1 /* $NetBSD: ip6_output.c,v 1.103 2006/10/12 01:32:39 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 1.103 2006/10/12 01:32:39 christos Exp $"); 66 67 #include "opt_inet.h" 68 #include "opt_inet6.h" 69 #include "opt_ipsec.h" 70 #include "opt_pfil_hooks.h" 71 72 #include <sys/param.h> 73 #include <sys/malloc.h> 74 #include <sys/mbuf.h> 75 #include <sys/errno.h> 76 #include <sys/protosw.h> 77 #include <sys/socket.h> 78 #include <sys/socketvar.h> 79 #include <sys/systm.h> 80 #include <sys/proc.h> 81 #include <sys/kauth.h> 82 83 #include <net/if.h> 84 #include <net/route.h> 85 #ifdef PFIL_HOOKS 86 #include <net/pfil.h> 87 #endif 88 89 #include <netinet/in.h> 90 #include <netinet/in_var.h> 91 #include <netinet/ip6.h> 92 #include <netinet/icmp6.h> 93 #include <netinet/in_offload.h> 94 #include <netinet6/ip6_var.h> 95 #include <netinet6/in6_pcb.h> 96 #include <netinet6/nd6.h> 97 #include <netinet6/ip6protosw.h> 98 #include <netinet6/scope6_var.h> 99 100 #ifdef IPSEC 101 #include <netinet6/ipsec.h> 102 #include <netkey/key.h> 103 #endif /* IPSEC */ 104 105 #include <net/net_osdep.h> 106 107 #ifdef PFIL_HOOKS 108 extern struct pfil_head inet6_pfil_hook; /* XXX */ 109 #endif 110 111 struct ip6_exthdrs { 112 struct mbuf *ip6e_ip6; 113 struct mbuf *ip6e_hbh; 114 struct mbuf *ip6e_dest1; 115 struct mbuf *ip6e_rthdr; 116 struct mbuf *ip6e_dest2; 117 }; 118 119 static int ip6_pcbopt __P((int, u_char *, int, struct ip6_pktopts **, 120 int, int)); 121 static int ip6_getpcbopt __P((struct ip6_pktopts *, int, struct mbuf **)); 122 static int ip6_setpktopt __P((int, u_char *, int, struct ip6_pktopts *, int, 123 int, int, int)); 124 static int ip6_setmoptions __P((int, struct ip6_moptions **, struct mbuf *)); 125 static int ip6_getmoptions __P((int, struct ip6_moptions *, struct mbuf **)); 126 static int ip6_copyexthdr __P((struct mbuf **, caddr_t, int)); 127 static int ip6_insertfraghdr __P((struct mbuf *, struct mbuf *, int, 128 struct ip6_frag **)); 129 static int ip6_insert_jumboopt __P((struct ip6_exthdrs *, u_int32_t)); 130 static int ip6_splithdr __P((struct mbuf *, struct ip6_exthdrs *)); 131 static int ip6_getpmtu __P((struct route_in6 *, struct route_in6 *, 132 struct ifnet *, struct in6_addr *, u_long *, int *)); 133 static int copypktopts __P((struct ip6_pktopts *, struct ip6_pktopts *, int)); 134 135 #ifdef RFC2292 136 static int ip6_pcbopts __P((struct ip6_pktopts **, struct mbuf *, 137 struct socket *)); 138 #endif 139 140 #define IN6_NEED_CHECKSUM(ifp, csum_flags) \ 141 (__predict_true(((ifp)->if_flags & IFF_LOOPBACK) == 0 || \ 142 (((csum_flags) & M_CSUM_UDPv6) != 0 && udp_do_loopback_cksum) || \ 143 (((csum_flags) & M_CSUM_TCPv6) != 0 && tcp_do_loopback_cksum))) 144 145 /* 146 * IP6 output. The packet in mbuf chain m contains a skeletal IP6 147 * header (with pri, len, nxt, hlim, src, dst). 148 * This function may modify ver and hlim only. 149 * The mbuf chain containing the packet will be freed. 150 * The mbuf opt, if present, will not be freed. 151 * 152 * type of "mtu": rt_rmx.rmx_mtu is u_long, ifnet.ifr_mtu is int, and 153 * nd_ifinfo.linkmtu is u_int32_t. so we use u_long to hold largest one, 154 * which is rt_rmx.rmx_mtu. 155 */ 156 int 157 ip6_output( 158 struct mbuf *m0, 159 struct ip6_pktopts *opt, 160 struct route_in6 *ro, 161 int flags, 162 struct ip6_moptions *im6o, 163 struct socket *so __unused, 164 struct ifnet **ifpp /* XXX: just for statistics */ 165 ) 166 { 167 struct ip6_hdr *ip6, *mhip6; 168 struct ifnet *ifp, *origifp; 169 struct mbuf *m = m0; 170 int hlen, tlen, len, off; 171 struct route_in6 ip6route; 172 struct rtentry *rt = NULL; 173 struct sockaddr_in6 *dst, src_sa, dst_sa; 174 int error = 0; 175 struct in6_ifaddr *ia = NULL; 176 u_long mtu; 177 int alwaysfrag, dontfrag; 178 u_int32_t optlen = 0, plen = 0, unfragpartlen = 0; 179 struct ip6_exthdrs exthdrs; 180 struct in6_addr finaldst, src0, dst0; 181 u_int32_t zone; 182 struct route_in6 *ro_pmtu = NULL; 183 int hdrsplit = 0; 184 int needipsec = 0; 185 #ifdef IPSEC 186 int needipsectun = 0; 187 struct secpolicy *sp = NULL; 188 189 ip6 = mtod(m, struct ip6_hdr *); 190 #endif /* IPSEC */ 191 192 #ifdef DIAGNOSTIC 193 if ((m->m_flags & M_PKTHDR) == 0) 194 panic("ip6_output: no HDR"); 195 196 if ((m->m_pkthdr.csum_flags & 197 (M_CSUM_TCPv4|M_CSUM_UDPv4|M_CSUM_TSOv4)) != 0) { 198 panic("ip6_output: IPv4 checksum offload flags: %d", 199 m->m_pkthdr.csum_flags); 200 } 201 202 if ((m->m_pkthdr.csum_flags & (M_CSUM_TCPv6|M_CSUM_UDPv6)) == 203 (M_CSUM_TCPv6|M_CSUM_UDPv6)) { 204 panic("ip6_output: conflicting checksum offload flags: %d", 205 m->m_pkthdr.csum_flags); 206 } 207 #endif 208 209 M_CSUM_DATA_IPv6_HL_SET(m->m_pkthdr.csum_data, sizeof(struct ip6_hdr)); 210 211 #define MAKE_EXTHDR(hp, mp) \ 212 do { \ 213 if (hp) { \ 214 struct ip6_ext *eh = (struct ip6_ext *)(hp); \ 215 error = ip6_copyexthdr((mp), (caddr_t)(hp), \ 216 ((eh)->ip6e_len + 1) << 3); \ 217 if (error) \ 218 goto freehdrs; \ 219 } \ 220 } while (/*CONSTCOND*/ 0) 221 222 bzero(&exthdrs, sizeof(exthdrs)); 223 if (opt) { 224 /* Hop-by-Hop options header */ 225 MAKE_EXTHDR(opt->ip6po_hbh, &exthdrs.ip6e_hbh); 226 /* Destination options header(1st part) */ 227 MAKE_EXTHDR(opt->ip6po_dest1, &exthdrs.ip6e_dest1); 228 /* Routing header */ 229 MAKE_EXTHDR(opt->ip6po_rthdr, &exthdrs.ip6e_rthdr); 230 /* Destination options header(2nd part) */ 231 MAKE_EXTHDR(opt->ip6po_dest2, &exthdrs.ip6e_dest2); 232 } 233 234 #ifdef IPSEC 235 if ((flags & IPV6_FORWARDING) != 0) { 236 needipsec = 0; 237 goto skippolicycheck; 238 } 239 240 /* get a security policy for this packet */ 241 if (so == NULL) 242 sp = ipsec6_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND, 0, &error); 243 else { 244 if (IPSEC_PCB_SKIP_IPSEC(sotoinpcb_hdr(so)->inph_sp, 245 IPSEC_DIR_OUTBOUND)) { 246 needipsec = 0; 247 goto skippolicycheck; 248 } 249 sp = ipsec6_getpolicybysock(m, IPSEC_DIR_OUTBOUND, so, &error); 250 } 251 252 if (sp == NULL) { 253 ipsec6stat.out_inval++; 254 goto freehdrs; 255 } 256 257 error = 0; 258 259 /* check policy */ 260 switch (sp->policy) { 261 case IPSEC_POLICY_DISCARD: 262 /* 263 * This packet is just discarded. 264 */ 265 ipsec6stat.out_polvio++; 266 goto freehdrs; 267 268 case IPSEC_POLICY_BYPASS: 269 case IPSEC_POLICY_NONE: 270 /* no need to do IPsec. */ 271 needipsec = 0; 272 break; 273 274 case IPSEC_POLICY_IPSEC: 275 if (sp->req == NULL) { 276 /* XXX should be panic ? */ 277 printf("ip6_output: No IPsec request specified.\n"); 278 error = EINVAL; 279 goto freehdrs; 280 } 281 needipsec = 1; 282 break; 283 284 case IPSEC_POLICY_ENTRUST: 285 default: 286 printf("ip6_output: Invalid policy found. %d\n", sp->policy); 287 } 288 289 skippolicycheck:; 290 #endif /* IPSEC */ 291 292 if (needipsec && 293 (m->m_pkthdr.csum_flags & (M_CSUM_UDPv6|M_CSUM_TCPv6)) != 0) { 294 in6_delayed_cksum(m); 295 m->m_pkthdr.csum_flags &= ~(M_CSUM_UDPv6|M_CSUM_TCPv6); 296 } 297 298 /* 299 * Calculate the total length of the extension header chain. 300 * Keep the length of the unfragmentable part for fragmentation. 301 */ 302 optlen = 0; 303 if (exthdrs.ip6e_hbh) optlen += exthdrs.ip6e_hbh->m_len; 304 if (exthdrs.ip6e_dest1) optlen += exthdrs.ip6e_dest1->m_len; 305 if (exthdrs.ip6e_rthdr) optlen += exthdrs.ip6e_rthdr->m_len; 306 unfragpartlen = optlen + sizeof(struct ip6_hdr); 307 /* NOTE: we don't add AH/ESP length here. do that later. */ 308 if (exthdrs.ip6e_dest2) optlen += exthdrs.ip6e_dest2->m_len; 309 310 /* 311 * If we need IPsec, or there is at least one extension header, 312 * separate IP6 header from the payload. 313 */ 314 if ((needipsec || optlen) && !hdrsplit) { 315 if ((error = ip6_splithdr(m, &exthdrs)) != 0) { 316 m = NULL; 317 goto freehdrs; 318 } 319 m = exthdrs.ip6e_ip6; 320 hdrsplit++; 321 } 322 323 /* adjust pointer */ 324 ip6 = mtod(m, struct ip6_hdr *); 325 326 /* adjust mbuf packet header length */ 327 m->m_pkthdr.len += optlen; 328 plen = m->m_pkthdr.len - sizeof(*ip6); 329 330 /* If this is a jumbo payload, insert a jumbo payload option. */ 331 if (plen > IPV6_MAXPACKET) { 332 if (!hdrsplit) { 333 if ((error = ip6_splithdr(m, &exthdrs)) != 0) { 334 m = NULL; 335 goto freehdrs; 336 } 337 m = exthdrs.ip6e_ip6; 338 hdrsplit++; 339 } 340 /* adjust pointer */ 341 ip6 = mtod(m, struct ip6_hdr *); 342 if ((error = ip6_insert_jumboopt(&exthdrs, plen)) != 0) 343 goto freehdrs; 344 optlen += 8; /* XXX JUMBOOPTLEN */ 345 ip6->ip6_plen = 0; 346 } else 347 ip6->ip6_plen = htons(plen); 348 349 /* 350 * Concatenate headers and fill in next header fields. 351 * Here we have, on "m" 352 * IPv6 payload 353 * and we insert headers accordingly. Finally, we should be getting: 354 * IPv6 hbh dest1 rthdr ah* [esp* dest2 payload] 355 * 356 * during the header composing process, "m" points to IPv6 header. 357 * "mprev" points to an extension header prior to esp. 358 */ 359 { 360 u_char *nexthdrp = &ip6->ip6_nxt; 361 struct mbuf *mprev = m; 362 363 /* 364 * we treat dest2 specially. this makes IPsec processing 365 * much easier. the goal here is to make mprev point the 366 * mbuf prior to dest2. 367 * 368 * result: IPv6 dest2 payload 369 * m and mprev will point to IPv6 header. 370 */ 371 if (exthdrs.ip6e_dest2) { 372 if (!hdrsplit) 373 panic("assumption failed: hdr not split"); 374 exthdrs.ip6e_dest2->m_next = m->m_next; 375 m->m_next = exthdrs.ip6e_dest2; 376 *mtod(exthdrs.ip6e_dest2, u_char *) = ip6->ip6_nxt; 377 ip6->ip6_nxt = IPPROTO_DSTOPTS; 378 } 379 380 #define MAKE_CHAIN(m, mp, p, i)\ 381 do {\ 382 if (m) {\ 383 if (!hdrsplit) \ 384 panic("assumption failed: hdr not split"); \ 385 *mtod((m), u_char *) = *(p);\ 386 *(p) = (i);\ 387 p = mtod((m), u_char *);\ 388 (m)->m_next = (mp)->m_next;\ 389 (mp)->m_next = (m);\ 390 (mp) = (m);\ 391 }\ 392 } while (/*CONSTCOND*/ 0) 393 /* 394 * result: IPv6 hbh dest1 rthdr dest2 payload 395 * m will point to IPv6 header. mprev will point to the 396 * extension header prior to dest2 (rthdr in the above case). 397 */ 398 MAKE_CHAIN(exthdrs.ip6e_hbh, mprev, nexthdrp, IPPROTO_HOPOPTS); 399 MAKE_CHAIN(exthdrs.ip6e_dest1, mprev, nexthdrp, 400 IPPROTO_DSTOPTS); 401 MAKE_CHAIN(exthdrs.ip6e_rthdr, mprev, nexthdrp, 402 IPPROTO_ROUTING); 403 404 M_CSUM_DATA_IPv6_HL_SET(m->m_pkthdr.csum_data, 405 sizeof(struct ip6_hdr) + optlen); 406 407 #ifdef IPSEC 408 if (!needipsec) 409 goto skip_ipsec2; 410 411 /* 412 * pointers after IPsec headers are not valid any more. 413 * other pointers need a great care too. 414 * (IPsec routines should not mangle mbufs prior to AH/ESP) 415 */ 416 exthdrs.ip6e_dest2 = NULL; 417 418 { 419 struct ip6_rthdr *rh = NULL; 420 int segleft_org = 0; 421 struct ipsec_output_state state; 422 423 if (exthdrs.ip6e_rthdr) { 424 rh = mtod(exthdrs.ip6e_rthdr, struct ip6_rthdr *); 425 segleft_org = rh->ip6r_segleft; 426 rh->ip6r_segleft = 0; 427 } 428 429 bzero(&state, sizeof(state)); 430 state.m = m; 431 error = ipsec6_output_trans(&state, nexthdrp, mprev, sp, flags, 432 &needipsectun); 433 m = state.m; 434 if (error) { 435 /* mbuf is already reclaimed in ipsec6_output_trans. */ 436 m = NULL; 437 switch (error) { 438 case EHOSTUNREACH: 439 case ENETUNREACH: 440 case EMSGSIZE: 441 case ENOBUFS: 442 case ENOMEM: 443 break; 444 default: 445 printf("ip6_output (ipsec): error code %d\n", error); 446 /* FALLTHROUGH */ 447 case ENOENT: 448 /* don't show these error codes to the user */ 449 error = 0; 450 break; 451 } 452 goto bad; 453 } 454 if (exthdrs.ip6e_rthdr) { 455 /* ah6_output doesn't modify mbuf chain */ 456 rh->ip6r_segleft = segleft_org; 457 } 458 } 459 skip_ipsec2:; 460 #endif 461 } 462 463 /* 464 * If there is a routing header, replace destination address field 465 * with the first hop of the routing header. 