ip6_output.c revision 1.119
1 /* $NetBSD: ip6_output.c,v 1.119 2007/05/23 17:15:02 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 1.119 2007/05/23 17:15:02 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/in6_offload.h> 95 #include <netinet6/ip6_var.h> 96 #include <netinet6/in6_pcb.h> 97 #include <netinet6/nd6.h> 98 #include <netinet6/ip6protosw.h> 99 #include <netinet6/scope6_var.h> 100 101 #ifdef IPSEC 102 #include <netinet6/ipsec.h> 103 #include <netkey/key.h> 104 #endif /* IPSEC */ 105 106 #ifdef FAST_IPSEC 107 #include <netipsec/ipsec.h> 108 #include <netipsec/ipsec6.h> 109 #include <netipsec/key.h> 110 #include <netipsec/xform.h> 111 #endif 112 113 114 #include <net/net_osdep.h> 115 116 #ifdef PFIL_HOOKS 117 extern struct pfil_head inet6_pfil_hook; /* XXX */ 118 #endif 119 120 struct ip6_exthdrs { 121 struct mbuf *ip6e_ip6; 122 struct mbuf *ip6e_hbh; 123 struct mbuf *ip6e_dest1; 124 struct mbuf *ip6e_rthdr; 125 struct mbuf *ip6e_dest2; 126 }; 127 128 static int ip6_pcbopt __P((int, u_char *, int, struct ip6_pktopts **, 129 int, int)); 130 static int ip6_getpcbopt __P((struct ip6_pktopts *, int, struct mbuf **)); 131 static int ip6_setpktopt __P((int, u_char *, int, struct ip6_pktopts *, int, 132 int, int, int)); 133 static int ip6_setmoptions __P((int, struct ip6_moptions **, struct mbuf *)); 134 static int ip6_getmoptions __P((int, struct ip6_moptions *, struct mbuf **)); 135 static int ip6_copyexthdr __P((struct mbuf **, void *, int)); 136 static int ip6_insertfraghdr __P((struct mbuf *, struct mbuf *, int, 137 struct ip6_frag **)); 138 static int ip6_insert_jumboopt __P((struct ip6_exthdrs *, u_int32_t)); 139 static int ip6_splithdr __P((struct mbuf *, struct ip6_exthdrs *)); 140 static int ip6_getpmtu(struct route *, struct route *, struct ifnet *, 141 const struct in6_addr *, u_long *, int *); 142 static int copypktopts __P((struct ip6_pktopts *, struct ip6_pktopts *, int)); 143 144 #ifdef RFC2292 145 static int ip6_pcbopts __P((struct ip6_pktopts **, struct mbuf *, 146 struct socket *)); 147 #endif 148 149 #define IN6_NEED_CHECKSUM(ifp, csum_flags) \ 150 (__predict_true(((ifp)->if_flags & IFF_LOOPBACK) == 0 || \ 151 (((csum_flags) & M_CSUM_UDPv6) != 0 && udp_do_loopback_cksum) || \ 152 (((csum_flags) & M_CSUM_TCPv6) != 0 && tcp_do_loopback_cksum))) 153 154 /* 155 * IP6 output. The packet in mbuf chain m contains a skeletal IP6 156 * header (with pri, len, nxt, hlim, src, dst). 157 * This function may modify ver and hlim only. 158 * The mbuf chain containing the packet will be freed. 159 * The mbuf opt, if present, will not be freed. 160 * 161 * type of "mtu": rt_rmx.rmx_mtu is u_long, ifnet.ifr_mtu is int, and 162 * nd_ifinfo.linkmtu is u_int32_t. so we use u_long to hold largest one, 163 * which is rt_rmx.rmx_mtu. 164 */ 165 int 166 ip6_output( 167 struct mbuf *m0, 168 struct ip6_pktopts *opt, 169 struct route *ro, 170 int flags, 171 struct ip6_moptions *im6o, 172 struct socket *so, 173 struct ifnet **ifpp /* XXX: just for statistics */ 174 ) 175 { 176 struct ip6_hdr *ip6, *mhip6; 177 struct ifnet *ifp, *origifp; 178 struct mbuf *m = m0; 179 int hlen, tlen, len, off; 180 bool tso; 181 struct route ip6route; 182 struct rtentry *rt = NULL; 183 const struct sockaddr_in6 *dst = NULL; 184 struct sockaddr_in6 src_sa, dst_sa; 185 int error = 0; 186 struct in6_ifaddr *ia = NULL; 187 u_long mtu; 188 int alwaysfrag, dontfrag; 189 u_int32_t optlen = 0, plen = 0, unfragpartlen = 0; 190 struct ip6_exthdrs exthdrs; 191 struct in6_addr finaldst, src0, dst0; 192 u_int32_t zone; 193 struct route *ro_pmtu = NULL; 194 int hdrsplit = 0; 195 int needipsec = 0; 196 #ifdef IPSEC 197 int needipsectun = 0; 198 struct secpolicy *sp = NULL; 199 200 ip6 = mtod(m, struct ip6_hdr *); 201 #endif /* IPSEC */ 202 #ifdef FAST_IPSEC 203 struct secpolicy *sp = NULL; 204 int s; 205 #endif 206 207 208 #ifdef DIAGNOSTIC 209 if ((m->m_flags & M_PKTHDR) == 0) 210 panic("ip6_output: no HDR"); 211 212 if ((m->m_pkthdr.csum_flags & 213 (M_CSUM_TCPv4|M_CSUM_UDPv4|M_CSUM_TSOv4)) != 0) { 214 panic("ip6_output: IPv4 checksum offload flags: %d", 215 m->m_pkthdr.csum_flags); 216 } 217 218 if ((m->m_pkthdr.csum_flags & (M_CSUM_TCPv6|M_CSUM_UDPv6)) == 219 (M_CSUM_TCPv6|M_CSUM_UDPv6)) { 220 panic("ip6_output: conflicting checksum offload flags: %d", 221 m->m_pkthdr.csum_flags); 222 } 223 #endif 224 225 M_CSUM_DATA_IPv6_HL_SET(m->m_pkthdr.csum_data, sizeof(struct ip6_hdr)); 226 227 #define MAKE_EXTHDR(hp, mp) \ 228 do { \ 229 if (hp) { \ 230 struct ip6_ext *eh = (struct ip6_ext *)(hp); \ 231 error = ip6_copyexthdr((mp), (void *)(hp), \ 232 ((eh)->ip6e_len + 1) << 3); \ 233 if (error) \ 234 goto freehdrs; \ 235 } \ 236 } while (/*CONSTCOND*/ 0) 237 238 bzero(&exthdrs, sizeof(exthdrs)); 239 if (opt) { 240 /* Hop-by-Hop options header */ 241 MAKE_EXTHDR(opt->ip6po_hbh, &exthdrs.ip6e_hbh); 242 /* Destination options header(1st part) */ 243 MAKE_EXTHDR(opt->ip6po_dest1, &exthdrs.ip6e_dest1); 244 /* Routing header */ 245 MAKE_EXTHDR(opt->ip6po_rthdr, &exthdrs.ip6e_rthdr); 246 /* Destination options header(2nd part) */ 247 MAKE_EXTHDR(opt->ip6po_dest2, &exthdrs.ip6e_dest2); 248 } 249 250 #ifdef IPSEC 251 if ((flags & IPV6_FORWARDING) != 0) { 252 needipsec = 0; 253 goto skippolicycheck; 254 } 255 256 /* get a security policy for this packet */ 257 if (so == NULL) 258 sp = ipsec6_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND, 0, &error); 259 else { 260 if (IPSEC_PCB_SKIP_IPSEC(sotoinpcb_hdr(so)->inph_sp, 261 IPSEC_DIR_OUTBOUND)) { 262 needipsec = 0; 263 goto skippolicycheck; 264 } 265 sp = ipsec6_getpolicybysock(m, IPSEC_DIR_OUTBOUND, so, &error); 266 } 267 268 if (sp == NULL) { 269 ipsec6stat.out_inval++; 270 goto freehdrs; 271 } 272 273 error = 0; 274 275 /* check policy */ 276 switch (sp->policy) { 277 case IPSEC_POLICY_DISCARD: 278 /* 279 * This packet is just discarded. 280 */ 281 ipsec6stat.out_polvio++; 282 goto freehdrs; 283 284 case IPSEC_POLICY_BYPASS: 285 case IPSEC_POLICY_NONE: 286 /* no need to do IPsec. */ 287 needipsec = 0; 288 break; 289 290 case IPSEC_POLICY_IPSEC: 291 if (sp->req == NULL) { 292 /* XXX should be panic ? */ 293 printf("ip6_output: No IPsec request specified.\n"); 294 error = EINVAL; 295 goto freehdrs; 296 } 297 needipsec = 1; 298 break; 299 300 case IPSEC_POLICY_ENTRUST: 301 default: 302 printf("ip6_output: Invalid policy found. %d\n", sp->policy); 303 } 304 305 skippolicycheck:; 306 #endif /* IPSEC */ 307 308 /* 309 * Calculate the total length of the extension header chain. 310 * Keep the length of the unfragmentable part for fragmentation. 311 */ 312 optlen = 0; 313 if (exthdrs.ip6e_hbh) optlen += exthdrs.ip6e_hbh->m_len; 314 if (exthdrs.ip6e_dest1) optlen += exthdrs.ip6e_dest1->m_len; 315 if (exthdrs.ip6e_rthdr) optlen += exthdrs.ip6e_rthdr->m_len; 316 unfragpartlen = optlen + sizeof(struct ip6_hdr); 317 /* NOTE: we don't add AH/ESP length here. do that later. */ 318 if (exthdrs.ip6e_dest2) optlen += exthdrs.ip6e_dest2->m_len; 319 320 #ifdef FAST_IPSEC 321 /* Check the security policy (SP) for the packet */ 322 323 /* XXX For moment, we doesn't support packet with extented action */ 324 if (optlen !=0) 325 goto freehdrs; 326 327 sp = ipsec6_check_policy(m,so,flags,&needipsec,&error); 328 if (error != 0) { 329 /* 330 * Hack: -EINVAL is used to signal that a packet 331 * should be silently discarded. This is typically 332 * because we asked key management for an SA and 333 * it was delayed (e.g. kicked up to IKE). 334 */ 335 if (error == -EINVAL) 336 error = 0; 337 goto freehdrs; 338 } 339 #endif /* FAST_IPSEC */ 340 341 342 if (needipsec && 343 (m->m_pkthdr.csum_flags & (M_CSUM_UDPv6|M_CSUM_TCPv6)) != 0) { 344 in6_delayed_cksum(m); 345 m->m_pkthdr.csum_flags &= ~(M_CSUM_UDPv6|M_CSUM_TCPv6); 346 } 347 348 349 /* 350 * If we need IPsec, or there is at least one extension header, 351 * separate IP6 header from the payload. 352 */ 353 if ((needipsec || optlen) && !hdrsplit) { 354 if ((error = ip6_splithdr(m, &exthdrs)) != 0) { 355 m = NULL; 356 goto freehdrs; 357 } 358 m = exthdrs.ip6e_ip6; 359 hdrsplit++; 360 } 361 362 /* adjust pointer */ 363 ip6 = mtod(m, struct ip6_hdr *); 364 365 /* adjust mbuf packet header length */ 366 m->m_pkthdr.len += optlen; 367 plen = m->m_pkthdr.len - sizeof(*ip6); 368 369 /* If this is a jumbo payload, insert a jumbo payload option. */ 370 if (plen > IPV6_MAXPACKET) { 371 if (!hdrsplit) { 372 if ((error = ip6_splithdr(m, &exthdrs)) != 0) { 373 m = NULL; 374 goto freehdrs; 375 } 376 m = exthdrs.ip6e_ip6; 377 hdrsplit++; 378 } 379 /* adjust pointer */ 380 ip6 = mtod(m, struct ip6_hdr *); 381 if ((error = ip6_insert_jumboopt(&exthdrs, plen)) != 0) 382 goto freehdrs; 383 optlen += 8; /* XXX JUMBOOPTLEN */ 384 ip6->ip6_plen = 0; 385 } else 386 ip6->ip6_plen = htons(plen); 387 388 /* 389 * Concatenate headers and fill in next header fields. 390 * Here we have, on "m" 391 * IPv6 payload 392 * and we insert headers accordingly. Finally, we should be getting: 393 * IPv6 hbh dest1 rthdr ah* [esp* dest2 payload] 394 * 395 * during the header composing process, "m" points to IPv6 header. 396 * "mprev" points to an extension header prior to esp. 397 */ 398 { 399 u_char *nexthdrp = &ip6->ip6_nxt; 400 struct mbuf *mprev = m; 401 402 /* 403 * we treat dest2 specially. this makes IPsec processing 404 * much easier. the goal here is to make mprev point the 405 * mbuf prior to dest2. 406 * 407 * result: IPv6 dest2 payload 408 * m and mprev will point to IPv6 header. 409 */ 410 if (exthdrs.ip6e_dest2) { 411 if (!hdrsplit) 412 panic("assumption failed: hdr not split"); 413 exthdrs.ip6e_dest2->m_next = m->m_next; 414 m->m_next = exthdrs.ip6e_dest2; 415 *mtod(exthdrs.ip6e_dest2, u_char *) = ip6->ip6_nxt; 416 ip6->ip6_nxt = IPPROTO_DSTOPTS; 417 } 418 419 #define MAKE_CHAIN(m, mp, p, i)\ 420 do {\ 421 if (m) {\ 422 if (!hdrsplit) \ 423 panic("assumption failed: hdr not split"); \ 424 *mtod((m), u_char *) = *(p);\ 425 *(p) = (i);\ 426 p = mtod((m), u_char *);\ 427 (m)->m_next = (mp)->m_next;\ 428 (mp)->m_next = (m);\ 429 (mp) = (m);\ 430 }\ 431 } while (/*CONSTCOND*/ 0) 432 /* 433 * result: IPv6 hbh dest1 rthdr dest2 payload 434 * m will point to IPv6 header. mprev will point to the 435 * extension header prior to dest2 (rthdr in the above case). 