 /* Rule 5: Prefer outgoing interface */ 382 if (!(ND_IFINFO(ifp)->flags & ND6_IFF_NO_PREFER_IFACE)) { 383 if (ia_best->ia_ifp == ifp && ia->ia_ifp != ifp) 384 NEXT(5); 385 if (ia_best->ia_ifp != ifp && ia->ia_ifp == ifp) 386 REPLACE(5); 387 } 388 389 /* 390 * Rule 6: Prefer matching label 391 * Note that best_policy should be non-NULL here. 392 */ 393 if (dst_policy == NULL) 394 dst_policy = lookup_addrsel_policy(dstsock); 395 if (dst_policy->label != ADDR_LABEL_NOTAPP) { 396 new_policy = lookup_addrsel_policy(&ia->ia_addr); 397 if (dst_policy->label == best_policy->label && 398 dst_policy->label != new_policy->label) 399 NEXT(6); 400 if (dst_policy->label != best_policy->label && 401 dst_policy->label == new_policy->label) 402 REPLACE(6); 403 } 404 405 /* 406 * Rule 7: Prefer public addresses. 407 * We allow users to reverse the logic by configuring 408 * a sysctl variable, so that privacy conscious users can 409 * always prefer temporary addresses. 410 */ 411 if (opts == NULL || 412 opts->ip6po_prefer_tempaddr == IP6PO_TEMPADDR_SYSTEM) { 413 prefer_tempaddr = V_ip6_prefer_tempaddr; 414 } else if (opts->ip6po_prefer_tempaddr == 415 IP6PO_TEMPADDR_NOTPREFER) { 416 prefer_tempaddr = 0; 417 } else 418 prefer_tempaddr = 1; 419 if (!(ia_best->ia6_flags & IN6_IFF_TEMPORARY) && 420 (ia->ia6_flags & IN6_IFF_TEMPORARY)) { 421 if (prefer_tempaddr) 422 REPLACE(7); 423 else 424 NEXT(7); 425 } 426 if ((ia_best->ia6_flags & IN6_IFF_TEMPORARY) && 427 !(ia->ia6_flags & IN6_IFF_TEMPORARY)) { 428 if (prefer_tempaddr) 429 NEXT(7); 430 else 431 REPLACE(7); 432 } 433 434 /* 435 * Rule 8: prefer addresses on alive interfaces. 436 * This is a KAME specific rule. 437 */ 438 if ((ia_best->ia_ifp->if_flags & IFF_UP) && 439 !(ia->ia_ifp->if_flags & IFF_UP)) 440 NEXT(8); 441 if (!(ia_best->ia_ifp->if_flags & IFF_UP) && 442 (ia->ia_ifp->if_flags & IFF_UP)) 443 REPLACE(8); 444 445 /* 446 * Rule 14: Use longest matching prefix. 447 * Note: in the address selection draft, this rule is 448 * documented as "Rule 8". However, since it is also 449 * documented that this rule can be overridden, we assign 450 * a large number so that it is easy to assign smaller numbers 451 * to more preferred rules. 452 */ 453 new_matchlen = in6_matchlen(&ia->ia_addr.sin6_addr, &dst); 454 if (best_matchlen < new_matchlen) 455 REPLACE(14); 456 if (new_matchlen < best_matchlen) 457 NEXT(14); 458 459 /* Rule 15 is reserved. */ 460 461 /* 462 * Last resort: just keep the current candidate. 463 * Or, do we need more rules? 464 */ 465 continue; 466 467 replace: 468 ia_best = ia; 469 best_scope = (new_scope >= 0 ? new_scope : 470 in6_addrscope(&ia_best->ia_addr.sin6_addr)); 471 best_policy = (new_policy ? new_policy : 472 lookup_addrsel_policy(&ia_best->ia_addr)); 473 best_matchlen = (new_matchlen >= 0 ? new_matchlen : 474 in6_matchlen(&ia_best->ia_addr.sin6_addr, 475 &dst)); 476 477 next: 478 continue; 479 480 out: 481 break; 482 } 483 484 if ((ia = ia_best) == NULL) { 485 IN6_IFADDR_RUNLOCK(); 486 IP6STAT_INC(ip6s_sources_none); 487 return (EADDRNOTAVAIL); 488 } 489 490 /* 491 * At this point at least one of the addresses belonged to the jail 492 * but it could still be, that we want to further restrict it, e.g. 493 * theoratically IN6_IS_ADDR_LOOPBACK. 