ip6_output.c revision 1.231.2.2
1 /* $NetBSD: ip6_output.c,v 1.231.2.2 2023/08/04 14:28:01 martin 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 1.231.2.2 2023/08/04 14:28:01 martin Exp $"); 66 67 #ifdef _KERNEL_OPT 68 #include "opt_inet.h" 69 #include "opt_inet6.h" 70 #include "opt_ipsec.h" 71 #endif 72 73 #include <sys/param.h> 74 #include <sys/malloc.h> 75 #include <sys/mbuf.h> 76 #include <sys/errno.h> 77 #include <sys/socket.h> 78 #include <sys/socketvar.h> 79 #include <sys/syslog.h> 80 #include <sys/systm.h> 81 #include <sys/proc.h> 82 #include <sys/kauth.h> 83 84 #include <net/if.h> 85 #include <net/route.h> 86 #include <net/pfil.h> 87 88 #include <netinet/in.h> 89 #include <netinet/in_var.h> 90 #include <netinet/ip6.h> 91 #include <netinet/ip_var.h> 92 #include <netinet/icmp6.h> 93 #include <netinet/in_offload.h> 94 #include <netinet/portalgo.h> 95 #include <netinet6/in6_offload.h> 96 #include <netinet6/ip6_var.h> 97 #include <netinet6/ip6_private.h> 98 #include <netinet6/in6_pcb.h> 99 #include <netinet6/nd6.h> 100 #include <netinet6/ip6protosw.h> 101 #include <netinet6/scope6_var.h> 102 103 #ifdef IPSEC 104 #include <netipsec/ipsec.h> 105 #include <netipsec/ipsec6.h> 106 #include <netipsec/key.h> 107 #endif 108 109 extern pfil_head_t *inet6_pfil_hook; /* XXX */ 110 111 struct ip6_exthdrs { 112 struct mbuf *ip6e_ip6; 113 struct mbuf *ip6e_hbh; 114 struct mbuf *ip6e_dest1; 115 struct mbuf *ip6e_rthdr; 116 struct mbuf *ip6e_dest2; 117 }; 118 119 static int ip6_pcbopt(int, u_char *, int, struct ip6_pktopts **, 120 kauth_cred_t, int); 121 static int ip6_getpcbopt(struct ip6_pktopts *, int, struct sockopt *); 122 static int ip6_setpktopt(int, u_char *, int, struct ip6_pktopts *, kauth_cred_t, 123 int, int, int); 124 static int ip6_setmoptions(const struct sockopt *, struct inpcb *); 125 static int ip6_getmoptions(struct sockopt *, struct inpcb *); 126 static int ip6_copyexthdr(struct mbuf **, void *, int); 127 static int ip6_insertfraghdr(struct mbuf *, struct mbuf *, int, 128 struct ip6_frag **); 129 static int ip6_insert_jumboopt(struct ip6_exthdrs *, u_int32_t); 130 static int ip6_splithdr(struct mbuf *, struct ip6_exthdrs *); 131 static int ip6_getpmtu(struct rtentry *, struct ifnet *, u_long *, int *); 132 static int copypktopts(struct ip6_pktopts *, struct ip6_pktopts *, int); 133 static int ip6_ifaddrvalid(const struct in6_addr *, const struct in6_addr *); 134 static int ip6_handle_rthdr(struct ip6_rthdr *, struct ip6_hdr *); 135 136 #ifdef RFC2292 137 static int ip6_pcbopts(struct ip6_pktopts **, struct socket *, struct sockopt *); 138 #endif 139 140 static int 141 ip6_handle_rthdr(struct ip6_rthdr *rh, struct ip6_hdr *ip6) 142 { 143 int error = 0; 144 145 switch (rh->ip6r_type) { 146 case IPV6_RTHDR_TYPE_0: 147 /* Dropped, RFC5095. */ 148 default: /* is it possible? */ 149 error = EINVAL; 150 } 151 152 return error; 153 } 154 155 /* 156 * Send an IP packet to a host. 157 */ 158 int 159 ip6_if_output(struct ifnet * const ifp, struct ifnet * const origifp, 160 struct mbuf * const m, const struct sockaddr_in6 * const dst, 161 const struct rtentry *rt) 162 { 163 int error = 0; 164 165 if (rt != NULL) { 166 error = rt_check_reject_route(rt, ifp); 167 if (error != 0) { 168 IP6_STATINC(IP6_STAT_RTREJECT); 169 m_freem(m); 170 return error; 171 } 172 } 173 174 /* discard the packet if IPv6 operation is disabled on the interface */ 175 if ((ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED)) { 176 m_freem(m); 177 return ENETDOWN; /* better error? */ 178 } 179 180 if ((ifp->if_flags & IFF_LOOPBACK) != 0) 181 error = if_output_lock(ifp, origifp, m, sin6tocsa(dst), rt); 182 else 183 error = if_output_lock(ifp, ifp, m, sin6tocsa(dst), rt); 184 return error; 185 } 186 187 /* 188 * IP6 output. The packet in mbuf chain m contains a skeletal IP6 189 * header (with pri, len, nxt, hlim, src, dst). 190 * 191 * This function may modify ver and hlim only. The mbuf chain containing the 192 * packet will be freed. The mbuf opt, if present, will not be freed. 193 * 194 * Type of "mtu": rt_rmx.rmx_mtu is u_long, ifnet.ifr_mtu is int, and 195 * nd_ifinfo.linkmtu is u_int32_t. So we use u_long to hold largest one, 196 * which is rt_rmx.rmx_mtu. 197 */ 198 int 199 ip6_output( 200 struct mbuf *m0, 201 struct ip6_pktopts *opt, 202 struct route *ro, 203 int flags, 204 struct ip6_moptions *im6o, 205 struct inpcb *inp, 206 struct ifnet **ifpp /* XXX: just for statistics */ 207 ) 208 { 209 struct ip6_hdr *ip6, *mhip6; 210 struct ifnet *ifp = NULL, *origifp = NULL; 211 struct mbuf *m = m0; 212 int tlen, len, off; 213 bool tso; 214 struct route ip6route; 215 struct rtentry *rt = NULL, *rt_pmtu; 216 const struct sockaddr_in6 *dst; 217 struct sockaddr_in6 src_sa, dst_sa; 218 int error = 0; 219 struct in6_ifaddr *ia = NULL; 220 u_long mtu; 221 int alwaysfrag, dontfrag; 222 u_int32_t optlen = 0, plen = 0, unfragpartlen = 0; 223 struct ip6_exthdrs exthdrs; 224 struct in6_addr finaldst, src0, dst0; 225 u_int32_t zone; 226 struct route *ro_pmtu = NULL; 227 int hdrsplit = 0; 228 int needipsec = 0; 229 #ifdef IPSEC 230 struct secpolicy *sp = NULL; 231 #endif 232 struct psref psref, psref_ia; 233 int bound = curlwp_bind(); 234 bool release_psref_ia = false; 235 236 #ifdef DIAGNOSTIC 237 if ((m->m_flags & M_PKTHDR) == 0) 238 panic("ip6_output: no HDR"); 239 if ((m->m_pkthdr.csum_flags & 240 (M_CSUM_TCPv4|M_CSUM_UDPv4|M_CSUM_TSOv4)) != 0) { 241 panic("ip6_output: IPv4 checksum offload flags: %d", 242 m->m_pkthdr.csum_flags); 243 } 244 if ((m->m_pkthdr.csum_flags & (M_CSUM_TCPv6|M_CSUM_UDPv6)) == 245 (M_CSUM_TCPv6|M_CSUM_UDPv6)) { 246 panic("ip6_output: conflicting checksum offload flags: %d", 247 m->m_pkthdr.csum_flags); 248 } 249 #endif 250 251 M_CSUM_DATA_IPv6_SET(m->m_pkthdr.csum_data, sizeof(struct ip6_hdr)); 252 253 #define MAKE_EXTHDR(hp, mp) \ 254 do { \ 255 if (hp) { \ 256 struct ip6_ext *eh = (struct ip6_ext *)(hp); \ 257 error = ip6_copyexthdr((mp), (void *)(hp), \ 258 ((eh)->ip6e_len + 1) << 3); \ 259 if (error) \ 260 goto freehdrs; \ 261 } \ 262 } while (/*CONSTCOND*/ 0) 263 264 memset(&exthdrs, 0, sizeof(exthdrs)); 265 if (opt) { 266 /* Hop-by-Hop options header */ 267 MAKE_EXTHDR(opt->ip6po_hbh, &exthdrs.ip6e_hbh); 268 /* Destination options header (1st part) */ 269 MAKE_EXTHDR(opt->ip6po_dest1, &exthdrs.ip6e_dest1); 270 /* Routing header */ 271 MAKE_EXTHDR(opt->ip6po_rthdr, &exthdrs.ip6e_rthdr); 272 /* Destination options header (2nd part) */ 273 MAKE_EXTHDR(opt->ip6po_dest2, &exthdrs.ip6e_dest2); 274 } 275 276 /* 277 * Calculate the total length of the extension header chain. 278 * Keep the length of the unfragmentable part for fragmentation. 279 */ 280 optlen = 0; 281 if (exthdrs.ip6e_hbh) 282 optlen += exthdrs.ip6e_hbh->m_len; 283 if (exthdrs.ip6e_dest1) 284 optlen += exthdrs.ip6e_dest1->m_len; 285 if (exthdrs.ip6e_rthdr) 286 optlen += exthdrs.ip6e_rthdr->m_len; 287 unfragpartlen = optlen + sizeof(struct ip6_hdr); 288 /* NOTE: we don't add AH/ESP length here. do that later. */ 289 if (exthdrs.ip6e_dest2) 290 optlen += exthdrs.ip6e_dest2->m_len; 291 292 #ifdef IPSEC 293 if (ipsec_used) { 294 /* Check the security policy (SP) for the packet */ 295 sp = ipsec6_check_policy(m, inp, flags, &needipsec, &error); 296 if (error != 0) { 297 /* 298 * Hack: -EINVAL is used to signal that a packet 299 * should be silently discarded. This is typically 300 * because we asked key management for an SA and 301 * it was delayed (e.g. kicked up to IKE). 302 */ 303 if (error == -EINVAL) 304 error = 0; 305 IP6_STATINC(IP6_STAT_IPSECDROP_OUT); 306 goto freehdrs; 307 } 308 } 309 #endif 310 311 if (needipsec && 312 (m->m_pkthdr.csum_flags & (M_CSUM_UDPv6|M_CSUM_TCPv6)) != 0) { 313 in6_undefer_cksum_tcpudp(m); 314 m->m_pkthdr.csum_flags &= ~(M_CSUM_UDPv6|M_CSUM_TCPv6); 315 } 316 317 /* 318 * If we need IPsec, or there is at least one extension header, 319 * separate IP6 header from the payload. 320 */ 321 if ((needipsec || optlen) && !hdrsplit) { 322 if ((error = ip6_splithdr(m, &exthdrs)) != 0) { 323 IP6_STATINC(IP6_STAT_ODROPPED); 324 m = NULL; 325 goto freehdrs; 326 } 327 m = exthdrs.ip6e_ip6; 328 hdrsplit++; 329 } 330 331 /* adjust pointer */ 332 ip6 = mtod(m, struct ip6_hdr *); 333 334 /* adjust mbuf packet header length */ 335 m->m_pkthdr.len += optlen; 336 plen = m->m_pkthdr.len - sizeof(*ip6); 337 338 /* If this is a jumbo payload, insert a jumbo payload option. */ 339 if (plen > IPV6_MAXPACKET) { 340 if (!hdrsplit) { 341 if ((error = ip6_splithdr(m, &exthdrs)) != 0) { 342 IP6_STATINC(IP6_STAT_ODROPPED); 343 m = NULL; 344 goto freehdrs; 345 } 346 m = exthdrs.ip6e_ip6; 347 hdrsplit++; 348 } 349 /* adjust pointer */ 350 ip6 = mtod(m, struct ip6_hdr *); 351 if ((error = ip6_insert_jumboopt(&exthdrs, plen)) != 0) { 352 IP6_STATINC(IP6_STAT_ODROPPED); 353 goto freehdrs; 354 } 355 optlen += 8; /* XXX JUMBOOPTLEN */ 356 ip6->ip6_plen = 0; 357 } else 358 ip6->ip6_plen = htons(plen); 359 360 /* 361 * Concatenate headers and fill in next header fields. 362 * Here we have, on "m" 363 * IPv6 payload 364 * and we insert headers accordingly. Finally, we should be getting: 365 * IPv6 hbh dest1 rthdr ah* [esp* dest2 payload] 366 * 367 * during the header composing process, "m" points to IPv6 header. 368 * "mprev" points to an extension header prior to esp. 369 */ 370 { 371 u_char *nexthdrp = &ip6->ip6_nxt; 372 struct mbuf *mprev = m; 373 374 /* 375 * we treat dest2 specially. this makes IPsec processing 376 * much easier. the goal here is to make mprev point the 377 * mbuf prior to dest2. 378 * 379 * result: IPv6 dest2 payload 380 * m and mprev will point to IPv6 header. 381 */ 382 if (exthdrs.ip6e_dest2) { 383 if (!hdrsplit) 384 panic("assumption failed: hdr not split"); 385 exthdrs.ip6e_dest2->m_next = m->m_next; 386 m->m_next = exthdrs.ip6e_dest2; 387 *mtod(exthdrs.ip6e_dest2, u_char *) = ip6->ip6_nxt; 388 ip6->ip6_nxt = IPPROTO_DSTOPTS; 389 } 390 391 #define MAKE_CHAIN(m, mp, p, i)\ 392 do {\ 393 if (m) {\ 394 if (!hdrsplit) \ 395 panic("assumption failed: hdr not split"); \ 396 *mtod((m), u_char *) = *(p);\ 397 *(p) = (i);\ 398 p = mtod((m), u_char *);\ 399 (m)->m_next = (mp)->m_next;\ 400 (mp)->m_next = (m);\ 401 (mp) = (m);\ 402 }\ 403 } while (/*CONSTCOND*/ 0) 404 /* 405 * result: IPv6 hbh dest1 rthdr dest2 payload 406 * m will point to IPv6 header. mprev will point to the 407 * extension header prior to dest2 (rthdr in the above case). 