1 /* $NetBSD: ip6_output.c,v 1.83 2004/06/11 04:10:10 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 1.83 2004/06/11 04:10:10 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 *)); 126 127 extern struct ifnet loif[NLOOP]; 128 129 /* 130 * IP6 output. The packet in mbuf chain m contains a skeletal IP6 131 * header (with pri, len, nxt, hlim, src, dst). 132 * This function may modify ver and hlim only. 133 * The mbuf chain containing the packet will be freed. 134 * The mbuf opt, if present, will not be freed. 135 * 136 * type of "mtu": rt_rmx.rmx_mtu is u_long, ifnet.ifr_mtu is int, and 137 * nd_ifinfo.linkmtu is u_int32_t. so we use u_long to hold largest one, 138 * which is rt_rmx.rmx_mtu. 139 */ 140 int 141 ip6_output(m0, opt, ro, flags, im6o, so, ifpp) 142 struct mbuf *m0; 143 struct ip6_pktopts *opt; 144 struct route_in6 *ro; 145 int flags; 146 struct ip6_moptions *im6o; 147 struct socket *so; 148 struct ifnet **ifpp; /* XXX: just for statistics */ 149 { 150 struct ip6_hdr *ip6, *mhip6; 151 struct ifnet *ifp, *origifp; 152 struct mbuf *m = m0; 153 int hlen, tlen, len, off; 154 struct route_in6 ip6route; 155 struct sockaddr_in6 *dst; 156 int error = 0; 157 u_long mtu; 158 int alwaysfrag, dontfrag; 159 u_int32_t optlen = 0, plen = 0, unfragpartlen = 0; 160 struct ip6_exthdrs exthdrs; 161 struct in6_addr finaldst; 162 struct route_in6 *ro_pmtu = NULL; 163 int hdrsplit = 0; 164 int needipsec = 0; 165 #ifdef IPSEC 166 int needipsectun = 0; 167 struct secpolicy *sp = NULL; 168 169 ip6 = mtod(m, struct ip6_hdr *); 170 #endif /* IPSEC */ 171 172 #define MAKE_EXTHDR(hp, mp) \ 173 do { \ 174 if (hp) { \ 175 struct ip6_ext *eh = (struct ip6_ext *)(hp); \ 176 error = ip6_copyexthdr((mp), (caddr_t)(hp), \ 177 ((eh)->ip6e_len + 1) << 3); \ 178 if (error) \ 179 goto freehdrs; \ 180 } \ 181 } while (/*CONSTCOND*/ 0) 182 183 bzero(&exthdrs, sizeof(exthdrs)); 184 if (opt) { 185 /* Hop-by-Hop options header */ 186 MAKE_EXTHDR(opt->ip6po_hbh, &exthdrs.ip6e_hbh); 187 /* Destination options header(1st part) */ 188 MAKE_EXTHDR(opt->ip6po_dest1, &exthdrs.ip6e_dest1); 189 /* Routing header */ 190 MAKE_EXTHDR(opt->ip6po_rthdr, &exthdrs.ip6e_rthdr); 191 /* Destination options header(2nd part) */ 192 MAKE_EXTHDR(opt->ip6po_dest2, &exthdrs.ip6e_dest2); 193 } 194 195 #ifdef IPSEC 196 if ((flags & IPV6_FORWARDING) != 0) { 197 needipsec = 0; 198 goto skippolicycheck; 199 } 200 201 /* get a security policy for this packet */ 202 if (so == NULL) 203 sp = ipsec6_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND, 0, &error); 204 else { 205 if (IPSEC_PCB_SKIP_IPSEC(sotoinpcb_hdr(so)->inph_sp, 206 IPSEC_DIR_OUTBOUND)) { 207 needipsec = 0; 208 goto skippolicycheck; 209 } 210 sp = ipsec6_getpolicybysock(m, IPSEC_DIR_OUTBOUND, so, &error); 211 } 212 213 if (sp == NULL) { 214 ipsec6stat.out_inval++; 215 goto freehdrs; 216 } 217 218 error = 0; 219 220 /* check policy */ 221 switch (sp->policy) { 222 case IPSEC_POLICY_DISCARD: 223 /* 224 * This packet is just discarded. 225 */ 226 ipsec6stat.out_polvio++; 227 goto freehdrs; 228 229 case IPSEC_POLICY_BYPASS: 230 case IPSEC_POLICY_NONE: 231 /* no need to do IPsec. */ 232 needipsec = 0; 233 break; 234 235 case IPSEC_POLICY_IPSEC: 236 if (sp->req == NULL) { 237 /* XXX should be panic ? */ 238 printf("ip6_output: No IPsec request specified.\n"); 239 error = EINVAL; 240 goto freehdrs; 241 } 242 needipsec = 1; 243 break; 244 245 case IPSEC_POLICY_ENTRUST: 246 default: 247 printf("ip6_output: Invalid policy found. %d\n", sp->policy); 248 } 249 250 skippolicycheck:; 251 #endif /* IPSEC */ 252 253 /* 254 * Calculate the total length of the extension header chain. 255 * Keep the length of the unfragmentable part for fragmentation. 256 */ 257 optlen = 0; 258 if (exthdrs.ip6e_hbh) optlen += exthdrs.ip6e_hbh->m_len; 259 if (exthdrs.ip6e_dest1) optlen += exthdrs.ip6e_dest1->m_len; 260 if (exthdrs.ip6e_rthdr) optlen += exthdrs.ip6e_rthdr->m_len; 261 unfragpartlen = optlen + sizeof(struct ip6_hdr); 262 /* NOTE: we don't add AH/ESP length here. do that later. */ 263 if (exthdrs.ip6e_dest2) optlen += exthdrs.ip6e_dest2->m_len; 264 265 /* 266 * If we need IPsec, or there is at least one extension header, 267 * separate IP6 header from the payload. 268 */ 269 if ((needipsec || optlen) && !hdrsplit) { 270 if ((error = ip6_splithdr(m, &exthdrs)) != 0) { 271 m = NULL; 272 goto freehdrs; 273 } 274 m = exthdrs.ip6e_ip6; 275 hdrsplit++; 276 } 277 278 /* adjust pointer */ 279 ip6 = mtod(m, struct ip6_hdr *); 280 281 /* adjust mbuf packet header length */ 282 m->m_pkthdr.len += optlen; 283 plen = m->m_pkthdr.len - sizeof(*ip6); 284 285 /* If this is a jumbo payload, insert a jumbo payload option. */ 286 if (plen > IPV6_MAXPACKET) { 287 if (!hdrsplit) { 288 if ((error = ip6_splithdr(m, &exthdrs)) != 0) { 289 m = NULL; 290 goto freehdrs; 291 } 292 m = exthdrs.ip6e_ip6; 293 hdrsplit++; 294 } 295 /* adjust pointer */ 296 ip6 = mtod(m, struct ip6_hdr *); 297 if ((error = ip6_insert_jumboopt(&exthdrs, plen)) != 0) 298 goto freehdrs; 299 ip6->ip6_plen = 0; 300 } else 301 ip6->ip6_plen = htons(plen); 302 303 /* 304 * Concatenate headers and fill in next header fields. 305 * Here we have, on "m" 306 * IPv6 payload 307 * and we insert headers accordingly. Finally, we should be getting: 308 * IPv6 hbh dest1 rthdr ah* [esp* dest2 payload] 309 * 310 * during the header composing process, "m" points to IPv6 header. 311 * "mprev" points to an extension header prior to esp. 312 */ 313 { 314 u_char *nexthdrp = &ip6->ip6_nxt; 315 struct mbuf *mprev = m; 316 317 /* 318 * we treat dest2 specially. this makes IPsec processing 319 * much easier. the goal here is to make mprev point the 320 * mbuf prior to dest2. 321 * 322 * result: IPv6 dest2 payload 323 * m and mprev will point to IPv6 header. 324 */ 325 if (exthdrs.ip6e_dest2) { 326 if (!hdrsplit) 327 panic("assumption failed: hdr not split"); 328 exthdrs.ip6e_dest2->m_next = m->m_next; 329 m->m_next = exthdrs.ip6e_dest2; 330 *mtod(exthdrs.ip6e_dest2, u_char *) = ip6->ip6_nxt; 331 ip6->ip6_nxt = IPPROTO_DSTOPTS; 332 } 333 334 #define MAKE_CHAIN(m, mp, p, i)\ 335 do {\ 336 if (m) {\ 337 if (!hdrsplit) \ 338 panic("assumption failed: hdr not split"); \ 339 *mtod((m), u_char *) = *(p);\ 340 *(p) = (i);\ 341 p = mtod((m), u_char *);\ 342 (m)->m_next = (mp)->m_next;\ 343 (mp)->m_next = (m);\ 344 (mp) = (m);\ 345 }\ 346 } while (/*CONSTCOND*/ 0) 347 /* 348 * result: IPv6 hbh dest1 rthdr dest2 payload 349 * m will point to IPv6 header. mprev will point to the 350 * extension header prior to dest2 (rthdr in the above case). 351 */ 352 MAKE_CHAIN(exthdrs.ip6e_hbh, mprev, nexthdrp, IPPROTO_HOPOPTS); 353 MAKE_CHAIN(exthdrs.ip6e_dest1, mprev, nexthdrp, 354 IPPROTO_DSTOPTS); 355 MAKE_CHAIN(exthdrs.ip6e_rthdr, mprev, nexthdrp, 356 IPPROTO_ROUTING); 357 358 #ifdef IPSEC 359 if (!needipsec) 360 goto skip_ipsec2; 361 362 /* 363 * pointers after IPsec headers are not valid any more. 364 * other pointers need a great care too. 365 * (IPsec routines should not mangle mbufs prior to AH/ESP) 366 */ 367 exthdrs.ip6e_dest2 = NULL; 368 369 { 370 struct ip6_rthdr *rh = NULL; 371 int segleft_org = 0; 372 struct ipsec_output_state state; 373 374 if (exthdrs.ip6e_rthdr) { 375 rh = mtod(exthdrs.ip6e_rthdr, struct ip6_rthdr *); 376 segleft_org = rh->ip6r_segleft; 377 rh->ip6r_segleft = 0; 378 } 379 380 bzero(&state, sizeof(state)); 381 state.m = m; 382 error = ipsec6_output_trans(&state, nexthdrp, mprev, sp, flags, 383 &needipsectun); 384 m = state.m; 385 if (error) { 386 /* mbuf is already reclaimed in ipsec6_output_trans. */ 387 m = NULL; 388 switch (error) { 389 case EHOSTUNREACH: 390 case ENETUNREACH: 391 case EMSGSIZE: 392 case ENOBUFS: 393 case ENOMEM: 394 break; 395 default: 396 printf("ip6_output (ipsec): error code %d\n", error); 397 /* FALLTHROUGH */ 398 case ENOENT: 399 /* don't show these error codes to the user */ 400 error = 0; 401 break; 402 } 403 goto bad; 404 } 405 if (exthdrs.ip6e_rthdr) { 406 /* ah6_output doesn't modify mbuf chain */ 407 rh->ip6r_segleft = segleft_org; 408 } 409 } 410 skip_ipsec2:; 411 #endif 412 } 413 414 /* 415 * If there is a routing header, replace destination address field 416 * with the first hop of the routing header. 