1 /* $FreeBSD: src/sys/netinet6/ip6_output.c,v 1.13.2.18 2003/01/24 05:11:35 sam Exp $ */ 2 /* $KAME: ip6_output.c,v 1.279 2002/01/26 06:12:30 jinmei 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 "opt_ip6fw.h" 65 #include "opt_inet.h" 66 #include "opt_inet6.h" 67 68 #include <sys/param.h> 69 #include <sys/malloc.h> 70 #include <sys/mbuf.h> 71 #include <sys/errno.h> 72 #include <sys/protosw.h> 73 #include <sys/socket.h> 74 #include <sys/socketvar.h> 75 #include <sys/systm.h> 76 #include <sys/kernel.h> 77 #include <sys/proc.h> 78 #include <sys/priv.h> 79 80 #include <sys/msgport2.h> 81 82 #include <net/if.h> 83 #include <net/route.h> 84 #include <net/pfil.h> 85 86 #include <netinet/in.h> 87 #include <netinet/in_var.h> 88 #include <netinet6/in6_var.h> 89 #include <netinet/ip6.h> 90 #include <netinet/icmp6.h> 91 #include <netinet6/ip6_var.h> 92 #include <netinet/in_pcb.h> 93 #include <netinet6/nd6.h> 94 #include <netinet6/ip6protosw.h> 95 96 #include <net/ip6fw/ip6_fw.h> 97 98 #include <net/net_osdep.h> 99 100 static MALLOC_DEFINE(M_IPMOPTS, "ip6_moptions", "internet multicast options"); 101 102 struct ip6_exthdrs { 103 struct mbuf *ip6e_ip6; 104 struct mbuf *ip6e_hbh; 105 struct mbuf *ip6e_dest1; 106 struct mbuf *ip6e_rthdr; 107 struct mbuf *ip6e_dest2; 108 }; 109 110 static int ip6_pcbopt (int, u_char *, int, struct ip6_pktopts **, int); 111 static int ip6_setpktoption (int, u_char *, int, struct ip6_pktopts *, 112 int, int, int, int); 113 static int ip6_pcbopts(struct ip6_pktopts **, struct mbuf *, struct socket *, 114 struct sockopt *); 115 static int ip6_getpcbopt(struct ip6_pktopts *, int, struct sockopt *); 116 static int ip6_setmoptions (int, struct ip6_moptions **, struct mbuf *); 117 static int ip6_getmoptions (int, struct ip6_moptions *, struct mbuf **); 118 static int ip6_getpmtu(struct route_in6 *, struct route_in6 *, 119 struct ifnet *, struct in6_addr *, u_long *, int *); 120 static int copyexthdr (void *, struct mbuf **); 121 static int ip6_insertfraghdr(struct mbuf *, struct mbuf *, int, 122 struct ip6_frag **); 123 static int ip6_insert_jumboopt (struct ip6_exthdrs *, u_int32_t); 124 static struct mbuf *ip6_splithdr (struct mbuf *); 125 static int copypktopts(struct ip6_pktopts *, struct ip6_pktopts *, 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(struct mbuf *m0, struct ip6_pktopts *opt, struct route_in6 *ro, 140 int flags, struct ip6_moptions *im6o, 141 struct ifnet **ifpp, /* XXX: just for statistics */ 142 struct inpcb *inp) 143 { 144 struct ip6_hdr *ip6, *mhip6; 145 struct ifnet *ifp, *origifp; 146 struct mbuf *m = m0; 147 struct mbuf *mprev; 148 u_char *nexthdrp; 149 int hlen, tlen, len, off; 150 struct route_in6 ip6route; 151 struct sockaddr_in6 *dst; 152 int error = 0; 153 struct in6_ifaddr *ia = NULL; 154 u_long mtu; 155 int alwaysfrag, dontfrag; 156 u_int32_t optlen, plen = 0, unfragpartlen; 157 struct ip6_exthdrs exthdrs; 158 struct in6_addr finaldst; 159 struct route_in6 *ro_pmtu = NULL; 160 boolean_t hdrsplit = FALSE; 161 162 bzero(&exthdrs, sizeof exthdrs); 163 164 if (opt) { 165 if ((error = copyexthdr(opt->ip6po_hbh, &exthdrs.ip6e_hbh))) 166 goto freehdrs; 167 if ((error = copyexthdr(opt->ip6po_dest1, &exthdrs.ip6e_dest1))) 168 goto freehdrs; 169 if ((error = copyexthdr(opt->ip6po_rthdr, &exthdrs.ip6e_rthdr))) 170 goto freehdrs; 171 if ((error = copyexthdr(opt->ip6po_dest2, &exthdrs.ip6e_dest2))) 172 goto freehdrs; 173 } 174 175 /* 176 * Calculate the total length of the extension header chain. 177 * Keep the length of the unfragmentable part for fragmentation. 178 */ 179 optlen = m_lengthm(exthdrs.ip6e_hbh, NULL) + 180 m_lengthm(exthdrs.ip6e_dest1, NULL) + 181 m_lengthm(exthdrs.ip6e_rthdr, NULL); 182 183 unfragpartlen = optlen + sizeof(struct ip6_hdr); 184 185 /* NOTE: we don't add AH/ESP length here. do that later. */ 186 optlen += m_lengthm(exthdrs.ip6e_dest2, NULL); 187 188 /* 189 * If there is at least one extension header, 190 * separate IP6 header from the payload. 191 */ 192 if (optlen && !hdrsplit) { 193 exthdrs.ip6e_ip6 = ip6_splithdr(m); 194 if (exthdrs.ip6e_ip6 == NULL) { 195 error = ENOBUFS; 196 goto freehdrs; 197 } 198 m = exthdrs.ip6e_ip6; 199 hdrsplit = TRUE; 200 } 201 202 /* adjust pointer */ 203 ip6 = mtod(m, struct ip6_hdr *); 204 205 /* adjust mbuf packet header length */ 206 m->m_pkthdr.len += optlen; 207 plen = m->m_pkthdr.len - sizeof(*ip6); 208 209 /* If this is a jumbo payload, insert a jumbo payload option. */ 210 if (plen > IPV6_MAXPACKET) { 211 if (!hdrsplit) { 212 exthdrs.ip6e_ip6 = ip6_splithdr(m); 213 if (exthdrs.ip6e_ip6 == NULL) { 214 error = ENOBUFS; 215 goto freehdrs; 216 } 217 m = exthdrs.ip6e_ip6; 218 hdrsplit = TRUE; 219 } 220 /* adjust pointer */ 221 ip6 = mtod(m, struct ip6_hdr *); 222 if ((error = ip6_insert_jumboopt(&exthdrs, plen)) != 0) 223 goto freehdrs; 224 ip6->ip6_plen = 0; 225 } else 226 ip6->ip6_plen = htons(plen); 227 228 /* 229 * Concatenate headers and fill in next header fields. 230 * Here we have, on "m" 231 * IPv6 payload 232 * and we insert headers accordingly. Finally, we should be getting: 233 * IPv6 hbh dest1 rthdr ah* [esp* dest2 payload] 234 * 235 * during the header composing process, "m" points to IPv6 header. 236 * "mprev" points to an extension header prior to esp. 237 */ 238 239 nexthdrp = &ip6->ip6_nxt; 240 mprev = m; 241 242 /* 243 * we treat dest2 specially. the goal here is to make mprev point the 244 * mbuf prior to dest2. 245 * 246 * result: IPv6 dest2 payload 247 * m and mprev will point to IPv6 header. 248 */ 249 if (exthdrs.ip6e_dest2) { 250 if (!hdrsplit) 251 panic("assumption failed: hdr not split"); 252 exthdrs.ip6e_dest2->m_next = m->m_next; 253 m->m_next = exthdrs.ip6e_dest2; 254 *mtod(exthdrs.ip6e_dest2, u_char *) = ip6->ip6_nxt; 255 ip6->ip6_nxt = IPPROTO_DSTOPTS; 256 } 257 258 /* 259 * Place m1 after mprev. 260 */ 261 #define MAKE_CHAIN(m1, mprev, nexthdrp, i)\ 262 do {\ 263 if (m1) {\ 264 if (!hdrsplit)\ 265 panic("assumption failed: hdr not split");\ 266 *mtod(m1, u_char *) = *nexthdrp;\ 267 *nexthdrp = (i);\ 268 nexthdrp = mtod(m1, u_char *);\ 269 m1->m_next = mprev->m_next;\ 270 mprev->m_next = m1;\ 271 mprev = m1;\ 272 }\ 273 } while (0) 274 275 /* 276 * result: IPv6 hbh dest1 rthdr dest2 payload 277 * m will point to IPv6 header. mprev will point to the 278 * extension header prior to dest2 (rthdr in the above case). 279 */ 280 MAKE_CHAIN(exthdrs.ip6e_hbh, mprev, nexthdrp, IPPROTO_HOPOPTS); 281 MAKE_CHAIN(exthdrs.ip6e_dest1, mprev, nexthdrp, IPPROTO_DSTOPTS); 282 MAKE_CHAIN(exthdrs.ip6e_rthdr, mprev, nexthdrp, IPPROTO_ROUTING); 283 284 /* 285 * If there is a routing header, replace the destination address field 286 * with the first hop of the routing header. 287 */ 288 if (exthdrs.ip6e_rthdr) { 289 struct ip6_rthdr *rh; 290 291 finaldst = ip6->ip6_dst; 292 rh = mtod(exthdrs.ip6e_rthdr, struct ip6_rthdr *); 293 switch (rh->ip6r_type) { 294 default: /* is it possible? */ 295 error = EINVAL; 296 goto bad; 297 } 298 } 299 300 /* Source address validation */ 301 if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) && 302 !(flags & IPV6_DADOUTPUT)) { 303 error = EOPNOTSUPP; 304 ip6stat.ip6s_badscope++; 305 goto bad; 306 } 307 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) { 308 error = EOPNOTSUPP; 309 ip6stat.ip6s_badscope++; 310 goto bad; 311 } 312 313 ip6stat.ip6s_localout++; 314 315 /* 316 * Route packet. 317 */ 318 if (ro == NULL) { 319 ro = &ip6route; 320 bzero(ro, sizeof(*ro)); 321 } 322 ro_pmtu = ro; 323 if (opt && opt->ip6po_rthdr) 324 ro = &opt->ip6po_route; 325 dst = (struct sockaddr_in6 *)&ro->ro_dst; 326 327 /* 328 * If there is a cached route, 329 * check that it is to the same destination 330 * and is still up. If not, free it and try again. 331 */ 332 if (ro->ro_rt != NULL && 333 (!(ro->ro_rt->rt_flags & RTF_UP) || dst->sin6_family != AF_INET6 || 334 !IN6_ARE_ADDR_EQUAL(&dst->sin6_addr, &ip6->ip6_dst))) { 335 RTFREE(ro->ro_rt); 336 ro->ro_rt = NULL; 337 } 338 if (ro->ro_rt == NULL) { 339 bzero(dst, sizeof(*dst)); 340 dst->sin6_family = AF_INET6; 341 dst->sin6_len = sizeof(struct sockaddr_in6); 342 dst->sin6_addr = ip6->ip6_dst; 343 } 344 if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) { 345 /* Unicast */ 346 347 #define ifatoia6(ifa) ((struct in6_ifaddr *)(ifa)) 348 #define sin6tosa(sin6) ((struct sockaddr *)(sin6)) 349 /* xxx 350 * interface selection comes here 351 * if an interface is specified from an upper layer, 352 * ifp must point it. 353 */ 354 if (ro->ro_rt == NULL) { 355 /* 356 * non-bsdi always clone routes, if parent is 357 * PRF_CLONING. 358 */ 359 rtalloc((struct route *)ro); 360 } 361 if (ro->ro_rt == NULL) { 362 ip6stat.ip6s_noroute++; 363 error = EHOSTUNREACH; 364 /* XXX in6_ifstat_inc(ifp, ifs6_out_discard); */ 365 goto bad; 366 } 367 ia = ifatoia6(ro->ro_rt->rt_ifa); 368 ifp = ro->ro_rt->rt_ifp; 369 ro->ro_rt->rt_use++; 370 if (ro->ro_rt->rt_flags & RTF_GATEWAY) 371 dst = (struct sockaddr_in6 *)ro->ro_rt->rt_gateway; 372 m->m_flags &= ~(M_BCAST | M_MCAST); /* just in case */ 373 374 in6_ifstat_inc(ifp, ifs6_out_request); 375 376 /* 377 * Check if the outgoing interface conflicts with 378 * the interface specified by ifi6_ifindex (if specified). 379 * Note that loopback interface is always okay. 380 * (this may happen when we are sending a packet to one of 381 * our own addresses.) 382 */ 383 if (opt && opt->ip6po_pktinfo 384 && opt->ip6po_pktinfo->ipi6_ifindex) { 385 if (!(ifp->if_flags & IFF_LOOPBACK) 386 && ifp->if_index != opt->ip6po_pktinfo->ipi6_ifindex) { 387 ip6stat.ip6s_noroute++; 388 in6_ifstat_inc(ifp, ifs6_out_discard); 389 error = EHOSTUNREACH; 390 goto bad; 391 } 392 } 393 394 if (opt && opt->ip6po_hlim != -1) 395 ip6->ip6_hlim = opt->ip6po_hlim & 0xff; 396 } else { 397 /* Multicast */ 398 struct in6_multi *in6m; 399 400 m->m_flags = (m->m_flags & ~M_BCAST) | M_MCAST; 401 402 /* 403 * See if the caller provided any multicast options 404 */ 405 ifp = NULL; 406 if (im6o != NULL) { 407 ip6->ip6_hlim = im6o->im6o_multicast_hlim; 408 if (im6o->im6o_multicast_ifp != NULL) 409 ifp = im6o->im6o_multicast_ifp; 410 } else 411 ip6->ip6_hlim = ip6_defmcasthlim; 412 413 /* 414 * See if the caller provided the outgoing interface 415 * as an ancillary data. 416 * Boundary check for ifindex is assumed to be already done. 417 */ 418 if (opt && opt->ip6po_pktinfo && opt->ip6po_pktinfo->ipi6_ifindex) 419 ifp = ifindex2ifnet[opt->ip6po_pktinfo->ipi6_ifindex]; 420 421 /* 422 * If the destination is a node-local scope multicast, 423 * the packet should be loop-backed only. 424 */ 425 if (IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst)) { 426 /* 427 * If the outgoing interface is already specified, 428 * it should be a loopback interface. 