ip6_output.c revision 1.186
1 /* $NetBSD: ip6_output.c,v 1.186 2017/02/22 07:46:00 ozaki-r Exp $ */ 2 /* $KAME: ip6_output.c,v 1.172 2001/03/25 09:55:56 itojun Exp $ */ 3 4 /* 5 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project. 6 * All rights reserved. 7 * 8 * Redistribution and use in source and binary forms, with or without 9 * modification, are permitted provided that the following conditions 10 * are met: 11 * 1. Redistributions of source code must retain the above copyright 12 * notice, this list of conditions and the following disclaimer. 13 * 2. Redistributions in binary form must reproduce the above copyright 14 * notice, this list of conditions and the following disclaimer in the 15 * documentation and/or other materials provided with the distribution. 16 * 3. Neither the name of the project nor the names of its contributors 17 * may be used to endorse or promote products derived from this software 18 * without specific prior written permission. 19 * 20 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND 21 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 22 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 23 * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE 24 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 25 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 26 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 27 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 28 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 29 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 30 * SUCH DAMAGE. 31 */ 32 33 /* 34 * Copyright (c) 1982, 1986, 1988, 1990, 1993 35 * The Regents of the University of California. All rights reserved. 36 * 37 * Redistribution and use in source and binary forms, with or without 38 * modification, are permitted provided that the following conditions 39 * are met: 40 * 1. Redistributions of source code must retain the above copyright 41 * notice, this list of conditions and the following disclaimer. 42 * 2. Redistributions in binary form must reproduce the above copyright 43 * notice, this list of conditions and the following disclaimer in the 44 * documentation and/or other materials provided with the distribution. 45 * 3. Neither the name of the University nor the names of its contributors 46 * may be used to endorse or promote products derived from this software 47 * without specific prior written permission. 48 * 49 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 50 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 51 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 52 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 53 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 54 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 55 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 56 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 57 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 58 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 59 * SUCH DAMAGE. 60 * 61 * @(#)ip_output.c 8.3 (Berkeley) 1/21/94 62 */ 63 64 #include <sys/cdefs.h> 65 __KERNEL_RCSID(0, "$NetBSD: ip6_output.c,v 1.186 2017/02/22 07:46:00 ozaki-r Exp $"); 66 67 #ifdef _KERNEL_OPT 68 #include "opt_inet.h" 69 #include "opt_inet6.h" 70 #include "opt_ipsec.h" 71 #endif 72 73 #include <sys/param.h> 74 #include <sys/malloc.h> 75 #include <sys/mbuf.h> 76 #include <sys/errno.h> 77 #include <sys/socket.h> 78 #include <sys/socketvar.h> 79 #include <sys/syslog.h> 80 #include <sys/systm.h> 81 #include <sys/proc.h> 82 #include <sys/kauth.h> 83 84 #include <net/if.h> 85 #include <net/route.h> 86 #include <net/pfil.h> 87 88 #include <netinet/in.h> 89 #include <netinet/in_var.h> 90 #include <netinet/ip6.h> 91 #include <netinet/ip_var.h> 92 #include <netinet/icmp6.h> 93 #include <netinet/in_offload.h> 94 #include <netinet/portalgo.h> 95 #include <netinet6/in6_offload.h> 96 #include <netinet6/ip6_var.h> 97 #include <netinet6/ip6_private.h> 98 #include <netinet6/in6_pcb.h> 99 #include <netinet6/nd6.h> 100 #include <netinet6/ip6protosw.h> 101 #include <netinet6/scope6_var.h> 102 103 #ifdef IPSEC 104 #include <netipsec/ipsec.h> 105 #include <netipsec/ipsec6.h> 106 #include <netipsec/key.h> 107 #include <netipsec/xform.h> 108 #endif 109 110 111 #include <net/net_osdep.h> 112 113 extern pfil_head_t *inet6_pfil_hook; /* XXX */ 114 115 struct ip6_exthdrs { 116 struct mbuf *ip6e_ip6; 117 struct mbuf *ip6e_hbh; 118 struct mbuf *ip6e_dest1; 119 struct mbuf *ip6e_rthdr; 120 struct mbuf *ip6e_dest2; 121 }; 122 123 static int ip6_pcbopt(int, u_char *, int, struct ip6_pktopts **, 124 kauth_cred_t, int); 125 static int ip6_getpcbopt(struct ip6_pktopts *, int, struct sockopt *); 126 static int ip6_setpktopt(int, u_char *, int, struct ip6_pktopts *, kauth_cred_t, 127 int, int, int); 128 static int ip6_setmoptions(const struct sockopt *, struct in6pcb *); 129 static int ip6_getmoptions(struct sockopt *, struct in6pcb *); 130 static int ip6_copyexthdr(struct mbuf **, void *, int); 131 static int ip6_insertfraghdr(struct mbuf *, struct mbuf *, int, 132 struct ip6_frag **); 133 static int ip6_insert_jumboopt(struct ip6_exthdrs *, u_int32_t); 134 static int ip6_splithdr(struct mbuf *, struct ip6_exthdrs *); 135 static int ip6_getpmtu(struct rtentry *, struct ifnet *, u_long *, int *); 136 static int copypktopts(struct ip6_pktopts *, struct ip6_pktopts *, int); 137 static int ip6_ifaddrvalid(const struct in6_addr *); 138 static int ip6_handle_rthdr(struct ip6_rthdr *, struct ip6_hdr *); 139 140 #ifdef RFC2292 141 static int ip6_pcbopts(struct ip6_pktopts **, struct socket *, struct sockopt *); 142 #endif 143 144 static int 145 ip6_handle_rthdr(struct ip6_rthdr *rh, struct ip6_hdr *ip6) 146 { 147 struct ip6_rthdr0 *rh0; 148 struct in6_addr *addr; 149 struct sockaddr_in6 sa; 150 int error = 0; 151 152 switch (rh->ip6r_type) { 153 case IPV6_RTHDR_TYPE_0: 154 rh0 = (struct ip6_rthdr0 *)rh; 155 addr = (struct in6_addr *)(rh0 + 1); 156 157 /* 158 * construct a sockaddr_in6 form of the first hop. 159 * 160 * XXX we may not have enough information about its scope zone; 161 * there is no standard API to pass the information from the 162 * application. 163 */ 164 sockaddr_in6_init(&sa, addr, 0, 0, 0); 165 error = sa6_embedscope(&sa, ip6_use_defzone); 166 if (error != 0) 167 break; 168 (void)memmove(&addr[0], &addr[1], 169 sizeof(struct in6_addr) * (rh0->ip6r0_segleft - 1)); 170 addr[rh0->ip6r0_segleft - 1] = ip6->ip6_dst; 171 ip6->ip6_dst = sa.sin6_addr; 172 /* XXX */ 173 in6_clearscope(addr + rh0->ip6r0_segleft - 1); 174 break; 175 default: /* is it possible? */ 176 error = EINVAL; 177 } 178 179 return error; 180 } 181 182 /* 183 * Send an IP packet to a host. 184 */ 185 int 186 ip6_if_output(struct ifnet * const ifp, struct ifnet * const origifp, 187 struct mbuf * const m, 188 const struct sockaddr_in6 * const dst, const struct rtentry *rt) 189 { 190 int error = 0; 191 192 if (rt != NULL) { 193 error = rt_check_reject_route(rt, ifp); 194 if (error != 0) { 195 m_freem(m); 196 return error; 197 } 198 } 199 200 if ((ifp->if_flags & IFF_LOOPBACK) != 0) 201 error = if_output_lock(ifp, origifp, m, sin6tocsa(dst), rt); 202 else 203 error = if_output_lock(ifp, ifp, m, sin6tocsa(dst), rt); 204 return error; 205 } 206 207 /* 208 * IP6 output. The packet in mbuf chain m contains a skeletal IP6 209 * header (with pri, len, nxt, hlim, src, dst). 210 * This function may modify ver and hlim only. 211 * The mbuf chain containing the packet will be freed. 212 * The mbuf opt, if present, will not be freed. 213 * 214 * type of "mtu": rt_rmx.rmx_mtu is u_long, ifnet.ifr_mtu is int, and 215 * nd_ifinfo.linkmtu is u_int32_t. so we use u_long to hold largest one, 216 * which is rt_rmx.rmx_mtu. 217 */ 218 int 219 ip6_output( 220 struct mbuf *m0, 221 struct ip6_pktopts *opt, 222 struct route *ro, 223 int flags, 224 struct ip6_moptions *im6o, 225 struct socket *so, 226 struct ifnet **ifpp /* XXX: just for statistics */ 227 ) 228 { 229 struct ip6_hdr *ip6, *mhip6; 230 struct ifnet *ifp = NULL, *origifp = NULL; 231 struct mbuf *m = m0; 232 int hlen, tlen, len, off; 233 bool tso; 234 struct route ip6route; 235 struct rtentry *rt = NULL, *rt_pmtu; 236 const struct sockaddr_in6 *dst; 237 struct sockaddr_in6 src_sa, dst_sa; 238 int error = 0; 239 struct in6_ifaddr *ia = NULL; 240 u_long mtu; 241 int alwaysfrag, dontfrag; 242 u_int32_t optlen = 0, plen = 0, unfragpartlen = 0; 243 struct ip6_exthdrs exthdrs; 244 struct in6_addr finaldst, src0, dst0; 245 u_int32_t zone; 246 struct route *ro_pmtu = NULL; 247 int hdrsplit = 0; 248 int needipsec = 0; 249 #ifdef IPSEC 250 struct secpolicy *sp = NULL; 251 #endif 252 struct psref psref, psref_ia; 253 int bound = curlwp_bind(); 254 bool release_psref_ia = false; 255 256 #ifdef DIAGNOSTIC 257 if ((m->m_flags & M_PKTHDR) == 0) 258 panic("ip6_output: no HDR"); 259 260 if ((m->m_pkthdr.csum_flags & 261 (M_CSUM_TCPv4|M_CSUM_UDPv4|M_CSUM_TSOv4)) != 0) { 262 panic("ip6_output: IPv4 checksum offload flags: %d", 263 m->m_pkthdr.csum_flags); 264 } 265 266 if ((m->m_pkthdr.csum_flags & (M_CSUM_TCPv6|M_CSUM_UDPv6)) == 267 (M_CSUM_TCPv6|M_CSUM_UDPv6)) { 268 panic("ip6_output: conflicting checksum offload flags: %d", 269 m->m_pkthdr.csum_flags); 270 } 271 #endif 272 273 M_CSUM_DATA_IPv6_HL_SET(m->m_pkthdr.csum_data, sizeof(struct ip6_hdr)); 274 275 #define MAKE_EXTHDR(hp, mp) \ 276 do { \ 277 if (hp) { \ 278 struct ip6_ext *eh = (struct ip6_ext *)(hp); \ 279 error = ip6_copyexthdr((mp), (void *)(hp), \ 280 ((eh)->ip6e_len + 1) << 3); \ 281 if (error) \ 282 goto freehdrs; \ 283 } \ 284 } while (/*CONSTCOND*/ 0) 285 286 memset(&exthdrs, 0, sizeof(exthdrs)); 287 if (opt) { 288 /* Hop-by-Hop options header */ 289 MAKE_EXTHDR(opt->ip6po_hbh, &exthdrs.ip6e_hbh); 290 /* Destination options header(1st part) */ 291 MAKE_EXTHDR(opt->ip6po_dest1, &exthdrs.ip6e_dest1); 292 /* Routing header */ 293 MAKE_EXTHDR(opt->ip6po_rthdr, &exthdrs.ip6e_rthdr); 294 /* Destination options header(2nd part) */ 295 MAKE_EXTHDR(opt->ip6po_dest2, &exthdrs.ip6e_dest2); 296 } 297 298 /* 299 * Calculate the total length of the extension header chain. 300 * Keep the length of the unfragmentable part for fragmentation. 301 */ 302 optlen = 0; 303 if (exthdrs.ip6e_hbh) optlen += exthdrs.ip6e_hbh->m_len; 304 if (exthdrs.ip6e_dest1) optlen += exthdrs.ip6e_dest1->m_len; 305 if (exthdrs.ip6e_rthdr) optlen += exthdrs.ip6e_rthdr->m_len; 306 unfragpartlen = optlen + sizeof(struct ip6_hdr); 307 /* NOTE: we don't add AH/ESP length here. do that later. */ 308 if (exthdrs.ip6e_dest2) optlen += exthdrs.ip6e_dest2->m_len; 309 310 #ifdef IPSEC 311 if (ipsec_used) { 312 /* Check the security policy (SP) for the packet */ 313 314 sp = ipsec6_check_policy(m, so, flags, &needipsec, &error); 315 if (error != 0) { 316 /* 317 * Hack: -EINVAL is used to signal that a packet 318 * should be silently discarded. This is typically 319 * because we asked key management for an SA and 320 * it was delayed (e.g. kicked up to IKE). 321 */ 322 if (error == -EINVAL) 323 error = 0; 324 goto freehdrs; 325 } 326 } 327 #endif /* IPSEC */ 328 329 330 if (needipsec && 331 (m->m_pkthdr.csum_flags & (M_CSUM_UDPv6|M_CSUM_TCPv6)) != 0) { 332 in6_delayed_cksum(m); 333 m->m_pkthdr.csum_flags &= ~(M_CSUM_UDPv6|M_CSUM_TCPv6); 334 } 335 336 337 /* 338 * If we need IPsec, or there is at least one extension header, 339 * separate IP6 header from the payload. 