ip6_output.c revision 1.63
1 /* $NetBSD: ip6_output.c,v 1.63 2003/08/22 20:20:09 jonathan 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 1.63 2003/08/22 20:20:09 jonathan 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 95 #ifdef IPSEC 96 #include <netinet6/ipsec.h> 97 #include <netkey/key.h> 98 #endif /* IPSEC */ 99 100 #include "loop.h" 101 102 #include <net/net_osdep.h> 103 104 #ifdef PFIL_HOOKS 105 extern struct pfil_head inet6_pfil_hook; /* XXX */ 106 #endif 107 108 struct ip6_exthdrs { 109 struct mbuf *ip6e_ip6; 110 struct mbuf *ip6e_hbh; 111 struct mbuf *ip6e_dest1; 112 struct mbuf *ip6e_rthdr; 113 struct mbuf *ip6e_dest2; 114 }; 115 116 static int ip6_pcbopts __P((struct ip6_pktopts **, struct mbuf *, 117 struct socket *)); 118 static int ip6_setmoptions __P((int, struct ip6_moptions **, struct mbuf *)); 119 static int ip6_getmoptions __P((int, struct ip6_moptions *, struct mbuf **)); 120 static int ip6_copyexthdr __P((struct mbuf **, caddr_t, int)); 121 static int ip6_insertfraghdr __P((struct mbuf *, struct mbuf *, int, 122 struct ip6_frag **)); 123 static int ip6_insert_jumboopt __P((struct ip6_exthdrs *, u_int32_t)); 124 static int ip6_splithdr __P((struct mbuf *, struct ip6_exthdrs *)); 125 static int ip6_getpmtu __P((struct route_in6 *, struct route_in6 *, 126 struct ifnet *, struct in6_addr *, u_long *)); 127 128 extern struct ifnet loif[NLOOP]; 129 130 /* 131 * IP6 output. The packet in mbuf chain m contains a skeletal IP6 132 * header (with pri, len, nxt, hlim, src, dst). 133 * This function may modify ver and hlim only. 134 * The mbuf chain containing the packet will be freed. 135 * The mbuf opt, if present, will not be freed. 136 * 137 * type of "mtu": rt_rmx.rmx_mtu is u_long, ifnet.ifr_mtu is int, and 138 * nd_ifinfo.linkmtu is u_int32_t. so we use u_long to hold largest one, 139 * which is rt_rmx.rmx_mtu. 140 */ 141 int 142 ip6_output(m0, opt, ro, flags, im6o, in6pcb, ifpp) 143 struct mbuf *m0; 144 struct ip6_pktopts *opt; 145 struct route_in6 *ro; 146 int flags; 147 struct ip6_moptions *im6o; 148 struct in6pcb *in6pcb; 149 struct ifnet **ifpp; /* XXX: just for statistics */ 150 { 151 struct ip6_hdr *ip6, *mhip6; 152 struct ifnet *ifp, *origifp; 153 struct mbuf *m = m0; 154 int hlen, tlen, len, off; 155 struct route_in6 ip6route; 156 struct sockaddr_in6 *dst; 157 int error = 0; 158 struct in6_ifaddr *ia; 159 u_long mtu; 160 u_int32_t optlen = 0, plen = 0, unfragpartlen = 0; 161 struct ip6_exthdrs exthdrs; 162 struct in6_addr finaldst; 163 struct route_in6 *ro_pmtu = NULL; 164 int hdrsplit = 0; 165 int needipsec = 0; 166 #ifdef IPSEC 167 int needipsectun = 0; 168 struct socket *so; 169 struct secpolicy *sp = NULL; 170 171 /* for AH processing. stupid to have "socket" variable in IP layer... */ 172 /* so = ipsec_getsocket(m); */ 173 so = in6pcb->in6p_socket; 174 (void)ipsec_setsocket(m, NULL); 175 ip6 = mtod(m, struct ip6_hdr *); 176 #endif /* IPSEC */ 177 178 #define MAKE_EXTHDR(hp, mp) \ 179 do { \ 180 if (hp) { \ 181 struct ip6_ext *eh = (struct ip6_ext *)(hp); \ 182 error = ip6_copyexthdr((mp), (caddr_t)(hp), \ 183 ((eh)->ip6e_len + 1) << 3); \ 184 if (error) \ 185 goto freehdrs; \ 186 } \ 187 } while (/*CONSTCOND*/ 0) 188 189 bzero(&exthdrs, sizeof(exthdrs)); 190 if (opt) { 191 /* Hop-by-Hop options header */ 192 MAKE_EXTHDR(opt->ip6po_hbh, &exthdrs.ip6e_hbh); 193 /* Destination options header(1st part) */ 194 MAKE_EXTHDR(opt->ip6po_dest1, &exthdrs.ip6e_dest1); 195 /* Routing header */ 196 MAKE_EXTHDR(opt->ip6po_rthdr, &exthdrs.ip6e_rthdr); 197 /* Destination options header(2nd part) */ 198 MAKE_EXTHDR(opt->ip6po_dest2, &exthdrs.ip6e_dest2); 199 } 200 201 #ifdef IPSEC 202 /* get a security policy for this packet */ 203 if (so == NULL) 204 sp = ipsec6_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND, 0, &error); 205 else 206 sp = ipsec6_getpolicybysock(m, IPSEC_DIR_OUTBOUND, so, &error); 207 208 if (sp == NULL) { 209 ipsec6stat.out_inval++; 210 goto freehdrs; 211 } 212 213 error = 0; 214 215 /* check policy */ 216 switch (sp->policy) { 217 case IPSEC_POLICY_DISCARD: 218 /* 219 * This packet is just discarded. 220 */ 221 ipsec6stat.out_polvio++; 222 goto freehdrs; 223 224 case IPSEC_POLICY_BYPASS: 225 case IPSEC_POLICY_NONE: 226 /* no need to do IPsec. */ 227 needipsec = 0; 228 break; 229 230 case IPSEC_POLICY_IPSEC: 231 if (sp->req == NULL) { 232 /* XXX should be panic ? */ 233 printf("ip6_output: No IPsec request specified.\n"); 234 error = EINVAL; 235 goto freehdrs; 236 } 237 needipsec = 1; 238 break; 239 240 case IPSEC_POLICY_ENTRUST: 241 default: 242 printf("ip6_output: Invalid policy found. %d\n", sp->policy); 243 } 244 #endif /* IPSEC */ 245 246 /* 247 * Calculate the total length of the extension header chain. 248 * Keep the length of the unfragmentable part for fragmentation. 249 */ 250 optlen = 0; 251 if (exthdrs.ip6e_hbh) optlen += exthdrs.ip6e_hbh->m_len; 252 if (exthdrs.ip6e_dest1) optlen += exthdrs.ip6e_dest1->m_len; 253 if (exthdrs.ip6e_rthdr) optlen += exthdrs.ip6e_rthdr->m_len; 254 unfragpartlen = optlen + sizeof(struct ip6_hdr); 255 /* NOTE: we don't add AH/ESP length here. do that later. */ 256 if (exthdrs.ip6e_dest2) optlen += exthdrs.ip6e_dest2->m_len; 257 258 /* 259 * If we need IPsec, or there is at least one extension header, 260 * separate IP6 header from the payload. 261 */ 262 if ((needipsec || optlen) && !hdrsplit) { 263 if ((error = ip6_splithdr(m, &exthdrs)) != 0) { 264 m = NULL; 265 goto freehdrs; 266 } 267 m = exthdrs.ip6e_ip6; 268 hdrsplit++; 269 } 270 271 /* adjust pointer */ 272 ip6 = mtod(m, struct ip6_hdr *); 273 274 /* adjust mbuf packet header length */ 275 m->m_pkthdr.len += optlen; 276 plen = m->m_pkthdr.len - sizeof(*ip6); 277 278 /* If this is a jumbo payload, insert a jumbo payload option. */ 279 if (plen > IPV6_MAXPACKET) { 280 if (!hdrsplit) { 281 if ((error = ip6_splithdr(m, &exthdrs)) != 0) { 282 m = NULL; 283 goto freehdrs; 284 } 285 m = exthdrs.ip6e_ip6; 286 hdrsplit++; 287 } 288 /* adjust pointer */ 289 ip6 = mtod(m, struct ip6_hdr *); 290 if ((error = ip6_insert_jumboopt(&exthdrs, plen)) != 0) 291 goto freehdrs; 292 ip6->ip6_plen = 0; 293 } else 294 ip6->ip6_plen = htons(plen); 295 296 /* 297 * Concatenate headers and fill in next header fields. 298 * Here we have, on "m" 299 * IPv6 payload 300 * and we insert headers accordingly. Finally, we should be getting: 301 * IPv6 hbh dest1 rthdr ah* [esp* dest2 payload] 302 * 303 * during the header composing process, "m" points to IPv6 header. 304 * "mprev" points to an extension header prior to esp. 305 */ 306 { 307 u_char *nexthdrp = &ip6->ip6_nxt; 308 struct mbuf *mprev = m; 309 310 /* 311 * we treat dest2 specially. this makes IPsec processing 312 * much easier. 313 * 314 * result: IPv6 dest2 payload 315 * m and mprev will point to IPv6 header. 316 */ 317 if (exthdrs.ip6e_dest2) { 318 if (!hdrsplit) 319 panic("assumption failed: hdr not split"); 320 exthdrs.ip6e_dest2->m_next = m->m_next; 321 m->m_next = exthdrs.ip6e_dest2; 322 *mtod(exthdrs.ip6e_dest2, u_char *) = ip6->ip6_nxt; 323 ip6->ip6_nxt = IPPROTO_DSTOPTS; 324 } 325 326 #define MAKE_CHAIN(m, mp, p, i)\ 327 do {\ 328 if (m) {\ 329 if (!hdrsplit) \ 330 panic("assumption failed: hdr not split"); \ 331 *mtod((m), u_char *) = *(p);\ 332 *(p) = (i);\ 333 p = mtod((m), u_char *);\ 334 (m)->m_next = (mp)->m_next;\ 335 (mp)->m_next = (m);\ 336 (mp) = (m);\ 337 }\ 338 } while (/*CONSTCOND*/ 0) 339 /* 340 * result: IPv6 hbh dest1 rthdr dest2 payload 341 * m will point to IPv6 header. mprev will point to the 342 * extension header prior to dest2 (rthdr in the above case). 