in6_src.c revision 271288
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: stable/10/sys/netinet6/in6_src.c 271288 2014-09-08 19:40:59Z ae $"); 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) 125static VNET_DEFINE(struct in6_addrpolicy, defaultaddrpolicy); 126#define V_defaultaddrpolicy VNET(defaultaddrpolicy) 127 128VNET_DEFINE(int, ip6_prefer_tempaddr) = 0; 129 130static int selectroute(struct sockaddr_in6 *, struct ip6_pktopts *, 131 struct ip6_moptions *, struct route_in6 *, struct ifnet **, 132 struct rtentry **, int, u_int); 133static int in6_selectif(struct sockaddr_in6 *, struct ip6_pktopts *, 134 struct ip6_moptions *, struct route_in6 *ro, struct ifnet **, 135 struct ifnet *, u_int); 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(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 IP6STAT_INC(ip6s_sources_rule[(r)]); \ 155 rule = (r); \ 156 /* { \ 157 char ip6buf[INET6_ADDRSTRLEN], ip6b[INET6_ADDRSTRLEN]; \ 158 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)); \ 159 } */ \ 160 goto replace; \ 161} while(0) 162#define NEXT(r) do {\ 163 /* { \ 164 char ip6buf[INET6_ADDRSTRLEN], ip6b[INET6_ADDRSTRLEN]; \ 165 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)); \ 166 } */ \ 167 goto next; /* XXX: we can't use 'continue' here */ \ 168} while(0) 169#define BREAK(r) do { \ 170 IP6STAT_INC(ip6s_sources_rule[(r)]); \ 171 rule = (r); \ 172 goto out; /* XXX: we can't use 'break' here */ \ 173} while(0) 174 175int 176in6_selectsrc(struct sockaddr_in6 *dstsock, struct ip6_pktopts *opts, 177 struct inpcb *inp, struct route_in6 *ro, struct ucred *cred, 178 struct ifnet **ifpp, struct in6_addr *srcp) 179{ 180 struct in6_addr dst, tmp; 181 struct ifnet *ifp = NULL, *oifp = NULL; 182 struct in6_ifaddr *ia = NULL, *ia_best = NULL; 183 struct in6_pktinfo *pi = NULL; 184 int dst_scope = -1, best_scope = -1, best_matchlen = -1; 185 struct in6_addrpolicy *dst_policy = NULL, *best_policy = NULL; 186 u_int32_t odstzone; 187 int prefer_tempaddr; 188 int error, rule; 189 struct ip6_moptions *mopts; 190 191 KASSERT(srcp != NULL, ("%s: srcp is NULL", __func__)); 192 193 dst = dstsock->sin6_addr; /* make a copy for local operation */ 194 if (ifpp) { 195 /* 196 * Save a possibly passed in ifp for in6_selectsrc. Only 197 * neighbor discovery code should use this feature, where 198 * we may know the interface but not the FIB number holding 199 * the connected subnet in case someone deleted it from the 200 * default FIB and we need to check the interface. 201 */ 202 if (*ifpp != NULL) 203 oifp = *ifpp; 204 *ifpp = NULL; 205 } 206 207 if (inp != NULL) { 208 INP_LOCK_ASSERT(inp); 209 mopts = inp->in6p_moptions; 210 } else { 211 mopts = NULL; 212 } 213 214 /* 215 * If the source address is explicitly specified by the caller, 216 * check if the requested source address is indeed a unicast address 217 * assigned to the node, and can be used as the packet's source 218 * address. If everything is okay, use the address as source. 219 */ 220 if (opts && (pi = opts->ip6po_pktinfo) && 221 !IN6_IS_ADDR_UNSPECIFIED(&pi->ipi6_addr)) { 222 struct sockaddr_in6 srcsock; 223 struct in6_ifaddr *ia6; 224 225 /* get the outgoing interface */ 226 if ((error = in6_selectif(dstsock, opts, mopts, ro, &ifp, oifp, 227 (inp != NULL) ? inp->inp_inc.inc_fibnum : RT_DEFAULT_FIB)) 228 != 0) 229 return (error); 230 231 /* 232 * determine the appropriate zone id of the source based on 233 * the zone of the destination and the outgoing interface. 234 * If the specified address is ambiguous wrt the scope zone, 235 * the interface must be specified; otherwise, ifa_ifwithaddr() 236 * will fail matching the address. 