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) 125 126 static VNET_DEFINE(struct in6_addrpolicy, defaultaddrpolicy); 127 VNET_DEFINE(int, ip6_prefer_tempaddr); 128 129 #define V_defaultaddrpolicy VNET(defaultaddrpolicy) 130 131 static int selectroute __P((struct sockaddr_in6 *, struct ip6_pktopts *, 132 struct ip6_moptions *, struct route_in6 *, struct ifnet **, 133 struct rtentry **, int)); 134 static int in6_selectif __P((struct sockaddr_in6 *, struct ip6_pktopts *, 135 struct ip6_moptions *, struct route_in6 *ro, struct ifnet **)); 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 __P((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 if ((r) < sizeof(V_ip6stat.ip6s_sources_rule) / \ 155 sizeof(V_ip6stat.ip6s_sources_rule[0])) /* check for safety */ \ 156 V_ip6stat.ip6s_sources_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 if ((r) < sizeof(V_ip6stat.ip6s_sources_rule) / \ 165 sizeof(V_ip6stat.ip6s_sources_rule[0])) /* check for safety */ \ 166 V_ip6stat.ip6s_sources_rule[(r)]++; \ 167 /* { \ 168 char ip6buf[INET6_ADDRSTRLEN], ip6b[INET6_ADDRSTRLEN]; \ 169 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)); \ 170 } */ \ 171 goto next; /* XXX: we can't use 'continue' here */ \ 172 } while(0) 173 #define BREAK(r) do { \ 174 if ((r) < sizeof(V_ip6stat.ip6s_sources_rule) / \ 175 sizeof(V_ip6stat.ip6s_sources_rule[0])) /* check for safety */ \ 176 V_ip6stat.ip6s_sources_rule[(r)]++; \ 177 goto out; /* XXX: we can't use 'break' here */ \ 178 } while(0) 179 180 int 181 in6_selectsrc(struct sockaddr_in6 *dstsock, struct ip6_pktopts *opts, 182 struct inpcb *inp, struct route_in6 *ro, struct ucred *cred, 183 struct ifnet **ifpp, struct in6_addr *srcp) 184 { 185 struct in6_addr dst; 186 struct ifnet *ifp = NULL; 187 struct in6_ifaddr *ia = NULL, *ia_best = NULL; 188 struct in6_pktinfo *pi = NULL; 189 int dst_scope = -1, best_scope = -1, best_matchlen = -1; 190 struct in6_addrpolicy *dst_policy = NULL, *best_policy = NULL; 191 u_int32_t odstzone; 192 int prefer_tempaddr; 193 int error; 194 struct ip6_moptions *mopts; 195 196 KASSERT(srcp != NULL, ("%s: srcp is NULL", __func__)); 197 198 dst = dstsock->sin6_addr; /* make a copy for local operation */ 199 if (ifpp) 200 *ifpp = NULL; 201 202 if (inp != NULL) { 203 INP_LOCK_ASSERT(inp); 204 mopts = inp->in6p_moptions; 205 } else { 206 mopts = NULL; 207 } 208 209 /* 210 * If the source address is explicitly specified by the caller, 211 * check if the requested source address is indeed a unicast address 212 * assigned to the node, and can be used as the packet's source 213 * address. If everything is okay, use the address as source. 214 */ 215 if (opts && (pi = opts->ip6po_pktinfo) && 216 !IN6_IS_ADDR_UNSPECIFIED(&pi->ipi6_addr)) { 217 struct sockaddr_in6 srcsock; 218 struct in6_ifaddr *ia6; 219 220 /* get the outgoing interface */ 221 if ((error = in6_selectif(dstsock, opts, mopts, ro, &ifp)) != 0) 222 return (error); 223 224 /* 225 * determine the appropriate zone id of the source based on 226 * the zone of the destination and the outgoing interface. 227 * If the specified address is ambiguous wrt the scope zone, 228 * the interface must be specified; otherwise, ifa_ifwithaddr() 229 * will fail matching the address. 