ip6_output.c revision 1.203.2.5
1 /* $NetBSD: ip6_output.c,v 1.203.2.5 2018/09/06 06:56:45 pgoyette 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 1.203.2.5 2018/09/06 06:56:45 pgoyette Exp $"); 66 67 #ifdef _KERNEL_OPT 68 #include "opt_inet.h" 69 #include "opt_inet6.h" 70 #include "opt_ipsec.h" 71 #endif 72 73 #include <sys/param.h> 74 #include <sys/malloc.h> 75 #include <sys/mbuf.h> 76 #include <sys/errno.h> 77 #include <sys/socket.h> 78 #include <sys/socketvar.h> 79 #include <sys/syslog.h> 80 #include <sys/systm.h> 81 #include <sys/proc.h> 82 #include <sys/kauth.h> 83 84 #include <net/if.h> 85 #include <net/route.h> 86 #include <net/pfil.h> 87 88 #include <netinet/in.h> 89 #include <netinet/in_var.h> 90 #include <netinet/ip6.h> 91 #include <netinet/ip_var.h> 92 #include <netinet/icmp6.h> 93 #include <netinet/in_offload.h> 94 #include <netinet/portalgo.h> 95 #include <netinet6/in6_offload.h> 96 #include <netinet6/ip6_var.h> 97 #include <netinet6/ip6_private.h> 98 #include <netinet6/in6_pcb.h> 99 #include <netinet6/nd6.h> 100 #include <netinet6/ip6protosw.h> 101 #include <netinet6/scope6_var.h> 102 103 #ifdef IPSEC 104 #include <netipsec/ipsec.h> 105 #include <netipsec/ipsec6.h> 106 #include <netipsec/key.h> 107 #endif 108 109 extern pfil_head_t *inet6_pfil_hook; /* XXX */ 110 111 struct ip6_exthdrs { 112 struct mbuf *ip6e_ip6; 113 struct mbuf *ip6e_hbh; 114 struct mbuf *ip6e_dest1; 115 struct mbuf *ip6e_rthdr; 116 struct mbuf *ip6e_dest2; 117 }; 118 119 static int ip6_pcbopt(int, u_char *, int, struct ip6_pktopts **, 120 kauth_cred_t, int); 121 static int ip6_getpcbopt(struct ip6_pktopts *, int, struct sockopt *); 122 static int ip6_setpktopt(int, u_char *, int, struct ip6_pktopts *, kauth_cred_t, 123 int, int, int); 124 static int ip6_setmoptions(const struct sockopt *, struct in6pcb *); 125 static int ip6_getmoptions(struct sockopt *, struct in6pcb *); 126 static int ip6_copyexthdr(struct mbuf **, void *, int); 127 static int ip6_insertfraghdr(struct mbuf *, struct mbuf *, int, 128 struct ip6_frag **); 129 static int ip6_insert_jumboopt(struct ip6_exthdrs *, u_int32_t); 130 static int ip6_splithdr(struct mbuf *, struct ip6_exthdrs *); 131 static int ip6_getpmtu(struct rtentry *, struct ifnet *, u_long *, int *); 132 static int copypktopts(struct ip6_pktopts *, struct ip6_pktopts *, int); 133 static int ip6_ifaddrvalid(const struct in6_addr *, const struct in6_addr *); 134 static int ip6_handle_rthdr(struct ip6_rthdr *, struct ip6_hdr *); 135 136 #ifdef RFC2292 137 static int ip6_pcbopts(struct ip6_pktopts **, struct socket *, struct sockopt *); 138 #endif 139 140 static int 141 ip6_handle_rthdr(struct ip6_rthdr *rh, struct ip6_hdr *ip6) 142 { 143 int error = 0; 144 145 switch (rh->ip6r_type) { 146 case IPV6_RTHDR_TYPE_0: 147 /* Dropped, RFC5095. */ 148 default: /* is it possible? */ 149 error = EINVAL; 150 } 151 152 return error; 153 } 154 155 /* 156 * Send an IP packet to a host. 157 */ 158 int 159 ip6_if_output(struct ifnet * const ifp, struct ifnet * const origifp, 160 struct mbuf * const m, const struct sockaddr_in6 * const dst, 161 const struct rtentry *rt) 162 { 163 int error = 0; 164 165 if (rt != NULL) { 166 error = rt_check_reject_route(rt, ifp); 167 if (error != 0) { 168 m_freem(m); 169 return error; 170 } 171 } 172 173 if ((ifp->if_flags & IFF_LOOPBACK) != 0) 174 error = if_output_lock(ifp, origifp, m, sin6tocsa(dst), rt); 175 else 176 error = if_output_lock(ifp, ifp, m, sin6tocsa(dst), rt); 177 return error; 178 } 179 180 /* 181 * IP6 output. The packet in mbuf chain m contains a skeletal IP6 182 * header (with pri, len, nxt, hlim, src, dst). 183 * 184 * This function may modify ver and hlim only. The mbuf chain containing the 185 * packet will be freed. The mbuf opt, if present, will not be freed. 186 * 187 * Type of "mtu": rt_rmx.rmx_mtu is u_long, ifnet.ifr_mtu is int, and 188 * nd_ifinfo.linkmtu is u_int32_t. So we use u_long to hold largest one, 189 * which is rt_rmx.rmx_mtu. 190 */ 191 int 192 ip6_output( 193 struct mbuf *m0, 194 struct ip6_pktopts *opt, 195 struct route *ro, 196 int flags, 197 struct ip6_moptions *im6o, 198 struct in6pcb *in6p, 199 struct ifnet **ifpp /* XXX: just for statistics */ 200 ) 201 { 202 struct ip6_hdr *ip6, *mhip6; 203 struct ifnet *ifp = NULL, *origifp = NULL; 204 struct mbuf *m = m0; 205 int tlen, len, off; 206 bool tso; 207 struct route ip6route; 208 struct rtentry *rt = NULL, *rt_pmtu; 209 const struct sockaddr_in6 *dst; 210 struct sockaddr_in6 src_sa, dst_sa; 211 int error = 0; 212 struct in6_ifaddr *ia = NULL; 213 u_long mtu; 214 int alwaysfrag, dontfrag; 215 u_int32_t optlen = 0, plen = 0, unfragpartlen = 0; 216 struct ip6_exthdrs exthdrs; 217 struct in6_addr finaldst, src0, dst0; 218 u_int32_t zone; 219 struct route *ro_pmtu = NULL; 220 int hdrsplit = 0; 221 int needipsec = 0; 222 #ifdef IPSEC 223 struct secpolicy *sp = NULL; 224 #endif 225 struct psref psref, psref_ia; 226 int bound = curlwp_bind(); 227 bool release_psref_ia = false; 228 229 #ifdef DIAGNOSTIC 230 if ((m->m_flags & M_PKTHDR) == 0) 231 panic("ip6_output: no HDR"); 232 if ((m->m_pkthdr.csum_flags & 233 (M_CSUM_TCPv4|M_CSUM_UDPv4|M_CSUM_TSOv4)) != 0) { 234 panic("ip6_output: IPv4 checksum offload flags: %d", 235 m->m_pkthdr.csum_flags); 236 } 237 if ((m->m_pkthdr.csum_flags & (M_CSUM_TCPv6|M_CSUM_UDPv6)) == 238 (M_CSUM_TCPv6|M_CSUM_UDPv6)) { 239 panic("ip6_output: conflicting checksum offload flags: %d", 240 m->m_pkthdr.csum_flags); 241 } 242 #endif 243 244 M_CSUM_DATA_IPv6_SET(m->m_pkthdr.csum_data, sizeof(struct ip6_hdr)); 245 246 #define MAKE_EXTHDR(hp, mp) \ 247 do { \ 248 if (hp) { \ 249 struct ip6_ext *eh = (struct ip6_ext *)(hp); \ 250 error = ip6_copyexthdr((mp), (void *)(hp), \ 251 ((eh)->ip6e_len + 1) << 3); \ 252 if (error) \ 253 goto freehdrs; \ 254 } \ 255 } while (/*CONSTCOND*/ 0) 256 257 memset(&exthdrs, 0, sizeof(exthdrs)); 258 if (opt) { 259 /* Hop-by-Hop options header */ 260 MAKE_EXTHDR(opt->ip6po_hbh, &exthdrs.ip6e_hbh); 261 /* Destination options header (1st part) */ 262 MAKE_EXTHDR(opt->ip6po_dest1, &exthdrs.ip6e_dest1); 263 /* Routing header */ 264 MAKE_EXTHDR(opt->ip6po_rthdr, &exthdrs.ip6e_rthdr); 265 /* Destination options header (2nd part) */ 266 MAKE_EXTHDR(opt->ip6po_dest2, &exthdrs.ip6e_dest2); 267 } 268 269 /* 270 * Calculate the total length of the extension header chain. 271 * Keep the length of the unfragmentable part for fragmentation. 272 */ 273 optlen = 0; 274 if (exthdrs.ip6e_hbh) 275 optlen += exthdrs.ip6e_hbh->m_len; 276 if (exthdrs.ip6e_dest1) 277 optlen += exthdrs.ip6e_dest1->m_len; 278 if (exthdrs.ip6e_rthdr) 279 optlen += exthdrs.ip6e_rthdr->m_len; 280 unfragpartlen = optlen + sizeof(struct ip6_hdr); 281 /* NOTE: we don't add AH/ESP length here. do that later. */ 282 if (exthdrs.ip6e_dest2) 283 optlen += exthdrs.ip6e_dest2->m_len; 284 285 #ifdef IPSEC 286 if (ipsec_used) { 287 /* Check the security policy (SP) for the packet */ 288 sp = ipsec6_check_policy(m, in6p, flags, &needipsec, &error); 289 if (error != 0) { 290 /* 291 * Hack: -EINVAL is used to signal that a packet 292 * should be silently discarded. This is typically 293 * because we asked key management for an SA and 294 * it was delayed (e.g. kicked up to IKE). 295 */ 296 if (error == -EINVAL) 297 error = 0; 298 goto freehdrs; 299 } 300 } 301 #endif 302 303 if (needipsec && 304 (m->m_pkthdr.csum_flags & (M_CSUM_UDPv6|M_CSUM_TCPv6)) != 0) { 305 in6_undefer_cksum_tcpudp(m); 306 m->m_pkthdr.csum_flags &= ~(M_CSUM_UDPv6|M_CSUM_TCPv6); 307 } 308 309 /* 310 * If we need IPsec, or there is at least one extension header, 311 * separate IP6 header from the payload. 312 */ 313 if ((needipsec || optlen) && !hdrsplit) { 314 if ((error = ip6_splithdr(m, &exthdrs)) != 0) { 315 m = NULL; 316 goto freehdrs; 317 } 318 m = exthdrs.ip6e_ip6; 319 hdrsplit++; 320 } 321 322 /* adjust pointer */ 323 ip6 = mtod(m, struct ip6_hdr *); 324 325 /* adjust mbuf packet header length */ 326 m->m_pkthdr.len += optlen; 327 plen = m->m_pkthdr.len - sizeof(*ip6); 328 329 /* If this is a jumbo payload, insert a jumbo payload option. */ 330 if (plen > IPV6_MAXPACKET) { 331 if (!hdrsplit) { 332 if ((error = ip6_splithdr(m, &exthdrs)) != 0) { 333 m = NULL; 334 goto freehdrs; 335 } 336 m = exthdrs.ip6e_ip6; 337 hdrsplit++; 338 } 339 /* adjust pointer */ 340 ip6 = mtod(m, struct ip6_hdr *); 341 if ((error = ip6_insert_jumboopt(&exthdrs, plen)) != 0) 342 goto freehdrs; 343 optlen += 8; /* XXX JUMBOOPTLEN */ 344 ip6->ip6_plen = 0; 345 } else 346 ip6->ip6_plen = htons(plen); 347 348 /* 349 * Concatenate headers and fill in next header fields. 350 * Here we have, on "m" 351 * IPv6 payload 352 * and we insert headers accordingly. Finally, we should be getting: 353 * IPv6 hbh dest1 rthdr ah* [esp* dest2 payload] 354 * 355 * during the header composing process, "m" points to IPv6 header. 356 * "mprev" points to an extension header prior to esp. 357 */ 358 { 359 u_char *nexthdrp = &ip6->ip6_nxt; 360 struct mbuf *mprev = m; 361 362 /* 363 * we treat dest2 specially. this makes IPsec processing 364 * much easier. the goal here is to make mprev point the 365 * mbuf prior to dest2. 366 * 367 * result: IPv6 dest2 payload 368 * m and mprev will point to IPv6 header. 369 */ 370 if (exthdrs.ip6e_dest2) { 371 if (!hdrsplit) 372 panic("assumption failed: hdr not split"); 373 exthdrs.ip6e_dest2->m_next = m->m_next; 374 m->m_next = exthdrs.ip6e_dest2; 375 *mtod(exthdrs.ip6e_dest2, u_char *) = ip6->ip6_nxt; 376 ip6->ip6_nxt = IPPROTO_DSTOPTS; 377 } 378 379 #define MAKE_CHAIN(m, mp, p, i)\ 380 do {\ 381 if (m) {\ 382 if (!hdrsplit) \ 383 panic("assumption failed: hdr not split"); \ 384 *mtod((m), u_char *) = *(p);\ 385 *(p) = (i);\ 386 p = mtod((m), u_char *);\ 387 (m)->m_next = (mp)->m_next;\ 388 (mp)->m_next = (m);\ 389 (mp) = (m);\ 390 }\ 391 } while (/*CONSTCOND*/ 0) 392 /* 393 * result: IPv6 hbh dest1 rthdr dest2 payload 394 * m will point to IPv6 header. mprev will point to the 395 * extension header prior to dest2 (rthdr in the above case). 396 */ 397 MAKE_CHAIN(exthdrs.ip6e_hbh, mprev, nexthdrp, IPPROTO_HOPOPTS); 398 MAKE_CHAIN(exthdrs.ip6e_dest1, mprev, nexthdrp, 399 IPPROTO_DSTOPTS); 400 MAKE_CHAIN(exthdrs.ip6e_rthdr, mprev, nexthdrp, 401 IPPROTO_ROUTING); 402 403 M_CSUM_DATA_IPv6_SET(m->m_pkthdr.csum_data, 404 sizeof(struct ip6_hdr) + optlen); 405 } 406 407 /* Need to save for pmtu */ 408 finaldst = ip6->ip6_dst; 409 410 /* 411 * If there is a routing header, replace destination address field 412 * with the first hop of the routing header. 413 */ 414 if (exthdrs.ip6e_rthdr) { 415 struct ip6_rthdr *rh; 416 417 rh = mtod(exthdrs.ip6e_rthdr, struct ip6_rthdr *); 418 419 error = ip6_handle_rthdr(rh, ip6); 420 if (error != 0) 421 goto bad; 422 } 423 424 /* Source address validation */ 425 if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) && 426 (flags & IPV6_UNSPECSRC) == 0) { 427 error = EOPNOTSUPP; 428 IP6_STATINC(IP6_STAT_BADSCOPE); 429 goto bad; 430 } 431 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) { 432 error = EOPNOTSUPP; 433 IP6_STATINC(IP6_STAT_BADSCOPE); 434 goto bad; 435 } 436 437 IP6_STATINC(IP6_STAT_LOCALOUT); 438 439 /* 440 * Route packet. 