ip6_output.c revision 1.180
1 /* $NetBSD: ip6_output.c,v 1.180 2017/01/11 13:08:29 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 1.180 2017/01/11 13:08:29 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 * IP6 output. The packet in mbuf chain m contains a skeletal IP6 184 * header (with pri, len, nxt, hlim, src, dst). 185 * This function may modify ver and hlim only. 186 * The mbuf chain containing the packet will be freed. 187 * The mbuf opt, if present, will not be freed. 188 * 189 * type of "mtu": rt_rmx.rmx_mtu is u_long, ifnet.ifr_mtu is int, and 190 * nd_ifinfo.linkmtu is u_int32_t. so we use u_long to hold largest one, 191 * which is rt_rmx.rmx_mtu. 192 */ 193 int 194 ip6_output( 195 struct mbuf *m0, 196 struct ip6_pktopts *opt, 197 struct route *ro, 198 int flags, 199 struct ip6_moptions *im6o, 200 struct socket *so, 201 struct ifnet **ifpp /* XXX: just for statistics */ 202 ) 203 { 204 struct ip6_hdr *ip6, *mhip6; 205 struct ifnet *ifp = NULL, *origifp = NULL; 206 struct mbuf *m = m0; 207 int hlen, tlen, len, off; 208 bool tso; 209 struct route ip6route; 210 struct rtentry *rt = NULL, *rt_pmtu; 211 const struct sockaddr_in6 *dst; 212 struct sockaddr_in6 src_sa, dst_sa; 213 int error = 0; 214 struct in6_ifaddr *ia = NULL; 215 u_long mtu; 216 int alwaysfrag, dontfrag; 217 u_int32_t optlen = 0, plen = 0, unfragpartlen = 0; 218 struct ip6_exthdrs exthdrs; 219 struct in6_addr finaldst, src0, dst0; 220 u_int32_t zone; 221 struct route *ro_pmtu = NULL; 222 int hdrsplit = 0; 223 int needipsec = 0; 224 #ifdef IPSEC 225 struct secpolicy *sp = NULL; 226 #endif 227 struct psref psref, psref_ia; 228 int bound = curlwp_bind(); 229 bool release_psref_ia = false; 230 231 #ifdef DIAGNOSTIC 232 if ((m->m_flags & M_PKTHDR) == 0) 233 panic("ip6_output: no HDR"); 234 235 if ((m->m_pkthdr.csum_flags & 236 (M_CSUM_TCPv4|M_CSUM_UDPv4|M_CSUM_TSOv4)) != 0) { 237 panic("ip6_output: IPv4 checksum offload flags: %d", 238 m->m_pkthdr.csum_flags); 239 } 240 241 if ((m->m_pkthdr.csum_flags & (M_CSUM_TCPv6|M_CSUM_UDPv6)) == 242 (M_CSUM_TCPv6|M_CSUM_UDPv6)) { 243 panic("ip6_output: conflicting checksum offload flags: %d", 244 m->m_pkthdr.csum_flags); 245 } 246 #endif 247 248 M_CSUM_DATA_IPv6_HL_SET(m->m_pkthdr.csum_data, sizeof(struct ip6_hdr)); 249 250 #define MAKE_EXTHDR(hp, mp) \ 251 do { \ 252 if (hp) { \ 253 struct ip6_ext *eh = (struct ip6_ext *)(hp); \ 254 error = ip6_copyexthdr((mp), (void *)(hp), \ 255 ((eh)->ip6e_len + 1) << 3); \ 256 if (error) \ 257 goto freehdrs; \ 258 } \ 259 } while (/*CONSTCOND*/ 0) 260 261 memset(&exthdrs, 0, sizeof(exthdrs)); 262 if (opt) { 263 /* Hop-by-Hop options header */ 264 MAKE_EXTHDR(opt->ip6po_hbh, &exthdrs.ip6e_hbh); 265 /* Destination options header(1st part) */ 266 MAKE_EXTHDR(opt->ip6po_dest1, &exthdrs.ip6e_dest1); 267 /* Routing header */ 268 MAKE_EXTHDR(opt->ip6po_rthdr, &exthdrs.ip6e_rthdr); 269 /* Destination options header(2nd part) */ 270 MAKE_EXTHDR(opt->ip6po_dest2, &exthdrs.ip6e_dest2); 271 } 272 273 /* 274 * Calculate the total length of the extension header chain. 275 * Keep the length of the unfragmentable part for fragmentation. 276 */ 277 optlen = 0; 278 if (exthdrs.ip6e_hbh) optlen += exthdrs.ip6e_hbh->m_len; 279 if (exthdrs.ip6e_dest1) optlen += exthdrs.ip6e_dest1->m_len; 280 if (exthdrs.ip6e_rthdr) optlen += exthdrs.ip6e_rthdr->m_len; 281 unfragpartlen = optlen + sizeof(struct ip6_hdr); 282 /* NOTE: we don't add AH/ESP length here. do that later. */ 283 if (exthdrs.ip6e_dest2) optlen += exthdrs.ip6e_dest2->m_len; 284 285 #ifdef IPSEC 286 if (ipsec_used) { 287 /* Check the security policy (SP) for the packet */ 288 289 sp = ipsec6_check_policy(m, so, flags, &needipsec, &error); 290 if (error != 0) { 291 /* 292 * Hack: -EINVAL is used to signal that a packet 293 * should be silently discarded. This is typically 294 * because we asked key management for an SA and 295 * it was delayed (e.g. kicked up to IKE). 296 */ 297 if (error == -EINVAL) 298 error = 0; 299 goto freehdrs; 300 } 301 } 302 #endif /* IPSEC */ 303 304 305 if (needipsec && 306 (m->m_pkthdr.csum_flags & (M_CSUM_UDPv6|M_CSUM_TCPv6)) != 0) { 307 in6_delayed_cksum(m); 308 m->m_pkthdr.csum_flags &= ~(M_CSUM_UDPv6|M_CSUM_TCPv6); 309 } 310 311 312 /* 313 * If we need IPsec, or there is at least one extension header, 314 * separate IP6 header from the payload. 315 */ 316 if ((needipsec || optlen) && !hdrsplit) { 317 if ((error = ip6_splithdr(m, &exthdrs)) != 0) { 318 m = NULL; 319 goto freehdrs; 320 } 321 m = exthdrs.ip6e_ip6; 322 hdrsplit++; 323 } 324 325 /* adjust pointer */ 326 ip6 = mtod(m, struct ip6_hdr *); 327 328 /* adjust mbuf packet header length */ 329 m->m_pkthdr.len += optlen; 330 plen = m->m_pkthdr.len - sizeof(*ip6); 331 332 /* If this is a jumbo payload, insert a jumbo payload option. */ 333 if (plen > IPV6_MAXPACKET) { 334 if (!hdrsplit) { 335 if ((error = ip6_splithdr(m, &exthdrs)) != 0) { 336 m = NULL; 337 goto freehdrs; 338 } 339 m = exthdrs.ip6e_ip6; 340 hdrsplit++; 341 } 342 /* adjust pointer */ 343 ip6 = mtod(m, struct ip6_hdr *); 344 if ((error = ip6_insert_jumboopt(&exthdrs, plen)) != 0) 345 goto freehdrs; 346 optlen += 8; /* XXX JUMBOOPTLEN */ 347 ip6->ip6_plen = 0; 348 } else 349 ip6->ip6_plen = htons(plen); 350 351 /* 352 * Concatenate headers and fill in next header fields. 353 * Here we have, on "m" 354 * IPv6 payload 355 * and we insert headers accordingly. Finally, we should be getting: 356 * IPv6 hbh dest1 rthdr ah* [esp* dest2 payload] 357 * 358 * during the header composing process, "m" points to IPv6 header. 359 * "mprev" points to an extension header prior to esp. 360 */ 361 { 362 u_char *nexthdrp = &ip6->ip6_nxt; 363 struct mbuf *mprev = m; 364 365 /* 366 * we treat dest2 specially. this makes IPsec processing 367 * much easier. the goal here is to make mprev point the 368 * mbuf prior to dest2. 369 * 370 * result: IPv6 dest2 payload 371 * m and mprev will point to IPv6 header. 372 */ 373 if (exthdrs.ip6e_dest2) { 374 if (!hdrsplit) 375 panic("assumption failed: hdr not split"); 376 exthdrs.ip6e_dest2->m_next = m->m_next; 377 m->m_next = exthdrs.ip6e_dest2; 378 *mtod(exthdrs.ip6e_dest2, u_char *) = ip6->ip6_nxt; 379 ip6->ip6_nxt = IPPROTO_DSTOPTS; 380 } 381 382 #define MAKE_CHAIN(m, mp, p, i)\ 383 do {\ 384 if (m) {\ 385 if (!hdrsplit) \ 386 panic("assumption failed: hdr not split"); \ 387 *mtod((m), u_char *) = *(p);\ 388 *(p) = (i);\ 389 p = mtod((m), u_char *);\ 390 (m)->m_next = (mp)->m_next;\ 391 (mp)->m_next = (m);\ 392 (mp) = (m);\ 393 }\ 394 } while (/*CONSTCOND*/ 0) 395 /* 396 * result: IPv6 hbh dest1 rthdr dest2 payload 397 * m will point to IPv6 header. mprev will point to the 398 * extension header prior to dest2 (rthdr in the above case). 399 */ 400 MAKE_CHAIN(exthdrs.ip6e_hbh, mprev, nexthdrp, IPPROTO_HOPOPTS); 401 MAKE_CHAIN(exthdrs.ip6e_dest1, mprev, nexthdrp, 402 IPPROTO_DSTOPTS); 403 MAKE_CHAIN(exthdrs.ip6e_rthdr, mprev, nexthdrp, 404 IPPROTO_ROUTING); 405 406 M_CSUM_DATA_IPv6_HL_SET(m->m_pkthdr.csum_data, 407 sizeof(struct ip6_hdr) + optlen); 408 } 409 410 /* Need to save for pmtu */ 411 finaldst = ip6->ip6_dst; 412 413 /* 414 * If there is a routing header, replace destination address field 415 * with the first hop of the routing header. 416 */ 417 if (exthdrs.ip6e_rthdr) { 418 struct ip6_rthdr *rh; 419 420 rh = (struct ip6_rthdr *)(mtod(exthdrs.ip6e_rthdr, 421 struct ip6_rthdr *)); 422 423 error = ip6_handle_rthdr(rh, ip6); 424 if (error != 0) 425 goto bad; 426 } 427 428 /* Source address validation */ 429 if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) && 430 (flags & IPV6_UNSPECSRC) == 0) { 431 error = EOPNOTSUPP; 432 IP6_STATINC(IP6_STAT_BADSCOPE); 433 goto bad; 434 } 435 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) { 436 error = EOPNOTSUPP; 437 IP6_STATINC(IP6_STAT_BADSCOPE); 438 goto bad; 439 } 440 441 IP6_STATINC(IP6_STAT_LOCALOUT); 442 443 /* 444 * Route packet. 445 */ 446 /* initialize cached route */ 447 if (ro == NULL) { 448 memset(&ip6route, 0, sizeof(ip6route)); 449 ro = &ip6route; 450 } 451 ro_pmtu = ro; 452 if (opt && opt->ip6po_rthdr) 453 ro = &opt->ip6po_route; 454 455 /* 456 * if specified, try to fill in the traffic class field. 457 * do not override if a non-zero value is already set. 458 * we check the diffserv field and the ecn field separately. 459 */ 460 if (opt && opt->ip6po_tclass >= 0) { 461 int mask = 0; 462 463 if ((ip6->ip6_flow & htonl(0xfc << 20)) == 0) 464 mask |= 0xfc; 465 if ((ip6->ip6_flow & htonl(0x03 << 20)) == 0) 466 mask |= 0x03; 467 if (mask != 0) 468 ip6->ip6_flow |= htonl((opt->ip6po_tclass & mask) << 20); 469 } 470 471 /* fill in or override the hop limit field, if necessary. */ 472 if (opt && opt->ip6po_hlim != -1) 473 ip6->ip6_hlim = opt->ip6po_hlim & 0xff; 474 else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) { 475 if (im6o != NULL) 476 ip6->ip6_hlim = im6o->im6o_multicast_hlim; 477 else 478 ip6->ip6_hlim = ip6_defmcasthlim; 479 } 480 481 #ifdef IPSEC 482 if (needipsec) { 483 int s = splsoftnet(); 484 error = ipsec6_process_packet(m, sp->req); 485 486 /* 487 * Preserve KAME behaviour: ENOENT can be returned 488 * when an SA acquire is in progress. Don't propagate 489 * this to user-level; it confuses applications. 490 * XXX this will go away when the SADB is redone. 491 */ 492 if (error == ENOENT) 493 error = 0; 494 splx(s); 495 goto done; 496 } 497 #endif /* IPSEC */ 498 499 /* adjust pointer */ 500 ip6 = mtod(m, struct ip6_hdr *); 501 502 sockaddr_in6_init(&dst_sa, &ip6->ip6_dst, 0, 0, 0); 503 504 /* We do not need a route for multicast */ 505 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) { 506 struct in6_pktinfo *pi = NULL; 507 508 /* 509 * If the outgoing interface for the address is specified by 510 * the caller, use it. 511 */ 512 if (opt && (pi = opt->ip6po_pktinfo) != NULL) { 513 /* XXX boundary check is assumed to be already done. */ 514 ifp = if_get_byindex(pi->ipi6_ifindex, &psref); 515 } else if (im6o != NULL) { 516 ifp = if_get_byindex(im6o->im6o_multicast_if_index, 517 &psref); 518 } 519 } 520 521 if (ifp == NULL) { 522 error = in6_selectroute(&dst_sa, opt, &ro, &rt, true); 523 if (error != 0) 524 goto bad; 525 ifp = if_get_byindex(rt->rt_ifp->if_index, &psref); 526 } 527 528 if (rt == NULL) { 529 /* 530 * If in6_selectroute() does not return a route entry, 531 * dst may not have been updated. 532 */ 533 error = rtcache_setdst(ro, sin6tosa(&dst_sa)); 534 if (error) { 535 goto bad; 536 } 537 } 538 539 /* 540 * then rt (for unicast) and ifp must be non-NULL valid values. 