ip6_output.c revision 1.206
1 /* $NetBSD: ip6_output.c,v 1.206 2018/04/26 19:50:09 maxv Exp $ */ 2 /* $KAME: ip6_output.c,v 1.172 2001/03/25 09:55:56 itojun Exp $ */ 3 4 /* 5 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project. 6 * All rights reserved. 7 * 8 * Redistribution and use in source and binary forms, with or without 9 * modification, are permitted provided that the following conditions 10 * are met: 11 * 1. Redistributions of source code must retain the above copyright 12 * notice, this list of conditions and the following disclaimer. 13 * 2. Redistributions in binary form must reproduce the above copyright 14 * notice, this list of conditions and the following disclaimer in the 15 * documentation and/or other materials provided with the distribution. 16 * 3. Neither the name of the project nor the names of its contributors 17 * may be used to endorse or promote products derived from this software 18 * without specific prior written permission. 19 * 20 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND 21 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 22 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 23 * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE 24 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 25 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 26 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 27 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 28 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 29 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 30 * SUCH DAMAGE. 31 */ 32 33 /* 34 * Copyright (c) 1982, 1986, 1988, 1990, 1993 35 * The Regents of the University of California. All rights reserved. 36 * 37 * Redistribution and use in source and binary forms, with or without 38 * modification, are permitted provided that the following conditions 39 * are met: 40 * 1. Redistributions of source code must retain the above copyright 41 * notice, this list of conditions and the following disclaimer. 42 * 2. Redistributions in binary form must reproduce the above copyright 43 * notice, this list of conditions and the following disclaimer in the 44 * documentation and/or other materials provided with the distribution. 45 * 3. Neither the name of the University nor the names of its contributors 46 * may be used to endorse or promote products derived from this software 47 * without specific prior written permission. 48 * 49 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 50 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 51 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 52 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 53 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 54 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 55 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 56 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 57 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 58 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 59 * SUCH DAMAGE. 60 * 61 * @(#)ip_output.c 8.3 (Berkeley) 1/21/94 62 */ 63 64 #include <sys/cdefs.h> 65 __KERNEL_RCSID(0, "$NetBSD: ip6_output.c,v 1.206 2018/04/26 19:50:09 maxv 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 108 109 110 #include <net/net_osdep.h> 111 112 extern pfil_head_t *inet6_pfil_hook; /* XXX */ 113 114 struct ip6_exthdrs { 115 struct mbuf *ip6e_ip6; 116 struct mbuf *ip6e_hbh; 117 struct mbuf *ip6e_dest1; 118 struct mbuf *ip6e_rthdr; 119 struct mbuf *ip6e_dest2; 120 }; 121 122 static int ip6_pcbopt(int, u_char *, int, struct ip6_pktopts **, 123 kauth_cred_t, int); 124 static int ip6_getpcbopt(struct ip6_pktopts *, int, struct sockopt *); 125 static int ip6_setpktopt(int, u_char *, int, struct ip6_pktopts *, kauth_cred_t, 126 int, int, int); 127 static int ip6_setmoptions(const struct sockopt *, struct in6pcb *); 128 static int ip6_getmoptions(struct sockopt *, struct in6pcb *); 129 static int ip6_copyexthdr(struct mbuf **, void *, int); 130 static int ip6_insertfraghdr(struct mbuf *, struct mbuf *, int, 131 struct ip6_frag **); 132 static int ip6_insert_jumboopt(struct ip6_exthdrs *, u_int32_t); 133 static int ip6_splithdr(struct mbuf *, struct ip6_exthdrs *); 134 static int ip6_getpmtu(struct rtentry *, struct ifnet *, u_long *, int *); 135 static int copypktopts(struct ip6_pktopts *, struct ip6_pktopts *, int); 136 static int ip6_ifaddrvalid(const struct in6_addr *, const struct in6_addr *); 137 static int ip6_handle_rthdr(struct ip6_rthdr *, struct ip6_hdr *); 138 139 #ifdef RFC2292 140 static int ip6_pcbopts(struct ip6_pktopts **, struct socket *, struct sockopt *); 141 #endif 142 143 static int 144 ip6_handle_rthdr(struct ip6_rthdr *rh, struct ip6_hdr *ip6) 145 { 146 int error = 0; 147 148 switch (rh->ip6r_type) { 149 case IPV6_RTHDR_TYPE_0: 150 /* Dropped, RFC5095. */ 151 default: /* is it possible? */ 152 error = EINVAL; 153 } 154 155 return error; 156 } 157 158 /* 159 * Send an IP packet to a host. 160 */ 161 int 162 ip6_if_output(struct ifnet * const ifp, struct ifnet * const origifp, 163 struct mbuf * const m, const struct sockaddr_in6 * const dst, 164 const struct rtentry *rt) 165 { 166 int error = 0; 167 168 if (rt != NULL) { 169 error = rt_check_reject_route(rt, ifp); 170 if (error != 0) { 171 m_freem(m); 172 return error; 173 } 174 } 175 176 if ((ifp->if_flags & IFF_LOOPBACK) != 0) 177 error = if_output_lock(ifp, origifp, m, sin6tocsa(dst), rt); 178 else 179 error = if_output_lock(ifp, ifp, m, sin6tocsa(dst), rt); 180 return error; 181 } 182 183 /* 184 * IP6 output. The packet in mbuf chain m contains a skeletal IP6 185 * header (with pri, len, nxt, hlim, src, dst). 186 * 187 * This function may modify ver and hlim only. The mbuf chain containing the 188 * packet will be freed. The mbuf opt, if present, will not be freed. 189 * 190 * Type of "mtu": rt_rmx.rmx_mtu is u_long, ifnet.ifr_mtu is int, and 191 * nd_ifinfo.linkmtu is u_int32_t. So we use u_long to hold largest one, 192 * which is rt_rmx.rmx_mtu. 193 */ 194 int 195 ip6_output( 196 struct mbuf *m0, 197 struct ip6_pktopts *opt, 198 struct route *ro, 199 int flags, 200 struct ip6_moptions *im6o, 201 struct in6pcb *in6p, 202 struct ifnet **ifpp /* XXX: just for statistics */ 203 ) 204 { 205 struct ip6_hdr *ip6, *mhip6; 206 struct ifnet *ifp = NULL, *origifp = NULL; 207 struct mbuf *m = m0; 208 int tlen, len, off; 209 bool tso; 210 struct route ip6route; 211 struct rtentry *rt = NULL, *rt_pmtu; 212 const struct sockaddr_in6 *dst; 213 struct sockaddr_in6 src_sa, dst_sa; 214 int error = 0; 215 struct in6_ifaddr *ia = NULL; 216 u_long mtu; 217 int alwaysfrag, dontfrag; 218 u_int32_t optlen = 0, plen = 0, unfragpartlen = 0; 219 struct ip6_exthdrs exthdrs; 220 struct in6_addr finaldst, src0, dst0; 221 u_int32_t zone; 222 struct route *ro_pmtu = NULL; 223 int hdrsplit = 0; 224 int needipsec = 0; 225 #ifdef IPSEC 226 struct secpolicy *sp = NULL; 227 #endif 228 struct psref psref, psref_ia; 229 int bound = curlwp_bind(); 230 bool release_psref_ia = false; 231 232 #ifdef DIAGNOSTIC 233 if ((m->m_flags & M_PKTHDR) == 0) 234 panic("ip6_output: no HDR"); 235 if ((m->m_pkthdr.csum_flags & 236 (M_CSUM_TCPv4|M_CSUM_UDPv4|M_CSUM_TSOv4)) != 0) { 237 panic("ip6_output: IPv4 checksum offload flags: %d", 238 m->m_pkthdr.csum_flags); 239 } 240 if ((m->m_pkthdr.csum_flags & (M_CSUM_TCPv6|M_CSUM_UDPv6)) == 241 (M_CSUM_TCPv6|M_CSUM_UDPv6)) { 242 panic("ip6_output: conflicting checksum offload flags: %d", 243 m->m_pkthdr.csum_flags); 244 } 245 #endif 246 247 M_CSUM_DATA_IPv6_HL_SET(m->m_pkthdr.csum_data, sizeof(struct ip6_hdr)); 248 249 #define MAKE_EXTHDR(hp, mp) \ 250 do { \ 251 if (hp) { \ 252 struct ip6_ext *eh = (struct ip6_ext *)(hp); \ 253 error = ip6_copyexthdr((mp), (void *)(hp), \ 254 ((eh)->ip6e_len + 1) << 3); \ 255 if (error) \ 256 goto freehdrs; \ 257 } \ 258 } while (/*CONSTCOND*/ 0) 259 260 memset(&exthdrs, 0, sizeof(exthdrs)); 261 if (opt) { 262 /* Hop-by-Hop options header */ 263 MAKE_EXTHDR(opt->ip6po_hbh, &exthdrs.ip6e_hbh); 264 /* Destination options header (1st part) */ 265 MAKE_EXTHDR(opt->ip6po_dest1, &exthdrs.ip6e_dest1); 266 /* Routing header */ 267 MAKE_EXTHDR(opt->ip6po_rthdr, &exthdrs.ip6e_rthdr); 268 /* Destination options header (2nd part) */ 269 MAKE_EXTHDR(opt->ip6po_dest2, &exthdrs.ip6e_dest2); 270 } 271 272 /* 273 * Calculate the total length of the extension header chain. 274 * Keep the length of the unfragmentable part for fragmentation. 275 */ 276 optlen = 0; 277 if (exthdrs.ip6e_hbh) 278 optlen += exthdrs.ip6e_hbh->m_len; 279 if (exthdrs.ip6e_dest1) 280 optlen += exthdrs.ip6e_dest1->m_len; 281 if (exthdrs.ip6e_rthdr) 282 optlen += exthdrs.ip6e_rthdr->m_len; 283 unfragpartlen = optlen + sizeof(struct ip6_hdr); 284 /* NOTE: we don't add AH/ESP length here. do that later. */ 285 if (exthdrs.ip6e_dest2) 286 optlen += exthdrs.ip6e_dest2->m_len; 287 288 #ifdef IPSEC 289 if (ipsec_used) { 290 /* Check the security policy (SP) for the packet */ 291 sp = ipsec6_check_policy(m, in6p, flags, &needipsec, &error); 292 if (error != 0) { 293 /* 294 * Hack: -EINVAL is used to signal that a packet 295 * should be silently discarded. This is typically 296 * because we asked key management for an SA and 297 * it was delayed (e.g. kicked up to IKE). 298 */ 299 if (error == -EINVAL) 300 error = 0; 301 goto freehdrs; 302 } 303 } 304 #endif 305 306 if (needipsec && 307 (m->m_pkthdr.csum_flags & (M_CSUM_UDPv6|M_CSUM_TCPv6)) != 0) { 308 in6_delayed_cksum(m); 309 m->m_pkthdr.csum_flags &= ~(M_CSUM_UDPv6|M_CSUM_TCPv6); 310 } 311 312 /* 313 * If we need IPsec, or there is at least one extension header, 314 * separate IP6 header from the payload. 315 */ 316 if ((needipsec || optlen) && !hdrsplit) { 317 if ((error = ip6_splithdr(m, &exthdrs)) != 0) { 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 m = NULL; 337 goto freehdrs; 338 } 339 m = exthdrs.ip6e_ip6; 340 hdrsplit++; 341 } 342 /* adjust pointer */ 343 ip6 = mtod(m, struct ip6_hdr *); 344 if ((error = ip6_insert_jumboopt(&exthdrs, plen)) != 0) 345 goto freehdrs; 346 optlen += 8; /* XXX JUMBOOPTLEN */ 347 ip6->ip6_plen = 0; 348 } else 349 ip6->ip6_plen = htons(plen); 350 351 /* 352 * Concatenate headers and fill in next header fields. 353 * Here we have, on "m" 354 * IPv6 payload 355 * and we insert headers accordingly. Finally, we should be getting: 356 * IPv6 hbh dest1 rthdr ah* [esp* dest2 payload] 357 * 358 * during the header composing process, "m" points to IPv6 header. 359 * "mprev" points to an extension header prior to esp. 360 */ 361 { 362 u_char *nexthdrp = &ip6->ip6_nxt; 363 struct mbuf *mprev = m; 364 365 /* 366 * we treat dest2 specially. this makes IPsec processing 367 * much easier. the goal here is to make mprev point the 368 * mbuf prior to dest2. 369 * 370 * result: IPv6 dest2 payload 371 * m and mprev will point to IPv6 header. 372 */ 373 if (exthdrs.ip6e_dest2) { 374 if (!hdrsplit) 375 panic("assumption failed: hdr not split"); 376 exthdrs.ip6e_dest2->m_next = m->m_next; 377 m->m_next = exthdrs.ip6e_dest2; 378 *mtod(exthdrs.ip6e_dest2, u_char *) = ip6->ip6_nxt; 379 ip6->ip6_nxt = IPPROTO_DSTOPTS; 380 } 381 382 #define MAKE_CHAIN(m, mp, p, i)\ 383 do {\ 384 if (m) {\ 385 if (!hdrsplit) \ 386 panic("assumption failed: hdr not split"); \ 387 *mtod((m), u_char *) = *(p);\ 388 *(p) = (i);\ 389 p = mtod((m), u_char *);\ 390 (m)->m_next = (mp)->m_next;\ 391 (mp)->m_next = (m);\ 392 (mp) = (m);\ 393 }\ 394 } while (/*CONSTCOND*/ 0) 395 /* 396 * result: IPv6 hbh dest1 rthdr dest2 payload 397 * m will point to IPv6 header. mprev will point to the 398 * extension header prior to dest2 (rthdr in the above case). 