ip6_output.c revision 1.117
1 /* $NetBSD: ip6_output.c,v 1.117 2007/03/04 06:03:26 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 1.117 2007/03/04 06:03:26 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_in6 *, struct route_in6 *, 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_in6 *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_in6 ip6route; 182 struct rtentry *rt = NULL; 183 struct sockaddr_in6 *dst, src_sa, dst_sa; 184 int error = 0; 185 struct in6_ifaddr *ia = NULL; 186 u_long mtu; 187 int alwaysfrag, dontfrag; 188 u_int32_t optlen = 0, plen = 0, unfragpartlen = 0; 189 struct ip6_exthdrs exthdrs; 190 struct in6_addr finaldst, src0, dst0; 191 u_int32_t zone; 192 struct route_in6 *ro_pmtu = NULL; 193 int hdrsplit = 0; 194 int needipsec = 0; 195 #ifdef IPSEC 196 int needipsectun = 0; 197 struct secpolicy *sp = NULL; 198 199 ip6 = mtod(m, struct ip6_hdr *); 200 #endif /* IPSEC */ 201 #ifdef FAST_IPSEC 202 struct secpolicy *sp = NULL; 203 int s; 204 #endif 205 206 207 #ifdef DIAGNOSTIC 208 if ((m->m_flags & M_PKTHDR) == 0) 209 panic("ip6_output: no HDR"); 210 211 if ((m->m_pkthdr.csum_flags & 212 (M_CSUM_TCPv4|M_CSUM_UDPv4|M_CSUM_TSOv4)) != 0) { 213 panic("ip6_output: IPv4 checksum offload flags: %d", 214 m->m_pkthdr.csum_flags); 215 } 216 217 if ((m->m_pkthdr.csum_flags & (M_CSUM_TCPv6|M_CSUM_UDPv6)) == 218 (M_CSUM_TCPv6|M_CSUM_UDPv6)) { 219 panic("ip6_output: conflicting checksum offload flags: %d", 220 m->m_pkthdr.csum_flags); 221 } 222 #endif 223 224 M_CSUM_DATA_IPv6_HL_SET(m->m_pkthdr.csum_data, sizeof(struct ip6_hdr)); 225 226 #define MAKE_EXTHDR(hp, mp) \ 227 do { \ 228 if (hp) { \ 229 struct ip6_ext *eh = (struct ip6_ext *)(hp); \ 230 error = ip6_copyexthdr((mp), (void *)(hp), \ 231 ((eh)->ip6e_len + 1) << 3); \ 232 if (error) \ 233 goto freehdrs; \ 234 } \ 235 } while (/*CONSTCOND*/ 0) 236 237 bzero(&exthdrs, sizeof(exthdrs)); 238 if (opt) { 239 /* Hop-by-Hop options header */ 240 MAKE_EXTHDR(opt->ip6po_hbh, &exthdrs.ip6e_hbh); 241 /* Destination options header(1st part) */ 242 MAKE_EXTHDR(opt->ip6po_dest1, &exthdrs.ip6e_dest1); 243 /* Routing header */ 244 MAKE_EXTHDR(opt->ip6po_rthdr, &exthdrs.ip6e_rthdr); 245 /* Destination options header(2nd part) */ 246 MAKE_EXTHDR(opt->ip6po_dest2, &exthdrs.ip6e_dest2); 247 } 248 249 #ifdef IPSEC 250 if ((flags & IPV6_FORWARDING) != 0) { 251 needipsec = 0; 252 goto skippolicycheck; 253 } 254 255 /* get a security policy for this packet */ 256 if (so == NULL) 257 sp = ipsec6_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND, 0, &error); 258 else { 259 if (IPSEC_PCB_SKIP_IPSEC(sotoinpcb_hdr(so)->inph_sp, 260 IPSEC_DIR_OUTBOUND)) { 261 needipsec = 0; 262 goto skippolicycheck; 263 } 264 sp = ipsec6_getpolicybysock(m, IPSEC_DIR_OUTBOUND, so, &error); 265 } 266 267 if (sp == NULL) { 268 ipsec6stat.out_inval++; 269 goto freehdrs; 270 } 271 272 error = 0; 273 274 /* check policy */ 275 switch (sp->policy) { 276 case IPSEC_POLICY_DISCARD: 277 /* 278 * This packet is just discarded. 279 */ 280 ipsec6stat.out_polvio++; 281 goto freehdrs; 282 283 case IPSEC_POLICY_BYPASS: 284 case IPSEC_POLICY_NONE: 285 /* no need to do IPsec. */ 286 needipsec = 0; 287 break; 288 289 case IPSEC_POLICY_IPSEC: 290 if (sp->req == NULL) { 291 /* XXX should be panic ? */ 292 printf("ip6_output: No IPsec request specified.\n"); 293 error = EINVAL; 294 goto freehdrs; 295 } 296 needipsec = 1; 297 break; 298 299 case IPSEC_POLICY_ENTRUST: 300 default: 301 printf("ip6_output: Invalid policy found. %d\n", sp->policy); 302 } 303 304 skippolicycheck:; 305 #endif /* IPSEC */ 306 307 /* 308 * Calculate the total length of the extension header chain. 309 * Keep the length of the unfragmentable part for fragmentation. 310 */ 311 optlen = 0; 312 if (exthdrs.ip6e_hbh) optlen += exthdrs.ip6e_hbh->m_len; 313 if (exthdrs.ip6e_dest1) optlen += exthdrs.ip6e_dest1->m_len; 314 if (exthdrs.ip6e_rthdr) optlen += exthdrs.ip6e_rthdr->m_len; 315 unfragpartlen = optlen + sizeof(struct ip6_hdr); 316 /* NOTE: we don't add AH/ESP length here. do that later. */ 317 if (exthdrs.ip6e_dest2) optlen += exthdrs.ip6e_dest2->m_len; 318 319 #ifdef FAST_IPSEC 320 /* Check the security policy (SP) for the packet */ 321 322 /* XXX For moment, we doesn't support packet with extented action */ 323 if (optlen !=0) 324 goto freehdrs; 325 326 sp = ipsec6_check_policy(m,so,flags,&needipsec,&error); 327 if (error != 0) { 328 /* 329 * Hack: -EINVAL is used to signal that a packet 330 * should be silently discarded. This is typically 331 * because we asked key management for an SA and 332 * it was delayed (e.g. kicked up to IKE). 333 */ 334 if (error == -EINVAL) 335 error = 0; 336 goto freehdrs; 337 } 338 #endif /* FAST_IPSEC */ 339 340 341 if (needipsec && 342 (m->m_pkthdr.csum_flags & (M_CSUM_UDPv6|M_CSUM_TCPv6)) != 0) { 343 in6_delayed_cksum(m); 344 m->m_pkthdr.csum_flags &= ~(M_CSUM_UDPv6|M_CSUM_TCPv6); 345 } 346 347 348 /* 349 * If we need IPsec, or there is at least one extension header, 350 * separate IP6 header from the payload. 351 */ 352 if ((needipsec || optlen) && !hdrsplit) { 353 if ((error = ip6_splithdr(m, &exthdrs)) != 0) { 354 m = NULL; 355 goto freehdrs; 356 } 357 m = exthdrs.ip6e_ip6; 358 hdrsplit++; 359 } 360 361 /* adjust pointer */ 362 ip6 = mtod(m, struct ip6_hdr *); 363 364 /* adjust mbuf packet header length */ 365 m->m_pkthdr.len += optlen; 366 plen = m->m_pkthdr.len - sizeof(*ip6); 367 368 /* If this is a jumbo payload, insert a jumbo payload option. */ 369 if (plen > IPV6_MAXPACKET) { 370 if (!hdrsplit) { 371 if ((error = ip6_splithdr(m, &exthdrs)) != 0) { 372 m = NULL; 373 goto freehdrs; 374 } 375 m = exthdrs.ip6e_ip6; 376 hdrsplit++; 377 } 378 /* adjust pointer */ 379 ip6 = mtod(m, struct ip6_hdr *); 380 if ((error = ip6_insert_jumboopt(&exthdrs, plen)) != 0) 381 goto freehdrs; 382 optlen += 8; /* XXX JUMBOOPTLEN */ 383 ip6->ip6_plen = 0; 384 } else 385 ip6->ip6_plen = htons(plen); 386 387 /* 388 * Concatenate headers and fill in next header fields. 389 * Here we have, on "m" 390 * IPv6 payload 391 * and we insert headers accordingly. Finally, we should be getting: 392 * IPv6 hbh dest1 rthdr ah* [esp* dest2 payload] 393 * 394 * during the header composing process, "m" points to IPv6 header. 395 * "mprev" points to an extension header prior to esp. 396 */ 397 { 398 u_char *nexthdrp = &ip6->ip6_nxt; 399 struct mbuf *mprev = m; 400 401 /* 402 * we treat dest2 specially. this makes IPsec processing 403 * much easier. the goal here is to make mprev point the 404 * mbuf prior to dest2. 405 * 406 * result: IPv6 dest2 payload 407 * m and mprev will point to IPv6 header. 408 */ 409 if (exthdrs.ip6e_dest2) { 410 if (!hdrsplit) 411 panic("assumption failed: hdr not split"); 412 exthdrs.ip6e_dest2->m_next = m->m_next; 413 m->m_next = exthdrs.ip6e_dest2; 414 *mtod(exthdrs.ip6e_dest2, u_char *) = ip6->ip6_nxt; 415 ip6->ip6_nxt = IPPROTO_DSTOPTS; 416 } 417 418 #define MAKE_CHAIN(m, mp, p, i)\ 419 do {\ 420 if (m) {\ 421 if (!hdrsplit) \ 422 panic("assumption failed: hdr not split"); \ 423 *mtod((m), u_char *) = *(p);\ 424 *(p) = (i);\ 425 p = mtod((m), u_char *);\ 426 (m)->m_next = (mp)->m_next;\ 427 (mp)->m_next = (m);\ 428 (mp) = (m);\ 429 }\ 430 } while (/*CONSTCOND*/ 0) 431 /* 432 * result: IPv6 hbh dest1 rthdr dest2 payload 433 * m will point to IPv6 header. mprev will point to the 434 * extension header prior to dest2 (rthdr in the above case). 435 */ 436 MAKE_CHAIN(exthdrs.ip6e_hbh, mprev, nexthdrp, IPPROTO_HOPOPTS); 437 MAKE_CHAIN(exthdrs.ip6e_dest1, mprev, nexthdrp, 438 IPPROTO_DSTOPTS); 439 MAKE_CHAIN(exthdrs.ip6e_rthdr, mprev, nexthdrp, 440 IPPROTO_ROUTING); 441 442 M_CSUM_DATA_IPv6_HL_SET(m->m_pkthdr.csum_data, 443 sizeof(struct ip6_hdr) + optlen); 444 445 #ifdef IPSEC 446 if (!needipsec) 447 goto skip_ipsec2; 448 449 /* 450 * pointers after IPsec headers are not valid any more. 451 * other pointers need a great care too. 452 * (IPsec routines should not mangle mbufs prior to AH/ESP) 453 */ 454 exthdrs.ip6e_dest2 = NULL; 455 456 { 457 struct ip6_rthdr *rh = NULL; 458 int segleft_org = 0; 459 struct ipsec_output_state state; 460 461 if (exthdrs.ip6e_rthdr) { 462 rh = mtod(exthdrs.ip6e_rthdr, struct ip6_rthdr *); 463 segleft_org = rh->ip6r_segleft; 464 rh->ip6r_segleft = 0; 465 } 466 467 bzero(&state, sizeof(state)); 468 state.m = m; 469 error = ipsec6_output_trans(&state, nexthdrp, mprev, sp, flags, 470 &needipsectun); 471 m = state.m; 472 if (error) { 473 rh = mtod(exthdrs.ip6e_rthdr, struct ip6_rthdr *); 474 /* mbuf is already reclaimed in ipsec6_output_trans. */ 475 m = NULL; 476 switch (error) { 477 case EHOSTUNREACH: 478 case ENETUNREACH: 479 case EMSGSIZE: 480 case ENOBUFS: 481 case ENOMEM: 482 break; 483 default: 484 printf("ip6_output (ipsec): error code %d\n", error); 485 /* FALLTHROUGH */ 486 case ENOENT: 487 /* don't show these error codes to the user */ 488 error = 0; 489 break; 490 } 491 goto bad; 492 } 493 if (exthdrs.ip6e_rthdr) { 494 /* ah6_output doesn't modify mbuf chain */ 495 rh->ip6r_segleft = segleft_org; 496 } 497 } 498 skip_ipsec2:; 499 #endif 500 } 501 502 /* 503 * If there is a routing header, replace destination address field 504 * with the first hop of the routing header. 505 */ 506 if (exthdrs.ip6e_rthdr) { 507 struct ip6_rthdr *rh; 508 struct ip6_rthdr0 *rh0; 509 struct in6_addr *addr; 510 struct sockaddr_in6 sa; 511 512 rh = (struct ip6_rthdr *)(mtod(exthdrs.ip6e_rthdr, 513 struct ip6_rthdr *)); 514 finaldst = ip6->ip6_dst; 515 switch (rh->ip6r_type) { 516 case IPV6_RTHDR_TYPE_0: 517 rh0 = (struct ip6_rthdr0 *)rh; 518 addr = (struct in6_addr *)(rh0 + 1); 519 520 /* 521 * construct a sockaddr_in6 form of 522 * the first hop. 523 * 524 * XXX: we may not have enough 525 * information about its scope zone; 526 * there is no standard API to pass 527 * the information from the 528 * application. 529 */ 530 bzero(&sa, sizeof(sa)); 531 sa.sin6_family = AF_INET6; 532 sa.sin6_len = sizeof(sa); 533 sa.sin6_addr = addr[0]; 534 if ((error = sa6_embedscope(&sa, 535 ip6_use_defzone)) != 0) { 536 goto bad; 537 } 538 ip6->ip6_dst = sa.sin6_addr; 539 (void)memmove(&addr[0], &addr[1], 540 sizeof(struct in6_addr) * 541 (rh0->ip6r0_segleft - 1)); 542 addr[rh0->ip6r0_segleft - 1] = finaldst; 543 /* XXX */ 544 in6_clearscope(addr + rh0->ip6r0_segleft - 1); 545 break; 546 default: /* is it possible? */ 547 error = EINVAL; 548 goto bad; 549 } 550 } 551 552 /* Source address validation */ 553 if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) && 554 (flags & IPV6_UNSPECSRC) == 0) { 555 error = EOPNOTSUPP; 556 ip6stat.ip6s_badscope++; 557 goto bad; 558 } 559 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) { 560 error = EOPNOTSUPP; 561 ip6stat.ip6s_badscope++; 562 goto bad; 563 } 564 565 ip6stat.ip6s_localout++; 566 567 /* 568 * Route packet. 