in6_src.c revision 194777
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 194777 2009-06-23 22:08:55Z 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#include <sys/vimage.h> 87 88#include <net/if.h> 89#include <net/route.h> 90#include <net/if_llatbl.h> 91#ifdef RADIX_MPATH 92#include <net/radix_mpath.h> 93#endif 94 95#include <netinet/in.h> 96#include <netinet/in_var.h> 97#include <netinet/in_systm.h> 98#include <netinet/ip.h> 99#include <netinet/in_pcb.h> 100#include <netinet/ip_var.h> 101#include <netinet/udp.h> 102#include <netinet/udp_var.h> 103#include <netinet/vinet.h> 104 105#include <netinet6/in6_var.h> 106#include <netinet/ip6.h> 107#include <netinet6/in6_pcb.h> 108#include <netinet6/ip6_var.h> 109#include <netinet6/scope6_var.h> 110#include <netinet6/nd6.h> 111#include <netinet6/vinet6.h> 112 113static struct mtx addrsel_lock; 114#define ADDRSEL_LOCK_INIT() mtx_init(&addrsel_lock, "addrsel_lock", NULL, MTX_DEF) 115#define ADDRSEL_LOCK() mtx_lock(&addrsel_lock) 116#define ADDRSEL_UNLOCK() mtx_unlock(&addrsel_lock) 117#define ADDRSEL_LOCK_ASSERT() mtx_assert(&addrsel_lock, MA_OWNED) 118 119static struct sx addrsel_sxlock; 120#define ADDRSEL_SXLOCK_INIT() sx_init(&addrsel_sxlock, "addrsel_sxlock") 121#define ADDRSEL_SLOCK() sx_slock(&addrsel_sxlock) 122#define ADDRSEL_SUNLOCK() sx_sunlock(&addrsel_sxlock) 123#define ADDRSEL_XLOCK() sx_xlock(&addrsel_sxlock) 124#define ADDRSEL_XUNLOCK() sx_xunlock(&addrsel_sxlock) 125 126#define ADDR_LABEL_NOTAPP (-1) 127 128#ifdef VIMAGE_GLOBALS 129struct in6_addrpolicy defaultaddrpolicy; 130int ip6_prefer_tempaddr; 131#endif 132 133static int selectroute __P((struct sockaddr_in6 *, struct ip6_pktopts *, 134 struct ip6_moptions *, struct route_in6 *, struct ifnet **, 135 struct rtentry **, int)); 136static int in6_selectif __P((struct sockaddr_in6 *, struct ip6_pktopts *, 137 struct ip6_moptions *, struct route_in6 *ro, struct ifnet **)); 138 139static struct in6_addrpolicy *lookup_addrsel_policy(struct sockaddr_in6 *); 140 141static void init_policy_queue(void); 142static int add_addrsel_policyent(struct in6_addrpolicy *); 143static int delete_addrsel_policyent(struct in6_addrpolicy *); 144static int walk_addrsel_policy __P((int (*)(struct in6_addrpolicy *, void *), 145 void *)); 146static int dump_addrsel_policyent(struct in6_addrpolicy *, void *); 147static struct in6_addrpolicy *match_addrsel_policy(struct sockaddr_in6 *); 148 149/* 150 * Return an IPv6 address, which is the most appropriate for a given 151 * destination and user specified options. 152 * If necessary, this function lookups the routing table and returns 153 * an entry to the caller for later use. 154 */ 155#define REPLACE(r) do {\ 156 if ((r) < sizeof(V_ip6stat.ip6s_sources_rule) / \ 157 sizeof(V_ip6stat.ip6s_sources_rule[0])) /* check for safety */ \ 158 V_ip6stat.ip6s_sources_rule[(r)]++; \ 159 /* { \ 160 char ip6buf[INET6_ADDRSTRLEN], ip6b[INET6_ADDRSTRLEN]; \ 161 printf("in6_selectsrc: replace %s with %s by %d\n", ia_best ? ip6_sprintf(ip6buf, &ia_best->ia_addr.sin6_addr) : "none", ip6_sprintf(ip6b, &ia->ia_addr.sin6_addr), (r)); \ 162 } */ \ 163 goto replace; \ 164} while(0) 165#define NEXT(r) do {\ 166 if ((r) < sizeof(V_ip6stat.ip6s_sources_rule) / \ 167 sizeof(V_ip6stat.ip6s_sources_rule[0])) /* check for safety */ \ 168 V_ip6stat.ip6s_sources_rule[(r)]++; \ 169 /* { \ 170 char ip6buf[INET6_ADDRSTRLEN], ip6b[INET6_ADDRSTRLEN]; \ 171 printf("in6_selectsrc: keep %s against %s by %d\n", ia_best ? ip6_sprintf(ip6buf, &ia_best->ia_addr.sin6_addr) : "none", ip6_sprintf(ip6b, &ia->ia_addr.sin6_addr), (r)); \ 172 } */ \ 173 goto next; /* XXX: we can't use 'continue' here */ \ 174} while(0) 175#define BREAK(r) do { \ 176 if ((r) < sizeof(V_ip6stat.ip6s_sources_rule) / \ 177 sizeof(V_ip6stat.ip6s_sources_rule[0])) /* check for safety */ \ 178 V_ip6stat.ip6s_sources_rule[(r)]++; \ 179 goto out; /* XXX: we can't use 'break' here */ \ 180} while(0) 181 182int 183in6_selectsrc(struct