1 /* $NetBSD: ip6_output.c,v 1.233 2023/03/20 09:15:52 ozaki-r 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 1.233 2023/03/20 09:15:52 ozaki-r Exp $"); 66 67 #ifdef _KERNEL_OPT 68 #include "opt_inet.h" 69 #include "opt_inet6.h" 70 #include "opt_ipsec.h" 71 #endif 72 73 #include <sys/param.h> 74 #include <sys/malloc.h> 75 #include <sys/mbuf.h> 76 #include <sys/errno.h> 77 #include <sys/socket.h> 78 #include <sys/socketvar.h> 79 #include <sys/syslog.h> 80 #include <sys/systm.h> 81 #include <sys/proc.h> 82 #include <sys/kauth.h> 83 84 #include <net/if.h> 85 #include <net/route.h> 86 #include <net/pfil.h> 87 88 #include <netinet/in.h> 89 #include <netinet/in_var.h> 90 #include <netinet/ip6.h> 91 #include <netinet/ip_var.h> 92 #include <netinet/icmp6.h> 93 #include <netinet/in_offload.h> 94 #include <netinet/portalgo.h> 95 #include <netinet6/in6_offload.h> 96 #include <netinet6/ip6_var.h> 97 #include <netinet6/ip6_private.h> 98 #include <netinet6/in6_pcb.h> 99 #include <netinet6/nd6.h> 100 #include <netinet6/ip6protosw.h> 101 #include <netinet6/scope6_var.h> 102 103 #ifdef IPSEC 104 #include <netipsec/ipsec.h> 105 #include <netipsec/ipsec6.h> 106 #include <netipsec/key.h> 107 #endif 108 109 extern pfil_head_t *inet6_pfil_hook; /* XXX */ 110 111 struct ip6_exthdrs { 112 struct mbuf *ip6e_ip6; 113 struct mbuf *ip6e_hbh; 114 struct mbuf *ip6e_dest1; 115 struct mbuf *ip6e_rthdr; 116 struct mbuf *ip6e_dest2; 117 }; 118 119 static int ip6_pcbopt(int, u_char *, int, struct ip6_pktopts **, 120 kauth_cred_t, int); 121 static int ip6_getpcbopt(struct ip6_pktopts *, int, struct sockopt *); 122 static int ip6_setpktopt(int, u_char *, int, struct ip6_pktopts *, kauth_cred_t, 123 int, int, int); 124 static int ip6_setmoptions(const struct sockopt *, struct inpcb *); 125 static int ip6_getmoptions(struct sockopt *, struct inpcb *); 126 static int ip6_copyexthdr(struct mbuf **, void *, int); 127 static int ip6_insertfraghdr(struct mbuf *, struct mbuf *, int, 128 struct ip6_frag **); 129 static int ip6_insert_jumboopt(struct ip6_exthdrs *, u_int32_t); 130 static int ip6_splithdr(struct mbuf *, struct ip6_exthdrs *); 131 static int ip6_getpmtu(struct rtentry *, struct ifnet *, u_long *, int *); 132 static int copypktopts(struct ip6_pktopts *, struct ip6_pktopts *, int); 133 static int ip6_ifaddrvalid(const struct in6_addr *, const struct in6_addr *); 134 static int ip6_handle_rthdr(struct ip6_rthdr *, struct ip6_hdr *); 135 136 #ifdef RFC2292 137 static int ip6_pcbopts(struct ip6_pktopts **, struct socket *, struct sockopt *); 138 #endif 139 140 static int 141 ip6_handle_rthdr(struct ip6_rthdr *rh, struct ip6_hdr *ip6) 142 { 143 int error = 0; 144 145 switch (rh->ip6r_type) { 146 case IPV6_RTHDR_TYPE_0: 147 /* Dropped, RFC5095. */ 148 default: /* is it possible? */ 149 error = EINVAL; 150 } 151 152 return error; 153 } 154 155 /* 156 * Send an IP packet to a host. 157 */ 158 int 159 ip6_if_output(struct ifnet * const ifp, struct ifnet * const origifp, 160 struct mbuf * const m, const struct sockaddr_in6 * const dst, 161 const struct rtentry *rt) 162 { 163 int error = 0; 164 165 if (rt != NULL) { 166 error = rt_check_reject_route(rt, ifp); 167 if (error != 0) { 168 IP6_STATINC(IP6_STAT_RTREJECT); 169 m_freem(m); 170 return error; 171 } 172 } 173 174 if ((ifp->if_flags & IFF_LOOPBACK) != 0) 175 error = if_output_lock(ifp, origifp, m, sin6tocsa(dst), rt); 176 else 177 error = if_output_lock(ifp, ifp, m, sin6tocsa(dst), rt); 178 return error; 179 } 180 181 /* 182 * IP6 output. The packet in mbuf chain m contains a skeletal IP6 183 * header (with pri, len, nxt, hlim, src, dst). 184 * 185 * This function may modify ver and hlim only. The mbuf chain containing the 186 * packet will be freed. The mbuf opt, if present, will not be freed. 187 * 188 * Type of "mtu": rt_rmx.rmx_mtu is u_long, ifnet.ifr_mtu is int, and 189 * nd_ifinfo.linkmtu is u_int32_t. So we use u_long to hold largest one, 190 * which is rt_rmx.rmx_mtu. 191 */ 192 int 193 ip6_output( 194 struct mbuf *m0, 195 struct ip6_pktopts *opt, 196 struct route *ro, 197 int flags, 198 struct ip6_moptions *im6o, 199 struct inpcb *inp, 200 struct ifnet **ifpp /* XXX: just for statistics */ 201 ) 202 { 203 struct ip6_hdr *ip6, *mhip6; 204 struct ifnet *ifp = NULL, *origifp = NULL; 205 struct mbuf *m = m0; 206 int tlen, len, off; 207 bool tso; 208 struct route ip6route; 209 struct rtentry *rt = NULL, *rt_pmtu; 210 const struct sockaddr_in6 *dst; 211 struct sockaddr_in6 src_sa, dst_sa; 212 int error = 0; 213 struct in6_ifaddr *ia = NULL; 214 u_long mtu; 215 int alwaysfrag, dontfrag; 216 u_int32_t optlen = 0, plen = 0, unfragpartlen = 0; 217 struct ip6_exthdrs exthdrs; 218 struct in6_addr finaldst, src0, dst0; 219 u_int32_t zone; 220 struct route *ro_pmtu = NULL; 221 int hdrsplit = 0; 222 int needipsec = 0; 223 #ifdef IPSEC 224 struct secpolicy *sp = NULL; 225 #endif 226 struct psref psref, psref_ia; 227 int bound = curlwp_bind(); 228 bool release_psref_ia = false; 229 230 #ifdef DIAGNOSTIC 231 if ((m->m_flags & M_PKTHDR) == 0) 232 panic("ip6_output: no HDR"); 233 if ((m->m_pkthdr.csum_flags & 234 (M_CSUM_TCPv4|M_CSUM_UDPv4|M_CSUM_TSOv4)) != 0) { 235 panic("ip6_output: IPv4 checksum offload flags: %d", 236 m->m_pkthdr.csum_flags); 237 } 238 if ((m->m_pkthdr.csum_flags & (M_CSUM_TCPv6|M_CSUM_UDPv6)) == 239 (M_CSUM_TCPv6|M_CSUM_UDPv6)) { 240 panic("ip6_output: conflicting checksum offload flags: %d", 241 m->m_pkthdr.csum_flags); 242 } 243 #endif 244 245 M_CSUM_DATA_IPv6_SET(m->m_pkthdr.csum_data, sizeof(struct ip6_hdr)); 246 247 #define MAKE_EXTHDR(hp, mp) \ 248 do { \ 249 if (hp) { \ 250 struct ip6_ext *eh = (struct ip6_ext *)(hp); \ 251 error = ip6_copyexthdr((mp), (void *)(hp), \ 252 ((eh)->ip6e_len + 1) << 3); \ 253 if (error) \ 254 goto freehdrs; \ 255 } \ 256 } while (/*CONSTCOND*/ 0) 257 258 memset(&exthdrs, 0, sizeof(exthdrs)); 259 if (opt) { 260 /* Hop-by-Hop options header */ 261 MAKE_EXTHDR(opt->ip6po_hbh, &exthdrs.ip6e_hbh); 262 /* Destination options header (1st part) */ 263 MAKE_EXTHDR(opt->ip6po_dest1, &exthdrs.ip6e_dest1); 264 /* Routing header */ 265 MAKE_EXTHDR(opt->ip6po_rthdr, &exthdrs.ip6e_rthdr); 266 /* Destination options header (2nd part) */ 267 MAKE_EXTHDR(opt->ip6po_dest2, &exthdrs.ip6e_dest2); 268 } 269 270 /* 271 * Calculate the total length of the extension header chain. 272 * Keep the length of the unfragmentable part for fragmentation. 273 */ 274 optlen = 0; 275 if (exthdrs.ip6e_hbh) 276 optlen += exthdrs.ip6e_hbh->m_len; 277 if (exthdrs.ip6e_dest1) 278 optlen += exthdrs.ip6e_dest1->m_len; 279 if (exthdrs.ip6e_rthdr) 280 optlen += exthdrs.ip6e_rthdr->m_len; 281 unfragpartlen = optlen + sizeof(struct ip6_hdr); 282 /* NOTE: we don't add AH/ESP length here. do that later. */ 283 if (exthdrs.ip6e_dest2) 284 optlen += exthdrs.ip6e_dest2->m_len; 285 286 #ifdef IPSEC 287 if (ipsec_used) { 288 /* Check the security policy (SP) for the packet */ 289 sp = ipsec6_check_policy(m, inp, flags, &needipsec, &error); 290 if (error != 0) { 291 /* 292 * Hack: -EINVAL is used to signal that a packet 293 * should be silently discarded. This is typically 294 * because we asked key management for an SA and 295 * it was delayed (e.g. kicked up to IKE). 296 */ 297 if (error == -EINVAL) 298 error = 0; 299 IP6_STATINC(IP6_STAT_IPSECDROP_OUT); 300 goto freehdrs; 301 } 302 } 303 #endif 304 305 if (needipsec && 306 (m->m_pkthdr.csum_flags & (M_CSUM_UDPv6|M_CSUM_TCPv6)) != 0) { 307 in6_undefer_cksum_tcpudp(m); 308 m->m_pkthdr.csum_flags &= ~(M_CSUM_UDPv6|M_CSUM_TCPv6); 309 } 310 311 /* 312 * If we need IPsec, or there is at least one extension header, 313 * separate IP6 header from the payload. 314 */ 315 if ((needipsec || optlen) && !hdrsplit) { 316 if ((error = ip6_splithdr(m, &exthdrs)) != 0) { 317 IP6_STATINC(IP6_STAT_ODROPPED); 318 m = NULL; 319 goto freehdrs; 320 } 321 m = exthdrs.ip6e_ip6; 322 hdrsplit++; 323 } 324 325 /* adjust pointer */ 326 ip6 = mtod(m, struct ip6_hdr *); 327 328 /* adjust mbuf packet header length */ 329 m->m_pkthdr.len += optlen; 330 plen = m->m_pkthdr.len - sizeof(*ip6); 331 332 /* If this is a jumbo payload, insert a jumbo payload option. */ 333 if (plen > IPV6_MAXPACKET) { 334 if (!hdrsplit) { 335 if ((error = ip6_splithdr(m, &exthdrs)) != 0) { 336 IP6_STATINC(IP6_STAT_ODROPPED); 337 m = NULL; 338 goto freehdrs; 339 } 340 m = exthdrs.ip6e_ip6; 341 hdrsplit++; 342 } 343 /* adjust pointer */ 344 ip6 = mtod(m, struct ip6_hdr *); 345 if ((error = ip6_insert_jumboopt(&exthdrs, plen)) != 0) { 346 IP6_STATINC(IP6_STAT_ODROPPED); 347 goto freehdrs; 348 } 349 optlen += 8; /* XXX JUMBOOPTLEN */ 350 ip6->ip6_plen = 0; 351 } else 352 ip6->ip6_plen = htons(plen); 353 354 /* 355 * Concatenate headers and fill in next header fields. 356 * Here we have, on "m" 357 * IPv6 payload 358 * and we insert headers accordingly. Finally, we should be getting: 359 * IPv6 hbh dest1 rthdr ah* [esp* dest2 payload] 360 * 361 * during the header composing process, "m" points to IPv6 header. 362 * "mprev" points to an extension header prior to esp. 363 */ 364 { 365 u_char *nexthdrp = &ip6->ip6_nxt; 366 struct mbuf *mprev = m; 367 368 /* 369 * we treat dest2 specially. this makes IPsec processing 370 * much easier. the goal here is to make mprev point the 371 * mbuf prior to dest2. 372 * 373 * result: IPv6 dest2 payload 374 * m and mprev will point to IPv6 header. 375 */ 376 if (exthdrs.ip6e_dest2) { 377 if (!hdrsplit) 378 panic("assumption failed: hdr not split"); 379 exthdrs.ip6e_dest2->m_next = m->m_next; 380 m->m_next = exthdrs.ip6e_dest2; 381 *mtod(exthdrs.ip6e_dest2, u_char *) = ip6->ip6_nxt; 382 ip6->ip6_nxt = IPPROTO_DSTOPTS; 383 } 384 385 #define MAKE_CHAIN(m, mp, p, i)\ 386 do {\ 387 if (m) {\ 388 if (!hdrsplit) \ 389 panic("assumption failed: hdr not split"); \ 390 *mtod((m), u_char *) = *(p);\ 391 *(p) = (i);\ 392 p = mtod((m), u_char *);\ 393 (m)->m_next = (mp)->m_next;\ 394 (mp)->m_next = (m);\ 395 (mp) = (m);\ 396 }\ 397 } while (/*CONSTCOND*/ 0) 398 /* 399 * result: IPv6 hbh dest1 rthdr dest2 payload 400 * m will point to IPv6 header. mprev will point to the 401 * extension header prior to dest2 (rthdr in the above case). 402 */ 403 MAKE_CHAIN(exthdrs.ip6e_hbh, mprev, nexthdrp, IPPROTO_HOPOPTS); 404 MAKE_CHAIN(exthdrs.ip6e_dest1, mprev, nexthdrp, 405 IPPROTO_DSTOPTS); 406 MAKE_CHAIN(exthdrs.ip6e_rthdr, mprev, nexthdrp, 407 IPPROTO_ROUTING); 408 409 M_CSUM_DATA_IPv6_SET(m->m_pkthdr.csum_data, 410 sizeof(struct ip6_hdr) + optlen); 411 } 412 413 /* Need to save for pmtu */ 414 finaldst = ip6->ip6_dst; 415 416 /* 417 * If there is a routing header, replace destination address field 418 * with the first hop of the routing header. 419 */ 420 if (exthdrs.ip6e_rthdr) { 421 struct ip6_rthdr *rh; 422 423 rh = mtod(exthdrs.ip6e_rthdr, struct ip6_rthdr *); 424 425 error = ip6_handle_rthdr(rh, ip6); 426 if (error != 0) { 427 IP6_STATINC(IP6_STAT_ODROPPED); 428 goto bad; 429 } 430 } 431 432 /* Source address validation */ 433 if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) && 434 (flags & IPV6_UNSPECSRC) == 0) { 435 error = EOPNOTSUPP; 436 IP6_STATINC(IP6_STAT_BADSCOPE); 437 goto bad; 438 } 439 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) { 440 error = EOPNOTSUPP; 441 IP6_STATINC(IP6_STAT_BADSCOPE); 442 goto bad; 443 } 444 445 IP6_STATINC(IP6_STAT_LOCALOUT); 446 447 /* 448 * Route packet. 