1 /* $NetBSD: ip6_output.c,v 1.158 2014/08/16 17:27:09 maxv Exp $ */ 2 /* $KAME: ip6_output.c,v 1.172 2001/03/25 09:55:56 itojun Exp $ */ 3 4 /* 5 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project. 6 * All rights reserved. 7 * 8 * Redistribution and use in source and binary forms, with or without 9 * modification, are permitted provided that the following conditions 10 * are met: 11 * 1. Redistributions of source code must retain the above copyright 12 * notice, this list of conditions and the following disclaimer. 13 * 2. Redistributions in binary form must reproduce the above copyright 14 * notice, this list of conditions and the following disclaimer in the 15 * documentation and/or other materials provided with the distribution. 16 * 3. Neither the name of the project nor the names of its contributors 17 * may be used to endorse or promote products derived from this software 18 * without specific prior written permission. 19 * 20 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND 21 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 22 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 23 * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE 24 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 25 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 26 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 27 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 28 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 29 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 30 * SUCH DAMAGE. 31 */ 32 33 /* 34 * Copyright (c) 1982, 1986, 1988, 1990, 1993 35 * The Regents of the University of California. All rights reserved. 36 * 37 * Redistribution and use in source and binary forms, with or without 38 * modification, are permitted provided that the following conditions 39 * are met: 40 * 1. Redistributions of source code must retain the above copyright 41 * notice, this list of conditions and the following disclaimer. 42 * 2. Redistributions in binary form must reproduce the above copyright 43 * notice, this list of conditions and the following disclaimer in the 44 * documentation and/or other materials provided with the distribution. 45 * 3. Neither the name of the University nor the names of its contributors 46 * may be used to endorse or promote products derived from this software 47 * without specific prior written permission. 48 * 49 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 50 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 51 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 52 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 53 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 54 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 55 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 56 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 57 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 58 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 59 * SUCH DAMAGE. 60 * 61 * @(#)ip_output.c 8.3 (Berkeley) 1/21/94 62 */ 63 64 #include <sys/cdefs.h> 65 __KERNEL_RCSID(0, "$NetBSD: ip6_output.c,v 1.158 2014/08/16 17:27:09 maxv Exp $"); 66 67 #include "opt_inet.h" 68 #include "opt_inet6.h" 69 #include "opt_ipsec.h" 70 71 #include <sys/param.h> 72 #include <sys/malloc.h> 73 #include <sys/mbuf.h> 74 #include <sys/errno.h> 75 #include <sys/protosw.h> 76 #include <sys/socket.h> 77 #include <sys/socketvar.h> 78 #include <sys/systm.h> 79 #include <sys/proc.h> 80 #include <sys/kauth.h> 81 82 #include <net/if.h> 83 #include <net/route.h> 84 #include <net/pfil.h> 85 86 #include <netinet/in.h> 87 #include <netinet/in_var.h> 88 #include <netinet/ip6.h> 89 #include <netinet/icmp6.h> 90 #include <netinet/in_offload.h> 91 #include <netinet/portalgo.h> 92 #include <netinet6/in6_offload.h> 93 #include <netinet6/ip6_var.h> 94 #include <netinet6/ip6_private.h> 95 #include <netinet6/in6_pcb.h> 96 #include <netinet6/nd6.h> 97 #include <netinet6/ip6protosw.h> 98 #include <netinet6/scope6_var.h> 99 100 #ifdef IPSEC 101 #include <netipsec/ipsec.h> 102 #include <netipsec/ipsec6.h> 103 #include <netipsec/key.h> 104 #include <netipsec/xform.h> 105 #endif 106 107 108 #include <net/net_osdep.h> 109 110 extern pfil_head_t *inet6_pfil_hook; /* XXX */ 111 112 struct ip6_exthdrs { 113 struct mbuf *ip6e_ip6; 114 struct mbuf *ip6e_hbh; 115 struct mbuf *ip6e_dest1; 116 struct mbuf *ip6e_rthdr; 117 struct mbuf *ip6e_dest2; 118 }; 119 120 static int ip6_pcbopt(int, u_char *, int, struct ip6_pktopts **, 121 kauth_cred_t, int); 122 static int ip6_getpcbopt(struct ip6_pktopts *, int, struct sockopt *); 123 static int ip6_setpktopt(int, u_char *, int, struct ip6_pktopts *, kauth_cred_t, 124 int, int, int); 125 static int ip6_setmoptions(const struct sockopt *, struct ip6_moptions **); 126 static int ip6_getmoptions(struct sockopt *, struct ip6_moptions *); 127 static int ip6_copyexthdr(struct mbuf **, void *, int); 128 static int ip6_insertfraghdr(struct mbuf *, struct mbuf *, int, 129 struct ip6_frag **); 130 static int ip6_insert_jumboopt(struct ip6_exthdrs *, u_int32_t); 131 static int ip6_splithdr(struct mbuf *, struct ip6_exthdrs *); 132 static int ip6_getpmtu(struct route *, struct route *, struct ifnet *, 133 const struct in6_addr *, u_long *, int *); 134 static int copypktopts(struct ip6_pktopts *, struct ip6_pktopts *, int); 135 136 #ifdef RFC2292 137 static int ip6_pcbopts(struct ip6_pktopts **, struct socket *, struct sockopt *); 138 #endif 139 140 /* 141 * IP6 output. The packet in mbuf chain m contains a skeletal IP6 142 * header (with pri, len, nxt, hlim, src, dst). 143 * This function may modify ver and hlim only. 144 * The mbuf chain containing the packet will be freed. 145 * The mbuf opt, if present, will not be freed. 146 * 147 * type of "mtu": rt_rmx.rmx_mtu is u_long, ifnet.ifr_mtu is int, and 148 * nd_ifinfo.linkmtu is u_int32_t. so we use u_long to hold largest one, 149 * which is rt_rmx.rmx_mtu. 150 */ 151 int 152 ip6_output( 153 struct mbuf *m0, 154 struct ip6_pktopts *opt, 155 struct route *ro, 156 int flags, 157 struct ip6_moptions *im6o, 158 struct socket *so, 159 struct ifnet **ifpp /* XXX: just for statistics */ 160 ) 161 { 162 struct ip6_hdr *ip6, *mhip6; 163 struct ifnet *ifp, *origifp; 164 struct mbuf *m = m0; 165 int hlen, tlen, len, off; 166 bool tso; 167 struct route ip6route; 168 struct rtentry *rt = NULL; 169 const struct sockaddr_in6 *dst = NULL; 170 struct sockaddr_in6 src_sa, dst_sa; 171 int error = 0; 172 struct in6_ifaddr *ia = NULL; 173 u_long mtu; 174 int alwaysfrag, dontfrag; 175 u_int32_t optlen = 0, plen = 0, unfragpartlen = 0; 176 struct ip6_exthdrs exthdrs; 177 struct in6_addr finaldst, src0, dst0; 178 u_int32_t zone; 179 struct route *ro_pmtu = NULL; 180 int hdrsplit = 0; 181 int needipsec = 0; 182 #ifdef IPSEC 183 struct secpolicy *sp = NULL; 184 #endif 185 186 memset(&ip6route, 0, sizeof(ip6route)); 187 188 #ifdef DIAGNOSTIC 189 if ((m->m_flags & M_PKTHDR) == 0) 190 panic("ip6_output: no HDR"); 191 192 if ((m->m_pkthdr.csum_flags & 193 (M_CSUM_TCPv4|M_CSUM_UDPv4|M_CSUM_TSOv4)) != 0) { 194 panic("ip6_output: IPv4 checksum offload flags: %d", 195 m->m_pkthdr.csum_flags); 196 } 197 198 if ((m->m_pkthdr.csum_flags & (M_CSUM_TCPv6|M_CSUM_UDPv6)) == 199 (M_CSUM_TCPv6|M_CSUM_UDPv6)) { 200 panic("ip6_output: conflicting checksum offload flags: %d", 201 m->m_pkthdr.csum_flags); 202 } 203 #endif 204 205 M_CSUM_DATA_IPv6_HL_SET(m->m_pkthdr.csum_data, sizeof(struct ip6_hdr)); 206 207 #define MAKE_EXTHDR(hp, mp) \ 208 do { \ 209 if (hp) { \ 210 struct ip6_ext *eh = (struct ip6_ext *)(hp); \ 211 error = ip6_copyexthdr((mp), (void *)(hp), \ 212 ((eh)->ip6e_len + 1) << 3); \ 213 if (error) \ 214 goto freehdrs; \ 215 } \ 216 } while (/*CONSTCOND*/ 0) 217 218 memset(&exthdrs, 0, sizeof(exthdrs)); 219 if (opt) { 220 /* Hop-by-Hop options header */ 221 MAKE_EXTHDR(opt->ip6po_hbh, &exthdrs.ip6e_hbh); 222 /* Destination options header(1st part) */ 223 MAKE_EXTHDR(opt->ip6po_dest1, &exthdrs.ip6e_dest1); 224 /* Routing header */ 225 MAKE_EXTHDR(opt->ip6po_rthdr, &exthdrs.ip6e_rthdr); 226 /* Destination options header(2nd part) */ 227 MAKE_EXTHDR(opt->ip6po_dest2, &exthdrs.ip6e_dest2); 228 } 229 230 /* 231 * Calculate the total length of the extension header chain. 232 * Keep the length of the unfragmentable part for fragmentation. 233 */ 234 optlen = 0; 235 if (exthdrs.ip6e_hbh) optlen += exthdrs.ip6e_hbh->m_len; 236 if (exthdrs.ip6e_dest1) optlen += exthdrs.ip6e_dest1->m_len; 237 if (exthdrs.ip6e_rthdr) optlen += exthdrs.ip6e_rthdr->m_len; 238 unfragpartlen = optlen + sizeof(struct ip6_hdr); 239 /* NOTE: we don't add AH/ESP length here. do that later. */ 240 if (exthdrs.ip6e_dest2) optlen += exthdrs.ip6e_dest2->m_len; 241 242 #ifdef IPSEC 243 if (ipsec_used) { 244 /* Check the security policy (SP) for the packet */ 245 246 sp = ipsec6_check_policy(m, so, flags, &needipsec, &error); 247 if (error != 0) { 248 /* 249 * Hack: -EINVAL is used to signal that a packet 250 * should be silently discarded. This is typically 251 * because we asked key management for an SA and 252 * it was delayed (e.g. kicked up to IKE). 253 */ 254 if (error == -EINVAL) 255 error = 0; 256 goto freehdrs; 257 } 258 } 259 #endif /* IPSEC */ 260 261 262 if (needipsec && 263 (m->m_pkthdr.csum_flags & (M_CSUM_UDPv6|M_CSUM_TCPv6)) != 0) { 264 in6_delayed_cksum(m); 265 m->m_pkthdr.csum_flags &= ~(M_CSUM_UDPv6|M_CSUM_TCPv6); 266 } 267 268 269 /* 270 * If we need IPsec, or there is at least one extension header, 271 * separate IP6 header from the payload. 272 */ 273 if ((needipsec || optlen) && !hdrsplit) { 274 if ((error = ip6_splithdr(m, &exthdrs)) != 0) { 275 m = NULL; 276 goto freehdrs; 277 } 278 m = exthdrs.ip6e_ip6; 279 hdrsplit++; 280 } 281 282 /* adjust pointer */ 283 ip6 = mtod(m, struct ip6_hdr *); 284 285 /* adjust mbuf packet header length */ 286 m->m_pkthdr.len += optlen; 287 plen = m->m_pkthdr.len - sizeof(*ip6); 288 289 /* If this is a jumbo payload, insert a jumbo payload option. */ 290 if (plen > IPV6_MAXPACKET) { 291 if (!hdrsplit) { 292 if ((error = ip6_splithdr(m, &exthdrs)) != 0) { 293 m = NULL; 294 goto freehdrs; 295 } 296 m = exthdrs.ip6e_ip6; 297 hdrsplit++; 298 } 299 /* adjust pointer */ 300 ip6 = mtod(m, struct ip6_hdr *); 301 if ((error = ip6_insert_jumboopt(&exthdrs, plen)) != 0) 302 goto freehdrs; 303 optlen += 8; /* XXX JUMBOOPTLEN */ 304 ip6->ip6_plen = 0; 305 } else 306 ip6->ip6_plen = htons(plen); 307 308 /* 309 * Concatenate headers and fill in next header fields. 310 * Here we have, on "m" 311 * IPv6 payload 312 * and we insert headers accordingly. Finally, we should be getting: 313 * IPv6 hbh dest1 rthdr ah* [esp* dest2 payload] 314 * 315 * during the header composing process, "m" points to IPv6 header. 