1 /* $NetBSD: ip6_output.c,v 1.87 2004/12/21 11:37:47 drochner 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 1.87 2004/12/21 11:37:47 drochner 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 <net/net_osdep.h> 102 103 #ifdef PFIL_HOOKS 104 extern struct pfil_head inet6_pfil_hook; /* XXX */ 105 #endif 106 107 struct ip6_exthdrs { 108 struct mbuf *ip6e_ip6; 109 struct mbuf *ip6e_hbh; 110 struct mbuf *ip6e_dest1; 111 struct mbuf *ip6e_rthdr; 112 struct mbuf *ip6e_dest2; 113 }; 114 115 static int ip6_pcbopts __P((struct ip6_pktopts **, struct mbuf *, 116 struct socket *)); 117 static int ip6_setmoptions __P((int, struct ip6_moptions **, struct mbuf *)); 118 static int ip6_getmoptions __P((int, struct ip6_moptions *, struct mbuf **)); 119 static int ip6_copyexthdr __P((struct mbuf **, caddr_t, int)); 120 static int ip6_insertfraghdr __P((struct mbuf *, struct mbuf *, int, 121 struct ip6_frag **)); 122 static int ip6_insert_jumboopt __P((struct ip6_exthdrs *, u_int32_t)); 123 static int ip6_splithdr __P((struct mbuf *, struct ip6_exthdrs *)); 124 125 /* 126 * IP6 output. The packet in mbuf chain m contains a skeletal IP6 127 * header (with pri, len, nxt, hlim, src, dst). 128 * This function may modify ver and hlim only. 129 * The mbuf chain containing the packet will be freed. 130 * The mbuf opt, if present, will not be freed. 131 * 132 * type of "mtu": rt_rmx.rmx_mtu is u_long, ifnet.ifr_mtu is int, and 133 * nd_ifinfo.linkmtu is u_int32_t. so we use u_long to hold largest one, 134 * which is rt_rmx.rmx_mtu. 135 */ 136 int 137 ip6_output(m0, opt, ro, flags, im6o, so, ifpp) 138 struct mbuf *m0; 139 struct ip6_pktopts *opt; 140 struct route_in6 *ro; 141 int flags; 142 struct ip6_moptions *im6o; 143 struct socket *so; 144 struct ifnet **ifpp; /* XXX: just for statistics */ 145 { 146 struct ip6_hdr *ip6, *mhip6; 147 struct ifnet *ifp, *origifp; 148 struct mbuf *m = m0; 149 int hlen, tlen, len, off; 150 struct route_in6 ip6route; 151 struct sockaddr_in6 *dst; 152 int error = 0; 153 u_long mtu; 154 int alwaysfrag, dontfrag; 155 u_int32_t optlen = 0, plen = 0, unfragpartlen = 0; 156 struct ip6_exthdrs exthdrs; 157 struct in6_addr finaldst; 158 struct route_in6 *ro_pmtu = NULL; 159 int hdrsplit = 0; 160 int needipsec = 0; 161 #ifdef IPSEC 162 int needipsectun = 0; 163 struct secpolicy *sp = NULL; 164 165 ip6 = mtod(m, struct ip6_hdr *); 166 #endif /* IPSEC */ 167 168 #define MAKE_EXTHDR(hp, mp) \ 169 do { \ 170 if (hp) { \ 171 struct ip6_ext *eh = (struct ip6_ext *)(hp); \ 172 error = ip6_copyexthdr((mp), (caddr_t)(hp), \ 173 ((eh)->ip6e_len + 1) << 3); \ 174 if (error) \ 175 goto freehdrs; \ 176 } \ 177 } while (/*CONSTCOND*/ 0) 178 179 bzero(&exthdrs, sizeof(exthdrs)); 180 if (opt) { 181 /* Hop-by-Hop options header */ 182 MAKE_EXTHDR(opt->ip6po_hbh, &exthdrs.ip6e_hbh); 183 /* Destination options header(1st part) */ 184 MAKE_EXTHDR(opt->ip6po_dest1, &exthdrs.ip6e_dest1); 185 /* Routing header */ 186 MAKE_EXTHDR(opt->ip6po_rthdr, &exthdrs.ip6e_rthdr); 187 /* Destination options header(2nd part) */ 188 MAKE_EXTHDR(opt->ip6po_dest2, &exthdrs.ip6e_dest2); 189 } 190 191 #ifdef IPSEC 192 if ((flags & IPV6_FORWARDING) != 0) { 193 needipsec = 0; 194 goto skippolicycheck; 195 } 196 197 /* get a security policy for this packet */ 198 if (so == NULL) 199 sp = ipsec6_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND, 0, &error); 200 else { 201 if (IPSEC_PCB_SKIP_IPSEC(sotoinpcb_hdr(so)->inph_sp, 202 IPSEC_DIR_OUTBOUND)) { 203 needipsec = 0; 204 goto skippolicycheck; 205 } 206 sp = ipsec6_getpolicybysock(m, IPSEC_DIR_OUTBOUND, so, &error); 207 } 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_pmtu = 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 = lo0ifp; 625 } 626 627 if (opt && opt->ip6po_hlim != -1) 628 ip6->ip6_hlim = opt->ip6po_hlim & 0xff; 629 630 /* 631 * If caller did not provide an interface lookup a 632 * default in the routing table. This is either a 633 * default for the speicfied group (i.e. a host 634 * route), or a multicast default (a route for the 635 * ``net'' ff00::/8). 636 */ 637 if (ifp == NULL) { 638 if (ro->ro_rt == 0) { 639 ro->ro_rt = rtalloc1((struct sockaddr *) 640 &ro->ro_dst, 0); 641 } 642 if (ro->ro_rt == 0) { 643 ip6stat.ip6s_noroute++; 644 error = EHOSTUNREACH; 645 /* XXX in6_ifstat_inc(ifp, ifs6_out_discard) */ 646 goto bad; 647 } 648 ifp = ro->ro_rt->rt_ifp; 649 ro->ro_rt->rt_use++; 650 } 651 652 if ((flags & IPV6_FORWARDING) == 0) 653 in6_ifstat_inc(ifp, ifs6_out_request); 654 in6_ifstat_inc(ifp, ifs6_out_mcast); 655 656 /* 657 * Confirm that the outgoing interface supports multicast. 658 */ 659 if ((ifp->if_flags & IFF_MULTICAST) == 0) { 660 ip6stat.ip6s_noroute++; 661 in6_ifstat_inc(ifp, ifs6_out_discard); 662 error = ENETUNREACH; 663 goto bad; 664 } 665 IN6_LOOKUP_MULTI(ip6->ip6_dst, ifp, in6m); 666 if (in6m != NULL && 667 (im6o == NULL || im6o->im6o_multicast_loop)) { 668 /* 669 * If we belong to the destination multicast group 670 * on the outgoing interface, and the caller did not 671 * forbid loopback, loop back a copy. 672 */ 673 ip6_mloopback(ifp, m, dst); 674 } else { 675 /* 676 * If we are acting as a multicast router, perform 677 * multicast forwarding as if the packet had just 678 * arrived on the interface to which we are about 679 * to send. The multicast forwarding function 680 * recursively calls this function, using the 681 * IPV6_FORWARDING flag to prevent infinite recursion. 682 * 683 * Multicasts that are looped back by ip6_mloopback(), 684 * above, will be forwarded by the ip6_input() routine, 685 * if necessary. 686 */ 687 if (ip6_mrouter && (flags & IPV6_FORWARDING) == 0) { 688 if (ip6_mforward(ip6, ifp, m) != 0) { 689 m_freem(m); 690 goto done; 691 } 692 } 693 } 694 /* 695 * Multicasts with a hoplimit of zero may be looped back, 696 * above, but must not be transmitted on a network. 