340 */ 341 if ((needipsec || optlen) && !hdrsplit) { 342 if ((error = ip6_splithdr(m, &exthdrs)) != 0) { 343 m = NULL; 344 goto freehdrs; 345 } 346 m = exthdrs.ip6e_ip6; 347 hdrsplit++; 348 } 349 350 /* adjust pointer */ 351 ip6 = mtod(m, struct ip6_hdr *); 352 353 /* adjust mbuf packet header length */ 354 m->m_pkthdr.len += optlen; 355 plen = m->m_pkthdr.len - sizeof(*ip6); 356 357 /* If this is a jumbo payload, insert a jumbo payload option. */ 358 if (plen > IPV6_MAXPACKET) { 359 if (!hdrsplit) { 360 if ((error = ip6_splithdr(m, &exthdrs)) != 0) { 361 m = NULL; 362 goto freehdrs; 363 } 364 m = exthdrs.ip6e_ip6; 365 hdrsplit++; 366 } 367 /* adjust pointer */ 368 ip6 = mtod(m, struct ip6_hdr *); 369 if ((error = ip6_insert_jumboopt(&exthdrs, plen)) != 0) 370 goto freehdrs; 371 optlen += 8; /* XXX JUMBOOPTLEN */ 372 ip6->ip6_plen = 0; 373 } else 374 ip6->ip6_plen = htons(plen); 375 376 /* 377 * Concatenate headers and fill in next header fields. 378 * Here we have, on "m" 379 * IPv6 payload 380 * and we insert headers accordingly. Finally, we should be getting: 381 * IPv6 hbh dest1 rthdr ah* [esp* dest2 payload] 382 * 383 * during the header composing process, "m" points to IPv6 header. 384 * "mprev" points to an extension header prior to esp. 385 */ 386 { 387 u_char *nexthdrp = &ip6->ip6_nxt; 388 struct mbuf *mprev = m; 389 390 /* 391 * we treat dest2 specially. this makes IPsec processing 392 * much easier. the goal here is to make mprev point the 393 * mbuf prior to dest2. 394 * 395 * result: IPv6 dest2 payload 396 * m and mprev will point to IPv6 header. 397 */ 398 if (exthdrs.ip6e_dest2) { 399 if (!hdrsplit) 400 panic("assumption failed: hdr not split"); 401 exthdrs.ip6e_dest2->m_next = m->m_next; 402 m->m_next = exthdrs.ip6e_dest2; 403 *mtod(exthdrs.ip6e_dest2, u_char *) = ip6->ip6_nxt; 404 ip6->ip6_nxt = IPPROTO_DSTOPTS; 405 } 406 407 #define MAKE_CHAIN(m, mp, p, i)\ 408 do {\ 409 if (m) {\ 410 if (!hdrsplit) \ 411 panic("assumption failed: hdr not split"); \ 412 *mtod((m), u_char *) = *(p);\ 413 *(p) = (i);\ 414 p = mtod((m), u_char *);\ 415 (m)->m_next = (mp)->m_next;\ 416 (mp)->m_next = (m);\ 417 (mp) = (m);\ 418 }\ 419 } while (/*CONSTCOND*/ 0) 420 /* 421 * result: IPv6 hbh dest1 rthdr dest2 payload 422 * m will point to IPv6 header. mprev will point to the 423 * extension header prior to dest2 (rthdr in the above case). 424 */ 425 MAKE_CHAIN(exthdrs.ip6e_hbh, mprev, nexthdrp, IPPROTO_HOPOPTS); 426 MAKE_CHAIN(exthdrs.ip6e_dest1, mprev, nexthdrp, 427 IPPROTO_DSTOPTS); 428 MAKE_CHAIN(exthdrs.ip6e_rthdr, mprev, nexthdrp, 429 IPPROTO_ROUTING); 430 431 M_CSUM_DATA_IPv6_HL_SET(m->m_pkthdr.csum_data, 432 sizeof(struct ip6_hdr) + optlen); 433 } 434 435 /* Need to save for pmtu */ 436 finaldst = ip6->ip6_dst; 437 438 /* 439 * If there is a routing header, replace destination address field 440 * with the first hop of the routing header. 441 */ 442 if (exthdrs.ip6e_rthdr) { 443 struct ip6_rthdr *rh; 444 445 rh = (struct ip6_rthdr *)(mtod(exthdrs.ip6e_rthdr, 446 struct ip6_rthdr *)); 447 448 error = ip6_handle_rthdr(rh, ip6); 449 if (error != 0) 450 goto bad; 451 } 452 453 /* Source address validation */ 454 if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) && 455 (flags & IPV6_UNSPECSRC) == 0) { 456 error = EOPNOTSUPP; 457 IP6_STATINC(IP6_STAT_BADSCOPE); 458 goto bad; 459 } 460 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) { 461 error = EOPNOTSUPP; 462 IP6_STATINC(IP6_STAT_BADSCOPE); 463 goto bad; 464 } 465 466 IP6_STATINC(IP6_STAT_LOCALOUT); 467 468 /* 469 * Route packet. 470 */ 471 /* initialize cached route */ 472 if (ro == NULL) { 473 memset(&ip6route, 0, sizeof(ip6route)); 474 ro = &ip6route; 475 } 476 ro_pmtu = ro; 477 if (opt && opt->ip6po_rthdr) 478 ro = &opt->ip6po_route; 479 480 /* 481 * if specified, try to fill in the traffic class field. 482 * do not override if a non-zero value is already set. 483 * we check the diffserv field and the ecn field separately. 484 */ 485 if (opt && opt->ip6po_tclass >= 0) { 486 int mask = 0; 487 488 if ((ip6->ip6_flow & htonl(0xfc << 20)) == 0) 489 mask |= 0xfc; 490 if ((ip6->ip6_flow & htonl(0x03 << 20)) == 0) 491 mask |= 0x03; 492 if (mask != 0) 493 ip6->ip6_flow |= htonl((opt->ip6po_tclass & mask) << 20); 494 } 495 496 /* fill in or override the hop limit field, if necessary. */ 497 if (opt && opt->ip6po_hlim != -1) 498 ip6->ip6_hlim = opt->ip6po_hlim & 0xff; 499 else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) { 500 if (im6o != NULL) 501 ip6->ip6_hlim = im6o->im6o_multicast_hlim; 502 else 503 ip6->ip6_hlim = ip6_defmcasthlim; 504 } 505 506 #ifdef IPSEC 507 if (needipsec) { 508 int s = splsoftnet(); 509 error = ipsec6_process_packet(m, sp->req); 510 511 /* 512 * Preserve KAME behaviour: ENOENT can be returned 513 * when an SA acquire is in progress. Don't propagate 514 * this to user-level; it confuses applications. 515 * XXX this will go away when the SADB is redone. 516 */ 517 if (error == ENOENT) 518 error = 0; 519 splx(s); 520 goto done; 521 } 522 #endif /* IPSEC */ 523 524 /* adjust pointer */ 525 ip6 = mtod(m, struct ip6_hdr *); 526 527 sockaddr_in6_init(&dst_sa, &ip6->ip6_dst, 0, 0, 0); 528 529 /* We do not need a route for multicast */ 530 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) { 531 struct in6_pktinfo *pi = NULL; 532 533 /* 534 * If the outgoing interface for the address is specified by 535 * the caller, use it. 536 */ 537 if (opt && (pi = opt->ip6po_pktinfo) != NULL) { 538 /* XXX boundary check is assumed to be already done. */ 539 ifp = if_get_byindex(pi->ipi6_ifindex, &psref); 540 } else if (im6o != NULL) { 541 ifp = if_get_byindex(im6o->im6o_multicast_if_index, 542 &psref); 543 } 544 } 545 546 if (ifp == NULL) { 547 error = in6_selectroute(&dst_sa, opt, &ro, &rt, true); 548 if (error != 0) 549 goto bad; 550 ifp = if_get_byindex(rt->rt_ifp->if_index, &psref); 551 } 552 553 if (rt == NULL) { 554 /* 555 * If in6_selectroute() does not return a route entry, 556 * dst may not have been updated. 557 */ 558 error = rtcache_setdst(ro, sin6tosa(&dst_sa)); 559 if (error) { 560 goto bad; 561 } 562 } 563 564 /* 565 * then rt (for unicast) and ifp must be non-NULL valid values. 566 */ 567 if ((flags & IPV6_FORWARDING) == 0) { 568 /* XXX: the FORWARDING flag can be set for mrouting. */ 569 in6_ifstat_inc(ifp, ifs6_out_request); 570 } 571 if (rt != NULL) { 572 ia = (struct in6_ifaddr *)(rt->rt_ifa); 573 rt->rt_use++; 574 } 575 576 /* 577 * The outgoing interface must be in the zone of source and 578 * destination addresses. We should use ia_ifp to support the 579 * case of sending packets to an address of our own. 580 */ 581 if (ia != NULL && ia->ia_ifp) { 582 origifp = ia->ia_ifp; 583 if (if_is_deactivated(origifp)) 584 goto bad; 585 if_acquire(origifp, &psref_ia); 586 release_psref_ia = true; 587 } else 588 origifp = ifp; 589 590 src0 = ip6->ip6_src; 591 if (in6_setscope(&src0, origifp, &zone)) 592 goto badscope; 593 sockaddr_in6_init(&src_sa, &ip6->ip6_src, 0, 0, 0); 594 if (sa6_recoverscope(&src_sa) || zone != src_sa.sin6_scope_id) 595 goto badscope; 596 597 dst0 = ip6->ip6_dst; 598 if (in6_setscope(&dst0, origifp, &zone)) 599 goto badscope; 600 /* re-initialize to be sure */ 601 sockaddr_in6_init(&dst_sa, &ip6->ip6_dst, 0, 0, 0); 602 if (sa6_recoverscope(&dst_sa) || zone != dst_sa.sin6_scope_id) 603 goto badscope; 604 605 /* scope check is done. */ 606 607 /* Ensure we only send from a valid address. */ 608 if ((error = ip6_ifaddrvalid(&src0)) != 0) { 609 char ip6buf[INET6_ADDRSTRLEN]; 610 nd6log(LOG_ERR, 611 "refusing to send from invalid address %s (pid %d)\n", 612 IN6_PRINT(ip6buf, &src0), curproc->p_pid); 613 IP6_STATINC(IP6_STAT_ODROPPED); 614 in6_ifstat_inc(origifp, ifs6_out_discard); 615 if (error == 1) 616 /* 617 * Address exists, but is tentative or detached. 618 * We can't send from it because it's invalid, 619 * so we drop the packet. 620 */ 621 error = 0; 622 else 623 error = EADDRNOTAVAIL; 624 goto bad; 625 } 626 627 if (rt != NULL && (rt->rt_flags & RTF_GATEWAY) && 628 !IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) 629 dst = satocsin6(rt->rt_gateway); 630 else 631 dst = satocsin6(rtcache_getdst(ro)); 632 633 /* 634 * XXXXXX: original code follows: 635 */ 636 if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) 637 m->m_flags &= ~(M_BCAST | M_MCAST); /* just in case */ 638 else { 639 struct in6_multi *in6m; 640 641 m->m_flags = (m->m_flags & ~M_BCAST) | M_MCAST; 642 643 in6_ifstat_inc(ifp, ifs6_out_mcast); 644 645 /* 646 * Confirm that the outgoing interface supports multicast. 647 */ 648 if (!(ifp->if_flags & IFF_MULTICAST)) { 649 IP6_STATINC(IP6_STAT_NOROUTE); 650 in6_ifstat_inc(ifp, ifs6_out_discard); 651 error = ENETUNREACH; 652 goto bad; 653 } 654 655 in6m = in6_lookup_multi(&ip6->ip6_dst, ifp); 656 if (in6m != NULL && 657 (im6o == NULL || im6o->im6o_multicast_loop)) { 658 /* 659 * If we belong to the destination multicast group 660 * on the outgoing interface, and the caller did not 661 * forbid loopback, loop back a copy. 662 */ 663 KASSERT(dst != NULL); 664 ip6_mloopback(ifp, m, dst); 665 } else { 666 /* 667 * If we are acting as a multicast router, perform 668 * multicast forwarding as if the packet had just 669 * arrived on the interface to which we are about 670 * to send. The multicast forwarding function 671 * recursively calls this function, using the 672 * IPV6_FORWARDING flag to prevent infinite recursion. 673 * 674 * Multicasts that are looped back by ip6_mloopback(), 675 * above, will be forwarded by the ip6_input() routine, 676 * if necessary. 677 */ 678 if (ip6_mrouter && (flags & IPV6_FORWARDING) == 0) { 679 if (ip6_mforward(ip6, ifp, m) != 0) { 680 m_freem(m); 681 goto done; 682 } 683 } 684 } 685 /* 686 * Multicasts with a hoplimit of zero may be looped back, 687 * above, but must not be transmitted on a network. 688 * Also, multicasts addressed to the loopback interface 689 * are not sent -- the above call to ip6_mloopback() will 690 * loop back a copy if this host actually belongs to the 691 * destination group on the loopback interface. 692 */ 693 if (ip6->ip6_hlim == 0 || (ifp->if_flags & IFF_LOOPBACK) || 694 IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst)) { 695 m_freem(m); 696 goto done; 697 } 698 } 699 700 /* 701 * Fill the outgoing inteface to tell the upper layer 702 * to increment per-interface statistics. 703 */ 704 if (ifpp) 705 *ifpp = ifp; 706 707 /* Determine path MTU. */ 708 /* 709 * ro_pmtu represent final destination while 710 * ro might represent immediate destination. 711 * Use ro_pmtu destination since MTU might differ. 712 */ 713 if (ro_pmtu != ro) { 714 union { 715 struct sockaddr dst; 716 struct sockaddr_in6 dst6; 717 } u; 718 719 /* ro_pmtu may not have a cache */ 720 sockaddr_in6_init(&u.dst6, &finaldst, 0, 0, 0); 721 rt_pmtu = rtcache_lookup(ro_pmtu, &u.dst); 722 } else 723 rt_pmtu = rt; 724 error = ip6_getpmtu(rt_pmtu, ifp, &mtu, &alwaysfrag); 725 if (rt_pmtu != NULL && rt_pmtu != rt) 726 rtcache_unref(rt_pmtu, ro_pmtu); 727 if (error != 0) 728 goto bad; 729 730 /* 731 * The caller of this function may specify to use the minimum MTU 732 * in some cases. 733 * An advanced API option (IPV6_USE_MIN_MTU) can also override MTU 734 * setting. The logic is a bit complicated; by default, unicast 735 * packets will follow path MTU while multicast packets will be sent at 736 * the minimum MTU. If IP6PO_MINMTU_ALL is specified, all packets 737 * including unicast ones will be sent at the minimum MTU. Multicast 738 * packets will always be sent at the minimum MTU unless 739 * IP6PO_MINMTU_DISABLE is explicitly specified. 740 * See RFC 3542 for more details. 