417 */ 418 if (exthdrs.ip6e_rthdr) { 419 struct ip6_rthdr *rh; 420 struct ip6_rthdr0 *rh0; 421 struct in6_addr *addr; 422 423 rh = (struct ip6_rthdr *)(mtod(exthdrs.ip6e_rthdr, 424 struct ip6_rthdr *)); 425 finaldst = ip6->ip6_dst; 426 switch (rh->ip6r_type) { 427 case IPV6_RTHDR_TYPE_0: 428 rh0 = (struct ip6_rthdr0 *)rh; 429 addr = (struct in6_addr *)(rh0 + 1); 430 ip6->ip6_dst = addr[0]; 431 bcopy(&addr[1], &addr[0], 432 sizeof(struct in6_addr) * (rh0->ip6r0_segleft - 1)); 433 addr[rh0->ip6r0_segleft - 1] = finaldst; 434 break; 435 default: /* is it possible? */ 436 error = EINVAL; 437 goto bad; 438 } 439 } 440 441 /* Source address validation */ 442 if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) && 443 (flags & IPV6_UNSPECSRC) == 0) { 444 error = EOPNOTSUPP; 445 ip6stat.ip6s_badscope++; 446 goto bad; 447 } 448 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) { 449 error = EOPNOTSUPP; 450 ip6stat.ip6s_badscope++; 451 goto bad; 452 } 453 454 ip6stat.ip6s_localout++; 455 456 /* 457 * Route packet. 458 */ 459 /* initialize cached route */ 460 if (ro == 0) { 461 ro = &ip6route; 462 bzero((caddr_t)ro, sizeof(*ro)); 463 } 464 ro_pmtu = ro; 465 if (opt && opt->ip6po_rthdr) 466 ro = &opt->ip6po_route; 467 dst = (struct sockaddr_in6 *)&ro->ro_dst; 468 /* 469 * If there is a cached route, 470 * check that it is to the same destination 471 * and is still up. If not, free it and try again. 472 */ 473 if (ro->ro_rt && ((ro->ro_rt->rt_flags & RTF_UP) == 0 || 474 dst->sin6_family != AF_INET6 || 475 !IN6_ARE_ADDR_EQUAL(&dst->sin6_addr, &ip6->ip6_dst))) { 476 RTFREE(ro->ro_rt); 477 ro->ro_rt = (struct rtentry *)0; 478 } 479 if (ro->ro_rt == 0) { 480 bzero(dst, sizeof(*dst)); 481 dst->sin6_family = AF_INET6; 482 dst->sin6_len = sizeof(struct sockaddr_in6); 483 dst->sin6_addr = ip6->ip6_dst; 484 } 485 #ifdef IPSEC 486 if (needipsec && needipsectun) { 487 struct ipsec_output_state state; 488 489 /* 490 * All the extension headers will become inaccessible 491 * (since they can be encrypted). 492 * Don't panic, we need no more updates to extension headers 493 * on inner IPv6 packet (since they are now encapsulated). 494 * 495 * IPv6 [ESP|AH] IPv6 [extension headers] payload 496 */ 497 bzero(&exthdrs, sizeof(exthdrs)); 498 exthdrs.ip6e_ip6 = m; 499 500 bzero(&state, sizeof(state)); 501 state.m = m; 502 state.ro = (struct route *)ro; 503 state.dst = (struct sockaddr *)dst; 504 505 error = ipsec6_output_tunnel(&state, sp, flags); 506 507 m = state.m; 508 ro = (struct route_in6 *)state.ro; 509 dst = (struct sockaddr_in6 *)state.dst; 510 if (error) { 511 /* mbuf is already reclaimed in ipsec6_output_tunnel. */ 512 m0 = m = NULL; 513 m = NULL; 514 switch (error) { 515 case EHOSTUNREACH: 516 case ENETUNREACH: 517 case EMSGSIZE: 518 case ENOBUFS: 519 case ENOMEM: 520 break; 521 default: 522 printf("ip6_output (ipsec): error code %d\n", error); 523 /* FALLTHROUGH */ 524 case ENOENT: 525 /* don't show these error codes to the user */ 526 error = 0; 527 break; 528 } 529 goto bad; 530 } 531 532 exthdrs.ip6e_ip6 = m; 533 } 534 #endif /* IPSEC */ 535 536 if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) { 537 /* Unicast */ 538 539 #define ifatoia6(ifa) ((struct in6_ifaddr *)(ifa)) 540 #define sin6tosa(sin6) ((struct sockaddr *)(sin6)) 541 /* xxx 542 * interface selection comes here 543 * if an interface is specified from an upper layer, 544 * ifp must point it. 545 */ 546 if (ro->ro_rt == 0) { 547 /* 548 * non-bsdi always clone routes, if parent is 549 * PRF_CLONING. 550 */ 551 rtalloc((struct route *)ro); 552 } 553 if (ro->ro_rt == 0) { 554 ip6stat.ip6s_noroute++; 555 error = EHOSTUNREACH; 556 /* XXX in6_ifstat_inc(ifp, ifs6_out_discard); */ 557 goto bad; 558 } 559 ifp = ro->ro_rt->rt_ifp; 560 ro->ro_rt->rt_use++; 561 if (ro->ro_rt->rt_flags & RTF_GATEWAY) 562 dst = (struct sockaddr_in6 *)ro->ro_rt->rt_gateway; 563 m->m_flags &= ~(M_BCAST | M_MCAST); /* just in case */ 564 565 in6_ifstat_inc(ifp, ifs6_out_request); 566 567 /* 568 * Check if the outgoing interface conflicts with 569 * the interface specified by ifi6_ifindex (if specified). 570 * Note that loopback interface is always okay. 571 * (this may happen when we are sending a packet to one of 572 * our own addresses.) 573 */ 574 if (opt && opt->ip6po_pktinfo && 575 opt->ip6po_pktinfo->ipi6_ifindex) { 576 if (!(ifp->if_flags & IFF_LOOPBACK) && 577 ifp->if_index != opt->ip6po_pktinfo->ipi6_ifindex) { 578 ip6stat.ip6s_noroute++; 579 in6_ifstat_inc(ifp, ifs6_out_discard); 580 error = EHOSTUNREACH; 581 goto bad; 582 } 583 } 584 585 if (opt && opt->ip6po_hlim != -1) 586 ip6->ip6_hlim = opt->ip6po_hlim & 0xff; 587 } else { 588 /* Multicast */ 589 struct in6_multi *in6m; 590 591 m->m_flags = (m->m_flags & ~M_BCAST) | M_MCAST; 592 593 /* 594 * See if the caller provided any multicast options 595 */ 596 ifp = NULL; 597 if (im6o != NULL) { 598 ip6->ip6_hlim = im6o->im6o_multicast_hlim; 599 if (im6o->im6o_multicast_ifp != NULL) 600 ifp = im6o->im6o_multicast_ifp; 601 } else 602 ip6->ip6_hlim = ip6_defmcasthlim; 603 604 /* 605 * See if the caller provided the outgoing interface 606 * as an ancillary data. 607 * Boundary check for ifindex is assumed to be already done. 608 */ 609 if (opt && opt->ip6po_pktinfo && opt->ip6po_pktinfo->ipi6_ifindex) 610 ifp = ifindex2ifnet[opt->ip6po_pktinfo->ipi6_ifindex]; 611 612 /* 613 * If the destination is a node-local scope multicast, 614 * the packet should be loop-backed only. 615 */ 616 if (IN6_IS_ADDR_MC_NODELOCAL(&ip6->ip6_dst)) { 617 /* 618 * If the outgoing interface is already specified, 619 * it should be a loopback interface. 620 */ 621 if (ifp && (ifp->if_flags & IFF_LOOPBACK) == 0) { 622 ip6stat.ip6s_badscope++; 623 error = ENETUNREACH; /* XXX: better error? */ 624 /* XXX correct ifp? */ 625 in6_ifstat_inc(ifp, ifs6_out_discard); 626 goto bad; 627 } else { 628 ifp = &loif[0]; 629 } 630 } 631 632 if (opt && opt->ip6po_hlim != -1) 633 ip6->ip6_hlim = opt->ip6po_hlim & 0xff; 634 635 /* 636 * If caller did not provide an interface lookup a 637 * default in the routing table. This is either a 638 * default for the speicfied group (i.e. a host 639 * route), or a multicast default (a route for the 640 * ``net'' ff00::/8). 641 */ 642 if (ifp == NULL) { 643 if (ro->ro_rt == 0) { 644 ro->ro_rt = rtalloc1((struct sockaddr *) 645 &ro->ro_dst, 0); 646 } 647 if (ro->ro_rt == 0) { 648 ip6stat.ip6s_noroute++; 649 error = EHOSTUNREACH; 650 /* XXX in6_ifstat_inc(ifp, ifs6_out_discard) */ 651 goto bad; 652 } 653 ifp = ro->ro_rt->rt_ifp; 654 ro->ro_rt->rt_use++; 655 } 656 657 if ((flags & IPV6_FORWARDING) == 0) 658 in6_ifstat_inc(ifp, ifs6_out_request); 659 in6_ifstat_inc(ifp, ifs6_out_mcast); 660 661 /* 662 * Confirm that the outgoing interface supports multicast. 