1 /* $NetBSD: ip6_output.c,v 1.60 2002/11/02 07:30:58 perry 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. All advertising materials mentioning features or use of this software 46 * must display the following acknowledgement: 47 * This product includes software developed by the University of 48 * California, Berkeley and its contributors. 49 * 4. Neither the name of the University nor the names of its contributors 50 * may be used to endorse or promote products derived from this software 51 * without specific prior written permission. 52 * 53 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 54 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 55 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 56 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 57 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 58 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 59 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 60 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 61 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 62 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 63 * SUCH DAMAGE. 64 * 65 * @(#)ip_output.c 8.3 (Berkeley) 1/21/94 66 */ 67 68 #include <sys/cdefs.h> 69 __KERNEL_RCSID(0, "$NetBSD: ip6_output.c,v 1.60 2002/11/02 07:30:58 perry Exp $"); 70 71 #include "opt_inet.h" 72 #include "opt_ipsec.h" 73 #include "opt_pfil_hooks.h" 74 75 #include <sys/param.h> 76 #include <sys/malloc.h> 77 #include <sys/mbuf.h> 78 #include <sys/errno.h> 79 #include <sys/protosw.h> 80 #include <sys/socket.h> 81 #include <sys/socketvar.h> 82 #include <sys/systm.h> 83 #include <sys/proc.h> 84 85 #include <net/if.h> 86 #include <net/route.h> 87 #ifdef PFIL_HOOKS 88 #include <net/pfil.h> 89 #endif 90 91 #include <netinet/in.h> 92 #include <netinet/in_var.h> 93 #include <netinet/ip6.h> 94 #include <netinet/icmp6.h> 95 #include <netinet6/ip6_var.h> 96 #include <netinet6/in6_pcb.h> 97 #include <netinet6/nd6.h> 98 99 #ifdef IPSEC 100 #include <netinet6/ipsec.h> 101 #include <netkey/key.h> 102 #endif /* IPSEC */ 103 104 #include "loop.h" 105 106 #include <net/net_osdep.h> 107 108 #ifdef PFIL_HOOKS 109 extern struct pfil_head inet6_pfil_hook; /* XXX */ 110 #endif 111 112 struct ip6_exthdrs { 113 struct mbuf *ip6e_ip6; 114 struct mbuf *ip6e_hbh; 115 struct mbuf *ip6e_dest1; 116 struct mbuf *ip6e_rthdr; 117 struct mbuf *ip6e_dest2; 118 }; 119 120 static int ip6_pcbopts __P((struct ip6_pktopts **, struct mbuf *, 121 struct socket *)); 122 static int ip6_setmoptions __P((int, struct ip6_moptions **, struct mbuf *)); 123 static int ip6_getmoptions __P((int, struct ip6_moptions *, struct mbuf **)); 124 static int ip6_copyexthdr __P((struct mbuf **, caddr_t, int)); 125 static int ip6_insertfraghdr __P((struct mbuf *, struct mbuf *, int, 126 struct ip6_frag **)); 127 static int ip6_insert_jumboopt __P((struct ip6_exthdrs *, u_int32_t)); 128 static int ip6_splithdr __P((struct mbuf *, struct ip6_exthdrs *)); 129 static int ip6_getpmtu __P((struct route_in6 *, struct route_in6 *, 130 struct ifnet *, struct in6_addr *, u_long *)); 131 132 extern struct ifnet loif[NLOOP]; 133 134 /* 135 * IP6 output. The packet in mbuf chain m contains a skeletal IP6 136 * header (with pri, len, nxt, hlim, src, dst). 137 * This function may modify ver and hlim only. 138 * The mbuf chain containing the packet will be freed. 139 * The mbuf opt, if present, will not be freed. 140 * 141 * type of "mtu": rt_rmx.rmx_mtu is u_long, ifnet.ifr_mtu is int, and 142 * nd_ifinfo.linkmtu is u_int32_t. so we use u_long to hold largest one, 143 * which is rt_rmx.rmx_mtu. 144 */ 145 int 146 ip6_output(m0, opt, ro, flags, im6o, ifpp) 147 struct mbuf *m0; 148 struct ip6_pktopts *opt; 149 struct route_in6 *ro; 150 int flags; 151 struct ip6_moptions *im6o; 152 struct ifnet **ifpp; /* XXX: just for statistics */ 153 { 154 struct ip6_hdr *ip6, *mhip6; 155 struct ifnet *ifp, *origifp; 156 struct mbuf *m = m0; 157 int hlen, tlen, len, off; 158 struct route_in6 ip6route; 159 struct sockaddr_in6 *dst; 160 int error = 0; 161 struct in6_ifaddr *ia; 162 u_long mtu; 163 u_int32_t optlen = 0, plen = 0, unfragpartlen = 0; 164 struct ip6_exthdrs exthdrs; 165 struct in6_addr finaldst; 166 struct route_in6 *ro_pmtu = NULL; 167 int hdrsplit = 0; 168 int needipsec = 0; 169 #ifdef IPSEC 170 int needipsectun = 0; 171 struct socket *so; 172 struct secpolicy *sp = NULL; 173 174 /* for AH processing. stupid to have "socket" variable in IP layer... */ 175 so = ipsec_getsocket(m); 176 (void)ipsec_setsocket(m, NULL); 177 ip6 = mtod(m, struct ip6_hdr *); 178 #endif /* IPSEC */ 179 180 #define MAKE_EXTHDR(hp, mp) \ 181 do { \ 182 if (hp) { \ 183 struct ip6_ext *eh = (struct ip6_ext *)(hp); \ 184 error = ip6_copyexthdr((mp), (caddr_t)(hp), \ 185 ((eh)->ip6e_len + 1) << 3); \ 186 if (error) \ 187 goto freehdrs; \ 188 } \ 189 } while (/*CONSTCOND*/ 0) 190 191 bzero(&exthdrs, sizeof(exthdrs)); 192 if (opt) { 193 /* Hop-by-Hop options header */ 194 MAKE_EXTHDR(opt->ip6po_hbh, &exthdrs.ip6e_hbh); 195 /* Destination options header(1st part) */ 196 MAKE_EXTHDR(opt->ip6po_dest1, &exthdrs.ip6e_dest1); 197 /* Routing header */ 198 MAKE_EXTHDR(opt->ip6po_rthdr, &exthdrs.ip6e_rthdr); 199 /* Destination options header(2nd part) */ 200 MAKE_EXTHDR(opt->ip6po_dest2, &exthdrs.ip6e_dest2); 201 } 202 203 #ifdef IPSEC 204 /* get a security policy for this packet */ 205 if (so == NULL) 206 sp = ipsec6_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND, 0, &error); 207 else 208 sp = ipsec6_getpolicybysock(m, IPSEC_DIR_OUTBOUND, so, &error); 209 210 if (sp == NULL) { 211 ipsec6stat.out_inval++; 212 goto freehdrs; 213 } 214 215 error = 0; 216 217 /* check policy */ 218 switch (sp->policy) { 219 case IPSEC_POLICY_DISCARD: 220 /* 221 * This packet is just discarded. 222 */ 223 ipsec6stat.out_polvio++; 224 goto freehdrs; 225 226 case IPSEC_POLICY_BYPASS: 227 case IPSEC_POLICY_NONE: 228 /* no need to do IPsec. */ 229 needipsec = 0; 230 break; 231 232 case IPSEC_POLICY_IPSEC: 233 if (sp->req == NULL) { 234 /* XXX should be panic ? */ 235 printf("ip6_output: No IPsec request specified.\n"); 236 error = EINVAL; 237 goto freehdrs; 238 } 239 needipsec = 1; 240 break; 241 242 case IPSEC_POLICY_ENTRUST: 243 default: 244 printf("ip6_output: Invalid policy found. %d\n", sp->policy); 245 } 246 #endif /* IPSEC */ 247 248 /* 249 * Calculate the total length of the extension header chain. 250 * Keep the length of the unfragmentable part for fragmentation. 251 */ 252 optlen = 0; 253 if (exthdrs.ip6e_hbh) optlen += exthdrs.ip6e_hbh->m_len; 254 if (exthdrs.ip6e_dest1) optlen += exthdrs.ip6e_dest1->m_len; 255 if (exthdrs.ip6e_rthdr) optlen += exthdrs.ip6e_rthdr->m_len; 256 unfragpartlen = optlen + sizeof(struct ip6_hdr); 257 /* NOTE: we don't add AH/ESP length here. do that later. */ 258 if (exthdrs.ip6e_dest2) optlen += exthdrs.ip6e_dest2->m_len; 259 260 /* 261 * If we need IPsec, or there is at least one extension header, 262 * separate IP6 header from the payload. 263 */ 264 if ((needipsec || optlen) && !hdrsplit) { 265 if ((error = ip6_splithdr(m, &exthdrs)) != 0) { 266 m = NULL; 267 goto freehdrs; 268 } 269 m = exthdrs.ip6e_ip6; 270 hdrsplit++; 271 } 272 273 /* adjust pointer */ 274 ip6 = mtod(m, struct ip6_hdr *); 275 276 /* adjust mbuf packet header length */ 277 m->m_pkthdr.len += optlen; 278 plen = m->m_pkthdr.len - sizeof(*ip6); 279 280 /* If this is a jumbo payload, insert a jumbo payload option. */ 281 if (plen > IPV6_MAXPACKET) { 282 if (!hdrsplit) { 283 if ((error = ip6_splithdr(m, &exthdrs)) != 0) { 284 m = NULL; 285 goto freehdrs; 286 } 287 m = exthdrs.ip6e_ip6; 288 hdrsplit++; 289 } 290 /* adjust pointer */ 291 ip6 = mtod(m, struct ip6_hdr *); 292 if ((error = ip6_insert_jumboopt(&exthdrs, plen)) != 0) 293 goto freehdrs; 294 ip6->ip6_plen = 0; 295 } else 296 ip6->ip6_plen = htons(plen); 297 298 /* 299 * Concatenate headers and fill in next header fields. 