316 * "mprev" points to an extension header prior to esp. 317 */ 318 { 319 u_char *nexthdrp = &ip6->ip6_nxt; 320 struct mbuf *mprev = m; 321 322 /* 323 * we treat dest2 specially. this makes IPsec processing 324 * much easier. the goal here is to make mprev point the 325 * mbuf prior to dest2. 326 * 327 * result: IPv6 dest2 payload 328 * m and mprev will point to IPv6 header. 329 */ 330 if (exthdrs.ip6e_dest2) { 331 if (!hdrsplit) 332 panic("assumption failed: hdr not split"); 333 exthdrs.ip6e_dest2->m_next = m->m_next; 334 m->m_next = exthdrs.ip6e_dest2; 335 *mtod(exthdrs.ip6e_dest2, u_char *) = ip6->ip6_nxt; 336 ip6->ip6_nxt = IPPROTO_DSTOPTS; 337 } 338 339 #define MAKE_CHAIN(m, mp, p, i)\ 340 do {\ 341 if (m) {\ 342 if (!hdrsplit) \ 343 panic("assumption failed: hdr not split"); \ 344 *mtod((m), u_char *) = *(p);\ 345 *(p) = (i);\ 346 p = mtod((m), u_char *);\ 347 (m)->m_next = (mp)->m_next;\ 348 (mp)->m_next = (m);\ 349 (mp) = (m);\ 350 }\ 351 } while (/*CONSTCOND*/ 0) 352 /* 353 * result: IPv6 hbh dest1 rthdr dest2 payload 354 * m will point to IPv6 header. mprev will point to the 355 * extension header prior to dest2 (rthdr in the above case). 356 */ 357 MAKE_CHAIN(exthdrs.ip6e_hbh, mprev, nexthdrp, IPPROTO_HOPOPTS); 358 MAKE_CHAIN(exthdrs.ip6e_dest1, mprev, nexthdrp, 359 IPPROTO_DSTOPTS); 360 MAKE_CHAIN(exthdrs.ip6e_rthdr, mprev, nexthdrp, 361 IPPROTO_ROUTING); 362 363 M_CSUM_DATA_IPv6_HL_SET(m->m_pkthdr.csum_data, 364 sizeof(struct ip6_hdr) + optlen); 365 } 366 367 /* 368 * If there is a routing header, replace destination address field 369 * with the first hop of the routing header. 370 */ 371 if (exthdrs.ip6e_rthdr) { 372 struct ip6_rthdr *rh; 373 struct ip6_rthdr0 *rh0; 374 struct in6_addr *addr; 375 struct sockaddr_in6 sa; 376 377 rh = (struct ip6_rthdr *)(mtod(exthdrs.ip6e_rthdr, 378 struct ip6_rthdr *)); 379 finaldst = ip6->ip6_dst; 380 switch (rh->ip6r_type) { 381 case IPV6_RTHDR_TYPE_0: 382 rh0 = (struct ip6_rthdr0 *)rh; 383 addr = (struct in6_addr *)(rh0 + 1); 384 385 /* 386 * construct a sockaddr_in6 form of 387 * the first hop. 388 * 389 * XXX: we may not have enough 390 * information about its scope zone; 391 * there is no standard API to pass 392 * the information from the 393 * application. 394 */ 395 sockaddr_in6_init(&sa, addr, 0, 0, 0); 396 if ((error = sa6_embedscope(&sa, 397 ip6_use_defzone)) != 0) { 398 goto bad; 399 } 400 ip6->ip6_dst = sa.sin6_addr; 401 (void)memmove(&addr[0], &addr[1], 402 sizeof(struct in6_addr) * 403 (rh0->ip6r0_segleft - 1)); 404 addr[rh0->ip6r0_segleft - 1] = finaldst; 405 /* XXX */ 406 in6_clearscope(addr + rh0->ip6r0_segleft - 1); 407 break; 408 default: /* is it possible? */ 409 error = EINVAL; 410 goto bad; 411 } 412 } 413 414 /* Source address validation */ 415 if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) && 416 (flags & IPV6_UNSPECSRC) == 0) { 417 error = EOPNOTSUPP; 418 IP6_STATINC(IP6_STAT_BADSCOPE); 419 goto bad; 420 } 421 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) { 422 error = EOPNOTSUPP; 423 IP6_STATINC(IP6_STAT_BADSCOPE); 424 goto bad; 425 } 426 427 IP6_STATINC(IP6_STAT_LOCALOUT); 428 429 /* 430 * Route packet. 431 */ 432 /* initialize cached route */ 433 if (ro == NULL) { 434 ro = &ip6route; 435 } 436 ro_pmtu = ro; 437 if (opt && opt->ip6po_rthdr) 438 ro = &opt->ip6po_route; 439 440 /* 441 * if specified, try to fill in the traffic class field. 442 * do not override if a non-zero value is already set. 443 * we check the diffserv field and the ecn field separately. 444 */ 445 if (opt && opt->ip6po_tclass >= 0) { 446 int mask = 0; 447 448 if ((ip6->ip6_flow & htonl(0xfc << 20)) == 0) 449 mask |= 0xfc; 450 if ((ip6->ip6_flow & htonl(0x03 << 20)) == 0) 451 mask |= 0x03; 452 if (mask != 0) 453 ip6->ip6_flow |= htonl((opt->ip6po_tclass & mask) << 20); 454 } 455 456 /* fill in or override the hop limit field, if necessary. */ 457 if (opt && opt->ip6po_hlim != -1) 458 ip6->ip6_hlim = opt->ip6po_hlim & 0xff; 459 else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) { 460 if (im6o != NULL) 461 ip6->ip6_hlim = im6o->im6o_multicast_hlim; 462 else 463 ip6->ip6_hlim = ip6_defmcasthlim; 464 } 465 466 #ifdef IPSEC 467 if (needipsec) { 468 int s = splsoftnet(); 469 error = ipsec6_process_packet(m, sp->req); 470 471 /* 472 * Preserve KAME behaviour: ENOENT can be returned 473 * when an SA acquire is in progress. Don't propagate 474 * this to user-level; it confuses applications. 475 * XXX this will go away when the SADB is redone. 476 */ 477 if (error == ENOENT) 478 error = 0; 479 splx(s); 480 goto done; 481 } 482 #endif /* IPSEC */ 483 484 /* adjust pointer */ 485 ip6 = mtod(m, struct ip6_hdr *); 486 487 sockaddr_in6_init(&dst_sa, &ip6->ip6_dst, 0, 0, 0); 488 if ((error = in6_selectroute(&dst_sa, opt, im6o, ro, 489 &ifp, &rt, 0)) != 0) { 490 if (ifp != NULL) 491 in6_ifstat_inc(ifp, ifs6_out_discard); 492 goto bad; 493 } 494 if (rt == NULL) { 495 /* 496 * If in6_selectroute() does not return a route entry, 497 * dst may not have been updated. 498 */ 499 error = rtcache_setdst(ro, sin6tosa(&dst_sa)); 500 if (error) { 501 goto bad; 502 } 503 } 504 505 /* 506 * then rt (for unicast) and ifp must be non-NULL valid values. 507 */ 508 if ((flags & IPV6_FORWARDING) == 0) { 509 /* XXX: the FORWARDING flag can be set for mrouting. */ 510 in6_ifstat_inc(ifp, ifs6_out_request); 511 } 512 if (rt != NULL) { 513 ia = (struct in6_ifaddr *)(rt->rt_ifa); 514 rt->rt_use++; 515 } 516 517 /* 518 * The outgoing interface must be in the zone of source and 519 * destination addresses. We should use ia_ifp to support the 520 * case of sending packets to an address of our own. 521 */ 522 if (ia != NULL && ia->ia_ifp) 523 origifp = ia->ia_ifp; 524 else 525 origifp = ifp; 526 527 src0 = ip6->ip6_src; 528 if (in6_setscope(&src0, origifp, &zone)) 529 goto badscope; 530 sockaddr_in6_init(&src_sa, &ip6->ip6_src, 0, 0, 0); 531 if (sa6_recoverscope(&src_sa) || zone != src_sa.sin6_scope_id) 532 goto badscope; 533 534 dst0 = ip6->ip6_dst; 535 if (in6_setscope(&dst0, origifp, &zone)) 536 goto badscope; 537 /* re-initialize to be sure */ 538 sockaddr_in6_init(&dst_sa, &ip6->ip6_dst, 0, 0, 0); 539 if (sa6_recoverscope(&dst_sa) || zone != dst_sa.sin6_scope_id) 540 goto badscope; 541 542 /* scope check is done. */ 543 544 if (rt == NULL || IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) { 545 if (dst == NULL) 546 dst = satocsin6(rtcache_getdst(ro)); 547 KASSERT(dst != NULL); 548 } else if (opt && rtcache_validate(&opt->ip6po_nextroute) != NULL) { 549 /* 550 * The nexthop is explicitly specified by the 551 * application. We assume the next hop is an IPv6 552 * address. 553 */ 554 dst = (struct sockaddr_in6 *)opt->ip6po_nexthop; 555 } else if ((rt->rt_flags & RTF_GATEWAY)) 556 dst = (struct sockaddr_in6 *)rt->rt_gateway; 557 else if (dst == NULL) 558 dst = satocsin6(rtcache_getdst(ro)); 559 560 /* 561 * XXXXXX: original code follows: 562 */ 563 if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) 564 m->m_flags &= ~(M_BCAST | M_MCAST); /* just in case */ 565 else { 566 struct in6_multi *in6m; 567 568 m->m_flags = (m->m_flags & ~M_BCAST) | M_MCAST; 569 570 in6_ifstat_inc(ifp, ifs6_out_mcast); 571 572 /* 573 * Confirm that the outgoing interface supports multicast. 574 */ 575 if (!(ifp->if_flags & IFF_MULTICAST)) { 576 IP6_STATINC(IP6_STAT_NOROUTE); 577 in6_ifstat_inc(ifp, ifs6_out_discard); 578 error = ENETUNREACH; 579 goto bad; 580 } 581 582 IN6_LOOKUP_MULTI(ip6->ip6_dst, ifp, in6m); 583 if (in6m != NULL && 584 (im6o == NULL || im6o->im6o_multicast_loop)) { 585 /* 586 * If we belong to the destination multicast group 587 * on the outgoing interface, and the caller did not 588 * forbid loopback, loop back a copy. 589 */ 590 KASSERT(dst != NULL); 591 ip6_mloopback(ifp, m, dst); 592 } else { 593 /* 594 * If we are acting as a multicast router, perform 595 * multicast forwarding as if the packet had just 596 * arrived on the interface to which we are about 597 * to send. The multicast forwarding function 598 * recursively calls this function, using the 599 * IPV6_FORWARDING flag to prevent infinite recursion. 600 * 601 * Multicasts that are looped back by ip6_mloopback(), 602 * above, will be forwarded by the ip6_input() routine, 603 * if necessary. 604 */ 605 if (ip6_mrouter && (flags & IPV6_FORWARDING) == 0) { 606 if (ip6_mforward(ip6, ifp, m) != 0) { 607 m_freem(m); 608 goto done; 609 } 610 } 611 } 612 /* 613 * Multicasts with a hoplimit of zero may be looped back, 614 * above, but must not be transmitted on a network. 615 * Also, multicasts addressed to the loopback interface 616 * are not sent -- the above call to ip6_mloopback() will 617 * loop back a copy if this host actually belongs to the 618 * destination group on the loopback interface. 619 */ 620 if (ip6->ip6_hlim == 0 || (ifp->if_flags & IFF_LOOPBACK) || 621 IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst)) { 622 m_freem(m); 623 goto done; 624 } 625 } 626 627 /* 628 * Fill the outgoing inteface to tell the upper layer 629 * to increment per-interface statistics. 630 */ 631 if (ifpp) 632 *ifpp = ifp; 633 634 /* Determine path MTU. */ 635 if ((error = ip6_getpmtu(ro_pmtu, ro, ifp, &finaldst, &mtu, 636 &alwaysfrag)) != 0) 637 goto bad; 638 639 /* 640 * The caller of this function may specify to use the minimum MTU 641 * in some cases. 642 * An advanced API option (IPV6_USE_MIN_MTU) can also override MTU 643 * setting. The logic is a bit complicated; by default, unicast 644 * packets will follow path MTU while multicast packets will be sent at 645 * the minimum MTU. If IP6PO_MINMTU_ALL is specified, all packets 646 * including unicast ones will be sent at the minimum MTU. Multicast 647 * packets will always be sent at the minimum MTU unless 648 * IP6PO_MINMTU_DISABLE is explicitly specified. 649 * See RFC 3542 for more details. 650 */ 651 if (mtu > IPV6_MMTU) { 652 if ((flags & IPV6_MINMTU)) 653 mtu = IPV6_MMTU; 654 else if (opt && opt->ip6po_minmtu == IP6PO_MINMTU_ALL) 655 mtu = IPV6_MMTU; 656 else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) && 657 (opt == NULL || 658 opt->ip6po_minmtu != IP6PO_MINMTU_DISABLE)) { 659 mtu = IPV6_MMTU; 660 } 661 } 662 663 /* 664 * clear embedded scope identifiers if necessary. 665 * in6_clearscope will touch the addresses only when necessary. 666 */ 667 in6_clearscope(&ip6->ip6_src); 668 in6_clearscope(&ip6->ip6_dst); 669 670 /* 671 * If the outgoing packet contains a hop-by-hop options header, 672 * it must be examined and processed even by the source node. 673 * (RFC 2460, section 4.) 