569 */ 570 /* initialize cached route */ 571 if (ro == NULL) { 572 memset(&ip6route, 0, sizeof(ip6route)); 573 ro = &ip6route; 574 } 575 ro_pmtu = ro; 576 if (opt && opt->ip6po_rthdr) 577 ro = &opt->ip6po_route; 578 dst = (struct sockaddr_in6 *)&ro->ro_dst; 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 = (struct route *)ro; 624 state.dst = (struct sockaddr *)dst; 625 626 error = ipsec6_output_tunnel(&state, sp, flags); 627 628 m = state.m; 629 ro_pmtu = ro = (struct route_in6 *)state.ro; 630 dst = (struct sockaddr_in6 *)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, (struct route *)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 *dst = dst_sa; /* XXX */ 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 goto routefound; 750 751 badscope: 752 ip6stat.ip6s_badscope++; 753 in6_ifstat_inc(origifp, ifs6_out_discard); 754 if (error == 0) 755 error = EHOSTUNREACH; /* XXX */ 756 goto bad; 757 758 routefound: 759 if (rt && !IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) { 760 if (opt && opt->ip6po_nextroute.ro_rt != NULL) { 761 /* 762 * The nexthop is explicitly specified by the 763 * application. We assume the next hop is an IPv6 764 * address. 765 */ 766 dst = (struct sockaddr_in6 *)opt->ip6po_nexthop; 767 } else if ((rt->rt_flags & RTF_GATEWAY)) 768 dst = (struct sockaddr_in6 *)rt->rt_gateway; 769 } 770 771 /* 772 * XXXXXX: original code follows: 773 */ 774 if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) 775 m->m_flags &= ~(M_BCAST | M_MCAST); /* just in case */ 776 else { 777 struct in6_multi *in6m; 778 779 m->m_flags = (m->m_flags & ~M_BCAST) | M_MCAST; 780 781 in6_ifstat_inc(ifp, ifs6_out_mcast); 782 783 /* 784 * Confirm that the outgoing interface supports multicast. 785 */ 786 if (!(ifp->if_flags & IFF_MULTICAST)) { 787 ip6stat.ip6s_noroute++; 788 in6_ifstat_inc(ifp, ifs6_out_discard); 789 error = ENETUNREACH; 790 goto bad; 791 } 792 793 IN6_LOOKUP_MULTI(ip6->ip6_dst, ifp, in6m); 794 if (in6m != NULL && 795 (im6o == NULL || im6o->im6o_multicast_loop)) { 796 /* 797 * If we belong to the destination multicast group 798 * on the outgoing interface, and the caller did not 799 * forbid loopback, loop back a copy. 800 */ 801 ip6_mloopback(ifp, m, dst); 802 } else { 803 /* 804 * If we are acting as a multicast router, perform 805 * multicast forwarding as if the packet had just 806 * arrived on the interface to which we are about 807 * to send. The multicast forwarding function 808 * recursively calls this function, using the 809 * IPV6_FORWARDING flag to prevent infinite recursion. 810 * 811 * Multicasts that are looped back by ip6_mloopback(), 812 * above, will be forwarded by the ip6_input() routine, 813 * if necessary. 814 */ 815 if (ip6_mrouter && (flags & IPV6_FORWARDING) == 0) { 816 if (ip6_mforward(ip6, ifp, m) != 0) { 817 m_freem(m); 818 goto done; 819 } 820 } 821 } 822 /* 823 * Multicasts with a hoplimit of zero may be looped back, 824 * above, but must not be transmitted on a network. 825 * Also, multicasts addressed to the loopback interface 826 * are not sent -- the above call to ip6_mloopback() will 827 * loop back a copy if this host actually belongs to the 828 * destination group on the loopback interface. 829 */ 830 if (ip6->ip6_hlim == 0 || (ifp->if_flags & IFF_LOOPBACK) || 831 IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst)) { 832 m_freem(m); 833 goto done; 834 } 835 } 836 837 /* 838 * Fill the outgoing inteface to tell the upper layer 839 * to increment per-interface statistics. 840 */ 841 if (ifpp) 842 *ifpp = ifp; 843 844 /* Determine path MTU. */ 845 if ((error = ip6_getpmtu(ro_pmtu, ro, ifp, &finaldst, &mtu, 846 &alwaysfrag)) != 0) 847 goto bad; 848 #ifdef IPSEC 849 if (needipsectun) 850 mtu = IPV6_MMTU; 851 #endif 852 853 /* 854 * The caller of this function may specify to use the minimum MTU 855 * in some cases. 856 * An advanced API option (IPV6_USE_MIN_MTU) can also override MTU 857 * setting. The logic is a bit complicated; by default, unicast 858 * packets will follow path MTU while multicast packets will be sent at 859 * the minimum MTU. If IP6PO_MINMTU_ALL is specified, all packets 860 * including unicast ones will be sent at the minimum MTU. Multicast 861 * packets will always be sent at the minimum MTU unless 862 * IP6PO_MINMTU_DISABLE is explicitly specified. 863 * See RFC 3542 for more details. 864 */ 865 if (mtu > IPV6_MMTU) { 866 if ((flags & IPV6_MINMTU)) 867 mtu = IPV6_MMTU; 868 else if (opt && opt->ip6po_minmtu == IP6PO_MINMTU_ALL) 869 mtu = IPV6_MMTU; 870 else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) && 871 (opt == NULL || 872 opt->ip6po_minmtu != IP6PO_MINMTU_DISABLE)) { 873 mtu = IPV6_MMTU; 874 } 875 } 876 877 /* 878 * clear embedded scope identifiers if necessary. 879 * in6_clearscope will touch the addresses only when necessary. 880 */ 881 in6_clearscope(&ip6->ip6_src); 882 in6_clearscope(&ip6->ip6_dst); 883 884 /* 885 * If the outgoing packet contains a hop-by-hop options header, 886 * it must be examined and processed even by the source node. 887 * (RFC 2460, section 4.) 888 */ 889 if (exthdrs.ip6e_hbh) { 890 struct ip6_hbh *hbh = mtod(exthdrs.ip6e_hbh, struct ip6_hbh *); 891 u_int32_t dummy1; /* XXX unused */ 892 u_int32_t dummy2; /* XXX unused */ 893 894 /* 895 * XXX: if we have to send an ICMPv6 error to the sender, 896 * we need the M_LOOP flag since icmp6_error() expects 897 * the IPv6 and the hop-by-hop options header are 898 * continuous unless the flag is set. 899 */ 900 m->m_flags |= M_LOOP; 901 m->m_pkthdr.rcvif = ifp; 902 if (ip6_process_hopopts(m, (u_int8_t *)(hbh + 1), 903 ((hbh->ip6h_len + 1) << 3) - sizeof(struct ip6_hbh), 904 &dummy1, &dummy2) < 0) { 905 /* m was already freed at this point */ 906 error = EINVAL;/* better error? */ 907 goto done; 908 } 909 m->m_flags &= ~M_LOOP; /* XXX */ 910 m->m_pkthdr.rcvif = NULL; 911 } 912 913 #ifdef PFIL_HOOKS 914 /* 915 * Run through list of hooks for output packets. 916 */ 917 if ((error = pfil_run_hooks(&inet6_pfil_hook, &m, ifp, PFIL_OUT)) != 0) 918 goto done; 919 if (m == NULL) 920 goto done; 921 ip6 = mtod(m, struct ip6_hdr *); 922 #endif /* PFIL_HOOKS */ 923 /* 924 * Send the packet to the outgoing interface. 925 * If necessary, do IPv6 fragmentation before sending. 926 * 927 * the logic here is rather complex: 928 * 1: normal case (dontfrag == 0, alwaysfrag == 0) 929 * 1-a: send as is if tlen <= path mtu 930 * 1-b: fragment if tlen > path mtu 931 * 932 * 2: if user asks us not to fragment (dontfrag == 1) 933 * 2-a: send as is if tlen <= interface mtu 934 * 2-b: error if tlen > interface mtu 935 * 936 * 3: if we always need to attach fragment header (alwaysfrag == 1) 937 * always fragment 938 * 939 * 4: if dontfrag == 1 && alwaysfrag == 1 940 * error, as we cannot handle this conflicting request 941 */ 942 tlen = m->m_pkthdr.len; 943 tso = (m->m_pkthdr.csum_flags & M_CSUM_TSOv6) != 0; 944 if (opt && (opt->ip6po_flags & IP6PO_DONTFRAG)) 945 dontfrag = 1; 946 else 947 dontfrag = 0; 948 949 if (dontfrag && alwaysfrag) { /* case 4 */ 950 /* conflicting request - can't transmit */ 951 error = EMSGSIZE; 952 goto bad; 953 } 954 if (dontfrag && (!tso && tlen > IN6_LINKMTU(ifp))) { /* case 2-b */ 955 /* 956 * Even if the DONTFRAG option is specified, we cannot send the 957 * packet when the data length is larger than the MTU of the 958 * outgoing interface. 959 * Notify the error by sending IPV6_PATHMTU ancillary data as 960 * well as returning an error code (the latter is not described 961 * in the API spec.) 962 */ 963 u_int32_t mtu32; 964 struct ip6ctlparam ip6cp; 965 966 mtu32 = (u_int32_t)mtu; 967 bzero(&ip6cp, sizeof(ip6cp)); 968 ip6cp.ip6c_cmdarg = (void *)&mtu32; 969 pfctlinput2(PRC_MSGSIZE, 970 rtcache_getdst((struct route *)ro_pmtu), &ip6cp); 971 972 error = EMSGSIZE; 973 goto bad; 974 } 975 976 /* 977 * transmit packet without fragmentation 978 */ 979 if (dontfrag || (!alwaysfrag && (tlen <= mtu || tso))) { 980 /* case 1-a and 2-a */ 981 struct in6_ifaddr *ia6; 982 int sw_csum; 983 984 ip6 = mtod(m, struct ip6_hdr *); 985 ia6 = in6_ifawithifp(ifp, &ip6->ip6_src); 986 if (ia6) { 987 /* Record statistics for this interface address. */ 988 ia6->ia_ifa.ifa_data.ifad_outbytes += m->m_pkthdr.len; 989 } 990 #ifdef IPSEC 991 /* clean ipsec history once it goes out of the node */ 992 ipsec_delaux(m); 993 #endif 994 995 sw_csum = m->m_pkthdr.csum_flags & ~ifp->if_csum_flags_tx; 996 if ((sw_csum & (M_CSUM_UDPv6|M_CSUM_TCPv6)) != 0) { 997 if (IN6_NEED_CHECKSUM(ifp, 998 sw_csum & (M_CSUM_UDPv6|M_CSUM_TCPv6))) { 999 in6_delayed_cksum(m); 1000 } 1001 m->m_pkthdr.csum_flags &= ~(M_CSUM_UDPv6|M_CSUM_TCPv6); 1002 } 1003 1004 if (__predict_true(!tso || 1005 (ifp->if_capenable & IFCAP_TSOv6) != 0)) { 1006 error = nd6_output(ifp, origifp, m, dst, rt); 1007 } else { 1008 error = ip6_tso_output(ifp, origifp, m, dst, rt); 1009 } 1010 goto done; 1011 } 1012 1013 if (tso) { 1014 error = EINVAL; /* XXX */ 1015 goto bad; 1016 } 1017 1018 /* 1019 * try to fragment the packet. case 1-b and 3 1020 */ 1021 if (mtu < IPV6_MMTU) { 1022 /* path MTU cannot be less than IPV6_MMTU */ 1023 error = EMSGSIZE; 1024 in6_ifstat_inc(ifp, ifs6_out_fragfail); 1025 goto bad; 1026 } else if (ip6->ip6_plen == 0) { 1027 /* jumbo payload cannot be fragmented */ 1028 error = EMSGSIZE; 1029 in6_ifstat_inc(ifp, ifs6_out_fragfail); 1030 goto bad; 1031 } else { 1032 struct mbuf **mnext, *m_frgpart; 1033 struct ip6_frag *ip6f; 1034 u_int32_t id = htonl(ip6_randomid()); 1035 u_char nextproto; 1036 #if 0 /* see below */ 1037 struct ip6ctlparam ip6cp; 1038 u_int32_t mtu32; 1039 #endif 1040 1041 /* 1042 * Too large for the destination or interface; 1043 * fragment if possible. 1044 * Must be able to put at least 8 bytes per fragment. 