ip6_output.c revision 1.97
1 /* $NetBSD: ip6_output.c,v 1.97 2006/05/05 00:03:22 rpaulo 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 1.97 2006/05/05 00:03:22 rpaulo 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 82 #include <net/if.h> 83 #include <net/route.h> 84 #ifdef PFIL_HOOKS 85 #include <net/pfil.h> 86 #endif 87 88 #include <netinet/in.h> 89 #include <netinet/in_var.h> 90 #include <netinet/ip6.h> 91 #include <netinet/icmp6.h> 92 #include <netinet/in_offload.h> 93 #include <netinet6/ip6_var.h> 94 #include <netinet6/in6_pcb.h> 95 #include <netinet6/nd6.h> 96 #include <netinet6/ip6protosw.h> 97 #include <netinet6/scope6_var.h> 98 99 #ifdef IPSEC 100 #include <netinet6/ipsec.h> 101 #include <netkey/key.h> 102 #endif /* IPSEC */ 103 104 #include <net/net_osdep.h> 105 106 #ifdef PFIL_HOOKS 107 extern struct pfil_head inet6_pfil_hook; /* XXX */ 108 #endif 109 110 struct ip6_exthdrs { 111 struct mbuf *ip6e_ip6; 112 struct mbuf *ip6e_hbh; 113 struct mbuf *ip6e_dest1; 114 struct mbuf *ip6e_rthdr; 115 struct mbuf *ip6e_dest2; 116 }; 117 118 static int ip6_pcbopt __P((int, u_char *, int, struct ip6_pktopts **, 119 int, int)); 120 static int ip6_getpcbopt __P((struct ip6_pktopts *, int, struct mbuf **)); 121 static int ip6_setpktopt __P((int, u_char *, int, struct ip6_pktopts *, int, 122 int, int, int)); 123 static int ip6_setmoptions __P((int, struct ip6_moptions **, struct mbuf *)); 124 static int ip6_getmoptions __P((int, struct ip6_moptions *, struct mbuf **)); 125 static int ip6_copyexthdr __P((struct mbuf **, caddr_t, int)); 126 static int ip6_insertfraghdr __P((struct mbuf *, struct mbuf *, int, 127 struct ip6_frag **)); 128 static int ip6_insert_jumboopt __P((struct ip6_exthdrs *, u_int32_t)); 129 static int ip6_splithdr __P((struct mbuf *, struct ip6_exthdrs *)); 130 static int ip6_getpmtu __P((struct route_in6 *, struct route_in6 *, 131 struct ifnet *, struct in6_addr *, u_long *, int *)); 132 static int copypktopts __P((struct ip6_pktopts *, struct ip6_pktopts *, int)); 133 134 #ifdef RFC2292 135 static int ip6_pcbopts __P((struct ip6_pktopts **, struct mbuf *, 136 struct socket *)); 137 #endif 138 139 #define IN6_NEED_CHECKSUM(ifp, csum_flags) \ 140 (__predict_true(((ifp)->if_flags & IFF_LOOPBACK) == 0 || \ 141 (((csum_flags) & M_CSUM_UDPv6) != 0 && udp_do_loopback_cksum) || \ 142 (((csum_flags) & M_CSUM_TCPv6) != 0 && tcp_do_loopback_cksum))) 143 144 /* 145 * IP6 output. The packet in mbuf chain m contains a skeletal IP6 146 * header (with pri, len, nxt, hlim, src, dst). 147 * This function may modify ver and hlim only. 148 * The mbuf chain containing the packet will be freed. 149 * The mbuf opt, if present, will not be freed. 150 * 151 * type of "mtu": rt_rmx.rmx_mtu is u_long, ifnet.ifr_mtu is int, and 152 * nd_ifinfo.linkmtu is u_int32_t. so we use u_long to hold largest one, 153 * which is rt_rmx.rmx_mtu. 154 */ 155 int 156 ip6_output(m0, opt, ro, flags, im6o, so, ifpp) 157 struct mbuf *m0; 158 struct ip6_pktopts *opt; 159 struct route_in6 *ro; 160 int flags; 161 struct ip6_moptions *im6o; 162 struct socket *so; 163 struct ifnet **ifpp; /* XXX: just for statistics */ 164 { 165 struct ip6_hdr *ip6, *mhip6; 166 struct ifnet *ifp, *origifp; 167 struct mbuf *m = m0; 168 int hlen, tlen, len, off; 169 struct route_in6 ip6route; 170 struct rtentry *rt = NULL; 171 struct sockaddr_in6 *dst, src_sa, dst_sa; 172 int error = 0; 173 struct in6_ifaddr *ia = NULL; 174 u_long mtu; 175 int alwaysfrag, dontfrag; 176 u_int32_t optlen = 0, plen = 0, unfragpartlen = 0; 177 struct ip6_exthdrs exthdrs; 178 struct in6_addr finaldst, src0, dst0; 179 u_int32_t zone; 180 struct route_in6 *ro_pmtu = NULL; 181 int hdrsplit = 0; 182 int needipsec = 0; 183 #ifdef IPSEC 184 int needipsectun = 0; 185 struct secpolicy *sp = NULL; 186 187 ip6 = mtod(m, struct ip6_hdr *); 188 #endif /* IPSEC */ 189 190 M_CSUM_DATA_IPv6_HL_SET(m->m_pkthdr.csum_data, sizeof(struct ip6_hdr)); 191 192 #define MAKE_EXTHDR(hp, mp) \ 193 do { \ 194 if (hp) { \ 195 struct ip6_ext *eh = (struct ip6_ext *)(hp); \ 196 error = ip6_copyexthdr((mp), (caddr_t)(hp), \ 197 ((eh)->ip6e_len + 1) << 3); \ 198 if (error) \ 199 goto freehdrs; \ 200 } \ 201 } while (/*CONSTCOND*/ 0) 202 203 bzero(&exthdrs, sizeof(exthdrs)); 204 if (opt) { 205 /* Hop-by-Hop options header */ 206 MAKE_EXTHDR(opt->ip6po_hbh, &exthdrs.ip6e_hbh); 207 /* Destination options header(1st part) */ 208 MAKE_EXTHDR(opt->ip6po_dest1, &exthdrs.ip6e_dest1); 209 /* Routing header */ 210 MAKE_EXTHDR(opt->ip6po_rthdr, &exthdrs.ip6e_rthdr); 211 /* Destination options header(2nd part) */ 212 MAKE_EXTHDR(opt->ip6po_dest2, &exthdrs.ip6e_dest2); 213 } 214 215 #ifdef IPSEC 216 if ((flags & IPV6_FORWARDING) != 0) { 217 needipsec = 0; 218 goto skippolicycheck; 219 } 220 221 /* get a security policy for this packet */ 222 if (so == NULL) 223 sp = ipsec6_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND, 0, &error); 224 else { 225 if (IPSEC_PCB_SKIP_IPSEC(sotoinpcb_hdr(so)->inph_sp, 226 IPSEC_DIR_OUTBOUND)) { 227 needipsec = 0; 228 goto skippolicycheck; 229 } 230 sp = ipsec6_getpolicybysock(m, IPSEC_DIR_OUTBOUND, so, &error); 231 } 232 233 if (sp == NULL) { 234 ipsec6stat.out_inval++; 235 goto freehdrs; 236 } 237 238 error = 0; 239 240 /* check policy */ 241 switch (sp->policy) { 242 case IPSEC_POLICY_DISCARD: 243 /* 244 * This packet is just discarded. 245 */ 246 ipsec6stat.out_polvio++; 247 goto freehdrs; 248 249 case IPSEC_POLICY_BYPASS: 250 case IPSEC_POLICY_NONE: 251 /* no need to do IPsec. */ 252 needipsec = 0; 253 break; 254 255 case IPSEC_POLICY_IPSEC: 256 if (sp->req == NULL) { 257 /* XXX should be panic ? */ 258 printf("ip6_output: No IPsec request specified.\n"); 259 error = EINVAL; 260 goto freehdrs; 261 } 262 needipsec = 1; 263 break; 264 265 case IPSEC_POLICY_ENTRUST: 266 default: 267 printf("ip6_output: Invalid policy found. %d\n", sp->policy); 268 } 269 270 skippolicycheck:; 271 #endif /* IPSEC */ 272 273 if (needipsec && 274 (m->m_pkthdr.csum_flags & (M_CSUM_UDPv6|M_CSUM_TCPv6)) != 0) { 275 in6_delayed_cksum(m); 276 m->m_pkthdr.csum_flags &= ~(M_CSUM_UDPv6|M_CSUM_TCPv6); 277 } 278 279 /* 280 * Calculate the total length of the extension header chain. 281 * Keep the length of the unfragmentable part for fragmentation. 282 */ 283 optlen = 0; 284 if (exthdrs.ip6e_hbh) optlen += exthdrs.ip6e_hbh->m_len; 285 if (exthdrs.ip6e_dest1) optlen += exthdrs.ip6e_dest1->m_len; 286 if (exthdrs.ip6e_rthdr) optlen += exthdrs.ip6e_rthdr->m_len; 287 unfragpartlen = optlen + sizeof(struct ip6_hdr); 288 /* NOTE: we don't add AH/ESP length here. do that later. */ 289 if (exthdrs.ip6e_dest2) optlen += exthdrs.ip6e_dest2->m_len; 290 291 /* 292 * If we need IPsec, or there is at least one extension header, 293 * separate IP6 header from the payload. 294 */ 295 if ((needipsec || optlen) && !hdrsplit) { 296 if ((error = ip6_splithdr(m, &exthdrs)) != 0) { 297 m = NULL; 298 goto freehdrs; 299 } 300 m = exthdrs.ip6e_ip6; 301 hdrsplit++; 302 } 303 304 /* adjust pointer */ 305 ip6 = mtod(m, struct ip6_hdr *); 306 307 /* adjust mbuf packet header length */ 308 m->m_pkthdr.len += optlen; 309 plen = m->m_pkthdr.len - sizeof(*ip6); 310 311 /* If this is a jumbo payload, insert a jumbo payload option. */ 312 if (plen > IPV6_MAXPACKET) { 313 if (!hdrsplit) { 314 if ((error = ip6_splithdr(m, &exthdrs)) != 0) { 315 m = NULL; 316 goto freehdrs; 317 } 318 m = exthdrs.ip6e_ip6; 319 hdrsplit++; 320 } 321 /* adjust pointer */ 322 ip6 = mtod(m, struct ip6_hdr *); 323 if ((error = ip6_insert_jumboopt(&exthdrs, plen)) != 0) 324 goto freehdrs; 325 optlen += 8; /* XXX JUMBOOPTLEN */ 326 ip6->ip6_plen = 0; 327 } else 328 ip6->ip6_plen = htons(plen); 329 330 /* 331 * Concatenate headers and fill in next header fields. 332 * Here we have, on "m" 333 * IPv6 payload 334 * and we insert headers accordingly. Finally, we should be getting: 335 * IPv6 hbh dest1 rthdr ah* [esp* dest2 payload] 336 * 337 * during the header composing process, "m" points to IPv6 header. 338 * "mprev" points to an extension header prior to esp. 339 */ 340 { 341 u_char *nexthdrp = &ip6->ip6_nxt; 342 struct mbuf *mprev = m; 343 344 /* 345 * we treat dest2 specially. this makes IPsec processing 346 * much easier. the goal here is to make mprev point the 347 * mbuf prior to dest2. 348 * 349 * result: IPv6 dest2 payload 350 * m and mprev will point to IPv6 header. 351 */ 352 if (exthdrs.ip6e_dest2) { 353 if (!hdrsplit) 354 panic("assumption failed: hdr not split"); 355 exthdrs.ip6e_dest2->m_next = m->m_next; 356 m->m_next = exthdrs.ip6e_dest2; 357 *mtod(exthdrs.ip6e_dest2, u_char *) = ip6->ip6_nxt; 358 ip6->ip6_nxt = IPPROTO_DSTOPTS; 359 } 360 361 #define MAKE_CHAIN(m, mp, p, i)\ 362 do {\ 363 if (m) {\ 364 if (!hdrsplit) \ 365 panic("assumption failed: hdr not split"); \ 366 *mtod((m), u_char *) = *(p);\ 367 *(p) = (i);\ 368 p = mtod((m), u_char *);\ 369 (m)->m_next = (mp)->m_next;\ 370 (mp)->m_next = (m);\ 371 (mp) = (m);\ 372 }\ 373 } while (/*CONSTCOND*/ 0) 374 /* 375 * result: IPv6 hbh dest1 rthdr dest2 payload 376 * m will point to IPv6 header. mprev will point to the 377 * extension header prior to dest2 (rthdr in the above case). 378 */ 379 MAKE_CHAIN(exthdrs.ip6e_hbh, mprev, nexthdrp, IPPROTO_HOPOPTS); 380 MAKE_CHAIN(exthdrs.ip6e_dest1, mprev, nexthdrp, 381 IPPROTO_DSTOPTS); 382 MAKE_CHAIN(exthdrs.ip6e_rthdr, mprev, nexthdrp, 383 IPPROTO_ROUTING); 384 385 M_CSUM_DATA_IPv6_HL_SET(m->m_pkthdr.csum_data, 386 sizeof(struct ip6_hdr) + optlen); 387 388 #ifdef IPSEC 389 if (!needipsec) 390 goto skip_ipsec2; 391 392 /* 393 * pointers after IPsec headers are not valid any more. 394 * other pointers need a great care too. 395 * (IPsec routines should not mangle mbufs prior to AH/ESP) 396 */ 397 exthdrs.ip6e_dest2 = NULL; 398 399 { 400 struct ip6_rthdr *rh = NULL; 401 int segleft_org = 0; 402 struct ipsec_output_state state; 403 404 if (exthdrs.ip6e_rthdr) { 405 rh = mtod(exthdrs.ip6e_rthdr, struct ip6_rthdr *); 406 segleft_org = rh->ip6r_segleft; 407 rh->ip6r_segleft = 0; 408 } 409 410 bzero(&state, sizeof(state)); 411 state.m = m; 412 error = ipsec6_output_trans(&state, nexthdrp, mprev, sp, flags, 413 &needipsectun); 414 m = state.m; 415 if (error) { 416 /* mbuf is already reclaimed in ipsec6_output_trans. */ 417 m = NULL; 418 switch (error) { 419 case EHOSTUNREACH: 420 case ENETUNREACH: 421 case EMSGSIZE: 422 case ENOBUFS: 423 case ENOMEM: 424 break; 425 default: 426 printf("ip6_output (ipsec): error code %d\n", error); 427 /* FALLTHROUGH */ 428 case ENOENT: 429 /* don't show these error codes to the user */ 430 error = 0; 431 break; 432 } 433 goto bad; 434 } 435 if (exthdrs.ip6e_rthdr) { 436 /* ah6_output doesn't modify mbuf chain */ 437 rh->ip6r_segleft = segleft_org; 438 } 439 } 440 skip_ipsec2:; 441 #endif 442 } 443 444 /* 445 * If there is a routing header, replace destination address field 446 * with the first hop of the routing header. 