ip6_output.c revision 1.128
1 /* $NetBSD: ip6_output.c,v 1.128 2008/04/15 03:57:04 thorpej 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 1.128 2008/04/15 03:57:04 thorpej Exp $"); 66 67 #include "opt_inet.h" 68 #include "opt_inet6.h" 69 #include "opt_ipsec.h" 70 #include "opt_pfil_hooks.h" 71 72 #include <sys/param.h> 73 #include <sys/malloc.h> 74 #include <sys/mbuf.h> 75 #include <sys/errno.h> 76 #include <sys/protosw.h> 77 #include <sys/socket.h> 78 #include <sys/socketvar.h> 79 #include <sys/systm.h> 80 #include <sys/proc.h> 81 #include <sys/kauth.h> 82 83 #include <net/if.h> 84 #include <net/route.h> 85 #ifdef PFIL_HOOKS 86 #include <net/pfil.h> 87 #endif 88 89 #include <netinet/in.h> 90 #include <netinet/in_var.h> 91 #include <netinet/ip6.h> 92 #include <netinet/icmp6.h> 93 #include <netinet/in_offload.h> 94 #include <netinet6/in6_offload.h> 95 #include <netinet6/ip6_var.h> 96 #include <netinet6/ip6_private.h> 97 #include <netinet6/in6_pcb.h> 98 #include <netinet6/nd6.h> 99 #include <netinet6/ip6protosw.h> 100 #include <netinet6/scope6_var.h> 101 102 #ifdef IPSEC 103 #include <netinet6/ipsec.h> 104 #include <netkey/key.h> 105 #endif /* IPSEC */ 106 107 #ifdef FAST_IPSEC 108 #include <netipsec/ipsec.h> 109 #include <netipsec/ipsec6.h> 110 #include <netipsec/key.h> 111 #include <netipsec/xform.h> 112 #endif 113 114 115 #include <net/net_osdep.h> 116 117 #ifdef PFIL_HOOKS 118 extern struct pfil_head inet6_pfil_hook; /* XXX */ 119 #endif 120 121 struct ip6_exthdrs { 122 struct mbuf *ip6e_ip6; 123 struct mbuf *ip6e_hbh; 124 struct mbuf *ip6e_dest1; 125 struct mbuf *ip6e_rthdr; 126 struct mbuf *ip6e_dest2; 127 }; 128 129 static int ip6_pcbopt(int, u_char *, int, struct ip6_pktopts **, 130 int, int); 131 static int ip6_getpcbopt(struct ip6_pktopts *, int, struct mbuf **); 132 static int ip6_setpktopt(int, u_char *, int, struct ip6_pktopts *, int, 133 int, int, int); 134 static int ip6_setmoptions(int, struct ip6_moptions **, struct mbuf *); 135 static int ip6_getmoptions(int, struct ip6_moptions *, struct mbuf **); 136 static int ip6_copyexthdr(struct mbuf **, void *, int); 137 static int ip6_insertfraghdr(struct mbuf *, struct mbuf *, int, 138 struct ip6_frag **); 139 static int ip6_insert_jumboopt(struct ip6_exthdrs *, u_int32_t); 140 static int ip6_splithdr(struct mbuf *, struct ip6_exthdrs *); 141 static int ip6_getpmtu(struct route *, struct route *, struct ifnet *, 142 const struct in6_addr *, u_long *, int *); 143 static int copypktopts(struct ip6_pktopts *, struct ip6_pktopts *, int); 144 145 #ifdef RFC2292 146 static int ip6_pcbopts(struct ip6_pktopts **, struct mbuf *, 147 struct socket *); 148 #endif 149 150 #define IN6_NEED_CHECKSUM(ifp, csum_flags) \ 151 (__predict_true(((ifp)->if_flags & IFF_LOOPBACK) == 0 || \ 152 (((csum_flags) & M_CSUM_UDPv6) != 0 && udp_do_loopback_cksum) || \ 153 (((csum_flags) & M_CSUM_TCPv6) != 0 && tcp_do_loopback_cksum))) 154 155 /* 156 * IP6 output. The packet in mbuf chain m contains a skeletal IP6 157 * header (with pri, len, nxt, hlim, src, dst). 158 * This function may modify ver and hlim only. 159 * The mbuf chain containing the packet will be freed. 160 * The mbuf opt, if present, will not be freed. 161 * 162 * type of "mtu": rt_rmx.rmx_mtu is u_long, ifnet.ifr_mtu is int, and 163 * nd_ifinfo.linkmtu is u_int32_t. so we use u_long to hold largest one, 164 * which is rt_rmx.rmx_mtu. 165 */ 166 int 167 ip6_output( 168 struct mbuf *m0, 169 struct ip6_pktopts *opt, 170 struct route *ro, 171 int flags, 172 struct ip6_moptions *im6o, 173 struct socket *so, 174 struct ifnet **ifpp /* XXX: just for statistics */ 175 ) 176 { 177 struct ip6_hdr *ip6, *mhip6; 178 struct ifnet *ifp, *origifp; 179 struct mbuf *m = m0; 180 int hlen, tlen, len, off; 181 bool tso; 182 struct route ip6route; 183 struct rtentry *rt = NULL; 184 const struct sockaddr_in6 *dst = NULL; 185 struct sockaddr_in6 src_sa, dst_sa; 186 int error = 0; 187 struct in6_ifaddr *ia = NULL; 188 u_long mtu; 189 int alwaysfrag, dontfrag; 190 u_int32_t optlen = 0, plen = 0, unfragpartlen = 0; 191 struct ip6_exthdrs exthdrs; 192 struct in6_addr finaldst, src0, dst0; 193 u_int32_t zone; 194 struct route *ro_pmtu = NULL; 195 int hdrsplit = 0; 196 int needipsec = 0; 197 #ifdef IPSEC 198 int needipsectun = 0; 199 struct secpolicy *sp = NULL; 200 201 ip6 = mtod(m, struct ip6_hdr *); 202 #endif /* IPSEC */ 203 #ifdef FAST_IPSEC 204 struct secpolicy *sp = NULL; 205 int s; 206 #endif 207 208 memset(&ip6route, 0, sizeof(ip6route)); 209 210 #ifdef DIAGNOSTIC 211 if ((m->m_flags & M_PKTHDR) == 0) 212 panic("ip6_output: no HDR"); 213 214 if ((m->m_pkthdr.csum_flags & 215 (M_CSUM_TCPv4|M_CSUM_UDPv4|M_CSUM_TSOv4)) != 0) { 216 panic("ip6_output: IPv4 checksum offload flags: %d", 217 m->m_pkthdr.csum_flags); 218 } 219 220 if ((m->m_pkthdr.csum_flags & (M_CSUM_TCPv6|M_CSUM_UDPv6)) == 221 (M_CSUM_TCPv6|M_CSUM_UDPv6)) { 222 panic("ip6_output: conflicting checksum offload flags: %d", 223 m->m_pkthdr.csum_flags); 224 } 225 #endif 226 227 M_CSUM_DATA_IPv6_HL_SET(m->m_pkthdr.csum_data, sizeof(struct ip6_hdr)); 228 229 #define MAKE_EXTHDR(hp, mp) \ 230 do { \ 231 if (hp) { \ 232 struct ip6_ext *eh = (struct ip6_ext *)(hp); \ 233 error = ip6_copyexthdr((mp), (void *)(hp), \ 234 ((eh)->ip6e_len + 1) << 3); \ 235 if (error) \ 236 goto freehdrs; \ 237 } \ 238 } while (/*CONSTCOND*/ 0) 239 240 bzero(&exthdrs, sizeof(exthdrs)); 241 if (opt) { 242 /* Hop-by-Hop options header */ 243 MAKE_EXTHDR(opt->ip6po_hbh, &exthdrs.ip6e_hbh); 244 /* Destination options header(1st part) */ 245 MAKE_EXTHDR(opt->ip6po_dest1, &exthdrs.ip6e_dest1); 246 /* Routing header */ 247 MAKE_EXTHDR(opt->ip6po_rthdr, &exthdrs.ip6e_rthdr); 248 /* Destination options header(2nd part) */ 249 MAKE_EXTHDR(opt->ip6po_dest2, &exthdrs.ip6e_dest2); 250 } 251 252 #ifdef IPSEC 253 if ((flags & IPV6_FORWARDING) != 0) { 254 needipsec = 0; 255 goto skippolicycheck; 256 } 257 258 /* get a security policy for this packet */ 259 if (so == NULL) 260 sp = ipsec6_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND, 0, &error); 261 else { 262 if (IPSEC_PCB_SKIP_IPSEC(sotoinpcb_hdr(so)->inph_sp, 263 IPSEC_DIR_OUTBOUND)) { 264 needipsec = 0; 265 goto skippolicycheck; 266 } 267 sp = ipsec6_getpolicybysock(m, IPSEC_DIR_OUTBOUND, so, &error); 268 } 269 270 if (sp == NULL) { 271 ipsec6stat.out_inval++; 272 goto freehdrs; 273 } 274 275 error = 0; 276 277 /* check policy */ 278 switch (sp->policy) { 279 case IPSEC_POLICY_DISCARD: 280 /* 281 * This packet is just discarded. 282 */ 283 ipsec6stat.out_polvio++; 284 goto freehdrs; 285 286 case IPSEC_POLICY_BYPASS: 287 case IPSEC_POLICY_NONE: 288 /* no need to do IPsec. */ 289 needipsec = 0; 290 break; 291 292 case IPSEC_POLICY_IPSEC: 293 if (sp->req == NULL) { 294 /* XXX should be panic ? */ 295 printf("ip6_output: No IPsec request specified.\n"); 296 error = EINVAL; 297 goto freehdrs; 298 } 299 needipsec = 1; 300 break; 301 302 case IPSEC_POLICY_ENTRUST: 303 default: 304 printf("ip6_output: Invalid policy found. %d\n", sp->policy); 305 } 306 307 skippolicycheck:; 308 #endif /* IPSEC */ 309 310 /* 311 * Calculate the total length of the extension header chain. 312 * Keep the length of the unfragmentable part for fragmentation. 313 */ 314 optlen = 0; 315 if (exthdrs.ip6e_hbh) optlen += exthdrs.ip6e_hbh->m_len; 316 if (exthdrs.ip6e_dest1) optlen += exthdrs.ip6e_dest1->m_len; 317 if (exthdrs.ip6e_rthdr) optlen += exthdrs.ip6e_rthdr->m_len; 318 unfragpartlen = optlen + sizeof(struct ip6_hdr); 319 /* NOTE: we don't add AH/ESP length here. do that later. */ 320 if (exthdrs.ip6e_dest2) optlen += exthdrs.ip6e_dest2->m_len; 321 322 #ifdef FAST_IPSEC 323 /* Check the security policy (SP) for the packet */ 324 325 /* XXX For moment, we doesn't support packet with extented action */ 326 if (optlen !=0) 327 goto freehdrs; 328 329 sp = ipsec6_check_policy(m,so,flags,&needipsec,&error); 330 if (error != 0) { 331 /* 332 * Hack: -EINVAL is used to signal that a packet 333 * should be silently discarded. This is typically 334 * because we asked key management for an SA and 335 * it was delayed (e.g. kicked up to IKE). 336 */ 337 if (error == -EINVAL) 338 error = 0; 339 goto freehdrs; 340 } 341 #endif /* FAST_IPSEC */ 342 343 344 if (needipsec && 345 (m->m_pkthdr.csum_flags & (M_CSUM_UDPv6|M_CSUM_TCPv6)) != 0) { 346 in6_delayed_cksum(m); 347 m->m_pkthdr.csum_flags &= ~(M_CSUM_UDPv6|M_CSUM_TCPv6); 348 } 349 350 351 /* 352 * If we need IPsec, or there is at least one extension header, 353 * separate IP6 header from the payload. 354 */ 355 if ((needipsec || optlen) && !hdrsplit) { 356 if ((error = ip6_splithdr(m, &exthdrs)) != 0) { 357 m = NULL; 358 goto freehdrs; 359 } 360 m = exthdrs.ip6e_ip6; 361 hdrsplit++; 362 } 363 364 /* adjust pointer */ 365 ip6 = mtod(m, struct ip6_hdr *); 366 367 /* adjust mbuf packet header length */ 368 m->m_pkthdr.len += optlen; 369 plen = m->m_pkthdr.len - sizeof(*ip6); 370 371 /* If this is a jumbo payload, insert a jumbo payload option. */ 372 if (plen > IPV6_MAXPACKET) { 373 if (!hdrsplit) { 374 if ((error = ip6_splithdr(m, &exthdrs)) != 0) { 375 m = NULL; 376 goto freehdrs; 377 } 378 m = exthdrs.ip6e_ip6; 379 hdrsplit++; 380 } 381 /* adjust pointer */ 382 ip6 = mtod(m, struct ip6_hdr *); 383 if ((error = ip6_insert_jumboopt(&exthdrs, plen)) != 0) 384 goto freehdrs; 385 optlen += 8; /* XXX JUMBOOPTLEN */ 386 ip6->ip6_plen = 0; 387 } else 388 ip6->ip6_plen = htons(plen); 389 390 /* 391 * Concatenate headers and fill in next header fields. 392 * Here we have, on "m" 393 * IPv6 payload 394 * and we insert headers accordingly. Finally, we should be getting: 395 * IPv6 hbh dest1 rthdr ah* [esp* dest2 payload] 396 * 397 * during the header composing process, "m" points to IPv6 header. 398 * "mprev" points to an extension header prior to esp. 399 */ 400 { 401 u_char *nexthdrp = &ip6->ip6_nxt; 402 struct mbuf *mprev = m; 403 404 /* 405 * we treat dest2 specially. this makes IPsec processing 406 * much easier. the goal here is to make mprev point the 407 * mbuf prior to dest2. 408 * 409 * result: IPv6 dest2 payload 410 * m and mprev will point to IPv6 header. 411 */ 412 if (exthdrs.ip6e_dest2) { 413 if (!hdrsplit) 414 panic("assumption failed: hdr not split"); 415 exthdrs.ip6e_dest2->m_next = m->m_next; 416 m->m_next = exthdrs.ip6e_dest2; 417 *mtod(exthdrs.ip6e_dest2, u_char *) = ip6->ip6_nxt; 418 ip6->ip6_nxt = IPPROTO_DSTOPTS; 419 } 420 421 #define MAKE_CHAIN(m, mp, p, i)\ 422 do {\ 423 if (m) {\ 424 if (!hdrsplit) \ 425 panic("assumption failed: hdr not split"); \ 426 *mtod((m), u_char *) = *(p);\ 427 *(p) = (i);\ 428 p = mtod((m), u_char *);\ 429 (m)->m_next = (mp)->m_next;\ 430 (mp)->m_next = (m);\ 431 (mp) = (m);\ 432 }\ 433 } while (/*CONSTCOND*/ 0) 434 /* 435 * result: IPv6 hbh dest1 rthdr dest2 payload 436 * m will point to IPv6 header. mprev will point to the 437 * extension header prior to dest2 (rthdr in the above case). 