1 /* $OpenBSD: ip6_output.c,v 1.257 2021/05/12 08:09:33 mvs Exp $ */ 2 /* $KAME: ip6_output.c,v 1.172 2001/03/25 09:55:56 itojun Exp $ */ 3 4 /* 5 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project. 6 * All rights reserved. 7 * 8 * Redistribution and use in source and binary forms, with or without 9 * modification, are permitted provided that the following conditions 10 * are met: 11 * 1. Redistributions of source code must retain the above copyright 12 * notice, this list of conditions and the following disclaimer. 13 * 2. Redistributions in binary form must reproduce the above copyright 14 * notice, this list of conditions and the following disclaimer in the 15 * documentation and/or other materials provided with the distribution. 16 * 3. Neither the name of the project nor the names of its contributors 17 * may be used to endorse or promote products derived from this software 18 * without specific prior written permission. 19 * 20 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND 21 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 22 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 23 * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE 24 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 25 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 26 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 27 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 28 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 29 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 30 * SUCH DAMAGE. 31 */ 32 33 /* 34 * Copyright (c) 1982, 1986, 1988, 1990, 1993 35 * The Regents of the University of California. All rights reserved. 36 * 37 * Redistribution and use in source and binary forms, with or without 38 * modification, are permitted provided that the following conditions 39 * are met: 40 * 1. Redistributions of source code must retain the above copyright 41 * notice, this list of conditions and the following disclaimer. 42 * 2. Redistributions in binary form must reproduce the above copyright 43 * notice, this list of conditions and the following disclaimer in the 44 * documentation and/or other materials provided with the distribution. 45 * 3. Neither the name of the University nor the names of its contributors 46 * may be used to endorse or promote products derived from this software 47 * without specific prior written permission. 48 * 49 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 50 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 51 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 52 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 53 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 54 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 55 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 56 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 57 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 58 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 59 * SUCH DAMAGE. 60 * 61 * @(#)ip_output.c 8.3 (Berkeley) 1/21/94 62 */ 63 64 #include "pf.h" 65 66 #include <sys/param.h> 67 #include <sys/malloc.h> 68 #include <sys/mbuf.h> 69 #include <sys/errno.h> 70 #include <sys/protosw.h> 71 #include <sys/socket.h> 72 #include <sys/socketvar.h> 73 #include <sys/proc.h> 74 #include <sys/systm.h> 75 76 #include <net/if.h> 77 #include <net/if_var.h> 78 #include <net/if_enc.h> 79 #include <net/route.h> 80 81 #include <netinet/in.h> 82 #include <netinet/ip.h> 83 #include <netinet/in_pcb.h> 84 #include <netinet/udp.h> 85 #include <netinet/tcp.h> 86 87 #include <netinet/ip_var.h> 88 #include <netinet/tcp_timer.h> 89 #include <netinet/tcp_var.h> 90 #include <netinet/udp_var.h> 91 92 #include <netinet6/in6_var.h> 93 #include <netinet/ip6.h> 94 #include <netinet/icmp6.h> 95 #include <netinet6/ip6_var.h> 96 #include <netinet6/nd6.h> 97 #include <netinet6/ip6protosw.h> 98 99 #include <crypto/idgen.h> 100 101 #if NPF > 0 102 #include <net/pfvar.h> 103 #endif 104 105 #ifdef IPSEC 106 #include <netinet/ip_ipsp.h> 107 #include <netinet/ip_ah.h> 108 #include <netinet/ip_esp.h> 109 110 #ifdef ENCDEBUG 111 #define DPRINTF(x) do { if (encdebug) printf x ; } while (0) 112 #else 113 #define DPRINTF(x) 114 #endif 115 #endif /* IPSEC */ 116 117 struct ip6_exthdrs { 118 struct mbuf *ip6e_ip6; 119 struct mbuf *ip6e_hbh; 120 struct mbuf *ip6e_dest1; 121 struct mbuf *ip6e_rthdr; 122 struct mbuf *ip6e_dest2; 123 }; 124 125 int ip6_pcbopt(int, u_char *, int, struct ip6_pktopts **, int, int); 126 int ip6_getpcbopt(struct ip6_pktopts *, int, struct mbuf *); 127 int ip6_setpktopt(int, u_char *, int, struct ip6_pktopts *, int, int, int); 128 int ip6_setmoptions(int, struct ip6_moptions **, struct mbuf *, unsigned int); 129 int ip6_getmoptions(int, struct ip6_moptions *, struct mbuf *); 130 int ip6_copyexthdr(struct mbuf **, caddr_t, int); 131 int ip6_insertfraghdr(struct mbuf *, struct mbuf *, int, 132 struct ip6_frag **); 133 int ip6_insert_jumboopt(struct ip6_exthdrs *, u_int32_t); 134 int ip6_splithdr(struct mbuf *, struct ip6_exthdrs *); 135 int ip6_getpmtu(struct rtentry *, struct ifnet *, u_long *); 136 int copypktopts(struct ip6_pktopts *, struct ip6_pktopts *); 137 static __inline u_int16_t __attribute__((__unused__)) 138 in6_cksum_phdr(const struct in6_addr *, const struct in6_addr *, 139 u_int32_t, u_int32_t); 140 void in6_delayed_cksum(struct mbuf *, u_int8_t); 141 142 /* Context for non-repeating IDs */ 143 struct idgen32_ctx ip6_id_ctx; 144 145 /* 146 * IP6 output. The packet in mbuf chain m contains a skeletal IP6 147 * header (with pri, len, nxt, hlim, src, dst). 148 * This function may modify ver and hlim only. 149 * The mbuf chain containing the packet will be freed. 150 * The mbuf opt, if present, will not be freed. 151 * 152 * type of "mtu": rt_mtu is u_long, ifnet.ifr_mtu is int, and 153 * nd_ifinfo.linkmtu is u_int32_t. so we use u_long to hold largest one, 154 * which is rt_mtu. 155 */ 156 int 157 ip6_output(struct mbuf *m, struct ip6_pktopts *opt, struct route_in6 *ro, 158 int flags, struct ip6_moptions *im6o, struct inpcb *inp) 159 { 160 struct ip6_hdr *ip6; 161 struct ifnet *ifp = NULL; 162 struct mbuf_list fml; 163 int hlen, tlen; 164 struct route_in6 ip6route; 165 struct rtentry *rt = NULL; 166 struct sockaddr_in6 *dst, dstsock; 167 int error = 0; 168 u_long mtu; 169 int dontfrag; 170 u_int16_t src_scope, dst_scope; 171 u_int32_t optlen = 0, plen = 0, unfragpartlen = 0; 172 struct ip6_exthdrs exthdrs; 173 struct in6_addr finaldst; 174 struct route_in6 *ro_pmtu = NULL; 175 int hdrsplit = 0; 176 u_int8_t sproto = 0; 177 u_char nextproto; 178 #ifdef IPSEC 179 struct tdb *tdb = NULL; 180 #endif /* IPSEC */ 181 182 #ifdef IPSEC 183 if (inp && (inp->inp_flags & INP_IPV6) == 0) 184 panic("%s: IPv4 pcb is passed", __func__); 185 #endif /* IPSEC */ 186 187 ip6 = mtod(m, struct ip6_hdr *); 188 finaldst = ip6->ip6_dst; 189 190 #define MAKE_EXTHDR(hp, mp) \ 191 do { \ 192 if (hp) { \ 193 struct ip6_ext *eh = (struct ip6_ext *)(hp); \ 194 error = ip6_copyexthdr((mp), (caddr_t)(hp), \ 195 ((eh)->ip6e_len + 1) << 3); \ 196 if (error) \ 197 goto freehdrs; \ 198 } \ 199 } while (0) 200 201 bzero(&exthdrs, sizeof(exthdrs)); 202 203 if (opt) { 204 /* Hop-by-Hop options header */ 205 MAKE_EXTHDR(opt->ip6po_hbh, &exthdrs.ip6e_hbh); 206 /* Destination options header(1st part) */ 207 MAKE_EXTHDR(opt->ip6po_dest1, &exthdrs.ip6e_dest1); 208 /* Routing header */ 209 MAKE_EXTHDR(opt->ip6po_rthdr, &exthdrs.ip6e_rthdr); 210 /* Destination options header(2nd part) */ 211 MAKE_EXTHDR(opt->ip6po_dest2, &exthdrs.ip6e_dest2); 212 } 213 214 #ifdef IPSEC 215 if (ipsec_in_use || inp) { 216 tdb = ip6_output_ipsec_lookup(m, &error, inp); 217 if (error != 0) { 218 /* 219 * -EINVAL is used to indicate that the packet should 220 * be silently dropped, typically because we've asked 221 * key management for an SA. 222 */ 223 if (error == -EINVAL) /* Should silently drop packet */ 224 error = 0; 225 226 goto freehdrs; 227 } 228 } 229 #endif /* IPSEC */ 230 231 /* 232 * Calculate the total length of the extension header chain. 233 * Keep the length of the unfragmentable part for fragmentation. 234 */ 235 optlen = 0; 236 if (exthdrs.ip6e_hbh) optlen += exthdrs.ip6e_hbh->m_len; 237 if (exthdrs.ip6e_dest1) optlen += exthdrs.ip6e_dest1->m_len; 238 if (exthdrs.ip6e_rthdr) optlen += exthdrs.ip6e_rthdr->m_len; 239 unfragpartlen = optlen + sizeof(struct ip6_hdr); 240 /* NOTE: we don't add AH/ESP length here. do that later. */ 241 if (exthdrs.ip6e_dest2) optlen += exthdrs.ip6e_dest2->m_len; 242 243 /* 244 * If we need IPsec, or there is at least one extension header, 245 * separate IP6 header from the payload. 246 */ 247 if ((sproto || optlen) && !hdrsplit) { 248 if ((error = ip6_splithdr(m, &exthdrs)) != 0) { 249 m = NULL; 250 goto freehdrs; 251 } 252 m = exthdrs.ip6e_ip6; 253 hdrsplit++; 254 } 255 256 /* adjust pointer */ 257 ip6 = mtod(m, struct ip6_hdr *); 258 259 /* adjust mbuf packet header length */ 260 m->m_pkthdr.len += optlen; 261 plen = m->m_pkthdr.len - sizeof(*ip6); 262 263 /* If this is a jumbo payload, insert a jumbo payload option. */ 264 if (plen > IPV6_MAXPACKET) { 265 if (!hdrsplit) { 266 if ((error = ip6_splithdr(m, &exthdrs)) != 0) { 267 m = NULL; 268 goto freehdrs; 269 } 270 m = exthdrs.ip6e_ip6; 271 hdrsplit++; 272 } 273 /* adjust pointer */ 274 ip6 = mtod(m, struct ip6_hdr *); 275 if ((error = ip6_insert_jumboopt(&exthdrs, plen)) != 0) 276 goto freehdrs; 277 ip6->ip6_plen = 0; 278 } else 279 ip6->ip6_plen = htons(plen); 280 281 /* 282 * Concatenate headers and fill in next header fields. 