sockaddr_in6 *dstsock, struct ip6_pktopts *opts, 184 struct inpcb *inp, struct route_in6 *ro, struct ucred *cred, 185 struct ifnet **ifpp, struct in6_addr *srcp) 186{ 187 INIT_VNET_INET6(curvnet); 188 struct in6_addr dst; 189 struct ifnet *ifp = NULL; 190 struct in6_ifaddr *ia = NULL, *ia_best = NULL; 191 struct in6_pktinfo *pi = NULL; 192 int dst_scope = -1, best_scope = -1, best_matchlen = -1; 193 struct in6_addrpolicy *dst_policy = NULL, *best_policy = NULL; 194 u_int32_t odstzone; 195 int prefer_tempaddr; 196 int error; 197 struct ip6_moptions *mopts; 198 199 KASSERT(srcp != NULL, ("%s: srcp is NULL", __func__)); 200 201 dst = dstsock->sin6_addr; /* make a copy for local operation */ 202 if (ifpp) 203 *ifpp = NULL; 204 205 if (inp != NULL) { 206 INP_LOCK_ASSERT(inp); 207 mopts = inp->in6p_moptions; 208 } else { 209 mopts = NULL; 210 } 211 212 /* 213 * If the source address is explicitly specified by the caller, 214 * check if the requested source address is indeed a unicast address 215 * assigned to the node, and can be used as the packet's source 216 * address. If everything is okay, use the address as source. 217 */ 218 if (opts && (pi = opts->ip6po_pktinfo) && 219 !IN6_IS_ADDR_UNSPECIFIED(&pi->ipi6_addr)) { 220 struct sockaddr_in6 srcsock; 221 struct in6_ifaddr *ia6; 222 223 /* get the outgoing interface */ 224 if ((error = in6_selectif(dstsock, opts, mopts, ro, &ifp)) != 0) 225 return (error); 226 227 /* 228 * determine the appropriate zone id of the source based on 229 * the zone of the destination and the outgoing interface. 230 * If the specified address is ambiguous wrt the scope zone, 231 * the interface must be specified; otherwise, ifa_ifwithaddr() 232 * will fail matching the address. 233 */ 234 bzero(&srcsock, sizeof(srcsock)); 235 srcsock.sin6_family = AF_INET6; 236 srcsock.sin6_len = sizeof(srcsock); 237 srcsock.sin6_addr = pi->ipi6_addr; 238 if (ifp) { 239 error = in6_setscope(&srcsock.sin6_addr, ifp, NULL); 240 if (error) 241 return (error); 242 } 243 if (cred != NULL && (error = prison_local_ip6(cred, 244 &srcsock.sin6_addr, (inp != NULL && 245 (inp->inp_flags & IN6P_IPV6_V6ONLY) != 0))) != 0) 246 return (error); 247 248 ia6 = (struct in6_ifaddr *)ifa_ifwithaddr( 249 (struct sockaddr *)&srcsock); 250 if (ia6 == NULL || 251 (ia6->ia6_flags & (IN6_IFF_ANYCAST | IN6_IFF_NOTREADY))) { 252 if (ia6 != NULL) 253 ifa_free(&ia6->ia_ifa); 254 return (EADDRNOTAVAIL); 255 } 256 pi->ipi6_addr = srcsock.sin6_addr; /* XXX: this overrides pi */ 257 if (ifpp) 258 *ifpp = ifp; 259 bcopy(&ia6->ia_addr.sin6_addr, srcp, sizeof(*srcp)); 260 ifa_free(&ia6->ia_ifa); 261 return (0); 262 } 263 264 /* 265 * Otherwise, if the socket has already bound the source, just use it. 266 */ 267 if (inp != NULL && !IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) { 268 if (cred != NULL && 269 (error = prison_local_ip6(cred, &inp->in6p_laddr, 270 ((inp->inp_flags & IN6P_IPV6_V6ONLY) != 0))) != 0) 271 return (error); 272 bcopy(&inp->in6p_laddr, srcp, sizeof(*srcp)); 273 return (0); 274 } 275 276 /* 277 * If the address is not specified, choose the best one based on 278 * the outgoing interface and the destination address. 279 */ 280 /* get the outgoing interface */ 281 if ((error = in6_selectif(dstsock, opts, mopts, ro, &ifp)) != 0) 282 return (error); 283 284#ifdef DIAGNOSTIC 285 if (ifp == NULL) /* this should not happen */ 286 panic("in6_selectsrc: NULL ifp"); 287#endif 288 error = in6_setscope(&dst, ifp, &odstzone); 289 if (error) 290 return (error); 291 292 for (ia = V_in6_ifaddr; ia; ia = ia->ia_next) { 293 int new_scope = -1, new_matchlen = -1; 294 struct in6_addrpolicy *new_policy = NULL; 295 u_int32_t srczone, osrczone, dstzone; 296 struct in6_addr src; 297 struct ifnet *ifp1 = ia->ia_ifp; 298 299 /* 300 * We'll never take an address that breaks the scope zone 301 * of the destination. We also skip an address if its zone 302 * does not contain the outgoing interface. 