in6_src.c revision 160981
1/* $FreeBSD: head/sys/netinet6/in6_src.c 160981 2006-08-04 21:27:40Z brooks $ */ 2/* $KAME: in6_src.c,v 1.132 2003/08/26 04:42:27 keiichi Exp $ */ 3 4/*- 5 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project. 6 * All rights reserved. 7 * 8 * Redistribution and use in source and binary forms, with or without 9 * modification, are permitted provided that the following conditions 10 * are met: 11 * 1. Redistributions of source code must retain the above copyright 12 * notice, this list of conditions and the following disclaimer. 13 * 2. Redistributions in binary form must reproduce the above copyright 14 * notice, this list of conditions and the following disclaimer in the 15 * documentation and/or other materials provided with the distribution. 16 * 3. Neither the name of the project nor the names of its contributors 17 * may be used to endorse or promote products derived from this software 18 * without specific prior written permission. 19 * 20 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND 21 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 22 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 23 * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE 24 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 25 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 26 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 27 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 28 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 29 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 30 * SUCH DAMAGE. 31 */ 32 33/*- 34 * Copyright (c) 1982, 1986, 1991, 1993 35 * The Regents of the University of California. All rights reserved. 36 * 37 * Redistribution and use in source and binary forms, with or without 38 * modification, are permitted provided that the following conditions 39 * are met: 40 * 1. Redistributions of source code must retain the above copyright 41 * notice, this list of conditions and the following disclaimer. 42 * 2. Redistributions in binary form must reproduce the above copyright 43 * notice, this list of conditions and the following disclaimer in the 44 * documentation and/or other materials provided with the distribution. 45 * 4. Neither the name of the University nor the names of its contributors 46 * may be used to endorse or promote products derived from this software 47 * without specific prior written permission. 48 * 49 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 50 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 51 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 52 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 53 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 54 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 55 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 56 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 57 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 58 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 59 * SUCH DAMAGE. 60 * 61 * @(#)in_pcb.c 8.2 (Berkeley) 1/4/94 62 */ 63 64#include "opt_inet.h" 65#include "opt_inet6.h" 66 67#include <sys/param.h> 68#include <sys/systm.h> 69#include <sys/malloc.h> 70#include <sys/mbuf.h> 71#include <sys/protosw.h> 72#include <sys/socket.h> 73#include <sys/socketvar.h> 74#include <sys/sockio.h> 75#include <sys/sysctl.h> 76#include <sys/errno.h> 77#include <sys/time.h> 78#include <sys/kernel.h> 79#include <sys/sx.h> 80 81#include <net/if.h> 82#include <net/route.h> 83 84#include <netinet/in.h> 85#include <netinet/in_var.h> 86#include <netinet/in_systm.h> 87#include <netinet/ip.h> 88#include <netinet/in_pcb.h> 89#include <netinet6/in6_var.h> 90#include <netinet/ip6.h> 91#include <netinet6/in6_pcb.h> 92#include <netinet6/ip6_var.h> 93#include <netinet6/scope6_var.h> 94#include <netinet6/nd6.h> 95 96static struct mtx addrsel_lock; 97#define ADDRSEL_LOCK_INIT() mtx_init(&addrsel_lock, "addrsel_lock", NULL, MTX_DEF) 98#define ADDRSEL_LOCK() mtx_lock(&addrsel_lock) 99#define ADDRSEL_UNLOCK() mtx_unlock(&addrsel_lock) 