ip6_output.c revision 1.81
1 /* $NetBSD: ip6_output.c,v 1.81 2004/03/02 02:28:28 thorpej Exp $ */ 2 /* $KAME: ip6_output.c,v 1.172 2001/03/25 09:55:56 itojun Exp $ */ 3 4 /* 5 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project. 6 * All rights reserved. 7 * 8 * Redistribution and use in source and binary forms, with or without 9 * modification, are permitted provided that the following conditions 10 * are met: 11 * 1. Redistributions of source code must retain the above copyright 12 * notice, this list of conditions and the following disclaimer. 13 * 2. Redistributions in binary form must reproduce the above copyright 14 * notice, this list of conditions and the following disclaimer in the 15 * documentation and/or other materials provided with the distribution. 16 * 3. Neither the name of the project nor the names of its contributors 17 * may be used to endorse or promote products derived from this software 18 * without specific prior written permission. 19 * 20 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND 21 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 22 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 23 * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE 24 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 25 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 26 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 27 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 28 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 29 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 30 * SUCH DAMAGE. 31 */ 32 33 /* 34 * Copyright (c) 1982, 1986, 1988, 1990, 1993 35 * The Regents of the University of California. All rights reserved. 36 * 37 * Redistribution and use in source and binary forms, with or without 38 * modification, are permitted provided that the following conditions 39 * are met: 40 * 1. Redistributions of source code must retain the above copyright 41 * notice, this list of conditions and the following disclaimer. 42 * 2. Redistributions in binary form must reproduce the above copyright 43 * notice, this list of conditions and the following disclaimer in the 44 * documentation and/or other materials provided with the distribution. 45 * 3. Neither the name of the University nor the names of its contributors 46 * may be used to endorse or promote products derived from this software 47 * without specific prior written permission. 48 * 49 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 50 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 51 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 52 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 53 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 54 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 55 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 56 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 57 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 58 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 59 * SUCH DAMAGE. 60 * 61 * @(#)ip_output.c 8.3 (Berkeley) 1/21/94 62 */ 63 64 #include <sys/cdefs.h> 65 __KERNEL_RCSID(0, "$NetBSD: ip6_output.c,v 1.81 2004/03/02 02:28:28 thorpej Exp $"); 66 67 #include "opt_inet.h" 68 #include "opt_ipsec.h" 69 #include "opt_pfil_hooks.h" 70 71 #include <sys/param.h> 72 #include <sys/malloc.h> 73 #include <sys/mbuf.h> 74 #include <sys/errno.h> 75 #include <sys/protosw.h> 76 #include <sys/socket.h> 77 #include <sys/socketvar.h> 78 #include <sys/systm.h> 79 #include <sys/proc.h> 80 81 #include <net/if.h> 82 #include <net/route.h> 83 #ifdef PFIL_HOOKS 84 #include <net/pfil.h> 85 #endif 86 87 #include <netinet/in.h> 88 #include <netinet/in_var.h> 89 #include <netinet/ip6.h> 90 #include <netinet/icmp6.h> 91 #include <netinet6/ip6_var.h> 92 #include <netinet6/in6_pcb.h> 93 #include <netinet6/nd6.h> 94 #include <netinet6/ip6protosw.h> 95 96 #ifdef IPSEC 97 #include <netinet6/ipsec.h> 98 #include <netkey/key.h> 99 #endif /* IPSEC */ 100 101 #include "loop.h" 102 103 #include <net/net_osdep.h> 104 105 #ifdef PFIL_HOOKS 106 extern struct pfil_head inet6_pfil_hook; /* XXX */ 107 #endif 108 109 struct ip6_exthdrs { 110 struct mbuf *ip6e_ip6; 111 struct mbuf *ip6e_hbh; 112 struct mbuf *ip6e_dest1; 113 struct mbuf *ip6e_rthdr; 114 struct mbuf *ip6e_dest2; 115 }; 116 117 static int ip6_pcbopts __P((struct ip6_pktopts **, struct mbuf *, 118 struct socket *)); 119 static int ip6_setmoptions __P((int, struct ip6_moptions **, struct mbuf *)); 120 static int ip6_getmoptions __P((int, struct ip6_moptions *, struct mbuf **)); 121 static int ip6_copyexthdr __P((struct mbuf **, caddr_t, int)); 122 static int ip6_insertfraghdr __P((struct mbuf *, struct mbuf *, int, 123 struct ip6_frag **)); 124 static int ip6_insert_jumboopt __P((struct ip6_exthdrs *, u_int32_t)); 125 static int ip6_splithdr __P((struct mbuf *, struct ip6_exthdrs *)); 126 static int ip6_getpmtu __P((struct route_in6 *, struct route_in6 *, 127 struct ifnet *, struct in6_addr *, u_long *, int *)); 128 129 extern struct ifnet loif[NLOOP]; 130 131 /* 132 * IP6 output. The packet in mbuf chain m contains a skeletal IP6 133 * header (with pri, len, nxt, hlim, src, dst). 134 * This function may modify ver and hlim only. 135 * The mbuf chain containing the packet will be freed. 136 * The mbuf opt, if present, will not be freed. 137 * 138 * type of "mtu": rt_rmx.rmx_mtu is u_long, ifnet.ifr_mtu is int, and 139 * nd_ifinfo.linkmtu is u_int32_t. so we use u_long to hold largest one, 140 * which is rt_rmx.rmx_mtu. 141 */ 142 int 143 ip6_output(m0, opt, ro, flags, im6o, so, ifpp) 144 struct mbuf *m0; 145 struct ip6_pktopts *opt; 146 struct route_in6 *ro; 147 int flags; 148 struct ip6_moptions *im6o; 149 struct socket *so; 150 struct ifnet **ifpp; /* XXX: just for statistics */ 151 { 152 struct ip6_hdr *ip6, *mhip6; 153 struct ifnet *ifp, *origifp; 154 struct mbuf *m = m0; 155 int hlen, tlen, len, off; 156 struct route_in6 ip6route; 157 struct sockaddr_in6 *dst; 158 int error = 0; 159 u_long mtu; 160 int alwaysfrag, dontfrag; 161 u_int32_t optlen = 0, plen = 0, unfragpartlen = 0; 162 struct ip6_exthdrs exthdrs; 163 struct in6_addr finaldst; 164 struct route_in6 *ro_pmtu = NULL; 165 int hdrsplit = 0; 166 int needipsec = 0; 167 #ifdef IPSEC 168 int needipsectun = 0; 169 struct secpolicy *sp = NULL; 170 171 ip6 = mtod(m, struct ip6_hdr *); 172 #endif /* IPSEC */ 173 174 #define MAKE_EXTHDR(hp, mp) \ 175 do { \ 176 if (hp) { \ 177 struct ip6_ext *eh = (struct ip6_ext *)(hp); \ 178 error = ip6_copyexthdr((mp), (caddr_t)(hp), \ 179 ((eh)->ip6e_len + 1) << 3); \ 180 if (error) \ 181 goto freehdrs; \ 182 } \ 183 } while (/*CONSTCOND*/ 0) 184 185 bzero(&exthdrs, sizeof(exthdrs)); 186 if (opt) { 187 /* Hop-by-Hop options header */ 188 MAKE_EXTHDR(opt->ip6po_hbh, &exthdrs.ip6e_hbh); 189 /* Destination options header(1st part) */ 190 MAKE_EXTHDR(opt->ip6po_dest1, &exthdrs.ip6e_dest1); 191 /* Routing header */ 192 MAKE_EXTHDR(opt->ip6po_rthdr, &exthdrs.ip6e_rthdr); 193 /* Destination options header(2nd part) */ 194 MAKE_EXTHDR(opt->ip6po_dest2, &exthdrs.ip6e_dest2); 195 } 196 197 #ifdef IPSEC 198 if ((flags & IPV6_FORWARDING) != 0) { 199 needipsec = 0; 200 goto skippolicycheck; 201 } 202 203 /* get a security policy for this packet */ 204 if (so == NULL) 205 sp = ipsec6_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND, 0, &error); 206 else { 207 if (IPSEC_PCB_SKIP_IPSEC(sotoinpcb_hdr(so)->inph_sp, 208 IPSEC_DIR_OUTBOUND)) { 209 needipsec = 0; 210 goto skippolicycheck; 211 } 212 sp = ipsec6_getpolicybysock(m, IPSEC_DIR_OUTBOUND, so, &error); 213 } 214 215 if (sp == NULL) { 216 ipsec6stat.out_inval++; 217 goto freehdrs; 218 } 219 220 error = 0; 221 222 /* check policy */ 223 switch (sp->policy) { 224 case IPSEC_POLICY_DISCARD: 225 /* 226 * This packet is just discarded. 227 */ 228 ipsec6stat.out_polvio++; 229 goto freehdrs; 230 231 case IPSEC_POLICY_BYPASS: 232 case IPSEC_POLICY_NONE: 233 /* no need to do IPsec. */ 234 needipsec = 0; 235 break; 236 237 case IPSEC_POLICY_IPSEC: 238 if (sp->req == NULL) { 239 /* XXX should be panic ? */ 240 printf("ip6_output: No IPsec request specified.