303 * XXX: we should probably use sin6_scope_id here. 304 */ 305 if (in6_setscope(&dst, ifp1, &dstzone) || 306 odstzone != dstzone) { 307 continue; 308 } 309 src = ia->ia_addr.sin6_addr; 310 if (in6_setscope(&src, ifp, &osrczone) || 311 in6_setscope(&src, ifp1, &srczone) || 312 osrczone != srczone) { 313 continue; 314 } 315 316 /* avoid unusable addresses */ 317 if ((ia->ia6_flags & 318 (IN6_IFF_NOTREADY | IN6_IFF_ANYCAST | IN6_IFF_DETACHED))) { 319 continue; 320 } 321 if (!V_ip6_use_deprecated && IFA6_IS_DEPRECATED(ia)) 322 continue; 323 324 if (cred != NULL && 325 prison_local_ip6(cred, &ia->ia_addr.sin6_addr, 326 (inp != NULL && 327 (inp->inp_flags & IN6P_IPV6_V6ONLY) != 0)) != 0) 328 continue; 329 330 /* Rule 1: Prefer same address */ 331 if (IN6_ARE_ADDR_EQUAL(&dst, &ia->ia_addr.sin6_addr)) { 332 ia_best = ia; 333 BREAK(1); /* there should be no better candidate */ 334 } 335 336 if (ia_best == NULL) 337 REPLACE(0); 338 339 /* Rule 2: Prefer appropriate scope */ 340 if (dst_scope < 0) 341 dst_scope = in6_addrscope(&dst); 342 new_scope = in6_addrscope(&ia->ia_addr.sin6_addr); 343 if (IN6_ARE_SCOPE_CMP(best_scope, new_scope) < 0) { 344 if (IN6_ARE_SCOPE_CMP(best_scope, dst_scope) < 0) 345 REPLACE(2); 346 NEXT(2); 347 } else if (IN6_ARE_SCOPE_CMP(new_scope, best_scope) < 0) { 348 if (IN6_ARE_SCOPE_CMP(new_scope, dst_scope) < 0) 349 NEXT(2); 350 REPLACE(2); 351 } 352 353 /* 354 * Rule 3: Avoid deprecated addresses. Note that the case of 355 * !ip6_use_deprecated is already rejected above. 356 */ 357 if (!IFA6_IS_DEPRECATED(ia_best) && IFA6_IS_DEPRECATED(ia)) 358 NEXT(3); 359 if (IFA6_IS_DEPRECATED(ia_best) && !IFA6_IS_DEPRECATED(ia)) 360 REPLACE(3); 361 362 /* Rule 4: Prefer home addresses */ 363 /* 364 * XXX: This is a TODO. We should probably merge the MIP6 365 * case above. 366 */ 367 368 /* Rule 5: Prefer outgoing interface */ 369 if (ia_best->ia_ifp == ifp && ia->ia_ifp != ifp) 370 NEXT(5); 371 if (ia_best->ia_ifp != ifp && ia->ia_ifp == ifp) 372 REPLACE(5); 373 374 /* 375 * Rule 6: Prefer matching label 376 * Note that best_policy should be non-NULL here. 377 */ 378 if (dst_policy == NULL) 379 dst_policy = lookup_addrsel_policy(dstsock); 380 if (dst_policy->label != ADDR_LABEL_NOTAPP) { 381 new_policy = lookup_addrsel_policy(&ia->ia_addr); 382 if (dst_policy->label == best_policy->label && 383 dst_policy->label != new_policy->label) 384 NEXT(6); 385 if (dst_policy->label != best_policy->label && 386 dst_policy->label == new_policy->label) 387 REPLACE(6); 388 } 389 390 /* 391 * Rule 7: Prefer public addresses. 392 * We allow users to reverse the logic by configuring 393 * a sysctl variable, so that privacy conscious users can 394 * always prefer temporary addresses. 395 */ 396 if (opts == NULL || 397 opts->ip6po_prefer_tempaddr == IP6PO_TEMPADDR_SYSTEM) { 398 prefer_tempaddr = V_ip6_prefer_tempaddr; 399 } else if (opts->ip6po_prefer_tempaddr == 400 IP6PO_TEMPADDR_NOTPREFER) { 401 prefer_tempaddr = 0; 402 } else 403 prefer_tempaddr = 1; 404 if (!(ia_best->ia6_flags & IN6_IFF_TEMPORARY) && 405 (ia->ia6_flags & IN6_IFF_TEMPORARY)) { 406 if (prefer_tempaddr) 407 REPLACE(7); 408 else 409 NEXT(7); 410 } 411 if ((ia_best->ia6_flags & IN6_IFF_TEMPORARY) && 412 !(ia->ia6_flags & IN6_IFF_TEMPORARY)) { 413 if (prefer_tempaddr) 414 NEXT(7); 415 else 416 REPLACE(7); 417 } 418 419 /* 420 * Rule 8: prefer addresses on alive interfaces. 421 * This is a KAME specific rule. 422 */ 423 if ((ia_best->ia_ifp->if_flags & IFF_UP) && 424 !(ia->ia_ifp->if_flags & IFF_UP)) 425 NEXT(8); 426 if (!(ia_best->ia_ifp->if_flags & IFF_UP) && 427 (ia->ia_ifp->if_flags & IFF_UP)) 428 REPLACE(8); 429 430 /* 431 * Rule 14: Use longest matching prefix. 432 * Note: in the address selection draft, this rule is 433 * documented as "Rule 8". However, since it is also 434 * documented that this rule can be overridden, we assign 435 * a large number so that it is easy to assign smaller numbers 436 * to more preferred rules. 