343 */ 344 MAKE_CHAIN(exthdrs.ip6e_hbh, mprev, nexthdrp, IPPROTO_HOPOPTS); 345 MAKE_CHAIN(exthdrs.ip6e_dest1, mprev, nexthdrp, 346 IPPROTO_DSTOPTS); 347 MAKE_CHAIN(exthdrs.ip6e_rthdr, mprev, nexthdrp, 348 IPPROTO_ROUTING); 349 350 #ifdef IPSEC 351 if (!needipsec) 352 goto skip_ipsec2; 353 354 /* 355 * pointers after IPsec headers are not valid any more. 356 * other pointers need a great care too. 357 * (IPsec routines should not mangle mbufs prior to AH/ESP) 358 */ 359 exthdrs.ip6e_dest2 = NULL; 360 361 { 362 struct ip6_rthdr *rh = NULL; 363 int segleft_org = 0; 364 struct ipsec_output_state state; 365 366 if (exthdrs.ip6e_rthdr) { 367 rh = mtod(exthdrs.ip6e_rthdr, struct ip6_rthdr *); 368 segleft_org = rh->ip6r_segleft; 369 rh->ip6r_segleft = 0; 370 } 371 372 bzero(&state, sizeof(state)); 373 state.m = m; 374 error = ipsec6_output_trans(&state, nexthdrp, mprev, sp, flags, 375 &needipsectun); 376 m = state.m; 377 if (error) { 378 /* mbuf is already reclaimed in ipsec6_output_trans. */ 379 m = NULL; 380 switch (error) { 381 case EHOSTUNREACH: 382 case ENETUNREACH: 383 case EMSGSIZE: 384 case ENOBUFS: 385 case ENOMEM: 386 break; 387 default: 388 printf("ip6_output (ipsec): error code %d\n", error); 389 /* FALLTHROUGH */ 390 case ENOENT: 391 /* don't show these error codes to the user */ 392 error = 0; 393 break; 394 } 395 goto bad; 396 } 397 if (exthdrs.ip6e_rthdr) { 398 /* ah6_output doesn't modify mbuf chain */ 399 rh->ip6r_segleft = segleft_org; 400 } 401 } 402 skip_ipsec2:; 403 #endif 404 } 405 406 /* 407 * If there is a routing header, replace destination address field 408 * with the first hop of the routing header. 409 */ 410 if (exthdrs.ip6e_rthdr) { 411 struct ip6_rthdr *rh; 412 struct ip6_rthdr0 *rh0; 413 struct in6_addr *addr; 414 415 rh = (struct ip6_rthdr *)(mtod(exthdrs.ip6e_rthdr, 416 struct ip6_rthdr *)); 417 finaldst = ip6->ip6_dst; 418 switch (rh->ip6r_type) { 419 case IPV6_RTHDR_TYPE_0: 420 rh0 = (struct ip6_rthdr0 *)rh; 421 addr = (struct in6_addr *)(rh0 + 1); 422 ip6->ip6_dst = addr[0]; 423 bcopy(&addr[1], &addr[0], 424 sizeof(struct in6_addr) * (rh0->ip6r0_segleft - 1)); 425 addr[rh0->ip6r0_segleft - 1] = finaldst; 426 break; 427 default: /* is it possible? */ 428 error = EINVAL; 429 goto bad; 430 } 431 } 432 433 /* Source address validation */ 434 if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) && 435 (flags & IPV6_UNSPECSRC) == 0) { 436 error = EOPNOTSUPP; 437 ip6stat.ip6s_badscope++; 438 goto bad; 439 } 440 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) { 441 error = EOPNOTSUPP; 442 ip6stat.ip6s_badscope++; 443 goto bad; 444 } 445 446 ip6stat.ip6s_localout++; 447 448 /* 449 * Route packet. 450 */ 451 if (ro == 0) { 452 ro = &ip6route; 453 bzero((caddr_t)ro, sizeof(*ro)); 454 } 455 ro_pmtu = ro; 456 if (opt && opt->ip6po_rthdr) 457 ro = &opt->ip6po_route; 458 dst = (struct sockaddr_in6 *)&ro->ro_dst; 459 /* 460 * If there is a cached route, 461 * check that it is to the same destination 462 * and is still up. If not, free it and try again. 463 */ 464 if (ro->ro_rt && ((ro->ro_rt->rt_flags & RTF_UP) == 0 || 465 dst->sin6_family != AF_INET6 || 466 !IN6_ARE_ADDR_EQUAL(&dst->sin6_addr, &ip6->ip6_dst))) { 467 RTFREE(ro->ro_rt); 468 ro->ro_rt = (struct rtentry *)0; 469 } 470 if (ro->ro_rt == 0) { 471 bzero(dst, sizeof(*dst)); 472 dst->sin6_family = AF_INET6; 473 dst->sin6_len = sizeof(struct sockaddr_in6); 474 dst->sin6_addr = ip6->ip6_dst; 475 } 476 #ifdef IPSEC 477 if (needipsec && needipsectun) { 478 struct ipsec_output_state state; 479 480 /* 481 * All the extension headers will become inaccessible 482 * (since they can be encrypted). 483 * Don't panic, we need no more updates to extension headers 484 * on inner IPv6 packet (since they are now encapsulated). 485 * 486 * IPv6 [ESP|AH] IPv6 [extension headers] payload 487 */ 488 bzero(&exthdrs, sizeof(exthdrs)); 489 exthdrs.ip6e_ip6 = m; 490 491 bzero(&state, sizeof(state)); 492 state.m = m; 493 state.ro = (struct route *)ro; 494 state.dst = (struct sockaddr *)dst; 495 496 error = ipsec6_output_tunnel(&state, sp, flags); 497 498 m = state.m; 499 ro = (struct route_in6 *)state.ro; 500 dst = (struct sockaddr_in6 *)state.dst; 501 if (error) { 502 /* mbuf is already reclaimed in ipsec6_output_tunnel. */ 503 m0 = m = NULL; 504 m = NULL; 505 switch (error) { 506 case EHOSTUNREACH: 507 case ENETUNREACH: 508 case EMSGSIZE: 509 case ENOBUFS: 510 case ENOMEM: 511 break; 512 default: 513 printf("ip6_output (ipsec): error code %d\n", error); 514 /* FALLTHROUGH */ 515 case ENOENT: 516 /* don't show these error codes to the user */ 517 error = 0; 518 break; 519 } 520 goto bad; 521 } 522 523 exthdrs.ip6e_ip6 = m; 524 } 525 #endif /* IPSEC */ 526 527 if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) { 528 /* Unicast */ 529 530 #define ifatoia6(ifa) ((struct in6_ifaddr *)(ifa)) 531 #define sin6tosa(sin6) ((struct sockaddr *)(sin6)) 532 /* xxx 533 * interface selection comes here 534 * if an interface is specified from an upper layer, 535 * ifp must point it. 536 */ 537 if (ro->ro_rt == 0) { 538 /* 539 * non-bsdi always clone routes, if parent is 540 * PRF_CLONING. 541 */ 542 rtalloc((struct route *)ro); 543 } 544 if (ro->ro_rt == 0) { 545 ip6stat.ip6s_noroute++; 546 error = EHOSTUNREACH; 547 /* XXX in6_ifstat_inc(ifp, ifs6_out_discard); */ 548 goto bad; 549 } 550 ia = ifatoia6(ro->ro_rt->rt_ifa); 551 ifp = ro->ro_rt->rt_ifp; 552 ro->ro_rt->rt_use++; 553 if (ro->ro_rt->rt_flags & RTF_GATEWAY) 554 dst = (struct sockaddr_in6 *)ro->ro_rt->rt_gateway; 555 m->m_flags &= ~(M_BCAST | M_MCAST); /* just in case */ 556 557 in6_ifstat_inc(ifp, ifs6_out_request); 558 559 /* 560 * Check if the outgoing interface conflicts with 561 * the interface specified by ifi6_ifindex (if specified). 562 * Note that loopback interface is always okay. 563 * (this may happen when we are sending a packet to one of 564 * our own addresses.) 565 */ 566 if (opt && opt->ip6po_pktinfo 567 && opt->ip6po_pktinfo->ipi6_ifindex) { 568 if (!(ifp->if_flags & IFF_LOOPBACK) 569 && ifp->if_index != opt->ip6po_pktinfo->ipi6_ifindex) { 570 ip6stat.ip6s_noroute++; 571 in6_ifstat_inc(ifp, ifs6_out_discard); 572 error = EHOSTUNREACH; 573 goto bad; 574 } 575 } 576 577 if (opt && opt->ip6po_hlim != -1) 578 ip6->ip6_hlim = opt->ip6po_hlim & 0xff; 579 } else { 580 /* Multicast */ 581 struct in6_multi *in6m; 582 583 m->m_flags = (m->m_flags & ~M_BCAST) | M_MCAST; 584 585 /* 586 * See if the caller provided any multicast options 587 */ 588 ifp = NULL; 589 if (im6o != NULL) { 590 ip6->ip6_hlim = im6o->im6o_multicast_hlim; 591 if (im6o->im6o_multicast_ifp != NULL) 592 ifp = im6o->im6o_multicast_ifp; 593 } else 594 ip6->ip6_hlim = ip6_defmcasthlim; 595 596 /* 597 * See if the caller provided the outgoing interface 598 * as an ancillary data. 599 * Boundary check for ifindex is assumed to be already done. 600 */ 601 if (opt && opt->ip6po_pktinfo && opt->ip6po_pktinfo->ipi6_ifindex) 602 ifp = ifindex2ifnet[opt->ip6po_pktinfo->ipi6_ifindex]; 603 604 /* 605 * If the destination is a node-local scope multicast, 606 * the packet should be loop-backed only. 607 */ 608 if (IN6_IS_ADDR_MC_NODELOCAL(&ip6->ip6_dst)) { 609 /* 610 * If the outgoing interface is already specified, 611 * it should be a loopback interface. 