\n"); 241 error = EINVAL; 242 goto freehdrs; 243 } 244 needipsec = 1; 245 break; 246 247 case IPSEC_POLICY_ENTRUST: 248 default: 249 printf("ip6_output: Invalid policy found. %d\n", sp->policy); 250 } 251 252 skippolicycheck:; 253 #endif /* IPSEC */ 254 255 /* 256 * Calculate the total length of the extension header chain. 257 * Keep the length of the unfragmentable part for fragmentation. 258 */ 259 optlen = 0; 260 if (exthdrs.ip6e_hbh) optlen += exthdrs.ip6e_hbh->m_len; 261 if (exthdrs.ip6e_dest1) optlen += exthdrs.ip6e_dest1->m_len; 262 if (exthdrs.ip6e_rthdr) optlen += exthdrs.ip6e_rthdr->m_len; 263 unfragpartlen = optlen + sizeof(struct ip6_hdr); 264 /* NOTE: we don't add AH/ESP length here. do that later. */ 265 if (exthdrs.ip6e_dest2) optlen += exthdrs.ip6e_dest2->m_len; 266 267 /* 268 * If we need IPsec, or there is at least one extension header, 269 * separate IP6 header from the payload. 270 */ 271 if ((needipsec || optlen) && !hdrsplit) { 272 if ((error = ip6_splithdr(m, &exthdrs)) != 0) { 273 m = NULL; 274 goto freehdrs; 275 } 276 m = exthdrs.ip6e_ip6; 277 hdrsplit++; 278 } 279 280 /* adjust pointer */ 281 ip6 = mtod(m, struct ip6_hdr *); 282 283 /* adjust mbuf packet header length */ 284 m->m_pkthdr.len += optlen; 285 plen = m->m_pkthdr.len - sizeof(*ip6); 286 287 /* If this is a jumbo payload, insert a jumbo payload option. */ 288 if (plen > IPV6_MAXPACKET) { 289 if (!hdrsplit) { 290 if ((error = ip6_splithdr(m, &exthdrs)) != 0) { 291 m = NULL; 292 goto freehdrs; 293 } 294 m = exthdrs.ip6e_ip6; 295 hdrsplit++; 296 } 297 /* adjust pointer */ 298 ip6 = mtod(m, struct ip6_hdr *); 299 if ((error = ip6_insert_jumboopt(&exthdrs, plen)) != 0) 300 goto freehdrs; 301 ip6->ip6_plen = 0; 302 } else 303 ip6->ip6_plen = htons(plen); 304 305 /* 306 * Concatenate headers and fill in next header fields. 307 * Here we have, on "m" 308 * IPv6 payload 309 * and we insert headers accordingly. Finally, we should be getting: 310 * IPv6 hbh dest1 rthdr ah* [esp* dest2 payload] 311 * 312 * during the header composing process, "m" points to IPv6 header. 313 * "mprev" points to an extension header prior to esp. 314 */ 315 { 316 u_char *nexthdrp = &ip6->ip6_nxt; 317 struct mbuf *mprev = m; 318 319 /* 320 * we treat dest2 specially. this makes IPsec processing 321 * much easier. the goal here is to make mprev point the 322 * mbuf prior to dest2. 323 * 324 * result: IPv6 dest2 payload 325 * m and mprev will point to IPv6 header. 326 */ 327 if (exthdrs.ip6e_dest2) { 328 if (!hdrsplit) 329 panic("assumption failed: hdr not split"); 330 exthdrs.ip6e_dest2->m_next = m->m_next; 331 m->m_next = exthdrs.ip6e_dest2; 332 *mtod(exthdrs.ip6e_dest2, u_char *) = ip6->ip6_nxt; 333 ip6->ip6_nxt = IPPROTO_DSTOPTS; 334 } 335 336 #define MAKE_CHAIN(m, mp, p, i)\ 337 do {\ 338 if (m) {\ 339 if (!hdrsplit) \ 340 panic("assumption failed: hdr not split"); \ 341 *mtod((m), u_char *) = *(p);\ 342 *(p) = (i);\ 343 p = mtod((m), u_char *);\ 344 (m)->m_next = (mp)->m_next;\ 345 (mp)->m_next = (m);\ 346 (mp) = (m);\ 347 }\ 348 } while (/*CONSTCOND*/ 0) 349 /* 350 * result: IPv6 hbh dest1 rthdr dest2 payload 351 * m will point to IPv6 header. mprev will point to the 352 * extension header prior to dest2 (rthdr in the above case). 353 */ 354 MAKE_CHAIN(exthdrs.ip6e_hbh, mprev, nexthdrp, IPPROTO_HOPOPTS); 355 MAKE_CHAIN(exthdrs.ip6e_dest1, mprev, nexthdrp, 356 IPPROTO_DSTOPTS); 357 MAKE_CHAIN(exthdrs.ip6e_rthdr, mprev, nexthdrp, 358 IPPROTO_ROUTING); 359 360 #ifdef IPSEC 361 if (!needipsec) 362 goto skip_ipsec2; 363 364 /* 365 * pointers after IPsec headers are not valid any more. 366 * other pointers need a great care too. 367 * (IPsec routines should not mangle mbufs prior to AH/ESP) 368 */ 369 exthdrs.ip6e_dest2 = NULL; 370 371 { 372 struct ip6_rthdr *rh = NULL; 373 int segleft_org = 0; 374 struct ipsec_output_state state; 375 376 if (exthdrs.ip6e_rthdr) { 377 rh = mtod(exthdrs.ip6e_rthdr, struct ip6_rthdr *); 378 segleft_org = rh->ip6r_segleft; 379 rh->ip6r_segleft = 0; 380 } 381 382 bzero(&state, sizeof(state)); 383 state.m = m; 384 error = ipsec6_output_trans(&state, nexthdrp, mprev, sp, flags, 385 &needipsectun); 386 m = state.m; 387 if (error) { 388 /* mbuf is already reclaimed in ipsec6_output_trans. */ 389 m = NULL; 390 switch (error) { 391 case EHOSTUNREACH: 392 case ENETUNREACH: 393 case EMSGSIZE: 394 case ENOBUFS: 395 case ENOMEM: 396 break; 397 default: 398 printf("ip6_output (ipsec): error code %d\n", error); 399 /* FALLTHROUGH */ 400 case ENOENT: 401 /* don't show these error codes to the user */ 402 error = 0; 403 break; 404 } 405 goto bad; 406 } 407 if (exthdrs.ip6e_rthdr) { 408 /* ah6_output doesn't modify mbuf chain */ 409 rh->ip6r_segleft = segleft_org; 410 } 411 } 412 skip_ipsec2:; 413 #endif 414 } 415 416 /* 417 * If there is a routing header, replace destination address field 418 * with the first hop of the routing header. 419 */ 420 if (exthdrs.ip6e_rthdr) { 421 struct ip6_rthdr *rh; 422 struct ip6_rthdr0 *rh0; 423 struct in6_addr *addr; 424 425 rh = (struct ip6_rthdr *)(mtod(exthdrs.ip6e_rthdr, 426 struct ip6_rthdr *)); 427 finaldst = ip6->ip6_dst; 428 switch (rh->ip6r_type) { 429 case IPV6_RTHDR_TYPE_0: 430 rh0 = (struct ip6_rthdr0 *)rh; 431 addr = (struct in6_addr *)(rh0 + 1); 432 ip6->ip6_dst = addr[0]; 433 bcopy(&addr[1], &addr[0], 434 sizeof(struct in6_addr) * (rh0->ip6r0_segleft - 1)); 435 addr[rh0->ip6r0_segleft - 1] = finaldst; 436 break; 437 default: /* is it possible? */ 438 error = EINVAL; 439 goto bad; 440 } 441 } 442 443 /* Source address validation */ 444 if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) && 445 (flags & IPV6_UNSPECSRC) == 0) { 446 error = EOPNOTSUPP; 447 ip6stat.ip6s_badscope++; 448 goto bad; 449 } 450 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) { 451 error = EOPNOTSUPP; 452 ip6stat.ip6s_badscope++; 453 goto bad; 454 } 455 456 ip6stat.ip6s_localout++; 457 458 /* 459 * Route packet. 460 */ 461 /* initialize cached route */ 462 if (ro == 0) { 463 ro = &ip6route; 464 bzero((caddr_t)ro, sizeof(*ro)); 465 } 466 ro_pmtu = ro; 467 if (opt && opt->ip6po_rthdr) 468 ro = &opt->ip6po_route; 469 dst = (struct sockaddr_in6 *)&ro->ro_dst; 470 /* 471 * If there is a cached route, 472 * check that it is to the same destination 473 * and is still up. If not, free it and try again. 474 */ 475 if (ro->ro_rt && ((ro->ro_rt->rt_flags & RTF_UP) == 0 || 476 dst->sin6_family != AF_INET6 || 477 !IN6_ARE_ADDR_EQUAL(&dst->sin6_addr, &ip6->ip6_dst))) { 478 RTFREE(ro->ro_rt); 479 ro->ro_rt = (struct rtentry *)0; 480 } 481 if (ro->ro_rt == 0) { 482 bzero(dst, sizeof(*dst)); 483 dst->sin6_family = AF_INET6; 484 dst->sin6_len = sizeof(struct sockaddr_in6); 485 dst->sin6_addr = ip6->ip6_dst; 486 } 487 #ifdef IPSEC 488 if (needipsec && needipsectun) { 489 struct ipsec_output_state state; 490 491 /* 492 * All the extension headers will become inaccessible 493 * (since they can be encrypted). 494 * Don't panic, we need no more updates to extension headers 495 * on inner IPv6 packet (since they are now encapsulated). 496 * 497 * IPv6 [ESP|AH] IPv6 [extension headers] payload 498 */ 499 bzero(&exthdrs, sizeof(exthdrs)); 500 exthdrs.ip6e_ip6 = m; 501 502 bzero(&state, sizeof(state)); 503 state.m = m; 504 state.ro = (struct route *)ro; 505 state.dst = (struct sockaddr *)dst; 506 507 error = ipsec6_output_tunnel(&state, sp, flags); 508 509 m = state.m; 510 ro = (struct route_in6 *)state.ro; 511 dst = (struct sockaddr_in6 *)state.dst; 512 if (error) { 513 /* mbuf is already reclaimed in ipsec6_output_tunnel. */ 514 m0 = m = NULL; 515 m = NULL; 516 switch (error) { 517 case EHOSTUNREACH: 518 case ENETUNREACH: 519 case EMSGSIZE: 520 case ENOBUFS: 521 case ENOMEM: 522 break; 523 default: 524 printf("ip6_output (ipsec): error code %d\n", error); 525 /* FALLTHROUGH */ 526 case ENOENT: 527 /* don't show these error codes to the user */ 528 error = 0; 529 break; 530 } 531 goto bad; 532 } 533 534 exthdrs.ip6e_ip6 = m; 535 } 536 #endif /* IPSEC */ 537 538 if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) { 539 /* Unicast */ 540 541 #define ifatoia6(ifa) ((struct in6_ifaddr *)(ifa)) 542 #define sin6tosa(sin6) ((struct sockaddr *)(sin6)) 543 /* xxx 544 * interface selection comes here 545 * if an interface is specified from an upper layer, 546 * ifp must point it. 547 */ 548 if (ro->ro_rt == 0) { 549 /* 550 * non-bsdi always clone routes, if parent is 551 * PRF_CLONING. 