447 */ 448 if (exthdrs.ip6e_rthdr) { 449 struct ip6_rthdr *rh; 450 struct ip6_rthdr0 *rh0; 451 struct in6_addr *addr; 452 struct sockaddr_in6 sa; 453 454 rh = (struct ip6_rthdr *)(mtod(exthdrs.ip6e_rthdr, 455 struct ip6_rthdr *)); 456 finaldst = ip6->ip6_dst; 457 switch (rh->ip6r_type) { 458 case IPV6_RTHDR_TYPE_0: 459 rh0 = (struct ip6_rthdr0 *)rh; 460 addr = (struct in6_addr *)(rh0 + 1); 461 462 /* 463 * construct a sockaddr_in6 form of 464 * the first hop. 465 * 466 * XXX: we may not have enough 467 * information about its scope zone; 468 * there is no standard API to pass 469 * the information from the 470 * application. 471 */ 472 bzero(&sa, sizeof(sa)); 473 sa.sin6_family = AF_INET6; 474 sa.sin6_len = sizeof(sa); 475 sa.sin6_addr = addr[0]; 476 if ((error = sa6_embedscope(&sa, 477 ip6_use_defzone)) != 0) { 478 goto bad; 479 } 480 ip6->ip6_dst = sa.sin6_addr; 481 (void)memmove(&addr[0], &addr[1], 482 sizeof(struct in6_addr) * 483 (rh0->ip6r0_segleft - 1)); 484 addr[rh0->ip6r0_segleft - 1] = finaldst; 485 /* XXX */ 486 in6_clearscope(addr + rh0->ip6r0_segleft - 1); 487 break; 488 default: /* is it possible? */ 489 error = EINVAL; 490 goto bad; 491 } 492 } 493 494 /* Source address validation */ 495 if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) && 496 (flags & IPV6_UNSPECSRC) == 0) { 497 error = EOPNOTSUPP; 498 ip6stat.ip6s_badscope++; 499 goto bad; 500 } 501 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) { 502 error = EOPNOTSUPP; 503 ip6stat.ip6s_badscope++; 504 goto bad; 505 } 506 507 ip6stat.ip6s_localout++; 508 509 /* 510 * Route packet. 511 */ 512 /* initialize cached route */ 513 if (ro == 0) { 514 ro = &ip6route; 515 bzero((caddr_t)ro, sizeof(*ro)); 516 } 517 ro_pmtu = ro; 518 if (opt && opt->ip6po_rthdr) 519 ro = &opt->ip6po_route; 520 dst = (struct sockaddr_in6 *)&ro->ro_dst; 521 522 /* 523 * if specified, try to fill in the traffic class field. 524 * do not override if a non-zero value is already set. 525 * we check the diffserv field and the ecn field separately. 526 */ 527 if (opt && opt->ip6po_tclass >= 0) { 528 int mask = 0; 529 530 if ((ip6->ip6_flow & htonl(0xfc << 20)) == 0) 531 mask |= 0xfc; 532 if ((ip6->ip6_flow & htonl(0x03 << 20)) == 0) 533 mask |= 0x03; 534 if (mask != 0) 535 ip6->ip6_flow |= htonl((opt->ip6po_tclass & mask) << 20); 536 } 537 538 /* fill in or override the hop limit field, if necessary. */ 539 if (opt && opt->ip6po_hlim != -1) 540 ip6->ip6_hlim = opt->ip6po_hlim & 0xff; 541 else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) { 542 if (im6o != NULL) 543 ip6->ip6_hlim = im6o->im6o_multicast_hlim; 544 else 545 ip6->ip6_hlim = ip6_defmcasthlim; 546 } 547 548 #ifdef IPSEC 549 if (needipsec && needipsectun) { 550 struct ipsec_output_state state; 551 552 /* 553 * All the extension headers will become inaccessible 554 * (since they can be encrypted). 555 * Don't panic, we need no more updates to extension headers 556 * on inner IPv6 packet (since they are now encapsulated). 557 * 558 * IPv6 [ESP|AH] IPv6 [extension headers] payload 559 */ 560 bzero(&exthdrs, sizeof(exthdrs)); 561 exthdrs.ip6e_ip6 = m; 562 563 bzero(&state, sizeof(state)); 564 state.m = m; 565 state.ro = (struct route *)ro; 566 state.dst = (struct sockaddr *)dst; 567 568 error = ipsec6_output_tunnel(&state, sp, flags); 569 570 m = state.m; 571 ro_pmtu = ro = (struct route_in6 *)state.ro; 572 dst = (struct sockaddr_in6 *)state.dst; 573 if (error) { 574 /* mbuf is already reclaimed in ipsec6_output_tunnel. */ 575 m0 = m = NULL; 576 m = NULL; 577 switch (error) { 578 case EHOSTUNREACH: 579 case ENETUNREACH: 580 case EMSGSIZE: 581 case ENOBUFS: 582 case ENOMEM: 583 break; 584 default: 585 printf("ip6_output (ipsec): error code %d\n", error); 586 /* FALLTHROUGH */ 587 case ENOENT: 588 /* don't show these error codes to the user */ 589 error = 0; 590 break; 591 } 592 goto bad; 593 } 594 595 exthdrs.ip6e_ip6 = m; 596 } 597 #endif /* IPSEC */ 598 599 /* adjust pointer */ 600 ip6 = mtod(m, struct ip6_hdr *); 601 602 bzero(&dst_sa, sizeof(dst_sa)); 603 dst_sa.sin6_family = AF_INET6; 604 dst_sa.sin6_len = sizeof(dst_sa); 605 dst_sa.sin6_addr = ip6->ip6_dst; 606 if ((error = in6_selectroute(&dst_sa, opt, im6o, ro, &ifp, &rt, 0)) 607 != 0) { 608 switch (error) { 609 case EHOSTUNREACH: 610 ip6stat.ip6s_noroute++; 611 break; 612 case EADDRNOTAVAIL: 613 default: 614 break; /* XXX statistics? */ 615 } 616 if (ifp != NULL) 617 in6_ifstat_inc(ifp, ifs6_out_discard); 618 goto bad; 619 } 620 if (rt == NULL) { 621 /* 622 * If in6_selectroute() does not return a route entry, 623 * dst may not have been updated. 624 */ 625 *dst = dst_sa; /* XXX */ 626 } 627 628 /* 629 * then rt (for unicast) and ifp must be non-NULL valid values. 630 */ 631 if ((flags & IPV6_FORWARDING) == 0) { 632 /* XXX: the FORWARDING flag can be set for mrouting. */ 633 in6_ifstat_inc(ifp, ifs6_out_request); 634 } 635 if (rt != NULL) { 636 ia = (struct in6_ifaddr *)(rt->rt_ifa); 637 rt->rt_use++; 638 } 639 640 /* 641 * The outgoing interface must be in the zone of source and 642 * destination addresses. We should use ia_ifp to support the 643 * case of sending packets to an address of our own. 644 */ 645 if (ia != NULL && ia->ia_ifp) 646 origifp = ia->ia_ifp; 647 else 648 origifp = ifp; 649 650 src0 = ip6->ip6_src; 651 if (in6_setscope(&src0, origifp, &zone)) 652 goto badscope; 653 bzero(&src_sa, sizeof(src_sa)); 654 src_sa.sin6_family = AF_INET6; 655 src_sa.sin6_len = sizeof(src_sa); 656 src_sa.sin6_addr = ip6->ip6_src; 657 if (sa6_recoverscope(&src_sa) || zone != src_sa.sin6_scope_id) 658 goto badscope; 659 660 dst0 = ip6->ip6_dst; 661 if (in6_setscope(&dst0, origifp, &zone)) 662 goto badscope; 663 /* re-initialize to be sure */ 664 bzero(&dst_sa, sizeof(dst_sa)); 665 dst_sa.sin6_family = AF_INET6; 666 dst_sa.sin6_len = sizeof(dst_sa); 667 dst_sa.sin6_addr = ip6->ip6_dst; 668 if (sa6_recoverscope(&dst_sa) || zone != dst_sa.sin6_scope_id) 669 goto badscope; 670 671 /* scope check is done. */ 672 goto routefound; 673 674 badscope: 675 ip6stat.ip6s_badscope++; 676 in6_ifstat_inc(origifp, ifs6_out_discard); 677 if (error == 0) 678 error = EHOSTUNREACH; /* XXX */ 679 goto bad; 680 681 routefound: 682 if (rt && !IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) { 683 if (opt && opt->ip6po_nextroute.ro_rt) { 684 /* 685 * The nexthop is explicitly specified by the 686 * application. We assume the next hop is an IPv6 687 * address. 688 */ 689 dst = (struct sockaddr_in6 *)opt->ip6po_nexthop; 690 } else if ((rt->rt_flags & RTF_GATEWAY)) 691 dst = (struct sockaddr_in6 *)rt->rt_gateway; 692 } 693 694 /* 695 * XXXXXX: original code follows: 696 */ 697 if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) 698 m->m_flags &= ~(M_BCAST | M_MCAST); /* just in case */ 699 else { 700 struct in6_multi *in6m; 701 702 m->m_flags = (m->m_flags & ~M_BCAST) | M_MCAST; 703 704 in6_ifstat_inc(ifp, ifs6_out_mcast); 705 706 /* 707 * Confirm that the outgoing interface supports multicast. 708 */ 709 if (!(ifp->if_flags & IFF_MULTICAST)) { 710 ip6stat.ip6s_noroute++; 711 in6_ifstat_inc(ifp, ifs6_out_discard); 712 error = ENETUNREACH; 713 goto bad; 714 } 715 716 IN6_LOOKUP_MULTI(ip6->ip6_dst, ifp, in6m); 717 if (in6m != NULL && 718 (im6o == NULL || im6o->im6o_multicast_loop)) { 719 /* 720 * If we belong to the destination multicast group 721 * on the outgoing interface, and the caller did not 722 * forbid loopback, loop back a copy. 723 */ 724 ip6_mloopback(ifp, m, dst); 725 } else { 726 /* 727 * If we are acting as a multicast router, perform 728 * multicast forwarding as if the packet had just 729 * arrived on the interface to which we are about 730 * to send. The multicast forwarding function 731 * recursively calls this function, using the 732 * IPV6_FORWARDING flag to prevent infinite recursion. 733 * 734 * Multicasts that are looped back by ip6_mloopback(), 735 * above, will be forwarded by the ip6_input() routine, 736 * if necessary. 737 */ 738 if (ip6_mrouter && (flags & IPV6_FORWARDING) == 0) { 739 if (ip6_mforward(ip6, ifp, m) != 0) { 740 m_freem(m); 741 goto done; 742 } 743 } 744 } 745 /* 746 * Multicasts with a hoplimit of zero may be looped back, 747 * above, but must not be transmitted on a network. 748 * Also, multicasts addressed to the loopback interface 749 * are not sent -- the above call to ip6_mloopback() will 750 * loop back a copy if this host actually belongs to the 751 * destination group on the loopback interface. 752 */ 753 if (ip6->ip6_hlim == 0 || (ifp->if_flags & IFF_LOOPBACK) || 754 IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst)) { 755 m_freem(m); 756 goto done; 757 } 758 } 759 760 /* 761 * Fill the outgoing inteface to tell the upper layer 762 * to increment per-interface statistics. 763 */ 764 if (ifpp) 765 *ifpp = ifp; 766 767 /* Determine path MTU. */ 768 if ((error = ip6_getpmtu(ro_pmtu, ro, ifp, &finaldst, &mtu, 769 &alwaysfrag)) != 0) 770 goto bad; 771 #ifdef IPSEC 772 if (needipsectun) 773 mtu = IPV6_MMTU; 774 #endif 775 776 /* 777 * The caller of this function may specify to use the minimum MTU 778 * in some cases. 779 * An advanced API option (IPV6_USE_MIN_MTU) can also override MTU 780 * setting. The logic is a bit complicated; by default, unicast 781 * packets will follow path MTU while multicast packets will be sent at 782 * the minimum MTU. If IP6PO_MINMTU_ALL is specified, all packets 783 * including unicast ones will be sent at the minimum MTU. Multicast 784 * packets will always be sent at the minimum MTU unless 785 * IP6PO_MINMTU_DISABLE is explicitly specified. 786 * See RFC 3542 for more details. 787 */ 788 if (mtu > IPV6_MMTU) { 789 if ((flags & IPV6_MINMTU)) 790 mtu = IPV6_MMTU; 791 else if (opt && opt->ip6po_minmtu == IP6PO_MINMTU_ALL) 792 mtu = IPV6_MMTU; 793 else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) && 794 (opt == NULL || 795 opt->ip6po_minmtu != IP6PO_MINMTU_DISABLE)) { 796 mtu = IPV6_MMTU; 797 } 798 } 799 800 /* 801 * clear embedded scope identifiers if necessary. 802 * in6_clearscope will touch the addresses only when necessary. 803 */ 804 in6_clearscope(&ip6->ip6_src); 805 in6_clearscope(&ip6->ip6_dst); 806 807 /* 808 * If the outgoing packet contains a hop-by-hop options header, 809 * it must be examined and processed even by the source node. 810 * (RFC 2460, section 4.) 811 */ 812 if (exthdrs.ip6e_hbh) { 813 struct ip6_hbh *hbh = mtod(exthdrs.ip6e_hbh, struct ip6_hbh *); 814 u_int32_t dummy1; /* XXX unused */ 815 u_int32_t dummy2; /* XXX unused */ 816 817 /* 818 * XXX: if we have to send an ICMPv6 error to the sender, 819 * we need the M_LOOP flag since icmp6_error() expects 820 * the IPv6 and the hop-by-hop options header are 821 * continuous unless the flag is set. 822 */ 823 m->m_flags |= M_LOOP; 824 m->m_pkthdr.rcvif = ifp; 825 if (ip6_process_hopopts(m, (u_int8_t *)(hbh + 1), 826 ((hbh->ip6h_len + 1) << 3) - sizeof(struct ip6_hbh), 827 &dummy1, &dummy2) < 0) { 828 /* m was already freed at this point */ 829 error = EINVAL;/* better error? */ 830 goto done; 831 } 832 m->m_flags &= ~M_LOOP; /* XXX */ 833 m->m_pkthdr.rcvif = NULL; 834 } 835 836 #ifdef PFIL_HOOKS 837 /* 838 * Run through list of hooks for output packets. 