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 #include <sys/vimage.h> 87 88 #include <net/if.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 #include <netinet/vinet.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 #include <netinet6/vinet6.h> 112 113 static struct mtx addrsel_lock; 114 #define ADDRSEL_LOCK_INIT() mtx_init(&addrsel_lock, "addrsel_lock", NULL, MTX_DEF) 115 #define ADDRSEL_LOCK() mtx_lock(&addrsel_lock) 116 #define ADDRSEL_UNLOCK() mtx_unlock(&addrsel_lock) 117 #define ADDRSEL_LOCK_ASSERT() mtx_assert(&addrsel_lock, MA_OWNED) 118 119 static struct sx addrsel_sxlock; 120 #define ADDRSEL_SXLOCK_INIT() sx_init(&addrsel_sxlock, "addrsel_sxlock") 121 #define ADDRSEL_SLOCK() sx_slock(&addrsel_sxlock) 122 #define ADDRSEL_SUNLOCK() sx_sunlock(&addrsel_sxlock) 123 #define ADDRSEL_XLOCK() sx_xlock(&addrsel_sxlock) 124 #define ADDRSEL_XUNLOCK() sx_xunlock(&addrsel_sxlock) 125 126 #define ADDR_LABEL_NOTAPP (-1) 127 128 #ifdef VIMAGE_GLOBALS 129 struct in6_addrpolicy defaultaddrpolicy; 130 int ip6_prefer_tempaddr; 131 #endif 132 133 static int selectroute __P((struct sockaddr_in6 *, struct ip6_pktopts *, 134 struct ip6_moptions *, struct route_in6 *, struct ifnet **, 135 struct rtentry **, int)); 136 static int in6_selectif __P((struct sockaddr_in6 *, struct ip6_pktopts *, 137 struct ip6_moptions *, struct route_in6 *ro, struct ifnet **)); 138 139 static struct in6_addrpolicy *lookup_addrsel_policy(struct sockaddr_in6 *); 140 141 static void init_policy_queue(void); 142 static int add_addrsel_policyent(struct in6_addrpolicy *); 143 static int delete_addrsel_policyent(struct in6_addrpolicy *); 144 static int walk_addrsel_policy __P((int (*)(struct in6_addrpolicy *, void *), 145 void *)); 146 static int dump_addrsel_policyent(struct in6_addrpolicy *, void *); 147 static struct in6_addrpolicy *match_addrsel_policy(struct sockaddr_in6 *); 148 149 /* 150 * Return an IPv6 address, which is the most appropriate for a given 151 * destination and user specified options. 152 * If necessary, this function lookups the routing table and returns 153 * an entry to the caller for later use. 154 */ 155 #define REPLACE(r) do {\ 156 if ((r) < sizeof(V_ip6stat.ip6s_sources_rule) / \ 157 sizeof(V_ip6stat.ip6s_sources_rule[0])) /* check for safety */ \ 158 V_ip6stat.ip6s_sources_rule[(r)]++; \ 159 /* { \ 160 char ip6buf[INET6_ADDRSTRLEN], ip6b[INET6_ADDRSTRLEN]; \ 161 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)); \ 162 } */ \ 163 goto replace; \ 164 } while(0) 165 #define NEXT(r) do {\ 166 if ((r) < sizeof(V_ip6stat.ip6s_sources_rule) / \ 167 sizeof(V_ip6stat.ip6s_sources_rule[0])) /* check for safety */ \ 168 V_ip6stat.ip6s_sources_rule[(r)]++; \ 169 /* { \ 170 char ip6buf[INET6_ADDRSTRLEN], ip6b[INET6_ADDRSTRLEN]; \ 171 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)); \ 172 } */ \ 173 goto next; /* XXX: we can't use 'continue' here */ \ 174 } while(0) 175 #define BREAK(r) do { \ 176 if ((r) < sizeof(V_ip6stat.ip6s_sources_rule) / \ 177 sizeof(V_ip6stat.ip6s_sources_rule[0])) /* check for safety */ \ 178 V_ip6stat.ip6s_sources_rule[(r)]++; \ 179 goto out; /* XXX: we can't use 'break' here */ \ 180 } while(0) 181 182 struct in6_addr * 183 in6_selectsrc(struct sockaddr_in6 *dstsock, struct ip6_pktopts *opts, 184 struct inpcb *inp, struct route_in6 *ro, struct ucred *cred, 185 struct ifnet **ifpp, int *errorp) 186 { 187 INIT_VNET_INET6(curvnet); 188 struct in6_addr dst; 189 struct ifnet *ifp = NULL; 190 struct in6_ifaddr *ia = NULL, *ia_best = NULL; 191 struct in6_pktinfo *pi = NULL; 192 int dst_scope = -1, best_scope = -1, best_matchlen = -1; 193 struct in6_addrpolicy *dst_policy = NULL, *best_policy = NULL; 194 u_int32_t odstzone; 195 int prefer_tempaddr; 196 struct ip6_moptions *mopts; 197 198 dst = dstsock->sin6_addr; /* make a copy for local operation */ 199 *errorp = 0; 200 if (ifpp) 201 *ifpp = NULL; 202 203 if (inp != NULL) { 204 INP_LOCK_ASSERT(inp); 205 mopts = inp->in6p_moptions; 206 } else { 207 mopts = NULL; 208 } 209 210 /* 211 * If the source address is explicitly specified by the caller, 212 * check if the requested source address is indeed a unicast address 213 * assigned to the node, and can be used as the packet's source 214 * address. If everything is okay, use the address as source. 