in6_src.c revision 207276
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: head/sys/netinet6/in6_src.c 207276 2010-04-27 15:05:03Z bz $"); 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 111static 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 117static 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 126static VNET_DEFINE(struct in6_addrpolicy, defaultaddrpolicy); 127VNET_DEFINE(int, ip6_prefer_tempaddr); 128 129#define V_defaultaddrpolicy VNET(defaultaddrpolicy) 130 131static int selectroute __P((struct sockaddr_in6 *, struct ip6_pktopts *, 132 struct ip6_moptions *, struct route_in6 *, struct ifnet **, 133 struct rtentry **, int)); 134static int in6_selectif __P((struct sockaddr_in6 *, struct ip6_pktopts *, 135 struct ip6_moptions *, struct route_in6 *ro, struct ifnet **)); 136 137static struct in6_addrpolicy *lookup_addrsel_policy(struct sockaddr_in6 *); 138 139static void init_policy_queue(void); 140static int add_addrsel_policyent(struct in6_addrpolicy *); 141static int delete_addrsel_policyent(struct in6_addrpolicy *); 142static int walk_addrsel_policy __P((int (*)(struct in6_addrpolicy *, void *), 143 void *)); 144static int dump_addrsel_policyent(struct in6_addrpolicy *, void *); 145static 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 180int 181in6_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, tmp; 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 * Bypass source address selection and use the primary jail IP 275 * if requested. 276 */ 277 if (cred != NULL && !prison_saddrsel_ip6(cred, srcp)) 278 return (0); 279 280 /* 281 * If the address is not specified, choose the best one based on 282 * the outgoing interface and the destination address. 283 */ 284 /* get the outgoing interface */ 285 if ((error = in6_selectif(dstsock, opts, mopts, ro, &ifp)) != 0) 286 return (error); 287 288#ifdef DIAGNOSTIC 289 if (ifp == NULL) /* this should not happen */ 290 panic("in6_selectsrc: NULL ifp"); 291#endif 292 error = in6_setscope(&dst, ifp, &odstzone); 293 if (error) 294 return (error); 295 296 IN6_IFADDR_RLOCK(); 297 TAILQ_FOREACH(ia, &V_in6_ifaddrhead, ia_link) { 298 int new_scope = -1, new_matchlen = -1; 299 struct in6_addrpolicy *new_policy = NULL; 300 u_int32_t srczone, osrczone, dstzone; 301 struct in6_addr src; 302 struct ifnet *ifp1 = ia->ia_ifp; 303 304 /* 305 * We'll never take an address that breaks the scope zone 306 * of the destination. We also skip an address if its zone 307 * does not contain the outgoing interface. 308 * XXX: we should probably use sin6_scope_id here. 309 */ 310 if (in6_setscope(&dst, ifp1, &dstzone) || 311 odstzone != dstzone) { 312 continue; 313 } 314 src = ia->ia_addr.sin6_addr; 315 if (in6_setscope(&src, ifp, &osrczone) || 316 in6_setscope(&src, ifp1, &srczone) || 317 osrczone != srczone) { 318 continue; 319 } 320 321 /* avoid unusable addresses */ 322 if ((ia->ia6_flags & 323 (IN6_IFF_NOTREADY | IN6_IFF_ANYCAST | IN6_IFF_DETACHED))) { 324 continue; 325 } 326 if (!V_ip6_use_deprecated && IFA6_IS_DEPRECATED(ia)) 327 continue; 328 329 /* If jailed only take addresses of the jail into account. */ 330 if (cred != NULL && 331 prison_check_ip6(cred, &ia->ia_addr.sin6_addr) != 0) 332 continue; 333 334 /* Rule 1: Prefer same address */ 335 if (IN6_ARE_ADDR_EQUAL(&dst, &ia->ia_addr.sin6_addr)) { 336 ia_best = ia; 337 BREAK(1); /* there should