429 */ 430 if (ifp && !(ifp->if_flags & IFF_LOOPBACK)) { 431 ip6stat.ip6s_badscope++; 432 error = ENETUNREACH; /* XXX: better error? */ 433 /* XXX correct ifp? */ 434 in6_ifstat_inc(ifp, ifs6_out_discard); 435 goto bad; 436 } else { 437 ifp = loif; 438 } 439 } 440 441 if (opt && opt->ip6po_hlim != -1) 442 ip6->ip6_hlim = opt->ip6po_hlim & 0xff; 443 444 /* 445 * If caller did not provide an interface lookup a 446 * default in the routing table. This is either a 447 * default for the speicfied group (i.e. a host 448 * route), or a multicast default (a route for the 449 * ``net'' ff00::/8). 450 */ 451 if (ifp == NULL) { 452 if (ro->ro_rt == NULL) { 453 ro->ro_rt = 454 rtpurelookup((struct sockaddr *)&ro->ro_dst); 455 } 456 if (ro->ro_rt == NULL) { 457 ip6stat.ip6s_noroute++; 458 error = EHOSTUNREACH; 459 /* XXX in6_ifstat_inc(ifp, ifs6_out_discard) */ 460 goto bad; 461 } 462 ia = ifatoia6(ro->ro_rt->rt_ifa); 463 ifp = ro->ro_rt->rt_ifp; 464 ro->ro_rt->rt_use++; 465 } 466 467 if (!(flags & IPV6_FORWARDING)) 468 in6_ifstat_inc(ifp, ifs6_out_request); 469 in6_ifstat_inc(ifp, ifs6_out_mcast); 470 471 /* 472 * Confirm that the outgoing interface supports multicast. 473 */ 474 if (!(ifp->if_flags & IFF_MULTICAST)) { 475 ip6stat.ip6s_noroute++; 476 in6_ifstat_inc(ifp, ifs6_out_discard); 477 error = ENETUNREACH; 478 goto bad; 479 } 480 in6m = IN6_LOOKUP_MULTI(&ip6->ip6_dst, ifp); 481 if (in6m != NULL && 482 (im6o == NULL || im6o->im6o_multicast_loop)) { 483 /* 484 * If we belong to the destination multicast group 485 * on the outgoing interface, and the caller did not 486 * forbid loopback, loop back a copy. 487 */ 488 ip6_mloopback(ifp, m, dst); 489 } else { 490 /* 491 * If we are acting as a multicast router, perform 492 * multicast forwarding as if the packet had just 493 * arrived on the interface to which we are about 494 * to send. The multicast forwarding function 495 * recursively calls this function, using the 496 * IPV6_FORWARDING flag to prevent infinite recursion. 497 * 498 * Multicasts that are looped back by ip6_mloopback(), 499 * above, will be forwarded by the ip6_input() routine, 500 * if necessary. 501 */ 502 if (ip6_mrouter && !(flags & IPV6_FORWARDING)) { 503 if (ip6_mforward(ip6, ifp, m) != 0) { 504 m_freem(m); 505 goto done; 506 } 507 } 508 } 509 /* 510 * Multicasts with a hoplimit of zero may be looped back, 511 * above, but must not be transmitted on a network. 512 * Also, multicasts addressed to the loopback interface 513 * are not sent -- the above call to ip6_mloopback() will 514 * loop back a copy if this host actually belongs to the 515 * destination group on the loopback interface. 516 */ 517 if (ip6->ip6_hlim == 0 || (ifp->if_flags & IFF_LOOPBACK)) { 518 m_freem(m); 519 goto done; 520 } 521 } 522 523 /* 524 * Fill the outgoing inteface to tell the upper layer 525 * to increment per-interface statistics. 526 */ 527 if (ifpp) 528 *ifpp = ifp; 529 530 /* Determine path MTU. */ 531 if ((error = ip6_getpmtu(ro_pmtu, ro, ifp, &finaldst, &mtu, 532 &alwaysfrag)) != 0) 533 goto bad; 534 535 /* 536 * The caller of this function may specify to use the minimum MTU 537 * in some cases. 538 * An advanced API option (IPV6_USE_MIN_MTU) can also override MTU 539 * setting. The logic is a bit complicated; by default, unicast 540 * packets will follow path MTU while multicast packets will be sent at 541 * the minimum MTU. If IP6PO_MINMTU_ALL is specified, all packets 542 * including unicast ones will be sent at the minimum MTU. Multicast 543 * packets will always be sent at the minimum MTU unless 544 * IP6PO_MINMTU_DISABLE is explicitly specified. 545 * See RFC 3542 for more details. 546 */ 547 if (mtu > IPV6_MMTU) { 548 if ((flags & IPV6_MINMTU)) 549 mtu = IPV6_MMTU; 550 else if (opt && opt->ip6po_minmtu == IP6PO_MINMTU_ALL) 551 mtu = IPV6_MMTU; 552 else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) && 553 (opt == NULL || 554 opt->ip6po_minmtu != IP6PO_MINMTU_DISABLE)) { 555 mtu = IPV6_MMTU; 556 } 557 } 558 559 /* Fake scoped addresses */ 560 if ((ifp->if_flags & IFF_LOOPBACK) != 0) { 561 /* 562 * If source or destination address is a scoped address, and 563 * the packet is going to be sent to a loopback interface, 564 * we should keep the original interface. 565 */ 566 567 /* 568 * XXX: this is a very experimental and temporary solution. 569 * We eventually have sockaddr_in6 and use the sin6_scope_id 570 * field of the structure here. 571 * We rely on the consistency between two scope zone ids 572 * of source and destination, which should already be assured. 573 * Larger scopes than link will be supported in the future. 574 */ 575 origifp = NULL; 576 if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src)) 577 origifp = ifindex2ifnet[ntohs(ip6->ip6_src.s6_addr16[1])]; 578 else if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst)) 579 origifp = ifindex2ifnet[ntohs(ip6->ip6_dst.s6_addr16[1])]; 580 /* 581 * XXX: origifp can be NULL even in those two cases above. 582 * For example, if we remove the (only) link-local address 583 * from the loopback interface, and try to send a link-local 584 * address without link-id information. Then the source 585 * address is ::1, and the destination address is the 586 * link-local address with its s6_addr16[1] being zero. 587 * What is worse, if the packet goes to the loopback interface 588 * by a default rejected route, the null pointer would be 589 * passed to looutput, and the kernel would hang. 590 * The following last resort would prevent such disaster. 591 */ 592 if (origifp == NULL) 593 origifp = ifp; 594 } 595 else 596 origifp = ifp; 597 /* 598 * clear embedded scope identifiers if necessary. 599 * in6_clearscope will touch the addresses only when necessary. 600 */ 601 in6_clearscope(&ip6->ip6_src); 602 in6_clearscope(&ip6->ip6_dst); 603 604 /* 605 * Check with the firewall... 606 */ 607 if (ip6_fw_enable && ip6_fw_chk_ptr) { 608 u_short port = 0; 609 610 m->m_pkthdr.rcvif = NULL; /* XXX */ 611 /* If ipfw says divert, we have to just drop packet */ 612 if ((*ip6_fw_chk_ptr)(&ip6, ifp, &port, &m)) { 613 m_freem(m); 614 goto done; 615 } 616 if (!m) { 617 error = EACCES; 618 goto done; 619 } 620 } 621 622 /* 623 * If the outgoing packet contains a hop-by-hop options header, 624 * it must be examined and processed even by the source node. 625 * (RFC 2460, section 4.) 626 */ 627 if (exthdrs.ip6e_hbh) { 628 struct ip6_hbh *hbh = mtod(exthdrs.ip6e_hbh, struct ip6_hbh *); 629 u_int32_t dummy1; /* XXX unused */ 630 u_int32_t dummy2; /* XXX unused */ 631 632 #ifdef DIAGNOSTIC 633 if ((hbh->ip6h_len + 1) << 3 > exthdrs.ip6e_hbh->m_len) 634 panic("ip6e_hbh is not continuous"); 635 #endif 636 /* 637 * XXX: if we have to send an ICMPv6 error to the sender, 638 * we need the M_LOOP flag since icmp6_error() expects 639 * the IPv6 and the hop-by-hop options header are 640 * continuous unless the flag is set. 641 */ 642 m->m_flags |= M_LOOP; 643 m->m_pkthdr.rcvif = ifp; 644 if (ip6_process_hopopts(m, (u_int8_t *)(hbh + 1), 645 ((hbh->ip6h_len + 1) << 3) - sizeof(struct ip6_hbh), 646 &dummy1, &dummy2) < 0) { 647 /* m was already freed at this point */ 648 error = EINVAL;/* better error? */ 649 goto done; 650 } 651 m->m_flags &= ~M_LOOP; /* XXX */ 652 m->m_pkthdr.rcvif = NULL; 653 } 654 655 /* 656 * Run through list of hooks for output packets. 657 */ 658 if (pfil_has_hooks(&inet6_pfil_hook)) { 659 error = pfil_run_hooks(&inet6_pfil_hook, &m, ifp, PFIL_OUT); 660 if (error != 0 || m == NULL) 661 goto done; 662 ip6 = mtod(m, struct ip6_hdr *); 663 } 664 665 /* 666 * Send the packet to the outgoing interface. 667 * If necessary, do IPv6 fragmentation before sending. 668 * 669 * the logic here is rather complex: 670 * 1: normal case (dontfrag == 0, alwaysfrag == 0) 671 * 1-a: send as is if tlen <= path mtu 672 * 1-b: fragment if tlen > path mtu 673 * 674 * 2: if user asks us not to fragment (dontfrag == 1) 675 * 2-a: send as is if tlen <= interface mtu 676 * 2-b: error if tlen > interface mtu 677 * 678 * 3: if we always need to attach fragment header (alwaysfrag == 1) 679 * always fragment 680 * 681 * 4: if dontfrag == 1 && alwaysfrag == 1 682 * error, as we cannot handle this conflicting request 683 */ 684 tlen = m->m_pkthdr.len; 685 686 if (opt && (opt->ip6po_flags & IP6PO_DONTFRAG)) 687 dontfrag = 1; 688 else 689 dontfrag = 0; 690 if (dontfrag && alwaysfrag) { /* case 4 */ 691 /* conflicting request - can't transmit */ 692 error = EMSGSIZE; 693 goto bad; 694 } 695 if (dontfrag && tlen > IN6_LINKMTU(ifp)) { /* case 2-b */ 696 /* 697 * Even if the DONTFRAG option is specified, we cannot send the 698 * packet when the data length is larger than the MTU of the 699 * outgoing interface. 700 * Notify the error by sending IPV6_PATHMTU ancillary data as 701 * well as returning an error code (the latter is not described 702 * in the API spec.) 703 */ 704 u_int32_t mtu32; 705 struct ip6ctlparam ip6cp; 706 707 mtu32 = (u_int32_t)mtu; 708 bzero(&ip6cp, sizeof(ip6cp)); 709 ip6cp.ip6c_cmdarg = (void *)&mtu32; 710 kpfctlinput2(PRC_MSGSIZE, (struct sockaddr *)&ro_pmtu->ro_dst, 711 (void *)&ip6cp); 712 713 error = EMSGSIZE; 714 goto bad; 715 } 716 717 /* 718 * transmit packet without fragmentation 719 */ 720 if (dontfrag || (!alwaysfrag && tlen <= mtu)) { /* case 1-a and 2-a */ 721 struct in6_ifaddr *ia6; 722 723 ip6 = mtod(m, struct ip6_hdr *); 724 ia6 = in6_ifawithifp(ifp, &ip6->ip6_src); 725 if (ia6) { 726 /* Record statistics for this interface address. */ 727 IFA_STAT_INC(&ia6->ia_ifa, opackets, 1); 728 IFA_STAT_INC(&ia6->ia_ifa, obytes, m->m_pkthdr.len); 729 } 730 error = nd6_output(ifp, origifp, m, dst, ro->ro_rt); 731 goto done; 732 } 733 734 /* 735 * try to fragment the packet. case 1-b and 3 736 */ 737 if (mtu < IPV6_MMTU) { 738 /* 739 * note that path MTU is never less than IPV6_MMTU 740 * (see icmp6_input). 