283 * Here we have, on "m" 284 * IPv6 payload 285 * and we insert headers accordingly. Finally, we should be getting: 286 * IPv6 hbh dest1 rthdr ah* [esp* dest2 payload] 287 * 288 * during the header composing process, "m" points to IPv6 header. 289 * "mprev" points to an extension header prior to esp. 290 */ 291 { 292 u_char *nexthdrp = &ip6->ip6_nxt; 293 struct mbuf *mprev = m; 294 295 /* 296 * we treat dest2 specially. this makes IPsec processing 297 * much easier. the goal here is to make mprev point the 298 * mbuf prior to dest2. 299 * 300 * result: IPv6 dest2 payload 301 * m and mprev will point to IPv6 header. 302 */ 303 if (exthdrs.ip6e_dest2) { 304 if (!hdrsplit) 305 panic("%s: assumption failed: hdr not split", 306 __func__); 307 exthdrs.ip6e_dest2->m_next = m->m_next; 308 m->m_next = exthdrs.ip6e_dest2; 309 *mtod(exthdrs.ip6e_dest2, u_char *) = ip6->ip6_nxt; 310 ip6->ip6_nxt = IPPROTO_DSTOPTS; 311 } 312 313 #define MAKE_CHAIN(m, mp, p, i)\ 314 do {\ 315 if (m) {\ 316 if (!hdrsplit) \ 317 panic("assumption failed: hdr not split"); \ 318 *mtod((m), u_char *) = *(p);\ 319 *(p) = (i);\ 320 p = mtod((m), u_char *);\ 321 (m)->m_next = (mp)->m_next;\ 322 (mp)->m_next = (m);\ 323 (mp) = (m);\ 324 }\ 325 } while (0) 326 /* 327 * result: IPv6 hbh dest1 rthdr dest2 payload 328 * m will point to IPv6 header. mprev will point to the 329 * extension header prior to dest2 (rthdr in the above case). 330 */ 331 MAKE_CHAIN(exthdrs.ip6e_hbh, mprev, nexthdrp, IPPROTO_HOPOPTS); 332 MAKE_CHAIN(exthdrs.ip6e_dest1, mprev, nexthdrp, 333 IPPROTO_DSTOPTS); 334 MAKE_CHAIN(exthdrs.ip6e_rthdr, mprev, nexthdrp, 335 IPPROTO_ROUTING); 336 } 337 338 /* 339 * If there is a routing header, replace the destination address field 340 * with the first hop of the routing header. 341 */ 342 if (exthdrs.ip6e_rthdr) { 343 struct ip6_rthdr *rh; 344 struct ip6_rthdr0 *rh0; 345 struct in6_addr *addr; 346 347 rh = (struct ip6_rthdr *)(mtod(exthdrs.ip6e_rthdr, 348 struct ip6_rthdr *)); 349 switch (rh->ip6r_type) { 350 case IPV6_RTHDR_TYPE_0: 351 rh0 = (struct ip6_rthdr0 *)rh; 352 addr = (struct in6_addr *)(rh0 + 1); 353 ip6->ip6_dst = addr[0]; 354 bcopy(&addr[1], &addr[0], 355 sizeof(struct in6_addr) * (rh0->ip6r0_segleft - 1)); 356 addr[rh0->ip6r0_segleft - 1] = finaldst; 357 break; 358 default: /* is it possible? */ 359 error = EINVAL; 360 goto bad; 361 } 362 } 363 364 /* Source address validation */ 365 if (!(flags & IPV6_UNSPECSRC) && 366 IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src)) { 367 /* 368 * XXX: we can probably assume validation in the caller, but 369 * we explicitly check the address here for safety. 370 */ 371 error = EOPNOTSUPP; 372 ip6stat_inc(ip6s_badscope); 373 goto bad; 374 } 375 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) { 376 error = EOPNOTSUPP; 377 ip6stat_inc(ip6s_badscope); 378 goto bad; 379 } 380 381 ip6stat_inc(ip6s_localout); 382 383 /* 384 * Route packet. 385 */ 386 #if NPF > 0 387 reroute: 388 #endif 389 390 /* initialize cached route */ 391 if (ro == NULL) { 392 ro = &ip6route; 393 bzero((caddr_t)ro, sizeof(*ro)); 394 } 395 ro_pmtu = ro; 396 if (opt && opt->ip6po_rthdr) 397 ro = &opt->ip6po_route; 398 dst = &ro->ro_dst; 399 400 /* 401 * if specified, try to fill in the traffic class field. 402 * do not override if a non-zero value is already set. 403 * we check the diffserv field and the ecn field separately. 404 */ 405 if (opt && opt->ip6po_tclass >= 0) { 406 int mask = 0; 407 408 if ((ip6->ip6_flow & htonl(0xfc << 20)) == 0) 409 mask |= 0xfc; 410 if ((ip6->ip6_flow & htonl(0x03 << 20)) == 0) 411 mask |= 0x03; 412 if (mask != 0) 413 ip6->ip6_flow |= 414 htonl((opt->ip6po_tclass & mask) << 20); 415 } 416 417 /* fill in or override the hop limit field, if necessary. */ 418 if (opt && opt->ip6po_hlim != -1) 419 ip6->ip6_hlim = opt->ip6po_hlim & 0xff; 420 else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) { 421 if (im6o != NULL) 422 ip6->ip6_hlim = im6o->im6o_hlim; 423 else 424 ip6->ip6_hlim = ip6_defmcasthlim; 425 } 426 427 #ifdef IPSEC 428 if (tdb) { 429 /* 430 * XXX what should we do if ip6_hlim == 0 and the 431 * packet gets tunneled? 432 */ 433 /* 434 * if we are source-routing, do not attempt to tunnel the 435 * packet just because ip6_dst is different from what tdb has. 436 * XXX 437 */ 438 error = ip6_output_ipsec_send(tdb, m, ro, 439 exthdrs.ip6e_rthdr ? 1 : 0, 0); 440 goto done; 441 } 442 #endif /* IPSEC */ 443 444 bzero(&dstsock, sizeof(dstsock)); 445 dstsock.sin6_family = AF_INET6; 446 dstsock.sin6_addr = ip6->ip6_dst; 447 dstsock.sin6_len = sizeof(dstsock); 448 ro->ro_tableid = m->m_pkthdr.ph_rtableid; 449 450 if (IN6_IS_ADDR_MULTICAST(&dstsock.sin6_addr)) { 451 struct in6_pktinfo *pi = NULL; 452 453 /* 454 * If the caller specify the outgoing interface 455 * explicitly, use it. 456 */ 457 if (opt != NULL && (pi = opt->ip6po_pktinfo) != NULL) 458 ifp = if_get(pi->ipi6_ifindex); 459 460 if (ifp == NULL && im6o != NULL) 461 ifp = if_get(im6o->im6o_ifidx); 462 } 463 464 if (ifp == NULL) { 465 rt = in6_selectroute(&dstsock, opt, ro, ro->ro_tableid); 466 if (rt == NULL) { 467 ip6stat_inc(ip6s_noroute); 468 error = EHOSTUNREACH; 469 goto bad; 470 } 471 if (ISSET(rt->rt_flags, RTF_LOCAL)) 472 ifp = if_get(rtable_loindex(m->m_pkthdr.ph_rtableid)); 473 else 474 ifp = if_get(rt->rt_ifidx); 475 /* 476 * We aren't using rtisvalid() here because the UP/DOWN state 477 * machine is broken with some Ethernet drivers like em(4). 478 * As a result we might try to use an invalid cached route 479 * entry while an interface is being detached. 480 */ 481 if (ifp == NULL) { 482 ip6stat_inc(ip6s_noroute); 483 error = EHOSTUNREACH; 484 goto bad; 485 } 486 } else { 487 *dst = dstsock; 488 } 489 490 if (rt && (rt->rt_flags & RTF_GATEWAY) && 491 !IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) 492 dst = satosin6(rt->rt_gateway); 493 494 if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) { 495 /* Unicast */ 496 497 m->m_flags &= ~(M_BCAST | M_MCAST); /* just in case */ 498 } else { 499 /* Multicast */ 500 501 m->m_flags = (m->m_flags & ~M_BCAST) | M_MCAST; 502 503 /* 504 * Confirm that the outgoing interface supports multicast. 505 */ 506 if ((ifp->if_flags & IFF_MULTICAST) == 0) { 507 ip6stat_inc(ip6s_noroute); 508 error = ENETUNREACH; 509 goto bad; 510 } 511 512 if ((im6o == NULL || im6o->im6o_loop) && 513 in6_hasmulti(&ip6->ip6_dst, ifp)) { 514 /* 515 * If we belong to the destination multicast group 516 * on the outgoing interface, and the caller did not 517 * forbid loopback, loop back a copy. 518 * Can't defer TCP/UDP checksumming, do the 519 * computation now. 520 */ 521 in6_proto_cksum_out(m, NULL); 522 ip6_mloopback(ifp, m, dst); 523 } 524 #ifdef MROUTING 525 else { 526 /* 527 * If we are acting as a multicast router, perform 528 * multicast forwarding as if the packet had just 529 * arrived on the interface to which we are about 530 * to send. The multicast forwarding function 531 * recursively calls this function, using the 532 * IPV6_FORWARDING flag to prevent infinite recursion. 533 * 534 * Multicasts that are looped back by ip6_mloopback(), 535 * above, will be forwarded by the ip6_input() routine, 536 * if necessary. 537 */ 538 if (ip6_mforwarding && ip6_mrouter[ifp->if_rdomain] && 539 (flags & IPV6_FORWARDING) == 0) { 540 if (ip6_mforward(ip6, ifp, m) != 0) { 541 m_freem(m); 542 goto done; 543 } 544 } 545 } 546 #endif 547 /* 548 * Multicasts with a hoplimit of zero may be looped back, 549 * above, but must not be transmitted on a network. 550 * Also, multicasts addressed to the loopback interface 551 * are not sent -- the above call to ip6_mloopback() will 552 * loop back a copy if this host actually belongs to the 553 * destination group on the loopback interface. 554 */ 555 if (ip6->ip6_hlim == 0 || (ifp->if_flags & IFF_LOOPBACK) || 556 IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst)) { 557 m_freem(m); 558 goto done; 559 } 560 } 561 562 /* 563 * If this packet is going through a loopback interface we won't 564 * be able to restore its scope ID using the interface index. 565 */ 566 if (IN6_IS_SCOPE_EMBED(&ip6->ip6_src)) { 567 if (ifp->if_flags & IFF_LOOPBACK) 568 src_scope = ip6->ip6_src.s6_addr16[1]; 569 ip6->ip6_src.s6_addr16[1] = 0; 570 } 571 if (IN6_IS_SCOPE_EMBED(&ip6->ip6_dst)) { 572 if (ifp->if_flags & IFF_LOOPBACK) 573 dst_scope = ip6->ip6_dst.s6_addr16[1]; 574 ip6->ip6_dst.s6_addr16[1] = 0; 575 } 576 577 /* Determine path MTU. */ 578 if ((error = ip6_getpmtu(ro_pmtu->ro_rt, ifp, &mtu)) != 0) 579 goto bad; 580 581 /* 582 * The caller of this function may specify to use the minimum MTU 583 * in some cases. 584 * An advanced API option (IPV6_USE_MIN_MTU) can also override MTU 585 * setting. The logic is a bit complicated; by default, unicast 586 * packets will follow path MTU while multicast packets will be sent at 587 * the minimum MTU. If IP6PO_MINMTU_ALL is specified, all packets 588 * including unicast ones will be sent at the minimum MTU. Multicast 589 * packets will always be sent at the minimum MTU unless 590 * IP6PO_MINMTU_DISABLE is explicitly specified. 591 * See RFC 3542 for more details. 592 */ 593 if (mtu > IPV6_MMTU) { 594 if ((flags & IPV6_MINMTU)) 595 mtu = IPV6_MMTU; 596 else if (opt && opt->ip6po_minmtu == IP6PO_MINMTU_ALL) 597 mtu = IPV6_MMTU; 598 else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) && (opt == NULL || 599 opt->ip6po_minmtu != IP6PO_MINMTU_DISABLE)) { 600 mtu = IPV6_MMTU; 601 } 602 } 603 604 /* 605 * If the outgoing packet contains a hop-by-hop options header, 606 * it must be examined and processed even by the source node. 607 * (RFC 2460, section 4.) 