674 */ 675 if (ip6->ip6_nxt == IPV6_HOPOPTS) { 676 u_int32_t dummy1; /* XXX unused */ 677 u_int32_t dummy2; /* XXX unused */ 678 int hoff = sizeof(struct ip6_hdr); 679 680 if (ip6_hopopts_input(&dummy1, &dummy2, &m, &hoff)) { 681 /* m was already freed at this point */ 682 error = EINVAL;/* better error? */ 683 goto done; 684 } 685 686 ip6 = mtod(m, struct ip6_hdr *); 687 } 688 689 /* 690 * Run through list of hooks for output packets. 691 */ 692 if ((error = pfil_run_hooks(inet6_pfil_hook, &m, ifp, PFIL_OUT)) != 0) 693 goto done; 694 if (m == NULL) 695 goto done; 696 ip6 = mtod(m, struct ip6_hdr *); 697 698 /* 699 * Send the packet to the outgoing interface. 700 * If necessary, do IPv6 fragmentation before sending. 701 * 702 * the logic here is rather complex: 703 * 1: normal case (dontfrag == 0, alwaysfrag == 0) 704 * 1-a: send as is if tlen <= path mtu 705 * 1-b: fragment if tlen > path mtu 706 * 707 * 2: if user asks us not to fragment (dontfrag == 1) 708 * 2-a: send as is if tlen <= interface mtu 709 * 2-b: error if tlen > interface mtu 710 * 711 * 3: if we always need to attach fragment header (alwaysfrag == 1) 712 * always fragment 713 * 714 * 4: if dontfrag == 1 && alwaysfrag == 1 715 * error, as we cannot handle this conflicting request 716 */ 717 tlen = m->m_pkthdr.len; 718 tso = (m->m_pkthdr.csum_flags & M_CSUM_TSOv6) != 0; 719 if (opt && (opt->ip6po_flags & IP6PO_DONTFRAG)) 720 dontfrag = 1; 721 else 722 dontfrag = 0; 723 724 if (dontfrag && alwaysfrag) { /* case 4 */ 725 /* conflicting request - can't transmit */ 726 error = EMSGSIZE; 727 goto bad; 728 } 729 if (dontfrag && (!tso && tlen > IN6_LINKMTU(ifp))) { /* case 2-b */ 730 /* 731 * Even if the DONTFRAG option is specified, we cannot send the 732 * packet when the data length is larger than the MTU of the 733 * outgoing interface. 734 * Notify the error by sending IPV6_PATHMTU ancillary data as 735 * well as returning an error code (the latter is not described 736 * in the API spec.) 737 */ 738 u_int32_t mtu32; 739 struct ip6ctlparam ip6cp; 740 741 mtu32 = (u_int32_t)mtu; 742 memset(&ip6cp, 0, sizeof(ip6cp)); 743 ip6cp.ip6c_cmdarg = (void *)&mtu32; 744 pfctlinput2(PRC_MSGSIZE, 745 rtcache_getdst(ro_pmtu), &ip6cp); 746 747 error = EMSGSIZE; 748 goto bad; 749 } 750 751 /* 752 * transmit packet without fragmentation 753 */ 754 if (dontfrag || (!alwaysfrag && (tlen <= mtu || tso))) { 755 /* case 1-a and 2-a */ 756 struct in6_ifaddr *ia6; 757 int sw_csum; 758 759 ip6 = mtod(m, struct ip6_hdr *); 760 ia6 = in6_ifawithifp(ifp, &ip6->ip6_src); 761 if (ia6) { 762 /* Record statistics for this interface address. */ 763 ia6->ia_ifa.ifa_data.ifad_outbytes += m->m_pkthdr.len; 764 } 765 766 sw_csum = m->m_pkthdr.csum_flags & ~ifp->if_csum_flags_tx; 767 if ((sw_csum & (M_CSUM_UDPv6|M_CSUM_TCPv6)) != 0) { 768 if (IN6_NEED_CHECKSUM(ifp, 769 sw_csum & (M_CSUM_UDPv6|M_CSUM_TCPv6))) { 770 in6_delayed_cksum(m); 771 } 772 m->m_pkthdr.csum_flags &= ~(M_CSUM_UDPv6|M_CSUM_TCPv6); 773 } 774 775 KASSERT(dst != NULL); 776 if (__predict_true(!tso || 777 (ifp->if_capenable & IFCAP_TSOv6) != 0)) { 778 error = nd6_output(ifp, origifp, m, dst, rt); 779 } else { 780 error = ip6_tso_output(ifp, origifp, m, dst, rt); 781 } 782 goto done; 783 } 784 785 if (tso) { 786 error = EINVAL; /* XXX */ 787 goto bad; 788 } 789 790 /* 791 * try to fragment the packet. case 1-b and 3 792 */ 793 if (mtu < IPV6_MMTU) { 794 /* path MTU cannot be less than IPV6_MMTU */ 795 error = EMSGSIZE; 796 in6_ifstat_inc(ifp, ifs6_out_fragfail); 797 goto bad; 798 } else if (ip6->ip6_plen == 0) { 799 /* jumbo payload cannot be fragmented */ 800 error = EMSGSIZE; 801 in6_ifstat_inc(ifp, ifs6_out_fragfail); 802 goto bad; 803 } else { 804 struct mbuf **mnext, *m_frgpart; 805 struct ip6_frag *ip6f; 806 u_int32_t id = htonl(ip6_randomid()); 807 u_char nextproto; 808 #if 0 /* see below */ 809 struct ip6ctlparam ip6cp; 810 u_int32_t mtu32; 811 #endif 812 813 /* 814 * Too large for the destination or interface; 815 * fragment if possible. 816 * Must be able to put at least 8 bytes per fragment. 817 */ 818 hlen = unfragpartlen; 819 if (mtu > IPV6_MAXPACKET) 820 mtu = IPV6_MAXPACKET; 821 822 #if 0 823 /* 824 * It is believed this code is a leftover from the 825 * development of the IPV6_RECVPATHMTU sockopt and 826 * associated work to implement RFC3542. 827 * It's not entirely clear what the intent of the API 828 * is at this point, so disable this code for now. 829 * The IPV6_RECVPATHMTU sockopt and/or IPV6_DONTFRAG 830 * will send notifications if the application requests. 831 */ 832 833 /* Notify a proper path MTU to applications. */ 834 mtu32 = (u_int32_t)mtu; 835 memset(&ip6cp, 0, sizeof(ip6cp)); 836 ip6cp.ip6c_cmdarg = (void *)&mtu32; 837 pfctlinput2(PRC_MSGSIZE, 838 rtcache_getdst(ro_pmtu), &ip6cp); 839 #endif 840 841 len = (mtu - hlen - sizeof(struct ip6_frag)) & ~7; 842 if (len < 8) { 843 error = EMSGSIZE; 844 in6_ifstat_inc(ifp, ifs6_out_fragfail); 845 goto bad; 846 } 847 848 mnext = &m->m_nextpkt; 849 850 /* 851 * Change the next header field of the last header in the 852 * unfragmentable part. 853 */ 854 if (exthdrs.ip6e_rthdr) { 855 nextproto = *mtod(exthdrs.ip6e_rthdr, u_char *); 856 *mtod(exthdrs.ip6e_rthdr, u_char *) = IPPROTO_FRAGMENT; 857 } else if (exthdrs.ip6e_dest1) { 858 nextproto = *mtod(exthdrs.ip6e_dest1, u_char *); 859 *mtod(exthdrs.ip6e_dest1, u_char *) = IPPROTO_FRAGMENT; 860 } else if (exthdrs.ip6e_hbh) { 861 nextproto = *mtod(exthdrs.ip6e_hbh, u_char *); 862 *mtod(exthdrs.ip6e_hbh, u_char *) = IPPROTO_FRAGMENT; 863 } else { 864 nextproto = ip6->ip6_nxt; 865 ip6->ip6_nxt = IPPROTO_FRAGMENT; 866 } 867 868 if ((m->m_pkthdr.csum_flags & (M_CSUM_UDPv6|M_CSUM_TCPv6)) 869 != 0) { 870 if (IN6_NEED_CHECKSUM(ifp, 871 m->m_pkthdr.csum_flags & 872 (M_CSUM_UDPv6|M_CSUM_TCPv6))) { 873 in6_delayed_cksum(m); 874 } 875 m->m_pkthdr.csum_flags &= ~(M_CSUM_UDPv6|M_CSUM_TCPv6); 876 } 877 878 /* 879 * Loop through length of segment after first fragment, 880 * make new header and copy data of each part and link onto 881 * chain. 882 */ 883 m0 = m; 884 for (off = hlen; off < tlen; off += len) { 885 struct mbuf *mlast; 886 887 MGETHDR(m, M_DONTWAIT, MT_HEADER); 888 if (!m) { 889 error = ENOBUFS; 890 IP6_STATINC(IP6_STAT_ODROPPED); 891 goto sendorfree; 892 } 893 m->m_pkthdr.rcvif = NULL; 894 m->m_flags = m0->m_flags & M_COPYFLAGS; 895 *mnext = m; 896 mnext = &m->m_nextpkt; 897 m->m_data += max_linkhdr; 898 mhip6 = mtod(m, struct ip6_hdr *); 899 *mhip6 = *ip6; 900 m->m_len = sizeof(*mhip6); 901 /* 902 * ip6f must be valid if error is 0. But how 903 * can a compiler be expected to infer this? 904 */ 905 ip6f = NULL; 906 error = ip6_insertfraghdr(m0, m, hlen, &ip6f); 907 if (error) { 908 IP6_STATINC(IP6_STAT_ODROPPED); 909 goto sendorfree; 910 } 911 ip6f->ip6f_offlg = htons((u_int16_t)((off - hlen) & ~7)); 912 if (off + len >= tlen) 913 len = tlen - off; 914 else 915 ip6f->ip6f_offlg |= IP6F_MORE_FRAG; 916 mhip6->ip6_plen = htons((u_int16_t)(len + hlen + 917 sizeof(*ip6f) - sizeof(struct ip6_hdr))); 918 if ((m_frgpart = m_copy(m0, off, len)) == 0) { 919 error = ENOBUFS; 920 IP6_STATINC(IP6_STAT_ODROPPED); 921 goto sendorfree; 922 } 923 for (mlast = m; mlast->m_next; mlast = mlast->m_next) 924 ; 925 mlast->m_next = m_frgpart; 926 m->m_pkthdr.len = len + hlen + sizeof(*ip6f); 927 m->m_pkthdr.rcvif = NULL; 928 ip6f->ip6f_reserved = 0; 929 ip6f->ip6f_ident = id; 930 ip6f->ip6f_nxt = nextproto; 931 IP6_STATINC(IP6_STAT_OFRAGMENTS); 932 in6_ifstat_inc(ifp, ifs6_out_fragcreat); 933 } 934 935 in6_ifstat_inc(ifp, ifs6_out_fragok); 936 } 937 938 /* 939 * Remove leading garbages. 940 */ 941 sendorfree: 942 m = m0->m_nextpkt; 943 m0->m_nextpkt = 0; 944 m_freem(m0); 945 for (m0 = m; m; m = m0) { 946 m0 = m->m_nextpkt; 947 m->m_nextpkt = 0; 948 if (error == 0) { 949 struct in6_ifaddr *ia6; 950 ip6 = mtod(m, struct ip6_hdr *); 951 ia6 = in6_ifawithifp(ifp, &ip6->ip6_src); 952 if (ia6) { 953 /* 954 * Record statistics for this interface 955 * address. 956 */ 957 ia6->ia_ifa.ifa_data.ifad_outbytes += 958 m->m_pkthdr.len; 959 } 960 KASSERT(dst != NULL); 961 error = nd6_output(ifp, origifp, m, dst, rt); 962 } else 963 m_freem(m); 964 } 965 966 if (error == 0) 967 IP6_STATINC(IP6_STAT_FRAGMENTED); 968 969 done: 970 rtcache_free(&ip6route); 971 972 #ifdef IPSEC 973 if (sp != NULL) 974 KEY_FREESP(&sp); 975 #endif /* IPSEC */ 976 977 978 return (error); 979 980 freehdrs: 981 m_freem(exthdrs.ip6e_hbh); /* m_freem will check if mbuf is 0 */ 982 m_freem(exthdrs.ip6e_dest1); 983 m_freem(exthdrs.ip6e_rthdr); 984 m_freem(exthdrs.ip6e_dest2); 985 /* FALLTHROUGH */ 986 bad: 987 m_freem(m); 988 goto done; 989 badscope: 990 IP6_STATINC(IP6_STAT_BADSCOPE); 991 in6_ifstat_inc(origifp, ifs6_out_discard); 992 if (error == 0) 993 error = EHOSTUNREACH; /* XXX */ 994 goto bad; 995 } 996 997 static int 998 ip6_copyexthdr(struct mbuf **mp, void *hdr, int hlen) 999 { 1000 struct mbuf *m; 1001 1002 if (hlen > MCLBYTES) 1003 return (ENOBUFS); /* XXX */ 1004 1005 MGET(m, M_DONTWAIT, MT_DATA); 1006 if (!m) 1007 return (ENOBUFS); 1008 1009 if (hlen > MLEN) { 1010 MCLGET(m, M_DONTWAIT); 1011 if ((m->m_flags & M_EXT) == 0) { 1012 m_free(m); 1013 return (ENOBUFS); 1014 } 1015 } 1016 m->m_len = hlen; 1017 if (hdr) 1018 bcopy(hdr, mtod(m, void *), hlen); 1019 1020 *mp = m; 1021 return (0); 1022 } 1023 1024 /* 1025 * Process a delayed payload checksum calculation. 1026 */ 1027 void 1028 in6_delayed_cksum(struct mbuf *m) 1029 { 1030 uint16_t csum, offset; 1031 1032 KASSERT((m->m_pkthdr.csum_flags & (M_CSUM_UDPv6|M_CSUM_TCPv6)) != 0); 1033 KASSERT((~m->m_pkthdr.csum_flags & (M_CSUM_UDPv6|M_CSUM_TCPv6)) != 0); 1034 KASSERT((m->m_pkthdr.csum_flags 1035 & (M_CSUM_UDPv4|M_CSUM_TCPv4|M_CSUM_TSOv4)) == 0); 1036 1037 offset = M_CSUM_DATA_IPv6_HL(m->m_pkthdr.csum_data); 1038 csum = in6_cksum(m, 0, offset, m->m_pkthdr.len - offset); 1039 if (csum == 0 && (m->m_pkthdr.csum_flags & M_CSUM_UDPv6) != 0) { 1040 csum = 0xffff; 1041 } 1042 1043 offset += M_CSUM_DATA_IPv6_OFFSET(m->m_pkthdr.csum_data); 1044 if ((offset + sizeof(csum)) > m->m_len) { 1045 m_copyback(m, offset, sizeof(csum), &csum); 1046 } else { 1047 *(uint16_t *)(mtod(m, char *) + offset) = csum; 1048 } 1049 } 1050 1051 /* 1052 * Insert jumbo payload option. 1053 */ 1054 static int 1055 ip6_insert_jumboopt(struct ip6_exthdrs *exthdrs, u_int32_t plen) 1056 { 1057 struct mbuf *mopt; 1058 u_int8_t *optbuf; 1059 u_int32_t v; 1060 1061 #define JUMBOOPTLEN 8 /* length of jumbo payload option and padding */ 1062 1063 /* 1064 * If there is no hop-by-hop options header, allocate new one. 1065 * If there is one but it doesn't have enough space to store the 1066 * jumbo payload option, allocate a cluster to store the whole options. 1067 * Otherwise, use it to store the options. 