237 */ 238 bzero(&srcsock, sizeof(srcsock)); 239 srcsock.sin6_family = AF_INET6; 240 srcsock.sin6_len = sizeof(srcsock); 241 srcsock.sin6_addr = pi->ipi6_addr; 242 if (ifp) { 243 error = in6_setscope(&srcsock.sin6_addr, ifp, NULL); 244 if (error) 245 return (error); 246 } 247 if (cred != NULL && (error = prison_local_ip6(cred, 248 &srcsock.sin6_addr, (inp != NULL && 249 (inp->inp_flags & IN6P_IPV6_V6ONLY) != 0))) != 0) 250 return (error); 251 252 ia6 = (struct in6_ifaddr *)ifa_ifwithaddr( 253 (struct sockaddr *)&srcsock); 254 if (ia6 == NULL || 255 (ia6->ia6_flags & (IN6_IFF_ANYCAST | IN6_IFF_NOTREADY))) { 256 if (ia6 != NULL) 257 ifa_free(&ia6->ia_ifa); 258 return (EADDRNOTAVAIL); 259 } 260 pi->ipi6_addr = srcsock.sin6_addr; /* XXX: this overrides pi */ 261 if (ifpp) 262 *ifpp = ifp; 263 bcopy(&ia6->ia_addr.sin6_addr, srcp, sizeof(*srcp)); 264 ifa_free(&ia6->ia_ifa); 265 return (0); 266 } 267 268 /* 269 * Otherwise, if the socket has already bound the source, just use it. 270 */ 271 if (inp != NULL && !IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) { 272 if (cred != NULL && 273 (error = prison_local_ip6(cred, &inp->in6p_laddr, 274 ((inp->inp_flags & IN6P_IPV6_V6ONLY) != 0))) != 0) 275 return (error); 276 bcopy(&inp->in6p_laddr, srcp, sizeof(*srcp)); 277 return (0); 278 } 279 280 /* 281 * Bypass source address selection and use the primary jail IP 282 * if requested. 283 */ 284 if (cred != NULL && !prison_saddrsel_ip6(cred, srcp)) 285 return (0); 286 287 /* 288 * If the address is not specified, choose the best one based on 289 * the outgoing interface and the destination address. 290 */ 291 /* get the outgoing interface */ 292 if ((error = in6_selectif(dstsock, opts, mopts, ro, &ifp, oifp, 293 (inp != NULL) ? inp->inp_inc.inc_fibnum : RT_DEFAULT_FIB)) != 0) 294 return (error); 295 296#ifdef DIAGNOSTIC 297 if (ifp == NULL) /* this should not happen */ 298 panic("in6_selectsrc: NULL ifp"); 299#endif 300 error = in6_setscope(&dst, ifp, &odstzone); 301 if (error) 302 return (error); 303 304 rule = 0; 305 IN6_IFADDR_RLOCK(); 306 TAILQ_FOREACH(ia, &V_in6_ifaddrhead, ia_link) { 307 int new_scope = -1, new_matchlen = -1; 308 struct in6_addrpolicy *new_policy = NULL; 309 u_int32_t srczone, osrczone, dstzone; 310 struct in6_addr src; 311 struct ifnet *ifp1 = ia->ia_ifp; 312 313 /* 314 * We'll never take an address that breaks the scope zone 315 * of the destination. We also skip an address if its zone 316 * does not contain the outgoing interface. 317 * XXX: we should probably use sin6_scope_id here. 318 */ 319 if (in6_setscope(&dst, ifp1, &dstzone) || 320 odstzone != dstzone) { 321 continue; 322 } 323 src = ia->ia_addr.sin6_addr; 324 if (in6_setscope(&src, ifp, &osrczone) || 325 in6_setscope(&src, ifp1, &srczone) || 326 osrczone != srczone) { 327 continue; 328 } 329 330 /* avoid unusable addresses */ 331 if ((ia->ia6_flags & 332 (IN6_IFF_NOTREADY | IN6_IFF_ANYCAST | IN6_IFF_DETACHED))) { 333 continue; 334 } 335 if (!V_ip6_use_deprecated && IFA6_IS_DEPRECATED(ia)) 336 continue; 337 338 /* If jailed only take addresses of the jail into account. */ 339 if (cred != NULL && 340 prison_check_ip6(cred, &ia->ia_addr.sin6_addr) != 0) 341 continue; 342 343 /* Rule 1: Prefer same address */ 344 if (IN6_ARE_ADDR_EQUAL(&dst, &ia->ia_addr.sin6_addr)) { 345 ia_best = ia; 346 BREAK(1); /* there should be no better candidate */ 347 } 348 349 if (ia_best == NULL) 350 REPLACE(0); 351 352 /* Rule 2: Prefer appropriate scope */ 353 if (dst_scope < 0) 354 dst_scope = in6_addrscope(&dst); 355 new_scope = in6_addrscope(&ia->ia_addr.sin6_addr); 356 if (IN6_ARE_SCOPE_CMP(best_scope, new_scope) < 0) { 357 if (IN6_ARE_SCOPE_CMP(best_scope, dst_scope) < 0) 358 REPLACE(2); 359 NEXT(2); 360 } else if (IN6_ARE_SCOPE_CMP(new_scope, best_scope) < 0) { 361 if (IN6_ARE_SCOPE_CMP(new_scope, dst_scope) < 0) 362 NEXT(2); 363 REPLACE(2); 364 } 365 366 /* 367 * Rule 3: Avoid deprecated addresses. Note that the case of 368 * !ip6_use_deprecated is already rejected above. 