441 */ 442 /* initialize cached route */ 443 if (ro == NULL) { 444 memset(&ip6route, 0, sizeof(ip6route)); 445 ro = &ip6route; 446 } 447 ro_pmtu = ro; 448 if (opt && opt->ip6po_rthdr) 449 ro = &opt->ip6po_route; 450 451 /* 452 * if specified, try to fill in the traffic class field. 453 * do not override if a non-zero value is already set. 454 * we check the diffserv field and the ecn field separately. 455 */ 456 if (opt && opt->ip6po_tclass >= 0) { 457 int mask = 0; 458 459 if ((ip6->ip6_flow & htonl(0xfc << 20)) == 0) 460 mask |= 0xfc; 461 if ((ip6->ip6_flow & htonl(0x03 << 20)) == 0) 462 mask |= 0x03; 463 if (mask != 0) 464 ip6->ip6_flow |= htonl((opt->ip6po_tclass & mask) << 20); 465 } 466 467 /* fill in or override the hop limit field, if necessary. */ 468 if (opt && opt->ip6po_hlim != -1) 469 ip6->ip6_hlim = opt->ip6po_hlim & 0xff; 470 else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) { 471 if (im6o != NULL) 472 ip6->ip6_hlim = im6o->im6o_multicast_hlim; 473 else 474 ip6->ip6_hlim = ip6_defmcasthlim; 475 } 476 477 #ifdef IPSEC 478 if (needipsec) { 479 int s = splsoftnet(); 480 error = ipsec6_process_packet(m, sp->req); 481 splx(s); 482 483 /* 484 * Preserve KAME behaviour: ENOENT can be returned 485 * when an SA acquire is in progress. Don't propagate 486 * this to user-level; it confuses applications. 487 * XXX this will go away when the SADB is redone. 488 */ 489 if (error == ENOENT) 490 error = 0; 491 492 goto done; 493 } 494 #endif 495 496 /* adjust pointer */ 497 ip6 = mtod(m, struct ip6_hdr *); 498 499 sockaddr_in6_init(&dst_sa, &ip6->ip6_dst, 0, 0, 0); 500 501 /* We do not need a route for multicast */ 502 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) { 503 struct in6_pktinfo *pi = NULL; 504 505 /* 506 * If the outgoing interface for the address is specified by 507 * the caller, use it. 508 */ 509 if (opt && (pi = opt->ip6po_pktinfo) != NULL) { 510 /* XXX boundary check is assumed to be already done. */ 511 ifp = if_get_byindex(pi->ipi6_ifindex, &psref); 512 } else if (im6o != NULL) { 513 ifp = if_get_byindex(im6o->im6o_multicast_if_index, 514 &psref); 515 } 516 } 517 518 if (ifp == NULL) { 519 error = in6_selectroute(&dst_sa, opt, &ro, &rt, true); 520 if (error != 0) 521 goto bad; 522 ifp = if_get_byindex(rt->rt_ifp->if_index, &psref); 523 } 524 525 if (rt == NULL) { 526 /* 527 * If in6_selectroute() does not return a route entry, 528 * dst may not have been updated. 529 */ 530 error = rtcache_setdst(ro, sin6tosa(&dst_sa)); 531 if (error) { 532 goto bad; 533 } 534 } 535 536 /* 537 * then rt (for unicast) and ifp must be non-NULL valid values. 538 */ 539 if ((flags & IPV6_FORWARDING) == 0) { 540 /* XXX: the FORWARDING flag can be set for mrouting. */ 541 in6_ifstat_inc(ifp, ifs6_out_request); 542 } 543 if (rt != NULL) { 544 ia = (struct in6_ifaddr *)(rt->rt_ifa); 545 rt->rt_use++; 546 } 547 548 /* 549 * The outgoing interface must be in the zone of source and 550 * destination addresses. We should use ia_ifp to support the 551 * case of sending packets to an address of our own. 552 */ 553 if (ia != NULL && ia->ia_ifp) { 554 origifp = ia->ia_ifp; 555 if (if_is_deactivated(origifp)) 556 goto bad; 557 if_acquire(origifp, &psref_ia); 558 release_psref_ia = true; 559 } else 560 origifp = ifp; 561 562 src0 = ip6->ip6_src; 563 if (in6_setscope(&src0, origifp, &zone)) 564 goto badscope; 565 sockaddr_in6_init(&src_sa, &ip6->ip6_src, 0, 0, 0); 566 if (sa6_recoverscope(&src_sa) || zone != src_sa.sin6_scope_id) 567 goto badscope; 568 569 dst0 = ip6->ip6_dst; 570 if (in6_setscope(&dst0, origifp, &zone)) 571 goto badscope; 572 /* re-initialize to be sure */ 573 sockaddr_in6_init(&dst_sa, &ip6->ip6_dst, 0, 0, 0); 574 if (sa6_recoverscope(&dst_sa) || zone != dst_sa.sin6_scope_id) 575 goto badscope; 576 577 /* scope check is done. */ 578 579 /* Ensure we only send from a valid address. */ 580 if ((ifp->if_flags & IFF_LOOPBACK) == 0 && 581 (error = ip6_ifaddrvalid(&src0, &dst0)) != 0) 582 { 583 char ip6buf[INET6_ADDRSTRLEN]; 584 nd6log(LOG_ERR, 585 "refusing to send from invalid address %s (pid %d)\n", 586 IN6_PRINT(ip6buf, &src0), curproc->p_pid); 587 IP6_STATINC(IP6_STAT_ODROPPED); 588 in6_ifstat_inc(origifp, ifs6_out_discard); 589 if (error == 1) 590 /* 591 * Address exists, but is tentative or detached. 592 * We can't send from it because it's invalid, 593 * so we drop the packet. 594 */ 595 error = 0; 596 else 597 error = EADDRNOTAVAIL; 598 goto bad; 599 } 600 601 if (rt != NULL && (rt->rt_flags & RTF_GATEWAY) && 602 !IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) 603 dst = satocsin6(rt->rt_gateway); 604 else 605 dst = satocsin6(rtcache_getdst(ro)); 606 607 /* 608 * XXXXXX: original code follows: 609 */ 610 if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) 611 m->m_flags &= ~(M_BCAST | M_MCAST); /* just in case */ 612 else { 613 bool ingroup; 614 615 m->m_flags = (m->m_flags & ~M_BCAST) | M_MCAST; 616 617 in6_ifstat_inc(ifp, ifs6_out_mcast); 618 619 /* 620 * Confirm that the outgoing interface supports multicast. 621 */ 622 if (!(ifp->if_flags & IFF_MULTICAST)) { 623 IP6_STATINC(IP6_STAT_NOROUTE); 624 in6_ifstat_inc(ifp, ifs6_out_discard); 625 error = ENETUNREACH; 626 goto bad; 627 } 628 629 ingroup = in6_multi_group(&ip6->ip6_dst, ifp); 630 if (ingroup && (im6o == NULL || im6o->im6o_multicast_loop)) { 631 /* 632 * If we belong to the destination multicast group 633 * on the outgoing interface, and the caller did not 634 * forbid loopback, loop back a copy. 635 */ 636 KASSERT(dst != NULL); 637 ip6_mloopback(ifp, m, dst); 638 } else { 639 /* 640 * If we are acting as a multicast router, perform 641 * multicast forwarding as if the packet had just 642 * arrived on the interface to which we are about 643 * to send. The multicast forwarding function 644 * recursively calls this function, using the 645 * IPV6_FORWARDING flag to prevent infinite recursion. 646 * 647 * Multicasts that are looped back by ip6_mloopback(), 648 * above, will be forwarded by the ip6_input() routine, 649 * if necessary. 650 */ 651 if (ip6_mrouter && (flags & IPV6_FORWARDING) == 0) { 652 if (ip6_mforward(ip6, ifp, m) != 0) { 653 m_freem(m); 654 goto done; 655 } 656 } 657 } 658 /* 659 * Multicasts with a hoplimit of zero may be looped back, 660 * above, but must not be transmitted on a network. 661 * Also, multicasts addressed to the loopback interface 662 * are not sent -- the above call to ip6_mloopback() will 663 * loop back a copy if this host actually belongs to the 664 * destination group on the loopback interface. 665 */ 666 if (ip6->ip6_hlim == 0 || (ifp->if_flags & IFF_LOOPBACK) || 667 IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst)) { 668 m_freem(m); 669 goto done; 670 } 671 } 672 673 /* 674 * Fill the outgoing inteface to tell the upper layer 675 * to increment per-interface statistics. 676 */ 677 if (ifpp) 678 *ifpp = ifp; 679 680 /* Determine path MTU. */ 681 /* 682 * ro_pmtu represent final destination while 683 * ro might represent immediate destination. 684 * Use ro_pmtu destination since MTU might differ. 685 */ 686 if (ro_pmtu != ro) { 687 union { 688 struct sockaddr dst; 689 struct sockaddr_in6 dst6; 690 } u; 691 692 /* ro_pmtu may not have a cache */ 693 sockaddr_in6_init(&u.dst6, &finaldst, 0, 0, 0); 694 rt_pmtu = rtcache_lookup(ro_pmtu, &u.dst); 695 } else 696 rt_pmtu = rt; 697 error = ip6_getpmtu(rt_pmtu, ifp, &mtu, &alwaysfrag); 698 if (rt_pmtu != NULL && rt_pmtu != rt) 699 rtcache_unref(rt_pmtu, ro_pmtu); 700 if (error != 0) 701 goto bad; 702 703 /* 704 * The caller of this function may specify to use the minimum MTU 705 * in some cases. 706 * An advanced API option (IPV6_USE_MIN_MTU) can also override MTU 707 * setting. The logic is a bit complicated; by default, unicast 708 * packets will follow path MTU while multicast packets will be sent at 709 * the minimum MTU. If IP6PO_MINMTU_ALL is specified, all packets 710 * including unicast ones will be sent at the minimum MTU. Multicast 711 * packets will always be sent at the minimum MTU unless 712 * IP6PO_MINMTU_DISABLE is explicitly specified. 713 * See RFC 3542 for more details. 714 */ 715 if (mtu > IPV6_MMTU) { 716 if ((flags & IPV6_MINMTU)) 717 mtu = IPV6_MMTU; 718 else if (opt && opt->ip6po_minmtu == IP6PO_MINMTU_ALL) 719 mtu = IPV6_MMTU; 720 else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) && 721 (opt == NULL || 722 opt->ip6po_minmtu != IP6PO_MINMTU_DISABLE)) { 723 mtu = IPV6_MMTU; 724 } 725 } 726 727 /* 728 * clear embedded scope identifiers if necessary. 729 * in6_clearscope will touch the addresses only when necessary. 730 */ 731 in6_clearscope(&ip6->ip6_src); 732 in6_clearscope(&ip6->ip6_dst); 733 734 /* 735 * If the outgoing packet contains a hop-by-hop options header, 736 * it must be examined and processed even by the source node. 737 * (RFC 2460, section 4.) 738 * 739 * XXX Is this really necessary? 740 */ 741 if (ip6->ip6_nxt == IPPROTO_HOPOPTS) { 742 u_int32_t dummy1; /* XXX unused */ 743 u_int32_t dummy2; /* XXX unused */ 744 int hoff = sizeof(struct ip6_hdr); 745 746 if (ip6_hopopts_input(&dummy1, &dummy2, &m, &hoff)) { 747 /* m was already freed at this point */ 748 error = EINVAL; 749 goto done; 750 } 751 752 ip6 = mtod(m, struct ip6_hdr *); 753 } 754 755 /* 756 * Run through list of hooks for output packets. 757 */ 758 if ((error = pfil_run_hooks(inet6_pfil_hook, &m, ifp, PFIL_OUT)) != 0) 759 goto done; 760 if (m == NULL) 761 goto done; 762 ip6 = mtod(m, struct ip6_hdr *); 763 764 /* 765 * Send the packet to the outgoing interface. 766 * If necessary, do IPv6 fragmentation before sending. 767 * 768 * the logic here is rather complex: 769 * 1: normal case (dontfrag == 0, alwaysfrag == 0) 770 * 1-a: send as is if tlen <= path mtu 771 * 1-b: fragment if tlen > path mtu 772 * 773 * 2: if user asks us not to fragment (dontfrag == 1) 774 * 2-a: send as is if tlen <= interface mtu 775 * 2-b: error if tlen > interface mtu 776 * 777 * 3: if we always need to attach fragment header (alwaysfrag == 1) 778 * always fragment 779 * 780 * 4: if dontfrag == 1 && alwaysfrag == 1 781 * error, as we cannot handle this conflicting request 782 */ 783 tlen = m->m_pkthdr.len; 784 tso = (m->m_pkthdr.csum_flags & M_CSUM_TSOv6) != 0; 785 if (opt && (opt->ip6po_flags & IP6PO_DONTFRAG)) 786 dontfrag = 1; 787 else 788 dontfrag = 0; 789 790 if (dontfrag && alwaysfrag) { /* case 4 */ 791 /* conflicting request - can't transmit */ 792 error = EMSGSIZE; 793 goto bad; 794 } 795 if (dontfrag && (!tso && tlen > IN6_LINKMTU(ifp))) { /* case 2-b */ 796 /* 797 * Even if the DONTFRAG option is specified, we cannot send the 798 * packet when the data length is larger than the MTU of the 799 * outgoing interface. 