438 */ 439 MAKE_CHAIN(exthdrs.ip6e_hbh, mprev, nexthdrp, IPPROTO_HOPOPTS); 440 MAKE_CHAIN(exthdrs.ip6e_dest1, mprev, nexthdrp, 441 IPPROTO_DSTOPTS); 442 MAKE_CHAIN(exthdrs.ip6e_rthdr, mprev, nexthdrp, 443 IPPROTO_ROUTING); 444 445 M_CSUM_DATA_IPv6_HL_SET(m->m_pkthdr.csum_data, 446 sizeof(struct ip6_hdr) + optlen); 447 448 #ifdef IPSEC 449 if (!needipsec) 450 goto skip_ipsec2; 451 452 /* 453 * pointers after IPsec headers are not valid any more. 454 * other pointers need a great care too. 455 * (IPsec routines should not mangle mbufs prior to AH/ESP) 456 */ 457 exthdrs.ip6e_dest2 = NULL; 458 459 { 460 struct ip6_rthdr *rh = NULL; 461 int segleft_org = 0; 462 struct ipsec_output_state state; 463 464 if (exthdrs.ip6e_rthdr) { 465 rh = mtod(exthdrs.ip6e_rthdr, struct ip6_rthdr *); 466 segleft_org = rh->ip6r_segleft; 467 rh->ip6r_segleft = 0; 468 } 469 470 bzero(&state, sizeof(state)); 471 state.m = m; 472 error = ipsec6_output_trans(&state, nexthdrp, mprev, sp, flags, 473 &needipsectun); 474 m = state.m; 475 if (error) { 476 rh = mtod(exthdrs.ip6e_rthdr, struct ip6_rthdr *); 477 /* mbuf is already reclaimed in ipsec6_output_trans. */ 478 m = NULL; 479 switch (error) { 480 case EHOSTUNREACH: 481 case ENETUNREACH: 482 case EMSGSIZE: 483 case ENOBUFS: 484 case ENOMEM: 485 break; 486 default: 487 printf("ip6_output (ipsec): error code %d\n", error); 488 /* FALLTHROUGH */ 489 case ENOENT: 490 /* don't show these error codes to the user */ 491 error = 0; 492 break; 493 } 494 goto bad; 495 } 496 if (exthdrs.ip6e_rthdr) { 497 /* ah6_output doesn't modify mbuf chain */ 498 rh->ip6r_segleft = segleft_org; 499 } 500 } 501 skip_ipsec2:; 502 #endif 503 } 504 505 /* 506 * If there is a routing header, replace destination address field 507 * with the first hop of the routing header. 508 */ 509 if (exthdrs.ip6e_rthdr) { 510 struct ip6_rthdr *rh; 511 struct ip6_rthdr0 *rh0; 512 struct in6_addr *addr; 513 struct sockaddr_in6 sa; 514 515 rh = (struct ip6_rthdr *)(mtod(exthdrs.ip6e_rthdr, 516 struct ip6_rthdr *)); 517 finaldst = ip6->ip6_dst; 518 switch (rh->ip6r_type) { 519 case IPV6_RTHDR_TYPE_0: 520 rh0 = (struct ip6_rthdr0 *)rh; 521 addr = (struct in6_addr *)(rh0 + 1); 522 523 /* 524 * construct a sockaddr_in6 form of 525 * the first hop. 526 * 527 * XXX: we may not have enough 528 * information about its scope zone; 529 * there is no standard API to pass 530 * the information from the 531 * application. 532 */ 533 sockaddr_in6_init(&sa, addr, 0, 0, 0); 534 if ((error = sa6_embedscope(&sa, 535 ip6_use_defzone)) != 0) { 536 goto bad; 537 } 538 ip6->ip6_dst = sa.sin6_addr; 539 (void)memmove(&addr[0], &addr[1], 540 sizeof(struct in6_addr) * 541 (rh0->ip6r0_segleft - 1)); 542 addr[rh0->ip6r0_segleft - 1] = finaldst; 543 /* XXX */ 544 in6_clearscope(addr + rh0->ip6r0_segleft - 1); 545 break; 546 default: /* is it possible? */ 547 error = EINVAL; 548 goto bad; 549 } 550 } 551 552 /* Source address validation */ 553 if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) && 554 (flags & IPV6_UNSPECSRC) == 0) { 555 error = EOPNOTSUPP; 556 IP6_STATINC(IP6_STAT_BADSCOPE); 557 goto bad; 558 } 559 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) { 560 error = EOPNOTSUPP; 561 IP6_STATINC(IP6_STAT_BADSCOPE); 562 goto bad; 563 } 564 565 IP6_STATINC(IP6_STAT_LOCALOUT); 566 567 /* 568 * Route packet. 569 */ 570 /* initialize cached route */ 571 if (ro == NULL) { 572 ro = &ip6route; 573 } 574 ro_pmtu = ro; 575 if (opt && opt->ip6po_rthdr) 576 ro = &opt->ip6po_route; 577 578 /* 579 * if specified, try to fill in the traffic class field. 580 * do not override if a non-zero value is already set. 581 * we check the diffserv field and the ecn field separately. 582 */ 583 if (opt && opt->ip6po_tclass >= 0) { 584 int mask = 0; 585 586 if ((ip6->ip6_flow & htonl(0xfc << 20)) == 0) 587 mask |= 0xfc; 588 if ((ip6->ip6_flow & htonl(0x03 << 20)) == 0) 589 mask |= 0x03; 590 if (mask != 0) 591 ip6->ip6_flow |= htonl((opt->ip6po_tclass & mask) << 20); 592 } 593 594 /* fill in or override the hop limit field, if necessary. */ 595 if (opt && opt->ip6po_hlim != -1) 596 ip6->ip6_hlim = opt->ip6po_hlim & 0xff; 597 else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) { 598 if (im6o != NULL) 599 ip6->ip6_hlim = im6o->im6o_multicast_hlim; 600 else 601 ip6->ip6_hlim = ip6_defmcasthlim; 602 } 603 604 #ifdef IPSEC 605 if (needipsec && needipsectun) { 606 struct ipsec_output_state state; 607 608 /* 609 * All the extension headers will become inaccessible 610 * (since they can be encrypted). 611 * Don't panic, we need no more updates to extension headers 612 * on inner IPv6 packet (since they are now encapsulated). 613 * 614 * IPv6 [ESP|AH] IPv6 [extension headers] payload 615 */ 616 bzero(&exthdrs, sizeof(exthdrs)); 617 exthdrs.ip6e_ip6 = m; 618 619 bzero(&state, sizeof(state)); 620 state.m = m; 621 state.ro = ro; 622 state.dst = rtcache_getdst(ro); 623 624 error = ipsec6_output_tunnel(&state, sp, flags); 625 626 m = state.m; 627 ro_pmtu = ro = state.ro; 628 dst = satocsin6(state.dst); 629 if (error) { 630 /* mbuf is already reclaimed in ipsec6_output_tunnel. */ 631 m0 = m = NULL; 632 m = NULL; 633 switch (error) { 634 case EHOSTUNREACH: 635 case ENETUNREACH: 636 case EMSGSIZE: 637 case ENOBUFS: 638 case ENOMEM: 639 break; 640 default: 641 printf("ip6_output (ipsec): error code %d\n", error); 642 /* FALLTHROUGH */ 643 case ENOENT: 644 /* don't show these error codes to the user */ 645 error = 0; 646 break; 647 } 648 goto bad; 649 } 650 651 exthdrs.ip6e_ip6 = m; 652 } 653 #endif /* IPSEC */ 654 #ifdef FAST_IPSEC 655 if (needipsec) { 656 s = splsoftnet(); 657 error = ipsec6_process_packet(m,sp->req); 658 659 /* 660 * Preserve KAME behaviour: ENOENT can be returned 661 * when an SA acquire is in progress. Don't propagate 662 * this to user-level; it confuses applications. 663 * XXX this will go away when the SADB is redone. 664 */ 665 if (error == ENOENT) 666 error = 0; 667 splx(s); 668 goto done; 669 } 670 #endif /* FAST_IPSEC */ 671 672 673 674 /* adjust pointer */ 675 ip6 = mtod(m, struct ip6_hdr *); 676 677 sockaddr_in6_init(&dst_sa, &ip6->ip6_dst, 0, 0, 0); 678 if ((error = in6_selectroute(&dst_sa, opt, im6o, ro, 679 &ifp, &rt, 0)) != 0) { 680 if (ifp != NULL) 681 in6_ifstat_inc(ifp, ifs6_out_discard); 682 goto bad; 683 } 684 if (rt == NULL) { 685 /* 686 * If in6_selectroute() does not return a route entry, 687 * dst may not have been updated. 688 */ 689 rtcache_setdst(ro, sin6tosa(&dst_sa)); 690 } 691 692 /* 693 * then rt (for unicast) and ifp must be non-NULL valid values. 694 */ 695 if ((flags & IPV6_FORWARDING) == 0) { 696 /* XXX: the FORWARDING flag can be set for mrouting. */ 697 in6_ifstat_inc(ifp, ifs6_out_request); 698 } 699 if (rt != NULL) { 700 ia = (struct in6_ifaddr *)(rt->rt_ifa); 701 rt->rt_use++; 702 } 703 704 /* 705 * The outgoing interface must be in the zone of source and 706 * destination addresses. We should use ia_ifp to support the 707 * case of sending packets to an address of our own. 708 */ 709 if (ia != NULL && ia->ia_ifp) 710 origifp = ia->ia_ifp; 711 else 712 origifp = ifp; 713 714 src0 = ip6->ip6_src; 715 if (in6_setscope(&src0, origifp, &zone)) 716 goto badscope; 717 sockaddr_in6_init(&src_sa, &ip6->ip6_src, 0, 0, 0); 718 if (sa6_recoverscope(&src_sa) || zone != src_sa.sin6_scope_id) 719 goto badscope; 720 721 dst0 = ip6->ip6_dst; 722 if (in6_setscope(&dst0, origifp, &zone)) 723 goto badscope; 724 /* re-initialize to be sure */ 725 sockaddr_in6_init(&dst_sa, &ip6->ip6_dst, 0, 0, 0); 726 if (sa6_recoverscope(&dst_sa) || zone != dst_sa.sin6_scope_id) 727 goto badscope; 728 729 /* scope check is done. */ 730 731 if (rt == NULL || IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) { 732 if (dst == NULL) 733 dst = satocsin6(rtcache_getdst(ro)); 734 KASSERT(dst != NULL); 735 } else if (opt && rtcache_validate(&opt->ip6po_nextroute) != NULL) { 736 /* 737 * The nexthop is explicitly specified by the 738 * application. We assume the next hop is an IPv6 739 * address. 740 */ 741 dst = (struct sockaddr_in6 *)opt->ip6po_nexthop; 742 } else if ((rt->rt_flags & RTF_GATEWAY)) 743 dst = (struct sockaddr_in6 *)rt->rt_gateway; 744 else if (dst == NULL) 745 dst = satocsin6(rtcache_getdst(ro)); 746 747 /* 748 * XXXXXX: original code follows: 749 */ 750 if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) 751 m->m_flags &= ~(M_BCAST | M_MCAST); /* just in case */ 752 else { 753 struct in6_multi *in6m; 754 755 m->m_flags = (m->m_flags & ~M_BCAST) | M_MCAST; 756 757 in6_ifstat_inc(ifp, ifs6_out_mcast); 758 759 /* 760 * Confirm that the outgoing interface supports multicast. 761 */ 762 if (!(ifp->if_flags & IFF_MULTICAST)) { 763 IP6_STATINC(IP6_STAT_NOROUTE); 764 in6_ifstat_inc(ifp, ifs6_out_discard); 765 error = ENETUNREACH; 766 goto bad; 767 } 768 769 IN6_LOOKUP_MULTI(ip6->ip6_dst, ifp, in6m); 770 if (in6m != NULL && 771 (im6o == NULL || im6o->im6o_multicast_loop)) { 772 /* 773 * If we belong to the destination multicast group 774 * on the outgoing interface, and the caller did not 775 * forbid loopback, loop back a copy. 776 */ 777 KASSERT(dst != NULL); 778 ip6_mloopback(ifp, m, dst); 779 } else { 780 /* 781 * If we are acting as a multicast router, perform 782 * multicast forwarding as if the packet had just 783 * arrived on the interface to which we are about 784 * to send. The multicast forwarding function 785 * recursively calls this function, using the 786 * IPV6_FORWARDING flag to prevent infinite recursion. 787 * 788 * Multicasts that are looped back by ip6_mloopback(), 789 * above, will be forwarded by the ip6_input() routine, 790 * if necessary. 791 */ 792 if (ip6_mrouter && (flags & IPV6_FORWARDING) == 0) { 793 if (ip6_mforward(ip6, ifp, m) != 0) { 794 m_freem(m); 795 goto done; 796 } 797 } 798 } 799 /* 800 * Multicasts with a hoplimit of zero may be looped back, 801 * above, but must not be transmitted on a network. 802 * Also, multicasts addressed to the loopback interface 803 * are not sent -- the above call to ip6_mloopback() will 804 * loop back a copy if this host actually belongs to the 805 * destination group on the loopback interface. 806 */ 807 if (ip6->ip6_hlim == 0 || (ifp->if_flags & IFF_LOOPBACK) || 808 IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst)) { 809 m_freem(m); 810 goto done; 811 } 812 } 813 814 /* 815 * Fill the outgoing inteface to tell the upper layer 816 * to increment per-interface statistics. 817 */ 818 if (ifpp) 819 *ifpp = ifp; 820 821 /* Determine path MTU. */ 822 if ((error = ip6_getpmtu(ro_pmtu, ro, ifp, &finaldst, &mtu, 823 &alwaysfrag)) != 0) 824 goto bad; 825 #ifdef IPSEC 826 if (needipsectun) 827 mtu = IPV6_MMTU; 828 #endif 829 830 /* 831 * The caller of this function may specify to use the minimum MTU 832 * in some cases. 833 * An advanced API option (IPV6_USE_MIN_MTU) can also override MTU 834 * setting. The logic is a bit complicated; by default, unicast 835 * packets will follow path MTU while multicast packets will be sent at 836 * the minimum MTU. If IP6PO_MINMTU_ALL is specified, all packets 837 * including unicast ones will be sent at the minimum MTU. Multicast 838 * packets will always be sent at the minimum MTU unless 839 * IP6PO_MINMTU_DISABLE is explicitly specified. 840 * See RFC 3542 for more details. 