ip6_output.c revision 1.89
1 /* $NetBSD: ip6_output.c,v 1.89 2005/08/10 12:58:37 yamt 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 1.89 2005/08/10 12:58:37 yamt Exp $"); 66 67 #include "opt_inet.h" 68 #include "opt_ipsec.h" 69 #include "opt_pfil_hooks.h" 70 71 #include <sys/param.h> 72 #include <sys/malloc.h> 73 #include <sys/mbuf.h> 74 #include <sys/errno.h> 75 #include <sys/protosw.h> 76 #include <sys/socket.h> 77 #include <sys/socketvar.h> 78 #include <sys/systm.h> 79 #include <sys/proc.h> 80 81 #include <net/if.h> 82 #include <net/route.h> 83 #ifdef PFIL_HOOKS 84 #include <net/pfil.h> 85 #endif 86 87 #include <netinet/in.h> 88 #include <netinet/in_var.h> 89 #include <netinet/ip6.h> 90 #include <netinet/icmp6.h> 91 #include <netinet6/ip6_var.h> 92 #include <netinet6/in6_pcb.h> 93 #include <netinet6/nd6.h> 94 #include <netinet6/ip6protosw.h> 95 96 #ifdef IPSEC 97 #include <netinet6/ipsec.h> 98 #include <netkey/key.h> 99 #endif /* IPSEC */ 100 101 #include <net/net_osdep.h> 102 103 #ifdef PFIL_HOOKS 104 extern struct pfil_head inet6_pfil_hook; /* XXX */ 105 #endif 106 107 struct ip6_exthdrs { 108 struct mbuf *ip6e_ip6; 109 struct mbuf *ip6e_hbh; 110 struct mbuf *ip6e_dest1; 111 struct mbuf *ip6e_rthdr; 112 struct mbuf *ip6e_dest2; 113 }; 114 115 static int ip6_pcbopts __P((struct ip6_pktopts **, struct mbuf *, 116 struct socket *)); 117 static int ip6_setmoptions __P((int, struct ip6_moptions **, struct mbuf *)); 118 static int ip6_getmoptions __P((int, struct ip6_moptions *, struct mbuf **)); 119 static int ip6_copyexthdr __P((struct mbuf **, caddr_t, int)); 120 static int ip6_insertfraghdr __P((struct mbuf *, struct mbuf *, int, 121 struct ip6_frag **)); 122 static int ip6_insert_jumboopt __P((struct ip6_exthdrs *, u_int32_t)); 123 static int ip6_splithdr __P((struct mbuf *, struct ip6_exthdrs *)); 124 static int ip6_getpmtu __P((struct route_in6 *, struct route_in6 *, 125 struct ifnet *, struct in6_addr *, u_long *, int *)); 126 127 /* 128 * IP6 output. The packet in mbuf chain m contains a skeletal IP6 129 * header (with pri, len, nxt, hlim, src, dst). 130 * This function may modify ver and hlim only. 131 * The mbuf chain containing the packet will be freed. 132 * The mbuf opt, if present, will not be freed. 133 * 134 * type of "mtu": rt_rmx.rmx_mtu is u_long, ifnet.ifr_mtu is int, and 135 * nd_ifinfo.linkmtu is u_int32_t. so we use u_long to hold largest one, 136 * which is rt_rmx.rmx_mtu. 137 */ 138 int 139 ip6_output(m0, opt, ro, flags, im6o, so, ifpp) 140 struct mbuf *m0; 141 struct ip6_pktopts *opt; 142 struct route_in6 *ro; 143 int flags; 144 struct ip6_moptions *im6o; 145 struct socket *so; 146 struct ifnet **ifpp; /* XXX: just for statistics */ 147 { 148 struct ip6_hdr *ip6, *mhip6; 149 struct ifnet *ifp, *origifp; 150 struct mbuf *m = m0; 151 int hlen, tlen, len, off; 152 struct route_in6 ip6route; 153 struct sockaddr_in6 *dst; 154 int error = 0; 155 u_long mtu; 156 int alwaysfrag, dontfrag; 157 u_int32_t optlen = 0, plen = 0, unfragpartlen = 0; 158 struct ip6_exthdrs exthdrs; 159 struct in6_addr finaldst; 160 struct route_in6 *ro_pmtu = NULL; 161 int hdrsplit = 0; 162 int needipsec = 0; 163 #ifdef IPSEC 164 int needipsectun = 0; 165 struct secpolicy *sp = NULL; 166 167 ip6 = mtod(m, struct ip6_hdr *); 168 #endif /* IPSEC */ 169 170 M_CSUM_DATA_IPv6_HL_SET(m->m_pkthdr.csum_data, sizeof(struct ip6_hdr)); 171 172 #define MAKE_EXTHDR(hp, mp) \ 173 do { \ 174 if (hp) { \ 175 struct ip6_ext *eh = (struct ip6_ext *)(hp); \ 176 error = ip6_copyexthdr((mp), (caddr_t)(hp), \ 177 ((eh)->ip6e_len + 1) << 3); \ 178 if (error) \ 179 goto freehdrs; \ 180 } \ 181 } while (/*CONSTCOND*/ 0) 182 183 bzero(&exthdrs, sizeof(exthdrs)); 184 if (opt) { 185 /* Hop-by-Hop options header */ 186 MAKE_EXTHDR(opt->ip6po_hbh, &exthdrs.ip6e_hbh); 187 /* Destination options header(1st part) */ 188 MAKE_EXTHDR(opt->ip6po_dest1, &exthdrs.ip6e_dest1); 189 /* Routing header */ 190 MAKE_EXTHDR(opt->ip6po_rthdr, &exthdrs.ip6e_rthdr); 191 /* Destination options header(2nd part) */ 192 MAKE_EXTHDR(opt->ip6po_dest2, &exthdrs.ip6e_dest2); 193 } 194 195 #ifdef IPSEC 196 if ((flags & IPV6_FORWARDING) != 0) { 197 needipsec = 0; 198 goto skippolicycheck; 199 } 200 201 /* get a security policy for this packet */ 202 if (so == NULL) 203 sp = ipsec6_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND, 0, &error); 204 else { 205 if (IPSEC_PCB_SKIP_IPSEC(sotoinpcb_hdr(so)->inph_sp, 206 IPSEC_DIR_OUTBOUND)) { 207 needipsec = 0; 208 goto skippolicycheck; 209 } 210 sp = ipsec6_getpolicybysock(m, IPSEC_DIR_OUTBOUND, so, &error); 211 } 212 213 if (sp == NULL) { 214 ipsec6stat.out_inval++; 215 goto freehdrs; 216 } 217 218 error = 0; 219 220 /* check policy */ 221 switch (sp->policy) { 222 case IPSEC_POLICY_DISCARD: 223 /* 224 * This packet is just discarded. 225 */ 226 ipsec6stat.out_polvio++; 227 goto freehdrs; 228 229 case IPSEC_POLICY_BYPASS: 230 case IPSEC_POLICY_NONE: 231 /* no need to do IPsec. */ 232 needipsec = 0; 233 break; 234 235 case IPSEC_POLICY_IPSEC: 236 if (sp->req == NULL) { 237 /* XXX should be panic ? */ 238 printf("ip6_output: No IPsec request specified.\n"); 239 error = EINVAL; 240 goto freehdrs; 241 } 242 needipsec = 1; 243 break; 244 245 case IPSEC_POLICY_ENTRUST: 246 default: 247 printf("ip6_output: Invalid policy found. %d\n", sp->policy); 248 } 249 250 skippolicycheck:; 251 #endif /* IPSEC */ 252 253 if (needipsec && 254 (m->m_pkthdr.csum_flags & (M_CSUM_UDPv6|M_CSUM_TCPv6)) != 0) { 255 in6_delayed_cksum(m); 256 m->m_pkthdr.csum_flags &= ~(M_CSUM_UDPv6|M_CSUM_TCPv6); 257 } 258 259 /* 260 * Calculate the total length of the extension header chain. 261 * Keep the length of the unfragmentable part for fragmentation. 262 */ 263 optlen = 0; 264 if (exthdrs.ip6e_hbh) optlen += exthdrs.ip6e_hbh->m_len; 265 if (exthdrs.ip6e_dest1) optlen += exthdrs.ip6e_dest1->m_len; 266 if (exthdrs.ip6e_rthdr) optlen += exthdrs.ip6e_rthdr->m_len; 267 unfragpartlen = optlen + sizeof(struct ip6_hdr); 268 /* NOTE: we don't add AH/ESP length here. do that later. */ 269 if (exthdrs.ip6e_dest2) optlen += exthdrs.ip6e_dest2->m_len; 270 271 /* 272 * If we need IPsec, or there is at least one extension header, 273 * separate IP6 header from the payload. 274 */ 275 if ((needipsec || optlen) && !hdrsplit) { 276 if ((error = ip6_splithdr(m, &exthdrs)) != 0) { 277 m = NULL; 278 goto freehdrs; 279 } 280 m = exthdrs.ip6e_ip6; 281 hdrsplit++; 282 } 283 284 /* adjust pointer */ 285 ip6 = mtod(m, struct ip6_hdr *); 286 287 /* adjust mbuf packet header length */ 288 m->m_pkthdr.len += optlen; 289 plen = m->m_pkthdr.len - sizeof(*ip6); 290 291 /* If this is a jumbo payload, insert a jumbo payload option. */ 292 if (plen > IPV6_MAXPACKET) { 293 if (!hdrsplit) { 294 if ((error = ip6_splithdr(m, &exthdrs)) != 0) { 295 m = NULL; 296 goto freehdrs; 297 } 298 m = exthdrs.ip6e_ip6; 299 hdrsplit++; 300 } 301 /* adjust pointer */ 302 ip6 = mtod(m, struct ip6_hdr *); 303 if ((error = ip6_insert_jumboopt(&exthdrs, plen)) != 0) 304 goto freehdrs; 305 optlen += 8; /* XXX JUMBOOPTLEN */ 306 ip6->ip6_plen = 0; 307 } else 308 ip6->ip6_plen = htons(plen); 309 310 /* 311 * Concatenate headers and fill in next header fields. 312 * Here we have, on "m" 313 * IPv6 payload 314 * and we insert headers accordingly. Finally, we should be getting: 315 * IPv6 hbh dest1 rthdr ah* [esp* dest2 payload] 316 * 317 * during the header composing process, "m" points to IPv6 header. 318 * "mprev" points to an extension header prior to esp. 319 */ 320 { 321 u_char *nexthdrp = &ip6->ip6_nxt; 322 struct mbuf *mprev = m; 323 324 /* 325 * we treat dest2 specially. this makes IPsec processing 326 * much easier. the goal here is to make mprev point the 327 * mbuf prior to dest2. 328 * 329 * result: IPv6 dest2 payload 330 * m and mprev will point to IPv6 header. 