1 /* $NetBSD: ip6_output.c,v 1.79 2004/02/06 08:07:55 itojun 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 1.79 2004/02/06 08:07:55 itojun 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 <netinet6/ip6_var.h> 92 #include <netinet6/in6_pcb.h> 93 #include <netinet6/nd6.h> 94 #include <netinet6/ip6protosw.h> 95 96 #ifdef IPSEC 97 #include <netinet6/ipsec.h> 98 #include <netkey/key.h> 99 #endif /* IPSEC */ 100 101 #include "loop.h" 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 extern struct ifnet loif[NLOOP]; 130 131 /* 132 * IP6 output. The packet in mbuf chain m contains a skeletal IP6 133 * header (with pri, len, nxt, hlim, src, dst). 134 * This function may modify ver and hlim only. 135 * The mbuf chain containing the packet will be freed. 136 * The mbuf opt, if present, will not be freed. 137 * 138 * type of "mtu": rt_rmx.rmx_mtu is u_long, ifnet.ifr_mtu is int, and 139 * nd_ifinfo.linkmtu is u_int32_t. so we use u_long to hold largest one, 140 * which is rt_rmx.rmx_mtu. 141 */ 142 int 143 ip6_output(m0, opt, ro, flags, im6o, so, ifpp) 144 struct mbuf *m0; 145 struct ip6_pktopts *opt; 146 struct route_in6 *ro; 147 int flags; 148 struct ip6_moptions *im6o; 149 struct socket *so; 150 struct ifnet **ifpp; /* XXX: just for statistics */ 151 { 152 struct ip6_hdr *ip6, *mhip6; 153 struct ifnet *ifp, *origifp; 154 struct mbuf *m = m0; 155 int hlen, tlen, len, off; 156 struct route_in6 ip6route; 157 struct sockaddr_in6 *dst; 158 int error = 0; 159 u_long mtu; 160 int alwaysfrag, dontfrag; 161 u_int32_t optlen = 0, plen = 0, unfragpartlen = 0; 162 struct ip6_exthdrs exthdrs; 163 struct in6_addr finaldst; 164 struct route_in6 *ro_pmtu = NULL; 165 int hdrsplit = 0; 166 int needipsec = 0; 167 #ifdef IPSEC 168 int needipsectun = 0; 169 struct secpolicy *sp = NULL; 170 171 ip6 = mtod(m, struct ip6_hdr *); 172 #endif /* IPSEC */ 173 174 #define MAKE_EXTHDR(hp, mp) \ 175 do { \ 176 if (hp) { \ 177 struct ip6_ext *eh = (struct ip6_ext *)(hp); \ 178 error = ip6_copyexthdr((mp), (caddr_t)(hp), \ 179 ((eh)->ip6e_len + 1) << 3); \ 180 if (error) \ 181 goto freehdrs; \ 182 } \ 183 } while (/*CONSTCOND*/ 0) 184 185 bzero(&exthdrs, sizeof(exthdrs)); 186 if (opt) { 187 /* Hop-by-Hop options header */ 188 MAKE_EXTHDR(opt->ip6po_hbh, &exthdrs.ip6e_hbh); 189 /* Destination options header(1st part) */ 190 MAKE_EXTHDR(opt->ip6po_dest1, &exthdrs.ip6e_dest1); 191 /* Routing header */ 192 MAKE_EXTHDR(opt->ip6po_rthdr, &exthdrs.ip6e_rthdr); 193 /* Destination options header(2nd part) */ 194 MAKE_EXTHDR(opt->ip6po_dest2, &exthdrs.ip6e_dest2); 195 } 196 197 #ifdef IPSEC 198 if ((flags & IPV6_FORWARDING) != 0) { 199 needipsec = 0; 200 goto skippolicycheck; 201 } 202 203 /* get a security policy for this packet */ 204 if (so == NULL) 205 sp = ipsec6_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND, 0, &error); 206 else 207 sp = ipsec6_getpolicybysock(m, IPSEC_DIR_OUTBOUND, so, &error); 208 209 if (sp == NULL) { 210 ipsec6stat.out_inval++; 211 goto freehdrs; 212 } 213 214 error = 0; 215 216 /* check policy */ 217 switch (sp->policy) { 218 case IPSEC_POLICY_DISCARD: 219 /* 220 * This packet is just discarded. 221 */ 222 ipsec6stat.out_polvio++; 223 goto freehdrs; 224 225 case IPSEC_POLICY_BYPASS: 226 case IPSEC_POLICY_NONE: 227 /* no need to do IPsec. */ 228 needipsec = 0; 229 break; 230 231 case IPSEC_POLICY_IPSEC: 232 if (sp->req == NULL) { 233 /* XXX should be panic ? */ 234 printf("ip6_output: No IPsec request specified.\n"); 235 error = EINVAL; 236 goto freehdrs; 237 } 238 needipsec = 1; 239 break; 240 241 case IPSEC_POLICY_ENTRUST: 242 default: 243 printf("ip6_output: Invalid policy found. %d\n", sp->policy); 244 } 245 246 skippolicycheck:; 247 #endif /* IPSEC */ 248 249 /* 250 * Calculate the total length of the extension header chain. 251 * Keep the length of the unfragmentable part for fragmentation. 252 */ 253 optlen = 0; 254 if (exthdrs.ip6e_hbh) optlen += exthdrs.ip6e_hbh->m_len; 255 if (exthdrs.ip6e_dest1) optlen += exthdrs.ip6e_dest1->m_len; 256 if (exthdrs.ip6e_rthdr) optlen += exthdrs.ip6e_rthdr->m_len; 257 unfragpartlen = optlen + sizeof(struct ip6_hdr); 258 /* NOTE: we don't add AH/ESP length here. do that later. */ 259 if (exthdrs.ip6e_dest2) optlen += exthdrs.ip6e_dest2->m_len; 260 261 /* 262 * If we need IPsec, or there is at least one extension header, 263 * separate IP6 header from the payload. 264 */ 265 if ((needipsec || optlen) && !hdrsplit) { 266 if ((error = ip6_splithdr(m, &exthdrs)) != 0) { 267 m = NULL; 268 goto freehdrs; 269 } 270 m = exthdrs.ip6e_ip6; 271 hdrsplit++; 272 } 273 274 /* adjust pointer */ 275 ip6 = mtod(m, struct ip6_hdr *); 276 277 /* adjust mbuf packet header length */ 278 m->m_pkthdr.len += optlen; 279 plen = m->m_pkthdr.len - sizeof(*ip6); 280 281 /* If this is a jumbo payload, insert a jumbo payload option. */ 282 if (plen > IPV6_MAXPACKET) { 283 if (!hdrsplit) { 284 if ((error = ip6_splithdr(m, &exthdrs)) != 0) { 285 m = NULL; 286 goto freehdrs; 287 } 288 m = exthdrs.ip6e_ip6; 289 hdrsplit++; 290 } 291 /* adjust pointer */ 292 ip6 = mtod(m, struct ip6_hdr *); 293 if ((error = ip6_insert_jumboopt(&exthdrs, plen)) != 0) 294 goto freehdrs; 295 ip6->ip6_plen = 0; 296 } else 297 ip6->ip6_plen = htons(plen); 298 299 /* 300 * Concatenate headers and fill in next header fields. 301 * Here we have, on "m" 302 * IPv6 payload 303 * and we insert headers accordingly. Finally, we should be getting: 304 * IPv6 hbh dest1 rthdr ah* [esp* dest2 payload] 305 * 306 * during the header composing process, "m" points to IPv6 header. 307 * "mprev" points to an extension header prior to esp. 308 */ 309 { 310 u_char *nexthdrp = &ip6->ip6_nxt; 311 struct mbuf *mprev = m; 312 313 /* 314 * we treat dest2 specially. this makes IPsec processing 315 * much easier. the goal here is to make mprev point the 316 * mbuf prior to dest2. 317 * 318 * result: IPv6 dest2 payload 319 * m and mprev will point to IPv6 header. 320 */ 321 if (exthdrs.ip6e_dest2) { 322 if (!hdrsplit) 323 panic("assumption failed: hdr not split"); 324 exthdrs.ip6e_dest2->m_next = m->m_next; 325 m->m_next = exthdrs.ip6e_dest2; 326 *mtod(exthdrs.ip6e_dest2, u_char *) = ip6->ip6_nxt; 327 ip6->ip6_nxt = IPPROTO_DSTOPTS; 328 } 329 330 #define MAKE_CHAIN(m, mp, p, i)\ 331 do {\ 332 if (m) {\ 333 if (!hdrsplit) \ 334 panic("assumption failed: hdr not split"); \ 335 *mtod((m), u_char *) = *(p);\ 336 *(p) = (i);\ 337 p = mtod((m), u_char *);\ 338 (m)->m_next = (mp)->m_next;\ 339 (mp)->m_next = (m);\ 340 (mp) = (m);\ 341 }\ 342 } while (/*CONSTCOND*/ 0) 343 /* 344 * result: IPv6 hbh dest1 rthdr dest2 payload 345 * m will point to IPv6 header. mprev will point to the 346 * extension header prior to dest2 (rthdr in the above case). 347 */ 348 MAKE_CHAIN(exthdrs.ip6e_hbh, mprev, nexthdrp, IPPROTO_HOPOPTS); 349 MAKE_CHAIN(exthdrs.ip6e_dest1, mprev, nexthdrp, 350 IPPROTO_DSTOPTS); 351 MAKE_CHAIN(exthdrs.ip6e_rthdr, mprev, nexthdrp, 352 IPPROTO_ROUTING); 353 354 #ifdef IPSEC 355 if (!needipsec) 356 goto skip_ipsec2; 357 358 /* 359 * pointers after IPsec headers are not valid any more. 360 * other pointers need a great care too. 361 * (IPsec routines should not mangle mbufs prior to AH/ESP) 362 */ 363 exthdrs.ip6e_dest2 = NULL; 364 365 { 366 struct ip6_rthdr *rh = NULL; 367 int segleft_org = 0; 368 struct ipsec_output_state state; 369 370 if (exthdrs.ip6e_rthdr) { 371 rh = mtod(exthdrs.ip6e_rthdr, struct ip6_rthdr *); 372 segleft_org = rh->ip6r_segleft; 373 rh->ip6r_segleft = 0; 374 } 375 376 bzero(&state, sizeof(state)); 377 state.m = m; 378 error = ipsec6_output_trans(&state, nexthdrp, mprev, sp, flags, 379 &needipsectun); 380 m = state.m; 381 if (error) { 382 /* mbuf is already reclaimed in ipsec6_output_trans. */ 383 m = NULL; 384 switch (error) { 385 case EHOSTUNREACH: 386 case ENETUNREACH: 387 case EMSGSIZE: 388 case ENOBUFS: 389 case ENOMEM: 390 break; 391 default: 392 printf("ip6_output (ipsec): error code %d\n", error); 393 /* FALLTHROUGH */ 394 case ENOENT: 395 /* don't show these error codes to the user */ 396 error = 0; 397 break; 398 } 399 goto bad; 400 } 401 if (exthdrs.ip6e_rthdr) { 402 /* ah6_output doesn't modify mbuf chain */ 403 rh->ip6r_segleft = segleft_org; 404 } 405 } 406 skip_ipsec2:; 407 #endif 408 } 409 410 /* 411 * If there is a routing header, replace destination address field 412 * with the first hop of the routing header. 