436 */ 437 MAKE_CHAIN(exthdrs.ip6e_hbh, mprev, nexthdrp, IPPROTO_HOPOPTS); 438 MAKE_CHAIN(exthdrs.ip6e_dest1, mprev, nexthdrp, 439 IPPROTO_DSTOPTS); 440 MAKE_CHAIN(exthdrs.ip6e_rthdr, mprev, nexthdrp, 441 IPPROTO_ROUTING); 442 443 M_CSUM_DATA_IPv6_HL_SET(m->m_pkthdr.csum_data, 444 sizeof(struct ip6_hdr) + optlen); 445 446 #ifdef IPSEC 447 if (!needipsec) 448 goto skip_ipsec2; 449 450 /* 451 * pointers after IPsec headers are not valid any more. 452 * other pointers need a great care too. 453 * (IPsec routines should not mangle mbufs prior to AH/ESP) 454 */ 455 exthdrs.ip6e_dest2 = NULL; 456 457 { 458 struct ip6_rthdr *rh = NULL; 459 int segleft_org = 0; 460 struct ipsec_output_state state; 461 462 if (exthdrs.ip6e_rthdr) { 463 rh = mtod(exthdrs.ip6e_rthdr, struct ip6_rthdr *); 464 segleft_org = rh->ip6r_segleft; 465 rh->ip6r_segleft = 0; 466 } 467 468 bzero(&state, sizeof(state)); 469 state.m = m; 470 error = ipsec6_output_trans(&state, nexthdrp, mprev, sp, flags, 471 &needipsectun); 472 m = state.m; 473 if (error) { 474 rh = mtod(exthdrs.ip6e_rthdr, struct ip6_rthdr *); 475 /* mbuf is already reclaimed in ipsec6_output_trans. */ 476 m = NULL; 477 switch (error) { 478 case EHOSTUNREACH: 479 case ENETUNREACH: 480 case EMSGSIZE: 481 case ENOBUFS: 482 case ENOMEM: 483 break; 484 default: 485 printf("ip6_output (ipsec): error code %d\n", error); 486 /* FALLTHROUGH */ 487 case ENOENT: 488 /* don't show these error codes to the user */ 489 error = 0; 490 break; 491 } 492 goto bad; 493 } 494 if (exthdrs.ip6e_rthdr) { 495 /* ah6_output doesn't modify mbuf chain */ 496 rh->ip6r_segleft = segleft_org; 497 } 498 } 499 skip_ipsec2:; 500 #endif 501 } 502 503 /* 504 * If there is a routing header, replace destination address field 505 * with the first hop of the routing header. 506 */ 507 if (exthdrs.ip6e_rthdr) { 508 struct ip6_rthdr *rh; 509 struct ip6_rthdr0 *rh0; 510 struct in6_addr *addr; 511 struct sockaddr_in6 sa; 512 513 rh = (struct ip6_rthdr *)(mtod(exthdrs.ip6e_rthdr, 514 struct ip6_rthdr *)); 515 finaldst = ip6->ip6_dst; 516 switch (rh->ip6r_type) { 517 case IPV6_RTHDR_TYPE_0: 518 rh0 = (struct ip6_rthdr0 *)rh; 519 addr = (struct in6_addr *)(rh0 + 1); 520 521 /* 522 * construct a sockaddr_in6 form of 523 * the first hop. 524 * 525 * XXX: we may not have enough 526 * information about its scope zone; 527 * there is no standard API to pass 528 * the information from the 529 * application. 530 */ 531 bzero(&sa, sizeof(sa)); 532 sa.sin6_family = AF_INET6; 533 sa.sin6_len = sizeof(sa); 534 sa.sin6_addr = addr[0]; 535 if ((error = sa6_embedscope(&sa, 536 ip6_use_defzone)) != 0) { 537 goto bad; 538 } 539 ip6->ip6_dst = sa.sin6_addr; 540 (void)memmove(&addr[0], &addr[1], 541 sizeof(struct in6_addr) * 542 (rh0->ip6r0_segleft - 1)); 543 addr[rh0->ip6r0_segleft - 1] = finaldst; 544 /* XXX */ 545 in6_clearscope(addr + rh0->ip6r0_segleft - 1); 546 break; 547 default: /* is it possible? */ 548 error = EINVAL; 549 goto bad; 550 } 551 } 552 553 /* Source address validation */ 554 if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) && 555 (flags & IPV6_UNSPECSRC) == 0) { 556 error = EOPNOTSUPP; 557 ip6stat.ip6s_badscope++; 558 goto bad; 559 } 560 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) { 561 error = EOPNOTSUPP; 562 ip6stat.ip6s_badscope++; 563 goto bad; 564 } 565 566 ip6stat.ip6s_localout++; 567 568 /* 569 * Route packet. 570 */ 571 /* initialize cached route */ 572 if (ro == NULL) { 573 memset(&ip6route, 0, sizeof(ip6route)); 574 ro = &ip6route; 575 } 576 ro_pmtu = ro; 577 if (opt && opt->ip6po_rthdr) 578 ro = &opt->ip6po_route; 579 580 /* 581 * if specified, try to fill in the traffic class field. 582 * do not override if a non-zero value is already set. 583 * we check the diffserv field and the ecn field separately. 584 */ 585 if (opt && opt->ip6po_tclass >= 0) { 586 int mask = 0; 587 588 if ((ip6->ip6_flow & htonl(0xfc << 20)) == 0) 589 mask |= 0xfc; 590 if ((ip6->ip6_flow & htonl(0x03 << 20)) == 0) 591 mask |= 0x03; 592 if (mask != 0) 593 ip6->ip6_flow |= htonl((opt->ip6po_tclass & mask) << 20); 594 } 595 596 /* fill in or override the hop limit field, if necessary. */ 597 if (opt && opt->ip6po_hlim != -1) 598 ip6->ip6_hlim = opt->ip6po_hlim & 0xff; 599 else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) { 600 if (im6o != NULL) 601 ip6->ip6_hlim = im6o->im6o_multicast_hlim; 602 else 603 ip6->ip6_hlim = ip6_defmcasthlim; 604 } 605 606 #ifdef IPSEC 607 if (needipsec && needipsectun) { 608 struct ipsec_output_state state; 609 610 /* 611 * All the extension headers will become inaccessible 612 * (since they can be encrypted). 613 * Don't panic, we need no more updates to extension headers 614 * on inner IPv6 packet (since they are now encapsulated). 615 * 616 * IPv6 [ESP|AH] IPv6 [extension headers] payload 617 */ 618 bzero(&exthdrs, sizeof(exthdrs)); 619 exthdrs.ip6e_ip6 = m; 620 621 bzero(&state, sizeof(state)); 622 state.m = m; 623 state.ro = ro; 624 state.dst = rtcache_getdst(ro); 625 626 error = ipsec6_output_tunnel(&state, sp, flags); 627 628 m = state.m; 629 ro_pmtu = ro = state.ro; 630 dst = satocsin6(state.dst); 631 if (error) { 632 /* mbuf is already reclaimed in ipsec6_output_tunnel. */ 633 m0 = m = NULL; 634 m = NULL; 635 switch (error) { 636 case EHOSTUNREACH: 637 case ENETUNREACH: 638 case EMSGSIZE: 639 case ENOBUFS: 640 case ENOMEM: 641 break; 642 default: 643 printf("ip6_output (ipsec): error code %d\n", error); 644 /* FALLTHROUGH */ 645 case ENOENT: 646 /* don't show these error codes to the user */ 647 error = 0; 648 break; 649 } 650 goto bad; 651 } 652 653 exthdrs.ip6e_ip6 = m; 654 } 655 #endif /* IPSEC */ 656 #ifdef FAST_IPSEC 657 if (needipsec) { 658 s = splsoftnet(); 659 error = ipsec6_process_packet(m,sp->req); 660 661 /* 662 * Preserve KAME behaviour: ENOENT can be returned 663 * when an SA acquire is in progress. Don't propagate 664 * this to user-level; it confuses applications. 665 * XXX this will go away when the SADB is redone. 666 */ 667 if (error == ENOENT) 668 error = 0; 669 splx(s); 670 goto done; 671 } 672 #endif /* FAST_IPSEC */ 673 674 675 676 /* adjust pointer */ 677 ip6 = mtod(m, struct ip6_hdr *); 678 679 bzero(&dst_sa, sizeof(dst_sa)); 680 dst_sa.sin6_family = AF_INET6; 681 dst_sa.sin6_len = sizeof(dst_sa); 682 dst_sa.sin6_addr = ip6->ip6_dst; 683 if ((error = in6_selectroute(&dst_sa, opt, im6o, ro, 684 &ifp, &rt, 0)) != 0) { 685 switch (error) { 686 case EHOSTUNREACH: 687 ip6stat.ip6s_noroute++; 688 break; 689 case EADDRNOTAVAIL: 690 default: 691 break; /* XXX statistics? */ 692 } 693 if (ifp != NULL) 694 in6_ifstat_inc(ifp, ifs6_out_discard); 695 goto bad; 696 } 697 if (rt == NULL) { 698 /* 699 * If in6_selectroute() does not return a route entry, 700 * dst may not have been updated. 701 */ 702 rtcache_setdst(ro, sin6tosa(&dst_sa)); 703 } 704 705 /* 706 * then rt (for unicast) and ifp must be non-NULL valid values. 707 */ 708 if ((flags & IPV6_FORWARDING) == 0) { 709 /* XXX: the FORWARDING flag can be set for mrouting. */ 710 in6_ifstat_inc(ifp, ifs6_out_request); 711 } 712 if (rt != NULL) { 713 ia = (struct in6_ifaddr *)(rt->rt_ifa); 714 rt->rt_use++; 715 } 716 717 /* 718 * The outgoing interface must be in the zone of source and 719 * destination addresses. We should use ia_ifp to support the 720 * case of sending packets to an address of our own. 721 */ 722 if (ia != NULL && ia->ia_ifp) 723 origifp = ia->ia_ifp; 724 else 725 origifp = ifp; 726 727 src0 = ip6->ip6_src; 728 if (in6_setscope(&src0, origifp, &zone)) 729 goto badscope; 730 bzero(&src_sa, sizeof(src_sa)); 731 src_sa.sin6_family = AF_INET6; 732 src_sa.sin6_len = sizeof(src_sa); 733 src_sa.sin6_addr = ip6->ip6_src; 734 if (sa6_recoverscope(&src_sa) || zone != src_sa.sin6_scope_id) 735 goto badscope; 736 737 dst0 = ip6->ip6_dst; 738 if (in6_setscope(&dst0, origifp, &zone)) 739 goto badscope; 740 /* re-initialize to be sure */ 741 bzero(&dst_sa, sizeof(dst_sa)); 742 dst_sa.sin6_family = AF_INET6; 743 dst_sa.sin6_len = sizeof(dst_sa); 744 dst_sa.sin6_addr = ip6->ip6_dst; 745 if (sa6_recoverscope(&dst_sa) || zone != dst_sa.sin6_scope_id) 746 goto badscope; 747 748 /* scope check is done. */ 749 750 if (rt == NULL || IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) { 751 if (dst == NULL) 752 dst = satocsin6(rtcache_getdst(ro)); 753 KASSERT(dst != NULL); 754 } else if (opt && opt->ip6po_nextroute.ro_rt != NULL) { 755 /* 756 * The nexthop is explicitly specified by the 757 * application. We assume the next hop is an IPv6 758 * address. 759 */ 760 dst = (struct sockaddr_in6 *)opt->ip6po_nexthop; 761 } else if ((rt->rt_flags & RTF_GATEWAY)) 762 dst = (struct sockaddr_in6 *)rt->rt_gateway; 763 else if (dst == NULL) 764 dst = satocsin6(rtcache_getdst(ro)); 765 766 /* 767 * XXXXXX: original code follows: 768 */ 769 if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) 770 m->m_flags &= ~(M_BCAST | M_MCAST); /* just in case */ 771 else { 772 struct in6_multi *in6m; 773 774 m->m_flags = (m->m_flags & ~M_BCAST) | M_MCAST; 775 776 in6_ifstat_inc(ifp, ifs6_out_mcast); 777 778 /* 779 * Confirm that the outgoing interface supports multicast. 780 */ 781 if (!(ifp->if_flags & IFF_MULTICAST)) { 782 ip6stat.ip6s_noroute++; 783 in6_ifstat_inc(ifp, ifs6_out_discard); 784 error = ENETUNREACH; 785 goto bad; 786 } 787 788 IN6_LOOKUP_MULTI(ip6->ip6_dst, ifp, in6m); 789 if (in6m != NULL && 790 (im6o == NULL || im6o->im6o_multicast_loop)) { 791 /* 792 * If we belong to the destination multicast group 793 * on the outgoing interface, and the caller did not 794 * forbid loopback, loop back a copy. 795 */ 796 KASSERT(dst != NULL); 797 ip6_mloopback(ifp, m, dst); 798 } else { 799 /* 800 * If we are acting as a multicast router, perform 801 * multicast forwarding as if the packet had just 802 * arrived on the interface to which we are about 803 * to send. The multicast forwarding function 804 * recursively calls this function, using the 805 * IPV6_FORWARDING flag to prevent infinite recursion. 806 * 807 * Multicasts that are looped back by ip6_mloopback(), 808 * above, will be forwarded by the ip6_input() routine, 809 * if necessary. 810 */ 811 if (ip6_mrouter && (flags & IPV6_FORWARDING) == 0) { 812 if (ip6_mforward(ip6, ifp, m) != 0) { 813 m_freem(m); 814 goto done; 815 } 816 } 817 } 818 /* 819 * Multicasts with a hoplimit of zero may be looped back, 820 * above, but must not be transmitted on a network. 821 * Also, multicasts addressed to the loopback interface 822 * are not sent -- the above call to ip6_mloopback() will 823 * loop back a copy if this host actually belongs to the 824 * destination group on the loopback interface. 825 */ 826 if (ip6->ip6_hlim == 0 || (ifp->if_flags & IFF_LOOPBACK) || 827 IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst)) { 828 m_freem(m); 829 goto done; 830 } 831 } 832 833 /* 834 * Fill the outgoing inteface to tell the upper layer 835 * to increment per-interface statistics. 836 */ 837 if (ifpp) 838 *ifpp = ifp; 839 840 /* Determine path MTU. */ 841 if ((error = ip6_getpmtu(ro_pmtu, ro, ifp, &finaldst, &mtu, 842 &alwaysfrag)) != 0) 843 goto bad; 844 #ifdef IPSEC 845 if (needipsectun) 846 mtu = IPV6_MMTU; 847 #endif 848 849 /* 850 * The caller of this function may specify to use the minimum MTU 851 * in some cases. 852 * An advanced API option (IPV6_USE_MIN_MTU) can also override MTU 853 * setting. The logic is a bit complicated; by default, unicast 854 * packets will follow path MTU while multicast packets will be sent at 855 * the minimum MTU. If IP6PO_MINMTU_ALL is specified, all packets 856 * including unicast ones will be sent at the minimum MTU. Multicast 857 * packets will always be sent at the minimum MTU unless 858 * IP6PO_MINMTU_DISABLE is explicitly specified. 859 * See RFC 3542 for more details. 