1045 */ 1046 hlen = unfragpartlen; 1047 if (mtu > IPV6_MAXPACKET) 1048 mtu = IPV6_MAXPACKET; 1049 1050 #if 0 1051 /* 1052 * It is believed this code is a leftover from the 1053 * development of the IPV6_RECVPATHMTU sockopt and 1054 * associated work to implement RFC3542. 1055 * It's not entirely clear what the intent of the API 1056 * is at this point, so disable this code for now. 1057 * The IPV6_RECVPATHMTU sockopt and/or IPV6_DONTFRAG 1058 * will send notifications if the application requests. 1059 */ 1060 1061 /* Notify a proper path MTU to applications. */ 1062 mtu32 = (u_int32_t)mtu; 1063 bzero(&ip6cp, sizeof(ip6cp)); 1064 ip6cp.ip6c_cmdarg = (void *)&mtu32; 1065 pfctlinput2(PRC_MSGSIZE, 1066 rtcache_getdst((struct route *)ro_pmtu), &ip6cp); 1067 #endif 1068 1069 len = (mtu - hlen - sizeof(struct ip6_frag)) & ~7; 1070 if (len < 8) { 1071 error = EMSGSIZE; 1072 in6_ifstat_inc(ifp, ifs6_out_fragfail); 1073 goto bad; 1074 } 1075 1076 mnext = &m->m_nextpkt; 1077 1078 /* 1079 * Change the next header field of the last header in the 1080 * unfragmentable part. 1081 */ 1082 if (exthdrs.ip6e_rthdr) { 1083 nextproto = *mtod(exthdrs.ip6e_rthdr, u_char *); 1084 *mtod(exthdrs.ip6e_rthdr, u_char *) = IPPROTO_FRAGMENT; 1085 } else if (exthdrs.ip6e_dest1) { 1086 nextproto = *mtod(exthdrs.ip6e_dest1, u_char *); 1087 *mtod(exthdrs.ip6e_dest1, u_char *) = IPPROTO_FRAGMENT; 1088 } else if (exthdrs.ip6e_hbh) { 1089 nextproto = *mtod(exthdrs.ip6e_hbh, u_char *); 1090 *mtod(exthdrs.ip6e_hbh, u_char *) = IPPROTO_FRAGMENT; 1091 } else { 1092 nextproto = ip6->ip6_nxt; 1093 ip6->ip6_nxt = IPPROTO_FRAGMENT; 1094 } 1095 1096 if ((m->m_pkthdr.csum_flags & (M_CSUM_UDPv6|M_CSUM_TCPv6)) 1097 != 0) { 1098 if (IN6_NEED_CHECKSUM(ifp, 1099 m->m_pkthdr.csum_flags & 1100 (M_CSUM_UDPv6|M_CSUM_TCPv6))) { 1101 in6_delayed_cksum(m); 1102 } 1103 m->m_pkthdr.csum_flags &= ~(M_CSUM_UDPv6|M_CSUM_TCPv6); 1104 } 1105 1106 /* 1107 * Loop through length of segment after first fragment, 1108 * make new header and copy data of each part and link onto 1109 * chain. 1110 */ 1111 m0 = m; 1112 for (off = hlen; off < tlen; off += len) { 1113 struct mbuf *mlast; 1114 1115 MGETHDR(m, M_DONTWAIT, MT_HEADER); 1116 if (!m) { 1117 error = ENOBUFS; 1118 ip6stat.ip6s_odropped++; 1119 goto sendorfree; 1120 } 1121 m->m_pkthdr.rcvif = NULL; 1122 m->m_flags = m0->m_flags & M_COPYFLAGS; 1123 *mnext = m; 1124 mnext = &m->m_nextpkt; 1125 m->m_data += max_linkhdr; 1126 mhip6 = mtod(m, struct ip6_hdr *); 1127 *mhip6 = *ip6; 1128 m->m_len = sizeof(*mhip6); 1129 error = ip6_insertfraghdr(m0, m, hlen, &ip6f); 1130 if (error) { 1131 ip6stat.ip6s_odropped++; 1132 goto sendorfree; 1133 } 1134 ip6f->ip6f_offlg = htons((u_int16_t)((off - hlen) & ~7)); 1135 if (off + len >= tlen) 1136 len = tlen - off; 1137 else 1138 ip6f->ip6f_offlg |= IP6F_MORE_FRAG; 1139 mhip6->ip6_plen = htons((u_int16_t)(len + hlen + 1140 sizeof(*ip6f) - sizeof(struct ip6_hdr))); 1141 if ((m_frgpart = m_copy(m0, off, len)) == 0) { 1142 error = ENOBUFS; 1143 ip6stat.ip6s_odropped++; 1144 goto sendorfree; 1145 } 1146 for (mlast = m; mlast->m_next; mlast = mlast->m_next) 1147 ; 1148 mlast->m_next = m_frgpart; 1149 m->m_pkthdr.len = len + hlen + sizeof(*ip6f); 1150 m->m_pkthdr.rcvif = (struct ifnet *)0; 1151 ip6f->ip6f_reserved = 0; 1152 ip6f->ip6f_ident = id; 1153 ip6f->ip6f_nxt = nextproto; 1154 ip6stat.ip6s_ofragments++; 1155 in6_ifstat_inc(ifp, ifs6_out_fragcreat); 1156 } 1157 1158 in6_ifstat_inc(ifp, ifs6_out_fragok); 1159 } 1160 1161 /* 1162 * Remove leading garbages. 1163 */ 1164 sendorfree: 1165 m = m0->m_nextpkt; 1166 m0->m_nextpkt = 0; 1167 m_freem(m0); 1168 for (m0 = m; m; m = m0) { 1169 m0 = m->m_nextpkt; 1170 m->m_nextpkt = 0; 1171 if (error == 0) { 1172 struct in6_ifaddr *ia6; 1173 ip6 = mtod(m, struct ip6_hdr *); 1174 ia6 = in6_ifawithifp(ifp, &ip6->ip6_src); 1175 if (ia6) { 1176 /* 1177 * Record statistics for this interface 1178 * address. 1179 */ 1180 ia6->ia_ifa.ifa_data.ifad_outbytes += 1181 m->m_pkthdr.len; 1182 } 1183 #ifdef IPSEC 1184 /* clean ipsec history once it goes out of the node */ 1185 ipsec_delaux(m); 1186 #endif 1187 error = nd6_output(ifp, origifp, m, dst, rt); 1188 } else 1189 m_freem(m); 1190 } 1191 1192 if (error == 0) 1193 ip6stat.ip6s_fragmented++; 1194 1195 done: 1196 /* XXX Second if is invariant? */ 1197 if (ro == &ip6route) 1198 rtcache_free((struct route *)ro); 1199 else if (ro_pmtu == &ip6route) 1200 rtcache_free((struct route *)ro_pmtu); 1201 1202 #ifdef IPSEC 1203 if (sp != NULL) 1204 key_freesp(sp); 1205 #endif /* IPSEC */ 1206 #ifdef FAST_IPSEC 1207 if (sp != NULL) 1208 KEY_FREESP(&sp); 1209 #endif /* FAST_IPSEC */ 1210 1211 1212 return (error); 1213 1214 freehdrs: 1215 m_freem(exthdrs.ip6e_hbh); /* m_freem will check if mbuf is 0 */ 1216 m_freem(exthdrs.ip6e_dest1); 1217 m_freem(exthdrs.ip6e_rthdr); 1218 m_freem(exthdrs.ip6e_dest2); 1219 /* FALLTHROUGH */ 1220 bad: 1221 m_freem(m); 1222 goto done; 1223 } 1224 1225 static int 1226 ip6_copyexthdr(mp, hdr, hlen) 1227 struct mbuf **mp; 1228 void *hdr; 1229 int hlen; 1230 { 1231 struct mbuf *m; 1232 1233 if (hlen > MCLBYTES) 1234 return (ENOBUFS); /* XXX */ 1235 1236 MGET(m, M_DONTWAIT, MT_DATA); 1237 if (!m) 1238 return (ENOBUFS); 1239 1240 if (hlen > MLEN) { 1241 MCLGET(m, M_DONTWAIT); 1242 if ((m->m_flags & M_EXT) == 0) { 1243 m_free(m); 1244 return (ENOBUFS); 1245 } 1246 } 1247 m->m_len = hlen; 1248 if (hdr) 1249 bcopy(hdr, mtod(m, void *), hlen); 1250 1251 *mp = m; 1252 return (0); 1253 } 1254 1255 /* 1256 * Process a delayed payload checksum calculation. 1257 */ 1258 void 1259 in6_delayed_cksum(struct mbuf *m) 1260 { 1261 uint16_t csum, offset; 1262 1263 KASSERT((m->m_pkthdr.csum_flags & (M_CSUM_UDPv6|M_CSUM_TCPv6)) != 0); 1264 KASSERT((~m->m_pkthdr.csum_flags & (M_CSUM_UDPv6|M_CSUM_TCPv6)) != 0); 1265 KASSERT((m->m_pkthdr.csum_flags 1266 & (M_CSUM_UDPv4|M_CSUM_TCPv4|M_CSUM_TSOv4)) == 0); 1267 1268 offset = M_CSUM_DATA_IPv6_HL(m->m_pkthdr.csum_data); 1269 csum = in6_cksum(m, 0, offset, m->m_pkthdr.len - offset); 1270 if (csum == 0 && (m->m_pkthdr.csum_flags & M_CSUM_UDPv6) != 0) { 1271 csum = 0xffff; 1272 } 1273 1274 offset += M_CSUM_DATA_IPv6_OFFSET(m->m_pkthdr.csum_data); 1275 if ((offset + sizeof(csum)) > m->m_len) { 1276 m_copyback(m, offset, sizeof(csum), &csum); 1277 } else { 1278 *(uint16_t *)(mtod(m, char *) + offset) = csum; 1279 } 1280 } 1281 1282 /* 1283 * Insert jumbo payload option. 1284 */ 1285 static int 1286 ip6_insert_jumboopt(exthdrs, plen) 1287 struct ip6_exthdrs *exthdrs; 1288 u_int32_t plen; 1289 { 1290 struct mbuf *mopt; 1291 u_int8_t *optbuf; 1292 u_int32_t v; 1293 1294 #define JUMBOOPTLEN 8 /* length of jumbo payload option and padding */ 1295 1296 /* 1297 * If there is no hop-by-hop options header, allocate new one. 1298 * If there is one but it doesn't have enough space to store the 1299 * jumbo payload option, allocate a cluster to store the whole options. 1300 * Otherwise, use it to store the options. 1301 */ 1302 if (exthdrs->ip6e_hbh == 0) { 1303 MGET(mopt, M_DONTWAIT, MT_DATA); 1304 if (mopt == 0) 1305 return (ENOBUFS); 1306 mopt->m_len = JUMBOOPTLEN; 1307 optbuf = mtod(mopt, u_int8_t *); 1308 optbuf[1] = 0; /* = ((JUMBOOPTLEN) >> 3) - 1 */ 1309 exthdrs->ip6e_hbh = mopt; 1310 } else { 1311 struct ip6_hbh *hbh; 1312 1313 mopt = exthdrs->ip6e_hbh; 1314 if (M_TRAILINGSPACE(mopt) < JUMBOOPTLEN) { 1315 /* 1316 * XXX assumption: 1317 * - exthdrs->ip6e_hbh is not referenced from places 1318 * other than exthdrs. 1319 * - exthdrs->ip6e_hbh is not an mbuf chain. 1320 */ 1321 int oldoptlen = mopt->m_len; 1322 struct mbuf *n; 1323 1324 /* 1325 * XXX: give up if the whole (new) hbh header does 1326 * not fit even in an mbuf cluster. 1327 */ 1328 if (oldoptlen + JUMBOOPTLEN > MCLBYTES) 1329 return (ENOBUFS); 1330 1331 /* 1332 * As a consequence, we must always prepare a cluster 1333 * at this point. 1334 */ 1335 MGET(n, M_DONTWAIT, MT_DATA); 1336 if (n) { 1337 MCLGET(n, M_DONTWAIT); 1338 if ((n->m_flags & M_EXT) == 0) { 1339 m_freem(n); 1340 n = NULL; 1341 } 1342 } 1343 if (!n) 1344 return (ENOBUFS); 1345 n->m_len = oldoptlen + JUMBOOPTLEN; 1346 bcopy(mtod(mopt, void *), mtod(n, void *), 1347 oldoptlen); 1348 optbuf = mtod(n, u_int8_t *) + oldoptlen; 1349 m_freem(mopt); 1350 mopt = exthdrs->ip6e_hbh = n; 1351 } else { 1352 optbuf = mtod(mopt, u_int8_t *) + mopt->m_len; 1353 mopt->m_len += JUMBOOPTLEN; 1354 } 1355 optbuf[0] = IP6OPT_PADN; 1356 optbuf[1] = 0; 1357 1358 /* 1359 * Adjust the header length according to the pad and 1360 * the jumbo payload option. 1361 */ 1362 hbh = mtod(mopt, struct ip6_hbh *); 1363 hbh->ip6h_len += (JUMBOOPTLEN >> 3); 1364 } 1365 1366 /* fill in the option. */ 1367 optbuf[2] = IP6OPT_JUMBO; 1368 optbuf[3] = 4; 1369 v = (u_int32_t)htonl(plen + JUMBOOPTLEN); 1370 bcopy(&v, &optbuf[4], sizeof(u_int32_t)); 1371 1372 /* finally, adjust the packet header length */ 1373 exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN; 1374 1375 return (0); 1376 #undef JUMBOOPTLEN 1377 } 1378 1379 /* 1380 * Insert fragment header and copy unfragmentable header portions. 1381 */ 1382 static int 1383 ip6_insertfraghdr(m0, m, hlen, frghdrp) 1384 struct mbuf *m0, *m; 1385 int hlen; 1386 struct ip6_frag **frghdrp; 1387 { 1388 struct mbuf *n, *mlast; 1389 1390 if (hlen > sizeof(struct ip6_hdr)) { 1391 n = m_copym(m0, sizeof(struct ip6_hdr), 1392 hlen - sizeof(struct ip6_hdr), M_DONTWAIT); 1393 if (n == 0) 1394 return (ENOBUFS); 1395 m->m_next = n; 1396 } else 1397 n = m; 1398 1399 /* Search for the last mbuf of unfragmentable part. */ 1400 for (mlast = n; mlast->m_next; mlast = mlast->m_next) 1401 ; 1402 1403 if ((mlast->m_flags & M_EXT) == 0 && 1404 M_TRAILINGSPACE(mlast) >= sizeof(struct ip6_frag)) { 1405 /* use the trailing space of the last mbuf for the fragment hdr */ 1406 *frghdrp = (struct ip6_frag *)(mtod(mlast, char *) + 1407 mlast->m_len); 1408 mlast->m_len += sizeof(struct ip6_frag); 1409 m->m_pkthdr.len += sizeof(struct ip6_frag); 1410 } else { 1411 /* allocate a new mbuf for the fragment header */ 1412 struct mbuf *mfrg; 1413 1414 MGET(mfrg, M_DONTWAIT, MT_DATA); 1415 if (mfrg == 0) 1416 return (ENOBUFS); 1417 mfrg->m_len = sizeof(struct ip6_frag); 1418 *frghdrp = mtod(mfrg, struct ip6_frag *); 1419 mlast->m_next = mfrg; 1420 } 1421 1422 return (0); 1423 } 1424 1425 static int 1426 ip6_getpmtu(struct route_in6 *ro_pmtu, struct route_in6 *ro, struct ifnet *ifp, 1427 const struct in6_addr *dst, u_long *mtup, int *alwaysfragp) 1428 { 1429 u_int32_t mtu = 0; 1430 int alwaysfrag = 0; 1431 int error = 0; 1432 const struct sockaddr_in6 *cdst; 1433 1434 if (ro_pmtu != ro) { 1435 /* The first hop and the final destination may differ. */ 1436 cdst = (const struct sockaddr_in6 *)rtcache_getdst((struct route *)ro_pmtu); 1437 if (!IN6_ARE_ADDR_EQUAL(&cdst->sin6_addr, dst)) 1438 rtcache_free((struct route *)ro_pmtu); 1439 else 1440 rtcache_check((struct route *)ro_pmtu); 1441 if (ro_pmtu->ro_rt == NULL) { 1442 struct sockaddr_in6 *sa6_dst = &ro_pmtu->ro_dst; 1443 memset(sa6_dst, 0, sizeof(*sa6_dst)); /* for safety */ 1444 sa6_dst->sin6_family = AF_INET6; 1445 sa6_dst->sin6_len = sizeof(struct sockaddr_in6); 1446 sa6_dst->sin6_addr = *dst; 1447 rtcache_init((struct route *)ro_pmtu); 1448 } 1449 } 1450 if (ro_pmtu->ro_rt != NULL) { 1451 u_int32_t ifmtu; 1452 1453 if (ifp == NULL) 1454 ifp = ro_pmtu->ro_rt->rt_ifp; 1455 ifmtu = IN6_LINKMTU(ifp); 1456 mtu = ro_pmtu->ro_rt->rt_rmx.rmx_mtu; 1457 if (mtu == 0) 1458 mtu = ifmtu; 1459 else if (mtu < IPV6_MMTU) { 1460 /* 1461 * RFC2460 section 5, last paragraph: 1462 * if we record ICMPv6 too big message with 1463 * mtu < IPV6_MMTU, transmit packets sized IPV6_MMTU 1464 * or smaller, with fragment header attached. 