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/caps.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 || 1156 caps_priv_check_td(td, SYSCAP_RESTRICTEDROOT)) ? 1157 0 : 1; 1158 1159 if (level == IPPROTO_IPV6) { 1160 switch (op) { 1161 1162 case SOPT_SET: 1163 switch (optname) { 1164 case IPV6_2292PKTOPTIONS: 1165 #ifdef IPV6_PKTOPTIONS 1166 case IPV6_PKTOPTIONS: 1167 #endif 1168 { 1169 struct mbuf *m; 1170 1171 error = soopt_getm(sopt, &m); /* XXX */ 1172 if (error != 0) 1173 break; 1174 soopt_to_mbuf(sopt, m); /* XXX */ 1175 error = ip6_pcbopts(&in6p->in6p_outputopts, 1176 m, so, sopt); 1177 m_freem(m); /* XXX */ 1178 break; 1179 } 1180 1181 /* 1182 * Use of some Hop-by-Hop options or some 1183 * Destination options, might require special 1184 * privilege. That is, normal applications 1185 * (without special privilege) might be forbidden 1186 * from setting certain options in outgoing packets, 1187 * and might never see certain options in received 1188 * packets. [RFC 2292 Section 6] 1189 * KAME specific note: 1190 * KAME prevents non-privileged users from sending or 1191 * receiving ANY hbh/dst options in order to avoid 1192 * overhead of parsing options in the kernel. 1193 */ 1194 case IPV6_RECVHOPOPTS: 1195 case IPV6_RECVDSTOPTS: 1196 case IPV6_RECVRTHDRDSTOPTS: 1197 if (!privileged) 1198 return (EPERM); 1199 case IPV6_RECVPKTINFO: 1200 case IPV6_RECVHOPLIMIT: 1201 case IPV6_RECVRTHDR: 1202 case IPV6_RECVPATHMTU: 1203 case IPV6_RECVTCLASS: 1204 case IPV6_AUTOFLOWLABEL: 1205 case IPV6_HOPLIMIT: 1206 /* FALLTHROUGH */ 1207 case IPV6_UNICAST_HOPS: 1208 1209 case IPV6_V6ONLY: 1210 if (optlen != sizeof(int)) { 1211 error = EINVAL; 1212 break; 1213 } 1214 error = soopt_to_kbuf(sopt, &optval, 1215 sizeof optval, sizeof optval); 1216 if (error) 1217 break; 1218 switch (optname) { 1219 1220 case IPV6_UNICAST_HOPS: 1221 if (optval < -1 || optval >= 256) 1222 error = EINVAL; 1223 else { 1224 /* -1 = kernel default */ 1225 in6p->in6p_hops = optval; 1226 } 1227 break; 1228 #define OPTSET(bit) \ 1229 do { \ 1230 if (optval) \ 1231 in6p->in6p_flags |= (bit); \ 1232 else \ 1233 in6p->in6p_flags &= ~(bit); \ 1234 } while (0) 1235 #define OPTBIT(bit) (in6p->in6p_flags & (bit) ? 1 : 0) 1236 /* 1237 * Although changed to RFC3542, It's better to also support RFC2292 API 1238 */ 1239 #define OPTSET2292(bit) \ 1240 do { \ 1241 in6p->in6p_flags |= IN6P_RFC2292; \ 1242 if (optval) \ 1243 in6p->in6p_flags |= (bit); \ 1244 else \ 1245 in6p->in6p_flags &= ~(bit); \ 1246 } while (/*CONSTCOND*/ 0) 1247 1248 case IPV6_RECVPKTINFO: 1249 /* cannot mix with RFC2292 */ 1250 if (OPTBIT(IN6P_RFC2292)) { 1251 error = EINVAL; 1252 break; 1253 } 1254 OPTSET(IN6P_PKTINFO); 1255 break; 1256 1257 case IPV6_HOPLIMIT: 1258 { 1259 struct ip6_pktopts **optp; 1260 1261 /* cannot mix with RFC2292 */ 1262 if (OPTBIT(IN6P_RFC2292)) { 1263 error = EINVAL; 1264 break; 1265 } 1266 optp = &in6p->in6p_outputopts; 1267 error = ip6_pcbopt(IPV6_HOPLIMIT, 1268 (u_char *)&optval, sizeof(optval), 1269 optp, uproto); 1270 break; 1271 } 1272 1273 case IPV6_RECVHOPLIMIT: 1274 /* cannot mix with RFC2292 */ 1275 if (OPTBIT(IN6P_RFC2292)) { 1276 error = EINVAL; 1277 break; 1278 } 1279 OPTSET(IN6P_HOPLIMIT); 1280 break; 1281 1282 case IPV6_RECVHOPOPTS: 1283 /* cannot mix with RFC2292 */ 1284 if (OPTBIT(IN6P_RFC2292)) { 1285 error = EINVAL; 1286 break; 1287 } 1288 OPTSET(IN6P_HOPOPTS); 1289 break; 1290 1291 case IPV6_RECVDSTOPTS: 1292 /* cannot mix with RFC2292 */ 1293 if (OPTBIT(IN6P_RFC2292)) { 1294 error = EINVAL; 1295 break; 1296 } 1297 OPTSET(IN6P_DSTOPTS); 1298 break; 1299 1300 case IPV6_RECVRTHDRDSTOPTS: 1301 /* cannot mix with RFC2292 */ 1302 if (OPTBIT(IN6P_RFC2292)) { 1303 error = EINVAL; 1304 break; 1305 } 1306 OPTSET(IN6P_RTHDRDSTOPTS); 1307 break; 1308 1309 case IPV6_RECVRTHDR: 1310 /* cannot mix with RFC2292 */ 1311 if (OPTBIT(IN6P_RFC2292)) { 1312 error = EINVAL; 1313 break; 1314 } 1315 OPTSET(IN6P_RTHDR); 1316 break; 1317 1318 case IPV6_RECVPATHMTU: 1319 /* 1320 * We ignore this option for TCP 1321 * sockets. 1322 * (RFC3542 leaves this case 1323 * unspecified.) 1324 */ 1325 if (uproto != IPPROTO_TCP) 1326 OPTSET(IN6P_MTU); 1327 break; 1328 1329 case IPV6_RECVTCLASS: 1330 /* cannot mix with RFC2292 XXX */ 1331 if (OPTBIT(IN6P_RFC2292)) { 1332 error = EINVAL; 1333 break; 1334 } 1335 OPTSET(IN6P_TCLASS); 1336 break; 1337 1338 case IPV6_AUTOFLOWLABEL: 1339 OPTSET(IN6P_AUTOFLOWLABEL); 1340 break; 1341 1342 case IPV6_V6ONLY: 1343 /* 1344 * make setsockopt(IPV6_V6ONLY) 1345 * available only prior to bind(2). 1346 */ 1347 if (in6p->in6p_lport || 1348 !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr)) 1349 { 1350 error = EINVAL; 1351 break; 1352 } 1353 1354 /* 1355 * Since we don't support v6->v4 1356 * mapping any more this option does 1357 * nothing. But apparently some 1358 * ports and libraries (e.g. libuv) 1359 * actually try to set the value to 1360 * 0 so just silently ignore the value 1361 * entirely. 1362 * 1363 * (also fixes named which uses libuv, 1364 * and a few other apps) 1365 */ 1366 #if 0 1367 if (!optval) { 1368 /* Don't allow v4-mapped */ 1369 error = EOPNOTSUPP; 1370 } 1371 #endif 1372 break; 1373 } 1374 break; 1375 1376 case IPV6_TCLASS: 1377 case IPV6_DONTFRAG: 1378 case IPV6_USE_MIN_MTU: 1379 case IPV6_PREFER_TEMPADDR: 1380 if (optlen != sizeof(optval)) { 1381 error = EINVAL; 1382 break; 1383 } 1384 error = soopt_to_kbuf(sopt, &optval, 1385 sizeof optval, sizeof optval); 1386 if (error) 1387 break; 1388 { 1389 struct ip6_pktopts **optp; 1390 optp = &in6p->in6p_outputopts; 1391 error = ip6_pcbopt(optname, 1392 (u_char *)&optval, sizeof(optval), 1393 optp, uproto); 1394 break; 1395 } 1396 1397 case IPV6_2292PKTINFO: 1398 case IPV6_2292HOPLIMIT: 1399 case IPV6_2292HOPOPTS: 1400 case IPV6_2292DSTOPTS: 1401 case IPV6_2292RTHDR: 1402 /* RFC 2292 */ 1403 if (optlen != sizeof(int)) { 1404 error = EINVAL; 1405 break; 1406 } 1407 error = soopt_to_kbuf(sopt, &optval, 1408 sizeof optval, sizeof optval); 1409 if (error) 1410 break; 1411 switch (optname) { 1412 case IPV6_2292PKTINFO: 1413 OPTSET2292(IN6P_PKTINFO); 1414 break; 1415 case IPV6_2292HOPLIMIT: 1416 OPTSET2292(IN6P_HOPLIMIT); 1417 break; 1418 case IPV6_2292HOPOPTS: 1419 /* 1420 * Check super-user privilege. 1421 * See comments for IPV6_RECVHOPOPTS. 