800 * Notify the error by sending IPV6_PATHMTU ancillary data as 801 * well as returning an error code (the latter is not described 802 * in the API spec.) 803 */ 804 u_int32_t mtu32; 805 struct ip6ctlparam ip6cp; 806 807 mtu32 = (u_int32_t)mtu; 808 memset(&ip6cp, 0, sizeof(ip6cp)); 809 ip6cp.ip6c_cmdarg = (void *)&mtu32; 810 pfctlinput2(PRC_MSGSIZE, 811 rtcache_getdst(ro_pmtu), &ip6cp); 812 813 error = EMSGSIZE; 814 goto bad; 815 } 816 817 /* 818 * transmit packet without fragmentation 819 */ 820 if (dontfrag || (!alwaysfrag && (tlen <= mtu || tso))) { 821 /* case 1-a and 2-a */ 822 struct in6_ifaddr *ia6; 823 int sw_csum; 824 int s; 825 826 ip6 = mtod(m, struct ip6_hdr *); 827 s = pserialize_read_enter(); 828 ia6 = in6_ifawithifp(ifp, &ip6->ip6_src); 829 if (ia6) { 830 /* Record statistics for this interface address. */ 831 ia6->ia_ifa.ifa_data.ifad_outbytes += m->m_pkthdr.len; 832 } 833 pserialize_read_exit(s); 834 835 sw_csum = m->m_pkthdr.csum_flags & ~ifp->if_csum_flags_tx; 836 if ((sw_csum & (M_CSUM_UDPv6|M_CSUM_TCPv6)) != 0) { 837 if (IN6_NEED_CHECKSUM(ifp, 838 sw_csum & (M_CSUM_UDPv6|M_CSUM_TCPv6))) { 839 in6_undefer_cksum_tcpudp(m); 840 } 841 m->m_pkthdr.csum_flags &= ~(M_CSUM_UDPv6|M_CSUM_TCPv6); 842 } 843 844 KASSERT(dst != NULL); 845 if (__predict_true(!tso || 846 (ifp->if_capenable & IFCAP_TSOv6) != 0)) { 847 error = ip6_if_output(ifp, origifp, m, dst, rt); 848 } else { 849 error = ip6_tso_output(ifp, origifp, m, dst, rt); 850 } 851 goto done; 852 } 853 854 if (tso) { 855 error = EINVAL; /* XXX */ 856 goto bad; 857 } 858 859 /* 860 * try to fragment the packet. case 1-b and 3 861 */ 862 if (mtu < IPV6_MMTU) { 863 /* path MTU cannot be less than IPV6_MMTU */ 864 error = EMSGSIZE; 865 in6_ifstat_inc(ifp, ifs6_out_fragfail); 866 goto bad; 867 } else if (ip6->ip6_plen == 0) { 868 /* jumbo payload cannot be fragmented */ 869 error = EMSGSIZE; 870 in6_ifstat_inc(ifp, ifs6_out_fragfail); 871 goto bad; 872 } else { 873 const u_int32_t id = htonl(ip6_randomid()); 874 struct mbuf **mnext, *m_frgpart; 875 const int hlen = unfragpartlen; 876 struct ip6_frag *ip6f; 877 u_char nextproto; 878 879 if (mtu > IPV6_MAXPACKET) 880 mtu = IPV6_MAXPACKET; 881 882 /* 883 * Must be able to put at least 8 bytes per fragment. 884 */ 885 len = (mtu - hlen - sizeof(struct ip6_frag)) & ~7; 886 if (len < 8) { 887 error = EMSGSIZE; 888 in6_ifstat_inc(ifp, ifs6_out_fragfail); 889 goto bad; 890 } 891 892 mnext = &m->m_nextpkt; 893 894 /* 895 * Change the next header field of the last header in the 896 * unfragmentable part. 897 */ 898 if (exthdrs.ip6e_rthdr) { 899 nextproto = *mtod(exthdrs.ip6e_rthdr, u_char *); 900 *mtod(exthdrs.ip6e_rthdr, u_char *) = IPPROTO_FRAGMENT; 901 } else if (exthdrs.ip6e_dest1) { 902 nextproto = *mtod(exthdrs.ip6e_dest1, u_char *); 903 *mtod(exthdrs.ip6e_dest1, u_char *) = IPPROTO_FRAGMENT; 904 } else if (exthdrs.ip6e_hbh) { 905 nextproto = *mtod(exthdrs.ip6e_hbh, u_char *); 906 *mtod(exthdrs.ip6e_hbh, u_char *) = IPPROTO_FRAGMENT; 907 } else { 908 nextproto = ip6->ip6_nxt; 909 ip6->ip6_nxt = IPPROTO_FRAGMENT; 910 } 911 912 if ((m->m_pkthdr.csum_flags & (M_CSUM_UDPv6|M_CSUM_TCPv6)) 913 != 0) { 914 if (IN6_NEED_CHECKSUM(ifp, 915 m->m_pkthdr.csum_flags & 916 (M_CSUM_UDPv6|M_CSUM_TCPv6))) { 917 in6_undefer_cksum_tcpudp(m); 918 } 919 m->m_pkthdr.csum_flags &= ~(M_CSUM_UDPv6|M_CSUM_TCPv6); 920 } 921 922 /* 923 * Loop through length of segment after first fragment, 924 * make new header and copy data of each part and link onto 925 * chain. 926 */ 927 m0 = m; 928 for (off = hlen; off < tlen; off += len) { 929 struct mbuf *mlast; 930 931 MGETHDR(m, M_DONTWAIT, MT_HEADER); 932 if (!m) { 933 error = ENOBUFS; 934 IP6_STATINC(IP6_STAT_ODROPPED); 935 goto sendorfree; 936 } 937 m_reset_rcvif(m); 938 m->m_flags = m0->m_flags & M_COPYFLAGS; 939 *mnext = m; 940 mnext = &m->m_nextpkt; 941 m->m_data += max_linkhdr; 942 mhip6 = mtod(m, struct ip6_hdr *); 943 *mhip6 = *ip6; 944 m->m_len = sizeof(*mhip6); 945 946 ip6f = NULL; 947 error = ip6_insertfraghdr(m0, m, hlen, &ip6f); 948 if (error) { 949 IP6_STATINC(IP6_STAT_ODROPPED); 950 goto sendorfree; 951 } 952 953 /* Fill in the Frag6 Header */ 954 ip6f->ip6f_offlg = htons((u_int16_t)((off - hlen) & ~7)); 955 if (off + len >= tlen) 956 len = tlen - off; 957 else 958 ip6f->ip6f_offlg |= IP6F_MORE_FRAG; 959 ip6f->ip6f_reserved = 0; 960 ip6f->ip6f_ident = id; 961 ip6f->ip6f_nxt = nextproto; 962 963 mhip6->ip6_plen = htons((u_int16_t)(len + hlen + 964 sizeof(*ip6f) - sizeof(struct ip6_hdr))); 965 if ((m_frgpart = m_copym(m0, off, len, M_DONTWAIT)) == NULL) { 966 error = ENOBUFS; 967 IP6_STATINC(IP6_STAT_ODROPPED); 968 goto sendorfree; 969 } 970 for (mlast = m; mlast->m_next; mlast = mlast->m_next) 971 ; 972 mlast->m_next = m_frgpart; 973 974 m->m_pkthdr.len = len + hlen + sizeof(*ip6f); 975 m_reset_rcvif(m); 976 IP6_STATINC(IP6_STAT_OFRAGMENTS); 977 in6_ifstat_inc(ifp, ifs6_out_fragcreat); 978 } 979 980 in6_ifstat_inc(ifp, ifs6_out_fragok); 981 } 982 983 sendorfree: 984 m = m0->m_nextpkt; 985 m0->m_nextpkt = 0; 986 m_freem(m0); 987 for (m0 = m; m; m = m0) { 988 m0 = m->m_nextpkt; 989 m->m_nextpkt = 0; 990 if (error == 0) { 991 struct in6_ifaddr *ia6; 992 int s; 993 ip6 = mtod(m, struct ip6_hdr *); 994 s = pserialize_read_enter(); 995 ia6 = in6_ifawithifp(ifp, &ip6->ip6_src); 996 if (ia6) { 997 /* 998 * Record statistics for this interface 999 * address. 1000 */ 1001 ia6->ia_ifa.ifa_data.ifad_outbytes += 1002 m->m_pkthdr.len; 1003 } 1004 pserialize_read_exit(s); 1005 KASSERT(dst != NULL); 1006 error = ip6_if_output(ifp, origifp, m, dst, rt); 1007 } else 1008 m_freem(m); 1009 } 1010 1011 if (error == 0) 1012 IP6_STATINC(IP6_STAT_FRAGMENTED); 1013 1014 done: 1015 rtcache_unref(rt, ro); 1016 if (ro == &ip6route) 1017 rtcache_free(&ip6route); 1018 #ifdef IPSEC 1019 if (sp != NULL) 1020 KEY_SP_UNREF(&sp); 1021 #endif 1022 if_put(ifp, &psref); 1023 if (release_psref_ia) 1024 if_put(origifp, &psref_ia); 1025 curlwp_bindx(bound); 1026 1027 return error; 1028 1029 freehdrs: 1030 m_freem(exthdrs.ip6e_hbh); 1031 m_freem(exthdrs.ip6e_dest1); 1032 m_freem(exthdrs.ip6e_rthdr); 1033 m_freem(exthdrs.ip6e_dest2); 1034 /* FALLTHROUGH */ 1035 bad: 1036 m_freem(m); 1037 goto done; 1038 1039 badscope: 1040 IP6_STATINC(IP6_STAT_BADSCOPE); 1041 in6_ifstat_inc(origifp, ifs6_out_discard); 1042 if (error == 0) 1043 error = EHOSTUNREACH; /* XXX */ 1044 goto bad; 1045 } 1046 1047 static int 1048 ip6_copyexthdr(struct mbuf **mp, void *hdr, int hlen) 1049 { 1050 struct mbuf *m; 1051 1052 if (hlen > MCLBYTES) 1053 return ENOBUFS; /* XXX */ 1054 1055 MGET(m, M_DONTWAIT, MT_DATA); 1056 if (!m) 1057 return ENOBUFS; 1058 1059 if (hlen > MLEN) { 1060 MCLGET(m, M_DONTWAIT); 1061 if ((m->m_flags & M_EXT) == 0) { 1062 m_free(m); 1063 return ENOBUFS; 1064 } 1065 } 1066 m->m_len = hlen; 1067 if (hdr) 1068 memcpy(mtod(m, void *), hdr, hlen); 1069 1070 *mp = m; 1071 return 0; 1072 } 1073 1074 /* 1075 * Insert jumbo payload option. 1076 */ 1077 static int 1078 ip6_insert_jumboopt(struct ip6_exthdrs *exthdrs, u_int32_t plen) 1079 { 1080 struct mbuf *mopt; 1081 u_int8_t *optbuf; 1082 u_int32_t v; 1083 1084 #define JUMBOOPTLEN 8 /* length of jumbo payload option and padding */ 1085 1086 /* 1087 * If there is no hop-by-hop options header, allocate new one. 1088 * If there is one but it doesn't have enough space to store the 1089 * jumbo payload option, allocate a cluster to store the whole options. 1090 * Otherwise, use it to store the options. 1091 */ 1092 if (exthdrs->ip6e_hbh == NULL) { 1093 MGET(mopt, M_DONTWAIT, MT_DATA); 1094 if (mopt == 0) 1095 return (ENOBUFS); 1096 mopt->m_len = JUMBOOPTLEN; 1097 optbuf = mtod(mopt, u_int8_t *); 1098 optbuf[1] = 0; /* = ((JUMBOOPTLEN) >> 3) - 1 */ 1099 exthdrs->ip6e_hbh = mopt; 1100 } else { 1101 struct ip6_hbh *hbh; 1102 1103 mopt = exthdrs->ip6e_hbh; 1104 if (M_TRAILINGSPACE(mopt) < JUMBOOPTLEN) { 1105 const int oldoptlen = mopt->m_len; 1106 struct mbuf *n; 1107 1108 /* 1109 * Assumptions: 1110 * - exthdrs->ip6e_hbh is not referenced from places 1111 * other than exthdrs. 1112 * - exthdrs->ip6e_hbh is not an mbuf chain. 1113 */ 1114 KASSERT(mopt->m_next == NULL); 1115 1116 /* 1117 * Give up if the whole (new) hbh header does not fit 1118 * even in an mbuf cluster. 1119 */ 1120 if (oldoptlen + JUMBOOPTLEN > MCLBYTES) 1121 return ENOBUFS; 1122 1123 /* 1124 * At this point, we must always prepare a cluster. 1125 */ 1126 MGET(n, M_DONTWAIT, MT_DATA); 1127 if (n) { 1128 MCLGET(n, M_DONTWAIT); 1129 if ((n->m_flags & M_EXT) == 0) { 1130 m_freem(n); 1131 n = NULL; 1132 } 1133 } 1134 if (!n) 1135 return ENOBUFS; 1136 1137 n->m_len = oldoptlen + JUMBOOPTLEN; 1138 bcopy(mtod(mopt, void *), mtod(n, void *), 1139 oldoptlen); 1140 optbuf = mtod(n, u_int8_t *) + oldoptlen; 1141 m_freem(mopt); 1142 mopt = exthdrs->ip6e_hbh = n; 1143 } else { 1144 optbuf = mtod(mopt, u_int8_t *) + mopt->m_len; 1145 mopt->m_len += JUMBOOPTLEN; 1146 } 1147 optbuf[0] = IP6OPT_PADN; 1148 optbuf[1] = 0; 1149 1150 /* 1151 * Adjust the header length according to the pad and 1152 * the jumbo payload option. 1153 */ 1154 hbh = mtod(mopt, struct ip6_hbh *); 1155 hbh->ip6h_len += (JUMBOOPTLEN >> 3); 1156 } 1157 1158 /* fill in the option. */ 1159 optbuf[2] = IP6OPT_JUMBO; 1160 optbuf[3] = 4; 1161 v = (u_int32_t)htonl(plen + JUMBOOPTLEN); 1162 memcpy(&optbuf[4], &v, sizeof(u_int32_t)); 1163 1164 /* finally, adjust the packet header length */ 1165 exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN; 1166 1167 return 0; 1168 #undef JUMBOOPTLEN 1169 } 1170 1171 /* 1172 * Insert fragment header and copy unfragmentable header portions. 1173 * 1174 * *frghdrp will not be read, and it is guaranteed that either an 1175 * error is returned or that *frghdrp will point to space allocated 1176 * for the fragment header. 1177 * 1178 * On entry, m contains: 1179 * IPv6 Header 1180 * On exit, it contains: 1181 * IPv6 Header -> Unfragmentable Part -> Frag6 Header 1182 */ 1183 static int 1184 ip6_insertfraghdr(struct mbuf *m0, struct mbuf *m, int hlen, 1185 struct ip6_frag **frghdrp) 1186 { 1187 struct mbuf *n, *mlast; 1188 1189 if (hlen > sizeof(struct ip6_hdr)) { 1190 n = m_copym(m0, sizeof(struct ip6_hdr), 1191 hlen - sizeof(struct ip6_hdr), M_DONTWAIT); 1192 if (n == NULL) 1193 return ENOBUFS; 1194 m->m_next = n; 1195 } else 1196 n = m; 1197 1198 /* Search for the last mbuf of unfragmentable part. */ 1199 for (mlast = n; mlast->m_next; mlast = mlast->m_next) 1200 ; 1201 1202 if ((mlast->m_flags & M_EXT) == 0 && 1203 M_TRAILINGSPACE(mlast) >= sizeof(struct ip6_frag)) { 1204 /* use the trailing space of the last mbuf for the fragment hdr */ 1205 *frghdrp = (struct ip6_frag *)(mtod(mlast, char *) + 1206 mlast->m_len); 1207 mlast->m_len += sizeof(struct ip6_frag); 1208 } else { 1209 /* allocate a new mbuf for the fragment header */ 1210 struct mbuf *mfrg; 1211 1212 MGET(mfrg, M_DONTWAIT, MT_DATA); 1213 if (mfrg == NULL) 1214 return ENOBUFS; 1215 mfrg->m_len = sizeof(struct ip6_frag); 1216 *frghdrp = mtod(mfrg, struct ip6_frag *); 1217 mlast->m_next = mfrg; 1218 } 1219 1220 return 0; 1221 } 1222 1223 static int 1224 ip6_getpmtu(struct rtentry *rt, struct ifnet *ifp, u_long *mtup, 1225 int *alwaysfragp) 1226 { 1227 u_int32_t mtu = 0; 1228 int alwaysfrag = 0; 1229 int error = 0; 1230 1231 if (rt != NULL) { 1232 u_int32_t ifmtu; 1233 1234 if (ifp == NULL) 1235 ifp = rt->rt_ifp; 1236 ifmtu = IN6_LINKMTU(ifp); 1237 mtu = rt->rt_rmx.rmx_mtu; 1238 if (mtu == 0) 1239 mtu = ifmtu; 1240 else if (mtu < IPV6_MMTU) { 1241 /* 1242 * RFC2460 section 5, last paragraph: 1243 * if we record ICMPv6 too big message with 1244 * mtu < IPV6_MMTU, transmit packets sized IPV6_MMTU 1245 * or smaller, with fragment header attached. 