be no better candidate */ 338 } 339 340 if (ia_best == NULL) 341 REPLACE(0); 342 343 /* Rule 2: Prefer appropriate scope */ 344 if (dst_scope < 0) 345 dst_scope = in6_addrscope(&dst); 346 new_scope = in6_addrscope(&ia->ia_addr.sin6_addr); 347 if (IN6_ARE_SCOPE_CMP(best_scope, new_scope) < 0) { 348 if (IN6_ARE_SCOPE_CMP(best_scope, dst_scope) < 0) 349 REPLACE(2); 350 NEXT(2); 351 } else if (IN6_ARE_SCOPE_CMP(new_scope, best_scope) < 0) { 352 if (IN6_ARE_SCOPE_CMP(new_scope, dst_scope) < 0) 353 NEXT(2); 354 REPLACE(2); 355 } 356 357 /* 358 * Rule 3: Avoid deprecated addresses. Note that the case of 359 * !ip6_use_deprecated is already rejected above. 360 */ 361 if (!IFA6_IS_DEPRECATED(ia_best) && IFA6_IS_DEPRECATED(ia)) 362 NEXT(3); 363 if (IFA6_IS_DEPRECATED(ia_best) && !IFA6_IS_DEPRECATED(ia)) 364 REPLACE(3); 365 366 /* Rule 4: Prefer home addresses */ 367 /* 368 * XXX: This is a TODO. We should probably merge the MIP6 369 * case above. 370 */ 371 372 /* Rule 5: Prefer outgoing interface */ 373 if (ia_best->ia_ifp == ifp && ia->ia_ifp != ifp) 374 NEXT(5); 375 if (ia_best->ia_ifp != ifp && ia->ia_ifp == ifp) 376 REPLACE(5); 377 378 /* 379 * Rule 6: Prefer matching label 380 * Note that best_policy should be non-NULL here. 381 */ 382 if (dst_policy == NULL) 383 dst_policy = lookup_addrsel_policy(dstsock); 384 if (dst_policy->label != ADDR_LABEL_NOTAPP) { 385 new_policy = lookup_addrsel_policy(&ia->ia_addr); 386 if (dst_policy->label == best_policy->label && 387 dst_policy->label != new_policy->label) 388 NEXT(6); 389 if (dst_policy->label != best_policy->label && 390 dst_policy->label == new_policy->label) 391 REPLACE(6); 392 } 393 394 /* 395 * Rule 7: Prefer public addresses. 396 * We allow users to reverse the logic by configuring 397 * a sysctl variable, so that privacy conscious users can 398 * always prefer temporary addresses. 399 */ 400 if (opts == NULL || 401 opts->ip6po_prefer_tempaddr == IP6PO_TEMPADDR_SYSTEM) { 402 prefer_tempaddr = V_ip6_prefer_tempaddr; 403 } else if (opts->ip6po_prefer_tempaddr == 404 IP6PO_TEMPADDR_NOTPREFER) { 405 prefer_tempaddr = 0; 406 } else 407 prefer_tempaddr = 1; 408 if (!(ia_best->ia6_flags & IN6_IFF_TEMPORARY) && 409 (ia->ia6_flags & IN6_IFF_TEMPORARY)) { 410 if (prefer_tempaddr) 411 REPLACE(7); 412 else 413 NEXT(7); 414 } 415 if ((ia_best->ia6_flags & IN6_IFF_TEMPORARY) && 416 !(ia->ia6_flags & IN6_IFF_TEMPORARY)) { 417 if (prefer_tempaddr) 418 NEXT(7); 419 else 420 REPLACE(7); 421 } 422 423 /* 424 * Rule 8: prefer addresses on alive interfaces. 425 * This is a KAME specific rule. 426 */ 427 if ((ia_best->ia_ifp->if_flags & IFF_UP) && 428 !(ia->ia_ifp->if_flags & IFF_UP)) 429 NEXT(8); 430 if (!(ia_best->ia_ifp->if_flags & IFF_UP) && 431 (ia->ia_ifp->if_flags & IFF_UP)) 432 REPLACE(8); 433 434 /* 435 * Rule 14: Use longest matching prefix. 436 * Note: in the address selection draft, this rule is 437 * documented as "Rule 8". However, since it is also 438 * documented that this rule can be overridden, we assign 439 * a large number so that it is easy to assign smaller numbers 440 * to more preferred rules. 441 */ 442 new_matchlen = in6_matchlen(&ia->ia_addr.sin6_addr, &dst); 443 if (best_matchlen < new_matchlen) 444 REPLACE(14); 445 if (new_matchlen < best_matchlen) 446 NEXT(14); 447 448 /* Rule 15 is reserved. */ 449 450 /* 451 * Last resort: just keep the current candidate. 