466 */ 467 if (exthdrs.ip6e_rthdr) { 468 struct ip6_rthdr *rh; 469 struct ip6_rthdr0 *rh0; 470 struct in6_addr *addr; 471 struct sockaddr_in6 sa; 472 473 rh = (struct ip6_rthdr *)(mtod(exthdrs.ip6e_rthdr, 474 struct ip6_rthdr *)); 475 finaldst = ip6->ip6_dst; 476 switch (rh->ip6r_type) { 477 case IPV6_RTHDR_TYPE_0: 478 rh0 = (struct ip6_rthdr0 *)rh; 479 addr = (struct in6_addr *)(rh0 + 1); 480 481 /* 482 * construct a sockaddr_in6 form of 483 * the first hop. 484 * 485 * XXX: we may not have enough 486 * information about its scope zone; 487 * there is no standard API to pass 488 * the information from the 489 * application. 490 */ 491 bzero(&sa, sizeof(sa)); 492 sa.sin6_family = AF_INET6; 493 sa.sin6_len = sizeof(sa); 494 sa.sin6_addr = addr[0]; 495 if ((error = sa6_embedscope(&sa, 496 ip6_use_defzone)) != 0) { 497 goto bad; 498 } 499 ip6->ip6_dst = sa.sin6_addr; 500 (void)memmove(&addr[0], &addr[1], 501 sizeof(struct in6_addr) * 502 (rh0->ip6r0_segleft - 1)); 503 addr[rh0->ip6r0_segleft - 1] = finaldst; 504 /* XXX */ 505 in6_clearscope(addr + rh0->ip6r0_segleft - 1); 506 break; 507 default: /* is it possible? */ 508 error = EINVAL; 509 goto bad; 510 } 511 } 512 513 /* Source address validation */ 514 if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) && 515 (flags & IPV6_UNSPECSRC) == 0) { 516 error = EOPNOTSUPP; 517 ip6stat.ip6s_badscope++; 518 goto bad; 519 } 520 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) { 521 error = EOPNOTSUPP; 522 ip6stat.ip6s_badscope++; 523 goto bad; 524 } 525 526 ip6stat.ip6s_localout++; 527 528 /* 529 * Route packet. 530 */ 531 /* initialize cached route */ 532 if (ro == 0) { 533 ro = &ip6route; 534 bzero((caddr_t)ro, sizeof(*ro)); 535 } 536 ro_pmtu = ro; 537 if (opt && opt->ip6po_rthdr) 538 ro = &opt->ip6po_route; 539 dst = (struct sockaddr_in6 *)&ro->ro_dst; 540 541 /* 542 * if specified, try to fill in the traffic class field. 543 * do not override if a non-zero value is already set. 544 * we check the diffserv field and the ecn field separately. 545 */ 546 if (opt && opt->ip6po_tclass >= 0) { 547 int mask = 0; 548 549 if ((ip6->ip6_flow & htonl(0xfc << 20)) == 0) 550 mask |= 0xfc; 551 if ((ip6->ip6_flow & htonl(0x03 << 20)) == 0) 552 mask |= 0x03; 553 if (mask != 0) 554 ip6->ip6_flow |= htonl((opt->ip6po_tclass & mask) << 20); 555 } 556 557 /* fill in or override the hop limit field, if necessary. */ 558 if (opt && opt->ip6po_hlim != -1) 559 ip6->ip6_hlim = opt->ip6po_hlim & 0xff; 560 else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) { 561 if (im6o != NULL) 562 ip6->ip6_hlim = im6o->im6o_multicast_hlim; 563 else 564 ip6->ip6_hlim = ip6_defmcasthlim; 565 } 566 567 #ifdef IPSEC 568 if (needipsec && needipsectun) { 569 struct ipsec_output_state state; 570 571 /* 572 * All the extension headers will become inaccessible 573 * (since they can be encrypted). 574 * Don't panic, we need no more updates to extension headers 575 * on inner IPv6 packet (since they are now encapsulated). 576 * 577 * IPv6 [ESP|AH] IPv6 [extension headers] payload 578 */ 579 bzero(&exthdrs, sizeof(exthdrs)); 580 exthdrs.ip6e_ip6 = m; 581 582 bzero(&state, sizeof(state)); 583 state.m = m; 584 state.ro = (struct route *)ro; 585 state.dst = (struct sockaddr *)dst; 586 587 error = ipsec6_output_tunnel(&state, sp, flags); 588 589 m = state.m; 590 ro_pmtu = ro = (struct route_in6 *)state.ro; 591 dst = (struct sockaddr_in6 *)state.dst; 592 if (error) { 593 /* mbuf is already reclaimed in ipsec6_output_tunnel. */ 594 m0 = m = NULL; 595 m = NULL; 596 switch (error) { 597 case EHOSTUNREACH: 598 case ENETUNREACH: 599 case EMSGSIZE: 600 case ENOBUFS: 601 case ENOMEM: 602 break; 603 default: 604 printf("ip6_output (ipsec): error code %d\n", error); 605 /* FALLTHROUGH */ 606 case ENOENT: 607 /* don't show these error codes to the user */ 608 error = 0; 609 break; 610 } 611 goto bad; 612 } 613 614 exthdrs.ip6e_ip6 = m; 615 } 616 #endif /* IPSEC */ 617 618 /* adjust pointer */ 619 ip6 = mtod(m, struct ip6_hdr *); 620 621 bzero(&dst_sa, sizeof(dst_sa)); 622 dst_sa.sin6_family = AF_INET6; 623 dst_sa.sin6_len = sizeof(dst_sa); 624 dst_sa.sin6_addr = ip6->ip6_dst; 625 if ((error = in6_selectroute(&dst_sa, opt, im6o, ro, &ifp, &rt, 0)) 626 != 0) { 627 switch (error) { 628 case EHOSTUNREACH: 629 ip6stat.ip6s_noroute++; 630 break; 631 case EADDRNOTAVAIL: 632 default: 633 break; /* XXX statistics? */ 634 } 635 if (ifp != NULL) 636 in6_ifstat_inc(ifp, ifs6_out_discard); 637 goto bad; 638 } 639 if (rt == NULL) { 640 /* 641 * If in6_selectroute() does not return a route entry, 642 * dst may not have been updated. 643 */ 644 *dst = dst_sa; /* XXX */ 645 } 646 647 /* 648 * then rt (for unicast) and ifp must be non-NULL valid values. 649 */ 650 if ((flags & IPV6_FORWARDING) == 0) { 651 /* XXX: the FORWARDING flag can be set for mrouting. */ 652 in6_ifstat_inc(ifp, ifs6_out_request); 653 } 654 if (rt != NULL) { 655 ia = (struct in6_ifaddr *)(rt->rt_ifa); 656 rt->rt_use++; 657 } 658 659 /* 660 * The outgoing interface must be in the zone of source and 661 * destination addresses. We should use ia_ifp to support the 662 * case of sending packets to an address of our own. 663 */ 664 if (ia != NULL && ia->ia_ifp) 665 origifp = ia->ia_ifp; 666 else 667 origifp = ifp; 668 669 src0 = ip6->ip6_src; 670 if (in6_setscope(&src0, origifp, &zone)) 671 goto badscope; 672 bzero(&src_sa, sizeof(src_sa)); 673 src_sa.sin6_family = AF_INET6; 674 src_sa.sin6_len = sizeof(src_sa); 675 src_sa.sin6_addr = ip6->ip6_src; 676 if (sa6_recoverscope(&src_sa) || zone != src_sa.sin6_scope_id) 677 goto badscope; 678 679 dst0 = ip6->ip6_dst; 680 if (in6_setscope(&dst0, origifp, &zone)) 681 goto badscope; 682 /* re-initialize to be sure */ 683 bzero(&dst_sa, sizeof(dst_sa)); 684 dst_sa.sin6_family = AF_INET6; 685 dst_sa.sin6_len = sizeof(dst_sa); 686 dst_sa.sin6_addr = ip6->ip6_dst; 687 if (sa6_recoverscope(&dst_sa) || zone != dst_sa.sin6_scope_id) 688 goto badscope; 689 690 /* scope check is done. */ 691 goto routefound; 692 693 badscope: 694 ip6stat.ip6s_badscope++; 695 in6_ifstat_inc(origifp, ifs6_out_discard); 696 if (error == 0) 697 error = EHOSTUNREACH; /* XXX */ 698 goto bad; 699 700 routefound: 701 if (rt && !IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) { 702 if (opt && opt->ip6po_nextroute.ro_rt) { 703 /* 704 * The nexthop is explicitly specified by the 705 * application. We assume the next hop is an IPv6 706 * address. 707 */ 708 dst = (struct sockaddr_in6 *)opt->ip6po_nexthop; 709 } else if ((rt->rt_flags & RTF_GATEWAY)) 710 dst = (struct sockaddr_in6 *)rt->rt_gateway; 711 } 712 713 /* 714 * XXXXXX: original code follows: 715 */ 716 if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) 717 m->m_flags &= ~(M_BCAST | M_MCAST); /* just in case */ 718 else { 719 struct in6_multi *in6m; 720 721 m->m_flags = (m->m_flags & ~M_BCAST) | M_MCAST; 722 723 in6_ifstat_inc(ifp, ifs6_out_mcast); 724 725 /* 726 * Confirm that the outgoing interface supports multicast. 727 */ 728 if (!(ifp->if_flags & IFF_MULTICAST)) { 729 ip6stat.ip6s_noroute++; 730 in6_ifstat_inc(ifp, ifs6_out_discard); 731 error = ENETUNREACH; 732 goto bad; 733 } 734 735 IN6_LOOKUP_MULTI(ip6->ip6_dst, ifp, in6m); 736 if (in6m != NULL && 737 (im6o == NULL || im6o->im6o_multicast_loop)) { 738 /* 739 * If we belong to the destination multicast group 740 * on the outgoing interface, and the caller did not 741 * forbid loopback, loop back a copy. 742 */ 743 ip6_mloopback(ifp, m, dst); 744 } else { 745 /* 746 * If we are acting as a multicast router, perform 747 * multicast forwarding as if the packet had just 748 * arrived on the interface to which we are about 749 * to send. The multicast forwarding function 750 * recursively calls this function, using the 751 * IPV6_FORWARDING flag to prevent infinite recursion. 752 * 753 * Multicasts that are looped back by ip6_mloopback(), 754 * above, will be forwarded by the ip6_input() routine, 755 * if necessary. 756 */ 757 if (ip6_mrouter && (flags & IPV6_FORWARDING) == 0) { 758 if (ip6_mforward(ip6, ifp, m) != 0) { 759 m_freem(m); 760 goto done; 761 } 762 } 763 } 764 /* 765 * Multicasts with a hoplimit of zero may be looped back, 766 * above, but must not be transmitted on a network. 767 * Also, multicasts addressed to the loopback interface 768 * are not sent -- the above call to ip6_mloopback() will 769 * loop back a copy if this host actually belongs to the 770 * destination group on the loopback interface. 771 */ 772 if (ip6->ip6_hlim == 0 || (ifp->if_flags & IFF_LOOPBACK) || 773 IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst)) { 774 m_freem(m); 775 goto done; 776 } 777 } 778 779 /* 780 * Fill the outgoing inteface to tell the upper layer 781 * to increment per-interface statistics. 782 */ 783 if (ifpp) 784 *ifpp = ifp; 785 786 /* Determine path MTU. */ 787 if ((error = ip6_getpmtu(ro_pmtu, ro, ifp, &finaldst, &mtu, 788 &alwaysfrag)) != 0) 789 goto bad; 790 #ifdef IPSEC 791 if (needipsectun) 792 mtu = IPV6_MMTU; 793 #endif 794 795 /* 796 * The caller of this function may specify to use the minimum MTU 797 * in some cases. 798 * An advanced API option (IPV6_USE_MIN_MTU) can also override MTU 799 * setting. The logic is a bit complicated; by default, unicast 800 * packets will follow path MTU while multicast packets will be sent at 801 * the minimum MTU. If IP6PO_MINMTU_ALL is specified, all packets 802 * including unicast ones will be sent at the minimum MTU. Multicast 803 * packets will always be sent at the minimum MTU unless 804 * IP6PO_MINMTU_DISABLE is explicitly specified. 805 * See RFC 3542 for more details. 806 */ 807 if (mtu > IPV6_MMTU) { 808 if ((flags & IPV6_MINMTU)) 809 mtu = IPV6_MMTU; 810 else if (opt && opt->ip6po_minmtu == IP6PO_MINMTU_ALL) 811 mtu = IPV6_MMTU; 812 else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) && 813 (opt == NULL || 814 opt->ip6po_minmtu != IP6PO_MINMTU_DISABLE)) { 815 mtu = IPV6_MMTU; 816 } 817 } 818 819 /* 820 * clear embedded scope identifiers if necessary. 821 * in6_clearscope will touch the addresses only when necessary. 822 */ 823 in6_clearscope(&ip6->ip6_src); 824 in6_clearscope(&ip6->ip6_dst); 825 826 /* 827 * If the outgoing packet contains a hop-by-hop options header, 828 * it must be examined and processed even by the source node. 829 * (RFC 2460, section 4.) 