408 */ 409 MAKE_CHAIN(exthdrs.ip6e_hbh, mprev, nexthdrp, IPPROTO_HOPOPTS); 410 MAKE_CHAIN(exthdrs.ip6e_dest1, mprev, nexthdrp, 411 IPPROTO_DSTOPTS); 412 MAKE_CHAIN(exthdrs.ip6e_rthdr, mprev, nexthdrp, 413 IPPROTO_ROUTING); 414 415 M_CSUM_DATA_IPv6_SET(m->m_pkthdr.csum_data, 416 sizeof(struct ip6_hdr) + optlen); 417 } 418 419 /* Need to save for pmtu */ 420 finaldst = ip6->ip6_dst; 421 422 /* 423 * If there is a routing header, replace destination address field 424 * with the first hop of the routing header. 425 */ 426 if (exthdrs.ip6e_rthdr) { 427 struct ip6_rthdr *rh; 428 429 rh = mtod(exthdrs.ip6e_rthdr, struct ip6_rthdr *); 430 431 error = ip6_handle_rthdr(rh, ip6); 432 if (error != 0) { 433 IP6_STATINC(IP6_STAT_ODROPPED); 434 goto bad; 435 } 436 } 437 438 /* Source address validation */ 439 if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) && 440 (flags & IPV6_UNSPECSRC) == 0) { 441 error = EOPNOTSUPP; 442 IP6_STATINC(IP6_STAT_BADSCOPE); 443 goto bad; 444 } 445 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) { 446 error = EOPNOTSUPP; 447 IP6_STATINC(IP6_STAT_BADSCOPE); 448 goto bad; 449 } 450 451 IP6_STATINC(IP6_STAT_LOCALOUT); 452 453 /* 454 * Route packet. 455 */ 456 /* initialize cached route */ 457 if (ro == NULL) { 458 memset(&ip6route, 0, sizeof(ip6route)); 459 ro = &ip6route; 460 } 461 ro_pmtu = ro; 462 if (opt && opt->ip6po_rthdr) 463 ro = &opt->ip6po_route; 464 465 /* 466 * if specified, try to fill in the traffic class field. 467 * do not override if a non-zero value is already set. 468 * we check the diffserv field and the ecn field separately. 469 */ 470 if (opt && opt->ip6po_tclass >= 0) { 471 int mask = 0; 472 473 if ((ip6->ip6_flow & htonl(0xfc << 20)) == 0) 474 mask |= 0xfc; 475 if ((ip6->ip6_flow & htonl(0x03 << 20)) == 0) 476 mask |= 0x03; 477 if (mask != 0) 478 ip6->ip6_flow |= htonl((opt->ip6po_tclass & mask) << 20); 479 } 480 481 /* fill in or override the hop limit field, if necessary. */ 482 if (opt && opt->ip6po_hlim != -1) 483 ip6->ip6_hlim = opt->ip6po_hlim & 0xff; 484 else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) { 485 if (im6o != NULL) 486 ip6->ip6_hlim = im6o->im6o_multicast_hlim; 487 else 488 ip6->ip6_hlim = ip6_defmcasthlim; 489 } 490 491 #ifdef IPSEC 492 if (needipsec) { 493 int s = splsoftnet(); 494 error = ipsec6_process_packet(m, sp->req, flags); 495 splx(s); 496 497 /* 498 * Preserve KAME behaviour: ENOENT can be returned 499 * when an SA acquire is in progress. Don't propagate 500 * this to user-level; it confuses applications. 501 * XXX this will go away when the SADB is redone. 502 */ 503 if (error == ENOENT) 504 error = 0; 505 506 goto done; 507 } 508 #endif 509 510 /* adjust pointer */ 511 ip6 = mtod(m, struct ip6_hdr *); 512 513 sockaddr_in6_init(&dst_sa, &ip6->ip6_dst, 0, 0, 0); 514 515 /* We do not need a route for multicast */ 516 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) { 517 struct in6_pktinfo *pi = NULL; 518 519 /* 520 * If the outgoing interface for the address is specified by 521 * the caller, use it. 522 */ 523 if (opt && (pi = opt->ip6po_pktinfo) != NULL) { 524 /* XXX boundary check is assumed to be already done. */ 525 ifp = if_get_byindex(pi->ipi6_ifindex, &psref); 526 } else if (im6o != NULL) { 527 ifp = if_get_byindex(im6o->im6o_multicast_if_index, 528 &psref); 529 } 530 } 531 532 if (ifp == NULL) { 533 error = in6_selectroute(&dst_sa, opt, &ro, &rt, true); 534 if (error != 0) 535 goto bad; 536 ifp = if_get_byindex(rt->rt_ifp->if_index, &psref); 537 } 538 539 if (rt == NULL) { 540 /* 541 * If in6_selectroute() does not return a route entry, 542 * dst may not have been updated. 543 */ 544 error = rtcache_setdst(ro, sin6tosa(&dst_sa)); 545 if (error) { 546 IP6_STATINC(IP6_STAT_ODROPPED); 547 goto bad; 548 } 549 } 550 551 /* 552 * then rt (for unicast) and ifp must be non-NULL valid values. 553 */ 554 if ((flags & IPV6_FORWARDING) == 0) { 555 /* XXX: the FORWARDING flag can be set for mrouting. */ 556 in6_ifstat_inc(ifp, ifs6_out_request); 557 } 558 if (rt != NULL) { 559 ia = (struct in6_ifaddr *)(rt->rt_ifa); 560 rt->rt_use++; 561 } 562 563 /* 564 * The outgoing interface must be in the zone of source and 565 * destination addresses. We should use ia_ifp to support the 566 * case of sending packets to an address of our own. 567 */ 568 if (ia != NULL) { 569 origifp = ia->ia_ifp; 570 if (if_is_deactivated(origifp)) { 571 IP6_STATINC(IP6_STAT_ODROPPED); 572 goto bad; 573 } 574 if_acquire(origifp, &psref_ia); 575 release_psref_ia = true; 576 } else 577 origifp = ifp; 578 579 src0 = ip6->ip6_src; 580 if (in6_setscope(&src0, origifp, &zone)) 581 goto badscope; 582 sockaddr_in6_init(&src_sa, &ip6->ip6_src, 0, 0, 0); 583 if (sa6_recoverscope(&src_sa) || zone != src_sa.sin6_scope_id) 584 goto badscope; 585 586 dst0 = ip6->ip6_dst; 587 if (in6_setscope(&dst0, origifp, &zone)) 588 goto badscope; 589 /* re-initialize to be sure */ 590 sockaddr_in6_init(&dst_sa, &ip6->ip6_dst, 0, 0, 0); 591 if (sa6_recoverscope(&dst_sa) || zone != dst_sa.sin6_scope_id) 592 goto badscope; 593 594 /* scope check is done. */ 595 596 /* Ensure we only send from a valid address. */ 597 if ((ifp->if_flags & IFF_LOOPBACK) == 0 && 598 (flags & IPV6_FORWARDING) == 0 && 599 (error = ip6_ifaddrvalid(&src0, &dst0)) != 0) 600 { 601 char ip6buf[INET6_ADDRSTRLEN]; 602 nd6log(LOG_ERR, 603 "refusing to send from invalid address %s (pid %d)\n", 604 IN6_PRINT(ip6buf, &src0), curproc->p_pid); 605 IP6_STATINC(IP6_STAT_ODROPPED); 606 in6_ifstat_inc(origifp, ifs6_out_discard); 607 if (error == 1) 608 /* 609 * Address exists, but is tentative or detached. 610 * We can't send from it because it's invalid, 611 * so we drop the packet. 612 */ 613 error = 0; 614 else 615 error = EADDRNOTAVAIL; 616 goto bad; 617 } 618 619 if (rt != NULL && (rt->rt_flags & RTF_GATEWAY) && 620 !IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) 621 dst = satocsin6(rt->rt_gateway); 622 else 623 dst = satocsin6(rtcache_getdst(ro)); 624 625 /* 626 * XXXXXX: original code follows: 627 */ 628 if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) 629 m->m_flags &= ~(M_BCAST | M_MCAST); /* just in case */ 630 else { 631 bool ingroup; 632 633 m->m_flags = (m->m_flags & ~M_BCAST) | M_MCAST; 634 635 in6_ifstat_inc(ifp, ifs6_out_mcast); 636 637 /* 638 * Confirm that the outgoing interface supports multicast. 639 */ 640 if (!(ifp->if_flags & IFF_MULTICAST)) { 641 IP6_STATINC(IP6_STAT_NOROUTE); 642 in6_ifstat_inc(ifp, ifs6_out_discard); 643 error = ENETUNREACH; 644 goto bad; 645 } 646 647 ingroup = in6_multi_group(&ip6->ip6_dst, ifp); 648 if (ingroup && (im6o == NULL || im6o->im6o_multicast_loop)) { 649 /* 650 * If we belong to the destination multicast group 651 * on the outgoing interface, and the caller did not 652 * forbid loopback, loop back a copy. 653 */ 654 KASSERT(dst != NULL); 655 ip6_mloopback(ifp, m, dst); 656 } else { 657 /* 658 * If we are acting as a multicast router, perform 659 * multicast forwarding as if the packet had just 660 * arrived on the interface to which we are about 661 * to send. The multicast forwarding function 662 * recursively calls this function, using the 663 * IPV6_FORWARDING flag to prevent infinite recursion. 664 * 665 * Multicasts that are looped back by ip6_mloopback(), 666 * above, will be forwarded by the ip6_input() routine, 667 * if necessary. 668 */ 669 if (ip6_mrouter && (flags & IPV6_FORWARDING) == 0) { 670 if (ip6_mforward(ip6, ifp, m) != 0) { 671 m_freem(m); 672 goto done; 673 } 674 } 675 } 676 /* 677 * Multicasts with a hoplimit of zero may be looped back, 678 * above, but must not be transmitted on a network. 679 * Also, multicasts addressed to the loopback interface 680 * are not sent -- the above call to ip6_mloopback() will 681 * loop back a copy if this host actually belongs to the 682 * destination group on the loopback interface. 683 */ 684 if (ip6->ip6_hlim == 0 || (ifp->if_flags & IFF_LOOPBACK) || 685 IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst)) { 686 m_freem(m); 687 goto done; 688 } 689 } 690 691 /* 692 * Fill the outgoing interface to tell the upper layer 693 * to increment per-interface statistics. 694 */ 695 if (ifpp) 696 *ifpp = ifp; 697 698 /* Determine path MTU. */ 699 /* 700 * ro_pmtu represent final destination while 701 * ro might represent immediate destination. 702 * Use ro_pmtu destination since MTU might differ. 703 */ 704 if (ro_pmtu != ro) { 705 union { 706 struct sockaddr dst; 707 struct sockaddr_in6 dst6; 708 } u; 709 710 /* ro_pmtu may not have a cache */ 711 sockaddr_in6_init(&u.dst6, &finaldst, 0, 0, 0); 712 rt_pmtu = rtcache_lookup(ro_pmtu, &u.dst); 713 } else 714 rt_pmtu = rt; 715 error = ip6_getpmtu(rt_pmtu, ifp, &mtu, &alwaysfrag); 716 if (rt_pmtu != NULL && rt_pmtu != rt) 717 rtcache_unref(rt_pmtu, ro_pmtu); 718 KASSERT(error == 0); /* ip6_getpmtu never fail if ifp is passed */ 719 720 /* 721 * The caller of this function may specify to use the minimum MTU 722 * in some cases. 723 * An advanced API option (IPV6_USE_MIN_MTU) can also override MTU 724 * setting. The logic is a bit complicated; by default, unicast 725 * packets will follow path MTU while multicast packets will be sent at 726 * the minimum MTU. If IP6PO_MINMTU_ALL is specified, all packets 727 * including unicast ones will be sent at the minimum MTU. Multicast 728 * packets will always be sent at the minimum MTU unless 729 * IP6PO_MINMTU_DISABLE is explicitly specified. 730 * See RFC 3542 for more details. 731 */ 732 if (mtu > IPV6_MMTU) { 733 if ((flags & IPV6_MINMTU)) 734 mtu = IPV6_MMTU; 735 else if (opt && opt->ip6po_minmtu == IP6PO_MINMTU_ALL) 736 mtu = IPV6_MMTU; 737 else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) && 738 (opt == NULL || 739 opt->ip6po_minmtu != IP6PO_MINMTU_DISABLE)) { 740 mtu = IPV6_MMTU; 741 } 742 } 743 744 /* 745 * clear embedded scope identifiers if necessary. 746 * in6_clearscope will touch the addresses only when necessary. 747 */ 748 in6_clearscope(&ip6->ip6_src); 749 in6_clearscope(&ip6->ip6_dst); 750 751 /* 752 * If the outgoing packet contains a hop-by-hop options header, 753 * it must be examined and processed even by the source node. 754 * (RFC 2460, section 4.) 755 * 756 * XXX Is this really necessary? 757 */ 758 if (ip6->ip6_nxt == IPPROTO_HOPOPTS) { 759 u_int32_t dummy1; /* XXX unused */ 760 u_int32_t dummy2; /* XXX unused */ 761 int hoff = sizeof(struct ip6_hdr); 762 763 if (ip6_hopopts_input(&dummy1, &dummy2, &m, &hoff)) { 764 /* m was already freed at this point */ 765 error = EINVAL; 766 goto done; 767 } 768 769 ip6 = mtod(m, struct ip6_hdr *); 770 } 771 772 /* 773 * Run through list of hooks for output packets. 