741 */ 742 error = EMSGSIZE; 743 in6_ifstat_inc(ifp, ifs6_out_fragfail); 744 goto bad; 745 } else if (ip6->ip6_plen == 0) { 746 /* jumbo payload cannot be fragmented */ 747 error = EMSGSIZE; 748 in6_ifstat_inc(ifp, ifs6_out_fragfail); 749 goto bad; 750 } else { 751 struct mbuf **mnext, *m_frgpart; 752 struct ip6_frag *ip6f; 753 u_int32_t id = htonl(ip6_id++); 754 u_char nextproto; 755 756 /* 757 * Too large for the destination or interface; 758 * fragment if possible. 759 * Must be able to put at least 8 bytes per fragment. 760 */ 761 hlen = unfragpartlen; 762 if (mtu > IPV6_MAXPACKET) 763 mtu = IPV6_MAXPACKET; 764 765 len = (mtu - hlen - sizeof(struct ip6_frag)) & ~7; 766 if (len < 8) { 767 error = EMSGSIZE; 768 in6_ifstat_inc(ifp, ifs6_out_fragfail); 769 goto bad; 770 } 771 772 mnext = &m->m_nextpkt; 773 774 /* 775 * Change the next header field of the last header in the 776 * unfragmentable part. 777 */ 778 if (exthdrs.ip6e_rthdr) { 779 nextproto = *mtod(exthdrs.ip6e_rthdr, u_char *); 780 *mtod(exthdrs.ip6e_rthdr, u_char *) = IPPROTO_FRAGMENT; 781 } else if (exthdrs.ip6e_dest1) { 782 nextproto = *mtod(exthdrs.ip6e_dest1, u_char *); 783 *mtod(exthdrs.ip6e_dest1, u_char *) = IPPROTO_FRAGMENT; 784 } else if (exthdrs.ip6e_hbh) { 785 nextproto = *mtod(exthdrs.ip6e_hbh, u_char *); 786 *mtod(exthdrs.ip6e_hbh, u_char *) = IPPROTO_FRAGMENT; 787 } else { 788 nextproto = ip6->ip6_nxt; 789 ip6->ip6_nxt = IPPROTO_FRAGMENT; 790 } 791 792 /* 793 * Loop through length of segment after first fragment, 794 * make new header and copy data of each part and link onto 795 * chain. 796 */ 797 m0 = m; 798 for (off = hlen; off < tlen; off += len) { 799 MGETHDR(m, M_NOWAIT, MT_HEADER); 800 if (!m) { 801 error = ENOBUFS; 802 ip6stat.ip6s_odropped++; 803 goto sendorfree; 804 } 805 m->m_pkthdr.rcvif = NULL; 806 m->m_flags = m0->m_flags & M_COPYFLAGS; 807 *mnext = m; 808 mnext = &m->m_nextpkt; 809 m->m_data += max_linkhdr; 810 mhip6 = mtod(m, struct ip6_hdr *); 811 *mhip6 = *ip6; 812 m->m_len = sizeof(*mhip6); 813 error = ip6_insertfraghdr(m0, m, hlen, &ip6f); 814 if (error) { 815 ip6stat.ip6s_odropped++; 816 goto sendorfree; 817 } 818 ip6f->ip6f_offlg = htons((u_short)((off - hlen) & ~7)); 819 if (off + len >= tlen) 820 len = tlen - off; 821 else 822 ip6f->ip6f_offlg |= IP6F_MORE_FRAG; 823 mhip6->ip6_plen = htons((u_short)(len + hlen + 824 sizeof(*ip6f) - sizeof(struct ip6_hdr))); 825 if ((m_frgpart = m_copy(m0, off, len)) == NULL) { 826 error = ENOBUFS; 827 ip6stat.ip6s_odropped++; 828 goto sendorfree; 829 } 830 m_cat(m, m_frgpart); 831 m->m_pkthdr.len = len + hlen + sizeof(*ip6f); 832 m->m_pkthdr.rcvif = NULL; 833 ip6f->ip6f_reserved = 0; 834 ip6f->ip6f_ident = id; 835 ip6f->ip6f_nxt = nextproto; 836 ip6stat.ip6s_ofragments++; 837 in6_ifstat_inc(ifp, ifs6_out_fragcreat); 838 } 839 840 in6_ifstat_inc(ifp, ifs6_out_fragok); 841 } 842 843 /* 844 * Remove leading garbages. 845 */ 846 sendorfree: 847 m = m0->m_nextpkt; 848 m0->m_nextpkt = NULL; 849 m_freem(m0); 850 for (m0 = m; m; m = m0) { 851 m0 = m->m_nextpkt; 852 m->m_nextpkt = NULL; 853 if (error == 0) { 854 /* Record statistics for this interface address. */ 855 if (ia) { 856 IFA_STAT_INC(&ia->ia_ifa, opackets, 1); 857 IFA_STAT_INC(&ia->ia_ifa, obytes, 858 m->m_pkthdr.len); 859 } 860 error = nd6_output(ifp, origifp, m, dst, ro->ro_rt); 861 } else 862 m_freem(m); 863 } 864 865 if (error == 0) 866 ip6stat.ip6s_fragmented++; 867 868 done: 869 if (ro == &ip6route && ro->ro_rt) { /* brace necessary for RTFREE */ 870 RTFREE(ro->ro_rt); 871 } else if (ro_pmtu == &ip6route && ro_pmtu->ro_rt) { 872 RTFREE(ro_pmtu->ro_rt); 873 } 874 875 return (error); 876 877 freehdrs: 878 m_freem(exthdrs.ip6e_hbh); /* m_freem will check if mbuf is 0 */ 879 m_freem(exthdrs.ip6e_dest1); 880 m_freem(exthdrs.ip6e_rthdr); 881 m_freem(exthdrs.ip6e_dest2); 882 /* FALLTHROUGH */ 883 bad: 884 m_freem(m); 885 goto done; 886 } 887 888 static int 889 copyexthdr(void *h, struct mbuf **mp) 890 { 891 struct ip6_ext *hdr = h; 892 int hlen; 893 struct mbuf *m; 894 895 if (hdr == NULL) 896 return 0; 897 898 hlen = (hdr->ip6e_len + 1) * 8; 899 if (hlen > MCLBYTES) 900 return ENOBUFS; /* XXX */ 901 902 m = m_getb(hlen, M_NOWAIT, MT_DATA, 0); 903 if (!m) 904 return ENOBUFS; 905 m->m_len = hlen; 906 907 bcopy(hdr, mtod(m, caddr_t), hlen); 908 909 *mp = m; 910 return 0; 911 } 912 913 /* 914 * Insert jumbo payload option. 915 */ 916 static int 917 ip6_insert_jumboopt(struct ip6_exthdrs *exthdrs, u_int32_t plen) 918 { 919 struct mbuf *mopt; 920 u_char *optbuf; 921 u_int32_t v; 922 923 #define JUMBOOPTLEN 8 /* length of jumbo payload option and padding */ 924 925 /* 926 * If there is no hop-by-hop options header, allocate new one. 927 * If there is one but it doesn't have enough space to store the 928 * jumbo payload option, allocate a cluster to store the whole options. 929 * Otherwise, use it to store the options. 930 */ 931 if (exthdrs->ip6e_hbh == NULL) { 932 MGET(mopt, M_NOWAIT, MT_DATA); 933 if (mopt == NULL) 934 return (ENOBUFS); 935 mopt->m_len = JUMBOOPTLEN; 936 optbuf = mtod(mopt, u_char *); 937 optbuf[1] = 0; /* = ((JUMBOOPTLEN) >> 3) - 1 */ 938 exthdrs->ip6e_hbh = mopt; 939 } else { 940 struct ip6_hbh *hbh; 941 942 mopt = exthdrs->ip6e_hbh; 943 if (M_TRAILINGSPACE(mopt) < JUMBOOPTLEN) { 944 /* 945 * XXX assumption: 946 * - exthdrs->ip6e_hbh is not referenced from places 947 * other than exthdrs. 948 * - exthdrs->ip6e_hbh is not an mbuf chain. 949 */ 950 int oldoptlen = mopt->m_len; 951 struct mbuf *n; 952 953 /* 954 * XXX: give up if the whole (new) hbh header does 955 * not fit even in an mbuf cluster. 956 */ 957 if (oldoptlen + JUMBOOPTLEN > MCLBYTES) 958 return (ENOBUFS); 959 960 /* 961 * As a consequence, we must always prepare a cluster 962 * at this point. 963 */ 964 n = m_getcl(M_NOWAIT, MT_DATA, 0); 965 if (!n) 966 return (ENOBUFS); 967 n->m_len = oldoptlen + JUMBOOPTLEN; 968 bcopy(mtod(mopt, caddr_t), mtod(n, caddr_t), oldoptlen); 969 optbuf = mtod(n, caddr_t) + oldoptlen; 970 m_freem(mopt); 971 mopt = exthdrs->ip6e_hbh = n; 972 } else { 973 optbuf = mtod(mopt, u_char *) + mopt->m_len; 974 mopt->m_len += JUMBOOPTLEN; 975 } 976 optbuf[0] = IP6OPT_PADN; 977 optbuf[1] = 1; 978 979 /* 980 * Adjust the header length according to the pad and 981 * the jumbo payload option. 982 */ 983 hbh = mtod(mopt, struct ip6_hbh *); 984 hbh->ip6h_len += (JUMBOOPTLEN >> 3); 985 } 986 987 /* fill in the option. */ 988 optbuf[2] = IP6OPT_JUMBO; 989 optbuf[3] = 4; 990 v = (u_int32_t)htonl(plen + JUMBOOPTLEN); 991 bcopy(&v, &optbuf[4], sizeof(u_int32_t)); 992 993 /* finally, adjust the packet header length */ 994 exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN; 995 996 return (0); 997 #undef JUMBOOPTLEN 998 } 999 1000 /* 1001 * Insert fragment header and copy unfragmentable header portions. 1002 */ 1003 static int 1004 ip6_insertfraghdr(struct mbuf *m0, struct mbuf *m, int hlen, 1005 struct ip6_frag **frghdrp) 1006 { 1007 struct mbuf *n, *mlast; 1008 1009 if (hlen > sizeof(struct ip6_hdr)) { 1010 n = m_copym(m0, sizeof(struct ip6_hdr), 1011 hlen - sizeof(struct ip6_hdr), M_NOWAIT); 1012 if (n == NULL) 1013 return (ENOBUFS); 1014 m->m_next = n; 1015 } else 1016 n = m; 1017 1018 /* Search for the last mbuf of unfragmentable part. */ 1019 for (mlast = n; mlast->m_next; mlast = mlast->m_next) 1020 ; 1021 1022 if (!(mlast->m_flags & M_EXT) && 1023 M_TRAILINGSPACE(mlast) >= sizeof(struct ip6_frag)) { 1024 /* use the trailing space of the last mbuf for the fragment hdr */ 1025 *frghdrp = (struct ip6_frag *) 1026 (mtod(mlast, caddr_t) + mlast->m_len); 1027 mlast->m_len += sizeof(struct ip6_frag); 1028 m->m_pkthdr.len += sizeof(struct ip6_frag); 1029 } else { 1030 /* allocate a new mbuf for the fragment header */ 1031 struct mbuf *mfrg; 1032 1033 MGET(mfrg, M_NOWAIT, MT_DATA); 1034 if (mfrg == NULL) 1035 return (ENOBUFS); 1036 mfrg->m_len = sizeof(struct ip6_frag); 1037 *frghdrp = mtod(mfrg, struct ip6_frag *); 1038 mlast->m_next = mfrg; 1039 } 1040 1041 return (0); 1042 } 1043 1044 static int 1045 ip6_getpmtu(struct route_in6 *ro_pmtu, struct route_in6 *ro, 1046 struct ifnet *ifp, struct in6_addr *dst, u_long *mtup, 1047 int *alwaysfragp) 1048 { 1049 u_int32_t mtu = 0; 1050 int alwaysfrag = 0; 1051 int error = 0; 1052 1053 if (ro_pmtu != ro) { 1054 /* The first hop and the final destination may differ. */ 1055 struct sockaddr_in6 *sa6_dst = 1056 (struct sockaddr_in6 *)&ro_pmtu->ro_dst; 1057 if (ro_pmtu->ro_rt && 1058 ((ro_pmtu->ro_rt->rt_flags & RTF_UP) == 0 || 1059 !IN6_ARE_ADDR_EQUAL(&sa6_dst->sin6_addr, dst))) { 1060 RTFREE(ro_pmtu->ro_rt); 1061 ro_pmtu->ro_rt = NULL; 1062 } 1063 if (ro_pmtu->ro_rt == NULL) { 1064 bzero(sa6_dst, sizeof(*sa6_dst)); 1065 sa6_dst->sin6_family = AF_INET6; 1066 sa6_dst->sin6_len = sizeof(struct sockaddr_in6); 1067 sa6_dst->sin6_addr = *dst; 1068 1069 rtalloc((struct route *)ro_pmtu); 1070 } 1071 } 1072 if (ro_pmtu->ro_rt) { 1073 u_int32_t ifmtu; 1074 1075 if (ifp == NULL) 1076 ifp = ro_pmtu->ro_rt->rt_ifp; 1077 ifmtu = IN6_LINKMTU(ifp); 1078 mtu = ro_pmtu->ro_rt->rt_rmx.rmx_mtu; 1079 if (mtu == 0) { 1080 mtu = ifmtu; 1081 } else if (mtu < IPV6_MMTU) { 1082 /* 1083 * RFC2460 section 5, last paragraph: 1084 * if we record ICMPv6 too big message with 1085 * mtu < IPV6_MMTU, transmit packets sized IPV6_MMTU 1086 * or smaller, with framgent header attached. 1087 * (fragment header is needed regardless from the 1088 * packet size, for translators to identify packets) 1089 */ 1090 alwaysfrag = 1; 1091 mtu = IPV6_MMTU; 1092 } else if (mtu > ifmtu) { 1093 /* 1094 * The MTU on the route is larger than the MTU on 1095 * the interface! This shouldn't happen, unless the 1096 * MTU of the interface has been changed after the 1097 * interface was brought up. Change the MTU in the 1098 * route to match the interface MTU (as long as the 1099 * field isn't locked). 1100 */ 1101 mtu = ifmtu; 1102 ro_pmtu->ro_rt->rt_rmx.rmx_mtu = mtu; 1103 } 1104 } else if (ifp) { 1105 mtu = IN6_LINKMTU(ifp); 1106 } else { 1107 error = EHOSTUNREACH; /* XXX */ 1108 } 1109 1110 *mtup = mtu; 1111 if (alwaysfragp) 1112 *alwaysfragp = alwaysfrag; 1113 return (error); 1114 } 1115 1116 /* 1117 * IP6 socket option processing. 