860 */ 861 if (mtu > IPV6_MMTU) { 862 if ((flags & IPV6_MINMTU)) 863 mtu = IPV6_MMTU; 864 else if (opt && opt->ip6po_minmtu == IP6PO_MINMTU_ALL) 865 mtu = IPV6_MMTU; 866 else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) && 867 (opt == NULL || 868 opt->ip6po_minmtu != IP6PO_MINMTU_DISABLE)) { 869 mtu = IPV6_MMTU; 870 } 871 } 872 873 /* 874 * clear embedded scope identifiers if necessary. 875 * in6_clearscope will touch the addresses only when necessary. 876 */ 877 in6_clearscope(&ip6->ip6_src); 878 in6_clearscope(&ip6->ip6_dst); 879 880 /* 881 * If the outgoing packet contains a hop-by-hop options header, 882 * it must be examined and processed even by the source node. 883 * (RFC 2460, section 4.) 884 */ 885 if (exthdrs.ip6e_hbh) { 886 struct ip6_hbh *hbh = mtod(exthdrs.ip6e_hbh, struct ip6_hbh *); 887 u_int32_t dummy1; /* XXX unused */ 888 u_int32_t dummy2; /* XXX unused */ 889 890 /* 891 * XXX: if we have to send an ICMPv6 error to the sender, 892 * we need the M_LOOP flag since icmp6_error() expects 893 * the IPv6 and the hop-by-hop options header are 894 * continuous unless the flag is set. 895 */ 896 m->m_flags |= M_LOOP; 897 m->m_pkthdr.rcvif = ifp; 898 if (ip6_process_hopopts(m, (u_int8_t *)(hbh + 1), 899 ((hbh->ip6h_len + 1) << 3) - sizeof(struct ip6_hbh), 900 &dummy1, &dummy2) < 0) { 901 /* m was already freed at this point */ 902 error = EINVAL;/* better error? */ 903 goto done; 904 } 905 m->m_flags &= ~M_LOOP; /* XXX */ 906 m->m_pkthdr.rcvif = NULL; 907 } 908 909 #ifdef PFIL_HOOKS 910 /* 911 * Run through list of hooks for output packets. 912 */ 913 if ((error = pfil_run_hooks(&inet6_pfil_hook, &m, ifp, PFIL_OUT)) != 0) 914 goto done; 915 if (m == NULL) 916 goto done; 917 ip6 = mtod(m, struct ip6_hdr *); 918 #endif /* PFIL_HOOKS */ 919 /* 920 * Send the packet to the outgoing interface. 921 * If necessary, do IPv6 fragmentation before sending. 922 * 923 * the logic here is rather complex: 924 * 1: normal case (dontfrag == 0, alwaysfrag == 0) 925 * 1-a: send as is if tlen <= path mtu 926 * 1-b: fragment if tlen > path mtu 927 * 928 * 2: if user asks us not to fragment (dontfrag == 1) 929 * 2-a: send as is if tlen <= interface mtu 930 * 2-b: error if tlen > interface mtu 931 * 932 * 3: if we always need to attach fragment header (alwaysfrag == 1) 933 * always fragment 934 * 935 * 4: if dontfrag == 1 && alwaysfrag == 1 936 * error, as we cannot handle this conflicting request 937 */ 938 tlen = m->m_pkthdr.len; 939 tso = (m->m_pkthdr.csum_flags & M_CSUM_TSOv6) != 0; 940 if (opt && (opt->ip6po_flags & IP6PO_DONTFRAG)) 941 dontfrag = 1; 942 else 943 dontfrag = 0; 944 945 if (dontfrag && alwaysfrag) { /* case 4 */ 946 /* conflicting request - can't transmit */ 947 error = EMSGSIZE; 948 goto bad; 949 } 950 if (dontfrag && (!tso && tlen > IN6_LINKMTU(ifp))) { /* case 2-b */ 951 /* 952 * Even if the DONTFRAG option is specified, we cannot send the 953 * packet when the data length is larger than the MTU of the 954 * outgoing interface. 955 * Notify the error by sending IPV6_PATHMTU ancillary data as 956 * well as returning an error code (the latter is not described 957 * in the API spec.) 958 */ 959 u_int32_t mtu32; 960 struct ip6ctlparam ip6cp; 961 962 mtu32 = (u_int32_t)mtu; 963 bzero(&ip6cp, sizeof(ip6cp)); 964 ip6cp.ip6c_cmdarg = (void *)&mtu32; 965 pfctlinput2(PRC_MSGSIZE, 966 rtcache_getdst(ro_pmtu), &ip6cp); 967 968 error = EMSGSIZE; 969 goto bad; 970 } 971 972 /* 973 * transmit packet without fragmentation 974 */ 975 if (dontfrag || (!alwaysfrag && (tlen <= mtu || tso))) { 976 /* case 1-a and 2-a */ 977 struct in6_ifaddr *ia6; 978 int sw_csum; 979 980 ip6 = mtod(m, struct ip6_hdr *); 981 ia6 = in6_ifawithifp(ifp, &ip6->ip6_src); 982 if (ia6) { 983 /* Record statistics for this interface address. */ 984 ia6->ia_ifa.ifa_data.ifad_outbytes += m->m_pkthdr.len; 985 } 986 #ifdef IPSEC 987 /* clean ipsec history once it goes out of the node */ 988 ipsec_delaux(m); 989 #endif 990 991 sw_csum = m->m_pkthdr.csum_flags & ~ifp->if_csum_flags_tx; 992 if ((sw_csum & (M_CSUM_UDPv6|M_CSUM_TCPv6)) != 0) { 993 if (IN6_NEED_CHECKSUM(ifp, 994 sw_csum & (M_CSUM_UDPv6|M_CSUM_TCPv6))) { 995 in6_delayed_cksum(m); 996 } 997 m->m_pkthdr.csum_flags &= ~(M_CSUM_UDPv6|M_CSUM_TCPv6); 998 } 999 1000 KASSERT(dst != NULL); 1001 if (__predict_true(!tso || 1002 (ifp->if_capenable & IFCAP_TSOv6) != 0)) { 1003 error = nd6_output(ifp, origifp, m, dst, rt); 1004 } else { 1005 error = ip6_tso_output(ifp, origifp, m, dst, rt); 1006 } 1007 goto done; 1008 } 1009 1010 if (tso) { 1011 error = EINVAL; /* XXX */ 1012 goto bad; 1013 } 1014 1015 /* 1016 * try to fragment the packet. case 1-b and 3 1017 */ 1018 if (mtu < IPV6_MMTU) { 1019 /* path MTU cannot be less than IPV6_MMTU */ 1020 error = EMSGSIZE; 1021 in6_ifstat_inc(ifp, ifs6_out_fragfail); 1022 goto bad; 1023 } else if (ip6->ip6_plen == 0) { 1024 /* jumbo payload cannot be fragmented */ 1025 error = EMSGSIZE; 1026 in6_ifstat_inc(ifp, ifs6_out_fragfail); 1027 goto bad; 1028 } else { 1029 struct mbuf **mnext, *m_frgpart; 1030 struct ip6_frag *ip6f; 1031 u_int32_t id = htonl(ip6_randomid()); 1032 u_char nextproto; 1033 #if 0 /* see below */ 1034 struct ip6ctlparam ip6cp; 1035 u_int32_t mtu32; 1036 #endif 1037 1038 /* 1039 * Too large for the destination or interface; 1040 * fragment if possible. 1041 * Must be able to put at least 8 bytes per fragment. 1042 */ 1043 hlen = unfragpartlen; 1044 if (mtu > IPV6_MAXPACKET) 1045 mtu = IPV6_MAXPACKET; 1046 1047 #if 0 1048 /* 1049 * It is believed this code is a leftover from the 1050 * development of the IPV6_RECVPATHMTU sockopt and 1051 * associated work to implement RFC3542. 1052 * It's not entirely clear what the intent of the API 1053 * is at this point, so disable this code for now. 1054 * The IPV6_RECVPATHMTU sockopt and/or IPV6_DONTFRAG 1055 * will send notifications if the application requests. 1056 */ 1057 1058 /* Notify a proper path MTU to applications. */ 1059 mtu32 = (u_int32_t)mtu; 1060 bzero(&ip6cp, sizeof(ip6cp)); 1061 ip6cp.ip6c_cmdarg = (void *)&mtu32; 1062 pfctlinput2(PRC_MSGSIZE, 1063 rtcache_getdst(ro_pmtu), &ip6cp); 1064 #endif 1065 1066 len = (mtu - hlen - sizeof(struct ip6_frag)) & ~7; 1067 if (len < 8) { 1068 error = EMSGSIZE; 1069 in6_ifstat_inc(ifp, ifs6_out_fragfail); 1070 goto bad; 1071 } 1072 1073 mnext = &m->m_nextpkt; 1074 1075 /* 1076 * Change the next header field of the last header in the 1077 * unfragmentable part. 1078 */ 1079 if (exthdrs.ip6e_rthdr) { 1080 nextproto = *mtod(exthdrs.ip6e_rthdr, u_char *); 1081 *mtod(exthdrs.ip6e_rthdr, u_char *) = IPPROTO_FRAGMENT; 1082 } else if (exthdrs.ip6e_dest1) { 1083 nextproto = *mtod(exthdrs.ip6e_dest1, u_char *); 1084 *mtod(exthdrs.ip6e_dest1, u_char *) = IPPROTO_FRAGMENT; 1085 } else if (exthdrs.ip6e_hbh) { 1086 nextproto = *mtod(exthdrs.ip6e_hbh, u_char *); 1087 *mtod(exthdrs.ip6e_hbh, u_char *) = IPPROTO_FRAGMENT; 1088 } else { 1089 nextproto = ip6->ip6_nxt; 1090 ip6->ip6_nxt = IPPROTO_FRAGMENT; 1091 } 1092 1093 if ((m->m_pkthdr.csum_flags & (M_CSUM_UDPv6|M_CSUM_TCPv6)) 1094 != 0) { 1095 if (IN6_NEED_CHECKSUM(ifp, 1096 m->m_pkthdr.csum_flags & 1097 (M_CSUM_UDPv6|M_CSUM_TCPv6))) { 1098 in6_delayed_cksum(m); 1099 } 1100 m->m_pkthdr.csum_flags &= ~(M_CSUM_UDPv6|M_CSUM_TCPv6); 1101 } 1102 1103 /* 1104 * Loop through length of segment after first fragment, 1105 * make new header and copy data of each part and link onto 1106 * chain. 1107 */ 1108 m0 = m; 1109 for (off = hlen; off < tlen; off += len) { 1110 struct mbuf *mlast; 1111 1112 MGETHDR(m, M_DONTWAIT, MT_HEADER); 1113 if (!m) { 1114 error = ENOBUFS; 1115 ip6stat.ip6s_odropped++; 1116 goto sendorfree; 1117 } 1118 m->m_pkthdr.rcvif = NULL; 1119 m->m_flags = m0->m_flags & M_COPYFLAGS; 1120 *mnext = m; 1121 mnext = &m->m_nextpkt; 1122 m->m_data += max_linkhdr; 1123 mhip6 = mtod(m, struct ip6_hdr *); 1124 *mhip6 = *ip6; 1125 m->m_len = sizeof(*mhip6); 1126 error = ip6_insertfraghdr(m0, m, hlen, &ip6f); 1127 if (error) { 1128 ip6stat.ip6s_odropped++; 1129 goto sendorfree; 1130 } 1131 ip6f->ip6f_offlg = htons((u_int16_t)((off - hlen) & ~7)); 1132 if (off + len >= tlen) 1133 len = tlen - off; 1134 else 1135 ip6f->ip6f_offlg |= IP6F_MORE_FRAG; 1136 mhip6->ip6_plen = htons((u_int16_t)(len + hlen + 1137 sizeof(*ip6f) - sizeof(struct ip6_hdr))); 1138 if ((m_frgpart = m_copy(m0, off, len)) == 0) { 1139 error = ENOBUFS; 1140 ip6stat.ip6s_odropped++; 1141 goto sendorfree; 1142 } 1143 for (mlast = m; mlast->m_next; mlast = mlast->m_next) 1144 ; 1145 mlast->m_next = m_frgpart; 1146 m->m_pkthdr.len = len + hlen + sizeof(*ip6f); 1147 m->m_pkthdr.rcvif = (struct ifnet *)0; 1148 ip6f->ip6f_reserved = 0; 1149 ip6f->ip6f_ident = id; 1150 ip6f->ip6f_nxt = nextproto; 1151 ip6stat.ip6s_ofragments++; 1152 in6_ifstat_inc(ifp, ifs6_out_fragcreat); 1153 } 1154 1155 in6_ifstat_inc(ifp, ifs6_out_fragok); 1156 } 1157 1158 /* 1159 * Remove leading garbages. 1160 */ 1161 sendorfree: 1162 m = m0->m_nextpkt; 1163 m0->m_nextpkt = 0; 1164 m_freem(m0); 1165 for (m0 = m; m; m = m0) { 1166 m0 = m->m_nextpkt; 1167 m->m_nextpkt = 0; 1168 if (error == 0) { 1169 struct in6_ifaddr *ia6; 1170 ip6 = mtod(m, struct ip6_hdr *); 1171 ia6 = in6_ifawithifp(ifp, &ip6->ip6_src); 1172 if (ia6) { 1173 /* 1174 * Record statistics for this interface 1175 * address. 1176 */ 1177 ia6->ia_ifa.ifa_data.ifad_outbytes += 1178 m->m_pkthdr.len; 1179 } 1180 #ifdef IPSEC 1181 /* clean ipsec history once it goes out of the node */ 1182 ipsec_delaux(m); 1183 #endif 1184 KASSERT(dst != NULL); 1185 error = nd6_output(ifp, origifp, m, dst, rt); 1186 } else 1187 m_freem(m); 1188 } 1189 1190 if (error == 0) 1191 ip6stat.ip6s_fragmented++; 1192 1193 done: 1194 /* XXX Second if is invariant? */ 1195 if (ro == &ip6route) 1196 rtcache_free(ro); 1197 else if (ro_pmtu == &ip6route) 1198 rtcache_free(ro_pmtu); 1199 1200 #ifdef IPSEC 1201 if (sp != NULL) 1202 key_freesp(sp); 1203 #endif /* IPSEC */ 1204 #ifdef FAST_IPSEC 1205 if (sp != NULL) 1206 KEY_FREESP(&sp); 1207 #endif /* FAST_IPSEC */ 1208 1209 1210 return (error); 1211 1212 freehdrs: 1213 m_freem(exthdrs.ip6e_hbh); /* m_freem will check if mbuf is 0 */ 1214 m_freem(exthdrs.ip6e_dest1); 1215 m_freem(exthdrs.ip6e_rthdr); 1216 m_freem(exthdrs.ip6e_dest2); 1217 /* FALLTHROUGH */ 1218 bad: 1219 m_freem(m); 1220 goto done; 1221 badscope: 1222 ip6stat.ip6s_badscope++; 1223 in6_ifstat_inc(origifp, ifs6_out_discard); 1224 if (error == 0) 1225 error = EHOSTUNREACH; /* XXX */ 1226 goto bad; 1227 } 1228 1229 static int 1230 ip6_copyexthdr(struct mbuf **mp, void *hdr, int hlen) 1231 { 1232 struct mbuf *m; 1233 1234 if (hlen > MCLBYTES) 1235 return (ENOBUFS); /* XXX */ 1236 1237 MGET(m, M_DONTWAIT, MT_DATA); 1238 if (!m) 1239 return (ENOBUFS); 1240 1241 if (hlen > MLEN) { 1242 MCLGET(m, M_DONTWAIT); 1243 if ((m->m_flags & M_EXT) == 0) { 1244 m_free(m); 1245 return (ENOBUFS); 1246 } 1247 } 1248 m->m_len = hlen; 1249 if (hdr) 1250 bcopy(hdr, mtod(m, void *), hlen); 1251 1252 *mp = m; 1253 return (0); 1254 } 1255 1256 /* 1257 * Process a delayed payload checksum calculation. 