1 /* $NetBSD: ip6_output.c,v 1.140 2011/04/25 22:20:59 yamt 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 1.140 2011/04/25 22:20:59 yamt 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/in6_offload.h> 95 #include <netinet6/ip6_var.h> 96 #include <netinet6/ip6_private.h> 97 #include <netinet6/in6_pcb.h> 98 #include <netinet6/nd6.h> 99 #include <netinet6/ip6protosw.h> 100 #include <netinet6/scope6_var.h> 101 102 #ifdef IPSEC 103 #include <netinet6/ipsec.h> 104 #include <netinet6/ipsec_private.h> 105 #include <netkey/key.h> 106 #endif /* IPSEC */ 107 108 #ifdef FAST_IPSEC 109 #include <netipsec/ipsec.h> 110 #include <netipsec/ipsec6.h> 111 #include <netipsec/key.h> 112 #include <netipsec/xform.h> 113 #endif 114 115 116 #include <net/net_osdep.h> 117 118 #ifdef PFIL_HOOKS 119 extern struct pfil_head inet6_pfil_hook; /* XXX */ 120 #endif 121 122 struct ip6_exthdrs { 123 struct mbuf *ip6e_ip6; 124 struct mbuf *ip6e_hbh; 125 struct mbuf *ip6e_dest1; 126 struct mbuf *ip6e_rthdr; 127 struct mbuf *ip6e_dest2; 128 }; 129 130 static int ip6_pcbopt(int, u_char *, int, struct ip6_pktopts **, 131 kauth_cred_t, int); 132 static int ip6_getpcbopt(struct ip6_pktopts *, int, struct sockopt *); 133 static int ip6_setpktopt(int, u_char *, int, struct ip6_pktopts *, kauth_cred_t, 134 int, int, int); 135 static int ip6_setmoptions(const struct sockopt *, struct ip6_moptions **); 136 static int ip6_getmoptions(struct sockopt *, struct ip6_moptions *); 137 static int ip6_copyexthdr(struct mbuf **, void *, int); 138 static int ip6_insertfraghdr(struct mbuf *, struct mbuf *, int, 139 struct ip6_frag **); 140 static int ip6_insert_jumboopt(struct ip6_exthdrs *, u_int32_t); 141 static int ip6_splithdr(struct mbuf *, struct ip6_exthdrs *); 142 static int ip6_getpmtu(struct route *, struct route *, struct ifnet *, 143 const struct in6_addr *, u_long *, int *); 144 static int copypktopts(struct ip6_pktopts *, struct ip6_pktopts *, int); 145 146 #ifdef RFC2292 147 static int ip6_pcbopts(struct ip6_pktopts **, struct socket *, struct sockopt *); 148 #endif 149 150 /* 151 * IP6 output. The packet in mbuf chain m contains a skeletal IP6 152 * header (with pri, len, nxt, hlim, src, dst). 153 * This function may modify ver and hlim only. 154 * The mbuf chain containing the packet will be freed. 155 * The mbuf opt, if present, will not be freed. 156 * 157 * type of "mtu": rt_rmx.rmx_mtu is u_long, ifnet.ifr_mtu is int, and 158 * nd_ifinfo.linkmtu is u_int32_t. so we use u_long to hold largest one, 159 * which is rt_rmx.rmx_mtu. 160 */ 161 int 162 ip6_output( 163 struct mbuf *m0, 164 struct ip6_pktopts *opt, 165 struct route *ro, 166 int flags, 167 struct ip6_moptions *im6o, 168 struct socket *so, 169 struct ifnet **ifpp /* XXX: just for statistics */ 170 ) 171 { 172 struct ip6_hdr *ip6, *mhip6; 173 struct ifnet *ifp, *origifp; 174 struct mbuf *m = m0; 175 int hlen, tlen, len, off; 176 bool tso; 177 struct route ip6route; 178 struct rtentry *rt = NULL; 179 const struct sockaddr_in6 *dst = NULL; 180 struct sockaddr_in6 src_sa, dst_sa; 181 int error = 0; 182 struct in6_ifaddr *ia = NULL; 183 u_long mtu; 184 int alwaysfrag, dontfrag; 185 u_int32_t optlen = 0, plen = 0, unfragpartlen = 0; 186 struct ip6_exthdrs exthdrs; 187 struct in6_addr finaldst, src0, dst0; 188 u_int32_t zone; 189 struct route *ro_pmtu = NULL; 190 int hdrsplit = 0; 191 int needipsec = 0; 192 #ifdef IPSEC 193 int needipsectun = 0; 194 struct secpolicy *sp = NULL; 195 196 ip6 = mtod(m, struct ip6_hdr *); 197 #endif /* IPSEC */ 198 #ifdef FAST_IPSEC 199 struct secpolicy *sp = NULL; 200 int s; 201 #endif 202 203 memset(&ip6route, 0, sizeof(ip6route)); 204 205 #ifdef DIAGNOSTIC 206 if ((m->m_flags & M_PKTHDR) == 0) 207 panic("ip6_output: no HDR"); 208 209 if ((m->m_pkthdr.csum_flags & 210 (M_CSUM_TCPv4|M_CSUM_UDPv4|M_CSUM_TSOv4)) != 0) { 211 panic("ip6_output: IPv4 checksum offload flags: %d", 212 m->m_pkthdr.csum_flags); 213 } 214 215 if ((m->m_pkthdr.csum_flags & (M_CSUM_TCPv6|M_CSUM_UDPv6)) == 216 (M_CSUM_TCPv6|M_CSUM_UDPv6)) { 217 panic("ip6_output: conflicting checksum offload flags: %d", 218 m->m_pkthdr.csum_flags); 219 } 220 #endif 221 222 M_CSUM_DATA_IPv6_HL_SET(m->m_pkthdr.csum_data, sizeof(struct ip6_hdr)); 223 224 #define MAKE_EXTHDR(hp, mp) \ 225 do { \ 226 if (hp) { \ 227 struct ip6_ext *eh = (struct ip6_ext *)(hp); \ 228 error = ip6_copyexthdr((mp), (void *)(hp), \ 229 ((eh)->ip6e_len + 1) << 3); \ 230 if (error) \ 231 goto freehdrs; \ 232 } \ 233 } while (/*CONSTCOND*/ 0) 234 235 memset(&exthdrs, 0, sizeof(exthdrs)); 236 if (opt) { 237 /* Hop-by-Hop options header */ 238 MAKE_EXTHDR(opt->ip6po_hbh, &exthdrs.ip6e_hbh); 239 /* Destination options header(1st part) */ 240 MAKE_EXTHDR(opt->ip6po_dest1, &exthdrs.ip6e_dest1); 241 /* Routing header */ 242 MAKE_EXTHDR(opt->ip6po_rthdr, &exthdrs.ip6e_rthdr); 243 /* Destination options header(2nd part) */ 244 MAKE_EXTHDR(opt->ip6po_dest2, &exthdrs.ip6e_dest2); 245 } 246 247 #ifdef IPSEC 248 if ((flags & IPV6_FORWARDING) != 0) { 249 needipsec = 0; 250 goto skippolicycheck; 251 } 252 253 /* get a security policy for this packet */ 254 if (so == NULL) 255 sp = ipsec6_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND, 0, &error); 256 else { 257 if (IPSEC_PCB_SKIP_IPSEC(sotoinpcb_hdr(so)->inph_sp, 258 IPSEC_DIR_OUTBOUND)) { 259 needipsec = 0; 260 goto skippolicycheck; 261 } 262 sp = ipsec6_getpolicybysock(m, IPSEC_DIR_OUTBOUND, so, &error); 263 } 264 265 if (sp == NULL) { 266 IPSEC6_STATINC(IPSEC_STAT_OUT_INVAL); 267 goto freehdrs; 268 } 269 270 error = 0; 271 272 /* check policy */ 273 switch (sp->policy) { 274 case IPSEC_POLICY_DISCARD: 275 /* 276 * This packet is just discarded. 277 */ 278 IPSEC6_STATINC(IPSEC_STAT_OUT_POLVIO); 279 goto freehdrs; 280 281 case IPSEC_POLICY_BYPASS: 282 case IPSEC_POLICY_NONE: 283 /* no need to do IPsec. */ 284 needipsec = 0; 285 break; 286 287 case IPSEC_POLICY_IPSEC: 288 if (sp->req == NULL) { 289 /* XXX should be panic ? */ 290 printf("ip6_output: No IPsec request specified.\n"); 291 error = EINVAL; 292 goto freehdrs; 293 } 294 needipsec = 1; 295 break; 296 297 case IPSEC_POLICY_ENTRUST: 298 default: 299 printf("ip6_output: Invalid policy found. %d\n", sp->policy); 300 } 301 302 skippolicycheck:; 303 #endif /* IPSEC */ 304 305 /* 306 * Calculate the total length of the extension header chain. 307 * Keep the length of the unfragmentable part for fragmentation. 308 */ 309 optlen = 0; 310 if (exthdrs.ip6e_hbh) optlen += exthdrs.ip6e_hbh->m_len; 311 if (exthdrs.ip6e_dest1) optlen += exthdrs.ip6e_dest1->m_len; 312 if (exthdrs.ip6e_rthdr) optlen += exthdrs.ip6e_rthdr->m_len; 313 unfragpartlen = optlen + sizeof(struct ip6_hdr); 314 /* NOTE: we don't add AH/ESP length here. do that later. */ 315 if (exthdrs.ip6e_dest2) optlen += exthdrs.ip6e_dest2->m_len; 316 317 #ifdef FAST_IPSEC 318 /* Check the security policy (SP) for the packet */ 319 320 /* XXX For moment, we doesn't support packet with extented action */ 321 if (optlen !=0) 322 goto freehdrs; 323 324 sp = ipsec6_check_policy(m,so,flags,&needipsec,&error); 325 if (error != 0) { 326 /* 327 * Hack: -EINVAL is used to signal that a packet 328 * should be silently discarded. This is typically 329 * because we asked key management for an SA and 330 * it was delayed (e.g. kicked up to IKE). 331 */ 332 if (error == -EINVAL) 333 error = 0; 334 goto freehdrs; 335 } 336 #endif /* FAST_IPSEC */ 337 338 339 if (needipsec && 340 (m->m_pkthdr.csum_flags & (M_CSUM_UDPv6|M_CSUM_TCPv6)) != 0) { 341 in6_delayed_cksum(m); 342 m->m_pkthdr.csum_flags &= ~(M_CSUM_UDPv6|M_CSUM_TCPv6); 343 } 344 345 346 /* 347 * If we need IPsec, or there is at least one extension header, 348 * separate IP6 header from the payload. 349 */ 350 if ((needipsec || optlen) && !hdrsplit) { 351 if ((error = ip6_splithdr(m, &exthdrs)) != 0) { 352 m = NULL; 353 goto freehdrs; 354 } 355 m = exthdrs.ip6e_ip6; 356 hdrsplit++; 357 } 358 359 /* adjust pointer */ 360 ip6 = mtod(m, struct ip6_hdr *); 361 362 /* adjust mbuf packet header length */ 363 m->m_pkthdr.len += optlen; 364 plen = m->m_pkthdr.len - sizeof(*ip6); 365 366 /* If this is a jumbo payload, insert a jumbo payload option. */ 367 if (plen > IPV6_MAXPACKET) { 368 if (!hdrsplit) { 369 if ((error = ip6_splithdr(m, &exthdrs)) != 0) { 370 m = NULL; 371 goto freehdrs; 372 } 373 m = exthdrs.ip6e_ip6; 374 hdrsplit++; 375 } 376 /* adjust pointer */ 377 ip6 = mtod(m, struct ip6_hdr *); 378 if ((error = ip6_insert_jumboopt(&exthdrs, plen)) != 0) 379 goto freehdrs; 380 optlen += 8; /* XXX JUMBOOPTLEN */ 381 ip6->ip6_plen = 0; 382 } else 383 ip6->ip6_plen = htons(plen); 384 385 /* 386 * Concatenate headers and fill in next header fields. 387 * Here we have, on "m" 388 * IPv6 payload 389 * and we insert headers accordingly. Finally, we should be getting: 390 * IPv6 hbh dest1 rthdr ah* [esp* dest2 payload] 391 * 392 * during the header composing process, "m" points to IPv6 header. 393 * "mprev" points to an extension header prior to esp. 394 */ 395 { 396 u_char *nexthdrp = &ip6->ip6_nxt; 397 struct mbuf *mprev = m; 398 399 /* 400 * we treat dest2 specially. this makes IPsec processing 401 * much easier. the goal here is to make mprev point the 402 * mbuf prior to dest2. 403 * 404 * result: IPv6 dest2 payload 405 * m and mprev will point to IPv6 header. 406 */ 407 if (exthdrs.ip6e_dest2) { 408 if (!hdrsplit) 409 panic("assumption failed: hdr not split"); 410 exthdrs.ip6e_dest2->m_next = m->m_next; 411 m->m_next = exthdrs.ip6e_dest2; 412 *mtod(exthdrs.ip6e_dest2, u_char *) = ip6->ip6_nxt; 413 ip6->ip6_nxt = IPPROTO_DSTOPTS; 414 } 415 416 #define MAKE_CHAIN(m, mp, p, i)\ 417 do {\ 418 if (m) {\ 419 if (!hdrsplit) \ 420 panic("assumption failed: hdr not split"); \ 421 *mtod((m), u_char *) = *(p);\ 422 *(p) = (i);\ 423 p = mtod((m), u_char *);\ 424 (m)->m_next = (mp)->m_next;\ 425 (mp)->m_next = (m);\ 426 (mp) = (m);\ 427 }\ 428 } while (/*CONSTCOND*/ 0) 429 /* 430 * result: IPv6 hbh dest1 rthdr dest2 payload 431 * m will point to IPv6 header. mprev will point to the 432 * extension header prior to dest2 (rthdr in the above case). 