ip6_output.c revision 1.159
1 /* $NetBSD: ip6_output.c,v 1.159 2014/10/11 20:53:16 christos 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 1.159 2014/10/11 20:53:16 christos Exp $"); 66 67 #include "opt_inet.h" 68 #include "opt_inet6.h" 69 #include "opt_ipsec.h" 70 71 #include <sys/param.h> 72 #include <sys/malloc.h> 73 #include <sys/mbuf.h> 74 #include <sys/errno.h> 75 #include <sys/protosw.h> 76 #include <sys/socket.h> 77 #include <sys/socketvar.h> 78 #include <sys/systm.h> 79 #include <sys/proc.h> 80 #include <sys/kauth.h> 81 82 #include <net/if.h> 83 #include <net/route.h> 84 #include <net/pfil.h> 85 86 #include <netinet/in.h> 87 #include <netinet/in_var.h> 88 #include <netinet/ip6.h> 89 #include <netinet/ip_var.h> 90 #include <netinet/icmp6.h> 91 #include <netinet/in_offload.h> 92 #include <netinet/portalgo.h> 93 #include <netinet6/in6_offload.h> 94 #include <netinet6/ip6_var.h> 95 #include <netinet6/ip6_private.h> 96 #include <netinet6/in6_pcb.h> 97 #include <netinet6/nd6.h> 98 #include <netinet6/ip6protosw.h> 99 #include <netinet6/scope6_var.h> 100 101 #ifdef IPSEC 102 #include <netipsec/ipsec.h> 103 #include <netipsec/ipsec6.h> 104 #include <netipsec/key.h> 105 #include <netipsec/xform.h> 106 #endif 107 108 109 #include <net/net_osdep.h> 110 111 extern pfil_head_t *inet6_pfil_hook; /* XXX */ 112 113 struct ip6_exthdrs { 114 struct mbuf *ip6e_ip6; 115 struct mbuf *ip6e_hbh; 116 struct mbuf *ip6e_dest1; 117 struct mbuf *ip6e_rthdr; 118 struct mbuf *ip6e_dest2; 119 }; 120 121 static int ip6_pcbopt(int, u_char *, int, struct ip6_pktopts **, 122 kauth_cred_t, int); 123 static int ip6_getpcbopt(struct ip6_pktopts *, int, struct sockopt *); 124 static int ip6_setpktopt(int, u_char *, int, struct ip6_pktopts *, kauth_cred_t, 125 int, int, int); 126 static int ip6_setmoptions(const struct sockopt *, struct ip6_moptions **); 127 static int ip6_getmoptions(struct sockopt *, struct ip6_moptions *); 128 static int ip6_copyexthdr(struct mbuf **, void *, int); 129 static int ip6_insertfraghdr(struct mbuf *, struct mbuf *, int, 130 struct ip6_frag **); 131 static int ip6_insert_jumboopt(struct ip6_exthdrs *, u_int32_t); 132 static int ip6_splithdr(struct mbuf *, struct ip6_exthdrs *); 133 static int ip6_getpmtu(struct route *, struct route *, struct ifnet *, 134 const struct in6_addr *, u_long *, int *); 135 static int copypktopts(struct ip6_pktopts *, struct ip6_pktopts *, int); 136 137 #ifdef RFC2292 138 static int ip6_pcbopts(struct ip6_pktopts **, struct socket *, struct sockopt *); 139 #endif 140 141 /* 142 * IP6 output. The packet in mbuf chain m contains a skeletal IP6 143 * header (with pri, len, nxt, hlim, src, dst). 144 * This function may modify ver and hlim only. 145 * The mbuf chain containing the packet will be freed. 146 * The mbuf opt, if present, will not be freed. 147 * 148 * type of "mtu": rt_rmx.rmx_mtu is u_long, ifnet.ifr_mtu is int, and 149 * nd_ifinfo.linkmtu is u_int32_t. so we use u_long to hold largest one, 150 * which is rt_rmx.rmx_mtu. 151 */ 152 int 153 ip6_output( 154 struct mbuf *m0, 155 struct ip6_pktopts *opt, 156 struct route *ro, 157 int flags, 158 struct ip6_moptions *im6o, 159 struct socket *so, 160 struct ifnet **ifpp /* XXX: just for statistics */ 161 ) 162 { 163 struct ip6_hdr *ip6, *mhip6; 164 struct ifnet *ifp, *origifp; 165 struct mbuf *m = m0; 166 int hlen, tlen, len, off; 167 bool tso; 168 struct route ip6route; 169 struct rtentry *rt = NULL; 170 const struct sockaddr_in6 *dst = NULL; 171 struct sockaddr_in6 src_sa, dst_sa; 172 int error = 0; 173 struct in6_ifaddr *ia = NULL; 174 u_long mtu; 175 int alwaysfrag, dontfrag; 176 u_int32_t optlen = 0, plen = 0, unfragpartlen = 0; 177 struct ip6_exthdrs exthdrs; 178 struct in6_addr finaldst, src0, dst0; 179 u_int32_t zone; 180 struct route *ro_pmtu = NULL; 181 int hdrsplit = 0; 182 int needipsec = 0; 183 #ifdef IPSEC 184 struct secpolicy *sp = NULL; 185 #endif 186 187 memset(&ip6route, 0, sizeof(ip6route)); 188 189 #ifdef DIAGNOSTIC 190 if ((m->m_flags & M_PKTHDR) == 0) 191 panic("ip6_output: no HDR"); 192 193 if ((m->m_pkthdr.csum_flags & 194 (M_CSUM_TCPv4|M_CSUM_UDPv4|M_CSUM_TSOv4)) != 0) { 195 panic("ip6_output: IPv4 checksum offload flags: %d", 196 m->m_pkthdr.csum_flags); 197 } 198 199 if ((m->m_pkthdr.csum_flags & (M_CSUM_TCPv6|M_CSUM_UDPv6)) == 200 (M_CSUM_TCPv6|M_CSUM_UDPv6)) { 201 panic("ip6_output: conflicting checksum offload flags: %d", 202 m->m_pkthdr.csum_flags); 203 } 204 #endif 205 206 M_CSUM_DATA_IPv6_HL_SET(m->m_pkthdr.csum_data, sizeof(struct ip6_hdr)); 207 208 #define MAKE_EXTHDR(hp, mp) \ 209 do { \ 210 if (hp) { \ 211 struct ip6_ext *eh = (struct ip6_ext *)(hp); \ 212 error = ip6_copyexthdr((mp), (void *)(hp), \ 213 ((eh)->ip6e_len + 1) << 3); \ 214 if (error) \ 215 goto freehdrs; \ 216 } \ 217 } while (/*CONSTCOND*/ 0) 218 219 memset(&exthdrs, 0, sizeof(exthdrs)); 220 if (opt) { 221 /* Hop-by-Hop options header */ 222 MAKE_EXTHDR(opt->ip6po_hbh, &exthdrs.ip6e_hbh); 223 /* Destination options header(1st part) */ 224 MAKE_EXTHDR(opt->ip6po_dest1, &exthdrs.ip6e_dest1); 225 /* Routing header */ 226 MAKE_EXTHDR(opt->ip6po_rthdr, &exthdrs.ip6e_rthdr); 227 /* Destination options header(2nd part) */ 228 MAKE_EXTHDR(opt->ip6po_dest2, &exthdrs.ip6e_dest2); 229 } 230 231 /* 232 * Calculate the total length of the extension header chain. 233 * Keep the length of the unfragmentable part for fragmentation. 234 */ 235 optlen = 0; 236 if (exthdrs.ip6e_hbh) optlen += exthdrs.ip6e_hbh->m_len; 237 if (exthdrs.ip6e_dest1) optlen += exthdrs.ip6e_dest1->m_len; 238 if (exthdrs.ip6e_rthdr) optlen += exthdrs.ip6e_rthdr->m_len; 239 unfragpartlen = optlen + sizeof(struct ip6_hdr); 240 /* NOTE: we don't add AH/ESP length here. do that later. */ 241 if (exthdrs.ip6e_dest2) optlen += exthdrs.ip6e_dest2->m_len; 242 243 #ifdef IPSEC 244 if (ipsec_used) { 245 /* Check the security policy (SP) for the packet */ 246 247 sp = ipsec6_check_policy(m, so, flags, &needipsec, &error); 248 if (error != 0) { 249 /* 250 * Hack: -EINVAL is used to signal that a packet 251 * should be silently discarded. This is typically 252 * because we asked key management for an SA and 253 * it was delayed (e.g. kicked up to IKE). 254 */ 255 if (error == -EINVAL) 256 error = 0; 257 goto freehdrs; 258 } 259 } 260 #endif /* IPSEC */ 261 262 263 if (needipsec && 264 (m->m_pkthdr.csum_flags & (M_CSUM_UDPv6|M_CSUM_TCPv6)) != 0) { 265 in6_delayed_cksum(m); 266 m->m_pkthdr.csum_flags &= ~(M_CSUM_UDPv6|M_CSUM_TCPv6); 267 } 268 269 270 /* 271 * If we need IPsec, or there is at least one extension header, 272 * separate IP6 header from the payload. 273 */ 274 if ((needipsec || optlen) && !hdrsplit) { 275 if ((error = ip6_splithdr(m, &exthdrs)) != 0) { 276 m = NULL; 277 goto freehdrs; 278 } 279 m = exthdrs.ip6e_ip6; 280 hdrsplit++; 281 } 282 283 /* adjust pointer */ 284 ip6 = mtod(m, struct ip6_hdr *); 285 286 /* adjust mbuf packet header length */ 287 m->m_pkthdr.len += optlen; 288 plen = m->m_pkthdr.len - sizeof(*ip6); 289 290 /* If this is a jumbo payload, insert a jumbo payload option. */ 291 if (plen > IPV6_MAXPACKET) { 292 if (!hdrsplit) { 293 if ((error = ip6_splithdr(m, &exthdrs)) != 0) { 294 m = NULL; 295 goto freehdrs; 296 } 297 m = exthdrs.ip6e_ip6; 298 hdrsplit++; 299 } 300 /* adjust pointer */ 301 ip6 = mtod(m, struct ip6_hdr *); 302 if ((error = ip6_insert_jumboopt(&exthdrs, plen)) != 0) 303 goto freehdrs; 304 optlen += 8; /* XXX JUMBOOPTLEN */ 305 ip6->ip6_plen = 0; 306 } else 307 ip6->ip6_plen = htons(plen); 308 309 /* 310 * Concatenate headers and fill in next header fields. 311 * Here we have, on "m" 312 * IPv6 payload 313 * and we insert headers accordingly. Finally, we should be getting: 314 * IPv6 hbh dest1 rthdr ah* [esp* dest2 payload] 315 * 316 * during the header composing process, "m" points to IPv6 header. 317 * "mprev" points to an extension header prior to esp. 318 */ 319 { 320 u_char *nexthdrp = &ip6->ip6_nxt; 321 struct mbuf *mprev = m; 322 323 /* 324 * we treat dest2 specially. this makes IPsec processing 325 * much easier. the goal here is to make mprev point the 326 * mbuf prior to dest2. 327 * 328 * result: IPv6 dest2 payload 329 * m and mprev will point to IPv6 header. 330 */ 331 if (exthdrs.ip6e_dest2) { 332 if (!hdrsplit) 333 panic("assumption failed: hdr not split"); 334 exthdrs.ip6e_dest2->m_next = m->m_next; 335 m->m_next = exthdrs.ip6e_dest2; 336 *mtod(exthdrs.ip6e_dest2, u_char *) = ip6->ip6_nxt; 337 ip6->ip6_nxt = IPPROTO_DSTOPTS; 338 } 339 340 #define MAKE_CHAIN(m, mp, p, i)\ 341 do {\ 342 if (m) {\ 343 if (!hdrsplit) \ 344 panic("assumption failed: hdr not split"); \ 345 *mtod((m), u_char *) = *(p);\ 346 *(p) = (i);\ 347 p = mtod((m), u_char *);\ 348 (m)->m_next = (mp)->m_next;\ 349 (mp)->m_next = (m);\ 350 (mp) = (m);\ 351 }\ 352 } while (/*CONSTCOND*/ 0) 353 /* 354 * result: IPv6 hbh dest1 rthdr dest2 payload 355 * m will point to IPv6 header. mprev will point to the 356 * extension header prior to dest2 (rthdr in the above case). 357 */ 358 MAKE_CHAIN(exthdrs.ip6e_hbh, mprev, nexthdrp, IPPROTO_HOPOPTS); 359 MAKE_CHAIN(exthdrs.ip6e_dest1, mprev, nexthdrp, 360 IPPROTO_DSTOPTS); 361 MAKE_CHAIN(exthdrs.ip6e_rthdr, mprev, nexthdrp, 362 IPPROTO_ROUTING); 363 364 M_CSUM_DATA_IPv6_HL_SET(m->m_pkthdr.csum_data, 365 sizeof(struct ip6_hdr) + optlen); 366 } 367 368 /* 369 * If there is a routing header, replace destination address field 370 * with the first hop of the routing header. 371 */ 372 if (exthdrs.ip6e_rthdr) { 373 struct ip6_rthdr *rh; 374 struct ip6_rthdr0 *rh0; 375 struct in6_addr *addr; 376 struct sockaddr_in6 sa; 377 378 rh = (struct ip6_rthdr *)(mtod(exthdrs.ip6e_rthdr, 379 struct ip6_rthdr *)); 380 finaldst = ip6->ip6_dst; 381 switch (rh->ip6r_type) { 382 case IPV6_RTHDR_TYPE_0: 383 rh0 = (struct ip6_rthdr0 *)rh; 384 addr = (struct in6_addr *)(rh0 + 1); 385 386 /* 387 * construct a sockaddr_in6 form of 388 * the first hop. 389 * 390 * XXX: we may not have enough 391 * information about its scope zone; 392 * there is no standard API to pass 393 * the information from the 394 * application. 