1 /* $OpenBSD: ip6_output.c,v 1.294 2025/01/03 21:27:40 bluhm 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 98 #include <crypto/idgen.h> 99 100 #if NPF > 0 101 #include <net/pfvar.h> 102 #endif 103 104 #ifdef IPSEC 105 #include <netinet/ip_ipsp.h> 106 #include <netinet/ip_ah.h> 107 #include <netinet/ip_esp.h> 108 109 #ifdef ENCDEBUG 110 #define DPRINTF(fmt, args...) \ 111 do { \ 112 if (encdebug) \ 113 printf("%s: " fmt "\n", __func__, ## args); \ 114 } while (0) 115 #else 116 #define DPRINTF(fmt, args...) \ 117 do { } while (0) 118 #endif 119 #endif /* IPSEC */ 120 121 struct ip6_exthdrs { 122 struct mbuf *ip6e_ip6; 123 struct mbuf *ip6e_hbh; 124 struct mbuf *ip6e_dest1; 125 struct mbuf *ip6e_rthdr; 126 struct mbuf *ip6e_dest2; 127 }; 128 129 int ip6_pcbopt(int, u_char *, int, struct ip6_pktopts **, int, int); 130 int ip6_getpcbopt(struct ip6_pktopts *, int, struct mbuf *); 131 int ip6_setpktopt(int, u_char *, int, struct ip6_pktopts *, int, int, int); 132 int ip6_setmoptions(int, struct ip6_moptions **, struct mbuf *, unsigned int); 133 int ip6_getmoptions(int, struct ip6_moptions *, struct mbuf *); 134 int ip6_copyexthdr(struct mbuf **, caddr_t, int); 135 int ip6_insertfraghdr(struct mbuf *, struct mbuf *, int, 136 struct ip6_frag **); 137 int ip6_insert_jumboopt(struct ip6_exthdrs *, u_int32_t); 138 int ip6_splithdr(struct mbuf *, struct ip6_exthdrs *); 139 int ip6_getpmtu(struct rtentry *, struct ifnet *, u_long *); 140 int copypktopts(struct ip6_pktopts *, struct ip6_pktopts *); 141 static __inline u_int16_t __attribute__((__unused__)) 142 in6_cksum_phdr(const struct in6_addr *, const struct in6_addr *, 143 u_int32_t, u_int32_t); 144 void in6_delayed_cksum(struct mbuf *, u_int8_t); 145 146 int ip6_output_ipsec_pmtu_update(struct tdb *, struct route *, 147 struct in6_addr *, int, int, int); 148 149 /* Context for non-repeating IDs */ 150 struct idgen32_ctx ip6_id_ctx; 151 152 /* 153 * IP6 output. The packet in mbuf chain m contains a skeletal IP6 154 * header (with pri, len, nxt, hlim, src, dst). 155 * This function may modify ver and hlim only. 156 * The mbuf chain containing the packet will be freed. 157 * The mbuf opt, if present, will not be freed. 158 * 159 * type of "mtu": rt_mtu is u_int, ifnet.ifr_mtu is int. 160 * We use u_long to hold largest one. XXX should be u_int 161 */ 162 int 163 ip6_output(struct mbuf *m, struct ip6_pktopts *opt, struct route *ro, 164 int flags, struct ip6_moptions *im6o, const struct ipsec_level *seclevel) 165 { 166 struct ip6_hdr *ip6; 167 struct ifnet *ifp = NULL; 168 struct mbuf_list ml; 169 int hlen, tlen; 170 struct route iproute; 171 struct rtentry *rt = NULL; 172 struct sockaddr_in6 *dst; 173 int error = 0; 174 u_long mtu; 175 int dontfrag; 176 u_int16_t src_scope, dst_scope; 177 u_int32_t optlen = 0, plen = 0, unfragpartlen = 0; 178 struct ip6_exthdrs exthdrs; 179 struct in6_addr finaldst; 180 struct route *ro_pmtu = NULL; 181 int hdrsplit = 0; 182 u_int8_t sproto = 0; 183 u_char nextproto; 184 #ifdef IPSEC 185 struct tdb *tdb = NULL; 