230 */ 231 bzero(&srcsock, sizeof(srcsock)); 232 srcsock.sin6_family = AF_INET6; 233 srcsock.sin6_len = sizeof(srcsock); 234 srcsock.sin6_addr = pi->ipi6_addr; 235 if (ifp) { 236 error = in6_setscope(&srcsock.sin6_addr, ifp, NULL); 237 if (error) 238 return (error); 239 } 240 if (cred != NULL && (error = prison_local_ip6(cred, 241 &srcsock.sin6_addr, (inp != NULL && 242 (inp->inp_flags & IN6P_IPV6_V6ONLY) != 0))) != 0) 243 return (error); 244 245 ia6 = (struct in6_ifaddr *)ifa_ifwithaddr( 246 (struct sockaddr *)&srcsock); 247 if (ia6 == NULL || 248 (ia6->ia6_flags & (IN6_IFF_ANYCAST | IN6_IFF_NOTREADY))) { 249 if (ia6 != NULL) 250 ifa_free(&ia6->ia_ifa); 251 return (EADDRNOTAVAIL); 252 } 253 pi->ipi6_addr = srcsock.sin6_addr; /* XXX: this overrides pi */ 254 if (ifpp) 255 *ifpp = ifp; 256 bcopy(&ia6->ia_addr.sin6_addr, srcp, sizeof(*srcp)); 257 ifa_free(&ia6->ia_ifa); 258 return (0); 259 } 260 261 /* 262 * Otherwise, if the socket has already bound the source, just use it. 263 */ 264 if (inp != NULL && !IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) { 265 if (cred != NULL && 266 (error = prison_local_ip6(cred, &inp->in6p_laddr, 267 ((inp->inp_flags & IN6P_IPV6_V6ONLY) != 0))) != 0) 268 return (error); 269 bcopy(&inp->in6p_laddr, srcp, sizeof(*srcp)); 270 return (0); 271 } 272 273 /* 274 * If the address is not specified, choose the best one based on 275 * the outgoing interface and the destination address. 276 */ 277 /* get the outgoing interface */ 278 if ((error = in6_selectif(dstsock, opts, mopts, ro, &ifp)) != 0) 279 return (error); 280 281 #ifdef DIAGNOSTIC 282 if (ifp == NULL) /* this should not happen */ 283 panic("in6_selectsrc: NULL ifp"); 284 #endif 285 error = in6_setscope(&dst, ifp, &odstzone); 286 if (error) 287 return (error); 288 289 IN6_IFADDR_RLOCK(); 290 TAILQ_FOREACH(ia, &V_in6_ifaddrhead, ia_link) { 291 int new_scope = -1, new_matchlen = -1; 292 struct in6_addrpolicy *new_policy = NULL; 293 u_int32_t srczone, osrczone, dstzone; 294 struct in6_addr src; 295 struct ifnet *ifp1 = ia->ia_ifp; 296 297 /* 298 * We'll never take an address that breaks the scope zone 299 * of the destination. We also skip an address if its zone 300 * does not contain the outgoing interface. 301 * XXX: we should probably use sin6_scope_id here. 302 */ 303 if (in6_setscope(&dst, ifp1, &dstzone) || 304 odstzone != dstzone) { 305 continue; 306 } 307 src = ia->ia_addr.sin6_addr; 308 if (in6_setscope(&src, ifp, &osrczone) || 309 in6_setscope(&src, ifp1, &srczone) || 310 osrczone != srczone) { 311 continue; 312 } 313 314 /* avoid unusable addresses */ 315 if ((ia->ia6_flags & 316 (IN6_IFF_NOTREADY | IN6_IFF_ANYCAST | IN6_IFF_DETACHED))) { 317 continue; 318 } 319 if (!V_ip6_use_deprecated && IFA6_IS_DEPRECATED(ia)) 320 continue; 321 322 if (cred != NULL && 323 prison_local_ip6(cred, &ia->ia_addr.sin6_addr, 324 (inp != NULL && 325 (inp->inp_flags & IN6P_IPV6_V6ONLY) != 0)) != 0) 326 continue; 327 328 /* Rule 1: Prefer same address */ 329 if (IN6_ARE_ADDR_EQUAL(&dst, &ia->ia_addr.sin6_addr)) { 330 ia_best = ia; 331 BREAK(1); /* there should be no better candidate */ 332 } 333 334 if (ia_best == NULL) 335 REPLACE(0); 336 337 /* Rule 2: Prefer appropriate scope */ 338 if (dst_scope < 0) 339 dst_scope = in6_addrscope(&dst); 340 new_scope = in6_addrscope(&ia->ia_addr.sin6_addr); 341 if (IN6_ARE_SCOPE_CMP(best_scope, new_scope) < 0) { 342 if (IN6_ARE_SCOPE_CMP(best_scope, dst_scope) < 0) 343 REPLACE(2); 344 NEXT(2); 345 } else if (IN6_ARE_SCOPE_CMP(new_scope, best_scope) < 0) { 346 if (IN6_ARE_SCOPE_CMP(new_scope, dst_scope) < 0) 347 NEXT(2); 348 REPLACE(2); 349 } 350 351 /* 352 * Rule 3: Avoid deprecated addresses. Note that the case of 353 * !ip6_use_deprecated is already rejected above. 354 */ 355 if (!IFA6_IS_DEPRECATED(ia_best) && IFA6_IS_DEPRECATED(ia)) 356 NEXT(3); 357 if (IFA6_IS_DEPRECATED(ia_best) && !IFA6_IS_DEPRECATED(ia)) 358 REPLACE(3); 359 360 /* Rule 4: Prefer home addresses */ 361 /* 362 * XXX: This is a TODO. We should probably merge the MIP6 363 * case above. 