ip6_output.c revision 1.94.4.1
1 /* $NetBSD: ip6_output.c,v 1.94.4.1 2006/04/22 11:40:12 simonb 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 1.94.4.1 2006/04/22 11:40:12 simonb Exp $"); 66 67 #include "opt_inet.h" 68 #include "opt_ipsec.h" 69 #include "opt_pfil_hooks.h" 70 71 #include <sys/param.h> 72 #include <sys/malloc.h> 73 #include <sys/mbuf.h> 74 #include <sys/errno.h> 75 #include <sys/protosw.h> 76 #include <sys/socket.h> 77 #include <sys/socketvar.h> 78 #include <sys/systm.h> 79 #include <sys/proc.h> 80 81 #include <net/if.h> 82 #include <net/route.h> 83 #ifdef PFIL_HOOKS 84 #include <net/pfil.h> 85 #endif 86 87 #include <netinet/in.h> 88 #include <netinet/in_var.h> 89 #include <netinet/ip6.h> 90 #include <netinet/icmp6.h> 91 #include <netinet/in_offload.h> 92 #include <netinet6/ip6_var.h> 93 #include <netinet6/in6_pcb.h> 94 #include <netinet6/nd6.h> 95 #include <netinet6/ip6protosw.h> 96 #include <netinet6/scope6_var.h> 97 98 #ifdef IPSEC 99 #include <netinet6/ipsec.h> 100 #include <netkey/key.h> 101 #endif /* IPSEC */ 102 103 #include <net/net_osdep.h> 104 105 #ifdef PFIL_HOOKS 106 extern struct pfil_head inet6_pfil_hook; /* XXX */ 107 #endif 108 109 struct ip6_exthdrs { 110 struct mbuf *ip6e_ip6; 111 struct mbuf *ip6e_hbh; 112 struct mbuf *ip6e_dest1; 113 struct mbuf *ip6e_rthdr; 114 struct mbuf *ip6e_dest2; 115 }; 116 117 static int ip6_pcbopts __P((struct ip6_pktopts **, struct mbuf *, 118 struct socket *)); 119 static int ip6_setmoptions __P((int, struct ip6_moptions **, struct mbuf *)); 120 static int ip6_getmoptions __P((int, struct ip6_moptions *, struct mbuf **)); 121 static int ip6_copyexthdr __P((struct mbuf **, caddr_t, int)); 122 static int ip6_insertfraghdr __P((struct mbuf *, struct mbuf *, int, 123 struct ip6_frag **)); 124 static int ip6_insert_jumboopt __P((struct ip6_exthdrs *, u_int32_t)); 125 static int ip6_splithdr __P((struct mbuf *, struct ip6_exthdrs *)); 126 static int ip6_getpmtu __P((struct route_in6 *, struct route_in6 *, 127 struct ifnet *, struct in6_addr *, u_long *, int *)); 128 129 #define IN6_NEED_CHECKSUM(ifp, csum_flags) \ 130 (__predict_true(((ifp)->if_flags & IFF_LOOPBACK) == 0 || \ 131 (((csum_flags) & M_CSUM_UDPv6) != 0 && udp_do_loopback_cksum) || \ 132 (((csum_flags) & M_CSUM_TCPv6) != 0 && tcp_do_loopback_cksum))) 133 134 /* 135 * IP6 output. The packet in mbuf chain m contains a skeletal IP6 136 * header (with pri, len, nxt, hlim, src, dst). 137 * This function may modify ver and hlim only. 138 * The mbuf chain containing the packet will be freed. 139 * The mbuf opt, if present, will not be freed. 140 * 141 * type of "mtu": rt_rmx.rmx_mtu is u_long, ifnet.ifr_mtu is int, and 142 * nd_ifinfo.linkmtu is u_int32_t. so we use u_long to hold largest one, 143 * which is rt_rmx.rmx_mtu. 144 */ 145 int 146 ip6_output(m0, opt, ro, flags, im6o, so, ifpp) 147 struct mbuf *m0; 148 struct ip6_pktopts *opt; 149 struct route_in6 *ro; 150 int flags; 151 struct ip6_moptions *im6o; 152 struct socket *so; 153 struct ifnet **ifpp; /* XXX: just for statistics */ 154 { 155 struct ip6_hdr *ip6, *mhip6; 156 struct ifnet *ifp, *origifp; 157 struct mbuf *m = m0; 158 int hlen, tlen, len, off; 159 struct route_in6 ip6route; 160 struct rtentry *rt = NULL; 161 struct sockaddr_in6 *dst, src_sa, dst_sa; 162 int error = 0; 163 struct in6_ifaddr *ia = NULL; 164 u_long mtu; 165 int alwaysfrag, dontfrag; 166 u_int32_t optlen = 0, plen = 0, unfragpartlen = 0; 167 struct ip6_exthdrs exthdrs; 168 struct in6_addr finaldst, src0, dst0; 169 u_int32_t zone; 170 struct route_in6 *ro_pmtu = NULL; 171 int hdrsplit = 0; 172 int needipsec = 0; 173 #ifdef IPSEC 174 int needipsectun = 0; 175 struct secpolicy *sp = NULL; 176 177 ip6 = mtod(m, struct ip6_hdr *); 178 #endif /* IPSEC */ 179 180 M_CSUM_DATA_IPv6_HL_SET(m->m_pkthdr.csum_data, sizeof(struct ip6_hdr)); 181 182 #define MAKE_EXTHDR(hp, mp) \ 183 do { \ 184 if (hp) { \ 185 struct ip6_ext *eh = (struct ip6_ext *)(hp); \ 186 error = ip6_copyexthdr((mp), (caddr_t)(hp), \ 187 ((eh)->ip6e_len + 1) << 3); \ 188 if (error) \ 189 goto freehdrs; \ 190 } \ 191 } while (/*CONSTCOND*/ 0) 192 193 bzero(&exthdrs, sizeof(exthdrs)); 194 if (opt) { 195 /* Hop-by-Hop options header */ 196 MAKE_EXTHDR(opt->ip6po_hbh, &exthdrs.ip6e_hbh); 197 /* Destination options header(1st part) */ 198 MAKE_EXTHDR(opt->ip6po_dest1, &exthdrs.ip6e_dest1); 199 /* Routing header */ 200 MAKE_EXTHDR(opt->ip6po_rthdr, &exthdrs.ip6e_rthdr); 201 /* Destination options header(2nd part) */ 202 MAKE_EXTHDR(opt->ip6po_dest2, &exthdrs.ip6e_dest2); 203 } 204 205 #ifdef IPSEC 206 if ((flags & IPV6_FORWARDING) != 0) { 207 needipsec = 0; 208 goto skippolicycheck; 209 } 210 211 /* get a security policy for this packet */ 212 if (so == NULL) 213 sp = ipsec6_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND, 0, &error); 214 else { 215 if (IPSEC_PCB_SKIP_IPSEC(sotoinpcb_hdr(so)->inph_sp, 216 IPSEC_DIR_OUTBOUND)) { 217 needipsec = 0; 218 goto skippolicycheck; 219 } 220 sp = ipsec6_getpolicybysock(m, IPSEC_DIR_OUTBOUND, so, &error); 221 } 222 223 if (sp == NULL) { 224 ipsec6stat.out_inval++; 225 goto freehdrs; 226 } 227 228 error = 0; 229 230 /* check policy */ 231 switch (sp->policy) { 232 case IPSEC_POLICY_DISCARD: 233 /* 234 * This packet is just discarded. 235 */ 236 ipsec6stat.out_polvio++; 237 goto freehdrs; 238 239 case IPSEC_POLICY_BYPASS: 240 case IPSEC_POLICY_NONE: 241 /* no need to do IPsec. */ 242 needipsec = 0; 243 break; 244 245 case IPSEC_POLICY_IPSEC: 246 if (sp->req == NULL) { 247 /* XXX should be panic ? */ 248 printf("ip6_output: No IPsec request specified.\n"); 249 error = EINVAL; 250 goto freehdrs; 251 } 252 needipsec = 1; 253 break; 254 255 case IPSEC_POLICY_ENTRUST: 256 default: 257 printf("ip6_output: Invalid policy found. %d\n", sp->policy); 258 } 259 260 skippolicycheck:; 261 #endif /* IPSEC */ 262 263 if (needipsec && 264 (m->m_pkthdr.csum_flags & (M_CSUM_UDPv6|M_CSUM_TCPv6)) != 0) { 265 in6_delayed_cksum(m); 266 m->m_pkthdr.csum_flags &= ~(M_CSUM_UDPv6|M_CSUM_TCPv6); 267 } 268 269 /* 270 * Calculate the total length of the extension header chain. 271 * Keep the length of the unfragmentable part for fragmentation. 272 */ 273 optlen = 0; 274 if (exthdrs.ip6e_hbh) optlen += exthdrs.ip6e_hbh->m_len; 275 if (exthdrs.ip6e_dest1) optlen += exthdrs.ip6e_dest1->m_len; 276 if (exthdrs.ip6e_rthdr) optlen += exthdrs.ip6e_rthdr->m_len; 277 unfragpartlen = optlen + sizeof(struct ip6_hdr); 278 /* NOTE: we don't add AH/ESP length here. do that later. */ 279 if (exthdrs.ip6e_dest2) optlen += exthdrs.ip6e_dest2->m_len; 280 281 /* 282 * If we need IPsec, or there is at least one extension header, 283 * separate IP6 header from the payload. 284 */ 285 if ((needipsec || optlen) && !hdrsplit) { 286 if ((error = ip6_splithdr(m, &exthdrs)) != 0) { 287 m = NULL; 288 goto freehdrs; 289 } 290 m = exthdrs.ip6e_ip6; 291 hdrsplit++; 292 } 293 294 /* adjust pointer */ 295 ip6 = mtod(m, struct ip6_hdr *); 296 297 /* adjust mbuf packet header length */ 298 m->m_pkthdr.len += optlen; 299 plen = m->m_pkthdr.len - sizeof(*ip6); 300 301 /* If this is a jumbo payload, insert a jumbo payload option. */ 302 if (plen > IPV6_MAXPACKET) { 303 if (!hdrsplit) { 304 if ((error = ip6_splithdr(m, &exthdrs)) != 0) { 305 m = NULL; 306 goto freehdrs; 307 } 308 m = exthdrs.ip6e_ip6; 309 hdrsplit++; 310 } 311 /* adjust pointer */ 312 ip6 = mtod(m, struct ip6_hdr *); 313 if ((error = ip6_insert_jumboopt(&exthdrs, plen)) != 0) 314 goto freehdrs; 315 optlen += 8; /* XXX JUMBOOPTLEN */ 316 ip6->ip6_plen = 0; 317 } else 318 ip6->ip6_plen = htons(plen); 319 320 /* 321 * Concatenate headers and fill in next header fields. 322 * Here we have, on "m" 323 * IPv6 payload 324 * and we insert headers accordingly. Finally, we should be getting: 325 * IPv6 hbh dest1 rthdr ah* [esp* dest2 payload] 326 * 327 * during the header composing process, "m" points to IPv6 header. 