1246 * (fragment header is needed regardless from the 1247 * packet size, for translators to identify packets) 1248 */ 1249 alwaysfrag = 1; 1250 mtu = IPV6_MMTU; 1251 } else if (mtu > ifmtu) { 1252 /* 1253 * The MTU on the route is larger than the MTU on 1254 * the interface! This shouldn't happen, unless the 1255 * MTU of the interface has been changed after the 1256 * interface was brought up. Change the MTU in the 1257 * route to match the interface MTU (as long as the 1258 * field isn't locked). 1259 */ 1260 mtu = ifmtu; 1261 if (!(rt->rt_rmx.rmx_locks & RTV_MTU)) 1262 rt->rt_rmx.rmx_mtu = mtu; 1263 } 1264 } else if (ifp) { 1265 mtu = IN6_LINKMTU(ifp); 1266 } else 1267 error = EHOSTUNREACH; /* XXX */ 1268 1269 *mtup = mtu; 1270 if (alwaysfragp) 1271 *alwaysfragp = alwaysfrag; 1272 return (error); 1273 } 1274 1275 /* 1276 * IP6 socket option processing. 1277 */ 1278 int 1279 ip6_ctloutput(int op, struct socket *so, struct sockopt *sopt) 1280 { 1281 int optdatalen, uproto; 1282 void *optdata; 1283 struct in6pcb *in6p = sotoin6pcb(so); 1284 struct ip_moptions **mopts; 1285 int error, optval; 1286 int level, optname; 1287 1288 KASSERT(solocked(so)); 1289 KASSERT(sopt != NULL); 1290 1291 level = sopt->sopt_level; 1292 optname = sopt->sopt_name; 1293 1294 error = optval = 0; 1295 uproto = (int)so->so_proto->pr_protocol; 1296 1297 switch (level) { 1298 case IPPROTO_IP: 1299 switch (optname) { 1300 case IP_ADD_MEMBERSHIP: 1301 case IP_DROP_MEMBERSHIP: 1302 case IP_MULTICAST_IF: 1303 case IP_MULTICAST_LOOP: 1304 case IP_MULTICAST_TTL: 1305 mopts = &in6p->in6p_v4moptions; 1306 switch (op) { 1307 case PRCO_GETOPT: 1308 return ip_getmoptions(*mopts, sopt); 1309 case PRCO_SETOPT: 1310 return ip_setmoptions(mopts, sopt); 1311 default: 1312 return EINVAL; 1313 } 1314 default: 1315 return ENOPROTOOPT; 1316 } 1317 case IPPROTO_IPV6: 1318 break; 1319 default: 1320 return ENOPROTOOPT; 1321 } 1322 switch (op) { 1323 case PRCO_SETOPT: 1324 switch (optname) { 1325 #ifdef RFC2292 1326 case IPV6_2292PKTOPTIONS: 1327 error = ip6_pcbopts(&in6p->in6p_outputopts, so, sopt); 1328 break; 1329 #endif 1330 1331 /* 1332 * Use of some Hop-by-Hop options or some 1333 * Destination options, might require special 1334 * privilege. That is, normal applications 1335 * (without special privilege) might be forbidden 1336 * from setting certain options in outgoing packets, 1337 * and might never see certain options in received 1338 * packets. [RFC 2292 Section 6] 1339 * KAME specific note: 1340 * KAME prevents non-privileged users from sending or 1341 * receiving ANY hbh/dst options in order to avoid 1342 * overhead of parsing options in the kernel. 1343 */ 1344 case IPV6_RECVHOPOPTS: 1345 case IPV6_RECVDSTOPTS: 1346 case IPV6_RECVRTHDRDSTOPTS: 1347 error = kauth_authorize_network(kauth_cred_get(), 1348 KAUTH_NETWORK_IPV6, KAUTH_REQ_NETWORK_IPV6_HOPBYHOP, 1349 NULL, NULL, NULL); 1350 if (error) 1351 break; 1352 /* FALLTHROUGH */ 1353 case IPV6_UNICAST_HOPS: 1354 case IPV6_HOPLIMIT: 1355 case IPV6_FAITH: 1356 1357 case IPV6_RECVPKTINFO: 1358 case IPV6_RECVHOPLIMIT: 1359 case IPV6_RECVRTHDR: 1360 case IPV6_RECVPATHMTU: 1361 case IPV6_RECVTCLASS: 1362 case IPV6_V6ONLY: 1363 error = sockopt_getint(sopt, &optval); 1364 if (error) 1365 break; 1366 switch (optname) { 1367 case IPV6_UNICAST_HOPS: 1368 if (optval < -1 || optval >= 256) 1369 error = EINVAL; 1370 else { 1371 /* -1 = kernel default */ 1372 in6p->in6p_hops = optval; 1373 } 1374 break; 1375 #define OPTSET(bit) \ 1376 do { \ 1377 if (optval) \ 1378 in6p->in6p_flags |= (bit); \ 1379 else \ 1380 in6p->in6p_flags &= ~(bit); \ 1381 } while (/*CONSTCOND*/ 0) 1382 1383 #ifdef RFC2292 1384 #define OPTSET2292(bit) \ 1385 do { \ 1386 in6p->in6p_flags |= IN6P_RFC2292; \ 1387 if (optval) \ 1388 in6p->in6p_flags |= (bit); \ 1389 else \ 1390 in6p->in6p_flags &= ~(bit); \ 1391 } while (/*CONSTCOND*/ 0) 1392 #endif 1393 1394 #define OPTBIT(bit) (in6p->in6p_flags & (bit) ? 1 : 0) 1395 1396 case IPV6_RECVPKTINFO: 1397 #ifdef RFC2292 1398 /* cannot mix with RFC2292 */ 1399 if (OPTBIT(IN6P_RFC2292)) { 1400 error = EINVAL; 1401 break; 1402 } 1403 #endif 1404 OPTSET(IN6P_PKTINFO); 1405 break; 1406 1407 case IPV6_HOPLIMIT: 1408 { 1409 struct ip6_pktopts **optp; 1410 1411 #ifdef RFC2292 1412 /* cannot mix with RFC2292 */ 1413 if (OPTBIT(IN6P_RFC2292)) { 1414 error = EINVAL; 1415 break; 1416 } 1417 #endif 1418 optp = &in6p->in6p_outputopts; 1419 error = ip6_pcbopt(IPV6_HOPLIMIT, 1420 (u_char *)&optval, 1421 sizeof(optval), 1422 optp, 1423 kauth_cred_get(), uproto); 1424 break; 1425 } 1426 1427 case IPV6_RECVHOPLIMIT: 1428 #ifdef RFC2292 1429 /* cannot mix with RFC2292 */ 1430 if (OPTBIT(IN6P_RFC2292)) { 1431 error = EINVAL; 1432 break; 1433 } 1434 #endif 1435 OPTSET(IN6P_HOPLIMIT); 1436 break; 1437 1438 case IPV6_RECVHOPOPTS: 1439 #ifdef RFC2292 1440 /* cannot mix with RFC2292 */ 1441 if (OPTBIT(IN6P_RFC2292)) { 1442 error = EINVAL; 1443 break; 1444 } 1445 #endif 1446 OPTSET(IN6P_HOPOPTS); 1447 break; 1448 1449 case IPV6_RECVDSTOPTS: 1450 #ifdef RFC2292 1451 /* cannot mix with RFC2292 */ 1452 if (OPTBIT(IN6P_RFC2292)) { 1453 error = EINVAL; 1454 break; 1455 } 1456 #endif 1457 OPTSET(IN6P_DSTOPTS); 1458 break; 1459 1460 case IPV6_RECVRTHDRDSTOPTS: 1461 #ifdef RFC2292 1462 /* cannot mix with RFC2292 */ 1463 if (OPTBIT(IN6P_RFC2292)) { 1464 error = EINVAL; 1465 break; 1466 } 1467 #endif 1468 OPTSET(IN6P_RTHDRDSTOPTS); 1469 break; 1470 1471 case IPV6_RECVRTHDR: 1472 #ifdef RFC2292 1473 /* cannot mix with RFC2292 */ 1474 if (OPTBIT(IN6P_RFC2292)) { 1475 error = EINVAL; 1476 break; 1477 } 1478 #endif 1479 OPTSET(IN6P_RTHDR); 1480 break; 1481 1482 case IPV6_FAITH: 1483 OPTSET(IN6P_FAITH); 1484 break; 1485 1486 case IPV6_RECVPATHMTU: 1487 /* 1488 * We ignore this option for TCP 1489 * sockets. 1490 * (RFC3542 leaves this case 1491 * unspecified.) 1492 */ 1493 if (uproto != IPPROTO_TCP) 1494 OPTSET(IN6P_MTU); 1495 break; 1496 1497 case IPV6_V6ONLY: 1498 /* 1499 * make setsockopt(IPV6_V6ONLY) 1500 * available only prior to bind(2). 1501 * see ipng mailing list, Jun 22 2001. 1502 */ 1503 if (in6p->in6p_lport || 1504 !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr)) { 1505 error = EINVAL; 1506 break; 1507 } 1508 #ifdef INET6_BINDV6ONLY 1509 if (!optval) 1510 error = EINVAL; 1511 #else 1512 OPTSET(IN6P_IPV6_V6ONLY); 1513 #endif 1514 break; 1515 case IPV6_RECVTCLASS: 1516 #ifdef RFC2292 1517 /* cannot mix with RFC2292 XXX */ 1518 if (OPTBIT(IN6P_RFC2292)) { 1519 error = EINVAL; 1520 break; 1521 } 1522 #endif 1523 OPTSET(IN6P_TCLASS); 1524 break; 1525 1526 } 1527 break; 1528 1529 case IPV6_OTCLASS: 1530 { 1531 struct ip6_pktopts **optp; 1532 u_int8_t tclass; 1533 1534 error = sockopt_get(sopt, &tclass, sizeof(tclass)); 1535 if (error) 1536 break; 1537 optp = &in6p->in6p_outputopts; 1538 error = ip6_pcbopt(optname, 1539 (u_char *)&tclass, 1540 sizeof(tclass), 1541 optp, 1542 kauth_cred_get(), uproto); 1543 break; 1544 } 1545 1546 case IPV6_TCLASS: 1547 case IPV6_DONTFRAG: 1548 case IPV6_USE_MIN_MTU: 1549 case IPV6_PREFER_TEMPADDR: 1550 error = sockopt_getint(sopt, &optval); 1551 if (error) 1552 break; 1553 { 1554 struct ip6_pktopts **optp; 1555 optp = &in6p->in6p_outputopts; 1556 error = ip6_pcbopt(optname, 1557 (u_char *)&optval, 1558 sizeof(optval), 1559 optp, 1560 kauth_cred_get(), uproto); 1561 break; 1562 } 1563 1564 #ifdef RFC2292 1565 case IPV6_2292PKTINFO: 1566 case IPV6_2292HOPLIMIT: 1567 case IPV6_2292HOPOPTS: 1568 case IPV6_2292DSTOPTS: 1569 case IPV6_2292RTHDR: 1570 /* RFC 2292 */ 1571 error = sockopt_getint(sopt, &optval); 1572 if (error) 1573 break; 1574 1575 switch (optname) { 1576 case IPV6_2292PKTINFO: 1577 OPTSET2292(IN6P_PKTINFO); 1578 break; 1579 case IPV6_2292HOPLIMIT: 1580 OPTSET2292(IN6P_HOPLIMIT); 1581 break; 1582 case IPV6_2292HOPOPTS: 1583 /* 1584 * Check super-user privilege. 1585 * See comments for IPV6_RECVHOPOPTS. 1586 */ 1587 error = 1588 kauth_authorize_network(kauth_cred_get(), 1589 KAUTH_NETWORK_IPV6, 1590 KAUTH_REQ_NETWORK_IPV6_HOPBYHOP, NULL, 1591 NULL, NULL); 1592 if (error) 1593 return (error); 1594 OPTSET2292(IN6P_HOPOPTS); 1595 break; 1596 case IPV6_2292DSTOPTS: 1597 error = 1598 kauth_authorize_network(kauth_cred_get(), 1599 KAUTH_NETWORK_IPV6, 1600 KAUTH_REQ_NETWORK_IPV6_HOPBYHOP, NULL, 1601 NULL, NULL); 1602 if (error) 1603 return (error); 1604 OPTSET2292(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); /* XXX */ 1605 break; 1606 case IPV6_2292RTHDR: 1607 OPTSET2292(IN6P_RTHDR); 1608 break; 1609 } 1610 break; 1611 #endif 1612 case IPV6_PKTINFO: 1613 case IPV6_HOPOPTS: 1614 case IPV6_RTHDR: 1615 case IPV6_DSTOPTS: 1616 case IPV6_RTHDRDSTOPTS: 1617 case IPV6_NEXTHOP: { 1618 /* new advanced API (RFC3542) */ 1619 void *optbuf; 1620 int optbuflen; 1621 struct ip6_pktopts **optp; 1622 1623 #ifdef RFC2292 1624 /* cannot mix with RFC2292 */ 1625 if (OPTBIT(IN6P_RFC2292)) { 1626 error = EINVAL; 1627 break; 1628 } 1629 #endif 1630 1631 optbuflen = sopt->sopt_size; 1632 optbuf = malloc(optbuflen, M_IP6OPT, M_NOWAIT); 1633 if (optbuf == NULL) { 1634 error = ENOBUFS; 1635 break; 1636 } 1637 1638 error = sockopt_get(sopt, optbuf, optbuflen); 1639 if (error) { 1640 free(optbuf, M_IP6OPT); 1641 break; 1642 } 1643 optp = &in6p->in6p_outputopts; 1644 error = ip6_pcbopt(optname, optbuf, optbuflen, 1645 optp, kauth_cred_get(), uproto); 1646 1647 free(optbuf, M_IP6OPT); 1648 break; 1649 } 1650 #undef OPTSET 1651 1652 case IPV6_MULTICAST_IF: 1653 case IPV6_MULTICAST_HOPS: 1654 case IPV6_MULTICAST_LOOP: 1655 case IPV6_JOIN_GROUP: 1656 case IPV6_LEAVE_GROUP: 1657 error = ip6_setmoptions(sopt, in6p); 1658 break; 1659 1660 case IPV6_PORTRANGE: 1661 error = sockopt_getint(sopt, &optval); 1662 if (error) 1663 break; 1664 1665 switch (optval) { 1666 case IPV6_PORTRANGE_DEFAULT: 1667 in6p->in6p_flags &= ~(IN6P_LOWPORT); 1668 in6p->in6p_flags &= ~(IN6P_HIGHPORT); 1669 break; 1670 1671 case IPV6_PORTRANGE_HIGH: 1672 in6p->in6p_flags &= ~(IN6P_LOWPORT); 1673 in6p->in6p_flags |= IN6P_HIGHPORT; 1674 break; 1675 1676 case IPV6_PORTRANGE_LOW: 1677 in6p->in6p_flags &= ~(IN6P_HIGHPORT); 1678 in6p->in6p_flags |= IN6P_LOWPORT; 1679 break; 1680 1681 default: 1682 error = EINVAL; 1683 break; 1684 } 1685 break; 1686 1687 case IPV6_PORTALGO: 1688 error = sockopt_getint(sopt, &optval); 1689 if (error) 1690 break; 1691 1692 error = portalgo_algo_index_select( 1693 (struct inpcb_hdr *)in6p, optval); 1694 break; 1695 1696 #if defined(IPSEC) 1697 case IPV6_IPSEC_POLICY: 1698 if (ipsec_enabled) { 1699 error = ipsec_set_policy(in6p, 1700 sopt->sopt_data, sopt->sopt_size, 1701 kauth_cred_get()); 1702 break; 1703 } 1704 /*FALLTHROUGH*/ 1705 #endif /* IPSEC */ 1706 1707 default: 1708 error = ENOPROTOOPT; 1709 break; 1710 } 1711 break; 1712 1713 case PRCO_GETOPT: 1714 switch (optname) { 1715 #ifdef RFC2292 1716 case IPV6_2292PKTOPTIONS: 1717 /* 1718 * RFC3542 (effectively) deprecated the 1719 * semantics of the 2292-style pktoptions. 