413 */ 414 if (exthdrs.ip6e_rthdr) { 415 struct ip6_rthdr *rh; 416 struct ip6_rthdr0 *rh0; 417 struct in6_addr *addr; 418 419 rh = (struct ip6_rthdr *)(mtod(exthdrs.ip6e_rthdr, 420 struct ip6_rthdr *)); 421 finaldst = ip6->ip6_dst; 422 switch (rh->ip6r_type) { 423 case IPV6_RTHDR_TYPE_0: 424 rh0 = (struct ip6_rthdr0 *)rh; 425 addr = (struct in6_addr *)(rh0 + 1); 426 ip6->ip6_dst = addr[0]; 427 bcopy(&addr[1], &addr[0], 428 sizeof(struct in6_addr) * (rh0->ip6r0_segleft - 1)); 429 addr[rh0->ip6r0_segleft - 1] = finaldst; 430 break; 431 default: /* is it possible? */ 432 error = EINVAL; 433 goto bad; 434 } 435 } 436 437 /* Source address validation */ 438 if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) && 439 (flags & IPV6_UNSPECSRC) == 0) { 440 error = EOPNOTSUPP; 441 ip6stat.ip6s_badscope++; 442 goto bad; 443 } 444 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) { 445 error = EOPNOTSUPP; 446 ip6stat.ip6s_badscope++; 447 goto bad; 448 } 449 450 ip6stat.ip6s_localout++; 451 452 /* 453 * Route packet. 454 */ 455 /* initialize cached route */ 456 if (ro == 0) { 457 ro = &ip6route; 458 bzero((caddr_t)ro, sizeof(*ro)); 459 } 460 ro_pmtu = ro; 461 if (opt && opt->ip6po_rthdr) 462 ro = &opt->ip6po_route; 463 dst = (struct sockaddr_in6 *)&ro->ro_dst; 464 /* 465 * If there is a cached route, 466 * check that it is to the same destination 467 * and is still up. If not, free it and try again. 468 */ 469 if (ro->ro_rt && ((ro->ro_rt->rt_flags & RTF_UP) == 0 || 470 dst->sin6_family != AF_INET6 || 471 !IN6_ARE_ADDR_EQUAL(&dst->sin6_addr, &ip6->ip6_dst))) { 472 RTFREE(ro->ro_rt); 473 ro->ro_rt = (struct rtentry *)0; 474 } 475 if (ro->ro_rt == 0) { 476 bzero(dst, sizeof(*dst)); 477 dst->sin6_family = AF_INET6; 478 dst->sin6_len = sizeof(struct sockaddr_in6); 479 dst->sin6_addr = ip6->ip6_dst; 480 } 481 #ifdef IPSEC 482 if (needipsec && needipsectun) { 483 struct ipsec_output_state state; 484 485 /* 486 * All the extension headers will become inaccessible 487 * (since they can be encrypted). 488 * Don't panic, we need no more updates to extension headers 489 * on inner IPv6 packet (since they are now encapsulated). 490 * 491 * IPv6 [ESP|AH] IPv6 [extension headers] payload 492 */ 493 bzero(&exthdrs, sizeof(exthdrs)); 494 exthdrs.ip6e_ip6 = m; 495 496 bzero(&state, sizeof(state)); 497 state.m = m; 498 state.ro = (struct route *)ro; 499 state.dst = (struct sockaddr *)dst; 500 501 error = ipsec6_output_tunnel(&state, sp, flags); 502 503 m = state.m; 504 ro = (struct route_in6 *)state.ro; 505 dst = (struct sockaddr_in6 *)state.dst; 506 if (error) { 507 /* mbuf is already reclaimed in ipsec6_output_tunnel. */ 508 m0 = m = NULL; 509 m = NULL; 510 switch (error) { 511 case EHOSTUNREACH: 512 case ENETUNREACH: 513 case EMSGSIZE: 514 case ENOBUFS: 515 case ENOMEM: 516 break; 517 default: 518 printf("ip6_output (ipsec): error code %d\n", error); 519 /* FALLTHROUGH */ 520 case ENOENT: 521 /* don't show these error codes to the user */ 522 error = 0; 523 break; 524 } 525 goto bad; 526 } 527 528 exthdrs.ip6e_ip6 = m; 529 } 530 #endif /* IPSEC */ 531 532 if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) { 533 /* Unicast */ 534 535 #define ifatoia6(ifa) ((struct in6_ifaddr *)(ifa)) 536 #define sin6tosa(sin6) ((struct sockaddr *)(sin6)) 537 /* xxx 538 * interface selection comes here 539 * if an interface is specified from an upper layer, 540 * ifp must point it. 541 */ 542 if (ro->ro_rt == 0) { 543 /* 544 * non-bsdi always clone routes, if parent is 545 * PRF_CLONING. 546 */ 547 rtalloc((struct route *)ro); 548 } 549 if (ro->ro_rt == 0) { 550 ip6stat.ip6s_noroute++; 551 error = EHOSTUNREACH; 552 /* XXX in6_ifstat_inc(ifp, ifs6_out_discard); */ 553 goto bad; 554 } 555 ifp = ro->ro_rt->rt_ifp; 556 ro->ro_rt->rt_use++; 557 if (ro->ro_rt->rt_flags & RTF_GATEWAY) 558 dst = (struct sockaddr_in6 *)ro->ro_rt->rt_gateway; 559 m->m_flags &= ~(M_BCAST | M_MCAST); /* just in case */ 560 561 in6_ifstat_inc(ifp, ifs6_out_request); 562 563 /* 564 * Check if the outgoing interface conflicts with 565 * the interface specified by ifi6_ifindex (if specified). 566 * Note that loopback interface is always okay. 567 * (this may happen when we are sending a packet to one of 568 * our own addresses.) 569 */ 570 if (opt && opt->ip6po_pktinfo 571 && opt->ip6po_pktinfo->ipi6_ifindex) { 572 if (!(ifp->if_flags & IFF_LOOPBACK) 573 && ifp->if_index != opt->ip6po_pktinfo->ipi6_ifindex) { 574 ip6stat.ip6s_noroute++; 575 in6_ifstat_inc(ifp, ifs6_out_discard); 576 error = EHOSTUNREACH; 577 goto bad; 578 } 579 } 580 581 if (opt && opt->ip6po_hlim != -1) 582 ip6->ip6_hlim = opt->ip6po_hlim & 0xff; 583 } else { 584 /* Multicast */ 585 struct in6_multi *in6m; 586 587 m->m_flags = (m->m_flags & ~M_BCAST) | M_MCAST; 588 589 /* 590 * See if the caller provided any multicast options 591 */ 592 ifp = NULL; 593 if (im6o != NULL) { 594 ip6->ip6_hlim = im6o->im6o_multicast_hlim; 595 if (im6o->im6o_multicast_ifp != NULL) 596 ifp = im6o->im6o_multicast_ifp; 597 } else 598 ip6->ip6_hlim = ip6_defmcasthlim; 599 600 /* 601 * See if the caller provided the outgoing interface 602 * as an ancillary data. 603 * Boundary check for ifindex is assumed to be already done. 604 */ 605 if (opt && opt->ip6po_pktinfo && opt->ip6po_pktinfo->ipi6_ifindex) 606 ifp = ifindex2ifnet[opt->ip6po_pktinfo->ipi6_ifindex]; 607 608 /* 609 * If the destination is a node-local scope multicast, 610 * the packet should be loop-backed only. 611 */ 612 if (IN6_IS_ADDR_MC_NODELOCAL(&ip6->ip6_dst)) { 613 /* 614 * If the outgoing interface is already specified, 615 * it should be a loopback interface. 616 */ 617 if (ifp && (ifp->if_flags & IFF_LOOPBACK) == 0) { 618 ip6stat.ip6s_badscope++; 619 error = ENETUNREACH; /* XXX: better error? */ 620 /* XXX correct ifp? */ 621 in6_ifstat_inc(ifp, ifs6_out_discard); 622 goto bad; 623 } else { 624 ifp = &loif[0]; 625 } 626 } 627 628 if (opt && opt->ip6po_hlim != -1) 629 ip6->ip6_hlim = opt->ip6po_hlim & 0xff; 630 631 /* 632 * If caller did not provide an interface lookup a 633 * default in the routing table. This is either a 634 * default for the speicfied group (i.e. a host 635 * route), or a multicast default (a route for the 636 * ``net'' ff00::/8). 637 */ 638 if (ifp == NULL) { 639 if (ro->ro_rt == 0) { 640 ro->ro_rt = rtalloc1((struct sockaddr *) 641 &ro->ro_dst, 0); 642 } 643 if (ro->ro_rt == 0) { 644 ip6stat.ip6s_noroute++; 645 error = EHOSTUNREACH; 646 /* XXX in6_ifstat_inc(ifp, ifs6_out_discard) */ 647 goto bad; 648 } 649 ifp = ro->ro_rt->rt_ifp; 650 ro->ro_rt->rt_use++; 651 } 652 653 if ((flags & IPV6_FORWARDING) == 0) 654 in6_ifstat_inc(ifp, ifs6_out_request); 655 in6_ifstat_inc(ifp, ifs6_out_mcast); 656 657 /* 658 * Confirm that the outgoing interface supports multicast. 