328 * "mprev" points to an extension header prior to esp. 329 */ 330 { 331 u_char *nexthdrp = &ip6->ip6_nxt; 332 struct mbuf *mprev = m; 333 334 /* 335 * we treat dest2 specially. this makes IPsec processing 336 * much easier. the goal here is to make mprev point the 337 * mbuf prior to dest2. 338 * 339 * result: IPv6 dest2 payload 340 * m and mprev will point to IPv6 header. 341 */ 342 if (exthdrs.ip6e_dest2) { 343 if (!hdrsplit) 344 panic("assumption failed: hdr not split"); 345 exthdrs.ip6e_dest2->m_next = m->m_next; 346 m->m_next = exthdrs.ip6e_dest2; 347 *mtod(exthdrs.ip6e_dest2, u_char *) = ip6->ip6_nxt; 348 ip6->ip6_nxt = IPPROTO_DSTOPTS; 349 } 350 351 #define MAKE_CHAIN(m, mp, p, i)\ 352 do {\ 353 if (m) {\ 354 if (!hdrsplit) \ 355 panic("assumption failed: hdr not split"); \ 356 *mtod((m), u_char *) = *(p);\ 357 *(p) = (i);\ 358 p = mtod((m), u_char *);\ 359 (m)->m_next = (mp)->m_next;\ 360 (mp)->m_next = (m);\ 361 (mp) = (m);\ 362 }\ 363 } while (/*CONSTCOND*/ 0) 364 /* 365 * result: IPv6 hbh dest1 rthdr dest2 payload 366 * m will point to IPv6 header. mprev will point to the 367 * extension header prior to dest2 (rthdr in the above case). 368 */ 369 MAKE_CHAIN(exthdrs.ip6e_hbh, mprev, nexthdrp, IPPROTO_HOPOPTS); 370 MAKE_CHAIN(exthdrs.ip6e_dest1, mprev, nexthdrp, 371 IPPROTO_DSTOPTS); 372 MAKE_CHAIN(exthdrs.ip6e_rthdr, mprev, nexthdrp, 373 IPPROTO_ROUTING); 374 375 M_CSUM_DATA_IPv6_HL_SET(m->m_pkthdr.csum_data, 376 sizeof(struct ip6_hdr) + optlen); 377 378 #ifdef IPSEC 379 if (!needipsec) 380 goto skip_ipsec2; 381 382 /* 383 * pointers after IPsec headers are not valid any more. 384 * other pointers need a great care too. 385 * (IPsec routines should not mangle mbufs prior to AH/ESP) 386 */ 387 exthdrs.ip6e_dest2 = NULL; 388 389 { 390 struct ip6_rthdr *rh = NULL; 391 int segleft_org = 0; 392 struct ipsec_output_state state; 393 394 if (exthdrs.ip6e_rthdr) { 395 rh = mtod(exthdrs.ip6e_rthdr, struct ip6_rthdr *); 396 segleft_org = rh->ip6r_segleft; 397 rh->ip6r_segleft = 0; 398 } 399 400 bzero(&state, sizeof(state)); 401 state.m = m; 402 error = ipsec6_output_trans(&state, nexthdrp, mprev, sp, flags, 403 &needipsectun); 404 m = state.m; 405 if (error) { 406 /* mbuf is already reclaimed in ipsec6_output_trans. */ 407 m = NULL; 408 switch (error) { 409 case EHOSTUNREACH: 410 case ENETUNREACH: 411 case EMSGSIZE: 412 case ENOBUFS: 413 case ENOMEM: 414 break; 415 default: 416 printf("ip6_output (ipsec): error code %d\n", error); 417 /* FALLTHROUGH */ 418 case ENOENT: 419 /* don't show these error codes to the user */ 420 error = 0; 421 break; 422 } 423 goto bad; 424 } 425 if (exthdrs.ip6e_rthdr) { 426 /* ah6_output doesn't modify mbuf chain */ 427 rh->ip6r_segleft = segleft_org; 428 } 429 } 430 skip_ipsec2:; 431 #endif 432 } 433 434 /* 435 * If there is a routing header, replace destination address field 436 * with the first hop of the routing header. 437 */ 438 if (exthdrs.ip6e_rthdr) { 439 struct ip6_rthdr *rh; 440 struct ip6_rthdr0 *rh0; 441 struct in6_addr *addr; 442 struct sockaddr_in6 sa; 443 444 rh = (struct ip6_rthdr *)(mtod(exthdrs.ip6e_rthdr, 445 struct ip6_rthdr *)); 446 finaldst = ip6->ip6_dst; 447 switch (rh->ip6r_type) { 448 case IPV6_RTHDR_TYPE_0: 449 rh0 = (struct ip6_rthdr0 *)rh; 450 addr = (struct in6_addr *)(rh0 + 1); 451 452 /* 453 * construct a sockaddr_in6 form of 454 * the first hop. 455 * 456 * XXX: we may not have enough 457 * information about its scope zone; 458 * there is no standard API to pass 459 * the information from the 460 * application. 461 */ 462 bzero(&sa, sizeof(sa)); 463 sa.sin6_family = AF_INET6; 464 sa.sin6_len = sizeof(sa); 465 sa.sin6_addr = addr[0]; 466 if ((error = sa6_embedscope(&sa, 467 ip6_use_defzone)) != 0) { 468 goto bad; 469 } 470 ip6->ip6_dst = sa.sin6_addr; 471 (void)memmove(&addr[0], &addr[1], 472 sizeof(struct in6_addr) * 473 (rh0->ip6r0_segleft - 1)); 474 addr[rh0->ip6r0_segleft - 1] = finaldst; 475 /* XXX */ 476 in6_clearscope(addr + rh0->ip6r0_segleft - 1); 477 break; 478 default: /* is it possible? */ 479 error = EINVAL; 480 goto bad; 481 } 482 } 483 484 /* Source address validation */ 485 if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) && 486 (flags & IPV6_UNSPECSRC) == 0) { 487 error = EOPNOTSUPP; 488 ip6stat.ip6s_badscope++; 489 goto bad; 490 } 491 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) { 492 error = EOPNOTSUPP; 493 ip6stat.ip6s_badscope++; 494 goto bad; 495 } 496 497 ip6stat.ip6s_localout++; 498 499 /* 500 * Route packet. 501 */ 502 /* initialize cached route */ 503 if (ro == 0) { 504 ro = &ip6route; 505 bzero((caddr_t)ro, sizeof(*ro)); 506 } 507 ro_pmtu = ro; 508 if (opt && opt->ip6po_rthdr) 509 ro = &opt->ip6po_route; 510 dst = (struct sockaddr_in6 *)&ro->ro_dst; 511 512 #ifdef notyet /* this will be available with the RFC3542 support */ 513 /* 514 * if specified, try to fill in the traffic class field. 515 * do not override if a non-zero value is already set. 516 * we check the diffserv field and the ecn field separately. 517 */ 518 if (opt && opt->ip6po_tclass >= 0) { 519 int mask = 0; 520 521 if ((ip6->ip6_flow & htonl(0xfc << 20)) == 0) 522 mask |= 0xfc; 523 if ((ip6->ip6_flow & htonl(0x03 << 20)) == 0) 524 mask |= 0x03; 525 if (mask != 0) 526 ip6->ip6_flow |= htonl((opt->ip6po_tclass & mask) << 20); 527 } 528 #endif 529 530 /* fill in or override the hop limit field, if necessary. */ 531 if (opt && opt->ip6po_hlim != -1) 532 ip6->ip6_hlim = opt->ip6po_hlim & 0xff; 533 else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) { 534 if (im6o != NULL) 535 ip6->ip6_hlim = im6o->im6o_multicast_hlim; 536 else 537 ip6->ip6_hlim = ip6_defmcasthlim; 538 } 539 540 #ifdef IPSEC 541 if (needipsec && needipsectun) { 542 struct ipsec_output_state state; 543 544 /* 545 * All the extension headers will become inaccessible 546 * (since they can be encrypted). 547 * Don't panic, we need no more updates to extension headers 548 * on inner IPv6 packet (since they are now encapsulated). 549 * 550 * IPv6 [ESP|AH] IPv6 [extension headers] payload 551 */ 552 bzero(&exthdrs, sizeof(exthdrs)); 553 exthdrs.ip6e_ip6 = m; 554 555 bzero(&state, sizeof(state)); 556 state.m = m; 557 state.ro = (struct route *)ro; 558 state.dst = (struct sockaddr *)dst; 559 560 error = ipsec6_output_tunnel(&state, sp, flags); 561 562 m = state.m; 563 ro_pmtu = ro = (struct route_in6 *)state.ro; 564 dst = (struct sockaddr_in6 *)state.dst; 565 if (error) { 566 /* mbuf is already reclaimed in ipsec6_output_tunnel. */ 567 m0 = m = NULL; 568 m = NULL; 569 switch (error) { 570 case EHOSTUNREACH: 571 case ENETUNREACH: 572 case EMSGSIZE: 573 case ENOBUFS: 574 case ENOMEM: 575 break; 576 default: 577 printf("ip6_output (ipsec): error code %d\n", error); 578 /* FALLTHROUGH */ 579 case ENOENT: 580 /* don't show these error codes to the user */ 581 error = 0; 582 break; 583 } 584 goto bad; 585 } 586 587 exthdrs.ip6e_ip6 = m; 588 } 589 #endif /* IPSEC */ 590 591 /* adjust pointer */ 592 ip6 = mtod(m, struct ip6_hdr *); 593 594 bzero(&dst_sa, sizeof(dst_sa)); 595 dst_sa.sin6_family = AF_INET6; 596 dst_sa.sin6_len = sizeof(dst_sa); 597 dst_sa.sin6_addr = ip6->ip6_dst; 598 if ((error = in6_selectroute(&dst_sa, opt, im6o, ro, &ifp, &rt, 0)) 599 != 0) { 600 switch (error) { 601 case EHOSTUNREACH: 602 ip6stat.ip6s_noroute++; 603 break; 604 case EADDRNOTAVAIL: 605 default: 606 break; /* XXX statistics? */ 607 } 608 if (ifp != NULL) 609 in6_ifstat_inc(ifp, ifs6_out_discard); 610 goto bad; 611 } 612 if (rt == NULL) { 613 /* 614 * If in6_selectroute() does not return a route entry, 615 * dst may not have been updated. 616 */ 617 *dst = dst_sa; /* XXX */ 618 } 619 620 /* 621 * then rt (for unicast) and ifp must be non-NULL valid values. 