663 */ 664 if ((ifp->if_flags & IFF_MULTICAST) == 0) { 665 ip6stat.ip6s_noroute++; 666 in6_ifstat_inc(ifp, ifs6_out_discard); 667 error = ENETUNREACH; 668 goto bad; 669 } 670 IN6_LOOKUP_MULTI(ip6->ip6_dst, ifp, in6m); 671 if (in6m != NULL && 672 (im6o == NULL || im6o->im6o_multicast_loop)) { 673 /* 674 * If we belong to the destination multicast group 675 * on the outgoing interface, and the caller did not 676 * forbid loopback, loop back a copy. 677 */ 678 ip6_mloopback(ifp, m, dst); 679 } else { 680 /* 681 * If we are acting as a multicast router, perform 682 * multicast forwarding as if the packet had just 683 * arrived on the interface to which we are about 684 * to send. The multicast forwarding function 685 * recursively calls this function, using the 686 * IPV6_FORWARDING flag to prevent infinite recursion. 687 * 688 * Multicasts that are looped back by ip6_mloopback(), 689 * above, will be forwarded by the ip6_input() routine, 690 * if necessary. 691 */ 692 if (ip6_mrouter && (flags & IPV6_FORWARDING) == 0) { 693 if (ip6_mforward(ip6, ifp, m) != 0) { 694 m_freem(m); 695 goto done; 696 } 697 } 698 } 699 /* 700 * Multicasts with a hoplimit of zero may be looped back, 701 * above, but must not be transmitted on a network. 702 * Also, multicasts addressed to the loopback interface 703 * are not sent -- the above call to ip6_mloopback() will 704 * loop back a copy if this host actually belongs to the 705 * destination group on the loopback interface. 706 */ 707 if (ip6->ip6_hlim == 0 || (ifp->if_flags & IFF_LOOPBACK)) { 708 m_freem(m); 709 goto done; 710 } 711 } 712 713 /* 714 * Fill the outgoing inteface to tell the upper layer 715 * to increment per-interface statistics. 716 */ 717 if (ifpp) 718 *ifpp = ifp; 719 720 /* Determine path MTU. */ 721 if ((error = ip6_getpmtu(ro_pmtu, ro, ifp, &finaldst, &mtu, 722 &alwaysfrag)) != 0) 723 goto bad; 724 725 /* 726 * The caller of this function may specify to use the minimum MTU 727 * in some cases. 728 */ 729 if (mtu > IPV6_MMTU) { 730 if ((flags & IPV6_MINMTU)) 731 mtu = IPV6_MMTU; 732 } 733 734 /* Fake scoped addresses */ 735 if ((ifp->if_flags & IFF_LOOPBACK) != 0) { 736 /* 737 * If source or destination address is a scoped address, and 738 * the packet is going to be sent to a loopback interface, 739 * we should keep the original interface. 740 */ 741 742 /* 743 * XXX: this is a very experimental and temporary solution. 744 * We eventually have sockaddr_in6 and use the sin6_scope_id 745 * field of the structure here. 746 * We rely on the consistency between two scope zone ids 747 * of source add destination, which should already be assured 748 * Larger scopes than link will be supported in the near 749 * future. 750 */ 751 origifp = NULL; 752 if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src)) 753 origifp = ifindex2ifnet[ntohs(ip6->ip6_src.s6_addr16[1])]; 754 else if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst)) 755 origifp = ifindex2ifnet[ntohs(ip6->ip6_dst.s6_addr16[1])]; 756 /* 757 * XXX: origifp can be NULL even in those two cases above. 758 * For example, if we remove the (only) link-local address 759 * from the loopback interface, and try to send a link-local 760 * address without link-id information. Then the source 761 * address is ::1, and the destination address is the 762 * link-local address with its s6_addr16[1] being zero. 763 * What is worse, if the packet goes to the loopback interface 764 * by a default rejected route, the null pointer would be 765 * passed to looutput, and the kernel would hang. 766 * The following last resort would prevent such disaster. 767 */ 768 if (origifp == NULL) 769 origifp = ifp; 770 } else 771 origifp = ifp; 772 if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src)) 773 ip6->ip6_src.s6_addr16[1] = 0; 774 if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst)) 775 ip6->ip6_dst.s6_addr16[1] = 0; 776 777 /* 778 * If the outgoing packet contains a hop-by-hop options header, 779 * it must be examined and processed even by the source node. 780 * (RFC 2460, section 4.) 781 */ 782 if (exthdrs.ip6e_hbh) { 783 struct ip6_hbh *hbh = mtod(exthdrs.ip6e_hbh, struct ip6_hbh *); 784 u_int32_t dummy1; /* XXX unused */ 785 u_int32_t dummy2; /* XXX unused */ 786 787 /* 788 * XXX: if we have to send an ICMPv6 error to the sender, 789 * we need the M_LOOP flag since icmp6_error() expects 790 * the IPv6 and the hop-by-hop options header are 791 * continuous unless the flag is set. 792 */ 793 m->m_flags |= M_LOOP; 794 m->m_pkthdr.rcvif = ifp; 795 if (ip6_process_hopopts(m, (u_int8_t *)(hbh + 1), 796 ((hbh->ip6h_len + 1) << 3) - sizeof(struct ip6_hbh), 797 &dummy1, &dummy2) < 0) { 798 /* m was already freed at this point */ 799 error = EINVAL;/* better error? */ 800 goto done; 801 } 802 m->m_flags &= ~M_LOOP; /* XXX */ 803 m->m_pkthdr.rcvif = NULL; 804 } 805 806 #ifdef PFIL_HOOKS 807 /* 808 * Run through list of hooks for output packets. 809 */ 810 if ((error = pfil_run_hooks(&inet6_pfil_hook, &m, ifp, PFIL_OUT)) != 0) 811 goto done; 812 if (m == NULL) 813 goto done; 814 ip6 = mtod(m, struct ip6_hdr *); 815 #endif /* PFIL_HOOKS */ 816 /* 817 * Send the packet to the outgoing interface. 818 * If necessary, do IPv6 fragmentation before sending. 819 * 820 * the logic here is rather complex: 821 * 1: normal case (dontfrag == 0, alwaysfrag == 0) 822 * 1-a: send as is if tlen <= path mtu 823 * 1-b: fragment if tlen > path mtu 824 * 825 * 2: if user asks us not to fragment (dontfrag == 1) 826 * 2-a: send as is if tlen <= interface mtu 827 * 2-b: error if tlen > interface mtu 828 * 829 * 3: if we always need to attach fragment header (alwaysfrag == 1) 830 * always fragment 831 * 832 * 4: if dontfrag == 1 && alwaysfrag == 1 833 * error, as we cannot handle this conflicting request 834 */ 835 tlen = m->m_pkthdr.len; 836 837 dontfrag = 0; 838 if (dontfrag && alwaysfrag) { /* case 4 */ 839 /* conflicting request - can't transmit */ 840 error = EMSGSIZE; 841 goto bad; 842 } 843 if (dontfrag && tlen > IN6_LINKMTU(ifp)) { /* case 2-b */ 844 /* 845 * Even if the DONTFRAG option is specified, we cannot send the 846 * packet when the data length is larger than the MTU of the 847 * outgoing interface. 848 * Notify the error by sending IPV6_PATHMTU ancillary data as 849 * well as returning an error code (the latter is not described 850 * in the API spec.) 851 */ 852 u_int32_t mtu32; 853 struct ip6ctlparam ip6cp; 854 855 mtu32 = (u_int32_t)mtu; 856 bzero(&ip6cp, sizeof(ip6cp)); 857 ip6cp.ip6c_cmdarg = (void *)&mtu32; 858 pfctlinput2(PRC_MSGSIZE, (struct sockaddr *)&ro_pmtu->ro_dst, 859 (void *)&ip6cp); 860 861 error = EMSGSIZE; 862 goto bad; 863 } 864 865 /* 866 * transmit packet without fragmentation 867 */ 868 if (dontfrag || (!alwaysfrag && tlen <= mtu)) { /* case 1-a and 2-a */ 869 struct in6_ifaddr *ia6; 870 871 ip6 = mtod(m, struct ip6_hdr *); 872 ia6 = in6_ifawithifp(ifp, &ip6->ip6_src); 873 if (ia6) { 874 /* Record statistics for this interface address. */ 875 ia6->ia_ifa.ifa_data.ifad_outbytes += m->m_pkthdr.len; 876 } 877 #ifdef IPSEC 878 /* clean ipsec history once it goes out of the node */ 879 ipsec_delaux(m); 880 #endif 881 error = nd6_output(ifp, origifp, m, dst, ro->ro_rt); 882 goto done; 883 } 884 885 /* 886 * try to fragment the packet. case 1-b and 3 887 */ 888 if (mtu < IPV6_MMTU) { 889 /* path MTU cannot be less than IPV6_MMTU */ 890 error = EMSGSIZE; 891 in6_ifstat_inc(ifp, ifs6_out_fragfail); 892 goto bad; 893 } else if (ip6->ip6_plen == 0) { 894 /* jumbo payload cannot be fragmented */ 895 error = EMSGSIZE; 896 in6_ifstat_inc(ifp, ifs6_out_fragfail); 897 goto bad; 898 } else { 899 struct mbuf **mnext, *m_frgpart; 900 struct ip6_frag *ip6f; 901 u_int32_t id = htonl(ip6_randomid()); 902 u_char nextproto; 903 struct ip6ctlparam ip6cp; 904 u_int32_t mtu32; 905 906 /* 907 * Too large for the destination or interface; 908 * fragment if possible. 