1720 * Since it was not reliable in nature (i.e., 1721 * applications had to expect the lack of some 1722 * information after all), it would make sense 1723 * to simplify this part by always returning 1724 * empty data. 1725 */ 1726 break; 1727 #endif 1728 1729 case IPV6_RECVHOPOPTS: 1730 case IPV6_RECVDSTOPTS: 1731 case IPV6_RECVRTHDRDSTOPTS: 1732 case IPV6_UNICAST_HOPS: 1733 case IPV6_RECVPKTINFO: 1734 case IPV6_RECVHOPLIMIT: 1735 case IPV6_RECVRTHDR: 1736 case IPV6_RECVPATHMTU: 1737 1738 case IPV6_FAITH: 1739 case IPV6_V6ONLY: 1740 case IPV6_PORTRANGE: 1741 case IPV6_RECVTCLASS: 1742 switch (optname) { 1743 1744 case IPV6_RECVHOPOPTS: 1745 optval = OPTBIT(IN6P_HOPOPTS); 1746 break; 1747 1748 case IPV6_RECVDSTOPTS: 1749 optval = OPTBIT(IN6P_DSTOPTS); 1750 break; 1751 1752 case IPV6_RECVRTHDRDSTOPTS: 1753 optval = OPTBIT(IN6P_RTHDRDSTOPTS); 1754 break; 1755 1756 case IPV6_UNICAST_HOPS: 1757 optval = in6p->in6p_hops; 1758 break; 1759 1760 case IPV6_RECVPKTINFO: 1761 optval = OPTBIT(IN6P_PKTINFO); 1762 break; 1763 1764 case IPV6_RECVHOPLIMIT: 1765 optval = OPTBIT(IN6P_HOPLIMIT); 1766 break; 1767 1768 case IPV6_RECVRTHDR: 1769 optval = OPTBIT(IN6P_RTHDR); 1770 break; 1771 1772 case IPV6_RECVPATHMTU: 1773 optval = OPTBIT(IN6P_MTU); 1774 break; 1775 1776 case IPV6_FAITH: 1777 optval = OPTBIT(IN6P_FAITH); 1778 break; 1779 1780 case IPV6_V6ONLY: 1781 optval = OPTBIT(IN6P_IPV6_V6ONLY); 1782 break; 1783 1784 case IPV6_PORTRANGE: 1785 { 1786 int flags; 1787 flags = in6p->in6p_flags; 1788 if (flags & IN6P_HIGHPORT) 1789 optval = IPV6_PORTRANGE_HIGH; 1790 else if (flags & IN6P_LOWPORT) 1791 optval = IPV6_PORTRANGE_LOW; 1792 else 1793 optval = 0; 1794 break; 1795 } 1796 case IPV6_RECVTCLASS: 1797 optval = OPTBIT(IN6P_TCLASS); 1798 break; 1799 1800 } 1801 if (error) 1802 break; 1803 error = sockopt_setint(sopt, optval); 1804 break; 1805 1806 case IPV6_PATHMTU: 1807 { 1808 u_long pmtu = 0; 1809 struct ip6_mtuinfo mtuinfo; 1810 struct route *ro = &in6p->in6p_route; 1811 struct rtentry *rt; 1812 union { 1813 struct sockaddr dst; 1814 struct sockaddr_in6 dst6; 1815 } u; 1816 1817 if (!(so->so_state & SS_ISCONNECTED)) 1818 return (ENOTCONN); 1819 /* 1820 * XXX: we dot not consider the case of source 1821 * routing, or optional information to specify 1822 * the outgoing interface. 1823 */ 1824 sockaddr_in6_init(&u.dst6, &in6p->in6p_faddr, 0, 0, 0); 1825 rt = rtcache_lookup(ro, &u.dst); 1826 error = ip6_getpmtu(rt, NULL, &pmtu, NULL); 1827 rtcache_unref(rt, ro); 1828 if (error) 1829 break; 1830 if (pmtu > IPV6_MAXPACKET) 1831 pmtu = IPV6_MAXPACKET; 1832 1833 memset(&mtuinfo, 0, sizeof(mtuinfo)); 1834 mtuinfo.ip6m_mtu = (u_int32_t)pmtu; 1835 optdata = (void *)&mtuinfo; 1836 optdatalen = sizeof(mtuinfo); 1837 if (optdatalen > MCLBYTES) 1838 return (EMSGSIZE); /* XXX */ 1839 error = sockopt_set(sopt, optdata, optdatalen); 1840 break; 1841 } 1842 1843 #ifdef RFC2292 1844 case IPV6_2292PKTINFO: 1845 case IPV6_2292HOPLIMIT: 1846 case IPV6_2292HOPOPTS: 1847 case IPV6_2292RTHDR: 1848 case IPV6_2292DSTOPTS: 1849 switch (optname) { 1850 case IPV6_2292PKTINFO: 1851 optval = OPTBIT(IN6P_PKTINFO); 1852 break; 1853 case IPV6_2292HOPLIMIT: 1854 optval = OPTBIT(IN6P_HOPLIMIT); 1855 break; 1856 case IPV6_2292HOPOPTS: 1857 optval = OPTBIT(IN6P_HOPOPTS); 1858 break; 1859 case IPV6_2292RTHDR: 1860 optval = OPTBIT(IN6P_RTHDR); 1861 break; 1862 case IPV6_2292DSTOPTS: 1863 optval = OPTBIT(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); 1864 break; 1865 } 1866 error = sockopt_setint(sopt, optval); 1867 break; 1868 #endif 1869 case IPV6_PKTINFO: 1870 case IPV6_HOPOPTS: 1871 case IPV6_RTHDR: 1872 case IPV6_DSTOPTS: 1873 case IPV6_RTHDRDSTOPTS: 1874 case IPV6_NEXTHOP: 1875 case IPV6_OTCLASS: 1876 case IPV6_TCLASS: 1877 case IPV6_DONTFRAG: 1878 case IPV6_USE_MIN_MTU: 1879 case IPV6_PREFER_TEMPADDR: 1880 error = ip6_getpcbopt(in6p->in6p_outputopts, 1881 optname, sopt); 1882 break; 1883 1884 case IPV6_MULTICAST_IF: 1885 case IPV6_MULTICAST_HOPS: 1886 case IPV6_MULTICAST_LOOP: 1887 case IPV6_JOIN_GROUP: 1888 case IPV6_LEAVE_GROUP: 1889 error = ip6_getmoptions(sopt, in6p); 1890 break; 1891 1892 case IPV6_PORTALGO: 1893 optval = ((struct inpcb_hdr *)in6p)->inph_portalgo; 1894 error = sockopt_setint(sopt, optval); 1895 break; 1896 1897 #if defined(IPSEC) 1898 case IPV6_IPSEC_POLICY: 1899 if (ipsec_used) { 1900 struct mbuf *m = NULL; 1901 1902 /* 1903 * XXX: this will return EINVAL as sopt is 1904 * empty 1905 */ 1906 error = ipsec_get_policy(in6p, sopt->sopt_data, 1907 sopt->sopt_size, &m); 1908 if (!error) 1909 error = sockopt_setmbuf(sopt, m); 1910 break; 1911 } 1912 /*FALLTHROUGH*/ 1913 #endif /* IPSEC */ 1914 1915 default: 1916 error = ENOPROTOOPT; 1917 break; 1918 } 1919 break; 1920 } 1921 return (error); 1922 } 1923 1924 int 1925 ip6_raw_ctloutput(int op, struct socket *so, struct sockopt *sopt) 1926 { 1927 int error = 0, optval; 1928 const int icmp6off = offsetof(struct icmp6_hdr, icmp6_cksum); 1929 struct in6pcb *in6p = sotoin6pcb(so); 1930 int level, optname; 1931 1932 KASSERT(sopt != NULL); 1933 1934 level = sopt->sopt_level; 1935 optname = sopt->sopt_name; 1936 1937 if (level != IPPROTO_IPV6) { 1938 return ENOPROTOOPT; 1939 } 1940 1941 switch (optname) { 1942 case IPV6_CHECKSUM: 1943 /* 1944 * For ICMPv6 sockets, no modification allowed for checksum 1945 * offset, permit "no change" values to help existing apps. 1946 * 1947 * XXX RFC3542 says: "An attempt to set IPV6_CHECKSUM 1948 * for an ICMPv6 socket will fail." The current 1949 * behavior does not meet RFC3542. 1950 */ 1951 switch (op) { 1952 case PRCO_SETOPT: 1953 error = sockopt_getint(sopt, &optval); 1954 if (error) 1955 break; 1956 if ((optval % 2) != 0) { 1957 /* the API assumes even offset values */ 1958 error = EINVAL; 1959 } else if (so->so_proto->pr_protocol == 1960 IPPROTO_ICMPV6) { 1961 if (optval != icmp6off) 1962 error = EINVAL; 1963 } else 1964 in6p->in6p_cksum = optval; 1965 break; 1966 1967 case PRCO_GETOPT: 1968 if (so->so_proto->pr_protocol == IPPROTO_ICMPV6) 1969 optval = icmp6off; 1970 else 1971 optval = in6p->in6p_cksum; 1972 1973 error = sockopt_setint(sopt, optval); 1974 break; 1975 1976 default: 1977 error = EINVAL; 1978 break; 1979 } 1980 break; 1981 1982 default: 1983 error = ENOPROTOOPT; 1984 break; 1985 } 1986 1987 return (error); 1988 } 1989 1990 #ifdef RFC2292 1991 /* 1992 * Set up IP6 options in pcb for insertion in output packets or 1993 * specifying behavior of outgoing packets. 1994 */ 1995 static int 1996 ip6_pcbopts(struct ip6_pktopts **pktopt, struct socket *so, 1997 struct sockopt *sopt) 1998 { 1999 struct ip6_pktopts *opt = *pktopt; 2000 struct mbuf *m; 2001 int error = 0; 2002 2003 KASSERT(solocked(so)); 2004 2005 /* turn off any old options. */ 2006 if (opt) { 2007 #ifdef DIAGNOSTIC 2008 if (opt->ip6po_pktinfo || opt->ip6po_nexthop || 2009 opt->ip6po_hbh || opt->ip6po_dest1 || opt->ip6po_dest2 || 2010 opt->ip6po_rhinfo.ip6po_rhi_rthdr) 2011 printf("ip6_pcbopts: all specified options are cleared.\n"); 2012 #endif 2013 ip6_clearpktopts(opt, -1); 2014 } else { 2015 opt = malloc(sizeof(*opt), M_IP6OPT, M_NOWAIT); 2016 if (opt == NULL) 2017 return (ENOBUFS); 2018 } 2019 *pktopt = NULL; 2020 2021 if (sopt == NULL || sopt->sopt_size == 0) { 2022 /* 2023 * Only turning off any previous options, regardless of 2024 * whether the opt is just created or given. 2025 */ 2026 free(opt, M_IP6OPT); 2027 return (0); 2028 } 2029 2030 /* set options specified by user. */ 2031 m = sockopt_getmbuf(sopt); 2032 if (m == NULL) { 2033 free(opt, M_IP6OPT); 2034 return (ENOBUFS); 2035 } 2036 2037 error = ip6_setpktopts(m, opt, NULL, kauth_cred_get(), 2038 so->so_proto->pr_protocol); 2039 m_freem(m); 2040 if (error != 0) { 2041 ip6_clearpktopts(opt, -1); /* XXX: discard all options */ 2042 free(opt, M_IP6OPT); 2043 return (error); 2044 } 2045 *pktopt = opt; 2046 return (0); 2047 } 2048 #endif 2049 2050 /* 2051 * initialize ip6_pktopts. beware that there are non-zero default values in 2052 * the struct. 2053 */ 2054 void 2055 ip6_initpktopts(struct ip6_pktopts *opt) 2056 { 2057 2058 memset(opt, 0, sizeof(*opt)); 2059 opt->ip6po_hlim = -1; /* -1 means default hop limit */ 2060 opt->ip6po_tclass = -1; /* -1 means default traffic class */ 2061 opt->ip6po_minmtu = IP6PO_MINMTU_MCASTONLY; 2062 opt->ip6po_prefer_tempaddr = IP6PO_TEMPADDR_SYSTEM; 2063 } 2064 2065 #define sin6tosa(sin6) ((struct sockaddr *)(sin6)) /* XXX */ 2066 static int 2067 ip6_pcbopt(int optname, u_char *buf, int len, struct ip6_pktopts **pktopt, 2068 kauth_cred_t cred, int uproto) 2069 { 2070 struct ip6_pktopts *opt; 2071 2072 if (*pktopt == NULL) { 2073 *pktopt = malloc(sizeof(struct ip6_pktopts), M_IP6OPT, 2074 M_NOWAIT); 2075 if (*pktopt == NULL) 2076 return (ENOBUFS); 2077 2078 ip6_initpktopts(*pktopt); 2079 } 2080 opt = *pktopt; 2081 2082 return (ip6_setpktopt(optname, buf, len, opt, cred, 1, 0, uproto)); 2083 } 2084 2085 static int 2086 ip6_getpcbopt(struct ip6_pktopts *pktopt, int optname, struct sockopt *sopt) 2087 { 2088 void *optdata = NULL; 2089 int optdatalen = 0; 2090 struct ip6_ext *ip6e; 2091 int error = 0; 2092 struct in6_pktinfo null_pktinfo; 2093 int deftclass = 0, on; 2094 int defminmtu = IP6PO_MINMTU_MCASTONLY; 2095 int defpreftemp = IP6PO_TEMPADDR_SYSTEM; 2096 2097 switch (optname) { 2098 case IPV6_PKTINFO: 2099 if (pktopt && pktopt->ip6po_pktinfo) 2100 optdata = (void *)pktopt->ip6po_pktinfo; 2101 else { 2102 /* XXX: we don't have to do this every time... */ 2103 memset(&null_pktinfo, 0, sizeof(null_pktinfo)); 2104 optdata = (void *)&null_pktinfo; 2105 } 2106 optdatalen = sizeof(struct in6_pktinfo); 2107 break; 2108 case IPV6_OTCLASS: 2109 /* XXX */ 2110 return (EINVAL); 2111 case IPV6_TCLASS: 2112 if (pktopt && pktopt->ip6po_tclass >= 0) 2113 optdata = (void *)&pktopt->ip6po_tclass; 2114 else 2115 optdata = (void *)&deftclass; 2116 optdatalen = sizeof(int); 2117 break; 2118 case IPV6_HOPOPTS: 2119 if (pktopt && pktopt->ip6po_hbh) { 2120 optdata = (void *)pktopt->ip6po_hbh; 2121 ip6e = (struct ip6_ext *)pktopt->ip6po_hbh; 2122 optdatalen = (ip6e->ip6e_len + 1) << 3; 2123 } 2124 break; 2125 case IPV6_RTHDR: 2126 if (pktopt && pktopt->ip6po_rthdr) { 2127 optdata = (void *)pktopt->ip6po_rthdr; 2128 ip6e = (struct ip6_ext *)pktopt->ip6po_rthdr; 2129 optdatalen = (ip6e->ip6e_len + 1) << 3; 2130 } 2131 break; 2132 case IPV6_RTHDRDSTOPTS: 2133 if (pktopt && pktopt->ip6po_dest1) { 2134 optdata = (void *)pktopt->ip6po_dest1; 