1068 */ 1069 if (exthdrs->ip6e_hbh == 0) { 1070 MGET(mopt, M_DONTWAIT, MT_DATA); 1071 if (mopt == 0) 1072 return (ENOBUFS); 1073 mopt->m_len = JUMBOOPTLEN; 1074 optbuf = mtod(mopt, u_int8_t *); 1075 optbuf[1] = 0; /* = ((JUMBOOPTLEN) >> 3) - 1 */ 1076 exthdrs->ip6e_hbh = mopt; 1077 } else { 1078 struct ip6_hbh *hbh; 1079 1080 mopt = exthdrs->ip6e_hbh; 1081 if (M_TRAILINGSPACE(mopt) < JUMBOOPTLEN) { 1082 /* 1083 * XXX assumption: 1084 * - exthdrs->ip6e_hbh is not referenced from places 1085 * other than exthdrs. 1086 * - exthdrs->ip6e_hbh is not an mbuf chain. 1087 */ 1088 int oldoptlen = mopt->m_len; 1089 struct mbuf *n; 1090 1091 /* 1092 * XXX: give up if the whole (new) hbh header does 1093 * not fit even in an mbuf cluster. 1094 */ 1095 if (oldoptlen + JUMBOOPTLEN > MCLBYTES) 1096 return (ENOBUFS); 1097 1098 /* 1099 * As a consequence, we must always prepare a cluster 1100 * at this point. 1101 */ 1102 MGET(n, M_DONTWAIT, MT_DATA); 1103 if (n) { 1104 MCLGET(n, M_DONTWAIT); 1105 if ((n->m_flags & M_EXT) == 0) { 1106 m_freem(n); 1107 n = NULL; 1108 } 1109 } 1110 if (!n) 1111 return (ENOBUFS); 1112 n->m_len = oldoptlen + JUMBOOPTLEN; 1113 bcopy(mtod(mopt, void *), mtod(n, void *), 1114 oldoptlen); 1115 optbuf = mtod(n, u_int8_t *) + oldoptlen; 1116 m_freem(mopt); 1117 mopt = exthdrs->ip6e_hbh = n; 1118 } else { 1119 optbuf = mtod(mopt, u_int8_t *) + mopt->m_len; 1120 mopt->m_len += JUMBOOPTLEN; 1121 } 1122 optbuf[0] = IP6OPT_PADN; 1123 optbuf[1] = 0; 1124 1125 /* 1126 * Adjust the header length according to the pad and 1127 * the jumbo payload option. 1128 */ 1129 hbh = mtod(mopt, struct ip6_hbh *); 1130 hbh->ip6h_len += (JUMBOOPTLEN >> 3); 1131 } 1132 1133 /* fill in the option. */ 1134 optbuf[2] = IP6OPT_JUMBO; 1135 optbuf[3] = 4; 1136 v = (u_int32_t)htonl(plen + JUMBOOPTLEN); 1137 bcopy(&v, &optbuf[4], sizeof(u_int32_t)); 1138 1139 /* finally, adjust the packet header length */ 1140 exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN; 1141 1142 return (0); 1143 #undef JUMBOOPTLEN 1144 } 1145 1146 /* 1147 * Insert fragment header and copy unfragmentable header portions. 1148 * 1149 * *frghdrp will not be read, and it is guaranteed that either an 1150 * error is returned or that *frghdrp will point to space allocated 1151 * for the fragment header. 1152 */ 1153 static int 1154 ip6_insertfraghdr(struct mbuf *m0, struct mbuf *m, int hlen, 1155 struct ip6_frag **frghdrp) 1156 { 1157 struct mbuf *n, *mlast; 1158 1159 if (hlen > sizeof(struct ip6_hdr)) { 1160 n = m_copym(m0, sizeof(struct ip6_hdr), 1161 hlen - sizeof(struct ip6_hdr), M_DONTWAIT); 1162 if (n == 0) 1163 return (ENOBUFS); 1164 m->m_next = n; 1165 } else 1166 n = m; 1167 1168 /* Search for the last mbuf of unfragmentable part. */ 1169 for (mlast = n; mlast->m_next; mlast = mlast->m_next) 1170 ; 1171 1172 if ((mlast->m_flags & M_EXT) == 0 && 1173 M_TRAILINGSPACE(mlast) >= sizeof(struct ip6_frag)) { 1174 /* use the trailing space of the last mbuf for the fragment hdr */ 1175 *frghdrp = (struct ip6_frag *)(mtod(mlast, char *) + 1176 mlast->m_len); 1177 mlast->m_len += sizeof(struct ip6_frag); 1178 m->m_pkthdr.len += sizeof(struct ip6_frag); 1179 } else { 1180 /* allocate a new mbuf for the fragment header */ 1181 struct mbuf *mfrg; 1182 1183 MGET(mfrg, M_DONTWAIT, MT_DATA); 1184 if (mfrg == 0) 1185 return (ENOBUFS); 1186 mfrg->m_len = sizeof(struct ip6_frag); 1187 *frghdrp = mtod(mfrg, struct ip6_frag *); 1188 mlast->m_next = mfrg; 1189 } 1190 1191 return (0); 1192 } 1193 1194 static int 1195 ip6_getpmtu(struct route *ro_pmtu, struct route *ro, struct ifnet *ifp, 1196 const struct in6_addr *dst, u_long *mtup, int *alwaysfragp) 1197 { 1198 struct rtentry *rt; 1199 u_int32_t mtu = 0; 1200 int alwaysfrag = 0; 1201 int error = 0; 1202 1203 if (ro_pmtu != ro) { 1204 union { 1205 struct sockaddr dst; 1206 struct sockaddr_in6 dst6; 1207 } u; 1208 1209 /* The first hop and the final destination may differ. */ 1210 sockaddr_in6_init(&u.dst6, dst, 0, 0, 0); 1211 rt = rtcache_lookup(ro_pmtu, &u.dst); 1212 } else 1213 rt = rtcache_validate(ro_pmtu); 1214 if (rt != NULL) { 1215 u_int32_t ifmtu; 1216 1217 if (ifp == NULL) 1218 ifp = rt->rt_ifp; 1219 ifmtu = IN6_LINKMTU(ifp); 1220 mtu = rt->rt_rmx.rmx_mtu; 1221 if (mtu == 0) 1222 mtu = ifmtu; 1223 else if (mtu < IPV6_MMTU) { 1224 /* 1225 * RFC2460 section 5, last paragraph: 1226 * if we record ICMPv6 too big message with 1227 * mtu < IPV6_MMTU, transmit packets sized IPV6_MMTU 1228 * or smaller, with fragment header attached. 1229 * (fragment header is needed regardless from the 1230 * packet size, for translators to identify packets) 1231 */ 1232 alwaysfrag = 1; 1233 mtu = IPV6_MMTU; 1234 } else if (mtu > ifmtu) { 1235 /* 1236 * The MTU on the route is larger than the MTU on 1237 * the interface! This shouldn't happen, unless the 1238 * MTU of the interface has been changed after the 1239 * interface was brought up. Change the MTU in the 1240 * route to match the interface MTU (as long as the 1241 * field isn't locked). 1242 */ 1243 mtu = ifmtu; 1244 if (!(rt->rt_rmx.rmx_locks & RTV_MTU)) 1245 rt->rt_rmx.rmx_mtu = mtu; 1246 } 1247 } else if (ifp) { 1248 mtu = IN6_LINKMTU(ifp); 1249 } else 1250 error = EHOSTUNREACH; /* XXX */ 1251 1252 *mtup = mtu; 1253 if (alwaysfragp) 1254 *alwaysfragp = alwaysfrag; 1255 return (error); 1256 } 1257 1258 /* 1259 * IP6 socket option processing. 1260 */ 1261 int 1262 ip6_ctloutput(int op, struct socket *so, struct sockopt *sopt) 1263 { 1264 int optdatalen, uproto; 1265 void *optdata; 1266 struct in6pcb *in6p = sotoin6pcb(so); 1267 int error, optval; 1268 int level, optname; 1269 1270 KASSERT(sopt != NULL); 1271 1272 level = sopt->sopt_level; 1273 optname = sopt->sopt_name; 1274 1275 error = optval = 0; 1276 uproto = (int)so->so_proto->pr_protocol; 1277 1278 if (level != IPPROTO_IPV6) { 1279 return ENOPROTOOPT; 1280 } 1281 switch (op) { 1282 case PRCO_SETOPT: 1283 switch (optname) { 1284 #ifdef RFC2292 1285 case IPV6_2292PKTOPTIONS: 1286 error = ip6_pcbopts(&in6p->in6p_outputopts, so, sopt); 1287 break; 1288 #endif 1289 1290 /* 1291 * Use of some Hop-by-Hop options or some 1292 * Destination options, might require special 1293 * privilege. That is, normal applications 1294 * (without special privilege) might be forbidden 1295 * from setting certain options in outgoing packets, 1296 * and might never see certain options in received 1297 * packets. [RFC 2292 Section 6] 1298 * KAME specific note: 1299 * KAME prevents non-privileged users from sending or 1300 * receiving ANY hbh/dst options in order to avoid 1301 * overhead of parsing options in the kernel. 1302 */ 1303 case IPV6_RECVHOPOPTS: 1304 case IPV6_RECVDSTOPTS: 1305 case IPV6_RECVRTHDRDSTOPTS: 1306 error = kauth_authorize_network(kauth_cred_get(), 1307 KAUTH_NETWORK_IPV6, KAUTH_REQ_NETWORK_IPV6_HOPBYHOP, 1308 NULL, NULL, NULL); 1309 if (error) 1310 break; 1311 /* FALLTHROUGH */ 1312 case IPV6_UNICAST_HOPS: 1313 case IPV6_HOPLIMIT: 1314 case IPV6_FAITH: 1315 1316 case IPV6_RECVPKTINFO: 1317 case IPV6_RECVHOPLIMIT: 1318 case IPV6_RECVRTHDR: 1319 case IPV6_RECVPATHMTU: 1320 case IPV6_RECVTCLASS: 1321 case IPV6_V6ONLY: 1322 error = sockopt_getint(sopt, &optval); 1323 if (error) 1324 break; 1325 switch (optname) { 1326 case IPV6_UNICAST_HOPS: 1327 if (optval < -1 || optval >= 256) 1328 error = EINVAL; 1329 else { 1330 /* -1 = kernel default */ 1331 in6p->in6p_hops = optval; 1332 } 1333 break; 1334 #define OPTSET(bit) \ 1335 do { \ 1336 if (optval) \ 1337 in6p->in6p_flags |= (bit); \ 1338 else \ 1339 in6p->in6p_flags &= ~(bit); \ 1340 } while (/*CONSTCOND*/ 0) 1341 1342 #ifdef RFC2292 1343 #define OPTSET2292(bit) \ 1344 do { \ 1345 in6p->in6p_flags |= IN6P_RFC2292; \ 1346 if (optval) \ 1347 in6p->in6p_flags |= (bit); \ 1348 else \ 1349 in6p->in6p_flags &= ~(bit); \ 1350 } while (/*CONSTCOND*/ 0) 1351 #endif 1352 1353 #define OPTBIT(bit) (in6p->in6p_flags & (bit) ? 1 : 0) 1354 1355 case IPV6_RECVPKTINFO: 1356 #ifdef RFC2292 1357 /* cannot mix with RFC2292 */ 1358 if (OPTBIT(IN6P_RFC2292)) { 1359 error = EINVAL; 1360 break; 1361 } 1362 #endif 1363 OPTSET(IN6P_PKTINFO); 1364 break; 1365 1366 case IPV6_HOPLIMIT: 1367 { 1368 struct ip6_pktopts **optp; 1369 1370 #ifdef RFC2292 1371 /* cannot mix with RFC2292 */ 1372 if (OPTBIT(IN6P_RFC2292)) { 1373 error = EINVAL; 1374 break; 1375 } 1376 #endif 1377 optp = &in6p->in6p_outputopts; 1378 error = ip6_pcbopt(IPV6_HOPLIMIT, 1379 (u_char *)&optval, 1380 sizeof(optval), 1381 optp, 1382 kauth_cred_get(), uproto); 1383 break; 1384 } 1385 1386 case IPV6_RECVHOPLIMIT: 1387 #ifdef RFC2292 1388 /* cannot mix with RFC2292 */ 1389 if (OPTBIT(IN6P_RFC2292)) { 1390 error = EINVAL; 1391 break; 1392 } 1393 #endif 1394 OPTSET(IN6P_HOPLIMIT); 1395 break; 1396 1397 case IPV6_RECVHOPOPTS: 1398 #ifdef RFC2292 1399 /* cannot mix with RFC2292 */ 1400 if (OPTBIT(IN6P_RFC2292)) { 1401 error = EINVAL; 1402 break; 1403 } 1404 #endif 1405 OPTSET(IN6P_HOPOPTS); 1406 break; 1407 1408 case IPV6_RECVDSTOPTS: 1409 #ifdef RFC2292 1410 /* cannot mix with RFC2292 */ 1411 if (OPTBIT(IN6P_RFC2292)) { 1412 error = EINVAL; 1413 break; 1414 } 1415 #endif 1416 OPTSET(IN6P_DSTOPTS); 1417 break; 1418 1419 case IPV6_RECVRTHDRDSTOPTS: 1420 #ifdef RFC2292 1421 /* cannot mix with RFC2292 */ 1422 if (OPTBIT(IN6P_RFC2292)) { 1423 error = EINVAL; 1424 break; 1425 } 1426 #endif 1427 OPTSET(IN6P_RTHDRDSTOPTS); 1428 break; 1429 1430 case IPV6_RECVRTHDR: 1431 #ifdef RFC2292 1432 /* cannot mix with RFC2292 */ 1433 if (OPTBIT(IN6P_RFC2292)) { 1434 error = EINVAL; 1435 break; 1436 } 1437 #endif 1438 OPTSET(IN6P_RTHDR); 1439 break; 1440 1441 case IPV6_FAITH: 1442 OPTSET(IN6P_FAITH); 1443 break; 1444 1445 case IPV6_RECVPATHMTU: 1446 /* 1447 * We ignore this option for TCP 1448 * sockets. 1449 * (RFC3542 leaves this case 1450 * unspecified.) 1451 */ 1452 if (uproto != IPPROTO_TCP) 1453 OPTSET(IN6P_MTU); 1454 break; 1455 1456 case IPV6_V6ONLY: 1457 /* 1458 * make setsockopt(IPV6_V6ONLY) 1459 * available only prior to bind(2). 1460 * see ipng mailing list, Jun 22 2001. 1461 */ 1462 if (in6p->in6p_lport || 1463 !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr)) { 1464 error = EINVAL; 1465 break; 1466 } 1467 #ifdef INET6_BINDV6ONLY 1468 if (!optval) 1469 error = EINVAL; 1470 #else 1471 OPTSET(IN6P_IPV6_V6ONLY); 1472 #endif 1473 break; 1474 case IPV6_RECVTCLASS: 1475 #ifdef RFC2292 1476 /* cannot mix with RFC2292 XXX */ 1477 if (OPTBIT(IN6P_RFC2292)) { 1478 error = EINVAL; 1479 break; 1480 } 1481 #endif 1482 OPTSET(IN6P_TCLASS); 1483 break; 1484 1485 } 1486 break; 1487 1488 case IPV6_OTCLASS: 1489 { 1490 struct ip6_pktopts **optp; 1491 u_int8_t tclass; 1492 1493 error = sockopt_get(sopt, &tclass, sizeof(tclass)); 1494 if (error) 1495 break; 1496 optp = &in6p->in6p_outputopts; 1497 error = ip6_pcbopt(optname, 1498 (u_char *)&tclass, 1499 sizeof(tclass), 1500 optp, 1501 kauth_cred_get(), uproto); 1502 break; 1503 } 1504 1505 case IPV6_TCLASS: 1506 case IPV6_DONTFRAG: 1507 case IPV6_USE_MIN_MTU: 1508 error = sockopt_getint(sopt, &optval); 1509 if (error) 1510 break; 1511 { 1512 struct ip6_pktopts **optp; 1513 optp = &in6p->in6p_outputopts; 1514 error = ip6_pcbopt(optname, 1515 (u_char *)&optval, 1516 sizeof(optval), 1517 optp, 1518 kauth_cred_get(), uproto); 1519 break; 1520 } 1521 1522 #ifdef RFC2292 1523 case IPV6_2292PKTINFO: 1524 case IPV6_2292HOPLIMIT: 1525 case IPV6_2292HOPOPTS: 1526 case IPV6_2292DSTOPTS: 1527 case IPV6_2292RTHDR: 1528 /* RFC 2292 */ 1529 error = sockopt_getint(sopt, &optval); 1530 if (error) 1531 break; 1532 1533 switch (optname) { 1534 case IPV6_2292PKTINFO: 1535 OPTSET2292(IN6P_PKTINFO); 1536 break; 1537 case IPV6_2292HOPLIMIT: 1538 OPTSET2292(IN6P_HOPLIMIT); 1539 break; 1540 case IPV6_2292HOPOPTS: 1541 /* 1542 * Check super-user privilege. 