ip6_output.c revision 1.114
1 /* $NetBSD: ip6_output.c,v 1.114 2007/02/10 09:43:05 degroote 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 1.114 2007/02/10 09:43:05 degroote 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 **, caddr_t, 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 __P((struct route_in6 *, struct route_in6 *, 141 struct ifnet *, 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 boolean_t 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), (caddr_t)(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 /* mbuf is already reclaimed in ipsec6_output_trans. */ 474 m = NULL; 475 switch (error) { 476 case EHOSTUNREACH: 477 case ENETUNREACH: 478 case EMSGSIZE: 479 case ENOBUFS: 480 case ENOMEM: 481 break; 482 default: 483 printf("ip6_output (ipsec): error code %d\n", error); 484 /* FALLTHROUGH */ 485 case ENOENT: 486 /* don't show these error codes to the user */ 487 error = 0; 488 break; 489 } 490 goto bad; 491 } 492 if (exthdrs.ip6e_rthdr) { 493 /* ah6_output doesn't modify mbuf chain */ 494 rh->ip6r_segleft = segleft_org; 495 } 496 } 497 skip_ipsec2:; 498 #endif 499 } 500 501 /* 502 * If there is a routing header, replace destination address field 503 * with the first hop of the routing header. 504 */ 505 if (exthdrs.ip6e_rthdr) { 506 struct ip6_rthdr *rh; 507 struct ip6_rthdr0 *rh0; 508 struct in6_addr *addr; 509 struct sockaddr_in6 sa; 510 511 rh = (struct ip6_rthdr *)(mtod(exthdrs.ip6e_rthdr, 512 struct ip6_rthdr *)); 513 finaldst = ip6->ip6_dst; 514 switch (rh->ip6r_type) { 515 case IPV6_RTHDR_TYPE_0: 516 rh0 = (struct ip6_rthdr0 *)rh; 517 addr = (struct in6_addr *)(rh0 + 1); 518 519 /* 520 * construct a sockaddr_in6 form of 521 * the first hop. 522 * 523 * XXX: we may not have enough 524 * information about its scope zone; 525 * there is no standard API to pass 526 * the information from the 527 * application. 528 */ 529 bzero(&sa, sizeof(sa)); 530 sa.sin6_family = AF_INET6; 531 sa.sin6_len = sizeof(sa); 532 sa.sin6_addr = addr[0]; 533 if ((error = sa6_embedscope(&sa, 534 ip6_use_defzone)) != 0) { 535 goto bad; 536 } 537 ip6->ip6_dst = sa.sin6_addr; 538 (void)memmove(&addr[0], &addr[1], 539 sizeof(struct in6_addr) * 540 (rh0->ip6r0_segleft - 1)); 541 addr[rh0->ip6r0_segleft - 1] = finaldst; 542 /* XXX */ 543 in6_clearscope(addr + rh0->ip6r0_segleft - 1); 544 break; 545 default: /* is it possible? */ 546 error = EINVAL; 547 goto bad; 548 } 549 } 550 551 /* Source address validation */ 552 if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) && 553 (flags & IPV6_UNSPECSRC) == 0) { 554 error = EOPNOTSUPP; 555 ip6stat.ip6s_badscope++; 556 goto bad; 557 } 558 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) { 559 error = EOPNOTSUPP; 560 ip6stat.ip6s_badscope++; 561 goto bad; 562 } 563 564 ip6stat.ip6s_localout++; 565 566 /* 567 * Route packet. 568 */ 569 /* initialize cached route */ 570 if (ro == NULL) { 571 memset(&ip6route, 0, sizeof(ip6route)); 572 ro = &ip6route; 573 } 574 ro_pmtu = ro; 575 if (opt && opt->ip6po_rthdr) 576 ro = &opt->ip6po_route; 577 dst = (struct sockaddr_in6 *)&ro->ro_dst; 578 579 /* 580 * if specified, try to fill in the traffic class field. 581 * do not override if a non-zero value is already set. 582 * we check the diffserv field and the ecn field separately. 583 */ 584 if (opt && opt->ip6po_tclass >= 0) { 585 int mask = 0; 586 587 if ((ip6->ip6_flow & htonl(0xfc << 20)) == 0) 588 mask |= 0xfc; 589 if ((ip6->ip6_flow & htonl(0x03 << 20)) == 0) 590 mask |= 0x03; 591 if (mask != 0) 592 ip6->ip6_flow |= htonl((opt->ip6po_tclass & mask) << 20); 593 } 594 595 /* fill in or override the hop limit field, if necessary. */ 596 if (opt && opt->ip6po_hlim != -1) 597 ip6->ip6_hlim = opt->ip6po_hlim & 0xff; 598 else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) { 599 if (im6o != NULL) 600 ip6->ip6_hlim = im6o->im6o_multicast_hlim; 601 else 602 ip6->ip6_hlim = ip6_defmcasthlim; 603 } 604 605 #ifdef IPSEC 606 if (needipsec && needipsectun) { 607 struct ipsec_output_state state; 608 609 /* 610 * All the extension headers will become inaccessible 611 * (since they can be encrypted). 612 * Don't panic, we need no more updates to extension headers 613 * on inner IPv6 packet (since they are now encapsulated). 614 * 615 * IPv6 [ESP|AH] IPv6 [extension headers] payload 616 */ 617 bzero(&exthdrs, sizeof(exthdrs)); 618 exthdrs.ip6e_ip6 = m; 619 620 bzero(&state, sizeof(state)); 621 state.m = m; 622 state.ro = (struct route *)ro; 623 state.dst = (struct sockaddr *)dst; 624 625 error = ipsec6_output_tunnel(&state, sp, flags); 626 627 m = state.m; 628 ro_pmtu = ro = (struct route_in6 *)state.ro; 629 dst = (struct sockaddr_in6 *)state.dst; 630 if (error) { 631 /* mbuf is already reclaimed in ipsec6_output_tunnel. */ 632 m0 = m = NULL; 633 m = NULL; 634 switch (error) { 635 case EHOSTUNREACH: 636 case ENETUNREACH: 637 case EMSGSIZE: 638 case ENOBUFS: 639 case ENOMEM: 640 break; 641 default: 642 printf("ip6_output (ipsec): error code %d\n", error); 643 /* FALLTHROUGH */ 644 case ENOENT: 645 /* don't show these error codes to the user */ 646 error = 0; 647 break; 648 } 649 goto bad; 650 } 651 652 exthdrs.ip6e_ip6 = m; 653 } 654 #endif /* IPSEC */ 655 #ifdef FAST_IPSEC 656 if (needipsec) { 657 s = splsoftnet(); 658 error = ipsec6_process_packet(m,sp->req); 659 660 /* 661 * Preserve KAME behaviour: ENOENT can be returned 662 * when an SA acquire is in progress. Don't propagate 663 * this to user-level; it confuses applications. 664 * XXX this will go away when the SADB is redone. 665 */ 666 if (error == ENOENT) 667 error = 0; 668 splx(s); 669 goto done; 670 } 671 #endif /* FAST_IPSEC */ 672 673 674 675 /* adjust pointer */ 676 ip6 = mtod(m, struct ip6_hdr *); 677 678 bzero(&dst_sa, sizeof(dst_sa)); 679 dst_sa.sin6_family = AF_INET6; 680 dst_sa.sin6_len = sizeof(dst_sa); 681 dst_sa.sin6_addr = ip6->ip6_dst; 682 if ((error = in6_selectroute(&dst_sa, opt, im6o, ro, &ifp, &rt, 0)) 683 != 0) { 684 switch (error) { 685 case EHOSTUNREACH: 686 ip6stat.ip6s_noroute++; 687 break; 688 case EADDRNOTAVAIL: 689 default: 690 break; /* XXX statistics? */ 691 } 692 if (ifp != NULL) 693 in6_ifstat_inc(ifp, ifs6_out_discard); 694 goto bad; 695 } 696 if (rt == NULL) { 697 /* 698 * If in6_selectroute() does not return a route entry, 699 * dst may not have been updated. 700 */ 701 *dst = dst_sa; /* XXX */ 702 } 703 704 /* 705 * then rt (for unicast) and ifp must be non-NULL valid values. 706 */ 707 if ((flags & IPV6_FORWARDING) == 0) { 708 /* XXX: the FORWARDING flag can be set for mrouting. */ 709 in6_ifstat_inc(ifp, ifs6_out_request); 710 } 711 if (rt != NULL) { 712 ia = (struct in6_ifaddr *)(rt->rt_ifa); 713 rt->rt_use++; 714 } 715 716 /* 717 * The outgoing interface must be in the zone of source and 718 * destination addresses. We should use ia_ifp to support the 719 * case of sending packets to an address of our own. 720 */ 721 if (ia != NULL && ia->ia_ifp) 722 origifp = ia->ia_ifp; 723 else 724 origifp = ifp; 725 726 src0 = ip6->ip6_src; 727 if (in6_setscope(&src0, origifp, &zone)) 728 goto badscope; 729 bzero(&src_sa, sizeof(src_sa)); 730 src_sa.sin6_family = AF_INET6; 731 src_sa.sin6_len = sizeof(src_sa); 732 src_sa.sin6_addr = ip6->ip6_src; 733 if (sa6_recoverscope(&src_sa) || zone != src_sa.sin6_scope_id) 734 goto badscope; 735 736 dst0 = ip6->ip6_dst; 737 if (in6_setscope(&dst0, origifp, &zone)) 738 goto badscope; 739 /* re-initialize to be sure */ 740 bzero(&dst_sa, sizeof(dst_sa)); 741 dst_sa.sin6_family = AF_INET6; 742 dst_sa.sin6_len = sizeof(dst_sa); 743 dst_sa.sin6_addr = ip6->ip6_dst; 744 if (sa6_recoverscope(&dst_sa) || zone != dst_sa.sin6_scope_id) 745 goto badscope; 746 747 /* scope check is done. */ 748 goto routefound; 749 750 badscope: 751 ip6stat.ip6s_badscope++; 752 in6_ifstat_inc(origifp, ifs6_out_discard); 753 if (error == 0) 754 error = EHOSTUNREACH; /* XXX */ 755 goto bad; 756 757 routefound: 758 if (rt && !IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) { 759 if (opt && opt->ip6po_nextroute.ro_rt != NULL) { 760 /* 761 * The nexthop is explicitly specified by the 762 * application. We assume the next hop is an IPv6 763 * address. 764 */ 765 dst = (struct sockaddr_in6 *)opt->ip6po_nexthop; 766 } else if ((rt->rt_flags & RTF_GATEWAY)) 767 dst = (struct sockaddr_in6 *)rt->rt_gateway; 768 } 769 770 /* 771 * XXXXXX: original code follows: 772 */ 773 if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) 774 m->m_flags &= ~(M_BCAST | M_MCAST); /* just in case */ 775 else { 776 struct in6_multi *in6m; 777 778 m->m_flags = (m->m_flags & ~M_BCAST) | M_MCAST; 779 780 in6_ifstat_inc(ifp, ifs6_out_mcast); 781 782 /* 783 * Confirm that the outgoing interface supports multicast. 784 */ 785 if (!(ifp->if_flags & IFF_MULTICAST)) { 786 ip6stat.ip6s_noroute++; 787 in6_ifstat_inc(ifp, ifs6_out_discard); 788 error = ENETUNREACH; 789 goto bad; 790 } 791 792 IN6_LOOKUP_MULTI(ip6->ip6_dst, ifp, in6m); 793 if (in6m != NULL && 794 (im6o == NULL || im6o->im6o_multicast_loop)) { 795 /* 796 * If we belong to the destination multicast group 797 * on the outgoing interface, and the caller did not 798 * forbid loopback, loop back a copy. 799 */ 800 ip6_mloopback(ifp, m, dst); 801 } else { 802 /* 803 * If we are acting as a multicast router, perform 804 * multicast forwarding as if the packet had just 805 * arrived on the interface to which we are about 806 * to send. The multicast forwarding function 807 * recursively calls this function, using the 808 * IPV6_FORWARDING flag to prevent infinite recursion. 809 * 810 * Multicasts that are looped back by ip6_mloopback(), 811 * above, will be forwarded by the ip6_input() routine, 812 * if necessary. 813 */ 814 if (ip6_mrouter && (flags & IPV6_FORWARDING) == 0) { 815 if (ip6_mforward(ip6, ifp, m) != 0) { 816 m_freem(m); 817 goto done; 818 } 819 } 820 } 821 /* 822 * Multicasts with a hoplimit of zero may be looped back, 823 * above, but must not be transmitted on a network. 