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_var.h> 89 #include <net/if_dl.h> 90 #include <net/route.h> 91 #include <net/if_llatbl.h> 92 #ifdef RADIX_MPATH 93 #include <net/radix_mpath.h> 94 #endif 95 96 #include <netinet/in.h> 97 #include <netinet/in_var.h> 98 #include <netinet/in_systm.h> 99 #include <netinet/ip.h> 100 #include <netinet/in_pcb.h> 101 #include <netinet/ip_var.h> 102 #include <netinet/udp.h> 103 #include <netinet/udp_var.h> 104 105 #include <netinet6/in6_var.h> 106 #include <netinet/ip6.h> 107 #include <netinet6/in6_pcb.h> 108 #include <netinet6/ip6_var.h> 109 #include <netinet6/scope6_var.h> 110 #include <netinet6/nd6.h> 111 112 static struct mtx addrsel_lock; 113 #define ADDRSEL_LOCK_INIT() mtx_init(&addrsel_lock, "addrsel_lock", NULL, MTX_DEF) 114 #define ADDRSEL_LOCK() mtx_lock(&addrsel_lock) 115 #define ADDRSEL_UNLOCK() mtx_unlock(&addrsel_lock) 116 #define ADDRSEL_LOCK_ASSERT() mtx_assert(&addrsel_lock, MA_OWNED) 117 118 static struct sx addrsel_sxlock; 119 #define ADDRSEL_SXLOCK_INIT() sx_init(&addrsel_sxlock, "addrsel_sxlock") 120 #define ADDRSEL_SLOCK() sx_slock(&addrsel_sxlock) 121 #define ADDRSEL_SUNLOCK() sx_sunlock(&addrsel_sxlock) 122 #define ADDRSEL_XLOCK() sx_xlock(&addrsel_sxlock) 123 #define ADDRSEL_XUNLOCK() sx_xunlock(&addrsel_sxlock) 124 125 #define ADDR_LABEL_NOTAPP (-1) 126 static VNET_DEFINE(struct in6_addrpolicy, defaultaddrpolicy); 127 #define V_defaultaddrpolicy VNET(defaultaddrpolicy) 128 129 VNET_DEFINE(int, ip6_prefer_tempaddr) = 0; 130 131 static int selectroute(struct sockaddr_in6 *, struct ip6_pktopts *, 132 struct ip6_moptions *, struct route_in6 *, struct ifnet **, 133 struct rtentry **, int, u_int); 134 static int in6_selectif(struct sockaddr_in6 *, struct ip6_pktopts *, 135 struct ip6_moptions *, struct route_in6 *ro, struct ifnet **, 136 struct ifnet *, u_int); 137 138 static struct in6_addrpolicy *lookup_addrsel_policy(struct sockaddr_in6 *); 139 140 static void init_policy_queue(void); 141 static int add_addrsel_policyent(struct in6_addrpolicy *); 142 static int delete_addrsel_policyent(struct in6_addrpolicy *); 143 static int walk_addrsel_policy(int (*)(struct in6_addrpolicy *, void *), 144 void *); 145 static int dump_addrsel_policyent(struct in6_addrpolicy *, void *); 146 static struct in6_addrpolicy *match_addrsel_policy(struct sockaddr_in6 *); 147 148 /* 149 * Return an IPv6 address, which is the most appropriate for a given 150 * destination and user specified options. 151 * If necessary, this function lookups the routing table and returns 152 * an entry to the caller for later use. 153 */ 154 #define REPLACE(r) do {\ 155 IP6STAT_INC(ip6s_sources_rule[(r)]); \ 156 rule = (r); \ 157 /* { \ 158 char ip6buf[INET6_ADDRSTRLEN], ip6b[INET6_ADDRSTRLEN]; \ 159 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)); \ 160 } */ \ 161 goto replace; \ 162 } while(0) 163 #define NEXT(r) do {\ 164 /* { \ 165 char ip6buf[INET6_ADDRSTRLEN], ip6b[INET6_ADDRSTRLEN]; \ 166 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)); \ 167 } */ \ 168 goto next; /* XXX: we can't use 'continue' here */ \ 169 } while(0) 170 #define BREAK(r) do { \ 171 IP6STAT_INC(ip6s_sources_rule[(r)]); \ 172 rule = (r); \ 173 goto out; /* XXX: we can't use 'break' here */ \ 174 } while(0) 175 176 int 177 in6_selectsrc(struct sockaddr_in6 *dstsock, struct ip6_pktopts *opts, 178 struct inpcb *inp, struct route_in6 *ro, struct ucred *cred, 179 struct ifnet **ifpp, struct in6_addr *srcp) 180 { 181 struct in6_addr dst, tmp; 182 struct ifnet *ifp = NULL, *oifp = NULL; 183 struct in6_ifaddr *ia = NULL, *ia_best = NULL; 184 struct in6_pktinfo *pi = NULL; 185 int dst_scope = -1, best_scope = -1, best_matchlen = -1; 186 struct in6_addrpolicy *dst_policy = NULL, *best_policy = NULL; 187 u_int32_t odstzone; 188 int prefer_tempaddr; 189 int error, rule; 190 struct ip6_moptions *mopts; 191 192 KASSERT(srcp != NULL, ("%s: srcp is NULL", __func__)); 193 194 dst = dstsock->sin6_addr; /* make a copy for local operation */ 195 if (ifpp) { 196 /* 197 * Save a possibly passed in ifp for in6_selectsrc. Only 198 * neighbor discovery code should use this feature, where 199 * we may know the interface but not the FIB number holding 200 * the connected subnet in case someone deleted it from the 201 * default FIB and we need to check the interface. 