1543 * See comments for IPV6_RECVHOPOPTS. 1544 */ 1545 error = 1546 kauth_authorize_network(kauth_cred_get(), 1547 KAUTH_NETWORK_IPV6, 1548 KAUTH_REQ_NETWORK_IPV6_HOPBYHOP, NULL, 1549 NULL, NULL); 1550 if (error) 1551 return (error); 1552 OPTSET2292(IN6P_HOPOPTS); 1553 break; 1554 case IPV6_2292DSTOPTS: 1555 error = 1556 kauth_authorize_network(kauth_cred_get(), 1557 KAUTH_NETWORK_IPV6, 1558 KAUTH_REQ_NETWORK_IPV6_HOPBYHOP, NULL, 1559 NULL, NULL); 1560 if (error) 1561 return (error); 1562 OPTSET2292(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); /* XXX */ 1563 break; 1564 case IPV6_2292RTHDR: 1565 OPTSET2292(IN6P_RTHDR); 1566 break; 1567 } 1568 break; 1569 #endif 1570 case IPV6_PKTINFO: 1571 case IPV6_HOPOPTS: 1572 case IPV6_RTHDR: 1573 case IPV6_DSTOPTS: 1574 case IPV6_RTHDRDSTOPTS: 1575 case IPV6_NEXTHOP: { 1576 /* new advanced API (RFC3542) */ 1577 void *optbuf; 1578 int optbuflen; 1579 struct ip6_pktopts **optp; 1580 1581 #ifdef RFC2292 1582 /* cannot mix with RFC2292 */ 1583 if (OPTBIT(IN6P_RFC2292)) { 1584 error = EINVAL; 1585 break; 1586 } 1587 #endif 1588 1589 optbuflen = sopt->sopt_size; 1590 optbuf = malloc(optbuflen, M_IP6OPT, M_NOWAIT); 1591 if (optbuf == NULL) { 1592 error = ENOBUFS; 1593 break; 1594 } 1595 1596 error = sockopt_get(sopt, optbuf, optbuflen); 1597 if (error) { 1598 free(optbuf, M_IP6OPT); 1599 break; 1600 } 1601 optp = &in6p->in6p_outputopts; 1602 error = ip6_pcbopt(optname, optbuf, optbuflen, 1603 optp, kauth_cred_get(), uproto); 1604 1605 free(optbuf, M_IP6OPT); 1606 break; 1607 } 1608 #undef OPTSET 1609 1610 case IPV6_MULTICAST_IF: 1611 case IPV6_MULTICAST_HOPS: 1612 case IPV6_MULTICAST_LOOP: 1613 case IPV6_JOIN_GROUP: 1614 case IPV6_LEAVE_GROUP: 1615 error = ip6_setmoptions(sopt, &in6p->in6p_moptions); 1616 break; 1617 1618 case IPV6_PORTRANGE: 1619 error = sockopt_getint(sopt, &optval); 1620 if (error) 1621 break; 1622 1623 switch (optval) { 1624 case IPV6_PORTRANGE_DEFAULT: 1625 in6p->in6p_flags &= ~(IN6P_LOWPORT); 1626 in6p->in6p_flags &= ~(IN6P_HIGHPORT); 1627 break; 1628 1629 case IPV6_PORTRANGE_HIGH: 1630 in6p->in6p_flags &= ~(IN6P_LOWPORT); 1631 in6p->in6p_flags |= IN6P_HIGHPORT; 1632 break; 1633 1634 case IPV6_PORTRANGE_LOW: 1635 in6p->in6p_flags &= ~(IN6P_HIGHPORT); 1636 in6p->in6p_flags |= IN6P_LOWPORT; 1637 break; 1638 1639 default: 1640 error = EINVAL; 1641 break; 1642 } 1643 break; 1644 1645 case IPV6_PORTALGO: 1646 error = sockopt_getint(sopt, &optval); 1647 if (error) 1648 break; 1649 1650 error = portalgo_algo_index_select( 1651 (struct inpcb_hdr *)in6p, optval); 1652 break; 1653 1654 #if defined(IPSEC) 1655 case IPV6_IPSEC_POLICY: 1656 if (ipsec_enabled) { 1657 error = ipsec6_set_policy(in6p, optname, 1658 sopt->sopt_data, sopt->sopt_size, 1659 kauth_cred_get()); 1660 break; 1661 } 1662 /*FALLTHROUGH*/ 1663 #endif /* IPSEC */ 1664 1665 default: 1666 error = ENOPROTOOPT; 1667 break; 1668 } 1669 break; 1670 1671 case PRCO_GETOPT: 1672 switch (optname) { 1673 #ifdef RFC2292 1674 case IPV6_2292PKTOPTIONS: 1675 /* 1676 * RFC3542 (effectively) deprecated the 1677 * semantics of the 2292-style pktoptions. 1678 * Since it was not reliable in nature (i.e., 1679 * applications had to expect the lack of some 1680 * information after all), it would make sense 1681 * to simplify this part by always returning 1682 * empty data. 1683 */ 1684 break; 1685 #endif 1686 1687 case IPV6_RECVHOPOPTS: 1688 case IPV6_RECVDSTOPTS: 1689 case IPV6_RECVRTHDRDSTOPTS: 1690 case IPV6_UNICAST_HOPS: 1691 case IPV6_RECVPKTINFO: 1692 case IPV6_RECVHOPLIMIT: 1693 case IPV6_RECVRTHDR: 1694 case IPV6_RECVPATHMTU: 1695 1696 case IPV6_FAITH: 1697 case IPV6_V6ONLY: 1698 case IPV6_PORTRANGE: 1699 case IPV6_RECVTCLASS: 1700 switch (optname) { 1701 1702 case IPV6_RECVHOPOPTS: 1703 optval = OPTBIT(IN6P_HOPOPTS); 1704 break; 1705 1706 case IPV6_RECVDSTOPTS: 1707 optval = OPTBIT(IN6P_DSTOPTS); 1708 break; 1709 1710 case IPV6_RECVRTHDRDSTOPTS: 1711 optval = OPTBIT(IN6P_RTHDRDSTOPTS); 1712 break; 1713 1714 case IPV6_UNICAST_HOPS: 1715 optval = in6p->in6p_hops; 1716 break; 1717 1718 case IPV6_RECVPKTINFO: 1719 optval = OPTBIT(IN6P_PKTINFO); 1720 break; 1721 1722 case IPV6_RECVHOPLIMIT: 1723 optval = OPTBIT(IN6P_HOPLIMIT); 1724 break; 1725 1726 case IPV6_RECVRTHDR: 1727 optval = OPTBIT(IN6P_RTHDR); 1728 break; 1729 1730 case IPV6_RECVPATHMTU: 1731 optval = OPTBIT(IN6P_MTU); 1732 break; 1733 1734 case IPV6_FAITH: 1735 optval = OPTBIT(IN6P_FAITH); 1736 break; 1737 1738 case IPV6_V6ONLY: 1739 optval = OPTBIT(IN6P_IPV6_V6ONLY); 1740 break; 1741 1742 case IPV6_PORTRANGE: 1743 { 1744 int flags; 1745 flags = in6p->in6p_flags; 1746 if (flags & IN6P_HIGHPORT) 1747 optval = IPV6_PORTRANGE_HIGH; 1748 else if (flags & IN6P_LOWPORT) 1749 optval = IPV6_PORTRANGE_LOW; 1750 else 1751 optval = 0; 1752 break; 1753 } 1754 case IPV6_RECVTCLASS: 1755 optval = OPTBIT(IN6P_TCLASS); 1756 break; 1757 1758 } 1759 if (error) 1760 break; 1761 error = sockopt_setint(sopt, optval); 1762 break; 1763 1764 case IPV6_PATHMTU: 1765 { 1766 u_long pmtu = 0; 1767 struct ip6_mtuinfo mtuinfo; 1768 struct route *ro = &in6p->in6p_route; 1769 1770 if (!(so->so_state & SS_ISCONNECTED)) 1771 return (ENOTCONN); 1772 /* 1773 * XXX: we dot not consider the case of source 1774 * routing, or optional information to specify 1775 * the outgoing interface. 1776 */ 1777 error = ip6_getpmtu(ro, NULL, NULL, 1778 &in6p->in6p_faddr, &pmtu, NULL); 1779 if (error) 1780 break; 1781 if (pmtu > IPV6_MAXPACKET) 1782 pmtu = IPV6_MAXPACKET; 1783 1784 memset(&mtuinfo, 0, sizeof(mtuinfo)); 1785 mtuinfo.ip6m_mtu = (u_int32_t)pmtu; 1786 optdata = (void *)&mtuinfo; 1787 optdatalen = sizeof(mtuinfo); 1788 if (optdatalen > MCLBYTES) 1789 return (EMSGSIZE); /* XXX */ 1790 error = sockopt_set(sopt, optdata, optdatalen); 1791 break; 1792 } 1793 1794 #ifdef RFC2292 1795 case IPV6_2292PKTINFO: 1796 case IPV6_2292HOPLIMIT: 1797 case IPV6_2292HOPOPTS: 1798 case IPV6_2292RTHDR: 1799 case IPV6_2292DSTOPTS: 1800 switch (optname) { 1801 case IPV6_2292PKTINFO: 1802 optval = OPTBIT(IN6P_PKTINFO); 1803 break; 1804 case IPV6_2292HOPLIMIT: 1805 optval = OPTBIT(IN6P_HOPLIMIT); 1806 break; 1807 case IPV6_2292HOPOPTS: 1808 optval = OPTBIT(IN6P_HOPOPTS); 1809 break; 1810 case IPV6_2292RTHDR: 1811 optval = OPTBIT(IN6P_RTHDR); 1812 break; 1813 case IPV6_2292DSTOPTS: 1814 optval = OPTBIT(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); 1815 break; 1816 } 1817 error = sockopt_setint(sopt, optval); 1818 break; 1819 #endif 1820 case IPV6_PKTINFO: 1821 case IPV6_HOPOPTS: 1822 case IPV6_RTHDR: 1823 case IPV6_DSTOPTS: 1824 case IPV6_RTHDRDSTOPTS: 1825 case IPV6_NEXTHOP: 1826 case IPV6_OTCLASS: 1827 case IPV6_TCLASS: 1828 case IPV6_DONTFRAG: 1829 case IPV6_USE_MIN_MTU: 1830 error = ip6_getpcbopt(in6p->in6p_outputopts, 1831 optname, sopt); 1832 break; 1833 1834 case IPV6_MULTICAST_IF: 1835 case IPV6_MULTICAST_HOPS: 1836 case IPV6_MULTICAST_LOOP: 1837 case IPV6_JOIN_GROUP: 1838 case IPV6_LEAVE_GROUP: 1839 error = ip6_getmoptions(sopt, in6p->in6p_moptions); 1840 break; 1841 1842 case IPV6_PORTALGO: 1843 optval = ((struct inpcb_hdr *)in6p)->inph_portalgo; 1844 error = sockopt_setint(sopt, optval); 1845 break; 1846 1847 #if defined(IPSEC) 1848 case IPV6_IPSEC_POLICY: 1849 if (ipsec_used) { 1850 struct mbuf *m = NULL; 1851 1852 /* 1853 * XXX: this will return EINVAL as sopt is 1854 * empty 1855 */ 1856 error = ipsec6_get_policy(in6p, sopt->sopt_data, 1857 sopt->sopt_size, &m); 1858 if (!error) 1859 error = sockopt_setmbuf(sopt, m); 1860 break; 1861 } 1862 /*FALLTHROUGH*/ 1863 #endif /* IPSEC */ 1864 1865 default: 1866 error = ENOPROTOOPT; 1867 break; 1868 } 1869 break; 1870 } 1871 return (error); 1872 } 1873 1874 int 1875 ip6_raw_ctloutput(int op, struct socket *so, struct sockopt *sopt) 1876 { 1877 int error = 0, optval; 1878 const int icmp6off = offsetof(struct icmp6_hdr, icmp6_cksum); 1879 struct in6pcb *in6p = sotoin6pcb(so); 1880 int level, optname; 1881 1882 KASSERT(sopt != NULL); 1883 1884 level = sopt->sopt_level; 1885 optname = sopt->sopt_name; 1886 1887 if (level != IPPROTO_IPV6) { 1888 return ENOPROTOOPT; 1889 } 1890 1891 switch (optname) { 1892 case IPV6_CHECKSUM: 1893 /* 1894 * For ICMPv6 sockets, no modification allowed for checksum 1895 * offset, permit "no change" values to help existing apps. 1896 * 1897 * XXX RFC3542 says: "An attempt to set IPV6_CHECKSUM 1898 * for an ICMPv6 socket will fail." The current 1899 * behavior does not meet RFC3542. 1900 */ 1901 switch (op) { 1902 case PRCO_SETOPT: 1903 error = sockopt_getint(sopt, &optval); 1904 if (error) 1905 break; 1906 if ((optval % 2) != 0) { 1907 /* the API assumes even offset values */ 1908 error = EINVAL; 1909 } else if (so->so_proto->pr_protocol == 1910 IPPROTO_ICMPV6) { 1911 if (optval != icmp6off) 1912 error = EINVAL; 1913 } else 1914 in6p->in6p_cksum = optval; 1915 break; 1916 1917 case PRCO_GETOPT: 1918 if (so->so_proto->pr_protocol == IPPROTO_ICMPV6) 1919 optval = icmp6off; 1920 else 1921 optval = in6p->in6p_cksum; 1922 1923 error = sockopt_setint(sopt, optval); 1924 break; 1925 1926 default: 1927 error = EINVAL; 1928 break; 1929 } 1930 break; 1931 1932 default: 1933 error = ENOPROTOOPT; 1934 break; 1935 } 1936 1937 return (error); 1938 } 1939 1940 #ifdef RFC2292 1941 /* 1942 * Set up IP6 options in pcb for insertion in output packets or 1943 * specifying behavior of outgoing packets. 1944 */ 1945 static int 1946 ip6_pcbopts(struct ip6_pktopts **pktopt, struct socket *so, 1947 struct sockopt *sopt) 1948 { 1949 struct ip6_pktopts *opt = *pktopt; 1950 struct mbuf *m; 1951 int error = 0; 1952 1953 /* turn off any old options. */ 1954 if (opt) { 1955 #ifdef DIAGNOSTIC 1956 if (opt->ip6po_pktinfo || opt->ip6po_nexthop || 1957 opt->ip6po_hbh || opt->ip6po_dest1 || opt->ip6po_dest2 || 1958 opt->ip6po_rhinfo.ip6po_rhi_rthdr) 1959 printf("ip6_pcbopts: all specified options are cleared.