433 */ 434 MAKE_CHAIN(exthdrs.ip6e_hbh, mprev, nexthdrp, IPPROTO_HOPOPTS); 435 MAKE_CHAIN(exthdrs.ip6e_dest1, mprev, nexthdrp, 436 IPPROTO_DSTOPTS); 437 MAKE_CHAIN(exthdrs.ip6e_rthdr, mprev, nexthdrp, 438 IPPROTO_ROUTING); 439 440 M_CSUM_DATA_IPv6_HL_SET(m->m_pkthdr.csum_data, 441 sizeof(struct ip6_hdr) + optlen); 442 443 #ifdef IPSEC 444 if (!needipsec) 445 goto skip_ipsec2; 446 447 /* 448 * pointers after IPsec headers are not valid any more. 449 * other pointers need a great care too. 450 * (IPsec routines should not mangle mbufs prior to AH/ESP) 451 */ 452 exthdrs.ip6e_dest2 = NULL; 453 454 { 455 struct ip6_rthdr *rh = NULL; 456 int segleft_org = 0; 457 struct ipsec_output_state state; 458 459 if (exthdrs.ip6e_rthdr) { 460 rh = mtod(exthdrs.ip6e_rthdr, struct ip6_rthdr *); 461 segleft_org = rh->ip6r_segleft; 462 rh->ip6r_segleft = 0; 463 } 464 465 memset(&state, 0, sizeof(state)); 466 state.m = m; 467 error = ipsec6_output_trans(&state, nexthdrp, mprev, sp, flags, 468 &needipsectun); 469 m = state.m; 470 if (error) { 471 rh = mtod(exthdrs.ip6e_rthdr, struct ip6_rthdr *); 472 /* mbuf is already reclaimed in ipsec6_output_trans. */ 473 m = NULL; 474 switch (error) { 475 case EHOSTUNREACH: 476 case ENETUNREACH: 477 case EMSGSIZE: 478 case ENOBUFS: 479 case ENOMEM: 480 break; 481 default: 482 printf("ip6_output (ipsec): error code %d\n", error); 483 /* FALLTHROUGH */ 484 case ENOENT: 485 /* don't show these error codes to the user */ 486 error = 0; 487 break; 488 } 489 goto bad; 490 } 491 if (exthdrs.ip6e_rthdr) { 492 /* ah6_output doesn't modify mbuf chain */ 493 rh->ip6r_segleft = segleft_org; 494 } 495 } 496 skip_ipsec2:; 497 #endif 498 } 499 500 /* 501 * If there is a routing header, replace destination address field 502 * with the first hop of the routing header. 503 */ 504 if (exthdrs.ip6e_rthdr) { 505 struct ip6_rthdr *rh; 506 struct ip6_rthdr0 *rh0; 507 struct in6_addr *addr; 508 struct sockaddr_in6 sa; 509 510 rh = (struct ip6_rthdr *)(mtod(exthdrs.ip6e_rthdr, 511 struct ip6_rthdr *)); 512 finaldst = ip6->ip6_dst; 513 switch (rh->ip6r_type) { 514 case IPV6_RTHDR_TYPE_0: 515 rh0 = (struct ip6_rthdr0 *)rh; 516 addr = (struct in6_addr *)(rh0 + 1); 517 518 /* 519 * construct a sockaddr_in6 form of 520 * the first hop. 521 * 522 * XXX: we may not have enough 523 * information about its scope zone; 524 * there is no standard API to pass 525 * the information from the 526 * application. 527 */ 528 sockaddr_in6_init(&sa, addr, 0, 0, 0); 529 if ((error = sa6_embedscope(&sa, 530 ip6_use_defzone)) != 0) { 531 goto bad; 532 } 533 ip6->ip6_dst = sa.sin6_addr; 534 (void)memmove(&addr[0], &addr[1], 535 sizeof(struct in6_addr) * 536 (rh0->ip6r0_segleft - 1)); 537 addr[rh0->ip6r0_segleft - 1] = finaldst; 538 /* XXX */ 539 in6_clearscope(addr + rh0->ip6r0_segleft - 1); 540 break; 541 default: /* is it possible? */ 542 error = EINVAL; 543 goto bad; 544 } 545 } 546 547 /* Source address validation */ 548 if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) && 549 (flags & IPV6_UNSPECSRC) == 0) { 550 error = EOPNOTSUPP; 551 IP6_STATINC(IP6_STAT_BADSCOPE); 552 goto bad; 553 } 554 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) { 555 error = EOPNOTSUPP; 556 IP6_STATINC(IP6_STAT_BADSCOPE); 557 goto bad; 558 } 559 560 IP6_STATINC(IP6_STAT_LOCALOUT); 561 562 /* 563 * Route packet. 564 */ 565 /* initialize cached route */ 566 if (ro == NULL) { 567 ro = &ip6route; 568 } 569 ro_pmtu = ro; 570 if (opt && opt->ip6po_rthdr) 571 ro = &opt->ip6po_route; 572 573 /* 574 * if specified, try to fill in the traffic class field. 575 * do not override if a non-zero value is already set. 576 * we check the diffserv field and the ecn field separately. 577 */ 578 if (opt && opt->ip6po_tclass >= 0) { 579 int mask = 0; 580 581 if ((ip6->ip6_flow & htonl(0xfc << 20)) == 0) 582 mask |= 0xfc; 583 if ((ip6->ip6_flow & htonl(0x03 << 20)) == 0) 584 mask |= 0x03; 585 if (mask != 0) 586 ip6->ip6_flow |= htonl((opt->ip6po_tclass & mask) << 20); 587 } 588 589 /* fill in or override the hop limit field, if necessary. */ 590 if (opt && opt->ip6po_hlim != -1) 591 ip6->ip6_hlim = opt->ip6po_hlim & 0xff; 592 else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) { 593 if (im6o != NULL) 594 ip6->ip6_hlim = im6o->im6o_multicast_hlim; 595 else 596 ip6->ip6_hlim = ip6_defmcasthlim; 597 } 598 599 #ifdef IPSEC 600 if (needipsec && needipsectun) { 601 struct ipsec_output_state state; 602 603 /* 604 * All the extension headers will become inaccessible 605 * (since they can be encrypted). 606 * Don't panic, we need no more updates to extension headers 607 * on inner IPv6 packet (since they are now encapsulated). 608 * 609 * IPv6 [ESP|AH] IPv6 [extension headers] payload 610 */ 611 memset(&exthdrs, 0, sizeof(exthdrs)); 612 exthdrs.ip6e_ip6 = m; 613 614 memset(&state, 0, sizeof(state)); 615 state.m = m; 616 state.ro = ro; 617 state.dst = rtcache_getdst(ro); 618 619 error = ipsec6_output_tunnel(&state, sp, flags); 620 621 m = state.m; 622 ro_pmtu = ro = state.ro; 623 dst = satocsin6(state.dst); 624 if (error) { 625 /* mbuf is already reclaimed in ipsec6_output_tunnel. */ 626 m0 = m = NULL; 627 m = NULL; 628 switch (error) { 629 case EHOSTUNREACH: 630 case ENETUNREACH: 631 case EMSGSIZE: 632 case ENOBUFS: 633 case ENOMEM: 634 break; 635 default: 636 printf("ip6_output (ipsec): error code %d\n", error); 637 /* FALLTHROUGH */ 638 case ENOENT: 639 /* don't show these error codes to the user */ 640 error = 0; 641 break; 642 } 643 goto bad; 644 } 645 646 exthdrs.ip6e_ip6 = m; 647 } 648 #endif /* IPSEC */ 649 #ifdef FAST_IPSEC 650 if (needipsec) { 651 s = splsoftnet(); 652 error = ipsec6_process_packet(m,sp->req); 653 654 /* 655 * Preserve KAME behaviour: ENOENT can be returned 656 * when an SA acquire is in progress. Don't propagate 657 * this to user-level; it confuses applications. 658 * XXX this will go away when the SADB is redone. 659 */ 660 if (error == ENOENT) 661 error = 0; 662 splx(s); 663 goto done; 664 } 665 #endif /* FAST_IPSEC */ 666 667 668 669 /* adjust pointer */ 670 ip6 = mtod(m, struct ip6_hdr *); 671 672 sockaddr_in6_init(&dst_sa, &ip6->ip6_dst, 0, 0, 0); 673 if ((error = in6_selectroute(&dst_sa, opt, im6o, ro, 674 &ifp, &rt, 0)) != 0) { 675 if (ifp != NULL) 676 in6_ifstat_inc(ifp, ifs6_out_discard); 677 goto bad; 678 } 679 if (rt == NULL) { 680 /* 681 * If in6_selectroute() does not return a route entry, 682 * dst may not have been updated. 683 */ 684 rtcache_setdst(ro, sin6tosa(&dst_sa)); 685 } 686 687 /* 688 * then rt (for unicast) and ifp must be non-NULL valid values. 689 */ 690 if ((flags & IPV6_FORWARDING) == 0) { 691 /* XXX: the FORWARDING flag can be set for mrouting. */ 692 in6_ifstat_inc(ifp, ifs6_out_request); 693 } 694 if (rt != NULL) { 695 ia = (struct in6_ifaddr *)(rt->rt_ifa); 696 rt->rt_use++; 697 } 698 699 /* 700 * The outgoing interface must be in the zone of source and 701 * destination addresses. We should use ia_ifp to support the 702 * case of sending packets to an address of our own. 703 */ 704 if (ia != NULL && ia->ia_ifp) 705 origifp = ia->ia_ifp; 706 else 707 origifp = ifp; 708 709 src0 = ip6->ip6_src; 710 if (in6_setscope(&src0, origifp, &zone)) 711 goto badscope; 712 sockaddr_in6_init(&src_sa, &ip6->ip6_src, 0, 0, 0); 713 if (sa6_recoverscope(&src_sa) || zone != src_sa.sin6_scope_id) 714 goto badscope; 715 716 dst0 = ip6->ip6_dst; 717 if (in6_setscope(&dst0, origifp, &zone)) 718 goto badscope; 719 /* re-initialize to be sure */ 720 sockaddr_in6_init(&dst_sa, &ip6->ip6_dst, 0, 0, 0); 721 if (sa6_recoverscope(&dst_sa) || zone != dst_sa.sin6_scope_id) 722 goto badscope; 723 724 /* scope check is done. */ 725 726 if (rt == NULL || IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) { 727 if (dst == NULL) 728 dst = satocsin6(rtcache_getdst(ro)); 729 KASSERT(dst != NULL); 730 } else if (opt && rtcache_validate(&opt->ip6po_nextroute) != NULL) { 731 /* 732 * The nexthop is explicitly specified by the 733 * application. We assume the next hop is an IPv6 734 * address. 735 */ 736 dst = (struct sockaddr_in6 *)opt->ip6po_nexthop; 737 } else if ((rt->rt_flags & RTF_GATEWAY)) 738 dst = (struct sockaddr_in6 *)rt->rt_gateway; 739 else if (dst == NULL) 740 dst = satocsin6(rtcache_getdst(ro)); 741 742 /* 743 * XXXXXX: original code follows: 744 */ 745 if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) 746 m->m_flags &= ~(M_BCAST | M_MCAST); /* just in case */ 747 else { 748 struct in6_multi *in6m; 749 750 m->m_flags = (m->m_flags & ~M_BCAST) | M_MCAST; 751 752 in6_ifstat_inc(ifp, ifs6_out_mcast); 753 754 /* 755 * Confirm that the outgoing interface supports multicast. 756 */ 757 if (!(ifp->if_flags & IFF_MULTICAST)) { 758 IP6_STATINC(IP6_STAT_NOROUTE); 759 in6_ifstat_inc(ifp, ifs6_out_discard); 760 error = ENETUNREACH; 761 goto bad; 762 } 763 764 IN6_LOOKUP_MULTI(ip6->ip6_dst, ifp, in6m); 765 if (in6m != NULL && 766 (im6o == NULL || im6o->im6o_multicast_loop)) { 767 /* 768 * If we belong to the destination multicast group 769 * on the outgoing interface, and the caller did not 770 * forbid loopback, loop back a copy. 771 */ 772 KASSERT(dst != NULL); 773 ip6_mloopback(ifp, m, dst); 774 } else { 775 /* 776 * If we are acting as a multicast router, perform 777 * multicast forwarding as if the packet had just 778 * arrived on the interface to which we are about 779 * to send. The multicast forwarding function 780 * recursively calls this function, using the 781 * IPV6_FORWARDING flag to prevent infinite recursion. 782 * 783 * Multicasts that are looped back by ip6_mloopback(), 784 * above, will be forwarded by the ip6_input() routine, 785 * if necessary. 786 */ 787 if (ip6_mrouter && (flags & IPV6_FORWARDING) == 0) { 788 if (ip6_mforward(ip6, ifp, m) != 0) { 789 m_freem(m); 790 goto done; 791 } 792 } 793 } 794 /* 795 * Multicasts with a hoplimit of zero may be looped back, 796 * above, but must not be transmitted on a network. 797 * Also, multicasts addressed to the loopback interface 798 * are not sent -- the above call to ip6_mloopback() will 799 * loop back a copy if this host actually belongs to the 800 * destination group on the loopback interface. 801 */ 802 if (ip6->ip6_hlim == 0 || (ifp->if_flags & IFF_LOOPBACK) || 803 IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst)) { 804 m_freem(m); 805 goto done; 806 } 807 } 808 809 /* 810 * Fill the outgoing inteface to tell the upper layer 811 * to increment per-interface statistics. 812 */ 813 if (ifpp) 814 *ifpp = ifp; 815 816 /* Determine path MTU. */ 817 if ((error = ip6_getpmtu(ro_pmtu, ro, ifp, &finaldst, &mtu, 818 &alwaysfrag)) != 0) 819 goto bad; 820 #ifdef IPSEC 821 if (needipsectun) 822 mtu = IPV6_MMTU; 823 #endif 824 825 /* 826 * The caller of this function may specify to use the minimum MTU 827 * in some cases. 828 * An advanced API option (IPV6_USE_MIN_MTU) can also override MTU 829 * setting. The logic is a bit complicated; by default, unicast 830 * packets will follow path MTU while multicast packets will be sent at 831 * the minimum MTU. If IP6PO_MINMTU_ALL is specified, all packets 832 * including unicast ones will be sent at the minimum MTU. Multicast 833 * packets will always be sent at the minimum MTU unless 834 * IP6PO_MINMTU_DISABLE is explicitly specified. 835 * See RFC 3542 for more details. 836 */ 837 if (mtu > IPV6_MMTU) { 838 if ((flags & IPV6_MINMTU)) 839 mtu = IPV6_MMTU; 840 else if (opt && opt->ip6po_minmtu == IP6PO_MINMTU_ALL) 841 mtu = IPV6_MMTU; 842 else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) && 843 (opt == NULL || 844 opt->ip6po_minmtu != IP6PO_MINMTU_DISABLE)) { 845 mtu = IPV6_MMTU; 846 } 847 } 848 849 /* 850 * clear embedded scope identifiers if necessary. 