612 */ 613 if (ifp && (ifp->if_flags & IFF_LOOPBACK) == 0) { 614 ip6stat.ip6s_badscope++; 615 error = ENETUNREACH; /* XXX: better error? */ 616 /* XXX correct ifp? */ 617 in6_ifstat_inc(ifp, ifs6_out_discard); 618 goto bad; 619 } else { 620 ifp = &loif[0]; 621 } 622 } 623 624 if (opt && opt->ip6po_hlim != -1) 625 ip6->ip6_hlim = opt->ip6po_hlim & 0xff; 626 627 /* 628 * If caller did not provide an interface lookup a 629 * default in the routing table. This is either a 630 * default for the speicfied group (i.e. a host 631 * route), or a multicast default (a route for the 632 * ``net'' ff00::/8). 633 */ 634 if (ifp == NULL) { 635 if (ro->ro_rt == 0) { 636 ro->ro_rt = rtalloc1((struct sockaddr *) 637 &ro->ro_dst, 0); 638 } 639 if (ro->ro_rt == 0) { 640 ip6stat.ip6s_noroute++; 641 error = EHOSTUNREACH; 642 /* XXX in6_ifstat_inc(ifp, ifs6_out_discard) */ 643 goto bad; 644 } 645 ia = ifatoia6(ro->ro_rt->rt_ifa); 646 ifp = ro->ro_rt->rt_ifp; 647 ro->ro_rt->rt_use++; 648 } 649 650 if ((flags & IPV6_FORWARDING) == 0) 651 in6_ifstat_inc(ifp, ifs6_out_request); 652 in6_ifstat_inc(ifp, ifs6_out_mcast); 653 654 /* 655 * Confirm that the outgoing interface supports multicast. 656 */ 657 if ((ifp->if_flags & IFF_MULTICAST) == 0) { 658 ip6stat.ip6s_noroute++; 659 in6_ifstat_inc(ifp, ifs6_out_discard); 660 error = ENETUNREACH; 661 goto bad; 662 } 663 IN6_LOOKUP_MULTI(ip6->ip6_dst, ifp, in6m); 664 if (in6m != NULL && 665 (im6o == NULL || im6o->im6o_multicast_loop)) { 666 /* 667 * If we belong to the destination multicast group 668 * on the outgoing interface, and the caller did not 669 * forbid loopback, loop back a copy. 670 */ 671 ip6_mloopback(ifp, m, dst); 672 } else { 673 /* 674 * If we are acting as a multicast router, perform 675 * multicast forwarding as if the packet had just 676 * arrived on the interface to which we are about 677 * to send. The multicast forwarding function 678 * recursively calls this function, using the 679 * IPV6_FORWARDING flag to prevent infinite recursion. 680 * 681 * Multicasts that are looped back by ip6_mloopback(), 682 * above, will be forwarded by the ip6_input() routine, 683 * if necessary. 684 */ 685 if (ip6_mrouter && (flags & IPV6_FORWARDING) == 0) { 686 if (ip6_mforward(ip6, ifp, m) != 0) { 687 m_freem(m); 688 goto done; 689 } 690 } 691 } 692 /* 693 * Multicasts with a hoplimit of zero may be looped back, 694 * above, but must not be transmitted on a network. 695 * Also, multicasts addressed to the loopback interface 696 * are not sent -- the above call to ip6_mloopback() will 697 * loop back a copy if this host actually belongs to the 698 * destination group on the loopback interface. 699 */ 700 if (ip6->ip6_hlim == 0 || (ifp->if_flags & IFF_LOOPBACK)) { 701 m_freem(m); 702 goto done; 703 } 704 } 705 706 /* 707 * Fill the outgoing inteface to tell the upper layer 708 * to increment per-interface statistics. 709 */ 710 if (ifpp) 711 *ifpp = ifp; 712 713 /* Determine path MTU. */ 714 if ((error = ip6_getpmtu(ro_pmtu, ro, ifp, &finaldst, &mtu)) != 0) 715 goto bad; 716 717 /* 718 * The caller of this function may specify to use the minimum MTU 719 * in some cases. 720 */ 721 if (mtu > IPV6_MMTU) { 722 if ((flags & IPV6_MINMTU)) 723 mtu = IPV6_MMTU; 724 } 725 726 /* Fake scoped addresses */ 727 if ((ifp->if_flags & IFF_LOOPBACK) != 0) { 728 /* 729 * If source or destination address is a scoped address, and 730 * the packet is going to be sent to a loopback interface, 731 * we should keep the original interface. 732 */ 733 734 /* 735 * XXX: this is a very experimental and temporary solution. 736 * We eventually have sockaddr_in6 and use the sin6_scope_id 737 * field of the structure here. 738 * We rely on the consistency between two scope zone ids 739 * of source add destination, which should already be assured 740 * Larger scopes than link will be supported in the near 741 * future. 742 */ 743 origifp = NULL; 744 if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src)) 745 origifp = ifindex2ifnet[ntohs(ip6->ip6_src.s6_addr16[1])]; 746 else if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst)) 747 origifp = ifindex2ifnet[ntohs(ip6->ip6_dst.s6_addr16[1])]; 748 /* 749 * XXX: origifp can be NULL even in those two cases above. 750 * For example, if we remove the (only) link-local address 751 * from the loopback interface, and try to send a link-local 752 * address without link-id information. Then the source 753 * address is ::1, and the destination address is the 754 * link-local address with its s6_addr16[1] being zero. 755 * What is worse, if the packet goes to the loopback interface 756 * by a default rejected route, the null pointer would be 757 * passed to looutput, and the kernel would hang. 758 * The following last resort would prevent such disaster. 759 */ 760 if (origifp == NULL) 761 origifp = ifp; 762 } else 763 origifp = ifp; 764 if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src)) 765 ip6->ip6_src.s6_addr16[1] = 0; 766 if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst)) 767 ip6->ip6_dst.s6_addr16[1] = 0; 768 769 /* 770 * If the outgoing packet contains a hop-by-hop options header, 771 * it must be examined and processed even by the source node. 772 * (RFC 2460, section 4.) 773 */ 774 if (exthdrs.ip6e_hbh) { 775 struct ip6_hbh *hbh = mtod(exthdrs.ip6e_hbh, struct ip6_hbh *); 776 u_int32_t dummy1; /* XXX unused */ 777 u_int32_t dummy2; /* XXX unused */ 778 779 /* 780 * XXX: if we have to send an ICMPv6 error to the sender, 781 * we need the M_LOOP flag since icmp6_error() expects 782 * the IPv6 and the hop-by-hop options header are 783 * continuous unless the flag is set. 784 */ 785 m->m_flags |= M_LOOP; 786 m->m_pkthdr.rcvif = ifp; 787 if (ip6_process_hopopts(m, (u_int8_t *)(hbh + 1), 788 ((hbh->ip6h_len + 1) << 3) - sizeof(struct ip6_hbh), 789 &dummy1, &dummy2) < 0) { 790 /* m was already freed at this point */ 791 error = EINVAL;/* better error? */ 792 goto done; 793 } 794 m->m_flags &= ~M_LOOP; /* XXX */ 795 m->m_pkthdr.rcvif = NULL; 796 } 797 798 #ifdef PFIL_HOOKS 799 /* 800 * Run through list of hooks for output packets. 801 */ 802 if ((error = pfil_run_hooks(&inet6_pfil_hook, &m, ifp, PFIL_OUT)) != 0) 803 goto done; 804 if (m == NULL) 805 goto done; 806 ip6 = mtod(m, struct ip6_hdr *); 807 #endif /* PFIL_HOOKS */ 808 /* 809 * Send the packet to the outgoing interface. 810 * If necessary, do IPv6 fragmentation before sending. 811 */ 812 tlen = m->m_pkthdr.len; 813 if (tlen <= mtu) { 814 #ifdef IFA_STATS 815 struct in6_ifaddr *ia6; 816 ip6 = mtod(m, struct ip6_hdr *); 817 ia6 = in6_ifawithifp(ifp, &ip6->ip6_src); 818 if (ia6) { 819 /* Record statistics for this interface address. */ 820 ia6->ia_ifa.ifa_data.ifad_outbytes += 821 m->m_pkthdr.len; 822 } 823 #endif 824 #ifdef IPSEC 825 /* clean ipsec history once it goes out of the node */ 826 ipsec_delaux(m); 827 #endif 828 error = nd6_output(ifp, origifp, m, dst, ro->ro_rt); 829 goto done; 830 } else if (mtu < IPV6_MMTU) { 831 /* 832 * note that path MTU is never less than IPV6_MMTU 833 * (see icmp6_input). 