552 */ 553 rtalloc((struct route *)ro); 554 } 555 if (ro->ro_rt == 0) { 556 ip6stat.ip6s_noroute++; 557 error = EHOSTUNREACH; 558 /* XXX in6_ifstat_inc(ifp, ifs6_out_discard); */ 559 goto bad; 560 } 561 ifp = ro->ro_rt->rt_ifp; 562 ro->ro_rt->rt_use++; 563 if (ro->ro_rt->rt_flags & RTF_GATEWAY) 564 dst = (struct sockaddr_in6 *)ro->ro_rt->rt_gateway; 565 m->m_flags &= ~(M_BCAST | M_MCAST); /* just in case */ 566 567 in6_ifstat_inc(ifp, ifs6_out_request); 568 569 /* 570 * Check if the outgoing interface conflicts with 571 * the interface specified by ifi6_ifindex (if specified). 572 * Note that loopback interface is always okay. 573 * (this may happen when we are sending a packet to one of 574 * our own addresses.) 575 */ 576 if (opt && opt->ip6po_pktinfo && 577 opt->ip6po_pktinfo->ipi6_ifindex) { 578 if (!(ifp->if_flags & IFF_LOOPBACK) && 579 ifp->if_index != opt->ip6po_pktinfo->ipi6_ifindex) { 580 ip6stat.ip6s_noroute++; 581 in6_ifstat_inc(ifp, ifs6_out_discard); 582 error = EHOSTUNREACH; 583 goto bad; 584 } 585 } 586 587 if (opt && opt->ip6po_hlim != -1) 588 ip6->ip6_hlim = opt->ip6po_hlim & 0xff; 589 } else { 590 /* Multicast */ 591 struct in6_multi *in6m; 592 593 m->m_flags = (m->m_flags & ~M_BCAST) | M_MCAST; 594 595 /* 596 * See if the caller provided any multicast options 597 */ 598 ifp = NULL; 599 if (im6o != NULL) { 600 ip6->ip6_hlim = im6o->im6o_multicast_hlim; 601 if (im6o->im6o_multicast_ifp != NULL) 602 ifp = im6o->im6o_multicast_ifp; 603 } else 604 ip6->ip6_hlim = ip6_defmcasthlim; 605 606 /* 607 * See if the caller provided the outgoing interface 608 * as an ancillary data. 609 * Boundary check for ifindex is assumed to be already done. 610 */ 611 if (opt && opt->ip6po_pktinfo && opt->ip6po_pktinfo->ipi6_ifindex) 612 ifp = ifindex2ifnet[opt->ip6po_pktinfo->ipi6_ifindex]; 613 614 /* 615 * If the destination is a node-local scope multicast, 616 * the packet should be loop-backed only. 617 */ 618 if (IN6_IS_ADDR_MC_NODELOCAL(&ip6->ip6_dst)) { 619 /* 620 * If the outgoing interface is already specified, 621 * it should be a loopback interface. 622 */ 623 if (ifp && (ifp->if_flags & IFF_LOOPBACK) == 0) { 624 ip6stat.ip6s_badscope++; 625 error = ENETUNREACH; /* XXX: better error? */ 626 /* XXX correct ifp? */ 627 in6_ifstat_inc(ifp, ifs6_out_discard); 628 goto bad; 629 } else { 630 ifp = &loif[0]; 631 } 632 } 633 634 if (opt && opt->ip6po_hlim != -1) 635 ip6->ip6_hlim = opt->ip6po_hlim & 0xff; 636 637 /* 638 * If caller did not provide an interface lookup a 639 * default in the routing table. This is either a 640 * default for the speicfied group (i.e. a host 641 * route), or a multicast default (a route for the 642 * ``net'' ff00::/8). 643 */ 644 if (ifp == NULL) { 645 if (ro->ro_rt == 0) { 646 ro->ro_rt = rtalloc1((struct sockaddr *) 647 &ro->ro_dst, 0); 648 } 649 if (ro->ro_rt == 0) { 650 ip6stat.ip6s_noroute++; 651 error = EHOSTUNREACH; 652 /* XXX in6_ifstat_inc(ifp, ifs6_out_discard) */ 653 goto bad; 654 } 655 ifp = ro->ro_rt->rt_ifp; 656 ro->ro_rt->rt_use++; 657 } 658 659 if ((flags & IPV6_FORWARDING) == 0) 660 in6_ifstat_inc(ifp, ifs6_out_request); 661 in6_ifstat_inc(ifp, ifs6_out_mcast); 662 663 /* 664 * Confirm that the outgoing interface supports multicast. 665 */ 666 if ((ifp->if_flags & IFF_MULTICAST) == 0) { 667 ip6stat.ip6s_noroute++; 668 in6_ifstat_inc(ifp, ifs6_out_discard); 669 error = ENETUNREACH; 670 goto bad; 671 } 672 IN6_LOOKUP_MULTI(ip6->ip6_dst, ifp, in6m); 673 if (in6m != NULL && 674 (im6o == NULL || im6o->im6o_multicast_loop)) { 675 /* 676 * If we belong to the destination multicast group 677 * on the outgoing interface, and the caller did not 678 * forbid loopback, loop back a copy. 679 */ 680 ip6_mloopback(ifp, m, dst); 681 } else { 682 /* 683 * If we are acting as a multicast router, perform 684 * multicast forwarding as if the packet had just 685 * arrived on the interface to which we are about 686 * to send. The multicast forwarding function 687 * recursively calls this function, using the 688 * IPV6_FORWARDING flag to prevent infinite recursion. 689 * 690 * Multicasts that are looped back by ip6_mloopback(), 691 * above, will be forwarded by the ip6_input() routine, 692 * if necessary. 693 */ 694 if (ip6_mrouter && (flags & IPV6_FORWARDING) == 0) { 695 if (ip6_mforward(ip6, ifp, m) != 0) { 696 m_freem(m); 697 goto done; 698 } 699 } 700 } 701 /* 702 * Multicasts with a hoplimit of zero may be looped back, 703 * above, but must not be transmitted on a network. 704 * Also, multicasts addressed to the loopback interface 705 * are not sent -- the above call to ip6_mloopback() will 706 * loop back a copy if this host actually belongs to the 707 * destination group on the loopback interface. 708 */ 709 if (ip6->ip6_hlim == 0 || (ifp->if_flags & IFF_LOOPBACK)) { 710 m_freem(m); 711 goto done; 712 } 713 } 714 715 /* 716 * Fill the outgoing inteface to tell the upper layer 717 * to increment per-interface statistics. 718 */ 719 if (ifpp) 720 *ifpp = ifp; 721 722 /* Determine path MTU. */ 723 if ((error = ip6_getpmtu(ro_pmtu, ro, ifp, &finaldst, &mtu, 724 &alwaysfrag)) != 0) 725 goto bad; 726 727 /* 728 * The caller of this function may specify to use the minimum MTU 729 * in some cases. 730 */ 731 if (mtu > IPV6_MMTU) { 732 if ((flags & IPV6_MINMTU)) 733 mtu = IPV6_MMTU; 734 } 735 736 /* Fake scoped addresses */ 737 if ((ifp->if_flags & IFF_LOOPBACK) != 0) { 738 /* 739 * If source or destination address is a scoped address, and 740 * the packet is going to be sent to a loopback interface, 741 * we should keep the original interface. 742 */ 743 744 /* 745 * XXX: this is a very experimental and temporary solution. 746 * We eventually have sockaddr_in6 and use the sin6_scope_id 747 * field of the structure here. 748 * We rely on the consistency between two scope zone ids 749 * of source add destination, which should already be assured 750 * Larger scopes than link will be supported in the near 751 * future. 752 */ 753 origifp = NULL; 754 if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src)) 755 origifp = ifindex2ifnet[ntohs(ip6->ip6_src.s6_addr16[1])]; 756 else if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst)) 757 origifp = ifindex2ifnet[ntohs(ip6->ip6_dst.s6_addr16[1])]; 758 /* 759 * XXX: origifp can be NULL even in those two cases above. 760 * For example, if we remove the (only) link-local address 761 * from the loopback interface, and try to send a link-local 762 * address without link-id information. Then the source 763 * address is ::1, and the destination address is the 764 * link-local address with its s6_addr16[1] being zero. 765 * What is worse, if the packet goes to the loopback interface 766 * by a default rejected route, the null pointer would be 767 * passed to looutput, and the kernel would hang. 768 * The following last resort would prevent such disaster. 769 */ 770 if (origifp == NULL) 771 origifp = ifp; 772 } else 773 origifp = ifp; 774 if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src)) 775 ip6->ip6_src.s6_addr16[1] = 0; 776 if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst)) 777 ip6->ip6_dst.s6_addr16[1] = 0; 778 779 /* 780 * If the outgoing packet contains a hop-by-hop options header, 781 * it must be examined and processed even by the source node. 782 * (RFC 2460, section 4.) 783 */ 784 if (exthdrs.ip6e_hbh) { 785 struct ip6_hbh *hbh = mtod(exthdrs.ip6e_hbh, struct ip6_hbh *); 786 u_int32_t dummy1; /* XXX unused */ 787 u_int32_t dummy2; /* XXX unused */ 788 789 /* 790 * XXX: if we have to send an ICMPv6 error to the sender, 791 * we need the M_LOOP flag since icmp6_error() expects 792 * the IPv6 and the hop-by-hop options header are 793 * continuous unless the flag is set. 794 */ 795 m->m_flags |= M_LOOP; 796 m->m_pkthdr.rcvif = ifp; 797 if (ip6_process_hopopts(m, (u_int8_t *)(hbh + 1), 798 ((hbh->ip6h_len + 1) << 3) - sizeof(struct ip6_hbh), 799 &dummy1, &dummy2) < 0) { 800 /* m was already freed at this point */ 801 error = EINVAL;/* better error? */ 802 goto done; 803 } 804 m->m_flags &= ~M_LOOP; /* XXX */ 805 m->m_pkthdr.rcvif = NULL; 806 } 807 808 #ifdef PFIL_HOOKS 809 /* 810 * Run through list of hooks for output packets. 