741 */ 742 if (mtu > IPV6_MMTU) { 743 if ((flags & IPV6_MINMTU)) 744 mtu = IPV6_MMTU; 745 else if (opt && opt->ip6po_minmtu == IP6PO_MINMTU_ALL) 746 mtu = IPV6_MMTU; 747 else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) && 748 (opt == NULL || 749 opt->ip6po_minmtu != IP6PO_MINMTU_DISABLE)) { 750 mtu = IPV6_MMTU; 751 } 752 } 753 754 /* 755 * clear embedded scope identifiers if necessary. 756 * in6_clearscope will touch the addresses only when necessary. 757 */ 758 in6_clearscope(&ip6->ip6_src); 759 in6_clearscope(&ip6->ip6_dst); 760 761 /* 762 * If the outgoing packet contains a hop-by-hop options header, 763 * it must be examined and processed even by the source node. 764 * (RFC 2460, section 4.) 765 */ 766 if (ip6->ip6_nxt == IPV6_HOPOPTS) { 767 u_int32_t dummy1; /* XXX unused */ 768 u_int32_t dummy2; /* XXX unused */ 769 int hoff = sizeof(struct ip6_hdr); 770 771 if (ip6_hopopts_input(&dummy1, &dummy2, &m, &hoff)) { 772 /* m was already freed at this point */ 773 error = EINVAL;/* better error? */ 774 goto done; 775 } 776 777 ip6 = mtod(m, struct ip6_hdr *); 778 } 779 780 /* 781 * Run through list of hooks for output packets. 782 */ 783 if ((error = pfil_run_hooks(inet6_pfil_hook, &m, ifp, PFIL_OUT)) != 0) 784 goto done; 785 if (m == NULL) 786 goto done; 787 ip6 = mtod(m, struct ip6_hdr *); 788 789 /* 790 * Send the packet to the outgoing interface. 791 * If necessary, do IPv6 fragmentation before sending. 792 * 793 * the logic here is rather complex: 794 * 1: normal case (dontfrag == 0, alwaysfrag == 0) 795 * 1-a: send as is if tlen <= path mtu 796 * 1-b: fragment if tlen > path mtu 797 * 798 * 2: if user asks us not to fragment (dontfrag == 1) 799 * 2-a: send as is if tlen <= interface mtu 800 * 2-b: error if tlen > interface mtu 801 * 802 * 3: if we always need to attach fragment header (alwaysfrag == 1) 803 * always fragment 804 * 805 * 4: if dontfrag == 1 && alwaysfrag == 1 806 * error, as we cannot handle this conflicting request 807 */ 808 tlen = m->m_pkthdr.len; 809 tso = (m->m_pkthdr.csum_flags & M_CSUM_TSOv6) != 0; 810 if (opt && (opt->ip6po_flags & IP6PO_DONTFRAG)) 811 dontfrag = 1; 812 else 813 dontfrag = 0; 814 815 if (dontfrag && alwaysfrag) { /* case 4 */ 816 /* conflicting request - can't transmit */ 817 error = EMSGSIZE; 818 goto bad; 819 } 820 if (dontfrag && (!tso && tlen > IN6_LINKMTU(ifp))) { /* case 2-b */ 821 /* 822 * Even if the DONTFRAG option is specified, we cannot send the 823 * packet when the data length is larger than the MTU of the 824 * outgoing interface. 825 * Notify the error by sending IPV6_PATHMTU ancillary data as 826 * well as returning an error code (the latter is not described 827 * in the API spec.) 828 */ 829 u_int32_t mtu32; 830 struct ip6ctlparam ip6cp; 831 832 mtu32 = (u_int32_t)mtu; 833 memset(&ip6cp, 0, sizeof(ip6cp)); 834 ip6cp.ip6c_cmdarg = (void *)&mtu32; 835 pfctlinput2(PRC_MSGSIZE, 836 rtcache_getdst(ro_pmtu), &ip6cp); 837 838 error = EMSGSIZE; 839 goto bad; 840 } 841 842 /* 843 * transmit packet without fragmentation 844 */ 845 if (dontfrag || (!alwaysfrag && (tlen <= mtu || tso))) { 846 /* case 1-a and 2-a */ 847 struct in6_ifaddr *ia6; 848 int sw_csum; 849 int s; 850 851 ip6 = mtod(m, struct ip6_hdr *); 852 s = pserialize_read_enter(); 853 ia6 = in6_ifawithifp(ifp, &ip6->ip6_src); 854 if (ia6) { 855 /* Record statistics for this interface address. */ 856 ia6->ia_ifa.ifa_data.ifad_outbytes += m->m_pkthdr.len; 857 } 858 pserialize_read_exit(s); 859 860 sw_csum = m->m_pkthdr.csum_flags & ~ifp->if_csum_flags_tx; 861 if ((sw_csum & (M_CSUM_UDPv6|M_CSUM_TCPv6)) != 0) { 862 if (IN6_NEED_CHECKSUM(ifp, 863 sw_csum & (M_CSUM_UDPv6|M_CSUM_TCPv6))) { 864 in6_delayed_cksum(m); 865 } 866 m->m_pkthdr.csum_flags &= ~(M_CSUM_UDPv6|M_CSUM_TCPv6); 867 } 868 869 KASSERT(dst != NULL); 870 if (__predict_true(!tso || 871 (ifp->if_capenable & IFCAP_TSOv6) != 0)) { 872 error = ip6_if_output(ifp, origifp, m, dst, rt); 873 } else { 874 error = ip6_tso_output(ifp, origifp, m, dst, rt); 875 } 876 goto done; 877 } 878 879 if (tso) { 880 error = EINVAL; /* XXX */ 881 goto bad; 882 } 883 884 /* 885 * try to fragment the packet. case 1-b and 3 886 */ 887 if (mtu < IPV6_MMTU) { 888 /* path MTU cannot be less than IPV6_MMTU */ 889 error = EMSGSIZE; 890 in6_ifstat_inc(ifp, ifs6_out_fragfail); 891 goto bad; 892 } else if (ip6->ip6_plen == 0) { 893 /* jumbo payload cannot be fragmented */ 894 error = EMSGSIZE; 895 in6_ifstat_inc(ifp, ifs6_out_fragfail); 896 goto bad; 897 } else { 898 struct mbuf **mnext, *m_frgpart; 899 struct ip6_frag *ip6f; 900 u_int32_t id = htonl(ip6_randomid()); 901 u_char nextproto; 902 #if 0 /* see below */ 903 struct ip6ctlparam ip6cp; 904 u_int32_t mtu32; 905 #endif 906 907 /* 908 * Too large for the destination or interface; 909 * fragment if possible. 910 * Must be able to put at least 8 bytes per fragment. 911 */ 912 hlen = unfragpartlen; 913 if (mtu > IPV6_MAXPACKET) 914 mtu = IPV6_MAXPACKET; 915 916 #if 0 917 /* 918 * It is believed this code is a leftover from the 919 * development of the IPV6_RECVPATHMTU sockopt and 920 * associated work to implement RFC3542. 921 * It's not entirely clear what the intent of the API 922 * is at this point, so disable this code for now. 923 * The IPV6_RECVPATHMTU sockopt and/or IPV6_DONTFRAG 924 * will send notifications if the application requests. 925 */ 926 927 /* Notify a proper path MTU to applications. */ 928 mtu32 = (u_int32_t)mtu; 929 memset(&ip6cp, 0, sizeof(ip6cp)); 930 ip6cp.ip6c_cmdarg = (void *)&mtu32; 931 pfctlinput2(PRC_MSGSIZE, 932 rtcache_getdst(ro_pmtu), &ip6cp); 933 #endif 934 935 len = (mtu - hlen - sizeof(struct ip6_frag)) & ~7; 936 if (len < 8) { 937 error = EMSGSIZE; 938 in6_ifstat_inc(ifp, ifs6_out_fragfail); 939 goto bad; 940 } 941 942 mnext = &m->m_nextpkt; 943 944 /* 945 * Change the next header field of the last header in the 946 * unfragmentable part. 947 */ 948 if (exthdrs.ip6e_rthdr) { 949 nextproto = *mtod(exthdrs.ip6e_rthdr, u_char *); 950 *mtod(exthdrs.ip6e_rthdr, u_char *) = IPPROTO_FRAGMENT; 951 } else if (exthdrs.ip6e_dest1) { 952 nextproto = *mtod(exthdrs.ip6e_dest1, u_char *); 953 *mtod(exthdrs.ip6e_dest1, u_char *) = IPPROTO_FRAGMENT; 954 } else if (exthdrs.ip6e_hbh) { 955 nextproto = *mtod(exthdrs.ip6e_hbh, u_char *); 956 *mtod(exthdrs.ip6e_hbh, u_char *) = IPPROTO_FRAGMENT; 957 } else { 958 nextproto = ip6->ip6_nxt; 959 ip6->ip6_nxt = IPPROTO_FRAGMENT; 960 } 961 962 if ((m->m_pkthdr.csum_flags & (M_CSUM_UDPv6|M_CSUM_TCPv6)) 963 != 0) { 964 if (IN6_NEED_CHECKSUM(ifp, 965 m->m_pkthdr.csum_flags & 966 (M_CSUM_UDPv6|M_CSUM_TCPv6))) { 967 in6_delayed_cksum(m); 968 } 969 m->m_pkthdr.csum_flags &= ~(M_CSUM_UDPv6|M_CSUM_TCPv6); 970 } 971 972 /* 973 * Loop through length of segment after first fragment, 974 * make new header and copy data of each part and link onto 975 * chain. 976 */ 977 m0 = m; 978 for (off = hlen; off < tlen; off += len) { 979 struct mbuf *mlast; 980 981 MGETHDR(m, M_DONTWAIT, MT_HEADER); 982 if (!m) { 983 error = ENOBUFS; 984 IP6_STATINC(IP6_STAT_ODROPPED); 985 goto sendorfree; 986 } 987 m_reset_rcvif(m); 988 m->m_flags = m0->m_flags & M_COPYFLAGS; 989 *mnext = m; 990 mnext = &m->m_nextpkt; 991 m->m_data += max_linkhdr; 992 mhip6 = mtod(m, struct ip6_hdr *); 993 *mhip6 = *ip6; 994 m->m_len = sizeof(*mhip6); 995 /* 996 * ip6f must be valid if error is 0. But how 997 * can a compiler be expected to infer this? 998 */ 999 ip6f = NULL; 1000 error = ip6_insertfraghdr(m0, m, hlen, &ip6f); 1001 if (error) { 1002 IP6_STATINC(IP6_STAT_ODROPPED); 1003 goto sendorfree; 1004 } 1005 ip6f->ip6f_offlg = htons((u_int16_t)((off - hlen) & ~7)); 1006 if (off + len >= tlen) 1007 len = tlen - off; 1008 else 1009 ip6f->ip6f_offlg |= IP6F_MORE_FRAG; 1010 mhip6->ip6_plen = htons((u_int16_t)(len + hlen + 1011 sizeof(*ip6f) - sizeof(struct ip6_hdr))); 1012 if ((m_frgpart = m_copy(m0, off, len)) == 0) { 1013 error = ENOBUFS; 1014 IP6_STATINC(IP6_STAT_ODROPPED); 1015 goto sendorfree; 1016 } 1017 for (mlast = m; mlast->m_next; mlast = mlast->m_next) 1018 ; 1019 mlast->m_next = m_frgpart; 1020 m->m_pkthdr.len = len + hlen + sizeof(*ip6f); 1021 m_reset_rcvif(m); 1022 ip6f->ip6f_reserved = 0; 1023 ip6f->ip6f_ident = id; 1024 ip6f->ip6f_nxt = nextproto; 1025 IP6_STATINC(IP6_STAT_OFRAGMENTS); 1026 in6_ifstat_inc(ifp, ifs6_out_fragcreat); 1027 } 1028 1029 in6_ifstat_inc(ifp, ifs6_out_fragok); 1030 } 1031 1032 /* 1033 * Remove leading garbages. 1034 */ 1035 sendorfree: 1036 m = m0->m_nextpkt; 1037 m0->m_nextpkt = 0; 1038 m_freem(m0); 1039 for (m0 = m; m; m = m0) { 1040 m0 = m->m_nextpkt; 1041 m->m_nextpkt = 0; 1042 if (error == 0) { 1043 struct in6_ifaddr *ia6; 1044 int s; 1045 ip6 = mtod(m, struct ip6_hdr *); 1046 s = pserialize_read_enter(); 1047 ia6 = in6_ifawithifp(ifp, &ip6->ip6_src); 1048 if (ia6) { 1049 /* 1050 * Record statistics for this interface 1051 * address. 1052 */ 1053 ia6->ia_ifa.ifa_data.ifad_outbytes += 1054 m->m_pkthdr.len; 1055 } 1056 pserialize_read_exit(s); 1057 KASSERT(dst != NULL); 1058 error = ip6_if_output(ifp, origifp, m, dst, rt); 1059 } else 1060 m_freem(m); 1061 } 1062 1063 if (error == 0) 1064 IP6_STATINC(IP6_STAT_FRAGMENTED); 1065 1066 done: 1067 rtcache_unref(rt, ro); 1068 if (ro == &ip6route) 1069 rtcache_free(&ip6route); 1070 1071 #ifdef IPSEC 1072 if (sp != NULL) 1073 KEY_FREESP(&sp); 1074 #endif /* IPSEC */ 1075 1076 if_put(ifp, &psref); 1077 if (release_psref_ia) 1078 if_put(origifp, &psref_ia); 1079 curlwp_bindx(bound); 1080 1081 return (error); 1082 1083 freehdrs: 1084 m_freem(exthdrs.ip6e_hbh); /* m_freem will check if mbuf is 0 */ 1085 m_freem(exthdrs.ip6e_dest1); 1086 m_freem(exthdrs.ip6e_rthdr); 1087 m_freem(exthdrs.ip6e_dest2); 1088 /* FALLTHROUGH */ 1089 bad: 1090 m_freem(m); 1091 goto done; 1092 badscope: 1093 IP6_STATINC(IP6_STAT_BADSCOPE); 1094 in6_ifstat_inc(origifp, ifs6_out_discard); 1095 if (error == 0) 1096 error = EHOSTUNREACH; /* XXX */ 1097 goto bad; 1098 } 1099 1100 static int 1101 ip6_copyexthdr(struct mbuf **mp, void *hdr, int hlen) 1102 { 1103 struct mbuf *m; 1104 1105 if (hlen > MCLBYTES) 1106 return (ENOBUFS); /* XXX */ 1107 1108 MGET(m, M_DONTWAIT, MT_DATA); 1109 if (!m) 1110 return (ENOBUFS); 1111 1112 if (hlen > MLEN) { 1113 MCLGET(m, M_DONTWAIT); 1114 if ((m->m_flags & M_EXT) == 0) { 1115 m_free(m); 1116 return (ENOBUFS); 1117 } 1118 } 1119 m->m_len = hlen; 1120 if (hdr) 1121 bcopy(hdr, mtod(m, void *), hlen); 1122 1123 *mp = m; 1124 return (0); 1125 } 1126 1127 /* 1128 * Process a delayed payload checksum calculation. 1129 */ 1130 void 1131 in6_delayed_cksum(struct mbuf *m) 1132 { 1133 uint16_t csum, offset; 1134 1135 KASSERT((m->m_pkthdr.csum_flags & (M_CSUM_UDPv6|M_CSUM_TCPv6)) != 0); 1136 KASSERT((~m->m_pkthdr.csum_flags & (M_CSUM_UDPv6|M_CSUM_TCPv6)) != 0); 1137 KASSERT((m->m_pkthdr.csum_flags 1138 & (M_CSUM_UDPv4|M_CSUM_TCPv4|M_CSUM_TSOv4)) == 0); 1139 1140 offset = M_CSUM_DATA_IPv6_HL(m->m_pkthdr.csum_data); 1141 csum = in6_cksum(m, 0, offset, m->m_pkthdr.len - offset); 1142 if (csum == 0 && (m->m_pkthdr.csum_flags & M_CSUM_UDPv6) != 0) { 1143 csum = 0xffff; 1144 } 1145 1146 offset += M_CSUM_DATA_IPv6_OFFSET(m->m_pkthdr.csum_data); 1147 if ((offset + sizeof(csum)) > m->m_len) { 1148 m_copyback(m, offset, sizeof(csum), &csum); 1149 } else { 1150 *(uint16_t *)(mtod(m, char *) + offset) = csum; 1151 } 1152 } 1153 1154 /* 1155 * Insert jumbo payload option. 1156 */ 1157 static int 1158 ip6_insert_jumboopt(struct ip6_exthdrs *exthdrs, u_int32_t plen) 1159 { 1160 struct mbuf *mopt; 1161 u_int8_t *optbuf; 1162 u_int32_t v; 1163 1164 #define JUMBOOPTLEN 8 /* length of jumbo payload option and padding */ 1165 1166 /* 1167 * If there is no hop-by-hop options header, allocate new one. 1168 * If there is one but it doesn't have enough space to store the 1169 * jumbo payload option, allocate a cluster to store the whole options. 