ip6_output.c revision 1.44.4.4
1 /* $NetBSD: ip6_output.c,v 1.44.4.4 2004/02/07 20:06:04 jmc 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 1.44.4.4 2004/02/07 20:06:04 jmc 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 #include <netinet6/ip6protosw.h> 99 100 #ifdef IPSEC 101 #include <netinet6/ipsec.h> 102 #include <netkey/key.h> 103 #endif /* IPSEC */ 104 105 #include "loop.h" 106 107 #include <net/net_osdep.h> 108 109 #ifdef PFIL_HOOKS 110 extern struct pfil_head inet6_pfil_hook; /* XXX */ 111 #endif 112 113 struct ip6_exthdrs { 114 struct mbuf *ip6e_ip6; 115 struct mbuf *ip6e_hbh; 116 struct mbuf *ip6e_dest1; 117 struct mbuf *ip6e_rthdr; 118 struct mbuf *ip6e_dest2; 119 }; 120 121 static int ip6_pcbopts __P((struct ip6_pktopts **, struct mbuf *, 122 struct socket *)); 123 static int ip6_setmoptions __P((int, struct ip6_moptions **, struct mbuf *)); 124 static int ip6_getmoptions __P((int, struct ip6_moptions *, struct mbuf **)); 125 static int ip6_copyexthdr __P((struct mbuf **, caddr_t, int)); 126 static int ip6_insertfraghdr __P((struct mbuf *, struct mbuf *, int, 127 struct ip6_frag **)); 128 static int ip6_insert_jumboopt __P((struct ip6_exthdrs *, u_int32_t)); 129 static int ip6_splithdr __P((struct mbuf *, struct ip6_exthdrs *)); 130 static int ip6_getpmtu __P((struct route_in6 *, struct route_in6 *, 131 struct ifnet *, struct in6_addr *, u_long *, int *)); 132 133 extern struct ifnet loif[NLOOP]; 134 135 /* 136 * IP6 output. The packet in mbuf chain m contains a skeletal IP6 137 * header (with pri, len, nxt, hlim, src, dst). 138 * This function may modify ver and hlim only. 139 * The mbuf chain containing the packet will be freed. 140 * The mbuf opt, if present, will not be freed. 141 */ 142 int 143 ip6_output(m0, opt, ro, flags, im6o, ifpp) 144 struct mbuf *m0; 145 struct ip6_pktopts *opt; 146 struct route_in6 *ro; 147 int flags; 148 struct ip6_moptions *im6o; 149 struct ifnet **ifpp; /* XXX: just for statistics */ 150 { 151 struct ip6_hdr *ip6, *mhip6; 152 struct ifnet *ifp, *origifp; 153 struct mbuf *m = m0; 154 int hlen, tlen, len, off; 155 struct route_in6 ip6route; 156 struct sockaddr_in6 *dst; 157 int error = 0; 158 struct in6_ifaddr *ia; 159 u_long mtu; 160 int alwaysfrag, dontfrag; 161 u_int32_t optlen = 0, plen = 0, unfragpartlen = 0; 162 struct ip6_exthdrs exthdrs; 163 struct in6_addr finaldst; 164 struct route_in6 *ro_pmtu = NULL; 165 int hdrsplit = 0; 166 int needipsec = 0; 167 #ifdef IPSEC 168 int needipsectun = 0; 169 struct socket *so; 170 struct secpolicy *sp = NULL; 171 172 /* for AH processing. stupid to have "socket" variable in IP layer... */ 173 so = ipsec_getsocket(m); 174 (void)ipsec_setsocket(m, NULL); 175 ip6 = mtod(m, struct ip6_hdr *); 176 #endif /* IPSEC */ 177 178 #define MAKE_EXTHDR(hp, mp) \ 179 do { \ 180 if (hp) { \ 181 struct ip6_ext *eh = (struct ip6_ext *)(hp); \ 182 error = ip6_copyexthdr((mp), (caddr_t)(hp), \ 183 ((eh)->ip6e_len + 1) << 3); \ 184 if (error) \ 185 goto freehdrs; \ 186 } \ 187 } while (0) 188 189 bzero(&exthdrs, sizeof(exthdrs)); 190 if (opt) { 191 /* Hop-by-Hop options header */ 192 MAKE_EXTHDR(opt->ip6po_hbh, &exthdrs.ip6e_hbh); 193 /* Destination options header(1st part) */ 194 MAKE_EXTHDR(opt->ip6po_dest1, &exthdrs.ip6e_dest1); 195 /* Routing header */ 196 MAKE_EXTHDR(opt->ip6po_rthdr, &exthdrs.ip6e_rthdr); 197 /* Destination options header(2nd part) */ 198 MAKE_EXTHDR(opt->ip6po_dest2, &exthdrs.ip6e_dest2); 199 } 200 201 #ifdef IPSEC 202 /* get a security policy for this packet */ 203 if (so == NULL) 204 sp = ipsec6_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND, 0, &error); 205 else 206 sp = ipsec6_getpolicybysock(m, IPSEC_DIR_OUTBOUND, so, &error); 207 208 if (sp == NULL) { 209 ipsec6stat.out_inval++; 210 goto freehdrs; 211 } 212 213 error = 0; 214 215 /* check policy */ 216 switch (sp->policy) { 217 case IPSEC_POLICY_DISCARD: 218 /* 219 * This packet is just discarded. 220 */ 221 ipsec6stat.out_polvio++; 222 goto freehdrs; 223 224 case IPSEC_POLICY_BYPASS: 225 case IPSEC_POLICY_NONE: 226 /* no need to do IPsec. */ 227 needipsec = 0; 228 break; 229 230 case IPSEC_POLICY_IPSEC: 231 if (sp->req == NULL) { 232 /* XXX should be panic ? */ 233 printf("ip6_output: No IPsec request specified.\n"); 234 error = EINVAL; 235 goto freehdrs; 236 } 237 needipsec = 1; 238 break; 239 240 case IPSEC_POLICY_ENTRUST: 241 default: 242 printf("ip6_output: Invalid policy found. %d\n", sp->policy); 243 } 244 #endif /* IPSEC */ 245 246 /* 247 * Calculate the total length of the extension header chain. 248 * Keep the length of the unfragmentable part for fragmentation. 