\n"); 1960 #endif 1961 ip6_clearpktopts(opt, -1); 1962 } else { 1963 opt = malloc(sizeof(*opt), M_IP6OPT, M_NOWAIT); 1964 if (opt == NULL) 1965 return (ENOBUFS); 1966 } 1967 *pktopt = NULL; 1968 1969 if (sopt == NULL || sopt->sopt_size == 0) { 1970 /* 1971 * Only turning off any previous options, regardless of 1972 * whether the opt is just created or given. 1973 */ 1974 free(opt, M_IP6OPT); 1975 return (0); 1976 } 1977 1978 /* set options specified by user. */ 1979 m = sockopt_getmbuf(sopt); 1980 if (m == NULL) { 1981 free(opt, M_IP6OPT); 1982 return (ENOBUFS); 1983 } 1984 1985 error = ip6_setpktopts(m, opt, NULL, kauth_cred_get(), 1986 so->so_proto->pr_protocol); 1987 m_freem(m); 1988 if (error != 0) { 1989 ip6_clearpktopts(opt, -1); /* XXX: discard all options */ 1990 free(opt, M_IP6OPT); 1991 return (error); 1992 } 1993 *pktopt = opt; 1994 return (0); 1995 } 1996 #endif 1997 1998 /* 1999 * initialize ip6_pktopts. beware that there are non-zero default values in 2000 * the struct. 2001 */ 2002 void 2003 ip6_initpktopts(struct ip6_pktopts *opt) 2004 { 2005 2006 memset(opt, 0, sizeof(*opt)); 2007 opt->ip6po_hlim = -1; /* -1 means default hop limit */ 2008 opt->ip6po_tclass = -1; /* -1 means default traffic class */ 2009 opt->ip6po_minmtu = IP6PO_MINMTU_MCASTONLY; 2010 } 2011 2012 #define sin6tosa(sin6) ((struct sockaddr *)(sin6)) /* XXX */ 2013 static int 2014 ip6_pcbopt(int optname, u_char *buf, int len, struct ip6_pktopts **pktopt, 2015 kauth_cred_t cred, int uproto) 2016 { 2017 struct ip6_pktopts *opt; 2018 2019 if (*pktopt == NULL) { 2020 *pktopt = malloc(sizeof(struct ip6_pktopts), M_IP6OPT, 2021 M_NOWAIT); 2022 if (*pktopt == NULL) 2023 return (ENOBUFS); 2024 2025 ip6_initpktopts(*pktopt); 2026 } 2027 opt = *pktopt; 2028 2029 return (ip6_setpktopt(optname, buf, len, opt, cred, 1, 0, uproto)); 2030 } 2031 2032 static int 2033 ip6_getpcbopt(struct ip6_pktopts *pktopt, int optname, struct sockopt *sopt) 2034 { 2035 void *optdata = NULL; 2036 int optdatalen = 0; 2037 struct ip6_ext *ip6e; 2038 int error = 0; 2039 struct in6_pktinfo null_pktinfo; 2040 int deftclass = 0, on; 2041 int defminmtu = IP6PO_MINMTU_MCASTONLY; 2042 2043 switch (optname) { 2044 case IPV6_PKTINFO: 2045 if (pktopt && pktopt->ip6po_pktinfo) 2046 optdata = (void *)pktopt->ip6po_pktinfo; 2047 else { 2048 /* XXX: we don't have to do this every time... */ 2049 memset(&null_pktinfo, 0, sizeof(null_pktinfo)); 2050 optdata = (void *)&null_pktinfo; 2051 } 2052 optdatalen = sizeof(struct in6_pktinfo); 2053 break; 2054 case IPV6_OTCLASS: 2055 /* XXX */ 2056 return (EINVAL); 2057 case IPV6_TCLASS: 2058 if (pktopt && pktopt->ip6po_tclass >= 0) 2059 optdata = (void *)&pktopt->ip6po_tclass; 2060 else 2061 optdata = (void *)&deftclass; 2062 optdatalen = sizeof(int); 2063 break; 2064 case IPV6_HOPOPTS: 2065 if (pktopt && pktopt->ip6po_hbh) { 2066 optdata = (void *)pktopt->ip6po_hbh; 2067 ip6e = (struct ip6_ext *)pktopt->ip6po_hbh; 2068 optdatalen = (ip6e->ip6e_len + 1) << 3; 2069 } 2070 break; 2071 case IPV6_RTHDR: 2072 if (pktopt && pktopt->ip6po_rthdr) { 2073 optdata = (void *)pktopt->ip6po_rthdr; 2074 ip6e = (struct ip6_ext *)pktopt->ip6po_rthdr; 2075 optdatalen = (ip6e->ip6e_len + 1) << 3; 2076 } 2077 break; 2078 case IPV6_RTHDRDSTOPTS: 2079 if (pktopt && pktopt->ip6po_dest1) { 2080 optdata = (void *)pktopt->ip6po_dest1; 2081 ip6e = (struct ip6_ext *)pktopt->ip6po_dest1; 2082 optdatalen = (ip6e->ip6e_len + 1) << 3; 2083 } 2084 break; 2085 case IPV6_DSTOPTS: 2086 if (pktopt && pktopt->ip6po_dest2) { 2087 optdata = (void *)pktopt->ip6po_dest2; 2088 ip6e = (struct ip6_ext *)pktopt->ip6po_dest2; 2089 optdatalen = (ip6e->ip6e_len + 1) << 3; 2090 } 2091 break; 2092 case IPV6_NEXTHOP: 2093 if (pktopt && pktopt->ip6po_nexthop) { 2094 optdata = (void *)pktopt->ip6po_nexthop; 2095 optdatalen = pktopt->ip6po_nexthop->sa_len; 2096 } 2097 break; 2098 case IPV6_USE_MIN_MTU: 2099 if (pktopt) 2100 optdata = (void *)&pktopt->ip6po_minmtu; 2101 else 2102 optdata = (void *)&defminmtu; 2103 optdatalen = sizeof(int); 2104 break; 2105 case IPV6_DONTFRAG: 2106 if (pktopt && ((pktopt->ip6po_flags) & IP6PO_DONTFRAG)) 2107 on = 1; 2108 else 2109 on = 0; 2110 optdata = (void *)&on; 2111 optdatalen = sizeof(on); 2112 break; 2113 default: /* should not happen */ 2114 #ifdef DIAGNOSTIC 2115 panic("ip6_getpcbopt: unexpected option\n"); 2116 #endif 2117 return (ENOPROTOOPT); 2118 } 2119 2120 error = sockopt_set(sopt, optdata, optdatalen); 2121 2122 return (error); 2123 } 2124 2125 void 2126 ip6_clearpktopts(struct ip6_pktopts *pktopt, int optname) 2127 { 2128 if (optname == -1 || optname == IPV6_PKTINFO) { 2129 if (pktopt->ip6po_pktinfo) 2130 free(pktopt->ip6po_pktinfo, M_IP6OPT); 2131 pktopt->ip6po_pktinfo = NULL; 2132 } 2133 if (optname == -1 || optname == IPV6_HOPLIMIT) 2134 pktopt->ip6po_hlim = -1; 2135 if (optname == -1 || optname == IPV6_TCLASS) 2136 pktopt->ip6po_tclass = -1; 2137 if (optname == -1 || optname == IPV6_NEXTHOP) { 2138 rtcache_free(&pktopt->ip6po_nextroute); 2139 if (pktopt->ip6po_nexthop) 2140 free(pktopt->ip6po_nexthop, M_IP6OPT); 2141 pktopt->ip6po_nexthop = NULL; 2142 } 2143 if (optname == -1 || optname == IPV6_HOPOPTS) { 2144 if (pktopt->ip6po_hbh) 2145 free(pktopt->ip6po_hbh, M_IP6OPT); 2146 pktopt->ip6po_hbh = NULL; 2147 } 2148 if (optname == -1 || optname == IPV6_RTHDRDSTOPTS) { 2149 if (pktopt->ip6po_dest1) 2150 free(pktopt->ip6po_dest1, M_IP6OPT); 2151 pktopt->ip6po_dest1 = NULL; 2152 } 2153 if (optname == -1 || optname == IPV6_RTHDR) { 2154 if (pktopt->ip6po_rhinfo.ip6po_rhi_rthdr) 2155 free(pktopt->ip6po_rhinfo.ip6po_rhi_rthdr, M_IP6OPT); 2156 pktopt->ip6po_rhinfo.ip6po_rhi_rthdr = NULL; 2157 rtcache_free(&pktopt->ip6po_route); 2158 } 2159 if (optname == -1 || optname == IPV6_DSTOPTS) { 2160 if (pktopt->ip6po_dest2) 2161 free(pktopt->ip6po_dest2, M_IP6OPT); 2162 pktopt->ip6po_dest2 = NULL; 2163 } 2164 } 2165 2166 #define PKTOPT_EXTHDRCPY(type) \ 2167 do { \ 2168 if (src->type) { \ 2169 int hlen = (((struct ip6_ext *)src->type)->ip6e_len + 1) << 3;\ 2170 dst->type = malloc(hlen, M_IP6OPT, canwait); \ 2171 if (dst->type == NULL) \ 2172 goto bad; \ 2173 memcpy(dst->type, src->type, hlen); \ 2174 } \ 2175 } while (/*CONSTCOND*/ 0) 2176 2177 static int 2178 copypktopts(struct ip6_pktopts *dst, struct ip6_pktopts *src, int canwait) 2179 { 2180 dst->ip6po_hlim = src->ip6po_hlim; 2181 dst->ip6po_tclass = src->ip6po_tclass; 2182 dst->ip6po_flags = src->ip6po_flags; 2183 if (src->ip6po_pktinfo) { 2184 dst->ip6po_pktinfo = malloc(sizeof(*dst->ip6po_pktinfo), 2185 M_IP6OPT, canwait); 2186 if (dst->ip6po_pktinfo == NULL) 2187 goto bad; 2188 *dst->ip6po_pktinfo = *src->ip6po_pktinfo; 2189 } 2190 if (src->ip6po_nexthop) { 2191 dst->ip6po_nexthop = malloc(src->ip6po_nexthop->sa_len, 2192 M_IP6OPT, canwait); 2193 if (dst->ip6po_nexthop == NULL) 2194 goto bad; 2195 memcpy(dst->ip6po_nexthop, src->ip6po_nexthop, 2196 src->ip6po_nexthop->sa_len); 2197 } 2198 PKTOPT_EXTHDRCPY(ip6po_hbh); 2199 PKTOPT_EXTHDRCPY(ip6po_dest1); 2200 PKTOPT_EXTHDRCPY(ip6po_dest2); 2201 PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */ 2202 return (0); 2203 2204 bad: 2205 if (dst->ip6po_pktinfo) free(dst->ip6po_pktinfo, M_IP6OPT); 2206 if (dst->ip6po_nexthop) free(dst->ip6po_nexthop, M_IP6OPT); 2207 if (dst->ip6po_hbh) free(dst->ip6po_hbh, M_IP6OPT); 2208 if (dst->ip6po_dest1) free(dst->ip6po_dest1, M_IP6OPT); 2209 if (dst->ip6po_dest2) free(dst->ip6po_dest2, M_IP6OPT); 2210 if (dst->ip6po_rthdr) free(dst->ip6po_rthdr, M_IP6OPT); 2211 2212 return (ENOBUFS); 2213 } 2214 #undef PKTOPT_EXTHDRCPY 2215 2216 struct ip6_pktopts * 2217 ip6_copypktopts(struct ip6_pktopts *src, int canwait) 2218 { 2219 int error; 2220 struct ip6_pktopts *dst; 2221 2222 dst = malloc(sizeof(*dst), M_IP6OPT, canwait); 2223 if (dst == NULL) 2224 return (NULL); 2225 ip6_initpktopts(dst); 2226 2227 if ((error = copypktopts(dst, src, canwait)) != 0) { 2228 free(dst, M_IP6OPT); 2229 return (NULL); 2230 } 2231 2232 return (dst); 2233 } 2234 2235 void 2236 ip6_freepcbopts(struct ip6_pktopts *pktopt) 2237 { 2238 if (pktopt == NULL) 2239 return; 2240 2241 ip6_clearpktopts(pktopt, -1); 2242 2243 free(pktopt, M_IP6OPT); 2244 } 2245 2246 /* 2247 * Set the IP6 multicast options in response to user setsockopt(). 2248 */ 2249 static int 2250 ip6_setmoptions(const struct sockopt *sopt, struct ip6_moptions **im6op) 2251 { 2252 int error = 0; 2253 u_int loop, ifindex; 2254 struct ipv6_mreq mreq; 2255 struct ifnet *ifp; 2256 struct ip6_moptions *im6o = *im6op; 2257 struct route ro; 2258 struct in6_multi_mship *imm; 2259 struct lwp *l = curlwp; /* XXX */ 2260 2261 if (im6o == NULL) { 2262 /* 2263 * No multicast option buffer attached to the pcb; 2264 * allocate one and initialize to default values. 2265 */ 2266 im6o = malloc(sizeof(*im6o), M_IPMOPTS, M_NOWAIT); 2267 if (im6o == NULL) 2268 return (ENOBUFS); 2269 2270 *im6op = im6o; 2271 im6o->im6o_multicast_ifp = NULL; 2272 im6o->im6o_multicast_hlim = ip6_defmcasthlim; 2273 im6o->im6o_multicast_loop = IPV6_DEFAULT_MULTICAST_LOOP; 2274 LIST_INIT(&im6o->im6o_memberships); 2275 } 2276 2277 switch (sopt->sopt_name) { 2278 2279 case IPV6_MULTICAST_IF: 2280 /* 2281 * Select the interface for outgoing multicast packets. 2282 */ 2283 error = sockopt_get(sopt, &ifindex, sizeof(ifindex)); 2284 if (error != 0) 2285 break; 2286 2287 if (ifindex != 0) { 2288 if ((ifp = if_byindex(ifindex)) == NULL) { 2289 error = ENXIO; /* XXX EINVAL? */ 2290 break; 2291 } 2292 if ((ifp->if_flags & IFF_MULTICAST) == 0) { 2293 error = EADDRNOTAVAIL; 2294 break; 2295 } 2296 } else 2297 ifp = NULL; 2298 im6o->im6o_multicast_ifp = ifp; 2299 break; 2300 2301 case IPV6_MULTICAST_HOPS: 2302 { 2303 /* 2304 * Set the IP6 hoplimit for outgoing multicast packets. 2305 */ 2306 int optval; 2307 2308 error = sockopt_getint(sopt, &optval); 2309 if (error != 0) 2310 break; 2311 2312 if (optval < -1 || optval >= 256) 2313 error = EINVAL; 2314 else if (optval == -1) 2315 im6o->im6o_multicast_hlim = ip6_defmcasthlim; 2316 else 2317 im6o->im6o_multicast_hlim = optval; 2318 break; 2319 } 2320 2321 case IPV6_MULTICAST_LOOP: 2322 /* 2323 * Set the loopback flag for outgoing multicast packets. 2324 * Must be zero or one. 2325 */ 2326 error = sockopt_get(sopt, &loop, sizeof(loop)); 2327 if (error != 0) 2328 break; 2329 if (loop > 1) { 2330 error = EINVAL; 2331 break; 2332 } 2333 im6o->im6o_multicast_loop = loop; 2334 break; 2335 2336 case IPV6_JOIN_GROUP: 2337 /* 2338 * Add a multicast group membership. 2339 * Group must be a valid IP6 multicast address. 2340 */ 2341 error = sockopt_get(sopt, &mreq, sizeof(mreq)); 2342 if (error != 0) 2343 break; 2344 2345 if (IN6_IS_ADDR_UNSPECIFIED(&mreq.ipv6mr_multiaddr)) { 2346 /* 2347 * We use the unspecified address to specify to accept 2348 * all multicast addresses. Only super user is allowed 2349 * to do this. 2350 */ 2351 if (kauth_authorize_network(l->l_cred, KAUTH_NETWORK_IPV6, 2352 KAUTH_REQ_NETWORK_IPV6_JOIN_MULTICAST, NULL, NULL, NULL)) 2353 { 2354 error = EACCES; 2355 break; 2356 } 2357 } else if (!IN6_IS_ADDR_MULTICAST(&mreq.ipv6mr_multiaddr)) { 2358 error = EINVAL; 2359 break; 2360 } 2361 2362 /* 2363 * If no interface was explicitly specified, choose an 2364 * appropriate one according to the given multicast address. 