1118 */ 1119 void 1120 ip6_ctloutput_dispatch(netmsg_t msg) 1121 { 1122 int error; 1123 1124 error = ip6_ctloutput(msg->ctloutput.base.nm_so, 1125 msg->ctloutput.nm_sopt); 1126 lwkt_replymsg(&msg->ctloutput.base.lmsg, error); 1127 } 1128 1129 int 1130 ip6_ctloutput(struct socket *so, struct sockopt *sopt) 1131 { 1132 int optdatalen,uproto; 1133 int privileged; 1134 struct inpcb *in6p = so->so_pcb; 1135 void *optdata; 1136 int error, optval; 1137 int level, op, optname; 1138 int optlen; 1139 struct thread *td; 1140 1141 if (sopt) { 1142 level = sopt->sopt_level; 1143 op = sopt->sopt_dir; 1144 optname = sopt->sopt_name; 1145 optlen = sopt->sopt_valsize; 1146 td = sopt->sopt_td; 1147 } else { 1148 panic("ip6_ctloutput: arg soopt is NULL"); 1149 /* NOT REACHED */ 1150 td = NULL; 1151 } 1152 error = optval = 0; 1153 1154 uproto = (int)so->so_proto->pr_protocol; 1155 privileged = (td == NULL || priv_check(td, PRIV_ROOT)) ? 0 : 1; 1156 1157 if (level == IPPROTO_IPV6) { 1158 switch (op) { 1159 1160 case SOPT_SET: 1161 switch (optname) { 1162 case IPV6_2292PKTOPTIONS: 1163 #ifdef IPV6_PKTOPTIONS 1164 case IPV6_PKTOPTIONS: 1165 #endif 1166 { 1167 struct mbuf *m; 1168 1169 error = soopt_getm(sopt, &m); /* XXX */ 1170 if (error != 0) 1171 break; 1172 soopt_to_mbuf(sopt, m); /* XXX */ 1173 error = ip6_pcbopts(&in6p->in6p_outputopts, 1174 m, so, sopt); 1175 m_freem(m); /* XXX */ 1176 break; 1177 } 1178 1179 /* 1180 * Use of some Hop-by-Hop options or some 1181 * Destination options, might require special 1182 * privilege. That is, normal applications 1183 * (without special privilege) might be forbidden 1184 * from setting certain options in outgoing packets, 1185 * and might never see certain options in received 1186 * packets. [RFC 2292 Section 6] 1187 * KAME specific note: 1188 * KAME prevents non-privileged users from sending or 1189 * receiving ANY hbh/dst options in order to avoid 1190 * overhead of parsing options in the kernel. 1191 */ 1192 case IPV6_RECVHOPOPTS: 1193 case IPV6_RECVDSTOPTS: 1194 case IPV6_RECVRTHDRDSTOPTS: 1195 if (!privileged) 1196 return (EPERM); 1197 case IPV6_RECVPKTINFO: 1198 case IPV6_RECVHOPLIMIT: 1199 case IPV6_RECVRTHDR: 1200 case IPV6_RECVPATHMTU: 1201 case IPV6_RECVTCLASS: 1202 case IPV6_AUTOFLOWLABEL: 1203 case IPV6_HOPLIMIT: 1204 /* FALLTHROUGH */ 1205 case IPV6_UNICAST_HOPS: 1206 1207 case IPV6_V6ONLY: 1208 if (optlen != sizeof(int)) { 1209 error = EINVAL; 1210 break; 1211 } 1212 error = soopt_to_kbuf(sopt, &optval, 1213 sizeof optval, sizeof optval); 1214 if (error) 1215 break; 1216 switch (optname) { 1217 1218 case IPV6_UNICAST_HOPS: 1219 if (optval < -1 || optval >= 256) 1220 error = EINVAL; 1221 else { 1222 /* -1 = kernel default */ 1223 in6p->in6p_hops = optval; 1224 } 1225 break; 1226 #define OPTSET(bit) \ 1227 do { \ 1228 if (optval) \ 1229 in6p->in6p_flags |= (bit); \ 1230 else \ 1231 in6p->in6p_flags &= ~(bit); \ 1232 } while (0) 1233 #define OPTBIT(bit) (in6p->in6p_flags & (bit) ? 1 : 0) 1234 /* 1235 * Although changed to RFC3542, It's better to also support RFC2292 API 1236 */ 1237 #define OPTSET2292(bit) \ 1238 do { \ 1239 in6p->in6p_flags |= IN6P_RFC2292; \ 1240 if (optval) \ 1241 in6p->in6p_flags |= (bit); \ 1242 else \ 1243 in6p->in6p_flags &= ~(bit); \ 1244 } while (/*CONSTCOND*/ 0) 1245 1246 case IPV6_RECVPKTINFO: 1247 /* cannot mix with RFC2292 */ 1248 if (OPTBIT(IN6P_RFC2292)) { 1249 error = EINVAL; 1250 break; 1251 } 1252 OPTSET(IN6P_PKTINFO); 1253 break; 1254 1255 case IPV6_HOPLIMIT: 1256 { 1257 struct ip6_pktopts **optp; 1258 1259 /* cannot mix with RFC2292 */ 1260 if (OPTBIT(IN6P_RFC2292)) { 1261 error = EINVAL; 1262 break; 1263 } 1264 optp = &in6p->in6p_outputopts; 1265 error = ip6_pcbopt(IPV6_HOPLIMIT, 1266 (u_char *)&optval, sizeof(optval), 1267 optp, uproto); 1268 break; 1269 } 1270 1271 case IPV6_RECVHOPLIMIT: 1272 /* cannot mix with RFC2292 */ 1273 if (OPTBIT(IN6P_RFC2292)) { 1274 error = EINVAL; 1275 break; 1276 } 1277 OPTSET(IN6P_HOPLIMIT); 1278 break; 1279 1280 case IPV6_RECVHOPOPTS: 1281 /* cannot mix with RFC2292 */ 1282 if (OPTBIT(IN6P_RFC2292)) { 1283 error = EINVAL; 1284 break; 1285 } 1286 OPTSET(IN6P_HOPOPTS); 1287 break; 1288 1289 case IPV6_RECVDSTOPTS: 1290 /* cannot mix with RFC2292 */ 1291 if (OPTBIT(IN6P_RFC2292)) { 1292 error = EINVAL; 1293 break; 1294 } 1295 OPTSET(IN6P_DSTOPTS); 1296 break; 1297 1298 case IPV6_RECVRTHDRDSTOPTS: 1299 /* cannot mix with RFC2292 */ 1300 if (OPTBIT(IN6P_RFC2292)) { 1301 error = EINVAL; 1302 break; 1303 } 1304 OPTSET(IN6P_RTHDRDSTOPTS); 1305 break; 1306 1307 case IPV6_RECVRTHDR: 1308 /* cannot mix with RFC2292 */ 1309 if (OPTBIT(IN6P_RFC2292)) { 1310 error = EINVAL; 1311 break; 1312 } 1313 OPTSET(IN6P_RTHDR); 1314 break; 1315 1316 case IPV6_RECVPATHMTU: 1317 /* 1318 * We ignore this option for TCP 1319 * sockets. 1320 * (RFC3542 leaves this case 1321 * unspecified.) 1322 */ 1323 if (uproto != IPPROTO_TCP) 1324 OPTSET(IN6P_MTU); 1325 break; 1326 1327 case IPV6_RECVTCLASS: 1328 /* cannot mix with RFC2292 XXX */ 1329 if (OPTBIT(IN6P_RFC2292)) { 1330 error = EINVAL; 1331 break; 1332 } 1333 OPTSET(IN6P_TCLASS); 1334 break; 1335 1336 case IPV6_AUTOFLOWLABEL: 1337 OPTSET(IN6P_AUTOFLOWLABEL); 1338 break; 1339 1340 case IPV6_V6ONLY: 1341 /* 1342 * make setsockopt(IPV6_V6ONLY) 1343 * available only prior to bind(2). 1344 */ 1345 if (in6p->in6p_lport || 1346 !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr)) 1347 { 1348 error = EINVAL; 1349 break; 1350 } 1351 1352 /* 1353 * Since we don't support v6->v4 1354 * mapping any more this option does 1355 * nothing. But apparently some 1356 * ports and libraries (e.g. libuv) 1357 * actually try to set the value to 1358 * 0 so just silently ignore the value 1359 * entirely. 1360 * 1361 * (also fixes named which uses libuv, 1362 * and a few other apps) 1363 */ 1364 #if 0 1365 if (!optval) { 1366 /* Don't allow v4-mapped */ 1367 error = EOPNOTSUPP; 1368 } 1369 #endif 1370 break; 1371 } 1372 break; 1373 1374 case IPV6_TCLASS: 1375 case IPV6_DONTFRAG: 1376 case IPV6_USE_MIN_MTU: 1377 case IPV6_PREFER_TEMPADDR: 1378 if (optlen != sizeof(optval)) { 1379 error = EINVAL; 1380 break; 1381 } 1382 error = soopt_to_kbuf(sopt, &optval, 1383 sizeof optval, sizeof optval); 1384 if (error) 1385 break; 1386 { 1387 struct ip6_pktopts **optp; 1388 optp = &in6p->in6p_outputopts; 1389 error = ip6_pcbopt(optname, 1390 (u_char *)&optval, sizeof(optval), 1391 optp, uproto); 1392 break; 1393 } 1394 1395 case IPV6_2292PKTINFO: 1396 case IPV6_2292HOPLIMIT: 1397 case IPV6_2292HOPOPTS: 1398 case IPV6_2292DSTOPTS: 1399 case IPV6_2292RTHDR: 1400 /* RFC 2292 */ 1401 if (optlen != sizeof(int)) { 1402 error = EINVAL; 1403 break; 1404 } 1405 error = soopt_to_kbuf(sopt, &optval, 1406 sizeof optval, sizeof optval); 1407 if (error) 1408 break; 1409 switch (optname) { 1410 case IPV6_2292PKTINFO: 1411 OPTSET2292(IN6P_PKTINFO); 1412 break; 1413 case IPV6_2292HOPLIMIT: 1414 OPTSET2292(IN6P_HOPLIMIT); 1415 break; 1416 case IPV6_2292HOPOPTS: 1417 /* 1418 * Check super-user privilege. 1419 * See comments for IPV6_RECVHOPOPTS. 1420 */ 1421 if (!privileged) 1422 return (EPERM); 1423 OPTSET2292(IN6P_HOPOPTS); 1424 break; 1425 case IPV6_2292DSTOPTS: 1426 if (!privileged) 1427 return (EPERM); 1428 OPTSET2292(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); /* XXX */ 1429 break; 1430 case IPV6_2292RTHDR: 1431 OPTSET2292(IN6P_RTHDR); 1432 break; 1433 } 1434 break; 1435 1436 case IPV6_PKTINFO: 1437 case IPV6_HOPOPTS: 1438 case IPV6_RTHDR: 1439 case IPV6_DSTOPTS: 1440 case IPV6_RTHDRDSTOPTS: 1441 case IPV6_NEXTHOP: 1442 { 1443 /* 1444 * New advanced API (RFC3542) 1445 */ 1446 u_char *optbuf; 1447 u_char optbuf_storage[MCLBYTES]; 1448 int optlen; 1449 struct ip6_pktopts **optp; 1450 1451 /* cannot mix with RFC2292 */ 1452 if (OPTBIT(IN6P_RFC2292)) { 1453 error = EINVAL; 1454 break; 1455 } 1456 1457 /* 1458 * We only ensure valsize is not too large 1459 * here. Further validation will be done 1460 * later. 1461 */ 1462 error = soopt_to_kbuf(sopt, optbuf_storage, 1463 sizeof(optbuf_storage), 0); 1464 if (error) 1465 break; 1466 optlen = sopt->sopt_valsize; 1467 optbuf = optbuf_storage; 1468 optp = &in6p->in6p_outputopts; 1469 error = ip6_pcbopt(optname, optbuf, optlen, 1470 optp, uproto); 1471 break; 1472 } 1473 #undef OPTSET 1474 1475 case IPV6_MULTICAST_IF: 1476 case IPV6_MULTICAST_HOPS: 1477 case IPV6_MULTICAST_LOOP: 1478 case IPV6_JOIN_GROUP: 1479 case IPV6_LEAVE_GROUP: 1480 { 1481 struct mbuf *m; 1482 1483 if (sopt->sopt_valsize > MLEN) { 1484 error = EMSGSIZE; 1485 break; 1486 } 1487 /* XXX */ 1488 MGET(m, sopt->sopt_td ? M_WAITOK : M_NOWAIT, MT_HEADER); 1489 if (m == NULL) { 1490 error = ENOBUFS; 1491 break; 1492 } 1493 m->m_len = sopt->sopt_valsize; 1494 error = soopt_to_kbuf(sopt, mtod(m, char *), 1495 m->m_len, m->m_len); 1496 error = ip6_setmoptions(sopt->sopt_name, 1497 &in6p->in6p_moptions, 1498 m); 1499 m_free(m); 1500 } 1501 break; 1502 1503 case IPV6_PORTRANGE: 1504 error = soopt_to_kbuf(sopt, &optval, 1505 sizeof optval, sizeof optval); 1506 if (error) 1507 break; 1508 1509 switch (optval) { 1510 case IPV6_PORTRANGE_DEFAULT: 1511 in6p->in6p_flags &= ~(IN6P_LOWPORT); 1512 in6p->in6p_flags &= ~(IN6P_HIGHPORT); 1513 break; 1514 1515 case IPV6_PORTRANGE_HIGH: 1516 in6p->in6p_flags &= ~(IN6P_LOWPORT); 1517 in6p->in6p_flags |= IN6P_HIGHPORT; 1518 break; 1519 1520 case IPV6_PORTRANGE_LOW: 1521 in6p->in6p_flags &= ~(IN6P_HIGHPORT); 1522 in6p->in6p_flags |= IN6P_LOWPORT; 1523 break; 1524 1525 default: 1526 error = EINVAL; 1527 break; 1528 } 1529 break; 1530 1531 case IPV6_FW_ADD: 1532 case IPV6_FW_DEL: 1533 case IPV6_FW_FLUSH: 1534 case IPV6_FW_ZERO: 1535 { 1536 struct mbuf *m; 1537 struct mbuf **mp = &m; 1538 1539 if (ip6_fw_ctl_ptr == NULL) 1540 return EINVAL; 1541 /* XXX */ 1542 if ((error = soopt_getm(sopt, &m)) != 0) 1543 break; 1544 /* XXX */ 1545 soopt_to_mbuf(sopt, m); 1546 error = (*ip6_fw_ctl_ptr)(optname, mp); 1547 m = *mp; 1548 } 1549 break; 1550 1551 default: 1552 error = ENOPROTOOPT; 1553 break; 1554 } 1555 break; 1556 1557 case SOPT_GET: 1558 switch (optname) { 1559 case IPV6_2292PKTOPTIONS: 1560 #ifdef IPV6_PKTOPTIONS 1561 case IPV6_PKTOPTIONS: 1562 #endif 1563 /* 1564 * RFC3542 (effectively) deprecated the 1565 * semantics of the 2292-style pktoptions. 