249 */ 250 optlen = 0; 251 if (exthdrs.ip6e_hbh) optlen += exthdrs.ip6e_hbh->m_len; 252 if (exthdrs.ip6e_dest1) optlen += exthdrs.ip6e_dest1->m_len; 253 if (exthdrs.ip6e_rthdr) optlen += exthdrs.ip6e_rthdr->m_len; 254 unfragpartlen = optlen + sizeof(struct ip6_hdr); 255 /* NOTE: we don't add AH/ESP length here. do that later. */ 256 if (exthdrs.ip6e_dest2) optlen += exthdrs.ip6e_dest2->m_len; 257 258 /* 259 * If we need IPsec, or there is at least one extension header, 260 * separate IP6 header from the payload. 261 */ 262 if ((needipsec || optlen) && !hdrsplit) { 263 if ((error = ip6_splithdr(m, &exthdrs)) != 0) { 264 m = NULL; 265 goto freehdrs; 266 } 267 m = exthdrs.ip6e_ip6; 268 hdrsplit++; 269 } 270 271 /* adjust pointer */ 272 ip6 = mtod(m, struct ip6_hdr *); 273 274 /* adjust mbuf packet header length */ 275 m->m_pkthdr.len += optlen; 276 plen = m->m_pkthdr.len - sizeof(*ip6); 277 278 /* If this is a jumbo payload, insert a jumbo payload option. */ 279 if (plen > IPV6_MAXPACKET) { 280 if (!hdrsplit) { 281 if ((error = ip6_splithdr(m, &exthdrs)) != 0) { 282 m = NULL; 283 goto freehdrs; 284 } 285 m = exthdrs.ip6e_ip6; 286 hdrsplit++; 287 } 288 /* adjust pointer */ 289 ip6 = mtod(m, struct ip6_hdr *); 290 if ((error = ip6_insert_jumboopt(&exthdrs, plen)) != 0) 291 goto freehdrs; 292 ip6->ip6_plen = 0; 293 } else 294 ip6->ip6_plen = htons(plen); 295 296 /* 297 * Concatenate headers and fill in next header fields. 298 * Here we have, on "m" 299 * IPv6 payload 300 * and we insert headers accordingly. Finally, we should be getting: 301 * IPv6 hbh dest1 rthdr ah* [esp* dest2 payload] 302 * 303 * during the header composing process, "m" points to IPv6 header. 304 * "mprev" points to an extension header prior to esp. 305 */ 306 { 307 u_char *nexthdrp = &ip6->ip6_nxt; 308 struct mbuf *mprev = m; 309 310 /* 311 * we treat dest2 specially. this makes IPsec processing 312 * much easier. 313 * 314 * result: IPv6 dest2 payload 315 * m and mprev will point to IPv6 header. 316 */ 317 if (exthdrs.ip6e_dest2) { 318 if (!hdrsplit) 319 panic("assumption failed: hdr not split"); 320 exthdrs.ip6e_dest2->m_next = m->m_next; 321 m->m_next = exthdrs.ip6e_dest2; 322 *mtod(exthdrs.ip6e_dest2, u_char *) = ip6->ip6_nxt; 323 ip6->ip6_nxt = IPPROTO_DSTOPTS; 324 } 325 326 #define MAKE_CHAIN(m, mp, p, i)\ 327 do {\ 328 if (m) {\ 329 if (!hdrsplit) \ 330 panic("assumption failed: hdr not split"); \ 331 *mtod((m), u_char *) = *(p);\ 332 *(p) = (i);\ 333 p = mtod((m), u_char *);\ 334 (m)->m_next = (mp)->m_next;\ 335 (mp)->m_next = (m);\ 336 (mp) = (m);\ 337 }\ 338 } while (0) 339 /* 340 * result: IPv6 hbh dest1 rthdr dest2 payload 341 * m will point to IPv6 header. mprev will point to the 342 * extension header prior to dest2 (rthdr in the above case). 343 */ 344 MAKE_CHAIN(exthdrs.ip6e_hbh, mprev, 345 nexthdrp, IPPROTO_HOPOPTS); 346 MAKE_CHAIN(exthdrs.ip6e_dest1, mprev, 347 nexthdrp, IPPROTO_DSTOPTS); 348 MAKE_CHAIN(exthdrs.ip6e_rthdr, mprev, 349 nexthdrp, IPPROTO_ROUTING); 350 351 #ifdef IPSEC 352 if (!needipsec) 353 goto skip_ipsec2; 354 355 /* 356 * pointers after IPsec headers are not valid any more. 357 * other pointers need a great care too. 358 * (IPsec routines should not mangle mbufs prior to AH/ESP) 359 */ 360 exthdrs.ip6e_dest2 = NULL; 361 362 { 363 struct ip6_rthdr *rh = NULL; 364 int segleft_org = 0; 365 struct ipsec_output_state state; 366 367 if (exthdrs.ip6e_rthdr) { 368 rh = mtod(exthdrs.ip6e_rthdr, struct ip6_rthdr *); 369 segleft_org = rh->ip6r_segleft; 370 rh->ip6r_segleft = 0; 371 } 372 373 bzero(&state, sizeof(state)); 374 state.m = m; 375 error = ipsec6_output_trans(&state, nexthdrp, mprev, sp, flags, 376 &needipsectun); 377 m = state.m; 378 if (error) { 379 /* mbuf is already reclaimed in ipsec6_output_trans. */ 380 m = NULL; 381 switch (error) { 382 case EHOSTUNREACH: 383 case ENETUNREACH: 384 case EMSGSIZE: 385 case ENOBUFS: 386 case ENOMEM: 387 break; 388 default: 389 printf("ip6_output (ipsec): error code %d\n", error); 390 /* fall through */ 391 case ENOENT: 392 /* don't show these error codes to the user */ 393 error = 0; 394 break; 395 } 396 goto bad; 397 } 398 if (exthdrs.ip6e_rthdr) { 399 /* ah6_output doesn't modify mbuf chain */ 400 rh->ip6r_segleft = segleft_org; 401 } 402 } 403 skip_ipsec2:; 404 #endif 405 } 406 407 /* 408 * If there is a routing header, replace destination address field 409 * with the first hop of the routing header. 