331 */ 332 if (exthdrs.ip6e_dest2) { 333 if (!hdrsplit) 334 panic("assumption failed: hdr not split"); 335 exthdrs.ip6e_dest2->m_next = m->m_next; 336 m->m_next = exthdrs.ip6e_dest2; 337 *mtod(exthdrs.ip6e_dest2, u_char *) = ip6->ip6_nxt; 338 ip6->ip6_nxt = IPPROTO_DSTOPTS; 339 } 340 341 #define MAKE_CHAIN(m, mp, p, i)\ 342 do {\ 343 if (m) {\ 344 if (!hdrsplit) \ 345 panic("assumption failed: hdr not split"); \ 346 *mtod((m), u_char *) = *(p);\ 347 *(p) = (i);\ 348 p = mtod((m), u_char *);\ 349 (m)->m_next = (mp)->m_next;\ 350 (mp)->m_next = (m);\ 351 (mp) = (m);\ 352 }\ 353 } while (/*CONSTCOND*/ 0) 354 /* 355 * result: IPv6 hbh dest1 rthdr dest2 payload 356 * m will point to IPv6 header. mprev will point to the 357 * extension header prior to dest2 (rthdr in the above case). 358 */ 359 MAKE_CHAIN(exthdrs.ip6e_hbh, mprev, nexthdrp, IPPROTO_HOPOPTS); 360 MAKE_CHAIN(exthdrs.ip6e_dest1, mprev, nexthdrp, 361 IPPROTO_DSTOPTS); 362 MAKE_CHAIN(exthdrs.ip6e_rthdr, mprev, nexthdrp, 363 IPPROTO_ROUTING); 364 365 M_CSUM_DATA_IPv6_HL_SET(m->m_pkthdr.csum_data, 366 sizeof(struct ip6_hdr) + optlen); 367 368 #ifdef IPSEC 369 if (!needipsec) 370 goto skip_ipsec2; 371 372 /* 373 * pointers after IPsec headers are not valid any more. 374 * other pointers need a great care too. 375 * (IPsec routines should not mangle mbufs prior to AH/ESP) 376 */ 377 exthdrs.ip6e_dest2 = NULL; 378 379 { 380 struct ip6_rthdr *rh = NULL; 381 int segleft_org = 0; 382 struct ipsec_output_state state; 383 384 if (exthdrs.ip6e_rthdr) { 385 rh = mtod(exthdrs.ip6e_rthdr, struct ip6_rthdr *); 386 segleft_org = rh->ip6r_segleft; 387 rh->ip6r_segleft = 0; 388 } 389 390 bzero(&state, sizeof(state)); 391 state.m = m; 392 error = ipsec6_output_trans(&state, nexthdrp, mprev, sp, flags, 393 &needipsectun); 394 m = state.m; 395 if (error) { 396 /* mbuf is already reclaimed in ipsec6_output_trans. */ 397 m = NULL; 398 switch (error) { 399 case EHOSTUNREACH: 400 case ENETUNREACH: 401 case EMSGSIZE: 402 case ENOBUFS: 403 case ENOMEM: 404 break; 405 default: 406 printf("ip6_output (ipsec): error code %d\n", error); 407 /* FALLTHROUGH */ 408 case ENOENT: 409 /* don't show these error codes to the user */ 410 error = 0; 411 break; 412 } 413 goto bad; 414 } 415 if (exthdrs.ip6e_rthdr) { 416 /* ah6_output doesn't modify mbuf chain */ 417 rh->ip6r_segleft = segleft_org; 418 } 419 } 420 skip_ipsec2:; 421 #endif 422 } 423 424 /* 425 * If there is a routing header, replace destination address field 426 * with the first hop of the routing header. 427 */ 428 if (exthdrs.ip6e_rthdr) { 429 struct ip6_rthdr *rh; 430 struct ip6_rthdr0 *rh0; 431 struct in6_addr *addr; 432 433 rh = (struct ip6_rthdr *)(mtod(exthdrs.ip6e_rthdr, 434 struct ip6_rthdr *)); 435 finaldst = ip6->ip6_dst; 436 switch (rh->ip6r_type) { 437 case IPV6_RTHDR_TYPE_0: 438 rh0 = (struct ip6_rthdr0 *)rh; 439 addr = (struct in6_addr *)(rh0 + 1); 440 ip6->ip6_dst = addr[0]; 441 bcopy(&addr[1], &addr[0], 442 sizeof(struct in6_addr) * (rh0->ip6r0_segleft - 1)); 443 addr[rh0->ip6r0_segleft - 1] = finaldst; 444 break; 445 default: /* is it possible? */ 446 error = EINVAL; 447 goto bad; 448 } 449 } 450 451 /* Source address validation */ 452 if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) && 453 (flags & IPV6_UNSPECSRC) == 0) { 454 error = EOPNOTSUPP; 455 ip6stat.ip6s_badscope++; 456 goto bad; 457 } 458 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) { 459 error = EOPNOTSUPP; 460 ip6stat.ip6s_badscope++; 461 goto bad; 462 } 463 464 ip6stat.ip6s_localout++; 465 466 /* 467 * Route packet. 468 */ 469 /* initialize cached route */ 470 if (ro == 0) { 471 ro = &ip6route; 472 bzero((caddr_t)ro, sizeof(*ro)); 473 } 474 ro_pmtu = ro; 475 if (opt && opt->ip6po_rthdr) 476 ro = &opt->ip6po_route; 477 dst = (struct sockaddr_in6 *)&ro->ro_dst; 478 /* 479 * If there is a cached route, 480 * check that it is to the same destination 481 * and is still up. If not, free it and try again. 482 */ 483 if (ro->ro_rt && ((ro->ro_rt->rt_flags & RTF_UP) == 0 || 484 dst->sin6_family != AF_INET6 || 485 !IN6_ARE_ADDR_EQUAL(&dst->sin6_addr, &ip6->ip6_dst))) { 486 RTFREE(ro->ro_rt); 487 ro->ro_rt = (struct rtentry *)0; 488 } 489 if (ro->ro_rt == 0) { 490 bzero(dst, sizeof(*dst)); 491 dst->sin6_family = AF_INET6; 492 dst->sin6_len = sizeof(struct sockaddr_in6); 493 dst->sin6_addr = ip6->ip6_dst; 494 } 495 #ifdef IPSEC 496 if (needipsec && needipsectun) { 497 struct ipsec_output_state state; 498 499 /* 500 * All the extension headers will become inaccessible 501 * (since they can be encrypted). 502 * Don't panic, we need no more updates to extension headers 503 * on inner IPv6 packet (since they are now encapsulated). 504 * 505 * IPv6 [ESP|AH] IPv6 [extension headers] payload 506 */ 507 bzero(&exthdrs, sizeof(exthdrs)); 508 exthdrs.ip6e_ip6 = m; 509 510 bzero(&state, sizeof(state)); 511 state.m = m; 512 state.ro = (struct route *)ro; 513 state.dst = (struct sockaddr *)dst; 514 515 error = ipsec6_output_tunnel(&state, sp, flags); 516 517 m = state.m; 518 ro_pmtu = ro = (struct route_in6 *)state.ro; 519 dst = (struct sockaddr_in6 *)state.dst; 520 if (error) { 521 /* mbuf is already reclaimed in ipsec6_output_tunnel. */ 522 m0 = m = NULL; 523 m = NULL; 524 switch (error) { 525 case EHOSTUNREACH: 526 case ENETUNREACH: 527 case EMSGSIZE: 528 case ENOBUFS: 529 case ENOMEM: 530 break; 531 default: 532 printf("ip6_output (ipsec): error code %d\n", error); 533 /* FALLTHROUGH */ 534 case ENOENT: 535 /* don't show these error codes to the user */ 536 error = 0; 537 break; 538 } 539 goto bad; 540 } 541 542 exthdrs.ip6e_ip6 = m; 543 } 544 #endif /* IPSEC */ 545 546 if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) { 547 /* Unicast */ 548 549 #define ifatoia6(ifa) ((struct in6_ifaddr *)(ifa)) 550 #define sin6tosa(sin6) ((struct sockaddr *)(sin6)) 551 /* xxx 552 * interface selection comes here 553 * if an interface is specified from an upper layer, 554 * ifp must point it. 555 */ 556 if (ro->ro_rt == 0) { 557 /* 558 * non-bsdi always clone routes, if parent is 559 * PRF_CLONING. 560 */ 561 rtalloc((struct route *)ro); 562 } 563 if (ro->ro_rt == 0) { 564 ip6stat.ip6s_noroute++; 565 error = EHOSTUNREACH; 566 /* XXX in6_ifstat_inc(ifp, ifs6_out_discard); */ 567 goto bad; 568 } 569 ifp = ro->ro_rt->rt_ifp; 570 ro->ro_rt->rt_use++; 571 if (ro->ro_rt->rt_flags & RTF_GATEWAY) 572 dst = (struct sockaddr_in6 *)ro->ro_rt->rt_gateway; 573 m->m_flags &= ~(M_BCAST | M_MCAST); /* just in case */ 574 575 in6_ifstat_inc(ifp, ifs6_out_request); 576 577 /* 578 * Check if the outgoing interface conflicts with 579 * the interface specified by ifi6_ifindex (if specified). 580 * Note that loopback interface is always okay. 581 * (this may happen when we are sending a packet to one of 582 * our own addresses.) 583 */ 584 if (opt && opt->ip6po_pktinfo && 585 opt->ip6po_pktinfo->ipi6_ifindex) { 586 if (!(ifp->if_flags & IFF_LOOPBACK) && 587 ifp->if_index != opt->ip6po_pktinfo->ipi6_ifindex) { 588 ip6stat.ip6s_noroute++; 589 in6_ifstat_inc(ifp, ifs6_out_discard); 590 error = EHOSTUNREACH; 591 goto bad; 592 } 593 } 594 595 if (opt && opt->ip6po_hlim != -1) 596 ip6->ip6_hlim = opt->ip6po_hlim & 0xff; 597 } else { 598 /* Multicast */ 599 struct in6_multi *in6m; 600 601 m->m_flags = (m->m_flags & ~M_BCAST) | M_MCAST; 602 603 /* 604 * See if the caller provided any multicast options 605 */ 606 ifp = NULL; 607 if (im6o != NULL) { 608 ip6->ip6_hlim = im6o->im6o_multicast_hlim; 609 if (im6o->im6o_multicast_ifp != NULL) 610 ifp = im6o->im6o_multicast_ifp; 611 } else 612 ip6->ip6_hlim = ip6_defmcasthlim; 613 614 /* 615 * See if the caller provided the outgoing interface 616 * as an ancillary data. 617 * Boundary check for ifindex is assumed to be already done. 618 */ 619 if (opt && opt->ip6po_pktinfo && opt->ip6po_pktinfo->ipi6_ifindex) 620 ifp = ifindex2ifnet[opt->ip6po_pktinfo->ipi6_ifindex]; 621 622 /* 623 * If the destination is a node-local scope multicast, 624 * the packet should be loop-backed only. 625 */ 626 if (IN6_IS_ADDR_MC_NODELOCAL(&ip6->ip6_dst)) { 627 /* 628 * If the outgoing interface is already specified, 629 * it should be a loopback interface. 