774 */ 775 error = pfil_run_hooks(inet6_pfil_hook, &m, ifp, PFIL_OUT); 776 if (error != 0 || m == NULL) { 777 IP6_STATINC(IP6_STAT_PFILDROP_OUT); 778 goto done; 779 } 780 ip6 = mtod(m, struct ip6_hdr *); 781 782 /* 783 * Send the packet to the outgoing interface. 784 * If necessary, do IPv6 fragmentation before sending. 785 * 786 * the logic here is rather complex: 787 * 1: normal case (dontfrag == 0, alwaysfrag == 0) 788 * 1-a: send as is if tlen <= path mtu 789 * 1-b: fragment if tlen > path mtu 790 * 791 * 2: if user asks us not to fragment (dontfrag == 1) 792 * 2-a: send as is if tlen <= interface mtu 793 * 2-b: error if tlen > interface mtu 794 * 795 * 3: if we always need to attach fragment header (alwaysfrag == 1) 796 * always fragment 797 * 798 * 4: if dontfrag == 1 && alwaysfrag == 1 799 * error, as we cannot handle this conflicting request 800 */ 801 tlen = m->m_pkthdr.len; 802 tso = (m->m_pkthdr.csum_flags & M_CSUM_TSOv6) != 0; 803 if (opt && (opt->ip6po_flags & IP6PO_DONTFRAG)) 804 dontfrag = 1; 805 else 806 dontfrag = 0; 807 808 if (dontfrag && alwaysfrag) { /* case 4 */ 809 /* conflicting request - can't transmit */ 810 IP6_STATINC(IP6_STAT_CANTFRAG); 811 error = EMSGSIZE; 812 goto bad; 813 } 814 if (dontfrag && (!tso && tlen > ifp->if_mtu)) { /* case 2-b */ 815 /* 816 * Even if the DONTFRAG option is specified, we cannot send the 817 * packet when the data length is larger than the MTU of the 818 * outgoing interface. 819 * Notify the error by sending IPV6_PATHMTU ancillary data as 820 * well as returning an error code (the latter is not described 821 * in the API spec.) 822 */ 823 u_int32_t mtu32; 824 struct ip6ctlparam ip6cp; 825 826 mtu32 = (u_int32_t)mtu; 827 memset(&ip6cp, 0, sizeof(ip6cp)); 828 ip6cp.ip6c_cmdarg = (void *)&mtu32; 829 pfctlinput2(PRC_MSGSIZE, 830 rtcache_getdst(ro_pmtu), &ip6cp); 831 832 IP6_STATINC(IP6_STAT_CANTFRAG); 833 error = EMSGSIZE; 834 goto bad; 835 } 836 837 /* 838 * transmit packet without fragmentation 839 */ 840 if (dontfrag || (!alwaysfrag && (tlen <= mtu || tso))) { 841 /* case 1-a and 2-a */ 842 struct in6_ifaddr *ia6; 843 int sw_csum; 844 int s; 845 846 ip6 = mtod(m, struct ip6_hdr *); 847 s = pserialize_read_enter(); 848 ia6 = in6_ifawithifp(ifp, &ip6->ip6_src); 849 if (ia6) { 850 /* Record statistics for this interface address. */ 851 ia6->ia_ifa.ifa_data.ifad_outbytes += m->m_pkthdr.len; 852 } 853 pserialize_read_exit(s); 854 855 sw_csum = m->m_pkthdr.csum_flags & ~ifp->if_csum_flags_tx; 856 if ((sw_csum & (M_CSUM_UDPv6|M_CSUM_TCPv6)) != 0) { 857 if (IN6_NEED_CHECKSUM(ifp, 858 sw_csum & (M_CSUM_UDPv6|M_CSUM_TCPv6))) { 859 in6_undefer_cksum_tcpudp(m); 860 } 861 m->m_pkthdr.csum_flags &= ~(M_CSUM_UDPv6|M_CSUM_TCPv6); 862 } 863 864 KASSERT(dst != NULL); 865 if (__predict_false(sw_csum & M_CSUM_TSOv6)) { 866 /* 867 * TSO6 is required by a packet, but disabled for 868 * the interface. 869 */ 870 error = ip6_tso_output(ifp, origifp, m, dst, rt); 871 } else 872 error = ip6_if_output(ifp, origifp, m, dst, rt); 873 goto done; 874 } 875 876 if (tso) { 877 IP6_STATINC(IP6_STAT_CANTFRAG); /* XXX */ 878 error = EINVAL; /* XXX */ 879 goto bad; 880 } 881 882 /* 883 * try to fragment the packet. case 1-b and 3 884 */ 885 if (mtu < IPV6_MMTU) { 886 /* path MTU cannot be less than IPV6_MMTU */ 887 IP6_STATINC(IP6_STAT_CANTFRAG); 888 error = EMSGSIZE; 889 in6_ifstat_inc(ifp, ifs6_out_fragfail); 890 goto bad; 891 } else if (ip6->ip6_plen == 0) { 892 /* jumbo payload cannot be fragmented */ 893 IP6_STATINC(IP6_STAT_CANTFRAG); 894 error = EMSGSIZE; 895 in6_ifstat_inc(ifp, ifs6_out_fragfail); 896 goto bad; 897 } else { 898 const uint32_t id = ip6_randomid(); 899 struct mbuf **mnext, *m_frgpart; 900 const int hlen = unfragpartlen; 901 struct ip6_frag *ip6f; 902 u_char nextproto; 903 904 if (mtu > IPV6_MAXPACKET) 905 mtu = IPV6_MAXPACKET; 906 907 /* 908 * Must be able to put at least 8 bytes per fragment. 909 */ 910 len = (mtu - hlen - sizeof(struct ip6_frag)) & ~7; 911 if (len < 8) { 912 IP6_STATINC(IP6_STAT_CANTFRAG); 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_undefer_cksum_tcpudp(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 * Insert jumbo payload option. 1102 */ 1103 static int 1104 ip6_insert_jumboopt(struct ip6_exthdrs *exthdrs, u_int32_t plen) 1105 { 1106 struct mbuf *mopt; 1107 u_int8_t *optbuf; 1108 u_int32_t v; 1109 1110 #define JUMBOOPTLEN 8 /* length of jumbo payload option and padding */ 1111 1112 /* 1113 * If there is no hop-by-hop options header, allocate new one. 1114 * If there is one but it doesn't have enough space to store the 1115 * jumbo payload option, allocate a cluster to store the whole options. 1116 * Otherwise, use it to store the options. 1117 */ 1118 if (exthdrs->ip6e_hbh == NULL) { 1119 MGET(mopt, M_DONTWAIT, MT_DATA); 1120 if (mopt == 0) 1121 return (ENOBUFS); 1122 mopt->m_len = JUMBOOPTLEN; 1123 optbuf = mtod(mopt, u_int8_t *); 1124 optbuf[1] = 0; /* = ((JUMBOOPTLEN) >> 3) - 1 */ 1125 exthdrs->ip6e_hbh = mopt; 1126 } else { 1127 struct ip6_hbh *hbh; 1128 1129 mopt = exthdrs->ip6e_hbh; 1130 if (M_TRAILINGSPACE(mopt) < JUMBOOPTLEN) { 1131 const int oldoptlen = mopt->m_len; 1132 struct mbuf *n; 1133 1134 /* 1135 * Assumptions: 1136 * - exthdrs->ip6e_hbh is not referenced from places 1137 * other than exthdrs. 1138 * - exthdrs->ip6e_hbh is not an mbuf chain. 1139 */ 1140 KASSERT(mopt->m_next == NULL); 1141 1142 /* 1143 * Give up if the whole (new) hbh header does not fit 1144 * even in an mbuf cluster. 1145 */ 1146 if (oldoptlen + JUMBOOPTLEN > MCLBYTES) 1147 return ENOBUFS; 1148 1149 /* 1150 * At this point, we must always prepare a cluster. 1151 */ 1152 MGET(n, M_DONTWAIT, MT_DATA); 1153 if (n) { 1154 MCLGET(n, M_DONTWAIT); 1155 if ((n->m_flags & M_EXT) == 0) { 1156 m_freem(n); 1157 n = NULL; 1158 } 1159 } 1160 if (!n) 1161 return ENOBUFS; 1162 1163 n->m_len = oldoptlen + JUMBOOPTLEN; 1164 bcopy(mtod(mopt, void *), mtod(n, void *), 1165 oldoptlen); 1166 optbuf = mtod(n, u_int8_t *) + oldoptlen; 1167 m_freem(mopt); 1168 mopt = exthdrs->ip6e_hbh = n; 1169 } else { 1170 optbuf = mtod(mopt, u_int8_t *) + mopt->m_len; 1171 mopt->m_len += JUMBOOPTLEN; 1172 } 1173 optbuf[0] = IP6OPT_PADN; 1174 optbuf[1] = 0; 1175 1176 /* 1177 * Adjust the header length according to the pad and 1178 * the jumbo payload option. 1179 */ 1180 hbh = mtod(mopt, struct ip6_hbh *); 1181 hbh->ip6h_len += (JUMBOOPTLEN >> 3); 1182 } 1183 1184 /* fill in the option. */ 1185 optbuf[2] = IP6OPT_JUMBO; 1186 optbuf[3] = 4; 1187 v = (u_int32_t)htonl(plen + JUMBOOPTLEN); 1188 memcpy(&optbuf[4], &v, sizeof(u_int32_t)); 1189 1190 /* finally, adjust the packet header length */ 1191 exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN; 1192 1193 return 0; 1194 #undef JUMBOOPTLEN 1195 } 1196 1197 /* 1198 * Insert fragment header and copy unfragmentable header portions. 1199 * 1200 * *frghdrp will not be read, and it is guaranteed that either an 1201 * error is returned or that *frghdrp will point to space allocated 1202 * for the fragment header. 1203 * 1204 * On entry, m contains: 1205 * IPv6 Header 1206 * On exit, it contains: 1207 * IPv6 Header -> Unfragmentable Part -> Frag6 Header 1208 */ 1209 static int 1210 ip6_insertfraghdr(struct mbuf *m0, struct mbuf *m, int hlen, 1211 struct ip6_frag **frghdrp) 1212 { 1213 struct mbuf *n, *mlast; 1214 1215 if (hlen > sizeof(struct ip6_hdr)) { 1216 n = m_copym(m0, sizeof(struct ip6_hdr), 1217 hlen - sizeof(struct ip6_hdr), M_DONTWAIT); 1218 if (n == NULL) 1219 return ENOBUFS; 1220 m->m_next = n; 1221 } else 1222 n = m; 1223 1224 /* Search for the last mbuf of unfragmentable part. */ 1225 for (mlast = n; mlast->m_next; mlast = mlast->m_next) 1226 ; 1227 1228 if ((mlast->m_flags & M_EXT) == 0 && 1229 M_TRAILINGSPACE(mlast) >= sizeof(struct ip6_frag)) { 1230 /* use the trailing space of the last mbuf for the fragment hdr */ 1231 *frghdrp = (struct ip6_frag *)(mtod(mlast, char *) + 1232 mlast->m_len); 1233 mlast->m_len += sizeof(struct ip6_frag); 1234 } else { 1235 /* allocate a new mbuf for the fragment header */ 1236 struct mbuf *mfrg; 1237 1238 MGET(mfrg, M_DONTWAIT, MT_DATA); 1239 if (mfrg == NULL) 1240 return ENOBUFS; 1241 mfrg->m_len = sizeof(struct ip6_frag); 1242 *frghdrp = mtod(mfrg, struct ip6_frag *); 1243 mlast->m_next = mfrg; 1244 } 1245 1246 return 0; 1247 } 1248 1249 static int 1250 ip6_getpmtu(struct rtentry *rt, struct ifnet *ifp, u_long *mtup, 1251 int *alwaysfragp) 1252 { 1253 u_int32_t mtu = 0; 1254 int alwaysfrag = 0; 1255 int error = 0; 1256 1257 if (rt != NULL) { 1258 if (ifp == NULL) 1259 ifp = rt->rt_ifp; 1260 mtu = rt->rt_rmx.rmx_mtu; 1261 if (mtu == 0) 1262 mtu = ifp->if_mtu; 1263 else if (mtu < IPV6_MMTU) { 1264 /* 1265 * RFC2460 section 5, last paragraph: 1266 * if we record ICMPv6 too big message with 1267 * mtu < IPV6_MMTU, transmit packets sized IPV6_MMTU 1268 * or smaller, with fragment header attached. 1269 * (fragment header is needed regardless from the 1270 * packet size, for translators to identify packets) 1271 */ 1272 alwaysfrag = 1; 1273 mtu = IPV6_MMTU; 1274 } else if (mtu > ifp->if_mtu) { 1275 /* 1276 * The MTU on the route is larger than the MTU on 1277 * the interface! This shouldn't happen, unless the 1278 * MTU of the interface has been changed after the 1279 * interface was brought up. Change the MTU in the 1280 * route to match the interface MTU (as long as the 1281 * field isn't locked). 1282 */ 1283 mtu = ifp->if_mtu; 1284 if (!(rt->rt_rmx.rmx_locks & RTV_MTU)) 1285 rt->rt_rmx.rmx_mtu = mtu; 1286 } 1287 } else if (ifp) { 1288 mtu = ifp->if_mtu; 1289 } else 1290 error = EHOSTUNREACH; /* XXX */ 1291 1292 *mtup = mtu; 1293 if (alwaysfragp) 1294 *alwaysfragp = alwaysfrag; 1295 return (error); 1296 } 1297 1298 /* 1299 * IP6 socket option processing. 