622 */ 623 if ((flags & IPV6_FORWARDING) == 0) { 624 /* XXX: the FORWARDING flag can be set for mrouting. */ 625 in6_ifstat_inc(ifp, ifs6_out_request); 626 } 627 if (rt != NULL) { 628 ia = (struct in6_ifaddr *)(rt->rt_ifa); 629 rt->rt_use++; 630 } 631 632 /* 633 * The outgoing interface must be in the zone of source and 634 * destination addresses. We should use ia_ifp to support the 635 * case of sending packets to an address of our own. 636 */ 637 if (ia != NULL && ia->ia_ifp) 638 origifp = ia->ia_ifp; 639 else 640 origifp = ifp; 641 642 src0 = ip6->ip6_src; 643 if (in6_setscope(&src0, origifp, &zone)) 644 goto badscope; 645 bzero(&src_sa, sizeof(src_sa)); 646 src_sa.sin6_family = AF_INET6; 647 src_sa.sin6_len = sizeof(src_sa); 648 src_sa.sin6_addr = ip6->ip6_src; 649 if (sa6_recoverscope(&src_sa) || zone != src_sa.sin6_scope_id) 650 goto badscope; 651 652 dst0 = ip6->ip6_dst; 653 if (in6_setscope(&dst0, origifp, &zone)) 654 goto badscope; 655 /* re-initialize to be sure */ 656 bzero(&dst_sa, sizeof(dst_sa)); 657 dst_sa.sin6_family = AF_INET6; 658 dst_sa.sin6_len = sizeof(dst_sa); 659 dst_sa.sin6_addr = ip6->ip6_dst; 660 if (sa6_recoverscope(&dst_sa) || zone != dst_sa.sin6_scope_id) 661 goto badscope; 662 663 /* scope check is done. */ 664 goto routefound; 665 666 badscope: 667 ip6stat.ip6s_badscope++; 668 in6_ifstat_inc(origifp, ifs6_out_discard); 669 if (error == 0) 670 error = EHOSTUNREACH; /* XXX */ 671 goto bad; 672 673 routefound: 674 if (rt && !IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) { 675 #ifdef notyet /* this will be available with the RFC3542 support */ 676 if (opt && opt->ip6po_nextroute.ro_rt) { 677 /* 678 * The nexthop is explicitly specified by the 679 * application. We assume the next hop is an IPv6 680 * address. 681 */ 682 dst = (struct sockaddr_in6 *)opt->ip6po_nexthop; 683 } else 684 #endif 685 if ((rt->rt_flags & RTF_GATEWAY)) 686 dst = (struct sockaddr_in6 *)rt->rt_gateway; 687 } 688 689 /* 690 * XXXXXX: original code follows: 691 */ 692 if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) 693 m->m_flags &= ~(M_BCAST | M_MCAST); /* just in case */ 694 else { 695 struct in6_multi *in6m; 696 697 m->m_flags = (m->m_flags & ~M_BCAST) | M_MCAST; 698 699 in6_ifstat_inc(ifp, ifs6_out_mcast); 700 701 /* 702 * Confirm that the outgoing interface supports multicast. 703 */ 704 if (!(ifp->if_flags & IFF_MULTICAST)) { 705 ip6stat.ip6s_noroute++; 706 in6_ifstat_inc(ifp, ifs6_out_discard); 707 error = ENETUNREACH; 708 goto bad; 709 } 710 711 IN6_LOOKUP_MULTI(ip6->ip6_dst, ifp, in6m); 712 if (in6m != NULL && 713 (im6o == NULL || im6o->im6o_multicast_loop)) { 714 /* 715 * If we belong to the destination multicast group 716 * on the outgoing interface, and the caller did not 717 * forbid loopback, loop back a copy. 718 */ 719 ip6_mloopback(ifp, m, dst); 720 } else { 721 /* 722 * If we are acting as a multicast router, perform 723 * multicast forwarding as if the packet had just 724 * arrived on the interface to which we are about 725 * to send. The multicast forwarding function 726 * recursively calls this function, using the 727 * IPV6_FORWARDING flag to prevent infinite recursion. 728 * 729 * Multicasts that are looped back by ip6_mloopback(), 730 * above, will be forwarded by the ip6_input() routine, 731 * if necessary. 732 */ 733 if (ip6_mrouter && (flags & IPV6_FORWARDING) == 0) { 734 if (ip6_mforward(ip6, ifp, m) != 0) { 735 m_freem(m); 736 goto done; 737 } 738 } 739 } 740 /* 741 * Multicasts with a hoplimit of zero may be looped back, 742 * above, but must not be transmitted on a network. 743 * Also, multicasts addressed to the loopback interface 744 * are not sent -- the above call to ip6_mloopback() will 745 * loop back a copy if this host actually belongs to the 746 * destination group on the loopback interface. 747 */ 748 if (ip6->ip6_hlim == 0 || (ifp->if_flags & IFF_LOOPBACK) || 749 IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst)) { 750 m_freem(m); 751 goto done; 752 } 753 } 754 755 /* 756 * Fill the outgoing inteface to tell the upper layer 757 * to increment per-interface statistics. 758 */ 759 if (ifpp) 760 *ifpp = ifp; 761 762 /* Determine path MTU. */ 763 if ((error = ip6_getpmtu(ro_pmtu, ro, ifp, &finaldst, &mtu, 764 &alwaysfrag)) != 0) 765 goto bad; 766 #ifdef IPSEC 767 if (needipsectun) 768 mtu = IPV6_MMTU; 769 #endif 770 771 /* 772 * The caller of this function may specify to use the minimum MTU 773 * in some cases. 774 */ 775 if (mtu > IPV6_MMTU) { 776 if ((flags & IPV6_MINMTU)) 777 mtu = IPV6_MMTU; 778 } 779 780 /* 781 * clear embedded scope identifiers if necessary. 782 * in6_clearscope will touch the addresses only when necessary. 783 */ 784 in6_clearscope(&ip6->ip6_src); 785 in6_clearscope(&ip6->ip6_dst); 786 787 /* 788 * If the outgoing packet contains a hop-by-hop options header, 789 * it must be examined and processed even by the source node. 790 * (RFC 2460, section 4.) 791 */ 792 if (exthdrs.ip6e_hbh) { 793 struct ip6_hbh *hbh = mtod(exthdrs.ip6e_hbh, struct ip6_hbh *); 794 u_int32_t dummy1; /* XXX unused */ 795 u_int32_t dummy2; /* XXX unused */ 796 797 /* 798 * XXX: if we have to send an ICMPv6 error to the sender, 799 * we need the M_LOOP flag since icmp6_error() expects 800 * the IPv6 and the hop-by-hop options header are 801 * continuous unless the flag is set. 802 */ 803 m->m_flags |= M_LOOP; 804 m->m_pkthdr.rcvif = ifp; 805 if (ip6_process_hopopts(m, (u_int8_t *)(hbh + 1), 806 ((hbh->ip6h_len + 1) << 3) - sizeof(struct ip6_hbh), 807 &dummy1, &dummy2) < 0) { 808 /* m was already freed at this point */ 809 error = EINVAL;/* better error? */ 810 goto done; 811 } 812 m->m_flags &= ~M_LOOP; /* XXX */ 813 m->m_pkthdr.rcvif = NULL; 814 } 815 816 #ifdef PFIL_HOOKS 817 /* 818 * Run through list of hooks for output packets. 819 */ 820 if ((error = pfil_run_hooks(&inet6_pfil_hook, &m, ifp, PFIL_OUT)) != 0) 821 goto done; 822 if (m == NULL) 823 goto done; 824 ip6 = mtod(m, struct ip6_hdr *); 825 #endif /* PFIL_HOOKS */ 826 /* 827 * Send the packet to the outgoing interface. 828 * If necessary, do IPv6 fragmentation before sending. 829 * 830 * the logic here is rather complex: 831 * 1: normal case (dontfrag == 0, alwaysfrag == 0) 832 * 1-a: send as is if tlen <= path mtu 833 * 1-b: fragment if tlen > path mtu 834 * 835 * 2: if user asks us not to fragment (dontfrag == 1) 836 * 2-a: send as is if tlen <= interface mtu 837 * 2-b: error if tlen > interface mtu 838 * 839 * 3: if we always need to attach fragment header (alwaysfrag == 1) 840 * always fragment 841 * 842 * 4: if dontfrag == 1 && alwaysfrag == 1 843 * error, as we cannot handle this conflicting request 844 */ 845 tlen = m->m_pkthdr.len; 846 847 dontfrag = 0; 848 #ifdef notdef 849 if (dontfrag && alwaysfrag) { /* case 4 */ 850 /* conflicting request - can't transmit */ 851 error = EMSGSIZE; 852 goto bad; 853 } 854 if (dontfrag && tlen > IN6_LINKMTU(ifp)) { /* case 2-b */ 855 /* 856 * Even if the DONTFRAG option is specified, we cannot send the 857 * packet when the data length is larger than the MTU of the 858 * outgoing interface. 859 * Notify the error by sending IPV6_PATHMTU ancillary data as 860 * well as returning an error code (the latter is not described 861 * in the API spec.) 862 */ 863 u_int32_t mtu32; 864 struct ip6ctlparam ip6cp; 865 866 mtu32 = (u_int32_t)mtu; 867 bzero(&ip6cp, sizeof(ip6cp)); 868 ip6cp.ip6c_cmdarg = (void *)&mtu32; 869 pfctlinput2(PRC_MSGSIZE, (struct sockaddr *)&ro_pmtu->ro_dst, 870 (void *)&ip6cp); 871 872 error = EMSGSIZE; 873 goto bad; 874 } 875 #endif 876 /* 877 * transmit packet without fragmentation 878 */ 879 if (dontfrag || (!alwaysfrag && tlen <= mtu)) { /* case 1-a and 2-a */ 880 struct in6_ifaddr *ia6; 881 int sw_csum; 882 883 ip6 = mtod(m, struct ip6_hdr *); 884 ia6 = in6_ifawithifp(ifp, &ip6->ip6_src); 885 if (ia6) { 886 /* Record statistics for this interface address. */ 887 ia6->ia_ifa.ifa_data.ifad_outbytes += m->m_pkthdr.len; 888 } 889 #ifdef IPSEC 890 /* clean ipsec history once it goes out of the node */ 891 ipsec_delaux(m); 892 #endif 893 894 sw_csum = m->m_pkthdr.csum_flags & ~ifp->if_csum_flags_tx; 895 if ((sw_csum & (M_CSUM_UDPv6|M_CSUM_TCPv6)) != 0) { 896 if (IN6_NEED_CHECKSUM(ifp, 897 sw_csum & (M_CSUM_UDPv6|M_CSUM_TCPv6))) { 898 in6_delayed_cksum(m); 899 } 900 m->m_pkthdr.csum_flags &= ~(M_CSUM_UDPv6|M_CSUM_TCPv6); 901 } 902 903 error = nd6_output(ifp, origifp, m, dst, ro->ro_rt); 904 goto done; 905 } 906 907 /* 908 * try to fragment the packet. case 1-b and 3 909 */ 910 if (mtu < IPV6_MMTU) { 911 /* path MTU cannot be less than IPV6_MMTU */ 912 error = EMSGSIZE; 913 in6_ifstat_inc(ifp, ifs6_out_fragfail); 914 goto bad; 915 } else if (ip6->ip6_plen == 0) { 916 /* jumbo payload cannot be fragmented */ 917 error = EMSGSIZE; 918 in6_ifstat_inc(ifp, ifs6_out_fragfail); 919 goto bad; 920 } else { 921 struct mbuf **mnext, *m_frgpart; 922 struct ip6_frag *ip6f; 923 u_int32_t id = htonl(ip6_randomid()); 924 u_char nextproto; 925 struct ip6ctlparam ip6cp; 926 u_int32_t mtu32; 927 928 /* 929 * Too large for the destination or interface; 930 * fragment if possible. 