1 /* $NetBSD: ip6_output.c,v 1.93 2005/12/11 12:25:02 christos Exp $ */ 2 /* $KAME: ip6_output.c,v 1.172 2001/03/25 09:55:56 itojun Exp $ */ 3 4 /* 5 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project. 6 * All rights reserved. 7 * 8 * Redistribution and use in source and binary forms, with or without 9 * modification, are permitted provided that the following conditions 10 * are met: 11 * 1. Redistributions of source code must retain the above copyright 12 * notice, this list of conditions and the following disclaimer. 13 * 2. Redistributions in binary form must reproduce the above copyright 14 * notice, this list of conditions and the following disclaimer in the 15 * documentation and/or other materials provided with the distribution. 16 * 3. Neither the name of the project nor the names of its contributors 17 * may be used to endorse or promote products derived from this software 18 * without specific prior written permission. 19 * 20 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND 21 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 22 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 23 * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE 24 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 25 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 26 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 27 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 28 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 29 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 30 * SUCH DAMAGE. 31 */ 32 33 /* 34 * Copyright (c) 1982, 1986, 1988, 1990, 1993 35 * The Regents of the University of California. All rights reserved. 36 * 37 * Redistribution and use in source and binary forms, with or without 38 * modification, are permitted provided that the following conditions 39 * are met: 40 * 1. Redistributions of source code must retain the above copyright 41 * notice, this list of conditions and the following disclaimer. 42 * 2. Redistributions in binary form must reproduce the above copyright 43 * notice, this list of conditions and the following disclaimer in the 44 * documentation and/or other materials provided with the distribution. 45 * 3. Neither the name of the University nor the names of its contributors 46 * may be used to endorse or promote products derived from this software 47 * without specific prior written permission. 48 * 49 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 50 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 51 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 52 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 53 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 54 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 55 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 56 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 57 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 58 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 59 * SUCH DAMAGE. 60 * 61 * @(#)ip_output.c 8.3 (Berkeley) 1/21/94 62 */ 63 64 #include <sys/cdefs.h> 65 __KERNEL_RCSID(0, "$NetBSD: ip6_output.c,v 1.93 2005/12/11 12:25:02 christos 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 97 #ifdef IPSEC 98 #include <netinet6/ipsec.h> 99 #include <netkey/key.h> 100 #endif /* IPSEC */ 101 102 #include <net/net_osdep.h> 103 104 #ifdef PFIL_HOOKS 105 extern struct pfil_head inet6_pfil_hook; /* XXX */ 106 #endif 107 108 struct ip6_exthdrs { 109 struct mbuf *ip6e_ip6; 110 struct mbuf *ip6e_hbh; 111 struct mbuf *ip6e_dest1; 112 struct mbuf *ip6e_rthdr; 113 struct mbuf *ip6e_dest2; 114 }; 115 116 static int ip6_pcbopts __P((struct ip6_pktopts **, struct mbuf *, 117 struct socket *)); 118 static int ip6_setmoptions __P((int, struct ip6_moptions **, struct mbuf *)); 119 static int ip6_getmoptions __P((int, struct ip6_moptions *, struct mbuf **)); 120 static int ip6_copyexthdr __P((struct mbuf **, caddr_t, int)); 121 static int ip6_insertfraghdr __P((struct mbuf *, struct mbuf *, int, 122 struct ip6_frag **)); 123 static int ip6_insert_jumboopt __P((struct ip6_exthdrs *, u_int32_t)); 124 static int ip6_splithdr __P((struct mbuf *, struct ip6_exthdrs *)); 125 static int ip6_getpmtu __P((struct route_in6 *, struct route_in6 *, 126 struct ifnet *, struct in6_addr *, u_long *, int *)); 127 128 #define IN6_NEED_CHECKSUM(ifp, csum_flags) \ 129 (__predict_true(((ifp)->if_flags & IFF_LOOPBACK) == 0 || \ 130 (((csum_flags) & M_CSUM_UDPv6) != 0 && udp_do_loopback_cksum) || \ 131 (((csum_flags) & M_CSUM_TCPv6) != 0 && tcp_do_loopback_cksum))) 132 133 /* 134 * IP6 output. The packet in mbuf chain m contains a skeletal IP6 135 * header (with pri, len, nxt, hlim, src, dst). 136 * This function may modify ver and hlim only. 137 * The mbuf chain containing the packet will be freed. 138 * The mbuf opt, if present, will not be freed. 139 * 140 * type of "mtu": rt_rmx.rmx_mtu is u_long, ifnet.ifr_mtu is int, and 141 * nd_ifinfo.linkmtu is u_int32_t. so we use u_long to hold largest one, 142 * which is rt_rmx.rmx_mtu. 143 */ 144 int 145 ip6_output(m0, opt, ro, flags, im6o, so, ifpp) 146 struct mbuf *m0; 147 struct ip6_pktopts *opt; 148 struct route_in6 *ro; 149 int flags; 150 struct ip6_moptions *im6o; 151 struct socket *so; 152 struct ifnet **ifpp; /* XXX: just for statistics */ 153 { 154 struct ip6_hdr *ip6, *mhip6; 155 struct ifnet *ifp, *origifp; 156 struct mbuf *m = m0; 157 int hlen, tlen, len, off; 158 struct route_in6 ip6route; 159 struct sockaddr_in6 *dst; 160 int error = 0; 161 u_long mtu; 162 int alwaysfrag, dontfrag; 163 u_int32_t optlen = 0, plen = 0, unfragpartlen = 0; 164 struct ip6_exthdrs exthdrs; 165 struct in6_addr finaldst; 166 struct route_in6 *ro_pmtu = NULL; 167 int hdrsplit = 0; 168 int needipsec = 0; 169 #ifdef IPSEC 170 int needipsectun = 0; 171 struct secpolicy *sp = NULL; 172 173 ip6 = mtod(m, struct ip6_hdr *); 174 #endif /* IPSEC */ 175 176 M_CSUM_DATA_IPv6_HL_SET(m->m_pkthdr.csum_data, sizeof(struct ip6_hdr)); 177 178 #define MAKE_EXTHDR(hp, mp) \ 179 do { \ 180 if (hp) { \ 181 struct ip6_ext *eh = (struct ip6_ext *)(hp); \ 182 error = ip6_copyexthdr((mp), (caddr_t)(hp), \ 183 ((eh)->ip6e_len + 1) << 3); \ 184 if (error) \ 185 goto freehdrs; \ 186 } \ 187 } while (/*CONSTCOND*/ 0) 188 189 bzero(&exthdrs, sizeof(exthdrs)); 190 if (opt) { 191 /* Hop-by-Hop options header */ 192 MAKE_EXTHDR(opt->ip6po_hbh, &exthdrs.ip6e_hbh); 193 /* Destination options header(1st part) */ 194 MAKE_EXTHDR(opt->ip6po_dest1, &exthdrs.ip6e_dest1); 195 /* Routing header */ 196 MAKE_EXTHDR(opt->ip6po_rthdr, &exthdrs.ip6e_rthdr); 197 /* Destination options header(2nd part) */ 198 MAKE_EXTHDR(opt->ip6po_dest2, &exthdrs.ip6e_dest2); 199 } 200 201 #ifdef IPSEC 202 if ((flags & IPV6_FORWARDING) != 0) { 203 needipsec = 0; 204 goto skippolicycheck; 205 } 206 207 /* get a security policy for this packet */ 208 if (so == NULL) 209 sp = ipsec6_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND, 0, &error); 210 else { 211 if (IPSEC_PCB_SKIP_IPSEC(sotoinpcb_hdr(so)->inph_sp, 212 IPSEC_DIR_OUTBOUND)) { 213 needipsec = 0; 214 goto skippolicycheck; 215 } 216 sp = ipsec6_getpolicybysock(m, IPSEC_DIR_OUTBOUND, so, &error); 217 } 218 219 if (sp == NULL) { 220 ipsec6stat.out_inval++; 221 goto freehdrs; 222 } 223 224 error = 0; 225 226 /* check policy */ 227 switch (sp->policy) { 228 case IPSEC_POLICY_DISCARD: 229 /* 230 * This packet is just discarded. 231 */ 232 ipsec6stat.out_polvio++; 233 goto freehdrs; 234 235 case IPSEC_POLICY_BYPASS: 236 case IPSEC_POLICY_NONE: 237 /* no need to do IPsec. */ 238 needipsec = 0; 239 break; 240 241 case IPSEC_POLICY_IPSEC: 242 if (sp->req == NULL) { 243 /* XXX should be panic ? */ 244 printf("ip6_output: No IPsec request specified.\n"); 245 error = EINVAL; 246 goto freehdrs; 247 } 248 needipsec = 1; 249 break; 250 251 case IPSEC_POLICY_ENTRUST: 252 default: 253 printf("ip6_output: Invalid policy found. %d\n", sp->policy); 254 } 255 256 skippolicycheck:; 257 #endif /* IPSEC */ 258 259 if (needipsec && 260 (m->m_pkthdr.csum_flags & (M_CSUM_UDPv6|M_CSUM_TCPv6)) != 0) { 261 in6_delayed_cksum(m); 262 m->m_pkthdr.csum_flags &= ~(M_CSUM_UDPv6|M_CSUM_TCPv6); 263 } 264 265 /* 266 * Calculate the total length of the extension header chain. 267 * Keep the length of the unfragmentable part for fragmentation. 