834 */ 835 error = EMSGSIZE; 836 in6_ifstat_inc(ifp, ifs6_out_fragfail); 837 goto bad; 838 } else if (ip6->ip6_plen == 0) { /* jumbo payload cannot be fragmented */ 839 error = EMSGSIZE; 840 in6_ifstat_inc(ifp, ifs6_out_fragfail); 841 goto bad; 842 } else { 843 struct mbuf **mnext, *m_frgpart; 844 struct ip6_frag *ip6f; 845 u_int32_t id = htonl(ip6_id++); 846 u_char nextproto; 847 848 /* 849 * Too large for the destination or interface; 850 * fragment if possible. 851 * Must be able to put at least 8 bytes per fragment. 852 */ 853 hlen = unfragpartlen; 854 if (mtu > IPV6_MAXPACKET) 855 mtu = IPV6_MAXPACKET; 856 len = (mtu - hlen - sizeof(struct ip6_frag)) & ~7; 857 if (len < 8) { 858 error = EMSGSIZE; 859 in6_ifstat_inc(ifp, ifs6_out_fragfail); 860 goto bad; 861 } 862 863 mnext = &m->m_nextpkt; 864 865 /* 866 * Change the next header field of the last header in the 867 * unfragmentable part. 868 */ 869 if (exthdrs.ip6e_rthdr) { 870 nextproto = *mtod(exthdrs.ip6e_rthdr, u_char *); 871 *mtod(exthdrs.ip6e_rthdr, u_char *) = IPPROTO_FRAGMENT; 872 } else if (exthdrs.ip6e_dest1) { 873 nextproto = *mtod(exthdrs.ip6e_dest1, u_char *); 874 *mtod(exthdrs.ip6e_dest1, u_char *) = IPPROTO_FRAGMENT; 875 } else if (exthdrs.ip6e_hbh) { 876 nextproto = *mtod(exthdrs.ip6e_hbh, u_char *); 877 *mtod(exthdrs.ip6e_hbh, u_char *) = IPPROTO_FRAGMENT; 878 } else { 879 nextproto = ip6->ip6_nxt; 880 ip6->ip6_nxt = IPPROTO_FRAGMENT; 881 } 882 883 /* 884 * Loop through length of segment after first fragment, 885 * make new header and copy data of each part and link onto 886 * chain. 887 */ 888 m0 = m; 889 for (off = hlen; off < tlen; off += len) { 890 MGETHDR(m, M_DONTWAIT, MT_HEADER); 891 if (!m) { 892 error = ENOBUFS; 893 ip6stat.ip6s_odropped++; 894 goto sendorfree; 895 } 896 m->m_flags = m0->m_flags & M_COPYFLAGS; 897 *mnext = m; 898 mnext = &m->m_nextpkt; 899 m->m_data += max_linkhdr; 900 mhip6 = mtod(m, struct ip6_hdr *); 901 *mhip6 = *ip6; 902 m->m_len = sizeof(*mhip6); 903 error = ip6_insertfraghdr(m0, m, hlen, &ip6f); 904 if (error) { 905 ip6stat.ip6s_odropped++; 906 goto sendorfree; 907 } 908 ip6f->ip6f_offlg = htons((u_short)((off - hlen) & ~7)); 909 if (off + len >= tlen) 910 len = tlen - off; 911 else 912 ip6f->ip6f_offlg |= IP6F_MORE_FRAG; 913 mhip6->ip6_plen = htons((u_short)(len + hlen + 914 sizeof(*ip6f) - sizeof(struct ip6_hdr))); 915 if ((m_frgpart = m_copy(m0, off, len)) == 0) { 916 error = ENOBUFS; 917 ip6stat.ip6s_odropped++; 918 goto sendorfree; 919 } 920 m_cat(m, m_frgpart); 921 m->m_pkthdr.len = len + hlen + sizeof(*ip6f); 922 m->m_pkthdr.rcvif = (struct ifnet *)0; 923 ip6f->ip6f_reserved = 0; 924 ip6f->ip6f_ident = id; 925 ip6f->ip6f_nxt = nextproto; 926 ip6stat.ip6s_ofragments++; 927 in6_ifstat_inc(ifp, ifs6_out_fragcreat); 928 } 929 930 in6_ifstat_inc(ifp, ifs6_out_fragok); 931 } 932 933 /* 934 * Remove leading garbages. 935 */ 936 sendorfree: 937 m = m0->m_nextpkt; 938 m0->m_nextpkt = 0; 939 m_freem(m0); 940 for (m0 = m; m; m = m0) { 941 m0 = m->m_nextpkt; 942 m->m_nextpkt = 0; 943 if (error == 0) { 944 #ifdef IFA_STATS 945 struct in6_ifaddr *ia6; 946 ip6 = mtod(m, struct ip6_hdr *); 947 ia6 = in6_ifawithifp(ifp, &ip6->ip6_src); 948 if (ia6) { 949 /* 950 * Record statistics for this interface 951 * address. 952 */ 953 ia6->ia_ifa.ifa_data.ifad_outbytes += 954 m->m_pkthdr.len; 955 } 956 #endif 957 #ifdef IPSEC 958 /* clean ipsec history once it goes out of the node */ 959 ipsec_delaux(m); 960 #endif 961 error = nd6_output(ifp, origifp, m, dst, ro->ro_rt); 962 } else 963 m_freem(m); 964 } 965 966 if (error == 0) 967 ip6stat.ip6s_fragmented++; 968 969 done: 970 if (ro == &ip6route && ro->ro_rt) { /* brace necessary for RTFREE */ 971 RTFREE(ro->ro_rt); 972 } else if (ro_pmtu == &ip6route && ro_pmtu->ro_rt) { 973 RTFREE(ro_pmtu->ro_rt); 974 } 975 976 #ifdef IPSEC 977 if (sp != NULL) 978 key_freesp(sp); 979 #endif /* IPSEC */ 980 981 return (error); 982 983 freehdrs: 984 m_freem(exthdrs.ip6e_hbh); /* m_freem will check if mbuf is 0 */ 985 m_freem(exthdrs.ip6e_dest1); 986 m_freem(exthdrs.ip6e_rthdr); 987 m_freem(exthdrs.ip6e_dest2); 988 /* FALLTHROUGH */ 989 bad: 990 m_freem(m); 991 goto done; 992 } 993 994 static int 995 ip6_copyexthdr(mp, hdr, hlen) 996 struct mbuf **mp; 997 caddr_t hdr; 998 int hlen; 999 { 1000 struct mbuf *m; 1001 1002 if (hlen > MCLBYTES) 1003 return (ENOBUFS); /* XXX */ 1004 1005 MGET(m, M_DONTWAIT, MT_DATA); 1006 if (!m) 1007 return (ENOBUFS); 1008 1009 if (hlen > MLEN) { 1010 MCLGET(m, M_DONTWAIT); 1011 if ((m->m_flags & M_EXT) == 0) { 1012 m_free(m); 1013 return (ENOBUFS); 1014 } 1015 } 1016 m->m_len = hlen; 1017 if (hdr) 1018 bcopy(hdr, mtod(m, caddr_t), hlen); 1019 1020 *mp = m; 1021 return (0); 1022 } 1023 1024 /* 1025 * Insert jumbo payload option. 1026 */ 1027 static int 1028 ip6_insert_jumboopt(exthdrs, plen) 1029 struct ip6_exthdrs *exthdrs; 1030 u_int32_t plen; 1031 { 1032 struct mbuf *mopt; 1033 u_int8_t *optbuf; 1034 u_int32_t v; 1035 1036 #define JUMBOOPTLEN 8 /* length of jumbo payload option and padding */ 1037 1038 /* 1039 * If there is no hop-by-hop options header, allocate new one. 1040 * If there is one but it doesn't have enough space to store the 1041 * jumbo payload option, allocate a cluster to store the whole options. 1042 * Otherwise, use it to store the options. 1043 */ 1044 if (exthdrs->ip6e_hbh == 0) { 1045 MGET(mopt, M_DONTWAIT, MT_DATA); 1046 if (mopt == 0) 1047 return (ENOBUFS); 1048 mopt->m_len = JUMBOOPTLEN; 1049 optbuf = mtod(mopt, u_int8_t *); 1050 optbuf[1] = 0; /* = ((JUMBOOPTLEN) >> 3) - 1 */ 1051 exthdrs->ip6e_hbh = mopt; 1052 } else { 1053 struct ip6_hbh *hbh; 1054 1055 mopt = exthdrs->ip6e_hbh; 1056 if (M_TRAILINGSPACE(mopt) < JUMBOOPTLEN) { 1057 /* 1058 * XXX assumption: 1059 * - exthdrs->ip6e_hbh is not referenced from places 1060 * other than exthdrs. 1061 * - exthdrs->ip6e_hbh is not an mbuf chain. 1062 */ 1063 int oldoptlen = mopt->m_len; 1064 struct mbuf *n; 1065 1066 /* 1067 * XXX: give up if the whole (new) hbh header does 1068 * not fit even in an mbuf cluster. 1069 */ 1070 if (oldoptlen + JUMBOOPTLEN > MCLBYTES) 1071 return (ENOBUFS); 1072 1073 /* 1074 * As a consequence, we must always prepare a cluster 1075 * at this point. 1076 */ 1077 MGET(n, M_DONTWAIT, MT_DATA); 1078 if (n) { 1079 MCLGET(n, M_DONTWAIT); 1080 if ((n->m_flags & M_EXT) == 0) { 1081 m_freem(n); 1082 n = NULL; 1083 } 1084 } 1085 if (!n) 1086 return (ENOBUFS); 1087 n->m_len = oldoptlen + JUMBOOPTLEN; 1088 bcopy(mtod(mopt, caddr_t), mtod(n, caddr_t), 1089 oldoptlen); 1090 optbuf = mtod(n, u_int8_t *) + oldoptlen; 1091 m_freem(mopt); 1092 mopt = exthdrs->ip6e_hbh = n; 1093 } else { 1094 optbuf = mtod(mopt, u_int8_t *) + mopt->m_len; 1095 mopt->m_len += JUMBOOPTLEN; 1096 } 1097 optbuf[0] = IP6OPT_PADN; 1098 optbuf[1] = 0; 1099 1100 /* 1101 * Adjust the header length according to the pad and 1102 * the jumbo payload option. 1103 */ 1104 hbh = mtod(mopt, struct ip6_hbh *); 1105 hbh->ip6h_len += (JUMBOOPTLEN >> 3); 1106 } 1107 1108 /* fill in the option. */ 1109 optbuf[2] = IP6OPT_JUMBO; 1110 optbuf[3] = 4; 1111 v = (u_int32_t)htonl(plen + JUMBOOPTLEN); 1112 bcopy(&v, &optbuf[4], sizeof(u_int32_t)); 1113 1114 /* finally, adjust the packet header length */ 1115 exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN; 1116 1117 return (0); 1118 #undef JUMBOOPTLEN 1119 } 1120 1121 /* 1122 * Insert fragment header and copy unfragmentable header portions. 