215 */ 216 if (opts && (pi = opts->ip6po_pktinfo) && 217 !IN6_IS_ADDR_UNSPECIFIED(&pi->ipi6_addr)) { 218 struct sockaddr_in6 srcsock; 219 struct in6_ifaddr *ia6; 220 221 /* get the outgoing interface */ 222 if ((*errorp = in6_selectif(dstsock, opts, mopts, ro, &ifp)) 223 != 0) { 224 return (NULL); 225 } 226 227 /* 228 * determine the appropriate zone id of the source based on 229 * the zone of the destination and the outgoing interface. 230 * If the specified address is ambiguous wrt the scope zone, 231 * the interface must be specified; otherwise, ifa_ifwithaddr() 232 * will fail matching the address. 233 */ 234 bzero(&srcsock, sizeof(srcsock)); 235 srcsock.sin6_family = AF_INET6; 236 srcsock.sin6_len = sizeof(srcsock); 237 srcsock.sin6_addr = pi->ipi6_addr; 238 if (ifp) { 239 *errorp = in6_setscope(&srcsock.sin6_addr, ifp, NULL); 240 if (*errorp != 0) 241 return (NULL); 242 } 243 if (cred != NULL && (*errorp = prison_local_ip6(cred, 244 &srcsock.sin6_addr, (inp != NULL && 245 (inp->inp_flags & IN6P_IPV6_V6ONLY) != 0))) != 0) 246 return (NULL); 247 248 ia6 = (struct in6_ifaddr *)ifa_ifwithaddr((struct sockaddr *)(&srcsock)); 249 if (ia6 == NULL || 250 (ia6->ia6_flags & (IN6_IFF_ANYCAST | IN6_IFF_NOTREADY))) { 251 *errorp = EADDRNOTAVAIL; 252 return (NULL); 253 } 254 pi->ipi6_addr = srcsock.sin6_addr; /* XXX: this overrides pi */ 255 if (ifpp) 256 *ifpp = ifp; 257 return (&ia6->ia_addr.sin6_addr); 258 } 259 260 /* 261 * Otherwise, if the socket has already bound the source, just use it. 262 */ 263 if (inp != NULL && !IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) { 264 if (cred != NULL && 265 (*errorp = prison_local_ip6(cred, &inp->in6p_laddr, 266 ((inp->inp_flags & IN6P_IPV6_V6ONLY) != 0))) != 0) 267 return (NULL); 268 return (&inp->in6p_laddr); 269 } 270 271 /* 272 * If the address is not specified, choose the best one based on 273 * the outgoing interface and the destination address. 274 */ 275 /* get the outgoing interface */ 276 if ((*errorp = in6_selectif(dstsock, opts, mopts, ro, &ifp)) != 0) 277 return (NULL); 278 279 #ifdef DIAGNOSTIC 280 if (ifp == NULL) /* this should not happen */ 281 panic("in6_selectsrc: NULL ifp"); 282 #endif 283 *errorp = in6_setscope(&dst, ifp, &odstzone); 284 if (*errorp != 0) 285 return (NULL); 286 287 for (ia = V_in6_ifaddr; ia; ia = ia->ia_next) { 288 int new_scope = -1, new_matchlen = -1; 289 struct in6_addrpolicy *new_policy = NULL; 290 u_int32_t srczone, osrczone, dstzone; 291 struct in6_addr src; 292 struct ifnet *ifp1 = ia->ia_ifp; 293 294 /* 295 * We'll never take an address that breaks the scope zone 296 * of the destination. We also skip an address if its zone 297 * does not contain the outgoing interface. 298 * XXX: we should probably use sin6_scope_id here. 299 */ 300 if (in6_setscope(&dst, ifp1, &dstzone) || 301 odstzone != dstzone) { 302 continue; 303 } 304 src = ia->ia_addr.sin6_addr; 305 if (in6_setscope(&src, ifp, &osrczone) || 306 in6_setscope(&src, ifp1, &srczone) || 307 osrczone != srczone) { 308 continue; 309 } 310 311 /* avoid unusable addresses */ 312 if ((ia->ia6_flags & 313 (IN6_IFF_NOTREADY | IN6_IFF_ANYCAST | IN6_IFF_DETACHED))) { 314 continue; 315 } 316 if (!V_ip6_use_deprecated && IFA6_IS_DEPRECATED(ia)) 317 continue; 318 319 if (cred != NULL && 320 prison_local_ip6(cred, &ia->ia_addr.sin6_addr, 321 (inp != NULL && 322 (inp->inp_flags & IN6P_IPV6_V6ONLY) != 0)) != 0) 323 continue; 324 325 /* Rule 1: Prefer same address */ 326 if (IN6_ARE_ADDR_EQUAL(&dst, &ia->ia_addr.sin6_addr)) { 327 ia_best = ia; 328 BREAK(1); /* there should be no better candidate */ 329 } 330 331 if (ia_best == NULL) 332 REPLACE(0); 333 334 /* Rule 2: Prefer appropriate scope */ 335 if (dst_scope < 0) 336 dst_scope = in6_addrscope(&dst); 337 new_scope = in6_addrscope(&ia->ia_addr.sin6_addr); 338 if (IN6_ARE_SCOPE_CMP(best_scope, new_scope) < 0) { 339 if (IN6_ARE_SCOPE_CMP(best_scope, dst_scope) < 0) 340 REPLACE(2); 341 NEXT(2); 342 } else if (IN6_ARE_SCOPE_CMP(new_scope, best_scope) < 0) { 343 if (IN6_ARE_SCOPE_CMP(new_scope, dst_scope) < 0) 344 NEXT(2); 345 REPLACE(2); 346 } 347 348 /* 349 * Rule 3: Avoid deprecated addresses. Note that the case of 350 * !ip6_use_deprecated is already rejected above. 351 */ 352 if (!IFA6_IS_DEPRECATED(ia_best) && IFA6_IS_DEPRECATED(ia)) 353 NEXT(3); 354 if (IFA6_IS_DEPRECATED(ia_best) && !IFA6_IS_DEPRECATED(ia)) 355 REPLACE(3); 356 357 /* Rule 4: Prefer home addresses */ 358 /* 359 * XXX: This is a TODO. We should probably merge the MIP6 360 * case above. 361 */ 362 363 /* Rule 5: Prefer outgoing interface */ 364 if (ia_best->ia_ifp == ifp && ia->ia_ifp != ifp) 365 NEXT(5); 366 if (ia_best->ia_ifp != ifp && ia->ia_ifp == ifp) 367 REPLACE(5); 368 369 /* 370 * Rule 6: Prefer matching label 371 * Note that best_policy should be non-NULL here. 372 */ 373 if (dst_policy == NULL) 374 dst_policy = lookup_addrsel_policy(dstsock); 375 if (dst_policy->label != ADDR_LABEL_NOTAPP) { 376 new_policy = lookup_addrsel_policy(&ia->ia_addr); 377 if (dst_policy->label == best_policy->label && 378 dst_policy->label != new_policy->label) 379 NEXT(6); 380 if (dst_policy->label != best_policy->label && 381 dst_policy->label == new_policy->label) 382 REPLACE(6); 383 } 384 385 /* 386 * Rule 7: Prefer public addresses. 387 * We allow users to reverse the logic by configuring 388 * a sysctl variable, so that privacy conscious users can 389 * always prefer temporary addresses. 390 */ 391 if (opts == NULL || 392 opts->ip6po_prefer_tempaddr == IP6PO_TEMPADDR_SYSTEM) { 393 prefer_tempaddr = V_ip6_prefer_tempaddr; 394 } else if (opts->ip6po_prefer_tempaddr == 395 IP6PO_TEMPADDR_NOTPREFER) { 396 prefer_tempaddr = 0; 397 } else 398 prefer_tempaddr = 1; 399 if (!(ia_best->ia6_flags & IN6_IFF_TEMPORARY) && 400 (ia->ia6_flags & IN6_IFF_TEMPORARY)) { 401 if (prefer_tempaddr) 402 REPLACE(7); 403 else 404 NEXT(7); 405 } 406 if ((ia_best->ia6_flags & IN6_IFF_TEMPORARY) && 407 !(ia->ia6_flags & IN6_IFF_TEMPORARY)) { 408 if (prefer_tempaddr) 409 NEXT(7); 410 else 411 REPLACE(7); 412 } 413 414 /* 415 * Rule 8: prefer addresses on alive interfaces. 416 * This is a KAME specific rule. 417 */ 418 if ((ia_best->ia_ifp->if_flags & IFF_UP) && 419 !(ia->ia_ifp->if_flags & IFF_UP)) 420 NEXT(8); 421 if (!(ia_best->ia_ifp->if_flags & IFF_UP) && 422 (ia->ia_ifp->if_flags & IFF_UP)) 423 REPLACE(8); 424 425 /* 426 * Rule 14: Use longest matching prefix. 427 * Note: in the address selection draft, this rule is 428 * documented as "Rule 8". However, since it is also 429 * documented that this rule can be overridden, we assign 430 * a large number so that it is easy to assign smaller numbers 431 * to more preferred rules. 432 */ 433 new_matchlen = in6_matchlen(&ia->ia_addr.sin6_addr, &dst); 434 if (best_matchlen < new_matchlen) 435 REPLACE(14); 436 if (new_matchlen < best_matchlen) 437 NEXT(14); 438 439 /* Rule 15 is reserved. */ 440 441 /* 442 * Last resort: just keep the current candidate. 443 * Or, do we need more rules? 