830 */ 831 if (exthdrs.ip6e_hbh) { 832 struct ip6_hbh *hbh = mtod(exthdrs.ip6e_hbh, struct ip6_hbh *); 833 u_int32_t dummy1; /* XXX unused */ 834 u_int32_t dummy2; /* XXX unused */ 835 836 /* 837 * XXX: if we have to send an ICMPv6 error to the sender, 838 * we need the M_LOOP flag since icmp6_error() expects 839 * the IPv6 and the hop-by-hop options header are 840 * continuous unless the flag is set. 841 */ 842 m->m_flags |= M_LOOP; 843 m->m_pkthdr.rcvif = ifp; 844 if (ip6_process_hopopts(m, (u_int8_t *)(hbh + 1), 845 ((hbh->ip6h_len + 1) << 3) - sizeof(struct ip6_hbh), 846 &dummy1, &dummy2) < 0) { 847 /* m was already freed at this point */ 848 error = EINVAL;/* better error? */ 849 goto done; 850 } 851 m->m_flags &= ~M_LOOP; /* XXX */ 852 m->m_pkthdr.rcvif = NULL; 853 } 854 855 #ifdef PFIL_HOOKS 856 /* 857 * Run through list of hooks for output packets. 858 */ 859 if ((error = pfil_run_hooks(&inet6_pfil_hook, &m, ifp, PFIL_OUT)) != 0) 860 goto done; 861 if (m == NULL) 862 goto done; 863 ip6 = mtod(m, struct ip6_hdr *); 864 #endif /* PFIL_HOOKS */ 865 /* 866 * Send the packet to the outgoing interface. 867 * If necessary, do IPv6 fragmentation before sending. 868 * 869 * the logic here is rather complex: 870 * 1: normal case (dontfrag == 0, alwaysfrag == 0) 871 * 1-a: send as is if tlen <= path mtu 872 * 1-b: fragment if tlen > path mtu 873 * 874 * 2: if user asks us not to fragment (dontfrag == 1) 875 * 2-a: send as is if tlen <= interface mtu 876 * 2-b: error if tlen > interface mtu 877 * 878 * 3: if we always need to attach fragment header (alwaysfrag == 1) 879 * always fragment 880 * 881 * 4: if dontfrag == 1 && alwaysfrag == 1 882 * error, as we cannot handle this conflicting request 883 */ 884 tlen = m->m_pkthdr.len; 885 886 if (opt && (opt->ip6po_flags & IP6PO_DONTFRAG)) 887 dontfrag = 1; 888 else 889 dontfrag = 0; 890 891 if (dontfrag && alwaysfrag) { /* case 4 */ 892 /* conflicting request - can't transmit */ 893 error = EMSGSIZE; 894 goto bad; 895 } 896 if (dontfrag && tlen > IN6_LINKMTU(ifp)) { /* case 2-b */ 897 /* 898 * Even if the DONTFRAG option is specified, we cannot send the 899 * packet when the data length is larger than the MTU of the 900 * outgoing interface. 901 * Notify the error by sending IPV6_PATHMTU ancillary data as 902 * well as returning an error code (the latter is not described 903 * in the API spec.) 904 */ 905 u_int32_t mtu32; 906 struct ip6ctlparam ip6cp; 907 908 mtu32 = (u_int32_t)mtu; 909 bzero(&ip6cp, sizeof(ip6cp)); 910 ip6cp.ip6c_cmdarg = (void *)&mtu32; 911 pfctlinput2(PRC_MSGSIZE, (struct sockaddr *)&ro_pmtu->ro_dst, 912 (void *)&ip6cp); 913 914 error = EMSGSIZE; 915 goto bad; 916 } 917 918 /* 919 * transmit packet without fragmentation 920 */ 921 if (dontfrag || (!alwaysfrag && tlen <= mtu)) { /* case 1-a and 2-a */ 922 struct in6_ifaddr *ia6; 923 int sw_csum; 924 925 ip6 = mtod(m, struct ip6_hdr *); 926 ia6 = in6_ifawithifp(ifp, &ip6->ip6_src); 927 if (ia6) { 928 /* Record statistics for this interface address. */ 929 ia6->ia_ifa.ifa_data.ifad_outbytes += m->m_pkthdr.len; 930 } 931 #ifdef IPSEC 932 /* clean ipsec history once it goes out of the node */ 933 ipsec_delaux(m); 934 #endif 935 936 sw_csum = m->m_pkthdr.csum_flags & ~ifp->if_csum_flags_tx; 937 if ((sw_csum & (M_CSUM_UDPv6|M_CSUM_TCPv6)) != 0) { 938 if (IN6_NEED_CHECKSUM(ifp, 939 sw_csum & (M_CSUM_UDPv6|M_CSUM_TCPv6))) { 940 in6_delayed_cksum(m); 941 } 942 m->m_pkthdr.csum_flags &= ~(M_CSUM_UDPv6|M_CSUM_TCPv6); 943 } 944 945 error = nd6_output(ifp, origifp, m, dst, rt); 946 goto done; 947 } 948 949 /* 950 * try to fragment the packet. case 1-b and 3 951 */ 952 if (mtu < IPV6_MMTU) { 953 /* path MTU cannot be less than IPV6_MMTU */ 954 error = EMSGSIZE; 955 in6_ifstat_inc(ifp, ifs6_out_fragfail); 956 goto bad; 957 } else if (ip6->ip6_plen == 0) { 958 /* jumbo payload cannot be fragmented */ 959 error = EMSGSIZE; 960 in6_ifstat_inc(ifp, ifs6_out_fragfail); 961 goto bad; 962 } else { 963 struct mbuf **mnext, *m_frgpart; 964 struct ip6_frag *ip6f; 965 u_int32_t id = htonl(ip6_randomid()); 966 u_char nextproto; 967 #if 0 /* see below */ 968 struct ip6ctlparam ip6cp; 969 u_int32_t mtu32; 970 #endif 971 972 /* 973 * Too large for the destination or interface; 974 * fragment if possible. 975 * Must be able to put at least 8 bytes per fragment. 976 */ 977 hlen = unfragpartlen; 978 if (mtu > IPV6_MAXPACKET) 979 mtu = IPV6_MAXPACKET; 980 981 #if 0 982 /* 983 * It is believed this code is a leftover from the 984 * development of the IPV6_RECVPATHMTU sockopt and 985 * associated work to implement RFC3542. 986 * It's not entirely clear what the intent of the API 987 * is at this point, so disable this code for now. 988 * The IPV6_RECVPATHMTU sockopt and/or IPV6_DONTFRAG 989 * will send notifications if the application requests. 990 */ 991 992 /* Notify a proper path MTU to applications. */ 993 mtu32 = (u_int32_t)mtu; 994 bzero(&ip6cp, sizeof(ip6cp)); 995 ip6cp.ip6c_cmdarg = (void *)&mtu32; 996 pfctlinput2(PRC_MSGSIZE, (struct sockaddr *)&ro_pmtu->ro_dst, 997 (void *)&ip6cp); 998 #endif 999 1000 len = (mtu - hlen - sizeof(struct ip6_frag)) & ~7; 1001 if (len < 8) { 1002 error = EMSGSIZE; 1003 in6_ifstat_inc(ifp, ifs6_out_fragfail); 1004 goto bad; 1005 } 1006 1007 mnext = &m->m_nextpkt; 1008 1009 /* 1010 * Change the next header field of the last header in the 1011 * unfragmentable part. 1012 */ 1013 if (exthdrs.ip6e_rthdr) { 1014 nextproto = *mtod(exthdrs.ip6e_rthdr, u_char *); 1015 *mtod(exthdrs.ip6e_rthdr, u_char *) = IPPROTO_FRAGMENT; 1016 } else if (exthdrs.ip6e_dest1) { 1017 nextproto = *mtod(exthdrs.ip6e_dest1, u_char *); 1018 *mtod(exthdrs.ip6e_dest1, u_char *) = IPPROTO_FRAGMENT; 1019 } else if (exthdrs.ip6e_hbh) { 1020 nextproto = *mtod(exthdrs.ip6e_hbh, u_char *); 1021 *mtod(exthdrs.ip6e_hbh, u_char *) = IPPROTO_FRAGMENT; 1022 } else { 1023 nextproto = ip6->ip6_nxt; 1024 ip6->ip6_nxt = IPPROTO_FRAGMENT; 1025 } 1026 1027 if ((m->m_pkthdr.csum_flags & (M_CSUM_UDPv6|M_CSUM_TCPv6)) 1028 != 0) { 1029 if (IN6_NEED_CHECKSUM(ifp, 1030 m->m_pkthdr.csum_flags & 1031 (M_CSUM_UDPv6|M_CSUM_TCPv6))) { 1032 in6_delayed_cksum(m); 1033 } 1034 m->m_pkthdr.csum_flags &= ~(M_CSUM_UDPv6|M_CSUM_TCPv6); 1035 } 1036 1037 /* 1038 * Loop through length of segment after first fragment, 1039 * make new header and copy data of each part and link onto 1040 * chain. 1041 */ 1042 m0 = m; 1043 for (off = hlen; off < tlen; off += len) { 1044 struct mbuf *mlast; 1045 1046 MGETHDR(m, M_DONTWAIT, MT_HEADER); 1047 if (!m) { 1048 error = ENOBUFS; 1049 ip6stat.ip6s_odropped++; 1050 goto sendorfree; 1051 } 1052 m->m_pkthdr.rcvif = NULL; 1053 m->m_flags = m0->m_flags & M_COPYFLAGS; 1054 *mnext = m; 1055 mnext = &m->m_nextpkt; 1056 m->m_data += max_linkhdr; 1057 mhip6 = mtod(m, struct ip6_hdr *); 1058 *mhip6 = *ip6; 1059 m->m_len = sizeof(*mhip6); 1060 error = ip6_insertfraghdr(m0, m, hlen, &ip6f); 1061 if (error) { 1062 ip6stat.ip6s_odropped++; 1063 goto sendorfree; 1064 } 1065 ip6f->ip6f_offlg = htons((u_int16_t)((off - hlen) & ~7)); 1066 if (off + len >= tlen) 1067 len = tlen - off; 1068 else 1069 ip6f->ip6f_offlg |= IP6F_MORE_FRAG; 1070 mhip6->ip6_plen = htons((u_int16_t)(len + hlen + 1071 sizeof(*ip6f) - sizeof(struct ip6_hdr))); 1072 if ((m_frgpart = m_copy(m0, off, len)) == 0) { 1073 error = ENOBUFS; 1074 ip6stat.ip6s_odropped++; 1075 goto sendorfree; 1076 } 1077 for (mlast = m; mlast->m_next; mlast = mlast->m_next) 1078 ; 1079 mlast->m_next = m_frgpart; 1080 m->m_pkthdr.len = len + hlen + sizeof(*ip6f); 1081 m->m_pkthdr.rcvif = (struct ifnet *)0; 1082 ip6f->ip6f_reserved = 0; 1083 ip6f->ip6f_ident = id; 1084 ip6f->ip6f_nxt = nextproto; 1085 ip6stat.ip6s_ofragments++; 1086 in6_ifstat_inc(ifp, ifs6_out_fragcreat); 1087 } 1088 1089 in6_ifstat_inc(ifp, ifs6_out_fragok); 1090 } 1091 1092 /* 1093 * Remove leading garbages. 1094 */ 1095 sendorfree: 1096 m = m0->m_nextpkt; 1097 m0->m_nextpkt = 0; 1098 m_freem(m0); 1099 for (m0 = m; m; m = m0) { 1100 m0 = m->m_nextpkt; 1101 m->m_nextpkt = 0; 1102 if (error == 0) { 1103 struct in6_ifaddr *ia6; 1104 ip6 = mtod(m, struct ip6_hdr *); 1105 ia6 = in6_ifawithifp(ifp, &ip6->ip6_src); 1106 if (ia6) { 1107 /* 1108 * Record statistics for this interface 1109 * address. 1110 */ 1111 ia6->ia_ifa.ifa_data.ifad_outbytes += 1112 m->m_pkthdr.len; 1113 } 1114 #ifdef IPSEC 1115 /* clean ipsec history once it goes out of the node */ 1116 ipsec_delaux(m); 1117 #endif 1118 error = nd6_output(ifp, origifp, m, dst, rt); 1119 } else 1120 m_freem(m); 1121 } 1122 1123 if (error == 0) 1124 ip6stat.ip6s_fragmented++; 1125 1126 done: 1127 if (ro == &ip6route && ro->ro_rt) { /* brace necessary for RTFREE */ 1128 RTFREE(ro->ro_rt); 1129 } else if (ro_pmtu == &ip6route && ro_pmtu->ro_rt) { 1130 RTFREE(ro_pmtu->ro_rt); 1131 } 1132 1133 #ifdef IPSEC 1134 if (sp != NULL) 1135 key_freesp(sp); 1136 #endif /* IPSEC */ 1137 1138 return (error); 1139 1140 freehdrs: 1141 m_freem(exthdrs.ip6e_hbh); /* m_freem will check if mbuf is 0 */ 1142 m_freem(exthdrs.ip6e_dest1); 1143 m_freem(exthdrs.ip6e_rthdr); 1144 m_freem(exthdrs.ip6e_dest2); 1145 /* FALLTHROUGH */ 1146 bad: 1147 m_freem(m); 1148 goto done; 1149 } 1150 1151 static int 1152 ip6_copyexthdr(mp, hdr, hlen) 1153 struct mbuf **mp; 1154 caddr_t hdr; 1155 int hlen; 1156 { 1157 struct mbuf *m; 1158 1159 if (hlen > MCLBYTES) 1160 return (ENOBUFS); /* XXX */ 1161 1162 MGET(m, M_DONTWAIT, MT_DATA); 1163 if (!m) 1164 return (ENOBUFS); 1165 1166 if (hlen > MLEN) { 1167 MCLGET(m, M_DONTWAIT); 1168 if ((m->m_flags & M_EXT) == 0) { 1169 m_free(m); 1170 return (ENOBUFS); 1171 } 1172 } 1173 m->m_len = hlen; 1174 if (hdr) 1175 bcopy(hdr, mtod(m, caddr_t), hlen); 1176 1177 *mp = m; 1178 return (0); 1179 } 1180 1181 /* 1182 * Process a delayed payload checksum calculation. 1183 */ 1184 void 1185 in6_delayed_cksum(struct mbuf *m) 1186 { 1187 uint16_t csum, offset; 1188 1189 KASSERT((m->m_pkthdr.csum_flags & (M_CSUM_UDPv6|M_CSUM_TCPv6)) != 0); 1190 KASSERT((~m->m_pkthdr.csum_flags & (M_CSUM_UDPv6|M_CSUM_TCPv6)) != 0); 1191 KASSERT((m->m_pkthdr.csum_flags 1192 & (M_CSUM_UDPv4|M_CSUM_TCPv4|M_CSUM_TSOv4)) == 0); 1193 1194 offset = M_CSUM_DATA_IPv6_HL(m->m_pkthdr.csum_data); 1195 csum = in6_cksum(m, 0, offset, m->m_pkthdr.len - offset); 1196 if (csum == 0 && (m->m_pkthdr.csum_flags & M_CSUM_UDPv6) != 0) { 1197 csum = 0xffff; 1198 } 1199 1200 offset += M_CSUM_DATA_IPv6_OFFSET(m->m_pkthdr.csum_data); 1201 if ((offset + sizeof(csum)) > m->m_len) { 1202 m_copyback(m, offset, sizeof(csum), &csum); 1203 } else { 1204 *(uint16_t *)(mtod(m, caddr_t) + offset) = csum; 1205 } 1206 } 1207 1208 /* 1209 * Insert jumbo payload option. 1210 */ 1211 static int 1212 ip6_insert_jumboopt(exthdrs, plen) 1213 struct ip6_exthdrs *exthdrs; 1214 u_int32_t plen; 1215 { 1216 struct mbuf *mopt; 1217 u_int8_t *optbuf; 1218 u_int32_t v; 1219 1220 #define JUMBOOPTLEN 8 /* length of jumbo payload option and padding */ 1221 1222 /* 1223 * If there is no hop-by-hop options header, allocate new one. 1224 * If there is one but it doesn't have enough space to store the 1225 * jumbo payload option, allocate a cluster to store the whole options. 