268 */ 269 optlen = 0; 270 if (exthdrs.ip6e_hbh) optlen += exthdrs.ip6e_hbh->m_len; 271 if (exthdrs.ip6e_dest1) optlen += exthdrs.ip6e_dest1->m_len; 272 if (exthdrs.ip6e_rthdr) optlen += exthdrs.ip6e_rthdr->m_len; 273 unfragpartlen = optlen + sizeof(struct ip6_hdr); 274 /* NOTE: we don't add AH/ESP length here. do that later. */ 275 if (exthdrs.ip6e_dest2) optlen += exthdrs.ip6e_dest2->m_len; 276 277 /* 278 * If we need IPsec, or there is at least one extension header, 279 * separate IP6 header from the payload. 280 */ 281 if ((needipsec || optlen) && !hdrsplit) { 282 if ((error = ip6_splithdr(m, &exthdrs)) != 0) { 283 m = NULL; 284 goto freehdrs; 285 } 286 m = exthdrs.ip6e_ip6; 287 hdrsplit++; 288 } 289 290 /* adjust pointer */ 291 ip6 = mtod(m, struct ip6_hdr *); 292 293 /* adjust mbuf packet header length */ 294 m->m_pkthdr.len += optlen; 295 plen = m->m_pkthdr.len - sizeof(*ip6); 296 297 /* If this is a jumbo payload, insert a jumbo payload option. */ 298 if (plen > IPV6_MAXPACKET) { 299 if (!hdrsplit) { 300 if ((error = ip6_splithdr(m, &exthdrs)) != 0) { 301 m = NULL; 302 goto freehdrs; 303 } 304 m = exthdrs.ip6e_ip6; 305 hdrsplit++; 306 } 307 /* adjust pointer */ 308 ip6 = mtod(m, struct ip6_hdr *); 309 if ((error = ip6_insert_jumboopt(&exthdrs, plen)) != 0) 310 goto freehdrs; 311 optlen += 8; /* XXX JUMBOOPTLEN */ 312 ip6->ip6_plen = 0; 313 } else 314 ip6->ip6_plen = htons(plen); 315 316 /* 317 * Concatenate headers and fill in next header fields. 318 * Here we have, on "m" 319 * IPv6 payload 320 * and we insert headers accordingly. Finally, we should be getting: 321 * IPv6 hbh dest1 rthdr ah* [esp* dest2 payload] 322 * 323 * during the header composing process, "m" points to IPv6 header. 324 * "mprev" points to an extension header prior to esp. 325 */ 326 { 327 u_char *nexthdrp = &ip6->ip6_nxt; 328 struct mbuf *mprev = m; 329 330 /* 331 * we treat dest2 specially. this makes IPsec processing 332 * much easier. the goal here is to make mprev point the 333 * mbuf prior to dest2. 334 * 335 * result: IPv6 dest2 payload 336 * m and mprev will point to IPv6 header. 337 */ 338 if (exthdrs.ip6e_dest2) { 339 if (!hdrsplit) 340 panic("assumption failed: hdr not split"); 341 exthdrs.ip6e_dest2->m_next = m->m_next; 342 m->m_next = exthdrs.ip6e_dest2; 343 *mtod(exthdrs.ip6e_dest2, u_char *) = ip6->ip6_nxt; 344 ip6->ip6_nxt = IPPROTO_DSTOPTS; 345 } 346 347 #define MAKE_CHAIN(m, mp, p, i)\ 348 do {\ 349 if (m) {\ 350 if (!hdrsplit) \ 351 panic("assumption failed: hdr not split"); \ 352 *mtod((m), u_char *) = *(p);\ 353 *(p) = (i);\ 354 p = mtod((m), u_char *);\ 355 (m)->m_next = (mp)->m_next;\ 356 (mp)->m_next = (m);\ 357 (mp) = (m);\ 358 }\ 359 } while (/*CONSTCOND*/ 0) 360 /* 361 * result: IPv6 hbh dest1 rthdr dest2 payload 362 * m will point to IPv6 header. mprev will point to the 363 * extension header prior to dest2 (rthdr in the above case). 364 */ 365 MAKE_CHAIN(exthdrs.ip6e_hbh, mprev, nexthdrp, IPPROTO_HOPOPTS); 366 MAKE_CHAIN(exthdrs.ip6e_dest1, mprev, nexthdrp, 367 IPPROTO_DSTOPTS); 368 MAKE_CHAIN(exthdrs.ip6e_rthdr, mprev, nexthdrp, 369 IPPROTO_ROUTING); 370 371 M_CSUM_DATA_IPv6_HL_SET(m->m_pkthdr.csum_data, 372 sizeof(struct ip6_hdr) + optlen); 373 374 #ifdef IPSEC 375 if (!needipsec) 376 goto skip_ipsec2; 377 378 /* 379 * pointers after IPsec headers are not valid any more. 380 * other pointers need a great care too. 381 * (IPsec routines should not mangle mbufs prior to AH/ESP) 382 */ 383 exthdrs.ip6e_dest2 = NULL; 384 385 { 386 struct ip6_rthdr *rh = NULL; 387 int segleft_org = 0; 388 struct ipsec_output_state state; 389 390 if (exthdrs.ip6e_rthdr) { 391 rh = mtod(exthdrs.ip6e_rthdr, struct ip6_rthdr *); 392 segleft_org = rh->ip6r_segleft; 393 rh->ip6r_segleft = 0; 394 } 395 396 bzero(&state, sizeof(state)); 397 state.m = m; 398 error = ipsec6_output_trans(&state, nexthdrp, mprev, sp, flags, 399 &needipsectun); 400 m = state.m; 401 if (error) { 402 /* mbuf is already reclaimed in ipsec6_output_trans. */ 403 m = NULL; 404 switch (error) { 405 case EHOSTUNREACH: 406 case ENETUNREACH: 407 case EMSGSIZE: 408 case ENOBUFS: 409 case ENOMEM: 410 break; 411 default: 412 printf("ip6_output (ipsec): error code %d\n", error); 413 /* FALLTHROUGH */ 414 case ENOENT: 415 /* don't show these error codes to the user */ 416 error = 0; 417 break; 418 } 419 goto bad; 420 } 421 if (exthdrs.ip6e_rthdr) { 422 /* ah6_output doesn't modify mbuf chain */ 423 rh->ip6r_segleft = segleft_org; 424 } 425 } 426 skip_ipsec2:; 427 #endif 428 } 429 430 /* 431 * If there is a routing header, replace destination address field 432 * with the first hop of the routing header. 433 */ 434 if (exthdrs.ip6e_rthdr) { 435 struct ip6_rthdr *rh; 436 struct ip6_rthdr0 *rh0; 437 struct in6_addr *addr; 438 439 rh = (struct ip6_rthdr *)(mtod(exthdrs.ip6e_rthdr, 440 struct ip6_rthdr *)); 441 finaldst = ip6->ip6_dst; 442 switch (rh->ip6r_type) { 443 case IPV6_RTHDR_TYPE_0: 444 rh0 = (struct ip6_rthdr0 *)rh; 445 addr = (struct in6_addr *)(rh0 + 1); 446 ip6->ip6_dst = addr[0]; 447 (void)memmove(&addr[0], &addr[1], 448 sizeof(struct in6_addr) * (rh0->ip6r0_segleft - 1)); 449 addr[rh0->ip6r0_segleft - 1] = finaldst; 450 break; 451 default: /* is it possible? */ 452 error = EINVAL; 453 goto bad; 454 } 455 } 456 457 /* Source address validation */ 458 if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) && 459 (flags & IPV6_UNSPECSRC) == 0) { 460 error = EOPNOTSUPP; 461 ip6stat.ip6s_badscope++; 462 goto bad; 463 } 464 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) { 465 error = EOPNOTSUPP; 466 ip6stat.ip6s_badscope++; 467 goto bad; 468 } 469 470 ip6stat.ip6s_localout++; 471 472 /* 473 * Route packet. 474 */ 475 /* initialize cached route */ 476 if (ro == 0) { 477 ro = &ip6route; 478 bzero((caddr_t)ro, sizeof(*ro)); 479 } 480 ro_pmtu = ro; 481 if (opt && opt->ip6po_rthdr) 482 ro = &opt->ip6po_route; 483 dst = (struct sockaddr_in6 *)&ro->ro_dst; 484 /* 485 * If there is a cached route, 486 * check that it is to the same destination 487 * and is still up. If not, free it and try again. 488 */ 489 if (ro->ro_rt && ((ro->ro_rt->rt_flags & RTF_UP) == 0 || 490 dst->sin6_family != AF_INET6 || 491 !IN6_ARE_ADDR_EQUAL(&dst->sin6_addr, &ip6->ip6_dst))) { 492 RTFREE(ro->ro_rt); 493 ro->ro_rt = (struct rtentry *)0; 494 } 495 if (ro->ro_rt == 0) { 496 bzero(dst, sizeof(*dst)); 497 dst->sin6_family = AF_INET6; 498 dst->sin6_len = sizeof(struct sockaddr_in6); 499 dst->sin6_addr = ip6->ip6_dst; 500 } 501 #ifdef IPSEC 502 if (needipsec && needipsectun) { 503 struct ipsec_output_state state; 504 505 /* 506 * All the extension headers will become inaccessible 507 * (since they can be encrypted). 508 * Don't panic, we need no more updates to extension headers 509 * on inner IPv6 packet (since they are now encapsulated). 510 * 511 * IPv6 [ESP|AH] IPv6 [extension headers] payload 512 */ 513 bzero(&exthdrs, sizeof(exthdrs)); 514 exthdrs.ip6e_ip6 = m; 515 516 bzero(&state, sizeof(state)); 517 state.m = m; 518 state.ro = (struct route *)ro; 519 state.dst = (struct sockaddr *)dst; 520 521 error = ipsec6_output_tunnel(&state, sp, flags); 522 523 m = state.m; 524 ro_pmtu = ro = (struct route_in6 *)state.ro; 525 dst = (struct sockaddr_in6 *)state.dst; 526 if (error) { 527 /* mbuf is already reclaimed in ipsec6_output_tunnel. */ 528 m0 = m = NULL; 529 m = NULL; 530 switch (error) { 531 case EHOSTUNREACH: 532 case ENETUNREACH: 533 case EMSGSIZE: 534 case ENOBUFS: 535 case ENOMEM: 536 break; 537 default: 538 printf("ip6_output (ipsec): error code %d\n", error); 539 /* FALLTHROUGH */ 540 case ENOENT: 541 /* don't show these error codes to the user */ 542 error = 0; 543 break; 544 } 545 goto bad; 546 } 547 548 exthdrs.ip6e_ip6 = m; 549 } 550 #endif /* IPSEC */ 551 552 if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) { 553 /* Unicast */ 554 555 #define ifatoia6(ifa) ((struct in6_ifaddr *)(ifa)) 556 #define sin6tosa(sin6) ((struct sockaddr *)(sin6)) 557 /* xxx 558 * interface selection comes here 559 * if an interface is specified from an upper layer, 560 * ifp must point it. 561 */ 562 if (ro->ro_rt == 0) { 563 /* 564 * non-bsdi always clone routes, if parent is 565 * PRF_CLONING. 