931 * Must be able to put at least 8 bytes per fragment. 932 */ 933 hlen = unfragpartlen; 934 if (mtu > IPV6_MAXPACKET) 935 mtu = IPV6_MAXPACKET; 936 937 /* Notify a proper path MTU to applications. */ 938 mtu32 = (u_int32_t)mtu; 939 bzero(&ip6cp, sizeof(ip6cp)); 940 ip6cp.ip6c_cmdarg = (void *)&mtu32; 941 pfctlinput2(PRC_MSGSIZE, (struct sockaddr *)&ro_pmtu->ro_dst, 942 (void *)&ip6cp); 943 944 len = (mtu - hlen - sizeof(struct ip6_frag)) & ~7; 945 if (len < 8) { 946 error = EMSGSIZE; 947 in6_ifstat_inc(ifp, ifs6_out_fragfail); 948 goto bad; 949 } 950 951 mnext = &m->m_nextpkt; 952 953 /* 954 * Change the next header field of the last header in the 955 * unfragmentable part. 956 */ 957 if (exthdrs.ip6e_rthdr) { 958 nextproto = *mtod(exthdrs.ip6e_rthdr, u_char *); 959 *mtod(exthdrs.ip6e_rthdr, u_char *) = IPPROTO_FRAGMENT; 960 } else if (exthdrs.ip6e_dest1) { 961 nextproto = *mtod(exthdrs.ip6e_dest1, u_char *); 962 *mtod(exthdrs.ip6e_dest1, u_char *) = IPPROTO_FRAGMENT; 963 } else if (exthdrs.ip6e_hbh) { 964 nextproto = *mtod(exthdrs.ip6e_hbh, u_char *); 965 *mtod(exthdrs.ip6e_hbh, u_char *) = IPPROTO_FRAGMENT; 966 } else { 967 nextproto = ip6->ip6_nxt; 968 ip6->ip6_nxt = IPPROTO_FRAGMENT; 969 } 970 971 if ((m->m_pkthdr.csum_flags & (M_CSUM_UDPv6|M_CSUM_TCPv6)) 972 != 0) { 973 if (IN6_NEED_CHECKSUM(ifp, 974 m->m_pkthdr.csum_flags & 975 (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 /* 982 * Loop through length of segment after first fragment, 983 * make new header and copy data of each part and link onto 984 * chain. 985 */ 986 m0 = m; 987 for (off = hlen; off < tlen; off += len) { 988 struct mbuf *mlast; 989 990 MGETHDR(m, M_DONTWAIT, MT_HEADER); 991 if (!m) { 992 error = ENOBUFS; 993 ip6stat.ip6s_odropped++; 994 goto sendorfree; 995 } 996 m->m_pkthdr.rcvif = NULL; 997 m->m_flags = m0->m_flags & M_COPYFLAGS; 998 *mnext = m; 999 mnext = &m->m_nextpkt; 1000 m->m_data += max_linkhdr; 1001 mhip6 = mtod(m, struct ip6_hdr *); 1002 *mhip6 = *ip6; 1003 m->m_len = sizeof(*mhip6); 1004 error = ip6_insertfraghdr(m0, m, hlen, &ip6f); 1005 if (error) { 1006 ip6stat.ip6s_odropped++; 1007 goto sendorfree; 1008 } 1009 ip6f->ip6f_offlg = htons((u_int16_t)((off - hlen) & ~7)); 1010 if (off + len >= tlen) 1011 len = tlen - off; 1012 else 1013 ip6f->ip6f_offlg |= IP6F_MORE_FRAG; 1014 mhip6->ip6_plen = htons((u_int16_t)(len + hlen + 1015 sizeof(*ip6f) - sizeof(struct ip6_hdr))); 1016 if ((m_frgpart = m_copy(m0, off, len)) == 0) { 1017 error = ENOBUFS; 1018 ip6stat.ip6s_odropped++; 1019 goto sendorfree; 1020 } 1021 for (mlast = m; mlast->m_next; mlast = mlast->m_next) 1022 ; 1023 mlast->m_next = m_frgpart; 1024 m->m_pkthdr.len = len + hlen + sizeof(*ip6f); 1025 m->m_pkthdr.rcvif = (struct ifnet *)0; 1026 ip6f->ip6f_reserved = 0; 1027 ip6f->ip6f_ident = id; 1028 ip6f->ip6f_nxt = nextproto; 1029 ip6stat.ip6s_ofragments++; 1030 in6_ifstat_inc(ifp, ifs6_out_fragcreat); 1031 } 1032 1033 in6_ifstat_inc(ifp, ifs6_out_fragok); 1034 } 1035 1036 /* 1037 * Remove leading garbages. 1038 */ 1039 sendorfree: 1040 m = m0->m_nextpkt; 1041 m0->m_nextpkt = 0; 1042 m_freem(m0); 1043 for (m0 = m; m; m = m0) { 1044 m0 = m->m_nextpkt; 1045 m->m_nextpkt = 0; 1046 if (error == 0) { 1047 struct in6_ifaddr *ia6; 1048 ip6 = mtod(m, struct ip6_hdr *); 1049 ia6 = in6_ifawithifp(ifp, &ip6->ip6_src); 1050 if (ia6) { 1051 /* 1052 * Record statistics for this interface 1053 * address. 1054 */ 1055 ia6->ia_ifa.ifa_data.ifad_outbytes += 1056 m->m_pkthdr.len; 1057 } 1058 #ifdef IPSEC 1059 /* clean ipsec history once it goes out of the node */ 1060 ipsec_delaux(m); 1061 #endif 1062 error = nd6_output(ifp, origifp, m, dst, ro->ro_rt); 1063 } else 1064 m_freem(m); 1065 } 1066 1067 if (error == 0) 1068 ip6stat.ip6s_fragmented++; 1069 1070 done: 1071 if (ro == &ip6route && ro->ro_rt) { /* brace necessary for RTFREE */ 1072 RTFREE(ro->ro_rt); 1073 } else if (ro_pmtu == &ip6route && ro_pmtu->ro_rt) { 1074 RTFREE(ro_pmtu->ro_rt); 1075 } 1076 1077 #ifdef IPSEC 1078 if (sp != NULL) 1079 key_freesp(sp); 1080 #endif /* IPSEC */ 1081 1082 return (error); 1083 1084 freehdrs: 1085 m_freem(exthdrs.ip6e_hbh); /* m_freem will check if mbuf is 0 */ 1086 m_freem(exthdrs.ip6e_dest1); 1087 m_freem(exthdrs.ip6e_rthdr); 1088 m_freem(exthdrs.ip6e_dest2); 1089 /* FALLTHROUGH */ 1090 bad: 1091 m_freem(m); 1092 goto done; 1093 } 1094 1095 static int 1096 ip6_copyexthdr(mp, hdr, hlen) 1097 struct mbuf **mp; 1098 caddr_t hdr; 1099 int hlen; 1100 { 1101 struct mbuf *m; 1102 1103 if (hlen > MCLBYTES) 1104 return (ENOBUFS); /* XXX */ 1105 1106 MGET(m, M_DONTWAIT, MT_DATA); 1107 if (!m) 1108 return (ENOBUFS); 1109 1110 if (hlen > MLEN) { 1111 MCLGET(m, M_DONTWAIT); 1112 if ((m->m_flags & M_EXT) == 0) { 1113 m_free(m); 1114 return (ENOBUFS); 1115 } 1116 } 1117 m->m_len = hlen; 1118 if (hdr) 1119 bcopy(hdr, mtod(m, caddr_t), hlen); 1120 1121 *mp = m; 1122 return (0); 1123 } 1124 1125 /* 1126 * Process a delayed payload checksum calculation. 1127 */ 1128 void 1129 in6_delayed_cksum(struct mbuf *m) 1130 { 1131 uint16_t csum, offset; 1132 1133 KASSERT((m->m_pkthdr.csum_flags & (M_CSUM_UDPv6|M_CSUM_TCPv6)) != 0); 1134 KASSERT((~m->m_pkthdr.csum_flags & (M_CSUM_UDPv6|M_CSUM_TCPv6)) != 0); 1135 KASSERT((m->m_pkthdr.csum_flags 1136 & (M_CSUM_UDPv4|M_CSUM_TCPv4|M_CSUM_TSOv4)) == 0); 1137 1138 offset = M_CSUM_DATA_IPv6_HL(m->m_pkthdr.csum_data); 1139 csum = in6_cksum(m, 0, offset, m->m_pkthdr.len - offset); 1140 if (csum == 0 && (m->m_pkthdr.csum_flags & M_CSUM_UDPv6) != 0) { 1141 csum = 0xffff; 1142 } 1143 1144 offset += M_CSUM_DATA_IPv6_OFFSET(m->m_pkthdr.csum_data); 1145 if ((offset + sizeof(csum)) > m->m_len) { 1146 m_copyback(m, offset, sizeof(csum), &csum); 1147 } else { 1148 *(uint16_t *)(mtod(m, caddr_t) + offset) = csum; 1149 } 1150 } 1151 1152 /* 1153 * Insert jumbo payload option. 1154 */ 1155 static int 1156 ip6_insert_jumboopt(exthdrs, plen) 1157 struct ip6_exthdrs *exthdrs; 1158 u_int32_t plen; 1159 { 1160 struct mbuf *mopt; 1161 u_int8_t *optbuf; 1162 u_int32_t v; 1163 1164 #define JUMBOOPTLEN 8 /* length of jumbo payload option and padding */ 1165 1166 /* 1167 * If there is no hop-by-hop options header, allocate new one. 1168 * If there is one but it doesn't have enough space to store the 1169 * jumbo payload option, allocate a cluster to store the whole options. 1170 * Otherwise, use it to store the options. 1171 */ 1172 if (exthdrs->ip6e_hbh == 0) { 1173 MGET(mopt, M_DONTWAIT, MT_DATA); 1174 if (mopt == 0) 1175 return (ENOBUFS); 1176 mopt->m_len = JUMBOOPTLEN; 1177 optbuf = mtod(mopt, u_int8_t *); 1178 optbuf[1] = 0; /* = ((JUMBOOPTLEN) >> 3) - 1 */ 1179 exthdrs->ip6e_hbh = mopt; 1180 } else { 1181 struct ip6_hbh *hbh; 1182 1183 mopt = exthdrs->ip6e_hbh; 1184 if (M_TRAILINGSPACE(mopt) < JUMBOOPTLEN) { 1185 /* 1186 * XXX assumption: 1187 * - exthdrs->ip6e_hbh is not referenced from places 1188 * other than exthdrs. 1189 * - exthdrs->ip6e_hbh is not an mbuf chain. 1190 */ 1191 int oldoptlen = mopt->m_len; 1192 struct mbuf *n; 1193 1194 /* 1195 * XXX: give up if the whole (new) hbh header does 1196 * not fit even in an mbuf cluster. 1197 */ 1198 if (oldoptlen + JUMBOOPTLEN > MCLBYTES) 1199 return (ENOBUFS); 1200 1201 /* 1202 * As a consequence, we must always prepare a cluster 1203 * at this point. 