909 * Must be able to put at least 8 bytes per fragment. 910 */ 911 hlen = unfragpartlen; 912 if (mtu > IPV6_MAXPACKET) 913 mtu = IPV6_MAXPACKET; 914 915 /* Notify a proper path MTU to applications. */ 916 mtu32 = (u_int32_t)mtu; 917 bzero(&ip6cp, sizeof(ip6cp)); 918 ip6cp.ip6c_cmdarg = (void *)&mtu32; 919 pfctlinput2(PRC_MSGSIZE, (struct sockaddr *)&ro_pmtu->ro_dst, 920 (void *)&ip6cp); 921 922 len = (mtu - hlen - sizeof(struct ip6_frag)) & ~7; 923 if (len < 8) { 924 error = EMSGSIZE; 925 in6_ifstat_inc(ifp, ifs6_out_fragfail); 926 goto bad; 927 } 928 929 mnext = &m->m_nextpkt; 930 931 /* 932 * Change the next header field of the last header in the 933 * unfragmentable part. 934 */ 935 if (exthdrs.ip6e_rthdr) { 936 nextproto = *mtod(exthdrs.ip6e_rthdr, u_char *); 937 *mtod(exthdrs.ip6e_rthdr, u_char *) = IPPROTO_FRAGMENT; 938 } else if (exthdrs.ip6e_dest1) { 939 nextproto = *mtod(exthdrs.ip6e_dest1, u_char *); 940 *mtod(exthdrs.ip6e_dest1, u_char *) = IPPROTO_FRAGMENT; 941 } else if (exthdrs.ip6e_hbh) { 942 nextproto = *mtod(exthdrs.ip6e_hbh, u_char *); 943 *mtod(exthdrs.ip6e_hbh, u_char *) = IPPROTO_FRAGMENT; 944 } else { 945 nextproto = ip6->ip6_nxt; 946 ip6->ip6_nxt = IPPROTO_FRAGMENT; 947 } 948 949 /* 950 * Loop through length of segment after first fragment, 951 * make new header and copy data of each part and link onto 952 * chain. 953 */ 954 m0 = m; 955 for (off = hlen; off < tlen; off += len) { 956 struct mbuf *mlast; 957 958 MGETHDR(m, M_DONTWAIT, MT_HEADER); 959 if (!m) { 960 error = ENOBUFS; 961 ip6stat.ip6s_odropped++; 962 goto sendorfree; 963 } 964 m->m_pkthdr.rcvif = NULL; 965 m->m_flags = m0->m_flags & M_COPYFLAGS; 966 *mnext = m; 967 mnext = &m->m_nextpkt; 968 m->m_data += max_linkhdr; 969 mhip6 = mtod(m, struct ip6_hdr *); 970 *mhip6 = *ip6; 971 m->m_len = sizeof(*mhip6); 972 error = ip6_insertfraghdr(m0, m, hlen, &ip6f); 973 if (error) { 974 ip6stat.ip6s_odropped++; 975 goto sendorfree; 976 } 977 ip6f->ip6f_offlg = htons((u_int16_t)((off - hlen) & ~7)); 978 if (off + len >= tlen) 979 len = tlen - off; 980 else 981 ip6f->ip6f_offlg |= IP6F_MORE_FRAG; 982 mhip6->ip6_plen = htons((u_int16_t)(len + hlen + 983 sizeof(*ip6f) - sizeof(struct ip6_hdr))); 984 if ((m_frgpart = m_copy(m0, off, len)) == 0) { 985 error = ENOBUFS; 986 ip6stat.ip6s_odropped++; 987 goto sendorfree; 988 } 989 for (mlast = m; mlast->m_next; mlast = mlast->m_next) 990 ; 991 mlast->m_next = m_frgpart; 992 m->m_pkthdr.len = len + hlen + sizeof(*ip6f); 993 m->m_pkthdr.rcvif = (struct ifnet *)0; 994 ip6f->ip6f_reserved = 0; 995 ip6f->ip6f_ident = id; 996 ip6f->ip6f_nxt = nextproto; 997 ip6stat.ip6s_ofragments++; 998 in6_ifstat_inc(ifp, ifs6_out_fragcreat); 999 } 1000 1001 in6_ifstat_inc(ifp, ifs6_out_fragok); 1002 } 1003 1004 /* 1005 * Remove leading garbages. 1006 */ 1007 sendorfree: 1008 m = m0->m_nextpkt; 1009 m0->m_nextpkt = 0; 1010 m_freem(m0); 1011 for (m0 = m; m; m = m0) { 1012 m0 = m->m_nextpkt; 1013 m->m_nextpkt = 0; 1014 if (error == 0) { 1015 struct in6_ifaddr *ia6; 1016 ip6 = mtod(m, struct ip6_hdr *); 1017 ia6 = in6_ifawithifp(ifp, &ip6->ip6_src); 1018 if (ia6) { 1019 /* 1020 * Record statistics for this interface 1021 * address. 1022 */ 1023 ia6->ia_ifa.ifa_data.ifad_outbytes += 1024 m->m_pkthdr.len; 1025 } 1026 #ifdef IPSEC 1027 /* clean ipsec history once it goes out of the node */ 1028 ipsec_delaux(m); 1029 #endif 1030 error = nd6_output(ifp, origifp, m, dst, ro->ro_rt); 1031 } else 1032 m_freem(m); 1033 } 1034 1035 if (error == 0) 1036 ip6stat.ip6s_fragmented++; 1037 1038 done: 1039 if (ro == &ip6route && ro->ro_rt) { /* brace necessary for RTFREE */ 1040 RTFREE(ro->ro_rt); 1041 } else if (ro_pmtu == &ip6route && ro_pmtu->ro_rt) { 1042 RTFREE(ro_pmtu->ro_rt); 1043 } 1044 1045 #ifdef IPSEC 1046 if (sp != NULL) 1047 key_freesp(sp); 1048 #endif /* IPSEC */ 1049 1050 return (error); 1051 1052 freehdrs: 1053 m_freem(exthdrs.ip6e_hbh); /* m_freem will check if mbuf is 0 */ 1054 m_freem(exthdrs.ip6e_dest1); 1055 m_freem(exthdrs.ip6e_rthdr); 1056 m_freem(exthdrs.ip6e_dest2); 1057 /* FALLTHROUGH */ 1058 bad: 1059 m_freem(m); 1060 goto done; 1061 } 1062 1063 static int 1064 ip6_copyexthdr(mp, hdr, hlen) 1065 struct mbuf **mp; 1066 caddr_t hdr; 1067 int hlen; 1068 { 1069 struct mbuf *m; 1070 1071 if (hlen > MCLBYTES) 1072 return (ENOBUFS); /* XXX */ 1073 1074 MGET(m, M_DONTWAIT, MT_DATA); 1075 if (!m) 1076 return (ENOBUFS); 1077 1078 if (hlen > MLEN) { 1079 MCLGET(m, M_DONTWAIT); 1080 if ((m->m_flags & M_EXT) == 0) { 1081 m_free(m); 1082 return (ENOBUFS); 1083 } 1084 } 1085 m->m_len = hlen; 1086 if (hdr) 1087 bcopy(hdr, mtod(m, caddr_t), hlen); 1088 1089 *mp = m; 1090 return (0); 1091 } 1092 1093 /* 1094 * Insert jumbo payload option. 1095 */ 1096 static int 1097 ip6_insert_jumboopt(exthdrs, plen) 1098 struct ip6_exthdrs *exthdrs; 1099 u_int32_t plen; 1100 { 1101 struct mbuf *mopt; 1102 u_int8_t *optbuf; 1103 u_int32_t v; 1104 1105 #define JUMBOOPTLEN 8 /* length of jumbo payload option and padding */ 1106 1107 /* 1108 * If there is no hop-by-hop options header, allocate new one. 1109 * If there is one but it doesn't have enough space to store the 1110 * jumbo payload option, allocate a cluster to store the whole options. 1111 * Otherwise, use it to store the options. 1112 */ 1113 if (exthdrs->ip6e_hbh == 0) { 1114 MGET(mopt, M_DONTWAIT, MT_DATA); 1115 if (mopt == 0) 1116 return (ENOBUFS); 1117 mopt->m_len = JUMBOOPTLEN; 1118 optbuf = mtod(mopt, u_int8_t *); 1119 optbuf[1] = 0; /* = ((JUMBOOPTLEN) >> 3) - 1 */ 1120 exthdrs->ip6e_hbh = mopt; 1121 } else { 1122 struct ip6_hbh *hbh; 1123 1124 mopt = exthdrs->ip6e_hbh; 1125 if (M_TRAILINGSPACE(mopt) < JUMBOOPTLEN) { 1126 /* 1127 * XXX assumption: 1128 * - exthdrs->ip6e_hbh is not referenced from places 1129 * other than exthdrs. 1130 * - exthdrs->ip6e_hbh is not an mbuf chain. 1131 */ 1132 int oldoptlen = mopt->m_len; 1133 struct mbuf *n; 1134 1135 /* 1136 * XXX: give up if the whole (new) hbh header does 1137 * not fit even in an mbuf cluster. 1138 */ 1139 if (oldoptlen + JUMBOOPTLEN > MCLBYTES) 1140 return (ENOBUFS); 1141 1142 /* 1143 * As a consequence, we must always prepare a cluster 1144 * at this point. 1145 */ 1146 MGET(n, M_DONTWAIT, MT_DATA); 1147 if (n) { 1148 MCLGET(n, M_DONTWAIT); 1149 if ((n->m_flags & M_EXT) == 0) { 1150 m_freem(n); 1151 n = NULL; 1152 } 1153 } 1154 if (!n) 1155 return (ENOBUFS); 1156 n->m_len = oldoptlen + JUMBOOPTLEN; 1157 bcopy(mtod(mopt, caddr_t), mtod(n, caddr_t), 1158 oldoptlen); 1159 optbuf = mtod(n, u_int8_t *) + oldoptlen; 1160 m_freem(mopt); 1161 mopt = exthdrs->ip6e_hbh = n; 1162 } else { 1163 optbuf = mtod(mopt, u_int8_t *) + mopt->m_len; 1164 mopt->m_len += JUMBOOPTLEN; 1165 } 1166 optbuf[0] = IP6OPT_PADN; 1167 optbuf[1] = 0; 1168 1169 /* 1170 * Adjust the header length according to the pad and 1171 * the jumbo payload option. 1172 */ 1173 hbh = mtod(mopt, struct ip6_hbh *); 1174 hbh->ip6h_len += (JUMBOOPTLEN >> 3); 1175 } 1176 1177 /* fill in the option. */ 1178 optbuf[2] = IP6OPT_JUMBO; 1179 optbuf[3] = 4; 1180 v = (u_int32_t)htonl(plen + JUMBOOPTLEN); 1181 bcopy(&v, &optbuf[4], sizeof(u_int32_t)); 1182 1183 /* finally, adjust the packet header length */ 1184 exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN; 1185 1186 return (0); 1187 #undef JUMBOOPTLEN 1188 } 1189 1190 /* 1191 * Insert fragment header and copy unfragmentable header portions. 