ip6_output.c revision 1.225
1 /* $NetBSD: ip6_output.c,v 1.225 2020/08/28 06:32:24 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 1.225 2020/08/28 06:32:24 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 in6pcb *); 125 static int ip6_getmoptions(struct sockopt *, struct in6pcb *); 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 in6pcb *in6p, 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, in6p, 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 int s = splsoftnet(); 488 error = ipsec6_process_packet(m, sp->req, flags); 489 splx(s); 490 491 /* 492 * Preserve KAME behaviour: ENOENT can be returned 493 * when an SA acquire is in progress. Don't propagate 494 * this to user-level; it confuses applications. 495 * XXX this will go away when the SADB is redone. 496 */ 497 if (error == ENOENT) 498 error = 0; 499 500 goto done; 501 } 502 #endif 503 504 /* adjust pointer */ 505 ip6 = mtod(m, struct ip6_hdr *); 506 507 sockaddr_in6_init(&dst_sa, &ip6->ip6_dst, 0, 0, 0); 508 509 /* We do not need a route for multicast */ 510 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) { 511 struct in6_pktinfo *pi = NULL; 512 513 /* 514 * If the outgoing interface for the address is specified by 515 * the caller, use it. 516 */ 517 if (opt && (pi = opt->ip6po_pktinfo) != NULL) { 518 /* XXX boundary check is assumed to be already done. */ 519 ifp = if_get_byindex(pi->ipi6_ifindex, &psref); 520 } else if (im6o != NULL) { 521 ifp = if_get_byindex(im6o->im6o_multicast_if_index, 522 &psref); 523 } 524 } 525 526 if (ifp == NULL) { 527 error = in6_selectroute(&dst_sa, opt, &ro, &rt, true); 528 if (error != 0) 529 goto bad; 530 ifp = if_get_byindex(rt->rt_ifp->if_index, &psref); 531 } 532 533 if (rt == NULL) { 534 /* 535 * If in6_selectroute() does not return a route entry, 536 * dst may not have been updated. 537 */ 538 error = rtcache_setdst(ro, sin6tosa(&dst_sa)); 539 if (error) { 540 IP6_STATINC(IP6_STAT_ODROPPED); 541 goto bad; 542 } 543 } 544 545 /* 546 * then rt (for unicast) and ifp must be non-NULL valid values. 547 */ 548 if ((flags & IPV6_FORWARDING) == 0) { 549 /* XXX: the FORWARDING flag can be set for mrouting. */ 550 in6_ifstat_inc(ifp, ifs6_out_request); 551 } 552 if (rt != NULL) { 553 ia = (struct in6_ifaddr *)(rt->rt_ifa); 554 rt->rt_use++; 555 } 556 557 /* 558 * The outgoing interface must be in the zone of source and 559 * destination addresses. We should use ia_ifp to support the 560 * case of sending packets to an address of our own. 561 */ 562 if (ia != NULL) { 563 origifp = ia->ia_ifp; 564 if (if_is_deactivated(origifp)) { 565 IP6_STATINC(IP6_STAT_ODROPPED); 566 goto bad; 567 } 568 if_acquire(origifp, &psref_ia); 569 release_psref_ia = true; 570 } else 571 origifp = ifp; 572 573 src0 = ip6->ip6_src; 574 if (in6_setscope(&src0, origifp, &zone)) 575 goto badscope; 576 sockaddr_in6_init(&src_sa, &ip6->ip6_src, 0, 0, 0); 577 if (sa6_recoverscope(&src_sa) || zone != src_sa.sin6_scope_id) 578 goto badscope; 579 580 dst0 = ip6->ip6_dst; 581 if (in6_setscope(&dst0, origifp, &zone)) 582 goto badscope; 583 /* re-initialize to be sure */ 584 sockaddr_in6_init(&dst_sa, &ip6->ip6_dst, 0, 0, 0); 585 if (sa6_recoverscope(&dst_sa) || zone != dst_sa.sin6_scope_id) 586 goto badscope; 587 588 /* scope check is done. */ 589 590 /* Ensure we only send from a valid address. */ 591 if ((ifp->if_flags & IFF_LOOPBACK) == 0 && 592 (flags & IPV6_FORWARDING) == 0 && 593 (error = ip6_ifaddrvalid(&src0, &dst0)) != 0) 594 { 595 char ip6buf[INET6_ADDRSTRLEN]; 596 nd6log(LOG_ERR, 597 "refusing to send from invalid address %s (pid %d)\n", 598 IN6_PRINT(ip6buf, &src0), curproc->p_pid); 599 IP6_STATINC(IP6_STAT_ODROPPED); 600 in6_ifstat_inc(origifp, ifs6_out_discard); 601 if (error == 1) 602 /* 603 * Address exists, but is tentative or detached. 604 * We can't send from it because it's invalid, 605 * so we drop the packet. 606 */ 607 error = 0; 608 else 609 error = EADDRNOTAVAIL; 610 goto bad; 611 } 612 613 if (rt != NULL && (rt->rt_flags & RTF_GATEWAY) && 614 !IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) 615 dst = satocsin6(rt->rt_gateway); 616 else 617 dst = satocsin6(rtcache_getdst(ro)); 618 619 /* 620 * XXXXXX: original code follows: 621 */ 622 if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) 623 m->m_flags &= ~(M_BCAST | M_MCAST); /* just in case */ 624 else { 625 bool ingroup; 626 627 m->m_flags = (m->m_flags & ~M_BCAST) | M_MCAST; 628 629 in6_ifstat_inc(ifp, ifs6_out_mcast); 630 631 /* 632 * Confirm that the outgoing interface supports multicast. 633 */ 634 if (!(ifp->if_flags & IFF_MULTICAST)) { 635 IP6_STATINC(IP6_STAT_NOROUTE); 636 in6_ifstat_inc(ifp, ifs6_out_discard); 637 error = ENETUNREACH; 638 goto bad; 639 } 640 641 ingroup = in6_multi_group(&ip6->ip6_dst, ifp); 642 if (ingroup && (im6o == NULL || im6o->im6o_multicast_loop)) { 643 /* 644 * If we belong to the destination multicast group 645 * on the outgoing interface, and the caller did not 646 * forbid loopback, loop back a copy. 647 */ 648 KASSERT(dst != NULL); 649 ip6_mloopback(ifp, m, dst); 650 } else { 651 /* 652 * If we are acting as a multicast router, perform 653 * multicast forwarding as if the packet had just 654 * arrived on the interface to which we are about 655 * to send. The multicast forwarding function 656 * recursively calls this function, using the 657 * IPV6_FORWARDING flag to prevent infinite recursion. 658 * 659 * Multicasts that are looped back by ip6_mloopback(), 660 * above, will be forwarded by the ip6_input() routine, 661 * if necessary. 662 */ 663 if (ip6_mrouter && (flags & IPV6_FORWARDING) == 0) { 664 if (ip6_mforward(ip6, ifp, m) != 0) { 665 m_freem(m); 666 goto done; 667 } 668 } 669 } 670 /* 671 * Multicasts with a hoplimit of zero may be looped back, 672 * above, but must not be transmitted on a network. 673 * Also, multicasts addressed to the loopback interface 674 * are not sent -- the above call to ip6_mloopback() will 675 * loop back a copy if this host actually belongs to the 676 * destination group on the loopback interface. 677 */ 678 if (ip6->ip6_hlim == 0 || (ifp->if_flags & IFF_LOOPBACK) || 679 IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst)) { 680 m_freem(m); 681 goto done; 682 } 683 } 684 685 /* 686 * Fill the outgoing inteface to tell the upper layer 687 * to increment per-interface statistics. 688 */ 689 if (ifpp) 690 *ifpp = ifp; 691 692 /* Determine path MTU. */ 693 /* 694 * ro_pmtu represent final destination while 695 * ro might represent immediate destination. 696 * Use ro_pmtu destination since MTU might differ. 697 */ 698 if (ro_pmtu != ro) { 699 union { 700 struct sockaddr dst; 701 struct sockaddr_in6 dst6; 702 } u; 703 704 /* ro_pmtu may not have a cache */ 705 sockaddr_in6_init(&u.dst6, &finaldst, 0, 0, 0); 706 rt_pmtu = rtcache_lookup(ro_pmtu, &u.dst); 707 } else 708 rt_pmtu = rt; 709 error = ip6_getpmtu(rt_pmtu, ifp, &mtu, &alwaysfrag); 710 if (rt_pmtu != NULL && rt_pmtu != rt) 711 rtcache_unref(rt_pmtu, ro_pmtu); 712 KASSERT(error == 0); /* ip6_getpmtu never fail if ifp is passed */ 713 714 /* 715 * The caller of this function may specify to use the minimum MTU 716 * in some cases. 717 * An advanced API option (IPV6_USE_MIN_MTU) can also override MTU 718 * setting. The logic is a bit complicated; by default, unicast 719 * packets will follow path MTU while multicast packets will be sent at 720 * the minimum MTU. If IP6PO_MINMTU_ALL is specified, all packets 721 * including unicast ones will be sent at the minimum MTU. Multicast 722 * packets will always be sent at the minimum MTU unless 723 * IP6PO_MINMTU_DISABLE is explicitly specified. 724 * See RFC 3542 for more details. 725 */ 726 if (mtu > IPV6_MMTU) { 727 if ((flags & IPV6_MINMTU)) 728 mtu = IPV6_MMTU; 729 else if (opt && opt->ip6po_minmtu == IP6PO_MINMTU_ALL) 730 mtu = IPV6_MMTU; 731 else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) && 732 (opt == NULL || 733 opt->ip6po_minmtu != IP6PO_MINMTU_DISABLE)) { 734 mtu = IPV6_MMTU; 735 } 736 } 737 738 /* 739 * clear embedded scope identifiers if necessary. 740 * in6_clearscope will touch the addresses only when necessary. 741 */ 742 in6_clearscope(&ip6->ip6_src); 743 in6_clearscope(&ip6->ip6_dst); 744 745 /* 746 * If the outgoing packet contains a hop-by-hop options header, 747 * it must be examined and processed even by the source node. 748 * (RFC 2460, section 4.) 749 * 750 * XXX Is this really necessary? 751 */ 752 if (ip6->ip6_nxt == IPPROTO_HOPOPTS) { 753 u_int32_t dummy1; /* XXX unused */ 754 u_int32_t dummy2; /* XXX unused */ 755 int hoff = sizeof(struct ip6_hdr); 756 757 if (ip6_hopopts_input(&dummy1, &dummy2, &m, &hoff)) { 758 /* m was already freed at this point */ 759 error = EINVAL; 760 goto done; 761 } 762 763 ip6 = mtod(m, struct ip6_hdr *); 764 } 765 766 /* 767 * Run through list of hooks for output packets. 768 */ 769 error = pfil_run_hooks(inet6_pfil_hook, &m, ifp, PFIL_OUT); 770 if (error != 0 || m == NULL) { 771 IP6_STATINC(IP6_STAT_PFILDROP_OUT); 772 goto done; 773 } 774 ip6 = mtod(m, struct ip6_hdr *); 775 776 /* 777 * Send the packet to the outgoing interface. 778 * If necessary, do IPv6 fragmentation before sending. 