395 */ 396 sockaddr_in6_init(&sa, addr, 0, 0, 0); 397 if ((error = sa6_embedscope(&sa, 398 ip6_use_defzone)) != 0) { 399 goto bad; 400 } 401 ip6->ip6_dst = sa.sin6_addr; 402 (void)memmove(&addr[0], &addr[1], 403 sizeof(struct in6_addr) * 404 (rh0->ip6r0_segleft - 1)); 405 addr[rh0->ip6r0_segleft - 1] = finaldst; 406 /* XXX */ 407 in6_clearscope(addr + rh0->ip6r0_segleft - 1); 408 break; 409 default: /* is it possible? */ 410 error = EINVAL; 411 goto bad; 412 } 413 } 414 415 /* Source address validation */ 416 if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) && 417 (flags & IPV6_UNSPECSRC) == 0) { 418 error = EOPNOTSUPP; 419 IP6_STATINC(IP6_STAT_BADSCOPE); 420 goto bad; 421 } 422 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) { 423 error = EOPNOTSUPP; 424 IP6_STATINC(IP6_STAT_BADSCOPE); 425 goto bad; 426 } 427 428 IP6_STATINC(IP6_STAT_LOCALOUT); 429 430 /* 431 * Route packet. 432 */ 433 /* initialize cached route */ 434 if (ro == NULL) { 435 ro = &ip6route; 436 } 437 ro_pmtu = ro; 438 if (opt && opt->ip6po_rthdr) 439 ro = &opt->ip6po_route; 440 441 /* 442 * if specified, try to fill in the traffic class field. 443 * do not override if a non-zero value is already set. 444 * we check the diffserv field and the ecn field separately. 445 */ 446 if (opt && opt->ip6po_tclass >= 0) { 447 int mask = 0; 448 449 if ((ip6->ip6_flow & htonl(0xfc << 20)) == 0) 450 mask |= 0xfc; 451 if ((ip6->ip6_flow & htonl(0x03 << 20)) == 0) 452 mask |= 0x03; 453 if (mask != 0) 454 ip6->ip6_flow |= htonl((opt->ip6po_tclass & mask) << 20); 455 } 456 457 /* fill in or override the hop limit field, if necessary. */ 458 if (opt && opt->ip6po_hlim != -1) 459 ip6->ip6_hlim = opt->ip6po_hlim & 0xff; 460 else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) { 461 if (im6o != NULL) 462 ip6->ip6_hlim = im6o->im6o_multicast_hlim; 463 else 464 ip6->ip6_hlim = ip6_defmcasthlim; 465 } 466 467 #ifdef IPSEC 468 if (needipsec) { 469 int s = splsoftnet(); 470 error = ipsec6_process_packet(m, sp->req); 471 472 /* 473 * Preserve KAME behaviour: ENOENT can be returned 474 * when an SA acquire is in progress. Don't propagate 475 * this to user-level; it confuses applications. 476 * XXX this will go away when the SADB is redone. 477 */ 478 if (error == ENOENT) 479 error = 0; 480 splx(s); 481 goto done; 482 } 483 #endif /* IPSEC */ 484 485 /* adjust pointer */ 486 ip6 = mtod(m, struct ip6_hdr *); 487 488 sockaddr_in6_init(&dst_sa, &ip6->ip6_dst, 0, 0, 0); 489 if ((error = in6_selectroute(&dst_sa, opt, im6o, ro, 490 &ifp, &rt, 0)) != 0) { 491 if (ifp != NULL) 492 in6_ifstat_inc(ifp, ifs6_out_discard); 493 goto bad; 494 } 495 if (rt == NULL) { 496 /* 497 * If in6_selectroute() does not return a route entry, 498 * dst may not have been updated. 499 */ 500 error = rtcache_setdst(ro, sin6tosa(&dst_sa)); 501 if (error) { 502 goto bad; 503 } 504 } 505 506 /* 507 * then rt (for unicast) and ifp must be non-NULL valid values. 508 */ 509 if ((flags & IPV6_FORWARDING) == 0) { 510 /* XXX: the FORWARDING flag can be set for mrouting. */ 511 in6_ifstat_inc(ifp, ifs6_out_request); 512 } 513 if (rt != NULL) { 514 ia = (struct in6_ifaddr *)(rt->rt_ifa); 515 rt->rt_use++; 516 } 517 518 /* 519 * The outgoing interface must be in the zone of source and 520 * destination addresses. We should use ia_ifp to support the 521 * case of sending packets to an address of our own. 522 */ 523 if (ia != NULL && ia->ia_ifp) 524 origifp = ia->ia_ifp; 525 else 526 origifp = ifp; 527 528 src0 = ip6->ip6_src; 529 if (in6_setscope(&src0, origifp, &zone)) 530 goto badscope; 531 sockaddr_in6_init(&src_sa, &ip6->ip6_src, 0, 0, 0); 532 if (sa6_recoverscope(&src_sa) || zone != src_sa.sin6_scope_id) 533 goto badscope; 534 535 dst0 = ip6->ip6_dst; 536 if (in6_setscope(&dst0, origifp, &zone)) 537 goto badscope; 538 /* re-initialize to be sure */ 539 sockaddr_in6_init(&dst_sa, &ip6->ip6_dst, 0, 0, 0); 540 if (sa6_recoverscope(&dst_sa) || zone != dst_sa.sin6_scope_id) 541 goto badscope; 542 543 /* scope check is done. */ 544 545 if (rt == NULL || IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) { 546 if (dst == NULL) 547 dst = satocsin6(rtcache_getdst(ro)); 548 KASSERT(dst != NULL); 549 } else if (opt && rtcache_validate(&opt->ip6po_nextroute) != NULL) { 550 /* 551 * The nexthop is explicitly specified by the 552 * application. We assume the next hop is an IPv6 553 * address. 554 */ 555 dst = (struct sockaddr_in6 *)opt->ip6po_nexthop; 556 } else if ((rt->rt_flags & RTF_GATEWAY)) 557 dst = (struct sockaddr_in6 *)rt->rt_gateway; 558 else if (dst == NULL) 559 dst = satocsin6(rtcache_getdst(ro)); 560 561 /* 562 * XXXXXX: original code follows: 563 */ 564 if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) 565 m->m_flags &= ~(M_BCAST | M_MCAST); /* just in case */ 566 else { 567 struct in6_multi *in6m; 568 569 m->m_flags = (m->m_flags & ~M_BCAST) | M_MCAST; 570 571 in6_ifstat_inc(ifp, ifs6_out_mcast); 572 573 /* 574 * Confirm that the outgoing interface supports multicast. 575 */ 576 if (!(ifp->if_flags & IFF_MULTICAST)) { 577 IP6_STATINC(IP6_STAT_NOROUTE); 578 in6_ifstat_inc(ifp, ifs6_out_discard); 579 error = ENETUNREACH; 580 goto bad; 581 } 582 583 IN6_LOOKUP_MULTI(ip6->ip6_dst, ifp, in6m); 584 if (in6m != NULL && 585 (im6o == NULL || im6o->im6o_multicast_loop)) { 586 /* 587 * If we belong to the destination multicast group 588 * on the outgoing interface, and the caller did not 589 * forbid loopback, loop back a copy. 590 */ 591 KASSERT(dst != NULL); 592 ip6_mloopback(ifp, m, dst); 593 } else { 594 /* 595 * If we are acting as a multicast router, perform 596 * multicast forwarding as if the packet had just 597 * arrived on the interface to which we are about 598 * to send. The multicast forwarding function 599 * recursively calls this function, using the 600 * IPV6_FORWARDING flag to prevent infinite recursion. 601 * 602 * Multicasts that are looped back by ip6_mloopback(), 603 * above, will be forwarded by the ip6_input() routine, 604 * if necessary. 605 */ 606 if (ip6_mrouter && (flags & IPV6_FORWARDING) == 0) { 607 if (ip6_mforward(ip6, ifp, m) != 0) { 608 m_freem(m); 609 goto done; 610 } 611 } 612 } 613 /* 614 * Multicasts with a hoplimit of zero may be looped back, 615 * above, but must not be transmitted on a network. 616 * Also, multicasts addressed to the loopback interface 617 * are not sent -- the above call to ip6_mloopback() will 618 * loop back a copy if this host actually belongs to the 619 * destination group on the loopback interface. 620 */ 621 if (ip6->ip6_hlim == 0 || (ifp->if_flags & IFF_LOOPBACK) || 622 IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst)) { 623 m_freem(m); 624 goto done; 625 } 626 } 627 628 /* 629 * Fill the outgoing inteface to tell the upper layer 630 * to increment per-interface statistics. 631 */ 632 if (ifpp) 633 *ifpp = ifp; 634 635 /* Determine path MTU. */ 636 if ((error = ip6_getpmtu(ro_pmtu, ro, ifp, &finaldst, &mtu, 637 &alwaysfrag)) != 0) 638 goto bad; 639 640 /* 641 * The caller of this function may specify to use the minimum MTU 642 * in some cases. 643 * An advanced API option (IPV6_USE_MIN_MTU) can also override MTU 644 * setting. The logic is a bit complicated; by default, unicast 645 * packets will follow path MTU while multicast packets will be sent at 646 * the minimum MTU. If IP6PO_MINMTU_ALL is specified, all packets 647 * including unicast ones will be sent at the minimum MTU. Multicast 648 * packets will always be sent at the minimum MTU unless 649 * IP6PO_MINMTU_DISABLE is explicitly specified. 650 * See RFC 3542 for more details. 651 */ 652 if (mtu > IPV6_MMTU) { 653 if ((flags & IPV6_MINMTU)) 654 mtu = IPV6_MMTU; 655 else if (opt && opt->ip6po_minmtu == IP6PO_MINMTU_ALL) 656 mtu = IPV6_MMTU; 657 else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) && 658 (opt == NULL || 659 opt->ip6po_minmtu != IP6PO_MINMTU_DISABLE)) { 660 mtu = IPV6_MMTU; 661 } 662 } 663 664 /* 665 * clear embedded scope identifiers if necessary. 666 * in6_clearscope will touch the addresses only when necessary. 667 */ 668 in6_clearscope(&ip6->ip6_src); 669 in6_clearscope(&ip6->ip6_dst); 670 671 /* 672 * If the outgoing packet contains a hop-by-hop options header, 673 * it must be examined and processed even by the source node. 674 * (RFC 2460, section 4.) 675 */ 676 if (ip6->ip6_nxt == IPV6_HOPOPTS) { 677 u_int32_t dummy1; /* XXX unused */ 678 u_int32_t dummy2; /* XXX unused */ 679 int hoff = sizeof(struct ip6_hdr); 680 681 if (ip6_hopopts_input(&dummy1, &dummy2, &m, &hoff)) { 682 /* m was already freed at this point */ 683 error = EINVAL;/* better error? */ 684 goto done; 685 } 686 687 ip6 = mtod(m, struct ip6_hdr *); 688 } 689 690 /* 691 * Run through list of hooks for output packets. 692 */ 693 if ((error = pfil_run_hooks(inet6_pfil_hook, &m, ifp, PFIL_OUT)) != 0) 694 goto done; 695 if (m == NULL) 696 goto done; 697 ip6 = mtod(m, struct ip6_hdr *); 698 699 /* 700 * Send the packet to the outgoing interface. 701 * If necessary, do IPv6 fragmentation before sending. 702 * 703 * the logic here is rather complex: 704 * 1: normal case (dontfrag == 0, alwaysfrag == 0) 705 * 1-a: send as is if tlen <= path mtu 706 * 1-b: fragment if tlen > path mtu 707 * 708 * 2: if user asks us not to fragment (dontfrag == 1) 709 * 2-a: send as is if tlen <= interface mtu 710 * 2-b: error if tlen > interface mtu 711 * 712 * 3: if we always need to attach fragment header (alwaysfrag == 1) 713 * always fragment 714 * 715 * 4: if dontfrag == 1 && alwaysfrag == 1 716 * error, as we cannot handle this conflicting request 717 */ 718 tlen = m->m_pkthdr.len; 719 tso = (m->m_pkthdr.csum_flags & M_CSUM_TSOv6) != 0; 720 if (opt && (opt->ip6po_flags & IP6PO_DONTFRAG)) 721 dontfrag = 1; 722 else 723 dontfrag = 0; 724 725 if (dontfrag && alwaysfrag) { /* case 4 */ 726 /* conflicting request - can't transmit */ 727 error = EMSGSIZE; 728 goto bad; 729 } 730 if (dontfrag && (!tso && tlen > IN6_LINKMTU(ifp))) { /* case 2-b */ 731 /* 732 * Even if the DONTFRAG option is specified, we cannot send the 733 * packet when the data length is larger than the MTU of the 734 * outgoing interface. 