1123 */ 1124 static int 1125 ip6_insertfraghdr(m0, m, hlen, frghdrp) 1126 struct mbuf *m0, *m; 1127 int hlen; 1128 struct ip6_frag **frghdrp; 1129 { 1130 struct mbuf *n, *mlast; 1131 1132 if (hlen > sizeof(struct ip6_hdr)) { 1133 n = m_copym(m0, sizeof(struct ip6_hdr), 1134 hlen - sizeof(struct ip6_hdr), M_DONTWAIT); 1135 if (n == 0) 1136 return (ENOBUFS); 1137 m->m_next = n; 1138 } else 1139 n = m; 1140 1141 /* Search for the last mbuf of unfragmentable part. */ 1142 for (mlast = n; mlast->m_next; mlast = mlast->m_next) 1143 ; 1144 1145 if ((mlast->m_flags & M_EXT) == 0 && 1146 M_TRAILINGSPACE(mlast) >= sizeof(struct ip6_frag)) { 1147 /* use the trailing space of the last mbuf for the fragment hdr */ 1148 *frghdrp = (struct ip6_frag *)(mtod(mlast, caddr_t) + 1149 mlast->m_len); 1150 mlast->m_len += sizeof(struct ip6_frag); 1151 m->m_pkthdr.len += sizeof(struct ip6_frag); 1152 } else { 1153 /* allocate a new mbuf for the fragment header */ 1154 struct mbuf *mfrg; 1155 1156 MGET(mfrg, M_DONTWAIT, MT_DATA); 1157 if (mfrg == 0) 1158 return (ENOBUFS); 1159 mfrg->m_len = sizeof(struct ip6_frag); 1160 *frghdrp = mtod(mfrg, struct ip6_frag *); 1161 mlast->m_next = mfrg; 1162 } 1163 1164 return (0); 1165 } 1166 1167 static int 1168 ip6_getpmtu(ro_pmtu, ro, ifp, dst, mtup) 1169 struct route_in6 *ro_pmtu, *ro; 1170 struct ifnet *ifp; 1171 struct in6_addr *dst; 1172 u_long *mtup; 1173 { 1174 u_int32_t mtu = 0; 1175 int error = 0; 1176 1177 if (ro_pmtu != ro) { 1178 /* The first hop and the final destination may differ. */ 1179 struct sockaddr_in6 *sa6_dst = 1180 (struct sockaddr_in6 *)&ro_pmtu->ro_dst; 1181 if (ro_pmtu->ro_rt && 1182 ((ro_pmtu->ro_rt->rt_flags & RTF_UP) == 0 || 1183 !IN6_ARE_ADDR_EQUAL(&sa6_dst->sin6_addr, dst))) { 1184 RTFREE(ro_pmtu->ro_rt); 1185 ro_pmtu->ro_rt = (struct rtentry *)0; 1186 } 1187 if (ro_pmtu->ro_rt == 0) { 1188 bzero(sa6_dst, sizeof(*sa6_dst)); 1189 sa6_dst->sin6_family = AF_INET6; 1190 sa6_dst->sin6_len = sizeof(struct sockaddr_in6); 1191 sa6_dst->sin6_addr = *dst; 1192 1193 rtalloc((struct route *)ro_pmtu); 1194 } 1195 } 1196 if (ro_pmtu->ro_rt) { 1197 u_int32_t ifmtu; 1198 1199 if (ifp == NULL) 1200 ifp = ro_pmtu->ro_rt->rt_ifp; 1201 ifmtu = IN6_LINKMTU(ifp); 1202 mtu = ro_pmtu->ro_rt->rt_rmx.rmx_mtu; 1203 if (mtu == 0) 1204 mtu = ifmtu; 1205 else if (mtu > ifmtu) { 1206 /* 1207 * The MTU on the route is larger than the MTU on 1208 * the interface! This shouldn't happen, unless the 1209 * MTU of the interface has been changed after the 1210 * interface was brought up. Change the MTU in the 1211 * route to match the interface MTU (as long as the 1212 * field isn't locked). 1213 * 1214 * if MTU on the route is 0, we need to fix the MTU. 1215 * this case happens with path MTU discovery timeouts. 1216 */ 1217 mtu = ifmtu; 1218 if (!(ro_pmtu->ro_rt->rt_rmx.rmx_locks & RTV_MTU)) 1219 ro_pmtu->ro_rt->rt_rmx.rmx_mtu = mtu; 1220 } 1221 } else if (ifp) { 1222 mtu = IN6_LINKMTU(ifp); 1223 } else 1224 error = EHOSTUNREACH; /* XXX */ 1225 1226 *mtup = mtu; 1227 return (error); 1228 } 1229 1230 /* 1231 * IP6 socket option processing. 1232 */ 1233 int 1234 ip6_ctloutput(op, so, level, optname, mp) 1235 int op; 1236 struct socket *so; 1237 int level, optname; 1238 struct mbuf **mp; 1239 { 1240 struct in6pcb *in6p = sotoin6pcb(so); 1241 struct mbuf *m = *mp; 1242 int optval = 0; 1243 int error = 0; 1244 struct proc *p = curproc; /* XXX */ 1245 1246 if (level == IPPROTO_IPV6) { 1247 switch (op) { 1248 case PRCO_SETOPT: 1249 switch (optname) { 1250 case IPV6_PKTOPTIONS: 1251 /* m is freed in ip6_pcbopts */ 1252 return (ip6_pcbopts(&in6p->in6p_outputopts, 1253 m, so)); 1254 case IPV6_HOPOPTS: 1255 case IPV6_DSTOPTS: 1256 if (p == 0 || suser(p->p_ucred, &p->p_acflag)) { 1257 error = EPERM; 1258 break; 1259 } 1260 /* FALLTHROUGH */ 1261 case IPV6_UNICAST_HOPS: 1262 case IPV6_RECVOPTS: 1263 case IPV6_RECVRETOPTS: 1264 case IPV6_RECVDSTADDR: 1265 case IPV6_PKTINFO: 1266 case IPV6_HOPLIMIT: 1267 case IPV6_RTHDR: 1268 case IPV6_FAITH: 1269 case IPV6_V6ONLY: 1270 if (!m || m->m_len != sizeof(int)) { 1271 error = EINVAL; 1272 break; 1273 } 1274 optval = *mtod(m, int *); 1275 switch (optname) { 1276 1277 case IPV6_UNICAST_HOPS: 1278 if (optval < -1 || optval >= 256) 1279 error = EINVAL; 1280 else { 1281 /* -1 = kernel default */ 1282 in6p->in6p_hops = optval; 1283 } 1284 break; 1285 #define OPTSET(bit) \ 1286 do { \ 1287 if (optval) \ 1288 in6p->in6p_flags |= (bit); \ 1289 else \ 1290 in6p->in6p_flags &= ~(bit); \ 1291 } while (/*CONSTCOND*/ 0) 1292 1293 case IPV6_RECVOPTS: 1294 OPTSET(IN6P_RECVOPTS); 1295 break; 1296 1297 case IPV6_RECVRETOPTS: 1298 OPTSET(IN6P_RECVRETOPTS); 1299 break; 1300 1301 case IPV6_RECVDSTADDR: 1302 OPTSET(IN6P_RECVDSTADDR); 1303 break; 1304 1305 case IPV6_PKTINFO: 1306 OPTSET(IN6P_PKTINFO); 1307 break; 1308 1309 case IPV6_HOPLIMIT: 1310 OPTSET(IN6P_HOPLIMIT); 1311 break; 1312 1313 case IPV6_HOPOPTS: 1314 OPTSET(IN6P_HOPOPTS); 1315 break; 1316 1317 case IPV6_DSTOPTS: 1318 OPTSET(IN6P_DSTOPTS); 1319 break; 1320 1321 case IPV6_RTHDR: 1322 OPTSET(IN6P_RTHDR); 1323 break; 1324 1325 case IPV6_FAITH: 1326 OPTSET(IN6P_FAITH); 1327 break; 1328 1329 case IPV6_V6ONLY: 1330 /* 1331 * make setsockopt(IPV6_V6ONLY) 1332 * available only prior to bind(2). 1333 * see ipng mailing list, Jun 22 2001. 1334 */ 1335 if (in6p->in6p_lport || 1336 !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr)) { 1337 error = EINVAL; 1338 break; 1339 } 1340 #ifdef INET6_BINDV6ONLY 1341 if (!optval) 1342 error = EINVAL; 1343 #else 1344 OPTSET(IN6P_IPV6_V6ONLY); 1345 #endif 1346 break; 1347 } 1348 break; 1349 #undef OPTSET 1350 1351 case IPV6_MULTICAST_IF: 1352 case IPV6_MULTICAST_HOPS: 1353 case IPV6_MULTICAST_LOOP: 1354 case IPV6_JOIN_GROUP: 1355 case IPV6_LEAVE_GROUP: 1356 error = ip6_setmoptions(optname, 1357 &in6p->in6p_moptions, m); 1358 break; 1359 1360 case IPV6_PORTRANGE: 1361 optval = *mtod(m, int *); 1362 1363 switch (optval) { 1364 case IPV6_PORTRANGE_DEFAULT: 1365 in6p->in6p_flags &= ~(IN6P_LOWPORT); 1366 in6p->in6p_flags &= ~(IN6P_HIGHPORT); 1367 break; 1368 1369 case IPV6_PORTRANGE_HIGH: 1370 in6p->in6p_flags &= ~(IN6P_LOWPORT); 1371 in6p->in6p_flags |= IN6P_HIGHPORT; 1372 break; 1373 1374 case IPV6_PORTRANGE_LOW: 1375 in6p->in6p_flags &= ~(IN6P_HIGHPORT); 1376 in6p->in6p_flags |= IN6P_LOWPORT; 1377 break; 1378 1379 default: 1380 error = EINVAL; 1381 break; 1382 } 1383 break; 1384 1385 #ifdef IPSEC 1386 case IPV6_IPSEC_POLICY: 1387 { 1388 caddr_t req = NULL; 1389 size_t len = 0; 1390 1391 int priv = 0; 1392 if (p == 0 || suser(p->p_ucred, &p->p_acflag)) 1393 priv = 0; 1394 else 1395 priv = 1; 1396 if (m) { 1397 req = mtod(m, caddr_t); 1398 len = m->m_len; 1399 } 1400 error = ipsec6_set_policy(in6p, 1401 optname, req, len, priv); 1402 } 1403 break; 1404 #endif /* IPSEC */ 1405 1406 default: 1407 error = ENOPROTOOPT; 1408 break; 1409 } 1410 if (m) 1411 (void)m_free(m); 1412 break; 1413 1414 case PRCO_GETOPT: 1415 switch (optname) { 1416 1417 case IPV6_OPTIONS: 1418 case IPV6_RETOPTS: 1419 error = ENOPROTOOPT; 1420 break; 1421 1422 case IPV6_PKTOPTIONS: 1423 if (in6p->in6p_options) { 1424 *mp = m_copym(in6p->in6p_options, 0, 1425 M_COPYALL, M_WAIT); 1426 } else { 1427 *mp = m_get(M_WAIT, MT_SOOPTS); 1428 (*mp)->m_len = 0; 1429 } 1430 break; 1431 1432 case IPV6_HOPOPTS: 1433 case IPV6_DSTOPTS: 1434 if (p == 0 || suser(p->p_ucred, &p->p_acflag)) { 1435 error = EPERM; 1436 break; 1437 } 1438 /* FALLTHROUGH */ 1439 case IPV6_UNICAST_HOPS: 1440 case IPV6_RECVOPTS: 1441 case IPV6_RECVRETOPTS: 1442 case IPV6_RECVDSTADDR: 1443 case IPV6_PORTRANGE: 1444 case IPV6_PKTINFO: 1445 case IPV6_HOPLIMIT: 1446 case IPV6_RTHDR: 1447 case IPV6_FAITH: 1448 case IPV6_V6ONLY: 1449 *mp = m = m_get(M_WAIT, MT_SOOPTS); 1450 m->m_len = sizeof(int); 1451 switch (optname) { 1452 1453 case IPV6_UNICAST_HOPS: 1454 optval = in6p->in6p_hops; 1455 break; 1456 1457 #define OPTBIT(bit) (in6p->in6p_flags & bit ? 