364 */ 365 366 /* Rule 5: Prefer outgoing interface */ 367 if (ia_best->ia_ifp == ifp && ia->ia_ifp != ifp) 368 NEXT(5); 369 if (ia_best->ia_ifp != ifp && ia->ia_ifp == ifp) 370 REPLACE(5); 371 372 /* 373 * Rule 6: Prefer matching label 374 * Note that best_policy should be non-NULL here. 375 */ 376 if (dst_policy == NULL) 377 dst_policy = lookup_addrsel_policy(dstsock); 378 if (dst_policy->label != ADDR_LABEL_NOTAPP) { 379 new_policy = lookup_addrsel_policy(&ia->ia_addr); 380 if (dst_policy->label == best_policy->label && 381 dst_policy->label != new_policy->label) 382 NEXT(6); 383 if (dst_policy->label != best_policy->label && 384 dst_policy->label == new_policy->label) 385 REPLACE(6); 386 } 387 388 /* 389 * Rule 7: Prefer public addresses. 390 * We allow users to reverse the logic by configuring 391 * a sysctl variable, so that privacy conscious users can 392 * always prefer temporary addresses. 393 */ 394 if (opts == NULL || 395 opts->ip6po_prefer_tempaddr == IP6PO_TEMPADDR_SYSTEM) { 396 prefer_tempaddr = V_ip6_prefer_tempaddr; 397 } else if (opts->ip6po_prefer_tempaddr == 398 IP6PO_TEMPADDR_NOTPREFER) { 399 prefer_tempaddr = 0; 400 } else 401 prefer_tempaddr = 1; 402 if (!(ia_best->ia6_flags & IN6_IFF_TEMPORARY) && 403 (ia->ia6_flags & IN6_IFF_TEMPORARY)) { 404 if (prefer_tempaddr) 405 REPLACE(7); 406 else 407 NEXT(7); 408 } 409 if ((ia_best->ia6_flags & IN6_IFF_TEMPORARY) && 410 !(ia->ia6_flags & IN6_IFF_TEMPORARY)) { 411 if (prefer_tempaddr) 412 NEXT(7); 413 else 414 REPLACE(7); 415 } 416 417 /* 418 * Rule 8: prefer addresses on alive interfaces. 419 * This is a KAME specific rule. 420 */ 421 if ((ia_best->ia_ifp->if_flags & IFF_UP) && 422 !(ia->ia_ifp->if_flags & IFF_UP)) 423 NEXT(8); 424 if (!(ia_best->ia_ifp->if_flags & IFF_UP) && 425 (ia->ia_ifp->if_flags & IFF_UP)) 426 REPLACE(8); 427 428 /* 429 * Rule 14: Use longest matching prefix. 430 * Note: in the address selection draft, this rule is 431 * documented as "Rule 8". However, since it is also 432 * documented that this rule can be overridden, we assign 433 * a large number so that it is easy to assign smaller numbers 434 * to more preferred rules. 435 */ 436 new_matchlen = in6_matchlen(&ia->ia_addr.sin6_addr, &dst); 437 if (best_matchlen < new_matchlen) 438 REPLACE(14); 439 if (new_matchlen < best_matchlen) 440 NEXT(14); 441 442 /* Rule 15 is reserved. */ 443 444 /* 445 * Last resort: just keep the current candidate. 446 * Or, do we need more rules? 447 */ 448 continue; 449 450 replace: 451 ia_best = ia; 452 best_scope = (new_scope >= 0 ? new_scope : 453 in6_addrscope(&ia_best->ia_addr.sin6_addr)); 454 best_policy = (new_policy ? new_policy : 455 lookup_addrsel_policy(&ia_best->ia_addr)); 456 best_matchlen = (new_matchlen >= 0 ? new_matchlen : 457 in6_matchlen(&ia_best->ia_addr.sin6_addr, 458 &dst)); 459 460 next: 461 continue; 462 463 out: 464 break; 465 } 466 467 if ((ia = ia_best) == NULL) { 468 IN6_IFADDR_RUNLOCK(); 469 return (EADDRNOTAVAIL); 470 } 471 472 if (ifpp) 473 *ifpp = ifp; 474 475 bcopy(&ia->ia_addr.sin6_addr, srcp, sizeof(*srcp)); 476 IN6_IFADDR_RUNLOCK(); 477 return (0); 478 } 479 480 /* 481 * clone - meaningful only for bsdi and freebsd 482 */ 483 static int 484 selectroute(struct sockaddr_in6 *dstsock, struct ip6_pktopts *opts, 485 struct ip6_moptions *mopts, struct route_in6 *ro, 486 struct ifnet **retifp, struct rtentry **retrt, int norouteok) 487 { 488 int error = 0; 489 struct ifnet *ifp = NULL; 490 struct rtentry *rt = NULL; 491 struct sockaddr_in6 *sin6_next; 492 struct in6_pktinfo *pi = NULL; 493 struct in6_addr *dst = &dstsock->sin6_addr; 494 #if 0 495 char ip6buf[INET6_ADDRSTRLEN]; 496 497 if (dstsock->sin6_addr.s6_addr32[0] == 0 && 498 dstsock->sin6_addr.s6_addr32[1] == 