779 * 780 * the logic here is rather complex: 781 * 1: normal case (dontfrag == 0, alwaysfrag == 0) 782 * 1-a: send as is if tlen <= path mtu 783 * 1-b: fragment if tlen > path mtu 784 * 785 * 2: if user asks us not to fragment (dontfrag == 1) 786 * 2-a: send as is if tlen <= interface mtu 787 * 2-b: error if tlen > interface mtu 788 * 789 * 3: if we always need to attach fragment header (alwaysfrag == 1) 790 * always fragment 791 * 792 * 4: if dontfrag == 1 && alwaysfrag == 1 793 * error, as we cannot handle this conflicting request 794 */ 795 tlen = m->m_pkthdr.len; 796 tso = (m->m_pkthdr.csum_flags & M_CSUM_TSOv6) != 0; 797 if (opt && (opt->ip6po_flags & IP6PO_DONTFRAG)) 798 dontfrag = 1; 799 else 800 dontfrag = 0; 801 802 if (dontfrag && alwaysfrag) { /* case 4 */ 803 /* conflicting request - can't transmit */ 804 IP6_STATINC(IP6_STAT_CANTFRAG); 805 error = EMSGSIZE; 806 goto bad; 807 } 808 if (dontfrag && (!tso && tlen > ifp->if_mtu)) { /* case 2-b */ 809 /* 810 * Even if the DONTFRAG option is specified, we cannot send the 811 * packet when the data length is larger than the MTU of the 812 * outgoing interface. 813 * Notify the error by sending IPV6_PATHMTU ancillary data as 814 * well as returning an error code (the latter is not described 815 * in the API spec.) 816 */ 817 u_int32_t mtu32; 818 struct ip6ctlparam ip6cp; 819 820 mtu32 = (u_int32_t)mtu; 821 memset(&ip6cp, 0, sizeof(ip6cp)); 822 ip6cp.ip6c_cmdarg = (void *)&mtu32; 823 pfctlinput2(PRC_MSGSIZE, 824 rtcache_getdst(ro_pmtu), &ip6cp); 825 826 IP6_STATINC(IP6_STAT_CANTFRAG); 827 error = EMSGSIZE; 828 goto bad; 829 } 830 831 /* 832 * transmit packet without fragmentation 833 */ 834 if (dontfrag || (!alwaysfrag && (tlen <= mtu || tso))) { 835 /* case 1-a and 2-a */ 836 struct in6_ifaddr *ia6; 837 int sw_csum; 838 int s; 839 840 ip6 = mtod(m, struct ip6_hdr *); 841 s = pserialize_read_enter(); 842 ia6 = in6_ifawithifp(ifp, &ip6->ip6_src); 843 if (ia6) { 844 /* Record statistics for this interface address. */ 845 ia6->ia_ifa.ifa_data.ifad_outbytes += m->m_pkthdr.len; 846 } 847 pserialize_read_exit(s); 848 849 sw_csum = m->m_pkthdr.csum_flags & ~ifp->if_csum_flags_tx; 850 if ((sw_csum & (M_CSUM_UDPv6|M_CSUM_TCPv6)) != 0) { 851 if (IN6_NEED_CHECKSUM(ifp, 852 sw_csum & (M_CSUM_UDPv6|M_CSUM_TCPv6))) { 853 in6_undefer_cksum_tcpudp(m); 854 } 855 m->m_pkthdr.csum_flags &= ~(M_CSUM_UDPv6|M_CSUM_TCPv6); 856 } 857 858 KASSERT(dst != NULL); 859 if (__predict_false(sw_csum & M_CSUM_TSOv6)) { 860 /* 861 * TSO6 is required by a packet, but disabled for 862 * the interface. 863 */ 864 error = ip6_tso_output(ifp, origifp, m, dst, rt); 865 } else 866 error = ip6_if_output(ifp, origifp, m, dst, rt); 867 goto done; 868 } 869 870 if (tso) { 871 IP6_STATINC(IP6_STAT_CANTFRAG); /* XXX */ 872 error = EINVAL; /* XXX */ 873 goto bad; 874 } 875 876 /* 877 * try to fragment the packet. case 1-b and 3 878 */ 879 if (mtu < IPV6_MMTU) { 880 /* path MTU cannot be less than IPV6_MMTU */ 881 IP6_STATINC(IP6_STAT_CANTFRAG); 882 error = EMSGSIZE; 883 in6_ifstat_inc(ifp, ifs6_out_fragfail); 884 goto bad; 885 } else if (ip6->ip6_plen == 0) { 886 /* jumbo payload cannot be fragmented */ 887 IP6_STATINC(IP6_STAT_CANTFRAG); 888 error = EMSGSIZE; 889 in6_ifstat_inc(ifp, ifs6_out_fragfail); 890 goto bad; 891 } else { 892 const u_int32_t id = htonl(ip6_randomid()); 893 struct mbuf **mnext, *m_frgpart; 894 const int hlen = unfragpartlen; 895 struct ip6_frag *ip6f; 896 u_char nextproto; 897 898 if (mtu > IPV6_MAXPACKET) 899 mtu = IPV6_MAXPACKET; 900 901 /* 902 * Must be able to put at least 8 bytes per fragment. 903 */ 904 len = (mtu - hlen - sizeof(struct ip6_frag)) & ~7; 905 if (len < 8) { 906 IP6_STATINC(IP6_STAT_CANTFRAG); 907 error = EMSGSIZE; 908 in6_ifstat_inc(ifp, ifs6_out_fragfail); 909 goto bad; 910 } 911 912 mnext = &m->m_nextpkt; 913 914 /* 915 * Change the next header field of the last header in the 916 * unfragmentable part. 917 */ 918 if (exthdrs.ip6e_rthdr) { 919 nextproto = *mtod(exthdrs.ip6e_rthdr, u_char *); 920 *mtod(exthdrs.ip6e_rthdr, u_char *) = IPPROTO_FRAGMENT; 921 } else if (exthdrs.ip6e_dest1) { 922 nextproto = *mtod(exthdrs.ip6e_dest1, u_char *); 923 *mtod(exthdrs.ip6e_dest1, u_char *) = IPPROTO_FRAGMENT; 924 } else if (exthdrs.ip6e_hbh) { 925 nextproto = *mtod(exthdrs.ip6e_hbh, u_char *); 926 *mtod(exthdrs.ip6e_hbh, u_char *) = IPPROTO_FRAGMENT; 927 } else { 928 nextproto = ip6->ip6_nxt; 929 ip6->ip6_nxt = IPPROTO_FRAGMENT; 930 } 931 932 if ((m->m_pkthdr.csum_flags & (M_CSUM_UDPv6|M_CSUM_TCPv6)) 933 != 0) { 934 if (IN6_NEED_CHECKSUM(ifp, 935 m->m_pkthdr.csum_flags & 936 (M_CSUM_UDPv6|M_CSUM_TCPv6))) { 937 in6_undefer_cksum_tcpudp(m); 938 } 939 m->m_pkthdr.csum_flags &= ~(M_CSUM_UDPv6|M_CSUM_TCPv6); 940 } 941 942 /* 943 * Loop through length of segment after first fragment, 944 * make new header and copy data of each part and link onto 945 * chain. 946 */ 947 m0 = m; 948 for (off = hlen; off < tlen; off += len) { 949 struct mbuf *mlast; 950 951 MGETHDR(m, M_DONTWAIT, MT_HEADER); 952 if (!m) { 953 error = ENOBUFS; 954 IP6_STATINC(IP6_STAT_ODROPPED); 955 goto sendorfree; 956 } 957 m_reset_rcvif(m); 958 m->m_flags = m0->m_flags & M_COPYFLAGS; 959 *mnext = m; 960 mnext = &m->m_nextpkt; 961 m->m_data += max_linkhdr; 962 mhip6 = mtod(m, struct ip6_hdr *); 963 *mhip6 = *ip6; 964 m->m_len = sizeof(*mhip6); 965 966 ip6f = NULL; 967 error = ip6_insertfraghdr(m0, m, hlen, &ip6f); 968 if (error) { 969 IP6_STATINC(IP6_STAT_ODROPPED); 970 goto sendorfree; 971 } 972 973 /* Fill in the Frag6 Header */ 974 ip6f->ip6f_offlg = htons((u_int16_t)((off - hlen) & ~7)); 975 if (off + len >= tlen) 976 len = tlen - off; 977 else 978 ip6f->ip6f_offlg |= IP6F_MORE_FRAG; 979 ip6f->ip6f_reserved = 0; 980 ip6f->ip6f_ident = id; 981 ip6f->ip6f_nxt = nextproto; 982 983 mhip6->ip6_plen = htons((u_int16_t)(len + hlen + 984 sizeof(*ip6f) - sizeof(struct ip6_hdr))); 985 if ((m_frgpart = m_copym(m0, off, len, M_DONTWAIT)) == NULL) { 986 error = ENOBUFS; 987 IP6_STATINC(IP6_STAT_ODROPPED); 988 goto sendorfree; 989 } 990 for (mlast = m; mlast->m_next; mlast = mlast->m_next) 991 ; 992 mlast->m_next = m_frgpart; 993 994 m->m_pkthdr.len = len + hlen + sizeof(*ip6f); 995 m_reset_rcvif(m); 996 IP6_STATINC(IP6_STAT_OFRAGMENTS); 997 in6_ifstat_inc(ifp, ifs6_out_fragcreat); 998 } 999 1000 in6_ifstat_inc(ifp, ifs6_out_fragok); 1001 } 1002 1003 sendorfree: 1004 m = m0->m_nextpkt; 1005 m0->m_nextpkt = 0; 1006 m_freem(m0); 1007 for (m0 = m; m; m = m0) { 1008 m0 = m->m_nextpkt; 1009 m->m_nextpkt = 0; 1010 if (error == 0) { 1011 struct in6_ifaddr *ia6; 1012 int s; 1013 ip6 = mtod(m, struct ip6_hdr *); 1014 s = pserialize_read_enter(); 1015 ia6 = in6_ifawithifp(ifp, &ip6->ip6_src); 1016 if (ia6) { 1017 /* 1018 * Record statistics for this interface 1019 * address. 1020 */ 1021 ia6->ia_ifa.ifa_data.ifad_outbytes += 1022 m->m_pkthdr.len; 1023 } 1024 pserialize_read_exit(s); 1025 KASSERT(dst != NULL); 1026 error = ip6_if_output(ifp, origifp, m, dst, rt); 1027 } else 1028 m_freem(m); 1029 } 1030 1031 if (error == 0) 1032 IP6_STATINC(IP6_STAT_FRAGMENTED); 1033 1034 done: 1035 rtcache_unref(rt, ro); 1036 if (ro == &ip6route) 1037 rtcache_free(&ip6route); 1038 #ifdef IPSEC 1039 if (sp != NULL) 1040 KEY_SP_UNREF(&sp); 1041 #endif 1042 if_put(ifp, &psref); 1043 if (release_psref_ia) 1044 if_put(origifp, &psref_ia); 1045 curlwp_bindx(bound); 1046 1047 return error; 1048 1049 freehdrs: 1050 m_freem(exthdrs.ip6e_hbh); 1051 m_freem(exthdrs.ip6e_dest1); 1052 m_freem(exthdrs.ip6e_rthdr); 1053 m_freem(exthdrs.ip6e_dest2); 1054 /* FALLTHROUGH */ 1055 bad: 1056 m_freem(m); 1057 goto done; 1058 1059 badscope: 1060 IP6_STATINC(IP6_STAT_BADSCOPE); 1061 in6_ifstat_inc(origifp, ifs6_out_discard); 1062 if (error == 0) 1063 error = EHOSTUNREACH; /* XXX */ 1064 goto bad; 1065 } 1066 1067 static int 1068 ip6_copyexthdr(struct mbuf **mp, void *hdr, int hlen) 1069 { 1070 struct mbuf *m; 1071 1072 if (hlen > MCLBYTES) 1073 return ENOBUFS; /* XXX */ 1074 1075 MGET(m, M_DONTWAIT, MT_DATA); 1076 if (!m) 1077 return ENOBUFS; 1078 1079 if (hlen > MLEN) { 1080 MCLGET(m, M_DONTWAIT); 1081 if ((m->m_flags & M_EXT) == 0) { 1082 m_free(m); 1083 return ENOBUFS; 1084 } 1085 } 1086 m->m_len = hlen; 1087 if (hdr) 1088 memcpy(mtod(m, void *), hdr, hlen); 1089 1090 *mp = m; 1091 return 0; 1092 } 1093 1094 /* 1095 * Insert jumbo payload option. 1096 */ 1097 static int 1098 ip6_insert_jumboopt(struct ip6_exthdrs *exthdrs, u_int32_t plen) 1099 { 1100 struct mbuf *mopt; 1101 u_int8_t *optbuf; 1102 u_int32_t v; 1103 1104 #define JUMBOOPTLEN 8 /* length of jumbo payload option and padding */ 1105 1106 /* 1107 * If there is no hop-by-hop options header, allocate new one. 1108 * If there is one but it doesn't have enough space to store the 1109 * jumbo payload option, allocate a cluster to store the whole options. 1110 * Otherwise, use it to store the options. 1111 */ 1112 if (exthdrs->ip6e_hbh == NULL) { 1113 MGET(mopt, M_DONTWAIT, MT_DATA); 1114 if (mopt == 0) 1115 return (ENOBUFS); 1116 mopt->m_len = JUMBOOPTLEN; 1117 optbuf = mtod(mopt, u_int8_t *); 1118 optbuf[1] = 0; /* = ((JUMBOOPTLEN) >> 3) - 1 */ 1119 exthdrs->ip6e_hbh = mopt; 1120 } else { 1121 struct ip6_hbh *hbh; 1122 1123 mopt = exthdrs->ip6e_hbh; 1124 if (M_TRAILINGSPACE(mopt) < JUMBOOPTLEN) { 1125 const int oldoptlen = mopt->m_len; 1126 struct mbuf *n; 1127 1128 /* 1129 * Assumptions: 1130 * - exthdrs->ip6e_hbh is not referenced from places 1131 * other than exthdrs. 