839 */ 840 if ((error = pfil_run_hooks(&inet6_pfil_hook, &m, ifp, PFIL_OUT)) != 0) 841 goto done; 842 if (m == NULL) 843 goto done; 844 ip6 = mtod(m, struct ip6_hdr *); 845 #endif /* PFIL_HOOKS */ 846 /* 847 * Send the packet to the outgoing interface. 848 * If necessary, do IPv6 fragmentation before sending. 849 * 850 * the logic here is rather complex: 851 * 1: normal case (dontfrag == 0, alwaysfrag == 0) 852 * 1-a: send as is if tlen <= path mtu 853 * 1-b: fragment if tlen > path mtu 854 * 855 * 2: if user asks us not to fragment (dontfrag == 1) 856 * 2-a: send as is if tlen <= interface mtu 857 * 2-b: error if tlen > interface mtu 858 * 859 * 3: if we always need to attach fragment header (alwaysfrag == 1) 860 * always fragment 861 * 862 * 4: if dontfrag == 1 && alwaysfrag == 1 863 * error, as we cannot handle this conflicting request 864 */ 865 tlen = m->m_pkthdr.len; 866 867 if (opt && (opt->ip6po_flags & IP6PO_DONTFRAG)) 868 dontfrag = 1; 869 else 870 dontfrag = 0; 871 872 if (dontfrag && alwaysfrag) { /* case 4 */ 873 /* conflicting request - can't transmit */ 874 error = EMSGSIZE; 875 goto bad; 876 } 877 if (dontfrag && tlen > IN6_LINKMTU(ifp)) { /* case 2-b */ 878 /* 879 * Even if the DONTFRAG option is specified, we cannot send the 880 * packet when the data length is larger than the MTU of the 881 * outgoing interface. 882 * Notify the error by sending IPV6_PATHMTU ancillary data as 883 * well as returning an error code (the latter is not described 884 * in the API spec.) 885 */ 886 u_int32_t mtu32; 887 struct ip6ctlparam ip6cp; 888 889 mtu32 = (u_int32_t)mtu; 890 bzero(&ip6cp, sizeof(ip6cp)); 891 ip6cp.ip6c_cmdarg = (void *)&mtu32; 892 pfctlinput2(PRC_MSGSIZE, (struct sockaddr *)&ro_pmtu->ro_dst, 893 (void *)&ip6cp); 894 895 error = EMSGSIZE; 896 goto bad; 897 } 898 899 /* 900 * transmit packet without fragmentation 901 */ 902 if (dontfrag || (!alwaysfrag && tlen <= mtu)) { /* case 1-a and 2-a */ 903 struct in6_ifaddr *ia6; 904 int sw_csum; 905 906 ip6 = mtod(m, struct ip6_hdr *); 907 ia6 = in6_ifawithifp(ifp, &ip6->ip6_src); 908 if (ia6) { 909 /* Record statistics for this interface address. */ 910 ia6->ia_ifa.ifa_data.ifad_outbytes += m->m_pkthdr.len; 911 } 912 #ifdef IPSEC 913 /* clean ipsec history once it goes out of the node */ 914 ipsec_delaux(m); 915 #endif 916 917 sw_csum = m->m_pkthdr.csum_flags & ~ifp->if_csum_flags_tx; 918 if ((sw_csum & (M_CSUM_UDPv6|M_CSUM_TCPv6)) != 0) { 919 if (IN6_NEED_CHECKSUM(ifp, 920 sw_csum & (M_CSUM_UDPv6|M_CSUM_TCPv6))) { 921 in6_delayed_cksum(m); 922 } 923 m->m_pkthdr.csum_flags &= ~(M_CSUM_UDPv6|M_CSUM_TCPv6); 924 } 925 926 error = nd6_output(ifp, origifp, m, dst, rt); 927 goto done; 928 } 929 930 /* 931 * try to fragment the packet. case 1-b and 3 932 */ 933 if (mtu < IPV6_MMTU) { 934 /* path MTU cannot be less than IPV6_MMTU */ 935 error = EMSGSIZE; 936 in6_ifstat_inc(ifp, ifs6_out_fragfail); 937 goto bad; 938 } else if (ip6->ip6_plen == 0) { 939 /* jumbo payload cannot be fragmented */ 940 error = EMSGSIZE; 941 in6_ifstat_inc(ifp, ifs6_out_fragfail); 942 goto bad; 943 } else { 944 struct mbuf **mnext, *m_frgpart; 945 struct ip6_frag *ip6f; 946 u_int32_t id = htonl(ip6_randomid()); 947 u_char nextproto; 948 struct ip6ctlparam ip6cp; 949 u_int32_t mtu32; 950 951 /* 952 * Too large for the destination or interface; 953 * fragment if possible. 954 * Must be able to put at least 8 bytes per fragment. 955 */ 956 hlen = unfragpartlen; 957 if (mtu > IPV6_MAXPACKET) 958 mtu = IPV6_MAXPACKET; 959 960 /* Notify a proper path MTU to applications. */ 961 mtu32 = (u_int32_t)mtu; 962 bzero(&ip6cp, sizeof(ip6cp)); 963 ip6cp.ip6c_cmdarg = (void *)&mtu32; 964 pfctlinput2(PRC_MSGSIZE, (struct sockaddr *)&ro_pmtu->ro_dst, 965 (void *)&ip6cp); 966 967 len = (mtu - hlen - sizeof(struct ip6_frag)) & ~7; 968 if (len < 8) { 969 error = EMSGSIZE; 970 in6_ifstat_inc(ifp, ifs6_out_fragfail); 971 goto bad; 972 } 973 974 mnext = &m->m_nextpkt; 975 976 /* 977 * Change the next header field of the last header in the 978 * unfragmentable part. 979 */ 980 if (exthdrs.ip6e_rthdr) { 981 nextproto = *mtod(exthdrs.ip6e_rthdr, u_char *); 982 *mtod(exthdrs.ip6e_rthdr, u_char *) = IPPROTO_FRAGMENT; 983 } else if (exthdrs.ip6e_dest1) { 984 nextproto = *mtod(exthdrs.ip6e_dest1, u_char *); 985 *mtod(exthdrs.ip6e_dest1, u_char *) = IPPROTO_FRAGMENT; 986 } else if (exthdrs.ip6e_hbh) { 987 nextproto = *mtod(exthdrs.ip6e_hbh, u_char *); 988 *mtod(exthdrs.ip6e_hbh, u_char *) = IPPROTO_FRAGMENT; 989 } else { 990 nextproto = ip6->ip6_nxt; 991 ip6->ip6_nxt = IPPROTO_FRAGMENT; 992 } 993 994 if ((m->m_pkthdr.csum_flags & (M_CSUM_UDPv6|M_CSUM_TCPv6)) 995 != 0) { 996 if (IN6_NEED_CHECKSUM(ifp, 997 m->m_pkthdr.csum_flags & 998 (M_CSUM_UDPv6|M_CSUM_TCPv6))) { 999 in6_delayed_cksum(m); 1000 } 1001 m->m_pkthdr.csum_flags &= ~(M_CSUM_UDPv6|M_CSUM_TCPv6); 1002 } 1003 1004 /* 1005 * Loop through length of segment after first fragment, 1006 * make new header and copy data of each part and link onto 1007 * chain. 1008 */ 1009 m0 = m; 1010 for (off = hlen; off < tlen; off += len) { 1011 struct mbuf *mlast; 1012 1013 MGETHDR(m, M_DONTWAIT, MT_HEADER); 1014 if (!m) { 1015 error = ENOBUFS; 1016 ip6stat.ip6s_odropped++; 1017 goto sendorfree; 1018 } 1019 m->m_pkthdr.rcvif = NULL; 1020 m->m_flags = m0->m_flags & M_COPYFLAGS; 1021 *mnext = m; 1022 mnext = &m->m_nextpkt; 1023 m->m_data += max_linkhdr; 1024 mhip6 = mtod(m, struct ip6_hdr *); 1025 *mhip6 = *ip6; 1026 m->m_len = sizeof(*mhip6); 1027 error = ip6_insertfraghdr(m0, m, hlen, &ip6f); 1028 if (error) { 1029 ip6stat.ip6s_odropped++; 1030 goto sendorfree; 1031 } 1032 ip6f->ip6f_offlg = htons((u_int16_t)((off - hlen) & ~7)); 1033 if (off + len >= tlen) 1034 len = tlen - off; 1035 else 1036 ip6f->ip6f_offlg |= IP6F_MORE_FRAG; 1037 mhip6->ip6_plen = htons((u_int16_t)(len + hlen + 1038 sizeof(*ip6f) - sizeof(struct ip6_hdr))); 1039 if ((m_frgpart = m_copy(m0, off, len)) == 0) { 1040 error = ENOBUFS; 1041 ip6stat.ip6s_odropped++; 1042 goto sendorfree; 1043 } 1044 for (mlast = m; mlast->m_next; mlast = mlast->m_next) 1045 ; 1046 mlast->m_next = m_frgpart; 1047 m->m_pkthdr.len = len + hlen + sizeof(*ip6f); 1048 m->m_pkthdr.rcvif = (struct ifnet *)0; 1049 ip6f->ip6f_reserved = 0; 1050 ip6f->ip6f_ident = id; 1051 ip6f->ip6f_nxt = nextproto; 1052 ip6stat.ip6s_ofragments++; 1053 in6_ifstat_inc(ifp, ifs6_out_fragcreat); 1054 } 1055 1056 in6_ifstat_inc(ifp, ifs6_out_fragok); 1057 } 1058 1059 /* 1060 * Remove leading garbages. 1061 */ 1062 sendorfree: 1063 m = m0->m_nextpkt; 1064 m0->m_nextpkt = 0; 1065 m_freem(m0); 1066 for (m0 = m; m; m = m0) { 1067 m0 = m->m_nextpkt; 1068 m->m_nextpkt = 0; 1069 if (error == 0) { 1070 struct in6_ifaddr *ia6; 1071 ip6 = mtod(m, struct ip6_hdr *); 1072 ia6 = in6_ifawithifp(ifp, &ip6->ip6_src); 1073 if (ia6) { 1074 /* 1075 * Record statistics for this interface 1076 * address. 1077 */ 1078 ia6->ia_ifa.ifa_data.ifad_outbytes += 1079 m->m_pkthdr.len; 1080 } 1081 #ifdef IPSEC 1082 /* clean ipsec history once it goes out of the node */ 1083 ipsec_delaux(m); 1084 #endif 1085 error = nd6_output(ifp, origifp, m, dst, rt); 1086 } else 1087 m_freem(m); 1088 } 1089 1090 if (error == 0) 1091 ip6stat.ip6s_fragmented++; 1092 1093 done: 1094 if (ro == &ip6route && ro->ro_rt) { /* brace necessary for RTFREE */ 1095 RTFREE(ro->ro_rt); 1096 } else if (ro_pmtu == &ip6route && ro_pmtu->ro_rt) { 1097 RTFREE(ro_pmtu->ro_rt); 1098 } 1099 1100 #ifdef IPSEC 1101 if (sp != NULL) 1102 key_freesp(sp); 1103 #endif /* IPSEC */ 1104 1105 return (error); 1106 1107 freehdrs: 1108 m_freem(exthdrs.ip6e_hbh); /* m_freem will check if mbuf is 0 */ 1109 m_freem(exthdrs.ip6e_dest1); 1110 m_freem(exthdrs.ip6e_rthdr); 1111 m_freem(exthdrs.ip6e_dest2); 1112 /* FALLTHROUGH */ 1113 bad: 1114 m_freem(m); 1115 goto done; 1116 } 1117 1118 static int 1119 ip6_copyexthdr(mp, hdr, hlen) 1120 struct mbuf **mp; 1121 caddr_t hdr; 1122 int hlen; 1123 { 1124 struct mbuf *m; 1125 1126 if (hlen > MCLBYTES) 1127 return (ENOBUFS); /* XXX */ 1128 1129 MGET(m, M_DONTWAIT, MT_DATA); 1130 if (!m) 1131 return (ENOBUFS); 1132 1133 if (hlen > MLEN) { 1134 MCLGET(m, M_DONTWAIT); 1135 if ((m->m_flags & M_EXT) == 0) { 1136 m_free(m); 1137 return (ENOBUFS); 1138 } 1139 } 1140 m->m_len = hlen; 1141 if (hdr) 1142 bcopy(hdr, mtod(m, caddr_t), hlen); 1143 1144 *mp = m; 1145 return (0); 1146 } 1147 1148 /* 1149 * Process a delayed payload checksum calculation. 1150 */ 1151 void 1152 in6_delayed_cksum(struct mbuf *m) 1153 { 1154 uint16_t csum, offset; 1155 1156 KASSERT((m->m_pkthdr.csum_flags & (M_CSUM_UDPv6|M_CSUM_TCPv6)) != 0); 1157 KASSERT((~m->m_pkthdr.csum_flags & (M_CSUM_UDPv6|M_CSUM_TCPv6)) != 0); 1158 KASSERT((m->m_pkthdr.csum_flags 1159 & (M_CSUM_UDPv4|M_CSUM_TCPv4|M_CSUM_TSOv4)) == 0); 1160 1161 offset = M_CSUM_DATA_IPv6_HL(m->m_pkthdr.csum_data); 1162 csum = in6_cksum(m, 0, offset, m->m_pkthdr.len - offset); 1163 if (csum == 0 && (m->m_pkthdr.csum_flags & M_CSUM_UDPv6) != 0) { 1164 csum = 0xffff; 1165 } 1166 1167 offset += M_CSUM_DATA_IPv6_OFFSET(m->m_pkthdr.csum_data); 1168 if ((offset + sizeof(csum)) > m->m_len) { 1169 m_copyback(m, offset, sizeof(csum), &csum); 1170 } else { 1171 *(uint16_t *)(mtod(m, caddr_t) + offset) = csum; 1172 } 1173 } 1174 1175 /* 1176 * Insert jumbo payload option. 1177 */ 1178 static int 1179 ip6_insert_jumboopt(exthdrs, plen) 1180 struct ip6_exthdrs *exthdrs; 1181 u_int32_t plen; 1182 { 1183 struct mbuf *mopt; 1184 u_int8_t *optbuf; 1185 u_int32_t v; 1186 1187 #define JUMBOOPTLEN 8 /* length of jumbo payload option and padding */ 1188 1189 /* 1190 * If there is no hop-by-hop options header, allocate new one. 1191 * If there is one but it doesn't have enough space to store the 1192 * jumbo payload option, allocate a cluster to store the whole options. 1193 * Otherwise, use it to store the options. 1194 */ 1195 if (exthdrs->ip6e_hbh == 0) { 1196 MGET(mopt, M_DONTWAIT, MT_DATA); 1197 if (mopt == 0) 1198 return (ENOBUFS); 1199 mopt->m_len = JUMBOOPTLEN; 1200 optbuf = mtod(mopt, u_int8_t *); 1201 optbuf[1] = 0; /* = ((JUMBOOPTLEN) >> 3) - 1 */ 1202 exthdrs->ip6e_hbh = mopt; 1203 } else { 1204 struct ip6_hbh *hbh; 1205 1206 mopt = exthdrs->ip6e_hbh; 1207 if (M_TRAILINGSPACE(mopt) < JUMBOOPTLEN) { 1208 /* 1209 * XXX assumption: 1210 * - exthdrs->ip6e_hbh is not referenced from places 1211 * other than exthdrs. 1212 * - exthdrs->ip6e_hbh is not an mbuf chain. 1213 */ 1214 int oldoptlen = mopt->m_len; 1215 struct mbuf *n; 1216 1217 /* 1218 * XXX: give up if the whole (new) hbh header does 1219 * not fit even in an mbuf cluster. 