0 && 499 !IN6_IS_ADDR_LOOPBACK(&dstsock->sin6_addr)) { 500 printf("in6_selectroute: strange destination %s\n", 501 ip6_sprintf(ip6buf, &dstsock->sin6_addr)); 502 } else { 503 printf("in6_selectroute: destination = %s%%%d\n", 504 ip6_sprintf(ip6buf, &dstsock->sin6_addr), 505 dstsock->sin6_scope_id); /* for debug */ 506 } 507 #endif 508 509 /* If the caller specify the outgoing interface explicitly, use it. */ 510 if (opts && (pi = opts->ip6po_pktinfo) != NULL && pi->ipi6_ifindex) { 511 /* XXX boundary check is assumed to be already done. */ 512 ifp = ifnet_byindex(pi->ipi6_ifindex); 513 if (ifp != NULL && 514 (norouteok || retrt == NULL || 515 IN6_IS_ADDR_MULTICAST(dst))) { 516 /* 517 * we do not have to check or get the route for 518 * multicast. 519 */ 520 goto done; 521 } else 522 goto getroute; 523 } 524 525 /* 526 * If the destination address is a multicast address and the outgoing 527 * interface for the address is specified by the caller, use it. 528 */ 529 if (IN6_IS_ADDR_MULTICAST(dst) && 530 mopts != NULL && (ifp = mopts->im6o_multicast_ifp) != NULL) { 531 goto done; /* we do not need a route for multicast. */ 532 } 533 534 getroute: 535 /* 536 * If the next hop address for the packet is specified by the caller, 537 * use it as the gateway. 538 */ 539 if (opts && opts->ip6po_nexthop) { 540 struct route_in6 *ron; 541 struct llentry *la; 542 543 sin6_next = satosin6(opts->ip6po_nexthop); 544 545 /* at this moment, we only support AF_INET6 next hops */ 546 if (sin6_next->sin6_family != AF_INET6) { 547 error = EAFNOSUPPORT; /* or should we proceed? */ 548 goto done; 549 } 550 551 /* 552 * If the next hop is an IPv6 address, then the node identified 553 * by that address must be a neighbor of the sending host. 554 */ 555 ron = &opts->ip6po_nextroute; 556 /* 557 * XXX what do we do here? 558 * PLZ to be fixing 559 */ 560 561 562 if (ron->ro_rt == NULL) { 563 rtalloc((struct route *)ron); /* multi path case? */ 564 if (ron->ro_rt == NULL) { 565 if (ron->ro_rt) { 566 RTFREE(ron->ro_rt); 567 ron->ro_rt = NULL; 568 } 569 error = EHOSTUNREACH; 570 goto done; 571 } 572 } 573 574 rt = ron->ro_rt; 575 ifp = rt->rt_ifp; 576 IF_AFDATA_LOCK(ifp); 577 la = lla_lookup(LLTABLE6(ifp), 0, (struct sockaddr *)&sin6_next->sin6_addr); 578 IF_AFDATA_UNLOCK(ifp); 579 if (la != NULL) 580 LLE_RUNLOCK(la); 581 else { 582 error = EHOSTUNREACH; 583 goto done; 584 } 585 #if 0 586 if ((ron->ro_rt && 587 (ron->ro_rt->rt_flags & (RTF_UP | RTF_LLINFO)) != 588 (RTF_UP | RTF_LLINFO)) || 589 !IN6_ARE_ADDR_EQUAL(&satosin6(&ron->ro_dst)->sin6_addr, 590 &sin6_next->sin6_addr)) { 591 if (ron->ro_rt) { 592 RTFREE(ron->ro_rt); 593 ron->ro_rt = NULL; 594 } 595 *satosin6(&ron->ro_dst) = *sin6_next; 596 } 597 if (ron->ro_rt == NULL) { 598 rtalloc((struct route *)ron); /* multi path case? */ 599 if (ron->ro_rt == NULL || 600 !(ron->ro_rt->rt_flags & RTF_LLINFO)) { 601 if (ron->ro_rt) { 602 RTFREE(ron->ro_rt); 603 ron->ro_rt = NULL; 604 } 605 error = EHOSTUNREACH; 606 goto done; 607 } 608 } 609 #endif 610 611 /* 612 * When cloning is required, try to allocate a route to the 613 * destination so that the caller can store path MTU 614 * information. 615 */ 616 goto done; 617 } 618 619 /* 620 * Use a cached route if it exists and is valid, else try to allocate 621 * a new one. Note that we should check the address family of the 622 * cached destination, in case of sharing the cache with IPv4. 623 */ 624 if (ro) { 625 if (ro->ro_rt && 626 (!