444 */ 445 continue; 446 447 replace: 448 ia_best = ia; 449 best_scope = (new_scope >= 0 ? new_scope : 450 in6_addrscope(&ia_best->ia_addr.sin6_addr)); 451 best_policy = (new_policy ? new_policy : 452 lookup_addrsel_policy(&ia_best->ia_addr)); 453 best_matchlen = (new_matchlen >= 0 ? new_matchlen : 454 in6_matchlen(&ia_best->ia_addr.sin6_addr, 455 &dst)); 456 457 next: 458 continue; 459 460 out: 461 break; 462 } 463 464 if ((ia = ia_best) == NULL) { 465 *errorp = EADDRNOTAVAIL; 466 return (NULL); 467 } 468 469 if (ifpp) 470 *ifpp = ifp; 471 472 return (&ia->ia_addr.sin6_addr); 473 } 474 475 /* 476 * clone - meaningful only for bsdi and freebsd 477 */ 478 static int 479 selectroute(struct sockaddr_in6 *dstsock, struct ip6_pktopts *opts, 480 struct ip6_moptions *mopts, struct route_in6 *ro, 481 struct ifnet **retifp, struct rtentry **retrt, int norouteok) 482 { 483 INIT_VNET_INET6(curvnet); 484 int error = 0; 485 struct ifnet *ifp = NULL; 486 struct rtentry *rt = NULL; 487 struct sockaddr_in6 *sin6_next; 488 struct in6_pktinfo *pi = NULL; 489 struct in6_addr *dst = &dstsock->sin6_addr; 490 #if 0 491 char ip6buf[INET6_ADDRSTRLEN]; 492 493 if (dstsock->sin6_addr.s6_addr32[0] == 0 && 494 dstsock->sin6_addr.s6_addr32[1] == 0 && 495 !IN6_IS_ADDR_LOOPBACK(&dstsock->sin6_addr)) { 496 printf("in6_selectroute: strange destination %s\n", 497 ip6_sprintf(ip6buf, &dstsock->sin6_addr)); 498 } else { 499 printf("in6_selectroute: destination = %s%%%d\n", 500 ip6_sprintf(ip6buf, &dstsock->sin6_addr), 501 dstsock->sin6_scope_id); /* for debug */ 502 } 503 #endif 504 505 /* If the caller specify the outgoing interface explicitly, use it. */ 506 if (opts && (pi = opts->ip6po_pktinfo) != NULL && pi->ipi6_ifindex) { 507 /* XXX boundary check is assumed to be already done. */ 508 ifp = ifnet_byindex(pi->ipi6_ifindex); 509 if (ifp != NULL && 510 (norouteok || retrt == NULL || 511 IN6_IS_ADDR_MULTICAST(dst))) { 512 /* 513 * we do not have to check or get the route for 514 * multicast. 515 */ 516 goto done; 517 } else 518 goto getroute; 519 } 520 521 /* 522 * If the destination address is a multicast address and the outgoing 523 * interface for the address is specified by the caller, use it. 524 */ 525 if (IN6_IS_ADDR_MULTICAST(dst) && 526 mopts != NULL && (ifp = mopts->im6o_multicast_ifp) != NULL) { 527 goto done; /* we do not need a route for multicast. */ 528 } 529 530 getroute: 531 /* 532 * If the next hop address for the packet is specified by the caller, 533 * use it as the gateway. 534 */ 535 if (opts && opts->ip6po_nexthop) { 536 struct route_in6 *ron; 537 struct llentry *la; 538 539 sin6_next = satosin6(opts->ip6po_nexthop); 540 541 /* at this moment, we only support AF_INET6 next hops */ 542 if (sin6_next->sin6_family != AF_INET6) { 543 error = EAFNOSUPPORT; /* or should we proceed? */ 544 goto done; 545 } 546 547 /* 548 * If the next hop is an IPv6 address, then the node identified 549 * by that address must be a neighbor of the sending host. 550 */ 551 ron = &opts->ip6po_nextroute; 552 /* 553 * XXX what do we do here? 554 * PLZ to be fixing 555 */ 556 557 558 if (ron->ro_rt == NULL) { 559 rtalloc((struct route *)ron); /* multi path case? */ 560 if (ron->ro_rt == NULL) { 561 if (ron->ro_rt) { 562 RTFREE(ron->ro_rt); 563 ron->ro_rt = NULL; 564 } 565 error = EHOSTUNREACH; 566 goto done; 567 } 568 } 569 570 rt = ron->ro_rt; 571 ifp = rt->rt_ifp; 572 IF_AFDATA_LOCK(ifp); 573 la = lla_lookup(LLTABLE6(ifp), 0, (struct sockaddr *)&sin6_next->sin6_addr); 574 IF_AFDATA_UNLOCK(ifp); 575 if (la != NULL) 576 LLE_RUNLOCK(la); 577 else { 578 error = EHOSTUNREACH; 579 goto done; 580 } 581 #if 0 582 if ((ron->ro_rt && 583 (ron->ro_rt->rt_flags & (RTF_UP | RTF_LLINFO)) != 584 (RTF_UP | RTF_LLINFO)) || 585 !IN6_ARE_ADDR_EQUAL(&satosin6(&ron->ro_dst)->sin6_addr, 586 &sin6_next->sin6_addr)) { 587 if (ron->ro_rt) { 588 RTFREE(ron->ro_rt); 589 ron->ro_rt = NULL; 590 } 591 *satosin6(&ron->ro_dst) = *sin6_next; 592 } 593 if (ron->ro_rt == NULL) { 594 rtalloc((struct route *)ron); /* multi path case? */ 595 if (ron->ro_rt == NULL || 596 !