437 */ 438 new_matchlen = in6_matchlen(&ia->ia_addr.sin6_addr, &dst); 439 if (best_matchlen < new_matchlen) 440 REPLACE(14); 441 if (new_matchlen < best_matchlen) 442 NEXT(14); 443 444 /* Rule 15 is reserved. */ 445 446 /* 447 * Last resort: just keep the current candidate. 448 * Or, do we need more rules? 449 */ 450 continue; 451 452 replace: 453 ia_best = ia; 454 best_scope = (new_scope >= 0 ? new_scope : 455 in6_addrscope(&ia_best->ia_addr.sin6_addr)); 456 best_policy = (new_policy ? new_policy : 457 lookup_addrsel_policy(&ia_best->ia_addr)); 458 best_matchlen = (new_matchlen >= 0 ? new_matchlen : 459 in6_matchlen(&ia_best->ia_addr.sin6_addr, 460 &dst)); 461 462 next: 463 continue; 464 465 out: 466 break; 467 } 468 469 if ((ia = ia_best) == NULL) 470 return (EADDRNOTAVAIL); 471 472 if (ifpp) 473 *ifpp = ifp; 474 475 bcopy(&ia->ia_addr.sin6_addr, srcp, sizeof(*srcp)); 476 return (0); 477} 478 479/* 480 * clone - meaningful only for bsdi and freebsd 481 */ 482static int 483selectroute(struct sockaddr_in6 *dstsock, struct ip6_pktopts *opts, 484 struct ip6_moptions *mopts, struct route_in6 *ro, 485 struct ifnet **retifp, struct rtentry **retrt, int norouteok) 486{ 487 INIT_VNET_INET6(curvnet); 488 int error = 0; 489 struct ifnet *ifp = NULL; 490 struct rtentry *rt = NULL; 491 struct sockaddr_in6 *sin6_next; 492 struct in6_pktinfo *pi = NULL; 493 struct in6_addr *dst = &dstsock->sin6_addr; 494#if 0 495 char ip6buf[INET6_ADDRSTRLEN]; 496 497 if (dstsock->sin6_addr.s6_addr32[0] == 0 && 498 dstsock->sin6_addr.s6_addr32[1] == 0 && 499 !IN6_IS_ADDR_LOOPBACK(&dstsock->sin6_addr)) { 500 printf("in6_selectroute: strange destination %s\n", 501 ip6_sprintf(ip6buf, &dstsock->sin6_addr)); 502 } else { 503 printf("in6_selectroute: destination = %s%%%d\n", 504 ip6_sprintf(ip6buf, &dstsock->sin6_addr), 505 dstsock->sin6_scope_id); /* for debug */ 506 } 507#endif 508 509 /* If the caller specify the outgoing interface explicitly, use it. */ 510 if (opts && (pi = opts->ip6po_pktinfo) != NULL && pi->ipi6_ifindex) { 511 /* XXX boundary check is assumed to be already done. */ 512 ifp = ifnet_byindex(pi->ipi6_ifindex); 513 if (ifp != NULL && 514 (norouteok || retrt == NULL || 515 IN6_IS_ADDR_MULTICAST(dst))) { 516 /* 517 * we do not have to check or get the route for 518 * multicast. 519 */ 520 goto done; 521 } else 522 goto getroute; 523 } 524 525 /* 526 * If the destination address is a multicast address and the outgoing 527 * interface for the address is specified by the caller, use it. 528 */ 529 if (IN6_IS_ADDR_MULTICAST(dst) && 530 mopts != NULL && (ifp = mopts->im6o_multicast_ifp) != NULL) { 531 goto done; /* we do not need a route for multicast. */ 532 } 533 534 getroute: 535 /* 536 * If the next hop address for the packet is specified by the caller, 537 * use it as the gateway. 538 */ 539 if (opts && opts->ip6po_nexthop) { 540 struct route_in6 *ron; 541 struct llentry *la; 542 543 sin6_next = satosin6(opts->ip6po_nexthop); 544 545 /* at this moment, we only support AF_INET6 next hops */ 546 if (sin6_next->sin6_family != AF_INET6) { 547 error = EAFNOSUPPORT; /* or should we proceed? */ 548 goto done; 549 } 550 551 /* 552 * If the next hop is an IPv6 address, then the node identified 553 * by that address must be a neighbor of the sending host. 554 */ 555 ron = &opts->ip6po_nextroute; 556 /* 557 * XXX what do we do here? 558 * PLZ to be fixing 559 */ 560 561 562 if (ron->ro_rt == NULL) { 563 rtalloc((struct route *)ron); /* multi path case? */ 564 if (ron->ro_rt == NULL) { 565 if (ron->ro_rt) { 566 RTFREE(ron->ro_rt); 567 ron->ro_rt = NULL; 568 } 569 error = EHOSTUNREACH; 570 goto done; 571 } 572 } 573 574 rt = ron->ro_rt; 575 ifp = rt->rt_ifp; 576 IF_AFDATA_LOCK(ifp); 577 la = lla_lookup(LLTABLE6(ifp), 0, (struct sockaddr *)&sin6_next->sin6_addr); 578 IF_AFDATA_UNLOCK(ifp); 579 if (la != NULL) 580 LLE_RUNLOCK(la); 581 else { 582 error = EHOSTUNREACH; 583 goto done; 584 } 585#if 0 586 if ((ron->ro_rt && 587 (ron->ro_rt->rt_flags & (RTF_UP | RTF_LLINFO)) != 588 (RTF_UP | RTF_LLINFO)) || 589 !IN6_ARE_ADDR_EQUAL(&satosin6(&ron->ro_dst)->sin6_addr, 590 &sin6_next->sin6_addr)) { 591 if (ron->ro_rt) { 592 RTFREE(ron->ro_rt); 593 ron->ro_rt = NULL; 594 } 595 *satosin6(&ron->ro_dst) = *sin6_next; 596 } 597 if (ron->ro_rt == NULL) { 598 rtalloc((struct route *)ron); /* multi path case? */ 599 if (ron->ro_rt == NULL || 600 !