841 */ 842 if (mtu > IPV6_MMTU) { 843 if ((flags & IPV6_MINMTU)) 844 mtu = IPV6_MMTU; 845 else if (opt && opt->ip6po_minmtu == IP6PO_MINMTU_ALL) 846 mtu = IPV6_MMTU; 847 else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) && 848 (opt == NULL || 849 opt->ip6po_minmtu != IP6PO_MINMTU_DISABLE)) { 850 mtu = IPV6_MMTU; 851 } 852 } 853 854 /* 855 * clear embedded scope identifiers if necessary. 856 * in6_clearscope will touch the addresses only when necessary. 857 */ 858 in6_clearscope(&ip6->ip6_src); 859 in6_clearscope(&ip6->ip6_dst); 860 861 /* 862 * If the outgoing packet contains a hop-by-hop options header, 863 * it must be examined and processed even by the source node. 864 * (RFC 2460, section 4.) 865 */ 866 if (exthdrs.ip6e_hbh) { 867 struct ip6_hbh *hbh = mtod(exthdrs.ip6e_hbh, struct ip6_hbh *); 868 u_int32_t dummy1; /* XXX unused */ 869 u_int32_t dummy2; /* XXX unused */ 870 871 /* 872 * XXX: if we have to send an ICMPv6 error to the sender, 873 * we need the M_LOOP flag since icmp6_error() expects 874 * the IPv6 and the hop-by-hop options header are 875 * continuous unless the flag is set. 876 */ 877 m->m_flags |= M_LOOP; 878 m->m_pkthdr.rcvif = ifp; 879 if (ip6_process_hopopts(m, (u_int8_t *)(hbh + 1), 880 ((hbh->ip6h_len + 1) << 3) - sizeof(struct ip6_hbh), 881 &dummy1, &dummy2) < 0) { 882 /* m was already freed at this point */ 883 error = EINVAL;/* better error? */ 884 goto done; 885 } 886 m->m_flags &= ~M_LOOP; /* XXX */ 887 m->m_pkthdr.rcvif = NULL; 888 } 889 890 #ifdef PFIL_HOOKS 891 /* 892 * Run through list of hooks for output packets. 893 */ 894 if ((error = pfil_run_hooks(&inet6_pfil_hook, &m, ifp, PFIL_OUT)) != 0) 895 goto done; 896 if (m == NULL) 897 goto done; 898 ip6 = mtod(m, struct ip6_hdr *); 899 #endif /* PFIL_HOOKS */ 900 /* 901 * Send the packet to the outgoing interface. 902 * If necessary, do IPv6 fragmentation before sending. 903 * 904 * the logic here is rather complex: 905 * 1: normal case (dontfrag == 0, alwaysfrag == 0) 906 * 1-a: send as is if tlen <= path mtu 907 * 1-b: fragment if tlen > path mtu 908 * 909 * 2: if user asks us not to fragment (dontfrag == 1) 910 * 2-a: send as is if tlen <= interface mtu 911 * 2-b: error if tlen > interface mtu 912 * 913 * 3: if we always need to attach fragment header (alwaysfrag == 1) 914 * always fragment 915 * 916 * 4: if dontfrag == 1 && alwaysfrag == 1 917 * error, as we cannot handle this conflicting request 918 */ 919 tlen = m->m_pkthdr.len; 920 tso = (m->m_pkthdr.csum_flags & M_CSUM_TSOv6) != 0; 921 if (opt && (opt->ip6po_flags & IP6PO_DONTFRAG)) 922 dontfrag = 1; 923 else 924 dontfrag = 0; 925 926 if (dontfrag && alwaysfrag) { /* case 4 */ 927 /* conflicting request - can't transmit */ 928 error = EMSGSIZE; 929 goto bad; 930 } 931 if (dontfrag && (!tso && tlen > IN6_LINKMTU(ifp))) { /* case 2-b */ 932 /* 933 * Even if the DONTFRAG option is specified, we cannot send the 934 * packet when the data length is larger than the MTU of the 935 * outgoing interface. 936 * Notify the error by sending IPV6_PATHMTU ancillary data as 937 * well as returning an error code (the latter is not described 938 * in the API spec.) 939 */ 940 u_int32_t mtu32; 941 struct ip6ctlparam ip6cp; 942 943 mtu32 = (u_int32_t)mtu; 944 bzero(&ip6cp, sizeof(ip6cp)); 945 ip6cp.ip6c_cmdarg = (void *)&mtu32; 946 pfctlinput2(PRC_MSGSIZE, 947 rtcache_getdst(ro_pmtu), &ip6cp); 948 949 error = EMSGSIZE; 950 goto bad; 951 } 952 953 /* 954 * transmit packet without fragmentation 955 */ 956 if (dontfrag || (!alwaysfrag && (tlen <= mtu || tso))) { 957 /* case 1-a and 2-a */ 958 struct in6_ifaddr *ia6; 959 int sw_csum; 960 961 ip6 = mtod(m, struct ip6_hdr *); 962 ia6 = in6_ifawithifp(ifp, &ip6->ip6_src); 963 if (ia6) { 964 /* Record statistics for this interface address. */ 965 ia6->ia_ifa.ifa_data.ifad_outbytes += m->m_pkthdr.len; 966 } 967 #ifdef IPSEC 968 /* clean ipsec history once it goes out of the node */ 969 ipsec_delaux(m); 970 #endif 971 972 sw_csum = m->m_pkthdr.csum_flags & ~ifp->if_csum_flags_tx; 973 if ((sw_csum & (M_CSUM_UDPv6|M_CSUM_TCPv6)) != 0) { 974 if (IN6_NEED_CHECKSUM(ifp, 975 sw_csum & (M_CSUM_UDPv6|M_CSUM_TCPv6))) { 976 in6_delayed_cksum(m); 977 } 978 m->m_pkthdr.csum_flags &= ~(M_CSUM_UDPv6|M_CSUM_TCPv6); 979 } 980 981 KASSERT(dst != NULL); 982 if (__predict_true(!tso || 983 (ifp->if_capenable & IFCAP_TSOv6) != 0)) { 984 error = nd6_output(ifp, origifp, m, dst, rt); 985 } else { 986 error = ip6_tso_output(ifp, origifp, m, dst, rt); 987 } 988 goto done; 989 } 990 991 if (tso) { 992 error = EINVAL; /* XXX */ 993 goto bad; 994 } 995 996 /* 997 * try to fragment the packet. case 1-b and 3 998 */ 999 if (mtu < IPV6_MMTU) { 1000 /* path MTU cannot be less than IPV6_MMTU */ 1001 error = EMSGSIZE; 1002 in6_ifstat_inc(ifp, ifs6_out_fragfail); 1003 goto bad; 1004 } else if (ip6->ip6_plen == 0) { 1005 /* jumbo payload cannot be fragmented */ 1006 error = EMSGSIZE; 1007 in6_ifstat_inc(ifp, ifs6_out_fragfail); 1008 goto bad; 1009 } else { 1010 struct mbuf **mnext, *m_frgpart; 1011 struct ip6_frag *ip6f; 1012 u_int32_t id = htonl(ip6_randomid()); 1013 u_char nextproto; 1014 #if 0 /* see below */ 1015 struct ip6ctlparam ip6cp; 1016 u_int32_t mtu32; 1017 #endif 1018 1019 /* 1020 * Too large for the destination or interface; 1021 * fragment if possible. 1022 * Must be able to put at least 8 bytes per fragment. 1023 */ 1024 hlen = unfragpartlen; 1025 if (mtu > IPV6_MAXPACKET) 1026 mtu = IPV6_MAXPACKET; 1027 1028 #if 0 1029 /* 1030 * It is believed this code is a leftover from the 1031 * development of the IPV6_RECVPATHMTU sockopt and 1032 * associated work to implement RFC3542. 1033 * It's not entirely clear what the intent of the API 1034 * is at this point, so disable this code for now. 1035 * The IPV6_RECVPATHMTU sockopt and/or IPV6_DONTFRAG 1036 * will send notifications if the application requests. 1037 */ 1038 1039 /* Notify a proper path MTU to applications. */ 1040 mtu32 = (u_int32_t)mtu; 1041 bzero(&ip6cp, sizeof(ip6cp)); 1042 ip6cp.ip6c_cmdarg = (void *)&mtu32; 1043 pfctlinput2(PRC_MSGSIZE, 1044 rtcache_getdst(ro_pmtu), &ip6cp); 1045 #endif 1046 1047 len = (mtu - hlen - sizeof(struct ip6_frag)) & ~7; 1048 if (len < 8) { 1049 error = EMSGSIZE; 1050 in6_ifstat_inc(ifp, ifs6_out_fragfail); 1051 goto bad; 1052 } 1053 1054 mnext = &m->m_nextpkt; 1055 1056 /* 1057 * Change the next header field of the last header in the 1058 * unfragmentable part. 1059 */ 1060 if (exthdrs.ip6e_rthdr) { 1061 nextproto = *mtod(exthdrs.ip6e_rthdr, u_char *); 1062 *mtod(exthdrs.ip6e_rthdr, u_char *) = IPPROTO_FRAGMENT; 1063 } else if (exthdrs.ip6e_dest1) { 1064 nextproto = *mtod(exthdrs.ip6e_dest1, u_char *); 1065 *mtod(exthdrs.ip6e_dest1, u_char *) = IPPROTO_FRAGMENT; 1066 } else if (exthdrs.ip6e_hbh) { 1067 nextproto = *mtod(exthdrs.ip6e_hbh, u_char *); 1068 *mtod(exthdrs.ip6e_hbh, u_char *) = IPPROTO_FRAGMENT; 1069 } else { 1070 nextproto = ip6->ip6_nxt; 1071 ip6->ip6_nxt = IPPROTO_FRAGMENT; 1072 } 1073 1074 if ((m->m_pkthdr.csum_flags & (M_CSUM_UDPv6|M_CSUM_TCPv6)) 1075 != 0) { 1076 if (IN6_NEED_CHECKSUM(ifp, 1077 m->m_pkthdr.csum_flags & 1078 (M_CSUM_UDPv6|M_CSUM_TCPv6))) { 1079 in6_delayed_cksum(m); 1080 } 1081 m->m_pkthdr.csum_flags &= ~(M_CSUM_UDPv6|M_CSUM_TCPv6); 1082 } 1083 1084 /* 1085 * Loop through length of segment after first fragment, 1086 * make new header and copy data of each part and link onto 1087 * chain. 1088 */ 1089 m0 = m; 1090 for (off = hlen; off < tlen; off += len) { 1091 struct mbuf *mlast; 1092 1093 MGETHDR(m, M_DONTWAIT, MT_HEADER); 1094 if (!m) { 1095 error = ENOBUFS; 1096 IP6_STATINC(IP6_STAT_ODROPPED); 1097 goto sendorfree; 1098 } 1099 m->m_pkthdr.rcvif = NULL; 1100 m->m_flags = m0->m_flags & M_COPYFLAGS; 1101 *mnext = m; 1102 mnext = &m->m_nextpkt; 1103 m->m_data += max_linkhdr; 1104 mhip6 = mtod(m, struct ip6_hdr *); 1105 *mhip6 = *ip6; 1106 m->m_len = sizeof(*mhip6); 1107 error = ip6_insertfraghdr(m0, m, hlen, &ip6f); 1108 if (error) { 1109 IP6_STATINC(IP6_STAT_ODROPPED); 1110 goto sendorfree; 1111 } 1112 ip6f->ip6f_offlg = htons((u_int16_t)((off - hlen) & ~7)); 1113 if (off + len >= tlen) 1114 len = tlen - off; 1115 else 1116 ip6f->ip6f_offlg |= IP6F_MORE_FRAG; 1117 mhip6->ip6_plen = htons((u_int16_t)(len + hlen + 1118 sizeof(*ip6f) - sizeof(struct ip6_hdr))); 1119 if ((m_frgpart = m_copy(m0, off, len)) == 0) { 1120 error = ENOBUFS; 1121 IP6_STATINC(IP6_STAT_ODROPPED); 1122 goto sendorfree; 1123 } 1124 for (mlast = m; mlast->m_next; mlast = mlast->m_next) 1125 ; 1126 mlast->m_next = m_frgpart; 1127 m->m_pkthdr.len = len + hlen + sizeof(*ip6f); 1128 m->m_pkthdr.rcvif = (struct ifnet *)0; 1129 ip6f->ip6f_reserved = 0; 1130 ip6f->ip6f_ident = id; 1131 ip6f->ip6f_nxt = nextproto; 1132 IP6_STATINC(IP6_STAT_OFRAGMENTS); 1133 in6_ifstat_inc(ifp, ifs6_out_fragcreat); 1134 } 1135 1136 in6_ifstat_inc(ifp, ifs6_out_fragok); 1137 } 1138 1139 /* 1140 * Remove leading garbages. 1141 */ 1142 sendorfree: 1143 m = m0->m_nextpkt; 1144 m0->m_nextpkt = 0; 1145 m_freem(m0); 1146 for (m0 = m; m; m = m0) { 1147 m0 = m->m_nextpkt; 1148 m->m_nextpkt = 0; 1149 if (error == 0) { 1150 struct in6_ifaddr *ia6; 1151 ip6 = mtod(m, struct ip6_hdr *); 1152 ia6 = in6_ifawithifp(ifp, &ip6->ip6_src); 1153 if (ia6) { 1154 /* 1155 * Record statistics for this interface 1156 * address. 1157 */ 1158 ia6->ia_ifa.ifa_data.ifad_outbytes += 1159 m->m_pkthdr.len; 1160 } 1161 #ifdef IPSEC 1162 /* clean ipsec history once it goes out of the node */ 1163 ipsec_delaux(m); 1164 #endif 1165 KASSERT(dst != NULL); 1166 error = nd6_output(ifp, origifp, m, dst, rt); 1167 } else 1168 m_freem(m); 1169 } 1170 1171 if (error == 0) 1172 IP6_STATINC(IP6_STAT_FRAGMENTED); 1173 1174 done: 1175 rtcache_free(&ip6route); 1176 1177 #ifdef IPSEC 1178 if (sp != NULL) 1179 key_freesp(sp); 1180 #endif /* IPSEC */ 1181 #ifdef FAST_IPSEC 1182 if (sp != NULL) 1183 KEY_FREESP(&sp); 1184 #endif /* FAST_IPSEC */ 1185 1186 1187 return (error); 1188 1189 freehdrs: 1190 m_freem(exthdrs.ip6e_hbh); /* m_freem will check if mbuf is 0 */ 1191 m_freem(exthdrs.ip6e_dest1); 1192 m_freem(exthdrs.ip6e_rthdr); 1193 m_freem(exthdrs.ip6e_dest2); 1194 /* FALLTHROUGH */ 1195 bad: 1196 m_freem(m); 1197 goto done; 1198 badscope: 1199 IP6_STATINC(IP6_STAT_BADSCOPE); 1200 in6_ifstat_inc(origifp, ifs6_out_discard); 1201 if (error == 0) 1202 error = EHOSTUNREACH; /* XXX */ 1203 goto bad; 1204 } 1205 1206 static int 1207 ip6_copyexthdr(struct mbuf **mp, void *hdr, int hlen) 1208 { 1209 struct mbuf *m; 1210 1211 if (hlen > MCLBYTES) 1212 return (ENOBUFS); /* XXX */ 1213 1214 MGET(m, M_DONTWAIT, MT_DATA); 1215 if (!m) 1216 return (ENOBUFS); 1217 1218 if (hlen > MLEN) { 1219 MCLGET(m, M_DONTWAIT); 1220 if ((m->m_flags & M_EXT) == 0) { 1221 m_free(m); 1222 return (ENOBUFS); 1223 } 1224 } 1225 m->m_len = hlen; 1226 if (hdr) 1227 bcopy(hdr, mtod(m, void *), hlen); 1228 1229 *mp = m; 1230 return (0); 1231 } 1232 1233 /* 1234 * Process a delayed payload checksum calculation. 