1258 */ 1259 void 1260 in6_delayed_cksum(struct mbuf *m) 1261 { 1262 uint16_t csum, offset; 1263 1264 KASSERT((m->m_pkthdr.csum_flags & (M_CSUM_UDPv6|M_CSUM_TCPv6)) != 0); 1265 KASSERT((~m->m_pkthdr.csum_flags & (M_CSUM_UDPv6|M_CSUM_TCPv6)) != 0); 1266 KASSERT((m->m_pkthdr.csum_flags 1267 & (M_CSUM_UDPv4|M_CSUM_TCPv4|M_CSUM_TSOv4)) == 0); 1268 1269 offset = M_CSUM_DATA_IPv6_HL(m->m_pkthdr.csum_data); 1270 csum = in6_cksum(m, 0, offset, m->m_pkthdr.len - offset); 1271 if (csum == 0 && (m->m_pkthdr.csum_flags & M_CSUM_UDPv6) != 0) { 1272 csum = 0xffff; 1273 } 1274 1275 offset += M_CSUM_DATA_IPv6_OFFSET(m->m_pkthdr.csum_data); 1276 if ((offset + sizeof(csum)) > m->m_len) { 1277 m_copyback(m, offset, sizeof(csum), &csum); 1278 } else { 1279 *(uint16_t *)(mtod(m, char *) + offset) = csum; 1280 } 1281 } 1282 1283 /* 1284 * Insert jumbo payload option. 1285 */ 1286 static int 1287 ip6_insert_jumboopt(struct ip6_exthdrs *exthdrs, u_int32_t plen) 1288 { 1289 struct mbuf *mopt; 1290 u_int8_t *optbuf; 1291 u_int32_t v; 1292 1293 #define JUMBOOPTLEN 8 /* length of jumbo payload option and padding */ 1294 1295 /* 1296 * If there is no hop-by-hop options header, allocate new one. 1297 * If there is one but it doesn't have enough space to store the 1298 * jumbo payload option, allocate a cluster to store the whole options. 1299 * Otherwise, use it to store the options. 1300 */ 1301 if (exthdrs->ip6e_hbh == 0) { 1302 MGET(mopt, M_DONTWAIT, MT_DATA); 1303 if (mopt == 0) 1304 return (ENOBUFS); 1305 mopt->m_len = JUMBOOPTLEN; 1306 optbuf = mtod(mopt, u_int8_t *); 1307 optbuf[1] = 0; /* = ((JUMBOOPTLEN) >> 3) - 1 */ 1308 exthdrs->ip6e_hbh = mopt; 1309 } else { 1310 struct ip6_hbh *hbh; 1311 1312 mopt = exthdrs->ip6e_hbh; 1313 if (M_TRAILINGSPACE(mopt) < JUMBOOPTLEN) { 1314 /* 1315 * XXX assumption: 1316 * - exthdrs->ip6e_hbh is not referenced from places 1317 * other than exthdrs. 1318 * - exthdrs->ip6e_hbh is not an mbuf chain. 1319 */ 1320 int oldoptlen = mopt->m_len; 1321 struct mbuf *n; 1322 1323 /* 1324 * XXX: give up if the whole (new) hbh header does 1325 * not fit even in an mbuf cluster. 1326 */ 1327 if (oldoptlen + JUMBOOPTLEN > MCLBYTES) 1328 return (ENOBUFS); 1329 1330 /* 1331 * As a consequence, we must always prepare a cluster 1332 * at this point. 1333 */ 1334 MGET(n, M_DONTWAIT, MT_DATA); 1335 if (n) { 1336 MCLGET(n, M_DONTWAIT); 1337 if ((n->m_flags & M_EXT) == 0) { 1338 m_freem(n); 1339 n = NULL; 1340 } 1341 } 1342 if (!n) 1343 return (ENOBUFS); 1344 n->m_len = oldoptlen + JUMBOOPTLEN; 1345 bcopy(mtod(mopt, void *), mtod(n, void *), 1346 oldoptlen); 1347 optbuf = mtod(n, u_int8_t *) + oldoptlen; 1348 m_freem(mopt); 1349 mopt = exthdrs->ip6e_hbh = n; 1350 } else { 1351 optbuf = mtod(mopt, u_int8_t *) + mopt->m_len; 1352 mopt->m_len += JUMBOOPTLEN; 1353 } 1354 optbuf[0] = IP6OPT_PADN; 1355 optbuf[1] = 0; 1356 1357 /* 1358 * Adjust the header length according to the pad and 1359 * the jumbo payload option. 1360 */ 1361 hbh = mtod(mopt, struct ip6_hbh *); 1362 hbh->ip6h_len += (JUMBOOPTLEN >> 3); 1363 } 1364 1365 /* fill in the option. */ 1366 optbuf[2] = IP6OPT_JUMBO; 1367 optbuf[3] = 4; 1368 v = (u_int32_t)htonl(plen + JUMBOOPTLEN); 1369 bcopy(&v, &optbuf[4], sizeof(u_int32_t)); 1370 1371 /* finally, adjust the packet header length */ 1372 exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN; 1373 1374 return (0); 1375 #undef JUMBOOPTLEN 1376 } 1377 1378 /* 1379 * Insert fragment header and copy unfragmentable header portions. 1380 */ 1381 static int 1382 ip6_insertfraghdr(struct mbuf *m0, struct mbuf *m, int hlen, 1383 struct ip6_frag **frghdrp) 1384 { 1385 struct mbuf *n, *mlast; 1386 1387 if (hlen > sizeof(struct ip6_hdr)) { 1388 n = m_copym(m0, sizeof(struct ip6_hdr), 1389 hlen - sizeof(struct ip6_hdr), M_DONTWAIT); 1390 if (n == 0) 1391 return (ENOBUFS); 1392 m->m_next = n; 1393 } else 1394 n = m; 1395 1396 /* Search for the last mbuf of unfragmentable part. */ 1397 for (mlast = n; mlast->m_next; mlast = mlast->m_next) 1398 ; 1399 1400 if ((mlast->m_flags & M_EXT) == 0 && 1401 M_TRAILINGSPACE(mlast) >= sizeof(struct ip6_frag)) { 1402 /* use the trailing space of the last mbuf for the fragment hdr */ 1403 *frghdrp = (struct ip6_frag *)(mtod(mlast, char *) + 1404 mlast->m_len); 1405 mlast->m_len += sizeof(struct ip6_frag); 1406 m->m_pkthdr.len += sizeof(struct ip6_frag); 1407 } else { 1408 /* allocate a new mbuf for the fragment header */ 1409 struct mbuf *mfrg; 1410 1411 MGET(mfrg, M_DONTWAIT, MT_DATA); 1412 if (mfrg == 0) 1413 return (ENOBUFS); 1414 mfrg->m_len = sizeof(struct ip6_frag); 1415 *frghdrp = mtod(mfrg, struct ip6_frag *); 1416 mlast->m_next = mfrg; 1417 } 1418 1419 return (0); 1420 } 1421 1422 static int 1423 ip6_getpmtu(struct route *ro_pmtu, struct route *ro, struct ifnet *ifp, 1424 const struct in6_addr *dst, u_long *mtup, int *alwaysfragp) 1425 { 1426 u_int32_t mtu = 0; 1427 int alwaysfrag = 0; 1428 int error = 0; 1429 1430 if (ro_pmtu != ro) { 1431 union { 1432 struct sockaddr dst; 1433 struct sockaddr_in6 dst6; 1434 } u; 1435 1436 /* The first hop and the final destination may differ. */ 1437 sockaddr_in6_init(&u.dst6, dst, 0, 0, 0); 1438 rtcache_lookup(ro_pmtu, &u.dst); 1439 } 1440 if (ro_pmtu->ro_rt != NULL) { 1441 u_int32_t ifmtu; 1442 1443 if (ifp == NULL) 1444 ifp = ro_pmtu->ro_rt->rt_ifp; 1445 ifmtu = IN6_LINKMTU(ifp); 1446 mtu = ro_pmtu->ro_rt->rt_rmx.rmx_mtu; 1447 if (mtu == 0) 1448 mtu = ifmtu; 1449 else if (mtu < IPV6_MMTU) { 1450 /* 1451 * RFC2460 section 5, last paragraph: 1452 * if we record ICMPv6 too big message with 1453 * mtu < IPV6_MMTU, transmit packets sized IPV6_MMTU 1454 * or smaller, with fragment header attached. 1455 * (fragment header is needed regardless from the 1456 * packet size, for translators to identify packets) 1457 */ 1458 alwaysfrag = 1; 1459 mtu = IPV6_MMTU; 1460 } else if (mtu > ifmtu) { 1461 /* 1462 * The MTU on the route is larger than the MTU on 1463 * the interface! This shouldn't happen, unless the 1464 * MTU of the interface has been changed after the 1465 * interface was brought up. Change the MTU in the 1466 * route to match the interface MTU (as long as the 1467 * field isn't locked). 1468 */ 1469 mtu = ifmtu; 1470 if (!(ro_pmtu->ro_rt->rt_rmx.rmx_locks & RTV_MTU)) 1471 ro_pmtu->ro_rt->rt_rmx.rmx_mtu = mtu; 1472 } 1473 } else if (ifp) { 1474 mtu = IN6_LINKMTU(ifp); 1475 } else 1476 error = EHOSTUNREACH; /* XXX */ 1477 1478 *mtup = mtu; 1479 if (alwaysfragp) 1480 *alwaysfragp = alwaysfrag; 1481 return (error); 1482 } 1483 1484 /* 1485 * IP6 socket option processing. 1486 */ 1487 int 1488 ip6_ctloutput(int op, struct socket *so, int level, int optname, 1489 struct mbuf **mp) 1490 { 1491 int privileged, optdatalen, uproto; 1492 void *optdata; 1493 struct in6pcb *in6p = sotoin6pcb(so); 1494 struct mbuf *m = *mp; 1495 int error, optval; 1496 int optlen; 1497 struct lwp *l = curlwp; /* XXX */ 1498 1499 optlen = m ? m->m_len : 0; 1500 error = optval = 0; 1501 privileged = (l == 0 || kauth_authorize_generic(l->l_cred, 1502 KAUTH_GENERIC_ISSUSER, NULL)) ? 0 : 1; 1503 uproto = (int)so->so_proto->pr_protocol; 1504 1505 if (level == IPPROTO_IPV6) { 1506 switch (op) { 1507 case PRCO_SETOPT: 1508 switch (optname) { 1509 #ifdef RFC2292 1510 case IPV6_2292PKTOPTIONS: 1511 /* m is freed in ip6_pcbopts */ 1512 error = ip6_pcbopts(&in6p->in6p_outputopts, 1513 m, so); 1514 break; 1515 #endif 1516 1517 /* 1518 * Use of some Hop-by-Hop options or some 1519 * Destination options, might require special 1520 * privilege. That is, normal applications 1521 * (without special privilege) might be forbidden 1522 * from setting certain options in outgoing packets, 1523 * and might never see certain options in received 1524 * packets. [RFC 2292 Section 6] 1525 * KAME specific note: 1526 * KAME prevents non-privileged users from sending or 1527 * receiving ANY hbh/dst options in order to avoid 1528 * overhead of parsing options in the kernel. 1529 */ 1530 case IPV6_RECVHOPOPTS: 1531 case IPV6_RECVDSTOPTS: 1532 case IPV6_RECVRTHDRDSTOPTS: 1533 if (!privileged) { 1534 error = EPERM; 1535 break; 1536 } 1537 /* FALLTHROUGH */ 1538 case IPV6_UNICAST_HOPS: 1539 case IPV6_HOPLIMIT: 1540 case IPV6_FAITH: 1541 1542 case IPV6_RECVPKTINFO: 1543 case IPV6_RECVHOPLIMIT: 1544 case IPV6_RECVRTHDR: 1545 case IPV6_RECVPATHMTU: 1546 case IPV6_RECVTCLASS: 1547 case IPV6_V6ONLY: 1548 if (optlen != sizeof(int)) { 1549 error = EINVAL; 1550 break; 1551 } 1552 optval = *mtod(m, int *); 1553 switch (optname) { 1554 1555 case IPV6_UNICAST_HOPS: 1556 if (optval < -1 || optval >= 256) 1557 error = EINVAL; 1558 else { 1559 /* -1 = kernel default */ 1560 in6p->in6p_hops = optval; 1561 } 1562 break; 1563 #define OPTSET(bit) \ 1564 do { \ 1565 if (optval) \ 1566 in6p->in6p_flags |= (bit); \ 1567 else \ 1568 in6p->in6p_flags &= ~(bit); \ 1569 } while (/*CONSTCOND*/ 0) 1570 1571 #ifdef RFC2292 1572 #define OPTSET2292(bit) \ 1573 do { \ 1574 in6p->in6p_flags |= IN6P_RFC2292; \ 1575 if (optval) \ 1576 in6p->in6p_flags |= (bit); \ 1577 else \ 1578 in6p->in6p_flags &= ~(bit); \ 1579 } while (/*CONSTCOND*/ 0) 1580 #endif 1581 1582 #define OPTBIT(bit) (in6p->in6p_flags & (bit) ? 1 : 0) 1583 1584 case IPV6_RECVPKTINFO: 1585 #ifdef RFC2292 1586 /* cannot mix with RFC2292 */ 1587 if (OPTBIT(IN6P_RFC2292)) { 1588 error = EINVAL; 1589 break; 1590 } 1591 #endif 1592 OPTSET(IN6P_PKTINFO); 1593 break; 1594 1595 case IPV6_HOPLIMIT: 1596 { 1597 struct ip6_pktopts **optp; 1598 1599 #ifdef RFC2292 1600 /* cannot mix with RFC2292 */ 1601 if (OPTBIT(IN6P_RFC2292)) { 1602 error = EINVAL; 1603 break; 1604 } 1605 #endif 1606 optp = &in6p->in6p_outputopts; 1607 error = ip6_pcbopt(IPV6_HOPLIMIT, 1608 (u_char *)&optval, 1609 sizeof(optval), 1610 optp, 1611 privileged, uproto); 1612 break; 1613 } 1614 1615 case IPV6_RECVHOPLIMIT: 1616 #ifdef RFC2292 1617 /* cannot mix with RFC2292 */ 1618 if (OPTBIT(IN6P_RFC2292)) { 1619 error = EINVAL; 1620 break; 1621 } 1622 #endif 1623 OPTSET(IN6P_HOPLIMIT); 1624 break; 1625 1626 case IPV6_RECVHOPOPTS: 1627 #ifdef RFC2292 1628 /* cannot mix with RFC2292 */ 1629 if (OPTBIT(IN6P_RFC2292)) { 1630 error = EINVAL; 1631 break; 1632 } 1633 #endif 1634 OPTSET(IN6P_HOPOPTS); 1635 break; 1636 1637 case IPV6_RECVDSTOPTS: 1638 #ifdef RFC2292 1639 /* cannot mix with RFC2292 */ 1640 if (OPTBIT(IN6P_RFC2292)) { 1641 error = EINVAL; 1642 break; 1643 } 1644 #endif 1645 OPTSET(IN6P_DSTOPTS); 1646 break; 1647 1648 case IPV6_RECVRTHDRDSTOPTS: 1649 #ifdef RFC2292 1650 /* cannot mix with RFC2292 */ 1651 if (OPTBIT(IN6P_RFC2292)) { 1652 error = EINVAL; 1653 break; 1654 } 1655 #endif 1656 OPTSET(IN6P_RTHDRDSTOPTS); 1657 break; 1658 1659 case IPV6_RECVRTHDR: 1660 #ifdef RFC2292 1661 /* cannot mix with RFC2292 */ 1662 if (OPTBIT(IN6P_RFC2292)) { 1663 error = EINVAL; 1664 break; 1665 } 1666 #endif 1667 OPTSET(IN6P_RTHDR); 1668 break; 1669 1670 case IPV6_FAITH: 1671 OPTSET(IN6P_FAITH); 1672 break; 1673 1674 case IPV6_RECVPATHMTU: 1675 /* 1676 * We ignore this option for TCP 1677 * sockets. 1678 * (RFC3542 leaves this case 1679 * unspecified.) 1680 */ 1681 if (uproto != IPPROTO_TCP) 1682 OPTSET(IN6P_MTU); 1683 break; 1684 1685 case IPV6_V6ONLY: 1686 /* 1687 * make setsockopt(IPV6_V6ONLY) 1688 * available only prior to bind(2). 1689 * see ipng mailing list, Jun 22 2001. 1690 */ 1691 if (in6p->in6p_lport || 1692 !