851 * in6_clearscope will touch the addresses only when necessary. 852 */ 853 in6_clearscope(&ip6->ip6_src); 854 in6_clearscope(&ip6->ip6_dst); 855 856 /* 857 * If the outgoing packet contains a hop-by-hop options header, 858 * it must be examined and processed even by the source node. 859 * (RFC 2460, section 4.) 860 */ 861 if (exthdrs.ip6e_hbh) { 862 struct ip6_hbh *hbh = mtod(exthdrs.ip6e_hbh, struct ip6_hbh *); 863 u_int32_t dummy1; /* XXX unused */ 864 u_int32_t dummy2; /* XXX unused */ 865 866 /* 867 * XXX: if we have to send an ICMPv6 error to the sender, 868 * we need the M_LOOP flag since icmp6_error() expects 869 * the IPv6 and the hop-by-hop options header are 870 * continuous unless the flag is set. 871 */ 872 m->m_flags |= M_LOOP; 873 m->m_pkthdr.rcvif = ifp; 874 if (ip6_process_hopopts(m, (u_int8_t *)(hbh + 1), 875 ((hbh->ip6h_len + 1) << 3) - sizeof(struct ip6_hbh), 876 &dummy1, &dummy2) < 0) { 877 /* m was already freed at this point */ 878 error = EINVAL;/* better error? */ 879 goto done; 880 } 881 m->m_flags &= ~M_LOOP; /* XXX */ 882 m->m_pkthdr.rcvif = NULL; 883 } 884 885 #ifdef PFIL_HOOKS 886 /* 887 * Run through list of hooks for output packets. 888 */ 889 if ((error = pfil_run_hooks(&inet6_pfil_hook, &m, ifp, PFIL_OUT)) != 0) 890 goto done; 891 if (m == NULL) 892 goto done; 893 ip6 = mtod(m, struct ip6_hdr *); 894 #endif /* PFIL_HOOKS */ 895 /* 896 * Send the packet to the outgoing interface. 897 * If necessary, do IPv6 fragmentation before sending. 898 * 899 * the logic here is rather complex: 900 * 1: normal case (dontfrag == 0, alwaysfrag == 0) 901 * 1-a: send as is if tlen <= path mtu 902 * 1-b: fragment if tlen > path mtu 903 * 904 * 2: if user asks us not to fragment (dontfrag == 1) 905 * 2-a: send as is if tlen <= interface mtu 906 * 2-b: error if tlen > interface mtu 907 * 908 * 3: if we always need to attach fragment header (alwaysfrag == 1) 909 * always fragment 910 * 911 * 4: if dontfrag == 1 && alwaysfrag == 1 912 * error, as we cannot handle this conflicting request 913 */ 914 tlen = m->m_pkthdr.len; 915 tso = (m->m_pkthdr.csum_flags & M_CSUM_TSOv6) != 0; 916 if (opt && (opt->ip6po_flags & IP6PO_DONTFRAG)) 917 dontfrag = 1; 918 else 919 dontfrag = 0; 920 921 if (dontfrag && alwaysfrag) { /* case 4 */ 922 /* conflicting request - can't transmit */ 923 error = EMSGSIZE; 924 goto bad; 925 } 926 if (dontfrag && (!tso && tlen > IN6_LINKMTU(ifp))) { /* case 2-b */ 927 /* 928 * Even if the DONTFRAG option is specified, we cannot send the 929 * packet when the data length is larger than the MTU of the 930 * outgoing interface. 931 * Notify the error by sending IPV6_PATHMTU ancillary data as 932 * well as returning an error code (the latter is not described 933 * in the API spec.) 934 */ 935 u_int32_t mtu32; 936 struct ip6ctlparam ip6cp; 937 938 mtu32 = (u_int32_t)mtu; 939 memset(&ip6cp, 0, sizeof(ip6cp)); 940 ip6cp.ip6c_cmdarg = (void *)&mtu32; 941 pfctlinput2(PRC_MSGSIZE, 942 rtcache_getdst(ro_pmtu), &ip6cp); 943 944 error = EMSGSIZE; 945 goto bad; 946 } 947 948 /* 949 * transmit packet without fragmentation 950 */ 951 if (dontfrag || (!alwaysfrag && (tlen <= mtu || tso))) { 952 /* case 1-a and 2-a */ 953 struct in6_ifaddr *ia6; 954 int sw_csum; 955 956 ip6 = mtod(m, struct ip6_hdr *); 957 ia6 = in6_ifawithifp(ifp, &ip6->ip6_src); 958 if (ia6) { 959 /* Record statistics for this interface address. */ 960 ia6->ia_ifa.ifa_data.ifad_outbytes += m->m_pkthdr.len; 961 } 962 #ifdef IPSEC 963 /* clean ipsec history once it goes out of the node */ 964 ipsec_delaux(m); 965 #endif 966 967 sw_csum = m->m_pkthdr.csum_flags & ~ifp->if_csum_flags_tx; 968 if ((sw_csum & (M_CSUM_UDPv6|M_CSUM_TCPv6)) != 0) { 969 if (IN6_NEED_CHECKSUM(ifp, 970 sw_csum & (M_CSUM_UDPv6|M_CSUM_TCPv6))) { 971 in6_delayed_cksum(m); 972 } 973 m->m_pkthdr.csum_flags &= ~(M_CSUM_UDPv6|M_CSUM_TCPv6); 974 } 975 976 KASSERT(dst != NULL); 977 if (__predict_true(!tso || 978 (ifp->if_capenable & IFCAP_TSOv6) != 0)) { 979 error = nd6_output(ifp, origifp, m, dst, rt); 980 } else { 981 error = ip6_tso_output(ifp, origifp, m, dst, rt); 982 } 983 goto done; 984 } 985 986 if (tso) { 987 error = EINVAL; /* XXX */ 988 goto bad; 989 } 990 991 /* 992 * try to fragment the packet. case 1-b and 3 993 */ 994 if (mtu < IPV6_MMTU) { 995 /* path MTU cannot be less than IPV6_MMTU */ 996 error = EMSGSIZE; 997 in6_ifstat_inc(ifp, ifs6_out_fragfail); 998 goto bad; 999 } else if (ip6->ip6_plen == 0) { 1000 /* jumbo payload cannot be fragmented */ 1001 error = EMSGSIZE; 1002 in6_ifstat_inc(ifp, ifs6_out_fragfail); 1003 goto bad; 1004 } else { 1005 struct mbuf **mnext, *m_frgpart; 1006 struct ip6_frag *ip6f; 1007 u_int32_t id = htonl(ip6_randomid()); 1008 u_char nextproto; 1009 #if 0 /* see below */ 1010 struct ip6ctlparam ip6cp; 1011 u_int32_t mtu32; 1012 #endif 1013 1014 /* 1015 * Too large for the destination or interface; 1016 * fragment if possible. 1017 * Must be able to put at least 8 bytes per fragment. 1018 */ 1019 hlen = unfragpartlen; 1020 if (mtu > IPV6_MAXPACKET) 1021 mtu = IPV6_MAXPACKET; 1022 1023 #if 0 1024 /* 1025 * It is believed this code is a leftover from the 1026 * development of the IPV6_RECVPATHMTU sockopt and 1027 * associated work to implement RFC3542. 1028 * It's not entirely clear what the intent of the API 1029 * is at this point, so disable this code for now. 1030 * The IPV6_RECVPATHMTU sockopt and/or IPV6_DONTFRAG 1031 * will send notifications if the application requests. 1032 */ 1033 1034 /* Notify a proper path MTU to applications. */ 1035 mtu32 = (u_int32_t)mtu; 1036 memset(&ip6cp, 0, sizeof(ip6cp)); 1037 ip6cp.ip6c_cmdarg = (void *)&mtu32; 1038 pfctlinput2(PRC_MSGSIZE, 1039 rtcache_getdst(ro_pmtu), &ip6cp); 1040 #endif 1041 1042 len = (mtu - hlen - sizeof(struct ip6_frag)) & ~7; 1043 if (len < 8) { 1044 error = EMSGSIZE; 1045 in6_ifstat_inc(ifp, ifs6_out_fragfail); 1046 goto bad; 1047 } 1048 1049 mnext = &m->m_nextpkt; 1050 1051 /* 1052 * Change the next header field of the last header in the 1053 * unfragmentable part. 1054 */ 1055 if (exthdrs.ip6e_rthdr) { 1056 nextproto = *mtod(exthdrs.ip6e_rthdr, u_char *); 1057 *mtod(exthdrs.ip6e_rthdr, u_char *) = IPPROTO_FRAGMENT; 1058 } else if (exthdrs.ip6e_dest1) { 1059 nextproto = *mtod(exthdrs.ip6e_dest1, u_char *); 1060 *mtod(exthdrs.ip6e_dest1, u_char *) = IPPROTO_FRAGMENT; 1061 } else if (exthdrs.ip6e_hbh) { 1062 nextproto = *mtod(exthdrs.ip6e_hbh, u_char *); 1063 *mtod(exthdrs.ip6e_hbh, u_char *) = IPPROTO_FRAGMENT; 1064 } else { 1065 nextproto = ip6->ip6_nxt; 1066 ip6->ip6_nxt = IPPROTO_FRAGMENT; 1067 } 1068 1069 if ((m->m_pkthdr.csum_flags & (M_CSUM_UDPv6|M_CSUM_TCPv6)) 1070 != 0) { 1071 if (IN6_NEED_CHECKSUM(ifp, 1072 m->m_pkthdr.csum_flags & 1073 (M_CSUM_UDPv6|M_CSUM_TCPv6))) { 1074 in6_delayed_cksum(m); 1075 } 1076 m->m_pkthdr.csum_flags &= ~(M_CSUM_UDPv6|M_CSUM_TCPv6); 1077 } 1078 1079 /* 1080 * Loop through length of segment after first fragment, 1081 * make new header and copy data of each part and link onto 1082 * chain. 1083 */ 1084 m0 = m; 1085 for (off = hlen; off < tlen; off += len) { 1086 struct mbuf *mlast; 1087 1088 MGETHDR(m, M_DONTWAIT, MT_HEADER); 1089 if (!m) { 1090 error = ENOBUFS; 1091 IP6_STATINC(IP6_STAT_ODROPPED); 1092 goto sendorfree; 1093 } 1094 m->m_pkthdr.rcvif = NULL; 1095 m->m_flags = m0->m_flags & M_COPYFLAGS; 1096 *mnext = m; 1097 mnext = &m->m_nextpkt; 1098 m->m_data += max_linkhdr; 1099 mhip6 = mtod(m, struct ip6_hdr *); 1100 *mhip6 = *ip6; 1101 m->m_len = sizeof(*mhip6); 1102 error = ip6_insertfraghdr(m0, m, hlen, &ip6f); 1103 if (error) { 1104 IP6_STATINC(IP6_STAT_ODROPPED); 1105 goto sendorfree; 1106 } 1107 ip6f->ip6f_offlg = htons((u_int16_t)((off - hlen) & ~7)); 1108 if (off + len >= tlen) 1109 len = tlen - off; 1110 else 1111 ip6f->ip6f_offlg |= IP6F_MORE_FRAG; 1112 mhip6->ip6_plen = htons((u_int16_t)(len + hlen + 1113 sizeof(*ip6f) - sizeof(struct ip6_hdr))); 1114 if ((m_frgpart = m_copy(m0, off, len)) == 0) { 1115 error = ENOBUFS; 1116 IP6_STATINC(IP6_STAT_ODROPPED); 1117 goto sendorfree; 1118 } 1119 for (mlast = m; mlast->m_next; mlast = mlast->m_next) 1120 ; 1121 mlast->m_next = m_frgpart; 1122 m->m_pkthdr.len = len + hlen + sizeof(*ip6f); 1123 m->m_pkthdr.rcvif = (struct ifnet *)0; 1124 ip6f->ip6f_reserved = 0; 1125 ip6f->ip6f_ident = id; 1126 ip6f->ip6f_nxt = nextproto; 1127 IP6_STATINC(IP6_STAT_OFRAGMENTS); 1128 in6_ifstat_inc(ifp, ifs6_out_fragcreat); 1129 } 1130 1131 in6_ifstat_inc(ifp, ifs6_out_fragok); 1132 } 1133 1134 /* 1135 * Remove leading garbages. 1136 */ 1137 sendorfree: 1138 m = m0->m_nextpkt; 1139 m0->m_nextpkt = 0; 1140 m_freem(m0); 1141 for (m0 = m; m; m = m0) { 1142 m0 = m->m_nextpkt; 1143 m->m_nextpkt = 0; 1144 if (error == 0) { 1145 struct in6_ifaddr *ia6; 1146 ip6 = mtod(m, struct ip6_hdr *); 1147 ia6 = in6_ifawithifp(ifp, &ip6->ip6_src); 1148 if (ia6) { 1149 /* 1150 * Record statistics for this interface 1151 * address. 1152 */ 1153 ia6->ia_ifa.ifa_data.ifad_outbytes += 1154 m->m_pkthdr.len; 1155 } 1156 #ifdef IPSEC 1157 /* clean ipsec history once it goes out of the node */ 1158 ipsec_delaux(m); 1159 #endif 1160 KASSERT(dst != NULL); 1161 error = nd6_output(ifp, origifp, m, dst, rt); 1162 } else 1163 m_freem(m); 1164 } 1165 1166 if (error == 0) 1167 IP6_STATINC(IP6_STAT_FRAGMENTED); 1168 1169 done: 1170 rtcache_free(&ip6route); 1171 1172 #ifdef IPSEC 1173 if (sp != NULL) 1174 key_freesp(sp); 1175 #endif /* IPSEC */ 1176 #ifdef FAST_IPSEC 1177 if (sp != NULL) 1178 KEY_FREESP(&sp); 1179 #endif /* FAST_IPSEC */ 1180 1181 1182 return (error); 1183 1184 freehdrs: 1185 m_freem(exthdrs.ip6e_hbh); /* m_freem will check if mbuf is 0 */ 1186 m_freem(exthdrs.ip6e_dest1); 1187 m_freem(exthdrs.ip6e_rthdr); 1188 m_freem(exthdrs.ip6e_dest2); 1189 /* FALLTHROUGH */ 1190 bad: 1191 m_freem(m); 1192 goto done; 1193 badscope: 1194 IP6_STATINC(IP6_STAT_BADSCOPE); 1195 in6_ifstat_inc(origifp, ifs6_out_discard); 1196 if (error == 0) 1197 error = EHOSTUNREACH; /* XXX */ 1198 goto bad; 1199 } 1200 1201 static int 1202 ip6_copyexthdr(struct mbuf **mp, void *hdr, int hlen) 1203 { 1204 struct mbuf *m; 1205 1206 if (hlen > MCLBYTES) 1207 return (ENOBUFS); /* XXX */ 1208 1209 MGET(m, M_DONTWAIT, MT_DATA); 1210 if (!m) 1211 return (ENOBUFS); 1212 1213 if (hlen > MLEN) { 1214 MCLGET(m, M_DONTWAIT); 1215 if ((m->m_flags & M_EXT) == 0) { 1216 m_free(m); 1217 return (ENOBUFS); 1218 } 1219 } 1220 m->m_len = hlen; 1221 if (hdr) 1222 bcopy(hdr, mtod(m, void *), hlen); 1223 1224 *mp = m; 1225 return (0); 1226 } 1227 1228 /* 1229 * Process a delayed payload checksum calculation. 