1226 * Otherwise, use it to store the options. 1227 */ 1228 if (exthdrs->ip6e_hbh == 0) { 1229 MGET(mopt, M_DONTWAIT, MT_DATA); 1230 if (mopt == 0) 1231 return (ENOBUFS); 1232 mopt->m_len = JUMBOOPTLEN; 1233 optbuf = mtod(mopt, u_int8_t *); 1234 optbuf[1] = 0; /* = ((JUMBOOPTLEN) >> 3) - 1 */ 1235 exthdrs->ip6e_hbh = mopt; 1236 } else { 1237 struct ip6_hbh *hbh; 1238 1239 mopt = exthdrs->ip6e_hbh; 1240 if (M_TRAILINGSPACE(mopt) < JUMBOOPTLEN) { 1241 /* 1242 * XXX assumption: 1243 * - exthdrs->ip6e_hbh is not referenced from places 1244 * other than exthdrs. 1245 * - exthdrs->ip6e_hbh is not an mbuf chain. 1246 */ 1247 int oldoptlen = mopt->m_len; 1248 struct mbuf *n; 1249 1250 /* 1251 * XXX: give up if the whole (new) hbh header does 1252 * not fit even in an mbuf cluster. 1253 */ 1254 if (oldoptlen + JUMBOOPTLEN > MCLBYTES) 1255 return (ENOBUFS); 1256 1257 /* 1258 * As a consequence, we must always prepare a cluster 1259 * at this point. 1260 */ 1261 MGET(n, M_DONTWAIT, MT_DATA); 1262 if (n) { 1263 MCLGET(n, M_DONTWAIT); 1264 if ((n->m_flags & M_EXT) == 0) { 1265 m_freem(n); 1266 n = NULL; 1267 } 1268 } 1269 if (!n) 1270 return (ENOBUFS); 1271 n->m_len = oldoptlen + JUMBOOPTLEN; 1272 bcopy(mtod(mopt, caddr_t), mtod(n, caddr_t), 1273 oldoptlen); 1274 optbuf = mtod(n, u_int8_t *) + oldoptlen; 1275 m_freem(mopt); 1276 mopt = exthdrs->ip6e_hbh = n; 1277 } else { 1278 optbuf = mtod(mopt, u_int8_t *) + mopt->m_len; 1279 mopt->m_len += JUMBOOPTLEN; 1280 } 1281 optbuf[0] = IP6OPT_PADN; 1282 optbuf[1] = 0; 1283 1284 /* 1285 * Adjust the header length according to the pad and 1286 * the jumbo payload option. 1287 */ 1288 hbh = mtod(mopt, struct ip6_hbh *); 1289 hbh->ip6h_len += (JUMBOOPTLEN >> 3); 1290 } 1291 1292 /* fill in the option. */ 1293 optbuf[2] = IP6OPT_JUMBO; 1294 optbuf[3] = 4; 1295 v = (u_int32_t)htonl(plen + JUMBOOPTLEN); 1296 bcopy(&v, &optbuf[4], sizeof(u_int32_t)); 1297 1298 /* finally, adjust the packet header length */ 1299 exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN; 1300 1301 return (0); 1302 #undef JUMBOOPTLEN 1303 } 1304 1305 /* 1306 * Insert fragment header and copy unfragmentable header portions. 1307 */ 1308 static int 1309 ip6_insertfraghdr(m0, m, hlen, frghdrp) 1310 struct mbuf *m0, *m; 1311 int hlen; 1312 struct ip6_frag **frghdrp; 1313 { 1314 struct mbuf *n, *mlast; 1315 1316 if (hlen > sizeof(struct ip6_hdr)) { 1317 n = m_copym(m0, sizeof(struct ip6_hdr), 1318 hlen - sizeof(struct ip6_hdr), M_DONTWAIT); 1319 if (n == 0) 1320 return (ENOBUFS); 1321 m->m_next = n; 1322 } else 1323 n = m; 1324 1325 /* Search for the last mbuf of unfragmentable part. */ 1326 for (mlast = n; mlast->m_next; mlast = mlast->m_next) 1327 ; 1328 1329 if ((mlast->m_flags & M_EXT) == 0 && 1330 M_TRAILINGSPACE(mlast) >= sizeof(struct ip6_frag)) { 1331 /* use the trailing space of the last mbuf for the fragment hdr */ 1332 *frghdrp = (struct ip6_frag *)(mtod(mlast, caddr_t) + 1333 mlast->m_len); 1334 mlast->m_len += sizeof(struct ip6_frag); 1335 m->m_pkthdr.len += sizeof(struct ip6_frag); 1336 } else { 1337 /* allocate a new mbuf for the fragment header */ 1338 struct mbuf *mfrg; 1339 1340 MGET(mfrg, M_DONTWAIT, MT_DATA); 1341 if (mfrg == 0) 1342 return (ENOBUFS); 1343 mfrg->m_len = sizeof(struct ip6_frag); 1344 *frghdrp = mtod(mfrg, struct ip6_frag *); 1345 mlast->m_next = mfrg; 1346 } 1347 1348 return (0); 1349 } 1350 1351 static int 1352 ip6_getpmtu(ro_pmtu, ro, ifp, dst, mtup, alwaysfragp) 1353 struct route_in6 *ro_pmtu, *ro; 1354 struct ifnet *ifp; 1355 struct in6_addr *dst; 1356 u_long *mtup; 1357 int *alwaysfragp; 1358 { 1359 u_int32_t mtu = 0; 1360 int alwaysfrag = 0; 1361 int error = 0; 1362 1363 if (ro_pmtu != ro) { 1364 /* The first hop and the final destination may differ. */ 1365 struct sockaddr_in6 *sa6_dst = 1366 (struct sockaddr_in6 *)&ro_pmtu->ro_dst; 1367 if (ro_pmtu->ro_rt && 1368 ((ro_pmtu->ro_rt->rt_flags & RTF_UP) == 0 || 1369 !IN6_ARE_ADDR_EQUAL(&sa6_dst->sin6_addr, dst))) { 1370 RTFREE(ro_pmtu->ro_rt); 1371 ro_pmtu->ro_rt = (struct rtentry *)NULL; 1372 } 1373 if (ro_pmtu->ro_rt == NULL) { 1374 bzero(sa6_dst, sizeof(*sa6_dst)); /* for safety */ 1375 sa6_dst->sin6_family = AF_INET6; 1376 sa6_dst->sin6_len = sizeof(struct sockaddr_in6); 1377 sa6_dst->sin6_addr = *dst; 1378 1379 rtalloc((struct route *)ro_pmtu); 1380 } 1381 } 1382 if (ro_pmtu->ro_rt) { 1383 u_int32_t ifmtu; 1384 1385 if (ifp == NULL) 1386 ifp = ro_pmtu->ro_rt->rt_ifp; 1387 ifmtu = IN6_LINKMTU(ifp); 1388 mtu = ro_pmtu->ro_rt->rt_rmx.rmx_mtu; 1389 if (mtu == 0) 1390 mtu = ifmtu; 1391 else if (mtu < IPV6_MMTU) { 1392 /* 1393 * RFC2460 section 5, last paragraph: 1394 * if we record ICMPv6 too big message with 1395 * mtu < IPV6_MMTU, transmit packets sized IPV6_MMTU 1396 * or smaller, with fragment header attached. 1397 * (fragment header is needed regardless from the 1398 * packet size, for translators to identify packets) 1399 */ 1400 alwaysfrag = 1; 1401 mtu = IPV6_MMTU; 1402 } else if (mtu > ifmtu) { 1403 /* 1404 * The MTU on the route is larger than the MTU on 1405 * the interface! This shouldn't happen, unless the 1406 * MTU of the interface has been changed after the 1407 * interface was brought up. Change the MTU in the 1408 * route to match the interface MTU (as long as the 1409 * field isn't locked). 1410 */ 1411 mtu = ifmtu; 1412 if (!(ro_pmtu->ro_rt->rt_rmx.rmx_locks & RTV_MTU)) 1413 ro_pmtu->ro_rt->rt_rmx.rmx_mtu = mtu; 1414 } 1415 } else if (ifp) { 1416 mtu = IN6_LINKMTU(ifp); 1417 } else 1418 error = EHOSTUNREACH; /* XXX */ 1419 1420 *mtup = mtu; 1421 if (alwaysfragp) 1422 *alwaysfragp = alwaysfrag; 1423 return (error); 1424 } 1425 1426 /* 1427 * IP6 socket option processing. 1428 */ 1429 int 1430 ip6_ctloutput(op, so, level, optname, mp) 1431 int op; 1432 struct socket *so; 1433 int level, optname; 1434 struct mbuf **mp; 1435 { 1436 int privileged, optdatalen, uproto; 1437 void *optdata; 1438 struct in6pcb *in6p = sotoin6pcb(so); 1439 struct mbuf *m = *mp; 1440 int error, optval; 1441 int optlen; 1442 struct lwp *l = curlwp; /* XXX */ 1443 1444 optlen = m ? m->m_len : 0; 1445 error = optval = 0; 1446 privileged = (l == 0 || kauth_authorize_generic(l->l_cred, 1447 KAUTH_GENERIC_ISSUSER, &l->l_acflag)) ? 0 : 1; 1448 uproto = (int)so->so_proto->pr_protocol; 1449 1450 if (level == IPPROTO_IPV6) { 1451 switch (op) { 1452 case PRCO_SETOPT: 1453 switch (optname) { 1454 #ifdef RFC2292 1455 case IPV6_2292PKTOPTIONS: 1456 /* m is freed in ip6_pcbopts */ 1457 error = ip6_pcbopts(&in6p->in6p_outputopts, 1458 m, so); 1459 break; 1460 #endif 1461 1462 /* 1463 * Use of some Hop-by-Hop options or some 1464 * Destination options, might require special 1465 * privilege. That is, normal applications 1466 * (without special privilege) might be forbidden 1467 * from setting certain options in outgoing packets, 1468 * and might never see certain options in received 1469 * packets. [RFC 2292 Section 6] 1470 * KAME specific note: 1471 * KAME prevents non-privileged users from sending or 1472 * receiving ANY hbh/dst options in order to avoid 1473 * overhead of parsing options in the kernel. 1474 */ 1475 case IPV6_RECVHOPOPTS: 1476 case IPV6_RECVDSTOPTS: 1477 case IPV6_RECVRTHDRDSTOPTS: 1478 if (!privileged) { 1479 error = EPERM; 1480 break; 1481 } 1482 /* FALLTHROUGH */ 1483 case IPV6_UNICAST_HOPS: 1484 case IPV6_HOPLIMIT: 1485 case IPV6_FAITH: 1486 1487 case IPV6_RECVPKTINFO: 1488 case IPV6_RECVHOPLIMIT: 1489 case IPV6_RECVRTHDR: 1490 case IPV6_RECVPATHMTU: 1491 case IPV6_RECVTCLASS: 1492 case IPV6_V6ONLY: 1493 if (optlen != sizeof(int)) { 1494 error = EINVAL; 1495 break; 1496 } 1497 optval = *mtod(m, int *); 1498 switch (optname) { 1499 1500 case IPV6_UNICAST_HOPS: 1501 if (optval < -1 || optval >= 256) 1502 error = EINVAL; 1503 else { 1504 /* -1 = kernel default */ 1505 in6p->in6p_hops = optval; 1506 } 1507 break; 1508 #define OPTSET(bit) \ 1509 do { \ 1510 if (optval) \ 1511 in6p->in6p_flags |= (bit); \ 1512 else \ 1513 in6p->in6p_flags &= ~(bit); \ 1514 } while (/*CONSTCOND*/ 0) 1515 1516 #ifdef RFC2292 1517 #define OPTSET2292(bit) \ 1518 do { \ 1519 in6p->in6p_flags |= IN6P_RFC2292; \ 1520 if (optval) \ 1521 in6p->in6p_flags |= (bit); \ 1522 else \ 1523 in6p->in6p_flags &= ~(bit); \ 1524 } while (/*CONSTCOND*/ 0) 1525 #endif 1526 1527 #define OPTBIT(bit) (in6p->in6p_flags & (bit) ? 1 : 0) 1528 1529 case IPV6_RECVPKTINFO: 1530 #ifdef RFC2292 1531 /* cannot mix with RFC2292 */ 1532 if (OPTBIT(IN6P_RFC2292)) { 1533 error = EINVAL; 1534 break; 1535 } 1536 #endif 1537 OPTSET(IN6P_PKTINFO); 1538 break; 1539 1540 case IPV6_HOPLIMIT: 1541 { 1542 struct ip6_pktopts **optp; 1543 1544 #ifdef RFC2292 1545 /* cannot mix with RFC2292 */ 1546 if (OPTBIT(IN6P_RFC2292)) { 1547 error = EINVAL; 1548 break; 1549 } 1550 #endif 1551 optp = &in6p->in6p_outputopts; 1552 error = ip6_pcbopt(IPV6_HOPLIMIT, 1553 (u_char *)&optval, 1554 sizeof(optval), 1555 optp, 1556 privileged, uproto); 1557 break; 1558 } 1559 1560 case IPV6_RECVHOPLIMIT: 1561 #ifdef RFC2292 1562 /* cannot mix with RFC2292 */ 1563 if (OPTBIT(IN6P_RFC2292)) { 1564 error = EINVAL; 1565 break; 1566 } 1567 #endif 1568 OPTSET(IN6P_HOPLIMIT); 1569 break; 1570 1571 case IPV6_RECVHOPOPTS: 1572 #ifdef RFC2292 1573 /* cannot mix with RFC2292 */ 1574 if (OPTBIT(IN6P_RFC2292)) { 1575 error = EINVAL; 1576 break; 1577 } 1578 #endif 1579 OPTSET(IN6P_HOPOPTS); 1580 break; 1581 1582 case IPV6_RECVDSTOPTS: 1583 #ifdef RFC2292 1584 /* cannot mix with RFC2292 */ 1585 if (OPTBIT(IN6P_RFC2292)) { 1586 error = EINVAL; 1587 break; 1588 } 1589 #endif 1590 OPTSET(IN6P_DSTOPTS); 1591 break; 1592 1593 case IPV6_RECVRTHDRDSTOPTS: 1594 #ifdef RFC2292 1595 /* cannot mix with RFC2292 */ 1596 if (OPTBIT(IN6P_RFC2292)) { 1597 error = EINVAL; 1598 break; 1599 } 1600 #endif 1601 OPTSET(IN6P_RTHDRDSTOPTS); 1602 break; 1603 1604 case IPV6_RECVRTHDR: 1605 #ifdef RFC2292 1606 /* cannot mix with RFC2292 */ 1607 if (OPTBIT(IN6P_RFC2292)) { 1608 error = EINVAL; 1609 break; 1610 } 1611 #endif 1612 OPTSET(IN6P_RTHDR); 1613 break; 1614 1615 case IPV6_FAITH: 1616 OPTSET(IN6P_FAITH); 1617 break; 1618 1619 case IPV6_RECVPATHMTU: 1620 /* 1621 * We ignore this option for TCP 1622 * sockets. 1623 * (RFC3542 leaves this case 1624 * unspecified.) 1625 */ 1626 if (uproto != IPPROTO_TCP) 1627 OPTSET(IN6P_MTU); 1628 break; 1629 1630 case IPV6_V6ONLY: 1631 /* 1632 * make setsockopt(IPV6_V6ONLY) 1633 * available only prior to bind(2). 1634 * see ipng mailing list, Jun 22 2001. 1635 */ 1636 if (in6p->in6p_lport || 1637 !