300 * Here we have, on "m" 301 * IPv6 payload 302 * and we insert headers accordingly. Finally, we should be getting: 303 * IPv6 hbh dest1 rthdr ah* [esp* dest2 payload] 304 * 305 * during the header composing process, "m" points to IPv6 header. 306 * "mprev" points to an extension header prior to esp. 307 */ 308 { 309 u_char *nexthdrp = &ip6->ip6_nxt; 310 struct mbuf *mprev = m; 311 312 /* 313 * we treat dest2 specially. this makes IPsec processing 314 * much easier. 315 * 316 * result: IPv6 dest2 payload 317 * m and mprev will point to IPv6 header. 318 */ 319 if (exthdrs.ip6e_dest2) { 320 if (!hdrsplit) 321 panic("assumption failed: hdr not split"); 322 exthdrs.ip6e_dest2->m_next = m->m_next; 323 m->m_next = exthdrs.ip6e_dest2; 324 *mtod(exthdrs.ip6e_dest2, u_char *) = ip6->ip6_nxt; 325 ip6->ip6_nxt = IPPROTO_DSTOPTS; 326 } 327 328 #define MAKE_CHAIN(m, mp, p, i)\ 329 do {\ 330 if (m) {\ 331 if (!hdrsplit) \ 332 panic("assumption failed: hdr not split"); \ 333 *mtod((m), u_char *) = *(p);\ 334 *(p) = (i);\ 335 p = mtod((m), u_char *);\ 336 (m)->m_next = (mp)->m_next;\ 337 (mp)->m_next = (m);\ 338 (mp) = (m);\ 339 }\ 340 } while (/*CONSTCOND*/ 0) 341 /* 342 * result: IPv6 hbh dest1 rthdr dest2 payload 343 * m will point to IPv6 header. mprev will point to the 344 * extension header prior to dest2 (rthdr in the above case). 345 */ 346 MAKE_CHAIN(exthdrs.ip6e_hbh, mprev, nexthdrp, IPPROTO_HOPOPTS); 347 MAKE_CHAIN(exthdrs.ip6e_dest1, mprev, nexthdrp, 348 IPPROTO_DSTOPTS); 349 MAKE_CHAIN(exthdrs.ip6e_rthdr, mprev, nexthdrp, 350 IPPROTO_ROUTING); 351 352 #ifdef IPSEC 353 if (!needipsec) 354 goto skip_ipsec2; 355 356 /* 357 * pointers after IPsec headers are not valid any more. 358 * other pointers need a great care too. 359 * (IPsec routines should not mangle mbufs prior to AH/ESP) 360 */ 361 exthdrs.ip6e_dest2 = NULL; 362 363 { 364 struct ip6_rthdr *rh = NULL; 365 int segleft_org = 0; 366 struct ipsec_output_state state; 367 368 if (exthdrs.ip6e_rthdr) { 369 rh = mtod(exthdrs.ip6e_rthdr, struct ip6_rthdr *); 370 segleft_org = rh->ip6r_segleft; 371 rh->ip6r_segleft = 0; 372 } 373 374 bzero(&state, sizeof(state)); 375 state.m = m; 376 error = ipsec6_output_trans(&state, nexthdrp, mprev, sp, flags, 377 &needipsectun); 378 m = state.m; 379 if (error) { 380 /* mbuf is already reclaimed in ipsec6_output_trans. */ 381 m = NULL; 382 switch (error) { 383 case EHOSTUNREACH: 384 case ENETUNREACH: 385 case EMSGSIZE: 386 case ENOBUFS: 387 case ENOMEM: 388 break; 389 default: 390 printf("ip6_output (ipsec): error code %d\n", error); 391 /* FALLTHROUGH */ 392 case ENOENT: 393 /* don't show these error codes to the user */ 394 error = 0; 395 break; 396 } 397 goto bad; 398 } 399 if (exthdrs.ip6e_rthdr) { 400 /* ah6_output doesn't modify mbuf chain */ 401 rh->ip6r_segleft = segleft_org; 402 } 403 } 404 skip_ipsec2:; 405 #endif 406 } 407 408 /* 409 * If there is a routing header, replace destination address field 410 * with the first hop of the routing header. 411 */ 412 if (exthdrs.ip6e_rthdr) { 413 struct ip6_rthdr *rh; 414 struct ip6_rthdr0 *rh0; 415 416 rh = (struct ip6_rthdr *)(mtod(exthdrs.ip6e_rthdr, 417 struct ip6_rthdr *)); 418 finaldst = ip6->ip6_dst; 419 switch (rh->ip6r_type) { 420 case IPV6_RTHDR_TYPE_0: 421 rh0 = (struct ip6_rthdr0 *)rh; 422 ip6->ip6_dst = rh0->ip6r0_addr[0]; 423 bcopy((caddr_t)&rh0->ip6r0_addr[1], 424 (caddr_t)&rh0->ip6r0_addr[0], 425 sizeof(struct in6_addr) * (rh0->ip6r0_segleft - 1)); 426 rh0->ip6r0_addr[rh0->ip6r0_segleft - 1] = finaldst; 427 break; 428 default: /* is it possible? */ 429 error = EINVAL; 430 goto bad; 431 } 432 } 433 434 /* Source address validation */ 435 if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) && 436 (flags & IPV6_UNSPECSRC) == 0) { 437 error = EOPNOTSUPP; 438 ip6stat.ip6s_badscope++; 439 goto bad; 440 } 441 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) { 442 error = EOPNOTSUPP; 443 ip6stat.ip6s_badscope++; 444 goto bad; 445 } 446 447 ip6stat.ip6s_localout++; 448 449 /* 450 * Route packet. 451 */ 452 if (ro == 0) { 453 ro = &ip6route; 454 bzero((caddr_t)ro, sizeof(*ro)); 455 } 456 ro_pmtu = ro; 457 if (opt && opt->ip6po_rthdr) 458 ro = &opt->ip6po_route; 459 dst = (struct sockaddr_in6 *)&ro->ro_dst; 460 /* 461 * If there is a cached route, 462 * check that it is to the same destination 463 * and is still up. If not, free it and try again. 464 */ 465 if (ro->ro_rt && ((ro->ro_rt->rt_flags & RTF_UP) == 0 || 466 dst->sin6_family != AF_INET6 || 467 !IN6_ARE_ADDR_EQUAL(&dst->sin6_addr, &ip6->ip6_dst))) { 468 RTFREE(ro->ro_rt); 469 ro->ro_rt = (struct rtentry *)0; 470 } 471 if (ro->ro_rt == 0) { 472 bzero(dst, sizeof(*dst)); 473 dst->sin6_family = AF_INET6; 474 dst->sin6_len = sizeof(struct sockaddr_in6); 475 dst->sin6_addr = ip6->ip6_dst; 476 } 477 #ifdef IPSEC 478 if (needipsec && needipsectun) { 479 struct ipsec_output_state state; 480 481 /* 482 * All the extension headers will become inaccessible 483 * (since they can be encrypted). 484 * Don't panic, we need no more updates to extension headers 485 * on inner IPv6 packet (since they are now encapsulated). 486 * 487 * IPv6 [ESP|AH] IPv6 [extension headers] payload 488 */ 489 bzero(&exthdrs, sizeof(exthdrs)); 490 exthdrs.ip6e_ip6 = m; 491 492 bzero(&state, sizeof(state)); 493 state.m = m; 494 state.ro = (struct route *)ro; 495 state.dst = (struct sockaddr *)dst; 496 497 error = ipsec6_output_tunnel(&state, sp, flags); 498 499 m = state.m; 500 ro = (struct route_in6 *)state.ro; 501 dst = (struct sockaddr_in6 *)state.dst; 502 if (error) { 503 /* mbuf is already reclaimed in ipsec6_output_tunnel. */ 504 m0 = m = NULL; 505 m = NULL; 506 switch (error) { 507 case EHOSTUNREACH: 508 case ENETUNREACH: 509 case EMSGSIZE: 510 case ENOBUFS: 511 case ENOMEM: 512 break; 513 default: 514 printf("ip6_output (ipsec): error code %d\n", error); 515 /* FALLTHROUGH */ 516 case ENOENT: 517 /* don't show these error codes to the user */ 518 error = 0; 519 break; 520 } 521 goto bad; 522 } 523 524 exthdrs.ip6e_ip6 = m; 525 } 526 #endif /* IPSEC */ 527 528 if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) { 529 /* Unicast */ 530 531 #define ifatoia6(ifa) ((struct in6_ifaddr *)(ifa)) 532 #define sin6tosa(sin6) ((struct sockaddr *)(sin6)) 533 /* xxx 534 * interface selection comes here 535 * if an interface is specified from an upper layer, 536 * ifp must point it. 537 */ 538 if (ro->ro_rt == 0) { 539 /* 540 * non-bsdi always clone routes, if parent is 541 * PRF_CLONING. 542 */ 543 rtalloc((struct route *)ro); 544 } 545 if (ro->ro_rt == 0) { 546 ip6stat.ip6s_noroute++; 547 error = EHOSTUNREACH; 548 /* XXX in6_ifstat_inc(ifp, ifs6_out_discard); */ 549 goto bad; 550 } 551 ia = ifatoia6(ro->ro_rt->rt_ifa); 552 ifp = ro->ro_rt->rt_ifp; 553 ro->ro_rt->rt_use++; 554 if (ro->ro_rt->rt_flags & RTF_GATEWAY) 555 dst = (struct sockaddr_in6 *)ro->ro_rt->rt_gateway; 556 m->m_flags &= ~(M_BCAST | M_MCAST); /* just in case */ 557 558 in6_ifstat_inc(ifp, ifs6_out_request); 559 560 /* 561 * Check if the outgoing interface conflicts with 562 * the interface specified by ifi6_ifindex (if specified). 