735 * Notify the error by sending IPV6_PATHMTU ancillary data as 736 * well as returning an error code (the latter is not described 737 * in the API spec.) 738 */ 739 u_int32_t mtu32; 740 struct ip6ctlparam ip6cp; 741 742 mtu32 = (u_int32_t)mtu; 743 memset(&ip6cp, 0, sizeof(ip6cp)); 744 ip6cp.ip6c_cmdarg = (void *)&mtu32; 745 pfctlinput2(PRC_MSGSIZE, 746 rtcache_getdst(ro_pmtu), &ip6cp); 747 748 error = EMSGSIZE; 749 goto bad; 750 } 751 752 /* 753 * transmit packet without fragmentation 754 */ 755 if (dontfrag || (!alwaysfrag && (tlen <= mtu || tso))) { 756 /* case 1-a and 2-a */ 757 struct in6_ifaddr *ia6; 758 int sw_csum; 759 760 ip6 = mtod(m, struct ip6_hdr *); 761 ia6 = in6_ifawithifp(ifp, &ip6->ip6_src); 762 if (ia6) { 763 /* Record statistics for this interface address. */ 764 ia6->ia_ifa.ifa_data.ifad_outbytes += m->m_pkthdr.len; 765 } 766 767 sw_csum = m->m_pkthdr.csum_flags & ~ifp->if_csum_flags_tx; 768 if ((sw_csum & (M_CSUM_UDPv6|M_CSUM_TCPv6)) != 0) { 769 if (IN6_NEED_CHECKSUM(ifp, 770 sw_csum & (M_CSUM_UDPv6|M_CSUM_TCPv6))) { 771 in6_delayed_cksum(m); 772 } 773 m->m_pkthdr.csum_flags &= ~(M_CSUM_UDPv6|M_CSUM_TCPv6); 774 } 775 776 KASSERT(dst != NULL); 777 if (__predict_true(!tso || 778 (ifp->if_capenable & IFCAP_TSOv6) != 0)) { 779 error = nd6_output(ifp, origifp, m, dst, rt); 780 } else { 781 error = ip6_tso_output(ifp, origifp, m, dst, rt); 782 } 783 goto done; 784 } 785 786 if (tso) { 787 error = EINVAL; /* XXX */ 788 goto bad; 789 } 790 791 /* 792 * try to fragment the packet. case 1-b and 3 793 */ 794 if (mtu < IPV6_MMTU) { 795 /* path MTU cannot be less than IPV6_MMTU */ 796 error = EMSGSIZE; 797 in6_ifstat_inc(ifp, ifs6_out_fragfail); 798 goto bad; 799 } else if (ip6->ip6_plen == 0) { 800 /* jumbo payload cannot be fragmented */ 801 error = EMSGSIZE; 802 in6_ifstat_inc(ifp, ifs6_out_fragfail); 803 goto bad; 804 } else { 805 struct mbuf **mnext, *m_frgpart; 806 struct ip6_frag *ip6f; 807 u_int32_t id = htonl(ip6_randomid()); 808 u_char nextproto; 809 #if 0 /* see below */ 810 struct ip6ctlparam ip6cp; 811 u_int32_t mtu32; 812 #endif 813 814 /* 815 * Too large for the destination or interface; 816 * fragment if possible. 817 * Must be able to put at least 8 bytes per fragment. 818 */ 819 hlen = unfragpartlen; 820 if (mtu > IPV6_MAXPACKET) 821 mtu = IPV6_MAXPACKET; 822 823 #if 0 824 /* 825 * It is believed this code is a leftover from the 826 * development of the IPV6_RECVPATHMTU sockopt and 827 * associated work to implement RFC3542. 828 * It's not entirely clear what the intent of the API 829 * is at this point, so disable this code for now. 830 * The IPV6_RECVPATHMTU sockopt and/or IPV6_DONTFRAG 831 * will send notifications if the application requests. 832 */ 833 834 /* Notify a proper path MTU to applications. */ 835 mtu32 = (u_int32_t)mtu; 836 memset(&ip6cp, 0, sizeof(ip6cp)); 837 ip6cp.ip6c_cmdarg = (void *)&mtu32; 838 pfctlinput2(PRC_MSGSIZE, 839 rtcache_getdst(ro_pmtu), &ip6cp); 840 #endif 841 842 len = (mtu - hlen - sizeof(struct ip6_frag)) & ~7; 843 if (len < 8) { 844 error = EMSGSIZE; 845 in6_ifstat_inc(ifp, ifs6_out_fragfail); 846 goto bad; 847 } 848 849 mnext = &m->m_nextpkt; 850 851 /* 852 * Change the next header field of the last header in the 853 * unfragmentable part. 854 */ 855 if (exthdrs.ip6e_rthdr) { 856 nextproto = *mtod(exthdrs.ip6e_rthdr, u_char *); 857 *mtod(exthdrs.ip6e_rthdr, u_char *) = IPPROTO_FRAGMENT; 858 } else if (exthdrs.ip6e_dest1) { 859 nextproto = *mtod(exthdrs.ip6e_dest1, u_char *); 860 *mtod(exthdrs.ip6e_dest1, u_char *) = IPPROTO_FRAGMENT; 861 } else if (exthdrs.ip6e_hbh) { 862 nextproto = *mtod(exthdrs.ip6e_hbh, u_char *); 863 *mtod(exthdrs.ip6e_hbh, u_char *) = IPPROTO_FRAGMENT; 864 } else { 865 nextproto = ip6->ip6_nxt; 866 ip6->ip6_nxt = IPPROTO_FRAGMENT; 867 } 868 869 if ((m->m_pkthdr.csum_flags & (M_CSUM_UDPv6|M_CSUM_TCPv6)) 870 != 0) { 871 if (IN6_NEED_CHECKSUM(ifp, 872 m->m_pkthdr.csum_flags & 873 (M_CSUM_UDPv6|M_CSUM_TCPv6))) { 874 in6_delayed_cksum(m); 875 } 876 m->m_pkthdr.csum_flags &= ~(M_CSUM_UDPv6|M_CSUM_TCPv6); 877 } 878 879 /* 880 * Loop through length of segment after first fragment, 881 * make new header and copy data of each part and link onto 882 * chain. 883 */ 884 m0 = m; 885 for (off = hlen; off < tlen; off += len) { 886 struct mbuf *mlast; 887 888 MGETHDR(m, M_DONTWAIT, MT_HEADER); 889 if (!m) { 890 error = ENOBUFS; 891 IP6_STATINC(IP6_STAT_ODROPPED); 892 goto sendorfree; 893 } 894 m->m_pkthdr.rcvif = NULL; 895 m->m_flags = m0->m_flags & M_COPYFLAGS; 896 *mnext = m; 897 mnext = &m->m_nextpkt; 898 m->m_data += max_linkhdr; 899 mhip6 = mtod(m, struct ip6_hdr *); 900 *mhip6 = *ip6; 901 m->m_len = sizeof(*mhip6); 902 /* 903 * ip6f must be valid if error is 0. But how 904 * can a compiler be expected to infer this? 905 */ 906 ip6f = NULL; 907 error = ip6_insertfraghdr(m0, m, hlen, &ip6f); 908 if (error) { 909 IP6_STATINC(IP6_STAT_ODROPPED); 910 goto sendorfree; 911 } 912 ip6f->ip6f_offlg = htons((u_int16_t)((off - hlen) & ~7)); 913 if (off + len >= tlen) 914 len = tlen - off; 915 else 916 ip6f->ip6f_offlg |= IP6F_MORE_FRAG; 917 mhip6->ip6_plen = htons((u_int16_t)(len + hlen + 918 sizeof(*ip6f) - sizeof(struct ip6_hdr))); 919 if ((m_frgpart = m_copy(m0, off, len)) == 0) { 920 error = ENOBUFS; 921 IP6_STATINC(IP6_STAT_ODROPPED); 922 goto sendorfree; 923 } 924 for (mlast = m; mlast->m_next; mlast = mlast->m_next) 925 ; 926 mlast->m_next = m_frgpart; 927 m->m_pkthdr.len = len + hlen + sizeof(*ip6f); 928 m->m_pkthdr.rcvif = NULL; 929 ip6f->ip6f_reserved = 0; 930 ip6f->ip6f_ident = id; 931 ip6f->ip6f_nxt = nextproto; 932 IP6_STATINC(IP6_STAT_OFRAGMENTS); 933 in6_ifstat_inc(ifp, ifs6_out_fragcreat); 934 } 935 936 in6_ifstat_inc(ifp, ifs6_out_fragok); 937 } 938 939 /* 940 * Remove leading garbages. 941 */ 942 sendorfree: 943 m = m0->m_nextpkt; 944 m0->m_nextpkt = 0; 945 m_freem(m0); 946 for (m0 = m; m; m = m0) { 947 m0 = m->m_nextpkt; 948 m->m_nextpkt = 0; 949 if (error == 0) { 950 struct in6_ifaddr *ia6; 951 ip6 = mtod(m, struct ip6_hdr *); 952 ia6 = in6_ifawithifp(ifp, &ip6->ip6_src); 953 if (ia6) { 954 /* 955 * Record statistics for this interface 956 * address. 957 */ 958 ia6->ia_ifa.ifa_data.ifad_outbytes += 959 m->m_pkthdr.len; 960 } 961 KASSERT(dst != NULL); 962 error = nd6_output(ifp, origifp, m, dst, rt); 963 } else 964 m_freem(m); 965 } 966 967 if (error == 0) 968 IP6_STATINC(IP6_STAT_FRAGMENTED); 969 970 done: 971 rtcache_free(&ip6route); 972 973 #ifdef IPSEC 974 if (sp != NULL) 975 KEY_FREESP(&sp); 976 #endif /* IPSEC */ 977 978 979 return (error); 980 981 freehdrs: 982 m_freem(exthdrs.ip6e_hbh); /* m_freem will check if mbuf is 0 */ 983 m_freem(exthdrs.ip6e_dest1); 984 m_freem(exthdrs.ip6e_rthdr); 985 m_freem(exthdrs.ip6e_dest2); 986 /* FALLTHROUGH */ 987 bad: 988 m_freem(m); 989 goto done; 990 badscope: 991 IP6_STATINC(IP6_STAT_BADSCOPE); 992 in6_ifstat_inc(origifp, ifs6_out_discard); 993 if (error == 0) 994 error = EHOSTUNREACH; /* XXX */ 995 goto bad; 996 } 997 998 static int 999 ip6_copyexthdr(struct mbuf **mp, void *hdr, int hlen) 1000 { 1001 struct mbuf *m; 1002 1003 if (hlen > MCLBYTES) 1004 return (ENOBUFS); /* XXX */ 1005 1006 MGET(m, M_DONTWAIT, MT_DATA); 1007 if (!m) 1008 return (ENOBUFS); 1009 1010 if (hlen > MLEN) { 1011 MCLGET(m, M_DONTWAIT); 1012 if ((m->m_flags & M_EXT) == 0) { 1013 m_free(m); 1014 return (ENOBUFS); 1015 } 1016 } 1017 m->m_len = hlen; 1018 if (hdr) 1019 bcopy(hdr, mtod(m, void *), hlen); 1020 1021 *mp = m; 1022 return (0); 1023 } 1024 1025 /* 1026 * Process a delayed payload checksum calculation. 1027 */ 1028 void 1029 in6_delayed_cksum(struct mbuf *m) 1030 { 1031 uint16_t csum, offset; 1032 1033 KASSERT((m->m_pkthdr.csum_flags & (M_CSUM_UDPv6|M_CSUM_TCPv6)) != 0); 1034 KASSERT((~m->m_pkthdr.csum_flags & (M_CSUM_UDPv6|M_CSUM_TCPv6)) != 0); 1035 KASSERT((m->m_pkthdr.csum_flags 1036 & (M_CSUM_UDPv4|M_CSUM_TCPv4|M_CSUM_TSOv4)) == 0); 1037 1038 offset = M_CSUM_DATA_IPv6_HL(m->m_pkthdr.csum_data); 1039 csum = in6_cksum(m, 0, offset, m->m_pkthdr.len - offset); 1040 if (csum == 0 && (m->m_pkthdr.csum_flags & M_CSUM_UDPv6) != 0) { 1041 csum = 0xffff; 1042 } 1043 1044 offset += M_CSUM_DATA_IPv6_OFFSET(m->m_pkthdr.csum_data); 1045 if ((offset + sizeof(csum)) > m->m_len) { 1046 m_copyback(m, offset, sizeof(csum), &csum); 1047 } else { 1048 *(uint16_t *)(mtod(m, char *) + offset) = csum; 1049 } 1050 } 1051 1052 /* 1053 * Insert jumbo payload option. 1054 */ 1055 static int 1056 ip6_insert_jumboopt(struct ip6_exthdrs *exthdrs, u_int32_t plen) 1057 { 1058 struct mbuf *mopt; 1059 u_int8_t *optbuf; 1060 u_int32_t v; 1061 1062 #define JUMBOOPTLEN 8 /* length of jumbo payload option and padding */ 1063 1064 /* 1065 * If there is no hop-by-hop options header, allocate new one. 1066 * If there is one but it doesn't have enough space to store the 1067 * jumbo payload option, allocate a cluster to store the whole options. 1068 * Otherwise, use it to store the options. 1069 */ 1070 if (exthdrs->ip6e_hbh == 0) { 1071 MGET(mopt, M_DONTWAIT, MT_DATA); 1072 if (mopt == 0) 1073 return (ENOBUFS); 1074 mopt->m_len = JUMBOOPTLEN; 1075 optbuf = mtod(mopt, u_int8_t *); 1076 optbuf[1] = 0; /* = ((JUMBOOPTLEN) >> 3) - 1 */ 1077 exthdrs->ip6e_hbh = mopt; 1078 } else { 1079 struct ip6_hbh *hbh; 1080 1081 mopt = exthdrs->ip6e_hbh; 1082 if (M_TRAILINGSPACE(mopt) < JUMBOOPTLEN) { 1083 /* 1084 * XXX assumption: 1085 * - exthdrs->ip6e_hbh is not referenced from places 1086 * other than exthdrs. 1087 * - exthdrs->ip6e_hbh is not an mbuf chain. 1088 */ 1089 int oldoptlen = mopt->m_len; 1090 struct mbuf *n; 1091 1092 /* 1093 * XXX: give up if the whole (new) hbh header does 1094 * not fit even in an mbuf cluster. 1095 */ 1096 if (oldoptlen + JUMBOOPTLEN > MCLBYTES) 1097 return (ENOBUFS); 1098 1099 /* 1100 * As a consequence, we must always prepare a cluster 1101 * at this point. 1102 */ 1103 MGET(n, M_DONTWAIT, MT_DATA); 1104 if (n) { 1105 MCLGET(n, M_DONTWAIT); 1106 if ((n->m_flags & M_EXT) == 0) { 1107 m_freem(n); 1108 n = NULL; 1109 } 1110 } 1111 if (!n) 1112 return (ENOBUFS); 1113 n->m_len = oldoptlen + JUMBOOPTLEN; 1114 bcopy(mtod(mopt, void *), mtod(n, void *), 1115 oldoptlen); 1116 optbuf = mtod(n, u_int8_t *) + oldoptlen; 1117 m_freem(mopt); 1118 mopt = exthdrs->ip6e_hbh = n; 1119 } else { 1120 optbuf = mtod(mopt, u_int8_t *) + mopt->m_len; 1121 mopt->m_len += JUMBOOPTLEN; 1122 } 1123 optbuf[0] = IP6OPT_PADN; 1124 optbuf[1] = 0; 1125 1126 /* 1127 * Adjust the header length according to the pad and 1128 * the jumbo payload option. 1129 */ 1130 hbh = mtod(mopt, struct ip6_hbh *); 1131 hbh->ip6h_len += (JUMBOOPTLEN >> 3); 1132 } 1133 1134 /* fill in the option. */ 1135 optbuf[2] = IP6OPT_JUMBO; 1136 optbuf[3] = 4; 1137 v = (u_int32_t)htonl(plen + JUMBOOPTLEN); 1138 bcopy(&v, &optbuf[4], sizeof(u_int32_t)); 1139 1140 /* finally, adjust the packet header length */ 1141 exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN; 1142 1143 return (0); 1144 #undef JUMBOOPTLEN 1145 } 1146 1147 /* 1148 * Insert fragment header and copy unfragmentable header portions. 1149 * 1150 * *frghdrp will not be read, and it is guaranteed that either an 1151 * error is returned or that *frghdrp will point to space allocated 1152 * for the fragment header. 