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr)) { 1638 error = EINVAL; 1639 break; 1640 } 1641 #ifdef INET6_BINDV6ONLY 1642 if (!optval) 1643 error = EINVAL; 1644 #else 1645 OPTSET(IN6P_IPV6_V6ONLY); 1646 #endif 1647 break; 1648 case IPV6_RECVTCLASS: 1649 #ifdef RFC2292 1650 /* cannot mix with RFC2292 XXX */ 1651 if (OPTBIT(IN6P_RFC2292)) { 1652 error = EINVAL; 1653 break; 1654 } 1655 #endif 1656 OPTSET(IN6P_TCLASS); 1657 break; 1658 1659 } 1660 break; 1661 1662 case IPV6_OTCLASS: 1663 { 1664 struct ip6_pktopts **optp; 1665 u_int8_t tclass; 1666 1667 if (optlen != sizeof(tclass)) { 1668 error = EINVAL; 1669 break; 1670 } 1671 tclass = *mtod(m, u_int8_t *); 1672 optp = &in6p->in6p_outputopts; 1673 error = ip6_pcbopt(optname, 1674 (u_char *)&tclass, 1675 sizeof(tclass), 1676 optp, 1677 privileged, uproto); 1678 break; 1679 } 1680 1681 case IPV6_TCLASS: 1682 case IPV6_DONTFRAG: 1683 case IPV6_USE_MIN_MTU: 1684 if (optlen != sizeof(optval)) { 1685 error = EINVAL; 1686 break; 1687 } 1688 optval = *mtod(m, int *); 1689 { 1690 struct ip6_pktopts **optp; 1691 optp = &in6p->in6p_outputopts; 1692 error = ip6_pcbopt(optname, 1693 (u_char *)&optval, 1694 sizeof(optval), 1695 optp, 1696 privileged, uproto); 1697 break; 1698 } 1699 1700 #ifdef RFC2292 1701 case IPV6_2292PKTINFO: 1702 case IPV6_2292HOPLIMIT: 1703 case IPV6_2292HOPOPTS: 1704 case IPV6_2292DSTOPTS: 1705 case IPV6_2292RTHDR: 1706 /* RFC 2292 */ 1707 if (optlen != sizeof(int)) { 1708 error = EINVAL; 1709 break; 1710 } 1711 optval = *mtod(m, int *); 1712 switch (optname) { 1713 case IPV6_2292PKTINFO: 1714 OPTSET2292(IN6P_PKTINFO); 1715 break; 1716 case IPV6_2292HOPLIMIT: 1717 OPTSET2292(IN6P_HOPLIMIT); 1718 break; 1719 case IPV6_2292HOPOPTS: 1720 /* 1721 * Check super-user privilege. 1722 * See comments for IPV6_RECVHOPOPTS. 1723 */ 1724 if (!privileged) 1725 return (EPERM); 1726 OPTSET2292(IN6P_HOPOPTS); 1727 break; 1728 case IPV6_2292DSTOPTS: 1729 if (!privileged) 1730 return (EPERM); 1731 OPTSET2292(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); /* XXX */ 1732 break; 1733 case IPV6_2292RTHDR: 1734 OPTSET2292(IN6P_RTHDR); 1735 break; 1736 } 1737 break; 1738 #endif 1739 case IPV6_PKTINFO: 1740 case IPV6_HOPOPTS: 1741 case IPV6_RTHDR: 1742 case IPV6_DSTOPTS: 1743 case IPV6_RTHDRDSTOPTS: 1744 case IPV6_NEXTHOP: 1745 { 1746 /* new advanced API (RFC3542) */ 1747 u_char *optbuf; 1748 int optbuflen; 1749 struct ip6_pktopts **optp; 1750 1751 #ifdef RFC2292 1752 /* cannot mix with RFC2292 */ 1753 if (OPTBIT(IN6P_RFC2292)) { 1754 error = EINVAL; 1755 break; 1756 } 1757 #endif 1758 1759 if (m && m->m_next) { 1760 error = EINVAL; /* XXX */ 1761 break; 1762 } 1763 if (m) { 1764 optbuf = mtod(m, u_char *); 1765 optbuflen = m->m_len; 1766 } else { 1767 optbuf = NULL; 1768 optbuflen = 0; 1769 } 1770 optp = &in6p->in6p_outputopts; 1771 error = ip6_pcbopt(optname, 1772 optbuf, optbuflen, 1773 optp, privileged, uproto); 1774 break; 1775 } 1776 #undef OPTSET 1777 1778 case IPV6_MULTICAST_IF: 1779 case IPV6_MULTICAST_HOPS: 1780 case IPV6_MULTICAST_LOOP: 1781 case IPV6_JOIN_GROUP: 1782 case IPV6_LEAVE_GROUP: 1783 error = ip6_setmoptions(optname, 1784 &in6p->in6p_moptions, m); 1785 break; 1786 1787 case IPV6_PORTRANGE: 1788 optval = *mtod(m, int *); 1789 1790 switch (optval) { 1791 case IPV6_PORTRANGE_DEFAULT: 1792 in6p->in6p_flags &= ~(IN6P_LOWPORT); 1793 in6p->in6p_flags &= ~(IN6P_HIGHPORT); 1794 break; 1795 1796 case IPV6_PORTRANGE_HIGH: 1797 in6p->in6p_flags &= ~(IN6P_LOWPORT); 1798 in6p->in6p_flags |= IN6P_HIGHPORT; 1799 break; 1800 1801 case IPV6_PORTRANGE_LOW: 1802 in6p->in6p_flags &= ~(IN6P_HIGHPORT); 1803 in6p->in6p_flags |= IN6P_LOWPORT; 1804 break; 1805 1806 default: 1807 error = EINVAL; 1808 break; 1809 } 1810 break; 1811 1812 #ifdef IPSEC 1813 case IPV6_IPSEC_POLICY: 1814 { 1815 caddr_t req = NULL; 1816 size_t len = 0; 1817 if (m) { 1818 req = mtod(m, caddr_t); 1819 len = m->m_len; 1820 } 1821 error = ipsec6_set_policy(in6p, optname, req, 1822 len, privileged); 1823 } 1824 break; 1825 #endif /* IPSEC */ 1826 1827 default: 1828 error = ENOPROTOOPT; 1829 break; 1830 } 1831 if (m) 1832 (void)m_free(m); 1833 break; 1834 1835 case PRCO_GETOPT: 1836 switch (optname) { 1837 #ifdef RFC2292 1838 case IPV6_2292PKTOPTIONS: 1839 /* 1840 * RFC3542 (effectively) deprecated the 1841 * semantics of the 2292-style pktoptions. 1842 * Since it was not reliable in nature (i.e., 1843 * applications had to expect the lack of some 1844 * information after all), it would make sense 1845 * to simplify this part by always returning 1846 * empty data. 1847 */ 1848 *mp = m_get(M_WAIT, MT_SOOPTS); 1849 (*mp)->m_len = 0; 1850 break; 1851 #endif 1852 1853 case IPV6_RECVHOPOPTS: 1854 case IPV6_RECVDSTOPTS: 1855 case IPV6_RECVRTHDRDSTOPTS: 1856 case IPV6_UNICAST_HOPS: 1857 case IPV6_RECVPKTINFO: 1858 case IPV6_RECVHOPLIMIT: 1859 case IPV6_RECVRTHDR: 1860 case IPV6_RECVPATHMTU: 1861 1862 case IPV6_FAITH: 1863 case IPV6_V6ONLY: 1864 case IPV6_PORTRANGE: 1865 case IPV6_RECVTCLASS: 1866 switch (optname) { 1867 1868 case IPV6_RECVHOPOPTS: 1869 optval = OPTBIT(IN6P_HOPOPTS); 1870 break; 1871 1872 case IPV6_RECVDSTOPTS: 1873 optval = OPTBIT(IN6P_DSTOPTS); 1874 break; 1875 1876 case IPV6_RECVRTHDRDSTOPTS: 1877 optval = OPTBIT(IN6P_RTHDRDSTOPTS); 1878 break; 1879 1880 case IPV6_UNICAST_HOPS: 1881 optval = in6p->in6p_hops; 1882 break; 1883 1884 case IPV6_RECVPKTINFO: 1885 optval = OPTBIT(IN6P_PKTINFO); 1886 break; 1887 1888 case IPV6_RECVHOPLIMIT: 1889 optval = OPTBIT(IN6P_HOPLIMIT); 1890 break; 1891 1892 case IPV6_RECVRTHDR: 1893 optval = OPTBIT(IN6P_RTHDR); 1894 break; 1895 1896 case IPV6_RECVPATHMTU: 1897 optval = OPTBIT(IN6P_MTU); 1898 break; 1899 1900 case IPV6_FAITH: 1901 optval = OPTBIT(IN6P_FAITH); 1902 break; 1903 1904 case IPV6_V6ONLY: 1905 optval = OPTBIT(IN6P_IPV6_V6ONLY); 1906 break; 1907 1908 case IPV6_PORTRANGE: 1909 { 1910 int flags; 1911 flags = in6p->in6p_flags; 1912 if (flags & IN6P_HIGHPORT) 1913 optval = IPV6_PORTRANGE_HIGH; 1914 else if (flags & IN6P_LOWPORT) 1915 optval = IPV6_PORTRANGE_LOW; 1916 else 1917 optval = 0; 1918 break; 1919 } 1920 case IPV6_RECVTCLASS: 1921 optval = OPTBIT(IN6P_TCLASS); 1922 break; 1923 1924 } 1925 if (error) 1926 break; 1927 *mp = m = m_get(M_WAIT, MT_SOOPTS); 1928 m->m_len = sizeof(int); 1929 *mtod(m, int *) = optval; 1930 break; 1931 1932 case IPV6_PATHMTU: 1933 { 1934 u_long pmtu = 0; 1935 struct ip6_mtuinfo mtuinfo; 1936 struct route_in6 *ro = (struct route_in6 *)&in6p 1937 ->in6p_route; 1938 1939 if (!(so->so_state & SS_ISCONNECTED)) 1940 return (ENOTCONN); 1941 /* 1942 * XXX: we dot not consider the case of source 1943 * routing, or optional information to specify 1944 * the outgoing interface. 1945 */ 1946 error = ip6_getpmtu(ro, NULL, NULL, 1947 &in6p->in6p_faddr, &pmtu, NULL); 1948 if (error) 1949 break; 1950 if (pmtu > IPV6_MAXPACKET) 1951 pmtu = IPV6_MAXPACKET; 1952 1953 memset(&mtuinfo, 0, sizeof(mtuinfo)); 1954 mtuinfo.ip6m_mtu = (u_int32_t)pmtu; 1955 optdata = (void *)&mtuinfo; 1956 optdatalen = sizeof(mtuinfo); 1957 if (optdatalen > MCLBYTES) 1958 return (EMSGSIZE); /* XXX */ 1959 *mp = m = m_get(M_WAIT, MT_SOOPTS); 1960 if (optdatalen > MLEN) 1961 MCLGET(m, M_WAIT); 1962 m->m_len = optdatalen; 1963 memcpy(mtod(m, void *), optdata, optdatalen); 1964 break; 1965 } 1966 1967 #ifdef RFC2292 1968 case IPV6_2292PKTINFO: 1969 case IPV6_2292HOPLIMIT: 1970 case IPV6_2292HOPOPTS: 1971 case IPV6_2292RTHDR: 1972 case IPV6_2292DSTOPTS: 1973 switch (optname) { 1974 case IPV6_2292PKTINFO: 1975 optval = OPTBIT(IN6P_PKTINFO); 1976 break; 1977 case IPV6_2292HOPLIMIT: 1978 optval = OPTBIT(IN6P_HOPLIMIT); 1979 break; 1980 case IPV6_2292HOPOPTS: 1981 optval = OPTBIT(IN6P_HOPOPTS); 1982 break; 1983 case IPV6_2292RTHDR: 1984 optval = OPTBIT(IN6P_RTHDR); 1985 break; 1986 case IPV6_2292DSTOPTS: 1987 optval = OPTBIT(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); 1988 break; 1989 } 1990 *mp = m = m_get(M_WAIT, MT_SOOPTS); 1991 m->m_len = sizeof(int); 1992 *mtod(m, int *) = optval; 1993 break; 1994 #endif 1995 case IPV6_PKTINFO: 1996 case IPV6_HOPOPTS: 1997 case IPV6_RTHDR: 1998 case IPV6_DSTOPTS: 1999 case IPV6_RTHDRDSTOPTS: 2000 case IPV6_NEXTHOP: 2001 case IPV6_OTCLASS: 2002 case IPV6_TCLASS: 2003 case IPV6_DONTFRAG: 2004 case IPV6_USE_MIN_MTU: 2005 error = ip6_getpcbopt(in6p->in6p_outputopts, 2006 optname, mp); 2007 break; 2008 2009 case IPV6_MULTICAST_IF: 2010 case IPV6_MULTICAST_HOPS: 2011 case IPV6_MULTICAST_LOOP: 2012 case IPV6_JOIN_GROUP: 2013 case IPV6_LEAVE_GROUP: 2014 error = ip6_getmoptions(optname, 2015 in6p->in6p_moptions, mp); 2016 break; 2017 2018 #ifdef IPSEC 2019 case IPV6_IPSEC_POLICY: 2020 { 2021 caddr_t req = NULL; 2022 size_t len = 0; 2023 if (m) { 2024 req = mtod(m, caddr_t); 2025 len = m->m_len; 2026 } 2027 error = ipsec6_get_policy(in6p, req, len, mp); 2028 break; 2029 } 2030 #endif /* IPSEC */ 2031 2032 2033 2034 2035 default: 2036 error = ENOPROTOOPT; 2037 break; 2038 } 2039 break; 2040 } 2041 } else { 2042 error = EINVAL; 2043 if (op == PRCO_SETOPT && *mp) 2044 (void)m_free(*mp); 2045 } 2046 return (error); 2047 } 2048 2049 int 2050 ip6_raw_ctloutput(op, so, level, optname, mp) 2051 int op; 2052 struct socket *so; 2053 int level, optname; 2054 struct mbuf **mp; 2055 { 2056 int error = 0, optval, optlen; 2057 const int icmp6off = offsetof(struct icmp6_hdr, icmp6_cksum); 2058 struct in6pcb *in6p = sotoin6pcb(so); 2059 struct mbuf *m = *mp; 2060 2061 optlen = m ? m->m_len : 0; 2062 2063 if (level != IPPROTO_IPV6) { 2064 if (op == PRCO_SETOPT && *mp) 2065 (void)m_free(*mp); 2066 return (EINVAL); 2067 } 2068 2069 switch (optname) { 2070 case IPV6_CHECKSUM: 2071 /* 2072 * For ICMPv6 sockets, no modification allowed for checksum 2073 * offset, permit "no change" values to help existing apps. 2074 * 2075 * XXX RFC3542 says: "An attempt to set IPV6_CHECKSUM 2076 * for an ICMPv6 socket will fail." The current 2077 * behavior does not meet RFC3542. 2078 */ 2079 switch (op) { 2080 case PRCO_SETOPT: 2081 if (optlen != sizeof(int)) { 2082 error = EINVAL; 2083 break; 2084 } 2085 optval = *mtod(m, int *); 2086 if ((optval % 2) != 0) { 2087 /* the API assumes even offset values */ 2088 error = EINVAL; 2089 } else if (so->so_proto->pr_protocol == 2090 IPPROTO_ICMPV6) { 2091 if (optval != icmp6off) 2092 error = EINVAL; 2093 } else 2094 in6p->in6p_cksum = optval; 2095 break; 2096 2097 case PRCO_GETOPT: 2098 if (so->so_proto->pr_protocol == IPPROTO_ICMPV6) 2099 optval = icmp6off; 2100 else 2101 optval = in6p->in6p_cksum; 2102 2103 *mp = m = m_get(M_WAIT, MT_SOOPTS); 2104 m->m_len = sizeof(int); 2105 *mtod(m, int *) = optval; 2106 break; 2107 2108 default: 2109 error = EINVAL; 2110 break; 2111 } 2112 break; 2113 2114 default: 2115 error = ENOPROTOOPT; 2116 break; 2117 } 2118 2119 if (op == PRCO_SETOPT && m) 2120 (void)m_free(m); 2121 2122 return (error); 2123 } 2124 2125 #ifdef RFC2292 2126 /* 2127 * Set up IP6 options in pcb for insertion in output packets or 2128 * specifying behavior of outgoing packets. 2129 */ 2130 static int 2131 ip6_pcbopts(pktopt, m, so) 2132 struct ip6_pktopts **pktopt; 2133 struct mbuf *m; 2134 struct socket *so; 2135 { 2136 struct ip6_pktopts *opt = *pktopt; 2137 int error = 0; 2138 struct lwp *l = curlwp; /* XXX */ 2139 int priv = 0; 2140 2141 /* turn off any old options. */ 2142 if (opt) { 2143 #ifdef DIAGNOSTIC 2144 if (opt->ip6po_pktinfo || opt->ip6po_nexthop || 2145 opt->ip6po_hbh || opt->ip6po_dest1 || opt->ip6po_dest2 || 2146 opt->ip6po_rhinfo.ip6po_rhi_rthdr) 2147 printf("ip6_pcbopts: all specified options are cleared.