541 */ 542 if ((flags & IPV6_FORWARDING) == 0) { 543 /* XXX: the FORWARDING flag can be set for mrouting. */ 544 in6_ifstat_inc(ifp, ifs6_out_request); 545 } 546 if (rt != NULL) { 547 ia = (struct in6_ifaddr *)(rt->rt_ifa); 548 rt->rt_use++; 549 } 550 551 /* 552 * The outgoing interface must be in the zone of source and 553 * destination addresses. We should use ia_ifp to support the 554 * case of sending packets to an address of our own. 555 */ 556 if (ia != NULL && ia->ia_ifp) { 557 origifp = ia->ia_ifp; 558 if (if_is_deactivated(origifp)) 559 goto bad; 560 if_acquire_NOMPSAFE(origifp, &psref_ia); 561 release_psref_ia = true; 562 } else 563 origifp = ifp; 564 565 src0 = ip6->ip6_src; 566 if (in6_setscope(&src0, origifp, &zone)) 567 goto badscope; 568 sockaddr_in6_init(&src_sa, &ip6->ip6_src, 0, 0, 0); 569 if (sa6_recoverscope(&src_sa) || zone != src_sa.sin6_scope_id) 570 goto badscope; 571 572 dst0 = ip6->ip6_dst; 573 if (in6_setscope(&dst0, origifp, &zone)) 574 goto badscope; 575 /* re-initialize to be sure */ 576 sockaddr_in6_init(&dst_sa, &ip6->ip6_dst, 0, 0, 0); 577 if (sa6_recoverscope(&dst_sa) || zone != dst_sa.sin6_scope_id) 578 goto badscope; 579 580 /* scope check is done. */ 581 582 /* Ensure we only send from a valid address. */ 583 if ((error = ip6_ifaddrvalid(&src0)) != 0) { 584 nd6log(LOG_ERR, 585 "refusing to send from invalid address %s (pid %d)\n", 586 ip6_sprintf(&src0), curproc->p_pid); 587 IP6_STATINC(IP6_STAT_ODROPPED); 588 in6_ifstat_inc(origifp, ifs6_out_discard); 589 if (error == 1) 590 /* 591 * Address exists, but is tentative or detached. 592 * We can't send from it because it's invalid, 593 * so we drop the packet. 594 */ 595 error = 0; 596 else 597 error = EADDRNOTAVAIL; 598 goto bad; 599 } 600 601 if (rt != NULL && (rt->rt_flags & RTF_GATEWAY) && 602 !IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) 603 dst = satocsin6(rt->rt_gateway); 604 else 605 dst = satocsin6(rtcache_getdst(ro)); 606 607 /* 608 * XXXXXX: original code follows: 609 */ 610 if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) 611 m->m_flags &= ~(M_BCAST | M_MCAST); /* just in case */ 612 else { 613 struct in6_multi *in6m; 614 615 m->m_flags = (m->m_flags & ~M_BCAST) | M_MCAST; 616 617 in6_ifstat_inc(ifp, ifs6_out_mcast); 618 619 /* 620 * Confirm that the outgoing interface supports multicast. 621 */ 622 if (!(ifp->if_flags & IFF_MULTICAST)) { 623 IP6_STATINC(IP6_STAT_NOROUTE); 624 in6_ifstat_inc(ifp, ifs6_out_discard); 625 error = ENETUNREACH; 626 goto bad; 627 } 628 629 IN6_LOOKUP_MULTI(ip6->ip6_dst, ifp, in6m); 630 if (in6m != NULL && 631 (im6o == NULL || im6o->im6o_multicast_loop)) { 632 /* 633 * If we belong to the destination multicast group 634 * on the outgoing interface, and the caller did not 635 * forbid loopback, loop back a copy. 636 */ 637 KASSERT(dst != NULL); 638 ip6_mloopback(ifp, m, dst); 639 } else { 640 /* 641 * If we are acting as a multicast router, perform 642 * multicast forwarding as if the packet had just 643 * arrived on the interface to which we are about 644 * to send. The multicast forwarding function 645 * recursively calls this function, using the 646 * IPV6_FORWARDING flag to prevent infinite recursion. 647 * 648 * Multicasts that are looped back by ip6_mloopback(), 649 * above, will be forwarded by the ip6_input() routine, 650 * if necessary. 651 */ 652 if (ip6_mrouter && (flags & IPV6_FORWARDING) == 0) { 653 if (ip6_mforward(ip6, ifp, m) != 0) { 654 m_freem(m); 655 goto done; 656 } 657 } 658 } 659 /* 660 * Multicasts with a hoplimit of zero may be looped back, 661 * above, but must not be transmitted on a network. 662 * Also, multicasts addressed to the loopback interface 663 * are not sent -- the above call to ip6_mloopback() will 664 * loop back a copy if this host actually belongs to the 665 * destination group on the loopback interface. 666 */ 667 if (ip6->ip6_hlim == 0 || (ifp->if_flags & IFF_LOOPBACK) || 668 IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst)) { 669 m_freem(m); 670 goto done; 671 } 672 } 673 674 /* 675 * Fill the outgoing inteface to tell the upper layer 676 * to increment per-interface statistics. 677 */ 678 if (ifpp) 679 *ifpp = ifp; 680 681 /* Determine path MTU. */ 682 /* 683 * ro_pmtu represent final destination while 684 * ro might represent immediate destination. 685 * Use ro_pmtu destination since MTU might differ. 686 */ 687 if (ro_pmtu != ro) { 688 union { 689 struct sockaddr dst; 690 struct sockaddr_in6 dst6; 691 } u; 692 693 /* ro_pmtu may not have a cache */ 694 sockaddr_in6_init(&u.dst6, &finaldst, 0, 0, 0); 695 rt_pmtu = rtcache_lookup(ro_pmtu, &u.dst); 696 } else 697 rt_pmtu = rt; 698 error = ip6_getpmtu(rt_pmtu, ifp, &mtu, &alwaysfrag); 699 if (rt_pmtu != NULL && rt_pmtu != rt) 700 rtcache_unref(rt_pmtu, ro_pmtu); 701 if (error != 0) 702 goto bad; 703 704 /* 705 * The caller of this function may specify to use the minimum MTU 706 * in some cases. 707 * An advanced API option (IPV6_USE_MIN_MTU) can also override MTU 708 * setting. The logic is a bit complicated; by default, unicast 709 * packets will follow path MTU while multicast packets will be sent at 710 * the minimum MTU. If IP6PO_MINMTU_ALL is specified, all packets 711 * including unicast ones will be sent at the minimum MTU. Multicast 712 * packets will always be sent at the minimum MTU unless 713 * IP6PO_MINMTU_DISABLE is explicitly specified. 714 * See RFC 3542 for more details. 715 */ 716 if (mtu > IPV6_MMTU) { 717 if ((flags & IPV6_MINMTU)) 718 mtu = IPV6_MMTU; 719 else if (opt && opt->ip6po_minmtu == IP6PO_MINMTU_ALL) 720 mtu = IPV6_MMTU; 721 else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) && 722 (opt == NULL || 723 opt->ip6po_minmtu != IP6PO_MINMTU_DISABLE)) { 724 mtu = IPV6_MMTU; 725 } 726 } 727 728 /* 729 * clear embedded scope identifiers if necessary. 730 * in6_clearscope will touch the addresses only when necessary. 731 */ 732 in6_clearscope(&ip6->ip6_src); 733 in6_clearscope(&ip6->ip6_dst); 734 735 /* 736 * If the outgoing packet contains a hop-by-hop options header, 737 * it must be examined and processed even by the source node. 738 * (RFC 2460, section 4.) 739 */ 740 if (ip6->ip6_nxt == IPV6_HOPOPTS) { 741 u_int32_t dummy1; /* XXX unused */ 742 u_int32_t dummy2; /* XXX unused */ 743 int hoff = sizeof(struct ip6_hdr); 744 745 if (ip6_hopopts_input(&dummy1, &dummy2, &m, &hoff)) { 746 /* m was already freed at this point */ 747 error = EINVAL;/* better error? */ 748 goto done; 749 } 750 751 ip6 = mtod(m, struct ip6_hdr *); 752 } 753 754 /* 755 * Run through list of hooks for output packets. 756 */ 757 if ((error = pfil_run_hooks(inet6_pfil_hook, &m, ifp, PFIL_OUT)) != 0) 758 goto done; 759 if (m == NULL) 760 goto done; 761 ip6 = mtod(m, struct ip6_hdr *); 762 763 /* 764 * Send the packet to the outgoing interface. 765 * If necessary, do IPv6 fragmentation before sending. 766 * 767 * the logic here is rather complex: 768 * 1: normal case (dontfrag == 0, alwaysfrag == 0) 769 * 1-a: send as is if tlen <= path mtu 770 * 1-b: fragment if tlen > path mtu 771 * 772 * 2: if user asks us not to fragment (dontfrag == 1) 773 * 2-a: send as is if tlen <= interface mtu 774 * 2-b: error if tlen > interface mtu 775 * 776 * 3: if we always need to attach fragment header (alwaysfrag == 1) 777 * always fragment 778 * 779 * 4: if dontfrag == 1 && alwaysfrag == 1 780 * error, as we cannot handle this conflicting request 781 */ 782 tlen = m->m_pkthdr.len; 783 tso = (m->m_pkthdr.csum_flags & M_CSUM_TSOv6) != 0; 784 if (opt && (opt->ip6po_flags & IP6PO_DONTFRAG)) 785 dontfrag = 1; 786 else 787 dontfrag = 0; 788 789 if (dontfrag && alwaysfrag) { /* case 4 */ 790 /* conflicting request - can't transmit */ 791 error = EMSGSIZE; 792 goto bad; 793 } 794 if (dontfrag && (!tso && tlen > IN6_LINKMTU(ifp))) { /* case 2-b */ 795 /* 796 * Even if the DONTFRAG option is specified, we cannot send the 797 * packet when the data length is larger than the MTU of the 798 * outgoing interface. 799 * Notify the error by sending IPV6_PATHMTU ancillary data as 800 * well as returning an error code (the latter is not described 801 * in the API spec.) 802 */ 803 u_int32_t mtu32; 804 struct ip6ctlparam ip6cp; 805 806 mtu32 = (u_int32_t)mtu; 807 memset(&ip6cp, 0, sizeof(ip6cp)); 808 ip6cp.ip6c_cmdarg = (void *)&mtu32; 809 pfctlinput2(PRC_MSGSIZE, 810 rtcache_getdst(ro_pmtu), &ip6cp); 811 812 error = EMSGSIZE; 813 goto bad; 814 } 815 816 /* 817 * transmit packet without fragmentation 818 */ 819 if (dontfrag || (!alwaysfrag && (tlen <= mtu || tso))) { 820 /* case 1-a and 2-a */ 821 struct in6_ifaddr *ia6; 822 int sw_csum; 823 int s; 824 825 ip6 = mtod(m, struct ip6_hdr *); 826 s = pserialize_read_enter(); 827 ia6 = in6_ifawithifp(ifp, &ip6->ip6_src); 828 if (ia6) { 829 /* Record statistics for this interface address. */ 830 ia6->ia_ifa.ifa_data.ifad_outbytes += m->m_pkthdr.len; 831 } 832 pserialize_read_exit(s); 833 834 sw_csum = m->m_pkthdr.csum_flags & ~ifp->if_csum_flags_tx; 835 if ((sw_csum & (M_CSUM_UDPv6|M_CSUM_TCPv6)) != 0) { 836 if (IN6_NEED_CHECKSUM(ifp, 837 sw_csum & (M_CSUM_UDPv6|M_CSUM_TCPv6))) { 838 in6_delayed_cksum(m); 839 } 840 m->m_pkthdr.csum_flags &= ~(M_CSUM_UDPv6|M_CSUM_TCPv6); 841 } 842 843 KASSERT(dst != NULL); 844 if (__predict_true(!tso || 845 (ifp->if_capenable & IFCAP_TSOv6) != 0)) { 846 error = nd6_output(ifp, origifp, m, dst, rt); 847 } else { 848 error = ip6_tso_output(ifp, origifp, m, dst, rt); 849 } 850 goto done; 851 } 852 853 if (tso) { 854 error = EINVAL; /* XXX */ 855 goto bad; 856 } 857 858 /* 859 * try to fragment the packet. case 1-b and 3 860 */ 861 if (mtu < IPV6_MMTU) { 862 /* path MTU cannot be less than IPV6_MMTU */ 863 error = EMSGSIZE; 864 in6_ifstat_inc(ifp, ifs6_out_fragfail); 865 goto bad; 866 } else if (ip6->ip6_plen == 0) { 867 /* jumbo payload cannot be fragmented */ 868 error = EMSGSIZE; 869 in6_ifstat_inc(ifp, ifs6_out_fragfail); 870 goto bad; 871 } else { 872 struct mbuf **mnext, *m_frgpart; 873 struct ip6_frag *ip6f; 874 u_int32_t id = htonl(ip6_randomid()); 875 u_char nextproto; 876 #if 0 /* see below */ 877 struct ip6ctlparam ip6cp; 878 u_int32_t mtu32; 879 #endif 880 881 /* 882 * Too large for the destination or interface; 883 * fragment if possible. 884 * Must be able to put at least 8 bytes per fragment. 885 */ 886 hlen = unfragpartlen; 887 if (mtu > IPV6_MAXPACKET) 888 mtu = IPV6_MAXPACKET; 889 890 #if 0 891 /* 892 * It is believed this code is a leftover from the 893 * development of the IPV6_RECVPATHMTU sockopt and 894 * associated work to implement RFC3542. 895 * It's not entirely clear what the intent of the API 896 * is at this point, so disable this code for now. 897 * The IPV6_RECVPATHMTU sockopt and/or IPV6_DONTFRAG 898 * will send notifications if the application requests. 