1422 */ 1423 if (!privileged) 1424 return (EPERM); 1425 OPTSET2292(IN6P_HOPOPTS); 1426 break; 1427 case IPV6_2292DSTOPTS: 1428 if (!privileged) 1429 return (EPERM); 1430 OPTSET2292(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); /* XXX */ 1431 break; 1432 case IPV6_2292RTHDR: 1433 OPTSET2292(IN6P_RTHDR); 1434 break; 1435 } 1436 break; 1437 1438 case IPV6_PKTINFO: 1439 case IPV6_HOPOPTS: 1440 case IPV6_RTHDR: 1441 case IPV6_DSTOPTS: 1442 case IPV6_RTHDRDSTOPTS: 1443 case IPV6_NEXTHOP: 1444 { 1445 /* 1446 * New advanced API (RFC3542) 1447 */ 1448 u_char *optbuf; 1449 u_char optbuf_storage[MCLBYTES]; 1450 int optlen; 1451 struct ip6_pktopts **optp; 1452 1453 /* cannot mix with RFC2292 */ 1454 if (OPTBIT(IN6P_RFC2292)) { 1455 error = EINVAL; 1456 break; 1457 } 1458 1459 /* 1460 * We only ensure valsize is not too large 1461 * here. Further validation will be done 1462 * later. 1463 */ 1464 error = soopt_to_kbuf(sopt, optbuf_storage, 1465 sizeof(optbuf_storage), 0); 1466 if (error) 1467 break; 1468 optlen = sopt->sopt_valsize; 1469 optbuf = optbuf_storage; 1470 optp = &in6p->in6p_outputopts; 1471 error = ip6_pcbopt(optname, optbuf, optlen, 1472 optp, uproto); 1473 break; 1474 } 1475 #undef OPTSET 1476 1477 case IPV6_MULTICAST_IF: 1478 case IPV6_MULTICAST_HOPS: 1479 case IPV6_MULTICAST_LOOP: 1480 case IPV6_JOIN_GROUP: 1481 case IPV6_LEAVE_GROUP: 1482 { 1483 struct mbuf *m; 1484 1485 if (sopt->sopt_valsize > MLEN) { 1486 error = EMSGSIZE; 1487 break; 1488 } 1489 /* XXX */ 1490 MGET(m, sopt->sopt_td ? M_WAITOK : M_NOWAIT, MT_HEADER); 1491 if (m == NULL) { 1492 error = ENOBUFS; 1493 break; 1494 } 1495 m->m_len = sopt->sopt_valsize; 1496 error = soopt_to_kbuf(sopt, mtod(m, char *), 1497 m->m_len, m->m_len); 1498 error = ip6_setmoptions(sopt->sopt_name, 1499 &in6p->in6p_moptions, 1500 m); 1501 m_free(m); 1502 } 1503 break; 1504 1505 case IPV6_PORTRANGE: 1506 error = soopt_to_kbuf(sopt, &optval, 1507 sizeof optval, sizeof optval); 1508 if (error) 1509 break; 1510 1511 switch (optval) { 1512 case IPV6_PORTRANGE_DEFAULT: 1513 in6p->in6p_flags &= ~(IN6P_LOWPORT); 1514 in6p->in6p_flags &= ~(IN6P_HIGHPORT); 1515 break; 1516 1517 case IPV6_PORTRANGE_HIGH: 1518 in6p->in6p_flags &= ~(IN6P_LOWPORT); 1519 in6p->in6p_flags |= IN6P_HIGHPORT; 1520 break; 1521 1522 case IPV6_PORTRANGE_LOW: 1523 in6p->in6p_flags &= ~(IN6P_HIGHPORT); 1524 in6p->in6p_flags |= IN6P_LOWPORT; 1525 break; 1526 1527 default: 1528 error = EINVAL; 1529 break; 1530 } 1531 break; 1532 1533 case IPV6_FW_ADD: 1534 case IPV6_FW_DEL: 1535 case IPV6_FW_FLUSH: 1536 case IPV6_FW_ZERO: 1537 { 1538 struct mbuf *m; 1539 struct mbuf **mp = &m; 1540 1541 if (ip6_fw_ctl_ptr == NULL) 1542 return EINVAL; 1543 /* XXX */ 1544 if ((error = soopt_getm(sopt, &m)) != 0) 1545 break; 1546 /* XXX */ 1547 soopt_to_mbuf(sopt, m); 1548 error = (*ip6_fw_ctl_ptr)(optname, mp); 1549 m = *mp; 1550 } 1551 break; 1552 1553 default: 1554 error = ENOPROTOOPT; 1555 break; 1556 } 1557 break; 1558 1559 case SOPT_GET: 1560 switch (optname) { 1561 case IPV6_2292PKTOPTIONS: 1562 #ifdef IPV6_PKTOPTIONS 1563 case IPV6_PKTOPTIONS: 1564 #endif 1565 /* 1566 * RFC3542 (effectively) deprecated the 1567 * semantics of the 2292-style pktoptions. 1568 * Since it was not reliable in nature (i.e., 1569 * applications had to expect the lack of some 1570 * information after all), it would make sense 1571 * to simplify this part by always returning 1572 * empty data. 1573 */ 1574 if (in6p->in6p_options) { 1575 struct mbuf *m; 1576 m = m_copym(in6p->in6p_options, 1577 0, M_COPYALL, M_WAITOK); 1578 error = soopt_from_mbuf(sopt, m); 1579 if (error == 0) 1580 m_freem(m); 1581 } else 1582 sopt->sopt_valsize = 0; 1583 break; 1584 1585 case IPV6_RECVHOPOPTS: 1586 case IPV6_RECVDSTOPTS: 1587 case IPV6_RECVRTHDRDSTOPTS: 1588 case IPV6_UNICAST_HOPS: 1589 case IPV6_RECVPKTINFO: 1590 case IPV6_RECVHOPLIMIT: 1591 case IPV6_RECVRTHDR: 1592 case IPV6_RECVPATHMTU: 1593 case IPV6_RECVTCLASS: 1594 case IPV6_AUTOFLOWLABEL: 1595 case IPV6_V6ONLY: 1596 case IPV6_PORTRANGE: 1597 switch (optname) { 1598 1599 case IPV6_RECVHOPOPTS: 1600 optval = OPTBIT(IN6P_HOPOPTS); 1601 break; 1602 1603 case IPV6_RECVDSTOPTS: 1604 optval = OPTBIT(IN6P_DSTOPTS); 1605 break; 1606 1607 case IPV6_RECVRTHDRDSTOPTS: 1608 optval = OPTBIT(IN6P_RTHDRDSTOPTS); 1609 break; 1610 1611 case IPV6_RECVPKTINFO: 1612 optval = OPTBIT(IN6P_PKTINFO); 1613 break; 1614 1615 case IPV6_RECVHOPLIMIT: 1616 optval = OPTBIT(IN6P_HOPLIMIT); 1617 break; 1618 1619 case IPV6_RECVRTHDR: 1620 optval = OPTBIT(IN6P_RTHDR); 1621 break; 1622 1623 case IPV6_RECVPATHMTU: 1624 optval = OPTBIT(IN6P_MTU); 1625 break; 1626 1627 case IPV6_RECVTCLASS: 1628 optval = OPTBIT(IN6P_TCLASS); 1629 break; 1630 1631 case IPV6_AUTOFLOWLABEL: 1632 optval = OPTBIT(IN6P_AUTOFLOWLABEL); 1633 break; 1634 1635 1636 case IPV6_UNICAST_HOPS: 1637 optval = in6p->in6p_hops; 1638 break; 1639 1640 case IPV6_V6ONLY: 1641 optval = 1; 1642 break; 1643 1644 case IPV6_PORTRANGE: 1645 { 1646 int flags; 1647 flags = in6p->in6p_flags; 1648 if (flags & IN6P_HIGHPORT) 1649 optval = IPV6_PORTRANGE_HIGH; 1650 else if (flags & IN6P_LOWPORT) 1651 optval = IPV6_PORTRANGE_LOW; 1652 else 1653 optval = 0; 1654 break; 1655 } 1656 } 1657 soopt_from_kbuf(sopt, &optval, 1658 sizeof optval); 1659 break; 1660 1661 case IPV6_PATHMTU: 1662 { 1663 u_long pmtu = 0; 1664 struct ip6_mtuinfo mtuinfo; 1665 struct route_in6 sro; 1666 1667 bzero(&sro, sizeof(sro)); 1668 1669 if (!(so->so_state & SS_ISCONNECTED)) 1670 return (ENOTCONN); 1671 /* 1672 * XXX: we dot not consider the case of source 1673 * routing, or optional information to specify 1674 * the outgoing interface. 