202 */ 203 if (*ifpp != NULL) 204 oifp = *ifpp; 205 *ifpp = NULL; 206 } 207 208 if (inp != NULL) { 209 INP_LOCK_ASSERT(inp); 210 mopts = inp->in6p_moptions; 211 } else { 212 mopts = NULL; 213 } 214 215 /* 216 * If the source address is explicitly specified by the caller, 217 * check if the requested source address is indeed a unicast address 218 * assigned to the node, and can be used as the packet's source 219 * address. If everything is okay, use the address as source. 220 */ 221 if (opts && (pi = opts->ip6po_pktinfo) && 222 !IN6_IS_ADDR_UNSPECIFIED(&pi->ipi6_addr)) { 223 struct sockaddr_in6 srcsock; 224 struct in6_ifaddr *ia6; 225 226 /* get the outgoing interface */ 227 if ((error = in6_selectif(dstsock, opts, mopts, ro, &ifp, oifp, 228 (inp != NULL) ? inp->inp_inc.inc_fibnum : RT_DEFAULT_FIB)) 229 != 0) 230 return (error); 231 232 /* 233 * determine the appropriate zone id of the source based on 234 * the zone of the destination and the outgoing interface. 235 * If the specified address is ambiguous wrt the scope zone, 236 * the interface must be specified; otherwise, ifa_ifwithaddr() 237 * will fail matching the address. 238 */ 239 bzero(&srcsock, sizeof(srcsock)); 240 srcsock.sin6_family = AF_INET6; 241 srcsock.sin6_len = sizeof(srcsock); 242 srcsock.sin6_addr = pi->ipi6_addr; 243 if (ifp) { 244 error = in6_setscope(&srcsock.sin6_addr, ifp, NULL); 245 if (error) 246 return (error); 247 } 248 if (cred != NULL && (error = prison_local_ip6(cred, 249 &srcsock.sin6_addr, (inp != NULL && 250 (inp->inp_flags & IN6P_IPV6_V6ONLY) != 0))) != 0) 251 return (error); 252 253 ia6 = (struct in6_ifaddr *)ifa_ifwithaddr( 254 (struct sockaddr *)&srcsock); 255 if (ia6 == NULL || 256 (ia6->ia6_flags & (IN6_IFF_ANYCAST | IN6_IFF_NOTREADY))) { 257 if (ia6 != NULL) 258 ifa_free(&ia6->ia_ifa); 259 return (EADDRNOTAVAIL); 260 } 261 pi->ipi6_addr = srcsock.sin6_addr; /* XXX: this overrides pi */ 262 if (ifpp) 263 *ifpp = ifp; 264 bcopy(&ia6->ia_addr.sin6_addr, srcp, sizeof(*srcp)); 265 ifa_free(&ia6->ia_ifa); 266 return (0); 267 } 268 269 /* 270 * Otherwise, if the socket has already bound the source, just use it. 271 */ 272 if (inp != NULL && !IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) { 273 if (cred != NULL && 274 (error = prison_local_ip6(cred, &inp->in6p_laddr, 275 ((inp->inp_flags & IN6P_IPV6_V6ONLY) != 0))) != 0) 276 return (error); 277 bcopy(&inp->in6p_laddr, srcp, sizeof(*srcp)); 278 return (0); 279 } 280 281 /* 282 * Bypass source address selection and use the primary jail IP 283 * if requested. 284 */ 285 if (cred != NULL && !prison_saddrsel_ip6(cred, srcp)) 286 return (0); 287 288 /* 289 * If the address is not specified, choose the best one based on 290 * the outgoing interface and the destination address. 291 */ 292 /* get the outgoing interface */ 293 if ((error = in6_selectif(dstsock, opts, mopts, ro, &ifp, oifp, 294 (inp != NULL) ? inp->inp_inc.inc_fibnum : RT_DEFAULT_FIB)) != 0) 295 return (error); 296 297 #ifdef DIAGNOSTIC 298 if (ifp == NULL) /* this should not happen */ 299 panic("in6_selectsrc: NULL ifp"); 300 #endif 301 error = in6_setscope(&dst, ifp, &odstzone); 302 if (error) 303 return (error); 304 305 rule = 0; 306 IN6_IFADDR_RLOCK(); 307 TAILQ_FOREACH(ia, &V_in6_ifaddrhead, ia_link) { 308 int new_scope = -1, new_matchlen = -1; 309 struct in6_addrpolicy *new_policy = NULL; 310 u_int32_t srczone, osrczone, dstzone; 311 struct in6_addr src; 312 struct ifnet *ifp1 = ia->ia_ifp; 313 314 /* 315 * We'll never take an address that breaks the scope zone 316 * of the destination. We also skip an address if its zone 317 * does not contain the outgoing interface. 