697 * Also, multicasts addressed to the loopback interface 698 * are not sent -- the above call to ip6_mloopback() will 699 * loop back a copy if this host actually belongs to the 700 * destination group on the loopback interface. 701 */ 702 if (ip6->ip6_hlim == 0 || (ifp->if_flags & IFF_LOOPBACK)) { 703 m_freem(m); 704 goto done; 705 } 706 } 707 708 /* 709 * Fill the outgoing inteface to tell the upper layer 710 * to increment per-interface statistics. 711 */ 712 if (ifpp) 713 *ifpp = ifp; 714 715 /* Determine path MTU. */ 716 if ((error = ip6_getpmtu(ro_pmtu, ro, ifp, &finaldst, &mtu, 717 &alwaysfrag)) != 0) 718 goto bad; 719 #ifdef IPSEC 720 if (needipsectun) 721 mtu = IPV6_MMTU; 722 #endif 723 724 /* 725 * The caller of this function may specify to use the minimum MTU 726 * in some cases. 727 */ 728 if (mtu > IPV6_MMTU) { 729 if ((flags & IPV6_MINMTU)) 730 mtu = IPV6_MMTU; 731 } 732 733 /* Fake scoped addresses */ 734 if ((ifp->if_flags & IFF_LOOPBACK) != 0) { 735 /* 736 * If source or destination address is a scoped address, and 737 * the packet is going to be sent to a loopback interface, 738 * we should keep the original interface. 739 */ 740 741 /* 742 * XXX: this is a very experimental and temporary solution. 743 * We eventually have sockaddr_in6 and use the sin6_scope_id 744 * field of the structure here. 745 * We rely on the consistency between two scope zone ids 746 * of source add destination, which should already be assured 747 * Larger scopes than link will be supported in the near 748 * future. 749 */ 750 origifp = NULL; 751 if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src)) 752 origifp = ifindex2ifnet[ntohs(ip6->ip6_src.s6_addr16[1])]; 753 else if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst)) 754 origifp = ifindex2ifnet[ntohs(ip6->ip6_dst.s6_addr16[1])]; 755 /* 756 * XXX: origifp can be NULL even in those two cases above. 757 * For example, if we remove the (only) link-local address 758 * from the loopback interface, and try to send a link-local 759 * address without link-id information. Then the source 760 * address is ::1, and the destination address is the 761 * link-local address with its s6_addr16[1] being zero. 762 * What is worse, if the packet goes to the loopback interface 763 * by a default rejected route, the null pointer would be 764 * passed to looutput, and the kernel would hang. 765 * The following last resort would prevent such disaster. 766 */ 767 if (origifp == NULL) 768 origifp = ifp; 769 } else 770 origifp = ifp; 771 if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src)) 772 ip6->ip6_src.s6_addr16[1] = 0; 773 if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst)) 774 ip6->ip6_dst.s6_addr16[1] = 0; 775 776 /* 777 * If the outgoing packet contains a hop-by-hop options header, 778 * it must be examined and processed even by the source node. 779 * (RFC 2460, section 4.) 780 */ 781 if (exthdrs.ip6e_hbh) { 782 struct ip6_hbh *hbh = mtod(exthdrs.ip6e_hbh, struct ip6_hbh *); 783 u_int32_t dummy1; /* XXX unused */ 784 u_int32_t dummy2; /* XXX unused */ 785 786 /* 787 * XXX: if we have to send an ICMPv6 error to the sender, 788 * we need the M_LOOP flag since icmp6_error() expects 789 * the IPv6 and the hop-by-hop options header are 790 * continuous unless the flag is set. 791 */ 792 m->m_flags |= M_LOOP; 793 m->m_pkthdr.rcvif = ifp; 794 if (ip6_process_hopopts(m, (u_int8_t *)(hbh + 1), 795 ((hbh->ip6h_len + 1) << 3) - sizeof(struct ip6_hbh), 796 &dummy1, &dummy2) < 0) { 797 /* m was already freed at this point */ 798 error = EINVAL;/* better error? */ 799 goto done; 800 } 801 m->m_flags &= ~M_LOOP; /* XXX */ 802 m->m_pkthdr.rcvif = NULL; 803 } 804 805 #ifdef PFIL_HOOKS 806 /* 807 * Run through list of hooks for output packets. 808 */ 809 if ((error = pfil_run_hooks(&inet6_pfil_hook, &m, ifp, PFIL_OUT)) != 0) 810 goto done; 811 if (m == NULL) 812 goto done; 813 ip6 = mtod(m, struct ip6_hdr *); 814 #endif /* PFIL_HOOKS */ 815 /* 816 * Send the packet to the outgoing interface. 817 * If necessary, do IPv6 fragmentation before sending. 818 * 819 * the logic here is rather complex: 820 * 1: normal case (dontfrag == 0, alwaysfrag == 0) 821 * 1-a: send as is if tlen <= path mtu 822 * 1-b: fragment if tlen > path mtu 823 * 824 * 2: if user asks us not to fragment (dontfrag == 1) 825 * 2-a: send as is if tlen <= interface mtu 826 * 2-b: error if tlen > interface mtu 827 * 828 * 3: if we always need to attach fragment header (alwaysfrag == 1) 829 * always fragment 830 * 831 * 4: if dontfrag == 1 && alwaysfrag == 1 832 * error, as we cannot handle this conflicting request 833 */ 834 tlen = m->m_pkthdr.len; 835 836 dontfrag = 0; 837 if (dontfrag && alwaysfrag) { /* case 4 */ 838 /* conflicting request - can't transmit */ 839 error = EMSGSIZE; 840 goto bad; 841 } 842 if (dontfrag && tlen > IN6_LINKMTU(ifp)) { /* case 2-b */ 843 /* 844 * Even if the DONTFRAG option is specified, we cannot send the 845 * packet when the data length is larger than the MTU of the 846 * outgoing interface. 