1170 * Otherwise, use it to store the options. 1171 */ 1172 if (exthdrs->ip6e_hbh == 0) { 1173 MGET(mopt, M_DONTWAIT, MT_DATA); 1174 if (mopt == 0) 1175 return (ENOBUFS); 1176 mopt->m_len = JUMBOOPTLEN; 1177 optbuf = mtod(mopt, u_int8_t *); 1178 optbuf[1] = 0; /* = ((JUMBOOPTLEN) >> 3) - 1 */ 1179 exthdrs->ip6e_hbh = mopt; 1180 } else { 1181 struct ip6_hbh *hbh; 1182 1183 mopt = exthdrs->ip6e_hbh; 1184 if (M_TRAILINGSPACE(mopt) < JUMBOOPTLEN) { 1185 /* 1186 * XXX assumption: 1187 * - exthdrs->ip6e_hbh is not referenced from places 1188 * other than exthdrs. 1189 * - exthdrs->ip6e_hbh is not an mbuf chain. 1190 */ 1191 int oldoptlen = mopt->m_len; 1192 struct mbuf *n; 1193 1194 /* 1195 * XXX: give up if the whole (new) hbh header does 1196 * not fit even in an mbuf cluster. 1197 */ 1198 if (oldoptlen + JUMBOOPTLEN > MCLBYTES) 1199 return (ENOBUFS); 1200 1201 /* 1202 * As a consequence, we must always prepare a cluster 1203 * at this point. 1204 */ 1205 MGET(n, M_DONTWAIT, MT_DATA); 1206 if (n) { 1207 MCLGET(n, M_DONTWAIT); 1208 if ((n->m_flags & M_EXT) == 0) { 1209 m_freem(n); 1210 n = NULL; 1211 } 1212 } 1213 if (!n) 1214 return (ENOBUFS); 1215 n->m_len = oldoptlen + JUMBOOPTLEN; 1216 bcopy(mtod(mopt, void *), mtod(n, void *), 1217 oldoptlen); 1218 optbuf = mtod(n, u_int8_t *) + oldoptlen; 1219 m_freem(mopt); 1220 mopt = exthdrs->ip6e_hbh = n; 1221 } else { 1222 optbuf = mtod(mopt, u_int8_t *) + mopt->m_len; 1223 mopt->m_len += JUMBOOPTLEN; 1224 } 1225 optbuf[0] = IP6OPT_PADN; 1226 optbuf[1] = 0; 1227 1228 /* 1229 * Adjust the header length according to the pad and 1230 * the jumbo payload option. 1231 */ 1232 hbh = mtod(mopt, struct ip6_hbh *); 1233 hbh->ip6h_len += (JUMBOOPTLEN >> 3); 1234 } 1235 1236 /* fill in the option. */ 1237 optbuf[2] = IP6OPT_JUMBO; 1238 optbuf[3] = 4; 1239 v = (u_int32_t)htonl(plen + JUMBOOPTLEN); 1240 bcopy(&v, &optbuf[4], sizeof(u_int32_t)); 1241 1242 /* finally, adjust the packet header length */ 1243 exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN; 1244 1245 return (0); 1246 #undef JUMBOOPTLEN 1247 } 1248 1249 /* 1250 * Insert fragment header and copy unfragmentable header portions. 1251 * 1252 * *frghdrp will not be read, and it is guaranteed that either an 1253 * error is returned or that *frghdrp will point to space allocated 1254 * for the fragment header. 1255 */ 1256 static int 1257 ip6_insertfraghdr(struct mbuf *m0, struct mbuf *m, int hlen, 1258 struct ip6_frag **frghdrp) 1259 { 1260 struct mbuf *n, *mlast; 1261 1262 if (hlen > sizeof(struct ip6_hdr)) { 1263 n = m_copym(m0, sizeof(struct ip6_hdr), 1264 hlen - sizeof(struct ip6_hdr), M_DONTWAIT); 1265 if (n == 0) 1266 return (ENOBUFS); 1267 m->m_next = n; 1268 } else 1269 n = m; 1270 1271 /* Search for the last mbuf of unfragmentable part. */ 1272 for (mlast = n; mlast->m_next; mlast = mlast->m_next) 1273 ; 1274 1275 if ((mlast->m_flags & M_EXT) == 0 && 1276 M_TRAILINGSPACE(mlast) >= sizeof(struct ip6_frag)) { 1277 /* use the trailing space of the last mbuf for the fragment hdr */ 1278 *frghdrp = (struct ip6_frag *)(mtod(mlast, char *) + 1279 mlast->m_len); 1280 mlast->m_len += sizeof(struct ip6_frag); 1281 m->m_pkthdr.len += sizeof(struct ip6_frag); 1282 } else { 1283 /* allocate a new mbuf for the fragment header */ 1284 struct mbuf *mfrg; 1285 1286 MGET(mfrg, M_DONTWAIT, MT_DATA); 1287 if (mfrg == 0) 1288 return (ENOBUFS); 1289 mfrg->m_len = sizeof(struct ip6_frag); 1290 *frghdrp = mtod(mfrg, struct ip6_frag *); 1291 mlast->m_next = mfrg; 1292 } 1293 1294 return (0); 1295 } 1296 1297 static int 1298 ip6_getpmtu(struct rtentry *rt, struct ifnet *ifp, u_long *mtup, 1299 int *alwaysfragp) 1300 { 1301 u_int32_t mtu = 0; 1302 int alwaysfrag = 0; 1303 int error = 0; 1304 1305 if (rt != NULL) { 1306 u_int32_t ifmtu; 1307 1308 if (ifp == NULL) 1309 ifp = rt->rt_ifp; 1310 ifmtu = IN6_LINKMTU(ifp); 1311 mtu = rt->rt_rmx.rmx_mtu; 1312 if (mtu == 0) 1313 mtu = ifmtu; 1314 else if (mtu < IPV6_MMTU) { 1315 /* 1316 * RFC2460 section 5, last paragraph: 1317 * if we record ICMPv6 too big message with 1318 * mtu < IPV6_MMTU, transmit packets sized IPV6_MMTU 1319 * or smaller, with fragment header attached. 1320 * (fragment header is needed regardless from the 1321 * packet size, for translators to identify packets) 1322 */ 1323 alwaysfrag = 1; 1324 mtu = IPV6_MMTU; 1325 } else if (mtu > ifmtu) { 1326 /* 1327 * The MTU on the route is larger than the MTU on 1328 * the interface! This shouldn't happen, unless the 1329 * MTU of the interface has been changed after the 1330 * interface was brought up. Change the MTU in the 1331 * route to match the interface MTU (as long as the 1332 * field isn't locked). 1333 */ 1334 mtu = ifmtu; 1335 if (!(rt->rt_rmx.rmx_locks & RTV_MTU)) 1336 rt->rt_rmx.rmx_mtu = mtu; 1337 } 1338 } else if (ifp) { 1339 mtu = IN6_LINKMTU(ifp); 1340 } else 1341 error = EHOSTUNREACH; /* XXX */ 1342 1343 *mtup = mtu; 1344 if (alwaysfragp) 1345 *alwaysfragp = alwaysfrag; 1346 return (error); 1347 } 1348 1349 /* 1350 * IP6 socket option processing. 1351 */ 1352 int 1353 ip6_ctloutput(int op, struct socket *so, struct sockopt *sopt) 1354 { 1355 int optdatalen, uproto; 1356 void *optdata; 1357 struct in6pcb *in6p = sotoin6pcb(so); 1358 struct ip_moptions **mopts; 1359 int error, optval; 1360 int level, optname; 1361 1362 KASSERT(solocked(so)); 1363 KASSERT(sopt != NULL); 1364 1365 level = sopt->sopt_level; 1366 optname = sopt->sopt_name; 1367 1368 error = optval = 0; 1369 uproto = (int)so->so_proto->pr_protocol; 1370 1371 switch (level) { 1372 case IPPROTO_IP: 1373 switch (optname) { 1374 case IP_ADD_MEMBERSHIP: 1375 case IP_DROP_MEMBERSHIP: 1376 case IP_MULTICAST_IF: 1377 case IP_MULTICAST_LOOP: 1378 case IP_MULTICAST_TTL: 1379 mopts = &in6p->in6p_v4moptions; 1380 switch (op) { 1381 case PRCO_GETOPT: 1382 return ip_getmoptions(*mopts, sopt); 1383 case PRCO_SETOPT: 1384 return ip_setmoptions(mopts, sopt); 1385 default: 1386 return EINVAL; 1387 } 1388 default: 1389 return ENOPROTOOPT; 1390 } 1391 case IPPROTO_IPV6: 1392 break; 1393 default: 1394 return ENOPROTOOPT; 1395 } 1396 switch (op) { 1397 case PRCO_SETOPT: 1398 switch (optname) { 1399 #ifdef RFC2292 1400 case IPV6_2292PKTOPTIONS: 1401 error = ip6_pcbopts(&in6p->in6p_outputopts, so, sopt); 1402 break; 1403 #endif 1404 1405 /* 1406 * Use of some Hop-by-Hop options or some 1407 * Destination options, might require special 1408 * privilege. That is, normal applications 1409 * (without special privilege) might be forbidden 1410 * from setting certain options in outgoing packets, 1411 * and might never see certain options in received 1412 * packets. [RFC 2292 Section 6] 1413 * KAME specific note: 1414 * KAME prevents non-privileged users from sending or 1415 * receiving ANY hbh/dst options in order to avoid 1416 * overhead of parsing options in the kernel. 1417 */ 1418 case IPV6_RECVHOPOPTS: 1419 case IPV6_RECVDSTOPTS: 1420 case IPV6_RECVRTHDRDSTOPTS: 1421 error = kauth_authorize_network(kauth_cred_get(), 1422 KAUTH_NETWORK_IPV6, KAUTH_REQ_NETWORK_IPV6_HOPBYHOP, 1423 NULL, NULL, NULL); 1424 if (error) 1425 break; 1426 /* FALLTHROUGH */ 1427 case IPV6_UNICAST_HOPS: 1428 case IPV6_HOPLIMIT: 1429 case IPV6_FAITH: 1430 1431 case IPV6_RECVPKTINFO: 1432 case IPV6_RECVHOPLIMIT: 1433 case IPV6_RECVRTHDR: 1434 case IPV6_RECVPATHMTU: 1435 case IPV6_RECVTCLASS: 1436 case IPV6_V6ONLY: 1437 error = sockopt_getint(sopt, &optval); 1438 if (error) 1439 break; 1440 switch (optname) { 1441 case IPV6_UNICAST_HOPS: 1442 if (optval < -1 || optval >= 256) 1443 error = EINVAL; 1444 else { 1445 /* -1 = kernel default */ 1446 in6p->in6p_hops = optval; 1447 } 1448 break; 1449 #define OPTSET(bit) \ 1450 do { \ 1451 if (optval) \ 1452 in6p->in6p_flags |= (bit); \ 1453 else \ 1454 in6p->in6p_flags &= ~(bit); \ 1455 } while (/*CONSTCOND*/ 0) 1456 1457 #ifdef RFC2292 1458 #define OPTSET2292(bit) \ 1459 do { \ 1460 in6p->in6p_flags |= IN6P_RFC2292; \ 1461 if (optval) \ 1462 in6p->in6p_flags |= (bit); \ 1463 else \ 1464 in6p->in6p_flags &= ~(bit); \ 1465 } while (/*CONSTCOND*/ 0) 1466 #endif 1467 1468 #define OPTBIT(bit) (in6p->in6p_flags & (bit) ? 1 : 0) 1469 1470 case IPV6_RECVPKTINFO: 1471 #ifdef RFC2292 1472 /* cannot mix with RFC2292 */ 1473 if (OPTBIT(IN6P_RFC2292)) { 1474 error = EINVAL; 1475 break; 1476 } 1477 #endif 1478 OPTSET(IN6P_PKTINFO); 1479 break; 1480 1481 case IPV6_HOPLIMIT: 1482 { 1483 struct ip6_pktopts **optp; 1484 1485 #ifdef RFC2292 1486 /* cannot mix with RFC2292 */ 1487 if (OPTBIT(IN6P_RFC2292)) { 1488 error = EINVAL; 1489 break; 1490 } 1491 #endif 1492 optp = &in6p->in6p_outputopts; 1493 error = ip6_pcbopt(IPV6_HOPLIMIT, 1494 (u_char *)&optval, 1495 sizeof(optval), 1496 optp, 1497 kauth_cred_get(), uproto); 1498 break; 1499 } 1500 1501 case IPV6_RECVHOPLIMIT: 1502 #ifdef RFC2292 1503 /* cannot mix with RFC2292 */ 1504 if (OPTBIT(IN6P_RFC2292)) { 1505 error = EINVAL; 1506 break; 1507 } 1508 #endif 1509 OPTSET(IN6P_HOPLIMIT); 1510 break; 1511 1512 case IPV6_RECVHOPOPTS: 1513 #ifdef RFC2292 1514 /* cannot mix with RFC2292 */ 1515 if (OPTBIT(IN6P_RFC2292)) { 1516 error = EINVAL; 1517 break; 1518 } 1519 #endif 1520 OPTSET(IN6P_HOPOPTS); 1521 break; 1522 1523 case IPV6_RECVDSTOPTS: 1524 #ifdef RFC2292 1525 /* cannot mix with RFC2292 */ 1526 if (OPTBIT(IN6P_RFC2292)) { 1527 error = EINVAL; 1528 break; 1529 } 1530 #endif 1531 OPTSET(IN6P_DSTOPTS); 1532 break; 1533 1534 case IPV6_RECVRTHDRDSTOPTS: 1535 #ifdef RFC2292 1536 /* cannot mix with RFC2292 */ 1537 if (OPTBIT(IN6P_RFC2292)) { 1538 error = EINVAL; 1539 break; 1540 } 1541 #endif 1542 OPTSET(IN6P_RTHDRDSTOPTS); 1543 break; 1544 1545 case IPV6_RECVRTHDR: 1546 #ifdef RFC2292 1547 /* cannot mix with RFC2292 */ 1548 if (OPTBIT(IN6P_RFC2292)) { 1549 error = EINVAL; 1550 break; 1551 } 1552 #endif 1553 OPTSET(IN6P_RTHDR); 1554 break; 1555 1556 case IPV6_FAITH: 1557 OPTSET(IN6P_FAITH); 1558 break; 1559 1560 case IPV6_RECVPATHMTU: 1561 /* 1562 * We ignore this option for TCP 1563 * sockets. 1564 * (RFC3542 leaves this case 1565 * unspecified.) 1566 */ 1567 if (uproto != IPPROTO_TCP) 1568 OPTSET(IN6P_MTU); 1569 break; 1570 1571 case IPV6_V6ONLY: 1572 /* 1573 * make setsockopt(IPV6_V6ONLY) 1574 * available only prior to bind(2). 1575 * see ipng mailing list, Jun 22 2001. 1576 */ 1577 if (in6p->in6p_lport || 1578 !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr)) { 1579 error = EINVAL; 1580 break; 1581 } 1582 #ifdef INET6_BINDV6ONLY 1583 if (!optval) 1584 error = EINVAL; 1585 #else 1586 OPTSET(IN6P_IPV6_V6ONLY); 1587 #endif 1588 break; 1589 case IPV6_RECVTCLASS: 1590 #ifdef RFC2292 1591 /* cannot mix with RFC2292 XXX */ 1592 if (OPTBIT(IN6P_RFC2292)) { 1593 error = EINVAL; 1594 break; 1595 } 1596 #endif 1597 OPTSET(IN6P_TCLASS); 1598 break; 1599 1600 } 1601 break; 1602 1603 case IPV6_OTCLASS: 1604 { 1605 struct ip6_pktopts **optp; 1606 u_int8_t tclass; 1607 1608 error = sockopt_get(sopt, &tclass, sizeof(tclass)); 1609 if (error) 1610 break; 1611 optp = &in6p->in6p_outputopts; 1612 error = ip6_pcbopt(optname, 1613 (u_char *)&tclass, 1614 sizeof(tclass), 1615 optp, 1616 kauth_cred_get(), uproto); 1617 break; 1618 } 1619 1620 case IPV6_TCLASS: 1621 case IPV6_DONTFRAG: 1622 case IPV6_USE_MIN_MTU: 1623 case IPV6_PREFER_TEMPADDR: 1624 error = sockopt_getint(sopt, &optval); 1625 if (error) 1626 break; 1627 { 1628 struct ip6_pktopts **optp; 1629 optp = &in6p->in6p_outputopts; 1630 error = ip6_pcbopt(optname, 1631 (u_char *)&optval, 1632 sizeof(optval), 1633 optp, 1634 kauth_cred_get(), uproto); 1635 break; 1636 } 1637 1638 #ifdef RFC2292 1639 case IPV6_2292PKTINFO: 1640 case IPV6_2292HOPLIMIT: 1641 case IPV6_2292HOPOPTS: 1642 case IPV6_2292DSTOPTS: 1643 case IPV6_2292RTHDR: 1644 /* RFC 2292 */ 1645 error = sockopt_getint(sopt, &optval); 1646 if (error) 1647 break; 1648 1649 switch (optname) { 1650 case IPV6_2292PKTINFO: 1651 OPTSET2292(IN6P_PKTINFO); 1652 break; 1653 case IPV6_2292HOPLIMIT: 1654 OPTSET2292(IN6P_HOPLIMIT); 1655 break; 1656 case IPV6_2292HOPOPTS: 1657 /* 1658 * Check super-user privilege. 