\n"); 2148 #endif 2149 ip6_clearpktopts(opt, -1); 2150 } else 2151 opt = malloc(sizeof(*opt), M_IP6OPT, M_WAITOK); 2152 *pktopt = NULL; 2153 2154 if (!m || m->m_len == 0) { 2155 /* 2156 * Only turning off any previous options, regardless of 2157 * whether the opt is just created or given. 2158 */ 2159 free(opt, M_IP6OPT); 2160 return (0); 2161 } 2162 2163 /* set options specified by user. */ 2164 if (l && !kauth_authorize_generic(l->l_cred, KAUTH_GENERIC_ISSUSER, 2165 &l->l_acflag)) 2166 priv = 1; 2167 if ((error = ip6_setpktopts(m, opt, NULL, priv, 2168 so->so_proto->pr_protocol)) != 0) { 2169 ip6_clearpktopts(opt, -1); /* XXX: discard all options */ 2170 free(opt, M_IP6OPT); 2171 return (error); 2172 } 2173 *pktopt = opt; 2174 return (0); 2175 } 2176 #endif 2177 2178 /* 2179 * initialize ip6_pktopts. beware that there are non-zero default values in 2180 * the struct. 2181 */ 2182 void 2183 ip6_initpktopts(struct ip6_pktopts *opt) 2184 { 2185 2186 memset(opt, 0, sizeof(*opt)); 2187 opt->ip6po_hlim = -1; /* -1 means default hop limit */ 2188 opt->ip6po_tclass = -1; /* -1 means default traffic class */ 2189 opt->ip6po_minmtu = IP6PO_MINMTU_MCASTONLY; 2190 } 2191 2192 #define sin6tosa(sin6) ((struct sockaddr *)(sin6)) /* XXX */ 2193 static int 2194 ip6_pcbopt(int optname, u_char *buf, int len, struct ip6_pktopts **pktopt, 2195 int priv, int uproto) 2196 { 2197 struct ip6_pktopts *opt; 2198 2199 if (*pktopt == NULL) { 2200 *pktopt = malloc(sizeof(struct ip6_pktopts), M_IP6OPT, 2201 M_WAITOK); 2202 ip6_initpktopts(*pktopt); 2203 } 2204 opt = *pktopt; 2205 2206 return (ip6_setpktopt(optname, buf, len, opt, priv, 1, 0, uproto)); 2207 } 2208 2209 static int 2210 ip6_getpcbopt(struct ip6_pktopts *pktopt, int optname, struct mbuf **mp) 2211 { 2212 void *optdata = NULL; 2213 int optdatalen = 0; 2214 struct ip6_ext *ip6e; 2215 int error = 0; 2216 struct in6_pktinfo null_pktinfo; 2217 int deftclass = 0, on; 2218 int defminmtu = IP6PO_MINMTU_MCASTONLY; 2219 struct mbuf *m; 2220 2221 switch (optname) { 2222 case IPV6_PKTINFO: 2223 if (pktopt && pktopt->ip6po_pktinfo) 2224 optdata = (void *)pktopt->ip6po_pktinfo; 2225 else { 2226 /* XXX: we don't have to do this every time... */ 2227 memset(&null_pktinfo, 0, sizeof(null_pktinfo)); 2228 optdata = (void *)&null_pktinfo; 2229 } 2230 optdatalen = sizeof(struct in6_pktinfo); 2231 break; 2232 case IPV6_OTCLASS: 2233 /* XXX */ 2234 return (EINVAL); 2235 case IPV6_TCLASS: 2236 if (pktopt && pktopt->ip6po_tclass >= 0) 2237 optdata = (void *)&pktopt->ip6po_tclass; 2238 else 2239 optdata = (void *)&deftclass; 2240 optdatalen = sizeof(int); 2241 break; 2242 case IPV6_HOPOPTS: 2243 if (pktopt && pktopt->ip6po_hbh) { 2244 optdata = (void *)pktopt->ip6po_hbh; 2245 ip6e = (struct ip6_ext *)pktopt->ip6po_hbh; 2246 optdatalen = (ip6e->ip6e_len + 1) << 3; 2247 } 2248 break; 2249 case IPV6_RTHDR: 2250 if (pktopt && pktopt->ip6po_rthdr) { 2251 optdata = (void *)pktopt->ip6po_rthdr; 2252 ip6e = (struct ip6_ext *)pktopt->ip6po_rthdr; 2253 optdatalen = (ip6e->ip6e_len + 1) << 3; 2254 } 2255 break; 2256 case IPV6_RTHDRDSTOPTS: 2257 if (pktopt && pktopt->ip6po_dest1) { 2258 optdata = (void *)pktopt->ip6po_dest1; 2259 ip6e = (struct ip6_ext *)pktopt->ip6po_dest1; 2260 optdatalen = (ip6e->ip6e_len + 1) << 3; 2261 } 2262 break; 2263 case IPV6_DSTOPTS: 2264 if (pktopt && pktopt->ip6po_dest2) { 2265 optdata = (void *)pktopt->ip6po_dest2; 2266 ip6e = (struct ip6_ext *)pktopt->ip6po_dest2; 2267 optdatalen = (ip6e->ip6e_len + 1) << 3; 2268 } 2269 break; 2270 case IPV6_NEXTHOP: 2271 if (pktopt && pktopt->ip6po_nexthop) { 2272 optdata = (void *)pktopt->ip6po_nexthop; 2273 optdatalen = pktopt->ip6po_nexthop->sa_len; 2274 } 2275 break; 2276 case IPV6_USE_MIN_MTU: 2277 if (pktopt) 2278 optdata = (void *)&pktopt->ip6po_minmtu; 2279 else 2280 optdata = (void *)&defminmtu; 2281 optdatalen = sizeof(int); 2282 break; 2283 case IPV6_DONTFRAG: 2284 if (pktopt && ((pktopt->ip6po_flags) & IP6PO_DONTFRAG)) 2285 on = 1; 2286 else 2287 on = 0; 2288 optdata = (void *)&on; 2289 optdatalen = sizeof(on); 2290 break; 2291 default: /* should not happen */ 2292 #ifdef DIAGNOSTIC 2293 panic("ip6_getpcbopt: unexpected option\n"); 2294 #endif 2295 return (ENOPROTOOPT); 2296 } 2297 2298 if (optdatalen > MCLBYTES) 2299 return (EMSGSIZE); /* XXX */ 2300 *mp = m = m_get(M_WAIT, MT_SOOPTS); 2301 if (optdatalen > MLEN) 2302 MCLGET(m, M_WAIT); 2303 m->m_len = optdatalen; 2304 if (optdatalen) 2305 memcpy(mtod(m, void *), optdata, optdatalen); 2306 2307 return (error); 2308 } 2309 2310 void 2311 ip6_clearpktopts(struct ip6_pktopts *pktopt, int optname) 2312 { 2313 if (optname == -1 || optname == IPV6_PKTINFO) { 2314 if (pktopt->ip6po_pktinfo) 2315 free(pktopt->ip6po_pktinfo, M_IP6OPT); 2316 pktopt->ip6po_pktinfo = NULL; 2317 } 2318 if (optname == -1 || optname == IPV6_HOPLIMIT) 2319 pktopt->ip6po_hlim = -1; 2320 if (optname == -1 || optname == IPV6_TCLASS) 2321 pktopt->ip6po_tclass = -1; 2322 if (optname == -1 || optname == IPV6_NEXTHOP) { 2323 if (pktopt->ip6po_nextroute.ro_rt) { 2324 RTFREE(pktopt->ip6po_nextroute.ro_rt); 2325 pktopt->ip6po_nextroute.ro_rt = NULL; 2326 } 2327 if (pktopt->ip6po_nexthop) 2328 free(pktopt->ip6po_nexthop, M_IP6OPT); 2329 pktopt->ip6po_nexthop = NULL; 2330 } 2331 if (optname == -1 || optname == IPV6_HOPOPTS) { 2332 if (pktopt->ip6po_hbh) 2333 free(pktopt->ip6po_hbh, M_IP6OPT); 2334 pktopt->ip6po_hbh = NULL; 2335 } 2336 if (optname == -1 || optname == IPV6_RTHDRDSTOPTS) { 2337 if (pktopt->ip6po_dest1) 2338 free(pktopt->ip6po_dest1, M_IP6OPT); 2339 pktopt->ip6po_dest1 = NULL; 2340 } 2341 if (optname == -1 || optname == IPV6_RTHDR) { 2342 if (pktopt->ip6po_rhinfo.ip6po_rhi_rthdr) 2343 free(pktopt->ip6po_rhinfo.ip6po_rhi_rthdr, M_IP6OPT); 2344 pktopt->ip6po_rhinfo.ip6po_rhi_rthdr = NULL; 2345 if (pktopt->ip6po_route.ro_rt) { 2346 RTFREE(pktopt->ip6po_route.ro_rt); 2347 pktopt->ip6po_route.ro_rt = NULL; 2348 } 2349 } 2350 if (optname == -1 || optname == IPV6_DSTOPTS) { 2351 if (pktopt->ip6po_dest2) 2352 free(pktopt->ip6po_dest2, M_IP6OPT); 2353 pktopt->ip6po_dest2 = NULL; 2354 } 2355 } 2356 2357 #define PKTOPT_EXTHDRCPY(type) \ 2358 do { \ 2359 if (src->type) { \ 2360 int hlen = (((struct ip6_ext *)src->type)->ip6e_len + 1) << 3;\ 2361 dst->type = malloc(hlen, M_IP6OPT, canwait); \ 2362 if (dst->type == NULL && canwait == M_NOWAIT) \ 2363 goto bad; \ 2364 memcpy(dst->type, src->type, hlen); \ 2365 } \ 2366 } while (/*CONSTCOND*/ 0) 2367 2368 static int 2369 copypktopts(struct ip6_pktopts *dst, struct ip6_pktopts *src, int canwait) 2370 { 2371 dst->ip6po_hlim = src->ip6po_hlim; 2372 dst->ip6po_tclass = src->ip6po_tclass; 2373 dst->ip6po_flags = src->ip6po_flags; 2374 if (src->ip6po_pktinfo) { 2375 dst->ip6po_pktinfo = malloc(sizeof(*dst->ip6po_pktinfo), 2376 M_IP6OPT, canwait); 2377 if (dst->ip6po_pktinfo == NULL && canwait == M_NOWAIT) 2378 goto bad; 2379 *dst->ip6po_pktinfo = *src->ip6po_pktinfo; 2380 } 2381 if (src->ip6po_nexthop) { 2382 dst->ip6po_nexthop = malloc(src->ip6po_nexthop->sa_len, 2383 M_IP6OPT, canwait); 2384 if (dst->ip6po_nexthop == NULL && canwait == M_NOWAIT) 2385 goto bad; 2386 memcpy(dst->ip6po_nexthop, src->ip6po_nexthop, 2387 src->ip6po_nexthop->sa_len); 2388 } 2389 PKTOPT_EXTHDRCPY(ip6po_hbh); 2390 PKTOPT_EXTHDRCPY(ip6po_dest1); 2391 PKTOPT_EXTHDRCPY(ip6po_dest2); 2392 PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */ 2393 return (0); 2394 2395 bad: 2396 if (dst->ip6po_pktinfo) free(dst->ip6po_pktinfo, M_IP6OPT); 2397 if (dst->ip6po_nexthop) free(dst->ip6po_nexthop, M_IP6OPT); 2398 if (dst->ip6po_hbh) free(dst->ip6po_hbh, M_IP6OPT); 2399 if (dst->ip6po_dest1) free(dst->ip6po_dest1, M_IP6OPT); 2400 if (dst->ip6po_dest2) free(dst->ip6po_dest2, M_IP6OPT); 2401 if (dst->ip6po_rthdr) free(dst->ip6po_rthdr, M_IP6OPT); 2402 2403 return (ENOBUFS); 2404 } 2405 #undef PKTOPT_EXTHDRCPY 2406 2407 struct ip6_pktopts * 2408 ip6_copypktopts(struct ip6_pktopts *src, int canwait) 2409 { 2410 int error; 2411 struct ip6_pktopts *dst; 2412 2413 dst = malloc(sizeof(*dst), M_IP6OPT, canwait); 2414 if (dst == NULL && canwait == M_NOWAIT) 2415 return (NULL); 2416 ip6_initpktopts(dst); 2417 2418 if ((error = copypktopts(dst, src, canwait)) != 0) { 2419 free(dst, M_IP6OPT); 2420 return (NULL); 2421 } 2422 2423 return (dst); 2424 } 2425 2426 void 2427 ip6_freepcbopts(struct ip6_pktopts *pktopt) 2428 { 2429 if (pktopt == NULL) 2430 return; 2431 2432 ip6_clearpktopts(pktopt, -1); 2433 2434 free(pktopt, M_IP6OPT); 2435 } 2436 2437 /* 2438 * Set the IP6 multicast options in response to user setsockopt(). 2439 */ 2440 static int 2441 ip6_setmoptions(optname, im6op, m) 2442 int optname; 2443 struct ip6_moptions **im6op; 2444 struct mbuf *m; 2445 { 2446 int error = 0; 2447 u_int loop, ifindex; 2448 struct ipv6_mreq *mreq; 2449 struct ifnet *ifp; 2450 struct ip6_moptions *im6o = *im6op; 2451 struct route_in6 ro; 2452 struct in6_multi_mship *imm; 2453 struct lwp *l = curlwp; /* XXX */ 2454 2455 if (im6o == NULL) { 2456 /* 2457 * No multicast option buffer attached to the pcb; 2458 * allocate one and initialize to default values. 2459 */ 2460 im6o = (struct ip6_moptions *) 2461 malloc(sizeof(*im6o), M_IPMOPTS, M_WAITOK); 2462 2463 if (im6o == NULL) 2464 return (ENOBUFS); 2465 *im6op = im6o; 2466 im6o->im6o_multicast_ifp = NULL; 2467 im6o->im6o_multicast_hlim = ip6_defmcasthlim; 2468 im6o->im6o_multicast_loop = IPV6_DEFAULT_MULTICAST_LOOP; 2469 LIST_INIT(&im6o->im6o_memberships); 2470 } 2471 2472 switch (optname) { 2473 2474 case IPV6_MULTICAST_IF: 2475 /* 2476 * Select the interface for outgoing multicast packets. 2477 */ 2478 if (m == NULL || m->m_len != sizeof(u_int)) { 2479 error = EINVAL; 2480 break; 2481 } 2482 bcopy(mtod(m, u_int *), &ifindex, sizeof(ifindex)); 2483 if (ifindex != 0) { 2484 if (if_indexlim <= ifindex || !ifindex2ifnet[ifindex]) { 2485 error = ENXIO; /* XXX EINVAL? */ 2486 break; 2487 } 2488 ifp = ifindex2ifnet[ifindex]; 2489 if ((ifp->if_flags & IFF_MULTICAST) == 0) { 2490 error = EADDRNOTAVAIL; 2491 break; 2492 } 2493 } else 2494 ifp = NULL; 2495 im6o->im6o_multicast_ifp = ifp; 2496 break; 2497 2498 case IPV6_MULTICAST_HOPS: 2499 { 2500 /* 2501 * Set the IP6 hoplimit for outgoing multicast packets. 2502 */ 2503 int optval; 2504 if (m == NULL || m->m_len != sizeof(int)) { 2505 error = EINVAL; 2506 break; 2507 } 2508 bcopy(mtod(m, u_int *), &optval, sizeof(optval)); 2509 if (optval < -1 || optval >= 256) 2510 error = EINVAL; 2511 else if (optval == -1) 2512 im6o->im6o_multicast_hlim = ip6_defmcasthlim; 2513 else 2514 im6o->im6o_multicast_hlim = optval; 2515 break; 2516 } 2517 2518 case IPV6_MULTICAST_LOOP: 2519 /* 2520 * Set the loopback flag for outgoing multicast packets. 2521 * Must be zero or one. 2522 */ 2523 if (m == NULL || m->m_len != sizeof(u_int)) { 2524 error = EINVAL; 2525 break; 2526 } 2527 bcopy(mtod(m, u_int *), &loop, sizeof(loop)); 2528 if (loop > 1) { 2529 error = EINVAL; 2530 break; 2531 } 2532 im6o->im6o_multicast_loop = loop; 2533 break; 2534 2535 case IPV6_JOIN_GROUP: 2536 /* 2537 * Add a multicast group membership. 2538 * Group must be a valid IP6 multicast address. 