566 */ 567 rtalloc((struct route *)ro); 568 } 569 if (ro->ro_rt == 0) { 570 ip6stat.ip6s_noroute++; 571 error = EHOSTUNREACH; 572 /* XXX in6_ifstat_inc(ifp, ifs6_out_discard); */ 573 goto bad; 574 } 575 ifp = ro->ro_rt->rt_ifp; 576 ro->ro_rt->rt_use++; 577 if (ro->ro_rt->rt_flags & RTF_GATEWAY) 578 dst = (struct sockaddr_in6 *)ro->ro_rt->rt_gateway; 579 m->m_flags &= ~(M_BCAST | M_MCAST); /* just in case */ 580 581 in6_ifstat_inc(ifp, ifs6_out_request); 582 583 /* 584 * Check if the outgoing interface conflicts with 585 * the interface specified by ifi6_ifindex (if specified). 586 * Note that loopback interface is always okay. 587 * (this may happen when we are sending a packet to one of 588 * our own addresses.) 589 */ 590 if (opt && opt->ip6po_pktinfo && 591 opt->ip6po_pktinfo->ipi6_ifindex) { 592 if (!(ifp->if_flags & IFF_LOOPBACK) && 593 ifp->if_index != opt->ip6po_pktinfo->ipi6_ifindex) { 594 ip6stat.ip6s_noroute++; 595 in6_ifstat_inc(ifp, ifs6_out_discard); 596 error = EHOSTUNREACH; 597 goto bad; 598 } 599 } 600 601 if (opt && opt->ip6po_hlim != -1) 602 ip6->ip6_hlim = opt->ip6po_hlim & 0xff; 603 } else { 604 /* Multicast */ 605 struct in6_multi *in6m; 606 607 m->m_flags = (m->m_flags & ~M_BCAST) | M_MCAST; 608 609 /* 610 * See if the caller provided any multicast options 611 */ 612 ifp = NULL; 613 if (im6o != NULL) { 614 ip6->ip6_hlim = im6o->im6o_multicast_hlim; 615 if (im6o->im6o_multicast_ifp != NULL) 616 ifp = im6o->im6o_multicast_ifp; 617 } else 618 ip6->ip6_hlim = ip6_defmcasthlim; 619 620 /* 621 * See if the caller provided the outgoing interface 622 * as an ancillary data. 623 * Boundary check for ifindex is assumed to be already done. 624 */ 625 if (opt && opt->ip6po_pktinfo && opt->ip6po_pktinfo->ipi6_ifindex) 626 ifp = ifindex2ifnet[opt->ip6po_pktinfo->ipi6_ifindex]; 627 628 /* 629 * If the destination is a node-local scope multicast, 630 * the packet should be loop-backed only. 631 */ 632 if (IN6_IS_ADDR_MC_NODELOCAL(&ip6->ip6_dst)) { 633 /* 634 * If the outgoing interface is already specified, 635 * it should be a loopback interface. 636 */ 637 if (ifp && (ifp->if_flags & IFF_LOOPBACK) == 0) { 638 ip6stat.ip6s_badscope++; 639 error = ENETUNREACH; /* XXX: better error? */ 640 /* XXX correct ifp? */ 641 in6_ifstat_inc(ifp, ifs6_out_discard); 642 goto bad; 643 } else 644 ifp = lo0ifp; 645 } 646 647 if (opt && opt->ip6po_hlim != -1) 648 ip6->ip6_hlim = opt->ip6po_hlim & 0xff; 649 650 /* 651 * If caller did not provide an interface lookup a 652 * default in the routing table. This is either a 653 * default for the speicfied group (i.e. a host 654 * route), or a multicast default (a route for the 655 * ``net'' ff00::/8). 656 */ 657 if (ifp == NULL) { 658 if (ro->ro_rt == 0) { 659 ro->ro_rt = rtalloc1((struct sockaddr *) 660 &ro->ro_dst, 0); 661 } 662 if (ro->ro_rt == 0) { 663 ip6stat.ip6s_noroute++; 664 error = EHOSTUNREACH; 665 /* XXX in6_ifstat_inc(ifp, ifs6_out_discard) */ 666 goto bad; 667 } 668 ifp = ro->ro_rt->rt_ifp; 669 ro->ro_rt->rt_use++; 670 } 671 672 if ((flags & IPV6_FORWARDING) == 0) 673 in6_ifstat_inc(ifp, ifs6_out_request); 674 in6_ifstat_inc(ifp, ifs6_out_mcast); 675 676 /* 677 * Confirm that the outgoing interface supports multicast. 678 */ 679 if ((ifp->if_flags & IFF_MULTICAST) == 0) { 680 ip6stat.ip6s_noroute++; 681 in6_ifstat_inc(ifp, ifs6_out_discard); 682 error = ENETUNREACH; 683 goto bad; 684 } 685 IN6_LOOKUP_MULTI(ip6->ip6_dst, ifp, in6m); 686 if (in6m != NULL && 687 (im6o == NULL || im6o->im6o_multicast_loop)) { 688 /* 689 * If we belong to the destination multicast group 690 * on the outgoing interface, and the caller did not 691 * forbid loopback, loop back a copy. 692 */ 693 ip6_mloopback(ifp, m, dst); 694 } else { 695 /* 696 * If we are acting as a multicast router, perform 697 * multicast forwarding as if the packet had just 698 * arrived on the interface to which we are about 699 * to send. The multicast forwarding function 700 * recursively calls this function, using the 701 * IPV6_FORWARDING flag to prevent infinite recursion. 702 * 703 * Multicasts that are looped back by ip6_mloopback(), 704 * above, will be forwarded by the ip6_input() routine, 705 * if necessary. 706 */ 707 if (ip6_mrouter && (flags & IPV6_FORWARDING) == 0) { 708 if (ip6_mforward(ip6, ifp, m) != 0) { 709 m_freem(m); 710 goto done; 711 } 712 } 713 } 714 /* 715 * Multicasts with a hoplimit of zero may be looped back, 716 * above, but must not be transmitted on a network. 717 * Also, multicasts addressed to the loopback interface 718 * are not sent -- the above call to ip6_mloopback() will 719 * loop back a copy if this host actually belongs to the 720 * destination group on the loopback interface. 721 */ 722 if (ip6->ip6_hlim == 0 || (ifp->if_flags & IFF_LOOPBACK)) { 723 m_freem(m); 724 goto done; 725 } 726 } 727 728 /* 729 * Fill the outgoing inteface to tell the upper layer 730 * to increment per-interface statistics. 731 */ 732 if (ifpp) 733 *ifpp = ifp; 734 735 /* Determine path MTU. */ 736 if ((error = ip6_getpmtu(ro_pmtu, ro, ifp, &finaldst, &mtu, 737 &alwaysfrag)) != 0) 738 goto bad; 739 #ifdef IPSEC 740 if (needipsectun) 741 mtu = IPV6_MMTU; 742 #endif 743 744 /* 745 * The caller of this function may specify to use the minimum MTU 746 * in some cases. 747 */ 748 if (mtu > IPV6_MMTU) { 749 if ((flags & IPV6_MINMTU)) 750 mtu = IPV6_MMTU; 751 } 752 753 /* Fake scoped addresses */ 754 if ((ifp->if_flags & IFF_LOOPBACK) != 0) { 755 /* 756 * If source or destination address is a scoped address, and 757 * the packet is going to be sent to a loopback interface, 758 * we should keep the original interface. 759 */ 760 761 /* 762 * XXX: this is a very experimental and temporary solution. 763 * We eventually have sockaddr_in6 and use the sin6_scope_id 764 * field of the structure here. 765 * We rely on the consistency between two scope zone ids 766 * of source add destination, which should already be assured 767 * Larger scopes than link will be supported in the near 768 * future. 769 */ 770 origifp = NULL; 771 if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src)) 772 origifp = ifindex2ifnet[ntohs(ip6->ip6_src.s6_addr16[1])]; 773 else if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst)) 774 origifp = ifindex2ifnet[ntohs(ip6->ip6_dst.s6_addr16[1])]; 775 /* 776 * XXX: origifp can be NULL even in those two cases above. 777 * For example, if we remove the (only) link-local address 778 * from the loopback interface, and try to send a link-local 779 * address without link-id information. Then the source 780 * address is ::1, and the destination address is the 781 * link-local address with its s6_addr16[1] being zero. 782 * What is worse, if the packet goes to the loopback interface 783 * by a default rejected route, the null pointer would be 784 * passed to looutput, and the kernel would hang. 785 * The following last resort would prevent such disaster. 786 */ 787 if (origifp == NULL) 788 origifp = ifp; 789 } else 790 origifp = ifp; 791 if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src)) 792 ip6->ip6_src.s6_addr16[1] = 0; 793 if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst)) 794 ip6->ip6_dst.s6_addr16[1] = 0; 795 796 /* 797 * If the outgoing packet contains a hop-by-hop options header, 798 * it must be examined and processed even by the source node. 799 * (RFC 2460, section 4.) 800 */ 801 if (exthdrs.ip6e_hbh) { 802 struct ip6_hbh *hbh = mtod(exthdrs.ip6e_hbh, struct ip6_hbh *); 803 u_int32_t dummy1; /* XXX unused */ 804 u_int32_t dummy2; /* XXX unused */ 805 806 /* 807 * XXX: if we have to send an ICMPv6 error to the sender, 808 * we need the M_LOOP flag since icmp6_error() expects 809 * the IPv6 and the hop-by-hop options header are 810 * continuous unless the flag is set. 811 */ 812 m->m_flags |= M_LOOP; 813 m->m_pkthdr.rcvif = ifp; 814 if (ip6_process_hopopts(m, (u_int8_t *)(hbh + 1), 815 ((hbh->ip6h_len + 1) << 3) - sizeof(struct ip6_hbh), 816 &dummy1, &dummy2) < 0) { 817 /* m was already freed at this point */ 818 error = EINVAL;/* better error? */ 819 goto done; 820 } 821 m->m_flags &= ~M_LOOP; /* XXX */ 822 m->m_pkthdr.rcvif = NULL; 823 } 824 825 #ifdef PFIL_HOOKS 826 /* 827 * Run through list of hooks for output packets. 828 */ 829 if ((error = pfil_run_hooks(&inet6_pfil_hook, &m, ifp, PFIL_OUT)) != 0) 830 goto done; 831 if (m == NULL) 832 goto done; 833 ip6 = mtod(m, struct ip6_hdr *); 834 #endif /* PFIL_HOOKS */ 835 /* 836 * Send the packet to the outgoing interface. 837 * If necessary, do IPv6 fragmentation before sending. 