449 */ 450 /* initialize cached route */ 451 if (ro == NULL) { 452 memset(&ip6route, 0, sizeof(ip6route)); 453 ro = &ip6route; 454 } 455 ro_pmtu = ro; 456 if (opt && opt->ip6po_rthdr) 457 ro = &opt->ip6po_route; 458 459 /* 460 * if specified, try to fill in the traffic class field. 461 * do not override if a non-zero value is already set. 462 * we check the diffserv field and the ecn field separately. 463 */ 464 if (opt && opt->ip6po_tclass >= 0) { 465 int mask = 0; 466 467 if ((ip6->ip6_flow & htonl(0xfc << 20)) == 0) 468 mask |= 0xfc; 469 if ((ip6->ip6_flow & htonl(0x03 << 20)) == 0) 470 mask |= 0x03; 471 if (mask != 0) 472 ip6->ip6_flow |= htonl((opt->ip6po_tclass & mask) << 20); 473 } 474 475 /* fill in or override the hop limit field, if necessary. */ 476 if (opt && opt->ip6po_hlim != -1) 477 ip6->ip6_hlim = opt->ip6po_hlim & 0xff; 478 else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) { 479 if (im6o != NULL) 480 ip6->ip6_hlim = im6o->im6o_multicast_hlim; 481 else 482 ip6->ip6_hlim = ip6_defmcasthlim; 483 } 484 485 #ifdef IPSEC 486 if (needipsec) { 487 error = ipsec6_process_packet(m, sp->req, flags); 488 489 /* 490 * Preserve KAME behaviour: ENOENT can be returned 491 * when an SA acquire is in progress. Don't propagate 492 * this to user-level; it confuses applications. 493 * XXX this will go away when the SADB is redone. 494 */ 495 if (error == ENOENT) 496 error = 0; 497 498 goto done; 499 } 500 #endif 501 502 /* adjust pointer */ 503 ip6 = mtod(m, struct ip6_hdr *); 504 505 sockaddr_in6_init(&dst_sa, &ip6->ip6_dst, 0, 0, 0); 506 507 /* We do not need a route for multicast */ 508 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) { 509 struct in6_pktinfo *pi = NULL; 510 511 /* 512 * If the outgoing interface for the address is specified by 513 * the caller, use it. 514 */ 515 if (opt && (pi = opt->ip6po_pktinfo) != NULL) { 516 /* XXX boundary check is assumed to be already done. */ 517 ifp = if_get_byindex(pi->ipi6_ifindex, &psref); 518 } else if (im6o != NULL) { 519 ifp = if_get_byindex(im6o->im6o_multicast_if_index, 520 &psref); 521 } 522 } 523 524 if (ifp == NULL) { 525 error = in6_selectroute(&dst_sa, opt, &ro, &rt, true); 526 if (error != 0) 527 goto bad; 528 ifp = if_get_byindex(rt->rt_ifp->if_index, &psref); 529 } 530 531 if (rt == NULL) { 532 /* 533 * If in6_selectroute() does not return a route entry, 534 * dst may not have been updated. 535 */ 536 error = rtcache_setdst(ro, sin6tosa(&dst_sa)); 537 if (error) { 538 IP6_STATINC(IP6_STAT_ODROPPED); 539 goto bad; 540 } 541 } 542 543 /* 544 * then rt (for unicast) and ifp must be non-NULL valid values. 545 */ 546 if ((flags & IPV6_FORWARDING) == 0) { 547 /* XXX: the FORWARDING flag can be set for mrouting. */ 548 in6_ifstat_inc(ifp, ifs6_out_request); 549 } 550 if (rt != NULL) { 551 ia = (struct in6_ifaddr *)(rt->rt_ifa); 552 rt->rt_use++; 553 } 554 555 /* 556 * The outgoing interface must be in the zone of source and 557 * destination addresses. We should use ia_ifp to support the 558 * case of sending packets to an address of our own. 559 */ 560 if (ia != NULL) { 561 origifp = ia->ia_ifp; 562 if (if_is_deactivated(origifp)) { 563 IP6_STATINC(IP6_STAT_ODROPPED); 564 goto bad; 565 } 566 if_acquire(origifp, &psref_ia); 567 release_psref_ia = true; 568 } else 569 origifp = ifp; 570 571 src0 = ip6->ip6_src; 572 if (in6_setscope(&src0, origifp, &zone)) 573 goto badscope; 574 sockaddr_in6_init(&src_sa, &ip6->ip6_src, 0, 0, 0); 575 if (sa6_recoverscope(&src_sa) || zone != src_sa.sin6_scope_id) 576 goto badscope; 577 578 dst0 = ip6->ip6_dst; 579 if (in6_setscope(&dst0, origifp, &zone)) 580 goto badscope; 581 /* re-initialize to be sure */ 582 sockaddr_in6_init(&dst_sa, &ip6->ip6_dst, 0, 0, 0); 583 if (sa6_recoverscope(&dst_sa) || zone != dst_sa.sin6_scope_id) 584 goto badscope; 585 586 /* scope check is done. */ 587 588 /* Ensure we only send from a valid address. */ 589 if ((ifp->if_flags & IFF_LOOPBACK) == 0 && 590 (flags & IPV6_FORWARDING) == 0 && 591 (error = ip6_ifaddrvalid(&src0, &dst0)) != 0) 592 { 593 char ip6buf[INET6_ADDRSTRLEN]; 594 nd6log(LOG_ERR, 595 "refusing to send from invalid address %s (pid %d)\n", 596 IN6_PRINT(ip6buf, &src0), curproc->p_pid); 597 IP6_STATINC(IP6_STAT_ODROPPED); 598 in6_ifstat_inc(origifp, ifs6_out_discard); 599 if (error == 1) 600 /* 601 * Address exists, but is tentative or detached. 602 * We can't send from it because it's invalid, 603 * so we drop the packet. 604 */ 605 error = 0; 606 else 607 error = EADDRNOTAVAIL; 608 goto bad; 609 } 610 611 if (rt != NULL && (rt->rt_flags & RTF_GATEWAY) && 612 !IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) 613 dst = satocsin6(rt->rt_gateway); 614 else 615 dst = satocsin6(rtcache_getdst(ro)); 616 617 /* 618 * XXXXXX: original code follows: 619 */ 620 if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) 621 m->m_flags &= ~(M_BCAST | M_MCAST); /* just in case */ 622 else { 623 bool ingroup; 624 625 m->m_flags = (m->m_flags & ~M_BCAST) | M_MCAST; 626 627 in6_ifstat_inc(ifp, ifs6_out_mcast); 628 629 /* 630 * Confirm that the outgoing interface supports multicast. 631 */ 632 if (!(ifp->if_flags & IFF_MULTICAST)) { 633 IP6_STATINC(IP6_STAT_NOROUTE); 634 in6_ifstat_inc(ifp, ifs6_out_discard); 635 error = ENETUNREACH; 636 goto bad; 637 } 638 639 ingroup = in6_multi_group(&ip6->ip6_dst, ifp); 640 if (ingroup && (im6o == NULL || im6o->im6o_multicast_loop)) { 641 /* 642 * If we belong to the destination multicast group 643 * on the outgoing interface, and the caller did not 644 * forbid loopback, loop back a copy. 645 */ 646 KASSERT(dst != NULL); 647 ip6_mloopback(ifp, m, dst); 648 } else { 649 /* 650 * If we are acting as a multicast router, perform 651 * multicast forwarding as if the packet had just 652 * arrived on the interface to which we are about 653 * to send. The multicast forwarding function 654 * recursively calls this function, using the 655 * IPV6_FORWARDING flag to prevent infinite recursion. 656 * 657 * Multicasts that are looped back by ip6_mloopback(), 658 * above, will be forwarded by the ip6_input() routine, 659 * if necessary. 660 */ 661 if (ip6_mrouter && (flags & IPV6_FORWARDING) == 0) { 662 if (ip6_mforward(ip6, ifp, m) != 0) { 663 m_freem(m); 664 goto done; 665 } 666 } 667 } 668 /* 669 * Multicasts with a hoplimit of zero may be looped back, 670 * above, but must not be transmitted on a network. 671 * Also, multicasts addressed to the loopback interface 672 * are not sent -- the above call to ip6_mloopback() will 673 * loop back a copy if this host actually belongs to the 674 * destination group on the loopback interface. 675 */ 676 if (ip6->ip6_hlim == 0 || (ifp->if_flags & IFF_LOOPBACK) || 677 IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst)) { 678 m_freem(m); 679 goto done; 680 } 681 } 682 683 /* 684 * Fill the outgoing interface to tell the upper layer 685 * to increment per-interface statistics. 686 */ 687 if (ifpp) 688 *ifpp = ifp; 689 690 /* Determine path MTU. */ 691 /* 692 * ro_pmtu represent final destination while 693 * ro might represent immediate destination. 694 * Use ro_pmtu destination since MTU might differ. 695 */ 696 if (ro_pmtu != ro) { 697 union { 698 struct sockaddr dst; 699 struct sockaddr_in6 dst6; 700 } u; 701 702 /* ro_pmtu may not have a cache */ 703 sockaddr_in6_init(&u.dst6, &finaldst, 0, 0, 0); 704 rt_pmtu = rtcache_lookup(ro_pmtu, &u.dst); 705 } else 706 rt_pmtu = rt; 707 error = ip6_getpmtu(rt_pmtu, ifp, &mtu, &alwaysfrag); 708 if (rt_pmtu != NULL && rt_pmtu != rt) 709 rtcache_unref(rt_pmtu, ro_pmtu); 710 KASSERT(error == 0); /* ip6_getpmtu never fail if ifp is passed */ 711 712 /* 713 * The caller of this function may specify to use the minimum MTU 714 * in some cases. 715 * An advanced API option (IPV6_USE_MIN_MTU) can also override MTU 716 * setting. The logic is a bit complicated; by default, unicast 717 * packets will follow path MTU while multicast packets will be sent at 718 * the minimum MTU. If IP6PO_MINMTU_ALL is specified, all packets 719 * including unicast ones will be sent at the minimum MTU. Multicast 720 * packets will always be sent at the minimum MTU unless 721 * IP6PO_MINMTU_DISABLE is explicitly specified. 722 * See RFC 3542 for more details. 723 */ 724 if (mtu > IPV6_MMTU) { 725 if ((flags & IPV6_MINMTU)) 726 mtu = IPV6_MMTU; 727 else if (opt && opt->ip6po_minmtu == IP6PO_MINMTU_ALL) 728 mtu = IPV6_MMTU; 729 else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) && 730 (opt == NULL || 731 opt->ip6po_minmtu != IP6PO_MINMTU_DISABLE)) { 732 mtu = IPV6_MMTU; 733 } 734 } 735 736 /* 737 * clear embedded scope identifiers if necessary. 738 * in6_clearscope will touch the addresses only when necessary. 739 */ 740 in6_clearscope(&ip6->ip6_src); 741 in6_clearscope(&ip6->ip6_dst); 742 743 /* 744 * If the outgoing packet contains a hop-by-hop options header, 745 * it must be examined and processed even by the source node. 746 * (RFC 2460, section 4.) 747 * 748 * XXX Is this really necessary? 749 */ 750 if (ip6->ip6_nxt == IPPROTO_HOPOPTS) { 751 u_int32_t dummy1; /* XXX unused */ 752 u_int32_t dummy2; /* XXX unused */ 753 int hoff = sizeof(struct ip6_hdr); 754 755 if (ip6_hopopts_input(&dummy1, &dummy2, &m, &hoff)) { 756 /* m was already freed at this point */ 757 error = EINVAL; 758 goto done; 759 } 760 761 ip6 = mtod(m, struct ip6_hdr *); 762 } 763 764 /* 765 * Run through list of hooks for output packets. 766 */ 767 error = pfil_run_hooks(inet6_pfil_hook, &m, ifp, PFIL_OUT); 768 if (error != 0 || m == NULL) { 769 IP6_STATINC(IP6_STAT_PFILDROP_OUT); 770 goto done; 771 } 772 ip6 = mtod(m, struct ip6_hdr *); 773 774 /* 775 * Send the packet to the outgoing interface. 776 * If necessary, do IPv6 fragmentation before sending. 777 * 778 * the logic here is rather complex: 779 * 1: normal case (dontfrag == 0, alwaysfrag == 0) 780 * 1-a: send as is if tlen <= path mtu 781 * 1-b: fragment if tlen > path mtu 782 * 783 * 2: if user asks us not to fragment (dontfrag == 1) 784 * 2-a: send as is if tlen <= interface mtu 785 * 2-b: error if tlen > interface mtu 786 * 787 * 3: if we always need to attach fragment header (alwaysfrag == 1) 788 * always fragment 789 * 790 * 4: if dontfrag == 1 && alwaysfrag == 1 791 * error, as we cannot handle this conflicting request 792 */ 793 tlen = m->m_pkthdr.len; 794 tso = (m->m_pkthdr.csum_flags & M_CSUM_TSOv6) != 0; 795 if (opt && (opt->ip6po_flags & IP6PO_DONTFRAG)) 796 dontfrag = 1; 797 else 798 dontfrag = 0; 799 800 if (dontfrag && alwaysfrag) { /* case 4 */ 801 /* conflicting request - can't transmit */ 802 IP6_STATINC(IP6_STAT_CANTFRAG); 803 error = EMSGSIZE; 804 goto bad; 805 } 806 if (dontfrag && (!tso && tlen > ifp->if_mtu)) { /* case 2-b */ 807 /* 808 * Even if the DONTFRAG option is specified, we cannot send the 809 * packet when the data length is larger than the MTU of the 810 * outgoing interface. 