1 : 0) 1458 1459 case IPV6_RECVOPTS: 1460 optval = OPTBIT(IN6P_RECVOPTS); 1461 break; 1462 1463 case IPV6_RECVRETOPTS: 1464 optval = OPTBIT(IN6P_RECVRETOPTS); 1465 break; 1466 1467 case IPV6_RECVDSTADDR: 1468 optval = OPTBIT(IN6P_RECVDSTADDR); 1469 break; 1470 1471 case IPV6_PORTRANGE: 1472 { 1473 int flags; 1474 flags = in6p->in6p_flags; 1475 if (flags & IN6P_HIGHPORT) 1476 optval = IPV6_PORTRANGE_HIGH; 1477 else if (flags & IN6P_LOWPORT) 1478 optval = IPV6_PORTRANGE_LOW; 1479 else 1480 optval = 0; 1481 break; 1482 } 1483 1484 case IPV6_PKTINFO: 1485 optval = OPTBIT(IN6P_PKTINFO); 1486 break; 1487 1488 case IPV6_HOPLIMIT: 1489 optval = OPTBIT(IN6P_HOPLIMIT); 1490 break; 1491 1492 case IPV6_HOPOPTS: 1493 optval = OPTBIT(IN6P_HOPOPTS); 1494 break; 1495 1496 case IPV6_DSTOPTS: 1497 optval = OPTBIT(IN6P_DSTOPTS); 1498 break; 1499 1500 case IPV6_RTHDR: 1501 optval = OPTBIT(IN6P_RTHDR); 1502 break; 1503 1504 case IPV6_FAITH: 1505 optval = OPTBIT(IN6P_FAITH); 1506 break; 1507 1508 case IPV6_V6ONLY: 1509 optval = OPTBIT(IN6P_IPV6_V6ONLY); 1510 break; 1511 } 1512 *mtod(m, int *) = optval; 1513 break; 1514 1515 case IPV6_MULTICAST_IF: 1516 case IPV6_MULTICAST_HOPS: 1517 case IPV6_MULTICAST_LOOP: 1518 case IPV6_JOIN_GROUP: 1519 case IPV6_LEAVE_GROUP: 1520 error = ip6_getmoptions(optname, in6p->in6p_moptions, mp); 1521 break; 1522 1523 #ifdef IPSEC 1524 case IPV6_IPSEC_POLICY: 1525 { 1526 caddr_t req = NULL; 1527 size_t len = 0; 1528 1529 if (m) { 1530 req = mtod(m, caddr_t); 1531 len = m->m_len; 1532 } 1533 error = ipsec6_get_policy(in6p, req, len, mp); 1534 break; 1535 } 1536 #endif /* IPSEC */ 1537 1538 default: 1539 error = ENOPROTOOPT; 1540 break; 1541 } 1542 break; 1543 } 1544 } else { 1545 error = EINVAL; 1546 if (op == PRCO_SETOPT && *mp) 1547 (void)m_free(*mp); 1548 } 1549 return (error); 1550 } 1551 1552 int 1553 ip6_raw_ctloutput(op, so, level, optname, mp) 1554 int op; 1555 struct socket *so; 1556 int level, optname; 1557 struct mbuf **mp; 1558 { 1559 int error = 0, optval, optlen; 1560 const int icmp6off = offsetof(struct icmp6_hdr, icmp6_cksum); 1561 struct in6pcb *in6p = sotoin6pcb(so); 1562 struct mbuf *m = *mp; 1563 1564 optlen = m ? m->m_len : 0; 1565 1566 if (level != IPPROTO_IPV6) { 1567 if (op == PRCO_SETOPT && *mp) 1568 (void)m_free(*mp); 1569 return (EINVAL); 1570 } 1571 1572 switch (optname) { 1573 case IPV6_CHECKSUM: 1574 /* 1575 * For ICMPv6 sockets, no modification allowed for checksum 1576 * offset, permit "no change" values to help existing apps. 1577 * 1578 * XXX 2292bis says: "An attempt to set IPV6_CHECKSUM 1579 * for an ICMPv6 socket will fail." 1580 * The current behavior does not meet 2292bis. 1581 */ 1582 switch (op) { 1583 case PRCO_SETOPT: 1584 if (optlen != sizeof(int)) { 1585 error = EINVAL; 1586 break; 1587 } 1588 optval = *mtod(m, int *); 1589 if ((optval % 2) != 0) { 1590 /* the API assumes even offset values */ 1591 error = EINVAL; 1592 } else if (so->so_proto->pr_protocol == 1593 IPPROTO_ICMPV6) { 1594 if (optval != icmp6off) 1595 error = EINVAL; 1596 } else 1597 in6p->in6p_cksum = optval; 1598 break; 1599 1600 case PRCO_GETOPT: 1601 if (so->so_proto->pr_protocol == IPPROTO_ICMPV6) 1602 optval = icmp6off; 1603 else 1604 optval = in6p->in6p_cksum; 1605 1606 *mp = m = m_get(M_WAIT, MT_SOOPTS); 1607 m->m_len = sizeof(int); 1608 *mtod(m, int *) = optval; 1609 break; 1610 1611 default: 1612 error = EINVAL; 1613 break; 1614 } 1615 break; 1616 1617 default: 1618 error = ENOPROTOOPT; 1619 break; 1620 } 1621 1622 if (op == PRCO_SETOPT && m) 1623 (void)m_free(m); 1624 1625 return (error); 1626 } 1627 1628 /* 1629 * Set up IP6 options in pcb for insertion in output packets. 1630 * Store in mbuf with pointer in pcbopt, adding pseudo-option 1631 * with destination address if source routed. 1632 */ 1633 static int 1634 ip6_pcbopts(pktopt, m, so) 1635 struct ip6_pktopts **pktopt; 1636 struct mbuf *m; 1637 struct socket *so; 1638 { 1639 struct ip6_pktopts *opt = *pktopt; 1640 int error = 0; 1641 struct proc *p = curproc; /* XXX */ 1642 int priv = 0; 1643 1644 /* turn off any old options. */ 1645 if (opt) { 1646 if (opt->ip6po_m) 1647 (void)m_free(opt->ip6po_m); 1648 } else 1649 opt = malloc(sizeof(*opt), M_IP6OPT, M_WAITOK); 1650 *pktopt = 0; 1651 1652 if (!m || m->m_len == 0) { 1653 /* 1654 * Only turning off any previous options. 1655 */ 1656 free(opt, M_IP6OPT); 1657 if (m) 1658 (void)m_free(m); 1659 return (0); 1660 } 1661 1662 /* set options specified by user. */ 1663 if (p && !suser(p->p_ucred, &p->p_acflag)) 1664 priv = 1; 1665 if ((error = ip6_setpktoptions(m, opt, priv)) != 0) { 1666 (void)m_free(m); 1667 free(opt, M_IP6OPT); 1668 return (error); 1669 } 1670 *pktopt = opt; 1671 return (0); 1672 } 1673 1674 /* 1675 * Set the IP6 multicast options in response to user setsockopt(). 1676 */ 1677 static int 1678 ip6_setmoptions(optname, im6op, m) 1679 int optname; 1680 struct ip6_moptions **im6op; 1681 struct mbuf *m; 1682 { 1683 int error = 0; 1684 u_int loop, ifindex; 1685 struct ipv6_mreq *mreq; 1686 struct ifnet *ifp; 1687 struct ip6_moptions *im6o = *im6op; 1688 struct route_in6 ro; 1689 struct sockaddr_in6 *dst; 1690 struct in6_multi_mship *imm; 1691 struct proc *p = curproc; /* XXX */ 1692 1693 if (im6o == NULL) { 1694 /* 1695 * No multicast option buffer attached to the pcb; 1696 * allocate one and initialize to default values. 1697 */ 1698 im6o = (struct ip6_moptions *) 1699 malloc(sizeof(*im6o), M_IPMOPTS, M_WAITOK); 1700 1701 if (im6o == NULL) 1702 return (ENOBUFS); 1703 *im6op = im6o; 1704 im6o->im6o_multicast_ifp = NULL; 1705 im6o->im6o_multicast_hlim = ip6_defmcasthlim; 1706 im6o->im6o_multicast_loop = IPV6_DEFAULT_MULTICAST_LOOP; 1707 LIST_INIT(&im6o->im6o_memberships); 1708 } 1709 1710 switch (optname) { 1711 1712 case IPV6_MULTICAST_IF: 1713 /* 1714 * Select the interface for outgoing multicast packets. 1715 */ 1716 if (m == NULL || m->m_len != sizeof(u_int)) { 1717 error = EINVAL; 1718 break; 1719 } 1720 bcopy(mtod(m, u_int *), &ifindex, sizeof(ifindex)); 1721 if (ifindex < 0 || if_index < ifindex) { 1722 error = ENXIO; /* XXX EINVAL? */ 1723 break; 1724 } 1725 ifp = ifindex2ifnet[ifindex]; 1726 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) { 1727 error = EADDRNOTAVAIL; 1728 break; 1729 } 1730 im6o->im6o_multicast_ifp = ifp; 1731 break; 1732 1733 case IPV6_MULTICAST_HOPS: 1734 { 1735 /* 1736 * Set the IP6 hoplimit for outgoing multicast packets. 1737 */ 1738 int optval; 1739 if (m == NULL || m->m_len != sizeof(int)) { 1740 error = EINVAL; 1741 break; 1742 } 1743 bcopy(mtod(m, u_int *), &optval, sizeof(optval)); 1744 if (optval < -1 || optval >= 256) 1745 error = EINVAL; 1746 else if (optval == -1) 1747 im6o->im6o_multicast_hlim = ip6_defmcasthlim; 1748 else 1749 im6o->im6o_multicast_hlim = optval; 1750 break; 1751 } 1752 1753 case IPV6_MULTICAST_LOOP: 1754 /* 1755 * Set the loopback flag for outgoing multicast packets. 