847 * Notify the error by sending IPV6_PATHMTU ancillary data as 848 * well as returning an error code (the latter is not described 849 * in the API spec.) 850 */ 851 u_int32_t mtu32; 852 struct ip6ctlparam ip6cp; 853 854 mtu32 = (u_int32_t)mtu; 855 bzero(&ip6cp, sizeof(ip6cp)); 856 ip6cp.ip6c_cmdarg = (void *)&mtu32; 857 pfctlinput2(PRC_MSGSIZE, (struct sockaddr *)&ro_pmtu->ro_dst, 858 (void *)&ip6cp); 859 860 error = EMSGSIZE; 861 goto bad; 862 } 863 864 /* 865 * transmit packet without fragmentation 866 */ 867 if (dontfrag || (!alwaysfrag && tlen <= mtu)) { /* case 1-a and 2-a */ 868 struct in6_ifaddr *ia6; 869 870 ip6 = mtod(m, struct ip6_hdr *); 871 ia6 = in6_ifawithifp(ifp, &ip6->ip6_src); 872 if (ia6) { 873 /* Record statistics for this interface address. */ 874 ia6->ia_ifa.ifa_data.ifad_outbytes += m->m_pkthdr.len; 875 } 876 #ifdef IPSEC 877 /* clean ipsec history once it goes out of the node */ 878 ipsec_delaux(m); 879 #endif 880 error = nd6_output(ifp, origifp, m, dst, ro->ro_rt); 881 goto done; 882 } 883 884 /* 885 * try to fragment the packet. case 1-b and 3 886 */ 887 if (mtu < IPV6_MMTU) { 888 /* path MTU cannot be less than IPV6_MMTU */ 889 error = EMSGSIZE; 890 in6_ifstat_inc(ifp, ifs6_out_fragfail); 891 goto bad; 892 } else if (ip6->ip6_plen == 0) { 893 /* jumbo payload cannot be fragmented */ 894 error = EMSGSIZE; 895 in6_ifstat_inc(ifp, ifs6_out_fragfail); 896 goto bad; 897 } else { 898 struct mbuf **mnext, *m_frgpart; 899 struct ip6_frag *ip6f; 900 u_int32_t id = htonl(ip6_randomid()); 901 u_char nextproto; 902 struct ip6ctlparam ip6cp; 903 u_int32_t mtu32; 904 905 /* 906 * Too large for the destination or interface; 907 * fragment if possible. 908 * Must be able to put at least 8 bytes per fragment. 909 */ 910 hlen = unfragpartlen; 911 if (mtu > IPV6_MAXPACKET) 912 mtu = IPV6_MAXPACKET; 913 914 /* Notify a proper path MTU to applications. */ 915 mtu32 = (u_int32_t)mtu; 916 bzero(&ip6cp, sizeof(ip6cp)); 917 ip6cp.ip6c_cmdarg = (void *)&mtu32; 918 pfctlinput2(PRC_MSGSIZE, (struct sockaddr *)&ro_pmtu->ro_dst, 919 (void *)&ip6cp); 920 921 len = (mtu - hlen - sizeof(struct ip6_frag)) & ~7; 922 if (len < 8) { 923 error = EMSGSIZE; 924 in6_ifstat_inc(ifp, ifs6_out_fragfail); 925 goto bad; 926 } 927 928 mnext = &m->m_nextpkt; 929 930 /* 931 * Change the next header field of the last header in the 932 * unfragmentable part. 933 */ 934 if (exthdrs.ip6e_rthdr) { 935 nextproto = *mtod(exthdrs.ip6e_rthdr, u_char *); 936 *mtod(exthdrs.ip6e_rthdr, u_char *) = IPPROTO_FRAGMENT; 937 } else if (exthdrs.ip6e_dest1) { 938 nextproto = *mtod(exthdrs.ip6e_dest1, u_char *); 939 *mtod(exthdrs.ip6e_dest1, u_char *) = IPPROTO_FRAGMENT; 940 } else if (exthdrs.ip6e_hbh) { 941 nextproto = *mtod(exthdrs.ip6e_hbh, u_char *); 942 *mtod(exthdrs.ip6e_hbh, u_char *) = IPPROTO_FRAGMENT; 943 } else { 944 nextproto = ip6->ip6_nxt; 945 ip6->ip6_nxt = IPPROTO_FRAGMENT; 946 } 947 948 /* 949 * Loop through length of segment after first fragment, 950 * make new header and copy data of each part and link onto 951 * chain. 952 */ 953 m0 = m; 954 for (off = hlen; off < tlen; off += len) { 955 struct mbuf *mlast; 956 957 MGETHDR(m, M_DONTWAIT, MT_HEADER); 958 if (!m) { 959 error = ENOBUFS; 960 ip6stat.ip6s_odropped++; 961 goto sendorfree; 962 } 963 m->m_pkthdr.rcvif = NULL; 964 m->m_flags = m0->m_flags & M_COPYFLAGS; 965 *mnext = m; 966 mnext = &m->m_nextpkt; 967 m->m_data += max_linkhdr; 968 mhip6 = mtod(m, struct ip6_hdr *); 969 *mhip6 = *ip6; 970 m->m_len = sizeof(*mhip6); 971 error = ip6_insertfraghdr(m0, m, hlen, &ip6f); 972 if (error) { 973 ip6stat.ip6s_odropped++; 974 goto sendorfree; 975 } 976 ip6f->ip6f_offlg = htons((u_int16_t)((off - hlen) & ~7)); 977 if (off + len >= tlen) 978 len = tlen - off; 979 else 980 ip6f->ip6f_offlg |= IP6F_MORE_FRAG; 981 mhip6->ip6_plen = htons((u_int16_t)(len + hlen + 982 sizeof(*ip6f) - sizeof(struct ip6_hdr))); 983 if ((m_frgpart = m_copy(m0, off, len)) == 0) { 984 error = ENOBUFS; 985 ip6stat.ip6s_odropped++; 986 goto sendorfree; 987 } 988 for (mlast = m; mlast->m_next; mlast = mlast->m_next) 989 ; 990 mlast->m_next = m_frgpart; 991 m->m_pkthdr.len = len + hlen + sizeof(*ip6f); 992 m->m_pkthdr.rcvif = (struct ifnet *)0; 993 ip6f->ip6f_reserved = 0; 994 ip6f->ip6f_ident = id; 995 ip6f->ip6f_nxt = nextproto; 996 ip6stat.ip6s_ofragments++; 997 in6_ifstat_inc(ifp, ifs6_out_fragcreat); 998 } 999 1000 in6_ifstat_inc(ifp, ifs6_out_fragok); 1001 } 1002 1003 /* 1004 * Remove leading garbages. 1005 */ 1006 sendorfree: 1007 m = m0->m_nextpkt; 1008 m0->m_nextpkt = 0; 1009 m_freem(m0); 1010 for (m0 = m; m; m = m0) { 1011 m0 = m->m_nextpkt; 1012 m->m_nextpkt = 0; 1013 if (error == 0) { 1014 struct in6_ifaddr *ia6; 1015 ip6 = mtod(m, struct ip6_hdr *); 1016 ia6 = in6_ifawithifp(ifp, &ip6->ip6_src); 1017 if (ia6) { 1018 /* 1019 * Record statistics for this interface 1020 * address. 1021 */ 1022 ia6->ia_ifa.ifa_data.ifad_outbytes += 1023 m->m_pkthdr.len; 1024 } 1025 #ifdef IPSEC 1026 /* clean ipsec history once it goes out of the node */ 1027 ipsec_delaux(m); 1028 #endif 1029 error = nd6_output(ifp, origifp, m, dst, ro->ro_rt); 1030 } else 1031 m_freem(m); 1032 } 1033 1034 if (error == 0) 1035 ip6stat.ip6s_fragmented++; 1036 1037 done: 1038 if (ro == &ip6route && ro->ro_rt) { /* brace necessary for RTFREE */ 1039 RTFREE(ro->ro_rt); 1040 } else if (ro_pmtu == &ip6route && ro_pmtu->ro_rt) { 1041 RTFREE(ro_pmtu->ro_rt); 1042 } 1043 1044 #ifdef IPSEC 1045 if (sp != NULL) 1046 key_freesp(sp); 1047 #endif /* IPSEC */ 1048 1049 return (error); 1050 1051 freehdrs: 1052 m_freem(exthdrs.ip6e_hbh); /* m_freem will check if mbuf is 0 */ 1053 m_freem(exthdrs.ip6e_dest1); 1054 m_freem(exthdrs.ip6e_rthdr); 1055 m_freem(exthdrs.ip6e_dest2); 1056 /* FALLTHROUGH */ 1057 bad: 1058 m_freem(m); 1059 goto done; 1060 } 1061 1062 static int 1063 ip6_copyexthdr(mp, hdr, hlen) 1064 struct mbuf **mp; 1065 caddr_t hdr; 1066 int hlen; 1067 { 1068 struct mbuf *m; 1069 1070 if (hlen > MCLBYTES) 1071 return (ENOBUFS); /* XXX */ 1072 1073 MGET(m, M_DONTWAIT, MT_DATA); 1074 if (!m) 1075 return (ENOBUFS); 1076 1077 if (hlen > MLEN) { 1078 MCLGET(m, M_DONTWAIT); 1079 if ((m->m_flags & M_EXT) == 0) { 1080 m_free(m); 1081 return (ENOBUFS); 1082 } 1083 } 1084 m->m_len = hlen; 1085 if (hdr) 1086 bcopy(hdr, mtod(m, caddr_t), hlen); 1087 1088 *mp = m; 1089 return (0); 1090 } 1091 1092 /* 1093 * Insert jumbo payload option. 