1192 */ 1193 static int 1194 ip6_insertfraghdr(m0, m, hlen, frghdrp) 1195 struct mbuf *m0, *m; 1196 int hlen; 1197 struct ip6_frag **frghdrp; 1198 { 1199 struct mbuf *n, *mlast; 1200 1201 if (hlen > sizeof(struct ip6_hdr)) { 1202 n = m_copym(m0, sizeof(struct ip6_hdr), 1203 hlen - sizeof(struct ip6_hdr), M_DONTWAIT); 1204 if (n == 0) 1205 return (ENOBUFS); 1206 m->m_next = n; 1207 } else 1208 n = m; 1209 1210 /* Search for the last mbuf of unfragmentable part. */ 1211 for (mlast = n; mlast->m_next; mlast = mlast->m_next) 1212 ; 1213 1214 if ((mlast->m_flags & M_EXT) == 0 && 1215 M_TRAILINGSPACE(mlast) >= sizeof(struct ip6_frag)) { 1216 /* use the trailing space of the last mbuf for the fragment hdr */ 1217 *frghdrp = (struct ip6_frag *)(mtod(mlast, caddr_t) + 1218 mlast->m_len); 1219 mlast->m_len += sizeof(struct ip6_frag); 1220 m->m_pkthdr.len += sizeof(struct ip6_frag); 1221 } else { 1222 /* allocate a new mbuf for the fragment header */ 1223 struct mbuf *mfrg; 1224 1225 MGET(mfrg, M_DONTWAIT, MT_DATA); 1226 if (mfrg == 0) 1227 return (ENOBUFS); 1228 mfrg->m_len = sizeof(struct ip6_frag); 1229 *frghdrp = mtod(mfrg, struct ip6_frag *); 1230 mlast->m_next = mfrg; 1231 } 1232 1233 return (0); 1234 } 1235 1236 int 1237 ip6_getpmtu(ro_pmtu, ro, ifp, dst, mtup, alwaysfragp) 1238 struct route_in6 *ro_pmtu, *ro; 1239 struct ifnet *ifp; 1240 struct in6_addr *dst; 1241 u_long *mtup; 1242 int *alwaysfragp; 1243 { 1244 u_int32_t mtu = 0; 1245 int alwaysfrag = 0; 1246 int error = 0; 1247 1248 if (ro_pmtu != ro) { 1249 /* The first hop and the final destination may differ. */ 1250 struct sockaddr_in6 *sa6_dst = 1251 (struct sockaddr_in6 *)&ro_pmtu->ro_dst; 1252 if (ro_pmtu->ro_rt && 1253 ((ro_pmtu->ro_rt->rt_flags & RTF_UP) == 0 || 1254 !IN6_ARE_ADDR_EQUAL(&sa6_dst->sin6_addr, dst))) { 1255 RTFREE(ro_pmtu->ro_rt); 1256 ro_pmtu->ro_rt = (struct rtentry *)NULL; 1257 } 1258 if (ro_pmtu->ro_rt == NULL) { 1259 bzero(sa6_dst, sizeof(*sa6_dst)); /* for safety */ 1260 sa6_dst->sin6_family = AF_INET6; 1261 sa6_dst->sin6_len = sizeof(struct sockaddr_in6); 1262 sa6_dst->sin6_addr = *dst; 1263 1264 rtalloc((struct route *)ro_pmtu); 1265 } 1266 } 1267 if (ro_pmtu->ro_rt) { 1268 u_int32_t ifmtu; 1269 1270 if (ifp == NULL) 1271 ifp = ro_pmtu->ro_rt->rt_ifp; 1272 ifmtu = IN6_LINKMTU(ifp); 1273 mtu = ro_pmtu->ro_rt->rt_rmx.rmx_mtu; 1274 if (mtu == 0) 1275 mtu = ifmtu; 1276 else if (mtu < IPV6_MMTU) { 1277 /* 1278 * RFC2460 section 5, last paragraph: 1279 * if we record ICMPv6 too big message with 1280 * mtu < IPV6_MMTU, transmit packets sized IPV6_MMTU 1281 * or smaller, with fragment header attached. 1282 * (fragment header is needed regardless from the 1283 * packet size, for translators to identify packets) 1284 */ 1285 alwaysfrag = 1; 1286 mtu = IPV6_MMTU; 1287 } else if (mtu > ifmtu) { 1288 /* 1289 * The MTU on the route is larger than the MTU on 1290 * the interface! This shouldn't happen, unless the 1291 * MTU of the interface has been changed after the 1292 * interface was brought up. Change the MTU in the 1293 * route to match the interface MTU (as long as the 1294 * field isn't locked). 1295 */ 1296 mtu = ifmtu; 1297 if (!(ro_pmtu->ro_rt->rt_rmx.rmx_locks & RTV_MTU)) 1298 ro_pmtu->ro_rt->rt_rmx.rmx_mtu = mtu; 1299 } 1300 } else if (ifp) { 1301 mtu = IN6_LINKMTU(ifp); 1302 } else 1303 error = EHOSTUNREACH; /* XXX */ 1304 1305 *mtup = mtu; 1306 if (alwaysfragp) 1307 *alwaysfragp = alwaysfrag; 1308 return (error); 1309 } 1310 1311 /* 1312 * IP6 socket option processing. 1313 */ 1314 int 1315 ip6_ctloutput(op, so, level, optname, mp) 1316 int op; 1317 struct socket *so; 1318 int level, optname; 1319 struct mbuf **mp; 1320 { 1321 struct in6pcb *in6p = sotoin6pcb(so); 1322 struct mbuf *m = *mp; 1323 int optval = 0; 1324 int error = 0; 1325 struct proc *p = curproc; /* XXX */ 1326 1327 if (level == IPPROTO_IPV6) { 1328 switch (op) { 1329 case PRCO_SETOPT: 1330 switch (optname) { 1331 case IPV6_PKTOPTIONS: 1332 /* m is freed in ip6_pcbopts */ 1333 return (ip6_pcbopts(&in6p->in6p_outputopts, 1334 m, so)); 1335 case IPV6_HOPOPTS: 1336 case IPV6_DSTOPTS: 1337 if (p == 0 || suser(p->p_ucred, &p->p_acflag)) { 1338 error = EPERM; 1339 break; 1340 } 1341 /* FALLTHROUGH */ 1342 case IPV6_UNICAST_HOPS: 1343 case IPV6_RECVOPTS: 1344 case IPV6_RECVRETOPTS: 1345 case IPV6_RECVDSTADDR: 1346 case IPV6_PKTINFO: 1347 case IPV6_HOPLIMIT: 1348 case IPV6_RTHDR: 1349 case IPV6_FAITH: 1350 case IPV6_V6ONLY: 1351 case IPV6_USE_MIN_MTU: 1352 if (!m || m->m_len != sizeof(int)) { 1353 error = EINVAL; 1354 break; 1355 } 1356 optval = *mtod(m, int *); 1357 switch (optname) { 1358 1359 case IPV6_UNICAST_HOPS: 1360 if (optval < -1 || optval >= 256) 1361 error = EINVAL; 1362 else { 1363 /* -1 = kernel default */ 1364 in6p->in6p_hops = optval; 1365 } 1366 break; 1367 #define OPTSET(bit) \ 1368 do { \ 1369 if (optval) \ 1370 in6p->in6p_flags |= (bit); \ 1371 else \ 1372 in6p->in6p_flags &= ~(bit); \ 1373 } while (/*CONSTCOND*/ 0) 1374 1375 case IPV6_RECVOPTS: 1376 OPTSET(IN6P_RECVOPTS); 1377 break; 1378 1379 case IPV6_RECVRETOPTS: 1380 OPTSET(IN6P_RECVRETOPTS); 1381 break; 1382 1383 case IPV6_RECVDSTADDR: 1384 OPTSET(IN6P_RECVDSTADDR); 1385 break; 1386 1387 case IPV6_PKTINFO: 1388 OPTSET(IN6P_PKTINFO); 1389 break; 1390 1391 case IPV6_HOPLIMIT: 1392 OPTSET(IN6P_HOPLIMIT); 1393 break; 1394 1395 case IPV6_HOPOPTS: 1396 OPTSET(IN6P_HOPOPTS); 1397 break; 1398 1399 case IPV6_DSTOPTS: 1400 OPTSET(IN6P_DSTOPTS); 1401 break; 1402 1403 case IPV6_RTHDR: 1404 OPTSET(IN6P_RTHDR); 1405 break; 1406 1407 case IPV6_FAITH: 1408 OPTSET(IN6P_FAITH); 1409 break; 1410 1411 case IPV6_USE_MIN_MTU: 1412 OPTSET(IN6P_MINMTU); 1413 break; 1414 1415 case IPV6_V6ONLY: 1416 /* 1417 * make setsockopt(IPV6_V6ONLY) 1418 * available only prior to bind(2). 1419 * see ipng mailing list, Jun 22 2001. 1420 */ 1421 if (in6p->in6p_lport || 1422 !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr)) { 1423 error = EINVAL; 1424 break; 1425 } 1426 #ifdef INET6_BINDV6ONLY 1427 if (!optval) 1428 error = EINVAL; 1429 #else 1430 OPTSET(IN6P_IPV6_V6ONLY); 1431 #endif 1432 break; 1433 } 1434 break; 1435 #undef OPTSET 1436 1437 case IPV6_MULTICAST_IF: 1438 case IPV6_MULTICAST_HOPS: 1439 case IPV6_MULTICAST_LOOP: 1440 case IPV6_JOIN_GROUP: 1441 case IPV6_LEAVE_GROUP: 1442 error = ip6_setmoptions(optname, 1443 &in6p->in6p_moptions, m); 1444 break; 1445 1446 case IPV6_PORTRANGE: 1447 optval = *mtod(m, int *); 1448 1449 switch (optval) { 1450 case IPV6_PORTRANGE_DEFAULT: 1451 in6p->in6p_flags &= ~(IN6P_LOWPORT); 1452 in6p->in6p_flags &= ~(IN6P_HIGHPORT); 1453 break; 1454 1455 case IPV6_PORTRANGE_HIGH: 1456 in6p->in6p_flags &= ~(IN6P_LOWPORT); 1457 in6p->in6p_flags |= IN6P_HIGHPORT; 1458 break; 1459 1460 case IPV6_PORTRANGE_LOW: 1461 in6p->in6p_flags &= ~(IN6P_HIGHPORT); 1462 in6p->in6p_flags |= IN6P_LOWPORT; 1463 break; 1464 1465 default: 1466 error = EINVAL; 1467 break; 1468 } 1469 break; 1470 1471 #ifdef IPSEC 1472 case IPV6_IPSEC_POLICY: 1473 { 1474 caddr_t req = NULL; 1475 size_t len = 0; 1476 1477 int priv = 0; 1478 if (p == 0 || suser(p->p_ucred, &p->p_acflag)) 1479 priv = 0; 1480 else 1481 priv = 1; 1482 if (m) { 1483 req = mtod(m, caddr_t); 1484 len = m->m_len; 1485 } 1486 error = ipsec6_set_policy(in6p, 1487 optname, req, len, priv); 1488 } 1489 break; 1490 #endif /* IPSEC */ 1491 1492 default: 1493 error = ENOPROTOOPT; 1494 break; 1495 } 1496 if (m) 1497 (void)m_free(m); 1498 break; 1499 1500 case PRCO_GETOPT: 1501 switch (optname) { 1502 1503 case IPV6_OPTIONS: 1504 case IPV6_RETOPTS: 1505 error = ENOPROTOOPT; 1506 break; 1507 1508 case IPV6_PKTOPTIONS: 1509 if (in6p->in6p_options) { 1510 *mp = m_copym(in6p->in6p_options, 0, 1511 M_COPYALL, M_WAIT); 1512 } else { 1513 *mp = m_get(M_WAIT, MT_SOOPTS); 1514 (*mp)->m_len = 0; 1515 } 1516 break; 1517 1518 case IPV6_HOPOPTS: 1519 case IPV6_DSTOPTS: 1520 if (p == 0 || suser(p->p_ucred, &p->p_acflag)) { 1521 error = EPERM; 1522 break; 1523 } 1524 /* FALLTHROUGH */ 1525 case IPV6_UNICAST_HOPS: 1526 case IPV6_RECVOPTS: 1527 case IPV6_RECVRETOPTS: 1528 case IPV6_RECVDSTADDR: 1529 case IPV6_PORTRANGE: 1530 case IPV6_PKTINFO: 1531 case IPV6_HOPLIMIT: 1532 case IPV6_RTHDR: 1533 case IPV6_FAITH: 1534 case IPV6_V6ONLY: 1535 case IPV6_USE_MIN_MTU: 1536 *mp = m = m_get(M_WAIT, MT_SOOPTS); 1537 m->m_len = sizeof(int); 1538 switch (optname) { 1539 1540 case IPV6_UNICAST_HOPS: 1541 optval = in6p->in6p_hops; 1542 break; 1543 1544 #define OPTBIT(bit) (in6p->in6p_flags & bit ? 