811 * Notify the error by sending IPV6_PATHMTU ancillary data as 812 * well as returning an error code (the latter is not described 813 * in the API spec.) 814 */ 815 u_int32_t mtu32; 816 struct ip6ctlparam ip6cp; 817 818 mtu32 = (u_int32_t)mtu; 819 memset(&ip6cp, 0, sizeof(ip6cp)); 820 ip6cp.ip6c_cmdarg = (void *)&mtu32; 821 pfctlinput2(PRC_MSGSIZE, 822 rtcache_getdst(ro_pmtu), &ip6cp); 823 824 IP6_STATINC(IP6_STAT_CANTFRAG); 825 error = EMSGSIZE; 826 goto bad; 827 } 828 829 /* 830 * transmit packet without fragmentation 831 */ 832 if (dontfrag || (!alwaysfrag && (tlen <= mtu || tso))) { 833 /* case 1-a and 2-a */ 834 struct in6_ifaddr *ia6; 835 int sw_csum; 836 int s; 837 838 ip6 = mtod(m, struct ip6_hdr *); 839 s = pserialize_read_enter(); 840 ia6 = in6_ifawithifp(ifp, &ip6->ip6_src); 841 if (ia6) { 842 /* Record statistics for this interface address. */ 843 ia6->ia_ifa.ifa_data.ifad_outbytes += m->m_pkthdr.len; 844 } 845 pserialize_read_exit(s); 846 847 sw_csum = m->m_pkthdr.csum_flags & ~ifp->if_csum_flags_tx; 848 if ((sw_csum & (M_CSUM_UDPv6|M_CSUM_TCPv6)) != 0) { 849 if (IN6_NEED_CHECKSUM(ifp, 850 sw_csum & (M_CSUM_UDPv6|M_CSUM_TCPv6))) { 851 in6_undefer_cksum_tcpudp(m); 852 } 853 m->m_pkthdr.csum_flags &= ~(M_CSUM_UDPv6|M_CSUM_TCPv6); 854 } 855 856 KASSERT(dst != NULL); 857 if (__predict_false(sw_csum & M_CSUM_TSOv6)) { 858 /* 859 * TSO6 is required by a packet, but disabled for 860 * the interface. 861 */ 862 error = ip6_tso_output(ifp, origifp, m, dst, rt); 863 } else 864 error = ip6_if_output(ifp, origifp, m, dst, rt); 865 goto done; 866 } 867 868 if (tso) { 869 IP6_STATINC(IP6_STAT_CANTFRAG); /* XXX */ 870 error = EINVAL; /* XXX */ 871 goto bad; 872 } 873 874 /* 875 * try to fragment the packet. case 1-b and 3 876 */ 877 if (mtu < IPV6_MMTU) { 878 /* path MTU cannot be less than IPV6_MMTU */ 879 IP6_STATINC(IP6_STAT_CANTFRAG); 880 error = EMSGSIZE; 881 in6_ifstat_inc(ifp, ifs6_out_fragfail); 882 goto bad; 883 } else if (ip6->ip6_plen == 0) { 884 /* jumbo payload cannot be fragmented */ 885 IP6_STATINC(IP6_STAT_CANTFRAG); 886 error = EMSGSIZE; 887 in6_ifstat_inc(ifp, ifs6_out_fragfail); 888 goto bad; 889 } else { 890 const uint32_t id = ip6_randomid(); 891 struct mbuf **mnext, *m_frgpart; 892 const int hlen = unfragpartlen; 893 struct ip6_frag *ip6f; 894 u_char nextproto; 895 896 if (mtu > IPV6_MAXPACKET) 897 mtu = IPV6_MAXPACKET; 898 899 /* 900 * Must be able to put at least 8 bytes per fragment. 901 */ 902 len = (mtu - hlen - sizeof(struct ip6_frag)) & ~7; 903 if (len < 8) { 904 IP6_STATINC(IP6_STAT_CANTFRAG); 905 error = EMSGSIZE; 906 in6_ifstat_inc(ifp, ifs6_out_fragfail); 907 goto bad; 908 } 909 910 mnext = &m->m_nextpkt; 911 912 /* 913 * Change the next header field of the last header in the 914 * unfragmentable part. 915 */ 916 if (exthdrs.ip6e_rthdr) { 917 nextproto = *mtod(exthdrs.ip6e_rthdr, u_char *); 918 *mtod(exthdrs.ip6e_rthdr, u_char *) = IPPROTO_FRAGMENT; 919 } else if (exthdrs.ip6e_dest1) { 920 nextproto = *mtod(exthdrs.ip6e_dest1, u_char *); 921 *mtod(exthdrs.ip6e_dest1, u_char *) = IPPROTO_FRAGMENT; 922 } else if (exthdrs.ip6e_hbh) { 923 nextproto = *mtod(exthdrs.ip6e_hbh, u_char *); 924 *mtod(exthdrs.ip6e_hbh, u_char *) = IPPROTO_FRAGMENT; 925 } else { 926 nextproto = ip6->ip6_nxt; 927 ip6->ip6_nxt = IPPROTO_FRAGMENT; 928 } 929 930 if ((m->m_pkthdr.csum_flags & (M_CSUM_UDPv6|M_CSUM_TCPv6)) 931 != 0) { 932 if (IN6_NEED_CHECKSUM(ifp, 933 m->m_pkthdr.csum_flags & 934 (M_CSUM_UDPv6|M_CSUM_TCPv6))) { 935 in6_undefer_cksum_tcpudp(m); 936 } 937 m->m_pkthdr.csum_flags &= ~(M_CSUM_UDPv6|M_CSUM_TCPv6); 938 } 939 940 /* 941 * Loop through length of segment after first fragment, 942 * make new header and copy data of each part and link onto 943 * chain. 944 */ 945 m0 = m; 946 for (off = hlen; off < tlen; off += len) { 947 struct mbuf *mlast; 948 949 MGETHDR(m, M_DONTWAIT, MT_HEADER); 950 if (!m) { 951 error = ENOBUFS; 952 IP6_STATINC(IP6_STAT_ODROPPED); 953 goto sendorfree; 954 } 955 m_reset_rcvif(m); 956 m->m_flags = m0->m_flags & M_COPYFLAGS; 957 *mnext = m; 958 mnext = &m->m_nextpkt; 959 m->m_data += max_linkhdr; 960 mhip6 = mtod(m, struct ip6_hdr *); 961 *mhip6 = *ip6; 962 m->m_len = sizeof(*mhip6); 963 964 ip6f = NULL; 965 error = ip6_insertfraghdr(m0, m, hlen, &ip6f); 966 if (error) { 967 IP6_STATINC(IP6_STAT_ODROPPED); 968 goto sendorfree; 969 } 970 971 /* Fill in the Frag6 Header */ 972 ip6f->ip6f_offlg = htons((u_int16_t)((off - hlen) & ~7)); 973 if (off + len >= tlen) 974 len = tlen - off; 975 else 976 ip6f->ip6f_offlg |= IP6F_MORE_FRAG; 977 ip6f->ip6f_reserved = 0; 978 ip6f->ip6f_ident = id; 979 ip6f->ip6f_nxt = nextproto; 980 981 mhip6->ip6_plen = htons((u_int16_t)(len + hlen + 982 sizeof(*ip6f) - sizeof(struct ip6_hdr))); 983 if ((m_frgpart = m_copym(m0, off, len, M_DONTWAIT)) == NULL) { 984 error = ENOBUFS; 985 IP6_STATINC(IP6_STAT_ODROPPED); 986 goto sendorfree; 987 } 988 for (mlast = m; mlast->m_next; mlast = mlast->m_next) 989 ; 990 mlast->m_next = m_frgpart; 991 992 m->m_pkthdr.len = len + hlen + sizeof(*ip6f); 993 m_reset_rcvif(m); 994 IP6_STATINC(IP6_STAT_OFRAGMENTS); 995 in6_ifstat_inc(ifp, ifs6_out_fragcreat); 996 } 997 998 in6_ifstat_inc(ifp, ifs6_out_fragok); 999 } 1000 1001 sendorfree: 1002 m = m0->m_nextpkt; 1003 m0->m_nextpkt = 0; 1004 m_freem(m0); 1005 for (m0 = m; m; m = m0) { 1006 m0 = m->m_nextpkt; 1007 m->m_nextpkt = 0; 1008 if (error == 0) { 1009 struct in6_ifaddr *ia6; 1010 int s; 1011 ip6 = mtod(m, struct ip6_hdr *); 1012 s = pserialize_read_enter(); 1013 ia6 = in6_ifawithifp(ifp, &ip6->ip6_src); 1014 if (ia6) { 1015 /* 1016 * Record statistics for this interface 1017 * address. 1018 */ 1019 ia6->ia_ifa.ifa_data.ifad_outbytes += 1020 m->m_pkthdr.len; 1021 } 1022 pserialize_read_exit(s); 1023 KASSERT(dst != NULL); 1024 error = ip6_if_output(ifp, origifp, m, dst, rt); 1025 } else 1026 m_freem(m); 1027 } 1028 1029 if (error == 0) 1030 IP6_STATINC(IP6_STAT_FRAGMENTED); 1031 1032 done: 1033 rtcache_unref(rt, ro); 1034 if (ro == &ip6route) 1035 rtcache_free(&ip6route); 1036 #ifdef IPSEC 1037 if (sp != NULL) 1038 KEY_SP_UNREF(&sp); 1039 #endif 1040 if_put(ifp, &psref); 1041 if (release_psref_ia) 1042 if_put(origifp, &psref_ia); 1043 curlwp_bindx(bound); 1044 1045 return error; 1046 1047 freehdrs: 1048 m_freem(exthdrs.ip6e_hbh); 1049 m_freem(exthdrs.ip6e_dest1); 1050 m_freem(exthdrs.ip6e_rthdr); 1051 m_freem(exthdrs.ip6e_dest2); 1052 /* FALLTHROUGH */ 1053 bad: 1054 m_freem(m); 1055 goto done; 1056 1057 badscope: 1058 IP6_STATINC(IP6_STAT_BADSCOPE); 1059 in6_ifstat_inc(origifp, ifs6_out_discard); 1060 if (error == 0) 1061 error = EHOSTUNREACH; /* XXX */ 1062 goto bad; 1063 } 1064 1065 static int 1066 ip6_copyexthdr(struct mbuf **mp, void *hdr, int hlen) 1067 { 1068 struct mbuf *m; 1069 1070 if (hlen > MCLBYTES) 1071 return ENOBUFS; /* XXX */ 1072 1073 MGET(m, M_DONTWAIT, MT_DATA); 1074 if (!m) 1075 return ENOBUFS; 1076 1077 if (hlen > MLEN) { 1078 MCLGET(m, M_DONTWAIT); 1079 if ((m->m_flags & M_EXT) == 0) { 1080 m_free(m); 1081 return ENOBUFS; 1082 } 1083 } 1084 m->m_len = hlen; 1085 if (hdr) 1086 memcpy(mtod(m, void *), hdr, hlen); 1087 1088 *mp = m; 1089 return 0; 1090 } 1091 1092 /* 1093 * Insert jumbo payload option. 1094 */ 1095 static int 1096 ip6_insert_jumboopt(struct ip6_exthdrs *exthdrs, u_int32_t plen) 1097 { 1098 struct mbuf *mopt; 1099 u_int8_t *optbuf; 1100 u_int32_t v; 1101 1102 #define JUMBOOPTLEN 8 /* length of jumbo payload option and padding */ 1103 1104 /* 1105 * If there is no hop-by-hop options header, allocate new one. 1106 * If there is one but it doesn't have enough space to store the 1107 * jumbo payload option, allocate a cluster to store the whole options. 1108 * Otherwise, use it to store the options. 1109 */ 1110 if (exthdrs->ip6e_hbh == NULL) { 1111 MGET(mopt, M_DONTWAIT, MT_DATA); 1112 if (mopt == 0) 1113 return (ENOBUFS); 1114 mopt->m_len = JUMBOOPTLEN; 1115 optbuf = mtod(mopt, u_int8_t *); 1116 optbuf[1] = 0; /* = ((JUMBOOPTLEN) >> 3) - 1 */ 1117 exthdrs->ip6e_hbh = mopt; 1118 } else { 1119 struct ip6_hbh *hbh; 1120 1121 mopt = exthdrs->ip6e_hbh; 1122 if (M_TRAILINGSPACE(mopt) < JUMBOOPTLEN) { 1123 const int oldoptlen = mopt->m_len; 1124 struct mbuf *n; 1125 1126 /* 1127 * Assumptions: 1128 * - exthdrs->ip6e_hbh is not referenced from places 1129 * other than exthdrs. 1130 * - exthdrs->ip6e_hbh is not an mbuf chain. 1131 */ 1132 KASSERT(mopt->m_next == NULL); 1133 1134 /* 1135 * Give up if the whole (new) hbh header does not fit 1136 * even in an mbuf cluster. 1137 */ 1138 if (oldoptlen + JUMBOOPTLEN > MCLBYTES) 1139 return ENOBUFS; 1140 1141 /* 1142 * At this point, we must always prepare a cluster. 1143 */ 1144 MGET(n, M_DONTWAIT, MT_DATA); 1145 if (n) { 1146 MCLGET(n, M_DONTWAIT); 1147 if ((n->m_flags & M_EXT) == 0) { 1148 m_freem(n); 1149 n = NULL; 1150 } 1151 } 1152 if (!n) 1153 return ENOBUFS; 1154 1155 n->m_len = oldoptlen + JUMBOOPTLEN; 1156 bcopy(mtod(mopt, void *), mtod(n, void *), 1157 oldoptlen); 1158 optbuf = mtod(n, u_int8_t *) + oldoptlen; 1159 m_freem(mopt); 1160 mopt = exthdrs->ip6e_hbh = n; 1161 } else { 1162 optbuf = mtod(mopt, u_int8_t *) + mopt->m_len; 1163 mopt->m_len += JUMBOOPTLEN; 1164 } 1165 optbuf[0] = IP6OPT_PADN; 1166 optbuf[1] = 0; 1167 1168 /* 1169 * Adjust the header length according to the pad and 1170 * the jumbo payload option. 1171 */ 1172 hbh = mtod(mopt, struct ip6_hbh *); 1173 hbh->ip6h_len += (JUMBOOPTLEN >> 3); 1174 } 1175 1176 /* fill in the option. */ 1177 optbuf[2] = IP6OPT_JUMBO; 1178 optbuf[3] = 4; 1179 v = (u_int32_t)htonl(plen + JUMBOOPTLEN); 1180 memcpy(&optbuf[4], &v, sizeof(u_int32_t)); 1181 1182 /* finally, adjust the packet header length */ 1183 exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN; 1184 1185 return 0; 1186 #undef JUMBOOPTLEN 1187 } 1188 1189 /* 1190 * Insert fragment header and copy unfragmentable header portions. 1191 * 1192 * *frghdrp will not be read, and it is guaranteed that either an 1193 * error is returned or that *frghdrp will point to space allocated 1194 * for the fragment header. 