100#define ADDRSEL_LOCK_ASSERT() mtx_assert(&addrsel_lock, MA_OWNED) 101 102static struct sx addrsel_sxlock; 103#define ADDRSEL_SXLOCK_INIT() sx_init(&addrsel_sxlock, "addrsel_sxlock") 104#define ADDRSEL_SLOCK() sx_slock(&addrsel_sxlock) 105#define ADDRSEL_SUNLOCK() sx_sunlock(&addrsel_sxlock) 106#define ADDRSEL_XLOCK() sx_xlock(&addrsel_sxlock) 107#define ADDRSEL_XUNLOCK() sx_xunlock(&addrsel_sxlock) 108 109#define ADDR_LABEL_NOTAPP (-1) 110struct in6_addrpolicy defaultaddrpolicy; 111 112int ip6_prefer_tempaddr = 0; 113 114static int selectroute __P((struct sockaddr_in6 *, struct ip6_pktopts *, 115 struct ip6_moptions *, struct route_in6 *, struct ifnet **, 116 struct rtentry **, int, int)); 117static int in6_selectif __P((struct sockaddr_in6 *, struct ip6_pktopts *, 118 struct ip6_moptions *, struct route_in6 *ro, struct ifnet **)); 119 120static struct in6_addrpolicy *lookup_addrsel_policy __P((struct sockaddr_in6 *)); 121 122static void init_policy_queue __P((void)); 123static int add_addrsel_policyent __P((struct in6_addrpolicy *)); 124static int delete_addrsel_policyent __P((struct in6_addrpolicy *)); 125static int walk_addrsel_policy __P((int (*)(struct in6_addrpolicy *, void *), 126 void *)); 127static int dump_addrsel_policyent __P((struct in6_addrpolicy *, void *)); 128static struct in6_addrpolicy *match_addrsel_policy __P((struct sockaddr_in6 *)); 129 130/* 131 * Return an IPv6 address, which is the most appropriate for a given 132 * destination and user specified options. 133 * If necessary, this function lookups the routing table and returns 134 * an entry to the caller for later use. 135 */ 136#define REPLACE(r) do {\ 137 if ((r) < sizeof(ip6stat.ip6s_sources_rule) / \ 138 sizeof(ip6stat.ip6s_sources_rule[0])) /* check for safety */ \ 139 ip6stat.ip6s_sources_rule[(r)]++; \ 140 /* printf("in6_selectsrc: replace %s with %s by %d\n", ia_best ? ip6_sprintf(&ia_best->ia_addr.sin6_addr) : "none", ip6_sprintf(&ia->ia_addr.sin6_addr), (r)); */ \ 141 goto replace; \ 142} while(0) 143#define NEXT(r) do {\ 144 if ((r) < sizeof(ip6stat.ip6s_sources_rule) / \ 145 sizeof(ip6stat.ip6s_sources_rule[0])) /* check for safety */ \ 146 ip6stat.ip6s_sources_rule[(r)]++; \ 147 /* printf("in6_selectsrc: keep %s against %s by %d\n", ia_best ? ip6_sprintf(&ia_best->ia_addr.sin6_addr) : "none", ip6_sprintf(&ia->ia_addr.sin6_addr), (r)); */ \ 148 goto next; /* XXX: we can't use 'continue' here */ \ 149} while(0) 150#define BREAK(r) do { \ 151 if ((r) < sizeof(ip6stat.ip6s_sources_rule) / \ 152 sizeof(ip6stat.ip6s_sources_rule[0])) /* check for safety */ \ 153 ip6stat.ip6s_sources_rule[(r)]++; \ 154 goto out; /* XXX: we can't use 'break' here */ \ 155} while(0) 156 157struct in6_addr * 158in6_selectsrc(dstsock, opts, mopts, ro, laddr, ifpp, errorp) 159 struct sockaddr_in6 *dstsock; 160 struct ip6_pktopts *opts; 161 struct ip6_moptions *mopts; 162 struct route_in6 *ro; 163 struct in6_addr *laddr; 164 struct ifnet **ifpp; 165 int *errorp; 166{ 167 struct in6_addr dst; 168 struct ifnet *ifp = NULL; 169 struct in6_ifaddr *ia = NULL, *ia_best = NULL; 170 struct in6_pktinfo *pi = NULL; 171 int dst_scope = -1, best_scope = -1, best_matchlen = -1; 172 struct in6_addrpolicy *dst_policy = NULL, *best_policy = NULL; 173 u_int32_t odstzone; 174 int prefer_tempaddr; 175 176 dst = dstsock->sin6_addr; /* make a copy for local operation */ 177 *errorp = 0; 178 if (ifpp) 179 *ifpp = NULL; 180 181 /* 182 * If the source address is explicitly specified by the caller, 183 * check if the requested source address is indeed a unicast address 184 * assigned to the node, and can be used as the packet's source 185 * address. If everything is okay, use the address as source. 186 */ 187 if (opts && (pi = opts->ip6po_pktinfo) && 188 !IN6_IS_ADDR_UNSPECIFIED(&pi->ipi6_addr)) { 189 struct sockaddr_in6 srcsock; 190 struct in6_ifaddr *ia6; 191 192 /* get the outgoing interface */ 193 if ((*errorp = in6_selectif(dstsock, opts, mopts, ro, &ifp)) 194 != 0) { 195 return (NULL); 196 } 197 198 /* 199 * determine the appropriate zone id of the source based on 200 * the zone of the destination and the outgoing interface. 