2365 */ 2366 if (mreq.ipv6mr_interface == 0) { 2367 struct rtentry *rt; 2368 union { 2369 struct sockaddr dst; 2370 struct sockaddr_in6 dst6; 2371 } u; 2372 2373 /* 2374 * Look up the routing table for the 2375 * address, and choose the outgoing interface. 2376 * XXX: is it a good approach? 2377 */ 2378 memset(&ro, 0, sizeof(ro)); 2379 sockaddr_in6_init(&u.dst6, &mreq.ipv6mr_multiaddr, 0, 2380 0, 0); 2381 rtcache_setdst(&ro, &u.dst); 2382 ifp = (rt = rtcache_init(&ro)) != NULL ? rt->rt_ifp 2383 : NULL; 2384 rtcache_free(&ro); 2385 } else { 2386 /* 2387 * If the interface is specified, validate it. 2388 */ 2389 if ((ifp = if_byindex(mreq.ipv6mr_interface)) == NULL) { 2390 error = ENXIO; /* XXX EINVAL? */ 2391 break; 2392 } 2393 } 2394 2395 /* 2396 * See if we found an interface, and confirm that it 2397 * supports multicast 2398 */ 2399 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) { 2400 error = EADDRNOTAVAIL; 2401 break; 2402 } 2403 2404 if (in6_setscope(&mreq.ipv6mr_multiaddr, ifp, NULL)) { 2405 error = EADDRNOTAVAIL; /* XXX: should not happen */ 2406 break; 2407 } 2408 2409 /* 2410 * See if the membership already exists. 2411 */ 2412 for (imm = im6o->im6o_memberships.lh_first; 2413 imm != NULL; imm = imm->i6mm_chain.le_next) 2414 if (imm->i6mm_maddr->in6m_ifp == ifp && 2415 IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr, 2416 &mreq.ipv6mr_multiaddr)) 2417 break; 2418 if (imm != NULL) { 2419 error = EADDRINUSE; 2420 break; 2421 } 2422 /* 2423 * Everything looks good; add a new record to the multicast 2424 * address list for the given interface. 2425 */ 2426 imm = in6_joingroup(ifp, &mreq.ipv6mr_multiaddr, &error, 0); 2427 if (imm == NULL) 2428 break; 2429 LIST_INSERT_HEAD(&im6o->im6o_memberships, imm, i6mm_chain); 2430 break; 2431 2432 case IPV6_LEAVE_GROUP: 2433 /* 2434 * Drop a multicast group membership. 2435 * Group must be a valid IP6 multicast address. 2436 */ 2437 error = sockopt_get(sopt, &mreq, sizeof(mreq)); 2438 if (error != 0) 2439 break; 2440 2441 /* 2442 * If an interface address was specified, get a pointer 2443 * to its ifnet structure. 2444 */ 2445 if (mreq.ipv6mr_interface != 0) { 2446 if ((ifp = if_byindex(mreq.ipv6mr_interface)) == NULL) { 2447 error = ENXIO; /* XXX EINVAL? */ 2448 break; 2449 } 2450 } else 2451 ifp = NULL; 2452 2453 /* Fill in the scope zone ID */ 2454 if (ifp) { 2455 if (in6_setscope(&mreq.ipv6mr_multiaddr, ifp, NULL)) { 2456 /* XXX: should not happen */ 2457 error = EADDRNOTAVAIL; 2458 break; 2459 } 2460 } else if (mreq.ipv6mr_interface != 0) { 2461 /* 2462 * XXX: This case would happens when the (positive) 2463 * index is in the valid range, but the corresponding 2464 * interface has been detached dynamically. The above 2465 * check probably avoids such case to happen here, but 2466 * we check it explicitly for safety. 2467 */ 2468 error = EADDRNOTAVAIL; 2469 break; 2470 } else { /* ipv6mr_interface == 0 */ 2471 struct sockaddr_in6 sa6_mc; 2472 2473 /* 2474 * The API spec says as follows: 2475 * If the interface index is specified as 0, the 2476 * system may choose a multicast group membership to 2477 * drop by matching the multicast address only. 2478 * On the other hand, we cannot disambiguate the scope 2479 * zone unless an interface is provided. Thus, we 2480 * check if there's ambiguity with the default scope 2481 * zone as the last resort. 2482 */ 2483 sockaddr_in6_init(&sa6_mc, &mreq.ipv6mr_multiaddr, 2484 0, 0, 0); 2485 error = sa6_embedscope(&sa6_mc, ip6_use_defzone); 2486 if (error != 0) 2487 break; 2488 mreq.ipv6mr_multiaddr = sa6_mc.sin6_addr; 2489 } 2490 2491 /* 2492 * Find the membership in the membership list. 2493 */ 2494 for (imm = im6o->im6o_memberships.lh_first; 2495 imm != NULL; imm = imm->i6mm_chain.le_next) { 2496 if ((ifp == NULL || imm->i6mm_maddr->in6m_ifp == ifp) && 2497 IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr, 2498 &mreq.ipv6mr_multiaddr)) 2499 break; 2500 } 2501 if (imm == NULL) { 2502 /* Unable to resolve interface */ 2503 error = EADDRNOTAVAIL; 2504 break; 2505 } 2506 /* 2507 * Give up the multicast address record to which the 2508 * membership points. 2509 */ 2510 LIST_REMOVE(imm, i6mm_chain); 2511 in6_leavegroup(imm); 2512 break; 2513 2514 default: 2515 error = EOPNOTSUPP; 2516 break; 2517 } 2518 2519 /* 2520 * If all options have default values, no need to keep the mbuf. 2521 */ 2522 if (im6o->im6o_multicast_ifp == NULL && 2523 im6o->im6o_multicast_hlim == ip6_defmcasthlim && 2524 im6o->im6o_multicast_loop == IPV6_DEFAULT_MULTICAST_LOOP && 2525 im6o->im6o_memberships.lh_first == NULL) { 2526 free(*im6op, M_IPMOPTS); 2527 *im6op = NULL; 2528 } 2529 2530 return (error); 2531 } 2532 2533 /* 2534 * Return the IP6 multicast options in response to user getsockopt(). 2535 */ 2536 static int 2537 ip6_getmoptions(struct sockopt *sopt, struct ip6_moptions *im6o) 2538 { 2539 u_int optval; 2540 int error; 2541 2542 switch (sopt->sopt_name) { 2543 case IPV6_MULTICAST_IF: 2544 if (im6o == NULL || im6o->im6o_multicast_ifp == NULL) 2545 optval = 0; 2546 else 2547 optval = im6o->im6o_multicast_ifp->if_index; 2548 2549 error = sockopt_set(sopt, &optval, sizeof(optval)); 2550 break; 2551 2552 case IPV6_MULTICAST_HOPS: 2553 if (im6o == NULL) 2554 optval = ip6_defmcasthlim; 2555 else 2556 optval = im6o->im6o_multicast_hlim; 2557 2558 error = sockopt_set(sopt, &optval, sizeof(optval)); 2559 break; 2560 2561 case IPV6_MULTICAST_LOOP: 2562 if (im6o == NULL) 2563 optval = IPV6_DEFAULT_MULTICAST_LOOP; 2564 else 2565 optval = im6o->im6o_multicast_loop; 2566 2567 error = sockopt_set(sopt, &optval, sizeof(optval)); 2568 break; 2569 2570 default: 2571 error = EOPNOTSUPP; 2572 } 2573 2574 return (error); 2575 } 2576 2577 /* 2578 * Discard the IP6 multicast options. 2579 */ 2580 void 2581 ip6_freemoptions(struct ip6_moptions *im6o) 2582 { 2583 struct in6_multi_mship *imm; 2584 2585 if (im6o == NULL) 2586 return; 2587 2588 while ((imm = im6o->im6o_memberships.lh_first) != NULL) { 2589 LIST_REMOVE(imm, i6mm_chain); 2590 in6_leavegroup(imm); 2591 } 2592 free(im6o, M_IPMOPTS); 2593 } 2594 2595 /* 2596 * Set IPv6 outgoing packet options based on advanced API. 2597 */ 2598 int 2599 ip6_setpktopts(struct mbuf *control, struct ip6_pktopts *opt, 2600 struct ip6_pktopts *stickyopt, kauth_cred_t cred, int uproto) 2601 { 2602 struct cmsghdr *cm = 0; 2603 2604 if (control == NULL || opt == NULL) 2605 return (EINVAL); 2606 2607 ip6_initpktopts(opt); 2608 if (stickyopt) { 2609 int error; 2610 2611 /* 2612 * If stickyopt is provided, make a local copy of the options 2613 * for this particular packet, then override them by ancillary 2614 * objects. 2615 * XXX: copypktopts() does not copy the cached route to a next 2616 * hop (if any). This is not very good in terms of efficiency, 2617 * but we can allow this since this option should be rarely 2618 * used. 2619 */ 2620 if ((error = copypktopts(opt, stickyopt, M_NOWAIT)) != 0) 2621 return (error); 2622 } 2623 2624 /* 2625 * XXX: Currently, we assume all the optional information is stored 2626 * in a single mbuf. 2627 */ 2628 if (control->m_next) 2629 return (EINVAL); 2630 2631 /* XXX if cm->cmsg_len is not aligned, control->m_len can become <0 */ 2632 for (; control->m_len > 0; control->m_data += CMSG_ALIGN(cm->cmsg_len), 2633 control->m_len -= CMSG_ALIGN(cm->cmsg_len)) { 2634 int error; 2635 2636 if (control->m_len < CMSG_LEN(0)) 2637 return (EINVAL); 2638 2639 cm = mtod(control, struct cmsghdr *); 2640 if (cm->cmsg_len == 0 || cm->cmsg_len > control->m_len) 2641 return (EINVAL); 2642 if (cm->cmsg_level != IPPROTO_IPV6) 2643 continue; 2644 2645 error = ip6_setpktopt(cm->cmsg_type, CMSG_DATA(cm), 2646 cm->cmsg_len - CMSG_LEN(0), opt, cred, 0, 1, uproto); 2647 if (error) 2648 return (error); 2649 } 2650 2651 return (0); 2652 } 2653 2654 /* 2655 * Set a particular packet option, as a sticky option or an ancillary data 2656 * item. "len" can be 0 only when it's a sticky option. 2657 * We have 4 cases of combination of "sticky" and "cmsg": 2658 * "sticky=0, cmsg=0": impossible 2659 * "sticky=0, cmsg=1": RFC2292 or RFC3542 ancillary data 2660 * "sticky=1, cmsg=0": RFC3542 socket option 2661 * "sticky=1, cmsg=1": RFC2292 socket option 2662 */ 2663 static int 2664 ip6_setpktopt(int optname, u_char *buf, int len, struct ip6_pktopts *opt, 2665 kauth_cred_t cred, int sticky, int cmsg, int uproto) 2666 { 2667 int minmtupolicy; 2668 int error; 2669 2670 if (!sticky && !cmsg) { 2671 #ifdef DIAGNOSTIC 2672 printf("ip6_setpktopt: impossible case\n"); 2673 #endif 2674 return (EINVAL); 2675 } 2676 2677 /* 2678 * IPV6_2292xxx is for backward compatibility to RFC2292, and should 2679 * not be specified in the context of RFC3542. Conversely, 2680 * RFC3542 types should not be specified in the context of RFC2292. 2681 */ 2682 if (!cmsg) { 2683 switch (optname) { 2684 case IPV6_2292PKTINFO: 2685 case IPV6_2292HOPLIMIT: 2686 case IPV6_2292NEXTHOP: 2687 case IPV6_2292HOPOPTS: 2688 case IPV6_2292DSTOPTS: 2689 case IPV6_2292RTHDR: 2690 case IPV6_2292PKTOPTIONS: 2691 return (ENOPROTOOPT); 2692 } 2693 } 2694 if (sticky && cmsg) { 2695 switch (optname) { 2696 case IPV6_PKTINFO: 2697 case IPV6_HOPLIMIT: 2698 case IPV6_NEXTHOP: 2699 case IPV6_HOPOPTS: 2700 case IPV6_DSTOPTS: 2701 case IPV6_RTHDRDSTOPTS: 2702 case IPV6_RTHDR: 2703 case IPV6_USE_MIN_MTU: 2704 case IPV6_DONTFRAG: 2705 case IPV6_OTCLASS: 2706 case IPV6_TCLASS: 2707 return (ENOPROTOOPT); 2708 } 2709 } 2710 2711 switch (optname) { 2712 #ifdef RFC2292 2713 case IPV6_2292PKTINFO: 2714 #endif 2715 case IPV6_PKTINFO: 2716 { 2717 struct ifnet *ifp = NULL; 2718 struct in6_pktinfo *pktinfo; 2719 2720 if (len != sizeof(struct in6_pktinfo)) 2721 return (EINVAL); 2722 2723 pktinfo = (struct in6_pktinfo *)buf; 2724 2725 /* 2726 * An application can clear any sticky IPV6_PKTINFO option by 2727 * doing a "regular" setsockopt with ipi6_addr being 2728 * in6addr_any and ipi6_ifindex being zero. 2729 * [RFC 3542, Section 6] 2730 */ 2731 if (optname == IPV6_PKTINFO && opt->ip6po_pktinfo && 2732 pktinfo->ipi6_ifindex == 0 && 2733 IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) { 2734 ip6_clearpktopts(opt, optname); 2735 break; 2736 } 2737 2738 if (uproto == IPPROTO_TCP && optname == IPV6_PKTINFO && 2739 sticky && !