2135 ip6e = (struct ip6_ext *)pktopt->ip6po_dest1; 2136 optdatalen = (ip6e->ip6e_len + 1) << 3; 2137 } 2138 break; 2139 case IPV6_DSTOPTS: 2140 if (pktopt && pktopt->ip6po_dest2) { 2141 optdata = (void *)pktopt->ip6po_dest2; 2142 ip6e = (struct ip6_ext *)pktopt->ip6po_dest2; 2143 optdatalen = (ip6e->ip6e_len + 1) << 3; 2144 } 2145 break; 2146 case IPV6_NEXTHOP: 2147 if (pktopt && pktopt->ip6po_nexthop) { 2148 optdata = (void *)pktopt->ip6po_nexthop; 2149 optdatalen = pktopt->ip6po_nexthop->sa_len; 2150 } 2151 break; 2152 case IPV6_USE_MIN_MTU: 2153 if (pktopt) 2154 optdata = (void *)&pktopt->ip6po_minmtu; 2155 else 2156 optdata = (void *)&defminmtu; 2157 optdatalen = sizeof(int); 2158 break; 2159 case IPV6_DONTFRAG: 2160 if (pktopt && ((pktopt->ip6po_flags) & IP6PO_DONTFRAG)) 2161 on = 1; 2162 else 2163 on = 0; 2164 optdata = (void *)&on; 2165 optdatalen = sizeof(on); 2166 break; 2167 case IPV6_PREFER_TEMPADDR: 2168 if (pktopt) 2169 optdata = (void *)&pktopt->ip6po_prefer_tempaddr; 2170 else 2171 optdata = (void *)&defpreftemp; 2172 optdatalen = sizeof(int); 2173 break; 2174 default: /* should not happen */ 2175 #ifdef DIAGNOSTIC 2176 panic("ip6_getpcbopt: unexpected option\n"); 2177 #endif 2178 return (ENOPROTOOPT); 2179 } 2180 2181 error = sockopt_set(sopt, optdata, optdatalen); 2182 2183 return (error); 2184 } 2185 2186 void 2187 ip6_clearpktopts(struct ip6_pktopts *pktopt, int optname) 2188 { 2189 if (optname == -1 || optname == IPV6_PKTINFO) { 2190 if (pktopt->ip6po_pktinfo) 2191 free(pktopt->ip6po_pktinfo, M_IP6OPT); 2192 pktopt->ip6po_pktinfo = NULL; 2193 } 2194 if (optname == -1 || optname == IPV6_HOPLIMIT) 2195 pktopt->ip6po_hlim = -1; 2196 if (optname == -1 || optname == IPV6_TCLASS) 2197 pktopt->ip6po_tclass = -1; 2198 if (optname == -1 || optname == IPV6_NEXTHOP) { 2199 rtcache_free(&pktopt->ip6po_nextroute); 2200 if (pktopt->ip6po_nexthop) 2201 free(pktopt->ip6po_nexthop, M_IP6OPT); 2202 pktopt->ip6po_nexthop = NULL; 2203 } 2204 if (optname == -1 || optname == IPV6_HOPOPTS) { 2205 if (pktopt->ip6po_hbh) 2206 free(pktopt->ip6po_hbh, M_IP6OPT); 2207 pktopt->ip6po_hbh = NULL; 2208 } 2209 if (optname == -1 || optname == IPV6_RTHDRDSTOPTS) { 2210 if (pktopt->ip6po_dest1) 2211 free(pktopt->ip6po_dest1, M_IP6OPT); 2212 pktopt->ip6po_dest1 = NULL; 2213 } 2214 if (optname == -1 || optname == IPV6_RTHDR) { 2215 if (pktopt->ip6po_rhinfo.ip6po_rhi_rthdr) 2216 free(pktopt->ip6po_rhinfo.ip6po_rhi_rthdr, M_IP6OPT); 2217 pktopt->ip6po_rhinfo.ip6po_rhi_rthdr = NULL; 2218 rtcache_free(&pktopt->ip6po_route); 2219 } 2220 if (optname == -1 || optname == IPV6_DSTOPTS) { 2221 if (pktopt->ip6po_dest2) 2222 free(pktopt->ip6po_dest2, M_IP6OPT); 2223 pktopt->ip6po_dest2 = NULL; 2224 } 2225 } 2226 2227 #define PKTOPT_EXTHDRCPY(type) \ 2228 do { \ 2229 if (src->type) { \ 2230 int hlen = (((struct ip6_ext *)src->type)->ip6e_len + 1) << 3;\ 2231 dst->type = malloc(hlen, M_IP6OPT, canwait); \ 2232 if (dst->type == NULL) \ 2233 goto bad; \ 2234 memcpy(dst->type, src->type, hlen); \ 2235 } \ 2236 } while (/*CONSTCOND*/ 0) 2237 2238 static int 2239 copypktopts(struct ip6_pktopts *dst, struct ip6_pktopts *src, int canwait) 2240 { 2241 dst->ip6po_hlim = src->ip6po_hlim; 2242 dst->ip6po_tclass = src->ip6po_tclass; 2243 dst->ip6po_flags = src->ip6po_flags; 2244 dst->ip6po_minmtu = src->ip6po_minmtu; 2245 dst->ip6po_prefer_tempaddr = src->ip6po_prefer_tempaddr; 2246 if (src->ip6po_pktinfo) { 2247 dst->ip6po_pktinfo = malloc(sizeof(*dst->ip6po_pktinfo), 2248 M_IP6OPT, canwait); 2249 if (dst->ip6po_pktinfo == NULL) 2250 goto bad; 2251 *dst->ip6po_pktinfo = *src->ip6po_pktinfo; 2252 } 2253 if (src->ip6po_nexthop) { 2254 dst->ip6po_nexthop = malloc(src->ip6po_nexthop->sa_len, 2255 M_IP6OPT, canwait); 2256 if (dst->ip6po_nexthop == NULL) 2257 goto bad; 2258 memcpy(dst->ip6po_nexthop, src->ip6po_nexthop, 2259 src->ip6po_nexthop->sa_len); 2260 } 2261 PKTOPT_EXTHDRCPY(ip6po_hbh); 2262 PKTOPT_EXTHDRCPY(ip6po_dest1); 2263 PKTOPT_EXTHDRCPY(ip6po_dest2); 2264 PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */ 2265 return (0); 2266 2267 bad: 2268 if (dst->ip6po_pktinfo) free(dst->ip6po_pktinfo, M_IP6OPT); 2269 if (dst->ip6po_nexthop) free(dst->ip6po_nexthop, M_IP6OPT); 2270 if (dst->ip6po_hbh) free(dst->ip6po_hbh, M_IP6OPT); 2271 if (dst->ip6po_dest1) free(dst->ip6po_dest1, M_IP6OPT); 2272 if (dst->ip6po_dest2) free(dst->ip6po_dest2, M_IP6OPT); 2273 if (dst->ip6po_rthdr) free(dst->ip6po_rthdr, M_IP6OPT); 2274 2275 return (ENOBUFS); 2276 } 2277 #undef PKTOPT_EXTHDRCPY 2278 2279 struct ip6_pktopts * 2280 ip6_copypktopts(struct ip6_pktopts *src, int canwait) 2281 { 2282 int error; 2283 struct ip6_pktopts *dst; 2284 2285 dst = malloc(sizeof(*dst), M_IP6OPT, canwait); 2286 if (dst == NULL) 2287 return (NULL); 2288 ip6_initpktopts(dst); 2289 2290 if ((error = copypktopts(dst, src, canwait)) != 0) { 2291 free(dst, M_IP6OPT); 2292 return (NULL); 2293 } 2294 2295 return (dst); 2296 } 2297 2298 void 2299 ip6_freepcbopts(struct ip6_pktopts *pktopt) 2300 { 2301 if (pktopt == NULL) 2302 return; 2303 2304 ip6_clearpktopts(pktopt, -1); 2305 2306 free(pktopt, M_IP6OPT); 2307 } 2308 2309 int 2310 ip6_get_membership(const struct sockopt *sopt, struct ifnet **ifp, 2311 struct psref *psref, void *v, size_t l) 2312 { 2313 struct ipv6_mreq mreq; 2314 int error; 2315 struct in6_addr *ia = &mreq.ipv6mr_multiaddr; 2316 struct in_addr *ia4 = (void *)&ia->s6_addr32[3]; 2317 2318 error = sockopt_get(sopt, &mreq, sizeof(mreq)); 2319 if (error != 0) 2320 return error; 2321 2322 if (IN6_IS_ADDR_UNSPECIFIED(ia)) { 2323 /* 2324 * We use the unspecified address to specify to accept 2325 * all multicast addresses. Only super user is allowed 2326 * to do this. 2327 */ 2328 if (kauth_authorize_network(curlwp->l_cred, KAUTH_NETWORK_IPV6, 2329 KAUTH_REQ_NETWORK_IPV6_JOIN_MULTICAST, NULL, NULL, NULL)) 2330 return EACCES; 2331 } else if (IN6_IS_ADDR_V4MAPPED(ia)) { 2332 // Don't bother if we are not going to use ifp. 2333 if (l == sizeof(*ia)) { 2334 memcpy(v, ia, l); 2335 return 0; 2336 } 2337 } else if (!IN6_IS_ADDR_MULTICAST(ia)) { 2338 return EINVAL; 2339 } 2340 2341 /* 2342 * If no interface was explicitly specified, choose an 2343 * appropriate one according to the given multicast address. 2344 */ 2345 if (mreq.ipv6mr_interface == 0) { 2346 struct rtentry *rt; 2347 union { 2348 struct sockaddr dst; 2349 struct sockaddr_in dst4; 2350 struct sockaddr_in6 dst6; 2351 } u; 2352 struct route ro; 2353 2354 /* 2355 * Look up the routing table for the 2356 * address, and choose the outgoing interface. 2357 * XXX: is it a good approach? 2358 */ 2359 memset(&ro, 0, sizeof(ro)); 2360 if (IN6_IS_ADDR_V4MAPPED(ia)) 2361 sockaddr_in_init(&u.dst4, ia4, 0); 2362 else 2363 sockaddr_in6_init(&u.dst6, ia, 0, 0, 0); 2364 error = rtcache_setdst(&ro, &u.dst); 2365 if (error != 0) 2366 return error; 2367 rt = rtcache_init(&ro); 2368 *ifp = rt != NULL ? 2369 if_get_byindex(rt->rt_ifp->if_index, psref) : NULL; 2370 rtcache_unref(rt, &ro); 2371 rtcache_free(&ro); 2372 } else { 2373 /* 2374 * If the interface is specified, validate it. 2375 */ 2376 *ifp = if_get_byindex(mreq.ipv6mr_interface, psref); 2377 if (*ifp == NULL) 2378 return ENXIO; /* XXX EINVAL? */ 2379 } 2380 if (sizeof(*ia) == l) 2381 memcpy(v, ia, l); 2382 else 2383 memcpy(v, ia4, l); 2384 return 0; 2385 } 2386 2387 /* 2388 * Set the IP6 multicast options in response to user setsockopt(). 2389 */ 2390 static int 2391 ip6_setmoptions(const struct sockopt *sopt, struct in6pcb *in6p) 2392 { 2393 int error = 0; 2394 u_int loop, ifindex; 2395 struct ipv6_mreq mreq; 2396 struct in6_addr ia; 2397 struct ifnet *ifp; 2398 struct ip6_moptions *im6o = in6p->in6p_moptions; 2399 struct in6_multi_mship *imm; 2400 2401 KASSERT(in6p_locked(in6p)); 2402 2403 if (im6o == NULL) { 2404 /* 2405 * No multicast option buffer attached to the pcb; 2406 * allocate one and initialize to default values. 2407 */ 2408 im6o = malloc(sizeof(*im6o), M_IPMOPTS, M_NOWAIT); 2409 if (im6o == NULL) 2410 return (ENOBUFS); 2411 in6p->in6p_moptions = im6o; 2412 im6o->im6o_multicast_if_index = 0; 2413 im6o->im6o_multicast_hlim = ip6_defmcasthlim; 2414 im6o->im6o_multicast_loop = IPV6_DEFAULT_MULTICAST_LOOP; 2415 LIST_INIT(&im6o->im6o_memberships); 2416 } 2417 2418 switch (sopt->sopt_name) { 2419 2420 case IPV6_MULTICAST_IF: { 2421 int s; 2422 /* 2423 * Select the interface for outgoing multicast packets. 2424 */ 2425 error = sockopt_get(sopt, &ifindex, sizeof(ifindex)); 2426 if (error != 0) 2427 break; 2428 2429 s = pserialize_read_enter(); 2430 if (ifindex != 0) { 2431 if ((ifp = if_byindex(ifindex)) == NULL) { 2432 pserialize_read_exit(s); 2433 error = ENXIO; /* XXX EINVAL? */ 2434 break; 2435 } 2436 if ((ifp->if_flags & IFF_MULTICAST) == 0) { 2437 pserialize_read_exit(s); 2438 error = EADDRNOTAVAIL; 2439 break; 2440 } 2441 } else 2442 ifp = NULL; 2443 im6o->im6o_multicast_if_index = if_get_index(ifp); 2444 pserialize_read_exit(s); 2445 break; 2446 } 2447 2448 case IPV6_MULTICAST_HOPS: 2449 { 2450 /* 2451 * Set the IP6 hoplimit for outgoing multicast packets. 2452 */ 2453 int optval; 2454 2455 error = sockopt_getint(sopt, &optval); 2456 if (error != 0) 2457 break; 2458 2459 if (optval < -1 || optval >= 256) 2460 error = EINVAL; 2461 else if (optval == -1) 2462 im6o->im6o_multicast_hlim = ip6_defmcasthlim; 2463 else 2464 im6o->im6o_multicast_hlim = optval; 2465 break; 2466 } 2467 2468 case IPV6_MULTICAST_LOOP: 2469 /* 2470 * Set the loopback flag for outgoing multicast packets. 2471 * Must be zero or one. 2472 */ 2473 error = sockopt_get(sopt, &loop, sizeof(loop)); 2474 if (error != 0) 2475 break; 2476 if (loop > 1) { 2477 error = EINVAL; 2478 break; 2479 } 2480 im6o->im6o_multicast_loop = loop; 2481 break; 2482 2483 case IPV6_JOIN_GROUP: { 2484 int bound; 2485 struct psref psref; 2486 /* 2487 * Add a multicast group membership. 2488 * Group must be a valid IP6 multicast address. 2489 */ 2490 bound = curlwp_bind(); 2491 ifp = NULL; 2492 error = ip6_get_membership(sopt, &ifp, &psref, &ia, sizeof(ia)); 2493 if (error != 0) { 2494 KASSERT(ifp == NULL); 2495 curlwp_bindx(bound); 2496 return error; 2497 } 2498 2499 if (IN6_IS_ADDR_V4MAPPED(&ia)) { 2500 error = ip_setmoptions(&in6p->in6p_v4moptions, sopt); 2501 goto put_break; 2502 } 2503 /* 2504 * See if we found an interface, and confirm that it 2505 * supports multicast 2506 */ 2507 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) { 2508 error = EADDRNOTAVAIL; 2509 goto put_break; 2510 } 2511 2512 if (in6_setscope(&ia, ifp, NULL)) { 2513 error = EADDRNOTAVAIL; /* XXX: should not happen */ 2514 goto put_break; 2515 } 2516 2517 /* 2518 * See if the membership already exists. 2519 */ 2520 LIST_FOREACH(imm, &im6o->im6o_memberships, i6mm_chain) { 2521 if (imm->i6mm_maddr->in6m_ifp == ifp && 2522 IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr, 2523 &ia)) 2524 goto put_break; 2525 } 2526 if (imm != NULL) { 2527 error = EADDRINUSE; 2528 goto put_break; 2529 } 2530 /* 2531 * Everything looks good; add a new record to the multicast 2532 * address list for the given interface. 