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr)) { 1693 error = EINVAL; 1694 break; 1695 } 1696 #ifdef INET6_BINDV6ONLY 1697 if (!optval) 1698 error = EINVAL; 1699 #else 1700 OPTSET(IN6P_IPV6_V6ONLY); 1701 #endif 1702 break; 1703 case IPV6_RECVTCLASS: 1704 #ifdef RFC2292 1705 /* cannot mix with RFC2292 XXX */ 1706 if (OPTBIT(IN6P_RFC2292)) { 1707 error = EINVAL; 1708 break; 1709 } 1710 #endif 1711 OPTSET(IN6P_TCLASS); 1712 break; 1713 1714 } 1715 break; 1716 1717 case IPV6_OTCLASS: 1718 { 1719 struct ip6_pktopts **optp; 1720 u_int8_t tclass; 1721 1722 if (optlen != sizeof(tclass)) { 1723 error = EINVAL; 1724 break; 1725 } 1726 tclass = *mtod(m, u_int8_t *); 1727 optp = &in6p->in6p_outputopts; 1728 error = ip6_pcbopt(optname, 1729 (u_char *)&tclass, 1730 sizeof(tclass), 1731 optp, 1732 privileged, uproto); 1733 break; 1734 } 1735 1736 case IPV6_TCLASS: 1737 case IPV6_DONTFRAG: 1738 case IPV6_USE_MIN_MTU: 1739 if (optlen != sizeof(optval)) { 1740 error = EINVAL; 1741 break; 1742 } 1743 optval = *mtod(m, int *); 1744 { 1745 struct ip6_pktopts **optp; 1746 optp = &in6p->in6p_outputopts; 1747 error = ip6_pcbopt(optname, 1748 (u_char *)&optval, 1749 sizeof(optval), 1750 optp, 1751 privileged, uproto); 1752 break; 1753 } 1754 1755 #ifdef RFC2292 1756 case IPV6_2292PKTINFO: 1757 case IPV6_2292HOPLIMIT: 1758 case IPV6_2292HOPOPTS: 1759 case IPV6_2292DSTOPTS: 1760 case IPV6_2292RTHDR: 1761 /* RFC 2292 */ 1762 if (optlen != sizeof(int)) { 1763 error = EINVAL; 1764 break; 1765 } 1766 optval = *mtod(m, int *); 1767 switch (optname) { 1768 case IPV6_2292PKTINFO: 1769 OPTSET2292(IN6P_PKTINFO); 1770 break; 1771 case IPV6_2292HOPLIMIT: 1772 OPTSET2292(IN6P_HOPLIMIT); 1773 break; 1774 case IPV6_2292HOPOPTS: 1775 /* 1776 * Check super-user privilege. 1777 * See comments for IPV6_RECVHOPOPTS. 1778 */ 1779 if (!privileged) 1780 return (EPERM); 1781 OPTSET2292(IN6P_HOPOPTS); 1782 break; 1783 case IPV6_2292DSTOPTS: 1784 if (!privileged) 1785 return (EPERM); 1786 OPTSET2292(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); /* XXX */ 1787 break; 1788 case IPV6_2292RTHDR: 1789 OPTSET2292(IN6P_RTHDR); 1790 break; 1791 } 1792 break; 1793 #endif 1794 case IPV6_PKTINFO: 1795 case IPV6_HOPOPTS: 1796 case IPV6_RTHDR: 1797 case IPV6_DSTOPTS: 1798 case IPV6_RTHDRDSTOPTS: 1799 case IPV6_NEXTHOP: 1800 { 1801 /* new advanced API (RFC3542) */ 1802 u_char *optbuf; 1803 int optbuflen; 1804 struct ip6_pktopts **optp; 1805 if (!m) { 1806 error = EINVAL; 1807 break; 1808 } 1809 1810 #ifdef RFC2292 1811 /* cannot mix with RFC2292 */ 1812 if (OPTBIT(IN6P_RFC2292)) { 1813 error = EINVAL; 1814 break; 1815 } 1816 #endif 1817 1818 if (m && m->m_next) { 1819 error = EINVAL; /* XXX */ 1820 break; 1821 } 1822 1823 optbuf = mtod(m, u_char *); 1824 optbuflen = m->m_len; 1825 optp = &in6p->in6p_outputopts; 1826 error = ip6_pcbopt(optname, optbuf, optbuflen, 1827 optp, privileged, uproto); 1828 break; 1829 } 1830 #undef OPTSET 1831 1832 case IPV6_MULTICAST_IF: 1833 case IPV6_MULTICAST_HOPS: 1834 case IPV6_MULTICAST_LOOP: 1835 case IPV6_JOIN_GROUP: 1836 case IPV6_LEAVE_GROUP: 1837 error = ip6_setmoptions(optname, 1838 &in6p->in6p_moptions, m); 1839 break; 1840 1841 case IPV6_PORTRANGE: 1842 if (!m) { 1843 error = EINVAL; 1844 break; 1845 } 1846 optval = *mtod(m, int *); 1847 1848 switch (optval) { 1849 case IPV6_PORTRANGE_DEFAULT: 1850 in6p->in6p_flags &= ~(IN6P_LOWPORT); 1851 in6p->in6p_flags &= ~(IN6P_HIGHPORT); 1852 break; 1853 1854 case IPV6_PORTRANGE_HIGH: 1855 in6p->in6p_flags &= ~(IN6P_LOWPORT); 1856 in6p->in6p_flags |= IN6P_HIGHPORT; 1857 break; 1858 1859 case IPV6_PORTRANGE_LOW: 1860 in6p->in6p_flags &= ~(IN6P_HIGHPORT); 1861 in6p->in6p_flags |= IN6P_LOWPORT; 1862 break; 1863 1864 default: 1865 error = EINVAL; 1866 break; 1867 } 1868 break; 1869 1870 1871 #if defined(IPSEC) || defined(FAST_IPSEC) 1872 case IPV6_IPSEC_POLICY: 1873 { 1874 void *req = NULL; 1875 size_t len = 0; 1876 if (m) { 1877 req = mtod(m, void *); 1878 len = m->m_len; 1879 } 1880 error = ipsec6_set_policy(in6p, optname, req, 1881 len, privileged); 1882 } 1883 break; 1884 #endif /* IPSEC */ 1885 1886 default: 1887 error = ENOPROTOOPT; 1888 break; 1889 } 1890 if (m) 1891 (void)m_free(m); 1892 break; 1893 1894 case PRCO_GETOPT: 1895 switch (optname) { 1896 #ifdef RFC2292 1897 case IPV6_2292PKTOPTIONS: 1898 /* 1899 * RFC3542 (effectively) deprecated the 1900 * semantics of the 2292-style pktoptions. 1901 * Since it was not reliable in nature (i.e., 1902 * applications had to expect the lack of some 1903 * information after all), it would make sense 1904 * to simplify this part by always returning 1905 * empty data. 1906 */ 1907 *mp = m_get(M_WAIT, MT_SOOPTS); 1908 (*mp)->m_len = 0; 1909 break; 1910 #endif 1911 1912 case IPV6_RECVHOPOPTS: 1913 case IPV6_RECVDSTOPTS: 1914 case IPV6_RECVRTHDRDSTOPTS: 1915 case IPV6_UNICAST_HOPS: 1916 case IPV6_RECVPKTINFO: 1917 case IPV6_RECVHOPLIMIT: 1918 case IPV6_RECVRTHDR: 1919 case IPV6_RECVPATHMTU: 1920 1921 case IPV6_FAITH: 1922 case IPV6_V6ONLY: 1923 case IPV6_PORTRANGE: 1924 case IPV6_RECVTCLASS: 1925 switch (optname) { 1926 1927 case IPV6_RECVHOPOPTS: 1928 optval = OPTBIT(IN6P_HOPOPTS); 1929 break; 1930 1931 case IPV6_RECVDSTOPTS: 1932 optval = OPTBIT(IN6P_DSTOPTS); 1933 break; 1934 1935 case IPV6_RECVRTHDRDSTOPTS: 1936 optval = OPTBIT(IN6P_RTHDRDSTOPTS); 1937 break; 1938 1939 case IPV6_UNICAST_HOPS: 1940 optval = in6p->in6p_hops; 1941 break; 1942 1943 case IPV6_RECVPKTINFO: 1944 optval = OPTBIT(IN6P_PKTINFO); 1945 break; 1946 1947 case IPV6_RECVHOPLIMIT: 1948 optval = OPTBIT(IN6P_HOPLIMIT); 1949 break; 1950 1951 case IPV6_RECVRTHDR: 1952 optval = OPTBIT(IN6P_RTHDR); 1953 break; 1954 1955 case IPV6_RECVPATHMTU: 1956 optval = OPTBIT(IN6P_MTU); 1957 break; 1958 1959 case IPV6_FAITH: 1960 optval = OPTBIT(IN6P_FAITH); 1961 break; 1962 1963 case IPV6_V6ONLY: 1964 optval = OPTBIT(IN6P_IPV6_V6ONLY); 1965 break; 1966 1967 case IPV6_PORTRANGE: 1968 { 1969 int flags; 1970 flags = in6p->in6p_flags; 1971 if (flags & IN6P_HIGHPORT) 1972 optval = IPV6_PORTRANGE_HIGH; 1973 else if (flags & IN6P_LOWPORT) 1974 optval = IPV6_PORTRANGE_LOW; 1975 else 1976 optval = 0; 1977 break; 1978 } 1979 case IPV6_RECVTCLASS: 1980 optval = OPTBIT(IN6P_TCLASS); 1981 break; 1982 1983 } 1984 if (error) 1985 break; 1986 *mp = m = m_get(M_WAIT, MT_SOOPTS); 1987 m->m_len = sizeof(int); 1988 *mtod(m, int *) = optval; 1989 break; 1990 1991 case IPV6_PATHMTU: 1992 { 1993 u_long pmtu = 0; 1994 struct ip6_mtuinfo mtuinfo; 1995 struct route *ro = &in6p->in6p_route; 1996 1997 if (!(so->so_state & SS_ISCONNECTED)) 1998 return (ENOTCONN); 1999 /* 2000 * XXX: we dot not consider the case of source 2001 * routing, or optional information to specify 2002 * the outgoing interface. 2003 */ 2004 error = ip6_getpmtu(ro, NULL, NULL, 2005 &in6p->in6p_faddr, &pmtu, NULL); 2006 if (error) 2007 break; 2008 if (pmtu > IPV6_MAXPACKET) 2009 pmtu = IPV6_MAXPACKET; 2010 2011 memset(&mtuinfo, 0, sizeof(mtuinfo)); 2012 mtuinfo.ip6m_mtu = (u_int32_t)pmtu; 2013 optdata = (void *)&mtuinfo; 2014 optdatalen = sizeof(mtuinfo); 2015 if (optdatalen > MCLBYTES) 2016 return (EMSGSIZE); /* XXX */ 2017 *mp = m = m_get(M_WAIT, MT_SOOPTS); 2018 if (optdatalen > MLEN) 2019 MCLGET(m, M_WAIT); 2020 m->m_len = optdatalen; 2021 memcpy(mtod(m, void *), optdata, optdatalen); 2022 break; 2023 } 2024 2025 #ifdef RFC2292 2026 case IPV6_2292PKTINFO: 2027 case IPV6_2292HOPLIMIT: 2028 case IPV6_2292HOPOPTS: 2029 case IPV6_2292RTHDR: 2030 case IPV6_2292DSTOPTS: 2031 switch (optname) { 2032 case IPV6_2292PKTINFO: 2033 optval = OPTBIT(IN6P_PKTINFO); 2034 break; 2035 case IPV6_2292HOPLIMIT: 2036 optval = OPTBIT(IN6P_HOPLIMIT); 2037 break; 2038 case IPV6_2292HOPOPTS: 2039 optval = OPTBIT(IN6P_HOPOPTS); 2040 break; 2041 case IPV6_2292RTHDR: 2042 optval = OPTBIT(IN6P_RTHDR); 2043 break; 2044 case IPV6_2292DSTOPTS: 2045 optval = OPTBIT(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); 2046 break; 2047 } 2048 *mp = m = m_get(M_WAIT, MT_SOOPTS); 2049 m->m_len = sizeof(int); 2050 *mtod(m, int *) = optval; 2051 break; 2052 #endif 2053 case IPV6_PKTINFO: 2054 case IPV6_HOPOPTS: 2055 case IPV6_RTHDR: 2056 case IPV6_DSTOPTS: 2057 case IPV6_RTHDRDSTOPTS: 2058 case IPV6_NEXTHOP: 2059 case IPV6_OTCLASS: 2060 case IPV6_TCLASS: 2061 case IPV6_DONTFRAG: 2062 case IPV6_USE_MIN_MTU: 2063 error = ip6_getpcbopt(in6p->in6p_outputopts, 2064 optname, mp); 2065 break; 2066 2067 case IPV6_MULTICAST_IF: 2068 case IPV6_MULTICAST_HOPS: 2069 case IPV6_MULTICAST_LOOP: 2070 case IPV6_JOIN_GROUP: 2071 case IPV6_LEAVE_GROUP: 2072 error = ip6_getmoptions(optname, 2073 in6p->in6p_moptions, mp); 2074 break; 2075 2076 #if defined(IPSEC) || defined(FAST_IPSEC) 2077 case IPV6_IPSEC_POLICY: 2078 { 2079 void *req = NULL; 2080 size_t len = 0; 2081 if (m) { 2082 req = mtod(m, void *); 2083 len = m->m_len; 2084 } 2085 error = ipsec6_get_policy(in6p, req, len, mp); 2086 break; 2087 } 2088 #endif /* IPSEC */ 2089 2090 2091 2092 2093 default: 2094 error = ENOPROTOOPT; 2095 break; 2096 } 2097 break; 2098 } 2099 } else { 2100 error = EINVAL; 2101 if (op == PRCO_SETOPT && *mp) 2102 (void)m_free(*mp); 2103 } 2104 return (error); 2105 } 2106 2107 int 2108 ip6_raw_ctloutput(int op, struct socket *so, int level, int optname, 2109 struct mbuf **mp) 2110 { 2111 int error = 0, optval, optlen; 2112 const int icmp6off = offsetof(struct icmp6_hdr, icmp6_cksum); 2113 struct in6pcb *in6p = sotoin6pcb(so); 2114 struct mbuf *m = *mp; 2115 2116 optlen = m ? m->m_len : 0; 2117 2118 if (level != IPPROTO_IPV6) { 2119 if (op == PRCO_SETOPT && *mp) 2120 (void)m_free(*mp); 2121 return (EINVAL); 2122 } 2123 2124 switch (optname) { 2125 case IPV6_CHECKSUM: 2126 /* 2127 * For ICMPv6 sockets, no modification allowed for checksum 2128 * offset, permit "no change" values to help existing apps. 2129 * 2130 * XXX RFC3542 says: "An attempt to set IPV6_CHECKSUM 2131 * for an ICMPv6 socket will fail." The current 2132 * behavior does not meet RFC3542. 2133 */ 2134 switch (op) { 2135 case PRCO_SETOPT: 2136 if (optlen != sizeof(int)) { 2137 error = EINVAL; 2138 break; 2139 } 2140 optval = *mtod(m, int *); 2141 if ((optval % 2) != 0) { 2142 /* the API assumes even offset values */ 2143 error = EINVAL; 2144 } else if (so->so_proto->pr_protocol == 2145 IPPROTO_ICMPV6) { 2146 if (optval != icmp6off) 2147 error = EINVAL; 2148 } else 2149 in6p->in6p_cksum = optval; 2150 break; 2151 2152 case PRCO_GETOPT: 2153 if (so->so_proto->pr_protocol == IPPROTO_ICMPV6) 2154 optval = icmp6off; 2155 else 2156 optval = in6p->in6p_cksum; 2157 2158 *mp = m = m_get(M_WAIT, MT_SOOPTS); 2159 m->m_len = sizeof(int); 2160 *mtod(m, int *) = optval; 2161 break; 2162 2163 default: 2164 error = EINVAL; 2165 break; 2166 } 2167 break; 2168 2169 default: 2170 error = ENOPROTOOPT; 2171 break; 2172 } 2173 2174 if (op == PRCO_SETOPT && m) 2175 (void)m_free(m); 2176 2177 return (error); 2178 } 2179 2180 #ifdef RFC2292 2181 /* 2182 * Set up IP6 options in pcb for insertion in output packets or 2183 * specifying behavior of outgoing packets. 