452 * Or, do we need more rules? 453 */ 454 continue; 455 456 replace: 457 ia_best = ia; 458 best_scope = (new_scope >= 0 ? new_scope : 459 in6_addrscope(&ia_best->ia_addr.sin6_addr)); 460 best_policy = (new_policy ? new_policy : 461 lookup_addrsel_policy(&ia_best->ia_addr)); 462 best_matchlen = (new_matchlen >= 0 ? new_matchlen : 463 in6_matchlen(&ia_best->ia_addr.sin6_addr, 464 &dst)); 465 466 next: 467 continue; 468 469 out: 470 break; 471 } 472 473 if ((ia = ia_best) == NULL) { 474 IN6_IFADDR_RUNLOCK(); 475 return (EADDRNOTAVAIL); 476 } 477 478 /* 479 * At this point at least one of the addresses belonged to the jail 480 * but it could still be, that we want to further restrict it, e.g. 481 * theoratically IN6_IS_ADDR_LOOPBACK. 482 * It must not be IN6_IS_ADDR_UNSPECIFIED anymore. 483 * prison_local_ip6() will fix an IN6_IS_ADDR_LOOPBACK but should 484 * let all others previously selected pass. 485 * Use tmp to not change ::1 on lo0 to the primary jail address. 486 */ 487 tmp = ia->ia_addr.sin6_addr; 488 if (cred != NULL && prison_local_ip6(cred, &tmp, (inp != NULL && 489 (inp->inp_flags & IN6P_IPV6_V6ONLY) != 0)) != 0) { 490 IN6_IFADDR_RUNLOCK(); 491 return (EADDRNOTAVAIL); 492 } 493 494 if (ifpp) 495 *ifpp = ifp; 496 497 bcopy(&tmp, srcp, sizeof(*srcp)); 498 IN6_IFADDR_RUNLOCK(); 499 return (0); 500} 501 502/* 503 * clone - meaningful only for bsdi and freebsd 504 */ 505static int 506selectroute(struct sockaddr_in6 *dstsock, struct ip6_pktopts *opts, 507 struct ip6_moptions *mopts, struct route_in6 *ro, 508 struct ifnet **retifp, struct rtentry **retrt, int norouteok) 509{ 510 int error = 0; 511 struct ifnet *ifp = NULL; 512 struct rtentry *rt = NULL; 513 struct sockaddr_in6 *sin6_next; 514 struct in6_pktinfo *pi = NULL; 515 struct in6_addr *dst = &dstsock->sin6_addr; 516#if 0 517 char ip6buf[INET6_ADDRSTRLEN]; 518 519 if (dstsock->sin6_addr.s6_addr32[0] == 0 && 520 dstsock->sin6_addr.s6_addr32[1] == 0 && 521 !IN6_IS_ADDR_LOOPBACK(&dstsock->sin6_addr)) { 522 printf("in6_selectroute: strange destination %s\n", 523 ip6_sprintf(ip6buf, &dstsock->sin6_addr)); 524 } else { 525 printf("in6_selectroute: destination = %s%%%d\n", 526 ip6_sprintf(ip6buf, &dstsock->sin6_addr), 527 dstsock->sin6_scope_id); /* for debug */ 528 } 529#endif 530 531 /* If the caller specify the outgoing interface explicitly, use it. */ 532 if (opts && (pi = opts->ip6po_pktinfo) != NULL && pi->ipi6_ifindex) { 533 /* XXX boundary check is assumed to be already done. */ 534 ifp = ifnet_byindex(pi->ipi6_ifindex); 535 if (ifp != NULL && 536 (norouteok || retrt == NULL || 537 IN6_IS_ADDR_MULTICAST(dst))) { 538 /* 539 * we do not have to check or get the route for 540 * multicast. 541 */ 542 goto done; 543 } else 544 goto getroute; 545 } 546 547 /* 548 * If the destination address is a multicast address and the outgoing 549 * interface for the address is specified by the caller, use it. 550 */ 551 if (IN6_IS_ADDR_MULTICAST(dst) && 552 mopts != NULL && (ifp = mopts->im6o_multicast_ifp) != NULL) { 553 goto done; /* we do not need a route for multicast. */ 554 } 555 556 getroute: 557 /* 558 * If the next hop address for the packet is specified by the caller, 559 * use it as the gateway. 