838 * 839 * the logic here is rather complex: 840 * 1: normal case (dontfrag == 0, alwaysfrag == 0) 841 * 1-a: send as is if tlen <= path mtu 842 * 1-b: fragment if tlen > path mtu 843 * 844 * 2: if user asks us not to fragment (dontfrag == 1) 845 * 2-a: send as is if tlen <= interface mtu 846 * 2-b: error if tlen > interface mtu 847 * 848 * 3: if we always need to attach fragment header (alwaysfrag == 1) 849 * always fragment 850 * 851 * 4: if dontfrag == 1 && alwaysfrag == 1 852 * error, as we cannot handle this conflicting request 853 */ 854 tlen = m->m_pkthdr.len; 855 856 dontfrag = 0; 857 if (dontfrag && alwaysfrag) { /* case 4 */ 858 /* conflicting request - can't transmit */ 859 error = EMSGSIZE; 860 goto bad; 861 } 862 if (dontfrag && tlen > IN6_LINKMTU(ifp)) { /* case 2-b */ 863 /* 864 * Even if the DONTFRAG option is specified, we cannot send the 865 * packet when the data length is larger than the MTU of the 866 * outgoing interface. 867 * Notify the error by sending IPV6_PATHMTU ancillary data as 868 * well as returning an error code (the latter is not described 869 * in the API spec.) 870 */ 871 u_int32_t mtu32; 872 struct ip6ctlparam ip6cp; 873 874 mtu32 = (u_int32_t)mtu; 875 bzero(&ip6cp, sizeof(ip6cp)); 876 ip6cp.ip6c_cmdarg = (void *)&mtu32; 877 pfctlinput2(PRC_MSGSIZE, (struct sockaddr *)&ro_pmtu->ro_dst, 878 (void *)&ip6cp); 879 880 error = EMSGSIZE; 881 goto bad; 882 } 883 884 /* 885 * transmit packet without fragmentation 886 */ 887 if (dontfrag || (!alwaysfrag && tlen <= mtu)) { /* case 1-a and 2-a */ 888 struct in6_ifaddr *ia6; 889 int sw_csum; 890 891 ip6 = mtod(m, struct ip6_hdr *); 892 ia6 = in6_ifawithifp(ifp, &ip6->ip6_src); 893 if (ia6) { 894 /* Record statistics for this interface address. */ 895 ia6->ia_ifa.ifa_data.ifad_outbytes += m->m_pkthdr.len; 896 } 897 #ifdef IPSEC 898 /* clean ipsec history once it goes out of the node */ 899 ipsec_delaux(m); 900 #endif 901 902 sw_csum = m->m_pkthdr.csum_flags & ~ifp->if_csum_flags_tx; 903 if ((sw_csum & (M_CSUM_UDPv6|M_CSUM_TCPv6)) != 0) { 904 if (IN6_NEED_CHECKSUM(ifp, 905 sw_csum & (M_CSUM_UDPv6|M_CSUM_TCPv6))) { 906 in6_delayed_cksum(m); 907 } 908 m->m_pkthdr.csum_flags &= ~(M_CSUM_UDPv6|M_CSUM_TCPv6); 909 } 910 911 error = nd6_output(ifp, origifp, m, dst, ro->ro_rt); 912 goto done; 913 } 914 915 /* 916 * try to fragment the packet. case 1-b and 3 917 */ 918 if (mtu < IPV6_MMTU) { 919 /* path MTU cannot be less than IPV6_MMTU */ 920 error = EMSGSIZE; 921 in6_ifstat_inc(ifp, ifs6_out_fragfail); 922 goto bad; 923 } else if (ip6->ip6_plen == 0) { 924 /* jumbo payload cannot be fragmented */ 925 error = EMSGSIZE; 926 in6_ifstat_inc(ifp, ifs6_out_fragfail); 927 goto bad; 928 } else { 929 struct mbuf **mnext, *m_frgpart; 930 struct ip6_frag *ip6f; 931 u_int32_t id = htonl(ip6_randomid()); 932 u_char nextproto; 933 struct ip6ctlparam ip6cp; 934 u_int32_t mtu32; 935 936 /* 937 * Too large for the destination or interface; 938 * fragment if possible. 939 * Must be able to put at least 8 bytes per fragment. 940 */ 941 hlen = unfragpartlen; 942 if (mtu > IPV6_MAXPACKET) 943 mtu = IPV6_MAXPACKET; 944 945 /* Notify a proper path MTU to applications. */ 946 mtu32 = (u_int32_t)mtu; 947 bzero(&ip6cp, sizeof(ip6cp)); 948 ip6cp.ip6c_cmdarg = (void *)&mtu32; 949 pfctlinput2(PRC_MSGSIZE, (struct sockaddr *)&ro_pmtu->ro_dst, 950 (void *)&ip6cp); 951 952 len = (mtu - hlen - sizeof(struct ip6_frag)) & ~7; 953 if (len < 8) { 954 error = EMSGSIZE; 955 in6_ifstat_inc(ifp, ifs6_out_fragfail); 956 goto bad; 957 } 958 959 mnext = &m->m_nextpkt; 960 961 /* 962 * Change the next header field of the last header in the 963 * unfragmentable part. 964 */ 965 if (exthdrs.ip6e_rthdr) { 966 nextproto = *mtod(exthdrs.ip6e_rthdr, u_char *); 967 *mtod(exthdrs.ip6e_rthdr, u_char *) = IPPROTO_FRAGMENT; 968 } else if (exthdrs.ip6e_dest1) { 969 nextproto = *mtod(exthdrs.ip6e_dest1, u_char *); 970 *mtod(exthdrs.ip6e_dest1, u_char *) = IPPROTO_FRAGMENT; 971 } else if (exthdrs.ip6e_hbh) { 972 nextproto = *mtod(exthdrs.ip6e_hbh, u_char *); 973 *mtod(exthdrs.ip6e_hbh, u_char *) = IPPROTO_FRAGMENT; 974 } else { 975 nextproto = ip6->ip6_nxt; 976 ip6->ip6_nxt = IPPROTO_FRAGMENT; 977 } 978 979 if ((m->m_pkthdr.csum_flags & (M_CSUM_UDPv6|M_CSUM_TCPv6)) 980 != 0) { 981 if (IN6_NEED_CHECKSUM(ifp, 982 m->m_pkthdr.csum_flags & 983 (M_CSUM_UDPv6|M_CSUM_TCPv6))) { 984 in6_delayed_cksum(m); 985 } 986 m->m_pkthdr.csum_flags &= ~(M_CSUM_UDPv6|M_CSUM_TCPv6); 987 } 988 989 /* 990 * Loop through length of segment after first fragment, 991 * make new header and copy data of each part and link onto 992 * chain. 993 */ 994 m0 = m; 995 for (off = hlen; off < tlen; off += len) { 996 struct mbuf *mlast; 997 998 MGETHDR(m, M_DONTWAIT, MT_HEADER); 999 if (!m) { 1000 error = ENOBUFS; 1001 ip6stat.ip6s_odropped++; 1002 goto sendorfree; 1003 } 1004 m->m_pkthdr.rcvif = NULL; 1005 m->m_flags = m0->m_flags & M_COPYFLAGS; 1006 *mnext = m; 1007 mnext = &m->m_nextpkt; 1008 m->m_data += max_linkhdr; 1009 mhip6 = mtod(m, struct ip6_hdr *); 1010 *mhip6 = *ip6; 1011 m->m_len = sizeof(*mhip6); 1012 error = ip6_insertfraghdr(m0, m, hlen, &ip6f); 1013 if (error) { 1014 ip6stat.ip6s_odropped++; 1015 goto sendorfree; 1016 } 1017 ip6f->ip6f_offlg = htons((u_int16_t)((off - hlen) & ~7)); 1018 if (off + len >= tlen) 1019 len = tlen - off; 1020 else 1021 ip6f->ip6f_offlg |= IP6F_MORE_FRAG; 1022 mhip6->ip6_plen = htons((u_int16_t)(len + hlen + 1023 sizeof(*ip6f) - sizeof(struct ip6_hdr))); 1024 if ((m_frgpart = m_copy(m0, off, len)) == 0) { 1025 error = ENOBUFS; 1026 ip6stat.ip6s_odropped++; 1027 goto sendorfree; 1028 } 1029 for (mlast = m; mlast->m_next; mlast = mlast->m_next) 1030 ; 1031 mlast->m_next = m_frgpart; 1032 m->m_pkthdr.len = len + hlen + sizeof(*ip6f); 1033 m->m_pkthdr.rcvif = (struct ifnet *)0; 1034 ip6f->ip6f_reserved = 0; 1035 ip6f->ip6f_ident = id; 1036 ip6f->ip6f_nxt = nextproto; 1037 ip6stat.ip6s_ofragments++; 1038 in6_ifstat_inc(ifp, ifs6_out_fragcreat); 1039 } 1040 1041 in6_ifstat_inc(ifp, ifs6_out_fragok); 1042 } 1043 1044 /* 1045 * Remove leading garbages. 1046 */ 1047 sendorfree: 1048 m = m0->m_nextpkt; 1049 m0->m_nextpkt = 0; 1050 m_freem(m0); 1051 for (m0 = m; m; m = m0) { 1052 m0 = m->m_nextpkt; 1053 m->m_nextpkt = 0; 1054 if (error == 0) { 1055 struct in6_ifaddr *ia6; 1056 ip6 = mtod(m, struct ip6_hdr *); 1057 ia6 = in6_ifawithifp(ifp, &ip6->ip6_src); 1058 if (ia6) { 1059 /* 1060 * Record statistics for this interface 1061 * address. 1062 */ 1063 ia6->ia_ifa.ifa_data.ifad_outbytes += 1064 m->m_pkthdr.len; 1065 } 1066 #ifdef IPSEC 1067 /* clean ipsec history once it goes out of the node */ 1068 ipsec_delaux(m); 1069 #endif 1070 error = nd6_output(ifp, origifp, m, dst, ro->ro_rt); 1071 } else 1072 m_freem(m); 1073 } 1074 1075 if (error == 0) 1076 ip6stat.ip6s_fragmented++; 1077 1078 done: 1079 if (ro == &ip6route && ro->ro_rt) { /* brace necessary for RTFREE */ 1080 RTFREE(ro->ro_rt); 1081 } else if (ro_pmtu == &ip6route && ro_pmtu->ro_rt) { 1082 RTFREE(ro_pmtu->ro_rt); 1083 } 1084 1085 #ifdef IPSEC 1086 if (sp != NULL) 1087 key_freesp(sp); 1088 #endif /* IPSEC */ 1089 1090 return (error); 1091 1092 freehdrs: 1093 m_freem(exthdrs.ip6e_hbh); /* m_freem will check if mbuf is 0 */ 1094 m_freem(exthdrs.ip6e_dest1); 1095 m_freem(exthdrs.ip6e_rthdr); 1096 m_freem(exthdrs.ip6e_dest2); 1097 /* FALLTHROUGH */ 1098 bad: 1099 m_freem(m); 1100 goto done; 1101 } 1102 1103 static int 1104 ip6_copyexthdr(mp, hdr, hlen) 1105 struct mbuf **mp; 1106 caddr_t hdr; 1107 int hlen; 1108 { 1109 struct mbuf *m; 1110 1111 if (hlen > MCLBYTES) 1112 return (ENOBUFS); /* XXX */ 1113 1114 MGET(m, M_DONTWAIT, MT_DATA); 1115 if (!m) 1116 return (ENOBUFS); 1117 1118 if (hlen > MLEN) { 1119 MCLGET(m, M_DONTWAIT); 1120 if ((m->m_flags & M_EXT) == 0) { 1121 m_free(m); 1122 return (ENOBUFS); 1123 } 1124 } 1125 m->m_len = hlen; 1126 if (hdr) 1127 bcopy(hdr, mtod(m, caddr_t), hlen); 1128 1129 *mp = m; 1130 return (0); 1131 } 1132 1133 /* 1134 * Process a delayed payload checksum calculation. 1135 */ 1136 void 1137 in6_delayed_cksum(struct mbuf *m) 1138 { 1139 uint16_t csum, offset; 1140 1141 KASSERT((m->m_pkthdr.csum_flags & (M_CSUM_UDPv6|M_CSUM_TCPv6)) != 0); 1142 KASSERT((~m->m_pkthdr.csum_flags & (M_CSUM_UDPv6|M_CSUM_TCPv6)) != 0); 1143 KASSERT((m->m_pkthdr.csum_flags 1144 & (M_CSUM_UDPv4|M_CSUM_TCPv4|M_CSUM_TSOv4)) == 0); 1145 1146 offset = M_CSUM_DATA_IPv6_HL(m->m_pkthdr.csum_data); 1147 csum = in6_cksum(m, 0, offset, m->m_pkthdr.len - offset); 1148 if (csum == 0 && (m->m_pkthdr.csum_flags & M_CSUM_UDPv6) != 0) { 1149 csum = 0xffff; 1150 } 1151 1152 offset += M_CSUM_DATA_IPv6_OFFSET(m->m_pkthdr.csum_data); 1153 if ((offset + sizeof(csum)) > m->m_len) { 1154 m_copyback(m, offset, sizeof(csum), &csum); 1155 } else { 1156 *(uint16_t *)(mtod(m, caddr_t) + offset) = csum; 1157 } 1158 } 1159 1160 /* 1161 * Insert jumbo payload option. 