1659 * See comments for IPV6_RECVHOPOPTS. 1660 */ 1661 error = 1662 kauth_authorize_network(kauth_cred_get(), 1663 KAUTH_NETWORK_IPV6, 1664 KAUTH_REQ_NETWORK_IPV6_HOPBYHOP, NULL, 1665 NULL, NULL); 1666 if (error) 1667 return (error); 1668 OPTSET2292(IN6P_HOPOPTS); 1669 break; 1670 case IPV6_2292DSTOPTS: 1671 error = 1672 kauth_authorize_network(kauth_cred_get(), 1673 KAUTH_NETWORK_IPV6, 1674 KAUTH_REQ_NETWORK_IPV6_HOPBYHOP, NULL, 1675 NULL, NULL); 1676 if (error) 1677 return (error); 1678 OPTSET2292(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); /* XXX */ 1679 break; 1680 case IPV6_2292RTHDR: 1681 OPTSET2292(IN6P_RTHDR); 1682 break; 1683 } 1684 break; 1685 #endif 1686 case IPV6_PKTINFO: 1687 case IPV6_HOPOPTS: 1688 case IPV6_RTHDR: 1689 case IPV6_DSTOPTS: 1690 case IPV6_RTHDRDSTOPTS: 1691 case IPV6_NEXTHOP: { 1692 /* new advanced API (RFC3542) */ 1693 void *optbuf; 1694 int optbuflen; 1695 struct ip6_pktopts **optp; 1696 1697 #ifdef RFC2292 1698 /* cannot mix with RFC2292 */ 1699 if (OPTBIT(IN6P_RFC2292)) { 1700 error = EINVAL; 1701 break; 1702 } 1703 #endif 1704 1705 optbuflen = sopt->sopt_size; 1706 optbuf = malloc(optbuflen, M_IP6OPT, M_NOWAIT); 1707 if (optbuf == NULL) { 1708 error = ENOBUFS; 1709 break; 1710 } 1711 1712 error = sockopt_get(sopt, optbuf, optbuflen); 1713 if (error) { 1714 free(optbuf, M_IP6OPT); 1715 break; 1716 } 1717 optp = &in6p->in6p_outputopts; 1718 error = ip6_pcbopt(optname, optbuf, optbuflen, 1719 optp, kauth_cred_get(), uproto); 1720 1721 free(optbuf, M_IP6OPT); 1722 break; 1723 } 1724 #undef OPTSET 1725 1726 case IPV6_MULTICAST_IF: 1727 case IPV6_MULTICAST_HOPS: 1728 case IPV6_MULTICAST_LOOP: 1729 case IPV6_JOIN_GROUP: 1730 case IPV6_LEAVE_GROUP: 1731 error = ip6_setmoptions(sopt, in6p); 1732 break; 1733 1734 case IPV6_PORTRANGE: 1735 error = sockopt_getint(sopt, &optval); 1736 if (error) 1737 break; 1738 1739 switch (optval) { 1740 case IPV6_PORTRANGE_DEFAULT: 1741 in6p->in6p_flags &= ~(IN6P_LOWPORT); 1742 in6p->in6p_flags &= ~(IN6P_HIGHPORT); 1743 break; 1744 1745 case IPV6_PORTRANGE_HIGH: 1746 in6p->in6p_flags &= ~(IN6P_LOWPORT); 1747 in6p->in6p_flags |= IN6P_HIGHPORT; 1748 break; 1749 1750 case IPV6_PORTRANGE_LOW: 1751 in6p->in6p_flags &= ~(IN6P_HIGHPORT); 1752 in6p->in6p_flags |= IN6P_LOWPORT; 1753 break; 1754 1755 default: 1756 error = EINVAL; 1757 break; 1758 } 1759 break; 1760 1761 case IPV6_PORTALGO: 1762 error = sockopt_getint(sopt, &optval); 1763 if (error) 1764 break; 1765 1766 error = portalgo_algo_index_select( 1767 (struct inpcb_hdr *)in6p, optval); 1768 break; 1769 1770 #if defined(IPSEC) 1771 case IPV6_IPSEC_POLICY: 1772 if (ipsec_enabled) { 1773 error = ipsec6_set_policy(in6p, optname, 1774 sopt->sopt_data, sopt->sopt_size, 1775 kauth_cred_get()); 1776 break; 1777 } 1778 /*FALLTHROUGH*/ 1779 #endif /* IPSEC */ 1780 1781 default: 1782 error = ENOPROTOOPT; 1783 break; 1784 } 1785 break; 1786 1787 case PRCO_GETOPT: 1788 switch (optname) { 1789 #ifdef RFC2292 1790 case IPV6_2292PKTOPTIONS: 1791 /* 1792 * RFC3542 (effectively) deprecated the 1793 * semantics of the 2292-style pktoptions. 1794 * Since it was not reliable in nature (i.e., 1795 * applications had to expect the lack of some 1796 * information after all), it would make sense 1797 * to simplify this part by always returning 1798 * empty data. 1799 */ 1800 break; 1801 #endif 1802 1803 case IPV6_RECVHOPOPTS: 1804 case IPV6_RECVDSTOPTS: 1805 case IPV6_RECVRTHDRDSTOPTS: 1806 case IPV6_UNICAST_HOPS: 1807 case IPV6_RECVPKTINFO: 1808 case IPV6_RECVHOPLIMIT: 1809 case IPV6_RECVRTHDR: 1810 case IPV6_RECVPATHMTU: 1811 1812 case IPV6_FAITH: 1813 case IPV6_V6ONLY: 1814 case IPV6_PORTRANGE: 1815 case IPV6_RECVTCLASS: 1816 switch (optname) { 1817 1818 case IPV6_RECVHOPOPTS: 1819 optval = OPTBIT(IN6P_HOPOPTS); 1820 break; 1821 1822 case IPV6_RECVDSTOPTS: 1823 optval = OPTBIT(IN6P_DSTOPTS); 1824 break; 1825 1826 case IPV6_RECVRTHDRDSTOPTS: 1827 optval = OPTBIT(IN6P_RTHDRDSTOPTS); 1828 break; 1829 1830 case IPV6_UNICAST_HOPS: 1831 optval = in6p->in6p_hops; 1832 break; 1833 1834 case IPV6_RECVPKTINFO: 1835 optval = OPTBIT(IN6P_PKTINFO); 1836 break; 1837 1838 case IPV6_RECVHOPLIMIT: 1839 optval = OPTBIT(IN6P_HOPLIMIT); 1840 break; 1841 1842 case IPV6_RECVRTHDR: 1843 optval = OPTBIT(IN6P_RTHDR); 1844 break; 1845 1846 case IPV6_RECVPATHMTU: 1847 optval = OPTBIT(IN6P_MTU); 1848 break; 1849 1850 case IPV6_FAITH: 1851 optval = OPTBIT(IN6P_FAITH); 1852 break; 1853 1854 case IPV6_V6ONLY: 1855 optval = OPTBIT(IN6P_IPV6_V6ONLY); 1856 break; 1857 1858 case IPV6_PORTRANGE: 1859 { 1860 int flags; 1861 flags = in6p->in6p_flags; 1862 if (flags & IN6P_HIGHPORT) 1863 optval = IPV6_PORTRANGE_HIGH; 1864 else if (flags & IN6P_LOWPORT) 1865 optval = IPV6_PORTRANGE_LOW; 1866 else 1867 optval = 0; 1868 break; 1869 } 1870 case IPV6_RECVTCLASS: 1871 optval = OPTBIT(IN6P_TCLASS); 1872 break; 1873 1874 } 1875 if (error) 1876 break; 1877 error = sockopt_setint(sopt, optval); 1878 break; 1879 1880 case IPV6_PATHMTU: 1881 { 1882 u_long pmtu = 0; 1883 struct ip6_mtuinfo mtuinfo; 1884 struct route *ro = &in6p->in6p_route; 1885 struct rtentry *rt; 1886 union { 1887 struct sockaddr dst; 1888 struct sockaddr_in6 dst6; 1889 } u; 1890 1891 if (!(so->so_state & SS_ISCONNECTED)) 1892 return (ENOTCONN); 1893 /* 1894 * XXX: we dot not consider the case of source 1895 * routing, or optional information to specify 1896 * the outgoing interface. 1897 */ 1898 sockaddr_in6_init(&u.dst6, &in6p->in6p_faddr, 0, 0, 0); 1899 rt = rtcache_lookup(ro, &u.dst); 1900 error = ip6_getpmtu(rt, NULL, &pmtu, NULL); 1901 rtcache_unref(rt, ro); 1902 if (error) 1903 break; 1904 if (pmtu > IPV6_MAXPACKET) 1905 pmtu = IPV6_MAXPACKET; 1906 1907 memset(&mtuinfo, 0, sizeof(mtuinfo)); 1908 mtuinfo.ip6m_mtu = (u_int32_t)pmtu; 1909 optdata = (void *)&mtuinfo; 1910 optdatalen = sizeof(mtuinfo); 1911 if (optdatalen > MCLBYTES) 1912 return (EMSGSIZE); /* XXX */ 1913 error = sockopt_set(sopt, optdata, optdatalen); 1914 break; 1915 } 1916 1917 #ifdef RFC2292 1918 case IPV6_2292PKTINFO: 1919 case IPV6_2292HOPLIMIT: 1920 case IPV6_2292HOPOPTS: 1921 case IPV6_2292RTHDR: 1922 case IPV6_2292DSTOPTS: 1923 switch (optname) { 1924 case IPV6_2292PKTINFO: 1925 optval = OPTBIT(IN6P_PKTINFO); 1926 break; 1927 case IPV6_2292HOPLIMIT: 1928 optval = OPTBIT(IN6P_HOPLIMIT); 1929 break; 1930 case IPV6_2292HOPOPTS: 1931 optval = OPTBIT(IN6P_HOPOPTS); 1932 break; 1933 case IPV6_2292RTHDR: 1934 optval = OPTBIT(IN6P_RTHDR); 1935 break; 1936 case IPV6_2292DSTOPTS: 1937 optval = OPTBIT(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); 1938 break; 1939 } 1940 error = sockopt_setint(sopt, optval); 1941 break; 1942 #endif 1943 case IPV6_PKTINFO: 1944 case IPV6_HOPOPTS: 1945 case IPV6_RTHDR: 1946 case IPV6_DSTOPTS: 1947 case IPV6_RTHDRDSTOPTS: 1948 case IPV6_NEXTHOP: 1949 case IPV6_OTCLASS: 1950 case IPV6_TCLASS: 1951 case IPV6_DONTFRAG: 1952 case IPV6_USE_MIN_MTU: 1953 case IPV6_PREFER_TEMPADDR: 1954 error = ip6_getpcbopt(in6p->in6p_outputopts, 1955 optname, sopt); 1956 break; 1957 1958 case IPV6_MULTICAST_IF: 1959 case IPV6_MULTICAST_HOPS: 1960 case IPV6_MULTICAST_LOOP: 1961 case IPV6_JOIN_GROUP: 1962 case IPV6_LEAVE_GROUP: 1963 error = ip6_getmoptions(sopt, in6p); 1964 break; 1965 1966 case IPV6_PORTALGO: 1967 optval = ((struct inpcb_hdr *)in6p)->inph_portalgo; 1968 error = sockopt_setint(sopt, optval); 1969 break; 1970 1971 #if defined(IPSEC) 1972 case IPV6_IPSEC_POLICY: 1973 if (ipsec_used) { 1974 struct mbuf *m = NULL; 1975 1976 /* 1977 * XXX: this will return EINVAL as sopt is 1978 * empty 1979 */ 1980 error = ipsec6_get_policy(in6p, sopt->sopt_data, 1981 sopt->sopt_size, &m); 1982 if (!error) 1983 error = sockopt_setmbuf(sopt, m); 1984 break; 1985 } 1986 /*FALLTHROUGH*/ 1987 #endif /* IPSEC */ 1988 1989 default: 1990 error = ENOPROTOOPT; 1991 break; 1992 } 1993 break; 1994 } 1995 return (error); 1996 } 1997 1998 int 1999 ip6_raw_ctloutput(int op, struct socket *so, struct sockopt *sopt) 2000 { 2001 int error = 0, optval; 2002 const int icmp6off = offsetof(struct icmp6_hdr, icmp6_cksum); 2003 struct in6pcb *in6p = sotoin6pcb(so); 2004 int level, optname; 2005 2006 KASSERT(sopt != NULL); 2007 2008 level = sopt->sopt_level; 2009 optname = sopt->sopt_name; 2010 2011 if (level != IPPROTO_IPV6) { 2012 return ENOPROTOOPT; 2013 } 2014 2015 switch (optname) { 2016 case IPV6_CHECKSUM: 2017 /* 2018 * For ICMPv6 sockets, no modification allowed for checksum 2019 * offset, permit "no change" values to help existing apps. 2020 * 2021 * XXX RFC3542 says: "An attempt to set IPV6_CHECKSUM 2022 * for an ICMPv6 socket will fail." The current 2023 * behavior does not meet RFC3542. 2024 */ 2025 switch (op) { 2026 case PRCO_SETOPT: 2027 error = sockopt_getint(sopt, &optval); 2028 if (error) 2029 break; 2030 if ((optval % 2) != 0) { 2031 /* the API assumes even offset values */ 2032 error = EINVAL; 2033 } else if (so->so_proto->pr_protocol == 2034 IPPROTO_ICMPV6) { 2035 if (optval != icmp6off) 2036 error = EINVAL; 2037 } else 2038 in6p->in6p_cksum = optval; 2039 break; 2040 2041 case PRCO_GETOPT: 2042 if (so->so_proto->pr_protocol == IPPROTO_ICMPV6) 2043 optval = icmp6off; 2044 else 2045 optval = in6p->in6p_cksum; 2046 2047 error = sockopt_setint(sopt, optval); 2048 break; 2049 2050 default: 2051 error = EINVAL; 2052 break; 2053 } 2054 break; 2055 2056 default: 2057 error = ENOPROTOOPT; 2058 break; 2059 } 2060 2061 return (error); 2062 } 2063 2064 #ifdef RFC2292 2065 /* 2066 * Set up IP6 options in pcb for insertion in output packets or 2067 * specifying behavior of outgoing packets. 2068 */ 2069 static int 2070 ip6_pcbopts(struct ip6_pktopts **pktopt, struct socket *so, 2071 struct sockopt *sopt) 2072 { 2073 struct ip6_pktopts *opt = *pktopt; 2074 struct mbuf *m; 2075 int error = 0; 2076 2077 KASSERT(solocked(so)); 2078 2079 /* turn off any old options. */ 2080 if (opt) { 2081 #ifdef DIAGNOSTIC 2082 if (opt->ip6po_pktinfo || opt->ip6po_nexthop || 2083 opt->ip6po_hbh || opt->ip6po_dest1 || opt->ip6po_dest2 || 2084 opt->ip6po_rhinfo.ip6po_rhi_rthdr) 2085 printf("ip6_pcbopts: all specified options are cleared.\n"); 2086 #endif 2087 ip6_clearpktopts(opt, -1); 2088 } else { 2089 opt = malloc(sizeof(*opt), M_IP6OPT, M_NOWAIT); 2090 if (opt == NULL) 2091 return (ENOBUFS); 2092 } 2093 *pktopt = NULL; 2094 2095 if (sopt == NULL || sopt->sopt_size == 0) { 2096 /* 2097 * Only turning off any previous options, regardless of 2098 * whether the opt is just created or given. 2099 */ 2100 free(opt, M_IP6OPT); 2101 return (0); 2102 } 2103 2104 /* set options specified by user. */ 2105 m = sockopt_getmbuf(sopt); 2106 if (m == NULL) { 2107 free(opt, M_IP6OPT); 2108 return (ENOBUFS); 2109 } 2110 2111 error = ip6_setpktopts(m, opt, NULL, kauth_cred_get(), 2112 so->so_proto->pr_protocol); 2113 m_freem(m); 2114 if (error != 0) { 2115 ip6_clearpktopts(opt, -1); /* XXX: discard all options */ 2116 free(opt, M_IP6OPT); 2117 return (error); 2118 } 2119 *pktopt = opt; 2120 return (0); 2121 } 2122 #endif 2123 2124 /* 2125 * initialize ip6_pktopts. beware that there are non-zero default values in 2126 * the struct. 