1300 */ 1301 int 1302 ip6_ctloutput(int op, struct socket *so, struct sockopt *sopt) 1303 { 1304 int optdatalen, uproto; 1305 void *optdata; 1306 struct inpcb *inp = sotoinpcb(so); 1307 struct ip_moptions **mopts; 1308 int error, optval; 1309 int level, optname; 1310 1311 KASSERT(solocked(so)); 1312 KASSERT(sopt != NULL); 1313 1314 level = sopt->sopt_level; 1315 optname = sopt->sopt_name; 1316 1317 error = optval = 0; 1318 uproto = (int)so->so_proto->pr_protocol; 1319 1320 switch (level) { 1321 case IPPROTO_IP: 1322 switch (optname) { 1323 case IP_ADD_MEMBERSHIP: 1324 case IP_DROP_MEMBERSHIP: 1325 case IP_MULTICAST_IF: 1326 case IP_MULTICAST_LOOP: 1327 case IP_MULTICAST_TTL: 1328 mopts = &inp->inp_moptions; 1329 switch (op) { 1330 case PRCO_GETOPT: 1331 return ip_getmoptions(*mopts, sopt); 1332 case PRCO_SETOPT: 1333 return ip_setmoptions(mopts, sopt); 1334 default: 1335 return EINVAL; 1336 } 1337 default: 1338 return ENOPROTOOPT; 1339 } 1340 case IPPROTO_IPV6: 1341 break; 1342 default: 1343 return ENOPROTOOPT; 1344 } 1345 switch (op) { 1346 case PRCO_SETOPT: 1347 switch (optname) { 1348 #ifdef RFC2292 1349 case IPV6_2292PKTOPTIONS: 1350 error = ip6_pcbopts(&in6p_outputopts(inp), so, sopt); 1351 break; 1352 #endif 1353 1354 /* 1355 * Use of some Hop-by-Hop options or some 1356 * Destination options, might require special 1357 * privilege. That is, normal applications 1358 * (without special privilege) might be forbidden 1359 * from setting certain options in outgoing packets, 1360 * and might never see certain options in received 1361 * packets. [RFC 2292 Section 6] 1362 * KAME specific note: 1363 * KAME prevents non-privileged users from sending or 1364 * receiving ANY hbh/dst options in order to avoid 1365 * overhead of parsing options in the kernel. 1366 */ 1367 case IPV6_RECVHOPOPTS: 1368 case IPV6_RECVDSTOPTS: 1369 case IPV6_RECVRTHDRDSTOPTS: 1370 error = kauth_authorize_network( 1371 kauth_cred_get(), 1372 KAUTH_NETWORK_IPV6, KAUTH_REQ_NETWORK_IPV6_HOPBYHOP, 1373 NULL, NULL, NULL); 1374 if (error) 1375 break; 1376 /* FALLTHROUGH */ 1377 case IPV6_UNICAST_HOPS: 1378 case IPV6_HOPLIMIT: 1379 case IPV6_FAITH: 1380 1381 case IPV6_RECVPKTINFO: 1382 case IPV6_RECVHOPLIMIT: 1383 case IPV6_RECVRTHDR: 1384 case IPV6_RECVPATHMTU: 1385 case IPV6_RECVTCLASS: 1386 case IPV6_V6ONLY: 1387 case IPV6_BINDANY: 1388 error = sockopt_getint(sopt, &optval); 1389 if (error) 1390 break; 1391 switch (optname) { 1392 case IPV6_UNICAST_HOPS: 1393 if (optval < -1 || optval >= 256) 1394 error = EINVAL; 1395 else { 1396 /* -1 = kernel default */ 1397 in6p_hops6(inp) = optval; 1398 } 1399 break; 1400 #define OPTSET(bit) \ 1401 do { \ 1402 if (optval) \ 1403 inp->inp_flags |= (bit); \ 1404 else \ 1405 inp->inp_flags &= ~(bit); \ 1406 } while (/*CONSTCOND*/ 0) 1407 1408 #ifdef RFC2292 1409 #define OPTSET2292(bit) \ 1410 do { \ 1411 inp->inp_flags |= IN6P_RFC2292; \ 1412 if (optval) \ 1413 inp->inp_flags |= (bit); \ 1414 else \ 1415 inp->inp_flags &= ~(bit); \ 1416 } while (/*CONSTCOND*/ 0) 1417 #endif 1418 1419 #define OPTBIT(bit) (inp->inp_flags & (bit) ? 1 : 0) 1420 1421 case IPV6_RECVPKTINFO: 1422 #ifdef RFC2292 1423 /* cannot mix with RFC2292 */ 1424 if (OPTBIT(IN6P_RFC2292)) { 1425 error = EINVAL; 1426 break; 1427 } 1428 #endif 1429 OPTSET(IN6P_PKTINFO); 1430 break; 1431 1432 case IPV6_HOPLIMIT: 1433 { 1434 struct ip6_pktopts **optp; 1435 1436 #ifdef RFC2292 1437 /* cannot mix with RFC2292 */ 1438 if (OPTBIT(IN6P_RFC2292)) { 1439 error = EINVAL; 1440 break; 1441 } 1442 #endif 1443 optp = &in6p_outputopts(inp); 1444 error = ip6_pcbopt(IPV6_HOPLIMIT, 1445 (u_char *)&optval, 1446 sizeof(optval), 1447 optp, 1448 kauth_cred_get(), uproto); 1449 break; 1450 } 1451 1452 case IPV6_RECVHOPLIMIT: 1453 #ifdef RFC2292 1454 /* cannot mix with RFC2292 */ 1455 if (OPTBIT(IN6P_RFC2292)) { 1456 error = EINVAL; 1457 break; 1458 } 1459 #endif 1460 OPTSET(IN6P_HOPLIMIT); 1461 break; 1462 1463 case IPV6_RECVHOPOPTS: 1464 #ifdef RFC2292 1465 /* cannot mix with RFC2292 */ 1466 if (OPTBIT(IN6P_RFC2292)) { 1467 error = EINVAL; 1468 break; 1469 } 1470 #endif 1471 OPTSET(IN6P_HOPOPTS); 1472 break; 1473 1474 case IPV6_RECVDSTOPTS: 1475 #ifdef RFC2292 1476 /* cannot mix with RFC2292 */ 1477 if (OPTBIT(IN6P_RFC2292)) { 1478 error = EINVAL; 1479 break; 1480 } 1481 #endif 1482 OPTSET(IN6P_DSTOPTS); 1483 break; 1484 1485 case IPV6_RECVRTHDRDSTOPTS: 1486 #ifdef RFC2292 1487 /* cannot mix with RFC2292 */ 1488 if (OPTBIT(IN6P_RFC2292)) { 1489 error = EINVAL; 1490 break; 1491 } 1492 #endif 1493 OPTSET(IN6P_RTHDRDSTOPTS); 1494 break; 1495 1496 case IPV6_RECVRTHDR: 1497 #ifdef RFC2292 1498 /* cannot mix with RFC2292 */ 1499 if (OPTBIT(IN6P_RFC2292)) { 1500 error = EINVAL; 1501 break; 1502 } 1503 #endif 1504 OPTSET(IN6P_RTHDR); 1505 break; 1506 1507 case IPV6_FAITH: 1508 OPTSET(IN6P_FAITH); 1509 break; 1510 1511 case IPV6_RECVPATHMTU: 1512 /* 1513 * We ignore this option for TCP 1514 * sockets. 1515 * (RFC3542 leaves this case 1516 * unspecified.) 1517 */ 1518 if (uproto != IPPROTO_TCP) 1519 OPTSET(IN6P_MTU); 1520 break; 1521 1522 case IPV6_V6ONLY: 1523 /* 1524 * make setsockopt(IPV6_V6ONLY) 1525 * available only prior to bind(2). 1526 * see ipng mailing list, Jun 22 2001. 1527 */ 1528 if (inp->inp_lport || 1529 !IN6_IS_ADDR_UNSPECIFIED(&in6p_laddr(inp))) { 1530 error = EINVAL; 1531 break; 1532 } 1533 #ifdef INET6_BINDV6ONLY 1534 if (!optval) 1535 error = EINVAL; 1536 #else 1537 OPTSET(IN6P_IPV6_V6ONLY); 1538 #endif 1539 break; 1540 1541 case IPV6_RECVTCLASS: 1542 #ifdef RFC2292 1543 /* cannot mix with RFC2292 XXX */ 1544 if (OPTBIT(IN6P_RFC2292)) { 1545 error = EINVAL; 1546 break; 1547 } 1548 #endif 1549 OPTSET(IN6P_TCLASS); 1550 break; 1551 1552 case IPV6_BINDANY: 1553 error = kauth_authorize_network( 1554 kauth_cred_get(), KAUTH_NETWORK_BIND, 1555 KAUTH_REQ_NETWORK_BIND_ANYADDR, so, NULL, 1556 NULL); 1557 if (error) 1558 break; 1559 OPTSET(IN6P_BINDANY); 1560 break; 1561 } 1562 break; 1563 1564 case IPV6_OTCLASS: 1565 { 1566 struct ip6_pktopts **optp; 1567 u_int8_t tclass; 1568 1569 error = sockopt_get(sopt, &tclass, sizeof(tclass)); 1570 if (error) 1571 break; 1572 optp = &in6p_outputopts(inp); 1573 error = ip6_pcbopt(optname, 1574 (u_char *)&tclass, 1575 sizeof(tclass), 1576 optp, 1577 kauth_cred_get(), uproto); 1578 break; 1579 } 1580 1581 case IPV6_TCLASS: 1582 case IPV6_DONTFRAG: 1583 case IPV6_USE_MIN_MTU: 1584 case IPV6_PREFER_TEMPADDR: 1585 error = sockopt_getint(sopt, &optval); 1586 if (error) 1587 break; 1588 { 1589 struct ip6_pktopts **optp; 1590 optp = &in6p_outputopts(inp); 1591 error = ip6_pcbopt(optname, 1592 (u_char *)&optval, 1593 sizeof(optval), 1594 optp, 1595 kauth_cred_get(), uproto); 1596 break; 1597 } 1598 1599 #ifdef RFC2292 1600 case IPV6_2292PKTINFO: 1601 case IPV6_2292HOPLIMIT: 1602 case IPV6_2292HOPOPTS: 1603 case IPV6_2292DSTOPTS: 1604 case IPV6_2292RTHDR: 1605 /* RFC 2292 */ 1606 error = sockopt_getint(sopt, &optval); 1607 if (error) 1608 break; 1609 1610 switch (optname) { 1611 case IPV6_2292PKTINFO: 1612 OPTSET2292(IN6P_PKTINFO); 1613 break; 1614 case IPV6_2292HOPLIMIT: 1615 OPTSET2292(IN6P_HOPLIMIT); 1616 break; 1617 case IPV6_2292HOPOPTS: 1618 /* 1619 * Check super-user privilege. 1620 * See comments for IPV6_RECVHOPOPTS. 1621 */ 1622 error = kauth_authorize_network( 1623 kauth_cred_get(), 1624 KAUTH_NETWORK_IPV6, 1625 KAUTH_REQ_NETWORK_IPV6_HOPBYHOP, NULL, 1626 NULL, NULL); 1627 if (error) 1628 return (error); 1629 OPTSET2292(IN6P_HOPOPTS); 1630 break; 1631 case IPV6_2292DSTOPTS: 1632 error = kauth_authorize_network( 1633 kauth_cred_get(), 1634 KAUTH_NETWORK_IPV6, 1635 KAUTH_REQ_NETWORK_IPV6_HOPBYHOP, NULL, 1636 NULL, NULL); 1637 if (error) 1638 return (error); 1639 OPTSET2292(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); /* XXX */ 1640 break; 1641 case IPV6_2292RTHDR: 1642 OPTSET2292(IN6P_RTHDR); 1643 break; 1644 } 1645 break; 1646 #endif 1647 case IPV6_PKTINFO: 1648 case IPV6_HOPOPTS: 1649 case IPV6_RTHDR: 1650 case IPV6_DSTOPTS: 1651 case IPV6_RTHDRDSTOPTS: 1652 case IPV6_NEXTHOP: { 1653 /* new advanced API (RFC3542) */ 1654 void *optbuf; 1655 int optbuflen; 1656 struct ip6_pktopts **optp; 1657 1658 #ifdef RFC2292 1659 /* cannot mix with RFC2292 */ 1660 if (OPTBIT(IN6P_RFC2292)) { 1661 error = EINVAL; 1662 break; 1663 } 1664 #endif 1665 1666 optbuflen = sopt->sopt_size; 1667 optbuf = malloc(optbuflen, M_IP6OPT, M_NOWAIT); 1668 if (optbuf == NULL) { 1669 error = ENOBUFS; 1670 break; 1671 } 1672 1673 error = sockopt_get(sopt, optbuf, optbuflen); 1674 if (error) { 1675 free(optbuf, M_IP6OPT); 1676 break; 1677 } 1678 optp = &in6p_outputopts(inp); 1679 error = ip6_pcbopt(optname, optbuf, optbuflen, 1680 optp, kauth_cred_get(), uproto); 1681 1682 free(optbuf, M_IP6OPT); 1683 break; 1684 } 1685 #undef OPTSET 1686 1687 case IPV6_MULTICAST_IF: 1688 case IPV6_MULTICAST_HOPS: 1689 case IPV6_MULTICAST_LOOP: 1690 case IPV6_JOIN_GROUP: 1691 case IPV6_LEAVE_GROUP: 1692 error = ip6_setmoptions(sopt, inp); 1693 break; 1694 1695 case IPV6_PORTRANGE: 1696 error = sockopt_getint(sopt, &optval); 1697 if (error) 1698 break; 1699 1700 switch (optval) { 1701 case IPV6_PORTRANGE_DEFAULT: 1702 inp->inp_flags &= ~(IN6P_LOWPORT); 1703 inp->inp_flags &= ~(IN6P_HIGHPORT); 1704 break; 1705 1706 case IPV6_PORTRANGE_HIGH: 1707 inp->inp_flags &= ~(IN6P_LOWPORT); 1708 inp->inp_flags |= IN6P_HIGHPORT; 1709 break; 1710 1711 case IPV6_PORTRANGE_LOW: 1712 inp->inp_flags &= ~(IN6P_HIGHPORT); 1713 inp->inp_flags |= IN6P_LOWPORT; 1714 break; 1715 1716 default: 1717 error = EINVAL; 1718 break; 1719 } 1720 break; 1721 1722 case IPV6_PORTALGO: 1723 error = sockopt_getint(sopt, &optval); 1724 if (error) 1725 break; 1726 1727 error = portalgo_algo_index_select(inp, optval); 1728 break; 1729 1730 #if defined(IPSEC) 1731 case IPV6_IPSEC_POLICY: 1732 if (ipsec_enabled) { 1733 error = ipsec_set_policy(inp, 1734 sopt->sopt_data, sopt->sopt_size, 1735 kauth_cred_get()); 1736 } else 1737 error = ENOPROTOOPT; 1738 break; 1739 #endif /* IPSEC */ 1740 1741 default: 1742 error = ENOPROTOOPT; 1743 break; 1744 } 1745 break; 1746 1747 case PRCO_GETOPT: 1748 switch (optname) { 1749 #ifdef RFC2292 1750 case IPV6_2292PKTOPTIONS: 1751 /* 1752 * RFC3542 (effectively) deprecated the 1753 * semantics of the 2292-style pktoptions. 