369 */ 370 if (!IFA6_IS_DEPRECATED(ia_best) && IFA6_IS_DEPRECATED(ia)) 371 NEXT(3); 372 if (IFA6_IS_DEPRECATED(ia_best) && !IFA6_IS_DEPRECATED(ia)) 373 REPLACE(3); 374 375 /* Rule 4: Prefer home addresses */ 376 /* 377 * XXX: This is a TODO. We should probably merge the MIP6 378 * case above. 379 */ 380 381 /* Rule 5: Prefer outgoing interface */ 382 if (!(ND_IFINFO(ifp)->flags & ND6_IFF_NO_PREFER_IFACE)) { 383 if (ia_best->ia_ifp == ifp && ia->ia_ifp != ifp) 384 NEXT(5); 385 if (ia_best->ia_ifp != ifp && ia->ia_ifp == ifp) 386 REPLACE(5); 387 } 388 389 /* 390 * Rule 6: Prefer matching label 391 * Note that best_policy should be non-NULL here. 392 */ 393 if (dst_policy == NULL) 394 dst_policy = lookup_addrsel_policy(dstsock); 395 if (dst_policy->label != ADDR_LABEL_NOTAPP) { 396 new_policy = lookup_addrsel_policy(&ia->ia_addr); 397 if (dst_policy->label == best_policy->label && 398 dst_policy->label != new_policy->label) 399 NEXT(6); 400 if (dst_policy->label != best_policy->label && 401 dst_policy->label == new_policy->label) 402 REPLACE(6); 403 } 404 405 /* 406 * Rule 7: Prefer public addresses. 407 * We allow users to reverse the logic by configuring 408 * a sysctl variable, so that privacy conscious users can 409 * always prefer temporary addresses. 410 */ 411 if (opts == NULL || 412 opts->ip6po_prefer_tempaddr == IP6PO_TEMPADDR_SYSTEM) { 413 prefer_tempaddr = V_ip6_prefer_tempaddr; 414 } else if (opts->ip6po_prefer_tempaddr == 415 IP6PO_TEMPADDR_NOTPREFER) { 416 prefer_tempaddr = 0; 417 } else 418 prefer_tempaddr = 1; 419 if (!(ia_best->ia6_flags & IN6_IFF_TEMPORARY) && 420 (ia->ia6_flags & IN6_IFF_TEMPORARY)) { 421 if (prefer_tempaddr) 422 REPLACE(7); 423 else 424 NEXT(7); 425 } 426 if ((ia_best->ia6_flags & IN6_IFF_TEMPORARY) && 427 !(ia->ia6_flags & IN6_IFF_TEMPORARY)) { 428 if (prefer_tempaddr) 429 NEXT(7); 430 else 431 REPLACE(7); 432 } 433 434 /* 435 * Rule 8: prefer addresses on alive interfaces. 436 * This is a KAME specific rule. 437 */ 438 if ((ia_best->ia_ifp->if_flags & IFF_UP) && 439 !(ia->ia_ifp->if_flags & IFF_UP)) 440 NEXT(8); 441 if (!(ia_best->ia_ifp->if_flags & IFF_UP) && 442 (ia->ia_ifp->if_flags & IFF_UP)) 443 REPLACE(8); 444 445 /* 446 * Rule 9: prefer address with better virtual status. 447 */ 448 if (ifa_preferred(&ia_best->ia_ifa, &ia->ia_ifa)) 449 REPLACE(9); 450 if (ifa_preferred(&ia->ia_ifa, &ia_best->ia_ifa)) 451 NEXT(9); 452 453 /* 454 * Rule 14: Use longest matching prefix. 455 * Note: in the address selection draft, this rule is 456 * documented as "Rule 8". However, since it is also 457 * documented that this rule can be overridden, we assign 458 * a large number so that it is easy to assign smaller numbers 459 * to more preferred rules. 460 */ 461 new_matchlen = in6_matchlen(&ia->ia_addr.sin6_addr, &dst); 462 if (best_matchlen < new_matchlen) 463 REPLACE(14); 464 if (new_matchlen < best_matchlen) 465 NEXT(14); 466 467 /* Rule 15 is reserved. */ 468 469 /* 470 * Last resort: just keep the current candidate. 471 * Or, do we need more rules? 472 */ 473 continue; 474 475 replace: 476 ia_best = ia; 477 best_scope = (new_scope >= 0 ? new_scope : 478 in6_addrscope(&ia_best->ia_addr.sin6_addr)); 479 best_policy = (new_policy ? new_policy : 480 lookup_addrsel_policy(&ia_best->ia_addr)); 481 best_matchlen = (new_matchlen >= 0 ? new_matchlen : 482 in6_matchlen(&ia_best->ia_addr.sin6_addr, 483 &dst)); 484 485 next: 486 continue; 487 488 out: 489 break; 490 } 491 492 if ((ia = ia_best) == NULL) { 493 IN6_IFADDR_RUNLOCK(); 494 IP6STAT_INC(ip6s_sources_none); 495 return (EADDRNOTAVAIL); 496 } 497 498 /* 499 * At this point at least one of the addresses belonged to the jail 500 * but it could still be, that we want to further restrict it, e.g. 501 * theoratically IN6_IS_ADDR_LOOPBACK. 