318 * XXX: we should probably use sin6_scope_id here. 319 */ 320 if (in6_setscope(&dst, ifp1, &dstzone) || 321 odstzone != dstzone) { 322 continue; 323 } 324 src = ia->ia_addr.sin6_addr; 325 if (in6_setscope(&src, ifp, &osrczone) || 326 in6_setscope(&src, ifp1, &srczone) || 327 osrczone != srczone) { 328 continue; 329 } 330 331 /* avoid unusable addresses */ 332 if ((ia->ia6_flags & 333 (IN6_IFF_NOTREADY | IN6_IFF_ANYCAST | IN6_IFF_DETACHED))) { 334 continue; 335 } 336 if (!V_ip6_use_deprecated && IFA6_IS_DEPRECATED(ia)) 337 continue; 338 339 /* If jailed only take addresses of the jail into account. */ 340 if (cred != NULL && 341 prison_check_ip6(cred, &ia->ia_addr.sin6_addr) != 0) 342 continue; 343 344 /* Rule 1: Prefer same address */ 345 if (IN6_ARE_ADDR_EQUAL(&dst, &ia->ia_addr.sin6_addr)) { 346 ia_best = ia; 347 BREAK(1); /* there should be no better candidate */ 348 } 349 350 if (ia_best == NULL) 351 REPLACE(0); 352 353 /* Rule 2: Prefer appropriate scope */ 354 if (dst_scope < 0) 355 dst_scope = in6_addrscope(&dst); 356 new_scope = in6_addrscope(&ia->ia_addr.sin6_addr); 357 if (IN6_ARE_SCOPE_CMP(best_scope, new_scope) < 0) { 358 if (IN6_ARE_SCOPE_CMP(best_scope, dst_scope) < 0) 359 REPLACE(2); 360 NEXT(2); 361 } else if (IN6_ARE_SCOPE_CMP(new_scope, best_scope) < 0) { 362 if (IN6_ARE_SCOPE_CMP(new_scope, dst_scope) < 0) 363 NEXT(2); 364 REPLACE(2); 365 } 366 367 /* 368 * Rule 3: Avoid deprecated addresses. Note that the case of 369 * !ip6_use_deprecated is already rejected above. 370 */ 371 if (!IFA6_IS_DEPRECATED(ia_best) && IFA6_IS_DEPRECATED(ia)) 372 NEXT(3); 373 if (IFA6_IS_DEPRECATED(ia_best) && !IFA6_IS_DEPRECATED(ia)) 374 REPLACE(3); 375 376 /* Rule 4: Prefer home addresses */ 377 /* 378 * XXX: This is a TODO. We should probably merge the MIP6 379 * case above. 380 */ 381 382 /* Rule 5: Prefer outgoing interface */ 383 if (!(ND_IFINFO(ifp)->flags & ND6_IFF_NO_PREFER_IFACE)) { 384 if (ia_best->ia_ifp == ifp && ia->ia_ifp != ifp) 385 NEXT(5); 386 if (ia_best->ia_ifp != ifp && ia->ia_ifp == ifp) 387 REPLACE(5); 388 } 389 390 /* 391 * Rule 6: Prefer matching label 392 * Note that best_policy should be non-NULL here. 393 */ 394 if (dst_policy == NULL) 395 dst_policy = lookup_addrsel_policy(dstsock); 396 if (dst_policy->label != ADDR_LABEL_NOTAPP) { 397 new_policy = lookup_addrsel_policy(&ia->ia_addr); 398 if (dst_policy->label == best_policy->label && 399 dst_policy->label != new_policy->label) 400 NEXT(6); 401 if (dst_policy->label != best_policy->label && 402 dst_policy->label == new_policy->label) 403 REPLACE(6); 404 } 405 406 /* 407 * Rule 7: Prefer public addresses. 408 * We allow users to reverse the logic by configuring 409 * a sysctl variable, so that privacy conscious users can 410 * always prefer temporary addresses. 411 */ 412 if (opts == NULL || 413 opts->ip6po_prefer_tempaddr == IP6PO_TEMPADDR_SYSTEM) { 414 prefer_tempaddr = V_ip6_prefer_tempaddr; 415 } else if (opts->ip6po_prefer_tempaddr == 416 IP6PO_TEMPADDR_NOTPREFER) { 417 prefer_tempaddr = 0; 418 } else 419 prefer_tempaddr = 1; 420 if (!(ia_best->ia6_flags & IN6_IFF_TEMPORARY) && 421 (ia->ia6_flags & IN6_IFF_TEMPORARY)) { 422 if (prefer_tempaddr) 423 REPLACE(7); 424 else 425 NEXT(7); 426 } 427 if ((ia_best->ia6_flags & IN6_IFF_TEMPORARY) && 428 !(ia->ia6_flags & IN6_IFF_TEMPORARY)) { 429 if (prefer_tempaddr) 430 NEXT(7); 431 else 432 REPLACE(7); 433 } 434 435 /* 436 * Rule 8: prefer addresses on alive interfaces. 437 * This is a KAME specific rule. 438 */ 439 if ((ia_best->ia_ifp->if_flags & IFF_UP) && 440 !(ia->ia_ifp->if_flags & IFF_UP)) 441 NEXT(8); 442 if (!(ia_best->ia_ifp->if_flags & IFF_UP) && 443 (ia->ia_ifp->if_flags & IFF_UP)) 444 REPLACE(8); 445 446 /* 447 * Rule 14: Use longest matching prefix. 448 * Note: in the address selection draft, this rule is 449 * documented as "Rule 8". However, since it is also 450 * documented that this rule can be overridden, we assign 451 * a large number so that it is easy to assign smaller numbers 452 * to more preferred rules. 