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) { 2740 return (EINVAL); 2741 } 2742 2743 /* Validate the interface index if specified. */ 2744 if (pktinfo->ipi6_ifindex) { 2745 ifp = if_byindex(pktinfo->ipi6_ifindex); 2746 if (ifp == NULL) 2747 return (ENXIO); 2748 } 2749 2750 /* 2751 * We store the address anyway, and let in6_selectsrc() 2752 * validate the specified address. This is because ipi6_addr 2753 * may not have enough information about its scope zone, and 2754 * we may need additional information (such as outgoing 2755 * interface or the scope zone of a destination address) to 2756 * disambiguate the scope. 2757 * XXX: the delay of the validation may confuse the 2758 * application when it is used as a sticky option. 2759 */ 2760 if (opt->ip6po_pktinfo == NULL) { 2761 opt->ip6po_pktinfo = malloc(sizeof(*pktinfo), 2762 M_IP6OPT, M_NOWAIT); 2763 if (opt->ip6po_pktinfo == NULL) 2764 return (ENOBUFS); 2765 } 2766 memcpy(opt->ip6po_pktinfo, pktinfo, sizeof(*pktinfo)); 2767 break; 2768 } 2769 2770 #ifdef RFC2292 2771 case IPV6_2292HOPLIMIT: 2772 #endif 2773 case IPV6_HOPLIMIT: 2774 { 2775 int *hlimp; 2776 2777 /* 2778 * RFC 3542 deprecated the usage of sticky IPV6_HOPLIMIT 2779 * to simplify the ordering among hoplimit options. 2780 */ 2781 if (optname == IPV6_HOPLIMIT && sticky) 2782 return (ENOPROTOOPT); 2783 2784 if (len != sizeof(int)) 2785 return (EINVAL); 2786 hlimp = (int *)buf; 2787 if (*hlimp < -1 || *hlimp > 255) 2788 return (EINVAL); 2789 2790 opt->ip6po_hlim = *hlimp; 2791 break; 2792 } 2793 2794 case IPV6_OTCLASS: 2795 if (len != sizeof(u_int8_t)) 2796 return (EINVAL); 2797 2798 opt->ip6po_tclass = *(u_int8_t *)buf; 2799 break; 2800 2801 case IPV6_TCLASS: 2802 { 2803 int tclass; 2804 2805 if (len != sizeof(int)) 2806 return (EINVAL); 2807 tclass = *(int *)buf; 2808 if (tclass < -1 || tclass > 255) 2809 return (EINVAL); 2810 2811 opt->ip6po_tclass = tclass; 2812 break; 2813 } 2814 2815 #ifdef RFC2292 2816 case IPV6_2292NEXTHOP: 2817 #endif 2818 case IPV6_NEXTHOP: 2819 error = kauth_authorize_network(cred, KAUTH_NETWORK_IPV6, 2820 KAUTH_REQ_NETWORK_IPV6_HOPBYHOP, NULL, NULL, NULL); 2821 if (error) 2822 return (error); 2823 2824 if (len == 0) { /* just remove the option */ 2825 ip6_clearpktopts(opt, IPV6_NEXTHOP); 2826 break; 2827 } 2828 2829 /* check if cmsg_len is large enough for sa_len */ 2830 if (len < sizeof(struct sockaddr) || len < *buf) 2831 return (EINVAL); 2832 2833 switch (((struct sockaddr *)buf)->sa_family) { 2834 case AF_INET6: 2835 { 2836 struct sockaddr_in6 *sa6 = (struct sockaddr_in6 *)buf; 2837 2838 if (sa6->sin6_len != sizeof(struct sockaddr_in6)) 2839 return (EINVAL); 2840 2841 if (IN6_IS_ADDR_UNSPECIFIED(&sa6->sin6_addr) || 2842 IN6_IS_ADDR_MULTICAST(&sa6->sin6_addr)) { 2843 return (EINVAL); 2844 } 2845 if ((error = sa6_embedscope(sa6, ip6_use_defzone)) 2846 != 0) { 2847 return (error); 2848 } 2849 break; 2850 } 2851 case AF_LINK: /* eventually be supported? */ 2852 default: 2853 return (EAFNOSUPPORT); 2854 } 2855 2856 /* turn off the previous option, then set the new option. */ 2857 ip6_clearpktopts(opt, IPV6_NEXTHOP); 2858 opt->ip6po_nexthop = malloc(*buf, M_IP6OPT, M_NOWAIT); 2859 if (opt->ip6po_nexthop == NULL) 2860 return (ENOBUFS); 2861 memcpy(opt->ip6po_nexthop, buf, *buf); 2862 break; 2863 2864 #ifdef RFC2292 2865 case IPV6_2292HOPOPTS: 2866 #endif 2867 case IPV6_HOPOPTS: 2868 { 2869 struct ip6_hbh *hbh; 2870 int hbhlen; 2871 2872 /* 2873 * XXX: We don't allow a non-privileged user to set ANY HbH 2874 * options, since per-option restriction has too much 2875 * overhead. 2876 */ 2877 error = kauth_authorize_network(cred, KAUTH_NETWORK_IPV6, 2878 KAUTH_REQ_NETWORK_IPV6_HOPBYHOP, NULL, NULL, NULL); 2879 if (error) 2880 return (error); 2881 2882 if (len == 0) { 2883 ip6_clearpktopts(opt, IPV6_HOPOPTS); 2884 break; /* just remove the option */ 2885 } 2886 2887 /* message length validation */ 2888 if (len < sizeof(struct ip6_hbh)) 2889 return (EINVAL); 2890 hbh = (struct ip6_hbh *)buf; 2891 hbhlen = (hbh->ip6h_len + 1) << 3; 2892 if (len != hbhlen) 2893 return (EINVAL); 2894 2895 /* turn off the previous option, then set the new option. */ 2896 ip6_clearpktopts(opt, IPV6_HOPOPTS); 2897 opt->ip6po_hbh = malloc(hbhlen, M_IP6OPT, M_NOWAIT); 2898 if (opt->ip6po_hbh == NULL) 2899 return (ENOBUFS); 2900 memcpy(opt->ip6po_hbh, hbh, hbhlen); 2901 2902 break; 2903 } 2904 2905 #ifdef RFC2292 2906 case IPV6_2292DSTOPTS: 2907 #endif 2908 case IPV6_DSTOPTS: 2909 case IPV6_RTHDRDSTOPTS: 2910 { 2911 struct ip6_dest *dest, **newdest = NULL; 2912 int destlen; 2913 2914 /* XXX: see the comment for IPV6_HOPOPTS */ 2915 error = kauth_authorize_network(cred, KAUTH_NETWORK_IPV6, 2916 KAUTH_REQ_NETWORK_IPV6_HOPBYHOP, NULL, NULL, NULL); 2917 if (error) 2918 return (error); 2919 2920 if (len == 0) { 2921 ip6_clearpktopts(opt, optname); 2922 break; /* just remove the option */ 2923 } 2924 2925 /* message length validation */ 2926 if (len < sizeof(struct ip6_dest)) 2927 return (EINVAL); 2928 dest = (struct ip6_dest *)buf; 2929 destlen = (dest->ip6d_len + 1) << 3; 2930 if (len != destlen) 2931 return (EINVAL); 2932 /* 2933 * Determine the position that the destination options header 2934 * should be inserted; before or after the routing header. 2935 */ 2936 switch (optname) { 2937 case IPV6_2292DSTOPTS: 2938 /* 2939 * The old advanced API is ambiguous on this point. 2940 * Our approach is to determine the position based 2941 * according to the existence of a routing header. 2942 * Note, however, that this depends on the order of the 2943 * extension headers in the ancillary data; the 1st 2944 * part of the destination options header must appear 2945 * before the routing header in the ancillary data, 2946 * too. 2947 * RFC3542 solved the ambiguity by introducing 2948 * separate ancillary data or option types. 2949 */ 2950 if (opt->ip6po_rthdr == NULL) 2951 newdest = &opt->ip6po_dest1; 2952 else 2953 newdest = &opt->ip6po_dest2; 2954 break; 2955 case IPV6_RTHDRDSTOPTS: 2956 newdest = &opt->ip6po_dest1; 2957 break; 2958 case IPV6_DSTOPTS: 2959 newdest = &opt->ip6po_dest2; 2960 break; 2961 } 2962 2963 /* turn off the previous option, then set the new option. */ 2964 ip6_clearpktopts(opt, optname); 2965 *newdest = malloc(destlen, M_IP6OPT, M_NOWAIT); 2966 if (*newdest == NULL) 2967 return (ENOBUFS); 2968 memcpy(*newdest, dest, destlen); 2969 2970 break; 2971 } 2972 2973 #ifdef RFC2292 2974 case IPV6_2292RTHDR: 2975 #endif 2976 case IPV6_RTHDR: 2977 { 2978 struct ip6_rthdr *rth; 2979 int rthlen; 2980 2981 if (len == 0) { 2982 ip6_clearpktopts(opt, IPV6_RTHDR); 2983 break; /* just remove the option */ 2984 } 2985 2986 /* message length validation */ 2987 if (len < sizeof(struct ip6_rthdr)) 2988 return (EINVAL); 2989 rth = (struct ip6_rthdr *)buf; 2990 rthlen = (rth->ip6r_len + 1) << 3; 2991 if (len != rthlen) 2992 return (EINVAL); 2993 switch (rth->ip6r_type) { 2994 case IPV6_RTHDR_TYPE_0: 2995 if (rth->ip6r_len == 0) /* must contain one addr */ 2996 return (EINVAL); 2997 if (rth->ip6r_len % 2) /* length must be even */ 2998 return (EINVAL); 2999 if (rth->ip6r_len / 2 != rth->ip6r_segleft) 3000 return (EINVAL); 3001 break; 3002 default: 3003 return (EINVAL); /* not supported */ 3004 } 3005 /* turn off the previous option */ 3006 ip6_clearpktopts(opt, IPV6_RTHDR); 3007 opt->ip6po_rthdr = malloc(rthlen, M_IP6OPT, M_NOWAIT); 3008 if (opt->ip6po_rthdr == NULL) 3009 return (ENOBUFS); 3010 memcpy(opt->ip6po_rthdr, rth, rthlen); 3011 break; 3012 } 3013 3014 case IPV6_USE_MIN_MTU: 3015 if (len != sizeof(int)) 3016 return (EINVAL); 3017 minmtupolicy = *(int *)buf; 3018 if (minmtupolicy != IP6PO_MINMTU_MCASTONLY && 3019 minmtupolicy != IP6PO_MINMTU_DISABLE && 3020 minmtupolicy != IP6PO_MINMTU_ALL) { 3021 return (EINVAL); 3022 } 3023 opt->ip6po_minmtu = minmtupolicy; 3024 break; 3025 3026 case IPV6_DONTFRAG: 3027 if (len != sizeof(int)) 3028 return (EINVAL); 3029 3030 if (uproto == IPPROTO_TCP || *(int *)buf == 0) { 3031 /* 3032 * we ignore this option for TCP sockets. 3033 * (RFC3542 leaves this case unspecified.) 3034 */ 3035 opt->ip6po_flags &= ~IP6PO_DONTFRAG; 3036 } else 3037 opt->ip6po_flags |= IP6PO_DONTFRAG; 3038 break; 3039 3040 default: 3041 return (ENOPROTOOPT); 3042 } /* end of switch */ 3043 3044 return (0); 3045 } 3046 3047 /* 3048 * Routine called from ip6_output() to loop back a copy of an IP6 multicast 3049 * packet to the input queue of a specified interface. Note that this 3050 * calls the output routine of the loopback "driver", but with an interface 3051 * pointer that might NOT be lo0ifp -- easier than replicating that code here. 3052 */ 3053 void 3054 ip6_mloopback(struct ifnet *ifp, struct mbuf *m, 3055 const struct sockaddr_in6 *dst) 3056 { 3057 struct mbuf *copym; 3058 struct ip6_hdr *ip6; 3059 3060 copym = m_copy(m, 0, M_COPYALL); 3061 if (copym == NULL) 3062 return; 3063 3064 /* 3065 * Make sure to deep-copy IPv6 header portion in case the data 3066 * is in an mbuf cluster, so that we can safely override the IPv6 3067 * header portion later. 3068 */ 3069 if ((copym->m_flags & M_EXT) != 0 || 3070 copym->m_len < sizeof(struct ip6_hdr)) { 3071 copym = m_pullup(copym, sizeof(struct ip6_hdr)); 3072 if (copym == NULL) 3073 return; 3074 } 3075 3076 #ifdef DIAGNOSTIC 3077 if (copym->m_len < sizeof(*ip6)) { 3078 m_freem(copym); 3079 return; 3080 } 3081 #endif 3082 3083 ip6 = mtod(copym, struct ip6_hdr *); 3084 /* 3085 * clear embedded scope identifiers if necessary. 3086 * in6_clearscope will touch the addresses only when necessary. 3087 */ 3088 in6_clearscope(&ip6->ip6_src); 3089 in6_clearscope(&ip6->ip6_dst); 3090 3091 (void)looutput(ifp, copym, (const struct sockaddr *)dst, NULL); 3092 } 3093 3094 /* 3095 * Chop IPv6 header off from the payload. 3096 */ 3097 static int 3098 ip6_splithdr(struct mbuf *m, struct ip6_exthdrs *exthdrs) 3099 { 3100 struct mbuf *mh; 3101 struct ip6_hdr *ip6; 3102 3103 ip6 = mtod(m, struct ip6_hdr *); 3104 if (m->m_len > sizeof(*ip6)) { 3105 MGETHDR(mh, M_DONTWAIT, MT_HEADER); 3106 if (mh == 0) { 3107 m_freem(m); 3108 return ENOBUFS; 3109 } 3110 M_MOVE_PKTHDR(mh, m); 3111 MH_ALIGN(mh, sizeof(*ip6)); 3112 m->m_len -= sizeof(*ip6); 3113 m->m_data += sizeof(*ip6); 3114 mh->m_next = m; 3115 m = mh; 3116 m->m_len = sizeof(*ip6); 3117 bcopy((void *)ip6, mtod(m, void *), sizeof(*ip6)); 3118 } 3119 exthdrs->ip6e_ip6 = m; 3120 return 0; 3121 } 3122 3123 /* 3124 * Compute IPv6 extension header length. 3125 */ 3126 int 3127 ip6_optlen(struct in6pcb *in6p) 3128 { 3129 int len; 3130 3131 if (!in6p->in6p_outputopts) 3132 return 0; 3133 3134 len = 0; 3135 #define elen(x) \ 3136 (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0) 3137 3138 len += elen(in6p->in6p_outputopts->ip6po_hbh); 3139 len += elen(in6p->in6p_outputopts->ip6po_dest1); 3140 len += elen(in6p->in6p_outputopts->ip6po_rthdr); 3141 len += elen(in6p->in6p_outputopts->ip6po_dest2); 3142 return len; 3143 #undef elen 3144 } 3145