1094 */ 1095 static int 1096 ip6_insert_jumboopt(exthdrs, plen) 1097 struct ip6_exthdrs *exthdrs; 1098 u_int32_t plen; 1099 { 1100 struct mbuf *mopt; 1101 u_int8_t *optbuf; 1102 u_int32_t v; 1103 1104 #define JUMBOOPTLEN 8 /* length of jumbo payload option and padding */ 1105 1106 /* 1107 * If there is no hop-by-hop options header, allocate new one. 1108 * If there is one but it doesn't have enough space to store the 1109 * jumbo payload option, allocate a cluster to store the whole options. 1110 * Otherwise, use it to store the options. 1111 */ 1112 if (exthdrs->ip6e_hbh == 0) { 1113 MGET(mopt, M_DONTWAIT, MT_DATA); 1114 if (mopt == 0) 1115 return (ENOBUFS); 1116 mopt->m_len = JUMBOOPTLEN; 1117 optbuf = mtod(mopt, u_int8_t *); 1118 optbuf[1] = 0; /* = ((JUMBOOPTLEN) >> 3) - 1 */ 1119 exthdrs->ip6e_hbh = mopt; 1120 } else { 1121 struct ip6_hbh *hbh; 1122 1123 mopt = exthdrs->ip6e_hbh; 1124 if (M_TRAILINGSPACE(mopt) < JUMBOOPTLEN) { 1125 /* 1126 * XXX assumption: 1127 * - exthdrs->ip6e_hbh is not referenced from places 1128 * other than exthdrs. 1129 * - exthdrs->ip6e_hbh is not an mbuf chain. 1130 */ 1131 int oldoptlen = mopt->m_len; 1132 struct mbuf *n; 1133 1134 /* 1135 * XXX: give up if the whole (new) hbh header does 1136 * not fit even in an mbuf cluster. 1137 */ 1138 if (oldoptlen + JUMBOOPTLEN > MCLBYTES) 1139 return (ENOBUFS); 1140 1141 /* 1142 * As a consequence, we must always prepare a cluster 1143 * at this point. 1144 */ 1145 MGET(n, M_DONTWAIT, MT_DATA); 1146 if (n) { 1147 MCLGET(n, M_DONTWAIT); 1148 if ((n->m_flags & M_EXT) == 0) { 1149 m_freem(n); 1150 n = NULL; 1151 } 1152 } 1153 if (!n) 1154 return (ENOBUFS); 1155 n->m_len = oldoptlen + JUMBOOPTLEN; 1156 bcopy(mtod(mopt, caddr_t), mtod(n, caddr_t), 1157 oldoptlen); 1158 optbuf = mtod(n, u_int8_t *) + oldoptlen; 1159 m_freem(mopt); 1160 mopt = exthdrs->ip6e_hbh = n; 1161 } else { 1162 optbuf = mtod(mopt, u_int8_t *) + mopt->m_len; 1163 mopt->m_len += JUMBOOPTLEN; 1164 } 1165 optbuf[0] = IP6OPT_PADN; 1166 optbuf[1] = 0; 1167 1168 /* 1169 * Adjust the header length according to the pad and 1170 * the jumbo payload option. 1171 */ 1172 hbh = mtod(mopt, struct ip6_hbh *); 1173 hbh->ip6h_len += (JUMBOOPTLEN >> 3); 1174 } 1175 1176 /* fill in the option. */ 1177 optbuf[2] = IP6OPT_JUMBO; 1178 optbuf[3] = 4; 1179 v = (u_int32_t)htonl(plen + JUMBOOPTLEN); 1180 bcopy(&v, &optbuf[4], sizeof(u_int32_t)); 1181 1182 /* finally, adjust the packet header length */ 1183 exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN; 1184 1185 return (0); 1186 #undef JUMBOOPTLEN 1187 } 1188 1189 /* 1190 * Insert fragment header and copy unfragmentable header portions. 1191 */ 1192 static int 1193 ip6_insertfraghdr(m0, m, hlen, frghdrp) 1194 struct mbuf *m0, *m; 1195 int hlen; 1196 struct ip6_frag **frghdrp; 1197 { 1198 struct mbuf *n, *mlast; 1199 1200 if (hlen > sizeof(struct ip6_hdr)) { 1201 n = m_copym(m0, sizeof(struct ip6_hdr), 1202 hlen - sizeof(struct ip6_hdr), M_DONTWAIT); 1203 if (n == 0) 1204 return (ENOBUFS); 1205 m->m_next = n; 1206 } else 1207 n = m; 1208 1209 /* Search for the last mbuf of unfragmentable part. */ 1210 for (mlast = n; mlast->m_next; mlast = mlast->m_next) 1211 ; 1212 1213 if ((mlast->m_flags & M_EXT) == 0 && 1214 M_TRAILINGSPACE(mlast) >= sizeof(struct ip6_frag)) { 1215 /* use the trailing space of the last mbuf for the fragment hdr */ 1216 *frghdrp = (struct ip6_frag *)(mtod(mlast, caddr_t) + 1217 mlast->m_len); 1218 mlast->m_len += sizeof(struct ip6_frag); 1219 m->m_pkthdr.len += sizeof(struct ip6_frag); 1220 } else { 1221 /* allocate a new mbuf for the fragment header */ 1222 struct mbuf *mfrg; 1223 1224 MGET(mfrg, M_DONTWAIT, MT_DATA); 1225 if (mfrg == 0) 1226 return (ENOBUFS); 1227 mfrg->m_len = sizeof(struct ip6_frag); 1228 *frghdrp = mtod(mfrg, struct ip6_frag *); 1229 mlast->m_next = mfrg; 1230 } 1231 1232 return (0); 1233 } 1234 1235 int 1236 ip6_getpmtu(ro_pmtu, ro, ifp, dst, mtup, alwaysfragp) 1237 struct route_in6 *ro_pmtu, *ro; 1238 struct ifnet *ifp; 1239 struct in6_addr *dst; 1240 u_long *mtup; 1241 int *alwaysfragp; 1242 { 1243 u_int32_t mtu = 0; 1244 int alwaysfrag = 0; 1245 int error = 0; 1246 1247 if (ro_pmtu != ro) { 1248 /* The first hop and the final destination may differ. */ 1249 struct sockaddr_in6 *sa6_dst = 1250 (struct sockaddr_in6 *)&ro_pmtu->ro_dst; 1251 if (ro_pmtu->ro_rt && 1252 ((ro_pmtu->ro_rt->rt_flags & RTF_UP) == 0 || 1253 !IN6_ARE_ADDR_EQUAL(&sa6_dst->sin6_addr, dst))) { 1254 RTFREE(ro_pmtu->ro_rt); 1255 ro_pmtu->ro_rt = (struct rtentry *)NULL; 1256 } 1257 if (ro_pmtu->ro_rt == NULL) { 1258 bzero(sa6_dst, sizeof(*sa6_dst)); /* for safety */ 1259 sa6_dst->sin6_family = AF_INET6; 1260 sa6_dst->sin6_len = sizeof(struct sockaddr_in6); 1261 sa6_dst->sin6_addr = *dst; 1262 1263 rtalloc((struct route *)ro_pmtu); 1264 } 1265 } 1266 if (ro_pmtu->ro_rt) { 1267 u_int32_t ifmtu; 1268 1269 if (ifp == NULL) 1270 ifp = ro_pmtu->ro_rt->rt_ifp; 1271 ifmtu = IN6_LINKMTU(ifp); 1272 mtu = ro_pmtu->ro_rt->rt_rmx.rmx_mtu; 1273 if (mtu == 0) 1274 mtu = ifmtu; 1275 else if (mtu < IPV6_MMTU) { 1276 /* 1277 * RFC2460 section 5, last paragraph: 1278 * if we record ICMPv6 too big message with 1279 * mtu < IPV6_MMTU, transmit packets sized IPV6_MMTU 1280 * or smaller, with fragment header attached. 1281 * (fragment header is needed regardless from the 1282 * packet size, for translators to identify packets) 1283 */ 1284 alwaysfrag = 1; 1285 mtu = IPV6_MMTU; 1286 } else if (mtu > ifmtu) { 1287 /* 1288 * The MTU on the route is larger than the MTU on 1289 * the interface! This shouldn't happen, unless the 1290 * MTU of the interface has been changed after the 1291 * interface was brought up. Change the MTU in the 1292 * route to match the interface MTU (as long as the 1293 * field isn't locked). 