(ron->ro_rt->rt_flags & RTF_LLINFO)) { 597 if (ron->ro_rt) { 598 RTFREE(ron->ro_rt); 599 ron->ro_rt = NULL; 600 } 601 error = EHOSTUNREACH; 602 goto done; 603 } 604 } 605 #endif 606 607 /* 608 * When cloning is required, try to allocate a route to the 609 * destination so that the caller can store path MTU 610 * information. 611 */ 612 goto done; 613 } 614 615 /* 616 * Use a cached route if it exists and is valid, else try to allocate 617 * a new one. Note that we should check the address family of the 618 * cached destination, in case of sharing the cache with IPv4. 619 */ 620 if (ro) { 621 if (ro->ro_rt && 622 (!(ro->ro_rt->rt_flags & RTF_UP) || 623 ((struct sockaddr *)(&ro->ro_dst))->sa_family != AF_INET6 || 624 !IN6_ARE_ADDR_EQUAL(&satosin6(&ro->ro_dst)->sin6_addr, 625 dst))) { 626 RTFREE(ro->ro_rt); 627 ro->ro_rt = (struct rtentry *)NULL; 628 } 629 if (ro->ro_rt == (struct rtentry *)NULL) { 630 struct sockaddr_in6 *sa6; 631 632 /* No route yet, so try to acquire one */ 633 bzero(&ro->ro_dst, sizeof(struct sockaddr_in6)); 634 sa6 = (struct sockaddr_in6 *)&ro->ro_dst; 635 *sa6 = *dstsock; 636 sa6->sin6_scope_id = 0; 637 638 #ifdef RADIX_MPATH 639 rtalloc_mpath((struct route *)ro, 640 ntohl(sa6->sin6_addr.s6_addr32[3])); 641 #else 642 ro->ro_rt = rtalloc1(&((struct route *)ro) 643 ->ro_dst, 0, 0UL); 644 if (ro->ro_rt) 645 RT_UNLOCK(ro->ro_rt); 646 #endif 647 } 648 649 /* 650 * do not care about the result if we have the nexthop 651 * explicitly specified. 652 */ 653 if (opts && opts->ip6po_nexthop) 654 goto done; 655 656 if (ro->ro_rt) { 657 ifp = ro->ro_rt->rt_ifp; 658 659 if (ifp == NULL) { /* can this really happen? */ 660 RTFREE(ro->ro_rt); 661 ro->ro_rt = NULL; 662 } 663 } 664 if (ro->ro_rt == NULL) 665 error = EHOSTUNREACH; 666 rt = ro->ro_rt; 667 668 /* 669 * Check if the outgoing interface conflicts with 670 * the interface specified by ipi6_ifindex (if specified). 671 * Note that loopback interface is always okay. 672 * (this may happen when we are sending a packet to one of 673 * our own addresses.) 674 */ 675 if (ifp && opts && opts->ip6po_pktinfo && 676 opts->ip6po_pktinfo->ipi6_ifindex) { 677 if (!(ifp->if_flags & IFF_LOOPBACK) && 678 ifp->if_index != 679 opts->ip6po_pktinfo->ipi6_ifindex) { 680 error = EHOSTUNREACH; 681 goto done; 682 } 683 } 684 } 685 686 done: 687 if (ifp == NULL && rt == NULL) { 688 /* 689 * This can happen if the caller did not pass a cached route 690 * nor any other hints. We treat this case an error. 691 */ 692 error = EHOSTUNREACH; 693 } 694 if (error == EHOSTUNREACH) 695 V_ip6stat.ip6s_noroute++; 696 697 if (retifp != NULL) 698 *retifp = ifp; 699 if (retrt != NULL) 700 *retrt = rt; /* rt may be NULL */ 701 702 return (error); 703 } 704 705 static int 706 in6_selectif(struct sockaddr_in6 *dstsock, struct ip6_pktopts *opts, 707 struct ip6_moptions *mopts, struct route_in6 *ro, struct ifnet **retifp) 708 { 709 int error; 710 struct route_in6 sro; 711 struct rtentry *rt = NULL; 712 713 if (ro == NULL) { 714 bzero(&sro, sizeof(sro)); 715 ro = &sro; 716 } 717 718 if ((error = selectroute(dstsock, opts, mopts, ro, retifp, 719 &rt, 1)) != 0) { 720 if (ro == &sro && rt && rt == sro.ro_rt) 721 RTFREE(rt); 722 return (error); 723 } 724 725 /* 726 * do not use a rejected or black hole route. 727 * XXX: this check should be done in the L2 output routine. 728 * However, if we skipped this check here, we'd see the following 729 * scenario: 730 * - install a rejected route for a scoped address prefix 731 * (like fe80::/10) 732 * - send a packet to a destination that matches the scoped prefix, 733 * with ambiguity about the scope zone. 734 * - pick the outgoing interface from the route, and disambiguate the 735 * scope zone with the interface. 736 * - ip6_output() would try to get another route with the "new" 737 * destination, which may be valid. 738 * - we'd see no error on output. 739 * Although this may not be very harmful, it should still be confusing. 