560 */ 561 if (opts && opts->ip6po_nexthop) { 562 struct route_in6 *ron; 563 struct llentry *la; 564 565 sin6_next = satosin6(opts->ip6po_nexthop); 566 567 /* at this moment, we only support AF_INET6 next hops */ 568 if (sin6_next->sin6_family != AF_INET6) { 569 error = EAFNOSUPPORT; /* or should we proceed? */ 570 goto done; 571 } 572 573 /* 574 * If the next hop is an IPv6 address, then the node identified 575 * by that address must be a neighbor of the sending host. 576 */ 577 ron = &opts->ip6po_nextroute; 578 /* 579 * XXX what do we do here? 580 * PLZ to be fixing 581 */ 582 583 584 if (ron->ro_rt == NULL) { 585 rtalloc((struct route *)ron); /* multi path case? */ 586 if (ron->ro_rt == NULL) { 587 if (ron->ro_rt) { 588 RTFREE(ron->ro_rt); 589 ron->ro_rt = NULL; 590 } 591 error = EHOSTUNREACH; 592 goto done; 593 } 594 } 595 596 rt = ron->ro_rt; 597 ifp = rt->rt_ifp; 598 IF_AFDATA_LOCK(ifp); 599 la = lla_lookup(LLTABLE6(ifp), 0, (struct sockaddr *)&sin6_next->sin6_addr); 600 IF_AFDATA_UNLOCK(ifp); 601 if (la != NULL) 602 LLE_RUNLOCK(la); 603 else { 604 error = EHOSTUNREACH; 605 goto done; 606 } 607#if 0 608 if ((ron->ro_rt && 609 (ron->ro_rt->rt_flags & (RTF_UP | RTF_LLINFO)) != 610 (RTF_UP | RTF_LLINFO)) || 611 !IN6_ARE_ADDR_EQUAL(&satosin6(&ron->ro_dst)->sin6_addr, 612 &sin6_next->sin6_addr)) { 613 if (ron->ro_rt) { 614 RTFREE(ron->ro_rt); 615 ron->ro_rt = NULL; 616 } 617 *satosin6(&ron->ro_dst) = *sin6_next; 618 } 619 if (ron->ro_rt == NULL) { 620 rtalloc((struct route *)ron); /* multi path case? */ 621 if (ron->ro_rt == NULL || 622 !(ron->ro_rt->rt_flags & RTF_LLINFO)) { 623 if (ron->ro_rt) { 624 RTFREE(ron->ro_rt); 625 ron->ro_rt = NULL; 626 } 627 error = EHOSTUNREACH; 628 goto done; 629 } 630 } 631#endif 632 633 /* 634 * When cloning is required, try to allocate a route to the 635 * destination so that the caller can store path MTU 636 * information. 637 */ 638 goto done; 639 } 640 641 /* 642 * Use a cached route if it exists and is valid, else try to allocate 643 * a new one. Note that we should check the address family of the 644 * cached destination, in case of sharing the cache with IPv4. 645 */ 646 if (ro) { 647 if (ro->ro_rt && 648 (!(ro->ro_rt->rt_flags & RTF_UP) || 649 ((struct sockaddr *)(&ro->ro_dst))->sa_family != AF_INET6 || 650 !IN6_ARE_ADDR_EQUAL(&satosin6(&ro->ro_dst)->sin6_addr, 651 dst))) { 652 RTFREE(ro->ro_rt); 653 ro->ro_rt = (struct rtentry *)NULL; 654 } 655 if (ro->ro_rt == (struct rtentry *)NULL) { 656 struct sockaddr_in6 *sa6; 657 658 /* No route yet, so try to acquire one */ 659 bzero(&ro->ro_dst, sizeof(struct sockaddr_in6)); 660 sa6 = (struct sockaddr_in6 *)&ro->ro_dst; 661 *sa6 = *dstsock; 662 sa6->sin6_scope_id = 0; 663 664#ifdef RADIX_MPATH 665 rtalloc_mpath((struct route *)ro, 666 ntohl(sa6->sin6_addr.s6_addr32[3])); 667#else 668 ro->ro_rt = rtalloc1(&((struct route *)ro) 669 ->ro_dst, 0, 0UL); 670 if (ro->ro_rt) 671 RT_UNLOCK(ro->ro_rt); 672#endif 673 } 674 675 /* 676 * do not care about the result if we have the nexthop 677 * explicitly specified. 678 */ 679 if (opts && opts->ip6po_nexthop) 680 goto done; 681 682 if (ro->ro_rt) { 683 ifp = ro->ro_rt->rt_ifp; 684 685 if (ifp == NULL) { /* can this really happen? */ 686 RTFREE(ro->ro_rt); 687 ro->ro_rt = NULL; 688 } 689 } 690 if (ro->ro_rt == NULL) 691 error = EHOSTUNREACH; 692 rt = ro->ro_rt; 693 694 /* 695 * Check if the outgoing interface conflicts with 696 * the interface specified by ipi6_ifindex (if specified). 