502 * It must not be IN6_IS_ADDR_UNSPECIFIED anymore. 503 * prison_local_ip6() will fix an IN6_IS_ADDR_LOOPBACK but should 504 * let all others previously selected pass. 505 * Use tmp to not change ::1 on lo0 to the primary jail address. 506 */ 507 tmp = ia->ia_addr.sin6_addr; 508 if (cred != NULL && prison_local_ip6(cred, &tmp, (inp != NULL && 509 (inp->inp_flags & IN6P_IPV6_V6ONLY) != 0)) != 0) { 510 IN6_IFADDR_RUNLOCK(); 511 IP6STAT_INC(ip6s_sources_none); 512 return (EADDRNOTAVAIL); 513 } 514 515 if (ifpp) 516 *ifpp = ifp; 517 518 bcopy(&tmp, srcp, sizeof(*srcp)); 519 if (ia->ia_ifp == ifp) 520 IP6STAT_INC(ip6s_sources_sameif[best_scope]); 521 else 522 IP6STAT_INC(ip6s_sources_otherif[best_scope]); 523 if (dst_scope == best_scope) 524 IP6STAT_INC(ip6s_sources_samescope[best_scope]); 525 else 526 IP6STAT_INC(ip6s_sources_otherscope[best_scope]); 527 if (IFA6_IS_DEPRECATED(ia)) 528 IP6STAT_INC(ip6s_sources_deprecated[best_scope]); 529 IN6_IFADDR_RUNLOCK(); 530 return (0); 531} 532 533/* 534 * clone - meaningful only for bsdi and freebsd 535 */ 536static int 537selectroute(struct sockaddr_in6 *dstsock, struct ip6_pktopts *opts, 538 struct ip6_moptions *mopts, struct route_in6 *ro, 539 struct ifnet **retifp, struct rtentry **retrt, int norouteok, u_int fibnum) 540{ 541 int error = 0; 542 struct ifnet *ifp = NULL; 543 struct rtentry *rt = NULL; 544 struct sockaddr_in6 *sin6_next; 545 struct in6_pktinfo *pi = NULL; 546 struct in6_addr *dst = &dstsock->sin6_addr; 547#if 0 548 char ip6buf[INET6_ADDRSTRLEN]; 549 550 if (dstsock->sin6_addr.s6_addr32[0] == 0 && 551 dstsock->sin6_addr.s6_addr32[1] == 0 && 552 !IN6_IS_ADDR_LOOPBACK(&dstsock->sin6_addr)) { 553 printf("in6_selectroute: strange destination %s\n", 554 ip6_sprintf(ip6buf, &dstsock->sin6_addr)); 555 } else { 556 printf("in6_selectroute: destination = %s%%%d\n", 557 ip6_sprintf(ip6buf, &dstsock->sin6_addr), 558 dstsock->sin6_scope_id); /* for debug */ 559 } 560#endif 561 562 /* If the caller specify the outgoing interface explicitly, use it. */ 563 if (opts && (pi = opts->ip6po_pktinfo) != NULL && pi->ipi6_ifindex) { 564 /* XXX boundary check is assumed to be already done. */ 565 ifp = ifnet_byindex(pi->ipi6_ifindex); 566 if (ifp != NULL && 567 (norouteok || retrt == NULL || 568 IN6_IS_ADDR_MULTICAST(dst))) { 569 /* 570 * we do not have to check or get the route for 571 * multicast. 572 */ 573 goto done; 574 } else 575 goto getroute; 576 } 577 578 /* 579 * If the destination address is a multicast address and the outgoing 580 * interface for the address is specified by the caller, use it. 581 */ 582 if (IN6_IS_ADDR_MULTICAST(dst) && 583 mopts != NULL && (ifp = mopts->im6o_multicast_ifp) != NULL) { 584 goto done; /* we do not need a route for multicast. */ 585 } 586 587 getroute: 588 /* 589 * If the next hop address for the packet is specified by the caller, 590 * use it as the gateway. 591 */ 592 if (opts && opts->ip6po_nexthop) { 593 struct route_in6 *ron; 594 struct llentry *la; 595 596 sin6_next = satosin6(opts->ip6po_nexthop); 597 598 /* at this moment, we only support AF_INET6 next hops */ 599 if (sin6_next->sin6_family != AF_INET6) { 600 error = EAFNOSUPPORT; /* or should we proceed? */ 601 goto done; 602 } 603 604 /* 605 * If the next hop is an IPv6 address, then the node identified 606 * by that address must be a neighbor of the sending host. 607 */ 608 ron = &opts->ip6po_nextroute; 609 /* 610 * XXX what do we do here? 611 * PLZ to be fixing 612 */ 613 614 615 if (ron->ro_rt == NULL) { 616 in6_rtalloc(ron, fibnum); /* multi path case? */ 617 if (ron->ro_rt == NULL) { 618 /* XXX-BZ WT.? */ 619 if (ron->ro_rt) { 620 RTFREE(ron->ro_rt); 621 ron->ro_rt = NULL; 