608 */ 609 if (exthdrs.ip6e_hbh) { 610 struct ip6_hbh *hbh = mtod(exthdrs.ip6e_hbh, struct ip6_hbh *); 611 u_int32_t rtalert; /* returned value is ignored */ 612 u_int32_t plen = 0; /* no more than 1 jumbo payload option! */ 613 614 m->m_pkthdr.ph_ifidx = ifp->if_index; 615 if (ip6_process_hopopts(m, (u_int8_t *)(hbh + 1), 616 ((hbh->ip6h_len + 1) << 3) - sizeof(struct ip6_hbh), 617 &rtalert, &plen) < 0) { 618 /* m was already freed at this point */ 619 error = EINVAL;/* better error? */ 620 goto done; 621 } 622 m->m_pkthdr.ph_ifidx = 0; 623 } 624 625 #if NPF > 0 626 if (pf_test(AF_INET6, PF_OUT, ifp, &m) != PF_PASS) { 627 error = EACCES; 628 m_freem(m); 629 goto done; 630 } 631 if (m == NULL) 632 goto done; 633 ip6 = mtod(m, struct ip6_hdr *); 634 if ((m->m_pkthdr.pf.flags & (PF_TAG_REROUTE | PF_TAG_GENERATED)) == 635 (PF_TAG_REROUTE | PF_TAG_GENERATED)) { 636 /* already rerun the route lookup, go on */ 637 m->m_pkthdr.pf.flags &= ~(PF_TAG_GENERATED | PF_TAG_REROUTE); 638 } else if (m->m_pkthdr.pf.flags & PF_TAG_REROUTE) { 639 /* tag as generated to skip over pf_test on rerun */ 640 m->m_pkthdr.pf.flags |= PF_TAG_GENERATED; 641 finaldst = ip6->ip6_dst; 642 ro = NULL; 643 if_put(ifp); /* drop reference since destination changed */ 644 ifp = NULL; 645 goto reroute; 646 } 647 #endif 648 649 /* 650 * If the packet is not going on the wire it can be destinated 651 * to any local address. In this case do not clear its scopes 652 * to let ip6_input() find a matching local route. 653 */ 654 if (ifp->if_flags & IFF_LOOPBACK) { 655 if (IN6_IS_SCOPE_EMBED(&ip6->ip6_src)) 656 ip6->ip6_src.s6_addr16[1] = src_scope; 657 if (IN6_IS_SCOPE_EMBED(&ip6->ip6_dst)) 658 ip6->ip6_dst.s6_addr16[1] = dst_scope; 659 } 660 661 in6_proto_cksum_out(m, ifp); 662 663 /* 664 * Send the packet to the outgoing interface. 665 * If necessary, do IPv6 fragmentation before sending. 666 * 667 * the logic here is rather complex: 668 * 1: normal case (dontfrag == 0) 669 * 1-a: send as is if tlen <= path mtu 670 * 1-b: fragment if tlen > path mtu 671 * 672 * 2: if user asks us not to fragment (dontfrag == 1) 673 * 2-a: send as is if tlen <= interface mtu 674 * 2-b: error if tlen > interface mtu 675 */ 676 tlen = m->m_pkthdr.len; 677 678 if (ISSET(m->m_pkthdr.csum_flags, M_IPV6_DF_OUT)) { 679 CLR(m->m_pkthdr.csum_flags, M_IPV6_DF_OUT); 680 dontfrag = 1; 681 } else if (opt && ISSET(opt->ip6po_flags, IP6PO_DONTFRAG)) 682 dontfrag = 1; 683 else 684 dontfrag = 0; 685 if (dontfrag && tlen > ifp->if_mtu) { /* case 2-b */ 686 #ifdef IPSEC 687 if (ip_mtudisc) 688 ipsec_adjust_mtu(m, mtu); 689 #endif 690 error = EMSGSIZE; 691 goto bad; 692 } 693 694 /* 695 * transmit packet without fragmentation 696 */ 697 if (dontfrag || (tlen <= mtu)) { /* case 1-a and 2-a */ 698 error = ifp->if_output(ifp, m, sin6tosa(dst), ro->ro_rt); 699 goto done; 700 } 701 702 /* 703 * try to fragment the packet. case 1-b 704 */ 705 if (mtu < IPV6_MMTU) { 706 /* path MTU cannot be less than IPV6_MMTU */ 707 error = EMSGSIZE; 708 goto bad; 709 } else if (ip6->ip6_plen == 0) { 710 /* jumbo payload cannot be fragmented */ 711 error = EMSGSIZE; 712 goto bad; 713 } 714 715 /* 716 * Too large for the destination or interface; 717 * fragment if possible. 718 * Must be able to put at least 8 bytes per fragment. 719 */ 720 hlen = unfragpartlen; 721 if (mtu > IPV6_MAXPACKET) 722 mtu = IPV6_MAXPACKET; 723 724 /* 725 * Change the next header field of the last header in the 726 * unfragmentable part. 727 */ 728 if (exthdrs.ip6e_rthdr) { 729 nextproto = *mtod(exthdrs.ip6e_rthdr, u_char *); 730 *mtod(exthdrs.ip6e_rthdr, u_char *) = IPPROTO_FRAGMENT; 731 } else if (exthdrs.ip6e_dest1) { 732 nextproto = *mtod(exthdrs.ip6e_dest1, u_char *); 733 *mtod(exthdrs.ip6e_dest1, u_char *) = IPPROTO_FRAGMENT; 734 } else if (exthdrs.ip6e_hbh) { 735 nextproto = *mtod(exthdrs.ip6e_hbh, u_char *); 736 *mtod(exthdrs.ip6e_hbh, u_char *) = IPPROTO_FRAGMENT; 737 } else { 738 nextproto = ip6->ip6_nxt; 739 ip6->ip6_nxt = IPPROTO_FRAGMENT; 740 } 741 742 error = ip6_fragment(m, &fml, hlen, nextproto, mtu); 743 if (error) 744 goto done; 745 746 while ((m = ml_dequeue(&fml)) != NULL) { 747 error = ifp->if_output(ifp, m, sin6tosa(dst), ro->ro_rt); 748 if (error) 749 break; 750 } 751 if (error) 752 ml_purge(&fml); 753 else 754 ip6stat_inc(ip6s_fragmented); 755 756 done: 757 if_put(ifp); 758 if (ro == &ip6route && ro->ro_rt) { 759 rtfree(ro->ro_rt); 760 } else if (ro_pmtu == &ip6route && ro_pmtu->ro_rt) { 761 rtfree(ro_pmtu->ro_rt); 762 } 763 return (error); 764 765 freehdrs: 766 m_freem(exthdrs.ip6e_hbh); /* m_freem will check if mbuf is 0 */ 767 m_freem(exthdrs.ip6e_dest1); 768 m_freem(exthdrs.ip6e_rthdr); 769 m_freem(exthdrs.ip6e_dest2); 770 /* FALLTHROUGH */ 771 bad: 772 m_freem(m); 773 goto done; 774 } 775 776 int 777 ip6_fragment(struct mbuf *m0, struct mbuf_list *fml, int hlen, 778 u_char nextproto, u_long mtu) 779 { 780 struct mbuf *m, *m_frgpart; 781 struct ip6_hdr *mhip6; 782 struct ip6_frag *ip6f; 783 u_int32_t id; 784 int tlen, len, off; 785 int error; 786 787 ml_init(fml); 788 789 tlen = m0->m_pkthdr.len; 790 len = (mtu - hlen - sizeof(struct ip6_frag)) & ~7; 791 if (len < 8) { 792 error = EMSGSIZE; 793 goto bad; 794 } 795 796 id = htonl(ip6_randomid()); 797 798 /* 799 * Loop through length of segment after first fragment, 800 * make new header and copy data of each part and link onto chain. 801 */ 802 for (off = hlen; off < tlen; off += len) { 803 struct mbuf *mlast; 804 805 MGETHDR(m, M_DONTWAIT, MT_HEADER); 806 if (m == NULL) { 807 error = ENOBUFS; 808 goto bad; 809 } 810 ml_enqueue(fml, m); 811 if ((error = m_dup_pkthdr(m, m0, M_DONTWAIT)) != 0) 812 goto bad; 813 m->m_data += max_linkhdr; 814 mhip6 = mtod(m, struct ip6_hdr *); 815 *mhip6 = *mtod(m0, struct ip6_hdr *); 816 m->m_len = sizeof(*mhip6); 817 if ((error = ip6_insertfraghdr(m0, m, hlen, &ip6f)) != 0) 818 goto bad; 819 ip6f->ip6f_offlg = htons((u_int16_t)((off - hlen) & ~7)); 820 if (off + len >= tlen) 821 len = tlen - off; 822 else 823 ip6f->ip6f_offlg |= IP6F_MORE_FRAG; 824 mhip6->ip6_plen = htons((u_int16_t)(len + hlen + 825 sizeof(*ip6f) - sizeof(struct ip6_hdr))); 826 if ((m_frgpart = m_copym(m0, off, len, M_DONTWAIT)) == NULL) { 827 error = ENOBUFS; 828 goto bad; 829 } 830 for (mlast = m; mlast->m_next; mlast = mlast->m_next) 831 ; 832 mlast->m_next = m_frgpart; 833 m->m_pkthdr.len = len + hlen + sizeof(*ip6f); 834 ip6f->ip6f_reserved = 0; 835 ip6f->ip6f_ident = id; 836 ip6f->ip6f_nxt = nextproto; 837 } 838 839 ip6stat_add(ip6s_ofragments, ml_len(fml)); 840 m_freem(m0); 841 return (0); 842 843 bad: 844 ip6stat_inc(ip6s_odropped); 845 ml_purge(fml); 846 m_freem(m0); 847 return (error); 848 } 849 850 int 851 ip6_copyexthdr(struct mbuf **mp, caddr_t hdr, int hlen) 852 { 853 struct mbuf *m; 854 855 if (hlen > MCLBYTES) 856 return (ENOBUFS); /* XXX */ 857 858 MGET(m, M_DONTWAIT, MT_DATA); 859 if (!m) 860 return (ENOBUFS); 861 862 if (hlen > MLEN) { 863 MCLGET(m, M_DONTWAIT); 864 if ((m->m_flags & M_EXT) == 0) { 865 m_free(m); 866 return (ENOBUFS); 867 } 868 } 869 m->m_len = hlen; 870 if (hdr) 871 memcpy(mtod(m, caddr_t), hdr, hlen); 872 873 *mp = m; 874 return (0); 875 } 876 877 /* 878 * Insert jumbo payload option. 879 */ 880 int 881 ip6_insert_jumboopt(struct ip6_exthdrs *exthdrs, u_int32_t plen) 882 { 883 struct mbuf *mopt; 884 u_int8_t *optbuf; 885 u_int32_t v; 886 887 #define JUMBOOPTLEN 8 /* length of jumbo payload option and padding */ 888 889 /* 890 * If there is no hop-by-hop options header, allocate new one. 891 * If there is one but it doesn't have enough space to store the 892 * jumbo payload option, allocate a cluster to store the whole options. 893 * Otherwise, use it to store the options. 894 */ 895 if (exthdrs->ip6e_hbh == 0) { 896 MGET(mopt, M_DONTWAIT, MT_DATA); 897 if (mopt == NULL) 898 return (ENOBUFS); 899 mopt->m_len = JUMBOOPTLEN; 900 optbuf = mtod(mopt, u_int8_t *); 901 optbuf[1] = 0; /* = ((JUMBOOPTLEN) >> 3) - 1 */ 902 exthdrs->ip6e_hbh = mopt; 903 } else { 904 struct ip6_hbh *hbh; 905 906 mopt = exthdrs->ip6e_hbh; 907 if (m_trailingspace(mopt) < JUMBOOPTLEN) { 908 /* 909 * XXX assumption: 910 * - exthdrs->ip6e_hbh is not referenced from places 911 * other than exthdrs. 912 * - exthdrs->ip6e_hbh is not an mbuf chain. 913 */ 914 int oldoptlen = mopt->m_len; 915 struct mbuf *n; 916 917 /* 918 * XXX: give up if the whole (new) hbh header does 919 * not fit even in an mbuf cluster. 920 */ 921 if (oldoptlen + JUMBOOPTLEN > MCLBYTES) 922 return (ENOBUFS); 923 924 /* 925 * As a consequence, we must always prepare a cluster 926 * at this point. 927 */ 928 MGET(n, M_DONTWAIT, MT_DATA); 929 if (n) { 930 MCLGET(n, M_DONTWAIT); 931 if ((n->m_flags & M_EXT) == 0) { 932 m_freem(n); 933 n = NULL; 934 } 935 } 936 if (!n) 937 return (ENOBUFS); 938 n->m_len = oldoptlen + JUMBOOPTLEN; 939 memcpy(mtod(n, caddr_t), mtod(mopt, caddr_t), 940 oldoptlen); 941 optbuf = mtod(n, u_int8_t *) + oldoptlen; 942 m_freem(mopt); 943 mopt = exthdrs->ip6e_hbh = n; 944 } else { 945 optbuf = mtod(mopt, u_int8_t *) + mopt->m_len; 946 mopt->m_len += JUMBOOPTLEN; 947 } 948 optbuf[0] = IP6OPT_PADN; 949 optbuf[1] = 0; 950 951 /* 952 * Adjust the header length according to the pad and 953 * the jumbo payload option. 954 */ 955 hbh = mtod(mopt, struct ip6_hbh *); 956 hbh->ip6h_len += (JUMBOOPTLEN >> 3); 957 } 958 959 /* fill in the option. */ 960 optbuf[2] = IP6OPT_JUMBO; 961 optbuf[3] = 4; 962 v = (u_int32_t)htonl(plen + JUMBOOPTLEN); 963 memcpy(&optbuf[4], &v, sizeof(u_int32_t)); 964 965 /* finally, adjust the packet header length */ 966 exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN; 967 968 return (0); 969 #undef JUMBOOPTLEN 970 } 971 972 /* 973 * Insert fragment header and copy unfragmentable header portions. 