899 */ 900 901 /* Notify a proper path MTU to applications. */ 902 mtu32 = (u_int32_t)mtu; 903 memset(&ip6cp, 0, sizeof(ip6cp)); 904 ip6cp.ip6c_cmdarg = (void *)&mtu32; 905 pfctlinput2(PRC_MSGSIZE, 906 rtcache_getdst(ro_pmtu), &ip6cp); 907 #endif 908 909 len = (mtu - hlen - sizeof(struct ip6_frag)) & ~7; 910 if (len < 8) { 911 error = EMSGSIZE; 912 in6_ifstat_inc(ifp, ifs6_out_fragfail); 913 goto bad; 914 } 915 916 mnext = &m->m_nextpkt; 917 918 /* 919 * Change the next header field of the last header in the 920 * unfragmentable part. 921 */ 922 if (exthdrs.ip6e_rthdr) { 923 nextproto = *mtod(exthdrs.ip6e_rthdr, u_char *); 924 *mtod(exthdrs.ip6e_rthdr, u_char *) = IPPROTO_FRAGMENT; 925 } else if (exthdrs.ip6e_dest1) { 926 nextproto = *mtod(exthdrs.ip6e_dest1, u_char *); 927 *mtod(exthdrs.ip6e_dest1, u_char *) = IPPROTO_FRAGMENT; 928 } else if (exthdrs.ip6e_hbh) { 929 nextproto = *mtod(exthdrs.ip6e_hbh, u_char *); 930 *mtod(exthdrs.ip6e_hbh, u_char *) = IPPROTO_FRAGMENT; 931 } else { 932 nextproto = ip6->ip6_nxt; 933 ip6->ip6_nxt = IPPROTO_FRAGMENT; 934 } 935 936 if ((m->m_pkthdr.csum_flags & (M_CSUM_UDPv6|M_CSUM_TCPv6)) 937 != 0) { 938 if (IN6_NEED_CHECKSUM(ifp, 939 m->m_pkthdr.csum_flags & 940 (M_CSUM_UDPv6|M_CSUM_TCPv6))) { 941 in6_delayed_cksum(m); 942 } 943 m->m_pkthdr.csum_flags &= ~(M_CSUM_UDPv6|M_CSUM_TCPv6); 944 } 945 946 /* 947 * Loop through length of segment after first fragment, 948 * make new header and copy data of each part and link onto 949 * chain. 950 */ 951 m0 = m; 952 for (off = hlen; off < tlen; off += len) { 953 struct mbuf *mlast; 954 955 MGETHDR(m, M_DONTWAIT, MT_HEADER); 956 if (!m) { 957 error = ENOBUFS; 958 IP6_STATINC(IP6_STAT_ODROPPED); 959 goto sendorfree; 960 } 961 m_reset_rcvif(m); 962 m->m_flags = m0->m_flags & M_COPYFLAGS; 963 *mnext = m; 964 mnext = &m->m_nextpkt; 965 m->m_data += max_linkhdr; 966 mhip6 = mtod(m, struct ip6_hdr *); 967 *mhip6 = *ip6; 968 m->m_len = sizeof(*mhip6); 969 /* 970 * ip6f must be valid if error is 0. But how 971 * can a compiler be expected to infer this? 972 */ 973 ip6f = NULL; 974 error = ip6_insertfraghdr(m0, m, hlen, &ip6f); 975 if (error) { 976 IP6_STATINC(IP6_STAT_ODROPPED); 977 goto sendorfree; 978 } 979 ip6f->ip6f_offlg = htons((u_int16_t)((off - hlen) & ~7)); 980 if (off + len >= tlen) 981 len = tlen - off; 982 else 983 ip6f->ip6f_offlg |= IP6F_MORE_FRAG; 984 mhip6->ip6_plen = htons((u_int16_t)(len + hlen + 985 sizeof(*ip6f) - sizeof(struct ip6_hdr))); 986 if ((m_frgpart = m_copy(m0, off, len)) == 0) { 987 error = ENOBUFS; 988 IP6_STATINC(IP6_STAT_ODROPPED); 989 goto sendorfree; 990 } 991 for (mlast = m; mlast->m_next; mlast = mlast->m_next) 992 ; 993 mlast->m_next = m_frgpart; 994 m->m_pkthdr.len = len + hlen + sizeof(*ip6f); 995 m_reset_rcvif(m); 996 ip6f->ip6f_reserved = 0; 997 ip6f->ip6f_ident = id; 998 ip6f->ip6f_nxt = nextproto; 999 IP6_STATINC(IP6_STAT_OFRAGMENTS); 1000 in6_ifstat_inc(ifp, ifs6_out_fragcreat); 1001 } 1002 1003 in6_ifstat_inc(ifp, ifs6_out_fragok); 1004 } 1005 1006 /* 1007 * Remove leading garbages. 1008 */ 1009 sendorfree: 1010 m = m0->m_nextpkt; 1011 m0->m_nextpkt = 0; 1012 m_freem(m0); 1013 for (m0 = m; m; m = m0) { 1014 m0 = m->m_nextpkt; 1015 m->m_nextpkt = 0; 1016 if (error == 0) { 1017 struct in6_ifaddr *ia6; 1018 int s; 1019 ip6 = mtod(m, struct ip6_hdr *); 1020 s = pserialize_read_enter(); 1021 ia6 = in6_ifawithifp(ifp, &ip6->ip6_src); 1022 if (ia6) { 1023 /* 1024 * Record statistics for this interface 1025 * address. 1026 */ 1027 ia6->ia_ifa.ifa_data.ifad_outbytes += 1028 m->m_pkthdr.len; 1029 } 1030 pserialize_read_exit(s); 1031 KASSERT(dst != NULL); 1032 error = nd6_output(ifp, origifp, m, dst, rt); 1033 } else 1034 m_freem(m); 1035 } 1036 1037 if (error == 0) 1038 IP6_STATINC(IP6_STAT_FRAGMENTED); 1039 1040 done: 1041 rtcache_unref(rt, ro); 1042 if (ro == &ip6route) 1043 rtcache_free(&ip6route); 1044 1045 #ifdef IPSEC 1046 if (sp != NULL) 1047 KEY_FREESP(&sp); 1048 #endif /* IPSEC */ 1049 1050 if_put(ifp, &psref); 1051 if (release_psref_ia) 1052 if_put(origifp, &psref_ia); 1053 curlwp_bindx(bound); 1054 1055 return (error); 1056 1057 freehdrs: 1058 m_freem(exthdrs.ip6e_hbh); /* m_freem will check if mbuf is 0 */ 1059 m_freem(exthdrs.ip6e_dest1); 1060 m_freem(exthdrs.ip6e_rthdr); 1061 m_freem(exthdrs.ip6e_dest2); 1062 /* FALLTHROUGH */ 1063 bad: 1064 m_freem(m); 1065 goto done; 1066 badscope: 1067 IP6_STATINC(IP6_STAT_BADSCOPE); 1068 in6_ifstat_inc(origifp, ifs6_out_discard); 1069 if (error == 0) 1070 error = EHOSTUNREACH; /* XXX */ 1071 goto bad; 1072 } 1073 1074 static int 1075 ip6_copyexthdr(struct mbuf **mp, void *hdr, int hlen) 1076 { 1077 struct mbuf *m; 1078 1079 if (hlen > MCLBYTES) 1080 return (ENOBUFS); /* XXX */ 1081 1082 MGET(m, M_DONTWAIT, MT_DATA); 1083 if (!m) 1084 return (ENOBUFS); 1085 1086 if (hlen > MLEN) { 1087 MCLGET(m, M_DONTWAIT); 1088 if ((m->m_flags & M_EXT) == 0) { 1089 m_free(m); 1090 return (ENOBUFS); 1091 } 1092 } 1093 m->m_len = hlen; 1094 if (hdr) 1095 bcopy(hdr, mtod(m, void *), hlen); 1096 1097 *mp = m; 1098 return (0); 1099 } 1100 1101 /* 1102 * Process a delayed payload checksum calculation. 1103 */ 1104 void 1105 in6_delayed_cksum(struct mbuf *m) 1106 { 1107 uint16_t csum, offset; 1108 1109 KASSERT((m->m_pkthdr.csum_flags & (M_CSUM_UDPv6|M_CSUM_TCPv6)) != 0); 1110 KASSERT((~m->m_pkthdr.csum_flags & (M_CSUM_UDPv6|M_CSUM_TCPv6)) != 0); 1111 KASSERT((m->m_pkthdr.csum_flags 1112 & (M_CSUM_UDPv4|M_CSUM_TCPv4|M_CSUM_TSOv4)) == 0); 1113 1114 offset = M_CSUM_DATA_IPv6_HL(m->m_pkthdr.csum_data); 1115 csum = in6_cksum(m, 0, offset, m->m_pkthdr.len - offset); 1116 if (csum == 0 && (m->m_pkthdr.csum_flags & M_CSUM_UDPv6) != 0) { 1117 csum = 0xffff; 1118 } 1119 1120 offset += M_CSUM_DATA_IPv6_OFFSET(m->m_pkthdr.csum_data); 1121 if ((offset + sizeof(csum)) > m->m_len) { 1122 m_copyback(m, offset, sizeof(csum), &csum); 1123 } else { 1124 *(uint16_t *)(mtod(m, char *) + offset) = csum; 1125 } 1126 } 1127 1128 /* 1129 * Insert jumbo payload option. 1130 */ 1131 static int 1132 ip6_insert_jumboopt(struct ip6_exthdrs *exthdrs, u_int32_t plen) 1133 { 1134 struct mbuf *mopt; 1135 u_int8_t *optbuf; 1136 u_int32_t v; 1137 1138 #define JUMBOOPTLEN 8 /* length of jumbo payload option and padding */ 1139 1140 /* 1141 * If there is no hop-by-hop options header, allocate new one. 1142 * If there is one but it doesn't have enough space to store the 1143 * jumbo payload option, allocate a cluster to store the whole options. 1144 * Otherwise, use it to store the options. 1145 */ 1146 if (exthdrs->ip6e_hbh == 0) { 1147 MGET(mopt, M_DONTWAIT, MT_DATA); 1148 if (mopt == 0) 1149 return (ENOBUFS); 1150 mopt->m_len = JUMBOOPTLEN; 1151 optbuf = mtod(mopt, u_int8_t *); 1152 optbuf[1] = 0; /* = ((JUMBOOPTLEN) >> 3) - 1 */ 1153 exthdrs->ip6e_hbh = mopt; 1154 } else { 1155 struct ip6_hbh *hbh; 1156 1157 mopt = exthdrs->ip6e_hbh; 1158 if (M_TRAILINGSPACE(mopt) < JUMBOOPTLEN) { 1159 /* 1160 * XXX assumption: 1161 * - exthdrs->ip6e_hbh is not referenced from places 1162 * other than exthdrs. 1163 * - exthdrs->ip6e_hbh is not an mbuf chain. 1164 */ 1165 int oldoptlen = mopt->m_len; 1166 struct mbuf *n; 1167 1168 /* 1169 * XXX: give up if the whole (new) hbh header does 1170 * not fit even in an mbuf cluster. 1171 */ 1172 if (oldoptlen + JUMBOOPTLEN > MCLBYTES) 1173 return (ENOBUFS); 1174 1175 /* 1176 * As a consequence, we must always prepare a cluster 1177 * at this point. 1178 */ 1179 MGET(n, M_DONTWAIT, MT_DATA); 1180 if (n) { 1181 MCLGET(n, M_DONTWAIT); 1182 if ((n->m_flags & M_EXT) == 0) { 1183 m_freem(n); 1184 n = NULL; 1185 } 1186 } 1187 if (!n) 1188 return (ENOBUFS); 1189 n->m_len = oldoptlen + JUMBOOPTLEN; 1190 bcopy(mtod(mopt, void *), mtod(n, void *), 1191 oldoptlen); 1192 optbuf = mtod(n, u_int8_t *) + oldoptlen; 1193 m_freem(mopt); 1194 mopt = exthdrs->ip6e_hbh = n; 1195 } else { 1196 optbuf = mtod(mopt, u_int8_t *) + mopt->m_len; 1197 mopt->m_len += JUMBOOPTLEN; 1198 } 1199 optbuf[0] = IP6OPT_PADN; 1200 optbuf[1] = 0; 1201 1202 /* 1203 * Adjust the header length according to the pad and 1204 * the jumbo payload option. 1205 */ 1206 hbh = mtod(mopt, struct ip6_hbh *); 1207 hbh->ip6h_len += (JUMBOOPTLEN >> 3); 1208 } 1209 1210 /* fill in the option. */ 1211 optbuf[2] = IP6OPT_JUMBO; 1212 optbuf[3] = 4; 1213 v = (u_int32_t)htonl(plen + JUMBOOPTLEN); 1214 bcopy(&v, &optbuf[4], sizeof(u_int32_t)); 1215 1216 /* finally, adjust the packet header length */ 1217 exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN; 1218 1219 return (0); 1220 #undef JUMBOOPTLEN 1221 } 1222 1223 /* 1224 * Insert fragment header and copy unfragmentable header portions. 1225 * 1226 * *frghdrp will not be read, and it is guaranteed that either an 1227 * error is returned or that *frghdrp will point to space allocated 1228 * for the fragment header. 1229 */ 1230 static int 1231 ip6_insertfraghdr(struct mbuf *m0, struct mbuf *m, int hlen, 1232 struct ip6_frag **frghdrp) 1233 { 1234 struct mbuf *n, *mlast; 1235 1236 if (hlen > sizeof(struct ip6_hdr)) { 1237 n = m_copym(m0, sizeof(struct ip6_hdr), 1238 hlen - sizeof(struct ip6_hdr), M_DONTWAIT); 1239 if (n == 0) 1240 return (ENOBUFS); 1241 m->m_next = n; 1242 } else 1243 n = m; 1244 1245 /* Search for the last mbuf of unfragmentable part. */ 1246 for (mlast = n; mlast->m_next; mlast = mlast->m_next) 1247 ; 1248 1249 if ((mlast->m_flags & M_EXT) == 0 && 1250 M_TRAILINGSPACE(mlast) >= sizeof(struct ip6_frag)) { 1251 /* use the trailing space of the last mbuf for the fragment hdr */ 1252 *frghdrp = (struct ip6_frag *)(mtod(mlast, char *) + 1253 mlast->m_len); 1254 mlast->m_len += sizeof(struct ip6_frag); 1255 m->m_pkthdr.len += sizeof(struct ip6_frag); 1256 } else { 1257 /* allocate a new mbuf for the fragment header */ 1258 struct mbuf *mfrg; 1259 1260 MGET(mfrg, M_DONTWAIT, MT_DATA); 1261 if (mfrg == 0) 1262 return (ENOBUFS); 1263 mfrg->m_len = sizeof(struct ip6_frag); 1264 *frghdrp = mtod(mfrg, struct ip6_frag *); 1265 mlast->m_next = mfrg; 1266 } 1267 1268 return (0); 1269 } 1270 1271 static int 1272 ip6_getpmtu(struct rtentry *rt, struct ifnet *ifp, u_long *mtup, 1273 int *alwaysfragp) 1274 { 1275 u_int32_t mtu = 0; 1276 int alwaysfrag = 0; 1277 int error = 0; 1278 1279 if (rt != NULL) { 1280 u_int32_t ifmtu; 1281 1282 if (ifp == NULL) 1283 ifp = rt->rt_ifp; 1284 ifmtu = IN6_LINKMTU(ifp); 1285 mtu = rt->rt_rmx.rmx_mtu; 1286 if (mtu == 0) 1287 mtu = ifmtu; 1288 else if (mtu < IPV6_MMTU) { 1289 /* 1290 * RFC2460 section 5, last paragraph: 1291 * if we record ICMPv6 too big message with 1292 * mtu < IPV6_MMTU, transmit packets sized IPV6_MMTU 1293 * or smaller, with fragment header attached. 