824 * Also, multicasts addressed to the loopback interface 825 * are not sent -- the above call to ip6_mloopback() will 826 * loop back a copy if this host actually belongs to the 827 * destination group on the loopback interface. 828 */ 829 if (ip6->ip6_hlim == 0 || (ifp->if_flags & IFF_LOOPBACK) || 830 IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst)) { 831 m_freem(m); 832 goto done; 833 } 834 } 835 836 /* 837 * Fill the outgoing inteface to tell the upper layer 838 * to increment per-interface statistics. 839 */ 840 if (ifpp) 841 *ifpp = ifp; 842 843 /* Determine path MTU. */ 844 if ((error = ip6_getpmtu(ro_pmtu, ro, ifp, &finaldst, &mtu, 845 &alwaysfrag)) != 0) 846 goto bad; 847 #ifdef IPSEC 848 if (needipsectun) 849 mtu = IPV6_MMTU; 850 #endif 851 852 /* 853 * The caller of this function may specify to use the minimum MTU 854 * in some cases. 855 * An advanced API option (IPV6_USE_MIN_MTU) can also override MTU 856 * setting. The logic is a bit complicated; by default, unicast 857 * packets will follow path MTU while multicast packets will be sent at 858 * the minimum MTU. If IP6PO_MINMTU_ALL is specified, all packets 859 * including unicast ones will be sent at the minimum MTU. Multicast 860 * packets will always be sent at the minimum MTU unless 861 * IP6PO_MINMTU_DISABLE is explicitly specified. 862 * See RFC 3542 for more details. 863 */ 864 if (mtu > IPV6_MMTU) { 865 if ((flags & IPV6_MINMTU)) 866 mtu = IPV6_MMTU; 867 else if (opt && opt->ip6po_minmtu == IP6PO_MINMTU_ALL) 868 mtu = IPV6_MMTU; 869 else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) && 870 (opt == NULL || 871 opt->ip6po_minmtu != IP6PO_MINMTU_DISABLE)) { 872 mtu = IPV6_MMTU; 873 } 874 } 875 876 /* 877 * clear embedded scope identifiers if necessary. 878 * in6_clearscope will touch the addresses only when necessary. 879 */ 880 in6_clearscope(&ip6->ip6_src); 881 in6_clearscope(&ip6->ip6_dst); 882 883 /* 884 * If the outgoing packet contains a hop-by-hop options header, 885 * it must be examined and processed even by the source node. 886 * (RFC 2460, section 4.) 887 */ 888 if (exthdrs.ip6e_hbh) { 889 struct ip6_hbh *hbh = mtod(exthdrs.ip6e_hbh, struct ip6_hbh *); 890 u_int32_t dummy1; /* XXX unused */ 891 u_int32_t dummy2; /* XXX unused */ 892 893 /* 894 * XXX: if we have to send an ICMPv6 error to the sender, 895 * we need the M_LOOP flag since icmp6_error() expects 896 * the IPv6 and the hop-by-hop options header are 897 * continuous unless the flag is set. 898 */ 899 m->m_flags |= M_LOOP; 900 m->m_pkthdr.rcvif = ifp; 901 if (ip6_process_hopopts(m, (u_int8_t *)(hbh + 1), 902 ((hbh->ip6h_len + 1) << 3) - sizeof(struct ip6_hbh), 903 &dummy1, &dummy2) < 0) { 904 /* m was already freed at this point */ 905 error = EINVAL;/* better error? */ 906 goto done; 907 } 908 m->m_flags &= ~M_LOOP; /* XXX */ 909 m->m_pkthdr.rcvif = NULL; 910 } 911 912 #ifdef PFIL_HOOKS 913 /* 914 * Run through list of hooks for output packets. 915 */ 916 if ((error = pfil_run_hooks(&inet6_pfil_hook, &m, ifp, PFIL_OUT)) != 0) 917 goto done; 918 if (m == NULL) 919 goto done; 920 ip6 = mtod(m, struct ip6_hdr *); 921 #endif /* PFIL_HOOKS */ 922 /* 923 * Send the packet to the outgoing interface. 924 * If necessary, do IPv6 fragmentation before sending. 925 * 926 * the logic here is rather complex: 927 * 1: normal case (dontfrag == 0, alwaysfrag == 0) 928 * 1-a: send as is if tlen <= path mtu 929 * 1-b: fragment if tlen > path mtu 930 * 931 * 2: if user asks us not to fragment (dontfrag == 1) 932 * 2-a: send as is if tlen <= interface mtu 933 * 2-b: error if tlen > interface mtu 934 * 935 * 3: if we always need to attach fragment header (alwaysfrag == 1) 936 * always fragment 937 * 938 * 4: if dontfrag == 1 && alwaysfrag == 1 939 * error, as we cannot handle this conflicting request 940 */ 941 tlen = m->m_pkthdr.len; 942 tso = (m->m_pkthdr.csum_flags & M_CSUM_TSOv6) != 0; 943 if (opt && (opt->ip6po_flags & IP6PO_DONTFRAG)) 944 dontfrag = 1; 945 else 946 dontfrag = 0; 947 948 if (dontfrag && alwaysfrag) { /* case 4 */ 949 /* conflicting request - can't transmit */ 950 error = EMSGSIZE; 951 goto bad; 952 } 953 if (dontfrag && (!tso && tlen > IN6_LINKMTU(ifp))) { /* case 2-b */ 954 /* 955 * Even if the DONTFRAG option is specified, we cannot send the 956 * packet when the data length is larger than the MTU of the 957 * outgoing interface. 958 * Notify the error by sending IPV6_PATHMTU ancillary data as 959 * well as returning an error code (the latter is not described 960 * in the API spec.) 961 */ 962 u_int32_t mtu32; 963 struct ip6ctlparam ip6cp; 964 965 mtu32 = (u_int32_t)mtu; 966 bzero(&ip6cp, sizeof(ip6cp)); 967 ip6cp.ip6c_cmdarg = (void *)&mtu32; 968 pfctlinput2(PRC_MSGSIZE, (struct sockaddr *)&ro_pmtu->ro_dst, 969 (void *)&ip6cp); 970 971 error = EMSGSIZE; 972 goto bad; 973 } 974 975 /* 976 * transmit packet without fragmentation 977 */ 978 if (dontfrag || (!alwaysfrag && (tlen <= mtu || tso))) { 979 /* case 1-a and 2-a */ 980 struct in6_ifaddr *ia6; 981 int sw_csum; 982 983 ip6 = mtod(m, struct ip6_hdr *); 984 ia6 = in6_ifawithifp(ifp, &ip6->ip6_src); 985 if (ia6) { 986 /* Record statistics for this interface address. */ 987 ia6->ia_ifa.ifa_data.ifad_outbytes += m->m_pkthdr.len; 988 } 989 #ifdef IPSEC 990 /* clean ipsec history once it goes out of the node */ 991 ipsec_delaux(m); 992 #endif 993 994 sw_csum = m->m_pkthdr.csum_flags & ~ifp->if_csum_flags_tx; 995 if ((sw_csum & (M_CSUM_UDPv6|M_CSUM_TCPv6)) != 0) { 996 if (IN6_NEED_CHECKSUM(ifp, 997 sw_csum & (M_CSUM_UDPv6|M_CSUM_TCPv6))) { 998 in6_delayed_cksum(m); 999 } 1000 m->m_pkthdr.csum_flags &= ~(M_CSUM_UDPv6|M_CSUM_TCPv6); 1001 } 1002 1003 if (__predict_true(!tso || 1004 (ifp->if_capenable & IFCAP_TSOv6) != 0)) { 1005 error = nd6_output(ifp, origifp, m, dst, rt); 1006 } else { 1007 error = ip6_tso_output(ifp, origifp, m, dst, rt); 1008 } 1009 goto done; 1010 } 1011 1012 if (tso) { 1013 error = EINVAL; /* XXX */ 1014 goto bad; 1015 } 1016 1017 /* 1018 * try to fragment the packet. case 1-b and 3 1019 */ 1020 if (mtu < IPV6_MMTU) { 1021 /* path MTU cannot be less than IPV6_MMTU */ 1022 error = EMSGSIZE; 1023 in6_ifstat_inc(ifp, ifs6_out_fragfail); 1024 goto bad; 1025 } else if (ip6->ip6_plen == 0) { 1026 /* jumbo payload cannot be fragmented */ 1027 error = EMSGSIZE; 1028 in6_ifstat_inc(ifp, ifs6_out_fragfail); 1029 goto bad; 1030 } else { 1031 struct mbuf **mnext, *m_frgpart; 1032 struct ip6_frag *ip6f; 1033 u_int32_t id = htonl(ip6_randomid()); 1034 u_char nextproto; 1035 #if 0 /* see below */ 1036 struct ip6ctlparam ip6cp; 1037 u_int32_t mtu32; 1038 #endif 1039 1040 /* 1041 * Too large for the destination or interface; 1042 * fragment if possible. 1043 * Must be able to put at least 8 bytes per fragment. 1044 */ 1045 hlen = unfragpartlen; 1046 if (mtu > IPV6_MAXPACKET) 1047 mtu = IPV6_MAXPACKET; 1048 1049 #if 0 1050 /* 1051 * It is believed this code is a leftover from the 1052 * development of the IPV6_RECVPATHMTU sockopt and 1053 * associated work to implement RFC3542. 1054 * It's not entirely clear what the intent of the API 1055 * is at this point, so disable this code for now. 1056 * The IPV6_RECVPATHMTU sockopt and/or IPV6_DONTFRAG 1057 * will send notifications if the application requests. 1058 */ 1059 1060 /* Notify a proper path MTU to applications. */ 1061 mtu32 = (u_int32_t)mtu; 1062 bzero(&ip6cp, sizeof(ip6cp)); 1063 ip6cp.ip6c_cmdarg = (void *)&mtu32; 1064 pfctlinput2(PRC_MSGSIZE, (struct sockaddr *)&ro_pmtu->ro_dst, 1065 (void *)&ip6cp); 1066 #endif 1067 1068 len = (mtu - hlen - sizeof(struct ip6_frag)) & ~7; 1069 if (len < 8) { 1070 error = EMSGSIZE; 1071 in6_ifstat_inc(ifp, ifs6_out_fragfail); 1072 goto bad; 1073 } 1074 1075 mnext = &m->m_nextpkt; 1076 1077 /* 1078 * Change the next header field of the last header in the 1079 * unfragmentable part. 1080 */ 1081 if (exthdrs.ip6e_rthdr) { 1082 nextproto = *mtod(exthdrs.ip6e_rthdr, u_char *); 1083 *mtod(exthdrs.ip6e_rthdr, u_char *) = IPPROTO_FRAGMENT; 1084 } else if (exthdrs.ip6e_dest1) { 1085 nextproto = *mtod(exthdrs.ip6e_dest1, u_char *); 1086 *mtod(exthdrs.ip6e_dest1, u_char *) = IPPROTO_FRAGMENT; 1087 } else if (exthdrs.ip6e_hbh) { 1088 nextproto = *mtod(exthdrs.ip6e_hbh, u_char *); 1089 *mtod(exthdrs.ip6e_hbh, u_char *) = IPPROTO_FRAGMENT; 1090 } else { 1091 nextproto = ip6->ip6_nxt; 1092 ip6->ip6_nxt = IPPROTO_FRAGMENT; 1093 } 1094 1095 if ((m->m_pkthdr.csum_flags & (M_CSUM_UDPv6|M_CSUM_TCPv6)) 1096 != 0) { 1097 if (IN6_NEED_CHECKSUM(ifp, 1098 m->m_pkthdr.csum_flags & 1099 (M_CSUM_UDPv6|M_CSUM_TCPv6))) { 1100 in6_delayed_cksum(m); 1101 } 1102 m->m_pkthdr.csum_flags &= ~(M_CSUM_UDPv6|M_CSUM_TCPv6); 1103 } 1104 1105 /* 1106 * Loop through length of segment after first fragment, 1107 * make new header and copy data of each part and link onto 1108 * chain. 1109 */ 1110 m0 = m; 1111 for (off = hlen; off < tlen; off += len) { 1112 struct mbuf *mlast; 1113 1114 MGETHDR(m, M_DONTWAIT, MT_HEADER); 1115 if (!m) { 1116 error = ENOBUFS; 1117 ip6stat.ip6s_odropped++; 1118 goto sendorfree; 1119 } 1120 m->m_pkthdr.rcvif = NULL; 1121 m->m_flags = m0->m_flags & M_COPYFLAGS; 1122 *mnext = m; 1123 mnext = &m->m_nextpkt; 1124 m->m_data += max_linkhdr; 1125 mhip6 = mtod(m, struct ip6_hdr *); 1126 *mhip6 = *ip6; 1127 m->m_len = sizeof(*mhip6); 1128 error = ip6_insertfraghdr(m0, m, hlen, &ip6f); 1129 if (error) { 1130 ip6stat.ip6s_odropped++; 1131 goto sendorfree; 1132 } 1133 ip6f->ip6f_offlg = htons((u_int16_t)((off - hlen) & ~7)); 1134 if (off + len >= tlen) 1135 len = tlen - off; 1136 else 1137 ip6f->ip6f_offlg |= IP6F_MORE_FRAG; 1138 mhip6->ip6_plen = htons((u_int16_t)(len + hlen + 1139 sizeof(*ip6f) - sizeof(struct ip6_hdr))); 1140 if ((m_frgpart = m_copy(m0, off, len)) == 0) { 1141 error = ENOBUFS; 1142 ip6stat.ip6s_odropped++; 1143 goto sendorfree; 1144 } 1145 for (mlast = m; mlast->m_next; mlast = mlast->m_next) 1146 ; 1147 mlast->m_next = m_frgpart; 1148 m->m_pkthdr.len = len + hlen + sizeof(*ip6f); 1149 m->m_pkthdr.rcvif = (struct ifnet *)0; 1150 ip6f->ip6f_reserved = 0; 1151 ip6f->ip6f_ident = id; 1152 ip6f->ip6f_nxt = nextproto; 1153 ip6stat.ip6s_ofragments++; 1154 in6_ifstat_inc(ifp, ifs6_out_fragcreat); 1155 } 1156 1157 in6_ifstat_inc(ifp, ifs6_out_fragok); 1158 } 1159 1160 /* 1161 * Remove leading garbages. 