(ron->ro_rt->rt_flags & RTF_LLINFO)) { 601 if (ron->ro_rt) { 602 RTFREE(ron->ro_rt); 603 ron->ro_rt = NULL; 604 } 605 error = EHOSTUNREACH; 606 goto done; 607 } 608 } 609#endif 610 611 /* 612 * When cloning is required, try to allocate a route to the 613 * destination so that the caller can store path MTU 614 * information. 615 */ 616 goto done; 617 } 618 619 /* 620 * Use a cached route if it exists and is valid, else try to allocate 621 * a new one. Note that we should check the address family of the 622 * cached destination, in case of sharing the cache with IPv4. 623 */ 624 if (ro) { 625 if (ro->ro_rt && 626 (!(ro->ro_rt->rt_flags & RTF_UP) || 627 ((struct sockaddr *)(&ro->ro_dst))->sa_family != AF_INET6 || 628 !IN6_ARE_ADDR_EQUAL(&satosin6(&ro->ro_dst)->sin6_addr, 629 dst))) { 630 RTFREE(ro->ro_rt); 631 ro->ro_rt = (struct rtentry *)NULL; 632 } 633 if (ro->ro_rt == (struct rtentry *)NULL) { 634 struct sockaddr_in6 *sa6; 635 636 /* No route yet, so try to acquire one */ 637 bzero(&ro->ro_dst, sizeof(struct sockaddr_in6)); 638 sa6 = (struct sockaddr_in6 *)&ro->ro_dst; 639 *sa6 = *dstsock; 640 sa6->sin6_scope_id = 0; 641 642#ifdef RADIX_MPATH 643 rtalloc_mpath((struct route *)ro, 644 ntohl(sa6->sin6_addr.s6_addr32[3])); 645#else 646 ro->ro_rt = rtalloc1(&((struct route *)ro) 647 ->ro_dst, 0, 0UL); 648 if (ro->ro_rt) 649 RT_UNLOCK(ro->ro_rt); 650#endif 651 } 652 653 /* 654 * do not care about the result if we have the nexthop 655 * explicitly specified. 656 */ 657 if (opts && opts->ip6po_nexthop) 658 goto done; 659 660 if (ro->ro_rt) { 661 ifp = ro->ro_rt->rt_ifp; 662 663 if (ifp == NULL) { /* can this really happen? */ 664 RTFREE(ro->ro_rt); 665 ro->ro_rt = NULL; 666 } 667 } 668 if (ro->ro_rt == NULL) 669 error = EHOSTUNREACH; 670 rt = ro->ro_rt; 671 672 /* 673 * Check if the outgoing interface conflicts with 674 * the interface specified by ipi6_ifindex (if specified). 675 * Note that loopback interface is always okay. 676 * (this may happen when we are sending a packet to one of 677 * our own addresses.) 678 */ 679 if (ifp && opts && opts->ip6po_pktinfo && 680 opts->ip6po_pktinfo->ipi6_ifindex) { 681 if (!(ifp->if_flags & IFF_LOOPBACK) && 682 ifp->if_index != 683 opts->ip6po_pktinfo->ipi6_ifindex) { 684 error = EHOSTUNREACH; 685 goto done; 686 } 687 } 688 } 689 690 done: 691 if (ifp == NULL && rt == NULL) { 692 /* 693 * This can happen if the caller did not pass a cached route 694 * nor any other hints. We treat this case an error. 695 */ 696 error = EHOSTUNREACH; 697 } 698 if (error == EHOSTUNREACH) 699 V_ip6stat.ip6s_noroute++; 700 701 if (retifp != NULL) 702 *retifp = ifp; 703 if (retrt != NULL) 704 *retrt = rt; /* rt may be NULL */ 705 706 return (error); 707} 708 709static int 710in6_selectif(struct sockaddr_in6 *dstsock, struct ip6_pktopts *opts, 711 struct ip6_moptions *mopts, struct route_in6 *ro, struct ifnet **retifp) 712{ 713 int error; 714 struct route_in6 sro; 715 struct rtentry *rt = NULL; 716 717 if (ro == NULL) { 718 bzero(&sro, sizeof(sro)); 719 ro = &sro; 720 } 721 722 if ((error = selectroute(dstsock, opts, mopts, ro, retifp, 723 &rt, 1)) != 0) { 724 if (ro == &sro && rt && rt == sro.ro_rt) 725 RTFREE(rt); 726 return (error); 727 } 728 729 /* 730 * do not use a rejected or black hole route. 731 * XXX: this check should be done in the L2 output routine. 732 * However, if we skipped this check here, we'd see the following 733 * scenario: 734 * - install a rejected route for a scoped address prefix 735 * (like fe80::/10) 736 * - send a packet to a destination that matches the scoped prefix, 737 * with ambiguity about the scope zone. 738 * - pick the outgoing interface from the route, and disambiguate the 739 * scope zone with the interface. 