1195 * 1196 * On entry, m contains: 1197 * IPv6 Header 1198 * On exit, it contains: 1199 * IPv6 Header -> Unfragmentable Part -> Frag6 Header 1200 */ 1201 static int 1202 ip6_insertfraghdr(struct mbuf *m0, struct mbuf *m, int hlen, 1203 struct ip6_frag **frghdrp) 1204 { 1205 struct mbuf *n, *mlast; 1206 1207 if (hlen > sizeof(struct ip6_hdr)) { 1208 n = m_copym(m0, sizeof(struct ip6_hdr), 1209 hlen - sizeof(struct ip6_hdr), M_DONTWAIT); 1210 if (n == NULL) 1211 return ENOBUFS; 1212 m->m_next = n; 1213 } else 1214 n = m; 1215 1216 /* Search for the last mbuf of unfragmentable part. */ 1217 for (mlast = n; mlast->m_next; mlast = mlast->m_next) 1218 ; 1219 1220 if ((mlast->m_flags & M_EXT) == 0 && 1221 M_TRAILINGSPACE(mlast) >= sizeof(struct ip6_frag)) { 1222 /* use the trailing space of the last mbuf for the fragment hdr */ 1223 *frghdrp = (struct ip6_frag *)(mtod(mlast, char *) + 1224 mlast->m_len); 1225 mlast->m_len += sizeof(struct ip6_frag); 1226 } else { 1227 /* allocate a new mbuf for the fragment header */ 1228 struct mbuf *mfrg; 1229 1230 MGET(mfrg, M_DONTWAIT, MT_DATA); 1231 if (mfrg == NULL) 1232 return ENOBUFS; 1233 mfrg->m_len = sizeof(struct ip6_frag); 1234 *frghdrp = mtod(mfrg, struct ip6_frag *); 1235 mlast->m_next = mfrg; 1236 } 1237 1238 return 0; 1239 } 1240 1241 static int 1242 ip6_getpmtu(struct rtentry *rt, struct ifnet *ifp, u_long *mtup, 1243 int *alwaysfragp) 1244 { 1245 u_int32_t mtu = 0; 1246 int alwaysfrag = 0; 1247 int error = 0; 1248 1249 if (rt != NULL) { 1250 if (ifp == NULL) 1251 ifp = rt->rt_ifp; 1252 mtu = rt->rt_rmx.rmx_mtu; 1253 if (mtu == 0) 1254 mtu = ifp->if_mtu; 1255 else if (mtu < IPV6_MMTU) { 1256 /* 1257 * RFC2460 section 5, last paragraph: 1258 * if we record ICMPv6 too big message with 1259 * mtu < IPV6_MMTU, transmit packets sized IPV6_MMTU 1260 * or smaller, with fragment header attached. 1261 * (fragment header is needed regardless from the 1262 * packet size, for translators to identify packets) 1263 */ 1264 alwaysfrag = 1; 1265 mtu = IPV6_MMTU; 1266 } else if (mtu > ifp->if_mtu) { 1267 /* 1268 * The MTU on the route is larger than the MTU on 1269 * the interface! This shouldn't happen, unless the 1270 * MTU of the interface has been changed after the 1271 * interface was brought up. Change the MTU in the 1272 * route to match the interface MTU (as long as the 1273 * field isn't locked). 1274 */ 1275 mtu = ifp->if_mtu; 1276 if (!(rt->rt_rmx.rmx_locks & RTV_MTU)) 1277 rt->rt_rmx.rmx_mtu = mtu; 1278 } 1279 } else if (ifp) { 1280 mtu = ifp->if_mtu; 1281 } else 1282 error = EHOSTUNREACH; /* XXX */ 1283 1284 *mtup = mtu; 1285 if (alwaysfragp) 1286 *alwaysfragp = alwaysfrag; 1287 return (error); 1288 } 1289 1290 /* 1291 * IP6 socket option processing. 1292 */ 1293 int 1294 ip6_ctloutput(int op, struct socket *so, struct sockopt *sopt) 1295 { 1296 int optdatalen, uproto; 1297 void *optdata; 1298 struct inpcb *inp = sotoinpcb(so); 1299 struct ip_moptions **mopts; 1300 int error, optval; 1301 int level, optname; 1302 1303 KASSERT(solocked(so)); 1304 KASSERT(sopt != NULL); 1305 1306 level = sopt->sopt_level; 1307 optname = sopt->sopt_name; 1308 1309 error = optval = 0; 1310 uproto = (int)so->so_proto->pr_protocol; 1311 1312 switch (level) { 1313 case IPPROTO_IP: 1314 switch (optname) { 1315 case IP_ADD_MEMBERSHIP: 1316 case IP_DROP_MEMBERSHIP: 1317 case IP_MULTICAST_IF: 1318 case IP_MULTICAST_LOOP: 1319 case IP_MULTICAST_TTL: 1320 mopts = &inp->inp_moptions; 1321 switch (op) { 1322 case PRCO_GETOPT: 1323 return ip_getmoptions(*mopts, sopt); 1324 case PRCO_SETOPT: 1325 return ip_setmoptions(mopts, sopt); 1326 default: 1327 return EINVAL; 1328 } 1329 default: 1330 return ENOPROTOOPT; 1331 } 1332 case IPPROTO_IPV6: 1333 break; 1334 default: 1335 return ENOPROTOOPT; 1336 } 1337 switch (op) { 1338 case PRCO_SETOPT: 1339 switch (optname) { 1340 #ifdef RFC2292 1341 case IPV6_2292PKTOPTIONS: 1342 error = ip6_pcbopts(&in6p_outputopts(inp), so, sopt); 1343 break; 1344 #endif 1345 1346 /* 1347 * Use of some Hop-by-Hop options or some 1348 * Destination options, might require special 1349 * privilege. That is, normal applications 1350 * (without special privilege) might be forbidden 1351 * from setting certain options in outgoing packets, 1352 * and might never see certain options in received 1353 * packets. [RFC 2292 Section 6] 1354 * KAME specific note: 1355 * KAME prevents non-privileged users from sending or 1356 * receiving ANY hbh/dst options in order to avoid 1357 * overhead of parsing options in the kernel. 1358 */ 1359 case IPV6_RECVHOPOPTS: 1360 case IPV6_RECVDSTOPTS: 1361 case IPV6_RECVRTHDRDSTOPTS: 1362 error = kauth_authorize_network( 1363 kauth_cred_get(), 1364 KAUTH_NETWORK_IPV6, KAUTH_REQ_NETWORK_IPV6_HOPBYHOP, 1365 NULL, NULL, NULL); 1366 if (error) 1367 break; 1368 /* FALLTHROUGH */ 1369 case IPV6_UNICAST_HOPS: 1370 case IPV6_HOPLIMIT: 1371 case IPV6_FAITH: 1372 1373 case IPV6_RECVPKTINFO: 1374 case IPV6_RECVHOPLIMIT: 1375 case IPV6_RECVRTHDR: 1376 case IPV6_RECVPATHMTU: 1377 case IPV6_RECVTCLASS: 1378 case IPV6_V6ONLY: 1379 case IPV6_BINDANY: 1380 error = sockopt_getint(sopt, &optval); 1381 if (error) 1382 break; 1383 switch (optname) { 1384 case IPV6_UNICAST_HOPS: 1385 if (optval < -1 || optval >= 256) 1386 error = EINVAL; 1387 else { 1388 /* -1 = kernel default */ 1389 in6p_hops6(inp) = optval; 1390 } 1391 break; 1392 #define OPTSET(bit) \ 1393 do { \ 1394 if (optval) \ 1395 inp->inp_flags |= (bit); \ 1396 else \ 1397 inp->inp_flags &= ~(bit); \ 1398 } while (/*CONSTCOND*/ 0) 1399 1400 #ifdef RFC2292 1401 #define OPTSET2292(bit) \ 1402 do { \ 1403 inp->inp_flags |= IN6P_RFC2292; \ 1404 if (optval) \ 1405 inp->inp_flags |= (bit); \ 1406 else \ 1407 inp->inp_flags &= ~(bit); \ 1408 } while (/*CONSTCOND*/ 0) 1409 #endif 1410 1411 #define OPTBIT(bit) (inp->inp_flags & (bit) ? 1 : 0) 1412 1413 case IPV6_RECVPKTINFO: 1414 #ifdef RFC2292 1415 /* cannot mix with RFC2292 */ 1416 if (OPTBIT(IN6P_RFC2292)) { 1417 error = EINVAL; 1418 break; 1419 } 1420 #endif 1421 OPTSET(IN6P_PKTINFO); 1422 break; 1423 1424 case IPV6_HOPLIMIT: 1425 { 1426 struct ip6_pktopts **optp; 1427 1428 #ifdef RFC2292 1429 /* cannot mix with RFC2292 */ 1430 if (OPTBIT(IN6P_RFC2292)) { 1431 error = EINVAL; 1432 break; 1433 } 1434 #endif 1435 optp = &in6p_outputopts(inp); 1436 error = ip6_pcbopt(IPV6_HOPLIMIT, 1437 (u_char *)&optval, 1438 sizeof(optval), 1439 optp, 1440 kauth_cred_get(), uproto); 1441 break; 1442 } 1443 1444 case IPV6_RECVHOPLIMIT: 1445 #ifdef RFC2292 1446 /* cannot mix with RFC2292 */ 1447 if (OPTBIT(IN6P_RFC2292)) { 1448 error = EINVAL; 1449 break; 1450 } 1451 #endif 1452 OPTSET(IN6P_HOPLIMIT); 1453 break; 1454 1455 case IPV6_RECVHOPOPTS: 1456 #ifdef RFC2292 1457 /* cannot mix with RFC2292 */ 1458 if (OPTBIT(IN6P_RFC2292)) { 1459 error = EINVAL; 1460 break; 1461 } 1462 #endif 1463 OPTSET(IN6P_HOPOPTS); 1464 break; 1465 1466 case IPV6_RECVDSTOPTS: 1467 #ifdef RFC2292 1468 /* cannot mix with RFC2292 */ 1469 if (OPTBIT(IN6P_RFC2292)) { 1470 error = EINVAL; 1471 break; 1472 } 1473 #endif 1474 OPTSET(IN6P_DSTOPTS); 1475 break; 1476 1477 case IPV6_RECVRTHDRDSTOPTS: 1478 #ifdef RFC2292 1479 /* cannot mix with RFC2292 */ 1480 if (OPTBIT(IN6P_RFC2292)) { 1481 error = EINVAL; 1482 break; 1483 } 1484 #endif 1485 OPTSET(IN6P_RTHDRDSTOPTS); 1486 break; 1487 1488 case IPV6_RECVRTHDR: 1489 #ifdef RFC2292 1490 /* cannot mix with RFC2292 */ 1491 if (OPTBIT(IN6P_RFC2292)) { 1492 error = EINVAL; 1493 break; 1494 } 1495 #endif 1496 OPTSET(IN6P_RTHDR); 1497 break; 1498 1499 case IPV6_FAITH: 1500 OPTSET(IN6P_FAITH); 1501 break; 1502 1503 case IPV6_RECVPATHMTU: 1504 /* 1505 * We ignore this option for TCP 1506 * sockets. 1507 * (RFC3542 leaves this case 1508 * unspecified.) 1509 */ 1510 if (uproto != IPPROTO_TCP) 1511 OPTSET(IN6P_MTU); 1512 break; 1513 1514 case IPV6_V6ONLY: 1515 /* 1516 * make setsockopt(IPV6_V6ONLY) 1517 * available only prior to bind(2). 1518 * see ipng mailing list, Jun 22 2001. 1519 */ 1520 if (inp->inp_lport || 1521 !IN6_IS_ADDR_UNSPECIFIED(&in6p_laddr(inp))) { 1522 error = EINVAL; 1523 break; 1524 } 1525 #ifdef INET6_BINDV6ONLY 1526 if (!optval) 1527 error = EINVAL; 1528 #else 1529 OPTSET(IN6P_IPV6_V6ONLY); 1530 #endif 1531 break; 1532 1533 case IPV6_RECVTCLASS: 1534 #ifdef RFC2292 1535 /* cannot mix with RFC2292 XXX */ 1536 if (OPTBIT(IN6P_RFC2292)) { 1537 error = EINVAL; 1538 break; 1539 } 1540 #endif 1541 OPTSET(IN6P_TCLASS); 1542 break; 1543 1544 case IPV6_BINDANY: 1545 error = kauth_authorize_network( 1546 kauth_cred_get(), KAUTH_NETWORK_BIND, 1547 KAUTH_REQ_NETWORK_BIND_ANYADDR, so, NULL, 1548 NULL); 1549 if (error) 1550 break; 1551 OPTSET(IN6P_BINDANY); 1552 break; 1553 } 1554 break; 1555 1556 case IPV6_OTCLASS: 1557 { 1558 struct ip6_pktopts **optp; 1559 u_int8_t tclass; 1560 1561 error = sockopt_get(sopt, &tclass, sizeof(tclass)); 1562 if (error) 1563 break; 1564 optp = &in6p_outputopts(inp); 1565 error = ip6_pcbopt(optname, 1566 (u_char *)&tclass, 1567 sizeof(tclass), 1568 optp, 1569 kauth_cred_get(), uproto); 1570 break; 1571 } 1572 1573 case IPV6_TCLASS: 1574 case IPV6_DONTFRAG: 1575 case IPV6_USE_MIN_MTU: 1576 case IPV6_PREFER_TEMPADDR: 1577 error = sockopt_getint(sopt, &optval); 1578 if (error) 1579 break; 1580 { 1581 struct ip6_pktopts **optp; 1582 optp = &in6p_outputopts(inp); 1583 error = ip6_pcbopt(optname, 1584 (u_char *)&optval, 1585 sizeof(optval), 1586 optp, 1587 kauth_cred_get(), uproto); 1588 break; 1589 } 1590 1591 #ifdef RFC2292 1592 case IPV6_2292PKTINFO: 1593 case IPV6_2292HOPLIMIT: 1594 case IPV6_2292HOPOPTS: 1595 case IPV6_2292DSTOPTS: 1596 case IPV6_2292RTHDR: 1597 /* RFC 2292 */ 1598 error = sockopt_getint(sopt, &optval); 1599 if (error) 1600 break; 1601 1602 switch (optname) { 1603 case IPV6_2292PKTINFO: 1604 OPTSET2292(IN6P_PKTINFO); 1605 break; 1606 case IPV6_2292HOPLIMIT: 1607 OPTSET2292(IN6P_HOPLIMIT); 1608 break; 1609 case IPV6_2292HOPOPTS: 1610 /* 1611 * Check super-user privilege. 1612 * See comments for IPV6_RECVHOPOPTS. 1613 */ 1614 error = kauth_authorize_network( 1615 kauth_cred_get(), 1616 KAUTH_NETWORK_IPV6, 1617 KAUTH_REQ_NETWORK_IPV6_HOPBYHOP, NULL, 1618 NULL, NULL); 1619 if (error) 1620 return (error); 1621 OPTSET2292(IN6P_HOPOPTS); 1622 break; 1623 case IPV6_2292DSTOPTS: 1624 error = kauth_authorize_network( 1625 kauth_cred_get(), 1626 KAUTH_NETWORK_IPV6, 1627 KAUTH_REQ_NETWORK_IPV6_HOPBYHOP, NULL, 1628 NULL, NULL); 1629 if (error) 1630 return (error); 1631 OPTSET2292(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); /* XXX */ 1632 break; 1633 case IPV6_2292RTHDR: 1634 OPTSET2292(IN6P_RTHDR); 1635 break; 1636 } 1637 break; 1638 #endif 1639 case IPV6_PKTINFO: 1640 case IPV6_HOPOPTS: 1641 case IPV6_RTHDR: 1642 case IPV6_DSTOPTS: 1643 case IPV6_RTHDRDSTOPTS: 1644 case IPV6_NEXTHOP: { 1645 /* new advanced API (RFC3542) */ 1646 void *optbuf; 1647 int optbuflen; 1648 struct ip6_pktopts **optp; 1649 1650 #ifdef RFC2292 1651 /* cannot mix with RFC2292 */ 1652 if (OPTBIT(IN6P_RFC2292)) { 1653 error = EINVAL; 1654 break; 1655 } 1656 #endif 1657 1658 optbuflen = sopt->sopt_size; 1659 optbuf = malloc(optbuflen, M_IP6OPT, M_NOWAIT); 1660 if (optbuf == NULL) { 1661 error = ENOBUFS; 1662 break; 1663 } 1664 1665 error = sockopt_get(sopt, optbuf, optbuflen); 1666 if (error) { 1667 free(optbuf, M_IP6OPT); 1668 break; 1669 } 1670 optp = &in6p_outputopts(inp); 1671 error = ip6_pcbopt(optname, optbuf, optbuflen, 1672 optp, kauth_cred_get(), uproto); 1673 1674 free(optbuf, M_IP6OPT); 1675 break; 1676 } 1677 #undef OPTSET 1678 1679 case IPV6_MULTICAST_IF: 1680 case IPV6_MULTICAST_HOPS: 1681 case IPV6_MULTICAST_LOOP: 1682 case IPV6_JOIN_GROUP: 1683 case IPV6_LEAVE_GROUP: 1684 error = ip6_setmoptions(sopt, inp); 1685 break; 1686 1687 case IPV6_PORTRANGE: 1688 error = sockopt_getint(sopt, &optval); 1689 if (error) 1690 break; 1691 1692 switch (optval) { 1693 case IPV6_PORTRANGE_DEFAULT: 1694 inp->inp_flags &= ~(IN6P_LOWPORT); 1695 inp->inp_flags &= ~(IN6P_HIGHPORT); 1696 break; 1697 1698 case IPV6_PORTRANGE_HIGH: 1699 inp->inp_flags &= ~(IN6P_LOWPORT); 1700 inp->inp_flags |= IN6P_HIGHPORT; 1701 break; 1702 1703 case IPV6_PORTRANGE_LOW: 1704 inp->inp_flags &= ~(IN6P_HIGHPORT); 1705 inp->inp_flags |= IN6P_LOWPORT; 1706 break; 1707 1708 default: 1709 error = EINVAL; 1710 break; 1711 } 1712 break; 1713 1714 case IPV6_PORTALGO: 1715 error = sockopt_getint(sopt, &optval); 1716 if (error) 1717 break; 1718 1719 error = portalgo_algo_index_select(inp, optval); 1720 break; 1721 1722 #if defined(IPSEC) 1723 case IPV6_IPSEC_POLICY: 1724 if (ipsec_enabled) { 1725 error = ipsec_set_policy(inp, 1726 sopt->sopt_data, sopt->sopt_size, 1727 kauth_cred_get()); 1728 } else 1729 error = ENOPROTOOPT; 1730 break; 1731 #endif /* IPSEC */ 1732 1733 default: 1734 error = ENOPROTOOPT; 1735 break; 1736 } 1737 break; 1738 1739 case PRCO_GETOPT: 1740 switch (optname) { 1741 #ifdef RFC2292 1742 case IPV6_2292PKTOPTIONS: 1743 /* 1744 * RFC3542 (effectively) deprecated the 1745 * semantics of the 2292-style pktoptions. 