659 */ 660 if ((ifp->if_flags & IFF_MULTICAST) == 0) { 661 ip6stat.ip6s_noroute++; 662 in6_ifstat_inc(ifp, ifs6_out_discard); 663 error = ENETUNREACH; 664 goto bad; 665 } 666 IN6_LOOKUP_MULTI(ip6->ip6_dst, ifp, in6m); 667 if (in6m != NULL && 668 (im6o == NULL || im6o->im6o_multicast_loop)) { 669 /* 670 * If we belong to the destination multicast group 671 * on the outgoing interface, and the caller did not 672 * forbid loopback, loop back a copy. 673 */ 674 ip6_mloopback(ifp, m, dst); 675 } else { 676 /* 677 * If we are acting as a multicast router, perform 678 * multicast forwarding as if the packet had just 679 * arrived on the interface to which we are about 680 * to send. The multicast forwarding function 681 * recursively calls this function, using the 682 * IPV6_FORWARDING flag to prevent infinite recursion. 683 * 684 * Multicasts that are looped back by ip6_mloopback(), 685 * above, will be forwarded by the ip6_input() routine, 686 * if necessary. 687 */ 688 if (ip6_mrouter && (flags & IPV6_FORWARDING) == 0) { 689 if (ip6_mforward(ip6, ifp, m) != 0) { 690 m_freem(m); 691 goto done; 692 } 693 } 694 } 695 /* 696 * Multicasts with a hoplimit of zero may be looped back, 697 * above, but must not be transmitted on a network. 698 * Also, multicasts addressed to the loopback interface 699 * are not sent -- the above call to ip6_mloopback() will 700 * loop back a copy if this host actually belongs to the 701 * destination group on the loopback interface. 702 */ 703 if (ip6->ip6_hlim == 0 || (ifp->if_flags & IFF_LOOPBACK)) { 704 m_freem(m); 705 goto done; 706 } 707 } 708 709 /* 710 * Fill the outgoing inteface to tell the upper layer 711 * to increment per-interface statistics. 712 */ 713 if (ifpp) 714 *ifpp = ifp; 715 716 /* Determine path MTU. */ 717 if ((error = ip6_getpmtu(ro_pmtu, ro, ifp, &finaldst, &mtu, 718 &alwaysfrag)) != 0) 719 goto bad; 720 721 /* 722 * The caller of this function may specify to use the minimum MTU 723 * in some cases. 724 */ 725 if (mtu > IPV6_MMTU) { 726 if ((flags & IPV6_MINMTU)) 727 mtu = IPV6_MMTU; 728 } 729 730 /* Fake scoped addresses */ 731 if ((ifp->if_flags & IFF_LOOPBACK) != 0) { 732 /* 733 * If source or destination address is a scoped address, and 734 * the packet is going to be sent to a loopback interface, 735 * we should keep the original interface. 736 */ 737 738 /* 739 * XXX: this is a very experimental and temporary solution. 740 * We eventually have sockaddr_in6 and use the sin6_scope_id 741 * field of the structure here. 742 * We rely on the consistency between two scope zone ids 743 * of source add destination, which should already be assured 744 * Larger scopes than link will be supported in the near 745 * future. 746 */ 747 origifp = NULL; 748 if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src)) 749 origifp = ifindex2ifnet[ntohs(ip6->ip6_src.s6_addr16[1])]; 750 else if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst)) 751 origifp = ifindex2ifnet[ntohs(ip6->ip6_dst.s6_addr16[1])]; 752 /* 753 * XXX: origifp can be NULL even in those two cases above. 754 * For example, if we remove the (only) link-local address 755 * from the loopback interface, and try to send a link-local 756 * address without link-id information. Then the source 757 * address is ::1, and the destination address is the 758 * link-local address with its s6_addr16[1] being zero. 759 * What is worse, if the packet goes to the loopback interface 760 * by a default rejected route, the null pointer would be 761 * passed to looutput, and the kernel would hang. 762 * The following last resort would prevent such disaster. 763 */ 764 if (origifp == NULL) 765 origifp = ifp; 766 } else 767 origifp = ifp; 768 if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src)) 769 ip6->ip6_src.s6_addr16[1] = 0; 770 if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst)) 771 ip6->ip6_dst.s6_addr16[1] = 0; 772 773 /* 774 * If the outgoing packet contains a hop-by-hop options header, 775 * it must be examined and processed even by the source node. 776 * (RFC 2460, section 4.) 777 */ 778 if (exthdrs.ip6e_hbh) { 779 struct ip6_hbh *hbh = mtod(exthdrs.ip6e_hbh, struct ip6_hbh *); 780 u_int32_t dummy1; /* XXX unused */ 781 u_int32_t dummy2; /* XXX unused */ 782 783 /* 784 * XXX: if we have to send an ICMPv6 error to the sender, 785 * we need the M_LOOP flag since icmp6_error() expects 786 * the IPv6 and the hop-by-hop options header are 787 * continuous unless the flag is set. 788 */ 789 m->m_flags |= M_LOOP; 790 m->m_pkthdr.rcvif = ifp; 791 if (ip6_process_hopopts(m, (u_int8_t *)(hbh + 1), 792 ((hbh->ip6h_len + 1) << 3) - sizeof(struct ip6_hbh), 793 &dummy1, &dummy2) < 0) { 794 /* m was already freed at this point */ 795 error = EINVAL;/* better error? */ 796 goto done; 797 } 798 m->m_flags &= ~M_LOOP; /* XXX */ 799 m->m_pkthdr.rcvif = NULL; 800 } 801 802 #ifdef PFIL_HOOKS 803 /* 804 * Run through list of hooks for output packets. 805 */ 806 if ((error = pfil_run_hooks(&inet6_pfil_hook, &m, ifp, PFIL_OUT)) != 0) 807 goto done; 808 if (m == NULL) 809 goto done; 810 ip6 = mtod(m, struct ip6_hdr *); 811 #endif /* PFIL_HOOKS */ 812 /* 813 * Send the packet to the outgoing interface. 814 * If necessary, do IPv6 fragmentation before sending. 815 * 816 * the logic here is rather complex: 817 * 1: normal case (dontfrag == 0, alwaysfrag == 0) 818 * 1-a: send as is if tlen <= path mtu 819 * 1-b: fragment if tlen > path mtu 820 * 821 * 2: if user asks us not to fragment (dontfrag == 1) 822 * 2-a: send as is if tlen <= interface mtu 823 * 2-b: error if tlen > interface mtu 824 * 825 * 3: if we always need to attach fragment header (alwaysfrag == 1) 826 * always fragment 827 * 828 * 4: if dontfrag == 1 && alwaysfrag == 1 829 * error, as we cannot handle this conflicting request 830 */ 831 tlen = m->m_pkthdr.len; 832 833 dontfrag = 0; 834 if (dontfrag && alwaysfrag) { /* case 4 */ 835 /* conflicting request - can't transmit */ 836 error = EMSGSIZE; 837 goto bad; 838 } 839 if (dontfrag && tlen > IN6_LINKMTU(ifp)) { /* case 2-b */ 840 /* 841 * Even if the DONTFRAG option is specified, we cannot send the 842 * packet when the data length is larger than the MTU of the 843 * outgoing interface. 844 * Notify the error by sending IPV6_PATHMTU ancillary data as 845 * well as returning an error code (the latter is not described 846 * in the API spec.) 847 */ 848 u_int32_t mtu32; 849 struct ip6ctlparam ip6cp; 850 851 mtu32 = (u_int32_t)mtu; 852 bzero(&ip6cp, sizeof(ip6cp)); 853 ip6cp.ip6c_cmdarg = (void *)&mtu32; 854 pfctlinput2(PRC_MSGSIZE, (struct sockaddr *)&ro_pmtu->ro_dst, 855 (void *)&ip6cp); 856 857 error = EMSGSIZE; 858 goto bad; 859 } 860 861 /* 862 * transmit packet without fragmentation 863 */ 864 if (dontfrag || (!alwaysfrag && tlen <= mtu)) { /* case 1-a and 2-a */ 865 struct in6_ifaddr *ia6; 866 867 ip6 = mtod(m, struct ip6_hdr *); 868 ia6 = in6_ifawithifp(ifp, &ip6->ip6_src); 869 if (ia6) { 870 /* Record statistics for this interface address. */ 871 ia6->ia_ifa.ifa_data.ifad_outbytes += m->m_pkthdr.len; 872 } 873 #ifdef IPSEC 874 /* clean ipsec history once it goes out of the node */ 875 ipsec_delaux(m); 876 #endif 877 error = nd6_output(ifp, origifp, m, dst, ro->ro_rt); 878 goto done; 879 } 880 881 /* 882 * try to fragment the packet. case 1-b and 3 883 */ 884 if (mtu < IPV6_MMTU) { 885 /* path MTU cannot be less than IPV6_MMTU */ 886 error = EMSGSIZE; 887 in6_ifstat_inc(ifp, ifs6_out_fragfail); 888 goto bad; 889 } else if (ip6->ip6_plen == 0) { 890 /* jumbo payload cannot be fragmented */ 891 error = EMSGSIZE; 892 in6_ifstat_inc(ifp, ifs6_out_fragfail); 893 goto bad; 894 } else { 895 struct mbuf **mnext, *m_frgpart; 896 struct ip6_frag *ip6f; 897 u_int32_t id = htonl(ip6_randomid()); 898 u_char nextproto; 899 struct ip6ctlparam ip6cp; 900 u_int32_t mtu32; 901 902 /* 903 * Too large for the destination or interface; 904 * fragment if possible. 