1675 */ 1676 error = ip6_getpmtu(&sro, NULL, NULL, 1677 &in6p->in6p_faddr, &pmtu, NULL); 1678 if (sro.ro_rt) 1679 RTFREE(sro.ro_rt); 1680 if (error) 1681 break; 1682 if (pmtu > IPV6_MAXPACKET) 1683 pmtu = IPV6_MAXPACKET; 1684 1685 bzero(&mtuinfo, sizeof(mtuinfo)); 1686 mtuinfo.ip6m_mtu = (u_int32_t)pmtu; 1687 optdata = (void *)&mtuinfo; 1688 optdatalen = sizeof(mtuinfo); 1689 soopt_from_kbuf(sopt, optdata, 1690 optdatalen); 1691 break; 1692 } 1693 1694 case IPV6_2292PKTINFO: 1695 case IPV6_2292HOPLIMIT: 1696 case IPV6_2292HOPOPTS: 1697 case IPV6_2292RTHDR: 1698 case IPV6_2292DSTOPTS: 1699 if (optname == IPV6_2292HOPOPTS || 1700 optname == IPV6_2292DSTOPTS || 1701 !privileged) 1702 return (EPERM); 1703 switch (optname) { 1704 case IPV6_2292PKTINFO: 1705 optval = OPTBIT(IN6P_PKTINFO); 1706 break; 1707 case IPV6_2292HOPLIMIT: 1708 optval = OPTBIT(IN6P_HOPLIMIT); 1709 break; 1710 case IPV6_2292HOPOPTS: 1711 if (!privileged) 1712 return (EPERM); 1713 optval = OPTBIT(IN6P_HOPOPTS); 1714 break; 1715 case IPV6_2292RTHDR: 1716 optval = OPTBIT(IN6P_RTHDR); 1717 break; 1718 case IPV6_2292DSTOPTS: 1719 if (!privileged) 1720 return (EPERM); 1721 optval = OPTBIT(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); 1722 break; 1723 } 1724 soopt_from_kbuf(sopt, &optval, 1725 sizeof optval); 1726 break; 1727 1728 case IPV6_PKTINFO: 1729 case IPV6_HOPOPTS: 1730 case IPV6_RTHDR: 1731 case IPV6_DSTOPTS: 1732 case IPV6_RTHDRDSTOPTS: 1733 case IPV6_NEXTHOP: 1734 case IPV6_TCLASS: 1735 case IPV6_DONTFRAG: 1736 case IPV6_USE_MIN_MTU: 1737 case IPV6_PREFER_TEMPADDR: 1738 error = ip6_getpcbopt(in6p->in6p_outputopts, 1739 optname, sopt); 1740 break; 1741 1742 case IPV6_MULTICAST_IF: 1743 case IPV6_MULTICAST_HOPS: 1744 case IPV6_MULTICAST_LOOP: 1745 case IPV6_JOIN_GROUP: 1746 case IPV6_LEAVE_GROUP: 1747 { 1748 struct mbuf *m; 1749 error = ip6_getmoptions(sopt->sopt_name, 1750 in6p->in6p_moptions, &m); 1751 if (error == 0) { 1752 soopt_from_kbuf(sopt, 1753 mtod(m, char *), m->m_len); 1754 } 1755 m_freem(m); 1756 } 1757 break; 1758 1759 case IPV6_FW_GET: 1760 { 1761 struct mbuf *m; 1762 struct mbuf **mp = &m; 1763 1764 if (ip6_fw_ctl_ptr == NULL) 1765 { 1766 return EINVAL; 1767 } 1768 error = (*ip6_fw_ctl_ptr)(optname, mp); 1769 if (error == 0) 1770 error = soopt_from_mbuf(sopt, m); /* XXX */ 1771 if (error == 0 && m != NULL) 1772 m_freem(m); 1773 } 1774 break; 1775 1776 default: 1777 error = ENOPROTOOPT; 1778 break; 1779 } 1780 break; 1781 } 1782 } else { 1783 error = EINVAL; 1784 } 1785 return (error); 1786 } 1787 1788 int 1789 ip6_raw_ctloutput(struct socket *so, struct sockopt *sopt) 1790 { 1791 int error = 0, optval, optlen; 1792 const int icmp6off = offsetof(struct icmp6_hdr, icmp6_cksum); 1793 struct in6pcb *in6p = sotoin6pcb(so); 1794 int level, op, optname; 1795 1796 if (sopt) { 1797 level = sopt->sopt_level; 1798 op = sopt->sopt_dir; 1799 optname = sopt->sopt_name; 1800 optlen = sopt->sopt_valsize; 1801 } else 1802 panic("ip6_raw_ctloutput: arg soopt is NULL"); 1803 1804 if (level != IPPROTO_IPV6) { 1805 return (EINVAL); 1806 } 1807 1808 switch (optname) { 1809 case IPV6_CHECKSUM: 1810 /* 1811 * For ICMPv6 sockets, no modification allowed for checksum 1812 * offset, permit "no change" values to help existing apps. 1813 * 1814 * RFC3542 says: "An attempt to set IPV6_CHECKSUM 1815 * for an ICMPv6 socket will fail." 1816 * The current behavior does not meet RFC3542. 1817 */ 1818 switch (op) { 1819 case SOPT_SET: 1820 if (optlen != sizeof(int)) { 1821 error = EINVAL; 1822 break; 1823 } 1824 error = soopt_to_kbuf(sopt, &optval, 1825 sizeof optval, sizeof optval); 1826 if (error) 1827 break; 1828 if ((optval % 2) != 0) { 1829 /* the API assumes even offset values */ 1830 error = EINVAL; 1831 } else if (so->so_proto->pr_protocol == 1832 IPPROTO_ICMPV6) { 1833 if (optval != icmp6off) 1834 error = EINVAL; 1835 } else 1836 in6p->in6p_cksum = optval; 1837 break; 1838 1839 case SOPT_GET: 1840 if (so->so_proto->pr_protocol == IPPROTO_ICMPV6) 1841 optval = icmp6off; 1842 else 1843 optval = in6p->in6p_cksum; 1844 1845 soopt_from_kbuf(sopt, &optval, sizeof(optval)); 1846 break; 1847 1848 default: 1849 error = EINVAL; 1850 break; 1851 } 1852 break; 1853 1854 default: 1855 error = ENOPROTOOPT; 1856 break; 1857 } 1858 1859 return (error); 1860 } 1861 1862 /* 1863 * Set up IP6 options in pcb for insertion in output packets or 1864 * specifying behavior of outgoing packets. 1865 */ 1866 static int 1867 ip6_pcbopts(struct ip6_pktopts **pktopt, struct mbuf *m, 1868 struct socket *so, struct sockopt *sopt) 1869 { 1870 int priv = 0; 1871 struct ip6_pktopts *opt = *pktopt; 1872 int error = 0; 1873 1874 /* turn off any old options. */ 1875 if (opt) { 1876 #ifdef DIAGNOSTIC 1877 if (opt->ip6po_pktinfo || opt->ip6po_nexthop || 1878 opt->ip6po_hbh || opt->ip6po_dest1 || opt->ip6po_dest2 || 1879 opt->ip6po_rhinfo.ip6po_rhi_rthdr) 1880 kprintf("ip6_pcbopts: all specified options are cleared.\n"); 1881 #endif 1882 ip6_clearpktopts(opt, -1); 1883 } else 1884 opt = kmalloc(sizeof(*opt), M_IP6OPT, M_WAITOK); 1885 *pktopt = NULL; 1886 1887 if (!m || m->m_len == 0) { 1888 /* 1889 * Only turning off any previous options, regardless of 1890 * whether the opt is just created or given. 1891 */ 1892 kfree(opt, M_IP6OPT); 1893 return (0); 1894 } 1895 1896 /* set options specified by user. */ 1897 if ((error = ip6_setpktoptions(m, opt, NULL, so->so_proto->pr_protocol, priv)) != 0) { 1898 ip6_clearpktopts(opt, -1); /* XXX: discard all options */ 1899 kfree(opt, M_IP6OPT); 1900 return (error); 1901 } 1902 *pktopt = opt; 1903 return (0); 1904 } 1905 1906 1907 /* 1908 * Below three functions are introduced by merge to RFC3542 1909 */ 1910 1911 static int 1912 ip6_getpcbopt(struct ip6_pktopts *pktopt, int optname, struct sockopt *sopt) 1913 { 1914 void *optdata = NULL; 1915 int optdatalen = 0; 1916 struct ip6_ext *ip6e; 1917 int error = 0; 1918 struct in6_pktinfo null_pktinfo; 1919 int deftclass = 0, on; 1920 int defminmtu = IP6PO_MINMTU_MCASTONLY; 1921 int defpreftemp = IP6PO_TEMPADDR_SYSTEM; 1922 1923 switch (optname) { 1924 case IPV6_PKTINFO: 1925 if (pktopt && pktopt->ip6po_pktinfo) 1926 optdata = (void *)pktopt->ip6po_pktinfo; 1927 else { 1928 /* XXX: we don't have to do this every time... */ 1929 bzero(&null_pktinfo, sizeof(null_pktinfo)); 1930 optdata = (void *)&null_pktinfo; 1931 } 1932 optdatalen = sizeof(struct in6_pktinfo); 1933 break; 1934 case IPV6_TCLASS: 1935 if (pktopt && pktopt->ip6po_tclass >= 0) 1936 optdata = (void *)&pktopt->ip6po_tclass; 1937 else 1938 optdata = (void *)&deftclass; 1939 optdatalen = sizeof(int); 1940 break; 1941 case IPV6_HOPOPTS: 1942 if (pktopt && pktopt->ip6po_hbh) { 1943 optdata = (void *)pktopt->ip6po_hbh; 1944 ip6e = (struct ip6_ext *)pktopt->ip6po_hbh; 1945 optdatalen = (ip6e->ip6e_len + 1) << 3; 1946 } 1947 break; 1948 case IPV6_RTHDR: 1949 if (pktopt && pktopt->ip6po_rthdr) { 1950 optdata = (void *)pktopt->ip6po_rthdr; 1951 ip6e = (struct ip6_ext *)pktopt->ip6po_rthdr; 1952 optdatalen = (ip6e->ip6e_len + 1) << 3; 1953 } 1954 break; 1955 case IPV6_RTHDRDSTOPTS: 1956 if (pktopt && pktopt->ip6po_dest1) { 1957 optdata = (void *)pktopt->ip6po_dest1; 1958 ip6e = (struct ip6_ext *)pktopt->ip6po_dest1; 