453 */ 454 new_matchlen = in6_matchlen(&ia->ia_addr.sin6_addr, &dst); 455 if (best_matchlen < new_matchlen) 456 REPLACE(14); 457 if (new_matchlen < best_matchlen) 458 NEXT(14); 459 460 /* Rule 15 is reserved. */ 461 462 /* 463 * Last resort: just keep the current candidate. 464 * Or, do we need more rules? 465 */ 466 continue; 467 468 replace: 469 ia_best = ia; 470 best_scope = (new_scope >= 0 ? new_scope : 471 in6_addrscope(&ia_best->ia_addr.sin6_addr)); 472 best_policy = (new_policy ? new_policy : 473 lookup_addrsel_policy(&ia_best->ia_addr)); 474 best_matchlen = (new_matchlen >= 0 ? new_matchlen : 475 in6_matchlen(&ia_best->ia_addr.sin6_addr, 476 &dst)); 477 478 next: 479 continue; 480 481 out: 482 break; 483 } 484 485 if ((ia = ia_best) == NULL) { 486 IN6_IFADDR_RUNLOCK(); 487 IP6STAT_INC(ip6s_sources_none); 488 return (EADDRNOTAVAIL); 489 } 490 491 /* 492 * At this point at least one of the addresses belonged to the jail 493 * but it could still be, that we want to further restrict it, e.g. 494 * theoratically IN6_IS_ADDR_LOOPBACK. 495 * It must not be IN6_IS_ADDR_UNSPECIFIED anymore. 496 * prison_local_ip6() will fix an IN6_IS_ADDR_LOOPBACK but should 497 * let all others previously selected pass. 498 * Use tmp to not change ::1 on lo0 to the primary jail address. 499 */ 500 tmp = ia->ia_addr.sin6_addr; 501 if (cred != NULL && prison_local_ip6(cred, &tmp, (inp != NULL && 502 (inp->inp_flags & IN6P_IPV6_V6ONLY) != 0)) != 0) { 503 IN6_IFADDR_RUNLOCK(); 504 IP6STAT_INC(ip6s_sources_none); 505 return (EADDRNOTAVAIL); 506 } 507 508 if (ifpp) 509 *ifpp = ifp; 510 511 bcopy(&tmp, srcp, sizeof(*srcp)); 512 if (ia->ia_ifp == ifp) 513 IP6STAT_INC(ip6s_sources_sameif[best_scope]); 514 else 515 IP6STAT_INC(ip6s_sources_otherif[best_scope]); 516 if (dst_scope == best_scope) 517 IP6STAT_INC(ip6s_sources_samescope[best_scope]); 518 else 519 IP6STAT_INC(ip6s_sources_otherscope[best_scope]); 520 if (IFA6_IS_DEPRECATED(ia)) 521 IP6STAT_INC(ip6s_sources_deprecated[best_scope]); 522 IN6_IFADDR_RUNLOCK(); 523 return (0); 524 } 525 526 /* 527 * clone - meaningful only for bsdi and freebsd 528 */ 529 static int 530 selectroute(struct sockaddr_in6 *dstsock, struct ip6_pktopts *opts, 531 struct ip6_moptions *mopts, struct route_in6 *ro, 532 struct ifnet **retifp, struct rtentry **retrt, int norouteok, u_int fibnum) 533 { 534 int error = 0; 535 struct ifnet *ifp = NULL; 536 struct rtentry *rt = NULL; 537 struct sockaddr_in6 *sin6_next; 538 struct in6_pktinfo *pi = NULL; 539 struct in6_addr *dst = &dstsock->sin6_addr; 540 #if 0 541 char ip6buf[INET6_ADDRSTRLEN]; 542 543 if (dstsock->sin6_addr.s6_addr32[0] == 0 && 544 dstsock->sin6_addr.s6_addr32[1] == 0 && 545 !IN6_IS_ADDR_LOOPBACK(&dstsock->sin6_addr)) { 546 printf("in6_selectroute: strange destination %s\n", 547 ip6_sprintf(ip6buf, &dstsock->sin6_addr)); 548 } else { 549 printf("in6_selectroute: destination = %s%%%d\n", 550 ip6_sprintf(ip6buf, &dstsock->sin6_addr), 551 dstsock->sin6_scope_id); /* for debug */ 552 } 553 #endif 554 555 /* If the caller specify the outgoing interface explicitly, use it. */ 556 if (opts && (pi = opts->ip6po_pktinfo) != NULL && pi->ipi6_ifindex) { 557 /* XXX boundary check is assumed to be already done. */ 558 ifp = ifnet_byindex(pi->ipi6_ifindex); 559 if (ifp != NULL && 560 (norouteok || retrt == NULL || 561 IN6_IS_ADDR_MULTICAST(dst))) { 562 /* 563 * we do not have to check or get the route for 564 * multicast. 565 */ 566 goto done; 567 } else 568 goto getroute; 569 } 570 571 /* 572 * If the destination address is a multicast address and the outgoing 573 * interface for the address is specified by the caller, use it. 574 */ 575 if (IN6_IS_ADDR_MULTICAST(dst) && 576 mopts != NULL && (ifp = mopts->im6o_multicast_ifp) != NULL) { 577 goto done; /* we do not need a route for multicast. */ 578 } 579 580 getroute: 581 /* 582 * If the next hop address for the packet is specified by the caller, 583 * use it as the gateway. 