1153 */ 1154 static int 1155 ip6_insertfraghdr(struct mbuf *m0, struct mbuf *m, int hlen, 1156 struct ip6_frag **frghdrp) 1157 { 1158 struct mbuf *n, *mlast; 1159 1160 if (hlen > sizeof(struct ip6_hdr)) { 1161 n = m_copym(m0, sizeof(struct ip6_hdr), 1162 hlen - sizeof(struct ip6_hdr), M_DONTWAIT); 1163 if (n == 0) 1164 return (ENOBUFS); 1165 m->m_next = n; 1166 } else 1167 n = m; 1168 1169 /* Search for the last mbuf of unfragmentable part. */ 1170 for (mlast = n; mlast->m_next; mlast = mlast->m_next) 1171 ; 1172 1173 if ((mlast->m_flags & M_EXT) == 0 && 1174 M_TRAILINGSPACE(mlast) >= sizeof(struct ip6_frag)) { 1175 /* use the trailing space of the last mbuf for the fragment hdr */ 1176 *frghdrp = (struct ip6_frag *)(mtod(mlast, char *) + 1177 mlast->m_len); 1178 mlast->m_len += sizeof(struct ip6_frag); 1179 m->m_pkthdr.len += sizeof(struct ip6_frag); 1180 } else { 1181 /* allocate a new mbuf for the fragment header */ 1182 struct mbuf *mfrg; 1183 1184 MGET(mfrg, M_DONTWAIT, MT_DATA); 1185 if (mfrg == 0) 1186 return (ENOBUFS); 1187 mfrg->m_len = sizeof(struct ip6_frag); 1188 *frghdrp = mtod(mfrg, struct ip6_frag *); 1189 mlast->m_next = mfrg; 1190 } 1191 1192 return (0); 1193 } 1194 1195 static int 1196 ip6_getpmtu(struct route *ro_pmtu, struct route *ro, struct ifnet *ifp, 1197 const struct in6_addr *dst, u_long *mtup, int *alwaysfragp) 1198 { 1199 struct rtentry *rt; 1200 u_int32_t mtu = 0; 1201 int alwaysfrag = 0; 1202 int error = 0; 1203 1204 if (ro_pmtu != ro) { 1205 union { 1206 struct sockaddr dst; 1207 struct sockaddr_in6 dst6; 1208 } u; 1209 1210 /* The first hop and the final destination may differ. */ 1211 sockaddr_in6_init(&u.dst6, dst, 0, 0, 0); 1212 rt = rtcache_lookup(ro_pmtu, &u.dst); 1213 } else 1214 rt = rtcache_validate(ro_pmtu); 1215 if (rt != NULL) { 1216 u_int32_t ifmtu; 1217 1218 if (ifp == NULL) 1219 ifp = rt->rt_ifp; 1220 ifmtu = IN6_LINKMTU(ifp); 1221 mtu = rt->rt_rmx.rmx_mtu; 1222 if (mtu == 0) 1223 mtu = ifmtu; 1224 else if (mtu < IPV6_MMTU) { 1225 /* 1226 * RFC2460 section 5, last paragraph: 1227 * if we record ICMPv6 too big message with 1228 * mtu < IPV6_MMTU, transmit packets sized IPV6_MMTU 1229 * or smaller, with fragment header attached. 1230 * (fragment header is needed regardless from the 1231 * packet size, for translators to identify packets) 1232 */ 1233 alwaysfrag = 1; 1234 mtu = IPV6_MMTU; 1235 } else if (mtu > ifmtu) { 1236 /* 1237 * The MTU on the route is larger than the MTU on 1238 * the interface! This shouldn't happen, unless the 1239 * MTU of the interface has been changed after the 1240 * interface was brought up. Change the MTU in the 1241 * route to match the interface MTU (as long as the 1242 * field isn't locked). 1243 */ 1244 mtu = ifmtu; 1245 if (!(rt->rt_rmx.rmx_locks & RTV_MTU)) 1246 rt->rt_rmx.rmx_mtu = mtu; 1247 } 1248 } else if (ifp) { 1249 mtu = IN6_LINKMTU(ifp); 1250 } else 1251 error = EHOSTUNREACH; /* XXX */ 1252 1253 *mtup = mtu; 1254 if (alwaysfragp) 1255 *alwaysfragp = alwaysfrag; 1256 return (error); 1257 } 1258 1259 /* 1260 * IP6 socket option processing. 1261 */ 1262 int 1263 ip6_ctloutput(int op, struct socket *so, struct sockopt *sopt) 1264 { 1265 int optdatalen, uproto; 1266 void *optdata; 1267 struct in6pcb *in6p = sotoin6pcb(so); 1268 struct ip_moptions **mopts; 1269 int error, optval; 1270 int level, optname; 1271 1272 KASSERT(sopt != NULL); 1273 1274 level = sopt->sopt_level; 1275 optname = sopt->sopt_name; 1276 1277 error = optval = 0; 1278 uproto = (int)so->so_proto->pr_protocol; 1279 1280 switch (level) { 1281 case IPPROTO_IP: 1282 switch (optname) { 1283 case IP_ADD_MEMBERSHIP: 1284 case IP_DROP_MEMBERSHIP: 1285 case IP_MULTICAST_IF: 1286 case IP_MULTICAST_LOOP: 1287 case IP_MULTICAST_TTL: 1288 mopts = &in6p->in6p_v4moptions; 1289 switch (op) { 1290 case PRCO_GETOPT: 1291 return ip_getmoptions(*mopts, sopt); 1292 case PRCO_SETOPT: 1293 return ip_setmoptions(mopts, sopt); 1294 default: 1295 return EINVAL; 1296 } 1297 default: 1298 return ENOPROTOOPT; 1299 } 1300 case IPPROTO_IPV6: 1301 break; 1302 default: 1303 return ENOPROTOOPT; 1304 } 1305 switch (op) { 1306 case PRCO_SETOPT: 1307 switch (optname) { 1308 #ifdef RFC2292 1309 case IPV6_2292PKTOPTIONS: 1310 error = ip6_pcbopts(&in6p->in6p_outputopts, so, sopt); 1311 break; 1312 #endif 1313 1314 /* 1315 * Use of some Hop-by-Hop options or some 1316 * Destination options, might require special 1317 * privilege. That is, normal applications 1318 * (without special privilege) might be forbidden 1319 * from setting certain options in outgoing packets, 1320 * and might never see certain options in received 1321 * packets. [RFC 2292 Section 6] 1322 * KAME specific note: 1323 * KAME prevents non-privileged users from sending or 1324 * receiving ANY hbh/dst options in order to avoid 1325 * overhead of parsing options in the kernel. 1326 */ 1327 case IPV6_RECVHOPOPTS: 1328 case IPV6_RECVDSTOPTS: 1329 case IPV6_RECVRTHDRDSTOPTS: 1330 error = kauth_authorize_network(kauth_cred_get(), 1331 KAUTH_NETWORK_IPV6, KAUTH_REQ_NETWORK_IPV6_HOPBYHOP, 1332 NULL, NULL, NULL); 1333 if (error) 1334 break; 1335 /* FALLTHROUGH */ 1336 case IPV6_UNICAST_HOPS: 1337 case IPV6_HOPLIMIT: 1338 case IPV6_FAITH: 1339 1340 case IPV6_RECVPKTINFO: 1341 case IPV6_RECVHOPLIMIT: 1342 case IPV6_RECVRTHDR: 1343 case IPV6_RECVPATHMTU: 1344 case IPV6_RECVTCLASS: 1345 case IPV6_V6ONLY: 1346 error = sockopt_getint(sopt, &optval); 1347 if (error) 1348 break; 1349 switch (optname) { 1350 case IPV6_UNICAST_HOPS: 1351 if (optval < -1 || optval >= 256) 1352 error = EINVAL; 1353 else { 1354 /* -1 = kernel default */ 1355 in6p->in6p_hops = optval; 1356 } 1357 break; 1358 #define OPTSET(bit) \ 1359 do { \ 1360 if (optval) \ 1361 in6p->in6p_flags |= (bit); \ 1362 else \ 1363 in6p->in6p_flags &= ~(bit); \ 1364 } while (/*CONSTCOND*/ 0) 1365 1366 #ifdef RFC2292 1367 #define OPTSET2292(bit) \ 1368 do { \ 1369 in6p->in6p_flags |= IN6P_RFC2292; \ 1370 if (optval) \ 1371 in6p->in6p_flags |= (bit); \ 1372 else \ 1373 in6p->in6p_flags &= ~(bit); \ 1374 } while (/*CONSTCOND*/ 0) 1375 #endif 1376 1377 #define OPTBIT(bit) (in6p->in6p_flags & (bit) ? 1 : 0) 1378 1379 case IPV6_RECVPKTINFO: 1380 #ifdef RFC2292 1381 /* cannot mix with RFC2292 */ 1382 if (OPTBIT(IN6P_RFC2292)) { 1383 error = EINVAL; 1384 break; 1385 } 1386 #endif 1387 OPTSET(IN6P_PKTINFO); 1388 break; 1389 1390 case IPV6_HOPLIMIT: 1391 { 1392 struct ip6_pktopts **optp; 1393 1394 #ifdef RFC2292 1395 /* cannot mix with RFC2292 */ 1396 if (OPTBIT(IN6P_RFC2292)) { 1397 error = EINVAL; 1398 break; 1399 } 1400 #endif 1401 optp = &in6p->in6p_outputopts; 1402 error = ip6_pcbopt(IPV6_HOPLIMIT, 1403 (u_char *)&optval, 1404 sizeof(optval), 1405 optp, 1406 kauth_cred_get(), uproto); 1407 break; 1408 } 1409 1410 case IPV6_RECVHOPLIMIT: 1411 #ifdef RFC2292 1412 /* cannot mix with RFC2292 */ 1413 if (OPTBIT(IN6P_RFC2292)) { 1414 error = EINVAL; 1415 break; 1416 } 1417 #endif 1418 OPTSET(IN6P_HOPLIMIT); 1419 break; 1420 1421 case IPV6_RECVHOPOPTS: 1422 #ifdef RFC2292 1423 /* cannot mix with RFC2292 */ 1424 if (OPTBIT(IN6P_RFC2292)) { 1425 error = EINVAL; 1426 break; 1427 } 1428 #endif 1429 OPTSET(IN6P_HOPOPTS); 1430 break; 1431 1432 case IPV6_RECVDSTOPTS: 1433 #ifdef RFC2292 1434 /* cannot mix with RFC2292 */ 1435 if (OPTBIT(IN6P_RFC2292)) { 1436 error = EINVAL; 1437 break; 1438 } 1439 #endif 1440 OPTSET(IN6P_DSTOPTS); 1441 break; 1442 1443 case IPV6_RECVRTHDRDSTOPTS: 1444 #ifdef RFC2292 1445 /* cannot mix with RFC2292 */ 1446 if (OPTBIT(IN6P_RFC2292)) { 1447 error = EINVAL; 1448 break; 1449 } 1450 #endif 1451 OPTSET(IN6P_RTHDRDSTOPTS); 1452 break; 1453 1454 case IPV6_RECVRTHDR: 1455 #ifdef RFC2292 1456 /* cannot mix with RFC2292 */ 1457 if (OPTBIT(IN6P_RFC2292)) { 1458 error = EINVAL; 1459 break; 1460 } 1461 #endif 1462 OPTSET(IN6P_RTHDR); 1463 break; 1464 1465 case IPV6_FAITH: 1466 OPTSET(IN6P_FAITH); 1467 break; 1468 1469 case IPV6_RECVPATHMTU: 1470 /* 1471 * We ignore this option for TCP 1472 * sockets. 1473 * (RFC3542 leaves this case 1474 * unspecified.) 1475 */ 1476 if (uproto != IPPROTO_TCP) 1477 OPTSET(IN6P_MTU); 1478 break; 1479 1480 case IPV6_V6ONLY: 1481 /* 1482 * make setsockopt(IPV6_V6ONLY) 1483 * available only prior to bind(2). 1484 * see ipng mailing list, Jun 22 2001. 1485 */ 1486 if (in6p->in6p_lport || 1487 !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr)) { 1488 error = EINVAL; 1489 break; 1490 } 1491 #ifdef INET6_BINDV6ONLY 1492 if (!optval) 1493 error = EINVAL; 1494 #else 1495 OPTSET(IN6P_IPV6_V6ONLY); 1496 #endif 1497 break; 1498 case IPV6_RECVTCLASS: 1499 #ifdef RFC2292 1500 /* cannot mix with RFC2292 XXX */ 1501 if (OPTBIT(IN6P_RFC2292)) { 1502 error = EINVAL; 1503 break; 1504 } 1505 #endif 1506 OPTSET(IN6P_TCLASS); 1507 break; 1508 1509 } 1510 break; 1511 1512 case IPV6_OTCLASS: 1513 { 1514 struct ip6_pktopts **optp; 1515 u_int8_t tclass; 1516 1517 error = sockopt_get(sopt, &tclass, sizeof(tclass)); 1518 if (error) 1519 break; 1520 optp = &in6p->in6p_outputopts; 1521 error = ip6_pcbopt(optname, 1522 (u_char *)&tclass, 1523 sizeof(tclass), 1524 optp, 1525 kauth_cred_get(), uproto); 1526 break; 1527 } 1528 1529 case IPV6_TCLASS: 1530 case IPV6_DONTFRAG: 1531 case IPV6_USE_MIN_MTU: 1532 error = sockopt_getint(sopt, &optval); 1533 if (error) 1534 break; 1535 { 1536 struct ip6_pktopts **optp; 1537 optp = &in6p->in6p_outputopts; 1538 error = ip6_pcbopt(optname, 1539 (u_char *)&optval, 1540 sizeof(optval), 1541 optp, 1542 kauth_cred_get(), uproto); 1543 break; 1544 } 1545 1546 #ifdef RFC2292 1547 case IPV6_2292PKTINFO: 1548 case IPV6_2292HOPLIMIT: 1549 case IPV6_2292HOPOPTS: 1550 case IPV6_2292DSTOPTS: 1551 case IPV6_2292RTHDR: 1552 /* RFC 2292 */ 1553 error = sockopt_getint(sopt, &optval); 1554 if (error) 1555 break; 1556 1557 switch (optname) { 1558 case IPV6_2292PKTINFO: 1559 OPTSET2292(IN6P_PKTINFO); 1560 break; 1561 case IPV6_2292HOPLIMIT: 1562 OPTSET2292(IN6P_HOPLIMIT); 1563 break; 1564 case IPV6_2292HOPOPTS: 1565 /* 1566 * Check super-user privilege. 1567 * See comments for IPV6_RECVHOPOPTS. 