1204 */ 1205 MGET(n, M_DONTWAIT, MT_DATA); 1206 if (n) { 1207 MCLGET(n, M_DONTWAIT); 1208 if ((n->m_flags & M_EXT) == 0) { 1209 m_freem(n); 1210 n = NULL; 1211 } 1212 } 1213 if (!n) 1214 return (ENOBUFS); 1215 n->m_len = oldoptlen + JUMBOOPTLEN; 1216 bcopy(mtod(mopt, caddr_t), mtod(n, caddr_t), 1217 oldoptlen); 1218 optbuf = mtod(n, u_int8_t *) + oldoptlen; 1219 m_freem(mopt); 1220 mopt = exthdrs->ip6e_hbh = n; 1221 } else { 1222 optbuf = mtod(mopt, u_int8_t *) + mopt->m_len; 1223 mopt->m_len += JUMBOOPTLEN; 1224 } 1225 optbuf[0] = IP6OPT_PADN; 1226 optbuf[1] = 0; 1227 1228 /* 1229 * Adjust the header length according to the pad and 1230 * the jumbo payload option. 1231 */ 1232 hbh = mtod(mopt, struct ip6_hbh *); 1233 hbh->ip6h_len += (JUMBOOPTLEN >> 3); 1234 } 1235 1236 /* fill in the option. */ 1237 optbuf[2] = IP6OPT_JUMBO; 1238 optbuf[3] = 4; 1239 v = (u_int32_t)htonl(plen + JUMBOOPTLEN); 1240 bcopy(&v, &optbuf[4], sizeof(u_int32_t)); 1241 1242 /* finally, adjust the packet header length */ 1243 exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN; 1244 1245 return (0); 1246 #undef JUMBOOPTLEN 1247 } 1248 1249 /* 1250 * Insert fragment header and copy unfragmentable header portions. 1251 */ 1252 static int 1253 ip6_insertfraghdr(m0, m, hlen, frghdrp) 1254 struct mbuf *m0, *m; 1255 int hlen; 1256 struct ip6_frag **frghdrp; 1257 { 1258 struct mbuf *n, *mlast; 1259 1260 if (hlen > sizeof(struct ip6_hdr)) { 1261 n = m_copym(m0, sizeof(struct ip6_hdr), 1262 hlen - sizeof(struct ip6_hdr), M_DONTWAIT); 1263 if (n == 0) 1264 return (ENOBUFS); 1265 m->m_next = n; 1266 } else 1267 n = m; 1268 1269 /* Search for the last mbuf of unfragmentable part. */ 1270 for (mlast = n; mlast->m_next; mlast = mlast->m_next) 1271 ; 1272 1273 if ((mlast->m_flags & M_EXT) == 0 && 1274 M_TRAILINGSPACE(mlast) >= sizeof(struct ip6_frag)) { 1275 /* use the trailing space of the last mbuf for the fragment hdr */ 1276 *frghdrp = (struct ip6_frag *)(mtod(mlast, caddr_t) + 1277 mlast->m_len); 1278 mlast->m_len += sizeof(struct ip6_frag); 1279 m->m_pkthdr.len += sizeof(struct ip6_frag); 1280 } else { 1281 /* allocate a new mbuf for the fragment header */ 1282 struct mbuf *mfrg; 1283 1284 MGET(mfrg, M_DONTWAIT, MT_DATA); 1285 if (mfrg == 0) 1286 return (ENOBUFS); 1287 mfrg->m_len = sizeof(struct ip6_frag); 1288 *frghdrp = mtod(mfrg, struct ip6_frag *); 1289 mlast->m_next = mfrg; 1290 } 1291 1292 return (0); 1293 } 1294 1295 static int 1296 ip6_getpmtu(ro_pmtu, ro, ifp, dst, mtup, alwaysfragp) 1297 struct route_in6 *ro_pmtu, *ro; 1298 struct ifnet *ifp; 1299 struct in6_addr *dst; 1300 u_long *mtup; 1301 int *alwaysfragp; 1302 { 1303 u_int32_t mtu = 0; 1304 int alwaysfrag = 0; 1305 int error = 0; 1306 1307 if (ro_pmtu != ro) { 1308 /* The first hop and the final destination may differ. */ 1309 struct sockaddr_in6 *sa6_dst = 1310 (struct sockaddr_in6 *)&ro_pmtu->ro_dst; 1311 if (ro_pmtu->ro_rt && 1312 ((ro_pmtu->ro_rt->rt_flags & RTF_UP) == 0 || 1313 !IN6_ARE_ADDR_EQUAL(&sa6_dst->sin6_addr, dst))) { 1314 RTFREE(ro_pmtu->ro_rt); 1315 ro_pmtu->ro_rt = (struct rtentry *)NULL; 1316 } 1317 if (ro_pmtu->ro_rt == NULL) { 1318 bzero(sa6_dst, sizeof(*sa6_dst)); /* for safety */ 1319 sa6_dst->sin6_family = AF_INET6; 1320 sa6_dst->sin6_len = sizeof(struct sockaddr_in6); 1321 sa6_dst->sin6_addr = *dst; 1322 1323 rtalloc((struct route *)ro_pmtu); 1324 } 1325 } 1326 if (ro_pmtu->ro_rt) { 1327 u_int32_t ifmtu; 1328 1329 if (ifp == NULL) 1330 ifp = ro_pmtu->ro_rt->rt_ifp; 1331 ifmtu = IN6_LINKMTU(ifp); 1332 mtu = ro_pmtu->ro_rt->rt_rmx.rmx_mtu; 1333 if (mtu == 0) 1334 mtu = ifmtu; 1335 else if (mtu < IPV6_MMTU) { 1336 /* 1337 * RFC2460 section 5, last paragraph: 1338 * if we record ICMPv6 too big message with 1339 * mtu < IPV6_MMTU, transmit packets sized IPV6_MMTU 1340 * or smaller, with fragment header attached. 1341 * (fragment header is needed regardless from the 1342 * packet size, for translators to identify packets) 1343 */ 1344 alwaysfrag = 1; 1345 mtu = IPV6_MMTU; 1346 } else if (mtu > ifmtu) { 1347 /* 1348 * The MTU on the route is larger than the MTU on 1349 * the interface! This shouldn't happen, unless the 1350 * MTU of the interface has been changed after the 1351 * interface was brought up. Change the MTU in the 1352 * route to match the interface MTU (as long as the 1353 * field isn't locked). 1354 */ 1355 mtu = ifmtu; 1356 if (!(ro_pmtu->ro_rt->rt_rmx.rmx_locks & RTV_MTU)) 1357 ro_pmtu->ro_rt->rt_rmx.rmx_mtu = mtu; 1358 } 1359 } else if (ifp) { 1360 mtu = IN6_LINKMTU(ifp); 1361 } else 1362 error = EHOSTUNREACH; /* XXX */ 1363 1364 *mtup = mtu; 1365 if (alwaysfragp) 1366 *alwaysfragp = alwaysfrag; 1367 return (error); 1368 } 1369 1370 /* 1371 * IP6 socket option processing. 1372 */ 1373 int 1374 ip6_ctloutput(op, so, level, optname, mp) 1375 int op; 1376 struct socket *so; 1377 int level, optname; 1378 struct mbuf **mp; 1379 { 1380 struct in6pcb *in6p = sotoin6pcb(so); 1381 struct mbuf *m = *mp; 1382 int optval = 0; 1383 int error = 0; 1384 struct proc *p = curproc; /* XXX */ 1385 1386 if (level == IPPROTO_IPV6) { 1387 switch (op) { 1388 case PRCO_SETOPT: 1389 switch (optname) { 1390 case IPV6_PKTOPTIONS: 1391 /* m is freed in ip6_pcbopts */ 1392 return (ip6_pcbopts(&in6p->in6p_outputopts, 1393 m, so)); 1394 case IPV6_HOPOPTS: 1395 case IPV6_DSTOPTS: 1396 if (p == 0 || suser(p->p_ucred, &p->p_acflag)) { 1397 error = EPERM; 1398 break; 1399 } 1400 /* FALLTHROUGH */ 1401 case IPV6_UNICAST_HOPS: 1402 case IPV6_RECVOPTS: 1403 case IPV6_RECVRETOPTS: 1404 case IPV6_RECVDSTADDR: 1405 case IPV6_PKTINFO: 1406 case IPV6_HOPLIMIT: 1407 case IPV6_RTHDR: 1408 case IPV6_FAITH: 1409 case IPV6_V6ONLY: 1410 case IPV6_USE_MIN_MTU: 1411 if (!m || m->m_len != sizeof(int)) { 1412 error = EINVAL; 1413 break; 1414 } 1415 optval = *mtod(m, int *); 1416 switch (optname) { 1417 1418 case IPV6_UNICAST_HOPS: 1419 if (optval < -1 || optval >= 256) 1420 error = EINVAL; 1421 else { 1422 /* -1 = kernel default */ 1423 in6p->in6p_hops = optval; 1424 } 1425 break; 1426 #define OPTSET(bit) \ 1427 do { \ 1428 if (optval) \ 1429 in6p->in6p_flags |= (bit); \ 1430 else \ 1431 in6p->in6p_flags &= ~(bit); \ 1432 } while (/*CONSTCOND*/ 0) 1433 1434 case IPV6_RECVOPTS: 1435 OPTSET(IN6P_RECVOPTS); 1436 break; 1437 1438 case IPV6_RECVRETOPTS: 1439 OPTSET(IN6P_RECVRETOPTS); 1440 break; 1441 1442 case IPV6_RECVDSTADDR: 1443 OPTSET(IN6P_RECVDSTADDR); 1444 break; 1445 1446 case IPV6_PKTINFO: 1447 OPTSET(IN6P_PKTINFO); 1448 break; 1449 1450 case IPV6_HOPLIMIT: 1451 OPTSET(IN6P_HOPLIMIT); 1452 break; 1453 1454 case IPV6_HOPOPTS: 1455 OPTSET(IN6P_HOPOPTS); 1456 break; 1457 1458 case IPV6_DSTOPTS: 1459 OPTSET(IN6P_DSTOPTS); 1460 break; 1461 1462 case IPV6_RTHDR: 1463 OPTSET(IN6P_RTHDR); 1464 break; 1465 1466 case IPV6_FAITH: 1467 OPTSET(IN6P_FAITH); 1468 break; 1469 1470 case IPV6_USE_MIN_MTU: 1471 OPTSET(IN6P_MINMTU); 1472 break; 1473 1474 case IPV6_V6ONLY: 1475 /* 1476 * make setsockopt(IPV6_V6ONLY) 1477 * available only prior to bind(2). 1478 * see ipng mailing list, Jun 22 2001. 1479 */ 1480 if (in6p->in6p_lport || 1481 !