1959 optdatalen = (ip6e->ip6e_len + 1) << 3; 1960 } 1961 break; 1962 case IPV6_DSTOPTS: 1963 if (pktopt && pktopt->ip6po_dest2) { 1964 optdata = (void *)pktopt->ip6po_dest2; 1965 ip6e = (struct ip6_ext *)pktopt->ip6po_dest2; 1966 optdatalen = (ip6e->ip6e_len + 1) << 3; 1967 } 1968 break; 1969 case IPV6_NEXTHOP: 1970 if (pktopt && pktopt->ip6po_nexthop) { 1971 optdata = (void *)pktopt->ip6po_nexthop; 1972 optdatalen = pktopt->ip6po_nexthop->sa_len; 1973 } 1974 break; 1975 case IPV6_USE_MIN_MTU: 1976 if (pktopt) 1977 optdata = (void *)&pktopt->ip6po_minmtu; 1978 else 1979 optdata = (void *)&defminmtu; 1980 optdatalen = sizeof(int); 1981 break; 1982 case IPV6_DONTFRAG: 1983 if (pktopt && ((pktopt->ip6po_flags) & IP6PO_DONTFRAG)) 1984 on = 1; 1985 else 1986 on = 0; 1987 optdata = (void *)&on; 1988 optdatalen = sizeof(on); 1989 break; 1990 case IPV6_PREFER_TEMPADDR: 1991 if (pktopt) 1992 optdata = (void *)&pktopt->ip6po_prefer_tempaddr; 1993 else 1994 optdata = (void *)&defpreftemp; 1995 optdatalen = sizeof(int); 1996 break; 1997 default: /* should not happen */ 1998 #ifdef DIAGNOSTIC 1999 panic("ip6_getpcbopt: unexpected option"); 2000 #endif 2001 return (ENOPROTOOPT); 2002 } 2003 2004 soopt_from_kbuf(sopt, optdata, optdatalen); 2005 2006 return (error); 2007 } 2008 2009 /* 2010 * initialize ip6_pktopts. beware that there are non-zero default values in 2011 * the struct. 2012 */ 2013 2014 static int 2015 ip6_pcbopt(int optname, u_char *buf, int len, struct ip6_pktopts **pktopt, int uproto) 2016 { 2017 struct ip6_pktopts *opt; 2018 int priv =0; 2019 if (*pktopt == NULL) { 2020 *pktopt = kmalloc(sizeof(*opt), M_IP6OPT, M_WAITOK); 2021 init_ip6pktopts(*pktopt); 2022 } 2023 opt = *pktopt; 2024 2025 return (ip6_setpktoption(optname, buf, len, opt, 1, 0, uproto, priv)); 2026 } 2027 2028 /* 2029 * initialize ip6_pktopts. beware that there are non-zero default values in 2030 * the struct. 2031 */ 2032 void 2033 init_ip6pktopts(struct ip6_pktopts *opt) 2034 { 2035 2036 bzero(opt, sizeof(*opt)); 2037 opt->ip6po_hlim = -1; /* -1 means default hop limit */ 2038 opt->ip6po_tclass = -1; /* -1 means default traffic class */ 2039 opt->ip6po_minmtu = IP6PO_MINMTU_MCASTONLY; 2040 opt->ip6po_prefer_tempaddr = IP6PO_TEMPADDR_SYSTEM; 2041 } 2042 2043 void 2044 ip6_clearpktopts(struct ip6_pktopts *pktopt, int optname) 2045 { 2046 if (pktopt == NULL) 2047 return; 2048 2049 if (optname == -1 || optname == IPV6_PKTINFO) { 2050 if (pktopt->ip6po_pktinfo) 2051 kfree(pktopt->ip6po_pktinfo, M_IP6OPT); 2052 pktopt->ip6po_pktinfo = NULL; 2053 } 2054 if (optname == -1 || optname == IPV6_HOPLIMIT) 2055 pktopt->ip6po_hlim = -1; 2056 if (optname == -1 || optname == IPV6_TCLASS) 2057 pktopt->ip6po_tclass = -1; 2058 if (optname == -1 || optname == IPV6_NEXTHOP) { 2059 if (pktopt->ip6po_nextroute.ro_rt) { 2060 RTFREE(pktopt->ip6po_nextroute.ro_rt); 2061 pktopt->ip6po_nextroute.ro_rt = NULL; 2062 } 2063 if (pktopt->ip6po_nexthop) 2064 kfree(pktopt->ip6po_nexthop, M_IP6OPT); 2065 pktopt->ip6po_nexthop = NULL; 2066 } 2067 if (optname == -1 || optname == IPV6_HOPOPTS) { 2068 if (pktopt->ip6po_hbh) 2069 kfree(pktopt->ip6po_hbh, M_IP6OPT); 2070 pktopt->ip6po_hbh = NULL; 2071 } 2072 if (optname == -1 || optname == IPV6_RTHDRDSTOPTS) { 2073 if (pktopt->ip6po_dest1) 2074 kfree(pktopt->ip6po_dest1, M_IP6OPT); 2075 pktopt->ip6po_dest1 = NULL; 2076 } 2077 if (optname == -1 || optname == IPV6_RTHDR) { 2078 if (pktopt->ip6po_rhinfo.ip6po_rhi_rthdr) 2079 kfree(pktopt->ip6po_rhinfo.ip6po_rhi_rthdr, M_IP6OPT); 2080 pktopt->ip6po_rhinfo.ip6po_rhi_rthdr = NULL; 2081 if (pktopt->ip6po_route.ro_rt) { 2082 RTFREE(pktopt->ip6po_route.ro_rt); 2083 pktopt->ip6po_route.ro_rt = NULL; 2084 } 2085 } 2086 if (optname == -1 || optname == IPV6_DSTOPTS) { 2087 if (pktopt->ip6po_dest2) 2088 kfree(pktopt->ip6po_dest2, M_IP6OPT); 2089 pktopt->ip6po_dest2 = NULL; 2090 } 2091 } 2092 2093 #define PKTOPT_EXTHDRCPY(type) \ 2094 do {\ 2095 if (src->type) {\ 2096 int hlen = (((struct ip6_ext *)src->type)->ip6e_len + 1) << 3;\ 2097 dst->type = kmalloc(hlen, M_IP6OPT, canwait);\ 2098 if (dst->type == NULL)\ 2099 goto bad;\ 2100 bcopy(src->type, dst->type, hlen);\ 2101 }\ 2102 } while (0) 2103 2104 struct ip6_pktopts * 2105 ip6_copypktopts(struct ip6_pktopts *src, int canwait) 2106 { 2107 struct ip6_pktopts *dst; 2108 2109 if (src == NULL) { 2110 kprintf("ip6_clearpktopts: invalid argument\n"); 2111 return (NULL); 2112 } 2113 2114 dst = kmalloc(sizeof(*dst), M_IP6OPT, canwait | M_ZERO); 2115 if (dst == NULL) 2116 return (NULL); 2117 2118 dst->ip6po_hlim = src->ip6po_hlim; 2119 if (src->ip6po_pktinfo) { 2120 dst->ip6po_pktinfo = kmalloc(sizeof(*dst->ip6po_pktinfo), 2121 M_IP6OPT, canwait); 2122 if (dst->ip6po_pktinfo == NULL) 2123 goto bad; 2124 *dst->ip6po_pktinfo = *src->ip6po_pktinfo; 2125 } 2126 if (src->ip6po_nexthop) { 2127 dst->ip6po_nexthop = kmalloc(src->ip6po_nexthop->sa_len, 2128 M_IP6OPT, canwait); 2129 if (dst->ip6po_nexthop == NULL) 2130 goto bad; 2131 bcopy(src->ip6po_nexthop, dst->ip6po_nexthop, 2132 src->ip6po_nexthop->sa_len); 2133 } 2134 PKTOPT_EXTHDRCPY(ip6po_hbh); 2135 PKTOPT_EXTHDRCPY(ip6po_dest1); 2136 PKTOPT_EXTHDRCPY(ip6po_dest2); 2137 PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */ 2138 return (dst); 2139 2140 bad: 2141 if (dst->ip6po_pktinfo) kfree(dst->ip6po_pktinfo, M_IP6OPT); 2142 if (dst->ip6po_nexthop) kfree(dst->ip6po_nexthop, M_IP6OPT); 2143 if (dst->ip6po_hbh) kfree(dst->ip6po_hbh, M_IP6OPT); 2144 if (dst->ip6po_dest1) kfree(dst->ip6po_dest1, M_IP6OPT); 2145 if (dst->ip6po_dest2) kfree(dst->ip6po_dest2, M_IP6OPT); 2146 if (dst->ip6po_rthdr) kfree(dst->ip6po_rthdr, M_IP6OPT); 2147 kfree(dst, M_IP6OPT); 2148 return (NULL); 2149 } 2150 2151 static int 2152 copypktopts(struct ip6_pktopts *dst, struct ip6_pktopts *src, int canwait) 2153 { 2154 if (dst == NULL || src == NULL) { 2155 #ifdef DIAGNOSTIC 2156 kprintf("ip6_clearpktopts: invalid argument\n"); 2157 #endif 2158 return (EINVAL); 2159 } 2160 2161 dst->ip6po_hlim = src->ip6po_hlim; 2162 dst->ip6po_tclass = src->ip6po_tclass; 2163 dst->ip6po_flags = src->ip6po_flags; 2164 if (src->ip6po_pktinfo) { 2165 dst->ip6po_pktinfo = kmalloc(sizeof(*dst->ip6po_pktinfo), 2166 M_IP6OPT, canwait); 2167 if (dst->ip6po_pktinfo == NULL) 2168 goto bad; 2169 *dst->ip6po_pktinfo = *src->ip6po_pktinfo; 2170 } 2171 if (src->ip6po_nexthop) { 2172 dst->ip6po_nexthop = kmalloc(src->ip6po_nexthop->sa_len, 2173 M_IP6OPT, canwait); 2174 if (dst->ip6po_nexthop == NULL) 2175 goto bad; 2176 bcopy(src->ip6po_nexthop, dst->ip6po_nexthop, 2177 src->ip6po_nexthop->sa_len); 2178 } 2179 PKTOPT_EXTHDRCPY(ip6po_hbh); 2180 PKTOPT_EXTHDRCPY(ip6po_dest1); 2181 PKTOPT_EXTHDRCPY(ip6po_dest2); 2182 PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */ 2183 return (0); 2184 2185 bad: 2186 ip6_clearpktopts(dst, -1); 2187 return (ENOBUFS); 2188 } 2189 #undef PKTOPT_EXTHDRCPY 2190 2191 void 2192 ip6_freepcbopts(struct ip6_pktopts *pktopt) 2193 { 2194 if (pktopt == NULL) 2195 return; 2196 2197 ip6_clearpktopts(pktopt, -1); 2198 2199 kfree(pktopt, M_IP6OPT); 2200 } 2201 2202 /* 2203 * Set the IP6 multicast options in response to user setsockopt(). 