(ro->ro_rt->rt_flags & RTF_UP) || 627 ((struct sockaddr *)(&ro->ro_dst))->sa_family != AF_INET6 || 628 !IN6_ARE_ADDR_EQUAL(&satosin6(&ro->ro_dst)->sin6_addr, 629 dst))) { 630 RTFREE(ro->ro_rt); 631 ro->ro_rt = (struct rtentry *)NULL; 632 } 633 if (ro->ro_rt == (struct rtentry *)NULL) { 634 struct sockaddr_in6 *sa6; 635 636 /* No route yet, so try to acquire one */ 637 bzero(&ro->ro_dst, sizeof(struct sockaddr_in6)); 638 sa6 = (struct sockaddr_in6 *)&ro->ro_dst; 639 *sa6 = *dstsock; 640 sa6->sin6_scope_id = 0; 641 642 #ifdef RADIX_MPATH 643 rtalloc_mpath((struct route *)ro, 644 ntohl(sa6->sin6_addr.s6_addr32[3])); 645 #else 646 ro->ro_rt = rtalloc1(&((struct route *)ro) 647 ->ro_dst, 0, 0UL); 648 if (ro->ro_rt) 649 RT_UNLOCK(ro->ro_rt); 650 #endif 651 } 652 653 /* 654 * do not care about the result if we have the nexthop 655 * explicitly specified. 656 */ 657 if (opts && opts->ip6po_nexthop) 658 goto done; 659 660 if (ro->ro_rt) { 661 ifp = ro->ro_rt->rt_ifp; 662 663 if (ifp == NULL) { /* can this really happen? */ 664 RTFREE(ro->ro_rt); 665 ro->ro_rt = NULL; 666 } 667 } 668 if (ro->ro_rt == NULL) 669 error = EHOSTUNREACH; 670 rt = ro->ro_rt; 671 672 /* 673 * Check if the outgoing interface conflicts with 674 * the interface specified by ipi6_ifindex (if specified). 675 * Note that loopback interface is always okay. 676 * (this may happen when we are sending a packet to one of 677 * our own addresses.) 678 */ 679 if (ifp && opts && opts->ip6po_pktinfo && 680 opts->ip6po_pktinfo->ipi6_ifindex) { 681 if (!(ifp->if_flags & IFF_LOOPBACK) && 682 ifp->if_index != 683 opts->ip6po_pktinfo->ipi6_ifindex) { 684 error = EHOSTUNREACH; 685 goto done; 686 } 687 } 688 } 689 690 done: 691 if (ifp == NULL && rt == NULL) { 692 /* 693 * This can happen if the caller did not pass a cached route 694 * nor any other hints. We treat this case an error. 695 */ 696 error = EHOSTUNREACH; 697 } 698 if (error == EHOSTUNREACH) 699 V_ip6stat.ip6s_noroute++; 700 701 if (retifp != NULL) { 702 *retifp = ifp; 703 704 /* 705 * Adjust the "outgoing" interface. If we're going to loop 706 * the packet back to ourselves, the ifp would be the loopback 707 * interface. However, we'd rather know the interface associated 708 * to the destination address (which should probably be one of 709 * our own addresses.) 710 */ 711 if (rt) { 712 if ((rt->rt_ifp->if_flags & IFF_LOOPBACK) && 713 (rt->rt_gateway->sa_family == AF_LINK)) 714 *retifp = 715 ifnet_byindex(((struct sockaddr_dl *) 716 rt->rt_gateway)->sdl_index); 717 } 718 } 719 720 if (retrt != NULL) 721 *retrt = rt; /* rt may be NULL */ 722 723 return (error); 724 } 725 726 static int 727 in6_selectif(struct sockaddr_in6 *dstsock, struct ip6_pktopts *opts, 728 struct ip6_moptions *mopts, struct route_in6 *ro, struct ifnet **retifp) 729 { 730 int error; 731 struct route_in6 sro; 732 struct rtentry *rt = NULL; 733 734 if (ro == NULL) { 735 bzero(&sro, sizeof(sro)); 736 ro = &sro; 737 } 738 739 if ((error = selectroute(dstsock, opts, mopts, ro, retifp, 740 &rt, 1)) != 0) { 741 if (ro == &sro && rt && rt == sro.ro_rt) 742 RTFREE(rt); 743 return (error); 744 } 745 746 /* 747 * do not use a rejected or black hole route. 748 * XXX: this check should be done in the L2 output routine. 749 * However, if we skipped this check here, we'd see the following 750 * scenario: 751 * - install a rejected route for a scoped address prefix 752 * (like fe80::/10) 753 * - send a packet to a destination that matches the scoped prefix, 754 * with ambiguity about the scope zone. 755 * - pick the outgoing interface from the route, and disambiguate the 756 * scope zone with the interface. 757 * - ip6_output() would try to get another route with the "new" 758 * destination, which may be valid. 759 * - we'd see no error on output. 760 * Although this may not be very harmful, it should still be confusing. 