1746 * Since it was not reliable in nature (i.e., 1747 * applications had to expect the lack of some 1748 * information after all), it would make sense 1749 * to simplify this part by always returning 1750 * empty data. 1751 */ 1752 break; 1753 #endif 1754 1755 case IPV6_RECVHOPOPTS: 1756 case IPV6_RECVDSTOPTS: 1757 case IPV6_RECVRTHDRDSTOPTS: 1758 case IPV6_UNICAST_HOPS: 1759 case IPV6_RECVPKTINFO: 1760 case IPV6_RECVHOPLIMIT: 1761 case IPV6_RECVRTHDR: 1762 case IPV6_RECVPATHMTU: 1763 1764 case IPV6_FAITH: 1765 case IPV6_V6ONLY: 1766 case IPV6_PORTRANGE: 1767 case IPV6_RECVTCLASS: 1768 case IPV6_BINDANY: 1769 switch (optname) { 1770 1771 case IPV6_RECVHOPOPTS: 1772 optval = OPTBIT(IN6P_HOPOPTS); 1773 break; 1774 1775 case IPV6_RECVDSTOPTS: 1776 optval = OPTBIT(IN6P_DSTOPTS); 1777 break; 1778 1779 case IPV6_RECVRTHDRDSTOPTS: 1780 optval = OPTBIT(IN6P_RTHDRDSTOPTS); 1781 break; 1782 1783 case IPV6_UNICAST_HOPS: 1784 optval = in6p_hops6(inp); 1785 break; 1786 1787 case IPV6_RECVPKTINFO: 1788 optval = OPTBIT(IN6P_PKTINFO); 1789 break; 1790 1791 case IPV6_RECVHOPLIMIT: 1792 optval = OPTBIT(IN6P_HOPLIMIT); 1793 break; 1794 1795 case IPV6_RECVRTHDR: 1796 optval = OPTBIT(IN6P_RTHDR); 1797 break; 1798 1799 case IPV6_RECVPATHMTU: 1800 optval = OPTBIT(IN6P_MTU); 1801 break; 1802 1803 case IPV6_FAITH: 1804 optval = OPTBIT(IN6P_FAITH); 1805 break; 1806 1807 case IPV6_V6ONLY: 1808 optval = OPTBIT(IN6P_IPV6_V6ONLY); 1809 break; 1810 1811 case IPV6_PORTRANGE: 1812 { 1813 int flags; 1814 flags = inp->inp_flags; 1815 if (flags & IN6P_HIGHPORT) 1816 optval = IPV6_PORTRANGE_HIGH; 1817 else if (flags & IN6P_LOWPORT) 1818 optval = IPV6_PORTRANGE_LOW; 1819 else 1820 optval = 0; 1821 break; 1822 } 1823 case IPV6_RECVTCLASS: 1824 optval = OPTBIT(IN6P_TCLASS); 1825 break; 1826 1827 case IPV6_BINDANY: 1828 optval = OPTBIT(IN6P_BINDANY); 1829 break; 1830 } 1831 1832 if (error) 1833 break; 1834 error = sockopt_setint(sopt, optval); 1835 break; 1836 1837 case IPV6_PATHMTU: 1838 { 1839 u_long pmtu = 0; 1840 struct ip6_mtuinfo mtuinfo; 1841 struct route *ro = &inp->inp_route; 1842 struct rtentry *rt; 1843 union { 1844 struct sockaddr dst; 1845 struct sockaddr_in6 dst6; 1846 } u; 1847 1848 if (!(so->so_state & SS_ISCONNECTED)) 1849 return (ENOTCONN); 1850 /* 1851 * XXX: we dot not consider the case of source 1852 * routing, or optional information to specify 1853 * the outgoing interface. 1854 */ 1855 sockaddr_in6_init(&u.dst6, &in6p_faddr(inp), 0, 0, 0); 1856 rt = rtcache_lookup(ro, &u.dst); 1857 error = ip6_getpmtu(rt, NULL, &pmtu, NULL); 1858 rtcache_unref(rt, ro); 1859 if (error) 1860 break; 1861 if (pmtu > IPV6_MAXPACKET) 1862 pmtu = IPV6_MAXPACKET; 1863 1864 memset(&mtuinfo, 0, sizeof(mtuinfo)); 1865 mtuinfo.ip6m_mtu = (u_int32_t)pmtu; 1866 optdata = (void *)&mtuinfo; 1867 optdatalen = sizeof(mtuinfo); 1868 if (optdatalen > MCLBYTES) 1869 return (EMSGSIZE); /* XXX */ 1870 error = sockopt_set(sopt, optdata, optdatalen); 1871 break; 1872 } 1873 1874 #ifdef RFC2292 1875 case IPV6_2292PKTINFO: 1876 case IPV6_2292HOPLIMIT: 1877 case IPV6_2292HOPOPTS: 1878 case IPV6_2292RTHDR: 1879 case IPV6_2292DSTOPTS: 1880 switch (optname) { 1881 case IPV6_2292PKTINFO: 1882 optval = OPTBIT(IN6P_PKTINFO); 1883 break; 1884 case IPV6_2292HOPLIMIT: 1885 optval = OPTBIT(IN6P_HOPLIMIT); 1886 break; 1887 case IPV6_2292HOPOPTS: 1888 optval = OPTBIT(IN6P_HOPOPTS); 1889 break; 1890 case IPV6_2292RTHDR: 1891 optval = OPTBIT(IN6P_RTHDR); 1892 break; 1893 case IPV6_2292DSTOPTS: 1894 optval = OPTBIT(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); 1895 break; 1896 } 1897 error = sockopt_setint(sopt, optval); 1898 break; 1899 #endif 1900 case IPV6_PKTINFO: 1901 case IPV6_HOPOPTS: 1902 case IPV6_RTHDR: 1903 case IPV6_DSTOPTS: 1904 case IPV6_RTHDRDSTOPTS: 1905 case IPV6_NEXTHOP: 1906 case IPV6_OTCLASS: 1907 case IPV6_TCLASS: 1908 case IPV6_DONTFRAG: 1909 case IPV6_USE_MIN_MTU: 1910 case IPV6_PREFER_TEMPADDR: 1911 error = ip6_getpcbopt(in6p_outputopts(inp), 1912 optname, sopt); 1913 break; 1914 1915 case IPV6_MULTICAST_IF: 1916 case IPV6_MULTICAST_HOPS: 1917 case IPV6_MULTICAST_LOOP: 1918 case IPV6_JOIN_GROUP: 1919 case IPV6_LEAVE_GROUP: 1920 error = ip6_getmoptions(sopt, inp); 1921 break; 1922 1923 case IPV6_PORTALGO: 1924 optval = inp->inp_portalgo; 1925 error = sockopt_setint(sopt, optval); 1926 break; 1927 1928 #if defined(IPSEC) 1929 case IPV6_IPSEC_POLICY: 1930 if (ipsec_used) { 1931 struct mbuf *m = NULL; 1932 1933 /* 1934 * XXX: this will return EINVAL as sopt is 1935 * empty 1936 */ 1937 error = ipsec_get_policy(inp, sopt->sopt_data, 1938 sopt->sopt_size, &m); 1939 if (!error) 1940 error = sockopt_setmbuf(sopt, m); 1941 } else 1942 error = ENOPROTOOPT; 1943 break; 1944 #endif /* IPSEC */ 1945 1946 default: 1947 error = ENOPROTOOPT; 1948 break; 1949 } 1950 break; 1951 } 1952 return (error); 1953 } 1954 1955 int 1956 ip6_raw_ctloutput(int op, struct socket *so, struct sockopt *sopt) 1957 { 1958 int error = 0, optval; 1959 const int icmp6off = offsetof(struct icmp6_hdr, icmp6_cksum); 1960 struct inpcb *inp = sotoinpcb(so); 1961 int level, optname; 1962 1963 KASSERT(sopt != NULL); 1964 1965 level = sopt->sopt_level; 1966 optname = sopt->sopt_name; 1967 1968 if (level != IPPROTO_IPV6) { 1969 return ENOPROTOOPT; 1970 } 1971 1972 switch (optname) { 1973 case IPV6_CHECKSUM: 1974 /* 1975 * For ICMPv6 sockets, no modification allowed for checksum 1976 * offset, permit "no change" values to help existing apps. 1977 * 1978 * XXX RFC3542 says: "An attempt to set IPV6_CHECKSUM 1979 * for an ICMPv6 socket will fail." The current 1980 * behavior does not meet RFC3542. 1981 */ 1982 switch (op) { 1983 case PRCO_SETOPT: 1984 error = sockopt_getint(sopt, &optval); 1985 if (error) 1986 break; 1987 if (optval < -1 || 1988 (optval > 0 && (optval % 2) != 0)) { 1989 /* 1990 * The API assumes non-negative even offset 1991 * values or -1 as a special value. 1992 */ 1993 error = EINVAL; 1994 } else if (so->so_proto->pr_protocol == 1995 IPPROTO_ICMPV6) { 1996 if (optval != icmp6off) 1997 error = EINVAL; 1998 } else 1999 in6p_cksum(inp) = optval; 2000 break; 2001 2002 case PRCO_GETOPT: 2003 if (so->so_proto->pr_protocol == IPPROTO_ICMPV6) 2004 optval = icmp6off; 2005 else 2006 optval = in6p_cksum(inp); 2007 2008 error = sockopt_setint(sopt, optval); 2009 break; 2010 2011 default: 2012 error = EINVAL; 2013 break; 2014 } 2015 break; 2016 2017 default: 2018 error = ENOPROTOOPT; 2019 break; 2020 } 2021 2022 return (error); 2023 } 2024 2025 #ifdef RFC2292 2026 /* 2027 * Set up IP6 options in pcb for insertion in output packets or 2028 * specifying behavior of outgoing packets. 2029 */ 2030 static int 2031 ip6_pcbopts(struct ip6_pktopts **pktopt, struct socket *so, 2032 struct sockopt *sopt) 2033 { 2034 struct ip6_pktopts *opt = *pktopt; 2035 struct mbuf *m; 2036 int error = 0; 2037 2038 KASSERT(solocked(so)); 2039 2040 /* turn off any old options. */ 2041 if (opt) { 2042 #ifdef DIAGNOSTIC 2043 if (opt->ip6po_pktinfo || opt->ip6po_nexthop || 2044 opt->ip6po_hbh || opt->ip6po_dest1 || opt->ip6po_dest2 || 2045 opt->ip6po_rhinfo.ip6po_rhi_rthdr) 2046 printf("ip6_pcbopts: all specified options are cleared.\n"); 2047 #endif 2048 ip6_clearpktopts(opt, -1); 2049 } else { 2050 opt = malloc(sizeof(*opt), M_IP6OPT, M_NOWAIT); 2051 if (opt == NULL) 2052 return (ENOBUFS); 2053 } 2054 *pktopt = NULL; 2055 2056 if (sopt == NULL || sopt->sopt_size == 0) { 2057 /* 2058 * Only turning off any previous options, regardless of 2059 * whether the opt is just created or given. 2060 */ 2061 free(opt, M_IP6OPT); 2062 return (0); 2063 } 2064 2065 /* set options specified by user. */ 2066 m = sockopt_getmbuf(sopt); 2067 if (m == NULL) { 2068 free(opt, M_IP6OPT); 2069 return (ENOBUFS); 2070 } 2071 2072 error = ip6_setpktopts(m, opt, NULL, kauth_cred_get(), 2073 so->so_proto->pr_protocol); 2074 m_freem(m); 2075 if (error != 0) { 2076 ip6_clearpktopts(opt, -1); /* XXX: discard all options */ 2077 free(opt, M_IP6OPT); 2078 return (error); 2079 } 2080 *pktopt = opt; 2081 return (0); 2082 } 2083 #endif 2084 2085 /* 2086 * initialize ip6_pktopts. beware that there are non-zero default values in 2087 * the struct. 2088 */ 2089 void 2090 ip6_initpktopts(struct ip6_pktopts *opt) 2091 { 2092 2093 memset(opt, 0, sizeof(*opt)); 2094 opt->ip6po_hlim = -1; /* -1 means default hop limit */ 2095 opt->ip6po_tclass = -1; /* -1 means default traffic class */ 2096 opt->ip6po_minmtu = IP6PO_MINMTU_MCASTONLY; 2097 opt->ip6po_prefer_tempaddr = IP6PO_TEMPADDR_SYSTEM; 2098 } 2099 2100 #define sin6tosa(sin6) ((struct sockaddr *)(sin6)) /* XXX */ 2101 static int 2102 ip6_pcbopt(int optname, u_char *buf, int len, struct ip6_pktopts **pktopt, 2103 kauth_cred_t cred, int uproto) 2104 { 2105 struct ip6_pktopts *opt; 2106 2107 if (*pktopt == NULL) { 2108 *pktopt = malloc(sizeof(struct ip6_pktopts), M_IP6OPT, 2109 M_NOWAIT); 2110 if (*pktopt == NULL) 2111 return (ENOBUFS); 2112 2113 ip6_initpktopts(*pktopt); 2114 } 2115 opt = *pktopt; 2116 2117 return (ip6_setpktopt(optname, buf, len, opt, cred, 1, 0, uproto)); 2118 } 2119 2120 static int 2121 ip6_getpcbopt(struct ip6_pktopts *pktopt, int optname, struct sockopt *sopt) 2122 { 2123 void *optdata = NULL; 2124 int optdatalen = 0; 2125 struct ip6_ext *ip6e; 2126 int error = 0; 2127 struct in6_pktinfo null_pktinfo; 2128 int deftclass = 0, on; 2129 int defminmtu = IP6PO_MINMTU_MCASTONLY; 2130 int defpreftemp = IP6PO_TEMPADDR_SYSTEM; 2131 2132 switch (optname) { 2133 case IPV6_PKTINFO: 2134 if (pktopt && pktopt->ip6po_pktinfo) 2135 optdata = (void *)pktopt->ip6po_pktinfo; 2136 else { 2137 /* XXX: we don't have to do this every time... */ 2138 memset(&null_pktinfo, 0, sizeof(null_pktinfo)); 2139 optdata = (void *)&null_pktinfo; 2140 } 2141 optdatalen = sizeof(struct in6_pktinfo); 2142 break; 2143 case IPV6_OTCLASS: 2144 /* XXX */ 2145 return (EINVAL); 2146 case IPV6_TCLASS: 2147 if (pktopt && pktopt->ip6po_tclass >= 0) 2148 optdata = (void *)&pktopt->ip6po_tclass; 2149 else 2150 optdata = (void *)&deftclass; 