399 */ 400 MAKE_CHAIN(exthdrs.ip6e_hbh, mprev, nexthdrp, IPPROTO_HOPOPTS); 401 MAKE_CHAIN(exthdrs.ip6e_dest1, mprev, nexthdrp, 402 IPPROTO_DSTOPTS); 403 MAKE_CHAIN(exthdrs.ip6e_rthdr, mprev, nexthdrp, 404 IPPROTO_ROUTING); 405 406 M_CSUM_DATA_IPv6_HL_SET(m->m_pkthdr.csum_data, 407 sizeof(struct ip6_hdr) + optlen); 408 } 409 410 /* Need to save for pmtu */ 411 finaldst = ip6->ip6_dst; 412 413 /* 414 * If there is a routing header, replace destination address field 415 * with the first hop of the routing header. 416 */ 417 if (exthdrs.ip6e_rthdr) { 418 struct ip6_rthdr *rh; 419 420 rh = mtod(exthdrs.ip6e_rthdr, struct ip6_rthdr *); 421 422 error = ip6_handle_rthdr(rh, ip6); 423 if (error != 0) 424 goto bad; 425 } 426 427 /* Source address validation */ 428 if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) && 429 (flags & IPV6_UNSPECSRC) == 0) { 430 error = EOPNOTSUPP; 431 IP6_STATINC(IP6_STAT_BADSCOPE); 432 goto bad; 433 } 434 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) { 435 error = EOPNOTSUPP; 436 IP6_STATINC(IP6_STAT_BADSCOPE); 437 goto bad; 438 } 439 440 IP6_STATINC(IP6_STAT_LOCALOUT); 441 442 /* 443 * Route packet. 444 */ 445 /* initialize cached route */ 446 if (ro == NULL) { 447 memset(&ip6route, 0, sizeof(ip6route)); 448 ro = &ip6route; 449 } 450 ro_pmtu = ro; 451 if (opt && opt->ip6po_rthdr) 452 ro = &opt->ip6po_route; 453 454 /* 455 * if specified, try to fill in the traffic class field. 456 * do not override if a non-zero value is already set. 457 * we check the diffserv field and the ecn field separately. 458 */ 459 if (opt && opt->ip6po_tclass >= 0) { 460 int mask = 0; 461 462 if ((ip6->ip6_flow & htonl(0xfc << 20)) == 0) 463 mask |= 0xfc; 464 if ((ip6->ip6_flow & htonl(0x03 << 20)) == 0) 465 mask |= 0x03; 466 if (mask != 0) 467 ip6->ip6_flow |= htonl((opt->ip6po_tclass & mask) << 20); 468 } 469 470 /* fill in or override the hop limit field, if necessary. */ 471 if (opt && opt->ip6po_hlim != -1) 472 ip6->ip6_hlim = opt->ip6po_hlim & 0xff; 473 else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) { 474 if (im6o != NULL) 475 ip6->ip6_hlim = im6o->im6o_multicast_hlim; 476 else 477 ip6->ip6_hlim = ip6_defmcasthlim; 478 } 479 480 #ifdef IPSEC 481 if (needipsec) { 482 int s = splsoftnet(); 483 error = ipsec6_process_packet(m, sp->req); 484 splx(s); 485 486 /* 487 * Preserve KAME behaviour: ENOENT can be returned 488 * when an SA acquire is in progress. Don't propagate 489 * this to user-level; it confuses applications. 490 * XXX this will go away when the SADB is redone. 491 */ 492 if (error == ENOENT) 493 error = 0; 494 495 goto done; 496 } 497 #endif 498 499 /* adjust pointer */ 500 ip6 = mtod(m, struct ip6_hdr *); 501 502 sockaddr_in6_init(&dst_sa, &ip6->ip6_dst, 0, 0, 0); 503 504 /* We do not need a route for multicast */ 505 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) { 506 struct in6_pktinfo *pi = NULL; 507 508 /* 509 * If the outgoing interface for the address is specified by 510 * the caller, use it. 511 */ 512 if (opt && (pi = opt->ip6po_pktinfo) != NULL) { 513 /* XXX boundary check is assumed to be already done. */ 514 ifp = if_get_byindex(pi->ipi6_ifindex, &psref); 515 } else if (im6o != NULL) { 516 ifp = if_get_byindex(im6o->im6o_multicast_if_index, 517 &psref); 518 } 519 } 520 521 if (ifp == NULL) { 522 error = in6_selectroute(&dst_sa, opt, &ro, &rt, true); 523 if (error != 0) 524 goto bad; 525 ifp = if_get_byindex(rt->rt_ifp->if_index, &psref); 526 } 527 528 if (rt == NULL) { 529 /* 530 * If in6_selectroute() does not return a route entry, 531 * dst may not have been updated. 532 */ 533 error = rtcache_setdst(ro, sin6tosa(&dst_sa)); 534 if (error) { 535 goto bad; 536 } 537 } 538 539 /* 540 * then rt (for unicast) and ifp must be non-NULL valid values. 541 */ 542 if ((flags & IPV6_FORWARDING) == 0) { 543 /* XXX: the FORWARDING flag can be set for mrouting. */ 544 in6_ifstat_inc(ifp, ifs6_out_request); 545 } 546 if (rt != NULL) { 547 ia = (struct in6_ifaddr *)(rt->rt_ifa); 548 rt->rt_use++; 549 } 550 551 /* 552 * The outgoing interface must be in the zone of source and 553 * destination addresses. We should use ia_ifp to support the 554 * case of sending packets to an address of our own. 555 */ 556 if (ia != NULL && ia->ia_ifp) { 557 origifp = ia->ia_ifp; 558 if (if_is_deactivated(origifp)) 559 goto bad; 560 if_acquire(origifp, &psref_ia); 561 release_psref_ia = true; 562 } else 563 origifp = ifp; 564 565 src0 = ip6->ip6_src; 566 if (in6_setscope(&src0, origifp, &zone)) 567 goto badscope; 568 sockaddr_in6_init(&src_sa, &ip6->ip6_src, 0, 0, 0); 569 if (sa6_recoverscope(&src_sa) || zone != src_sa.sin6_scope_id) 570 goto badscope; 571 572 dst0 = ip6->ip6_dst; 573 if (in6_setscope(&dst0, origifp, &zone)) 574 goto badscope; 575 /* re-initialize to be sure */ 576 sockaddr_in6_init(&dst_sa, &ip6->ip6_dst, 0, 0, 0); 577 if (sa6_recoverscope(&dst_sa) || zone != dst_sa.sin6_scope_id) 578 goto badscope; 579 580 /* scope check is done. */ 581 582 /* Ensure we only send from a valid address. */ 583 if ((ifp->if_flags & IFF_LOOPBACK) == 0 && 584 (error = ip6_ifaddrvalid(&src0, &dst0)) != 0) 585 { 586 char ip6buf[INET6_ADDRSTRLEN]; 587 nd6log(LOG_ERR, 588 "refusing to send from invalid address %s (pid %d)\n", 589 IN6_PRINT(ip6buf, &src0), curproc->p_pid); 590 IP6_STATINC(IP6_STAT_ODROPPED); 591 in6_ifstat_inc(origifp, ifs6_out_discard); 592 if (error == 1) 593 /* 594 * Address exists, but is tentative or detached. 595 * We can't send from it because it's invalid, 596 * so we drop the packet. 597 */ 598 error = 0; 599 else 600 error = EADDRNOTAVAIL; 601 goto bad; 602 } 603 604 if (rt != NULL && (rt->rt_flags & RTF_GATEWAY) && 605 !IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) 606 dst = satocsin6(rt->rt_gateway); 607 else 608 dst = satocsin6(rtcache_getdst(ro)); 609 610 /* 611 * XXXXXX: original code follows: 612 */ 613 if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) 614 m->m_flags &= ~(M_BCAST | M_MCAST); /* just in case */ 615 else { 616 bool ingroup; 617 618 m->m_flags = (m->m_flags & ~M_BCAST) | M_MCAST; 619 620 in6_ifstat_inc(ifp, ifs6_out_mcast); 621 622 /* 623 * Confirm that the outgoing interface supports multicast. 624 */ 625 if (!(ifp->if_flags & IFF_MULTICAST)) { 626 IP6_STATINC(IP6_STAT_NOROUTE); 627 in6_ifstat_inc(ifp, ifs6_out_discard); 628 error = ENETUNREACH; 629 goto bad; 630 } 631 632 ingroup = in6_multi_group(&ip6->ip6_dst, ifp); 633 if (ingroup && (im6o == NULL || im6o->im6o_multicast_loop)) { 634 /* 635 * If we belong to the destination multicast group 636 * on the outgoing interface, and the caller did not 637 * forbid loopback, loop back a copy. 638 */ 639 KASSERT(dst != NULL); 640 ip6_mloopback(ifp, m, dst); 641 } else { 642 /* 643 * If we are acting as a multicast router, perform 644 * multicast forwarding as if the packet had just 645 * arrived on the interface to which we are about 646 * to send. The multicast forwarding function 647 * recursively calls this function, using the 648 * IPV6_FORWARDING flag to prevent infinite recursion. 649 * 650 * Multicasts that are looped back by ip6_mloopback(), 651 * above, will be forwarded by the ip6_input() routine, 652 * if necessary. 653 */ 654 if (ip6_mrouter && (flags & IPV6_FORWARDING) == 0) { 655 if (ip6_mforward(ip6, ifp, m) != 0) { 656 m_freem(m); 657 goto done; 658 } 659 } 660 } 661 /* 662 * Multicasts with a hoplimit of zero may be looped back, 663 * above, but must not be transmitted on a network. 664 * Also, multicasts addressed to the loopback interface 665 * are not sent -- the above call to ip6_mloopback() will 666 * loop back a copy if this host actually belongs to the 667 * destination group on the loopback interface. 668 */ 669 if (ip6->ip6_hlim == 0 || (ifp->if_flags & IFF_LOOPBACK) || 670 IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst)) { 671 m_freem(m); 672 goto done; 673 } 674 } 675 676 /* 677 * Fill the outgoing inteface to tell the upper layer 678 * to increment per-interface statistics. 679 */ 680 if (ifpp) 681 *ifpp = ifp; 682 683 /* Determine path MTU. */ 684 /* 685 * ro_pmtu represent final destination while 686 * ro might represent immediate destination. 687 * Use ro_pmtu destination since MTU might differ. 688 */ 689 if (ro_pmtu != ro) { 690 union { 691 struct sockaddr dst; 692 struct sockaddr_in6 dst6; 693 } u; 694 695 /* ro_pmtu may not have a cache */ 696 sockaddr_in6_init(&u.dst6, &finaldst, 0, 0, 0); 697 rt_pmtu = rtcache_lookup(ro_pmtu, &u.dst); 698 } else 699 rt_pmtu = rt; 700 error = ip6_getpmtu(rt_pmtu, ifp, &mtu, &alwaysfrag); 701 if (rt_pmtu != NULL && rt_pmtu != rt) 702 rtcache_unref(rt_pmtu, ro_pmtu); 703 if (error != 0) 704 goto bad; 705 706 /* 707 * The caller of this function may specify to use the minimum MTU 708 * in some cases. 709 * An advanced API option (IPV6_USE_MIN_MTU) can also override MTU 710 * setting. The logic is a bit complicated; by default, unicast 711 * packets will follow path MTU while multicast packets will be sent at 712 * the minimum MTU. If IP6PO_MINMTU_ALL is specified, all packets 713 * including unicast ones will be sent at the minimum MTU. Multicast 714 * packets will always be sent at the minimum MTU unless 715 * IP6PO_MINMTU_DISABLE is explicitly specified. 716 * See RFC 3542 for more details. 717 */ 718 if (mtu > IPV6_MMTU) { 719 if ((flags & IPV6_MINMTU)) 720 mtu = IPV6_MMTU; 721 else if (opt && opt->ip6po_minmtu == IP6PO_MINMTU_ALL) 722 mtu = IPV6_MMTU; 723 else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) && 724 (opt == NULL || 725 opt->ip6po_minmtu != IP6PO_MINMTU_DISABLE)) { 726 mtu = IPV6_MMTU; 727 } 728 } 729 730 /* 731 * clear embedded scope identifiers if necessary. 732 * in6_clearscope will touch the addresses only when necessary. 733 */ 734 in6_clearscope(&ip6->ip6_src); 735 in6_clearscope(&ip6->ip6_dst); 736 737 /* 738 * If the outgoing packet contains a hop-by-hop options header, 739 * it must be examined and processed even by the source node. 740 * (RFC 2460, section 4.) 741 * 742 * XXX Is this really necessary? 743 */ 744 if (ip6->ip6_nxt == IPPROTO_HOPOPTS) { 745 u_int32_t dummy1; /* XXX unused */ 746 u_int32_t dummy2; /* XXX unused */ 747 int hoff = sizeof(struct ip6_hdr); 748 749 if (ip6_hopopts_input(&dummy1, &dummy2, &m, &hoff)) { 750 /* m was already freed at this point */ 751 error = EINVAL; 752 goto done; 753 } 754 755 ip6 = mtod(m, struct ip6_hdr *); 756 } 757 758 /* 759 * Run through list of hooks for output packets. 760 */ 761 if ((error = pfil_run_hooks(inet6_pfil_hook, &m, ifp, PFIL_OUT)) != 0) 762 goto done; 763 if (m == NULL) 764 goto done; 765 ip6 = mtod(m, struct ip6_hdr *); 766 767 /* 768 * Send the packet to the outgoing interface. 769 * If necessary, do IPv6 fragmentation before sending. 