1230 */ 1231 void 1232 in6_delayed_cksum(struct mbuf *m) 1233 { 1234 uint16_t csum, offset; 1235 1236 KASSERT((m->m_pkthdr.csum_flags & (M_CSUM_UDPv6|M_CSUM_TCPv6)) != 0); 1237 KASSERT((~m->m_pkthdr.csum_flags & (M_CSUM_UDPv6|M_CSUM_TCPv6)) != 0); 1238 KASSERT((m->m_pkthdr.csum_flags 1239 & (M_CSUM_UDPv4|M_CSUM_TCPv4|M_CSUM_TSOv4)) == 0); 1240 1241 offset = M_CSUM_DATA_IPv6_HL(m->m_pkthdr.csum_data); 1242 csum = in6_cksum(m, 0, offset, m->m_pkthdr.len - offset); 1243 if (csum == 0 && (m->m_pkthdr.csum_flags & M_CSUM_UDPv6) != 0) { 1244 csum = 0xffff; 1245 } 1246 1247 offset += M_CSUM_DATA_IPv6_OFFSET(m->m_pkthdr.csum_data); 1248 if ((offset + sizeof(csum)) > m->m_len) { 1249 m_copyback(m, offset, sizeof(csum), &csum); 1250 } else { 1251 *(uint16_t *)(mtod(m, char *) + offset) = csum; 1252 } 1253 } 1254 1255 /* 1256 * Insert jumbo payload option. 1257 */ 1258 static int 1259 ip6_insert_jumboopt(struct ip6_exthdrs *exthdrs, u_int32_t plen) 1260 { 1261 struct mbuf *mopt; 1262 u_int8_t *optbuf; 1263 u_int32_t v; 1264 1265 #define JUMBOOPTLEN 8 /* length of jumbo payload option and padding */ 1266 1267 /* 1268 * If there is no hop-by-hop options header, allocate new one. 1269 * If there is one but it doesn't have enough space to store the 1270 * jumbo payload option, allocate a cluster to store the whole options. 1271 * Otherwise, use it to store the options. 1272 */ 1273 if (exthdrs->ip6e_hbh == 0) { 1274 MGET(mopt, M_DONTWAIT, MT_DATA); 1275 if (mopt == 0) 1276 return (ENOBUFS); 1277 mopt->m_len = JUMBOOPTLEN; 1278 optbuf = mtod(mopt, u_int8_t *); 1279 optbuf[1] = 0; /* = ((JUMBOOPTLEN) >> 3) - 1 */ 1280 exthdrs->ip6e_hbh = mopt; 1281 } else { 1282 struct ip6_hbh *hbh; 1283 1284 mopt = exthdrs->ip6e_hbh; 1285 if (M_TRAILINGSPACE(mopt) < JUMBOOPTLEN) { 1286 /* 1287 * XXX assumption: 1288 * - exthdrs->ip6e_hbh is not referenced from places 1289 * other than exthdrs. 1290 * - exthdrs->ip6e_hbh is not an mbuf chain. 1291 */ 1292 int oldoptlen = mopt->m_len; 1293 struct mbuf *n; 1294 1295 /* 1296 * XXX: give up if the whole (new) hbh header does 1297 * not fit even in an mbuf cluster. 1298 */ 1299 if (oldoptlen + JUMBOOPTLEN > MCLBYTES) 1300 return (ENOBUFS); 1301 1302 /* 1303 * As a consequence, we must always prepare a cluster 1304 * at this point. 1305 */ 1306 MGET(n, M_DONTWAIT, MT_DATA); 1307 if (n) { 1308 MCLGET(n, M_DONTWAIT); 1309 if ((n->m_flags & M_EXT) == 0) { 1310 m_freem(n); 1311 n = NULL; 1312 } 1313 } 1314 if (!n) 1315 return (ENOBUFS); 1316 n->m_len = oldoptlen + JUMBOOPTLEN; 1317 bcopy(mtod(mopt, void *), mtod(n, void *), 1318 oldoptlen); 1319 optbuf = mtod(n, u_int8_t *) + oldoptlen; 1320 m_freem(mopt); 1321 mopt = exthdrs->ip6e_hbh = n; 1322 } else { 1323 optbuf = mtod(mopt, u_int8_t *) + mopt->m_len; 1324 mopt->m_len += JUMBOOPTLEN; 1325 } 1326 optbuf[0] = IP6OPT_PADN; 1327 optbuf[1] = 0; 1328 1329 /* 1330 * Adjust the header length according to the pad and 1331 * the jumbo payload option. 1332 */ 1333 hbh = mtod(mopt, struct ip6_hbh *); 1334 hbh->ip6h_len += (JUMBOOPTLEN >> 3); 1335 } 1336 1337 /* fill in the option. */ 1338 optbuf[2] = IP6OPT_JUMBO; 1339 optbuf[3] = 4; 1340 v = (u_int32_t)htonl(plen + JUMBOOPTLEN); 1341 bcopy(&v, &optbuf[4], sizeof(u_int32_t)); 1342 1343 /* finally, adjust the packet header length */ 1344 exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN; 1345 1346 return (0); 1347 #undef JUMBOOPTLEN 1348 } 1349 1350 /* 1351 * Insert fragment header and copy unfragmentable header portions. 1352 */ 1353 static int 1354 ip6_insertfraghdr(struct mbuf *m0, struct mbuf *m, int hlen, 1355 struct ip6_frag **frghdrp) 1356 { 1357 struct mbuf *n, *mlast; 1358 1359 if (hlen > sizeof(struct ip6_hdr)) { 1360 n = m_copym(m0, sizeof(struct ip6_hdr), 1361 hlen - sizeof(struct ip6_hdr), M_DONTWAIT); 1362 if (n == 0) 1363 return (ENOBUFS); 1364 m->m_next = n; 1365 } else 1366 n = m; 1367 1368 /* Search for the last mbuf of unfragmentable part. */ 1369 for (mlast = n; mlast->m_next; mlast = mlast->m_next) 1370 ; 1371 1372 if ((mlast->m_flags & M_EXT) == 0 && 1373 M_TRAILINGSPACE(mlast) >= sizeof(struct ip6_frag)) { 1374 /* use the trailing space of the last mbuf for the fragment hdr */ 1375 *frghdrp = (struct ip6_frag *)(mtod(mlast, char *) + 1376 mlast->m_len); 1377 mlast->m_len += sizeof(struct ip6_frag); 1378 m->m_pkthdr.len += sizeof(struct ip6_frag); 1379 } else { 1380 /* allocate a new mbuf for the fragment header */ 1381 struct mbuf *mfrg; 1382 1383 MGET(mfrg, M_DONTWAIT, MT_DATA); 1384 if (mfrg == 0) 1385 return (ENOBUFS); 1386 mfrg->m_len = sizeof(struct ip6_frag); 1387 *frghdrp = mtod(mfrg, struct ip6_frag *); 1388 mlast->m_next = mfrg; 1389 } 1390 1391 return (0); 1392 } 1393 1394 static int 1395 ip6_getpmtu(struct route *ro_pmtu, struct route *ro, struct ifnet *ifp, 1396 const struct in6_addr *dst, u_long *mtup, int *alwaysfragp) 1397 { 1398 struct rtentry *rt; 1399 u_int32_t mtu = 0; 1400 int alwaysfrag = 0; 1401 int error = 0; 1402 1403 if (ro_pmtu != ro) { 1404 union { 1405 struct sockaddr dst; 1406 struct sockaddr_in6 dst6; 1407 } u; 1408 1409 /* The first hop and the final destination may differ. */ 1410 sockaddr_in6_init(&u.dst6, dst, 0, 0, 0); 1411 rt = rtcache_lookup(ro_pmtu, &u.dst); 1412 } else 1413 rt = rtcache_validate(ro_pmtu); 1414 if (rt != NULL) { 1415 u_int32_t ifmtu; 1416 1417 if (ifp == NULL) 1418 ifp = rt->rt_ifp; 1419 ifmtu = IN6_LINKMTU(ifp); 1420 mtu = rt->rt_rmx.rmx_mtu; 1421 if (mtu == 0) 1422 mtu = ifmtu; 1423 else if (mtu < IPV6_MMTU) { 1424 /* 1425 * RFC2460 section 5, last paragraph: 1426 * if we record ICMPv6 too big message with 1427 * mtu < IPV6_MMTU, transmit packets sized IPV6_MMTU 1428 * or smaller, with fragment header attached. 1429 * (fragment header is needed regardless from the 1430 * packet size, for translators to identify packets) 1431 */ 1432 alwaysfrag = 1; 1433 mtu = IPV6_MMTU; 1434 } else if (mtu > ifmtu) { 1435 /* 1436 * The MTU on the route is larger than the MTU on 1437 * the interface! This shouldn't happen, unless the 1438 * MTU of the interface has been changed after the 1439 * interface was brought up. Change the MTU in the 1440 * route to match the interface MTU (as long as the 1441 * field isn't locked). 1442 */ 1443 mtu = ifmtu; 1444 if (!(rt->rt_rmx.rmx_locks & RTV_MTU)) 1445 rt->rt_rmx.rmx_mtu = mtu; 1446 } 1447 } else if (ifp) { 1448 mtu = IN6_LINKMTU(ifp); 1449 } else 1450 error = EHOSTUNREACH; /* XXX */ 1451 1452 *mtup = mtu; 1453 if (alwaysfragp) 1454 *alwaysfragp = alwaysfrag; 1455 return (error); 1456 } 1457 1458 /* 1459 * IP6 socket option processing. 1460 */ 1461 int 1462 ip6_ctloutput(int op, struct socket *so, struct sockopt *sopt) 1463 { 1464 int optdatalen, uproto; 1465 void *optdata; 1466 struct in6pcb *in6p = sotoin6pcb(so); 1467 int error, optval; 1468 int level, optname; 1469 1470 KASSERT(sopt != NULL); 1471 1472 level = sopt->sopt_level; 1473 optname = sopt->sopt_name; 1474 1475 error = optval = 0; 1476 uproto = (int)so->so_proto->pr_protocol; 1477 1478 if (level != IPPROTO_IPV6) { 1479 return ENOPROTOOPT; 1480 } 1481 switch (op) { 1482 case PRCO_SETOPT: 1483 switch (optname) { 1484 #ifdef RFC2292 1485 case IPV6_2292PKTOPTIONS: 1486 error = ip6_pcbopts(&in6p->in6p_outputopts, so, sopt); 1487 break; 1488 #endif 1489 1490 /* 1491 * Use of some Hop-by-Hop options or some 1492 * Destination options, might require special 1493 * privilege. That is, normal applications 1494 * (without special privilege) might be forbidden 1495 * from setting certain options in outgoing packets, 1496 * and might never see certain options in received 1497 * packets. [RFC 2292 Section 6] 1498 * KAME specific note: 1499 * KAME prevents non-privileged users from sending or 1500 * receiving ANY hbh/dst options in order to avoid 1501 * overhead of parsing options in the kernel. 1502 */ 1503 case IPV6_RECVHOPOPTS: 1504 case IPV6_RECVDSTOPTS: 1505 case IPV6_RECVRTHDRDSTOPTS: 1506 error = kauth_authorize_generic(kauth_cred_get(), 1507 KAUTH_GENERIC_ISSUSER, NULL); 1508 if (error) 1509 break; 1510 /* FALLTHROUGH */ 1511 case IPV6_UNICAST_HOPS: 1512 case IPV6_HOPLIMIT: 1513 case IPV6_FAITH: 1514 1515 case IPV6_RECVPKTINFO: 1516 case IPV6_RECVHOPLIMIT: 1517 case IPV6_RECVRTHDR: 1518 case IPV6_RECVPATHMTU: 1519 case IPV6_RECVTCLASS: 1520 case IPV6_V6ONLY: 1521 error = sockopt_getint(sopt, &optval); 1522 if (error) 1523 break; 1524 switch (optname) { 1525 case IPV6_UNICAST_HOPS: 1526 if (optval < -1 || optval >= 256) 1527 error = EINVAL; 1528 else { 1529 /* -1 = kernel default */ 1530 in6p->in6p_hops = optval; 1531 } 1532 break; 1533 #define OPTSET(bit) \ 1534 do { \ 1535 if (optval) \ 1536 in6p->in6p_flags |= (bit); \ 1537 else \ 1538 in6p->in6p_flags &= ~(bit); \ 1539 } while (/*CONSTCOND*/ 0) 1540 1541 #ifdef RFC2292 1542 #define OPTSET2292(bit) \ 1543 do { \ 1544 in6p->in6p_flags |= IN6P_RFC2292; \ 1545 if (optval) \ 1546 in6p->in6p_flags |= (bit); \ 1547 else \ 1548 in6p->in6p_flags &= ~(bit); \ 1549 } while (/*CONSTCOND*/ 0) 1550 #endif 1551 1552 #define OPTBIT(bit) (in6p->in6p_flags & (bit) ? 1 : 0) 1553 1554 case IPV6_RECVPKTINFO: 1555 #ifdef RFC2292 1556 /* cannot mix with RFC2292 */ 1557 if (OPTBIT(IN6P_RFC2292)) { 1558 error = EINVAL; 1559 break; 1560 } 1561 #endif 1562 OPTSET(IN6P_PKTINFO); 1563 break; 1564 1565 case IPV6_HOPLIMIT: 1566 { 1567 struct ip6_pktopts **optp; 1568 1569 #ifdef RFC2292 1570 /* cannot mix with RFC2292 */ 1571 if (OPTBIT(IN6P_RFC2292)) { 1572 error = EINVAL; 1573 break; 1574 } 1575 #endif 1576 optp = &in6p->in6p_outputopts; 1577 error = ip6_pcbopt(IPV6_HOPLIMIT, 1578 (u_char *)&optval, 1579 sizeof(optval), 1580 optp, 1581 kauth_cred_get(), uproto); 1582 break; 1583 } 1584 1585 case IPV6_RECVHOPLIMIT: 1586 #ifdef RFC2292 1587 /* cannot mix with RFC2292 */ 1588 if (OPTBIT(IN6P_RFC2292)) { 1589 error = EINVAL; 1590 break; 1591 } 1592 #endif 1593 OPTSET(IN6P_HOPLIMIT); 1594 break; 1595 1596 case IPV6_RECVHOPOPTS: 1597 #ifdef RFC2292 1598 /* cannot mix with RFC2292 */ 1599 if (OPTBIT(IN6P_RFC2292)) { 1600 error = EINVAL; 1601 break; 1602 } 1603 #endif 1604 OPTSET(IN6P_HOPOPTS); 1605 break; 1606 1607 case IPV6_RECVDSTOPTS: 1608 #ifdef RFC2292 1609 /* cannot mix with RFC2292 */ 1610 if (OPTBIT(IN6P_RFC2292)) { 1611 error = EINVAL; 1612 break; 1613 } 1614 #endif 1615 OPTSET(IN6P_DSTOPTS); 1616 break; 1617 1618 case IPV6_RECVRTHDRDSTOPTS: 1619 #ifdef RFC2292 1620 /* cannot mix with RFC2292 */ 1621 if (OPTBIT(IN6P_RFC2292)) { 1622 error = EINVAL; 1623 break; 1624 } 1625 #endif 1626 OPTSET(IN6P_RTHDRDSTOPTS); 1627 break; 1628 1629 case IPV6_RECVRTHDR: 1630 #ifdef RFC2292 1631 /* cannot mix with RFC2292 */ 1632 if (OPTBIT(IN6P_RFC2292)) { 1633 error = EINVAL; 1634 break; 1635 } 1636 #endif 1637 OPTSET(IN6P_RTHDR); 1638 break; 1639 1640 case IPV6_FAITH: 1641 OPTSET(IN6P_FAITH); 1642 break; 1643 1644 case IPV6_RECVPATHMTU: 1645 /* 1646 * We ignore this option for TCP 1647 * sockets. 1648 * (RFC3542 leaves this case 1649 * unspecified.) 1650 */ 1651 if (uproto != IPPROTO_TCP) 1652 OPTSET(IN6P_MTU); 1653 break; 1654 1655 case IPV6_V6ONLY: 1656 /* 1657 * make setsockopt(IPV6_V6ONLY) 1658 * available only prior to bind(2). 1659 * see ipng mailing list, Jun 22 2001. 1660 */ 1661 if (in6p->in6p_lport || 1662 !