494 * It must not be IN6_IS_ADDR_UNSPECIFIED anymore. 495 * prison_local_ip6() will fix an IN6_IS_ADDR_LOOPBACK but should 496 * let all others previously selected pass. 497 * Use tmp to not change ::1 on lo0 to the primary jail address. 498 */ 499 tmp = ia->ia_addr.sin6_addr; 500 if (cred != NULL && prison_local_ip6(cred, &tmp, (inp != NULL && 501 (inp->inp_flags & IN6P_IPV6_V6ONLY) != 0)) != 0) { 502 IN6_IFADDR_RUNLOCK(); 503 IP6STAT_INC(ip6s_sources_none); 504 return (EADDRNOTAVAIL); 505 } 506 507 if (ifpp) 508 *ifpp = ifp; 509 510 bcopy(&tmp, srcp, sizeof(*srcp)); 511 if (ia->ia_ifp == ifp) 512 IP6STAT_INC(ip6s_sources_sameif[best_scope]); 513 else 514 IP6STAT_INC(ip6s_sources_otherif[best_scope]); 515 if (dst_scope == best_scope) 516 IP6STAT_INC(ip6s_sources_samescope[best_scope]); 517 else 518 IP6STAT_INC(ip6s_sources_otherscope[best_scope]); 519 if (IFA6_IS_DEPRECATED(ia)) 520 IP6STAT_INC(ip6s_sources_deprecated[best_scope]); 521 IN6_IFADDR_RUNLOCK(); 522 return (0); 523 } 524 525 /* 526 * clone - meaningful only for bsdi and freebsd 527 */ 528 static int 529 selectroute(struct sockaddr_in6 *dstsock, struct ip6_pktopts *opts, 530 struct ip6_moptions *mopts, struct route_in6 *ro, 531 struct ifnet **retifp, struct rtentry **retrt, int norouteok, u_int fibnum) 532 { 533 int error = 0; 534 struct ifnet *ifp = NULL; 535 struct rtentry *rt = NULL; 536 struct sockaddr_in6 *sin6_next; 537 struct in6_pktinfo *pi = NULL; 538 struct in6_addr *dst = &dstsock->sin6_addr; 539 #if 0 540 char ip6buf[INET6_ADDRSTRLEN]; 541 542 if (dstsock->sin6_addr.s6_addr32[0] == 0 && 543 dstsock->sin6_addr.s6_addr32[1] == 0 && 544 !IN6_IS_ADDR_LOOPBACK(&dstsock->sin6_addr)) { 545 printf("in6_selectroute: strange destination %s\n", 546 ip6_sprintf(ip6buf, &dstsock->sin6_addr)); 547 } else { 548 printf("in6_selectroute: destination = %s%%%d\n", 549 ip6_sprintf(ip6buf, &dstsock->sin6_addr), 550 dstsock->sin6_scope_id); /* for debug */ 551 } 552 #endif 553 554 /* If the caller specify the outgoing interface explicitly, use it. */ 555 if (opts && (pi = opts->ip6po_pktinfo) != NULL && pi->ipi6_ifindex) { 556 /* XXX boundary check is assumed to be already done. */ 557 ifp = ifnet_byindex(pi->ipi6_ifindex); 558 if (ifp != NULL && 559 (norouteok || retrt == NULL || 560 IN6_IS_ADDR_MULTICAST(dst))) { 561 /* 562 * we do not have to check or get the route for 563 * multicast. 564 */ 565 goto done; 566 } else 567 goto getroute; 568 } 569 570 /* 571 * If the destination address is a multicast address and the outgoing 572 * interface for the address is specified by the caller, use it. 573 */ 574 if (IN6_IS_ADDR_MULTICAST(dst) && 575 mopts != NULL && (ifp = mopts->im6o_multicast_ifp) != NULL) { 576 goto done; /* we do not need a route for multicast. */ 577 } 578 579 getroute: 580 /* 581 * If the next hop address for the packet is specified by the caller, 582 * use it as the gateway. 