563 * Note that loopback interface is always okay. 564 * (this may happen when we are sending a packet to one of 565 * our own addresses.) 566 */ 567 if (opt && opt->ip6po_pktinfo 568 && opt->ip6po_pktinfo->ipi6_ifindex) { 569 if (!(ifp->if_flags & IFF_LOOPBACK) 570 && ifp->if_index != opt->ip6po_pktinfo->ipi6_ifindex) { 571 ip6stat.ip6s_noroute++; 572 in6_ifstat_inc(ifp, ifs6_out_discard); 573 error = EHOSTUNREACH; 574 goto bad; 575 } 576 } 577 578 if (opt && opt->ip6po_hlim != -1) 579 ip6->ip6_hlim = opt->ip6po_hlim & 0xff; 580 } else { 581 /* Multicast */ 582 struct in6_multi *in6m; 583 584 m->m_flags = (m->m_flags & ~M_BCAST) | M_MCAST; 585 586 /* 587 * See if the caller provided any multicast options 588 */ 589 ifp = NULL; 590 if (im6o != NULL) { 591 ip6->ip6_hlim = im6o->im6o_multicast_hlim; 592 if (im6o->im6o_multicast_ifp != NULL) 593 ifp = im6o->im6o_multicast_ifp; 594 } else 595 ip6->ip6_hlim = ip6_defmcasthlim; 596 597 /* 598 * See if the caller provided the outgoing interface 599 * as an ancillary data. 600 * Boundary check for ifindex is assumed to be already done. 601 */ 602 if (opt && opt->ip6po_pktinfo && opt->ip6po_pktinfo->ipi6_ifindex) 603 ifp = ifindex2ifnet[opt->ip6po_pktinfo->ipi6_ifindex]; 604 605 /* 606 * If the destination is a node-local scope multicast, 607 * the packet should be loop-backed only. 608 */ 609 if (IN6_IS_ADDR_MC_NODELOCAL(&ip6->ip6_dst)) { 610 /* 611 * If the outgoing interface is already specified, 612 * it should be a loopback interface. 613 */ 614 if (ifp && (ifp->if_flags & IFF_LOOPBACK) == 0) { 615 ip6stat.ip6s_badscope++; 616 error = ENETUNREACH; /* XXX: better error? */ 617 /* XXX correct ifp? */ 618 in6_ifstat_inc(ifp, ifs6_out_discard); 619 goto bad; 620 } else { 621 ifp = &loif[0]; 622 } 623 } 624 625 if (opt && opt->ip6po_hlim != -1) 626 ip6->ip6_hlim = opt->ip6po_hlim & 0xff; 627 628 /* 629 * If caller did not provide an interface lookup a 630 * default in the routing table. This is either a 631 * default for the speicfied group (i.e. a host 632 * route), or a multicast default (a route for the 633 * ``net'' ff00::/8). 634 */ 635 if (ifp == NULL) { 636 if (ro->ro_rt == 0) { 637 ro->ro_rt = rtalloc1((struct sockaddr *) 638 &ro->ro_dst, 0); 639 } 640 if (ro->ro_rt == 0) { 641 ip6stat.ip6s_noroute++; 642 error = EHOSTUNREACH; 643 /* XXX in6_ifstat_inc(ifp, ifs6_out_discard) */ 644 goto bad; 645 } 646 ia = ifatoia6(ro->ro_rt->rt_ifa); 647 ifp = ro->ro_rt->rt_ifp; 648 ro->ro_rt->rt_use++; 649 } 650 651 if ((flags & IPV6_FORWARDING) == 0) 652 in6_ifstat_inc(ifp, ifs6_out_request); 653 in6_ifstat_inc(ifp, ifs6_out_mcast); 654 655 /* 656 * Confirm that the outgoing interface supports multicast. 657 */ 658 if ((ifp->if_flags & IFF_MULTICAST) == 0) { 659 ip6stat.ip6s_noroute++; 660 in6_ifstat_inc(ifp, ifs6_out_discard); 661 error = ENETUNREACH; 662 goto bad; 663 } 664 IN6_LOOKUP_MULTI(ip6->ip6_dst, ifp, in6m); 665 if (in6m != NULL && 666 (im6o == NULL || im6o->im6o_multicast_loop)) { 667 /* 668 * If we belong to the destination multicast group 669 * on the outgoing interface, and the caller did not 670 * forbid loopback, loop back a copy. 671 */ 672 ip6_mloopback(ifp, m, dst); 673 } else { 674 /* 675 * If we are acting as a multicast router, perform 676 * multicast forwarding as if the packet had just 677 * arrived on the interface to which we are about 678 * to send. The multicast forwarding function 679 * recursively calls this function, using the 680 * IPV6_FORWARDING flag to prevent infinite recursion. 681 * 682 * Multicasts that are looped back by ip6_mloopback(), 683 * above, will be forwarded by the ip6_input() routine, 684 * if necessary. 685 */ 686 if (ip6_mrouter && (flags & IPV6_FORWARDING) == 0) { 687 if (ip6_mforward(ip6, ifp, m) != 0) { 688 m_freem(m); 689 goto done; 690 } 691 } 692 } 693 /* 694 * Multicasts with a hoplimit of zero may be looped back, 695 * above, but must not be transmitted on a network. 696 * Also, multicasts addressed to the loopback interface 697 * are not sent -- the above call to ip6_mloopback() will 698 * loop back a copy if this host actually belongs to the 699 * destination group on the loopback interface. 700 */ 701 if (ip6->ip6_hlim == 0 || (ifp->if_flags & IFF_LOOPBACK)) { 702 m_freem(m); 703 goto done; 704 } 705 } 706 707 /* 708 * Fill the outgoing inteface to tell the upper layer 709 * to increment per-interface statistics. 710 */ 711 if (ifpp) 712 *ifpp = ifp; 713 714 /* Determine path MTU. */ 715 if ((error = ip6_getpmtu(ro_pmtu, ro, ifp, &finaldst, &mtu)) != 0) 716 goto bad; 717 718 /* 719 * The caller of this function may specify to use the minimum MTU 720 * in some cases. 721 */ 722 if (mtu > IPV6_MMTU) { 723 if ((flags & IPV6_MINMTU)) 724 mtu = IPV6_MMTU; 725 } 726 727 /* Fake scoped addresses */ 728 if ((ifp->if_flags & IFF_LOOPBACK) != 0) { 729 /* 730 * If source or destination address is a scoped address, and 731 * the packet is going to be sent to a loopback interface, 732 * we should keep the original interface. 733 */ 734 735 /* 736 * XXX: this is a very experimental and temporary solution. 737 * We eventually have sockaddr_in6 and use the sin6_scope_id 738 * field of the structure here. 739 * We rely on the consistency between two scope zone ids 740 * of source add destination, which should already be assured 741 * Larger scopes than link will be supported in the near 742 * future. 743 */ 744 origifp = NULL; 745 if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src)) 746 origifp = ifindex2ifnet[ntohs(ip6->ip6_src.s6_addr16[1])]; 747 else if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst)) 748 origifp = ifindex2ifnet[ntohs(ip6->ip6_dst.s6_addr16[1])]; 749 /* 750 * XXX: origifp can be NULL even in those two cases above. 751 * For example, if we remove the (only) link-local address 752 * from the loopback interface, and try to send a link-local 753 * address without link-id information. Then the source 754 * address is ::1, and the destination address is the 755 * link-local address with its s6_addr16[1] being zero. 756 * What is worse, if the packet goes to the loopback interface 757 * by a default rejected route, the null pointer would be 758 * passed to looutput, and the kernel would hang. 759 * The following last resort would prevent such disaster. 760 */ 761 if (origifp == NULL) 762 origifp = ifp; 763 } else 764 origifp = ifp; 765 if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src)) 766 ip6->ip6_src.s6_addr16[1] = 0; 767 if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst)) 768 ip6->ip6_dst.s6_addr16[1] = 0; 769 770 /* 771 * If the outgoing packet contains a hop-by-hop options header, 772 * it must be examined and processed even by the source node. 773 * (RFC 2460, section 4.) 774 */ 775 if (exthdrs.ip6e_hbh) { 776 struct ip6_hbh *hbh = mtod(exthdrs.ip6e_hbh, struct ip6_hbh *); 777 u_int32_t dummy1; /* XXX unused */ 778 u_int32_t dummy2; /* XXX unused */ 779 780 /* 781 * XXX: if we have to send an ICMPv6 error to the sender, 782 * we need the M_LOOP flag since icmp6_error() expects 783 * the IPv6 and the hop-by-hop options header are 784 * continuous unless the flag is set. 785 */ 786 m->m_flags |= M_LOOP; 787 m->m_pkthdr.rcvif = ifp; 788 if (ip6_process_hopopts(m, (u_int8_t *)(hbh + 1), 789 ((hbh->ip6h_len + 1) << 3) - sizeof(struct ip6_hbh), 790 &dummy1, &dummy2) < 0) { 791 /* m was already freed at this point */ 792 error = EINVAL;/* better error? */ 793 goto done; 794 } 795 m->m_flags &= ~M_LOOP; /* XXX */ 796 m->m_pkthdr.rcvif = NULL; 797 } 798 799 #ifdef PFIL_HOOKS 800 /* 801 * Run through list of hooks for output packets. 