630 */ 631 if (ifp && (ifp->if_flags & IFF_LOOPBACK) == 0) { 632 ip6stat.ip6s_badscope++; 633 error = ENETUNREACH; /* XXX: better error? */ 634 /* XXX correct ifp? */ 635 in6_ifstat_inc(ifp, ifs6_out_discard); 636 goto bad; 637 } else 638 ifp = lo0ifp; 639 } 640 641 if (opt && opt->ip6po_hlim != -1) 642 ip6->ip6_hlim = opt->ip6po_hlim & 0xff; 643 644 /* 645 * If caller did not provide an interface lookup a 646 * default in the routing table. This is either a 647 * default for the speicfied group (i.e. a host 648 * route), or a multicast default (a route for the 649 * ``net'' ff00::/8). 650 */ 651 if (ifp == NULL) { 652 if (ro->ro_rt == 0) { 653 ro->ro_rt = rtalloc1((struct sockaddr *) 654 &ro->ro_dst, 0); 655 } 656 if (ro->ro_rt == 0) { 657 ip6stat.ip6s_noroute++; 658 error = EHOSTUNREACH; 659 /* XXX in6_ifstat_inc(ifp, ifs6_out_discard) */ 660 goto bad; 661 } 662 ifp = ro->ro_rt->rt_ifp; 663 ro->ro_rt->rt_use++; 664 } 665 666 if ((flags & IPV6_FORWARDING) == 0) 667 in6_ifstat_inc(ifp, ifs6_out_request); 668 in6_ifstat_inc(ifp, ifs6_out_mcast); 669 670 /* 671 * Confirm that the outgoing interface supports multicast. 672 */ 673 if ((ifp->if_flags & IFF_MULTICAST) == 0) { 674 ip6stat.ip6s_noroute++; 675 in6_ifstat_inc(ifp, ifs6_out_discard); 676 error = ENETUNREACH; 677 goto bad; 678 } 679 IN6_LOOKUP_MULTI(ip6->ip6_dst, ifp, in6m); 680 if (in6m != NULL && 681 (im6o == NULL || im6o->im6o_multicast_loop)) { 682 /* 683 * If we belong to the destination multicast group 684 * on the outgoing interface, and the caller did not 685 * forbid loopback, loop back a copy. 686 */ 687 ip6_mloopback(ifp, m, dst); 688 } else { 689 /* 690 * If we are acting as a multicast router, perform 691 * multicast forwarding as if the packet had just 692 * arrived on the interface to which we are about 693 * to send. The multicast forwarding function 694 * recursively calls this function, using the 695 * IPV6_FORWARDING flag to prevent infinite recursion. 696 * 697 * Multicasts that are looped back by ip6_mloopback(), 698 * above, will be forwarded by the ip6_input() routine, 699 * if necessary. 700 */ 701 if (ip6_mrouter && (flags & IPV6_FORWARDING) == 0) { 702 if (ip6_mforward(ip6, ifp, m) != 0) { 703 m_freem(m); 704 goto done; 705 } 706 } 707 } 708 /* 709 * Multicasts with a hoplimit of zero may be looped back, 710 * above, but must not be transmitted on a network. 711 * Also, multicasts addressed to the loopback interface 712 * are not sent -- the above call to ip6_mloopback() will 713 * loop back a copy if this host actually belongs to the 714 * destination group on the loopback interface. 715 */ 716 if (ip6->ip6_hlim == 0 || (ifp->if_flags & IFF_LOOPBACK)) { 717 m_freem(m); 718 goto done; 719 } 720 } 721 722 /* 723 * Fill the outgoing inteface to tell the upper layer 724 * to increment per-interface statistics. 725 */ 726 if (ifpp) 727 *ifpp = ifp; 728 729 /* Determine path MTU. */ 730 if ((error = ip6_getpmtu(ro_pmtu, ro, ifp, &finaldst, &mtu, 731 &alwaysfrag)) != 0) 732 goto bad; 733 #ifdef IPSEC 734 if (needipsectun) 735 mtu = IPV6_MMTU; 736 #endif 737 738 /* 739 * The caller of this function may specify to use the minimum MTU 740 * in some cases. 741 */ 742 if (mtu > IPV6_MMTU) { 743 if ((flags & IPV6_MINMTU)) 744 mtu = IPV6_MMTU; 745 } 746 747 /* Fake scoped addresses */ 748 if ((ifp->if_flags & IFF_LOOPBACK) != 0) { 749 /* 750 * If source or destination address is a scoped address, and 751 * the packet is going to be sent to a loopback interface, 752 * we should keep the original interface. 753 */ 754 755 /* 756 * XXX: this is a very experimental and temporary solution. 757 * We eventually have sockaddr_in6 and use the sin6_scope_id 758 * field of the structure here. 759 * We rely on the consistency between two scope zone ids 760 * of source add destination, which should already be assured 761 * Larger scopes than link will be supported in the near 762 * future. 763 */ 764 origifp = NULL; 765 if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src)) 766 origifp = ifindex2ifnet[ntohs(ip6->ip6_src.s6_addr16[1])]; 767 else if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst)) 768 origifp = ifindex2ifnet[ntohs(ip6->ip6_dst.s6_addr16[1])]; 769 /* 770 * XXX: origifp can be NULL even in those two cases above. 771 * For example, if we remove the (only) link-local address 772 * from the loopback interface, and try to send a link-local 773 * address without link-id information. Then the source 774 * address is ::1, and the destination address is the 775 * link-local address with its s6_addr16[1] being zero. 776 * What is worse, if the packet goes to the loopback interface 777 * by a default rejected route, the null pointer would be 778 * passed to looutput, and the kernel would hang. 779 * The following last resort would prevent such disaster. 780 */ 781 if (origifp == NULL) 782 origifp = ifp; 783 } else 784 origifp = ifp; 785 if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src)) 786 ip6->ip6_src.s6_addr16[1] = 0; 787 if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst)) 788 ip6->ip6_dst.s6_addr16[1] = 0; 789 790 /* 791 * If the outgoing packet contains a hop-by-hop options header, 792 * it must be examined and processed even by the source node. 793 * (RFC 2460, section 4.) 794 */ 795 if (exthdrs.ip6e_hbh) { 796 struct ip6_hbh *hbh = mtod(exthdrs.ip6e_hbh, struct ip6_hbh *); 797 u_int32_t dummy1; /* XXX unused */ 798 u_int32_t dummy2; /* XXX unused */ 799 800 /* 801 * XXX: if we have to send an ICMPv6 error to the sender, 802 * we need the M_LOOP flag since icmp6_error() expects 803 * the IPv6 and the hop-by-hop options header are 804 * continuous unless the flag is set. 805 */ 806 m->m_flags |= M_LOOP; 807 m->m_pkthdr.rcvif = ifp; 808 if (ip6_process_hopopts(m, (u_int8_t *)(hbh + 1), 809 ((hbh->ip6h_len + 1) << 3) - sizeof(struct ip6_hbh), 810 &dummy1, &dummy2) < 0) { 811 /* m was already freed at this point */ 812 error = EINVAL;/* better error? */ 813 goto done; 814 } 815 m->m_flags &= ~M_LOOP; /* XXX */ 816 m->m_pkthdr.rcvif = NULL; 817 } 818 819 #ifdef PFIL_HOOKS 820 /* 821 * Run through list of hooks for output packets. 822 */ 823 if ((error = pfil_run_hooks(&inet6_pfil_hook, &m, ifp, PFIL_OUT)) != 0) 824 goto done; 825 if (m == NULL) 826 goto done; 827 ip6 = mtod(m, struct ip6_hdr *); 828 #endif /* PFIL_HOOKS */ 829 /* 830 * Send the packet to the outgoing interface. 831 * If necessary, do IPv6 fragmentation before sending. 832 * 833 * the logic here is rather complex: 834 * 1: normal case (dontfrag == 0, alwaysfrag == 0) 835 * 1-a: send as is if tlen <= path mtu 836 * 1-b: fragment if tlen > path mtu 837 * 838 * 2: if user asks us not to fragment (dontfrag == 1) 839 * 2-a: send as is if tlen <= interface mtu 840 * 2-b: error if tlen > interface mtu 841 * 842 * 3: if we always need to attach fragment header (alwaysfrag == 1) 843 * always fragment 844 * 845 * 4: if dontfrag == 1 && alwaysfrag == 1 846 * error, as we cannot handle this conflicting request 847 */ 848 tlen = m->m_pkthdr.len; 849 850 dontfrag = 0; 851 if (dontfrag && alwaysfrag) { /* case 4 */ 852 /* conflicting request - can't transmit */ 853 error = EMSGSIZE; 854 goto bad; 855 } 856 if (dontfrag && tlen > IN6_LINKMTU(ifp)) { /* case 2-b */ 857 /* 858 * Even if the DONTFRAG option is specified, we cannot send the 859 * packet when the data length is larger than the MTU of the 860 * outgoing interface. 