740 * - ip6_output() would try to get another route with the "new" 741 * destination, which may be valid. 742 * - we'd see no error on output. 743 * Although this may not be very harmful, it should still be confusing. 744 * We thus reject the case here. 745 */ 746 if (rt && (rt->rt_flags & (RTF_REJECT | RTF_BLACKHOLE))) { 747 int flags = (rt->rt_flags & RTF_HOST ? EHOSTUNREACH : ENETUNREACH); 748 749 if (ro == &sro && rt && rt == sro.ro_rt) 750 RTFREE(rt); 751 return (flags); 752 } 753 754 /* 755 * Adjust the "outgoing" interface. If we're going to loop the packet 756 * back to ourselves, the ifp would be the loopback interface. 757 * However, we'd rather know the interface associated to the 758 * destination address (which should probably be one of our own 759 * addresses.) 760 */ 761 if (rt && rt->rt_ifa && rt->rt_ifa->ifa_ifp) 762 *retifp = rt->rt_ifa->ifa_ifp; 763 764 if (ro == &sro && rt && rt == sro.ro_rt) 765 RTFREE(rt); 766 return (0); 767} 768 769/* 770 * clone - meaningful only for bsdi and freebsd 771 */ 772int 773in6_selectroute(struct sockaddr_in6 *dstsock, struct ip6_pktopts *opts, 774 struct ip6_moptions *mopts, struct route_in6 *ro, 775 struct ifnet **retifp, struct rtentry **retrt) 776{ 777 778 return (selectroute(dstsock, opts, mopts, ro, retifp, 779 retrt, 0)); 780} 781 782/* 783 * Default hop limit selection. The precedence is as follows: 784 * 1. Hoplimit value specified via ioctl. 785 * 2. (If the outgoing interface is detected) the current 786 * hop limit of the interface specified by router advertisement. 787 * 3. The system default hoplimit. 788 */ 789int 790in6_selecthlim(struct inpcb *in6p, struct ifnet *ifp) 791{ 792 INIT_VNET_INET6(curvnet); 793 794 if (in6p && in6p->in6p_hops >= 0) 795 return (in6p->in6p_hops); 796 else if (ifp) 797 return (ND_IFINFO(ifp)->chlim); 798 else if (in6p && !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr)) { 799 struct route_in6 ro6; 800 struct ifnet *lifp; 801 802 bzero(&ro6, sizeof(ro6)); 803 ro6.ro_dst.sin6_family = AF_INET6; 804 ro6.ro_dst.sin6_len = sizeof(struct sockaddr_in6); 805 ro6.ro_dst.sin6_addr = in6p->in6p_faddr; 806 rtalloc((struct route *)&ro6); 807 if (ro6.ro_rt) { 808 lifp = ro6.ro_rt->rt_ifp; 809 RTFREE(ro6.ro_rt); 810 if (lifp) 811 return (ND_IFINFO(lifp)->chlim); 812 } else 813 return (V_ip6_defhlim); 814 } 815 return (V_ip6_defhlim); 816} 817 818/* 819 * XXX: this is borrowed from in6_pcbbind(). If possible, we should 820 * share this function by all *bsd*... 821 */ 822int 823in6_pcbsetport(struct in6_addr *laddr, struct inpcb *inp, struct ucred *cred) 824{ 825 INIT_VNET_INET(curvnet); 826 struct socket *so = inp->inp_socket; 827 u_int16_t lport = 0, first, last, *lastport; 828 int count, error, wild = 0, dorandom; 829 struct inpcbinfo *pcbinfo = inp->inp_pcbinfo; 830 831 INP_INFO_WLOCK_ASSERT(pcbinfo); 832 INP_WLOCK_ASSERT(inp); 833 834 error = prison_local_ip6(cred, laddr, 835 ((inp->inp_flags & IN6P_IPV6_V6ONLY) != 0)); 836 if (error) 837 return(error); 838 839 /* XXX: this is redundant when called from in6_pcbbind */ 840 if ((so->so_options & (SO_REUSEADDR|SO_REUSEPORT)) == 0) 841 wild = INPLOOKUP_WILDCARD; 842 843 inp->inp_flags |= INP_ANONPORT; 844 845 if (inp->inp_flags & INP_HIGHPORT) { 846 first = V_ipport_hifirstauto; /* sysctl */ 847 last = V_ipport_hilastauto; 848 lastport = &pcbinfo->ipi_lasthi; 849 } else if (inp->inp_flags & INP_LOWPORT) { 850 error = priv_check_cred(cred, PRIV_NETINET_RESERVEDPORT, 0); 851 if (error) 852 return error; 853 first = V_ipport_lowfirstauto; /* 1023 */ 854 last = V_ipport_lowlastauto; /* 600 */ 855 lastport = &pcbinfo->ipi_lastlow; 856 } else { 857 first = V_ipport_firstauto; /* sysctl */ 858 last = V_ipport_lastauto; 859 lastport = &pcbinfo->ipi_lastport; 860 } 861 862 /* 863 * For UDP, use random port allocation as long as the user 864 * allows it. For TCP (and as of yet unknown) connections, 865 * use random port allocation only if the user allows it AND 866 * ipport_tick() allows it. 867 */ 868 if (V_ipport_randomized && 869 (!V_ipport_stoprandom || pcbinfo == &V_udbinfo)) 870 dorandom = 1; 871 else 872 dorandom = 0; 873 /* 874 * It makes no sense to do random port allocation if 875 * we have the only port available. 