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr)) { 1663 error = EINVAL; 1664 break; 1665 } 1666 #ifdef INET6_BINDV6ONLY 1667 if (!optval) 1668 error = EINVAL; 1669 #else 1670 OPTSET(IN6P_IPV6_V6ONLY); 1671 #endif 1672 break; 1673 case IPV6_RECVTCLASS: 1674 #ifdef RFC2292 1675 /* cannot mix with RFC2292 XXX */ 1676 if (OPTBIT(IN6P_RFC2292)) { 1677 error = EINVAL; 1678 break; 1679 } 1680 #endif 1681 OPTSET(IN6P_TCLASS); 1682 break; 1683 1684 } 1685 break; 1686 1687 case IPV6_OTCLASS: 1688 { 1689 struct ip6_pktopts **optp; 1690 u_int8_t tclass; 1691 1692 error = sockopt_get(sopt, &tclass, sizeof(tclass)); 1693 if (error) 1694 break; 1695 optp = &in6p->in6p_outputopts; 1696 error = ip6_pcbopt(optname, 1697 (u_char *)&tclass, 1698 sizeof(tclass), 1699 optp, 1700 kauth_cred_get(), uproto); 1701 break; 1702 } 1703 1704 case IPV6_TCLASS: 1705 case IPV6_DONTFRAG: 1706 case IPV6_USE_MIN_MTU: 1707 error = sockopt_getint(sopt, &optval); 1708 if (error) 1709 break; 1710 { 1711 struct ip6_pktopts **optp; 1712 optp = &in6p->in6p_outputopts; 1713 error = ip6_pcbopt(optname, 1714 (u_char *)&optval, 1715 sizeof(optval), 1716 optp, 1717 kauth_cred_get(), uproto); 1718 break; 1719 } 1720 1721 #ifdef RFC2292 1722 case IPV6_2292PKTINFO: 1723 case IPV6_2292HOPLIMIT: 1724 case IPV6_2292HOPOPTS: 1725 case IPV6_2292DSTOPTS: 1726 case IPV6_2292RTHDR: 1727 /* RFC 2292 */ 1728 error = sockopt_getint(sopt, &optval); 1729 if (error) 1730 break; 1731 1732 switch (optname) { 1733 case IPV6_2292PKTINFO: 1734 OPTSET2292(IN6P_PKTINFO); 1735 break; 1736 case IPV6_2292HOPLIMIT: 1737 OPTSET2292(IN6P_HOPLIMIT); 1738 break; 1739 case IPV6_2292HOPOPTS: 1740 /* 1741 * Check super-user privilege. 1742 * See comments for IPV6_RECVHOPOPTS. 1743 */ 1744 error = 1745 kauth_authorize_generic(kauth_cred_get(), 1746 KAUTH_GENERIC_ISSUSER, NULL); 1747 if (error) 1748 return (error); 1749 OPTSET2292(IN6P_HOPOPTS); 1750 break; 1751 case IPV6_2292DSTOPTS: 1752 error = 1753 kauth_authorize_generic(kauth_cred_get(), 1754 KAUTH_GENERIC_ISSUSER, NULL); 1755 if (error) 1756 return (error); 1757 OPTSET2292(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); /* XXX */ 1758 break; 1759 case IPV6_2292RTHDR: 1760 OPTSET2292(IN6P_RTHDR); 1761 break; 1762 } 1763 break; 1764 #endif 1765 case IPV6_PKTINFO: 1766 case IPV6_HOPOPTS: 1767 case IPV6_RTHDR: 1768 case IPV6_DSTOPTS: 1769 case IPV6_RTHDRDSTOPTS: 1770 case IPV6_NEXTHOP: { 1771 /* new advanced API (RFC3542) */ 1772 void *optbuf; 1773 int optbuflen; 1774 struct ip6_pktopts **optp; 1775 1776 #ifdef RFC2292 1777 /* cannot mix with RFC2292 */ 1778 if (OPTBIT(IN6P_RFC2292)) { 1779 error = EINVAL; 1780 break; 1781 } 1782 #endif 1783 1784 optbuflen = sopt->sopt_size; 1785 optbuf = malloc(optbuflen, M_IP6OPT, M_NOWAIT); 1786 if (optbuf == NULL) { 1787 error = ENOBUFS; 1788 break; 1789 } 1790 1791 sockopt_get(sopt, optbuf, optbuflen); 1792 optp = &in6p->in6p_outputopts; 1793 error = ip6_pcbopt(optname, optbuf, optbuflen, 1794 optp, kauth_cred_get(), uproto); 1795 break; 1796 } 1797 #undef OPTSET 1798 1799 case IPV6_MULTICAST_IF: 1800 case IPV6_MULTICAST_HOPS: 1801 case IPV6_MULTICAST_LOOP: 1802 case IPV6_JOIN_GROUP: 1803 case IPV6_LEAVE_GROUP: 1804 error = ip6_setmoptions(sopt, &in6p->in6p_moptions); 1805 break; 1806 1807 case IPV6_PORTRANGE: 1808 error = sockopt_getint(sopt, &optval); 1809 if (error) 1810 break; 1811 1812 switch (optval) { 1813 case IPV6_PORTRANGE_DEFAULT: 1814 in6p->in6p_flags &= ~(IN6P_LOWPORT); 1815 in6p->in6p_flags &= ~(IN6P_HIGHPORT); 1816 break; 1817 1818 case IPV6_PORTRANGE_HIGH: 1819 in6p->in6p_flags &= ~(IN6P_LOWPORT); 1820 in6p->in6p_flags |= IN6P_HIGHPORT; 1821 break; 1822 1823 case IPV6_PORTRANGE_LOW: 1824 in6p->in6p_flags &= ~(IN6P_HIGHPORT); 1825 in6p->in6p_flags |= IN6P_LOWPORT; 1826 break; 1827 1828 default: 1829 error = EINVAL; 1830 break; 1831 } 1832 break; 1833 1834 1835 #if defined(IPSEC) || defined(FAST_IPSEC) 1836 case IPV6_IPSEC_POLICY: 1837 error = ipsec6_set_policy(in6p, optname, 1838 sopt->sopt_data, sopt->sopt_size, kauth_cred_get()); 1839 break; 1840 #endif /* IPSEC */ 1841 1842 default: 1843 error = ENOPROTOOPT; 1844 break; 1845 } 1846 break; 1847 1848 case PRCO_GETOPT: 1849 switch (optname) { 1850 #ifdef RFC2292 1851 case IPV6_2292PKTOPTIONS: 1852 /* 1853 * RFC3542 (effectively) deprecated the 1854 * semantics of the 2292-style pktoptions. 1855 * Since it was not reliable in nature (i.e., 1856 * applications had to expect the lack of some 1857 * information after all), it would make sense 1858 * to simplify this part by always returning 1859 * empty data. 1860 */ 1861 break; 1862 #endif 1863 1864 case IPV6_RECVHOPOPTS: 1865 case IPV6_RECVDSTOPTS: 1866 case IPV6_RECVRTHDRDSTOPTS: 1867 case IPV6_UNICAST_HOPS: 1868 case IPV6_RECVPKTINFO: 1869 case IPV6_RECVHOPLIMIT: 1870 case IPV6_RECVRTHDR: 1871 case IPV6_RECVPATHMTU: 1872 1873 case IPV6_FAITH: 1874 case IPV6_V6ONLY: 1875 case IPV6_PORTRANGE: 1876 case IPV6_RECVTCLASS: 1877 switch (optname) { 1878 1879 case IPV6_RECVHOPOPTS: 1880 optval = OPTBIT(IN6P_HOPOPTS); 1881 break; 1882 1883 case IPV6_RECVDSTOPTS: 1884 optval = OPTBIT(IN6P_DSTOPTS); 1885 break; 1886 1887 case IPV6_RECVRTHDRDSTOPTS: 1888 optval = OPTBIT(IN6P_RTHDRDSTOPTS); 1889 break; 1890 1891 case IPV6_UNICAST_HOPS: 1892 optval = in6p->in6p_hops; 1893 break; 1894 1895 case IPV6_RECVPKTINFO: 1896 optval = OPTBIT(IN6P_PKTINFO); 1897 break; 1898 1899 case IPV6_RECVHOPLIMIT: 1900 optval = OPTBIT(IN6P_HOPLIMIT); 1901 break; 1902 1903 case IPV6_RECVRTHDR: 1904 optval = OPTBIT(IN6P_RTHDR); 1905 break; 1906 1907 case IPV6_RECVPATHMTU: 1908 optval = OPTBIT(IN6P_MTU); 1909 break; 1910 1911 case IPV6_FAITH: 1912 optval = OPTBIT(IN6P_FAITH); 1913 break; 1914 1915 case IPV6_V6ONLY: 1916 optval = OPTBIT(IN6P_IPV6_V6ONLY); 1917 break; 1918 1919 case IPV6_PORTRANGE: 1920 { 1921 int flags; 1922 flags = in6p->in6p_flags; 1923 if (flags & IN6P_HIGHPORT) 1924 optval = IPV6_PORTRANGE_HIGH; 1925 else if (flags & IN6P_LOWPORT) 1926 optval = IPV6_PORTRANGE_LOW; 1927 else 1928 optval = 0; 1929 break; 1930 } 1931 case IPV6_RECVTCLASS: 1932 optval = OPTBIT(IN6P_TCLASS); 1933 break; 1934 1935 } 1936 if (error) 1937 break; 1938 error = sockopt_setint(sopt, optval); 1939 break; 1940 1941 case IPV6_PATHMTU: 1942 { 1943 u_long pmtu = 0; 1944 struct ip6_mtuinfo mtuinfo; 1945 struct route *ro = &in6p->in6p_route; 1946 1947 if (!(so->so_state & SS_ISCONNECTED)) 1948 return (ENOTCONN); 1949 /* 1950 * XXX: we dot not consider the case of source 1951 * routing, or optional information to specify 1952 * the outgoing interface. 1953 */ 1954 error = ip6_getpmtu(ro, NULL, NULL, 1955 &in6p->in6p_faddr, &pmtu, NULL); 1956 if (error) 1957 break; 1958 if (pmtu > IPV6_MAXPACKET) 1959 pmtu = IPV6_MAXPACKET; 1960 1961 memset(&mtuinfo, 0, sizeof(mtuinfo)); 1962 mtuinfo.ip6m_mtu = (u_int32_t)pmtu; 1963 optdata = (void *)&mtuinfo; 1964 optdatalen = sizeof(mtuinfo); 1965 if (optdatalen > MCLBYTES) 1966 return (EMSGSIZE); /* XXX */ 1967 error = sockopt_set(sopt, optdata, optdatalen); 1968 break; 1969 } 1970 1971 #ifdef RFC2292 1972 case IPV6_2292PKTINFO: 1973 case IPV6_2292HOPLIMIT: 1974 case IPV6_2292HOPOPTS: 1975 case IPV6_2292RTHDR: 1976 case IPV6_2292DSTOPTS: 1977 switch (optname) { 1978 case IPV6_2292PKTINFO: 1979 optval = OPTBIT(IN6P_PKTINFO); 1980 break; 1981 case IPV6_2292HOPLIMIT: 1982 optval = OPTBIT(IN6P_HOPLIMIT); 1983 break; 1984 case IPV6_2292HOPOPTS: 1985 optval = OPTBIT(IN6P_HOPOPTS); 1986 break; 1987 case IPV6_2292RTHDR: 1988 optval = OPTBIT(IN6P_RTHDR); 1989 break; 1990 case IPV6_2292DSTOPTS: 1991 optval = OPTBIT(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); 1992 break; 1993 } 1994 error = sockopt_setint(sopt, optval); 1995 break; 1996 #endif 1997 case IPV6_PKTINFO: 1998 case IPV6_HOPOPTS: 1999 case IPV6_RTHDR: 2000 case IPV6_DSTOPTS: 2001 case IPV6_RTHDRDSTOPTS: 2002 case IPV6_NEXTHOP: 2003 case IPV6_OTCLASS: 2004 case IPV6_TCLASS: 2005 case IPV6_DONTFRAG: 2006 case IPV6_USE_MIN_MTU: 2007 error = ip6_getpcbopt(in6p->in6p_outputopts, 2008 optname, sopt); 2009 break; 2010 2011 case IPV6_MULTICAST_IF: 2012 case IPV6_MULTICAST_HOPS: 2013 case IPV6_MULTICAST_LOOP: 2014 case IPV6_JOIN_GROUP: 2015 case IPV6_LEAVE_GROUP: 2016 error = ip6_getmoptions(sopt, in6p->in6p_moptions); 2017 break; 2018 2019 #if defined(IPSEC) || defined(FAST_IPSEC) 2020 case IPV6_IPSEC_POLICY: 2021 { 2022 struct mbuf *m = NULL; 2023 2024 /* XXX this will return EINVAL as sopt is empty */ 2025 error = ipsec6_get_policy(in6p, sopt->sopt_data, 2026 sopt->sopt_size, &m); 2027 if (!error) 2028 error = sockopt_setmbuf(sopt, m); 2029 2030 break; 2031 } 2032 #endif /* IPSEC */ 2033 2034 default: 2035 error = ENOPROTOOPT; 2036 break; 2037 } 2038 break; 2039 } 2040 return (error); 2041 } 2042 2043 int 2044 ip6_raw_ctloutput(int op, struct socket *so, struct sockopt *sopt) 2045 { 2046 int error = 0, optval; 2047 const int icmp6off = offsetof(struct icmp6_hdr, icmp6_cksum); 2048 struct in6pcb *in6p = sotoin6pcb(so); 2049 int level, optname; 2050 2051 KASSERT(sopt != NULL); 2052 2053 level = sopt->sopt_level; 2054 optname = sopt->sopt_name; 2055 2056 if (level != IPPROTO_IPV6) { 2057 return ENOPROTOOPT; 2058 } 2059 2060 switch (optname) { 2061 case IPV6_CHECKSUM: 2062 /* 2063 * For ICMPv6 sockets, no modification allowed for checksum 2064 * offset, permit "no change" values to help existing apps. 2065 * 2066 * XXX RFC3542 says: "An attempt to set IPV6_CHECKSUM 2067 * for an ICMPv6 socket will fail." The current 2068 * behavior does not meet RFC3542. 2069 */ 2070 switch (op) { 2071 case PRCO_SETOPT: 2072 error = sockopt_getint(sopt, &optval); 2073 if (error) 2074 break; 2075 if ((optval % 2) != 0) { 2076 /* the API assumes even offset values */ 2077 error = EINVAL; 2078 } else if (so->so_proto->pr_protocol == 2079 IPPROTO_ICMPV6) { 2080 if (optval != icmp6off) 2081 error = EINVAL; 2082 } else 2083 in6p->in6p_cksum = optval; 2084 break; 2085 2086 case PRCO_GETOPT: 2087 if (so->so_proto->pr_protocol == IPPROTO_ICMPV6) 2088 optval = icmp6off; 2089 else 2090 optval = in6p->in6p_cksum; 2091 2092 error = sockopt_setint(sopt, optval); 2093 break; 2094 2095 default: 2096 error = EINVAL; 2097 break; 2098 } 2099 break; 2100 2101 default: 2102 error = ENOPROTOOPT; 2103 break; 2104 } 2105 2106 return (error); 2107 } 2108 2109 #ifdef RFC2292 2110 /* 2111 * Set up IP6 options in pcb for insertion in output packets or 2112 * specifying behavior of outgoing packets. 2113 */ 2114 static int 2115 ip6_pcbopts(struct ip6_pktopts **pktopt, struct socket *so, 2116 struct sockopt *sopt) 2117 { 2118 struct ip6_pktopts *opt = *pktopt; 2119 struct mbuf *m; 2120 int error = 0; 2121 2122 /* turn off any old options. */ 2123 if (opt) { 2124 #ifdef DIAGNOSTIC 2125 if (opt->ip6po_pktinfo || opt->ip6po_nexthop || 2126 opt->ip6po_hbh || opt->ip6po_dest1 || opt->ip6po_dest2 || 2127 opt->ip6po_rhinfo.ip6po_rhi_rthdr) 2128 printf("ip6_pcbopts: all specified options are cleared.