697 * Note that loopback interface is always okay. 698 * (this may happen when we are sending a packet to one of 699 * our own addresses.) 700 */ 701 if (ifp && opts && opts->ip6po_pktinfo && 702 opts->ip6po_pktinfo->ipi6_ifindex) { 703 if (!(ifp->if_flags & IFF_LOOPBACK) && 704 ifp->if_index != 705 opts->ip6po_pktinfo->ipi6_ifindex) { 706 error = EHOSTUNREACH; 707 goto done; 708 } 709 } 710 } 711 712 done: 713 if (ifp == NULL && rt == NULL) { 714 /* 715 * This can happen if the caller did not pass a cached route 716 * nor any other hints. We treat this case an error. 717 */ 718 error = EHOSTUNREACH; 719 } 720 if (error == EHOSTUNREACH) 721 V_ip6stat.ip6s_noroute++; 722 723 if (retifp != NULL) { 724 *retifp = ifp; 725 726 /* 727 * Adjust the "outgoing" interface. If we're going to loop 728 * the packet back to ourselves, the ifp would be the loopback 729 * interface. However, we'd rather know the interface associated 730 * to the destination address (which should probably be one of 731 * our own addresses.) 732 */ 733 if (rt) { 734 if ((rt->rt_ifp->if_flags & IFF_LOOPBACK) && 735 (rt->rt_gateway->sa_family == AF_LINK)) 736 *retifp = 737 ifnet_byindex(((struct sockaddr_dl *) 738 rt->rt_gateway)->sdl_index); 739 } 740 } 741 742 if (retrt != NULL) 743 *retrt = rt; /* rt may be NULL */ 744 745 return (error); 746} 747 748static int 749in6_selectif(struct sockaddr_in6 *dstsock, struct ip6_pktopts *opts, 750 struct ip6_moptions *mopts, struct route_in6 *ro, struct ifnet **retifp) 751{ 752 int error; 753 struct route_in6 sro; 754 struct rtentry *rt = NULL; 755 756 if (ro == NULL) { 757 bzero(&sro, sizeof(sro)); 758 ro = &sro; 759 } 760 761 if ((error = selectroute(dstsock, opts, mopts, ro, retifp, 762 &rt, 1)) != 0) { 763 if (ro == &sro && rt && rt == sro.ro_rt) 764 RTFREE(rt); 765 return (error); 766 } 767 768 /* 769 * do not use a rejected or black hole route. 770 * XXX: this check should be done in the L2 output routine. 771 * However, if we skipped this check here, we'd see the following 772 * scenario: 773 * - install a rejected route for a scoped address prefix 774 * (like fe80::/10) 775 * - send a packet to a destination that matches the scoped prefix, 776 * with ambiguity about the scope zone. 777 * - pick the outgoing interface from the route, and disambiguate the 778 * scope zone with the interface. 779 * - ip6_output() would try to get another route with the "new" 780 * destination, which may be valid. 781 * - we'd see no error on output. 782 * Although this may not be very harmful, it should still be confusing. 783 * We thus reject the case here. 784 */ 785 if (rt && (rt->rt_flags & (RTF_REJECT | RTF_BLACKHOLE))) { 786 int flags = (rt->rt_flags & RTF_HOST ? EHOSTUNREACH : ENETUNREACH); 787 788 if (ro == &sro && rt && rt == sro.ro_rt) 789 RTFREE(rt); 790 return (flags); 791 } 792 793 if (ro == &sro && rt && rt == sro.ro_rt) 794 RTFREE(rt); 795 return (0); 796} 797 798/* 799 * clone - meaningful only for bsdi and freebsd 800 */ 801int 802in6_selectroute(struct sockaddr_in6 *dstsock, struct ip6_pktopts *opts, 803 struct ip6_moptions *mopts, struct route_in6 *ro, 804 struct ifnet **retifp, struct rtentry **retrt) 805{ 806 807 return (selectroute(dstsock, opts, mopts, ro, retifp, 808 retrt, 0)); 809} 810 811/* 812 * Default hop limit selection. The precedence is as follows: 813 * 1. Hoplimit value specified via ioctl. 