1162 */ 1163 sendorfree: 1164 m = m0->m_nextpkt; 1165 m0->m_nextpkt = 0; 1166 m_freem(m0); 1167 for (m0 = m; m; m = m0) { 1168 m0 = m->m_nextpkt; 1169 m->m_nextpkt = 0; 1170 if (error == 0) { 1171 struct in6_ifaddr *ia6; 1172 ip6 = mtod(m, struct ip6_hdr *); 1173 ia6 = in6_ifawithifp(ifp, &ip6->ip6_src); 1174 if (ia6) { 1175 /* 1176 * Record statistics for this interface 1177 * address. 1178 */ 1179 ia6->ia_ifa.ifa_data.ifad_outbytes += 1180 m->m_pkthdr.len; 1181 } 1182 #ifdef IPSEC 1183 /* clean ipsec history once it goes out of the node */ 1184 ipsec_delaux(m); 1185 #endif 1186 error = nd6_output(ifp, origifp, m, dst, rt); 1187 } else 1188 m_freem(m); 1189 } 1190 1191 if (error == 0) 1192 ip6stat.ip6s_fragmented++; 1193 1194 done: 1195 /* XXX Second if is invariant? */ 1196 if (ro == &ip6route) 1197 rtcache_free((struct route *)ro); 1198 else if (ro_pmtu == &ip6route) 1199 rtcache_free((struct route *)ro_pmtu); 1200 1201 #ifdef IPSEC 1202 if (sp != NULL) 1203 key_freesp(sp); 1204 #endif /* IPSEC */ 1205 #ifdef FAST_IPSEC 1206 if (sp != NULL) 1207 KEY_FREESP(&sp); 1208 #endif /* FAST_IPSEC */ 1209 1210 1211 return (error); 1212 1213 freehdrs: 1214 m_freem(exthdrs.ip6e_hbh); /* m_freem will check if mbuf is 0 */ 1215 m_freem(exthdrs.ip6e_dest1); 1216 m_freem(exthdrs.ip6e_rthdr); 1217 m_freem(exthdrs.ip6e_dest2); 1218 /* FALLTHROUGH */ 1219 bad: 1220 m_freem(m); 1221 goto done; 1222 } 1223 1224 static int 1225 ip6_copyexthdr(mp, hdr, hlen) 1226 struct mbuf **mp; 1227 caddr_t hdr; 1228 int hlen; 1229 { 1230 struct mbuf *m; 1231 1232 if (hlen > MCLBYTES) 1233 return (ENOBUFS); /* XXX */ 1234 1235 MGET(m, M_DONTWAIT, MT_DATA); 1236 if (!m) 1237 return (ENOBUFS); 1238 1239 if (hlen > MLEN) { 1240 MCLGET(m, M_DONTWAIT); 1241 if ((m->m_flags & M_EXT) == 0) { 1242 m_free(m); 1243 return (ENOBUFS); 1244 } 1245 } 1246 m->m_len = hlen; 1247 if (hdr) 1248 bcopy(hdr, mtod(m, caddr_t), hlen); 1249 1250 *mp = m; 1251 return (0); 1252 } 1253 1254 /* 1255 * Process a delayed payload checksum calculation. 1256 */ 1257 void 1258 in6_delayed_cksum(struct mbuf *m) 1259 { 1260 uint16_t csum, offset; 1261 1262 KASSERT((m->m_pkthdr.csum_flags & (M_CSUM_UDPv6|M_CSUM_TCPv6)) != 0); 1263 KASSERT((~m->m_pkthdr.csum_flags & (M_CSUM_UDPv6|M_CSUM_TCPv6)) != 0); 1264 KASSERT((m->m_pkthdr.csum_flags 1265 & (M_CSUM_UDPv4|M_CSUM_TCPv4|M_CSUM_TSOv4)) == 0); 1266 1267 offset = M_CSUM_DATA_IPv6_HL(m->m_pkthdr.csum_data); 1268 csum = in6_cksum(m, 0, offset, m->m_pkthdr.len - offset); 1269 if (csum == 0 && (m->m_pkthdr.csum_flags & M_CSUM_UDPv6) != 0) { 1270 csum = 0xffff; 1271 } 1272 1273 offset += M_CSUM_DATA_IPv6_OFFSET(m->m_pkthdr.csum_data); 1274 if ((offset + sizeof(csum)) > m->m_len) { 1275 m_copyback(m, offset, sizeof(csum), &csum); 1276 } else { 1277 *(uint16_t *)(mtod(m, caddr_t) + offset) = csum; 1278 } 1279 } 1280 1281 /* 1282 * Insert jumbo payload option. 1283 */ 1284 static int 1285 ip6_insert_jumboopt(exthdrs, plen) 1286 struct ip6_exthdrs *exthdrs; 1287 u_int32_t plen; 1288 { 1289 struct mbuf *mopt; 1290 u_int8_t *optbuf; 1291 u_int32_t v; 1292 1293 #define JUMBOOPTLEN 8 /* length of jumbo payload option and padding */ 1294 1295 /* 1296 * If there is no hop-by-hop options header, allocate new one. 1297 * If there is one but it doesn't have enough space to store the 1298 * jumbo payload option, allocate a cluster to store the whole options. 1299 * Otherwise, use it to store the options. 1300 */ 1301 if (exthdrs->ip6e_hbh == 0) { 1302 MGET(mopt, M_DONTWAIT, MT_DATA); 1303 if (mopt == 0) 1304 return (ENOBUFS); 1305 mopt->m_len = JUMBOOPTLEN; 1306 optbuf = mtod(mopt, u_int8_t *); 1307 optbuf[1] = 0; /* = ((JUMBOOPTLEN) >> 3) - 1 */ 1308 exthdrs->ip6e_hbh = mopt; 1309 } else { 1310 struct ip6_hbh *hbh; 1311 1312 mopt = exthdrs->ip6e_hbh; 1313 if (M_TRAILINGSPACE(mopt) < JUMBOOPTLEN) { 1314 /* 1315 * XXX assumption: 1316 * - exthdrs->ip6e_hbh is not referenced from places 1317 * other than exthdrs. 1318 * - exthdrs->ip6e_hbh is not an mbuf chain. 1319 */ 1320 int oldoptlen = mopt->m_len; 1321 struct mbuf *n; 1322 1323 /* 1324 * XXX: give up if the whole (new) hbh header does 1325 * not fit even in an mbuf cluster. 1326 */ 1327 if (oldoptlen + JUMBOOPTLEN > MCLBYTES) 1328 return (ENOBUFS); 1329 1330 /* 1331 * As a consequence, we must always prepare a cluster 1332 * at this point. 1333 */ 1334 MGET(n, M_DONTWAIT, MT_DATA); 1335 if (n) { 1336 MCLGET(n, M_DONTWAIT); 1337 if ((n->m_flags & M_EXT) == 0) { 1338 m_freem(n); 1339 n = NULL; 1340 } 1341 } 1342 if (!n) 1343 return (ENOBUFS); 1344 n->m_len = oldoptlen + JUMBOOPTLEN; 1345 bcopy(mtod(mopt, caddr_t), mtod(n, caddr_t), 1346 oldoptlen); 1347 optbuf = mtod(n, u_int8_t *) + oldoptlen; 1348 m_freem(mopt); 1349 mopt = exthdrs->ip6e_hbh = n; 1350 } else { 1351 optbuf = mtod(mopt, u_int8_t *) + mopt->m_len; 1352 mopt->m_len += JUMBOOPTLEN; 1353 } 1354 optbuf[0] = IP6OPT_PADN; 1355 optbuf[1] = 0; 1356 1357 /* 1358 * Adjust the header length according to the pad and 1359 * the jumbo payload option. 1360 */ 1361 hbh = mtod(mopt, struct ip6_hbh *); 1362 hbh->ip6h_len += (JUMBOOPTLEN >> 3); 1363 } 1364 1365 /* fill in the option. */ 1366 optbuf[2] = IP6OPT_JUMBO; 1367 optbuf[3] = 4; 1368 v = (u_int32_t)htonl(plen + JUMBOOPTLEN); 1369 bcopy(&v, &optbuf[4], sizeof(u_int32_t)); 1370 1371 /* finally, adjust the packet header length */ 1372 exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN; 1373 1374 return (0); 1375 #undef JUMBOOPTLEN 1376 } 1377 1378 /* 1379 * Insert fragment header and copy unfragmentable header portions. 1380 */ 1381 static int 1382 ip6_insertfraghdr(m0, m, hlen, frghdrp) 1383 struct mbuf *m0, *m; 1384 int hlen; 1385 struct ip6_frag **frghdrp; 1386 { 1387 struct mbuf *n, *mlast; 1388 1389 if (hlen > sizeof(struct ip6_hdr)) { 1390 n = m_copym(m0, sizeof(struct ip6_hdr), 1391 hlen - sizeof(struct ip6_hdr), M_DONTWAIT); 1392 if (n == 0) 1393 return (ENOBUFS); 1394 m->m_next = n; 1395 } else 1396 n = m; 1397 1398 /* Search for the last mbuf of unfragmentable part. */ 1399 for (mlast = n; mlast->m_next; mlast = mlast->m_next) 1400 ; 1401 1402 if ((mlast->m_flags & M_EXT) == 0 && 1403 M_TRAILINGSPACE(mlast) >= sizeof(struct ip6_frag)) { 1404 /* use the trailing space of the last mbuf for the fragment hdr */ 1405 *frghdrp = (struct ip6_frag *)(mtod(mlast, caddr_t) + 1406 mlast->m_len); 1407 mlast->m_len += sizeof(struct ip6_frag); 1408 m->m_pkthdr.len += sizeof(struct ip6_frag); 1409 } else { 1410 /* allocate a new mbuf for the fragment header */ 1411 struct mbuf *mfrg; 1412 1413 MGET(mfrg, M_DONTWAIT, MT_DATA); 1414 if (mfrg == 0) 1415 return (ENOBUFS); 1416 mfrg->m_len = sizeof(struct ip6_frag); 1417 *frghdrp = mtod(mfrg, struct ip6_frag *); 1418 mlast->m_next = mfrg; 1419 } 1420 1421 return (0); 1422 } 1423 1424 static int 1425 ip6_getpmtu(ro_pmtu, ro, ifp, dst, mtup, alwaysfragp) 1426 struct route_in6 *ro_pmtu, *ro; 1427 struct ifnet *ifp; 1428 struct in6_addr *dst; 1429 u_long *mtup; 1430 int *alwaysfragp; 1431 { 1432 u_int32_t mtu = 0; 1433 int alwaysfrag = 0; 1434 int error = 0; 1435 1436 if (ro_pmtu != ro) { 1437 /* The first hop and the final destination may differ. */ 1438 struct sockaddr_in6 *sa6_dst = 1439 (struct sockaddr_in6 *)&ro_pmtu->ro_dst; 1440 if (!IN6_ARE_ADDR_EQUAL(&sa6_dst->sin6_addr, dst)) 1441 rtcache_free((struct route *)ro_pmtu); 1442 else 1443 rtcache_check((struct route *)ro_pmtu); 1444 if (ro_pmtu->ro_rt == NULL) { 1445 memset(sa6_dst, 0, sizeof(*sa6_dst)); /* for safety */ 1446 sa6_dst->sin6_family = AF_INET6; 1447 sa6_dst->sin6_len = sizeof(struct sockaddr_in6); 1448 sa6_dst->sin6_addr = *dst; 1449 rtcache_init((struct route *)ro_pmtu); 1450 } 1451 } 1452 if (ro_pmtu->ro_rt != NULL) { 1453 u_int32_t ifmtu; 1454 1455 if (ifp == NULL) 1456 ifp = ro_pmtu->ro_rt->rt_ifp; 1457 ifmtu = IN6_LINKMTU(ifp); 1458 mtu = ro_pmtu->ro_rt->rt_rmx.rmx_mtu; 1459 if (mtu == 0) 1460 mtu = ifmtu; 1461 else if (mtu < IPV6_MMTU) { 1462 /* 1463 * RFC2460 section 5, last paragraph: 1464 * if we record ICMPv6 too big message with 1465 * mtu < IPV6_MMTU, transmit packets sized IPV6_MMTU 1466 * or smaller, with fragment header attached. 1467 * (fragment header is needed regardless from the 1468 * packet size, for translators to identify packets) 1469 */ 1470 alwaysfrag = 1; 1471 mtu = IPV6_MMTU; 1472 } else if (mtu > ifmtu) { 1473 /* 1474 * The MTU on the route is larger than the MTU on 1475 * the interface! This shouldn't happen, unless the 1476 * MTU of the interface has been changed after the 1477 * interface was brought up. Change the MTU in the 1478 * route to match the interface MTU (as long as the 1479 * field isn't locked). 1480 */ 1481 mtu = ifmtu; 1482 if (!(ro_pmtu->ro_rt->rt_rmx.rmx_locks & RTV_MTU)) 1483 ro_pmtu->ro_rt->rt_rmx.rmx_mtu = mtu; 1484 } 1485 } else if (ifp) { 1486 mtu = IN6_LINKMTU(ifp); 1487 } else 1488 error = EHOSTUNREACH; /* XXX */ 1489 1490 *mtup = mtu; 1491 if (alwaysfragp) 1492 *alwaysfragp = alwaysfrag; 1493 return (error); 1494 } 1495 1496 /* 1497 * IP6 socket option processing. 1498 */ 1499 int 1500 ip6_ctloutput(op, so, level, optname, mp) 1501 int op; 1502 struct socket *so; 1503 int level, optname; 1504 struct mbuf **mp; 1505 { 1506 int privileged, optdatalen, uproto; 1507 void *optdata; 1508 struct in6pcb *in6p = sotoin6pcb(so); 1509 struct mbuf *m = *mp; 1510 int error, optval; 1511 int optlen; 1512 struct lwp *l = curlwp; /* XXX */ 1513 1514 optlen = m ? m->m_len : 0; 1515 error = optval = 0; 1516 privileged = (l == 0 || kauth_authorize_generic(l->l_cred, 1517 KAUTH_GENERIC_ISSUSER, NULL)) ? 