2151 optdatalen = sizeof(int); 2152 break; 2153 case IPV6_HOPOPTS: 2154 if (pktopt && pktopt->ip6po_hbh) { 2155 optdata = (void *)pktopt->ip6po_hbh; 2156 ip6e = (struct ip6_ext *)pktopt->ip6po_hbh; 2157 optdatalen = (ip6e->ip6e_len + 1) << 3; 2158 } 2159 break; 2160 case IPV6_RTHDR: 2161 if (pktopt && pktopt->ip6po_rthdr) { 2162 optdata = (void *)pktopt->ip6po_rthdr; 2163 ip6e = (struct ip6_ext *)pktopt->ip6po_rthdr; 2164 optdatalen = (ip6e->ip6e_len + 1) << 3; 2165 } 2166 break; 2167 case IPV6_RTHDRDSTOPTS: 2168 if (pktopt && pktopt->ip6po_dest1) { 2169 optdata = (void *)pktopt->ip6po_dest1; 2170 ip6e = (struct ip6_ext *)pktopt->ip6po_dest1; 2171 optdatalen = (ip6e->ip6e_len + 1) << 3; 2172 } 2173 break; 2174 case IPV6_DSTOPTS: 2175 if (pktopt && pktopt->ip6po_dest2) { 2176 optdata = (void *)pktopt->ip6po_dest2; 2177 ip6e = (struct ip6_ext *)pktopt->ip6po_dest2; 2178 optdatalen = (ip6e->ip6e_len + 1) << 3; 2179 } 2180 break; 2181 case IPV6_NEXTHOP: 2182 if (pktopt && pktopt->ip6po_nexthop) { 2183 optdata = (void *)pktopt->ip6po_nexthop; 2184 optdatalen = pktopt->ip6po_nexthop->sa_len; 2185 } 2186 break; 2187 case IPV6_USE_MIN_MTU: 2188 if (pktopt) 2189 optdata = (void *)&pktopt->ip6po_minmtu; 2190 else 2191 optdata = (void *)&defminmtu; 2192 optdatalen = sizeof(int); 2193 break; 2194 case IPV6_DONTFRAG: 2195 if (pktopt && ((pktopt->ip6po_flags) & IP6PO_DONTFRAG)) 2196 on = 1; 2197 else 2198 on = 0; 2199 optdata = (void *)&on; 2200 optdatalen = sizeof(on); 2201 break; 2202 case IPV6_PREFER_TEMPADDR: 2203 if (pktopt) 2204 optdata = (void *)&pktopt->ip6po_prefer_tempaddr; 2205 else 2206 optdata = (void *)&defpreftemp; 2207 optdatalen = sizeof(int); 2208 break; 2209 default: /* should not happen */ 2210 #ifdef DIAGNOSTIC 2211 panic("ip6_getpcbopt: unexpected option\n"); 2212 #endif 2213 return (ENOPROTOOPT); 2214 } 2215 2216 error = sockopt_set(sopt, optdata, optdatalen); 2217 2218 return (error); 2219 } 2220 2221 void 2222 ip6_clearpktopts(struct ip6_pktopts *pktopt, int optname) 2223 { 2224 if (optname == -1 || optname == IPV6_PKTINFO) { 2225 if (pktopt->ip6po_pktinfo) 2226 free(pktopt->ip6po_pktinfo, M_IP6OPT); 2227 pktopt->ip6po_pktinfo = NULL; 2228 } 2229 if (optname == -1 || optname == IPV6_HOPLIMIT) 2230 pktopt->ip6po_hlim = -1; 2231 if (optname == -1 || optname == IPV6_TCLASS) 2232 pktopt->ip6po_tclass = -1; 2233 if (optname == -1 || optname == IPV6_NEXTHOP) { 2234 rtcache_free(&pktopt->ip6po_nextroute); 2235 if (pktopt->ip6po_nexthop) 2236 free(pktopt->ip6po_nexthop, M_IP6OPT); 2237 pktopt->ip6po_nexthop = NULL; 2238 } 2239 if (optname == -1 || optname == IPV6_HOPOPTS) { 2240 if (pktopt->ip6po_hbh) 2241 free(pktopt->ip6po_hbh, M_IP6OPT); 2242 pktopt->ip6po_hbh = NULL; 2243 } 2244 if (optname == -1 || optname == IPV6_RTHDRDSTOPTS) { 2245 if (pktopt->ip6po_dest1) 2246 free(pktopt->ip6po_dest1, M_IP6OPT); 2247 pktopt->ip6po_dest1 = NULL; 2248 } 2249 if (optname == -1 || optname == IPV6_RTHDR) { 2250 if (pktopt->ip6po_rhinfo.ip6po_rhi_rthdr) 2251 free(pktopt->ip6po_rhinfo.ip6po_rhi_rthdr, M_IP6OPT); 2252 pktopt->ip6po_rhinfo.ip6po_rhi_rthdr = NULL; 2253 rtcache_free(&pktopt->ip6po_route); 2254 } 2255 if (optname == -1 || optname == IPV6_DSTOPTS) { 2256 if (pktopt->ip6po_dest2) 2257 free(pktopt->ip6po_dest2, M_IP6OPT); 2258 pktopt->ip6po_dest2 = NULL; 2259 } 2260 } 2261 2262 #define PKTOPT_EXTHDRCPY(type) \ 2263 do { \ 2264 if (src->type) { \ 2265 int hlen = (((struct ip6_ext *)src->type)->ip6e_len + 1) << 3;\ 2266 dst->type = malloc(hlen, M_IP6OPT, canwait); \ 2267 if (dst->type == NULL) \ 2268 goto bad; \ 2269 memcpy(dst->type, src->type, hlen); \ 2270 } \ 2271 } while (/*CONSTCOND*/ 0) 2272 2273 static int 2274 copypktopts(struct ip6_pktopts *dst, struct ip6_pktopts *src, int canwait) 2275 { 2276 dst->ip6po_hlim = src->ip6po_hlim; 2277 dst->ip6po_tclass = src->ip6po_tclass; 2278 dst->ip6po_flags = src->ip6po_flags; 2279 dst->ip6po_minmtu = src->ip6po_minmtu; 2280 dst->ip6po_prefer_tempaddr = src->ip6po_prefer_tempaddr; 2281 if (src->ip6po_pktinfo) { 2282 dst->ip6po_pktinfo = malloc(sizeof(*dst->ip6po_pktinfo), 2283 M_IP6OPT, canwait); 2284 if (dst->ip6po_pktinfo == NULL) 2285 goto bad; 2286 *dst->ip6po_pktinfo = *src->ip6po_pktinfo; 2287 } 2288 if (src->ip6po_nexthop) { 2289 dst->ip6po_nexthop = malloc(src->ip6po_nexthop->sa_len, 2290 M_IP6OPT, canwait); 2291 if (dst->ip6po_nexthop == NULL) 2292 goto bad; 2293 memcpy(dst->ip6po_nexthop, src->ip6po_nexthop, 2294 src->ip6po_nexthop->sa_len); 2295 } 2296 PKTOPT_EXTHDRCPY(ip6po_hbh); 2297 PKTOPT_EXTHDRCPY(ip6po_dest1); 2298 PKTOPT_EXTHDRCPY(ip6po_dest2); 2299 PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */ 2300 return (0); 2301 2302 bad: 2303 if (dst->ip6po_pktinfo) free(dst->ip6po_pktinfo, M_IP6OPT); 2304 if (dst->ip6po_nexthop) free(dst->ip6po_nexthop, M_IP6OPT); 2305 if (dst->ip6po_hbh) free(dst->ip6po_hbh, M_IP6OPT); 2306 if (dst->ip6po_dest1) free(dst->ip6po_dest1, M_IP6OPT); 2307 if (dst->ip6po_dest2) free(dst->ip6po_dest2, M_IP6OPT); 2308 if (dst->ip6po_rthdr) free(dst->ip6po_rthdr, M_IP6OPT); 2309 2310 return (ENOBUFS); 2311 } 2312 #undef PKTOPT_EXTHDRCPY 2313 2314 struct ip6_pktopts * 2315 ip6_copypktopts(struct ip6_pktopts *src, int canwait) 2316 { 2317 int error; 2318 struct ip6_pktopts *dst; 2319 2320 dst = malloc(sizeof(*dst), M_IP6OPT, canwait); 2321 if (dst == NULL) 2322 return (NULL); 2323 ip6_initpktopts(dst); 2324 2325 if ((error = copypktopts(dst, src, canwait)) != 0) { 2326 free(dst, M_IP6OPT); 2327 return (NULL); 2328 } 2329 2330 return (dst); 2331 } 2332 2333 void 2334 ip6_freepcbopts(struct ip6_pktopts *pktopt) 2335 { 2336 if (pktopt == NULL) 2337 return; 2338 2339 ip6_clearpktopts(pktopt, -1); 2340 2341 free(pktopt, M_IP6OPT); 2342 } 2343 2344 int 2345 ip6_get_membership(const struct sockopt *sopt, struct ifnet **ifp, 2346 struct psref *psref, void *v, size_t l) 2347 { 2348 struct ipv6_mreq mreq; 2349 int error; 2350 struct in6_addr *ia = &mreq.ipv6mr_multiaddr; 2351 struct in_addr *ia4 = (void *)&ia->s6_addr32[3]; 2352 2353 error = sockopt_get(sopt, &mreq, sizeof(mreq)); 2354 if (error != 0) 2355 return error; 2356 2357 if (IN6_IS_ADDR_UNSPECIFIED(ia)) { 2358 /* 2359 * We use the unspecified address to specify to accept 2360 * all multicast addresses. Only super user is allowed 2361 * to do this. 2362 */ 2363 if (kauth_authorize_network(kauth_cred_get(), 2364 KAUTH_NETWORK_IPV6, 2365 KAUTH_REQ_NETWORK_IPV6_JOIN_MULTICAST, NULL, NULL, NULL)) 2366 return EACCES; 2367 } else if (IN6_IS_ADDR_V4MAPPED(ia)) { 2368 // Don't bother if we are not going to use ifp. 2369 if (l == sizeof(*ia)) { 2370 memcpy(v, ia, l); 2371 return 0; 2372 } 2373 } else if (!IN6_IS_ADDR_MULTICAST(ia)) { 2374 return EINVAL; 2375 } 2376 2377 /* 2378 * If no interface was explicitly specified, choose an 2379 * appropriate one according to the given multicast address. 2380 */ 2381 if (mreq.ipv6mr_interface == 0) { 2382 struct rtentry *rt; 2383 union { 2384 struct sockaddr dst; 2385 struct sockaddr_in dst4; 2386 struct sockaddr_in6 dst6; 2387 } u; 2388 struct route ro; 2389 2390 /* 2391 * Look up the routing table for the 2392 * address, and choose the outgoing interface. 2393 * XXX: is it a good approach? 2394 */ 2395 memset(&ro, 0, sizeof(ro)); 2396 if (IN6_IS_ADDR_V4MAPPED(ia)) 2397 sockaddr_in_init(&u.dst4, ia4, 0); 2398 else 2399 sockaddr_in6_init(&u.dst6, ia, 0, 0, 0); 2400 error = rtcache_setdst(&ro, &u.dst); 2401 if (error != 0) 2402 return error; 2403 rt = rtcache_init(&ro); 2404 *ifp = rt != NULL ? 2405 if_get_byindex(rt->rt_ifp->if_index, psref) : NULL; 2406 rtcache_unref(rt, &ro); 2407 rtcache_free(&ro); 2408 } else { 2409 /* 2410 * If the interface is specified, validate it. 2411 */ 2412 *ifp = if_get_byindex(mreq.ipv6mr_interface, psref); 2413 if (*ifp == NULL) 2414 return ENXIO; /* XXX EINVAL? */ 2415 } 2416 if (sizeof(*ia) == l) 2417 memcpy(v, ia, l); 2418 else 2419 memcpy(v, ia4, l); 2420 return 0; 2421 } 2422 2423 /* 2424 * Set the IP6 multicast options in response to user setsockopt(). 2425 */ 2426 static int 2427 ip6_setmoptions(const struct sockopt *sopt, struct inpcb *inp) 2428 { 2429 int error = 0; 2430 u_int loop, ifindex; 2431 struct ipv6_mreq mreq; 2432 struct in6_addr ia; 2433 struct ifnet *ifp; 2434 struct ip6_moptions *im6o = in6p_moptions(inp); 2435 struct in6_multi_mship *imm; 2436 2437 KASSERT(inp_locked(inp)); 2438 2439 if (im6o == NULL) { 2440 /* 2441 * No multicast option buffer attached to the pcb; 2442 * allocate one and initialize to default values. 2443 */ 2444 im6o = malloc(sizeof(*im6o), M_IPMOPTS, M_NOWAIT); 2445 if (im6o == NULL) 2446 return (ENOBUFS); 2447 in6p_moptions(inp) = im6o; 2448 im6o->im6o_multicast_if_index = 0; 2449 im6o->im6o_multicast_hlim = ip6_defmcasthlim; 2450 im6o->im6o_multicast_loop = IPV6_DEFAULT_MULTICAST_LOOP; 2451 LIST_INIT(&im6o->im6o_memberships); 2452 } 2453 2454 switch (sopt->sopt_name) { 2455 2456 case IPV6_MULTICAST_IF: { 2457 int s; 2458 /* 2459 * Select the interface for outgoing multicast packets. 2460 */ 2461 error = sockopt_get(sopt, &ifindex, sizeof(ifindex)); 2462 if (error != 0) 2463 break; 2464 2465 s = pserialize_read_enter(); 2466 if (ifindex != 0) { 2467 if ((ifp = if_byindex(ifindex)) == NULL) { 2468 pserialize_read_exit(s); 2469 error = ENXIO; /* XXX EINVAL? */ 2470 break; 2471 } 2472 if ((ifp->if_flags & IFF_MULTICAST) == 0) { 2473 pserialize_read_exit(s); 2474 error = EADDRNOTAVAIL; 2475 break; 2476 } 2477 } else 2478 ifp = NULL; 2479 im6o->im6o_multicast_if_index = if_get_index(ifp); 2480 pserialize_read_exit(s); 2481 break; 2482 } 2483 2484 case IPV6_MULTICAST_HOPS: 2485 { 2486 /* 2487 * Set the IP6 hoplimit for outgoing multicast packets. 2488 */ 2489 int optval; 2490 2491 error = sockopt_getint(sopt, &optval); 2492 if (error != 0) 2493 break; 2494 2495 if (optval < -1 || optval >= 256) 2496 error = EINVAL; 2497 else if (optval == -1) 2498 im6o->im6o_multicast_hlim = ip6_defmcasthlim; 2499 else 2500 im6o->im6o_multicast_hlim = optval; 2501 break; 2502 } 2503 2504 case IPV6_MULTICAST_LOOP: 2505 /* 2506 * Set the loopback flag for outgoing multicast packets. 2507 * Must be zero or one. 2508 */ 2509 error = sockopt_get(sopt, &loop, sizeof(loop)); 2510 if (error != 0) 2511 break; 2512 if (loop > 1) { 2513 error = EINVAL; 2514 break; 2515 } 2516 im6o->im6o_multicast_loop = loop; 2517 break; 2518 2519 case IPV6_JOIN_GROUP: { 2520 int bound; 2521 struct psref psref; 2522 /* 2523 * Add a multicast group membership. 2524 * Group must be a valid IP6 multicast address. 2525 */ 2526 bound = curlwp_bind(); 2527 ifp = NULL; 2528 error = ip6_get_membership(sopt, &ifp, &psref, &ia, sizeof(ia)); 2529 if (error != 0) { 2530 KASSERT(ifp == NULL); 2531 curlwp_bindx(bound); 2532 return error; 2533 } 2534 2535 if (IN6_IS_ADDR_V4MAPPED(&ia)) { 2536 error = ip_setmoptions(&inp->inp_moptions, sopt); 2537 goto put_break; 2538 } 2539 /* 2540 * See if we found an interface, and confirm that it 2541 * supports multicast 2542 */ 2543 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) { 2544 error = EADDRNOTAVAIL; 2545 goto put_break; 2546 } 2547 2548 if (in6_setscope(&ia, ifp, NULL)) { 2549 error = EADDRNOTAVAIL; /* XXX: should not happen */ 2550 goto put_break; 2551 } 2552 2553 /* 2554 * See if the membership already exists. 