1465 * (fragment header is needed regardless from the 1466 * packet size, for translators to identify packets) 1467 */ 1468 alwaysfrag = 1; 1469 mtu = IPV6_MMTU; 1470 } else if (mtu > ifmtu) { 1471 /* 1472 * The MTU on the route is larger than the MTU on 1473 * the interface! This shouldn't happen, unless the 1474 * MTU of the interface has been changed after the 1475 * interface was brought up. Change the MTU in the 1476 * route to match the interface MTU (as long as the 1477 * field isn't locked). 1478 */ 1479 mtu = ifmtu; 1480 if (!(ro_pmtu->ro_rt->rt_rmx.rmx_locks & RTV_MTU)) 1481 ro_pmtu->ro_rt->rt_rmx.rmx_mtu = mtu; 1482 } 1483 } else if (ifp) { 1484 mtu = IN6_LINKMTU(ifp); 1485 } else 1486 error = EHOSTUNREACH; /* XXX */ 1487 1488 *mtup = mtu; 1489 if (alwaysfragp) 1490 *alwaysfragp = alwaysfrag; 1491 return (error); 1492 } 1493 1494 /* 1495 * IP6 socket option processing. 1496 */ 1497 int 1498 ip6_ctloutput(int op, struct socket *so, int level, int optname, 1499 struct mbuf **mp) 1500 { 1501 int privileged, optdatalen, uproto; 1502 void *optdata; 1503 struct in6pcb *in6p = sotoin6pcb(so); 1504 struct mbuf *m = *mp; 1505 int error, optval; 1506 int optlen; 1507 struct lwp *l = curlwp; /* XXX */ 1508 1509 optlen = m ? m->m_len : 0; 1510 error = optval = 0; 1511 privileged = (l == 0 || kauth_authorize_generic(l->l_cred, 1512 KAUTH_GENERIC_ISSUSER, NULL)) ? 0 : 1; 1513 uproto = (int)so->so_proto->pr_protocol; 1514 1515 if (level == IPPROTO_IPV6) { 1516 switch (op) { 1517 case PRCO_SETOPT: 1518 switch (optname) { 1519 #ifdef RFC2292 1520 case IPV6_2292PKTOPTIONS: 1521 /* m is freed in ip6_pcbopts */ 1522 error = ip6_pcbopts(&in6p->in6p_outputopts, 1523 m, so); 1524 break; 1525 #endif 1526 1527 /* 1528 * Use of some Hop-by-Hop options or some 1529 * Destination options, might require special 1530 * privilege. That is, normal applications 1531 * (without special privilege) might be forbidden 1532 * from setting certain options in outgoing packets, 1533 * and might never see certain options in received 1534 * packets. [RFC 2292 Section 6] 1535 * KAME specific note: 1536 * KAME prevents non-privileged users from sending or 1537 * receiving ANY hbh/dst options in order to avoid 1538 * overhead of parsing options in the kernel. 1539 */ 1540 case IPV6_RECVHOPOPTS: 1541 case IPV6_RECVDSTOPTS: 1542 case IPV6_RECVRTHDRDSTOPTS: 1543 if (!privileged) { 1544 error = EPERM; 1545 break; 1546 } 1547 /* FALLTHROUGH */ 1548 case IPV6_UNICAST_HOPS: 1549 case IPV6_HOPLIMIT: 1550 case IPV6_FAITH: 1551 1552 case IPV6_RECVPKTINFO: 1553 case IPV6_RECVHOPLIMIT: 1554 case IPV6_RECVRTHDR: 1555 case IPV6_RECVPATHMTU: 1556 case IPV6_RECVTCLASS: 1557 case IPV6_V6ONLY: 1558 if (optlen != sizeof(int)) { 1559 error = EINVAL; 1560 break; 1561 } 1562 optval = *mtod(m, int *); 1563 switch (optname) { 1564 1565 case IPV6_UNICAST_HOPS: 1566 if (optval < -1 || optval >= 256) 1567 error = EINVAL; 1568 else { 1569 /* -1 = kernel default */ 1570 in6p->in6p_hops = optval; 1571 } 1572 break; 1573 #define OPTSET(bit) \ 1574 do { \ 1575 if (optval) \ 1576 in6p->in6p_flags |= (bit); \ 1577 else \ 1578 in6p->in6p_flags &= ~(bit); \ 1579 } while (/*CONSTCOND*/ 0) 1580 1581 #ifdef RFC2292 1582 #define OPTSET2292(bit) \ 1583 do { \ 1584 in6p->in6p_flags |= IN6P_RFC2292; \ 1585 if (optval) \ 1586 in6p->in6p_flags |= (bit); \ 1587 else \ 1588 in6p->in6p_flags &= ~(bit); \ 1589 } while (/*CONSTCOND*/ 0) 1590 #endif 1591 1592 #define OPTBIT(bit) (in6p->in6p_flags & (bit) ? 1 : 0) 1593 1594 case IPV6_RECVPKTINFO: 1595 #ifdef RFC2292 1596 /* cannot mix with RFC2292 */ 1597 if (OPTBIT(IN6P_RFC2292)) { 1598 error = EINVAL; 1599 break; 1600 } 1601 #endif 1602 OPTSET(IN6P_PKTINFO); 1603 break; 1604 1605 case IPV6_HOPLIMIT: 1606 { 1607 struct ip6_pktopts **optp; 1608 1609 #ifdef RFC2292 1610 /* cannot mix with RFC2292 */ 1611 if (OPTBIT(IN6P_RFC2292)) { 1612 error = EINVAL; 1613 break; 1614 } 1615 #endif 1616 optp = &in6p->in6p_outputopts; 1617 error = ip6_pcbopt(IPV6_HOPLIMIT, 1618 (u_char *)&optval, 1619 sizeof(optval), 1620 optp, 1621 privileged, uproto); 1622 break; 1623 } 1624 1625 case IPV6_RECVHOPLIMIT: 1626 #ifdef RFC2292 1627 /* cannot mix with RFC2292 */ 1628 if (OPTBIT(IN6P_RFC2292)) { 1629 error = EINVAL; 1630 break; 1631 } 1632 #endif 1633 OPTSET(IN6P_HOPLIMIT); 1634 break; 1635 1636 case IPV6_RECVHOPOPTS: 1637 #ifdef RFC2292 1638 /* cannot mix with RFC2292 */ 1639 if (OPTBIT(IN6P_RFC2292)) { 1640 error = EINVAL; 1641 break; 1642 } 1643 #endif 1644 OPTSET(IN6P_HOPOPTS); 1645 break; 1646 1647 case IPV6_RECVDSTOPTS: 1648 #ifdef RFC2292 1649 /* cannot mix with RFC2292 */ 1650 if (OPTBIT(IN6P_RFC2292)) { 1651 error = EINVAL; 1652 break; 1653 } 1654 #endif 1655 OPTSET(IN6P_DSTOPTS); 1656 break; 1657 1658 case IPV6_RECVRTHDRDSTOPTS: 1659 #ifdef RFC2292 1660 /* cannot mix with RFC2292 */ 1661 if (OPTBIT(IN6P_RFC2292)) { 1662 error = EINVAL; 1663 break; 1664 } 1665 #endif 1666 OPTSET(IN6P_RTHDRDSTOPTS); 1667 break; 1668 1669 case IPV6_RECVRTHDR: 1670 #ifdef RFC2292 1671 /* cannot mix with RFC2292 */ 1672 if (OPTBIT(IN6P_RFC2292)) { 1673 error = EINVAL; 1674 break; 1675 } 1676 #endif 1677 OPTSET(IN6P_RTHDR); 1678 break; 1679 1680 case IPV6_FAITH: 1681 OPTSET(IN6P_FAITH); 1682 break; 1683 1684 case IPV6_RECVPATHMTU: 1685 /* 1686 * We ignore this option for TCP 1687 * sockets. 1688 * (RFC3542 leaves this case 1689 * unspecified.) 1690 */ 1691 if (uproto != IPPROTO_TCP) 1692 OPTSET(IN6P_MTU); 1693 break; 1694 1695 case IPV6_V6ONLY: 1696 /* 1697 * make setsockopt(IPV6_V6ONLY) 1698 * available only prior to bind(2). 1699 * see ipng mailing list, Jun 22 2001. 1700 */ 1701 if (in6p->in6p_lport || 1702 !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr)) { 1703 error = EINVAL; 1704 break; 1705 } 1706 #ifdef INET6_BINDV6ONLY 1707 if (!optval) 1708 error = EINVAL; 1709 #else 1710 OPTSET(IN6P_IPV6_V6ONLY); 1711 #endif 1712 break; 1713 case IPV6_RECVTCLASS: 1714 #ifdef RFC2292 1715 /* cannot mix with RFC2292 XXX */ 1716 if (OPTBIT(IN6P_RFC2292)) { 1717 error = EINVAL; 1718 break; 1719 } 1720 #endif 1721 OPTSET(IN6P_TCLASS); 1722 break; 1723 1724 } 1725 break; 1726 1727 case IPV6_OTCLASS: 1728 { 1729 struct ip6_pktopts **optp; 1730 u_int8_t tclass; 1731 1732 if (optlen != sizeof(tclass)) { 1733 error = EINVAL; 1734 break; 1735 } 1736 tclass = *mtod(m, u_int8_t *); 1737 optp = &in6p->in6p_outputopts; 1738 error = ip6_pcbopt(optname, 1739 (u_char *)&tclass, 1740 sizeof(tclass), 1741 optp, 1742 privileged, uproto); 1743 break; 1744 } 1745 1746 case IPV6_TCLASS: 1747 case IPV6_DONTFRAG: 1748 case IPV6_USE_MIN_MTU: 1749 if (optlen != sizeof(optval)) { 1750 error = EINVAL; 1751 break; 1752 } 1753 optval = *mtod(m, int *); 1754 { 1755 struct ip6_pktopts **optp; 1756 optp = &in6p->in6p_outputopts; 1757 error = ip6_pcbopt(optname, 1758 (u_char *)&optval, 1759 sizeof(optval), 1760 optp, 1761 privileged, uproto); 1762 break; 1763 } 1764 1765 #ifdef RFC2292 1766 case IPV6_2292PKTINFO: 1767 case IPV6_2292HOPLIMIT: 1768 case IPV6_2292HOPOPTS: 1769 case IPV6_2292DSTOPTS: 1770 case IPV6_2292RTHDR: 1771 /* RFC 2292 */ 1772 if (optlen != sizeof(int)) { 1773 error = EINVAL; 1774 break; 1775 } 1776 optval = *mtod(m, int *); 1777 switch (optname) { 1778 case IPV6_2292PKTINFO: 1779 OPTSET2292(IN6P_PKTINFO); 1780 break; 1781 case IPV6_2292HOPLIMIT: 1782 OPTSET2292(IN6P_HOPLIMIT); 1783 break; 1784 case IPV6_2292HOPOPTS: 1785 /* 1786 * Check super-user privilege. 1787 * See comments for IPV6_RECVHOPOPTS. 1788 */ 1789 if (!privileged) 1790 return (EPERM); 1791 OPTSET2292(IN6P_HOPOPTS); 1792 break; 1793 case IPV6_2292DSTOPTS: 1794 if (!privileged) 1795 return (EPERM); 1796 OPTSET2292(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); /* XXX */ 1797 break; 1798 case IPV6_2292RTHDR: 1799 OPTSET2292(IN6P_RTHDR); 1800 break; 1801 } 1802 break; 1803 #endif 1804 case IPV6_PKTINFO: 1805 case IPV6_HOPOPTS: 1806 case IPV6_RTHDR: 1807 case IPV6_DSTOPTS: 1808 case IPV6_RTHDRDSTOPTS: 1809 case IPV6_NEXTHOP: 1810 { 1811 /* new advanced API (RFC3542) */ 1812 u_char *optbuf; 1813 int optbuflen; 1814 struct ip6_pktopts **optp; 1815 if (!m) { 1816 error = EINVAL; 1817 break; 1818 } 1819 1820 #ifdef RFC2292 1821 /* cannot mix with RFC2292 */ 1822 if (OPTBIT(IN6P_RFC2292)) { 1823 error = EINVAL; 1824 break; 1825 } 1826 #endif 1827 1828 if (m && m->m_next) { 1829 error = EINVAL; /* XXX */ 1830 break; 1831 } 1832 1833 optbuf = mtod(m, u_char *); 1834 optbuflen = m->m_len; 1835 optp = &in6p->in6p_outputopts; 1836 error = ip6_pcbopt(optname, optbuf, optbuflen, 1837 optp, privileged, uproto); 1838 break; 1839 } 1840 #undef OPTSET 1841 1842 case IPV6_MULTICAST_IF: 1843 case IPV6_MULTICAST_HOPS: 1844 case IPV6_MULTICAST_LOOP: 1845 case IPV6_JOIN_GROUP: 1846 case IPV6_LEAVE_GROUP: 1847 error = ip6_setmoptions(optname, 1848 &in6p->in6p_moptions, m); 1849 break; 1850 1851 case IPV6_PORTRANGE: 1852 if (!m) { 1853 error = EINVAL; 1854 break; 1855 } 1856 optval = *mtod(m, int *); 1857 1858 switch (optval) { 1859 case IPV6_PORTRANGE_DEFAULT: 1860 in6p->in6p_flags &= ~(IN6P_LOWPORT); 1861 in6p->in6p_flags &= ~(IN6P_HIGHPORT); 1862 break; 1863 1864 case IPV6_PORTRANGE_HIGH: 1865 in6p->in6p_flags &= ~(IN6P_LOWPORT); 1866 in6p->in6p_flags |= IN6P_HIGHPORT; 1867 break; 1868 1869 case IPV6_PORTRANGE_LOW: 1870 in6p->in6p_flags &= ~(IN6P_HIGHPORT); 1871 in6p->in6p_flags |= IN6P_LOWPORT; 1872 break; 1873 1874 default: 1875 error = EINVAL; 1876 break; 1877 } 1878 break; 1879 1880 1881 #if defined(IPSEC) || defined(FAST_IPSEC) 1882 case IPV6_IPSEC_POLICY: 1883 { 1884 void *req = NULL; 1885 size_t len = 0; 1886 if (m) { 1887 req = mtod(m, void *); 1888 len = m->m_len; 1889 } 1890 error = ipsec6_set_policy(in6p, optname, req, 1891 len, privileged); 1892 } 1893 break; 1894 #endif /* IPSEC */ 1895 1896 default: 1897 error = ENOPROTOOPT; 1898 break; 1899 } 1900 if (m) 1901 (void)m_free(m); 1902 break; 1903 1904 case PRCO_GETOPT: 1905 switch (optname) { 1906 #ifdef RFC2292 1907 case IPV6_2292PKTOPTIONS: 1908 /* 1909 * RFC3542 (effectively) deprecated the 1910 * semantics of the 2292-style pktoptions. 