2539 */ 2540 if (m == NULL || m->m_len != sizeof(struct ipv6_mreq)) { 2541 error = EINVAL; 2542 break; 2543 } 2544 mreq = mtod(m, struct ipv6_mreq *); 2545 if (IN6_IS_ADDR_UNSPECIFIED(&mreq->ipv6mr_multiaddr)) { 2546 /* 2547 * We use the unspecified address to specify to accept 2548 * all multicast addresses. Only super user is allowed 2549 * to do this. 2550 */ 2551 if (kauth_authorize_generic(l->l_cred, 2552 KAUTH_GENERIC_ISSUSER, &l->l_acflag)) 2553 { 2554 error = EACCES; 2555 break; 2556 } 2557 } else if (!IN6_IS_ADDR_MULTICAST(&mreq->ipv6mr_multiaddr)) { 2558 error = EINVAL; 2559 break; 2560 } 2561 2562 /* 2563 * If no interface was explicitly specified, choose an 2564 * appropriate one according to the given multicast address. 2565 */ 2566 if (mreq->ipv6mr_interface == 0) { 2567 struct sockaddr_in6 *dst; 2568 2569 /* 2570 * Look up the routing table for the 2571 * address, and choose the outgoing interface. 2572 * XXX: is it a good approach? 2573 */ 2574 ro.ro_rt = NULL; 2575 dst = (struct sockaddr_in6 *)&ro.ro_dst; 2576 bzero(dst, sizeof(*dst)); 2577 dst->sin6_family = AF_INET6; 2578 dst->sin6_len = sizeof(*dst); 2579 dst->sin6_addr = mreq->ipv6mr_multiaddr; 2580 rtalloc((struct route *)&ro); 2581 if (ro.ro_rt == NULL) { 2582 error = EADDRNOTAVAIL; 2583 break; 2584 } 2585 ifp = ro.ro_rt->rt_ifp; 2586 rtfree(ro.ro_rt); 2587 } else { 2588 /* 2589 * If the interface is specified, validate it. 2590 */ 2591 if (if_indexlim <= mreq->ipv6mr_interface || 2592 !ifindex2ifnet[mreq->ipv6mr_interface]) { 2593 error = ENXIO; /* XXX EINVAL? */ 2594 break; 2595 } 2596 ifp = ifindex2ifnet[mreq->ipv6mr_interface]; 2597 } 2598 2599 /* 2600 * See if we found an interface, and confirm that it 2601 * supports multicast 2602 */ 2603 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) { 2604 error = EADDRNOTAVAIL; 2605 break; 2606 } 2607 2608 if (in6_setscope(&mreq->ipv6mr_multiaddr, ifp, NULL)) { 2609 error = EADDRNOTAVAIL; /* XXX: should not happen */ 2610 break; 2611 } 2612 2613 /* 2614 * See if the membership already exists. 2615 */ 2616 for (imm = im6o->im6o_memberships.lh_first; 2617 imm != NULL; imm = imm->i6mm_chain.le_next) 2618 if (imm->i6mm_maddr->in6m_ifp == ifp && 2619 IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr, 2620 &mreq->ipv6mr_multiaddr)) 2621 break; 2622 if (imm != NULL) { 2623 error = EADDRINUSE; 2624 break; 2625 } 2626 /* 2627 * Everything looks good; add a new record to the multicast 2628 * address list for the given interface. 2629 */ 2630 imm = in6_joingroup(ifp, &mreq->ipv6mr_multiaddr, &error, 0); 2631 if (imm == NULL) 2632 break; 2633 LIST_INSERT_HEAD(&im6o->im6o_memberships, imm, i6mm_chain); 2634 break; 2635 2636 case IPV6_LEAVE_GROUP: 2637 /* 2638 * Drop a multicast group membership. 2639 * Group must be a valid IP6 multicast address. 2640 */ 2641 if (m == NULL || m->m_len != sizeof(struct ipv6_mreq)) { 2642 error = EINVAL; 2643 break; 2644 } 2645 mreq = mtod(m, struct ipv6_mreq *); 2646 2647 /* 2648 * If an interface address was specified, get a pointer 2649 * to its ifnet structure. 2650 */ 2651 if (mreq->ipv6mr_interface != 0) { 2652 if (if_indexlim <= mreq->ipv6mr_interface || 2653 !ifindex2ifnet[mreq->ipv6mr_interface]) { 2654 error = ENXIO; /* XXX EINVAL? */ 2655 break; 2656 } 2657 ifp = ifindex2ifnet[mreq->ipv6mr_interface]; 2658 } else 2659 ifp = NULL; 2660 2661 /* Fill in the scope zone ID */ 2662 if (ifp) { 2663 if (in6_setscope(&mreq->ipv6mr_multiaddr, ifp, NULL)) { 2664 /* XXX: should not happen */ 2665 error = EADDRNOTAVAIL; 2666 break; 2667 } 2668 } else if (mreq->ipv6mr_interface != 0) { 2669 /* 2670 * XXX: This case would happens when the (positive) 2671 * index is in the valid range, but the corresponding 2672 * interface has been detached dynamically. The above 2673 * check probably avoids such case to happen here, but 2674 * we check it explicitly for safety. 2675 */ 2676 error = EADDRNOTAVAIL; 2677 break; 2678 } else { /* ipv6mr_interface == 0 */ 2679 struct sockaddr_in6 sa6_mc; 2680 2681 /* 2682 * The API spec says as follows: 2683 * If the interface index is specified as 0, the 2684 * system may choose a multicast group membership to 2685 * drop by matching the multicast address only. 2686 * On the other hand, we cannot disambiguate the scope 2687 * zone unless an interface is provided. Thus, we 2688 * check if there's ambiguity with the default scope 2689 * zone as the last resort. 2690 */ 2691 bzero(&sa6_mc, sizeof(sa6_mc)); 2692 sa6_mc.sin6_family = AF_INET6; 2693 sa6_mc.sin6_len = sizeof(sa6_mc); 2694 sa6_mc.sin6_addr = mreq->ipv6mr_multiaddr; 2695 error = sa6_embedscope(&sa6_mc, ip6_use_defzone); 2696 if (error != 0) 2697 break; 2698 mreq->ipv6mr_multiaddr = sa6_mc.sin6_addr; 2699 } 2700 2701 /* 2702 * Find the membership in the membership list. 2703 */ 2704 for (imm = im6o->im6o_memberships.lh_first; 2705 imm != NULL; imm = imm->i6mm_chain.le_next) { 2706 if ((ifp == NULL || imm->i6mm_maddr->in6m_ifp == ifp) && 2707 IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr, 2708 &mreq->ipv6mr_multiaddr)) 2709 break; 2710 } 2711 if (imm == NULL) { 2712 /* Unable to resolve interface */ 2713 error = EADDRNOTAVAIL; 2714 break; 2715 } 2716 /* 2717 * Give up the multicast address record to which the 2718 * membership points. 2719 */ 2720 LIST_REMOVE(imm, i6mm_chain); 2721 in6_leavegroup(imm); 2722 break; 2723 2724 default: 2725 error = EOPNOTSUPP; 2726 break; 2727 } 2728 2729 /* 2730 * If all options have default values, no need to keep the mbuf. 2731 */ 2732 if (im6o->im6o_multicast_ifp == NULL && 2733 im6o->im6o_multicast_hlim == ip6_defmcasthlim && 2734 im6o->im6o_multicast_loop == IPV6_DEFAULT_MULTICAST_LOOP && 2735 im6o->im6o_memberships.lh_first == NULL) { 2736 free(*im6op, M_IPMOPTS); 2737 *im6op = NULL; 2738 } 2739 2740 return (error); 2741 } 2742 2743 /* 2744 * Return the IP6 multicast options in response to user getsockopt(). 2745 */ 2746 static int 2747 ip6_getmoptions(optname, im6o, mp) 2748 int optname; 2749 struct ip6_moptions *im6o; 2750 struct mbuf **mp; 2751 { 2752 u_int *hlim, *loop, *ifindex; 2753 2754 *mp = m_get(M_WAIT, MT_SOOPTS); 2755 2756 switch (optname) { 2757 2758 case IPV6_MULTICAST_IF: 2759 ifindex = mtod(*mp, u_int *); 2760 (*mp)->m_len = sizeof(u_int); 2761 if (im6o == NULL || im6o->im6o_multicast_ifp == NULL) 2762 *ifindex = 0; 2763 else 2764 *ifindex = im6o->im6o_multicast_ifp->if_index; 2765 return (0); 2766 2767 case IPV6_MULTICAST_HOPS: 2768 hlim = mtod(*mp, u_int *); 2769 (*mp)->m_len = sizeof(u_int); 2770 if (im6o == NULL) 2771 *hlim = ip6_defmcasthlim; 2772 else 2773 *hlim = im6o->im6o_multicast_hlim; 2774 return (0); 2775 2776 case IPV6_MULTICAST_LOOP: 2777 loop = mtod(*mp, u_int *); 2778 (*mp)->m_len = sizeof(u_int); 2779 if (im6o == NULL) 2780 *loop = ip6_defmcasthlim; 2781 else 2782 *loop = im6o->im6o_multicast_loop; 2783 return (0); 2784 2785 default: 2786 return (EOPNOTSUPP); 2787 } 2788 } 2789 2790 /* 2791 * Discard the IP6 multicast options. 2792 */ 2793 void 2794 ip6_freemoptions(im6o) 2795 struct ip6_moptions *im6o; 2796 { 2797 struct in6_multi_mship *imm; 2798 2799 if (im6o == NULL) 2800 return; 2801 2802 while ((imm = im6o->im6o_memberships.lh_first) != NULL) { 2803 LIST_REMOVE(imm, i6mm_chain); 2804 in6_leavegroup(imm); 2805 } 2806 free(im6o, M_IPMOPTS); 2807 } 2808 2809 /* 2810 * Set IPv6 outgoing packet options based on advanced API. 2811 */ 2812 int 2813 ip6_setpktopts(control, opt, stickyopt, priv, uproto) 2814 struct mbuf *control; 2815 struct ip6_pktopts *opt, *stickyopt; 2816 int priv, uproto; 2817 { 2818 struct cmsghdr *cm = 0; 2819 2820 if (control == NULL || opt == NULL) 2821 return (EINVAL); 2822 2823 ip6_initpktopts(opt); 2824 if (stickyopt) { 2825 int error; 2826 2827 /* 2828 * If stickyopt is provided, make a local copy of the options 2829 * for this particular packet, then override them by ancillary 2830 * objects. 2831 * XXX: copypktopts() does not copy the cached route to a next 2832 * hop (if any). This is not very good in terms of efficiency, 2833 * but we can allow this since this option should be rarely 2834 * used. 2835 */ 2836 if ((error = copypktopts(opt, stickyopt, M_NOWAIT)) != 0) 2837 return (error); 2838 } 2839 2840 /* 2841 * XXX: Currently, we assume all the optional information is stored 2842 * in a single mbuf. 2843 */ 2844 if (control->m_next) 2845 return (EINVAL); 2846 2847 for (; control->m_len; control->m_data += CMSG_ALIGN(cm->cmsg_len), 2848 control->m_len -= CMSG_ALIGN(cm->cmsg_len)) { 2849 int error; 2850 2851 if (control->m_len < CMSG_LEN(0)) 2852 return (EINVAL); 2853 2854 cm = mtod(control, struct cmsghdr *); 2855 if (cm->cmsg_len == 0 || cm->cmsg_len > control->m_len) 2856 return (EINVAL); 2857 if (cm->cmsg_level != IPPROTO_IPV6) 2858 continue; 2859 2860 error = ip6_setpktopt(cm->cmsg_type, CMSG_DATA(cm), 2861 cm->cmsg_len - CMSG_LEN(0), opt, priv, 0, 1, uproto); 2862 if (error) 2863 return (error); 2864 } 2865 2866 return (0); 2867 } 2868 2869 /* 2870 * Set a particular packet option, as a sticky option or an ancillary data 2871 * item. "len" can be 0 only when it's a sticky option. 2872 * We have 4 cases of combination of "sticky" and "cmsg": 2873 * "sticky=0, cmsg=0": impossible 2874 * "sticky=0, cmsg=1": RFC2292 or RFC3542 ancillary data 2875 * "sticky=1, cmsg=0": RFC3542 socket option 2876 * "sticky=1, cmsg=1": RFC2292 socket option 2877 */ 2878 static int 2879 ip6_setpktopt(int optname, u_char *buf, int len, struct ip6_pktopts *opt, 2880 int priv, int sticky, int cmsg, int uproto) 2881 { 2882 int minmtupolicy; 2883 2884 if (!sticky && !cmsg) { 2885 #ifdef DIAGNOSTIC 2886 printf("ip6_setpktopt: impossible case\n"); 2887 #endif 2888 return (EINVAL); 2889 } 2890 2891 /* 2892 * IPV6_2292xxx is for backward compatibility to RFC2292, and should 2893 * not be specified in the context of RFC3542. Conversely, 2894 * RFC3542 types should not be specified in the context of RFC2292. 2895 */ 2896 if (!cmsg) { 2897 switch (optname) { 2898 case IPV6_2292PKTINFO: 2899 case IPV6_2292HOPLIMIT: 2900 case IPV6_2292NEXTHOP: 2901 case IPV6_2292HOPOPTS: 2902 case IPV6_2292DSTOPTS: 2903 case IPV6_2292RTHDR: 2904 case IPV6_2292PKTOPTIONS: 2905 return (ENOPROTOOPT); 2906 } 2907 } 2908 if (sticky && cmsg) { 2909 switch (optname) { 2910 case IPV6_PKTINFO: 2911 case IPV6_HOPLIMIT: 2912 case IPV6_NEXTHOP: 2913 case IPV6_HOPOPTS: 2914 case IPV6_DSTOPTS: 2915 case IPV6_RTHDRDSTOPTS: 2916 case IPV6_RTHDR: 2917 case IPV6_USE_MIN_MTU: 2918 case IPV6_DONTFRAG: 2919 case IPV6_OTCLASS: 2920 case IPV6_TCLASS: 2921 return (ENOPROTOOPT); 2922 } 2923 } 2924 2925 switch (optname) { 2926 #ifdef RFC2292 2927 case IPV6_2292PKTINFO: 2928 #endif 2929 case IPV6_PKTINFO: 2930 { 2931 struct ifnet *ifp = NULL; 2932 struct in6_pktinfo *pktinfo; 2933 2934 if (len != sizeof(struct in6_pktinfo)) 2935 return (EINVAL); 2936 2937 pktinfo = (struct in6_pktinfo *)buf; 2938 2939 /* 2940 * An application can clear any sticky IPV6_PKTINFO option by 2941 * doing a "regular" setsockopt with ipi6_addr being 2942 * in6addr_any and ipi6_ifindex being zero. 2943 * [RFC 3542, Section 6] 2944 */ 2945 if (optname == IPV6_PKTINFO && opt->ip6po_pktinfo && 2946 pktinfo->ipi6_ifindex == 0 && 2947 IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) { 2948 ip6_clearpktopts(opt, optname); 2949 break; 2950 } 2951 2952 if (uproto == IPPROTO_TCP && optname == IPV6_PKTINFO && 2953 sticky && !