1132 * - exthdrs->ip6e_hbh is not an mbuf chain. 1133 */ 1134 KASSERT(mopt->m_next == NULL); 1135 1136 /* 1137 * Give up if the whole (new) hbh header does not fit 1138 * even in an mbuf cluster. 1139 */ 1140 if (oldoptlen + JUMBOOPTLEN > MCLBYTES) 1141 return ENOBUFS; 1142 1143 /* 1144 * At this point, we must always prepare a cluster. 1145 */ 1146 MGET(n, M_DONTWAIT, MT_DATA); 1147 if (n) { 1148 MCLGET(n, M_DONTWAIT); 1149 if ((n->m_flags & M_EXT) == 0) { 1150 m_freem(n); 1151 n = NULL; 1152 } 1153 } 1154 if (!n) 1155 return ENOBUFS; 1156 1157 n->m_len = oldoptlen + JUMBOOPTLEN; 1158 bcopy(mtod(mopt, void *), mtod(n, void *), 1159 oldoptlen); 1160 optbuf = mtod(n, u_int8_t *) + oldoptlen; 1161 m_freem(mopt); 1162 mopt = exthdrs->ip6e_hbh = n; 1163 } else { 1164 optbuf = mtod(mopt, u_int8_t *) + mopt->m_len; 1165 mopt->m_len += JUMBOOPTLEN; 1166 } 1167 optbuf[0] = IP6OPT_PADN; 1168 optbuf[1] = 0; 1169 1170 /* 1171 * Adjust the header length according to the pad and 1172 * the jumbo payload option. 1173 */ 1174 hbh = mtod(mopt, struct ip6_hbh *); 1175 hbh->ip6h_len += (JUMBOOPTLEN >> 3); 1176 } 1177 1178 /* fill in the option. */ 1179 optbuf[2] = IP6OPT_JUMBO; 1180 optbuf[3] = 4; 1181 v = (u_int32_t)htonl(plen + JUMBOOPTLEN); 1182 memcpy(&optbuf[4], &v, sizeof(u_int32_t)); 1183 1184 /* finally, adjust the packet header length */ 1185 exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN; 1186 1187 return 0; 1188 #undef JUMBOOPTLEN 1189 } 1190 1191 /* 1192 * Insert fragment header and copy unfragmentable header portions. 1193 * 1194 * *frghdrp will not be read, and it is guaranteed that either an 1195 * error is returned or that *frghdrp will point to space allocated 1196 * for the fragment header. 1197 * 1198 * On entry, m contains: 1199 * IPv6 Header 1200 * On exit, it contains: 1201 * IPv6 Header -> Unfragmentable Part -> Frag6 Header 1202 */ 1203 static int 1204 ip6_insertfraghdr(struct mbuf *m0, struct mbuf *m, int hlen, 1205 struct ip6_frag **frghdrp) 1206 { 1207 struct mbuf *n, *mlast; 1208 1209 if (hlen > sizeof(struct ip6_hdr)) { 1210 n = m_copym(m0, sizeof(struct ip6_hdr), 1211 hlen - sizeof(struct ip6_hdr), M_DONTWAIT); 1212 if (n == NULL) 1213 return ENOBUFS; 1214 m->m_next = n; 1215 } else 1216 n = m; 1217 1218 /* Search for the last mbuf of unfragmentable part. */ 1219 for (mlast = n; mlast->m_next; mlast = mlast->m_next) 1220 ; 1221 1222 if ((mlast->m_flags & M_EXT) == 0 && 1223 M_TRAILINGSPACE(mlast) >= sizeof(struct ip6_frag)) { 1224 /* use the trailing space of the last mbuf for the fragment hdr */ 1225 *frghdrp = (struct ip6_frag *)(mtod(mlast, char *) + 1226 mlast->m_len); 1227 mlast->m_len += sizeof(struct ip6_frag); 1228 } else { 1229 /* allocate a new mbuf for the fragment header */ 1230 struct mbuf *mfrg; 1231 1232 MGET(mfrg, M_DONTWAIT, MT_DATA); 1233 if (mfrg == NULL) 1234 return ENOBUFS; 1235 mfrg->m_len = sizeof(struct ip6_frag); 1236 *frghdrp = mtod(mfrg, struct ip6_frag *); 1237 mlast->m_next = mfrg; 1238 } 1239 1240 return 0; 1241 } 1242 1243 static int 1244 ip6_getpmtu(struct rtentry *rt, struct ifnet *ifp, u_long *mtup, 1245 int *alwaysfragp) 1246 { 1247 u_int32_t mtu = 0; 1248 int alwaysfrag = 0; 1249 int error = 0; 1250 1251 if (rt != NULL) { 1252 if (ifp == NULL) 1253 ifp = rt->rt_ifp; 1254 mtu = rt->rt_rmx.rmx_mtu; 1255 if (mtu == 0) 1256 mtu = ifp->if_mtu; 1257 else if (mtu < IPV6_MMTU) { 1258 /* 1259 * RFC2460 section 5, last paragraph: 1260 * if we record ICMPv6 too big message with 1261 * mtu < IPV6_MMTU, transmit packets sized IPV6_MMTU 1262 * or smaller, with fragment header attached. 1263 * (fragment header is needed regardless from the 1264 * packet size, for translators to identify packets) 1265 */ 1266 alwaysfrag = 1; 1267 mtu = IPV6_MMTU; 1268 } else if (mtu > ifp->if_mtu) { 1269 /* 1270 * The MTU on the route is larger than the MTU on 1271 * the interface! This shouldn't happen, unless the 1272 * MTU of the interface has been changed after the 1273 * interface was brought up. Change the MTU in the 1274 * route to match the interface MTU (as long as the 1275 * field isn't locked). 1276 */ 1277 mtu = ifp->if_mtu; 1278 if (!(rt->rt_rmx.rmx_locks & RTV_MTU)) 1279 rt->rt_rmx.rmx_mtu = mtu; 1280 } 1281 } else if (ifp) { 1282 mtu = ifp->if_mtu; 1283 } else 1284 error = EHOSTUNREACH; /* XXX */ 1285 1286 *mtup = mtu; 1287 if (alwaysfragp) 1288 *alwaysfragp = alwaysfrag; 1289 return (error); 1290 } 1291 1292 /* 1293 * IP6 socket option processing. 1294 */ 1295 int 1296 ip6_ctloutput(int op, struct socket *so, struct sockopt *sopt) 1297 { 1298 int optdatalen, uproto; 1299 void *optdata; 1300 struct in6pcb *in6p = sotoin6pcb(so); 1301 struct ip_moptions **mopts; 1302 int error, optval; 1303 int level, optname; 1304 1305 KASSERT(solocked(so)); 1306 KASSERT(sopt != NULL); 1307 1308 level = sopt->sopt_level; 1309 optname = sopt->sopt_name; 1310 1311 error = optval = 0; 1312 uproto = (int)so->so_proto->pr_protocol; 1313 1314 switch (level) { 1315 case IPPROTO_IP: 1316 switch (optname) { 1317 case IP_ADD_MEMBERSHIP: 1318 case IP_DROP_MEMBERSHIP: 1319 case IP_MULTICAST_IF: 1320 case IP_MULTICAST_LOOP: 1321 case IP_MULTICAST_TTL: 1322 mopts = &in6p->in6p_v4moptions; 1323 switch (op) { 1324 case PRCO_GETOPT: 1325 return ip_getmoptions(*mopts, sopt); 1326 case PRCO_SETOPT: 1327 return ip_setmoptions(mopts, sopt); 1328 default: 1329 return EINVAL; 1330 } 1331 default: 1332 return ENOPROTOOPT; 1333 } 1334 case IPPROTO_IPV6: 1335 break; 1336 default: 1337 return ENOPROTOOPT; 1338 } 1339 switch (op) { 1340 case PRCO_SETOPT: 1341 switch (optname) { 1342 #ifdef RFC2292 1343 case IPV6_2292PKTOPTIONS: 1344 error = ip6_pcbopts(&in6p->in6p_outputopts, so, sopt); 1345 break; 1346 #endif 1347 1348 /* 1349 * Use of some Hop-by-Hop options or some 1350 * Destination options, might require special 1351 * privilege. That is, normal applications 1352 * (without special privilege) might be forbidden 1353 * from setting certain options in outgoing packets, 1354 * and might never see certain options in received 1355 * packets. [RFC 2292 Section 6] 1356 * KAME specific note: 1357 * KAME prevents non-privileged users from sending or 1358 * receiving ANY hbh/dst options in order to avoid 1359 * overhead of parsing options in the kernel. 1360 */ 1361 case IPV6_RECVHOPOPTS: 1362 case IPV6_RECVDSTOPTS: 1363 case IPV6_RECVRTHDRDSTOPTS: 1364 error = kauth_authorize_network(kauth_cred_get(), 1365 KAUTH_NETWORK_IPV6, KAUTH_REQ_NETWORK_IPV6_HOPBYHOP, 1366 NULL, NULL, NULL); 1367 if (error) 1368 break; 1369 /* FALLTHROUGH */ 1370 case IPV6_UNICAST_HOPS: 1371 case IPV6_HOPLIMIT: 1372 case IPV6_FAITH: 1373 1374 case IPV6_RECVPKTINFO: 1375 case IPV6_RECVHOPLIMIT: 1376 case IPV6_RECVRTHDR: 1377 case IPV6_RECVPATHMTU: 1378 case IPV6_RECVTCLASS: 1379 case IPV6_V6ONLY: 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->in6p_hops = optval; 1390 } 1391 break; 1392 #define OPTSET(bit) \ 1393 do { \ 1394 if (optval) \ 1395 in6p->in6p_flags |= (bit); \ 1396 else \ 1397 in6p->in6p_flags &= ~(bit); \ 1398 } while (/*CONSTCOND*/ 0) 1399 1400 #ifdef RFC2292 1401 #define OPTSET2292(bit) \ 1402 do { \ 1403 in6p->in6p_flags |= IN6P_RFC2292; \ 1404 if (optval) \ 1405 in6p->in6p_flags |= (bit); \ 1406 else \ 1407 in6p->in6p_flags &= ~(bit); \ 1408 } while (/*CONSTCOND*/ 0) 1409 #endif 1410 1411 #define OPTBIT(bit) (in6p->in6p_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->in6p_outputopts; 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 (in6p->in6p_lport || 1521 !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr)) { 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 case IPV6_RECVTCLASS: 1533 #ifdef RFC2292 1534 /* cannot mix with RFC2292 XXX */ 1535 if (OPTBIT(IN6P_RFC2292)) { 1536 error = EINVAL; 1537 break; 1538 } 1539 #endif 1540 OPTSET(IN6P_TCLASS); 1541 break; 1542 1543 } 1544 break; 1545 1546 case IPV6_OTCLASS: 1547 { 1548 struct ip6_pktopts **optp; 1549 u_int8_t tclass; 1550 1551 error = sockopt_get(sopt, &tclass, sizeof(tclass)); 1552 if (error) 1553 break; 1554 optp = &in6p->in6p_outputopts; 1555 error = ip6_pcbopt(optname, 1556 (u_char *)&tclass, 1557 sizeof(tclass), 1558 optp, 1559 kauth_cred_get(), uproto); 1560 break; 1561 } 1562 1563 case IPV6_TCLASS: 1564 case IPV6_DONTFRAG: 1565 case IPV6_USE_MIN_MTU: 1566 case IPV6_PREFER_TEMPADDR: 1567 error = sockopt_getint(sopt, &optval); 1568 if (error) 1569 break; 1570 { 1571 struct ip6_pktopts **optp; 1572 optp = &in6p->in6p_outputopts; 1573 error = ip6_pcbopt(optname, 1574 (u_char *)&optval, 1575 sizeof(optval), 1576 optp, 1577 kauth_cred_get(), uproto); 1578 break; 1579 } 1580 1581 #ifdef RFC2292 1582 case IPV6_2292PKTINFO: 1583 case IPV6_2292HOPLIMIT: 1584 case IPV6_2292HOPOPTS: 1585 case IPV6_2292DSTOPTS: 1586 case IPV6_2292RTHDR: 1587 /* RFC 2292 */ 1588 error = sockopt_getint(sopt, &optval); 1589 if (error) 1590 break; 1591 1592 switch (optname) { 1593 case IPV6_2292PKTINFO: 1594 OPTSET2292(IN6P_PKTINFO); 1595 break; 1596 case IPV6_2292HOPLIMIT: 1597 OPTSET2292(IN6P_HOPLIMIT); 1598 break; 1599 case IPV6_2292HOPOPTS: 1600 /* 1601 * Check super-user privilege. 