583 */ 584 if (opts && opts->ip6po_nexthop) { 585 struct route_in6 *ron; 586 struct llentry *la; 587 588 sin6_next = satosin6(opts->ip6po_nexthop); 589 590 /* at this moment, we only support AF_INET6 next hops */ 591 if (sin6_next->sin6_family != AF_INET6) { 592 error = EAFNOSUPPORT; /* or should we proceed? */ 593 goto done; 594 } 595 596 /* 597 * If the next hop is an IPv6 address, then the node identified 598 * by that address must be a neighbor of the sending host. 599 */ 600 ron = &opts->ip6po_nextroute; 601 /* 602 * XXX what do we do here? 603 * PLZ to be fixing 604 */ 605 606 607 if (ron->ro_rt == NULL) { 608 in6_rtalloc(ron, fibnum); /* multi path case? */ 609 if (ron->ro_rt == NULL) { 610 /* XXX-BZ WT.? */ 611 if (ron->ro_rt) { 612 RTFREE(ron->ro_rt); 613 ron->ro_rt = NULL; 614 } 615 error = EHOSTUNREACH; 616 goto done; 617 } 618 } 619 620 rt = ron->ro_rt; 621 ifp = rt->rt_ifp; 622 IF_AFDATA_RLOCK(ifp); 623 la = lla_lookup(LLTABLE6(ifp), 0, (struct sockaddr *)&sin6_next->sin6_addr); 624 IF_AFDATA_RUNLOCK(ifp); 625 if (la != NULL) 626 LLE_RUNLOCK(la); 627 else { 628 error = EHOSTUNREACH; 629 goto done; 630 } 631 #if 0 632 if ((ron->ro_rt && 633 (ron->ro_rt->rt_flags & (RTF_UP | RTF_LLINFO)) != 634 (RTF_UP | RTF_LLINFO)) || 635 !IN6_ARE_ADDR_EQUAL(&satosin6(&ron->ro_dst)->sin6_addr, 636 &sin6_next->sin6_addr)) { 637 if (ron->ro_rt) { 638 RTFREE(ron->ro_rt); 639 ron->ro_rt = NULL; 640 } 641 *satosin6(&ron->ro_dst) = *sin6_next; 642 } 643 if (ron->ro_rt == NULL) { 644 in6_rtalloc(ron, fibnum); /* multi path case? */ 645 if (ron->ro_rt == NULL || 646 !(ron->ro_rt->rt_flags & RTF_LLINFO)) { 647 if (ron->ro_rt) { 648 RTFREE(ron->ro_rt); 649 ron->ro_rt = NULL; 650 } 651 error = EHOSTUNREACH; 652 goto done; 653 } 654 } 655 #endif 656 657 /* 658 * When cloning is required, try to allocate a route to the 659 * destination so that the caller can store path MTU 660 * information. 661 */ 662 goto done; 663 } 664 665 /* 666 * Use a cached route if it exists and is valid, else try to allocate 667 * a new one. Note that we should check the address family of the 668 * cached destination, in case of sharing the cache with IPv4. 669 */ 670 if (ro) { 671 if (ro->ro_rt && 672 (!(ro->ro_rt->rt_flags & RTF_UP) || 673 ((struct sockaddr *)(&ro->ro_dst))->sa_family != AF_INET6 || 674 !IN6_ARE_ADDR_EQUAL(&satosin6(&ro->ro_dst)->sin6_addr, 675 dst))) { 676 RTFREE(ro->ro_rt); 677 ro->ro_rt = (struct rtentry *)NULL; 678 } 679 if (ro->ro_rt == (struct rtentry *)NULL) { 680 struct sockaddr_in6 *sa6; 681 682 /* No route yet, so try to acquire one */ 683 bzero(&ro->ro_dst, sizeof(struct sockaddr_in6)); 684 sa6 = (struct sockaddr_in6 *)&ro->ro_dst; 685 *sa6 = *dstsock; 686 sa6->sin6_scope_id = 0; 687 688 #ifdef RADIX_MPATH 689 rtalloc_mpath_fib((struct route *)ro, 690 ntohl(sa6->sin6_addr.s6_addr32[3]), fibnum); 691 #else 692 ro->ro_rt = in6_rtalloc1((struct sockaddr *) 693 &ro->ro_dst, 0, 0UL, fibnum); 694 if (ro->ro_rt) 695 RT_UNLOCK(ro->ro_rt); 696 #endif 697 } 698 699 /* 700 * do not care about the result if we have the nexthop 701 * explicitly specified. 702 */ 703 if (opts && opts->ip6po_nexthop) 704 goto done; 705 706 if (ro->ro_rt) { 707 ifp = ro->ro_rt->rt_ifp; 708 709 if (ifp == NULL) { /* can this really happen? */ 710 RTFREE(ro->ro_rt); 711 ro->ro_rt = NULL; 712 } 713 } 714 if (ro->ro_rt == NULL) 715 error = EHOSTUNREACH; 716 rt = ro->ro_rt; 717 718 /* 719 * Check if the outgoing interface conflicts with 720 * the interface specified by ipi6_ifindex (if specified). 