814 * 2. (If the outgoing interface is detected) the current 815 * hop limit of the interface specified by router advertisement. 816 * 3. The system default hoplimit. 817 */ 818int 819in6_selecthlim(struct inpcb *in6p, struct ifnet *ifp) 820{ 821 822 if (in6p && in6p->in6p_hops >= 0) 823 return (in6p->in6p_hops); 824 else if (ifp) 825 return (ND_IFINFO(ifp)->chlim); 826 else if (in6p && !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr)) { 827 struct route_in6 ro6; 828 struct ifnet *lifp; 829 830 bzero(&ro6, sizeof(ro6)); 831 ro6.ro_dst.sin6_family = AF_INET6; 832 ro6.ro_dst.sin6_len = sizeof(struct sockaddr_in6); 833 ro6.ro_dst.sin6_addr = in6p->in6p_faddr; 834 rtalloc((struct route *)&ro6); 835 if (ro6.ro_rt) { 836 lifp = ro6.ro_rt->rt_ifp; 837 RTFREE(ro6.ro_rt); 838 if (lifp) 839 return (ND_IFINFO(lifp)->chlim); 840 } else 841 return (V_ip6_defhlim); 842 } 843 return (V_ip6_defhlim); 844} 845 846/* 847 * XXX: this is borrowed from in6_pcbbind(). If possible, we should 848 * share this function by all *bsd*... 849 */ 850int 851in6_pcbsetport(struct in6_addr *laddr, struct inpcb *inp, struct ucred *cred) 852{ 853 struct socket *so = inp->inp_socket; 854 u_int16_t lport = 0, first, last, *lastport; 855 int count, error, wild = 0, dorandom; 856 struct inpcbinfo *pcbinfo = inp->inp_pcbinfo; 857 858 INP_INFO_WLOCK_ASSERT(pcbinfo); 859 INP_WLOCK_ASSERT(inp); 860 861 error = prison_local_ip6(cred, laddr, 862 ((inp->inp_flags & IN6P_IPV6_V6ONLY) != 0)); 863 if (error) 864 return(error); 865 866 /* XXX: this is redundant when called from in6_pcbbind */ 867 if ((so->so_options & (SO_REUSEADDR|SO_REUSEPORT)) == 0) 868 wild = INPLOOKUP_WILDCARD; 869 870 inp->inp_flags |= INP_ANONPORT; 871 872 if (inp->inp_flags & INP_HIGHPORT) { 873 first = V_ipport_hifirstauto; /* sysctl */ 874 last = V_ipport_hilastauto; 875 lastport = &pcbinfo->ipi_lasthi; 876 } else if (inp->inp_flags & INP_LOWPORT) { 877 error = priv_check_cred(cred, PRIV_NETINET_RESERVEDPORT, 0); 878 if (error) 879 return error; 880 first = V_ipport_lowfirstauto; /* 1023 */ 881 last = V_ipport_lowlastauto; /* 600 */ 882 lastport = &pcbinfo->ipi_lastlow; 883 } else { 884 first = V_ipport_firstauto; /* sysctl */ 885 last = V_ipport_lastauto; 886 lastport = &pcbinfo->ipi_lastport; 887 } 888 889 /* 890 * For UDP, use random port allocation as long as the user 891 * allows it. For TCP (and as of yet unknown) connections, 892 * use random port allocation only if the user allows it AND 893 * ipport_tick() allows it. 894 */ 895 if (V_ipport_randomized && 896 (!V_ipport_stoprandom || pcbinfo == &V_udbinfo)) 897 dorandom = 1; 898 else 899 dorandom = 0; 900 /* 901 * It makes no sense to do random port allocation if 902 * we have the only port available. 903 */ 904 if (first == last) 905 dorandom = 0; 906 /* Make sure to not include UDP packets in the count. */ 907 if (pcbinfo != &V_udbinfo) 908 V_ipport_tcpallocs++; 909 910 /* 911 * Instead of having two loops further down counting up or down 912 * make sure that first is always <= last and go with only one 913 * code path implementing all logic. 