186 #endif /* IPSEC */ 187 188 ip6 = mtod(m, struct ip6_hdr *); 189 finaldst = ip6->ip6_dst; 190 191 #define MAKE_EXTHDR(hp, mp) \ 192 do { \ 193 if (hp) { \ 194 struct ip6_ext *eh = (struct ip6_ext *)(hp); \ 195 error = ip6_copyexthdr((mp), (caddr_t)(hp), \ 196 ((eh)->ip6e_len + 1) << 3); \ 197 if (error) \ 198 goto freehdrs; \ 199 } \ 200 } while (0) 201 202 bzero(&exthdrs, sizeof(exthdrs)); 203 204 if (opt) { 205 /* Hop-by-Hop options header */ 206 MAKE_EXTHDR(opt->ip6po_hbh, &exthdrs.ip6e_hbh); 207 /* Destination options header(1st part) */ 208 MAKE_EXTHDR(opt->ip6po_dest1, &exthdrs.ip6e_dest1); 209 /* Routing header */ 210 MAKE_EXTHDR(opt->ip6po_rthdr, &exthdrs.ip6e_rthdr); 211 /* Destination options header(2nd part) */ 212 MAKE_EXTHDR(opt->ip6po_dest2, &exthdrs.ip6e_dest2); 213 } 214 215 #ifdef IPSEC 216 if (ipsec_in_use || seclevel != NULL) { 217 error = ip6_output_ipsec_lookup(m, seclevel, &tdb); 218 if (error) { 219 /* 220 * -EINVAL is used to indicate that the packet should 221 * be silently dropped, typically because we've asked 222 * key management for an SA. 223 */ 224 if (error == -EINVAL) /* Should silently drop packet */ 225 error = 0; 226 227 goto freehdrs; 228 } 229 } 230 #endif /* IPSEC */ 231 232 /* 233 * Calculate the total length of the extension header chain. 234 * Keep the length of the unfragmentable part for fragmentation. 235 */ 236 optlen = 0; 237 if (exthdrs.ip6e_hbh) optlen += exthdrs.ip6e_hbh->m_len; 238 if (exthdrs.ip6e_dest1) optlen += exthdrs.ip6e_dest1->m_len; 239 if (exthdrs.ip6e_rthdr) optlen += exthdrs.ip6e_rthdr->m_len; 240 unfragpartlen = optlen + sizeof(struct ip6_hdr); 241 /* NOTE: we don't add AH/ESP length here. do that later. */ 242 if (exthdrs.ip6e_dest2) optlen += exthdrs.ip6e_dest2->m_len; 243 244 /* 245 * If we need IPsec, or there is at least one extension header, 246 * separate IP6 header from the payload. 247 */ 248 if ((sproto || optlen) && !hdrsplit) { 249 if ((error = ip6_splithdr(m, &exthdrs)) != 0) { 250 m = NULL; 251 goto freehdrs; 252 } 253 m = exthdrs.ip6e_ip6; 254 hdrsplit++; 255 } 256 257 /* adjust pointer */ 258 ip6 = mtod(m, struct ip6_hdr *); 259 260 /* adjust mbuf packet header length */ 261 m->m_pkthdr.len += optlen; 262 plen = m->m_pkthdr.len - sizeof(*ip6); 263 264 /* If this is a jumbo payload, insert a jumbo payload option. */ 265 if (plen > IPV6_MAXPACKET) { 266 if (!hdrsplit) { 267 if ((error = ip6_splithdr(m, &exthdrs)) != 0) { 268 m = NULL; 269 goto freehdrs; 270 } 271 m = exthdrs.ip6e_ip6; 272 hdrsplit++; 273 } 274 /* adjust pointer */ 275 ip6 = mtod(m, struct ip6_hdr *); 276 if ((error = ip6_insert_jumboopt(&exthdrs, plen)) != 0) 277 goto freehdrs; 278 ip6->ip6_plen = 0; 279 } else 280 ip6->ip6_plen = htons(plen); 281 282 /* 283 * Concatenate headers and fill in next header fields. 284 * Here we have, on "m" 285 * IPv6 payload 286 * and we insert headers accordingly. Finally, we should be getting: 287 * IPv6 hbh dest1 rthdr ah* [esp* dest2 payload] 288 * 289 * during the header composing process, "m" points to IPv6 header. 