770 * 771 * the logic here is rather complex: 772 * 1: normal case (dontfrag == 0, alwaysfrag == 0) 773 * 1-a: send as is if tlen <= path mtu 774 * 1-b: fragment if tlen > path mtu 775 * 776 * 2: if user asks us not to fragment (dontfrag == 1) 777 * 2-a: send as is if tlen <= interface mtu 778 * 2-b: error if tlen > interface mtu 779 * 780 * 3: if we always need to attach fragment header (alwaysfrag == 1) 781 * always fragment 782 * 783 * 4: if dontfrag == 1 && alwaysfrag == 1 784 * error, as we cannot handle this conflicting request 785 */ 786 tlen = m->m_pkthdr.len; 787 tso = (m->m_pkthdr.csum_flags & M_CSUM_TSOv6) != 0; 788 if (opt && (opt->ip6po_flags & IP6PO_DONTFRAG)) 789 dontfrag = 1; 790 else 791 dontfrag = 0; 792 793 if (dontfrag && alwaysfrag) { /* case 4 */ 794 /* conflicting request - can't transmit */ 795 error = EMSGSIZE; 796 goto bad; 797 } 798 if (dontfrag && (!tso && tlen > IN6_LINKMTU(ifp))) { /* case 2-b */ 799 /* 800 * Even if the DONTFRAG option is specified, we cannot send the 801 * packet when the data length is larger than the MTU of the 802 * outgoing interface. 803 * Notify the error by sending IPV6_PATHMTU ancillary data as 804 * well as returning an error code (the latter is not described 805 * in the API spec.) 806 */ 807 u_int32_t mtu32; 808 struct ip6ctlparam ip6cp; 809 810 mtu32 = (u_int32_t)mtu; 811 memset(&ip6cp, 0, sizeof(ip6cp)); 812 ip6cp.ip6c_cmdarg = (void *)&mtu32; 813 pfctlinput2(PRC_MSGSIZE, 814 rtcache_getdst(ro_pmtu), &ip6cp); 815 816 error = EMSGSIZE; 817 goto bad; 818 } 819 820 /* 821 * transmit packet without fragmentation 822 */ 823 if (dontfrag || (!alwaysfrag && (tlen <= mtu || tso))) { 824 /* case 1-a and 2-a */ 825 struct in6_ifaddr *ia6; 826 int sw_csum; 827 int s; 828 829 ip6 = mtod(m, struct ip6_hdr *); 830 s = pserialize_read_enter(); 831 ia6 = in6_ifawithifp(ifp, &ip6->ip6_src); 832 if (ia6) { 833 /* Record statistics for this interface address. */ 834 ia6->ia_ifa.ifa_data.ifad_outbytes += m->m_pkthdr.len; 835 } 836 pserialize_read_exit(s); 837 838 sw_csum = m->m_pkthdr.csum_flags & ~ifp->if_csum_flags_tx; 839 if ((sw_csum & (M_CSUM_UDPv6|M_CSUM_TCPv6)) != 0) { 840 if (IN6_NEED_CHECKSUM(ifp, 841 sw_csum & (M_CSUM_UDPv6|M_CSUM_TCPv6))) { 842 in6_delayed_cksum(m); 843 } 844 m->m_pkthdr.csum_flags &= ~(M_CSUM_UDPv6|M_CSUM_TCPv6); 845 } 846 847 KASSERT(dst != NULL); 848 if (__predict_true(!tso || 849 (ifp->if_capenable & IFCAP_TSOv6) != 0)) { 850 error = ip6_if_output(ifp, origifp, m, dst, rt); 851 } else { 852 error = ip6_tso_output(ifp, origifp, m, dst, rt); 853 } 854 goto done; 855 } 856 857 if (tso) { 858 error = EINVAL; /* XXX */ 859 goto bad; 860 } 861 862 /* 863 * try to fragment the packet. case 1-b and 3 864 */ 865 if (mtu < IPV6_MMTU) { 866 /* path MTU cannot be less than IPV6_MMTU */ 867 error = EMSGSIZE; 868 in6_ifstat_inc(ifp, ifs6_out_fragfail); 869 goto bad; 870 } else if (ip6->ip6_plen == 0) { 871 /* jumbo payload cannot be fragmented */ 872 error = EMSGSIZE; 873 in6_ifstat_inc(ifp, ifs6_out_fragfail); 874 goto bad; 875 } else { 876 const u_int32_t id = htonl(ip6_randomid()); 877 struct mbuf **mnext, *m_frgpart; 878 const int hlen = unfragpartlen; 879 struct ip6_frag *ip6f; 880 u_char nextproto; 881 #if 0 /* see below */ 882 struct ip6ctlparam ip6cp; 883 u_int32_t mtu32; 884 #endif 885 886 if (mtu > IPV6_MAXPACKET) 887 mtu = IPV6_MAXPACKET; 888 889 #if 0 890 /* 891 * It is believed this code is a leftover from the 892 * development of the IPV6_RECVPATHMTU sockopt and 893 * associated work to implement RFC3542. 894 * It's not entirely clear what the intent of the API 895 * is at this point, so disable this code for now. 896 * The IPV6_RECVPATHMTU sockopt and/or IPV6_DONTFRAG 897 * will send notifications if the application requests. 898 */ 899 900 /* Notify a proper path MTU to applications. */ 901 mtu32 = (u_int32_t)mtu; 902 memset(&ip6cp, 0, sizeof(ip6cp)); 903 ip6cp.ip6c_cmdarg = (void *)&mtu32; 904 pfctlinput2(PRC_MSGSIZE, 905 rtcache_getdst(ro_pmtu), &ip6cp); 906 #endif 907 908 /* 909 * Must be able to put at least 8 bytes per fragment. 910 */ 911 len = (mtu - hlen - sizeof(struct ip6_frag)) & ~7; 912 if (len < 8) { 913 error = EMSGSIZE; 914 in6_ifstat_inc(ifp, ifs6_out_fragfail); 915 goto bad; 916 } 917 918 mnext = &m->m_nextpkt; 919 920 /* 921 * Change the next header field of the last header in the 922 * unfragmentable part. 923 */ 924 if (exthdrs.ip6e_rthdr) { 925 nextproto = *mtod(exthdrs.ip6e_rthdr, u_char *); 926 *mtod(exthdrs.ip6e_rthdr, u_char *) = IPPROTO_FRAGMENT; 927 } else if (exthdrs.ip6e_dest1) { 928 nextproto = *mtod(exthdrs.ip6e_dest1, u_char *); 929 *mtod(exthdrs.ip6e_dest1, u_char *) = IPPROTO_FRAGMENT; 930 } else if (exthdrs.ip6e_hbh) { 931 nextproto = *mtod(exthdrs.ip6e_hbh, u_char *); 932 *mtod(exthdrs.ip6e_hbh, u_char *) = IPPROTO_FRAGMENT; 933 } else { 934 nextproto = ip6->ip6_nxt; 935 ip6->ip6_nxt = IPPROTO_FRAGMENT; 936 } 937 938 if ((m->m_pkthdr.csum_flags & (M_CSUM_UDPv6|M_CSUM_TCPv6)) 939 != 0) { 940 if (IN6_NEED_CHECKSUM(ifp, 941 m->m_pkthdr.csum_flags & 942 (M_CSUM_UDPv6|M_CSUM_TCPv6))) { 943 in6_delayed_cksum(m); 944 } 945 m->m_pkthdr.csum_flags &= ~(M_CSUM_UDPv6|M_CSUM_TCPv6); 946 } 947 948 /* 949 * Loop through length of segment after first fragment, 950 * make new header and copy data of each part and link onto 951 * chain. 952 */ 953 m0 = m; 954 for (off = hlen; off < tlen; off += len) { 955 struct mbuf *mlast; 956 957 MGETHDR(m, M_DONTWAIT, MT_HEADER); 958 if (!m) { 959 error = ENOBUFS; 960 IP6_STATINC(IP6_STAT_ODROPPED); 961 goto sendorfree; 962 } 963 m_reset_rcvif(m); 964 m->m_flags = m0->m_flags & M_COPYFLAGS; 965 *mnext = m; 966 mnext = &m->m_nextpkt; 967 m->m_data += max_linkhdr; 968 mhip6 = mtod(m, struct ip6_hdr *); 969 *mhip6 = *ip6; 970 m->m_len = sizeof(*mhip6); 971 972 ip6f = NULL; 973 error = ip6_insertfraghdr(m0, m, hlen, &ip6f); 974 if (error) { 975 IP6_STATINC(IP6_STAT_ODROPPED); 976 goto sendorfree; 977 } 978 979 /* Fill in the Frag6 Header */ 980 ip6f->ip6f_offlg = htons((u_int16_t)((off - hlen) & ~7)); 981 if (off + len >= tlen) 982 len = tlen - off; 983 else 984 ip6f->ip6f_offlg |= IP6F_MORE_FRAG; 985 ip6f->ip6f_reserved = 0; 986 ip6f->ip6f_ident = id; 987 ip6f->ip6f_nxt = nextproto; 988 989 mhip6->ip6_plen = htons((u_int16_t)(len + hlen + 990 sizeof(*ip6f) - sizeof(struct ip6_hdr))); 991 if ((m_frgpart = m_copym(m0, off, len, M_DONTWAIT)) == NULL) { 992 error = ENOBUFS; 993 IP6_STATINC(IP6_STAT_ODROPPED); 994 goto sendorfree; 995 } 996 for (mlast = m; mlast->m_next; mlast = mlast->m_next) 997 ; 998 mlast->m_next = m_frgpart; 999 1000 m->m_pkthdr.len = len + hlen + sizeof(*ip6f); 1001 m_reset_rcvif(m); 1002 IP6_STATINC(IP6_STAT_OFRAGMENTS); 1003 in6_ifstat_inc(ifp, ifs6_out_fragcreat); 1004 } 1005 1006 in6_ifstat_inc(ifp, ifs6_out_fragok); 1007 } 1008 1009 sendorfree: 1010 m = m0->m_nextpkt; 1011 m0->m_nextpkt = 0; 1012 m_freem(m0); 1013 for (m0 = m; m; m = m0) { 1014 m0 = m->m_nextpkt; 1015 m->m_nextpkt = 0; 1016 if (error == 0) { 1017 struct in6_ifaddr *ia6; 1018 int s; 1019 ip6 = mtod(m, struct ip6_hdr *); 1020 s = pserialize_read_enter(); 1021 ia6 = in6_ifawithifp(ifp, &ip6->ip6_src); 1022 if (ia6) { 1023 /* 1024 * Record statistics for this interface 1025 * address. 1026 */ 1027 ia6->ia_ifa.ifa_data.ifad_outbytes += 1028 m->m_pkthdr.len; 1029 } 1030 pserialize_read_exit(s); 1031 KASSERT(dst != NULL); 1032 error = ip6_if_output(ifp, origifp, m, dst, rt); 1033 } else 1034 m_freem(m); 1035 } 1036 1037 if (error == 0) 1038 IP6_STATINC(IP6_STAT_FRAGMENTED); 1039 1040 done: 1041 rtcache_unref(rt, ro); 1042 if (ro == &ip6route) 1043 rtcache_free(&ip6route); 1044 #ifdef IPSEC 1045 if (sp != NULL) 1046 KEY_SP_UNREF(&sp); 1047 #endif 1048 if_put(ifp, &psref); 1049 if (release_psref_ia) 1050 if_put(origifp, &psref_ia); 1051 curlwp_bindx(bound); 1052 1053 return error; 1054 1055 freehdrs: 1056 m_freem(exthdrs.ip6e_hbh); 1057 m_freem(exthdrs.ip6e_dest1); 1058 m_freem(exthdrs.ip6e_rthdr); 1059 m_freem(exthdrs.ip6e_dest2); 1060 /* FALLTHROUGH */ 1061 bad: 1062 m_freem(m); 1063 goto done; 1064 1065 badscope: 1066 IP6_STATINC(IP6_STAT_BADSCOPE); 1067 in6_ifstat_inc(origifp, ifs6_out_discard); 1068 if (error == 0) 1069 error = EHOSTUNREACH; /* XXX */ 1070 goto bad; 1071 } 1072 1073 static int 1074 ip6_copyexthdr(struct mbuf **mp, void *hdr, int hlen) 1075 { 1076 struct mbuf *m; 1077 1078 if (hlen > MCLBYTES) 1079 return ENOBUFS; /* XXX */ 1080 1081 MGET(m, M_DONTWAIT, MT_DATA); 1082 if (!m) 1083 return ENOBUFS; 1084 1085 if (hlen > MLEN) { 1086 MCLGET(m, M_DONTWAIT); 1087 if ((m->m_flags & M_EXT) == 0) { 1088 m_free(m); 1089 return ENOBUFS; 1090 } 1091 } 1092 m->m_len = hlen; 1093 if (hdr) 1094 memcpy(mtod(m, void *), hdr, hlen); 1095 1096 *mp = m; 1097 return 0; 1098 } 1099 1100 /* 1101 * Process a delayed payload checksum calculation. 1102 */ 1103 void 1104 in6_delayed_cksum(struct mbuf *m) 1105 { 1106 uint16_t csum, offset; 1107 1108 KASSERT((m->m_pkthdr.csum_flags & (M_CSUM_UDPv6|M_CSUM_TCPv6)) != 0); 1109 KASSERT((~m->m_pkthdr.csum_flags & (M_CSUM_UDPv6|M_CSUM_TCPv6)) != 0); 1110 KASSERT((m->m_pkthdr.csum_flags 1111 & (M_CSUM_UDPv4|M_CSUM_TCPv4|M_CSUM_TSOv4)) == 0); 1112 1113 offset = M_CSUM_DATA_IPv6_HL(m->m_pkthdr.csum_data); 1114 csum = in6_cksum(m, 0, offset, m->m_pkthdr.len - offset); 1115 if (csum == 0 && (m->m_pkthdr.csum_flags & M_CSUM_UDPv6) != 0) { 1116 csum = 0xffff; 1117 } 1118 1119 offset += M_CSUM_DATA_IPv6_OFFSET(m->m_pkthdr.csum_data); 1120 if ((offset + sizeof(csum)) > m->m_len) { 1121 m_copyback(m, offset, sizeof(csum), &csum); 1122 } else { 1123 *(uint16_t *)(mtod(m, char *) + offset) = csum; 1124 } 1125 } 1126 1127 /* 1128 * Insert jumbo payload option. 1129 */ 1130 static int 1131 ip6_insert_jumboopt(struct ip6_exthdrs *exthdrs, u_int32_t plen) 1132 { 1133 struct mbuf *mopt; 1134 u_int8_t *optbuf; 1135 u_int32_t v; 1136 1137 #define JUMBOOPTLEN 8 /* length of jumbo payload option and padding */ 1138 1139 /* 1140 * If there is no hop-by-hop options header, allocate new one. 1141 * If there is one but it doesn't have enough space to store the 1142 * jumbo payload option, allocate a cluster to store the whole options. 1143 * Otherwise, use it to store the options. 1144 */ 1145 if (exthdrs->ip6e_hbh == NULL) { 1146 MGET(mopt, M_DONTWAIT, MT_DATA); 1147 if (mopt == 0) 1148 return (ENOBUFS); 1149 mopt->m_len = JUMBOOPTLEN; 1150 optbuf = mtod(mopt, u_int8_t *); 1151 optbuf[1] = 0; /* = ((JUMBOOPTLEN) >> 3) - 1 */ 1152 exthdrs->ip6e_hbh = mopt; 1153 } else { 1154 struct ip6_hbh *hbh; 1155 1156 mopt = exthdrs->ip6e_hbh; 1157 if (M_TRAILINGSPACE(mopt) < JUMBOOPTLEN) { 1158 const int oldoptlen = mopt->m_len; 1159 struct mbuf *n; 1160 1161 /* 1162 * Assumptions: 1163 * - exthdrs->ip6e_hbh is not referenced from places 1164 * other than exthdrs. 1165 * - exthdrs->ip6e_hbh is not an mbuf chain. 1166 */ 1167 KASSERT(mopt->m_next == NULL); 1168 1169 /* 1170 * Give up if the whole (new) hbh header does not fit 1171 * even in an mbuf cluster. 1172 */ 1173 if (oldoptlen + JUMBOOPTLEN > MCLBYTES) 1174 return ENOBUFS; 1175 1176 /* 1177 * At this point, we must always prepare a cluster. 1178 */ 1179 MGET(n, M_DONTWAIT, MT_DATA); 1180 if (n) { 1181 MCLGET(n, M_DONTWAIT); 1182 if ((n->m_flags & M_EXT) == 0) { 1183 m_freem(n); 1184 n = NULL; 1185 } 1186 } 1187 if (!n) 1188 return ENOBUFS; 1189 1190 n->m_len = oldoptlen + JUMBOOPTLEN; 1191 bcopy(mtod(mopt, void *), mtod(n, void *), 1192 oldoptlen); 1193 optbuf = mtod(n, u_int8_t *) + oldoptlen; 1194 m_freem(mopt); 1195 mopt = exthdrs->ip6e_hbh = n; 1196 } else { 1197 optbuf = mtod(mopt, u_int8_t *) + mopt->m_len; 1198 mopt->m_len += JUMBOOPTLEN; 1199 } 1200 optbuf[0] = IP6OPT_PADN; 1201 optbuf[1] = 0; 1202 1203 /* 1204 * Adjust the header length according to the pad and 1205 * the jumbo payload option. 