1235 */ 1236 void 1237 in6_delayed_cksum(struct mbuf *m) 1238 { 1239 uint16_t csum, offset; 1240 1241 KASSERT((m->m_pkthdr.csum_flags & (M_CSUM_UDPv6|M_CSUM_TCPv6)) != 0); 1242 KASSERT((~m->m_pkthdr.csum_flags & (M_CSUM_UDPv6|M_CSUM_TCPv6)) != 0); 1243 KASSERT((m->m_pkthdr.csum_flags 1244 & (M_CSUM_UDPv4|M_CSUM_TCPv4|M_CSUM_TSOv4)) == 0); 1245 1246 offset = M_CSUM_DATA_IPv6_HL(m->m_pkthdr.csum_data); 1247 csum = in6_cksum(m, 0, offset, m->m_pkthdr.len - offset); 1248 if (csum == 0 && (m->m_pkthdr.csum_flags & M_CSUM_UDPv6) != 0) { 1249 csum = 0xffff; 1250 } 1251 1252 offset += M_CSUM_DATA_IPv6_OFFSET(m->m_pkthdr.csum_data); 1253 if ((offset + sizeof(csum)) > m->m_len) { 1254 m_copyback(m, offset, sizeof(csum), &csum); 1255 } else { 1256 *(uint16_t *)(mtod(m, char *) + offset) = csum; 1257 } 1258 } 1259 1260 /* 1261 * Insert jumbo payload option. 1262 */ 1263 static int 1264 ip6_insert_jumboopt(struct ip6_exthdrs *exthdrs, u_int32_t plen) 1265 { 1266 struct mbuf *mopt; 1267 u_int8_t *optbuf; 1268 u_int32_t v; 1269 1270 #define JUMBOOPTLEN 8 /* length of jumbo payload option and padding */ 1271 1272 /* 1273 * If there is no hop-by-hop options header, allocate new one. 1274 * If there is one but it doesn't have enough space to store the 1275 * jumbo payload option, allocate a cluster to store the whole options. 1276 * Otherwise, use it to store the options. 1277 */ 1278 if (exthdrs->ip6e_hbh == 0) { 1279 MGET(mopt, M_DONTWAIT, MT_DATA); 1280 if (mopt == 0) 1281 return (ENOBUFS); 1282 mopt->m_len = JUMBOOPTLEN; 1283 optbuf = mtod(mopt, u_int8_t *); 1284 optbuf[1] = 0; /* = ((JUMBOOPTLEN) >> 3) - 1 */ 1285 exthdrs->ip6e_hbh = mopt; 1286 } else { 1287 struct ip6_hbh *hbh; 1288 1289 mopt = exthdrs->ip6e_hbh; 1290 if (M_TRAILINGSPACE(mopt) < JUMBOOPTLEN) { 1291 /* 1292 * XXX assumption: 1293 * - exthdrs->ip6e_hbh is not referenced from places 1294 * other than exthdrs. 1295 * - exthdrs->ip6e_hbh is not an mbuf chain. 1296 */ 1297 int oldoptlen = mopt->m_len; 1298 struct mbuf *n; 1299 1300 /* 1301 * XXX: give up if the whole (new) hbh header does 1302 * not fit even in an mbuf cluster. 1303 */ 1304 if (oldoptlen + JUMBOOPTLEN > MCLBYTES) 1305 return (ENOBUFS); 1306 1307 /* 1308 * As a consequence, we must always prepare a cluster 1309 * at this point. 1310 */ 1311 MGET(n, M_DONTWAIT, MT_DATA); 1312 if (n) { 1313 MCLGET(n, M_DONTWAIT); 1314 if ((n->m_flags & M_EXT) == 0) { 1315 m_freem(n); 1316 n = NULL; 1317 } 1318 } 1319 if (!n) 1320 return (ENOBUFS); 1321 n->m_len = oldoptlen + JUMBOOPTLEN; 1322 bcopy(mtod(mopt, void *), mtod(n, void *), 1323 oldoptlen); 1324 optbuf = mtod(n, u_int8_t *) + oldoptlen; 1325 m_freem(mopt); 1326 mopt = exthdrs->ip6e_hbh = n; 1327 } else { 1328 optbuf = mtod(mopt, u_int8_t *) + mopt->m_len; 1329 mopt->m_len += JUMBOOPTLEN; 1330 } 1331 optbuf[0] = IP6OPT_PADN; 1332 optbuf[1] = 0; 1333 1334 /* 1335 * Adjust the header length according to the pad and 1336 * the jumbo payload option. 1337 */ 1338 hbh = mtod(mopt, struct ip6_hbh *); 1339 hbh->ip6h_len += (JUMBOOPTLEN >> 3); 1340 } 1341 1342 /* fill in the option. */ 1343 optbuf[2] = IP6OPT_JUMBO; 1344 optbuf[3] = 4; 1345 v = (u_int32_t)htonl(plen + JUMBOOPTLEN); 1346 bcopy(&v, &optbuf[4], sizeof(u_int32_t)); 1347 1348 /* finally, adjust the packet header length */ 1349 exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN; 1350 1351 return (0); 1352 #undef JUMBOOPTLEN 1353 } 1354 1355 /* 1356 * Insert fragment header and copy unfragmentable header portions. 1357 */ 1358 static int 1359 ip6_insertfraghdr(struct mbuf *m0, struct mbuf *m, int hlen, 1360 struct ip6_frag **frghdrp) 1361 { 1362 struct mbuf *n, *mlast; 1363 1364 if (hlen > sizeof(struct ip6_hdr)) { 1365 n = m_copym(m0, sizeof(struct ip6_hdr), 1366 hlen - sizeof(struct ip6_hdr), M_DONTWAIT); 1367 if (n == 0) 1368 return (ENOBUFS); 1369 m->m_next = n; 1370 } else 1371 n = m; 1372 1373 /* Search for the last mbuf of unfragmentable part. */ 1374 for (mlast = n; mlast->m_next; mlast = mlast->m_next) 1375 ; 1376 1377 if ((mlast->m_flags & M_EXT) == 0 && 1378 M_TRAILINGSPACE(mlast) >= sizeof(struct ip6_frag)) { 1379 /* use the trailing space of the last mbuf for the fragment hdr */ 1380 *frghdrp = (struct ip6_frag *)(mtod(mlast, char *) + 1381 mlast->m_len); 1382 mlast->m_len += sizeof(struct ip6_frag); 1383 m->m_pkthdr.len += sizeof(struct ip6_frag); 1384 } else { 1385 /* allocate a new mbuf for the fragment header */ 1386 struct mbuf *mfrg; 1387 1388 MGET(mfrg, M_DONTWAIT, MT_DATA); 1389 if (mfrg == 0) 1390 return (ENOBUFS); 1391 mfrg->m_len = sizeof(struct ip6_frag); 1392 *frghdrp = mtod(mfrg, struct ip6_frag *); 1393 mlast->m_next = mfrg; 1394 } 1395 1396 return (0); 1397 } 1398 1399 static int 1400 ip6_getpmtu(struct route *ro_pmtu, struct route *ro, struct ifnet *ifp, 1401 const struct in6_addr *dst, u_long *mtup, int *alwaysfragp) 1402 { 1403 struct rtentry *rt; 1404 u_int32_t mtu = 0; 1405 int alwaysfrag = 0; 1406 int error = 0; 1407 1408 if (ro_pmtu != ro) { 1409 union { 1410 struct sockaddr dst; 1411 struct sockaddr_in6 dst6; 1412 } u; 1413 1414 /* The first hop and the final destination may differ. */ 1415 sockaddr_in6_init(&u.dst6, dst, 0, 0, 0); 1416 rt = rtcache_lookup(ro_pmtu, &u.dst); 1417 } else 1418 rt = rtcache_validate(ro_pmtu); 1419 if (rt != NULL) { 1420 u_int32_t ifmtu; 1421 1422 if (ifp == NULL) 1423 ifp = rt->rt_ifp; 1424 ifmtu = IN6_LINKMTU(ifp); 1425 mtu = rt->rt_rmx.rmx_mtu; 1426 if (mtu == 0) 1427 mtu = ifmtu; 1428 else if (mtu < IPV6_MMTU) { 1429 /* 1430 * RFC2460 section 5, last paragraph: 1431 * if we record ICMPv6 too big message with 1432 * mtu < IPV6_MMTU, transmit packets sized IPV6_MMTU 1433 * or smaller, with fragment header attached. 1434 * (fragment header is needed regardless from the 1435 * packet size, for translators to identify packets) 1436 */ 1437 alwaysfrag = 1; 1438 mtu = IPV6_MMTU; 1439 } else if (mtu > ifmtu) { 1440 /* 1441 * The MTU on the route is larger than the MTU on 1442 * the interface! This shouldn't happen, unless the 1443 * MTU of the interface has been changed after the 1444 * interface was brought up. Change the MTU in the 1445 * route to match the interface MTU (as long as the 1446 * field isn't locked). 1447 */ 1448 mtu = ifmtu; 1449 if (!(rt->rt_rmx.rmx_locks & RTV_MTU)) 1450 rt->rt_rmx.rmx_mtu = mtu; 1451 } 1452 } else if (ifp) { 1453 mtu = IN6_LINKMTU(ifp); 1454 } else 1455 error = EHOSTUNREACH; /* XXX */ 1456 1457 *mtup = mtu; 1458 if (alwaysfragp) 1459 *alwaysfragp = alwaysfrag; 1460 return (error); 1461 } 1462 1463 /* 1464 * IP6 socket option processing. 1465 */ 1466 int 1467 ip6_ctloutput(int op, struct socket *so, int level, int optname, 1468 struct mbuf **mp) 1469 { 1470 int privileged, optdatalen, uproto; 1471 void *optdata; 1472 struct in6pcb *in6p = sotoin6pcb(so); 1473 struct mbuf *m = *mp; 1474 int error, optval; 1475 int optlen; 1476 struct lwp *l = curlwp; /* XXX */ 1477 1478 optlen = m ? m->m_len : 0; 1479 error = optval = 0; 1480 privileged = (l == 0 || kauth_authorize_generic(l->l_cred, 1481 KAUTH_GENERIC_ISSUSER, NULL)) ? 0 : 1; 1482 uproto = (int)so->so_proto->pr_protocol; 1483 1484 if (level != IPPROTO_IPV6) { 1485 if (op == PRCO_SETOPT && *mp) 1486 (void)m_free(*mp); 1487 return ENOPROTOOPT; 1488 } 1489 switch (op) { 1490 case PRCO_SETOPT: 1491 switch (optname) { 1492 #ifdef RFC2292 1493 case IPV6_2292PKTOPTIONS: 1494 /* m is freed in ip6_pcbopts */ 1495 error = ip6_pcbopts(&in6p->in6p_outputopts, 1496 m, so); 1497 break; 1498 #endif 1499 1500 /* 1501 * Use of some Hop-by-Hop options or some 1502 * Destination options, might require special 1503 * privilege. That is, normal applications 1504 * (without special privilege) might be forbidden 1505 * from setting certain options in outgoing packets, 1506 * and might never see certain options in received 1507 * packets. [RFC 2292 Section 6] 1508 * KAME specific note: 1509 * KAME prevents non-privileged users from sending or 1510 * receiving ANY hbh/dst options in order to avoid 1511 * overhead of parsing options in the kernel. 1512 */ 1513 case IPV6_RECVHOPOPTS: 1514 case IPV6_RECVDSTOPTS: 1515 case IPV6_RECVRTHDRDSTOPTS: 1516 if (!privileged) { 1517 error = EPERM; 1518 break; 1519 } 1520 /* FALLTHROUGH */ 1521 case IPV6_UNICAST_HOPS: 1522 case IPV6_HOPLIMIT: 1523 case IPV6_FAITH: 1524 1525 case IPV6_RECVPKTINFO: 1526 case IPV6_RECVHOPLIMIT: 1527 case IPV6_RECVRTHDR: 1528 case IPV6_RECVPATHMTU: 1529 case IPV6_RECVTCLASS: 1530 case IPV6_V6ONLY: 1531 if (optlen != sizeof(int)) { 1532 error = EINVAL; 1533 break; 1534 } 1535 optval = *mtod(m, int *); 1536 switch (optname) { 1537 1538 case IPV6_UNICAST_HOPS: 1539 if (optval < -1 || optval >= 256) 1540 error = EINVAL; 1541 else { 1542 /* -1 = kernel default */ 1543 in6p->in6p_hops = optval; 1544 } 1545 break; 1546 #define OPTSET(bit) \ 1547 do { \ 1548 if (optval) \ 1549 in6p->in6p_flags |= (bit); \ 1550 else \ 1551 in6p->in6p_flags &= ~(bit); \ 1552 } while (/*CONSTCOND*/ 0) 1553 1554 #ifdef RFC2292 1555 #define OPTSET2292(bit) \ 1556 do { \ 1557 in6p->in6p_flags |= IN6P_RFC2292; \ 1558 if (optval) \ 1559 in6p->in6p_flags |= (bit); \ 1560 else \ 1561 in6p->in6p_flags &= ~(bit); \ 1562 } while (/*CONSTCOND*/ 0) 1563 #endif 1564 1565 #define OPTBIT(bit) (in6p->in6p_flags & (bit) ? 1 : 0) 1566 1567 case IPV6_RECVPKTINFO: 1568 #ifdef RFC2292 1569 /* cannot mix with RFC2292 */ 1570 if (OPTBIT(IN6P_RFC2292)) { 1571 error = EINVAL; 1572 break; 1573 } 1574 #endif 1575 OPTSET(IN6P_PKTINFO); 1576 break; 1577 1578 case IPV6_HOPLIMIT: 1579 { 1580 struct ip6_pktopts **optp; 1581 1582 #ifdef RFC2292 1583 /* cannot mix with RFC2292 */ 1584 if (OPTBIT(IN6P_RFC2292)) { 1585 error = EINVAL; 1586 break; 1587 } 1588 #endif 1589 optp = &in6p->in6p_outputopts; 1590 error = ip6_pcbopt(IPV6_HOPLIMIT, 1591 (u_char *)&optval, 1592 sizeof(optval), 1593 optp, 1594 privileged, uproto); 1595 break; 1596 } 1597 1598 case IPV6_RECVHOPLIMIT: 1599 #ifdef RFC2292 1600 /* cannot mix with RFC2292 */ 1601 if (OPTBIT(IN6P_RFC2292)) { 1602 error = EINVAL; 1603 break; 1604 } 1605 #endif 1606 OPTSET(IN6P_HOPLIMIT); 1607 break; 1608 1609 case IPV6_RECVHOPOPTS: 1610 #ifdef RFC2292 1611 /* cannot mix with RFC2292 */ 1612 if (OPTBIT(IN6P_RFC2292)) { 1613 error = EINVAL; 1614 break; 1615 } 1616 #endif 1617 OPTSET(IN6P_HOPOPTS); 1618 break; 1619 1620 case IPV6_RECVDSTOPTS: 1621 #ifdef RFC2292 1622 /* cannot mix with RFC2292 */ 1623 if (OPTBIT(IN6P_RFC2292)) { 1624 error = EINVAL; 1625 break; 1626 } 1627 #endif 1628 OPTSET(IN6P_DSTOPTS); 1629 break; 1630 1631 case IPV6_RECVRTHDRDSTOPTS: 1632 #ifdef RFC2292 1633 /* cannot mix with RFC2292 */ 1634 if (OPTBIT(IN6P_RFC2292)) { 1635 error = EINVAL; 1636 break; 1637 } 1638 #endif 1639 OPTSET(IN6P_RTHDRDSTOPTS); 1640 break; 1641 1642 case IPV6_RECVRTHDR: 1643 #ifdef RFC2292 1644 /* cannot mix with RFC2292 */ 1645 if (OPTBIT(IN6P_RFC2292)) { 1646 error = EINVAL; 1647 break; 1648 } 1649 #endif 1650 OPTSET(IN6P_RTHDR); 1651 break; 1652 1653 case IPV6_FAITH: 1654 OPTSET(IN6P_FAITH); 1655 break; 1656 1657 case IPV6_RECVPATHMTU: 1658 /* 1659 * We ignore this option for TCP 1660 * sockets. 1661 * (RFC3542 leaves this case 1662 * unspecified.) 1663 */ 1664 if (uproto != IPPROTO_TCP) 1665 OPTSET(IN6P_MTU); 1666 break; 1667 1668 case IPV6_V6ONLY: 1669 /* 1670 * make setsockopt(IPV6_V6ONLY) 1671 * available only prior to bind(2). 