1602 * See comments for IPV6_RECVHOPOPTS. 1603 */ 1604 error = 1605 kauth_authorize_network(kauth_cred_get(), 1606 KAUTH_NETWORK_IPV6, 1607 KAUTH_REQ_NETWORK_IPV6_HOPBYHOP, NULL, 1608 NULL, NULL); 1609 if (error) 1610 return (error); 1611 OPTSET2292(IN6P_HOPOPTS); 1612 break; 1613 case IPV6_2292DSTOPTS: 1614 error = 1615 kauth_authorize_network(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_DSTOPTS|IN6P_RTHDRDSTOPTS); /* XXX */ 1622 break; 1623 case IPV6_2292RTHDR: 1624 OPTSET2292(IN6P_RTHDR); 1625 break; 1626 } 1627 break; 1628 #endif 1629 case IPV6_PKTINFO: 1630 case IPV6_HOPOPTS: 1631 case IPV6_RTHDR: 1632 case IPV6_DSTOPTS: 1633 case IPV6_RTHDRDSTOPTS: 1634 case IPV6_NEXTHOP: { 1635 /* new advanced API (RFC3542) */ 1636 void *optbuf; 1637 int optbuflen; 1638 struct ip6_pktopts **optp; 1639 1640 #ifdef RFC2292 1641 /* cannot mix with RFC2292 */ 1642 if (OPTBIT(IN6P_RFC2292)) { 1643 error = EINVAL; 1644 break; 1645 } 1646 #endif 1647 1648 optbuflen = sopt->sopt_size; 1649 optbuf = malloc(optbuflen, M_IP6OPT, M_NOWAIT); 1650 if (optbuf == NULL) { 1651 error = ENOBUFS; 1652 break; 1653 } 1654 1655 error = sockopt_get(sopt, optbuf, optbuflen); 1656 if (error) { 1657 free(optbuf, M_IP6OPT); 1658 break; 1659 } 1660 optp = &in6p->in6p_outputopts; 1661 error = ip6_pcbopt(optname, optbuf, optbuflen, 1662 optp, kauth_cred_get(), uproto); 1663 1664 free(optbuf, M_IP6OPT); 1665 break; 1666 } 1667 #undef OPTSET 1668 1669 case IPV6_MULTICAST_IF: 1670 case IPV6_MULTICAST_HOPS: 1671 case IPV6_MULTICAST_LOOP: 1672 case IPV6_JOIN_GROUP: 1673 case IPV6_LEAVE_GROUP: 1674 error = ip6_setmoptions(sopt, in6p); 1675 break; 1676 1677 case IPV6_PORTRANGE: 1678 error = sockopt_getint(sopt, &optval); 1679 if (error) 1680 break; 1681 1682 switch (optval) { 1683 case IPV6_PORTRANGE_DEFAULT: 1684 in6p->in6p_flags &= ~(IN6P_LOWPORT); 1685 in6p->in6p_flags &= ~(IN6P_HIGHPORT); 1686 break; 1687 1688 case IPV6_PORTRANGE_HIGH: 1689 in6p->in6p_flags &= ~(IN6P_LOWPORT); 1690 in6p->in6p_flags |= IN6P_HIGHPORT; 1691 break; 1692 1693 case IPV6_PORTRANGE_LOW: 1694 in6p->in6p_flags &= ~(IN6P_HIGHPORT); 1695 in6p->in6p_flags |= IN6P_LOWPORT; 1696 break; 1697 1698 default: 1699 error = EINVAL; 1700 break; 1701 } 1702 break; 1703 1704 case IPV6_PORTALGO: 1705 error = sockopt_getint(sopt, &optval); 1706 if (error) 1707 break; 1708 1709 error = portalgo_algo_index_select( 1710 (struct inpcb_hdr *)in6p, optval); 1711 break; 1712 1713 #if defined(IPSEC) 1714 case IPV6_IPSEC_POLICY: 1715 if (ipsec_enabled) { 1716 error = ipsec_set_policy(in6p, 1717 sopt->sopt_data, sopt->sopt_size, 1718 kauth_cred_get()); 1719 } else 1720 error = ENOPROTOOPT; 1721 break; 1722 #endif /* IPSEC */ 1723 1724 default: 1725 error = ENOPROTOOPT; 1726 break; 1727 } 1728 break; 1729 1730 case PRCO_GETOPT: 1731 switch (optname) { 1732 #ifdef RFC2292 1733 case IPV6_2292PKTOPTIONS: 1734 /* 1735 * RFC3542 (effectively) deprecated the 1736 * semantics of the 2292-style pktoptions. 1737 * Since it was not reliable in nature (i.e., 1738 * applications had to expect the lack of some 1739 * information after all), it would make sense 1740 * to simplify this part by always returning 1741 * empty data. 1742 */ 1743 break; 1744 #endif 1745 1746 case IPV6_RECVHOPOPTS: 1747 case IPV6_RECVDSTOPTS: 1748 case IPV6_RECVRTHDRDSTOPTS: 1749 case IPV6_UNICAST_HOPS: 1750 case IPV6_RECVPKTINFO: 1751 case IPV6_RECVHOPLIMIT: 1752 case IPV6_RECVRTHDR: 1753 case IPV6_RECVPATHMTU: 1754 1755 case IPV6_FAITH: 1756 case IPV6_V6ONLY: 1757 case IPV6_PORTRANGE: 1758 case IPV6_RECVTCLASS: 1759 switch (optname) { 1760 1761 case IPV6_RECVHOPOPTS: 1762 optval = OPTBIT(IN6P_HOPOPTS); 1763 break; 1764 1765 case IPV6_RECVDSTOPTS: 1766 optval = OPTBIT(IN6P_DSTOPTS); 1767 break; 1768 1769 case IPV6_RECVRTHDRDSTOPTS: 1770 optval = OPTBIT(IN6P_RTHDRDSTOPTS); 1771 break; 1772 1773 case IPV6_UNICAST_HOPS: 1774 optval = in6p->in6p_hops; 1775 break; 1776 1777 case IPV6_RECVPKTINFO: 1778 optval = OPTBIT(IN6P_PKTINFO); 1779 break; 1780 1781 case IPV6_RECVHOPLIMIT: 1782 optval = OPTBIT(IN6P_HOPLIMIT); 1783 break; 1784 1785 case IPV6_RECVRTHDR: 1786 optval = OPTBIT(IN6P_RTHDR); 1787 break; 1788 1789 case IPV6_RECVPATHMTU: 1790 optval = OPTBIT(IN6P_MTU); 1791 break; 1792 1793 case IPV6_FAITH: 1794 optval = OPTBIT(IN6P_FAITH); 1795 break; 1796 1797 case IPV6_V6ONLY: 1798 optval = OPTBIT(IN6P_IPV6_V6ONLY); 1799 break; 1800 1801 case IPV6_PORTRANGE: 1802 { 1803 int flags; 1804 flags = in6p->in6p_flags; 1805 if (flags & IN6P_HIGHPORT) 1806 optval = IPV6_PORTRANGE_HIGH; 1807 else if (flags & IN6P_LOWPORT) 1808 optval = IPV6_PORTRANGE_LOW; 1809 else 1810 optval = 0; 1811 break; 1812 } 1813 case IPV6_RECVTCLASS: 1814 optval = OPTBIT(IN6P_TCLASS); 1815 break; 1816 1817 } 1818 if (error) 1819 break; 1820 error = sockopt_setint(sopt, optval); 1821 break; 1822 1823 case IPV6_PATHMTU: 1824 { 1825 u_long pmtu = 0; 1826 struct ip6_mtuinfo mtuinfo; 1827 struct route *ro = &in6p->in6p_route; 1828 struct rtentry *rt; 1829 union { 1830 struct sockaddr dst; 1831 struct sockaddr_in6 dst6; 1832 } u; 1833 1834 if (!(so->so_state & SS_ISCONNECTED)) 1835 return (ENOTCONN); 1836 /* 1837 * XXX: we dot not consider the case of source 1838 * routing, or optional information to specify 1839 * the outgoing interface. 1840 */ 1841 sockaddr_in6_init(&u.dst6, &in6p->in6p_faddr, 0, 0, 0); 1842 rt = rtcache_lookup(ro, &u.dst); 1843 error = ip6_getpmtu(rt, NULL, &pmtu, NULL); 1844 rtcache_unref(rt, ro); 1845 if (error) 1846 break; 1847 if (pmtu > IPV6_MAXPACKET) 1848 pmtu = IPV6_MAXPACKET; 1849 1850 memset(&mtuinfo, 0, sizeof(mtuinfo)); 1851 mtuinfo.ip6m_mtu = (u_int32_t)pmtu; 1852 optdata = (void *)&mtuinfo; 1853 optdatalen = sizeof(mtuinfo); 1854 if (optdatalen > MCLBYTES) 1855 return (EMSGSIZE); /* XXX */ 1856 error = sockopt_set(sopt, optdata, optdatalen); 1857 break; 1858 } 1859 1860 #ifdef RFC2292 1861 case IPV6_2292PKTINFO: 1862 case IPV6_2292HOPLIMIT: 1863 case IPV6_2292HOPOPTS: 1864 case IPV6_2292RTHDR: 1865 case IPV6_2292DSTOPTS: 1866 switch (optname) { 1867 case IPV6_2292PKTINFO: 1868 optval = OPTBIT(IN6P_PKTINFO); 1869 break; 1870 case IPV6_2292HOPLIMIT: 1871 optval = OPTBIT(IN6P_HOPLIMIT); 1872 break; 1873 case IPV6_2292HOPOPTS: 1874 optval = OPTBIT(IN6P_HOPOPTS); 1875 break; 1876 case IPV6_2292RTHDR: 1877 optval = OPTBIT(IN6P_RTHDR); 1878 break; 1879 case IPV6_2292DSTOPTS: 1880 optval = OPTBIT(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); 1881 break; 1882 } 1883 error = sockopt_setint(sopt, optval); 1884 break; 1885 #endif 1886 case IPV6_PKTINFO: 1887 case IPV6_HOPOPTS: 1888 case IPV6_RTHDR: 1889 case IPV6_DSTOPTS: 1890 case IPV6_RTHDRDSTOPTS: 1891 case IPV6_NEXTHOP: 1892 case IPV6_OTCLASS: 1893 case IPV6_TCLASS: 1894 case IPV6_DONTFRAG: 1895 case IPV6_USE_MIN_MTU: 1896 case IPV6_PREFER_TEMPADDR: 1897 error = ip6_getpcbopt(in6p->in6p_outputopts, 1898 optname, sopt); 1899 break; 1900 1901 case IPV6_MULTICAST_IF: 1902 case IPV6_MULTICAST_HOPS: 1903 case IPV6_MULTICAST_LOOP: 1904 case IPV6_JOIN_GROUP: 1905 case IPV6_LEAVE_GROUP: 1906 error = ip6_getmoptions(sopt, in6p); 1907 break; 1908 1909 case IPV6_PORTALGO: 1910 optval = ((struct inpcb_hdr *)in6p)->inph_portalgo; 1911 error = sockopt_setint(sopt, optval); 1912 break; 1913 1914 #if defined(IPSEC) 1915 case IPV6_IPSEC_POLICY: 1916 if (ipsec_used) { 1917 struct mbuf *m = NULL; 1918 1919 /* 1920 * XXX: this will return EINVAL as sopt is 1921 * empty 1922 */ 1923 error = ipsec_get_policy(in6p, sopt->sopt_data, 1924 sopt->sopt_size, &m); 1925 if (!error) 1926 error = sockopt_setmbuf(sopt, m); 1927 } else 1928 error = ENOPROTOOPT; 1929 break; 1930 #endif /* IPSEC */ 1931 1932 default: 1933 error = ENOPROTOOPT; 1934 break; 1935 } 1936 break; 1937 } 1938 return (error); 1939 } 1940 1941 int 1942 ip6_raw_ctloutput(int op, struct socket *so, struct sockopt *sopt) 1943 { 1944 int error = 0, optval; 1945 const int icmp6off = offsetof(struct icmp6_hdr, icmp6_cksum); 1946 struct in6pcb *in6p = sotoin6pcb(so); 1947 int level, optname; 1948 1949 KASSERT(sopt != NULL); 1950 1951 level = sopt->sopt_level; 1952 optname = sopt->sopt_name; 1953 1954 if (level != IPPROTO_IPV6) { 1955 return ENOPROTOOPT; 1956 } 1957 1958 switch (optname) { 1959 case IPV6_CHECKSUM: 1960 /* 1961 * For ICMPv6 sockets, no modification allowed for checksum 1962 * offset, permit "no change" values to help existing apps. 1963 * 1964 * XXX RFC3542 says: "An attempt to set IPV6_CHECKSUM 1965 * for an ICMPv6 socket will fail." The current 1966 * behavior does not meet RFC3542. 1967 */ 1968 switch (op) { 1969 case PRCO_SETOPT: 1970 error = sockopt_getint(sopt, &optval); 1971 if (error) 1972 break; 1973 if ((optval % 2) != 0) { 1974 /* the API assumes even offset values */ 1975 error = EINVAL; 1976 } else if (so->so_proto->pr_protocol == 1977 IPPROTO_ICMPV6) { 1978 if (optval != icmp6off) 1979 error = EINVAL; 1980 } else 1981 in6p->in6p_cksum = optval; 1982 break; 1983 1984 case PRCO_GETOPT: 1985 if (so->so_proto->pr_protocol == IPPROTO_ICMPV6) 1986 optval = icmp6off; 1987 else 1988 optval = in6p->in6p_cksum; 1989 1990 error = sockopt_setint(sopt, optval); 1991 break; 1992 1993 default: 1994 error = EINVAL; 1995 break; 1996 } 1997 break; 1998 1999 default: 2000 error = ENOPROTOOPT; 2001 break; 2002 } 2003 2004 return (error); 2005 } 2006 2007 #ifdef RFC2292 2008 /* 2009 * Set up IP6 options in pcb for insertion in output packets or 2010 * specifying behavior of outgoing packets. 