290 * "mprev" points to an extension header prior to esp. 291 */ 292 { 293 u_char *nexthdrp = &ip6->ip6_nxt; 294 struct mbuf *mprev = m; 295 296 /* 297 * we treat dest2 specially. this makes IPsec processing 298 * much easier. the goal here is to make mprev point the 299 * mbuf prior to dest2. 300 * 301 * result: IPv6 dest2 payload 302 * m and mprev will point to IPv6 header. 303 */ 304 if (exthdrs.ip6e_dest2) { 305 if (!hdrsplit) 306 panic("%s: assumption failed: hdr not split", 307 __func__); 308 exthdrs.ip6e_dest2->m_next = m->m_next; 309 m->m_next = exthdrs.ip6e_dest2; 310 *mtod(exthdrs.ip6e_dest2, u_char *) = ip6->ip6_nxt; 311 ip6->ip6_nxt = IPPROTO_DSTOPTS; 312 } 313 314 #define MAKE_CHAIN(m, mp, p, i)\ 315 do {\ 316 if (m) {\ 317 if (!hdrsplit) \ 318 panic("assumption failed: hdr not split"); \ 319 *mtod((m), u_char *) = *(p);\ 320 *(p) = (i);\ 321 p = mtod((m), u_char *);\ 322 (m)->m_next = (mp)->m_next;\ 323 (mp)->m_next = (m);\ 324 (mp) = (m);\ 325 }\ 326 } while (0) 327 /* 328 * result: IPv6 hbh dest1 rthdr dest2 payload 329 * m will point to IPv6 header. mprev will point to the 330 * extension header prior to dest2 (rthdr in the above case). 331 */ 332 MAKE_CHAIN(exthdrs.ip6e_hbh, mprev, nexthdrp, IPPROTO_HOPOPTS); 333 MAKE_CHAIN(exthdrs.ip6e_dest1, mprev, nexthdrp, 334 IPPROTO_DSTOPTS); 335 MAKE_CHAIN(exthdrs.ip6e_rthdr, mprev, nexthdrp, 336 IPPROTO_ROUTING); 337 } 338 339 /* 340 * If there is a routing header, replace the destination address field 341 * with the first hop of the routing header. 342 */ 343 if (exthdrs.ip6e_rthdr) { 344 struct ip6_rthdr *rh; 345 struct ip6_rthdr0 *rh0; 346 struct in6_addr *addr; 347 348 rh = (struct ip6_rthdr *)(mtod(exthdrs.ip6e_rthdr, 349 struct ip6_rthdr *)); 350 switch (rh->ip6r_type) { 351 case IPV6_RTHDR_TYPE_0: 352 rh0 = (struct ip6_rthdr0 *)rh; 353 addr = (struct in6_addr *)(rh0 + 1); 354 ip6->ip6_dst = addr[0]; 355 bcopy(&addr[1], &addr[0], 356 sizeof(struct in6_addr) * (rh0->ip6r0_segleft - 1)); 357 addr[rh0->ip6r0_segleft - 1] = finaldst; 358 break; 359 default: /* is it possible? */ 360 error = EINVAL; 361 goto bad; 362 } 363 } 364 365 /* Source address validation */ 366 if (!(flags & IPV6_UNSPECSRC) && 367 IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src)) { 368 /* 369 * XXX: we can probably assume validation in the caller, but 370 * we explicitly check the address here for safety. 371 */ 372 error = EOPNOTSUPP; 373 ip6stat_inc(ip6s_badscope); 374 goto bad; 375 } 376 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) { 377 error = EOPNOTSUPP; 378 ip6stat_inc(ip6s_badscope); 379 goto bad; 380 } 381 382 ip6stat_inc(ip6s_localout); 383 384 /* 385 * Route packet. 386 */ 387 #if NPF > 0 388 reroute: 389 #endif 390 391 /* initialize cached route */ 392 if (ro == NULL) { 393 ro = &iproute; 394 ro->ro_rt = NULL; 395 } 396 ro_pmtu = ro; 397 if (opt && opt->ip6po_rthdr) 398 ro = &opt->ip6po_route; 399 dst = &ro->ro_dstsin6; 400 401 /* 402 * if specified, try to fill in the traffic class field. 