2204 */ 2205 static int 2206 ip6_setmoptions(int optname, struct ip6_moptions **im6op, struct mbuf *m) 2207 { 2208 int error = 0; 2209 u_int loop, ifindex; 2210 struct ipv6_mreq *mreq; 2211 struct ifnet *ifp; 2212 struct ip6_moptions *im6o = *im6op; 2213 struct route_in6 ro; 2214 struct sockaddr_in6 *dst; 2215 struct in6_multi_mship *imm; 2216 2217 if (im6o == NULL) { 2218 /* 2219 * No multicast option buffer attached to the pcb; 2220 * allocate one and initialize to default values. 2221 */ 2222 im6o = (struct ip6_moptions *) 2223 kmalloc(sizeof(*im6o), M_IPMOPTS, M_WAITOK); 2224 2225 *im6op = im6o; 2226 im6o->im6o_multicast_ifp = NULL; 2227 im6o->im6o_multicast_hlim = ip6_defmcasthlim; 2228 im6o->im6o_multicast_loop = IPV6_DEFAULT_MULTICAST_LOOP; 2229 LIST_INIT(&im6o->im6o_memberships); 2230 } 2231 2232 switch (optname) { 2233 2234 case IPV6_MULTICAST_IF: 2235 /* 2236 * Select the interface for outgoing multicast packets. 2237 */ 2238 if (m == NULL || m->m_len != sizeof(u_int)) { 2239 error = EINVAL; 2240 break; 2241 } 2242 bcopy(mtod(m, u_int *), &ifindex, sizeof(ifindex)); 2243 if (ifindex < 0 || if_index < ifindex) { 2244 error = ENXIO; /* XXX EINVAL? */ 2245 break; 2246 } 2247 ifp = ifindex2ifnet[ifindex]; 2248 if (ifp == NULL || !(ifp->if_flags & IFF_MULTICAST)) { 2249 error = EADDRNOTAVAIL; 2250 break; 2251 } 2252 im6o->im6o_multicast_ifp = ifp; 2253 break; 2254 2255 case IPV6_MULTICAST_HOPS: 2256 { 2257 /* 2258 * Set the IP6 hoplimit for outgoing multicast packets. 2259 */ 2260 int optval; 2261 if (m == NULL || m->m_len != sizeof(int)) { 2262 error = EINVAL; 2263 break; 2264 } 2265 bcopy(mtod(m, u_int *), &optval, sizeof(optval)); 2266 if (optval < -1 || optval >= 256) 2267 error = EINVAL; 2268 else if (optval == -1) 2269 im6o->im6o_multicast_hlim = ip6_defmcasthlim; 2270 else 2271 im6o->im6o_multicast_hlim = optval; 2272 break; 2273 } 2274 2275 case IPV6_MULTICAST_LOOP: 2276 /* 2277 * Set the loopback flag for outgoing multicast packets. 2278 * Must be zero or one. 2279 */ 2280 if (m == NULL || m->m_len != sizeof(u_int)) { 2281 error = EINVAL; 2282 break; 2283 } 2284 bcopy(mtod(m, u_int *), &loop, sizeof(loop)); 2285 if (loop > 1) { 2286 error = EINVAL; 2287 break; 2288 } 2289 im6o->im6o_multicast_loop = loop; 2290 break; 2291 2292 case IPV6_JOIN_GROUP: 2293 /* 2294 * Add a multicast group membership. 2295 * Group must be a valid IP6 multicast address. 2296 */ 2297 if (m == NULL || m->m_len != sizeof(struct ipv6_mreq)) { 2298 error = EINVAL; 2299 break; 2300 } 2301 mreq = mtod(m, struct ipv6_mreq *); 2302 if (IN6_IS_ADDR_UNSPECIFIED(&mreq->ipv6mr_multiaddr)) { 2303 /* 2304 * We use the unspecified address to specify to accept 2305 * all multicast addresses. Only super user is allowed 2306 * to do this. 2307 */ 2308 if (caps_priv_check_self(SYSCAP_RESTRICTEDROOT)) { 2309 error = EACCES; 2310 break; 2311 } 2312 } else if (!IN6_IS_ADDR_MULTICAST(&mreq->ipv6mr_multiaddr)) { 2313 error = EINVAL; 2314 break; 2315 } 2316 2317 /* 2318 * If the interface is specified, validate it. 2319 */ 2320 if (mreq->ipv6mr_interface < 0 2321 || if_index < mreq->ipv6mr_interface) { 2322 error = ENXIO; /* XXX EINVAL? */ 2323 break; 2324 } 2325 /* 2326 * If no interface was explicitly specified, choose an 2327 * appropriate one according to the given multicast address. 2328 */ 2329 if (mreq->ipv6mr_interface == 0) { 2330 /* 2331 * If the multicast address is in node-local scope, 2332 * the interface should be a loopback interface. 2333 * Otherwise, look up the routing table for the 2334 * address, and choose the outgoing interface. 2335 * XXX: is it a good approach? 2336 */ 2337 if (IN6_IS_ADDR_MC_INTFACELOCAL(&mreq->ipv6mr_multiaddr)) { 2338 ifp = loif; 2339 } else { 2340 ro.ro_rt = NULL; 2341 dst = (struct sockaddr_in6 *)&ro.ro_dst; 2342 bzero(dst, sizeof(*dst)); 2343 dst->sin6_len = sizeof(struct sockaddr_in6); 2344 dst->sin6_family = AF_INET6; 2345 dst->sin6_addr = mreq->ipv6mr_multiaddr; 2346 rtalloc((struct route *)&ro); 2347 if (ro.ro_rt == NULL) { 2348 error = EADDRNOTAVAIL; 2349 break; 2350 } 2351 ifp = ro.ro_rt->rt_ifp; 2352 rtfree(ro.ro_rt); 2353 } 2354 } else 2355 ifp = ifindex2ifnet[mreq->ipv6mr_interface]; 2356 2357 /* 2358 * See if we found an interface, and confirm that it 2359 * supports multicast 2360 */ 2361 if (ifp == NULL || !(ifp->if_flags & IFF_MULTICAST)) { 2362 error = EADDRNOTAVAIL; 2363 break; 2364 } 2365 /* 2366 * Put interface index into the multicast address, 2367 * if the address has link-local scope. 2368 */ 2369 if (IN6_IS_ADDR_MC_LINKLOCAL(&mreq->ipv6mr_multiaddr)) { 2370 mreq->ipv6mr_multiaddr.s6_addr16[1] 2371 = htons(mreq->ipv6mr_interface); 2372 } 2373 /* 2374 * See if the membership already exists. 2375 */ 2376 for (imm = im6o->im6o_memberships.lh_first; 2377 imm != NULL; imm = imm->i6mm_chain.le_next) 2378 if (imm->i6mm_maddr->in6m_ifp == ifp && 2379 IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr, 2380 &mreq->ipv6mr_multiaddr)) 2381 break; 2382 if (imm != NULL) { 2383 error = EADDRINUSE; 2384 break; 2385 } 2386 /* 2387 * Everything looks good; add a new record to the multicast 2388 * address list for the given interface. 2389 */ 2390 imm = kmalloc(sizeof(*imm), M_IPMADDR, M_WAITOK); 2391 if ((imm->i6mm_maddr = 2392 in6_addmulti(&mreq->ipv6mr_multiaddr, ifp, &error)) == NULL) { 2393 kfree(imm, M_IPMADDR); 2394 break; 2395 } 2396 LIST_INSERT_HEAD(&im6o->im6o_memberships, imm, i6mm_chain); 2397 break; 2398 2399 case IPV6_LEAVE_GROUP: 2400 /* 2401 * Drop a multicast group membership. 2402 * Group must be a valid IP6 multicast address. 