1568 */ 1569 error = 1570 kauth_authorize_network(kauth_cred_get(), 1571 KAUTH_NETWORK_IPV6, 1572 KAUTH_REQ_NETWORK_IPV6_HOPBYHOP, NULL, 1573 NULL, NULL); 1574 if (error) 1575 return (error); 1576 OPTSET2292(IN6P_HOPOPTS); 1577 break; 1578 case IPV6_2292DSTOPTS: 1579 error = 1580 kauth_authorize_network(kauth_cred_get(), 1581 KAUTH_NETWORK_IPV6, 1582 KAUTH_REQ_NETWORK_IPV6_HOPBYHOP, NULL, 1583 NULL, NULL); 1584 if (error) 1585 return (error); 1586 OPTSET2292(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); /* XXX */ 1587 break; 1588 case IPV6_2292RTHDR: 1589 OPTSET2292(IN6P_RTHDR); 1590 break; 1591 } 1592 break; 1593 #endif 1594 case IPV6_PKTINFO: 1595 case IPV6_HOPOPTS: 1596 case IPV6_RTHDR: 1597 case IPV6_DSTOPTS: 1598 case IPV6_RTHDRDSTOPTS: 1599 case IPV6_NEXTHOP: { 1600 /* new advanced API (RFC3542) */ 1601 void *optbuf; 1602 int optbuflen; 1603 struct ip6_pktopts **optp; 1604 1605 #ifdef RFC2292 1606 /* cannot mix with RFC2292 */ 1607 if (OPTBIT(IN6P_RFC2292)) { 1608 error = EINVAL; 1609 break; 1610 } 1611 #endif 1612 1613 optbuflen = sopt->sopt_size; 1614 optbuf = malloc(optbuflen, M_IP6OPT, M_NOWAIT); 1615 if (optbuf == NULL) { 1616 error = ENOBUFS; 1617 break; 1618 } 1619 1620 error = sockopt_get(sopt, optbuf, optbuflen); 1621 if (error) { 1622 free(optbuf, M_IP6OPT); 1623 break; 1624 } 1625 optp = &in6p->in6p_outputopts; 1626 error = ip6_pcbopt(optname, optbuf, optbuflen, 1627 optp, kauth_cred_get(), uproto); 1628 1629 free(optbuf, M_IP6OPT); 1630 break; 1631 } 1632 #undef OPTSET 1633 1634 case IPV6_MULTICAST_IF: 1635 case IPV6_MULTICAST_HOPS: 1636 case IPV6_MULTICAST_LOOP: 1637 case IPV6_JOIN_GROUP: 1638 case IPV6_LEAVE_GROUP: 1639 error = ip6_setmoptions(sopt, &in6p->in6p_moptions); 1640 break; 1641 1642 case IPV6_PORTRANGE: 1643 error = sockopt_getint(sopt, &optval); 1644 if (error) 1645 break; 1646 1647 switch (optval) { 1648 case IPV6_PORTRANGE_DEFAULT: 1649 in6p->in6p_flags &= ~(IN6P_LOWPORT); 1650 in6p->in6p_flags &= ~(IN6P_HIGHPORT); 1651 break; 1652 1653 case IPV6_PORTRANGE_HIGH: 1654 in6p->in6p_flags &= ~(IN6P_LOWPORT); 1655 in6p->in6p_flags |= IN6P_HIGHPORT; 1656 break; 1657 1658 case IPV6_PORTRANGE_LOW: 1659 in6p->in6p_flags &= ~(IN6P_HIGHPORT); 1660 in6p->in6p_flags |= IN6P_LOWPORT; 1661 break; 1662 1663 default: 1664 error = EINVAL; 1665 break; 1666 } 1667 break; 1668 1669 case IPV6_PORTALGO: 1670 error = sockopt_getint(sopt, &optval); 1671 if (error) 1672 break; 1673 1674 error = portalgo_algo_index_select( 1675 (struct inpcb_hdr *)in6p, optval); 1676 break; 1677 1678 #if defined(IPSEC) 1679 case IPV6_IPSEC_POLICY: 1680 if (ipsec_enabled) { 1681 error = ipsec6_set_policy(in6p, optname, 1682 sopt->sopt_data, sopt->sopt_size, 1683 kauth_cred_get()); 1684 break; 1685 } 1686 /*FALLTHROUGH*/ 1687 #endif /* IPSEC */ 1688 1689 default: 1690 error = ENOPROTOOPT; 1691 break; 1692 } 1693 break; 1694 1695 case PRCO_GETOPT: 1696 switch (optname) { 1697 #ifdef RFC2292 1698 case IPV6_2292PKTOPTIONS: 1699 /* 1700 * RFC3542 (effectively) deprecated the 1701 * semantics of the 2292-style pktoptions. 1702 * Since it was not reliable in nature (i.e., 1703 * applications had to expect the lack of some 1704 * information after all), it would make sense 1705 * to simplify this part by always returning 1706 * empty data. 1707 */ 1708 break; 1709 #endif 1710 1711 case IPV6_RECVHOPOPTS: 1712 case IPV6_RECVDSTOPTS: 1713 case IPV6_RECVRTHDRDSTOPTS: 1714 case IPV6_UNICAST_HOPS: 1715 case IPV6_RECVPKTINFO: 1716 case IPV6_RECVHOPLIMIT: 1717 case IPV6_RECVRTHDR: 1718 case IPV6_RECVPATHMTU: 1719 1720 case IPV6_FAITH: 1721 case IPV6_V6ONLY: 1722 case IPV6_PORTRANGE: 1723 case IPV6_RECVTCLASS: 1724 switch (optname) { 1725 1726 case IPV6_RECVHOPOPTS: 1727 optval = OPTBIT(IN6P_HOPOPTS); 1728 break; 1729 1730 case IPV6_RECVDSTOPTS: 1731 optval = OPTBIT(IN6P_DSTOPTS); 1732 break; 1733 1734 case IPV6_RECVRTHDRDSTOPTS: 1735 optval = OPTBIT(IN6P_RTHDRDSTOPTS); 1736 break; 1737 1738 case IPV6_UNICAST_HOPS: 1739 optval = in6p->in6p_hops; 1740 break; 1741 1742 case IPV6_RECVPKTINFO: 1743 optval = OPTBIT(IN6P_PKTINFO); 1744 break; 1745 1746 case IPV6_RECVHOPLIMIT: 1747 optval = OPTBIT(IN6P_HOPLIMIT); 1748 break; 1749 1750 case IPV6_RECVRTHDR: 1751 optval = OPTBIT(IN6P_RTHDR); 1752 break; 1753 1754 case IPV6_RECVPATHMTU: 1755 optval = OPTBIT(IN6P_MTU); 1756 break; 1757 1758 case IPV6_FAITH: 1759 optval = OPTBIT(IN6P_FAITH); 1760 break; 1761 1762 case IPV6_V6ONLY: 1763 optval = OPTBIT(IN6P_IPV6_V6ONLY); 1764 break; 1765 1766 case IPV6_PORTRANGE: 1767 { 1768 int flags; 1769 flags = in6p->in6p_flags; 1770 if (flags & IN6P_HIGHPORT) 1771 optval = IPV6_PORTRANGE_HIGH; 1772 else if (flags & IN6P_LOWPORT) 1773 optval = IPV6_PORTRANGE_LOW; 1774 else 1775 optval = 0; 1776 break; 1777 } 1778 case IPV6_RECVTCLASS: 1779 optval = OPTBIT(IN6P_TCLASS); 1780 break; 1781 1782 } 1783 if (error) 1784 break; 1785 error = sockopt_setint(sopt, optval); 1786 break; 1787 1788 case IPV6_PATHMTU: 1789 { 1790 u_long pmtu = 0; 1791 struct ip6_mtuinfo mtuinfo; 1792 struct route *ro = &in6p->in6p_route; 1793 1794 if (!(so->so_state & SS_ISCONNECTED)) 1795 return (ENOTCONN); 1796 /* 1797 * XXX: we dot not consider the case of source 1798 * routing, or optional information to specify 1799 * the outgoing interface. 1800 */ 1801 error = ip6_getpmtu(ro, NULL, NULL, 1802 &in6p->in6p_faddr, &pmtu, NULL); 1803 if (error) 1804 break; 1805 if (pmtu > IPV6_MAXPACKET) 1806 pmtu = IPV6_MAXPACKET; 1807 1808 memset(&mtuinfo, 0, sizeof(mtuinfo)); 1809 mtuinfo.ip6m_mtu = (u_int32_t)pmtu; 1810 optdata = (void *)&mtuinfo; 1811 optdatalen = sizeof(mtuinfo); 1812 if (optdatalen > MCLBYTES) 1813 return (EMSGSIZE); /* XXX */ 1814 error = sockopt_set(sopt, optdata, optdatalen); 1815 break; 1816 } 1817 1818 #ifdef RFC2292 1819 case IPV6_2292PKTINFO: 1820 case IPV6_2292HOPLIMIT: 1821 case IPV6_2292HOPOPTS: 1822 case IPV6_2292RTHDR: 1823 case IPV6_2292DSTOPTS: 1824 switch (optname) { 1825 case IPV6_2292PKTINFO: 1826 optval = OPTBIT(IN6P_PKTINFO); 1827 break; 1828 case IPV6_2292HOPLIMIT: 1829 optval = OPTBIT(IN6P_HOPLIMIT); 1830 break; 1831 case IPV6_2292HOPOPTS: 1832 optval = OPTBIT(IN6P_HOPOPTS); 1833 break; 1834 case IPV6_2292RTHDR: 1835 optval = OPTBIT(IN6P_RTHDR); 1836 break; 1837 case IPV6_2292DSTOPTS: 1838 optval = OPTBIT(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); 1839 break; 1840 } 1841 error = sockopt_setint(sopt, optval); 1842 break; 1843 #endif 1844 case IPV6_PKTINFO: 1845 case IPV6_HOPOPTS: 1846 case IPV6_RTHDR: 1847 case IPV6_DSTOPTS: 1848 case IPV6_RTHDRDSTOPTS: 1849 case IPV6_NEXTHOP: 1850 case IPV6_OTCLASS: 1851 case IPV6_TCLASS: 1852 case IPV6_DONTFRAG: 1853 case IPV6_USE_MIN_MTU: 1854 error = ip6_getpcbopt(in6p->in6p_outputopts, 1855 optname, sopt); 1856 break; 1857 1858 case IPV6_MULTICAST_IF: 1859 case IPV6_MULTICAST_HOPS: 1860 case IPV6_MULTICAST_LOOP: 1861 case IPV6_JOIN_GROUP: 1862 case IPV6_LEAVE_GROUP: 1863 error = ip6_getmoptions(sopt, in6p->in6p_moptions); 1864 break; 1865 1866 case IPV6_PORTALGO: 1867 optval = ((struct inpcb_hdr *)in6p)->inph_portalgo; 1868 error = sockopt_setint(sopt, optval); 1869 break; 1870 1871 #if defined(IPSEC) 1872 case IPV6_IPSEC_POLICY: 1873 if (ipsec_used) { 1874 struct mbuf *m = NULL; 1875 1876 /* 1877 * XXX: this will return EINVAL as sopt is 1878 * empty 1879 */ 1880 error = ipsec6_get_policy(in6p, sopt->sopt_data, 1881 sopt->sopt_size, &m); 1882 if (!error) 1883 error = sockopt_setmbuf(sopt, m); 1884 break; 1885 } 1886 /*FALLTHROUGH*/ 1887 #endif /* IPSEC */ 1888 1889 default: 1890 error = ENOPROTOOPT; 1891 break; 1892 } 1893 break; 1894 } 1895 return (error); 1896 } 1897 1898 int 1899 ip6_raw_ctloutput(int op, struct socket *so, struct sockopt *sopt) 1900 { 1901 int error = 0, optval; 1902 const int icmp6off = offsetof(struct icmp6_hdr, icmp6_cksum); 1903 struct in6pcb *in6p = sotoin6pcb(so); 1904 int level, optname; 1905 1906 KASSERT(sopt != NULL); 1907 1908 level = sopt->sopt_level; 1909 optname = sopt->sopt_name; 1910 1911 if (level != IPPROTO_IPV6) { 1912 return ENOPROTOOPT; 1913 } 1914 1915 switch (optname) { 1916 case IPV6_CHECKSUM: 1917 /* 1918 * For ICMPv6 sockets, no modification allowed for checksum 1919 * offset, permit "no change" values to help existing apps. 1920 * 1921 * XXX RFC3542 says: "An attempt to set IPV6_CHECKSUM 1922 * for an ICMPv6 socket will fail." The current 1923 * behavior does not meet RFC3542. 1924 */ 1925 switch (op) { 1926 case PRCO_SETOPT: 1927 error = sockopt_getint(sopt, &optval); 1928 if (error) 1929 break; 1930 if ((optval % 2) != 0) { 1931 /* the API assumes even offset values */ 1932 error = EINVAL; 1933 } else if (so->so_proto->pr_protocol == 1934 IPPROTO_ICMPV6) { 1935 if (optval != icmp6off) 1936 error = EINVAL; 1937 } else 1938 in6p->in6p_cksum = optval; 1939 break; 1940 1941 case PRCO_GETOPT: 1942 if (so->so_proto->pr_protocol == IPPROTO_ICMPV6) 1943 optval = icmp6off; 1944 else 1945 optval = in6p->in6p_cksum; 1946 1947 error = sockopt_setint(sopt, optval); 1948 break; 1949 1950 default: 1951 error = EINVAL; 1952 break; 1953 } 1954 break; 1955 1956 default: 1957 error = ENOPROTOOPT; 1958 break; 1959 } 1960 1961 return (error); 1962 } 1963 1964 #ifdef RFC2292 1965 /* 1966 * Set up IP6 options in pcb for insertion in output packets or 1967 * specifying behavior of outgoing packets. 1968 */ 1969 static int 1970 ip6_pcbopts(struct ip6_pktopts **pktopt, struct socket *so, 1971 struct sockopt *sopt) 1972 { 1973 struct ip6_pktopts *opt = *pktopt; 1974 struct mbuf *m; 1975 int error = 0; 1976 1977 /* turn off any old options. */ 1978 if (opt) { 1979 #ifdef DIAGNOSTIC 1980 if (opt->ip6po_pktinfo || opt->ip6po_nexthop || 1981 opt->ip6po_hbh || opt->ip6po_dest1 || opt->ip6po_dest2 || 1982 opt->ip6po_rhinfo.ip6po_rhi_rthdr) 1983 printf("ip6_pcbopts: all specified options are cleared.\n"); 1984 #endif 1985 ip6_clearpktopts(opt, -1); 1986 } else { 1987 opt = malloc(sizeof(*opt), M_IP6OPT, M_NOWAIT); 1988 if (opt == NULL) 1989 return (ENOBUFS); 1990 } 1991 *pktopt = NULL; 1992 1993 if (sopt == NULL || sopt->sopt_size == 0) { 1994 /* 1995 * Only turning off any previous options, regardless of 1996 * whether the opt is just created or given. 1997 */ 1998 free(opt, M_IP6OPT); 1999 return (0); 2000 } 2001 2002 /* set options specified by user. */ 2003 m = sockopt_getmbuf(sopt); 2004 if (m == NULL) { 2005 free(opt, M_IP6OPT); 2006 return (ENOBUFS); 2007 } 2008 2009 error = ip6_setpktopts(m, opt, NULL, kauth_cred_get(), 2010 so->so_proto->pr_protocol); 2011 m_freem(m); 2012 if (error != 0) { 2013 ip6_clearpktopts(opt, -1); /* XXX: discard all options */ 2014 free(opt, M_IP6OPT); 2015 return (error); 2016 } 2017 *pktopt = opt; 2018 return (0); 2019 } 2020 #endif 2021 2022 /* 2023 * initialize ip6_pktopts. beware that there are non-zero default values in 2024 * the struct. 