1206 */ 1207 hbh = mtod(mopt, struct ip6_hbh *); 1208 hbh->ip6h_len += (JUMBOOPTLEN >> 3); 1209 } 1210 1211 /* fill in the option. */ 1212 optbuf[2] = IP6OPT_JUMBO; 1213 optbuf[3] = 4; 1214 v = (u_int32_t)htonl(plen + JUMBOOPTLEN); 1215 memcpy(&optbuf[4], &v, sizeof(u_int32_t)); 1216 1217 /* finally, adjust the packet header length */ 1218 exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN; 1219 1220 return 0; 1221 #undef JUMBOOPTLEN 1222 } 1223 1224 /* 1225 * Insert fragment header and copy unfragmentable header portions. 1226 * 1227 * *frghdrp will not be read, and it is guaranteed that either an 1228 * error is returned or that *frghdrp will point to space allocated 1229 * for the fragment header. 1230 * 1231 * On entry, m contains: 1232 * IPv6 Header 1233 * On exit, it contains: 1234 * IPv6 Header -> Unfragmentable Part -> Frag6 Header 1235 */ 1236 static int 1237 ip6_insertfraghdr(struct mbuf *m0, struct mbuf *m, int hlen, 1238 struct ip6_frag **frghdrp) 1239 { 1240 struct mbuf *n, *mlast; 1241 1242 if (hlen > sizeof(struct ip6_hdr)) { 1243 n = m_copym(m0, sizeof(struct ip6_hdr), 1244 hlen - sizeof(struct ip6_hdr), M_DONTWAIT); 1245 if (n == NULL) 1246 return ENOBUFS; 1247 m->m_next = n; 1248 } else 1249 n = m; 1250 1251 /* Search for the last mbuf of unfragmentable part. */ 1252 for (mlast = n; mlast->m_next; mlast = mlast->m_next) 1253 ; 1254 1255 if ((mlast->m_flags & M_EXT) == 0 && 1256 M_TRAILINGSPACE(mlast) >= sizeof(struct ip6_frag)) { 1257 /* use the trailing space of the last mbuf for the fragment hdr */ 1258 *frghdrp = (struct ip6_frag *)(mtod(mlast, char *) + 1259 mlast->m_len); 1260 mlast->m_len += sizeof(struct ip6_frag); 1261 } else { 1262 /* allocate a new mbuf for the fragment header */ 1263 struct mbuf *mfrg; 1264 1265 MGET(mfrg, M_DONTWAIT, MT_DATA); 1266 if (mfrg == NULL) 1267 return ENOBUFS; 1268 mfrg->m_len = sizeof(struct ip6_frag); 1269 *frghdrp = mtod(mfrg, struct ip6_frag *); 1270 mlast->m_next = mfrg; 1271 } 1272 1273 return 0; 1274 } 1275 1276 static int 1277 ip6_getpmtu(struct rtentry *rt, struct ifnet *ifp, u_long *mtup, 1278 int *alwaysfragp) 1279 { 1280 u_int32_t mtu = 0; 1281 int alwaysfrag = 0; 1282 int error = 0; 1283 1284 if (rt != NULL) { 1285 u_int32_t ifmtu; 1286 1287 if (ifp == NULL) 1288 ifp = rt->rt_ifp; 1289 ifmtu = IN6_LINKMTU(ifp); 1290 mtu = rt->rt_rmx.rmx_mtu; 1291 if (mtu == 0) 1292 mtu = ifmtu; 1293 else if (mtu < IPV6_MMTU) { 1294 /* 1295 * RFC2460 section 5, last paragraph: 1296 * if we record ICMPv6 too big message with 1297 * mtu < IPV6_MMTU, transmit packets sized IPV6_MMTU 1298 * or smaller, with fragment header attached. 1299 * (fragment header is needed regardless from the 1300 * packet size, for translators to identify packets) 1301 */ 1302 alwaysfrag = 1; 1303 mtu = IPV6_MMTU; 1304 } else if (mtu > ifmtu) { 1305 /* 1306 * The MTU on the route is larger than the MTU on 1307 * the interface! This shouldn't happen, unless the 1308 * MTU of the interface has been changed after the 1309 * interface was brought up. Change the MTU in the 1310 * route to match the interface MTU (as long as the 1311 * field isn't locked). 1312 */ 1313 mtu = ifmtu; 1314 if (!(rt->rt_rmx.rmx_locks & RTV_MTU)) 1315 rt->rt_rmx.rmx_mtu = mtu; 1316 } 1317 } else if (ifp) { 1318 mtu = IN6_LINKMTU(ifp); 1319 } else 1320 error = EHOSTUNREACH; /* XXX */ 1321 1322 *mtup = mtu; 1323 if (alwaysfragp) 1324 *alwaysfragp = alwaysfrag; 1325 return (error); 1326 } 1327 1328 /* 1329 * IP6 socket option processing. 1330 */ 1331 int 1332 ip6_ctloutput(int op, struct socket *so, struct sockopt *sopt) 1333 { 1334 int optdatalen, uproto; 1335 void *optdata; 1336 struct in6pcb *in6p = sotoin6pcb(so); 1337 struct ip_moptions **mopts; 1338 int error, optval; 1339 int level, optname; 1340 1341 KASSERT(solocked(so)); 1342 KASSERT(sopt != NULL); 1343 1344 level = sopt->sopt_level; 1345 optname = sopt->sopt_name; 1346 1347 error = optval = 0; 1348 uproto = (int)so->so_proto->pr_protocol; 1349 1350 switch (level) { 1351 case IPPROTO_IP: 1352 switch (optname) { 1353 case IP_ADD_MEMBERSHIP: 1354 case IP_DROP_MEMBERSHIP: 1355 case IP_MULTICAST_IF: 1356 case IP_MULTICAST_LOOP: 1357 case IP_MULTICAST_TTL: 1358 mopts = &in6p->in6p_v4moptions; 1359 switch (op) { 1360 case PRCO_GETOPT: 1361 return ip_getmoptions(*mopts, sopt); 1362 case PRCO_SETOPT: 1363 return ip_setmoptions(mopts, sopt); 1364 default: 1365 return EINVAL; 1366 } 1367 default: 1368 return ENOPROTOOPT; 1369 } 1370 case IPPROTO_IPV6: 1371 break; 1372 default: 1373 return ENOPROTOOPT; 1374 } 1375 switch (op) { 1376 case PRCO_SETOPT: 1377 switch (optname) { 1378 #ifdef RFC2292 1379 case IPV6_2292PKTOPTIONS: 1380 error = ip6_pcbopts(&in6p->in6p_outputopts, so, sopt); 1381 break; 1382 #endif 1383 1384 /* 1385 * Use of some Hop-by-Hop options or some 1386 * Destination options, might require special 1387 * privilege. That is, normal applications 1388 * (without special privilege) might be forbidden 1389 * from setting certain options in outgoing packets, 1390 * and might never see certain options in received 1391 * packets. [RFC 2292 Section 6] 1392 * KAME specific note: 1393 * KAME prevents non-privileged users from sending or 1394 * receiving ANY hbh/dst options in order to avoid 1395 * overhead of parsing options in the kernel. 1396 */ 1397 case IPV6_RECVHOPOPTS: 1398 case IPV6_RECVDSTOPTS: 1399 case IPV6_RECVRTHDRDSTOPTS: 1400 error = kauth_authorize_network(kauth_cred_get(), 1401 KAUTH_NETWORK_IPV6, KAUTH_REQ_NETWORK_IPV6_HOPBYHOP, 1402 NULL, NULL, NULL); 1403 if (error) 1404 break; 1405 /* FALLTHROUGH */ 1406 case IPV6_UNICAST_HOPS: 1407 case IPV6_HOPLIMIT: 1408 case IPV6_FAITH: 1409 1410 case IPV6_RECVPKTINFO: 1411 case IPV6_RECVHOPLIMIT: 1412 case IPV6_RECVRTHDR: 1413 case IPV6_RECVPATHMTU: 1414 case IPV6_RECVTCLASS: 1415 case IPV6_V6ONLY: 1416 error = sockopt_getint(sopt, &optval); 1417 if (error) 1418 break; 1419 switch (optname) { 1420 case IPV6_UNICAST_HOPS: 1421 if (optval < -1 || optval >= 256) 1422 error = EINVAL; 1423 else { 1424 /* -1 = kernel default */ 1425 in6p->in6p_hops = optval; 1426 } 1427 break; 1428 #define OPTSET(bit) \ 1429 do { \ 1430 if (optval) \ 1431 in6p->in6p_flags |= (bit); \ 1432 else \ 1433 in6p->in6p_flags &= ~(bit); \ 1434 } while (/*CONSTCOND*/ 0) 1435 1436 #ifdef RFC2292 1437 #define OPTSET2292(bit) \ 1438 do { \ 1439 in6p->in6p_flags |= IN6P_RFC2292; \ 1440 if (optval) \ 1441 in6p->in6p_flags |= (bit); \ 1442 else \ 1443 in6p->in6p_flags &= ~(bit); \ 1444 } while (/*CONSTCOND*/ 0) 1445 #endif 1446 1447 #define OPTBIT(bit) (in6p->in6p_flags & (bit) ? 1 : 0) 1448 1449 case IPV6_RECVPKTINFO: 1450 #ifdef RFC2292 1451 /* cannot mix with RFC2292 */ 1452 if (OPTBIT(IN6P_RFC2292)) { 1453 error = EINVAL; 1454 break; 1455 } 1456 #endif 1457 OPTSET(IN6P_PKTINFO); 1458 break; 1459 1460 case IPV6_HOPLIMIT: 1461 { 1462 struct ip6_pktopts **optp; 1463 1464 #ifdef RFC2292 1465 /* cannot mix with RFC2292 */ 1466 if (OPTBIT(IN6P_RFC2292)) { 1467 error = EINVAL; 1468 break; 1469 } 1470 #endif 1471 optp = &in6p->in6p_outputopts; 1472 error = ip6_pcbopt(IPV6_HOPLIMIT, 1473 (u_char *)&optval, 1474 sizeof(optval), 1475 optp, 1476 kauth_cred_get(), uproto); 1477 break; 1478 } 1479 1480 case IPV6_RECVHOPLIMIT: 1481 #ifdef RFC2292 1482 /* cannot mix with RFC2292 */ 1483 if (OPTBIT(IN6P_RFC2292)) { 1484 error = EINVAL; 1485 break; 1486 } 1487 #endif 1488 OPTSET(IN6P_HOPLIMIT); 1489 break; 1490 1491 case IPV6_RECVHOPOPTS: 1492 #ifdef RFC2292 1493 /* cannot mix with RFC2292 */ 1494 if (OPTBIT(IN6P_RFC2292)) { 1495 error = EINVAL; 1496 break; 1497 } 1498 #endif 1499 OPTSET(IN6P_HOPOPTS); 1500 break; 1501 1502 case IPV6_RECVDSTOPTS: 1503 #ifdef RFC2292 1504 /* cannot mix with RFC2292 */ 1505 if (OPTBIT(IN6P_RFC2292)) { 1506 error = EINVAL; 1507 break; 1508 } 1509 #endif 1510 OPTSET(IN6P_DSTOPTS); 1511 break; 1512 1513 case IPV6_RECVRTHDRDSTOPTS: 1514 #ifdef RFC2292 1515 /* cannot mix with RFC2292 */ 1516 if (OPTBIT(IN6P_RFC2292)) { 1517 error = EINVAL; 1518 break; 1519 } 1520 #endif 1521 OPTSET(IN6P_RTHDRDSTOPTS); 1522 break; 1523 1524 case IPV6_RECVRTHDR: 1525 #ifdef RFC2292 1526 /* cannot mix with RFC2292 */ 1527 if (OPTBIT(IN6P_RFC2292)) { 1528 error = EINVAL; 1529 break; 1530 } 1531 #endif 1532 OPTSET(IN6P_RTHDR); 1533 break; 1534 1535 case IPV6_FAITH: 1536 OPTSET(IN6P_FAITH); 1537 break; 1538 1539 case IPV6_RECVPATHMTU: 1540 /* 1541 * We ignore this option for TCP 1542 * sockets. 1543 * (RFC3542 leaves this case 1544 * unspecified.) 1545 */ 1546 if (uproto != IPPROTO_TCP) 1547 OPTSET(IN6P_MTU); 1548 break; 1549 1550 case IPV6_V6ONLY: 1551 /* 1552 * make setsockopt(IPV6_V6ONLY) 1553 * available only prior to bind(2). 1554 * see ipng mailing list, Jun 22 2001. 1555 */ 1556 if (in6p->in6p_lport || 1557 !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr)) { 1558 error = EINVAL; 1559 break; 1560 } 1561 #ifdef INET6_BINDV6ONLY 1562 if (!optval) 1563 error = EINVAL; 1564 #else 1565 OPTSET(IN6P_IPV6_V6ONLY); 1566 #endif 1567 break; 1568 case IPV6_RECVTCLASS: 1569 #ifdef RFC2292 1570 /* cannot mix with RFC2292 XXX */ 1571 if (OPTBIT(IN6P_RFC2292)) { 1572 error = EINVAL; 1573 break; 1574 } 1575 #endif 1576 OPTSET(IN6P_TCLASS); 1577 break; 1578 1579 } 1580 break; 1581 1582 case IPV6_OTCLASS: 1583 { 1584 struct ip6_pktopts **optp; 1585 u_int8_t tclass; 1586 1587 error = sockopt_get(sopt, &tclass, sizeof(tclass)); 1588 if (error) 1589 break; 1590 optp = &in6p->in6p_outputopts; 1591 error = ip6_pcbopt(optname, 1592 (u_char *)&tclass, 1593 sizeof(tclass), 1594 optp, 1595 kauth_cred_get(), uproto); 1596 break; 1597 } 1598 1599 case IPV6_TCLASS: 1600 case IPV6_DONTFRAG: 1601 case IPV6_USE_MIN_MTU: 1602 case IPV6_PREFER_TEMPADDR: 1603 error = sockopt_getint(sopt, &optval); 1604 if (error) 1605 break; 1606 { 1607 struct ip6_pktopts **optp; 1608 optp = &in6p->in6p_outputopts; 1609 error = ip6_pcbopt(optname, 1610 (u_char *)&optval, 1611 sizeof(optval), 1612 optp, 1613 kauth_cred_get(), uproto); 1614 break; 1615 } 1616 1617 #ifdef RFC2292 1618 case IPV6_2292PKTINFO: 1619 case IPV6_2292HOPLIMIT: 1620 case IPV6_2292HOPOPTS: 1621 case IPV6_2292DSTOPTS: 1622 case IPV6_2292RTHDR: 1623 /* RFC 2292 */ 1624 error = sockopt_getint(sopt, &optval); 1625 if (error) 1626 break; 1627 1628 switch (optname) { 1629 case IPV6_2292PKTINFO: 1630 OPTSET2292(IN6P_PKTINFO); 1631 break; 1632 case IPV6_2292HOPLIMIT: 1633 OPTSET2292(IN6P_HOPLIMIT); 1634 break; 1635 case IPV6_2292HOPOPTS: 1636 /* 1637 * Check super-user privilege. 1638 * See comments for IPV6_RECVHOPOPTS. 