802 */ 803 if ((error = pfil_run_hooks(&inet6_pfil_hook, &m, ifp, PFIL_OUT)) != 0) 804 goto done; 805 if (m == NULL) 806 goto done; 807 ip6 = mtod(m, struct ip6_hdr *); 808 #endif /* PFIL_HOOKS */ 809 /* 810 * Send the packet to the outgoing interface. 811 * If necessary, do IPv6 fragmentation before sending. 812 */ 813 tlen = m->m_pkthdr.len; 814 if (tlen <= mtu) { 815 #ifdef IFA_STATS 816 struct in6_ifaddr *ia6; 817 ip6 = mtod(m, struct ip6_hdr *); 818 ia6 = in6_ifawithifp(ifp, &ip6->ip6_src); 819 if (ia6) { 820 /* Record statistics for this interface address. */ 821 ia6->ia_ifa.ifa_data.ifad_outbytes += 822 m->m_pkthdr.len; 823 } 824 #endif 825 #ifdef IPSEC 826 /* clean ipsec history once it goes out of the node */ 827 ipsec_delaux(m); 828 #endif 829 error = nd6_output(ifp, origifp, m, dst, ro->ro_rt); 830 goto done; 831 } else if (mtu < IPV6_MMTU) { 832 /* 833 * note that path MTU is never less than IPV6_MMTU 834 * (see icmp6_input). 835 */ 836 error = EMSGSIZE; 837 in6_ifstat_inc(ifp, ifs6_out_fragfail); 838 goto bad; 839 } else if (ip6->ip6_plen == 0) { /* jumbo payload cannot be fragmented */ 840 error = EMSGSIZE; 841 in6_ifstat_inc(ifp, ifs6_out_fragfail); 842 goto bad; 843 } else { 844 struct mbuf **mnext, *m_frgpart; 845 struct ip6_frag *ip6f; 846 u_int32_t id = htonl(ip6_id++); 847 u_char nextproto; 848 849 /* 850 * Too large for the destination or interface; 851 * fragment if possible. 852 * Must be able to put at least 8 bytes per fragment. 853 */ 854 hlen = unfragpartlen; 855 if (mtu > IPV6_MAXPACKET) 856 mtu = IPV6_MAXPACKET; 857 len = (mtu - hlen - sizeof(struct ip6_frag)) & ~7; 858 if (len < 8) { 859 error = EMSGSIZE; 860 in6_ifstat_inc(ifp, ifs6_out_fragfail); 861 goto bad; 862 } 863 864 mnext = &m->m_nextpkt; 865 866 /* 867 * Change the next header field of the last header in the 868 * unfragmentable part. 869 */ 870 if (exthdrs.ip6e_rthdr) { 871 nextproto = *mtod(exthdrs.ip6e_rthdr, u_char *); 872 *mtod(exthdrs.ip6e_rthdr, u_char *) = IPPROTO_FRAGMENT; 873 } else if (exthdrs.ip6e_dest1) { 874 nextproto = *mtod(exthdrs.ip6e_dest1, u_char *); 875 *mtod(exthdrs.ip6e_dest1, u_char *) = IPPROTO_FRAGMENT; 876 } else if (exthdrs.ip6e_hbh) { 877 nextproto = *mtod(exthdrs.ip6e_hbh, u_char *); 878 *mtod(exthdrs.ip6e_hbh, u_char *) = IPPROTO_FRAGMENT; 879 } else { 880 nextproto = ip6->ip6_nxt; 881 ip6->ip6_nxt = IPPROTO_FRAGMENT; 882 } 883 884 /* 885 * Loop through length of segment after first fragment, 886 * make new header and copy data of each part and link onto 887 * chain. 888 */ 889 m0 = m; 890 for (off = hlen; off < tlen; off += len) { 891 MGETHDR(m, M_DONTWAIT, MT_HEADER); 892 if (!m) { 893 error = ENOBUFS; 894 ip6stat.ip6s_odropped++; 895 goto sendorfree; 896 } 897 m->m_flags = m0->m_flags & M_COPYFLAGS; 898 *mnext = m; 899 mnext = &m->m_nextpkt; 900 m->m_data += max_linkhdr; 901 mhip6 = mtod(m, struct ip6_hdr *); 902 *mhip6 = *ip6; 903 m->m_len = sizeof(*mhip6); 904 error = ip6_insertfraghdr(m0, m, hlen, &ip6f); 905 if (error) { 906 ip6stat.ip6s_odropped++; 907 goto sendorfree; 908 } 909 ip6f->ip6f_offlg = htons((u_short)((off - hlen) & ~7)); 910 if (off + len >= tlen) 911 len = tlen - off; 912 else 913 ip6f->ip6f_offlg |= IP6F_MORE_FRAG; 914 mhip6->ip6_plen = htons((u_short)(len + hlen + 915 sizeof(*ip6f) - sizeof(struct ip6_hdr))); 916 if ((m_frgpart = m_copy(m0, off, len)) == 0) { 917 error = ENOBUFS; 918 ip6stat.ip6s_odropped++; 919 goto sendorfree; 920 } 921 m_cat(m, m_frgpart); 922 m->m_pkthdr.len = len + hlen + sizeof(*ip6f); 923 m->m_pkthdr.rcvif = (struct ifnet *)0; 924 ip6f->ip6f_reserved = 0; 925 ip6f->ip6f_ident = id; 926 ip6f->ip6f_nxt = nextproto; 927 ip6stat.ip6s_ofragments++; 928 in6_ifstat_inc(ifp, ifs6_out_fragcreat); 929 } 930 931 in6_ifstat_inc(ifp, ifs6_out_fragok); 932 } 933 934 /* 935 * Remove leading garbages. 936 */ 937 sendorfree: 938 m = m0->m_nextpkt; 939 m0->m_nextpkt = 0; 940 m_freem(m0); 941 for (m0 = m; m; m = m0) { 942 m0 = m->m_nextpkt; 943 m->m_nextpkt = 0; 944 if (error == 0) { 945 #ifdef IFA_STATS 946 struct in6_ifaddr *ia6; 947 ip6 = mtod(m, struct ip6_hdr *); 948 ia6 = in6_ifawithifp(ifp, &ip6->ip6_src); 949 if (ia6) { 950 /* 951 * Record statistics for this interface 952 * address. 953 */ 954 ia6->ia_ifa.ifa_data.ifad_outbytes += 955 m->m_pkthdr.len; 956 } 957 #endif 958 #ifdef IPSEC 959 /* clean ipsec history once it goes out of the node */ 960 ipsec_delaux(m); 961 #endif 962 error = nd6_output(ifp, origifp, m, dst, ro->ro_rt); 963 } else 964 m_freem(m); 965 } 966 967 if (error == 0) 968 ip6stat.ip6s_fragmented++; 969 970 done: 971 if (ro == &ip6route && ro->ro_rt) { /* brace necessary for RTFREE */ 972 RTFREE(ro->ro_rt); 973 } else if (ro_pmtu == &ip6route && ro_pmtu->ro_rt) { 974 RTFREE(ro_pmtu->ro_rt); 975 } 976 977 #ifdef IPSEC 978 if (sp != NULL) 979 key_freesp(sp); 980 #endif /* IPSEC */ 981 982 return (error); 983 984 freehdrs: 985 m_freem(exthdrs.ip6e_hbh); /* m_freem will check if mbuf is 0 */ 986 m_freem(exthdrs.ip6e_dest1); 987 m_freem(exthdrs.ip6e_rthdr); 988 m_freem(exthdrs.ip6e_dest2); 989 /* FALLTHROUGH */ 990 bad: 991 m_freem(m); 992 goto done; 993 } 994 995 static int 996 ip6_copyexthdr(mp, hdr, hlen) 997 struct mbuf **mp; 998 caddr_t hdr; 999 int hlen; 1000 { 1001 struct mbuf *m; 1002 1003 if (hlen > MCLBYTES) 1004 return (ENOBUFS); /* XXX */ 1005 1006 MGET(m, M_DONTWAIT, MT_DATA); 1007 if (!m) 1008 return (ENOBUFS); 1009 1010 if (hlen > MLEN) { 1011 MCLGET(m, M_DONTWAIT); 1012 if ((m->m_flags & M_EXT) == 0) { 1013 m_free(m); 1014 return (ENOBUFS); 1015 } 1016 } 1017 m->m_len = hlen; 1018 if (hdr) 1019 bcopy(hdr, mtod(m, caddr_t), hlen); 1020 1021 *mp = m; 1022 return (0); 1023 } 1024 1025 /* 1026 * Insert jumbo payload option. 1027 */ 1028 static int 1029 ip6_insert_jumboopt(exthdrs, plen) 1030 struct ip6_exthdrs *exthdrs; 1031 u_int32_t plen; 1032 { 1033 struct mbuf *mopt; 1034 u_int8_t *optbuf; 1035 u_int32_t v; 1036 1037 #define JUMBOOPTLEN 8 /* length of jumbo payload option and padding */ 1038 1039 /* 1040 * If there is no hop-by-hop options header, allocate new one. 1041 * If there is one but it doesn't have enough space to store the 1042 * jumbo payload option, allocate a cluster to store the whole options. 1043 * Otherwise, use it to store the options. 1044 */ 1045 if (exthdrs->ip6e_hbh == 0) { 1046 MGET(mopt, M_DONTWAIT, MT_DATA); 1047 if (mopt == 0) 1048 return (ENOBUFS); 1049 mopt->m_len = JUMBOOPTLEN; 1050 optbuf = mtod(mopt, u_int8_t *); 1051 optbuf[1] = 0; /* = ((JUMBOOPTLEN) >> 3) - 1 */ 1052 exthdrs->ip6e_hbh = mopt; 1053 } else { 1054 struct ip6_hbh *hbh; 1055 1056 mopt = exthdrs->ip6e_hbh; 1057 if (M_TRAILINGSPACE(mopt) < JUMBOOPTLEN) { 1058 /* 1059 * XXX assumption: 1060 * - exthdrs->ip6e_hbh is not referenced from places 1061 * other than exthdrs. 1062 * - exthdrs->ip6e_hbh is not an mbuf chain. 1063 */ 1064 int oldoptlen = mopt->m_len; 1065 struct mbuf *n; 1066 1067 /* 1068 * XXX: give up if the whole (new) hbh header does 1069 * not fit even in an mbuf cluster. 1070 */ 1071 if (oldoptlen + JUMBOOPTLEN > MCLBYTES) 1072 return (ENOBUFS); 1073 1074 /* 1075 * As a consequence, we must always prepare a cluster 1076 * at this point. 1077 */ 1078 MGET(n, M_DONTWAIT, MT_DATA); 1079 if (n) { 1080 MCLGET(n, M_DONTWAIT); 1081 if ((n->m_flags & M_EXT) == 0) { 1082 m_freem(n); 1083 n = NULL; 1084 } 1085 } 1086 if (!n) 1087 return (ENOBUFS); 1088 n->m_len = oldoptlen + JUMBOOPTLEN; 1089 bcopy(mtod(mopt, caddr_t), mtod(n, caddr_t), 1090 oldoptlen); 1091 optbuf = mtod(n, u_int8_t *) + oldoptlen; 1092 m_freem(mopt); 1093 mopt = exthdrs->ip6e_hbh = n; 1094 } else { 1095 optbuf = mtod(mopt, u_int8_t *) + mopt->m_len; 1096 mopt->m_len += JUMBOOPTLEN; 1097 } 1098 optbuf[0] = IP6OPT_PADN; 1099 optbuf[1] = 0; 1100 1101 /* 1102 * Adjust the header length according to the pad and 1103 * the jumbo payload option. 