1220 */ 1221 if (oldoptlen + JUMBOOPTLEN > MCLBYTES) 1222 return (ENOBUFS); 1223 1224 /* 1225 * As a consequence, we must always prepare a cluster 1226 * at this point. 1227 */ 1228 MGET(n, M_DONTWAIT, MT_DATA); 1229 if (n) { 1230 MCLGET(n, M_DONTWAIT); 1231 if ((n->m_flags & M_EXT) == 0) { 1232 m_freem(n); 1233 n = NULL; 1234 } 1235 } 1236 if (!n) 1237 return (ENOBUFS); 1238 n->m_len = oldoptlen + JUMBOOPTLEN; 1239 bcopy(mtod(mopt, caddr_t), mtod(n, caddr_t), 1240 oldoptlen); 1241 optbuf = mtod(n, u_int8_t *) + oldoptlen; 1242 m_freem(mopt); 1243 mopt = exthdrs->ip6e_hbh = n; 1244 } else { 1245 optbuf = mtod(mopt, u_int8_t *) + mopt->m_len; 1246 mopt->m_len += JUMBOOPTLEN; 1247 } 1248 optbuf[0] = IP6OPT_PADN; 1249 optbuf[1] = 0; 1250 1251 /* 1252 * Adjust the header length according to the pad and 1253 * the jumbo payload option. 1254 */ 1255 hbh = mtod(mopt, struct ip6_hbh *); 1256 hbh->ip6h_len += (JUMBOOPTLEN >> 3); 1257 } 1258 1259 /* fill in the option. */ 1260 optbuf[2] = IP6OPT_JUMBO; 1261 optbuf[3] = 4; 1262 v = (u_int32_t)htonl(plen + JUMBOOPTLEN); 1263 bcopy(&v, &optbuf[4], sizeof(u_int32_t)); 1264 1265 /* finally, adjust the packet header length */ 1266 exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN; 1267 1268 return (0); 1269 #undef JUMBOOPTLEN 1270 } 1271 1272 /* 1273 * Insert fragment header and copy unfragmentable header portions. 1274 */ 1275 static int 1276 ip6_insertfraghdr(m0, m, hlen, frghdrp) 1277 struct mbuf *m0, *m; 1278 int hlen; 1279 struct ip6_frag **frghdrp; 1280 { 1281 struct mbuf *n, *mlast; 1282 1283 if (hlen > sizeof(struct ip6_hdr)) { 1284 n = m_copym(m0, sizeof(struct ip6_hdr), 1285 hlen - sizeof(struct ip6_hdr), M_DONTWAIT); 1286 if (n == 0) 1287 return (ENOBUFS); 1288 m->m_next = n; 1289 } else 1290 n = m; 1291 1292 /* Search for the last mbuf of unfragmentable part. */ 1293 for (mlast = n; mlast->m_next; mlast = mlast->m_next) 1294 ; 1295 1296 if ((mlast->m_flags & M_EXT) == 0 && 1297 M_TRAILINGSPACE(mlast) >= sizeof(struct ip6_frag)) { 1298 /* use the trailing space of the last mbuf for the fragment hdr */ 1299 *frghdrp = (struct ip6_frag *)(mtod(mlast, caddr_t) + 1300 mlast->m_len); 1301 mlast->m_len += sizeof(struct ip6_frag); 1302 m->m_pkthdr.len += sizeof(struct ip6_frag); 1303 } else { 1304 /* allocate a new mbuf for the fragment header */ 1305 struct mbuf *mfrg; 1306 1307 MGET(mfrg, M_DONTWAIT, MT_DATA); 1308 if (mfrg == 0) 1309 return (ENOBUFS); 1310 mfrg->m_len = sizeof(struct ip6_frag); 1311 *frghdrp = mtod(mfrg, struct ip6_frag *); 1312 mlast->m_next = mfrg; 1313 } 1314 1315 return (0); 1316 } 1317 1318 static int 1319 ip6_getpmtu(ro_pmtu, ro, ifp, dst, mtup, alwaysfragp) 1320 struct route_in6 *ro_pmtu, *ro; 1321 struct ifnet *ifp; 1322 struct in6_addr *dst; 1323 u_long *mtup; 1324 int *alwaysfragp; 1325 { 1326 u_int32_t mtu = 0; 1327 int alwaysfrag = 0; 1328 int error = 0; 1329 1330 if (ro_pmtu != ro) { 1331 /* The first hop and the final destination may differ. */ 1332 struct sockaddr_in6 *sa6_dst = 1333 (struct sockaddr_in6 *)&ro_pmtu->ro_dst; 1334 if (ro_pmtu->ro_rt && 1335 ((ro_pmtu->ro_rt->rt_flags & RTF_UP) == 0 || 1336 !IN6_ARE_ADDR_EQUAL(&sa6_dst->sin6_addr, dst))) { 1337 RTFREE(ro_pmtu->ro_rt); 1338 ro_pmtu->ro_rt = (struct rtentry *)NULL; 1339 } 1340 if (ro_pmtu->ro_rt == NULL) { 1341 bzero(sa6_dst, sizeof(*sa6_dst)); /* for safety */ 1342 sa6_dst->sin6_family = AF_INET6; 1343 sa6_dst->sin6_len = sizeof(struct sockaddr_in6); 1344 sa6_dst->sin6_addr = *dst; 1345 1346 rtalloc((struct route *)ro_pmtu); 1347 } 1348 } 1349 if (ro_pmtu->ro_rt) { 1350 u_int32_t ifmtu; 1351 1352 if (ifp == NULL) 1353 ifp = ro_pmtu->ro_rt->rt_ifp; 1354 ifmtu = IN6_LINKMTU(ifp); 1355 mtu = ro_pmtu->ro_rt->rt_rmx.rmx_mtu; 1356 if (mtu == 0) 1357 mtu = ifmtu; 1358 else if (mtu < IPV6_MMTU) { 1359 /* 1360 * RFC2460 section 5, last paragraph: 1361 * if we record ICMPv6 too big message with 1362 * mtu < IPV6_MMTU, transmit packets sized IPV6_MMTU 1363 * or smaller, with fragment header attached. 1364 * (fragment header is needed regardless from the 1365 * packet size, for translators to identify packets) 1366 */ 1367 alwaysfrag = 1; 1368 mtu = IPV6_MMTU; 1369 } else if (mtu > ifmtu) { 1370 /* 1371 * The MTU on the route is larger than the MTU on 1372 * the interface! This shouldn't happen, unless the 1373 * MTU of the interface has been changed after the 1374 * interface was brought up. Change the MTU in the 1375 * route to match the interface MTU (as long as the 1376 * field isn't locked). 1377 */ 1378 mtu = ifmtu; 1379 if (!(ro_pmtu->ro_rt->rt_rmx.rmx_locks & RTV_MTU)) 1380 ro_pmtu->ro_rt->rt_rmx.rmx_mtu = mtu; 1381 } 1382 } else if (ifp) { 1383 mtu = IN6_LINKMTU(ifp); 1384 } else 1385 error = EHOSTUNREACH; /* XXX */ 1386 1387 *mtup = mtu; 1388 if (alwaysfragp) 1389 *alwaysfragp = alwaysfrag; 1390 return (error); 1391 } 1392 1393 /* 1394 * IP6 socket option processing. 1395 */ 1396 int 1397 ip6_ctloutput(op, so, level, optname, mp) 1398 int op; 1399 struct socket *so; 1400 int level, optname; 1401 struct mbuf **mp; 1402 { 1403 int privileged, optdatalen, uproto; 1404 void *optdata; 1405 struct in6pcb *in6p = sotoin6pcb(so); 1406 struct mbuf *m = *mp; 1407 int error, optval; 1408 int optlen; 1409 struct proc *p = curproc; /* XXX */ 1410 1411 optlen = m ? m->m_len : 0; 1412 error = optval = 0; 1413 privileged = (p == 0 || suser(p->p_ucred, &p->p_acflag)) ? 0 : 1; 1414 uproto = (int)so->so_proto->pr_protocol; 1415 1416 if (level == IPPROTO_IPV6) { 1417 switch (op) { 1418 case PRCO_SETOPT: 1419 switch (optname) { 1420 #ifdef RFC2292 1421 case IPV6_2292PKTOPTIONS: 1422 /* m is freed in ip6_pcbopts */ 1423 error = ip6_pcbopts(&in6p->in6p_outputopts, 1424 m, so); 1425 break; 1426 #endif 1427 1428 /* 1429 * Use of some Hop-by-Hop options or some 1430 * Destination options, might require special 1431 * privilege. That is, normal applications 1432 * (without special privilege) might be forbidden 1433 * from setting certain options in outgoing packets, 1434 * and might never see certain options in received 1435 * packets. [RFC 2292 Section 6] 1436 * KAME specific note: 1437 * KAME prevents non-privileged users from sending or 1438 * receiving ANY hbh/dst options in order to avoid 1439 * overhead of parsing options in the kernel. 1440 */ 1441 case IPV6_RECVHOPOPTS: 1442 case IPV6_RECVDSTOPTS: 1443 case IPV6_RECVRTHDRDSTOPTS: 1444 if (!privileged) { 1445 error = EPERM; 1446 break; 1447 } 1448 /* FALLTHROUGH */ 1449 case IPV6_UNICAST_HOPS: 1450 case IPV6_HOPLIMIT: 1451 case IPV6_FAITH: 1452 1453 case IPV6_RECVPKTINFO: 1454 case IPV6_RECVHOPLIMIT: 1455 case IPV6_RECVRTHDR: 1456 case IPV6_RECVPATHMTU: 1457 case IPV6_RECVTCLASS: 1458 case IPV6_V6ONLY: 1459 if (optlen != sizeof(int)) { 1460 error = EINVAL; 1461 break; 1462 } 1463 optval = *mtod(m, int *); 1464 switch (optname) { 1465 1466 case IPV6_UNICAST_HOPS: 1467 if (optval < -1 || optval >= 256) 1468 error = EINVAL; 1469 else { 1470 /* -1 = kernel default */ 1471 in6p->in6p_hops = optval; 1472 } 1473 break; 1474 #define OPTSET(bit) \ 1475 do { \ 1476 if (optval) \ 1477 in6p->in6p_flags |= (bit); \ 1478 else \ 1479 in6p->in6p_flags &= ~(bit); \ 1480 } while (/*CONSTCOND*/ 0) 1481 1482 #ifdef RFC2292 1483 #define OPTSET2292(bit) \ 1484 do { \ 1485 in6p->in6p_flags |= IN6P_RFC2292; \ 1486 if (optval) \ 1487 in6p->in6p_flags |= (bit); \ 1488 else \ 1489 in6p->in6p_flags &= ~(bit); \ 1490 } while (/*CONSTCOND*/ 0) 1491 #endif 1492 1493 #define OPTBIT(bit) (in6p->in6p_flags & (bit) ? 1 : 0) 1494 1495 case IPV6_RECVPKTINFO: 1496 #ifdef RFC2292 1497 /* cannot mix with RFC2292 */ 1498 if (OPTBIT(IN6P_RFC2292)) { 1499 error = EINVAL; 1500 break; 1501 } 1502 #endif 1503 OPTSET(IN6P_PKTINFO); 1504 break; 1505 1506 case IPV6_HOPLIMIT: 1507 { 1508 struct ip6_pktopts **optp; 1509 1510 #ifdef RFC2292 1511 /* cannot mix with RFC2292 */ 1512 if (OPTBIT(IN6P_RFC2292)) { 1513 error = EINVAL; 1514 break; 1515 } 1516 #endif 1517 optp = &in6p->in6p_outputopts; 1518 error = ip6_pcbopt(IPV6_HOPLIMIT, 1519 (u_char *)&optval, 1520 sizeof(optval), 1521 optp, 1522 privileged, uproto); 1523 break; 1524 } 1525 1526 case IPV6_RECVHOPLIMIT: 1527 #ifdef RFC2292 1528 /* cannot mix with RFC2292 */ 1529 if (OPTBIT(IN6P_RFC2292)) { 1530 error = EINVAL; 1531 break; 1532 } 1533 #endif 1534 OPTSET(IN6P_HOPLIMIT); 1535 break; 1536 1537 case IPV6_RECVHOPOPTS: 1538 #ifdef RFC2292 1539 /* cannot mix with RFC2292 */ 1540 if (OPTBIT(IN6P_RFC2292)) { 1541 error = EINVAL; 1542 break; 1543 } 1544 #endif 1545 OPTSET(IN6P_HOPOPTS); 1546 break; 1547 1548 case IPV6_RECVDSTOPTS: 1549 #ifdef RFC2292 1550 /* cannot mix with RFC2292 */ 1551 if (OPTBIT(IN6P_RFC2292)) { 1552 error = EINVAL; 1553 break; 1554 } 1555 #endif 1556 OPTSET(IN6P_DSTOPTS); 1557 break; 1558 1559 case IPV6_RECVRTHDRDSTOPTS: 1560 #ifdef RFC2292 1561 /* cannot mix with RFC2292 */ 1562 if (OPTBIT(IN6P_RFC2292)) { 1563 error = EINVAL; 1564 break; 1565 } 1566 #endif 1567 OPTSET(IN6P_RTHDRDSTOPTS); 1568 break; 1569 1570 case IPV6_RECVRTHDR: 1571 #ifdef RFC2292 1572 /* cannot mix with RFC2292 */ 1573 if (OPTBIT(IN6P_RFC2292)) { 1574 error = EINVAL; 1575 break; 1576 } 1577 #endif 1578 OPTSET(IN6P_RTHDR); 1579 break; 1580 1581 case IPV6_FAITH: 1582 OPTSET(IN6P_FAITH); 1583 break; 1584 1585 case IPV6_RECVPATHMTU: 1586 /* 1587 * We ignore this option for TCP 1588 * sockets. 1589 * (RFC3542 leaves this case 1590 * unspecified.) 1591 */ 1592 if (uproto != IPPROTO_TCP) 1593 OPTSET(IN6P_MTU); 1594 break; 1595 1596 case IPV6_V6ONLY: 1597 /* 1598 * make setsockopt(IPV6_V6ONLY) 1599 * available only prior to bind(2). 1600 * see ipng mailing list, Jun 22 2001. 1601 */ 1602 if (in6p->in6p_lport || 1603 !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr)) { 1604 error = EINVAL; 1605 break; 1606 } 1607 #ifdef INET6_BINDV6ONLY 1608 if (!optval) 1609 error = EINVAL; 1610 #else 1611 OPTSET(IN6P_IPV6_V6ONLY); 1612 #endif 1613 break; 1614 case IPV6_RECVTCLASS: 1615 #ifdef RFC2292 1616 /* cannot mix with RFC2292 XXX */ 1617 if (OPTBIT(IN6P_RFC2292)) { 1618 error = EINVAL; 1619 break; 1620 } 1621 #endif 1622 OPTSET(IN6P_TCLASS); 1623 break; 1624 1625 } 1626 break; 1627 1628 case IPV6_OTCLASS: 1629 { 1630 struct ip6_pktopts **optp; 1631 u_int8_t tclass; 1632 1633 if (optlen != sizeof(tclass)) { 1634 error = EINVAL; 1635 break; 1636 } 1637 tclass = *mtod(m, u_int8_t *); 1638 optp = &in6p->in6p_outputopts; 1639 error = ip6_pcbopt(optname, 1640 (u_char *)&tclass, 1641 sizeof(tclass), 1642 optp, 1643 privileged, uproto); 1644 break; 1645 } 1646 1647 case IPV6_TCLASS: 1648 case IPV6_DONTFRAG: 1649 case IPV6_USE_MIN_MTU: 1650 if (optlen != sizeof(optval)) { 1651 error = EINVAL; 1652 break; 1653 } 1654 optval = *mtod(m, int *); 1655 { 1656 struct ip6_pktopts **optp; 1657 optp = &in6p->in6p_outputopts; 1658 error = ip6_pcbopt(optname, 1659 (u_char *)&optval, 1660 sizeof(optval), 1661 optp, 1662 privileged, uproto); 1663 break; 1664 } 1665 1666 #ifdef RFC2292 1667 case IPV6_2292PKTINFO: 1668 case IPV6_2292HOPLIMIT: 1669 case IPV6_2292HOPOPTS: 1670 case IPV6_2292DSTOPTS: 1671 case IPV6_2292RTHDR: 1672 /* RFC 2292 */ 1673 if (optlen != sizeof(int)) { 1674 error = EINVAL; 1675 break; 1676 } 1677 optval = *mtod(m, int *); 1678 switch (optname) { 1679 case IPV6_2292PKTINFO: 1680 OPTSET2292(IN6P_PKTINFO); 1681 break; 1682 case IPV6_2292HOPLIMIT: 1683 OPTSET2292(IN6P_HOPLIMIT); 1684 break; 1685 case IPV6_2292HOPOPTS: 1686 /* 1687 * Check super-user privilege. 