905 * Must be able to put at least 8 bytes per fragment. 906 */ 907 hlen = unfragpartlen; 908 if (mtu > IPV6_MAXPACKET) 909 mtu = IPV6_MAXPACKET; 910 911 /* Notify a proper path MTU to applications. */ 912 mtu32 = (u_int32_t)mtu; 913 bzero(&ip6cp, sizeof(ip6cp)); 914 ip6cp.ip6c_cmdarg = (void *)&mtu32; 915 pfctlinput2(PRC_MSGSIZE, (struct sockaddr *)&ro_pmtu->ro_dst, 916 (void *)&ip6cp); 917 918 len = (mtu - hlen - sizeof(struct ip6_frag)) & ~7; 919 if (len < 8) { 920 error = EMSGSIZE; 921 in6_ifstat_inc(ifp, ifs6_out_fragfail); 922 goto bad; 923 } 924 925 mnext = &m->m_nextpkt; 926 927 /* 928 * Change the next header field of the last header in the 929 * unfragmentable part. 930 */ 931 if (exthdrs.ip6e_rthdr) { 932 nextproto = *mtod(exthdrs.ip6e_rthdr, u_char *); 933 *mtod(exthdrs.ip6e_rthdr, u_char *) = IPPROTO_FRAGMENT; 934 } else if (exthdrs.ip6e_dest1) { 935 nextproto = *mtod(exthdrs.ip6e_dest1, u_char *); 936 *mtod(exthdrs.ip6e_dest1, u_char *) = IPPROTO_FRAGMENT; 937 } else if (exthdrs.ip6e_hbh) { 938 nextproto = *mtod(exthdrs.ip6e_hbh, u_char *); 939 *mtod(exthdrs.ip6e_hbh, u_char *) = IPPROTO_FRAGMENT; 940 } else { 941 nextproto = ip6->ip6_nxt; 942 ip6->ip6_nxt = IPPROTO_FRAGMENT; 943 } 944 945 /* 946 * Loop through length of segment after first fragment, 947 * make new header and copy data of each part and link onto 948 * chain. 949 */ 950 m0 = m; 951 for (off = hlen; off < tlen; off += len) { 952 struct mbuf *mlast; 953 954 MGETHDR(m, M_DONTWAIT, MT_HEADER); 955 if (!m) { 956 error = ENOBUFS; 957 ip6stat.ip6s_odropped++; 958 goto sendorfree; 959 } 960 m->m_pkthdr.rcvif = NULL; 961 m->m_flags = m0->m_flags & M_COPYFLAGS; 962 *mnext = m; 963 mnext = &m->m_nextpkt; 964 m->m_data += max_linkhdr; 965 mhip6 = mtod(m, struct ip6_hdr *); 966 *mhip6 = *ip6; 967 m->m_len = sizeof(*mhip6); 968 error = ip6_insertfraghdr(m0, m, hlen, &ip6f); 969 if (error) { 970 ip6stat.ip6s_odropped++; 971 goto sendorfree; 972 } 973 ip6f->ip6f_offlg = htons((u_int16_t)((off - hlen) & ~7)); 974 if (off + len >= tlen) 975 len = tlen - off; 976 else 977 ip6f->ip6f_offlg |= IP6F_MORE_FRAG; 978 mhip6->ip6_plen = htons((u_int16_t)(len + hlen + 979 sizeof(*ip6f) - sizeof(struct ip6_hdr))); 980 if ((m_frgpart = m_copy(m0, off, len)) == 0) { 981 error = ENOBUFS; 982 ip6stat.ip6s_odropped++; 983 goto sendorfree; 984 } 985 for (mlast = m; mlast->m_next; mlast = mlast->m_next) 986 ; 987 mlast->m_next = m_frgpart; 988 m->m_pkthdr.len = len + hlen + sizeof(*ip6f); 989 m->m_pkthdr.rcvif = (struct ifnet *)0; 990 ip6f->ip6f_reserved = 0; 991 ip6f->ip6f_ident = id; 992 ip6f->ip6f_nxt = nextproto; 993 ip6stat.ip6s_ofragments++; 994 in6_ifstat_inc(ifp, ifs6_out_fragcreat); 995 } 996 997 in6_ifstat_inc(ifp, ifs6_out_fragok); 998 } 999 1000 /* 1001 * Remove leading garbages. 1002 */ 1003 sendorfree: 1004 m = m0->m_nextpkt; 1005 m0->m_nextpkt = 0; 1006 m_freem(m0); 1007 for (m0 = m; m; m = m0) { 1008 m0 = m->m_nextpkt; 1009 m->m_nextpkt = 0; 1010 if (error == 0) { 1011 struct in6_ifaddr *ia6; 1012 ip6 = mtod(m, struct ip6_hdr *); 1013 ia6 = in6_ifawithifp(ifp, &ip6->ip6_src); 1014 if (ia6) { 1015 /* 1016 * Record statistics for this interface 1017 * address. 1018 */ 1019 ia6->ia_ifa.ifa_data.ifad_outbytes += 1020 m->m_pkthdr.len; 1021 } 1022 #ifdef IPSEC 1023 /* clean ipsec history once it goes out of the node */ 1024 ipsec_delaux(m); 1025 #endif 1026 error = nd6_output(ifp, origifp, m, dst, ro->ro_rt); 1027 } else 1028 m_freem(m); 1029 } 1030 1031 if (error == 0) 1032 ip6stat.ip6s_fragmented++; 1033 1034 done: 1035 if (ro == &ip6route && ro->ro_rt) { /* brace necessary for RTFREE */ 1036 RTFREE(ro->ro_rt); 1037 } else if (ro_pmtu == &ip6route && ro_pmtu->ro_rt) { 1038 RTFREE(ro_pmtu->ro_rt); 1039 } 1040 1041 #ifdef IPSEC 1042 if (sp != NULL) 1043 key_freesp(sp); 1044 #endif /* IPSEC */ 1045 1046 return (error); 1047 1048 freehdrs: 1049 m_freem(exthdrs.ip6e_hbh); /* m_freem will check if mbuf is 0 */ 1050 m_freem(exthdrs.ip6e_dest1); 1051 m_freem(exthdrs.ip6e_rthdr); 1052 m_freem(exthdrs.ip6e_dest2); 1053 /* FALLTHROUGH */ 1054 bad: 1055 m_freem(m); 1056 goto done; 1057 } 1058 1059 static int 1060 ip6_copyexthdr(mp, hdr, hlen) 1061 struct mbuf **mp; 1062 caddr_t hdr; 1063 int hlen; 1064 { 1065 struct mbuf *m; 1066 1067 if (hlen > MCLBYTES) 1068 return (ENOBUFS); /* XXX */ 1069 1070 MGET(m, M_DONTWAIT, MT_DATA); 1071 if (!m) 1072 return (ENOBUFS); 1073 1074 if (hlen > MLEN) { 1075 MCLGET(m, M_DONTWAIT); 1076 if ((m->m_flags & M_EXT) == 0) { 1077 m_free(m); 1078 return (ENOBUFS); 1079 } 1080 } 1081 m->m_len = hlen; 1082 if (hdr) 1083 bcopy(hdr, mtod(m, caddr_t), hlen); 1084 1085 *mp = m; 1086 return (0); 1087 } 1088 1089 /* 1090 * Insert jumbo payload option. 1091 */ 1092 static int 1093 ip6_insert_jumboopt(exthdrs, plen) 1094 struct ip6_exthdrs *exthdrs; 1095 u_int32_t plen; 1096 { 1097 struct mbuf *mopt; 1098 u_int8_t *optbuf; 1099 u_int32_t v; 1100 1101 #define JUMBOOPTLEN 8 /* length of jumbo payload option and padding */ 1102 1103 /* 1104 * If there is no hop-by-hop options header, allocate new one. 1105 * If there is one but it doesn't have enough space to store the 1106 * jumbo payload option, allocate a cluster to store the whole options. 1107 * Otherwise, use it to store the options. 1108 */ 1109 if (exthdrs->ip6e_hbh == 0) { 1110 MGET(mopt, M_DONTWAIT, MT_DATA); 1111 if (mopt == 0) 1112 return (ENOBUFS); 1113 mopt->m_len = JUMBOOPTLEN; 1114 optbuf = mtod(mopt, u_int8_t *); 1115 optbuf[1] = 0; /* = ((JUMBOOPTLEN) >> 3) - 1 */ 1116 exthdrs->ip6e_hbh = mopt; 1117 } else { 1118 struct ip6_hbh *hbh; 1119 1120 mopt = exthdrs->ip6e_hbh; 1121 if (M_TRAILINGSPACE(mopt) < JUMBOOPTLEN) { 1122 /* 1123 * XXX assumption: 1124 * - exthdrs->ip6e_hbh is not referenced from places 1125 * other than exthdrs. 1126 * - exthdrs->ip6e_hbh is not an mbuf chain. 1127 */ 1128 int oldoptlen = mopt->m_len; 1129 struct mbuf *n; 1130 1131 /* 1132 * XXX: give up if the whole (new) hbh header does 1133 * not fit even in an mbuf cluster. 1134 */ 1135 if (oldoptlen + JUMBOOPTLEN > MCLBYTES) 1136 return (ENOBUFS); 1137 1138 /* 1139 * As a consequence, we must always prepare a cluster 1140 * at this point. 1141 */ 1142 MGET(n, M_DONTWAIT, MT_DATA); 1143 if (n) { 1144 MCLGET(n, M_DONTWAIT); 1145 if ((n->m_flags & M_EXT) == 0) { 1146 m_freem(n); 1147 n = NULL; 1148 } 1149 } 1150 if (!n) 1151 return (ENOBUFS); 1152 n->m_len = oldoptlen + JUMBOOPTLEN; 1153 bcopy(mtod(mopt, caddr_t), mtod(n, caddr_t), 1154 oldoptlen); 1155 optbuf = mtod(n, u_int8_t *) + oldoptlen; 1156 m_freem(mopt); 1157 mopt = exthdrs->ip6e_hbh = n; 1158 } else { 1159 optbuf = mtod(mopt, u_int8_t *) + mopt->m_len; 1160 mopt->m_len += JUMBOOPTLEN; 1161 } 1162 optbuf[0] = IP6OPT_PADN; 1163 optbuf[1] = 0; 1164 1165 /* 1166 * Adjust the header length according to the pad and 1167 * the jumbo payload option. 1168 */ 1169 hbh = mtod(mopt, struct ip6_hbh *); 1170 hbh->ip6h_len += (JUMBOOPTLEN >> 3); 1171 } 1172 1173 /* fill in the option. */ 1174 optbuf[2] = IP6OPT_JUMBO; 1175 optbuf[3] = 4; 1176 v = (u_int32_t)htonl(plen + JUMBOOPTLEN); 1177 bcopy(&v, &optbuf[4], sizeof(u_int32_t)); 1178 1179 /* finally, adjust the packet header length */ 1180 exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN; 1181 1182 return (0); 1183 #undef JUMBOOPTLEN 1184 } 1185 1186 /* 1187 * Insert fragment header and copy unfragmentable header portions. 