2127 */ 2128 void 2129 ip6_initpktopts(struct ip6_pktopts *opt) 2130 { 2131 2132 memset(opt, 0, sizeof(*opt)); 2133 opt->ip6po_hlim = -1; /* -1 means default hop limit */ 2134 opt->ip6po_tclass = -1; /* -1 means default traffic class */ 2135 opt->ip6po_minmtu = IP6PO_MINMTU_MCASTONLY; 2136 opt->ip6po_prefer_tempaddr = IP6PO_TEMPADDR_SYSTEM; 2137 } 2138 2139 #define sin6tosa(sin6) ((struct sockaddr *)(sin6)) /* XXX */ 2140 static int 2141 ip6_pcbopt(int optname, u_char *buf, int len, struct ip6_pktopts **pktopt, 2142 kauth_cred_t cred, int uproto) 2143 { 2144 struct ip6_pktopts *opt; 2145 2146 if (*pktopt == NULL) { 2147 *pktopt = malloc(sizeof(struct ip6_pktopts), M_IP6OPT, 2148 M_NOWAIT); 2149 if (*pktopt == NULL) 2150 return (ENOBUFS); 2151 2152 ip6_initpktopts(*pktopt); 2153 } 2154 opt = *pktopt; 2155 2156 return (ip6_setpktopt(optname, buf, len, opt, cred, 1, 0, uproto)); 2157 } 2158 2159 static int 2160 ip6_getpcbopt(struct ip6_pktopts *pktopt, int optname, struct sockopt *sopt) 2161 { 2162 void *optdata = NULL; 2163 int optdatalen = 0; 2164 struct ip6_ext *ip6e; 2165 int error = 0; 2166 struct in6_pktinfo null_pktinfo; 2167 int deftclass = 0, on; 2168 int defminmtu = IP6PO_MINMTU_MCASTONLY; 2169 int defpreftemp = IP6PO_TEMPADDR_SYSTEM; 2170 2171 switch (optname) { 2172 case IPV6_PKTINFO: 2173 if (pktopt && pktopt->ip6po_pktinfo) 2174 optdata = (void *)pktopt->ip6po_pktinfo; 2175 else { 2176 /* XXX: we don't have to do this every time... */ 2177 memset(&null_pktinfo, 0, sizeof(null_pktinfo)); 2178 optdata = (void *)&null_pktinfo; 2179 } 2180 optdatalen = sizeof(struct in6_pktinfo); 2181 break; 2182 case IPV6_OTCLASS: 2183 /* XXX */ 2184 return (EINVAL); 2185 case IPV6_TCLASS: 2186 if (pktopt && pktopt->ip6po_tclass >= 0) 2187 optdata = (void *)&pktopt->ip6po_tclass; 2188 else 2189 optdata = (void *)&deftclass; 2190 optdatalen = sizeof(int); 2191 break; 2192 case IPV6_HOPOPTS: 2193 if (pktopt && pktopt->ip6po_hbh) { 2194 optdata = (void *)pktopt->ip6po_hbh; 2195 ip6e = (struct ip6_ext *)pktopt->ip6po_hbh; 2196 optdatalen = (ip6e->ip6e_len + 1) << 3; 2197 } 2198 break; 2199 case IPV6_RTHDR: 2200 if (pktopt && pktopt->ip6po_rthdr) { 2201 optdata = (void *)pktopt->ip6po_rthdr; 2202 ip6e = (struct ip6_ext *)pktopt->ip6po_rthdr; 2203 optdatalen = (ip6e->ip6e_len + 1) << 3; 2204 } 2205 break; 2206 case IPV6_RTHDRDSTOPTS: 2207 if (pktopt && pktopt->ip6po_dest1) { 2208 optdata = (void *)pktopt->ip6po_dest1; 2209 ip6e = (struct ip6_ext *)pktopt->ip6po_dest1; 2210 optdatalen = (ip6e->ip6e_len + 1) << 3; 2211 } 2212 break; 2213 case IPV6_DSTOPTS: 2214 if (pktopt && pktopt->ip6po_dest2) { 2215 optdata = (void *)pktopt->ip6po_dest2; 2216 ip6e = (struct ip6_ext *)pktopt->ip6po_dest2; 2217 optdatalen = (ip6e->ip6e_len + 1) << 3; 2218 } 2219 break; 2220 case IPV6_NEXTHOP: 2221 if (pktopt && pktopt->ip6po_nexthop) { 2222 optdata = (void *)pktopt->ip6po_nexthop; 2223 optdatalen = pktopt->ip6po_nexthop->sa_len; 2224 } 2225 break; 2226 case IPV6_USE_MIN_MTU: 2227 if (pktopt) 2228 optdata = (void *)&pktopt->ip6po_minmtu; 2229 else 2230 optdata = (void *)&defminmtu; 2231 optdatalen = sizeof(int); 2232 break; 2233 case IPV6_DONTFRAG: 2234 if (pktopt && ((pktopt->ip6po_flags) & IP6PO_DONTFRAG)) 2235 on = 1; 2236 else 2237 on = 0; 2238 optdata = (void *)&on; 2239 optdatalen = sizeof(on); 2240 break; 2241 case IPV6_PREFER_TEMPADDR: 2242 if (pktopt) 2243 optdata = (void *)&pktopt->ip6po_prefer_tempaddr; 2244 else 2245 optdata = (void *)&defpreftemp; 2246 optdatalen = sizeof(int); 2247 break; 2248 default: /* should not happen */ 2249 #ifdef DIAGNOSTIC 2250 panic("ip6_getpcbopt: unexpected option\n"); 2251 #endif 2252 return (ENOPROTOOPT); 2253 } 2254 2255 error = sockopt_set(sopt, optdata, optdatalen); 2256 2257 return (error); 2258 } 2259 2260 void 2261 ip6_clearpktopts(struct ip6_pktopts *pktopt, int optname) 2262 { 2263 if (optname == -1 || optname == IPV6_PKTINFO) { 2264 if (pktopt->ip6po_pktinfo) 2265 free(pktopt->ip6po_pktinfo, M_IP6OPT); 2266 pktopt->ip6po_pktinfo = NULL; 2267 } 2268 if (optname == -1 || optname == IPV6_HOPLIMIT) 2269 pktopt->ip6po_hlim = -1; 2270 if (optname == -1 || optname == IPV6_TCLASS) 2271 pktopt->ip6po_tclass = -1; 2272 if (optname == -1 || optname == IPV6_NEXTHOP) { 2273 rtcache_free(&pktopt->ip6po_nextroute); 2274 if (pktopt->ip6po_nexthop) 2275 free(pktopt->ip6po_nexthop, M_IP6OPT); 2276 pktopt->ip6po_nexthop = NULL; 2277 } 2278 if (optname == -1 || optname == IPV6_HOPOPTS) { 2279 if (pktopt->ip6po_hbh) 2280 free(pktopt->ip6po_hbh, M_IP6OPT); 2281 pktopt->ip6po_hbh = NULL; 2282 } 2283 if (optname == -1 || optname == IPV6_RTHDRDSTOPTS) { 2284 if (pktopt->ip6po_dest1) 2285 free(pktopt->ip6po_dest1, M_IP6OPT); 2286 pktopt->ip6po_dest1 = NULL; 2287 } 2288 if (optname == -1 || optname == IPV6_RTHDR) { 2289 if (pktopt->ip6po_rhinfo.ip6po_rhi_rthdr) 2290 free(pktopt->ip6po_rhinfo.ip6po_rhi_rthdr, M_IP6OPT); 2291 pktopt->ip6po_rhinfo.ip6po_rhi_rthdr = NULL; 2292 rtcache_free(&pktopt->ip6po_route); 2293 } 2294 if (optname == -1 || optname == IPV6_DSTOPTS) { 2295 if (pktopt->ip6po_dest2) 2296 free(pktopt->ip6po_dest2, M_IP6OPT); 2297 pktopt->ip6po_dest2 = NULL; 2298 } 2299 } 2300 2301 #define PKTOPT_EXTHDRCPY(type) \ 2302 do { \ 2303 if (src->type) { \ 2304 int hlen = (((struct ip6_ext *)src->type)->ip6e_len + 1) << 3;\ 2305 dst->type = malloc(hlen, M_IP6OPT, canwait); \ 2306 if (dst->type == NULL) \ 2307 goto bad; \ 2308 memcpy(dst->type, src->type, hlen); \ 2309 } \ 2310 } while (/*CONSTCOND*/ 0) 2311 2312 static int 2313 copypktopts(struct ip6_pktopts *dst, struct ip6_pktopts *src, int canwait) 2314 { 2315 dst->ip6po_hlim = src->ip6po_hlim; 2316 dst->ip6po_tclass = src->ip6po_tclass; 2317 dst->ip6po_flags = src->ip6po_flags; 2318 dst->ip6po_minmtu = src->ip6po_minmtu; 2319 dst->ip6po_prefer_tempaddr = src->ip6po_prefer_tempaddr; 2320 if (src->ip6po_pktinfo) { 2321 dst->ip6po_pktinfo = malloc(sizeof(*dst->ip6po_pktinfo), 2322 M_IP6OPT, canwait); 2323 if (dst->ip6po_pktinfo == NULL) 2324 goto bad; 2325 *dst->ip6po_pktinfo = *src->ip6po_pktinfo; 2326 } 2327 if (src->ip6po_nexthop) { 2328 dst->ip6po_nexthop = malloc(src->ip6po_nexthop->sa_len, 2329 M_IP6OPT, canwait); 2330 if (dst->ip6po_nexthop == NULL) 2331 goto bad; 2332 memcpy(dst->ip6po_nexthop, src->ip6po_nexthop, 2333 src->ip6po_nexthop->sa_len); 2334 } 2335 PKTOPT_EXTHDRCPY(ip6po_hbh); 2336 PKTOPT_EXTHDRCPY(ip6po_dest1); 2337 PKTOPT_EXTHDRCPY(ip6po_dest2); 2338 PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */ 2339 return (0); 2340 2341 bad: 2342 if (dst->ip6po_pktinfo) free(dst->ip6po_pktinfo, M_IP6OPT); 2343 if (dst->ip6po_nexthop) free(dst->ip6po_nexthop, M_IP6OPT); 2344 if (dst->ip6po_hbh) free(dst->ip6po_hbh, M_IP6OPT); 2345 if (dst->ip6po_dest1) free(dst->ip6po_dest1, M_IP6OPT); 2346 if (dst->ip6po_dest2) free(dst->ip6po_dest2, M_IP6OPT); 2347 if (dst->ip6po_rthdr) free(dst->ip6po_rthdr, M_IP6OPT); 2348 2349 return (ENOBUFS); 2350 } 2351 #undef PKTOPT_EXTHDRCPY 2352 2353 struct ip6_pktopts * 2354 ip6_copypktopts(struct ip6_pktopts *src, int canwait) 2355 { 2356 int error; 2357 struct ip6_pktopts *dst; 2358 2359 dst = malloc(sizeof(*dst), M_IP6OPT, canwait); 2360 if (dst == NULL) 2361 return (NULL); 2362 ip6_initpktopts(dst); 2363 2364 if ((error = copypktopts(dst, src, canwait)) != 0) { 2365 free(dst, M_IP6OPT); 2366 return (NULL); 2367 } 2368 2369 return (dst); 2370 } 2371 2372 void 2373 ip6_freepcbopts(struct ip6_pktopts *pktopt) 2374 { 2375 if (pktopt == NULL) 2376 return; 2377 2378 ip6_clearpktopts(pktopt, -1); 2379 2380 free(pktopt, M_IP6OPT); 2381 } 2382 2383 int 2384 ip6_get_membership(const struct sockopt *sopt, struct ifnet **ifp, void *v, 2385 size_t l) 2386 { 2387 struct ipv6_mreq mreq; 2388 int error; 2389 struct in6_addr *ia = &mreq.ipv6mr_multiaddr; 2390 struct in_addr *ia4 = (void *)&ia->s6_addr32[3]; 2391 error = sockopt_get(sopt, &mreq, sizeof(mreq)); 2392 if (error != 0) 2393 return error; 2394 2395 if (IN6_IS_ADDR_UNSPECIFIED(ia)) { 2396 /* 2397 * We use the unspecified address to specify to accept 2398 * all multicast addresses. Only super user is allowed 2399 * to do this. 2400 */ 2401 if (kauth_authorize_network(curlwp->l_cred, KAUTH_NETWORK_IPV6, 2402 KAUTH_REQ_NETWORK_IPV6_JOIN_MULTICAST, NULL, NULL, NULL)) 2403 return EACCES; 2404 } else if (IN6_IS_ADDR_V4MAPPED(ia)) { 2405 // Don't bother if we are not going to use ifp. 2406 if (l == sizeof(*ia)) { 2407 memcpy(v, ia, l); 2408 return 0; 2409 } 2410 } else if (!IN6_IS_ADDR_MULTICAST(ia)) { 2411 return EINVAL; 2412 } 2413 2414 /* 2415 * If no interface was explicitly specified, choose an 2416 * appropriate one according to the given multicast address. 2417 */ 2418 if (mreq.ipv6mr_interface == 0) { 2419 struct rtentry *rt; 2420 union { 2421 struct sockaddr dst; 2422 struct sockaddr_in dst4; 2423 struct sockaddr_in6 dst6; 2424 } u; 2425 struct route ro; 2426 2427 /* 2428 * Look up the routing table for the 2429 * address, and choose the outgoing interface. 2430 * XXX: is it a good approach? 2431 */ 2432 memset(&ro, 0, sizeof(ro)); 2433 if (IN6_IS_ADDR_V4MAPPED(ia)) 2434 sockaddr_in_init(&u.dst4, ia4, 0); 2435 else 2436 sockaddr_in6_init(&u.dst6, ia, 0, 0, 0); 2437 error = rtcache_setdst(&ro, &u.dst); 2438 if (error != 0) 2439 return error; 2440 rt = rtcache_init(&ro); 2441 *ifp = rt != NULL ? rt->rt_ifp : NULL; 2442 /* FIXME *ifp is NOMPSAFE */ 2443 rtcache_unref(rt, &ro); 2444 rtcache_free(&ro); 2445 } else { 2446 /* 2447 * If the interface is specified, validate it. 2448 */ 2449 if ((*ifp = if_byindex(mreq.ipv6mr_interface)) == NULL) 2450 return ENXIO; /* XXX EINVAL? */ 2451 } 2452 if (sizeof(*ia) == l) 2453 memcpy(v, ia, l); 2454 else 2455 memcpy(v, ia4, l); 2456 return 0; 2457 } 2458 2459 /* 2460 * Set the IP6 multicast options in response to user setsockopt(). 2461 */ 2462 static int 2463 ip6_setmoptions(const struct sockopt *sopt, struct in6pcb *in6p) 2464 { 2465 int error = 0; 2466 u_int loop, ifindex; 2467 struct ipv6_mreq mreq; 2468 struct in6_addr ia; 2469 struct ifnet *ifp; 2470 struct ip6_moptions *im6o = in6p->in6p_moptions; 2471 struct in6_multi_mship *imm; 2472 2473 KASSERT(in6plocked(in6p)); 2474 2475 if (im6o == NULL) { 2476 /* 2477 * No multicast option buffer attached to the pcb; 2478 * allocate one and initialize to default values. 2479 */ 2480 im6o = malloc(sizeof(*im6o), M_IPMOPTS, M_NOWAIT); 2481 if (im6o == NULL) 2482 return (ENOBUFS); 2483 in6p->in6p_moptions = im6o; 2484 im6o->im6o_multicast_if_index = 0; 2485 im6o->im6o_multicast_hlim = ip6_defmcasthlim; 2486 im6o->im6o_multicast_loop = IPV6_DEFAULT_MULTICAST_LOOP; 2487 LIST_INIT(&im6o->im6o_memberships); 2488 } 2489 2490 switch (sopt->sopt_name) { 2491 2492 case IPV6_MULTICAST_IF: 2493 /* 2494 * Select the interface for outgoing multicast packets. 2495 */ 2496 error = sockopt_get(sopt, &ifindex, sizeof(ifindex)); 2497 if (error != 0) 2498 break; 2499 2500 if (ifindex != 0) { 2501 if ((ifp = if_byindex(ifindex)) == NULL) { 2502 error = ENXIO; /* XXX EINVAL? */ 2503 break; 2504 } 2505 if ((ifp->if_flags & IFF_MULTICAST) == 0) { 2506 error = EADDRNOTAVAIL; 2507 break; 2508 } 2509 } else 2510 ifp = NULL; 2511 im6o->im6o_multicast_if_index = if_get_index(ifp); 2512 break; 2513 2514 case IPV6_MULTICAST_HOPS: 2515 { 2516 /* 2517 * Set the IP6 hoplimit for outgoing multicast packets. 