410 */ 411 if (exthdrs.ip6e_rthdr) { 412 struct ip6_rthdr *rh = 413 (struct ip6_rthdr *)(mtod(exthdrs.ip6e_rthdr, 414 struct ip6_rthdr *)); 415 struct ip6_rthdr0 *rh0; 416 417 finaldst = ip6->ip6_dst; 418 switch (rh->ip6r_type) { 419 case IPV6_RTHDR_TYPE_0: 420 rh0 = (struct ip6_rthdr0 *)rh; 421 ip6->ip6_dst = rh0->ip6r0_addr[0]; 422 bcopy((caddr_t)&rh0->ip6r0_addr[1], 423 (caddr_t)&rh0->ip6r0_addr[0], 424 sizeof(struct in6_addr) * (rh0->ip6r0_segleft - 1) 425 ); 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 /* fall through */ 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 } 641 if (ro->ro_rt == 0) { 642 ip6stat.ip6s_noroute++; 643 error = EHOSTUNREACH; 644 /* XXX in6_ifstat_inc(ifp, ifs6_out_discard) */ 645 goto bad; 646 } 647 ia = ifatoia6(ro->ro_rt->rt_ifa); 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; 719 720 /* 721 * The caller of this function may specify to use the minimum MTU 722 * in some cases. 723 */ 724 if (mtu > IPV6_MMTU) { 725 if ((flags & IPV6_MINMTU)) 726 mtu = IPV6_MMTU; 727 } 728 729 /* Fake scoped addresses */ 730 if ((ifp->if_flags & IFF_LOOPBACK) != 0) { 731 /* 732 * If source or destination address is a scoped address, and 733 * the packet is going to be sent to a loopback interface, 734 * we should keep the original interface. 735 */ 736 737 /* 738 * XXX: this is a very experimental and temporary solution. 739 * We eventually have sockaddr_in6 and use the sin6_scope_id 740 * field of the structure here. 741 * We rely on the consistency between two scope zone ids 742 * of source add destination, which should already be assured 743 * Larger scopes than link will be supported in the near 744 * future. 745 */ 746 origifp = NULL; 747 if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src)) 748 origifp = ifindex2ifnet[ntohs(ip6->ip6_src.s6_addr16[1])]; 749 else if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst)) 750 origifp = ifindex2ifnet[ntohs(ip6->ip6_dst.s6_addr16[1])]; 751 /* 752 * XXX: origifp can be NULL even in those two cases above. 753 * For example, if we remove the (only) link-local address 754 * from the loopback interface, and try to send a link-local 755 * address without link-id information. Then the source 756 * address is ::1, and the destination address is the 757 * link-local address with its s6_addr16[1] being zero. 758 * What is worse, if the packet goes to the loopback interface 759 * by a default rejected route, the null pointer would be 760 * passed to looutput, and the kernel would hang. 761 * The following last resort would prevent such disaster. 762 */ 763 if (origifp == NULL) 764 origifp = ifp; 765 } 766 else 767 origifp = ifp; 768 if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src)) 769 ip6->ip6_src.s6_addr16[1] = 0; 770 if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst)) 771 ip6->ip6_dst.s6_addr16[1] = 0; 772 773 /* 774 * If the outgoing packet contains a hop-by-hop options header, 775 * it must be examined and processed even by the source node. 776 * (RFC 2460, section 4.) 777 */ 778 if (exthdrs.ip6e_hbh) { 779 struct ip6_hbh *hbh = mtod(exthdrs.ip6e_hbh, struct ip6_hbh *); 780 u_int32_t dummy1; /* XXX unused */ 781 u_int32_t dummy2; /* XXX unused */ 782 783 /* 784 * XXX: if we have to send an ICMPv6 error to the sender, 785 * we need the M_LOOP flag since icmp6_error() expects 786 * the IPv6 and the hop-by-hop options header are 787 * continuous unless the flag is set. 788 */ 789 m->m_flags |= M_LOOP; 790 m->m_pkthdr.rcvif = ifp; 791 if (ip6_process_hopopts(m, 792 (u_int8_t *)(hbh + 1), 793 ((hbh->ip6h_len + 1) << 3) - 794 sizeof(struct ip6_hbh), 795 &dummy1, &dummy2) < 0) { 796 /* m was already freed at this point */ 797 error = EINVAL;/* better error? */ 798 goto done; 799 } 800 m->m_flags &= ~M_LOOP; /* XXX */ 801 m->m_pkthdr.rcvif = NULL; 802 } 803 804 #ifdef PFIL_HOOKS 805 /* 806 * Run through list of hooks for output packets. 807 */ 808 if ((error = pfil_run_hooks(&inet6_pfil_hook, &m, ifp, 809 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 > nd_ifinfo[ifp->if_index].linkmtu) { /* 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_id++); 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 /* fall through */ 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_char *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_char *); 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, caddr_t) + oldoptlen; 1159 m_freem(mopt); 1160 mopt = exthdrs->ip6e_hbh = n; 1161 } else { 1162 optbuf = mtod(mopt, u_char *) + 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 = 1217 (struct ip6_frag *)(mtod(mlast, caddr_t) + 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 static 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 = nd_ifinfo[ifp->if_index].linkmtu; 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 = nd_ifinfo[ifp->if_index].linkmtu; 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 1329 case PRCO_SETOPT: 1330 switch (optname) { 1331 case IPV6_PKTOPTIONS: 1332 /* m is freed in ip6_pcbopts */ 1333 return(ip6_pcbopts(&in6p->in6p_outputopts, 1334 m, so)); 1335 case IPV6_HOPOPTS: 1336 case IPV6_DSTOPTS: 1337 if (p == 0 || suser(p->p_ucred, &p->p_acflag)) { 1338 error = EPERM; 1339 break; 1340 } 1341 /* fall through */ 1342 case IPV6_UNICAST_HOPS: 1343 case IPV6_RECVOPTS: 1344 case IPV6_RECVRETOPTS: 1345 case IPV6_RECVDSTADDR: 1346 case IPV6_PKTINFO: 1347 case IPV6_HOPLIMIT: 1348 case IPV6_RTHDR: 1349 case IPV6_FAITH: 1350 case IPV6_V6ONLY: 1351 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 if (optval) \ 1368 in6p->in6p_flags |= bit; \ 1369 else \ 1370 in6p->in6p_flags &= ~bit; 1371 1372 case IPV6_RECVOPTS: 1373 OPTSET(IN6P_RECVOPTS); 1374 break; 1375 1376 case IPV6_RECVRETOPTS: 1377 OPTSET(IN6P_RECVRETOPTS); 1378 break; 1379 1380 case IPV6_RECVDSTADDR: 1381 OPTSET(IN6P_RECVDSTADDR); 1382 break; 1383 1384 case IPV6_PKTINFO: 1385 OPTSET(IN6P_PKTINFO); 1386 break; 1387 1388 case IPV6_HOPLIMIT: 1389 OPTSET(IN6P_HOPLIMIT); 1390 break; 1391 1392 case IPV6_HOPOPTS: 1393 OPTSET(IN6P_HOPOPTS); 1394 break; 1395 1396 case IPV6_DSTOPTS: 1397 OPTSET(IN6P_DSTOPTS); 1398 break; 1399 1400 case IPV6_RTHDR: 1401 OPTSET(IN6P_RTHDR); 1402 break; 1403 1404 case IPV6_FAITH: 1405 OPTSET(IN6P_FAITH); 1406 break; 1407 1408 case IPV6_V6ONLY: 1409 /* 1410 * make setsockopt(IPV6_V6ONLY) 1411 * available only prior to bind(2). 1412 * see ipng mailing list, Jun 22 2001. 1413 */ 1414 if (in6p->in6p_lport || 1415 !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr)) 1416 { 1417 error = EINVAL; 1418 break; 1419 } 1420 #ifdef INET6_BINDV6ONLY 1421 if (!optval) 1422 error = EINVAL; 1423 #else 1424 OPTSET(IN6P_IPV6_V6ONLY); 1425 #endif 1426 break; 1427 } 1428 break; 1429 #undef OPTSET 1430 1431 case IPV6_MULTICAST_IF: 1432 case IPV6_MULTICAST_HOPS: 1433 case IPV6_MULTICAST_LOOP: 1434 case IPV6_JOIN_GROUP: 1435 case IPV6_LEAVE_GROUP: 1436 error = ip6_setmoptions(optname, &in6p->in6p_moptions, m); 1437 break; 1438 1439 case IPV6_PORTRANGE: 1440 optval = *mtod(m, int *); 1441 1442 switch (optval) { 1443 case IPV6_PORTRANGE_DEFAULT: 1444 in6p->in6p_flags &= ~(IN6P_LOWPORT); 1445 in6p->in6p_flags &= ~(IN6P_HIGHPORT); 1446 break; 1447 1448 case IPV6_PORTRANGE_HIGH: 1449 in6p->in6p_flags &= ~(IN6P_LOWPORT); 1450 in6p->in6p_flags |= IN6P_HIGHPORT; 1451 break; 1452 1453 case IPV6_PORTRANGE_LOW: 1454 in6p->in6p_flags &= ~(IN6P_HIGHPORT); 1455 in6p->in6p_flags |= IN6P_LOWPORT; 1456 break; 1457 1458 default: 1459 error = EINVAL; 1460 break; 1461 } 1462 break; 1463 1464 #ifdef IPSEC 1465 case IPV6_IPSEC_POLICY: 1466 { 1467 caddr_t req = NULL; 1468 size_t len = 0; 1469 1470 int priv = 0; 1471 if (p == 0 || suser(p->p_ucred, &p->p_acflag)) 1472 priv = 0; 1473 else 1474 priv = 1; 1475 if (m) { 1476 req = mtod(m, caddr_t); 1477 len = m->m_len; 1478 } 1479 error = ipsec6_set_policy(in6p, 1480 optname, req, len, priv); 1481 } 1482 break; 1483 #endif /* IPSEC */ 1484 1485 default: 1486 error = ENOPROTOOPT; 1487 break; 1488 } 1489 if (m) 1490 (void)m_free(m); 1491 break; 1492 1493 case PRCO_GETOPT: 1494 switch (optname) { 1495 1496 case IPV6_OPTIONS: 1497 case IPV6_RETOPTS: 1498 #if 0 1499 *mp = m = m_get(M_WAIT, MT_SOOPTS); 1500 if (in6p->in6p_options) { 1501 m->m_len = in6p->in6p_options->m_len; 1502 bcopy(mtod(in6p->in6p_options, caddr_t), 1503 mtod(m, caddr_t), 1504 (unsigned)m->m_len); 1505 } else 1506 m->m_len = 0; 1507 break; 1508 #else 1509 error = ENOPROTOOPT; 1510 break; 1511 #endif 1512 1513 case IPV6_PKTOPTIONS: 1514 if (in6p->in6p_options) { 1515 *mp = m_copym(in6p->in6p_options, 0, 1516 M_COPYALL, M_WAIT); 1517 } else { 1518 *mp = m_get(M_WAIT, MT_SOOPTS); 1519 (*mp)->m_len = 0; 1520 } 1521 break; 1522 1523 case IPV6_HOPOPTS: 1524 case IPV6_DSTOPTS: 1525 if (p == 0 || suser(p->p_ucred, &p->p_acflag)) { 1526 error = EPERM; 1527 break; 1528 } 1529 /* fall through */ 1530 case IPV6_UNICAST_HOPS: 1531 case IPV6_RECVOPTS: 1532 case IPV6_RECVRETOPTS: 1533 case IPV6_RECVDSTADDR: 1534 case IPV6_PORTRANGE: 1535 case IPV6_PKTINFO: 1536 case IPV6_HOPLIMIT: 1537 case IPV6_RTHDR: 1538 case IPV6_FAITH: 1539 case IPV6_V6ONLY: 1540 *mp = m = m_get(M_WAIT, MT_SOOPTS); 1541 m->m_len = sizeof(int); 1542 switch (optname) { 1543 1544 case IPV6_UNICAST_HOPS: 1545 optval = in6p->in6p_hops; 1546 break; 1547 1548 #define OPTBIT(bit) (in6p->in6p_flags & bit ? 