1566 * Since it was not reliable in nature (i.e., 1567 * applications had to expect the lack of some 1568 * information after all), it would make sense 1569 * to simplify this part by always returning 1570 * empty data. 1571 */ 1572 if (in6p->in6p_options) { 1573 struct mbuf *m; 1574 m = m_copym(in6p->in6p_options, 1575 0, M_COPYALL, M_WAITOK); 1576 error = soopt_from_mbuf(sopt, m); 1577 if (error == 0) 1578 m_freem(m); 1579 } else 1580 sopt->sopt_valsize = 0; 1581 break; 1582 1583 case IPV6_RECVHOPOPTS: 1584 case IPV6_RECVDSTOPTS: 1585 case IPV6_RECVRTHDRDSTOPTS: 1586 case IPV6_UNICAST_HOPS: 1587 case IPV6_RECVPKTINFO: 1588 case IPV6_RECVHOPLIMIT: 1589 case IPV6_RECVRTHDR: 1590 case IPV6_RECVPATHMTU: 1591 case IPV6_RECVTCLASS: 1592 case IPV6_AUTOFLOWLABEL: 1593 case IPV6_V6ONLY: 1594 case IPV6_PORTRANGE: 1595 switch (optname) { 1596 1597 case IPV6_RECVHOPOPTS: 1598 optval = OPTBIT(IN6P_HOPOPTS); 1599 break; 1600 1601 case IPV6_RECVDSTOPTS: 1602 optval = OPTBIT(IN6P_DSTOPTS); 1603 break; 1604 1605 case IPV6_RECVRTHDRDSTOPTS: 1606 optval = OPTBIT(IN6P_RTHDRDSTOPTS); 1607 break; 1608 1609 case IPV6_RECVPKTINFO: 1610 optval = OPTBIT(IN6P_PKTINFO); 1611 break; 1612 1613 case IPV6_RECVHOPLIMIT: 1614 optval = OPTBIT(IN6P_HOPLIMIT); 1615 break; 1616 1617 case IPV6_RECVRTHDR: 1618 optval = OPTBIT(IN6P_RTHDR); 1619 break; 1620 1621 case IPV6_RECVPATHMTU: 1622 optval = OPTBIT(IN6P_MTU); 1623 break; 1624 1625 case IPV6_RECVTCLASS: 1626 optval = OPTBIT(IN6P_TCLASS); 1627 break; 1628 1629 case IPV6_AUTOFLOWLABEL: 1630 optval = OPTBIT(IN6P_AUTOFLOWLABEL); 1631 break; 1632 1633 1634 case IPV6_UNICAST_HOPS: 1635 optval = in6p->in6p_hops; 1636 break; 1637 1638 case IPV6_V6ONLY: 1639 optval = 1; 1640 break; 1641 1642 case IPV6_PORTRANGE: 1643 { 1644 int flags; 1645 flags = in6p->in6p_flags; 1646 if (flags & IN6P_HIGHPORT) 1647 optval = IPV6_PORTRANGE_HIGH; 1648 else if (flags & IN6P_LOWPORT) 1649 optval = IPV6_PORTRANGE_LOW; 1650 else 1651 optval = 0; 1652 break; 1653 } 1654 } 1655 soopt_from_kbuf(sopt, &optval, 1656 sizeof optval); 1657 break; 1658 1659 case IPV6_PATHMTU: 1660 { 1661 u_long pmtu = 0; 1662 struct ip6_mtuinfo mtuinfo; 1663 struct route_in6 sro; 1664 1665 bzero(&sro, sizeof(sro)); 1666 1667 if (!(so->so_state & SS_ISCONNECTED)) 1668 return (ENOTCONN); 1669 /* 1670 * XXX: we dot not consider the case of source 1671 * routing, or optional information to specify 1672 * the outgoing interface. 1673 */ 1674 error = ip6_getpmtu(&sro, NULL, NULL, 1675 &in6p->in6p_faddr, &pmtu, NULL); 1676 if (sro.ro_rt) 1677 RTFREE(sro.ro_rt); 1678 if (error) 1679 break; 1680 if (pmtu > IPV6_MAXPACKET) 1681 pmtu = IPV6_MAXPACKET; 1682 1683 bzero(&mtuinfo, sizeof(mtuinfo)); 1684 mtuinfo.ip6m_mtu = (u_int32_t)pmtu; 1685 optdata = (void *)&mtuinfo; 1686 optdatalen = sizeof(mtuinfo); 1687 soopt_from_kbuf(sopt, optdata, 1688 optdatalen); 1689 break; 1690 } 1691 1692 case IPV6_2292PKTINFO: 1693 case IPV6_2292HOPLIMIT: 1694 case IPV6_2292HOPOPTS: 1695 case IPV6_2292RTHDR: 1696 case IPV6_2292DSTOPTS: 1697 if (optname == IPV6_2292HOPOPTS || 1698 optname == IPV6_2292DSTOPTS || 1699 !privileged) 1700 return (EPERM); 1701 switch (optname) { 1702 case IPV6_2292PKTINFO: 1703 optval = OPTBIT(IN6P_PKTINFO); 1704 break; 1705 case IPV6_2292HOPLIMIT: 1706 optval = OPTBIT(IN6P_HOPLIMIT); 1707 break; 1708 case IPV6_2292HOPOPTS: 1709 if (!privileged) 1710 return (EPERM); 1711 optval = OPTBIT(IN6P_HOPOPTS); 1712 break; 1713 case IPV6_2292RTHDR: 1714 optval = OPTBIT(IN6P_RTHDR); 1715 break; 1716 case IPV6_2292DSTOPTS: 1717 if (!privileged) 1718 return (EPERM); 1719 optval = OPTBIT(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); 1720 break; 1721 } 1722 soopt_from_kbuf(sopt, &optval, 1723 sizeof optval); 1724 break; 1725 1726 case IPV6_PKTINFO: 1727 case IPV6_HOPOPTS: 1728 case IPV6_RTHDR: 1729 case IPV6_DSTOPTS: 1730 case IPV6_RTHDRDSTOPTS: 1731 case IPV6_NEXTHOP: 1732 case IPV6_TCLASS: 1733 case IPV6_DONTFRAG: 1734 case IPV6_USE_MIN_MTU: 1735 case IPV6_PREFER_TEMPADDR: 1736 error = ip6_getpcbopt(in6p->in6p_outputopts, 1737 optname, sopt); 1738 break; 1739 1740 case IPV6_MULTICAST_IF: 1741 case IPV6_MULTICAST_HOPS: 1742 case IPV6_MULTICAST_LOOP: 1743 case IPV6_JOIN_GROUP: 1744 case IPV6_LEAVE_GROUP: 1745 { 1746 struct mbuf *m; 1747 error = ip6_getmoptions(sopt->sopt_name, 1748 in6p->in6p_moptions, &m); 1749 if (error == 0) { 1750 soopt_from_kbuf(sopt, 1751 mtod(m, char *), m->m_len); 1752 } 1753 m_freem(m); 1754 } 1755 break; 1756 1757 case IPV6_FW_GET: 1758 { 1759 struct mbuf *m; 1760 struct mbuf **mp = &m; 1761 1762 if (ip6_fw_ctl_ptr == NULL) 1763 { 1764 return EINVAL; 1765 } 1766 error = (*ip6_fw_ctl_ptr)(optname, mp); 1767 if (error == 0) 1768 error = soopt_from_mbuf(sopt, m); /* XXX */ 1769 if (error == 0 && m != NULL) 1770 m_freem(m); 1771 } 1772 break; 1773 1774 default: 1775 error = ENOPROTOOPT; 1776 break; 1777 } 1778 break; 1779 } 1780 } else { 1781 error = EINVAL; 1782 } 1783 return (error); 1784 } 1785 1786 int 1787 ip6_raw_ctloutput(struct socket *so, struct sockopt *sopt) 1788 { 1789 int error = 0, optval, optlen; 1790 const int icmp6off = offsetof(struct icmp6_hdr, icmp6_cksum); 1791 struct in6pcb *in6p = sotoin6pcb(so); 1792 int level, op, optname; 1793 1794 if (sopt) { 1795 level = sopt->sopt_level; 1796 op = sopt->sopt_dir; 1797 optname = sopt->sopt_name; 1798 optlen = sopt->sopt_valsize; 1799 } else 1800 panic("ip6_raw_ctloutput: arg soopt is NULL"); 1801 1802 if (level != IPPROTO_IPV6) { 1803 return (EINVAL); 1804 } 1805 1806 switch (optname) { 1807 case IPV6_CHECKSUM: 1808 /* 1809 * For ICMPv6 sockets, no modification allowed for checksum 1810 * offset, permit "no change" values to help existing apps. 1811 * 1812 * RFC3542 says: "An attempt to set IPV6_CHECKSUM 1813 * for an ICMPv6 socket will fail." 1814 * The current behavior does not meet RFC3542. 1815 */ 1816 switch (op) { 1817 case SOPT_SET: 1818 if (optlen != sizeof(int)) { 1819 error = EINVAL; 1820 break; 1821 } 1822 error = soopt_to_kbuf(sopt, &optval, 1823 sizeof optval, sizeof optval); 1824 if (error) 1825 break; 1826 if ((optval % 2) != 0) { 1827 /* the API assumes even offset values */ 1828 error = EINVAL; 1829 } else if (so->so_proto->pr_protocol == 1830 IPPROTO_ICMPV6) { 1831 if (optval != icmp6off) 1832 error = EINVAL; 1833 } else 1834 in6p->in6p_cksum = optval; 1835 break; 1836 1837 case SOPT_GET: 1838 if (so->so_proto->pr_protocol == IPPROTO_ICMPV6) 1839 optval = icmp6off; 1840 else 1841 optval = in6p->in6p_cksum; 1842 1843 soopt_from_kbuf(sopt, &optval, sizeof(optval)); 1844 break; 1845 1846 default: 1847 error = EINVAL; 1848 break; 1849 } 1850 break; 1851 1852 default: 1853 error = ENOPROTOOPT; 1854 break; 1855 } 1856 1857 return (error); 1858 } 1859 1860 /* 1861 * Set up IP6 options in pcb for insertion in output packets or 1862 * specifying behavior of outgoing packets. 1863 */ 1864 static int 1865 ip6_pcbopts(struct ip6_pktopts **pktopt, struct mbuf *m, 1866 struct socket *so, struct sockopt *sopt) 1867 { 1868 int priv = 0; 1869 struct ip6_pktopts *opt = *pktopt; 1870 int error = 0; 1871 1872 /* turn off any old options. */ 1873 if (opt) { 1874 #ifdef DIAGNOSTIC 1875 if (opt->ip6po_pktinfo || opt->ip6po_nexthop || 1876 opt->ip6po_hbh || opt->ip6po_dest1 || opt->ip6po_dest2 || 1877 opt->ip6po_rhinfo.ip6po_rhi_rthdr) 1878 kprintf("ip6_pcbopts: all specified options are cleared.\n"); 1879 #endif 1880 ip6_clearpktopts(opt, -1); 1881 } else 1882 opt = kmalloc(sizeof(*opt), M_IP6OPT, M_WAITOK); 1883 *pktopt = NULL; 1884 1885 if (!m || m->m_len == 0) { 1886 /* 1887 * Only turning off any previous options, regardless of 1888 * whether the opt is just created or given. 1889 */ 1890 kfree(opt, M_IP6OPT); 1891 return (0); 1892 } 1893 1894 /* set options specified by user. */ 1895 if ((error = ip6_setpktoptions(m, opt, NULL, so->so_proto->pr_protocol, priv)) != 0) { 1896 ip6_clearpktopts(opt, -1); /* XXX: discard all options */ 1897 kfree(opt, M_IP6OPT); 1898 return (error); 1899 } 1900 *pktopt = opt; 1901 return (0); 1902 } 1903 1904 1905 /* 1906 * Below three functions are introduced by merge to RFC3542 1907 */ 1908 1909 static int 1910 ip6_getpcbopt(struct ip6_pktopts *pktopt, int optname, struct sockopt *sopt) 1911 { 1912 void *optdata = NULL; 1913 int optdatalen = 0; 1914 struct ip6_ext *ip6e; 1915 int error = 0; 1916 struct in6_pktinfo null_pktinfo; 1917 int deftclass = 0, on; 1918 int defminmtu = IP6PO_MINMTU_MCASTONLY; 1919 int defpreftemp = IP6PO_TEMPADDR_SYSTEM; 1920 1921 switch (optname) { 1922 case IPV6_PKTINFO: 1923 if (pktopt && pktopt->ip6po_pktinfo) 1924 optdata = (void *)pktopt->ip6po_pktinfo; 1925 else { 1926 /* XXX: we don't have to do this every time... */ 1927 bzero(&null_pktinfo, sizeof(null_pktinfo)); 1928 optdata = (void *)&null_pktinfo; 1929 } 1930 optdatalen = sizeof(struct in6_pktinfo); 1931 break; 1932 case IPV6_TCLASS: 1933 if (pktopt && pktopt->ip6po_tclass >= 0) 1934 optdata = (void *)&pktopt->ip6po_tclass; 1935 else 1936 optdata = (void *)&deftclass; 1937 optdatalen = sizeof(int); 1938 break; 1939 case IPV6_HOPOPTS: 1940 if (pktopt && pktopt->ip6po_hbh) { 1941 optdata = (void *)pktopt->ip6po_hbh; 1942 ip6e = (struct ip6_ext *)pktopt->ip6po_hbh; 1943 optdatalen = (ip6e->ip6e_len + 1) << 3; 1944 } 1945 break; 1946 case IPV6_RTHDR: 1947 if (pktopt && pktopt->ip6po_rthdr) { 1948 optdata = (void *)pktopt->ip6po_rthdr; 1949 ip6e = (struct ip6_ext *)pktopt->ip6po_rthdr; 1950 optdatalen = (ip6e->ip6e_len + 1) << 3; 1951 } 1952 break; 1953 case IPV6_RTHDRDSTOPTS: 1954 if (pktopt && pktopt->ip6po_dest1) { 1955 optdata = (void *)pktopt->ip6po_dest1; 1956 ip6e = (struct ip6_ext *)pktopt->ip6po_dest1; 1957 optdatalen = (ip6e->ip6e_len + 1) << 3; 1958 } 1959 break; 1960 case IPV6_DSTOPTS: 1961 if (pktopt && pktopt->ip6po_dest2) { 1962 optdata = (void *)pktopt->ip6po_dest2; 1963 ip6e = (struct ip6_ext *)pktopt->ip6po_dest2; 1964 optdatalen = (ip6e->ip6e_len + 1) << 3; 1965 } 1966 break; 1967 case IPV6_NEXTHOP: 1968 if (pktopt && pktopt->ip6po_nexthop) { 1969 optdata = (void *)pktopt->ip6po_nexthop; 1970 optdatalen = pktopt->ip6po_nexthop->sa_len; 1971 } 1972 break; 1973 case IPV6_USE_MIN_MTU: 1974 if (pktopt) 1975 optdata = (void *)&pktopt->ip6po_minmtu; 1976 else 1977 optdata = (void *)&defminmtu; 1978 optdatalen = sizeof(int); 1979 break; 1980 case IPV6_DONTFRAG: 1981 if (pktopt && ((pktopt->ip6po_flags) & IP6PO_DONTFRAG)) 1982 on = 1; 1983 else 1984 on = 0; 1985 optdata = (void *)&on; 1986 optdatalen = sizeof(on); 1987 break; 1988 case IPV6_PREFER_TEMPADDR: 1989 if (pktopt) 1990 optdata = (void *)&pktopt->ip6po_prefer_tempaddr; 1991 else 1992 optdata = (void *)&defpreftemp; 1993 optdatalen = sizeof(int); 1994 break; 1995 default: /* should not happen */ 1996 #ifdef DIAGNOSTIC 1997 panic("ip6_getpcbopt: unexpected option"); 1998 #endif 1999 return (ENOPROTOOPT); 2000 } 2001 2002 soopt_from_kbuf(sopt, optdata, optdatalen); 2003 2004 return (error); 2005 } 2006 2007 /* 2008 * initialize ip6_pktopts. beware that there are non-zero default values in 2009 * the struct. 