876 */ 877 if (first == last) 878 dorandom = 0; 879 /* Make sure to not include UDP packets in the count. */ 880 if (pcbinfo != &V_udbinfo) 881 V_ipport_tcpallocs++; 882 883 /* 884 * Instead of having two loops further down counting up or down 885 * make sure that first is always <= last and go with only one 886 * code path implementing all logic. 887 */ 888 if (first > last) { 889 u_int16_t aux; 890 891 aux = first; 892 first = last; 893 last = aux; 894 } 895 896 if (dorandom) 897 *lastport = first + (arc4random() % (last - first)); 898 899 count = last - first; 900 901 do { 902 if (count-- < 0) { /* completely used? */ 903 /* Undo an address bind that may have occurred. */ 904 inp->in6p_laddr = in6addr_any; 905 return (EADDRNOTAVAIL); 906 } 907 ++*lastport; 908 if (*lastport < first || *lastport > last) 909 *lastport = first; 910 lport = htons(*lastport); 911 } while (in6_pcblookup_local(pcbinfo, &inp->in6p_laddr, 912 lport, wild, cred)); 913 914 inp->inp_lport = lport; 915 if (in_pcbinshash(inp) != 0) { 916 inp->in6p_laddr = in6addr_any; 917 inp->inp_lport = 0; 918 return (EAGAIN); 919 } 920 921 return (0); 922} 923 924void 925addrsel_policy_init(void) 926{ 927 INIT_VNET_INET6(curvnet); 928 929 V_ip6_prefer_tempaddr = 0; 930 931 init_policy_queue(); 932 933 /* initialize the "last resort" policy */ 934 bzero(&V_defaultaddrpolicy, sizeof(V_defaultaddrpolicy)); 935 V_defaultaddrpolicy.label = ADDR_LABEL_NOTAPP; 936 937 if (!IS_DEFAULT_VNET(curvnet)) 938 return; 939 940 ADDRSEL_LOCK_INIT(); 941 ADDRSEL_SXLOCK_INIT(); 942} 943 944static struct in6_addrpolicy * 945lookup_addrsel_policy(struct sockaddr_in6 *key) 946{ 947 INIT_VNET_INET6(curvnet); 948 struct in6_addrpolicy *match = NULL; 949 950 ADDRSEL_LOCK(); 951 match = match_addrsel_policy(key); 952 953 if (match == NULL) 954 match = &V_defaultaddrpolicy; 955 else 956 match->use++; 957 ADDRSEL_UNLOCK(); 958 959 return (match); 960} 961 962/* 963 * Subroutines to manage the address selection policy table via sysctl. 964 */ 965struct walkarg { 966 struct sysctl_req *w_req; 967}; 968 969static int in6_src_sysctl(SYSCTL_HANDLER_ARGS); 970SYSCTL_DECL(_net_inet6_ip6); 971SYSCTL_NODE(_net_inet6_ip6, IPV6CTL_ADDRCTLPOLICY, addrctlpolicy, 972 CTLFLAG_RD, in6_src_sysctl, ""); 973 974static int 975in6_src_sysctl(SYSCTL_HANDLER_ARGS) 976{ 977 struct walkarg w; 978 979 if (req->newptr) 980 return EPERM; 981 982 bzero(&w, sizeof(w)); 983 w.w_req = req; 984 985 return (walk_addrsel_policy(dump_addrsel_policyent, &w)); 986} 987 988int 989in6_src_ioctl(u_long cmd, caddr_t data) 990{ 991 int i; 992 struct in6_addrpolicy ent0; 993 994 if (cmd != SIOCAADDRCTL_POLICY && cmd != SIOCDADDRCTL_POLICY) 995 return (EOPNOTSUPP); /* check for safety */ 996 997 ent0 = *(struct in6_addrpolicy *)data; 998 999 if (ent0.label == ADDR_LABEL_NOTAPP) 1000 return (EINVAL); 1001 /* check if the prefix mask is consecutive. */ 1002 if (in6_mask2len(&ent0.addrmask.sin6_addr, NULL) < 0) 1003 return (EINVAL); 1004 /* clear trailing garbages (if any) of the prefix address. */ 1005 for (i = 0; i < 4; i++) { 1006 ent0.addr.sin6_addr.s6_addr32[i] &= 1007 ent0.addrmask.sin6_addr.s6_addr32[i]; 1008 } 1009 ent0.use = 0; 1010 1011 switch (cmd) { 1012 case SIOCAADDRCTL_POLICY: 1013 return (add_addrsel_policyent(&ent0)); 1014 case SIOCDADDRCTL_POLICY: 1015 return (delete_addrsel_policyent(&ent0)); 1016 } 1017 1018 return (0); /* XXX: compromise compilers */ 1019} 1020 1021/* 1022 * The followings are implementation of the policy table using a 1023 * simple tail queue. 1024 * XXX such details should be hidden. 