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr)) { 1482 error = EINVAL; 1483 break; 1484 } 1485 #ifdef INET6_BINDV6ONLY 1486 if (!optval) 1487 error = EINVAL; 1488 #else 1489 OPTSET(IN6P_IPV6_V6ONLY); 1490 #endif 1491 break; 1492 } 1493 break; 1494 #undef OPTSET 1495 1496 case IPV6_MULTICAST_IF: 1497 case IPV6_MULTICAST_HOPS: 1498 case IPV6_MULTICAST_LOOP: 1499 case IPV6_JOIN_GROUP: 1500 case IPV6_LEAVE_GROUP: 1501 error = ip6_setmoptions(optname, 1502 &in6p->in6p_moptions, m); 1503 break; 1504 1505 case IPV6_PORTRANGE: 1506 optval = *mtod(m, int *); 1507 1508 switch (optval) { 1509 case IPV6_PORTRANGE_DEFAULT: 1510 in6p->in6p_flags &= ~(IN6P_LOWPORT); 1511 in6p->in6p_flags &= ~(IN6P_HIGHPORT); 1512 break; 1513 1514 case IPV6_PORTRANGE_HIGH: 1515 in6p->in6p_flags &= ~(IN6P_LOWPORT); 1516 in6p->in6p_flags |= IN6P_HIGHPORT; 1517 break; 1518 1519 case IPV6_PORTRANGE_LOW: 1520 in6p->in6p_flags &= ~(IN6P_HIGHPORT); 1521 in6p->in6p_flags |= IN6P_LOWPORT; 1522 break; 1523 1524 default: 1525 error = EINVAL; 1526 break; 1527 } 1528 break; 1529 1530 #ifdef IPSEC 1531 case IPV6_IPSEC_POLICY: 1532 { 1533 caddr_t req = NULL; 1534 size_t len = 0; 1535 1536 int priv = 0; 1537 if (p == 0 || suser(p->p_ucred, &p->p_acflag)) 1538 priv = 0; 1539 else 1540 priv = 1; 1541 if (m) { 1542 req = mtod(m, caddr_t); 1543 len = m->m_len; 1544 } 1545 error = ipsec6_set_policy(in6p, 1546 optname, req, len, priv); 1547 } 1548 break; 1549 #endif /* IPSEC */ 1550 1551 default: 1552 error = ENOPROTOOPT; 1553 break; 1554 } 1555 if (m) 1556 (void)m_free(m); 1557 break; 1558 1559 case PRCO_GETOPT: 1560 switch (optname) { 1561 1562 case IPV6_OPTIONS: 1563 case IPV6_RETOPTS: 1564 error = ENOPROTOOPT; 1565 break; 1566 1567 case IPV6_PKTOPTIONS: 1568 if (in6p->in6p_options) { 1569 *mp = m_copym(in6p->in6p_options, 0, 1570 M_COPYALL, M_WAIT); 1571 } else { 1572 *mp = m_get(M_WAIT, MT_SOOPTS); 1573 (*mp)->m_len = 0; 1574 } 1575 break; 1576 1577 case IPV6_HOPOPTS: 1578 case IPV6_DSTOPTS: 1579 if (p == 0 || suser(p->p_ucred, &p->p_acflag)) { 1580 error = EPERM; 1581 break; 1582 } 1583 /* FALLTHROUGH */ 1584 case IPV6_UNICAST_HOPS: 1585 case IPV6_RECVOPTS: 1586 case IPV6_RECVRETOPTS: 1587 case IPV6_RECVDSTADDR: 1588 case IPV6_PORTRANGE: 1589 case IPV6_PKTINFO: 1590 case IPV6_HOPLIMIT: 1591 case IPV6_RTHDR: 1592 case IPV6_FAITH: 1593 case IPV6_V6ONLY: 1594 case IPV6_USE_MIN_MTU: 1595 *mp = m = m_get(M_WAIT, MT_SOOPTS); 1596 m->m_len = sizeof(int); 1597 switch (optname) { 1598 1599 case IPV6_UNICAST_HOPS: 1600 optval = in6p->in6p_hops; 1601 break; 1602 1603 #define OPTBIT(bit) (in6p->in6p_flags & bit ? 1 : 0) 1604 1605 case IPV6_RECVOPTS: 1606 optval = OPTBIT(IN6P_RECVOPTS); 1607 break; 1608 1609 case IPV6_RECVRETOPTS: 1610 optval = OPTBIT(IN6P_RECVRETOPTS); 1611 break; 1612 1613 case IPV6_RECVDSTADDR: 1614 optval = OPTBIT(IN6P_RECVDSTADDR); 1615 break; 1616 1617 case IPV6_PORTRANGE: 1618 { 1619 int flags; 1620 flags = in6p->in6p_flags; 1621 if (flags & IN6P_HIGHPORT) 1622 optval = IPV6_PORTRANGE_HIGH; 1623 else if (flags & IN6P_LOWPORT) 1624 optval = IPV6_PORTRANGE_LOW; 1625 else 1626 optval = 0; 1627 break; 1628 } 1629 1630 case IPV6_PKTINFO: 1631 optval = OPTBIT(IN6P_PKTINFO); 1632 break; 1633 1634 case IPV6_HOPLIMIT: 1635 optval = OPTBIT(IN6P_HOPLIMIT); 1636 break; 1637 1638 case IPV6_HOPOPTS: 1639 optval = OPTBIT(IN6P_HOPOPTS); 1640 break; 1641 1642 case IPV6_DSTOPTS: 1643 optval = OPTBIT(IN6P_DSTOPTS); 1644 break; 1645 1646 case IPV6_RTHDR: 1647 optval = OPTBIT(IN6P_RTHDR); 1648 break; 1649 1650 case IPV6_FAITH: 1651 optval = OPTBIT(IN6P_FAITH); 1652 break; 1653 1654 case IPV6_V6ONLY: 1655 optval = OPTBIT(IN6P_IPV6_V6ONLY); 1656 break; 1657 1658 case IPV6_USE_MIN_MTU: 1659 optval = OPTBIT(IN6P_MINMTU); 1660 break; 1661 } 1662 *mtod(m, int *) = optval; 1663 break; 1664 1665 case IPV6_MULTICAST_IF: 1666 case IPV6_MULTICAST_HOPS: 1667 case IPV6_MULTICAST_LOOP: 1668 case IPV6_JOIN_GROUP: 1669 case IPV6_LEAVE_GROUP: 1670 error = ip6_getmoptions(optname, in6p->in6p_moptions, mp); 1671 break; 1672 1673 #if 0 /* defined(IPSEC) */ 1674 /* XXX: code broken */ 1675 case IPV6_IPSEC_POLICY: 1676 { 1677 caddr_t req = NULL; 1678 size_t len = 0; 1679 1680 if (m) { 1681 req = mtod(m, caddr_t); 1682 len = m->m_len; 1683 } 1684 error = ipsec6_get_policy(in6p, req, len, mp); 1685 break; 1686 } 1687 #endif /* IPSEC */ 1688 1689 default: 1690 error = ENOPROTOOPT; 1691 break; 1692 } 1693 break; 1694 } 1695 } else { 1696 error = EINVAL; 1697 if (op == PRCO_SETOPT && *mp) 1698 (void)m_free(*mp); 1699 } 1700 return (error); 1701 } 1702 1703 int 1704 ip6_raw_ctloutput(op, so, level, optname, mp) 1705 int op; 1706 struct socket *so; 1707 int level, optname; 1708 struct mbuf **mp; 1709 { 1710 int error = 0, optval, optlen; 1711 const int icmp6off = offsetof(struct icmp6_hdr, icmp6_cksum); 1712 struct in6pcb *in6p = sotoin6pcb(so); 1713 struct mbuf *m = *mp; 1714 1715 optlen = m ? m->m_len : 0; 1716 1717 if (level != IPPROTO_IPV6) { 1718 if (op == PRCO_SETOPT && *mp) 1719 (void)m_free(*mp); 1720 return (EINVAL); 1721 } 1722 1723 switch (optname) { 1724 case IPV6_CHECKSUM: 1725 /* 1726 * For ICMPv6 sockets, no modification allowed for checksum 1727 * offset, permit "no change" values to help existing apps. 1728 * 1729 * XXX 2292bis says: "An attempt to set IPV6_CHECKSUM 1730 * for an ICMPv6 socket will fail." 1731 * The current behavior does not meet 2292bis. 1732 */ 1733 switch (op) { 1734 case PRCO_SETOPT: 1735 if (optlen != sizeof(int)) { 1736 error = EINVAL; 1737 break; 1738 } 1739 optval = *mtod(m, int *); 1740 if ((optval % 2) != 0) { 1741 /* the API assumes even offset values */ 1742 error = EINVAL; 1743 } else if (so->so_proto->pr_protocol == 1744 IPPROTO_ICMPV6) { 1745 if (optval != icmp6off) 1746 error = EINVAL; 1747 } else 1748 in6p->in6p_cksum = optval; 1749 break; 1750 1751 case PRCO_GETOPT: 1752 if (so->so_proto->pr_protocol == IPPROTO_ICMPV6) 1753 optval = icmp6off; 1754 else 1755 optval = in6p->in6p_cksum; 1756 1757 *mp = m = m_get(M_WAIT, MT_SOOPTS); 1758 m->m_len = sizeof(int); 1759 *mtod(m, int *) = optval; 1760 break; 1761 1762 default: 1763 error = EINVAL; 1764 break; 1765 } 1766 break; 1767 1768 default: 1769 error = ENOPROTOOPT; 1770 break; 1771 } 1772 1773 if (op == PRCO_SETOPT && m) 1774 (void)m_free(m); 1775 1776 return (error); 1777 } 1778 1779 /* 1780 * Set up IP6 options in pcb for insertion in output packets. 1781 * Store in mbuf with pointer in pcbopt, adding pseudo-option 1782 * with destination address if source routed. 1783 */ 1784 static int 1785 ip6_pcbopts(pktopt, m, so) 1786 struct ip6_pktopts **pktopt; 1787 struct mbuf *m; 1788 struct socket *so; 1789 { 1790 struct ip6_pktopts *opt = *pktopt; 1791 int error = 0; 1792 struct proc *p = curproc; /* XXX */ 1793 int priv = 0; 1794 1795 /* turn off any old options. */ 1796 if (opt) { 1797 if (opt->ip6po_m) 1798 (void)m_free(opt->ip6po_m); 1799 } else 1800 opt = malloc(sizeof(*opt), M_IP6OPT, M_WAITOK); 1801 *pktopt = 0; 1802 1803 if (!m || m->m_len == 0) { 1804 /* 1805 * Only turning off any previous options. 1806 */ 1807 free(opt, M_IP6OPT); 1808 if (m) 1809 (void)m_free(m); 1810 return (0); 1811 } 1812 1813 /* set options specified by user. */ 1814 if (p && !suser(p->p_ucred, &p->p_acflag)) 1815 priv = 1; 1816 if ((error = ip6_setpktoptions(m, opt, priv)) != 0) { 1817 (void)m_free(m); 1818 free(opt, M_IP6OPT); 1819 return (error); 1820 } 1821 *pktopt = opt; 1822 return (0); 1823 } 1824 1825 /* 1826 * Set the IP6 multicast options in response to user setsockopt(). 1827 */ 1828 static int 1829 ip6_setmoptions(optname, im6op, m) 1830 int optname; 1831 struct ip6_moptions **im6op; 1832 struct mbuf *m; 1833 { 1834 int error = 0; 1835 u_int loop, ifindex; 1836 struct ipv6_mreq *mreq; 1837 struct ifnet *ifp; 1838 struct ip6_moptions *im6o = *im6op; 1839 struct route_in6 ro; 1840 struct in6_multi_mship *imm; 1841 struct proc *p = curproc; /* XXX */ 1842 1843 if (im6o == NULL) { 1844 /* 1845 * No multicast option buffer attached to the pcb; 1846 * allocate one and initialize to default values. 1847 */ 1848 im6o = (struct ip6_moptions *) 1849 malloc(sizeof(*im6o), M_IPMOPTS, M_WAITOK); 1850 1851 if (im6o == NULL) 1852 return (ENOBUFS); 1853 *im6op = im6o; 1854 im6o->im6o_multicast_ifp = NULL; 1855 im6o->im6o_multicast_hlim = ip6_defmcasthlim; 1856 im6o->im6o_multicast_loop = IPV6_DEFAULT_MULTICAST_LOOP; 1857 LIST_INIT(&im6o->im6o_memberships); 1858 } 1859 1860 switch (optname) { 1861 1862 case IPV6_MULTICAST_IF: 1863 /* 1864 * Select the interface for outgoing multicast packets. 1865 */ 1866 if (m == NULL || m->m_len != sizeof(u_int)) { 1867 error = EINVAL; 1868 break; 1869 } 1870 bcopy(mtod(m, u_int *), &ifindex, sizeof(ifindex)); 1871 if (ifindex != 0) { 1872 if (ifindex < 0 || if_indexlim <= ifindex || 1873 !ifindex2ifnet[ifindex]) { 1874 error = ENXIO; /* XXX EINVAL? */ 1875 break; 1876 } 1877 ifp = ifindex2ifnet[ifindex]; 1878 if ((ifp->if_flags & IFF_MULTICAST) == 0) { 1879 error = EADDRNOTAVAIL; 1880 break; 1881 } 1882 } else 1883 ifp = NULL; 1884 im6o->im6o_multicast_ifp = ifp; 1885 break; 1886 1887 case IPV6_MULTICAST_HOPS: 1888 { 1889 /* 1890 * Set the IP6 hoplimit for outgoing multicast packets. 