584 */ 585 if (opts && opts->ip6po_nexthop) { 586 struct route_in6 *ron; 587 struct llentry *la; 588 589 sin6_next = satosin6(opts->ip6po_nexthop); 590 591 /* at this moment, we only support AF_INET6 next hops */ 592 if (sin6_next->sin6_family != AF_INET6) { 593 error = EAFNOSUPPORT; /* or should we proceed? */ 594 goto done; 595 } 596 597 /* 598 * If the next hop is an IPv6 address, then the node identified 599 * by that address must be a neighbor of the sending host. 600 */ 601 ron = &opts->ip6po_nextroute; 602 /* 603 * XXX what do we do here? 604 * PLZ to be fixing 605 */ 606 607 608 if (ron->ro_rt == NULL) { 609 in6_rtalloc(ron, fibnum); /* multi path case? */ 610 if (ron->ro_rt == NULL) { 611 /* XXX-BZ WT.? */ 612 if (ron->ro_rt) { 613 RTFREE(ron->ro_rt); 614 ron->ro_rt = NULL; 615 } 616 error = EHOSTUNREACH; 617 goto done; 618 } 619 } 620 621 rt = ron->ro_rt; 622 ifp = rt->rt_ifp; 623 IF_AFDATA_RLOCK(ifp); 624 la = lla_lookup(LLTABLE6(ifp), 0, (struct sockaddr *)&sin6_next->sin6_addr); 625 IF_AFDATA_RUNLOCK(ifp); 626 if (la != NULL) 627 LLE_RUNLOCK(la); 628 else { 629 error = EHOSTUNREACH; 630 goto done; 631 } 632 #if 0 633 if ((ron->ro_rt && 634 (ron->ro_rt->rt_flags & (RTF_UP | RTF_LLINFO)) != 635 (RTF_UP | RTF_LLINFO)) || 636 !IN6_ARE_ADDR_EQUAL(&satosin6(&ron->ro_dst)->sin6_addr, 637 &sin6_next->sin6_addr)) { 638 if (ron->ro_rt) { 639 RTFREE(ron->ro_rt); 640 ron->ro_rt = NULL; 641 } 642 *satosin6(&ron->ro_dst) = *sin6_next; 643 } 644 if (ron->ro_rt == NULL) { 645 in6_rtalloc(ron, fibnum); /* multi path case? */ 646 if (ron->ro_rt == NULL || 647 !(ron->ro_rt->rt_flags & RTF_LLINFO)) { 648 if (ron->ro_rt) { 649 RTFREE(ron->ro_rt); 650 ron->ro_rt = NULL; 651 } 652 error = EHOSTUNREACH; 653 goto done; 654 } 655 } 656 #endif 657 658 /* 659 * When cloning is required, try to allocate a route to the 660 * destination so that the caller can store path MTU 661 * information. 662 */ 663 goto done; 664 } 665 666 /* 667 * Use a cached route if it exists and is valid, else try to allocate 668 * a new one. Note that we should check the address family of the 669 * cached destination, in case of sharing the cache with IPv4. 670 */ 671 if (ro) { 672 if (ro->ro_rt && 673 (!(ro->ro_rt->rt_flags & RTF_UP) || 674 ((struct sockaddr *)(&ro->ro_dst))->sa_family != AF_INET6 || 675 !IN6_ARE_ADDR_EQUAL(&satosin6(&ro->ro_dst)->sin6_addr, 676 dst))) { 677 RTFREE(ro->ro_rt); 678 ro->ro_rt = (struct rtentry *)NULL; 679 } 680 if (ro->ro_rt == (struct rtentry *)NULL) { 681 struct sockaddr_in6 *sa6; 682 683 /* No route yet, so try to acquire one */ 684 bzero(&ro->ro_dst, sizeof(struct sockaddr_in6)); 685 sa6 = (struct sockaddr_in6 *)&ro->ro_dst; 686 *sa6 = *dstsock; 687 sa6->sin6_scope_id = 0; 688 689 #ifdef RADIX_MPATH 690 rtalloc_mpath_fib((struct route *)ro, 691 ntohl(sa6->sin6_addr.s6_addr32[3]), fibnum); 692 #else 693 ro->ro_rt = in6_rtalloc1((struct sockaddr *) 694 &ro->ro_dst, 0, 0UL, fibnum); 695 if (ro->ro_rt) 696 RT_UNLOCK(ro->ro_rt); 697 #endif 698 } 699 700 /* 701 * do not care about the result if we have the nexthop 702 * explicitly specified. 703 */ 704 if (opts && opts->ip6po_nexthop) 705 goto done; 706 707 if (ro->ro_rt) { 708 ifp = ro->ro_rt->rt_ifp; 709 710 if (ifp == NULL) { /* can this really happen? */ 711 RTFREE(ro->ro_rt); 712 ro->ro_rt = NULL; 713 } 714 } 715 if (ro->ro_rt == NULL) 716 error = EHOSTUNREACH; 717 rt = ro->ro_rt; 718 719 /* 720 * Check if the outgoing interface conflicts with 721 * the interface specified by ipi6_ifindex (if specified). 722 * Note that loopback interface is always okay. 723 * (this may happen when we are sending a packet to one of 724 * our own addresses.) 725 */ 726 if (ifp && opts && opts->ip6po_pktinfo && 727 opts->ip6po_pktinfo->ipi6_ifindex) { 728 if (!(ifp->if_flags & IFF_LOOPBACK) && 729 ifp->if_index != 730 opts->ip6po_pktinfo->ipi6_ifindex) { 731 error = EHOSTUNREACH; 732 goto done; 733 } 734 } 735 } 736 737 done: 738 if (ifp == NULL && rt == NULL) { 739 /* 740 * This can happen if the caller did not pass a cached route 741 * nor any other hints. We treat this case an error. 