2533 */ 2534 IFNET_LOCK(ifp); 2535 imm = in6_joingroup(ifp, &ia, &error, 0); 2536 IFNET_UNLOCK(ifp); 2537 if (imm == NULL) 2538 goto put_break; 2539 LIST_INSERT_HEAD(&im6o->im6o_memberships, imm, i6mm_chain); 2540 put_break: 2541 if_put(ifp, &psref); 2542 curlwp_bindx(bound); 2543 break; 2544 } 2545 2546 case IPV6_LEAVE_GROUP: { 2547 struct ifnet *in6m_ifp; 2548 /* 2549 * Drop a multicast group membership. 2550 * Group must be a valid IP6 multicast address. 2551 */ 2552 error = sockopt_get(sopt, &mreq, sizeof(mreq)); 2553 if (error != 0) 2554 break; 2555 2556 if (IN6_IS_ADDR_V4MAPPED(&mreq.ipv6mr_multiaddr)) { 2557 error = ip_setmoptions(&in6p->in6p_v4moptions, sopt); 2558 break; 2559 } 2560 /* 2561 * If an interface address was specified, get a pointer 2562 * to its ifnet structure. 2563 */ 2564 if (mreq.ipv6mr_interface != 0) { 2565 if ((ifp = if_byindex(mreq.ipv6mr_interface)) == NULL) { 2566 error = ENXIO; /* XXX EINVAL? */ 2567 break; 2568 } 2569 } else 2570 ifp = NULL; 2571 2572 /* Fill in the scope zone ID */ 2573 if (ifp) { 2574 if (in6_setscope(&mreq.ipv6mr_multiaddr, ifp, NULL)) { 2575 /* XXX: should not happen */ 2576 error = EADDRNOTAVAIL; 2577 break; 2578 } 2579 } else if (mreq.ipv6mr_interface != 0) { 2580 /* 2581 * XXX: This case would happens when the (positive) 2582 * index is in the valid range, but the corresponding 2583 * interface has been detached dynamically. The above 2584 * check probably avoids such case to happen here, but 2585 * we check it explicitly for safety. 2586 */ 2587 error = EADDRNOTAVAIL; 2588 break; 2589 } else { /* ipv6mr_interface == 0 */ 2590 struct sockaddr_in6 sa6_mc; 2591 2592 /* 2593 * The API spec says as follows: 2594 * If the interface index is specified as 0, the 2595 * system may choose a multicast group membership to 2596 * drop by matching the multicast address only. 2597 * On the other hand, we cannot disambiguate the scope 2598 * zone unless an interface is provided. Thus, we 2599 * check if there's ambiguity with the default scope 2600 * zone as the last resort. 2601 */ 2602 sockaddr_in6_init(&sa6_mc, &mreq.ipv6mr_multiaddr, 2603 0, 0, 0); 2604 error = sa6_embedscope(&sa6_mc, ip6_use_defzone); 2605 if (error != 0) 2606 break; 2607 mreq.ipv6mr_multiaddr = sa6_mc.sin6_addr; 2608 } 2609 2610 /* 2611 * Find the membership in the membership list. 2612 */ 2613 LIST_FOREACH(imm, &im6o->im6o_memberships, i6mm_chain) { 2614 if ((ifp == NULL || imm->i6mm_maddr->in6m_ifp == ifp) && 2615 IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr, 2616 &mreq.ipv6mr_multiaddr)) 2617 break; 2618 } 2619 if (imm == NULL) { 2620 /* Unable to resolve interface */ 2621 error = EADDRNOTAVAIL; 2622 break; 2623 } 2624 /* 2625 * Give up the multicast address record to which the 2626 * membership points. 2627 */ 2628 LIST_REMOVE(imm, i6mm_chain); 2629 in6m_ifp = imm->i6mm_maddr->in6m_ifp; 2630 IFNET_LOCK(in6m_ifp); 2631 in6_leavegroup(imm); 2632 /* in6m_ifp should not leave thanks to in6p_lock */ 2633 IFNET_UNLOCK(in6m_ifp); 2634 break; 2635 } 2636 2637 default: 2638 error = EOPNOTSUPP; 2639 break; 2640 } 2641 2642 /* 2643 * If all options have default values, no need to keep the mbuf. 2644 */ 2645 if (im6o->im6o_multicast_if_index == 0 && 2646 im6o->im6o_multicast_hlim == ip6_defmcasthlim && 2647 im6o->im6o_multicast_loop == IPV6_DEFAULT_MULTICAST_LOOP && 2648 LIST_EMPTY(&im6o->im6o_memberships)) { 2649 free(in6p->in6p_moptions, M_IPMOPTS); 2650 in6p->in6p_moptions = NULL; 2651 } 2652 2653 return (error); 2654 } 2655 2656 /* 2657 * Return the IP6 multicast options in response to user getsockopt(). 2658 */ 2659 static int 2660 ip6_getmoptions(struct sockopt *sopt, struct in6pcb *in6p) 2661 { 2662 u_int optval; 2663 int error; 2664 struct ip6_moptions *im6o = in6p->in6p_moptions; 2665 2666 switch (sopt->sopt_name) { 2667 case IPV6_MULTICAST_IF: 2668 if (im6o == NULL || im6o->im6o_multicast_if_index == 0) 2669 optval = 0; 2670 else 2671 optval = im6o->im6o_multicast_if_index; 2672 2673 error = sockopt_set(sopt, &optval, sizeof(optval)); 2674 break; 2675 2676 case IPV6_MULTICAST_HOPS: 2677 if (im6o == NULL) 2678 optval = ip6_defmcasthlim; 2679 else 2680 optval = im6o->im6o_multicast_hlim; 2681 2682 error = sockopt_set(sopt, &optval, sizeof(optval)); 2683 break; 2684 2685 case IPV6_MULTICAST_LOOP: 2686 if (im6o == NULL) 2687 optval = IPV6_DEFAULT_MULTICAST_LOOP; 2688 else 2689 optval = im6o->im6o_multicast_loop; 2690 2691 error = sockopt_set(sopt, &optval, sizeof(optval)); 2692 break; 2693 2694 default: 2695 error = EOPNOTSUPP; 2696 } 2697 2698 return (error); 2699 } 2700 2701 /* 2702 * Discard the IP6 multicast options. 2703 */ 2704 void 2705 ip6_freemoptions(struct ip6_moptions *im6o) 2706 { 2707 struct in6_multi_mship *imm, *nimm; 2708 2709 if (im6o == NULL) 2710 return; 2711 2712 /* The owner of im6o (in6p) should be protected by solock */ 2713 LIST_FOREACH_SAFE(imm, &im6o->im6o_memberships, i6mm_chain, nimm) { 2714 struct ifnet *ifp; 2715 2716 LIST_REMOVE(imm, i6mm_chain); 2717 2718 ifp = imm->i6mm_maddr->in6m_ifp; 2719 IFNET_LOCK(ifp); 2720 in6_leavegroup(imm); 2721 /* ifp should not leave thanks to solock */ 2722 IFNET_UNLOCK(ifp); 2723 } 2724 free(im6o, M_IPMOPTS); 2725 } 2726 2727 /* 2728 * Set IPv6 outgoing packet options based on advanced API. 2729 */ 2730 int 2731 ip6_setpktopts(struct mbuf *control, struct ip6_pktopts *opt, 2732 struct ip6_pktopts *stickyopt, kauth_cred_t cred, int uproto) 2733 { 2734 struct cmsghdr *cm = 0; 2735 2736 if (control == NULL || opt == NULL) 2737 return (EINVAL); 2738 2739 ip6_initpktopts(opt); 2740 if (stickyopt) { 2741 int error; 2742 2743 /* 2744 * If stickyopt is provided, make a local copy of the options 2745 * for this particular packet, then override them by ancillary 2746 * objects. 2747 * XXX: copypktopts() does not copy the cached route to a next 2748 * hop (if any). This is not very good in terms of efficiency, 2749 * but we can allow this since this option should be rarely 2750 * used. 2751 */ 2752 if ((error = copypktopts(opt, stickyopt, M_NOWAIT)) != 0) 2753 return (error); 2754 } 2755 2756 /* 2757 * XXX: Currently, we assume all the optional information is stored 2758 * in a single mbuf. 2759 */ 2760 if (control->m_next) 2761 return (EINVAL); 2762 2763 /* XXX if cm->cmsg_len is not aligned, control->m_len can become <0 */ 2764 for (; control->m_len > 0; control->m_data += CMSG_ALIGN(cm->cmsg_len), 2765 control->m_len -= CMSG_ALIGN(cm->cmsg_len)) { 2766 int error; 2767 2768 if (control->m_len < CMSG_LEN(0)) 2769 return (EINVAL); 2770 2771 cm = mtod(control, struct cmsghdr *); 2772 if (cm->cmsg_len == 0 || cm->cmsg_len > control->m_len) 2773 return (EINVAL); 2774 if (cm->cmsg_level != IPPROTO_IPV6) 2775 continue; 2776 2777 error = ip6_setpktopt(cm->cmsg_type, CMSG_DATA(cm), 2778 cm->cmsg_len - CMSG_LEN(0), opt, cred, 0, 1, uproto); 2779 if (error) 2780 return (error); 2781 } 2782 2783 return (0); 2784 } 2785 2786 /* 2787 * Set a particular packet option, as a sticky option or an ancillary data 2788 * item. "len" can be 0 only when it's a sticky option. 2789 * We have 4 cases of combination of "sticky" and "cmsg": 2790 * "sticky=0, cmsg=0": impossible 2791 * "sticky=0, cmsg=1": RFC2292 or RFC3542 ancillary data 2792 * "sticky=1, cmsg=0": RFC3542 socket option 2793 * "sticky=1, cmsg=1": RFC2292 socket option 2794 */ 2795 static int 2796 ip6_setpktopt(int optname, u_char *buf, int len, struct ip6_pktopts *opt, 2797 kauth_cred_t cred, int sticky, int cmsg, int uproto) 2798 { 2799 int minmtupolicy; 2800 int error; 2801 2802 if (!sticky && !cmsg) { 2803 #ifdef DIAGNOSTIC 2804 printf("ip6_setpktopt: impossible case\n"); 2805 #endif 2806 return (EINVAL); 2807 } 2808 2809 /* 2810 * IPV6_2292xxx is for backward compatibility to RFC2292, and should 2811 * not be specified in the context of RFC3542. Conversely, 2812 * RFC3542 types should not be specified in the context of RFC2292. 2813 */ 2814 if (!cmsg) { 2815 switch (optname) { 2816 case IPV6_2292PKTINFO: 2817 case IPV6_2292HOPLIMIT: 2818 case IPV6_2292NEXTHOP: 2819 case IPV6_2292HOPOPTS: 2820 case IPV6_2292DSTOPTS: 2821 case IPV6_2292RTHDR: 2822 case IPV6_2292PKTOPTIONS: 2823 return (ENOPROTOOPT); 2824 } 2825 } 2826 if (sticky && cmsg) { 2827 switch (optname) { 2828 case IPV6_PKTINFO: 2829 case IPV6_HOPLIMIT: 2830 case IPV6_NEXTHOP: 2831 case IPV6_HOPOPTS: 2832 case IPV6_DSTOPTS: 2833 case IPV6_RTHDRDSTOPTS: 2834 case IPV6_RTHDR: 2835 case IPV6_USE_MIN_MTU: 2836 case IPV6_DONTFRAG: 2837 case IPV6_OTCLASS: 2838 case IPV6_TCLASS: 2839 case IPV6_PREFER_TEMPADDR: /* XXX not an RFC3542 option */ 2840 return (ENOPROTOOPT); 2841 } 2842 } 2843 2844 switch (optname) { 2845 #ifdef RFC2292 2846 case IPV6_2292PKTINFO: 2847 #endif 2848 case IPV6_PKTINFO: 2849 { 2850 struct in6_pktinfo *pktinfo; 2851 2852 if (len != sizeof(struct in6_pktinfo)) 2853 return (EINVAL); 2854 2855 pktinfo = (struct in6_pktinfo *)buf; 2856 2857 /* 2858 * An application can clear any sticky IPV6_PKTINFO option by 2859 * doing a "regular" setsockopt with ipi6_addr being 2860 * in6addr_any and ipi6_ifindex being zero. 2861 * [RFC 3542, Section 6] 2862 */ 2863 if (optname == IPV6_PKTINFO && opt->ip6po_pktinfo && 2864 pktinfo->ipi6_ifindex == 0 && 2865 IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) { 2866 ip6_clearpktopts(opt, optname); 2867 break; 2868 } 2869 2870 if (uproto == IPPROTO_TCP && optname == IPV6_PKTINFO && 2871 sticky && !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) { 2872 return (EINVAL); 2873 } 2874 2875 /* Validate the interface index if specified. */ 2876 if (pktinfo->ipi6_ifindex) { 2877 struct ifnet *ifp; 2878 int s = pserialize_read_enter(); 2879 ifp = if_byindex(pktinfo->ipi6_ifindex); 2880 if (ifp == NULL) { 2881 pserialize_read_exit(s); 2882 return ENXIO; 2883 } 2884 pserialize_read_exit(s); 2885 } 2886 2887 /* 2888 * We store the address anyway, and let in6_selectsrc() 2889 * validate the specified address. This is because ipi6_addr 2890 * may not have enough information about its scope zone, and 2891 * we may need additional information (such as outgoing 2892 * interface or the scope zone of a destination address) to 2893 * disambiguate the scope. 2894 * XXX: the delay of the validation may confuse the 2895 * application when it is used as a sticky option. 2896 */ 2897 if (opt->ip6po_pktinfo == NULL) { 2898 opt->ip6po_pktinfo = malloc(sizeof(*pktinfo), 2899 M_IP6OPT, M_NOWAIT); 2900 if (opt->ip6po_pktinfo == NULL) 2901 return (ENOBUFS); 2902 } 2903 memcpy(opt->ip6po_pktinfo, pktinfo, sizeof(*pktinfo)); 2904 break; 2905 } 2906 2907 #ifdef RFC2292 2908 case IPV6_2292HOPLIMIT: 2909 #endif 2910 case IPV6_HOPLIMIT: 2911 { 2912 int *hlimp; 2913 2914 /* 2915 * RFC 3542 deprecated the usage of sticky IPV6_HOPLIMIT 2916 * to simplify the ordering among hoplimit options. 