1104 */ 1105 hbh = mtod(mopt, struct ip6_hbh *); 1106 hbh->ip6h_len += (JUMBOOPTLEN >> 3); 1107 } 1108 1109 /* fill in the option. */ 1110 optbuf[2] = IP6OPT_JUMBO; 1111 optbuf[3] = 4; 1112 v = (u_int32_t)htonl(plen + JUMBOOPTLEN); 1113 bcopy(&v, &optbuf[4], sizeof(u_int32_t)); 1114 1115 /* finally, adjust the packet header length */ 1116 exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN; 1117 1118 return (0); 1119 #undef JUMBOOPTLEN 1120 } 1121 1122 /* 1123 * Insert fragment header and copy unfragmentable header portions. 1124 */ 1125 static int 1126 ip6_insertfraghdr(m0, m, hlen, frghdrp) 1127 struct mbuf *m0, *m; 1128 int hlen; 1129 struct ip6_frag **frghdrp; 1130 { 1131 struct mbuf *n, *mlast; 1132 1133 if (hlen > sizeof(struct ip6_hdr)) { 1134 n = m_copym(m0, sizeof(struct ip6_hdr), 1135 hlen - sizeof(struct ip6_hdr), M_DONTWAIT); 1136 if (n == 0) 1137 return (ENOBUFS); 1138 m->m_next = n; 1139 } else 1140 n = m; 1141 1142 /* Search for the last mbuf of unfragmentable part. */ 1143 for (mlast = n; mlast->m_next; mlast = mlast->m_next) 1144 ; 1145 1146 if ((mlast->m_flags & M_EXT) == 0 && 1147 M_TRAILINGSPACE(mlast) >= sizeof(struct ip6_frag)) { 1148 /* use the trailing space of the last mbuf for the fragment hdr */ 1149 *frghdrp = (struct ip6_frag *)(mtod(mlast, caddr_t) + 1150 mlast->m_len); 1151 mlast->m_len += sizeof(struct ip6_frag); 1152 m->m_pkthdr.len += sizeof(struct ip6_frag); 1153 } else { 1154 /* allocate a new mbuf for the fragment header */ 1155 struct mbuf *mfrg; 1156 1157 MGET(mfrg, M_DONTWAIT, MT_DATA); 1158 if (mfrg == 0) 1159 return (ENOBUFS); 1160 mfrg->m_len = sizeof(struct ip6_frag); 1161 *frghdrp = mtod(mfrg, struct ip6_frag *); 1162 mlast->m_next = mfrg; 1163 } 1164 1165 return (0); 1166 } 1167 1168 static int 1169 ip6_getpmtu(ro_pmtu, ro, ifp, dst, mtup) 1170 struct route_in6 *ro_pmtu, *ro; 1171 struct ifnet *ifp; 1172 struct in6_addr *dst; 1173 u_long *mtup; 1174 { 1175 u_int32_t mtu = 0; 1176 int error = 0; 1177 1178 if (ro_pmtu != ro) { 1179 /* The first hop and the final destination may differ. */ 1180 struct sockaddr_in6 *sa6_dst = 1181 (struct sockaddr_in6 *)&ro_pmtu->ro_dst; 1182 if (ro_pmtu->ro_rt && 1183 ((ro_pmtu->ro_rt->rt_flags & RTF_UP) == 0 || 1184 !IN6_ARE_ADDR_EQUAL(&sa6_dst->sin6_addr, dst))) { 1185 RTFREE(ro_pmtu->ro_rt); 1186 ro_pmtu->ro_rt = (struct rtentry *)0; 1187 } 1188 if (ro_pmtu->ro_rt == 0) { 1189 bzero(sa6_dst, sizeof(*sa6_dst)); 1190 sa6_dst->sin6_family = AF_INET6; 1191 sa6_dst->sin6_len = sizeof(struct sockaddr_in6); 1192 sa6_dst->sin6_addr = *dst; 1193 1194 rtalloc((struct route *)ro_pmtu); 1195 } 1196 } 1197 if (ro_pmtu->ro_rt) { 1198 u_int32_t ifmtu; 1199 1200 if (ifp == NULL) 1201 ifp = ro_pmtu->ro_rt->rt_ifp; 1202 ifmtu = IN6_LINKMTU(ifp); 1203 mtu = ro_pmtu->ro_rt->rt_rmx.rmx_mtu; 1204 if (mtu == 0) 1205 mtu = ifmtu; 1206 else if (mtu > ifmtu) { 1207 /* 1208 * The MTU on the route is larger than the MTU on 1209 * the interface! This shouldn't happen, unless the 1210 * MTU of the interface has been changed after the 1211 * interface was brought up. Change the MTU in the 1212 * route to match the interface MTU (as long as the 1213 * field isn't locked). 1214 * 1215 * if MTU on the route is 0, we need to fix the MTU. 1216 * this case happens with path MTU discovery timeouts. 1217 */ 1218 mtu = ifmtu; 1219 if (!(ro_pmtu->ro_rt->rt_rmx.rmx_locks & RTV_MTU)) 1220 ro_pmtu->ro_rt->rt_rmx.rmx_mtu = mtu; 1221 } 1222 } else if (ifp) { 1223 mtu = IN6_LINKMTU(ifp); 1224 } else 1225 error = EHOSTUNREACH; /* XXX */ 1226 1227 *mtup = mtu; 1228 return (error); 1229 } 1230 1231 /* 1232 * IP6 socket option processing. 1233 */ 1234 int 1235 ip6_ctloutput(op, so, level, optname, mp) 1236 int op; 1237 struct socket *so; 1238 int level, optname; 1239 struct mbuf **mp; 1240 { 1241 struct in6pcb *in6p = sotoin6pcb(so); 1242 struct mbuf *m = *mp; 1243 int optval = 0; 1244 int error = 0; 1245 struct proc *p = curproc; /* XXX */ 1246 1247 if (level == IPPROTO_IPV6) { 1248 switch (op) { 1249 case PRCO_SETOPT: 1250 switch (optname) { 1251 case IPV6_PKTOPTIONS: 1252 /* m is freed in ip6_pcbopts */ 1253 return (ip6_pcbopts(&in6p->in6p_outputopts, 1254 m, so)); 1255 case IPV6_HOPOPTS: 1256 case IPV6_DSTOPTS: 1257 if (p == 0 || suser(p->p_ucred, &p->p_acflag)) { 1258 error = EPERM; 1259 break; 1260 } 1261 /* FALLTHROUGH */ 1262 case IPV6_UNICAST_HOPS: 1263 case IPV6_RECVOPTS: 1264 case IPV6_RECVRETOPTS: 1265 case IPV6_RECVDSTADDR: 1266 case IPV6_PKTINFO: 1267 case IPV6_HOPLIMIT: 1268 case IPV6_RTHDR: 1269 case IPV6_FAITH: 1270 case IPV6_V6ONLY: 1271 if (!m || m->m_len != sizeof(int)) { 1272 error = EINVAL; 1273 break; 1274 } 1275 optval = *mtod(m, int *); 1276 switch (optname) { 1277 1278 case IPV6_UNICAST_HOPS: 1279 if (optval < -1 || optval >= 256) 1280 error = EINVAL; 1281 else { 1282 /* -1 = kernel default */ 1283 in6p->in6p_hops = optval; 1284 } 1285 break; 1286 #define OPTSET(bit) \ 1287 do { \ 1288 if (optval) \ 1289 in6p->in6p_flags |= (bit); \ 1290 else \ 1291 in6p->in6p_flags &= ~(bit); \ 1292 } while (/*CONSTCOND*/ 0) 1293 1294 case IPV6_RECVOPTS: 1295 OPTSET(IN6P_RECVOPTS); 1296 break; 1297 1298 case IPV6_RECVRETOPTS: 1299 OPTSET(IN6P_RECVRETOPTS); 1300 break; 1301 1302 case IPV6_RECVDSTADDR: 1303 OPTSET(IN6P_RECVDSTADDR); 1304 break; 1305 1306 case IPV6_PKTINFO: 1307 OPTSET(IN6P_PKTINFO); 1308 break; 1309 1310 case IPV6_HOPLIMIT: 1311 OPTSET(IN6P_HOPLIMIT); 1312 break; 1313 1314 case IPV6_HOPOPTS: 1315 OPTSET(IN6P_HOPOPTS); 1316 break; 1317 1318 case IPV6_DSTOPTS: 1319 OPTSET(IN6P_DSTOPTS); 1320 break; 1321 1322 case IPV6_RTHDR: 1323 OPTSET(IN6P_RTHDR); 1324 break; 1325 1326 case IPV6_FAITH: 1327 OPTSET(IN6P_FAITH); 1328 break; 1329 1330 case IPV6_V6ONLY: 1331 /* 1332 * make setsockopt(IPV6_V6ONLY) 1333 * available only prior to bind(2). 1334 * see ipng mailing list, Jun 22 2001. 1335 */ 1336 if (in6p->in6p_lport || 1337 !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr)) { 1338 error = EINVAL; 1339 break; 1340 } 1341 #ifdef INET6_BINDV6ONLY 1342 if (!optval) 1343 error = EINVAL; 1344 #else 1345 OPTSET(IN6P_IPV6_V6ONLY); 1346 #endif 1347 break; 1348 } 1349 break; 1350 #undef OPTSET 1351 1352 case IPV6_MULTICAST_IF: 1353 case IPV6_MULTICAST_HOPS: 1354 case IPV6_MULTICAST_LOOP: 1355 case IPV6_JOIN_GROUP: 1356 case IPV6_LEAVE_GROUP: 1357 error = ip6_setmoptions(optname, 1358 &in6p->in6p_moptions, m); 1359 break; 1360 1361 case IPV6_PORTRANGE: 1362 optval = *mtod(m, int *); 1363 1364 switch (optval) { 1365 case IPV6_PORTRANGE_DEFAULT: 1366 in6p->in6p_flags &= ~(IN6P_LOWPORT); 1367 in6p->in6p_flags &= ~(IN6P_HIGHPORT); 1368 break; 1369 1370 case IPV6_PORTRANGE_HIGH: 1371 in6p->in6p_flags &= ~(IN6P_LOWPORT); 1372 in6p->in6p_flags |= IN6P_HIGHPORT; 1373 break; 1374 1375 case IPV6_PORTRANGE_LOW: 1376 in6p->in6p_flags &= ~(IN6P_HIGHPORT); 1377 in6p->in6p_flags |= IN6P_LOWPORT; 1378 break; 1379 1380 default: 1381 error = EINVAL; 1382 break; 1383 } 1384 break; 1385 1386 #ifdef IPSEC 1387 case IPV6_IPSEC_POLICY: 1388 { 1389 caddr_t req = NULL; 1390 size_t len = 0; 1391 1392 int priv = 0; 1393 if (p == 0 || suser(p->p_ucred, &p->p_acflag)) 1394 priv = 0; 1395 else 1396 priv = 1; 1397 if (m) { 1398 req = mtod(m, caddr_t); 1399 len = m->m_len; 1400 } 1401 error = ipsec6_set_policy(in6p, 1402 optname, req, len, priv); 1403 } 1404 break; 1405 #endif /* IPSEC */ 1406 1407 default: 1408 error = ENOPROTOOPT; 1409 break; 1410 } 1411 if (m) 1412 (void)m_free(m); 1413 break; 1414 1415 case PRCO_GETOPT: 1416 switch (optname) { 1417 1418 case IPV6_OPTIONS: 1419 case IPV6_RETOPTS: 1420 error = ENOPROTOOPT; 1421 break; 1422 1423 case IPV6_PKTOPTIONS: 1424 if (in6p->in6p_options) { 1425 *mp = m_copym(in6p->in6p_options, 0, 1426 M_COPYALL, M_WAIT); 1427 } else { 1428 *mp = m_get(M_WAIT, MT_SOOPTS); 1429 (*mp)->m_len = 0; 1430 } 1431 break; 1432 1433 case IPV6_HOPOPTS: 1434 case IPV6_DSTOPTS: 1435 if (p == 0 || suser(p->p_ucred, &p->p_acflag)) { 1436 error = EPERM; 1437 break; 1438 } 1439 /* FALLTHROUGH */ 1440 case IPV6_UNICAST_HOPS: 1441 case IPV6_RECVOPTS: 1442 case IPV6_RECVRETOPTS: 1443 case IPV6_RECVDSTADDR: 1444 case IPV6_PORTRANGE: 1445 case IPV6_PKTINFO: 1446 case IPV6_HOPLIMIT: 1447 case IPV6_RTHDR: 1448 case IPV6_FAITH: 1449 case IPV6_V6ONLY: 1450 *mp = m = m_get(M_WAIT, MT_SOOPTS); 1451 m->m_len = sizeof(int); 1452 switch (optname) { 1453 1454 case IPV6_UNICAST_HOPS: 1455 optval = in6p->in6p_hops; 1456 break; 1457 1458 #define OPTBIT(bit) (in6p->in6p_flags & bit ? 1 : 0) 1459 1460 case IPV6_RECVOPTS: 1461 optval = OPTBIT(IN6P_RECVOPTS); 1462 break; 1463 1464 case IPV6_RECVRETOPTS: 1465 optval = OPTBIT(IN6P_RECVRETOPTS); 1466 break; 1467 1468 case IPV6_RECVDSTADDR: 1469 optval = OPTBIT(IN6P_RECVDSTADDR); 1470 break; 1471 1472 case IPV6_PORTRANGE: 1473 { 1474 int flags; 1475 flags = in6p->in6p_flags; 1476 if (flags & IN6P_HIGHPORT) 1477 optval = IPV6_PORTRANGE_HIGH; 1478 else if (flags & IN6P_LOWPORT) 1479 optval = IPV6_PORTRANGE_LOW; 1480 else 1481 optval = 0; 1482 break; 1483 } 1484 1485 case IPV6_PKTINFO: 1486 optval = OPTBIT(IN6P_PKTINFO); 1487 break; 1488 1489 case IPV6_HOPLIMIT: 1490 optval = OPTBIT(IN6P_HOPLIMIT); 1491 break; 1492 1493 case IPV6_HOPOPTS: 1494 optval = OPTBIT(IN6P_HOPOPTS); 1495 break; 1496 1497 case IPV6_DSTOPTS: 1498 optval = OPTBIT(IN6P_DSTOPTS); 1499 break; 1500 1501 case IPV6_RTHDR: 1502 optval = OPTBIT(IN6P_RTHDR); 1503 break; 1504 1505 case IPV6_FAITH: 1506 optval = OPTBIT(IN6P_FAITH); 1507 break; 1508 1509 case IPV6_V6ONLY: 1510 optval = OPTBIT(IN6P_IPV6_V6ONLY); 1511 break; 1512 } 1513 *mtod(m, int *) = optval; 1514 break; 1515 1516 case IPV6_MULTICAST_IF: 1517 case IPV6_MULTICAST_HOPS: 1518 case IPV6_MULTICAST_LOOP: 1519 case IPV6_JOIN_GROUP: 1520 case IPV6_LEAVE_GROUP: 1521 error = ip6_getmoptions(optname, in6p->in6p_moptions, mp); 1522 break; 1523 1524 #ifdef IPSEC 1525 case IPV6_IPSEC_POLICY: 1526 { 1527 caddr_t req = NULL; 1528 size_t len = 0; 1529 1530 if (m) { 1531 req = mtod(m, caddr_t); 1532 len = m->m_len; 1533 } 1534 error = ipsec6_get_policy(in6p, req, len, mp); 1535 break; 1536 } 1537 #endif /* IPSEC */ 1538 1539 default: 1540 error = ENOPROTOOPT; 1541 break; 1542 } 1543 break; 1544 } 1545 } else { 1546 error = EINVAL; 1547 if (op == PRCO_SETOPT && *mp) 1548 (void)m_free(*mp); 1549 } 1550 return (error); 1551 } 1552 1553 int 1554 ip6_raw_ctloutput(op, so, level, optname, mp) 1555 int op; 1556 struct socket *so; 1557 int level, optname; 1558 struct mbuf **mp; 1559 { 1560 int error = 0, optval, optlen; 1561 const int icmp6off = offsetof(struct icmp6_hdr, icmp6_cksum); 1562 struct in6pcb *in6p = sotoin6pcb(so); 1563 struct mbuf *m = *mp; 1564 1565 optlen = m ? m->m_len : 0; 1566 1567 if (level != IPPROTO_IPV6) { 1568 if (op == PRCO_SETOPT && *mp) 1569 (void)m_free(*mp); 1570 return (EINVAL); 1571 } 1572 1573 switch (optname) { 1574 case IPV6_CHECKSUM: 1575 /* 1576 * For ICMPv6 sockets, no modification allowed for checksum 1577 * offset, permit "no change" values to help existing apps. 1578 * 1579 * XXX 2292bis says: "An attempt to set IPV6_CHECKSUM 1580 * for an ICMPv6 socket will fail." 1581 * The current behavior does not meet 2292bis. 1582 */ 1583 switch (op) { 1584 case PRCO_SETOPT: 1585 if (optlen != sizeof(int)) { 1586 error = EINVAL; 1587 break; 1588 } 1589 optval = *mtod(m, int *); 1590 if ((optval % 2) != 0) { 1591 /* the API assumes even offset values */ 1592 error = EINVAL; 1593 } else if (so->so_proto->pr_protocol == 1594 IPPROTO_ICMPV6) { 1595 if (optval != icmp6off) 1596 error = EINVAL; 1597 } else 1598 in6p->in6p_cksum = optval; 1599 break; 1600 1601 case PRCO_GETOPT: 1602 if (so->so_proto->pr_protocol == IPPROTO_ICMPV6) 1603 optval = icmp6off; 1604 else 1605 optval = in6p->in6p_cksum; 1606 1607 *mp = m = m_get(M_WAIT, MT_SOOPTS); 1608 m->m_len = sizeof(int); 1609 *mtod(m, int *) = optval; 1610 break; 1611 1612 default: 1613 error = EINVAL; 1614 break; 1615 } 1616 break; 1617 1618 default: 1619 error = ENOPROTOOPT; 1620 break; 1621 } 1622 1623 if (op == PRCO_SETOPT && m) 1624 (void)m_free(m); 1625 1626 return (error); 1627 } 1628 1629 /* 1630 * Set up IP6 options in pcb for insertion in output packets. 1631 * Store in mbuf with pointer in pcbopt, adding pseudo-option 1632 * with destination address if source routed. 1633 */ 1634 static int 1635 ip6_pcbopts(pktopt, m, so) 1636 struct ip6_pktopts **pktopt; 1637 struct mbuf *m; 1638 struct socket *so; 1639 { 1640 struct ip6_pktopts *opt = *pktopt; 1641 int error = 0; 1642 struct proc *p = curproc; /* XXX */ 1643 int priv = 0; 1644 1645 /* turn off any old options. */ 1646 if (opt) { 1647 if (opt->ip6po_m) 1648 (void)m_free(opt->ip6po_m); 1649 } else 1650 opt = malloc(sizeof(*opt), M_IP6OPT, M_WAITOK); 1651 *pktopt = 0; 1652 1653 if (!m || m->m_len == 0) { 1654 /* 1655 * Only turning off any previous options. 1656 */ 1657 free(opt, M_IP6OPT); 1658 if (m) 1659 (void)m_free(m); 1660 return (0); 1661 } 1662 1663 /* set options specified by user. */ 1664 if (p && !suser(p->p_ucred, &p->p_acflag)) 1665 priv = 1; 1666 if ((error = ip6_setpktoptions(m, opt, priv)) != 0) { 1667 (void)m_free(m); 1668 free(opt, M_IP6OPT); 1669 return (error); 1670 } 1671 *pktopt = opt; 1672 return (0); 1673 } 1674 1675 /* 1676 * Set the IP6 multicast options in response to user setsockopt(). 1677 */ 1678 static int 1679 ip6_setmoptions(optname, im6op, m) 1680 int optname; 1681 struct ip6_moptions **im6op; 1682 struct mbuf *m; 1683 { 1684 int error = 0; 1685 u_int loop, ifindex; 1686 struct ipv6_mreq *mreq; 1687 struct ifnet *ifp; 1688 struct ip6_moptions *im6o = *im6op; 1689 struct route_in6 ro; 1690 struct sockaddr_in6 *dst; 1691 struct in6_multi_mship *imm; 1692 struct proc *p = curproc; /* XXX */ 1693 1694 if (im6o == NULL) { 1695 /* 1696 * No multicast option buffer attached to the pcb; 1697 * allocate one and initialize to default values. 1698 */ 1699 im6o = (struct ip6_moptions *) 1700 malloc(sizeof(*im6o), M_IPMOPTS, M_WAITOK); 1701 1702 if (im6o == NULL) 1703 return (ENOBUFS); 1704 *im6op = im6o; 1705 im6o->im6o_multicast_ifp = NULL; 1706 im6o->im6o_multicast_hlim = ip6_defmcasthlim; 1707 im6o->im6o_multicast_loop = IPV6_DEFAULT_MULTICAST_LOOP; 1708 LIST_INIT(&im6o->im6o_memberships); 1709 } 1710 1711 switch (optname) { 1712 1713 case IPV6_MULTICAST_IF: 1714 /* 1715 * Select the interface for outgoing multicast packets. 1716 */ 1717 if (m == NULL || m->m_len != sizeof(u_int)) { 1718 error = EINVAL; 1719 break; 1720 } 1721 bcopy(mtod(m, u_int *), &ifindex, sizeof(ifindex)); 1722 if (ifindex < 0 || if_index < ifindex) { 1723 error = ENXIO; /* XXX EINVAL? */ 1724 break; 1725 } 1726 ifp = ifindex2ifnet[ifindex]; 1727 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) { 1728 error = EADDRNOTAVAIL; 1729 break; 1730 } 1731 im6o->im6o_multicast_ifp = ifp; 1732 break; 1733 1734 case IPV6_MULTICAST_HOPS: 1735 { 1736 /* 1737 * Set the IP6 hoplimit for outgoing multicast packets. 