1188 */ 1189 static int 1190 ip6_insertfraghdr(m0, m, hlen, frghdrp) 1191 struct mbuf *m0, *m; 1192 int hlen; 1193 struct ip6_frag **frghdrp; 1194 { 1195 struct mbuf *n, *mlast; 1196 1197 if (hlen > sizeof(struct ip6_hdr)) { 1198 n = m_copym(m0, sizeof(struct ip6_hdr), 1199 hlen - sizeof(struct ip6_hdr), M_DONTWAIT); 1200 if (n == 0) 1201 return (ENOBUFS); 1202 m->m_next = n; 1203 } else 1204 n = m; 1205 1206 /* Search for the last mbuf of unfragmentable part. */ 1207 for (mlast = n; mlast->m_next; mlast = mlast->m_next) 1208 ; 1209 1210 if ((mlast->m_flags & M_EXT) == 0 && 1211 M_TRAILINGSPACE(mlast) >= sizeof(struct ip6_frag)) { 1212 /* use the trailing space of the last mbuf for the fragment hdr */ 1213 *frghdrp = (struct ip6_frag *)(mtod(mlast, caddr_t) + 1214 mlast->m_len); 1215 mlast->m_len += sizeof(struct ip6_frag); 1216 m->m_pkthdr.len += sizeof(struct ip6_frag); 1217 } else { 1218 /* allocate a new mbuf for the fragment header */ 1219 struct mbuf *mfrg; 1220 1221 MGET(mfrg, M_DONTWAIT, MT_DATA); 1222 if (mfrg == 0) 1223 return (ENOBUFS); 1224 mfrg->m_len = sizeof(struct ip6_frag); 1225 *frghdrp = mtod(mfrg, struct ip6_frag *); 1226 mlast->m_next = mfrg; 1227 } 1228 1229 return (0); 1230 } 1231 1232 static int 1233 ip6_getpmtu(ro_pmtu, ro, ifp, dst, mtup, alwaysfragp) 1234 struct route_in6 *ro_pmtu, *ro; 1235 struct ifnet *ifp; 1236 struct in6_addr *dst; 1237 u_long *mtup; 1238 int *alwaysfragp; 1239 { 1240 u_int32_t mtu = 0; 1241 int alwaysfrag = 0; 1242 int error = 0; 1243 1244 if (ro_pmtu != ro) { 1245 /* The first hop and the final destination may differ. */ 1246 struct sockaddr_in6 *sa6_dst = 1247 (struct sockaddr_in6 *)&ro_pmtu->ro_dst; 1248 if (ro_pmtu->ro_rt && 1249 ((ro_pmtu->ro_rt->rt_flags & RTF_UP) == 0 || 1250 !IN6_ARE_ADDR_EQUAL(&sa6_dst->sin6_addr, dst))) { 1251 RTFREE(ro_pmtu->ro_rt); 1252 ro_pmtu->ro_rt = (struct rtentry *)NULL; 1253 } 1254 if (ro_pmtu->ro_rt == NULL) { 1255 bzero(sa6_dst, sizeof(*sa6_dst)); /* for safety */ 1256 sa6_dst->sin6_family = AF_INET6; 1257 sa6_dst->sin6_len = sizeof(struct sockaddr_in6); 1258 sa6_dst->sin6_addr = *dst; 1259 1260 rtalloc((struct route *)ro_pmtu); 1261 } 1262 } 1263 if (ro_pmtu->ro_rt) { 1264 u_int32_t ifmtu; 1265 1266 if (ifp == NULL) 1267 ifp = ro_pmtu->ro_rt->rt_ifp; 1268 ifmtu = IN6_LINKMTU(ifp); 1269 mtu = ro_pmtu->ro_rt->rt_rmx.rmx_mtu; 1270 if (mtu == 0) 1271 mtu = ifmtu; 1272 else if (mtu < IPV6_MMTU) { 1273 /* 1274 * RFC2460 section 5, last paragraph: 1275 * if we record ICMPv6 too big message with 1276 * mtu < IPV6_MMTU, transmit packets sized IPV6_MMTU 1277 * or smaller, with fragment header attached. 1278 * (fragment header is needed regardless from the 1279 * packet size, for translators to identify packets) 1280 */ 1281 alwaysfrag = 1; 1282 mtu = IPV6_MMTU; 1283 } else if (mtu > ifmtu) { 1284 /* 1285 * The MTU on the route is larger than the MTU on 1286 * the interface! This shouldn't happen, unless the 1287 * MTU of the interface has been changed after the 1288 * interface was brought up. Change the MTU in the 1289 * route to match the interface MTU (as long as the 1290 * field isn't locked). 1291 */ 1292 mtu = ifmtu; 1293 if (!(ro_pmtu->ro_rt->rt_rmx.rmx_locks & RTV_MTU)) 1294 ro_pmtu->ro_rt->rt_rmx.rmx_mtu = mtu; 1295 } 1296 } else if (ifp) { 1297 mtu = IN6_LINKMTU(ifp); 1298 } else 1299 error = EHOSTUNREACH; /* XXX */ 1300 1301 *mtup = mtu; 1302 if (alwaysfragp) 1303 *alwaysfragp = alwaysfrag; 1304 return (error); 1305 } 1306 1307 /* 1308 * IP6 socket option processing. 1309 */ 1310 int 1311 ip6_ctloutput(op, so, level, optname, mp) 1312 int op; 1313 struct socket *so; 1314 int level, optname; 1315 struct mbuf **mp; 1316 { 1317 struct in6pcb *in6p = sotoin6pcb(so); 1318 struct mbuf *m = *mp; 1319 int optval = 0; 1320 int error = 0; 1321 struct proc *p = curproc; /* XXX */ 1322 1323 if (level == IPPROTO_IPV6) { 1324 switch (op) { 1325 case PRCO_SETOPT: 1326 switch (optname) { 1327 case IPV6_PKTOPTIONS: 1328 /* m is freed in ip6_pcbopts */ 1329 return (ip6_pcbopts(&in6p->in6p_outputopts, 1330 m, so)); 1331 case IPV6_HOPOPTS: 1332 case IPV6_DSTOPTS: 1333 if (p == 0 || suser(p->p_ucred, &p->p_acflag)) { 1334 error = EPERM; 1335 break; 1336 } 1337 /* FALLTHROUGH */ 1338 case IPV6_UNICAST_HOPS: 1339 case IPV6_RECVOPTS: 1340 case IPV6_RECVRETOPTS: 1341 case IPV6_RECVDSTADDR: 1342 case IPV6_PKTINFO: 1343 case IPV6_HOPLIMIT: 1344 case IPV6_RTHDR: 1345 case IPV6_FAITH: 1346 case IPV6_V6ONLY: 1347 if (!m || m->m_len != sizeof(int)) { 1348 error = EINVAL; 1349 break; 1350 } 1351 optval = *mtod(m, int *); 1352 switch (optname) { 1353 1354 case IPV6_UNICAST_HOPS: 1355 if (optval < -1 || optval >= 256) 1356 error = EINVAL; 1357 else { 1358 /* -1 = kernel default */ 1359 in6p->in6p_hops = optval; 1360 } 1361 break; 1362 #define OPTSET(bit) \ 1363 do { \ 1364 if (optval) \ 1365 in6p->in6p_flags |= (bit); \ 1366 else \ 1367 in6p->in6p_flags &= ~(bit); \ 1368 } while (/*CONSTCOND*/ 0) 1369 1370 case IPV6_RECVOPTS: 1371 OPTSET(IN6P_RECVOPTS); 1372 break; 1373 1374 case IPV6_RECVRETOPTS: 1375 OPTSET(IN6P_RECVRETOPTS); 1376 break; 1377 1378 case IPV6_RECVDSTADDR: 1379 OPTSET(IN6P_RECVDSTADDR); 1380 break; 1381 1382 case IPV6_PKTINFO: 1383 OPTSET(IN6P_PKTINFO); 1384 break; 1385 1386 case IPV6_HOPLIMIT: 1387 OPTSET(IN6P_HOPLIMIT); 1388 break; 1389 1390 case IPV6_HOPOPTS: 1391 OPTSET(IN6P_HOPOPTS); 1392 break; 1393 1394 case IPV6_DSTOPTS: 1395 OPTSET(IN6P_DSTOPTS); 1396 break; 1397 1398 case IPV6_RTHDR: 1399 OPTSET(IN6P_RTHDR); 1400 break; 1401 1402 case IPV6_FAITH: 1403 OPTSET(IN6P_FAITH); 1404 break; 1405 1406 case IPV6_V6ONLY: 1407 /* 1408 * make setsockopt(IPV6_V6ONLY) 1409 * available only prior to bind(2). 1410 * see ipng mailing list, Jun 22 2001. 1411 */ 1412 if (in6p->in6p_lport || 1413 !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr)) { 1414 error = EINVAL; 1415 break; 1416 } 1417 #ifdef INET6_BINDV6ONLY 1418 if (!optval) 1419 error = EINVAL; 1420 #else 1421 OPTSET(IN6P_IPV6_V6ONLY); 1422 #endif 1423 break; 1424 } 1425 break; 1426 #undef OPTSET 1427 1428 case IPV6_MULTICAST_IF: 1429 case IPV6_MULTICAST_HOPS: 1430 case IPV6_MULTICAST_LOOP: 1431 case IPV6_JOIN_GROUP: 1432 case IPV6_LEAVE_GROUP: 1433 error = ip6_setmoptions(optname, 1434 &in6p->in6p_moptions, m); 1435 break; 1436 1437 case IPV6_PORTRANGE: 1438 optval = *mtod(m, int *); 1439 1440 switch (optval) { 1441 case IPV6_PORTRANGE_DEFAULT: 1442 in6p->in6p_flags &= ~(IN6P_LOWPORT); 1443 in6p->in6p_flags &= ~(IN6P_HIGHPORT); 1444 break; 1445 1446 case IPV6_PORTRANGE_HIGH: 1447 in6p->in6p_flags &= ~(IN6P_LOWPORT); 1448 in6p->in6p_flags |= IN6P_HIGHPORT; 1449 break; 1450 1451 case IPV6_PORTRANGE_LOW: 1452 in6p->in6p_flags &= ~(IN6P_HIGHPORT); 1453 in6p->in6p_flags |= IN6P_LOWPORT; 1454 break; 1455 1456 default: 1457 error = EINVAL; 1458 break; 1459 } 1460 break; 1461 1462 #ifdef IPSEC 1463 case IPV6_IPSEC_POLICY: 1464 { 1465 caddr_t req = NULL; 1466 size_t len = 0; 1467 1468 int priv = 0; 1469 if (p == 0 || suser(p->p_ucred, &p->p_acflag)) 1470 priv = 0; 1471 else 1472 priv = 1; 1473 if (m) { 1474 req = mtod(m, caddr_t); 1475 len = m->m_len; 1476 } 1477 error = ipsec6_set_policy(in6p, 1478 optname, req, len, priv); 1479 } 1480 break; 1481 #endif /* IPSEC */ 1482 1483 default: 1484 error = ENOPROTOOPT; 1485 break; 1486 } 1487 if (m) 1488 (void)m_free(m); 1489 break; 1490 1491 case PRCO_GETOPT: 1492 switch (optname) { 1493 1494 case IPV6_OPTIONS: 1495 case IPV6_RETOPTS: 1496 error = ENOPROTOOPT; 1497 break; 1498 1499 case IPV6_PKTOPTIONS: 1500 if (in6p->in6p_options) { 1501 *mp = m_copym(in6p->in6p_options, 0, 1502 M_COPYALL, M_WAIT); 1503 } else { 1504 *mp = m_get(M_WAIT, MT_SOOPTS); 1505 (*mp)->m_len = 0; 1506 } 1507 break; 1508 1509 case IPV6_HOPOPTS: 1510 case IPV6_DSTOPTS: 1511 if (p == 0 || suser(p->p_ucred, &p->p_acflag)) { 1512 error = EPERM; 1513 break; 1514 } 1515 /* FALLTHROUGH */ 1516 case IPV6_UNICAST_HOPS: 1517 case IPV6_RECVOPTS: 1518 case IPV6_RECVRETOPTS: 1519 case IPV6_RECVDSTADDR: 1520 case IPV6_PORTRANGE: 1521 case IPV6_PKTINFO: 1522 case IPV6_HOPLIMIT: 1523 case IPV6_RTHDR: 1524 case IPV6_FAITH: 1525 case IPV6_V6ONLY: 1526 *mp = m = m_get(M_WAIT, MT_SOOPTS); 1527 m->m_len = sizeof(int); 1528 switch (optname) { 1529 1530 case IPV6_UNICAST_HOPS: 1531 optval = in6p->in6p_hops; 1532 break; 1533 1534 #define OPTBIT(bit) (in6p->in6p_flags & bit ? 