1672 * see ipng mailing list, Jun 22 2001. 1673 */ 1674 if (in6p->in6p_lport || 1675 !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr)) { 1676 error = EINVAL; 1677 break; 1678 } 1679 #ifdef INET6_BINDV6ONLY 1680 if (!optval) 1681 error = EINVAL; 1682 #else 1683 OPTSET(IN6P_IPV6_V6ONLY); 1684 #endif 1685 break; 1686 case IPV6_RECVTCLASS: 1687 #ifdef RFC2292 1688 /* cannot mix with RFC2292 XXX */ 1689 if (OPTBIT(IN6P_RFC2292)) { 1690 error = EINVAL; 1691 break; 1692 } 1693 #endif 1694 OPTSET(IN6P_TCLASS); 1695 break; 1696 1697 } 1698 break; 1699 1700 case IPV6_OTCLASS: 1701 { 1702 struct ip6_pktopts **optp; 1703 u_int8_t tclass; 1704 1705 if (optlen != sizeof(tclass)) { 1706 error = EINVAL; 1707 break; 1708 } 1709 tclass = *mtod(m, u_int8_t *); 1710 optp = &in6p->in6p_outputopts; 1711 error = ip6_pcbopt(optname, 1712 (u_char *)&tclass, 1713 sizeof(tclass), 1714 optp, 1715 privileged, uproto); 1716 break; 1717 } 1718 1719 case IPV6_TCLASS: 1720 case IPV6_DONTFRAG: 1721 case IPV6_USE_MIN_MTU: 1722 if (optlen != sizeof(optval)) { 1723 error = EINVAL; 1724 break; 1725 } 1726 optval = *mtod(m, int *); 1727 { 1728 struct ip6_pktopts **optp; 1729 optp = &in6p->in6p_outputopts; 1730 error = ip6_pcbopt(optname, 1731 (u_char *)&optval, 1732 sizeof(optval), 1733 optp, 1734 privileged, uproto); 1735 break; 1736 } 1737 1738 #ifdef RFC2292 1739 case IPV6_2292PKTINFO: 1740 case IPV6_2292HOPLIMIT: 1741 case IPV6_2292HOPOPTS: 1742 case IPV6_2292DSTOPTS: 1743 case IPV6_2292RTHDR: 1744 /* RFC 2292 */ 1745 if (optlen != sizeof(int)) { 1746 error = EINVAL; 1747 break; 1748 } 1749 optval = *mtod(m, int *); 1750 switch (optname) { 1751 case IPV6_2292PKTINFO: 1752 OPTSET2292(IN6P_PKTINFO); 1753 break; 1754 case IPV6_2292HOPLIMIT: 1755 OPTSET2292(IN6P_HOPLIMIT); 1756 break; 1757 case IPV6_2292HOPOPTS: 1758 /* 1759 * Check super-user privilege. 1760 * See comments for IPV6_RECVHOPOPTS. 1761 */ 1762 if (!privileged) 1763 return (EPERM); 1764 OPTSET2292(IN6P_HOPOPTS); 1765 break; 1766 case IPV6_2292DSTOPTS: 1767 if (!privileged) 1768 return (EPERM); 1769 OPTSET2292(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); /* XXX */ 1770 break; 1771 case IPV6_2292RTHDR: 1772 OPTSET2292(IN6P_RTHDR); 1773 break; 1774 } 1775 break; 1776 #endif 1777 case IPV6_PKTINFO: 1778 case IPV6_HOPOPTS: 1779 case IPV6_RTHDR: 1780 case IPV6_DSTOPTS: 1781 case IPV6_RTHDRDSTOPTS: 1782 case IPV6_NEXTHOP: 1783 { 1784 /* new advanced API (RFC3542) */ 1785 u_char *optbuf; 1786 int optbuflen; 1787 struct ip6_pktopts **optp; 1788 if (!m) { 1789 error = EINVAL; 1790 break; 1791 } 1792 1793 #ifdef RFC2292 1794 /* cannot mix with RFC2292 */ 1795 if (OPTBIT(IN6P_RFC2292)) { 1796 error = EINVAL; 1797 break; 1798 } 1799 #endif 1800 1801 if (m && m->m_next) { 1802 error = EINVAL; /* XXX */ 1803 break; 1804 } 1805 1806 optbuf = mtod(m, u_char *); 1807 optbuflen = m->m_len; 1808 optp = &in6p->in6p_outputopts; 1809 error = ip6_pcbopt(optname, optbuf, optbuflen, 1810 optp, privileged, uproto); 1811 break; 1812 } 1813 #undef OPTSET 1814 1815 case IPV6_MULTICAST_IF: 1816 case IPV6_MULTICAST_HOPS: 1817 case IPV6_MULTICAST_LOOP: 1818 case IPV6_JOIN_GROUP: 1819 case IPV6_LEAVE_GROUP: 1820 error = ip6_setmoptions(optname, 1821 &in6p->in6p_moptions, m); 1822 break; 1823 1824 case IPV6_PORTRANGE: 1825 if (!m) { 1826 error = EINVAL; 1827 break; 1828 } 1829 optval = *mtod(m, int *); 1830 1831 switch (optval) { 1832 case IPV6_PORTRANGE_DEFAULT: 1833 in6p->in6p_flags &= ~(IN6P_LOWPORT); 1834 in6p->in6p_flags &= ~(IN6P_HIGHPORT); 1835 break; 1836 1837 case IPV6_PORTRANGE_HIGH: 1838 in6p->in6p_flags &= ~(IN6P_LOWPORT); 1839 in6p->in6p_flags |= IN6P_HIGHPORT; 1840 break; 1841 1842 case IPV6_PORTRANGE_LOW: 1843 in6p->in6p_flags &= ~(IN6P_HIGHPORT); 1844 in6p->in6p_flags |= IN6P_LOWPORT; 1845 break; 1846 1847 default: 1848 error = EINVAL; 1849 break; 1850 } 1851 break; 1852 1853 1854 #if defined(IPSEC) || defined(FAST_IPSEC) 1855 case IPV6_IPSEC_POLICY: 1856 { 1857 void *req = NULL; 1858 size_t len = 0; 1859 if (m) { 1860 req = mtod(m, void *); 1861 len = m->m_len; 1862 } 1863 error = ipsec6_set_policy(in6p, optname, req, 1864 len, privileged); 1865 } 1866 break; 1867 #endif /* IPSEC */ 1868 1869 default: 1870 error = ENOPROTOOPT; 1871 break; 1872 } 1873 if (m) 1874 (void)m_free(m); 1875 break; 1876 1877 case PRCO_GETOPT: 1878 switch (optname) { 1879 #ifdef RFC2292 1880 case IPV6_2292PKTOPTIONS: 1881 /* 1882 * RFC3542 (effectively) deprecated the 1883 * semantics of the 2292-style pktoptions. 1884 * Since it was not reliable in nature (i.e., 1885 * applications had to expect the lack of some 1886 * information after all), it would make sense 1887 * to simplify this part by always returning 1888 * empty data. 1889 */ 1890 *mp = m_get(M_WAIT, MT_SOOPTS); 1891 (*mp)->m_len = 0; 1892 break; 1893 #endif 1894 1895 case IPV6_RECVHOPOPTS: 1896 case IPV6_RECVDSTOPTS: 1897 case IPV6_RECVRTHDRDSTOPTS: 1898 case IPV6_UNICAST_HOPS: 1899 case IPV6_RECVPKTINFO: 1900 case IPV6_RECVHOPLIMIT: 1901 case IPV6_RECVRTHDR: 1902 case IPV6_RECVPATHMTU: 1903 1904 case IPV6_FAITH: 1905 case IPV6_V6ONLY: 1906 case IPV6_PORTRANGE: 1907 case IPV6_RECVTCLASS: 1908 switch (optname) { 1909 1910 case IPV6_RECVHOPOPTS: 1911 optval = OPTBIT(IN6P_HOPOPTS); 1912 break; 1913 1914 case IPV6_RECVDSTOPTS: 1915 optval = OPTBIT(IN6P_DSTOPTS); 1916 break; 1917 1918 case IPV6_RECVRTHDRDSTOPTS: 1919 optval = OPTBIT(IN6P_RTHDRDSTOPTS); 1920 break; 1921 1922 case IPV6_UNICAST_HOPS: 1923 optval = in6p->in6p_hops; 1924 break; 1925 1926 case IPV6_RECVPKTINFO: 1927 optval = OPTBIT(IN6P_PKTINFO); 1928 break; 1929 1930 case IPV6_RECVHOPLIMIT: 1931 optval = OPTBIT(IN6P_HOPLIMIT); 1932 break; 1933 1934 case IPV6_RECVRTHDR: 1935 optval = OPTBIT(IN6P_RTHDR); 1936 break; 1937 1938 case IPV6_RECVPATHMTU: 1939 optval = OPTBIT(IN6P_MTU); 1940 break; 1941 1942 case IPV6_FAITH: 1943 optval = OPTBIT(IN6P_FAITH); 1944 break; 1945 1946 case IPV6_V6ONLY: 1947 optval = OPTBIT(IN6P_IPV6_V6ONLY); 1948 break; 1949 1950 case IPV6_PORTRANGE: 1951 { 1952 int flags; 1953 flags = in6p->in6p_flags; 1954 if (flags & IN6P_HIGHPORT) 1955 optval = IPV6_PORTRANGE_HIGH; 1956 else if (flags & IN6P_LOWPORT) 1957 optval = IPV6_PORTRANGE_LOW; 1958 else 1959 optval = 0; 1960 break; 1961 } 1962 case IPV6_RECVTCLASS: 1963 optval = OPTBIT(IN6P_TCLASS); 1964 break; 1965 1966 } 1967 if (error) 1968 break; 1969 *mp = m = m_get(M_WAIT, MT_SOOPTS); 1970 m->m_len = sizeof(int); 1971 *mtod(m, int *) = optval; 1972 break; 1973 1974 case IPV6_PATHMTU: 1975 { 1976 u_long pmtu = 0; 1977 struct ip6_mtuinfo mtuinfo; 1978 struct route *ro = &in6p->in6p_route; 1979 1980 if (!(so->so_state & SS_ISCONNECTED)) 1981 return (ENOTCONN); 1982 /* 1983 * XXX: we dot not consider the case of source 1984 * routing, or optional information to specify 1985 * the outgoing interface. 1986 */ 1987 error = ip6_getpmtu(ro, NULL, NULL, 1988 &in6p->in6p_faddr, &pmtu, NULL); 1989 if (error) 1990 break; 1991 if (pmtu > IPV6_MAXPACKET) 1992 pmtu = IPV6_MAXPACKET; 1993 1994 memset(&mtuinfo, 0, sizeof(mtuinfo)); 1995 mtuinfo.ip6m_mtu = (u_int32_t)pmtu; 1996 optdata = (void *)&mtuinfo; 1997 optdatalen = sizeof(mtuinfo); 1998 if (optdatalen > MCLBYTES) 1999 return (EMSGSIZE); /* XXX */ 2000 *mp = m = m_get(M_WAIT, MT_SOOPTS); 2001 if (optdatalen > MLEN) 2002 MCLGET(m, M_WAIT); 2003 m->m_len = optdatalen; 2004 memcpy(mtod(m, void *), optdata, optdatalen); 2005 break; 2006 } 2007 2008 #ifdef RFC2292 2009 case IPV6_2292PKTINFO: 2010 case IPV6_2292HOPLIMIT: 2011 case IPV6_2292HOPOPTS: 2012 case IPV6_2292RTHDR: 2013 case IPV6_2292DSTOPTS: 2014 switch (optname) { 2015 case IPV6_2292PKTINFO: 2016 optval = OPTBIT(IN6P_PKTINFO); 2017 break; 2018 case IPV6_2292HOPLIMIT: 2019 optval = OPTBIT(IN6P_HOPLIMIT); 2020 break; 2021 case IPV6_2292HOPOPTS: 2022 optval = OPTBIT(IN6P_HOPOPTS); 2023 break; 2024 case IPV6_2292RTHDR: 2025 optval = OPTBIT(IN6P_RTHDR); 2026 break; 2027 case IPV6_2292DSTOPTS: 2028 optval = OPTBIT(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); 2029 break; 2030 } 2031 *mp = m = m_get(M_WAIT, MT_SOOPTS); 2032 m->m_len = sizeof(int); 2033 *mtod(m, int *) = optval; 2034 break; 2035 #endif 2036 case IPV6_PKTINFO: 2037 case IPV6_HOPOPTS: 2038 case IPV6_RTHDR: 2039 case IPV6_DSTOPTS: 2040 case IPV6_RTHDRDSTOPTS: 2041 case IPV6_NEXTHOP: 2042 case IPV6_OTCLASS: 2043 case IPV6_TCLASS: 2044 case IPV6_DONTFRAG: 2045 case IPV6_USE_MIN_MTU: 2046 error = ip6_getpcbopt(in6p->in6p_outputopts, 2047 optname, mp); 2048 break; 2049 2050 case IPV6_MULTICAST_IF: 2051 case IPV6_MULTICAST_HOPS: 2052 case IPV6_MULTICAST_LOOP: 2053 case IPV6_JOIN_GROUP: 2054 case IPV6_LEAVE_GROUP: 2055 error = ip6_getmoptions(optname, 2056 in6p->in6p_moptions, mp); 2057 break; 2058 2059 #if defined(IPSEC) || defined(FAST_IPSEC) 2060 case IPV6_IPSEC_POLICY: 2061 { 2062 void *req = NULL; 2063 size_t len = 0; 2064 if (m) { 2065 req = mtod(m, void *); 2066 len = m->m_len; 2067 } 2068 error = ipsec6_get_policy(in6p, req, len, mp); 2069 break; 2070 } 2071 #endif /* IPSEC */ 2072 2073 2074 2075 2076 default: 2077 error = ENOPROTOOPT; 2078 break; 2079 } 2080 break; 2081 } 2082 return (error); 2083 } 2084 2085 int 2086 ip6_raw_ctloutput(int op, struct socket *so, int level, int optname, 2087 struct mbuf **mp) 2088 { 2089 int error = 0, optval, optlen; 2090 const int icmp6off = offsetof(struct icmp6_hdr, icmp6_cksum); 2091 struct in6pcb *in6p = sotoin6pcb(so); 2092 struct mbuf *m = *mp; 2093 2094 optlen = m ? m->m_len : 0; 2095 2096 if (level != IPPROTO_IPV6) { 2097 if (op == PRCO_SETOPT && *mp) 2098 (void)m_free(*mp); 2099 return ENOPROTOOPT; 2100 } 2101 2102 switch (optname) { 2103 case IPV6_CHECKSUM: 2104 /* 2105 * For ICMPv6 sockets, no modification allowed for checksum 2106 * offset, permit "no change" values to help existing apps. 2107 * 2108 * XXX RFC3542 says: "An attempt to set IPV6_CHECKSUM 2109 * for an ICMPv6 socket will fail." The current 2110 * behavior does not meet RFC3542. 2111 */ 2112 switch (op) { 2113 case PRCO_SETOPT: 2114 if (optlen != sizeof(int)) { 2115 error = EINVAL; 2116 break; 2117 } 2118 optval = *mtod(m, int *); 2119 if ((optval % 2) != 0) { 2120 /* the API assumes even offset values */ 2121 error = EINVAL; 2122 } else if (so->so_proto->pr_protocol == 2123 IPPROTO_ICMPV6) { 2124 if (optval != icmp6off) 2125 error = EINVAL; 2126 } else 2127 in6p->in6p_cksum = optval; 2128 break; 2129 2130 case PRCO_GETOPT: 2131 if (so->so_proto->pr_protocol == IPPROTO_ICMPV6) 2132 optval = icmp6off; 2133 else 2134 optval = in6p->in6p_cksum; 2135 2136 *mp = m = m_get(M_WAIT, MT_SOOPTS); 2137 m->m_len = sizeof(int); 2138 *mtod(m, int *) = optval; 2139 break; 2140 2141 default: 2142 error = EINVAL; 2143 break; 2144 } 2145 break; 2146 2147 default: 2148 error = ENOPROTOOPT; 2149 break; 2150 } 2151 2152 if (op == PRCO_SETOPT && m) 2153 (void)m_free(m); 2154 2155 return (error); 2156 } 2157 2158 #ifdef RFC2292 2159 /* 2160 * Set up IP6 options in pcb for insertion in output packets or 2161 * specifying behavior of outgoing packets. 