1 : 0) 1549 1550 case IPV6_RECVOPTS: 1551 optval = OPTBIT(IN6P_RECVOPTS); 1552 break; 1553 1554 case IPV6_RECVRETOPTS: 1555 optval = OPTBIT(IN6P_RECVRETOPTS); 1556 break; 1557 1558 case IPV6_RECVDSTADDR: 1559 optval = OPTBIT(IN6P_RECVDSTADDR); 1560 break; 1561 1562 case IPV6_PORTRANGE: 1563 { 1564 int flags; 1565 flags = in6p->in6p_flags; 1566 if (flags & IN6P_HIGHPORT) 1567 optval = IPV6_PORTRANGE_HIGH; 1568 else if (flags & IN6P_LOWPORT) 1569 optval = IPV6_PORTRANGE_LOW; 1570 else 1571 optval = 0; 1572 break; 1573 } 1574 1575 case IPV6_PKTINFO: 1576 optval = OPTBIT(IN6P_PKTINFO); 1577 break; 1578 1579 case IPV6_HOPLIMIT: 1580 optval = OPTBIT(IN6P_HOPLIMIT); 1581 break; 1582 1583 case IPV6_HOPOPTS: 1584 optval = OPTBIT(IN6P_HOPOPTS); 1585 break; 1586 1587 case IPV6_DSTOPTS: 1588 optval = OPTBIT(IN6P_DSTOPTS); 1589 break; 1590 1591 case IPV6_RTHDR: 1592 optval = OPTBIT(IN6P_RTHDR); 1593 break; 1594 1595 case IPV6_FAITH: 1596 optval = OPTBIT(IN6P_FAITH); 1597 break; 1598 1599 case IPV6_V6ONLY: 1600 optval = OPTBIT(IN6P_IPV6_V6ONLY); 1601 break; 1602 } 1603 *mtod(m, int *) = optval; 1604 break; 1605 1606 case IPV6_MULTICAST_IF: 1607 case IPV6_MULTICAST_HOPS: 1608 case IPV6_MULTICAST_LOOP: 1609 case IPV6_JOIN_GROUP: 1610 case IPV6_LEAVE_GROUP: 1611 error = ip6_getmoptions(optname, in6p->in6p_moptions, mp); 1612 break; 1613 1614 #ifdef IPSEC 1615 case IPV6_IPSEC_POLICY: 1616 { 1617 caddr_t req = NULL; 1618 size_t len = 0; 1619 1620 if (m) { 1621 req = mtod(m, caddr_t); 1622 len = m->m_len; 1623 } 1624 error = ipsec6_get_policy(in6p, req, len, mp); 1625 break; 1626 } 1627 #endif /* IPSEC */ 1628 1629 default: 1630 error = ENOPROTOOPT; 1631 break; 1632 } 1633 break; 1634 } 1635 } else { 1636 error = EINVAL; 1637 if (op == PRCO_SETOPT && *mp) 1638 (void)m_free(*mp); 1639 } 1640 return(error); 1641 } 1642 1643 /* 1644 * Set up IP6 options in pcb for insertion in output packets. 1645 * Store in mbuf with pointer in pcbopt, adding pseudo-option 1646 * with destination address if source routed. 1647 */ 1648 static int 1649 ip6_pcbopts(pktopt, m, so) 1650 struct ip6_pktopts **pktopt; 1651 struct mbuf *m; 1652 struct socket *so; 1653 { 1654 struct ip6_pktopts *opt = *pktopt; 1655 int error = 0; 1656 struct proc *p = curproc; /* XXX */ 1657 int priv = 0; 1658 1659 /* turn off any old options. */ 1660 if (opt) { 1661 if (opt->ip6po_m) 1662 (void)m_free(opt->ip6po_m); 1663 } else 1664 opt = malloc(sizeof(*opt), M_IP6OPT, M_WAITOK); 1665 *pktopt = 0; 1666 1667 if (!m || m->m_len == 0) { 1668 /* 1669 * Only turning off any previous options. 1670 */ 1671 if (opt) 1672 free(opt, M_IP6OPT); 1673 if (m) 1674 (void)m_free(m); 1675 return(0); 1676 } 1677 1678 /* set options specified by user. */ 1679 if (p && !suser(p->p_ucred, &p->p_acflag)) 1680 priv = 1; 1681 if ((error = ip6_setpktoptions(m, opt, priv)) != 0) { 1682 (void)m_free(m); 1683 return(error); 1684 } 1685 *pktopt = opt; 1686 return(0); 1687 } 1688 1689 /* 1690 * Set the IP6 multicast options in response to user setsockopt(). 1691 */ 1692 static int 1693 ip6_setmoptions(optname, im6op, m) 1694 int optname; 1695 struct ip6_moptions **im6op; 1696 struct mbuf *m; 1697 { 1698 int error = 0; 1699 u_int loop, ifindex; 1700 struct ipv6_mreq *mreq; 1701 struct ifnet *ifp; 1702 struct ip6_moptions *im6o = *im6op; 1703 struct route_in6 ro; 1704 struct sockaddr_in6 *dst; 1705 struct in6_multi_mship *imm; 1706 struct proc *p = curproc; /* XXX */ 1707 1708 if (im6o == NULL) { 1709 /* 1710 * No multicast option buffer attached to the pcb; 1711 * allocate one and initialize to default values. 1712 */ 1713 im6o = (struct ip6_moptions *) 1714 malloc(sizeof(*im6o), M_IPMOPTS, M_WAITOK); 1715 1716 if (im6o == NULL) 1717 return(ENOBUFS); 1718 *im6op = im6o; 1719 im6o->im6o_multicast_ifp = NULL; 1720 im6o->im6o_multicast_hlim = ip6_defmcasthlim; 1721 im6o->im6o_multicast_loop = IPV6_DEFAULT_MULTICAST_LOOP; 1722 LIST_INIT(&im6o->im6o_memberships); 1723 } 1724 1725 switch (optname) { 1726 1727 case IPV6_MULTICAST_IF: 1728 /* 1729 * Select the interface for outgoing multicast packets. 1730 */ 1731 if (m == NULL || m->m_len != sizeof(u_int)) { 1732 error = EINVAL; 1733 break; 1734 } 1735 bcopy(mtod(m, u_int *), &ifindex, sizeof(ifindex)); 1736 if (ifindex < 0 || if_index < ifindex) { 1737 error = ENXIO; /* XXX EINVAL? */ 1738 break; 1739 } 1740 ifp = ifindex2ifnet[ifindex]; 1741 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) { 1742 error = EADDRNOTAVAIL; 1743 break; 1744 } 1745 im6o->im6o_multicast_ifp = ifp; 1746 break; 1747 1748 case IPV6_MULTICAST_HOPS: 1749 { 1750 /* 1751 * Set the IP6 hoplimit for outgoing multicast packets. 1752 */ 1753 int optval; 1754 if (m == NULL || m->m_len != sizeof(int)) { 1755 error = EINVAL; 1756 break; 1757 } 1758 bcopy(mtod(m, u_int *), &optval, sizeof(optval)); 1759 if (optval < -1 || optval >= 256) 1760 error = EINVAL; 1761 else if (optval == -1) 1762 im6o->im6o_multicast_hlim = ip6_defmcasthlim; 1763 else 1764 im6o->im6o_multicast_hlim = optval; 1765 break; 1766 } 1767 1768 case IPV6_MULTICAST_LOOP: 1769 /* 1770 * Set the loopback flag for outgoing multicast packets. 1771 * Must be zero or one. 1772 */ 1773 if (m == NULL || m->m_len != sizeof(u_int)) { 1774 error = EINVAL; 1775 break; 1776 } 1777 bcopy(mtod(m, u_int *), &loop, sizeof(loop)); 1778 if (loop > 1) { 1779 error = EINVAL; 1780 break; 1781 } 1782 im6o->im6o_multicast_loop = loop; 1783 break; 1784 1785 case IPV6_JOIN_GROUP: 1786 /* 1787 * Add a multicast group membership. 