2010 */ 2011 2012 static int 2013 ip6_pcbopt(int optname, u_char *buf, int len, struct ip6_pktopts **pktopt, int uproto) 2014 { 2015 struct ip6_pktopts *opt; 2016 int priv =0; 2017 if (*pktopt == NULL) { 2018 *pktopt = kmalloc(sizeof(*opt), M_IP6OPT, M_WAITOK); 2019 init_ip6pktopts(*pktopt); 2020 } 2021 opt = *pktopt; 2022 2023 return (ip6_setpktoption(optname, buf, len, opt, 1, 0, uproto, priv)); 2024 } 2025 2026 /* 2027 * initialize ip6_pktopts. beware that there are non-zero default values in 2028 * the struct. 2029 */ 2030 void 2031 init_ip6pktopts(struct ip6_pktopts *opt) 2032 { 2033 2034 bzero(opt, sizeof(*opt)); 2035 opt->ip6po_hlim = -1; /* -1 means default hop limit */ 2036 opt->ip6po_tclass = -1; /* -1 means default traffic class */ 2037 opt->ip6po_minmtu = IP6PO_MINMTU_MCASTONLY; 2038 opt->ip6po_prefer_tempaddr = IP6PO_TEMPADDR_SYSTEM; 2039 } 2040 2041 void 2042 ip6_clearpktopts(struct ip6_pktopts *pktopt, int optname) 2043 { 2044 if (pktopt == NULL) 2045 return; 2046 2047 if (optname == -1 || optname == IPV6_PKTINFO) { 2048 if (pktopt->ip6po_pktinfo) 2049 kfree(pktopt->ip6po_pktinfo, M_IP6OPT); 2050 pktopt->ip6po_pktinfo = NULL; 2051 } 2052 if (optname == -1 || optname == IPV6_HOPLIMIT) 2053 pktopt->ip6po_hlim = -1; 2054 if (optname == -1 || optname == IPV6_TCLASS) 2055 pktopt->ip6po_tclass = -1; 2056 if (optname == -1 || optname == IPV6_NEXTHOP) { 2057 if (pktopt->ip6po_nextroute.ro_rt) { 2058 RTFREE(pktopt->ip6po_nextroute.ro_rt); 2059 pktopt->ip6po_nextroute.ro_rt = NULL; 2060 } 2061 if (pktopt->ip6po_nexthop) 2062 kfree(pktopt->ip6po_nexthop, M_IP6OPT); 2063 pktopt->ip6po_nexthop = NULL; 2064 } 2065 if (optname == -1 || optname == IPV6_HOPOPTS) { 2066 if (pktopt->ip6po_hbh) 2067 kfree(pktopt->ip6po_hbh, M_IP6OPT); 2068 pktopt->ip6po_hbh = NULL; 2069 } 2070 if (optname == -1 || optname == IPV6_RTHDRDSTOPTS) { 2071 if (pktopt->ip6po_dest1) 2072 kfree(pktopt->ip6po_dest1, M_IP6OPT); 2073 pktopt->ip6po_dest1 = NULL; 2074 } 2075 if (optname == -1 || optname == IPV6_RTHDR) { 2076 if (pktopt->ip6po_rhinfo.ip6po_rhi_rthdr) 2077 kfree(pktopt->ip6po_rhinfo.ip6po_rhi_rthdr, M_IP6OPT); 2078 pktopt->ip6po_rhinfo.ip6po_rhi_rthdr = NULL; 2079 if (pktopt->ip6po_route.ro_rt) { 2080 RTFREE(pktopt->ip6po_route.ro_rt); 2081 pktopt->ip6po_route.ro_rt = NULL; 2082 } 2083 } 2084 if (optname == -1 || optname == IPV6_DSTOPTS) { 2085 if (pktopt->ip6po_dest2) 2086 kfree(pktopt->ip6po_dest2, M_IP6OPT); 2087 pktopt->ip6po_dest2 = NULL; 2088 } 2089 } 2090 2091 #define PKTOPT_EXTHDRCPY(type) \ 2092 do {\ 2093 if (src->type) {\ 2094 int hlen = (((struct ip6_ext *)src->type)->ip6e_len + 1) << 3;\ 2095 dst->type = kmalloc(hlen, M_IP6OPT, canwait);\ 2096 if (dst->type == NULL)\ 2097 goto bad;\ 2098 bcopy(src->type, dst->type, hlen);\ 2099 }\ 2100 } while (0) 2101 2102 struct ip6_pktopts * 2103 ip6_copypktopts(struct ip6_pktopts *src, int canwait) 2104 { 2105 struct ip6_pktopts *dst; 2106 2107 if (src == NULL) { 2108 kprintf("ip6_clearpktopts: invalid argument\n"); 2109 return (NULL); 2110 } 2111 2112 dst = kmalloc(sizeof(*dst), M_IP6OPT, canwait | M_ZERO); 2113 if (dst == NULL) 2114 return (NULL); 2115 2116 dst->ip6po_hlim = src->ip6po_hlim; 2117 if (src->ip6po_pktinfo) { 2118 dst->ip6po_pktinfo = kmalloc(sizeof(*dst->ip6po_pktinfo), 2119 M_IP6OPT, canwait); 2120 if (dst->ip6po_pktinfo == NULL) 2121 goto bad; 2122 *dst->ip6po_pktinfo = *src->ip6po_pktinfo; 2123 } 2124 if (src->ip6po_nexthop) { 2125 dst->ip6po_nexthop = kmalloc(src->ip6po_nexthop->sa_len, 2126 M_IP6OPT, canwait); 2127 if (dst->ip6po_nexthop == NULL) 2128 goto bad; 2129 bcopy(src->ip6po_nexthop, dst->ip6po_nexthop, 2130 src->ip6po_nexthop->sa_len); 2131 } 2132 PKTOPT_EXTHDRCPY(ip6po_hbh); 2133 PKTOPT_EXTHDRCPY(ip6po_dest1); 2134 PKTOPT_EXTHDRCPY(ip6po_dest2); 2135 PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */ 2136 return (dst); 2137 2138 bad: 2139 if (dst->ip6po_pktinfo) kfree(dst->ip6po_pktinfo, M_IP6OPT); 2140 if (dst->ip6po_nexthop) kfree(dst->ip6po_nexthop, M_IP6OPT); 2141 if (dst->ip6po_hbh) kfree(dst->ip6po_hbh, M_IP6OPT); 2142 if (dst->ip6po_dest1) kfree(dst->ip6po_dest1, M_IP6OPT); 2143 if (dst->ip6po_dest2) kfree(dst->ip6po_dest2, M_IP6OPT); 2144 if (dst->ip6po_rthdr) kfree(dst->ip6po_rthdr, M_IP6OPT); 2145 kfree(dst, M_IP6OPT); 2146 return (NULL); 2147 } 2148 2149 static int 2150 copypktopts(struct ip6_pktopts *dst, struct ip6_pktopts *src, int canwait) 2151 { 2152 if (dst == NULL || src == NULL) { 2153 #ifdef DIAGNOSTIC 2154 kprintf("ip6_clearpktopts: invalid argument\n"); 2155 #endif 2156 return (EINVAL); 2157 } 2158 2159 dst->ip6po_hlim = src->ip6po_hlim; 2160 dst->ip6po_tclass = src->ip6po_tclass; 2161 dst->ip6po_flags = src->ip6po_flags; 2162 if (src->ip6po_pktinfo) { 2163 dst->ip6po_pktinfo = kmalloc(sizeof(*dst->ip6po_pktinfo), 2164 M_IP6OPT, canwait); 2165 if (dst->ip6po_pktinfo == NULL) 2166 goto bad; 2167 *dst->ip6po_pktinfo = *src->ip6po_pktinfo; 2168 } 2169 if (src->ip6po_nexthop) { 2170 dst->ip6po_nexthop = kmalloc(src->ip6po_nexthop->sa_len, 2171 M_IP6OPT, canwait); 2172 if (dst->ip6po_nexthop == NULL) 2173 goto bad; 2174 bcopy(src->ip6po_nexthop, dst->ip6po_nexthop, 2175 src->ip6po_nexthop->sa_len); 2176 } 2177 PKTOPT_EXTHDRCPY(ip6po_hbh); 2178 PKTOPT_EXTHDRCPY(ip6po_dest1); 2179 PKTOPT_EXTHDRCPY(ip6po_dest2); 2180 PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */ 2181 return (0); 2182 2183 bad: 2184 ip6_clearpktopts(dst, -1); 2185 return (ENOBUFS); 2186 } 2187 #undef PKTOPT_EXTHDRCPY 2188 2189 void 2190 ip6_freepcbopts(struct ip6_pktopts *pktopt) 2191 { 2192 if (pktopt == NULL) 2193 return; 2194 2195 ip6_clearpktopts(pktopt, -1); 2196 2197 kfree(pktopt, M_IP6OPT); 2198 } 2199 2200 /* 2201 * Set the IP6 multicast options in response to user setsockopt(). 2202 */ 2203 static int 2204 ip6_setmoptions(int optname, struct ip6_moptions **im6op, struct mbuf *m) 2205 { 2206 int error = 0; 2207 u_int loop, ifindex; 2208 struct ipv6_mreq *mreq; 2209 struct ifnet *ifp; 2210 struct ip6_moptions *im6o = *im6op; 2211 struct route_in6 ro; 2212 struct sockaddr_in6 *dst; 2213 struct in6_multi_mship *imm; 2214 struct thread *td = curthread; 2215 2216 if (im6o == NULL) { 2217 /* 2218 * No multicast option buffer attached to the pcb; 2219 * allocate one and initialize to default values. 2220 */ 2221 im6o = (struct ip6_moptions *) 2222 kmalloc(sizeof(*im6o), M_IPMOPTS, M_WAITOK); 2223 2224 *im6op = im6o; 2225 im6o->im6o_multicast_ifp = NULL; 2226 im6o->im6o_multicast_hlim = ip6_defmcasthlim; 2227 im6o->im6o_multicast_loop = IPV6_DEFAULT_MULTICAST_LOOP; 2228 LIST_INIT(&im6o->im6o_memberships); 2229 } 2230 2231 switch (optname) { 2232 2233 case IPV6_MULTICAST_IF: 2234 /* 2235 * Select the interface for outgoing multicast packets. 2236 */ 2237 if (m == NULL || m->m_len != sizeof(u_int)) { 2238 error = EINVAL; 2239 break; 2240 } 2241 bcopy(mtod(m, u_int *), &ifindex, sizeof(ifindex)); 2242 if (ifindex < 0 || if_index < ifindex) { 2243 error = ENXIO; /* XXX EINVAL? */ 2244 break; 2245 } 2246 ifp = ifindex2ifnet[ifindex]; 2247 if (ifp == NULL || !(ifp->if_flags & IFF_MULTICAST)) { 2248 error = EADDRNOTAVAIL; 2249 break; 2250 } 2251 im6o->im6o_multicast_ifp = ifp; 2252 break; 2253 2254 case IPV6_MULTICAST_HOPS: 2255 { 2256 /* 2257 * Set the IP6 hoplimit for outgoing multicast packets. 2258 */ 2259 int optval; 2260 if (m == NULL || m->m_len != sizeof(int)) { 2261 error = EINVAL; 2262 break; 2263 } 2264 bcopy(mtod(m, u_int *), &optval, sizeof(optval)); 2265 if (optval < -1 || optval >= 256) 2266 error = EINVAL; 2267 else if (optval == -1) 2268 im6o->im6o_multicast_hlim = ip6_defmcasthlim; 2269 else 2270 im6o->im6o_multicast_hlim = optval; 2271 break; 2272 } 2273 2274 case IPV6_MULTICAST_LOOP: 2275 /* 2276 * Set the loopback flag for outgoing multicast packets. 2277 * Must be zero or one. 2278 */ 2279 if (m == NULL || m->m_len != sizeof(u_int)) { 2280 error = EINVAL; 2281 break; 2282 } 2283 bcopy(mtod(m, u_int *), &loop, sizeof(loop)); 2284 if (loop > 1) { 2285 error = EINVAL; 2286 break; 2287 } 2288 im6o->im6o_multicast_loop = loop; 2289 break; 2290 2291 case IPV6_JOIN_GROUP: 2292 /* 2293 * Add a multicast group membership. 2294 * Group must be a valid IP6 multicast address. 2295 */ 2296 if (m == NULL || m->m_len != sizeof(struct ipv6_mreq)) { 2297 error = EINVAL; 2298 break; 2299 } 2300 mreq = mtod(m, struct ipv6_mreq *); 2301 if (IN6_IS_ADDR_UNSPECIFIED(&mreq->ipv6mr_multiaddr)) { 2302 /* 2303 * We use the unspecified address to specify to accept 2304 * all multicast addresses. Only super user is allowed 2305 * to do this. 2306 */ 2307 if (priv_check(td, PRIV_ROOT)) { 2308 error = EACCES; 2309 break; 2310 } 2311 } else if (!IN6_IS_ADDR_MULTICAST(&mreq->ipv6mr_multiaddr)) { 2312 error = EINVAL; 2313 break; 2314 } 2315 2316 /* 2317 * If the interface is specified, validate it. 2318 */ 2319 if (mreq->ipv6mr_interface < 0 2320 || if_index < mreq->ipv6mr_interface) { 2321 error = ENXIO; /* XXX EINVAL? */ 2322 break; 2323 } 2324 /* 2325 * If no interface was explicitly specified, choose an 2326 * appropriate one according to the given multicast address. 2327 */ 2328 if (mreq->ipv6mr_interface == 0) { 2329 /* 2330 * If the multicast address is in node-local scope, 2331 * the interface should be a loopback interface. 2332 * Otherwise, look up the routing table for the 2333 * address, and choose the outgoing interface. 2334 * XXX: is it a good approach? 2335 */ 2336 if (IN6_IS_ADDR_MC_INTFACELOCAL(&mreq->ipv6mr_multiaddr)) { 2337 ifp = loif; 2338 } else { 2339 ro.ro_rt = NULL; 2340 dst = (struct sockaddr_in6 *)&ro.ro_dst; 2341 bzero(dst, sizeof(*dst)); 2342 dst->sin6_len = sizeof(struct sockaddr_in6); 2343 dst->sin6_family = AF_INET6; 2344 dst->sin6_addr = mreq->ipv6mr_multiaddr; 2345 rtalloc((struct route *)&ro); 2346 if (ro.ro_rt == NULL) { 2347 error = EADDRNOTAVAIL; 2348 break; 2349 } 2350 ifp = ro.ro_rt->rt_ifp; 2351 rtfree(ro.ro_rt); 2352 } 2353 } else 2354 ifp = ifindex2ifnet[mreq->ipv6mr_interface]; 2355 2356 /* 2357 * See if we found an interface, and confirm that it 2358 * supports multicast 2359 */ 2360 if (ifp == NULL || !