811 */ 812 if ((error = pfil_run_hooks(&inet6_pfil_hook, &m, ifp, PFIL_OUT)) != 0) 813 goto done; 814 if (m == NULL) 815 goto done; 816 ip6 = mtod(m, struct ip6_hdr *); 817 #endif /* PFIL_HOOKS */ 818 /* 819 * Send the packet to the outgoing interface. 820 * If necessary, do IPv6 fragmentation before sending. 821 * 822 * the logic here is rather complex: 823 * 1: normal case (dontfrag == 0, alwaysfrag == 0) 824 * 1-a: send as is if tlen <= path mtu 825 * 1-b: fragment if tlen > path mtu 826 * 827 * 2: if user asks us not to fragment (dontfrag == 1) 828 * 2-a: send as is if tlen <= interface mtu 829 * 2-b: error if tlen > interface mtu 830 * 831 * 3: if we always need to attach fragment header (alwaysfrag == 1) 832 * always fragment 833 * 834 * 4: if dontfrag == 1 && alwaysfrag == 1 835 * error, as we cannot handle this conflicting request 836 */ 837 tlen = m->m_pkthdr.len; 838 839 dontfrag = 0; 840 if (dontfrag && alwaysfrag) { /* case 4 */ 841 /* conflicting request - can't transmit */ 842 error = EMSGSIZE; 843 goto bad; 844 } 845 if (dontfrag && tlen > IN6_LINKMTU(ifp)) { /* case 2-b */ 846 /* 847 * Even if the DONTFRAG option is specified, we cannot send the 848 * packet when the data length is larger than the MTU of the 849 * outgoing interface. 850 * Notify the error by sending IPV6_PATHMTU ancillary data as 851 * well as returning an error code (the latter is not described 852 * in the API spec.) 853 */ 854 u_int32_t mtu32; 855 struct ip6ctlparam ip6cp; 856 857 mtu32 = (u_int32_t)mtu; 858 bzero(&ip6cp, sizeof(ip6cp)); 859 ip6cp.ip6c_cmdarg = (void *)&mtu32; 860 pfctlinput2(PRC_MSGSIZE, (struct sockaddr *)&ro_pmtu->ro_dst, 861 (void *)&ip6cp); 862 863 error = EMSGSIZE; 864 goto bad; 865 } 866 867 /* 868 * transmit packet without fragmentation 869 */ 870 if (dontfrag || (!alwaysfrag && tlen <= mtu)) { /* case 1-a and 2-a */ 871 struct in6_ifaddr *ia6; 872 873 ip6 = mtod(m, struct ip6_hdr *); 874 ia6 = in6_ifawithifp(ifp, &ip6->ip6_src); 875 if (ia6) { 876 /* Record statistics for this interface address. */ 877 ia6->ia_ifa.ifa_data.ifad_outbytes += m->m_pkthdr.len; 878 } 879 #ifdef IPSEC 880 /* clean ipsec history once it goes out of the node */ 881 ipsec_delaux(m); 882 #endif 883 error = nd6_output(ifp, origifp, m, dst, ro->ro_rt); 884 goto done; 885 } 886 887 /* 888 * try to fragment the packet. case 1-b and 3 889 */ 890 if (mtu < IPV6_MMTU) { 891 /* path MTU cannot be less than IPV6_MMTU */ 892 error = EMSGSIZE; 893 in6_ifstat_inc(ifp, ifs6_out_fragfail); 894 goto bad; 895 } else if (ip6->ip6_plen == 0) { 896 /* jumbo payload cannot be fragmented */ 897 error = EMSGSIZE; 898 in6_ifstat_inc(ifp, ifs6_out_fragfail); 899 goto bad; 900 } else { 901 struct mbuf **mnext, *m_frgpart; 902 struct ip6_frag *ip6f; 903 u_int32_t id = htonl(ip6_randomid()); 904 u_char nextproto; 905 struct ip6ctlparam ip6cp; 906 u_int32_t mtu32; 907 908 /* 909 * Too large for the destination or interface; 910 * fragment if possible. 911 * Must be able to put at least 8 bytes per fragment. 912 */ 913 hlen = unfragpartlen; 914 if (mtu > IPV6_MAXPACKET) 915 mtu = IPV6_MAXPACKET; 916 917 /* Notify a proper path MTU to applications. */ 918 mtu32 = (u_int32_t)mtu; 919 bzero(&ip6cp, sizeof(ip6cp)); 920 ip6cp.ip6c_cmdarg = (void *)&mtu32; 921 pfctlinput2(PRC_MSGSIZE, (struct sockaddr *)&ro_pmtu->ro_dst, 922 (void *)&ip6cp); 923 924 len = (mtu - hlen - sizeof(struct ip6_frag)) & ~7; 925 if (len < 8) { 926 error = EMSGSIZE; 927 in6_ifstat_inc(ifp, ifs6_out_fragfail); 928 goto bad; 929 } 930 931 mnext = &m->m_nextpkt; 932 933 /* 934 * Change the next header field of the last header in the 935 * unfragmentable part. 936 */ 937 if (exthdrs.ip6e_rthdr) { 938 nextproto = *mtod(exthdrs.ip6e_rthdr, u_char *); 939 *mtod(exthdrs.ip6e_rthdr, u_char *) = IPPROTO_FRAGMENT; 940 } else if (exthdrs.ip6e_dest1) { 941 nextproto = *mtod(exthdrs.ip6e_dest1, u_char *); 942 *mtod(exthdrs.ip6e_dest1, u_char *) = IPPROTO_FRAGMENT; 943 } else if (exthdrs.ip6e_hbh) { 944 nextproto = *mtod(exthdrs.ip6e_hbh, u_char *); 945 *mtod(exthdrs.ip6e_hbh, u_char *) = IPPROTO_FRAGMENT; 946 } else { 947 nextproto = ip6->ip6_nxt; 948 ip6->ip6_nxt = IPPROTO_FRAGMENT; 949 } 950 951 /* 952 * Loop through length of segment after first fragment, 953 * make new header and copy data of each part and link onto 954 * chain. 955 */ 956 m0 = m; 957 for (off = hlen; off < tlen; off += len) { 958 struct mbuf *mlast; 959 960 MGETHDR(m, M_DONTWAIT, MT_HEADER); 961 if (!m) { 962 error = ENOBUFS; 963 ip6stat.ip6s_odropped++; 964 goto sendorfree; 965 } 966 m->m_pkthdr.rcvif = NULL; 967 m->m_flags = m0->m_flags & M_COPYFLAGS; 968 *mnext = m; 969 mnext = &m->m_nextpkt; 970 m->m_data += max_linkhdr; 971 mhip6 = mtod(m, struct ip6_hdr *); 972 *mhip6 = *ip6; 973 m->m_len = sizeof(*mhip6); 974 error = ip6_insertfraghdr(m0, m, hlen, &ip6f); 975 if (error) { 976 ip6stat.ip6s_odropped++; 977 goto sendorfree; 978 } 979 ip6f->ip6f_offlg = htons((u_int16_t)((off - hlen) & ~7)); 980 if (off + len >= tlen) 981 len = tlen - off; 982 else 983 ip6f->ip6f_offlg |= IP6F_MORE_FRAG; 984 mhip6->ip6_plen = htons((u_int16_t)(len + hlen + 985 sizeof(*ip6f) - sizeof(struct ip6_hdr))); 986 if ((m_frgpart = m_copy(m0, off, len)) == 0) { 987 error = ENOBUFS; 988 ip6stat.ip6s_odropped++; 989 goto sendorfree; 990 } 991 for (mlast = m; mlast->m_next; mlast = mlast->m_next) 992 ; 993 mlast->m_next = m_frgpart; 994 m->m_pkthdr.len = len + hlen + sizeof(*ip6f); 995 m->m_pkthdr.rcvif = (struct ifnet *)0; 996 ip6f->ip6f_reserved = 0; 997 ip6f->ip6f_ident = id; 998 ip6f->ip6f_nxt = nextproto; 999 ip6stat.ip6s_ofragments++; 1000 in6_ifstat_inc(ifp, ifs6_out_fragcreat); 1001 } 1002 1003 in6_ifstat_inc(ifp, ifs6_out_fragok); 1004 } 1005 1006 /* 1007 * Remove leading garbages. 1008 */ 1009 sendorfree: 1010 m = m0->m_nextpkt; 1011 m0->m_nextpkt = 0; 1012 m_freem(m0); 1013 for (m0 = m; m; m = m0) { 1014 m0 = m->m_nextpkt; 1015 m->m_nextpkt = 0; 1016 if (error == 0) { 1017 struct in6_ifaddr *ia6; 1018 ip6 = mtod(m, struct ip6_hdr *); 1019 ia6 = in6_ifawithifp(ifp, &ip6->ip6_src); 1020 if (ia6) { 1021 /* 1022 * Record statistics for this interface 1023 * address. 1024 */ 1025 ia6->ia_ifa.ifa_data.ifad_outbytes += 1026 m->m_pkthdr.len; 1027 } 1028 #ifdef IPSEC 1029 /* clean ipsec history once it goes out of the node */ 1030 ipsec_delaux(m); 1031 #endif 1032 error = nd6_output(ifp, origifp, m, dst, ro->ro_rt); 1033 } else 1034 m_freem(m); 1035 } 1036 1037 if (error == 0) 1038 ip6stat.ip6s_fragmented++; 1039 1040 done: 1041 if (ro == &ip6route && ro->ro_rt) { /* brace necessary for RTFREE */ 1042 RTFREE(ro->ro_rt); 1043 } else if (ro_pmtu == &ip6route && ro_pmtu->ro_rt) { 1044 RTFREE(ro_pmtu->ro_rt); 1045 } 1046 1047 #ifdef IPSEC 1048 if (sp != NULL) 1049 key_freesp(sp); 1050 #endif /* IPSEC */ 1051 1052 return (error); 1053 1054 freehdrs: 1055 m_freem(exthdrs.ip6e_hbh); /* m_freem will check if mbuf is 0 */ 1056 m_freem(exthdrs.ip6e_dest1); 1057 m_freem(exthdrs.ip6e_rthdr); 1058 m_freem(exthdrs.ip6e_dest2); 1059 /* FALLTHROUGH */ 1060 bad: 1061 m_freem(m); 1062 goto done; 1063 } 1064 1065 static int 1066 ip6_copyexthdr(mp, hdr, hlen) 1067 struct mbuf **mp; 1068 caddr_t hdr; 1069 int hlen; 1070 { 1071 struct mbuf *m; 1072 1073 if (hlen > MCLBYTES) 1074 return (ENOBUFS); /* XXX */ 1075 1076 MGET(m, M_DONTWAIT, MT_DATA); 1077 if (!m) 1078 return (ENOBUFS); 1079 1080 if (hlen > MLEN) { 1081 MCLGET(m, M_DONTWAIT); 1082 if ((m->m_flags & M_EXT) == 0) { 1083 m_free(m); 1084 return (ENOBUFS); 1085 } 1086 } 1087 m->m_len = hlen; 1088 if (hdr) 1089 bcopy(hdr, mtod(m, caddr_t), hlen); 1090 1091 *mp = m; 1092 return (0); 1093 } 1094 1095 /* 1096 * Insert jumbo payload option. 