1 : 0) 1545 1546 case IPV6_RECVOPTS: 1547 optval = OPTBIT(IN6P_RECVOPTS); 1548 break; 1549 1550 case IPV6_RECVRETOPTS: 1551 optval = OPTBIT(IN6P_RECVRETOPTS); 1552 break; 1553 1554 case IPV6_RECVDSTADDR: 1555 optval = OPTBIT(IN6P_RECVDSTADDR); 1556 break; 1557 1558 case IPV6_PORTRANGE: 1559 { 1560 int flags; 1561 flags = in6p->in6p_flags; 1562 if (flags & IN6P_HIGHPORT) 1563 optval = IPV6_PORTRANGE_HIGH; 1564 else if (flags & IN6P_LOWPORT) 1565 optval = IPV6_PORTRANGE_LOW; 1566 else 1567 optval = 0; 1568 break; 1569 } 1570 1571 case IPV6_PKTINFO: 1572 optval = OPTBIT(IN6P_PKTINFO); 1573 break; 1574 1575 case IPV6_HOPLIMIT: 1576 optval = OPTBIT(IN6P_HOPLIMIT); 1577 break; 1578 1579 case IPV6_HOPOPTS: 1580 optval = OPTBIT(IN6P_HOPOPTS); 1581 break; 1582 1583 case IPV6_DSTOPTS: 1584 optval = OPTBIT(IN6P_DSTOPTS); 1585 break; 1586 1587 case IPV6_RTHDR: 1588 optval = OPTBIT(IN6P_RTHDR); 1589 break; 1590 1591 case IPV6_FAITH: 1592 optval = OPTBIT(IN6P_FAITH); 1593 break; 1594 1595 case IPV6_V6ONLY: 1596 optval = OPTBIT(IN6P_IPV6_V6ONLY); 1597 break; 1598 1599 case IPV6_USE_MIN_MTU: 1600 optval = OPTBIT(IN6P_MINMTU); 1601 break; 1602 } 1603 *mtod(m, int *) = optval; 1604 break; 1605 1606 case IPV6_MULTICAST_IF: 1607 case IPV6_MULTICAST_HOPS: 1608 case IPV6_MULTICAST_LOOP: 1609 case IPV6_JOIN_GROUP: 1610 case IPV6_LEAVE_GROUP: 1611 error = ip6_getmoptions(optname, in6p->in6p_moptions, mp); 1612 break; 1613 1614 #ifdef IPSEC 1615 case IPV6_IPSEC_POLICY: 1616 { 1617 caddr_t req = NULL; 1618 size_t len = 0; 1619 1620 if (m) { 1621 req = mtod(m, caddr_t); 1622 len = m->m_len; 1623 } 1624 error = ipsec6_get_policy(in6p, req, len, mp); 1625 break; 1626 } 1627 #endif /* IPSEC */ 1628 1629 default: 1630 error = ENOPROTOOPT; 1631 break; 1632 } 1633 break; 1634 } 1635 } else { 1636 error = EINVAL; 1637 if (op == PRCO_SETOPT && *mp) 1638 (void)m_free(*mp); 1639 } 1640 return (error); 1641 } 1642 1643 int 1644 ip6_raw_ctloutput(op, so, level, optname, mp) 1645 int op; 1646 struct socket *so; 1647 int level, optname; 1648 struct mbuf **mp; 1649 { 1650 int error = 0, optval, optlen; 1651 const int icmp6off = offsetof(struct icmp6_hdr, icmp6_cksum); 1652 struct in6pcb *in6p = sotoin6pcb(so); 1653 struct mbuf *m = *mp; 1654 1655 optlen = m ? m->m_len : 0; 1656 1657 if (level != IPPROTO_IPV6) { 1658 if (op == PRCO_SETOPT && *mp) 1659 (void)m_free(*mp); 1660 return (EINVAL); 1661 } 1662 1663 switch (optname) { 1664 case IPV6_CHECKSUM: 1665 /* 1666 * For ICMPv6 sockets, no modification allowed for checksum 1667 * offset, permit "no change" values to help existing apps. 1668 * 1669 * XXX 2292bis says: "An attempt to set IPV6_CHECKSUM 1670 * for an ICMPv6 socket will fail." 1671 * The current behavior does not meet 2292bis. 1672 */ 1673 switch (op) { 1674 case PRCO_SETOPT: 1675 if (optlen != sizeof(int)) { 1676 error = EINVAL; 1677 break; 1678 } 1679 optval = *mtod(m, int *); 1680 if ((optval % 2) != 0) { 1681 /* the API assumes even offset values */ 1682 error = EINVAL; 1683 } else if (so->so_proto->pr_protocol == 1684 IPPROTO_ICMPV6) { 1685 if (optval != icmp6off) 1686 error = EINVAL; 1687 } else 1688 in6p->in6p_cksum = optval; 1689 break; 1690 1691 case PRCO_GETOPT: 1692 if (so->so_proto->pr_protocol == IPPROTO_ICMPV6) 1693 optval = icmp6off; 1694 else 1695 optval = in6p->in6p_cksum; 1696 1697 *mp = m = m_get(M_WAIT, MT_SOOPTS); 1698 m->m_len = sizeof(int); 1699 *mtod(m, int *) = optval; 1700 break; 1701 1702 default: 1703 error = EINVAL; 1704 break; 1705 } 1706 break; 1707 1708 default: 1709 error = ENOPROTOOPT; 1710 break; 1711 } 1712 1713 if (op == PRCO_SETOPT && m) 1714 (void)m_free(m); 1715 1716 return (error); 1717 } 1718 1719 /* 1720 * Set up IP6 options in pcb for insertion in output packets. 1721 * Store in mbuf with pointer in pcbopt, adding pseudo-option 1722 * with destination address if source routed. 1723 */ 1724 static int 1725 ip6_pcbopts(pktopt, m, so) 1726 struct ip6_pktopts **pktopt; 1727 struct mbuf *m; 1728 struct socket *so; 1729 { 1730 struct ip6_pktopts *opt = *pktopt; 1731 int error = 0; 1732 struct proc *p = curproc; /* XXX */ 1733 int priv = 0; 1734 1735 /* turn off any old options. */ 1736 if (opt) { 1737 if (opt->ip6po_m) 1738 (void)m_free(opt->ip6po_m); 1739 } else 1740 opt = malloc(sizeof(*opt), M_IP6OPT, M_WAITOK); 1741 *pktopt = 0; 1742 1743 if (!m || m->m_len == 0) { 1744 /* 1745 * Only turning off any previous options. 1746 */ 1747 free(opt, M_IP6OPT); 1748 if (m) 1749 (void)m_free(m); 1750 return (0); 1751 } 1752 1753 /* set options specified by user. */ 1754 if (p && !suser(p->p_ucred, &p->p_acflag)) 1755 priv = 1; 1756 if ((error = ip6_setpktoptions(m, opt, priv)) != 0) { 1757 (void)m_free(m); 1758 free(opt, M_IP6OPT); 1759 return (error); 1760 } 1761 *pktopt = opt; 1762 return (0); 1763 } 1764 1765 /* 1766 * Set the IP6 multicast options in response to user setsockopt(). 1767 */ 1768 static int 1769 ip6_setmoptions(optname, im6op, m) 1770 int optname; 1771 struct ip6_moptions **im6op; 1772 struct mbuf *m; 1773 { 1774 int error = 0; 1775 u_int loop, ifindex; 1776 struct ipv6_mreq *mreq; 1777 struct ifnet *ifp; 1778 struct ip6_moptions *im6o = *im6op; 1779 struct route_in6 ro; 1780 struct sockaddr_in6 *dst; 1781 struct in6_multi_mship *imm; 1782 struct proc *p = curproc; /* XXX */ 1783 1784 if (im6o == NULL) { 1785 /* 1786 * No multicast option buffer attached to the pcb; 1787 * allocate one and initialize to default values. 1788 */ 1789 im6o = (struct ip6_moptions *) 1790 malloc(sizeof(*im6o), M_IPMOPTS, M_WAITOK); 1791 1792 if (im6o == NULL) 1793 return (ENOBUFS); 1794 *im6op = im6o; 1795 im6o->im6o_multicast_ifp = NULL; 1796 im6o->im6o_multicast_hlim = ip6_defmcasthlim; 1797 im6o->im6o_multicast_loop = IPV6_DEFAULT_MULTICAST_LOOP; 1798 LIST_INIT(&im6o->im6o_memberships); 1799 } 1800 1801 switch (optname) { 1802 1803 case IPV6_MULTICAST_IF: 1804 /* 1805 * Select the interface for outgoing multicast packets. 1806 */ 1807 if (m == NULL || m->m_len != sizeof(u_int)) { 1808 error = EINVAL; 1809 break; 1810 } 1811 bcopy(mtod(m, u_int *), &ifindex, sizeof(ifindex)); 1812 if (ifindex < 0 || if_indexlim <= ifindex || 1813 !ifindex2ifnet[ifindex]) { 1814 error = ENXIO; /* XXX EINVAL? */ 1815 break; 1816 } 1817 ifp = ifindex2ifnet[ifindex]; 1818 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) { 1819 error = EADDRNOTAVAIL; 1820 break; 1821 } 1822 im6o->im6o_multicast_ifp = ifp; 1823 break; 1824 1825 case IPV6_MULTICAST_HOPS: 1826 { 1827 /* 1828 * Set the IP6 hoplimit for outgoing multicast packets. 