1639 */ 1640 error = 1641 kauth_authorize_network(kauth_cred_get(), 1642 KAUTH_NETWORK_IPV6, 1643 KAUTH_REQ_NETWORK_IPV6_HOPBYHOP, NULL, 1644 NULL, NULL); 1645 if (error) 1646 return (error); 1647 OPTSET2292(IN6P_HOPOPTS); 1648 break; 1649 case IPV6_2292DSTOPTS: 1650 error = 1651 kauth_authorize_network(kauth_cred_get(), 1652 KAUTH_NETWORK_IPV6, 1653 KAUTH_REQ_NETWORK_IPV6_HOPBYHOP, NULL, 1654 NULL, NULL); 1655 if (error) 1656 return (error); 1657 OPTSET2292(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); /* XXX */ 1658 break; 1659 case IPV6_2292RTHDR: 1660 OPTSET2292(IN6P_RTHDR); 1661 break; 1662 } 1663 break; 1664 #endif 1665 case IPV6_PKTINFO: 1666 case IPV6_HOPOPTS: 1667 case IPV6_RTHDR: 1668 case IPV6_DSTOPTS: 1669 case IPV6_RTHDRDSTOPTS: 1670 case IPV6_NEXTHOP: { 1671 /* new advanced API (RFC3542) */ 1672 void *optbuf; 1673 int optbuflen; 1674 struct ip6_pktopts **optp; 1675 1676 #ifdef RFC2292 1677 /* cannot mix with RFC2292 */ 1678 if (OPTBIT(IN6P_RFC2292)) { 1679 error = EINVAL; 1680 break; 1681 } 1682 #endif 1683 1684 optbuflen = sopt->sopt_size; 1685 optbuf = malloc(optbuflen, M_IP6OPT, M_NOWAIT); 1686 if (optbuf == NULL) { 1687 error = ENOBUFS; 1688 break; 1689 } 1690 1691 error = sockopt_get(sopt, optbuf, optbuflen); 1692 if (error) { 1693 free(optbuf, M_IP6OPT); 1694 break; 1695 } 1696 optp = &in6p->in6p_outputopts; 1697 error = ip6_pcbopt(optname, optbuf, optbuflen, 1698 optp, kauth_cred_get(), uproto); 1699 1700 free(optbuf, M_IP6OPT); 1701 break; 1702 } 1703 #undef OPTSET 1704 1705 case IPV6_MULTICAST_IF: 1706 case IPV6_MULTICAST_HOPS: 1707 case IPV6_MULTICAST_LOOP: 1708 case IPV6_JOIN_GROUP: 1709 case IPV6_LEAVE_GROUP: 1710 error = ip6_setmoptions(sopt, in6p); 1711 break; 1712 1713 case IPV6_PORTRANGE: 1714 error = sockopt_getint(sopt, &optval); 1715 if (error) 1716 break; 1717 1718 switch (optval) { 1719 case IPV6_PORTRANGE_DEFAULT: 1720 in6p->in6p_flags &= ~(IN6P_LOWPORT); 1721 in6p->in6p_flags &= ~(IN6P_HIGHPORT); 1722 break; 1723 1724 case IPV6_PORTRANGE_HIGH: 1725 in6p->in6p_flags &= ~(IN6P_LOWPORT); 1726 in6p->in6p_flags |= IN6P_HIGHPORT; 1727 break; 1728 1729 case IPV6_PORTRANGE_LOW: 1730 in6p->in6p_flags &= ~(IN6P_HIGHPORT); 1731 in6p->in6p_flags |= IN6P_LOWPORT; 1732 break; 1733 1734 default: 1735 error = EINVAL; 1736 break; 1737 } 1738 break; 1739 1740 case IPV6_PORTALGO: 1741 error = sockopt_getint(sopt, &optval); 1742 if (error) 1743 break; 1744 1745 error = portalgo_algo_index_select( 1746 (struct inpcb_hdr *)in6p, optval); 1747 break; 1748 1749 #if defined(IPSEC) 1750 case IPV6_IPSEC_POLICY: 1751 if (ipsec_enabled) { 1752 error = ipsec_set_policy(in6p, optname, 1753 sopt->sopt_data, sopt->sopt_size, 1754 kauth_cred_get()); 1755 break; 1756 } 1757 /*FALLTHROUGH*/ 1758 #endif /* IPSEC */ 1759 1760 default: 1761 error = ENOPROTOOPT; 1762 break; 1763 } 1764 break; 1765 1766 case PRCO_GETOPT: 1767 switch (optname) { 1768 #ifdef RFC2292 1769 case IPV6_2292PKTOPTIONS: 1770 /* 1771 * RFC3542 (effectively) deprecated the 1772 * semantics of the 2292-style pktoptions. 1773 * Since it was not reliable in nature (i.e., 1774 * applications had to expect the lack of some 1775 * information after all), it would make sense 1776 * to simplify this part by always returning 1777 * empty data. 1778 */ 1779 break; 1780 #endif 1781 1782 case IPV6_RECVHOPOPTS: 1783 case IPV6_RECVDSTOPTS: 1784 case IPV6_RECVRTHDRDSTOPTS: 1785 case IPV6_UNICAST_HOPS: 1786 case IPV6_RECVPKTINFO: 1787 case IPV6_RECVHOPLIMIT: 1788 case IPV6_RECVRTHDR: 1789 case IPV6_RECVPATHMTU: 1790 1791 case IPV6_FAITH: 1792 case IPV6_V6ONLY: 1793 case IPV6_PORTRANGE: 1794 case IPV6_RECVTCLASS: 1795 switch (optname) { 1796 1797 case IPV6_RECVHOPOPTS: 1798 optval = OPTBIT(IN6P_HOPOPTS); 1799 break; 1800 1801 case IPV6_RECVDSTOPTS: 1802 optval = OPTBIT(IN6P_DSTOPTS); 1803 break; 1804 1805 case IPV6_RECVRTHDRDSTOPTS: 1806 optval = OPTBIT(IN6P_RTHDRDSTOPTS); 1807 break; 1808 1809 case IPV6_UNICAST_HOPS: 1810 optval = in6p->in6p_hops; 1811 break; 1812 1813 case IPV6_RECVPKTINFO: 1814 optval = OPTBIT(IN6P_PKTINFO); 1815 break; 1816 1817 case IPV6_RECVHOPLIMIT: 1818 optval = OPTBIT(IN6P_HOPLIMIT); 1819 break; 1820 1821 case IPV6_RECVRTHDR: 1822 optval = OPTBIT(IN6P_RTHDR); 1823 break; 1824 1825 case IPV6_RECVPATHMTU: 1826 optval = OPTBIT(IN6P_MTU); 1827 break; 1828 1829 case IPV6_FAITH: 1830 optval = OPTBIT(IN6P_FAITH); 1831 break; 1832 1833 case IPV6_V6ONLY: 1834 optval = OPTBIT(IN6P_IPV6_V6ONLY); 1835 break; 1836 1837 case IPV6_PORTRANGE: 1838 { 1839 int flags; 1840 flags = in6p->in6p_flags; 1841 if (flags & IN6P_HIGHPORT) 1842 optval = IPV6_PORTRANGE_HIGH; 1843 else if (flags & IN6P_LOWPORT) 1844 optval = IPV6_PORTRANGE_LOW; 1845 else 1846 optval = 0; 1847 break; 1848 } 1849 case IPV6_RECVTCLASS: 1850 optval = OPTBIT(IN6P_TCLASS); 1851 break; 1852 1853 } 1854 if (error) 1855 break; 1856 error = sockopt_setint(sopt, optval); 1857 break; 1858 1859 case IPV6_PATHMTU: 1860 { 1861 u_long pmtu = 0; 1862 struct ip6_mtuinfo mtuinfo; 1863 struct route *ro = &in6p->in6p_route; 1864 struct rtentry *rt; 1865 union { 1866 struct sockaddr dst; 1867 struct sockaddr_in6 dst6; 1868 } u; 1869 1870 if (!(so->so_state & SS_ISCONNECTED)) 1871 return (ENOTCONN); 1872 /* 1873 * XXX: we dot not consider the case of source 1874 * routing, or optional information to specify 1875 * the outgoing interface. 1876 */ 1877 sockaddr_in6_init(&u.dst6, &in6p->in6p_faddr, 0, 0, 0); 1878 rt = rtcache_lookup(ro, &u.dst); 1879 error = ip6_getpmtu(rt, NULL, &pmtu, NULL); 1880 rtcache_unref(rt, ro); 1881 if (error) 1882 break; 1883 if (pmtu > IPV6_MAXPACKET) 1884 pmtu = IPV6_MAXPACKET; 1885 1886 memset(&mtuinfo, 0, sizeof(mtuinfo)); 1887 mtuinfo.ip6m_mtu = (u_int32_t)pmtu; 1888 optdata = (void *)&mtuinfo; 1889 optdatalen = sizeof(mtuinfo); 1890 if (optdatalen > MCLBYTES) 1891 return (EMSGSIZE); /* XXX */ 1892 error = sockopt_set(sopt, optdata, optdatalen); 1893 break; 1894 } 1895 1896 #ifdef RFC2292 1897 case IPV6_2292PKTINFO: 1898 case IPV6_2292HOPLIMIT: 1899 case IPV6_2292HOPOPTS: 1900 case IPV6_2292RTHDR: 1901 case IPV6_2292DSTOPTS: 1902 switch (optname) { 1903 case IPV6_2292PKTINFO: 1904 optval = OPTBIT(IN6P_PKTINFO); 1905 break; 1906 case IPV6_2292HOPLIMIT: 1907 optval = OPTBIT(IN6P_HOPLIMIT); 1908 break; 1909 case IPV6_2292HOPOPTS: 1910 optval = OPTBIT(IN6P_HOPOPTS); 1911 break; 1912 case IPV6_2292RTHDR: 1913 optval = OPTBIT(IN6P_RTHDR); 1914 break; 1915 case IPV6_2292DSTOPTS: 1916 optval = OPTBIT(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); 1917 break; 1918 } 1919 error = sockopt_setint(sopt, optval); 1920 break; 1921 #endif 1922 case IPV6_PKTINFO: 1923 case IPV6_HOPOPTS: 1924 case IPV6_RTHDR: 1925 case IPV6_DSTOPTS: 1926 case IPV6_RTHDRDSTOPTS: 1927 case IPV6_NEXTHOP: 1928 case IPV6_OTCLASS: 1929 case IPV6_TCLASS: 1930 case IPV6_DONTFRAG: 1931 case IPV6_USE_MIN_MTU: 1932 case IPV6_PREFER_TEMPADDR: 1933 error = ip6_getpcbopt(in6p->in6p_outputopts, 1934 optname, sopt); 1935 break; 1936 1937 case IPV6_MULTICAST_IF: 1938 case IPV6_MULTICAST_HOPS: 1939 case IPV6_MULTICAST_LOOP: 1940 case IPV6_JOIN_GROUP: 1941 case IPV6_LEAVE_GROUP: 1942 error = ip6_getmoptions(sopt, in6p); 1943 break; 1944 1945 case IPV6_PORTALGO: 1946 optval = ((struct inpcb_hdr *)in6p)->inph_portalgo; 1947 error = sockopt_setint(sopt, optval); 1948 break; 1949 1950 #if defined(IPSEC) 1951 case IPV6_IPSEC_POLICY: 1952 if (ipsec_used) { 1953 struct mbuf *m = NULL; 1954 1955 /* 1956 * XXX: this will return EINVAL as sopt is 1957 * empty 1958 */ 1959 error = ipsec_get_policy(in6p, sopt->sopt_data, 1960 sopt->sopt_size, &m); 1961 if (!error) 1962 error = sockopt_setmbuf(sopt, m); 1963 break; 1964 } 1965 /*FALLTHROUGH*/ 1966 #endif /* IPSEC */ 1967 1968 default: 1969 error = ENOPROTOOPT; 1970 break; 1971 } 1972 break; 1973 } 1974 return (error); 1975 } 1976 1977 int 1978 ip6_raw_ctloutput(int op, struct socket *so, struct sockopt *sopt) 1979 { 1980 int error = 0, optval; 1981 const int icmp6off = offsetof(struct icmp6_hdr, icmp6_cksum); 1982 struct in6pcb *in6p = sotoin6pcb(so); 1983 int level, optname; 1984 1985 KASSERT(sopt != NULL); 1986 1987 level = sopt->sopt_level; 1988 optname = sopt->sopt_name; 1989 1990 if (level != IPPROTO_IPV6) { 1991 return ENOPROTOOPT; 1992 } 1993 1994 switch (optname) { 1995 case IPV6_CHECKSUM: 1996 /* 1997 * For ICMPv6 sockets, no modification allowed for checksum 1998 * offset, permit "no change" values to help existing apps. 1999 * 2000 * XXX RFC3542 says: "An attempt to set IPV6_CHECKSUM 2001 * for an ICMPv6 socket will fail." The current 2002 * behavior does not meet RFC3542. 2003 */ 2004 switch (op) { 2005 case PRCO_SETOPT: 2006 error = sockopt_getint(sopt, &optval); 2007 if (error) 2008 break; 2009 if ((optval % 2) != 0) { 2010 /* the API assumes even offset values */ 2011 error = EINVAL; 2012 } else if (so->so_proto->pr_protocol == 2013 IPPROTO_ICMPV6) { 2014 if (optval != icmp6off) 2015 error = EINVAL; 2016 } else 2017 in6p->in6p_cksum = optval; 2018 break; 2019 2020 case PRCO_GETOPT: 2021 if (so->so_proto->pr_protocol == IPPROTO_ICMPV6) 2022 optval = icmp6off; 2023 else 2024 optval = in6p->in6p_cksum; 2025 2026 error = sockopt_setint(sopt, optval); 2027 break; 2028 2029 default: 2030 error = EINVAL; 2031 break; 2032 } 2033 break; 2034 2035 default: 2036 error = ENOPROTOOPT; 2037 break; 2038 } 2039 2040 return (error); 2041 } 2042 2043 #ifdef RFC2292 2044 /* 2045 * Set up IP6 options in pcb for insertion in output packets or 2046 * specifying behavior of outgoing packets. 2047 */ 2048 static int 2049 ip6_pcbopts(struct ip6_pktopts **pktopt, struct socket *so, 2050 struct sockopt *sopt) 2051 { 2052 struct ip6_pktopts *opt = *pktopt; 2053 struct mbuf *m; 2054 int error = 0; 2055 2056 KASSERT(solocked(so)); 2057 2058 /* turn off any old options. */ 2059 if (opt) { 2060 #ifdef DIAGNOSTIC 2061 if (opt->ip6po_pktinfo || opt->ip6po_nexthop || 2062 opt->ip6po_hbh || opt->ip6po_dest1 || opt->ip6po_dest2 || 2063 opt->ip6po_rhinfo.ip6po_rhi_rthdr) 2064 printf("ip6_pcbopts: all specified options are cleared.\n"); 2065 #endif 2066 ip6_clearpktopts(opt, -1); 2067 } else { 2068 opt = malloc(sizeof(*opt), M_IP6OPT, M_NOWAIT); 2069 if (opt == NULL) 2070 return (ENOBUFS); 2071 } 2072 *pktopt = NULL; 2073 2074 if (sopt == NULL || sopt->sopt_size == 0) { 2075 /* 2076 * Only turning off any previous options, regardless of 2077 * whether the opt is just created or given. 2078 */ 2079 free(opt, M_IP6OPT); 2080 return (0); 2081 } 2082 2083 /* set options specified by user. */ 2084 m = sockopt_getmbuf(sopt); 2085 if (m == NULL) { 2086 free(opt, M_IP6OPT); 2087 return (ENOBUFS); 2088 } 2089 2090 error = ip6_setpktopts(m, opt, NULL, kauth_cred_get(), 2091 so->so_proto->pr_protocol); 2092 m_freem(m); 2093 if (error != 0) { 2094 ip6_clearpktopts(opt, -1); /* XXX: discard all options */ 2095 free(opt, M_IP6OPT); 2096 return (error); 2097 } 2098 *pktopt = opt; 2099 return (0); 2100 } 2101 #endif 2102 2103 /* 2104 * initialize ip6_pktopts. beware that there are non-zero default values in 2105 * the struct. 