201 * If the specified address is ambiguous wrt the scope zone, 202 * the interface must be specified; otherwise, ifa_ifwithaddr() 203 * will fail matching the address. 204 */ 205 bzero(&srcsock, sizeof(srcsock)); 206 srcsock.sin6_family = AF_INET6; 207 srcsock.sin6_len = sizeof(srcsock); 208 srcsock.sin6_addr = pi->ipi6_addr; 209 if (ifp) { 210 *errorp = in6_setscope(&srcsock.sin6_addr, ifp, NULL); 211 if (*errorp != 0) 212 return (NULL); 213 } 214 215 ia6 = (struct in6_ifaddr *)ifa_ifwithaddr((struct sockaddr *)(&srcsock)); 216 if (ia6 == NULL || 217 (ia6->ia6_flags & (IN6_IFF_ANYCAST | IN6_IFF_NOTREADY))) { 218 *errorp = EADDRNOTAVAIL; 219 return (NULL); 220 } 221 pi->ipi6_addr = srcsock.sin6_addr; /* XXX: this overrides pi */ 222 if (ifpp) 223 *ifpp = ifp; 224 return (&ia6->ia_addr.sin6_addr); 225 } 226 227 /* 228 * Otherwise, if the socket has already bound the source, just use it. 229 */ 230 if (laddr && !IN6_IS_ADDR_UNSPECIFIED(laddr)) 231 return (laddr); 232 233 /* 234 * If the address is not specified, choose the best one based on 235 * the outgoing interface and the destination address. 236 */ 237 /* get the outgoing interface */ 238 if ((*errorp = in6_selectif(dstsock, opts, mopts, ro, &ifp)) != 0) 239 return (NULL); 240 241#ifdef DIAGNOSTIC 242 if (ifp == NULL) /* this should not happen */ 243 panic("in6_selectsrc: NULL ifp"); 244#endif 245 *errorp = in6_setscope(&dst, ifp, &odstzone); 246 if (*errorp != 0) 247 return (NULL); 248 249 for (ia = in6_ifaddr; ia; ia = ia->ia_next) { 250 int new_scope = -1, new_matchlen = -1; 251 struct in6_addrpolicy *new_policy = NULL; 252 u_int32_t srczone, osrczone, dstzone; 253 struct in6_addr src; 254 struct ifnet *ifp1 = ia->ia_ifp; 255 256 /* 257 * We'll never take an address that breaks the scope zone 258 * of the destination. We also skip an address if its zone 259 * does not contain the outgoing interface. 260 * XXX: we should probably use sin6_scope_id here. 261 */ 262 if (in6_setscope(&dst, ifp1, &dstzone) || 263 odstzone != dstzone) { 264 continue; 265 } 266 src = ia->ia_addr.sin6_addr; 267 if (in6_setscope(&src, ifp, &osrczone) || 268 in6_setscope(&src, ifp1, &srczone) || 269 osrczone != srczone) { 270 continue; 271 } 272 273 /* avoid unusable addresses */ 274 if ((ia->ia6_flags & 275 (IN6_IFF_NOTREADY | IN6_IFF_ANYCAST | IN6_IFF_DETACHED))) { 276 continue; 277 } 278 if (!ip6_use_deprecated && IFA6_IS_DEPRECATED(ia)) 279 continue; 280 281 /* Rule 1: Prefer same address */ 282 if (IN6_ARE_ADDR_EQUAL(&dst, &ia->ia_addr.sin6_addr)) { 283 ia_best = ia; 284 BREAK(1); /* there should be no better candidate */ 285 } 286 287 if (ia_best == NULL) 288 REPLACE(0); 289 290 /* Rule 2: Prefer appropriate scope */ 291 if (dst_scope < 0) 292 dst_scope = in6_addrscope(&dst); 293 new_scope = in6_addrscope(&ia->ia_addr.sin6_addr); 294 if (IN6_ARE_SCOPE_CMP(best_scope, new_scope) < 0) { 295 if (IN6_ARE_SCOPE_CMP(best_scope, dst_scope) < 0) 296 REPLACE(2); 297 NEXT(2); 298 } else if (IN6_ARE_SCOPE_CMP(new_scope, best_scope) < 0) { 299 if (IN6_ARE_SCOPE_CMP(new_scope, dst_scope) < 0) 300 NEXT(2); 301 REPLACE(2); 302 } 303 304 /* 305 * Rule 3: Avoid deprecated addresses. Note that the case of 306 * !ip6_use_deprecated is already rejected above. 307 */ 308 if (!IFA6_IS_DEPRECATED(ia_best) && IFA6_IS_DEPRECATED(ia)) 309 NEXT(3); 310 if (IFA6_IS_DEPRECATED(ia_best) && !IFA6_IS_DEPRECATED(ia)) 311 REPLACE(3); 312 313 /* Rule 4: Prefer home addresses */ 314 /* 315 * XXX: This is a TODO. We should probably merge the MIP6 316 * case above. 317 */ 318 319 /* Rule 5: Prefer outgoing interface */ 320 if (ia_best->ia_ifp == ifp && ia->ia_ifp != ifp) 321 NEXT(5); 322 if (ia_best->ia_ifp != ifp && ia->ia_ifp == ifp) 323 REPLACE(5); 324 325 /* 326 * Rule 6: Prefer matching label 327 * Note that best_policy should be non-NULL here. 328 */ 329 if (dst_policy == NULL) 330 dst_policy = lookup_addrsel_policy(dstsock); 331 if (dst_policy->label != ADDR_LABEL_NOTAPP) { 332 new_policy = lookup_addrsel_policy(&ia->ia_addr); 333 if (dst_policy->label == best_policy->label && 334 dst_policy->label != new_policy->label) 335 NEXT(6); 336 if (dst_policy->label != best_policy->label && 337 dst_policy->label == new_policy->label) 338 REPLACE(6); 339 } 340 341 /* 342 * Rule 7: Prefer public addresses. 