622 } 623 error = EHOSTUNREACH; 624 goto done; 625 } 626 } 627 628 rt = ron->ro_rt; 629 ifp = rt->rt_ifp; 630 IF_AFDATA_RLOCK(ifp); 631 la = lla_lookup(LLTABLE6(ifp), 0, (struct sockaddr *)&sin6_next->sin6_addr); 632 IF_AFDATA_RUNLOCK(ifp); 633 if (la != NULL) 634 LLE_RUNLOCK(la); 635 else { 636 error = EHOSTUNREACH; 637 goto done; 638 } 639#if 0 640 if ((ron->ro_rt && 641 (ron->ro_rt->rt_flags & (RTF_UP | RTF_LLINFO)) != 642 (RTF_UP | RTF_LLINFO)) || 643 !IN6_ARE_ADDR_EQUAL(&satosin6(&ron->ro_dst)->sin6_addr, 644 &sin6_next->sin6_addr)) { 645 if (ron->ro_rt) { 646 RTFREE(ron->ro_rt); 647 ron->ro_rt = NULL; 648 } 649 *satosin6(&ron->ro_dst) = *sin6_next; 650 } 651 if (ron->ro_rt == NULL) { 652 in6_rtalloc(ron, fibnum); /* multi path case? */ 653 if (ron->ro_rt == NULL || 654 !(ron->ro_rt->rt_flags & RTF_LLINFO)) { 655 if (ron->ro_rt) { 656 RTFREE(ron->ro_rt); 657 ron->ro_rt = NULL; 658 } 659 error = EHOSTUNREACH; 660 goto done; 661 } 662 } 663#endif 664 665 /* 666 * When cloning is required, try to allocate a route to the 667 * destination so that the caller can store path MTU 668 * information. 669 */ 670 goto done; 671 } 672 673 /* 674 * Use a cached route if it exists and is valid, else try to allocate 675 * a new one. Note that we should check the address family of the 676 * cached destination, in case of sharing the cache with IPv4. 677 */ 678 if (ro) { 679 if (ro->ro_rt && 680 (!(ro->ro_rt->rt_flags & RTF_UP) || 681 ((struct sockaddr *)(&ro->ro_dst))->sa_family != AF_INET6 || 682 !IN6_ARE_ADDR_EQUAL(&satosin6(&ro->ro_dst)->sin6_addr, 683 dst))) { 684 RTFREE(ro->ro_rt); 685 ro->ro_rt = (struct rtentry *)NULL; 686 } 687 if (ro->ro_rt == (struct rtentry *)NULL) { 688 struct sockaddr_in6 *sa6; 689 690 /* No route yet, so try to acquire one */ 691 bzero(&ro->ro_dst, sizeof(struct sockaddr_in6)); 692 sa6 = (struct sockaddr_in6 *)&ro->ro_dst; 693 *sa6 = *dstsock; 694 sa6->sin6_scope_id = 0; 695 696#ifdef RADIX_MPATH 697 rtalloc_mpath_fib((struct route *)ro, 698 ntohl(sa6->sin6_addr.s6_addr32[3]), fibnum); 699#else 700 ro->ro_rt = in6_rtalloc1((struct sockaddr *) 701 &ro->ro_dst, 0, 0UL, fibnum); 702 if (ro->ro_rt) 703 RT_UNLOCK(ro->ro_rt); 704#endif 705 } 706 707 /* 708 * do not care about the result if we have the nexthop 709 * explicitly specified. 710 */ 711 if (opts && opts->ip6po_nexthop) 712 goto done; 713 714 if (ro->ro_rt) { 715 ifp = ro->ro_rt->rt_ifp; 716 717 if (ifp == NULL) { /* can this really happen? */ 718 RTFREE(ro->ro_rt); 719 ro->ro_rt = NULL; 720 } 721 } 722 if (ro->ro_rt == NULL) 723 error = EHOSTUNREACH; 724 rt = ro->ro_rt; 725 726 /* 727 * Check if the outgoing interface conflicts with 728 * the interface specified by ipi6_ifindex (if specified). 729 * Note that loopback interface is always okay. 730 * (this may happen when we are sending a packet to one of 731 * our own addresses.) 732 */ 733 if (ifp && opts && opts->ip6po_pktinfo && 734 opts->ip6po_pktinfo->ipi6_ifindex) { 735 if (!(ifp->if_flags & IFF_LOOPBACK) && 736 ifp->if_index != 737 opts->ip6po_pktinfo->ipi6_ifindex) { 738 error = EHOSTUNREACH; 739 goto done; 740 } 741 } 742 } 743 744 done: 745 if (ifp == NULL && rt == NULL) { 746 /* 747 * This can happen if the caller did not pass a cached route 748 * nor any other hints. We treat this case an error. 749 */ 750 error = EHOSTUNREACH; 751 } 752 if (error == EHOSTUNREACH) 753 IP6STAT_INC(ip6s_noroute); 754 755 if (retifp != NULL) { 756 *retifp = ifp; 757 758 /* 759 * Adjust the "outgoing" interface. If we're going to loop 760 * the packet back to ourselves, the ifp would be the loopback 761 * interface. However, we'd rather know the interface associated 762 * to the destination address (which should probably be one of 763 * our own addresses.) 