2025 */ 2026 void 2027 ip6_initpktopts(struct ip6_pktopts *opt) 2028 { 2029 2030 memset(opt, 0, sizeof(*opt)); 2031 opt->ip6po_hlim = -1; /* -1 means default hop limit */ 2032 opt->ip6po_tclass = -1; /* -1 means default traffic class */ 2033 opt->ip6po_minmtu = IP6PO_MINMTU_MCASTONLY; 2034 } 2035 2036 #define sin6tosa(sin6) ((struct sockaddr *)(sin6)) /* XXX */ 2037 static int 2038 ip6_pcbopt(int optname, u_char *buf, int len, struct ip6_pktopts **pktopt, 2039 kauth_cred_t cred, int uproto) 2040 { 2041 struct ip6_pktopts *opt; 2042 2043 if (*pktopt == NULL) { 2044 *pktopt = malloc(sizeof(struct ip6_pktopts), M_IP6OPT, 2045 M_NOWAIT); 2046 if (*pktopt == NULL) 2047 return (ENOBUFS); 2048 2049 ip6_initpktopts(*pktopt); 2050 } 2051 opt = *pktopt; 2052 2053 return (ip6_setpktopt(optname, buf, len, opt, cred, 1, 0, uproto)); 2054 } 2055 2056 static int 2057 ip6_getpcbopt(struct ip6_pktopts *pktopt, int optname, struct sockopt *sopt) 2058 { 2059 void *optdata = NULL; 2060 int optdatalen = 0; 2061 struct ip6_ext *ip6e; 2062 int error = 0; 2063 struct in6_pktinfo null_pktinfo; 2064 int deftclass = 0, on; 2065 int defminmtu = IP6PO_MINMTU_MCASTONLY; 2066 2067 switch (optname) { 2068 case IPV6_PKTINFO: 2069 if (pktopt && pktopt->ip6po_pktinfo) 2070 optdata = (void *)pktopt->ip6po_pktinfo; 2071 else { 2072 /* XXX: we don't have to do this every time... */ 2073 memset(&null_pktinfo, 0, sizeof(null_pktinfo)); 2074 optdata = (void *)&null_pktinfo; 2075 } 2076 optdatalen = sizeof(struct in6_pktinfo); 2077 break; 2078 case IPV6_OTCLASS: 2079 /* XXX */ 2080 return (EINVAL); 2081 case IPV6_TCLASS: 2082 if (pktopt && pktopt->ip6po_tclass >= 0) 2083 optdata = (void *)&pktopt->ip6po_tclass; 2084 else 2085 optdata = (void *)&deftclass; 2086 optdatalen = sizeof(int); 2087 break; 2088 case IPV6_HOPOPTS: 2089 if (pktopt && pktopt->ip6po_hbh) { 2090 optdata = (void *)pktopt->ip6po_hbh; 2091 ip6e = (struct ip6_ext *)pktopt->ip6po_hbh; 2092 optdatalen = (ip6e->ip6e_len + 1) << 3; 2093 } 2094 break; 2095 case IPV6_RTHDR: 2096 if (pktopt && pktopt->ip6po_rthdr) { 2097 optdata = (void *)pktopt->ip6po_rthdr; 2098 ip6e = (struct ip6_ext *)pktopt->ip6po_rthdr; 2099 optdatalen = (ip6e->ip6e_len + 1) << 3; 2100 } 2101 break; 2102 case IPV6_RTHDRDSTOPTS: 2103 if (pktopt && pktopt->ip6po_dest1) { 2104 optdata = (void *)pktopt->ip6po_dest1; 2105 ip6e = (struct ip6_ext *)pktopt->ip6po_dest1; 2106 optdatalen = (ip6e->ip6e_len + 1) << 3; 2107 } 2108 break; 2109 case IPV6_DSTOPTS: 2110 if (pktopt && pktopt->ip6po_dest2) { 2111 optdata = (void *)pktopt->ip6po_dest2; 2112 ip6e = (struct ip6_ext *)pktopt->ip6po_dest2; 2113 optdatalen = (ip6e->ip6e_len + 1) << 3; 2114 } 2115 break; 2116 case IPV6_NEXTHOP: 2117 if (pktopt && pktopt->ip6po_nexthop) { 2118 optdata = (void *)pktopt->ip6po_nexthop; 2119 optdatalen = pktopt->ip6po_nexthop->sa_len; 2120 } 2121 break; 2122 case IPV6_USE_MIN_MTU: 2123 if (pktopt) 2124 optdata = (void *)&pktopt->ip6po_minmtu; 2125 else 2126 optdata = (void *)&defminmtu; 2127 optdatalen = sizeof(int); 2128 break; 2129 case IPV6_DONTFRAG: 2130 if (pktopt && ((pktopt->ip6po_flags) & IP6PO_DONTFRAG)) 2131 on = 1; 2132 else 2133 on = 0; 2134 optdata = (void *)&on; 2135 optdatalen = sizeof(on); 2136 break; 2137 default: /* should not happen */ 2138 #ifdef DIAGNOSTIC 2139 panic("ip6_getpcbopt: unexpected option\n"); 2140 #endif 2141 return (ENOPROTOOPT); 2142 } 2143 2144 error = sockopt_set(sopt, optdata, optdatalen); 2145 2146 return (error); 2147 } 2148 2149 void 2150 ip6_clearpktopts(struct ip6_pktopts *pktopt, int optname) 2151 { 2152 if (optname == -1 || optname == IPV6_PKTINFO) { 2153 if (pktopt->ip6po_pktinfo) 2154 free(pktopt->ip6po_pktinfo, M_IP6OPT); 2155 pktopt->ip6po_pktinfo = NULL; 2156 } 2157 if (optname == -1 || optname == IPV6_HOPLIMIT) 2158 pktopt->ip6po_hlim = -1; 2159 if (optname == -1 || optname == IPV6_TCLASS) 2160 pktopt->ip6po_tclass = -1; 2161 if (optname == -1 || optname == IPV6_NEXTHOP) { 2162 rtcache_free(&pktopt->ip6po_nextroute); 2163 if (pktopt->ip6po_nexthop) 2164 free(pktopt->ip6po_nexthop, M_IP6OPT); 2165 pktopt->ip6po_nexthop = NULL; 2166 } 2167 if (optname == -1 || optname == IPV6_HOPOPTS) { 2168 if (pktopt->ip6po_hbh) 2169 free(pktopt->ip6po_hbh, M_IP6OPT); 2170 pktopt->ip6po_hbh = NULL; 2171 } 2172 if (optname == -1 || optname == IPV6_RTHDRDSTOPTS) { 2173 if (pktopt->ip6po_dest1) 2174 free(pktopt->ip6po_dest1, M_IP6OPT); 2175 pktopt->ip6po_dest1 = NULL; 2176 } 2177 if (optname == -1 || optname == IPV6_RTHDR) { 2178 if (pktopt->ip6po_rhinfo.ip6po_rhi_rthdr) 2179 free(pktopt->ip6po_rhinfo.ip6po_rhi_rthdr, M_IP6OPT); 2180 pktopt->ip6po_rhinfo.ip6po_rhi_rthdr = NULL; 2181 rtcache_free(&pktopt->ip6po_route); 2182 } 2183 if (optname == -1 || optname == IPV6_DSTOPTS) { 2184 if (pktopt->ip6po_dest2) 2185 free(pktopt->ip6po_dest2, M_IP6OPT); 2186 pktopt->ip6po_dest2 = NULL; 2187 } 2188 } 2189 2190 #define PKTOPT_EXTHDRCPY(type) \ 2191 do { \ 2192 if (src->type) { \ 2193 int hlen = (((struct ip6_ext *)src->type)->ip6e_len + 1) << 3;\ 2194 dst->type = malloc(hlen, M_IP6OPT, canwait); \ 2195 if (dst->type == NULL) \ 2196 goto bad; \ 2197 memcpy(dst->type, src->type, hlen); \ 2198 } \ 2199 } while (/*CONSTCOND*/ 0) 2200 2201 static int 2202 copypktopts(struct ip6_pktopts *dst, struct ip6_pktopts *src, int canwait) 2203 { 2204 dst->ip6po_hlim = src->ip6po_hlim; 2205 dst->ip6po_tclass = src->ip6po_tclass; 2206 dst->ip6po_flags = src->ip6po_flags; 2207 if (src->ip6po_pktinfo) { 2208 dst->ip6po_pktinfo = malloc(sizeof(*dst->ip6po_pktinfo), 2209 M_IP6OPT, canwait); 2210 if (dst->ip6po_pktinfo == NULL) 2211 goto bad; 2212 *dst->ip6po_pktinfo = *src->ip6po_pktinfo; 2213 } 2214 if (src->ip6po_nexthop) { 2215 dst->ip6po_nexthop = malloc(src->ip6po_nexthop->sa_len, 2216 M_IP6OPT, canwait); 2217 if (dst->ip6po_nexthop == NULL) 2218 goto bad; 2219 memcpy(dst->ip6po_nexthop, src->ip6po_nexthop, 2220 src->ip6po_nexthop->sa_len); 2221 } 2222 PKTOPT_EXTHDRCPY(ip6po_hbh); 2223 PKTOPT_EXTHDRCPY(ip6po_dest1); 2224 PKTOPT_EXTHDRCPY(ip6po_dest2); 2225 PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */ 2226 return (0); 2227 2228 bad: 2229 if (dst->ip6po_pktinfo) free(dst->ip6po_pktinfo, M_IP6OPT); 2230 if (dst->ip6po_nexthop) free(dst->ip6po_nexthop, M_IP6OPT); 2231 if (dst->ip6po_hbh) free(dst->ip6po_hbh, M_IP6OPT); 2232 if (dst->ip6po_dest1) free(dst->ip6po_dest1, M_IP6OPT); 2233 if (dst->ip6po_dest2) free(dst->ip6po_dest2, M_IP6OPT); 2234 if (dst->ip6po_rthdr) free(dst->ip6po_rthdr, M_IP6OPT); 2235 2236 return (ENOBUFS); 2237 } 2238 #undef PKTOPT_EXTHDRCPY 2239 2240 struct ip6_pktopts * 2241 ip6_copypktopts(struct ip6_pktopts *src, int canwait) 2242 { 2243 int error; 2244 struct ip6_pktopts *dst; 2245 2246 dst = malloc(sizeof(*dst), M_IP6OPT, canwait); 2247 if (dst == NULL) 2248 return (NULL); 2249 ip6_initpktopts(dst); 2250 2251 if ((error = copypktopts(dst, src, canwait)) != 0) { 2252 free(dst, M_IP6OPT); 2253 return (NULL); 2254 } 2255 2256 return (dst); 2257 } 2258 2259 void 2260 ip6_freepcbopts(struct ip6_pktopts *pktopt) 2261 { 2262 if (pktopt == NULL) 2263 return; 2264 2265 ip6_clearpktopts(pktopt, -1); 2266 2267 free(pktopt, M_IP6OPT); 2268 } 2269 2270 /* 2271 * Set the IP6 multicast options in response to user setsockopt(). 2272 */ 2273 static int 2274 ip6_setmoptions(const struct sockopt *sopt, struct ip6_moptions **im6op) 2275 { 2276 int error = 0; 2277 u_int loop, ifindex; 2278 struct ipv6_mreq mreq; 2279 struct ifnet *ifp; 2280 struct ip6_moptions *im6o = *im6op; 2281 struct route ro; 2282 struct in6_multi_mship *imm; 2283 struct lwp *l = curlwp; /* XXX */ 2284 2285 if (im6o == NULL) { 2286 /* 2287 * No multicast option buffer attached to the pcb; 2288 * allocate one and initialize to default values. 2289 */ 2290 im6o = malloc(sizeof(*im6o), M_IPMOPTS, M_NOWAIT); 2291 if (im6o == NULL) 2292 return (ENOBUFS); 2293 2294 *im6op = im6o; 2295 im6o->im6o_multicast_ifp = NULL; 2296 im6o->im6o_multicast_hlim = ip6_defmcasthlim; 2297 im6o->im6o_multicast_loop = IPV6_DEFAULT_MULTICAST_LOOP; 2298 LIST_INIT(&im6o->im6o_memberships); 2299 } 2300 2301 switch (sopt->sopt_name) { 2302 2303 case IPV6_MULTICAST_IF: 2304 /* 2305 * Select the interface for outgoing multicast packets. 2306 */ 2307 error = sockopt_get(sopt, &ifindex, sizeof(ifindex)); 2308 if (error != 0) 2309 break; 2310 2311 if (ifindex != 0) { 2312 if ((ifp = if_byindex(ifindex)) == NULL) { 2313 error = ENXIO; /* XXX EINVAL? */ 2314 break; 2315 } 2316 if ((ifp->if_flags & IFF_MULTICAST) == 0) { 2317 error = EADDRNOTAVAIL; 2318 break; 2319 } 2320 } else 2321 ifp = NULL; 2322 im6o->im6o_multicast_ifp = ifp; 2323 break; 2324 2325 case IPV6_MULTICAST_HOPS: 2326 { 2327 /* 2328 * Set the IP6 hoplimit for outgoing multicast packets. 2329 */ 2330 int optval; 2331 2332 error = sockopt_getint(sopt, &optval); 2333 if (error != 0) 2334 break; 2335 2336 if (optval < -1 || optval >= 256) 2337 error = EINVAL; 2338 else if (optval == -1) 2339 im6o->im6o_multicast_hlim = ip6_defmcasthlim; 2340 else 2341 im6o->im6o_multicast_hlim = optval; 2342 break; 2343 } 2344 2345 case IPV6_MULTICAST_LOOP: 2346 /* 2347 * Set the loopback flag for outgoing multicast packets. 2348 * Must be zero or one. 2349 */ 2350 error = sockopt_get(sopt, &loop, sizeof(loop)); 2351 if (error != 0) 2352 break; 2353 if (loop > 1) { 2354 error = EINVAL; 2355 break; 2356 } 2357 im6o->im6o_multicast_loop = loop; 2358 break; 2359 2360 case IPV6_JOIN_GROUP: 2361 /* 2362 * Add a multicast group membership. 2363 * Group must be a valid IP6 multicast address. 2364 */ 2365 error = sockopt_get(sopt, &mreq, sizeof(mreq)); 2366 if (error != 0) 2367 break; 2368 2369 if (IN6_IS_ADDR_UNSPECIFIED(&mreq.ipv6mr_multiaddr)) { 2370 /* 2371 * We use the unspecified address to specify to accept 2372 * all multicast addresses. Only super user is allowed 2373 * to do this. 2374 */ 2375 if (kauth_authorize_network(l->l_cred, KAUTH_NETWORK_IPV6, 2376 KAUTH_REQ_NETWORK_IPV6_JOIN_MULTICAST, NULL, NULL, NULL)) 2377 { 2378 error = EACCES; 2379 break; 2380 } 2381 } else if (!IN6_IS_ADDR_MULTICAST(&mreq.ipv6mr_multiaddr)) { 2382 error = EINVAL; 2383 break; 2384 } 2385 2386 /* 2387 * If no interface was explicitly specified, choose an 2388 * appropriate one according to the given multicast address. 2389 */ 2390 if (mreq.ipv6mr_interface == 0) { 2391 struct rtentry *rt; 2392 union { 2393 struct sockaddr dst; 2394 struct sockaddr_in6 dst6; 2395 } u; 2396 2397 /* 2398 * Look up the routing table for the 2399 * address, and choose the outgoing interface. 2400 * XXX: is it a good approach? 