1688 * See comments for IPV6_RECVHOPOPTS. 1689 */ 1690 if (!privileged) 1691 return (EPERM); 1692 OPTSET2292(IN6P_HOPOPTS); 1693 break; 1694 case IPV6_2292DSTOPTS: 1695 if (!privileged) 1696 return (EPERM); 1697 OPTSET2292(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); /* XXX */ 1698 break; 1699 case IPV6_2292RTHDR: 1700 OPTSET2292(IN6P_RTHDR); 1701 break; 1702 } 1703 break; 1704 #endif 1705 case IPV6_PKTINFO: 1706 case IPV6_HOPOPTS: 1707 case IPV6_RTHDR: 1708 case IPV6_DSTOPTS: 1709 case IPV6_RTHDRDSTOPTS: 1710 case IPV6_NEXTHOP: 1711 { 1712 /* new advanced API (RFC3542) */ 1713 u_char *optbuf; 1714 int optbuflen; 1715 struct ip6_pktopts **optp; 1716 1717 #ifdef RFC2292 1718 /* cannot mix with RFC2292 */ 1719 if (OPTBIT(IN6P_RFC2292)) { 1720 error = EINVAL; 1721 break; 1722 } 1723 #endif 1724 1725 if (m && m->m_next) { 1726 error = EINVAL; /* XXX */ 1727 break; 1728 } 1729 if (m) { 1730 optbuf = mtod(m, u_char *); 1731 optbuflen = m->m_len; 1732 } else { 1733 optbuf = NULL; 1734 optbuflen = 0; 1735 } 1736 optp = &in6p->in6p_outputopts; 1737 error = ip6_pcbopt(optname, 1738 optbuf, optbuflen, 1739 optp, privileged, uproto); 1740 break; 1741 } 1742 #undef OPTSET 1743 1744 case IPV6_MULTICAST_IF: 1745 case IPV6_MULTICAST_HOPS: 1746 case IPV6_MULTICAST_LOOP: 1747 case IPV6_JOIN_GROUP: 1748 case IPV6_LEAVE_GROUP: 1749 error = ip6_setmoptions(optname, 1750 &in6p->in6p_moptions, m); 1751 break; 1752 1753 case IPV6_PORTRANGE: 1754 optval = *mtod(m, int *); 1755 1756 switch (optval) { 1757 case IPV6_PORTRANGE_DEFAULT: 1758 in6p->in6p_flags &= ~(IN6P_LOWPORT); 1759 in6p->in6p_flags &= ~(IN6P_HIGHPORT); 1760 break; 1761 1762 case IPV6_PORTRANGE_HIGH: 1763 in6p->in6p_flags &= ~(IN6P_LOWPORT); 1764 in6p->in6p_flags |= IN6P_HIGHPORT; 1765 break; 1766 1767 case IPV6_PORTRANGE_LOW: 1768 in6p->in6p_flags &= ~(IN6P_HIGHPORT); 1769 in6p->in6p_flags |= IN6P_LOWPORT; 1770 break; 1771 1772 default: 1773 error = EINVAL; 1774 break; 1775 } 1776 break; 1777 1778 #ifdef IPSEC 1779 case IPV6_IPSEC_POLICY: 1780 { 1781 caddr_t req = NULL; 1782 size_t len = 0; 1783 if (m) { 1784 req = mtod(m, caddr_t); 1785 len = m->m_len; 1786 } 1787 error = ipsec6_set_policy(in6p, optname, req, 1788 len, privileged); 1789 } 1790 break; 1791 #endif /* IPSEC */ 1792 1793 default: 1794 error = ENOPROTOOPT; 1795 break; 1796 } 1797 if (m) 1798 (void)m_free(m); 1799 break; 1800 1801 case PRCO_GETOPT: 1802 switch (optname) { 1803 #ifdef RFC2292 1804 case IPV6_2292PKTOPTIONS: 1805 /* 1806 * RFC3542 (effectively) deprecated the 1807 * semantics of the 2292-style pktoptions. 1808 * Since it was not reliable in nature (i.e., 1809 * applications had to expect the lack of some 1810 * information after all), it would make sense 1811 * to simplify this part by always returning 1812 * empty data. 1813 */ 1814 *mp = m_get(M_WAIT, MT_SOOPTS); 1815 (*mp)->m_len = 0; 1816 break; 1817 #endif 1818 1819 case IPV6_RECVHOPOPTS: 1820 case IPV6_RECVDSTOPTS: 1821 case IPV6_RECVRTHDRDSTOPTS: 1822 case IPV6_UNICAST_HOPS: 1823 case IPV6_RECVPKTINFO: 1824 case IPV6_RECVHOPLIMIT: 1825 case IPV6_RECVRTHDR: 1826 case IPV6_RECVPATHMTU: 1827 1828 case IPV6_FAITH: 1829 case IPV6_V6ONLY: 1830 case IPV6_PORTRANGE: 1831 case IPV6_RECVTCLASS: 1832 switch (optname) { 1833 1834 case IPV6_RECVHOPOPTS: 1835 optval = OPTBIT(IN6P_HOPOPTS); 1836 break; 1837 1838 case IPV6_RECVDSTOPTS: 1839 optval = OPTBIT(IN6P_DSTOPTS); 1840 break; 1841 1842 case IPV6_RECVRTHDRDSTOPTS: 1843 optval = OPTBIT(IN6P_RTHDRDSTOPTS); 1844 break; 1845 1846 case IPV6_UNICAST_HOPS: 1847 optval = in6p->in6p_hops; 1848 break; 1849 1850 case IPV6_RECVPKTINFO: 1851 optval = OPTBIT(IN6P_PKTINFO); 1852 break; 1853 1854 case IPV6_RECVHOPLIMIT: 1855 optval = OPTBIT(IN6P_HOPLIMIT); 1856 break; 1857 1858 case IPV6_RECVRTHDR: 1859 optval = OPTBIT(IN6P_RTHDR); 1860 break; 1861 1862 case IPV6_RECVPATHMTU: 1863 optval = OPTBIT(IN6P_MTU); 1864 break; 1865 1866 case IPV6_FAITH: 1867 optval = OPTBIT(IN6P_FAITH); 1868 break; 1869 1870 case IPV6_V6ONLY: 1871 optval = OPTBIT(IN6P_IPV6_V6ONLY); 1872 break; 1873 1874 case IPV6_PORTRANGE: 1875 { 1876 int flags; 1877 flags = in6p->in6p_flags; 1878 if (flags & IN6P_HIGHPORT) 1879 optval = IPV6_PORTRANGE_HIGH; 1880 else if (flags & IN6P_LOWPORT) 1881 optval = IPV6_PORTRANGE_LOW; 1882 else 1883 optval = 0; 1884 break; 1885 } 1886 case IPV6_RECVTCLASS: 1887 optval = OPTBIT(IN6P_TCLASS); 1888 break; 1889 1890 } 1891 if (error) 1892 break; 1893 *mp = m = m_get(M_WAIT, MT_SOOPTS); 1894 m->m_len = sizeof(int); 1895 *mtod(m, int *) = optval; 1896 break; 1897 1898 case IPV6_PATHMTU: 1899 { 1900 u_long pmtu = 0; 1901 struct ip6_mtuinfo mtuinfo; 1902 struct route_in6 *ro = (struct route_in6 *)&in6p 1903 ->in6p_route; 1904 1905 if (!(so->so_state & SS_ISCONNECTED)) 1906 return (ENOTCONN); 1907 /* 1908 * XXX: we dot not consider the case of source 1909 * routing, or optional information to specify 1910 * the outgoing interface. 1911 */ 1912 error = ip6_getpmtu(ro, NULL, NULL, 1913 &in6p->in6p_faddr, &pmtu, NULL); 1914 if (error) 1915 break; 1916 if (pmtu > IPV6_MAXPACKET) 1917 pmtu = IPV6_MAXPACKET; 1918 1919 memset(&mtuinfo, 0, sizeof(mtuinfo)); 1920 mtuinfo.ip6m_mtu = (u_int32_t)pmtu; 1921 optdata = (void *)&mtuinfo; 1922 optdatalen = sizeof(mtuinfo); 1923 if (optdatalen > MCLBYTES) 1924 return (EMSGSIZE); /* XXX */ 1925 *mp = m = m_get(M_WAIT, MT_SOOPTS); 1926 if (optdatalen > MLEN) 1927 MCLGET(m, M_WAIT); 1928 m->m_len = optdatalen; 1929 memcpy(mtod(m, void *), optdata, optdatalen); 1930 break; 1931 } 1932 1933 #ifdef RFC2292 1934 case IPV6_2292PKTINFO: 1935 case IPV6_2292HOPLIMIT: 1936 case IPV6_2292HOPOPTS: 1937 case IPV6_2292RTHDR: 1938 case IPV6_2292DSTOPTS: 1939 switch (optname) { 1940 case IPV6_2292PKTINFO: 1941 optval = OPTBIT(IN6P_PKTINFO); 1942 break; 1943 case IPV6_2292HOPLIMIT: 1944 optval = OPTBIT(IN6P_HOPLIMIT); 1945 break; 1946 case IPV6_2292HOPOPTS: 1947 optval = OPTBIT(IN6P_HOPOPTS); 1948 break; 1949 case IPV6_2292RTHDR: 1950 optval = OPTBIT(IN6P_RTHDR); 1951 break; 1952 case IPV6_2292DSTOPTS: 1953 optval = OPTBIT(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); 1954 break; 1955 } 1956 *mp = m = m_get(M_WAIT, MT_SOOPTS); 1957 m->m_len = sizeof(int); 1958 *mtod(m, int *) = optval; 1959 break; 1960 #endif 1961 case IPV6_PKTINFO: 1962 case IPV6_HOPOPTS: 1963 case IPV6_RTHDR: 1964 case IPV6_DSTOPTS: 1965 case IPV6_RTHDRDSTOPTS: 1966 case IPV6_NEXTHOP: 1967 case IPV6_OTCLASS: 1968 case IPV6_TCLASS: 1969 case IPV6_DONTFRAG: 1970 case IPV6_USE_MIN_MTU: 1971 error = ip6_getpcbopt(in6p->in6p_outputopts, 1972 optname, mp); 1973 break; 1974 1975 case IPV6_MULTICAST_IF: 1976 case IPV6_MULTICAST_HOPS: 1977 case IPV6_MULTICAST_LOOP: 1978 case IPV6_JOIN_GROUP: 1979 case IPV6_LEAVE_GROUP: 1980 error = ip6_getmoptions(optname, 1981 in6p->in6p_moptions, mp); 1982 break; 1983 1984 #ifdef IPSEC 1985 case IPV6_IPSEC_POLICY: 1986 { 1987 caddr_t req = NULL; 1988 size_t len = 0; 1989 if (m) { 1990 req = mtod(m, caddr_t); 1991 len = m->m_len; 1992 } 1993 error = ipsec6_get_policy(in6p, req, len, mp); 1994 break; 1995 } 1996 #endif /* IPSEC */ 1997 1998 1999 2000 2001 default: 2002 error = ENOPROTOOPT; 2003 break; 2004 } 2005 break; 2006 } 2007 } else { 2008 error = EINVAL; 2009 if (op == PRCO_SETOPT && *mp) 2010 (void)m_free(*mp); 2011 } 2012 return (error); 2013 } 2014 2015 int 2016 ip6_raw_ctloutput(op, so, level, optname, mp) 2017 int op; 2018 struct socket *so; 2019 int level, optname; 2020 struct mbuf **mp; 2021 { 2022 int error = 0, optval, optlen; 2023 const int icmp6off = offsetof(struct icmp6_hdr, icmp6_cksum); 2024 struct in6pcb *in6p = sotoin6pcb(so); 2025 struct mbuf *m = *mp; 2026 2027 optlen = m ? m->m_len : 0; 2028 2029 if (level != IPPROTO_IPV6) { 2030 if (op == PRCO_SETOPT && *mp) 2031 (void)m_free(*mp); 2032 return (EINVAL); 2033 } 2034 2035 switch (optname) { 2036 case IPV6_CHECKSUM: 2037 /* 2038 * For ICMPv6 sockets, no modification allowed for checksum 2039 * offset, permit "no change" values to help existing apps. 2040 * 2041 * XXX RFC3542 says: "An attempt to set IPV6_CHECKSUM 2042 * for an ICMPv6 socket will fail." The current 2043 * behavior does not meet RFC3542. 2044 */ 2045 switch (op) { 2046 case PRCO_SETOPT: 2047 if (optlen != sizeof(int)) { 2048 error = EINVAL; 2049 break; 2050 } 2051 optval = *mtod(m, int *); 2052 if ((optval % 2) != 0) { 2053 /* the API assumes even offset values */ 2054 error = EINVAL; 2055 } else if (so->so_proto->pr_protocol == 2056 IPPROTO_ICMPV6) { 2057 if (optval != icmp6off) 2058 error = EINVAL; 2059 } else 2060 in6p->in6p_cksum = optval; 2061 break; 2062 2063 case PRCO_GETOPT: 2064 if (so->so_proto->pr_protocol == IPPROTO_ICMPV6) 2065 optval = icmp6off; 2066 else 2067 optval = in6p->in6p_cksum; 2068 2069 *mp = m = m_get(M_WAIT, MT_SOOPTS); 2070 m->m_len = sizeof(int); 2071 *mtod(m, int *) = optval; 2072 break; 2073 2074 default: 2075 error = EINVAL; 2076 break; 2077 } 2078 break; 2079 2080 default: 2081 error = ENOPROTOOPT; 2082 break; 2083 } 2084 2085 if (op == PRCO_SETOPT && m) 2086 (void)m_free(m); 2087 2088 return (error); 2089 } 2090 2091 #ifdef RFC2292 2092 /* 2093 * Set up IP6 options in pcb for insertion in output packets or 2094 * specifying behavior of outgoing packets. 2095 */ 2096 static int 2097 ip6_pcbopts(pktopt, m, so) 2098 struct ip6_pktopts **pktopt; 2099 struct mbuf *m; 2100 struct socket *so; 2101 { 2102 struct ip6_pktopts *opt = *pktopt; 2103 int error = 0; 2104 struct proc *p = curproc; /* XXX */ 2105 int priv = 0; 2106 2107 /* turn off any old options. */ 2108 if (opt) { 2109 #ifdef DIAGNOSTIC 2110 if (opt->ip6po_pktinfo || opt->ip6po_nexthop || 2111 opt->ip6po_hbh || opt->ip6po_dest1 || opt->ip6po_dest2 || 2112 opt->ip6po_rhinfo.ip6po_rhi_rthdr) 2113 printf("ip6_pcbopts: all specified options are cleared.\n"); 2114 #endif 2115 ip6_clearpktopts(opt, -1); 2116 } else 2117 opt = malloc(sizeof(*opt), M_IP6OPT, M_WAITOK); 2118 *pktopt = NULL; 2119 2120 if (!m || m->m_len == 0) { 2121 /* 2122 * Only turning off any previous options, regardless of 2123 * whether the opt is just created or given. 2124 */ 2125 free(opt, M_IP6OPT); 2126 return (0); 2127 } 2128 2129 /* set options specified by user. */ 2130 if (p && !suser(p->p_ucred, &p->p_acflag)) 2131 priv = 1; 2132 if ((error = ip6_setpktopts(m, opt, NULL, priv, 2133 so->so_proto->pr_protocol)) != 0) { 2134 ip6_clearpktopts(opt, -1); /* XXX: discard all options */ 2135 free(opt, M_IP6OPT); 2136 return (error); 2137 } 2138 *pktopt = opt; 2139 return (0); 2140 } 2141 #endif 2142 2143 /* 2144 * initialize ip6_pktopts. beware that there are non-zero default values in 2145 * the struct. 