1294 * (fragment header is needed regardless from the 1295 * packet size, for translators to identify packets) 1296 */ 1297 alwaysfrag = 1; 1298 mtu = IPV6_MMTU; 1299 } else if (mtu > ifmtu) { 1300 /* 1301 * The MTU on the route is larger than the MTU on 1302 * the interface! This shouldn't happen, unless the 1303 * MTU of the interface has been changed after the 1304 * interface was brought up. Change the MTU in the 1305 * route to match the interface MTU (as long as the 1306 * field isn't locked). 1307 */ 1308 mtu = ifmtu; 1309 if (!(rt->rt_rmx.rmx_locks & RTV_MTU)) 1310 rt->rt_rmx.rmx_mtu = mtu; 1311 } 1312 } else if (ifp) { 1313 mtu = IN6_LINKMTU(ifp); 1314 } else 1315 error = EHOSTUNREACH; /* XXX */ 1316 1317 *mtup = mtu; 1318 if (alwaysfragp) 1319 *alwaysfragp = alwaysfrag; 1320 return (error); 1321 } 1322 1323 /* 1324 * IP6 socket option processing. 1325 */ 1326 int 1327 ip6_ctloutput(int op, struct socket *so, struct sockopt *sopt) 1328 { 1329 int optdatalen, uproto; 1330 void *optdata; 1331 struct in6pcb *in6p = sotoin6pcb(so); 1332 struct ip_moptions **mopts; 1333 int error, optval; 1334 int level, optname; 1335 1336 KASSERT(sopt != NULL); 1337 1338 level = sopt->sopt_level; 1339 optname = sopt->sopt_name; 1340 1341 error = optval = 0; 1342 uproto = (int)so->so_proto->pr_protocol; 1343 1344 switch (level) { 1345 case IPPROTO_IP: 1346 switch (optname) { 1347 case IP_ADD_MEMBERSHIP: 1348 case IP_DROP_MEMBERSHIP: 1349 case IP_MULTICAST_IF: 1350 case IP_MULTICAST_LOOP: 1351 case IP_MULTICAST_TTL: 1352 mopts = &in6p->in6p_v4moptions; 1353 switch (op) { 1354 case PRCO_GETOPT: 1355 return ip_getmoptions(*mopts, sopt); 1356 case PRCO_SETOPT: 1357 return ip_setmoptions(mopts, sopt); 1358 default: 1359 return EINVAL; 1360 } 1361 default: 1362 return ENOPROTOOPT; 1363 } 1364 case IPPROTO_IPV6: 1365 break; 1366 default: 1367 return ENOPROTOOPT; 1368 } 1369 switch (op) { 1370 case PRCO_SETOPT: 1371 switch (optname) { 1372 #ifdef RFC2292 1373 case IPV6_2292PKTOPTIONS: 1374 error = ip6_pcbopts(&in6p->in6p_outputopts, so, sopt); 1375 break; 1376 #endif 1377 1378 /* 1379 * Use of some Hop-by-Hop options or some 1380 * Destination options, might require special 1381 * privilege. That is, normal applications 1382 * (without special privilege) might be forbidden 1383 * from setting certain options in outgoing packets, 1384 * and might never see certain options in received 1385 * packets. [RFC 2292 Section 6] 1386 * KAME specific note: 1387 * KAME prevents non-privileged users from sending or 1388 * receiving ANY hbh/dst options in order to avoid 1389 * overhead of parsing options in the kernel. 1390 */ 1391 case IPV6_RECVHOPOPTS: 1392 case IPV6_RECVDSTOPTS: 1393 case IPV6_RECVRTHDRDSTOPTS: 1394 error = kauth_authorize_network(kauth_cred_get(), 1395 KAUTH_NETWORK_IPV6, KAUTH_REQ_NETWORK_IPV6_HOPBYHOP, 1396 NULL, NULL, NULL); 1397 if (error) 1398 break; 1399 /* FALLTHROUGH */ 1400 case IPV6_UNICAST_HOPS: 1401 case IPV6_HOPLIMIT: 1402 case IPV6_FAITH: 1403 1404 case IPV6_RECVPKTINFO: 1405 case IPV6_RECVHOPLIMIT: 1406 case IPV6_RECVRTHDR: 1407 case IPV6_RECVPATHMTU: 1408 case IPV6_RECVTCLASS: 1409 case IPV6_V6ONLY: 1410 error = sockopt_getint(sopt, &optval); 1411 if (error) 1412 break; 1413 switch (optname) { 1414 case IPV6_UNICAST_HOPS: 1415 if (optval < -1 || optval >= 256) 1416 error = EINVAL; 1417 else { 1418 /* -1 = kernel default */ 1419 in6p->in6p_hops = optval; 1420 } 1421 break; 1422 #define OPTSET(bit) \ 1423 do { \ 1424 if (optval) \ 1425 in6p->in6p_flags |= (bit); \ 1426 else \ 1427 in6p->in6p_flags &= ~(bit); \ 1428 } while (/*CONSTCOND*/ 0) 1429 1430 #ifdef RFC2292 1431 #define OPTSET2292(bit) \ 1432 do { \ 1433 in6p->in6p_flags |= IN6P_RFC2292; \ 1434 if (optval) \ 1435 in6p->in6p_flags |= (bit); \ 1436 else \ 1437 in6p->in6p_flags &= ~(bit); \ 1438 } while (/*CONSTCOND*/ 0) 1439 #endif 1440 1441 #define OPTBIT(bit) (in6p->in6p_flags & (bit) ? 1 : 0) 1442 1443 case IPV6_RECVPKTINFO: 1444 #ifdef RFC2292 1445 /* cannot mix with RFC2292 */ 1446 if (OPTBIT(IN6P_RFC2292)) { 1447 error = EINVAL; 1448 break; 1449 } 1450 #endif 1451 OPTSET(IN6P_PKTINFO); 1452 break; 1453 1454 case IPV6_HOPLIMIT: 1455 { 1456 struct ip6_pktopts **optp; 1457 1458 #ifdef RFC2292 1459 /* cannot mix with RFC2292 */ 1460 if (OPTBIT(IN6P_RFC2292)) { 1461 error = EINVAL; 1462 break; 1463 } 1464 #endif 1465 optp = &in6p->in6p_outputopts; 1466 error = ip6_pcbopt(IPV6_HOPLIMIT, 1467 (u_char *)&optval, 1468 sizeof(optval), 1469 optp, 1470 kauth_cred_get(), uproto); 1471 break; 1472 } 1473 1474 case IPV6_RECVHOPLIMIT: 1475 #ifdef RFC2292 1476 /* cannot mix with RFC2292 */ 1477 if (OPTBIT(IN6P_RFC2292)) { 1478 error = EINVAL; 1479 break; 1480 } 1481 #endif 1482 OPTSET(IN6P_HOPLIMIT); 1483 break; 1484 1485 case IPV6_RECVHOPOPTS: 1486 #ifdef RFC2292 1487 /* cannot mix with RFC2292 */ 1488 if (OPTBIT(IN6P_RFC2292)) { 1489 error = EINVAL; 1490 break; 1491 } 1492 #endif 1493 OPTSET(IN6P_HOPOPTS); 1494 break; 1495 1496 case IPV6_RECVDSTOPTS: 1497 #ifdef RFC2292 1498 /* cannot mix with RFC2292 */ 1499 if (OPTBIT(IN6P_RFC2292)) { 1500 error = EINVAL; 1501 break; 1502 } 1503 #endif 1504 OPTSET(IN6P_DSTOPTS); 1505 break; 1506 1507 case IPV6_RECVRTHDRDSTOPTS: 1508 #ifdef RFC2292 1509 /* cannot mix with RFC2292 */ 1510 if (OPTBIT(IN6P_RFC2292)) { 1511 error = EINVAL; 1512 break; 1513 } 1514 #endif 1515 OPTSET(IN6P_RTHDRDSTOPTS); 1516 break; 1517 1518 case IPV6_RECVRTHDR: 1519 #ifdef RFC2292 1520 /* cannot mix with RFC2292 */ 1521 if (OPTBIT(IN6P_RFC2292)) { 1522 error = EINVAL; 1523 break; 1524 } 1525 #endif 1526 OPTSET(IN6P_RTHDR); 1527 break; 1528 1529 case IPV6_FAITH: 1530 OPTSET(IN6P_FAITH); 1531 break; 1532 1533 case IPV6_RECVPATHMTU: 1534 /* 1535 * We ignore this option for TCP 1536 * sockets. 1537 * (RFC3542 leaves this case 1538 * unspecified.) 1539 */ 1540 if (uproto != IPPROTO_TCP) 1541 OPTSET(IN6P_MTU); 1542 break; 1543 1544 case IPV6_V6ONLY: 1545 /* 1546 * make setsockopt(IPV6_V6ONLY) 1547 * available only prior to bind(2). 1548 * see ipng mailing list, Jun 22 2001. 1549 */ 1550 if (in6p->in6p_lport || 1551 !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr)) { 1552 error = EINVAL; 1553 break; 1554 } 1555 #ifdef INET6_BINDV6ONLY 1556 if (!optval) 1557 error = EINVAL; 1558 #else 1559 OPTSET(IN6P_IPV6_V6ONLY); 1560 #endif 1561 break; 1562 case IPV6_RECVTCLASS: 1563 #ifdef RFC2292 1564 /* cannot mix with RFC2292 XXX */ 1565 if (OPTBIT(IN6P_RFC2292)) { 1566 error = EINVAL; 1567 break; 1568 } 1569 #endif 1570 OPTSET(IN6P_TCLASS); 1571 break; 1572 1573 } 1574 break; 1575 1576 case IPV6_OTCLASS: 1577 { 1578 struct ip6_pktopts **optp; 1579 u_int8_t tclass; 1580 1581 error = sockopt_get(sopt, &tclass, sizeof(tclass)); 1582 if (error) 1583 break; 1584 optp = &in6p->in6p_outputopts; 1585 error = ip6_pcbopt(optname, 1586 (u_char *)&tclass, 1587 sizeof(tclass), 1588 optp, 1589 kauth_cred_get(), uproto); 1590 break; 1591 } 1592 1593 case IPV6_TCLASS: 1594 case IPV6_DONTFRAG: 1595 case IPV6_USE_MIN_MTU: 1596 case IPV6_PREFER_TEMPADDR: 1597 error = sockopt_getint(sopt, &optval); 1598 if (error) 1599 break; 1600 { 1601 struct ip6_pktopts **optp; 1602 optp = &in6p->in6p_outputopts; 1603 error = ip6_pcbopt(optname, 1604 (u_char *)&optval, 1605 sizeof(optval), 1606 optp, 1607 kauth_cred_get(), uproto); 1608 break; 1609 } 1610 1611 #ifdef RFC2292 1612 case IPV6_2292PKTINFO: 1613 case IPV6_2292HOPLIMIT: 1614 case IPV6_2292HOPOPTS: 1615 case IPV6_2292DSTOPTS: 1616 case IPV6_2292RTHDR: 1617 /* RFC 2292 */ 1618 error = sockopt_getint(sopt, &optval); 1619 if (error) 1620 break; 1621 1622 switch (optname) { 1623 case IPV6_2292PKTINFO: 1624 OPTSET2292(IN6P_PKTINFO); 1625 break; 1626 case IPV6_2292HOPLIMIT: 1627 OPTSET2292(IN6P_HOPLIMIT); 1628 break; 1629 case IPV6_2292HOPOPTS: 1630 /* 1631 * Check super-user privilege. 1632 * See comments for IPV6_RECVHOPOPTS. 1633 */ 1634 error = 1635 kauth_authorize_network(kauth_cred_get(), 1636 KAUTH_NETWORK_IPV6, 1637 KAUTH_REQ_NETWORK_IPV6_HOPBYHOP, NULL, 1638 NULL, NULL); 1639 if (error) 1640 return (error); 1641 OPTSET2292(IN6P_HOPOPTS); 1642 break; 1643 case IPV6_2292DSTOPTS: 1644 error = 1645 kauth_authorize_network(kauth_cred_get(), 1646 KAUTH_NETWORK_IPV6, 1647 KAUTH_REQ_NETWORK_IPV6_HOPBYHOP, NULL, 1648 NULL, NULL); 1649 if (error) 1650 return (error); 1651 OPTSET2292(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); /* XXX */ 1652 break; 1653 case IPV6_2292RTHDR: 1654 OPTSET2292(IN6P_RTHDR); 1655 break; 1656 } 1657 break; 1658 #endif 1659 case IPV6_PKTINFO: 1660 case IPV6_HOPOPTS: 1661 case IPV6_RTHDR: 1662 case IPV6_DSTOPTS: 1663 case IPV6_RTHDRDSTOPTS: 1664 case IPV6_NEXTHOP: { 1665 /* new advanced API (RFC3542) */ 1666 void *optbuf; 1667 int optbuflen; 1668 struct ip6_pktopts **optp; 1669 1670 #ifdef RFC2292 1671 /* cannot mix with RFC2292 */ 1672 if (OPTBIT(IN6P_RFC2292)) { 1673 error = EINVAL; 1674 break; 1675 } 1676 #endif 1677 1678 optbuflen = sopt->sopt_size; 1679 optbuf = malloc(optbuflen, M_IP6OPT, M_NOWAIT); 1680 if (optbuf == NULL) { 1681 error = ENOBUFS; 1682 break; 1683 } 1684 1685 error = sockopt_get(sopt, optbuf, optbuflen); 1686 if (error) { 1687 free(optbuf, M_IP6OPT); 1688 break; 1689 } 1690 optp = &in6p->in6p_outputopts; 1691 error = ip6_pcbopt(optname, optbuf, optbuflen, 1692 optp, kauth_cred_get(), uproto); 1693 1694 free(optbuf, M_IP6OPT); 1695 break; 1696 } 1697 #undef OPTSET 1698 1699 case IPV6_MULTICAST_IF: 1700 case IPV6_MULTICAST_HOPS: 1701 case IPV6_MULTICAST_LOOP: 1702 case IPV6_JOIN_GROUP: 1703 case IPV6_LEAVE_GROUP: 1704 error = ip6_setmoptions(sopt, in6p); 1705 break; 1706 1707 case IPV6_PORTRANGE: 1708 error = sockopt_getint(sopt, &optval); 1709 if (error) 1710 break; 1711 1712 switch (optval) { 1713 case IPV6_PORTRANGE_DEFAULT: 1714 in6p->in6p_flags &= ~(IN6P_LOWPORT); 1715 in6p->in6p_flags &= ~(IN6P_HIGHPORT); 1716 break; 1717 1718 case IPV6_PORTRANGE_HIGH: 1719 in6p->in6p_flags &= ~(IN6P_LOWPORT); 1720 in6p->in6p_flags |= IN6P_HIGHPORT; 1721 break; 1722 1723 case IPV6_PORTRANGE_LOW: 1724 in6p->in6p_flags &= ~(IN6P_HIGHPORT); 1725 in6p->in6p_flags |= IN6P_LOWPORT; 1726 break; 1727 1728 default: 1729 error = EINVAL; 1730 break; 1731 } 1732 break; 1733 1734 case IPV6_PORTALGO: 1735 error = sockopt_getint(sopt, &optval); 1736 if (error) 1737 break; 1738 1739 error = portalgo_algo_index_select( 1740 (struct inpcb_hdr *)in6p, optval); 1741 break; 1742 1743 #if defined(IPSEC) 1744 case IPV6_IPSEC_POLICY: 1745 if (ipsec_enabled) { 1746 error = ipsec6_set_policy(in6p, optname, 1747 sopt->sopt_data, sopt->sopt_size, 1748 kauth_cred_get()); 1749 break; 1750 } 1751 /*FALLTHROUGH*/ 1752 #endif /* IPSEC */ 1753 1754 default: 1755 error = ENOPROTOOPT; 1756 break; 1757 } 1758 break; 1759 1760 case PRCO_GETOPT: 1761 switch (optname) { 1762 #ifdef RFC2292 1763 case IPV6_2292PKTOPTIONS: 1764 /* 1765 * RFC3542 (effectively) deprecated the 1766 * semantics of the 2292-style pktoptions. 