2162 */ 2163 static int 2164 ip6_pcbopts(struct ip6_pktopts **pktopt, struct mbuf *m, struct socket *so) 2165 { 2166 struct ip6_pktopts *opt = *pktopt; 2167 int error = 0; 2168 struct lwp *l = curlwp; /* XXX */ 2169 int priv = 0; 2170 2171 /* turn off any old options. */ 2172 if (opt) { 2173 #ifdef DIAGNOSTIC 2174 if (opt->ip6po_pktinfo || opt->ip6po_nexthop || 2175 opt->ip6po_hbh || opt->ip6po_dest1 || opt->ip6po_dest2 || 2176 opt->ip6po_rhinfo.ip6po_rhi_rthdr) 2177 printf("ip6_pcbopts: all specified options are cleared.\n"); 2178 #endif 2179 ip6_clearpktopts(opt, -1); 2180 } else 2181 opt = malloc(sizeof(*opt), M_IP6OPT, M_WAITOK); 2182 *pktopt = NULL; 2183 2184 if (!m || m->m_len == 0) { 2185 /* 2186 * Only turning off any previous options, regardless of 2187 * whether the opt is just created or given. 2188 */ 2189 free(opt, M_IP6OPT); 2190 return (0); 2191 } 2192 2193 /* set options specified by user. */ 2194 if (l && !kauth_authorize_generic(l->l_cred, KAUTH_GENERIC_ISSUSER, 2195 NULL)) 2196 priv = 1; 2197 if ((error = ip6_setpktopts(m, opt, NULL, priv, 2198 so->so_proto->pr_protocol)) != 0) { 2199 ip6_clearpktopts(opt, -1); /* XXX: discard all options */ 2200 free(opt, M_IP6OPT); 2201 return (error); 2202 } 2203 *pktopt = opt; 2204 return (0); 2205 } 2206 #endif 2207 2208 /* 2209 * initialize ip6_pktopts. beware that there are non-zero default values in 2210 * the struct. 2211 */ 2212 void 2213 ip6_initpktopts(struct ip6_pktopts *opt) 2214 { 2215 2216 memset(opt, 0, sizeof(*opt)); 2217 opt->ip6po_hlim = -1; /* -1 means default hop limit */ 2218 opt->ip6po_tclass = -1; /* -1 means default traffic class */ 2219 opt->ip6po_minmtu = IP6PO_MINMTU_MCASTONLY; 2220 } 2221 2222 #define sin6tosa(sin6) ((struct sockaddr *)(sin6)) /* XXX */ 2223 static int 2224 ip6_pcbopt(int optname, u_char *buf, int len, struct ip6_pktopts **pktopt, 2225 int priv, int uproto) 2226 { 2227 struct ip6_pktopts *opt; 2228 2229 if (*pktopt == NULL) { 2230 *pktopt = malloc(sizeof(struct ip6_pktopts), M_IP6OPT, 2231 M_WAITOK); 2232 ip6_initpktopts(*pktopt); 2233 } 2234 opt = *pktopt; 2235 2236 return (ip6_setpktopt(optname, buf, len, opt, priv, 1, 0, uproto)); 2237 } 2238 2239 static int 2240 ip6_getpcbopt(struct ip6_pktopts *pktopt, int optname, struct mbuf **mp) 2241 { 2242 void *optdata = NULL; 2243 int optdatalen = 0; 2244 struct ip6_ext *ip6e; 2245 int error = 0; 2246 struct in6_pktinfo null_pktinfo; 2247 int deftclass = 0, on; 2248 int defminmtu = IP6PO_MINMTU_MCASTONLY; 2249 struct mbuf *m; 2250 2251 switch (optname) { 2252 case IPV6_PKTINFO: 2253 if (pktopt && pktopt->ip6po_pktinfo) 2254 optdata = (void *)pktopt->ip6po_pktinfo; 2255 else { 2256 /* XXX: we don't have to do this every time... */ 2257 memset(&null_pktinfo, 0, sizeof(null_pktinfo)); 2258 optdata = (void *)&null_pktinfo; 2259 } 2260 optdatalen = sizeof(struct in6_pktinfo); 2261 break; 2262 case IPV6_OTCLASS: 2263 /* XXX */ 2264 return (EINVAL); 2265 case IPV6_TCLASS: 2266 if (pktopt && pktopt->ip6po_tclass >= 0) 2267 optdata = (void *)&pktopt->ip6po_tclass; 2268 else 2269 optdata = (void *)&deftclass; 2270 optdatalen = sizeof(int); 2271 break; 2272 case IPV6_HOPOPTS: 2273 if (pktopt && pktopt->ip6po_hbh) { 2274 optdata = (void *)pktopt->ip6po_hbh; 2275 ip6e = (struct ip6_ext *)pktopt->ip6po_hbh; 2276 optdatalen = (ip6e->ip6e_len + 1) << 3; 2277 } 2278 break; 2279 case IPV6_RTHDR: 2280 if (pktopt && pktopt->ip6po_rthdr) { 2281 optdata = (void *)pktopt->ip6po_rthdr; 2282 ip6e = (struct ip6_ext *)pktopt->ip6po_rthdr; 2283 optdatalen = (ip6e->ip6e_len + 1) << 3; 2284 } 2285 break; 2286 case IPV6_RTHDRDSTOPTS: 2287 if (pktopt && pktopt->ip6po_dest1) { 2288 optdata = (void *)pktopt->ip6po_dest1; 2289 ip6e = (struct ip6_ext *)pktopt->ip6po_dest1; 2290 optdatalen = (ip6e->ip6e_len + 1) << 3; 2291 } 2292 break; 2293 case IPV6_DSTOPTS: 2294 if (pktopt && pktopt->ip6po_dest2) { 2295 optdata = (void *)pktopt->ip6po_dest2; 2296 ip6e = (struct ip6_ext *)pktopt->ip6po_dest2; 2297 optdatalen = (ip6e->ip6e_len + 1) << 3; 2298 } 2299 break; 2300 case IPV6_NEXTHOP: 2301 if (pktopt && pktopt->ip6po_nexthop) { 2302 optdata = (void *)pktopt->ip6po_nexthop; 2303 optdatalen = pktopt->ip6po_nexthop->sa_len; 2304 } 2305 break; 2306 case IPV6_USE_MIN_MTU: 2307 if (pktopt) 2308 optdata = (void *)&pktopt->ip6po_minmtu; 2309 else 2310 optdata = (void *)&defminmtu; 2311 optdatalen = sizeof(int); 2312 break; 2313 case IPV6_DONTFRAG: 2314 if (pktopt && ((pktopt->ip6po_flags) & IP6PO_DONTFRAG)) 2315 on = 1; 2316 else 2317 on = 0; 2318 optdata = (void *)&on; 2319 optdatalen = sizeof(on); 2320 break; 2321 default: /* should not happen */ 2322 #ifdef DIAGNOSTIC 2323 panic("ip6_getpcbopt: unexpected option\n"); 2324 #endif 2325 return (ENOPROTOOPT); 2326 } 2327 2328 if (optdatalen > MCLBYTES) 2329 return (EMSGSIZE); /* XXX */ 2330 *mp = m = m_get(M_WAIT, MT_SOOPTS); 2331 if (optdatalen > MLEN) 2332 MCLGET(m, M_WAIT); 2333 m->m_len = optdatalen; 2334 if (optdatalen) 2335 memcpy(mtod(m, void *), optdata, optdatalen); 2336 2337 return (error); 2338 } 2339 2340 void 2341 ip6_clearpktopts(struct ip6_pktopts *pktopt, int optname) 2342 { 2343 if (optname == -1 || optname == IPV6_PKTINFO) { 2344 if (pktopt->ip6po_pktinfo) 2345 free(pktopt->ip6po_pktinfo, M_IP6OPT); 2346 pktopt->ip6po_pktinfo = NULL; 2347 } 2348 if (optname == -1 || optname == IPV6_HOPLIMIT) 2349 pktopt->ip6po_hlim = -1; 2350 if (optname == -1 || optname == IPV6_TCLASS) 2351 pktopt->ip6po_tclass = -1; 2352 if (optname == -1 || optname == IPV6_NEXTHOP) { 2353 rtcache_free(&pktopt->ip6po_nextroute); 2354 if (pktopt->ip6po_nexthop) 2355 free(pktopt->ip6po_nexthop, M_IP6OPT); 2356 pktopt->ip6po_nexthop = NULL; 2357 } 2358 if (optname == -1 || optname == IPV6_HOPOPTS) { 2359 if (pktopt->ip6po_hbh) 2360 free(pktopt->ip6po_hbh, M_IP6OPT); 2361 pktopt->ip6po_hbh = NULL; 2362 } 2363 if (optname == -1 || optname == IPV6_RTHDRDSTOPTS) { 2364 if (pktopt->ip6po_dest1) 2365 free(pktopt->ip6po_dest1, M_IP6OPT); 2366 pktopt->ip6po_dest1 = NULL; 2367 } 2368 if (optname == -1 || optname == IPV6_RTHDR) { 2369 if (pktopt->ip6po_rhinfo.ip6po_rhi_rthdr) 2370 free(pktopt->ip6po_rhinfo.ip6po_rhi_rthdr, M_IP6OPT); 2371 pktopt->ip6po_rhinfo.ip6po_rhi_rthdr = NULL; 2372 rtcache_free(&pktopt->ip6po_route); 2373 } 2374 if (optname == -1 || optname == IPV6_DSTOPTS) { 2375 if (pktopt->ip6po_dest2) 2376 free(pktopt->ip6po_dest2, M_IP6OPT); 2377 pktopt->ip6po_dest2 = NULL; 2378 } 2379 } 2380 2381 #define PKTOPT_EXTHDRCPY(type) \ 2382 do { \ 2383 if (src->type) { \ 2384 int hlen = (((struct ip6_ext *)src->type)->ip6e_len + 1) << 3;\ 2385 dst->type = malloc(hlen, M_IP6OPT, canwait); \ 2386 if (dst->type == NULL && canwait == M_NOWAIT) \ 2387 goto bad; \ 2388 memcpy(dst->type, src->type, hlen); \ 2389 } \ 2390 } while (/*CONSTCOND*/ 0) 2391 2392 static int 2393 copypktopts(struct ip6_pktopts *dst, struct ip6_pktopts *src, int canwait) 2394 { 2395 dst->ip6po_hlim = src->ip6po_hlim; 2396 dst->ip6po_tclass = src->ip6po_tclass; 2397 dst->ip6po_flags = src->ip6po_flags; 2398 if (src->ip6po_pktinfo) { 2399 dst->ip6po_pktinfo = malloc(sizeof(*dst->ip6po_pktinfo), 2400 M_IP6OPT, canwait); 2401 if (dst->ip6po_pktinfo == NULL && canwait == M_NOWAIT) 2402 goto bad; 2403 *dst->ip6po_pktinfo = *src->ip6po_pktinfo; 2404 } 2405 if (src->ip6po_nexthop) { 2406 dst->ip6po_nexthop = malloc(src->ip6po_nexthop->sa_len, 2407 M_IP6OPT, canwait); 2408 if (dst->ip6po_nexthop == NULL && canwait == M_NOWAIT) 2409 goto bad; 2410 memcpy(dst->ip6po_nexthop, src->ip6po_nexthop, 2411 src->ip6po_nexthop->sa_len); 2412 } 2413 PKTOPT_EXTHDRCPY(ip6po_hbh); 2414 PKTOPT_EXTHDRCPY(ip6po_dest1); 2415 PKTOPT_EXTHDRCPY(ip6po_dest2); 2416 PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */ 2417 return (0); 2418 2419 bad: 2420 if (dst->ip6po_pktinfo) free(dst->ip6po_pktinfo, M_IP6OPT); 2421 if (dst->ip6po_nexthop) free(dst->ip6po_nexthop, M_IP6OPT); 2422 if (dst->ip6po_hbh) free(dst->ip6po_hbh, M_IP6OPT); 2423 if (dst->ip6po_dest1) free(dst->ip6po_dest1, M_IP6OPT); 2424 if (dst->ip6po_dest2) free(dst->ip6po_dest2, M_IP6OPT); 2425 if (dst->ip6po_rthdr) free(dst->ip6po_rthdr, M_IP6OPT); 2426 2427 return (ENOBUFS); 2428 } 2429 #undef PKTOPT_EXTHDRCPY 2430 2431 struct ip6_pktopts * 2432 ip6_copypktopts(struct ip6_pktopts *src, int canwait) 2433 { 2434 int error; 2435 struct ip6_pktopts *dst; 2436 2437 dst = malloc(sizeof(*dst), M_IP6OPT, canwait); 2438 if (dst == NULL && canwait == M_NOWAIT) 2439 return (NULL); 2440 ip6_initpktopts(dst); 2441 2442 if ((error = copypktopts(dst, src, canwait)) != 0) { 2443 free(dst, M_IP6OPT); 2444 return (NULL); 2445 } 2446 2447 return (dst); 2448 } 2449 2450 void 2451 ip6_freepcbopts(struct ip6_pktopts *pktopt) 2452 { 2453 if (pktopt == NULL) 2454 return; 2455 2456 ip6_clearpktopts(pktopt, -1); 2457 2458 free(pktopt, M_IP6OPT); 2459 } 2460 2461 /* 2462 * Set the IP6 multicast options in response to user setsockopt(). 2463 */ 2464 static int 2465 ip6_setmoptions(int optname, struct ip6_moptions **im6op, struct mbuf *m) 2466 { 2467 int error = 0; 2468 u_int loop, ifindex; 2469 struct ipv6_mreq *mreq; 2470 struct ifnet *ifp; 2471 struct ip6_moptions *im6o = *im6op; 2472 struct route ro; 2473 struct in6_multi_mship *imm; 2474 struct lwp *l = curlwp; /* XXX */ 2475 2476 if (im6o == NULL) { 2477 /* 2478 * No multicast option buffer attached to the pcb; 2479 * allocate one and initialize to default values. 2480 */ 2481 im6o = (struct ip6_moptions *) 2482 malloc(sizeof(*im6o), M_IPMOPTS, M_WAITOK); 2483 2484 if (im6o == NULL) 2485 return (ENOBUFS); 2486 *im6op = im6o; 2487 im6o->im6o_multicast_ifp = NULL; 2488 im6o->im6o_multicast_hlim = ip6_defmcasthlim; 2489 im6o->im6o_multicast_loop = IPV6_DEFAULT_MULTICAST_LOOP; 2490 LIST_INIT(&im6o->im6o_memberships); 2491 } 2492 2493 switch (optname) { 2494 2495 case IPV6_MULTICAST_IF: 2496 /* 2497 * Select the interface for outgoing multicast packets. 2498 */ 2499 if (m == NULL || m->m_len != sizeof(u_int)) { 2500 error = EINVAL; 2501 break; 2502 } 2503 bcopy(mtod(m, u_int *), &ifindex, sizeof(ifindex)); 2504 if (ifindex != 0) { 2505 if (if_indexlim <= ifindex || !ifindex2ifnet[ifindex]) { 2506 error = ENXIO; /* XXX EINVAL? */ 2507 break; 2508 } 2509 ifp = ifindex2ifnet[ifindex]; 2510 if ((ifp->if_flags & IFF_MULTICAST) == 0) { 2511 error = EADDRNOTAVAIL; 2512 break; 2513 } 2514 } else 2515 ifp = NULL; 2516 im6o->im6o_multicast_ifp = ifp; 2517 break; 2518 2519 case IPV6_MULTICAST_HOPS: 2520 { 2521 /* 2522 * Set the IP6 hoplimit for outgoing multicast packets. 2523 */ 2524 int optval; 2525 if (m == NULL || m->m_len != sizeof(int)) { 2526 error = EINVAL; 2527 break; 2528 } 2529 bcopy(mtod(m, u_int *), &optval, sizeof(optval)); 2530 if (optval < -1 || optval >= 256) 2531 error = EINVAL; 2532 else if (optval == -1) 2533 im6o->im6o_multicast_hlim = ip6_defmcasthlim; 2534 else 2535 im6o->im6o_multicast_hlim = optval; 2536 break; 2537 } 2538 2539 case IPV6_MULTICAST_LOOP: 2540 /* 2541 * Set the loopback flag for outgoing multicast packets. 