1788 * Group must be a valid IP6 multicast address. 1789 */ 1790 if (m == NULL || m->m_len != sizeof(struct ipv6_mreq)) { 1791 error = EINVAL; 1792 break; 1793 } 1794 mreq = mtod(m, struct ipv6_mreq *); 1795 if (IN6_IS_ADDR_UNSPECIFIED(&mreq->ipv6mr_multiaddr)) { 1796 /* 1797 * We use the unspecified address to specify to accept 1798 * all multicast addresses. Only super user is allowed 1799 * to do this. 1800 */ 1801 if (suser(p->p_ucred, &p->p_acflag)) 1802 { 1803 error = EACCES; 1804 break; 1805 } 1806 } else if (!IN6_IS_ADDR_MULTICAST(&mreq->ipv6mr_multiaddr)) { 1807 error = EINVAL; 1808 break; 1809 } 1810 1811 /* 1812 * If the interface is specified, validate it. 1813 */ 1814 if (mreq->ipv6mr_interface < 0 1815 || if_index < mreq->ipv6mr_interface) { 1816 error = ENXIO; /* XXX EINVAL? */ 1817 break; 1818 } 1819 /* 1820 * If no interface was explicitly specified, choose an 1821 * appropriate one according to the given multicast address. 1822 */ 1823 if (mreq->ipv6mr_interface == 0) { 1824 /* 1825 * If the multicast address is in node-local scope, 1826 * the interface should be a loopback interface. 1827 * Otherwise, look up the routing table for the 1828 * address, and choose the outgoing interface. 1829 * XXX: is it a good approach? 1830 */ 1831 if (IN6_IS_ADDR_MC_NODELOCAL(&mreq->ipv6mr_multiaddr)) { 1832 ifp = &loif[0]; 1833 } else { 1834 ro.ro_rt = NULL; 1835 dst = (struct sockaddr_in6 *)&ro.ro_dst; 1836 bzero(dst, sizeof(*dst)); 1837 dst->sin6_len = sizeof(struct sockaddr_in6); 1838 dst->sin6_family = AF_INET6; 1839 dst->sin6_addr = mreq->ipv6mr_multiaddr; 1840 rtalloc((struct route *)&ro); 1841 if (ro.ro_rt == NULL) { 1842 error = EADDRNOTAVAIL; 1843 break; 1844 } 1845 ifp = ro.ro_rt->rt_ifp; 1846 rtfree(ro.ro_rt); 1847 } 1848 } else 1849 ifp = ifindex2ifnet[mreq->ipv6mr_interface]; 1850 1851 /* 1852 * See if we found an interface, and confirm that it 1853 * supports multicast 1854 */ 1855 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) { 1856 error = EADDRNOTAVAIL; 1857 break; 1858 } 1859 /* 1860 * Put interface index into the multicast address, 1861 * if the address has link-local scope. 1862 */ 1863 if (IN6_IS_ADDR_MC_LINKLOCAL(&mreq->ipv6mr_multiaddr)) { 1864 mreq->ipv6mr_multiaddr.s6_addr16[1] 1865 = htons(mreq->ipv6mr_interface); 1866 } 1867 /* 1868 * See if the membership already exists. 1869 */ 1870 for (imm = im6o->im6o_memberships.lh_first; 1871 imm != NULL; imm = imm->i6mm_chain.le_next) 1872 if (imm->i6mm_maddr->in6m_ifp == ifp && 1873 IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr, 1874 &mreq->ipv6mr_multiaddr)) 1875 break; 1876 if (imm != NULL) { 1877 error = EADDRINUSE; 1878 break; 1879 } 1880 /* 1881 * Everything looks good; add a new record to the multicast 1882 * address list for the given interface. 1883 */ 1884 imm = in6_joingroup(ifp, &mreq->ipv6mr_multiaddr, &error); 1885 if (!imm) 1886 break; 1887 LIST_INSERT_HEAD(&im6o->im6o_memberships, imm, i6mm_chain); 1888 break; 1889 1890 case IPV6_LEAVE_GROUP: 1891 /* 1892 * Drop a multicast group membership. 1893 * Group must be a valid IP6 multicast address. 1894 */ 1895 if (m == NULL || m->m_len != sizeof(struct ipv6_mreq)) { 1896 error = EINVAL; 1897 break; 1898 } 1899 mreq = mtod(m, struct ipv6_mreq *); 1900 if (IN6_IS_ADDR_UNSPECIFIED(&mreq->ipv6mr_multiaddr)) { 1901 if (suser(p->p_ucred, &p->p_acflag)) 1902 { 1903 error = EACCES; 1904 break; 1905 } 1906 } else if (!IN6_IS_ADDR_MULTICAST(&mreq->ipv6mr_multiaddr)) { 1907 error = EINVAL; 1908 break; 1909 } 1910 /* 1911 * If an interface address was specified, get a pointer 1912 * to its ifnet structure. 1913 */ 1914 if (mreq->ipv6mr_interface < 0 1915 || if_index < mreq->ipv6mr_interface) { 1916 error = ENXIO; /* XXX EINVAL? */ 1917 break; 1918 } 1919 ifp = ifindex2ifnet[mreq->ipv6mr_interface]; 1920 /* 1921 * Put interface index into the multicast address, 1922 * if the address has link-local scope. 1923 */ 1924 if (IN6_IS_ADDR_MC_LINKLOCAL(&mreq->ipv6mr_multiaddr)) { 1925 mreq->ipv6mr_multiaddr.s6_addr16[1] 1926 = htons(mreq->ipv6mr_interface); 1927 } 1928 /* 1929 * Find the membership in the membership list. 1930 */ 1931 for (imm = im6o->im6o_memberships.lh_first; 1932 imm != NULL; imm = imm->i6mm_chain.le_next) { 1933 if ((ifp == NULL || 1934 imm->i6mm_maddr->in6m_ifp == ifp) && 1935 IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr, 1936 &mreq->ipv6mr_multiaddr)) 1937 break; 1938 } 1939 if (imm == NULL) { 1940 /* Unable to resolve interface */ 1941 error = EADDRNOTAVAIL; 1942 break; 1943 } 1944 /* 1945 * Give up the multicast address record to which the 1946 * membership points. 