0 : 1; 1518 uproto = (int)so->so_proto->pr_protocol; 1519 1520 if (level == IPPROTO_IPV6) { 1521 switch (op) { 1522 case PRCO_SETOPT: 1523 switch (optname) { 1524 #ifdef RFC2292 1525 case IPV6_2292PKTOPTIONS: 1526 /* m is freed in ip6_pcbopts */ 1527 error = ip6_pcbopts(&in6p->in6p_outputopts, 1528 m, so); 1529 break; 1530 #endif 1531 1532 /* 1533 * Use of some Hop-by-Hop options or some 1534 * Destination options, might require special 1535 * privilege. That is, normal applications 1536 * (without special privilege) might be forbidden 1537 * from setting certain options in outgoing packets, 1538 * and might never see certain options in received 1539 * packets. [RFC 2292 Section 6] 1540 * KAME specific note: 1541 * KAME prevents non-privileged users from sending or 1542 * receiving ANY hbh/dst options in order to avoid 1543 * overhead of parsing options in the kernel. 1544 */ 1545 case IPV6_RECVHOPOPTS: 1546 case IPV6_RECVDSTOPTS: 1547 case IPV6_RECVRTHDRDSTOPTS: 1548 if (!privileged) { 1549 error = EPERM; 1550 break; 1551 } 1552 /* FALLTHROUGH */ 1553 case IPV6_UNICAST_HOPS: 1554 case IPV6_HOPLIMIT: 1555 case IPV6_FAITH: 1556 1557 case IPV6_RECVPKTINFO: 1558 case IPV6_RECVHOPLIMIT: 1559 case IPV6_RECVRTHDR: 1560 case IPV6_RECVPATHMTU: 1561 case IPV6_RECVTCLASS: 1562 case IPV6_V6ONLY: 1563 if (optlen != sizeof(int)) { 1564 error = EINVAL; 1565 break; 1566 } 1567 optval = *mtod(m, int *); 1568 switch (optname) { 1569 1570 case IPV6_UNICAST_HOPS: 1571 if (optval < -1 || optval >= 256) 1572 error = EINVAL; 1573 else { 1574 /* -1 = kernel default */ 1575 in6p->in6p_hops = optval; 1576 } 1577 break; 1578 #define OPTSET(bit) \ 1579 do { \ 1580 if (optval) \ 1581 in6p->in6p_flags |= (bit); \ 1582 else \ 1583 in6p->in6p_flags &= ~(bit); \ 1584 } while (/*CONSTCOND*/ 0) 1585 1586 #ifdef RFC2292 1587 #define OPTSET2292(bit) \ 1588 do { \ 1589 in6p->in6p_flags |= IN6P_RFC2292; \ 1590 if (optval) \ 1591 in6p->in6p_flags |= (bit); \ 1592 else \ 1593 in6p->in6p_flags &= ~(bit); \ 1594 } while (/*CONSTCOND*/ 0) 1595 #endif 1596 1597 #define OPTBIT(bit) (in6p->in6p_flags & (bit) ? 1 : 0) 1598 1599 case IPV6_RECVPKTINFO: 1600 #ifdef RFC2292 1601 /* cannot mix with RFC2292 */ 1602 if (OPTBIT(IN6P_RFC2292)) { 1603 error = EINVAL; 1604 break; 1605 } 1606 #endif 1607 OPTSET(IN6P_PKTINFO); 1608 break; 1609 1610 case IPV6_HOPLIMIT: 1611 { 1612 struct ip6_pktopts **optp; 1613 1614 #ifdef RFC2292 1615 /* cannot mix with RFC2292 */ 1616 if (OPTBIT(IN6P_RFC2292)) { 1617 error = EINVAL; 1618 break; 1619 } 1620 #endif 1621 optp = &in6p->in6p_outputopts; 1622 error = ip6_pcbopt(IPV6_HOPLIMIT, 1623 (u_char *)&optval, 1624 sizeof(optval), 1625 optp, 1626 privileged, uproto); 1627 break; 1628 } 1629 1630 case IPV6_RECVHOPLIMIT: 1631 #ifdef RFC2292 1632 /* cannot mix with RFC2292 */ 1633 if (OPTBIT(IN6P_RFC2292)) { 1634 error = EINVAL; 1635 break; 1636 } 1637 #endif 1638 OPTSET(IN6P_HOPLIMIT); 1639 break; 1640 1641 case IPV6_RECVHOPOPTS: 1642 #ifdef RFC2292 1643 /* cannot mix with RFC2292 */ 1644 if (OPTBIT(IN6P_RFC2292)) { 1645 error = EINVAL; 1646 break; 1647 } 1648 #endif 1649 OPTSET(IN6P_HOPOPTS); 1650 break; 1651 1652 case IPV6_RECVDSTOPTS: 1653 #ifdef RFC2292 1654 /* cannot mix with RFC2292 */ 1655 if (OPTBIT(IN6P_RFC2292)) { 1656 error = EINVAL; 1657 break; 1658 } 1659 #endif 1660 OPTSET(IN6P_DSTOPTS); 1661 break; 1662 1663 case IPV6_RECVRTHDRDSTOPTS: 1664 #ifdef RFC2292 1665 /* cannot mix with RFC2292 */ 1666 if (OPTBIT(IN6P_RFC2292)) { 1667 error = EINVAL; 1668 break; 1669 } 1670 #endif 1671 OPTSET(IN6P_RTHDRDSTOPTS); 1672 break; 1673 1674 case IPV6_RECVRTHDR: 1675 #ifdef RFC2292 1676 /* cannot mix with RFC2292 */ 1677 if (OPTBIT(IN6P_RFC2292)) { 1678 error = EINVAL; 1679 break; 1680 } 1681 #endif 1682 OPTSET(IN6P_RTHDR); 1683 break; 1684 1685 case IPV6_FAITH: 1686 OPTSET(IN6P_FAITH); 1687 break; 1688 1689 case IPV6_RECVPATHMTU: 1690 /* 1691 * We ignore this option for TCP 1692 * sockets. 1693 * (RFC3542 leaves this case 1694 * unspecified.) 1695 */ 1696 if (uproto != IPPROTO_TCP) 1697 OPTSET(IN6P_MTU); 1698 break; 1699 1700 case IPV6_V6ONLY: 1701 /* 1702 * make setsockopt(IPV6_V6ONLY) 1703 * available only prior to bind(2). 1704 * see ipng mailing list, Jun 22 2001. 1705 */ 1706 if (in6p->in6p_lport || 1707 !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr)) { 1708 error = EINVAL; 1709 break; 1710 } 1711 #ifdef INET6_BINDV6ONLY 1712 if (!optval) 1713 error = EINVAL; 1714 #else 1715 OPTSET(IN6P_IPV6_V6ONLY); 1716 #endif 1717 break; 1718 case IPV6_RECVTCLASS: 1719 #ifdef RFC2292 1720 /* cannot mix with RFC2292 XXX */ 1721 if (OPTBIT(IN6P_RFC2292)) { 1722 error = EINVAL; 1723 break; 1724 } 1725 #endif 1726 OPTSET(IN6P_TCLASS); 1727 break; 1728 1729 } 1730 break; 1731 1732 case IPV6_OTCLASS: 1733 { 1734 struct ip6_pktopts **optp; 1735 u_int8_t tclass; 1736 1737 if (optlen != sizeof(tclass)) { 1738 error = EINVAL; 1739 break; 1740 } 1741 tclass = *mtod(m, u_int8_t *); 1742 optp = &in6p->in6p_outputopts; 1743 error = ip6_pcbopt(optname, 1744 (u_char *)&tclass, 1745 sizeof(tclass), 1746 optp, 1747 privileged, uproto); 1748 break; 1749 } 1750 1751 case IPV6_TCLASS: 1752 case IPV6_DONTFRAG: 1753 case IPV6_USE_MIN_MTU: 1754 if (optlen != sizeof(optval)) { 1755 error = EINVAL; 1756 break; 1757 } 1758 optval = *mtod(m, int *); 1759 { 1760 struct ip6_pktopts **optp; 1761 optp = &in6p->in6p_outputopts; 1762 error = ip6_pcbopt(optname, 1763 (u_char *)&optval, 1764 sizeof(optval), 1765 optp, 1766 privileged, uproto); 1767 break; 1768 } 1769 1770 #ifdef RFC2292 1771 case IPV6_2292PKTINFO: 1772 case IPV6_2292HOPLIMIT: 1773 case IPV6_2292HOPOPTS: 1774 case IPV6_2292DSTOPTS: 1775 case IPV6_2292RTHDR: 1776 /* RFC 2292 */ 1777 if (optlen != sizeof(int)) { 1778 error = EINVAL; 1779 break; 1780 } 1781 optval = *mtod(m, int *); 1782 switch (optname) { 1783 case IPV6_2292PKTINFO: 1784 OPTSET2292(IN6P_PKTINFO); 1785 break; 1786 case IPV6_2292HOPLIMIT: 1787 OPTSET2292(IN6P_HOPLIMIT); 1788 break; 1789 case IPV6_2292HOPOPTS: 1790 /* 1791 * Check super-user privilege. 1792 * See comments for IPV6_RECVHOPOPTS. 1793 */ 1794 if (!privileged) 1795 return (EPERM); 1796 OPTSET2292(IN6P_HOPOPTS); 1797 break; 1798 case IPV6_2292DSTOPTS: 1799 if (!privileged) 1800 return (EPERM); 1801 OPTSET2292(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); /* XXX */ 1802 break; 1803 case IPV6_2292RTHDR: 1804 OPTSET2292(IN6P_RTHDR); 1805 break; 1806 } 1807 break; 1808 #endif 1809 case IPV6_PKTINFO: 1810 case IPV6_HOPOPTS: 1811 case IPV6_RTHDR: 1812 case IPV6_DSTOPTS: 1813 case IPV6_RTHDRDSTOPTS: 1814 case IPV6_NEXTHOP: 1815 { 1816 /* new advanced API (RFC3542) */ 1817 u_char *optbuf; 1818 int optbuflen; 1819 struct ip6_pktopts **optp; 1820 if (!m) { 1821 error = EINVAL; 1822 break; 1823 } 1824 1825 #ifdef RFC2292 1826 /* cannot mix with RFC2292 */ 1827 if (OPTBIT(IN6P_RFC2292)) { 1828 error = EINVAL; 1829 break; 1830 } 1831 #endif 1832 1833 if (m && m->m_next) { 1834 error = EINVAL; /* XXX */ 1835 break; 1836 } 1837 1838 optbuf = mtod(m, u_char *); 1839 optbuflen = m->m_len; 1840 optp = &in6p->in6p_outputopts; 1841 error = ip6_pcbopt(optname, optbuf, optbuflen, 1842 optp, privileged, uproto); 1843 break; 1844 } 1845 #undef OPTSET 1846 1847 case IPV6_MULTICAST_IF: 1848 case IPV6_MULTICAST_HOPS: 1849 case IPV6_MULTICAST_LOOP: 1850 case IPV6_JOIN_GROUP: 1851 case IPV6_LEAVE_GROUP: 1852 error = ip6_setmoptions(optname, 1853 &in6p->in6p_moptions, m); 1854 break; 1855 1856 case IPV6_PORTRANGE: 1857 if (!m) { 1858 error = EINVAL; 1859 break; 1860 } 1861 optval = *mtod(m, int *); 1862 1863 switch (optval) { 1864 case IPV6_PORTRANGE_DEFAULT: 1865 in6p->in6p_flags &= ~(IN6P_LOWPORT); 1866 in6p->in6p_flags &= ~(IN6P_HIGHPORT); 1867 break; 1868 1869 case IPV6_PORTRANGE_HIGH: 1870 in6p->in6p_flags &= ~(IN6P_LOWPORT); 1871 in6p->in6p_flags |= IN6P_HIGHPORT; 1872 break; 1873 1874 case IPV6_PORTRANGE_LOW: 1875 in6p->in6p_flags &= ~(IN6P_HIGHPORT); 1876 in6p->in6p_flags |= IN6P_LOWPORT; 1877 break; 1878 1879 default: 1880 error = EINVAL; 1881 break; 1882 } 1883 break; 1884 1885 1886 #if defined(IPSEC) || defined(FAST_IPSEC) 1887 case IPV6_IPSEC_POLICY: 1888 { 1889 caddr_t req = NULL; 1890 size_t len = 0; 1891 if (m) { 1892 req = mtod(m, caddr_t); 1893 len = m->m_len; 1894 } 1895 error = ipsec6_set_policy(in6p, optname, req, 1896 len, privileged); 1897 } 1898 break; 1899 #endif /* IPSEC */ 1900 1901 default: 1902 error = ENOPROTOOPT; 1903 break; 1904 } 1905 if (m) 1906 (void)m_free(m); 1907 break; 1908 1909 case PRCO_GETOPT: 1910 switch (optname) { 1911 #ifdef RFC2292 1912 case IPV6_2292PKTOPTIONS: 1913 /* 1914 * RFC3542 (effectively) deprecated the 1915 * semantics of the 2292-style pktoptions. 