764 */ 765 if (rt) { 766 if ((rt->rt_ifp->if_flags & IFF_LOOPBACK) && 767 (rt->rt_gateway->sa_family == AF_LINK)) 768 *retifp = 769 ifnet_byindex(((struct sockaddr_dl *) 770 rt->rt_gateway)->sdl_index); 771 } 772 } 773 774 if (retrt != NULL) 775 *retrt = rt; /* rt may be NULL */ 776 777 return (error); 778} 779 780static int 781in6_selectif(struct sockaddr_in6 *dstsock, struct ip6_pktopts *opts, 782 struct ip6_moptions *mopts, struct route_in6 *ro, struct ifnet **retifp, 783 struct ifnet *oifp, u_int fibnum) 784{ 785 int error; 786 struct route_in6 sro; 787 struct rtentry *rt = NULL; 788 789 KASSERT(retifp != NULL, ("%s: retifp is NULL", __func__)); 790 791 if (ro == NULL) { 792 bzero(&sro, sizeof(sro)); 793 ro = &sro; 794 } 795 796 if ((error = selectroute(dstsock, opts, mopts, ro, retifp, 797 &rt, 1, fibnum)) != 0) { 798 if (ro == &sro && rt && rt == sro.ro_rt) 799 RTFREE(rt); 800 /* Help ND. See oifp comment in in6_selectsrc(). */ 801 if (oifp != NULL && fibnum == RT_DEFAULT_FIB) { 802 *retifp = oifp; 803 error = 0; 804 } 805 return (error); 806 } 807 808 /* 809 * do not use a rejected or black hole route. 810 * XXX: this check should be done in the L2 output routine. 811 * However, if we skipped this check here, we'd see the following 812 * scenario: 813 * - install a rejected route for a scoped address prefix 814 * (like fe80::/10) 815 * - send a packet to a destination that matches the scoped prefix, 816 * with ambiguity about the scope zone. 817 * - pick the outgoing interface from the route, and disambiguate the 818 * scope zone with the interface. 819 * - ip6_output() would try to get another route with the "new" 820 * destination, which may be valid. 821 * - we'd see no error on output. 822 * Although this may not be very harmful, it should still be confusing. 823 * We thus reject the case here. 824 */ 825 if (rt && (rt->rt_flags & (RTF_REJECT | RTF_BLACKHOLE))) { 826 int flags = (rt->rt_flags & RTF_HOST ? EHOSTUNREACH : ENETUNREACH); 827 828 if (ro == &sro && rt && rt == sro.ro_rt) 829 RTFREE(rt); 830 return (flags); 831 } 832 833 if (ro == &sro && rt && rt == sro.ro_rt) 834 RTFREE(rt); 835 return (0); 836} 837 838/* 839 * Public wrapper function to selectroute(). 840 * 841 * XXX-BZ in6_selectroute() should and will grow the FIB argument. The 842 * in6_selectroute_fib() function is only there for backward compat on stable. 843 */ 844int 845in6_selectroute(struct sockaddr_in6 *dstsock, struct ip6_pktopts *opts, 846 struct ip6_moptions *mopts, struct route_in6 *ro, 847 struct ifnet **retifp, struct rtentry **retrt) 848{ 849 850 return (selectroute(dstsock, opts, mopts, ro, retifp, 851 retrt, 0, RT_DEFAULT_FIB)); 852} 853 854#ifndef BURN_BRIDGES 855int 856in6_selectroute_fib(struct sockaddr_in6 *dstsock, struct ip6_pktopts *opts, 857 struct ip6_moptions *mopts, struct route_in6 *ro, 858 struct ifnet **retifp, struct rtentry **retrt, u_int fibnum) 859{ 860 861 return (selectroute(dstsock, opts, mopts, ro, retifp, 862 retrt, 0, fibnum)); 863} 864#endif 865 866/* 867 * Default hop limit selection. The precedence is as follows: 868 * 1. Hoplimit value specified via ioctl. 869 * 2. (If the outgoing interface is detected) the current 870 * hop limit of the interface specified by router advertisement. 871 * 3. The system default hoplimit. 872 */ 873int 874in6_selecthlim(struct inpcb *in6p, struct ifnet *ifp) 875{ 876 877 if (in6p && in6p->in6p_hops >= 0) 878 return (in6p->in6p_hops); 879 else if (ifp) 880 return (ND_IFINFO(ifp)->chlim); 881 else if (in6p && !