1 : 0) 1535 1536 case IPV6_RECVOPTS: 1537 optval = OPTBIT(IN6P_RECVOPTS); 1538 break; 1539 1540 case IPV6_RECVRETOPTS: 1541 optval = OPTBIT(IN6P_RECVRETOPTS); 1542 break; 1543 1544 case IPV6_RECVDSTADDR: 1545 optval = OPTBIT(IN6P_RECVDSTADDR); 1546 break; 1547 1548 case IPV6_PORTRANGE: 1549 { 1550 int flags; 1551 flags = in6p->in6p_flags; 1552 if (flags & IN6P_HIGHPORT) 1553 optval = IPV6_PORTRANGE_HIGH; 1554 else if (flags & IN6P_LOWPORT) 1555 optval = IPV6_PORTRANGE_LOW; 1556 else 1557 optval = 0; 1558 break; 1559 } 1560 1561 case IPV6_PKTINFO: 1562 optval = OPTBIT(IN6P_PKTINFO); 1563 break; 1564 1565 case IPV6_HOPLIMIT: 1566 optval = OPTBIT(IN6P_HOPLIMIT); 1567 break; 1568 1569 case IPV6_HOPOPTS: 1570 optval = OPTBIT(IN6P_HOPOPTS); 1571 break; 1572 1573 case IPV6_DSTOPTS: 1574 optval = OPTBIT(IN6P_DSTOPTS); 1575 break; 1576 1577 case IPV6_RTHDR: 1578 optval = OPTBIT(IN6P_RTHDR); 1579 break; 1580 1581 case IPV6_FAITH: 1582 optval = OPTBIT(IN6P_FAITH); 1583 break; 1584 1585 case IPV6_V6ONLY: 1586 optval = OPTBIT(IN6P_IPV6_V6ONLY); 1587 break; 1588 } 1589 *mtod(m, int *) = optval; 1590 break; 1591 1592 case IPV6_MULTICAST_IF: 1593 case IPV6_MULTICAST_HOPS: 1594 case IPV6_MULTICAST_LOOP: 1595 case IPV6_JOIN_GROUP: 1596 case IPV6_LEAVE_GROUP: 1597 error = ip6_getmoptions(optname, in6p->in6p_moptions, mp); 1598 break; 1599 1600 #ifdef IPSEC 1601 case IPV6_IPSEC_POLICY: 1602 { 1603 caddr_t req = NULL; 1604 size_t len = 0; 1605 1606 if (m) { 1607 req = mtod(m, caddr_t); 1608 len = m->m_len; 1609 } 1610 error = ipsec6_get_policy(in6p, req, len, mp); 1611 break; 1612 } 1613 #endif /* IPSEC */ 1614 1615 default: 1616 error = ENOPROTOOPT; 1617 break; 1618 } 1619 break; 1620 } 1621 } else { 1622 error = EINVAL; 1623 if (op == PRCO_SETOPT && *mp) 1624 (void)m_free(*mp); 1625 } 1626 return (error); 1627 } 1628 1629 int 1630 ip6_raw_ctloutput(op, so, level, optname, mp) 1631 int op; 1632 struct socket *so; 1633 int level, optname; 1634 struct mbuf **mp; 1635 { 1636 int error = 0, optval, optlen; 1637 const int icmp6off = offsetof(struct icmp6_hdr, icmp6_cksum); 1638 struct in6pcb *in6p = sotoin6pcb(so); 1639 struct mbuf *m = *mp; 1640 1641 optlen = m ? m->m_len : 0; 1642 1643 if (level != IPPROTO_IPV6) { 1644 if (op == PRCO_SETOPT && *mp) 1645 (void)m_free(*mp); 1646 return (EINVAL); 1647 } 1648 1649 switch (optname) { 1650 case IPV6_CHECKSUM: 1651 /* 1652 * For ICMPv6 sockets, no modification allowed for checksum 1653 * offset, permit "no change" values to help existing apps. 1654 * 1655 * XXX 2292bis says: "An attempt to set IPV6_CHECKSUM 1656 * for an ICMPv6 socket will fail." 1657 * The current behavior does not meet 2292bis. 1658 */ 1659 switch (op) { 1660 case PRCO_SETOPT: 1661 if (optlen != sizeof(int)) { 1662 error = EINVAL; 1663 break; 1664 } 1665 optval = *mtod(m, int *); 1666 if ((optval % 2) != 0) { 1667 /* the API assumes even offset values */ 1668 error = EINVAL; 1669 } else if (so->so_proto->pr_protocol == 1670 IPPROTO_ICMPV6) { 1671 if (optval != icmp6off) 1672 error = EINVAL; 1673 } else 1674 in6p->in6p_cksum = optval; 1675 break; 1676 1677 case PRCO_GETOPT: 1678 if (so->so_proto->pr_protocol == IPPROTO_ICMPV6) 1679 optval = icmp6off; 1680 else 1681 optval = in6p->in6p_cksum; 1682 1683 *mp = m = m_get(M_WAIT, MT_SOOPTS); 1684 m->m_len = sizeof(int); 1685 *mtod(m, int *) = optval; 1686 break; 1687 1688 default: 1689 error = EINVAL; 1690 break; 1691 } 1692 break; 1693 1694 default: 1695 error = ENOPROTOOPT; 1696 break; 1697 } 1698 1699 if (op == PRCO_SETOPT && m) 1700 (void)m_free(m); 1701 1702 return (error); 1703 } 1704 1705 /* 1706 * Set up IP6 options in pcb for insertion in output packets. 1707 * Store in mbuf with pointer in pcbopt, adding pseudo-option 1708 * with destination address if source routed. 1709 */ 1710 static int 1711 ip6_pcbopts(pktopt, m, so) 1712 struct ip6_pktopts **pktopt; 1713 struct mbuf *m; 1714 struct socket *so; 1715 { 1716 struct ip6_pktopts *opt = *pktopt; 1717 int error = 0; 1718 struct proc *p = curproc; /* XXX */ 1719 int priv = 0; 1720 1721 /* turn off any old options. */ 1722 if (opt) { 1723 if (opt->ip6po_m) 1724 (void)m_free(opt->ip6po_m); 1725 } else 1726 opt = malloc(sizeof(*opt), M_IP6OPT, M_WAITOK); 1727 *pktopt = 0; 1728 1729 if (!m || m->m_len == 0) { 1730 /* 1731 * Only turning off any previous options. 1732 */ 1733 free(opt, M_IP6OPT); 1734 if (m) 1735 (void)m_free(m); 1736 return (0); 1737 } 1738 1739 /* set options specified by user. */ 1740 if (p && !suser(p->p_ucred, &p->p_acflag)) 1741 priv = 1; 1742 if ((error = ip6_setpktoptions(m, opt, priv)) != 0) { 1743 (void)m_free(m); 1744 free(opt, M_IP6OPT); 1745 return (error); 1746 } 1747 *pktopt = opt; 1748 return (0); 1749 } 1750 1751 /* 1752 * Set the IP6 multicast options in response to user setsockopt(). 1753 */ 1754 static int 1755 ip6_setmoptions(optname, im6op, m) 1756 int optname; 1757 struct ip6_moptions **im6op; 1758 struct mbuf *m; 1759 { 1760 int error = 0; 1761 u_int loop, ifindex; 1762 struct ipv6_mreq *mreq; 1763 struct ifnet *ifp; 1764 struct ip6_moptions *im6o = *im6op; 1765 struct route_in6 ro; 1766 struct sockaddr_in6 *dst; 1767 struct in6_multi_mship *imm; 1768 struct proc *p = curproc; /* XXX */ 1769 1770 if (im6o == NULL) { 1771 /* 1772 * No multicast option buffer attached to the pcb; 1773 * allocate one and initialize to default values. 1774 */ 1775 im6o = (struct ip6_moptions *) 1776 malloc(sizeof(*im6o), M_IPMOPTS, M_WAITOK); 1777 1778 if (im6o == NULL) 1779 return (ENOBUFS); 1780 *im6op = im6o; 1781 im6o->im6o_multicast_ifp = NULL; 1782 im6o->im6o_multicast_hlim = ip6_defmcasthlim; 1783 im6o->im6o_multicast_loop = IPV6_DEFAULT_MULTICAST_LOOP; 1784 LIST_INIT(&im6o->im6o_memberships); 1785 } 1786 1787 switch (optname) { 1788 1789 case IPV6_MULTICAST_IF: 1790 /* 1791 * Select the interface for outgoing multicast packets. 1792 */ 1793 if (m == NULL || m->m_len != sizeof(u_int)) { 1794 error = EINVAL; 1795 break; 1796 } 1797 bcopy(mtod(m, u_int *), &ifindex, sizeof(ifindex)); 1798 if (ifindex < 0 || if_indexlim <= ifindex || 1799 !ifindex2ifnet[ifindex]) { 1800 error = ENXIO; /* XXX EINVAL? */ 1801 break; 1802 } 1803 ifp = ifindex2ifnet[ifindex]; 1804 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) { 1805 error = EADDRNOTAVAIL; 1806 break; 1807 } 1808 im6o->im6o_multicast_ifp = ifp; 1809 break; 1810 1811 case IPV6_MULTICAST_HOPS: 1812 { 1813 /* 1814 * Set the IP6 hoplimit for outgoing multicast packets. 