2518 */ 2519 int optval; 2520 2521 error = sockopt_getint(sopt, &optval); 2522 if (error != 0) 2523 break; 2524 2525 if (optval < -1 || optval >= 256) 2526 error = EINVAL; 2527 else if (optval == -1) 2528 im6o->im6o_multicast_hlim = ip6_defmcasthlim; 2529 else 2530 im6o->im6o_multicast_hlim = optval; 2531 break; 2532 } 2533 2534 case IPV6_MULTICAST_LOOP: 2535 /* 2536 * Set the loopback flag for outgoing multicast packets. 2537 * Must be zero or one. 2538 */ 2539 error = sockopt_get(sopt, &loop, sizeof(loop)); 2540 if (error != 0) 2541 break; 2542 if (loop > 1) { 2543 error = EINVAL; 2544 break; 2545 } 2546 im6o->im6o_multicast_loop = loop; 2547 break; 2548 2549 case IPV6_JOIN_GROUP: 2550 /* 2551 * Add a multicast group membership. 2552 * Group must be a valid IP6 multicast address. 2553 */ 2554 if ((error = ip6_get_membership(sopt, &ifp, &ia, sizeof(ia)))) 2555 return error; 2556 2557 if (IN6_IS_ADDR_V4MAPPED(&ia)) { 2558 error = ip_setmoptions(&in6p->in6p_v4moptions, sopt); 2559 break; 2560 } 2561 /* 2562 * See if we found an interface, and confirm that it 2563 * supports multicast 2564 */ 2565 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) { 2566 error = EADDRNOTAVAIL; 2567 break; 2568 } 2569 2570 if (in6_setscope(&ia, ifp, NULL)) { 2571 error = EADDRNOTAVAIL; /* XXX: should not happen */ 2572 break; 2573 } 2574 2575 /* 2576 * See if the membership already exists. 2577 */ 2578 for (imm = im6o->im6o_memberships.lh_first; 2579 imm != NULL; imm = imm->i6mm_chain.le_next) 2580 if (imm->i6mm_maddr->in6m_ifp == ifp && 2581 IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr, 2582 &ia)) 2583 break; 2584 if (imm != NULL) { 2585 error = EADDRINUSE; 2586 break; 2587 } 2588 /* 2589 * Everything looks good; add a new record to the multicast 2590 * address list for the given interface. 2591 */ 2592 imm = in6_joingroup(ifp, &ia, &error, 0); 2593 if (imm == NULL) 2594 break; 2595 LIST_INSERT_HEAD(&im6o->im6o_memberships, imm, i6mm_chain); 2596 break; 2597 2598 case IPV6_LEAVE_GROUP: 2599 /* 2600 * Drop a multicast group membership. 2601 * Group must be a valid IP6 multicast address. 2602 */ 2603 error = sockopt_get(sopt, &mreq, sizeof(mreq)); 2604 if (error != 0) 2605 break; 2606 2607 if (IN6_IS_ADDR_V4MAPPED(&mreq.ipv6mr_multiaddr)) { 2608 error = ip_setmoptions(&in6p->in6p_v4moptions, sopt); 2609 break; 2610 } 2611 /* 2612 * If an interface address was specified, get a pointer 2613 * to its ifnet structure. 2614 */ 2615 if (mreq.ipv6mr_interface != 0) { 2616 if ((ifp = if_byindex(mreq.ipv6mr_interface)) == NULL) { 2617 error = ENXIO; /* XXX EINVAL? */ 2618 break; 2619 } 2620 } else 2621 ifp = NULL; 2622 2623 /* Fill in the scope zone ID */ 2624 if (ifp) { 2625 if (in6_setscope(&mreq.ipv6mr_multiaddr, ifp, NULL)) { 2626 /* XXX: should not happen */ 2627 error = EADDRNOTAVAIL; 2628 break; 2629 } 2630 } else if (mreq.ipv6mr_interface != 0) { 2631 /* 2632 * XXX: This case would happens when the (positive) 2633 * index is in the valid range, but the corresponding 2634 * interface has been detached dynamically. The above 2635 * check probably avoids such case to happen here, but 2636 * we check it explicitly for safety. 2637 */ 2638 error = EADDRNOTAVAIL; 2639 break; 2640 } else { /* ipv6mr_interface == 0 */ 2641 struct sockaddr_in6 sa6_mc; 2642 2643 /* 2644 * The API spec says as follows: 2645 * If the interface index is specified as 0, the 2646 * system may choose a multicast group membership to 2647 * drop by matching the multicast address only. 2648 * On the other hand, we cannot disambiguate the scope 2649 * zone unless an interface is provided. Thus, we 2650 * check if there's ambiguity with the default scope 2651 * zone as the last resort. 2652 */ 2653 sockaddr_in6_init(&sa6_mc, &mreq.ipv6mr_multiaddr, 2654 0, 0, 0); 2655 error = sa6_embedscope(&sa6_mc, ip6_use_defzone); 2656 if (error != 0) 2657 break; 2658 mreq.ipv6mr_multiaddr = sa6_mc.sin6_addr; 2659 } 2660 2661 /* 2662 * Find the membership in the membership list. 2663 */ 2664 for (imm = im6o->im6o_memberships.lh_first; 2665 imm != NULL; imm = imm->i6mm_chain.le_next) { 2666 if ((ifp == NULL || imm->i6mm_maddr->in6m_ifp == ifp) && 2667 IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr, 2668 &mreq.ipv6mr_multiaddr)) 2669 break; 2670 } 2671 if (imm == NULL) { 2672 /* Unable to resolve interface */ 2673 error = EADDRNOTAVAIL; 2674 break; 2675 } 2676 /* 2677 * Give up the multicast address record to which the 2678 * membership points. 2679 */ 2680 LIST_REMOVE(imm, i6mm_chain); 2681 in6_leavegroup(imm); 2682 break; 2683 2684 default: 2685 error = EOPNOTSUPP; 2686 break; 2687 } 2688 2689 /* 2690 * If all options have default values, no need to keep the mbuf. 2691 */ 2692 if (im6o->im6o_multicast_if_index == 0 && 2693 im6o->im6o_multicast_hlim == ip6_defmcasthlim && 2694 im6o->im6o_multicast_loop == IPV6_DEFAULT_MULTICAST_LOOP && 2695 im6o->im6o_memberships.lh_first == NULL) { 2696 free(in6p->in6p_moptions, M_IPMOPTS); 2697 in6p->in6p_moptions = NULL; 2698 } 2699 2700 return (error); 2701 } 2702 2703 /* 2704 * Return the IP6 multicast options in response to user getsockopt(). 2705 */ 2706 static int 2707 ip6_getmoptions(struct sockopt *sopt, struct in6pcb *in6p) 2708 { 2709 u_int optval; 2710 int error; 2711 struct ip6_moptions *im6o = in6p->in6p_moptions; 2712 2713 switch (sopt->sopt_name) { 2714 case IPV6_MULTICAST_IF: 2715 if (im6o == NULL || im6o->im6o_multicast_if_index == 0) 2716 optval = 0; 2717 else 2718 optval = im6o->im6o_multicast_if_index; 2719 2720 error = sockopt_set(sopt, &optval, sizeof(optval)); 2721 break; 2722 2723 case IPV6_MULTICAST_HOPS: 2724 if (im6o == NULL) 2725 optval = ip6_defmcasthlim; 2726 else 2727 optval = im6o->im6o_multicast_hlim; 2728 2729 error = sockopt_set(sopt, &optval, sizeof(optval)); 2730 break; 2731 2732 case IPV6_MULTICAST_LOOP: 2733 if (im6o == NULL) 2734 optval = IPV6_DEFAULT_MULTICAST_LOOP; 2735 else 2736 optval = im6o->im6o_multicast_loop; 2737 2738 error = sockopt_set(sopt, &optval, sizeof(optval)); 2739 break; 2740 2741 default: 2742 error = EOPNOTSUPP; 2743 } 2744 2745 return (error); 2746 } 2747 2748 /* 2749 * Discard the IP6 multicast options. 2750 */ 2751 void 2752 ip6_freemoptions(struct ip6_moptions *im6o) 2753 { 2754 struct in6_multi_mship *imm; 2755 2756 if (im6o == NULL) 2757 return; 2758 2759 while ((imm = im6o->im6o_memberships.lh_first) != NULL) { 2760 LIST_REMOVE(imm, i6mm_chain); 2761 in6_leavegroup(imm); 2762 } 2763 free(im6o, M_IPMOPTS); 2764 } 2765 2766 /* 2767 * Set IPv6 outgoing packet options based on advanced API. 2768 */ 2769 int 2770 ip6_setpktopts(struct mbuf *control, struct ip6_pktopts *opt, 2771 struct ip6_pktopts *stickyopt, kauth_cred_t cred, int uproto) 2772 { 2773 struct cmsghdr *cm = 0; 2774 2775 if (control == NULL || opt == NULL) 2776 return (EINVAL); 2777 2778 ip6_initpktopts(opt); 2779 if (stickyopt) { 2780 int error; 2781 2782 /* 2783 * If stickyopt is provided, make a local copy of the options 2784 * for this particular packet, then override them by ancillary 2785 * objects. 2786 * XXX: copypktopts() does not copy the cached route to a next 2787 * hop (if any). This is not very good in terms of efficiency, 2788 * but we can allow this since this option should be rarely 2789 * used. 2790 */ 2791 if ((error = copypktopts(opt, stickyopt, M_NOWAIT)) != 0) 2792 return (error); 2793 } 2794 2795 /* 2796 * XXX: Currently, we assume all the optional information is stored 2797 * in a single mbuf. 2798 */ 2799 if (control->m_next) 2800 return (EINVAL); 2801 2802 /* XXX if cm->cmsg_len is not aligned, control->m_len can become <0 */ 2803 for (; control->m_len > 0; control->m_data += CMSG_ALIGN(cm->cmsg_len), 2804 control->m_len -= CMSG_ALIGN(cm->cmsg_len)) { 2805 int error; 2806 2807 if (control->m_len < CMSG_LEN(0)) 2808 return (EINVAL); 2809 2810 cm = mtod(control, struct cmsghdr *); 2811 if (cm->cmsg_len == 0 || cm->cmsg_len > control->m_len) 2812 return (EINVAL); 2813 if (cm->cmsg_level != IPPROTO_IPV6) 2814 continue; 2815 2816 error = ip6_setpktopt(cm->cmsg_type, CMSG_DATA(cm), 2817 cm->cmsg_len - CMSG_LEN(0), opt, cred, 0, 1, uproto); 2818 if (error) 2819 return (error); 2820 } 2821 2822 return (0); 2823 } 2824 2825 /* 2826 * Set a particular packet option, as a sticky option or an ancillary data 2827 * item. "len" can be 0 only when it's a sticky option. 2828 * We have 4 cases of combination of "sticky" and "cmsg": 2829 * "sticky=0, cmsg=0": impossible 2830 * "sticky=0, cmsg=1": RFC2292 or RFC3542 ancillary data 2831 * "sticky=1, cmsg=0": RFC3542 socket option 2832 * "sticky=1, cmsg=1": RFC2292 socket option 2833 */ 2834 static int 2835 ip6_setpktopt(int optname, u_char *buf, int len, struct ip6_pktopts *opt, 2836 kauth_cred_t cred, int sticky, int cmsg, int uproto) 2837 { 2838 int minmtupolicy; 2839 int error; 2840 2841 if (!sticky && !cmsg) { 2842 #ifdef DIAGNOSTIC 2843 printf("ip6_setpktopt: impossible case\n"); 2844 #endif 2845 return (EINVAL); 2846 } 2847 2848 /* 2849 * IPV6_2292xxx is for backward compatibility to RFC2292, and should 2850 * not be specified in the context of RFC3542. Conversely, 2851 * RFC3542 types should not be specified in the context of RFC2292. 2852 */ 2853 if (!cmsg) { 2854 switch (optname) { 2855 case IPV6_2292PKTINFO: 2856 case IPV6_2292HOPLIMIT: 2857 case IPV6_2292NEXTHOP: 2858 case IPV6_2292HOPOPTS: 2859 case IPV6_2292DSTOPTS: 2860 case IPV6_2292RTHDR: 2861 case IPV6_2292PKTOPTIONS: 2862 return (ENOPROTOOPT); 2863 } 2864 } 2865 if (sticky && cmsg) { 2866 switch (optname) { 2867 case IPV6_PKTINFO: 2868 case IPV6_HOPLIMIT: 2869 case IPV6_NEXTHOP: 2870 case IPV6_HOPOPTS: 2871 case IPV6_DSTOPTS: 2872 case IPV6_RTHDRDSTOPTS: 2873 case IPV6_RTHDR: 2874 case IPV6_USE_MIN_MTU: 2875 case IPV6_DONTFRAG: 2876 case IPV6_OTCLASS: 2877 case IPV6_TCLASS: 2878 case IPV6_PREFER_TEMPADDR: /* XXX not an RFC3542 option */ 2879 return (ENOPROTOOPT); 2880 } 2881 } 2882 2883 switch (optname) { 2884 #ifdef RFC2292 2885 case IPV6_2292PKTINFO: 2886 #endif 2887 case IPV6_PKTINFO: 2888 { 2889 struct in6_pktinfo *pktinfo; 2890 2891 if (len != sizeof(struct in6_pktinfo)) 2892 return (EINVAL); 2893 2894 pktinfo = (struct in6_pktinfo *)buf; 2895 2896 /* 2897 * An application can clear any sticky IPV6_PKTINFO option by 2898 * doing a "regular" setsockopt with ipi6_addr being 2899 * in6addr_any and ipi6_ifindex being zero. 2900 * [RFC 3542, Section 6] 2901 */ 2902 if (optname == IPV6_PKTINFO && opt->ip6po_pktinfo && 2903 pktinfo->ipi6_ifindex == 0 && 2904 IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) { 2905 ip6_clearpktopts(opt, optname); 2906 break; 2907 } 2908 2909 if (uproto == IPPROTO_TCP && optname == IPV6_PKTINFO && 2910 sticky && !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) { 2911 return (EINVAL); 2912 } 2913 2914 /* Validate the interface index if specified. */ 2915 if (pktinfo->ipi6_ifindex) { 2916 struct ifnet *ifp; 2917 int s = pserialize_read_enter(); 2918 ifp = if_byindex(pktinfo->ipi6_ifindex); 2919 if (ifp == NULL) { 2920 pserialize_read_exit(s); 2921 return ENXIO; 2922 } 2923 pserialize_read_exit(s); 2924 } 2925 2926 /* 2927 * We store the address anyway, and let in6_selectsrc() 2928 * validate the specified address. This is because ipi6_addr 2929 * may not have enough information about its scope zone, and 2930 * we may need additional information (such as outgoing 2931 * interface or the scope zone of a destination address) to 2932 * disambiguate the scope. 2933 * XXX: the delay of the validation may confuse the 2934 * application when it is used as a sticky option. 