1916 * Since it was not reliable in nature (i.e., 1917 * applications had to expect the lack of some 1918 * information after all), it would make sense 1919 * to simplify this part by always returning 1920 * empty data. 1921 */ 1922 *mp = m_get(M_WAIT, MT_SOOPTS); 1923 (*mp)->m_len = 0; 1924 break; 1925 #endif 1926 1927 case IPV6_RECVHOPOPTS: 1928 case IPV6_RECVDSTOPTS: 1929 case IPV6_RECVRTHDRDSTOPTS: 1930 case IPV6_UNICAST_HOPS: 1931 case IPV6_RECVPKTINFO: 1932 case IPV6_RECVHOPLIMIT: 1933 case IPV6_RECVRTHDR: 1934 case IPV6_RECVPATHMTU: 1935 1936 case IPV6_FAITH: 1937 case IPV6_V6ONLY: 1938 case IPV6_PORTRANGE: 1939 case IPV6_RECVTCLASS: 1940 switch (optname) { 1941 1942 case IPV6_RECVHOPOPTS: 1943 optval = OPTBIT(IN6P_HOPOPTS); 1944 break; 1945 1946 case IPV6_RECVDSTOPTS: 1947 optval = OPTBIT(IN6P_DSTOPTS); 1948 break; 1949 1950 case IPV6_RECVRTHDRDSTOPTS: 1951 optval = OPTBIT(IN6P_RTHDRDSTOPTS); 1952 break; 1953 1954 case IPV6_UNICAST_HOPS: 1955 optval = in6p->in6p_hops; 1956 break; 1957 1958 case IPV6_RECVPKTINFO: 1959 optval = OPTBIT(IN6P_PKTINFO); 1960 break; 1961 1962 case IPV6_RECVHOPLIMIT: 1963 optval = OPTBIT(IN6P_HOPLIMIT); 1964 break; 1965 1966 case IPV6_RECVRTHDR: 1967 optval = OPTBIT(IN6P_RTHDR); 1968 break; 1969 1970 case IPV6_RECVPATHMTU: 1971 optval = OPTBIT(IN6P_MTU); 1972 break; 1973 1974 case IPV6_FAITH: 1975 optval = OPTBIT(IN6P_FAITH); 1976 break; 1977 1978 case IPV6_V6ONLY: 1979 optval = OPTBIT(IN6P_IPV6_V6ONLY); 1980 break; 1981 1982 case IPV6_PORTRANGE: 1983 { 1984 int flags; 1985 flags = in6p->in6p_flags; 1986 if (flags & IN6P_HIGHPORT) 1987 optval = IPV6_PORTRANGE_HIGH; 1988 else if (flags & IN6P_LOWPORT) 1989 optval = IPV6_PORTRANGE_LOW; 1990 else 1991 optval = 0; 1992 break; 1993 } 1994 case IPV6_RECVTCLASS: 1995 optval = OPTBIT(IN6P_TCLASS); 1996 break; 1997 1998 } 1999 if (error) 2000 break; 2001 *mp = m = m_get(M_WAIT, MT_SOOPTS); 2002 m->m_len = sizeof(int); 2003 *mtod(m, int *) = optval; 2004 break; 2005 2006 case IPV6_PATHMTU: 2007 { 2008 u_long pmtu = 0; 2009 struct ip6_mtuinfo mtuinfo; 2010 struct route_in6 *ro = (struct route_in6 *)&in6p 2011 ->in6p_route; 2012 2013 if (!(so->so_state & SS_ISCONNECTED)) 2014 return (ENOTCONN); 2015 /* 2016 * XXX: we dot not consider the case of source 2017 * routing, or optional information to specify 2018 * the outgoing interface. 2019 */ 2020 error = ip6_getpmtu(ro, NULL, NULL, 2021 &in6p->in6p_faddr, &pmtu, NULL); 2022 if (error) 2023 break; 2024 if (pmtu > IPV6_MAXPACKET) 2025 pmtu = IPV6_MAXPACKET; 2026 2027 memset(&mtuinfo, 0, sizeof(mtuinfo)); 2028 mtuinfo.ip6m_mtu = (u_int32_t)pmtu; 2029 optdata = (void *)&mtuinfo; 2030 optdatalen = sizeof(mtuinfo); 2031 if (optdatalen > MCLBYTES) 2032 return (EMSGSIZE); /* XXX */ 2033 *mp = m = m_get(M_WAIT, MT_SOOPTS); 2034 if (optdatalen > MLEN) 2035 MCLGET(m, M_WAIT); 2036 m->m_len = optdatalen; 2037 memcpy(mtod(m, void *), optdata, optdatalen); 2038 break; 2039 } 2040 2041 #ifdef RFC2292 2042 case IPV6_2292PKTINFO: 2043 case IPV6_2292HOPLIMIT: 2044 case IPV6_2292HOPOPTS: 2045 case IPV6_2292RTHDR: 2046 case IPV6_2292DSTOPTS: 2047 switch (optname) { 2048 case IPV6_2292PKTINFO: 2049 optval = OPTBIT(IN6P_PKTINFO); 2050 break; 2051 case IPV6_2292HOPLIMIT: 2052 optval = OPTBIT(IN6P_HOPLIMIT); 2053 break; 2054 case IPV6_2292HOPOPTS: 2055 optval = OPTBIT(IN6P_HOPOPTS); 2056 break; 2057 case IPV6_2292RTHDR: 2058 optval = OPTBIT(IN6P_RTHDR); 2059 break; 2060 case IPV6_2292DSTOPTS: 2061 optval = OPTBIT(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); 2062 break; 2063 } 2064 *mp = m = m_get(M_WAIT, MT_SOOPTS); 2065 m->m_len = sizeof(int); 2066 *mtod(m, int *) = optval; 2067 break; 2068 #endif 2069 case IPV6_PKTINFO: 2070 case IPV6_HOPOPTS: 2071 case IPV6_RTHDR: 2072 case IPV6_DSTOPTS: 2073 case IPV6_RTHDRDSTOPTS: 2074 case IPV6_NEXTHOP: 2075 case IPV6_OTCLASS: 2076 case IPV6_TCLASS: 2077 case IPV6_DONTFRAG: 2078 case IPV6_USE_MIN_MTU: 2079 error = ip6_getpcbopt(in6p->in6p_outputopts, 2080 optname, mp); 2081 break; 2082 2083 case IPV6_MULTICAST_IF: 2084 case IPV6_MULTICAST_HOPS: 2085 case IPV6_MULTICAST_LOOP: 2086 case IPV6_JOIN_GROUP: 2087 case IPV6_LEAVE_GROUP: 2088 error = ip6_getmoptions(optname, 2089 in6p->in6p_moptions, mp); 2090 break; 2091 2092 #if defined(IPSEC) || defined(FAST_IPSEC) 2093 case IPV6_IPSEC_POLICY: 2094 { 2095 caddr_t req = NULL; 2096 size_t len = 0; 2097 if (m) { 2098 req = mtod(m, caddr_t); 2099 len = m->m_len; 2100 } 2101 error = ipsec6_get_policy(in6p, req, len, mp); 2102 break; 2103 } 2104 #endif /* IPSEC */ 2105 2106 2107 2108 2109 default: 2110 error = ENOPROTOOPT; 2111 break; 2112 } 2113 break; 2114 } 2115 } else { 2116 error = EINVAL; 2117 if (op == PRCO_SETOPT && *mp) 2118 (void)m_free(*mp); 2119 } 2120 return (error); 2121 } 2122 2123 int 2124 ip6_raw_ctloutput(op, so, level, optname, mp) 2125 int op; 2126 struct socket *so; 2127 int level, optname; 2128 struct mbuf **mp; 2129 { 2130 int error = 0, optval, optlen; 2131 const int icmp6off = offsetof(struct icmp6_hdr, icmp6_cksum); 2132 struct in6pcb *in6p = sotoin6pcb(so); 2133 struct mbuf *m = *mp; 2134 2135 optlen = m ? m->m_len : 0; 2136 2137 if (level != IPPROTO_IPV6) { 2138 if (op == PRCO_SETOPT && *mp) 2139 (void)m_free(*mp); 2140 return (EINVAL); 2141 } 2142 2143 switch (optname) { 2144 case IPV6_CHECKSUM: 2145 /* 2146 * For ICMPv6 sockets, no modification allowed for checksum 2147 * offset, permit "no change" values to help existing apps. 2148 * 2149 * XXX RFC3542 says: "An attempt to set IPV6_CHECKSUM 2150 * for an ICMPv6 socket will fail." The current 2151 * behavior does not meet RFC3542. 2152 */ 2153 switch (op) { 2154 case PRCO_SETOPT: 2155 if (optlen != sizeof(int)) { 2156 error = EINVAL; 2157 break; 2158 } 2159 optval = *mtod(m, int *); 2160 if ((optval % 2) != 0) { 2161 /* the API assumes even offset values */ 2162 error = EINVAL; 2163 } else if (so->so_proto->pr_protocol == 2164 IPPROTO_ICMPV6) { 2165 if (optval != icmp6off) 2166 error = EINVAL; 2167 } else 2168 in6p->in6p_cksum = optval; 2169 break; 2170 2171 case PRCO_GETOPT: 2172 if (so->so_proto->pr_protocol == IPPROTO_ICMPV6) 2173 optval = icmp6off; 2174 else 2175 optval = in6p->in6p_cksum; 2176 2177 *mp = m = m_get(M_WAIT, MT_SOOPTS); 2178 m->m_len = sizeof(int); 2179 *mtod(m, int *) = optval; 2180 break; 2181 2182 default: 2183 error = EINVAL; 2184 break; 2185 } 2186 break; 2187 2188 default: 2189 error = ENOPROTOOPT; 2190 break; 2191 } 2192 2193 if (op == PRCO_SETOPT && m) 2194 (void)m_free(m); 2195 2196 return (error); 2197 } 2198 2199 #ifdef RFC2292 2200 /* 2201 * Set up IP6 options in pcb for insertion in output packets or 2202 * specifying behavior of outgoing packets. 2203 */ 2204 static int 2205 ip6_pcbopts(pktopt, m, so) 2206 struct ip6_pktopts **pktopt; 2207 struct mbuf *m; 2208 struct socket *so; 2209 { 2210 struct ip6_pktopts *opt = *pktopt; 2211 int error = 0; 2212 struct lwp *l = curlwp; /* XXX */ 2213 int priv = 0; 2214 2215 /* turn off any old options. */ 2216 if (opt) { 2217 #ifdef DIAGNOSTIC 2218 if (opt->ip6po_pktinfo || opt->ip6po_nexthop || 2219 opt->ip6po_hbh || opt->ip6po_dest1 || opt->ip6po_dest2 || 2220 opt->ip6po_rhinfo.ip6po_rhi_rthdr) 2221 printf("ip6_pcbopts: all specified options are cleared.\n"); 2222 #endif 2223 ip6_clearpktopts(opt, -1); 2224 } else 2225 opt = malloc(sizeof(*opt), M_IP6OPT, M_WAITOK); 2226 *pktopt = NULL; 2227 2228 if (!m || m->m_len == 0) { 2229 /* 2230 * Only turning off any previous options, regardless of 2231 * whether the opt is just created or given. 2232 */ 2233 free(opt, M_IP6OPT); 2234 return (0); 2235 } 2236 2237 /* set options specified by user. */ 2238 if (l && !kauth_authorize_generic(l->l_cred, KAUTH_GENERIC_ISSUSER, 2239 NULL)) 2240 priv = 1; 2241 if ((error = ip6_setpktopts(m, opt, NULL, priv, 2242 so->so_proto->pr_protocol)) != 0) { 2243 ip6_clearpktopts(opt, -1); /* XXX: discard all options */ 2244 free(opt, M_IP6OPT); 2245 return (error); 2246 } 2247 *pktopt = opt; 2248 return (0); 2249 } 2250 #endif 2251 2252 /* 2253 * initialize ip6_pktopts. beware that there are non-zero default values in 2254 * the struct. 2255 */ 2256 void 2257 ip6_initpktopts(struct ip6_pktopts *opt) 2258 { 2259 2260 memset(opt, 0, sizeof(*opt)); 2261 opt->ip6po_hlim = -1; /* -1 means default hop limit */ 2262 opt->ip6po_tclass = -1; /* -1 means default traffic class */ 2263 opt->ip6po_minmtu = IP6PO_MINMTU_MCASTONLY; 2264 } 2265 2266 #define sin6tosa(sin6) ((struct sockaddr *)(sin6)) /* XXX */ 2267 static int 2268 ip6_pcbopt(int optname, u_char *buf, int len, struct ip6_pktopts **pktopt, 2269 int priv, int uproto) 2270 { 2271 struct ip6_pktopts *opt; 2272 2273 if (*pktopt == NULL) { 2274 *pktopt = malloc(sizeof(struct ip6_pktopts), M_IP6OPT, 2275 M_WAITOK); 2276 ip6_initpktopts(*pktopt); 2277 } 2278 opt = *pktopt; 2279 2280 return (ip6_setpktopt(optname, buf, len, opt, priv, 1, 0, uproto)); 2281 } 2282 2283 static int 2284 ip6_getpcbopt(struct ip6_pktopts *pktopt, int optname, struct mbuf **mp) 2285 { 2286 void *optdata = NULL; 2287 int optdatalen = 0; 2288 struct ip6_ext *ip6e; 2289 int error = 0; 2290 struct in6_pktinfo null_pktinfo; 2291 int deftclass = 0, on; 2292 int defminmtu = IP6PO_MINMTU_MCASTONLY; 2293 struct mbuf *m; 2294 2295 switch (optname) { 2296 case IPV6_PKTINFO: 2297 if (pktopt && pktopt->ip6po_pktinfo) 2298 optdata = (void *)pktopt->ip6po_pktinfo; 2299 else { 2300 /* XXX: we don't have to do this every time... */ 2301 memset(&null_pktinfo, 0, sizeof(null_pktinfo)); 2302 optdata = (void *)&null_pktinfo; 2303 } 2304 optdatalen = sizeof(struct in6_pktinfo); 2305 break; 2306 case IPV6_OTCLASS: 2307 /* XXX */ 2308 return (EINVAL); 2309 case IPV6_TCLASS: 2310 if (pktopt && pktopt->ip6po_tclass >= 0) 2311 optdata = (void *)&pktopt->ip6po_tclass; 2312 else 2313 optdata = (void *)&deftclass; 2314 optdatalen = sizeof(int); 2315 break; 2316 case IPV6_HOPOPTS: 2317 if (pktopt && pktopt->ip6po_hbh) { 2318 optdata = (void *)pktopt->ip6po_hbh; 2319 ip6e = (struct ip6_ext *)pktopt->ip6po_hbh; 2320 optdatalen = (ip6e->ip6e_len + 1) << 3; 2321 } 2322 break; 2323 case IPV6_RTHDR: 2324 if (pktopt && pktopt->ip6po_rthdr) { 2325 optdata = (void *)pktopt->ip6po_rthdr; 2326 ip6e = (struct ip6_ext *)pktopt->ip6po_rthdr; 2327 optdatalen = (ip6e->ip6e_len + 1) << 3; 2328 } 2329 break; 2330 case IPV6_RTHDRDSTOPTS: 2331 if (pktopt && pktopt->ip6po_dest1) { 2332 optdata = (void *)pktopt->ip6po_dest1; 2333 ip6e = (struct ip6_ext *)pktopt->ip6po_dest1; 2334 optdatalen = (ip6e->ip6e_len + 1) << 3; 2335 } 2336 break; 2337 case IPV6_DSTOPTS: 2338 if (pktopt && pktopt->ip6po_dest2) { 2339 optdata = (void *)pktopt->ip6po_dest2; 2340 ip6e = (struct ip6_ext *)pktopt->ip6po_dest2; 2341 optdatalen = (ip6e->ip6e_len + 1) << 3; 2342 } 2343 