740 * We thus reject the case here. 741 */ 742 if (rt && (rt->rt_flags & (RTF_REJECT | RTF_BLACKHOLE))) { 743 int flags = (rt->rt_flags & RTF_HOST ? EHOSTUNREACH : ENETUNREACH); 744 745 if (ro == &sro && rt && rt == sro.ro_rt) 746 RTFREE(rt); 747 return (flags); 748 } 749 750 /* 751 * Adjust the "outgoing" interface. If we're going to loop the packet 752 * back to ourselves, the ifp would be the loopback interface. 753 * However, we'd rather know the interface associated to the 754 * destination address (which should probably be one of our own 755 * addresses.) 756 */ 757 if (rt && rt->rt_ifa && rt->rt_ifa->ifa_ifp) 758 *retifp = rt->rt_ifa->ifa_ifp; 759 760 if (ro == &sro && rt && rt == sro.ro_rt) 761 RTFREE(rt); 762 return (0); 763 } 764 765 /* 766 * clone - meaningful only for bsdi and freebsd 767 */ 768 int 769 in6_selectroute(struct sockaddr_in6 *dstsock, struct ip6_pktopts *opts, 770 struct ip6_moptions *mopts, struct route_in6 *ro, 771 struct ifnet **retifp, struct rtentry **retrt) 772 { 773 774 return (selectroute(dstsock, opts, mopts, ro, retifp, 775 retrt, 0)); 776 } 777 778 /* 779 * Default hop limit selection. The precedence is as follows: 780 * 1. Hoplimit value specified via ioctl. 781 * 2. (If the outgoing interface is detected) the current 782 * hop limit of the interface specified by router advertisement. 783 * 3. The system default hoplimit. 784 */ 785 int 786 in6_selecthlim(struct inpcb *in6p, struct ifnet *ifp) 787 { 788 INIT_VNET_INET6(curvnet); 789 790 if (in6p && in6p->in6p_hops >= 0) 791 return (in6p->in6p_hops); 792 else if (ifp) 793 return (ND_IFINFO(ifp)->chlim); 794 else if (in6p && !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr)) { 795 struct route_in6 ro6; 796 struct ifnet *lifp; 797 798 bzero(&ro6, sizeof(ro6)); 799 ro6.ro_dst.sin6_family = AF_INET6; 800 ro6.ro_dst.sin6_len = sizeof(struct sockaddr_in6); 801 ro6.ro_dst.sin6_addr = in6p->in6p_faddr; 802 rtalloc((struct route *)&ro6); 803 if (ro6.ro_rt) { 804 lifp = ro6.ro_rt->rt_ifp; 805 RTFREE(ro6.ro_rt); 806 if (lifp) 807 return (ND_IFINFO(lifp)->chlim); 808 } else 809 return (V_ip6_defhlim); 810 } 811 return (V_ip6_defhlim); 812 } 813 814 /* 815 * XXX: this is borrowed from in6_pcbbind(). If possible, we should 816 * share this function by all *bsd*... 817 */ 818 int 819 in6_pcbsetport(struct in6_addr *laddr, struct inpcb *inp, struct ucred *cred) 820 { 821 INIT_VNET_INET(curvnet); 822 struct socket *so = inp->inp_socket; 823 u_int16_t lport = 0, first, last, *lastport; 824 int count, error, wild = 0, dorandom; 825 struct inpcbinfo *pcbinfo = inp->inp_pcbinfo; 826 827 INP_INFO_WLOCK_ASSERT(pcbinfo); 828 INP_WLOCK_ASSERT(inp); 829 830 error = prison_local_ip6(cred, laddr, 831 ((inp->inp_flags & IN6P_IPV6_V6ONLY) != 0)); 832 if (error) 833 return(error); 834 835 /* XXX: this is redundant when called from in6_pcbbind */ 836 if ((so->so_options & (SO_REUSEADDR|SO_REUSEPORT)) == 0) 837 wild = INPLOOKUP_WILDCARD; 838 839 inp->inp_flags |= INP_ANONPORT; 840 841 if (inp->inp_flags & INP_HIGHPORT) { 842 first = V_ipport_hifirstauto; /* sysctl */ 843 last = V_ipport_hilastauto; 844 lastport = &pcbinfo->ipi_lasthi; 845 } else if (inp->inp_flags & INP_LOWPORT) { 846 error = priv_check_cred(cred, PRIV_NETINET_RESERVEDPORT, 0); 847 if (error) 848 return error; 849 first = V_ipport_lowfirstauto; /* 1023 */ 850 last = V_ipport_lowlastauto; /* 600 */ 851 lastport = &pcbinfo->ipi_lastlow; 852 } else { 853 first = V_ipport_firstauto; /* sysctl */ 854 last = V_ipport_lastauto; 855 lastport = &pcbinfo->ipi_lastport; 856 } 857 858 /* 859 * For UDP, use random port allocation as long as the user 860 * allows it. For TCP (and as of yet unknown) connections, 861 * use random port allocation only if the user allows it AND 862 * ipport_tick() allows it. 863 */ 864 if (V_ipport_randomized && 865 (!V_ipport_stoprandom || pcbinfo == &V_udbinfo)) 866 dorandom = 1; 867 else 868 dorandom = 0; 869 /* 870 * It makes no sense to do random port allocation if 871 * we have the only port available. 