2184 */ 2185 static int 2186 ip6_pcbopts(struct ip6_pktopts **pktopt, struct mbuf *m, struct socket *so) 2187 { 2188 struct ip6_pktopts *opt = *pktopt; 2189 int error = 0; 2190 struct lwp *l = curlwp; /* XXX */ 2191 int priv = 0; 2192 2193 /* turn off any old options. */ 2194 if (opt) { 2195 #ifdef DIAGNOSTIC 2196 if (opt->ip6po_pktinfo || opt->ip6po_nexthop || 2197 opt->ip6po_hbh || opt->ip6po_dest1 || opt->ip6po_dest2 || 2198 opt->ip6po_rhinfo.ip6po_rhi_rthdr) 2199 printf("ip6_pcbopts: all specified options are cleared.\n"); 2200 #endif 2201 ip6_clearpktopts(opt, -1); 2202 } else 2203 opt = malloc(sizeof(*opt), M_IP6OPT, M_WAITOK); 2204 *pktopt = NULL; 2205 2206 if (!m || m->m_len == 0) { 2207 /* 2208 * Only turning off any previous options, regardless of 2209 * whether the opt is just created or given. 2210 */ 2211 free(opt, M_IP6OPT); 2212 return (0); 2213 } 2214 2215 /* set options specified by user. */ 2216 if (l && !kauth_authorize_generic(l->l_cred, KAUTH_GENERIC_ISSUSER, 2217 NULL)) 2218 priv = 1; 2219 if ((error = ip6_setpktopts(m, opt, NULL, priv, 2220 so->so_proto->pr_protocol)) != 0) { 2221 ip6_clearpktopts(opt, -1); /* XXX: discard all options */ 2222 free(opt, M_IP6OPT); 2223 return (error); 2224 } 2225 *pktopt = opt; 2226 return (0); 2227 } 2228 #endif 2229 2230 /* 2231 * initialize ip6_pktopts. beware that there are non-zero default values in 2232 * the struct. 2233 */ 2234 void 2235 ip6_initpktopts(struct ip6_pktopts *opt) 2236 { 2237 2238 memset(opt, 0, sizeof(*opt)); 2239 opt->ip6po_hlim = -1; /* -1 means default hop limit */ 2240 opt->ip6po_tclass = -1; /* -1 means default traffic class */ 2241 opt->ip6po_minmtu = IP6PO_MINMTU_MCASTONLY; 2242 } 2243 2244 #define sin6tosa(sin6) ((struct sockaddr *)(sin6)) /* XXX */ 2245 static int 2246 ip6_pcbopt(int optname, u_char *buf, int len, struct ip6_pktopts **pktopt, 2247 int priv, int uproto) 2248 { 2249 struct ip6_pktopts *opt; 2250 2251 if (*pktopt == NULL) { 2252 *pktopt = malloc(sizeof(struct ip6_pktopts), M_IP6OPT, 2253 M_WAITOK); 2254 ip6_initpktopts(*pktopt); 2255 } 2256 opt = *pktopt; 2257 2258 return (ip6_setpktopt(optname, buf, len, opt, priv, 1, 0, uproto)); 2259 } 2260 2261 static int 2262 ip6_getpcbopt(struct ip6_pktopts *pktopt, int optname, struct mbuf **mp) 2263 { 2264 void *optdata = NULL; 2265 int optdatalen = 0; 2266 struct ip6_ext *ip6e; 2267 int error = 0; 2268 struct in6_pktinfo null_pktinfo; 2269 int deftclass = 0, on; 2270 int defminmtu = IP6PO_MINMTU_MCASTONLY; 2271 struct mbuf *m; 2272 2273 switch (optname) { 2274 case IPV6_PKTINFO: 2275 if (pktopt && pktopt->ip6po_pktinfo) 2276 optdata = (void *)pktopt->ip6po_pktinfo; 2277 else { 2278 /* XXX: we don't have to do this every time... */ 2279 memset(&null_pktinfo, 0, sizeof(null_pktinfo)); 2280 optdata = (void *)&null_pktinfo; 2281 } 2282 optdatalen = sizeof(struct in6_pktinfo); 2283 break; 2284 case IPV6_OTCLASS: 2285 /* XXX */ 2286 return (EINVAL); 2287 case IPV6_TCLASS: 2288 if (pktopt && pktopt->ip6po_tclass >= 0) 2289 optdata = (void *)&pktopt->ip6po_tclass; 2290 else 2291 optdata = (void *)&deftclass; 2292 optdatalen = sizeof(int); 2293 break; 2294 case IPV6_HOPOPTS: 2295 if (pktopt && pktopt->ip6po_hbh) { 2296 optdata = (void *)pktopt->ip6po_hbh; 2297 ip6e = (struct ip6_ext *)pktopt->ip6po_hbh; 2298 optdatalen = (ip6e->ip6e_len + 1) << 3; 2299 } 2300 break; 2301 case IPV6_RTHDR: 2302 if (pktopt && pktopt->ip6po_rthdr) { 2303 optdata = (void *)pktopt->ip6po_rthdr; 2304 ip6e = (struct ip6_ext *)pktopt->ip6po_rthdr; 2305 optdatalen = (ip6e->ip6e_len + 1) << 3; 2306 } 2307 break; 2308 case IPV6_RTHDRDSTOPTS: 2309 if (pktopt && pktopt->ip6po_dest1) { 2310 optdata = (void *)pktopt->ip6po_dest1; 2311 ip6e = (struct ip6_ext *)pktopt->ip6po_dest1; 2312 optdatalen = (ip6e->ip6e_len + 1) << 3; 2313 } 2314 break; 2315 case IPV6_DSTOPTS: 2316 if (pktopt && pktopt->ip6po_dest2) { 2317 optdata = (void *)pktopt->ip6po_dest2; 2318 ip6e = (struct ip6_ext *)pktopt->ip6po_dest2; 2319 optdatalen = (ip6e->ip6e_len + 1) << 3; 2320 } 2321 break; 2322 case IPV6_NEXTHOP: 2323 if (pktopt && pktopt->ip6po_nexthop) { 2324 optdata = (void *)pktopt->ip6po_nexthop; 2325 optdatalen = pktopt->ip6po_nexthop->sa_len; 2326 } 2327 break; 2328 case IPV6_USE_MIN_MTU: 2329 if (pktopt) 2330 optdata = (void *)&pktopt->ip6po_minmtu; 2331 else 2332 optdata = (void *)&defminmtu; 2333 optdatalen = sizeof(int); 2334 break; 2335 case IPV6_DONTFRAG: 2336 if (pktopt && ((pktopt->ip6po_flags) & IP6PO_DONTFRAG)) 2337 on = 1; 2338 else 2339 on = 0; 2340 optdata = (void *)&on; 2341 optdatalen = sizeof(on); 2342 break; 2343 default: /* should not happen */ 2344 #ifdef DIAGNOSTIC 2345 panic("ip6_getpcbopt: unexpected option\n"); 2346 #endif 2347 return (ENOPROTOOPT); 2348 } 2349 2350 if (optdatalen > MCLBYTES) 2351 return (EMSGSIZE); /* XXX */ 2352 *mp = m = m_get(M_WAIT, MT_SOOPTS); 2353 if (optdatalen > MLEN) 2354 MCLGET(m, M_WAIT); 2355 m->m_len = optdatalen; 2356 if (optdatalen) 2357 memcpy(mtod(m, void *), optdata, optdatalen); 2358 2359 return (error); 2360 } 2361 2362 void 2363 ip6_clearpktopts(struct ip6_pktopts *pktopt, int optname) 2364 { 2365 if (optname == -1 || optname == IPV6_PKTINFO) { 2366 if (pktopt->ip6po_pktinfo) 2367 free(pktopt->ip6po_pktinfo, M_IP6OPT); 2368 pktopt->ip6po_pktinfo = NULL; 2369 } 2370 if (optname == -1 || optname == IPV6_HOPLIMIT) 2371 pktopt->ip6po_hlim = -1; 2372 if (optname == -1 || optname == IPV6_TCLASS) 2373 pktopt->ip6po_tclass = -1; 2374 if (optname == -1 || optname == IPV6_NEXTHOP) { 2375 rtcache_free(&pktopt->ip6po_nextroute); 2376 if (pktopt->ip6po_nexthop) 2377 free(pktopt->ip6po_nexthop, M_IP6OPT); 2378 pktopt->ip6po_nexthop = NULL; 2379 } 2380 if (optname == -1 || optname == IPV6_HOPOPTS) { 2381 if (pktopt->ip6po_hbh) 2382 free(pktopt->ip6po_hbh, M_IP6OPT); 2383 pktopt->ip6po_hbh = NULL; 2384 } 2385 if (optname == -1 || optname == IPV6_RTHDRDSTOPTS) { 2386 if (pktopt->ip6po_dest1) 2387 free(pktopt->ip6po_dest1, M_IP6OPT); 2388 pktopt->ip6po_dest1 = NULL; 2389 } 2390 if (optname == -1 || optname == IPV6_RTHDR) { 2391 if (pktopt->ip6po_rhinfo.ip6po_rhi_rthdr) 2392 free(pktopt->ip6po_rhinfo.ip6po_rhi_rthdr, M_IP6OPT); 2393 pktopt->ip6po_rhinfo.ip6po_rhi_rthdr = NULL; 2394 rtcache_free(&pktopt->ip6po_route); 2395 } 2396 if (optname == -1 || optname == IPV6_DSTOPTS) { 2397 if (pktopt->ip6po_dest2) 2398 free(pktopt->ip6po_dest2, M_IP6OPT); 2399 pktopt->ip6po_dest2 = NULL; 2400 } 2401 } 2402 2403 #define PKTOPT_EXTHDRCPY(type) \ 2404 do { \ 2405 if (src->type) { \ 2406 int hlen = (((struct ip6_ext *)src->type)->ip6e_len + 1) << 3;\ 2407 dst->type = malloc(hlen, M_IP6OPT, canwait); \ 2408 if (dst->type == NULL && canwait == M_NOWAIT) \ 2409 goto bad; \ 2410 memcpy(dst->type, src->type, hlen); \ 2411 } \ 2412 } while (/*CONSTCOND*/ 0) 2413 2414 static int 2415 copypktopts(struct ip6_pktopts *dst, struct ip6_pktopts *src, int canwait) 2416 { 2417 dst->ip6po_hlim = src->ip6po_hlim; 2418 dst->ip6po_tclass = src->ip6po_tclass; 2419 dst->ip6po_flags = src->ip6po_flags; 2420 if (src->ip6po_pktinfo) { 2421 dst->ip6po_pktinfo = malloc(sizeof(*dst->ip6po_pktinfo), 2422 M_IP6OPT, canwait); 2423 if (dst->ip6po_pktinfo == NULL && canwait == M_NOWAIT) 2424 goto bad; 2425 *dst->ip6po_pktinfo = *src->ip6po_pktinfo; 2426 } 2427 if (src->ip6po_nexthop) { 2428 dst->ip6po_nexthop = malloc(src->ip6po_nexthop->sa_len, 2429 M_IP6OPT, canwait); 2430 if (dst->ip6po_nexthop == NULL && canwait == M_NOWAIT) 2431 goto bad; 2432 memcpy(dst->ip6po_nexthop, src->ip6po_nexthop, 2433 src->ip6po_nexthop->sa_len); 2434 } 2435 PKTOPT_EXTHDRCPY(ip6po_hbh); 2436 PKTOPT_EXTHDRCPY(ip6po_dest1); 2437 PKTOPT_EXTHDRCPY(ip6po_dest2); 2438 PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */ 2439 return (0); 2440 2441 bad: 2442 if (dst->ip6po_pktinfo) free(dst->ip6po_pktinfo, M_IP6OPT); 2443 if (dst->ip6po_nexthop) free(dst->ip6po_nexthop, M_IP6OPT); 2444 if (dst->ip6po_hbh) free(dst->ip6po_hbh, M_IP6OPT); 2445 if (dst->ip6po_dest1) free(dst->ip6po_dest1, M_IP6OPT); 2446 if (dst->ip6po_dest2) free(dst->ip6po_dest2, M_IP6OPT); 2447 if (dst->ip6po_rthdr) free(dst->ip6po_rthdr, M_IP6OPT); 2448 2449 return (ENOBUFS); 2450 } 2451 #undef PKTOPT_EXTHDRCPY 2452 2453 struct ip6_pktopts * 2454 ip6_copypktopts(struct ip6_pktopts *src, int canwait) 2455 { 2456 int error; 2457 struct ip6_pktopts *dst; 2458 2459 dst = malloc(sizeof(*dst), M_IP6OPT, canwait); 2460 if (dst == NULL && canwait == M_NOWAIT) 2461 return (NULL); 2462 ip6_initpktopts(dst); 2463 2464 if ((error = copypktopts(dst, src, canwait)) != 0) { 2465 free(dst, M_IP6OPT); 2466 return (NULL); 2467 } 2468 2469 return (dst); 2470 } 2471 2472 void 2473 ip6_freepcbopts(struct ip6_pktopts *pktopt) 2474 { 2475 if (pktopt == NULL) 2476 return; 2477 2478 ip6_clearpktopts(pktopt, -1); 2479 2480 free(pktopt, M_IP6OPT); 2481 } 2482 2483 /* 2484 * Set the IP6 multicast options in response to user setsockopt(). 2485 */ 2486 static int 2487 ip6_setmoptions(int optname, struct ip6_moptions **im6op, struct mbuf *m) 2488 { 2489 int error = 0; 2490 u_int loop, ifindex; 2491 struct ipv6_mreq *mreq; 2492 struct ifnet *ifp; 2493 struct ip6_moptions *im6o = *im6op; 2494 struct route ro; 2495 struct in6_multi_mship *imm; 2496 struct lwp *l = curlwp; /* XXX */ 2497 2498 if (im6o == NULL) { 2499 /* 2500 * No multicast option buffer attached to the pcb; 2501 * allocate one and initialize to default values. 2502 */ 2503 im6o = (struct ip6_moptions *) 2504 malloc(sizeof(*im6o), M_IPMOPTS, M_WAITOK); 2505 2506 if (im6o == NULL) 2507 return (ENOBUFS); 2508 *im6op = im6o; 2509 im6o->im6o_multicast_ifp = NULL; 2510 im6o->im6o_multicast_hlim = ip6_defmcasthlim; 2511 im6o->im6o_multicast_loop = IPV6_DEFAULT_MULTICAST_LOOP; 2512 LIST_INIT(&im6o->im6o_memberships); 2513 } 2514 2515 switch (optname) { 2516 2517 case IPV6_MULTICAST_IF: 2518 /* 2519 * Select the interface for outgoing multicast packets. 2520 */ 2521 if (m == NULL || m->m_len != sizeof(u_int)) { 2522 error = EINVAL; 2523 break; 2524 } 2525 bcopy(mtod(m, u_int *), &ifindex, sizeof(ifindex)); 2526 if (ifindex != 0) { 2527 if (if_indexlim <= ifindex || !ifindex2ifnet[ifindex]) { 2528 error = ENXIO; /* XXX EINVAL? */ 2529 break; 2530 } 2531 ifp = ifindex2ifnet[ifindex]; 2532 if ((ifp->if_flags & IFF_MULTICAST) == 0) { 2533 error = EADDRNOTAVAIL; 2534 break; 2535 } 2536 } else 2537 ifp = NULL; 2538 im6o->im6o_multicast_ifp = ifp; 2539 break; 2540 2541 case IPV6_MULTICAST_HOPS: 2542 { 2543 /* 2544 * Set the IP6 hoplimit for outgoing multicast packets. 2545 */ 2546 int optval; 2547 if (m == NULL || m->m_len != sizeof(int)) { 2548 error = EINVAL; 2549 break; 2550 } 2551 bcopy(mtod(m, u_int *), &optval, sizeof(optval)); 2552 if (optval < -1 || optval >= 256) 2553 error = EINVAL; 2554 else if (optval == -1) 2555 im6o->im6o_multicast_hlim = ip6_defmcasthlim; 2556 else 2557 im6o->im6o_multicast_hlim = optval; 2558 break; 2559 } 2560 2561 case IPV6_MULTICAST_LOOP: 2562 /* 2563 * Set the loopback flag for outgoing multicast packets. 