1891 */ 1892 int optval; 1893 if (m == NULL || m->m_len != sizeof(int)) { 1894 error = EINVAL; 1895 break; 1896 } 1897 bcopy(mtod(m, u_int *), &optval, sizeof(optval)); 1898 if (optval < -1 || optval >= 256) 1899 error = EINVAL; 1900 else if (optval == -1) 1901 im6o->im6o_multicast_hlim = ip6_defmcasthlim; 1902 else 1903 im6o->im6o_multicast_hlim = optval; 1904 break; 1905 } 1906 1907 case IPV6_MULTICAST_LOOP: 1908 /* 1909 * Set the loopback flag for outgoing multicast packets. 1910 * Must be zero or one. 1911 */ 1912 if (m == NULL || m->m_len != sizeof(u_int)) { 1913 error = EINVAL; 1914 break; 1915 } 1916 bcopy(mtod(m, u_int *), &loop, sizeof(loop)); 1917 if (loop > 1) { 1918 error = EINVAL; 1919 break; 1920 } 1921 im6o->im6o_multicast_loop = loop; 1922 break; 1923 1924 case IPV6_JOIN_GROUP: 1925 /* 1926 * Add a multicast group membership. 1927 * Group must be a valid IP6 multicast address. 1928 */ 1929 if (m == NULL || m->m_len != sizeof(struct ipv6_mreq)) { 1930 error = EINVAL; 1931 break; 1932 } 1933 mreq = mtod(m, struct ipv6_mreq *); 1934 if (IN6_IS_ADDR_UNSPECIFIED(&mreq->ipv6mr_multiaddr)) { 1935 /* 1936 * We use the unspecified address to specify to accept 1937 * all multicast addresses. Only super user is allowed 1938 * to do this. 1939 */ 1940 if (suser(p->p_ucred, &p->p_acflag)) 1941 { 1942 error = EACCES; 1943 break; 1944 } 1945 } else if (!IN6_IS_ADDR_MULTICAST(&mreq->ipv6mr_multiaddr)) { 1946 error = EINVAL; 1947 break; 1948 } 1949 1950 /* 1951 * If no interface was explicitly specified, choose an 1952 * appropriate one according to the given multicast address. 1953 */ 1954 if (mreq->ipv6mr_interface == 0) { 1955 struct sockaddr_in6 *dst; 1956 1957 /* 1958 * Look up the routing table for the 1959 * address, and choose the outgoing interface. 1960 * XXX: is it a good approach? 1961 */ 1962 ro.ro_rt = NULL; 1963 dst = (struct sockaddr_in6 *)&ro.ro_dst; 1964 bzero(dst, sizeof(*dst)); 1965 dst->sin6_family = AF_INET6; 1966 dst->sin6_len = sizeof(*dst); 1967 dst->sin6_addr = mreq->ipv6mr_multiaddr; 1968 rtalloc((struct route *)&ro); 1969 if (ro.ro_rt == NULL) { 1970 error = EADDRNOTAVAIL; 1971 break; 1972 } 1973 ifp = ro.ro_rt->rt_ifp; 1974 rtfree(ro.ro_rt); 1975 } else { 1976 /* 1977 * If the interface is specified, validate it. 1978 */ 1979 if (mreq->ipv6mr_interface < 0 || 1980 if_indexlim <= mreq->ipv6mr_interface || 1981 !ifindex2ifnet[mreq->ipv6mr_interface]) { 1982 error = ENXIO; /* XXX EINVAL? */ 1983 break; 1984 } 1985 ifp = ifindex2ifnet[mreq->ipv6mr_interface]; 1986 } 1987 1988 /* 1989 * See if we found an interface, and confirm that it 1990 * supports multicast 1991 */ 1992 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) { 1993 error = EADDRNOTAVAIL; 1994 break; 1995 } 1996 1997 if (in6_setscope(&mreq->ipv6mr_multiaddr, ifp, NULL)) { 1998 error = EADDRNOTAVAIL; /* XXX: should not happen */ 1999 break; 2000 } 2001 2002 /* 2003 * See if the membership already exists. 2004 */ 2005 for (imm = im6o->im6o_memberships.lh_first; 2006 imm != NULL; imm = imm->i6mm_chain.le_next) 2007 if (imm->i6mm_maddr->in6m_ifp == ifp && 2008 IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr, 2009 &mreq->ipv6mr_multiaddr)) 2010 break; 2011 if (imm != NULL) { 2012 error = EADDRINUSE; 2013 break; 2014 } 2015 /* 2016 * Everything looks good; add a new record to the multicast 2017 * address list for the given interface. 2018 */ 2019 imm = in6_joingroup(ifp, &mreq->ipv6mr_multiaddr, &error, 0); 2020 if (imm == NULL) 2021 break; 2022 LIST_INSERT_HEAD(&im6o->im6o_memberships, imm, i6mm_chain); 2023 break; 2024 2025 case IPV6_LEAVE_GROUP: 2026 /* 2027 * Drop a multicast group membership. 2028 * Group must be a valid IP6 multicast address. 2029 */ 2030 if (m == NULL || m->m_len != sizeof(struct ipv6_mreq)) { 2031 error = EINVAL; 2032 break; 2033 } 2034 mreq = mtod(m, struct ipv6_mreq *); 2035 2036 /* 2037 * If an interface address was specified, get a pointer 2038 * to its ifnet structure. 2039 */ 2040 if (mreq->ipv6mr_interface != 0) { 2041 if (mreq->ipv6mr_interface < 0 || 2042 if_indexlim <= mreq->ipv6mr_interface || 2043 !ifindex2ifnet[mreq->ipv6mr_interface]) { 2044 error = ENXIO; /* XXX EINVAL? */ 2045 break; 2046 } 2047 ifp = ifindex2ifnet[mreq->ipv6mr_interface]; 2048 } else 2049 ifp = NULL; 2050 2051 /* Fill in the scope zone ID */ 2052 if (ifp) { 2053 if (in6_setscope(&mreq->ipv6mr_multiaddr, ifp, NULL)) { 2054 /* XXX: should not happen */ 2055 error = EADDRNOTAVAIL; 2056 break; 2057 } 2058 } else if (mreq->ipv6mr_interface != 0) { 2059 /* 2060 * XXX: This case would happens when the (positive) 2061 * index is in the valid range, but the corresponding 2062 * interface has been detached dynamically. The above 2063 * check probably avoids such case to happen here, but 2064 * we check it explicitly for safety. 2065 */ 2066 error = EADDRNOTAVAIL; 2067 break; 2068 } else { /* ipv6mr_interface == 0 */ 2069 struct sockaddr_in6 sa6_mc; 2070 2071 /* 2072 * The API spec says as follows: 2073 * If the interface index is specified as 0, the 2074 * system may choose a multicast group membership to 2075 * drop by matching the multicast address only. 2076 * On the other hand, we cannot disambiguate the scope 2077 * zone unless an interface is provided. Thus, we 2078 * check if there's ambiguity with the default scope 2079 * zone as the last resort. 2080 */ 2081 bzero(&sa6_mc, sizeof(sa6_mc)); 2082 sa6_mc.sin6_family = AF_INET6; 2083 sa6_mc.sin6_len = sizeof(sa6_mc); 2084 sa6_mc.sin6_addr = mreq->ipv6mr_multiaddr; 2085 error = sa6_embedscope(&sa6_mc, ip6_use_defzone); 2086 if (error != 0) 2087 break; 2088 mreq->ipv6mr_multiaddr = sa6_mc.sin6_addr; 2089 } 2090 2091 /* 2092 * Find the membership in the membership list. 2093 */ 2094 for (imm = im6o->im6o_memberships.lh_first; 2095 imm != NULL; imm = imm->i6mm_chain.le_next) { 2096 if ((ifp == NULL || imm->i6mm_maddr->in6m_ifp == ifp) && 2097 IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr, 2098 &mreq->ipv6mr_multiaddr)) 2099 break; 2100 } 2101 if (imm == NULL) { 2102 /* Unable to resolve interface */ 2103 error = EADDRNOTAVAIL; 2104 break; 2105 } 2106 /* 2107 * Give up the multicast address record to which the 2108 * membership points. 2109 */ 2110 LIST_REMOVE(imm, i6mm_chain); 2111 in6_leavegroup(imm); 2112 break; 2113 2114 default: 2115 error = EOPNOTSUPP; 2116 break; 2117 } 2118 2119 /* 2120 * If all options have default values, no need to keep the mbuf. 2121 */ 2122 if (im6o->im6o_multicast_ifp == NULL && 2123 im6o->im6o_multicast_hlim == ip6_defmcasthlim && 2124 im6o->im6o_multicast_loop == IPV6_DEFAULT_MULTICAST_LOOP && 2125 im6o->im6o_memberships.lh_first == NULL) { 2126 free(*im6op, M_IPMOPTS); 2127 *im6op = NULL; 2128 } 2129 2130 return (error); 2131 } 2132 2133 /* 2134 * Return the IP6 multicast options in response to user getsockopt(). 2135 */ 2136 static int 2137 ip6_getmoptions(optname, im6o, mp) 2138 int optname; 2139 struct ip6_moptions *im6o; 2140 struct mbuf **mp; 2141 { 2142 u_int *hlim, *loop, *ifindex; 2143 2144 *mp = m_get(M_WAIT, MT_SOOPTS); 2145 2146 switch (optname) { 2147 2148 case IPV6_MULTICAST_IF: 2149 ifindex = mtod(*mp, u_int *); 2150 (*mp)->m_len = sizeof(u_int); 2151 if (im6o == NULL || im6o->im6o_multicast_ifp == NULL) 2152 *ifindex = 0; 2153 else 2154 *ifindex = im6o->im6o_multicast_ifp->if_index; 2155 return (0); 2156 2157 case IPV6_MULTICAST_HOPS: 2158 hlim = mtod(*mp, u_int *); 2159 (*mp)->m_len = sizeof(u_int); 2160 if (im6o == NULL) 2161 *hlim = ip6_defmcasthlim; 2162 else 2163 *hlim = im6o->im6o_multicast_hlim; 2164 return (0); 2165 2166 case IPV6_MULTICAST_LOOP: 2167 loop = mtod(*mp, u_int *); 2168 (*mp)->m_len = sizeof(u_int); 2169 if (im6o == NULL) 2170 *loop = ip6_defmcasthlim; 2171 else 2172 *loop = im6o->im6o_multicast_loop; 2173 return (0); 2174 2175 default: 2176 return (EOPNOTSUPP); 2177 } 2178 } 2179 2180 /* 2181 * Discard the IP6 multicast options. 