343 * We allow users to reverse the logic by configuring 344 * a sysctl variable, so that privacy conscious users can 345 * always prefer temporary addresses. 346 */ 347 if (opts == NULL || 348 opts->ip6po_prefer_tempaddr == IP6PO_TEMPADDR_SYSTEM) { 349 prefer_tempaddr = ip6_prefer_tempaddr; 350 } else if (opts->ip6po_prefer_tempaddr == 351 IP6PO_TEMPADDR_NOTPREFER) { 352 prefer_tempaddr = 0; 353 } else 354 prefer_tempaddr = 1; 355 if (!(ia_best->ia6_flags & IN6_IFF_TEMPORARY) && 356 (ia->ia6_flags & IN6_IFF_TEMPORARY)) { 357 if (prefer_tempaddr) 358 REPLACE(7); 359 else 360 NEXT(7); 361 } 362 if ((ia_best->ia6_flags & IN6_IFF_TEMPORARY) && 363 !(ia->ia6_flags & IN6_IFF_TEMPORARY)) { 364 if (prefer_tempaddr) 365 NEXT(7); 366 else 367 REPLACE(7); 368 } 369 370 /* 371 * Rule 8: prefer addresses on alive interfaces. 372 * This is a KAME specific rule. 373 */ 374 if ((ia_best->ia_ifp->if_flags & IFF_UP) && 375 !(ia->ia_ifp->if_flags & IFF_UP)) 376 NEXT(8); 377 if (!(ia_best->ia_ifp->if_flags & IFF_UP) && 378 (ia->ia_ifp->if_flags & IFF_UP)) 379 REPLACE(8); 380 381 /* 382 * Rule 14: Use longest matching prefix. 383 * Note: in the address selection draft, this rule is 384 * documented as "Rule 8". However, since it is also 385 * documented that this rule can be overridden, we assign 386 * a large number so that it is easy to assign smaller numbers 387 * to more preferred rules. 388 */ 389 new_matchlen = in6_matchlen(&ia->ia_addr.sin6_addr, &dst); 390 if (best_matchlen < new_matchlen) 391 REPLACE(14); 392 if (new_matchlen < best_matchlen) 393 NEXT(14); 394 395 /* Rule 15 is reserved. */ 396 397 /* 398 * Last resort: just keep the current candidate. 399 * Or, do we need more rules? 400 */ 401 continue; 402 403 replace: 404 ia_best = ia; 405 best_scope = (new_scope >= 0 ? new_scope : 406 in6_addrscope(&ia_best->ia_addr.sin6_addr)); 407 best_policy = (new_policy ? new_policy : 408 lookup_addrsel_policy(&ia_best->ia_addr)); 409 best_matchlen = (new_matchlen >= 0 ? new_matchlen : 410 in6_matchlen(&ia_best->ia_addr.sin6_addr, 411 &dst)); 412 413 next: 414 continue; 415 416 out: 417 break; 418 } 419 420 if ((ia = ia_best) == NULL) { 421 *errorp = EADDRNOTAVAIL; 422 return (NULL); 423 } 424 425 if (ifpp) 426 *ifpp = ifp; 427 428 return (&ia->ia_addr.sin6_addr); 429} 430 431static int 432selectroute(dstsock, opts, mopts, ro, retifp, retrt, clone, norouteok) 433 struct sockaddr_in6 *dstsock; 434 struct ip6_pktopts *opts; 435 struct ip6_moptions *mopts; 436 struct route_in6 *ro; 437 struct ifnet **retifp; 438 struct rtentry **retrt; 439 int clone; /* meaningful only for bsdi and freebsd. */ 440 int norouteok; 441{ 442 int error = 0; 443 struct ifnet *ifp = NULL; 444 struct rtentry *rt = NULL; 445 struct sockaddr_in6 *sin6_next; 446 struct in6_pktinfo *pi = NULL; 447 struct in6_addr *dst = &dstsock->sin6_addr; 448 449#if 0 450 if (dstsock->sin6_addr.s6_addr32[0] == 0 && 451 dstsock->sin6_addr.s6_addr32[1] == 0 && 452 !IN6_IS_ADDR_LOOPBACK(&dstsock->sin6_addr)) { 453 printf("in6_selectroute: strange destination %s\n", 454 ip6_sprintf(&dstsock->sin6_addr)); 455 } else { 456 printf("in6_selectroute: destination = %s%%%d\n", 457 ip6_sprintf(&dstsock->sin6_addr), 458 dstsock->sin6_scope_id); /* for debug */ 459 } 460#endif 461 462 /* If the caller specify the outgoing interface explicitly, use it. */ 463 if (opts && (pi = opts->ip6po_pktinfo) != NULL && pi->ipi6_ifindex) { 464 /* XXX boundary check is assumed to be already done. */ 465 ifp = ifnet_byindex(pi->ipi6_ifindex); 466 if (ifp != NULL && 467 (norouteok || retrt == NULL || 468 IN6_IS_ADDR_MULTICAST(dst))) { 469 /* 470 * we do not have to check or get the route for 471 * multicast. 