2401 */ 2402 memset(&ro, 0, sizeof(ro)); 2403 sockaddr_in6_init(&u.dst6, &mreq.ipv6mr_multiaddr, 0, 2404 0, 0); 2405 rtcache_setdst(&ro, &u.dst); 2406 ifp = (rt = rtcache_init(&ro)) != NULL ? rt->rt_ifp 2407 : NULL; 2408 rtcache_free(&ro); 2409 } else { 2410 /* 2411 * If the interface is specified, validate it. 2412 */ 2413 if ((ifp = if_byindex(mreq.ipv6mr_interface)) == NULL) { 2414 error = ENXIO; /* XXX EINVAL? */ 2415 break; 2416 } 2417 } 2418 2419 /* 2420 * See if we found an interface, and confirm that it 2421 * supports multicast 2422 */ 2423 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) { 2424 error = EADDRNOTAVAIL; 2425 break; 2426 } 2427 2428 if (in6_setscope(&mreq.ipv6mr_multiaddr, ifp, NULL)) { 2429 error = EADDRNOTAVAIL; /* XXX: should not happen */ 2430 break; 2431 } 2432 2433 /* 2434 * See if the membership already exists. 2435 */ 2436 for (imm = im6o->im6o_memberships.lh_first; 2437 imm != NULL; imm = imm->i6mm_chain.le_next) 2438 if (imm->i6mm_maddr->in6m_ifp == ifp && 2439 IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr, 2440 &mreq.ipv6mr_multiaddr)) 2441 break; 2442 if (imm != NULL) { 2443 error = EADDRINUSE; 2444 break; 2445 } 2446 /* 2447 * Everything looks good; add a new record to the multicast 2448 * address list for the given interface. 2449 */ 2450 imm = in6_joingroup(ifp, &mreq.ipv6mr_multiaddr, &error, 0); 2451 if (imm == NULL) 2452 break; 2453 LIST_INSERT_HEAD(&im6o->im6o_memberships, imm, i6mm_chain); 2454 break; 2455 2456 case IPV6_LEAVE_GROUP: 2457 /* 2458 * Drop a multicast group membership. 2459 * Group must be a valid IP6 multicast address. 2460 */ 2461 error = sockopt_get(sopt, &mreq, sizeof(mreq)); 2462 if (error != 0) 2463 break; 2464 2465 /* 2466 * If an interface address was specified, get a pointer 2467 * to its ifnet structure. 2468 */ 2469 if (mreq.ipv6mr_interface != 0) { 2470 if ((ifp = if_byindex(mreq.ipv6mr_interface)) == NULL) { 2471 error = ENXIO; /* XXX EINVAL? */ 2472 break; 2473 } 2474 } else 2475 ifp = NULL; 2476 2477 /* Fill in the scope zone ID */ 2478 if (ifp) { 2479 if (in6_setscope(&mreq.ipv6mr_multiaddr, ifp, NULL)) { 2480 /* XXX: should not happen */ 2481 error = EADDRNOTAVAIL; 2482 break; 2483 } 2484 } else if (mreq.ipv6mr_interface != 0) { 2485 /* 2486 * XXX: This case would happens when the (positive) 2487 * index is in the valid range, but the corresponding 2488 * interface has been detached dynamically. The above 2489 * check probably avoids such case to happen here, but 2490 * we check it explicitly for safety. 2491 */ 2492 error = EADDRNOTAVAIL; 2493 break; 2494 } else { /* ipv6mr_interface == 0 */ 2495 struct sockaddr_in6 sa6_mc; 2496 2497 /* 2498 * The API spec says as follows: 2499 * If the interface index is specified as 0, the 2500 * system may choose a multicast group membership to 2501 * drop by matching the multicast address only. 2502 * On the other hand, we cannot disambiguate the scope 2503 * zone unless an interface is provided. Thus, we 2504 * check if there's ambiguity with the default scope 2505 * zone as the last resort. 2506 */ 2507 sockaddr_in6_init(&sa6_mc, &mreq.ipv6mr_multiaddr, 2508 0, 0, 0); 2509 error = sa6_embedscope(&sa6_mc, ip6_use_defzone); 2510 if (error != 0) 2511 break; 2512 mreq.ipv6mr_multiaddr = sa6_mc.sin6_addr; 2513 } 2514 2515 /* 2516 * Find the membership in the membership list. 2517 */ 2518 for (imm = im6o->im6o_memberships.lh_first; 2519 imm != NULL; imm = imm->i6mm_chain.le_next) { 2520 if ((ifp == NULL || imm->i6mm_maddr->in6m_ifp == ifp) && 2521 IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr, 2522 &mreq.ipv6mr_multiaddr)) 2523 break; 2524 } 2525 if (imm == NULL) { 2526 /* Unable to resolve interface */ 2527 error = EADDRNOTAVAIL; 2528 break; 2529 } 2530 /* 2531 * Give up the multicast address record to which the 2532 * membership points. 2533 */ 2534 LIST_REMOVE(imm, i6mm_chain); 2535 in6_leavegroup(imm); 2536 break; 2537 2538 default: 2539 error = EOPNOTSUPP; 2540 break; 2541 } 2542 2543 /* 2544 * If all options have default values, no need to keep the mbuf. 2545 */ 2546 if (im6o->im6o_multicast_ifp == NULL && 2547 im6o->im6o_multicast_hlim == ip6_defmcasthlim && 2548 im6o->im6o_multicast_loop == IPV6_DEFAULT_MULTICAST_LOOP && 2549 im6o->im6o_memberships.lh_first == NULL) { 2550 free(*im6op, M_IPMOPTS); 2551 *im6op = NULL; 2552 } 2553 2554 return (error); 2555 } 2556 2557 /* 2558 * Return the IP6 multicast options in response to user getsockopt(). 2559 */ 2560 static int 2561 ip6_getmoptions(struct sockopt *sopt, struct ip6_moptions *im6o) 2562 { 2563 u_int optval; 2564 int error; 2565 2566 switch (sopt->sopt_name) { 2567 case IPV6_MULTICAST_IF: 2568 if (im6o == NULL || im6o->im6o_multicast_ifp == NULL) 2569 optval = 0; 2570 else 2571 optval = im6o->im6o_multicast_ifp->if_index; 2572 2573 error = sockopt_set(sopt, &optval, sizeof(optval)); 2574 break; 2575 2576 case IPV6_MULTICAST_HOPS: 2577 if (im6o == NULL) 2578 optval = ip6_defmcasthlim; 2579 else 2580 optval = im6o->im6o_multicast_hlim; 2581 2582 error = sockopt_set(sopt, &optval, sizeof(optval)); 2583 break; 2584 2585 case IPV6_MULTICAST_LOOP: 2586 if (im6o == NULL) 2587 optval = IPV6_DEFAULT_MULTICAST_LOOP; 2588 else 2589 optval = im6o->im6o_multicast_loop; 2590 2591 error = sockopt_set(sopt, &optval, sizeof(optval)); 2592 break; 2593 2594 default: 2595 error = EOPNOTSUPP; 2596 } 2597 2598 return (error); 2599 } 2600 2601 /* 2602 * Discard the IP6 multicast options. 2603 */ 2604 void 2605 ip6_freemoptions(struct ip6_moptions *im6o) 2606 { 2607 struct in6_multi_mship *imm; 2608 2609 if (im6o == NULL) 2610 return; 2611 2612 while ((imm = im6o->im6o_memberships.lh_first) != NULL) { 2613 LIST_REMOVE(imm, i6mm_chain); 2614 in6_leavegroup(imm); 2615 } 2616 free(im6o, M_IPMOPTS); 2617 } 2618 2619 /* 2620 * Set IPv6 outgoing packet options based on advanced API. 2621 */ 2622 int 2623 ip6_setpktopts(struct mbuf *control, struct ip6_pktopts *opt, 2624 struct ip6_pktopts *stickyopt, kauth_cred_t cred, int uproto) 2625 { 2626 struct cmsghdr *cm = 0; 2627 2628 if (control == NULL || opt == NULL) 2629 return (EINVAL); 2630 2631 ip6_initpktopts(opt); 2632 if (stickyopt) { 2633 int error; 2634 2635 /* 2636 * If stickyopt is provided, make a local copy of the options 2637 * for this particular packet, then override them by ancillary 2638 * objects. 2639 * XXX: copypktopts() does not copy the cached route to a next 2640 * hop (if any). This is not very good in terms of efficiency, 2641 * but we can allow this since this option should be rarely 2642 * used. 2643 */ 2644 if ((error = copypktopts(opt, stickyopt, M_NOWAIT)) != 0) 2645 return (error); 2646 } 2647 2648 /* 2649 * XXX: Currently, we assume all the optional information is stored 2650 * in a single mbuf. 2651 */ 2652 if (control->m_next) 2653 return (EINVAL); 2654 2655 /* XXX if cm->cmsg_len is not aligned, control->m_len can become <0 */ 2656 for (; control->m_len > 0; control->m_data += CMSG_ALIGN(cm->cmsg_len), 2657 control->m_len -= CMSG_ALIGN(cm->cmsg_len)) { 2658 int error; 2659 2660 if (control->m_len < CMSG_LEN(0)) 2661 return (EINVAL); 2662 2663 cm = mtod(control, struct cmsghdr *); 2664 if (cm->cmsg_len == 0 || cm->cmsg_len > control->m_len) 2665 return (EINVAL); 2666 if (cm->cmsg_level != IPPROTO_IPV6) 2667 continue; 2668 2669 error = ip6_setpktopt(cm->cmsg_type, CMSG_DATA(cm), 2670 cm->cmsg_len - CMSG_LEN(0), opt, cred, 0, 1, uproto); 2671 if (error) 2672 return (error); 2673 } 2674 2675 return (0); 2676 } 2677 2678 /* 2679 * Set a particular packet option, as a sticky option or an ancillary data 2680 * item. "len" can be 0 only when it's a sticky option. 2681 * We have 4 cases of combination of "sticky" and "cmsg": 2682 * "sticky=0, cmsg=0": impossible 2683 * "sticky=0, cmsg=1": RFC2292 or RFC3542 ancillary data 2684 * "sticky=1, cmsg=0": RFC3542 socket option 2685 * "sticky=1, cmsg=1": RFC2292 socket option 2686 */ 2687 static int 2688 ip6_setpktopt(int optname, u_char *buf, int len, struct ip6_pktopts *opt, 2689 kauth_cred_t cred, int sticky, int cmsg, int uproto) 2690 { 2691 int minmtupolicy; 2692 int error; 2693 2694 if (!sticky && !cmsg) { 2695 #ifdef DIAGNOSTIC 2696 printf("ip6_setpktopt: impossible case\n"); 2697 #endif 2698 return (EINVAL); 2699 } 2700 2701 /* 2702 * IPV6_2292xxx is for backward compatibility to RFC2292, and should 2703 * not be specified in the context of RFC3542. Conversely, 2704 * RFC3542 types should not be specified in the context of RFC2292. 2705 */ 2706 if (!cmsg) { 2707 switch (optname) { 2708 case IPV6_2292PKTINFO: 2709 case IPV6_2292HOPLIMIT: 2710 case IPV6_2292NEXTHOP: 2711 case IPV6_2292HOPOPTS: 2712 case IPV6_2292DSTOPTS: 2713 case IPV6_2292RTHDR: 2714 case IPV6_2292PKTOPTIONS: 2715 return (ENOPROTOOPT); 2716 } 2717 } 2718 if (sticky && cmsg) { 2719 switch (optname) { 2720 case IPV6_PKTINFO: 2721 case IPV6_HOPLIMIT: 2722 case IPV6_NEXTHOP: 2723 case IPV6_HOPOPTS: 2724 case IPV6_DSTOPTS: 2725 case IPV6_RTHDRDSTOPTS: 2726 case IPV6_RTHDR: 2727 case IPV6_USE_MIN_MTU: 2728 case IPV6_DONTFRAG: 2729 case IPV6_OTCLASS: 2730 case IPV6_TCLASS: 2731 return (ENOPROTOOPT); 2732 } 2733 } 2734 2735 switch (optname) { 2736 #ifdef RFC2292 2737 case IPV6_2292PKTINFO: 2738 #endif 2739 case IPV6_PKTINFO: 2740 { 2741 struct ifnet *ifp = NULL; 2742 struct in6_pktinfo *pktinfo; 2743 2744 if (len != sizeof(struct in6_pktinfo)) 2745 return (EINVAL); 2746 2747 pktinfo = (struct in6_pktinfo *)buf; 2748 2749 /* 2750 * An application can clear any sticky IPV6_PKTINFO option by 2751 * doing a "regular" setsockopt with ipi6_addr being 2752 * in6addr_any and ipi6_ifindex being zero. 2753 * [RFC 3542, Section 6] 2754 */ 2755 if (optname == IPV6_PKTINFO && opt->ip6po_pktinfo && 2756 pktinfo->ipi6_ifindex == 0 && 2757 IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) { 2758 ip6_clearpktopts(opt, optname); 2759 break; 2760 } 2761 2762 if (uproto == IPPROTO_TCP && optname == IPV6_PKTINFO && 2763 sticky && !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) { 2764 return (EINVAL); 2765 } 2766 2767 /* Validate the interface index if specified. */ 2768 if (pktinfo->ipi6_ifindex) { 2769 ifp = if_byindex(pktinfo->ipi6_ifindex); 2770 if (ifp == NULL) 2771 return (ENXIO); 2772 } 2773 2774 /* 2775 * We store the address anyway, and let in6_selectsrc() 2776 * validate the specified address. This is because ipi6_addr 2777 * may not have enough information about its scope zone, and 2778 * we may need additional information (such as outgoing 2779 * interface or the scope zone of a destination address) to 2780 * disambiguate the scope. 