403 * do not override if a non-zero value is already set. 404 * we check the diffserv field and the ecn field separately. 405 */ 406 if (opt && opt->ip6po_tclass >= 0) { 407 int mask = 0; 408 409 if ((ip6->ip6_flow & htonl(0xfc << 20)) == 0) 410 mask |= 0xfc; 411 if ((ip6->ip6_flow & htonl(0x03 << 20)) == 0) 412 mask |= 0x03; 413 if (mask != 0) 414 ip6->ip6_flow |= 415 htonl((opt->ip6po_tclass & mask) << 20); 416 } 417 418 /* fill in or override the hop limit field, if necessary. */ 419 if (opt && opt->ip6po_hlim != -1) 420 ip6->ip6_hlim = opt->ip6po_hlim & 0xff; 421 else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) { 422 if (im6o != NULL) 423 ip6->ip6_hlim = im6o->im6o_hlim; 424 else 425 ip6->ip6_hlim = ip6_defmcasthlim; 426 } 427 428 #ifdef IPSEC 429 if (tdb != NULL) { 430 /* 431 * XXX what should we do if ip6_hlim == 0 and the 432 * packet gets tunneled? 433 */ 434 /* 435 * if we are source-routing, do not attempt to tunnel the 436 * packet just because ip6_dst is different from what tdb has. 437 * XXX 438 */ 439 error = ip6_output_ipsec_send(tdb, m, ro, 440 exthdrs.ip6e_rthdr ? 1 : 0, 0); 441 goto done; 442 } 443 #endif /* IPSEC */ 444 445 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) { 446 struct in6_pktinfo *pi = NULL; 447 448 /* 449 * If the caller specify the outgoing interface 450 * explicitly, use it. 451 */ 452 if (opt != NULL && (pi = opt->ip6po_pktinfo) != NULL) 453 ifp = if_get(pi->ipi6_ifindex); 454 455 if (ifp == NULL && im6o != NULL) 456 ifp = if_get(im6o->im6o_ifidx); 457 } 458 459 if (ifp == NULL) { 460 rt = in6_selectroute(&ip6->ip6_dst, opt, ro, 461 m->m_pkthdr.ph_rtableid); 462 if (rt == NULL) { 463 ip6stat_inc(ip6s_noroute); 464 error = EHOSTUNREACH; 465 goto bad; 466 } 467 if (ISSET(rt->rt_flags, RTF_LOCAL)) 468 ifp = if_get(rtable_loindex(m->m_pkthdr.ph_rtableid)); 469 else 470 ifp = if_get(rt->rt_ifidx); 471 /* 472 * We aren't using rtisvalid() here because the UP/DOWN state 473 * machine is broken with some Ethernet drivers like em(4). 474 * As a result we might try to use an invalid cached route 475 * entry while an interface is being detached. 476 */ 477 if (ifp == NULL) { 478 ip6stat_inc(ip6s_noroute); 479 error = EHOSTUNREACH; 480 goto bad; 481 } 482 } else { 483 route6_cache(ro, &ip6->ip6_dst, NULL, m->m_pkthdr.ph_rtableid); 484 } 485 486 if (rt && (rt->rt_flags & RTF_GATEWAY) && 487 !IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) 488 dst = satosin6(rt->rt_gateway); 489 490 if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) { 491 /* Unicast */ 492 493 m->m_flags &= ~(M_BCAST | M_MCAST); /* just in case */ 494 } else { 495 /* Multicast */ 496 497 m->m_flags = (m->m_flags & ~M_BCAST) | M_MCAST; 498 499 /* 500 * Confirm that the outgoing interface supports multicast. 