2403 */ 2404 if (m == NULL || m->m_len != sizeof(struct ipv6_mreq)) { 2405 error = EINVAL; 2406 break; 2407 } 2408 mreq = mtod(m, struct ipv6_mreq *); 2409 if (IN6_IS_ADDR_UNSPECIFIED(&mreq->ipv6mr_multiaddr)) { 2410 if (caps_priv_check_self(SYSCAP_RESTRICTEDROOT)) { 2411 error = EACCES; 2412 break; 2413 } 2414 } else if (!IN6_IS_ADDR_MULTICAST(&mreq->ipv6mr_multiaddr)) { 2415 error = EINVAL; 2416 break; 2417 } 2418 /* 2419 * If an interface address was specified, get a pointer 2420 * to its ifnet structure. 2421 */ 2422 if (mreq->ipv6mr_interface < 0 2423 || if_index < mreq->ipv6mr_interface) { 2424 error = ENXIO; /* XXX EINVAL? */ 2425 break; 2426 } 2427 ifp = ifindex2ifnet[mreq->ipv6mr_interface]; 2428 /* 2429 * Put interface index into the multicast address, 2430 * if the address has link-local scope. 2431 */ 2432 if (IN6_IS_ADDR_MC_LINKLOCAL(&mreq->ipv6mr_multiaddr)) { 2433 mreq->ipv6mr_multiaddr.s6_addr16[1] 2434 = htons(mreq->ipv6mr_interface); 2435 } 2436 2437 /* 2438 * Find the membership in the membership list. 2439 */ 2440 for (imm = im6o->im6o_memberships.lh_first; 2441 imm != NULL; imm = imm->i6mm_chain.le_next) { 2442 if ((ifp == NULL || imm->i6mm_maddr->in6m_ifp == ifp) && 2443 IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr, 2444 &mreq->ipv6mr_multiaddr)) 2445 break; 2446 } 2447 if (imm == NULL) { 2448 /* Unable to resolve interface */ 2449 error = EADDRNOTAVAIL; 2450 break; 2451 } 2452 /* 2453 * Give up the multicast address record to which the 2454 * membership points. 2455 */ 2456 LIST_REMOVE(imm, i6mm_chain); 2457 in6_delmulti(imm->i6mm_maddr); 2458 kfree(imm, M_IPMADDR); 2459 break; 2460 2461 default: 2462 error = EOPNOTSUPP; 2463 break; 2464 } 2465 2466 /* 2467 * If all options have default values, no need to keep the mbuf. 2468 */ 2469 if (im6o->im6o_multicast_ifp == NULL && 2470 im6o->im6o_multicast_hlim == ip6_defmcasthlim && 2471 im6o->im6o_multicast_loop == IPV6_DEFAULT_MULTICAST_LOOP && 2472 im6o->im6o_memberships.lh_first == NULL) { 2473 kfree(*im6op, M_IPMOPTS); 2474 *im6op = NULL; 2475 } 2476 2477 return (error); 2478 } 2479 2480 /* 2481 * Return the IP6 multicast options in response to user getsockopt(). 2482 */ 2483 static int 2484 ip6_getmoptions(int optname, struct ip6_moptions *im6o, struct mbuf **mp) 2485 { 2486 u_int *hlim, *loop, *ifindex; 2487 2488 *mp = m_get(M_WAITOK, MT_HEADER); /* XXX */ 2489 2490 switch (optname) { 2491 2492 case IPV6_MULTICAST_IF: 2493 ifindex = mtod(*mp, u_int *); 2494 (*mp)->m_len = sizeof(u_int); 2495 if (im6o == NULL || im6o->im6o_multicast_ifp == NULL) 2496 *ifindex = 0; 2497 else 2498 *ifindex = im6o->im6o_multicast_ifp->if_index; 2499 return (0); 2500 2501 case IPV6_MULTICAST_HOPS: 2502 hlim = mtod(*mp, u_int *); 2503 (*mp)->m_len = sizeof(u_int); 2504 if (im6o == NULL) 2505 *hlim = ip6_defmcasthlim; 2506 else 2507 *hlim = im6o->im6o_multicast_hlim; 2508 return (0); 2509 2510 case IPV6_MULTICAST_LOOP: 2511 loop = mtod(*mp, u_int *); 2512 (*mp)->m_len = sizeof(u_int); 2513 if (im6o == NULL) 2514 *loop = ip6_defmcasthlim; 2515 else 2516 *loop = im6o->im6o_multicast_loop; 2517 return (0); 2518 2519 default: 2520 return (EOPNOTSUPP); 2521 } 2522 } 2523 2524 /* 2525 * Discard the IP6 multicast options. 2526 */ 2527 void 2528 ip6_freemoptions(struct ip6_moptions *im6o) 2529 { 2530 struct in6_multi_mship *imm; 2531 2532 if (im6o == NULL) 2533 return; 2534 2535 while ((imm = im6o->im6o_memberships.lh_first) != NULL) { 2536 LIST_REMOVE(imm, i6mm_chain); 2537 if (imm->i6mm_maddr) 2538 in6_delmulti(imm->i6mm_maddr); 2539 kfree(imm, M_IPMADDR); 2540 } 2541 kfree(im6o, M_IPMOPTS); 2542 } 2543 2544 /* 2545 * Set a particular packet option, as a sticky option or an ancillary data 2546 * item. "len" can be 0 only when it's a sticky option. 2547 * We have 4 cases of combination of "sticky" and "cmsg": 2548 * "sticky=0, cmsg=0": impossible 2549 * "sticky=0, cmsg=1": RFC2292 or RFC3542 ancillary data 2550 * "sticky=1, cmsg=0": RFC3542 socket option 2551 * "sticky=1, cmsg=1": RFC2292 socket option 2552 */ 2553 static int 2554 ip6_setpktoption(int optname, u_char *buf, int len, struct ip6_pktopts *opt, 2555 int sticky, int cmsg, int uproto, int priv) 2556 { 2557 int minmtupolicy, preftemp; 2558 //int error; 2559 2560 if (!sticky && !cmsg) { 2561 kprintf("ip6_setpktoption: impossible case\n"); 2562 return (EINVAL); 2563 } 2564 2565 /* 2566 * IPV6_2292xxx is for backward compatibility to RFC2292, and should 2567 * not be specified in the context of RFC3542. Conversely, 2568 * RFC3542 types should not be specified in the context of RFC2292. 2569 */ 2570 if (!cmsg) { 2571 switch (optname) { 2572 case IPV6_2292PKTINFO: 2573 case IPV6_2292HOPLIMIT: 2574 case IPV6_2292NEXTHOP: 2575 case IPV6_2292HOPOPTS: 2576 case IPV6_2292DSTOPTS: 2577 case IPV6_2292RTHDR: 2578 case IPV6_2292PKTOPTIONS: 2579 return (ENOPROTOOPT); 2580 } 2581 } 2582 if (sticky && cmsg) { 2583 switch (optname) { 2584 case IPV6_PKTINFO: 2585 case IPV6_HOPLIMIT: 2586 case IPV6_NEXTHOP: 2587 case IPV6_HOPOPTS: 2588 case IPV6_DSTOPTS: 2589 case IPV6_RTHDRDSTOPTS: 2590 case IPV6_RTHDR: 2591 case IPV6_USE_MIN_MTU: 2592 case IPV6_DONTFRAG: 2593 case IPV6_TCLASS: 2594 case IPV6_PREFER_TEMPADDR: /* XXX: not an RFC3542 option */ 2595 return (ENOPROTOOPT); 2596 } 2597 } 2598 2599 switch (optname) { 2600 case IPV6_2292PKTINFO: 2601 case IPV6_PKTINFO: 2602 { 2603 struct in6_pktinfo *pktinfo; 2604 if (len != sizeof(struct in6_pktinfo)) 2605 return (EINVAL); 2606 pktinfo = (struct in6_pktinfo *)buf; 2607 2608 /* 2609 * An application can clear any sticky IPV6_PKTINFO option by 2610 * doing a "regular" setsockopt with ipi6_addr being 2611 * in6addr_any and ipi6_ifindex being zero. 2612 * [RFC 3542, Section 6] 2613 */ 2614 if (optname == IPV6_PKTINFO && opt->ip6po_pktinfo && 2615 pktinfo->ipi6_ifindex == 0 && 2616 IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) { 2617 ip6_clearpktopts(opt, optname); 2618 break; 2619 } 2620 2621 if (uproto == IPPROTO_TCP && optname == IPV6_PKTINFO && 2622 sticky && !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) { 2623 return (EINVAL); 2624 } 2625 2626 /* validate the interface index if specified. */ 2627 if (pktinfo->ipi6_ifindex > if_index || 2628 pktinfo->ipi6_ifindex < 0) { 2629 return (ENXIO); 2630 } 2631 /* 2632 * Check if the requested source address is indeed a 2633 * unicast address assigned to the node, and can be 2634 * used as the packet's source address. 