1754 * Since it was not reliable in nature (i.e., 1755 * applications had to expect the lack of some 1756 * information after all), it would make sense 1757 * to simplify this part by always returning 1758 * empty data. 1759 */ 1760 break; 1761 #endif 1762 1763 case IPV6_RECVHOPOPTS: 1764 case IPV6_RECVDSTOPTS: 1765 case IPV6_RECVRTHDRDSTOPTS: 1766 case IPV6_UNICAST_HOPS: 1767 case IPV6_RECVPKTINFO: 1768 case IPV6_RECVHOPLIMIT: 1769 case IPV6_RECVRTHDR: 1770 case IPV6_RECVPATHMTU: 1771 1772 case IPV6_FAITH: 1773 case IPV6_V6ONLY: 1774 case IPV6_PORTRANGE: 1775 case IPV6_RECVTCLASS: 1776 case IPV6_BINDANY: 1777 switch (optname) { 1778 1779 case IPV6_RECVHOPOPTS: 1780 optval = OPTBIT(IN6P_HOPOPTS); 1781 break; 1782 1783 case IPV6_RECVDSTOPTS: 1784 optval = OPTBIT(IN6P_DSTOPTS); 1785 break; 1786 1787 case IPV6_RECVRTHDRDSTOPTS: 1788 optval = OPTBIT(IN6P_RTHDRDSTOPTS); 1789 break; 1790 1791 case IPV6_UNICAST_HOPS: 1792 optval = in6p_hops6(inp); 1793 break; 1794 1795 case IPV6_RECVPKTINFO: 1796 optval = OPTBIT(IN6P_PKTINFO); 1797 break; 1798 1799 case IPV6_RECVHOPLIMIT: 1800 optval = OPTBIT(IN6P_HOPLIMIT); 1801 break; 1802 1803 case IPV6_RECVRTHDR: 1804 optval = OPTBIT(IN6P_RTHDR); 1805 break; 1806 1807 case IPV6_RECVPATHMTU: 1808 optval = OPTBIT(IN6P_MTU); 1809 break; 1810 1811 case IPV6_FAITH: 1812 optval = OPTBIT(IN6P_FAITH); 1813 break; 1814 1815 case IPV6_V6ONLY: 1816 optval = OPTBIT(IN6P_IPV6_V6ONLY); 1817 break; 1818 1819 case IPV6_PORTRANGE: 1820 { 1821 int flags; 1822 flags = inp->inp_flags; 1823 if (flags & IN6P_HIGHPORT) 1824 optval = IPV6_PORTRANGE_HIGH; 1825 else if (flags & IN6P_LOWPORT) 1826 optval = IPV6_PORTRANGE_LOW; 1827 else 1828 optval = 0; 1829 break; 1830 } 1831 case IPV6_RECVTCLASS: 1832 optval = OPTBIT(IN6P_TCLASS); 1833 break; 1834 1835 case IPV6_BINDANY: 1836 optval = OPTBIT(IN6P_BINDANY); 1837 break; 1838 } 1839 1840 if (error) 1841 break; 1842 error = sockopt_setint(sopt, optval); 1843 break; 1844 1845 case IPV6_PATHMTU: 1846 { 1847 u_long pmtu = 0; 1848 struct ip6_mtuinfo mtuinfo; 1849 struct route *ro = &inp->inp_route; 1850 struct rtentry *rt; 1851 union { 1852 struct sockaddr dst; 1853 struct sockaddr_in6 dst6; 1854 } u; 1855 1856 if (!(so->so_state & SS_ISCONNECTED)) 1857 return (ENOTCONN); 1858 /* 1859 * XXX: we dot not consider the case of source 1860 * routing, or optional information to specify 1861 * the outgoing interface. 1862 */ 1863 sockaddr_in6_init(&u.dst6, &in6p_faddr(inp), 0, 0, 0); 1864 rt = rtcache_lookup(ro, &u.dst); 1865 error = ip6_getpmtu(rt, NULL, &pmtu, NULL); 1866 rtcache_unref(rt, ro); 1867 if (error) 1868 break; 1869 if (pmtu > IPV6_MAXPACKET) 1870 pmtu = IPV6_MAXPACKET; 1871 1872 memset(&mtuinfo, 0, sizeof(mtuinfo)); 1873 mtuinfo.ip6m_mtu = (u_int32_t)pmtu; 1874 optdata = (void *)&mtuinfo; 1875 optdatalen = sizeof(mtuinfo); 1876 if (optdatalen > MCLBYTES) 1877 return (EMSGSIZE); /* XXX */ 1878 error = sockopt_set(sopt, optdata, optdatalen); 1879 break; 1880 } 1881 1882 #ifdef RFC2292 1883 case IPV6_2292PKTINFO: 1884 case IPV6_2292HOPLIMIT: 1885 case IPV6_2292HOPOPTS: 1886 case IPV6_2292RTHDR: 1887 case IPV6_2292DSTOPTS: 1888 switch (optname) { 1889 case IPV6_2292PKTINFO: 1890 optval = OPTBIT(IN6P_PKTINFO); 1891 break; 1892 case IPV6_2292HOPLIMIT: 1893 optval = OPTBIT(IN6P_HOPLIMIT); 1894 break; 1895 case IPV6_2292HOPOPTS: 1896 optval = OPTBIT(IN6P_HOPOPTS); 1897 break; 1898 case IPV6_2292RTHDR: 1899 optval = OPTBIT(IN6P_RTHDR); 1900 break; 1901 case IPV6_2292DSTOPTS: 1902 optval = OPTBIT(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); 1903 break; 1904 } 1905 error = sockopt_setint(sopt, optval); 1906 break; 1907 #endif 1908 case IPV6_PKTINFO: 1909 case IPV6_HOPOPTS: 1910 case IPV6_RTHDR: 1911 case IPV6_DSTOPTS: 1912 case IPV6_RTHDRDSTOPTS: 1913 case IPV6_NEXTHOP: 1914 case IPV6_OTCLASS: 1915 case IPV6_TCLASS: 1916 case IPV6_DONTFRAG: 1917 case IPV6_USE_MIN_MTU: 1918 case IPV6_PREFER_TEMPADDR: 1919 error = ip6_getpcbopt(in6p_outputopts(inp), 1920 optname, sopt); 1921 break; 1922 1923 case IPV6_MULTICAST_IF: 1924 case IPV6_MULTICAST_HOPS: 1925 case IPV6_MULTICAST_LOOP: 1926 case IPV6_JOIN_GROUP: 1927 case IPV6_LEAVE_GROUP: 1928 error = ip6_getmoptions(sopt, inp); 1929 break; 1930 1931 case IPV6_PORTALGO: 1932 optval = inp->inp_portalgo; 1933 error = sockopt_setint(sopt, optval); 1934 break; 1935 1936 #if defined(IPSEC) 1937 case IPV6_IPSEC_POLICY: 1938 if (ipsec_used) { 1939 struct mbuf *m = NULL; 1940 1941 /* 1942 * XXX: this will return EINVAL as sopt is 1943 * empty 1944 */ 1945 error = ipsec_get_policy(inp, sopt->sopt_data, 1946 sopt->sopt_size, &m); 1947 if (!error) 1948 error = sockopt_setmbuf(sopt, m); 1949 } else 1950 error = ENOPROTOOPT; 1951 break; 1952 #endif /* IPSEC */ 1953 1954 default: 1955 error = ENOPROTOOPT; 1956 break; 1957 } 1958 break; 1959 } 1960 return (error); 1961 } 1962 1963 int 1964 ip6_raw_ctloutput(int op, struct socket *so, struct sockopt *sopt) 1965 { 1966 int error = 0, optval; 1967 const int icmp6off = offsetof(struct icmp6_hdr, icmp6_cksum); 1968 struct inpcb *inp = sotoinpcb(so); 1969 int level, optname; 1970 1971 KASSERT(sopt != NULL); 1972 1973 level = sopt->sopt_level; 1974 optname = sopt->sopt_name; 1975 1976 if (level != IPPROTO_IPV6) { 1977 return ENOPROTOOPT; 1978 } 1979 1980 switch (optname) { 1981 case IPV6_CHECKSUM: 1982 /* 1983 * For ICMPv6 sockets, no modification allowed for checksum 1984 * offset, permit "no change" values to help existing apps. 1985 * 1986 * XXX RFC3542 says: "An attempt to set IPV6_CHECKSUM 1987 * for an ICMPv6 socket will fail." The current 1988 * behavior does not meet RFC3542. 1989 */ 1990 switch (op) { 1991 case PRCO_SETOPT: 1992 error = sockopt_getint(sopt, &optval); 1993 if (error) 1994 break; 1995 if (optval < -1 || 1996 (optval > 0 && (optval % 2) != 0)) { 1997 /* 1998 * The API assumes non-negative even offset 1999 * values or -1 as a special value. 2000 */ 2001 error = EINVAL; 2002 } else if (so->so_proto->pr_protocol == 2003 IPPROTO_ICMPV6) { 2004 if (optval != icmp6off) 2005 error = EINVAL; 2006 } else 2007 in6p_cksum(inp) = optval; 2008 break; 2009 2010 case PRCO_GETOPT: 2011 if (so->so_proto->pr_protocol == IPPROTO_ICMPV6) 2012 optval = icmp6off; 2013 else 2014 optval = in6p_cksum(inp); 2015 2016 error = sockopt_setint(sopt, optval); 2017 break; 2018 2019 default: 2020 error = EINVAL; 2021 break; 2022 } 2023 break; 2024 2025 default: 2026 error = ENOPROTOOPT; 2027 break; 2028 } 2029 2030 return (error); 2031 } 2032 2033 #ifdef RFC2292 2034 /* 2035 * Set up IP6 options in pcb for insertion in output packets or 2036 * specifying behavior of outgoing packets. 2037 */ 2038 static int 2039 ip6_pcbopts(struct ip6_pktopts **pktopt, struct socket *so, 2040 struct sockopt *sopt) 2041 { 2042 struct ip6_pktopts *opt = *pktopt; 2043 struct mbuf *m; 2044 int error = 0; 2045 2046 KASSERT(solocked(so)); 2047 2048 /* turn off any old options. */ 2049 if (opt) { 2050 #ifdef DIAGNOSTIC 2051 if (opt->ip6po_pktinfo || opt->ip6po_nexthop || 2052 opt->ip6po_hbh || opt->ip6po_dest1 || opt->ip6po_dest2 || 2053 opt->ip6po_rhinfo.ip6po_rhi_rthdr) 2054 printf("ip6_pcbopts: all specified options are cleared.\n"); 2055 #endif 2056 ip6_clearpktopts(opt, -1); 2057 } else { 2058 opt = malloc(sizeof(*opt), M_IP6OPT, M_NOWAIT); 2059 if (opt == NULL) 2060 return (ENOBUFS); 2061 } 2062 *pktopt = NULL; 2063 2064 if (sopt == NULL || sopt->sopt_size == 0) { 2065 /* 2066 * Only turning off any previous options, regardless of 2067 * whether the opt is just created or given. 2068 */ 2069 free(opt, M_IP6OPT); 2070 return (0); 2071 } 2072 2073 /* set options specified by user. */ 2074 m = sockopt_getmbuf(sopt); 2075 if (m == NULL) { 2076 free(opt, M_IP6OPT); 2077 return (ENOBUFS); 2078 } 2079 2080 error = ip6_setpktopts(m, opt, NULL, kauth_cred_get(), 2081 so->so_proto->pr_protocol); 2082 m_freem(m); 2083 if (error != 0) { 2084 ip6_clearpktopts(opt, -1); /* XXX: discard all options */ 2085 free(opt, M_IP6OPT); 2086 return (error); 2087 } 2088 *pktopt = opt; 2089 return (0); 2090 } 2091 #endif 2092 2093 /* 2094 * initialize ip6_pktopts. beware that there are non-zero default values in 2095 * the struct. 2096 */ 2097 void 2098 ip6_initpktopts(struct ip6_pktopts *opt) 2099 { 2100 2101 memset(opt, 0, sizeof(*opt)); 2102 opt->ip6po_hlim = -1; /* -1 means default hop limit */ 2103 opt->ip6po_tclass = -1; /* -1 means default traffic class */ 2104 opt->ip6po_minmtu = IP6PO_MINMTU_MCASTONLY; 2105 opt->ip6po_prefer_tempaddr = IP6PO_TEMPADDR_SYSTEM; 2106 } 2107 2108 #define sin6tosa(sin6) ((struct sockaddr *)(sin6)) /* XXX */ 2109 static int 2110 ip6_pcbopt(int optname, u_char *buf, int len, struct ip6_pktopts **pktopt, 2111 kauth_cred_t cred, int uproto) 2112 { 2113 struct ip6_pktopts *opt; 2114 2115 if (*pktopt == NULL) { 2116 *pktopt = malloc(sizeof(struct ip6_pktopts), M_IP6OPT, 2117 M_NOWAIT); 2118 if (*pktopt == NULL) 2119 return (ENOBUFS); 2120 2121 ip6_initpktopts(*pktopt); 2122 } 2123 opt = *pktopt; 2124 2125 return (ip6_setpktopt(optname, buf, len, opt, cred, 1, 0, uproto)); 2126 } 2127 2128 static int 2129 ip6_getpcbopt(struct ip6_pktopts *pktopt, int optname, struct sockopt *sopt) 2130 { 2131 void *optdata = NULL; 2132 int optdatalen = 0; 2133 struct ip6_ext *ip6e; 2134 int error = 0; 2135 struct in6_pktinfo null_pktinfo; 2136 int deftclass = 0, on; 2137 int defminmtu = IP6PO_MINMTU_MCASTONLY; 2138 int defpreftemp = IP6PO_TEMPADDR_SYSTEM; 2139 2140 switch (optname) { 2141 case IPV6_PKTINFO: 2142 if (pktopt && pktopt->ip6po_pktinfo) 2143 optdata = (void *)pktopt->ip6po_pktinfo; 2144 else { 2145 /* XXX: we don't have to do this every time... */ 2146 memset(&null_pktinfo, 0, sizeof(null_pktinfo)); 2147 optdata = (void *)&null_pktinfo; 2148 } 2149 optdatalen = sizeof(struct in6_pktinfo); 2150 break; 2151 case IPV6_OTCLASS: 2152 /* XXX */ 2153 return (EINVAL); 2154 case IPV6_TCLASS: 2155 if (pktopt && pktopt->ip6po_tclass >= 0) 2156 optdata = (void *)&pktopt->ip6po_tclass; 2157 else 2158 optdata = (void *)&deftclass; 2159 optdatalen = sizeof(int); 2160 break; 2161 case IPV6_HOPOPTS: 2162 if (pktopt && pktopt->ip6po_hbh) { 2163 optdata = (void *)pktopt->ip6po_hbh; 2164 ip6e = (struct ip6_ext *)pktopt->ip6po_hbh; 2165 optdatalen = (ip6e->ip6e_len + 1) << 3; 2166 } 2167 break; 2168 case IPV6_RTHDR: 2169 if (pktopt && pktopt->ip6po_rthdr) { 2170 optdata = (void *)pktopt->ip6po_rthdr; 2171 ip6e = (struct ip6_ext *)pktopt->ip6po_rthdr; 2172 optdatalen = (ip6e->ip6e_len + 1) << 3; 2173 } 2174 break; 2175 case IPV6_RTHDRDSTOPTS: 2176 if (pktopt && pktopt->ip6po_dest1) { 2177 optdata = (void *)pktopt->ip6po_dest1; 2178 ip6e = (struct ip6_ext *)pktopt->ip6po_dest1; 2179 optdatalen = (ip6e->ip6e_len + 1) << 3; 2180 } 2181 break; 2182 case IPV6_DSTOPTS: 2183 if (pktopt && pktopt->ip6po_dest2) { 2184 optdata = (void *)pktopt->ip6po_dest2; 2185 ip6e = (struct ip6_ext *)pktopt->ip6po_dest2; 2186 optdatalen = (ip6e->ip6e_len + 1) << 3; 2187 } 2188 break; 2189 case IPV6_NEXTHOP: 2190 if (pktopt && pktopt->ip6po_nexthop) { 2191 optdata = (void *)pktopt->ip6po_nexthop; 2192 optdatalen = pktopt->ip6po_nexthop->sa_len; 2193 } 2194 break; 2195 case IPV6_USE_MIN_MTU: 2196 if (pktopt) 2197 optdata = (void *)&pktopt->ip6po_minmtu; 2198 else 2199 optdata = (void *)&defminmtu; 2200 optdatalen = sizeof(int); 2201 break; 2202 case IPV6_DONTFRAG: 2203 if (pktopt && ((pktopt->ip6po_flags) & IP6PO_DONTFRAG)) 2204 on = 1; 2205 else 2206 on = 0; 2207 optdata = (void *)&on; 2208 optdatalen = sizeof(on); 2209 break; 2210 case IPV6_PREFER_TEMPADDR: 2211 if (pktopt) 2212 optdata = (void *)&pktopt->ip6po_prefer_tempaddr; 