861 * Notify the error by sending IPV6_PATHMTU ancillary data as 862 * well as returning an error code (the latter is not described 863 * in the API spec.) 864 */ 865 u_int32_t mtu32; 866 struct ip6ctlparam ip6cp; 867 868 mtu32 = (u_int32_t)mtu; 869 bzero(&ip6cp, sizeof(ip6cp)); 870 ip6cp.ip6c_cmdarg = (void *)&mtu32; 871 pfctlinput2(PRC_MSGSIZE, (struct sockaddr *)&ro_pmtu->ro_dst, 872 (void *)&ip6cp); 873 874 error = EMSGSIZE; 875 goto bad; 876 } 877 878 /* 879 * transmit packet without fragmentation 880 */ 881 if (dontfrag || (!alwaysfrag && tlen <= mtu)) { /* case 1-a and 2-a */ 882 struct in6_ifaddr *ia6; 883 int sw_csum; 884 885 ip6 = mtod(m, struct ip6_hdr *); 886 ia6 = in6_ifawithifp(ifp, &ip6->ip6_src); 887 if (ia6) { 888 /* Record statistics for this interface address. */ 889 ia6->ia_ifa.ifa_data.ifad_outbytes += m->m_pkthdr.len; 890 } 891 #ifdef IPSEC 892 /* clean ipsec history once it goes out of the node */ 893 ipsec_delaux(m); 894 #endif 895 896 sw_csum = m->m_pkthdr.csum_flags & ~ifp->if_csum_flags_tx; 897 if ((sw_csum & (M_CSUM_UDPv6|M_CSUM_TCPv6)) != 0) { 898 in6_delayed_cksum(m); 899 m->m_pkthdr.csum_flags &= ~(M_CSUM_UDPv6|M_CSUM_TCPv6); 900 } 901 902 error = nd6_output(ifp, origifp, m, dst, ro->ro_rt); 903 goto done; 904 } 905 906 /* 907 * try to fragment the packet. case 1-b and 3 908 */ 909 if (mtu < IPV6_MMTU) { 910 /* path MTU cannot be less than IPV6_MMTU */ 911 error = EMSGSIZE; 912 in6_ifstat_inc(ifp, ifs6_out_fragfail); 913 goto bad; 914 } else if (ip6->ip6_plen == 0) { 915 /* jumbo payload cannot be fragmented */ 916 error = EMSGSIZE; 917 in6_ifstat_inc(ifp, ifs6_out_fragfail); 918 goto bad; 919 } else { 920 struct mbuf **mnext, *m_frgpart; 921 struct ip6_frag *ip6f; 922 u_int32_t id = htonl(ip6_randomid()); 923 u_char nextproto; 924 struct ip6ctlparam ip6cp; 925 u_int32_t mtu32; 926 927 /* 928 * Too large for the destination or interface; 929 * fragment if possible. 930 * Must be able to put at least 8 bytes per fragment. 931 */ 932 hlen = unfragpartlen; 933 if (mtu > IPV6_MAXPACKET) 934 mtu = IPV6_MAXPACKET; 935 936 /* Notify a proper path MTU to applications. */ 937 mtu32 = (u_int32_t)mtu; 938 bzero(&ip6cp, sizeof(ip6cp)); 939 ip6cp.ip6c_cmdarg = (void *)&mtu32; 940 pfctlinput2(PRC_MSGSIZE, (struct sockaddr *)&ro_pmtu->ro_dst, 941 (void *)&ip6cp); 942 943 len = (mtu - hlen - sizeof(struct ip6_frag)) & ~7; 944 if (len < 8) { 945 error = EMSGSIZE; 946 in6_ifstat_inc(ifp, ifs6_out_fragfail); 947 goto bad; 948 } 949 950 mnext = &m->m_nextpkt; 951 952 /* 953 * Change the next header field of the last header in the 954 * unfragmentable part. 955 */ 956 if (exthdrs.ip6e_rthdr) { 957 nextproto = *mtod(exthdrs.ip6e_rthdr, u_char *); 958 *mtod(exthdrs.ip6e_rthdr, u_char *) = IPPROTO_FRAGMENT; 959 } else if (exthdrs.ip6e_dest1) { 960 nextproto = *mtod(exthdrs.ip6e_dest1, u_char *); 961 *mtod(exthdrs.ip6e_dest1, u_char *) = IPPROTO_FRAGMENT; 962 } else if (exthdrs.ip6e_hbh) { 963 nextproto = *mtod(exthdrs.ip6e_hbh, u_char *); 964 *mtod(exthdrs.ip6e_hbh, u_char *) = IPPROTO_FRAGMENT; 965 } else { 966 nextproto = ip6->ip6_nxt; 967 ip6->ip6_nxt = IPPROTO_FRAGMENT; 968 } 969 970 if ((m->m_pkthdr.csum_flags & (M_CSUM_UDPv6|M_CSUM_TCPv6)) 971 != 0) { 972 in6_delayed_cksum(m); 973 m->m_pkthdr.csum_flags &= ~(M_CSUM_UDPv6|M_CSUM_TCPv6); 974 } 975 976 /* 977 * Loop through length of segment after first fragment, 978 * make new header and copy data of each part and link onto 979 * chain. 980 */ 981 m0 = m; 982 for (off = hlen; off < tlen; off += len) { 983 struct mbuf *mlast; 984 985 MGETHDR(m, M_DONTWAIT, MT_HEADER); 986 if (!m) { 987 error = ENOBUFS; 988 ip6stat.ip6s_odropped++; 989 goto sendorfree; 990 } 991 m->m_pkthdr.rcvif = NULL; 992 m->m_flags = m0->m_flags & M_COPYFLAGS; 993 *mnext = m; 994 mnext = &m->m_nextpkt; 995 m->m_data += max_linkhdr; 996 mhip6 = mtod(m, struct ip6_hdr *); 997 *mhip6 = *ip6; 998 m->m_len = sizeof(*mhip6); 999 error = ip6_insertfraghdr(m0, m, hlen, &ip6f); 1000 if (error) { 1001 ip6stat.ip6s_odropped++; 1002 goto sendorfree; 1003 } 1004 ip6f->ip6f_offlg = htons((u_int16_t)((off - hlen) & ~7)); 1005 if (off + len >= tlen) 1006 len = tlen - off; 1007 else 1008 ip6f->ip6f_offlg |= IP6F_MORE_FRAG; 1009 mhip6->ip6_plen = htons((u_int16_t)(len + hlen + 1010 sizeof(*ip6f) - sizeof(struct ip6_hdr))); 1011 if ((m_frgpart = m_copy(m0, off, len)) == 0) { 1012 error = ENOBUFS; 1013 ip6stat.ip6s_odropped++; 1014 goto sendorfree; 1015 } 1016 for (mlast = m; mlast->m_next; mlast = mlast->m_next) 1017 ; 1018 mlast->m_next = m_frgpart; 1019 m->m_pkthdr.len = len + hlen + sizeof(*ip6f); 1020 m->m_pkthdr.rcvif = (struct ifnet *)0; 1021 ip6f->ip6f_reserved = 0; 1022 ip6f->ip6f_ident = id; 1023 ip6f->ip6f_nxt = nextproto; 1024 ip6stat.ip6s_ofragments++; 1025 in6_ifstat_inc(ifp, ifs6_out_fragcreat); 1026 } 1027 1028 in6_ifstat_inc(ifp, ifs6_out_fragok); 1029 } 1030 1031 /* 1032 * Remove leading garbages. 1033 */ 1034 sendorfree: 1035 m = m0->m_nextpkt; 1036 m0->m_nextpkt = 0; 1037 m_freem(m0); 1038 for (m0 = m; m; m = m0) { 1039 m0 = m->m_nextpkt; 1040 m->m_nextpkt = 0; 1041 if (error == 0) { 1042 struct in6_ifaddr *ia6; 1043 ip6 = mtod(m, struct ip6_hdr *); 1044 ia6 = in6_ifawithifp(ifp, &ip6->ip6_src); 1045 if (ia6) { 1046 /* 1047 * Record statistics for this interface 1048 * address. 1049 */ 1050 ia6->ia_ifa.ifa_data.ifad_outbytes += 1051 m->m_pkthdr.len; 1052 } 1053 #ifdef IPSEC 1054 /* clean ipsec history once it goes out of the node */ 1055 ipsec_delaux(m); 1056 #endif 1057 error = nd6_output(ifp, origifp, m, dst, ro->ro_rt); 1058 } else 1059 m_freem(m); 1060 } 1061 1062 if (error == 0) 1063 ip6stat.ip6s_fragmented++; 1064 1065 done: 1066 if (ro == &ip6route && ro->ro_rt) { /* brace necessary for RTFREE */ 1067 RTFREE(ro->ro_rt); 1068 } else if (ro_pmtu == &ip6route && ro_pmtu->ro_rt) { 1069 RTFREE(ro_pmtu->ro_rt); 1070 } 1071 1072 #ifdef IPSEC 1073 if (sp != NULL) 1074 key_freesp(sp); 1075 #endif /* IPSEC */ 1076 1077 return (error); 1078 1079 freehdrs: 1080 m_freem(exthdrs.ip6e_hbh); /* m_freem will check if mbuf is 0 */ 1081 m_freem(exthdrs.ip6e_dest1); 1082 m_freem(exthdrs.ip6e_rthdr); 1083 m_freem(exthdrs.ip6e_dest2); 1084 /* FALLTHROUGH */ 1085 bad: 1086 m_freem(m); 1087 goto done; 1088 } 1089 1090 static int 1091 ip6_copyexthdr(mp, hdr, hlen) 1092 struct mbuf **mp; 1093 caddr_t hdr; 1094 int hlen; 1095 { 1096 struct mbuf *m; 1097 1098 if (hlen > MCLBYTES) 1099 return (ENOBUFS); /* XXX */ 1100 1101 MGET(m, M_DONTWAIT, MT_DATA); 1102 if (!m) 1103 return (ENOBUFS); 1104 1105 if (hlen > MLEN) { 1106 MCLGET(m, M_DONTWAIT); 1107 if ((m->m_flags & M_EXT) == 0) { 1108 m_free(m); 1109 return (ENOBUFS); 1110 } 1111 } 1112 m->m_len = hlen; 1113 if (hdr) 1114 bcopy(hdr, mtod(m, caddr_t), hlen); 1115 1116 *mp = m; 1117 return (0); 1118 } 1119 1120 /* 1121 * Process a delayed payload checksum calculation. 1122 */ 1123 void 1124 in6_delayed_cksum(struct mbuf *m) 1125 { 1126 uint16_t csum, offset; 1127 1128 KASSERT((m->m_pkthdr.csum_flags & (M_CSUM_UDPv6|M_CSUM_TCPv6)) != 0); 1129 KASSERT((~m->m_pkthdr.csum_flags & (M_CSUM_UDPv6|M_CSUM_TCPv6)) != 0); 1130 KASSERT((m->m_pkthdr.csum_flags 1131 & (M_CSUM_UDPv4|M_CSUM_TCPv4|M_CSUM_TSOv4)) == 0); 1132 1133 offset = M_CSUM_DATA_IPv6_HL(m->m_pkthdr.csum_data); 1134 csum = in6_cksum(m, 0, offset, m->m_pkthdr.len - offset); 1135 if (csum == 0 && (m->m_pkthdr.csum_flags & M_CSUM_UDPv6) != 0) { 1136 csum = 0xffff; 1137 } 1138 1139 offset += M_CSUM_DATA_IPv6_OFFSET(m->m_pkthdr.csum_data); 1140 if ((offset + sizeof(csum)) > m->m_len) { 1141 m_copyback(m, offset, sizeof(csum), &csum); 1142 } else { 1143 *(uint16_t *)(mtod(m, caddr_t) + offset) = csum; 1144 } 1145 } 1146 1147 /* 1148 * Insert jumbo payload option. 1149 */ 1150 static int 1151 ip6_insert_jumboopt(exthdrs, plen) 1152 struct ip6_exthdrs *exthdrs; 1153 u_int32_t plen; 1154 { 1155 struct mbuf *mopt; 1156 u_int8_t *optbuf; 1157 u_int32_t v; 1158 1159 #define JUMBOOPTLEN 8 /* length of jumbo payload option and padding */ 1160 1161 /* 1162 * If there is no hop-by-hop options header, allocate new one. 1163 * If there is one but it doesn't have enough space to store the 1164 * jumbo payload option, allocate a cluster to store the whole options. 1165 * Otherwise, use it to store the options. 1166 */ 1167 if (exthdrs->ip6e_hbh == 0) { 1168 MGET(mopt, M_DONTWAIT, MT_DATA); 1169 if (mopt == 0) 1170 return (ENOBUFS); 1171 mopt->m_len = JUMBOOPTLEN; 1172 optbuf = mtod(mopt, u_int8_t *); 1173 optbuf[1] = 0; /* = ((JUMBOOPTLEN) >> 3) - 1 */ 1174 exthdrs->ip6e_hbh = mopt; 1175 } else { 1176 struct ip6_hbh *hbh; 1177 1178 mopt = exthdrs->ip6e_hbh; 1179 if (M_TRAILINGSPACE(mopt) < JUMBOOPTLEN) { 1180 /* 1181 * XXX assumption: 1182 * - exthdrs->ip6e_hbh is not referenced from places 1183 * other than exthdrs. 