2555 */ 2556 LIST_FOREACH(imm, &im6o->im6o_memberships, i6mm_chain) { 2557 if (imm->i6mm_maddr->in6m_ifp == ifp && 2558 IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr, 2559 &ia)) 2560 goto put_break; 2561 } 2562 if (imm != NULL) { 2563 error = EADDRINUSE; 2564 goto put_break; 2565 } 2566 /* 2567 * Everything looks good; add a new record to the multicast 2568 * address list for the given interface. 2569 */ 2570 imm = in6_joingroup(ifp, &ia, &error, 0); 2571 if (imm == NULL) 2572 goto put_break; 2573 LIST_INSERT_HEAD(&im6o->im6o_memberships, imm, i6mm_chain); 2574 put_break: 2575 if_put(ifp, &psref); 2576 curlwp_bindx(bound); 2577 break; 2578 } 2579 2580 case IPV6_LEAVE_GROUP: { 2581 /* 2582 * Drop a multicast group membership. 2583 * Group must be a valid IP6 multicast address. 2584 */ 2585 error = sockopt_get(sopt, &mreq, sizeof(mreq)); 2586 if (error != 0) 2587 break; 2588 2589 if (IN6_IS_ADDR_V4MAPPED(&mreq.ipv6mr_multiaddr)) { 2590 error = ip_setmoptions(&inp->inp_moptions, sopt); 2591 break; 2592 } 2593 /* 2594 * If an interface address was specified, get a pointer 2595 * to its ifnet structure. 2596 */ 2597 if (mreq.ipv6mr_interface != 0) { 2598 if ((ifp = if_byindex(mreq.ipv6mr_interface)) == NULL) { 2599 error = ENXIO; /* XXX EINVAL? */ 2600 break; 2601 } 2602 } else 2603 ifp = NULL; 2604 2605 /* Fill in the scope zone ID */ 2606 if (ifp) { 2607 if (in6_setscope(&mreq.ipv6mr_multiaddr, ifp, NULL)) { 2608 /* XXX: should not happen */ 2609 error = EADDRNOTAVAIL; 2610 break; 2611 } 2612 } else if (mreq.ipv6mr_interface != 0) { 2613 /* 2614 * XXX: This case would happens when the (positive) 2615 * index is in the valid range, but the corresponding 2616 * interface has been detached dynamically. The above 2617 * check probably avoids such case to happen here, but 2618 * we check it explicitly for safety. 2619 */ 2620 error = EADDRNOTAVAIL; 2621 break; 2622 } else { /* ipv6mr_interface == 0 */ 2623 struct sockaddr_in6 sa6_mc; 2624 2625 /* 2626 * The API spec says as follows: 2627 * If the interface index is specified as 0, the 2628 * system may choose a multicast group membership to 2629 * drop by matching the multicast address only. 2630 * On the other hand, we cannot disambiguate the scope 2631 * zone unless an interface is provided. Thus, we 2632 * check if there's ambiguity with the default scope 2633 * zone as the last resort. 2634 */ 2635 sockaddr_in6_init(&sa6_mc, &mreq.ipv6mr_multiaddr, 2636 0, 0, 0); 2637 error = sa6_embedscope(&sa6_mc, ip6_use_defzone); 2638 if (error != 0) 2639 break; 2640 mreq.ipv6mr_multiaddr = sa6_mc.sin6_addr; 2641 } 2642 2643 /* 2644 * Find the membership in the membership list. 2645 */ 2646 LIST_FOREACH(imm, &im6o->im6o_memberships, i6mm_chain) { 2647 if ((ifp == NULL || imm->i6mm_maddr->in6m_ifp == ifp) && 2648 IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr, 2649 &mreq.ipv6mr_multiaddr)) 2650 break; 2651 } 2652 if (imm == NULL) { 2653 /* Unable to resolve interface */ 2654 error = EADDRNOTAVAIL; 2655 break; 2656 } 2657 /* 2658 * Give up the multicast address record to which the 2659 * membership points. 2660 */ 2661 LIST_REMOVE(imm, i6mm_chain); 2662 in6_leavegroup(imm); 2663 /* in6m_ifp should not leave thanks to inp_lock */ 2664 break; 2665 } 2666 2667 default: 2668 error = EOPNOTSUPP; 2669 break; 2670 } 2671 2672 /* 2673 * If all options have default values, no need to keep the mbuf. 2674 */ 2675 if (im6o->im6o_multicast_if_index == 0 && 2676 im6o->im6o_multicast_hlim == ip6_defmcasthlim && 2677 im6o->im6o_multicast_loop == IPV6_DEFAULT_MULTICAST_LOOP && 2678 LIST_EMPTY(&im6o->im6o_memberships)) { 2679 free(in6p_moptions(inp), M_IPMOPTS); 2680 in6p_moptions(inp) = NULL; 2681 } 2682 2683 return (error); 2684 } 2685 2686 /* 2687 * Return the IP6 multicast options in response to user getsockopt(). 2688 */ 2689 static int 2690 ip6_getmoptions(struct sockopt *sopt, struct inpcb *inp) 2691 { 2692 u_int optval; 2693 int error; 2694 struct ip6_moptions *im6o = in6p_moptions(inp); 2695 2696 switch (sopt->sopt_name) { 2697 case IPV6_MULTICAST_IF: 2698 if (im6o == NULL || im6o->im6o_multicast_if_index == 0) 2699 optval = 0; 2700 else 2701 optval = im6o->im6o_multicast_if_index; 2702 2703 error = sockopt_set(sopt, &optval, sizeof(optval)); 2704 break; 2705 2706 case IPV6_MULTICAST_HOPS: 2707 if (im6o == NULL) 2708 optval = ip6_defmcasthlim; 2709 else 2710 optval = im6o->im6o_multicast_hlim; 2711 2712 error = sockopt_set(sopt, &optval, sizeof(optval)); 2713 break; 2714 2715 case IPV6_MULTICAST_LOOP: 2716 if (im6o == NULL) 2717 optval = IPV6_DEFAULT_MULTICAST_LOOP; 2718 else 2719 optval = im6o->im6o_multicast_loop; 2720 2721 error = sockopt_set(sopt, &optval, sizeof(optval)); 2722 break; 2723 2724 default: 2725 error = EOPNOTSUPP; 2726 } 2727 2728 return (error); 2729 } 2730 2731 /* 2732 * Discard the IP6 multicast options. 2733 */ 2734 void 2735 ip6_freemoptions(struct ip6_moptions *im6o) 2736 { 2737 struct in6_multi_mship *imm, *nimm; 2738 2739 if (im6o == NULL) 2740 return; 2741 2742 /* The owner of im6o (inp) should be protected by solock */ 2743 LIST_FOREACH_SAFE(imm, &im6o->im6o_memberships, i6mm_chain, nimm) { 2744 LIST_REMOVE(imm, i6mm_chain); 2745 in6_leavegroup(imm); 2746 } 2747 free(im6o, M_IPMOPTS); 2748 } 2749 2750 /* 2751 * Set IPv6 outgoing packet options based on advanced API. 2752 */ 2753 int 2754 ip6_setpktopts(struct mbuf *control, struct ip6_pktopts *opt, 2755 struct ip6_pktopts *stickyopt, kauth_cred_t cred, int uproto) 2756 { 2757 struct cmsghdr *cm = 0; 2758 2759 if (control == NULL || opt == NULL) 2760 return (EINVAL); 2761 2762 ip6_initpktopts(opt); 2763 if (stickyopt) { 2764 int error; 2765 2766 /* 2767 * If stickyopt is provided, make a local copy of the options 2768 * for this particular packet, then override them by ancillary 2769 * objects. 2770 * XXX: copypktopts() does not copy the cached route to a next 2771 * hop (if any). This is not very good in terms of efficiency, 2772 * but we can allow this since this option should be rarely 2773 * used. 2774 */ 2775 if ((error = copypktopts(opt, stickyopt, M_NOWAIT)) != 0) 2776 return (error); 2777 } 2778 2779 /* 2780 * XXX: Currently, we assume all the optional information is stored 2781 * in a single mbuf. 2782 */ 2783 if (control->m_next) 2784 return (EINVAL); 2785 2786 /* XXX if cm->cmsg_len is not aligned, control->m_len can become <0 */ 2787 for (; control->m_len > 0; control->m_data += CMSG_ALIGN(cm->cmsg_len), 2788 control->m_len -= CMSG_ALIGN(cm->cmsg_len)) { 2789 int error; 2790 2791 if (control->m_len < CMSG_LEN(0)) 2792 return (EINVAL); 2793 2794 cm = mtod(control, struct cmsghdr *); 2795 if (cm->cmsg_len < CMSG_LEN(0) || cm->cmsg_len > control->m_len) 2796 return (EINVAL); 2797 if (cm->cmsg_level != IPPROTO_IPV6) 2798 continue; 2799 2800 error = ip6_setpktopt(cm->cmsg_type, CMSG_DATA(cm), 2801 cm->cmsg_len - CMSG_LEN(0), opt, cred, 0, 1, uproto); 2802 if (error) 2803 return (error); 2804 } 2805 2806 return (0); 2807 } 2808 2809 /* 2810 * Set a particular packet option, as a sticky option or an ancillary data 2811 * item. "len" can be 0 only when it's a sticky option. 2812 * We have 4 cases of combination of "sticky" and "cmsg": 2813 * "sticky=0, cmsg=0": impossible 2814 * "sticky=0, cmsg=1": RFC2292 or RFC3542 ancillary data 2815 * "sticky=1, cmsg=0": RFC3542 socket option 2816 * "sticky=1, cmsg=1": RFC2292 socket option 2817 */ 2818 static int 2819 ip6_setpktopt(int optname, u_char *buf, int len, struct ip6_pktopts *opt, 2820 kauth_cred_t cred, int sticky, int cmsg, int uproto) 2821 { 2822 int minmtupolicy; 2823 int error; 2824 2825 if (!sticky && !cmsg) { 2826 #ifdef DIAGNOSTIC 2827 printf("ip6_setpktopt: impossible case\n"); 2828 #endif 2829 return (EINVAL); 2830 } 2831 2832 /* 2833 * IPV6_2292xxx is for backward compatibility to RFC2292, and should 2834 * not be specified in the context of RFC3542. Conversely, 2835 * RFC3542 types should not be specified in the context of RFC2292. 2836 */ 2837 if (!cmsg) { 2838 switch (optname) { 2839 case IPV6_2292PKTINFO: 2840 case IPV6_2292HOPLIMIT: 2841 case IPV6_2292NEXTHOP: 2842 case IPV6_2292HOPOPTS: 2843 case IPV6_2292DSTOPTS: 2844 case IPV6_2292RTHDR: 2845 case IPV6_2292PKTOPTIONS: 2846 return (ENOPROTOOPT); 2847 } 2848 } 2849 if (sticky && cmsg) { 2850 switch (optname) { 2851 case IPV6_PKTINFO: 2852 case IPV6_HOPLIMIT: 2853 case IPV6_NEXTHOP: 2854 case IPV6_HOPOPTS: 2855 case IPV6_DSTOPTS: 2856 case IPV6_RTHDRDSTOPTS: 2857 case IPV6_RTHDR: 2858 case IPV6_USE_MIN_MTU: 2859 case IPV6_DONTFRAG: 2860 case IPV6_OTCLASS: 2861 case IPV6_TCLASS: 2862 case IPV6_PREFER_TEMPADDR: /* XXX not an RFC3542 option */ 2863 return (ENOPROTOOPT); 2864 } 2865 } 2866 2867 switch (optname) { 2868 #ifdef RFC2292 2869 case IPV6_2292PKTINFO: 2870 #endif 2871 case IPV6_PKTINFO: 2872 { 2873 struct in6_pktinfo *pktinfo; 2874 2875 if (len != sizeof(struct in6_pktinfo)) 2876 return (EINVAL); 2877 2878 pktinfo = (struct in6_pktinfo *)buf; 2879 2880 /* 2881 * An application can clear any sticky IPV6_PKTINFO option by 2882 * doing a "regular" setsockopt with ipi6_addr being 2883 * in6addr_any and ipi6_ifindex being zero. 2884 * [RFC 3542, Section 6] 2885 */ 2886 if (optname == IPV6_PKTINFO && opt->ip6po_pktinfo && 2887 pktinfo->ipi6_ifindex == 0 && 2888 IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) { 2889 ip6_clearpktopts(opt, optname); 2890 break; 2891 } 2892 2893 if (uproto == IPPROTO_TCP && optname == IPV6_PKTINFO && 2894 sticky && !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) { 2895 return (EINVAL); 2896 } 2897 2898 /* Validate the interface index if specified. */ 2899 if (pktinfo->ipi6_ifindex) { 2900 struct ifnet *ifp; 2901 int s = pserialize_read_enter(); 2902 ifp = if_byindex(pktinfo->ipi6_ifindex); 2903 if (ifp == NULL) { 2904 pserialize_read_exit(s); 2905 return ENXIO; 2906 } 2907 pserialize_read_exit(s); 2908 } 2909 2910 /* 2911 * We store the address anyway, and let in6_selectsrc() 2912 * validate the specified address. This is because ipi6_addr 2913 * may not have enough information about its scope zone, and 2914 * we may need additional information (such as outgoing 2915 * interface or the scope zone of a destination address) to 2916 * disambiguate the scope. 2917 * XXX: the delay of the validation may confuse the 2918 * application when it is used as a sticky option. 2919 */ 2920 if (opt->ip6po_pktinfo == NULL) { 2921 opt->ip6po_pktinfo = malloc(sizeof(*pktinfo), 2922 M_IP6OPT, M_NOWAIT); 2923 if (opt->ip6po_pktinfo == NULL) 2924 return (ENOBUFS); 2925 } 2926 memcpy(opt->ip6po_pktinfo, pktinfo, sizeof(*pktinfo)); 2927 break; 2928 } 2929 2930 #ifdef RFC2292 2931 case IPV6_2292HOPLIMIT: 2932 #endif 2933 case IPV6_HOPLIMIT: 2934 { 2935 int *hlimp; 2936 2937 /* 2938 * RFC 3542 deprecated the usage of sticky IPV6_HOPLIMIT 2939 * to simplify the ordering among hoplimit options. 