1097 */ 1098 static int 1099 ip6_insert_jumboopt(exthdrs, plen) 1100 struct ip6_exthdrs *exthdrs; 1101 u_int32_t plen; 1102 { 1103 struct mbuf *mopt; 1104 u_int8_t *optbuf; 1105 u_int32_t v; 1106 1107 #define JUMBOOPTLEN 8 /* length of jumbo payload option and padding */ 1108 1109 /* 1110 * If there is no hop-by-hop options header, allocate new one. 1111 * If there is one but it doesn't have enough space to store the 1112 * jumbo payload option, allocate a cluster to store the whole options. 1113 * Otherwise, use it to store the options. 1114 */ 1115 if (exthdrs->ip6e_hbh == 0) { 1116 MGET(mopt, M_DONTWAIT, MT_DATA); 1117 if (mopt == 0) 1118 return (ENOBUFS); 1119 mopt->m_len = JUMBOOPTLEN; 1120 optbuf = mtod(mopt, u_int8_t *); 1121 optbuf[1] = 0; /* = ((JUMBOOPTLEN) >> 3) - 1 */ 1122 exthdrs->ip6e_hbh = mopt; 1123 } else { 1124 struct ip6_hbh *hbh; 1125 1126 mopt = exthdrs->ip6e_hbh; 1127 if (M_TRAILINGSPACE(mopt) < JUMBOOPTLEN) { 1128 /* 1129 * XXX assumption: 1130 * - exthdrs->ip6e_hbh is not referenced from places 1131 * other than exthdrs. 1132 * - exthdrs->ip6e_hbh is not an mbuf chain. 1133 */ 1134 int oldoptlen = mopt->m_len; 1135 struct mbuf *n; 1136 1137 /* 1138 * XXX: give up if the whole (new) hbh header does 1139 * not fit even in an mbuf cluster. 1140 */ 1141 if (oldoptlen + JUMBOOPTLEN > MCLBYTES) 1142 return (ENOBUFS); 1143 1144 /* 1145 * As a consequence, we must always prepare a cluster 1146 * at this point. 1147 */ 1148 MGET(n, M_DONTWAIT, MT_DATA); 1149 if (n) { 1150 MCLGET(n, M_DONTWAIT); 1151 if ((n->m_flags & M_EXT) == 0) { 1152 m_freem(n); 1153 n = NULL; 1154 } 1155 } 1156 if (!n) 1157 return (ENOBUFS); 1158 n->m_len = oldoptlen + JUMBOOPTLEN; 1159 bcopy(mtod(mopt, caddr_t), mtod(n, caddr_t), 1160 oldoptlen); 1161 optbuf = mtod(n, u_int8_t *) + oldoptlen; 1162 m_freem(mopt); 1163 mopt = exthdrs->ip6e_hbh = n; 1164 } else { 1165 optbuf = mtod(mopt, u_int8_t *) + mopt->m_len; 1166 mopt->m_len += JUMBOOPTLEN; 1167 } 1168 optbuf[0] = IP6OPT_PADN; 1169 optbuf[1] = 0; 1170 1171 /* 1172 * Adjust the header length according to the pad and 1173 * the jumbo payload option. 1174 */ 1175 hbh = mtod(mopt, struct ip6_hbh *); 1176 hbh->ip6h_len += (JUMBOOPTLEN >> 3); 1177 } 1178 1179 /* fill in the option. */ 1180 optbuf[2] = IP6OPT_JUMBO; 1181 optbuf[3] = 4; 1182 v = (u_int32_t)htonl(plen + JUMBOOPTLEN); 1183 bcopy(&v, &optbuf[4], sizeof(u_int32_t)); 1184 1185 /* finally, adjust the packet header length */ 1186 exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN; 1187 1188 return (0); 1189 #undef JUMBOOPTLEN 1190 } 1191 1192 /* 1193 * Insert fragment header and copy unfragmentable header portions. 1194 */ 1195 static int 1196 ip6_insertfraghdr(m0, m, hlen, frghdrp) 1197 struct mbuf *m0, *m; 1198 int hlen; 1199 struct ip6_frag **frghdrp; 1200 { 1201 struct mbuf *n, *mlast; 1202 1203 if (hlen > sizeof(struct ip6_hdr)) { 1204 n = m_copym(m0, sizeof(struct ip6_hdr), 1205 hlen - sizeof(struct ip6_hdr), M_DONTWAIT); 1206 if (n == 0) 1207 return (ENOBUFS); 1208 m->m_next = n; 1209 } else 1210 n = m; 1211 1212 /* Search for the last mbuf of unfragmentable part. */ 1213 for (mlast = n; mlast->m_next; mlast = mlast->m_next) 1214 ; 1215 1216 if ((mlast->m_flags & M_EXT) == 0 && 1217 M_TRAILINGSPACE(mlast) >= sizeof(struct ip6_frag)) { 1218 /* use the trailing space of the last mbuf for the fragment hdr */ 1219 *frghdrp = (struct ip6_frag *)(mtod(mlast, caddr_t) + 1220 mlast->m_len); 1221 mlast->m_len += sizeof(struct ip6_frag); 1222 m->m_pkthdr.len += sizeof(struct ip6_frag); 1223 } else { 1224 /* allocate a new mbuf for the fragment header */ 1225 struct mbuf *mfrg; 1226 1227 MGET(mfrg, M_DONTWAIT, MT_DATA); 1228 if (mfrg == 0) 1229 return (ENOBUFS); 1230 mfrg->m_len = sizeof(struct ip6_frag); 1231 *frghdrp = mtod(mfrg, struct ip6_frag *); 1232 mlast->m_next = mfrg; 1233 } 1234 1235 return (0); 1236 } 1237 1238 static int 1239 ip6_getpmtu(ro_pmtu, ro, ifp, dst, mtup, alwaysfragp) 1240 struct route_in6 *ro_pmtu, *ro; 1241 struct ifnet *ifp; 1242 struct in6_addr *dst; 1243 u_long *mtup; 1244 int *alwaysfragp; 1245 { 1246 u_int32_t mtu = 0; 1247 int alwaysfrag = 0; 1248 int error = 0; 1249 1250 if (ro_pmtu != ro) { 1251 /* The first hop and the final destination may differ. */ 1252 struct sockaddr_in6 *sa6_dst = 1253 (struct sockaddr_in6 *)&ro_pmtu->ro_dst; 1254 if (ro_pmtu->ro_rt && 1255 ((ro_pmtu->ro_rt->rt_flags & RTF_UP) == 0 || 1256 !IN6_ARE_ADDR_EQUAL(&sa6_dst->sin6_addr, dst))) { 1257 RTFREE(ro_pmtu->ro_rt); 1258 ro_pmtu->ro_rt = (struct rtentry *)NULL; 1259 } 1260 if (ro_pmtu->ro_rt == NULL) { 1261 bzero(sa6_dst, sizeof(*sa6_dst)); /* for safety */ 1262 sa6_dst->sin6_family = AF_INET6; 1263 sa6_dst->sin6_len = sizeof(struct sockaddr_in6); 1264 sa6_dst->sin6_addr = *dst; 1265 1266 rtalloc((struct route *)ro_pmtu); 1267 } 1268 } 1269 if (ro_pmtu->ro_rt) { 1270 u_int32_t ifmtu; 1271 1272 if (ifp == NULL) 1273 ifp = ro_pmtu->ro_rt->rt_ifp; 1274 ifmtu = IN6_LINKMTU(ifp); 1275 mtu = ro_pmtu->ro_rt->rt_rmx.rmx_mtu; 1276 if (mtu == 0) 1277 mtu = ifmtu; 1278 else if (mtu < IPV6_MMTU) { 1279 /* 1280 * RFC2460 section 5, last paragraph: 1281 * if we record ICMPv6 too big message with 1282 * mtu < IPV6_MMTU, transmit packets sized IPV6_MMTU 1283 * or smaller, with fragment header attached. 1284 * (fragment header is needed regardless from the 1285 * packet size, for translators to identify packets) 1286 */ 1287 alwaysfrag = 1; 1288 mtu = IPV6_MMTU; 1289 } else if (mtu > ifmtu) { 1290 /* 1291 * The MTU on the route is larger than the MTU on 1292 * the interface! This shouldn't happen, unless the 1293 * MTU of the interface has been changed after the 1294 * interface was brought up. Change the MTU in the 1295 * route to match the interface MTU (as long as the 1296 * field isn't locked). 1297 */ 1298 mtu = ifmtu; 1299 if (!(ro_pmtu->ro_rt->rt_rmx.rmx_locks & RTV_MTU)) 1300 ro_pmtu->ro_rt->rt_rmx.rmx_mtu = mtu; 1301 } 1302 } else if (ifp) { 1303 mtu = IN6_LINKMTU(ifp); 1304 } else 1305 error = EHOSTUNREACH; /* XXX */ 1306 1307 *mtup = mtu; 1308 if (alwaysfragp) 1309 *alwaysfragp = alwaysfrag; 1310 return (error); 1311 } 1312 1313 /* 1314 * IP6 socket option processing. 1315 */ 1316 int 1317 ip6_ctloutput(op, so, level, optname, mp) 1318 int op; 1319 struct socket *so; 1320 int level, optname; 1321 struct mbuf **mp; 1322 { 1323 struct in6pcb *in6p = sotoin6pcb(so); 1324 struct mbuf *m = *mp; 1325 int optval = 0; 1326 int error = 0; 1327 struct proc *p = curproc; /* XXX */ 1328 1329 if (level == IPPROTO_IPV6) { 1330 switch (op) { 1331 case PRCO_SETOPT: 1332 switch (optname) { 1333 case IPV6_PKTOPTIONS: 1334 /* m is freed in ip6_pcbopts */ 1335 return (ip6_pcbopts(&in6p->in6p_outputopts, 1336 m, so)); 1337 case IPV6_HOPOPTS: 1338 case IPV6_DSTOPTS: 1339 if (p == 0 || suser(p->p_ucred, &p->p_acflag)) { 1340 error = EPERM; 1341 break; 1342 } 1343 /* FALLTHROUGH */ 1344 case IPV6_UNICAST_HOPS: 1345 case IPV6_RECVOPTS: 1346 case IPV6_RECVRETOPTS: 1347 case IPV6_RECVDSTADDR: 1348 case IPV6_PKTINFO: 1349 case IPV6_HOPLIMIT: 1350 case IPV6_RTHDR: 1351 case IPV6_FAITH: 1352 case IPV6_V6ONLY: 1353 if (!m || m->m_len != sizeof(int)) { 1354 error = EINVAL; 1355 break; 1356 } 1357 optval = *mtod(m, int *); 1358 switch (optname) { 1359 1360 case IPV6_UNICAST_HOPS: 1361 if (optval < -1 || optval >= 256) 1362 error = EINVAL; 1363 else { 1364 /* -1 = kernel default */ 1365 in6p->in6p_hops = optval; 1366 } 1367 break; 1368 #define OPTSET(bit) \ 1369 do { \ 1370 if (optval) \ 1371 in6p->in6p_flags |= (bit); \ 1372 else \ 1373 in6p->in6p_flags &= ~(bit); \ 1374 } while (/*CONSTCOND*/ 0) 1375 1376 case IPV6_RECVOPTS: 1377 OPTSET(IN6P_RECVOPTS); 1378 break; 1379 1380 case IPV6_RECVRETOPTS: 1381 OPTSET(IN6P_RECVRETOPTS); 1382 break; 1383 1384 case IPV6_RECVDSTADDR: 1385 OPTSET(IN6P_RECVDSTADDR); 1386 break; 1387 1388 case IPV6_PKTINFO: 1389 OPTSET(IN6P_PKTINFO); 1390 break; 1391 1392 case IPV6_HOPLIMIT: 1393 OPTSET(IN6P_HOPLIMIT); 1394 break; 1395 1396 case IPV6_HOPOPTS: 1397 OPTSET(IN6P_HOPOPTS); 1398 break; 1399 1400 case IPV6_DSTOPTS: 1401 OPTSET(IN6P_DSTOPTS); 1402 break; 1403 1404 case IPV6_RTHDR: 1405 OPTSET(IN6P_RTHDR); 1406 break; 1407 1408 case IPV6_FAITH: 1409 OPTSET(IN6P_FAITH); 1410 break; 1411 1412 case IPV6_V6ONLY: 1413 /* 1414 * make setsockopt(IPV6_V6ONLY) 1415 * available only prior to bind(2). 1416 * see ipng mailing list, Jun 22 2001. 