1162 */ 1163 static int 1164 ip6_insert_jumboopt(exthdrs, plen) 1165 struct ip6_exthdrs *exthdrs; 1166 u_int32_t plen; 1167 { 1168 struct mbuf *mopt; 1169 u_int8_t *optbuf; 1170 u_int32_t v; 1171 1172 #define JUMBOOPTLEN 8 /* length of jumbo payload option and padding */ 1173 1174 /* 1175 * If there is no hop-by-hop options header, allocate new one. 1176 * If there is one but it doesn't have enough space to store the 1177 * jumbo payload option, allocate a cluster to store the whole options. 1178 * Otherwise, use it to store the options. 1179 */ 1180 if (exthdrs->ip6e_hbh == 0) { 1181 MGET(mopt, M_DONTWAIT, MT_DATA); 1182 if (mopt == 0) 1183 return (ENOBUFS); 1184 mopt->m_len = JUMBOOPTLEN; 1185 optbuf = mtod(mopt, u_int8_t *); 1186 optbuf[1] = 0; /* = ((JUMBOOPTLEN) >> 3) - 1 */ 1187 exthdrs->ip6e_hbh = mopt; 1188 } else { 1189 struct ip6_hbh *hbh; 1190 1191 mopt = exthdrs->ip6e_hbh; 1192 if (M_TRAILINGSPACE(mopt) < JUMBOOPTLEN) { 1193 /* 1194 * XXX assumption: 1195 * - exthdrs->ip6e_hbh is not referenced from places 1196 * other than exthdrs. 1197 * - exthdrs->ip6e_hbh is not an mbuf chain. 1198 */ 1199 int oldoptlen = mopt->m_len; 1200 struct mbuf *n; 1201 1202 /* 1203 * XXX: give up if the whole (new) hbh header does 1204 * not fit even in an mbuf cluster. 1205 */ 1206 if (oldoptlen + JUMBOOPTLEN > MCLBYTES) 1207 return (ENOBUFS); 1208 1209 /* 1210 * As a consequence, we must always prepare a cluster 1211 * at this point. 1212 */ 1213 MGET(n, M_DONTWAIT, MT_DATA); 1214 if (n) { 1215 MCLGET(n, M_DONTWAIT); 1216 if ((n->m_flags & M_EXT) == 0) { 1217 m_freem(n); 1218 n = NULL; 1219 } 1220 } 1221 if (!n) 1222 return (ENOBUFS); 1223 n->m_len = oldoptlen + JUMBOOPTLEN; 1224 bcopy(mtod(mopt, caddr_t), mtod(n, caddr_t), 1225 oldoptlen); 1226 optbuf = mtod(n, u_int8_t *) + oldoptlen; 1227 m_freem(mopt); 1228 mopt = exthdrs->ip6e_hbh = n; 1229 } else { 1230 optbuf = mtod(mopt, u_int8_t *) + mopt->m_len; 1231 mopt->m_len += JUMBOOPTLEN; 1232 } 1233 optbuf[0] = IP6OPT_PADN; 1234 optbuf[1] = 0; 1235 1236 /* 1237 * Adjust the header length according to the pad and 1238 * the jumbo payload option. 1239 */ 1240 hbh = mtod(mopt, struct ip6_hbh *); 1241 hbh->ip6h_len += (JUMBOOPTLEN >> 3); 1242 } 1243 1244 /* fill in the option. */ 1245 optbuf[2] = IP6OPT_JUMBO; 1246 optbuf[3] = 4; 1247 v = (u_int32_t)htonl(plen + JUMBOOPTLEN); 1248 bcopy(&v, &optbuf[4], sizeof(u_int32_t)); 1249 1250 /* finally, adjust the packet header length */ 1251 exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN; 1252 1253 return (0); 1254 #undef JUMBOOPTLEN 1255 } 1256 1257 /* 1258 * Insert fragment header and copy unfragmentable header portions. 1259 */ 1260 static int 1261 ip6_insertfraghdr(m0, m, hlen, frghdrp) 1262 struct mbuf *m0, *m; 1263 int hlen; 1264 struct ip6_frag **frghdrp; 1265 { 1266 struct mbuf *n, *mlast; 1267 1268 if (hlen > sizeof(struct ip6_hdr)) { 1269 n = m_copym(m0, sizeof(struct ip6_hdr), 1270 hlen - sizeof(struct ip6_hdr), M_DONTWAIT); 1271 if (n == 0) 1272 return (ENOBUFS); 1273 m->m_next = n; 1274 } else 1275 n = m; 1276 1277 /* Search for the last mbuf of unfragmentable part. */ 1278 for (mlast = n; mlast->m_next; mlast = mlast->m_next) 1279 ; 1280 1281 if ((mlast->m_flags & M_EXT) == 0 && 1282 M_TRAILINGSPACE(mlast) >= sizeof(struct ip6_frag)) { 1283 /* use the trailing space of the last mbuf for the fragment hdr */ 1284 *frghdrp = (struct ip6_frag *)(mtod(mlast, caddr_t) + 1285 mlast->m_len); 1286 mlast->m_len += sizeof(struct ip6_frag); 1287 m->m_pkthdr.len += sizeof(struct ip6_frag); 1288 } else { 1289 /* allocate a new mbuf for the fragment header */ 1290 struct mbuf *mfrg; 1291 1292 MGET(mfrg, M_DONTWAIT, MT_DATA); 1293 if (mfrg == 0) 1294 return (ENOBUFS); 1295 mfrg->m_len = sizeof(struct ip6_frag); 1296 *frghdrp = mtod(mfrg, struct ip6_frag *); 1297 mlast->m_next = mfrg; 1298 } 1299 1300 return (0); 1301 } 1302 1303 static int 1304 ip6_getpmtu(ro_pmtu, ro, ifp, dst, mtup, alwaysfragp) 1305 struct route_in6 *ro_pmtu, *ro; 1306 struct ifnet *ifp; 1307 struct in6_addr *dst; 1308 u_long *mtup; 1309 int *alwaysfragp; 1310 { 1311 u_int32_t mtu = 0; 1312 int alwaysfrag = 0; 1313 int error = 0; 1314 1315 if (ro_pmtu != ro) { 1316 /* The first hop and the final destination may differ. */ 1317 struct sockaddr_in6 *sa6_dst = 1318 (struct sockaddr_in6 *)&ro_pmtu->ro_dst; 1319 if (ro_pmtu->ro_rt && 1320 ((ro_pmtu->ro_rt->rt_flags & RTF_UP) == 0 || 1321 !IN6_ARE_ADDR_EQUAL(&sa6_dst->sin6_addr, dst))) { 1322 RTFREE(ro_pmtu->ro_rt); 1323 ro_pmtu->ro_rt = (struct rtentry *)NULL; 1324 } 1325 if (ro_pmtu->ro_rt == NULL) { 1326 bzero(sa6_dst, sizeof(*sa6_dst)); /* for safety */ 1327 sa6_dst->sin6_family = AF_INET6; 1328 sa6_dst->sin6_len = sizeof(struct sockaddr_in6); 1329 sa6_dst->sin6_addr = *dst; 1330 1331 rtalloc((struct route *)ro_pmtu); 1332 } 1333 } 1334 if (ro_pmtu->ro_rt) { 1335 u_int32_t ifmtu; 1336 1337 if (ifp == NULL) 1338 ifp = ro_pmtu->ro_rt->rt_ifp; 1339 ifmtu = IN6_LINKMTU(ifp); 1340 mtu = ro_pmtu->ro_rt->rt_rmx.rmx_mtu; 1341 if (mtu == 0) 1342 mtu = ifmtu; 1343 else if (mtu < IPV6_MMTU) { 1344 /* 1345 * RFC2460 section 5, last paragraph: 1346 * if we record ICMPv6 too big message with 1347 * mtu < IPV6_MMTU, transmit packets sized IPV6_MMTU 1348 * or smaller, with fragment header attached. 1349 * (fragment header is needed regardless from the 1350 * packet size, for translators to identify packets) 1351 */ 1352 alwaysfrag = 1; 1353 mtu = IPV6_MMTU; 1354 } else if (mtu > ifmtu) { 1355 /* 1356 * The MTU on the route is larger than the MTU on 1357 * the interface! This shouldn't happen, unless the 1358 * MTU of the interface has been changed after the 1359 * interface was brought up. Change the MTU in the 1360 * route to match the interface MTU (as long as the 1361 * field isn't locked). 1362 */ 1363 mtu = ifmtu; 1364 if (!(ro_pmtu->ro_rt->rt_rmx.rmx_locks & RTV_MTU)) 1365 ro_pmtu->ro_rt->rt_rmx.rmx_mtu = mtu; 1366 } 1367 } else if (ifp) { 1368 mtu = IN6_LINKMTU(ifp); 1369 } else 1370 error = EHOSTUNREACH; /* XXX */ 1371 1372 *mtup = mtu; 1373 if (alwaysfragp) 1374 *alwaysfragp = alwaysfrag; 1375 return (error); 1376 } 1377 1378 /* 1379 * IP6 socket option processing. 1380 */ 1381 int 1382 ip6_ctloutput(op, so, level, optname, mp) 1383 int op; 1384 struct socket *so; 1385 int level, optname; 1386 struct mbuf **mp; 1387 { 1388 struct in6pcb *in6p = sotoin6pcb(so); 1389 struct mbuf *m = *mp; 1390 int optval = 0; 1391 int error = 0; 1392 struct proc *p = curproc; /* XXX */ 1393 1394 if (level == IPPROTO_IPV6) { 1395 switch (op) { 1396 case PRCO_SETOPT: 1397 switch (optname) { 1398 case IPV6_PKTOPTIONS: 1399 /* m is freed in ip6_pcbopts */ 1400 return (ip6_pcbopts(&in6p->in6p_outputopts, 1401 m, so)); 1402 case IPV6_HOPOPTS: 1403 case IPV6_DSTOPTS: 1404 if (p == 0 || suser(p->p_ucred, &p->p_acflag)) { 1405 error = EPERM; 1406 break; 1407 } 1408 /* FALLTHROUGH */ 1409 case IPV6_UNICAST_HOPS: 1410 case IPV6_RECVOPTS: 1411 case IPV6_RECVRETOPTS: 1412 case IPV6_RECVDSTADDR: 1413 case IPV6_PKTINFO: 1414 case IPV6_HOPLIMIT: 1415 case IPV6_RTHDR: 1416 case IPV6_FAITH: 1417 case IPV6_V6ONLY: 1418 case IPV6_USE_MIN_MTU: 1419 if (!m || m->m_len != sizeof(int)) { 1420 error = EINVAL; 1421 break; 1422 } 1423 optval = *mtod(m, int *); 1424 switch (optname) { 1425 1426 case IPV6_UNICAST_HOPS: 1427 if (optval < -1 || optval >= 256) 1428 error = EINVAL; 1429 else { 1430 /* -1 = kernel default */ 1431 in6p->in6p_hops = optval; 1432 } 1433 break; 1434 #define OPTSET(bit) \ 1435 do { \ 1436 if (optval) \ 1437 in6p->in6p_flags |= (bit); \ 1438 else \ 1439 in6p->in6p_flags &= ~(bit); \ 1440 } while (/*CONSTCOND*/ 0) 1441 1442 case IPV6_RECVOPTS: 1443 OPTSET(IN6P_RECVOPTS); 1444 break; 1445 1446 case IPV6_RECVRETOPTS: 1447 OPTSET(IN6P_RECVRETOPTS); 1448 break; 1449 1450 case IPV6_RECVDSTADDR: 1451 OPTSET(IN6P_RECVDSTADDR); 1452 break; 1453 1454 case IPV6_PKTINFO: 1455 OPTSET(IN6P_PKTINFO); 1456 break; 1457 1458 case IPV6_HOPLIMIT: 1459 OPTSET(IN6P_HOPLIMIT); 1460 break; 1461 1462 case IPV6_HOPOPTS: 1463 OPTSET(IN6P_HOPOPTS); 1464 break; 1465 1466 case IPV6_DSTOPTS: 1467 OPTSET(IN6P_DSTOPTS); 1468 break; 1469 1470 case IPV6_RTHDR: 1471 OPTSET(IN6P_RTHDR); 1472 break; 1473 1474 case IPV6_FAITH: 1475 OPTSET(IN6P_FAITH); 1476 break; 1477 1478 case IPV6_USE_MIN_MTU: 1479 OPTSET(IN6P_MINMTU); 1480 break; 1481 1482 case IPV6_V6ONLY: 1483 /* 1484 * make setsockopt(IPV6_V6ONLY) 1485 * available only prior to bind(2). 