472 */ 473 goto done; 474 } else 475 goto getroute; 476 } 477 478 /* 479 * If the destination address is a multicast address and the outgoing 480 * interface for the address is specified by the caller, use it. 481 */ 482 if (IN6_IS_ADDR_MULTICAST(dst) && 483 mopts != NULL && (ifp = mopts->im6o_multicast_ifp) != NULL) { 484 goto done; /* we do not need a route for multicast. */ 485 } 486 487 getroute: 488 /* 489 * If the next hop address for the packet is specified by the caller, 490 * use it as the gateway. 491 */ 492 if (opts && opts->ip6po_nexthop) { 493 struct route_in6 *ron; 494 495 sin6_next = satosin6(opts->ip6po_nexthop); 496 497 /* at this moment, we only support AF_INET6 next hops */ 498 if (sin6_next->sin6_family != AF_INET6) { 499 error = EAFNOSUPPORT; /* or should we proceed? */ 500 goto done; 501 } 502 503 /* 504 * If the next hop is an IPv6 address, then the node identified 505 * by that address must be a neighbor of the sending host. 506 */ 507 ron = &opts->ip6po_nextroute; 508 if ((ron->ro_rt && 509 (ron->ro_rt->rt_flags & (RTF_UP | RTF_LLINFO)) != 510 (RTF_UP | RTF_LLINFO)) || 511 !IN6_ARE_ADDR_EQUAL(&satosin6(&ron->ro_dst)->sin6_addr, 512 &sin6_next->sin6_addr)) { 513 if (ron->ro_rt) { 514 RTFREE(ron->ro_rt); 515 ron->ro_rt = NULL; 516 } 517 *satosin6(&ron->ro_dst) = *sin6_next; 518 } 519 if (ron->ro_rt == NULL) { 520 rtalloc((struct route *)ron); /* multi path case? */ 521 if (ron->ro_rt == NULL || 522 !(ron->ro_rt->rt_flags & RTF_LLINFO)) { 523 if (ron->ro_rt) { 524 RTFREE(ron->ro_rt); 525 ron->ro_rt = NULL; 526 } 527 error = EHOSTUNREACH; 528 goto done; 529 } 530 } 531 rt = ron->ro_rt; 532 ifp = rt->rt_ifp; 533 534 /* 535 * When cloning is required, try to allocate a route to the 536 * destination so that the caller can store path MTU 537 * information. 538 */ 539 if (!clone) 540 goto done; 541 } 542 543 /* 544 * Use a cached route if it exists and is valid, else try to allocate 545 * a new one. Note that we should check the address family of the 546 * cached destination, in case of sharing the cache with IPv4. 547 */ 548 if (ro) { 549 if (ro->ro_rt && 550 (!(ro->ro_rt->rt_flags & RTF_UP) || 551 ((struct sockaddr *)(&ro->ro_dst))->sa_family != AF_INET6 || 552 !IN6_ARE_ADDR_EQUAL(&satosin6(&ro->ro_dst)->sin6_addr, 553 dst))) { 554 RTFREE(ro->ro_rt); 555 ro->ro_rt = (struct rtentry *)NULL; 556 } 557 if (ro->ro_rt == (struct rtentry *)NULL) { 558 struct sockaddr_in6 *sa6; 559 560 /* No route yet, so try to acquire one */ 561 bzero(&ro->ro_dst, sizeof(struct sockaddr_in6)); 562 sa6 = (struct sockaddr_in6 *)&ro->ro_dst; 563 *sa6 = *dstsock; 564 sa6->sin6_scope_id = 0; 565 566 if (clone) { 567 rtalloc((struct route *)ro); 568 } else { 569 ro->ro_rt = rtalloc1(&((struct route *)ro) 570 ->ro_dst, 0, 0UL); 571 if (ro->ro_rt) 572 RT_UNLOCK(ro->ro_rt); 573 } 574 } 575 576 /* 577 * do not care about the result if we have the nexthop 578 * explicitly specified. 579 */ 580 if (opts && opts->ip6po_nexthop) 581 goto done; 582 583 if (ro->ro_rt) { 584 ifp = ro->ro_rt->rt_ifp; 585 586 if (ifp == NULL) { /* can this really happen? */ 587 RTFREE(ro->ro_rt); 588 ro->ro_rt = NULL; 589 } 590 } 591 if (ro->ro_rt == NULL) 592 error = EHOSTUNREACH; 593 rt = ro->ro_rt; 594 595 /* 596 * Check if the outgoing interface conflicts with 597 * the interface specified by ipi6_ifindex (if specified). 598 * Note that loopback interface is always okay. 599 * (this may happen when we are sending a packet to one of 600 * our own addresses.) 601 */ 602 if (ifp && opts && opts->ip6po_pktinfo && 603 opts->ip6po_pktinfo->ipi6_ifindex) { 604 if (!(ifp->if_flags & IFF_LOOPBACK) && 605 ifp->if_index != 606 opts->ip6po_pktinfo->ipi6_ifindex) { 607 error = EHOSTUNREACH; 608 goto done; 609 } 610 } 611 } 612 613 done: 614 if (ifp == NULL && rt == NULL) { 615 /* 616 * This can happen if the caller did not pass a cached route 617 * nor any other hints. We treat this case an error. 