(ifp->if_flags & IFF_MULTICAST)) { 2361 error = EADDRNOTAVAIL; 2362 break; 2363 } 2364 /* 2365 * Put interface index into the multicast address, 2366 * if the address has link-local scope. 2367 */ 2368 if (IN6_IS_ADDR_MC_LINKLOCAL(&mreq->ipv6mr_multiaddr)) { 2369 mreq->ipv6mr_multiaddr.s6_addr16[1] 2370 = htons(mreq->ipv6mr_interface); 2371 } 2372 /* 2373 * See if the membership already exists. 2374 */ 2375 for (imm = im6o->im6o_memberships.lh_first; 2376 imm != NULL; imm = imm->i6mm_chain.le_next) 2377 if (imm->i6mm_maddr->in6m_ifp == ifp && 2378 IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr, 2379 &mreq->ipv6mr_multiaddr)) 2380 break; 2381 if (imm != NULL) { 2382 error = EADDRINUSE; 2383 break; 2384 } 2385 /* 2386 * Everything looks good; add a new record to the multicast 2387 * address list for the given interface. 2388 */ 2389 imm = kmalloc(sizeof(*imm), M_IPMADDR, M_WAITOK); 2390 if ((imm->i6mm_maddr = 2391 in6_addmulti(&mreq->ipv6mr_multiaddr, ifp, &error)) == NULL) { 2392 kfree(imm, M_IPMADDR); 2393 break; 2394 } 2395 LIST_INSERT_HEAD(&im6o->im6o_memberships, imm, i6mm_chain); 2396 break; 2397 2398 case IPV6_LEAVE_GROUP: 2399 /* 2400 * Drop a multicast group membership. 2401 * Group must be a valid IP6 multicast address. 2402 */ 2403 if (m == NULL || m->m_len != sizeof(struct ipv6_mreq)) { 2404 error = EINVAL; 2405 break; 2406 } 2407 mreq = mtod(m, struct ipv6_mreq *); 2408 if (IN6_IS_ADDR_UNSPECIFIED(&mreq->ipv6mr_multiaddr)) { 2409 if (priv_check(td, PRIV_ROOT)) { 2410 error = EACCES; 2411 break; 2412 } 2413 } else if (!IN6_IS_ADDR_MULTICAST(&mreq->ipv6mr_multiaddr)) { 2414 error = EINVAL; 2415 break; 2416 } 2417 /* 2418 * If an interface address was specified, get a pointer 2419 * to its ifnet structure. 2420 */ 2421 if (mreq->ipv6mr_interface < 0 2422 || if_index < mreq->ipv6mr_interface) { 2423 error = ENXIO; /* XXX EINVAL? */ 2424 break; 2425 } 2426 ifp = ifindex2ifnet[mreq->ipv6mr_interface]; 2427 /* 2428 * Put interface index into the multicast address, 2429 * if the address has link-local scope. 2430 */ 2431 if (IN6_IS_ADDR_MC_LINKLOCAL(&mreq->ipv6mr_multiaddr)) { 2432 mreq->ipv6mr_multiaddr.s6_addr16[1] 2433 = htons(mreq->ipv6mr_interface); 2434 } 2435 2436 /* 2437 * Find the membership in the membership list. 2438 */ 2439 for (imm = im6o->im6o_memberships.lh_first; 2440 imm != NULL; imm = imm->i6mm_chain.le_next) { 2441 if ((ifp == NULL || imm->i6mm_maddr->in6m_ifp == ifp) && 2442 IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr, 2443 &mreq->ipv6mr_multiaddr)) 2444 break; 2445 } 2446 if (imm == NULL) { 2447 /* Unable to resolve interface */ 2448 error = EADDRNOTAVAIL; 2449 break; 2450 } 2451 /* 2452 * Give up the multicast address record to which the 2453 * membership points. 2454 */ 2455 LIST_REMOVE(imm, i6mm_chain); 2456 in6_delmulti(imm->i6mm_maddr); 2457 kfree(imm, M_IPMADDR); 2458 break; 2459 2460 default: 2461 error = EOPNOTSUPP; 2462 break; 2463 } 2464 2465 /* 2466 * If all options have default values, no need to keep the mbuf. 2467 */ 2468 if (im6o->im6o_multicast_ifp == NULL && 2469 im6o->im6o_multicast_hlim == ip6_defmcasthlim && 2470 im6o->im6o_multicast_loop == IPV6_DEFAULT_MULTICAST_LOOP && 2471 im6o->im6o_memberships.lh_first == NULL) { 2472 kfree(*im6op, M_IPMOPTS); 2473 *im6op = NULL; 2474 } 2475 2476 return (error); 2477 } 2478 2479 /* 2480 * Return the IP6 multicast options in response to user getsockopt(). 2481 */ 2482 static int 2483 ip6_getmoptions(int optname, struct ip6_moptions *im6o, struct mbuf **mp) 2484 { 2485 u_int *hlim, *loop, *ifindex; 2486 2487 *mp = m_get(M_WAITOK, MT_HEADER); /* XXX */ 2488 2489 switch (optname) { 2490 2491 case IPV6_MULTICAST_IF: 2492 ifindex = mtod(*mp, u_int *); 2493 (*mp)->m_len = sizeof(u_int); 2494 if (im6o == NULL || im6o->im6o_multicast_ifp == NULL) 2495 *ifindex = 0; 2496 else 2497 *ifindex = im6o->im6o_multicast_ifp->if_index; 2498 return (0); 2499 2500 case IPV6_MULTICAST_HOPS: 2501 hlim = mtod(*mp, u_int *); 2502 (*mp)->m_len = sizeof(u_int); 2503 if (im6o == NULL) 2504 *hlim = ip6_defmcasthlim; 2505 else 2506 *hlim = im6o->im6o_multicast_hlim; 2507 return (0); 2508 2509 case IPV6_MULTICAST_LOOP: 2510 loop = mtod(*mp, u_int *); 2511 (*mp)->m_len = sizeof(u_int); 2512 if (im6o == NULL) 2513 *loop = ip6_defmcasthlim; 2514 else 2515 *loop = im6o->im6o_multicast_loop; 2516 return (0); 2517 2518 default: 2519 return (EOPNOTSUPP); 2520 } 2521 } 2522 2523 /* 2524 * Discard the IP6 multicast options. 2525 */ 2526 void 2527 ip6_freemoptions(struct ip6_moptions *im6o) 2528 { 2529 struct in6_multi_mship *imm; 2530 2531 if (im6o == NULL) 2532 return; 2533 2534 while ((imm = im6o->im6o_memberships.lh_first) != NULL) { 2535 LIST_REMOVE(imm, i6mm_chain); 2536 if (imm->i6mm_maddr) 2537 in6_delmulti(imm->i6mm_maddr); 2538 kfree(imm, M_IPMADDR); 2539 } 2540 kfree(im6o, M_IPMOPTS); 2541 } 2542 2543 /* 2544 * Set a particular packet option, as a sticky option or an ancillary data 2545 * item. "len" can be 0 only when it's a sticky option. 2546 * We have 4 cases of combination of "sticky" and "cmsg": 2547 * "sticky=0, cmsg=0": impossible 2548 * "sticky=0, cmsg=1": RFC2292 or RFC3542 ancillary data 2549 * "sticky=1, cmsg=0": RFC3542 socket option 2550 * "sticky=1, cmsg=1": RFC2292 socket option 2551 */ 2552 static int 2553 ip6_setpktoption(int optname, u_char *buf, int len, struct ip6_pktopts *opt, 2554 int sticky, int cmsg, int uproto, int priv) 2555 { 2556 int minmtupolicy, preftemp; 2557 //int error; 2558 2559 if (!sticky && !cmsg) { 2560 kprintf("ip6_setpktoption: impossible case\n"); 2561 return (EINVAL); 2562 } 2563 2564 /* 2565 * IPV6_2292xxx is for backward compatibility to RFC2292, and should 2566 * not be specified in the context of RFC3542. Conversely, 2567 * RFC3542 types should not be specified in the context of RFC2292. 2568 */ 2569 if (!cmsg) { 2570 switch (optname) { 2571 case IPV6_2292PKTINFO: 2572 case IPV6_2292HOPLIMIT: 2573 case IPV6_2292NEXTHOP: 2574 case IPV6_2292HOPOPTS: 2575 case IPV6_2292DSTOPTS: 2576 case IPV6_2292RTHDR: 2577 case IPV6_2292PKTOPTIONS: 2578 return (ENOPROTOOPT); 2579 } 2580 } 2581 if (sticky && cmsg) { 2582 switch (optname) { 2583 case IPV6_PKTINFO: 2584 case IPV6_HOPLIMIT: 2585 case IPV6_NEXTHOP: 2586 case IPV6_HOPOPTS: 2587 case IPV6_DSTOPTS: 2588 case IPV6_RTHDRDSTOPTS: 2589 case IPV6_RTHDR: 2590 case IPV6_USE_MIN_MTU: 2591 case IPV6_DONTFRAG: 2592 case IPV6_TCLASS: 2593 case IPV6_PREFER_TEMPADDR: /* XXX: not an RFC3542 option */ 2594 return (ENOPROTOOPT); 2595 } 2596 } 2597 2598 switch (optname) { 2599 case IPV6_2292PKTINFO: 2600 case IPV6_PKTINFO: 2601 { 2602 struct in6_pktinfo *pktinfo; 2603 if (len != sizeof(struct in6_pktinfo)) 2604 return (EINVAL); 2605 pktinfo = (struct in6_pktinfo *)buf; 2606 2607 /* 2608 * An application can clear any sticky IPV6_PKTINFO option by 2609 * doing a "regular" setsockopt with ipi6_addr being 2610 * in6addr_any and ipi6_ifindex being zero. 2611 * [RFC 3542, Section 6] 2612 */ 2613 if (optname == IPV6_PKTINFO && opt->ip6po_pktinfo && 2614 pktinfo->ipi6_ifindex == 0 && 2615 IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) { 2616 ip6_clearpktopts(opt, optname); 2617 break; 2618 } 2619 2620 if (uproto == IPPROTO_TCP && optname == IPV6_PKTINFO && 2621 sticky && !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) { 2622 return (EINVAL); 2623 } 2624 2625 /* validate the interface index if specified. */ 2626 if (pktinfo->ipi6_ifindex > if_index || 2627 pktinfo->ipi6_ifindex < 0) { 2628 return (ENXIO); 2629 } 2630 /* 2631 * Check if the requested source address is indeed a 2632 * unicast address assigned to the node, and can be 2633 * used as the packet's source address. 2634 */ 2635 if (opt->ip6po_pktinfo != NULL && 2636 !IN6_IS_ADDR_UNSPECIFIED(&opt->ip6po_pktinfo->ipi6_addr)) { 2637 struct in6_ifaddr *ia6; 2638 struct sockaddr_in6 sin6; 2639 2640 bzero(&sin6, sizeof(sin6)); 2641 sin6.sin6_len = sizeof(sin6); 2642 sin6.sin6_family = AF_INET6; 2643 sin6.sin6_addr = 2644 opt->ip6po_pktinfo->ipi6_addr; 2645 ia6 = (struct in6_ifaddr *)ifa_ifwithaddr(sin6tosa(&sin6)); 2646 if (ia6 == NULL || 2647 (ia6->ia6_flags & (IN6_IFF_ANYCAST | 2648 IN6_IFF_NOTREADY)) != 0) 2649 return (EADDRNOTAVAIL); 2650 } 2651 2652 /* 2653 * We store the address anyway, and let in6_selectsrc() 2654 * validate the specified address. This is because ipi6_addr 2655 * may not have enough information about its scope zone, and 2656 * we may need additional information (such as outgoing 2657 * interface or the scope zone of a destination address) to 2658 * disambiguate the scope. 2659 * XXX: the delay of the validation may confuse the 2660 * application when it is used as a sticky option. 2661 */ 2662 if (opt->ip6po_pktinfo == NULL) { 2663 opt->ip6po_pktinfo = kmalloc(sizeof(*pktinfo), 2664 M_IP6OPT, M_NOWAIT); 2665 if (opt->ip6po_pktinfo == NULL) 2666 return (ENOBUFS); 2667 } 2668 bcopy(pktinfo, opt->ip6po_pktinfo, sizeof(*pktinfo)); 2669 break; 2670 } 2671 2672 case IPV6_2292HOPLIMIT: 2673 case IPV6_HOPLIMIT: 2674 { 2675 int *hlimp; 2676 2677 /* 2678 * RFC 3542 deprecated the usage of sticky IPV6_HOPLIMIT 2679 * to simplify the ordering among hoplimit options. 