1829 */ 1830 int optval; 1831 if (m == NULL || m->m_len != sizeof(int)) { 1832 error = EINVAL; 1833 break; 1834 } 1835 bcopy(mtod(m, u_int *), &optval, sizeof(optval)); 1836 if (optval < -1 || optval >= 256) 1837 error = EINVAL; 1838 else if (optval == -1) 1839 im6o->im6o_multicast_hlim = ip6_defmcasthlim; 1840 else 1841 im6o->im6o_multicast_hlim = optval; 1842 break; 1843 } 1844 1845 case IPV6_MULTICAST_LOOP: 1846 /* 1847 * Set the loopback flag for outgoing multicast packets. 1848 * Must be zero or one. 1849 */ 1850 if (m == NULL || m->m_len != sizeof(u_int)) { 1851 error = EINVAL; 1852 break; 1853 } 1854 bcopy(mtod(m, u_int *), &loop, sizeof(loop)); 1855 if (loop > 1) { 1856 error = EINVAL; 1857 break; 1858 } 1859 im6o->im6o_multicast_loop = loop; 1860 break; 1861 1862 case IPV6_JOIN_GROUP: 1863 /* 1864 * Add a multicast group membership. 1865 * Group must be a valid IP6 multicast address. 1866 */ 1867 if (m == NULL || m->m_len != sizeof(struct ipv6_mreq)) { 1868 error = EINVAL; 1869 break; 1870 } 1871 mreq = mtod(m, struct ipv6_mreq *); 1872 if (IN6_IS_ADDR_UNSPECIFIED(&mreq->ipv6mr_multiaddr)) { 1873 /* 1874 * We use the unspecified address to specify to accept 1875 * all multicast addresses. Only super user is allowed 1876 * to do this. 1877 */ 1878 if (suser(p->p_ucred, &p->p_acflag)) 1879 { 1880 error = EACCES; 1881 break; 1882 } 1883 } else if (!IN6_IS_ADDR_MULTICAST(&mreq->ipv6mr_multiaddr)) { 1884 error = EINVAL; 1885 break; 1886 } 1887 1888 /* 1889 * If the interface is specified, validate it. 1890 */ 1891 if (mreq->ipv6mr_interface < 0 || 1892 if_indexlim <= mreq->ipv6mr_interface || 1893 !ifindex2ifnet[mreq->ipv6mr_interface]) { 1894 error = ENXIO; /* XXX EINVAL? */ 1895 break; 1896 } 1897 /* 1898 * If no interface was explicitly specified, choose an 1899 * appropriate one according to the given multicast address. 1900 */ 1901 if (mreq->ipv6mr_interface == 0) { 1902 /* 1903 * If the multicast address is in node-local scope, 1904 * the interface should be a loopback interface. 1905 * Otherwise, look up the routing table for the 1906 * address, and choose the outgoing interface. 1907 * XXX: is it a good approach? 1908 */ 1909 if (IN6_IS_ADDR_MC_NODELOCAL(&mreq->ipv6mr_multiaddr)) { 1910 ifp = &loif[0]; 1911 } else { 1912 ro.ro_rt = NULL; 1913 dst = (struct sockaddr_in6 *)&ro.ro_dst; 1914 bzero(dst, sizeof(*dst)); 1915 dst->sin6_len = sizeof(struct sockaddr_in6); 1916 dst->sin6_family = AF_INET6; 1917 dst->sin6_addr = mreq->ipv6mr_multiaddr; 1918 rtalloc((struct route *)&ro); 1919 if (ro.ro_rt == NULL) { 1920 error = EADDRNOTAVAIL; 1921 break; 1922 } 1923 ifp = ro.ro_rt->rt_ifp; 1924 rtfree(ro.ro_rt); 1925 } 1926 } else 1927 ifp = ifindex2ifnet[mreq->ipv6mr_interface]; 1928 1929 /* 1930 * See if we found an interface, and confirm that it 1931 * supports multicast 1932 */ 1933 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) { 1934 error = EADDRNOTAVAIL; 1935 break; 1936 } 1937 /* 1938 * Put interface index into the multicast address, 1939 * if the address has link-local scope. 1940 */ 1941 if (IN6_IS_ADDR_MC_LINKLOCAL(&mreq->ipv6mr_multiaddr)) { 1942 mreq->ipv6mr_multiaddr.s6_addr16[1] = 1943 htons(ifp->if_index); 1944 } 1945 /* 1946 * See if the membership already exists. 1947 */ 1948 for (imm = im6o->im6o_memberships.lh_first; 1949 imm != NULL; imm = imm->i6mm_chain.le_next) 1950 if (imm->i6mm_maddr->in6m_ifp == ifp && 1951 IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr, 1952 &mreq->ipv6mr_multiaddr)) 1953 break; 1954 if (imm != NULL) { 1955 error = EADDRINUSE; 1956 break; 1957 } 1958 /* 1959 * Everything looks good; add a new record to the multicast 1960 * address list for the given interface. 1961 */ 1962 imm = in6_joingroup(ifp, &mreq->ipv6mr_multiaddr, &error); 1963 if (!imm) 1964 break; 1965 LIST_INSERT_HEAD(&im6o->im6o_memberships, imm, i6mm_chain); 1966 break; 1967 1968 case IPV6_LEAVE_GROUP: 1969 /* 1970 * Drop a multicast group membership. 1971 * Group must be a valid IP6 multicast address. 1972 */ 1973 if (m == NULL || m->m_len != sizeof(struct ipv6_mreq)) { 1974 error = EINVAL; 1975 break; 1976 } 1977 mreq = mtod(m, struct ipv6_mreq *); 1978 if (IN6_IS_ADDR_UNSPECIFIED(&mreq->ipv6mr_multiaddr)) { 1979 if (suser(p->p_ucred, &p->p_acflag)) 1980 { 1981 error = EACCES; 1982 break; 1983 } 1984 } else if (!IN6_IS_ADDR_MULTICAST(&mreq->ipv6mr_multiaddr)) { 1985 error = EINVAL; 1986 break; 1987 } 1988 /* 1989 * If an interface address was specified, get a pointer 1990 * to its ifnet structure. 1991 */ 1992 if (mreq->ipv6mr_interface < 0 || 1993 if_indexlim <= mreq->ipv6mr_interface || 1994 !ifindex2ifnet[mreq->ipv6mr_interface]) { 1995 error = ENXIO; /* XXX EINVAL? */ 1996 break; 1997 } 1998 ifp = ifindex2ifnet[mreq->ipv6mr_interface]; 1999 /* 2000 * Put interface index into the multicast address, 2001 * if the address has link-local scope. 2002 */ 2003 if (IN6_IS_ADDR_MC_LINKLOCAL(&mreq->ipv6mr_multiaddr)) { 2004 mreq->ipv6mr_multiaddr.s6_addr16[1] = 2005 htons(mreq->ipv6mr_interface); 2006 } 2007 /* 2008 * Find the membership in the membership list. 2009 */ 2010 for (imm = im6o->im6o_memberships.lh_first; 2011 imm != NULL; imm = imm->i6mm_chain.le_next) { 2012 if ((ifp == NULL || imm->i6mm_maddr->in6m_ifp == ifp) && 2013 IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr, 2014 &mreq->ipv6mr_multiaddr)) 2015 break; 2016 } 2017 if (imm == NULL) { 2018 /* Unable to resolve interface */ 2019 error = EADDRNOTAVAIL; 2020 break; 2021 } 2022 /* 2023 * Give up the multicast address record to which the 2024 * membership points. 2025 */ 2026 LIST_REMOVE(imm, i6mm_chain); 2027 in6_leavegroup(imm); 2028 break; 2029 2030 default: 2031 error = EOPNOTSUPP; 2032 break; 2033 } 2034 2035 /* 2036 * If all options have default values, no need to keep the mbuf. 2037 */ 2038 if (im6o->im6o_multicast_ifp == NULL && 2039 im6o->im6o_multicast_hlim == ip6_defmcasthlim && 2040 im6o->im6o_multicast_loop == IPV6_DEFAULT_MULTICAST_LOOP && 2041 im6o->im6o_memberships.lh_first == NULL) { 2042 free(*im6op, M_IPMOPTS); 2043 *im6op = NULL; 2044 } 2045 2046 return (error); 2047 } 2048 2049 /* 2050 * Return the IP6 multicast options in response to user getsockopt(). 2051 */ 2052 static int 2053 ip6_getmoptions(optname, im6o, mp) 2054 int optname; 2055 struct ip6_moptions *im6o; 2056 struct mbuf **mp; 2057 { 2058 u_int *hlim, *loop, *ifindex; 2059 2060 *mp = m_get(M_WAIT, MT_SOOPTS); 2061 2062 switch (optname) { 2063 2064 case IPV6_MULTICAST_IF: 2065 ifindex = mtod(*mp, u_int *); 2066 (*mp)->m_len = sizeof(u_int); 2067 if (im6o == NULL || im6o->im6o_multicast_ifp == NULL) 2068 *ifindex = 0; 2069 else 2070 *ifindex = im6o->im6o_multicast_ifp->if_index; 2071 return (0); 2072 2073 case IPV6_MULTICAST_HOPS: 2074 hlim = mtod(*mp, u_int *); 2075 (*mp)->m_len = sizeof(u_int); 2076 if (im6o == NULL) 2077 *hlim = ip6_defmcasthlim; 2078 else 2079 *hlim = im6o->im6o_multicast_hlim; 2080 return (0); 2081 2082 case IPV6_MULTICAST_LOOP: 2083 loop = mtod(*mp, u_int *); 2084 (*mp)->m_len = sizeof(u_int); 2085 if (im6o == NULL) 2086 *loop = ip6_defmcasthlim; 2087 else 2088 *loop = im6o->im6o_multicast_loop; 2089 return (0); 2090 2091 default: 2092 return (EOPNOTSUPP); 2093 } 2094 } 2095 2096 /* 2097 * Discard the IP6 multicast options. 