2213 else 2214 optdata = (void *)&defpreftemp; 2215 optdatalen = sizeof(int); 2216 break; 2217 default: /* should not happen */ 2218 #ifdef DIAGNOSTIC 2219 panic("ip6_getpcbopt: unexpected option\n"); 2220 #endif 2221 return (ENOPROTOOPT); 2222 } 2223 2224 error = sockopt_set(sopt, optdata, optdatalen); 2225 2226 return (error); 2227 } 2228 2229 void 2230 ip6_clearpktopts(struct ip6_pktopts *pktopt, int optname) 2231 { 2232 if (optname == -1 || optname == IPV6_PKTINFO) { 2233 if (pktopt->ip6po_pktinfo) 2234 free(pktopt->ip6po_pktinfo, M_IP6OPT); 2235 pktopt->ip6po_pktinfo = NULL; 2236 } 2237 if (optname == -1 || optname == IPV6_HOPLIMIT) 2238 pktopt->ip6po_hlim = -1; 2239 if (optname == -1 || optname == IPV6_TCLASS) 2240 pktopt->ip6po_tclass = -1; 2241 if (optname == -1 || optname == IPV6_NEXTHOP) { 2242 rtcache_free(&pktopt->ip6po_nextroute); 2243 if (pktopt->ip6po_nexthop) 2244 free(pktopt->ip6po_nexthop, M_IP6OPT); 2245 pktopt->ip6po_nexthop = NULL; 2246 } 2247 if (optname == -1 || optname == IPV6_HOPOPTS) { 2248 if (pktopt->ip6po_hbh) 2249 free(pktopt->ip6po_hbh, M_IP6OPT); 2250 pktopt->ip6po_hbh = NULL; 2251 } 2252 if (optname == -1 || optname == IPV6_RTHDRDSTOPTS) { 2253 if (pktopt->ip6po_dest1) 2254 free(pktopt->ip6po_dest1, M_IP6OPT); 2255 pktopt->ip6po_dest1 = NULL; 2256 } 2257 if (optname == -1 || optname == IPV6_RTHDR) { 2258 if (pktopt->ip6po_rhinfo.ip6po_rhi_rthdr) 2259 free(pktopt->ip6po_rhinfo.ip6po_rhi_rthdr, M_IP6OPT); 2260 pktopt->ip6po_rhinfo.ip6po_rhi_rthdr = NULL; 2261 rtcache_free(&pktopt->ip6po_route); 2262 } 2263 if (optname == -1 || optname == IPV6_DSTOPTS) { 2264 if (pktopt->ip6po_dest2) 2265 free(pktopt->ip6po_dest2, M_IP6OPT); 2266 pktopt->ip6po_dest2 = NULL; 2267 } 2268 } 2269 2270 #define PKTOPT_EXTHDRCPY(type) \ 2271 do { \ 2272 if (src->type) { \ 2273 int hlen = (((struct ip6_ext *)src->type)->ip6e_len + 1) << 3;\ 2274 dst->type = malloc(hlen, M_IP6OPT, canwait); \ 2275 if (dst->type == NULL) \ 2276 goto bad; \ 2277 memcpy(dst->type, src->type, hlen); \ 2278 } \ 2279 } while (/*CONSTCOND*/ 0) 2280 2281 static int 2282 copypktopts(struct ip6_pktopts *dst, struct ip6_pktopts *src, int canwait) 2283 { 2284 dst->ip6po_hlim = src->ip6po_hlim; 2285 dst->ip6po_tclass = src->ip6po_tclass; 2286 dst->ip6po_flags = src->ip6po_flags; 2287 dst->ip6po_minmtu = src->ip6po_minmtu; 2288 dst->ip6po_prefer_tempaddr = src->ip6po_prefer_tempaddr; 2289 if (src->ip6po_pktinfo) { 2290 dst->ip6po_pktinfo = malloc(sizeof(*dst->ip6po_pktinfo), 2291 M_IP6OPT, canwait); 2292 if (dst->ip6po_pktinfo == NULL) 2293 goto bad; 2294 *dst->ip6po_pktinfo = *src->ip6po_pktinfo; 2295 } 2296 if (src->ip6po_nexthop) { 2297 dst->ip6po_nexthop = malloc(src->ip6po_nexthop->sa_len, 2298 M_IP6OPT, canwait); 2299 if (dst->ip6po_nexthop == NULL) 2300 goto bad; 2301 memcpy(dst->ip6po_nexthop, src->ip6po_nexthop, 2302 src->ip6po_nexthop->sa_len); 2303 } 2304 PKTOPT_EXTHDRCPY(ip6po_hbh); 2305 PKTOPT_EXTHDRCPY(ip6po_dest1); 2306 PKTOPT_EXTHDRCPY(ip6po_dest2); 2307 PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */ 2308 return (0); 2309 2310 bad: 2311 if (dst->ip6po_pktinfo) free(dst->ip6po_pktinfo, M_IP6OPT); 2312 if (dst->ip6po_nexthop) free(dst->ip6po_nexthop, M_IP6OPT); 2313 if (dst->ip6po_hbh) free(dst->ip6po_hbh, M_IP6OPT); 2314 if (dst->ip6po_dest1) free(dst->ip6po_dest1, M_IP6OPT); 2315 if (dst->ip6po_dest2) free(dst->ip6po_dest2, M_IP6OPT); 2316 if (dst->ip6po_rthdr) free(dst->ip6po_rthdr, M_IP6OPT); 2317 2318 return (ENOBUFS); 2319 } 2320 #undef PKTOPT_EXTHDRCPY 2321 2322 struct ip6_pktopts * 2323 ip6_copypktopts(struct ip6_pktopts *src, int canwait) 2324 { 2325 int error; 2326 struct ip6_pktopts *dst; 2327 2328 dst = malloc(sizeof(*dst), M_IP6OPT, canwait); 2329 if (dst == NULL) 2330 return (NULL); 2331 ip6_initpktopts(dst); 2332 2333 if ((error = copypktopts(dst, src, canwait)) != 0) { 2334 free(dst, M_IP6OPT); 2335 return (NULL); 2336 } 2337 2338 return (dst); 2339 } 2340 2341 void 2342 ip6_freepcbopts(struct ip6_pktopts *pktopt) 2343 { 2344 if (pktopt == NULL) 2345 return; 2346 2347 ip6_clearpktopts(pktopt, -1); 2348 2349 free(pktopt, M_IP6OPT); 2350 } 2351 2352 int 2353 ip6_get_membership(const struct sockopt *sopt, struct ifnet **ifp, 2354 struct psref *psref, void *v, size_t l) 2355 { 2356 struct ipv6_mreq mreq; 2357 int error; 2358 struct in6_addr *ia = &mreq.ipv6mr_multiaddr; 2359 struct in_addr *ia4 = (void *)&ia->s6_addr32[3]; 2360 2361 error = sockopt_get(sopt, &mreq, sizeof(mreq)); 2362 if (error != 0) 2363 return error; 2364 2365 if (IN6_IS_ADDR_UNSPECIFIED(ia)) { 2366 /* 2367 * We use the unspecified address to specify to accept 2368 * all multicast addresses. Only super user is allowed 2369 * to do this. 2370 */ 2371 if (kauth_authorize_network(kauth_cred_get(), 2372 KAUTH_NETWORK_IPV6, 2373 KAUTH_REQ_NETWORK_IPV6_JOIN_MULTICAST, NULL, NULL, NULL)) 2374 return EACCES; 2375 } else if (IN6_IS_ADDR_V4MAPPED(ia)) { 2376 // Don't bother if we are not going to use ifp. 2377 if (l == sizeof(*ia)) { 2378 memcpy(v, ia, l); 2379 return 0; 2380 } 2381 } else if (!IN6_IS_ADDR_MULTICAST(ia)) { 2382 return EINVAL; 2383 } 2384 2385 /* 2386 * If no interface was explicitly specified, choose an 2387 * appropriate one according to the given multicast address. 2388 */ 2389 if (mreq.ipv6mr_interface == 0) { 2390 struct rtentry *rt; 2391 union { 2392 struct sockaddr dst; 2393 struct sockaddr_in dst4; 2394 struct sockaddr_in6 dst6; 2395 } u; 2396 struct route ro; 2397 2398 /* 2399 * Look up the routing table for the 2400 * address, and choose the outgoing interface. 2401 * XXX: is it a good approach? 2402 */ 2403 memset(&ro, 0, sizeof(ro)); 2404 if (IN6_IS_ADDR_V4MAPPED(ia)) 2405 sockaddr_in_init(&u.dst4, ia4, 0); 2406 else 2407 sockaddr_in6_init(&u.dst6, ia, 0, 0, 0); 2408 error = rtcache_setdst(&ro, &u.dst); 2409 if (error != 0) 2410 return error; 2411 rt = rtcache_init(&ro); 2412 *ifp = rt != NULL ? 2413 if_get_byindex(rt->rt_ifp->if_index, psref) : NULL; 2414 rtcache_unref(rt, &ro); 2415 rtcache_free(&ro); 2416 } else { 2417 /* 2418 * If the interface is specified, validate it. 2419 */ 2420 *ifp = if_get_byindex(mreq.ipv6mr_interface, psref); 2421 if (*ifp == NULL) 2422 return ENXIO; /* XXX EINVAL? */ 2423 } 2424 if (sizeof(*ia) == l) 2425 memcpy(v, ia, l); 2426 else 2427 memcpy(v, ia4, l); 2428 return 0; 2429 } 2430 2431 /* 2432 * Set the IP6 multicast options in response to user setsockopt(). 2433 */ 2434 static int 2435 ip6_setmoptions(const struct sockopt *sopt, struct inpcb *inp) 2436 { 2437 int error = 0; 2438 u_int loop, ifindex; 2439 struct ipv6_mreq mreq; 2440 struct in6_addr ia; 2441 struct ifnet *ifp; 2442 struct ip6_moptions *im6o = in6p_moptions(inp); 2443 struct in6_multi_mship *imm; 2444 2445 KASSERT(inp_locked(inp)); 2446 2447 if (im6o == NULL) { 2448 /* 2449 * No multicast option buffer attached to the pcb; 2450 * allocate one and initialize to default values. 2451 */ 2452 im6o = malloc(sizeof(*im6o), M_IPMOPTS, M_NOWAIT); 2453 if (im6o == NULL) 2454 return (ENOBUFS); 2455 in6p_moptions(inp) = im6o; 2456 im6o->im6o_multicast_if_index = 0; 2457 im6o->im6o_multicast_hlim = ip6_defmcasthlim; 2458 im6o->im6o_multicast_loop = IPV6_DEFAULT_MULTICAST_LOOP; 2459 LIST_INIT(&im6o->im6o_memberships); 2460 } 2461 2462 switch (sopt->sopt_name) { 2463 2464 case IPV6_MULTICAST_IF: { 2465 int s; 2466 /* 2467 * Select the interface for outgoing multicast packets. 2468 */ 2469 error = sockopt_get(sopt, &ifindex, sizeof(ifindex)); 2470 if (error != 0) 2471 break; 2472 2473 s = pserialize_read_enter(); 2474 if (ifindex != 0) { 2475 if ((ifp = if_byindex(ifindex)) == NULL) { 2476 pserialize_read_exit(s); 2477 error = ENXIO; /* XXX EINVAL? */ 2478 break; 2479 } 2480 if ((ifp->if_flags & IFF_MULTICAST) == 0) { 2481 pserialize_read_exit(s); 2482 error = EADDRNOTAVAIL; 2483 break; 2484 } 2485 } else 2486 ifp = NULL; 2487 im6o->im6o_multicast_if_index = if_get_index(ifp); 2488 pserialize_read_exit(s); 2489 break; 2490 } 2491 2492 case IPV6_MULTICAST_HOPS: 2493 { 2494 /* 2495 * Set the IP6 hoplimit for outgoing multicast packets. 2496 */ 2497 int optval; 2498 2499 error = sockopt_getint(sopt, &optval); 2500 if (error != 0) 2501 break; 2502 2503 if (optval < -1 || optval >= 256) 2504 error = EINVAL; 2505 else if (optval == -1) 2506 im6o->im6o_multicast_hlim = ip6_defmcasthlim; 2507 else 2508 im6o->im6o_multicast_hlim = optval; 2509 break; 2510 } 2511 2512 case IPV6_MULTICAST_LOOP: 2513 /* 2514 * Set the loopback flag for outgoing multicast packets. 2515 * Must be zero or one. 2516 */ 2517 error = sockopt_get(sopt, &loop, sizeof(loop)); 2518 if (error != 0) 2519 break; 2520 if (loop > 1) { 2521 error = EINVAL; 2522 break; 2523 } 2524 im6o->im6o_multicast_loop = loop; 2525 break; 2526 2527 case IPV6_JOIN_GROUP: { 2528 int bound; 2529 struct psref psref; 2530 /* 2531 * Add a multicast group membership. 2532 * Group must be a valid IP6 multicast address. 2533 */ 2534 bound = curlwp_bind(); 2535 ifp = NULL; 2536 error = ip6_get_membership(sopt, &ifp, &psref, &ia, sizeof(ia)); 2537 if (error != 0) { 2538 KASSERT(ifp == NULL); 2539 curlwp_bindx(bound); 2540 return error; 2541 } 2542 2543 if (IN6_IS_ADDR_V4MAPPED(&ia)) { 2544 error = ip_setmoptions(&inp->inp_moptions, sopt); 2545 goto put_break; 2546 } 2547 /* 2548 * See if we found an interface, and confirm that it 2549 * supports multicast 2550 */ 2551 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) { 2552 error = EADDRNOTAVAIL; 2553 goto put_break; 2554 } 2555 2556 if (in6_setscope(&ia, ifp, NULL)) { 2557 error = EADDRNOTAVAIL; /* XXX: should not happen */ 2558 goto put_break; 2559 } 2560 2561 /* 2562 * See if the membership already exists. 2563 */ 2564 LIST_FOREACH(imm, &im6o->im6o_memberships, i6mm_chain) { 2565 if (imm->i6mm_maddr->in6m_ifp == ifp && 2566 IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr, 2567 &ia)) 2568 goto put_break; 2569 } 2570 if (imm != NULL) { 2571 error = EADDRINUSE; 2572 goto put_break; 2573 } 2574 /* 2575 * Everything looks good; add a new record to the multicast 2576 * address list for the given interface. 2577 */ 2578 imm = in6_joingroup(ifp, &ia, &error, 0); 2579 if (imm == NULL) 2580 goto put_break; 2581 LIST_INSERT_HEAD(&im6o->im6o_memberships, imm, i6mm_chain); 2582 put_break: 2583 if_put(ifp, &psref); 2584 curlwp_bindx(bound); 2585 break; 2586 } 2587 2588 case IPV6_LEAVE_GROUP: { 2589 /* 2590 * Drop a multicast group membership. 2591 * Group must be a valid IP6 multicast address. 