2106 */ 2107 void 2108 ip6_initpktopts(struct ip6_pktopts *opt) 2109 { 2110 2111 memset(opt, 0, sizeof(*opt)); 2112 opt->ip6po_hlim = -1; /* -1 means default hop limit */ 2113 opt->ip6po_tclass = -1; /* -1 means default traffic class */ 2114 opt->ip6po_minmtu = IP6PO_MINMTU_MCASTONLY; 2115 opt->ip6po_prefer_tempaddr = IP6PO_TEMPADDR_SYSTEM; 2116 } 2117 2118 #define sin6tosa(sin6) ((struct sockaddr *)(sin6)) /* XXX */ 2119 static int 2120 ip6_pcbopt(int optname, u_char *buf, int len, struct ip6_pktopts **pktopt, 2121 kauth_cred_t cred, int uproto) 2122 { 2123 struct ip6_pktopts *opt; 2124 2125 if (*pktopt == NULL) { 2126 *pktopt = malloc(sizeof(struct ip6_pktopts), M_IP6OPT, 2127 M_NOWAIT); 2128 if (*pktopt == NULL) 2129 return (ENOBUFS); 2130 2131 ip6_initpktopts(*pktopt); 2132 } 2133 opt = *pktopt; 2134 2135 return (ip6_setpktopt(optname, buf, len, opt, cred, 1, 0, uproto)); 2136 } 2137 2138 static int 2139 ip6_getpcbopt(struct ip6_pktopts *pktopt, int optname, struct sockopt *sopt) 2140 { 2141 void *optdata = NULL; 2142 int optdatalen = 0; 2143 struct ip6_ext *ip6e; 2144 int error = 0; 2145 struct in6_pktinfo null_pktinfo; 2146 int deftclass = 0, on; 2147 int defminmtu = IP6PO_MINMTU_MCASTONLY; 2148 int defpreftemp = IP6PO_TEMPADDR_SYSTEM; 2149 2150 switch (optname) { 2151 case IPV6_PKTINFO: 2152 if (pktopt && pktopt->ip6po_pktinfo) 2153 optdata = (void *)pktopt->ip6po_pktinfo; 2154 else { 2155 /* XXX: we don't have to do this every time... */ 2156 memset(&null_pktinfo, 0, sizeof(null_pktinfo)); 2157 optdata = (void *)&null_pktinfo; 2158 } 2159 optdatalen = sizeof(struct in6_pktinfo); 2160 break; 2161 case IPV6_OTCLASS: 2162 /* XXX */ 2163 return (EINVAL); 2164 case IPV6_TCLASS: 2165 if (pktopt && pktopt->ip6po_tclass >= 0) 2166 optdata = (void *)&pktopt->ip6po_tclass; 2167 else 2168 optdata = (void *)&deftclass; 2169 optdatalen = sizeof(int); 2170 break; 2171 case IPV6_HOPOPTS: 2172 if (pktopt && pktopt->ip6po_hbh) { 2173 optdata = (void *)pktopt->ip6po_hbh; 2174 ip6e = (struct ip6_ext *)pktopt->ip6po_hbh; 2175 optdatalen = (ip6e->ip6e_len + 1) << 3; 2176 } 2177 break; 2178 case IPV6_RTHDR: 2179 if (pktopt && pktopt->ip6po_rthdr) { 2180 optdata = (void *)pktopt->ip6po_rthdr; 2181 ip6e = (struct ip6_ext *)pktopt->ip6po_rthdr; 2182 optdatalen = (ip6e->ip6e_len + 1) << 3; 2183 } 2184 break; 2185 case IPV6_RTHDRDSTOPTS: 2186 if (pktopt && pktopt->ip6po_dest1) { 2187 optdata = (void *)pktopt->ip6po_dest1; 2188 ip6e = (struct ip6_ext *)pktopt->ip6po_dest1; 2189 optdatalen = (ip6e->ip6e_len + 1) << 3; 2190 } 2191 break; 2192 case IPV6_DSTOPTS: 2193 if (pktopt && pktopt->ip6po_dest2) { 2194 optdata = (void *)pktopt->ip6po_dest2; 2195 ip6e = (struct ip6_ext *)pktopt->ip6po_dest2; 2196 optdatalen = (ip6e->ip6e_len + 1) << 3; 2197 } 2198 break; 2199 case IPV6_NEXTHOP: 2200 if (pktopt && pktopt->ip6po_nexthop) { 2201 optdata = (void *)pktopt->ip6po_nexthop; 2202 optdatalen = pktopt->ip6po_nexthop->sa_len; 2203 } 2204 break; 2205 case IPV6_USE_MIN_MTU: 2206 if (pktopt) 2207 optdata = (void *)&pktopt->ip6po_minmtu; 2208 else 2209 optdata = (void *)&defminmtu; 2210 optdatalen = sizeof(int); 2211 break; 2212 case IPV6_DONTFRAG: 2213 if (pktopt && ((pktopt->ip6po_flags) & IP6PO_DONTFRAG)) 2214 on = 1; 2215 else 2216 on = 0; 2217 optdata = (void *)&on; 2218 optdatalen = sizeof(on); 2219 break; 2220 case IPV6_PREFER_TEMPADDR: 2221 if (pktopt) 2222 optdata = (void *)&pktopt->ip6po_prefer_tempaddr; 2223 else 2224 optdata = (void *)&defpreftemp; 2225 optdatalen = sizeof(int); 2226 break; 2227 default: /* should not happen */ 2228 #ifdef DIAGNOSTIC 2229 panic("ip6_getpcbopt: unexpected option\n"); 2230 #endif 2231 return (ENOPROTOOPT); 2232 } 2233 2234 error = sockopt_set(sopt, optdata, optdatalen); 2235 2236 return (error); 2237 } 2238 2239 void 2240 ip6_clearpktopts(struct ip6_pktopts *pktopt, int optname) 2241 { 2242 if (optname == -1 || optname == IPV6_PKTINFO) { 2243 if (pktopt->ip6po_pktinfo) 2244 free(pktopt->ip6po_pktinfo, M_IP6OPT); 2245 pktopt->ip6po_pktinfo = NULL; 2246 } 2247 if (optname == -1 || optname == IPV6_HOPLIMIT) 2248 pktopt->ip6po_hlim = -1; 2249 if (optname == -1 || optname == IPV6_TCLASS) 2250 pktopt->ip6po_tclass = -1; 2251 if (optname == -1 || optname == IPV6_NEXTHOP) { 2252 rtcache_free(&pktopt->ip6po_nextroute); 2253 if (pktopt->ip6po_nexthop) 2254 free(pktopt->ip6po_nexthop, M_IP6OPT); 2255 pktopt->ip6po_nexthop = NULL; 2256 } 2257 if (optname == -1 || optname == IPV6_HOPOPTS) { 2258 if (pktopt->ip6po_hbh) 2259 free(pktopt->ip6po_hbh, M_IP6OPT); 2260 pktopt->ip6po_hbh = NULL; 2261 } 2262 if (optname == -1 || optname == IPV6_RTHDRDSTOPTS) { 2263 if (pktopt->ip6po_dest1) 2264 free(pktopt->ip6po_dest1, M_IP6OPT); 2265 pktopt->ip6po_dest1 = NULL; 2266 } 2267 if (optname == -1 || optname == IPV6_RTHDR) { 2268 if (pktopt->ip6po_rhinfo.ip6po_rhi_rthdr) 2269 free(pktopt->ip6po_rhinfo.ip6po_rhi_rthdr, M_IP6OPT); 2270 pktopt->ip6po_rhinfo.ip6po_rhi_rthdr = NULL; 2271 rtcache_free(&pktopt->ip6po_route); 2272 } 2273 if (optname == -1 || optname == IPV6_DSTOPTS) { 2274 if (pktopt->ip6po_dest2) 2275 free(pktopt->ip6po_dest2, M_IP6OPT); 2276 pktopt->ip6po_dest2 = NULL; 2277 } 2278 } 2279 2280 #define PKTOPT_EXTHDRCPY(type) \ 2281 do { \ 2282 if (src->type) { \ 2283 int hlen = (((struct ip6_ext *)src->type)->ip6e_len + 1) << 3;\ 2284 dst->type = malloc(hlen, M_IP6OPT, canwait); \ 2285 if (dst->type == NULL) \ 2286 goto bad; \ 2287 memcpy(dst->type, src->type, hlen); \ 2288 } \ 2289 } while (/*CONSTCOND*/ 0) 2290 2291 static int 2292 copypktopts(struct ip6_pktopts *dst, struct ip6_pktopts *src, int canwait) 2293 { 2294 dst->ip6po_hlim = src->ip6po_hlim; 2295 dst->ip6po_tclass = src->ip6po_tclass; 2296 dst->ip6po_flags = src->ip6po_flags; 2297 dst->ip6po_minmtu = src->ip6po_minmtu; 2298 dst->ip6po_prefer_tempaddr = src->ip6po_prefer_tempaddr; 2299 if (src->ip6po_pktinfo) { 2300 dst->ip6po_pktinfo = malloc(sizeof(*dst->ip6po_pktinfo), 2301 M_IP6OPT, canwait); 2302 if (dst->ip6po_pktinfo == NULL) 2303 goto bad; 2304 *dst->ip6po_pktinfo = *src->ip6po_pktinfo; 2305 } 2306 if (src->ip6po_nexthop) { 2307 dst->ip6po_nexthop = malloc(src->ip6po_nexthop->sa_len, 2308 M_IP6OPT, canwait); 2309 if (dst->ip6po_nexthop == NULL) 2310 goto bad; 2311 memcpy(dst->ip6po_nexthop, src->ip6po_nexthop, 2312 src->ip6po_nexthop->sa_len); 2313 } 2314 PKTOPT_EXTHDRCPY(ip6po_hbh); 2315 PKTOPT_EXTHDRCPY(ip6po_dest1); 2316 PKTOPT_EXTHDRCPY(ip6po_dest2); 2317 PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */ 2318 return (0); 2319 2320 bad: 2321 if (dst->ip6po_pktinfo) free(dst->ip6po_pktinfo, M_IP6OPT); 2322 if (dst->ip6po_nexthop) free(dst->ip6po_nexthop, M_IP6OPT); 2323 if (dst->ip6po_hbh) free(dst->ip6po_hbh, M_IP6OPT); 2324 if (dst->ip6po_dest1) free(dst->ip6po_dest1, M_IP6OPT); 2325 if (dst->ip6po_dest2) free(dst->ip6po_dest2, M_IP6OPT); 2326 if (dst->ip6po_rthdr) free(dst->ip6po_rthdr, M_IP6OPT); 2327 2328 return (ENOBUFS); 2329 } 2330 #undef PKTOPT_EXTHDRCPY 2331 2332 struct ip6_pktopts * 2333 ip6_copypktopts(struct ip6_pktopts *src, int canwait) 2334 { 2335 int error; 2336 struct ip6_pktopts *dst; 2337 2338 dst = malloc(sizeof(*dst), M_IP6OPT, canwait); 2339 if (dst == NULL) 2340 return (NULL); 2341 ip6_initpktopts(dst); 2342 2343 if ((error = copypktopts(dst, src, canwait)) != 0) { 2344 free(dst, M_IP6OPT); 2345 return (NULL); 2346 } 2347 2348 return (dst); 2349 } 2350 2351 void 2352 ip6_freepcbopts(struct ip6_pktopts *pktopt) 2353 { 2354 if (pktopt == NULL) 2355 return; 2356 2357 ip6_clearpktopts(pktopt, -1); 2358 2359 free(pktopt, M_IP6OPT); 2360 } 2361 2362 int 2363 ip6_get_membership(const struct sockopt *sopt, struct ifnet **ifp, 2364 struct psref *psref, void *v, size_t l) 2365 { 2366 struct ipv6_mreq mreq; 2367 int error; 2368 struct in6_addr *ia = &mreq.ipv6mr_multiaddr; 2369 struct in_addr *ia4 = (void *)&ia->s6_addr32[3]; 2370 2371 error = sockopt_get(sopt, &mreq, sizeof(mreq)); 2372 if (error != 0) 2373 return error; 2374 2375 if (IN6_IS_ADDR_UNSPECIFIED(ia)) { 2376 /* 2377 * We use the unspecified address to specify to accept 2378 * all multicast addresses. Only super user is allowed 2379 * to do this. 2380 */ 2381 if (kauth_authorize_network(curlwp->l_cred, KAUTH_NETWORK_IPV6, 2382 KAUTH_REQ_NETWORK_IPV6_JOIN_MULTICAST, NULL, NULL, NULL)) 2383 return EACCES; 2384 } else if (IN6_IS_ADDR_V4MAPPED(ia)) { 2385 // Don't bother if we are not going to use ifp. 2386 if (l == sizeof(*ia)) { 2387 memcpy(v, ia, l); 2388 return 0; 2389 } 2390 } else if (!IN6_IS_ADDR_MULTICAST(ia)) { 2391 return EINVAL; 2392 } 2393 2394 /* 2395 * If no interface was explicitly specified, choose an 2396 * appropriate one according to the given multicast address. 2397 */ 2398 if (mreq.ipv6mr_interface == 0) { 2399 struct rtentry *rt; 2400 union { 2401 struct sockaddr dst; 2402 struct sockaddr_in dst4; 2403 struct sockaddr_in6 dst6; 2404 } u; 2405 struct route ro; 2406 2407 /* 2408 * Look up the routing table for the 2409 * address, and choose the outgoing interface. 2410 * XXX: is it a good approach? 2411 */ 2412 memset(&ro, 0, sizeof(ro)); 2413 if (IN6_IS_ADDR_V4MAPPED(ia)) 2414 sockaddr_in_init(&u.dst4, ia4, 0); 2415 else 2416 sockaddr_in6_init(&u.dst6, ia, 0, 0, 0); 2417 error = rtcache_setdst(&ro, &u.dst); 2418 if (error != 0) 2419 return error; 2420 rt = rtcache_init(&ro); 2421 *ifp = rt != NULL ? 2422 if_get_byindex(rt->rt_ifp->if_index, psref) : NULL; 2423 rtcache_unref(rt, &ro); 2424 rtcache_free(&ro); 2425 } else { 2426 /* 2427 * If the interface is specified, validate it. 2428 */ 2429 *ifp = if_get_byindex(mreq.ipv6mr_interface, psref); 2430 if (*ifp == NULL) 2431 return ENXIO; /* XXX EINVAL? */ 2432 } 2433 if (sizeof(*ia) == l) 2434 memcpy(v, ia, l); 2435 else 2436 memcpy(v, ia4, l); 2437 return 0; 2438 } 2439 2440 /* 2441 * Set the IP6 multicast options in response to user setsockopt(). 2442 */ 2443 static int 2444 ip6_setmoptions(const struct sockopt *sopt, struct in6pcb *in6p) 2445 { 2446 int error = 0; 2447 u_int loop, ifindex; 2448 struct ipv6_mreq mreq; 2449 struct in6_addr ia; 2450 struct ifnet *ifp; 2451 struct ip6_moptions *im6o = in6p->in6p_moptions; 2452 struct in6_multi_mship *imm; 2453 2454 KASSERT(in6p_locked(in6p)); 2455 2456 if (im6o == NULL) { 2457 /* 2458 * No multicast option buffer attached to the pcb; 2459 * allocate one and initialize to default values. 2460 */ 2461 im6o = malloc(sizeof(*im6o), M_IPMOPTS, M_NOWAIT); 2462 if (im6o == NULL) 2463 return (ENOBUFS); 2464 in6p->in6p_moptions = im6o; 2465 im6o->im6o_multicast_if_index = 0; 2466 im6o->im6o_multicast_hlim = ip6_defmcasthlim; 2467 im6o->im6o_multicast_loop = IPV6_DEFAULT_MULTICAST_LOOP; 2468 LIST_INIT(&im6o->im6o_memberships); 2469 } 2470 2471 switch (sopt->sopt_name) { 2472 2473 case IPV6_MULTICAST_IF: { 2474 int s; 2475 /* 2476 * Select the interface for outgoing multicast packets. 2477 */ 2478 error = sockopt_get(sopt, &ifindex, sizeof(ifindex)); 2479 if (error != 0) 2480 break; 2481 2482 s = pserialize_read_enter(); 2483 if (ifindex != 0) { 2484 if ((ifp = if_byindex(ifindex)) == NULL) { 2485 pserialize_read_exit(s); 2486 error = ENXIO; /* XXX EINVAL? */ 2487 break; 2488 } 2489 if ((ifp->if_flags & IFF_MULTICAST) == 0) { 2490 pserialize_read_exit(s); 2491 error = EADDRNOTAVAIL; 2492 break; 2493 } 2494 } else 2495 ifp = NULL; 2496 im6o->im6o_multicast_if_index = if_get_index(ifp); 2497 pserialize_read_exit(s); 2498 break; 2499 } 2500 2501 case IPV6_MULTICAST_HOPS: 2502 { 2503 /* 2504 * Set the IP6 hoplimit for outgoing multicast packets. 2505 */ 2506 int optval; 2507 2508 error = sockopt_getint(sopt, &optval); 2509 if (error != 0) 2510 break; 2511 2512 if (optval < -1 || optval >= 256) 2513 error = EINVAL; 2514 else if (optval == -1) 2515 im6o->im6o_multicast_hlim = ip6_defmcasthlim; 2516 else 2517 im6o->im6o_multicast_hlim = optval; 2518 break; 2519 } 2520 2521 case IPV6_MULTICAST_LOOP: 2522 /* 2523 * Set the loopback flag for outgoing multicast packets. 2524 * Must be zero or one. 2525 */ 2526 error = sockopt_get(sopt, &loop, sizeof(loop)); 2527 if (error != 0) 2528 break; 2529 if (loop > 1) { 2530 error = EINVAL; 2531 break; 2532 } 2533 im6o->im6o_multicast_loop = loop; 2534 break; 2535 2536 case IPV6_JOIN_GROUP: { 2537 int bound; 2538 struct psref psref; 2539 /* 2540 * Add a multicast group membership. 2541 * Group must be a valid IP6 multicast address. 