618 */ 619 error = EHOSTUNREACH; 620 } 621 if (error == EHOSTUNREACH) 622 ip6stat.ip6s_noroute++; 623 624 if (retifp != NULL) 625 *retifp = ifp; 626 if (retrt != NULL) 627 *retrt = rt; /* rt may be NULL */ 628 629 return (error); 630} 631 632static int 633in6_selectif(dstsock, opts, mopts, ro, retifp) 634 struct sockaddr_in6 *dstsock; 635 struct ip6_pktopts *opts; 636 struct ip6_moptions *mopts; 637 struct route_in6 *ro; 638 struct ifnet **retifp; 639{ 640 int error; 641 struct route_in6 sro; 642 struct rtentry *rt = NULL; 643 644 if (ro == NULL) { 645 bzero(&sro, sizeof(sro)); 646 ro = &sro; 647 } 648 649 if ((error = selectroute(dstsock, opts, mopts, ro, retifp, 650 &rt, 0, 1)) != 0) { 651 if (ro == &sro && rt && rt == sro.ro_rt) 652 RTFREE(rt); 653 return (error); 654 } 655 656 /* 657 * do not use a rejected or black hole route. 658 * XXX: this check should be done in the L2 output routine. 659 * However, if we skipped this check here, we'd see the following 660 * scenario: 661 * - install a rejected route for a scoped address prefix 662 * (like fe80::/10) 663 * - send a packet to a destination that matches the scoped prefix, 664 * with ambiguity about the scope zone. 665 * - pick the outgoing interface from the route, and disambiguate the 666 * scope zone with the interface. 667 * - ip6_output() would try to get another route with the "new" 668 * destination, which may be valid. 669 * - we'd see no error on output. 670 * Although this may not be very harmful, it should still be confusing. 671 * We thus reject the case here. 672 */ 673 if (rt && (rt->rt_flags & (RTF_REJECT | RTF_BLACKHOLE))) { 674 int flags = (rt->rt_flags & RTF_HOST ? EHOSTUNREACH : ENETUNREACH); 675 676 if (ro == &sro && rt && rt == sro.ro_rt) 677 RTFREE(rt); 678 return (flags); 679 } 680 681 /* 682 * Adjust the "outgoing" interface. If we're going to loop the packet 683 * back to ourselves, the ifp would be the loopback interface. 684 * However, we'd rather know the interface associated to the 685 * destination address (which should probably be one of our own 686 * addresses.) 687 */ 688 if (rt && rt->rt_ifa && rt->rt_ifa->ifa_ifp) 689 *retifp = rt->rt_ifa->ifa_ifp; 690 691 if (ro == &sro && rt && rt == sro.ro_rt) 692 RTFREE(rt); 693 return (0); 694} 695 696int 697in6_selectroute(dstsock, opts, mopts, ro, retifp, retrt, clone) 698 struct sockaddr_in6 *dstsock; 699 struct ip6_pktopts *opts; 700 struct ip6_moptions *mopts; 701 struct route_in6 *ro; 702 struct ifnet **retifp; 703 struct rtentry **retrt; 704 int clone; /* meaningful only for bsdi and freebsd. */ 705{ 706 return (selectroute(dstsock, opts, mopts, ro, retifp, 707 retrt, clone, 0)); 708} 709 710/* 711 * Default hop limit selection. The precedence is as follows: 712 * 1. Hoplimit value specified via ioctl. 713 * 2. (If the outgoing interface is detected) the current 714 * hop limit of the interface specified by router advertisement. 715 * 3. The system default hoplimit. 716 */ 717int 718in6_selecthlim(in6p, ifp) 719 struct in6pcb *in6p; 720 struct ifnet *ifp; 721{ 722 if (in6p && in6p->in6p_hops >= 0) 723 return (in6p->in6p_hops); 724 else if (ifp) 725 return (ND_IFINFO(ifp)->chlim); 726 else if (in6p && !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr)) { 727 struct route_in6 ro6; 728 struct ifnet *lifp; 729 730 bzero(&ro6, sizeof(ro6)); 731 ro6.ro_dst.sin6_family = AF_INET6; 732 ro6.ro_dst.sin6_len = sizeof(struct sockaddr_in6); 733 ro6.ro_dst.sin6_addr = in6p->in6p_faddr; 734 rtalloc((struct route *)&ro6); 735 if (ro6.ro_rt) { 736 lifp = ro6.ro_rt->rt_ifp; 737 RTFREE(ro6.ro_rt); 738 if (lifp) 739 return (ND_IFINFO(lifp)->chlim); 740 } else 741 return (ip6_defhlim); 742 } 743 return (ip6_defhlim); 744} 745 746/* 747 * XXX: this is borrowed from in6_pcbbind(). If possible, we should 748 * share this function by all *bsd*... 