872 */ 873 if (first == last) 874 dorandom = 0; 875 /* Make sure to not include UDP packets in the count. */ 876 if (pcbinfo != &V_udbinfo) 877 V_ipport_tcpallocs++; 878 879 /* 880 * Instead of having two loops further down counting up or down 881 * make sure that first is always <= last and go with only one 882 * code path implementing all logic. 883 */ 884 if (first > last) { 885 u_int16_t aux; 886 887 aux = first; 888 first = last; 889 last = aux; 890 } 891 892 if (dorandom) 893 *lastport = first + (arc4random() % (last - first)); 894 895 count = last - first; 896 897 do { 898 if (count-- < 0) { /* completely used? */ 899 /* Undo an address bind that may have occurred. */ 900 inp->in6p_laddr = in6addr_any; 901 return (EADDRNOTAVAIL); 902 } 903 ++*lastport; 904 if (*lastport < first || *lastport > last) 905 *lastport = first; 906 lport = htons(*lastport); 907 } while (in6_pcblookup_local(pcbinfo, &inp->in6p_laddr, 908 lport, wild, cred)); 909 910 inp->inp_lport = lport; 911 if (in_pcbinshash(inp) != 0) { 912 inp->in6p_laddr = in6addr_any; 913 inp->inp_lport = 0; 914 return (EAGAIN); 915 } 916 917 return (0); 918 } 919 920 void 921 addrsel_policy_init(void) 922 { 923 ADDRSEL_LOCK_INIT(); 924 ADDRSEL_SXLOCK_INIT(); 925 INIT_VNET_INET6(curvnet); 926 927 V_ip6_prefer_tempaddr = 0; 928 929 init_policy_queue(); 930 931 /* initialize the "last resort" policy */ 932 bzero(&V_defaultaddrpolicy, sizeof(V_defaultaddrpolicy)); 933 V_defaultaddrpolicy.label = ADDR_LABEL_NOTAPP; 934 } 935 936 static struct in6_addrpolicy * 937 lookup_addrsel_policy(struct sockaddr_in6 *key) 938 { 939 INIT_VNET_INET6(curvnet); 940 struct in6_addrpolicy *match = NULL; 941 942 ADDRSEL_LOCK(); 943 match = match_addrsel_policy(key); 944 945 if (match == NULL) 946 match = &V_defaultaddrpolicy; 947 else 948 match->use++; 949 ADDRSEL_UNLOCK(); 950 951 return (match); 952 } 953 954 /* 955 * Subroutines to manage the address selection policy table via sysctl. 956 */ 957 struct walkarg { 958 struct sysctl_req *w_req; 959 }; 960 961 static int in6_src_sysctl(SYSCTL_HANDLER_ARGS); 962 SYSCTL_DECL(_net_inet6_ip6); 963 SYSCTL_NODE(_net_inet6_ip6, IPV6CTL_ADDRCTLPOLICY, addrctlpolicy, 964 CTLFLAG_RD, in6_src_sysctl, ""); 965 966 static int 967 in6_src_sysctl(SYSCTL_HANDLER_ARGS) 968 { 969 struct walkarg w; 970 971 if (req->newptr) 972 return EPERM; 973 974 bzero(&w, sizeof(w)); 975 w.w_req = req; 976 977 return (walk_addrsel_policy(dump_addrsel_policyent, &w)); 978 } 979 980 int 981 in6_src_ioctl(u_long cmd, caddr_t data) 982 { 983 int i; 984 struct in6_addrpolicy ent0; 985 986 if (cmd != SIOCAADDRCTL_POLICY && cmd != SIOCDADDRCTL_POLICY) 987 return (EOPNOTSUPP); /* check for safety */ 988 989 ent0 = *(struct in6_addrpolicy *)data; 990 991 if (ent0.label == ADDR_LABEL_NOTAPP) 992 return (EINVAL); 993 /* check if the prefix mask is consecutive. */ 994 if (in6_mask2len(&ent0.addrmask.sin6_addr, NULL) < 0) 995 return (EINVAL); 996 /* clear trailing garbages (if any) of the prefix address. */ 997 for (i = 0; i < 4; i++) { 998 ent0.addr.sin6_addr.s6_addr32[i] &= 999 ent0.addrmask.sin6_addr.s6_addr32[i]; 1000 } 1001 ent0.use = 0; 1002 1003 switch (cmd) { 1004 case SIOCAADDRCTL_POLICY: 1005 return (add_addrsel_policyent(&ent0)); 1006 case SIOCDADDRCTL_POLICY: 1007 return (delete_addrsel_policyent(&ent0)); 1008 } 1009 1010 return (0); /* XXX: compromise compilers */ 1011 } 1012 1013 /* 1014 * The followings are implementation of the policy table using a 1015 * simple tail queue. 1016 * XXX such details should be hidden. 1017 * XXX implementation using binary tree should be more efficient. 