2592 */ 2593 error = sockopt_get(sopt, &mreq, sizeof(mreq)); 2594 if (error != 0) 2595 break; 2596 2597 if (IN6_IS_ADDR_V4MAPPED(&mreq.ipv6mr_multiaddr)) { 2598 error = ip_setmoptions(&inp->inp_moptions, sopt); 2599 break; 2600 } 2601 /* 2602 * If an interface address was specified, get a pointer 2603 * to its ifnet structure. 2604 */ 2605 if (mreq.ipv6mr_interface != 0) { 2606 if ((ifp = if_byindex(mreq.ipv6mr_interface)) == NULL) { 2607 error = ENXIO; /* XXX EINVAL? */ 2608 break; 2609 } 2610 } else 2611 ifp = NULL; 2612 2613 /* Fill in the scope zone ID */ 2614 if (ifp) { 2615 if (in6_setscope(&mreq.ipv6mr_multiaddr, ifp, NULL)) { 2616 /* XXX: should not happen */ 2617 error = EADDRNOTAVAIL; 2618 break; 2619 } 2620 } else if (mreq.ipv6mr_interface != 0) { 2621 /* 2622 * XXX: This case would happens when the (positive) 2623 * index is in the valid range, but the corresponding 2624 * interface has been detached dynamically. The above 2625 * check probably avoids such case to happen here, but 2626 * we check it explicitly for safety. 2627 */ 2628 error = EADDRNOTAVAIL; 2629 break; 2630 } else { /* ipv6mr_interface == 0 */ 2631 struct sockaddr_in6 sa6_mc; 2632 2633 /* 2634 * The API spec says as follows: 2635 * If the interface index is specified as 0, the 2636 * system may choose a multicast group membership to 2637 * drop by matching the multicast address only. 2638 * On the other hand, we cannot disambiguate the scope 2639 * zone unless an interface is provided. Thus, we 2640 * check if there's ambiguity with the default scope 2641 * zone as the last resort. 2642 */ 2643 sockaddr_in6_init(&sa6_mc, &mreq.ipv6mr_multiaddr, 2644 0, 0, 0); 2645 error = sa6_embedscope(&sa6_mc, ip6_use_defzone); 2646 if (error != 0) 2647 break; 2648 mreq.ipv6mr_multiaddr = sa6_mc.sin6_addr; 2649 } 2650 2651 /* 2652 * Find the membership in the membership list. 2653 */ 2654 LIST_FOREACH(imm, &im6o->im6o_memberships, i6mm_chain) { 2655 if ((ifp == NULL || imm->i6mm_maddr->in6m_ifp == ifp) && 2656 IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr, 2657 &mreq.ipv6mr_multiaddr)) 2658 break; 2659 } 2660 if (imm == NULL) { 2661 /* Unable to resolve interface */ 2662 error = EADDRNOTAVAIL; 2663 break; 2664 } 2665 /* 2666 * Give up the multicast address record to which the 2667 * membership points. 2668 */ 2669 LIST_REMOVE(imm, i6mm_chain); 2670 in6_leavegroup(imm); 2671 /* in6m_ifp should not leave thanks to inp_lock */ 2672 break; 2673 } 2674 2675 default: 2676 error = EOPNOTSUPP; 2677 break; 2678 } 2679 2680 /* 2681 * If all options have default values, no need to keep the mbuf. 2682 */ 2683 if (im6o->im6o_multicast_if_index == 0 && 2684 im6o->im6o_multicast_hlim == ip6_defmcasthlim && 2685 im6o->im6o_multicast_loop == IPV6_DEFAULT_MULTICAST_LOOP && 2686 LIST_EMPTY(&im6o->im6o_memberships)) { 2687 free(in6p_moptions(inp), M_IPMOPTS); 2688 in6p_moptions(inp) = NULL; 2689 } 2690 2691 return (error); 2692 } 2693 2694 /* 2695 * Return the IP6 multicast options in response to user getsockopt(). 2696 */ 2697 static int 2698 ip6_getmoptions(struct sockopt *sopt, struct inpcb *inp) 2699 { 2700 u_int optval; 2701 int error; 2702 struct ip6_moptions *im6o = in6p_moptions(inp); 2703 2704 switch (sopt->sopt_name) { 2705 case IPV6_MULTICAST_IF: 2706 if (im6o == NULL || im6o->im6o_multicast_if_index == 0) 2707 optval = 0; 2708 else 2709 optval = im6o->im6o_multicast_if_index; 2710 2711 error = sockopt_set(sopt, &optval, sizeof(optval)); 2712 break; 2713 2714 case IPV6_MULTICAST_HOPS: 2715 if (im6o == NULL) 2716 optval = ip6_defmcasthlim; 2717 else 2718 optval = im6o->im6o_multicast_hlim; 2719 2720 error = sockopt_set(sopt, &optval, sizeof(optval)); 2721 break; 2722 2723 case IPV6_MULTICAST_LOOP: 2724 if (im6o == NULL) 2725 optval = IPV6_DEFAULT_MULTICAST_LOOP; 2726 else 2727 optval = im6o->im6o_multicast_loop; 2728 2729 error = sockopt_set(sopt, &optval, sizeof(optval)); 2730 break; 2731 2732 default: 2733 error = EOPNOTSUPP; 2734 } 2735 2736 return (error); 2737 } 2738 2739 /* 2740 * Discard the IP6 multicast options. 2741 */ 2742 void 2743 ip6_freemoptions(struct ip6_moptions *im6o) 2744 { 2745 struct in6_multi_mship *imm, *nimm; 2746 2747 if (im6o == NULL) 2748 return; 2749 2750 /* The owner of im6o (inp) should be protected by solock */ 2751 LIST_FOREACH_SAFE(imm, &im6o->im6o_memberships, i6mm_chain, nimm) { 2752 LIST_REMOVE(imm, i6mm_chain); 2753 in6_leavegroup(imm); 2754 } 2755 free(im6o, M_IPMOPTS); 2756 } 2757 2758 /* 2759 * Set IPv6 outgoing packet options based on advanced API. 2760 */ 2761 int 2762 ip6_setpktopts(struct mbuf *control, struct ip6_pktopts *opt, 2763 struct ip6_pktopts *stickyopt, kauth_cred_t cred, int uproto) 2764 { 2765 struct cmsghdr *cm = 0; 2766 2767 if (control == NULL || opt == NULL) 2768 return (EINVAL); 2769 2770 ip6_initpktopts(opt); 2771 if (stickyopt) { 2772 int error; 2773 2774 /* 2775 * If stickyopt is provided, make a local copy of the options 2776 * for this particular packet, then override them by ancillary 2777 * objects. 2778 * XXX: copypktopts() does not copy the cached route to a next 2779 * hop (if any). This is not very good in terms of efficiency, 2780 * but we can allow this since this option should be rarely 2781 * used. 2782 */ 2783 if ((error = copypktopts(opt, stickyopt, M_NOWAIT)) != 0) 2784 return (error); 2785 } 2786 2787 /* 2788 * XXX: Currently, we assume all the optional information is stored 2789 * in a single mbuf. 2790 */ 2791 if (control->m_next) 2792 return (EINVAL); 2793 2794 /* XXX if cm->cmsg_len is not aligned, control->m_len can become <0 */ 2795 for (; control->m_len > 0; control->m_data += CMSG_ALIGN(cm->cmsg_len), 2796 control->m_len -= CMSG_ALIGN(cm->cmsg_len)) { 2797 int error; 2798 2799 if (control->m_len < CMSG_LEN(0)) 2800 return (EINVAL); 2801 2802 cm = mtod(control, struct cmsghdr *); 2803 if (cm->cmsg_len < CMSG_LEN(0) || cm->cmsg_len > control->m_len) 2804 return (EINVAL); 2805 if (cm->cmsg_level != IPPROTO_IPV6) 2806 continue; 2807 2808 error = ip6_setpktopt(cm->cmsg_type, CMSG_DATA(cm), 2809 cm->cmsg_len - CMSG_LEN(0), opt, cred, 0, 1, uproto); 2810 if (error) 2811 return (error); 2812 } 2813 2814 return (0); 2815 } 2816 2817 /* 2818 * Set a particular packet option, as a sticky option or an ancillary data 2819 * item. "len" can be 0 only when it's a sticky option. 2820 * We have 4 cases of combination of "sticky" and "cmsg": 2821 * "sticky=0, cmsg=0": impossible 2822 * "sticky=0, cmsg=1": RFC2292 or RFC3542 ancillary data 2823 * "sticky=1, cmsg=0": RFC3542 socket option 2824 * "sticky=1, cmsg=1": RFC2292 socket option 2825 */ 2826 static int 2827 ip6_setpktopt(int optname, u_char *buf, int len, struct ip6_pktopts *opt, 2828 kauth_cred_t cred, int sticky, int cmsg, int uproto) 2829 { 2830 int minmtupolicy; 2831 int error; 2832 2833 if (!sticky && !cmsg) { 2834 #ifdef DIAGNOSTIC 2835 printf("ip6_setpktopt: impossible case\n"); 2836 #endif 2837 return (EINVAL); 2838 } 2839 2840 /* 2841 * IPV6_2292xxx is for backward compatibility to RFC2292, and should 2842 * not be specified in the context of RFC3542. Conversely, 2843 * RFC3542 types should not be specified in the context of RFC2292. 2844 */ 2845 if (!cmsg) { 2846 switch (optname) { 2847 case IPV6_2292PKTINFO: 2848 case IPV6_2292HOPLIMIT: 2849 case IPV6_2292NEXTHOP: 2850 case IPV6_2292HOPOPTS: 2851 case IPV6_2292DSTOPTS: 2852 case IPV6_2292RTHDR: 2853 case IPV6_2292PKTOPTIONS: 2854 return (ENOPROTOOPT); 2855 } 2856 } 2857 if (sticky && cmsg) { 2858 switch (optname) { 2859 case IPV6_PKTINFO: 2860 case IPV6_HOPLIMIT: 2861 case IPV6_NEXTHOP: 2862 case IPV6_HOPOPTS: 2863 case IPV6_DSTOPTS: 2864 case IPV6_RTHDRDSTOPTS: 2865 case IPV6_RTHDR: 2866 case IPV6_USE_MIN_MTU: 2867 case IPV6_DONTFRAG: 2868 case IPV6_OTCLASS: 2869 case IPV6_TCLASS: 2870 case IPV6_PREFER_TEMPADDR: /* XXX not an RFC3542 option */ 2871 return (ENOPROTOOPT); 2872 } 2873 } 2874 2875 switch (optname) { 2876 #ifdef RFC2292 2877 case IPV6_2292PKTINFO: 2878 #endif 2879 case IPV6_PKTINFO: 2880 { 2881 struct in6_pktinfo *pktinfo; 2882 2883 if (len != sizeof(struct in6_pktinfo)) 2884 return (EINVAL); 2885 2886 pktinfo = (struct in6_pktinfo *)buf; 2887 2888 /* 2889 * An application can clear any sticky IPV6_PKTINFO option by 2890 * doing a "regular" setsockopt with ipi6_addr being 2891 * in6addr_any and ipi6_ifindex being zero. 2892 * [RFC 3542, Section 6] 2893 */ 2894 if (optname == IPV6_PKTINFO && opt->ip6po_pktinfo && 2895 pktinfo->ipi6_ifindex == 0 && 2896 IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) { 2897 ip6_clearpktopts(opt, optname); 2898 break; 2899 } 2900 2901 if (uproto == IPPROTO_TCP && optname == IPV6_PKTINFO && 2902 sticky && !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) { 2903 return (EINVAL); 2904 } 2905 2906 /* Validate the interface index if specified. */ 2907 if (pktinfo->ipi6_ifindex) { 2908 struct ifnet *ifp; 2909 int s = pserialize_read_enter(); 2910 ifp = if_byindex(pktinfo->ipi6_ifindex); 2911 if (ifp == NULL) { 2912 pserialize_read_exit(s); 2913 return ENXIO; 2914 } 2915 pserialize_read_exit(s); 2916 } 2917 2918 /* 2919 * We store the address anyway, and let in6_selectsrc() 2920 * validate the specified address. This is because ipi6_addr 2921 * may not have enough information about its scope zone, and 2922 * we may need additional information (such as outgoing 2923 * interface or the scope zone of a destination address) to 2924 * disambiguate the scope. 2925 * XXX: the delay of the validation may confuse the 2926 * application when it is used as a sticky option. 2927 */ 2928 if (opt->ip6po_pktinfo == NULL) { 2929 opt->ip6po_pktinfo = malloc(sizeof(*pktinfo), 2930 M_IP6OPT, M_NOWAIT); 2931 if (opt->ip6po_pktinfo == NULL) 2932 return (ENOBUFS); 2933 } 2934 memcpy(opt->ip6po_pktinfo, pktinfo, sizeof(*pktinfo)); 2935 break; 2936 } 2937 2938 #ifdef RFC2292 2939 case IPV6_2292HOPLIMIT: 2940 #endif 2941 case IPV6_HOPLIMIT: 2942 { 2943 int *hlimp; 2944 2945 /* 2946 * RFC 3542 deprecated the usage of sticky IPV6_HOPLIMIT 2947 * to simplify the ordering among hoplimit options. 