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) { 2954 return (EINVAL); 2955 } 2956 2957 /* validate the interface index if specified. */ 2958 if (pktinfo->ipi6_ifindex >= if_indexlim) { 2959 return (ENXIO); 2960 } 2961 if (pktinfo->ipi6_ifindex) { 2962 ifp = ifindex2ifnet[pktinfo->ipi6_ifindex]; 2963 if (ifp == NULL) 2964 return (ENXIO); 2965 } 2966 2967 /* 2968 * We store the address anyway, and let in6_selectsrc() 2969 * validate the specified address. This is because ipi6_addr 2970 * may not have enough information about its scope zone, and 2971 * we may need additional information (such as outgoing 2972 * interface or the scope zone of a destination address) to 2973 * disambiguate the scope. 2974 * XXX: the delay of the validation may confuse the 2975 * application when it is used as a sticky option. 2976 */ 2977 if (opt->ip6po_pktinfo == NULL) { 2978 opt->ip6po_pktinfo = malloc(sizeof(*pktinfo), 2979 M_IP6OPT, M_NOWAIT); 2980 if (opt->ip6po_pktinfo == NULL) 2981 return (ENOBUFS); 2982 } 2983 memcpy(opt->ip6po_pktinfo, pktinfo, sizeof(*pktinfo)); 2984 break; 2985 } 2986 2987 #ifdef RFC2292 2988 case IPV6_2292HOPLIMIT: 2989 #endif 2990 case IPV6_HOPLIMIT: 2991 { 2992 int *hlimp; 2993 2994 /* 2995 * RFC 3542 deprecated the usage of sticky IPV6_HOPLIMIT 2996 * to simplify the ordering among hoplimit options. 2997 */ 2998 if (optname == IPV6_HOPLIMIT && sticky) 2999 return (ENOPROTOOPT); 3000 3001 if (len != sizeof(int)) 3002 return (EINVAL); 3003 hlimp = (int *)buf; 3004 if (*hlimp < -1 || *hlimp > 255) 3005 return (EINVAL); 3006 3007 opt->ip6po_hlim = *hlimp; 3008 break; 3009 } 3010 3011 case IPV6_OTCLASS: 3012 if (len != sizeof(u_int8_t)) 3013 return (EINVAL); 3014 3015 opt->ip6po_tclass = *(u_int8_t *)buf; 3016 break; 3017 3018 case IPV6_TCLASS: 3019 { 3020 int tclass; 3021 3022 if (len != sizeof(int)) 3023 return (EINVAL); 3024 tclass = *(int *)buf; 3025 if (tclass < -1 || tclass > 255) 3026 return (EINVAL); 3027 3028 opt->ip6po_tclass = tclass; 3029 break; 3030 } 3031 3032 #ifdef RFC2292 3033 case IPV6_2292NEXTHOP: 3034 #endif 3035 case IPV6_NEXTHOP: 3036 if (!priv) 3037 return (EPERM); 3038 3039 if (len == 0) { /* just remove the option */ 3040 ip6_clearpktopts(opt, IPV6_NEXTHOP); 3041 break; 3042 } 3043 3044 /* check if cmsg_len is large enough for sa_len */ 3045 if (len < sizeof(struct sockaddr) || len < *buf) 3046 return (EINVAL); 3047 3048 switch (((struct sockaddr *)buf)->sa_family) { 3049 case AF_INET6: 3050 { 3051 struct sockaddr_in6 *sa6 = (struct sockaddr_in6 *)buf; 3052 int error; 3053 3054 if (sa6->sin6_len != sizeof(struct sockaddr_in6)) 3055 return (EINVAL); 3056 3057 if (IN6_IS_ADDR_UNSPECIFIED(&sa6->sin6_addr) || 3058 IN6_IS_ADDR_MULTICAST(&sa6->sin6_addr)) { 3059 return (EINVAL); 3060 } 3061 if ((error = sa6_embedscope(sa6, ip6_use_defzone)) 3062 != 0) { 3063 return (error); 3064 } 3065 break; 3066 } 3067 case AF_LINK: /* eventually be supported? */ 3068 default: 3069 return (EAFNOSUPPORT); 3070 } 3071 3072 /* turn off the previous option, then set the new option. */ 3073 ip6_clearpktopts(opt, IPV6_NEXTHOP); 3074 opt->ip6po_nexthop = malloc(*buf, M_IP6OPT, M_NOWAIT); 3075 if (opt->ip6po_nexthop == NULL) 3076 return (ENOBUFS); 3077 memcpy(opt->ip6po_nexthop, buf, *buf); 3078 break; 3079 3080 #ifdef RFC2292 3081 case IPV6_2292HOPOPTS: 3082 #endif 3083 case IPV6_HOPOPTS: 3084 { 3085 struct ip6_hbh *hbh; 3086 int hbhlen; 3087 3088 /* 3089 * XXX: We don't allow a non-privileged user to set ANY HbH 3090 * options, since per-option restriction has too much 3091 * overhead. 3092 */ 3093 if (!priv) 3094 return (EPERM); 3095 3096 if (len == 0) { 3097 ip6_clearpktopts(opt, IPV6_HOPOPTS); 3098 break; /* just remove the option */ 3099 } 3100 3101 /* message length validation */ 3102 if (len < sizeof(struct ip6_hbh)) 3103 return (EINVAL); 3104 hbh = (struct ip6_hbh *)buf; 3105 hbhlen = (hbh->ip6h_len + 1) << 3; 3106 if (len != hbhlen) 3107 return (EINVAL); 3108 3109 /* turn off the previous option, then set the new option. */ 3110 ip6_clearpktopts(opt, IPV6_HOPOPTS); 3111 opt->ip6po_hbh = malloc(hbhlen, M_IP6OPT, M_NOWAIT); 3112 if (opt->ip6po_hbh == NULL) 3113 return (ENOBUFS); 3114 memcpy(opt->ip6po_hbh, hbh, hbhlen); 3115 3116 break; 3117 } 3118 3119 #ifdef RFC2292 3120 case IPV6_2292DSTOPTS: 3121 #endif 3122 case IPV6_DSTOPTS: 3123 case IPV6_RTHDRDSTOPTS: 3124 { 3125 struct ip6_dest *dest, **newdest = NULL; 3126 int destlen; 3127 3128 if (!priv) /* XXX: see the comment for IPV6_HOPOPTS */ 3129 return (EPERM); 3130 3131 if (len == 0) { 3132 ip6_clearpktopts(opt, optname); 3133 break; /* just remove the option */ 3134 } 3135 3136 /* message length validation */ 3137 if (len < sizeof(struct ip6_dest)) 3138 return (EINVAL); 3139 dest = (struct ip6_dest *)buf; 3140 destlen = (dest->ip6d_len + 1) << 3; 3141 if (len != destlen) 3142 return (EINVAL); 3143 /* 3144 * Determine the position that the destination options header 3145 * should be inserted; before or after the routing header. 3146 */ 3147 switch (optname) { 3148 case IPV6_2292DSTOPTS: 3149 /* 3150 * The old advanced API is ambiguous on this point. 3151 * Our approach is to determine the position based 3152 * according to the existence of a routing header. 3153 * Note, however, that this depends on the order of the 3154 * extension headers in the ancillary data; the 1st 3155 * part of the destination options header must appear 3156 * before the routing header in the ancillary data, 3157 * too. 3158 * RFC3542 solved the ambiguity by introducing 3159 * separate ancillary data or option types. 3160 */ 3161 if (opt->ip6po_rthdr == NULL) 3162 newdest = &opt->ip6po_dest1; 3163 else 3164 newdest = &opt->ip6po_dest2; 3165 break; 3166 case IPV6_RTHDRDSTOPTS: 3167 newdest = &opt->ip6po_dest1; 3168 break; 3169 case IPV6_DSTOPTS: 3170 newdest = &opt->ip6po_dest2; 3171 break; 3172 } 3173 3174 /* turn off the previous option, then set the new option. */ 3175 ip6_clearpktopts(opt, optname); 3176 *newdest = malloc(destlen, M_IP6OPT, M_NOWAIT); 3177 if (*newdest == NULL) 3178 return (ENOBUFS); 3179 memcpy(*newdest, dest, destlen); 3180 3181 break; 3182 } 3183 3184 #ifdef RFC2292 3185 case IPV6_2292RTHDR: 3186 #endif 3187 case IPV6_RTHDR: 3188 { 3189 struct ip6_rthdr *rth; 3190 int rthlen; 3191 3192 if (len == 0) { 3193 ip6_clearpktopts(opt, IPV6_RTHDR); 3194 break; /* just remove the option */ 3195 } 3196 3197 /* message length validation */ 3198 if (len < sizeof(struct ip6_rthdr)) 3199 return (EINVAL); 3200 rth = (struct ip6_rthdr *)buf; 3201 rthlen = (rth->ip6r_len + 1) << 3; 3202 if (len != rthlen) 3203 return (EINVAL); 3204 switch (rth->ip6r_type) { 3205 case IPV6_RTHDR_TYPE_0: 3206 if (rth->ip6r_len == 0) /* must contain one addr */ 3207 return (EINVAL); 3208 if (rth->ip6r_len % 2) /* length must be even */ 3209 return (EINVAL); 3210 if (rth->ip6r_len / 2 != rth->ip6r_segleft) 3211 return (EINVAL); 3212 break; 3213 default: 3214 return (EINVAL); /* not supported */ 3215 } 3216 /* turn off the previous option */ 3217 ip6_clearpktopts(opt, IPV6_RTHDR); 3218 opt->ip6po_rthdr = malloc(rthlen, M_IP6OPT, M_NOWAIT); 3219 if (opt->ip6po_rthdr == NULL) 3220 return (ENOBUFS); 3221 memcpy(opt->ip6po_rthdr, rth, rthlen); 3222 break; 3223 } 3224 3225 case IPV6_USE_MIN_MTU: 3226 if (len != sizeof(int)) 3227 return (EINVAL); 3228 minmtupolicy = *(int *)buf; 3229 if (minmtupolicy != IP6PO_MINMTU_MCASTONLY && 3230 minmtupolicy != IP6PO_MINMTU_DISABLE && 3231 minmtupolicy != IP6PO_MINMTU_ALL) { 3232 return (EINVAL); 3233 } 3234 opt->ip6po_minmtu = minmtupolicy; 3235 break; 3236 3237 case IPV6_DONTFRAG: 3238 if (len != sizeof(int)) 3239 return (EINVAL); 3240 3241 if (uproto == IPPROTO_TCP || *(int *)buf == 0) { 3242 /* 3243 * we ignore this option for TCP sockets. 3244 * (RFC3542 leaves this case unspecified.) 3245 */ 3246 opt->ip6po_flags &= ~IP6PO_DONTFRAG; 3247 } else 3248 opt->ip6po_flags |= IP6PO_DONTFRAG; 3249 break; 3250 3251 default: 3252 return (ENOPROTOOPT); 3253 } /* end of switch */ 3254 3255 return (0); 3256 } 3257 3258 /* 3259 * Routine called from ip6_output() to loop back a copy of an IP6 multicast 3260 * packet to the input queue of a specified interface. Note that this 3261 * calls the output routine of the loopback "driver", but with an interface 3262 * pointer that might NOT be lo0ifp -- easier than replicating that code here. 3263 */ 3264 void 3265 ip6_mloopback(ifp, m, dst) 3266 struct ifnet *ifp; 3267 struct mbuf *m; 3268 struct sockaddr_in6 *dst; 3269 { 3270 struct mbuf *copym; 3271 struct ip6_hdr *ip6; 3272 3273 copym = m_copy(m, 0, M_COPYALL); 3274 if (copym == NULL) 3275 return; 3276 3277 /* 3278 * Make sure to deep-copy IPv6 header portion in case the data 3279 * is in an mbuf cluster, so that we can safely override the IPv6 3280 * header portion later. 3281 */ 3282 if ((copym->m_flags & M_EXT) != 0 || 3283 copym->m_len < sizeof(struct ip6_hdr)) { 3284 copym = m_pullup(copym, sizeof(struct ip6_hdr)); 3285 if (copym == NULL) 3286 return; 3287 } 3288 3289 #ifdef DIAGNOSTIC 3290 if (copym->m_len < sizeof(*ip6)) { 3291 m_freem(copym); 3292 return; 3293 } 3294 #endif 3295 3296 ip6 = mtod(copym, struct ip6_hdr *); 3297 /* 3298 * clear embedded scope identifiers if necessary. 3299 * in6_clearscope will touch the addresses only when necessary. 3300 */ 3301 in6_clearscope(&ip6->ip6_src); 3302 in6_clearscope(&ip6->ip6_dst); 3303 3304 (void)looutput(ifp, copym, (struct sockaddr *)dst, NULL); 3305 } 3306 3307 /* 3308 * Chop IPv6 header off from the payload. 3309 */ 3310 static int 3311 ip6_splithdr(m, exthdrs) 3312 struct mbuf *m; 3313 struct ip6_exthdrs *exthdrs; 3314 { 3315 struct mbuf *mh; 3316 struct ip6_hdr *ip6; 3317 3318 ip6 = mtod(m, struct ip6_hdr *); 3319 if (m->m_len > sizeof(*ip6)) { 3320 MGETHDR(mh, M_DONTWAIT, MT_HEADER); 3321 if (mh == 0) { 3322 m_freem(m); 3323 return ENOBUFS; 3324 } 3325 M_MOVE_PKTHDR(mh, m); 3326 MH_ALIGN(mh, sizeof(*ip6)); 3327 m->m_len -= sizeof(*ip6); 3328 m->m_data += sizeof(*ip6); 3329 mh->m_next = m; 3330 m = mh; 3331 m->m_len = sizeof(*ip6); 3332 bcopy((caddr_t)ip6, mtod(m, caddr_t), sizeof(*ip6)); 3333 } 3334 exthdrs->ip6e_ip6 = m; 3335 return 0; 3336 } 3337 3338 /* 3339 * Compute IPv6 extension header length. 3340 */ 3341 int 3342 ip6_optlen(in6p) 3343 struct in6pcb *in6p; 3344 { 3345 int len; 3346 3347 if (!in6p->in6p_outputopts) 3348 return 0; 3349 3350 len = 0; 3351 #define elen(x) \ 3352 (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0) 3353 3354 len += elen(in6p->in6p_outputopts->ip6po_hbh); 3355 len += elen(in6p->in6p_outputopts->ip6po_dest1); 3356 len += elen(in6p->in6p_outputopts->ip6po_rthdr); 3357 len += elen(in6p->in6p_outputopts->ip6po_dest2); 3358 return len; 3359 #undef elen 3360 } 3361