2146 */ 2147 void 2148 ip6_initpktopts(struct ip6_pktopts *opt) 2149 { 2150 2151 memset(opt, 0, sizeof(*opt)); 2152 opt->ip6po_hlim = -1; /* -1 means default hop limit */ 2153 opt->ip6po_tclass = -1; /* -1 means default traffic class */ 2154 opt->ip6po_minmtu = IP6PO_MINMTU_MCASTONLY; 2155 } 2156 2157 #define sin6tosa(sin6) ((struct sockaddr *)(sin6)) /* XXX */ 2158 static int 2159 ip6_pcbopt(int optname, u_char *buf, int len, struct ip6_pktopts **pktopt, 2160 int priv, int uproto) 2161 { 2162 struct ip6_pktopts *opt; 2163 2164 if (*pktopt == NULL) { 2165 *pktopt = malloc(sizeof(struct ip6_pktopts), M_IP6OPT, 2166 M_WAITOK); 2167 ip6_initpktopts(*pktopt); 2168 } 2169 opt = *pktopt; 2170 2171 return (ip6_setpktopt(optname, buf, len, opt, priv, 1, 0, uproto)); 2172 } 2173 2174 static int 2175 ip6_getpcbopt(struct ip6_pktopts *pktopt, int optname, struct mbuf **mp) 2176 { 2177 void *optdata = NULL; 2178 int optdatalen = 0; 2179 struct ip6_ext *ip6e; 2180 int error = 0; 2181 struct in6_pktinfo null_pktinfo; 2182 int deftclass = 0, on; 2183 int defminmtu = IP6PO_MINMTU_MCASTONLY; 2184 struct mbuf *m; 2185 2186 switch (optname) { 2187 case IPV6_PKTINFO: 2188 if (pktopt && pktopt->ip6po_pktinfo) 2189 optdata = (void *)pktopt->ip6po_pktinfo; 2190 else { 2191 /* XXX: we don't have to do this every time... */ 2192 memset(&null_pktinfo, 0, sizeof(null_pktinfo)); 2193 optdata = (void *)&null_pktinfo; 2194 } 2195 optdatalen = sizeof(struct in6_pktinfo); 2196 break; 2197 case IPV6_OTCLASS: 2198 /* XXX */ 2199 return (EINVAL); 2200 case IPV6_TCLASS: 2201 if (pktopt && pktopt->ip6po_tclass >= 0) 2202 optdata = (void *)&pktopt->ip6po_tclass; 2203 else 2204 optdata = (void *)&deftclass; 2205 optdatalen = sizeof(int); 2206 break; 2207 case IPV6_HOPOPTS: 2208 if (pktopt && pktopt->ip6po_hbh) { 2209 optdata = (void *)pktopt->ip6po_hbh; 2210 ip6e = (struct ip6_ext *)pktopt->ip6po_hbh; 2211 optdatalen = (ip6e->ip6e_len + 1) << 3; 2212 } 2213 break; 2214 case IPV6_RTHDR: 2215 if (pktopt && pktopt->ip6po_rthdr) { 2216 optdata = (void *)pktopt->ip6po_rthdr; 2217 ip6e = (struct ip6_ext *)pktopt->ip6po_rthdr; 2218 optdatalen = (ip6e->ip6e_len + 1) << 3; 2219 } 2220 break; 2221 case IPV6_RTHDRDSTOPTS: 2222 if (pktopt && pktopt->ip6po_dest1) { 2223 optdata = (void *)pktopt->ip6po_dest1; 2224 ip6e = (struct ip6_ext *)pktopt->ip6po_dest1; 2225 optdatalen = (ip6e->ip6e_len + 1) << 3; 2226 } 2227 break; 2228 case IPV6_DSTOPTS: 2229 if (pktopt && pktopt->ip6po_dest2) { 2230 optdata = (void *)pktopt->ip6po_dest2; 2231 ip6e = (struct ip6_ext *)pktopt->ip6po_dest2; 2232 optdatalen = (ip6e->ip6e_len + 1) << 3; 2233 } 2234 break; 2235 case IPV6_NEXTHOP: 2236 if (pktopt && pktopt->ip6po_nexthop) { 2237 optdata = (void *)pktopt->ip6po_nexthop; 2238 optdatalen = pktopt->ip6po_nexthop->sa_len; 2239 } 2240 break; 2241 case IPV6_USE_MIN_MTU: 2242 if (pktopt) 2243 optdata = (void *)&pktopt->ip6po_minmtu; 2244 else 2245 optdata = (void *)&defminmtu; 2246 optdatalen = sizeof(int); 2247 break; 2248 case IPV6_DONTFRAG: 2249 if (pktopt && ((pktopt->ip6po_flags) & IP6PO_DONTFRAG)) 2250 on = 1; 2251 else 2252 on = 0; 2253 optdata = (void *)&on; 2254 optdatalen = sizeof(on); 2255 break; 2256 default: /* should not happen */ 2257 #ifdef DIAGNOSTIC 2258 panic("ip6_getpcbopt: unexpected option\n"); 2259 #endif 2260 return (ENOPROTOOPT); 2261 } 2262 2263 if (optdatalen > MCLBYTES) 2264 return (EMSGSIZE); /* XXX */ 2265 *mp = m = m_get(M_WAIT, MT_SOOPTS); 2266 if (optdatalen > MLEN) 2267 MCLGET(m, M_WAIT); 2268 m->m_len = optdatalen; 2269 if (optdatalen) 2270 memcpy(mtod(m, void *), optdata, optdatalen); 2271 2272 return (error); 2273 } 2274 2275 void 2276 ip6_clearpktopts(struct ip6_pktopts *pktopt, int optname) 2277 { 2278 if (optname == -1 || optname == IPV6_PKTINFO) { 2279 if (pktopt->ip6po_pktinfo) 2280 free(pktopt->ip6po_pktinfo, M_IP6OPT); 2281 pktopt->ip6po_pktinfo = NULL; 2282 } 2283 if (optname == -1 || optname == IPV6_HOPLIMIT) 2284 pktopt->ip6po_hlim = -1; 2285 if (optname == -1 || optname == IPV6_TCLASS) 2286 pktopt->ip6po_tclass = -1; 2287 if (optname == -1 || optname == IPV6_NEXTHOP) { 2288 if (pktopt->ip6po_nextroute.ro_rt) { 2289 RTFREE(pktopt->ip6po_nextroute.ro_rt); 2290 pktopt->ip6po_nextroute.ro_rt = NULL; 2291 } 2292 if (pktopt->ip6po_nexthop) 2293 free(pktopt->ip6po_nexthop, M_IP6OPT); 2294 pktopt->ip6po_nexthop = NULL; 2295 } 2296 if (optname == -1 || optname == IPV6_HOPOPTS) { 2297 if (pktopt->ip6po_hbh) 2298 free(pktopt->ip6po_hbh, M_IP6OPT); 2299 pktopt->ip6po_hbh = NULL; 2300 } 2301 if (optname == -1 || optname == IPV6_RTHDRDSTOPTS) { 2302 if (pktopt->ip6po_dest1) 2303 free(pktopt->ip6po_dest1, M_IP6OPT); 2304 pktopt->ip6po_dest1 = NULL; 2305 } 2306 if (optname == -1 || optname == IPV6_RTHDR) { 2307 if (pktopt->ip6po_rhinfo.ip6po_rhi_rthdr) 2308 free(pktopt->ip6po_rhinfo.ip6po_rhi_rthdr, M_IP6OPT); 2309 pktopt->ip6po_rhinfo.ip6po_rhi_rthdr = NULL; 2310 if (pktopt->ip6po_route.ro_rt) { 2311 RTFREE(pktopt->ip6po_route.ro_rt); 2312 pktopt->ip6po_route.ro_rt = NULL; 2313 } 2314 } 2315 if (optname == -1 || optname == IPV6_DSTOPTS) { 2316 if (pktopt->ip6po_dest2) 2317 free(pktopt->ip6po_dest2, M_IP6OPT); 2318 pktopt->ip6po_dest2 = NULL; 2319 } 2320 } 2321 2322 #define PKTOPT_EXTHDRCPY(type) \ 2323 do { \ 2324 if (src->type) { \ 2325 int hlen = (((struct ip6_ext *)src->type)->ip6e_len + 1) << 3;\ 2326 dst->type = malloc(hlen, M_IP6OPT, canwait); \ 2327 if (dst->type == NULL && canwait == M_NOWAIT) \ 2328 goto bad; \ 2329 memcpy(dst->type, src->type, hlen); \ 2330 } \ 2331 } while (/*CONSTCOND*/ 0) 2332 2333 static int 2334 copypktopts(struct ip6_pktopts *dst, struct ip6_pktopts *src, int canwait) 2335 { 2336 dst->ip6po_hlim = src->ip6po_hlim; 2337 dst->ip6po_tclass = src->ip6po_tclass; 2338 dst->ip6po_flags = src->ip6po_flags; 2339 if (src->ip6po_pktinfo) { 2340 dst->ip6po_pktinfo = malloc(sizeof(*dst->ip6po_pktinfo), 2341 M_IP6OPT, canwait); 2342 if (dst->ip6po_pktinfo == NULL && canwait == M_NOWAIT) 2343 goto bad; 2344 *dst->ip6po_pktinfo = *src->ip6po_pktinfo; 2345 } 2346 if (src->ip6po_nexthop) { 2347 dst->ip6po_nexthop = malloc(src->ip6po_nexthop->sa_len, 2348 M_IP6OPT, canwait); 2349 if (dst->ip6po_nexthop == NULL && canwait == M_NOWAIT) 2350 goto bad; 2351 memcpy(dst->ip6po_nexthop, src->ip6po_nexthop, 2352 src->ip6po_nexthop->sa_len); 2353 } 2354 PKTOPT_EXTHDRCPY(ip6po_hbh); 2355 PKTOPT_EXTHDRCPY(ip6po_dest1); 2356 PKTOPT_EXTHDRCPY(ip6po_dest2); 2357 PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */ 2358 return (0); 2359 2360 bad: 2361 if (dst->ip6po_pktinfo) free(dst->ip6po_pktinfo, M_IP6OPT); 2362 if (dst->ip6po_nexthop) free(dst->ip6po_nexthop, M_IP6OPT); 2363 if (dst->ip6po_hbh) free(dst->ip6po_hbh, M_IP6OPT); 2364 if (dst->ip6po_dest1) free(dst->ip6po_dest1, M_IP6OPT); 2365 if (dst->ip6po_dest2) free(dst->ip6po_dest2, M_IP6OPT); 2366 if (dst->ip6po_rthdr) free(dst->ip6po_rthdr, M_IP6OPT); 2367 2368 return (ENOBUFS); 2369 } 2370 #undef PKTOPT_EXTHDRCPY 2371 2372 struct ip6_pktopts * 2373 ip6_copypktopts(struct ip6_pktopts *src, int canwait) 2374 { 2375 int error; 2376 struct ip6_pktopts *dst; 2377 2378 dst = malloc(sizeof(*dst), M_IP6OPT, canwait); 2379 if (dst == NULL && canwait == M_NOWAIT) 2380 return (NULL); 2381 ip6_initpktopts(dst); 2382 2383 if ((error = copypktopts(dst, src, canwait)) != 0) { 2384 free(dst, M_IP6OPT); 2385 return (NULL); 2386 } 2387 2388 return (dst); 2389 } 2390 2391 void 2392 ip6_freepcbopts(struct ip6_pktopts *pktopt) 2393 { 2394 if (pktopt == NULL) 2395 return; 2396 2397 ip6_clearpktopts(pktopt, -1); 2398 2399 free(pktopt, M_IP6OPT); 2400 } 2401 2402 /* 2403 * Set the IP6 multicast options in response to user setsockopt(). 2404 */ 2405 static int 2406 ip6_setmoptions(optname, im6op, m) 2407 int optname; 2408 struct ip6_moptions **im6op; 2409 struct mbuf *m; 2410 { 2411 int error = 0; 2412 u_int loop, ifindex; 2413 struct ipv6_mreq *mreq; 2414 struct ifnet *ifp; 2415 struct ip6_moptions *im6o = *im6op; 2416 struct route_in6 ro; 2417 struct in6_multi_mship *imm; 2418 struct proc *p = curproc; /* XXX */ 2419 2420 if (im6o == NULL) { 2421 /* 2422 * No multicast option buffer attached to the pcb; 2423 * allocate one and initialize to default values. 2424 */ 2425 im6o = (struct ip6_moptions *) 2426 malloc(sizeof(*im6o), M_IPMOPTS, M_WAITOK); 2427 2428 if (im6o == NULL) 2429 return (ENOBUFS); 2430 *im6op = im6o; 2431 im6o->im6o_multicast_ifp = NULL; 2432 im6o->im6o_multicast_hlim = ip6_defmcasthlim; 2433 im6o->im6o_multicast_loop = IPV6_DEFAULT_MULTICAST_LOOP; 2434 LIST_INIT(&im6o->im6o_memberships); 2435 } 2436 2437 switch (optname) { 2438 2439 case IPV6_MULTICAST_IF: 2440 /* 2441 * Select the interface for outgoing multicast packets. 2442 */ 2443 if (m == NULL || m->m_len != sizeof(u_int)) { 2444 error = EINVAL; 2445 break; 2446 } 2447 bcopy(mtod(m, u_int *), &ifindex, sizeof(ifindex)); 2448 if (ifindex != 0) { 2449 if (ifindex < 0 || if_indexlim <= ifindex || 2450 !ifindex2ifnet[ifindex]) { 2451 error = ENXIO; /* XXX EINVAL? */ 2452 break; 2453 } 2454 ifp = ifindex2ifnet[ifindex]; 2455 if ((ifp->if_flags & IFF_MULTICAST) == 0) { 2456 error = EADDRNOTAVAIL; 2457 break; 2458 } 2459 } else 2460 ifp = NULL; 2461 im6o->im6o_multicast_ifp = ifp; 2462 break; 2463 2464 case IPV6_MULTICAST_HOPS: 2465 { 2466 /* 2467 * Set the IP6 hoplimit for outgoing multicast packets. 2468 */ 2469 int optval; 2470 if (m == NULL || m->m_len != sizeof(int)) { 2471 error = EINVAL; 2472 break; 2473 } 2474 bcopy(mtod(m, u_int *), &optval, sizeof(optval)); 2475 if (optval < -1 || optval >= 256) 2476 error = EINVAL; 2477 else if (optval == -1) 2478 im6o->im6o_multicast_hlim = ip6_defmcasthlim; 2479 else 2480 im6o->im6o_multicast_hlim = optval; 2481 break; 2482 } 2483 2484 case IPV6_MULTICAST_LOOP: 2485 /* 2486 * Set the loopback flag for outgoing multicast packets. 2487 * Must be zero or one. 2488 */ 2489 if (m == NULL || m->m_len != sizeof(u_int)) { 2490 error = EINVAL; 2491 break; 2492 } 2493 bcopy(mtod(m, u_int *), &loop, sizeof(loop)); 2494 if (loop > 1) { 2495 error = EINVAL; 2496 break; 2497 } 2498 im6o->im6o_multicast_loop = loop; 2499 break; 2500 2501 case IPV6_JOIN_GROUP: 2502 /* 2503 * Add a multicast group membership. 2504 * Group must be a valid IP6 multicast address. 2505 */ 2506 if (m == NULL || m->m_len != sizeof(struct ipv6_mreq)) { 2507 error = EINVAL; 2508 break; 2509 } 2510 mreq = mtod(m, struct ipv6_mreq *); 2511 if (IN6_IS_ADDR_UNSPECIFIED(&mreq->ipv6mr_multiaddr)) { 2512 /* 2513 * We use the unspecified address to specify to accept 2514 * all multicast addresses. Only super user is allowed 2515 * to do this. 2516 */ 2517 if (suser(p->p_ucred, &p->p_acflag)) 2518 { 2519 error = EACCES; 2520 break; 2521 } 2522 } else if (!IN6_IS_ADDR_MULTICAST(&mreq->ipv6mr_multiaddr)) { 2523 error = EINVAL; 2524 break; 2525 } 2526 2527 /* 2528 * If no interface was explicitly specified, choose an 2529 * appropriate one according to the given multicast address. 