1756 * Must be zero or one. 1757 */ 1758 if (m == NULL || m->m_len != sizeof(u_int)) { 1759 error = EINVAL; 1760 break; 1761 } 1762 bcopy(mtod(m, u_int *), &loop, sizeof(loop)); 1763 if (loop > 1) { 1764 error = EINVAL; 1765 break; 1766 } 1767 im6o->im6o_multicast_loop = loop; 1768 break; 1769 1770 case IPV6_JOIN_GROUP: 1771 /* 1772 * Add a multicast group membership. 1773 * Group must be a valid IP6 multicast address. 1774 */ 1775 if (m == NULL || m->m_len != sizeof(struct ipv6_mreq)) { 1776 error = EINVAL; 1777 break; 1778 } 1779 mreq = mtod(m, struct ipv6_mreq *); 1780 if (IN6_IS_ADDR_UNSPECIFIED(&mreq->ipv6mr_multiaddr)) { 1781 /* 1782 * We use the unspecified address to specify to accept 1783 * all multicast addresses. Only super user is allowed 1784 * to do this. 1785 */ 1786 if (suser(p->p_ucred, &p->p_acflag)) 1787 { 1788 error = EACCES; 1789 break; 1790 } 1791 } else if (!IN6_IS_ADDR_MULTICAST(&mreq->ipv6mr_multiaddr)) { 1792 error = EINVAL; 1793 break; 1794 } 1795 1796 /* 1797 * If the interface is specified, validate it. 1798 */ 1799 if (mreq->ipv6mr_interface < 0 1800 || if_index < mreq->ipv6mr_interface) { 1801 error = ENXIO; /* XXX EINVAL? */ 1802 break; 1803 } 1804 /* 1805 * If no interface was explicitly specified, choose an 1806 * appropriate one according to the given multicast address. 1807 */ 1808 if (mreq->ipv6mr_interface == 0) { 1809 /* 1810 * If the multicast address is in node-local scope, 1811 * the interface should be a loopback interface. 1812 * Otherwise, look up the routing table for the 1813 * address, and choose the outgoing interface. 1814 * XXX: is it a good approach? 1815 */ 1816 if (IN6_IS_ADDR_MC_NODELOCAL(&mreq->ipv6mr_multiaddr)) { 1817 ifp = &loif[0]; 1818 } else { 1819 ro.ro_rt = NULL; 1820 dst = (struct sockaddr_in6 *)&ro.ro_dst; 1821 bzero(dst, sizeof(*dst)); 1822 dst->sin6_len = sizeof(struct sockaddr_in6); 1823 dst->sin6_family = AF_INET6; 1824 dst->sin6_addr = mreq->ipv6mr_multiaddr; 1825 rtalloc((struct route *)&ro); 1826 if (ro.ro_rt == NULL) { 1827 error = EADDRNOTAVAIL; 1828 break; 1829 } 1830 ifp = ro.ro_rt->rt_ifp; 1831 rtfree(ro.ro_rt); 1832 } 1833 } else 1834 ifp = ifindex2ifnet[mreq->ipv6mr_interface]; 1835 1836 /* 1837 * See if we found an interface, and confirm that it 1838 * supports multicast 1839 */ 1840 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) { 1841 error = EADDRNOTAVAIL; 1842 break; 1843 } 1844 /* 1845 * Put interface index into the multicast address, 1846 * if the address has link-local scope. 1847 */ 1848 if (IN6_IS_ADDR_MC_LINKLOCAL(&mreq->ipv6mr_multiaddr)) { 1849 mreq->ipv6mr_multiaddr.s6_addr16[1] = 1850 htons(mreq->ipv6mr_interface); 1851 } 1852 /* 1853 * See if the membership already exists. 1854 */ 1855 for (imm = im6o->im6o_memberships.lh_first; 1856 imm != NULL; imm = imm->i6mm_chain.le_next) 1857 if (imm->i6mm_maddr->in6m_ifp == ifp && 1858 IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr, 1859 &mreq->ipv6mr_multiaddr)) 1860 break; 1861 if (imm != NULL) { 1862 error = EADDRINUSE; 1863 break; 1864 } 1865 /* 1866 * Everything looks good; add a new record to the multicast 1867 * address list for the given interface. 1868 */ 1869 imm = in6_joingroup(ifp, &mreq->ipv6mr_multiaddr, &error); 1870 if (!imm) 1871 break; 1872 LIST_INSERT_HEAD(&im6o->im6o_memberships, imm, i6mm_chain); 1873 break; 1874 1875 case IPV6_LEAVE_GROUP: 1876 /* 1877 * Drop a multicast group membership. 1878 * Group must be a valid IP6 multicast address. 1879 */ 1880 if (m == NULL || m->m_len != sizeof(struct ipv6_mreq)) { 1881 error = EINVAL; 1882 break; 1883 } 1884 mreq = mtod(m, struct ipv6_mreq *); 1885 if (IN6_IS_ADDR_UNSPECIFIED(&mreq->ipv6mr_multiaddr)) { 1886 if (suser(p->p_ucred, &p->p_acflag)) 1887 { 1888 error = EACCES; 1889 break; 1890 } 1891 } else if (!IN6_IS_ADDR_MULTICAST(&mreq->ipv6mr_multiaddr)) { 1892 error = EINVAL; 1893 break; 1894 } 1895 /* 1896 * If an interface address was specified, get a pointer 1897 * to its ifnet structure. 1898 */ 1899 if (mreq->ipv6mr_interface < 0 1900 || if_index < mreq->ipv6mr_interface) { 1901 error = ENXIO; /* XXX EINVAL? */ 1902 break; 1903 } 1904 ifp = ifindex2ifnet[mreq->ipv6mr_interface]; 1905 /* 1906 * Put interface index into the multicast address, 1907 * if the address has link-local scope. 1908 */ 1909 if (IN6_IS_ADDR_MC_LINKLOCAL(&mreq->ipv6mr_multiaddr)) { 1910 mreq->ipv6mr_multiaddr.s6_addr16[1] = 1911 htons(mreq->ipv6mr_interface); 1912 } 1913 /* 1914 * Find the membership in the membership list. 1915 */ 1916 for (imm = im6o->im6o_memberships.lh_first; 1917 imm != NULL; imm = imm->i6mm_chain.le_next) { 1918 if ((ifp == NULL || imm->i6mm_maddr->in6m_ifp == ifp) && 1919 IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr, 1920 &mreq->ipv6mr_multiaddr)) 1921 break; 1922 } 1923 if (imm == NULL) { 1924 /* Unable to resolve interface */ 1925 error = EADDRNOTAVAIL; 1926 break; 1927 } 1928 /* 1929 * Give up the multicast address record to which the 1930 * membership points. 1931 */ 1932 LIST_REMOVE(imm, i6mm_chain); 1933 in6_leavegroup(imm); 1934 break; 1935 1936 default: 1937 error = EOPNOTSUPP; 1938 break; 1939 } 1940 1941 /* 1942 * If all options have default values, no need to keep the mbuf. 1943 */ 1944 if (im6o->im6o_multicast_ifp == NULL && 1945 im6o->im6o_multicast_hlim == ip6_defmcasthlim && 1946 im6o->im6o_multicast_loop == IPV6_DEFAULT_MULTICAST_LOOP && 1947 im6o->im6o_memberships.lh_first == NULL) { 1948 free(*im6op, M_IPMOPTS); 1949 *im6op = NULL; 1950 } 1951 1952 return (error); 1953 } 1954 1955 /* 1956 * Return the IP6 multicast options in response to user getsockopt(). 1957 */ 1958 static int 1959 ip6_getmoptions(optname, im6o, mp) 1960 int optname; 1961 struct ip6_moptions *im6o; 1962 struct mbuf **mp; 1963 { 1964 u_int *hlim, *loop, *ifindex; 1965 1966 *mp = m_get(M_WAIT, MT_SOOPTS); 1967 1968 switch (optname) { 1969 1970 case IPV6_MULTICAST_IF: 1971 ifindex = mtod(*mp, u_int *); 1972 (*mp)->m_len = sizeof(u_int); 1973 if (im6o == NULL || im6o->im6o_multicast_ifp == NULL) 1974 *ifindex = 0; 1975 else 1976 *ifindex = im6o->im6o_multicast_ifp->if_index; 1977 return (0); 1978 1979 case IPV6_MULTICAST_HOPS: 1980 hlim = mtod(*mp, u_int *); 1981 (*mp)->m_len = sizeof(u_int); 1982 if (im6o == NULL) 1983 *hlim = ip6_defmcasthlim; 1984 else 1985 *hlim = im6o->im6o_multicast_hlim; 1986 return (0); 1987 1988 case IPV6_MULTICAST_LOOP: 1989 loop = mtod(*mp, u_int *); 1990 (*mp)->m_len = sizeof(u_int); 1991 if (im6o == NULL) 1992 *loop = ip6_defmcasthlim; 1993 else 1994 *loop = im6o->im6o_multicast_loop; 1995 return (0); 1996 1997 default: 1998 return (EOPNOTSUPP); 1999 } 2000 } 2001 2002 /* 2003 * Discard the IP6 multicast options. 