2542 * Must be zero or one. 2543 */ 2544 if (m == NULL || m->m_len != sizeof(u_int)) { 2545 error = EINVAL; 2546 break; 2547 } 2548 bcopy(mtod(m, u_int *), &loop, sizeof(loop)); 2549 if (loop > 1) { 2550 error = EINVAL; 2551 break; 2552 } 2553 im6o->im6o_multicast_loop = loop; 2554 break; 2555 2556 case IPV6_JOIN_GROUP: 2557 /* 2558 * Add a multicast group membership. 2559 * Group must be a valid IP6 multicast address. 2560 */ 2561 if (m == NULL || m->m_len != sizeof(struct ipv6_mreq)) { 2562 error = EINVAL; 2563 break; 2564 } 2565 mreq = mtod(m, struct ipv6_mreq *); 2566 if (IN6_IS_ADDR_UNSPECIFIED(&mreq->ipv6mr_multiaddr)) { 2567 /* 2568 * We use the unspecified address to specify to accept 2569 * all multicast addresses. Only super user is allowed 2570 * to do this. 2571 */ 2572 if (kauth_authorize_generic(l->l_cred, 2573 KAUTH_GENERIC_ISSUSER, NULL)) 2574 { 2575 error = EACCES; 2576 break; 2577 } 2578 } else if (!IN6_IS_ADDR_MULTICAST(&mreq->ipv6mr_multiaddr)) { 2579 error = EINVAL; 2580 break; 2581 } 2582 2583 /* 2584 * If no interface was explicitly specified, choose an 2585 * appropriate one according to the given multicast address. 2586 */ 2587 if (mreq->ipv6mr_interface == 0) { 2588 struct rtentry *rt; 2589 union { 2590 struct sockaddr dst; 2591 struct sockaddr_in6 dst6; 2592 } u; 2593 2594 /* 2595 * Look up the routing table for the 2596 * address, and choose the outgoing interface. 2597 * XXX: is it a good approach? 2598 */ 2599 memset(&ro, 0, sizeof(ro)); 2600 sockaddr_in6_init(&u.dst6, &mreq->ipv6mr_multiaddr, 0, 2601 0, 0); 2602 rtcache_setdst(&ro, &u.dst); 2603 ifp = (rt = rtcache_init(&ro)) != NULL ? rt->rt_ifp 2604 : NULL; 2605 rtcache_free(&ro); 2606 } else { 2607 /* 2608 * If the interface is specified, validate it. 2609 */ 2610 if (if_indexlim <= mreq->ipv6mr_interface || 2611 !ifindex2ifnet[mreq->ipv6mr_interface]) { 2612 error = ENXIO; /* XXX EINVAL? */ 2613 break; 2614 } 2615 ifp = ifindex2ifnet[mreq->ipv6mr_interface]; 2616 } 2617 2618 /* 2619 * See if we found an interface, and confirm that it 2620 * supports multicast 2621 */ 2622 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) { 2623 error = EADDRNOTAVAIL; 2624 break; 2625 } 2626 2627 if (in6_setscope(&mreq->ipv6mr_multiaddr, ifp, NULL)) { 2628 error = EADDRNOTAVAIL; /* XXX: should not happen */ 2629 break; 2630 } 2631 2632 /* 2633 * See if the membership already exists. 2634 */ 2635 for (imm = im6o->im6o_memberships.lh_first; 2636 imm != NULL; imm = imm->i6mm_chain.le_next) 2637 if (imm->i6mm_maddr->in6m_ifp == ifp && 2638 IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr, 2639 &mreq->ipv6mr_multiaddr)) 2640 break; 2641 if (imm != NULL) { 2642 error = EADDRINUSE; 2643 break; 2644 } 2645 /* 2646 * Everything looks good; add a new record to the multicast 2647 * address list for the given interface. 2648 */ 2649 imm = in6_joingroup(ifp, &mreq->ipv6mr_multiaddr, &error, 0); 2650 if (imm == NULL) 2651 break; 2652 LIST_INSERT_HEAD(&im6o->im6o_memberships, imm, i6mm_chain); 2653 break; 2654 2655 case IPV6_LEAVE_GROUP: 2656 /* 2657 * Drop a multicast group membership. 2658 * Group must be a valid IP6 multicast address. 2659 */ 2660 if (m == NULL || m->m_len != sizeof(struct ipv6_mreq)) { 2661 error = EINVAL; 2662 break; 2663 } 2664 mreq = mtod(m, struct ipv6_mreq *); 2665 2666 /* 2667 * If an interface address was specified, get a pointer 2668 * to its ifnet structure. 2669 */ 2670 if (mreq->ipv6mr_interface != 0) { 2671 if (if_indexlim <= mreq->ipv6mr_interface || 2672 !ifindex2ifnet[mreq->ipv6mr_interface]) { 2673 error = ENXIO; /* XXX EINVAL? */ 2674 break; 2675 } 2676 ifp = ifindex2ifnet[mreq->ipv6mr_interface]; 2677 } else 2678 ifp = NULL; 2679 2680 /* Fill in the scope zone ID */ 2681 if (ifp) { 2682 if (in6_setscope(&mreq->ipv6mr_multiaddr, ifp, NULL)) { 2683 /* XXX: should not happen */ 2684 error = EADDRNOTAVAIL; 2685 break; 2686 } 2687 } else if (mreq->ipv6mr_interface != 0) { 2688 /* 2689 * XXX: This case would happens when the (positive) 2690 * index is in the valid range, but the corresponding 2691 * interface has been detached dynamically. The above 2692 * check probably avoids such case to happen here, but 2693 * we check it explicitly for safety. 2694 */ 2695 error = EADDRNOTAVAIL; 2696 break; 2697 } else { /* ipv6mr_interface == 0 */ 2698 struct sockaddr_in6 sa6_mc; 2699 2700 /* 2701 * The API spec says as follows: 2702 * If the interface index is specified as 0, the 2703 * system may choose a multicast group membership to 2704 * drop by matching the multicast address only. 2705 * On the other hand, we cannot disambiguate the scope 2706 * zone unless an interface is provided. Thus, we 2707 * check if there's ambiguity with the default scope 2708 * zone as the last resort. 2709 */ 2710 sockaddr_in6_init(&sa6_mc, &mreq->ipv6mr_multiaddr, 2711 0, 0, 0); 2712 error = sa6_embedscope(&sa6_mc, ip6_use_defzone); 2713 if (error != 0) 2714 break; 2715 mreq->ipv6mr_multiaddr = sa6_mc.sin6_addr; 2716 } 2717 2718 /* 2719 * Find the membership in the membership list. 2720 */ 2721 for (imm = im6o->im6o_memberships.lh_first; 2722 imm != NULL; imm = imm->i6mm_chain.le_next) { 2723 if ((ifp == NULL || imm->i6mm_maddr->in6m_ifp == ifp) && 2724 IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr, 2725 &mreq->ipv6mr_multiaddr)) 2726 break; 2727 } 2728 if (imm == NULL) { 2729 /* Unable to resolve interface */ 2730 error = EADDRNOTAVAIL; 2731 break; 2732 } 2733 /* 2734 * Give up the multicast address record to which the 2735 * membership points. 2736 */ 2737 LIST_REMOVE(imm, i6mm_chain); 2738 in6_leavegroup(imm); 2739 break; 2740 2741 default: 2742 error = EOPNOTSUPP; 2743 break; 2744 } 2745 2746 /* 2747 * If all options have default values, no need to keep the mbuf. 2748 */ 2749 if (im6o->im6o_multicast_ifp == NULL && 2750 im6o->im6o_multicast_hlim == ip6_defmcasthlim && 2751 im6o->im6o_multicast_loop == IPV6_DEFAULT_MULTICAST_LOOP && 2752 im6o->im6o_memberships.lh_first == NULL) { 2753 free(*im6op, M_IPMOPTS); 2754 *im6op = NULL; 2755 } 2756 2757 return (error); 2758 } 2759 2760 /* 2761 * Return the IP6 multicast options in response to user getsockopt(). 2762 */ 2763 static int 2764 ip6_getmoptions(int optname, struct ip6_moptions *im6o, struct mbuf **mp) 2765 { 2766 u_int *hlim, *loop, *ifindex; 2767 2768 *mp = m_get(M_WAIT, MT_SOOPTS); 2769 2770 switch (optname) { 2771 2772 case IPV6_MULTICAST_IF: 2773 ifindex = mtod(*mp, u_int *); 2774 (*mp)->m_len = sizeof(u_int); 2775 if (im6o == NULL || im6o->im6o_multicast_ifp == NULL) 2776 *ifindex = 0; 2777 else 2778 *ifindex = im6o->im6o_multicast_ifp->if_index; 2779 return (0); 2780 2781 case IPV6_MULTICAST_HOPS: 2782 hlim = mtod(*mp, u_int *); 2783 (*mp)->m_len = sizeof(u_int); 2784 if (im6o == NULL) 2785 *hlim = ip6_defmcasthlim; 2786 else 2787 *hlim = im6o->im6o_multicast_hlim; 2788 return (0); 2789 2790 case IPV6_MULTICAST_LOOP: 2791 loop = mtod(*mp, u_int *); 2792 (*mp)->m_len = sizeof(u_int); 2793 if (im6o == NULL) 2794 *loop = ip6_defmcasthlim; 2795 else 2796 *loop = im6o->im6o_multicast_loop; 2797 return (0); 2798 2799 default: 2800 return (EOPNOTSUPP); 2801 } 2802 } 2803 2804 /* 2805 * Discard the IP6 multicast options. 2806 */ 2807 void 2808 ip6_freemoptions(struct ip6_moptions *im6o) 2809 { 2810 struct in6_multi_mship *imm; 2811 2812 if (im6o == NULL) 2813 return; 2814 2815 while ((imm = im6o->im6o_memberships.lh_first) != NULL) { 2816 LIST_REMOVE(imm, i6mm_chain); 2817 in6_leavegroup(imm); 2818 } 2819 free(im6o, M_IPMOPTS); 2820 } 2821 2822 /* 2823 * Set IPv6 outgoing packet options based on advanced API. 2824 */ 2825 int 2826 ip6_setpktopts(struct mbuf *control, struct ip6_pktopts *opt, 2827 struct ip6_pktopts *stickyopt, int priv, int uproto) 2828 { 2829 struct cmsghdr *cm = 0; 2830 2831 if (control == NULL || opt == NULL) 2832 return (EINVAL); 2833 2834 ip6_initpktopts(opt); 2835 if (stickyopt) { 2836 int error; 2837 2838 /* 2839 * If stickyopt is provided, make a local copy of the options 2840 * for this particular packet, then override them by ancillary 2841 * objects. 2842 * XXX: copypktopts() does not copy the cached route to a next 2843 * hop (if any). This is not very good in terms of efficiency, 2844 * but we can allow this since this option should be rarely 2845 * used. 2846 */ 2847 if ((error = copypktopts(opt, stickyopt, M_NOWAIT)) != 0) 2848 return (error); 2849 } 2850 2851 /* 2852 * XXX: Currently, we assume all the optional information is stored 2853 * in a single mbuf. 2854 */ 2855 if (control->m_next) 2856 return (EINVAL); 2857 2858 for (; control->m_len; control->m_data += CMSG_ALIGN(cm->cmsg_len), 2859 control->m_len -= CMSG_ALIGN(cm->cmsg_len)) { 2860 int error; 2861 2862 if (control->m_len < CMSG_LEN(0)) 2863 return (EINVAL); 2864 2865 cm = mtod(control, struct cmsghdr *); 2866 if (cm->cmsg_len == 0 || cm->cmsg_len > control->m_len) 2867 return (EINVAL); 2868 if (cm->cmsg_level != IPPROTO_IPV6) 2869 continue; 2870 2871 error = ip6_setpktopt(cm->cmsg_type, CMSG_DATA(cm), 2872 cm->cmsg_len - CMSG_LEN(0), opt, priv, 0, 1, uproto); 2873 if (error) 2874 return (error); 2875 } 2876 2877 return (0); 2878 } 2879 2880 /* 2881 * Set a particular packet option, as a sticky option or an ancillary data 2882 * item. "len" can be 0 only when it's a sticky option. 2883 * We have 4 cases of combination of "sticky" and "cmsg": 2884 * "sticky=0, cmsg=0": impossible 2885 * "sticky=0, cmsg=1": RFC2292 or RFC3542 ancillary data 2886 * "sticky=1, cmsg=0": RFC3542 socket option 2887 * "sticky=1, cmsg=1": RFC2292 socket option 2888 */ 2889 static int 2890 ip6_setpktopt(int optname, u_char *buf, int len, struct ip6_pktopts *opt, 2891 int priv, int sticky, int cmsg, int uproto) 2892 { 2893 int minmtupolicy; 2894 2895 if (!sticky && !cmsg) { 2896 #ifdef DIAGNOSTIC 2897 printf("ip6_setpktopt: impossible case\n"); 2898 #endif 2899 return (EINVAL); 2900 } 2901 2902 /* 2903 * IPV6_2292xxx is for backward compatibility to RFC2292, and should 2904 * not be specified in the context of RFC3542. Conversely, 2905 * RFC3542 types should not be specified in the context of RFC2292. 2906 */ 2907 if (!cmsg) { 2908 switch (optname) { 2909 case IPV6_2292PKTINFO: 2910 case IPV6_2292HOPLIMIT: 2911 case IPV6_2292NEXTHOP: 2912 case IPV6_2292HOPOPTS: 2913 case IPV6_2292DSTOPTS: 2914 case IPV6_2292RTHDR: 2915 case IPV6_2292PKTOPTIONS: 2916 return (ENOPROTOOPT); 2917 } 2918 } 2919 if (sticky && cmsg) { 2920 switch (optname) { 2921 case IPV6_PKTINFO: 2922 case IPV6_HOPLIMIT: 2923 case IPV6_NEXTHOP: 2924 case IPV6_HOPOPTS: 2925 case IPV6_DSTOPTS: 2926 case IPV6_RTHDRDSTOPTS: 2927 case IPV6_RTHDR: 2928 case IPV6_USE_MIN_MTU: 2929 case IPV6_DONTFRAG: 2930 case IPV6_OTCLASS: 2931 case IPV6_TCLASS: 2932 return (ENOPROTOOPT); 2933 } 2934 } 2935 2936 switch (optname) { 2937 #ifdef RFC2292 2938 case IPV6_2292PKTINFO: 2939 #endif 2940 case IPV6_PKTINFO: 2941 { 2942 struct ifnet *ifp = NULL; 2943 struct in6_pktinfo *pktinfo; 2944 2945 if (len != sizeof(struct in6_pktinfo)) 2946 return (EINVAL); 2947 2948 pktinfo = (struct in6_pktinfo *)buf; 2949 2950 /* 2951 * An application can clear any sticky IPV6_PKTINFO option by 2952 * doing a "regular" setsockopt with ipi6_addr being 2953 * in6addr_any and ipi6_ifindex being zero. 2954 * [RFC 3542, Section 6] 2955 */ 2956 if (optname == IPV6_PKTINFO && opt->ip6po_pktinfo && 2957 pktinfo->ipi6_ifindex == 0 && 2958 IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) { 2959 ip6_clearpktopts(opt, optname); 2960 break; 2961 } 2962 2963 if (uproto == IPPROTO_TCP && optname == IPV6_PKTINFO && 2964 sticky && !