742 */ 743 error = EHOSTUNREACH; 744 } 745 if (error == EHOSTUNREACH) 746 IP6STAT_INC(ip6s_noroute); 747 748 if (retifp != NULL) { 749 *retifp = ifp; 750 751 /* 752 * Adjust the "outgoing" interface. If we're going to loop 753 * the packet back to ourselves, the ifp would be the loopback 754 * interface. However, we'd rather know the interface associated 755 * to the destination address (which should probably be one of 756 * our own addresses.) 757 */ 758 if (rt) { 759 if ((rt->rt_ifp->if_flags & IFF_LOOPBACK) && 760 (rt->rt_gateway->sa_family == AF_LINK)) 761 *retifp = 762 ifnet_byindex(((struct sockaddr_dl *) 763 rt->rt_gateway)->sdl_index); 764 } 765 } 766 767 if (retrt != NULL) 768 *retrt = rt; /* rt may be NULL */ 769 770 return (error); 771 } 772 773 static int 774 in6_selectif(struct sockaddr_in6 *dstsock, struct ip6_pktopts *opts, 775 struct ip6_moptions *mopts, struct route_in6 *ro, struct ifnet **retifp, 776 struct ifnet *oifp, u_int fibnum) 777 { 778 int error; 779 struct route_in6 sro; 780 struct rtentry *rt = NULL; 781 782 KASSERT(retifp != NULL, ("%s: retifp is NULL", __func__)); 783 784 if (ro == NULL) { 785 bzero(&sro, sizeof(sro)); 786 ro = &sro; 787 } 788 789 if ((error = selectroute(dstsock, opts, mopts, ro, retifp, 790 &rt, 1, fibnum)) != 0) { 791 if (ro == &sro && rt && rt == sro.ro_rt) 792 RTFREE(rt); 793 /* Help ND. See oifp comment in in6_selectsrc(). */ 794 if (oifp != NULL && fibnum == RT_DEFAULT_FIB) { 795 *retifp = oifp; 796 error = 0; 797 } 798 return (error); 799 } 800 801 /* 802 * do not use a rejected or black hole route. 803 * XXX: this check should be done in the L2 output routine. 804 * However, if we skipped this check here, we'd see the following 805 * scenario: 806 * - install a rejected route for a scoped address prefix 807 * (like fe80::/10) 808 * - send a packet to a destination that matches the scoped prefix, 809 * with ambiguity about the scope zone. 810 * - pick the outgoing interface from the route, and disambiguate the 811 * scope zone with the interface. 812 * - ip6_output() would try to get another route with the "new" 813 * destination, which may be valid. 814 * - we'd see no error on output. 815 * Although this may not be very harmful, it should still be confusing. 816 * We thus reject the case here. 817 */ 818 if (rt && (rt->rt_flags & (RTF_REJECT | RTF_BLACKHOLE))) { 819 int flags = (rt->rt_flags & RTF_HOST ? EHOSTUNREACH : ENETUNREACH); 820 821 if (ro == &sro && rt && rt == sro.ro_rt) 822 RTFREE(rt); 823 return (flags); 824 } 825 826 if (ro == &sro && rt && rt == sro.ro_rt) 827 RTFREE(rt); 828 return (0); 829 } 830 831 /* 832 * Public wrapper function to selectroute(). 833 * 834 * XXX-BZ in6_selectroute() should and will grow the FIB argument. The 835 * in6_selectroute_fib() function is only there for backward compat on stable. 836 */ 837 int 838 in6_selectroute(struct sockaddr_in6 *dstsock, struct ip6_pktopts *opts, 839 struct ip6_moptions *mopts, struct route_in6 *ro, 840 struct ifnet **retifp, struct rtentry **retrt) 841 { 842 843 return (selectroute(dstsock, opts, mopts, ro, retifp, 844 retrt, 0, RT_DEFAULT_FIB)); 845 } 846 847 #ifndef BURN_BRIDGES 848 int 849 in6_selectroute_fib(struct sockaddr_in6 *dstsock, struct ip6_pktopts *opts, 850 struct ip6_moptions *mopts, struct route_in6 *ro, 851 struct ifnet **retifp, struct rtentry **retrt, u_int fibnum) 852 { 853 854 return (selectroute(dstsock, opts, mopts, ro, retifp, 855 retrt, 0, fibnum)); 856 } 857 #endif 858 859 /* 860 * Default hop limit selection. The precedence is as follows: 861 * 1. Hoplimit value specified via ioctl. 862 * 2. (If the outgoing interface is detected) the current 863 * hop limit of the interface specified by router advertisement. 864 * 3. The system default hoplimit. 865 */ 866 int 867 in6_selecthlim(struct inpcb *in6p, struct ifnet *ifp) 868 { 869 870 if (in6p && in6p->in6p_hops >= 0) 871 return (in6p->in6p_hops); 872 else if (ifp) 873 return (ND_IFINFO(ifp)->chlim); 874 else if (in6p && !