break; 2344 case IPV6_NEXTHOP: 2345 if (pktopt && pktopt->ip6po_nexthop) { 2346 optdata = (void *)pktopt->ip6po_nexthop; 2347 optdatalen = pktopt->ip6po_nexthop->sa_len; 2348 } 2349 break; 2350 case IPV6_USE_MIN_MTU: 2351 if (pktopt) 2352 optdata = (void *)&pktopt->ip6po_minmtu; 2353 else 2354 optdata = (void *)&defminmtu; 2355 optdatalen = sizeof(int); 2356 break; 2357 case IPV6_DONTFRAG: 2358 if (pktopt && ((pktopt->ip6po_flags) & IP6PO_DONTFRAG)) 2359 on = 1; 2360 else 2361 on = 0; 2362 optdata = (void *)&on; 2363 optdatalen = sizeof(on); 2364 break; 2365 default: /* should not happen */ 2366 #ifdef DIAGNOSTIC 2367 panic("ip6_getpcbopt: unexpected option\n"); 2368 #endif 2369 return (ENOPROTOOPT); 2370 } 2371 2372 if (optdatalen > MCLBYTES) 2373 return (EMSGSIZE); /* XXX */ 2374 *mp = m = m_get(M_WAIT, MT_SOOPTS); 2375 if (optdatalen > MLEN) 2376 MCLGET(m, M_WAIT); 2377 m->m_len = optdatalen; 2378 if (optdatalen) 2379 memcpy(mtod(m, void *), optdata, optdatalen); 2380 2381 return (error); 2382 } 2383 2384 void 2385 ip6_clearpktopts(struct ip6_pktopts *pktopt, int optname) 2386 { 2387 if (optname == -1 || optname == IPV6_PKTINFO) { 2388 if (pktopt->ip6po_pktinfo) 2389 free(pktopt->ip6po_pktinfo, M_IP6OPT); 2390 pktopt->ip6po_pktinfo = NULL; 2391 } 2392 if (optname == -1 || optname == IPV6_HOPLIMIT) 2393 pktopt->ip6po_hlim = -1; 2394 if (optname == -1 || optname == IPV6_TCLASS) 2395 pktopt->ip6po_tclass = -1; 2396 if (optname == -1 || optname == IPV6_NEXTHOP) { 2397 rtcache_free((struct route *)&pktopt->ip6po_nextroute); 2398 if (pktopt->ip6po_nexthop) 2399 free(pktopt->ip6po_nexthop, M_IP6OPT); 2400 pktopt->ip6po_nexthop = NULL; 2401 } 2402 if (optname == -1 || optname == IPV6_HOPOPTS) { 2403 if (pktopt->ip6po_hbh) 2404 free(pktopt->ip6po_hbh, M_IP6OPT); 2405 pktopt->ip6po_hbh = NULL; 2406 } 2407 if (optname == -1 || optname == IPV6_RTHDRDSTOPTS) { 2408 if (pktopt->ip6po_dest1) 2409 free(pktopt->ip6po_dest1, M_IP6OPT); 2410 pktopt->ip6po_dest1 = NULL; 2411 } 2412 if (optname == -1 || optname == IPV6_RTHDR) { 2413 if (pktopt->ip6po_rhinfo.ip6po_rhi_rthdr) 2414 free(pktopt->ip6po_rhinfo.ip6po_rhi_rthdr, M_IP6OPT); 2415 pktopt->ip6po_rhinfo.ip6po_rhi_rthdr = NULL; 2416 rtcache_free((struct route *)&pktopt->ip6po_route); 2417 } 2418 if (optname == -1 || optname == IPV6_DSTOPTS) { 2419 if (pktopt->ip6po_dest2) 2420 free(pktopt->ip6po_dest2, M_IP6OPT); 2421 pktopt->ip6po_dest2 = NULL; 2422 } 2423 } 2424 2425 #define PKTOPT_EXTHDRCPY(type) \ 2426 do { \ 2427 if (src->type) { \ 2428 int hlen = (((struct ip6_ext *)src->type)->ip6e_len + 1) << 3;\ 2429 dst->type = malloc(hlen, M_IP6OPT, canwait); \ 2430 if (dst->type == NULL && canwait == M_NOWAIT) \ 2431 goto bad; \ 2432 memcpy(dst->type, src->type, hlen); \ 2433 } \ 2434 } while (/*CONSTCOND*/ 0) 2435 2436 static int 2437 copypktopts(struct ip6_pktopts *dst, struct ip6_pktopts *src, int canwait) 2438 { 2439 dst->ip6po_hlim = src->ip6po_hlim; 2440 dst->ip6po_tclass = src->ip6po_tclass; 2441 dst->ip6po_flags = src->ip6po_flags; 2442 if (src->ip6po_pktinfo) { 2443 dst->ip6po_pktinfo = malloc(sizeof(*dst->ip6po_pktinfo), 2444 M_IP6OPT, canwait); 2445 if (dst->ip6po_pktinfo == NULL && canwait == M_NOWAIT) 2446 goto bad; 2447 *dst->ip6po_pktinfo = *src->ip6po_pktinfo; 2448 } 2449 if (src->ip6po_nexthop) { 2450 dst->ip6po_nexthop = malloc(src->ip6po_nexthop->sa_len, 2451 M_IP6OPT, canwait); 2452 if (dst->ip6po_nexthop == NULL && canwait == M_NOWAIT) 2453 goto bad; 2454 memcpy(dst->ip6po_nexthop, src->ip6po_nexthop, 2455 src->ip6po_nexthop->sa_len); 2456 } 2457 PKTOPT_EXTHDRCPY(ip6po_hbh); 2458 PKTOPT_EXTHDRCPY(ip6po_dest1); 2459 PKTOPT_EXTHDRCPY(ip6po_dest2); 2460 PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */ 2461 return (0); 2462 2463 bad: 2464 if (dst->ip6po_pktinfo) free(dst->ip6po_pktinfo, M_IP6OPT); 2465 if (dst->ip6po_nexthop) free(dst->ip6po_nexthop, M_IP6OPT); 2466 if (dst->ip6po_hbh) free(dst->ip6po_hbh, M_IP6OPT); 2467 if (dst->ip6po_dest1) free(dst->ip6po_dest1, M_IP6OPT); 2468 if (dst->ip6po_dest2) free(dst->ip6po_dest2, M_IP6OPT); 2469 if (dst->ip6po_rthdr) free(dst->ip6po_rthdr, M_IP6OPT); 2470 2471 return (ENOBUFS); 2472 } 2473 #undef PKTOPT_EXTHDRCPY 2474 2475 struct ip6_pktopts * 2476 ip6_copypktopts(struct ip6_pktopts *src, int canwait) 2477 { 2478 int error; 2479 struct ip6_pktopts *dst; 2480 2481 dst = malloc(sizeof(*dst), M_IP6OPT, canwait); 2482 if (dst == NULL && canwait == M_NOWAIT) 2483 return (NULL); 2484 ip6_initpktopts(dst); 2485 2486 if ((error = copypktopts(dst, src, canwait)) != 0) { 2487 free(dst, M_IP6OPT); 2488 return (NULL); 2489 } 2490 2491 return (dst); 2492 } 2493 2494 void 2495 ip6_freepcbopts(struct ip6_pktopts *pktopt) 2496 { 2497 if (pktopt == NULL) 2498 return; 2499 2500 ip6_clearpktopts(pktopt, -1); 2501 2502 free(pktopt, M_IP6OPT); 2503 } 2504 2505 /* 2506 * Set the IP6 multicast options in response to user setsockopt(). 2507 */ 2508 static int 2509 ip6_setmoptions(optname, im6op, m) 2510 int optname; 2511 struct ip6_moptions **im6op; 2512 struct mbuf *m; 2513 { 2514 int error = 0; 2515 u_int loop, ifindex; 2516 struct ipv6_mreq *mreq; 2517 struct ifnet *ifp; 2518 struct ip6_moptions *im6o = *im6op; 2519 struct route_in6 ro; 2520 struct in6_multi_mship *imm; 2521 struct lwp *l = curlwp; /* XXX */ 2522 2523 if (im6o == NULL) { 2524 /* 2525 * No multicast option buffer attached to the pcb; 2526 * allocate one and initialize to default values. 2527 */ 2528 im6o = (struct ip6_moptions *) 2529 malloc(sizeof(*im6o), M_IPMOPTS, M_WAITOK); 2530 2531 if (im6o == NULL) 2532 return (ENOBUFS); 2533 *im6op = im6o; 2534 im6o->im6o_multicast_ifp = NULL; 2535 im6o->im6o_multicast_hlim = ip6_defmcasthlim; 2536 im6o->im6o_multicast_loop = IPV6_DEFAULT_MULTICAST_LOOP; 2537 LIST_INIT(&im6o->im6o_memberships); 2538 } 2539 2540 switch (optname) { 2541 2542 case IPV6_MULTICAST_IF: 2543 /* 2544 * Select the interface for outgoing multicast packets. 2545 */ 2546 if (m == NULL || m->m_len != sizeof(u_int)) { 2547 error = EINVAL; 2548 break; 2549 } 2550 bcopy(mtod(m, u_int *), &ifindex, sizeof(ifindex)); 2551 if (ifindex != 0) { 2552 if (if_indexlim <= ifindex || !ifindex2ifnet[ifindex]) { 2553 error = ENXIO; /* XXX EINVAL? */ 2554 break; 2555 } 2556 ifp = ifindex2ifnet[ifindex]; 2557 if ((ifp->if_flags & IFF_MULTICAST) == 0) { 2558 error = EADDRNOTAVAIL; 2559 break; 2560 } 2561 } else 2562 ifp = NULL; 2563 im6o->im6o_multicast_ifp = ifp; 2564 break; 2565 2566 case IPV6_MULTICAST_HOPS: 2567 { 2568 /* 2569 * Set the IP6 hoplimit for outgoing multicast packets. 2570 */ 2571 int optval; 2572 if (m == NULL || m->m_len != sizeof(int)) { 2573 error = EINVAL; 2574 break; 2575 } 2576 bcopy(mtod(m, u_int *), &optval, sizeof(optval)); 2577 if (optval < -1 || optval >= 256) 2578 error = EINVAL; 2579 else if (optval == -1) 2580 im6o->im6o_multicast_hlim = ip6_defmcasthlim; 2581 else 2582 im6o->im6o_multicast_hlim = optval; 2583 break; 2584 } 2585 2586 case IPV6_MULTICAST_LOOP: 2587 /* 2588 * Set the loopback flag for outgoing multicast packets. 2589 * Must be zero or one. 2590 */ 2591 if (m == NULL || m->m_len != sizeof(u_int)) { 2592 error = EINVAL; 2593 break; 2594 } 2595 bcopy(mtod(m, u_int *), &loop, sizeof(loop)); 2596 if (loop > 1) { 2597 error = EINVAL; 2598 break; 2599 } 2600 im6o->im6o_multicast_loop = loop; 2601 break; 2602 2603 case IPV6_JOIN_GROUP: 2604 /* 2605 * Add a multicast group membership. 2606 * Group must be a valid IP6 multicast address. 2607 */ 2608 if (m == NULL || m->m_len != sizeof(struct ipv6_mreq)) { 2609 error = EINVAL; 2610 break; 2611 } 2612 mreq = mtod(m, struct ipv6_mreq *); 2613 if (IN6_IS_ADDR_UNSPECIFIED(&mreq->ipv6mr_multiaddr)) { 2614 /* 2615 * We use the unspecified address to specify to accept 2616 * all multicast addresses. Only super user is allowed 2617 * to do this. 2618 */ 2619 if (kauth_authorize_generic(l->l_cred, 2620 KAUTH_GENERIC_ISSUSER, NULL)) 2621 { 2622 error = EACCES; 2623 break; 2624 } 2625 } else if (!IN6_IS_ADDR_MULTICAST(&mreq->ipv6mr_multiaddr)) { 2626 error = EINVAL; 2627 break; 2628 } 2629 2630 /* 2631 * If no interface was explicitly specified, choose an 2632 * appropriate one according to the given multicast address. 2633 */ 2634 if (mreq->ipv6mr_interface == 0) { 2635 struct sockaddr_in6 *dst; 2636 2637 /* 2638 * Look up the routing table for the 2639 * address, and choose the outgoing interface. 2640 * XXX: is it a good approach? 2641 */ 2642 memset(&ro, 0, sizeof(ro)); 2643 dst = &ro.ro_dst; 2644 dst->sin6_family = AF_INET6; 2645 dst->sin6_len = sizeof(*dst); 2646 dst->sin6_addr = mreq->ipv6mr_multiaddr; 2647 rtcache_init((struct route *)&ro); 2648 ifp = (ro.ro_rt != NULL) ? ro.ro_rt->rt_ifp : NULL; 2649 rtcache_free((struct route *)&ro); 2650 } else { 2651 /* 2652 * If the interface is specified, validate it. 2653 */ 2654 if (if_indexlim <= mreq->ipv6mr_interface || 2655 !ifindex2ifnet[mreq->ipv6mr_interface]) { 2656 error = ENXIO; /* XXX EINVAL? */ 2657 break; 2658 } 2659 ifp = ifindex2ifnet[mreq->ipv6mr_interface]; 2660 } 2661 2662 /* 2663 * See if we found an interface, and confirm that it 2664 * supports multicast 2665 */ 2666 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) { 2667 error = EADDRNOTAVAIL; 2668 break; 2669 } 2670 2671 if (in6_setscope(&mreq->ipv6mr_multiaddr, ifp, NULL)) { 2672 error = EADDRNOTAVAIL; /* XXX: should not happen */ 2673 break; 2674 } 2675 2676 /* 2677 * See if the membership already exists. 2678 */ 2679 for (imm = im6o->im6o_memberships.lh_first; 2680 imm != NULL; imm = imm->i6mm_chain.le_next) 2681 if (imm->i6mm_maddr->in6m_ifp == ifp && 2682 IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr, 2683 &mreq->ipv6mr_multiaddr)) 2684 break; 2685 if (imm != NULL) { 2686 error = EADDRINUSE; 2687 break; 2688 } 2689 /* 2690 * Everything looks good; add a new record to the multicast 2691 * address list for the given interface. 