721 * Note that loopback interface is always okay. 722 * (this may happen when we are sending a packet to one of 723 * our own addresses.) 724 */ 725 if (ifp && opts && opts->ip6po_pktinfo && 726 opts->ip6po_pktinfo->ipi6_ifindex) { 727 if (!(ifp->if_flags & IFF_LOOPBACK) && 728 ifp->if_index != 729 opts->ip6po_pktinfo->ipi6_ifindex) { 730 error = EHOSTUNREACH; 731 goto done; 732 } 733 } 734 } 735 736 done: 737 if (ifp == NULL && rt == NULL) { 738 /* 739 * This can happen if the caller did not pass a cached route 740 * nor any other hints. We treat this case an error. 741 */ 742 error = EHOSTUNREACH; 743 } 744 if (error == EHOSTUNREACH) 745 IP6STAT_INC(ip6s_noroute); 746 747 if (retifp != NULL) { 748 *retifp = ifp; 749 750 /* 751 * Adjust the "outgoing" interface. If we're going to loop 752 * the packet back to ourselves, the ifp would be the loopback 753 * interface. However, we'd rather know the interface associated 754 * to the destination address (which should probably be one of 755 * our own addresses.) 756 */ 757 if (rt) { 758 if ((rt->rt_ifp->if_flags & IFF_LOOPBACK) && 759 (rt->rt_gateway->sa_family == AF_LINK)) 760 *retifp = 761 ifnet_byindex(((struct sockaddr_dl *) 762 rt->rt_gateway)->sdl_index); 763 } 764 } 765 766 if (retrt != NULL) 767 *retrt = rt; /* rt may be NULL */ 768 769 return (error); 770 } 771 772 static int 773 in6_selectif(struct sockaddr_in6 *dstsock, struct ip6_pktopts *opts, 774 struct ip6_moptions *mopts, struct route_in6 *ro, struct ifnet **retifp, 775 struct ifnet *oifp, u_int fibnum) 776 { 777 int error; 778 struct route_in6 sro; 779 struct rtentry *rt = NULL; 780 781 KASSERT(retifp != NULL, ("%s: retifp is NULL", __func__)); 782 783 if (ro == NULL) { 784 bzero(&sro, sizeof(sro)); 785 ro = &sro; 786 } 787 788 if ((error = selectroute(dstsock, opts, mopts, ro, retifp, 789 &rt, 1, fibnum)) != 0) { 790 if (ro == &sro && rt && rt == sro.ro_rt) 791 RTFREE(rt); 792 /* Help ND. See oifp comment in in6_selectsrc(). */ 793 if (oifp != NULL && fibnum == RT_DEFAULT_FIB) { 794 *retifp = oifp; 795 error = 0; 796 } 797 return (error); 798 } 799 800 /* 801 * do not use a rejected or black hole route. 802 * XXX: this check should be done in the L2 output routine. 803 * However, if we skipped this check here, we'd see the following 804 * scenario: 805 * - install a rejected route for a scoped address prefix 806 * (like fe80::/10) 807 * - send a packet to a destination that matches the scoped prefix, 808 * with ambiguity about the scope zone. 809 * - pick the outgoing interface from the route, and disambiguate the 810 * scope zone with the interface. 811 * - ip6_output() would try to get another route with the "new" 812 * destination, which may be valid. 813 * - we'd see no error on output. 814 * Although this may not be very harmful, it should still be confusing. 