2542 */ 2543 bound = curlwp_bind(); 2544 ifp = NULL; 2545 error = ip6_get_membership(sopt, &ifp, &psref, &ia, sizeof(ia)); 2546 if (error != 0) { 2547 KASSERT(ifp == NULL); 2548 curlwp_bindx(bound); 2549 return error; 2550 } 2551 2552 if (IN6_IS_ADDR_V4MAPPED(&ia)) { 2553 error = ip_setmoptions(&in6p->in6p_v4moptions, sopt); 2554 goto put_break; 2555 } 2556 /* 2557 * See if we found an interface, and confirm that it 2558 * supports multicast 2559 */ 2560 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) { 2561 error = EADDRNOTAVAIL; 2562 goto put_break; 2563 } 2564 2565 if (in6_setscope(&ia, ifp, NULL)) { 2566 error = EADDRNOTAVAIL; /* XXX: should not happen */ 2567 goto put_break; 2568 } 2569 2570 /* 2571 * See if the membership already exists. 2572 */ 2573 LIST_FOREACH(imm, &im6o->im6o_memberships, i6mm_chain) { 2574 if (imm->i6mm_maddr->in6m_ifp == ifp && 2575 IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr, 2576 &ia)) 2577 goto put_break; 2578 } 2579 if (imm != NULL) { 2580 error = EADDRINUSE; 2581 goto put_break; 2582 } 2583 /* 2584 * Everything looks good; add a new record to the multicast 2585 * address list for the given interface. 2586 */ 2587 IFNET_LOCK(ifp); 2588 imm = in6_joingroup(ifp, &ia, &error, 0); 2589 IFNET_UNLOCK(ifp); 2590 if (imm == NULL) 2591 goto put_break; 2592 LIST_INSERT_HEAD(&im6o->im6o_memberships, imm, i6mm_chain); 2593 put_break: 2594 if_put(ifp, &psref); 2595 curlwp_bindx(bound); 2596 break; 2597 } 2598 2599 case IPV6_LEAVE_GROUP: { 2600 struct ifnet *in6m_ifp; 2601 /* 2602 * Drop a multicast group membership. 2603 * Group must be a valid IP6 multicast address. 2604 */ 2605 error = sockopt_get(sopt, &mreq, sizeof(mreq)); 2606 if (error != 0) 2607 break; 2608 2609 if (IN6_IS_ADDR_V4MAPPED(&mreq.ipv6mr_multiaddr)) { 2610 error = ip_setmoptions(&in6p->in6p_v4moptions, sopt); 2611 break; 2612 } 2613 /* 2614 * If an interface address was specified, get a pointer 2615 * to its ifnet structure. 2616 */ 2617 if (mreq.ipv6mr_interface != 0) { 2618 if ((ifp = if_byindex(mreq.ipv6mr_interface)) == NULL) { 2619 error = ENXIO; /* XXX EINVAL? */ 2620 break; 2621 } 2622 } else 2623 ifp = NULL; 2624 2625 /* Fill in the scope zone ID */ 2626 if (ifp) { 2627 if (in6_setscope(&mreq.ipv6mr_multiaddr, ifp, NULL)) { 2628 /* XXX: should not happen */ 2629 error = EADDRNOTAVAIL; 2630 break; 2631 } 2632 } else if (mreq.ipv6mr_interface != 0) { 2633 /* 2634 * XXX: This case would happens when the (positive) 2635 * index is in the valid range, but the corresponding 2636 * interface has been detached dynamically. The above 2637 * check probably avoids such case to happen here, but 2638 * we check it explicitly for safety. 2639 */ 2640 error = EADDRNOTAVAIL; 2641 break; 2642 } else { /* ipv6mr_interface == 0 */ 2643 struct sockaddr_in6 sa6_mc; 2644 2645 /* 2646 * The API spec says as follows: 2647 * If the interface index is specified as 0, the 2648 * system may choose a multicast group membership to 2649 * drop by matching the multicast address only. 2650 * On the other hand, we cannot disambiguate the scope 2651 * zone unless an interface is provided. Thus, we 2652 * check if there's ambiguity with the default scope 2653 * zone as the last resort. 2654 */ 2655 sockaddr_in6_init(&sa6_mc, &mreq.ipv6mr_multiaddr, 2656 0, 0, 0); 2657 error = sa6_embedscope(&sa6_mc, ip6_use_defzone); 2658 if (error != 0) 2659 break; 2660 mreq.ipv6mr_multiaddr = sa6_mc.sin6_addr; 2661 } 2662 2663 /* 2664 * Find the membership in the membership list. 2665 */ 2666 LIST_FOREACH(imm, &im6o->im6o_memberships, i6mm_chain) { 2667 if ((ifp == NULL || imm->i6mm_maddr->in6m_ifp == ifp) && 2668 IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr, 2669 &mreq.ipv6mr_multiaddr)) 2670 break; 2671 } 2672 if (imm == NULL) { 2673 /* Unable to resolve interface */ 2674 error = EADDRNOTAVAIL; 2675 break; 2676 } 2677 /* 2678 * Give up the multicast address record to which the 2679 * membership points. 2680 */ 2681 LIST_REMOVE(imm, i6mm_chain); 2682 in6m_ifp = imm->i6mm_maddr->in6m_ifp; 2683 IFNET_LOCK(in6m_ifp); 2684 in6_leavegroup(imm); 2685 /* in6m_ifp should not leave thanks to in6p_lock */ 2686 IFNET_UNLOCK(in6m_ifp); 2687 break; 2688 } 2689 2690 default: 2691 error = EOPNOTSUPP; 2692 break; 2693 } 2694 2695 /* 2696 * If all options have default values, no need to keep the mbuf. 2697 */ 2698 if (im6o->im6o_multicast_if_index == 0 && 2699 im6o->im6o_multicast_hlim == ip6_defmcasthlim && 2700 im6o->im6o_multicast_loop == IPV6_DEFAULT_MULTICAST_LOOP && 2701 LIST_EMPTY(&im6o->im6o_memberships)) { 2702 free(in6p->in6p_moptions, M_IPMOPTS); 2703 in6p->in6p_moptions = NULL; 2704 } 2705 2706 return (error); 2707 } 2708 2709 /* 2710 * Return the IP6 multicast options in response to user getsockopt(). 2711 */ 2712 static int 2713 ip6_getmoptions(struct sockopt *sopt, struct in6pcb *in6p) 2714 { 2715 u_int optval; 2716 int error; 2717 struct ip6_moptions *im6o = in6p->in6p_moptions; 2718 2719 switch (sopt->sopt_name) { 2720 case IPV6_MULTICAST_IF: 2721 if (im6o == NULL || im6o->im6o_multicast_if_index == 0) 2722 optval = 0; 2723 else 2724 optval = im6o->im6o_multicast_if_index; 2725 2726 error = sockopt_set(sopt, &optval, sizeof(optval)); 2727 break; 2728 2729 case IPV6_MULTICAST_HOPS: 2730 if (im6o == NULL) 2731 optval = ip6_defmcasthlim; 2732 else 2733 optval = im6o->im6o_multicast_hlim; 2734 2735 error = sockopt_set(sopt, &optval, sizeof(optval)); 2736 break; 2737 2738 case IPV6_MULTICAST_LOOP: 2739 if (im6o == NULL) 2740 optval = IPV6_DEFAULT_MULTICAST_LOOP; 2741 else 2742 optval = im6o->im6o_multicast_loop; 2743 2744 error = sockopt_set(sopt, &optval, sizeof(optval)); 2745 break; 2746 2747 default: 2748 error = EOPNOTSUPP; 2749 } 2750 2751 return (error); 2752 } 2753 2754 /* 2755 * Discard the IP6 multicast options. 2756 */ 2757 void 2758 ip6_freemoptions(struct ip6_moptions *im6o) 2759 { 2760 struct in6_multi_mship *imm, *nimm; 2761 2762 if (im6o == NULL) 2763 return; 2764 2765 /* The owner of im6o (in6p) should be protected by solock */ 2766 LIST_FOREACH_SAFE(imm, &im6o->im6o_memberships, i6mm_chain, nimm) { 2767 struct ifnet *ifp; 2768 2769 LIST_REMOVE(imm, i6mm_chain); 2770 2771 ifp = imm->i6mm_maddr->in6m_ifp; 2772 IFNET_LOCK(ifp); 2773 in6_leavegroup(imm); 2774 /* ifp should not leave thanks to solock */ 2775 IFNET_UNLOCK(ifp); 2776 } 2777 free(im6o, M_IPMOPTS); 2778 } 2779 2780 /* 2781 * Set IPv6 outgoing packet options based on advanced API. 2782 */ 2783 int 2784 ip6_setpktopts(struct mbuf *control, struct ip6_pktopts *opt, 2785 struct ip6_pktopts *stickyopt, kauth_cred_t cred, int uproto) 2786 { 2787 struct cmsghdr *cm = 0; 2788 2789 if (control == NULL || opt == NULL) 2790 return (EINVAL); 2791 2792 ip6_initpktopts(opt); 2793 if (stickyopt) { 2794 int error; 2795 2796 /* 2797 * If stickyopt is provided, make a local copy of the options 2798 * for this particular packet, then override them by ancillary 2799 * objects. 2800 * XXX: copypktopts() does not copy the cached route to a next 2801 * hop (if any). This is not very good in terms of efficiency, 2802 * but we can allow this since this option should be rarely 2803 * used. 2804 */ 2805 if ((error = copypktopts(opt, stickyopt, M_NOWAIT)) != 0) 2806 return (error); 2807 } 2808 2809 /* 2810 * XXX: Currently, we assume all the optional information is stored 2811 * in a single mbuf. 2812 */ 2813 if (control->m_next) 2814 return (EINVAL); 2815 2816 /* XXX if cm->cmsg_len is not aligned, control->m_len can become <0 */ 2817 for (; control->m_len > 0; control->m_data += CMSG_ALIGN(cm->cmsg_len), 2818 control->m_len -= CMSG_ALIGN(cm->cmsg_len)) { 2819 int error; 2820 2821 if (control->m_len < CMSG_LEN(0)) 2822 return (EINVAL); 2823 2824 cm = mtod(control, struct cmsghdr *); 2825 if (cm->cmsg_len == 0 || cm->cmsg_len > control->m_len) 2826 return (EINVAL); 2827 if (cm->cmsg_level != IPPROTO_IPV6) 2828 continue; 2829 2830 error = ip6_setpktopt(cm->cmsg_type, CMSG_DATA(cm), 2831 cm->cmsg_len - CMSG_LEN(0), opt, cred, 0, 1, uproto); 2832 if (error) 2833 return (error); 2834 } 2835 2836 return (0); 2837 } 2838 2839 /* 2840 * Set a particular packet option, as a sticky option or an ancillary data 2841 * item. "len" can be 0 only when it's a sticky option. 2842 * We have 4 cases of combination of "sticky" and "cmsg": 2843 * "sticky=0, cmsg=0": impossible 2844 * "sticky=0, cmsg=1": RFC2292 or RFC3542 ancillary data 2845 * "sticky=1, cmsg=0": RFC3542 socket option 2846 * "sticky=1, cmsg=1": RFC2292 socket option 2847 */ 2848 static int 2849 ip6_setpktopt(int optname, u_char *buf, int len, struct ip6_pktopts *opt, 2850 kauth_cred_t cred, int sticky, int cmsg, int uproto) 2851 { 2852 int minmtupolicy; 2853 int error; 2854 2855 if (!sticky && !cmsg) { 2856 #ifdef DIAGNOSTIC 2857 printf("ip6_setpktopt: impossible case\n"); 2858 #endif 2859 return (EINVAL); 2860 } 2861 2862 /* 2863 * IPV6_2292xxx is for backward compatibility to RFC2292, and should 2864 * not be specified in the context of RFC3542. Conversely, 2865 * RFC3542 types should not be specified in the context of RFC2292. 2866 */ 2867 if (!cmsg) { 2868 switch (optname) { 2869 case IPV6_2292PKTINFO: 2870 case IPV6_2292HOPLIMIT: 2871 case IPV6_2292NEXTHOP: 2872 case IPV6_2292HOPOPTS: 2873 case IPV6_2292DSTOPTS: 2874 case IPV6_2292RTHDR: 2875 case IPV6_2292PKTOPTIONS: 2876 return (ENOPROTOOPT); 2877 } 2878 } 2879 if (sticky && cmsg) { 2880 switch (optname) { 2881 case IPV6_PKTINFO: 2882 case IPV6_HOPLIMIT: 2883 case IPV6_NEXTHOP: 2884 case IPV6_HOPOPTS: 2885 case IPV6_DSTOPTS: 2886 case IPV6_RTHDRDSTOPTS: 2887 case IPV6_RTHDR: 2888 case IPV6_USE_MIN_MTU: 2889 case IPV6_DONTFRAG: 2890 case IPV6_OTCLASS: 2891 case IPV6_TCLASS: 2892 case IPV6_PREFER_TEMPADDR: /* XXX not an RFC3542 option */ 2893 return (ENOPROTOOPT); 2894 } 2895 } 2896 2897 switch (optname) { 2898 #ifdef RFC2292 2899 case IPV6_2292PKTINFO: 2900 #endif 2901 case IPV6_PKTINFO: 2902 { 2903 struct in6_pktinfo *pktinfo; 2904 2905 if (len != sizeof(struct in6_pktinfo)) 2906 return (EINVAL); 2907 2908 pktinfo = (struct in6_pktinfo *)buf; 2909 2910 /* 2911 * An application can clear any sticky IPV6_PKTINFO option by 2912 * doing a "regular" setsockopt with ipi6_addr being 2913 * in6addr_any and ipi6_ifindex being zero. 2914 * [RFC 3542, Section 6] 2915 */ 2916 if (optname == IPV6_PKTINFO && opt->ip6po_pktinfo && 2917 pktinfo->ipi6_ifindex == 0 && 2918 IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) { 2919 ip6_clearpktopts(opt, optname); 2920 break; 2921 } 2922 2923 if (uproto == IPPROTO_TCP && optname == IPV6_PKTINFO && 2924 sticky && !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) { 2925 return (EINVAL); 2926 } 2927 2928 /* Validate the interface index if specified. */ 2929 if (pktinfo->ipi6_ifindex) { 2930 struct ifnet *ifp; 2931 int s = pserialize_read_enter(); 2932 ifp = if_byindex(pktinfo->ipi6_ifindex); 2933 if (ifp == NULL) { 2934 pserialize_read_exit(s); 2935 return ENXIO; 2936 } 2937 pserialize_read_exit(s); 2938 } 2939 2940 /* 2941 * We store the address anyway, and let in6_selectsrc() 2942 * validate the specified address. This is because ipi6_addr 2943 * may not have enough information about its scope zone, and 2944 * we may need additional information (such as outgoing 2945 * interface or the scope zone of a destination address) to 2946 * disambiguate the scope. 2947 * XXX: the delay of the validation may confuse the 2948 * application when it is used as a sticky option. 