2781 * XXX: the delay of the validation may confuse the 2782 * application when it is used as a sticky option. 2783 */ 2784 if (opt->ip6po_pktinfo == NULL) { 2785 opt->ip6po_pktinfo = malloc(sizeof(*pktinfo), 2786 M_IP6OPT, M_NOWAIT); 2787 if (opt->ip6po_pktinfo == NULL) 2788 return (ENOBUFS); 2789 } 2790 memcpy(opt->ip6po_pktinfo, pktinfo, sizeof(*pktinfo)); 2791 break; 2792 } 2793 2794 #ifdef RFC2292 2795 case IPV6_2292HOPLIMIT: 2796 #endif 2797 case IPV6_HOPLIMIT: 2798 { 2799 int *hlimp; 2800 2801 /* 2802 * RFC 3542 deprecated the usage of sticky IPV6_HOPLIMIT 2803 * to simplify the ordering among hoplimit options. 2804 */ 2805 if (optname == IPV6_HOPLIMIT && sticky) 2806 return (ENOPROTOOPT); 2807 2808 if (len != sizeof(int)) 2809 return (EINVAL); 2810 hlimp = (int *)buf; 2811 if (*hlimp < -1 || *hlimp > 255) 2812 return (EINVAL); 2813 2814 opt->ip6po_hlim = *hlimp; 2815 break; 2816 } 2817 2818 case IPV6_OTCLASS: 2819 if (len != sizeof(u_int8_t)) 2820 return (EINVAL); 2821 2822 opt->ip6po_tclass = *(u_int8_t *)buf; 2823 break; 2824 2825 case IPV6_TCLASS: 2826 { 2827 int tclass; 2828 2829 if (len != sizeof(int)) 2830 return (EINVAL); 2831 tclass = *(int *)buf; 2832 if (tclass < -1 || tclass > 255) 2833 return (EINVAL); 2834 2835 opt->ip6po_tclass = tclass; 2836 break; 2837 } 2838 2839 #ifdef RFC2292 2840 case IPV6_2292NEXTHOP: 2841 #endif 2842 case IPV6_NEXTHOP: 2843 error = kauth_authorize_network(cred, KAUTH_NETWORK_IPV6, 2844 KAUTH_REQ_NETWORK_IPV6_HOPBYHOP, NULL, NULL, NULL); 2845 if (error) 2846 return (error); 2847 2848 if (len == 0) { /* just remove the option */ 2849 ip6_clearpktopts(opt, IPV6_NEXTHOP); 2850 break; 2851 } 2852 2853 /* check if cmsg_len is large enough for sa_len */ 2854 if (len < sizeof(struct sockaddr) || len < *buf) 2855 return (EINVAL); 2856 2857 switch (((struct sockaddr *)buf)->sa_family) { 2858 case AF_INET6: 2859 { 2860 struct sockaddr_in6 *sa6 = (struct sockaddr_in6 *)buf; 2861 2862 if (sa6->sin6_len != sizeof(struct sockaddr_in6)) 2863 return (EINVAL); 2864 2865 if (IN6_IS_ADDR_UNSPECIFIED(&sa6->sin6_addr) || 2866 IN6_IS_ADDR_MULTICAST(&sa6->sin6_addr)) { 2867 return (EINVAL); 2868 } 2869 if ((error = sa6_embedscope(sa6, ip6_use_defzone)) 2870 != 0) { 2871 return (error); 2872 } 2873 break; 2874 } 2875 case AF_LINK: /* eventually be supported? */ 2876 default: 2877 return (EAFNOSUPPORT); 2878 } 2879 2880 /* turn off the previous option, then set the new option. */ 2881 ip6_clearpktopts(opt, IPV6_NEXTHOP); 2882 opt->ip6po_nexthop = malloc(*buf, M_IP6OPT, M_NOWAIT); 2883 if (opt->ip6po_nexthop == NULL) 2884 return (ENOBUFS); 2885 memcpy(opt->ip6po_nexthop, buf, *buf); 2886 break; 2887 2888 #ifdef RFC2292 2889 case IPV6_2292HOPOPTS: 2890 #endif 2891 case IPV6_HOPOPTS: 2892 { 2893 struct ip6_hbh *hbh; 2894 int hbhlen; 2895 2896 /* 2897 * XXX: We don't allow a non-privileged user to set ANY HbH 2898 * options, since per-option restriction has too much 2899 * overhead. 2900 */ 2901 error = kauth_authorize_network(cred, KAUTH_NETWORK_IPV6, 2902 KAUTH_REQ_NETWORK_IPV6_HOPBYHOP, NULL, NULL, NULL); 2903 if (error) 2904 return (error); 2905 2906 if (len == 0) { 2907 ip6_clearpktopts(opt, IPV6_HOPOPTS); 2908 break; /* just remove the option */ 2909 } 2910 2911 /* message length validation */ 2912 if (len < sizeof(struct ip6_hbh)) 2913 return (EINVAL); 2914 hbh = (struct ip6_hbh *)buf; 2915 hbhlen = (hbh->ip6h_len + 1) << 3; 2916 if (len != hbhlen) 2917 return (EINVAL); 2918 2919 /* turn off the previous option, then set the new option. */ 2920 ip6_clearpktopts(opt, IPV6_HOPOPTS); 2921 opt->ip6po_hbh = malloc(hbhlen, M_IP6OPT, M_NOWAIT); 2922 if (opt->ip6po_hbh == NULL) 2923 return (ENOBUFS); 2924 memcpy(opt->ip6po_hbh, hbh, hbhlen); 2925 2926 break; 2927 } 2928 2929 #ifdef RFC2292 2930 case IPV6_2292DSTOPTS: 2931 #endif 2932 case IPV6_DSTOPTS: 2933 case IPV6_RTHDRDSTOPTS: 2934 { 2935 struct ip6_dest *dest, **newdest = NULL; 2936 int destlen; 2937 2938 /* XXX: see the comment for IPV6_HOPOPTS */ 2939 error = kauth_authorize_network(cred, KAUTH_NETWORK_IPV6, 2940 KAUTH_REQ_NETWORK_IPV6_HOPBYHOP, NULL, NULL, NULL); 2941 if (error) 2942 return (error); 2943 2944 if (len == 0) { 2945 ip6_clearpktopts(opt, optname); 2946 break; /* just remove the option */ 2947 } 2948 2949 /* message length validation */ 2950 if (len < sizeof(struct ip6_dest)) 2951 return (EINVAL); 2952 dest = (struct ip6_dest *)buf; 2953 destlen = (dest->ip6d_len + 1) << 3; 2954 if (len != destlen) 2955 return (EINVAL); 2956 /* 2957 * Determine the position that the destination options header 2958 * should be inserted; before or after the routing header. 2959 */ 2960 switch (optname) { 2961 case IPV6_2292DSTOPTS: 2962 /* 2963 * The old advanced API is ambiguous on this point. 2964 * Our approach is to determine the position based 2965 * according to the existence of a routing header. 2966 * Note, however, that this depends on the order of the 2967 * extension headers in the ancillary data; the 1st 2968 * part of the destination options header must appear 2969 * before the routing header in the ancillary data, 2970 * too. 2971 * RFC3542 solved the ambiguity by introducing 2972 * separate ancillary data or option types. 2973 */ 2974 if (opt->ip6po_rthdr == NULL) 2975 newdest = &opt->ip6po_dest1; 2976 else 2977 newdest = &opt->ip6po_dest2; 2978 break; 2979 case IPV6_RTHDRDSTOPTS: 2980 newdest = &opt->ip6po_dest1; 2981 break; 2982 case IPV6_DSTOPTS: 2983 newdest = &opt->ip6po_dest2; 2984 break; 2985 } 2986 2987 /* turn off the previous option, then set the new option. */ 2988 ip6_clearpktopts(opt, optname); 2989 *newdest = malloc(destlen, M_IP6OPT, M_NOWAIT); 2990 if (*newdest == NULL) 2991 return (ENOBUFS); 2992 memcpy(*newdest, dest, destlen); 2993 2994 break; 2995 } 2996 2997 #ifdef RFC2292 2998 case IPV6_2292RTHDR: 2999 #endif 3000 case IPV6_RTHDR: 3001 { 3002 struct ip6_rthdr *rth; 3003 int rthlen; 3004 3005 if (len == 0) { 3006 ip6_clearpktopts(opt, IPV6_RTHDR); 3007 break; /* just remove the option */ 3008 } 3009 3010 /* message length validation */ 3011 if (len < sizeof(struct ip6_rthdr)) 3012 return (EINVAL); 3013 rth = (struct ip6_rthdr *)buf; 3014 rthlen = (rth->ip6r_len + 1) << 3; 3015 if (len != rthlen) 3016 return (EINVAL); 3017 switch (rth->ip6r_type) { 3018 case IPV6_RTHDR_TYPE_0: 3019 if (rth->ip6r_len == 0) /* must contain one addr */ 3020 return (EINVAL); 3021 if (rth->ip6r_len % 2) /* length must be even */ 3022 return (EINVAL); 3023 if (rth->ip6r_len / 2 != rth->ip6r_segleft) 3024 return (EINVAL); 3025 break; 3026 default: 3027 return (EINVAL); /* not supported */ 3028 } 3029 /* turn off the previous option */ 3030 ip6_clearpktopts(opt, IPV6_RTHDR); 3031 opt->ip6po_rthdr = malloc(rthlen, M_IP6OPT, M_NOWAIT); 3032 if (opt->ip6po_rthdr == NULL) 3033 return (ENOBUFS); 3034 memcpy(opt->ip6po_rthdr, rth, rthlen); 3035 break; 3036 } 3037 3038 case IPV6_USE_MIN_MTU: 3039 if (len != sizeof(int)) 3040 return (EINVAL); 3041 minmtupolicy = *(int *)buf; 3042 if (minmtupolicy != IP6PO_MINMTU_MCASTONLY && 3043 minmtupolicy != IP6PO_MINMTU_DISABLE && 3044 minmtupolicy != IP6PO_MINMTU_ALL) { 3045 return (EINVAL); 3046 } 3047 opt->ip6po_minmtu = minmtupolicy; 3048 break; 3049 3050 case IPV6_DONTFRAG: 3051 if (len != sizeof(int)) 3052 return (EINVAL); 3053 3054 if (uproto == IPPROTO_TCP || *(int *)buf == 0) { 3055 /* 3056 * we ignore this option for TCP sockets. 3057 * (RFC3542 leaves this case unspecified.) 3058 */ 3059 opt->ip6po_flags &= ~IP6PO_DONTFRAG; 3060 } else 3061 opt->ip6po_flags |= IP6PO_DONTFRAG; 3062 break; 3063 3064 default: 3065 return (ENOPROTOOPT); 3066 } /* end of switch */ 3067 3068 return (0); 3069 } 3070 3071 /* 3072 * Routine called from ip6_output() to loop back a copy of an IP6 multicast 3073 * packet to the input queue of a specified interface. Note that this 3074 * calls the output routine of the loopback "driver", but with an interface 3075 * pointer that might NOT be lo0ifp -- easier than replicating that code here. 3076 */ 3077 void 3078 ip6_mloopback(struct ifnet *ifp, struct mbuf *m, 3079 const struct sockaddr_in6 *dst) 3080 { 3081 struct mbuf *copym; 3082 struct ip6_hdr *ip6; 3083 3084 copym = m_copy(m, 0, M_COPYALL); 3085 if (copym == NULL) 3086 return; 3087 3088 /* 3089 * Make sure to deep-copy IPv6 header portion in case the data 3090 * is in an mbuf cluster, so that we can safely override the IPv6 3091 * header portion later. 3092 */ 3093 if ((copym->m_flags & M_EXT) != 0 || 3094 copym->m_len < sizeof(struct ip6_hdr)) { 3095 copym = m_pullup(copym, sizeof(struct ip6_hdr)); 3096 if (copym == NULL) 3097 return; 3098 } 3099 3100 #ifdef DIAGNOSTIC 3101 if (copym->m_len < sizeof(*ip6)) { 3102 m_freem(copym); 3103 return; 3104 } 3105 #endif 3106 3107 ip6 = mtod(copym, struct ip6_hdr *); 3108 /* 3109 * clear embedded scope identifiers if necessary. 3110 * in6_clearscope will touch the addresses only when necessary. 3111 */ 3112 in6_clearscope(&ip6->ip6_src); 3113 in6_clearscope(&ip6->ip6_dst); 3114 3115 (void)looutput(ifp, copym, (const struct sockaddr *)dst, NULL); 3116 } 3117 3118 /* 3119 * Chop IPv6 header off from the payload. 3120 */ 3121 static int 3122 ip6_splithdr(struct mbuf *m, struct ip6_exthdrs *exthdrs) 3123 { 3124 struct mbuf *mh; 3125 struct ip6_hdr *ip6; 3126 3127 ip6 = mtod(m, struct ip6_hdr *); 3128 if (m->m_len > sizeof(*ip6)) { 3129 MGETHDR(mh, M_DONTWAIT, MT_HEADER); 3130 if (mh == 0) { 3131 m_freem(m); 3132 return ENOBUFS; 3133 } 3134 M_MOVE_PKTHDR(mh, m); 3135 MH_ALIGN(mh, sizeof(*ip6)); 3136 m->m_len -= sizeof(*ip6); 3137 m->m_data += sizeof(*ip6); 3138 mh->m_next = m; 3139 m = mh; 3140 m->m_len = sizeof(*ip6); 3141 bcopy((void *)ip6, mtod(m, void *), sizeof(*ip6)); 3142 } 3143 exthdrs->ip6e_ip6 = m; 3144 return 0; 3145 } 3146 3147 /* 3148 * Compute IPv6 extension header length. 3149 */ 3150 int 3151 ip6_optlen(struct in6pcb *in6p) 3152 { 3153 int len; 3154 3155 if (!in6p->in6p_outputopts) 3156 return 0; 3157 3158 len = 0; 3159 #define elen(x) \ 3160 (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0) 3161 3162 len += elen(in6p->in6p_outputopts->ip6po_hbh); 3163 len += elen(in6p->in6p_outputopts->ip6po_dest1); 3164 len += elen(in6p->in6p_outputopts->ip6po_rthdr); 3165 len += elen(in6p->in6p_outputopts->ip6po_dest2); 3166 return len; 3167 #undef elen 3168 } 3169
Indexes created Wed Sep 24 05:09:52 GMT 2025