2692 */ 2693 imm = in6_joingroup(ifp, &mreq->ipv6mr_multiaddr, &error, 0); 2694 if (imm == NULL) 2695 break; 2696 LIST_INSERT_HEAD(&im6o->im6o_memberships, imm, i6mm_chain); 2697 break; 2698 2699 case IPV6_LEAVE_GROUP: 2700 /* 2701 * Drop a multicast group membership. 2702 * Group must be a valid IP6 multicast address. 2703 */ 2704 if (m == NULL || m->m_len != sizeof(struct ipv6_mreq)) { 2705 error = EINVAL; 2706 break; 2707 } 2708 mreq = mtod(m, struct ipv6_mreq *); 2709 2710 /* 2711 * If an interface address was specified, get a pointer 2712 * to its ifnet structure. 2713 */ 2714 if (mreq->ipv6mr_interface != 0) { 2715 if (if_indexlim <= mreq->ipv6mr_interface || 2716 !ifindex2ifnet[mreq->ipv6mr_interface]) { 2717 error = ENXIO; /* XXX EINVAL? */ 2718 break; 2719 } 2720 ifp = ifindex2ifnet[mreq->ipv6mr_interface]; 2721 } else 2722 ifp = NULL; 2723 2724 /* Fill in the scope zone ID */ 2725 if (ifp) { 2726 if (in6_setscope(&mreq->ipv6mr_multiaddr, ifp, NULL)) { 2727 /* XXX: should not happen */ 2728 error = EADDRNOTAVAIL; 2729 break; 2730 } 2731 } else if (mreq->ipv6mr_interface != 0) { 2732 /* 2733 * XXX: This case would happens when the (positive) 2734 * index is in the valid range, but the corresponding 2735 * interface has been detached dynamically. The above 2736 * check probably avoids such case to happen here, but 2737 * we check it explicitly for safety. 2738 */ 2739 error = EADDRNOTAVAIL; 2740 break; 2741 } else { /* ipv6mr_interface == 0 */ 2742 struct sockaddr_in6 sa6_mc; 2743 2744 /* 2745 * The API spec says as follows: 2746 * If the interface index is specified as 0, the 2747 * system may choose a multicast group membership to 2748 * drop by matching the multicast address only. 2749 * On the other hand, we cannot disambiguate the scope 2750 * zone unless an interface is provided. Thus, we 2751 * check if there's ambiguity with the default scope 2752 * zone as the last resort. 2753 */ 2754 bzero(&sa6_mc, sizeof(sa6_mc)); 2755 sa6_mc.sin6_family = AF_INET6; 2756 sa6_mc.sin6_len = sizeof(sa6_mc); 2757 sa6_mc.sin6_addr = mreq->ipv6mr_multiaddr; 2758 error = sa6_embedscope(&sa6_mc, ip6_use_defzone); 2759 if (error != 0) 2760 break; 2761 mreq->ipv6mr_multiaddr = sa6_mc.sin6_addr; 2762 } 2763 2764 /* 2765 * Find the membership in the membership list. 2766 */ 2767 for (imm = im6o->im6o_memberships.lh_first; 2768 imm != NULL; imm = imm->i6mm_chain.le_next) { 2769 if ((ifp == NULL || imm->i6mm_maddr->in6m_ifp == ifp) && 2770 IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr, 2771 &mreq->ipv6mr_multiaddr)) 2772 break; 2773 } 2774 if (imm == NULL) { 2775 /* Unable to resolve interface */ 2776 error = EADDRNOTAVAIL; 2777 break; 2778 } 2779 /* 2780 * Give up the multicast address record to which the 2781 * membership points. 2782 */ 2783 LIST_REMOVE(imm, i6mm_chain); 2784 in6_leavegroup(imm); 2785 break; 2786 2787 default: 2788 error = EOPNOTSUPP; 2789 break; 2790 } 2791 2792 /* 2793 * If all options have default values, no need to keep the mbuf. 2794 */ 2795 if (im6o->im6o_multicast_ifp == NULL && 2796 im6o->im6o_multicast_hlim == ip6_defmcasthlim && 2797 im6o->im6o_multicast_loop == IPV6_DEFAULT_MULTICAST_LOOP && 2798 im6o->im6o_memberships.lh_first == NULL) { 2799 free(*im6op, M_IPMOPTS); 2800 *im6op = NULL; 2801 } 2802 2803 return (error); 2804 } 2805 2806 /* 2807 * Return the IP6 multicast options in response to user getsockopt(). 2808 */ 2809 static int 2810 ip6_getmoptions(optname, im6o, mp) 2811 int optname; 2812 struct ip6_moptions *im6o; 2813 struct mbuf **mp; 2814 { 2815 u_int *hlim, *loop, *ifindex; 2816 2817 *mp = m_get(M_WAIT, MT_SOOPTS); 2818 2819 switch (optname) { 2820 2821 case IPV6_MULTICAST_IF: 2822 ifindex = mtod(*mp, u_int *); 2823 (*mp)->m_len = sizeof(u_int); 2824 if (im6o == NULL || im6o->im6o_multicast_ifp == NULL) 2825 *ifindex = 0; 2826 else 2827 *ifindex = im6o->im6o_multicast_ifp->if_index; 2828 return (0); 2829 2830 case IPV6_MULTICAST_HOPS: 2831 hlim = mtod(*mp, u_int *); 2832 (*mp)->m_len = sizeof(u_int); 2833 if (im6o == NULL) 2834 *hlim = ip6_defmcasthlim; 2835 else 2836 *hlim = im6o->im6o_multicast_hlim; 2837 return (0); 2838 2839 case IPV6_MULTICAST_LOOP: 2840 loop = mtod(*mp, u_int *); 2841 (*mp)->m_len = sizeof(u_int); 2842 if (im6o == NULL) 2843 *loop = ip6_defmcasthlim; 2844 else 2845 *loop = im6o->im6o_multicast_loop; 2846 return (0); 2847 2848 default: 2849 return (EOPNOTSUPP); 2850 } 2851 } 2852 2853 /* 2854 * Discard the IP6 multicast options. 2855 */ 2856 void 2857 ip6_freemoptions(im6o) 2858 struct ip6_moptions *im6o; 2859 { 2860 struct in6_multi_mship *imm; 2861 2862 if (im6o == NULL) 2863 return; 2864 2865 while ((imm = im6o->im6o_memberships.lh_first) != NULL) { 2866 LIST_REMOVE(imm, i6mm_chain); 2867 in6_leavegroup(imm); 2868 } 2869 free(im6o, M_IPMOPTS); 2870 } 2871 2872 /* 2873 * Set IPv6 outgoing packet options based on advanced API. 2874 */ 2875 int 2876 ip6_setpktopts(control, opt, stickyopt, priv, uproto) 2877 struct mbuf *control; 2878 struct ip6_pktopts *opt, *stickyopt; 2879 int priv, uproto; 2880 { 2881 struct cmsghdr *cm = 0; 2882 2883 if (control == NULL || opt == NULL) 2884 return (EINVAL); 2885 2886 ip6_initpktopts(opt); 2887 if (stickyopt) { 2888 int error; 2889 2890 /* 2891 * If stickyopt is provided, make a local copy of the options 2892 * for this particular packet, then override them by ancillary 2893 * objects. 2894 * XXX: copypktopts() does not copy the cached route to a next 2895 * hop (if any). This is not very good in terms of efficiency, 2896 * but we can allow this since this option should be rarely 2897 * used. 2898 */ 2899 if ((error = copypktopts(opt, stickyopt, M_NOWAIT)) != 0) 2900 return (error); 2901 } 2902 2903 /* 2904 * XXX: Currently, we assume all the optional information is stored 2905 * in a single mbuf. 2906 */ 2907 if (control->m_next) 2908 return (EINVAL); 2909 2910 for (; control->m_len; control->m_data += CMSG_ALIGN(cm->cmsg_len), 2911 control->m_len -= CMSG_ALIGN(cm->cmsg_len)) { 2912 int error; 2913 2914 if (control->m_len < CMSG_LEN(0)) 2915 return (EINVAL); 2916 2917 cm = mtod(control, struct cmsghdr *); 2918 if (cm->cmsg_len == 0 || cm->cmsg_len > control->m_len) 2919 return (EINVAL); 2920 if (cm->cmsg_level != IPPROTO_IPV6) 2921 continue; 2922 2923 error = ip6_setpktopt(cm->cmsg_type, CMSG_DATA(cm), 2924 cm->cmsg_len - CMSG_LEN(0), opt, priv, 0, 1, uproto); 2925 if (error) 2926 return (error); 2927 } 2928 2929 return (0); 2930 } 2931 2932 /* 2933 * Set a particular packet option, as a sticky option or an ancillary data 2934 * item. "len" can be 0 only when it's a sticky option. 2935 * We have 4 cases of combination of "sticky" and "cmsg": 2936 * "sticky=0, cmsg=0": impossible 2937 * "sticky=0, cmsg=1": RFC2292 or RFC3542 ancillary data 2938 * "sticky=1, cmsg=0": RFC3542 socket option 2939 * "sticky=1, cmsg=1": RFC2292 socket option 2940 */ 2941 static int 2942 ip6_setpktopt(int optname, u_char *buf, int len, struct ip6_pktopts *opt, 2943 int priv, int sticky, int cmsg, int uproto) 2944 { 2945 int minmtupolicy; 2946 2947 if (!sticky && !cmsg) { 2948 #ifdef DIAGNOSTIC 2949 printf("ip6_setpktopt: impossible case\n"); 2950 #endif 2951 return (EINVAL); 2952 } 2953 2954 /* 2955 * IPV6_2292xxx is for backward compatibility to RFC2292, and should 2956 * not be specified in the context of RFC3542. Conversely, 2957 * RFC3542 types should not be specified in the context of RFC2292. 2958 */ 2959 if (!cmsg) { 2960 switch (optname) { 2961 case IPV6_2292PKTINFO: 2962 case IPV6_2292HOPLIMIT: 2963 case IPV6_2292NEXTHOP: 2964 case IPV6_2292HOPOPTS: 2965 case IPV6_2292DSTOPTS: 2966 case IPV6_2292RTHDR: 2967 case IPV6_2292PKTOPTIONS: 2968 return (ENOPROTOOPT); 2969 } 2970 } 2971 if (sticky && cmsg) { 2972 switch (optname) { 2973 case IPV6_PKTINFO: 2974 case IPV6_HOPLIMIT: 2975 case IPV6_NEXTHOP: 2976 case IPV6_HOPOPTS: 2977 case IPV6_DSTOPTS: 2978 case IPV6_RTHDRDSTOPTS: 2979 case IPV6_RTHDR: 2980 case IPV6_USE_MIN_MTU: 2981 case IPV6_DONTFRAG: 2982 case IPV6_OTCLASS: 2983 case IPV6_TCLASS: 2984 return (ENOPROTOOPT); 2985 } 2986 } 2987 2988 switch (optname) { 2989 #ifdef RFC2292 2990 case IPV6_2292PKTINFO: 2991 #endif 2992 case IPV6_PKTINFO: 2993 { 2994 struct ifnet *ifp = NULL; 2995 struct in6_pktinfo *pktinfo; 2996 2997 if (len != sizeof(struct in6_pktinfo)) 2998 return (EINVAL); 2999 3000 pktinfo = (struct in6_pktinfo *)buf; 3001 3002 /* 3003 * An application can clear any sticky IPV6_PKTINFO option by 3004 * doing a "regular" setsockopt with ipi6_addr being 3005 * in6addr_any and ipi6_ifindex being zero. 3006 * [RFC 3542, Section 6] 3007 */ 3008 if (optname == IPV6_PKTINFO && opt->ip6po_pktinfo && 3009 pktinfo->ipi6_ifindex == 0 && 3010 IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) { 3011 ip6_clearpktopts(opt, optname); 3012 break; 3013 } 3014 3015 if (uproto == IPPROTO_TCP && optname == IPV6_PKTINFO && 3016 sticky && !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) { 3017 return (EINVAL); 3018 } 3019 3020 /* validate the interface index if specified. */ 3021 if (pktinfo->ipi6_ifindex >= if_indexlim) { 3022 return (ENXIO); 3023 } 3024 if (pktinfo->ipi6_ifindex) { 3025 ifp = ifindex2ifnet[pktinfo->ipi6_ifindex]; 3026 if (ifp == NULL) 3027 return (ENXIO); 3028 } 3029 3030 /* 3031 * We store the address anyway, and let in6_selectsrc() 3032 * validate the specified address. This is because ipi6_addr 3033 * may not have enough information about its scope zone, and 3034 * we may need additional information (such as outgoing 3035 * interface or the scope zone of a destination address) to 3036 * disambiguate the scope. 3037 * XXX: the delay of the validation may confuse the 3038 * application when it is used as a sticky option. 