761 * We thus reject the case here. 762 */ 763 if (rt && (rt->rt_flags & (RTF_REJECT | RTF_BLACKHOLE))) { 764 int flags = (rt->rt_flags & RTF_HOST ? EHOSTUNREACH : ENETUNREACH); 765 766 if (ro == &sro && rt && rt == sro.ro_rt) 767 RTFREE(rt); 768 return (flags); 769 } 770 771 if (ro == &sro && rt && rt == sro.ro_rt) 772 RTFREE(rt); 773 return (0); 774 } 775 776 /* 777 * clone - meaningful only for bsdi and freebsd 778 */ 779 int 780 in6_selectroute(struct sockaddr_in6 *dstsock, struct ip6_pktopts *opts, 781 struct ip6_moptions *mopts, struct route_in6 *ro, 782 struct ifnet **retifp, struct rtentry **retrt) 783 { 784 785 return (selectroute(dstsock, opts, mopts, ro, retifp, 786 retrt, 0)); 787 } 788 789 /* 790 * Default hop limit selection. The precedence is as follows: 791 * 1. Hoplimit value specified via ioctl. 792 * 2. (If the outgoing interface is detected) the current 793 * hop limit of the interface specified by router advertisement. 794 * 3. The system default hoplimit. 795 */ 796 int 797 in6_selecthlim(struct inpcb *in6p, struct ifnet *ifp) 798 { 799 800 if (in6p && in6p->in6p_hops >= 0) 801 return (in6p->in6p_hops); 802 else if (ifp) 803 return (ND_IFINFO(ifp)->chlim); 804 else if (in6p && !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr)) { 805 struct route_in6 ro6; 806 struct ifnet *lifp; 807 808 bzero(&ro6, sizeof(ro6)); 809 ro6.ro_dst.sin6_family = AF_INET6; 810 ro6.ro_dst.sin6_len = sizeof(struct sockaddr_in6); 811 ro6.ro_dst.sin6_addr = in6p->in6p_faddr; 812 rtalloc((struct route *)&ro6); 813 if (ro6.ro_rt) { 814 lifp = ro6.ro_rt->rt_ifp; 815 RTFREE(ro6.ro_rt); 816 if (lifp) 817 return (ND_IFINFO(lifp)->chlim); 818 } else 819 return (V_ip6_defhlim); 820 } 821 return (V_ip6_defhlim); 822 } 823 824 /* 825 * XXX: this is borrowed from in6_pcbbind(). If possible, we should 826 * share this function by all *bsd*... 827 */ 828 int 829 in6_pcbsetport(struct in6_addr *laddr, struct inpcb *inp, struct ucred *cred) 830 { 831 struct socket *so = inp->inp_socket; 832 u_int16_t lport = 0, first, last, *lastport; 833 int count, error, wild = 0, dorandom; 834 struct inpcbinfo *pcbinfo = inp->inp_pcbinfo; 835 836 INP_INFO_WLOCK_ASSERT(pcbinfo); 837 INP_WLOCK_ASSERT(inp); 838 839 error = prison_local_ip6(cred, laddr, 840 ((inp->inp_flags & IN6P_IPV6_V6ONLY) != 0)); 841 if (error) 842 return(error); 843 844 /* XXX: this is redundant when called from in6_pcbbind */ 845 if ((so->so_options & (SO_REUSEADDR|SO_REUSEPORT)) == 0) 846 wild = INPLOOKUP_WILDCARD; 847 848 inp->inp_flags |= INP_ANONPORT; 849 850 if (inp->inp_flags & INP_HIGHPORT) { 851 first = V_ipport_hifirstauto; /* sysctl */ 852 last = V_ipport_hilastauto; 853 lastport = &pcbinfo->ipi_lasthi; 854 } else if (inp->inp_flags & INP_LOWPORT) { 855 error = priv_check_cred(cred, PRIV_NETINET_RESERVEDPORT, 0); 856 if (error) 857 return error; 858 first = V_ipport_lowfirstauto; /* 1023 */ 859 last = V_ipport_lowlastauto; /* 600 */ 860 lastport = &pcbinfo->ipi_lastlow; 861 } else { 862 first = V_ipport_firstauto; /* sysctl */ 863 last = V_ipport_lastauto; 864 lastport = &pcbinfo->ipi_lastport; 865 } 866 867 /* 868 * For UDP, use random port allocation as long as the user 869 * allows it. For TCP (and as of yet unknown) connections, 870 * use random port allocation only if the user allows it AND 871 * ipport_tick() allows it. 872 */ 873 if (V_ipport_randomized && 874 (!V_ipport_stoprandom || pcbinfo == &V_udbinfo)) 875 dorandom = 1; 876 else 877 dorandom = 0; 878 /* 879 * It makes no sense to do random port allocation if 880 * we have the only port available. 