2949 */ 2950 if (opt->ip6po_pktinfo == NULL) { 2951 opt->ip6po_pktinfo = malloc(sizeof(*pktinfo), 2952 M_IP6OPT, M_NOWAIT); 2953 if (opt->ip6po_pktinfo == NULL) 2954 return (ENOBUFS); 2955 } 2956 memcpy(opt->ip6po_pktinfo, pktinfo, sizeof(*pktinfo)); 2957 break; 2958 } 2959 2960 #ifdef RFC2292 2961 case IPV6_2292HOPLIMIT: 2962 #endif 2963 case IPV6_HOPLIMIT: 2964 { 2965 int *hlimp; 2966 2967 /* 2968 * RFC 3542 deprecated the usage of sticky IPV6_HOPLIMIT 2969 * to simplify the ordering among hoplimit options. 2970 */ 2971 if (optname == IPV6_HOPLIMIT && sticky) 2972 return (ENOPROTOOPT); 2973 2974 if (len != sizeof(int)) 2975 return (EINVAL); 2976 hlimp = (int *)buf; 2977 if (*hlimp < -1 || *hlimp > 255) 2978 return (EINVAL); 2979 2980 opt->ip6po_hlim = *hlimp; 2981 break; 2982 } 2983 2984 case IPV6_OTCLASS: 2985 if (len != sizeof(u_int8_t)) 2986 return (EINVAL); 2987 2988 opt->ip6po_tclass = *(u_int8_t *)buf; 2989 break; 2990 2991 case IPV6_TCLASS: 2992 { 2993 int tclass; 2994 2995 if (len != sizeof(int)) 2996 return (EINVAL); 2997 tclass = *(int *)buf; 2998 if (tclass < -1 || tclass > 255) 2999 return (EINVAL); 3000 3001 opt->ip6po_tclass = tclass; 3002 break; 3003 } 3004 3005 #ifdef RFC2292 3006 case IPV6_2292NEXTHOP: 3007 #endif 3008 case IPV6_NEXTHOP: 3009 error = kauth_authorize_network(cred, KAUTH_NETWORK_IPV6, 3010 KAUTH_REQ_NETWORK_IPV6_HOPBYHOP, NULL, NULL, NULL); 3011 if (error) 3012 return (error); 3013 3014 if (len == 0) { /* just remove the option */ 3015 ip6_clearpktopts(opt, IPV6_NEXTHOP); 3016 break; 3017 } 3018 3019 /* check if cmsg_len is large enough for sa_len */ 3020 if (len < sizeof(struct sockaddr) || len < *buf) 3021 return (EINVAL); 3022 3023 switch (((struct sockaddr *)buf)->sa_family) { 3024 case AF_INET6: 3025 { 3026 struct sockaddr_in6 *sa6 = (struct sockaddr_in6 *)buf; 3027 3028 if (sa6->sin6_len != sizeof(struct sockaddr_in6)) 3029 return (EINVAL); 3030 3031 if (IN6_IS_ADDR_UNSPECIFIED(&sa6->sin6_addr) || 3032 IN6_IS_ADDR_MULTICAST(&sa6->sin6_addr)) { 3033 return (EINVAL); 3034 } 3035 if ((error = sa6_embedscope(sa6, ip6_use_defzone)) 3036 != 0) { 3037 return (error); 3038 } 3039 break; 3040 } 3041 case AF_LINK: /* eventually be supported? */ 3042 default: 3043 return (EAFNOSUPPORT); 3044 } 3045 3046 /* turn off the previous option, then set the new option. */ 3047 ip6_clearpktopts(opt, IPV6_NEXTHOP); 3048 opt->ip6po_nexthop = malloc(*buf, M_IP6OPT, M_NOWAIT); 3049 if (opt->ip6po_nexthop == NULL) 3050 return (ENOBUFS); 3051 memcpy(opt->ip6po_nexthop, buf, *buf); 3052 break; 3053 3054 #ifdef RFC2292 3055 case IPV6_2292HOPOPTS: 3056 #endif 3057 case IPV6_HOPOPTS: 3058 { 3059 struct ip6_hbh *hbh; 3060 int hbhlen; 3061 3062 /* 3063 * XXX: We don't allow a non-privileged user to set ANY HbH 3064 * options, since per-option restriction has too much 3065 * overhead. 3066 */ 3067 error = kauth_authorize_network(cred, KAUTH_NETWORK_IPV6, 3068 KAUTH_REQ_NETWORK_IPV6_HOPBYHOP, NULL, NULL, NULL); 3069 if (error) 3070 return (error); 3071 3072 if (len == 0) { 3073 ip6_clearpktopts(opt, IPV6_HOPOPTS); 3074 break; /* just remove the option */ 3075 } 3076 3077 /* message length validation */ 3078 if (len < sizeof(struct ip6_hbh)) 3079 return (EINVAL); 3080 hbh = (struct ip6_hbh *)buf; 3081 hbhlen = (hbh->ip6h_len + 1) << 3; 3082 if (len != hbhlen) 3083 return (EINVAL); 3084 3085 /* turn off the previous option, then set the new option. */ 3086 ip6_clearpktopts(opt, IPV6_HOPOPTS); 3087 opt->ip6po_hbh = malloc(hbhlen, M_IP6OPT, M_NOWAIT); 3088 if (opt->ip6po_hbh == NULL) 3089 return (ENOBUFS); 3090 memcpy(opt->ip6po_hbh, hbh, hbhlen); 3091 3092 break; 3093 } 3094 3095 #ifdef RFC2292 3096 case IPV6_2292DSTOPTS: 3097 #endif 3098 case IPV6_DSTOPTS: 3099 case IPV6_RTHDRDSTOPTS: 3100 { 3101 struct ip6_dest *dest, **newdest = NULL; 3102 int destlen; 3103 3104 /* XXX: see the comment for IPV6_HOPOPTS */ 3105 error = kauth_authorize_network(cred, KAUTH_NETWORK_IPV6, 3106 KAUTH_REQ_NETWORK_IPV6_HOPBYHOP, NULL, NULL, NULL); 3107 if (error) 3108 return (error); 3109 3110 if (len == 0) { 3111 ip6_clearpktopts(opt, optname); 3112 break; /* just remove the option */ 3113 } 3114 3115 /* message length validation */ 3116 if (len < sizeof(struct ip6_dest)) 3117 return (EINVAL); 3118 dest = (struct ip6_dest *)buf; 3119 destlen = (dest->ip6d_len + 1) << 3; 3120 if (len != destlen) 3121 return (EINVAL); 3122 /* 3123 * Determine the position that the destination options header 3124 * should be inserted; before or after the routing header. 3125 */ 3126 switch (optname) { 3127 case IPV6_2292DSTOPTS: 3128 /* 3129 * The old advanced API is ambiguous on this point. 3130 * Our approach is to determine the position based 3131 * according to the existence of a routing header. 3132 * Note, however, that this depends on the order of the 3133 * extension headers in the ancillary data; the 1st 3134 * part of the destination options header must appear 3135 * before the routing header in the ancillary data, 3136 * too. 3137 * RFC3542 solved the ambiguity by introducing 3138 * separate ancillary data or option types. 3139 */ 3140 if (opt->ip6po_rthdr == NULL) 3141 newdest = &opt->ip6po_dest1; 3142 else 3143 newdest = &opt->ip6po_dest2; 3144 break; 3145 case IPV6_RTHDRDSTOPTS: 3146 newdest = &opt->ip6po_dest1; 3147 break; 3148 case IPV6_DSTOPTS: 3149 newdest = &opt->ip6po_dest2; 3150 break; 3151 } 3152 3153 /* turn off the previous option, then set the new option. */ 3154 ip6_clearpktopts(opt, optname); 3155 *newdest = malloc(destlen, M_IP6OPT, M_NOWAIT); 3156 if (*newdest == NULL) 3157 return (ENOBUFS); 3158 memcpy(*newdest, dest, destlen); 3159 3160 break; 3161 } 3162 3163 #ifdef RFC2292 3164 case IPV6_2292RTHDR: 3165 #endif 3166 case IPV6_RTHDR: 3167 { 3168 struct ip6_rthdr *rth; 3169 int rthlen; 3170 3171 if (len == 0) { 3172 ip6_clearpktopts(opt, IPV6_RTHDR); 3173 break; /* just remove the option */ 3174 } 3175 3176 /* message length validation */ 3177 if (len < sizeof(struct ip6_rthdr)) 3178 return (EINVAL); 3179 rth = (struct ip6_rthdr *)buf; 3180 rthlen = (rth->ip6r_len + 1) << 3; 3181 if (len != rthlen) 3182 return (EINVAL); 3183 switch (rth->ip6r_type) { 3184 case IPV6_RTHDR_TYPE_0: 3185 /* Dropped, RFC5095. */ 3186 default: 3187 return (EINVAL); /* not supported */ 3188 } 3189 /* turn off the previous option */ 3190 ip6_clearpktopts(opt, IPV6_RTHDR); 3191 opt->ip6po_rthdr = malloc(rthlen, M_IP6OPT, M_NOWAIT); 3192 if (opt->ip6po_rthdr == NULL) 3193 return (ENOBUFS); 3194 memcpy(opt->ip6po_rthdr, rth, rthlen); 3195 break; 3196 } 3197 3198 case IPV6_USE_MIN_MTU: 3199 if (len != sizeof(int)) 3200 return (EINVAL); 3201 minmtupolicy = *(int *)buf; 3202 if (minmtupolicy != IP6PO_MINMTU_MCASTONLY && 3203 minmtupolicy != IP6PO_MINMTU_DISABLE && 3204 minmtupolicy != IP6PO_MINMTU_ALL) { 3205 return (EINVAL); 3206 } 3207 opt->ip6po_minmtu = minmtupolicy; 3208 break; 3209 3210 case IPV6_DONTFRAG: 3211 if (len != sizeof(int)) 3212 return (EINVAL); 3213 3214 if (uproto == IPPROTO_TCP || *(int *)buf == 0) { 3215 /* 3216 * we ignore this option for TCP sockets. 3217 * (RFC3542 leaves this case unspecified.) 3218 */ 3219 opt->ip6po_flags &= ~IP6PO_DONTFRAG; 3220 } else 3221 opt->ip6po_flags |= IP6PO_DONTFRAG; 3222 break; 3223 3224 case IPV6_PREFER_TEMPADDR: 3225 { 3226 int preftemp; 3227 3228 if (len != sizeof(int)) 3229 return (EINVAL); 3230 preftemp = *(int *)buf; 3231 switch (preftemp) { 3232 case IP6PO_TEMPADDR_SYSTEM: 3233 case IP6PO_TEMPADDR_NOTPREFER: 3234 case IP6PO_TEMPADDR_PREFER: 3235 break; 3236 default: 3237 return (EINVAL); 3238 } 3239 opt->ip6po_prefer_tempaddr = preftemp; 3240 break; 3241 } 3242 3243 default: 3244 return (ENOPROTOOPT); 3245 } /* end of switch */ 3246 3247 return (0); 3248 } 3249 3250 /* 3251 * Routine called from ip6_output() to loop back a copy of an IP6 multicast 3252 * packet to the input queue of a specified interface. Note that this 3253 * calls the output routine of the loopback "driver", but with an interface 3254 * pointer that might NOT be lo0ifp -- easier than replicating that code here. 3255 */ 3256 void 3257 ip6_mloopback(struct ifnet *ifp, struct mbuf *m, 3258 const struct sockaddr_in6 *dst) 3259 { 3260 struct mbuf *copym; 3261 struct ip6_hdr *ip6; 3262 3263 copym = m_copym(m, 0, M_COPYALL, M_DONTWAIT); 3264 if (copym == NULL) 3265 return; 3266 3267 /* 3268 * Make sure to deep-copy IPv6 header portion in case the data 3269 * is in an mbuf cluster, so that we can safely override the IPv6 3270 * header portion later. 3271 */ 3272 if ((copym->m_flags & M_EXT) != 0 || 3273 copym->m_len < sizeof(struct ip6_hdr)) { 3274 copym = m_pullup(copym, sizeof(struct ip6_hdr)); 3275 if (copym == NULL) 3276 return; 3277 } 3278 3279 #ifdef DIAGNOSTIC 3280 if (copym->m_len < sizeof(*ip6)) { 3281 m_freem(copym); 3282 return; 3283 } 3284 #endif 3285 3286 ip6 = mtod(copym, struct ip6_hdr *); 3287 /* 3288 * clear embedded scope identifiers if necessary. 3289 * in6_clearscope will touch the addresses only when necessary. 3290 */ 3291 in6_clearscope(&ip6->ip6_src); 3292 in6_clearscope(&ip6->ip6_dst); 3293 3294 (void)looutput(ifp, copym, (const struct sockaddr *)dst, NULL); 3295 } 3296 3297 /* 3298 * Chop IPv6 header off from the payload. 3299 */ 3300 static int 3301 ip6_splithdr(struct mbuf *m, struct ip6_exthdrs *exthdrs) 3302 { 3303 struct mbuf *mh; 3304 struct ip6_hdr *ip6; 3305 3306 ip6 = mtod(m, struct ip6_hdr *); 3307 if (m->m_len > sizeof(*ip6)) { 3308 MGETHDR(mh, M_DONTWAIT, MT_HEADER); 3309 if (mh == NULL) { 3310 m_freem(m); 3311 return ENOBUFS; 3312 } 3313 M_MOVE_PKTHDR(mh, m); 3314 MH_ALIGN(mh, sizeof(*ip6)); 3315 m->m_len -= sizeof(*ip6); 3316 m->m_data += sizeof(*ip6); 3317 mh->m_next = m; 3318 mh->m_len = sizeof(*ip6); 3319 memcpy(mtod(mh, void *), (void *)ip6, sizeof(*ip6)); 3320 m = mh; 3321 } 3322 exthdrs->ip6e_ip6 = m; 3323 return 0; 3324 } 3325 3326 /* 3327 * Compute IPv6 extension header length. 3328 */ 3329 int 3330 ip6_optlen(struct in6pcb *in6p) 3331 { 3332 int len; 3333 3334 if (!in6p->in6p_outputopts) 3335 return 0; 3336 3337 len = 0; 3338 #define elen(x) \ 3339 (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0) 3340 3341 len += elen(in6p->in6p_outputopts->ip6po_hbh); 3342 len += elen(in6p->in6p_outputopts->ip6po_dest1); 3343 len += elen(in6p->in6p_outputopts->ip6po_rthdr); 3344 len += elen(in6p->in6p_outputopts->ip6po_dest2); 3345 return len; 3346 #undef elen 3347 } 3348 3349 /* 3350 * Ensure sending address is valid. 3351 * Returns 0 on success, -1 if an error should be sent back or 1 3352 * if the packet could be dropped without error (protocol dependent). 3353 */ 3354 static int 3355 ip6_ifaddrvalid(const struct in6_addr *src, const struct in6_addr *dst) 3356 { 3357 struct sockaddr_in6 sin6; 3358 int s, error; 3359 struct ifaddr *ifa; 3360 struct in6_ifaddr *ia6; 3361 3362 if (IN6_IS_ADDR_UNSPECIFIED(src)) 3363 return 0; 3364 3365 memset(&sin6, 0, sizeof(sin6)); 3366 sin6.sin6_family = AF_INET6; 3367 sin6.sin6_len = sizeof(sin6); 3368 sin6.sin6_addr = *src; 3369 3370 s = pserialize_read_enter(); 3371 ifa = ifa_ifwithaddr(sin6tosa(&sin6)); 3372 if ((ia6 = ifatoia6(ifa)) == NULL || 3373 ia6->ia6_flags & (IN6_IFF_ANYCAST | IN6_IFF_DUPLICATED)) 3374 error = -1; 3375 else if (ia6->ia6_flags & IN6_IFF_TENTATIVE) 3376 error = 1; 3377 else if (ia6->ia6_flags & IN6_IFF_DETACHED && 3378 (sin6.sin6_addr = *dst, ifa_ifwithaddr(sin6tosa(&sin6)) == NULL)) 3379 /* Allow internal traffic to DETACHED addresses */ 3380 error = 1; 3381 else 3382 error = 0; 3383 pserialize_read_exit(s); 3384 3385 return error; 3386 } 3387
Indexes created Tue Sep 23 07:09:52 GMT 2025