1486 * see ipng mailing list, Jun 22 2001. 1487 */ 1488 if (in6p->in6p_lport || 1489 !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr)) { 1490 error = EINVAL; 1491 break; 1492 } 1493 #ifdef INET6_BINDV6ONLY 1494 if (!optval) 1495 error = EINVAL; 1496 #else 1497 OPTSET(IN6P_IPV6_V6ONLY); 1498 #endif 1499 break; 1500 } 1501 break; 1502 #undef OPTSET 1503 1504 case IPV6_MULTICAST_IF: 1505 case IPV6_MULTICAST_HOPS: 1506 case IPV6_MULTICAST_LOOP: 1507 case IPV6_JOIN_GROUP: 1508 case IPV6_LEAVE_GROUP: 1509 error = ip6_setmoptions(optname, 1510 &in6p->in6p_moptions, m); 1511 break; 1512 1513 case IPV6_PORTRANGE: 1514 optval = *mtod(m, int *); 1515 1516 switch (optval) { 1517 case IPV6_PORTRANGE_DEFAULT: 1518 in6p->in6p_flags &= ~(IN6P_LOWPORT); 1519 in6p->in6p_flags &= ~(IN6P_HIGHPORT); 1520 break; 1521 1522 case IPV6_PORTRANGE_HIGH: 1523 in6p->in6p_flags &= ~(IN6P_LOWPORT); 1524 in6p->in6p_flags |= IN6P_HIGHPORT; 1525 break; 1526 1527 case IPV6_PORTRANGE_LOW: 1528 in6p->in6p_flags &= ~(IN6P_HIGHPORT); 1529 in6p->in6p_flags |= IN6P_LOWPORT; 1530 break; 1531 1532 default: 1533 error = EINVAL; 1534 break; 1535 } 1536 break; 1537 1538 #ifdef IPSEC 1539 case IPV6_IPSEC_POLICY: 1540 { 1541 caddr_t req = NULL; 1542 size_t len = 0; 1543 1544 int priv = 0; 1545 if (p == 0 || suser(p->p_ucred, &p->p_acflag)) 1546 priv = 0; 1547 else 1548 priv = 1; 1549 if (m) { 1550 req = mtod(m, caddr_t); 1551 len = m->m_len; 1552 } 1553 error = ipsec6_set_policy(in6p, 1554 optname, req, len, priv); 1555 } 1556 break; 1557 #endif /* IPSEC */ 1558 1559 default: 1560 error = ENOPROTOOPT; 1561 break; 1562 } 1563 if (m) 1564 (void)m_free(m); 1565 break; 1566 1567 case PRCO_GETOPT: 1568 switch (optname) { 1569 1570 case IPV6_OPTIONS: 1571 case IPV6_RETOPTS: 1572 error = ENOPROTOOPT; 1573 break; 1574 1575 case IPV6_PKTOPTIONS: 1576 if (in6p->in6p_options) { 1577 *mp = m_copym(in6p->in6p_options, 0, 1578 M_COPYALL, M_WAIT); 1579 } else { 1580 *mp = m_get(M_WAIT, MT_SOOPTS); 1581 (*mp)->m_len = 0; 1582 } 1583 break; 1584 1585 case IPV6_HOPOPTS: 1586 case IPV6_DSTOPTS: 1587 if (p == 0 || suser(p->p_ucred, &p->p_acflag)) { 1588 error = EPERM; 1589 break; 1590 } 1591 /* FALLTHROUGH */ 1592 case IPV6_UNICAST_HOPS: 1593 case IPV6_RECVOPTS: 1594 case IPV6_RECVRETOPTS: 1595 case IPV6_RECVDSTADDR: 1596 case IPV6_PORTRANGE: 1597 case IPV6_PKTINFO: 1598 case IPV6_HOPLIMIT: 1599 case IPV6_RTHDR: 1600 case IPV6_FAITH: 1601 case IPV6_V6ONLY: 1602 case IPV6_USE_MIN_MTU: 1603 *mp = m = m_get(M_WAIT, MT_SOOPTS); 1604 m->m_len = sizeof(int); 1605 switch (optname) { 1606 1607 case IPV6_UNICAST_HOPS: 1608 optval = in6p->in6p_hops; 1609 break; 1610 1611 #define OPTBIT(bit) (in6p->in6p_flags & bit ? 1 : 0) 1612 1613 case IPV6_RECVOPTS: 1614 optval = OPTBIT(IN6P_RECVOPTS); 1615 break; 1616 1617 case IPV6_RECVRETOPTS: 1618 optval = OPTBIT(IN6P_RECVRETOPTS); 1619 break; 1620 1621 case IPV6_RECVDSTADDR: 1622 optval = OPTBIT(IN6P_RECVDSTADDR); 1623 break; 1624 1625 case IPV6_PORTRANGE: 1626 { 1627 int flags; 1628 flags = in6p->in6p_flags; 1629 if (flags & IN6P_HIGHPORT) 1630 optval = IPV6_PORTRANGE_HIGH; 1631 else if (flags & IN6P_LOWPORT) 1632 optval = IPV6_PORTRANGE_LOW; 1633 else 1634 optval = 0; 1635 break; 1636 } 1637 1638 case IPV6_PKTINFO: 1639 optval = OPTBIT(IN6P_PKTINFO); 1640 break; 1641 1642 case IPV6_HOPLIMIT: 1643 optval = OPTBIT(IN6P_HOPLIMIT); 1644 break; 1645 1646 case IPV6_HOPOPTS: 1647 optval = OPTBIT(IN6P_HOPOPTS); 1648 break; 1649 1650 case IPV6_DSTOPTS: 1651 optval = OPTBIT(IN6P_DSTOPTS); 1652 break; 1653 1654 case IPV6_RTHDR: 1655 optval = OPTBIT(IN6P_RTHDR); 1656 break; 1657 1658 case IPV6_FAITH: 1659 optval = OPTBIT(IN6P_FAITH); 1660 break; 1661 1662 case IPV6_V6ONLY: 1663 optval = OPTBIT(IN6P_IPV6_V6ONLY); 1664 break; 1665 1666 case IPV6_USE_MIN_MTU: 1667 optval = OPTBIT(IN6P_MINMTU); 1668 break; 1669 } 1670 *mtod(m, int *) = optval; 1671 break; 1672 1673 case IPV6_MULTICAST_IF: 1674 case IPV6_MULTICAST_HOPS: 1675 case IPV6_MULTICAST_LOOP: 1676 case IPV6_JOIN_GROUP: 1677 case IPV6_LEAVE_GROUP: 1678 error = ip6_getmoptions(optname, in6p->in6p_moptions, mp); 1679 break; 1680 1681 #if 0 /* defined(IPSEC) */ 1682 /* XXX: code broken */ 1683 case IPV6_IPSEC_POLICY: 1684 { 1685 caddr_t req = NULL; 1686 size_t len = 0; 1687 1688 if (m) { 1689 req = mtod(m, caddr_t); 1690 len = m->m_len; 1691 } 1692 error = ipsec6_get_policy(in6p, req, len, mp); 1693 break; 1694 } 1695 #endif /* IPSEC */ 1696 1697 default: 1698 error = ENOPROTOOPT; 1699 break; 1700 } 1701 break; 1702 } 1703 } else { 1704 error = EINVAL; 1705 if (op == PRCO_SETOPT && *mp) 1706 (void)m_free(*mp); 1707 } 1708 return (error); 1709 } 1710 1711 int 1712 ip6_raw_ctloutput(op, so, level, optname, mp) 1713 int op; 1714 struct socket *so; 1715 int level, optname; 1716 struct mbuf **mp; 1717 { 1718 int error = 0, optval, optlen; 1719 const int icmp6off = offsetof(struct icmp6_hdr, icmp6_cksum); 1720 struct in6pcb *in6p = sotoin6pcb(so); 1721 struct mbuf *m = *mp; 1722 1723 optlen = m ? m->m_len : 0; 1724 1725 if (level != IPPROTO_IPV6) { 1726 if (op == PRCO_SETOPT && *mp) 1727 (void)m_free(*mp); 1728 return (EINVAL); 1729 } 1730 1731 switch (optname) { 1732 case IPV6_CHECKSUM: 1733 /* 1734 * For ICMPv6 sockets, no modification allowed for checksum 1735 * offset, permit "no change" values to help existing apps. 1736 * 1737 * XXX 2292bis says: "An attempt to set IPV6_CHECKSUM 1738 * for an ICMPv6 socket will fail." 1739 * The current behavior does not meet 2292bis. 1740 */ 1741 switch (op) { 1742 case PRCO_SETOPT: 1743 if (optlen != sizeof(int)) { 1744 error = EINVAL; 1745 break; 1746 } 1747 optval = *mtod(m, int *); 1748 if ((optval % 2) != 0) { 1749 /* the API assumes even offset values */ 1750 error = EINVAL; 1751 } else if (so->so_proto->pr_protocol == 1752 IPPROTO_ICMPV6) { 1753 if (optval != icmp6off) 1754 error = EINVAL; 1755 } else 1756 in6p->in6p_cksum = optval; 1757 break; 1758 1759 case PRCO_GETOPT: 1760 if (so->so_proto->pr_protocol == IPPROTO_ICMPV6) 1761 optval = icmp6off; 1762 else 1763 optval = in6p->in6p_cksum; 1764 1765 *mp = m = m_get(M_WAIT, MT_SOOPTS); 1766 m->m_len = sizeof(int); 1767 *mtod(m, int *) = optval; 1768 break; 1769 1770 default: 1771 error = EINVAL; 1772 break; 1773 } 1774 break; 1775 1776 default: 1777 error = ENOPROTOOPT; 1778 break; 1779 } 1780 1781 if (op == PRCO_SETOPT && m) 1782 (void)m_free(m); 1783 1784 return (error); 1785 } 1786 1787 /* 1788 * Set up IP6 options in pcb for insertion in output packets. 1789 * Store in mbuf with pointer in pcbopt, adding pseudo-option 1790 * with destination address if source routed. 1791 */ 1792 static int 1793 ip6_pcbopts(pktopt, m, so) 1794 struct ip6_pktopts **pktopt; 1795 struct mbuf *m; 1796 struct socket *so; 1797 { 1798 struct ip6_pktopts *opt = *pktopt; 1799 int error = 0; 1800 struct proc *p = curproc; /* XXX */ 1801 int priv = 0; 1802 1803 /* turn off any old options. */ 1804 if (opt) { 1805 if (opt->ip6po_m) 1806 (void)m_free(opt->ip6po_m); 1807 } else 1808 opt = malloc(sizeof(*opt), M_IP6OPT, M_WAITOK); 1809 *pktopt = 0; 1810 1811 if (!m || m->m_len == 0) { 1812 /* 1813 * Only turning off any previous options. 1814 */ 1815 free(opt, M_IP6OPT); 1816 if (m) 1817 (void)m_free(m); 1818 return (0); 1819 } 1820 1821 /* set options specified by user. */ 1822 if (p && !suser(p->p_ucred, &p->p_acflag)) 1823 priv = 1; 1824 if ((error = ip6_setpktoptions(m, opt, priv)) != 0) { 1825 (void)m_free(m); 1826 free(opt, M_IP6OPT); 1827 return (error); 1828 } 1829 *pktopt = opt; 1830 return (0); 1831 } 1832 1833 /* 1834 * Set the IP6 multicast options in response to user setsockopt(). 