914 */ 915 if (first > last) { 916 u_int16_t aux; 917 918 aux = first; 919 first = last; 920 last = aux; 921 } 922 923 if (dorandom) 924 *lastport = first + (arc4random() % (last - first)); 925 926 count = last - first; 927 928 do { 929 if (count-- < 0) { /* completely used? */ 930 /* Undo an address bind that may have occurred. */ 931 inp->in6p_laddr = in6addr_any; 932 return (EADDRNOTAVAIL); 933 } 934 ++*lastport; 935 if (*lastport < first || *lastport > last) 936 *lastport = first; 937 lport = htons(*lastport); 938 } while (in6_pcblookup_local(pcbinfo, &inp->in6p_laddr, 939 lport, wild, cred)); 940 941 inp->inp_lport = lport; 942 if (in_pcbinshash(inp) != 0) { 943 inp->in6p_laddr = in6addr_any; 944 inp->inp_lport = 0; 945 return (EAGAIN); 946 } 947 948 return (0); 949} 950 951void 952addrsel_policy_init(void) 953{ 954 955 V_ip6_prefer_tempaddr = 0; 956 957 init_policy_queue(); 958 959 /* initialize the "last resort" policy */ 960 bzero(&V_defaultaddrpolicy, sizeof(V_defaultaddrpolicy)); 961 V_defaultaddrpolicy.label = ADDR_LABEL_NOTAPP; 962 963 if (!IS_DEFAULT_VNET(curvnet)) 964 return; 965 966 ADDRSEL_LOCK_INIT(); 967 ADDRSEL_SXLOCK_INIT(); 968} 969 970static struct in6_addrpolicy * 971lookup_addrsel_policy(struct sockaddr_in6 *key) 972{ 973 struct in6_addrpolicy *match = NULL; 974 975 ADDRSEL_LOCK(); 976 match = match_addrsel_policy(key); 977 978 if (match == NULL) 979 match = &V_defaultaddrpolicy; 980 else 981 match->use++; 982 ADDRSEL_UNLOCK(); 983 984 return (match); 985} 986 987/* 988 * Subroutines to manage the address selection policy table via sysctl. 989 */ 990struct walkarg { 991 struct sysctl_req *w_req; 992}; 993 994static int in6_src_sysctl(SYSCTL_HANDLER_ARGS); 995SYSCTL_DECL(_net_inet6_ip6); 996SYSCTL_NODE(_net_inet6_ip6, IPV6CTL_ADDRCTLPOLICY, addrctlpolicy, 997 CTLFLAG_RD, in6_src_sysctl, ""); 998 999static int 1000in6_src_sysctl(SYSCTL_HANDLER_ARGS) 1001{ 1002 struct walkarg w; 1003 1004 if (req->newptr) 1005 return EPERM; 1006 1007 bzero(&w, sizeof(w)); 1008 w.w_req = req; 1009 1010 return (walk_addrsel_policy(dump_addrsel_policyent, &w)); 1011} 1012 1013int 1014in6_src_ioctl(u_long cmd, caddr_t data) 1015{ 1016 int i; 1017 struct in6_addrpolicy ent0; 1018 1019 if (cmd != SIOCAADDRCTL_POLICY && cmd != SIOCDADDRCTL_POLICY) 1020 return (EOPNOTSUPP); /* check for safety */ 1021 1022 ent0 = *(struct in6_addrpolicy *)data; 1023 1024 if (ent0.label == ADDR_LABEL_NOTAPP) 1025 return (EINVAL); 1026 /* check if the prefix mask is consecutive. */ 1027 if (in6_mask2len(&ent0.addrmask.sin6_addr, NULL) < 0) 1028 return (EINVAL); 1029 /* clear trailing garbages (if any) of the prefix address. */ 1030 for (i = 0; i < 4; i++) { 1031 ent0.addr.sin6_addr.s6_addr32[i] &= 1032 ent0.addrmask.sin6_addr.s6_addr32[i]; 1033 } 1034 ent0.use = 0; 1035 1036 switch (cmd) { 1037 case SIOCAADDRCTL_POLICY: 1038 return (add_addrsel_policyent(&ent0)); 1039 case SIOCDADDRCTL_POLICY: 1040 return (delete_addrsel_policyent(&ent0)); 1041 } 1042 1043 return (0); /* XXX: compromise compilers */ 1044} 1045 1046/* 1047 * The followings are implementation of the policy table using a 1048 * simple tail queue. 1049 * XXX such details should be hidden. 1050 * XXX implementation using binary tree should be more efficient. 