1815 */ 1816 int optval; 1817 if (m == NULL || m->m_len != sizeof(int)) { 1818 error = EINVAL; 1819 break; 1820 } 1821 bcopy(mtod(m, u_int *), &optval, sizeof(optval)); 1822 if (optval < -1 || optval >= 256) 1823 error = EINVAL; 1824 else if (optval == -1) 1825 im6o->im6o_multicast_hlim = ip6_defmcasthlim; 1826 else 1827 im6o->im6o_multicast_hlim = optval; 1828 break; 1829 } 1830 1831 case IPV6_MULTICAST_LOOP: 1832 /* 1833 * Set the loopback flag for outgoing multicast packets. 1834 * Must be zero or one. 1835 */ 1836 if (m == NULL || m->m_len != sizeof(u_int)) { 1837 error = EINVAL; 1838 break; 1839 } 1840 bcopy(mtod(m, u_int *), &loop, sizeof(loop)); 1841 if (loop > 1) { 1842 error = EINVAL; 1843 break; 1844 } 1845 im6o->im6o_multicast_loop = loop; 1846 break; 1847 1848 case IPV6_JOIN_GROUP: 1849 /* 1850 * Add a multicast group membership. 1851 * Group must be a valid IP6 multicast address. 1852 */ 1853 if (m == NULL || m->m_len != sizeof(struct ipv6_mreq)) { 1854 error = EINVAL; 1855 break; 1856 } 1857 mreq = mtod(m, struct ipv6_mreq *); 1858 if (IN6_IS_ADDR_UNSPECIFIED(&mreq->ipv6mr_multiaddr)) { 1859 /* 1860 * We use the unspecified address to specify to accept 1861 * all multicast addresses. Only super user is allowed 1862 * to do this. 1863 */ 1864 if (suser(p->p_ucred, &p->p_acflag)) 1865 { 1866 error = EACCES; 1867 break; 1868 } 1869 } else if (!IN6_IS_ADDR_MULTICAST(&mreq->ipv6mr_multiaddr)) { 1870 error = EINVAL; 1871 break; 1872 } 1873 1874 /* 1875 * If the interface is specified, validate it. 1876 */ 1877 if (mreq->ipv6mr_interface < 0 || 1878 if_indexlim <= mreq->ipv6mr_interface || 1879 !ifindex2ifnet[mreq->ipv6mr_interface]) { 1880 error = ENXIO; /* XXX EINVAL? */ 1881 break; 1882 } 1883 /* 1884 * If no interface was explicitly specified, choose an 1885 * appropriate one according to the given multicast address. 1886 */ 1887 if (mreq->ipv6mr_interface == 0) { 1888 /* 1889 * If the multicast address is in node-local scope, 1890 * the interface should be a loopback interface. 1891 * Otherwise, look up the routing table for the 1892 * address, and choose the outgoing interface. 1893 * XXX: is it a good approach? 1894 */ 1895 if (IN6_IS_ADDR_MC_NODELOCAL(&mreq->ipv6mr_multiaddr)) { 1896 ifp = &loif[0]; 1897 } else { 1898 ro.ro_rt = NULL; 1899 dst = (struct sockaddr_in6 *)&ro.ro_dst; 1900 bzero(dst, sizeof(*dst)); 1901 dst->sin6_len = sizeof(struct sockaddr_in6); 1902 dst->sin6_family = AF_INET6; 1903 dst->sin6_addr = mreq->ipv6mr_multiaddr; 1904 rtalloc((struct route *)&ro); 1905 if (ro.ro_rt == NULL) { 1906 error = EADDRNOTAVAIL; 1907 break; 1908 } 1909 ifp = ro.ro_rt->rt_ifp; 1910 rtfree(ro.ro_rt); 1911 } 1912 } else 1913 ifp = ifindex2ifnet[mreq->ipv6mr_interface]; 1914 1915 /* 1916 * See if we found an interface, and confirm that it 1917 * supports multicast 1918 */ 1919 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) { 1920 error = EADDRNOTAVAIL; 1921 break; 1922 } 1923 /* 1924 * Put interface index into the multicast address, 1925 * if the address has link-local scope. 1926 */ 1927 if (IN6_IS_ADDR_MC_LINKLOCAL(&mreq->ipv6mr_multiaddr)) { 1928 mreq->ipv6mr_multiaddr.s6_addr16[1] = 1929 htons(ifp->if_index); 1930 } 1931 /* 1932 * See if the membership already exists. 1933 */ 1934 for (imm = im6o->im6o_memberships.lh_first; 1935 imm != NULL; imm = imm->i6mm_chain.le_next) 1936 if (imm->i6mm_maddr->in6m_ifp == ifp && 1937 IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr, 1938 &mreq->ipv6mr_multiaddr)) 1939 break; 1940 if (imm != NULL) { 1941 error = EADDRINUSE; 1942 break; 1943 } 1944 /* 1945 * Everything looks good; add a new record to the multicast 1946 * address list for the given interface. 1947 */ 1948 imm = in6_joingroup(ifp, &mreq->ipv6mr_multiaddr, &error); 1949 if (!imm) 1950 break; 1951 LIST_INSERT_HEAD(&im6o->im6o_memberships, imm, i6mm_chain); 1952 break; 1953 1954 case IPV6_LEAVE_GROUP: 1955 /* 1956 * Drop a multicast group membership. 1957 * Group must be a valid IP6 multicast address. 1958 */ 1959 if (m == NULL || m->m_len != sizeof(struct ipv6_mreq)) { 1960 error = EINVAL; 1961 break; 1962 } 1963 mreq = mtod(m, struct ipv6_mreq *); 1964 if (IN6_IS_ADDR_UNSPECIFIED(&mreq->ipv6mr_multiaddr)) { 1965 if (suser(p->p_ucred, &p->p_acflag)) 1966 { 1967 error = EACCES; 1968 break; 1969 } 1970 } else if (!IN6_IS_ADDR_MULTICAST(&mreq->ipv6mr_multiaddr)) { 1971 error = EINVAL; 1972 break; 1973 } 1974 /* 1975 * If an interface address was specified, get a pointer 1976 * to its ifnet structure. 1977 */ 1978 if (mreq->ipv6mr_interface < 0 || 1979 if_indexlim <= mreq->ipv6mr_interface || 1980 !ifindex2ifnet[mreq->ipv6mr_interface]) { 1981 error = ENXIO; /* XXX EINVAL? */ 1982 break; 1983 } 1984 ifp = ifindex2ifnet[mreq->ipv6mr_interface]; 1985 /* 1986 * Put interface index into the multicast address, 1987 * if the address has link-local scope. 1988 */ 1989 if (IN6_IS_ADDR_MC_LINKLOCAL(&mreq->ipv6mr_multiaddr)) { 1990 mreq->ipv6mr_multiaddr.s6_addr16[1] = 1991 htons(mreq->ipv6mr_interface); 1992 } 1993 /* 1994 * Find the membership in the membership list. 1995 */ 1996 for (imm = im6o->im6o_memberships.lh_first; 1997 imm != NULL; imm = imm->i6mm_chain.le_next) { 1998 if ((ifp == NULL || imm->i6mm_maddr->in6m_ifp == ifp) && 1999 IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr, 2000 &mreq->ipv6mr_multiaddr)) 2001 break; 2002 } 2003 if (imm == NULL) { 2004 /* Unable to resolve interface */ 2005 error = EADDRNOTAVAIL; 2006 break; 2007 } 2008 /* 2009 * Give up the multicast address record to which the 2010 * membership points. 2011 */ 2012 LIST_REMOVE(imm, i6mm_chain); 2013 in6_leavegroup(imm); 2014 break; 2015 2016 default: 2017 error = EOPNOTSUPP; 2018 break; 2019 } 2020 2021 /* 2022 * If all options have default values, no need to keep the mbuf. 2023 */ 2024 if (im6o->im6o_multicast_ifp == NULL && 2025 im6o->im6o_multicast_hlim == ip6_defmcasthlim && 2026 im6o->im6o_multicast_loop == IPV6_DEFAULT_MULTICAST_LOOP && 2027 im6o->im6o_memberships.lh_first == NULL) { 2028 free(*im6op, M_IPMOPTS); 2029 *im6op = NULL; 2030 } 2031 2032 return (error); 2033 } 2034 2035 /* 2036 * Return the IP6 multicast options in response to user getsockopt(). 2037 */ 2038 static int 2039 ip6_getmoptions(optname, im6o, mp) 2040 int optname; 2041 struct ip6_moptions *im6o; 2042 struct mbuf **mp; 2043 { 2044 u_int *hlim, *loop, *ifindex; 2045 2046 *mp = m_get(M_WAIT, MT_SOOPTS); 2047 2048 switch (optname) { 2049 2050 case IPV6_MULTICAST_IF: 2051 ifindex = mtod(*mp, u_int *); 2052 (*mp)->m_len = sizeof(u_int); 2053 if (im6o == NULL || im6o->im6o_multicast_ifp == NULL) 2054 *ifindex = 0; 2055 else 2056 *ifindex = im6o->im6o_multicast_ifp->if_index; 2057 return (0); 2058 2059 case IPV6_MULTICAST_HOPS: 2060 hlim = mtod(*mp, u_int *); 2061 (*mp)->m_len = sizeof(u_int); 2062 if (im6o == NULL) 2063 *hlim = ip6_defmcasthlim; 2064 else 2065 *hlim = im6o->im6o_multicast_hlim; 2066 return (0); 2067 2068 case IPV6_MULTICAST_LOOP: 2069 loop = mtod(*mp, u_int *); 2070 (*mp)->m_len = sizeof(u_int); 2071 if (im6o == NULL) 2072 *loop = ip6_defmcasthlim; 2073 else 2074 *loop = im6o->im6o_multicast_loop; 2075 return (0); 2076 2077 default: 2078 return (EOPNOTSUPP); 2079 } 2080 } 2081 2082 /* 2083 * Discard the IP6 multicast options. 