1025 * XXX implementation using binary tree should be more efficient. 1026 */ 1027struct addrsel_policyent { 1028 TAILQ_ENTRY(addrsel_policyent) ape_entry; 1029 struct in6_addrpolicy ape_policy; 1030}; 1031 1032TAILQ_HEAD(addrsel_policyhead, addrsel_policyent); 1033 1034#ifdef VIMAGE_GLOBALS 1035struct addrsel_policyhead addrsel_policytab; 1036#endif 1037 1038static void 1039init_policy_queue(void) 1040{ 1041 INIT_VNET_INET6(curvnet); 1042 1043 TAILQ_INIT(&V_addrsel_policytab); 1044} 1045 1046static int 1047add_addrsel_policyent(struct in6_addrpolicy *newpolicy) 1048{ 1049 INIT_VNET_INET6(curvnet); 1050 struct addrsel_policyent *new, *pol; 1051 1052 new = malloc(sizeof(*new), M_IFADDR, 1053 M_WAITOK); 1054 ADDRSEL_XLOCK(); 1055 ADDRSEL_LOCK(); 1056 1057 /* duplication check */ 1058 TAILQ_FOREACH(pol, &V_addrsel_policytab, ape_entry) { 1059 if (IN6_ARE_ADDR_EQUAL(&newpolicy->addr.sin6_addr, 1060 &pol->ape_policy.addr.sin6_addr) && 1061 IN6_ARE_ADDR_EQUAL(&newpolicy->addrmask.sin6_addr, 1062 &pol->ape_policy.addrmask.sin6_addr)) { 1063 ADDRSEL_UNLOCK(); 1064 ADDRSEL_XUNLOCK(); 1065 free(new, M_IFADDR); 1066 return (EEXIST); /* or override it? */ 1067 } 1068 } 1069 1070 bzero(new, sizeof(*new)); 1071 1072 /* XXX: should validate entry */ 1073 new->ape_policy = *newpolicy; 1074 1075 TAILQ_INSERT_TAIL(&V_addrsel_policytab, new, ape_entry); 1076 ADDRSEL_UNLOCK(); 1077 ADDRSEL_XUNLOCK(); 1078 1079 return (0); 1080} 1081 1082static int 1083delete_addrsel_policyent(struct in6_addrpolicy *key) 1084{ 1085 INIT_VNET_INET6(curvnet); 1086 struct addrsel_policyent *pol; 1087 1088 ADDRSEL_XLOCK(); 1089 ADDRSEL_LOCK(); 1090 1091 /* search for the entry in the table */ 1092 TAILQ_FOREACH(pol, &V_addrsel_policytab, ape_entry) { 1093 if (IN6_ARE_ADDR_EQUAL(&key->addr.sin6_addr, 1094 &pol->ape_policy.addr.sin6_addr) && 1095 IN6_ARE_ADDR_EQUAL(&key->addrmask.sin6_addr, 1096 &pol->ape_policy.addrmask.sin6_addr)) { 1097 break; 1098 } 1099 } 1100 if (pol == NULL) { 1101 ADDRSEL_UNLOCK(); 1102 ADDRSEL_XUNLOCK(); 1103 return (ESRCH); 1104 } 1105 1106 TAILQ_REMOVE(&V_addrsel_policytab, pol, ape_entry); 1107 ADDRSEL_UNLOCK(); 1108 ADDRSEL_XUNLOCK(); 1109 1110 return (0); 1111} 1112 1113static int 1114walk_addrsel_policy(int (*callback)(struct in6_addrpolicy *, void *), 1115 void *w) 1116{ 1117 INIT_VNET_INET6(curvnet); 1118 struct addrsel_policyent *pol; 1119 int error = 0; 1120 1121 ADDRSEL_SLOCK(); 1122 TAILQ_FOREACH(pol, &V_addrsel_policytab, ape_entry) { 1123 if ((error = (*callback)(&pol->ape_policy, w)) != 0) { 1124 ADDRSEL_SUNLOCK(); 1125 return (error); 1126 } 1127 } 1128 ADDRSEL_SUNLOCK(); 1129 return (error); 1130} 1131 1132static int 1133dump_addrsel_policyent(struct in6_addrpolicy *pol, void *arg) 1134{ 1135 int error = 0; 1136 struct walkarg *w = arg; 1137 1138 error = SYSCTL_OUT(w->w_req, pol, sizeof(*pol)); 1139 1140 return (error); 1141} 1142 1143static struct in6_addrpolicy * 1144match_addrsel_policy(struct sockaddr_in6 *key) 1145{ 1146 INIT_VNET_INET6(curvnet); 1147 struct addrsel_policyent *pent; 1148 struct in6_addrpolicy *bestpol = NULL, *pol; 1149 int matchlen, bestmatchlen = -1; 1150 u_char *mp, *ep, *k, *p, m; 1151 1152 TAILQ_FOREACH(pent, &V_addrsel_policytab, ape_entry) { 1153 matchlen = 0; 1154 1155 pol = &pent->ape_policy; 1156 mp = (u_char *)&pol->addrmask.sin6_addr; 1157 ep = mp + 16; /* XXX: scope field? */ 1158 k = (u_char *)&key->sin6_addr; 1159 p = (u_char *)&pol->addr.sin6_addr; 1160 for (; mp < ep && *mp; mp++, k++, p++) { 1161 m = *mp; 1162 if ((*k & m) != *p) 1163 goto next; /* not match */ 1164 if (m == 0xff) /* short cut for a typical case */ 1165 matchlen += 8; 1166 else { 1167 while (m >= 0x80) { 1168 matchlen++; 1169 m <<= 1; 1170 } 1171 } 1172 } 1173 1174 /* matched. check if this is better than the current best. */ 1175 if (bestpol == NULL || 1176 matchlen > bestmatchlen) { 1177 bestpol = pol; 1178 bestmatchlen = matchlen; 1179 } 1180 1181 next: 1182 continue; 1183 } 1184 1185 return (bestpol); 1186} 1187