1294 */ 1295 mtu = ifmtu; 1296 if (!(ro_pmtu->ro_rt->rt_rmx.rmx_locks & RTV_MTU)) 1297 ro_pmtu->ro_rt->rt_rmx.rmx_mtu = mtu; 1298 } 1299 } else if (ifp) { 1300 mtu = IN6_LINKMTU(ifp); 1301 } else 1302 error = EHOSTUNREACH; /* XXX */ 1303 1304 *mtup = mtu; 1305 if (alwaysfragp) 1306 *alwaysfragp = alwaysfrag; 1307 return (error); 1308 } 1309 1310 /* 1311 * IP6 socket option processing. 1312 */ 1313 int 1314 ip6_ctloutput(op, so, level, optname, mp) 1315 int op; 1316 struct socket *so; 1317 int level, optname; 1318 struct mbuf **mp; 1319 { 1320 struct in6pcb *in6p = sotoin6pcb(so); 1321 struct mbuf *m = *mp; 1322 int optval = 0; 1323 int error = 0; 1324 struct proc *p = curproc; /* XXX */ 1325 1326 if (level == IPPROTO_IPV6) { 1327 switch (op) { 1328 case PRCO_SETOPT: 1329 switch (optname) { 1330 case IPV6_PKTOPTIONS: 1331 /* m is freed in ip6_pcbopts */ 1332 return (ip6_pcbopts(&in6p->in6p_outputopts, 1333 m, so)); 1334 case IPV6_HOPOPTS: 1335 case IPV6_DSTOPTS: 1336 if (p == 0 || suser(p->p_ucred, &p->p_acflag)) { 1337 error = EPERM; 1338 break; 1339 } 1340 /* FALLTHROUGH */ 1341 case IPV6_UNICAST_HOPS: 1342 case IPV6_RECVOPTS: 1343 case IPV6_RECVRETOPTS: 1344 case IPV6_RECVDSTADDR: 1345 case IPV6_PKTINFO: 1346 case IPV6_HOPLIMIT: 1347 case IPV6_RTHDR: 1348 case IPV6_FAITH: 1349 case IPV6_V6ONLY: 1350 case IPV6_USE_MIN_MTU: 1351 if (!m || m->m_len != sizeof(int)) { 1352 error = EINVAL; 1353 break; 1354 } 1355 optval = *mtod(m, int *); 1356 switch (optname) { 1357 1358 case IPV6_UNICAST_HOPS: 1359 if (optval < -1 || optval >= 256) 1360 error = EINVAL; 1361 else { 1362 /* -1 = kernel default */ 1363 in6p->in6p_hops = optval; 1364 } 1365 break; 1366 #define OPTSET(bit) \ 1367 do { \ 1368 if (optval) \ 1369 in6p->in6p_flags |= (bit); \ 1370 else \ 1371 in6p->in6p_flags &= ~(bit); \ 1372 } while (/*CONSTCOND*/ 0) 1373 1374 case IPV6_RECVOPTS: 1375 OPTSET(IN6P_RECVOPTS); 1376 break; 1377 1378 case IPV6_RECVRETOPTS: 1379 OPTSET(IN6P_RECVRETOPTS); 1380 break; 1381 1382 case IPV6_RECVDSTADDR: 1383 OPTSET(IN6P_RECVDSTADDR); 1384 break; 1385 1386 case IPV6_PKTINFO: 1387 OPTSET(IN6P_PKTINFO); 1388 break; 1389 1390 case IPV6_HOPLIMIT: 1391 OPTSET(IN6P_HOPLIMIT); 1392 break; 1393 1394 case IPV6_HOPOPTS: 1395 OPTSET(IN6P_HOPOPTS); 1396 break; 1397 1398 case IPV6_DSTOPTS: 1399 OPTSET(IN6P_DSTOPTS); 1400 break; 1401 1402 case IPV6_RTHDR: 1403 OPTSET(IN6P_RTHDR); 1404 break; 1405 1406 case IPV6_FAITH: 1407 OPTSET(IN6P_FAITH); 1408 break; 1409 1410 case IPV6_USE_MIN_MTU: 1411 OPTSET(IN6P_MINMTU); 1412 break; 1413 1414 case IPV6_V6ONLY: 1415 /* 1416 * make setsockopt(IPV6_V6ONLY) 1417 * available only prior to bind(2). 1418 * see ipng mailing list, Jun 22 2001. 1419 */ 1420 if (in6p->in6p_lport || 1421 !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr)) { 1422 error = EINVAL; 1423 break; 1424 } 1425 #ifdef INET6_BINDV6ONLY 1426 if (!optval) 1427 error = EINVAL; 1428 #else 1429 OPTSET(IN6P_IPV6_V6ONLY); 1430 #endif 1431 break; 1432 } 1433 break; 1434 #undef OPTSET 1435 1436 case IPV6_MULTICAST_IF: 1437 case IPV6_MULTICAST_HOPS: 1438 case IPV6_MULTICAST_LOOP: 1439 case IPV6_JOIN_GROUP: 1440 case IPV6_LEAVE_GROUP: 1441 error = ip6_setmoptions(optname, 1442 &in6p->in6p_moptions, m); 1443 break; 1444 1445 case IPV6_PORTRANGE: 1446 optval = *mtod(m, int *); 1447 1448 switch (optval) { 1449 case IPV6_PORTRANGE_DEFAULT: 1450 in6p->in6p_flags &= ~(IN6P_LOWPORT); 1451 in6p->in6p_flags &= ~(IN6P_HIGHPORT); 1452 break; 1453 1454 case IPV6_PORTRANGE_HIGH: 1455 in6p->in6p_flags &= ~(IN6P_LOWPORT); 1456 in6p->in6p_flags |= IN6P_HIGHPORT; 1457 break; 1458 1459 case IPV6_PORTRANGE_LOW: 1460 in6p->in6p_flags &= ~(IN6P_HIGHPORT); 1461 in6p->in6p_flags |= IN6P_LOWPORT; 1462 break; 1463 1464 default: 1465 error = EINVAL; 1466 break; 1467 } 1468 break; 1469 1470 #ifdef IPSEC 1471 case IPV6_IPSEC_POLICY: 1472 { 1473 caddr_t req = NULL; 1474 size_t len = 0; 1475 1476 int priv = 0; 1477 if (p == 0 || suser(p->p_ucred, &p->p_acflag)) 1478 priv = 0; 1479 else 1480 priv = 1; 1481 if (m) { 1482 req = mtod(m, caddr_t); 1483 len = m->m_len; 1484 } 1485 error = ipsec6_set_policy(in6p, 1486 optname, req, len, priv); 1487 } 1488 break; 1489 #endif /* IPSEC */ 1490 1491 default: 1492 error = ENOPROTOOPT; 1493 break; 1494 } 1495 if (m) 1496 (void)m_free(m); 1497 break; 1498 1499 case PRCO_GETOPT: 1500 switch (optname) { 1501 1502 case IPV6_OPTIONS: 1503 case IPV6_RETOPTS: 1504 error = ENOPROTOOPT; 1505 break; 1506 1507 case IPV6_PKTOPTIONS: 1508 if (in6p->in6p_options) { 1509 *mp = m_copym(in6p->in6p_options, 0, 1510 M_COPYALL, M_WAIT); 1511 } else { 1512 *mp = m_get(M_WAIT, MT_SOOPTS); 1513 (*mp)->m_len = 0; 1514 } 1515 break; 1516 1517 case IPV6_HOPOPTS: 1518 case IPV6_DSTOPTS: 1519 if (p == 0 || suser(p->p_ucred, &p->p_acflag)) { 1520 error = EPERM; 1521 break; 1522 } 1523 /* FALLTHROUGH */ 1524 case IPV6_UNICAST_HOPS: 1525 case IPV6_RECVOPTS: 1526 case IPV6_RECVRETOPTS: 1527 case IPV6_RECVDSTADDR: 1528 case IPV6_PORTRANGE: 1529 case IPV6_PKTINFO: 1530 case IPV6_HOPLIMIT: 1531 case IPV6_RTHDR: 1532 case IPV6_FAITH: 1533 case IPV6_V6ONLY: 1534 case IPV6_USE_MIN_MTU: 1535 *mp = m = m_get(M_WAIT, MT_SOOPTS); 1536 m->m_len = sizeof(int); 1537 switch (optname) { 1538 1539 case IPV6_UNICAST_HOPS: 1540 optval = in6p->in6p_hops; 1541 break; 1542 1543 #define OPTBIT(bit) (in6p->in6p_flags & bit ? 