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) { 2965 return (EINVAL); 2966 } 2967 2968 /* validate the interface index if specified. */ 2969 if (pktinfo->ipi6_ifindex >= if_indexlim) { 2970 return (ENXIO); 2971 } 2972 if (pktinfo->ipi6_ifindex) { 2973 ifp = ifindex2ifnet[pktinfo->ipi6_ifindex]; 2974 if (ifp == NULL) 2975 return (ENXIO); 2976 } 2977 2978 /* 2979 * We store the address anyway, and let in6_selectsrc() 2980 * validate the specified address. This is because ipi6_addr 2981 * may not have enough information about its scope zone, and 2982 * we may need additional information (such as outgoing 2983 * interface or the scope zone of a destination address) to 2984 * disambiguate the scope. 2985 * XXX: the delay of the validation may confuse the 2986 * application when it is used as a sticky option. 2987 */ 2988 if (opt->ip6po_pktinfo == NULL) { 2989 opt->ip6po_pktinfo = malloc(sizeof(*pktinfo), 2990 M_IP6OPT, M_NOWAIT); 2991 if (opt->ip6po_pktinfo == NULL) 2992 return (ENOBUFS); 2993 } 2994 memcpy(opt->ip6po_pktinfo, pktinfo, sizeof(*pktinfo)); 2995 break; 2996 } 2997 2998 #ifdef RFC2292 2999 case IPV6_2292HOPLIMIT: 3000 #endif 3001 case IPV6_HOPLIMIT: 3002 { 3003 int *hlimp; 3004 3005 /* 3006 * RFC 3542 deprecated the usage of sticky IPV6_HOPLIMIT 3007 * to simplify the ordering among hoplimit options. 3008 */ 3009 if (optname == IPV6_HOPLIMIT && sticky) 3010 return (ENOPROTOOPT); 3011 3012 if (len != sizeof(int)) 3013 return (EINVAL); 3014 hlimp = (int *)buf; 3015 if (*hlimp < -1 || *hlimp > 255) 3016 return (EINVAL); 3017 3018 opt->ip6po_hlim = *hlimp; 3019 break; 3020 } 3021 3022 case IPV6_OTCLASS: 3023 if (len != sizeof(u_int8_t)) 3024 return (EINVAL); 3025 3026 opt->ip6po_tclass = *(u_int8_t *)buf; 3027 break; 3028 3029 case IPV6_TCLASS: 3030 { 3031 int tclass; 3032 3033 if (len != sizeof(int)) 3034 return (EINVAL); 3035 tclass = *(int *)buf; 3036 if (tclass < -1 || tclass > 255) 3037 return (EINVAL); 3038 3039 opt->ip6po_tclass = tclass; 3040 break; 3041 } 3042 3043 #ifdef RFC2292 3044 case IPV6_2292NEXTHOP: 3045 #endif 3046 case IPV6_NEXTHOP: 3047 if (!priv) 3048 return (EPERM); 3049 3050 if (len == 0) { /* just remove the option */ 3051 ip6_clearpktopts(opt, IPV6_NEXTHOP); 3052 break; 3053 } 3054 3055 /* check if cmsg_len is large enough for sa_len */ 3056 if (len < sizeof(struct sockaddr) || len < *buf) 3057 return (EINVAL); 3058 3059 switch (((struct sockaddr *)buf)->sa_family) { 3060 case AF_INET6: 3061 { 3062 struct sockaddr_in6 *sa6 = (struct sockaddr_in6 *)buf; 3063 int error; 3064 3065 if (sa6->sin6_len != sizeof(struct sockaddr_in6)) 3066 return (EINVAL); 3067 3068 if (IN6_IS_ADDR_UNSPECIFIED(&sa6->sin6_addr) || 3069 IN6_IS_ADDR_MULTICAST(&sa6->sin6_addr)) { 3070 return (EINVAL); 3071 } 3072 if ((error = sa6_embedscope(sa6, ip6_use_defzone)) 3073 != 0) { 3074 return (error); 3075 } 3076 break; 3077 } 3078 case AF_LINK: /* eventually be supported? */ 3079 default: 3080 return (EAFNOSUPPORT); 3081 } 3082 3083 /* turn off the previous option, then set the new option. */ 3084 ip6_clearpktopts(opt, IPV6_NEXTHOP); 3085 opt->ip6po_nexthop = malloc(*buf, M_IP6OPT, M_NOWAIT); 3086 if (opt->ip6po_nexthop == NULL) 3087 return (ENOBUFS); 3088 memcpy(opt->ip6po_nexthop, buf, *buf); 3089 break; 3090 3091 #ifdef RFC2292 3092 case IPV6_2292HOPOPTS: 3093 #endif 3094 case IPV6_HOPOPTS: 3095 { 3096 struct ip6_hbh *hbh; 3097 int hbhlen; 3098 3099 /* 3100 * XXX: We don't allow a non-privileged user to set ANY HbH 3101 * options, since per-option restriction has too much 3102 * overhead. 3103 */ 3104 if (!priv) 3105 return (EPERM); 3106 3107 if (len == 0) { 3108 ip6_clearpktopts(opt, IPV6_HOPOPTS); 3109 break; /* just remove the option */ 3110 } 3111 3112 /* message length validation */ 3113 if (len < sizeof(struct ip6_hbh)) 3114 return (EINVAL); 3115 hbh = (struct ip6_hbh *)buf; 3116 hbhlen = (hbh->ip6h_len + 1) << 3; 3117 if (len != hbhlen) 3118 return (EINVAL); 3119 3120 /* turn off the previous option, then set the new option. */ 3121 ip6_clearpktopts(opt, IPV6_HOPOPTS); 3122 opt->ip6po_hbh = malloc(hbhlen, M_IP6OPT, M_NOWAIT); 3123 if (opt->ip6po_hbh == NULL) 3124 return (ENOBUFS); 3125 memcpy(opt->ip6po_hbh, hbh, hbhlen); 3126 3127 break; 3128 } 3129 3130 #ifdef RFC2292 3131 case IPV6_2292DSTOPTS: 3132 #endif 3133 case IPV6_DSTOPTS: 3134 case IPV6_RTHDRDSTOPTS: 3135 { 3136 struct ip6_dest *dest, **newdest = NULL; 3137 int destlen; 3138 3139 if (!priv) /* XXX: see the comment for IPV6_HOPOPTS */ 3140 return (EPERM); 3141 3142 if (len == 0) { 3143 ip6_clearpktopts(opt, optname); 3144 break; /* just remove the option */ 3145 } 3146 3147 /* message length validation */ 3148 if (len < sizeof(struct ip6_dest)) 3149 return (EINVAL); 3150 dest = (struct ip6_dest *)buf; 3151 destlen = (dest->ip6d_len + 1) << 3; 3152 if (len != destlen) 3153 return (EINVAL); 3154 /* 3155 * Determine the position that the destination options header 3156 * should be inserted; before or after the routing header. 3157 */ 3158 switch (optname) { 3159 case IPV6_2292DSTOPTS: 3160 /* 3161 * The old advanced API is ambiguous on this point. 3162 * Our approach is to determine the position based 3163 * according to the existence of a routing header. 3164 * Note, however, that this depends on the order of the 3165 * extension headers in the ancillary data; the 1st 3166 * part of the destination options header must appear 3167 * before the routing header in the ancillary data, 3168 * too. 3169 * RFC3542 solved the ambiguity by introducing 3170 * separate ancillary data or option types. 3171 */ 3172 if (opt->ip6po_rthdr == NULL) 3173 newdest = &opt->ip6po_dest1; 3174 else 3175 newdest = &opt->ip6po_dest2; 3176 break; 3177 case IPV6_RTHDRDSTOPTS: 3178 newdest = &opt->ip6po_dest1; 3179 break; 3180 case IPV6_DSTOPTS: 3181 newdest = &opt->ip6po_dest2; 3182 break; 3183 } 3184 3185 /* turn off the previous option, then set the new option. */ 3186 ip6_clearpktopts(opt, optname); 3187 *newdest = malloc(destlen, M_IP6OPT, M_NOWAIT); 3188 if (*newdest == NULL) 3189 return (ENOBUFS); 3190 memcpy(*newdest, dest, destlen); 3191 3192 break; 3193 } 3194 3195 #ifdef RFC2292 3196 case IPV6_2292RTHDR: 3197 #endif 3198 case IPV6_RTHDR: 3199 { 3200 struct ip6_rthdr *rth; 3201 int rthlen; 3202 3203 if (len == 0) { 3204 ip6_clearpktopts(opt, IPV6_RTHDR); 3205 break; /* just remove the option */ 3206 } 3207 3208 /* message length validation */ 3209 if (len < sizeof(struct ip6_rthdr)) 3210 return (EINVAL); 3211 rth = (struct ip6_rthdr *)buf; 3212 rthlen = (rth->ip6r_len + 1) << 3; 3213 if (len != rthlen) 3214 return (EINVAL); 3215 switch (rth->ip6r_type) { 3216 case IPV6_RTHDR_TYPE_0: 3217 if (rth->ip6r_len == 0) /* must contain one addr */ 3218 return (EINVAL); 3219 if (rth->ip6r_len % 2) /* length must be even */ 3220 return (EINVAL); 3221 if (rth->ip6r_len / 2 != rth->ip6r_segleft) 3222 return (EINVAL); 3223 break; 3224 default: 3225 return (EINVAL); /* not supported */ 3226 } 3227 /* turn off the previous option */ 3228 ip6_clearpktopts(opt, IPV6_RTHDR); 3229 opt->ip6po_rthdr = malloc(rthlen, M_IP6OPT, M_NOWAIT); 3230 if (opt->ip6po_rthdr == NULL) 3231 return (ENOBUFS); 3232 memcpy(opt->ip6po_rthdr, rth, rthlen); 3233 break; 3234 } 3235 3236 case IPV6_USE_MIN_MTU: 3237 if (len != sizeof(int)) 3238 return (EINVAL); 3239 minmtupolicy = *(int *)buf; 3240 if (minmtupolicy != IP6PO_MINMTU_MCASTONLY && 3241 minmtupolicy != IP6PO_MINMTU_DISABLE && 3242 minmtupolicy != IP6PO_MINMTU_ALL) { 3243 return (EINVAL); 3244 } 3245 opt->ip6po_minmtu = minmtupolicy; 3246 break; 3247 3248 case IPV6_DONTFRAG: 3249 if (len != sizeof(int)) 3250 return (EINVAL); 3251 3252 if (uproto == IPPROTO_TCP || *(int *)buf == 0) { 3253 /* 3254 * we ignore this option for TCP sockets. 3255 * (RFC3542 leaves this case unspecified.) 3256 */ 3257 opt->ip6po_flags &= ~IP6PO_DONTFRAG; 3258 } else 3259 opt->ip6po_flags |= IP6PO_DONTFRAG; 3260 break; 3261 3262 default: 3263 return (ENOPROTOOPT); 3264 } /* end of switch */ 3265 3266 return (0); 3267 } 3268 3269 /* 3270 * Routine called from ip6_output() to loop back a copy of an IP6 multicast 3271 * packet to the input queue of a specified interface. Note that this 3272 * calls the output routine of the loopback "driver", but with an interface 3273 * pointer that might NOT be lo0ifp -- easier than replicating that code here. 3274 */ 3275 void 3276 ip6_mloopback(struct ifnet *ifp, struct mbuf *m, 3277 const struct sockaddr_in6 *dst) 3278 { 3279 struct mbuf *copym; 3280 struct ip6_hdr *ip6; 3281 3282 copym = m_copy(m, 0, M_COPYALL); 3283 if (copym == NULL) 3284 return; 3285 3286 /* 3287 * Make sure to deep-copy IPv6 header portion in case the data 3288 * is in an mbuf cluster, so that we can safely override the IPv6 3289 * header portion later. 3290 */ 3291 if ((copym->m_flags & M_EXT) != 0 || 3292 copym->m_len < sizeof(struct ip6_hdr)) { 3293 copym = m_pullup(copym, sizeof(struct ip6_hdr)); 3294 if (copym == NULL) 3295 return; 3296 } 3297 3298 #ifdef DIAGNOSTIC 3299 if (copym->m_len < sizeof(*ip6)) { 3300 m_freem(copym); 3301 return; 3302 } 3303 #endif 3304 3305 ip6 = mtod(copym, struct ip6_hdr *); 3306 /* 3307 * clear embedded scope identifiers if necessary. 3308 * in6_clearscope will touch the addresses only when necessary. 3309 */ 3310 in6_clearscope(&ip6->ip6_src); 3311 in6_clearscope(&ip6->ip6_dst); 3312 3313 (void)looutput(ifp, copym, (const struct sockaddr *)dst, NULL); 3314 } 3315 3316 /* 3317 * Chop IPv6 header off from the payload. 3318 */ 3319 static int 3320 ip6_splithdr(struct mbuf *m, struct ip6_exthdrs *exthdrs) 3321 { 3322 struct mbuf *mh; 3323 struct ip6_hdr *ip6; 3324 3325 ip6 = mtod(m, struct ip6_hdr *); 3326 if (m->m_len > sizeof(*ip6)) { 3327 MGETHDR(mh, M_DONTWAIT, MT_HEADER); 3328 if (mh == 0) { 3329 m_freem(m); 3330 return ENOBUFS; 3331 } 3332 M_MOVE_PKTHDR(mh, m); 3333 MH_ALIGN(mh, sizeof(*ip6)); 3334 m->m_len -= sizeof(*ip6); 3335 m->m_data += sizeof(*ip6); 3336 mh->m_next = m; 3337 m = mh; 3338 m->m_len = sizeof(*ip6); 3339 bcopy((void *)ip6, mtod(m, void *), sizeof(*ip6)); 3340 } 3341 exthdrs->ip6e_ip6 = m; 3342 return 0; 3343 } 3344 3345 /* 3346 * Compute IPv6 extension header length. 3347 */ 3348 int 3349 ip6_optlen(struct in6pcb *in6p) 3350 { 3351 int len; 3352 3353 if (!in6p->in6p_outputopts) 3354 return 0; 3355 3356 len = 0; 3357 #define elen(x) \ 3358 (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0) 3359 3360 len += elen(in6p->in6p_outputopts->ip6po_hbh); 3361 len += elen(in6p->in6p_outputopts->ip6po_dest1); 3362 len += elen(in6p->in6p_outputopts->ip6po_rthdr); 3363 len += elen(in6p->in6p_outputopts->ip6po_dest2); 3364 return len; 3365 #undef elen 3366 } 3367
Indexes created Sun Sep 21 20:09:37 GMT 2025