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr)) { 882 struct route_in6 ro6; 883 struct ifnet *lifp; 884 885 bzero(&ro6, sizeof(ro6)); 886 ro6.ro_dst.sin6_family = AF_INET6; 887 ro6.ro_dst.sin6_len = sizeof(struct sockaddr_in6); 888 ro6.ro_dst.sin6_addr = in6p->in6p_faddr; 889 in6_rtalloc(&ro6, in6p->inp_inc.inc_fibnum); 890 if (ro6.ro_rt) { 891 lifp = ro6.ro_rt->rt_ifp; 892 RTFREE(ro6.ro_rt); 893 if (lifp) 894 return (ND_IFINFO(lifp)->chlim); 895 } 896 } 897 return (V_ip6_defhlim); 898} 899 900/* 901 * XXX: this is borrowed from in6_pcbbind(). If possible, we should 902 * share this function by all *bsd*... 903 */ 904int 905in6_pcbsetport(struct in6_addr *laddr, struct inpcb *inp, struct ucred *cred) 906{ 907 struct socket *so = inp->inp_socket; 908 u_int16_t lport = 0; 909 int error, lookupflags = 0; 910#ifdef INVARIANTS 911 struct inpcbinfo *pcbinfo = inp->inp_pcbinfo; 912#endif 913 914 INP_WLOCK_ASSERT(inp); 915 INP_HASH_WLOCK_ASSERT(pcbinfo); 916 917 error = prison_local_ip6(cred, laddr, 918 ((inp->inp_flags & IN6P_IPV6_V6ONLY) != 0)); 919 if (error) 920 return(error); 921 922 /* XXX: this is redundant when called from in6_pcbbind */ 923 if ((so->so_options & (SO_REUSEADDR|SO_REUSEPORT)) == 0) 924 lookupflags = INPLOOKUP_WILDCARD; 925 926 inp->inp_flags |= INP_ANONPORT; 927 928 error = in_pcb_lport(inp, NULL, &lport, cred, lookupflags); 929 if (error != 0) 930 return (error); 931 932 inp->inp_lport = lport; 933 if (in_pcbinshash(inp) != 0) { 934 inp->in6p_laddr = in6addr_any; 935 inp->inp_lport = 0; 936 return (EAGAIN); 937 } 938 939 return (0); 940} 941 942void 943addrsel_policy_init(void) 944{ 945 946 init_policy_queue(); 947 948 /* initialize the "last resort" policy */ 949 bzero(&V_defaultaddrpolicy, sizeof(V_defaultaddrpolicy)); 950 V_defaultaddrpolicy.label = ADDR_LABEL_NOTAPP; 951 952 if (!IS_DEFAULT_VNET(curvnet)) 953 return; 954 955 ADDRSEL_LOCK_INIT(); 956 ADDRSEL_SXLOCK_INIT(); 957} 958 959static struct in6_addrpolicy * 960lookup_addrsel_policy(struct sockaddr_in6 *key) 961{ 962 struct in6_addrpolicy *match = NULL; 963 964 ADDRSEL_LOCK(); 965 match = match_addrsel_policy(key); 966 967 if (match == NULL) 968 match = &V_defaultaddrpolicy; 969 else 970 match->use++; 971 ADDRSEL_UNLOCK(); 972 973 return (match); 974} 975 976/* 977 * Subroutines to manage the address selection policy table via sysctl. 978 */ 979struct walkarg { 980 struct sysctl_req *w_req; 981}; 982 983static int in6_src_sysctl(SYSCTL_HANDLER_ARGS); 984SYSCTL_DECL(_net_inet6_ip6); 985static SYSCTL_NODE(_net_inet6_ip6, IPV6CTL_ADDRCTLPOLICY, addrctlpolicy, 986 CTLFLAG_RD, in6_src_sysctl, ""); 987 988static int 989in6_src_sysctl(SYSCTL_HANDLER_ARGS) 990{ 991 struct walkarg w; 992 993 if (req->newptr) 994 return EPERM; 995 996 bzero(&w, sizeof(w)); 997 w.w_req = req; 998 999 return (walk_addrsel_policy(dump_addrsel_policyent, &w)); 1000} 1001 1002int 1003in6_src_ioctl(u_long cmd, caddr_t data) 1004{ 1005 int i; 1006 struct in6_addrpolicy ent0; 1007 1008 if (cmd != SIOCAADDRCTL_POLICY && cmd != SIOCDADDRCTL_POLICY) 1009 return (EOPNOTSUPP); /* check for safety */ 1010 1011 ent0 = *(struct in6_addrpolicy *)data; 1012 1013 if (ent0.label == ADDR_LABEL_NOTAPP) 1014 return (EINVAL); 1015 /* check if the prefix mask is consecutive. */ 1016 if (in6_mask2len(&ent0.addrmask.sin6_addr, NULL) < 0) 1017 return (EINVAL); 1018 /* clear trailing garbages (if any) of the prefix address. */ 1019 for (i = 0; i < 4; i++) { 1020 ent0.addr.sin6_addr.s6_addr32[i] &= 1021 ent0.addrmask.sin6_addr.s6_addr32[i]; 1022 } 1023 ent0.use = 0; 1024 1025 switch (cmd) { 1026 case SIOCAADDRCTL_POLICY: 1027 return (add_addrsel_policyent(&ent0)); 1028 case SIOCDADDRCTL_POLICY: 1029 return (delete_addrsel_policyent(&ent0)); 1030 } 1031 1032 return (0); /* XXX: compromise compilers */ 1033} 1034 1035/* 1036 * The followings are implementation of the policy table using a 1037 * simple tail queue. 1038 * XXX such details should be hidden. 1039 * XXX implementation using binary tree should be more efficient. 