2011 */ 2012 static int 2013 ip6_pcbopts(struct ip6_pktopts **pktopt, struct socket *so, 2014 struct sockopt *sopt) 2015 { 2016 struct ip6_pktopts *opt = *pktopt; 2017 struct mbuf *m; 2018 int error = 0; 2019 2020 KASSERT(solocked(so)); 2021 2022 /* turn off any old options. */ 2023 if (opt) { 2024 #ifdef DIAGNOSTIC 2025 if (opt->ip6po_pktinfo || opt->ip6po_nexthop || 2026 opt->ip6po_hbh || opt->ip6po_dest1 || opt->ip6po_dest2 || 2027 opt->ip6po_rhinfo.ip6po_rhi_rthdr) 2028 printf("ip6_pcbopts: all specified options are cleared.\n"); 2029 #endif 2030 ip6_clearpktopts(opt, -1); 2031 } else { 2032 opt = malloc(sizeof(*opt), M_IP6OPT, M_NOWAIT); 2033 if (opt == NULL) 2034 return (ENOBUFS); 2035 } 2036 *pktopt = NULL; 2037 2038 if (sopt == NULL || sopt->sopt_size == 0) { 2039 /* 2040 * Only turning off any previous options, regardless of 2041 * whether the opt is just created or given. 2042 */ 2043 free(opt, M_IP6OPT); 2044 return (0); 2045 } 2046 2047 /* set options specified by user. */ 2048 m = sockopt_getmbuf(sopt); 2049 if (m == NULL) { 2050 free(opt, M_IP6OPT); 2051 return (ENOBUFS); 2052 } 2053 2054 error = ip6_setpktopts(m, opt, NULL, kauth_cred_get(), 2055 so->so_proto->pr_protocol); 2056 m_freem(m); 2057 if (error != 0) { 2058 ip6_clearpktopts(opt, -1); /* XXX: discard all options */ 2059 free(opt, M_IP6OPT); 2060 return (error); 2061 } 2062 *pktopt = opt; 2063 return (0); 2064 } 2065 #endif 2066 2067 /* 2068 * initialize ip6_pktopts. beware that there are non-zero default values in 2069 * the struct. 2070 */ 2071 void 2072 ip6_initpktopts(struct ip6_pktopts *opt) 2073 { 2074 2075 memset(opt, 0, sizeof(*opt)); 2076 opt->ip6po_hlim = -1; /* -1 means default hop limit */ 2077 opt->ip6po_tclass = -1; /* -1 means default traffic class */ 2078 opt->ip6po_minmtu = IP6PO_MINMTU_MCASTONLY; 2079 opt->ip6po_prefer_tempaddr = IP6PO_TEMPADDR_SYSTEM; 2080 } 2081 2082 #define sin6tosa(sin6) ((struct sockaddr *)(sin6)) /* XXX */ 2083 static int 2084 ip6_pcbopt(int optname, u_char *buf, int len, struct ip6_pktopts **pktopt, 2085 kauth_cred_t cred, int uproto) 2086 { 2087 struct ip6_pktopts *opt; 2088 2089 if (*pktopt == NULL) { 2090 *pktopt = malloc(sizeof(struct ip6_pktopts), M_IP6OPT, 2091 M_NOWAIT); 2092 if (*pktopt == NULL) 2093 return (ENOBUFS); 2094 2095 ip6_initpktopts(*pktopt); 2096 } 2097 opt = *pktopt; 2098 2099 return (ip6_setpktopt(optname, buf, len, opt, cred, 1, 0, uproto)); 2100 } 2101 2102 static int 2103 ip6_getpcbopt(struct ip6_pktopts *pktopt, int optname, struct sockopt *sopt) 2104 { 2105 void *optdata = NULL; 2106 int optdatalen = 0; 2107 struct ip6_ext *ip6e; 2108 int error = 0; 2109 struct in6_pktinfo null_pktinfo; 2110 int deftclass = 0, on; 2111 int defminmtu = IP6PO_MINMTU_MCASTONLY; 2112 int defpreftemp = IP6PO_TEMPADDR_SYSTEM; 2113 2114 switch (optname) { 2115 case IPV6_PKTINFO: 2116 if (pktopt && pktopt->ip6po_pktinfo) 2117 optdata = (void *)pktopt->ip6po_pktinfo; 2118 else { 2119 /* XXX: we don't have to do this every time... */ 2120 memset(&null_pktinfo, 0, sizeof(null_pktinfo)); 2121 optdata = (void *)&null_pktinfo; 2122 } 2123 optdatalen = sizeof(struct in6_pktinfo); 2124 break; 2125 case IPV6_OTCLASS: 2126 /* XXX */ 2127 return (EINVAL); 2128 case IPV6_TCLASS: 2129 if (pktopt && pktopt->ip6po_tclass >= 0) 2130 optdata = (void *)&pktopt->ip6po_tclass; 2131 else 2132 optdata = (void *)&deftclass; 2133 optdatalen = sizeof(int); 2134 break; 2135 case IPV6_HOPOPTS: 2136 if (pktopt && pktopt->ip6po_hbh) { 2137 optdata = (void *)pktopt->ip6po_hbh; 2138 ip6e = (struct ip6_ext *)pktopt->ip6po_hbh; 2139 optdatalen = (ip6e->ip6e_len + 1) << 3; 2140 } 2141 break; 2142 case IPV6_RTHDR: 2143 if (pktopt && pktopt->ip6po_rthdr) { 2144 optdata = (void *)pktopt->ip6po_rthdr; 2145 ip6e = (struct ip6_ext *)pktopt->ip6po_rthdr; 2146 optdatalen = (ip6e->ip6e_len + 1) << 3; 2147 } 2148 break; 2149 case IPV6_RTHDRDSTOPTS: 2150 if (pktopt && pktopt->ip6po_dest1) { 2151 optdata = (void *)pktopt->ip6po_dest1; 2152 ip6e = (struct ip6_ext *)pktopt->ip6po_dest1; 2153 optdatalen = (ip6e->ip6e_len + 1) << 3; 2154 } 2155 break; 2156 case IPV6_DSTOPTS: 2157 if (pktopt && pktopt->ip6po_dest2) { 2158 optdata = (void *)pktopt->ip6po_dest2; 2159 ip6e = (struct ip6_ext *)pktopt->ip6po_dest2; 2160 optdatalen = (ip6e->ip6e_len + 1) << 3; 2161 } 2162 break; 2163 case IPV6_NEXTHOP: 2164 if (pktopt && pktopt->ip6po_nexthop) { 2165 optdata = (void *)pktopt->ip6po_nexthop; 2166 optdatalen = pktopt->ip6po_nexthop->sa_len; 2167 } 2168 break; 2169 case IPV6_USE_MIN_MTU: 2170 if (pktopt) 2171 optdata = (void *)&pktopt->ip6po_minmtu; 2172 else 2173 optdata = (void *)&defminmtu; 2174 optdatalen = sizeof(int); 2175 break; 2176 case IPV6_DONTFRAG: 2177 if (pktopt && ((pktopt->ip6po_flags) & IP6PO_DONTFRAG)) 2178 on = 1; 2179 else 2180 on = 0; 2181 optdata = (void *)&on; 2182 optdatalen = sizeof(on); 2183 break; 2184 case IPV6_PREFER_TEMPADDR: 2185 if (pktopt) 2186 optdata = (void *)&pktopt->ip6po_prefer_tempaddr; 2187 else 2188 optdata = (void *)&defpreftemp; 2189 optdatalen = sizeof(int); 2190 break; 2191 default: /* should not happen */ 2192 #ifdef DIAGNOSTIC 2193 panic("ip6_getpcbopt: unexpected option\n"); 2194 #endif 2195 return (ENOPROTOOPT); 2196 } 2197 2198 error = sockopt_set(sopt, optdata, optdatalen); 2199 2200 return (error); 2201 } 2202 2203 void 2204 ip6_clearpktopts(struct ip6_pktopts *pktopt, int optname) 2205 { 2206 if (optname == -1 || optname == IPV6_PKTINFO) { 2207 if (pktopt->ip6po_pktinfo) 2208 free(pktopt->ip6po_pktinfo, M_IP6OPT); 2209 pktopt->ip6po_pktinfo = NULL; 2210 } 2211 if (optname == -1 || optname == IPV6_HOPLIMIT) 2212 pktopt->ip6po_hlim = -1; 2213 if (optname == -1 || optname == IPV6_TCLASS) 2214 pktopt->ip6po_tclass = -1; 2215 if (optname == -1 || optname == IPV6_NEXTHOP) { 2216 rtcache_free(&pktopt->ip6po_nextroute); 2217 if (pktopt->ip6po_nexthop) 2218 free(pktopt->ip6po_nexthop, M_IP6OPT); 2219 pktopt->ip6po_nexthop = NULL; 2220 } 2221 if (optname == -1 || optname == IPV6_HOPOPTS) { 2222 if (pktopt->ip6po_hbh) 2223 free(pktopt->ip6po_hbh, M_IP6OPT); 2224 pktopt->ip6po_hbh = NULL; 2225 } 2226 if (optname == -1 || optname == IPV6_RTHDRDSTOPTS) { 2227 if (pktopt->ip6po_dest1) 2228 free(pktopt->ip6po_dest1, M_IP6OPT); 2229 pktopt->ip6po_dest1 = NULL; 2230 } 2231 if (optname == -1 || optname == IPV6_RTHDR) { 2232 if (pktopt->ip6po_rhinfo.ip6po_rhi_rthdr) 2233 free(pktopt->ip6po_rhinfo.ip6po_rhi_rthdr, M_IP6OPT); 2234 pktopt->ip6po_rhinfo.ip6po_rhi_rthdr = NULL; 2235 rtcache_free(&pktopt->ip6po_route); 2236 } 2237 if (optname == -1 || optname == IPV6_DSTOPTS) { 2238 if (pktopt->ip6po_dest2) 2239 free(pktopt->ip6po_dest2, M_IP6OPT); 2240 pktopt->ip6po_dest2 = NULL; 2241 } 2242 } 2243 2244 #define PKTOPT_EXTHDRCPY(type) \ 2245 do { \ 2246 if (src->type) { \ 2247 int hlen = (((struct ip6_ext *)src->type)->ip6e_len + 1) << 3;\ 2248 dst->type = malloc(hlen, M_IP6OPT, canwait); \ 2249 if (dst->type == NULL) \ 2250 goto bad; \ 2251 memcpy(dst->type, src->type, hlen); \ 2252 } \ 2253 } while (/*CONSTCOND*/ 0) 2254 2255 static int 2256 copypktopts(struct ip6_pktopts *dst, struct ip6_pktopts *src, int canwait) 2257 { 2258 dst->ip6po_hlim = src->ip6po_hlim; 2259 dst->ip6po_tclass = src->ip6po_tclass; 2260 dst->ip6po_flags = src->ip6po_flags; 2261 dst->ip6po_minmtu = src->ip6po_minmtu; 2262 dst->ip6po_prefer_tempaddr = src->ip6po_prefer_tempaddr; 2263 if (src->ip6po_pktinfo) { 2264 dst->ip6po_pktinfo = malloc(sizeof(*dst->ip6po_pktinfo), 2265 M_IP6OPT, canwait); 2266 if (dst->ip6po_pktinfo == NULL) 2267 goto bad; 2268 *dst->ip6po_pktinfo = *src->ip6po_pktinfo; 2269 } 2270 if (src->ip6po_nexthop) { 2271 dst->ip6po_nexthop = malloc(src->ip6po_nexthop->sa_len, 2272 M_IP6OPT, canwait); 2273 if (dst->ip6po_nexthop == NULL) 2274 goto bad; 2275 memcpy(dst->ip6po_nexthop, src->ip6po_nexthop, 2276 src->ip6po_nexthop->sa_len); 2277 } 2278 PKTOPT_EXTHDRCPY(ip6po_hbh); 2279 PKTOPT_EXTHDRCPY(ip6po_dest1); 2280 PKTOPT_EXTHDRCPY(ip6po_dest2); 2281 PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */ 2282 return (0); 2283 2284 bad: 2285 if (dst->ip6po_pktinfo) free(dst->ip6po_pktinfo, M_IP6OPT); 2286 if (dst->ip6po_nexthop) free(dst->ip6po_nexthop, M_IP6OPT); 2287 if (dst->ip6po_hbh) free(dst->ip6po_hbh, M_IP6OPT); 2288 if (dst->ip6po_dest1) free(dst->ip6po_dest1, M_IP6OPT); 2289 if (dst->ip6po_dest2) free(dst->ip6po_dest2, M_IP6OPT); 2290 if (dst->ip6po_rthdr) free(dst->ip6po_rthdr, M_IP6OPT); 2291 2292 return (ENOBUFS); 2293 } 2294 #undef PKTOPT_EXTHDRCPY 2295 2296 struct ip6_pktopts * 2297 ip6_copypktopts(struct ip6_pktopts *src, int canwait) 2298 { 2299 int error; 2300 struct ip6_pktopts *dst; 2301 2302 dst = malloc(sizeof(*dst), M_IP6OPT, canwait); 2303 if (dst == NULL) 2304 return (NULL); 2305 ip6_initpktopts(dst); 2306 2307 if ((error = copypktopts(dst, src, canwait)) != 0) { 2308 free(dst, M_IP6OPT); 2309 return (NULL); 2310 } 2311 2312 return (dst); 2313 } 2314 2315 void 2316 ip6_freepcbopts(struct ip6_pktopts *pktopt) 2317 { 2318 if (pktopt == NULL) 2319 return; 2320 2321 ip6_clearpktopts(pktopt, -1); 2322 2323 free(pktopt, M_IP6OPT); 2324 } 2325 2326 int 2327 ip6_get_membership(const struct sockopt *sopt, struct ifnet **ifp, 2328 struct psref *psref, void *v, size_t l) 2329 { 2330 struct ipv6_mreq mreq; 2331 int error; 2332 struct in6_addr *ia = &mreq.ipv6mr_multiaddr; 2333 struct in_addr *ia4 = (void *)&ia->s6_addr32[3]; 2334 2335 error = sockopt_get(sopt, &mreq, sizeof(mreq)); 2336 if (error != 0) 2337 return error; 2338 2339 if (IN6_IS_ADDR_UNSPECIFIED(ia)) { 2340 /* 2341 * We use the unspecified address to specify to accept 2342 * all multicast addresses. Only super user is allowed 2343 * to do this. 2344 */ 2345 if (kauth_authorize_network(curlwp->l_cred, KAUTH_NETWORK_IPV6, 2346 KAUTH_REQ_NETWORK_IPV6_JOIN_MULTICAST, NULL, NULL, NULL)) 2347 return EACCES; 2348 } else if (IN6_IS_ADDR_V4MAPPED(ia)) { 2349 // Don't bother if we are not going to use ifp. 2350 if (l == sizeof(*ia)) { 2351 memcpy(v, ia, l); 2352 return 0; 2353 } 2354 } else if (!IN6_IS_ADDR_MULTICAST(ia)) { 2355 return EINVAL; 2356 } 2357 2358 /* 2359 * If no interface was explicitly specified, choose an 2360 * appropriate one according to the given multicast address. 