1184 * - exthdrs->ip6e_hbh is not an mbuf chain. 1185 */ 1186 int oldoptlen = mopt->m_len; 1187 struct mbuf *n; 1188 1189 /* 1190 * XXX: give up if the whole (new) hbh header does 1191 * not fit even in an mbuf cluster. 1192 */ 1193 if (oldoptlen + JUMBOOPTLEN > MCLBYTES) 1194 return (ENOBUFS); 1195 1196 /* 1197 * As a consequence, we must always prepare a cluster 1198 * at this point. 1199 */ 1200 MGET(n, M_DONTWAIT, MT_DATA); 1201 if (n) { 1202 MCLGET(n, M_DONTWAIT); 1203 if ((n->m_flags & M_EXT) == 0) { 1204 m_freem(n); 1205 n = NULL; 1206 } 1207 } 1208 if (!n) 1209 return (ENOBUFS); 1210 n->m_len = oldoptlen + JUMBOOPTLEN; 1211 bcopy(mtod(mopt, caddr_t), mtod(n, caddr_t), 1212 oldoptlen); 1213 optbuf = mtod(n, u_int8_t *) + oldoptlen; 1214 m_freem(mopt); 1215 mopt = exthdrs->ip6e_hbh = n; 1216 } else { 1217 optbuf = mtod(mopt, u_int8_t *) + mopt->m_len; 1218 mopt->m_len += JUMBOOPTLEN; 1219 } 1220 optbuf[0] = IP6OPT_PADN; 1221 optbuf[1] = 0; 1222 1223 /* 1224 * Adjust the header length according to the pad and 1225 * the jumbo payload option. 1226 */ 1227 hbh = mtod(mopt, struct ip6_hbh *); 1228 hbh->ip6h_len += (JUMBOOPTLEN >> 3); 1229 } 1230 1231 /* fill in the option. */ 1232 optbuf[2] = IP6OPT_JUMBO; 1233 optbuf[3] = 4; 1234 v = (u_int32_t)htonl(plen + JUMBOOPTLEN); 1235 bcopy(&v, &optbuf[4], sizeof(u_int32_t)); 1236 1237 /* finally, adjust the packet header length */ 1238 exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN; 1239 1240 return (0); 1241 #undef JUMBOOPTLEN 1242 } 1243 1244 /* 1245 * Insert fragment header and copy unfragmentable header portions. 1246 */ 1247 static int 1248 ip6_insertfraghdr(m0, m, hlen, frghdrp) 1249 struct mbuf *m0, *m; 1250 int hlen; 1251 struct ip6_frag **frghdrp; 1252 { 1253 struct mbuf *n, *mlast; 1254 1255 if (hlen > sizeof(struct ip6_hdr)) { 1256 n = m_copym(m0, sizeof(struct ip6_hdr), 1257 hlen - sizeof(struct ip6_hdr), M_DONTWAIT); 1258 if (n == 0) 1259 return (ENOBUFS); 1260 m->m_next = n; 1261 } else 1262 n = m; 1263 1264 /* Search for the last mbuf of unfragmentable part. */ 1265 for (mlast = n; mlast->m_next; mlast = mlast->m_next) 1266 ; 1267 1268 if ((mlast->m_flags & M_EXT) == 0 && 1269 M_TRAILINGSPACE(mlast) >= sizeof(struct ip6_frag)) { 1270 /* use the trailing space of the last mbuf for the fragment hdr */ 1271 *frghdrp = (struct ip6_frag *)(mtod(mlast, caddr_t) + 1272 mlast->m_len); 1273 mlast->m_len += sizeof(struct ip6_frag); 1274 m->m_pkthdr.len += sizeof(struct ip6_frag); 1275 } else { 1276 /* allocate a new mbuf for the fragment header */ 1277 struct mbuf *mfrg; 1278 1279 MGET(mfrg, M_DONTWAIT, MT_DATA); 1280 if (mfrg == 0) 1281 return (ENOBUFS); 1282 mfrg->m_len = sizeof(struct ip6_frag); 1283 *frghdrp = mtod(mfrg, struct ip6_frag *); 1284 mlast->m_next = mfrg; 1285 } 1286 1287 return (0); 1288 } 1289 1290 static int 1291 ip6_getpmtu(ro_pmtu, ro, ifp, dst, mtup, alwaysfragp) 1292 struct route_in6 *ro_pmtu, *ro; 1293 struct ifnet *ifp; 1294 struct in6_addr *dst; 1295 u_long *mtup; 1296 int *alwaysfragp; 1297 { 1298 u_int32_t mtu = 0; 1299 int alwaysfrag = 0; 1300 int error = 0; 1301 1302 if (ro_pmtu != ro) { 1303 /* The first hop and the final destination may differ. */ 1304 struct sockaddr_in6 *sa6_dst = 1305 (struct sockaddr_in6 *)&ro_pmtu->ro_dst; 1306 if (ro_pmtu->ro_rt && 1307 ((ro_pmtu->ro_rt->rt_flags & RTF_UP) == 0 || 1308 !IN6_ARE_ADDR_EQUAL(&sa6_dst->sin6_addr, dst))) { 1309 RTFREE(ro_pmtu->ro_rt); 1310 ro_pmtu->ro_rt = (struct rtentry *)NULL; 1311 } 1312 if (ro_pmtu->ro_rt == NULL) { 1313 bzero(sa6_dst, sizeof(*sa6_dst)); /* for safety */ 1314 sa6_dst->sin6_family = AF_INET6; 1315 sa6_dst->sin6_len = sizeof(struct sockaddr_in6); 1316 sa6_dst->sin6_addr = *dst; 1317 1318 rtalloc((struct route *)ro_pmtu); 1319 } 1320 } 1321 if (ro_pmtu->ro_rt) { 1322 u_int32_t ifmtu; 1323 1324 if (ifp == NULL) 1325 ifp = ro_pmtu->ro_rt->rt_ifp; 1326 ifmtu = IN6_LINKMTU(ifp); 1327 mtu = ro_pmtu->ro_rt->rt_rmx.rmx_mtu; 1328 if (mtu == 0) 1329 mtu = ifmtu; 1330 else if (mtu < IPV6_MMTU) { 1331 /* 1332 * RFC2460 section 5, last paragraph: 1333 * if we record ICMPv6 too big message with 1334 * mtu < IPV6_MMTU, transmit packets sized IPV6_MMTU 1335 * or smaller, with fragment header attached. 1336 * (fragment header is needed regardless from the 1337 * packet size, for translators to identify packets) 1338 */ 1339 alwaysfrag = 1; 1340 mtu = IPV6_MMTU; 1341 } else if (mtu > ifmtu) { 1342 /* 1343 * The MTU on the route is larger than the MTU on 1344 * the interface! This shouldn't happen, unless the 1345 * MTU of the interface has been changed after the 1346 * interface was brought up. Change the MTU in the 1347 * route to match the interface MTU (as long as the 1348 * field isn't locked). 1349 */ 1350 mtu = ifmtu; 1351 if (!(ro_pmtu->ro_rt->rt_rmx.rmx_locks & RTV_MTU)) 1352 ro_pmtu->ro_rt->rt_rmx.rmx_mtu = mtu; 1353 } 1354 } else if (ifp) { 1355 mtu = IN6_LINKMTU(ifp); 1356 } else 1357 error = EHOSTUNREACH; /* XXX */ 1358 1359 *mtup = mtu; 1360 if (alwaysfragp) 1361 *alwaysfragp = alwaysfrag; 1362 return (error); 1363 } 1364 1365 /* 1366 * IP6 socket option processing. 1367 */ 1368 int 1369 ip6_ctloutput(op, so, level, optname, mp) 1370 int op; 1371 struct socket *so; 1372 int level, optname; 1373 struct mbuf **mp; 1374 { 1375 struct in6pcb *in6p = sotoin6pcb(so); 1376 struct mbuf *m = *mp; 1377 int optval = 0; 1378 int error = 0; 1379 struct proc *p = curproc; /* XXX */ 1380 1381 if (level == IPPROTO_IPV6) { 1382 switch (op) { 1383 case PRCO_SETOPT: 1384 switch (optname) { 1385 case IPV6_PKTOPTIONS: 1386 /* m is freed in ip6_pcbopts */ 1387 return (ip6_pcbopts(&in6p->in6p_outputopts, 1388 m, so)); 1389 case IPV6_HOPOPTS: 1390 case IPV6_DSTOPTS: 1391 if (p == 0 || suser(p->p_ucred, &p->p_acflag)) { 1392 error = EPERM; 1393 break; 1394 } 1395 /* FALLTHROUGH */ 1396 case IPV6_UNICAST_HOPS: 1397 case IPV6_RECVOPTS: 1398 case IPV6_RECVRETOPTS: 1399 case IPV6_RECVDSTADDR: 1400 case IPV6_PKTINFO: 1401 case IPV6_HOPLIMIT: 1402 case IPV6_RTHDR: 1403 case IPV6_FAITH: 1404 case IPV6_V6ONLY: 1405 case IPV6_USE_MIN_MTU: 1406 if (!m || m->m_len != sizeof(int)) { 1407 error = EINVAL; 1408 break; 1409 } 1410 optval = *mtod(m, int *); 1411 switch (optname) { 1412 1413 case IPV6_UNICAST_HOPS: 1414 if (optval < -1 || optval >= 256) 1415 error = EINVAL; 1416 else { 1417 /* -1 = kernel default */ 1418 in6p->in6p_hops = optval; 1419 } 1420 break; 1421 #define OPTSET(bit) \ 1422 do { \ 1423 if (optval) \ 1424 in6p->in6p_flags |= (bit); \ 1425 else \ 1426 in6p->in6p_flags &= ~(bit); \ 1427 } while (/*CONSTCOND*/ 0) 1428 1429 case IPV6_RECVOPTS: 1430 OPTSET(IN6P_RECVOPTS); 1431 break; 1432 1433 case IPV6_RECVRETOPTS: 1434 OPTSET(IN6P_RECVRETOPTS); 1435 break; 1436 1437 case IPV6_RECVDSTADDR: 1438 OPTSET(IN6P_RECVDSTADDR); 1439 break; 1440 1441 case IPV6_PKTINFO: 1442 OPTSET(IN6P_PKTINFO); 1443 break; 1444 1445 case IPV6_HOPLIMIT: 1446 OPTSET(IN6P_HOPLIMIT); 1447 break; 1448 1449 case IPV6_HOPOPTS: 1450 OPTSET(IN6P_HOPOPTS); 1451 break; 1452 1453 case IPV6_DSTOPTS: 1454 OPTSET(IN6P_DSTOPTS); 1455 break; 1456 1457 case IPV6_RTHDR: 1458 OPTSET(IN6P_RTHDR); 1459 break; 1460 1461 case IPV6_FAITH: 1462 OPTSET(IN6P_FAITH); 1463 break; 1464 1465 case IPV6_USE_MIN_MTU: 1466 OPTSET(IN6P_MINMTU); 1467 break; 1468 1469 case IPV6_V6ONLY: 1470 /* 1471 * make setsockopt(IPV6_V6ONLY) 1472 * available only prior to bind(2). 1473 * see ipng mailing list, Jun 22 2001. 1474 */ 1475 if (in6p->in6p_lport || 1476 !