1 : 0) 1544 1545 case IPV6_RECVOPTS: 1546 optval = OPTBIT(IN6P_RECVOPTS); 1547 break; 1548 1549 case IPV6_RECVRETOPTS: 1550 optval = OPTBIT(IN6P_RECVRETOPTS); 1551 break; 1552 1553 case IPV6_RECVDSTADDR: 1554 optval = OPTBIT(IN6P_RECVDSTADDR); 1555 break; 1556 1557 case IPV6_PORTRANGE: 1558 { 1559 int flags; 1560 flags = in6p->in6p_flags; 1561 if (flags & IN6P_HIGHPORT) 1562 optval = IPV6_PORTRANGE_HIGH; 1563 else if (flags & IN6P_LOWPORT) 1564 optval = IPV6_PORTRANGE_LOW; 1565 else 1566 optval = 0; 1567 break; 1568 } 1569 1570 case IPV6_PKTINFO: 1571 optval = OPTBIT(IN6P_PKTINFO); 1572 break; 1573 1574 case IPV6_HOPLIMIT: 1575 optval = OPTBIT(IN6P_HOPLIMIT); 1576 break; 1577 1578 case IPV6_HOPOPTS: 1579 optval = OPTBIT(IN6P_HOPOPTS); 1580 break; 1581 1582 case IPV6_DSTOPTS: 1583 optval = OPTBIT(IN6P_DSTOPTS); 1584 break; 1585 1586 case IPV6_RTHDR: 1587 optval = OPTBIT(IN6P_RTHDR); 1588 break; 1589 1590 case IPV6_FAITH: 1591 optval = OPTBIT(IN6P_FAITH); 1592 break; 1593 1594 case IPV6_V6ONLY: 1595 optval = OPTBIT(IN6P_IPV6_V6ONLY); 1596 break; 1597 1598 case IPV6_USE_MIN_MTU: 1599 optval = OPTBIT(IN6P_MINMTU); 1600 break; 1601 } 1602 *mtod(m, int *) = optval; 1603 break; 1604 1605 case IPV6_MULTICAST_IF: 1606 case IPV6_MULTICAST_HOPS: 1607 case IPV6_MULTICAST_LOOP: 1608 case IPV6_JOIN_GROUP: 1609 case IPV6_LEAVE_GROUP: 1610 error = ip6_getmoptions(optname, in6p->in6p_moptions, mp); 1611 break; 1612 1613 #if 0 /* defined(IPSEC) */ 1614 /* XXX: code broken */ 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) { 1813 if (ifindex < 0 || if_indexlim <= ifindex || 1814 !ifindex2ifnet[ifindex]) { 1815 error = ENXIO; /* XXX EINVAL? */ 1816 break; 1817 } 1818 ifp = ifindex2ifnet[ifindex]; 1819 if ((ifp->if_flags & IFF_MULTICAST) == 0) { 1820 error = EADDRNOTAVAIL; 1821 break; 1822 } 1823 } else 1824 ifp = NULL; 1825 im6o->im6o_multicast_ifp = ifp; 1826 break; 1827 1828 case IPV6_MULTICAST_HOPS: 1829 { 1830 /* 1831 * Set the IP6 hoplimit for outgoing multicast packets. 1832 */ 1833 int optval; 1834 if (m == NULL || m->m_len != sizeof(int)) { 1835 error = EINVAL; 1836 break; 1837 } 1838 bcopy(mtod(m, u_int *), &optval, sizeof(optval)); 1839 if (optval < -1 || optval >= 256) 1840 error = EINVAL; 1841 else if (optval == -1) 1842 im6o->im6o_multicast_hlim = ip6_defmcasthlim; 1843 else 1844 im6o->im6o_multicast_hlim = optval; 1845 break; 1846 } 1847 1848 case IPV6_MULTICAST_LOOP: 1849 /* 1850 * Set the loopback flag for outgoing multicast packets. 1851 * Must be zero or one. 1852 */ 1853 if (m == NULL || m->m_len != sizeof(u_int)) { 1854 error = EINVAL; 1855 break; 1856 } 1857 bcopy(mtod(m, u_int *), &loop, sizeof(loop)); 1858 if (loop > 1) { 1859 error = EINVAL; 1860 break; 1861 } 1862 im6o->im6o_multicast_loop = loop; 1863 break; 1864 1865 case IPV6_JOIN_GROUP: 1866 /* 1867 * Add a multicast group membership. 1868 * Group must be a valid IP6 multicast address. 1869 */ 1870 if (m == NULL || m->m_len != sizeof(struct ipv6_mreq)) { 1871 error = EINVAL; 1872 break; 1873 } 1874 mreq = mtod(m, struct ipv6_mreq *); 1875 if (IN6_IS_ADDR_UNSPECIFIED(&mreq->ipv6mr_multiaddr)) { 1876 /* 1877 * We use the unspecified address to specify to accept 1878 * all multicast addresses. Only super user is allowed 1879 * to do this. 1880 */ 1881 if (suser(p->p_ucred, &p->p_acflag)) 1882 { 1883 error = EACCES; 1884 break; 1885 } 1886 } else if (!IN6_IS_ADDR_MULTICAST(&mreq->ipv6mr_multiaddr)) { 1887 error = EINVAL; 1888 break; 1889 } 1890 1891 /* 1892 * If the interface is specified, validate it. 1893 * If no interface was explicitly specified, choose an 1894 * appropriate one according to the given multicast address. 1895 */ 1896 if (mreq->ipv6mr_interface != 0) { 1897 if (mreq->ipv6mr_interface < 0 || 1898 if_indexlim <= mreq->ipv6mr_interface || 1899 !ifindex2ifnet[mreq->ipv6mr_interface]) { 1900 error = ENXIO; /* XXX EINVAL? */ 1901 break; 1902 } 1903 ifp = ifindex2ifnet[mreq->ipv6mr_interface]; 1904 } else { 1905 /* 1906 * If the multicast address is in node-local scope, 1907 * the interface should be a loopback interface. 1908 * Otherwise, look up the routing table for the 1909 * address, and choose the outgoing interface. 1910 * XXX: is it a good approach? 1911 */ 1912 if (IN6_IS_ADDR_MC_NODELOCAL(&mreq->ipv6mr_multiaddr)) { 1913 ifp = lo0ifp; 1914 } else { 1915 ro.ro_rt = NULL; 1916 dst = (struct sockaddr_in6 *)&ro.ro_dst; 1917 bzero(dst, sizeof(*dst)); 1918 dst->sin6_len = sizeof(struct sockaddr_in6); 1919 dst->sin6_family = AF_INET6; 1920 dst->sin6_addr = mreq->ipv6mr_multiaddr; 1921 rtalloc((struct route *)&ro); 1922 if (ro.ro_rt == NULL) { 1923 error = EADDRNOTAVAIL; 1924 break; 1925 } 1926 ifp = ro.ro_rt->rt_ifp; 1927 rtfree(ro.ro_rt); 1928 } 1929 } 1930 1931 /* 1932 * See if we found an interface, and confirm that it 1933 * supports multicast 1934 */ 1935 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) { 1936 error = EADDRNOTAVAIL; 1937 break; 1938 } 1939 /* 1940 * Put interface index into the multicast address, 1941 * if the address has link-local scope. 1942 */ 1943 if (IN6_IS_ADDR_MC_LINKLOCAL(&mreq->ipv6mr_multiaddr)) { 1944 mreq->ipv6mr_multiaddr.s6_addr16[1] = 1945 htons(ifp->if_index); 1946 } 1947 /* 1948 * See if the membership already exists. 1949 */ 1950 for (imm = im6o->im6o_memberships.lh_first; 1951 imm != NULL; imm = imm->i6mm_chain.le_next) 1952 if (imm->i6mm_maddr->in6m_ifp == ifp && 1953 IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr, 1954 &mreq->ipv6mr_multiaddr)) 1955 break; 1956 if (imm != NULL) { 1957 error = EADDRINUSE; 1958 break; 1959 } 1960 /* 1961 * Everything looks good; add a new record to the multicast 1962 * address list for the given interface. 1963 */ 1964 imm = in6_joingroup(ifp, &mreq->ipv6mr_multiaddr, &error); 1965 if (!imm) 1966 break; 1967 LIST_INSERT_HEAD(&im6o->im6o_memberships, imm, i6mm_chain); 1968 break; 1969 1970 case IPV6_LEAVE_GROUP: 1971 /* 1972 * Drop a multicast group membership. 