2361 */ 2362 if (mreq.ipv6mr_interface == 0) { 2363 struct rtentry *rt; 2364 union { 2365 struct sockaddr dst; 2366 struct sockaddr_in dst4; 2367 struct sockaddr_in6 dst6; 2368 } u; 2369 struct route ro; 2370 2371 /* 2372 * Look up the routing table for the 2373 * address, and choose the outgoing interface. 2374 * XXX: is it a good approach? 2375 */ 2376 memset(&ro, 0, sizeof(ro)); 2377 if (IN6_IS_ADDR_V4MAPPED(ia)) 2378 sockaddr_in_init(&u.dst4, ia4, 0); 2379 else 2380 sockaddr_in6_init(&u.dst6, ia, 0, 0, 0); 2381 error = rtcache_setdst(&ro, &u.dst); 2382 if (error != 0) 2383 return error; 2384 rt = rtcache_init(&ro); 2385 *ifp = rt != NULL ? 2386 if_get_byindex(rt->rt_ifp->if_index, psref) : NULL; 2387 rtcache_unref(rt, &ro); 2388 rtcache_free(&ro); 2389 } else { 2390 /* 2391 * If the interface is specified, validate it. 2392 */ 2393 *ifp = if_get_byindex(mreq.ipv6mr_interface, psref); 2394 if (*ifp == NULL) 2395 return ENXIO; /* XXX EINVAL? */ 2396 } 2397 if (sizeof(*ia) == l) 2398 memcpy(v, ia, l); 2399 else 2400 memcpy(v, ia4, l); 2401 return 0; 2402 } 2403 2404 /* 2405 * Set the IP6 multicast options in response to user setsockopt(). 2406 */ 2407 static int 2408 ip6_setmoptions(const struct sockopt *sopt, struct in6pcb *in6p) 2409 { 2410 int error = 0; 2411 u_int loop, ifindex; 2412 struct ipv6_mreq mreq; 2413 struct in6_addr ia; 2414 struct ifnet *ifp; 2415 struct ip6_moptions *im6o = in6p->in6p_moptions; 2416 struct in6_multi_mship *imm; 2417 2418 KASSERT(in6p_locked(in6p)); 2419 2420 if (im6o == NULL) { 2421 /* 2422 * No multicast option buffer attached to the pcb; 2423 * allocate one and initialize to default values. 2424 */ 2425 im6o = malloc(sizeof(*im6o), M_IPMOPTS, M_NOWAIT); 2426 if (im6o == NULL) 2427 return (ENOBUFS); 2428 in6p->in6p_moptions = im6o; 2429 im6o->im6o_multicast_if_index = 0; 2430 im6o->im6o_multicast_hlim = ip6_defmcasthlim; 2431 im6o->im6o_multicast_loop = IPV6_DEFAULT_MULTICAST_LOOP; 2432 LIST_INIT(&im6o->im6o_memberships); 2433 } 2434 2435 switch (sopt->sopt_name) { 2436 2437 case IPV6_MULTICAST_IF: { 2438 int s; 2439 /* 2440 * Select the interface for outgoing multicast packets. 2441 */ 2442 error = sockopt_get(sopt, &ifindex, sizeof(ifindex)); 2443 if (error != 0) 2444 break; 2445 2446 s = pserialize_read_enter(); 2447 if (ifindex != 0) { 2448 if ((ifp = if_byindex(ifindex)) == NULL) { 2449 pserialize_read_exit(s); 2450 error = ENXIO; /* XXX EINVAL? */ 2451 break; 2452 } 2453 if ((ifp->if_flags & IFF_MULTICAST) == 0) { 2454 pserialize_read_exit(s); 2455 error = EADDRNOTAVAIL; 2456 break; 2457 } 2458 } else 2459 ifp = NULL; 2460 im6o->im6o_multicast_if_index = if_get_index(ifp); 2461 pserialize_read_exit(s); 2462 break; 2463 } 2464 2465 case IPV6_MULTICAST_HOPS: 2466 { 2467 /* 2468 * Set the IP6 hoplimit for outgoing multicast packets. 2469 */ 2470 int optval; 2471 2472 error = sockopt_getint(sopt, &optval); 2473 if (error != 0) 2474 break; 2475 2476 if (optval < -1 || optval >= 256) 2477 error = EINVAL; 2478 else if (optval == -1) 2479 im6o->im6o_multicast_hlim = ip6_defmcasthlim; 2480 else 2481 im6o->im6o_multicast_hlim = optval; 2482 break; 2483 } 2484 2485 case IPV6_MULTICAST_LOOP: 2486 /* 2487 * Set the loopback flag for outgoing multicast packets. 2488 * Must be zero or one. 2489 */ 2490 error = sockopt_get(sopt, &loop, sizeof(loop)); 2491 if (error != 0) 2492 break; 2493 if (loop > 1) { 2494 error = EINVAL; 2495 break; 2496 } 2497 im6o->im6o_multicast_loop = loop; 2498 break; 2499 2500 case IPV6_JOIN_GROUP: { 2501 int bound; 2502 struct psref psref; 2503 /* 2504 * Add a multicast group membership. 2505 * Group must be a valid IP6 multicast address. 2506 */ 2507 bound = curlwp_bind(); 2508 ifp = NULL; 2509 error = ip6_get_membership(sopt, &ifp, &psref, &ia, sizeof(ia)); 2510 if (error != 0) { 2511 KASSERT(ifp == NULL); 2512 curlwp_bindx(bound); 2513 return error; 2514 } 2515 2516 if (IN6_IS_ADDR_V4MAPPED(&ia)) { 2517 error = ip_setmoptions(&in6p->in6p_v4moptions, sopt); 2518 goto put_break; 2519 } 2520 /* 2521 * See if we found an interface, and confirm that it 2522 * supports multicast 2523 */ 2524 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) { 2525 error = EADDRNOTAVAIL; 2526 goto put_break; 2527 } 2528 2529 if (in6_setscope(&ia, ifp, NULL)) { 2530 error = EADDRNOTAVAIL; /* XXX: should not happen */ 2531 goto put_break; 2532 } 2533 2534 /* 2535 * See if the membership already exists. 2536 */ 2537 LIST_FOREACH(imm, &im6o->im6o_memberships, i6mm_chain) { 2538 if (imm->i6mm_maddr->in6m_ifp == ifp && 2539 IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr, 2540 &ia)) 2541 goto put_break; 2542 } 2543 if (imm != NULL) { 2544 error = EADDRINUSE; 2545 goto put_break; 2546 } 2547 /* 2548 * Everything looks good; add a new record to the multicast 2549 * address list for the given interface. 2550 */ 2551 imm = in6_joingroup(ifp, &ia, &error, 0); 2552 if (imm == NULL) 2553 goto put_break; 2554 LIST_INSERT_HEAD(&im6o->im6o_memberships, imm, i6mm_chain); 2555 put_break: 2556 if_put(ifp, &psref); 2557 curlwp_bindx(bound); 2558 break; 2559 } 2560 2561 case IPV6_LEAVE_GROUP: { 2562 /* 2563 * Drop a multicast group membership. 2564 * Group must be a valid IP6 multicast address. 2565 */ 2566 error = sockopt_get(sopt, &mreq, sizeof(mreq)); 2567 if (error != 0) 2568 break; 2569 2570 if (IN6_IS_ADDR_V4MAPPED(&mreq.ipv6mr_multiaddr)) { 2571 error = ip_setmoptions(&in6p->in6p_v4moptions, sopt); 2572 break; 2573 } 2574 /* 2575 * If an interface address was specified, get a pointer 2576 * to its ifnet structure. 2577 */ 2578 if (mreq.ipv6mr_interface != 0) { 2579 if ((ifp = if_byindex(mreq.ipv6mr_interface)) == NULL) { 2580 error = ENXIO; /* XXX EINVAL? */ 2581 break; 2582 } 2583 } else 2584 ifp = NULL; 2585 2586 /* Fill in the scope zone ID */ 2587 if (ifp) { 2588 if (in6_setscope(&mreq.ipv6mr_multiaddr, ifp, NULL)) { 2589 /* XXX: should not happen */ 2590 error = EADDRNOTAVAIL; 2591 break; 2592 } 2593 } else if (mreq.ipv6mr_interface != 0) { 2594 /* 2595 * XXX: This case would happens when the (positive) 2596 * index is in the valid range, but the corresponding 2597 * interface has been detached dynamically. The above 2598 * check probably avoids such case to happen here, but 2599 * we check it explicitly for safety. 2600 */ 2601 error = EADDRNOTAVAIL; 2602 break; 2603 } else { /* ipv6mr_interface == 0 */ 2604 struct sockaddr_in6 sa6_mc; 2605 2606 /* 2607 * The API spec says as follows: 2608 * If the interface index is specified as 0, the 2609 * system may choose a multicast group membership to 2610 * drop by matching the multicast address only. 2611 * On the other hand, we cannot disambiguate the scope 2612 * zone unless an interface is provided. Thus, we 2613 * check if there's ambiguity with the default scope 2614 * zone as the last resort. 2615 */ 2616 sockaddr_in6_init(&sa6_mc, &mreq.ipv6mr_multiaddr, 2617 0, 0, 0); 2618 error = sa6_embedscope(&sa6_mc, ip6_use_defzone); 2619 if (error != 0) 2620 break; 2621 mreq.ipv6mr_multiaddr = sa6_mc.sin6_addr; 2622 } 2623 2624 /* 2625 * Find the membership in the membership list. 2626 */ 2627 LIST_FOREACH(imm, &im6o->im6o_memberships, i6mm_chain) { 2628 if ((ifp == NULL || imm->i6mm_maddr->in6m_ifp == ifp) && 2629 IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr, 2630 &mreq.ipv6mr_multiaddr)) 2631 break; 2632 } 2633 if (imm == NULL) { 2634 /* Unable to resolve interface */ 2635 error = EADDRNOTAVAIL; 2636 break; 2637 } 2638 /* 2639 * Give up the multicast address record to which the 2640 * membership points. 2641 */ 2642 LIST_REMOVE(imm, i6mm_chain); 2643 in6_leavegroup(imm); 2644 /* in6m_ifp should not leave thanks to in6p_lock */ 2645 break; 2646 } 2647 2648 default: 2649 error = EOPNOTSUPP; 2650 break; 2651 } 2652 2653 /* 2654 * If all options have default values, no need to keep the mbuf. 2655 */ 2656 if (im6o->im6o_multicast_if_index == 0 && 2657 im6o->im6o_multicast_hlim == ip6_defmcasthlim && 2658 im6o->im6o_multicast_loop == IPV6_DEFAULT_MULTICAST_LOOP && 2659 LIST_EMPTY(&im6o->im6o_memberships)) { 2660 free(in6p->in6p_moptions, M_IPMOPTS); 2661 in6p->in6p_moptions = NULL; 2662 } 2663 2664 return (error); 2665 } 2666 2667 /* 2668 * Return the IP6 multicast options in response to user getsockopt(). 