749 */ 750int 751in6_pcbsetport(laddr, inp, cred) 752 struct in6_addr *laddr; 753 struct inpcb *inp; 754 struct ucred *cred; 755{ 756 struct socket *so = inp->inp_socket; 757 u_int16_t lport = 0, first, last, *lastport; 758 int count, error = 0, wild = 0; 759 struct inpcbinfo *pcbinfo = inp->inp_pcbinfo; 760 761 INP_INFO_WLOCK_ASSERT(pcbinfo); 762 INP_LOCK_ASSERT(inp); 763 764 /* XXX: this is redundant when called from in6_pcbbind */ 765 if ((so->so_options & (SO_REUSEADDR|SO_REUSEPORT)) == 0) 766 wild = INPLOOKUP_WILDCARD; 767 768 inp->inp_flags |= INP_ANONPORT; 769 770 if (inp->inp_flags & INP_HIGHPORT) { 771 first = ipport_hifirstauto; /* sysctl */ 772 last = ipport_hilastauto; 773 lastport = &pcbinfo->lasthi; 774 } else if (inp->inp_flags & INP_LOWPORT) { 775 if ((error = suser_cred(cred, 0))) 776 return error; 777 first = ipport_lowfirstauto; /* 1023 */ 778 last = ipport_lowlastauto; /* 600 */ 779 lastport = &pcbinfo->lastlow; 780 } else { 781 first = ipport_firstauto; /* sysctl */ 782 last = ipport_lastauto; 783 lastport = &pcbinfo->lastport; 784 } 785 /* 786 * Simple check to ensure all ports are not used up causing 787 * a deadlock here. 788 * 789 * We split the two cases (up and down) so that the direction 790 * is not being tested on each round of the loop. 791 */ 792 if (first > last) { 793 /* 794 * counting down 795 */ 796 count = first - last; 797 798 do { 799 if (count-- < 0) { /* completely used? */ 800 /* 801 * Undo any address bind that may have 802 * occurred above. 803 */ 804 inp->in6p_laddr = in6addr_any; 805 return (EAGAIN); 806 } 807 --*lastport; 808 if (*lastport > first || *lastport < last) 809 *lastport = first; 810 lport = htons(*lastport); 811 } while (in6_pcblookup_local(pcbinfo, &inp->in6p_laddr, 812 lport, wild)); 813 } else { 814 /* 815 * counting up 816 */ 817 count = last - first; 818 819 do { 820 if (count-- < 0) { /* completely used? */ 821 /* 822 * Undo any address bind that may have 823 * occurred above. 824 */ 825 inp->in6p_laddr = in6addr_any; 826 return (EAGAIN); 827 } 828 ++*lastport; 829 if (*lastport < first || *lastport > last) 830 *lastport = first; 831 lport = htons(*lastport); 832 } while (in6_pcblookup_local(pcbinfo, 833 &inp->in6p_laddr, lport, wild)); 834 } 835 836 inp->inp_lport = lport; 837 if (in_pcbinshash(inp) != 0) { 838 inp->in6p_laddr = in6addr_any; 839 inp->inp_lport = 0; 840 return (EAGAIN); 841 } 842 843 return (0); 844} 845 846void 847addrsel_policy_init() 848{ 849 ADDRSEL_LOCK_INIT(); 850 ADDRSEL_SXLOCK_INIT(); 851 852 init_policy_queue(); 853 854 /* initialize the "last resort" policy */ 855 bzero(&defaultaddrpolicy, sizeof(defaultaddrpolicy)); 856 defaultaddrpolicy.label = ADDR_LABEL_NOTAPP; 857} 858 859static struct in6_addrpolicy * 860lookup_addrsel_policy(key) 861 struct sockaddr_in6 *key; 862{ 863 struct in6_addrpolicy *match = NULL; 864 865 ADDRSEL_LOCK(); 866 match = match_addrsel_policy(key); 867 868 if (match == NULL) 869 match = &defaultaddrpolicy; 870 else 871 match->use++; 872 ADDRSEL_UNLOCK(); 873 874 return (match); 875} 876 877/* 878 * Subroutines to manage the address selection policy table via sysctl. 879 */ 880struct walkarg { 881 struct sysctl_req *w_req; 882}; 883 884static int in6_src_sysctl(SYSCTL_HANDLER_ARGS); 885SYSCTL_DECL(_net_inet6_ip6); 886SYSCTL_NODE(_net_inet6_ip6, IPV6CTL_ADDRCTLPOLICY, addrctlpolicy, 887 CTLFLAG_RD, in6_src_sysctl, ""); 888 889static int 890in6_src_sysctl(SYSCTL_HANDLER_ARGS) 891{ 892 struct walkarg w; 893 894 if (req->newptr) 895 return EPERM; 896 897 bzero(&w, sizeof(w)); 898 w.w_req = req; 899 900 return (walk_addrsel_policy(dump_addrsel_policyent, &w)); 901} 902 903int 904in6_src_ioctl(cmd, data) 905 u_long cmd; 906 caddr_t data; 907{ 908 int i; 909 struct in6_addrpolicy ent0; 910 911 if (cmd != SIOCAADDRCTL_POLICY && cmd != SIOCDADDRCTL_POLICY) 912 return (EOPNOTSUPP); /* check for safety */ 913 914 ent0 = *(struct in6_addrpolicy *)data; 915 916 if (ent0.label == ADDR_LABEL_NOTAPP) 917 return (EINVAL); 918 /* check if the prefix mask is consecutive. */ 919 if (in6_mask2len(&ent0.addrmask.sin6_addr, NULL) < 0) 920 return (EINVAL); 921 /* clear trailing garbages (if any) of the prefix address. */ 922 for (i = 0; i < 4; i++) { 923 ent0.addr.sin6_addr.s6_addr32[i] &= 924 ent0.addrmask.sin6_addr.s6_addr32[i]; 925 } 926 ent0.use = 0; 927 928 switch (cmd) { 929 case SIOCAADDRCTL_POLICY: 930 return (add_addrsel_policyent(&ent0)); 931 case SIOCDADDRCTL_POLICY: 932 return (delete_addrsel_policyent(&ent0)); 933 } 934 935 return (0); /* XXX: compromise compilers */ 936} 937 938/* 939 * The followings are implementation of the policy table using a 940 * simple tail queue. 