1767 * Since it was not reliable in nature (i.e., 1768 * applications had to expect the lack of some 1769 * information after all), it would make sense 1770 * to simplify this part by always returning 1771 * empty data. 1772 */ 1773 break; 1774 #endif 1775 1776 case IPV6_RECVHOPOPTS: 1777 case IPV6_RECVDSTOPTS: 1778 case IPV6_RECVRTHDRDSTOPTS: 1779 case IPV6_UNICAST_HOPS: 1780 case IPV6_RECVPKTINFO: 1781 case IPV6_RECVHOPLIMIT: 1782 case IPV6_RECVRTHDR: 1783 case IPV6_RECVPATHMTU: 1784 1785 case IPV6_FAITH: 1786 case IPV6_V6ONLY: 1787 case IPV6_PORTRANGE: 1788 case IPV6_RECVTCLASS: 1789 switch (optname) { 1790 1791 case IPV6_RECVHOPOPTS: 1792 optval = OPTBIT(IN6P_HOPOPTS); 1793 break; 1794 1795 case IPV6_RECVDSTOPTS: 1796 optval = OPTBIT(IN6P_DSTOPTS); 1797 break; 1798 1799 case IPV6_RECVRTHDRDSTOPTS: 1800 optval = OPTBIT(IN6P_RTHDRDSTOPTS); 1801 break; 1802 1803 case IPV6_UNICAST_HOPS: 1804 optval = in6p->in6p_hops; 1805 break; 1806 1807 case IPV6_RECVPKTINFO: 1808 optval = OPTBIT(IN6P_PKTINFO); 1809 break; 1810 1811 case IPV6_RECVHOPLIMIT: 1812 optval = OPTBIT(IN6P_HOPLIMIT); 1813 break; 1814 1815 case IPV6_RECVRTHDR: 1816 optval = OPTBIT(IN6P_RTHDR); 1817 break; 1818 1819 case IPV6_RECVPATHMTU: 1820 optval = OPTBIT(IN6P_MTU); 1821 break; 1822 1823 case IPV6_FAITH: 1824 optval = OPTBIT(IN6P_FAITH); 1825 break; 1826 1827 case IPV6_V6ONLY: 1828 optval = OPTBIT(IN6P_IPV6_V6ONLY); 1829 break; 1830 1831 case IPV6_PORTRANGE: 1832 { 1833 int flags; 1834 flags = in6p->in6p_flags; 1835 if (flags & IN6P_HIGHPORT) 1836 optval = IPV6_PORTRANGE_HIGH; 1837 else if (flags & IN6P_LOWPORT) 1838 optval = IPV6_PORTRANGE_LOW; 1839 else 1840 optval = 0; 1841 break; 1842 } 1843 case IPV6_RECVTCLASS: 1844 optval = OPTBIT(IN6P_TCLASS); 1845 break; 1846 1847 } 1848 if (error) 1849 break; 1850 error = sockopt_setint(sopt, optval); 1851 break; 1852 1853 case IPV6_PATHMTU: 1854 { 1855 u_long pmtu = 0; 1856 struct ip6_mtuinfo mtuinfo; 1857 struct route *ro = &in6p->in6p_route; 1858 struct rtentry *rt; 1859 union { 1860 struct sockaddr dst; 1861 struct sockaddr_in6 dst6; 1862 } u; 1863 1864 if (!(so->so_state & SS_ISCONNECTED)) 1865 return (ENOTCONN); 1866 /* 1867 * XXX: we dot not consider the case of source 1868 * routing, or optional information to specify 1869 * the outgoing interface. 1870 */ 1871 sockaddr_in6_init(&u.dst6, &in6p->in6p_faddr, 0, 0, 0); 1872 rt = rtcache_lookup(ro, &u.dst); 1873 error = ip6_getpmtu(rt, NULL, &pmtu, NULL); 1874 rtcache_unref(rt, ro); 1875 if (error) 1876 break; 1877 if (pmtu > IPV6_MAXPACKET) 1878 pmtu = IPV6_MAXPACKET; 1879 1880 memset(&mtuinfo, 0, sizeof(mtuinfo)); 1881 mtuinfo.ip6m_mtu = (u_int32_t)pmtu; 1882 optdata = (void *)&mtuinfo; 1883 optdatalen = sizeof(mtuinfo); 1884 if (optdatalen > MCLBYTES) 1885 return (EMSGSIZE); /* XXX */ 1886 error = sockopt_set(sopt, optdata, optdatalen); 1887 break; 1888 } 1889 1890 #ifdef RFC2292 1891 case IPV6_2292PKTINFO: 1892 case IPV6_2292HOPLIMIT: 1893 case IPV6_2292HOPOPTS: 1894 case IPV6_2292RTHDR: 1895 case IPV6_2292DSTOPTS: 1896 switch (optname) { 1897 case IPV6_2292PKTINFO: 1898 optval = OPTBIT(IN6P_PKTINFO); 1899 break; 1900 case IPV6_2292HOPLIMIT: 1901 optval = OPTBIT(IN6P_HOPLIMIT); 1902 break; 1903 case IPV6_2292HOPOPTS: 1904 optval = OPTBIT(IN6P_HOPOPTS); 1905 break; 1906 case IPV6_2292RTHDR: 1907 optval = OPTBIT(IN6P_RTHDR); 1908 break; 1909 case IPV6_2292DSTOPTS: 1910 optval = OPTBIT(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); 1911 break; 1912 } 1913 error = sockopt_setint(sopt, optval); 1914 break; 1915 #endif 1916 case IPV6_PKTINFO: 1917 case IPV6_HOPOPTS: 1918 case IPV6_RTHDR: 1919 case IPV6_DSTOPTS: 1920 case IPV6_RTHDRDSTOPTS: 1921 case IPV6_NEXTHOP: 1922 case IPV6_OTCLASS: 1923 case IPV6_TCLASS: 1924 case IPV6_DONTFRAG: 1925 case IPV6_USE_MIN_MTU: 1926 case IPV6_PREFER_TEMPADDR: 1927 error = ip6_getpcbopt(in6p->in6p_outputopts, 1928 optname, sopt); 1929 break; 1930 1931 case IPV6_MULTICAST_IF: 1932 case IPV6_MULTICAST_HOPS: 1933 case IPV6_MULTICAST_LOOP: 1934 case IPV6_JOIN_GROUP: 1935 case IPV6_LEAVE_GROUP: 1936 error = ip6_getmoptions(sopt, in6p); 1937 break; 1938 1939 case IPV6_PORTALGO: 1940 optval = ((struct inpcb_hdr *)in6p)->inph_portalgo; 1941 error = sockopt_setint(sopt, optval); 1942 break; 1943 1944 #if defined(IPSEC) 1945 case IPV6_IPSEC_POLICY: 1946 if (ipsec_used) { 1947 struct mbuf *m = NULL; 1948 1949 /* 1950 * XXX: this will return EINVAL as sopt is 1951 * empty 1952 */ 1953 error = ipsec6_get_policy(in6p, sopt->sopt_data, 1954 sopt->sopt_size, &m); 1955 if (!error) 1956 error = sockopt_setmbuf(sopt, m); 1957 break; 1958 } 1959 /*FALLTHROUGH*/ 1960 #endif /* IPSEC */ 1961 1962 default: 1963 error = ENOPROTOOPT; 1964 break; 1965 } 1966 break; 1967 } 1968 return (error); 1969 } 1970 1971 int 1972 ip6_raw_ctloutput(int op, struct socket *so, struct sockopt *sopt) 1973 { 1974 int error = 0, optval; 1975 const int icmp6off = offsetof(struct icmp6_hdr, icmp6_cksum); 1976 struct in6pcb *in6p = sotoin6pcb(so); 1977 int level, optname; 1978 1979 KASSERT(sopt != NULL); 1980 1981 level = sopt->sopt_level; 1982 optname = sopt->sopt_name; 1983 1984 if (level != IPPROTO_IPV6) { 1985 return ENOPROTOOPT; 1986 } 1987 1988 switch (optname) { 1989 case IPV6_CHECKSUM: 1990 /* 1991 * For ICMPv6 sockets, no modification allowed for checksum 1992 * offset, permit "no change" values to help existing apps. 1993 * 1994 * XXX RFC3542 says: "An attempt to set IPV6_CHECKSUM 1995 * for an ICMPv6 socket will fail." The current 1996 * behavior does not meet RFC3542. 1997 */ 1998 switch (op) { 1999 case PRCO_SETOPT: 2000 error = sockopt_getint(sopt, &optval); 2001 if (error) 2002 break; 2003 if ((optval % 2) != 0) { 2004 /* the API assumes even offset values */ 2005 error = EINVAL; 2006 } else if (so->so_proto->pr_protocol == 2007 IPPROTO_ICMPV6) { 2008 if (optval != icmp6off) 2009 error = EINVAL; 2010 } else 2011 in6p->in6p_cksum = optval; 2012 break; 2013 2014 case PRCO_GETOPT: 2015 if (so->so_proto->pr_protocol == IPPROTO_ICMPV6) 2016 optval = icmp6off; 2017 else 2018 optval = in6p->in6p_cksum; 2019 2020 error = sockopt_setint(sopt, optval); 2021 break; 2022 2023 default: 2024 error = EINVAL; 2025 break; 2026 } 2027 break; 2028 2029 default: 2030 error = ENOPROTOOPT; 2031 break; 2032 } 2033 2034 return (error); 2035 } 2036 2037 #ifdef RFC2292 2038 /* 2039 * Set up IP6 options in pcb for insertion in output packets or 2040 * specifying behavior of outgoing packets. 2041 */ 2042 static int 2043 ip6_pcbopts(struct ip6_pktopts **pktopt, struct socket *so, 2044 struct sockopt *sopt) 2045 { 2046 struct ip6_pktopts *opt = *pktopt; 2047 struct mbuf *m; 2048 int error = 0; 2049 2050 /* turn off any old options. */ 2051 if (opt) { 2052 #ifdef DIAGNOSTIC 2053 if (opt->ip6po_pktinfo || opt->ip6po_nexthop || 2054 opt->ip6po_hbh || opt->ip6po_dest1 || opt->ip6po_dest2 || 2055 opt->ip6po_rhinfo.ip6po_rhi_rthdr) 2056 printf("ip6_pcbopts: all specified options are cleared.\n"); 2057 #endif 2058 ip6_clearpktopts(opt, -1); 2059 } else { 2060 opt = malloc(sizeof(*opt), M_IP6OPT, M_NOWAIT); 2061 if (opt == NULL) 2062 return (ENOBUFS); 2063 } 2064 *pktopt = NULL; 2065 2066 if (sopt == NULL || sopt->sopt_size == 0) { 2067 /* 2068 * Only turning off any previous options, regardless of 2069 * whether the opt is just created or given. 2070 */ 2071 free(opt, M_IP6OPT); 2072 return (0); 2073 } 2074 2075 /* set options specified by user. */ 2076 m = sockopt_getmbuf(sopt); 2077 if (m == NULL) { 2078 free(opt, M_IP6OPT); 2079 return (ENOBUFS); 2080 } 2081 2082 error = ip6_setpktopts(m, opt, NULL, kauth_cred_get(), 2083 so->so_proto->pr_protocol); 2084 m_freem(m); 2085 if (error != 0) { 2086 ip6_clearpktopts(opt, -1); /* XXX: discard all options */ 2087 free(opt, M_IP6OPT); 2088 return (error); 2089 } 2090 *pktopt = opt; 2091 return (0); 2092 } 2093 #endif 2094 2095 /* 2096 * initialize ip6_pktopts. beware that there are non-zero default values in 2097 * the struct. 2098 */ 2099 void 2100 ip6_initpktopts(struct ip6_pktopts *opt) 2101 { 2102 2103 memset(opt, 0, sizeof(*opt)); 2104 opt->ip6po_hlim = -1; /* -1 means default hop limit */ 2105 opt->ip6po_tclass = -1; /* -1 means default traffic class */ 2106 opt->ip6po_minmtu = IP6PO_MINMTU_MCASTONLY; 2107 opt->ip6po_prefer_tempaddr = IP6PO_TEMPADDR_SYSTEM; 2108 } 2109 2110 #define sin6tosa(sin6) ((struct sockaddr *)(sin6)) /* XXX */ 2111 static int 2112 ip6_pcbopt(int optname, u_char *buf, int len, struct ip6_pktopts **pktopt, 2113 kauth_cred_t cred, int uproto) 2114 { 2115 struct ip6_pktopts *opt; 2116 2117 if (*pktopt == NULL) { 2118 *pktopt = malloc(sizeof(struct ip6_pktopts), M_IP6OPT, 2119 M_NOWAIT); 2120 if (*pktopt == NULL) 2121 return (ENOBUFS); 2122 2123 ip6_initpktopts(*pktopt); 2124 } 2125 opt = *pktopt; 2126 2127 return (ip6_setpktopt(optname, buf, len, opt, cred, 1, 0, uproto)); 2128 } 2129 2130 static int 2131 ip6_getpcbopt(struct ip6_pktopts *pktopt, int optname, struct sockopt *sopt) 2132 { 2133 void *optdata = NULL; 2134 int optdatalen = 0; 2135 struct ip6_ext *ip6e; 2136 int error = 0; 2137 struct in6_pktinfo null_pktinfo; 2138 int deftclass = 0, on; 2139 int defminmtu = IP6PO_MINMTU_MCASTONLY; 2140 int defpreftemp = IP6PO_TEMPADDR_SYSTEM; 2141 2142 switch (optname) { 2143 case IPV6_PKTINFO: 2144 if (pktopt && pktopt->ip6po_pktinfo) 2145 optdata = (void *)pktopt->ip6po_pktinfo; 2146 else { 2147 /* XXX: we don't have to do this every time... */ 2148 memset(&null_pktinfo, 0, sizeof(null_pktinfo)); 2149 optdata = (void *)&null_pktinfo; 2150 } 2151 optdatalen = sizeof(struct in6_pktinfo); 2152 break; 2153 case IPV6_OTCLASS: 2154 /* XXX */ 2155 return (EINVAL); 2156 case IPV6_TCLASS: 2157 if (pktopt && pktopt->ip6po_tclass >= 0) 2158 optdata = (void *)&pktopt->ip6po_tclass; 2159 else 2160 optdata = (void *)&deftclass; 2161 optdatalen = sizeof(int); 2162 break; 2163 case IPV6_HOPOPTS: 2164 if (pktopt && pktopt->ip6po_hbh) { 2165 optdata = (void *)pktopt->ip6po_hbh; 2166 ip6e = (struct ip6_ext *)pktopt->ip6po_hbh; 2167 optdatalen = (ip6e->ip6e_len + 1) << 3; 2168 } 2169 break; 2170 case IPV6_RTHDR: 2171 if (pktopt && pktopt->ip6po_rthdr) { 2172 optdata = (void *)pktopt->ip6po_rthdr; 2173 ip6e = (struct ip6_ext *)pktopt->ip6po_rthdr; 2174 optdatalen = (ip6e->ip6e_len + 1) << 3; 2175 } 2176 break; 2177 case IPV6_RTHDRDSTOPTS: 2178 if (pktopt && pktopt->ip6po_dest1) { 2179 optdata = (void *)pktopt->ip6po_dest1; 2180 ip6e = (struct ip6_ext *)pktopt->ip6po_dest1; 2181 optdatalen = (ip6e->ip6e_len + 1) << 3; 2182 } 2183 break; 2184 case IPV6_DSTOPTS: 2185 if (pktopt && pktopt->ip6po_dest2) { 2186 optdata = (void *)pktopt->ip6po_dest2; 2187 ip6e = (struct ip6_ext *)pktopt->ip6po_dest2; 2188 optdatalen = (ip6e->ip6e_len + 1) << 3; 2189 } 2190 break; 2191 case IPV6_NEXTHOP: 2192 if (pktopt && pktopt->ip6po_nexthop) { 2193 optdata = (void *)pktopt->ip6po_nexthop; 2194 optdatalen = pktopt->ip6po_nexthop->sa_len; 2195 } 2196 break; 2197 case IPV6_USE_MIN_MTU: 2198 if (pktopt) 2199 optdata = (void *)&pktopt->ip6po_minmtu; 2200 else 2201 optdata = (void *)&defminmtu; 2202 optdatalen = sizeof(int); 2203 break; 2204 case IPV6_DONTFRAG: 2205 if (pktopt && ((pktopt->ip6po_flags) & IP6PO_DONTFRAG)) 2206 on = 1; 2207 else 2208 on = 0; 2209 optdata = (void *)&on; 2210 optdatalen = sizeof(on); 2211 break; 2212 case IPV6_PREFER_TEMPADDR: 2213 if (pktopt) 2214 optdata = (void *)&pktopt->ip6po_prefer_tempaddr; 2215 else 2216 optdata = (void *)&defpreftemp; 2217 optdatalen = sizeof(int); 2218 break; 2219 default: /* should not happen */ 2220 #ifdef DIAGNOSTIC 2221 panic("ip6_getpcbopt: unexpected option\n"); 2222 #endif 2223 return (ENOPROTOOPT); 2224 } 2225 2226 error = sockopt_set(sopt, optdata, optdatalen); 2227 2228 return (error); 2229 } 2230 2231 void 2232 ip6_clearpktopts(struct ip6_pktopts *pktopt, int optname) 2233 { 2234 if (optname == -1 || optname == IPV6_PKTINFO) { 2235 if (pktopt->ip6po_pktinfo) 2236 free(pktopt->ip6po_pktinfo, M_IP6OPT); 2237 pktopt->ip6po_pktinfo = NULL; 2238 } 2239 if (optname == -1 || optname == IPV6_HOPLIMIT) 2240 pktopt->ip6po_hlim = -1; 2241 if (optname == -1 || optname == IPV6_TCLASS) 2242 pktopt->ip6po_tclass = -1; 2243 if (optname == -1 || optname == IPV6_NEXTHOP) { 2244 rtcache_free(&pktopt->ip6po_nextroute); 2245 if (pktopt->ip6po_nexthop) 2246 free(pktopt->ip6po_nexthop, M_IP6OPT); 2247 pktopt->ip6po_nexthop = NULL; 2248 } 2249 if (optname == -1 || optname == IPV6_HOPOPTS) { 2250 if (pktopt->ip6po_hbh) 2251 free(pktopt->ip6po_hbh, M_IP6OPT); 2252 pktopt->ip6po_hbh = NULL; 2253 } 2254 if (optname == -1 || optname == IPV6_RTHDRDSTOPTS) { 2255 if (pktopt->ip6po_dest1) 2256 free(pktopt->ip6po_dest1, M_IP6OPT); 2257 pktopt->ip6po_dest1 = NULL; 2258 } 2259 if (optname == -1 || optname == IPV6_RTHDR) { 2260 if (pktopt->ip6po_rhinfo.ip6po_rhi_rthdr) 2261 free(pktopt->ip6po_rhinfo.ip6po_rhi_rthdr, M_IP6OPT); 2262 pktopt->ip6po_rhinfo.ip6po_rhi_rthdr = NULL; 2263 rtcache_free(&pktopt->ip6po_route); 2264 } 2265 if (optname == -1 || optname == IPV6_DSTOPTS) { 2266 if (pktopt->ip6po_dest2) 2267 free(pktopt->ip6po_dest2, M_IP6OPT); 2268 pktopt->ip6po_dest2 = NULL; 2269 } 2270 } 2271 2272 #define PKTOPT_EXTHDRCPY(type) \ 2273 do { \ 2274 if (src->type) { \ 2275 int hlen = (((struct ip6_ext *)src->type)->ip6e_len + 1) << 3;\ 2276 dst->type = malloc(hlen, M_IP6OPT, canwait); \ 2277 if (dst->type == NULL) \ 2278 goto bad; \ 2279 memcpy(dst->type, src->type, hlen); \ 2280 } \ 2281 } while (/*CONSTCOND*/ 0) 2282 2283 static int 2284 copypktopts(struct ip6_pktopts *dst, struct ip6_pktopts *src, int canwait) 2285 { 2286 dst->ip6po_hlim = src->ip6po_hlim; 2287 dst->ip6po_tclass = src->ip6po_tclass; 2288 dst->ip6po_flags = src->ip6po_flags; 2289 dst->ip6po_minmtu = src->ip6po_minmtu; 2290 dst->ip6po_prefer_tempaddr = src->ip6po_prefer_tempaddr; 2291 if (src->ip6po_pktinfo) { 2292 dst->ip6po_pktinfo = malloc(sizeof(*dst->ip6po_pktinfo), 2293 M_IP6OPT, canwait); 2294 if (dst->ip6po_pktinfo == NULL) 2295 goto bad; 2296 *dst->ip6po_pktinfo = *src->ip6po_pktinfo; 2297 } 2298 if (src->ip6po_nexthop) { 2299 dst->ip6po_nexthop = malloc(src->ip6po_nexthop->sa_len, 2300 M_IP6OPT, canwait); 2301 if (dst->ip6po_nexthop == NULL) 2302 goto bad; 2303 memcpy(dst->ip6po_nexthop, src->ip6po_nexthop, 2304 src->ip6po_nexthop->sa_len); 2305 } 2306 PKTOPT_EXTHDRCPY(ip6po_hbh); 2307 PKTOPT_EXTHDRCPY(ip6po_dest1); 2308 PKTOPT_EXTHDRCPY(ip6po_dest2); 2309 PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */ 2310 return (0); 2311 2312 bad: 2313 if (dst->ip6po_pktinfo) free(dst->ip6po_pktinfo, M_IP6OPT); 2314 if (dst->ip6po_nexthop) free(dst->ip6po_nexthop, M_IP6OPT); 2315 if (dst->ip6po_hbh) free(dst->ip6po_hbh, M_IP6OPT); 2316 if (dst->ip6po_dest1) free(dst->ip6po_dest1, M_IP6OPT); 2317 if (dst->ip6po_dest2) free(dst->ip6po_dest2, M_IP6OPT); 2318 if (dst->ip6po_rthdr) free(dst->ip6po_rthdr, M_IP6OPT); 2319 2320 return (ENOBUFS); 2321 } 2322 #undef PKTOPT_EXTHDRCPY 2323 2324 struct ip6_pktopts * 2325 ip6_copypktopts(struct ip6_pktopts *src, int canwait) 2326 { 2327 int error; 2328 struct ip6_pktopts *dst; 2329 2330 dst = malloc(sizeof(*dst), M_IP6OPT, canwait); 2331 if (dst == NULL) 2332 return (NULL); 2333 ip6_initpktopts(dst); 2334 2335 if ((error = copypktopts(dst, src, canwait)) != 0) { 2336 free(dst, M_IP6OPT); 2337 return (NULL); 2338 } 2339 2340 return (dst); 2341 } 2342 2343 void 2344 ip6_freepcbopts(struct ip6_pktopts *pktopt) 2345 { 2346 if (pktopt == NULL) 2347 return; 2348 2349 ip6_clearpktopts(pktopt, -1); 2350 2351 free(pktopt, M_IP6OPT); 2352 } 2353 2354 int 2355 ip6_get_membership(const struct sockopt *sopt, struct ifnet **ifp, void *v, 2356 size_t l) 2357 { 2358 struct ipv6_mreq mreq; 2359 int error; 2360 struct in6_addr *ia = &mreq.ipv6mr_multiaddr; 2361 struct in_addr *ia4 = (void *)&ia->s6_addr32[3]; 2362 error = sockopt_get(sopt, &mreq, sizeof(mreq)); 2363 if (error != 0) 2364 return error; 2365 2366 if (IN6_IS_ADDR_UNSPECIFIED(ia)) { 2367 /* 2368 * We use the unspecified address to specify to accept 2369 * all multicast addresses. Only super user is allowed 2370 * to do this. 2371 */ 2372 if (kauth_authorize_network(curlwp->l_cred, KAUTH_NETWORK_IPV6, 2373 KAUTH_REQ_NETWORK_IPV6_JOIN_MULTICAST, NULL, NULL, NULL)) 2374 return EACCES; 2375 } else if (IN6_IS_ADDR_V4MAPPED(ia)) { 2376 // Don't bother if we are not going to use ifp. 2377 if (l == sizeof(*ia)) { 2378 memcpy(v, ia, l); 2379 return 0; 2380 } 2381 } else if (!IN6_IS_ADDR_MULTICAST(ia)) { 2382 return EINVAL; 2383 } 2384 2385 /* 2386 * If no interface was explicitly specified, choose an 2387 * appropriate one according to the given multicast address. 2388 */ 2389 if (mreq.ipv6mr_interface == 0) { 2390 struct rtentry *rt; 2391 union { 2392 struct sockaddr dst; 2393 struct sockaddr_in dst4; 2394 struct sockaddr_in6 dst6; 2395 } u; 2396 struct route ro; 2397 2398 /* 2399 * Look up the routing table for the 2400 * address, and choose the outgoing interface. 2401 * XXX: is it a good approach? 2402 */ 2403 memset(&ro, 0, sizeof(ro)); 2404 if (IN6_IS_ADDR_V4MAPPED(ia)) 2405 sockaddr_in_init(&u.dst4, ia4, 0); 2406 else 2407 sockaddr_in6_init(&u.dst6, ia, 0, 0, 0); 2408 error = rtcache_setdst(&ro, &u.dst); 2409 if (error != 0) 2410 return error; 2411 rt = rtcache_init(&ro); 2412 *ifp = rt != NULL ? rt->rt_ifp : NULL; 2413 /* FIXME *ifp is NOMPSAFE */ 2414 rtcache_unref(rt, &ro); 2415 rtcache_free(&ro); 2416 } else { 2417 /* 2418 * If the interface is specified, validate it. 2419 */ 2420 if ((*ifp = if_byindex(mreq.ipv6mr_interface)) == NULL) 2421 return ENXIO; /* XXX EINVAL? */ 2422 } 2423 if (sizeof(*ia) == l) 2424 memcpy(v, ia, l); 2425 else 2426 memcpy(v, ia4, l); 2427 return 0; 2428 } 2429 2430 /* 2431 * Set the IP6 multicast options in response to user setsockopt(). 2432 */ 2433 static int 2434 ip6_setmoptions(const struct sockopt *sopt, struct in6pcb *in6p) 2435 { 2436 int error = 0; 2437 u_int loop, ifindex; 2438 struct ipv6_mreq mreq; 2439 struct in6_addr ia; 2440 struct ifnet *ifp; 2441 struct ip6_moptions *im6o = in6p->in6p_moptions; 2442 struct in6_multi_mship *imm; 2443 2444 if (im6o == NULL) { 2445 /* 2446 * No multicast option buffer attached to the pcb; 2447 * allocate one and initialize to default values. 2448 */ 2449 im6o = malloc(sizeof(*im6o), M_IPMOPTS, M_NOWAIT); 2450 if (im6o == NULL) 2451 return (ENOBUFS); 2452 in6p->in6p_moptions = im6o; 2453 im6o->im6o_multicast_if_index = 0; 2454 im6o->im6o_multicast_hlim = ip6_defmcasthlim; 2455 im6o->im6o_multicast_loop = IPV6_DEFAULT_MULTICAST_LOOP; 2456 LIST_INIT(&im6o->im6o_memberships); 2457 } 2458 2459 switch (sopt->sopt_name) { 2460 2461 case IPV6_MULTICAST_IF: 2462 /* 2463 * Select the interface for outgoing multicast packets. 2464 */ 2465 error = sockopt_get(sopt, &ifindex, sizeof(ifindex)); 2466 if (error != 0) 2467 break; 2468 2469 if (ifindex != 0) { 2470 if ((ifp = if_byindex(ifindex)) == NULL) { 2471 error = ENXIO; /* XXX EINVAL? */ 2472 break; 2473 } 2474 if ((ifp->if_flags & IFF_MULTICAST) == 0) { 2475 error = EADDRNOTAVAIL; 2476 break; 2477 } 2478 } else 2479 ifp = NULL; 2480 im6o->im6o_multicast_if_index = if_get_index(ifp); 2481 break; 2482 2483 case IPV6_MULTICAST_HOPS: 2484 { 2485 /* 2486 * Set the IP6 hoplimit for outgoing multicast packets. 2487 */ 2488 int optval; 2489 2490 error = sockopt_getint(sopt, &optval); 2491 if (error != 0) 2492 break; 2493 2494 if (optval < -1 || optval >= 256) 2495 error = EINVAL; 2496 else if (optval == -1) 2497 im6o->im6o_multicast_hlim = ip6_defmcasthlim; 2498 else 2499 im6o->im6o_multicast_hlim = optval; 2500 break; 2501 } 2502 2503 case IPV6_MULTICAST_LOOP: 2504 /* 2505 * Set the loopback flag for outgoing multicast packets. 2506 * Must be zero or one. 2507 */ 2508 error = sockopt_get(sopt, &loop, sizeof(loop)); 2509 if (error != 0) 2510 break; 2511 if (loop > 1) { 2512 error = EINVAL; 2513 break; 2514 } 2515 im6o->im6o_multicast_loop = loop; 2516 break; 2517 2518 case IPV6_JOIN_GROUP: 2519 /* 2520 * Add a multicast group membership. 2521 * Group must be a valid IP6 multicast address. 2522 */ 2523 if ((error = ip6_get_membership(sopt, &ifp, &ia, sizeof(ia)))) 2524 return error; 2525 2526 if (IN6_IS_ADDR_V4MAPPED(&ia)) { 2527 error = ip_setmoptions(&in6p->in6p_v4moptions, sopt); 2528 break; 2529 } 2530 /* 2531 * See if we found an interface, and confirm that it 2532 * supports multicast 2533 */ 2534 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) { 2535 error = EADDRNOTAVAIL; 2536 break; 2537 } 2538 2539 if (in6_setscope(&ia, ifp, NULL)) { 2540 error = EADDRNOTAVAIL; /* XXX: should not happen */ 2541 break; 2542 } 2543 2544 /* 2545 * See if the membership already exists. 2546 */ 2547 for (imm = im6o->im6o_memberships.lh_first; 2548 imm != NULL; imm = imm->i6mm_chain.le_next) 2549 if (imm->i6mm_maddr->in6m_ifp == ifp && 2550 IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr, 2551 &ia)) 2552 break; 2553 if (imm != NULL) { 2554 error = EADDRINUSE; 2555 break; 2556 } 2557 /* 2558 * Everything looks good; add a new record to the multicast 2559 * address list for the given interface. 