2635 */ 2636 if (opt->ip6po_pktinfo != NULL && 2637 !IN6_IS_ADDR_UNSPECIFIED(&opt->ip6po_pktinfo->ipi6_addr)) { 2638 struct in6_ifaddr *ia6; 2639 struct sockaddr_in6 sin6; 2640 2641 bzero(&sin6, sizeof(sin6)); 2642 sin6.sin6_len = sizeof(sin6); 2643 sin6.sin6_family = AF_INET6; 2644 sin6.sin6_addr = 2645 opt->ip6po_pktinfo->ipi6_addr; 2646 ia6 = (struct in6_ifaddr *)ifa_ifwithaddr(sin6tosa(&sin6)); 2647 if (ia6 == NULL || 2648 (ia6->ia6_flags & (IN6_IFF_ANYCAST | 2649 IN6_IFF_NOTREADY)) != 0) 2650 return (EADDRNOTAVAIL); 2651 } 2652 2653 /* 2654 * We store the address anyway, and let in6_selectsrc() 2655 * validate the specified address. This is because ipi6_addr 2656 * may not have enough information about its scope zone, and 2657 * we may need additional information (such as outgoing 2658 * interface or the scope zone of a destination address) to 2659 * disambiguate the scope. 2660 * XXX: the delay of the validation may confuse the 2661 * application when it is used as a sticky option. 2662 */ 2663 if (opt->ip6po_pktinfo == NULL) { 2664 opt->ip6po_pktinfo = kmalloc(sizeof(*pktinfo), 2665 M_IP6OPT, M_NOWAIT); 2666 if (opt->ip6po_pktinfo == NULL) 2667 return (ENOBUFS); 2668 } 2669 bcopy(pktinfo, opt->ip6po_pktinfo, sizeof(*pktinfo)); 2670 break; 2671 } 2672 2673 case IPV6_2292HOPLIMIT: 2674 case IPV6_HOPLIMIT: 2675 { 2676 int *hlimp; 2677 2678 /* 2679 * RFC 3542 deprecated the usage of sticky IPV6_HOPLIMIT 2680 * to simplify the ordering among hoplimit options. 2681 */ 2682 if (optname == IPV6_HOPLIMIT && sticky) 2683 return (ENOPROTOOPT); 2684 2685 if (len != sizeof(int)) 2686 return (EINVAL); 2687 hlimp = (int *)buf; 2688 if (*hlimp < -1 || *hlimp > 255) 2689 return (EINVAL); 2690 2691 opt->ip6po_hlim = *hlimp; 2692 break; 2693 } 2694 2695 case IPV6_TCLASS: 2696 { 2697 int tclass; 2698 2699 if (len != sizeof(int)) 2700 return (EINVAL); 2701 tclass = *(int *)buf; 2702 if (tclass < -1 || tclass > 255) 2703 return (EINVAL); 2704 2705 opt->ip6po_tclass = tclass; 2706 break; 2707 } 2708 2709 case IPV6_2292NEXTHOP: 2710 case IPV6_NEXTHOP: 2711 if (!priv) 2712 return (EPERM); 2713 2714 if (len == 0) { /* just remove the option */ 2715 ip6_clearpktopts(opt, IPV6_NEXTHOP); 2716 break; 2717 } 2718 2719 /* check if cmsg_len is large enough for sa_len */ 2720 if (len < sizeof(struct sockaddr) || len < *buf) 2721 return (EINVAL); 2722 2723 switch (((struct sockaddr *)buf)->sa_family) { 2724 case AF_INET6: 2725 { 2726 struct sockaddr_in6 *sa6 = (struct sockaddr_in6 *)buf; 2727 //int error; 2728 2729 if (sa6->sin6_len != sizeof(struct sockaddr_in6)) 2730 return (EINVAL); 2731 2732 if (IN6_IS_ADDR_UNSPECIFIED(&sa6->sin6_addr) || 2733 IN6_IS_ADDR_MULTICAST(&sa6->sin6_addr)) { 2734 return (EINVAL); 2735 } 2736 break; 2737 } 2738 case AF_LINK: /* should eventually be supported */ 2739 default: 2740 return (EAFNOSUPPORT); 2741 } 2742 2743 /* turn off the previous option, then set the new option. */ 2744 ip6_clearpktopts(opt, IPV6_NEXTHOP); 2745 opt->ip6po_nexthop = kmalloc(*buf, M_IP6OPT, M_NOWAIT); 2746 if (opt->ip6po_nexthop == NULL) 2747 return (ENOBUFS); 2748 bcopy(buf, opt->ip6po_nexthop, *buf); 2749 break; 2750 2751 case IPV6_2292HOPOPTS: 2752 case IPV6_HOPOPTS: 2753 { 2754 struct ip6_hbh *hbh; 2755 int hbhlen; 2756 2757 /* 2758 * XXX: We don't allow a non-privileged user to set ANY HbH 2759 * options, since per-option restriction has too much 2760 * overhead. 2761 */ 2762 if (!priv) 2763 return (EPERM); 2764 if (len == 0) { 2765 ip6_clearpktopts(opt, IPV6_HOPOPTS); 2766 break; /* just remove the option */ 2767 } 2768 2769 /* message length validation */ 2770 if (len < sizeof(struct ip6_hbh)) 2771 return (EINVAL); 2772 hbh = (struct ip6_hbh *)buf; 2773 hbhlen = (hbh->ip6h_len + 1) << 3; 2774 if (len != hbhlen) 2775 return (EINVAL); 2776 2777 /* turn off the previous option, then set the new option. */ 2778 ip6_clearpktopts(opt, IPV6_HOPOPTS); 2779 opt->ip6po_hbh = kmalloc(hbhlen, M_IP6OPT, M_NOWAIT); 2780 if (opt->ip6po_hbh == NULL) 2781 return (ENOBUFS); 2782 bcopy(hbh, opt->ip6po_hbh, hbhlen); 2783 2784 break; 2785 } 2786 2787 case IPV6_2292DSTOPTS: 2788 case IPV6_DSTOPTS: 2789 case IPV6_RTHDRDSTOPTS: 2790 { 2791 struct ip6_dest *dest, **newdest = NULL; 2792 int destlen; 2793 if (!priv) 2794 return (EPERM); 2795 2796 if (len == 0) { 2797 ip6_clearpktopts(opt, optname); 2798 break; /* just remove the option */ 2799 } 2800 2801 /* message length validation */ 2802 if (len < sizeof(struct ip6_dest)) 2803 return (EINVAL); 2804 dest = (struct ip6_dest *)buf; 2805 destlen = (dest->ip6d_len + 1) << 3; 2806 if (len != destlen) 2807 return (EINVAL); 2808 2809 /* 2810 * Determine the position that the destination options header 2811 * should be inserted; before or after the routing header. 2812 */ 2813 switch (optname) { 2814 case IPV6_2292DSTOPTS: 2815 /* 2816 * The old advacned API is ambiguous on this point. 2817 * Our approach is to determine the position based 2818 * according to the existence of a routing header. 2819 * Note, however, that this depends on the order of the 2820 * extension headers in the ancillary data; the 1st 2821 * part of the destination options header must appear 2822 * before the routing header in the ancillary data, 2823 * too. 2824 * RFC3542 solved the ambiguity by introducing 2825 * separate ancillary data or option types. 2826 */ 2827 if (opt->ip6po_rthdr == NULL) 2828 newdest = &opt->ip6po_dest1; 2829 else 2830 newdest = &opt->ip6po_dest2; 2831 break; 2832 case IPV6_RTHDRDSTOPTS: 2833 newdest = &opt->ip6po_dest1; 2834 break; 2835 case IPV6_DSTOPTS: 2836 newdest = &opt->ip6po_dest2; 2837 break; 2838 } 2839 2840 /* turn off the previous option, then set the new option. */ 2841 ip6_clearpktopts(opt, optname); 2842 *newdest = kmalloc(destlen, M_IP6OPT, M_NOWAIT); 2843 if (*newdest == NULL) 2844 return (ENOBUFS); 2845 bcopy(dest, *newdest, destlen); 2846 2847 break; 2848 } 2849 2850 case IPV6_2292RTHDR: 2851 case IPV6_RTHDR: 2852 { 2853 struct ip6_rthdr *rth; 2854 int rthlen; 2855 2856 if (len == 0) { 2857 ip6_clearpktopts(opt, IPV6_RTHDR); 2858 break; /* just remove the option */ 2859 } 2860 2861 /* message length validation */ 2862 if (len < sizeof(struct ip6_rthdr)) 2863 return (EINVAL); 2864 rth = (struct ip6_rthdr *)buf; 2865 rthlen = (rth->ip6r_len + 1) << 3; 2866 if (len != rthlen) 2867 return (EINVAL); 2868 2869 switch (rth->ip6r_type) { 2870 default: 2871 return (EINVAL); /* not supported */ 2872 } 2873 2874 /* turn off the previous option */ 2875 ip6_clearpktopts(opt, IPV6_RTHDR); 2876 opt->ip6po_rthdr = kmalloc(rthlen, M_IP6OPT, M_NOWAIT); 2877 if (opt->ip6po_rthdr == NULL) 2878 return (ENOBUFS); 2879 bcopy(rth, opt->ip6po_rthdr, rthlen); 2880 2881 break; 2882 } 2883 2884 case IPV6_USE_MIN_MTU: 2885 if (len != sizeof(int)) 2886 return (EINVAL); 2887 minmtupolicy = *(int *)buf; 2888 if (minmtupolicy != IP6PO_MINMTU_MCASTONLY && 2889 minmtupolicy != IP6PO_MINMTU_DISABLE && 2890 minmtupolicy != IP6PO_MINMTU_ALL) { 2891 return (EINVAL); 2892 } 2893 opt->ip6po_minmtu = minmtupolicy; 2894 break; 2895 2896 case IPV6_DONTFRAG: 2897 if (len != sizeof(int)) 2898 return (EINVAL); 2899 2900 if (uproto == IPPROTO_TCP || *(int *)buf == 0) { 2901 /* 2902 * we ignore this option for TCP sockets. 