1911 * Since it was not reliable in nature (i.e., 1912 * applications had to expect the lack of some 1913 * information after all), it would make sense 1914 * to simplify this part by always returning 1915 * empty data. 1916 */ 1917 *mp = m_get(M_WAIT, MT_SOOPTS); 1918 (*mp)->m_len = 0; 1919 break; 1920 #endif 1921 1922 case IPV6_RECVHOPOPTS: 1923 case IPV6_RECVDSTOPTS: 1924 case IPV6_RECVRTHDRDSTOPTS: 1925 case IPV6_UNICAST_HOPS: 1926 case IPV6_RECVPKTINFO: 1927 case IPV6_RECVHOPLIMIT: 1928 case IPV6_RECVRTHDR: 1929 case IPV6_RECVPATHMTU: 1930 1931 case IPV6_FAITH: 1932 case IPV6_V6ONLY: 1933 case IPV6_PORTRANGE: 1934 case IPV6_RECVTCLASS: 1935 switch (optname) { 1936 1937 case IPV6_RECVHOPOPTS: 1938 optval = OPTBIT(IN6P_HOPOPTS); 1939 break; 1940 1941 case IPV6_RECVDSTOPTS: 1942 optval = OPTBIT(IN6P_DSTOPTS); 1943 break; 1944 1945 case IPV6_RECVRTHDRDSTOPTS: 1946 optval = OPTBIT(IN6P_RTHDRDSTOPTS); 1947 break; 1948 1949 case IPV6_UNICAST_HOPS: 1950 optval = in6p->in6p_hops; 1951 break; 1952 1953 case IPV6_RECVPKTINFO: 1954 optval = OPTBIT(IN6P_PKTINFO); 1955 break; 1956 1957 case IPV6_RECVHOPLIMIT: 1958 optval = OPTBIT(IN6P_HOPLIMIT); 1959 break; 1960 1961 case IPV6_RECVRTHDR: 1962 optval = OPTBIT(IN6P_RTHDR); 1963 break; 1964 1965 case IPV6_RECVPATHMTU: 1966 optval = OPTBIT(IN6P_MTU); 1967 break; 1968 1969 case IPV6_FAITH: 1970 optval = OPTBIT(IN6P_FAITH); 1971 break; 1972 1973 case IPV6_V6ONLY: 1974 optval = OPTBIT(IN6P_IPV6_V6ONLY); 1975 break; 1976 1977 case IPV6_PORTRANGE: 1978 { 1979 int flags; 1980 flags = in6p->in6p_flags; 1981 if (flags & IN6P_HIGHPORT) 1982 optval = IPV6_PORTRANGE_HIGH; 1983 else if (flags & IN6P_LOWPORT) 1984 optval = IPV6_PORTRANGE_LOW; 1985 else 1986 optval = 0; 1987 break; 1988 } 1989 case IPV6_RECVTCLASS: 1990 optval = OPTBIT(IN6P_TCLASS); 1991 break; 1992 1993 } 1994 if (error) 1995 break; 1996 *mp = m = m_get(M_WAIT, MT_SOOPTS); 1997 m->m_len = sizeof(int); 1998 *mtod(m, int *) = optval; 1999 break; 2000 2001 case IPV6_PATHMTU: 2002 { 2003 u_long pmtu = 0; 2004 struct ip6_mtuinfo mtuinfo; 2005 struct route_in6 *ro = (struct route_in6 *)&in6p 2006 ->in6p_route; 2007 2008 if (!(so->so_state & SS_ISCONNECTED)) 2009 return (ENOTCONN); 2010 /* 2011 * XXX: we dot not consider the case of source 2012 * routing, or optional information to specify 2013 * the outgoing interface. 2014 */ 2015 error = ip6_getpmtu(ro, NULL, NULL, 2016 &in6p->in6p_faddr, &pmtu, NULL); 2017 if (error) 2018 break; 2019 if (pmtu > IPV6_MAXPACKET) 2020 pmtu = IPV6_MAXPACKET; 2021 2022 memset(&mtuinfo, 0, sizeof(mtuinfo)); 2023 mtuinfo.ip6m_mtu = (u_int32_t)pmtu; 2024 optdata = (void *)&mtuinfo; 2025 optdatalen = sizeof(mtuinfo); 2026 if (optdatalen > MCLBYTES) 2027 return (EMSGSIZE); /* XXX */ 2028 *mp = m = m_get(M_WAIT, MT_SOOPTS); 2029 if (optdatalen > MLEN) 2030 MCLGET(m, M_WAIT); 2031 m->m_len = optdatalen; 2032 memcpy(mtod(m, void *), optdata, optdatalen); 2033 break; 2034 } 2035 2036 #ifdef RFC2292 2037 case IPV6_2292PKTINFO: 2038 case IPV6_2292HOPLIMIT: 2039 case IPV6_2292HOPOPTS: 2040 case IPV6_2292RTHDR: 2041 case IPV6_2292DSTOPTS: 2042 switch (optname) { 2043 case IPV6_2292PKTINFO: 2044 optval = OPTBIT(IN6P_PKTINFO); 2045 break; 2046 case IPV6_2292HOPLIMIT: 2047 optval = OPTBIT(IN6P_HOPLIMIT); 2048 break; 2049 case IPV6_2292HOPOPTS: 2050 optval = OPTBIT(IN6P_HOPOPTS); 2051 break; 2052 case IPV6_2292RTHDR: 2053 optval = OPTBIT(IN6P_RTHDR); 2054 break; 2055 case IPV6_2292DSTOPTS: 2056 optval = OPTBIT(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); 2057 break; 2058 } 2059 *mp = m = m_get(M_WAIT, MT_SOOPTS); 2060 m->m_len = sizeof(int); 2061 *mtod(m, int *) = optval; 2062 break; 2063 #endif 2064 case IPV6_PKTINFO: 2065 case IPV6_HOPOPTS: 2066 case IPV6_RTHDR: 2067 case IPV6_DSTOPTS: 2068 case IPV6_RTHDRDSTOPTS: 2069 case IPV6_NEXTHOP: 2070 case IPV6_OTCLASS: 2071 case IPV6_TCLASS: 2072 case IPV6_DONTFRAG: 2073 case IPV6_USE_MIN_MTU: 2074 error = ip6_getpcbopt(in6p->in6p_outputopts, 2075 optname, mp); 2076 break; 2077 2078 case IPV6_MULTICAST_IF: 2079 case IPV6_MULTICAST_HOPS: 2080 case IPV6_MULTICAST_LOOP: 2081 case IPV6_JOIN_GROUP: 2082 case IPV6_LEAVE_GROUP: 2083 error = ip6_getmoptions(optname, 2084 in6p->in6p_moptions, mp); 2085 break; 2086 2087 #if defined(IPSEC) || defined(FAST_IPSEC) 2088 case IPV6_IPSEC_POLICY: 2089 { 2090 void *req = NULL; 2091 size_t len = 0; 2092 if (m) { 2093 req = mtod(m, void *); 2094 len = m->m_len; 2095 } 2096 error = ipsec6_get_policy(in6p, req, len, mp); 2097 break; 2098 } 2099 #endif /* IPSEC */ 2100 2101 2102 2103 2104 default: 2105 error = ENOPROTOOPT; 2106 break; 2107 } 2108 break; 2109 } 2110 } else { 2111 error = EINVAL; 2112 if (op == PRCO_SETOPT && *mp) 2113 (void)m_free(*mp); 2114 } 2115 return (error); 2116 } 2117 2118 int 2119 ip6_raw_ctloutput(op, so, level, optname, mp) 2120 int op; 2121 struct socket *so; 2122 int level, optname; 2123 struct mbuf **mp; 2124 { 2125 int error = 0, optval, optlen; 2126 const int icmp6off = offsetof(struct icmp6_hdr, icmp6_cksum); 2127 struct in6pcb *in6p = sotoin6pcb(so); 2128 struct mbuf *m = *mp; 2129 2130 optlen = m ? m->m_len : 0; 2131 2132 if (level != IPPROTO_IPV6) { 2133 if (op == PRCO_SETOPT && *mp) 2134 (void)m_free(*mp); 2135 return (EINVAL); 2136 } 2137 2138 switch (optname) { 2139 case IPV6_CHECKSUM: 2140 /* 2141 * For ICMPv6 sockets, no modification allowed for checksum 2142 * offset, permit "no change" values to help existing apps. 2143 * 2144 * XXX RFC3542 says: "An attempt to set IPV6_CHECKSUM 2145 * for an ICMPv6 socket will fail." The current 2146 * behavior does not meet RFC3542. 2147 */ 2148 switch (op) { 2149 case PRCO_SETOPT: 2150 if (optlen != sizeof(int)) { 2151 error = EINVAL; 2152 break; 2153 } 2154 optval = *mtod(m, int *); 2155 if ((optval % 2) != 0) { 2156 /* the API assumes even offset values */ 2157 error = EINVAL; 2158 } else if (so->so_proto->pr_protocol == 2159 IPPROTO_ICMPV6) { 2160 if (optval != icmp6off) 2161 error = EINVAL; 2162 } else 2163 in6p->in6p_cksum = optval; 2164 break; 2165 2166 case PRCO_GETOPT: 2167 if (so->so_proto->pr_protocol == IPPROTO_ICMPV6) 2168 optval = icmp6off; 2169 else 2170 optval = in6p->in6p_cksum; 2171 2172 *mp = m = m_get(M_WAIT, MT_SOOPTS); 2173 m->m_len = sizeof(int); 2174 *mtod(m, int *) = optval; 2175 break; 2176 2177 default: 2178 error = EINVAL; 2179 break; 2180 } 2181 break; 2182 2183 default: 2184 error = ENOPROTOOPT; 2185 break; 2186 } 2187 2188 if (op == PRCO_SETOPT && m) 2189 (void)m_free(m); 2190 2191 return (error); 2192 } 2193 2194 #ifdef RFC2292 2195 /* 2196 * Set up IP6 options in pcb for insertion in output packets or 2197 * specifying behavior of outgoing packets. 2198 */ 2199 static int 2200 ip6_pcbopts(pktopt, m, so) 2201 struct ip6_pktopts **pktopt; 2202 struct mbuf *m; 2203 struct socket *so; 2204 { 2205 struct ip6_pktopts *opt = *pktopt; 2206 int error = 0; 2207 struct lwp *l = curlwp; /* XXX */ 2208 int priv = 0; 2209 2210 /* turn off any old options. */ 2211 if (opt) { 2212 #ifdef DIAGNOSTIC 2213 if (opt->ip6po_pktinfo || opt->ip6po_nexthop || 2214 opt->ip6po_hbh || opt->ip6po_dest1 || opt->ip6po_dest2 || 2215 opt->ip6po_rhinfo.ip6po_rhi_rthdr) 2216 printf("ip6_pcbopts: all specified options are cleared.\n"); 2217 #endif 2218 ip6_clearpktopts(opt, -1); 2219 } else 2220 opt = malloc(sizeof(*opt), M_IP6OPT, M_WAITOK); 2221 *pktopt = NULL; 2222 2223 if (!m || m->m_len == 0) { 2224 /* 2225 * Only turning off any previous options, regardless of 2226 * whether the opt is just created or given. 2227 */ 2228 free(opt, M_IP6OPT); 2229 return (0); 2230 } 2231 2232 /* set options specified by user. */ 2233 if (l && !kauth_authorize_generic(l->l_cred, KAUTH_GENERIC_ISSUSER, 2234 NULL)) 2235 priv = 1; 2236 if ((error = ip6_setpktopts(m, opt, NULL, priv, 2237 so->so_proto->pr_protocol)) != 0) { 2238 ip6_clearpktopts(opt, -1); /* XXX: discard all options */ 2239 free(opt, M_IP6OPT); 2240 return (error); 2241 } 2242 *pktopt = opt; 2243 return (0); 2244 } 2245 #endif 2246 2247 /* 2248 * initialize ip6_pktopts. beware that there are non-zero default values in 2249 * the struct. 