1973 * Group must be a valid IP6 multicast address. 1974 */ 1975 if (m == NULL || m->m_len != sizeof(struct ipv6_mreq)) { 1976 error = EINVAL; 1977 break; 1978 } 1979 mreq = mtod(m, struct ipv6_mreq *); 1980 if (IN6_IS_ADDR_UNSPECIFIED(&mreq->ipv6mr_multiaddr)) { 1981 if (suser(p->p_ucred, &p->p_acflag)) 1982 { 1983 error = EACCES; 1984 break; 1985 } 1986 } else if (!IN6_IS_ADDR_MULTICAST(&mreq->ipv6mr_multiaddr)) { 1987 error = EINVAL; 1988 break; 1989 } 1990 /* 1991 * If an interface address was specified, get a pointer 1992 * to its ifnet structure. 1993 */ 1994 if (mreq->ipv6mr_interface != 0) { 1995 if (mreq->ipv6mr_interface < 0 || 1996 if_indexlim <= mreq->ipv6mr_interface || 1997 !ifindex2ifnet[mreq->ipv6mr_interface]) { 1998 error = ENXIO; /* XXX EINVAL? */ 1999 break; 2000 } 2001 ifp = ifindex2ifnet[mreq->ipv6mr_interface]; 2002 } else 2003 ifp = NULL; 2004 /* 2005 * Put interface index into the multicast address, 2006 * if the address has link-local scope. 2007 */ 2008 if (IN6_IS_ADDR_MC_LINKLOCAL(&mreq->ipv6mr_multiaddr)) { 2009 mreq->ipv6mr_multiaddr.s6_addr16[1] = 2010 htons(mreq->ipv6mr_interface); 2011 } 2012 /* 2013 * Find the membership in the membership list. 2014 */ 2015 for (imm = im6o->im6o_memberships.lh_first; 2016 imm != NULL; imm = imm->i6mm_chain.le_next) { 2017 if ((ifp == NULL || imm->i6mm_maddr->in6m_ifp == ifp) && 2018 IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr, 2019 &mreq->ipv6mr_multiaddr)) 2020 break; 2021 } 2022 if (imm == NULL) { 2023 /* Unable to resolve interface */ 2024 error = EADDRNOTAVAIL; 2025 break; 2026 } 2027 /* 2028 * Give up the multicast address record to which the 2029 * membership points. 2030 */ 2031 LIST_REMOVE(imm, i6mm_chain); 2032 in6_leavegroup(imm); 2033 break; 2034 2035 default: 2036 error = EOPNOTSUPP; 2037 break; 2038 } 2039 2040 /* 2041 * If all options have default values, no need to keep the mbuf. 2042 */ 2043 if (im6o->im6o_multicast_ifp == NULL && 2044 im6o->im6o_multicast_hlim == ip6_defmcasthlim && 2045 im6o->im6o_multicast_loop == IPV6_DEFAULT_MULTICAST_LOOP && 2046 im6o->im6o_memberships.lh_first == NULL) { 2047 free(*im6op, M_IPMOPTS); 2048 *im6op = NULL; 2049 } 2050 2051 return (error); 2052 } 2053 2054 /* 2055 * Return the IP6 multicast options in response to user getsockopt(). 2056 */ 2057 static int 2058 ip6_getmoptions(optname, im6o, mp) 2059 int optname; 2060 struct ip6_moptions *im6o; 2061 struct mbuf **mp; 2062 { 2063 u_int *hlim, *loop, *ifindex; 2064 2065 *mp = m_get(M_WAIT, MT_SOOPTS); 2066 2067 switch (optname) { 2068 2069 case IPV6_MULTICAST_IF: 2070 ifindex = mtod(*mp, u_int *); 2071 (*mp)->m_len = sizeof(u_int); 2072 if (im6o == NULL || im6o->im6o_multicast_ifp == NULL) 2073 *ifindex = 0; 2074 else 2075 *ifindex = im6o->im6o_multicast_ifp->if_index; 2076 return (0); 2077 2078 case IPV6_MULTICAST_HOPS: 2079 hlim = mtod(*mp, u_int *); 2080 (*mp)->m_len = sizeof(u_int); 2081 if (im6o == NULL) 2082 *hlim = ip6_defmcasthlim; 2083 else 2084 *hlim = im6o->im6o_multicast_hlim; 2085 return (0); 2086 2087 case IPV6_MULTICAST_LOOP: 2088 loop = mtod(*mp, u_int *); 2089 (*mp)->m_len = sizeof(u_int); 2090 if (im6o == NULL) 2091 *loop = ip6_defmcasthlim; 2092 else 2093 *loop = im6o->im6o_multicast_loop; 2094 return (0); 2095 2096 default: 2097 return (EOPNOTSUPP); 2098 } 2099 } 2100 2101 /* 2102 * Discard the IP6 multicast options. 2103 */ 2104 void 2105 ip6_freemoptions(im6o) 2106 struct ip6_moptions *im6o; 2107 { 2108 struct in6_multi_mship *imm; 2109 2110 if (im6o == NULL) 2111 return; 2112 2113 while ((imm = im6o->im6o_memberships.lh_first) != NULL) { 2114 LIST_REMOVE(imm, i6mm_chain); 2115 in6_leavegroup(imm); 2116 } 2117 free(im6o, M_IPMOPTS); 2118 } 2119 2120 /* 2121 * Set IPv6 outgoing packet options based on advanced API. 2122 */ 2123 int 2124 ip6_setpktoptions(control, opt, priv) 2125 struct mbuf *control; 2126 struct ip6_pktopts *opt; 2127 int priv; 2128 { 2129 struct cmsghdr *cm = 0; 2130 2131 if (control == 0 || opt == 0) 2132 return (EINVAL); 2133 2134 bzero(opt, sizeof(*opt)); 2135 opt->ip6po_hlim = -1; /* -1 means to use default hop limit */ 2136 2137 /* 2138 * XXX: Currently, we assume all the optional information is stored 2139 * in a single mbuf. 2140 */ 2141 if (control->m_next) 2142 return (EINVAL); 2143 2144 opt->ip6po_m = control; 2145 2146 for (; control->m_len; control->m_data += CMSG_ALIGN(cm->cmsg_len), 2147 control->m_len -= CMSG_ALIGN(cm->cmsg_len)) { 2148 cm = mtod(control, struct cmsghdr *); 2149 if (cm->cmsg_len == 0 || cm->cmsg_len > control->m_len) 2150 return (EINVAL); 2151 if (cm->cmsg_level != IPPROTO_IPV6) 2152 continue; 2153 2154 switch (cm->cmsg_type) { 2155 case IPV6_PKTINFO: 2156 if (cm->cmsg_len != CMSG_LEN(sizeof(struct in6_pktinfo))) 2157 return (EINVAL); 2158 opt->ip6po_pktinfo = (struct in6_pktinfo *)CMSG_DATA(cm); 2159 if (opt->ip6po_pktinfo->ipi6_ifindex && 2160 IN6_IS_ADDR_LINKLOCAL(&opt->ip6po_pktinfo->ipi6_addr)) 2161 opt->ip6po_pktinfo->ipi6_addr.s6_addr16[1] = 2162 htons(opt->ip6po_pktinfo->ipi6_ifindex); 2163 2164 if (opt->ip6po_pktinfo->ipi6_ifindex >= if_indexlim || 2165 opt->ip6po_pktinfo->ipi6_ifindex < 0) 2166 return (ENXIO); 2167 if (opt->ip6po_pktinfo->ipi6_ifindex > 0 && 2168 !ifindex2ifnet[opt->ip6po_pktinfo->ipi6_ifindex]) 2169 return (ENXIO); 2170 2171 /* 2172 * Check if the requested source address is indeed a 2173 * unicast address assigned to the node, and can be 2174 * used as the packet's source address. 