2917 */ 2918 if (optname == IPV6_HOPLIMIT && sticky) 2919 return (ENOPROTOOPT); 2920 2921 if (len != sizeof(int)) 2922 return (EINVAL); 2923 hlimp = (int *)buf; 2924 if (*hlimp < -1 || *hlimp > 255) 2925 return (EINVAL); 2926 2927 opt->ip6po_hlim = *hlimp; 2928 break; 2929 } 2930 2931 case IPV6_OTCLASS: 2932 if (len != sizeof(u_int8_t)) 2933 return (EINVAL); 2934 2935 opt->ip6po_tclass = *(u_int8_t *)buf; 2936 break; 2937 2938 case IPV6_TCLASS: 2939 { 2940 int tclass; 2941 2942 if (len != sizeof(int)) 2943 return (EINVAL); 2944 tclass = *(int *)buf; 2945 if (tclass < -1 || tclass > 255) 2946 return (EINVAL); 2947 2948 opt->ip6po_tclass = tclass; 2949 break; 2950 } 2951 2952 #ifdef RFC2292 2953 case IPV6_2292NEXTHOP: 2954 #endif 2955 case IPV6_NEXTHOP: 2956 error = kauth_authorize_network(cred, KAUTH_NETWORK_IPV6, 2957 KAUTH_REQ_NETWORK_IPV6_HOPBYHOP, NULL, NULL, NULL); 2958 if (error) 2959 return (error); 2960 2961 if (len == 0) { /* just remove the option */ 2962 ip6_clearpktopts(opt, IPV6_NEXTHOP); 2963 break; 2964 } 2965 2966 /* check if cmsg_len is large enough for sa_len */ 2967 if (len < sizeof(struct sockaddr) || len < *buf) 2968 return (EINVAL); 2969 2970 switch (((struct sockaddr *)buf)->sa_family) { 2971 case AF_INET6: 2972 { 2973 struct sockaddr_in6 *sa6 = (struct sockaddr_in6 *)buf; 2974 2975 if (sa6->sin6_len != sizeof(struct sockaddr_in6)) 2976 return (EINVAL); 2977 2978 if (IN6_IS_ADDR_UNSPECIFIED(&sa6->sin6_addr) || 2979 IN6_IS_ADDR_MULTICAST(&sa6->sin6_addr)) { 2980 return (EINVAL); 2981 } 2982 if ((error = sa6_embedscope(sa6, ip6_use_defzone)) 2983 != 0) { 2984 return (error); 2985 } 2986 break; 2987 } 2988 case AF_LINK: /* eventually be supported? */ 2989 default: 2990 return (EAFNOSUPPORT); 2991 } 2992 2993 /* turn off the previous option, then set the new option. */ 2994 ip6_clearpktopts(opt, IPV6_NEXTHOP); 2995 opt->ip6po_nexthop = malloc(*buf, M_IP6OPT, M_NOWAIT); 2996 if (opt->ip6po_nexthop == NULL) 2997 return (ENOBUFS); 2998 memcpy(opt->ip6po_nexthop, buf, *buf); 2999 break; 3000 3001 #ifdef RFC2292 3002 case IPV6_2292HOPOPTS: 3003 #endif 3004 case IPV6_HOPOPTS: 3005 { 3006 struct ip6_hbh *hbh; 3007 int hbhlen; 3008 3009 /* 3010 * XXX: We don't allow a non-privileged user to set ANY HbH 3011 * options, since per-option restriction has too much 3012 * overhead. 3013 */ 3014 error = kauth_authorize_network(cred, KAUTH_NETWORK_IPV6, 3015 KAUTH_REQ_NETWORK_IPV6_HOPBYHOP, NULL, NULL, NULL); 3016 if (error) 3017 return (error); 3018 3019 if (len == 0) { 3020 ip6_clearpktopts(opt, IPV6_HOPOPTS); 3021 break; /* just remove the option */ 3022 } 3023 3024 /* message length validation */ 3025 if (len < sizeof(struct ip6_hbh)) 3026 return (EINVAL); 3027 hbh = (struct ip6_hbh *)buf; 3028 hbhlen = (hbh->ip6h_len + 1) << 3; 3029 if (len != hbhlen) 3030 return (EINVAL); 3031 3032 /* turn off the previous option, then set the new option. */ 3033 ip6_clearpktopts(opt, IPV6_HOPOPTS); 3034 opt->ip6po_hbh = malloc(hbhlen, M_IP6OPT, M_NOWAIT); 3035 if (opt->ip6po_hbh == NULL) 3036 return (ENOBUFS); 3037 memcpy(opt->ip6po_hbh, hbh, hbhlen); 3038 3039 break; 3040 } 3041 3042 #ifdef RFC2292 3043 case IPV6_2292DSTOPTS: 3044 #endif 3045 case IPV6_DSTOPTS: 3046 case IPV6_RTHDRDSTOPTS: 3047 { 3048 struct ip6_dest *dest, **newdest = NULL; 3049 int destlen; 3050 3051 /* XXX: see the comment for IPV6_HOPOPTS */ 3052 error = kauth_authorize_network(cred, KAUTH_NETWORK_IPV6, 3053 KAUTH_REQ_NETWORK_IPV6_HOPBYHOP, NULL, NULL, NULL); 3054 if (error) 3055 return (error); 3056 3057 if (len == 0) { 3058 ip6_clearpktopts(opt, optname); 3059 break; /* just remove the option */ 3060 } 3061 3062 /* message length validation */ 3063 if (len < sizeof(struct ip6_dest)) 3064 return (EINVAL); 3065 dest = (struct ip6_dest *)buf; 3066 destlen = (dest->ip6d_len + 1) << 3; 3067 if (len != destlen) 3068 return (EINVAL); 3069 /* 3070 * Determine the position that the destination options header 3071 * should be inserted; before or after the routing header. 3072 */ 3073 switch (optname) { 3074 case IPV6_2292DSTOPTS: 3075 /* 3076 * The old advanced API is ambiguous on this point. 3077 * Our approach is to determine the position based 3078 * according to the existence of a routing header. 3079 * Note, however, that this depends on the order of the 3080 * extension headers in the ancillary data; the 1st 3081 * part of the destination options header must appear 3082 * before the routing header in the ancillary data, 3083 * too. 3084 * RFC3542 solved the ambiguity by introducing 3085 * separate ancillary data or option types. 3086 */ 3087 if (opt->ip6po_rthdr == NULL) 3088 newdest = &opt->ip6po_dest1; 3089 else 3090 newdest = &opt->ip6po_dest2; 3091 break; 3092 case IPV6_RTHDRDSTOPTS: 3093 newdest = &opt->ip6po_dest1; 3094 break; 3095 case IPV6_DSTOPTS: 3096 newdest = &opt->ip6po_dest2; 3097 break; 3098 } 3099 3100 /* turn off the previous option, then set the new option. */ 3101 ip6_clearpktopts(opt, optname); 3102 *newdest = malloc(destlen, M_IP6OPT, M_NOWAIT); 3103 if (*newdest == NULL) 3104 return (ENOBUFS); 3105 memcpy(*newdest, dest, destlen); 3106 3107 break; 3108 } 3109 3110 #ifdef RFC2292 3111 case IPV6_2292RTHDR: 3112 #endif 3113 case IPV6_RTHDR: 3114 { 3115 struct ip6_rthdr *rth; 3116 int rthlen; 3117 3118 if (len == 0) { 3119 ip6_clearpktopts(opt, IPV6_RTHDR); 3120 break; /* just remove the option */ 3121 } 3122 3123 /* message length validation */ 3124 if (len < sizeof(struct ip6_rthdr)) 3125 return (EINVAL); 3126 rth = (struct ip6_rthdr *)buf; 3127 rthlen = (rth->ip6r_len + 1) << 3; 3128 if (len != rthlen) 3129 return (EINVAL); 3130 switch (rth->ip6r_type) { 3131 case IPV6_RTHDR_TYPE_0: 3132 /* Dropped, RFC5095. */ 3133 default: 3134 return (EINVAL); /* not supported */ 3135 } 3136 /* turn off the previous option */ 3137 ip6_clearpktopts(opt, IPV6_RTHDR); 3138 opt->ip6po_rthdr = malloc(rthlen, M_IP6OPT, M_NOWAIT); 3139 if (opt->ip6po_rthdr == NULL) 3140 return (ENOBUFS); 3141 memcpy(opt->ip6po_rthdr, rth, rthlen); 3142 break; 3143 } 3144 3145 case IPV6_USE_MIN_MTU: 3146 if (len != sizeof(int)) 3147 return (EINVAL); 3148 minmtupolicy = *(int *)buf; 3149 if (minmtupolicy != IP6PO_MINMTU_MCASTONLY && 3150 minmtupolicy != IP6PO_MINMTU_DISABLE && 3151 minmtupolicy != IP6PO_MINMTU_ALL) { 3152 return (EINVAL); 3153 } 3154 opt->ip6po_minmtu = minmtupolicy; 3155 break; 3156 3157 case IPV6_DONTFRAG: 3158 if (len != sizeof(int)) 3159 return (EINVAL); 3160 3161 if (uproto == IPPROTO_TCP || *(int *)buf == 0) { 3162 /* 3163 * we ignore this option for TCP sockets. 3164 * (RFC3542 leaves this case unspecified.) 3165 */ 3166 opt->ip6po_flags &= ~IP6PO_DONTFRAG; 3167 } else 3168 opt->ip6po_flags |= IP6PO_DONTFRAG; 3169 break; 3170 3171 case IPV6_PREFER_TEMPADDR: 3172 { 3173 int preftemp; 3174 3175 if (len != sizeof(int)) 3176 return (EINVAL); 3177 preftemp = *(int *)buf; 3178 switch (preftemp) { 3179 case IP6PO_TEMPADDR_SYSTEM: 3180 case IP6PO_TEMPADDR_NOTPREFER: 3181 case IP6PO_TEMPADDR_PREFER: 3182 break; 3183 default: 3184 return (EINVAL); 3185 } 3186 opt->ip6po_prefer_tempaddr = preftemp; 3187 break; 3188 } 3189 3190 default: 3191 return (ENOPROTOOPT); 3192 } /* end of switch */ 3193 3194 return (0); 3195 } 3196 3197 /* 3198 * Routine called from ip6_output() to loop back a copy of an IP6 multicast 3199 * packet to the input queue of a specified interface. Note that this 3200 * calls the output routine of the loopback "driver", but with an interface 3201 * pointer that might NOT be lo0ifp -- easier than replicating that code here. 3202 */ 3203 void 3204 ip6_mloopback(struct ifnet *ifp, struct mbuf *m, 3205 const struct sockaddr_in6 *dst) 3206 { 3207 struct mbuf *copym; 3208 struct ip6_hdr *ip6; 3209 3210 copym = m_copypacket(m, M_DONTWAIT); 3211 if (copym == NULL) 3212 return; 3213 3214 /* 3215 * Make sure to deep-copy IPv6 header portion in case the data 3216 * is in an mbuf cluster, so that we can safely override the IPv6 3217 * header portion later. 3218 */ 3219 if ((copym->m_flags & M_EXT) != 0 || 3220 copym->m_len < sizeof(struct ip6_hdr)) { 3221 copym = m_pullup(copym, sizeof(struct ip6_hdr)); 3222 if (copym == NULL) 3223 return; 3224 } 3225 3226 #ifdef DIAGNOSTIC 3227 if (copym->m_len < sizeof(*ip6)) { 3228 m_freem(copym); 3229 return; 3230 } 3231 #endif 3232 3233 ip6 = mtod(copym, struct ip6_hdr *); 3234 /* 3235 * clear embedded scope identifiers if necessary. 3236 * in6_clearscope will touch the addresses only when necessary. 3237 */ 3238 in6_clearscope(&ip6->ip6_src); 3239 in6_clearscope(&ip6->ip6_dst); 3240 3241 (void)looutput(ifp, copym, (const struct sockaddr *)dst, NULL); 3242 } 3243 3244 /* 3245 * Chop IPv6 header off from the payload. 3246 */ 3247 static int 3248 ip6_splithdr(struct mbuf *m, struct ip6_exthdrs *exthdrs) 3249 { 3250 struct mbuf *mh; 3251 struct ip6_hdr *ip6; 3252 3253 ip6 = mtod(m, struct ip6_hdr *); 3254 if (m->m_len > sizeof(*ip6)) { 3255 MGETHDR(mh, M_DONTWAIT, MT_HEADER); 3256 if (mh == NULL) { 3257 m_freem(m); 3258 return ENOBUFS; 3259 } 3260 M_MOVE_PKTHDR(mh, m); 3261 MH_ALIGN(mh, sizeof(*ip6)); 3262 m->m_len -= sizeof(*ip6); 3263 m->m_data += sizeof(*ip6); 3264 mh->m_next = m; 3265 mh->m_len = sizeof(*ip6); 3266 memcpy(mtod(mh, void *), (void *)ip6, sizeof(*ip6)); 3267 m = mh; 3268 } 3269 exthdrs->ip6e_ip6 = m; 3270 return 0; 3271 } 3272 3273 /* 3274 * Compute IPv6 extension header length. 3275 */ 3276 int 3277 ip6_optlen(struct in6pcb *in6p) 3278 { 3279 int len; 3280 3281 if (!in6p->in6p_outputopts) 3282 return 0; 3283 3284 len = 0; 3285 #define elen(x) \ 3286 (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0) 3287 3288 len += elen(in6p->in6p_outputopts->ip6po_hbh); 3289 len += elen(in6p->in6p_outputopts->ip6po_dest1); 3290 len += elen(in6p->in6p_outputopts->ip6po_rthdr); 3291 len += elen(in6p->in6p_outputopts->ip6po_dest2); 3292 return len; 3293 #undef elen 3294 } 3295 3296 /* 3297 * Ensure sending address is valid. 3298 * Returns 0 on success, -1 if an error should be sent back or 1 3299 * if the packet could be dropped without error (protocol dependent). 3300 */ 3301 static int 3302 ip6_ifaddrvalid(const struct in6_addr *src, const struct in6_addr *dst) 3303 { 3304 struct sockaddr_in6 sin6; 3305 int s, error; 3306 struct ifaddr *ifa; 3307 struct in6_ifaddr *ia6; 3308 3309 if (IN6_IS_ADDR_UNSPECIFIED(src)) 3310 return 0; 3311 3312 memset(&sin6, 0, sizeof(sin6)); 3313 sin6.sin6_family = AF_INET6; 3314 sin6.sin6_len = sizeof(sin6); 3315 sin6.sin6_addr = *src; 3316 3317 s = pserialize_read_enter(); 3318 ifa = ifa_ifwithaddr(sin6tosa(&sin6)); 3319 if ((ia6 = ifatoia6(ifa)) == NULL || 3320 ia6->ia6_flags & (IN6_IFF_ANYCAST | IN6_IFF_DUPLICATED)) 3321 error = -1; 3322 else if (ia6->ia6_flags & IN6_IFF_TENTATIVE) 3323 error = 1; 3324 else if (ia6->ia6_flags & IN6_IFF_DETACHED && 3325 (sin6.sin6_addr = *dst, ifa_ifwithaddr(sin6tosa(&sin6)) == NULL)) 3326 /* Allow internal traffic to DETACHED addresses */ 3327 error = 1; 3328 else 3329 error = 0; 3330 pserialize_read_exit(s); 3331 3332 return error; 3333 } 3334
Indexes created Tue Sep 30 20:09:53 GMT 2025