2948 */ 2949 if (optname == IPV6_HOPLIMIT && sticky) 2950 return (ENOPROTOOPT); 2951 2952 if (len != sizeof(int)) 2953 return (EINVAL); 2954 hlimp = (int *)buf; 2955 if (*hlimp < -1 || *hlimp > 255) 2956 return (EINVAL); 2957 2958 opt->ip6po_hlim = *hlimp; 2959 break; 2960 } 2961 2962 case IPV6_OTCLASS: 2963 if (len != sizeof(u_int8_t)) 2964 return (EINVAL); 2965 2966 opt->ip6po_tclass = *(u_int8_t *)buf; 2967 break; 2968 2969 case IPV6_TCLASS: 2970 { 2971 int tclass; 2972 2973 if (len != sizeof(int)) 2974 return (EINVAL); 2975 tclass = *(int *)buf; 2976 if (tclass < -1 || tclass > 255) 2977 return (EINVAL); 2978 2979 opt->ip6po_tclass = tclass; 2980 break; 2981 } 2982 2983 #ifdef RFC2292 2984 case IPV6_2292NEXTHOP: 2985 #endif 2986 case IPV6_NEXTHOP: 2987 error = kauth_authorize_network(cred, 2988 KAUTH_NETWORK_IPV6, 2989 KAUTH_REQ_NETWORK_IPV6_HOPBYHOP, NULL, NULL, NULL); 2990 if (error) 2991 return (error); 2992 2993 if (len == 0) { /* just remove the option */ 2994 ip6_clearpktopts(opt, IPV6_NEXTHOP); 2995 break; 2996 } 2997 2998 /* check if cmsg_len is large enough for sa_len */ 2999 if (len < sizeof(struct sockaddr) || len < *buf) 3000 return (EINVAL); 3001 3002 switch (((struct sockaddr *)buf)->sa_family) { 3003 case AF_INET6: 3004 { 3005 struct sockaddr_in6 *sa6 = (struct sockaddr_in6 *)buf; 3006 3007 if (sa6->sin6_len != sizeof(struct sockaddr_in6)) 3008 return (EINVAL); 3009 3010 if (IN6_IS_ADDR_UNSPECIFIED(&sa6->sin6_addr) || 3011 IN6_IS_ADDR_MULTICAST(&sa6->sin6_addr)) { 3012 return (EINVAL); 3013 } 3014 if ((error = sa6_embedscope(sa6, ip6_use_defzone)) 3015 != 0) { 3016 return (error); 3017 } 3018 break; 3019 } 3020 case AF_LINK: /* eventually be supported? */ 3021 default: 3022 return (EAFNOSUPPORT); 3023 } 3024 3025 /* turn off the previous option, then set the new option. */ 3026 ip6_clearpktopts(opt, IPV6_NEXTHOP); 3027 opt->ip6po_nexthop = malloc(*buf, M_IP6OPT, M_NOWAIT); 3028 if (opt->ip6po_nexthop == NULL) 3029 return (ENOBUFS); 3030 memcpy(opt->ip6po_nexthop, buf, *buf); 3031 break; 3032 3033 #ifdef RFC2292 3034 case IPV6_2292HOPOPTS: 3035 #endif 3036 case IPV6_HOPOPTS: 3037 { 3038 struct ip6_hbh *hbh; 3039 int hbhlen; 3040 3041 /* 3042 * XXX: We don't allow a non-privileged user to set ANY HbH 3043 * options, since per-option restriction has too much 3044 * overhead. 3045 */ 3046 error = kauth_authorize_network(cred, 3047 KAUTH_NETWORK_IPV6, 3048 KAUTH_REQ_NETWORK_IPV6_HOPBYHOP, NULL, NULL, NULL); 3049 if (error) 3050 return (error); 3051 3052 if (len == 0) { 3053 ip6_clearpktopts(opt, IPV6_HOPOPTS); 3054 break; /* just remove the option */ 3055 } 3056 3057 /* message length validation */ 3058 if (len < sizeof(struct ip6_hbh)) 3059 return (EINVAL); 3060 hbh = (struct ip6_hbh *)buf; 3061 hbhlen = (hbh->ip6h_len + 1) << 3; 3062 if (len != hbhlen) 3063 return (EINVAL); 3064 3065 /* turn off the previous option, then set the new option. */ 3066 ip6_clearpktopts(opt, IPV6_HOPOPTS); 3067 opt->ip6po_hbh = malloc(hbhlen, M_IP6OPT, M_NOWAIT); 3068 if (opt->ip6po_hbh == NULL) 3069 return (ENOBUFS); 3070 memcpy(opt->ip6po_hbh, hbh, hbhlen); 3071 3072 break; 3073 } 3074 3075 #ifdef RFC2292 3076 case IPV6_2292DSTOPTS: 3077 #endif 3078 case IPV6_DSTOPTS: 3079 case IPV6_RTHDRDSTOPTS: 3080 { 3081 struct ip6_dest *dest, **newdest = NULL; 3082 int destlen; 3083 3084 /* XXX: see the comment for IPV6_HOPOPTS */ 3085 error = kauth_authorize_network(cred, 3086 KAUTH_NETWORK_IPV6, 3087 KAUTH_REQ_NETWORK_IPV6_HOPBYHOP, NULL, NULL, NULL); 3088 if (error) 3089 return (error); 3090 3091 if (len == 0) { 3092 ip6_clearpktopts(opt, optname); 3093 break; /* just remove the option */ 3094 } 3095 3096 /* message length validation */ 3097 if (len < sizeof(struct ip6_dest)) 3098 return (EINVAL); 3099 dest = (struct ip6_dest *)buf; 3100 destlen = (dest->ip6d_len + 1) << 3; 3101 if (len != destlen) 3102 return (EINVAL); 3103 /* 3104 * Determine the position that the destination options header 3105 * should be inserted; before or after the routing header. 3106 */ 3107 switch (optname) { 3108 case IPV6_2292DSTOPTS: 3109 /* 3110 * The old advanced API is ambiguous on this point. 3111 * Our approach is to determine the position based 3112 * according to the existence of a routing header. 3113 * Note, however, that this depends on the order of the 3114 * extension headers in the ancillary data; the 1st 3115 * part of the destination options header must appear 3116 * before the routing header in the ancillary data, 3117 * too. 3118 * RFC3542 solved the ambiguity by introducing 3119 * separate ancillary data or option types. 3120 */ 3121 if (opt->ip6po_rthdr == NULL) 3122 newdest = &opt->ip6po_dest1; 3123 else 3124 newdest = &opt->ip6po_dest2; 3125 break; 3126 case IPV6_RTHDRDSTOPTS: 3127 newdest = &opt->ip6po_dest1; 3128 break; 3129 case IPV6_DSTOPTS: 3130 newdest = &opt->ip6po_dest2; 3131 break; 3132 } 3133 3134 /* turn off the previous option, then set the new option. */ 3135 ip6_clearpktopts(opt, optname); 3136 *newdest = malloc(destlen, M_IP6OPT, M_NOWAIT); 3137 if (*newdest == NULL) 3138 return (ENOBUFS); 3139 memcpy(*newdest, dest, destlen); 3140 3141 break; 3142 } 3143 3144 #ifdef RFC2292 3145 case IPV6_2292RTHDR: 3146 #endif 3147 case IPV6_RTHDR: 3148 { 3149 struct ip6_rthdr *rth; 3150 int rthlen; 3151 3152 if (len == 0) { 3153 ip6_clearpktopts(opt, IPV6_RTHDR); 3154 break; /* just remove the option */ 3155 } 3156 3157 /* message length validation */ 3158 if (len < sizeof(struct ip6_rthdr)) 3159 return (EINVAL); 3160 rth = (struct ip6_rthdr *)buf; 3161 rthlen = (rth->ip6r_len + 1) << 3; 3162 if (len != rthlen) 3163 return (EINVAL); 3164 switch (rth->ip6r_type) { 3165 case IPV6_RTHDR_TYPE_0: 3166 /* Dropped, RFC5095. */ 3167 default: 3168 return (EINVAL); /* not supported */ 3169 } 3170 /* turn off the previous option */ 3171 ip6_clearpktopts(opt, IPV6_RTHDR); 3172 opt->ip6po_rthdr = malloc(rthlen, M_IP6OPT, M_NOWAIT); 3173 if (opt->ip6po_rthdr == NULL) 3174 return (ENOBUFS); 3175 memcpy(opt->ip6po_rthdr, rth, rthlen); 3176 break; 3177 } 3178 3179 case IPV6_USE_MIN_MTU: 3180 if (len != sizeof(int)) 3181 return (EINVAL); 3182 minmtupolicy = *(int *)buf; 3183 if (minmtupolicy != IP6PO_MINMTU_MCASTONLY && 3184 minmtupolicy != IP6PO_MINMTU_DISABLE && 3185 minmtupolicy != IP6PO_MINMTU_ALL) { 3186 return (EINVAL); 3187 } 3188 opt->ip6po_minmtu = minmtupolicy; 3189 break; 3190 3191 case IPV6_DONTFRAG: 3192 if (len != sizeof(int)) 3193 return (EINVAL); 3194 3195 if (uproto == IPPROTO_TCP || *(int *)buf == 0) { 3196 /* 3197 * we ignore this option for TCP sockets. 3198 * (RFC3542 leaves this case unspecified.) 3199 */ 3200 opt->ip6po_flags &= ~IP6PO_DONTFRAG; 3201 } else 3202 opt->ip6po_flags |= IP6PO_DONTFRAG; 3203 break; 3204 3205 case IPV6_PREFER_TEMPADDR: 3206 { 3207 int preftemp; 3208 3209 if (len != sizeof(int)) 3210 return (EINVAL); 3211 preftemp = *(int *)buf; 3212 switch (preftemp) { 3213 case IP6PO_TEMPADDR_SYSTEM: 3214 case IP6PO_TEMPADDR_NOTPREFER: 3215 case IP6PO_TEMPADDR_PREFER: 3216 break; 3217 default: 3218 return (EINVAL); 3219 } 3220 opt->ip6po_prefer_tempaddr = preftemp; 3221 break; 3222 } 3223 3224 default: 3225 return (ENOPROTOOPT); 3226 } /* end of switch */ 3227 3228 return (0); 3229 } 3230 3231 /* 3232 * Routine called from ip6_output() to loop back a copy of an IP6 multicast 3233 * packet to the input queue of a specified interface. Note that this 3234 * calls the output routine of the loopback "driver", but with an interface 3235 * pointer that might NOT be lo0ifp -- easier than replicating that code here. 3236 */ 3237 void 3238 ip6_mloopback(struct ifnet *ifp, struct mbuf *m, 3239 const struct sockaddr_in6 *dst) 3240 { 3241 struct mbuf *copym; 3242 struct ip6_hdr *ip6; 3243 3244 copym = m_copypacket(m, M_DONTWAIT); 3245 if (copym == NULL) 3246 return; 3247 3248 /* 3249 * Make sure to deep-copy IPv6 header portion in case the data 3250 * is in an mbuf cluster, so that we can safely override the IPv6 3251 * header portion later. 3252 */ 3253 if ((copym->m_flags & M_EXT) != 0 || 3254 copym->m_len < sizeof(struct ip6_hdr)) { 3255 copym = m_pullup(copym, sizeof(struct ip6_hdr)); 3256 if (copym == NULL) 3257 return; 3258 } 3259 3260 #ifdef DIAGNOSTIC 3261 if (copym->m_len < sizeof(*ip6)) { 3262 m_freem(copym); 3263 return; 3264 } 3265 #endif 3266 3267 ip6 = mtod(copym, struct ip6_hdr *); 3268 /* 3269 * clear embedded scope identifiers if necessary. 3270 * in6_clearscope will touch the addresses only when necessary. 3271 */ 3272 in6_clearscope(&ip6->ip6_src); 3273 in6_clearscope(&ip6->ip6_dst); 3274 3275 (void)looutput(ifp, copym, (const struct sockaddr *)dst, NULL); 3276 } 3277 3278 /* 3279 * Chop IPv6 header off from the payload. 3280 */ 3281 static int 3282 ip6_splithdr(struct mbuf *m, struct ip6_exthdrs *exthdrs) 3283 { 3284 struct mbuf *mh; 3285 struct ip6_hdr *ip6; 3286 3287 ip6 = mtod(m, struct ip6_hdr *); 3288 if (m->m_len > sizeof(*ip6)) { 3289 MGETHDR(mh, M_DONTWAIT, MT_HEADER); 3290 if (mh == NULL) { 3291 m_freem(m); 3292 return ENOBUFS; 3293 } 3294 m_move_pkthdr(mh, m); 3295 m_align(mh, sizeof(*ip6)); 3296 m->m_len -= sizeof(*ip6); 3297 m->m_data += sizeof(*ip6); 3298 mh->m_next = m; 3299 mh->m_len = sizeof(*ip6); 3300 memcpy(mtod(mh, void *), (void *)ip6, sizeof(*ip6)); 3301 m = mh; 3302 } 3303 exthdrs->ip6e_ip6 = m; 3304 return 0; 3305 } 3306 3307 /* 3308 * Compute IPv6 extension header length. 3309 */ 3310 int 3311 ip6_optlen(struct inpcb *inp) 3312 { 3313 int len; 3314 3315 if (!in6p_outputopts(inp)) 3316 return 0; 3317 3318 len = 0; 3319 #define elen(x) \ 3320 (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0) 3321 3322 len += elen(in6p_outputopts(inp)->ip6po_hbh); 3323 len += elen(in6p_outputopts(inp)->ip6po_dest1); 3324 len += elen(in6p_outputopts(inp)->ip6po_rthdr); 3325 len += elen(in6p_outputopts(inp)->ip6po_dest2); 3326 return len; 3327 #undef elen 3328 } 3329 3330 /* 3331 * Ensure sending address is valid. 3332 * Returns 0 on success, -1 if an error should be sent back or 1 3333 * if the packet could be dropped without error (protocol dependent). 3334 */ 3335 static int 3336 ip6_ifaddrvalid(const struct in6_addr *src, const struct in6_addr *dst) 3337 { 3338 struct sockaddr_in6 sin6; 3339 int s, error; 3340 struct ifaddr *ifa; 3341 struct in6_ifaddr *ia6; 3342 3343 if (IN6_IS_ADDR_UNSPECIFIED(src)) 3344 return 0; 3345 3346 memset(&sin6, 0, sizeof(sin6)); 3347 sin6.sin6_family = AF_INET6; 3348 sin6.sin6_len = sizeof(sin6); 3349 sin6.sin6_addr = *src; 3350 3351 s = pserialize_read_enter(); 3352 ifa = ifa_ifwithaddr(sin6tosa(&sin6)); 3353 if ((ia6 = ifatoia6(ifa)) == NULL || 3354 ia6->ia6_flags & (IN6_IFF_ANYCAST | IN6_IFF_DUPLICATED)) 3355 error = -1; 3356 else if (ia6->ia6_flags & IN6_IFF_TENTATIVE) 3357 error = 1; 3358 else if (ia6->ia6_flags & IN6_IFF_DETACHED && 3359 (sin6.sin6_addr = *dst, ifa_ifwithaddr(sin6tosa(&sin6)) == NULL)) 3360 /* Allow internal traffic to DETACHED addresses */ 3361 error = 1; 3362 else 3363 error = 0; 3364 pserialize_read_exit(s); 3365 3366 return error; 3367 } 3368
Indexes created Wed Oct 01 07:09:59 GMT 2025