2250 */ 2251 void 2252 ip6_initpktopts(struct ip6_pktopts *opt) 2253 { 2254 2255 memset(opt, 0, sizeof(*opt)); 2256 opt->ip6po_hlim = -1; /* -1 means default hop limit */ 2257 opt->ip6po_tclass = -1; /* -1 means default traffic class */ 2258 opt->ip6po_minmtu = IP6PO_MINMTU_MCASTONLY; 2259 } 2260 2261 #define sin6tosa(sin6) ((struct sockaddr *)(sin6)) /* XXX */ 2262 static int 2263 ip6_pcbopt(int optname, u_char *buf, int len, struct ip6_pktopts **pktopt, 2264 int priv, int uproto) 2265 { 2266 struct ip6_pktopts *opt; 2267 2268 if (*pktopt == NULL) { 2269 *pktopt = malloc(sizeof(struct ip6_pktopts), M_IP6OPT, 2270 M_WAITOK); 2271 ip6_initpktopts(*pktopt); 2272 } 2273 opt = *pktopt; 2274 2275 return (ip6_setpktopt(optname, buf, len, opt, priv, 1, 0, uproto)); 2276 } 2277 2278 static int 2279 ip6_getpcbopt(struct ip6_pktopts *pktopt, int optname, struct mbuf **mp) 2280 { 2281 void *optdata = NULL; 2282 int optdatalen = 0; 2283 struct ip6_ext *ip6e; 2284 int error = 0; 2285 struct in6_pktinfo null_pktinfo; 2286 int deftclass = 0, on; 2287 int defminmtu = IP6PO_MINMTU_MCASTONLY; 2288 struct mbuf *m; 2289 2290 switch (optname) { 2291 case IPV6_PKTINFO: 2292 if (pktopt && pktopt->ip6po_pktinfo) 2293 optdata = (void *)pktopt->ip6po_pktinfo; 2294 else { 2295 /* XXX: we don't have to do this every time... */ 2296 memset(&null_pktinfo, 0, sizeof(null_pktinfo)); 2297 optdata = (void *)&null_pktinfo; 2298 } 2299 optdatalen = sizeof(struct in6_pktinfo); 2300 break; 2301 case IPV6_OTCLASS: 2302 /* XXX */ 2303 return (EINVAL); 2304 case IPV6_TCLASS: 2305 if (pktopt && pktopt->ip6po_tclass >= 0) 2306 optdata = (void *)&pktopt->ip6po_tclass; 2307 else 2308 optdata = (void *)&deftclass; 2309 optdatalen = sizeof(int); 2310 break; 2311 case IPV6_HOPOPTS: 2312 if (pktopt && pktopt->ip6po_hbh) { 2313 optdata = (void *)pktopt->ip6po_hbh; 2314 ip6e = (struct ip6_ext *)pktopt->ip6po_hbh; 2315 optdatalen = (ip6e->ip6e_len + 1) << 3; 2316 } 2317 break; 2318 case IPV6_RTHDR: 2319 if (pktopt && pktopt->ip6po_rthdr) { 2320 optdata = (void *)pktopt->ip6po_rthdr; 2321 ip6e = (struct ip6_ext *)pktopt->ip6po_rthdr; 2322 optdatalen = (ip6e->ip6e_len + 1) << 3; 2323 } 2324 break; 2325 case IPV6_RTHDRDSTOPTS: 2326 if (pktopt && pktopt->ip6po_dest1) { 2327 optdata = (void *)pktopt->ip6po_dest1; 2328 ip6e = (struct ip6_ext *)pktopt->ip6po_dest1; 2329 optdatalen = (ip6e->ip6e_len + 1) << 3; 2330 } 2331 break; 2332 case IPV6_DSTOPTS: 2333 if (pktopt && pktopt->ip6po_dest2) { 2334 optdata = (void *)pktopt->ip6po_dest2; 2335 ip6e = (struct ip6_ext *)pktopt->ip6po_dest2; 2336 optdatalen = (ip6e->ip6e_len + 1) << 3; 2337 } 2338 break; 2339 case IPV6_NEXTHOP: 2340 if (pktopt && pktopt->ip6po_nexthop) { 2341 optdata = (void *)pktopt->ip6po_nexthop; 2342 optdatalen = pktopt->ip6po_nexthop->sa_len; 2343 } 2344 break; 2345 case IPV6_USE_MIN_MTU: 2346 if (pktopt) 2347 optdata = (void *)&pktopt->ip6po_minmtu; 2348 else 2349 optdata = (void *)&defminmtu; 2350 optdatalen = sizeof(int); 2351 break; 2352 case IPV6_DONTFRAG: 2353 if (pktopt && ((pktopt->ip6po_flags) & IP6PO_DONTFRAG)) 2354 on = 1; 2355 else 2356 on = 0; 2357 optdata = (void *)&on; 2358 optdatalen = sizeof(on); 2359 break; 2360 default: /* should not happen */ 2361 #ifdef DIAGNOSTIC 2362 panic("ip6_getpcbopt: unexpected option\n"); 2363 #endif 2364 return (ENOPROTOOPT); 2365 } 2366 2367 if (optdatalen > MCLBYTES) 2368 return (EMSGSIZE); /* XXX */ 2369 *mp = m = m_get(M_WAIT, MT_SOOPTS); 2370 if (optdatalen > MLEN) 2371 MCLGET(m, M_WAIT); 2372 m->m_len = optdatalen; 2373 if (optdatalen) 2374 memcpy(mtod(m, void *), optdata, optdatalen); 2375 2376 return (error); 2377 } 2378 2379 void 2380 ip6_clearpktopts(struct ip6_pktopts *pktopt, int optname) 2381 { 2382 if (optname == -1 || optname == IPV6_PKTINFO) { 2383 if (pktopt->ip6po_pktinfo) 2384 free(pktopt->ip6po_pktinfo, M_IP6OPT); 2385 pktopt->ip6po_pktinfo = NULL; 2386 } 2387 if (optname == -1 || optname == IPV6_HOPLIMIT) 2388 pktopt->ip6po_hlim = -1; 2389 if (optname == -1 || optname == IPV6_TCLASS) 2390 pktopt->ip6po_tclass = -1; 2391 if (optname == -1 || optname == IPV6_NEXTHOP) { 2392 rtcache_free((struct route *)&pktopt->ip6po_nextroute); 2393 if (pktopt->ip6po_nexthop) 2394 free(pktopt->ip6po_nexthop, M_IP6OPT); 2395 pktopt->ip6po_nexthop = NULL; 2396 } 2397 if (optname == -1 || optname == IPV6_HOPOPTS) { 2398 if (pktopt->ip6po_hbh) 2399 free(pktopt->ip6po_hbh, M_IP6OPT); 2400 pktopt->ip6po_hbh = NULL; 2401 } 2402 if (optname == -1 || optname == IPV6_RTHDRDSTOPTS) { 2403 if (pktopt->ip6po_dest1) 2404 free(pktopt->ip6po_dest1, M_IP6OPT); 2405 pktopt->ip6po_dest1 = NULL; 2406 } 2407 if (optname == -1 || optname == IPV6_RTHDR) { 2408 if (pktopt->ip6po_rhinfo.ip6po_rhi_rthdr) 2409 free(pktopt->ip6po_rhinfo.ip6po_rhi_rthdr, M_IP6OPT); 2410 pktopt->ip6po_rhinfo.ip6po_rhi_rthdr = NULL; 2411 rtcache_free((struct route *)&pktopt->ip6po_route); 2412 } 2413 if (optname == -1 || optname == IPV6_DSTOPTS) { 2414 if (pktopt->ip6po_dest2) 2415 free(pktopt->ip6po_dest2, M_IP6OPT); 2416 pktopt->ip6po_dest2 = NULL; 2417 } 2418 } 2419 2420 #define PKTOPT_EXTHDRCPY(type) \ 2421 do { \ 2422 if (src->type) { \ 2423 int hlen = (((struct ip6_ext *)src->type)->ip6e_len + 1) << 3;\ 2424 dst->type = malloc(hlen, M_IP6OPT, canwait); \ 2425 if (dst->type == NULL && canwait == M_NOWAIT) \ 2426 goto bad; \ 2427 memcpy(dst->type, src->type, hlen); \ 2428 } \ 2429 } while (/*CONSTCOND*/ 0) 2430 2431 static int 2432 copypktopts(struct ip6_pktopts *dst, struct ip6_pktopts *src, int canwait) 2433 { 2434 dst->ip6po_hlim = src->ip6po_hlim; 2435 dst->ip6po_tclass = src->ip6po_tclass; 2436 dst->ip6po_flags = src->ip6po_flags; 2437 if (src->ip6po_pktinfo) { 2438 dst->ip6po_pktinfo = malloc(sizeof(*dst->ip6po_pktinfo), 2439 M_IP6OPT, canwait); 2440 if (dst->ip6po_pktinfo == NULL && canwait == M_NOWAIT) 2441 goto bad; 2442 *dst->ip6po_pktinfo = *src->ip6po_pktinfo; 2443 } 2444 if (src->ip6po_nexthop) { 2445 dst->ip6po_nexthop = malloc(src->ip6po_nexthop->sa_len, 2446 M_IP6OPT, canwait); 2447 if (dst->ip6po_nexthop == NULL && canwait == M_NOWAIT) 2448 goto bad; 2449 memcpy(dst->ip6po_nexthop, src->ip6po_nexthop, 2450 src->ip6po_nexthop->sa_len); 2451 } 2452 PKTOPT_EXTHDRCPY(ip6po_hbh); 2453 PKTOPT_EXTHDRCPY(ip6po_dest1); 2454 PKTOPT_EXTHDRCPY(ip6po_dest2); 2455 PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */ 2456 return (0); 2457 2458 bad: 2459 if (dst->ip6po_pktinfo) free(dst->ip6po_pktinfo, M_IP6OPT); 2460 if (dst->ip6po_nexthop) free(dst->ip6po_nexthop, M_IP6OPT); 2461 if (dst->ip6po_hbh) free(dst->ip6po_hbh, M_IP6OPT); 2462 if (dst->ip6po_dest1) free(dst->ip6po_dest1, M_IP6OPT); 2463 if (dst->ip6po_dest2) free(dst->ip6po_dest2, M_IP6OPT); 2464 if (dst->ip6po_rthdr) free(dst->ip6po_rthdr, M_IP6OPT); 2465 2466 return (ENOBUFS); 2467 } 2468 #undef PKTOPT_EXTHDRCPY 2469 2470 struct ip6_pktopts * 2471 ip6_copypktopts(struct ip6_pktopts *src, int canwait) 2472 { 2473 int error; 2474 struct ip6_pktopts *dst; 2475 2476 dst = malloc(sizeof(*dst), M_IP6OPT, canwait); 2477 if (dst == NULL && canwait == M_NOWAIT) 2478 return (NULL); 2479 ip6_initpktopts(dst); 2480 2481 if ((error = copypktopts(dst, src, canwait)) != 0) { 2482 free(dst, M_IP6OPT); 2483 return (NULL); 2484 } 2485 2486 return (dst); 2487 } 2488 2489 void 2490 ip6_freepcbopts(struct ip6_pktopts *pktopt) 2491 { 2492 if (pktopt == NULL) 2493 return; 2494 2495 ip6_clearpktopts(pktopt, -1); 2496 2497 free(pktopt, M_IP6OPT); 2498 } 2499 2500 /* 2501 * Set the IP6 multicast options in response to user setsockopt(). 2502 */ 2503 static int 2504 ip6_setmoptions(optname, im6op, m) 2505 int optname; 2506 struct ip6_moptions **im6op; 2507 struct mbuf *m; 2508 { 2509 int error = 0; 2510 u_int loop, ifindex; 2511 struct ipv6_mreq *mreq; 2512 struct ifnet *ifp; 2513 struct ip6_moptions *im6o = *im6op; 2514 struct route_in6 ro; 2515 struct in6_multi_mship *imm; 2516 struct lwp *l = curlwp; /* XXX */ 2517 2518 if (im6o == NULL) { 2519 /* 2520 * No multicast option buffer attached to the pcb; 2521 * allocate one and initialize to default values. 2522 */ 2523 im6o = (struct ip6_moptions *) 2524 malloc(sizeof(*im6o), M_IPMOPTS, M_WAITOK); 2525 2526 if (im6o == NULL) 2527 return (ENOBUFS); 2528 *im6op = im6o; 2529 im6o->im6o_multicast_ifp = NULL; 2530 im6o->im6o_multicast_hlim = ip6_defmcasthlim; 2531 im6o->im6o_multicast_loop = IPV6_DEFAULT_MULTICAST_LOOP; 2532 LIST_INIT(&im6o->im6o_memberships); 2533 } 2534 2535 switch (optname) { 2536 2537 case IPV6_MULTICAST_IF: 2538 /* 2539 * Select the interface for outgoing multicast packets. 2540 */ 2541 if (m == NULL || m->m_len != sizeof(u_int)) { 2542 error = EINVAL; 2543 break; 2544 } 2545 bcopy(mtod(m, u_int *), &ifindex, sizeof(ifindex)); 2546 if (ifindex != 0) { 2547 if (if_indexlim <= ifindex || !ifindex2ifnet[ifindex]) { 2548 error = ENXIO; /* XXX EINVAL? */ 2549 break; 2550 } 2551 ifp = ifindex2ifnet[ifindex]; 2552 if ((ifp->if_flags & IFF_MULTICAST) == 0) { 2553 error = EADDRNOTAVAIL; 2554 break; 2555 } 2556 } else 2557 ifp = NULL; 2558 im6o->im6o_multicast_ifp = ifp; 2559 break; 2560 2561 case IPV6_MULTICAST_HOPS: 2562 { 2563 /* 2564 * Set the IP6 hoplimit for outgoing multicast packets. 2565 */ 2566 int optval; 2567 if (m == NULL || m->m_len != sizeof(int)) { 2568 error = EINVAL; 2569 break; 2570 } 2571 bcopy(mtod(m, u_int *), &optval, sizeof(optval)); 2572 if (optval < -1 || optval >= 256) 2573 error = EINVAL; 2574 else if (optval == -1) 2575 im6o->im6o_multicast_hlim = ip6_defmcasthlim; 2576 else 2577 im6o->im6o_multicast_hlim = optval; 2578 break; 2579 } 2580 2581 case IPV6_MULTICAST_LOOP: 2582 /* 2583 * Set the loopback flag for outgoing multicast packets. 2584 * Must be zero or one. 2585 */ 2586 if (m == NULL || m->m_len != sizeof(u_int)) { 2587 error = EINVAL; 2588 break; 2589 } 2590 bcopy(mtod(m, u_int *), &loop, sizeof(loop)); 2591 if (loop > 1) { 2592 error = EINVAL; 2593 break; 2594 } 2595 im6o->im6o_multicast_loop = loop; 2596 break; 2597 2598 case IPV6_JOIN_GROUP: 2599 /* 2600 * Add a multicast group membership. 2601 * Group must be a valid IP6 multicast address. 2602 */ 2603 if (m == NULL || m->m_len != sizeof(struct ipv6_mreq)) { 2604 error = EINVAL; 2605 break; 2606 } 2607 mreq = mtod(m, struct ipv6_mreq *); 2608 if (IN6_IS_ADDR_UNSPECIFIED(&mreq->ipv6mr_multiaddr)) { 2609 /* 2610 * We use the unspecified address to specify to accept 2611 * all multicast addresses. Only super user is allowed 2612 * to do this. 2613 */ 2614 if (kauth_authorize_generic(l->l_cred, 2615 KAUTH_GENERIC_ISSUSER, NULL)) 2616 { 2617 error = EACCES; 2618 break; 2619 } 2620 } else if (!IN6_IS_ADDR_MULTICAST(&mreq->ipv6mr_multiaddr)) { 2621 error = EINVAL; 2622 break; 2623 } 2624 2625 /* 2626 * If no interface was explicitly specified, choose an 2627 * appropriate one according to the given multicast address. 2628 */ 2629 if (mreq->ipv6mr_interface == 0) { 2630 struct sockaddr_in6 *dst; 2631 2632 /* 2633 * Look up the routing table for the 2634 * address, and choose the outgoing interface. 2635 * XXX: is it a good approach? 