2175 */ 2176 if (!IN6_IS_ADDR_UNSPECIFIED(&opt->ip6po_pktinfo->ipi6_addr)) { 2177 struct ifaddr *ia; 2178 struct in6_ifaddr *ia6; 2179 struct sockaddr_in6 sin6; 2180 2181 bzero(&sin6, sizeof(sin6)); 2182 sin6.sin6_len = sizeof(sin6); 2183 sin6.sin6_family = AF_INET6; 2184 sin6.sin6_addr = 2185 opt->ip6po_pktinfo->ipi6_addr; 2186 ia = ifa_ifwithaddr(sin6tosa(&sin6)); 2187 if (ia == NULL || 2188 (opt->ip6po_pktinfo->ipi6_ifindex && 2189 (ia->ifa_ifp->if_index != 2190 opt->ip6po_pktinfo->ipi6_ifindex))) { 2191 return (EADDRNOTAVAIL); 2192 } 2193 ia6 = (struct in6_ifaddr *)ia; 2194 if ((ia6->ia6_flags & (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY)) != 0) { 2195 return (EADDRNOTAVAIL); 2196 } 2197 2198 /* 2199 * Check if the requested source address is 2200 * indeed a unicast address assigned to the 2201 * node. 2202 */ 2203 if (IN6_IS_ADDR_MULTICAST(&opt->ip6po_pktinfo->ipi6_addr)) 2204 return (EADDRNOTAVAIL); 2205 } 2206 break; 2207 2208 case IPV6_HOPLIMIT: 2209 if (cm->cmsg_len != CMSG_LEN(sizeof(int))) 2210 return (EINVAL); 2211 else { 2212 int t; 2213 2214 bcopy(CMSG_DATA(cm), &t, sizeof(t)); 2215 if (t < -1 || t > 255) 2216 return (EINVAL); 2217 opt->ip6po_hlim = t; 2218 } 2219 break; 2220 2221 case IPV6_NEXTHOP: 2222 if (!priv) 2223 return (EPERM); 2224 2225 /* check if cmsg_len is large enough for sa_len */ 2226 if (cm->cmsg_len < sizeof(u_char) || 2227 cm->cmsg_len < CMSG_LEN(*CMSG_DATA(cm))) 2228 return (EINVAL); 2229 2230 opt->ip6po_nexthop = (struct sockaddr *)CMSG_DATA(cm); 2231 2232 break; 2233 2234 case IPV6_HOPOPTS: 2235 if (cm->cmsg_len < CMSG_LEN(sizeof(struct ip6_hbh))) 2236 return (EINVAL); 2237 else { 2238 struct ip6_hbh *t; 2239 2240 t = (struct ip6_hbh *)CMSG_DATA(cm); 2241 if (cm->cmsg_len != 2242 CMSG_LEN((t->ip6h_len + 1) << 3)) 2243 return (EINVAL); 2244 opt->ip6po_hbh = t; 2245 } 2246 break; 2247 2248 case IPV6_DSTOPTS: 2249 if (cm->cmsg_len < CMSG_LEN(sizeof(struct ip6_dest))) 2250 return (EINVAL); 2251 2252 /* 2253 * If there is no routing header yet, the destination 2254 * options header should be put on the 1st part. 2255 * Otherwise, the header should be on the 2nd part. 2256 * (See RFC 2460, section 4.1) 2257 */ 2258 if (opt->ip6po_rthdr == NULL) { 2259 struct ip6_dest *t; 2260 2261 t = (struct ip6_dest *)CMSG_DATA(cm); 2262 if (cm->cmsg_len != 2263 CMSG_LEN((t->ip6d_len + 1) << 3)); 2264 return (EINVAL); 2265 opt->ip6po_dest1 = t; 2266 } 2267 else { 2268 struct ip6_dest *t; 2269 2270 t = (struct ip6_dest *)CMSG_DATA(cm); 2271 if (cm->cmsg_len != 2272 CMSG_LEN((opt->ip6po_dest2->ip6d_len + 1) << 3)) 2273 return (EINVAL); 2274 opt->ip6po_dest2 = t; 2275 } 2276 break; 2277 2278 case IPV6_RTHDR: 2279 if (cm->cmsg_len < CMSG_LEN(sizeof(struct ip6_rthdr))) 2280 return (EINVAL); 2281 else { 2282 struct ip6_rthdr *t; 2283 2284 t = (struct ip6_rthdr *)CMSG_DATA(cm); 2285 if (cm->cmsg_len != 2286 CMSG_LEN((t->ip6r_len + 1) << 3)) 2287 return (EINVAL); 2288 switch (t->ip6r_type) { 2289 case IPV6_RTHDR_TYPE_0: 2290 if (t->ip6r_segleft == 0) 2291 return (EINVAL); 2292 break; 2293 default: 2294 return (EINVAL); 2295 } 2296 opt->ip6po_rthdr = t; 2297 } 2298 break; 2299 2300 default: 2301 return (ENOPROTOOPT); 2302 } 2303 } 2304 2305 return (0); 2306 } 2307 2308 /* 2309 * Routine called from ip6_output() to loop back a copy of an IP6 multicast 2310 * packet to the input queue of a specified interface. Note that this 2311 * calls the output routine of the loopback "driver", but with an interface 2312 * pointer that might NOT be lo0ifp -- easier than replicating that code here. 2313 */ 2314 void 2315 ip6_mloopback(ifp, m, dst) 2316 struct ifnet *ifp; 2317 struct mbuf *m; 2318 struct sockaddr_in6 *dst; 2319 { 2320 struct mbuf *copym; 2321 struct ip6_hdr *ip6; 2322 2323 copym = m_copy(m, 0, M_COPYALL); 2324 if (copym == NULL) 2325 return; 2326 2327 /* 2328 * Make sure to deep-copy IPv6 header portion in case the data 2329 * is in an mbuf cluster, so that we can safely override the IPv6 2330 * header portion later. 2331 */ 2332 if ((copym->m_flags & M_EXT) != 0 || 2333 copym->m_len < sizeof(struct ip6_hdr)) { 2334 copym = m_pullup(copym, sizeof(struct ip6_hdr)); 2335 if (copym == NULL) 2336 return; 2337 } 2338 2339 #ifdef DIAGNOSTIC 2340 if (copym->m_len < sizeof(*ip6)) { 2341 m_freem(copym); 2342 return; 2343 } 2344 #endif 2345 2346 ip6 = mtod(copym, struct ip6_hdr *); 2347 if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src)) 2348 ip6->ip6_src.s6_addr16[1] = 0; 2349 if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst)) 2350 ip6->ip6_dst.s6_addr16[1] = 0; 2351 2352 (void)looutput(ifp, copym, (struct sockaddr *)dst, NULL); 2353 } 2354 2355 /* 2356 * Chop IPv6 header off from the payload. 2357 */ 2358 static int 2359 ip6_splithdr(m, exthdrs) 2360 struct mbuf *m; 2361 struct ip6_exthdrs *exthdrs; 2362 { 2363 struct mbuf *mh; 2364 struct ip6_hdr *ip6; 2365 2366 ip6 = mtod(m, struct ip6_hdr *); 2367 if (m->m_len > sizeof(*ip6)) { 2368 MGETHDR(mh, M_DONTWAIT, MT_HEADER); 2369 if (mh == 0) { 2370 m_freem(m); 2371 return ENOBUFS; 2372 } 2373 M_COPY_PKTHDR(mh, m); 2374 MH_ALIGN(mh, sizeof(*ip6)); 2375 m_tag_delete_chain(m, NULL); 2376 m->m_flags &= ~M_PKTHDR; 2377 m->m_len -= sizeof(*ip6); 2378 m->m_data += sizeof(*ip6); 2379 mh->m_next = m; 2380 m = mh; 2381 m->m_len = sizeof(*ip6); 2382 bcopy((caddr_t)ip6, mtod(m, caddr_t), sizeof(*ip6)); 2383 } 2384 exthdrs->ip6e_ip6 = m; 2385 return 0; 2386 } 2387 2388 /* 2389 * Compute IPv6 extension header length. 2390 */ 2391 int 2392 ip6_optlen(in6p) 2393 struct in6pcb *in6p; 2394 { 2395 int len; 2396 2397 if (!in6p->in6p_outputopts) 2398 return 0; 2399 2400 len = 0; 2401 #define elen(x) \ 2402 (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0) 2403 2404 len += elen(in6p->in6p_outputopts->ip6po_hbh); 2405 len += elen(in6p->in6p_outputopts->ip6po_dest1); 2406 len += elen(in6p->in6p_outputopts->ip6po_rthdr); 2407 len += elen(in6p->in6p_outputopts->ip6po_dest2); 2408 return len; 2409 #undef elen 2410 } 2411