815 * We thus reject the case here. 816 */ 817 if (rt && (rt->rt_flags & (RTF_REJECT | RTF_BLACKHOLE))) { 818 int flags = (rt->rt_flags & RTF_HOST ? EHOSTUNREACH : ENETUNREACH); 819 820 if (ro == &sro && rt && rt == sro.ro_rt) 821 RTFREE(rt); 822 return (flags); 823 } 824 825 if (ro == &sro && rt && rt == sro.ro_rt) 826 RTFREE(rt); 827 return (0); 828 } 829 830 /* 831 * Public wrapper function to selectroute(). 832 * 833 * XXX-BZ in6_selectroute() should and will grow the FIB argument. The 834 * in6_selectroute_fib() function is only there for backward compat on stable. 835 */ 836 int 837 in6_selectroute(struct sockaddr_in6 *dstsock, struct ip6_pktopts *opts, 838 struct ip6_moptions *mopts, struct route_in6 *ro, 839 struct ifnet **retifp, struct rtentry **retrt) 840 { 841 842 return (selectroute(dstsock, opts, mopts, ro, retifp, 843 retrt, 0, RT_DEFAULT_FIB)); 844 } 845 846 #ifndef BURN_BRIDGES 847 int 848 in6_selectroute_fib(struct sockaddr_in6 *dstsock, struct ip6_pktopts *opts, 849 struct ip6_moptions *mopts, struct route_in6 *ro, 850 struct ifnet **retifp, struct rtentry **retrt, u_int fibnum) 851 { 852 853 return (selectroute(dstsock, opts, mopts, ro, retifp, 854 retrt, 0, fibnum)); 855 } 856 #endif 857 858 /* 859 * Default hop limit selection. The precedence is as follows: 860 * 1. Hoplimit value specified via ioctl. 861 * 2. (If the outgoing interface is detected) the current 862 * hop limit of the interface specified by router advertisement. 863 * 3. The system default hoplimit. 864 */ 865 int 866 in6_selecthlim(struct inpcb *in6p, struct ifnet *ifp) 867 { 868 869 if (in6p && in6p->in6p_hops >= 0) 870 return (in6p->in6p_hops); 871 else if (ifp) 872 return (ND_IFINFO(ifp)->chlim); 873 else if (in6p && !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr)) { 874 struct route_in6 ro6; 875 struct ifnet *lifp; 876 877 bzero(&ro6, sizeof(ro6)); 878 ro6.ro_dst.sin6_family = AF_INET6; 879 ro6.ro_dst.sin6_len = sizeof(struct sockaddr_in6); 880 ro6.ro_dst.sin6_addr = in6p->in6p_faddr; 881 in6_rtalloc(&ro6, in6p->inp_inc.inc_fibnum); 882 if (ro6.ro_rt) { 883 lifp = ro6.ro_rt->rt_ifp; 884 RTFREE(ro6.ro_rt); 885 if (lifp) 886 return (ND_IFINFO(lifp)->chlim); 887 } 888 } 889 return (V_ip6_defhlim); 890 } 891 892 /* 893 * XXX: this is borrowed from in6_pcbbind(). If possible, we should 894 * share this function by all *bsd*... 895 */ 896 int 897 in6_pcbsetport(struct in6_addr *laddr, struct inpcb *inp, struct ucred *cred) 898 { 899 struct socket *so = inp->inp_socket; 900 u_int16_t lport = 0; 901 int error, lookupflags = 0; 902 #ifdef INVARIANTS 903 struct inpcbinfo *pcbinfo = inp->inp_pcbinfo; 904 #endif 905 906 INP_WLOCK_ASSERT(inp); 907 INP_HASH_WLOCK_ASSERT(pcbinfo); 908 909 error = prison_local_ip6(cred, laddr, 910 ((inp->inp_flags & IN6P_IPV6_V6ONLY) != 0)); 911 if (error) 912 return(error); 913 914 /* XXX: this is redundant when called from in6_pcbbind */ 915 if ((so->so_options & (SO_REUSEADDR|SO_REUSEPORT)) == 0) 916 lookupflags = INPLOOKUP_WILDCARD; 917 918 inp->inp_flags |= INP_ANONPORT; 919 920 error = in_pcb_lport(inp, NULL, &lport, cred, lookupflags); 921 if (error != 0) 922 return (error); 923 924 inp->inp_lport = lport; 925 if (in_pcbinshash(inp) != 0) { 926 inp->in6p_laddr = in6addr_any; 927 inp->inp_lport = 0; 928 return (EAGAIN); 929 } 930 931 