2636 */ 2637 memset(&ro, 0, sizeof(ro)); 2638 dst = &ro.ro_dst; 2639 dst->sin6_family = AF_INET6; 2640 dst->sin6_len = sizeof(*dst); 2641 dst->sin6_addr = mreq->ipv6mr_multiaddr; 2642 rtcache_init((struct route *)&ro); 2643 ifp = (ro.ro_rt != NULL) ? ro.ro_rt->rt_ifp : NULL; 2644 rtcache_free((struct route *)&ro); 2645 } else { 2646 /* 2647 * If the interface is specified, validate it. 2648 */ 2649 if (if_indexlim <= mreq->ipv6mr_interface || 2650 !ifindex2ifnet[mreq->ipv6mr_interface]) { 2651 error = ENXIO; /* XXX EINVAL? */ 2652 break; 2653 } 2654 ifp = ifindex2ifnet[mreq->ipv6mr_interface]; 2655 } 2656 2657 /* 2658 * See if we found an interface, and confirm that it 2659 * supports multicast 2660 */ 2661 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) { 2662 error = EADDRNOTAVAIL; 2663 break; 2664 } 2665 2666 if (in6_setscope(&mreq->ipv6mr_multiaddr, ifp, NULL)) { 2667 error = EADDRNOTAVAIL; /* XXX: should not happen */ 2668 break; 2669 } 2670 2671 /* 2672 * See if the membership already exists. 2673 */ 2674 for (imm = im6o->im6o_memberships.lh_first; 2675 imm != NULL; imm = imm->i6mm_chain.le_next) 2676 if (imm->i6mm_maddr->in6m_ifp == ifp && 2677 IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr, 2678 &mreq->ipv6mr_multiaddr)) 2679 break; 2680 if (imm != NULL) { 2681 error = EADDRINUSE; 2682 break; 2683 } 2684 /* 2685 * Everything looks good; add a new record to the multicast 2686 * address list for the given interface. 2687 */ 2688 imm = in6_joingroup(ifp, &mreq->ipv6mr_multiaddr, &error, 0); 2689 if (imm == NULL) 2690 break; 2691 LIST_INSERT_HEAD(&im6o->im6o_memberships, imm, i6mm_chain); 2692 break; 2693 2694 case IPV6_LEAVE_GROUP: 2695 /* 2696 * Drop a multicast group membership. 2697 * Group must be a valid IP6 multicast address. 2698 */ 2699 if (m == NULL || m->m_len != sizeof(struct ipv6_mreq)) { 2700 error = EINVAL; 2701 break; 2702 } 2703 mreq = mtod(m, struct ipv6_mreq *); 2704 2705 /* 2706 * If an interface address was specified, get a pointer 2707 * to its ifnet structure. 2708 */ 2709 if (mreq->ipv6mr_interface != 0) { 2710 if (if_indexlim <= mreq->ipv6mr_interface || 2711 !ifindex2ifnet[mreq->ipv6mr_interface]) { 2712 error = ENXIO; /* XXX EINVAL? */ 2713 break; 2714 } 2715 ifp = ifindex2ifnet[mreq->ipv6mr_interface]; 2716 } else 2717 ifp = NULL; 2718 2719 /* Fill in the scope zone ID */ 2720 if (ifp) { 2721 if (in6_setscope(&mreq->ipv6mr_multiaddr, ifp, NULL)) { 2722 /* XXX: should not happen */ 2723 error = EADDRNOTAVAIL; 2724 break; 2725 } 2726 } else if (mreq->ipv6mr_interface != 0) { 2727 /* 2728 * XXX: This case would happens when the (positive) 2729 * index is in the valid range, but the corresponding 2730 * interface has been detached dynamically. The above 2731 * check probably avoids such case to happen here, but 2732 * we check it explicitly for safety. 2733 */ 2734 error = EADDRNOTAVAIL; 2735 break; 2736 } else { /* ipv6mr_interface == 0 */ 2737 struct sockaddr_in6 sa6_mc; 2738 2739 /* 2740 * The API spec says as follows: 2741 * If the interface index is specified as 0, the 2742 * system may choose a multicast group membership to 2743 * drop by matching the multicast address only. 2744 * On the other hand, we cannot disambiguate the scope 2745 * zone unless an interface is provided. Thus, we 2746 * check if there's ambiguity with the default scope 2747 * zone as the last resort. 2748 */ 2749 bzero(&sa6_mc, sizeof(sa6_mc)); 2750 sa6_mc.sin6_family = AF_INET6; 2751 sa6_mc.sin6_len = sizeof(sa6_mc); 2752 sa6_mc.sin6_addr = mreq->ipv6mr_multiaddr; 2753 error = sa6_embedscope(&sa6_mc, ip6_use_defzone); 2754 if (error != 0) 2755 break; 2756 mreq->ipv6mr_multiaddr = sa6_mc.sin6_addr; 2757 } 2758 2759 /* 2760 * Find the membership in the membership list. 2761 */ 2762 for (imm = im6o->im6o_memberships.lh_first; 2763 imm != NULL; imm = imm->i6mm_chain.le_next) { 2764 if ((ifp == NULL || imm->i6mm_maddr->in6m_ifp == ifp) && 2765 IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr, 2766 &mreq->ipv6mr_multiaddr)) 2767 break; 2768 } 2769 if (imm == NULL) { 2770 /* Unable to resolve interface */ 2771 error = EADDRNOTAVAIL; 2772 break; 2773 } 2774 /* 2775 * Give up the multicast address record to which the 2776 * membership points. 2777 */ 2778 LIST_REMOVE(imm, i6mm_chain); 2779 in6_leavegroup(imm); 2780 break; 2781 2782 default: 2783 error = EOPNOTSUPP; 2784 break; 2785 } 2786 2787 /* 2788 * If all options have default values, no need to keep the mbuf. 2789 */ 2790 if (im6o->im6o_multicast_ifp == NULL && 2791 im6o->im6o_multicast_hlim == ip6_defmcasthlim && 2792 im6o->im6o_multicast_loop == IPV6_DEFAULT_MULTICAST_LOOP && 2793 im6o->im6o_memberships.lh_first == NULL) { 2794 free(*im6op, M_IPMOPTS); 2795 *im6op = NULL; 2796 } 2797 2798 return (error); 2799 } 2800 2801 /* 2802 * Return the IP6 multicast options in response to user getsockopt(). 2803 */ 2804 static int 2805 ip6_getmoptions(optname, im6o, mp) 2806 int optname; 2807 struct ip6_moptions *im6o; 2808 struct mbuf **mp; 2809 { 2810 u_int *hlim, *loop, *ifindex; 2811 2812 *mp = m_get(M_WAIT, MT_SOOPTS); 2813 2814 switch (optname) { 2815 2816 case IPV6_MULTICAST_IF: 2817 ifindex = mtod(*mp, u_int *); 2818 (*mp)->m_len = sizeof(u_int); 2819 if (im6o == NULL || im6o->im6o_multicast_ifp == NULL) 2820 *ifindex = 0; 2821 else 2822 *ifindex = im6o->im6o_multicast_ifp->if_index; 2823 return (0); 2824 2825 case IPV6_MULTICAST_HOPS: 2826 hlim = mtod(*mp, u_int *); 2827 (*mp)->m_len = sizeof(u_int); 2828 if (im6o == NULL) 2829 *hlim = ip6_defmcasthlim; 2830 else 2831 *hlim = im6o->im6o_multicast_hlim; 2832 return (0); 2833 2834 case IPV6_MULTICAST_LOOP: 2835 loop = mtod(*mp, u_int *); 2836 (*mp)->m_len = sizeof(u_int); 2837 if (im6o == NULL) 2838 *loop = ip6_defmcasthlim; 2839 else 2840 *loop = im6o->im6o_multicast_loop; 2841 return (0); 2842 2843 default: 2844 return (EOPNOTSUPP); 2845 } 2846 } 2847 2848 /* 2849 * Discard the IP6 multicast options. 2850 */ 2851 void 2852 ip6_freemoptions(im6o) 2853 struct ip6_moptions *im6o; 2854 { 2855 struct in6_multi_mship *imm; 2856 2857 if (im6o == NULL) 2858 return; 2859 2860 while ((imm = im6o->im6o_memberships.lh_first) != NULL) { 2861 LIST_REMOVE(imm, i6mm_chain); 2862 in6_leavegroup(imm); 2863 } 2864 free(im6o, M_IPMOPTS); 2865 } 2866 2867 /* 2868 * Set IPv6 outgoing packet options based on advanced API. 2869 */ 2870 int 2871 ip6_setpktopts(control, opt, stickyopt, priv, uproto) 2872 struct mbuf *control; 2873 struct ip6_pktopts *opt, *stickyopt; 2874 int priv, uproto; 2875 { 2876 struct cmsghdr *cm = 0; 2877 2878 if (control == NULL || opt == NULL) 2879 return (EINVAL); 2880 2881 ip6_initpktopts(opt); 2882 if (stickyopt) { 2883 int error; 2884 2885 /* 2886 * If stickyopt is provided, make a local copy of the options 2887 * for this particular packet, then override them by ancillary 2888 * objects. 2889 * XXX: copypktopts() does not copy the cached route to a next 2890 * hop (if any). This is not very good in terms of efficiency, 2891 * but we can allow this since this option should be rarely 2892 * used. 2893 */ 2894 if ((error = copypktopts(opt, stickyopt, M_NOWAIT)) != 0) 2895 return (error); 2896 } 2897 2898 /* 2899 * XXX: Currently, we assume all the optional information is stored 2900 * in a single mbuf. 2901 */ 2902 if (control->m_next) 2903 return (EINVAL); 2904 2905 for (; control->m_len; control->m_data += CMSG_ALIGN(cm->cmsg_len), 2906 control->m_len -= CMSG_ALIGN(cm->cmsg_len)) { 2907 int error; 2908 2909 if (control->m_len < CMSG_LEN(0)) 2910 return (EINVAL); 2911 2912 cm = mtod(control, struct cmsghdr *); 2913 if (cm->cmsg_len == 0 || cm->cmsg_len > control->m_len) 2914 return (EINVAL); 2915 if (cm->cmsg_level != IPPROTO_IPV6) 2916 continue; 2917 2918 error = ip6_setpktopt(cm->cmsg_type, CMSG_DATA(cm), 2919 cm->cmsg_len - CMSG_LEN(0), opt, priv, 0, 1, uproto); 2920 if (error) 2921 return (error); 2922 } 2923 2924 return (0); 2925 } 2926 2927 /* 2928 * Set a particular packet option, as a sticky option or an ancillary data 2929 * item. "len" can be 0 only when it's a sticky option. 2930 * We have 4 cases of combination of "sticky" and "cmsg": 2931 * "sticky=0, cmsg=0": impossible 2932 * "sticky=0, cmsg=1": RFC2292 or RFC3542 ancillary data 2933 * "sticky=1, cmsg=0": RFC3542 socket option 2934 * "sticky=1, cmsg=1": RFC2292 socket option 2935 */ 2936 static int 2937 ip6_setpktopt(int optname, u_char *buf, int len, struct ip6_pktopts *opt, 2938 int priv, int sticky, int cmsg, int uproto) 2939 { 2940 int minmtupolicy; 2941 2942 if (!sticky && !cmsg) { 2943 #ifdef DIAGNOSTIC 2944 printf("ip6_setpktopt: impossible case\n"); 2945 #endif 2946 return (EINVAL); 2947 } 2948 2949 /* 2950 * IPV6_2292xxx is for backward compatibility to RFC2292, and should 2951 * not be specified in the context of RFC3542. Conversely, 2952 * RFC3542 types should not be specified in the context of RFC2292. 2953 */ 2954 if (!cmsg) { 2955 switch (optname) { 2956 case IPV6_2292PKTINFO: 2957 case IPV6_2292HOPLIMIT: 2958 case IPV6_2292NEXTHOP: 2959 case IPV6_2292HOPOPTS: 2960 case IPV6_2292DSTOPTS: 2961 case IPV6_2292RTHDR: 2962 case IPV6_2292PKTOPTIONS: 2963 return (ENOPROTOOPT); 2964 } 2965 } 2966 if (sticky && cmsg) { 2967 switch (optname) { 2968 case IPV6_PKTINFO: 2969 case IPV6_HOPLIMIT: 2970 case IPV6_NEXTHOP: 2971 case IPV6_HOPOPTS: 2972 case IPV6_DSTOPTS: 2973 case IPV6_RTHDRDSTOPTS: 2974 case IPV6_RTHDR: 2975 case IPV6_USE_MIN_MTU: 2976 case IPV6_DONTFRAG: 2977 case IPV6_OTCLASS: 2978 case IPV6_TCLASS: 2979 return (ENOPROTOOPT); 2980 } 2981 } 2982 2983 switch (optname) { 2984 #ifdef RFC2292 2985 case IPV6_2292PKTINFO: 2986 #endif 2987 case IPV6_PKTINFO: 2988 { 2989 struct ifnet *ifp = NULL; 2990 struct in6_pktinfo *pktinfo; 2991 2992 if (len != sizeof(struct in6_pktinfo)) 2993 return (EINVAL); 2994 2995 pktinfo = (struct in6_pktinfo *)buf; 2996 2997 /* 2998 * An application can clear any sticky IPV6_PKTINFO option by 2999 * doing a "regular" setsockopt with ipi6_addr being 3000 * in6addr_any and ipi6_ifindex being zero. 3001 * [RFC 3542, Section 6] 3002 */ 3003 if (optname == IPV6_PKTINFO && opt->ip6po_pktinfo && 3004 pktinfo->ipi6_ifindex == 0 && 3005 IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) { 3006 ip6_clearpktopts(opt, optname); 3007 break; 3008 } 3009 3010 if (uproto == IPPROTO_TCP && optname == IPV6_PKTINFO && 3011 sticky && !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) { 3012 return (EINVAL); 3013 } 3014 3015 /* validate the interface index if specified. */ 3016 if (pktinfo->ipi6_ifindex >= if_indexlim) { 3017 return (ENXIO); 3018 } 3019 if (pktinfo->ipi6_ifindex) { 3020 ifp = ifindex2ifnet[pktinfo->ipi6_ifindex]; 3021 if (ifp == NULL) 3022 return (ENXIO); 3023 } 3024 3025 /* 3026 * We store the address anyway, and let in6_selectsrc() 3027 * validate the specified address. This is because ipi6_addr 3028 * may not have enough information about its scope zone, and 3029 * we may need additional information (such as outgoing 3030 * interface or the scope zone of a destination address) to 3031 * disambiguate the scope. 