2903 * (RFC3542 leaves this case unspecified.) 2904 */ 2905 opt->ip6po_flags &= ~IP6PO_DONTFRAG; 2906 } else 2907 opt->ip6po_flags |= IP6PO_DONTFRAG; 2908 break; 2909 2910 case IPV6_PREFER_TEMPADDR: 2911 if (len != sizeof(int)) 2912 return (EINVAL); 2913 preftemp = *(int *)buf; 2914 if (preftemp != IP6PO_TEMPADDR_SYSTEM && 2915 preftemp != IP6PO_TEMPADDR_NOTPREFER && 2916 preftemp != IP6PO_TEMPADDR_PREFER) { 2917 return (EINVAL); 2918 } 2919 opt->ip6po_prefer_tempaddr = preftemp; 2920 break; 2921 2922 default: 2923 return (ENOPROTOOPT); 2924 } /* end of switch */ 2925 2926 return (0); 2927 } 2928 2929 2930 /* 2931 * Set IPv6 outgoing packet options based on advanced API. 2932 */ 2933 int 2934 ip6_setpktoptions(struct mbuf *control, struct ip6_pktopts *opt, 2935 struct ip6_pktopts *stickyopt, int uproto, int priv) 2936 { 2937 struct cmsghdr *cm = NULL; 2938 2939 if (control == NULL || opt == NULL) 2940 return (EINVAL); 2941 2942 init_ip6pktopts(opt); 2943 2944 /* 2945 * XXX: Currently, we assume all the optional information is stored 2946 * in a single mbuf. 2947 */ 2948 if (stickyopt) { 2949 int error; 2950 2951 /* 2952 * If stickyopt is provided, make a local copy of the options 2953 * for this particular packet, then override them by ancillary 2954 * objects. 2955 * XXX: copypktopts() does not copy the cached route to a next 2956 * hop (if any). This is not very good in terms of efficiency, 2957 * but we can allow this since this option should be rarely 2958 * used. 2959 */ 2960 if ((error = copypktopts(opt, stickyopt, M_NOWAIT)) != 0) 2961 return (error); 2962 } 2963 2964 /* 2965 * XXX: Currently, we assume all the optional information is stored 2966 * in a single mbuf. 2967 */ 2968 if (control->m_next) 2969 return (EINVAL); 2970 2971 for (;;) { 2972 int error; 2973 2974 if (control->m_len == 0) 2975 break; 2976 if (control->m_len < sizeof(*cm)) 2977 return EINVAL; 2978 2979 cm = mtod(control, struct cmsghdr *); 2980 if (cm->cmsg_len == 0 || cm->cmsg_len > control->m_len) 2981 return (EINVAL); 2982 if (cm->cmsg_level == IPPROTO_IPV6) { 2983 error = ip6_setpktoption(cm->cmsg_type, CMSG_DATA(cm), 2984 cm->cmsg_len - CMSG_LEN(0), 2985 opt, 0, 1, uproto, priv); 2986 if (error) 2987 return (error); 2988 } 2989 2990 /* 2991 * The cmsg fit, but the aligned step for the next one might 2992 * not. Check the case and terminate normally (allows the 2993 * cmsg_len to not be aligned). 2994 */ 2995 if (CMSG_ALIGN(cm->cmsg_len) >= control->m_len) { 2996 control->m_data += control->m_len; 2997 control->m_len = 0; 2998 break; 2999 } 3000 control->m_data += CMSG_ALIGN(cm->cmsg_len); 3001 control->m_len -= CMSG_ALIGN(cm->cmsg_len); 3002 } 3003 3004 return (0); 3005 } 3006 3007 /* 3008 * Routine called from ip6_output() to loop back a copy of an IP6 multicast 3009 * packet to the input queue of a specified interface. Note that this 3010 * calls the output routine of the loopback "driver", but with an interface 3011 * pointer that might NOT be loif -- easier than replicating that code here. 3012 */ 3013 void 3014 ip6_mloopback(struct ifnet *ifp, struct mbuf *m, struct sockaddr_in6 *dst) 3015 { 3016 struct mbuf *copym; 3017 struct ip6_hdr *ip6; 3018 3019 copym = m_copy(m, 0, M_COPYALL); 3020 if (copym == NULL) 3021 return; 3022 3023 /* 3024 * Make sure to deep-copy IPv6 header portion in case the data 3025 * is in an mbuf cluster, so that we can safely override the IPv6 3026 * header portion later. 3027 */ 3028 if ((copym->m_flags & M_EXT) != 0 || 3029 copym->m_len < sizeof(struct ip6_hdr)) { 3030 copym = m_pullup(copym, sizeof(struct ip6_hdr)); 3031 if (copym == NULL) 3032 return; 3033 } 3034 3035 #ifdef DIAGNOSTIC 3036 if (copym->m_len < sizeof(*ip6)) { 3037 m_freem(copym); 3038 return; 3039 } 3040 #endif 3041 3042 ip6 = mtod(copym, struct ip6_hdr *); 3043 /* 3044 * clear embedded scope identifiers if necessary. 3045 * in6_clearscope will touch the addresses only when necessary. 3046 */ 3047 in6_clearscope(&ip6->ip6_src); 3048 in6_clearscope(&ip6->ip6_dst); 3049 3050 if_simloop(ifp, copym, dst->sin6_family, 0); 3051 } 3052 3053 /* 3054 * Separate the IPv6 header from the payload into its own mbuf. 3055 * 3056 * Returns the new mbuf chain or the original mbuf if no payload. 3057 * Returns NULL if can't allocate new mbuf for header. 3058 */ 3059 static struct mbuf * 3060 ip6_splithdr(struct mbuf *m) 3061 { 3062 struct mbuf *mh; 3063 3064 if (m->m_len <= sizeof(struct ip6_hdr)) /* no payload */ 3065 return (m); 3066 3067 MGETHDR(mh, M_NOWAIT, MT_HEADER); 3068 if (mh == NULL) 3069 return (NULL); 3070 mh->m_len = sizeof(struct ip6_hdr); 3071 M_MOVE_PKTHDR(mh, m); 3072 MH_ALIGN(mh, sizeof(struct ip6_hdr)); 3073 bcopy(mtod(m, caddr_t), mtod(mh, caddr_t), sizeof(struct ip6_hdr)); 3074 m->m_data += sizeof(struct ip6_hdr); 3075 m->m_len -= sizeof(struct ip6_hdr); 3076 mh->m_next = m; 3077 return (mh); 3078 } 3079 3080 /* 3081 * Compute IPv6 extension header length. 3082 */ 3083 int 3084 ip6_optlen(struct in6pcb *in6p) 3085 { 3086 int len; 3087 3088 if (!in6p->in6p_outputopts) 3089 return 0; 3090 3091 len = 0; 3092 #define elen(x) \ 3093 (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0) 3094 3095 len += elen(in6p->in6p_outputopts->ip6po_hbh); 3096 if (in6p->in6p_outputopts->ip6po_rthdr) 3097 /* dest1 is valid with rthdr only */ 3098 len += elen(in6p->in6p_outputopts->ip6po_dest1); 3099 len += elen(in6p->in6p_outputopts->ip6po_rthdr); 3100 len += elen(in6p->in6p_outputopts->ip6po_dest2); 3101 return len; 3102 #undef elen 3103 } 3104