974 */ 975 int 976 ip6_insertfraghdr(struct mbuf *m0, struct mbuf *m, int hlen, 977 struct ip6_frag **frghdrp) 978 { 979 struct mbuf *n, *mlast; 980 981 if (hlen > sizeof(struct ip6_hdr)) { 982 n = m_copym(m0, sizeof(struct ip6_hdr), 983 hlen - sizeof(struct ip6_hdr), M_DONTWAIT); 984 if (n == NULL) 985 return (ENOBUFS); 986 m->m_next = n; 987 } else 988 n = m; 989 990 /* Search for the last mbuf of unfragmentable part. */ 991 for (mlast = n; mlast->m_next; mlast = mlast->m_next) 992 ; 993 994 if ((mlast->m_flags & M_EXT) == 0 && 995 m_trailingspace(mlast) >= sizeof(struct ip6_frag)) { 996 /* use the trailing space of the last mbuf for fragment hdr */ 997 *frghdrp = (struct ip6_frag *)(mtod(mlast, caddr_t) + 998 mlast->m_len); 999 mlast->m_len += sizeof(struct ip6_frag); 1000 m->m_pkthdr.len += sizeof(struct ip6_frag); 1001 } else { 1002 /* allocate a new mbuf for the fragment header */ 1003 struct mbuf *mfrg; 1004 1005 MGET(mfrg, M_DONTWAIT, MT_DATA); 1006 if (mfrg == NULL) 1007 return (ENOBUFS); 1008 mfrg->m_len = sizeof(struct ip6_frag); 1009 *frghdrp = mtod(mfrg, struct ip6_frag *); 1010 mlast->m_next = mfrg; 1011 } 1012 1013 return (0); 1014 } 1015 1016 int 1017 ip6_getpmtu(struct rtentry *rt, struct ifnet *ifp, u_long *mtup) 1018 { 1019 u_int32_t mtu = 0; 1020 int error = 0; 1021 1022 if (rt != NULL) { 1023 mtu = rt->rt_mtu; 1024 if (mtu == 0) 1025 mtu = ifp->if_mtu; 1026 else if (mtu < IPV6_MMTU) { 1027 /* RFC8021 IPv6 Atomic Fragments Considered Harmful */ 1028 mtu = IPV6_MMTU; 1029 } else if (mtu > ifp->if_mtu) { 1030 /* 1031 * The MTU on the route is larger than the MTU on 1032 * the interface! This shouldn't happen, unless the 1033 * MTU of the interface has been changed after the 1034 * interface was brought up. Change the MTU in the 1035 * route to match the interface MTU (as long as the 1036 * field isn't locked). 1037 */ 1038 mtu = ifp->if_mtu; 1039 if (!(rt->rt_locks & RTV_MTU)) 1040 rt->rt_mtu = mtu; 1041 } 1042 } else { 1043 mtu = ifp->if_mtu; 1044 } 1045 1046 *mtup = mtu; 1047 return (error); 1048 } 1049 1050 /* 1051 * IP6 socket option processing. 1052 */ 1053 int 1054 ip6_ctloutput(int op, struct socket *so, int level, int optname, 1055 struct mbuf *m) 1056 { 1057 int privileged, optdatalen, uproto; 1058 void *optdata; 1059 struct inpcb *inp = sotoinpcb(so); 1060 int error, optval; 1061 struct proc *p = curproc; /* For IPsec and rdomain */ 1062 u_int rtableid, rtid = 0; 1063 1064 error = optval = 0; 1065 1066 privileged = (inp->inp_socket->so_state & SS_PRIV); 1067 uproto = (int)so->so_proto->pr_protocol; 1068 1069 if (level != IPPROTO_IPV6) 1070 return (EINVAL); 1071 1072 rtableid = p->p_p->ps_rtableid; 1073 1074 switch (op) { 1075 case PRCO_SETOPT: 1076 switch (optname) { 1077 /* 1078 * Use of some Hop-by-Hop options or some 1079 * Destination options, might require special 1080 * privilege. That is, normal applications 1081 * (without special privilege) might be forbidden 1082 * from setting certain options in outgoing packets, 1083 * and might never see certain options in received 1084 * packets. [RFC 2292 Section 6] 1085 * KAME specific note: 1086 * KAME prevents non-privileged users from sending or 1087 * receiving ANY hbh/dst options in order to avoid 1088 * overhead of parsing options in the kernel. 1089 */ 1090 case IPV6_RECVHOPOPTS: 1091 case IPV6_RECVDSTOPTS: 1092 if (!privileged) { 1093 error = EPERM; 1094 break; 1095 } 1096 /* FALLTHROUGH */ 1097 case IPV6_UNICAST_HOPS: 1098 case IPV6_MINHOPCOUNT: 1099 case IPV6_HOPLIMIT: 1100 1101 case IPV6_RECVPKTINFO: 1102 case IPV6_RECVHOPLIMIT: 1103 case IPV6_RECVRTHDR: 1104 case IPV6_RECVPATHMTU: 1105 case IPV6_RECVTCLASS: 1106 case IPV6_V6ONLY: 1107 case IPV6_AUTOFLOWLABEL: 1108 case IPV6_RECVDSTPORT: 1109 if (m == NULL || m->m_len != sizeof(int)) { 1110 error = EINVAL; 1111 break; 1112 } 1113 optval = *mtod(m, int *); 1114 switch (optname) { 1115 1116 case IPV6_UNICAST_HOPS: 1117 if (optval < -1 || optval >= 256) 1118 error = EINVAL; 1119 else { 1120 /* -1 = kernel default */ 1121 inp->inp_hops = optval; 1122 } 1123 break; 1124 1125 case IPV6_MINHOPCOUNT: 1126 if (optval < 0 || optval > 255) 1127 error = EINVAL; 1128 else 1129 inp->inp_ip6_minhlim = optval; 1130 break; 1131 1132 #define OPTSET(bit) \ 1133 do { \ 1134 if (optval) \ 1135 inp->inp_flags |= (bit); \ 1136 else \ 1137 inp->inp_flags &= ~(bit); \ 1138 } while (/*CONSTCOND*/ 0) 1139 #define OPTBIT(bit) (inp->inp_flags & (bit) ? 1 : 0) 1140 1141 case IPV6_RECVPKTINFO: 1142 OPTSET(IN6P_PKTINFO); 1143 break; 1144 1145 case IPV6_HOPLIMIT: 1146 { 1147 struct ip6_pktopts **optp; 1148 1149 optp = &inp->inp_outputopts6; 1150 error = ip6_pcbopt(IPV6_HOPLIMIT, 1151 (u_char *)&optval, sizeof(optval), optp, 1152 privileged, uproto); 1153 break; 1154 } 1155 1156 case IPV6_RECVHOPLIMIT: 1157 OPTSET(IN6P_HOPLIMIT); 1158 break; 1159 1160 case IPV6_RECVHOPOPTS: 1161 OPTSET(IN6P_HOPOPTS); 1162 break; 1163 1164 case IPV6_RECVDSTOPTS: 1165 OPTSET(IN6P_DSTOPTS); 1166 break; 1167 1168 case IPV6_RECVRTHDR: 1169 OPTSET(IN6P_RTHDR); 1170 break; 1171 1172 case IPV6_RECVPATHMTU: 1173 /* 1174 * We ignore this option for TCP 1175 * sockets. 1176 * (RFC3542 leaves this case 1177 * unspecified.) 1178 */ 1179 if (uproto != IPPROTO_TCP) 1180 OPTSET(IN6P_MTU); 1181 break; 1182 1183 case IPV6_V6ONLY: 1184 /* 1185 * make setsockopt(IPV6_V6ONLY) 1186 * available only prior to bind(2). 1187 * see ipng mailing list, Jun 22 2001. 1188 */ 1189 if (inp->inp_lport || !IN6_IS_ADDR_UNSPECIFIED( 1190 &inp->inp_laddr6)) { 1191 error = EINVAL; 1192 break; 1193 } 1194 /* No support for IPv4-mapped addresses. */ 1195 if (!optval) 1196 error = EINVAL; 1197 else 1198 error = 0; 1199 break; 1200 case IPV6_RECVTCLASS: 1201 OPTSET(IN6P_TCLASS); 1202 break; 1203 case IPV6_AUTOFLOWLABEL: 1204 OPTSET(IN6P_AUTOFLOWLABEL); 1205 break; 1206 1207 case IPV6_RECVDSTPORT: 1208 OPTSET(IN6P_RECVDSTPORT); 1209 break; 1210 } 1211 break; 1212 1213 case IPV6_TCLASS: 1214 case IPV6_DONTFRAG: 1215 case IPV6_USE_MIN_MTU: 1216 if (m == NULL || m->m_len != sizeof(optval)) { 1217 error = EINVAL; 1218 break; 1219 } 1220 optval = *mtod(m, int *); 1221 { 1222 struct ip6_pktopts **optp; 1223 optp = &inp->inp_outputopts6; 1224 error = ip6_pcbopt(optname, (u_char *)&optval, 1225 sizeof(optval), optp, privileged, uproto); 1226 break; 1227 } 1228 1229 case IPV6_PKTINFO: 1230 case IPV6_HOPOPTS: 1231 case IPV6_RTHDR: 1232 case IPV6_DSTOPTS: 1233 case IPV6_RTHDRDSTOPTS: 1234 { 1235 /* new advanced API (RFC3542) */ 1236 u_char *optbuf; 1237 int optbuflen; 1238 struct ip6_pktopts **optp; 1239 1240 if (m && m->m_next) { 1241 error = EINVAL; /* XXX */ 1242 break; 1243 } 1244 if (m) { 1245 optbuf = mtod(m, u_char *); 1246 optbuflen = m->m_len; 1247 } else { 1248 optbuf = NULL; 1249 optbuflen = 0; 1250 } 1251 optp = &inp->inp_outputopts6; 1252 error = ip6_pcbopt(optname, optbuf, optbuflen, optp, 1253 privileged, uproto); 1254 break; 1255 } 1256 #undef OPTSET 1257 1258 case IPV6_MULTICAST_IF: 1259 case IPV6_MULTICAST_HOPS: 1260 case IPV6_MULTICAST_LOOP: 1261 case IPV6_JOIN_GROUP: 1262 case IPV6_LEAVE_GROUP: 1263 error = ip6_setmoptions(optname, 1264 &inp->inp_moptions6, 1265 m, inp->inp_rtableid); 1266 break; 1267 1268 case IPV6_PORTRANGE: 1269 if (m == NULL || m->m_len != sizeof(int)) { 1270 error = EINVAL; 1271 break; 1272 } 1273 optval = *mtod(m, int *); 1274 1275 switch (optval) { 1276 case IPV6_PORTRANGE_DEFAULT: 1277 inp->inp_flags &= ~(IN6P_LOWPORT); 1278 inp->inp_flags &= ~(IN6P_HIGHPORT); 1279 break; 1280 1281 case IPV6_PORTRANGE_HIGH: 1282 inp->inp_flags &= ~(IN6P_LOWPORT); 1283 inp->inp_flags |= IN6P_HIGHPORT; 1284 break; 1285 1286 case IPV6_PORTRANGE_LOW: 1287 inp->inp_flags &= ~(IN6P_HIGHPORT); 1288 inp->inp_flags |= IN6P_LOWPORT; 1289 break; 1290 1291 default: 1292 error = EINVAL; 1293 break; 1294 } 1295 break; 1296 1297 case IPSEC6_OUTSA: 1298 error = EINVAL; 1299 break; 1300 1301 case IPV6_AUTH_LEVEL: 1302 case IPV6_ESP_TRANS_LEVEL: 1303 case IPV6_ESP_NETWORK_LEVEL: 1304 case IPV6_IPCOMP_LEVEL: 1305 #ifndef IPSEC 1306 error = EINVAL; 1307 #else 1308 if (m == NULL || m->m_len != sizeof(int)) { 1309 error = EINVAL; 1310 break; 1311 } 1312 optval = *mtod(m, int *); 1313 1314 if (optval < IPSEC_LEVEL_BYPASS || 1315 optval > IPSEC_LEVEL_UNIQUE) { 1316 error = EINVAL; 1317 break; 1318 } 1319 1320 switch (optname) { 1321 case IPV6_AUTH_LEVEL: 1322 if (optval < IPSEC_AUTH_LEVEL_DEFAULT && 1323 suser(p)) { 1324 error = EACCES; 1325 break; 1326 } 1327 inp->inp_seclevel[SL_AUTH] = optval; 1328 break; 1329 1330 case IPV6_ESP_TRANS_LEVEL: 1331 if (optval < IPSEC_ESP_TRANS_LEVEL_DEFAULT && 1332 suser(p)) { 1333 error = EACCES; 1334 break; 1335 } 1336 inp->inp_seclevel[SL_ESP_TRANS] = optval; 1337 break; 1338 1339 case IPV6_ESP_NETWORK_LEVEL: 1340 if (optval < IPSEC_ESP_NETWORK_LEVEL_DEFAULT && 1341 suser(p)) { 1342 error = EACCES; 1343 break; 1344 } 1345 inp->inp_seclevel[SL_ESP_NETWORK] = optval; 1346 break; 1347 1348 case IPV6_IPCOMP_LEVEL: 1349 if (optval < IPSEC_IPCOMP_LEVEL_DEFAULT && 1350 suser(p)) { 1351 error = EACCES; 1352 break; 1353 } 1354 inp->inp_seclevel[SL_IPCOMP] = optval; 1355 break; 1356 } 1357 #endif 1358 break; 1359 case SO_RTABLE: 1360 if (m == NULL || m->m_len < sizeof(u_int)) { 1361 error = EINVAL; 1362 break; 1363 } 1364 rtid = *mtod(m, u_int *); 1365 if (inp->inp_rtableid == rtid) 1366 break; 1367 /* needs privileges to switch when already set */ 1368 if (rtableid != rtid && rtableid != 0 && 1369 (error = suser(p)) != 0) 1370 break; 1371 /* table must exist */ 1372 if (!rtable_exists(rtid)) { 1373 error = EINVAL; 1374 break; 1375 } 1376 if (inp->inp_lport) { 1377 error = EBUSY; 1378 break; 1379 } 1380 inp->inp_rtableid = rtid; 1381 in_pcbrehash(inp); 1382 break; 1383 case IPV6_PIPEX: 1384 if (m != NULL && m->m_len == sizeof(int)) 1385 inp->inp_pipex = *mtod(m, int *); 1386 else 1387 error = EINVAL; 1388 break; 1389 1390 default: 1391 error = ENOPROTOOPT; 1392 break; 1393 } 1394 break; 1395 1396 case PRCO_GETOPT: 1397 switch (optname) { 1398 1399 case IPV6_RECVHOPOPTS: 1400 case IPV6_RECVDSTOPTS: 1401 case IPV6_UNICAST_HOPS: 1402 case IPV6_MINHOPCOUNT: 1403 case IPV6_RECVPKTINFO: 1404 case IPV6_RECVHOPLIMIT: 1405 case IPV6_RECVRTHDR: 1406 case IPV6_RECVPATHMTU: 1407 1408 case IPV6_V6ONLY: 1409 case IPV6_PORTRANGE: 1410 case IPV6_RECVTCLASS: 1411 case IPV6_AUTOFLOWLABEL: 1412 case IPV6_RECVDSTPORT: 1413 switch (optname) { 1414 1415 case IPV6_RECVHOPOPTS: 1416 optval = OPTBIT(IN6P_HOPOPTS); 1417 break; 1418 1419 case IPV6_RECVDSTOPTS: 1420 optval = OPTBIT(IN6P_DSTOPTS); 1421 break; 1422 1423 case IPV6_UNICAST_HOPS: 1424 optval = inp->inp_hops; 1425 break; 1426 1427 case IPV6_MINHOPCOUNT: 1428 optval = inp->inp_ip6_minhlim; 1429 break; 1430 1431 case IPV6_RECVPKTINFO: 1432 optval = OPTBIT(IN6P_PKTINFO); 1433 break; 1434 1435 case IPV6_RECVHOPLIMIT: 1436 optval = OPTBIT(IN6P_HOPLIMIT); 1437 break; 1438 1439 case IPV6_RECVRTHDR: 1440 optval = OPTBIT(IN6P_RTHDR); 1441 break; 1442 1443 case IPV6_RECVPATHMTU: 1444 optval = OPTBIT(IN6P_MTU); 1445 break; 1446 1447 case IPV6_V6ONLY: 1448 optval = 1; 1449 break; 1450 1451 case IPV6_PORTRANGE: 1452 { 1453 int flags; 1454 flags = inp->inp_flags; 1455 if (flags & IN6P_HIGHPORT) 1456 optval = IPV6_PORTRANGE_HIGH; 1457 else if (flags & IN6P_LOWPORT) 1458 optval = IPV6_PORTRANGE_LOW; 1459 else 1460 optval = 0; 1461 break; 1462 } 1463 case IPV6_RECVTCLASS: 1464 optval = OPTBIT(IN6P_TCLASS); 1465 break; 1466 1467 case IPV6_AUTOFLOWLABEL: 1468 optval = OPTBIT(IN6P_AUTOFLOWLABEL); 1469 break; 1470 1471 case IPV6_RECVDSTPORT: 1472 optval = OPTBIT(IN6P_RECVDSTPORT); 1473 break; 1474 } 1475 if (error) 1476 break; 1477 m->m_len = sizeof(int); 1478 *mtod(m, int *) = optval; 1479 break; 1480 1481 case IPV6_PATHMTU: 1482 { 1483 u_long pmtu = 0; 1484 struct ip6_mtuinfo mtuinfo; 1485 struct ifnet *ifp; 1486 struct rtentry *rt; 1487 1488 if (!(so->so_state & SS_ISCONNECTED)) 1489 return (ENOTCONN); 1490 1491 rt = in_pcbrtentry(inp); 1492 if (!rtisvalid(rt)) 1493 return (EHOSTUNREACH); 1494 1495 ifp = if_get(rt->rt_ifidx); 1496 if (ifp == NULL) 1497 return (EHOSTUNREACH); 1498 /* 1499 * XXX: we dot not consider the case of source 1500 * routing, or optional information to specify 1501 * the outgoing interface. 1502 */ 1503 error = ip6_getpmtu(rt, ifp, &pmtu); 1504 if_put(ifp); 1505 if (error) 1506 break; 1507 if (pmtu > IPV6_MAXPACKET) 1508 pmtu = IPV6_MAXPACKET; 1509 1510 bzero(&mtuinfo, sizeof(mtuinfo)); 1511 mtuinfo.ip6m_mtu = (u_int32_t)pmtu; 1512 optdata = (void *)&mtuinfo; 1513 optdatalen = sizeof(mtuinfo); 1514 if (optdatalen > MCLBYTES) 1515 return (EMSGSIZE); /* XXX */ 1516 if (optdatalen > MLEN) 1517 MCLGET(m, M_WAIT); 1518 m->m_len = optdatalen; 1519 bcopy(optdata, mtod(m, void *), optdatalen); 1520 break; 1521 } 1522 1523 case IPV6_PKTINFO: 1524 case IPV6_HOPOPTS: 1525 case IPV6_RTHDR: 1526 case IPV6_DSTOPTS: 1527 case IPV6_RTHDRDSTOPTS: 1528 case IPV6_TCLASS: 1529 case IPV6_DONTFRAG: 1530 case IPV6_USE_MIN_MTU: 1531 error = ip6_getpcbopt(inp->inp_outputopts6, 1532 optname, m); 1533 break; 1534 1535 case IPV6_MULTICAST_IF: 1536 case IPV6_MULTICAST_HOPS: 1537 case IPV6_MULTICAST_LOOP: 1538 case IPV6_JOIN_GROUP: 1539 case IPV6_LEAVE_GROUP: 1540 error = ip6_getmoptions(optname, 1541 inp->inp_moptions6, m); 1542 break; 1543 1544 case IPSEC6_OUTSA: 1545 error = EINVAL; 1546 break; 1547 1548 case IPV6_AUTH_LEVEL: 1549 case IPV6_ESP_TRANS_LEVEL: 1550 case IPV6_ESP_NETWORK_LEVEL: 1551 case IPV6_IPCOMP_LEVEL: 1552 #ifndef IPSEC 1553 m->m_len = sizeof(int); 1554 *mtod(m, int *) = IPSEC_LEVEL_NONE; 1555 #else 1556 m->m_len = sizeof(int); 1557 switch (optname) { 1558 case IPV6_AUTH_LEVEL: 1559 optval = inp->inp_seclevel[SL_AUTH]; 1560 break; 1561 1562 case IPV6_ESP_TRANS_LEVEL: 1563 optval = 1564 inp->inp_seclevel[SL_ESP_TRANS]; 1565 break; 1566 1567 case IPV6_ESP_NETWORK_LEVEL: 1568 optval = 1569 inp->inp_seclevel[SL_ESP_NETWORK]; 1570 break; 1571 1572 case IPV6_IPCOMP_LEVEL: 1573 optval = inp->inp_seclevel[SL_IPCOMP]; 1574 break; 1575 } 1576 *mtod(m, int *) = optval; 1577 #endif 1578 break; 1579 case SO_RTABLE: 1580 m->m_len = sizeof(u_int); 1581 *mtod(m, u_int *) = inp->inp_rtableid; 1582 break; 1583 case IPV6_PIPEX: 1584 m->m_len = sizeof(int); 1585 *mtod(m, int *) = inp->inp_pipex; 1586 break; 1587 1588 default: 1589 error = ENOPROTOOPT; 1590 break; 1591 } 1592 break; 1593 } 1594 return (error); 1595 } 1596 1597 int 1598 ip6_raw_ctloutput(int op, struct socket *so, int level, int optname, 1599 struct mbuf *m) 1600 { 1601 int error = 0, optval; 1602 const int icmp6off = offsetof(struct icmp6_hdr, icmp6_cksum); 1603 struct inpcb *inp = sotoinpcb(so); 1604 1605 if (level != IPPROTO_IPV6) 1606 return (EINVAL); 1607 1608 switch (optname) { 1609 case IPV6_CHECKSUM: 1610 /* 1611 * For ICMPv6 sockets, no modification allowed for checksum 1612 * offset, permit "no change" values to help existing apps. 1613 * 1614 * RFC3542 says: "An attempt to set IPV6_CHECKSUM 1615 * for an ICMPv6 socket will fail." 1616 * The current behavior does not meet RFC3542. 1617 */ 1618 switch (op) { 1619 case PRCO_SETOPT: 1620 if (m == NULL || m->m_len != sizeof(int)) { 1621 error = EINVAL; 1622 break; 1623 } 1624 optval = *mtod(m, int *); 1625 if (optval < -1 || 1626 (optval > 0 && (optval % 2) != 0)) { 1627 /* 1628 * The API assumes non-negative even offset 1629 * values or -1 as a special value. 1630 */ 1631 error = EINVAL; 1632 } else if (so->so_proto->pr_protocol == 1633 IPPROTO_ICMPV6) { 1634 if (optval != icmp6off) 1635 error = EINVAL; 1636 } else 1637 inp->inp_cksum6 = optval; 1638 break; 1639 1640 case PRCO_GETOPT: 1641 if (so->so_proto->pr_protocol == IPPROTO_ICMPV6) 1642 optval = icmp6off; 1643 else 1644 optval = inp->inp_cksum6; 1645 1646 m->m_len = sizeof(int); 1647 *mtod(m, int *) = optval; 1648 break; 1649 1650 default: 1651 error = EINVAL; 1652 break; 1653 } 1654 break; 1655 1656 default: 1657 error = ENOPROTOOPT; 1658 break; 1659 } 1660 1661 return (error); 1662 } 1663 1664 /* 1665 * initialize ip6_pktopts. beware that there are non-zero default values in 1666 * the struct. 1667 */ 1668 void 1669 ip6_initpktopts(struct ip6_pktopts *opt) 1670 { 1671 bzero(opt, sizeof(*opt)); 1672 opt->ip6po_hlim = -1; /* -1 means default hop limit */ 1673 opt->ip6po_tclass = -1; /* -1 means default traffic class */ 1674 opt->ip6po_minmtu = IP6PO_MINMTU_MCASTONLY; 1675 } 1676 1677 int 1678 ip6_pcbopt(int optname, u_char *buf, int len, struct ip6_pktopts **pktopt, 1679 int priv, int uproto) 1680 { 1681 struct ip6_pktopts *opt; 1682 1683 if (*pktopt == NULL) { 1684 *pktopt = malloc(sizeof(struct ip6_pktopts), M_IP6OPT, 1685 M_WAITOK); 1686 ip6_initpktopts(*pktopt); 1687 } 1688 opt = *pktopt; 1689 1690 return (ip6_setpktopt(optname, buf, len, opt, priv, 1, uproto)); 1691 } 1692 1693 int 1694 ip6_getpcbopt(struct ip6_pktopts *pktopt, int optname, struct mbuf *m) 1695 { 1696 void *optdata = NULL; 1697 int optdatalen = 0; 1698 struct ip6_ext *ip6e; 1699 int error = 0; 1700 struct in6_pktinfo null_pktinfo; 1701 int deftclass = 0, on; 1702 int defminmtu = IP6PO_MINMTU_MCASTONLY; 1703 1704 switch (optname) { 1705 case IPV6_PKTINFO: 1706 if (pktopt && pktopt->ip6po_pktinfo) 1707 optdata = (void *)pktopt->ip6po_pktinfo; 1708 else { 1709 /* XXX: we don't have to do this every time... */ 1710 bzero(&null_pktinfo, sizeof(null_pktinfo)); 1711 optdata = (void *)&null_pktinfo; 1712 } 1713 optdatalen = sizeof(struct in6_pktinfo); 1714 break; 1715 case IPV6_TCLASS: 1716 if (pktopt && pktopt->ip6po_tclass >= 0) 1717 optdata = (void *)&pktopt->ip6po_tclass; 1718 else 1719 optdata = (void *)&deftclass; 1720 optdatalen = sizeof(int); 1721 break; 1722 case IPV6_HOPOPTS: 1723 if (pktopt && pktopt->ip6po_hbh) { 1724 optdata = (void *)pktopt->ip6po_hbh; 1725 ip6e = (struct ip6_ext *)pktopt->ip6po_hbh; 1726 optdatalen = (ip6e->ip6e_len + 1) << 3; 1727 } 1728 break; 1729 case IPV6_RTHDR: 1730 if (pktopt && pktopt->ip6po_rthdr) { 1731 optdata = (void *)pktopt->ip6po_rthdr; 1732 ip6e = (struct ip6_ext *)pktopt->ip6po_rthdr; 1733 optdatalen = (ip6e->ip6e_len + 1) << 3; 1734 } 1735 break; 1736 case IPV6_RTHDRDSTOPTS: 1737 if (pktopt && pktopt->ip6po_dest1) { 1738 optdata = (void *)pktopt->ip6po_dest1; 1739 ip6e = (struct ip6_ext *)pktopt->ip6po_dest1; 1740 optdatalen = (ip6e->ip6e_len + 1) << 3; 1741 } 1742 break; 1743 case IPV6_DSTOPTS: 1744 if (pktopt && pktopt->ip6po_dest2) { 1745 optdata = (void *)pktopt->ip6po_dest2; 1746 ip6e = (struct ip6_ext *)pktopt->ip6po_dest2; 1747 optdatalen = (ip6e->ip6e_len + 1) << 3; 1748 } 1749 break; 1750 case IPV6_USE_MIN_MTU: 1751 if (pktopt) 1752 optdata = (void *)&pktopt->ip6po_minmtu; 1753 else 1754 optdata = (void *)&defminmtu; 1755 optdatalen = sizeof(int); 1756 break; 1757 case IPV6_DONTFRAG: 1758 if (pktopt && ((pktopt->ip6po_flags) & IP6PO_DONTFRAG)) 1759 on = 1; 1760 else 1761 on = 0; 1762 optdata = (void *)&on; 1763 optdatalen = sizeof(on); 1764 break; 1765 default: /* should not happen */ 1766 #ifdef DIAGNOSTIC 1767 panic("%s: unexpected option", __func__); 1768 #endif 1769 return (ENOPROTOOPT); 1770 } 1771 1772 if (optdatalen > MCLBYTES) 1773 return (EMSGSIZE); /* XXX */ 1774 if (optdatalen > MLEN) 1775 MCLGET(m, M_WAIT); 1776 m->m_len = optdatalen; 1777 if (optdatalen) 1778 bcopy(optdata, mtod(m, void *), optdatalen); 1779 1780 return (error); 1781 } 1782 1783 void 1784 ip6_clearpktopts(struct ip6_pktopts *pktopt, int optname) 1785 { 1786 if (optname == -1 || optname == IPV6_PKTINFO) { 1787 if (pktopt->ip6po_pktinfo) 1788 free(pktopt->ip6po_pktinfo, M_IP6OPT, 0); 1789 pktopt->ip6po_pktinfo = NULL; 1790 } 1791 if (optname == -1 || optname == IPV6_HOPLIMIT) 1792 pktopt->ip6po_hlim = -1; 1793 if (optname == -1 || optname == IPV6_TCLASS) 1794 pktopt->ip6po_tclass = -1; 1795 if (optname == -1 || optname == IPV6_HOPOPTS) { 1796 if (pktopt->ip6po_hbh) 1797 free(pktopt->ip6po_hbh, M_IP6OPT, 0); 1798 pktopt->ip6po_hbh = NULL; 1799 } 1800 if (optname == -1 || optname == IPV6_RTHDRDSTOPTS) { 1801 if (pktopt->ip6po_dest1) 1802 free(pktopt->ip6po_dest1, M_IP6OPT, 0); 1803 pktopt->ip6po_dest1 = NULL; 1804 } 1805 if (optname == -1 || optname == IPV6_RTHDR) { 1806 if (pktopt->ip6po_rhinfo.ip6po_rhi_rthdr) 1807 free(pktopt->ip6po_rhinfo.ip6po_rhi_rthdr, M_IP6OPT, 0); 1808 pktopt->ip6po_rhinfo.ip6po_rhi_rthdr = NULL; 1809 if (pktopt->ip6po_route.ro_rt) { 1810 rtfree(pktopt->ip6po_route.ro_rt); 1811 pktopt->ip6po_route.ro_rt = NULL; 1812 } 1813 } 1814 if (optname == -1 || optname == IPV6_DSTOPTS) { 1815 if (pktopt->ip6po_dest2) 1816 free(pktopt->ip6po_dest2, M_IP6OPT, 0); 1817 pktopt->ip6po_dest2 = NULL; 1818 } 1819 } 1820 1821 #define PKTOPT_EXTHDRCPY(type) \ 1822 do {\ 1823 if (src->type) {\ 1824 size_t hlen;\ 1825 hlen = (((struct ip6_ext *)src->type)->ip6e_len + 1) << 3;\ 1826 dst->type = malloc(hlen, M_IP6OPT, M_NOWAIT);\ 1827 if (dst->type == NULL)\ 1828 goto bad;\ 1829 memcpy(dst->type, src->type, hlen);\ 1830 }\ 1831 } while (/*CONSTCOND*/ 0) 1832 1833 int 1834 copypktopts(struct ip6_pktopts *dst, struct ip6_pktopts *src) 1835 { 1836 dst->ip6po_hlim = src->ip6po_hlim; 1837 dst->ip6po_tclass = src->ip6po_tclass; 1838 dst->ip6po_flags = src->ip6po_flags; 1839 if (src->ip6po_pktinfo) { 1840 dst->ip6po_pktinfo = malloc(sizeof(*dst->ip6po_pktinfo), 1841 M_IP6OPT, M_NOWAIT); 1842 if (dst->ip6po_pktinfo == NULL) 1843 goto bad; 1844 *dst->ip6po_pktinfo = *src->ip6po_pktinfo; 1845 } 1846 PKTOPT_EXTHDRCPY(ip6po_hbh); 1847 PKTOPT_EXTHDRCPY(ip6po_dest1); 1848 PKTOPT_EXTHDRCPY(ip6po_dest2); 1849 PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */ 1850 return (0); 1851 1852 bad: 1853 ip6_clearpktopts(dst, -1); 1854 return (ENOBUFS); 1855 } 1856 #undef PKTOPT_EXTHDRCPY 1857 1858 void 1859 ip6_freepcbopts(struct ip6_pktopts *pktopt) 1860 { 1861 if (pktopt == NULL) 1862 return; 1863 1864 ip6_clearpktopts(pktopt, -1); 1865 1866 free(pktopt, M_IP6OPT, 0); 1867 } 1868 1869 /* 1870 * Set the IP6 multicast options in response to user setsockopt(). 1871 */ 1872 int 1873 ip6_setmoptions(int optname, struct ip6_moptions **im6op, struct mbuf *m, 1874 unsigned int rtableid) 1875 { 1876 int error = 0; 1877 u_int loop, ifindex; 1878 struct ipv6_mreq *mreq; 1879 struct ifnet *ifp; 1880 struct ip6_moptions *im6o = *im6op; 1881 struct in6_multi_mship *imm; 1882 struct proc *p = curproc; /* XXX */ 1883 1884 if (im6o == NULL) { 1885 /* 1886 * No multicast option buffer attached to the pcb; 1887 * allocate one and initialize to default values. 1888 */ 1889 im6o = malloc(sizeof(*im6o), M_IPMOPTS, M_WAITOK); 1890 if (im6o == NULL) 1891 return (ENOBUFS); 1892 *im6op = im6o; 1893 im6o->im6o_ifidx = 0; 1894 im6o->im6o_hlim = ip6_defmcasthlim; 1895 im6o->im6o_loop = IPV6_DEFAULT_MULTICAST_LOOP; 1896 LIST_INIT(&im6o->im6o_memberships); 1897 } 1898 1899 switch (optname) { 1900 1901 case IPV6_MULTICAST_IF: 1902 /* 1903 * Select the interface for outgoing multicast packets. 1904 */ 1905 if (m == NULL || m->m_len != sizeof(u_int)) { 1906 error = EINVAL; 1907 break; 1908 } 1909 memcpy(&ifindex, mtod(m, u_int *), sizeof(ifindex)); 1910 if (ifindex != 0) { 1911 ifp = if_get(ifindex); 1912 if (ifp == NULL) { 1913 error = ENXIO; /* XXX EINVAL? */ 1914 break; 1915 } 1916 if (ifp->if_rdomain != rtable_l2(rtableid) || 1917 (ifp->if_flags & IFF_MULTICAST) == 0) { 1918 error = EADDRNOTAVAIL; 1919 if_put(ifp); 1920 break; 1921 } 1922 if_put(ifp); 1923 } 1924 im6o->im6o_ifidx = ifindex; 1925 break; 1926 1927 case IPV6_MULTICAST_HOPS: 1928 { 1929 /* 1930 * Set the IP6 hoplimit for outgoing multicast packets. 1931 */ 1932 int optval; 1933 if (m == NULL || m->m_len != sizeof(int)) { 1934 error = EINVAL; 1935 break; 1936 } 1937 memcpy(&optval, mtod(m, u_int *), sizeof(optval)); 1938 if (optval < -1 || optval >= 256) 1939 error = EINVAL; 1940 else if (optval == -1) 1941 im6o->im6o_hlim = ip6_defmcasthlim; 1942 else 1943 im6o->im6o_hlim = optval; 1944 break; 1945 } 1946 1947 case IPV6_MULTICAST_LOOP: 1948 /* 1949 * Set the loopback flag for outgoing multicast packets. 1950 * Must be zero or one. 1951 */ 1952 if (m == NULL || m->m_len != sizeof(u_int)) { 1953 error = EINVAL; 1954 break; 1955 } 1956 memcpy(&loop, mtod(m, u_int *), sizeof(loop)); 1957 if (loop > 1) { 1958 error = EINVAL; 1959 break; 1960 } 1961 im6o->im6o_loop = loop; 1962 break; 1963 1964 case IPV6_JOIN_GROUP: 1965 /* 1966 * Add a multicast group membership. 1967 * Group must be a valid IP6 multicast address. 1968 */ 1969 if (m == NULL || m->m_len != sizeof(struct ipv6_mreq)) { 1970 error = EINVAL; 1971 break; 1972 } 1973 mreq = mtod(m, struct ipv6_mreq *); 1974 if (IN6_IS_ADDR_UNSPECIFIED(&mreq->ipv6mr_multiaddr)) { 1975 /* 1976 * We use the unspecified address to specify to accept 1977 * all multicast addresses. Only super user is allowed 1978 * to do this. 1979 */ 1980 if (suser(p)) 1981 { 1982 error = EACCES; 1983 break; 1984 } 1985 } else if (!IN6_IS_ADDR_MULTICAST(&mreq->ipv6mr_multiaddr)) { 1986 error = EINVAL; 1987 break; 1988 } 1989 1990 /* 1991 * If no interface was explicitly specified, choose an 1992 * appropriate one according to the given multicast address. 1993 */ 1994 if (mreq->ipv6mr_interface == 0) { 1995 struct rtentry *rt; 1996 struct sockaddr_in6 dst; 1997 1998 memset(&dst, 0, sizeof(dst)); 1999 dst.sin6_len = sizeof(dst); 2000 dst.sin6_family = AF_INET6; 2001 dst.sin6_addr = mreq->ipv6mr_multiaddr; 2002 rt = rtalloc(sin6tosa(&dst), RT_RESOLVE, rtableid); 2003 if (rt == NULL) { 2004 error = EADDRNOTAVAIL; 2005 break; 2006 } 2007 ifp = if_get(rt->rt_ifidx); 2008 rtfree(rt); 2009 } else { 2010 /* 2011 * If the interface is specified, validate it. 2012 */ 2013 ifp = if_get(mreq->ipv6mr_interface); 2014 if (ifp == NULL) { 2015 error = ENXIO; /* XXX EINVAL? */ 2016 break; 2017 } 2018 } 2019 2020 /* 2021 * See if we found an interface, and confirm that it 2022 * supports multicast 2023 */ 2024 if (ifp == NULL || ifp->if_rdomain != rtable_l2(rtableid) || 2025 (ifp->if_flags & IFF_MULTICAST) == 0) { 2026 if_put(ifp); 2027 error = EADDRNOTAVAIL; 2028 break; 2029 } 2030 /* 2031 * Put interface index into the multicast address, 2032 * if the address has link/interface-local scope. 2033 */ 2034 if (IN6_IS_SCOPE_EMBED(&mreq->ipv6mr_multiaddr)) { 2035 mreq->ipv6mr_multiaddr.s6_addr16[1] = 2036 htons(ifp->if_index); 2037 } 2038 /* 2039 * See if the membership already exists. 2040 */ 2041 LIST_FOREACH(imm, &im6o->im6o_memberships, i6mm_chain) 2042 if (imm->i6mm_maddr->in6m_ifidx == ifp->if_index && 2043 IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr, 2044 &mreq->ipv6mr_multiaddr)) 2045 break; 2046 if (imm != NULL) { 2047 if_put(ifp); 2048 error = EADDRINUSE; 2049 break; 2050 } 2051 /* 2052 * Everything looks good; add a new record to the multicast 2053 * address list for the given interface. 2054 */ 2055 imm = in6_joingroup(ifp, &mreq->ipv6mr_multiaddr, &error); 2056 if_put(ifp); 2057 if (!imm) 2058 break; 2059 LIST_INSERT_HEAD(&im6o->im6o_memberships, imm, i6mm_chain); 2060 break; 2061 2062 case IPV6_LEAVE_GROUP: 2063 /* 2064 * Drop a multicast group membership. 2065 * Group must be a valid IP6 multicast address. 2066 */ 2067 if (m == NULL || m->m_len != sizeof(struct ipv6_mreq)) { 2068 error = EINVAL; 2069 break; 2070 } 2071 mreq = mtod(m, struct ipv6_mreq *); 2072 if (IN6_IS_ADDR_UNSPECIFIED(&mreq->ipv6mr_multiaddr)) { 2073 if (suser(p)) { 2074 error = EACCES; 2075 break; 2076 } 2077 } else if (!IN6_IS_ADDR_MULTICAST(&mreq->ipv6mr_multiaddr)) { 2078 error = EINVAL; 2079 break; 2080 } 2081 2082 /* 2083 * Put interface index into the multicast address, 2084 * if the address has link-local scope. 2085 */ 2086 if (IN6_IS_ADDR_MC_LINKLOCAL(&mreq->ipv6mr_multiaddr)) { 2087 mreq->ipv6mr_multiaddr.s6_addr16[1] = 2088 htons(mreq->ipv6mr_interface); 2089 } 2090 2091 /* 2092 * If an interface address was specified, get a pointer 2093 * to its ifnet structure. 2094 */ 2095 if (mreq->ipv6mr_interface == 0) 2096 ifp = NULL; 2097 else { 2098 ifp = if_get(mreq->ipv6mr_interface); 2099 if (ifp == NULL) { 2100 error = ENXIO; /* XXX EINVAL? */ 2101 break; 2102 } 2103 } 2104 2105 /* 2106 * Find the membership in the membership list. 2107 */ 2108 LIST_FOREACH(imm, &im6o->im6o_memberships, i6mm_chain) { 2109 if ((ifp == NULL || 2110 imm->i6mm_maddr->in6m_ifidx == ifp->if_index) && 2111 IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr, 2112 &mreq->ipv6mr_multiaddr)) 2113 break; 2114 } 2115 2116 if_put(ifp); 2117 2118 if (imm == NULL) { 2119 /* Unable to resolve interface */ 2120 error = EADDRNOTAVAIL; 2121 break; 2122 } 2123 /* 2124 * Give up the multicast address record to which the 2125 * membership points. 2126 */ 2127 LIST_REMOVE(imm, i6mm_chain); 2128 in6_leavegroup(imm); 2129 break; 2130 2131 default: 2132 error = EOPNOTSUPP; 2133 break; 2134 } 2135 2136 /* 2137 * If all options have default values, no need to keep the option 2138 * structure. 2139 */ 2140 if (im6o->im6o_ifidx == 0 && 2141 im6o->im6o_hlim == ip6_defmcasthlim && 2142 im6o->im6o_loop == IPV6_DEFAULT_MULTICAST_LOOP && 2143 LIST_EMPTY(&im6o->im6o_memberships)) { 2144 free(*im6op, M_IPMOPTS, sizeof(**im6op)); 2145 *im6op = NULL; 2146 } 2147 2148 return (error); 2149 } 2150 2151 /* 2152 * Return the IP6 multicast options in response to user getsockopt(). 