1018 */ 1019 struct addrsel_policyent { 1020 TAILQ_ENTRY(addrsel_policyent) ape_entry; 1021 struct in6_addrpolicy ape_policy; 1022 }; 1023 1024 TAILQ_HEAD(addrsel_policyhead, addrsel_policyent); 1025 1026 #ifdef VIMAGE_GLOBALS 1027 struct addrsel_policyhead addrsel_policytab; 1028 #endif 1029 1030 static void 1031 init_policy_queue(void) 1032 { 1033 INIT_VNET_INET6(curvnet); 1034 1035 TAILQ_INIT(&V_addrsel_policytab); 1036 } 1037 1038 static int 1039 add_addrsel_policyent(struct in6_addrpolicy *newpolicy) 1040 { 1041 INIT_VNET_INET6(curvnet); 1042 struct addrsel_policyent *new, *pol; 1043 1044 new = malloc(sizeof(*new), M_IFADDR, 1045 M_WAITOK); 1046 ADDRSEL_XLOCK(); 1047 ADDRSEL_LOCK(); 1048 1049 /* duplication check */ 1050 TAILQ_FOREACH(pol, &V_addrsel_policytab, ape_entry) { 1051 if (IN6_ARE_ADDR_EQUAL(&newpolicy->addr.sin6_addr, 1052 &pol->ape_policy.addr.sin6_addr) && 1053 IN6_ARE_ADDR_EQUAL(&newpolicy->addrmask.sin6_addr, 1054 &pol->ape_policy.addrmask.sin6_addr)) { 1055 ADDRSEL_UNLOCK(); 1056 ADDRSEL_XUNLOCK(); 1057 free(new, M_IFADDR); 1058 return (EEXIST); /* or override it? */ 1059 } 1060 } 1061 1062 bzero(new, sizeof(*new)); 1063 1064 /* XXX: should validate entry */ 1065 new->ape_policy = *newpolicy; 1066 1067 TAILQ_INSERT_TAIL(&V_addrsel_policytab, new, ape_entry); 1068 ADDRSEL_UNLOCK(); 1069 ADDRSEL_XUNLOCK(); 1070 1071 return (0); 1072 } 1073 1074 static int 1075 delete_addrsel_policyent(struct in6_addrpolicy *key) 1076 { 1077 INIT_VNET_INET6(curvnet); 1078 struct addrsel_policyent *pol; 1079 1080 ADDRSEL_XLOCK(); 1081 ADDRSEL_LOCK(); 1082 1083 /* search for the entry in the table */ 1084 TAILQ_FOREACH(pol, &V_addrsel_policytab, ape_entry) { 1085 if (IN6_ARE_ADDR_EQUAL(&key->addr.sin6_addr, 1086 &pol->ape_policy.addr.sin6_addr) && 1087 IN6_ARE_ADDR_EQUAL(&key->addrmask.sin6_addr, 1088 &pol->ape_policy.addrmask.sin6_addr)) { 1089 break; 1090 } 1091 } 1092 if (pol == NULL) { 1093 ADDRSEL_UNLOCK(); 1094 ADDRSEL_XUNLOCK(); 1095 return (ESRCH); 1096 } 1097 1098 TAILQ_REMOVE(&V_addrsel_policytab, pol, ape_entry); 1099 ADDRSEL_UNLOCK(); 1100 ADDRSEL_XUNLOCK(); 1101 1102 return (0); 1103 } 1104 1105 static int 1106 walk_addrsel_policy(int (*callback)(struct in6_addrpolicy *, void *), 1107 void *w) 1108 { 1109 INIT_VNET_INET6(curvnet); 1110 struct addrsel_policyent *pol; 1111 int error = 0; 1112 1113 ADDRSEL_SLOCK(); 1114 TAILQ_FOREACH(pol, &V_addrsel_policytab, ape_entry) { 1115 if ((error = (*callback)(&pol->ape_policy, w)) != 0) { 1116 ADDRSEL_SUNLOCK(); 1117 return (error); 1118 } 1119 } 1120 ADDRSEL_SUNLOCK(); 1121 return (error); 1122 } 1123 1124 static int 1125 dump_addrsel_policyent(struct in6_addrpolicy *pol, void *arg) 1126 { 1127 int error = 0; 1128 struct walkarg *w = arg; 1129 1130 error = SYSCTL_OUT(w->w_req, pol, sizeof(*pol)); 1131 1132 return (error); 1133 } 1134 1135 static struct in6_addrpolicy * 1136 match_addrsel_policy(struct sockaddr_in6 *key) 1137 { 1138 INIT_VNET_INET6(curvnet); 1139 struct addrsel_policyent *pent; 1140 struct in6_addrpolicy *bestpol = NULL, *pol; 1141 int matchlen, bestmatchlen = -1; 1142 u_char *mp, *ep, *k, *p, m; 1143 1144 TAILQ_FOREACH(pent, &V_addrsel_policytab, ape_entry) { 1145 matchlen = 0; 1146 1147 pol = &pent->ape_policy; 1148 mp = (u_char *)&pol->addrmask.sin6_addr; 1149 ep = mp + 16; /* XXX: scope field? */ 1150 k = (u_char *)&key->sin6_addr; 1151 p = (u_char *)&pol->addr.sin6_addr; 1152 for (; mp < ep && *mp; mp++, k++, p++) { 1153 m = *mp; 1154 if ((*k & m) != *p) 1155 goto next; /* not match */ 1156 if (m == 0xff) /* short cut for a typical case */ 1157 matchlen += 8; 1158 else { 1159 while (m >= 0x80) { 1160 matchlen++; 1161 m <<= 1; 1162 } 1163 } 1164 } 1165 1166 /* matched. check if this is better than the current best. */ 1167 if (bestpol == NULL || 1168 matchlen > bestmatchlen) { 1169 bestpol = pol; 1170 bestmatchlen = matchlen; 1171 } 1172 1173 next: 1174 continue; 1175 } 1176 1177 return (bestpol); 1178 } 1179