2560 */ 2561 imm = in6_joingroup(ifp, &ia, &error, 0); 2562 if (imm == NULL) 2563 break; 2564 LIST_INSERT_HEAD(&im6o->im6o_memberships, imm, i6mm_chain); 2565 break; 2566 2567 case IPV6_LEAVE_GROUP: 2568 /* 2569 * Drop a multicast group membership. 2570 * Group must be a valid IP6 multicast address. 2571 */ 2572 error = sockopt_get(sopt, &mreq, sizeof(mreq)); 2573 if (error != 0) 2574 break; 2575 2576 if (IN6_IS_ADDR_V4MAPPED(&mreq.ipv6mr_multiaddr)) { 2577 error = ip_setmoptions(&in6p->in6p_v4moptions, sopt); 2578 break; 2579 } 2580 /* 2581 * If an interface address was specified, get a pointer 2582 * to its ifnet structure. 2583 */ 2584 if (mreq.ipv6mr_interface != 0) { 2585 if ((ifp = if_byindex(mreq.ipv6mr_interface)) == NULL) { 2586 error = ENXIO; /* XXX EINVAL? */ 2587 break; 2588 } 2589 } else 2590 ifp = NULL; 2591 2592 /* Fill in the scope zone ID */ 2593 if (ifp) { 2594 if (in6_setscope(&mreq.ipv6mr_multiaddr, ifp, NULL)) { 2595 /* XXX: should not happen */ 2596 error = EADDRNOTAVAIL; 2597 break; 2598 } 2599 } else if (mreq.ipv6mr_interface != 0) { 2600 /* 2601 * XXX: This case would happens when the (positive) 2602 * index is in the valid range, but the corresponding 2603 * interface has been detached dynamically. The above 2604 * check probably avoids such case to happen here, but 2605 * we check it explicitly for safety. 2606 */ 2607 error = EADDRNOTAVAIL; 2608 break; 2609 } else { /* ipv6mr_interface == 0 */ 2610 struct sockaddr_in6 sa6_mc; 2611 2612 /* 2613 * The API spec says as follows: 2614 * If the interface index is specified as 0, the 2615 * system may choose a multicast group membership to 2616 * drop by matching the multicast address only. 2617 * On the other hand, we cannot disambiguate the scope 2618 * zone unless an interface is provided. Thus, we 2619 * check if there's ambiguity with the default scope 2620 * zone as the last resort. 2621 */ 2622 sockaddr_in6_init(&sa6_mc, &mreq.ipv6mr_multiaddr, 2623 0, 0, 0); 2624 error = sa6_embedscope(&sa6_mc, ip6_use_defzone); 2625 if (error != 0) 2626 break; 2627 mreq.ipv6mr_multiaddr = sa6_mc.sin6_addr; 2628 } 2629 2630 /* 2631 * Find the membership in the membership list. 2632 */ 2633 for (imm = im6o->im6o_memberships.lh_first; 2634 imm != NULL; imm = imm->i6mm_chain.le_next) { 2635 if ((ifp == NULL || imm->i6mm_maddr->in6m_ifp == ifp) && 2636 IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr, 2637 &mreq.ipv6mr_multiaddr)) 2638 break; 2639 } 2640 if (imm == NULL) { 2641 /* Unable to resolve interface */ 2642 error = EADDRNOTAVAIL; 2643 break; 2644 } 2645 /* 2646 * Give up the multicast address record to which the 2647 * membership points. 2648 */ 2649 LIST_REMOVE(imm, i6mm_chain); 2650 in6_leavegroup(imm); 2651 break; 2652 2653 default: 2654 error = EOPNOTSUPP; 2655 break; 2656 } 2657 2658 /* 2659 * If all options have default values, no need to keep the mbuf. 2660 */ 2661 if (im6o->im6o_multicast_if_index == 0 && 2662 im6o->im6o_multicast_hlim == ip6_defmcasthlim && 2663 im6o->im6o_multicast_loop == IPV6_DEFAULT_MULTICAST_LOOP && 2664 im6o->im6o_memberships.lh_first == NULL) { 2665 free(in6p->in6p_moptions, M_IPMOPTS); 2666 in6p->in6p_moptions = NULL; 2667 } 2668 2669 return (error); 2670 } 2671 2672 /* 2673 * Return the IP6 multicast options in response to user getsockopt(). 2674 */ 2675 static int 2676 ip6_getmoptions(struct sockopt *sopt, struct in6pcb *in6p) 2677 { 2678 u_int optval; 2679 int error; 2680 struct ip6_moptions *im6o = in6p->in6p_moptions; 2681 2682 switch (sopt->sopt_name) { 2683 case IPV6_MULTICAST_IF: 2684 if (im6o == NULL || im6o->im6o_multicast_if_index == 0) 2685 optval = 0; 2686 else 2687 optval = im6o->im6o_multicast_if_index; 2688 2689 error = sockopt_set(sopt, &optval, sizeof(optval)); 2690 break; 2691 2692 case IPV6_MULTICAST_HOPS: 2693 if (im6o == NULL) 2694 optval = ip6_defmcasthlim; 2695 else 2696 optval = im6o->im6o_multicast_hlim; 2697 2698 error = sockopt_set(sopt, &optval, sizeof(optval)); 2699 break; 2700 2701 case IPV6_MULTICAST_LOOP: 2702 if (im6o == NULL) 2703 optval = IPV6_DEFAULT_MULTICAST_LOOP; 2704 else 2705 optval = im6o->im6o_multicast_loop; 2706 2707 error = sockopt_set(sopt, &optval, sizeof(optval)); 2708 break; 2709 2710 default: 2711 error = EOPNOTSUPP; 2712 } 2713 2714 return (error); 2715 } 2716 2717 /* 2718 * Discard the IP6 multicast options. 2719 */ 2720 void 2721 ip6_freemoptions(struct ip6_moptions *im6o) 2722 { 2723 struct in6_multi_mship *imm; 2724 2725 if (im6o == NULL) 2726 return; 2727 2728 while ((imm = im6o->im6o_memberships.lh_first) != NULL) { 2729 LIST_REMOVE(imm, i6mm_chain); 2730 in6_leavegroup(imm); 2731 } 2732 free(im6o, M_IPMOPTS); 2733 } 2734 2735 /* 2736 * Set IPv6 outgoing packet options based on advanced API. 2737 */ 2738 int 2739 ip6_setpktopts(struct mbuf *control, struct ip6_pktopts *opt, 2740 struct ip6_pktopts *stickyopt, kauth_cred_t cred, int uproto) 2741 { 2742 struct cmsghdr *cm = 0; 2743 2744 if (control == NULL || opt == NULL) 2745 return (EINVAL); 2746 2747 ip6_initpktopts(opt); 2748 if (stickyopt) { 2749 int error; 2750 2751 /* 2752 * If stickyopt is provided, make a local copy of the options 2753 * for this particular packet, then override them by ancillary 2754 * objects. 2755 * XXX: copypktopts() does not copy the cached route to a next 2756 * hop (if any). This is not very good in terms of efficiency, 2757 * but we can allow this since this option should be rarely 2758 * used. 2759 */ 2760 if ((error = copypktopts(opt, stickyopt, M_NOWAIT)) != 0) 2761 return (error); 2762 } 2763 2764 /* 2765 * XXX: Currently, we assume all the optional information is stored 2766 * in a single mbuf. 2767 */ 2768 if (control->m_next) 2769 return (EINVAL); 2770 2771 /* XXX if cm->cmsg_len is not aligned, control->m_len can become <0 */ 2772 for (; control->m_len > 0; control->m_data += CMSG_ALIGN(cm->cmsg_len), 2773 control->m_len -= CMSG_ALIGN(cm->cmsg_len)) { 2774 int error; 2775 2776 if (control->m_len < CMSG_LEN(0)) 2777 return (EINVAL); 2778 2779 cm = mtod(control, struct cmsghdr *); 2780 if (cm->cmsg_len == 0 || cm->cmsg_len > control->m_len) 2781 return (EINVAL); 2782 if (cm->cmsg_level != IPPROTO_IPV6) 2783 continue; 2784 2785 error = ip6_setpktopt(cm->cmsg_type, CMSG_DATA(cm), 2786 cm->cmsg_len - CMSG_LEN(0), opt, cred, 0, 1, uproto); 2787 if (error) 2788 return (error); 2789 } 2790 2791 return (0); 2792 } 2793 2794 /* 2795 * Set a particular packet option, as a sticky option or an ancillary data 2796 * item. "len" can be 0 only when it's a sticky option. 2797 * We have 4 cases of combination of "sticky" and "cmsg": 2798 * "sticky=0, cmsg=0": impossible 2799 * "sticky=0, cmsg=1": RFC2292 or RFC3542 ancillary data 2800 * "sticky=1, cmsg=0": RFC3542 socket option 2801 * "sticky=1, cmsg=1": RFC2292 socket option 2802 */ 2803 static int 2804 ip6_setpktopt(int optname, u_char *buf, int len, struct ip6_pktopts *opt, 2805 kauth_cred_t cred, int sticky, int cmsg, int uproto) 2806 { 2807 int minmtupolicy; 2808 int error; 2809 2810 if (!sticky && !cmsg) { 2811 #ifdef DIAGNOSTIC 2812 printf("ip6_setpktopt: impossible case\n"); 2813 #endif 2814 return (EINVAL); 2815 } 2816 2817 /* 2818 * IPV6_2292xxx is for backward compatibility to RFC2292, and should 2819 * not be specified in the context of RFC3542. Conversely, 2820 * RFC3542 types should not be specified in the context of RFC2292. 2821 */ 2822 if (!cmsg) { 2823 switch (optname) { 2824 case IPV6_2292PKTINFO: 2825 case IPV6_2292HOPLIMIT: 2826 case IPV6_2292NEXTHOP: 2827 case IPV6_2292HOPOPTS: 2828 case IPV6_2292DSTOPTS: 2829 case IPV6_2292RTHDR: 2830 case IPV6_2292PKTOPTIONS: 2831 return (ENOPROTOOPT); 2832 } 2833 } 2834 if (sticky && cmsg) { 2835 switch (optname) { 2836 case IPV6_PKTINFO: 2837 case IPV6_HOPLIMIT: 2838 case IPV6_NEXTHOP: 2839 case IPV6_HOPOPTS: 2840 case IPV6_DSTOPTS: 2841 case IPV6_RTHDRDSTOPTS: 2842 case IPV6_RTHDR: 2843 case IPV6_USE_MIN_MTU: 2844 case IPV6_DONTFRAG: 2845 case IPV6_OTCLASS: 2846 case IPV6_TCLASS: 2847 case IPV6_PREFER_TEMPADDR: /* XXX not an RFC3542 option */ 2848 return (ENOPROTOOPT); 2849 } 2850 } 2851 2852 switch (optname) { 2853 #ifdef RFC2292 2854 case IPV6_2292PKTINFO: 2855 #endif 2856 case IPV6_PKTINFO: 2857 { 2858 struct in6_pktinfo *pktinfo; 2859 2860 if (len != sizeof(struct in6_pktinfo)) 2861 return (EINVAL); 2862 2863 pktinfo = (struct in6_pktinfo *)buf; 2864 2865 /* 2866 * An application can clear any sticky IPV6_PKTINFO option by 2867 * doing a "regular" setsockopt with ipi6_addr being 2868 * in6addr_any and ipi6_ifindex being zero. 2869 * [RFC 3542, Section 6] 2870 */ 2871 if (optname == IPV6_PKTINFO && opt->ip6po_pktinfo && 2872 pktinfo->ipi6_ifindex == 0 && 2873 IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) { 2874 ip6_clearpktopts(opt, optname); 2875 break; 2876 } 2877 2878 if (uproto == IPPROTO_TCP && optname == IPV6_PKTINFO && 2879 sticky && !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) { 2880 return (EINVAL); 2881 } 2882 2883 /* Validate the interface index if specified. */ 2884 if (pktinfo->ipi6_ifindex) { 2885 struct ifnet *ifp; 2886 int s = pserialize_read_enter(); 2887 ifp = if_byindex(pktinfo->ipi6_ifindex); 2888 if (ifp == NULL) { 2889 pserialize_read_exit(s); 2890 return ENXIO; 2891 } 2892 pserialize_read_exit(s); 2893 } 2894 2895 /* 2896 * We store the address anyway, and let in6_selectsrc() 2897 * validate the specified address. This is because ipi6_addr 2898 * may not have enough information about its scope zone, and 2899 * we may need additional information (such as outgoing 2900 * interface or the scope zone of a destination address) to 2901 * disambiguate the scope. 2902 * XXX: the delay of the validation may confuse the 2903 * application when it is used as a sticky option. 2904 */ 2905 if (opt->ip6po_pktinfo == NULL) { 2906 opt->ip6po_pktinfo = malloc(sizeof(*pktinfo), 2907 M_IP6OPT, M_NOWAIT); 2908 if (opt->ip6po_pktinfo == NULL) 2909 return (ENOBUFS); 2910 } 2911 memcpy(opt->ip6po_pktinfo, pktinfo, sizeof(*pktinfo)); 2912 break; 2913 } 2914 2915 #ifdef RFC2292 2916 case IPV6_2292HOPLIMIT: 2917 #endif 2918 case IPV6_HOPLIMIT: 2919 { 2920 int *hlimp; 2921 2922 /* 2923 * RFC 3542 deprecated the usage of sticky IPV6_HOPLIMIT 2924 * to simplify the ordering among hoplimit options. 