1417 */ 1418 if (in6p->in6p_lport || 1419 !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr)) { 1420 error = EINVAL; 1421 break; 1422 } 1423 #ifdef INET6_BINDV6ONLY 1424 if (!optval) 1425 error = EINVAL; 1426 #else 1427 OPTSET(IN6P_IPV6_V6ONLY); 1428 #endif 1429 break; 1430 } 1431 break; 1432 #undef OPTSET 1433 1434 case IPV6_MULTICAST_IF: 1435 case IPV6_MULTICAST_HOPS: 1436 case IPV6_MULTICAST_LOOP: 1437 case IPV6_JOIN_GROUP: 1438 case IPV6_LEAVE_GROUP: 1439 error = ip6_setmoptions(optname, 1440 &in6p->in6p_moptions, m); 1441 break; 1442 1443 case IPV6_PORTRANGE: 1444 optval = *mtod(m, int *); 1445 1446 switch (optval) { 1447 case IPV6_PORTRANGE_DEFAULT: 1448 in6p->in6p_flags &= ~(IN6P_LOWPORT); 1449 in6p->in6p_flags &= ~(IN6P_HIGHPORT); 1450 break; 1451 1452 case IPV6_PORTRANGE_HIGH: 1453 in6p->in6p_flags &= ~(IN6P_LOWPORT); 1454 in6p->in6p_flags |= IN6P_HIGHPORT; 1455 break; 1456 1457 case IPV6_PORTRANGE_LOW: 1458 in6p->in6p_flags &= ~(IN6P_HIGHPORT); 1459 in6p->in6p_flags |= IN6P_LOWPORT; 1460 break; 1461 1462 default: 1463 error = EINVAL; 1464 break; 1465 } 1466 break; 1467 1468 #ifdef IPSEC 1469 case IPV6_IPSEC_POLICY: 1470 { 1471 caddr_t req = NULL; 1472 size_t len = 0; 1473 1474 int priv = 0; 1475 if (p == 0 || suser(p->p_ucred, &p->p_acflag)) 1476 priv = 0; 1477 else 1478 priv = 1; 1479 if (m) { 1480 req = mtod(m, caddr_t); 1481 len = m->m_len; 1482 } 1483 error = ipsec6_set_policy(in6p, 1484 optname, req, len, priv); 1485 } 1486 break; 1487 #endif /* IPSEC */ 1488 1489 default: 1490 error = ENOPROTOOPT; 1491 break; 1492 } 1493 if (m) 1494 (void)m_free(m); 1495 break; 1496 1497 case PRCO_GETOPT: 1498 switch (optname) { 1499 1500 case IPV6_OPTIONS: 1501 case IPV6_RETOPTS: 1502 error = ENOPROTOOPT; 1503 break; 1504 1505 case IPV6_PKTOPTIONS: 1506 if (in6p->in6p_options) { 1507 *mp = m_copym(in6p->in6p_options, 0, 1508 M_COPYALL, M_WAIT); 1509 } else { 1510 *mp = m_get(M_WAIT, MT_SOOPTS); 1511 (*mp)->m_len = 0; 1512 } 1513 break; 1514 1515 case IPV6_HOPOPTS: 1516 case IPV6_DSTOPTS: 1517 if (p == 0 || suser(p->p_ucred, &p->p_acflag)) { 1518 error = EPERM; 1519 break; 1520 } 1521 /* FALLTHROUGH */ 1522 case IPV6_UNICAST_HOPS: 1523 case IPV6_RECVOPTS: 1524 case IPV6_RECVRETOPTS: 1525 case IPV6_RECVDSTADDR: 1526 case IPV6_PORTRANGE: 1527 case IPV6_PKTINFO: 1528 case IPV6_HOPLIMIT: 1529 case IPV6_RTHDR: 1530 case IPV6_FAITH: 1531 case IPV6_V6ONLY: 1532 *mp = m = m_get(M_WAIT, MT_SOOPTS); 1533 m->m_len = sizeof(int); 1534 switch (optname) { 1535 1536 case IPV6_UNICAST_HOPS: 1537 optval = in6p->in6p_hops; 1538 break; 1539 1540 #define OPTBIT(bit) (in6p->in6p_flags & bit ? 1 : 0) 1541 1542 case IPV6_RECVOPTS: 1543 optval = OPTBIT(IN6P_RECVOPTS); 1544 break; 1545 1546 case IPV6_RECVRETOPTS: 1547 optval = OPTBIT(IN6P_RECVRETOPTS); 1548 break; 1549 1550 case IPV6_RECVDSTADDR: 1551 optval = OPTBIT(IN6P_RECVDSTADDR); 1552 break; 1553 1554 case IPV6_PORTRANGE: 1555 { 1556 int flags; 1557 flags = in6p->in6p_flags; 1558 if (flags & IN6P_HIGHPORT) 1559 optval = IPV6_PORTRANGE_HIGH; 1560 else if (flags & IN6P_LOWPORT) 1561 optval = IPV6_PORTRANGE_LOW; 1562 else 1563 optval = 0; 1564 break; 1565 } 1566 1567 case IPV6_PKTINFO: 1568 optval = OPTBIT(IN6P_PKTINFO); 1569 break; 1570 1571 case IPV6_HOPLIMIT: 1572 optval = OPTBIT(IN6P_HOPLIMIT); 1573 break; 1574 1575 case IPV6_HOPOPTS: 1576 optval = OPTBIT(IN6P_HOPOPTS); 1577 break; 1578 1579 case IPV6_DSTOPTS: 1580 optval = OPTBIT(IN6P_DSTOPTS); 1581 break; 1582 1583 case IPV6_RTHDR: 1584 optval = OPTBIT(IN6P_RTHDR); 1585 break; 1586 1587 case IPV6_FAITH: 1588 optval = OPTBIT(IN6P_FAITH); 1589 break; 1590 1591 case IPV6_V6ONLY: 1592 optval = OPTBIT(IN6P_IPV6_V6ONLY); 1593 break; 1594 } 1595 *mtod(m, int *) = optval; 1596 break; 1597 1598 case IPV6_MULTICAST_IF: 1599 case IPV6_MULTICAST_HOPS: 1600 case IPV6_MULTICAST_LOOP: 1601 case IPV6_JOIN_GROUP: 1602 case IPV6_LEAVE_GROUP: 1603 error = ip6_getmoptions(optname, in6p->in6p_moptions, mp); 1604 break; 1605 1606 #ifdef IPSEC 1607 case IPV6_IPSEC_POLICY: 1608 { 1609 caddr_t req = NULL; 1610 size_t len = 0; 1611 1612 if (m) { 1613 req = mtod(m, caddr_t); 1614 len = m->m_len; 1615 } 1616 error = ipsec6_get_policy(in6p, req, len, mp); 1617 break; 1618 } 1619 #endif /* IPSEC */ 1620 1621 default: 1622 error = ENOPROTOOPT; 1623 break; 1624 } 1625 break; 1626 } 1627 } else { 1628 error = EINVAL; 1629 if (op == PRCO_SETOPT && *mp) 1630 (void)m_free(*mp); 1631 } 1632 return (error); 1633 } 1634 1635 int 1636 ip6_raw_ctloutput(op, so, level, optname, mp) 1637 int op; 1638 struct socket *so; 1639 int level, optname; 1640 struct mbuf **mp; 1641 { 1642 int error = 0, optval, optlen; 1643 const int icmp6off = offsetof(struct icmp6_hdr, icmp6_cksum); 1644 struct in6pcb *in6p = sotoin6pcb(so); 1645 struct mbuf *m = *mp; 1646 1647 optlen = m ? m->m_len : 0; 1648 1649 if (level != IPPROTO_IPV6) { 1650 if (op == PRCO_SETOPT && *mp) 1651 (void)m_free(*mp); 1652 return (EINVAL); 1653 } 1654 1655 switch (optname) { 1656 case IPV6_CHECKSUM: 1657 /* 1658 * For ICMPv6 sockets, no modification allowed for checksum 1659 * offset, permit "no change" values to help existing apps. 1660 * 1661 * XXX 2292bis says: "An attempt to set IPV6_CHECKSUM 1662 * for an ICMPv6 socket will fail." 1663 * The current behavior does not meet 2292bis. 1664 */ 1665 switch (op) { 1666 case PRCO_SETOPT: 1667 if (optlen != sizeof(int)) { 1668 error = EINVAL; 1669 break; 1670 } 1671 optval = *mtod(m, int *); 1672 if ((optval % 2) != 0) { 1673 /* the API assumes even offset values */ 1674 error = EINVAL; 1675 } else if (so->so_proto->pr_protocol == 1676 IPPROTO_ICMPV6) { 1677 if (optval != icmp6off) 1678 error = EINVAL; 1679 } else 1680 in6p->in6p_cksum = optval; 1681 break; 1682 1683 case PRCO_GETOPT: 1684 if (so->so_proto->pr_protocol == IPPROTO_ICMPV6) 1685 optval = icmp6off; 1686 else 1687 optval = in6p->in6p_cksum; 1688 1689 *mp = m = m_get(M_WAIT, MT_SOOPTS); 1690 m->m_len = sizeof(int); 1691 *mtod(m, int *) = optval; 1692 break; 1693 1694 default: 1695 error = EINVAL; 1696 break; 1697 } 1698 break; 1699 1700 default: 1701 error = ENOPROTOOPT; 1702 break; 1703 } 1704 1705 if (op == PRCO_SETOPT && m) 1706 (void)m_free(m); 1707 1708 return (error); 1709 } 1710 1711 /* 1712 * Set up IP6 options in pcb for insertion in output packets. 1713 * Store in mbuf with pointer in pcbopt, adding pseudo-option 1714 * with destination address if source routed. 1715 */ 1716 static int 1717 ip6_pcbopts(pktopt, m, so) 1718 struct ip6_pktopts **pktopt; 1719 struct mbuf *m; 1720 struct socket *so; 1721 { 1722 struct ip6_pktopts *opt = *pktopt; 1723 int error = 0; 1724 struct proc *p = curproc; /* XXX */ 1725 int priv = 0; 1726 1727 /* turn off any old options. */ 1728 if (opt) { 1729 if (opt->ip6po_m) 1730 (void)m_free(opt->ip6po_m); 1731 } else 1732 opt = malloc(sizeof(*opt), M_IP6OPT, M_WAITOK); 1733 *pktopt = 0; 1734 1735 if (!m || m->m_len == 0) { 1736 /* 1737 * Only turning off any previous options. 1738 */ 1739 free(opt, M_IP6OPT); 1740 if (m) 1741 (void)m_free(m); 1742 return (0); 1743 } 1744 1745 /* set options specified by user. */ 1746 if (p && !suser(p->p_ucred, &p->p_acflag)) 1747 priv = 1; 1748 if ((error = ip6_setpktoptions(m, opt, priv)) != 0) { 1749 (void)m_free(m); 1750 free(opt, M_IP6OPT); 1751 return (error); 1752 } 1753 *pktopt = opt; 1754 return (0); 1755 } 1756 1757 /* 1758 * Set the IP6 multicast options in response to user setsockopt(). 1759 */ 1760 static int 1761 ip6_setmoptions(optname, im6op, m) 1762 int optname; 1763 struct ip6_moptions **im6op; 1764 struct mbuf *m; 1765 { 1766 int error = 0; 1767 u_int loop, ifindex; 1768 struct ipv6_mreq *mreq; 1769 struct ifnet *ifp; 1770 struct ip6_moptions *im6o = *im6op; 1771 struct route_in6 ro; 1772 struct sockaddr_in6 *dst; 1773 struct in6_multi_mship *imm; 1774 struct proc *p = curproc; /* XXX */ 1775 1776 if (im6o == NULL) { 1777 /* 1778 * No multicast option buffer attached to the pcb; 1779 * allocate one and initialize to default values. 