1738 */ 1739 int optval; 1740 if (m == NULL || m->m_len != sizeof(int)) { 1741 error = EINVAL; 1742 break; 1743 } 1744 bcopy(mtod(m, u_int *), &optval, sizeof(optval)); 1745 if (optval < -1 || optval >= 256) 1746 error = EINVAL; 1747 else if (optval == -1) 1748 im6o->im6o_multicast_hlim = ip6_defmcasthlim; 1749 else 1750 im6o->im6o_multicast_hlim = optval; 1751 break; 1752 } 1753 1754 case IPV6_MULTICAST_LOOP: 1755 /* 1756 * Set the loopback flag for outgoing multicast packets. 1757 * Must be zero or one. 1758 */ 1759 if (m == NULL || m->m_len != sizeof(u_int)) { 1760 error = EINVAL; 1761 break; 1762 } 1763 bcopy(mtod(m, u_int *), &loop, sizeof(loop)); 1764 if (loop > 1) { 1765 error = EINVAL; 1766 break; 1767 } 1768 im6o->im6o_multicast_loop = loop; 1769 break; 1770 1771 case IPV6_JOIN_GROUP: 1772 /* 1773 * Add a multicast group membership. 1774 * Group must be a valid IP6 multicast address. 1775 */ 1776 if (m == NULL || m->m_len != sizeof(struct ipv6_mreq)) { 1777 error = EINVAL; 1778 break; 1779 } 1780 mreq = mtod(m, struct ipv6_mreq *); 1781 if (IN6_IS_ADDR_UNSPECIFIED(&mreq->ipv6mr_multiaddr)) { 1782 /* 1783 * We use the unspecified address to specify to accept 1784 * all multicast addresses. Only super user is allowed 1785 * to do this. 1786 */ 1787 if (suser(p->p_ucred, &p->p_acflag)) 1788 { 1789 error = EACCES; 1790 break; 1791 } 1792 } else if (!IN6_IS_ADDR_MULTICAST(&mreq->ipv6mr_multiaddr)) { 1793 error = EINVAL; 1794 break; 1795 } 1796 1797 /* 1798 * If the interface is specified, validate it. 1799 */ 1800 if (mreq->ipv6mr_interface < 0 1801 || if_index < mreq->ipv6mr_interface) { 1802 error = ENXIO; /* XXX EINVAL? */ 1803 break; 1804 } 1805 /* 1806 * If no interface was explicitly specified, choose an 1807 * appropriate one according to the given multicast address. 1808 */ 1809 if (mreq->ipv6mr_interface == 0) { 1810 /* 1811 * If the multicast address is in node-local scope, 1812 * the interface should be a loopback interface. 1813 * Otherwise, look up the routing table for the 1814 * address, and choose the outgoing interface. 1815 * XXX: is it a good approach? 1816 */ 1817 if (IN6_IS_ADDR_MC_NODELOCAL(&mreq->ipv6mr_multiaddr)) { 1818 ifp = &loif[0]; 1819 } else { 1820 ro.ro_rt = NULL; 1821 dst = (struct sockaddr_in6 *)&ro.ro_dst; 1822 bzero(dst, sizeof(*dst)); 1823 dst->sin6_len = sizeof(struct sockaddr_in6); 1824 dst->sin6_family = AF_INET6; 1825 dst->sin6_addr = mreq->ipv6mr_multiaddr; 1826 rtalloc((struct route *)&ro); 1827 if (ro.ro_rt == NULL) { 1828 error = EADDRNOTAVAIL; 1829 break; 1830 } 1831 ifp = ro.ro_rt->rt_ifp; 1832 rtfree(ro.ro_rt); 1833 } 1834 } else 1835 ifp = ifindex2ifnet[mreq->ipv6mr_interface]; 1836 1837 /* 1838 * See if we found an interface, and confirm that it 1839 * supports multicast 1840 */ 1841 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) { 1842 error = EADDRNOTAVAIL; 1843 break; 1844 } 1845 /* 1846 * Put interface index into the multicast address, 1847 * if the address has link-local scope. 1848 */ 1849 if (IN6_IS_ADDR_MC_LINKLOCAL(&mreq->ipv6mr_multiaddr)) { 1850 mreq->ipv6mr_multiaddr.s6_addr16[1] = 1851 htons(mreq->ipv6mr_interface); 1852 } 1853 /* 1854 * See if the membership already exists. 1855 */ 1856 for (imm = im6o->im6o_memberships.lh_first; 1857 imm != NULL; imm = imm->i6mm_chain.le_next) 1858 if (imm->i6mm_maddr->in6m_ifp == ifp && 1859 IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr, 1860 &mreq->ipv6mr_multiaddr)) 1861 break; 1862 if (imm != NULL) { 1863 error = EADDRINUSE; 1864 break; 1865 } 1866 /* 1867 * Everything looks good; add a new record to the multicast 1868 * address list for the given interface. 1869 */ 1870 imm = in6_joingroup(ifp, &mreq->ipv6mr_multiaddr, &error); 1871 if (!imm) 1872 break; 1873 LIST_INSERT_HEAD(&im6o->im6o_memberships, imm, i6mm_chain); 1874 break; 1875 1876 case IPV6_LEAVE_GROUP: 1877 /* 1878 * Drop a multicast group membership. 1879 * Group must be a valid IP6 multicast address. 1880 */ 1881 if (m == NULL || m->m_len != sizeof(struct ipv6_mreq)) { 1882 error = EINVAL; 1883 break; 1884 } 1885 mreq = mtod(m, struct ipv6_mreq *); 1886 if (IN6_IS_ADDR_UNSPECIFIED(&mreq->ipv6mr_multiaddr)) { 1887 if (suser(p->p_ucred, &p->p_acflag)) 1888 { 1889 error = EACCES; 1890 break; 1891 } 1892 } else if (!IN6_IS_ADDR_MULTICAST(&mreq->ipv6mr_multiaddr)) { 1893 error = EINVAL; 1894 break; 1895 } 1896 /* 1897 * If an interface address was specified, get a pointer 1898 * to its ifnet structure. 1899 */ 1900 if (mreq->ipv6mr_interface < 0 1901 || if_index < mreq->ipv6mr_interface) { 1902 error = ENXIO; /* XXX EINVAL? */ 1903 break; 1904 } 1905 ifp = ifindex2ifnet[mreq->ipv6mr_interface]; 1906 /* 1907 * Put interface index into the multicast address, 1908 * if the address has link-local scope. 1909 */ 1910 if (IN6_IS_ADDR_MC_LINKLOCAL(&mreq->ipv6mr_multiaddr)) { 1911 mreq->ipv6mr_multiaddr.s6_addr16[1] = 1912 htons(mreq->ipv6mr_interface); 1913 } 1914 /* 1915 * Find the membership in the membership list. 1916 */ 1917 for (imm = im6o->im6o_memberships.lh_first; 1918 imm != NULL; imm = imm->i6mm_chain.le_next) { 1919 if ((ifp == NULL || imm->i6mm_maddr->in6m_ifp == ifp) && 1920 IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr, 1921 &mreq->ipv6mr_multiaddr)) 1922 break; 1923 } 1924 if (imm == NULL) { 1925 /* Unable to resolve interface */ 1926 error = EADDRNOTAVAIL; 1927 break; 1928 } 1929 /* 1930 * Give up the multicast address record to which the 1931 * membership points. 1932 */ 1933 LIST_REMOVE(imm, i6mm_chain); 1934 in6_leavegroup(imm); 1935 break; 1936 1937 default: 1938 error = EOPNOTSUPP; 1939 break; 1940 } 1941 1942 /* 1943 * If all options have default values, no need to keep the mbuf. 1944 */ 1945 if (im6o->im6o_multicast_ifp == NULL && 1946 im6o->im6o_multicast_hlim == ip6_defmcasthlim && 1947 im6o->im6o_multicast_loop == IPV6_DEFAULT_MULTICAST_LOOP && 1948 im6o->im6o_memberships.lh_first == NULL) { 1949 free(*im6op, M_IPMOPTS); 1950 *im6op = NULL; 1951 } 1952 1953 return (error); 1954 } 1955 1956 /* 1957 * Return the IP6 multicast options in response to user getsockopt(). 1958 */ 1959 static int 1960 ip6_getmoptions(optname, im6o, mp) 1961 int optname; 1962 struct ip6_moptions *im6o; 1963 struct mbuf **mp; 1964 { 1965 u_int *hlim, *loop, *ifindex; 1966 1967 *mp = m_get(M_WAIT, MT_SOOPTS); 1968 1969 switch (optname) { 1970 1971 case IPV6_MULTICAST_IF: 1972 ifindex = mtod(*mp, u_int *); 1973 (*mp)->m_len = sizeof(u_int); 1974 if (im6o == NULL || im6o->im6o_multicast_ifp == NULL) 1975 *ifindex = 0; 1976 else 1977 *ifindex = im6o->im6o_multicast_ifp->if_index; 1978 return (0); 1979 1980 case IPV6_MULTICAST_HOPS: 1981 hlim = mtod(*mp, u_int *); 1982 (*mp)->m_len = sizeof(u_int); 1983 if (im6o == NULL) 1984 *hlim = ip6_defmcasthlim; 1985 else 1986 *hlim = im6o->im6o_multicast_hlim; 1987 return (0); 1988 1989 case IPV6_MULTICAST_LOOP: 1990 loop = mtod(*mp, u_int *); 1991 (*mp)->m_len = sizeof(u_int); 1992 if (im6o == NULL) 1993 *loop = ip6_defmcasthlim; 1994 else 1995 *loop = im6o->im6o_multicast_loop; 1996 return (0); 1997 1998 default: 1999 return (EOPNOTSUPP); 2000 } 2001 } 2002 2003 /* 2004 * Discard the IP6 multicast options. 