941 * XXX such details should be hidden. 942 * XXX implementation using binary tree should be more efficient. 943 */ 944struct addrsel_policyent { 945 TAILQ_ENTRY(addrsel_policyent) ape_entry; 946 struct in6_addrpolicy ape_policy; 947}; 948 949TAILQ_HEAD(addrsel_policyhead, addrsel_policyent); 950 951struct addrsel_policyhead addrsel_policytab; 952 953static void 954init_policy_queue() 955{ 956 TAILQ_INIT(&addrsel_policytab); 957} 958 959static int 960add_addrsel_policyent(newpolicy) 961 struct in6_addrpolicy *newpolicy; 962{ 963 struct addrsel_policyent *new, *pol; 964 965 MALLOC(new, struct addrsel_policyent *, sizeof(*new), M_IFADDR, 966 M_WAITOK); 967 ADDRSEL_XLOCK(); 968 ADDRSEL_LOCK(); 969 970 /* duplication check */ 971 TAILQ_FOREACH(pol, &addrsel_policytab, ape_entry) { 972 if (IN6_ARE_ADDR_EQUAL(&newpolicy->addr.sin6_addr, 973 &pol->ape_policy.addr.sin6_addr) && 974 IN6_ARE_ADDR_EQUAL(&newpolicy->addrmask.sin6_addr, 975 &pol->ape_policy.addrmask.sin6_addr)) { 976 ADDRSEL_UNLOCK(); 977 ADDRSEL_XUNLOCK(); 978 FREE(new, M_IFADDR); 979 return (EEXIST); /* or override it? */ 980 } 981 } 982 983 bzero(new, sizeof(*new)); 984 985 /* XXX: should validate entry */ 986 new->ape_policy = *newpolicy; 987 988 TAILQ_INSERT_TAIL(&addrsel_policytab, new, ape_entry); 989 ADDRSEL_UNLOCK(); 990 ADDRSEL_XUNLOCK(); 991 992 return (0); 993} 994 995static int 996delete_addrsel_policyent(key) 997 struct in6_addrpolicy *key; 998{ 999 struct addrsel_policyent *pol; 1000 1001 ADDRSEL_XLOCK(); 1002 ADDRSEL_LOCK(); 1003 1004 /* search for the entry in the table */ 1005 TAILQ_FOREACH(pol, &addrsel_policytab, ape_entry) { 1006 if (IN6_ARE_ADDR_EQUAL(&key->addr.sin6_addr, 1007 &pol->ape_policy.addr.sin6_addr) && 1008 IN6_ARE_ADDR_EQUAL(&key->addrmask.sin6_addr, 1009 &pol->ape_policy.addrmask.sin6_addr)) { 1010 break; 1011 } 1012 } 1013 if (pol == NULL) { 1014 ADDRSEL_UNLOCK(); 1015 ADDRSEL_XUNLOCK(); 1016 return (ESRCH); 1017 } 1018 1019 TAILQ_REMOVE(&addrsel_policytab, pol, ape_entry); 1020 ADDRSEL_UNLOCK(); 1021 ADDRSEL_XUNLOCK(); 1022 1023 return (0); 1024} 1025 1026static int 1027walk_addrsel_policy(callback, w) 1028 int (*callback) __P((struct in6_addrpolicy *, void *)); 1029 void *w; 1030{ 1031 struct addrsel_policyent *pol; 1032 int error = 0; 1033 1034 ADDRSEL_SLOCK(); 1035 TAILQ_FOREACH(pol, &addrsel_policytab, ape_entry) { 1036 if ((error = (*callback)(&pol->ape_policy, w)) != 0) { 1037 ADDRSEL_SUNLOCK(); 1038 return (error); 1039 } 1040 } 1041 ADDRSEL_SUNLOCK(); 1042 return (error); 1043} 1044 1045static int 1046dump_addrsel_policyent(pol, arg) 1047 struct in6_addrpolicy *pol; 1048 void *arg; 1049{ 1050 int error = 0; 1051 struct walkarg *w = arg; 1052 1053 error = SYSCTL_OUT(w->w_req, pol, sizeof(*pol)); 1054 1055 return (error); 1056} 1057 1058static struct in6_addrpolicy * 1059match_addrsel_policy(key) 1060 struct sockaddr_in6 *key; 1061{ 1062 struct addrsel_policyent *pent; 1063 struct in6_addrpolicy *bestpol = NULL, *pol; 1064 int matchlen, bestmatchlen = -1; 1065 u_char *mp, *ep, *k, *p, m; 1066 1067 TAILQ_FOREACH(pent, &addrsel_policytab, ape_entry) { 1068 matchlen = 0; 1069 1070 pol = &pent->ape_policy; 1071 mp = (u_char *)&pol->addrmask.sin6_addr; 1072 ep = mp + 16; /* XXX: scope field? */ 1073 k = (u_char *)&key->sin6_addr; 1074 p = (u_char *)&pol->addr.sin6_addr; 1075 for (; mp < ep && *mp; mp++, k++, p++) { 1076 m = *mp; 1077 if ((*k & m) != *p) 1078 goto next; /* not match */ 1079 if (m == 0xff) /* short cut for a typical case */ 1080 matchlen += 8; 1081 else { 1082 while (m >= 0x80) { 1083 matchlen++; 1084 m <<= 1; 1085 } 1086 } 1087 } 1088 1089 /* matched. check if this is better than the current best. */ 1090 if (bestpol == NULL || 1091 matchlen > bestmatchlen) { 1092 bestpol = pol; 1093 bestmatchlen = matchlen; 1094 } 1095 1096 next: 1097 continue; 1098 } 1099 1100 return (bestpol); 1101} 1102