2098 */ 2099 void 2100 ip6_freemoptions(im6o) 2101 struct ip6_moptions *im6o; 2102 { 2103 struct in6_multi_mship *imm; 2104 2105 if (im6o == NULL) 2106 return; 2107 2108 while ((imm = im6o->im6o_memberships.lh_first) != NULL) { 2109 LIST_REMOVE(imm, i6mm_chain); 2110 in6_leavegroup(imm); 2111 } 2112 free(im6o, M_IPMOPTS); 2113 } 2114 2115 /* 2116 * Set IPv6 outgoing packet options based on advanced API. 2117 */ 2118 int 2119 ip6_setpktoptions(control, opt, priv) 2120 struct mbuf *control; 2121 struct ip6_pktopts *opt; 2122 int priv; 2123 { 2124 struct cmsghdr *cm = 0; 2125 2126 if (control == 0 || opt == 0) 2127 return (EINVAL); 2128 2129 bzero(opt, sizeof(*opt)); 2130 opt->ip6po_hlim = -1; /* -1 means to use default hop limit */ 2131 2132 /* 2133 * XXX: Currently, we assume all the optional information is stored 2134 * in a single mbuf. 2135 */ 2136 if (control->m_next) 2137 return (EINVAL); 2138 2139 opt->ip6po_m = control; 2140 2141 for (; control->m_len; control->m_data += CMSG_ALIGN(cm->cmsg_len), 2142 control->m_len -= CMSG_ALIGN(cm->cmsg_len)) { 2143 cm = mtod(control, struct cmsghdr *); 2144 if (cm->cmsg_len == 0 || cm->cmsg_len > control->m_len) 2145 return (EINVAL); 2146 if (cm->cmsg_level != IPPROTO_IPV6) 2147 continue; 2148 2149 switch (cm->cmsg_type) { 2150 case IPV6_PKTINFO: 2151 if (cm->cmsg_len != CMSG_LEN(sizeof(struct in6_pktinfo))) 2152 return (EINVAL); 2153 opt->ip6po_pktinfo = (struct in6_pktinfo *)CMSG_DATA(cm); 2154 if (opt->ip6po_pktinfo->ipi6_ifindex && 2155 IN6_IS_ADDR_LINKLOCAL(&opt->ip6po_pktinfo->ipi6_addr)) 2156 opt->ip6po_pktinfo->ipi6_addr.s6_addr16[1] = 2157 htons(opt->ip6po_pktinfo->ipi6_ifindex); 2158 2159 if (opt->ip6po_pktinfo->ipi6_ifindex >= if_indexlim || 2160 opt->ip6po_pktinfo->ipi6_ifindex < 0) 2161 return (ENXIO); 2162 if (opt->ip6po_pktinfo->ipi6_ifindex > 0 && 2163 !ifindex2ifnet[opt->ip6po_pktinfo->ipi6_ifindex]) 2164 return (ENXIO); 2165 2166 /* 2167 * Check if the requested source address is indeed a 2168 * unicast address assigned to the node, and can be 2169 * used as the packet's source address. 2170 */ 2171 if (!IN6_IS_ADDR_UNSPECIFIED(&opt->ip6po_pktinfo->ipi6_addr)) { 2172 struct ifaddr *ia; 2173 struct in6_ifaddr *ia6; 2174 struct sockaddr_in6 sin6; 2175 2176 bzero(&sin6, sizeof(sin6)); 2177 sin6.sin6_len = sizeof(sin6); 2178 sin6.sin6_family = AF_INET6; 2179 sin6.sin6_addr = 2180 opt->ip6po_pktinfo->ipi6_addr; 2181 ia = ifa_ifwithaddr(sin6tosa(&sin6)); 2182 if (ia == NULL || 2183 (opt->ip6po_pktinfo->ipi6_ifindex && 2184 (ia->ifa_ifp->if_index != 2185 opt->ip6po_pktinfo->ipi6_ifindex))) { 2186 return (EADDRNOTAVAIL); 2187 } 2188 ia6 = (struct in6_ifaddr *)ia; 2189 if ((ia6->ia6_flags & (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY)) != 0) { 2190 return (EADDRNOTAVAIL); 2191 } 2192 2193 /* 2194 * Check if the requested source address is 2195 * indeed a unicast address assigned to the 2196 * node. 2197 */ 2198 if (IN6_IS_ADDR_MULTICAST(&opt->ip6po_pktinfo->ipi6_addr)) 2199 return (EADDRNOTAVAIL); 2200 } 2201 break; 2202 2203 case IPV6_HOPLIMIT: 2204 if (cm->cmsg_len != CMSG_LEN(sizeof(int))) 2205 return (EINVAL); 2206 else { 2207 int t; 2208 2209 bcopy(CMSG_DATA(cm), &t, sizeof(t)); 2210 if (t < -1 || t > 255) 2211 return (EINVAL); 2212 opt->ip6po_hlim = t; 2213 } 2214 break; 2215 2216 case IPV6_NEXTHOP: 2217 if (!priv) 2218 return (EPERM); 2219 2220 /* check if cmsg_len is large enough for sa_len */ 2221 if (cm->cmsg_len < sizeof(u_char) || 2222 cm->cmsg_len < CMSG_LEN(*CMSG_DATA(cm))) 2223 return (EINVAL); 2224 2225 opt->ip6po_nexthop = (struct sockaddr *)CMSG_DATA(cm); 2226 2227 break; 2228 2229 case IPV6_HOPOPTS: 2230 if (cm->cmsg_len < CMSG_LEN(sizeof(struct ip6_hbh))) 2231 return (EINVAL); 2232 else { 2233 struct ip6_hbh *t; 2234 2235 t = (struct ip6_hbh *)CMSG_DATA(cm); 2236 if (cm->cmsg_len != 2237 CMSG_LEN((t->ip6h_len + 1) << 3)) 2238 return (EINVAL); 2239 opt->ip6po_hbh = t; 2240 } 2241 break; 2242 2243 case IPV6_DSTOPTS: 2244 if (cm->cmsg_len < CMSG_LEN(sizeof(struct ip6_dest))) 2245 return (EINVAL); 2246 2247 /* 2248 * If there is no routing header yet, the destination 2249 * options header should be put on the 1st part. 2250 * Otherwise, the header should be on the 2nd part. 2251 * (See RFC 2460, section 4.1) 2252 */ 2253 if (opt->ip6po_rthdr == NULL) { 2254 struct ip6_dest *t; 2255 2256 t = (struct ip6_dest *)CMSG_DATA(cm); 2257 if (cm->cmsg_len != 2258 CMSG_LEN((t->ip6d_len + 1) << 3)); 2259 return (EINVAL); 2260 opt->ip6po_dest1 = t; 2261 } 2262 else { 2263 struct ip6_dest *t; 2264 2265 t = (struct ip6_dest *)CMSG_DATA(cm); 2266 if (cm->cmsg_len != 2267 CMSG_LEN((opt->ip6po_dest2->ip6d_len + 1) << 3)) 2268 return (EINVAL); 2269 opt->ip6po_dest2 = t; 2270 } 2271 break; 2272 2273 case IPV6_RTHDR: 2274 if (cm->cmsg_len < CMSG_LEN(sizeof(struct ip6_rthdr))) 2275 return (EINVAL); 2276 else { 2277 struct ip6_rthdr *t; 2278 2279 t = (struct ip6_rthdr *)CMSG_DATA(cm); 2280 if (cm->cmsg_len != 2281 CMSG_LEN((t->ip6r_len + 1) << 3)) 2282 return (EINVAL); 2283 switch (t->ip6r_type) { 2284 case IPV6_RTHDR_TYPE_0: 2285 if (t->ip6r_segleft == 0) 2286 return (EINVAL); 2287 break; 2288 default: 2289 return (EINVAL); 2290 } 2291 opt->ip6po_rthdr = t; 2292 } 2293 break; 2294 2295 default: 2296 return (ENOPROTOOPT); 2297 } 2298 } 2299 2300 return (0); 2301 } 2302 2303 /* 2304 * Routine called from ip6_output() to loop back a copy of an IP6 multicast 2305 * packet to the input queue of a specified interface. Note that this 2306 * calls the output routine of the loopback "driver", but with an interface 2307 * pointer that might NOT be &loif -- easier than replicating that code here. 2308 */ 2309 void 2310 ip6_mloopback(ifp, m, dst) 2311 struct ifnet *ifp; 2312 struct mbuf *m; 2313 struct sockaddr_in6 *dst; 2314 { 2315 struct mbuf *copym; 2316 struct ip6_hdr *ip6; 2317 2318 copym = m_copy(m, 0, M_COPYALL); 2319 if (copym == NULL) 2320 return; 2321 2322 /* 2323 * Make sure to deep-copy IPv6 header portion in case the data 2324 * is in an mbuf cluster, so that we can safely override the IPv6 2325 * header portion later. 2326 */ 2327 if ((copym->m_flags & M_EXT) != 0 || 2328 copym->m_len < sizeof(struct ip6_hdr)) { 2329 copym = m_pullup(copym, sizeof(struct ip6_hdr)); 2330 if (copym == NULL) 2331 return; 2332 } 2333 2334 #ifdef DIAGNOSTIC 2335 if (copym->m_len < sizeof(*ip6)) { 2336 m_freem(copym); 2337 return; 2338 } 2339 #endif 2340 2341 ip6 = mtod(copym, struct ip6_hdr *); 2342 if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src)) 2343 ip6->ip6_src.s6_addr16[1] = 0; 2344 if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst)) 2345 ip6->ip6_dst.s6_addr16[1] = 0; 2346 2347 (void)looutput(ifp, copym, (struct sockaddr *)dst, NULL); 2348 } 2349 2350 /* 2351 * Chop IPv6 header off from the payload. 2352 */ 2353 static int 2354 ip6_splithdr(m, exthdrs) 2355 struct mbuf *m; 2356 struct ip6_exthdrs *exthdrs; 2357 { 2358 struct mbuf *mh; 2359 struct ip6_hdr *ip6; 2360 2361 ip6 = mtod(m, struct ip6_hdr *); 2362 if (m->m_len > sizeof(*ip6)) { 2363 MGETHDR(mh, M_DONTWAIT, MT_HEADER); 2364 if (mh == 0) { 2365 m_freem(m); 2366 return ENOBUFS; 2367 } 2368 M_COPY_PKTHDR(mh, m); 2369 MH_ALIGN(mh, sizeof(*ip6)); 2370 m_tag_delete_chain(m, NULL); 2371 m->m_flags &= ~M_PKTHDR; 2372 m->m_len -= sizeof(*ip6); 2373 m->m_data += sizeof(*ip6); 2374 mh->m_next = m; 2375 m = mh; 2376 m->m_len = sizeof(*ip6); 2377 bcopy((caddr_t)ip6, mtod(m, caddr_t), sizeof(*ip6)); 2378 } 2379 exthdrs->ip6e_ip6 = m; 2380 return 0; 2381 } 2382 2383 /* 2384 * Compute IPv6 extension header length. 2385 */ 2386 int 2387 ip6_optlen(in6p) 2388 struct in6pcb *in6p; 2389 { 2390 int len; 2391 2392 if (!in6p->in6p_outputopts) 2393 return 0; 2394 2395 len = 0; 2396 #define elen(x) \ 2397 (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0) 2398 2399 len += elen(in6p->in6p_outputopts->ip6po_hbh); 2400 len += elen(in6p->in6p_outputopts->ip6po_dest1); 2401 len += elen(in6p->in6p_outputopts->ip6po_rthdr); 2402 len += elen(in6p->in6p_outputopts->ip6po_dest2); 2403 return len; 2404 #undef elen 2405 } 2406