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr)) { 875 struct route_in6 ro6; 876 struct ifnet *lifp; 877 878 bzero(&ro6, sizeof(ro6)); 879 ro6.ro_dst.sin6_family = AF_INET6; 880 ro6.ro_dst.sin6_len = sizeof(struct sockaddr_in6); 881 ro6.ro_dst.sin6_addr = in6p->in6p_faddr; 882 in6_rtalloc(&ro6, in6p->inp_inc.inc_fibnum); 883 if (ro6.ro_rt) { 884 lifp = ro6.ro_rt->rt_ifp; 885 RTFREE(ro6.ro_rt); 886 if (lifp) 887 return (ND_IFINFO(lifp)->chlim); 888 } 889 } 890 return (V_ip6_defhlim); 891 } 892 893 /* 894 * XXX: this is borrowed from in6_pcbbind(). If possible, we should 895 * share this function by all *bsd*... 896 */ 897 int 898 in6_pcbsetport(struct in6_addr *laddr, struct inpcb *inp, struct ucred *cred) 899 { 900 struct socket *so = inp->inp_socket; 901 u_int16_t lport = 0; 902 int error, lookupflags = 0; 903 #ifdef INVARIANTS 904 struct inpcbinfo *pcbinfo = inp->inp_pcbinfo; 905 #endif 906 907 INP_WLOCK_ASSERT(inp); 908 INP_HASH_WLOCK_ASSERT(pcbinfo); 909 910 error = prison_local_ip6(cred, laddr, 911 ((inp->inp_flags & IN6P_IPV6_V6ONLY) != 0)); 912 if (error) 913 return(error); 914 915 /* XXX: this is redundant when called from in6_pcbbind */ 916 if ((so->so_options & (SO_REUSEADDR|SO_REUSEPORT)) == 0) 917 lookupflags = INPLOOKUP_WILDCARD; 918 919 inp->inp_flags |= INP_ANONPORT; 920 921 error = in_pcb_lport(inp, NULL, &lport, cred, lookupflags); 922 if (error != 0) 923 return (error); 924 925 inp->inp_lport = lport; 926 if (in_pcbinshash(inp) != 0) { 927 inp->in6p_laddr = in6addr_any; 928 inp->inp_lport = 0; 929 return (EAGAIN); 930 } 931 932 return (0); 933 } 934 935 void 936 addrsel_policy_init(void) 937 { 938 939 init_policy_queue(); 940 941 /* initialize the "last resort" policy */ 942 bzero(&V_defaultaddrpolicy, sizeof(V_defaultaddrpolicy)); 943 V_defaultaddrpolicy.label = ADDR_LABEL_NOTAPP; 944 945 if (!IS_DEFAULT_VNET(curvnet)) 946 return; 947 948 ADDRSEL_LOCK_INIT(); 949 ADDRSEL_SXLOCK_INIT(); 950 } 951 952 static struct in6_addrpolicy * 953 lookup_addrsel_policy(struct sockaddr_in6 *key) 954 { 955 struct in6_addrpolicy *match = NULL; 956 957 ADDRSEL_LOCK(); 958 match = match_addrsel_policy(key); 959 960 if (match == NULL) 961 match = &V_defaultaddrpolicy; 962 else 963 match->use++; 964 ADDRSEL_UNLOCK(); 965 966 return (match); 967 } 968 969 /* 970 * Subroutines to manage the address selection policy table via sysctl. 971 */ 972 struct walkarg { 973 struct sysctl_req *w_req; 974 }; 975 976 static int in6_src_sysctl(SYSCTL_HANDLER_ARGS); 977 SYSCTL_DECL(_net_inet6_ip6); 978 static SYSCTL_NODE(_net_inet6_ip6, IPV6CTL_ADDRCTLPOLICY, addrctlpolicy, 979 CTLFLAG_RD, in6_src_sysctl, ""); 980 981 static int 982 in6_src_sysctl(SYSCTL_HANDLER_ARGS) 983 { 984 struct walkarg w; 985 986 if (req->newptr) 987 return EPERM; 988 989 bzero(&w, sizeof(w)); 990 w.w_req = req; 991 992 return (walk_addrsel_policy(dump_addrsel_policyent, &w)); 993 } 994 995 int 996 in6_src_ioctl(u_long cmd, caddr_t data) 997 { 998 int i; 999 struct in6_addrpolicy ent0; 1000 1001 if (cmd != SIOCAADDRCTL_POLICY && cmd != SIOCDADDRCTL_POLICY) 1002 return (EOPNOTSUPP); /* check for safety */ 1003 1004 ent0 = *(struct in6_addrpolicy *)data; 1005 1006 if (ent0.label == ADDR_LABEL_NOTAPP) 1007 return (EINVAL); 1008 /* check if the prefix mask is consecutive. */ 1009 if (in6_mask2len(&ent0.addrmask.sin6_addr, NULL) < 0) 1010 return (EINVAL); 1011 /* clear trailing garbages (if any) of the prefix address. */ 1012 for (i = 0; i < 4; i++) { 1013 ent0.addr.sin6_addr.s6_addr32[i] &= 1014 ent0.addrmask.sin6_addr.s6_addr32[i]; 1015 } 1016 ent0.use = 0; 1017 1018 switch (cmd) { 1019 case SIOCAADDRCTL_POLICY: 1020 return (add_addrsel_policyent(&ent0)); 1021 case SIOCDADDRCTL_POLICY: 1022 return (delete_addrsel_policyent(&ent0)); 1023 } 1024 1025 return (0); /* XXX: compromise compilers */ 1026 } 1027 1028 /* 1029 * The followings are implementation of the policy table using a 1030 * simple tail queue. 