2680 */ 2681 if (optname == IPV6_HOPLIMIT && sticky) 2682 return (ENOPROTOOPT); 2683 2684 if (len != sizeof(int)) 2685 return (EINVAL); 2686 hlimp = (int *)buf; 2687 if (*hlimp < -1 || *hlimp > 255) 2688 return (EINVAL); 2689 2690 opt->ip6po_hlim = *hlimp; 2691 break; 2692 } 2693 2694 case IPV6_TCLASS: 2695 { 2696 int tclass; 2697 2698 if (len != sizeof(int)) 2699 return (EINVAL); 2700 tclass = *(int *)buf; 2701 if (tclass < -1 || tclass > 255) 2702 return (EINVAL); 2703 2704 opt->ip6po_tclass = tclass; 2705 break; 2706 } 2707 2708 case IPV6_2292NEXTHOP: 2709 case IPV6_NEXTHOP: 2710 if (!priv) 2711 return (EPERM); 2712 2713 if (len == 0) { /* just remove the option */ 2714 ip6_clearpktopts(opt, IPV6_NEXTHOP); 2715 break; 2716 } 2717 2718 /* check if cmsg_len is large enough for sa_len */ 2719 if (len < sizeof(struct sockaddr) || len < *buf) 2720 return (EINVAL); 2721 2722 switch (((struct sockaddr *)buf)->sa_family) { 2723 case AF_INET6: 2724 { 2725 struct sockaddr_in6 *sa6 = (struct sockaddr_in6 *)buf; 2726 //int error; 2727 2728 if (sa6->sin6_len != sizeof(struct sockaddr_in6)) 2729 return (EINVAL); 2730 2731 if (IN6_IS_ADDR_UNSPECIFIED(&sa6->sin6_addr) || 2732 IN6_IS_ADDR_MULTICAST(&sa6->sin6_addr)) { 2733 return (EINVAL); 2734 } 2735 break; 2736 } 2737 case AF_LINK: /* should eventually be supported */ 2738 default: 2739 return (EAFNOSUPPORT); 2740 } 2741 2742 /* turn off the previous option, then set the new option. */ 2743 ip6_clearpktopts(opt, IPV6_NEXTHOP); 2744 opt->ip6po_nexthop = kmalloc(*buf, M_IP6OPT, M_NOWAIT); 2745 if (opt->ip6po_nexthop == NULL) 2746 return (ENOBUFS); 2747 bcopy(buf, opt->ip6po_nexthop, *buf); 2748 break; 2749 2750 case IPV6_2292HOPOPTS: 2751 case IPV6_HOPOPTS: 2752 { 2753 struct ip6_hbh *hbh; 2754 int hbhlen; 2755 2756 /* 2757 * XXX: We don't allow a non-privileged user to set ANY HbH 2758 * options, since per-option restriction has too much 2759 * overhead. 2760 */ 2761 if (!priv) 2762 return (EPERM); 2763 if (len == 0) { 2764 ip6_clearpktopts(opt, IPV6_HOPOPTS); 2765 break; /* just remove the option */ 2766 } 2767 2768 /* message length validation */ 2769 if (len < sizeof(struct ip6_hbh)) 2770 return (EINVAL); 2771 hbh = (struct ip6_hbh *)buf; 2772 hbhlen = (hbh->ip6h_len + 1) << 3; 2773 if (len != hbhlen) 2774 return (EINVAL); 2775 2776 /* turn off the previous option, then set the new option. */ 2777 ip6_clearpktopts(opt, IPV6_HOPOPTS); 2778 opt->ip6po_hbh = kmalloc(hbhlen, M_IP6OPT, M_NOWAIT); 2779 if (opt->ip6po_hbh == NULL) 2780 return (ENOBUFS); 2781 bcopy(hbh, opt->ip6po_hbh, hbhlen); 2782 2783 break; 2784 } 2785 2786 case IPV6_2292DSTOPTS: 2787 case IPV6_DSTOPTS: 2788 case IPV6_RTHDRDSTOPTS: 2789 { 2790 struct ip6_dest *dest, **newdest = NULL; 2791 int destlen; 2792 if (!priv) 2793 return (EPERM); 2794 2795 if (len == 0) { 2796 ip6_clearpktopts(opt, optname); 2797 break; /* just remove the option */ 2798 } 2799 2800 /* message length validation */ 2801 if (len < sizeof(struct ip6_dest)) 2802 return (EINVAL); 2803 dest = (struct ip6_dest *)buf; 2804 destlen = (dest->ip6d_len + 1) << 3; 2805 if (len != destlen) 2806 return (EINVAL); 2807 2808 /* 2809 * Determine the position that the destination options header 2810 * should be inserted; before or after the routing header. 2811 */ 2812 switch (optname) { 2813 case IPV6_2292DSTOPTS: 2814 /* 2815 * The old advacned API is ambiguous on this point. 2816 * Our approach is to determine the position based 2817 * according to the existence of a routing header. 2818 * Note, however, that this depends on the order of the 2819 * extension headers in the ancillary data; the 1st 2820 * part of the destination options header must appear 2821 * before the routing header in the ancillary data, 2822 * too. 2823 * RFC3542 solved the ambiguity by introducing 2824 * separate ancillary data or option types. 2825 */ 2826 if (opt->ip6po_rthdr == NULL) 2827 newdest = &opt->ip6po_dest1; 2828 else 2829 newdest = &opt->ip6po_dest2; 2830 break; 2831 case IPV6_RTHDRDSTOPTS: 2832 newdest = &opt->ip6po_dest1; 2833 break; 2834 case IPV6_DSTOPTS: 2835 newdest = &opt->ip6po_dest2; 2836 break; 2837 } 2838 2839 /* turn off the previous option, then set the new option. */ 2840 ip6_clearpktopts(opt, optname); 2841 *newdest = kmalloc(destlen, M_IP6OPT, M_NOWAIT); 2842 if (*newdest == NULL) 2843 return (ENOBUFS); 2844 bcopy(dest, *newdest, destlen); 2845 2846 break; 2847 } 2848 2849 case IPV6_2292RTHDR: 2850 case IPV6_RTHDR: 2851 { 2852 struct ip6_rthdr *rth; 2853 int rthlen; 2854 2855 if (len == 0) { 2856 ip6_clearpktopts(opt, IPV6_RTHDR); 2857 break; /* just remove the option */ 2858 } 2859 2860 /* message length validation */ 2861 if (len < sizeof(struct ip6_rthdr)) 2862 return (EINVAL); 2863 rth = (struct ip6_rthdr *)buf; 2864 rthlen = (rth->ip6r_len + 1) << 3; 2865 if (len != rthlen) 2866 return (EINVAL); 2867 2868 switch (rth->ip6r_type) { 2869 default: 2870 return (EINVAL); /* not supported */ 2871 } 2872 2873 /* turn off the previous option */ 2874 ip6_clearpktopts(opt, IPV6_RTHDR); 2875 opt->ip6po_rthdr = kmalloc(rthlen, M_IP6OPT, M_NOWAIT); 2876 if (opt->ip6po_rthdr == NULL) 2877 return (ENOBUFS); 2878 bcopy(rth, opt->ip6po_rthdr, rthlen); 2879 2880 break; 2881 } 2882 2883 case IPV6_USE_MIN_MTU: 2884 if (len != sizeof(int)) 2885 return (EINVAL); 2886 minmtupolicy = *(int *)buf; 2887 if (minmtupolicy != IP6PO_MINMTU_MCASTONLY && 2888 minmtupolicy != IP6PO_MINMTU_DISABLE && 2889 minmtupolicy != IP6PO_MINMTU_ALL) { 2890 return (EINVAL); 2891 } 2892 opt->ip6po_minmtu = minmtupolicy; 2893 break; 2894 2895 case IPV6_DONTFRAG: 2896 if (len != sizeof(int)) 2897 return (EINVAL); 2898 2899 if (uproto == IPPROTO_TCP || *(int *)buf == 0) { 2900 /* 2901 * we ignore this option for TCP sockets. 2902 * (RFC3542 leaves this case unspecified.) 2903 */ 2904 opt->ip6po_flags &= ~IP6PO_DONTFRAG; 2905 } else 2906 opt->ip6po_flags |= IP6PO_DONTFRAG; 2907 break; 2908 2909 case IPV6_PREFER_TEMPADDR: 2910 if (len != sizeof(int)) 2911 return (EINVAL); 2912 preftemp = *(int *)buf; 2913 if (preftemp != IP6PO_TEMPADDR_SYSTEM && 2914 preftemp != IP6PO_TEMPADDR_NOTPREFER && 2915 preftemp != IP6PO_TEMPADDR_PREFER) { 2916 return (EINVAL); 2917 } 2918 opt->ip6po_prefer_tempaddr = preftemp; 2919 break; 2920 2921 default: 2922 return (ENOPROTOOPT); 2923 } /* end of switch */ 2924 2925 return (0); 2926 } 2927 2928 2929 /* 2930 * Set IPv6 outgoing packet options based on advanced API. 2931 */ 2932 int 2933 ip6_setpktoptions(struct mbuf *control, struct ip6_pktopts *opt, 2934 struct ip6_pktopts *stickyopt, int uproto, int priv) 2935 { 2936 struct cmsghdr *cm = NULL; 2937 2938 if (control == NULL || opt == NULL) 2939 return (EINVAL); 2940 2941 init_ip6pktopts(opt); 2942 2943 /* 2944 * XXX: Currently, we assume all the optional information is stored 2945 * in a single mbuf. 2946 */ 2947 if (stickyopt) { 2948 int error; 2949 2950 /* 2951 * If stickyopt is provided, make a local copy of the options 2952 * for this particular packet, then override them by ancillary 2953 * objects. 2954 * XXX: copypktopts() does not copy the cached route to a next 2955 * hop (if any). This is not very good in terms of efficiency, 2956 * but we can allow this since this option should be rarely 2957 * used. 2958 */ 2959 if ((error = copypktopts(opt, stickyopt, M_NOWAIT)) != 0) 2960 return (error); 2961 } 2962 2963 /* 2964 * XXX: Currently, we assume all the optional information is stored 2965 * in a single mbuf. 2966 */ 2967 if (control->m_next) 2968 return (EINVAL); 2969 2970 for (;;) { 2971 int error; 2972 2973 if (control->m_len == 0) 2974 break; 2975 if (control->m_len < sizeof(*cm)) 2976 return EINVAL; 2977 2978 cm = mtod(control, struct cmsghdr *); 2979 if (cm->cmsg_len == 0 || cm->cmsg_len > control->m_len) 2980 return (EINVAL); 2981 if (cm->cmsg_level == IPPROTO_IPV6) { 2982 error = ip6_setpktoption(cm->cmsg_type, CMSG_DATA(cm), 2983 cm->cmsg_len - CMSG_LEN(0), 2984 opt, 0, 1, uproto, priv); 2985 if (error) 2986 return (error); 2987 } 2988 2989 /* 2990 * The cmsg fit, but the aligned step for the next one might 2991 * not. Check the case and terminate normally (allows the 2992 * cmsg_len to not be aligned). 2993 */ 2994 if (CMSG_ALIGN(cm->cmsg_len) >= control->m_len) { 2995 control->m_data += control->m_len; 2996 control->m_len = 0; 2997 break; 2998 } 2999 control->m_data += CMSG_ALIGN(cm->cmsg_len); 3000 control->m_len -= CMSG_ALIGN(cm->cmsg_len); 3001 } 3002 3003 return (0); 3004 } 3005 3006 /* 3007 * Routine called from ip6_output() to loop back a copy of an IP6 multicast 3008 * packet to the input queue of a specified interface. Note that this 3009 * calls the output routine of the loopback "driver", but with an interface 3010 * pointer that might NOT be loif -- easier than replicating that code here. 3011 */ 3012 void 3013 ip6_mloopback(struct ifnet *ifp, struct mbuf *m, struct sockaddr_in6 *dst) 3014 { 3015 struct mbuf *copym; 3016 struct ip6_hdr *ip6; 3017 3018 copym = m_copy(m, 0, M_COPYALL); 3019 if (copym == NULL) 3020 return; 3021 3022 /* 3023 * Make sure to deep-copy IPv6 header portion in case the data 3024 * is in an mbuf cluster, so that we can safely override the IPv6 3025 * header portion later. 3026 */ 3027 if ((copym->m_flags & M_EXT) != 0 || 3028 copym->m_len < sizeof(struct ip6_hdr)) { 3029 copym = m_pullup(copym, sizeof(struct ip6_hdr)); 3030 if (copym == NULL) 3031 return; 3032 } 3033 3034 #ifdef DIAGNOSTIC 3035 if (copym->m_len < sizeof(*ip6)) { 3036 m_freem(copym); 3037 return; 3038 } 3039 #endif 3040 3041 ip6 = mtod(copym, struct ip6_hdr *); 3042 /* 3043 * clear embedded scope identifiers if necessary. 3044 * in6_clearscope will touch the addresses only when necessary. 3045 */ 3046 in6_clearscope(&ip6->ip6_src); 3047 in6_clearscope(&ip6->ip6_dst); 3048 3049 if_simloop(ifp, copym, dst->sin6_family, 0); 3050 } 3051 3052 /* 3053 * Separate the IPv6 header from the payload into its own mbuf. 3054 * 3055 * Returns the new mbuf chain or the original mbuf if no payload. 3056 * Returns NULL if can't allocate new mbuf for header. 3057 */ 3058 static struct mbuf * 3059 ip6_splithdr(struct mbuf *m) 3060 { 3061 struct mbuf *mh; 3062 3063 if (m->m_len <= sizeof(struct ip6_hdr)) /* no payload */ 3064 return (m); 3065 3066 MGETHDR(mh, M_NOWAIT, MT_HEADER); 3067 if (mh == NULL) 3068 return (NULL); 3069 mh->m_len = sizeof(struct ip6_hdr); 3070 M_MOVE_PKTHDR(mh, m); 3071 MH_ALIGN(mh, sizeof(struct ip6_hdr)); 3072 bcopy(mtod(m, caddr_t), mtod(mh, caddr_t), sizeof(struct ip6_hdr)); 3073 m->m_data += sizeof(struct ip6_hdr); 3074 m->m_len -= sizeof(struct ip6_hdr); 3075 mh->m_next = m; 3076 return (mh); 3077 } 3078 3079 /* 3080 * Compute IPv6 extension header length. 3081 */ 3082 int 3083 ip6_optlen(struct in6pcb *in6p) 3084 { 3085 int len; 3086 3087 if (!in6p->in6p_outputopts) 3088 return 0; 3089 3090 len = 0; 3091 #define elen(x) \ 3092 (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0) 3093 3094 len += elen(in6p->in6p_outputopts->ip6po_hbh); 3095 if (in6p->in6p_outputopts->ip6po_rthdr) 3096 /* dest1 is valid with rthdr only */ 3097 len += elen(in6p->in6p_outputopts->ip6po_dest1); 3098 len += elen(in6p->in6p_outputopts->ip6po_rthdr); 3099 len += elen(in6p->in6p_outputopts->ip6po_dest2); 3100 return len; 3101 #undef elen 3102 } 3103