3032 * XXX: the delay of the validation may confuse the 3033 * application when it is used as a sticky option. 3034 */ 3035 if (opt->ip6po_pktinfo == NULL) { 3036 opt->ip6po_pktinfo = malloc(sizeof(*pktinfo), 3037 M_IP6OPT, M_NOWAIT); 3038 if (opt->ip6po_pktinfo == NULL) 3039 return (ENOBUFS); 3040 } 3041 memcpy(opt->ip6po_pktinfo, pktinfo, sizeof(*pktinfo)); 3042 break; 3043 } 3044 3045 #ifdef RFC2292 3046 case IPV6_2292HOPLIMIT: 3047 #endif 3048 case IPV6_HOPLIMIT: 3049 { 3050 int *hlimp; 3051 3052 /* 3053 * RFC 3542 deprecated the usage of sticky IPV6_HOPLIMIT 3054 * to simplify the ordering among hoplimit options. 3055 */ 3056 if (optname == IPV6_HOPLIMIT && sticky) 3057 return (ENOPROTOOPT); 3058 3059 if (len != sizeof(int)) 3060 return (EINVAL); 3061 hlimp = (int *)buf; 3062 if (*hlimp < -1 || *hlimp > 255) 3063 return (EINVAL); 3064 3065 opt->ip6po_hlim = *hlimp; 3066 break; 3067 } 3068 3069 case IPV6_OTCLASS: 3070 if (len != sizeof(u_int8_t)) 3071 return (EINVAL); 3072 3073 opt->ip6po_tclass = *(u_int8_t *)buf; 3074 break; 3075 3076 case IPV6_TCLASS: 3077 { 3078 int tclass; 3079 3080 if (len != sizeof(int)) 3081 return (EINVAL); 3082 tclass = *(int *)buf; 3083 if (tclass < -1 || tclass > 255) 3084 return (EINVAL); 3085 3086 opt->ip6po_tclass = tclass; 3087 break; 3088 } 3089 3090 #ifdef RFC2292 3091 case IPV6_2292NEXTHOP: 3092 #endif 3093 case IPV6_NEXTHOP: 3094 if (!priv) 3095 return (EPERM); 3096 3097 if (len == 0) { /* just remove the option */ 3098 ip6_clearpktopts(opt, IPV6_NEXTHOP); 3099 break; 3100 } 3101 3102 /* check if cmsg_len is large enough for sa_len */ 3103 if (len < sizeof(struct sockaddr) || len < *buf) 3104 return (EINVAL); 3105 3106 switch (((struct sockaddr *)buf)->sa_family) { 3107 case AF_INET6: 3108 { 3109 struct sockaddr_in6 *sa6 = (struct sockaddr_in6 *)buf; 3110 int error; 3111 3112 if (sa6->sin6_len != sizeof(struct sockaddr_in6)) 3113 return (EINVAL); 3114 3115 if (IN6_IS_ADDR_UNSPECIFIED(&sa6->sin6_addr) || 3116 IN6_IS_ADDR_MULTICAST(&sa6->sin6_addr)) { 3117 return (EINVAL); 3118 } 3119 if ((error = sa6_embedscope(sa6, ip6_use_defzone)) 3120 != 0) { 3121 return (error); 3122 } 3123 break; 3124 } 3125 case AF_LINK: /* eventually be supported? */ 3126 default: 3127 return (EAFNOSUPPORT); 3128 } 3129 3130 /* turn off the previous option, then set the new option. */ 3131 ip6_clearpktopts(opt, IPV6_NEXTHOP); 3132 opt->ip6po_nexthop = malloc(*buf, M_IP6OPT, M_NOWAIT); 3133 if (opt->ip6po_nexthop == NULL) 3134 return (ENOBUFS); 3135 memcpy(opt->ip6po_nexthop, buf, *buf); 3136 break; 3137 3138 #ifdef RFC2292 3139 case IPV6_2292HOPOPTS: 3140 #endif 3141 case IPV6_HOPOPTS: 3142 { 3143 struct ip6_hbh *hbh; 3144 int hbhlen; 3145 3146 /* 3147 * XXX: We don't allow a non-privileged user to set ANY HbH 3148 * options, since per-option restriction has too much 3149 * overhead. 3150 */ 3151 if (!priv) 3152 return (EPERM); 3153 3154 if (len == 0) { 3155 ip6_clearpktopts(opt, IPV6_HOPOPTS); 3156 break; /* just remove the option */ 3157 } 3158 3159 /* message length validation */ 3160 if (len < sizeof(struct ip6_hbh)) 3161 return (EINVAL); 3162 hbh = (struct ip6_hbh *)buf; 3163 hbhlen = (hbh->ip6h_len + 1) << 3; 3164 if (len != hbhlen) 3165 return (EINVAL); 3166 3167 /* turn off the previous option, then set the new option. */ 3168 ip6_clearpktopts(opt, IPV6_HOPOPTS); 3169 opt->ip6po_hbh = malloc(hbhlen, M_IP6OPT, M_NOWAIT); 3170 if (opt->ip6po_hbh == NULL) 3171 return (ENOBUFS); 3172 memcpy(opt->ip6po_hbh, hbh, hbhlen); 3173 3174 break; 3175 } 3176 3177 #ifdef RFC2292 3178 case IPV6_2292DSTOPTS: 3179 #endif 3180 case IPV6_DSTOPTS: 3181 case IPV6_RTHDRDSTOPTS: 3182 { 3183 struct ip6_dest *dest, **newdest = NULL; 3184 int destlen; 3185 3186 if (!priv) /* XXX: see the comment for IPV6_HOPOPTS */ 3187 return (EPERM); 3188 3189 if (len == 0) { 3190 ip6_clearpktopts(opt, optname); 3191 break; /* just remove the option */ 3192 } 3193 3194 /* message length validation */ 3195 if (len < sizeof(struct ip6_dest)) 3196 return (EINVAL); 3197 dest = (struct ip6_dest *)buf; 3198 destlen = (dest->ip6d_len + 1) << 3; 3199 if (len != destlen) 3200 return (EINVAL); 3201 /* 3202 * Determine the position that the destination options header 3203 * should be inserted; before or after the routing header. 3204 */ 3205 switch (optname) { 3206 case IPV6_2292DSTOPTS: 3207 /* 3208 * The old advanced API is ambiguous on this point. 3209 * Our approach is to determine the position based 3210 * according to the existence of a routing header. 3211 * Note, however, that this depends on the order of the 3212 * extension headers in the ancillary data; the 1st 3213 * part of the destination options header must appear 3214 * before the routing header in the ancillary data, 3215 * too. 3216 * RFC3542 solved the ambiguity by introducing 3217 * separate ancillary data or option types. 3218 */ 3219 if (opt->ip6po_rthdr == NULL) 3220 newdest = &opt->ip6po_dest1; 3221 else 3222 newdest = &opt->ip6po_dest2; 3223 break; 3224 case IPV6_RTHDRDSTOPTS: 3225 newdest = &opt->ip6po_dest1; 3226 break; 3227 case IPV6_DSTOPTS: 3228 newdest = &opt->ip6po_dest2; 3229 break; 3230 } 3231 3232 /* turn off the previous option, then set the new option. */ 3233 ip6_clearpktopts(opt, optname); 3234 *newdest = malloc(destlen, M_IP6OPT, M_NOWAIT); 3235 if (*newdest == NULL) 3236 return (ENOBUFS); 3237 memcpy(*newdest, dest, destlen); 3238 3239 break; 3240 } 3241 3242 #ifdef RFC2292 3243 case IPV6_2292RTHDR: 3244 #endif 3245 case IPV6_RTHDR: 3246 { 3247 struct ip6_rthdr *rth; 3248 int rthlen; 3249 3250 if (len == 0) { 3251 ip6_clearpktopts(opt, IPV6_RTHDR); 3252 break; /* just remove the option */ 3253 } 3254 3255 /* message length validation */ 3256 if (len < sizeof(struct ip6_rthdr)) 3257 return (EINVAL); 3258 rth = (struct ip6_rthdr *)buf; 3259 rthlen = (rth->ip6r_len + 1) << 3; 3260 if (len != rthlen) 3261 return (EINVAL); 3262 switch (rth->ip6r_type) { 3263 case IPV6_RTHDR_TYPE_0: 3264 if (rth->ip6r_len == 0) /* must contain one addr */ 3265 return (EINVAL); 3266 if (rth->ip6r_len % 2) /* length must be even */ 3267 return (EINVAL); 3268 if (rth->ip6r_len / 2 != rth->ip6r_segleft) 3269 return (EINVAL); 3270 break; 3271 default: 3272 return (EINVAL); /* not supported */ 3273 } 3274 /* turn off the previous option */ 3275 ip6_clearpktopts(opt, IPV6_RTHDR); 3276 opt->ip6po_rthdr = malloc(rthlen, M_IP6OPT, M_NOWAIT); 3277 if (opt->ip6po_rthdr == NULL) 3278 return (ENOBUFS); 3279 memcpy(opt->ip6po_rthdr, rth, rthlen); 3280 break; 3281 } 3282 3283 case IPV6_USE_MIN_MTU: 3284 if (len != sizeof(int)) 3285 return (EINVAL); 3286 minmtupolicy = *(int *)buf; 3287 if (minmtupolicy != IP6PO_MINMTU_MCASTONLY && 3288 minmtupolicy != IP6PO_MINMTU_DISABLE && 3289 minmtupolicy != IP6PO_MINMTU_ALL) { 3290 return (EINVAL); 3291 } 3292 opt->ip6po_minmtu = minmtupolicy; 3293 break; 3294 3295 case IPV6_DONTFRAG: 3296 if (len != sizeof(int)) 3297 return (EINVAL); 3298 3299 if (uproto == IPPROTO_TCP || *(int *)buf == 0) { 3300 /* 3301 * we ignore this option for TCP sockets. 3302 * (RFC3542 leaves this case unspecified.) 3303 */ 3304 opt->ip6po_flags &= ~IP6PO_DONTFRAG; 3305 } else 3306 opt->ip6po_flags |= IP6PO_DONTFRAG; 3307 break; 3308 3309 default: 3310 return (ENOPROTOOPT); 3311 } /* end of switch */ 3312 3313 return (0); 3314 } 3315 3316 /* 3317 * Routine called from ip6_output() to loop back a copy of an IP6 multicast 3318 * packet to the input queue of a specified interface. Note that this 3319 * calls the output routine of the loopback "driver", but with an interface 3320 * pointer that might NOT be lo0ifp -- easier than replicating that code here. 3321 */ 3322 void 3323 ip6_mloopback(ifp, m, dst) 3324 struct ifnet *ifp; 3325 struct mbuf *m; 3326 const struct sockaddr_in6 *dst; 3327 { 3328 struct mbuf *copym; 3329 struct ip6_hdr *ip6; 3330 3331 copym = m_copy(m, 0, M_COPYALL); 3332 if (copym == NULL) 3333 return; 3334 3335 /* 3336 * Make sure to deep-copy IPv6 header portion in case the data 3337 * is in an mbuf cluster, so that we can safely override the IPv6 3338 * header portion later. 3339 */ 3340 if ((copym->m_flags & M_EXT) != 0 || 3341 copym->m_len < sizeof(struct ip6_hdr)) { 3342 copym = m_pullup(copym, sizeof(struct ip6_hdr)); 3343 if (copym == NULL) 3344 return; 3345 } 3346 3347 #ifdef DIAGNOSTIC 3348 if (copym->m_len < sizeof(*ip6)) { 3349 m_freem(copym); 3350 return; 3351 } 3352 #endif 3353 3354 ip6 = mtod(copym, struct ip6_hdr *); 3355 /* 3356 * clear embedded scope identifiers if necessary. 3357 * in6_clearscope will touch the addresses only when necessary. 3358 */ 3359 in6_clearscope(&ip6->ip6_src); 3360 in6_clearscope(&ip6->ip6_dst); 3361 3362 (void)looutput(ifp, copym, (const struct sockaddr *)dst, NULL); 3363 } 3364 3365 /* 3366 * Chop IPv6 header off from the payload. 3367 */ 3368 static int 3369 ip6_splithdr(m, exthdrs) 3370 struct mbuf *m; 3371 struct ip6_exthdrs *exthdrs; 3372 { 3373 struct mbuf *mh; 3374 struct ip6_hdr *ip6; 3375 3376 ip6 = mtod(m, struct ip6_hdr *); 3377 if (m->m_len > sizeof(*ip6)) { 3378 MGETHDR(mh, M_DONTWAIT, MT_HEADER); 3379 if (mh == 0) { 3380 m_freem(m); 3381 return ENOBUFS; 3382 } 3383 M_MOVE_PKTHDR(mh, m); 3384 MH_ALIGN(mh, sizeof(*ip6)); 3385 m->m_len -= sizeof(*ip6); 3386 m->m_data += sizeof(*ip6); 3387 mh->m_next = m; 3388 m = mh; 3389 m->m_len = sizeof(*ip6); 3390 bcopy((void *)ip6, mtod(m, void *), sizeof(*ip6)); 3391 } 3392 exthdrs->ip6e_ip6 = m; 3393 return 0; 3394 } 3395 3396 /* 3397 * Compute IPv6 extension header length. 3398 */ 3399 int 3400 ip6_optlen(in6p) 3401 struct in6pcb *in6p; 3402 { 3403 int len; 3404 3405 if (!in6p->in6p_outputopts) 3406 return 0; 3407 3408 len = 0; 3409 #define elen(x) \ 3410 (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0) 3411 3412 len += elen(in6p->in6p_outputopts->ip6po_hbh); 3413 len += elen(in6p->in6p_outputopts->ip6po_dest1); 3414 len += elen(in6p->in6p_outputopts->ip6po_rthdr); 3415 len += elen(in6p->in6p_outputopts->ip6po_dest2); 3416 return len; 3417 #undef elen 3418 } 3419
Indexes created Sun Sep 21 18:09:52 GMT 2025