501 */ 502 if ((ifp->if_flags & IFF_MULTICAST) == 0) { 503 ip6stat_inc(ip6s_noroute); 504 error = ENETUNREACH; 505 goto bad; 506 } 507 508 if ((im6o == NULL || im6o->im6o_loop) && 509 in6_hasmulti(&ip6->ip6_dst, ifp)) { 510 /* 511 * If we belong to the destination multicast group 512 * on the outgoing interface, and the caller did not 513 * forbid loopback, loop back a copy. 514 * Can't defer TCP/UDP checksumming, do the 515 * computation now. 516 */ 517 in6_proto_cksum_out(m, NULL); 518 ip6_mloopback(ifp, m, dst); 519 } 520 #ifdef MROUTING 521 else { 522 /* 523 * If we are acting as a multicast router, perform 524 * multicast forwarding as if the packet had just 525 * arrived on the interface to which we are about 526 * to send. The multicast forwarding function 527 * recursively calls this function, using the 528 * IPV6_FORWARDING flag to prevent infinite recursion. 529 * 530 * Multicasts that are looped back by ip6_mloopback(), 531 * above, will be forwarded by the ip6_input() routine, 532 * if necessary. 533 */ 534 if (ip6_mforwarding && ip6_mrouter[ifp->if_rdomain] && 535 (flags & IPV6_FORWARDING) == 0) { 536 if (ip6_mforward(ip6, ifp, m, flags) != 0) { 537 m_freem(m); 538 goto done; 539 } 540 } 541 } 542 #endif 543 /* 544 * Multicasts with a hoplimit of zero may be looped back, 545 * above, but must not be transmitted on a network. 546 * Also, multicasts addressed to the loopback interface 547 * are not sent -- the above call to ip6_mloopback() will 548 * loop back a copy if this host actually belongs to the 549 * destination group on the loopback interface. 550 */ 551 if (ip6->ip6_hlim == 0 || (ifp->if_flags & IFF_LOOPBACK) || 552 IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst)) { 553 m_freem(m); 554 goto done; 555 } 556 } 557 558 /* 559 * If this packet is going through a loopback interface we won't 560 * be able to restore its scope ID using the interface index. 561 */ 562 if (IN6_IS_SCOPE_EMBED(&ip6->ip6_src)) { 563 if (ifp->if_flags & IFF_LOOPBACK) 564 src_scope = ip6->ip6_src.s6_addr16[1]; 565 ip6->ip6_src.s6_addr16[1] = 0; 566 } 567 if (IN6_IS_SCOPE_EMBED(&ip6->ip6_dst)) { 568 if (ifp->if_flags & IFF_LOOPBACK) 569 dst_scope = ip6->ip6_dst.s6_addr16[1]; 570 ip6->ip6_dst.s6_addr16[1] = 0; 571 } 572 573 /* Determine path MTU. */ 574 if ((error = ip6_getpmtu(ro_pmtu->ro_rt, ifp, &mtu)) != 0) 575 goto bad; 576 577 /* 578 * The caller of this function may specify to use the minimum MTU 579 * in some cases. 580 * An advanced API option (IPV6_USE_MIN_MTU) can also override MTU 581 * setting. The logic is a bit complicated; by default, unicast 582 * packets will follow path MTU while multicast packets will be sent at 583 * the minimum MTU. If IP6PO_MINMTU_ALL is specified, all packets 584 * including unicast ones will be sent at the minimum MTU. Multicast 585 * packets will always be sent at the minimum MTU unless 586 * IP6PO_MINMTU_DISABLE is explicitly specified. 587 * See RFC 3542 for more details. 588 */ 589 if (mtu > IPV6_MMTU) { 590 if ((flags & IPV6_MINMTU)) 591 mtu = IPV6_MMTU; 592 else if (opt && opt->ip6po_minmtu == IP6PO_MINMTU_ALL) 593 mtu = IPV6_MMTU; 594 else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) && (opt == NULL || 595 opt->ip6po_minmtu != IP6PO_MINMTU_DISABLE)) { 596 mtu = IPV6_MMTU; 597 } 598 } 599 600 /* 601 * If the outgoing packet contains a hop-by-hop options header, 602 * it must be examined and processed even by the source node. 603 * (RFC 2460, section 4.) 