2530 */ 2531 if (mreq->ipv6mr_interface == 0) { 2532 struct sockaddr_in6 *dst; 2533 2534 /* 2535 * Look up the routing table for the 2536 * address, and choose the outgoing interface. 2537 * XXX: is it a good approach? 2538 */ 2539 ro.ro_rt = NULL; 2540 dst = (struct sockaddr_in6 *)&ro.ro_dst; 2541 bzero(dst, sizeof(*dst)); 2542 dst->sin6_family = AF_INET6; 2543 dst->sin6_len = sizeof(*dst); 2544 dst->sin6_addr = mreq->ipv6mr_multiaddr; 2545 rtalloc((struct route *)&ro); 2546 if (ro.ro_rt == NULL) { 2547 error = EADDRNOTAVAIL; 2548 break; 2549 } 2550 ifp = ro.ro_rt->rt_ifp; 2551 rtfree(ro.ro_rt); 2552 } else { 2553 /* 2554 * If the interface is specified, validate it. 2555 */ 2556 if (mreq->ipv6mr_interface < 0 || 2557 if_indexlim <= mreq->ipv6mr_interface || 2558 !ifindex2ifnet[mreq->ipv6mr_interface]) { 2559 error = ENXIO; /* XXX EINVAL? */ 2560 break; 2561 } 2562 ifp = ifindex2ifnet[mreq->ipv6mr_interface]; 2563 } 2564 2565 /* 2566 * See if we found an interface, and confirm that it 2567 * supports multicast 2568 */ 2569 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) { 2570 error = EADDRNOTAVAIL; 2571 break; 2572 } 2573 2574 if (in6_setscope(&mreq->ipv6mr_multiaddr, ifp, NULL)) { 2575 error = EADDRNOTAVAIL; /* XXX: should not happen */ 2576 break; 2577 } 2578 2579 /* 2580 * See if the membership already exists. 2581 */ 2582 for (imm = im6o->im6o_memberships.lh_first; 2583 imm != NULL; imm = imm->i6mm_chain.le_next) 2584 if (imm->i6mm_maddr->in6m_ifp == ifp && 2585 IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr, 2586 &mreq->ipv6mr_multiaddr)) 2587 break; 2588 if (imm != NULL) { 2589 error = EADDRINUSE; 2590 break; 2591 } 2592 /* 2593 * Everything looks good; add a new record to the multicast 2594 * address list for the given interface. 2595 */ 2596 imm = in6_joingroup(ifp, &mreq->ipv6mr_multiaddr, &error, 0); 2597 if (imm == NULL) 2598 break; 2599 LIST_INSERT_HEAD(&im6o->im6o_memberships, imm, i6mm_chain); 2600 break; 2601 2602 case IPV6_LEAVE_GROUP: 2603 /* 2604 * Drop a multicast group membership. 2605 * Group must be a valid IP6 multicast address. 2606 */ 2607 if (m == NULL || m->m_len != sizeof(struct ipv6_mreq)) { 2608 error = EINVAL; 2609 break; 2610 } 2611 mreq = mtod(m, struct ipv6_mreq *); 2612 2613 /* 2614 * If an interface address was specified, get a pointer 2615 * to its ifnet structure. 2616 */ 2617 if (mreq->ipv6mr_interface != 0) { 2618 if (mreq->ipv6mr_interface < 0 || 2619 if_indexlim <= mreq->ipv6mr_interface || 2620 !ifindex2ifnet[mreq->ipv6mr_interface]) { 2621 error = ENXIO; /* XXX EINVAL? */ 2622 break; 2623 } 2624 ifp = ifindex2ifnet[mreq->ipv6mr_interface]; 2625 } else 2626 ifp = NULL; 2627 2628 /* Fill in the scope zone ID */ 2629 if (ifp) { 2630 if (in6_setscope(&mreq->ipv6mr_multiaddr, ifp, NULL)) { 2631 /* XXX: should not happen */ 2632 error = EADDRNOTAVAIL; 2633 break; 2634 } 2635 } else if (mreq->ipv6mr_interface != 0) { 2636 /* 2637 * XXX: This case would happens when the (positive) 2638 * index is in the valid range, but the corresponding 2639 * interface has been detached dynamically. The above 2640 * check probably avoids such case to happen here, but 2641 * we check it explicitly for safety. 2642 */ 2643 error = EADDRNOTAVAIL; 2644 break; 2645 } else { /* ipv6mr_interface == 0 */ 2646 struct sockaddr_in6 sa6_mc; 2647 2648 /* 2649 * The API spec says as follows: 2650 * If the interface index is specified as 0, the 2651 * system may choose a multicast group membership to 2652 * drop by matching the multicast address only. 2653 * On the other hand, we cannot disambiguate the scope 2654 * zone unless an interface is provided. Thus, we 2655 * check if there's ambiguity with the default scope 2656 * zone as the last resort. 2657 */ 2658 bzero(&sa6_mc, sizeof(sa6_mc)); 2659 sa6_mc.sin6_family = AF_INET6; 2660 sa6_mc.sin6_len = sizeof(sa6_mc); 2661 sa6_mc.sin6_addr = mreq->ipv6mr_multiaddr; 2662 error = sa6_embedscope(&sa6_mc, ip6_use_defzone); 2663 if (error != 0) 2664 break; 2665 mreq->ipv6mr_multiaddr = sa6_mc.sin6_addr; 2666 } 2667 2668 /* 2669 * Find the membership in the membership list. 2670 */ 2671 for (imm = im6o->im6o_memberships.lh_first; 2672 imm != NULL; imm = imm->i6mm_chain.le_next) { 2673 if ((ifp == NULL || imm->i6mm_maddr->in6m_ifp == ifp) && 2674 IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr, 2675 &mreq->ipv6mr_multiaddr)) 2676 break; 2677 } 2678 if (imm == NULL) { 2679 /* Unable to resolve interface */ 2680 error = EADDRNOTAVAIL; 2681 break; 2682 } 2683 /* 2684 * Give up the multicast address record to which the 2685 * membership points. 2686 */ 2687 LIST_REMOVE(imm, i6mm_chain); 2688 in6_leavegroup(imm); 2689 break; 2690 2691 default: 2692 error = EOPNOTSUPP; 2693 break; 2694 } 2695 2696 /* 2697 * If all options have default values, no need to keep the mbuf. 2698 */ 2699 if (im6o->im6o_multicast_ifp == NULL && 2700 im6o->im6o_multicast_hlim == ip6_defmcasthlim && 2701 im6o->im6o_multicast_loop == IPV6_DEFAULT_MULTICAST_LOOP && 2702 im6o->im6o_memberships.lh_first == NULL) { 2703 free(*im6op, M_IPMOPTS); 2704 *im6op = NULL; 2705 } 2706 2707 return (error); 2708 } 2709 2710 /* 2711 * Return the IP6 multicast options in response to user getsockopt(). 2712 */ 2713 static int 2714 ip6_getmoptions(optname, im6o, mp) 2715 int optname; 2716 struct ip6_moptions *im6o; 2717 struct mbuf **mp; 2718 { 2719 u_int *hlim, *loop, *ifindex; 2720 2721 *mp = m_get(M_WAIT, MT_SOOPTS); 2722 2723 switch (optname) { 2724 2725 case IPV6_MULTICAST_IF: 2726 ifindex = mtod(*mp, u_int *); 2727 (*mp)->m_len = sizeof(u_int); 2728 if (im6o == NULL || im6o->im6o_multicast_ifp == NULL) 2729 *ifindex = 0; 2730 else 2731 *ifindex = im6o->im6o_multicast_ifp->if_index; 2732 return (0); 2733 2734 case IPV6_MULTICAST_HOPS: 2735 hlim = mtod(*mp, u_int *); 2736 (*mp)->m_len = sizeof(u_int); 2737 if (im6o == NULL) 2738 *hlim = ip6_defmcasthlim; 2739 else 2740 *hlim = im6o->im6o_multicast_hlim; 2741 return (0); 2742 2743 case IPV6_MULTICAST_LOOP: 2744 loop = mtod(*mp, u_int *); 2745 (*mp)->m_len = sizeof(u_int); 2746 if (im6o == NULL) 2747 *loop = ip6_defmcasthlim; 2748 else 2749 *loop = im6o->im6o_multicast_loop; 2750 return (0); 2751 2752 default: 2753 return (EOPNOTSUPP); 2754 } 2755 } 2756 2757 /* 2758 * Discard the IP6 multicast options. 2759 */ 2760 void 2761 ip6_freemoptions(im6o) 2762 struct ip6_moptions *im6o; 2763 { 2764 struct in6_multi_mship *imm; 2765 2766 if (im6o == NULL) 2767 return; 2768 2769 while ((imm = im6o->im6o_memberships.lh_first) != NULL) { 2770 LIST_REMOVE(imm, i6mm_chain); 2771 in6_leavegroup(imm); 2772 } 2773 free(im6o, M_IPMOPTS); 2774 } 2775 2776 /* 2777 * Set IPv6 outgoing packet options based on advanced API. 2778 */ 2779 int 2780 ip6_setpktopts(control, opt, stickyopt, priv, uproto) 2781 struct mbuf *control; 2782 struct ip6_pktopts *opt, *stickyopt; 2783 int priv, uproto; 2784 { 2785 struct cmsghdr *cm = 0; 2786 2787 if (control == NULL || opt == NULL) 2788 return (EINVAL); 2789 2790 ip6_initpktopts(opt); 2791 if (stickyopt) { 2792 int error; 2793 2794 /* 2795 * If stickyopt is provided, make a local copy of the options 2796 * for this particular packet, then override them by ancillary 2797 * objects. 2798 * XXX: copypktopts() does not copy the cached route to a next 2799 * hop (if any). This is not very good in terms of efficiency, 2800 * but we can allow this since this option should be rarely 2801 * used. 2802 */ 2803 if ((error = copypktopts(opt, stickyopt, M_NOWAIT)) != 0) 2804 return (error); 2805 } 2806 2807 /* 2808 * XXX: Currently, we assume all the optional information is stored 2809 * in a single mbuf. 2810 */ 2811 if (control->m_next) 2812 return (EINVAL); 2813 2814 for (; control->m_len; control->m_data += CMSG_ALIGN(cm->cmsg_len), 2815 control->m_len -= CMSG_ALIGN(cm->cmsg_len)) { 2816 int error; 2817 2818 if (control->m_len < CMSG_LEN(0)) 2819 return (EINVAL); 2820 2821 cm = mtod(control, struct cmsghdr *); 2822 if (cm->cmsg_len == 0 || cm->cmsg_len > control->m_len) 2823 return (EINVAL); 2824 if (cm->cmsg_level != IPPROTO_IPV6) 2825 continue; 2826 2827 error = ip6_setpktopt(cm->cmsg_type, CMSG_DATA(cm), 2828 cm->cmsg_len - CMSG_LEN(0), opt, priv, 0, 1, uproto); 2829 if (error) 2830 return (error); 2831 } 2832 2833 return (0); 2834 } 2835 2836 /* 2837 * Set a particular packet option, as a sticky option or an ancillary data 2838 * item. "len" can be 0 only when it's a sticky option. 2839 * We have 4 cases of combination of "sticky" and "cmsg": 2840 * "sticky=0, cmsg=0": impossible 2841 * "sticky=0, cmsg=1": RFC2292 or RFC3542 ancillary data 2842 * "sticky=1, cmsg=0": RFC3542 socket option 2843 * "sticky=1, cmsg=1": RFC2292 socket option 2844 */ 2845 static int 2846 ip6_setpktopt(int optname, u_char *buf, int len, struct ip6_pktopts *opt, 2847 int priv, int sticky, int cmsg, int uproto) 2848 { 2849 int minmtupolicy; 2850 2851 if (!sticky && !cmsg) { 2852 #ifdef DIAGNOSTIC 2853 printf("ip6_setpktopt: impossible case\n"); 2854 #endif 2855 return (EINVAL); 2856 } 2857 2858 /* 2859 * IPV6_2292xxx is for backward compatibility to RFC2292, and should 2860 * not be specified in the context of RFC3542. Conversely, 2861 * RFC3542 types should not be specified in the context of RFC2292. 2862 */ 2863 if (!cmsg) { 2864 switch (optname) { 2865 case IPV6_2292PKTINFO: 2866 case IPV6_2292HOPLIMIT: 2867 case IPV6_2292NEXTHOP: 2868 case IPV6_2292HOPOPTS: 2869 case IPV6_2292DSTOPTS: 2870 case IPV6_2292RTHDR: 2871 case IPV6_2292PKTOPTIONS: 2872 return (ENOPROTOOPT); 2873 } 2874 } 2875 if (sticky && cmsg) { 2876 switch (optname) { 2877 case IPV6_PKTINFO: 2878 case IPV6_HOPLIMIT: 2879 case IPV6_NEXTHOP: 2880 case IPV6_HOPOPTS: 2881 case IPV6_DSTOPTS: 2882 case IPV6_RTHDRDSTOPTS: 2883 case IPV6_RTHDR: 2884 case IPV6_USE_MIN_MTU: 2885 case IPV6_DONTFRAG: 2886 case IPV6_OTCLASS: 2887 case IPV6_TCLASS: 2888 return (ENOPROTOOPT); 2889 } 2890 } 2891 2892 switch (optname) { 2893 #ifdef RFC2292 2894 case IPV6_2292PKTINFO: 2895 #endif 2896 case IPV6_PKTINFO: 2897 { 2898 struct ifnet *ifp = NULL; 2899 struct in6_pktinfo *pktinfo; 2900 2901 if (len != sizeof(struct in6_pktinfo)) 2902 return (EINVAL); 2903 2904 pktinfo = (struct in6_pktinfo *)buf; 2905 2906 /* 2907 * An application can clear any sticky IPV6_PKTINFO option by 2908 * doing a "regular" setsockopt with ipi6_addr being 2909 * in6addr_any and ipi6_ifindex being zero. 2910 * [RFC 3542, Section 6] 2911 */ 2912 if (optname == IPV6_PKTINFO && opt->ip6po_pktinfo && 2913 pktinfo->ipi6_ifindex == 0 && 2914 IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) { 2915 ip6_clearpktopts(opt, optname); 2916 break; 2917 } 2918 2919 if (uproto == IPPROTO_TCP && optname == IPV6_PKTINFO && 2920 sticky && !