2925 */ 2926 if (optname == IPV6_HOPLIMIT && sticky) 2927 return (ENOPROTOOPT); 2928 2929 if (len != sizeof(int)) 2930 return (EINVAL); 2931 hlimp = (int *)buf; 2932 if (*hlimp < -1 || *hlimp > 255) 2933 return (EINVAL); 2934 2935 opt->ip6po_hlim = *hlimp; 2936 break; 2937 } 2938 2939 case IPV6_OTCLASS: 2940 if (len != sizeof(u_int8_t)) 2941 return (EINVAL); 2942 2943 opt->ip6po_tclass = *(u_int8_t *)buf; 2944 break; 2945 2946 case IPV6_TCLASS: 2947 { 2948 int tclass; 2949 2950 if (len != sizeof(int)) 2951 return (EINVAL); 2952 tclass = *(int *)buf; 2953 if (tclass < -1 || tclass > 255) 2954 return (EINVAL); 2955 2956 opt->ip6po_tclass = tclass; 2957 break; 2958 } 2959 2960 #ifdef RFC2292 2961 case IPV6_2292NEXTHOP: 2962 #endif 2963 case IPV6_NEXTHOP: 2964 error = kauth_authorize_network(cred, KAUTH_NETWORK_IPV6, 2965 KAUTH_REQ_NETWORK_IPV6_HOPBYHOP, NULL, NULL, NULL); 2966 if (error) 2967 return (error); 2968 2969 if (len == 0) { /* just remove the option */ 2970 ip6_clearpktopts(opt, IPV6_NEXTHOP); 2971 break; 2972 } 2973 2974 /* check if cmsg_len is large enough for sa_len */ 2975 if (len < sizeof(struct sockaddr) || len < *buf) 2976 return (EINVAL); 2977 2978 switch (((struct sockaddr *)buf)->sa_family) { 2979 case AF_INET6: 2980 { 2981 struct sockaddr_in6 *sa6 = (struct sockaddr_in6 *)buf; 2982 2983 if (sa6->sin6_len != sizeof(struct sockaddr_in6)) 2984 return (EINVAL); 2985 2986 if (IN6_IS_ADDR_UNSPECIFIED(&sa6->sin6_addr) || 2987 IN6_IS_ADDR_MULTICAST(&sa6->sin6_addr)) { 2988 return (EINVAL); 2989 } 2990 if ((error = sa6_embedscope(sa6, ip6_use_defzone)) 2991 != 0) { 2992 return (error); 2993 } 2994 break; 2995 } 2996 case AF_LINK: /* eventually be supported? */ 2997 default: 2998 return (EAFNOSUPPORT); 2999 } 3000 3001 /* turn off the previous option, then set the new option. */ 3002 ip6_clearpktopts(opt, IPV6_NEXTHOP); 3003 opt->ip6po_nexthop = malloc(*buf, M_IP6OPT, M_NOWAIT); 3004 if (opt->ip6po_nexthop == NULL) 3005 return (ENOBUFS); 3006 memcpy(opt->ip6po_nexthop, buf, *buf); 3007 break; 3008 3009 #ifdef RFC2292 3010 case IPV6_2292HOPOPTS: 3011 #endif 3012 case IPV6_HOPOPTS: 3013 { 3014 struct ip6_hbh *hbh; 3015 int hbhlen; 3016 3017 /* 3018 * XXX: We don't allow a non-privileged user to set ANY HbH 3019 * options, since per-option restriction has too much 3020 * overhead. 3021 */ 3022 error = kauth_authorize_network(cred, KAUTH_NETWORK_IPV6, 3023 KAUTH_REQ_NETWORK_IPV6_HOPBYHOP, NULL, NULL, NULL); 3024 if (error) 3025 return (error); 3026 3027 if (len == 0) { 3028 ip6_clearpktopts(opt, IPV6_HOPOPTS); 3029 break; /* just remove the option */ 3030 } 3031 3032 /* message length validation */ 3033 if (len < sizeof(struct ip6_hbh)) 3034 return (EINVAL); 3035 hbh = (struct ip6_hbh *)buf; 3036 hbhlen = (hbh->ip6h_len + 1) << 3; 3037 if (len != hbhlen) 3038 return (EINVAL); 3039 3040 /* turn off the previous option, then set the new option. */ 3041 ip6_clearpktopts(opt, IPV6_HOPOPTS); 3042 opt->ip6po_hbh = malloc(hbhlen, M_IP6OPT, M_NOWAIT); 3043 if (opt->ip6po_hbh == NULL) 3044 return (ENOBUFS); 3045 memcpy(opt->ip6po_hbh, hbh, hbhlen); 3046 3047 break; 3048 } 3049 3050 #ifdef RFC2292 3051 case IPV6_2292DSTOPTS: 3052 #endif 3053 case IPV6_DSTOPTS: 3054 case IPV6_RTHDRDSTOPTS: 3055 { 3056 struct ip6_dest *dest, **newdest = NULL; 3057 int destlen; 3058 3059 /* XXX: see the comment for IPV6_HOPOPTS */ 3060 error = kauth_authorize_network(cred, KAUTH_NETWORK_IPV6, 3061 KAUTH_REQ_NETWORK_IPV6_HOPBYHOP, NULL, NULL, NULL); 3062 if (error) 3063 return (error); 3064 3065 if (len == 0) { 3066 ip6_clearpktopts(opt, optname); 3067 break; /* just remove the option */ 3068 } 3069 3070 /* message length validation */ 3071 if (len < sizeof(struct ip6_dest)) 3072 return (EINVAL); 3073 dest = (struct ip6_dest *)buf; 3074 destlen = (dest->ip6d_len + 1) << 3; 3075 if (len != destlen) 3076 return (EINVAL); 3077 /* 3078 * Determine the position that the destination options header 3079 * should be inserted; before or after the routing header. 3080 */ 3081 switch (optname) { 3082 case IPV6_2292DSTOPTS: 3083 /* 3084 * The old advanced API is ambiguous on this point. 3085 * Our approach is to determine the position based 3086 * according to the existence of a routing header. 3087 * Note, however, that this depends on the order of the 3088 * extension headers in the ancillary data; the 1st 3089 * part of the destination options header must appear 3090 * before the routing header in the ancillary data, 3091 * too. 3092 * RFC3542 solved the ambiguity by introducing 3093 * separate ancillary data or option types. 3094 */ 3095 if (opt->ip6po_rthdr == NULL) 3096 newdest = &opt->ip6po_dest1; 3097 else 3098 newdest = &opt->ip6po_dest2; 3099 break; 3100 case IPV6_RTHDRDSTOPTS: 3101 newdest = &opt->ip6po_dest1; 3102 break; 3103 case IPV6_DSTOPTS: 3104 newdest = &opt->ip6po_dest2; 3105 break; 3106 } 3107 3108 /* turn off the previous option, then set the new option. */ 3109 ip6_clearpktopts(opt, optname); 3110 *newdest = malloc(destlen, M_IP6OPT, M_NOWAIT); 3111 if (*newdest == NULL) 3112 return (ENOBUFS); 3113 memcpy(*newdest, dest, destlen); 3114 3115 break; 3116 } 3117 3118 #ifdef RFC2292 3119 case IPV6_2292RTHDR: 3120 #endif 3121 case IPV6_RTHDR: 3122 { 3123 struct ip6_rthdr *rth; 3124 int rthlen; 3125 3126 if (len == 0) { 3127 ip6_clearpktopts(opt, IPV6_RTHDR); 3128 break; /* just remove the option */ 3129 } 3130 3131 /* message length validation */ 3132 if (len < sizeof(struct ip6_rthdr)) 3133 return (EINVAL); 3134 rth = (struct ip6_rthdr *)buf; 3135 rthlen = (rth->ip6r_len + 1) << 3; 3136 if (len != rthlen) 3137 return (EINVAL); 3138 switch (rth->ip6r_type) { 3139 case IPV6_RTHDR_TYPE_0: 3140 if (rth->ip6r_len == 0) /* must contain one addr */ 3141 return (EINVAL); 3142 if (rth->ip6r_len % 2) /* length must be even */ 3143 return (EINVAL); 3144 if (rth->ip6r_len / 2 != rth->ip6r_segleft) 3145 return (EINVAL); 3146 break; 3147 default: 3148 return (EINVAL); /* not supported */ 3149 } 3150 /* turn off the previous option */ 3151 ip6_clearpktopts(opt, IPV6_RTHDR); 3152 opt->ip6po_rthdr = malloc(rthlen, M_IP6OPT, M_NOWAIT); 3153 if (opt->ip6po_rthdr == NULL) 3154 return (ENOBUFS); 3155 memcpy(opt->ip6po_rthdr, rth, rthlen); 3156 break; 3157 } 3158 3159 case IPV6_USE_MIN_MTU: 3160 if (len != sizeof(int)) 3161 return (EINVAL); 3162 minmtupolicy = *(int *)buf; 3163 if (minmtupolicy != IP6PO_MINMTU_MCASTONLY && 3164 minmtupolicy != IP6PO_MINMTU_DISABLE && 3165 minmtupolicy != IP6PO_MINMTU_ALL) { 3166 return (EINVAL); 3167 } 3168 opt->ip6po_minmtu = minmtupolicy; 3169 break; 3170 3171 case IPV6_DONTFRAG: 3172 if (len != sizeof(int)) 3173 return (EINVAL); 3174 3175 if (uproto == IPPROTO_TCP || *(int *)buf == 0) { 3176 /* 3177 * we ignore this option for TCP sockets. 3178 * (RFC3542 leaves this case unspecified.) 3179 */ 3180 opt->ip6po_flags &= ~IP6PO_DONTFRAG; 3181 } else 3182 opt->ip6po_flags |= IP6PO_DONTFRAG; 3183 break; 3184 3185 case IPV6_PREFER_TEMPADDR: 3186 { 3187 int preftemp; 3188 3189 if (len != sizeof(int)) 3190 return (EINVAL); 3191 preftemp = *(int *)buf; 3192 switch (preftemp) { 3193 case IP6PO_TEMPADDR_SYSTEM: 3194 case IP6PO_TEMPADDR_NOTPREFER: 3195 case IP6PO_TEMPADDR_PREFER: 3196 break; 3197 default: 3198 return (EINVAL); 3199 } 3200 opt->ip6po_prefer_tempaddr = preftemp; 3201 break; 3202 } 3203 3204 default: 3205 return (ENOPROTOOPT); 3206 } /* end of switch */ 3207 3208 return (0); 3209 } 3210 3211 /* 3212 * Routine called from ip6_output() to loop back a copy of an IP6 multicast 3213 * packet to the input queue of a specified interface. Note that this 3214 * calls the output routine of the loopback "driver", but with an interface 3215 * pointer that might NOT be lo0ifp -- easier than replicating that code here. 3216 */ 3217 void 3218 ip6_mloopback(struct ifnet *ifp, struct mbuf *m, 3219 const struct sockaddr_in6 *dst) 3220 { 3221 struct mbuf *copym; 3222 struct ip6_hdr *ip6; 3223 3224 copym = m_copy(m, 0, M_COPYALL); 3225 if (copym == NULL) 3226 return; 3227 3228 /* 3229 * Make sure to deep-copy IPv6 header portion in case the data 3230 * is in an mbuf cluster, so that we can safely override the IPv6 3231 * header portion later. 3232 */ 3233 if ((copym->m_flags & M_EXT) != 0 || 3234 copym->m_len < sizeof(struct ip6_hdr)) { 3235 copym = m_pullup(copym, sizeof(struct ip6_hdr)); 3236 if (copym == NULL) 3237 return; 3238 } 3239 3240 #ifdef DIAGNOSTIC 3241 if (copym->m_len < sizeof(*ip6)) { 3242 m_freem(copym); 3243 return; 3244 } 3245 #endif 3246 3247 ip6 = mtod(copym, struct ip6_hdr *); 3248 /* 3249 * clear embedded scope identifiers if necessary. 3250 * in6_clearscope will touch the addresses only when necessary. 3251 */ 3252 in6_clearscope(&ip6->ip6_src); 3253 in6_clearscope(&ip6->ip6_dst); 3254 3255 (void)looutput(ifp, copym, (const struct sockaddr *)dst, NULL); 3256 } 3257 3258 /* 3259 * Chop IPv6 header off from the payload. 3260 */ 3261 static int 3262 ip6_splithdr(struct mbuf *m, struct ip6_exthdrs *exthdrs) 3263 { 3264 struct mbuf *mh; 3265 struct ip6_hdr *ip6; 3266 3267 ip6 = mtod(m, struct ip6_hdr *); 3268 if (m->m_len > sizeof(*ip6)) { 3269 MGETHDR(mh, M_DONTWAIT, MT_HEADER); 3270 if (mh == 0) { 3271 m_freem(m); 3272 return ENOBUFS; 3273 } 3274 M_MOVE_PKTHDR(mh, m); 3275 MH_ALIGN(mh, sizeof(*ip6)); 3276 m->m_len -= sizeof(*ip6); 3277 m->m_data += sizeof(*ip6); 3278 mh->m_next = m; 3279 m = mh; 3280 m->m_len = sizeof(*ip6); 3281 bcopy((void *)ip6, mtod(m, void *), sizeof(*ip6)); 3282 } 3283 exthdrs->ip6e_ip6 = m; 3284 return 0; 3285 } 3286 3287 /* 3288 * Compute IPv6 extension header length. 3289 */ 3290 int 3291 ip6_optlen(struct in6pcb *in6p) 3292 { 3293 int len; 3294 3295 if (!in6p->in6p_outputopts) 3296 return 0; 3297 3298 len = 0; 3299 #define elen(x) \ 3300 (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0) 3301 3302 len += elen(in6p->in6p_outputopts->ip6po_hbh); 3303 len += elen(in6p->in6p_outputopts->ip6po_dest1); 3304 len += elen(in6p->in6p_outputopts->ip6po_rthdr); 3305 len += elen(in6p->in6p_outputopts->ip6po_dest2); 3306 return len; 3307 #undef elen 3308 } 3309 3310 /* 3311 * Ensure sending address is valid. 3312 * Returns 0 on success, -1 if an error should be sent back or 1 3313 * if the packet could be dropped without error (protocol dependent). 3314 */ 3315 static int 3316 ip6_ifaddrvalid(const struct in6_addr *addr) 3317 { 3318 struct sockaddr_in6 sin6; 3319 int s, error; 3320 struct ifaddr *ifa; 3321 struct in6_ifaddr *ia6; 3322 3323 if (IN6_IS_ADDR_UNSPECIFIED(addr)) 3324 return 0; 3325 3326 memset(&sin6, 0, sizeof(sin6)); 3327 sin6.sin6_family = AF_INET6; 3328 sin6.sin6_len = sizeof(sin6); 3329 sin6.sin6_addr = *addr; 3330 3331 s = pserialize_read_enter(); 3332 ifa = ifa_ifwithaddr(sin6tosa(&sin6)); 3333 if ((ia6 = ifatoia6(ifa)) == NULL || 3334 ia6->ia6_flags & (IN6_IFF_ANYCAST | IN6_IFF_DUPLICATED)) 3335 error = -1; 3336 else if (ia6->ia6_flags & (IN6_IFF_TENTATIVE | IN6_IFF_DETACHED)) 3337 error = 1; 3338 else 3339 error = 0; 3340 pserialize_read_exit(s); 3341 3342 return error; 3343 } 3344
Indexes created Wed Oct 01 07:09:59 GMT 2025