2564 * Must be zero or one. 2565 */ 2566 if (m == NULL || m->m_len != sizeof(u_int)) { 2567 error = EINVAL; 2568 break; 2569 } 2570 bcopy(mtod(m, u_int *), &loop, sizeof(loop)); 2571 if (loop > 1) { 2572 error = EINVAL; 2573 break; 2574 } 2575 im6o->im6o_multicast_loop = loop; 2576 break; 2577 2578 case IPV6_JOIN_GROUP: 2579 /* 2580 * Add a multicast group membership. 2581 * Group must be a valid IP6 multicast address. 2582 */ 2583 if (m == NULL || m->m_len != sizeof(struct ipv6_mreq)) { 2584 error = EINVAL; 2585 break; 2586 } 2587 mreq = mtod(m, struct ipv6_mreq *); 2588 if (IN6_IS_ADDR_UNSPECIFIED(&mreq->ipv6mr_multiaddr)) { 2589 /* 2590 * We use the unspecified address to specify to accept 2591 * all multicast addresses. Only super user is allowed 2592 * to do this. 2593 */ 2594 if (kauth_authorize_generic(l->l_cred, 2595 KAUTH_GENERIC_ISSUSER, NULL)) 2596 { 2597 error = EACCES; 2598 break; 2599 } 2600 } else if (!IN6_IS_ADDR_MULTICAST(&mreq->ipv6mr_multiaddr)) { 2601 error = EINVAL; 2602 break; 2603 } 2604 2605 /* 2606 * If no interface was explicitly specified, choose an 2607 * appropriate one according to the given multicast address. 2608 */ 2609 if (mreq->ipv6mr_interface == 0) { 2610 union { 2611 struct sockaddr dst; 2612 struct sockaddr_in6 dst6; 2613 } u; 2614 2615 /* 2616 * Look up the routing table for the 2617 * address, and choose the outgoing interface. 2618 * XXX: is it a good approach? 2619 */ 2620 memset(&ro, 0, sizeof(ro)); 2621 sockaddr_in6_init(&u.dst6, &mreq->ipv6mr_multiaddr, 0, 2622 0, 0); 2623 rtcache_setdst(&ro, &u.dst); 2624 rtcache_init(&ro); 2625 ifp = (ro.ro_rt != NULL) ? ro.ro_rt->rt_ifp : NULL; 2626 rtcache_free(&ro); 2627 } else { 2628 /* 2629 * If the interface is specified, validate it. 2630 */ 2631 if (if_indexlim <= mreq->ipv6mr_interface || 2632 !ifindex2ifnet[mreq->ipv6mr_interface]) { 2633 error = ENXIO; /* XXX EINVAL? */ 2634 break; 2635 } 2636 ifp = ifindex2ifnet[mreq->ipv6mr_interface]; 2637 } 2638 2639 /* 2640 * See if we found an interface, and confirm that it 2641 * supports multicast 2642 */ 2643 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) { 2644 error = EADDRNOTAVAIL; 2645 break; 2646 } 2647 2648 if (in6_setscope(&mreq->ipv6mr_multiaddr, ifp, NULL)) { 2649 error = EADDRNOTAVAIL; /* XXX: should not happen */ 2650 break; 2651 } 2652 2653 /* 2654 * See if the membership already exists. 2655 */ 2656 for (imm = im6o->im6o_memberships.lh_first; 2657 imm != NULL; imm = imm->i6mm_chain.le_next) 2658 if (imm->i6mm_maddr->in6m_ifp == ifp && 2659 IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr, 2660 &mreq->ipv6mr_multiaddr)) 2661 break; 2662 if (imm != NULL) { 2663 error = EADDRINUSE; 2664 break; 2665 } 2666 /* 2667 * Everything looks good; add a new record to the multicast 2668 * address list for the given interface. 2669 */ 2670 imm = in6_joingroup(ifp, &mreq->ipv6mr_multiaddr, &error, 0); 2671 if (imm == NULL) 2672 break; 2673 LIST_INSERT_HEAD(&im6o->im6o_memberships, imm, i6mm_chain); 2674 break; 2675 2676 case IPV6_LEAVE_GROUP: 2677 /* 2678 * Drop a multicast group membership. 2679 * Group must be a valid IP6 multicast address. 2680 */ 2681 if (m == NULL || m->m_len != sizeof(struct ipv6_mreq)) { 2682 error = EINVAL; 2683 break; 2684 } 2685 mreq = mtod(m, struct ipv6_mreq *); 2686 2687 /* 2688 * If an interface address was specified, get a pointer 2689 * to its ifnet structure. 2690 */ 2691 if (mreq->ipv6mr_interface != 0) { 2692 if (if_indexlim <= mreq->ipv6mr_interface || 2693 !ifindex2ifnet[mreq->ipv6mr_interface]) { 2694 error = ENXIO; /* XXX EINVAL? */ 2695 break; 2696 } 2697 ifp = ifindex2ifnet[mreq->ipv6mr_interface]; 2698 } else 2699 ifp = NULL; 2700 2701 /* Fill in the scope zone ID */ 2702 if (ifp) { 2703 if (in6_setscope(&mreq->ipv6mr_multiaddr, ifp, NULL)) { 2704 /* XXX: should not happen */ 2705 error = EADDRNOTAVAIL; 2706 break; 2707 } 2708 } else if (mreq->ipv6mr_interface != 0) { 2709 /* 2710 * XXX: This case would happens when the (positive) 2711 * index is in the valid range, but the corresponding 2712 * interface has been detached dynamically. The above 2713 * check probably avoids such case to happen here, but 2714 * we check it explicitly for safety. 2715 */ 2716 error = EADDRNOTAVAIL; 2717 break; 2718 } else { /* ipv6mr_interface == 0 */ 2719 struct sockaddr_in6 sa6_mc; 2720 2721 /* 2722 * The API spec says as follows: 2723 * If the interface index is specified as 0, the 2724 * system may choose a multicast group membership to 2725 * drop by matching the multicast address only. 2726 * On the other hand, we cannot disambiguate the scope 2727 * zone unless an interface is provided. Thus, we 2728 * check if there's ambiguity with the default scope 2729 * zone as the last resort. 2730 */ 2731 bzero(&sa6_mc, sizeof(sa6_mc)); 2732 sa6_mc.sin6_family = AF_INET6; 2733 sa6_mc.sin6_len = sizeof(sa6_mc); 2734 sa6_mc.sin6_addr = mreq->ipv6mr_multiaddr; 2735 error = sa6_embedscope(&sa6_mc, ip6_use_defzone); 2736 if (error != 0) 2737 break; 2738 mreq->ipv6mr_multiaddr = sa6_mc.sin6_addr; 2739 } 2740 2741 /* 2742 * Find the membership in the membership list. 2743 */ 2744 for (imm = im6o->im6o_memberships.lh_first; 2745 imm != NULL; imm = imm->i6mm_chain.le_next) { 2746 if ((ifp == NULL || imm->i6mm_maddr->in6m_ifp == ifp) && 2747 IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr, 2748 &mreq->ipv6mr_multiaddr)) 2749 break; 2750 } 2751 if (imm == NULL) { 2752 /* Unable to resolve interface */ 2753 error = EADDRNOTAVAIL; 2754 break; 2755 } 2756 /* 2757 * Give up the multicast address record to which the 2758 * membership points. 2759 */ 2760 LIST_REMOVE(imm, i6mm_chain); 2761 in6_leavegroup(imm); 2762 break; 2763 2764 default: 2765 error = EOPNOTSUPP; 2766 break; 2767 } 2768 2769 /* 2770 * If all options have default values, no need to keep the mbuf. 2771 */ 2772 if (im6o->im6o_multicast_ifp == NULL && 2773 im6o->im6o_multicast_hlim == ip6_defmcasthlim && 2774 im6o->im6o_multicast_loop == IPV6_DEFAULT_MULTICAST_LOOP && 2775 im6o->im6o_memberships.lh_first == NULL) { 2776 free(*im6op, M_IPMOPTS); 2777 *im6op = NULL; 2778 } 2779 2780 return (error); 2781 } 2782 2783 /* 2784 * Return the IP6 multicast options in response to user getsockopt(). 2785 */ 2786 static int 2787 ip6_getmoptions(int optname, struct ip6_moptions *im6o, struct mbuf **mp) 2788 { 2789 u_int *hlim, *loop, *ifindex; 2790 2791 *mp = m_get(M_WAIT, MT_SOOPTS); 2792 2793 switch (optname) { 2794 2795 case IPV6_MULTICAST_IF: 2796 ifindex = mtod(*mp, u_int *); 2797 (*mp)->m_len = sizeof(u_int); 2798 if (im6o == NULL || im6o->im6o_multicast_ifp == NULL) 2799 *ifindex = 0; 2800 else 2801 *ifindex = im6o->im6o_multicast_ifp->if_index; 2802 return (0); 2803 2804 case IPV6_MULTICAST_HOPS: 2805 hlim = mtod(*mp, u_int *); 2806 (*mp)->m_len = sizeof(u_int); 2807 if (im6o == NULL) 2808 *hlim = ip6_defmcasthlim; 2809 else 2810 *hlim = im6o->im6o_multicast_hlim; 2811 return (0); 2812 2813 case IPV6_MULTICAST_LOOP: 2814 loop = mtod(*mp, u_int *); 2815 (*mp)->m_len = sizeof(u_int); 2816 if (im6o == NULL) 2817 *loop = ip6_defmcasthlim; 2818 else 2819 *loop = im6o->im6o_multicast_loop; 2820 return (0); 2821 2822 default: 2823 return (EOPNOTSUPP); 2824 } 2825 } 2826 2827 /* 2828 * Discard the IP6 multicast options. 2829 */ 2830 void 2831 ip6_freemoptions(struct ip6_moptions *im6o) 2832 { 2833 struct in6_multi_mship *imm; 2834 2835 if (im6o == NULL) 2836 return; 2837 2838 while ((imm = im6o->im6o_memberships.lh_first) != NULL) { 2839 LIST_REMOVE(imm, i6mm_chain); 2840 in6_leavegroup(imm); 2841 } 2842 free(im6o, M_IPMOPTS); 2843 } 2844 2845 /* 2846 * Set IPv6 outgoing packet options based on advanced API. 2847 */ 2848 int 2849 ip6_setpktopts(struct mbuf *control, struct ip6_pktopts *opt, 2850 struct ip6_pktopts *stickyopt, int priv, int uproto) 2851 { 2852 struct cmsghdr *cm = 0; 2853 2854 if (control == NULL || opt == NULL) 2855 return (EINVAL); 2856 2857 ip6_initpktopts(opt); 2858 if (stickyopt) { 2859 int error; 2860 2861 /* 2862 * If stickyopt is provided, make a local copy of the options 2863 * for this particular packet, then override them by ancillary 2864 * objects. 2865 * XXX: copypktopts() does not copy the cached route to a next 2866 * hop (if any). This is not very good in terms of efficiency, 2867 * but we can allow this since this option should be rarely 2868 * used. 2869 */ 2870 if ((error = copypktopts(opt, stickyopt, M_NOWAIT)) != 0) 2871 return (error); 2872 } 2873 2874 /* 2875 * XXX: Currently, we assume all the optional information is stored 2876 * in a single mbuf. 2877 */ 2878 if (control->m_next) 2879 return (EINVAL); 2880 2881 for (; control->m_len; control->m_data += CMSG_ALIGN(cm->cmsg_len), 2882 control->m_len -= CMSG_ALIGN(cm->cmsg_len)) { 2883 int error; 2884 2885 if (control->m_len < CMSG_LEN(0)) 2886 return (EINVAL); 2887 2888 cm = mtod(control, struct cmsghdr *); 2889 if (cm->cmsg_len == 0 || cm->cmsg_len > control->m_len) 2890 return (EINVAL); 2891 if (cm->cmsg_level != IPPROTO_IPV6) 2892 continue; 2893 2894 error = ip6_setpktopt(cm->cmsg_type, CMSG_DATA(cm), 2895 cm->cmsg_len - CMSG_LEN(0), opt, priv, 0, 1, uproto); 2896 if (error) 2897 return (error); 2898 } 2899 2900 return (0); 2901 } 2902 2903 /* 2904 * Set a particular packet option, as a sticky option or an ancillary data 2905 * item. "len" can be 0 only when it's a sticky option. 2906 * We have 4 cases of combination of "sticky" and "cmsg": 2907 * "sticky=0, cmsg=0": impossible 2908 * "sticky=0, cmsg=1": RFC2292 or RFC3542 ancillary data 2909 * "sticky=1, cmsg=0": RFC3542 socket option 2910 * "sticky=1, cmsg=1": RFC2292 socket option 2911 */ 2912 static int 2913 ip6_setpktopt(int optname, u_char *buf, int len, struct ip6_pktopts *opt, 2914 int priv, int sticky, int cmsg, int uproto) 2915 { 2916 int minmtupolicy; 2917 2918 if (!sticky && !cmsg) { 2919 #ifdef DIAGNOSTIC 2920 printf("ip6_setpktopt: impossible case\n"); 2921 #endif 2922 return (EINVAL); 2923 } 2924 2925 /* 2926 * IPV6_2292xxx is for backward compatibility to RFC2292, and should 2927 * not be specified in the context of RFC3542. Conversely, 2928 * RFC3542 types should not be specified in the context of RFC2292. 2929 */ 2930 if (!cmsg) { 2931 switch (optname) { 2932 case IPV6_2292PKTINFO: 2933 case IPV6_2292HOPLIMIT: 2934 case IPV6_2292NEXTHOP: 2935 case IPV6_2292HOPOPTS: 2936 case IPV6_2292DSTOPTS: 2937 case IPV6_2292RTHDR: 2938 case IPV6_2292PKTOPTIONS: 2939 return (ENOPROTOOPT); 2940 } 2941 } 2942 if (sticky && cmsg) { 2943 switch (optname) { 2944 case IPV6_PKTINFO: 2945 case IPV6_HOPLIMIT: 2946 case IPV6_NEXTHOP: 2947 case IPV6_HOPOPTS: 2948 case IPV6_DSTOPTS: 2949 case IPV6_RTHDRDSTOPTS: 2950 case IPV6_RTHDR: 2951 case IPV6_USE_MIN_MTU: 2952 case IPV6_DONTFRAG: 2953 case IPV6_OTCLASS: 2954 case IPV6_TCLASS: 2955 return (ENOPROTOOPT); 2956 } 2957 } 2958 2959 switch (optname) { 2960 #ifdef RFC2292 2961 case IPV6_2292PKTINFO: 2962 #endif 2963 case IPV6_PKTINFO: 2964 { 2965 struct ifnet *ifp = NULL; 2966 struct in6_pktinfo *pktinfo; 2967 2968 if (len != sizeof(struct in6_pktinfo)) 2969 return (EINVAL); 2970 2971 pktinfo = (struct in6_pktinfo *)buf; 2972 2973 /* 2974 * An application can clear any sticky IPV6_PKTINFO option by 2975 * doing a "regular" setsockopt with ipi6_addr being 2976 * in6addr_any and ipi6_ifindex being zero. 2977 * [RFC 3542, Section 6] 2978 */ 2979 if (optname == IPV6_PKTINFO && opt->ip6po_pktinfo && 2980 pktinfo->ipi6_ifindex == 0 && 2981 IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) { 2982 ip6_clearpktopts(opt, optname); 2983 break; 2984 } 2985 2986 if (uproto == IPPROTO_TCP && optname == IPV6_PKTINFO && 2987 sticky && !