1051 */ 1052struct addrsel_policyent { 1053 TAILQ_ENTRY(addrsel_policyent) ape_entry; 1054 struct in6_addrpolicy ape_policy; 1055}; 1056 1057TAILQ_HEAD(addrsel_policyhead, addrsel_policyent); 1058 1059static VNET_DEFINE(struct addrsel_policyhead, addrsel_policytab); 1060#define V_addrsel_policytab VNET(addrsel_policytab) 1061 1062static void 1063init_policy_queue(void) 1064{ 1065 1066 TAILQ_INIT(&V_addrsel_policytab); 1067} 1068 1069static int 1070add_addrsel_policyent(struct in6_addrpolicy *newpolicy) 1071{ 1072 struct addrsel_policyent *new, *pol; 1073 1074 new = malloc(sizeof(*new), M_IFADDR, 1075 M_WAITOK); 1076 ADDRSEL_XLOCK(); 1077 ADDRSEL_LOCK(); 1078 1079 /* duplication check */ 1080 TAILQ_FOREACH(pol, &V_addrsel_policytab, ape_entry) { 1081 if (IN6_ARE_ADDR_EQUAL(&newpolicy->addr.sin6_addr, 1082 &pol->ape_policy.addr.sin6_addr) && 1083 IN6_ARE_ADDR_EQUAL(&newpolicy->addrmask.sin6_addr, 1084 &pol->ape_policy.addrmask.sin6_addr)) { 1085 ADDRSEL_UNLOCK(); 1086 ADDRSEL_XUNLOCK(); 1087 free(new, M_IFADDR); 1088 return (EEXIST); /* or override it? */ 1089 } 1090 } 1091 1092 bzero(new, sizeof(*new)); 1093 1094 /* XXX: should validate entry */ 1095 new->ape_policy = *newpolicy; 1096 1097 TAILQ_INSERT_TAIL(&V_addrsel_policytab, new, ape_entry); 1098 ADDRSEL_UNLOCK(); 1099 ADDRSEL_XUNLOCK(); 1100 1101 return (0); 1102} 1103 1104static int 1105delete_addrsel_policyent(struct in6_addrpolicy *key) 1106{ 1107 struct addrsel_policyent *pol; 1108 1109 ADDRSEL_XLOCK(); 1110 ADDRSEL_LOCK(); 1111 1112 /* search for the entry in the table */ 1113 TAILQ_FOREACH(pol, &V_addrsel_policytab, ape_entry) { 1114 if (IN6_ARE_ADDR_EQUAL(&key->addr.sin6_addr, 1115 &pol->ape_policy.addr.sin6_addr) && 1116 IN6_ARE_ADDR_EQUAL(&key->addrmask.sin6_addr, 1117 &pol->ape_policy.addrmask.sin6_addr)) { 1118 break; 1119 } 1120 } 1121 if (pol == NULL) { 1122 ADDRSEL_UNLOCK(); 1123 ADDRSEL_XUNLOCK(); 1124 return (ESRCH); 1125 } 1126 1127 TAILQ_REMOVE(&V_addrsel_policytab, pol, ape_entry); 1128 ADDRSEL_UNLOCK(); 1129 ADDRSEL_XUNLOCK(); 1130 1131 return (0); 1132} 1133 1134static int 1135walk_addrsel_policy(int (*callback)(struct in6_addrpolicy *, void *), 1136 void *w) 1137{ 1138 struct addrsel_policyent *pol; 1139 int error = 0; 1140 1141 ADDRSEL_SLOCK(); 1142 TAILQ_FOREACH(pol, &V_addrsel_policytab, ape_entry) { 1143 if ((error = (*callback)(&pol->ape_policy, w)) != 0) { 1144 ADDRSEL_SUNLOCK(); 1145 return (error); 1146 } 1147 } 1148 ADDRSEL_SUNLOCK(); 1149 return (error); 1150} 1151 1152static int 1153dump_addrsel_policyent(struct in6_addrpolicy *pol, void *arg) 1154{ 1155 int error = 0; 1156 struct walkarg *w = arg; 1157 1158 error = SYSCTL_OUT(w->w_req, pol, sizeof(*pol)); 1159 1160 return (error); 1161} 1162 1163static struct in6_addrpolicy * 1164match_addrsel_policy(struct sockaddr_in6 *key) 1165{ 1166 struct addrsel_policyent *pent; 1167 struct in6_addrpolicy *bestpol = NULL, *pol; 1168 int matchlen, bestmatchlen = -1; 1169 u_char *mp, *ep, *k, *p, m; 1170 1171 TAILQ_FOREACH(pent, &V_addrsel_policytab, ape_entry) { 1172 matchlen = 0; 1173 1174 pol = &pent->ape_policy; 1175 mp = (u_char *)&pol->addrmask.sin6_addr; 1176 ep = mp + 16; /* XXX: scope field? */ 1177 k = (u_char *)&key->sin6_addr; 1178 p = (u_char *)&pol->addr.sin6_addr; 1179 for (; mp < ep && *mp; mp++, k++, p++) { 1180 m = *mp; 1181 if ((*k & m) != *p) 1182 goto next; /* not match */ 1183 if (m == 0xff) /* short cut for a typical case */ 1184 matchlen += 8; 1185 else { 1186 while (m >= 0x80) { 1187 matchlen++; 1188 m <<= 1; 1189 } 1190 } 1191 } 1192 1193 /* matched. check if this is better than the current best. */ 1194 if (bestpol == NULL || 1195 matchlen > bestmatchlen) { 1196 bestpol = pol; 1197 bestmatchlen = matchlen; 1198 } 1199 1200 next: 1201 continue; 1202 } 1203 1204 return (bestpol); 1205} 1206