2182 */ 2183 void 2184 ip6_freemoptions(im6o) 2185 struct ip6_moptions *im6o; 2186 { 2187 struct in6_multi_mship *imm; 2188 2189 if (im6o == NULL) 2190 return; 2191 2192 while ((imm = im6o->im6o_memberships.lh_first) != NULL) { 2193 LIST_REMOVE(imm, i6mm_chain); 2194 in6_leavegroup(imm); 2195 } 2196 free(im6o, M_IPMOPTS); 2197 } 2198 2199 /* 2200 * Set IPv6 outgoing packet options based on advanced API. 2201 */ 2202 int 2203 ip6_setpktoptions(control, opt, priv) 2204 struct mbuf *control; 2205 struct ip6_pktopts *opt; 2206 int priv; 2207 { 2208 struct cmsghdr *cm = 0; 2209 2210 if (control == 0 || opt == 0) 2211 return (EINVAL); 2212 2213 bzero(opt, sizeof(*opt)); 2214 opt->ip6po_hlim = -1; /* -1 means to use default hop limit */ 2215 2216 /* 2217 * XXX: Currently, we assume all the optional information is stored 2218 * in a single mbuf. 2219 */ 2220 if (control->m_next) 2221 return (EINVAL); 2222 2223 opt->ip6po_m = control; 2224 2225 for (; control->m_len; control->m_data += CMSG_ALIGN(cm->cmsg_len), 2226 control->m_len -= CMSG_ALIGN(cm->cmsg_len)) { 2227 cm = mtod(control, struct cmsghdr *); 2228 if (cm->cmsg_len == 0 || cm->cmsg_len > control->m_len) 2229 return (EINVAL); 2230 if (cm->cmsg_level != IPPROTO_IPV6) 2231 continue; 2232 2233 switch (cm->cmsg_type) { 2234 case IPV6_PKTINFO: 2235 if (cm->cmsg_len != CMSG_LEN(sizeof(struct in6_pktinfo))) 2236 return (EINVAL); 2237 opt->ip6po_pktinfo = (struct in6_pktinfo *)CMSG_DATA(cm); 2238 if (opt->ip6po_pktinfo->ipi6_ifindex >= if_indexlim || 2239 opt->ip6po_pktinfo->ipi6_ifindex < 0) 2240 return (ENXIO); 2241 if (opt->ip6po_pktinfo->ipi6_ifindex > 0 && 2242 !ifindex2ifnet[opt->ip6po_pktinfo->ipi6_ifindex]) 2243 return (ENXIO); 2244 2245 if (opt->ip6po_pktinfo->ipi6_ifindex) { 2246 struct ifnet *ifp; 2247 int error; 2248 2249 /* ipi6_ifindex must be valid here */ 2250 ifp = ifindex2ifnet[opt->ip6po_pktinfo->ipi6_ifindex]; 2251 error = in6_setscope(&opt->ip6po_pktinfo->ipi6_addr, 2252 ifp, NULL); 2253 if (error != 0) 2254 return (error); 2255 } 2256 2257 /* 2258 * Check if the requested source address is indeed a 2259 * unicast address assigned to the node, and can be 2260 * used as the packet's source address. 2261 */ 2262 if (!IN6_IS_ADDR_UNSPECIFIED(&opt->ip6po_pktinfo->ipi6_addr)) { 2263 struct ifaddr *ia; 2264 struct in6_ifaddr *ia6; 2265 struct sockaddr_in6 sin6; 2266 2267 bzero(&sin6, sizeof(sin6)); 2268 sin6.sin6_len = sizeof(sin6); 2269 sin6.sin6_family = AF_INET6; 2270 sin6.sin6_addr = 2271 opt->ip6po_pktinfo->ipi6_addr; 2272 ia = ifa_ifwithaddr(sin6tosa(&sin6)); 2273 if (ia == NULL || 2274 (opt->ip6po_pktinfo->ipi6_ifindex && 2275 (ia->ifa_ifp->if_index != 2276 opt->ip6po_pktinfo->ipi6_ifindex))) { 2277 return (EADDRNOTAVAIL); 2278 } 2279 ia6 = (struct in6_ifaddr *)ia; 2280 if ((ia6->ia6_flags & (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY)) != 0) { 2281 return (EADDRNOTAVAIL); 2282 } 2283 2284 /* 2285 * Check if the requested source address is 2286 * indeed a unicast address assigned to the 2287 * node. 2288 */ 2289 if (IN6_IS_ADDR_MULTICAST(&opt->ip6po_pktinfo->ipi6_addr)) 2290 return (EADDRNOTAVAIL); 2291 } 2292 break; 2293 2294 case IPV6_HOPLIMIT: 2295 if (cm->cmsg_len != CMSG_LEN(sizeof(int))) 2296 return (EINVAL); 2297 else { 2298 int t; 2299 2300 bcopy(CMSG_DATA(cm), &t, sizeof(t)); 2301 if (t < -1 || t > 255) 2302 return (EINVAL); 2303 opt->ip6po_hlim = t; 2304 } 2305 break; 2306 2307 case IPV6_NEXTHOP: 2308 if (!priv) 2309 return (EPERM); 2310 2311 /* check if cmsg_len is large enough for sa_len */ 2312 if (cm->cmsg_len < sizeof(u_char) || 2313 cm->cmsg_len < CMSG_LEN(*CMSG_DATA(cm))) 2314 return (EINVAL); 2315 2316 opt->ip6po_nexthop = (struct sockaddr *)CMSG_DATA(cm); 2317 2318 break; 2319 2320 case IPV6_HOPOPTS: 2321 if (cm->cmsg_len < CMSG_LEN(sizeof(struct ip6_hbh))) 2322 return (EINVAL); 2323 else { 2324 struct ip6_hbh *t; 2325 2326 t = (struct ip6_hbh *)CMSG_DATA(cm); 2327 if (cm->cmsg_len != 2328 CMSG_LEN((t->ip6h_len + 1) << 3)) 2329 return (EINVAL); 2330 opt->ip6po_hbh = t; 2331 } 2332 break; 2333 2334 case IPV6_DSTOPTS: 2335 if (cm->cmsg_len < CMSG_LEN(sizeof(struct ip6_dest))) 2336 return (EINVAL); 2337 2338 /* 2339 * If there is no routing header yet, the destination 2340 * options header should be put on the 1st part. 2341 * Otherwise, the header should be on the 2nd part. 2342 * (See RFC 2460, section 4.1) 2343 */ 2344 if (opt->ip6po_rthdr == NULL) { 2345 struct ip6_dest *t; 2346 2347 t = (struct ip6_dest *)CMSG_DATA(cm); 2348 if (cm->cmsg_len != 2349 CMSG_LEN((t->ip6d_len + 1) << 3)); 2350 return (EINVAL); 2351 opt->ip6po_dest1 = t; 2352 } 2353 else { 2354 struct ip6_dest *t; 2355 2356 t = (struct ip6_dest *)CMSG_DATA(cm); 2357 if (cm->cmsg_len != 2358 CMSG_LEN((opt->ip6po_dest2->ip6d_len + 1) << 3)) 2359 return (EINVAL); 2360 opt->ip6po_dest2 = t; 2361 } 2362 break; 2363 2364 case IPV6_RTHDR: 2365 if (cm->cmsg_len < CMSG_LEN(sizeof(struct ip6_rthdr))) 2366 return (EINVAL); 2367 else { 2368 struct ip6_rthdr *t; 2369 2370 t = (struct ip6_rthdr *)CMSG_DATA(cm); 2371 if (cm->cmsg_len != 2372 CMSG_LEN((t->ip6r_len + 1) << 3)) 2373 return (EINVAL); 2374 switch (t->ip6r_type) { 2375 case IPV6_RTHDR_TYPE_0: 2376 if (t->ip6r_segleft == 0) 2377 return (EINVAL); 2378 break; 2379 default: 2380 return (EINVAL); 2381 } 2382 opt->ip6po_rthdr = t; 2383 } 2384 break; 2385 2386 default: 2387 return (ENOPROTOOPT); 2388 } 2389 } 2390 2391 return (0); 2392 } 2393 2394 /* 2395 * Routine called from ip6_output() to loop back a copy of an IP6 multicast 2396 * packet to the input queue of a specified interface. Note that this 2397 * calls the output routine of the loopback "driver", but with an interface 2398 * pointer that might NOT be lo0ifp -- easier than replicating that code here. 2399 */ 2400 void 2401 ip6_mloopback(ifp, m, dst) 2402 struct ifnet *ifp; 2403 struct mbuf *m; 2404 struct sockaddr_in6 *dst; 2405 { 2406 struct mbuf *copym; 2407 struct ip6_hdr *ip6; 2408 2409 copym = m_copy(m, 0, M_COPYALL); 2410 if (copym == NULL) 2411 return; 2412 2413 /* 2414 * Make sure to deep-copy IPv6 header portion in case the data 2415 * is in an mbuf cluster, so that we can safely override the IPv6 2416 * header portion later. 2417 */ 2418 if ((copym->m_flags & M_EXT) != 0 || 2419 copym->m_len < sizeof(struct ip6_hdr)) { 2420 copym = m_pullup(copym, sizeof(struct ip6_hdr)); 2421 if (copym == NULL) 2422 return; 2423 } 2424 2425 #ifdef DIAGNOSTIC 2426 if (copym->m_len < sizeof(*ip6)) { 2427 m_freem(copym); 2428 return; 2429 } 2430 #endif 2431 2432 ip6 = mtod(copym, struct ip6_hdr *); 2433 /* 2434 * clear embedded scope identifiers if necessary. 2435 * in6_clearscope will touch the addresses only when necessary. 2436 */ 2437 in6_clearscope(&ip6->ip6_src); 2438 in6_clearscope(&ip6->ip6_dst); 2439 2440 (void)looutput(ifp, copym, (struct sockaddr *)dst, NULL); 2441 } 2442 2443 /* 2444 * Chop IPv6 header off from the payload. 2445 */ 2446 static int 2447 ip6_splithdr(m, exthdrs) 2448 struct mbuf *m; 2449 struct ip6_exthdrs *exthdrs; 2450 { 2451 struct mbuf *mh; 2452 struct ip6_hdr *ip6; 2453 2454 ip6 = mtod(m, struct ip6_hdr *); 2455 if (m->m_len > sizeof(*ip6)) { 2456 MGETHDR(mh, M_DONTWAIT, MT_HEADER); 2457 if (mh == 0) { 2458 m_freem(m); 2459 return ENOBUFS; 2460 } 2461 M_MOVE_PKTHDR(mh, m); 2462 MH_ALIGN(mh, sizeof(*ip6)); 2463 m->m_len -= sizeof(*ip6); 2464 m->m_data += sizeof(*ip6); 2465 mh->m_next = m; 2466 m = mh; 2467 m->m_len = sizeof(*ip6); 2468 bcopy((caddr_t)ip6, mtod(m, caddr_t), sizeof(*ip6)); 2469 } 2470 exthdrs->ip6e_ip6 = m; 2471 return 0; 2472 } 2473 2474 /* 2475 * Compute IPv6 extension header length. 2476 */ 2477 int 2478 ip6_optlen(in6p) 2479 struct in6pcb *in6p; 2480 { 2481 int len; 2482 2483 if (!in6p->in6p_outputopts) 2484 return 0; 2485 2486 len = 0; 2487 #define elen(x) \ 2488 (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0) 2489 2490 len += elen(in6p->in6p_outputopts->ip6po_hbh); 2491 len += elen(in6p->in6p_outputopts->ip6po_dest1); 2492 len += elen(in6p->in6p_outputopts->ip6po_rthdr); 2493 len += elen(in6p->in6p_outputopts->ip6po_dest2); 2494 return len; 2495 #undef elen 2496 } 2497
Indexes created Tue Sep 30 20:09:53 GMT 2025