\n"); 2129 #endif 2130 ip6_clearpktopts(opt, -1); 2131 } else { 2132 opt = malloc(sizeof(*opt), M_IP6OPT, M_NOWAIT); 2133 if (opt == NULL) 2134 return (ENOBUFS); 2135 } 2136 *pktopt = NULL; 2137 2138 if (sopt == NULL || sopt->sopt_size == 0) { 2139 /* 2140 * Only turning off any previous options, regardless of 2141 * whether the opt is just created or given. 2142 */ 2143 free(opt, M_IP6OPT); 2144 return (0); 2145 } 2146 2147 /* set options specified by user. */ 2148 m = sockopt_getmbuf(sopt); 2149 if (m == NULL) { 2150 free(opt, M_IP6OPT); 2151 return (ENOBUFS); 2152 } 2153 2154 error = ip6_setpktopts(m, opt, NULL, kauth_cred_get(), 2155 so->so_proto->pr_protocol); 2156 m_freem(m); 2157 if (error != 0) { 2158 ip6_clearpktopts(opt, -1); /* XXX: discard all options */ 2159 free(opt, M_IP6OPT); 2160 return (error); 2161 } 2162 *pktopt = opt; 2163 return (0); 2164 } 2165 #endif 2166 2167 /* 2168 * initialize ip6_pktopts. beware that there are non-zero default values in 2169 * the struct. 2170 */ 2171 void 2172 ip6_initpktopts(struct ip6_pktopts *opt) 2173 { 2174 2175 memset(opt, 0, sizeof(*opt)); 2176 opt->ip6po_hlim = -1; /* -1 means default hop limit */ 2177 opt->ip6po_tclass = -1; /* -1 means default traffic class */ 2178 opt->ip6po_minmtu = IP6PO_MINMTU_MCASTONLY; 2179 } 2180 2181 #define sin6tosa(sin6) ((struct sockaddr *)(sin6)) /* XXX */ 2182 static int 2183 ip6_pcbopt(int optname, u_char *buf, int len, struct ip6_pktopts **pktopt, 2184 kauth_cred_t cred, int uproto) 2185 { 2186 struct ip6_pktopts *opt; 2187 2188 if (*pktopt == NULL) { 2189 *pktopt = malloc(sizeof(struct ip6_pktopts), M_IP6OPT, 2190 M_NOWAIT); 2191 if (*pktopt == NULL) 2192 return (ENOBUFS); 2193 2194 ip6_initpktopts(*pktopt); 2195 } 2196 opt = *pktopt; 2197 2198 return (ip6_setpktopt(optname, buf, len, opt, cred, 1, 0, uproto)); 2199 } 2200 2201 static int 2202 ip6_getpcbopt(struct ip6_pktopts *pktopt, int optname, struct sockopt *sopt) 2203 { 2204 void *optdata = NULL; 2205 int optdatalen = 0; 2206 struct ip6_ext *ip6e; 2207 int error = 0; 2208 struct in6_pktinfo null_pktinfo; 2209 int deftclass = 0, on; 2210 int defminmtu = IP6PO_MINMTU_MCASTONLY; 2211 2212 switch (optname) { 2213 case IPV6_PKTINFO: 2214 if (pktopt && pktopt->ip6po_pktinfo) 2215 optdata = (void *)pktopt->ip6po_pktinfo; 2216 else { 2217 /* XXX: we don't have to do this every time... */ 2218 memset(&null_pktinfo, 0, sizeof(null_pktinfo)); 2219 optdata = (void *)&null_pktinfo; 2220 } 2221 optdatalen = sizeof(struct in6_pktinfo); 2222 break; 2223 case IPV6_OTCLASS: 2224 /* XXX */ 2225 return (EINVAL); 2226 case IPV6_TCLASS: 2227 if (pktopt && pktopt->ip6po_tclass >= 0) 2228 optdata = (void *)&pktopt->ip6po_tclass; 2229 else 2230 optdata = (void *)&deftclass; 2231 optdatalen = sizeof(int); 2232 break; 2233 case IPV6_HOPOPTS: 2234 if (pktopt && pktopt->ip6po_hbh) { 2235 optdata = (void *)pktopt->ip6po_hbh; 2236 ip6e = (struct ip6_ext *)pktopt->ip6po_hbh; 2237 optdatalen = (ip6e->ip6e_len + 1) << 3; 2238 } 2239 break; 2240 case IPV6_RTHDR: 2241 if (pktopt && pktopt->ip6po_rthdr) { 2242 optdata = (void *)pktopt->ip6po_rthdr; 2243 ip6e = (struct ip6_ext *)pktopt->ip6po_rthdr; 2244 optdatalen = (ip6e->ip6e_len + 1) << 3; 2245 } 2246 break; 2247 case IPV6_RTHDRDSTOPTS: 2248 if (pktopt && pktopt->ip6po_dest1) { 2249 optdata = (void *)pktopt->ip6po_dest1; 2250 ip6e = (struct ip6_ext *)pktopt->ip6po_dest1; 2251 optdatalen = (ip6e->ip6e_len + 1) << 3; 2252 } 2253 break; 2254 case IPV6_DSTOPTS: 2255 if (pktopt && pktopt->ip6po_dest2) { 2256 optdata = (void *)pktopt->ip6po_dest2; 2257 ip6e = (struct ip6_ext *)pktopt->ip6po_dest2; 2258 optdatalen = (ip6e->ip6e_len + 1) << 3; 2259 } 2260 break; 2261 case IPV6_NEXTHOP: 2262 if (pktopt && pktopt->ip6po_nexthop) { 2263 optdata = (void *)pktopt->ip6po_nexthop; 2264 optdatalen = pktopt->ip6po_nexthop->sa_len; 2265 } 2266 break; 2267 case IPV6_USE_MIN_MTU: 2268 if (pktopt) 2269 optdata = (void *)&pktopt->ip6po_minmtu; 2270 else 2271 optdata = (void *)&defminmtu; 2272 optdatalen = sizeof(int); 2273 break; 2274 case IPV6_DONTFRAG: 2275 if (pktopt && ((pktopt->ip6po_flags) & IP6PO_DONTFRAG)) 2276 on = 1; 2277 else 2278 on = 0; 2279 optdata = (void *)&on; 2280 optdatalen = sizeof(on); 2281 break; 2282 default: /* should not happen */ 2283 #ifdef DIAGNOSTIC 2284 panic("ip6_getpcbopt: unexpected option\n"); 2285 #endif 2286 return (ENOPROTOOPT); 2287 } 2288 2289 error = sockopt_set(sopt, optdata, optdatalen); 2290 2291 return (error); 2292 } 2293 2294 void 2295 ip6_clearpktopts(struct ip6_pktopts *pktopt, int optname) 2296 { 2297 if (optname == -1 || optname == IPV6_PKTINFO) { 2298 if (pktopt->ip6po_pktinfo) 2299 free(pktopt->ip6po_pktinfo, M_IP6OPT); 2300 pktopt->ip6po_pktinfo = NULL; 2301 } 2302 if (optname == -1 || optname == IPV6_HOPLIMIT) 2303 pktopt->ip6po_hlim = -1; 2304 if (optname == -1 || optname == IPV6_TCLASS) 2305 pktopt->ip6po_tclass = -1; 2306 if (optname == -1 || optname == IPV6_NEXTHOP) { 2307 rtcache_free(&pktopt->ip6po_nextroute); 2308 if (pktopt->ip6po_nexthop) 2309 free(pktopt->ip6po_nexthop, M_IP6OPT); 2310 pktopt->ip6po_nexthop = NULL; 2311 } 2312 if (optname == -1 || optname == IPV6_HOPOPTS) { 2313 if (pktopt->ip6po_hbh) 2314 free(pktopt->ip6po_hbh, M_IP6OPT); 2315 pktopt->ip6po_hbh = NULL; 2316 } 2317 if (optname == -1 || optname == IPV6_RTHDRDSTOPTS) { 2318 if (pktopt->ip6po_dest1) 2319 free(pktopt->ip6po_dest1, M_IP6OPT); 2320 pktopt->ip6po_dest1 = NULL; 2321 } 2322 if (optname == -1 || optname == IPV6_RTHDR) { 2323 if (pktopt->ip6po_rhinfo.ip6po_rhi_rthdr) 2324 free(pktopt->ip6po_rhinfo.ip6po_rhi_rthdr, M_IP6OPT); 2325 pktopt->ip6po_rhinfo.ip6po_rhi_rthdr = NULL; 2326 rtcache_free(&pktopt->ip6po_route); 2327 } 2328 if (optname == -1 || optname == IPV6_DSTOPTS) { 2329 if (pktopt->ip6po_dest2) 2330 free(pktopt->ip6po_dest2, M_IP6OPT); 2331 pktopt->ip6po_dest2 = NULL; 2332 } 2333 } 2334 2335 #define PKTOPT_EXTHDRCPY(type) \ 2336 do { \ 2337 if (src->type) { \ 2338 int hlen = (((struct ip6_ext *)src->type)->ip6e_len + 1) << 3;\ 2339 dst->type = malloc(hlen, M_IP6OPT, canwait); \ 2340 if (dst->type == NULL && canwait == M_NOWAIT) \ 2341 goto bad; \ 2342 memcpy(dst->type, src->type, hlen); \ 2343 } \ 2344 } while (/*CONSTCOND*/ 0) 2345 2346 static int 2347 copypktopts(struct ip6_pktopts *dst, struct ip6_pktopts *src, int canwait) 2348 { 2349 dst->ip6po_hlim = src->ip6po_hlim; 2350 dst->ip6po_tclass = src->ip6po_tclass; 2351 dst->ip6po_flags = src->ip6po_flags; 2352 if (src->ip6po_pktinfo) { 2353 dst->ip6po_pktinfo = malloc(sizeof(*dst->ip6po_pktinfo), 2354 M_IP6OPT, canwait); 2355 if (dst->ip6po_pktinfo == NULL && canwait == M_NOWAIT) 2356 goto bad; 2357 *dst->ip6po_pktinfo = *src->ip6po_pktinfo; 2358 } 2359 if (src->ip6po_nexthop) { 2360 dst->ip6po_nexthop = malloc(src->ip6po_nexthop->sa_len, 2361 M_IP6OPT, canwait); 2362 if (dst->ip6po_nexthop == NULL && canwait == M_NOWAIT) 2363 goto bad; 2364 memcpy(dst->ip6po_nexthop, src->ip6po_nexthop, 2365 src->ip6po_nexthop->sa_len); 2366 } 2367 PKTOPT_EXTHDRCPY(ip6po_hbh); 2368 PKTOPT_EXTHDRCPY(ip6po_dest1); 2369 PKTOPT_EXTHDRCPY(ip6po_dest2); 2370 PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */ 2371 return (0); 2372 2373 bad: 2374 if (dst->ip6po_pktinfo) free(dst->ip6po_pktinfo, M_IP6OPT); 2375 if (dst->ip6po_nexthop) free(dst->ip6po_nexthop, M_IP6OPT); 2376 if (dst->ip6po_hbh) free(dst->ip6po_hbh, M_IP6OPT); 2377 if (dst->ip6po_dest1) free(dst->ip6po_dest1, M_IP6OPT); 2378 if (dst->ip6po_dest2) free(dst->ip6po_dest2, M_IP6OPT); 2379 if (dst->ip6po_rthdr) free(dst->ip6po_rthdr, M_IP6OPT); 2380 2381 return (ENOBUFS); 2382 } 2383 #undef PKTOPT_EXTHDRCPY 2384 2385 struct ip6_pktopts * 2386 ip6_copypktopts(struct ip6_pktopts *src, int canwait) 2387 { 2388 int error; 2389 struct ip6_pktopts *dst; 2390 2391 dst = malloc(sizeof(*dst), M_IP6OPT, canwait); 2392 if (dst == NULL && canwait == M_NOWAIT) 2393 return (NULL); 2394 ip6_initpktopts(dst); 2395 2396 if ((error = copypktopts(dst, src, canwait)) != 0) { 2397 free(dst, M_IP6OPT); 2398 return (NULL); 2399 } 2400 2401 return (dst); 2402 } 2403 2404 void 2405 ip6_freepcbopts(struct ip6_pktopts *pktopt) 2406 { 2407 if (pktopt == NULL) 2408 return; 2409 2410 ip6_clearpktopts(pktopt, -1); 2411 2412 free(pktopt, M_IP6OPT); 2413 } 2414 2415 /* 2416 * Set the IP6 multicast options in response to user setsockopt(). 2417 */ 2418 static int 2419 ip6_setmoptions(const struct sockopt *sopt, struct ip6_moptions **im6op) 2420 { 2421 int error = 0; 2422 u_int loop, ifindex; 2423 struct ipv6_mreq mreq; 2424 struct ifnet *ifp; 2425 struct ip6_moptions *im6o = *im6op; 2426 struct route ro; 2427 struct in6_multi_mship *imm; 2428 struct lwp *l = curlwp; /* XXX */ 2429 2430 if (im6o == NULL) { 2431 /* 2432 * No multicast option buffer attached to the pcb; 2433 * allocate one and initialize to default values. 2434 */ 2435 im6o = malloc(sizeof(*im6o), M_IPMOPTS, M_NOWAIT); 2436 if (im6o == NULL) 2437 return (ENOBUFS); 2438 2439 *im6op = im6o; 2440 im6o->im6o_multicast_ifp = NULL; 2441 im6o->im6o_multicast_hlim = ip6_defmcasthlim; 2442 im6o->im6o_multicast_loop = IPV6_DEFAULT_MULTICAST_LOOP; 2443 LIST_INIT(&im6o->im6o_memberships); 2444 } 2445 2446 switch (sopt->sopt_name) { 2447 2448 case IPV6_MULTICAST_IF: 2449 /* 2450 * Select the interface for outgoing multicast packets. 2451 */ 2452 error = sockopt_get(sopt, &ifindex, sizeof(ifindex)); 2453 if (error != 0) 2454 break; 2455 2456 if (ifindex != 0) { 2457 if (if_indexlim <= ifindex || !ifindex2ifnet[ifindex]) { 2458 error = ENXIO; /* XXX EINVAL? */ 2459 break; 2460 } 2461 ifp = ifindex2ifnet[ifindex]; 2462 if ((ifp->if_flags & IFF_MULTICAST) == 0) { 2463 error = EADDRNOTAVAIL; 2464 break; 2465 } 2466 } else 2467 ifp = NULL; 2468 im6o->im6o_multicast_ifp = ifp; 2469 break; 2470 2471 case IPV6_MULTICAST_HOPS: 2472 { 2473 /* 2474 * Set the IP6 hoplimit for outgoing multicast packets. 2475 */ 2476 int optval; 2477 2478 error = sockopt_getint(sopt, &optval); 2479 if (error != 0) 2480 break; 2481 2482 if (optval < -1 || optval >= 256) 2483 error = EINVAL; 2484 else if (optval == -1) 2485 im6o->im6o_multicast_hlim = ip6_defmcasthlim; 2486 else 2487 im6o->im6o_multicast_hlim = optval; 2488 break; 2489 } 2490 2491 case IPV6_MULTICAST_LOOP: 2492 /* 2493 * Set the loopback flag for outgoing multicast packets. 2494 * Must be zero or one. 2495 */ 2496 error = sockopt_get(sopt, &loop, sizeof(loop)); 2497 if (error != 0) 2498 break; 2499 if (loop > 1) { 2500 error = EINVAL; 2501 break; 2502 } 2503 im6o->im6o_multicast_loop = loop; 2504 break; 2505 2506 case IPV6_JOIN_GROUP: 2507 /* 2508 * Add a multicast group membership. 2509 * Group must be a valid IP6 multicast address. 