1040 */ 1041struct addrsel_policyent { 1042 TAILQ_ENTRY(addrsel_policyent) ape_entry; 1043 struct in6_addrpolicy ape_policy; 1044}; 1045 1046TAILQ_HEAD(addrsel_policyhead, addrsel_policyent); 1047 1048static VNET_DEFINE(struct addrsel_policyhead, addrsel_policytab); 1049#define V_addrsel_policytab VNET(addrsel_policytab) 1050 1051static void 1052init_policy_queue(void) 1053{ 1054 1055 TAILQ_INIT(&V_addrsel_policytab); 1056} 1057 1058static int 1059add_addrsel_policyent(struct in6_addrpolicy *newpolicy) 1060{ 1061 struct addrsel_policyent *new, *pol; 1062 1063 new = malloc(sizeof(*new), M_IFADDR, 1064 M_WAITOK); 1065 ADDRSEL_XLOCK(); 1066 ADDRSEL_LOCK(); 1067 1068 /* duplication check */ 1069 TAILQ_FOREACH(pol, &V_addrsel_policytab, ape_entry) { 1070 if (IN6_ARE_ADDR_EQUAL(&newpolicy->addr.sin6_addr, 1071 &pol->ape_policy.addr.sin6_addr) && 1072 IN6_ARE_ADDR_EQUAL(&newpolicy->addrmask.sin6_addr, 1073 &pol->ape_policy.addrmask.sin6_addr)) { 1074 ADDRSEL_UNLOCK(); 1075 ADDRSEL_XUNLOCK(); 1076 free(new, M_IFADDR); 1077 return (EEXIST); /* or override it? */ 1078 } 1079 } 1080 1081 bzero(new, sizeof(*new)); 1082 1083 /* XXX: should validate entry */ 1084 new->ape_policy = *newpolicy; 1085 1086 TAILQ_INSERT_TAIL(&V_addrsel_policytab, new, ape_entry); 1087 ADDRSEL_UNLOCK(); 1088 ADDRSEL_XUNLOCK(); 1089 1090 return (0); 1091} 1092 1093static int 1094delete_addrsel_policyent(struct in6_addrpolicy *key) 1095{ 1096 struct addrsel_policyent *pol; 1097 1098 ADDRSEL_XLOCK(); 1099 ADDRSEL_LOCK(); 1100 1101 /* search for the entry in the table */ 1102 TAILQ_FOREACH(pol, &V_addrsel_policytab, ape_entry) { 1103 if (IN6_ARE_ADDR_EQUAL(&key->addr.sin6_addr, 1104 &pol->ape_policy.addr.sin6_addr) && 1105 IN6_ARE_ADDR_EQUAL(&key->addrmask.sin6_addr, 1106 &pol->ape_policy.addrmask.sin6_addr)) { 1107 break; 1108 } 1109 } 1110 if (pol == NULL) { 1111 ADDRSEL_UNLOCK(); 1112 ADDRSEL_XUNLOCK(); 1113 return (ESRCH); 1114 } 1115 1116 TAILQ_REMOVE(&V_addrsel_policytab, pol, ape_entry); 1117 ADDRSEL_UNLOCK(); 1118 ADDRSEL_XUNLOCK(); 1119 free(pol, M_IFADDR); 1120 1121 return (0); 1122} 1123 1124static int 1125walk_addrsel_policy(int (*callback)(struct in6_addrpolicy *, void *), void *w) 1126{ 1127 struct addrsel_policyent *pol; 1128 int error = 0; 1129 1130 ADDRSEL_SLOCK(); 1131 TAILQ_FOREACH(pol, &V_addrsel_policytab, ape_entry) { 1132 if ((error = (*callback)(&pol->ape_policy, w)) != 0) { 1133 ADDRSEL_SUNLOCK(); 1134 return (error); 1135 } 1136 } 1137 ADDRSEL_SUNLOCK(); 1138 return (error); 1139} 1140 1141static int 1142dump_addrsel_policyent(struct in6_addrpolicy *pol, void *arg) 1143{ 1144 int error = 0; 1145 struct walkarg *w = arg; 1146 1147 error = SYSCTL_OUT(w->w_req, pol, sizeof(*pol)); 1148 1149 return (error); 1150} 1151 1152static struct in6_addrpolicy * 1153match_addrsel_policy(struct sockaddr_in6 *key) 1154{ 1155 struct addrsel_policyent *pent; 1156 struct in6_addrpolicy *bestpol = NULL, *pol; 1157 int matchlen, bestmatchlen = -1; 1158 u_char *mp, *ep, *k, *p, m; 1159 1160 TAILQ_FOREACH(pent, &V_addrsel_policytab, ape_entry) { 1161 matchlen = 0; 1162 1163 pol = &pent->ape_policy; 1164 mp = (u_char *)&pol->addrmask.sin6_addr; 1165 ep = mp + 16; /* XXX: scope field? */ 1166 k = (u_char *)&key->sin6_addr; 1167 p = (u_char *)&pol->addr.sin6_addr; 1168 for (; mp < ep && *mp; mp++, k++, p++) { 1169 m = *mp; 1170 if ((*k & m) != *p) 1171 goto next; /* not match */ 1172 if (m == 0xff) /* short cut for a typical case */ 1173 matchlen += 8; 1174 else { 1175 while (m >= 0x80) { 1176 matchlen++; 1177 m <<= 1; 1178 } 1179 } 1180 } 1181 1182 /* matched. check if this is better than the current best. */ 1183 if (bestpol == NULL || 1184 matchlen > bestmatchlen) { 1185 bestpol = pol; 1186 bestmatchlen = matchlen; 1187 } 1188 1189 next: 1190 continue; 1191 } 1192 1193 return (bestpol); 1194} 1195