604 */ 605 if (exthdrs.ip6e_hbh) { 606 struct ip6_hbh *hbh = mtod(exthdrs.ip6e_hbh, struct ip6_hbh *); 607 u_int32_t rtalert; /* returned value is ignored */ 608 u_int32_t plen = 0; /* no more than 1 jumbo payload option! */ 609 610 m->m_pkthdr.ph_ifidx = ifp->if_index; 611 if (ip6_process_hopopts(&m, (u_int8_t *)(hbh + 1), 612 ((hbh->ip6h_len + 1) << 3) - sizeof(struct ip6_hbh), 613 &rtalert, &plen) < 0) { 614 /* m was already freed at this point */ 615 error = EINVAL;/* better error? */ 616 goto done; 617 } 618 m->m_pkthdr.ph_ifidx = 0; 619 } 620 621 #if NPF > 0 622 if (pf_test(AF_INET6, PF_OUT, ifp, &m) != PF_PASS) { 623 error = EACCES; 624 m_freem(m); 625 goto done; 626 } 627 if (m == NULL) 628 goto done; 629 ip6 = mtod(m, struct ip6_hdr *); 630 if ((m->m_pkthdr.pf.flags & (PF_TAG_REROUTE | PF_TAG_GENERATED)) == 631 (PF_TAG_REROUTE | PF_TAG_GENERATED)) { 632 /* already rerun the route lookup, go on */ 633 m->m_pkthdr.pf.flags &= ~(PF_TAG_GENERATED | PF_TAG_REROUTE); 634 } else if (m->m_pkthdr.pf.flags & PF_TAG_REROUTE) { 635 /* tag as generated to skip over pf_test on rerun */ 636 m->m_pkthdr.pf.flags |= PF_TAG_GENERATED; 637 finaldst = ip6->ip6_dst; 638 if (ro == &iproute) 639 rtfree(ro->ro_rt); 640 ro = NULL; 641 if_put(ifp); /* drop reference since destination changed */ 642 ifp = NULL; 643 goto reroute; 644 } 645 #endif 646 647 #ifdef IPSEC 648 if (ISSET(flags, IPV6_FORWARDING) && 649 ISSET(flags, IPV6_FORWARDING_IPSEC) && 650 !ISSET(m->m_pkthdr.ph_tagsset, PACKET_TAG_IPSEC_IN_DONE)) { 651 error = EHOSTUNREACH; 652 goto bad; 653 } 654 #endif 655 656 /* 657 * If the packet is not going on the wire it can be destined 658 * to any local address. In this case do not clear its scopes 659 * to let ip6_input() find a matching local route. 660 */ 661 if (ifp->if_flags & IFF_LOOPBACK) { 662 if (IN6_IS_SCOPE_EMBED(&ip6->ip6_src)) 663 ip6->ip6_src.s6_addr16[1] = src_scope; 664 if (IN6_IS_SCOPE_EMBED(&ip6->ip6_dst)) 665 ip6->ip6_dst.s6_addr16[1] = dst_scope; 666 } 667 668 /* 669 * Send the packet to the outgoing interface. 670 * If necessary, do IPv6 fragmentation before sending. 671 * 672 * the logic here is rather complex: 673 * 1: normal case (dontfrag == 0) 674 * 1-a: send as is if tlen <= path mtu 675 * 1-b: fragment if tlen > path mtu 676 * 677 * 2: if user asks us not to fragment (dontfrag == 1) 678 * 2-a: send as is if tlen <= interface mtu 679 * 2-b: error if tlen > interface mtu 680 */ 681 tlen = ISSET(m->m_pkthdr.csum_flags, M_TCP_TSO) ? 