1031 * XXX such details should be hidden. 1032 * XXX implementation using binary tree should be more efficient. 1033 */ 1034 struct addrsel_policyent { 1035 TAILQ_ENTRY(addrsel_policyent) ape_entry; 1036 struct in6_addrpolicy ape_policy; 1037 }; 1038 1039 TAILQ_HEAD(addrsel_policyhead, addrsel_policyent); 1040 1041 static VNET_DEFINE(struct addrsel_policyhead, addrsel_policytab); 1042 #define V_addrsel_policytab VNET(addrsel_policytab) 1043 1044 static void 1045 init_policy_queue(void) 1046 { 1047 1048 TAILQ_INIT(&V_addrsel_policytab); 1049 } 1050 1051 static int 1052 add_addrsel_policyent(struct in6_addrpolicy *newpolicy) 1053 { 1054 struct addrsel_policyent *new, *pol; 1055 1056 new = malloc(sizeof(*new), M_IFADDR, 1057 M_WAITOK); 1058 ADDRSEL_XLOCK(); 1059 ADDRSEL_LOCK(); 1060 1061 /* duplication check */ 1062 TAILQ_FOREACH(pol, &V_addrsel_policytab, ape_entry) { 1063 if (IN6_ARE_ADDR_EQUAL(&newpolicy->addr.sin6_addr, 1064 &pol->ape_policy.addr.sin6_addr) && 1065 IN6_ARE_ADDR_EQUAL(&newpolicy->addrmask.sin6_addr, 1066 &pol->ape_policy.addrmask.sin6_addr)) { 1067 ADDRSEL_UNLOCK(); 1068 ADDRSEL_XUNLOCK(); 1069 free(new, M_IFADDR); 1070 return (EEXIST); /* or override it? */ 1071 } 1072 } 1073 1074 bzero(new, sizeof(*new)); 1075 1076 /* XXX: should validate entry */ 1077 new->ape_policy = *newpolicy; 1078 1079 TAILQ_INSERT_TAIL(&V_addrsel_policytab, new, ape_entry); 1080 ADDRSEL_UNLOCK(); 1081 ADDRSEL_XUNLOCK(); 1082 1083 return (0); 1084 } 1085 1086 static int 1087 delete_addrsel_policyent(struct in6_addrpolicy *key) 1088 { 1089 struct addrsel_policyent *pol; 1090 1091 ADDRSEL_XLOCK(); 1092 ADDRSEL_LOCK(); 1093 1094 /* search for the entry in the table */ 1095 TAILQ_FOREACH(pol, &V_addrsel_policytab, ape_entry) { 1096 if (IN6_ARE_ADDR_EQUAL(&key->addr.sin6_addr, 1097 &pol->ape_policy.addr.sin6_addr) && 1098 IN6_ARE_ADDR_EQUAL(&key->addrmask.sin6_addr, 1099 &pol->ape_policy.addrmask.sin6_addr)) { 1100 break; 1101 } 1102 } 1103 if (pol == NULL) { 1104 ADDRSEL_UNLOCK(); 1105 ADDRSEL_XUNLOCK(); 1106 return (ESRCH); 1107 } 1108 1109 TAILQ_REMOVE(&V_addrsel_policytab, pol, ape_entry); 1110 ADDRSEL_UNLOCK(); 1111 ADDRSEL_XUNLOCK(); 1112 free(pol, M_IFADDR); 1113 1114 return (0); 1115 } 1116 1117 static int 1118 walk_addrsel_policy(int (*callback)(struct in6_addrpolicy *, void *), void *w) 1119 { 1120 struct addrsel_policyent *pol; 1121 int error = 0; 1122 1123 ADDRSEL_SLOCK(); 1124 TAILQ_FOREACH(pol, &V_addrsel_policytab, ape_entry) { 1125 if ((error = (*callback)(&pol->ape_policy, w)) != 0) { 1126 ADDRSEL_SUNLOCK(); 1127 return (error); 1128 } 1129 } 1130 ADDRSEL_SUNLOCK(); 1131 return (error); 1132 } 1133 1134 static int 1135 dump_addrsel_policyent(struct in6_addrpolicy *pol, void *arg) 1136 { 1137 int error = 0; 1138 struct walkarg *w = arg; 1139 1140 error = SYSCTL_OUT(w->w_req, pol, sizeof(*pol)); 1141 1142 return (error); 1143 } 1144 1145 static struct in6_addrpolicy * 1146 match_addrsel_policy(struct sockaddr_in6 *key) 1147 { 1148 struct addrsel_policyent *pent; 1149 struct in6_addrpolicy *bestpol = NULL, *pol; 1150 int matchlen, bestmatchlen = -1; 1151 u_char *mp, *ep, *k, *p, m; 1152 1153 TAILQ_FOREACH(pent, &V_addrsel_policytab, ape_entry) { 1154 matchlen = 0; 1155 1156 pol = &pent->ape_policy; 1157 mp = (u_char *)&pol->addrmask.sin6_addr; 1158 ep = mp + 16; /* XXX: scope field? */ 1159 k = (u_char *)&key->sin6_addr; 1160 p = (u_char *)&pol->addr.sin6_addr; 1161 for (; mp < ep && *mp; mp++, k++, p++) { 1162 m = *mp; 1163 if ((*k & m) != *p) 1164 goto next; /* not match */ 1165 if (m == 0xff) /* short cut for a typical case */ 1166 matchlen += 8; 1167 else { 1168 while (m >= 0x80) { 1169 matchlen++; 1170 m <<= 1; 1171 } 1172 } 1173 } 1174 1175 /* matched. check if this is better than the current best. */ 1176 if (bestpol == NULL || 1177 matchlen > bestmatchlen) { 1178 bestpol = pol; 1179 bestmatchlen = matchlen; 1180 } 1181 1182 next: 1183 continue; 1184 } 1185 1186 return (bestpol); 1187 } 1188