2510 */ 2511 error = sockopt_get(sopt, &mreq, sizeof(mreq)); 2512 if (error != 0) 2513 break; 2514 2515 if (IN6_IS_ADDR_UNSPECIFIED(&mreq.ipv6mr_multiaddr)) { 2516 /* 2517 * We use the unspecified address to specify to accept 2518 * all multicast addresses. Only super user is allowed 2519 * to do this. 2520 */ 2521 if (kauth_authorize_generic(l->l_cred, 2522 KAUTH_GENERIC_ISSUSER, NULL)) 2523 { 2524 error = EACCES; 2525 break; 2526 } 2527 } else if (!IN6_IS_ADDR_MULTICAST(&mreq.ipv6mr_multiaddr)) { 2528 error = EINVAL; 2529 break; 2530 } 2531 2532 /* 2533 * If no interface was explicitly specified, choose an 2534 * appropriate one according to the given multicast address. 2535 */ 2536 if (mreq.ipv6mr_interface == 0) { 2537 struct rtentry *rt; 2538 union { 2539 struct sockaddr dst; 2540 struct sockaddr_in6 dst6; 2541 } u; 2542 2543 /* 2544 * Look up the routing table for the 2545 * address, and choose the outgoing interface. 2546 * XXX: is it a good approach? 2547 */ 2548 memset(&ro, 0, sizeof(ro)); 2549 sockaddr_in6_init(&u.dst6, &mreq.ipv6mr_multiaddr, 0, 2550 0, 0); 2551 rtcache_setdst(&ro, &u.dst); 2552 ifp = (rt = rtcache_init(&ro)) != NULL ? rt->rt_ifp 2553 : NULL; 2554 rtcache_free(&ro); 2555 } else { 2556 /* 2557 * If the interface is specified, validate it. 2558 */ 2559 if (if_indexlim <= mreq.ipv6mr_interface || 2560 !ifindex2ifnet[mreq.ipv6mr_interface]) { 2561 error = ENXIO; /* XXX EINVAL? */ 2562 break; 2563 } 2564 ifp = ifindex2ifnet[mreq.ipv6mr_interface]; 2565 } 2566 2567 /* 2568 * See if we found an interface, and confirm that it 2569 * supports multicast 2570 */ 2571 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) { 2572 error = EADDRNOTAVAIL; 2573 break; 2574 } 2575 2576 if (in6_setscope(&mreq.ipv6mr_multiaddr, ifp, NULL)) { 2577 error = EADDRNOTAVAIL; /* XXX: should not happen */ 2578 break; 2579 } 2580 2581 /* 2582 * See if the membership already exists. 2583 */ 2584 for (imm = im6o->im6o_memberships.lh_first; 2585 imm != NULL; imm = imm->i6mm_chain.le_next) 2586 if (imm->i6mm_maddr->in6m_ifp == ifp && 2587 IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr, 2588 &mreq.ipv6mr_multiaddr)) 2589 break; 2590 if (imm != NULL) { 2591 error = EADDRINUSE; 2592 break; 2593 } 2594 /* 2595 * Everything looks good; add a new record to the multicast 2596 * address list for the given interface. 2597 */ 2598 imm = in6_joingroup(ifp, &mreq.ipv6mr_multiaddr, &error, 0); 2599 if (imm == NULL) 2600 break; 2601 LIST_INSERT_HEAD(&im6o->im6o_memberships, imm, i6mm_chain); 2602 break; 2603 2604 case IPV6_LEAVE_GROUP: 2605 /* 2606 * Drop a multicast group membership. 2607 * Group must be a valid IP6 multicast address. 2608 */ 2609 error = sockopt_get(sopt, &mreq, sizeof(mreq)); 2610 if (error != 0) 2611 break; 2612 2613 /* 2614 * If an interface address was specified, get a pointer 2615 * to its ifnet structure. 2616 */ 2617 if (mreq.ipv6mr_interface != 0) { 2618 if (if_indexlim <= mreq.ipv6mr_interface || 2619 !ifindex2ifnet[mreq.ipv6mr_interface]) { 2620 error = ENXIO; /* XXX EINVAL? */ 2621 break; 2622 } 2623 ifp = ifindex2ifnet[mreq.ipv6mr_interface]; 2624 } else 2625 ifp = NULL; 2626 2627 /* Fill in the scope zone ID */ 2628 if (ifp) { 2629 if (in6_setscope(&mreq.ipv6mr_multiaddr, ifp, NULL)) { 2630 /* XXX: should not happen */ 2631 error = EADDRNOTAVAIL; 2632 break; 2633 } 2634 } else if (mreq.ipv6mr_interface != 0) { 2635 /* 2636 * XXX: This case would happens when the (positive) 2637 * index is in the valid range, but the corresponding 2638 * interface has been detached dynamically. The above 2639 * check probably avoids such case to happen here, but 2640 * we check it explicitly for safety. 2641 */ 2642 error = EADDRNOTAVAIL; 2643 break; 2644 } else { /* ipv6mr_interface == 0 */ 2645 struct sockaddr_in6 sa6_mc; 2646 2647 /* 2648 * The API spec says as follows: 2649 * If the interface index is specified as 0, the 2650 * system may choose a multicast group membership to 2651 * drop by matching the multicast address only. 2652 * On the other hand, we cannot disambiguate the scope 2653 * zone unless an interface is provided. Thus, we 2654 * check if there's ambiguity with the default scope 2655 * zone as the last resort. 2656 */ 2657 sockaddr_in6_init(&sa6_mc, &mreq.ipv6mr_multiaddr, 2658 0, 0, 0); 2659 error = sa6_embedscope(&sa6_mc, ip6_use_defzone); 2660 if (error != 0) 2661 break; 2662 mreq.ipv6mr_multiaddr = sa6_mc.sin6_addr; 2663 } 2664 2665 /* 2666 * Find the membership in the membership list. 2667 */ 2668 for (imm = im6o->im6o_memberships.lh_first; 2669 imm != NULL; imm = imm->i6mm_chain.le_next) { 2670 if ((ifp == NULL || imm->i6mm_maddr->in6m_ifp == ifp) && 2671 IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr, 2672 &mreq.ipv6mr_multiaddr)) 2673 break; 2674 } 2675 if (imm == NULL) { 2676 /* Unable to resolve interface */ 2677 error = EADDRNOTAVAIL; 2678 break; 2679 } 2680 /* 2681 * Give up the multicast address record to which the 2682 * membership points. 2683 */ 2684 LIST_REMOVE(imm, i6mm_chain); 2685 in6_leavegroup(imm); 2686 break; 2687 2688 default: 2689 error = EOPNOTSUPP; 2690 break; 2691 } 2692 2693 /* 2694 * If all options have default values, no need to keep the mbuf. 2695 */ 2696 if (im6o->im6o_multicast_ifp == NULL && 2697 im6o->im6o_multicast_hlim == ip6_defmcasthlim && 2698 im6o->im6o_multicast_loop == IPV6_DEFAULT_MULTICAST_LOOP && 2699 im6o->im6o_memberships.lh_first == NULL) { 2700 free(*im6op, M_IPMOPTS); 2701 *im6op = NULL; 2702 } 2703 2704 return (error); 2705 } 2706 2707 /* 2708 * Return the IP6 multicast options in response to user getsockopt(). 2709 */ 2710 static int 2711 ip6_getmoptions(struct sockopt *sopt, struct ip6_moptions *im6o) 2712 { 2713 u_int optval; 2714 int error; 2715 2716 switch (sopt->sopt_name) { 2717 case IPV6_MULTICAST_IF: 2718 if (im6o == NULL || im6o->im6o_multicast_ifp == NULL) 2719 optval = 0; 2720 else 2721 optval = im6o->im6o_multicast_ifp->if_index; 2722 2723 error = sockopt_set(sopt, &optval, sizeof(optval)); 2724 break; 2725 2726 case IPV6_MULTICAST_HOPS: 2727 if (im6o == NULL) 2728 optval = ip6_defmcasthlim; 2729 else 2730 optval = im6o->im6o_multicast_hlim; 2731 2732 error = sockopt_set(sopt, &optval, sizeof(optval)); 2733 break; 2734 2735 case IPV6_MULTICAST_LOOP: 2736 if (im6o == NULL) 2737 optval = ip6_defmcasthlim; 2738 else 2739 optval = im6o->im6o_multicast_loop; 2740 2741 error = sockopt_set(sopt, &optval, sizeof(optval)); 2742 break; 2743 2744 default: 2745 error = EOPNOTSUPP; 2746 } 2747 2748 return (error); 2749 } 2750 2751 /* 2752 * Discard the IP6 multicast options. 2753 */ 2754 void 2755 ip6_freemoptions(struct ip6_moptions *im6o) 2756 { 2757 struct in6_multi_mship *imm; 2758 2759 if (im6o == NULL) 2760 return; 2761 2762 while ((imm = im6o->im6o_memberships.lh_first) != NULL) { 2763 LIST_REMOVE(imm, i6mm_chain); 2764 in6_leavegroup(imm); 2765 } 2766 free(im6o, M_IPMOPTS); 2767 } 2768 2769 /* 2770 * Set IPv6 outgoing packet options based on advanced API. 2771 */ 2772 int 2773 ip6_setpktopts(struct mbuf *control, struct ip6_pktopts *opt, 2774 struct ip6_pktopts *stickyopt, kauth_cred_t cred, int uproto) 2775 { 2776 struct cmsghdr *cm = 0; 2777 2778 if (control == NULL || opt == NULL) 2779 return (EINVAL); 2780 2781 ip6_initpktopts(opt); 2782 if (stickyopt) { 2783 int error; 2784 2785 /* 2786 * If stickyopt is provided, make a local copy of the options 2787 * for this particular packet, then override them by ancillary 2788 * objects. 2789 * XXX: copypktopts() does not copy the cached route to a next 2790 * hop (if any). This is not very good in terms of efficiency, 2791 * but we can allow this since this option should be rarely 2792 * used. 2793 */ 2794 if ((error = copypktopts(opt, stickyopt, M_NOWAIT)) != 0) 2795 return (error); 2796 } 2797 2798 /* 2799 * XXX: Currently, we assume all the optional information is stored 2800 * in a single mbuf. 2801 */ 2802 if (control->m_next) 2803 return (EINVAL); 2804 2805 /* XXX if cm->cmsg_len is not aligned, control->m_len can become <0 */ 2806 for (; control->m_len > 0; control->m_data += CMSG_ALIGN(cm->cmsg_len), 2807 control->m_len -= CMSG_ALIGN(cm->cmsg_len)) { 2808 int error; 2809 2810 if (control->m_len < CMSG_LEN(0)) 2811 return (EINVAL); 2812 2813 cm = mtod(control, struct cmsghdr *); 2814 if (cm->cmsg_len == 0 || cm->cmsg_len > control->m_len) 2815 return (EINVAL); 2816 if (cm->cmsg_level != IPPROTO_IPV6) 2817 continue; 2818 2819 error = ip6_setpktopt(cm->cmsg_type, CMSG_DATA(cm), 2820 cm->cmsg_len - CMSG_LEN(0), opt, cred, 0, 1, uproto); 2821 if (error) 2822 return (error); 2823 } 2824 2825 return (0); 2826 } 2827 2828 /* 2829 * Set a particular packet option, as a sticky option or an ancillary data 2830 * item. "len" can be 0 only when it's a sticky option. 2831 * We have 4 cases of combination of "sticky" and "cmsg": 2832 * "sticky=0, cmsg=0": impossible 2833 * "sticky=0, cmsg=1": RFC2292 or RFC3542 ancillary data 2834 * "sticky=1, cmsg=0": RFC3542 socket option 2835 * "sticky=1, cmsg=1": RFC2292 socket option 2836 */ 2837 static int 2838 ip6_setpktopt(int optname, u_char *buf, int len, struct ip6_pktopts *opt, 2839 kauth_cred_t cred, int sticky, int cmsg, int uproto) 2840 { 2841 int minmtupolicy; 2842 int error; 2843 2844 if (!sticky && !cmsg) { 2845 #ifdef DIAGNOSTIC 2846 printf("ip6_setpktopt: impossible case\n"); 2847 #endif 2848 return (EINVAL); 2849 } 2850 2851 /* 2852 * IPV6_2292xxx is for backward compatibility to RFC2292, and should 2853 * not be specified in the context of RFC3542. Conversely, 2854 * RFC3542 types should not be specified in the context of RFC2292. 2855 */ 2856 if (!cmsg) { 2857 switch (optname) { 2858 case IPV6_2292PKTINFO: 2859 case IPV6_2292HOPLIMIT: 2860 case IPV6_2292NEXTHOP: 2861 case IPV6_2292HOPOPTS: 2862 case IPV6_2292DSTOPTS: 2863 case IPV6_2292RTHDR: 2864 case IPV6_2292PKTOPTIONS: 2865 return (ENOPROTOOPT); 2866 } 2867 } 2868 if (sticky && cmsg) { 2869 switch (optname) { 2870 case IPV6_PKTINFO: 2871 case IPV6_HOPLIMIT: 2872 case IPV6_NEXTHOP: 2873 case IPV6_HOPOPTS: 2874 case IPV6_DSTOPTS: 2875 case IPV6_RTHDRDSTOPTS: 2876 case IPV6_RTHDR: 2877 case IPV6_USE_MIN_MTU: 2878 case IPV6_DONTFRAG: 2879 case IPV6_OTCLASS: 2880 case IPV6_TCLASS: 2881 return (ENOPROTOOPT); 2882 } 2883 } 2884 2885 switch (optname) { 2886 #ifdef RFC2292 2887 case IPV6_2292PKTINFO: 2888 #endif 2889 case IPV6_PKTINFO: 2890 { 2891 struct ifnet *ifp = NULL; 2892 struct in6_pktinfo *pktinfo; 2893 2894 if (len != sizeof(struct in6_pktinfo)) 2895 return (EINVAL); 2896 2897 pktinfo = (struct in6_pktinfo *)buf; 2898 2899 /* 2900 * An application can clear any sticky IPV6_PKTINFO option by 2901 * doing a "regular" setsockopt with ipi6_addr being 2902 * in6addr_any and ipi6_ifindex being zero. 2903 * [RFC 3542, Section 6] 2904 */ 2905 if (optname == IPV6_PKTINFO && opt->ip6po_pktinfo && 2906 pktinfo->ipi6_ifindex == 0 && 2907 IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) { 2908 ip6_clearpktopts(opt, optname); 2909 break; 2910 } 2911 2912 if (uproto == IPPROTO_TCP && optname == IPV6_PKTINFO && 2913 sticky && !