1780 */ 1781 im6o = (struct ip6_moptions *) 1782 malloc(sizeof(*im6o), M_IPMOPTS, M_WAITOK); 1783 1784 if (im6o == NULL) 1785 return (ENOBUFS); 1786 *im6op = im6o; 1787 im6o->im6o_multicast_ifp = NULL; 1788 im6o->im6o_multicast_hlim = ip6_defmcasthlim; 1789 im6o->im6o_multicast_loop = IPV6_DEFAULT_MULTICAST_LOOP; 1790 LIST_INIT(&im6o->im6o_memberships); 1791 } 1792 1793 switch (optname) { 1794 1795 case IPV6_MULTICAST_IF: 1796 /* 1797 * Select the interface for outgoing multicast packets. 1798 */ 1799 if (m == NULL || m->m_len != sizeof(u_int)) { 1800 error = EINVAL; 1801 break; 1802 } 1803 bcopy(mtod(m, u_int *), &ifindex, sizeof(ifindex)); 1804 if (ifindex < 0 || if_indexlim <= ifindex || 1805 !ifindex2ifnet[ifindex]) { 1806 error = ENXIO; /* XXX EINVAL? */ 1807 break; 1808 } 1809 ifp = ifindex2ifnet[ifindex]; 1810 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) { 1811 error = EADDRNOTAVAIL; 1812 break; 1813 } 1814 im6o->im6o_multicast_ifp = ifp; 1815 break; 1816 1817 case IPV6_MULTICAST_HOPS: 1818 { 1819 /* 1820 * Set the IP6 hoplimit for outgoing multicast packets. 1821 */ 1822 int optval; 1823 if (m == NULL || m->m_len != sizeof(int)) { 1824 error = EINVAL; 1825 break; 1826 } 1827 bcopy(mtod(m, u_int *), &optval, sizeof(optval)); 1828 if (optval < -1 || optval >= 256) 1829 error = EINVAL; 1830 else if (optval == -1) 1831 im6o->im6o_multicast_hlim = ip6_defmcasthlim; 1832 else 1833 im6o->im6o_multicast_hlim = optval; 1834 break; 1835 } 1836 1837 case IPV6_MULTICAST_LOOP: 1838 /* 1839 * Set the loopback flag for outgoing multicast packets. 1840 * Must be zero or one. 1841 */ 1842 if (m == NULL || m->m_len != sizeof(u_int)) { 1843 error = EINVAL; 1844 break; 1845 } 1846 bcopy(mtod(m, u_int *), &loop, sizeof(loop)); 1847 if (loop > 1) { 1848 error = EINVAL; 1849 break; 1850 } 1851 im6o->im6o_multicast_loop = loop; 1852 break; 1853 1854 case IPV6_JOIN_GROUP: 1855 /* 1856 * Add a multicast group membership. 1857 * Group must be a valid IP6 multicast address. 1858 */ 1859 if (m == NULL || m->m_len != sizeof(struct ipv6_mreq)) { 1860 error = EINVAL; 1861 break; 1862 } 1863 mreq = mtod(m, struct ipv6_mreq *); 1864 if (IN6_IS_ADDR_UNSPECIFIED(&mreq->ipv6mr_multiaddr)) { 1865 /* 1866 * We use the unspecified address to specify to accept 1867 * all multicast addresses. Only super user is allowed 1868 * to do this. 1869 */ 1870 if (suser(p->p_ucred, &p->p_acflag)) 1871 { 1872 error = EACCES; 1873 break; 1874 } 1875 } else if (!IN6_IS_ADDR_MULTICAST(&mreq->ipv6mr_multiaddr)) { 1876 error = EINVAL; 1877 break; 1878 } 1879 1880 /* 1881 * If the interface is specified, validate it. 1882 */ 1883 if (mreq->ipv6mr_interface < 0 || 1884 if_indexlim <= mreq->ipv6mr_interface || 1885 !ifindex2ifnet[mreq->ipv6mr_interface]) { 1886 error = ENXIO; /* XXX EINVAL? */ 1887 break; 1888 } 1889 /* 1890 * If no interface was explicitly specified, choose an 1891 * appropriate one according to the given multicast address. 1892 */ 1893 if (mreq->ipv6mr_interface == 0) { 1894 /* 1895 * If the multicast address is in node-local scope, 1896 * the interface should be a loopback interface. 1897 * Otherwise, look up the routing table for the 1898 * address, and choose the outgoing interface. 1899 * XXX: is it a good approach? 1900 */ 1901 if (IN6_IS_ADDR_MC_NODELOCAL(&mreq->ipv6mr_multiaddr)) { 1902 ifp = &loif[0]; 1903 } else { 1904 ro.ro_rt = NULL; 1905 dst = (struct sockaddr_in6 *)&ro.ro_dst; 1906 bzero(dst, sizeof(*dst)); 1907 dst->sin6_len = sizeof(struct sockaddr_in6); 1908 dst->sin6_family = AF_INET6; 1909 dst->sin6_addr = mreq->ipv6mr_multiaddr; 1910 rtalloc((struct route *)&ro); 1911 if (ro.ro_rt == NULL) { 1912 error = EADDRNOTAVAIL; 1913 break; 1914 } 1915 ifp = ro.ro_rt->rt_ifp; 1916 rtfree(ro.ro_rt); 1917 } 1918 } else 1919 ifp = ifindex2ifnet[mreq->ipv6mr_interface]; 1920 1921 /* 1922 * See if we found an interface, and confirm that it 1923 * supports multicast 1924 */ 1925 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) { 1926 error = EADDRNOTAVAIL; 1927 break; 1928 } 1929 /* 1930 * Put interface index into the multicast address, 1931 * if the address has link-local scope. 1932 */ 1933 if (IN6_IS_ADDR_MC_LINKLOCAL(&mreq->ipv6mr_multiaddr)) { 1934 mreq->ipv6mr_multiaddr.s6_addr16[1] = 1935 htons(ifp->if_index); 1936 } 1937 /* 1938 * See if the membership already exists. 1939 */ 1940 for (imm = im6o->im6o_memberships.lh_first; 1941 imm != NULL; imm = imm->i6mm_chain.le_next) 1942 if (imm->i6mm_maddr->in6m_ifp == ifp && 1943 IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr, 1944 &mreq->ipv6mr_multiaddr)) 1945 break; 1946 if (imm != NULL) { 1947 error = EADDRINUSE; 1948 break; 1949 } 1950 /* 1951 * Everything looks good; add a new record to the multicast 1952 * address list for the given interface. 1953 */ 1954 imm = in6_joingroup(ifp, &mreq->ipv6mr_multiaddr, &error); 1955 if (!imm) 1956 break; 1957 LIST_INSERT_HEAD(&im6o->im6o_memberships, imm, i6mm_chain); 1958 break; 1959 1960 case IPV6_LEAVE_GROUP: 1961 /* 1962 * Drop a multicast group membership. 1963 * Group must be a valid IP6 multicast address. 1964 */ 1965 if (m == NULL || m->m_len != sizeof(struct ipv6_mreq)) { 1966 error = EINVAL; 1967 break; 1968 } 1969 mreq = mtod(m, struct ipv6_mreq *); 1970 if (IN6_IS_ADDR_UNSPECIFIED(&mreq->ipv6mr_multiaddr)) { 1971 if (suser(p->p_ucred, &p->p_acflag)) 1972 { 1973 error = EACCES; 1974 break; 1975 } 1976 } else if (!IN6_IS_ADDR_MULTICAST(&mreq->ipv6mr_multiaddr)) { 1977 error = EINVAL; 1978 break; 1979 } 1980 /* 1981 * If an interface address was specified, get a pointer 1982 * to its ifnet structure. 1983 */ 1984 if (mreq->ipv6mr_interface < 0 || 1985 if_indexlim <= mreq->ipv6mr_interface || 1986 !ifindex2ifnet[mreq->ipv6mr_interface]) { 1987 error = ENXIO; /* XXX EINVAL? */ 1988 break; 1989 } 1990 ifp = ifindex2ifnet[mreq->ipv6mr_interface]; 1991 /* 1992 * Put interface index into the multicast address, 1993 * if the address has link-local scope. 1994 */ 1995 if (IN6_IS_ADDR_MC_LINKLOCAL(&mreq->ipv6mr_multiaddr)) { 1996 mreq->ipv6mr_multiaddr.s6_addr16[1] = 1997 htons(mreq->ipv6mr_interface); 1998 } 1999 /* 2000 * Find the membership in the membership list. 2001 */ 2002 for (imm = im6o->im6o_memberships.lh_first; 2003 imm != NULL; imm = imm->i6mm_chain.le_next) { 2004 if ((ifp == NULL || imm->i6mm_maddr->in6m_ifp == ifp) && 2005 IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr, 2006 &mreq->ipv6mr_multiaddr)) 2007 break; 2008 } 2009 if (imm == NULL) { 2010 /* Unable to resolve interface */ 2011 error = EADDRNOTAVAIL; 2012 break; 2013 } 2014 /* 2015 * Give up the multicast address record to which the 2016 * membership points. 2017 */ 2018 LIST_REMOVE(imm, i6mm_chain); 2019 in6_leavegroup(imm); 2020 break; 2021 2022 default: 2023 error = EOPNOTSUPP; 2024 break; 2025 } 2026 2027 /* 2028 * If all options have default values, no need to keep the mbuf. 2029 */ 2030 if (im6o->im6o_multicast_ifp == NULL && 2031 im6o->im6o_multicast_hlim == ip6_defmcasthlim && 2032 im6o->im6o_multicast_loop == IPV6_DEFAULT_MULTICAST_LOOP && 2033 im6o->im6o_memberships.lh_first == NULL) { 2034 free(*im6op, M_IPMOPTS); 2035 *im6op = NULL; 2036 } 2037 2038 return (error); 2039 } 2040 2041 /* 2042 * Return the IP6 multicast options in response to user getsockopt(). 2043 */ 2044 static int 2045 ip6_getmoptions(optname, im6o, mp) 2046 int optname; 2047 struct ip6_moptions *im6o; 2048 struct mbuf **mp; 2049 { 2050 u_int *hlim, *loop, *ifindex; 2051 2052 *mp = m_get(M_WAIT, MT_SOOPTS); 2053 2054 switch (optname) { 2055 2056 case IPV6_MULTICAST_IF: 2057 ifindex = mtod(*mp, u_int *); 2058 (*mp)->m_len = sizeof(u_int); 2059 if (im6o == NULL || im6o->im6o_multicast_ifp == NULL) 2060 *ifindex = 0; 2061 else 2062 *ifindex = im6o->im6o_multicast_ifp->if_index; 2063 return (0); 2064 2065 case IPV6_MULTICAST_HOPS: 2066 hlim = mtod(*mp, u_int *); 2067 (*mp)->m_len = sizeof(u_int); 2068 if (im6o == NULL) 2069 *hlim = ip6_defmcasthlim; 2070 else 2071 *hlim = im6o->im6o_multicast_hlim; 2072 return (0); 2073 2074 case IPV6_MULTICAST_LOOP: 2075 loop = mtod(*mp, u_int *); 2076 (*mp)->m_len = sizeof(u_int); 2077 if (im6o == NULL) 2078 *loop = ip6_defmcasthlim; 2079 else 2080 *loop = im6o->im6o_multicast_loop; 2081 return (0); 2082 2083 default: 2084 return (EOPNOTSUPP); 2085 } 2086 } 2087 2088 /* 2089 * Discard the IP6 multicast options. 