2669 */ 2670 static int 2671 ip6_getmoptions(struct sockopt *sopt, struct in6pcb *in6p) 2672 { 2673 u_int optval; 2674 int error; 2675 struct ip6_moptions *im6o = in6p->in6p_moptions; 2676 2677 switch (sopt->sopt_name) { 2678 case IPV6_MULTICAST_IF: 2679 if (im6o == NULL || im6o->im6o_multicast_if_index == 0) 2680 optval = 0; 2681 else 2682 optval = im6o->im6o_multicast_if_index; 2683 2684 error = sockopt_set(sopt, &optval, sizeof(optval)); 2685 break; 2686 2687 case IPV6_MULTICAST_HOPS: 2688 if (im6o == NULL) 2689 optval = ip6_defmcasthlim; 2690 else 2691 optval = im6o->im6o_multicast_hlim; 2692 2693 error = sockopt_set(sopt, &optval, sizeof(optval)); 2694 break; 2695 2696 case IPV6_MULTICAST_LOOP: 2697 if (im6o == NULL) 2698 optval = IPV6_DEFAULT_MULTICAST_LOOP; 2699 else 2700 optval = im6o->im6o_multicast_loop; 2701 2702 error = sockopt_set(sopt, &optval, sizeof(optval)); 2703 break; 2704 2705 default: 2706 error = EOPNOTSUPP; 2707 } 2708 2709 return (error); 2710 } 2711 2712 /* 2713 * Discard the IP6 multicast options. 2714 */ 2715 void 2716 ip6_freemoptions(struct ip6_moptions *im6o) 2717 { 2718 struct in6_multi_mship *imm, *nimm; 2719 2720 if (im6o == NULL) 2721 return; 2722 2723 /* The owner of im6o (in6p) should be protected by solock */ 2724 LIST_FOREACH_SAFE(imm, &im6o->im6o_memberships, i6mm_chain, nimm) { 2725 LIST_REMOVE(imm, i6mm_chain); 2726 in6_leavegroup(imm); 2727 } 2728 free(im6o, M_IPMOPTS); 2729 } 2730 2731 /* 2732 * Set IPv6 outgoing packet options based on advanced API. 2733 */ 2734 int 2735 ip6_setpktopts(struct mbuf *control, struct ip6_pktopts *opt, 2736 struct ip6_pktopts *stickyopt, kauth_cred_t cred, int uproto) 2737 { 2738 struct cmsghdr *cm = 0; 2739 2740 if (control == NULL || opt == NULL) 2741 return (EINVAL); 2742 2743 ip6_initpktopts(opt); 2744 if (stickyopt) { 2745 int error; 2746 2747 /* 2748 * If stickyopt is provided, make a local copy of the options 2749 * for this particular packet, then override them by ancillary 2750 * objects. 2751 * XXX: copypktopts() does not copy the cached route to a next 2752 * hop (if any). This is not very good in terms of efficiency, 2753 * but we can allow this since this option should be rarely 2754 * used. 2755 */ 2756 if ((error = copypktopts(opt, stickyopt, M_NOWAIT)) != 0) 2757 return (error); 2758 } 2759 2760 /* 2761 * XXX: Currently, we assume all the optional information is stored 2762 * in a single mbuf. 2763 */ 2764 if (control->m_next) 2765 return (EINVAL); 2766 2767 /* XXX if cm->cmsg_len is not aligned, control->m_len can become <0 */ 2768 for (; control->m_len > 0; control->m_data += CMSG_ALIGN(cm->cmsg_len), 2769 control->m_len -= CMSG_ALIGN(cm->cmsg_len)) { 2770 int error; 2771 2772 if (control->m_len < CMSG_LEN(0)) 2773 return (EINVAL); 2774 2775 cm = mtod(control, struct cmsghdr *); 2776 if (cm->cmsg_len < CMSG_LEN(0) || cm->cmsg_len > control->m_len) 2777 return (EINVAL); 2778 if (cm->cmsg_level != IPPROTO_IPV6) 2779 continue; 2780 2781 error = ip6_setpktopt(cm->cmsg_type, CMSG_DATA(cm), 2782 cm->cmsg_len - CMSG_LEN(0), opt, cred, 0, 1, uproto); 2783 if (error) 2784 return (error); 2785 } 2786 2787 return (0); 2788 } 2789 2790 /* 2791 * Set a particular packet option, as a sticky option or an ancillary data 2792 * item. "len" can be 0 only when it's a sticky option. 2793 * We have 4 cases of combination of "sticky" and "cmsg": 2794 * "sticky=0, cmsg=0": impossible 2795 * "sticky=0, cmsg=1": RFC2292 or RFC3542 ancillary data 2796 * "sticky=1, cmsg=0": RFC3542 socket option 2797 * "sticky=1, cmsg=1": RFC2292 socket option 2798 */ 2799 static int 2800 ip6_setpktopt(int optname, u_char *buf, int len, struct ip6_pktopts *opt, 2801 kauth_cred_t cred, int sticky, int cmsg, int uproto) 2802 { 2803 int minmtupolicy; 2804 int error; 2805 2806 if (!sticky && !cmsg) { 2807 #ifdef DIAGNOSTIC 2808 printf("ip6_setpktopt: impossible case\n"); 2809 #endif 2810 return (EINVAL); 2811 } 2812 2813 /* 2814 * IPV6_2292xxx is for backward compatibility to RFC2292, and should 2815 * not be specified in the context of RFC3542. Conversely, 2816 * RFC3542 types should not be specified in the context of RFC2292. 2817 */ 2818 if (!cmsg) { 2819 switch (optname) { 2820 case IPV6_2292PKTINFO: 2821 case IPV6_2292HOPLIMIT: 2822 case IPV6_2292NEXTHOP: 2823 case IPV6_2292HOPOPTS: 2824 case IPV6_2292DSTOPTS: 2825 case IPV6_2292RTHDR: 2826 case IPV6_2292PKTOPTIONS: 2827 return (ENOPROTOOPT); 2828 } 2829 } 2830 if (sticky && cmsg) { 2831 switch (optname) { 2832 case IPV6_PKTINFO: 2833 case IPV6_HOPLIMIT: 2834 case IPV6_NEXTHOP: 2835 case IPV6_HOPOPTS: 2836 case IPV6_DSTOPTS: 2837 case IPV6_RTHDRDSTOPTS: 2838 case IPV6_RTHDR: 2839 case IPV6_USE_MIN_MTU: 2840 case IPV6_DONTFRAG: 2841 case IPV6_OTCLASS: 2842 case IPV6_TCLASS: 2843 case IPV6_PREFER_TEMPADDR: /* XXX not an RFC3542 option */ 2844 return (ENOPROTOOPT); 2845 } 2846 } 2847 2848 switch (optname) { 2849 #ifdef RFC2292 2850 case IPV6_2292PKTINFO: 2851 #endif 2852 case IPV6_PKTINFO: 2853 { 2854 struct in6_pktinfo *pktinfo; 2855 2856 if (len != sizeof(struct in6_pktinfo)) 2857 return (EINVAL); 2858 2859 pktinfo = (struct in6_pktinfo *)buf; 2860 2861 /* 2862 * An application can clear any sticky IPV6_PKTINFO option by 2863 * doing a "regular" setsockopt with ipi6_addr being 2864 * in6addr_any and ipi6_ifindex being zero. 2865 * [RFC 3542, Section 6] 2866 */ 2867 if (optname == IPV6_PKTINFO && opt->ip6po_pktinfo && 2868 pktinfo->ipi6_ifindex == 0 && 2869 IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) { 2870 ip6_clearpktopts(opt, optname); 2871 break; 2872 } 2873 2874 if (uproto == IPPROTO_TCP && optname == IPV6_PKTINFO && 2875 sticky && !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) { 2876 return (EINVAL); 2877 } 2878 2879 /* Validate the interface index if specified. */ 2880 if (pktinfo->ipi6_ifindex) { 2881 struct ifnet *ifp; 2882 int s = pserialize_read_enter(); 2883 ifp = if_byindex(pktinfo->ipi6_ifindex); 2884 if (ifp == NULL) { 2885 pserialize_read_exit(s); 2886 return ENXIO; 2887 } 2888 pserialize_read_exit(s); 2889 } 2890 2891 /* 2892 * We store the address anyway, and let in6_selectsrc() 2893 * validate the specified address. This is because ipi6_addr 2894 * may not have enough information about its scope zone, and 2895 * we may need additional information (such as outgoing 2896 * interface or the scope zone of a destination address) to 2897 * disambiguate the scope. 2898 * XXX: the delay of the validation may confuse the 2899 * application when it is used as a sticky option. 2900 */ 2901 if (opt->ip6po_pktinfo == NULL) { 2902 opt->ip6po_pktinfo = malloc(sizeof(*pktinfo), 2903 M_IP6OPT, M_NOWAIT); 2904 if (opt->ip6po_pktinfo == NULL) 2905 return (ENOBUFS); 2906 } 2907 memcpy(opt->ip6po_pktinfo, pktinfo, sizeof(*pktinfo)); 2908 break; 2909 } 2910 2911 #ifdef RFC2292 2912 case IPV6_2292HOPLIMIT: 2913 #endif 2914 case IPV6_HOPLIMIT: 2915 { 2916 int *hlimp; 2917 2918 /* 2919 * RFC 3542 deprecated the usage of sticky IPV6_HOPLIMIT 2920 * to simplify the ordering among hoplimit options. 2921 */ 2922 if (optname == IPV6_HOPLIMIT && sticky) 2923 return (ENOPROTOOPT); 2924 2925 if (len != sizeof(int)) 2926 return (EINVAL); 2927 hlimp = (int *)buf; 2928 if (*hlimp < -1 || *hlimp > 255) 2929 return (EINVAL); 2930 2931 opt->ip6po_hlim = *hlimp; 2932 break; 2933 } 2934 2935 case IPV6_OTCLASS: 2936 if (len != sizeof(u_int8_t)) 2937 return (EINVAL); 2938 2939 opt->ip6po_tclass = *(u_int8_t *)buf; 2940 break; 2941 2942 case IPV6_TCLASS: 2943 { 2944 int tclass; 2945 2946 if (len != sizeof(int)) 2947 return (EINVAL); 2948 tclass = *(int *)buf; 2949 if (tclass < -1 || tclass > 255) 2950 return (EINVAL); 2951 2952 opt->ip6po_tclass = tclass; 2953 break; 2954 } 2955 2956 #ifdef RFC2292 2957 case IPV6_2292NEXTHOP: 2958 #endif 2959 case IPV6_NEXTHOP: 2960 error = kauth_authorize_network(cred, KAUTH_NETWORK_IPV6, 2961 KAUTH_REQ_NETWORK_IPV6_HOPBYHOP, NULL, NULL, NULL); 2962 if (error) 2963 return (error); 2964 2965 if (len == 0) { /* just remove the option */ 2966 ip6_clearpktopts(opt, IPV6_NEXTHOP); 2967 break; 2968 } 2969 2970 /* check if cmsg_len is large enough for sa_len */ 2971 if (len < sizeof(struct sockaddr) || len < *buf) 2972 return (EINVAL); 2973 2974 switch (((struct sockaddr *)buf)->sa_family) { 2975 case AF_INET6: 2976 { 2977 struct sockaddr_in6 *sa6 = (struct sockaddr_in6 *)buf; 2978 2979 if (sa6->sin6_len != sizeof(struct sockaddr_in6)) 2980 return (EINVAL); 2981 2982 if (IN6_IS_ADDR_UNSPECIFIED(&sa6->sin6_addr) || 2983 IN6_IS_ADDR_MULTICAST(&sa6->sin6_addr)) { 2984 return (EINVAL); 2985 } 2986 if ((error = sa6_embedscope(sa6, ip6_use_defzone)) 2987 != 0) { 2988 return (error); 2989 } 2990 break; 2991 } 2992 case AF_LINK: /* eventually be supported? */ 2993 default: 2994 return (EAFNOSUPPORT); 2995 } 2996 2997 /* turn off the previous option, then set the new option. */ 2998 ip6_clearpktopts(opt, IPV6_NEXTHOP); 2999 opt->ip6po_nexthop = malloc(*buf, M_IP6OPT, M_NOWAIT); 3000 if (opt->ip6po_nexthop == NULL) 3001 return (ENOBUFS); 3002 memcpy(opt->ip6po_nexthop, buf, *buf); 3003 break; 3004 3005 #ifdef RFC2292 3006 case IPV6_2292HOPOPTS: 3007 #endif 3008 case IPV6_HOPOPTS: 3009 { 3010 struct ip6_hbh *hbh; 3011 int hbhlen; 3012 3013 /* 3014 * XXX: We don't allow a non-privileged user to set ANY HbH 3015 * options, since per-option restriction has too much 3016 * overhead. 