return (0); 932 } 933 934 void 935 addrsel_policy_init(void) 936 { 937 938 init_policy_queue(); 939 940 /* initialize the "last resort" policy */ 941 bzero(&V_defaultaddrpolicy, sizeof(V_defaultaddrpolicy)); 942 V_defaultaddrpolicy.label = ADDR_LABEL_NOTAPP; 943 944 if (!IS_DEFAULT_VNET(curvnet)) 945 return; 946 947 ADDRSEL_LOCK_INIT(); 948 ADDRSEL_SXLOCK_INIT(); 949 } 950 951 static struct in6_addrpolicy * 952 lookup_addrsel_policy(struct sockaddr_in6 *key) 953 { 954 struct in6_addrpolicy *match = NULL; 955 956 ADDRSEL_LOCK(); 957 match = match_addrsel_policy(key); 958 959 if (match == NULL) 960 match = &V_defaultaddrpolicy; 961 else 962 match->use++; 963 ADDRSEL_UNLOCK(); 964 965 return (match); 966 } 967 968 /* 969 * Subroutines to manage the address selection policy table via sysctl. 970 */ 971 struct walkarg { 972 struct sysctl_req *w_req; 973 }; 974 975 static int in6_src_sysctl(SYSCTL_HANDLER_ARGS); 976 SYSCTL_DECL(_net_inet6_ip6); 977 static SYSCTL_NODE(_net_inet6_ip6, IPV6CTL_ADDRCTLPOLICY, addrctlpolicy, 978 CTLFLAG_RD, in6_src_sysctl, ""); 979 980 static int 981 in6_src_sysctl(SYSCTL_HANDLER_ARGS) 982 { 983 struct walkarg w; 984 985 if (req->newptr) 986 return EPERM; 987 988 bzero(&w, sizeof(w)); 989 w.w_req = req; 990 991 return (walk_addrsel_policy(dump_addrsel_policyent, &w)); 992 } 993 994 int 995 in6_src_ioctl(u_long cmd, caddr_t data) 996 { 997 int i; 998 struct in6_addrpolicy ent0; 999 1000 if (cmd != SIOCAADDRCTL_POLICY && cmd != SIOCDADDRCTL_POLICY) 1001 return (EOPNOTSUPP); /* check for safety */ 1002 1003 ent0 = *(struct in6_addrpolicy *)data; 1004 1005 if (ent0.label == ADDR_LABEL_NOTAPP) 1006 return (EINVAL); 1007 /* check if the prefix mask is consecutive. */ 1008 if (in6_mask2len(&ent0.addrmask.sin6_addr, NULL) < 0) 1009 return (EINVAL); 1010 /* clear trailing garbages (if any) of the prefix address. */ 1011 for (i = 0; i < 4; i++) { 1012 ent0.addr.sin6_addr.s6_addr32[i] &= 1013 ent0.addrmask.sin6_addr.s6_addr32[i]; 1014 } 1015 ent0.use = 0; 1016 1017 switch (cmd) { 1018 case SIOCAADDRCTL_POLICY: 1019 return (add_addrsel_policyent(&ent0)); 1020 case SIOCDADDRCTL_POLICY: 1021 return (delete_addrsel_policyent(&ent0)); 1022 } 1023 1024 return (0); /* XXX: compromise compilers */ 1025 } 1026 1027 /* 1028 * The followings are implementation of the policy table using a 1029 * simple tail queue. 1030 * XXX such details should be hidden. 1031 * XXX implementation using binary tree should be more efficient. 1032 */ 1033 struct addrsel_policyent { 1034 TAILQ_ENTRY(addrsel_policyent) ape_entry; 1035 struct in6_addrpolicy ape_policy; 1036 }; 1037 1038 TAILQ_HEAD(addrsel_policyhead, addrsel_policyent); 1039 1040 static VNET_DEFINE(struct addrsel_policyhead, addrsel_policytab); 1041 #define V_addrsel_policytab VNET(addrsel_policytab) 1042 1043 static void 1044 init_policy_queue(void) 1045 { 1046 1047 TAILQ_INIT(&V_addrsel_policytab); 1048 } 1049 1050 static int 1051 add_addrsel_policyent(struct in6_addrpolicy *newpolicy) 1052 { 1053 struct addrsel_policyent *new, *pol; 1054 1055 new = malloc(sizeof(*new), M_IFADDR, 1056 M_WAITOK); 1057 ADDRSEL_XLOCK(); 1058 ADDRSEL_LOCK(); 1059 1060 /* duplication