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr)) { 1477 error = EINVAL; 1478 break; 1479 } 1480 #ifdef INET6_BINDV6ONLY 1481 if (!optval) 1482 error = EINVAL; 1483 #else 1484 OPTSET(IN6P_IPV6_V6ONLY); 1485 #endif 1486 break; 1487 } 1488 break; 1489 #undef OPTSET 1490 1491 case IPV6_MULTICAST_IF: 1492 case IPV6_MULTICAST_HOPS: 1493 case IPV6_MULTICAST_LOOP: 1494 case IPV6_JOIN_GROUP: 1495 case IPV6_LEAVE_GROUP: 1496 error = ip6_setmoptions(optname, 1497 &in6p->in6p_moptions, m); 1498 break; 1499 1500 case IPV6_PORTRANGE: 1501 optval = *mtod(m, int *); 1502 1503 switch (optval) { 1504 case IPV6_PORTRANGE_DEFAULT: 1505 in6p->in6p_flags &= ~(IN6P_LOWPORT); 1506 in6p->in6p_flags &= ~(IN6P_HIGHPORT); 1507 break; 1508 1509 case IPV6_PORTRANGE_HIGH: 1510 in6p->in6p_flags &= ~(IN6P_LOWPORT); 1511 in6p->in6p_flags |= IN6P_HIGHPORT; 1512 break; 1513 1514 case IPV6_PORTRANGE_LOW: 1515 in6p->in6p_flags &= ~(IN6P_HIGHPORT); 1516 in6p->in6p_flags |= IN6P_LOWPORT; 1517 break; 1518 1519 default: 1520 error = EINVAL; 1521 break; 1522 } 1523 break; 1524 1525 #ifdef IPSEC 1526 case IPV6_IPSEC_POLICY: 1527 { 1528 caddr_t req = NULL; 1529 size_t len = 0; 1530 1531 int priv = 0; 1532 if (p == 0 || suser(p->p_ucred, &p->p_acflag)) 1533 priv = 0; 1534 else 1535 priv = 1; 1536 if (m) { 1537 req = mtod(m, caddr_t); 1538 len = m->m_len; 1539 } 1540 error = ipsec6_set_policy(in6p, 1541 optname, req, len, priv); 1542 } 1543 break; 1544 #endif /* IPSEC */ 1545 1546 default: 1547 error = ENOPROTOOPT; 1548 break; 1549 } 1550 if (m) 1551 (void)m_free(m); 1552 break; 1553 1554 case PRCO_GETOPT: 1555 switch (optname) { 1556 1557 case IPV6_OPTIONS: 1558 case IPV6_RETOPTS: 1559 error = ENOPROTOOPT; 1560 break; 1561 1562 case IPV6_PKTOPTIONS: 1563 if (in6p->in6p_options) { 1564 *mp = m_copym(in6p->in6p_options, 0, 1565 M_COPYALL, M_WAIT); 1566 } else { 1567 *mp = m_get(M_WAIT, MT_SOOPTS); 1568 (*mp)->m_len = 0; 1569 } 1570 break; 1571 1572 case IPV6_HOPOPTS: 1573 case IPV6_DSTOPTS: 1574 if (p == 0 || suser(p->p_ucred, &p->p_acflag)) { 1575 error = EPERM; 1576 break; 1577 } 1578 /* FALLTHROUGH */ 1579 case IPV6_UNICAST_HOPS: 1580 case IPV6_RECVOPTS: 1581 case IPV6_RECVRETOPTS: 1582 case IPV6_RECVDSTADDR: 1583 case IPV6_PORTRANGE: 1584 case IPV6_PKTINFO: 1585 case IPV6_HOPLIMIT: 1586 case IPV6_RTHDR: 1587 case IPV6_FAITH: 1588 case IPV6_V6ONLY: 1589 case IPV6_USE_MIN_MTU: 1590 *mp = m = m_get(M_WAIT, MT_SOOPTS); 1591 m->m_len = sizeof(int); 1592 switch (optname) { 1593 1594 case IPV6_UNICAST_HOPS: 1595 optval = in6p->in6p_hops; 1596 break; 1597 1598 #define OPTBIT(bit) (in6p->in6p_flags & bit ? 1 : 0) 1599 1600 case IPV6_RECVOPTS: 1601 optval = OPTBIT(IN6P_RECVOPTS); 1602 break; 1603 1604 case IPV6_RECVRETOPTS: 1605 optval = OPTBIT(IN6P_RECVRETOPTS); 1606 break; 1607 1608 case IPV6_RECVDSTADDR: 1609 optval = OPTBIT(IN6P_RECVDSTADDR); 1610 break; 1611 1612 case IPV6_PORTRANGE: 1613 { 1614 int flags; 1615 flags = in6p->in6p_flags; 1616 if (flags & IN6P_HIGHPORT) 1617 optval = IPV6_PORTRANGE_HIGH; 1618 else if (flags & IN6P_LOWPORT) 1619 optval = IPV6_PORTRANGE_LOW; 1620 else 1621 optval = 0; 1622 break; 1623 } 1624 1625 case IPV6_PKTINFO: 1626 optval = OPTBIT(IN6P_PKTINFO); 1627 break; 1628 1629 case IPV6_HOPLIMIT: 1630 optval = OPTBIT(IN6P_HOPLIMIT); 1631 break; 1632 1633 case IPV6_HOPOPTS: 1634 optval = OPTBIT(IN6P_HOPOPTS); 1635 break; 1636 1637 case IPV6_DSTOPTS: 1638 optval = OPTBIT(IN6P_DSTOPTS); 1639 break; 1640 1641 case IPV6_RTHDR: 1642 optval = OPTBIT(IN6P_RTHDR); 1643 break; 1644 1645 case IPV6_FAITH: 1646 optval = OPTBIT(IN6P_FAITH); 1647 break; 1648 1649 case IPV6_V6ONLY: 1650 optval = OPTBIT(IN6P_IPV6_V6ONLY); 1651 break; 1652 1653 case IPV6_USE_MIN_MTU: 1654 optval = OPTBIT(IN6P_MINMTU); 1655 break; 1656 } 1657 *mtod(m, int *) = optval; 1658 break; 1659 1660 case IPV6_MULTICAST_IF: 1661 case IPV6_MULTICAST_HOPS: 1662 case IPV6_MULTICAST_LOOP: 1663 case IPV6_JOIN_GROUP: 1664 case IPV6_LEAVE_GROUP: 1665 error = ip6_getmoptions(optname, in6p->in6p_moptions, mp); 1666 break; 1667 1668 #if 0 /* defined(IPSEC) */ 1669 /* XXX: code broken */ 1670 case IPV6_IPSEC_POLICY: 1671 { 1672 caddr_t req = NULL; 1673 size_t len = 0; 1674 1675 if (m) { 1676 req = mtod(m, caddr_t); 1677 len = m->m_len; 1678 } 1679 error = ipsec6_get_policy(in6p, req, len, mp); 1680 break; 1681 } 1682 #endif /* IPSEC */ 1683 1684 default: 1685 error = ENOPROTOOPT; 1686 break; 1687 } 1688 break; 1689 } 1690 } else { 1691 error = EINVAL; 1692 if (op == PRCO_SETOPT && *mp) 1693 (void)m_free(*mp); 1694 } 1695 return (error); 1696 } 1697 1698 int 1699 ip6_raw_ctloutput(op, so, level, optname, mp) 1700 int op; 1701 struct socket *so; 1702 int level, optname; 1703 struct mbuf **mp; 1704 { 1705 int error = 0, optval, optlen; 1706 const int icmp6off = offsetof(struct icmp6_hdr, icmp6_cksum); 1707 struct in6pcb *in6p = sotoin6pcb(so); 1708 struct mbuf *m = *mp; 1709 1710 optlen = m ? m->m_len : 0; 1711 1712 if (level != IPPROTO_IPV6) { 1713 if (op == PRCO_SETOPT && *mp) 1714 (void)m_free(*mp); 1715 return (EINVAL); 1716 } 1717 1718 switch (optname) { 1719 case IPV6_CHECKSUM: 1720 /* 1721 * For ICMPv6 sockets, no modification allowed for checksum 1722 * offset, permit "no change" values to help existing apps. 1723 * 1724 * XXX 2292bis says: "An attempt to set IPV6_CHECKSUM 1725 * for an ICMPv6 socket will fail." 1726 * The current behavior does not meet 2292bis. 1727 */ 1728 switch (op) { 1729 case PRCO_SETOPT: 1730 if (optlen != sizeof(int)) { 1731 error = EINVAL; 1732 break; 1733 } 1734 optval = *mtod(m, int *); 1735 if ((optval % 2) != 0) { 1736 /* the API assumes even offset values */ 1737 error = EINVAL; 1738 } else if (so->so_proto->pr_protocol == 1739 IPPROTO_ICMPV6) { 1740 if (optval != icmp6off) 1741 error = EINVAL; 1742 } else 1743 in6p->in6p_cksum = optval; 1744 break; 1745 1746 case PRCO_GETOPT: 1747 if (so->so_proto->pr_protocol == IPPROTO_ICMPV6) 1748 optval = icmp6off; 1749 else 1750 optval = in6p->in6p_cksum; 1751 1752 *mp = m = m_get(M_WAIT, MT_SOOPTS); 1753 m->m_len = sizeof(int); 1754 *mtod(m, int *) = optval; 1755 break; 1756 1757 default: 1758 error = EINVAL; 1759 break; 1760 } 1761 break; 1762 1763 default: 1764 error = ENOPROTOOPT; 1765 break; 1766 } 1767 1768 if (op == PRCO_SETOPT && m) 1769 (void)m_free(m); 1770 1771 return (error); 1772 } 1773 1774 /* 1775 * Set up IP6 options in pcb for insertion in output packets. 1776 * Store in mbuf with pointer in pcbopt, adding pseudo-option 1777 * with destination address if source routed. 1778 */ 1779 static int 1780 ip6_pcbopts(pktopt, m, so) 1781 struct ip6_pktopts **pktopt; 1782 struct mbuf *m; 1783 struct socket *so; 1784 { 1785 struct ip6_pktopts *opt = *pktopt; 1786 int error = 0; 1787 struct proc *p = curproc; /* XXX */ 1788 int priv = 0; 1789 1790 /* turn off any old options. */ 1791 if (opt) { 1792 if (opt->ip6po_m) 1793 (void)m_free(opt->ip6po_m); 1794 } else 1795 opt = malloc(sizeof(*opt), M_IP6OPT, M_WAITOK); 1796 *pktopt = 0; 1797 1798 if (!m || m->m_len == 0) { 1799 /* 1800 * Only turning off any previous options. 1801 */ 1802 free(opt, M_IP6OPT); 1803 if (m) 1804 (void)m_free(m); 1805 return (0); 1806 } 1807 1808 /* set options specified by user. */ 1809 if (p && !suser(p->p_ucred, &p->p_acflag)) 1810 priv = 1; 1811 if ((error = ip6_setpktoptions(m, opt, priv)) != 0) { 1812 (void)m_free(m); 1813 free(opt, M_IP6OPT); 1814 return (error); 1815 } 1816 *pktopt = opt; 1817 return (0); 1818 } 1819 1820 /* 1821 * Set the IP6 multicast options in response to user setsockopt(). 1822 */ 1823 static int 1824 ip6_setmoptions(optname, im6op, m) 1825 int optname; 1826 struct ip6_moptions **im6op; 1827 struct mbuf *m; 1828 { 1829 int error = 0; 1830 u_int loop, ifindex; 1831 struct ipv6_mreq *mreq; 1832 struct ifnet *ifp; 1833 struct ip6_moptions *im6o = *im6op; 1834 struct route_in6 ro; 1835 struct sockaddr_in6 *dst; 1836 struct in6_multi_mship *imm; 1837 struct proc *p = curproc; /* XXX */ 1838 1839 if (im6o == NULL) { 1840 /* 1841 * No multicast option buffer attached to the pcb; 1842 * allocate one and initialize to default values. 1843 */ 1844 im6o = (struct ip6_moptions *) 1845 malloc(sizeof(*im6o), M_IPMOPTS, M_WAITOK); 1846 1847 if (im6o == NULL) 1848 return (ENOBUFS); 1849 *im6op = im6o; 1850 im6o->im6o_multicast_ifp = NULL; 1851 im6o->im6o_multicast_hlim = ip6_defmcasthlim; 1852 im6o->im6o_multicast_loop = IPV6_DEFAULT_MULTICAST_LOOP; 1853 LIST_INIT(&im6o->im6o_memberships); 1854 } 1855 1856 switch (optname) { 1857 1858 case IPV6_MULTICAST_IF: 1859 /* 1860 * Select the interface for outgoing multicast packets. 