3039 */ 3040 if (opt->ip6po_pktinfo == NULL) { 3041 opt->ip6po_pktinfo = malloc(sizeof(*pktinfo), 3042 M_IP6OPT, M_NOWAIT); 3043 if (opt->ip6po_pktinfo == NULL) 3044 return (ENOBUFS); 3045 } 3046 memcpy(opt->ip6po_pktinfo, pktinfo, sizeof(*pktinfo)); 3047 break; 3048 } 3049 3050 #ifdef RFC2292 3051 case IPV6_2292HOPLIMIT: 3052 #endif 3053 case IPV6_HOPLIMIT: 3054 { 3055 int *hlimp; 3056 3057 /* 3058 * RFC 3542 deprecated the usage of sticky IPV6_HOPLIMIT 3059 * to simplify the ordering among hoplimit options. 3060 */ 3061 if (optname == IPV6_HOPLIMIT && sticky) 3062 return (ENOPROTOOPT); 3063 3064 if (len != sizeof(int)) 3065 return (EINVAL); 3066 hlimp = (int *)buf; 3067 if (*hlimp < -1 || *hlimp > 255) 3068 return (EINVAL); 3069 3070 opt->ip6po_hlim = *hlimp; 3071 break; 3072 } 3073 3074 case IPV6_OTCLASS: 3075 if (len != sizeof(u_int8_t)) 3076 return (EINVAL); 3077 3078 opt->ip6po_tclass = *(u_int8_t *)buf; 3079 break; 3080 3081 case IPV6_TCLASS: 3082 { 3083 int tclass; 3084 3085 if (len != sizeof(int)) 3086 return (EINVAL); 3087 tclass = *(int *)buf; 3088 if (tclass < -1 || tclass > 255) 3089 return (EINVAL); 3090 3091 opt->ip6po_tclass = tclass; 3092 break; 3093 } 3094 3095 #ifdef RFC2292 3096 case IPV6_2292NEXTHOP: 3097 #endif 3098 case IPV6_NEXTHOP: 3099 if (!priv) 3100 return (EPERM); 3101 3102 if (len == 0) { /* just remove the option */ 3103 ip6_clearpktopts(opt, IPV6_NEXTHOP); 3104 break; 3105 } 3106 3107 /* check if cmsg_len is large enough for sa_len */ 3108 if (len < sizeof(struct sockaddr) || len < *buf) 3109 return (EINVAL); 3110 3111 switch (((struct sockaddr *)buf)->sa_family) { 3112 case AF_INET6: 3113 { 3114 struct sockaddr_in6 *sa6 = (struct sockaddr_in6 *)buf; 3115 int error; 3116 3117 if (sa6->sin6_len != sizeof(struct sockaddr_in6)) 3118 return (EINVAL); 3119 3120 if (IN6_IS_ADDR_UNSPECIFIED(&sa6->sin6_addr) || 3121 IN6_IS_ADDR_MULTICAST(&sa6->sin6_addr)) { 3122 return (EINVAL); 3123 } 3124 if ((error = sa6_embedscope(sa6, ip6_use_defzone)) 3125 != 0) { 3126 return (error); 3127 } 3128 break; 3129 } 3130 case AF_LINK: /* eventually be supported? */ 3131 default: 3132 return (EAFNOSUPPORT); 3133 } 3134 3135 /* turn off the previous option, then set the new option. */ 3136 ip6_clearpktopts(opt, IPV6_NEXTHOP); 3137 opt->ip6po_nexthop = malloc(*buf, M_IP6OPT, M_NOWAIT); 3138 if (opt->ip6po_nexthop == NULL) 3139 return (ENOBUFS); 3140 memcpy(opt->ip6po_nexthop, buf, *buf); 3141 break; 3142 3143 #ifdef RFC2292 3144 case IPV6_2292HOPOPTS: 3145 #endif 3146 case IPV6_HOPOPTS: 3147 { 3148 struct ip6_hbh *hbh; 3149 int hbhlen; 3150 3151 /* 3152 * XXX: We don't allow a non-privileged user to set ANY HbH 3153 * options, since per-option restriction has too much 3154 * overhead. 3155 */ 3156 if (!priv) 3157 return (EPERM); 3158 3159 if (len == 0) { 3160 ip6_clearpktopts(opt, IPV6_HOPOPTS); 3161 break; /* just remove the option */ 3162 } 3163 3164 /* message length validation */ 3165 if (len < sizeof(struct ip6_hbh)) 3166 return (EINVAL); 3167 hbh = (struct ip6_hbh *)buf; 3168 hbhlen = (hbh->ip6h_len + 1) << 3; 3169 if (len != hbhlen) 3170 return (EINVAL); 3171 3172 /* turn off the previous option, then set the new option. */ 3173 ip6_clearpktopts(opt, IPV6_HOPOPTS); 3174 opt->ip6po_hbh = malloc(hbhlen, M_IP6OPT, M_NOWAIT); 3175 if (opt->ip6po_hbh == NULL) 3176 return (ENOBUFS); 3177 memcpy(opt->ip6po_hbh, hbh, hbhlen); 3178 3179 break; 3180 } 3181 3182 #ifdef RFC2292 3183 case IPV6_2292DSTOPTS: 3184 #endif 3185 case IPV6_DSTOPTS: 3186 case IPV6_RTHDRDSTOPTS: 3187 { 3188 struct ip6_dest *dest, **newdest = NULL; 3189 int destlen; 3190 3191 if (!priv) /* XXX: see the comment for IPV6_HOPOPTS */ 3192 return (EPERM); 3193 3194 if (len == 0) { 3195 ip6_clearpktopts(opt, optname); 3196 break; /* just remove the option */ 3197 } 3198 3199 /* message length validation */ 3200 if (len < sizeof(struct ip6_dest)) 3201 return (EINVAL); 3202 dest = (struct ip6_dest *)buf; 3203 destlen = (dest->ip6d_len + 1) << 3; 3204 if (len != destlen) 3205 return (EINVAL); 3206 /* 3207 * Determine the position that the destination options header 3208 * should be inserted; before or after the routing header. 3209 */ 3210 switch (optname) { 3211 case IPV6_2292DSTOPTS: 3212 /* 3213 * The old advanced API is ambiguous on this point. 3214 * Our approach is to determine the position based 3215 * according to the existence of a routing header. 3216 * Note, however, that this depends on the order of the 3217 * extension headers in the ancillary data; the 1st 3218 * part of the destination options header must appear 3219 * before the routing header in the ancillary data, 3220 * too. 3221 * RFC3542 solved the ambiguity by introducing 3222 * separate ancillary data or option types. 3223 */ 3224 if (opt->ip6po_rthdr == NULL) 3225 newdest = &opt->ip6po_dest1; 3226 else 3227 newdest = &opt->ip6po_dest2; 3228 break; 3229 case IPV6_RTHDRDSTOPTS: 3230 newdest = &opt->ip6po_dest1; 3231 break; 3232 case IPV6_DSTOPTS: 3233 newdest = &opt->ip6po_dest2; 3234 break; 3235 } 3236 3237 /* turn off the previous option, then set the new option. */ 3238 ip6_clearpktopts(opt, optname); 3239 *newdest = malloc(destlen, M_IP6OPT, M_NOWAIT); 3240 if (*newdest == NULL) 3241 return (ENOBUFS); 3242 memcpy(*newdest, dest, destlen); 3243 3244 break; 3245 } 3246 3247 #ifdef RFC2292 3248 case IPV6_2292RTHDR: 3249 #endif 3250 case IPV6_RTHDR: 3251 { 3252 struct ip6_rthdr *rth; 3253 int rthlen; 3254 3255 if (len == 0) { 3256 ip6_clearpktopts(opt, IPV6_RTHDR); 3257 break; /* just remove the option */ 3258 } 3259 3260 /* message length validation */ 3261 if (len < sizeof(struct ip6_rthdr)) 3262 return (EINVAL); 3263 rth = (struct ip6_rthdr *)buf; 3264 rthlen = (rth->ip6r_len + 1) << 3; 3265 if (len != rthlen) 3266 return (EINVAL); 3267 switch (rth->ip6r_type) { 3268 case IPV6_RTHDR_TYPE_0: 3269 if (rth->ip6r_len == 0) /* must contain one addr */ 3270 return (EINVAL); 3271 if (rth->ip6r_len % 2) /* length must be even */ 3272 return (EINVAL); 3273 if (rth->ip6r_len / 2 != rth->ip6r_segleft) 3274 return (EINVAL); 3275 break; 3276 default: 3277 return (EINVAL); /* not supported */ 3278 } 3279 /* turn off the previous option */ 3280 ip6_clearpktopts(opt, IPV6_RTHDR); 3281 opt->ip6po_rthdr = malloc(rthlen, M_IP6OPT, M_NOWAIT); 3282 if (opt->ip6po_rthdr == NULL) 3283 return (ENOBUFS); 3284 memcpy(opt->ip6po_rthdr, rth, rthlen); 3285 break; 3286 } 3287 3288 case IPV6_USE_MIN_MTU: 3289 if (len != sizeof(int)) 3290 return (EINVAL); 3291 minmtupolicy = *(int *)buf; 3292 if (minmtupolicy != IP6PO_MINMTU_MCASTONLY && 3293 minmtupolicy != IP6PO_MINMTU_DISABLE && 3294 minmtupolicy != IP6PO_MINMTU_ALL) { 3295 return (EINVAL); 3296 } 3297 opt->ip6po_minmtu = minmtupolicy; 3298 break; 3299 3300 case IPV6_DONTFRAG: 3301 if (len != sizeof(int)) 3302 return (EINVAL); 3303 3304 if (uproto == IPPROTO_TCP || *(int *)buf == 0) { 3305 /* 3306 * we ignore this option for TCP sockets. 3307 * (RFC3542 leaves this case unspecified.) 3308 */ 3309 opt->ip6po_flags &= ~IP6PO_DONTFRAG; 3310 } else 3311 opt->ip6po_flags |= IP6PO_DONTFRAG; 3312 break; 3313 3314 default: 3315 return (ENOPROTOOPT); 3316 } /* end of switch */ 3317 3318 return (0); 3319 } 3320 3321 /* 3322 * Routine called from ip6_output() to loop back a copy of an IP6 multicast 3323 * packet to the input queue of a specified interface. Note that this 3324 * calls the output routine of the loopback "driver", but with an interface 3325 * pointer that might NOT be lo0ifp -- easier than replicating that code here. 3326 */ 3327 void 3328 ip6_mloopback(ifp, m, dst) 3329 struct ifnet *ifp; 3330 struct mbuf *m; 3331 struct sockaddr_in6 *dst; 3332 { 3333 struct mbuf *copym; 3334 struct ip6_hdr *ip6; 3335 3336 copym = m_copy(m, 0, M_COPYALL); 3337 if (copym == NULL) 3338 return; 3339 3340 /* 3341 * Make sure to deep-copy IPv6 header portion in case the data 3342 * is in an mbuf cluster, so that we can safely override the IPv6 3343 * header portion later. 3344 */ 3345 if ((copym->m_flags & M_EXT) != 0 || 3346 copym->m_len < sizeof(struct ip6_hdr)) { 3347 copym = m_pullup(copym, sizeof(struct ip6_hdr)); 3348 if (copym == NULL) 3349 return; 3350 } 3351 3352 #ifdef DIAGNOSTIC 3353 if (copym->m_len < sizeof(*ip6)) { 3354 m_freem(copym); 3355 return; 3356 } 3357 #endif 3358 3359 ip6 = mtod(copym, struct ip6_hdr *); 3360 /* 3361 * clear embedded scope identifiers if necessary. 3362 * in6_clearscope will touch the addresses only when necessary. 3363 */ 3364 in6_clearscope(&ip6->ip6_src); 3365 in6_clearscope(&ip6->ip6_dst); 3366 3367 (void)looutput(ifp, copym, (struct sockaddr *)dst, NULL); 3368 } 3369 3370 /* 3371 * Chop IPv6 header off from the payload. 3372 */ 3373 static int 3374 ip6_splithdr(m, exthdrs) 3375 struct mbuf *m; 3376 struct ip6_exthdrs *exthdrs; 3377 { 3378 struct mbuf *mh; 3379 struct ip6_hdr *ip6; 3380 3381 ip6 = mtod(m, struct ip6_hdr *); 3382 if (m->m_len > sizeof(*ip6)) { 3383 MGETHDR(mh, M_DONTWAIT, MT_HEADER); 3384 if (mh == 0) { 3385 m_freem(m); 3386 return ENOBUFS; 3387 } 3388 M_MOVE_PKTHDR(mh, m); 3389 MH_ALIGN(mh, sizeof(*ip6)); 3390 m->m_len -= sizeof(*ip6); 3391 m->m_data += sizeof(*ip6); 3392 mh->m_next = m; 3393 m = mh; 3394 m->m_len = sizeof(*ip6); 3395 bcopy((caddr_t)ip6, mtod(m, caddr_t), sizeof(*ip6)); 3396 } 3397 exthdrs->ip6e_ip6 = m; 3398 return 0; 3399 } 3400 3401 /* 3402 * Compute IPv6 extension header length. 3403 */ 3404 int 3405 ip6_optlen(in6p) 3406 struct in6pcb *in6p; 3407 { 3408 int len; 3409 3410 if (!in6p->in6p_outputopts) 3411 return 0; 3412 3413 len = 0; 3414 #define elen(x) \ 3415 (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0) 3416 3417 len += elen(in6p->in6p_outputopts->ip6po_hbh); 3418 len += elen(in6p->in6p_outputopts->ip6po_dest1); 3419 len += elen(in6p->in6p_outputopts->ip6po_rthdr); 3420 len += elen(in6p->in6p_outputopts->ip6po_dest2); 3421 return len; 3422 #undef elen 3423 } 3424
Indexes created Tue Sep 30 20:09:53 GMT 2025