2153 */ 2154 int 2155 ip6_getmoptions(int optname, struct ip6_moptions *im6o, struct mbuf *m) 2156 { 2157 u_int *hlim, *loop, *ifindex; 2158 2159 switch (optname) { 2160 case IPV6_MULTICAST_IF: 2161 ifindex = mtod(m, u_int *); 2162 m->m_len = sizeof(u_int); 2163 if (im6o == NULL || im6o->im6o_ifidx == 0) 2164 *ifindex = 0; 2165 else 2166 *ifindex = im6o->im6o_ifidx; 2167 return (0); 2168 2169 case IPV6_MULTICAST_HOPS: 2170 hlim = mtod(m, u_int *); 2171 m->m_len = sizeof(u_int); 2172 if (im6o == NULL) 2173 *hlim = ip6_defmcasthlim; 2174 else 2175 *hlim = im6o->im6o_hlim; 2176 return (0); 2177 2178 case IPV6_MULTICAST_LOOP: 2179 loop = mtod(m, u_int *); 2180 m->m_len = sizeof(u_int); 2181 if (im6o == NULL) 2182 *loop = ip6_defmcasthlim; 2183 else 2184 *loop = im6o->im6o_loop; 2185 return (0); 2186 2187 default: 2188 return (EOPNOTSUPP); 2189 } 2190 } 2191 2192 /* 2193 * Discard the IP6 multicast options. 2194 */ 2195 void 2196 ip6_freemoptions(struct ip6_moptions *im6o) 2197 { 2198 struct in6_multi_mship *imm; 2199 2200 if (im6o == NULL) 2201 return; 2202 2203 while (!LIST_EMPTY(&im6o->im6o_memberships)) { 2204 imm = LIST_FIRST(&im6o->im6o_memberships); 2205 LIST_REMOVE(imm, i6mm_chain); 2206 in6_leavegroup(imm); 2207 } 2208 free(im6o, M_IPMOPTS, sizeof(*im6o)); 2209 } 2210 2211 /* 2212 * Set IPv6 outgoing packet options based on advanced API. 2213 */ 2214 int 2215 ip6_setpktopts(struct mbuf *control, struct ip6_pktopts *opt, 2216 struct ip6_pktopts *stickyopt, int priv, int uproto) 2217 { 2218 u_int clen; 2219 struct cmsghdr *cm = 0; 2220 caddr_t cmsgs; 2221 int error; 2222 2223 if (control == NULL || opt == NULL) 2224 return (EINVAL); 2225 2226 ip6_initpktopts(opt); 2227 if (stickyopt) { 2228 int error; 2229 2230 /* 2231 * If stickyopt is provided, make a local copy of the options 2232 * for this particular packet, then override them by ancillary 2233 * objects. 2234 * XXX: copypktopts() does not copy the cached route to a next 2235 * hop (if any). This is not very good in terms of efficiency, 2236 * but we can allow this since this option should be rarely 2237 * used. 2238 */ 2239 if ((error = copypktopts(opt, stickyopt)) != 0) 2240 return (error); 2241 } 2242 2243 /* 2244 * XXX: Currently, we assume all the optional information is stored 2245 * in a single mbuf. 2246 */ 2247 if (control->m_next) 2248 return (EINVAL); 2249 2250 clen = control->m_len; 2251 cmsgs = mtod(control, caddr_t); 2252 do { 2253 if (clen < CMSG_LEN(0)) 2254 return (EINVAL); 2255 cm = (struct cmsghdr *)cmsgs; 2256 if (cm->cmsg_len < CMSG_LEN(0) || cm->cmsg_len > clen || 2257 CMSG_ALIGN(cm->cmsg_len) > clen) 2258 return (EINVAL); 2259 if (cm->cmsg_level == IPPROTO_IPV6) { 2260 error = ip6_setpktopt(cm->cmsg_type, CMSG_DATA(cm), 2261 cm->cmsg_len - CMSG_LEN(0), opt, priv, 0, uproto); 2262 if (error) 2263 return (error); 2264 } 2265 2266 clen -= CMSG_ALIGN(cm->cmsg_len); 2267 cmsgs += CMSG_ALIGN(cm->cmsg_len); 2268 } while (clen); 2269 2270 return (0); 2271 } 2272 2273 /* 2274 * Set a particular packet option, as a sticky option or an ancillary data 2275 * item. "len" can be 0 only when it's a sticky option. 2276 */ 2277 int 2278 ip6_setpktopt(int optname, u_char *buf, int len, struct ip6_pktopts *opt, 2279 int priv, int sticky, int uproto) 2280 { 2281 int minmtupolicy; 2282 2283 switch (optname) { 2284 case IPV6_PKTINFO: 2285 { 2286 struct ifnet *ifp = NULL; 2287 struct in6_pktinfo *pktinfo; 2288 2289 if (len != sizeof(struct in6_pktinfo)) 2290 return (EINVAL); 2291 2292 pktinfo = (struct in6_pktinfo *)buf; 2293 2294 /* 2295 * An application can clear any sticky IPV6_PKTINFO option by 2296 * doing a "regular" setsockopt with ipi6_addr being 2297 * in6addr_any and ipi6_ifindex being zero. 2298 * [RFC 3542, Section 6] 2299 */ 2300 if (opt->ip6po_pktinfo && 2301 pktinfo->ipi6_ifindex == 0 && 2302 IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) { 2303 ip6_clearpktopts(opt, optname); 2304 break; 2305 } 2306 2307 if (uproto == IPPROTO_TCP && 2308 sticky && !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) { 2309 return (EINVAL); 2310 } 2311 2312 if (pktinfo->ipi6_ifindex) { 2313 ifp = if_get(pktinfo->ipi6_ifindex); 2314 if (ifp == NULL) 2315 return (ENXIO); 2316 if_put(ifp); 2317 } 2318 2319 /* 2320 * We store the address anyway, and let in6_selectsrc() 2321 * validate the specified address. This is because ipi6_addr 2322 * may not have enough information about its scope zone, and 2323 * we may need additional information (such as outgoing 2324 * interface or the scope zone of a destination address) to 2325 * disambiguate the scope. 2326 * XXX: the delay of the validation may confuse the 2327 * application when it is used as a sticky option. 2328 */ 2329 if (opt->ip6po_pktinfo == NULL) { 2330 opt->ip6po_pktinfo = malloc(sizeof(*pktinfo), 2331 M_IP6OPT, M_NOWAIT); 2332 if (opt->ip6po_pktinfo == NULL) 2333 return (ENOBUFS); 2334 } 2335 bcopy(pktinfo, opt->ip6po_pktinfo, sizeof(*pktinfo)); 2336 break; 2337 } 2338 2339 case IPV6_HOPLIMIT: 2340 { 2341 int *hlimp; 2342 2343 /* 2344 * RFC 3542 deprecated the usage of sticky IPV6_HOPLIMIT 2345 * to simplify the ordering among hoplimit options. 2346 */ 2347 if (sticky) 2348 return (ENOPROTOOPT); 2349 2350 if (len != sizeof(int)) 2351 return (EINVAL); 2352 hlimp = (int *)buf; 2353 if (*hlimp < -1 || *hlimp > 255) 2354 return (EINVAL); 2355 2356 opt->ip6po_hlim = *hlimp; 2357 break; 2358 } 2359 2360 case IPV6_TCLASS: 2361 { 2362 int tclass; 2363 2364 if (len != sizeof(int)) 2365 return (EINVAL); 2366 tclass = *(int *)buf; 2367 if (tclass < -1 || tclass > 255) 2368 return (EINVAL); 2369 2370 opt->ip6po_tclass = tclass; 2371 break; 2372 } 2373 case IPV6_HOPOPTS: 2374 { 2375 struct ip6_hbh *hbh; 2376 int hbhlen; 2377 2378 /* 2379 * XXX: We don't allow a non-privileged user to set ANY HbH 2380 * options, since per-option restriction has too much 2381 * overhead. 2382 */ 2383 if (!priv) 2384 return (EPERM); 2385 2386 if (len == 0) { 2387 ip6_clearpktopts(opt, IPV6_HOPOPTS); 2388 break; /* just remove the option */ 2389 } 2390 2391 /* message length validation */ 2392 if (len < sizeof(struct ip6_hbh)) 2393 return (EINVAL); 2394 hbh = (struct ip6_hbh *)buf; 2395 hbhlen = (hbh->ip6h_len + 1) << 3; 2396 if (len != hbhlen) 2397 return (EINVAL); 2398 2399 /* turn off the previous option, then set the new option. */ 2400 ip6_clearpktopts(opt, IPV6_HOPOPTS); 2401 opt->ip6po_hbh = malloc(hbhlen, M_IP6OPT, M_NOWAIT); 2402 if (opt->ip6po_hbh == NULL) 2403 return (ENOBUFS); 2404 memcpy(opt->ip6po_hbh, hbh, hbhlen); 2405 2406 break; 2407 } 2408 2409 case IPV6_DSTOPTS: 2410 case IPV6_RTHDRDSTOPTS: 2411 { 2412 struct ip6_dest *dest, **newdest = NULL; 2413 int destlen; 2414 2415 if (!priv) /* XXX: see the comment for IPV6_HOPOPTS */ 2416 return (EPERM); 2417 2418 if (len == 0) { 2419 ip6_clearpktopts(opt, optname); 2420 break; /* just remove the option */ 2421 } 2422 2423 /* message length validation */ 2424 if (len < sizeof(struct ip6_dest)) 2425 return (EINVAL); 2426 dest = (struct ip6_dest *)buf; 2427 destlen = (dest->ip6d_len + 1) << 3; 2428 if (len != destlen) 2429 return (EINVAL); 2430 /* 2431 * Determine the position that the destination options header 2432 * should be inserted; before or after the routing header. 2433 */ 2434 switch (optname) { 2435 case IPV6_RTHDRDSTOPTS: 2436 newdest = &opt->ip6po_dest1; 2437 break; 2438 case IPV6_DSTOPTS: 2439 newdest = &opt->ip6po_dest2; 2440 break; 2441 } 2442 2443 /* turn off the previous option, then set the new option. */ 2444 ip6_clearpktopts(opt, optname); 2445 *newdest = malloc(destlen, M_IP6OPT, M_NOWAIT); 2446 if (*newdest == NULL) 2447 return (ENOBUFS); 2448 memcpy(*newdest, dest, destlen); 2449 2450 break; 2451 } 2452 2453 case IPV6_RTHDR: 2454 { 2455 struct ip6_rthdr *rth; 2456 int rthlen; 2457 2458 if (len == 0) { 2459 ip6_clearpktopts(opt, IPV6_RTHDR); 2460 break; /* just remove the option */ 2461 } 2462 2463 /* message length validation */ 2464 if (len < sizeof(struct ip6_rthdr)) 2465 return (EINVAL); 2466 rth = (struct ip6_rthdr *)buf; 2467 rthlen = (rth->ip6r_len + 1) << 3; 2468 if (len != rthlen) 2469 return (EINVAL); 2470 2471 switch (rth->ip6r_type) { 2472 case IPV6_RTHDR_TYPE_0: 2473 if (rth->ip6r_len == 0) /* must contain one addr */ 2474 return (EINVAL); 2475 if (rth->ip6r_len % 2) /* length must be even */ 2476 return (EINVAL); 2477 if (rth->ip6r_len / 2 != rth->ip6r_segleft) 2478 return (EINVAL); 2479 break; 2480 default: 2481 return (EINVAL); /* not supported */ 2482 } 2483 /* turn off the previous option */ 2484 ip6_clearpktopts(opt, IPV6_RTHDR); 2485 opt->ip6po_rthdr = malloc(rthlen, M_IP6OPT, M_NOWAIT); 2486 if (opt->ip6po_rthdr == NULL) 2487 return (ENOBUFS); 2488 memcpy(opt->ip6po_rthdr, rth, rthlen); 2489 break; 2490 } 2491 2492 case IPV6_USE_MIN_MTU: 2493 if (len != sizeof(int)) 2494 return (EINVAL); 2495 minmtupolicy = *(int *)buf; 2496 if (minmtupolicy != IP6PO_MINMTU_MCASTONLY && 2497 minmtupolicy != IP6PO_MINMTU_DISABLE && 2498 minmtupolicy != IP6PO_MINMTU_ALL) { 2499 return (EINVAL); 2500 } 2501 opt->ip6po_minmtu = minmtupolicy; 2502 break; 2503 2504 case IPV6_DONTFRAG: 2505 if (len != sizeof(int)) 2506 return (EINVAL); 2507 2508 if (uproto == IPPROTO_TCP || *(int *)buf == 0) { 2509 /* 2510 * we ignore this option for TCP sockets. 2511 * (RFC3542 leaves this case unspecified.) 2512 */ 2513 opt->ip6po_flags &= ~IP6PO_DONTFRAG; 2514 } else 2515 opt->ip6po_flags |= IP6PO_DONTFRAG; 2516 break; 2517 2518 default: 2519 return (ENOPROTOOPT); 2520 } /* end of switch */ 2521 2522 return (0); 2523 } 2524 2525 /* 2526 * Routine called from ip6_output() to loop back a copy of an IP6 multicast 2527 * packet to the input queue of a specified interface. 2528 */ 2529 void 2530 ip6_mloopback(struct ifnet *ifp, struct mbuf *m, struct sockaddr_in6 *dst) 2531 { 2532 struct mbuf *copym; 2533 struct ip6_hdr *ip6; 2534 2535 /* 2536 * Duplicate the packet. 