2004 */ 2005 void 2006 ip6_freemoptions(im6o) 2007 struct ip6_moptions *im6o; 2008 { 2009 struct in6_multi_mship *imm; 2010 2011 if (im6o == NULL) 2012 return; 2013 2014 while ((imm = im6o->im6o_memberships.lh_first) != NULL) { 2015 LIST_REMOVE(imm, i6mm_chain); 2016 in6_leavegroup(imm); 2017 } 2018 free(im6o, M_IPMOPTS); 2019 } 2020 2021 /* 2022 * Set IPv6 outgoing packet options based on advanced API. 2023 */ 2024 int 2025 ip6_setpktoptions(control, opt, priv) 2026 struct mbuf *control; 2027 struct ip6_pktopts *opt; 2028 int priv; 2029 { 2030 struct cmsghdr *cm = 0; 2031 2032 if (control == 0 || opt == 0) 2033 return (EINVAL); 2034 2035 bzero(opt, sizeof(*opt)); 2036 opt->ip6po_hlim = -1; /* -1 means to use default hop limit */ 2037 2038 /* 2039 * XXX: Currently, we assume all the optional information is stored 2040 * in a single mbuf. 2041 */ 2042 if (control->m_next) 2043 return (EINVAL); 2044 2045 opt->ip6po_m = control; 2046 2047 for (; control->m_len; control->m_data += CMSG_ALIGN(cm->cmsg_len), 2048 control->m_len -= CMSG_ALIGN(cm->cmsg_len)) { 2049 cm = mtod(control, struct cmsghdr *); 2050 if (cm->cmsg_len == 0 || cm->cmsg_len > control->m_len) 2051 return (EINVAL); 2052 if (cm->cmsg_level != IPPROTO_IPV6) 2053 continue; 2054 2055 switch (cm->cmsg_type) { 2056 case IPV6_PKTINFO: 2057 if (cm->cmsg_len != CMSG_LEN(sizeof(struct in6_pktinfo))) 2058 return (EINVAL); 2059 opt->ip6po_pktinfo = (struct in6_pktinfo *)CMSG_DATA(cm); 2060 if (opt->ip6po_pktinfo->ipi6_ifindex && 2061 IN6_IS_ADDR_LINKLOCAL(&opt->ip6po_pktinfo->ipi6_addr)) 2062 opt->ip6po_pktinfo->ipi6_addr.s6_addr16[1] = 2063 htons(opt->ip6po_pktinfo->ipi6_ifindex); 2064 2065 if (opt->ip6po_pktinfo->ipi6_ifindex > if_index || 2066 opt->ip6po_pktinfo->ipi6_ifindex < 0) { 2067 return (ENXIO); 2068 } 2069 2070 /* 2071 * Check if the requested source address is indeed a 2072 * unicast address assigned to the node, and can be 2073 * used as the packet's source address. 2074 */ 2075 if (!IN6_IS_ADDR_UNSPECIFIED(&opt->ip6po_pktinfo->ipi6_addr)) { 2076 struct ifaddr *ia; 2077 struct in6_ifaddr *ia6; 2078 struct sockaddr_in6 sin6; 2079 2080 bzero(&sin6, sizeof(sin6)); 2081 sin6.sin6_len = sizeof(sin6); 2082 sin6.sin6_family = AF_INET6; 2083 sin6.sin6_addr = 2084 opt->ip6po_pktinfo->ipi6_addr; 2085 ia = ifa_ifwithaddr(sin6tosa(&sin6)); 2086 if (ia == NULL || 2087 (opt->ip6po_pktinfo->ipi6_ifindex && 2088 (ia->ifa_ifp->if_index != 2089 opt->ip6po_pktinfo->ipi6_ifindex))) { 2090 return (EADDRNOTAVAIL); 2091 } 2092 ia6 = (struct in6_ifaddr *)ia; 2093 if ((ia6->ia6_flags & (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY)) != 0) { 2094 return (EADDRNOTAVAIL); 2095 } 2096 2097 /* 2098 * Check if the requested source address is 2099 * indeed a unicast address assigned to the 2100 * node. 2101 */ 2102 if (IN6_IS_ADDR_MULTICAST(&opt->ip6po_pktinfo->ipi6_addr)) 2103 return (EADDRNOTAVAIL); 2104 } 2105 break; 2106 2107 case IPV6_HOPLIMIT: 2108 if (cm->cmsg_len != CMSG_LEN(sizeof(int))) 2109 return (EINVAL); 2110 2111 bcopy(CMSG_DATA(cm), &opt->ip6po_hlim, 2112 sizeof(opt->ip6po_hlim)); 2113 if (opt->ip6po_hlim < -1 || opt->ip6po_hlim > 255) 2114 return (EINVAL); 2115 break; 2116 2117 case IPV6_NEXTHOP: 2118 if (!priv) 2119 return (EPERM); 2120 2121 /* check if cmsg_len is large enough for sa_len */ 2122 if (cm->cmsg_len < sizeof(u_char) || 2123 cm->cmsg_len < CMSG_LEN(*CMSG_DATA(cm))) 2124 return (EINVAL); 2125 2126 opt->ip6po_nexthop = (struct sockaddr *)CMSG_DATA(cm); 2127 2128 break; 2129 2130 case IPV6_HOPOPTS: 2131 if (cm->cmsg_len < CMSG_LEN(sizeof(struct ip6_hbh))) 2132 return (EINVAL); 2133 opt->ip6po_hbh = (struct ip6_hbh *)CMSG_DATA(cm); 2134 if (cm->cmsg_len != 2135 CMSG_LEN((opt->ip6po_hbh->ip6h_len + 1) << 3)) 2136 return (EINVAL); 2137 break; 2138 2139 case IPV6_DSTOPTS: 2140 if (cm->cmsg_len < CMSG_LEN(sizeof(struct ip6_dest))) 2141 return (EINVAL); 2142 2143 /* 2144 * If there is no routing header yet, the destination 2145 * options header should be put on the 1st part. 2146 * Otherwise, the header should be on the 2nd part. 2147 * (See RFC 2460, section 4.1) 2148 */ 2149 if (opt->ip6po_rthdr == NULL) { 2150 opt->ip6po_dest1 = 2151 (struct ip6_dest *)CMSG_DATA(cm); 2152 if (cm->cmsg_len != 2153 CMSG_LEN((opt->ip6po_dest1->ip6d_len + 1) << 3)); 2154 return (EINVAL); 2155 } 2156 else { 2157 opt->ip6po_dest2 = 2158 (struct ip6_dest *)CMSG_DATA(cm); 2159 if (cm->cmsg_len != 2160 CMSG_LEN((opt->ip6po_dest2->ip6d_len + 1) << 3)) 2161 return (EINVAL); 2162 } 2163 break; 2164 2165 case IPV6_RTHDR: 2166 if (cm->cmsg_len < CMSG_LEN(sizeof(struct ip6_rthdr))) 2167 return (EINVAL); 2168 opt->ip6po_rthdr = (struct ip6_rthdr *)CMSG_DATA(cm); 2169 if (cm->cmsg_len != 2170 CMSG_LEN((opt->ip6po_rthdr->ip6r_len + 1) << 3)) 2171 return (EINVAL); 2172 switch (opt->ip6po_rthdr->ip6r_type) { 2173 case IPV6_RTHDR_TYPE_0: 2174 if (opt->ip6po_rthdr->ip6r_segleft == 0) 2175 return (EINVAL); 2176 break; 2177 default: 2178 return (EINVAL); 2179 } 2180 break; 2181 2182 default: 2183 return (ENOPROTOOPT); 2184 } 2185 } 2186 2187 return (0); 2188 } 2189 2190 /* 2191 * Routine called from ip6_output() to loop back a copy of an IP6 multicast 2192 * packet to the input queue of a specified interface. Note that this 2193 * calls the output routine of the loopback "driver", but with an interface 2194 * pointer that might NOT be &loif -- easier than replicating that code here. 2195 */ 2196 void 2197 ip6_mloopback(ifp, m, dst) 2198 struct ifnet *ifp; 2199 struct mbuf *m; 2200 struct sockaddr_in6 *dst; 2201 { 2202 struct mbuf *copym; 2203 struct ip6_hdr *ip6; 2204 2205 copym = m_copy(m, 0, M_COPYALL); 2206 if (copym == NULL) 2207 return; 2208 2209 /* 2210 * Make sure to deep-copy IPv6 header portion in case the data 2211 * is in an mbuf cluster, so that we can safely override the IPv6 2212 * header portion later. 2213 */ 2214 if ((copym->m_flags & M_EXT) != 0 || 2215 copym->m_len < sizeof(struct ip6_hdr)) { 2216 copym = m_pullup(copym, sizeof(struct ip6_hdr)); 2217 if (copym == NULL) 2218 return; 2219 } 2220 2221 #ifdef DIAGNOSTIC 2222 if (copym->m_len < sizeof(*ip6)) { 2223 m_freem(copym); 2224 return; 2225 } 2226 #endif 2227 2228 ip6 = mtod(copym, struct ip6_hdr *); 2229 if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src)) 2230 ip6->ip6_src.s6_addr16[1] = 0; 2231 if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst)) 2232 ip6->ip6_dst.s6_addr16[1] = 0; 2233 2234 (void)looutput(ifp, copym, (struct sockaddr *)dst, NULL); 2235 } 2236 2237 /* 2238 * Chop IPv6 header off from the payload. 2239 */ 2240 static int 2241 ip6_splithdr(m, exthdrs) 2242 struct mbuf *m; 2243 struct ip6_exthdrs *exthdrs; 2244 { 2245 struct mbuf *mh; 2246 struct ip6_hdr *ip6; 2247 2248 ip6 = mtod(m, struct ip6_hdr *); 2249 if (m->m_len > sizeof(*ip6)) { 2250 MGETHDR(mh, M_DONTWAIT, MT_HEADER); 2251 if (mh == 0) { 2252 m_freem(m); 2253 return ENOBUFS; 2254 } 2255 M_COPY_PKTHDR(mh, m); 2256 MH_ALIGN(mh, sizeof(*ip6)); 2257 m->m_flags &= ~M_PKTHDR; 2258 m->m_len -= sizeof(*ip6); 2259 m->m_data += sizeof(*ip6); 2260 mh->m_next = m; 2261 m = mh; 2262 m->m_len = sizeof(*ip6); 2263 bcopy((caddr_t)ip6, mtod(m, caddr_t), sizeof(*ip6)); 2264 } 2265 exthdrs->ip6e_ip6 = m; 2266 return 0; 2267 } 2268 2269 /* 2270 * Compute IPv6 extension header length. 2271 */ 2272 int 2273 ip6_optlen(in6p) 2274 struct in6pcb *in6p; 2275 { 2276 int len; 2277 2278 if (!in6p->in6p_outputopts) 2279 return 0; 2280 2281 len = 0; 2282 #define elen(x) \ 2283 (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0) 2284 2285 len += elen(in6p->in6p_outputopts->ip6po_hbh); 2286 len += elen(in6p->in6p_outputopts->ip6po_dest1); 2287 len += elen(in6p->in6p_outputopts->ip6po_rthdr); 2288 len += elen(in6p->in6p_outputopts->ip6po_dest2); 2289 return len; 2290 #undef elen 2291 } 2292
Indexes created Tue Sep 30 20:09:53 GMT 2025