2084 */ 2085 void 2086 ip6_freemoptions(im6o) 2087 struct ip6_moptions *im6o; 2088 { 2089 struct in6_multi_mship *imm; 2090 2091 if (im6o == NULL) 2092 return; 2093 2094 while ((imm = im6o->im6o_memberships.lh_first) != NULL) { 2095 LIST_REMOVE(imm, i6mm_chain); 2096 in6_leavegroup(imm); 2097 } 2098 free(im6o, M_IPMOPTS); 2099 } 2100 2101 /* 2102 * Set IPv6 outgoing packet options based on advanced API. 2103 */ 2104 int 2105 ip6_setpktoptions(control, opt, priv) 2106 struct mbuf *control; 2107 struct ip6_pktopts *opt; 2108 int priv; 2109 { 2110 struct cmsghdr *cm = 0; 2111 2112 if (control == 0 || opt == 0) 2113 return (EINVAL); 2114 2115 bzero(opt, sizeof(*opt)); 2116 opt->ip6po_hlim = -1; /* -1 means to use default hop limit */ 2117 2118 /* 2119 * XXX: Currently, we assume all the optional information is stored 2120 * in a single mbuf. 2121 */ 2122 if (control->m_next) 2123 return (EINVAL); 2124 2125 opt->ip6po_m = control; 2126 2127 for (; control->m_len; control->m_data += CMSG_ALIGN(cm->cmsg_len), 2128 control->m_len -= CMSG_ALIGN(cm->cmsg_len)) { 2129 cm = mtod(control, struct cmsghdr *); 2130 if (cm->cmsg_len == 0 || cm->cmsg_len > control->m_len) 2131 return (EINVAL); 2132 if (cm->cmsg_level != IPPROTO_IPV6) 2133 continue; 2134 2135 switch (cm->cmsg_type) { 2136 case IPV6_PKTINFO: 2137 if (cm->cmsg_len != CMSG_LEN(sizeof(struct in6_pktinfo))) 2138 return (EINVAL); 2139 opt->ip6po_pktinfo = (struct in6_pktinfo *)CMSG_DATA(cm); 2140 if (opt->ip6po_pktinfo->ipi6_ifindex && 2141 IN6_IS_ADDR_LINKLOCAL(&opt->ip6po_pktinfo->ipi6_addr)) 2142 opt->ip6po_pktinfo->ipi6_addr.s6_addr16[1] = 2143 htons(opt->ip6po_pktinfo->ipi6_ifindex); 2144 2145 if (opt->ip6po_pktinfo->ipi6_ifindex >= if_indexlim || 2146 opt->ip6po_pktinfo->ipi6_ifindex < 0) 2147 return (ENXIO); 2148 if (opt->ip6po_pktinfo->ipi6_ifindex > 0 && 2149 !ifindex2ifnet[opt->ip6po_pktinfo->ipi6_ifindex]) 2150 return (ENXIO); 2151 2152 /* 2153 * Check if the requested source address is indeed a 2154 * unicast address assigned to the node, and can be 2155 * used as the packet's source address. 2156 */ 2157 if (!IN6_IS_ADDR_UNSPECIFIED(&opt->ip6po_pktinfo->ipi6_addr)) { 2158 struct ifaddr *ia; 2159 struct in6_ifaddr *ia6; 2160 struct sockaddr_in6 sin6; 2161 2162 bzero(&sin6, sizeof(sin6)); 2163 sin6.sin6_len = sizeof(sin6); 2164 sin6.sin6_family = AF_INET6; 2165 sin6.sin6_addr = 2166 opt->ip6po_pktinfo->ipi6_addr; 2167 ia = ifa_ifwithaddr(sin6tosa(&sin6)); 2168 if (ia == NULL || 2169 (opt->ip6po_pktinfo->ipi6_ifindex && 2170 (ia->ifa_ifp->if_index != 2171 opt->ip6po_pktinfo->ipi6_ifindex))) { 2172 return (EADDRNOTAVAIL); 2173 } 2174 ia6 = (struct in6_ifaddr *)ia; 2175 if ((ia6->ia6_flags & (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY)) != 0) { 2176 return (EADDRNOTAVAIL); 2177 } 2178 2179 /* 2180 * Check if the requested source address is 2181 * indeed a unicast address assigned to the 2182 * node. 2183 */ 2184 if (IN6_IS_ADDR_MULTICAST(&opt->ip6po_pktinfo->ipi6_addr)) 2185 return (EADDRNOTAVAIL); 2186 } 2187 break; 2188 2189 case IPV6_HOPLIMIT: 2190 if (cm->cmsg_len != CMSG_LEN(sizeof(int))) 2191 return (EINVAL); 2192 else { 2193 int t; 2194 2195 bcopy(CMSG_DATA(cm), &t, sizeof(t)); 2196 if (t < -1 || t > 255) 2197 return (EINVAL); 2198 opt->ip6po_hlim = t; 2199 } 2200 break; 2201 2202 case IPV6_NEXTHOP: 2203 if (!priv) 2204 return (EPERM); 2205 2206 /* check if cmsg_len is large enough for sa_len */ 2207 if (cm->cmsg_len < sizeof(u_char) || 2208 cm->cmsg_len < CMSG_LEN(*CMSG_DATA(cm))) 2209 return (EINVAL); 2210 2211 opt->ip6po_nexthop = (struct sockaddr *)CMSG_DATA(cm); 2212 2213 break; 2214 2215 case IPV6_HOPOPTS: 2216 if (cm->cmsg_len < CMSG_LEN(sizeof(struct ip6_hbh))) 2217 return (EINVAL); 2218 else { 2219 struct ip6_hbh *t; 2220 2221 t = (struct ip6_hbh *)CMSG_DATA(cm); 2222 if (cm->cmsg_len != 2223 CMSG_LEN((t->ip6h_len + 1) << 3)) 2224 return (EINVAL); 2225 opt->ip6po_hbh = t; 2226 } 2227 break; 2228 2229 case IPV6_DSTOPTS: 2230 if (cm->cmsg_len < CMSG_LEN(sizeof(struct ip6_dest))) 2231 return (EINVAL); 2232 2233 /* 2234 * If there is no routing header yet, the destination 2235 * options header should be put on the 1st part. 2236 * Otherwise, the header should be on the 2nd part. 2237 * (See RFC 2460, section 4.1) 2238 */ 2239 if (opt->ip6po_rthdr == NULL) { 2240 struct ip6_dest *t; 2241 2242 t = (struct ip6_dest *)CMSG_DATA(cm); 2243 if (cm->cmsg_len != 2244 CMSG_LEN((t->ip6d_len + 1) << 3)); 2245 return (EINVAL); 2246 opt->ip6po_dest1 = t; 2247 } 2248 else { 2249 struct ip6_dest *t; 2250 2251 t = (struct ip6_dest *)CMSG_DATA(cm); 2252 if (cm->cmsg_len != 2253 CMSG_LEN((opt->ip6po_dest2->ip6d_len + 1) << 3)) 2254 return (EINVAL); 2255 opt->ip6po_dest2 = t; 2256 } 2257 break; 2258 2259 case IPV6_RTHDR: 2260 if (cm->cmsg_len < CMSG_LEN(sizeof(struct ip6_rthdr))) 2261 return (EINVAL); 2262 else { 2263 struct ip6_rthdr *t; 2264 2265 t = (struct ip6_rthdr *)CMSG_DATA(cm); 2266 if (cm->cmsg_len != 2267 CMSG_LEN((t->ip6r_len + 1) << 3)) 2268 return (EINVAL); 2269 switch (t->ip6r_type) { 2270 case IPV6_RTHDR_TYPE_0: 2271 if (t->ip6r_segleft == 0) 2272 return (EINVAL); 2273 break; 2274 default: 2275 return (EINVAL); 2276 } 2277 opt->ip6po_rthdr = t; 2278 } 2279 break; 2280 2281 default: 2282 return (ENOPROTOOPT); 2283 } 2284 } 2285 2286 return (0); 2287 } 2288 2289 /* 2290 * Routine called from ip6_output() to loop back a copy of an IP6 multicast 2291 * packet to the input queue of a specified interface. Note that this 2292 * calls the output routine of the loopback "driver", but with an interface 2293 * pointer that might NOT be &loif -- easier than replicating that code here. 2294 */ 2295 void 2296 ip6_mloopback(ifp, m, dst) 2297 struct ifnet *ifp; 2298 struct mbuf *m; 2299 struct sockaddr_in6 *dst; 2300 { 2301 struct mbuf *copym; 2302 struct ip6_hdr *ip6; 2303 2304 copym = m_copy(m, 0, M_COPYALL); 2305 if (copym == NULL) 2306 return; 2307 2308 /* 2309 * Make sure to deep-copy IPv6 header portion in case the data 2310 * is in an mbuf cluster, so that we can safely override the IPv6 2311 * header portion later. 2312 */ 2313 if ((copym->m_flags & M_EXT) != 0 || 2314 copym->m_len < sizeof(struct ip6_hdr)) { 2315 copym = m_pullup(copym, sizeof(struct ip6_hdr)); 2316 if (copym == NULL) 2317 return; 2318 } 2319 2320 #ifdef DIAGNOSTIC 2321 if (copym->m_len < sizeof(*ip6)) { 2322 m_freem(copym); 2323 return; 2324 } 2325 #endif 2326 2327 ip6 = mtod(copym, struct ip6_hdr *); 2328 if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src)) 2329 ip6->ip6_src.s6_addr16[1] = 0; 2330 if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst)) 2331 ip6->ip6_dst.s6_addr16[1] = 0; 2332 2333 (void)looutput(ifp, copym, (struct sockaddr *)dst, NULL); 2334 } 2335 2336 /* 2337 * Chop IPv6 header off from the payload. 2338 */ 2339 static int 2340 ip6_splithdr(m, exthdrs) 2341 struct mbuf *m; 2342 struct ip6_exthdrs *exthdrs; 2343 { 2344 struct mbuf *mh; 2345 struct ip6_hdr *ip6; 2346 2347 ip6 = mtod(m, struct ip6_hdr *); 2348 if (m->m_len > sizeof(*ip6)) { 2349 MGETHDR(mh, M_DONTWAIT, MT_HEADER); 2350 if (mh == 0) { 2351 m_freem(m); 2352 return ENOBUFS; 2353 } 2354 M_COPY_PKTHDR(mh, m); 2355 MH_ALIGN(mh, sizeof(*ip6)); 2356 m_tag_delete_chain(m, NULL); 2357 m->m_flags &= ~M_PKTHDR; 2358 m->m_len -= sizeof(*ip6); 2359 m->m_data += sizeof(*ip6); 2360 mh->m_next = m; 2361 m = mh; 2362 m->m_len = sizeof(*ip6); 2363 bcopy((caddr_t)ip6, mtod(m, caddr_t), sizeof(*ip6)); 2364 } 2365 exthdrs->ip6e_ip6 = m; 2366 return 0; 2367 } 2368 2369 /* 2370 * Compute IPv6 extension header length. 2371 */ 2372 int 2373 ip6_optlen(in6p) 2374 struct in6pcb *in6p; 2375 { 2376 int len; 2377 2378 if (!in6p->in6p_outputopts) 2379 return 0; 2380 2381 len = 0; 2382 #define elen(x) \ 2383 (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0) 2384 2385 len += elen(in6p->in6p_outputopts->ip6po_hbh); 2386 len += elen(in6p->in6p_outputopts->ip6po_dest1); 2387 len += elen(in6p->in6p_outputopts->ip6po_rthdr); 2388 len += elen(in6p->in6p_outputopts->ip6po_dest2); 2389 return len; 2390 #undef elen 2391 } 2392