2935 */ 2936 if (opt->ip6po_pktinfo == NULL) { 2937 opt->ip6po_pktinfo = malloc(sizeof(*pktinfo), 2938 M_IP6OPT, M_NOWAIT); 2939 if (opt->ip6po_pktinfo == NULL) 2940 return (ENOBUFS); 2941 } 2942 memcpy(opt->ip6po_pktinfo, pktinfo, sizeof(*pktinfo)); 2943 break; 2944 } 2945 2946 #ifdef RFC2292 2947 case IPV6_2292HOPLIMIT: 2948 #endif 2949 case IPV6_HOPLIMIT: 2950 { 2951 int *hlimp; 2952 2953 /* 2954 * RFC 3542 deprecated the usage of sticky IPV6_HOPLIMIT 2955 * to simplify the ordering among hoplimit options. 2956 */ 2957 if (optname == IPV6_HOPLIMIT && sticky) 2958 return (ENOPROTOOPT); 2959 2960 if (len != sizeof(int)) 2961 return (EINVAL); 2962 hlimp = (int *)buf; 2963 if (*hlimp < -1 || *hlimp > 255) 2964 return (EINVAL); 2965 2966 opt->ip6po_hlim = *hlimp; 2967 break; 2968 } 2969 2970 case IPV6_OTCLASS: 2971 if (len != sizeof(u_int8_t)) 2972 return (EINVAL); 2973 2974 opt->ip6po_tclass = *(u_int8_t *)buf; 2975 break; 2976 2977 case IPV6_TCLASS: 2978 { 2979 int tclass; 2980 2981 if (len != sizeof(int)) 2982 return (EINVAL); 2983 tclass = *(int *)buf; 2984 if (tclass < -1 || tclass > 255) 2985 return (EINVAL); 2986 2987 opt->ip6po_tclass = tclass; 2988 break; 2989 } 2990 2991 #ifdef RFC2292 2992 case IPV6_2292NEXTHOP: 2993 #endif 2994 case IPV6_NEXTHOP: 2995 error = kauth_authorize_network(cred, KAUTH_NETWORK_IPV6, 2996 KAUTH_REQ_NETWORK_IPV6_HOPBYHOP, NULL, NULL, NULL); 2997 if (error) 2998 return (error); 2999 3000 if (len == 0) { /* just remove the option */ 3001 ip6_clearpktopts(opt, IPV6_NEXTHOP); 3002 break; 3003 } 3004 3005 /* check if cmsg_len is large enough for sa_len */ 3006 if (len < sizeof(struct sockaddr) || len < *buf) 3007 return (EINVAL); 3008 3009 switch (((struct sockaddr *)buf)->sa_family) { 3010 case AF_INET6: 3011 { 3012 struct sockaddr_in6 *sa6 = (struct sockaddr_in6 *)buf; 3013 3014 if (sa6->sin6_len != sizeof(struct sockaddr_in6)) 3015 return (EINVAL); 3016 3017 if (IN6_IS_ADDR_UNSPECIFIED(&sa6->sin6_addr) || 3018 IN6_IS_ADDR_MULTICAST(&sa6->sin6_addr)) { 3019 return (EINVAL); 3020 } 3021 if ((error = sa6_embedscope(sa6, ip6_use_defzone)) 3022 != 0) { 3023 return (error); 3024 } 3025 break; 3026 } 3027 case AF_LINK: /* eventually be supported? */ 3028 default: 3029 return (EAFNOSUPPORT); 3030 } 3031 3032 /* turn off the previous option, then set the new option. */ 3033 ip6_clearpktopts(opt, IPV6_NEXTHOP); 3034 opt->ip6po_nexthop = malloc(*buf, M_IP6OPT, M_NOWAIT); 3035 if (opt->ip6po_nexthop == NULL) 3036 return (ENOBUFS); 3037 memcpy(opt->ip6po_nexthop, buf, *buf); 3038 break; 3039 3040 #ifdef RFC2292 3041 case IPV6_2292HOPOPTS: 3042 #endif 3043 case IPV6_HOPOPTS: 3044 { 3045 struct ip6_hbh *hbh; 3046 int hbhlen; 3047 3048 /* 3049 * XXX: We don't allow a non-privileged user to set ANY HbH 3050 * options, since per-option restriction has too much 3051 * overhead. 3052 */ 3053 error = kauth_authorize_network(cred, KAUTH_NETWORK_IPV6, 3054 KAUTH_REQ_NETWORK_IPV6_HOPBYHOP, NULL, NULL, NULL); 3055 if (error) 3056 return (error); 3057 3058 if (len == 0) { 3059 ip6_clearpktopts(opt, IPV6_HOPOPTS); 3060 break; /* just remove the option */ 3061 } 3062 3063 /* message length validation */ 3064 if (len < sizeof(struct ip6_hbh)) 3065 return (EINVAL); 3066 hbh = (struct ip6_hbh *)buf; 3067 hbhlen = (hbh->ip6h_len + 1) << 3; 3068 if (len != hbhlen) 3069 return (EINVAL); 3070 3071 /* turn off the previous option, then set the new option. */ 3072 ip6_clearpktopts(opt, IPV6_HOPOPTS); 3073 opt->ip6po_hbh = malloc(hbhlen, M_IP6OPT, M_NOWAIT); 3074 if (opt->ip6po_hbh == NULL) 3075 return (ENOBUFS); 3076 memcpy(opt->ip6po_hbh, hbh, hbhlen); 3077 3078 break; 3079 } 3080 3081 #ifdef RFC2292 3082 case IPV6_2292DSTOPTS: 3083 #endif 3084 case IPV6_DSTOPTS: 3085 case IPV6_RTHDRDSTOPTS: 3086 { 3087 struct ip6_dest *dest, **newdest = NULL; 3088 int destlen; 3089 3090 /* XXX: see the comment for IPV6_HOPOPTS */ 3091 error = kauth_authorize_network(cred, KAUTH_NETWORK_IPV6, 3092 KAUTH_REQ_NETWORK_IPV6_HOPBYHOP, NULL, NULL, NULL); 3093 if (error) 3094 return (error); 3095 3096 if (len == 0) { 3097 ip6_clearpktopts(opt, optname); 3098 break; /* just remove the option */ 3099 } 3100 3101 /* message length validation */ 3102 if (len < sizeof(struct ip6_dest)) 3103 return (EINVAL); 3104 dest = (struct ip6_dest *)buf; 3105 destlen = (dest->ip6d_len + 1) << 3; 3106 if (len != destlen) 3107 return (EINVAL); 3108 /* 3109 * Determine the position that the destination options header 3110 * should be inserted; before or after the routing header. 3111 */ 3112 switch (optname) { 3113 case IPV6_2292DSTOPTS: 3114 /* 3115 * The old advanced API is ambiguous on this point. 3116 * Our approach is to determine the position based 3117 * according to the existence of a routing header. 3118 * Note, however, that this depends on the order of the 3119 * extension headers in the ancillary data; the 1st 3120 * part of the destination options header must appear 3121 * before the routing header in the ancillary data, 3122 * too. 3123 * RFC3542 solved the ambiguity by introducing 3124 * separate ancillary data or option types. 3125 */ 3126 if (opt->ip6po_rthdr == NULL) 3127 newdest = &opt->ip6po_dest1; 3128 else 3129 newdest = &opt->ip6po_dest2; 3130 break; 3131 case IPV6_RTHDRDSTOPTS: 3132 newdest = &opt->ip6po_dest1; 3133 break; 3134 case IPV6_DSTOPTS: 3135 newdest = &opt->ip6po_dest2; 3136 break; 3137 } 3138 3139 /* turn off the previous option, then set the new option. */ 3140 ip6_clearpktopts(opt, optname); 3141 *newdest = malloc(destlen, M_IP6OPT, M_NOWAIT); 3142 if (*newdest == NULL) 3143 return (ENOBUFS); 3144 memcpy(*newdest, dest, destlen); 3145 3146 break; 3147 } 3148 3149 #ifdef RFC2292 3150 case IPV6_2292RTHDR: 3151 #endif 3152 case IPV6_RTHDR: 3153 { 3154 struct ip6_rthdr *rth; 3155 int rthlen; 3156 3157 if (len == 0) { 3158 ip6_clearpktopts(opt, IPV6_RTHDR); 3159 break; /* just remove the option */ 3160 } 3161 3162 /* message length validation */ 3163 if (len < sizeof(struct ip6_rthdr)) 3164 return (EINVAL); 3165 rth = (struct ip6_rthdr *)buf; 3166 rthlen = (rth->ip6r_len + 1) << 3; 3167 if (len != rthlen) 3168 return (EINVAL); 3169 switch (rth->ip6r_type) { 3170 case IPV6_RTHDR_TYPE_0: 3171 if (rth->ip6r_len == 0) /* must contain one addr */ 3172 return (EINVAL); 3173 if (rth->ip6r_len % 2) /* length must be even */ 3174 return (EINVAL); 3175 if (rth->ip6r_len / 2 != rth->ip6r_segleft) 3176 return (EINVAL); 3177 break; 3178 default: 3179 return (EINVAL); /* not supported */ 3180 } 3181 /* turn off the previous option */ 3182 ip6_clearpktopts(opt, IPV6_RTHDR); 3183 opt->ip6po_rthdr = malloc(rthlen, M_IP6OPT, M_NOWAIT); 3184 if (opt->ip6po_rthdr == NULL) 3185 return (ENOBUFS); 3186 memcpy(opt->ip6po_rthdr, rth, rthlen); 3187 break; 3188 } 3189 3190 case IPV6_USE_MIN_MTU: 3191 if (len != sizeof(int)) 3192 return (EINVAL); 3193 minmtupolicy = *(int *)buf; 3194 if (minmtupolicy != IP6PO_MINMTU_MCASTONLY && 3195 minmtupolicy != IP6PO_MINMTU_DISABLE && 3196 minmtupolicy != IP6PO_MINMTU_ALL) { 3197 return (EINVAL); 3198 } 3199 opt->ip6po_minmtu = minmtupolicy; 3200 break; 3201 3202 case IPV6_DONTFRAG: 3203 if (len != sizeof(int)) 3204 return (EINVAL); 3205 3206 if (uproto == IPPROTO_TCP || *(int *)buf == 0) { 3207 /* 3208 * we ignore this option for TCP sockets. 3209 * (RFC3542 leaves this case unspecified.) 3210 */ 3211 opt->ip6po_flags &= ~IP6PO_DONTFRAG; 3212 } else 3213 opt->ip6po_flags |= IP6PO_DONTFRAG; 3214 break; 3215 3216 case IPV6_PREFER_TEMPADDR: 3217 { 3218 int preftemp; 3219 3220 if (len != sizeof(int)) 3221 return (EINVAL); 3222 preftemp = *(int *)buf; 3223 switch (preftemp) { 3224 case IP6PO_TEMPADDR_SYSTEM: 3225 case IP6PO_TEMPADDR_NOTPREFER: 3226 case IP6PO_TEMPADDR_PREFER: 3227 break; 3228 default: 3229 return (EINVAL); 3230 } 3231 opt->ip6po_prefer_tempaddr = preftemp; 3232 break; 3233 } 3234 3235 default: 3236 return (ENOPROTOOPT); 3237 } /* end of switch */ 3238 3239 return (0); 3240 } 3241 3242 /* 3243 * Routine called from ip6_output() to loop back a copy of an IP6 multicast 3244 * packet to the input queue of a specified interface. Note that this 3245 * calls the output routine of the loopback "driver", but with an interface 3246 * pointer that might NOT be lo0ifp -- easier than replicating that code here. 3247 */ 3248 void 3249 ip6_mloopback(struct ifnet *ifp, struct mbuf *m, 3250 const struct sockaddr_in6 *dst) 3251 { 3252 struct mbuf *copym; 3253 struct ip6_hdr *ip6; 3254 3255 copym = m_copy(m, 0, M_COPYALL); 3256 if (copym == NULL) 3257 return; 3258 3259 /* 3260 * Make sure to deep-copy IPv6 header portion in case the data 3261 * is in an mbuf cluster, so that we can safely override the IPv6 3262 * header portion later. 3263 */ 3264 if ((copym->m_flags & M_EXT) != 0 || 3265 copym->m_len < sizeof(struct ip6_hdr)) { 3266 copym = m_pullup(copym, sizeof(struct ip6_hdr)); 3267 if (copym == NULL) 3268 return; 3269 } 3270 3271 #ifdef DIAGNOSTIC 3272 if (copym->m_len < sizeof(*ip6)) { 3273 m_freem(copym); 3274 return; 3275 } 3276 #endif 3277 3278 ip6 = mtod(copym, struct ip6_hdr *); 3279 /* 3280 * clear embedded scope identifiers if necessary. 3281 * in6_clearscope will touch the addresses only when necessary. 3282 */ 3283 in6_clearscope(&ip6->ip6_src); 3284 in6_clearscope(&ip6->ip6_dst); 3285 3286 (void)looutput(ifp, copym, (const struct sockaddr *)dst, NULL); 3287 } 3288 3289 /* 3290 * Chop IPv6 header off from the payload. 3291 */ 3292 static int 3293 ip6_splithdr(struct mbuf *m, struct ip6_exthdrs *exthdrs) 3294 { 3295 struct mbuf *mh; 3296 struct ip6_hdr *ip6; 3297 3298 ip6 = mtod(m, struct ip6_hdr *); 3299 if (m->m_len > sizeof(*ip6)) { 3300 MGETHDR(mh, M_DONTWAIT, MT_HEADER); 3301 if (mh == 0) { 3302 m_freem(m); 3303 return ENOBUFS; 3304 } 3305 M_MOVE_PKTHDR(mh, m); 3306 MH_ALIGN(mh, sizeof(*ip6)); 3307 m->m_len -= sizeof(*ip6); 3308 m->m_data += sizeof(*ip6); 3309 mh->m_next = m; 3310 m = mh; 3311 m->m_len = sizeof(*ip6); 3312 bcopy((void *)ip6, mtod(m, void *), sizeof(*ip6)); 3313 } 3314 exthdrs->ip6e_ip6 = m; 3315 return 0; 3316 } 3317 3318 /* 3319 * Compute IPv6 extension header length. 3320 */ 3321 int 3322 ip6_optlen(struct in6pcb *in6p) 3323 { 3324 int len; 3325 3326 if (!in6p->in6p_outputopts) 3327 return 0; 3328 3329 len = 0; 3330 #define elen(x) \ 3331 (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0) 3332 3333 len += elen(in6p->in6p_outputopts->ip6po_hbh); 3334 len += elen(in6p->in6p_outputopts->ip6po_dest1); 3335 len += elen(in6p->in6p_outputopts->ip6po_rthdr); 3336 len += elen(in6p->in6p_outputopts->ip6po_dest2); 3337 return len; 3338 #undef elen 3339 } 3340 3341 /* 3342 * Ensure sending address is valid. 3343 * Returns 0 on success, -1 if an error should be sent back or 1 3344 * if the packet could be dropped without error (protocol dependent). 3345 */ 3346 static int 3347 ip6_ifaddrvalid(const struct in6_addr *addr) 3348 { 3349 struct sockaddr_in6 sin6; 3350 int s, error; 3351 struct ifaddr *ifa; 3352 struct in6_ifaddr *ia6; 3353 3354 if (IN6_IS_ADDR_UNSPECIFIED(addr)) 3355 return 0; 3356 3357 memset(&sin6, 0, sizeof(sin6)); 3358 sin6.sin6_family = AF_INET6; 3359 sin6.sin6_len = sizeof(sin6); 3360 sin6.sin6_addr = *addr; 3361 3362 s = pserialize_read_enter(); 3363 ifa = ifa_ifwithaddr(sin6tosa(&sin6)); 3364 if ((ia6 = ifatoia6(ifa)) == NULL || 3365 ia6->ia6_flags & (IN6_IFF_ANYCAST | IN6_IFF_DUPLICATED)) 3366 error = -1; 3367 else if (ia6->ia6_flags & (IN6_IFF_TENTATIVE | IN6_IFF_DETACHED)) 3368 error = 1; 3369 else 3370 error = 0; 3371 pserialize_read_exit(s); 3372 3373 return error; 3374 } 3375
Indexes created Tue Sep 30 20:09:53 GMT 2025