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) { 2914 return (EINVAL); 2915 } 2916 2917 /* validate the interface index if specified. */ 2918 if (pktinfo->ipi6_ifindex >= if_indexlim) { 2919 return (ENXIO); 2920 } 2921 if (pktinfo->ipi6_ifindex) { 2922 ifp = ifindex2ifnet[pktinfo->ipi6_ifindex]; 2923 if (ifp == NULL) 2924 return (ENXIO); 2925 } 2926 2927 /* 2928 * We store the address anyway, and let in6_selectsrc() 2929 * validate the specified address. This is because ipi6_addr 2930 * may not have enough information about its scope zone, and 2931 * we may need additional information (such as outgoing 2932 * interface or the scope zone of a destination address) to 2933 * disambiguate the scope. 2934 * XXX: the delay of the validation may confuse the 2935 * application when it is used as a sticky option. 2936 */ 2937 if (opt->ip6po_pktinfo == NULL) { 2938 opt->ip6po_pktinfo = malloc(sizeof(*pktinfo), 2939 M_IP6OPT, M_NOWAIT); 2940 if (opt->ip6po_pktinfo == NULL) 2941 return (ENOBUFS); 2942 } 2943 memcpy(opt->ip6po_pktinfo, pktinfo, sizeof(*pktinfo)); 2944 break; 2945 } 2946 2947 #ifdef RFC2292 2948 case IPV6_2292HOPLIMIT: 2949 #endif 2950 case IPV6_HOPLIMIT: 2951 { 2952 int *hlimp; 2953 2954 /* 2955 * RFC 3542 deprecated the usage of sticky IPV6_HOPLIMIT 2956 * to simplify the ordering among hoplimit options. 2957 */ 2958 if (optname == IPV6_HOPLIMIT && sticky) 2959 return (ENOPROTOOPT); 2960 2961 if (len != sizeof(int)) 2962 return (EINVAL); 2963 hlimp = (int *)buf; 2964 if (*hlimp < -1 || *hlimp > 255) 2965 return (EINVAL); 2966 2967 opt->ip6po_hlim = *hlimp; 2968 break; 2969 } 2970 2971 case IPV6_OTCLASS: 2972 if (len != sizeof(u_int8_t)) 2973 return (EINVAL); 2974 2975 opt->ip6po_tclass = *(u_int8_t *)buf; 2976 break; 2977 2978 case IPV6_TCLASS: 2979 { 2980 int tclass; 2981 2982 if (len != sizeof(int)) 2983 return (EINVAL); 2984 tclass = *(int *)buf; 2985 if (tclass < -1 || tclass > 255) 2986 return (EINVAL); 2987 2988 opt->ip6po_tclass = tclass; 2989 break; 2990 } 2991 2992 #ifdef RFC2292 2993 case IPV6_2292NEXTHOP: 2994 #endif 2995 case IPV6_NEXTHOP: 2996 error = kauth_authorize_generic(cred, KAUTH_GENERIC_ISSUSER, 2997 NULL); 2998 if (error) 2999 return (error); 3000 3001 if (len == 0) { /* just remove the option */ 3002 ip6_clearpktopts(opt, IPV6_NEXTHOP); 3003 break; 3004 } 3005 3006 /* check if cmsg_len is large enough for sa_len */ 3007 if (len < sizeof(struct sockaddr) || len < *buf) 3008 return (EINVAL); 3009 3010 switch (((struct sockaddr *)buf)->sa_family) { 3011 case AF_INET6: 3012 { 3013 struct sockaddr_in6 *sa6 = (struct sockaddr_in6 *)buf; 3014 3015 if (sa6->sin6_len != sizeof(struct sockaddr_in6)) 3016 return (EINVAL); 3017 3018 if (IN6_IS_ADDR_UNSPECIFIED(&sa6->sin6_addr) || 3019 IN6_IS_ADDR_MULTICAST(&sa6->sin6_addr)) { 3020 return (EINVAL); 3021 } 3022 if ((error = sa6_embedscope(sa6, ip6_use_defzone)) 3023 != 0) { 3024 return (error); 3025 } 3026 break; 3027 } 3028 case AF_LINK: /* eventually be supported? */ 3029 default: 3030 return (EAFNOSUPPORT); 3031 } 3032 3033 /* turn off the previous option, then set the new option. */ 3034 ip6_clearpktopts(opt, IPV6_NEXTHOP); 3035 opt->ip6po_nexthop = malloc(*buf, M_IP6OPT, M_NOWAIT); 3036 if (opt->ip6po_nexthop == NULL) 3037 return (ENOBUFS); 3038 memcpy(opt->ip6po_nexthop, buf, *buf); 3039 break; 3040 3041 #ifdef RFC2292 3042 case IPV6_2292HOPOPTS: 3043 #endif 3044 case IPV6_HOPOPTS: 3045 { 3046 struct ip6_hbh *hbh; 3047 int hbhlen; 3048 3049 /* 3050 * XXX: We don't allow a non-privileged user to set ANY HbH 3051 * options, since per-option restriction has too much 3052 * overhead. 3053 */ 3054 error = kauth_authorize_generic(cred, KAUTH_GENERIC_ISSUSER, 3055 NULL); 3056 if (error) 3057 return (error); 3058 3059 if (len == 0) { 3060 ip6_clearpktopts(opt, IPV6_HOPOPTS); 3061 break; /* just remove the option */ 3062 } 3063 3064 /* message length validation */ 3065 if (len < sizeof(struct ip6_hbh)) 3066 return (EINVAL); 3067 hbh = (struct ip6_hbh *)buf; 3068 hbhlen = (hbh->ip6h_len + 1) << 3; 3069 if (len != hbhlen) 3070 return (EINVAL); 3071 3072 /* turn off the previous option, then set the new option. */ 3073 ip6_clearpktopts(opt, IPV6_HOPOPTS); 3074 opt->ip6po_hbh = malloc(hbhlen, M_IP6OPT, M_NOWAIT); 3075 if (opt->ip6po_hbh == NULL) 3076 return (ENOBUFS); 3077 memcpy(opt->ip6po_hbh, hbh, hbhlen); 3078 3079 break; 3080 } 3081 3082 #ifdef RFC2292 3083 case IPV6_2292DSTOPTS: 3084 #endif 3085 case IPV6_DSTOPTS: 3086 case IPV6_RTHDRDSTOPTS: 3087 { 3088 struct ip6_dest *dest, **newdest = NULL; 3089 int destlen; 3090 3091 /* XXX: see the comment for IPV6_HOPOPTS */ 3092 error = kauth_authorize_generic(cred, KAUTH_GENERIC_ISSUSER, 3093 NULL); 3094 if (error) 3095 return (error); 3096 3097 if (len == 0) { 3098 ip6_clearpktopts(opt, optname); 3099 break; /* just remove the option */ 3100 } 3101 3102 /* message length validation */ 3103 if (len < sizeof(struct ip6_dest)) 3104 return (EINVAL); 3105 dest = (struct ip6_dest *)buf; 3106 destlen = (dest->ip6d_len + 1) << 3; 3107 if (len != destlen) 3108 return (EINVAL); 3109 /* 3110 * Determine the position that the destination options header 3111 * should be inserted; before or after the routing header. 3112 */ 3113 switch (optname) { 3114 case IPV6_2292DSTOPTS: 3115 /* 3116 * The old advanced API is ambiguous on this point. 3117 * Our approach is to determine the position based 3118 * according to the existence of a routing header. 3119 * Note, however, that this depends on the order of the 3120 * extension headers in the ancillary data; the 1st 3121 * part of the destination options header must appear 3122 * before the routing header in the ancillary data, 3123 * too. 3124 * RFC3542 solved the ambiguity by introducing 3125 * separate ancillary data or option types. 3126 */ 3127 if (opt->ip6po_rthdr == NULL) 3128 newdest = &opt->ip6po_dest1; 3129 else 3130 newdest = &opt->ip6po_dest2; 3131 break; 3132 case IPV6_RTHDRDSTOPTS: 3133 newdest = &opt->ip6po_dest1; 3134 break; 3135 case IPV6_DSTOPTS: 3136 newdest = &opt->ip6po_dest2; 3137 break; 3138 } 3139 3140 /* turn off the previous option, then set the new option. */ 3141 ip6_clearpktopts(opt, optname); 3142 *newdest = malloc(destlen, M_IP6OPT, M_NOWAIT); 3143 if (*newdest == NULL) 3144 return (ENOBUFS); 3145 memcpy(*newdest, dest, destlen); 3146 3147 break; 3148 } 3149 3150 #ifdef RFC2292 3151 case IPV6_2292RTHDR: 3152 #endif 3153 case IPV6_RTHDR: 3154 { 3155 struct ip6_rthdr *rth; 3156 int rthlen; 3157 3158 if (len == 0) { 3159 ip6_clearpktopts(opt, IPV6_RTHDR); 3160 break; /* just remove the option */ 3161 } 3162 3163 /* message length validation */ 3164 if (len < sizeof(struct ip6_rthdr)) 3165 return (EINVAL); 3166 rth = (struct ip6_rthdr *)buf; 3167 rthlen = (rth->ip6r_len + 1) << 3; 3168 if (len != rthlen) 3169 return (EINVAL); 3170 switch (rth->ip6r_type) { 3171 case IPV6_RTHDR_TYPE_0: 3172 if (rth->ip6r_len == 0) /* must contain one addr */ 3173 return (EINVAL); 3174 if (rth->ip6r_len % 2) /* length must be even */ 3175 return (EINVAL); 3176 if (rth->ip6r_len / 2 != rth->ip6r_segleft) 3177 return (EINVAL); 3178 break; 3179 default: 3180 return (EINVAL); /* not supported */ 3181 } 3182 /* turn off the previous option */ 3183 ip6_clearpktopts(opt, IPV6_RTHDR); 3184 opt->ip6po_rthdr = malloc(rthlen, M_IP6OPT, M_NOWAIT); 3185 if (opt->ip6po_rthdr == NULL) 3186 return (ENOBUFS); 3187 memcpy(opt->ip6po_rthdr, rth, rthlen); 3188 break; 3189 } 3190 3191 case IPV6_USE_MIN_MTU: 3192 if (len != sizeof(int)) 3193 return (EINVAL); 3194 minmtupolicy = *(int *)buf; 3195 if (minmtupolicy != IP6PO_MINMTU_MCASTONLY && 3196 minmtupolicy != IP6PO_MINMTU_DISABLE && 3197 minmtupolicy != IP6PO_MINMTU_ALL) { 3198 return (EINVAL); 3199 } 3200 opt->ip6po_minmtu = minmtupolicy; 3201 break; 3202 3203 case IPV6_DONTFRAG: 3204 if (len != sizeof(int)) 3205 return (EINVAL); 3206 3207 if (uproto == IPPROTO_TCP || *(int *)buf == 0) { 3208 /* 3209 * we ignore this option for TCP sockets. 3210 * (RFC3542 leaves this case unspecified.) 3211 */ 3212 opt->ip6po_flags &= ~IP6PO_DONTFRAG; 3213 } else 3214 opt->ip6po_flags |= IP6PO_DONTFRAG; 3215 break; 3216 3217 default: 3218 return (ENOPROTOOPT); 3219 } /* end of switch */ 3220 3221 return (0); 3222 } 3223 3224 /* 3225 * Routine called from ip6_output() to loop back a copy of an IP6 multicast 3226 * packet to the input queue of a specified interface. Note that this 3227 * calls the output routine of the loopback "driver", but with an interface 3228 * pointer that might NOT be lo0ifp -- easier than replicating that code here. 3229 */ 3230 void 3231 ip6_mloopback(struct ifnet *ifp, struct mbuf *m, 3232 const struct sockaddr_in6 *dst) 3233 { 3234 struct mbuf *copym; 3235 struct ip6_hdr *ip6; 3236 3237 copym = m_copy(m, 0, M_COPYALL); 3238 if (copym == NULL) 3239 return; 3240 3241 /* 3242 * Make sure to deep-copy IPv6 header portion in case the data 3243 * is in an mbuf cluster, so that we can safely override the IPv6 3244 * header portion later. 3245 */ 3246 if ((copym->m_flags & M_EXT) != 0 || 3247 copym->m_len < sizeof(struct ip6_hdr)) { 3248 copym = m_pullup(copym, sizeof(struct ip6_hdr)); 3249 if (copym == NULL) 3250 return; 3251 } 3252 3253 #ifdef DIAGNOSTIC 3254 if (copym->m_len < sizeof(*ip6)) { 3255 m_freem(copym); 3256 return; 3257 } 3258 #endif 3259 3260 ip6 = mtod(copym, struct ip6_hdr *); 3261 /* 3262 * clear embedded scope identifiers if necessary. 3263 * in6_clearscope will touch the addresses only when necessary. 3264 */ 3265 in6_clearscope(&ip6->ip6_src); 3266 in6_clearscope(&ip6->ip6_dst); 3267 3268 (void)looutput(ifp, copym, (const struct sockaddr *)dst, NULL); 3269 } 3270 3271 /* 3272 * Chop IPv6 header off from the payload. 3273 */ 3274 static int 3275 ip6_splithdr(struct mbuf *m, struct ip6_exthdrs *exthdrs) 3276 { 3277 struct mbuf *mh; 3278 struct ip6_hdr *ip6; 3279 3280 ip6 = mtod(m, struct ip6_hdr *); 3281 if (m->m_len > sizeof(*ip6)) { 3282 MGETHDR(mh, M_DONTWAIT, MT_HEADER); 3283 if (mh == 0) { 3284 m_freem(m); 3285 return ENOBUFS; 3286 } 3287 M_MOVE_PKTHDR(mh, m); 3288 MH_ALIGN(mh, sizeof(*ip6)); 3289 m->m_len -= sizeof(*ip6); 3290 m->m_data += sizeof(*ip6); 3291 mh->m_next = m; 3292 m = mh; 3293 m->m_len = sizeof(*ip6); 3294 bcopy((void *)ip6, mtod(m, void *), sizeof(*ip6)); 3295 } 3296 exthdrs->ip6e_ip6 = m; 3297 return 0; 3298 } 3299 3300 /* 3301 * Compute IPv6 extension header length. 3302 */ 3303 int 3304 ip6_optlen(struct in6pcb *in6p) 3305 { 3306 int len; 3307 3308 if (!in6p->in6p_outputopts) 3309 return 0; 3310 3311 len = 0; 3312 #define elen(x) \ 3313 (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0) 3314 3315 len += elen(in6p->in6p_outputopts->ip6po_hbh); 3316 len += elen(in6p->in6p_outputopts->ip6po_dest1); 3317 len += elen(in6p->in6p_outputopts->ip6po_rthdr); 3318 len += elen(in6p->in6p_outputopts->ip6po_dest2); 3319 return len; 3320 #undef elen 3321 } 3322