1835 */ 1836 static int 1837 ip6_setmoptions(optname, im6op, m) 1838 int optname; 1839 struct ip6_moptions **im6op; 1840 struct mbuf *m; 1841 { 1842 int error = 0; 1843 u_int loop, ifindex; 1844 struct ipv6_mreq *mreq; 1845 struct ifnet *ifp; 1846 struct ip6_moptions *im6o = *im6op; 1847 struct route_in6 ro; 1848 struct sockaddr_in6 *dst; 1849 struct in6_multi_mship *imm; 1850 struct proc *p = curproc; /* XXX */ 1851 1852 if (im6o == NULL) { 1853 /* 1854 * No multicast option buffer attached to the pcb; 1855 * allocate one and initialize to default values. 1856 */ 1857 im6o = (struct ip6_moptions *) 1858 malloc(sizeof(*im6o), M_IPMOPTS, M_WAITOK); 1859 1860 if (im6o == NULL) 1861 return (ENOBUFS); 1862 *im6op = im6o; 1863 im6o->im6o_multicast_ifp = NULL; 1864 im6o->im6o_multicast_hlim = ip6_defmcasthlim; 1865 im6o->im6o_multicast_loop = IPV6_DEFAULT_MULTICAST_LOOP; 1866 LIST_INIT(&im6o->im6o_memberships); 1867 } 1868 1869 switch (optname) { 1870 1871 case IPV6_MULTICAST_IF: 1872 /* 1873 * Select the interface for outgoing multicast packets. 1874 */ 1875 if (m == NULL || m->m_len != sizeof(u_int)) { 1876 error = EINVAL; 1877 break; 1878 } 1879 bcopy(mtod(m, u_int *), &ifindex, sizeof(ifindex)); 1880 if (ifindex != 0) { 1881 if (ifindex < 0 || if_indexlim <= ifindex || 1882 !ifindex2ifnet[ifindex]) { 1883 error = ENXIO; /* XXX EINVAL? */ 1884 break; 1885 } 1886 ifp = ifindex2ifnet[ifindex]; 1887 if ((ifp->if_flags & IFF_MULTICAST) == 0) { 1888 error = EADDRNOTAVAIL; 1889 break; 1890 } 1891 } else 1892 ifp = NULL; 1893 im6o->im6o_multicast_ifp = ifp; 1894 break; 1895 1896 case IPV6_MULTICAST_HOPS: 1897 { 1898 /* 1899 * Set the IP6 hoplimit for outgoing multicast packets. 1900 */ 1901 int optval; 1902 if (m == NULL || m->m_len != sizeof(int)) { 1903 error = EINVAL; 1904 break; 1905 } 1906 bcopy(mtod(m, u_int *), &optval, sizeof(optval)); 1907 if (optval < -1 || optval >= 256) 1908 error = EINVAL; 1909 else if (optval == -1) 1910 im6o->im6o_multicast_hlim = ip6_defmcasthlim; 1911 else 1912 im6o->im6o_multicast_hlim = optval; 1913 break; 1914 } 1915 1916 case IPV6_MULTICAST_LOOP: 1917 /* 1918 * Set the loopback flag for outgoing multicast packets. 1919 * Must be zero or one. 1920 */ 1921 if (m == NULL || m->m_len != sizeof(u_int)) { 1922 error = EINVAL; 1923 break; 1924 } 1925 bcopy(mtod(m, u_int *), &loop, sizeof(loop)); 1926 if (loop > 1) { 1927 error = EINVAL; 1928 break; 1929 } 1930 im6o->im6o_multicast_loop = loop; 1931 break; 1932 1933 case IPV6_JOIN_GROUP: 1934 /* 1935 * Add a multicast group membership. 1936 * Group must be a valid IP6 multicast address. 1937 */ 1938 if (m == NULL || m->m_len != sizeof(struct ipv6_mreq)) { 1939 error = EINVAL; 1940 break; 1941 } 1942 mreq = mtod(m, struct ipv6_mreq *); 1943 if (IN6_IS_ADDR_UNSPECIFIED(&mreq->ipv6mr_multiaddr)) { 1944 /* 1945 * We use the unspecified address to specify to accept 1946 * all multicast addresses. Only super user is allowed 1947 * to do this. 1948 */ 1949 if (suser(p->p_ucred, &p->p_acflag)) 1950 { 1951 error = EACCES; 1952 break; 1953 } 1954 } else if (!IN6_IS_ADDR_MULTICAST(&mreq->ipv6mr_multiaddr)) { 1955 error = EINVAL; 1956 break; 1957 } 1958 1959 /* 1960 * If the interface is specified, validate it. 1961 * If no interface was explicitly specified, choose an 1962 * appropriate one according to the given multicast address. 1963 */ 1964 if (mreq->ipv6mr_interface != 0) { 1965 if (mreq->ipv6mr_interface < 0 || 1966 if_indexlim <= mreq->ipv6mr_interface || 1967 !ifindex2ifnet[mreq->ipv6mr_interface]) { 1968 error = ENXIO; /* XXX EINVAL? */ 1969 break; 1970 } 1971 ifp = ifindex2ifnet[mreq->ipv6mr_interface]; 1972 } else { 1973 /* 1974 * If the multicast address is in node-local scope, 1975 * the interface should be a loopback interface. 1976 * Otherwise, look up the routing table for the 1977 * address, and choose the outgoing interface. 1978 * XXX: is it a good approach? 1979 */ 1980 if (IN6_IS_ADDR_MC_NODELOCAL(&mreq->ipv6mr_multiaddr)) { 1981 ifp = lo0ifp; 1982 } else { 1983 ro.ro_rt = NULL; 1984 dst = (struct sockaddr_in6 *)&ro.ro_dst; 1985 bzero(dst, sizeof(*dst)); 1986 dst->sin6_len = sizeof(struct sockaddr_in6); 1987 dst->sin6_family = AF_INET6; 1988 dst->sin6_addr = mreq->ipv6mr_multiaddr; 1989 rtalloc((struct route *)&ro); 1990 if (ro.ro_rt == NULL) { 1991 error = EADDRNOTAVAIL; 1992 break; 1993 } 1994 ifp = ro.ro_rt->rt_ifp; 1995 rtfree(ro.ro_rt); 1996 } 1997 } 1998 1999 /* 2000 * See if we found an interface, and confirm that it 2001 * supports multicast 2002 */ 2003 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) { 2004 error = EADDRNOTAVAIL; 2005 break; 2006 } 2007 /* 2008 * Put interface index into the multicast address, 2009 * if the address has link-local scope. 2010 */ 2011 if (IN6_IS_ADDR_MC_LINKLOCAL(&mreq->ipv6mr_multiaddr)) { 2012 mreq->ipv6mr_multiaddr.s6_addr16[1] = 2013 htons(ifp->if_index); 2014 } 2015 /* 2016 * See if the membership already exists. 2017 */ 2018 for (imm = im6o->im6o_memberships.lh_first; 2019 imm != NULL; imm = imm->i6mm_chain.le_next) 2020 if (imm->i6mm_maddr->in6m_ifp == ifp && 2021 IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr, 2022 &mreq->ipv6mr_multiaddr)) 2023 break; 2024 if (imm != NULL) { 2025 error = EADDRINUSE; 2026 break; 2027 } 2028 /* 2029 * Everything looks good; add a new record to the multicast 2030 * address list for the given interface. 2031 */ 2032 imm = in6_joingroup(ifp, &mreq->ipv6mr_multiaddr, &error); 2033 if (!imm) 2034 break; 2035 LIST_INSERT_HEAD(&im6o->im6o_memberships, imm, i6mm_chain); 2036 break; 2037 2038 case IPV6_LEAVE_GROUP: 2039 /* 2040 * Drop a multicast group membership. 2041 * Group must be a valid IP6 multicast address. 2042 */ 2043 if (m == NULL || m->m_len != sizeof(struct ipv6_mreq)) { 2044 error = EINVAL; 2045 break; 2046 } 2047 mreq = mtod(m, struct ipv6_mreq *); 2048 if (IN6_IS_ADDR_UNSPECIFIED(&mreq->ipv6mr_multiaddr)) { 2049 if (suser(p->p_ucred, &p->p_acflag)) 2050 { 2051 error = EACCES; 2052 break; 2053 } 2054 } else if (!IN6_IS_ADDR_MULTICAST(&mreq->ipv6mr_multiaddr)) { 2055 error = EINVAL; 2056 break; 2057 } 2058 /* 2059 * If an interface address was specified, get a pointer 2060 * to its ifnet structure. 2061 */ 2062 if (mreq->ipv6mr_interface != 0) { 2063 if (mreq->ipv6mr_interface < 0 || 2064 if_indexlim <= mreq->ipv6mr_interface || 2065 !ifindex2ifnet[mreq->ipv6mr_interface]) { 2066 error = ENXIO; /* XXX EINVAL? */ 2067 break; 2068 } 2069 ifp = ifindex2ifnet[mreq->ipv6mr_interface]; 2070 } else 2071 ifp = NULL; 2072 /* 2073 * Put interface index into the multicast address, 2074 * if the address has link-local scope. 2075 */ 2076 if (IN6_IS_ADDR_MC_LINKLOCAL(&mreq->ipv6mr_multiaddr)) { 2077 mreq->ipv6mr_multiaddr.s6_addr16[1] = 2078 htons(mreq->ipv6mr_interface); 2079 } 2080 /* 2081 * Find the membership in the membership list. 2082 */ 2083 for (imm = im6o->im6o_memberships.lh_first; 2084 imm != NULL; imm = imm->i6mm_chain.le_next) { 2085 if ((ifp == NULL || imm->i6mm_maddr->in6m_ifp == ifp) && 2086 IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr, 2087 &mreq->ipv6mr_multiaddr)) 2088 break; 2089 } 2090 if (imm == NULL) { 2091 /* Unable to resolve interface */ 2092 error = EADDRNOTAVAIL; 2093 break; 2094 } 2095 /* 2096 * Give up the multicast address record to which the 2097 * membership points. 2098 */ 2099 LIST_REMOVE(imm, i6mm_chain); 2100 in6_leavegroup(imm); 2101 break; 2102 2103 default: 2104 error = EOPNOTSUPP; 2105 break; 2106 } 2107 2108 /* 2109 * If all options have default values, no need to keep the mbuf. 2110 */ 2111 if (im6o->im6o_multicast_ifp == NULL && 2112 im6o->im6o_multicast_hlim == ip6_defmcasthlim && 2113 im6o->im6o_multicast_loop == IPV6_DEFAULT_MULTICAST_LOOP && 2114 im6o->im6o_memberships.lh_first == NULL) { 2115 free(*im6op, M_IPMOPTS); 2116 *im6op = NULL; 2117 } 2118 2119 return (error); 2120 } 2121 2122 /* 2123 * Return the IP6 multicast options in response to user getsockopt(). 2124 */ 2125 static int 2126 ip6_getmoptions(optname, im6o, mp) 2127 int optname; 2128 struct ip6_moptions *im6o; 2129 struct mbuf **mp; 2130 { 2131 u_int *hlim, *loop, *ifindex; 2132 2133 *mp = m_get(M_WAIT, MT_SOOPTS); 2134 2135 switch (optname) { 2136 2137 case IPV6_MULTICAST_IF: 2138 ifindex = mtod(*mp, u_int *); 2139 (*mp)->m_len = sizeof(u_int); 2140 if (im6o == NULL || im6o->im6o_multicast_ifp == NULL) 2141 *ifindex = 0; 2142 else 2143 *ifindex = im6o->im6o_multicast_ifp->if_index; 2144 return (0); 2145 2146 case IPV6_MULTICAST_HOPS: 2147 hlim = mtod(*mp, u_int *); 2148 (*mp)->m_len = sizeof(u_int); 2149 if (im6o == NULL) 2150 *hlim = ip6_defmcasthlim; 2151 else 2152 *hlim = im6o->im6o_multicast_hlim; 2153 return (0); 2154 2155 case IPV6_MULTICAST_LOOP: 2156 loop = mtod(*mp, u_int *); 2157 (*mp)->m_len = sizeof(u_int); 2158 if (im6o == NULL) 2159 *loop = ip6_defmcasthlim; 2160 else 2161 *loop = im6o->im6o_multicast_loop; 2162 return (0); 2163 2164 default: 2165 return (EOPNOTSUPP); 2166 } 2167 } 2168 2169 /* 2170 * Discard the IP6 multicast options. 