2940 */ 2941 if (optname == IPV6_HOPLIMIT && sticky) 2942 return (ENOPROTOOPT); 2943 2944 if (len != sizeof(int)) 2945 return (EINVAL); 2946 hlimp = (int *)buf; 2947 if (*hlimp < -1 || *hlimp > 255) 2948 return (EINVAL); 2949 2950 opt->ip6po_hlim = *hlimp; 2951 break; 2952 } 2953 2954 case IPV6_OTCLASS: 2955 if (len != sizeof(u_int8_t)) 2956 return (EINVAL); 2957 2958 opt->ip6po_tclass = *(u_int8_t *)buf; 2959 break; 2960 2961 case IPV6_TCLASS: 2962 { 2963 int tclass; 2964 2965 if (len != sizeof(int)) 2966 return (EINVAL); 2967 tclass = *(int *)buf; 2968 if (tclass < -1 || tclass > 255) 2969 return (EINVAL); 2970 2971 opt->ip6po_tclass = tclass; 2972 break; 2973 } 2974 2975 #ifdef RFC2292 2976 case IPV6_2292NEXTHOP: 2977 #endif 2978 case IPV6_NEXTHOP: 2979 error = kauth_authorize_network(cred, 2980 KAUTH_NETWORK_IPV6, 2981 KAUTH_REQ_NETWORK_IPV6_HOPBYHOP, NULL, NULL, NULL); 2982 if (error) 2983 return (error); 2984 2985 if (len == 0) { /* just remove the option */ 2986 ip6_clearpktopts(opt, IPV6_NEXTHOP); 2987 break; 2988 } 2989 2990 /* check if cmsg_len is large enough for sa_len */ 2991 if (len < sizeof(struct sockaddr) || len < *buf) 2992 return (EINVAL); 2993 2994 switch (((struct sockaddr *)buf)->sa_family) { 2995 case AF_INET6: 2996 { 2997 struct sockaddr_in6 *sa6 = (struct sockaddr_in6 *)buf; 2998 2999 if (sa6->sin6_len != sizeof(struct sockaddr_in6)) 3000 return (EINVAL); 3001 3002 if (IN6_IS_ADDR_UNSPECIFIED(&sa6->sin6_addr) || 3003 IN6_IS_ADDR_MULTICAST(&sa6->sin6_addr)) { 3004 return (EINVAL); 3005 } 3006 if ((error = sa6_embedscope(sa6, ip6_use_defzone)) 3007 != 0) { 3008 return (error); 3009 } 3010 break; 3011 } 3012 case AF_LINK: /* eventually be supported? */ 3013 default: 3014 return (EAFNOSUPPORT); 3015 } 3016 3017 /* turn off the previous option, then set the new option. */ 3018 ip6_clearpktopts(opt, IPV6_NEXTHOP); 3019 opt->ip6po_nexthop = malloc(*buf, M_IP6OPT, M_NOWAIT); 3020 if (opt->ip6po_nexthop == NULL) 3021 return (ENOBUFS); 3022 memcpy(opt->ip6po_nexthop, buf, *buf); 3023 break; 3024 3025 #ifdef RFC2292 3026 case IPV6_2292HOPOPTS: 3027 #endif 3028 case IPV6_HOPOPTS: 3029 { 3030 struct ip6_hbh *hbh; 3031 int hbhlen; 3032 3033 /* 3034 * XXX: We don't allow a non-privileged user to set ANY HbH 3035 * options, since per-option restriction has too much 3036 * overhead. 3037 */ 3038 error = kauth_authorize_network(cred, 3039 KAUTH_NETWORK_IPV6, 3040 KAUTH_REQ_NETWORK_IPV6_HOPBYHOP, NULL, NULL, NULL); 3041 if (error) 3042 return (error); 3043 3044 if (len == 0) { 3045 ip6_clearpktopts(opt, IPV6_HOPOPTS); 3046 break; /* just remove the option */ 3047 } 3048 3049 /* message length validation */ 3050 if (len < sizeof(struct ip6_hbh)) 3051 return (EINVAL); 3052 hbh = (struct ip6_hbh *)buf; 3053 hbhlen = (hbh->ip6h_len + 1) << 3; 3054 if (len != hbhlen) 3055 return (EINVAL); 3056 3057 /* turn off the previous option, then set the new option. */ 3058 ip6_clearpktopts(opt, IPV6_HOPOPTS); 3059 opt->ip6po_hbh = malloc(hbhlen, M_IP6OPT, M_NOWAIT); 3060 if (opt->ip6po_hbh == NULL) 3061 return (ENOBUFS); 3062 memcpy(opt->ip6po_hbh, hbh, hbhlen); 3063 3064 break; 3065 } 3066 3067 #ifdef RFC2292 3068 case IPV6_2292DSTOPTS: 3069 #endif 3070 case IPV6_DSTOPTS: 3071 case IPV6_RTHDRDSTOPTS: 3072 { 3073 struct ip6_dest *dest, **newdest = NULL; 3074 int destlen; 3075 3076 /* XXX: see the comment for IPV6_HOPOPTS */ 3077 error = kauth_authorize_network(cred, 3078 KAUTH_NETWORK_IPV6, 3079 KAUTH_REQ_NETWORK_IPV6_HOPBYHOP, NULL, NULL, NULL); 3080 if (error) 3081 return (error); 3082 3083 if (len == 0) { 3084 ip6_clearpktopts(opt, optname); 3085 break; /* just remove the option */ 3086 } 3087 3088 /* message length validation */ 3089 if (len < sizeof(struct ip6_dest)) 3090 return (EINVAL); 3091 dest = (struct ip6_dest *)buf; 3092 destlen = (dest->ip6d_len + 1) << 3; 3093 if (len != destlen) 3094 return (EINVAL); 3095 /* 3096 * Determine the position that the destination options header 3097 * should be inserted; before or after the routing header. 3098 */ 3099 switch (optname) { 3100 case IPV6_2292DSTOPTS: 3101 /* 3102 * The old advanced API is ambiguous on this point. 3103 * Our approach is to determine the position based 3104 * according to the existence of a routing header. 3105 * Note, however, that this depends on the order of the 3106 * extension headers in the ancillary data; the 1st 3107 * part of the destination options header must appear 3108 * before the routing header in the ancillary data, 3109 * too. 3110 * RFC3542 solved the ambiguity by introducing 3111 * separate ancillary data or option types. 3112 */ 3113 if (opt->ip6po_rthdr == NULL) 3114 newdest = &opt->ip6po_dest1; 3115 else 3116 newdest = &opt->ip6po_dest2; 3117 break; 3118 case IPV6_RTHDRDSTOPTS: 3119 newdest = &opt->ip6po_dest1; 3120 break; 3121 case IPV6_DSTOPTS: 3122 newdest = &opt->ip6po_dest2; 3123 break; 3124 } 3125 3126 /* turn off the previous option, then set the new option. */ 3127 ip6_clearpktopts(opt, optname); 3128 *newdest = malloc(destlen, M_IP6OPT, M_NOWAIT); 3129 if (*newdest == NULL) 3130 return (ENOBUFS); 3131 memcpy(*newdest, dest, destlen); 3132 3133 break; 3134 } 3135 3136 #ifdef RFC2292 3137 case IPV6_2292RTHDR: 3138 #endif 3139 case IPV6_RTHDR: 3140 { 3141 struct ip6_rthdr *rth; 3142 int rthlen; 3143 3144 if (len == 0) { 3145 ip6_clearpktopts(opt, IPV6_RTHDR); 3146 break; /* just remove the option */ 3147 } 3148 3149 /* message length validation */ 3150 if (len < sizeof(struct ip6_rthdr)) 3151 return (EINVAL); 3152 rth = (struct ip6_rthdr *)buf; 3153 rthlen = (rth->ip6r_len + 1) << 3; 3154 if (len != rthlen) 3155 return (EINVAL); 3156 switch (rth->ip6r_type) { 3157 case IPV6_RTHDR_TYPE_0: 3158 /* Dropped, RFC5095. */ 3159 default: 3160 return (EINVAL); /* not supported */ 3161 } 3162 /* turn off the previous option */ 3163 ip6_clearpktopts(opt, IPV6_RTHDR); 3164 opt->ip6po_rthdr = malloc(rthlen, M_IP6OPT, M_NOWAIT); 3165 if (opt->ip6po_rthdr == NULL) 3166 return (ENOBUFS); 3167 memcpy(opt->ip6po_rthdr, rth, rthlen); 3168 break; 3169 } 3170 3171 case IPV6_USE_MIN_MTU: 3172 if (len != sizeof(int)) 3173 return (EINVAL); 3174 minmtupolicy = *(int *)buf; 3175 if (minmtupolicy != IP6PO_MINMTU_MCASTONLY && 3176 minmtupolicy != IP6PO_MINMTU_DISABLE && 3177 minmtupolicy != IP6PO_MINMTU_ALL) { 3178 return (EINVAL); 3179 } 3180 opt->ip6po_minmtu = minmtupolicy; 3181 break; 3182 3183 case IPV6_DONTFRAG: 3184 if (len != sizeof(int)) 3185 return (EINVAL); 3186 3187 if (uproto == IPPROTO_TCP || *(int *)buf == 0) { 3188 /* 3189 * we ignore this option for TCP sockets. 3190 * (RFC3542 leaves this case unspecified.) 3191 */ 3192 opt->ip6po_flags &= ~IP6PO_DONTFRAG; 3193 } else 3194 opt->ip6po_flags |= IP6PO_DONTFRAG; 3195 break; 3196 3197 case IPV6_PREFER_TEMPADDR: 3198 { 3199 int preftemp; 3200 3201 if (len != sizeof(int)) 3202 return (EINVAL); 3203 preftemp = *(int *)buf; 3204 switch (preftemp) { 3205 case IP6PO_TEMPADDR_SYSTEM: 3206 case IP6PO_TEMPADDR_NOTPREFER: 3207 case IP6PO_TEMPADDR_PREFER: 3208 break; 3209 default: 3210 return (EINVAL); 3211 } 3212 opt->ip6po_prefer_tempaddr = preftemp; 3213 break; 3214 } 3215 3216 default: 3217 return (ENOPROTOOPT); 3218 } /* end of switch */ 3219 3220 return (0); 3221 } 3222 3223 /* 3224 * Routine called from ip6_output() to loop back a copy of an IP6 multicast 3225 * packet to the input queue of a specified interface. Note that this 3226 * calls the output routine of the loopback "driver", but with an interface 3227 * pointer that might NOT be lo0ifp -- easier than replicating that code here. 3228 */ 3229 void 3230 ip6_mloopback(struct ifnet *ifp, struct mbuf *m, 3231 const struct sockaddr_in6 *dst) 3232 { 3233 struct mbuf *copym; 3234 struct ip6_hdr *ip6; 3235 3236 copym = m_copypacket(m, M_DONTWAIT); 3237 if (copym == NULL) 3238 return; 3239 3240 /* 3241 * Make sure to deep-copy IPv6 header portion in case the data 3242 * is in an mbuf cluster, so that we can safely override the IPv6 3243 * header portion later. 3244 */ 3245 if ((copym->m_flags & M_EXT) != 0 || 3246 copym->m_len < sizeof(struct ip6_hdr)) { 3247 copym = m_pullup(copym, sizeof(struct ip6_hdr)); 3248 if (copym == NULL) 3249 return; 3250 } 3251 3252 #ifdef DIAGNOSTIC 3253 if (copym->m_len < sizeof(*ip6)) { 3254 m_freem(copym); 3255 return; 3256 } 3257 #endif 3258 3259 ip6 = mtod(copym, struct ip6_hdr *); 3260 /* 3261 * clear embedded scope identifiers if necessary. 3262 * in6_clearscope will touch the addresses only when necessary. 3263 */ 3264 in6_clearscope(&ip6->ip6_src); 3265 in6_clearscope(&ip6->ip6_dst); 3266 3267 (void)looutput(ifp, copym, (const struct sockaddr *)dst, NULL); 3268 } 3269 3270 /* 3271 * Chop IPv6 header off from the payload. 3272 */ 3273 static int 3274 ip6_splithdr(struct mbuf *m, struct ip6_exthdrs *exthdrs) 3275 { 3276 struct mbuf *mh; 3277 struct ip6_hdr *ip6; 3278 3279 ip6 = mtod(m, struct ip6_hdr *); 3280 if (m->m_len > sizeof(*ip6)) { 3281 MGETHDR(mh, M_DONTWAIT, MT_HEADER); 3282 if (mh == NULL) { 3283 m_freem(m); 3284 return ENOBUFS; 3285 } 3286 m_move_pkthdr(mh, m); 3287 m_align(mh, sizeof(*ip6)); 3288 m->m_len -= sizeof(*ip6); 3289 m->m_data += sizeof(*ip6); 3290 mh->m_next = m; 3291 mh->m_len = sizeof(*ip6); 3292 memcpy(mtod(mh, void *), (void *)ip6, sizeof(*ip6)); 3293 m = mh; 3294 } 3295 exthdrs->ip6e_ip6 = m; 3296 return 0; 3297 } 3298 3299 /* 3300 * Compute IPv6 extension header length. 3301 */ 3302 int 3303 ip6_optlen(struct inpcb *inp) 3304 { 3305 int len; 3306 3307 if (!in6p_outputopts(inp)) 3308 return 0; 3309 3310 len = 0; 3311 #define elen(x) \ 3312 (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0) 3313 3314 len += elen(in6p_outputopts(inp)->ip6po_hbh); 3315 len += elen(in6p_outputopts(inp)->ip6po_dest1); 3316 len += elen(in6p_outputopts(inp)->ip6po_rthdr); 3317 len += elen(in6p_outputopts(inp)->ip6po_dest2); 3318 return len; 3319 #undef elen 3320 } 3321 3322 /* 3323 * Ensure sending address is valid. 3324 * Returns 0 on success, -1 if an error should be sent back or 1 3325 * if the packet could be dropped without error (protocol dependent). 3326 */ 3327 static int 3328 ip6_ifaddrvalid(const struct in6_addr *src, const struct in6_addr *dst) 3329 { 3330 struct sockaddr_in6 sin6; 3331 int s, error; 3332 struct ifaddr *ifa; 3333 struct in6_ifaddr *ia6; 3334 3335 if (IN6_IS_ADDR_UNSPECIFIED(src)) 3336 return 0; 3337 3338 memset(&sin6, 0, sizeof(sin6)); 3339 sin6.sin6_family = AF_INET6; 3340 sin6.sin6_len = sizeof(sin6); 3341 sin6.sin6_addr = *src; 3342 3343 s = pserialize_read_enter(); 3344 ifa = ifa_ifwithaddr(sin6tosa(&sin6)); 3345 if ((ia6 = ifatoia6(ifa)) == NULL || 3346 ia6->ia6_flags & (IN6_IFF_ANYCAST | IN6_IFF_DUPLICATED)) 3347 error = -1; 3348 else if (ia6->ia6_flags & IN6_IFF_TENTATIVE) 3349 error = 1; 3350 else if (ia6->ia6_flags & IN6_IFF_DETACHED && 3351 (sin6.sin6_addr = *dst, ifa_ifwithaddr(sin6tosa(&sin6)) == NULL)) 3352 /* Allow internal traffic to DETACHED addresses */ 3353 error = 1; 3354 else 3355 error = 0; 3356 pserialize_read_exit(s); 3357 3358 return error; 3359 } 3360