2537 */ 2538 copym = m_copym(m, 0, M_COPYALL, M_NOWAIT); 2539 if (copym == NULL) 2540 return; 2541 2542 /* 2543 * Make sure to deep-copy IPv6 header portion in case the data 2544 * is in an mbuf cluster, so that we can safely override the IPv6 2545 * header portion later. 2546 */ 2547 if ((copym->m_flags & M_EXT) != 0 || 2548 copym->m_len < sizeof(struct ip6_hdr)) { 2549 copym = m_pullup(copym, sizeof(struct ip6_hdr)); 2550 if (copym == NULL) 2551 return; 2552 } 2553 2554 #ifdef DIAGNOSTIC 2555 if (copym->m_len < sizeof(*ip6)) { 2556 m_freem(copym); 2557 return; 2558 } 2559 #endif 2560 2561 ip6 = mtod(copym, struct ip6_hdr *); 2562 if (IN6_IS_SCOPE_EMBED(&ip6->ip6_src)) 2563 ip6->ip6_src.s6_addr16[1] = 0; 2564 if (IN6_IS_SCOPE_EMBED(&ip6->ip6_dst)) 2565 ip6->ip6_dst.s6_addr16[1] = 0; 2566 2567 if_input_local(ifp, copym, dst->sin6_family); 2568 } 2569 2570 /* 2571 * Chop IPv6 header off from the payload. 2572 */ 2573 int 2574 ip6_splithdr(struct mbuf *m, struct ip6_exthdrs *exthdrs) 2575 { 2576 struct mbuf *mh; 2577 struct ip6_hdr *ip6; 2578 2579 ip6 = mtod(m, struct ip6_hdr *); 2580 if (m->m_len > sizeof(*ip6)) { 2581 MGET(mh, M_DONTWAIT, MT_HEADER); 2582 if (mh == NULL) { 2583 m_freem(m); 2584 return ENOBUFS; 2585 } 2586 M_MOVE_PKTHDR(mh, m); 2587 m_align(mh, sizeof(*ip6)); 2588 m->m_len -= sizeof(*ip6); 2589 m->m_data += sizeof(*ip6); 2590 mh->m_next = m; 2591 m = mh; 2592 m->m_len = sizeof(*ip6); 2593 bcopy((caddr_t)ip6, mtod(m, caddr_t), sizeof(*ip6)); 2594 } 2595 exthdrs->ip6e_ip6 = m; 2596 return 0; 2597 } 2598 2599 u_int32_t 2600 ip6_randomid(void) 2601 { 2602 return idgen32(&ip6_id_ctx); 2603 } 2604 2605 void 2606 ip6_randomid_init(void) 2607 { 2608 idgen32_init(&ip6_id_ctx); 2609 } 2610 2611 /* 2612 * Compute significant parts of the IPv6 checksum pseudo-header 2613 * for use in a delayed TCP/UDP checksum calculation. 2614 */ 2615 static __inline u_int16_t __attribute__((__unused__)) 2616 in6_cksum_phdr(const struct in6_addr *src, const struct in6_addr *dst, 2617 u_int32_t len, u_int32_t nxt) 2618 { 2619 u_int32_t sum = 0; 2620 const u_int16_t *w; 2621 2622 w = (const u_int16_t *) src; 2623 sum += w[0]; 2624 if (!IN6_IS_SCOPE_EMBED(src)) 2625 sum += w[1]; 2626 sum += w[2]; sum += w[3]; sum += w[4]; sum += w[5]; 2627 sum += w[6]; sum += w[7]; 2628 2629 w = (const u_int16_t *) dst; 2630 sum += w[0]; 2631 if (!IN6_IS_SCOPE_EMBED(dst)) 2632 sum += w[1]; 2633 sum += w[2]; sum += w[3]; sum += w[4]; sum += w[5]; 2634 sum += w[6]; sum += w[7]; 2635 2636 sum += (u_int16_t)(len >> 16) + (u_int16_t)(len /*& 0xffff*/); 2637 2638 sum += (u_int16_t)(nxt >> 16) + (u_int16_t)(nxt /*& 0xffff*/); 2639 2640 sum = (u_int16_t)(sum >> 16) + (u_int16_t)(sum /*& 0xffff*/); 2641 2642 if (sum > 0xffff) 2643 sum -= 0xffff; 2644 2645 return (sum); 2646 } 2647 2648 /* 2649 * Process a delayed payload checksum calculation. 2650 */ 2651 void 2652 in6_delayed_cksum(struct mbuf *m, u_int8_t nxt) 2653 { 2654 int nxtp, offset; 2655 u_int16_t csum; 2656 2657 offset = ip6_lasthdr(m, 0, IPPROTO_IPV6, &nxtp); 2658 if (offset <= 0 || nxtp != nxt) 2659 /* If the desired next protocol isn't found, punt. */ 2660 return; 2661 csum = (u_int16_t)(in6_cksum(m, 0, offset, m->m_pkthdr.len - offset)); 2662 2663 switch (nxt) { 2664 case IPPROTO_TCP: 2665 offset += offsetof(struct tcphdr, th_sum); 2666 break; 2667 2668 case IPPROTO_UDP: 2669 offset += offsetof(struct udphdr, uh_sum); 2670 if (csum == 0) 2671 csum = 0xffff; 2672 break; 2673 2674 case IPPROTO_ICMPV6: 2675 offset += offsetof(struct icmp6_hdr, icmp6_cksum); 2676 break; 2677 } 2678 2679 if ((offset + sizeof(u_int16_t)) > m->m_len) 2680 m_copyback(m, offset, sizeof(csum), &csum, M_NOWAIT); 2681 else 2682 *(u_int16_t *)(mtod(m, caddr_t) + offset) = csum; 2683 } 2684 2685 void 2686 in6_proto_cksum_out(struct mbuf *m, struct ifnet *ifp) 2687 { 2688 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *); 2689 2690 /* some hw and in6_delayed_cksum need the pseudo header cksum */ 2691 if (m->m_pkthdr.csum_flags & 2692 (M_TCP_CSUM_OUT|M_UDP_CSUM_OUT|M_ICMP_CSUM_OUT)) { 2693 int nxt, offset; 2694 u_int16_t csum; 2695 2696 offset = ip6_lasthdr(m, 0, IPPROTO_IPV6, &nxt); 2697 csum = in6_cksum_phdr(&ip6->ip6_src, &ip6->ip6_dst, 2698 htonl(m->m_pkthdr.len - offset), htonl(nxt)); 2699 if (nxt == IPPROTO_TCP) 2700 offset += offsetof(struct tcphdr, th_sum); 2701 else if (nxt == IPPROTO_UDP) 2702 offset += offsetof(struct udphdr, uh_sum); 2703 else if (nxt == IPPROTO_ICMPV6) 2704 offset += offsetof(struct icmp6_hdr, icmp6_cksum); 2705 if ((offset + sizeof(u_int16_t)) > m->m_len) 2706 m_copyback(m, offset, sizeof(csum), &csum, M_NOWAIT); 2707 else 2708 *(u_int16_t *)(mtod(m, caddr_t) + offset) = csum; 2709 } 2710 2711 if (m->m_pkthdr.csum_flags & M_TCP_CSUM_OUT) { 2712 if (!ifp || !(ifp->if_capabilities & IFCAP_CSUM_TCPv6) || 2713 ip6->ip6_nxt != IPPROTO_TCP || 2714 ifp->if_bridgeidx != 0) { 2715 tcpstat_inc(tcps_outswcsum); 2716 in6_delayed_cksum(m, IPPROTO_TCP); 2717 m->m_pkthdr.csum_flags &= ~M_TCP_CSUM_OUT; /* Clear */ 2718 } 2719 } else if (m->m_pkthdr.csum_flags & M_UDP_CSUM_OUT) { 2720 if (!ifp || !(ifp->if_capabilities & IFCAP_CSUM_UDPv6) || 2721 ip6->ip6_nxt != IPPROTO_UDP || 2722 ifp->if_bridgeidx != 0) { 2723 udpstat_inc(udps_outswcsum); 2724 in6_delayed_cksum(m, IPPROTO_UDP); 2725 m->m_pkthdr.csum_flags &= ~M_UDP_CSUM_OUT; /* Clear */ 2726 } 2727 } else if (m->m_pkthdr.csum_flags & M_ICMP_CSUM_OUT) { 2728 in6_delayed_cksum(m, IPPROTO_ICMPV6); 2729 m->m_pkthdr.csum_flags &= ~M_ICMP_CSUM_OUT; /* Clear */ 2730 } 2731 } 2732 2733 #ifdef IPSEC 2734 struct tdb * 2735 ip6_output_ipsec_lookup(struct mbuf *m, int *error, struct inpcb *inp) 2736 { 2737 struct tdb *tdb; 2738 struct m_tag *mtag; 2739 struct tdb_ident *tdbi; 2740 2741 /* 2742 * Check if there was an outgoing SA bound to the flow 2743 * from a transport protocol. 2744 */ 2745 2746 /* Do we have any pending SAs to apply ? */ 2747 tdb = ipsp_spd_lookup(m, AF_INET6, sizeof(struct ip6_hdr), 2748 error, IPSP_DIRECTION_OUT, NULL, inp, 0); 2749 2750 if (tdb == NULL) 2751 return NULL; 2752 /* Loop detection */ 2753 for (mtag = m_tag_first(m); mtag != NULL; mtag = m_tag_next(m, mtag)) { 2754 if (mtag->m_tag_id != PACKET_TAG_IPSEC_OUT_DONE) 2755 continue; 2756 tdbi = (struct tdb_ident *)(mtag + 1); 2757 if (tdbi->spi == tdb->tdb_spi && 2758 tdbi->proto == tdb->tdb_sproto && 2759 tdbi->rdomain == tdb->tdb_rdomain && 2760 !memcmp(&tdbi->dst, &tdb->tdb_dst, 2761 sizeof(union sockaddr_union))) { 2762 /* no IPsec needed */ 2763 return NULL; 2764 } 2765 } 2766 return tdb; 2767 } 2768 2769 int 2770 ip6_output_ipsec_send(struct tdb *tdb, struct mbuf *m, struct route_in6 *ro, 2771 int tunalready, int fwd) 2772 { 2773 #if NPF > 0 2774 struct ifnet *encif; 2775 #endif 2776 struct ip6_hdr *ip6; 2777 int error; 2778 2779 #if NPF > 0 2780 /* 2781 * Packet filter 2782 */ 2783 if ((encif = enc_getif(tdb->tdb_rdomain, tdb->tdb_tap)) == NULL || 2784 pf_test(AF_INET6, fwd ? PF_FWD : PF_OUT, encif, &m) != PF_PASS) { 2785 m_freem(m); 2786 return EACCES; 2787 } 2788 if (m == NULL) 2789 return 0; 2790 /* 2791 * PF_TAG_REROUTE handling or not... 2792 * Packet is entering IPsec so the routing is 2793 * already overruled by the IPsec policy. 2794 * Until now the change was not reconsidered. 2795 * What's the behaviour? 2796 */ 2797 in6_proto_cksum_out(m, encif); 2798 #endif 2799 2800 /* Check if we are allowed to fragment */ 2801 ip6 = mtod(m, struct ip6_hdr *); 2802 if (ip_mtudisc && tdb->tdb_mtu && 2803 sizeof(struct ip6_hdr) + ntohs(ip6->ip6_plen) > tdb->tdb_mtu && 2804 tdb->tdb_mtutimeout > gettime()) { 2805 struct rtentry *rt = NULL; 2806 int rt_mtucloned = 0; 2807 int transportmode = 0; 2808 2809 transportmode = (tdb->tdb_dst.sa.sa_family == AF_INET6) && 2810 (IN6_ARE_ADDR_EQUAL(&tdb->tdb_dst.sin6.sin6_addr, 2811 &ip6->ip6_dst)); 2812 2813 /* Find a host route to store the mtu in */ 2814 if (ro != NULL) 2815 rt = ro->ro_rt; 2816 /* but don't add a PMTU route for transport mode SAs */ 2817 if (transportmode) 2818 rt = NULL; 2819 else if (rt == NULL || (rt->rt_flags & RTF_HOST) == 0) { 2820 struct sockaddr_in6 sin6; 2821 2822 memset(&sin6, 0, sizeof(sin6)); 2823 sin6.sin6_family = AF_INET6; 2824 sin6.sin6_len = sizeof(sin6); 2825 sin6.sin6_addr = ip6->ip6_dst; 2826 sin6.sin6_scope_id = 2827 in6_addr2scopeid(m->m_pkthdr.ph_ifidx, 2828 &ip6->ip6_dst); 2829 error = in6_embedscope(&ip6->ip6_dst, &sin6, NULL); 2830 if (error) { 2831 /* should be impossible */ 2832 ipsecstat_inc(ipsec_odrops); 2833 m_freem(m); 2834 return error; 2835 } 2836 rt = icmp6_mtudisc_clone(&sin6, 2837 m->m_pkthdr.ph_rtableid, 1); 2838 rt_mtucloned = 1; 2839 } 2840 DPRINTF(("%s: spi %08x mtu %d rt %p cloned %d\n", __func__, 2841 ntohl(tdb->tdb_spi), tdb->tdb_mtu, rt, rt_mtucloned)); 2842 if (rt != NULL) { 2843 rt->rt_mtu = tdb->tdb_mtu; 2844 if (ro != NULL && ro->ro_rt != NULL) { 2845 rtfree(ro->ro_rt); 2846 ro->ro_rt = rtalloc(sin6tosa(&ro->ro_dst), 2847 RT_RESOLVE, m->m_pkthdr.ph_rtableid); 2848 } 2849 if (rt_mtucloned) 2850 rtfree(rt); 2851 } 2852 ipsec_adjust_mtu(m, tdb->tdb_mtu); 2853 m_freem(m); 2854 return EMSGSIZE; 2855 } 2856 /* propagate don't fragment for v6-over-v6 */ 2857 if (ip_mtudisc) 2858 SET(m->m_pkthdr.csum_flags, M_IPV6_DF_OUT); 2859 2860 /* 2861 * Clear these -- they'll be set in the recursive invocation 2862 * as needed. 2863 */ 2864 m->m_flags &= ~(M_BCAST | M_MCAST); 2865 2866 /* Callee frees mbuf */ 2867 error = ipsp_process_packet(m, tdb, AF_INET6, tunalready); 2868 if (error) { 2869 ipsecstat_inc(ipsec_odrops); 2870 tdb->tdb_odrops++; 2871 } 2872 return error; 2873 } 2874 #endif /* IPSEC */ 2875