2090 */ 2091 void 2092 ip6_freemoptions(im6o) 2093 struct ip6_moptions *im6o; 2094 { 2095 struct in6_multi_mship *imm; 2096 2097 if (im6o == NULL) 2098 return; 2099 2100 while ((imm = im6o->im6o_memberships.lh_first) != NULL) { 2101 LIST_REMOVE(imm, i6mm_chain); 2102 in6_leavegroup(imm); 2103 } 2104 free(im6o, M_IPMOPTS); 2105 } 2106 2107 /* 2108 * Set IPv6 outgoing packet options based on advanced API. 2109 */ 2110 int 2111 ip6_setpktoptions(control, opt, priv) 2112 struct mbuf *control; 2113 struct ip6_pktopts *opt; 2114 int priv; 2115 { 2116 struct cmsghdr *cm = 0; 2117 2118 if (control == 0 || opt == 0) 2119 return (EINVAL); 2120 2121 bzero(opt, sizeof(*opt)); 2122 opt->ip6po_hlim = -1; /* -1 means to use default hop limit */ 2123 2124 /* 2125 * XXX: Currently, we assume all the optional information is stored 2126 * in a single mbuf. 2127 */ 2128 if (control->m_next) 2129 return (EINVAL); 2130 2131 opt->ip6po_m = control; 2132 2133 for (; control->m_len; control->m_data += CMSG_ALIGN(cm->cmsg_len), 2134 control->m_len -= CMSG_ALIGN(cm->cmsg_len)) { 2135 cm = mtod(control, struct cmsghdr *); 2136 if (cm->cmsg_len == 0 || cm->cmsg_len > control->m_len) 2137 return (EINVAL); 2138 if (cm->cmsg_level != IPPROTO_IPV6) 2139 continue; 2140 2141 switch (cm->cmsg_type) { 2142 case IPV6_PKTINFO: 2143 if (cm->cmsg_len != CMSG_LEN(sizeof(struct in6_pktinfo))) 2144 return (EINVAL); 2145 opt->ip6po_pktinfo = (struct in6_pktinfo *)CMSG_DATA(cm); 2146 if (opt->ip6po_pktinfo->ipi6_ifindex && 2147 IN6_IS_ADDR_LINKLOCAL(&opt->ip6po_pktinfo->ipi6_addr)) 2148 opt->ip6po_pktinfo->ipi6_addr.s6_addr16[1] = 2149 htons(opt->ip6po_pktinfo->ipi6_ifindex); 2150 2151 if (opt->ip6po_pktinfo->ipi6_ifindex >= if_indexlim || 2152 opt->ip6po_pktinfo->ipi6_ifindex < 0) 2153 return (ENXIO); 2154 if (opt->ip6po_pktinfo->ipi6_ifindex > 0 && 2155 !ifindex2ifnet[opt->ip6po_pktinfo->ipi6_ifindex]) 2156 return (ENXIO); 2157 2158 /* 2159 * Check if the requested source address is indeed a 2160 * unicast address assigned to the node, and can be 2161 * used as the packet's source address. 2162 */ 2163 if (!IN6_IS_ADDR_UNSPECIFIED(&opt->ip6po_pktinfo->ipi6_addr)) { 2164 struct ifaddr *ia; 2165 struct in6_ifaddr *ia6; 2166 struct sockaddr_in6 sin6; 2167 2168 bzero(&sin6, sizeof(sin6)); 2169 sin6.sin6_len = sizeof(sin6); 2170 sin6.sin6_family = AF_INET6; 2171 sin6.sin6_addr = 2172 opt->ip6po_pktinfo->ipi6_addr; 2173 ia = ifa_ifwithaddr(sin6tosa(&sin6)); 2174 if (ia == NULL || 2175 (opt->ip6po_pktinfo->ipi6_ifindex && 2176 (ia->ifa_ifp->if_index != 2177 opt->ip6po_pktinfo->ipi6_ifindex))) { 2178 return (EADDRNOTAVAIL); 2179 } 2180 ia6 = (struct in6_ifaddr *)ia; 2181 if ((ia6->ia6_flags & (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY)) != 0) { 2182 return (EADDRNOTAVAIL); 2183 } 2184 2185 /* 2186 * Check if the requested source address is 2187 * indeed a unicast address assigned to the 2188 * node. 2189 */ 2190 if (IN6_IS_ADDR_MULTICAST(&opt->ip6po_pktinfo->ipi6_addr)) 2191 return (EADDRNOTAVAIL); 2192 } 2193 break; 2194 2195 case IPV6_HOPLIMIT: 2196 if (cm->cmsg_len != CMSG_LEN(sizeof(int))) 2197 return (EINVAL); 2198 else { 2199 int t; 2200 2201 bcopy(CMSG_DATA(cm), &t, sizeof(t)); 2202 if (t < -1 || t > 255) 2203 return (EINVAL); 2204 opt->ip6po_hlim = t; 2205 } 2206 break; 2207 2208 case IPV6_NEXTHOP: 2209 if (!priv) 2210 return (EPERM); 2211 2212 /* check if cmsg_len is large enough for sa_len */ 2213 if (cm->cmsg_len < sizeof(u_char) || 2214 cm->cmsg_len < CMSG_LEN(*CMSG_DATA(cm))) 2215 return (EINVAL); 2216 2217 opt->ip6po_nexthop = (struct sockaddr *)CMSG_DATA(cm); 2218 2219 break; 2220 2221 case IPV6_HOPOPTS: 2222 if (cm->cmsg_len < CMSG_LEN(sizeof(struct ip6_hbh))) 2223 return (EINVAL); 2224 else { 2225 struct ip6_hbh *t; 2226 2227 t = (struct ip6_hbh *)CMSG_DATA(cm); 2228 if (cm->cmsg_len != 2229 CMSG_LEN((t->ip6h_len + 1) << 3)) 2230 return (EINVAL); 2231 opt->ip6po_hbh = t; 2232 } 2233 break; 2234 2235 case IPV6_DSTOPTS: 2236 if (cm->cmsg_len < CMSG_LEN(sizeof(struct ip6_dest))) 2237 return (EINVAL); 2238 2239 /* 2240 * If there is no routing header yet, the destination 2241 * options header should be put on the 1st part. 2242 * Otherwise, the header should be on the 2nd part. 2243 * (See RFC 2460, section 4.1) 2244 */ 2245 if (opt->ip6po_rthdr == NULL) { 2246 struct ip6_dest *t; 2247 2248 t = (struct ip6_dest *)CMSG_DATA(cm); 2249 if (cm->cmsg_len != 2250 CMSG_LEN((t->ip6d_len + 1) << 3)); 2251 return (EINVAL); 2252 opt->ip6po_dest1 = t; 2253 } 2254 else { 2255 struct ip6_dest *t; 2256 2257 t = (struct ip6_dest *)CMSG_DATA(cm); 2258 if (cm->cmsg_len != 2259 CMSG_LEN((opt->ip6po_dest2->ip6d_len + 1) << 3)) 2260 return (EINVAL); 2261 opt->ip6po_dest2 = t; 2262 } 2263 break; 2264 2265 case IPV6_RTHDR: 2266 if (cm->cmsg_len < CMSG_LEN(sizeof(struct ip6_rthdr))) 2267 return (EINVAL); 2268 else { 2269 struct ip6_rthdr *t; 2270 2271 t = (struct ip6_rthdr *)CMSG_DATA(cm); 2272 if (cm->cmsg_len != 2273 CMSG_LEN((t->ip6r_len + 1) << 3)) 2274 return (EINVAL); 2275 switch (t->ip6r_type) { 2276 case IPV6_RTHDR_TYPE_0: 2277 if (t->ip6r_segleft == 0) 2278 return (EINVAL); 2279 break; 2280 default: 2281 return (EINVAL); 2282 } 2283 opt->ip6po_rthdr = t; 2284 } 2285 break; 2286 2287 default: 2288 return (ENOPROTOOPT); 2289 } 2290 } 2291 2292 return (0); 2293 } 2294 2295 /* 2296 * Routine called from ip6_output() to loop back a copy of an IP6 multicast 2297 * packet to the input queue of a specified interface. Note that this 2298 * calls the output routine of the loopback "driver", but with an interface 2299 * pointer that might NOT be &loif -- easier than replicating that code here. 2300 */ 2301 void 2302 ip6_mloopback(ifp, m, dst) 2303 struct ifnet *ifp; 2304 struct mbuf *m; 2305 struct sockaddr_in6 *dst; 2306 { 2307 struct mbuf *copym; 2308 struct ip6_hdr *ip6; 2309 2310 copym = m_copy(m, 0, M_COPYALL); 2311 if (copym == NULL) 2312 return; 2313 2314 /* 2315 * Make sure to deep-copy IPv6 header portion in case the data 2316 * is in an mbuf cluster, so that we can safely override the IPv6 2317 * header portion later. 2318 */ 2319 if ((copym->m_flags & M_EXT) != 0 || 2320 copym->m_len < sizeof(struct ip6_hdr)) { 2321 copym = m_pullup(copym, sizeof(struct ip6_hdr)); 2322 if (copym == NULL) 2323 return; 2324 } 2325 2326 #ifdef DIAGNOSTIC 2327 if (copym->m_len < sizeof(*ip6)) { 2328 m_freem(copym); 2329 return; 2330 } 2331 #endif 2332 2333 ip6 = mtod(copym, struct ip6_hdr *); 2334 if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src)) 2335 ip6->ip6_src.s6_addr16[1] = 0; 2336 if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst)) 2337 ip6->ip6_dst.s6_addr16[1] = 0; 2338 2339 (void)looutput(ifp, copym, (struct sockaddr *)dst, NULL); 2340 } 2341 2342 /* 2343 * Chop IPv6 header off from the payload. 2344 */ 2345 static int 2346 ip6_splithdr(m, exthdrs) 2347 struct mbuf *m; 2348 struct ip6_exthdrs *exthdrs; 2349 { 2350 struct mbuf *mh; 2351 struct ip6_hdr *ip6; 2352 2353 ip6 = mtod(m, struct ip6_hdr *); 2354 if (m->m_len > sizeof(*ip6)) { 2355 MGETHDR(mh, M_DONTWAIT, MT_HEADER); 2356 if (mh == 0) { 2357 m_freem(m); 2358 return ENOBUFS; 2359 } 2360 M_COPY_PKTHDR(mh, m); 2361 MH_ALIGN(mh, sizeof(*ip6)); 2362 m_tag_delete_chain(m, NULL); 2363 m->m_flags &= ~M_PKTHDR; 2364 m->m_len -= sizeof(*ip6); 2365 m->m_data += sizeof(*ip6); 2366 mh->m_next = m; 2367 m = mh; 2368 m->m_len = sizeof(*ip6); 2369 bcopy((caddr_t)ip6, mtod(m, caddr_t), sizeof(*ip6)); 2370 } 2371 exthdrs->ip6e_ip6 = m; 2372 return 0; 2373 } 2374 2375 /* 2376 * Compute IPv6 extension header length. 2377 */ 2378 int 2379 ip6_optlen(in6p) 2380 struct in6pcb *in6p; 2381 { 2382 int len; 2383 2384 if (!in6p->in6p_outputopts) 2385 return 0; 2386 2387 len = 0; 2388 #define elen(x) \ 2389 (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0) 2390 2391 len += elen(in6p->in6p_outputopts->ip6po_hbh); 2392 len += elen(in6p->in6p_outputopts->ip6po_dest1); 2393 len += elen(in6p->in6p_outputopts->ip6po_rthdr); 2394 len += elen(in6p->in6p_outputopts->ip6po_dest2); 2395 return len; 2396 #undef elen 2397 } 2398
Indexes created Wed Sep 24 14:09:57 GMT 2025