1861 */ 1862 if (m == NULL || m->m_len != sizeof(u_int)) { 1863 error = EINVAL; 1864 break; 1865 } 1866 bcopy(mtod(m, u_int *), &ifindex, sizeof(ifindex)); 1867 if (ifindex != 0) { 1868 if (ifindex < 0 || if_indexlim <= ifindex || 1869 !ifindex2ifnet[ifindex]) { 1870 error = ENXIO; /* XXX EINVAL? */ 1871 break; 1872 } 1873 ifp = ifindex2ifnet[ifindex]; 1874 if ((ifp->if_flags & IFF_MULTICAST) == 0) { 1875 error = EADDRNOTAVAIL; 1876 break; 1877 } 1878 } else 1879 ifp = NULL; 1880 im6o->im6o_multicast_ifp = ifp; 1881 break; 1882 1883 case IPV6_MULTICAST_HOPS: 1884 { 1885 /* 1886 * Set the IP6 hoplimit for outgoing multicast packets. 1887 */ 1888 int optval; 1889 if (m == NULL || m->m_len != sizeof(int)) { 1890 error = EINVAL; 1891 break; 1892 } 1893 bcopy(mtod(m, u_int *), &optval, sizeof(optval)); 1894 if (optval < -1 || optval >= 256) 1895 error = EINVAL; 1896 else if (optval == -1) 1897 im6o->im6o_multicast_hlim = ip6_defmcasthlim; 1898 else 1899 im6o->im6o_multicast_hlim = optval; 1900 break; 1901 } 1902 1903 case IPV6_MULTICAST_LOOP: 1904 /* 1905 * Set the loopback flag for outgoing multicast packets. 1906 * Must be zero or one. 1907 */ 1908 if (m == NULL || m->m_len != sizeof(u_int)) { 1909 error = EINVAL; 1910 break; 1911 } 1912 bcopy(mtod(m, u_int *), &loop, sizeof(loop)); 1913 if (loop > 1) { 1914 error = EINVAL; 1915 break; 1916 } 1917 im6o->im6o_multicast_loop = loop; 1918 break; 1919 1920 case IPV6_JOIN_GROUP: 1921 /* 1922 * Add a multicast group membership. 1923 * Group must be a valid IP6 multicast address. 1924 */ 1925 if (m == NULL || m->m_len != sizeof(struct ipv6_mreq)) { 1926 error = EINVAL; 1927 break; 1928 } 1929 mreq = mtod(m, struct ipv6_mreq *); 1930 if (IN6_IS_ADDR_UNSPECIFIED(&mreq->ipv6mr_multiaddr)) { 1931 /* 1932 * We use the unspecified address to specify to accept 1933 * all multicast addresses. Only super user is allowed 1934 * to do this. 1935 */ 1936 if (suser(p->p_ucred, &p->p_acflag)) 1937 { 1938 error = EACCES; 1939 break; 1940 } 1941 } else if (!IN6_IS_ADDR_MULTICAST(&mreq->ipv6mr_multiaddr)) { 1942 error = EINVAL; 1943 break; 1944 } 1945 1946 /* 1947 * If the interface is specified, validate it. 1948 * If no interface was explicitly specified, choose an 1949 * appropriate one according to the given multicast address. 1950 */ 1951 if (mreq->ipv6mr_interface != 0) { 1952 if (mreq->ipv6mr_interface < 0 || 1953 if_indexlim <= mreq->ipv6mr_interface || 1954 !ifindex2ifnet[mreq->ipv6mr_interface]) { 1955 error = ENXIO; /* XXX EINVAL? */ 1956 break; 1957 } 1958 ifp = ifindex2ifnet[mreq->ipv6mr_interface]; 1959 } else { 1960 /* 1961 * If the multicast address is in node-local scope, 1962 * the interface should be a loopback interface. 1963 * Otherwise, look up the routing table for the 1964 * address, and choose the outgoing interface. 1965 * XXX: is it a good approach? 1966 */ 1967 if (IN6_IS_ADDR_MC_NODELOCAL(&mreq->ipv6mr_multiaddr)) { 1968 ifp = lo0ifp; 1969 } else { 1970 ro.ro_rt = NULL; 1971 dst = (struct sockaddr_in6 *)&ro.ro_dst; 1972 bzero(dst, sizeof(*dst)); 1973 dst->sin6_len = sizeof(struct sockaddr_in6); 1974 dst->sin6_family = AF_INET6; 1975 dst->sin6_addr = mreq->ipv6mr_multiaddr; 1976 rtalloc((struct route *)&ro); 1977 if (ro.ro_rt == NULL) { 1978 error = EADDRNOTAVAIL; 1979 break; 1980 } 1981 ifp = ro.ro_rt->rt_ifp; 1982 rtfree(ro.ro_rt); 1983 } 1984 } 1985 1986 /* 1987 * See if we found an interface, and confirm that it 1988 * supports multicast 1989 */ 1990 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) { 1991 error = EADDRNOTAVAIL; 1992 break; 1993 } 1994 /* 1995 * Put interface index into the multicast address, 1996 * if the address has link-local scope. 1997 */ 1998 if (IN6_IS_ADDR_MC_LINKLOCAL(&mreq->ipv6mr_multiaddr)) { 1999 mreq->ipv6mr_multiaddr.s6_addr16[1] = 2000 htons(ifp->if_index); 2001 } 2002 /* 2003 * See if the membership already exists. 2004 */ 2005 for (imm = im6o->im6o_memberships.lh_first; 2006 imm != NULL; imm = imm->i6mm_chain.le_next) 2007 if (imm->i6mm_maddr->in6m_ifp == ifp && 2008 IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr, 2009 &mreq->ipv6mr_multiaddr)) 2010 break; 2011 if (imm != NULL) { 2012 error = EADDRINUSE; 2013 break; 2014 } 2015 /* 2016 * Everything looks good; add a new record to the multicast 2017 * address list for the given interface. 2018 */ 2019 imm = in6_joingroup(ifp, &mreq->ipv6mr_multiaddr, &error); 2020 if (!imm) 2021 break; 2022 LIST_INSERT_HEAD(&im6o->im6o_memberships, imm, i6mm_chain); 2023 break; 2024 2025 case IPV6_LEAVE_GROUP: 2026 /* 2027 * Drop a multicast group membership. 2028 * Group must be a valid IP6 multicast address. 2029 */ 2030 if (m == NULL || m->m_len != sizeof(struct ipv6_mreq)) { 2031 error = EINVAL; 2032 break; 2033 } 2034 mreq = mtod(m, struct ipv6_mreq *); 2035 if (IN6_IS_ADDR_UNSPECIFIED(&mreq->ipv6mr_multiaddr)) { 2036 if (suser(p->p_ucred, &p->p_acflag)) 2037 { 2038 error = EACCES; 2039 break; 2040 } 2041 } else if (!IN6_IS_ADDR_MULTICAST(&mreq->ipv6mr_multiaddr)) { 2042 error = EINVAL; 2043 break; 2044 } 2045 /* 2046 * If an interface address was specified, get a pointer 2047 * to its ifnet structure. 2048 */ 2049 if (mreq->ipv6mr_interface != 0) { 2050 if (mreq->ipv6mr_interface < 0 || 2051 if_indexlim <= mreq->ipv6mr_interface || 2052 !ifindex2ifnet[mreq->ipv6mr_interface]) { 2053 error = ENXIO; /* XXX EINVAL? */ 2054 break; 2055 } 2056 ifp = ifindex2ifnet[mreq->ipv6mr_interface]; 2057 } else 2058 ifp = NULL; 2059 /* 2060 * Put interface index into the multicast address, 2061 * if the address has link-local scope. 2062 */ 2063 if (IN6_IS_ADDR_MC_LINKLOCAL(&mreq->ipv6mr_multiaddr)) { 2064 mreq->ipv6mr_multiaddr.s6_addr16[1] = 2065 htons(mreq->ipv6mr_interface); 2066 } 2067 /* 2068 * Find the membership in the membership list. 2069 */ 2070 for (imm = im6o->im6o_memberships.lh_first; 2071 imm != NULL; imm = imm->i6mm_chain.le_next) { 2072 if ((ifp == NULL || imm->i6mm_maddr->in6m_ifp == ifp) && 2073 IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr, 2074 &mreq->ipv6mr_multiaddr)) 2075 break; 2076 } 2077 if (imm == NULL) { 2078 /* Unable to resolve interface */ 2079 error = EADDRNOTAVAIL; 2080 break; 2081 } 2082 /* 2083 * Give up the multicast address record to which the 2084 * membership points. 2085 */ 2086 LIST_REMOVE(imm, i6mm_chain); 2087 in6_leavegroup(imm); 2088 break; 2089 2090 default: 2091 error = EOPNOTSUPP; 2092 break; 2093 } 2094 2095 /* 2096 * If all options have default values, no need to keep the mbuf. 2097 */ 2098 if (im6o->im6o_multicast_ifp == NULL && 2099 im6o->im6o_multicast_hlim == ip6_defmcasthlim && 2100 im6o->im6o_multicast_loop == IPV6_DEFAULT_MULTICAST_LOOP && 2101 im6o->im6o_memberships.lh_first == NULL) { 2102 free(*im6op, M_IPMOPTS); 2103 *im6op = NULL; 2104 } 2105 2106 return (error); 2107 } 2108 2109 /* 2110 * Return the IP6 multicast options in response to user getsockopt(). 2111 */ 2112 static int 2113 ip6_getmoptions(optname, im6o, mp) 2114 int optname; 2115 struct ip6_moptions *im6o; 2116 struct mbuf **mp; 2117 { 2118 u_int *hlim, *loop, *ifindex; 2119 2120 *mp = m_get(M_WAIT, MT_SOOPTS); 2121 2122 switch (optname) { 2123 2124 case IPV6_MULTICAST_IF: 2125 ifindex = mtod(*mp, u_int *); 2126 (*mp)->m_len = sizeof(u_int); 2127 if (im6o == NULL || im6o->im6o_multicast_ifp == NULL) 2128 *ifindex = 0; 2129 else 2130 *ifindex = im6o->im6o_multicast_ifp->if_index; 2131 return (0); 2132 2133 case IPV6_MULTICAST_HOPS: 2134 hlim = mtod(*mp, u_int *); 2135 (*mp)->m_len = sizeof(u_int); 2136 if (im6o == NULL) 2137 *hlim = ip6_defmcasthlim; 2138 else 2139 *hlim = im6o->im6o_multicast_hlim; 2140 return (0); 2141 2142 case IPV6_MULTICAST_LOOP: 2143 loop = mtod(*mp, u_int *); 2144 (*mp)->m_len = sizeof(u_int); 2145 if (im6o == NULL) 2146 *loop = ip6_defmcasthlim; 2147 else 2148 *loop = im6o->im6o_multicast_loop; 2149 return (0); 2150 2151 default: 2152 return (EOPNOTSUPP); 2153 } 2154 } 2155 2156 /* 2157 * Discard the IP6 multicast options. 