881 */ 882 if (first == last) 883 dorandom = 0; 884 /* Make sure to not include UDP packets in the count. */ 885 if (pcbinfo != &V_udbinfo) 886 V_ipport_tcpallocs++; 887 888 /* 889 * Instead of having two loops further down counting up or down 890 * make sure that first is always <= last and go with only one 891 * code path implementing all logic. 892 */ 893 if (first > last) { 894 u_int16_t aux; 895 896 aux = first; 897 first = last; 898 last = aux; 899 } 900 901 if (dorandom) 902 *lastport = first + (arc4random() % (last - first)); 903 904 count = last - first; 905 906 do { 907 if (count-- < 0) { /* completely used? */ 908 /* Undo an address bind that may have occurred. */ 909 inp->in6p_laddr = in6addr_any; 910 return (EADDRNOTAVAIL); 911 } 912 ++*lastport; 913 if (*lastport < first || *lastport > last) 914 *lastport = first; 915 lport = htons(*lastport); 916 } while (in6_pcblookup_local(pcbinfo, &inp->in6p_laddr, 917 lport, wild, cred)); 918 919 inp->inp_lport = lport; 920 if (in_pcbinshash(inp) != 0) { 921 inp->in6p_laddr = in6addr_any; 922 inp->inp_lport = 0; 923 return (EAGAIN); 924 } 925 926 return (0); 927 } 928 929 void 930 addrsel_policy_init(void) 931 { 932 933 V_ip6_prefer_tempaddr = 0; 934 935 init_policy_queue(); 936 937 /* initialize the "last resort" policy */ 938 bzero(&V_defaultaddrpolicy, sizeof(V_defaultaddrpolicy)); 939 V_defaultaddrpolicy.label = ADDR_LABEL_NOTAPP; 940 941 if (!IS_DEFAULT_VNET(curvnet)) 942 return; 943 944 ADDRSEL_LOCK_INIT(); 945 ADDRSEL_SXLOCK_INIT(); 946 } 947 948 static struct in6_addrpolicy * 949 lookup_addrsel_policy(struct sockaddr_in6 *key) 950 { 951 struct in6_addrpolicy *match = NULL; 952 953 ADDRSEL_LOCK(); 954 match = match_addrsel_policy(key); 955 956 if (match == NULL) 957 match = &V_defaultaddrpolicy; 958 else 959 match->use++; 960 ADDRSEL_UNLOCK(); 961 962 return (match); 963 } 964 965 /* 966 * Subroutines to manage the address selection policy table via sysctl. 967 */ 968 struct walkarg { 969 struct sysctl_req *w_req; 970 }; 971 972 static int in6_src_sysctl(SYSCTL_HANDLER_ARGS); 973 SYSCTL_DECL(_net_inet6_ip6); 974 SYSCTL_NODE(_net_inet6_ip6, IPV6CTL_ADDRCTLPOLICY, addrctlpolicy, 975 CTLFLAG_RD, in6_src_sysctl, ""); 976 977 static int 978 in6_src_sysctl(SYSCTL_HANDLER_ARGS) 979 { 980 struct walkarg w; 981 982 if (req->newptr) 983 return EPERM; 984 985 bzero(&w, sizeof(w)); 986 w.w_req = req; 987 988 return (walk_addrsel_policy(dump_addrsel_policyent, &w)); 989 } 990 991 int 992 in6_src_ioctl(u_long cmd, caddr_t data) 993 { 994 int i; 995 struct in6_addrpolicy ent0; 996 997 if (cmd != SIOCAADDRCTL_POLICY && cmd != SIOCDADDRCTL_POLICY) 998 return (EOPNOTSUPP); /* check for safety */ 999 1000 ent0 = *(struct in6_addrpolicy *)data; 1001 1002 if (ent0.label == ADDR_LABEL_NOTAPP) 1003 return (EINVAL); 1004 /* check if the prefix mask is consecutive. */ 1005 if (in6_mask2len(&ent0.addrmask.sin6_addr, NULL) < 0) 1006 return (EINVAL); 1007 /* clear trailing garbages (if any) of the prefix address. */ 1008 for (i = 0; i < 4; i++) { 1009 ent0.addr.sin6_addr.s6_addr32[i] &= 1010 ent0.addrmask.sin6_addr.s6_addr32[i]; 1011 } 1012 ent0.use = 0; 1013 1014 switch (cmd) { 1015 case SIOCAADDRCTL_POLICY: 1016 return (add_addrsel_policyent(&ent0)); 1017 case SIOCDADDRCTL_POLICY: 1018 return (delete_addrsel_policyent(&ent0)); 1019 } 1020 1021 return (0); /* XXX: compromise compilers */ 1022 } 1023 1024 /* 1025 * The followings are implementation of the policy table using a 1026 * simple tail queue. 1027 * XXX such details should be hidden. 1028 * XXX implementation using binary tree should be more efficient. 