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) { 2988 return (EINVAL); 2989 } 2990 2991 /* validate the interface index if specified. */ 2992 if (pktinfo->ipi6_ifindex >= if_indexlim) { 2993 return (ENXIO); 2994 } 2995 if (pktinfo->ipi6_ifindex) { 2996 ifp = ifindex2ifnet[pktinfo->ipi6_ifindex]; 2997 if (ifp == NULL) 2998 return (ENXIO); 2999 } 3000 3001 /* 3002 * We store the address anyway, and let in6_selectsrc() 3003 * validate the specified address. This is because ipi6_addr 3004 * may not have enough information about its scope zone, and 3005 * we may need additional information (such as outgoing 3006 * interface or the scope zone of a destination address) to 3007 * disambiguate the scope. 3008 * XXX: the delay of the validation may confuse the 3009 * application when it is used as a sticky option. 3010 */ 3011 if (opt->ip6po_pktinfo == NULL) { 3012 opt->ip6po_pktinfo = malloc(sizeof(*pktinfo), 3013 M_IP6OPT, M_NOWAIT); 3014 if (opt->ip6po_pktinfo == NULL) 3015 return (ENOBUFS); 3016 } 3017 memcpy(opt->ip6po_pktinfo, pktinfo, sizeof(*pktinfo)); 3018 break; 3019 } 3020 3021 #ifdef RFC2292 3022 case IPV6_2292HOPLIMIT: 3023 #endif 3024 case IPV6_HOPLIMIT: 3025 { 3026 int *hlimp; 3027 3028 /* 3029 * RFC 3542 deprecated the usage of sticky IPV6_HOPLIMIT 3030 * to simplify the ordering among hoplimit options. 3031 */ 3032 if (optname == IPV6_HOPLIMIT && sticky) 3033 return (ENOPROTOOPT); 3034 3035 if (len != sizeof(int)) 3036 return (EINVAL); 3037 hlimp = (int *)buf; 3038 if (*hlimp < -1 || *hlimp > 255) 3039 return (EINVAL); 3040 3041 opt->ip6po_hlim = *hlimp; 3042 break; 3043 } 3044 3045 case IPV6_OTCLASS: 3046 if (len != sizeof(u_int8_t)) 3047 return (EINVAL); 3048 3049 opt->ip6po_tclass = *(u_int8_t *)buf; 3050 break; 3051 3052 case IPV6_TCLASS: 3053 { 3054 int tclass; 3055 3056 if (len != sizeof(int)) 3057 return (EINVAL); 3058 tclass = *(int *)buf; 3059 if (tclass < -1 || tclass > 255) 3060 return (EINVAL); 3061 3062 opt->ip6po_tclass = tclass; 3063 break; 3064 } 3065 3066 #ifdef RFC2292 3067 case IPV6_2292NEXTHOP: 3068 #endif 3069 case IPV6_NEXTHOP: 3070 if (!priv) 3071 return (EPERM); 3072 3073 if (len == 0) { /* just remove the option */ 3074 ip6_clearpktopts(opt, IPV6_NEXTHOP); 3075 break; 3076 } 3077 3078 /* check if cmsg_len is large enough for sa_len */ 3079 if (len < sizeof(struct sockaddr) || len < *buf) 3080 return (EINVAL); 3081 3082 switch (((struct sockaddr *)buf)->sa_family) { 3083 case AF_INET6: 3084 { 3085 struct sockaddr_in6 *sa6 = (struct sockaddr_in6 *)buf; 3086 int error; 3087 3088 if (sa6->sin6_len != sizeof(struct sockaddr_in6)) 3089 return (EINVAL); 3090 3091 if (IN6_IS_ADDR_UNSPECIFIED(&sa6->sin6_addr) || 3092 IN6_IS_ADDR_MULTICAST(&sa6->sin6_addr)) { 3093 return (EINVAL); 3094 } 3095 if ((error = sa6_embedscope(sa6, ip6_use_defzone)) 3096 != 0) { 3097 return (error); 3098 } 3099 break; 3100 } 3101 case AF_LINK: /* eventually be supported? */ 3102 default: 3103 return (EAFNOSUPPORT); 3104 } 3105 3106 /* turn off the previous option, then set the new option. */ 3107 ip6_clearpktopts(opt, IPV6_NEXTHOP); 3108 opt->ip6po_nexthop = malloc(*buf, M_IP6OPT, M_NOWAIT); 3109 if (opt->ip6po_nexthop == NULL) 3110 return (ENOBUFS); 3111 memcpy(opt->ip6po_nexthop, buf, *buf); 3112 break; 3113 3114 #ifdef RFC2292 3115 case IPV6_2292HOPOPTS: 3116 #endif 3117 case IPV6_HOPOPTS: 3118 { 3119 struct ip6_hbh *hbh; 3120 int hbhlen; 3121 3122 /* 3123 * XXX: We don't allow a non-privileged user to set ANY HbH 3124 * options, since per-option restriction has too much 3125 * overhead. 3126 */ 3127 if (!priv) 3128 return (EPERM); 3129 3130 if (len == 0) { 3131 ip6_clearpktopts(opt, IPV6_HOPOPTS); 3132 break; /* just remove the option */ 3133 } 3134 3135 /* message length validation */ 3136 if (len < sizeof(struct ip6_hbh)) 3137 return (EINVAL); 3138 hbh = (struct ip6_hbh *)buf; 3139 hbhlen = (hbh->ip6h_len + 1) << 3; 3140 if (len != hbhlen) 3141 return (EINVAL); 3142 3143 /* turn off the previous option, then set the new option. */ 3144 ip6_clearpktopts(opt, IPV6_HOPOPTS); 3145 opt->ip6po_hbh = malloc(hbhlen, M_IP6OPT, M_NOWAIT); 3146 if (opt->ip6po_hbh == NULL) 3147 return (ENOBUFS); 3148 memcpy(opt->ip6po_hbh, hbh, hbhlen); 3149 3150 break; 3151 } 3152 3153 #ifdef RFC2292 3154 case IPV6_2292DSTOPTS: 3155 #endif 3156 case IPV6_DSTOPTS: 3157 case IPV6_RTHDRDSTOPTS: 3158 { 3159 struct ip6_dest *dest, **newdest = NULL; 3160 int destlen; 3161 3162 if (!priv) /* XXX: see the comment for IPV6_HOPOPTS */ 3163 return (EPERM); 3164 3165 if (len == 0) { 3166 ip6_clearpktopts(opt, optname); 3167 break; /* just remove the option */ 3168 } 3169 3170 /* message length validation */ 3171 if (len < sizeof(struct ip6_dest)) 3172 return (EINVAL); 3173 dest = (struct ip6_dest *)buf; 3174 destlen = (dest->ip6d_len + 1) << 3; 3175 if (len != destlen) 3176 return (EINVAL); 3177 /* 3178 * Determine the position that the destination options header 3179 * should be inserted; before or after the routing header. 3180 */ 3181 switch (optname) { 3182 case IPV6_2292DSTOPTS: 3183 /* 3184 * The old advanced API is ambiguous on this point. 3185 * Our approach is to determine the position based 3186 * according to the existence of a routing header. 3187 * Note, however, that this depends on the order of the 3188 * extension headers in the ancillary data; the 1st 3189 * part of the destination options header must appear 3190 * before the routing header in the ancillary data, 3191 * too. 3192 * RFC3542 solved the ambiguity by introducing 3193 * separate ancillary data or option types. 3194 */ 3195 if (opt->ip6po_rthdr == NULL) 3196 newdest = &opt->ip6po_dest1; 3197 else 3198 newdest = &opt->ip6po_dest2; 3199 break; 3200 case IPV6_RTHDRDSTOPTS: 3201 newdest = &opt->ip6po_dest1; 3202 break; 3203 case IPV6_DSTOPTS: 3204 newdest = &opt->ip6po_dest2; 3205 break; 3206 } 3207 3208 /* turn off the previous option, then set the new option. */ 3209 ip6_clearpktopts(opt, optname); 3210 *newdest = malloc(destlen, M_IP6OPT, M_NOWAIT); 3211 if (*newdest == NULL) 3212 return (ENOBUFS); 3213 memcpy(*newdest, dest, destlen); 3214 3215 break; 3216 } 3217 3218 #ifdef RFC2292 3219 case IPV6_2292RTHDR: 3220 #endif 3221 case IPV6_RTHDR: 3222 { 3223 struct ip6_rthdr *rth; 3224 int rthlen; 3225 3226 if (len == 0) { 3227 ip6_clearpktopts(opt, IPV6_RTHDR); 3228 break; /* just remove the option */ 3229 } 3230 3231 /* message length validation */ 3232 if (len < sizeof(struct ip6_rthdr)) 3233 return (EINVAL); 3234 rth = (struct ip6_rthdr *)buf; 3235 rthlen = (rth->ip6r_len + 1) << 3; 3236 if (len != rthlen) 3237 return (EINVAL); 3238 switch (rth->ip6r_type) { 3239 case IPV6_RTHDR_TYPE_0: 3240 if (rth->ip6r_len == 0) /* must contain one addr */ 3241 return (EINVAL); 3242 if (rth->ip6r_len % 2) /* length must be even */ 3243 return (EINVAL); 3244 if (rth->ip6r_len / 2 != rth->ip6r_segleft) 3245 return (EINVAL); 3246 break; 3247 default: 3248 return (EINVAL); /* not supported */ 3249 } 3250 /* turn off the previous option */ 3251 ip6_clearpktopts(opt, IPV6_RTHDR); 3252 opt->ip6po_rthdr = malloc(rthlen, M_IP6OPT, M_NOWAIT); 3253 if (opt->ip6po_rthdr == NULL) 3254 return (ENOBUFS); 3255 memcpy(opt->ip6po_rthdr, rth, rthlen); 3256 break; 3257 } 3258 3259 case IPV6_USE_MIN_MTU: 3260 if (len != sizeof(int)) 3261 return (EINVAL); 3262 minmtupolicy = *(int *)buf; 3263 if (minmtupolicy != IP6PO_MINMTU_MCASTONLY && 3264 minmtupolicy != IP6PO_MINMTU_DISABLE && 3265 minmtupolicy != IP6PO_MINMTU_ALL) { 3266 return (EINVAL); 3267 } 3268 opt->ip6po_minmtu = minmtupolicy; 3269 break; 3270 3271 case IPV6_DONTFRAG: 3272 if (len != sizeof(int)) 3273 return (EINVAL); 3274 3275 if (uproto == IPPROTO_TCP || *(int *)buf == 0) { 3276 /* 3277 * we ignore this option for TCP sockets. 3278 * (RFC3542 leaves this case unspecified.) 3279 */ 3280 opt->ip6po_flags &= ~IP6PO_DONTFRAG; 3281 } else 3282 opt->ip6po_flags |= IP6PO_DONTFRAG; 3283 break; 3284 3285 default: 3286 return (ENOPROTOOPT); 3287 } /* end of switch */ 3288 3289 return (0); 3290 } 3291 3292 /* 3293 * Routine called from ip6_output() to loop back a copy of an IP6 multicast 3294 * packet to the input queue of a specified interface. Note that this 3295 * calls the output routine of the loopback "driver", but with an interface 3296 * pointer that might NOT be lo0ifp -- easier than replicating that code here. 3297 */ 3298 void 3299 ip6_mloopback(struct ifnet *ifp, struct mbuf *m, 3300 const struct sockaddr_in6 *dst) 3301 { 3302 struct mbuf *copym; 3303 struct ip6_hdr *ip6; 3304 3305 copym = m_copy(m, 0, M_COPYALL); 3306 if (copym == NULL) 3307 return; 3308 3309 /* 3310 * Make sure to deep-copy IPv6 header portion in case the data 3311 * is in an mbuf cluster, so that we can safely override the IPv6 3312 * header portion later. 3313 */ 3314 if ((copym->m_flags & M_EXT) != 0 || 3315 copym->m_len < sizeof(struct ip6_hdr)) { 3316 copym = m_pullup(copym, sizeof(struct ip6_hdr)); 3317 if (copym == NULL) 3318 return; 3319 } 3320 3321 #ifdef DIAGNOSTIC 3322 if (copym->m_len < sizeof(*ip6)) { 3323 m_freem(copym); 3324 return; 3325 } 3326 #endif 3327 3328 ip6 = mtod(copym, struct ip6_hdr *); 3329 /* 3330 * clear embedded scope identifiers if necessary. 3331 * in6_clearscope will touch the addresses only when necessary. 3332 */ 3333 in6_clearscope(&ip6->ip6_src); 3334 in6_clearscope(&ip6->ip6_dst); 3335 3336 (void)looutput(ifp, copym, (const struct sockaddr *)dst, NULL); 3337 } 3338 3339 /* 3340 * Chop IPv6 header off from the payload. 3341 */ 3342 static int 3343 ip6_splithdr(struct mbuf *m, struct ip6_exthdrs *exthdrs) 3344 { 3345 struct mbuf *mh; 3346 struct ip6_hdr *ip6; 3347 3348 ip6 = mtod(m, struct ip6_hdr *); 3349 if (m->m_len > sizeof(*ip6)) { 3350 MGETHDR(mh, M_DONTWAIT, MT_HEADER); 3351 if (mh == 0) { 3352 m_freem(m); 3353 return ENOBUFS; 3354 } 3355 M_MOVE_PKTHDR(mh, m); 3356 MH_ALIGN(mh, sizeof(*ip6)); 3357 m->m_len -= sizeof(*ip6); 3358 m->m_data += sizeof(*ip6); 3359 mh->m_next = m; 3360 m = mh; 3361 m->m_len = sizeof(*ip6); 3362 bcopy((void *)ip6, mtod(m, void *), sizeof(*ip6)); 3363 } 3364 exthdrs->ip6e_ip6 = m; 3365 return 0; 3366 } 3367 3368 /* 3369 * Compute IPv6 extension header length. 3370 */ 3371 int 3372 ip6_optlen(struct in6pcb *in6p) 3373 { 3374 int len; 3375 3376 if (!in6p->in6p_outputopts) 3377 return 0; 3378 3379 len = 0; 3380 #define elen(x) \ 3381 (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0) 3382 3383 len += elen(in6p->in6p_outputopts->ip6po_hbh); 3384 len += elen(in6p->in6p_outputopts->ip6po_dest1); 3385 len += elen(in6p->in6p_outputopts->ip6po_rthdr); 3386 len += elen(in6p->in6p_outputopts->ip6po_dest2); 3387 return len; 3388 #undef elen 3389 } 3390
Indexes created Tue Sep 30 20:09:53 GMT 2025