2158 */ 2159 void 2160 ip6_freemoptions(im6o) 2161 struct ip6_moptions *im6o; 2162 { 2163 struct in6_multi_mship *imm; 2164 2165 if (im6o == NULL) 2166 return; 2167 2168 while ((imm = im6o->im6o_memberships.lh_first) != NULL) { 2169 LIST_REMOVE(imm, i6mm_chain); 2170 in6_leavegroup(imm); 2171 } 2172 free(im6o, M_IPMOPTS); 2173 } 2174 2175 /* 2176 * Set IPv6 outgoing packet options based on advanced API. 2177 */ 2178 int 2179 ip6_setpktoptions(control, opt, priv) 2180 struct mbuf *control; 2181 struct ip6_pktopts *opt; 2182 int priv; 2183 { 2184 struct cmsghdr *cm = 0; 2185 2186 if (control == 0 || opt == 0) 2187 return (EINVAL); 2188 2189 bzero(opt, sizeof(*opt)); 2190 opt->ip6po_hlim = -1; /* -1 means to use default hop limit */ 2191 2192 /* 2193 * XXX: Currently, we assume all the optional information is stored 2194 * in a single mbuf. 2195 */ 2196 if (control->m_next) 2197 return (EINVAL); 2198 2199 opt->ip6po_m = control; 2200 2201 for (; control->m_len; control->m_data += CMSG_ALIGN(cm->cmsg_len), 2202 control->m_len -= CMSG_ALIGN(cm->cmsg_len)) { 2203 cm = mtod(control, struct cmsghdr *); 2204 if (cm->cmsg_len == 0 || cm->cmsg_len > control->m_len) 2205 return (EINVAL); 2206 if (cm->cmsg_level != IPPROTO_IPV6) 2207 continue; 2208 2209 switch (cm->cmsg_type) { 2210 case IPV6_PKTINFO: 2211 if (cm->cmsg_len != CMSG_LEN(sizeof(struct in6_pktinfo))) 2212 return (EINVAL); 2213 opt->ip6po_pktinfo = (struct in6_pktinfo *)CMSG_DATA(cm); 2214 if (opt->ip6po_pktinfo->ipi6_ifindex && 2215 IN6_IS_ADDR_LINKLOCAL(&opt->ip6po_pktinfo->ipi6_addr)) 2216 opt->ip6po_pktinfo->ipi6_addr.s6_addr16[1] = 2217 htons(opt->ip6po_pktinfo->ipi6_ifindex); 2218 2219 if (opt->ip6po_pktinfo->ipi6_ifindex >= if_indexlim || 2220 opt->ip6po_pktinfo->ipi6_ifindex < 0) 2221 return (ENXIO); 2222 if (opt->ip6po_pktinfo->ipi6_ifindex > 0 && 2223 !ifindex2ifnet[opt->ip6po_pktinfo->ipi6_ifindex]) 2224 return (ENXIO); 2225 2226 /* 2227 * Check if the requested source address is indeed a 2228 * unicast address assigned to the node, and can be 2229 * used as the packet's source address. 2230 */ 2231 if (!IN6_IS_ADDR_UNSPECIFIED(&opt->ip6po_pktinfo->ipi6_addr)) { 2232 struct ifaddr *ia; 2233 struct in6_ifaddr *ia6; 2234 struct sockaddr_in6 sin6; 2235 2236 bzero(&sin6, sizeof(sin6)); 2237 sin6.sin6_len = sizeof(sin6); 2238 sin6.sin6_family = AF_INET6; 2239 sin6.sin6_addr = 2240 opt->ip6po_pktinfo->ipi6_addr; 2241 ia = ifa_ifwithaddr(sin6tosa(&sin6)); 2242 if (ia == NULL || 2243 (opt->ip6po_pktinfo->ipi6_ifindex && 2244 (ia->ifa_ifp->if_index != 2245 opt->ip6po_pktinfo->ipi6_ifindex))) { 2246 return (EADDRNOTAVAIL); 2247 } 2248 ia6 = (struct in6_ifaddr *)ia; 2249 if ((ia6->ia6_flags & (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY)) != 0) { 2250 return (EADDRNOTAVAIL); 2251 } 2252 2253 /* 2254 * Check if the requested source address is 2255 * indeed a unicast address assigned to the 2256 * node. 2257 */ 2258 if (IN6_IS_ADDR_MULTICAST(&opt->ip6po_pktinfo->ipi6_addr)) 2259 return (EADDRNOTAVAIL); 2260 } 2261 break; 2262 2263 case IPV6_HOPLIMIT: 2264 if (cm->cmsg_len != CMSG_LEN(sizeof(int))) 2265 return (EINVAL); 2266 else { 2267 int t; 2268 2269 bcopy(CMSG_DATA(cm), &t, sizeof(t)); 2270 if (t < -1 || t > 255) 2271 return (EINVAL); 2272 opt->ip6po_hlim = t; 2273 } 2274 break; 2275 2276 case IPV6_NEXTHOP: 2277 if (!priv) 2278 return (EPERM); 2279 2280 /* check if cmsg_len is large enough for sa_len */ 2281 if (cm->cmsg_len < sizeof(u_char) || 2282 cm->cmsg_len < CMSG_LEN(*CMSG_DATA(cm))) 2283 return (EINVAL); 2284 2285 opt->ip6po_nexthop = (struct sockaddr *)CMSG_DATA(cm); 2286 2287 break; 2288 2289 case IPV6_HOPOPTS: 2290 if (cm->cmsg_len < CMSG_LEN(sizeof(struct ip6_hbh))) 2291 return (EINVAL); 2292 else { 2293 struct ip6_hbh *t; 2294 2295 t = (struct ip6_hbh *)CMSG_DATA(cm); 2296 if (cm->cmsg_len != 2297 CMSG_LEN((t->ip6h_len + 1) << 3)) 2298 return (EINVAL); 2299 opt->ip6po_hbh = t; 2300 } 2301 break; 2302 2303 case IPV6_DSTOPTS: 2304 if (cm->cmsg_len < CMSG_LEN(sizeof(struct ip6_dest))) 2305 return (EINVAL); 2306 2307 /* 2308 * If there is no routing header yet, the destination 2309 * options header should be put on the 1st part. 2310 * Otherwise, the header should be on the 2nd part. 2311 * (See RFC 2460, section 4.1) 2312 */ 2313 if (opt->ip6po_rthdr == NULL) { 2314 struct ip6_dest *t; 2315 2316 t = (struct ip6_dest *)CMSG_DATA(cm); 2317 if (cm->cmsg_len != 2318 CMSG_LEN((t->ip6d_len + 1) << 3)); 2319 return (EINVAL); 2320 opt->ip6po_dest1 = t; 2321 } 2322 else { 2323 struct ip6_dest *t; 2324 2325 t = (struct ip6_dest *)CMSG_DATA(cm); 2326 if (cm->cmsg_len != 2327 CMSG_LEN((opt->ip6po_dest2->ip6d_len + 1) << 3)) 2328 return (EINVAL); 2329 opt->ip6po_dest2 = t; 2330 } 2331 break; 2332 2333 case IPV6_RTHDR: 2334 if (cm->cmsg_len < CMSG_LEN(sizeof(struct ip6_rthdr))) 2335 return (EINVAL); 2336 else { 2337 struct ip6_rthdr *t; 2338 2339 t = (struct ip6_rthdr *)CMSG_DATA(cm); 2340 if (cm->cmsg_len != 2341 CMSG_LEN((t->ip6r_len + 1) << 3)) 2342 return (EINVAL); 2343 switch (t->ip6r_type) { 2344 case IPV6_RTHDR_TYPE_0: 2345 if (t->ip6r_segleft == 0) 2346 return (EINVAL); 2347 break; 2348 default: 2349 return (EINVAL); 2350 } 2351 opt->ip6po_rthdr = t; 2352 } 2353 break; 2354 2355 default: 2356 return (ENOPROTOOPT); 2357 } 2358 } 2359 2360 return (0); 2361 } 2362 2363 /* 2364 * Routine called from ip6_output() to loop back a copy of an IP6 multicast 2365 * packet to the input queue of a specified interface. Note that this 2366 * calls the output routine of the loopback "driver", but with an interface 2367 * pointer that might NOT be lo0ifp -- easier than replicating that code here. 2368 */ 2369 void 2370 ip6_mloopback(ifp, m, dst) 2371 struct ifnet *ifp; 2372 struct mbuf *m; 2373 struct sockaddr_in6 *dst; 2374 { 2375 struct mbuf *copym; 2376 struct ip6_hdr *ip6; 2377 2378 copym = m_copy(m, 0, M_COPYALL); 2379 if (copym == NULL) 2380 return; 2381 2382 /* 2383 * Make sure to deep-copy IPv6 header portion in case the data 2384 * is in an mbuf cluster, so that we can safely override the IPv6 2385 * header portion later. 2386 */ 2387 if ((copym->m_flags & M_EXT) != 0 || 2388 copym->m_len < sizeof(struct ip6_hdr)) { 2389 copym = m_pullup(copym, sizeof(struct ip6_hdr)); 2390 if (copym == NULL) 2391 return; 2392 } 2393 2394 #ifdef DIAGNOSTIC 2395 if (copym->m_len < sizeof(*ip6)) { 2396 m_freem(copym); 2397 return; 2398 } 2399 #endif 2400 2401 ip6 = mtod(copym, struct ip6_hdr *); 2402 if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src)) 2403 ip6->ip6_src.s6_addr16[1] = 0; 2404 if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst)) 2405 ip6->ip6_dst.s6_addr16[1] = 0; 2406 2407 (void)looutput(ifp, copym, (struct sockaddr *)dst, NULL); 2408 } 2409 2410 /* 2411 * Chop IPv6 header off from the payload. 2412 */ 2413 static int 2414 ip6_splithdr(m, exthdrs) 2415 struct mbuf *m; 2416 struct ip6_exthdrs *exthdrs; 2417 { 2418 struct mbuf *mh; 2419 struct ip6_hdr *ip6; 2420 2421 ip6 = mtod(m, struct ip6_hdr *); 2422 if (m->m_len > sizeof(*ip6)) { 2423 MGETHDR(mh, M_DONTWAIT, MT_HEADER); 2424 if (mh == 0) { 2425 m_freem(m); 2426 return ENOBUFS; 2427 } 2428 M_COPY_PKTHDR(mh, m); 2429 MH_ALIGN(mh, sizeof(*ip6)); 2430 m_tag_delete_chain(m, NULL); 2431 m->m_flags &= ~M_PKTHDR; 2432 m->m_len -= sizeof(*ip6); 2433 m->m_data += sizeof(*ip6); 2434 mh->m_next = m; 2435 m = mh; 2436 m->m_len = sizeof(*ip6); 2437 bcopy((caddr_t)ip6, mtod(m, caddr_t), sizeof(*ip6)); 2438 } 2439 exthdrs->ip6e_ip6 = m; 2440 return 0; 2441 } 2442 2443 /* 2444 * Compute IPv6 extension header length. 2445 */ 2446 int 2447 ip6_optlen(in6p) 2448 struct in6pcb *in6p; 2449 { 2450 int len; 2451 2452 if (!in6p->in6p_outputopts) 2453 return 0; 2454 2455 len = 0; 2456 #define elen(x) \ 2457 (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0) 2458 2459 len += elen(in6p->in6p_outputopts->ip6po_hbh); 2460 len += elen(in6p->in6p_outputopts->ip6po_dest1); 2461 len += elen(in6p->in6p_outputopts->ip6po_rthdr); 2462 len += elen(in6p->in6p_outputopts->ip6po_dest2); 2463 return len; 2464 #undef elen 2465 } 2466
Indexes created Wed Sep 24 15:09:51 GMT 2025