2171 */ 2172 void 2173 ip6_freemoptions(im6o) 2174 struct ip6_moptions *im6o; 2175 { 2176 struct in6_multi_mship *imm; 2177 2178 if (im6o == NULL) 2179 return; 2180 2181 while ((imm = im6o->im6o_memberships.lh_first) != NULL) { 2182 LIST_REMOVE(imm, i6mm_chain); 2183 in6_leavegroup(imm); 2184 } 2185 free(im6o, M_IPMOPTS); 2186 } 2187 2188 /* 2189 * Set IPv6 outgoing packet options based on advanced API. 2190 */ 2191 int 2192 ip6_setpktoptions(control, opt, priv) 2193 struct mbuf *control; 2194 struct ip6_pktopts *opt; 2195 int priv; 2196 { 2197 struct cmsghdr *cm = 0; 2198 2199 if (control == 0 || opt == 0) 2200 return (EINVAL); 2201 2202 bzero(opt, sizeof(*opt)); 2203 opt->ip6po_hlim = -1; /* -1 means to use default hop limit */ 2204 2205 /* 2206 * XXX: Currently, we assume all the optional information is stored 2207 * in a single mbuf. 2208 */ 2209 if (control->m_next) 2210 return (EINVAL); 2211 2212 opt->ip6po_m = control; 2213 2214 for (; control->m_len; control->m_data += CMSG_ALIGN(cm->cmsg_len), 2215 control->m_len -= CMSG_ALIGN(cm->cmsg_len)) { 2216 cm = mtod(control, struct cmsghdr *); 2217 if (cm->cmsg_len == 0 || cm->cmsg_len > control->m_len) 2218 return (EINVAL); 2219 if (cm->cmsg_level != IPPROTO_IPV6) 2220 continue; 2221 2222 switch (cm->cmsg_type) { 2223 case IPV6_PKTINFO: 2224 if (cm->cmsg_len != CMSG_LEN(sizeof(struct in6_pktinfo))) 2225 return (EINVAL); 2226 opt->ip6po_pktinfo = (struct in6_pktinfo *)CMSG_DATA(cm); 2227 if (opt->ip6po_pktinfo->ipi6_ifindex && 2228 IN6_IS_ADDR_LINKLOCAL(&opt->ip6po_pktinfo->ipi6_addr)) 2229 opt->ip6po_pktinfo->ipi6_addr.s6_addr16[1] = 2230 htons(opt->ip6po_pktinfo->ipi6_ifindex); 2231 2232 if (opt->ip6po_pktinfo->ipi6_ifindex >= if_indexlim || 2233 opt->ip6po_pktinfo->ipi6_ifindex < 0) 2234 return (ENXIO); 2235 if (opt->ip6po_pktinfo->ipi6_ifindex > 0 && 2236 !ifindex2ifnet[opt->ip6po_pktinfo->ipi6_ifindex]) 2237 return (ENXIO); 2238 2239 /* 2240 * Check if the requested source address is indeed a 2241 * unicast address assigned to the node, and can be 2242 * used as the packet's source address. 2243 */ 2244 if (!IN6_IS_ADDR_UNSPECIFIED(&opt->ip6po_pktinfo->ipi6_addr)) { 2245 struct ifaddr *ia; 2246 struct in6_ifaddr *ia6; 2247 struct sockaddr_in6 sin6; 2248 2249 bzero(&sin6, sizeof(sin6)); 2250 sin6.sin6_len = sizeof(sin6); 2251 sin6.sin6_family = AF_INET6; 2252 sin6.sin6_addr = 2253 opt->ip6po_pktinfo->ipi6_addr; 2254 ia = ifa_ifwithaddr(sin6tosa(&sin6)); 2255 if (ia == NULL || 2256 (opt->ip6po_pktinfo->ipi6_ifindex && 2257 (ia->ifa_ifp->if_index != 2258 opt->ip6po_pktinfo->ipi6_ifindex))) { 2259 return (EADDRNOTAVAIL); 2260 } 2261 ia6 = (struct in6_ifaddr *)ia; 2262 if ((ia6->ia6_flags & (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY)) != 0) { 2263 return (EADDRNOTAVAIL); 2264 } 2265 2266 /* 2267 * Check if the requested source address is 2268 * indeed a unicast address assigned to the 2269 * node. 2270 */ 2271 if (IN6_IS_ADDR_MULTICAST(&opt->ip6po_pktinfo->ipi6_addr)) 2272 return (EADDRNOTAVAIL); 2273 } 2274 break; 2275 2276 case IPV6_HOPLIMIT: 2277 if (cm->cmsg_len != CMSG_LEN(sizeof(int))) 2278 return (EINVAL); 2279 else { 2280 int t; 2281 2282 bcopy(CMSG_DATA(cm), &t, sizeof(t)); 2283 if (t < -1 || t > 255) 2284 return (EINVAL); 2285 opt->ip6po_hlim = t; 2286 } 2287 break; 2288 2289 case IPV6_NEXTHOP: 2290 if (!priv) 2291 return (EPERM); 2292 2293 /* check if cmsg_len is large enough for sa_len */ 2294 if (cm->cmsg_len < sizeof(u_char) || 2295 cm->cmsg_len < CMSG_LEN(*CMSG_DATA(cm))) 2296 return (EINVAL); 2297 2298 opt->ip6po_nexthop = (struct sockaddr *)CMSG_DATA(cm); 2299 2300 break; 2301 2302 case IPV6_HOPOPTS: 2303 if (cm->cmsg_len < CMSG_LEN(sizeof(struct ip6_hbh))) 2304 return (EINVAL); 2305 else { 2306 struct ip6_hbh *t; 2307 2308 t = (struct ip6_hbh *)CMSG_DATA(cm); 2309 if (cm->cmsg_len != 2310 CMSG_LEN((t->ip6h_len + 1) << 3)) 2311 return (EINVAL); 2312 opt->ip6po_hbh = t; 2313 } 2314 break; 2315 2316 case IPV6_DSTOPTS: 2317 if (cm->cmsg_len < CMSG_LEN(sizeof(struct ip6_dest))) 2318 return (EINVAL); 2319 2320 /* 2321 * If there is no routing header yet, the destination 2322 * options header should be put on the 1st part. 2323 * Otherwise, the header should be on the 2nd part. 2324 * (See RFC 2460, section 4.1) 2325 */ 2326 if (opt->ip6po_rthdr == NULL) { 2327 struct ip6_dest *t; 2328 2329 t = (struct ip6_dest *)CMSG_DATA(cm); 2330 if (cm->cmsg_len != 2331 CMSG_LEN((t->ip6d_len + 1) << 3)); 2332 return (EINVAL); 2333 opt->ip6po_dest1 = t; 2334 } 2335 else { 2336 struct ip6_dest *t; 2337 2338 t = (struct ip6_dest *)CMSG_DATA(cm); 2339 if (cm->cmsg_len != 2340 CMSG_LEN((opt->ip6po_dest2->ip6d_len + 1) << 3)) 2341 return (EINVAL); 2342 opt->ip6po_dest2 = t; 2343 } 2344 break; 2345 2346 case IPV6_RTHDR: 2347 if (cm->cmsg_len < CMSG_LEN(sizeof(struct ip6_rthdr))) 2348 return (EINVAL); 2349 else { 2350 struct ip6_rthdr *t; 2351 2352 t = (struct ip6_rthdr *)CMSG_DATA(cm); 2353 if (cm->cmsg_len != 2354 CMSG_LEN((t->ip6r_len + 1) << 3)) 2355 return (EINVAL); 2356 switch (t->ip6r_type) { 2357 case IPV6_RTHDR_TYPE_0: 2358 if (t->ip6r_segleft == 0) 2359 return (EINVAL); 2360 break; 2361 default: 2362 return (EINVAL); 2363 } 2364 opt->ip6po_rthdr = t; 2365 } 2366 break; 2367 2368 default: 2369 return (ENOPROTOOPT); 2370 } 2371 } 2372 2373 return (0); 2374 } 2375 2376 /* 2377 * Routine called from ip6_output() to loop back a copy of an IP6 multicast 2378 * packet to the input queue of a specified interface. Note that this 2379 * calls the output routine of the loopback "driver", but with an interface 2380 * pointer that might NOT be lo0ifp -- easier than replicating that code here. 2381 */ 2382 void 2383 ip6_mloopback(ifp, m, dst) 2384 struct ifnet *ifp; 2385 struct mbuf *m; 2386 struct sockaddr_in6 *dst; 2387 { 2388 struct mbuf *copym; 2389 struct ip6_hdr *ip6; 2390 2391 copym = m_copy(m, 0, M_COPYALL); 2392 if (copym == NULL) 2393 return; 2394 2395 /* 2396 * Make sure to deep-copy IPv6 header portion in case the data 2397 * is in an mbuf cluster, so that we can safely override the IPv6 2398 * header portion later. 2399 */ 2400 if ((copym->m_flags & M_EXT) != 0 || 2401 copym->m_len < sizeof(struct ip6_hdr)) { 2402 copym = m_pullup(copym, sizeof(struct ip6_hdr)); 2403 if (copym == NULL) 2404 return; 2405 } 2406 2407 #ifdef DIAGNOSTIC 2408 if (copym->m_len < sizeof(*ip6)) { 2409 m_freem(copym); 2410 return; 2411 } 2412 #endif 2413 2414 ip6 = mtod(copym, struct ip6_hdr *); 2415 if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src)) 2416 ip6->ip6_src.s6_addr16[1] = 0; 2417 if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst)) 2418 ip6->ip6_dst.s6_addr16[1] = 0; 2419 2420 (void)looutput(ifp, copym, (struct sockaddr *)dst, NULL); 2421 } 2422 2423 /* 2424 * Chop IPv6 header off from the payload. 2425 */ 2426 static int 2427 ip6_splithdr(m, exthdrs) 2428 struct mbuf *m; 2429 struct ip6_exthdrs *exthdrs; 2430 { 2431 struct mbuf *mh; 2432 struct ip6_hdr *ip6; 2433 2434 ip6 = mtod(m, struct ip6_hdr *); 2435 if (m->m_len > sizeof(*ip6)) { 2436 MGETHDR(mh, M_DONTWAIT, MT_HEADER); 2437 if (mh == 0) { 2438 m_freem(m); 2439 return ENOBUFS; 2440 } 2441 M_MOVE_PKTHDR(mh, m); 2442 MH_ALIGN(mh, sizeof(*ip6)); 2443 m->m_len -= sizeof(*ip6); 2444 m->m_data += sizeof(*ip6); 2445 mh->m_next = m; 2446 m = mh; 2447 m->m_len = sizeof(*ip6); 2448 bcopy((caddr_t)ip6, mtod(m, caddr_t), sizeof(*ip6)); 2449 } 2450 exthdrs->ip6e_ip6 = m; 2451 return 0; 2452 } 2453 2454 /* 2455 * Compute IPv6 extension header length. 2456 */ 2457 int 2458 ip6_optlen(in6p) 2459 struct in6pcb *in6p; 2460 { 2461 int len; 2462 2463 if (!in6p->in6p_outputopts) 2464 return 0; 2465 2466 len = 0; 2467 #define elen(x) \ 2468 (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0) 2469 2470 len += elen(in6p->in6p_outputopts->ip6po_hbh); 2471 len += elen(in6p->in6p_outputopts->ip6po_dest1); 2472 len += elen(in6p->in6p_outputopts->ip6po_rthdr); 2473 len += elen(in6p->in6p_outputopts->ip6po_dest2); 2474 return len; 2475 #undef elen 2476 } 2477