check */ 1061 TAILQ_FOREACH(pol, &V_addrsel_policytab, ape_entry) { 1062 if (IN6_ARE_ADDR_EQUAL(&newpolicy->addr.sin6_addr, 1063 &pol->ape_policy.addr.sin6_addr) && 1064 IN6_ARE_ADDR_EQUAL(&newpolicy->addrmask.sin6_addr, 1065 &pol->ape_policy.addrmask.sin6_addr)) { 1066 ADDRSEL_UNLOCK(); 1067 ADDRSEL_XUNLOCK(); 1068 free(new, M_IFADDR); 1069 return (EEXIST); /* or override it? */ 1070 } 1071 } 1072 1073 bzero(new, sizeof(*new)); 1074 1075 /* XXX: should validate entry */ 1076 new->ape_policy = *newpolicy; 1077 1078 TAILQ_INSERT_TAIL(&V_addrsel_policytab, new, ape_entry); 1079 ADDRSEL_UNLOCK(); 1080 ADDRSEL_XUNLOCK(); 1081 1082 return (0); 1083 } 1084 1085 static int 1086 delete_addrsel_policyent(struct in6_addrpolicy *key) 1087 { 1088 struct addrsel_policyent *pol; 1089 1090 ADDRSEL_XLOCK(); 1091 ADDRSEL_LOCK(); 1092 1093 /* search for the entry in the table */ 1094 TAILQ_FOREACH(pol, &V_addrsel_policytab, ape_entry) { 1095 if (IN6_ARE_ADDR_EQUAL(&key->addr.sin6_addr, 1096 &pol->ape_policy.addr.sin6_addr) && 1097 IN6_ARE_ADDR_EQUAL(&key->addrmask.sin6_addr, 1098 &pol->ape_policy.addrmask.sin6_addr)) { 1099 break; 1100 } 1101 } 1102 if (pol == NULL) { 1103 ADDRSEL_UNLOCK(); 1104 ADDRSEL_XUNLOCK(); 1105 return (ESRCH); 1106 } 1107 1108 TAILQ_REMOVE(&V_addrsel_policytab, pol, ape_entry); 1109 ADDRSEL_UNLOCK(); 1110 ADDRSEL_XUNLOCK(); 1111 free(pol, M_IFADDR); 1112 1113 return (0); 1114 } 1115 1116 static int 1117 walk_addrsel_policy(int (*callback)(struct in6_addrpolicy *, void *), void *w) 1118 { 1119 struct addrsel_policyent *pol; 1120 int error = 0; 1121 1122 ADDRSEL_SLOCK(); 1123 TAILQ_FOREACH(pol, &V_addrsel_policytab, ape_entry) { 1124 if ((error = (*callback)(&pol->ape_policy, w)) != 0) { 1125 ADDRSEL_SUNLOCK(); 1126 return (error); 1127 } 1128 } 1129 ADDRSEL_SUNLOCK(); 1130 return (error); 1131 } 1132 1133 static int 1134 dump_addrsel_policyent(struct in6_addrpolicy *pol, void *arg) 1135 { 1136 int error = 0; 1137 struct walkarg *w = arg; 1138 1139 error = SYSCTL_OUT(w->w_req, pol, sizeof(*pol)); 1140 1141 return (error); 1142 } 1143 1144 static struct in6_addrpolicy * 1145 match_addrsel_policy(struct sockaddr_in6 *key) 1146 { 1147 struct addrsel_policyent *pent; 1148 struct in6_addrpolicy *bestpol = NULL, *pol; 1149 int matchlen, bestmatchlen = -1; 1150 u_char *mp, *ep, *k, *p, m; 1151 1152 TAILQ_FOREACH(pent, &V_addrsel_policytab, ape_entry) { 1153 matchlen = 0; 1154 1155 pol = &pent->ape_policy; 1156 mp = (u_char *)&pol->addrmask.sin6_addr; 1157 ep = mp + 16; /* XXX: scope field? */ 1158 k = (u_char *)&key->sin6_addr; 1159 p = (u_char *)&pol->addr.sin6_addr; 1160 for (; mp < ep && *mp; mp++, k++, p++) { 1161 m = *mp; 1162 if ((*k & m) != *p) 1163 goto next; /* not match */ 1164 if (m == 0xff) /* short cut for a typical case */ 1165 matchlen += 8; 1166 else { 1167 while (m >= 0x80) { 1168 matchlen++; 1169 m <<= 1; 1170 } 1171 } 1172 } 1173 1174 /* matched. check if this is better than the current best. */ 1175 if (bestpol == NULL || 1176 matchlen > bestmatchlen) { 1177 bestpol = pol; 1178 bestmatchlen = matchlen; 1179 } 1180 1181 next: 1182 continue; 1183 } 1184 1185 return (bestpol); 1186 } 1187