2005 */ 2006 void 2007 ip6_freemoptions(im6o) 2008 struct ip6_moptions *im6o; 2009 { 2010 struct in6_multi_mship *imm; 2011 2012 if (im6o == NULL) 2013 return; 2014 2015 while ((imm = im6o->im6o_memberships.lh_first) != NULL) { 2016 LIST_REMOVE(imm, i6mm_chain); 2017 in6_leavegroup(imm); 2018 } 2019 free(im6o, M_IPMOPTS); 2020 } 2021 2022 /* 2023 * Set IPv6 outgoing packet options based on advanced API. 2024 */ 2025 int 2026 ip6_setpktoptions(control, opt, priv) 2027 struct mbuf *control; 2028 struct ip6_pktopts *opt; 2029 int priv; 2030 { 2031 struct cmsghdr *cm = 0; 2032 2033 if (control == 0 || opt == 0) 2034 return (EINVAL); 2035 2036 bzero(opt, sizeof(*opt)); 2037 opt->ip6po_hlim = -1; /* -1 means to use default hop limit */ 2038 2039 /* 2040 * XXX: Currently, we assume all the optional information is stored 2041 * in a single mbuf. 2042 */ 2043 if (control->m_next) 2044 return (EINVAL); 2045 2046 opt->ip6po_m = control; 2047 2048 for (; control->m_len; control->m_data += CMSG_ALIGN(cm->cmsg_len), 2049 control->m_len -= CMSG_ALIGN(cm->cmsg_len)) { 2050 cm = mtod(control, struct cmsghdr *); 2051 if (cm->cmsg_len == 0 || cm->cmsg_len > control->m_len) 2052 return (EINVAL); 2053 if (cm->cmsg_level != IPPROTO_IPV6) 2054 continue; 2055 2056 switch (cm->cmsg_type) { 2057 case IPV6_PKTINFO: 2058 if (cm->cmsg_len != CMSG_LEN(sizeof(struct in6_pktinfo))) 2059 return (EINVAL); 2060 opt->ip6po_pktinfo = (struct in6_pktinfo *)CMSG_DATA(cm); 2061 if (opt->ip6po_pktinfo->ipi6_ifindex && 2062 IN6_IS_ADDR_LINKLOCAL(&opt->ip6po_pktinfo->ipi6_addr)) 2063 opt->ip6po_pktinfo->ipi6_addr.s6_addr16[1] = 2064 htons(opt->ip6po_pktinfo->ipi6_ifindex); 2065 2066 if (opt->ip6po_pktinfo->ipi6_ifindex > if_index || 2067 opt->ip6po_pktinfo->ipi6_ifindex < 0) { 2068 return (ENXIO); 2069 } 2070 2071 /* 2072 * Check if the requested source address is indeed a 2073 * unicast address assigned to the node, and can be 2074 * used as the packet's source address. 2075 */ 2076 if (!IN6_IS_ADDR_UNSPECIFIED(&opt->ip6po_pktinfo->ipi6_addr)) { 2077 struct ifaddr *ia; 2078 struct in6_ifaddr *ia6; 2079 struct sockaddr_in6 sin6; 2080 2081 bzero(&sin6, sizeof(sin6)); 2082 sin6.sin6_len = sizeof(sin6); 2083 sin6.sin6_family = AF_INET6; 2084 sin6.sin6_addr = 2085 opt->ip6po_pktinfo->ipi6_addr; 2086 ia = ifa_ifwithaddr(sin6tosa(&sin6)); 2087 if (ia == NULL || 2088 (opt->ip6po_pktinfo->ipi6_ifindex && 2089 (ia->ifa_ifp->if_index != 2090 opt->ip6po_pktinfo->ipi6_ifindex))) { 2091 return (EADDRNOTAVAIL); 2092 } 2093 ia6 = (struct in6_ifaddr *)ia; 2094 if ((ia6->ia6_flags & (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY)) != 0) { 2095 return (EADDRNOTAVAIL); 2096 } 2097 2098 /* 2099 * Check if the requested source address is 2100 * indeed a unicast address assigned to the 2101 * node. 2102 */ 2103 if (IN6_IS_ADDR_MULTICAST(&opt->ip6po_pktinfo->ipi6_addr)) 2104 return (EADDRNOTAVAIL); 2105 } 2106 break; 2107 2108 case IPV6_HOPLIMIT: 2109 if (cm->cmsg_len != CMSG_LEN(sizeof(int))) 2110 return (EINVAL); 2111 2112 bcopy(CMSG_DATA(cm), &opt->ip6po_hlim, 2113 sizeof(opt->ip6po_hlim)); 2114 if (opt->ip6po_hlim < -1 || opt->ip6po_hlim > 255) 2115 return (EINVAL); 2116 break; 2117 2118 case IPV6_NEXTHOP: 2119 if (!priv) 2120 return (EPERM); 2121 2122 /* check if cmsg_len is large enough for sa_len */ 2123 if (cm->cmsg_len < sizeof(u_char) || 2124 cm->cmsg_len < CMSG_LEN(*CMSG_DATA(cm))) 2125 return (EINVAL); 2126 2127 opt->ip6po_nexthop = (struct sockaddr *)CMSG_DATA(cm); 2128 2129 break; 2130 2131 case IPV6_HOPOPTS: 2132 if (cm->cmsg_len < CMSG_LEN(sizeof(struct ip6_hbh))) 2133 return (EINVAL); 2134 opt->ip6po_hbh = (struct ip6_hbh *)CMSG_DATA(cm); 2135 if (cm->cmsg_len != 2136 CMSG_LEN((opt->ip6po_hbh->ip6h_len + 1) << 3)) 2137 return (EINVAL); 2138 break; 2139 2140 case IPV6_DSTOPTS: 2141 if (cm->cmsg_len < CMSG_LEN(sizeof(struct ip6_dest))) 2142 return (EINVAL); 2143 2144 /* 2145 * If there is no routing header yet, the destination 2146 * options header should be put on the 1st part. 2147 * Otherwise, the header should be on the 2nd part. 2148 * (See RFC 2460, section 4.1) 2149 */ 2150 if (opt->ip6po_rthdr == NULL) { 2151 opt->ip6po_dest1 = 2152 (struct ip6_dest *)CMSG_DATA(cm); 2153 if (cm->cmsg_len != 2154 CMSG_LEN((opt->ip6po_dest1->ip6d_len + 1) << 3)); 2155 return (EINVAL); 2156 } 2157 else { 2158 opt->ip6po_dest2 = 2159 (struct ip6_dest *)CMSG_DATA(cm); 2160 if (cm->cmsg_len != 2161 CMSG_LEN((opt->ip6po_dest2->ip6d_len + 1) << 3)) 2162 return (EINVAL); 2163 } 2164 break; 2165 2166 case IPV6_RTHDR: 2167 if (cm->cmsg_len < CMSG_LEN(sizeof(struct ip6_rthdr))) 2168 return (EINVAL); 2169 opt->ip6po_rthdr = (struct ip6_rthdr *)CMSG_DATA(cm); 2170 if (cm->cmsg_len != 2171 CMSG_LEN((opt->ip6po_rthdr->ip6r_len + 1) << 3)) 2172 return (EINVAL); 2173 switch (opt->ip6po_rthdr->ip6r_type) { 2174 case IPV6_RTHDR_TYPE_0: 2175 if (opt->ip6po_rthdr->ip6r_segleft == 0) 2176 return (EINVAL); 2177 break; 2178 default: 2179 return (EINVAL); 2180 } 2181 break; 2182 2183 default: 2184 return (ENOPROTOOPT); 2185 } 2186 } 2187 2188 return (0); 2189 } 2190 2191 /* 2192 * Routine called from ip6_output() to loop back a copy of an IP6 multicast 2193 * packet to the input queue of a specified interface. Note that this 2194 * calls the output routine of the loopback "driver", but with an interface 2195 * pointer that might NOT be &loif -- easier than replicating that code here. 2196 */ 2197 void 2198 ip6_mloopback(ifp, m, dst) 2199 struct ifnet *ifp; 2200 struct mbuf *m; 2201 struct sockaddr_in6 *dst; 2202 { 2203 struct mbuf *copym; 2204 struct ip6_hdr *ip6; 2205 2206 copym = m_copy(m, 0, M_COPYALL); 2207 if (copym == NULL) 2208 return; 2209 2210 /* 2211 * Make sure to deep-copy IPv6 header portion in case the data 2212 * is in an mbuf cluster, so that we can safely override the IPv6 2213 * header portion later. 2214 */ 2215 if ((copym->m_flags & M_EXT) != 0 || 2216 copym->m_len < sizeof(struct ip6_hdr)) { 2217 copym = m_pullup(copym, sizeof(struct ip6_hdr)); 2218 if (copym == NULL) 2219 return; 2220 } 2221 2222 #ifdef DIAGNOSTIC 2223 if (copym->m_len < sizeof(*ip6)) { 2224 m_freem(copym); 2225 return; 2226 } 2227 #endif 2228 2229 ip6 = mtod(copym, struct ip6_hdr *); 2230 if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src)) 2231 ip6->ip6_src.s6_addr16[1] = 0; 2232 if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst)) 2233 ip6->ip6_dst.s6_addr16[1] = 0; 2234 2235 (void)looutput(ifp, copym, (struct sockaddr *)dst, NULL); 2236 } 2237 2238 /* 2239 * Chop IPv6 header off from the payload. 2240 */ 2241 static int 2242 ip6_splithdr(m, exthdrs) 2243 struct mbuf *m; 2244 struct ip6_exthdrs *exthdrs; 2245 { 2246 struct mbuf *mh; 2247 struct ip6_hdr *ip6; 2248 2249 ip6 = mtod(m, struct ip6_hdr *); 2250 if (m->m_len > sizeof(*ip6)) { 2251 MGETHDR(mh, M_DONTWAIT, MT_HEADER); 2252 if (mh == 0) { 2253 m_freem(m); 2254 return ENOBUFS; 2255 } 2256 M_COPY_PKTHDR(mh, m); 2257 MH_ALIGN(mh, sizeof(*ip6)); 2258 m->m_flags &= ~M_PKTHDR; 2259 m->m_len -= sizeof(*ip6); 2260 m->m_data += sizeof(*ip6); 2261 mh->m_next = m; 2262 m = mh; 2263 m->m_len = sizeof(*ip6); 2264 bcopy((caddr_t)ip6, mtod(m, caddr_t), sizeof(*ip6)); 2265 } 2266 exthdrs->ip6e_ip6 = m; 2267 return 0; 2268 } 2269 2270 /* 2271 * Compute IPv6 extension header length. 2272 */ 2273 int 2274 ip6_optlen(in6p) 2275 struct in6pcb *in6p; 2276 { 2277 int len; 2278 2279 if (!in6p->in6p_outputopts) 2280 return 0; 2281 2282 len = 0; 2283 #define elen(x) \ 2284 (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0) 2285 2286 len += elen(in6p->in6p_outputopts->ip6po_hbh); 2287 len += elen(in6p->in6p_outputopts->ip6po_dest1); 2288 len += elen(in6p->in6p_outputopts->ip6po_rthdr); 2289 len += elen(in6p->in6p_outputopts->ip6po_dest2); 2290 return len; 2291 #undef elen 2292 } 2293