1947 */ 1948 LIST_REMOVE(imm, i6mm_chain); 1949 in6_leavegroup(imm); 1950 break; 1951 1952 default: 1953 error = EOPNOTSUPP; 1954 break; 1955 } 1956 1957 /* 1958 * If all options have default values, no need to keep the mbuf. 1959 */ 1960 if (im6o->im6o_multicast_ifp == NULL && 1961 im6o->im6o_multicast_hlim == ip6_defmcasthlim && 1962 im6o->im6o_multicast_loop == IPV6_DEFAULT_MULTICAST_LOOP && 1963 im6o->im6o_memberships.lh_first == NULL) { 1964 free(*im6op, M_IPMOPTS); 1965 *im6op = NULL; 1966 } 1967 1968 return(error); 1969 } 1970 1971 /* 1972 * Return the IP6 multicast options in response to user getsockopt(). 1973 */ 1974 static int 1975 ip6_getmoptions(optname, im6o, mp) 1976 int optname; 1977 struct ip6_moptions *im6o; 1978 struct mbuf **mp; 1979 { 1980 u_int *hlim, *loop, *ifindex; 1981 1982 *mp = m_get(M_WAIT, MT_SOOPTS); 1983 1984 switch (optname) { 1985 1986 case IPV6_MULTICAST_IF: 1987 ifindex = mtod(*mp, u_int *); 1988 (*mp)->m_len = sizeof(u_int); 1989 if (im6o == NULL || im6o->im6o_multicast_ifp == NULL) 1990 *ifindex = 0; 1991 else 1992 *ifindex = im6o->im6o_multicast_ifp->if_index; 1993 return(0); 1994 1995 case IPV6_MULTICAST_HOPS: 1996 hlim = mtod(*mp, u_int *); 1997 (*mp)->m_len = sizeof(u_int); 1998 if (im6o == NULL) 1999 *hlim = ip6_defmcasthlim; 2000 else 2001 *hlim = im6o->im6o_multicast_hlim; 2002 return(0); 2003 2004 case IPV6_MULTICAST_LOOP: 2005 loop = mtod(*mp, u_int *); 2006 (*mp)->m_len = sizeof(u_int); 2007 if (im6o == NULL) 2008 *loop = ip6_defmcasthlim; 2009 else 2010 *loop = im6o->im6o_multicast_loop; 2011 return(0); 2012 2013 default: 2014 return(EOPNOTSUPP); 2015 } 2016 } 2017 2018 /* 2019 * Discard the IP6 multicast options. 2020 */ 2021 void 2022 ip6_freemoptions(im6o) 2023 struct ip6_moptions *im6o; 2024 { 2025 struct in6_multi_mship *imm; 2026 2027 if (im6o == NULL) 2028 return; 2029 2030 while ((imm = im6o->im6o_memberships.lh_first) != NULL) { 2031 LIST_REMOVE(imm, i6mm_chain); 2032 in6_leavegroup(imm); 2033 } 2034 free(im6o, M_IPMOPTS); 2035 } 2036 2037 /* 2038 * Set IPv6 outgoing packet options based on advanced API. 2039 */ 2040 int 2041 ip6_setpktoptions(control, opt, priv) 2042 struct mbuf *control; 2043 struct ip6_pktopts *opt; 2044 int priv; 2045 { 2046 struct cmsghdr *cm = 0; 2047 2048 if (control == 0 || opt == 0) 2049 return(EINVAL); 2050 2051 bzero(opt, sizeof(*opt)); 2052 opt->ip6po_hlim = -1; /* -1 means to use default hop limit */ 2053 2054 /* 2055 * XXX: Currently, we assume all the optional information is stored 2056 * in a single mbuf. 2057 */ 2058 if (control->m_next) 2059 return(EINVAL); 2060 2061 opt->ip6po_m = control; 2062 2063 for (; control->m_len; control->m_data += CMSG_ALIGN(cm->cmsg_len), 2064 control->m_len -= CMSG_ALIGN(cm->cmsg_len)) { 2065 cm = mtod(control, struct cmsghdr *); 2066 if (cm->cmsg_len == 0 || cm->cmsg_len > control->m_len) 2067 return(EINVAL); 2068 if (cm->cmsg_level != IPPROTO_IPV6) 2069 continue; 2070 2071 switch (cm->cmsg_type) { 2072 case IPV6_PKTINFO: 2073 if (cm->cmsg_len != CMSG_LEN(sizeof(struct in6_pktinfo))) 2074 return(EINVAL); 2075 opt->ip6po_pktinfo = (struct in6_pktinfo *)CMSG_DATA(cm); 2076 if (opt->ip6po_pktinfo->ipi6_ifindex && 2077 IN6_IS_ADDR_LINKLOCAL(&opt->ip6po_pktinfo->ipi6_addr)) 2078 opt->ip6po_pktinfo->ipi6_addr.s6_addr16[1] = 2079 htons(opt->ip6po_pktinfo->ipi6_ifindex); 2080 2081 if (opt->ip6po_pktinfo->ipi6_ifindex > if_index 2082 || opt->ip6po_pktinfo->ipi6_ifindex < 0) { 2083 return(ENXIO); 2084 } 2085 2086 /* 2087 * Check if the requested source address is indeed a 2088 * unicast address assigned to the node, and can be 2089 * used as the packet's source address. 2090 */ 2091 if (!IN6_IS_ADDR_UNSPECIFIED(&opt->ip6po_pktinfo->ipi6_addr)) { 2092 struct ifaddr *ia; 2093 struct in6_ifaddr *ia6; 2094 struct sockaddr_in6 sin6; 2095 2096 bzero(&sin6, sizeof(sin6)); 2097 sin6.sin6_len = sizeof(sin6); 2098 sin6.sin6_family = AF_INET6; 2099 sin6.sin6_addr = 2100 opt->ip6po_pktinfo->ipi6_addr; 2101 ia = ifa_ifwithaddr(sin6tosa(&sin6)); 2102 if (ia == NULL || 2103 (opt->ip6po_pktinfo->ipi6_ifindex && 2104 (ia->ifa_ifp->if_index != 2105 opt->ip6po_pktinfo->ipi6_ifindex))) { 2106 return(EADDRNOTAVAIL); 2107 } 2108 ia6 = (struct in6_ifaddr *)ia; 2109 if ((ia6->ia6_flags & (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY)) != 0) { 2110 return(EADDRNOTAVAIL); 2111 } 2112 2113 /* 2114 * Check if the requested source address is 2115 * indeed a unicast address assigned to the 2116 * node. 2117 */ 2118 if (IN6_IS_ADDR_MULTICAST(&opt->ip6po_pktinfo->ipi6_addr)) 2119 return(EADDRNOTAVAIL); 2120 } 2121 break; 2122 2123 case IPV6_HOPLIMIT: 2124 if (cm->cmsg_len != CMSG_LEN(sizeof(int))) 2125 return(EINVAL); 2126 2127 bcopy(CMSG_DATA(cm), &opt->ip6po_hlim, 2128 sizeof(opt->ip6po_hlim)); 