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) { 2921 return (EINVAL); 2922 } 2923 2924 /* validate the interface index if specified. */ 2925 if (pktinfo->ipi6_ifindex >= if_indexlim || 2926 pktinfo->ipi6_ifindex < 0) { 2927 return (ENXIO); 2928 } 2929 if (pktinfo->ipi6_ifindex) { 2930 ifp = ifindex2ifnet[pktinfo->ipi6_ifindex]; 2931 if (ifp == NULL) 2932 return (ENXIO); 2933 } 2934 2935 /* 2936 * We store the address anyway, and let in6_selectsrc() 2937 * validate the specified address. This is because ipi6_addr 2938 * may not have enough information about its scope zone, and 2939 * we may need additional information (such as outgoing 2940 * interface or the scope zone of a destination address) to 2941 * disambiguate the scope. 2942 * XXX: the delay of the validation may confuse the 2943 * application when it is used as a sticky option. 2944 */ 2945 if (opt->ip6po_pktinfo == NULL) { 2946 opt->ip6po_pktinfo = malloc(sizeof(*pktinfo), 2947 M_IP6OPT, M_NOWAIT); 2948 if (opt->ip6po_pktinfo == NULL) 2949 return (ENOBUFS); 2950 } 2951 memcpy(opt->ip6po_pktinfo, pktinfo, sizeof(*pktinfo)); 2952 break; 2953 } 2954 2955 #ifdef RFC2292 2956 case IPV6_2292HOPLIMIT: 2957 #endif 2958 case IPV6_HOPLIMIT: 2959 { 2960 int *hlimp; 2961 2962 /* 2963 * RFC 3542 deprecated the usage of sticky IPV6_HOPLIMIT 2964 * to simplify the ordering among hoplimit options. 2965 */ 2966 if (optname == IPV6_HOPLIMIT && sticky) 2967 return (ENOPROTOOPT); 2968 2969 if (len != sizeof(int)) 2970 return (EINVAL); 2971 hlimp = (int *)buf; 2972 if (*hlimp < -1 || *hlimp > 255) 2973 return (EINVAL); 2974 2975 opt->ip6po_hlim = *hlimp; 2976 break; 2977 } 2978 2979 case IPV6_OTCLASS: 2980 if (len != sizeof(u_int8_t)) 2981 return (EINVAL); 2982 2983 opt->ip6po_tclass = *(u_int8_t *)buf; 2984 break; 2985 2986 case IPV6_TCLASS: 2987 { 2988 int tclass; 2989 2990 if (len != sizeof(int)) 2991 return (EINVAL); 2992 tclass = *(int *)buf; 2993 if (tclass < -1 || tclass > 255) 2994 return (EINVAL); 2995 2996 opt->ip6po_tclass = tclass; 2997 break; 2998 } 2999 3000 #ifdef RFC2292 3001 case IPV6_2292NEXTHOP: 3002 #endif 3003 case IPV6_NEXTHOP: 3004 if (!priv) 3005 return (EPERM); 3006 3007 if (len == 0) { /* just remove the option */ 3008 ip6_clearpktopts(opt, IPV6_NEXTHOP); 3009 break; 3010 } 3011 3012 /* check if cmsg_len is large enough for sa_len */ 3013 if (len < sizeof(struct sockaddr) || len < *buf) 3014 return (EINVAL); 3015 3016 switch (((struct sockaddr *)buf)->sa_family) { 3017 case AF_INET6: 3018 { 3019 struct sockaddr_in6 *sa6 = (struct sockaddr_in6 *)buf; 3020 int error; 3021 3022 if (sa6->sin6_len != sizeof(struct sockaddr_in6)) 3023 return (EINVAL); 3024 3025 if (IN6_IS_ADDR_UNSPECIFIED(&sa6->sin6_addr) || 3026 IN6_IS_ADDR_MULTICAST(&sa6->sin6_addr)) { 3027 return (EINVAL); 3028 } 3029 if ((error = sa6_embedscope(sa6, ip6_use_defzone)) 3030 != 0) { 3031 return (error); 3032 } 3033 break; 3034 } 3035 case AF_LINK: /* eventually be supported? */ 3036 default: 3037 return (EAFNOSUPPORT); 3038 } 3039 3040 /* turn off the previous option, then set the new option. */ 3041 ip6_clearpktopts(opt, IPV6_NEXTHOP); 3042 opt->ip6po_nexthop = malloc(*buf, M_IP6OPT, M_NOWAIT); 3043 if (opt->ip6po_nexthop == NULL) 3044 return (ENOBUFS); 3045 memcpy(opt->ip6po_nexthop, buf, *buf); 3046 break; 3047 3048 #ifdef RFC2292 3049 case IPV6_2292HOPOPTS: 3050 #endif 3051 case IPV6_HOPOPTS: 3052 { 3053 struct ip6_hbh *hbh; 3054 int hbhlen; 3055 3056 /* 3057 * XXX: We don't allow a non-privileged user to set ANY HbH 3058 * options, since per-option restriction has too much 3059 * overhead. 3060 */ 3061 if (!priv) 3062 return (EPERM); 3063 3064 if (len == 0) { 3065 ip6_clearpktopts(opt, IPV6_HOPOPTS); 3066 break; /* just remove the option */ 3067 } 3068 3069 /* message length validation */ 3070 if (len < sizeof(struct ip6_hbh)) 3071 return (EINVAL); 3072 hbh = (struct ip6_hbh *)buf; 3073 hbhlen = (hbh->ip6h_len + 1) << 3; 3074 if (len != hbhlen) 3075 return (EINVAL); 3076 3077 /* turn off the previous option, then set the new option. */ 3078 ip6_clearpktopts(opt, IPV6_HOPOPTS); 3079 opt->ip6po_hbh = malloc(hbhlen, M_IP6OPT, M_NOWAIT); 3080 if (opt->ip6po_hbh == NULL) 3081 return (ENOBUFS); 3082 memcpy(opt->ip6po_hbh, hbh, hbhlen); 3083 3084 break; 3085 } 3086 3087 #ifdef RFC2292 3088 case IPV6_2292DSTOPTS: 3089 #endif 3090 case IPV6_DSTOPTS: 3091 case IPV6_RTHDRDSTOPTS: 3092 { 3093 struct ip6_dest *dest, **newdest = NULL; 3094 int destlen; 3095 3096 if (!priv) /* XXX: see the comment for IPV6_HOPOPTS */ 3097 return (EPERM); 3098 3099 if (len == 0) { 3100 ip6_clearpktopts(opt, optname); 3101 break; /* just remove the option */ 3102 } 3103 3104 /* message length validation */ 3105 if (len < sizeof(struct ip6_dest)) 3106 return (EINVAL); 3107 dest = (struct ip6_dest *)buf; 3108 destlen = (dest->ip6d_len + 1) << 3; 3109 if (len != destlen) 3110 return (EINVAL); 3111 /* 3112 * Determine the position that the destination options header 3113 * should be inserted; before or after the routing header. 3114 */ 3115 switch (optname) { 3116 case IPV6_2292DSTOPTS: 3117 /* 3118 * The old advanced API is ambiguous on this point. 3119 * Our approach is to determine the position based 3120 * according to the existence of a routing header. 3121 * Note, however, that this depends on the order of the 3122 * extension headers in the ancillary data; the 1st 3123 * part of the destination options header must appear 3124 * before the routing header in the ancillary data, 3125 * too. 3126 * RFC3542 solved the ambiguity by introducing 3127 * separate ancillary data or option types. 3128 */ 3129 if (opt->ip6po_rthdr == NULL) 3130 newdest = &opt->ip6po_dest1; 3131 else 3132 newdest = &opt->ip6po_dest2; 3133 break; 3134 case IPV6_RTHDRDSTOPTS: 3135 newdest = &opt->ip6po_dest1; 3136 break; 3137 case IPV6_DSTOPTS: 3138 newdest = &opt->ip6po_dest2; 3139 break; 3140 } 3141 3142 /* turn off the previous option, then set the new option. */ 3143 ip6_clearpktopts(opt, optname); 3144 *newdest = malloc(destlen, M_IP6OPT, M_NOWAIT); 3145 if (*newdest == NULL) 3146 return (ENOBUFS); 3147 memcpy(*newdest, dest, destlen); 3148 3149 break; 3150 } 3151 3152 #ifdef RFC2292 3153 case IPV6_2292RTHDR: 3154 #endif 3155 case IPV6_RTHDR: 3156 { 3157 struct ip6_rthdr *rth; 3158 int rthlen; 3159 3160 if (len == 0) { 3161 ip6_clearpktopts(opt, IPV6_RTHDR); 3162 break; /* just remove the option */ 3163 } 3164 3165 /* message length validation */ 3166 if (len < sizeof(struct ip6_rthdr)) 3167 return (EINVAL); 3168 rth = (struct ip6_rthdr *)buf; 3169 rthlen = (rth->ip6r_len + 1) << 3; 3170 if (len != rthlen) 3171 return (EINVAL); 3172 switch (rth->ip6r_type) { 3173 case IPV6_RTHDR_TYPE_0: 3174 if (rth->ip6r_len == 0) /* must contain one addr */ 3175 return (EINVAL); 3176 if (rth->ip6r_len % 2) /* length must be even */ 3177 return (EINVAL); 3178 if (rth->ip6r_len / 2 != rth->ip6r_segleft) 3179 return (EINVAL); 3180 break; 3181 default: 3182 return (EINVAL); /* not supported */ 3183 } 3184 /* turn off the previous option */ 3185 ip6_clearpktopts(opt, IPV6_RTHDR); 3186 opt->ip6po_rthdr = malloc(rthlen, M_IP6OPT, M_NOWAIT); 3187 if (opt->ip6po_rthdr == NULL) 3188 return (ENOBUFS); 3189 memcpy(opt->ip6po_rthdr, rth, rthlen); 3190 break; 3191 } 3192 3193 case IPV6_USE_MIN_MTU: 3194 if (len != sizeof(int)) 3195 return (EINVAL); 3196 minmtupolicy = *(int *)buf; 3197 if (minmtupolicy != IP6PO_MINMTU_MCASTONLY && 3198 minmtupolicy != IP6PO_MINMTU_DISABLE && 3199 minmtupolicy != IP6PO_MINMTU_ALL) { 3200 return (EINVAL); 3201 } 3202 opt->ip6po_minmtu = minmtupolicy; 3203 break; 3204 3205 case IPV6_DONTFRAG: 3206 if (len != sizeof(int)) 3207 return (EINVAL); 3208 3209 if (uproto == IPPROTO_TCP || *(int *)buf == 0) { 3210 /* 3211 * we ignore this option for TCP sockets. 3212 * (RFC3542 leaves this case unspecified.) 3213 */ 3214 opt->ip6po_flags &= ~IP6PO_DONTFRAG; 3215 } else 3216 opt->ip6po_flags |= IP6PO_DONTFRAG; 3217 break; 3218 3219 default: 3220 return (ENOPROTOOPT); 3221 } /* end of switch */ 3222 3223 return (0); 3224 } 3225 3226 /* 3227 * Routine called from ip6_output() to loop back a copy of an IP6 multicast 3228 * packet to the input queue of a specified interface. Note that this 3229 * calls the output routine of the loopback "driver", but with an interface 3230 * pointer that might NOT be lo0ifp -- easier than replicating that code here. 3231 */ 3232 void 3233 ip6_mloopback(ifp, m, dst) 3234 struct ifnet *ifp; 3235 struct mbuf *m; 3236 struct sockaddr_in6 *dst; 3237 { 3238 struct mbuf *copym; 3239 struct ip6_hdr *ip6; 3240 3241 copym = m_copy(m, 0, M_COPYALL); 3242 if (copym == NULL) 3243 return; 3244 3245 /* 3246 * Make sure to deep-copy IPv6 header portion in case the data 3247 * is in an mbuf cluster, so that we can safely override the IPv6 3248 * header portion later. 3249 */ 3250 if ((copym->m_flags & M_EXT) != 0 || 3251 copym->m_len < sizeof(struct ip6_hdr)) { 3252 copym = m_pullup(copym, sizeof(struct ip6_hdr)); 3253 if (copym == NULL) 3254 return; 3255 } 3256 3257 #ifdef DIAGNOSTIC 3258 if (copym->m_len < sizeof(*ip6)) { 3259 m_freem(copym); 3260 return; 3261 } 3262 #endif 3263 3264 ip6 = mtod(copym, struct ip6_hdr *); 3265 /* 3266 * clear embedded scope identifiers if necessary. 3267 * in6_clearscope will touch the addresses only when necessary. 3268 */ 3269 in6_clearscope(&ip6->ip6_src); 3270 in6_clearscope(&ip6->ip6_dst); 3271 3272 (void)looutput(ifp, copym, (struct sockaddr *)dst, NULL); 3273 } 3274 3275 /* 3276 * Chop IPv6 header off from the payload. 3277 */ 3278 static int 3279 ip6_splithdr(m, exthdrs) 3280 struct mbuf *m; 3281 struct ip6_exthdrs *exthdrs; 3282 { 3283 struct mbuf *mh; 3284 struct ip6_hdr *ip6; 3285 3286 ip6 = mtod(m, struct ip6_hdr *); 3287 if (m->m_len > sizeof(*ip6)) { 3288 MGETHDR(mh, M_DONTWAIT, MT_HEADER); 3289 if (mh == 0) { 3290 m_freem(m); 3291 return ENOBUFS; 3292 } 3293 M_MOVE_PKTHDR(mh, m); 3294 MH_ALIGN(mh, sizeof(*ip6)); 3295 m->m_len -= sizeof(*ip6); 3296 m->m_data += sizeof(*ip6); 3297 mh->m_next = m; 3298 m = mh; 3299 m->m_len = sizeof(*ip6); 3300 bcopy((caddr_t)ip6, mtod(m, caddr_t), sizeof(*ip6)); 3301 } 3302 exthdrs->ip6e_ip6 = m; 3303 return 0; 3304 } 3305 3306 /* 3307 * Compute IPv6 extension header length. 3308 */ 3309 int 3310 ip6_optlen(in6p) 3311 struct in6pcb *in6p; 3312 { 3313 int len; 3314 3315 if (!in6p->in6p_outputopts) 3316 return 0; 3317 3318 len = 0; 3319 #define elen(x) \ 3320 (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0) 3321 3322 len += elen(in6p->in6p_outputopts->ip6po_hbh); 3323 len += elen(in6p->in6p_outputopts->ip6po_dest1); 3324 len += elen(in6p->in6p_outputopts->ip6po_rthdr); 3325 len += elen(in6p->in6p_outputopts->ip6po_dest2); 3326 return len; 3327 #undef elen 3328 } 3329
Indexes created Mon Sep 22 21:09:42 GMT 2025