1029 */ 1030 struct addrsel_policyent { 1031 TAILQ_ENTRY(addrsel_policyent) ape_entry; 1032 struct in6_addrpolicy ape_policy; 1033 }; 1034 1035 TAILQ_HEAD(addrsel_policyhead, addrsel_policyent); 1036 1037 static VNET_DEFINE(struct addrsel_policyhead, addrsel_policytab); 1038 #define V_addrsel_policytab VNET(addrsel_policytab) 1039 1040 static void 1041 init_policy_queue(void) 1042 { 1043 1044 TAILQ_INIT(&V_addrsel_policytab); 1045 } 1046 1047 static int 1048 add_addrsel_policyent(struct in6_addrpolicy *newpolicy) 1049 { 1050 struct addrsel_policyent *new, *pol; 1051 1052 new = malloc(sizeof(*new), M_IFADDR, 1053 M_WAITOK); 1054 ADDRSEL_XLOCK(); 1055 ADDRSEL_LOCK(); 1056 1057 /* duplication check */ 1058 TAILQ_FOREACH(pol, &V_addrsel_policytab, ape_entry) { 1059 if (IN6_ARE_ADDR_EQUAL(&newpolicy->addr.sin6_addr, 1060 &pol->ape_policy.addr.sin6_addr) && 1061 IN6_ARE_ADDR_EQUAL(&newpolicy->addrmask.sin6_addr, 1062 &pol->ape_policy.addrmask.sin6_addr)) { 1063 ADDRSEL_UNLOCK(); 1064 ADDRSEL_XUNLOCK(); 1065 free(new, M_IFADDR); 1066 return (EEXIST); /* or override it? */ 1067 } 1068 } 1069 1070 bzero(new, sizeof(*new)); 1071 1072 /* XXX: should validate entry */ 1073 new->ape_policy = *newpolicy; 1074 1075 TAILQ_INSERT_TAIL(&V_addrsel_policytab, new, ape_entry); 1076 ADDRSEL_UNLOCK(); 1077 ADDRSEL_XUNLOCK(); 1078 1079 return (0); 1080 } 1081 1082 static int 1083 delete_addrsel_policyent(struct in6_addrpolicy *key) 1084 { 1085 struct addrsel_policyent *pol; 1086 1087 ADDRSEL_XLOCK(); 1088 ADDRSEL_LOCK(); 1089 1090 /* search for the entry in the table */ 1091 TAILQ_FOREACH(pol, &V_addrsel_policytab, ape_entry) { 1092 if (IN6_ARE_ADDR_EQUAL(&key->addr.sin6_addr, 1093 &pol->ape_policy.addr.sin6_addr) && 1094 IN6_ARE_ADDR_EQUAL(&key->addrmask.sin6_addr, 1095 &pol->ape_policy.addrmask.sin6_addr)) { 1096 break; 1097 } 1098 } 1099 if (pol == NULL) { 1100 ADDRSEL_UNLOCK(); 1101 ADDRSEL_XUNLOCK(); 1102 return (ESRCH); 1103 } 1104 1105 TAILQ_REMOVE(&V_addrsel_policytab, pol, ape_entry); 1106 ADDRSEL_UNLOCK(); 1107 ADDRSEL_XUNLOCK(); 1108 1109 return (0); 1110 } 1111 1112 static int 1113 walk_addrsel_policy(int (*callback)(struct in6_addrpolicy *, void *), 1114 void *w) 1115 { 1116 struct addrsel_policyent *pol; 1117 int error = 0; 1118 1119 ADDRSEL_SLOCK(); 1120 TAILQ_FOREACH(pol, &V_addrsel_policytab, ape_entry) { 1121 if ((error = (*callback)(&pol->ape_policy, w)) != 0) { 1122 ADDRSEL_SUNLOCK(); 1123 return (error); 1124 } 1125 } 1126 ADDRSEL_SUNLOCK(); 1127 return (error); 1128 } 1129 1130 static int 1131 dump_addrsel_policyent(struct in6_addrpolicy *pol, void *arg) 1132 { 1133 int error = 0; 1134 struct walkarg *w = arg; 1135 1136 error = SYSCTL_OUT(w->w_req, pol, sizeof(*pol)); 1137 1138 return (error); 1139 } 1140 1141 static struct in6_addrpolicy * 1142 match_addrsel_policy(struct sockaddr_in6 *key) 1143 { 1144 struct addrsel_policyent *pent; 1145 struct in6_addrpolicy *bestpol = NULL, *pol; 1146 int matchlen, bestmatchlen = -1; 1147 u_char *mp, *ep, *k, *p, m; 1148 1149 TAILQ_FOREACH(pent, &V_addrsel_policytab, ape_entry) { 1150 matchlen = 0; 1151 1152 pol = &pent->ape_policy; 1153 mp = (u_char *)&pol->addrmask.sin6_addr; 1154 ep = mp + 16; /* XXX: scope field? */ 1155 k = (u_char *)&key->sin6_addr; 1156 p = (u_char *)&pol->addr.sin6_addr; 1157 for (; mp < ep && *mp; mp++, k++, p++) { 1158 m = *mp; 1159 if ((*k & m) != *p) 1160 goto next; /* not match */ 1161 if (m == 0xff) /* short cut for a typical case */ 1162 matchlen += 8; 1163 else { 1164 while (m >= 0x80) { 1165 matchlen++; 1166 m <<= 1; 1167 } 1168 } 1169 } 1170 1171 /* matched. check if this is better than the current best. */ 1172 if (bestpol == NULL || 1173 matchlen > bestmatchlen) { 1174 bestpol = pol; 1175 bestmatchlen = matchlen; 1176 } 1177 1178 next: 1179 continue; 1180 } 1181 1182 return (bestpol); 1183 } 1184