3017 */ 3018 error = kauth_authorize_network(cred, KAUTH_NETWORK_IPV6, 3019 KAUTH_REQ_NETWORK_IPV6_HOPBYHOP, NULL, NULL, NULL); 3020 if (error) 3021 return (error); 3022 3023 if (len == 0) { 3024 ip6_clearpktopts(opt, IPV6_HOPOPTS); 3025 break; /* just remove the option */ 3026 } 3027 3028 /* message length validation */ 3029 if (len < sizeof(struct ip6_hbh)) 3030 return (EINVAL); 3031 hbh = (struct ip6_hbh *)buf; 3032 hbhlen = (hbh->ip6h_len + 1) << 3; 3033 if (len != hbhlen) 3034 return (EINVAL); 3035 3036 /* turn off the previous option, then set the new option. */ 3037 ip6_clearpktopts(opt, IPV6_HOPOPTS); 3038 opt->ip6po_hbh = malloc(hbhlen, M_IP6OPT, M_NOWAIT); 3039 if (opt->ip6po_hbh == NULL) 3040 return (ENOBUFS); 3041 memcpy(opt->ip6po_hbh, hbh, hbhlen); 3042 3043 break; 3044 } 3045 3046 #ifdef RFC2292 3047 case IPV6_2292DSTOPTS: 3048 #endif 3049 case IPV6_DSTOPTS: 3050 case IPV6_RTHDRDSTOPTS: 3051 { 3052 struct ip6_dest *dest, **newdest = NULL; 3053 int destlen; 3054 3055 /* XXX: see the comment for IPV6_HOPOPTS */ 3056 error = kauth_authorize_network(cred, KAUTH_NETWORK_IPV6, 3057 KAUTH_REQ_NETWORK_IPV6_HOPBYHOP, NULL, NULL, NULL); 3058 if (error) 3059 return (error); 3060 3061 if (len == 0) { 3062 ip6_clearpktopts(opt, optname); 3063 break; /* just remove the option */ 3064 } 3065 3066 /* message length validation */ 3067 if (len < sizeof(struct ip6_dest)) 3068 return (EINVAL); 3069 dest = (struct ip6_dest *)buf; 3070 destlen = (dest->ip6d_len + 1) << 3; 3071 if (len != destlen) 3072 return (EINVAL); 3073 /* 3074 * Determine the position that the destination options header 3075 * should be inserted; before or after the routing header. 3076 */ 3077 switch (optname) { 3078 case IPV6_2292DSTOPTS: 3079 /* 3080 * The old advanced API is ambiguous on this point. 3081 * Our approach is to determine the position based 3082 * according to the existence of a routing header. 3083 * Note, however, that this depends on the order of the 3084 * extension headers in the ancillary data; the 1st 3085 * part of the destination options header must appear 3086 * before the routing header in the ancillary data, 3087 * too. 3088 * RFC3542 solved the ambiguity by introducing 3089 * separate ancillary data or option types. 3090 */ 3091 if (opt->ip6po_rthdr == NULL) 3092 newdest = &opt->ip6po_dest1; 3093 else 3094 newdest = &opt->ip6po_dest2; 3095 break; 3096 case IPV6_RTHDRDSTOPTS: 3097 newdest = &opt->ip6po_dest1; 3098 break; 3099 case IPV6_DSTOPTS: 3100 newdest = &opt->ip6po_dest2; 3101 break; 3102 } 3103 3104 /* turn off the previous option, then set the new option. */ 3105 ip6_clearpktopts(opt, optname); 3106 *newdest = malloc(destlen, M_IP6OPT, M_NOWAIT); 3107 if (*newdest == NULL) 3108 return (ENOBUFS); 3109 memcpy(*newdest, dest, destlen); 3110 3111 break; 3112 } 3113 3114 #ifdef RFC2292 3115 case IPV6_2292RTHDR: 3116 #endif 3117 case IPV6_RTHDR: 3118 { 3119 struct ip6_rthdr *rth; 3120 int rthlen; 3121 3122 if (len == 0) { 3123 ip6_clearpktopts(opt, IPV6_RTHDR); 3124 break; /* just remove the option */ 3125 } 3126 3127 /* message length validation */ 3128 if (len < sizeof(struct ip6_rthdr)) 3129 return (EINVAL); 3130 rth = (struct ip6_rthdr *)buf; 3131 rthlen = (rth->ip6r_len + 1) << 3; 3132 if (len != rthlen) 3133 return (EINVAL); 3134 switch (rth->ip6r_type) { 3135 case IPV6_RTHDR_TYPE_0: 3136 /* Dropped, RFC5095. */ 3137 default: 3138 return (EINVAL); /* not supported */ 3139 } 3140 /* turn off the previous option */ 3141 ip6_clearpktopts(opt, IPV6_RTHDR); 3142 opt->ip6po_rthdr = malloc(rthlen, M_IP6OPT, M_NOWAIT); 3143 if (opt->ip6po_rthdr == NULL) 3144 return (ENOBUFS); 3145 memcpy(opt->ip6po_rthdr, rth, rthlen); 3146 break; 3147 } 3148 3149 case IPV6_USE_MIN_MTU: 3150 if (len != sizeof(int)) 3151 return (EINVAL); 3152 minmtupolicy = *(int *)buf; 3153 if (minmtupolicy != IP6PO_MINMTU_MCASTONLY && 3154 minmtupolicy != IP6PO_MINMTU_DISABLE && 3155 minmtupolicy != IP6PO_MINMTU_ALL) { 3156 return (EINVAL); 3157 } 3158 opt->ip6po_minmtu = minmtupolicy; 3159 break; 3160 3161 case IPV6_DONTFRAG: 3162 if (len != sizeof(int)) 3163 return (EINVAL); 3164 3165 if (uproto == IPPROTO_TCP || *(int *)buf == 0) { 3166 /* 3167 * we ignore this option for TCP sockets. 3168 * (RFC3542 leaves this case unspecified.) 3169 */ 3170 opt->ip6po_flags &= ~IP6PO_DONTFRAG; 3171 } else 3172 opt->ip6po_flags |= IP6PO_DONTFRAG; 3173 break; 3174 3175 case IPV6_PREFER_TEMPADDR: 3176 { 3177 int preftemp; 3178 3179 if (len != sizeof(int)) 3180 return (EINVAL); 3181 preftemp = *(int *)buf; 3182 switch (preftemp) { 3183 case IP6PO_TEMPADDR_SYSTEM: 3184 case IP6PO_TEMPADDR_NOTPREFER: 3185 case IP6PO_TEMPADDR_PREFER: 3186 break; 3187 default: 3188 return (EINVAL); 3189 } 3190 opt->ip6po_prefer_tempaddr = preftemp; 3191 break; 3192 } 3193 3194 default: 3195 return (ENOPROTOOPT); 3196 } /* end of switch */ 3197 3198 return (0); 3199 } 3200 3201 /* 3202 * Routine called from ip6_output() to loop back a copy of an IP6 multicast 3203 * packet to the input queue of a specified interface. Note that this 3204 * calls the output routine of the loopback "driver", but with an interface 3205 * pointer that might NOT be lo0ifp -- easier than replicating that code here. 3206 */ 3207 void 3208 ip6_mloopback(struct ifnet *ifp, struct mbuf *m, 3209 const struct sockaddr_in6 *dst) 3210 { 3211 struct mbuf *copym; 3212 struct ip6_hdr *ip6; 3213 3214 copym = m_copypacket(m, M_DONTWAIT); 3215 if (copym == NULL) 3216 return; 3217 3218 /* 3219 * Make sure to deep-copy IPv6 header portion in case the data 3220 * is in an mbuf cluster, so that we can safely override the IPv6 3221 * header portion later. 3222 */ 3223 if ((copym->m_flags & M_EXT) != 0 || 3224 copym->m_len < sizeof(struct ip6_hdr)) { 3225 copym = m_pullup(copym, sizeof(struct ip6_hdr)); 3226 if (copym == NULL) 3227 return; 3228 } 3229 3230 #ifdef DIAGNOSTIC 3231 if (copym->m_len < sizeof(*ip6)) { 3232 m_freem(copym); 3233 return; 3234 } 3235 #endif 3236 3237 ip6 = mtod(copym, struct ip6_hdr *); 3238 /* 3239 * clear embedded scope identifiers if necessary. 3240 * in6_clearscope will touch the addresses only when necessary. 3241 */ 3242 in6_clearscope(&ip6->ip6_src); 3243 in6_clearscope(&ip6->ip6_dst); 3244 3245 (void)looutput(ifp, copym, (const struct sockaddr *)dst, NULL); 3246 } 3247 3248 /* 3249 * Chop IPv6 header off from the payload. 3250 */ 3251 static int 3252 ip6_splithdr(struct mbuf *m, struct ip6_exthdrs *exthdrs) 3253 { 3254 struct mbuf *mh; 3255 struct ip6_hdr *ip6; 3256 3257 ip6 = mtod(m, struct ip6_hdr *); 3258 if (m->m_len > sizeof(*ip6)) { 3259 MGETHDR(mh, M_DONTWAIT, MT_HEADER); 3260 if (mh == NULL) { 3261 m_freem(m); 3262 return ENOBUFS; 3263 } 3264 m_move_pkthdr(mh, m); 3265 m_align(mh, sizeof(*ip6)); 3266 m->m_len -= sizeof(*ip6); 3267 m->m_data += sizeof(*ip6); 3268 mh->m_next = m; 3269 mh->m_len = sizeof(*ip6); 3270 memcpy(mtod(mh, void *), (void *)ip6, sizeof(*ip6)); 3271 m = mh; 3272 } 3273 exthdrs->ip6e_ip6 = m; 3274 return 0; 3275 } 3276 3277 /* 3278 * Compute IPv6 extension header length. 3279 */ 3280 int 3281 ip6_optlen(struct in6pcb *in6p) 3282 { 3283 int len; 3284 3285 if (!in6p->in6p_outputopts) 3286 return 0; 3287 3288 len = 0; 3289 #define elen(x) \ 3290 (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0) 3291 3292 len += elen(in6p->in6p_outputopts->ip6po_hbh); 3293 len += elen(in6p->in6p_outputopts->ip6po_dest1); 3294 len += elen(in6p->in6p_outputopts->ip6po_rthdr); 3295 len += elen(in6p->in6p_outputopts->ip6po_dest2); 3296 return len; 3297 #undef elen 3298 } 3299 3300 /* 3301 * Ensure sending address is valid. 3302 * Returns 0 on success, -1 if an error should be sent back or 1 3303 * if the packet could be dropped without error (protocol dependent). 3304 */ 3305 static int 3306 ip6_ifaddrvalid(const struct in6_addr *src, const struct in6_addr *dst) 3307 { 3308 struct sockaddr_in6 sin6; 3309 int s, error; 3310 struct ifaddr *ifa; 3311 struct in6_ifaddr *ia6; 3312 3313 if (IN6_IS_ADDR_UNSPECIFIED(src)) 3314 return 0; 3315 3316 memset(&sin6, 0, sizeof(sin6)); 3317 sin6.sin6_family = AF_INET6; 3318 sin6.sin6_len = sizeof(sin6); 3319 sin6.sin6_addr = *src; 3320 3321 s = pserialize_read_enter(); 3322 ifa = ifa_ifwithaddr(sin6tosa(&sin6)); 3323 if ((ia6 = ifatoia6(ifa)) == NULL || 3324 ia6->ia6_flags & (IN6_IFF_ANYCAST | IN6_IFF_DUPLICATED)) 3325 error = -1; 3326 else if (ia6->ia6_flags & IN6_IFF_TENTATIVE) 3327 error = 1; 3328 else if (ia6->ia6_flags & IN6_IFF_DETACHED && 3329 (sin6.sin6_addr = *dst, ifa_ifwithaddr(sin6tosa(&sin6)) == NULL)) 3330 /* Allow internal traffic to DETACHED addresses */ 3331 error = 1; 3332 else 3333 error = 0; 3334 pserialize_read_exit(s); 3335 3336 return error; 3337 } 3338
Indexes created Mon Sep 22 13:09:51 GMT 2025