2129 if (opt->ip6po_hlim < -1 || opt->ip6po_hlim > 255) 2130 return(EINVAL); 2131 break; 2132 2133 case IPV6_NEXTHOP: 2134 if (!priv) 2135 return(EPERM); 2136 2137 if (cm->cmsg_len < sizeof(u_char) || 2138 /* check if cmsg_len is large enough for sa_len */ 2139 cm->cmsg_len < CMSG_LEN(*CMSG_DATA(cm))) 2140 return(EINVAL); 2141 2142 opt->ip6po_nexthop = (struct sockaddr *)CMSG_DATA(cm); 2143 2144 break; 2145 2146 case IPV6_HOPOPTS: 2147 if (cm->cmsg_len < CMSG_LEN(sizeof(struct ip6_hbh))) 2148 return(EINVAL); 2149 opt->ip6po_hbh = (struct ip6_hbh *)CMSG_DATA(cm); 2150 if (cm->cmsg_len != 2151 CMSG_LEN((opt->ip6po_hbh->ip6h_len + 1) << 3)) 2152 return(EINVAL); 2153 break; 2154 2155 case IPV6_DSTOPTS: 2156 if (cm->cmsg_len < CMSG_LEN(sizeof(struct ip6_dest))) 2157 return(EINVAL); 2158 2159 /* 2160 * If there is no routing header yet, the destination 2161 * options header should be put on the 1st part. 2162 * Otherwise, the header should be on the 2nd part. 2163 * (See RFC 2460, section 4.1) 2164 */ 2165 if (opt->ip6po_rthdr == NULL) { 2166 opt->ip6po_dest1 = 2167 (struct ip6_dest *)CMSG_DATA(cm); 2168 if (cm->cmsg_len != 2169 CMSG_LEN((opt->ip6po_dest1->ip6d_len + 1) 2170 << 3)) 2171 return(EINVAL); 2172 } 2173 else { 2174 opt->ip6po_dest2 = 2175 (struct ip6_dest *)CMSG_DATA(cm); 2176 if (cm->cmsg_len != 2177 CMSG_LEN((opt->ip6po_dest2->ip6d_len + 1) 2178 << 3)) 2179 return(EINVAL); 2180 } 2181 break; 2182 2183 case IPV6_RTHDR: 2184 if (cm->cmsg_len < CMSG_LEN(sizeof(struct ip6_rthdr))) 2185 return(EINVAL); 2186 opt->ip6po_rthdr = (struct ip6_rthdr *)CMSG_DATA(cm); 2187 if (cm->cmsg_len != 2188 CMSG_LEN((opt->ip6po_rthdr->ip6r_len + 1) << 3)) 2189 return(EINVAL); 2190 switch (opt->ip6po_rthdr->ip6r_type) { 2191 case IPV6_RTHDR_TYPE_0: 2192 if (opt->ip6po_rthdr->ip6r_segleft == 0) 2193 return(EINVAL); 2194 break; 2195 default: 2196 return(EINVAL); 2197 } 2198 break; 2199 2200 default: 2201 return(ENOPROTOOPT); 2202 } 2203 } 2204 2205 return(0); 2206 } 2207 2208 /* 2209 * Routine called from ip6_output() to loop back a copy of an IP6 multicast 2210 * packet to the input queue of a specified interface. Note that this 2211 * calls the output routine of the loopback "driver", but with an interface 2212 * pointer that might NOT be &loif -- easier than replicating that code here. 2213 */ 2214 void 2215 ip6_mloopback(ifp, m, dst) 2216 struct ifnet *ifp; 2217 struct mbuf *m; 2218 struct sockaddr_in6 *dst; 2219 { 2220 struct mbuf *copym; 2221 struct ip6_hdr *ip6; 2222 2223 copym = m_copy(m, 0, M_COPYALL); 2224 if (copym == NULL) 2225 return; 2226 2227 /* 2228 * Make sure to deep-copy IPv6 header portion in case the data 2229 * is in an mbuf cluster, so that we can safely override the IPv6 2230 * header portion later. 2231 */ 2232 if ((copym->m_flags & M_EXT) != 0 || 2233 copym->m_len < sizeof(struct ip6_hdr)) { 2234 copym = m_pullup(copym, sizeof(struct ip6_hdr)); 2235 if (copym == NULL) 2236 return; 2237 } 2238 2239 #ifdef DIAGNOSTIC 2240 if (copym->m_len < sizeof(*ip6)) { 2241 m_freem(copym); 2242 return; 2243 } 2244 #endif 2245 2246 ip6 = mtod(copym, struct ip6_hdr *); 2247 if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src)) 2248 ip6->ip6_src.s6_addr16[1] = 0; 2249 if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst)) 2250 ip6->ip6_dst.s6_addr16[1] = 0; 2251 2252 (void)looutput(ifp, copym, (struct sockaddr *)dst, NULL); 2253 } 2254 2255 /* 2256 * Chop IPv6 header off from the payload. 2257 */ 2258 static int 2259 ip6_splithdr(m, exthdrs) 2260 struct mbuf *m; 2261 struct ip6_exthdrs *exthdrs; 2262 { 2263 struct mbuf *mh; 2264 struct ip6_hdr *ip6; 2265 2266 ip6 = mtod(m, struct ip6_hdr *); 2267 if (m->m_len > sizeof(*ip6)) { 2268 MGETHDR(mh, M_DONTWAIT, MT_HEADER); 2269 if (mh == 0) { 2270 m_freem(m); 2271 return ENOBUFS; 2272 } 2273 M_COPY_PKTHDR(mh, m); 2274 MH_ALIGN(mh, sizeof(*ip6)); 2275 m->m_flags &= ~M_PKTHDR; 2276 m->m_len -= sizeof(*ip6); 2277 m->m_data += sizeof(*ip6); 2278 mh->m_next = m; 2279 m = mh; 2280 m->m_len = sizeof(*ip6); 2281 bcopy((caddr_t)ip6, mtod(m, caddr_t), sizeof(*ip6)); 2282 } 2283 exthdrs->ip6e_ip6 = m; 2284 return 0; 2285 } 2286 2287 /* 2288 * Compute IPv6 extension header length. 2289 */ 2290 int 2291 ip6_optlen(in6p) 2292 struct in6pcb *in6p; 2293 { 2294 int len; 2295 2296 if (!in6p->in6p_outputopts) 2297 return 0; 2298 2299 len = 0; 2300 #define elen(x) \ 2301 (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0) 2302 2303 len += elen(in6p->in6p_outputopts->ip6po_hbh); 2304 len += elen(in6p->in6p_outputopts->ip6po_dest1); 2305 len += elen(in6p->in6p_outputopts->ip6po_rthdr); 2306 len += elen(in6p->in6p_outputopts->ip6po_dest2); 2307 return len; 2308 #undef elen 2309 } 2310
Indexes created Tue Sep 30 20:09:53 GMT 2025