682 m->m_pkthdr.ph_mss : m->m_pkthdr.len; 683 684 if (ISSET(m->m_pkthdr.csum_flags, M_IPV6_DF_OUT)) { 685 CLR(m->m_pkthdr.csum_flags, M_IPV6_DF_OUT); 686 dontfrag = 1; 687 } else if (opt && ISSET(opt->ip6po_flags, IP6PO_DONTFRAG)) 688 dontfrag = 1; 689 else 690 dontfrag = 0; 691 692 if (dontfrag && tlen > ifp->if_mtu) { /* case 2-b */ 693 #ifdef IPSEC 694 if (ip_mtudisc) 695 ipsec_adjust_mtu(m, mtu); 696 #endif 697 error = EMSGSIZE; 698 goto bad; 699 } 700 701 /* 702 * transmit packet without fragmentation 703 */ 704 if (dontfrag || tlen <= mtu) { /* case 1-a and 2-a */ 705 error = if_output_tso(ifp, &m, sin6tosa(dst), ro->ro_rt, 706 ifp->if_mtu); 707 if (error || m == NULL) 708 goto done; 709 goto bad; /* should not happen */ 710 } 711 712 /* 713 * try to fragment the packet. case 1-b 714 */ 715 if (mtu < IPV6_MMTU) { 716 /* path MTU cannot be less than IPV6_MMTU */ 717 error = EMSGSIZE; 718 goto bad; 719 } else if (ip6->ip6_plen == 0) { 720 /* jumbo payload cannot be fragmented */ 721 error = EMSGSIZE; 722 goto bad; 723 } 724 725 /* 726 * Too large for the destination or interface; 727 * fragment if possible. 728 * Must be able to put at least 8 bytes per fragment. 729 */ 730 hlen = unfragpartlen; 731 if (mtu > IPV6_MAXPACKET) 732 mtu = IPV6_MAXPACKET; 733 734 /* 735 * If we are doing fragmentation, we can't defer TCP/UDP 736 * checksumming; compute the checksum and clear the flag. 737 */ 738 in6_proto_cksum_out(m, NULL); 739 740 /* 741 * Change the next header field of the last header in the 742 * unfragmentable part. 743 */ 744 if (exthdrs.ip6e_rthdr) { 745 nextproto = *mtod(exthdrs.ip6e_rthdr, u_char *); 746 *mtod(exthdrs.ip6e_rthdr, u_char *) = IPPROTO_FRAGMENT; 747 } else if (exthdrs.ip6e_dest1) { 748 nextproto = *mtod(exthdrs.ip6e_dest1, u_char *); 749 *mtod(exthdrs.ip6e_dest1, u_char *) = IPPROTO_FRAGMENT; 750 } else if (exthdrs.ip6e_hbh) { 751 nextproto = *mtod(exthdrs.ip6e_hbh, u_char *); 752 *mtod(exthdrs.ip6e_hbh, u_char *) = IPPROTO_FRAGMENT; 753 } else { 754 nextproto = ip6->ip6_nxt; 755 ip6->ip6_nxt = IPPROTO_FRAGMENT; 756 } 757 758 if ((error = ip6_fragment(m, &ml, hlen, nextproto, mtu)) || 759 (error = if_output_ml(ifp, &ml, sin6tosa(dst), ro->ro_rt))) 760 goto done; 761 ip6stat_inc(ip6s_fragmented); 762 goto done; 763 764 freehdrs: 765 m_freem(exthdrs.ip6e_hbh); /* m_freem will check if mbuf is 0 */ 766 m_freem(exthdrs.ip6e_dest1); 767 m_freem(exthdrs.ip6e_rthdr); 768 m_freem(exthdrs.ip6e_dest2); 769 bad: 770 m_freem(m); 771 done: 772 if (ro == &iproute) 773 rtfree(ro->ro_rt); 774 else if (ro_pmtu == &iproute) 775 rtfree(ro_pmtu->ro_rt); 776 if_put(ifp); 777 #ifdef IPSEC 778 tdb_unref(tdb); 779 #endif /* IPSEC */ 780 return (error); 781 } 782 783 int 784 ip6_fragment(struct mbuf *m0, struct mbuf_list *ml, int hlen, u_char nextproto, 785 u_long mtu) 786 { 787 struct ip6_hdr *ip6; 788 u_int32_t id; 789 int tlen, len, off; 790 int error; 791 792 ml_init(ml); 793 794 ip6 = mtod(m0, struct ip6_hdr *); 795 tlen = m0->m_pkthdr.len; 796 len = (mtu - hlen - sizeof(struct ip6_frag)) & ~7; 797 if (len < 8) { 798 error = EMSGSIZE; 799 goto bad; 800 } 801 id = htonl(ip6_randomid()); 802 803 /* 804 * Loop through length of payload, 805 * make new header and copy data of each part and link onto chain. 806 */ 807 for (off = hlen; off < tlen; off += len) { 808 struct mbuf *m; 809 struct mbuf *mlast; 810 struct ip6_hdr *mhip6; 811 struct ip6_frag *ip6f; 812 813 MGETHDR(m, M_DONTWAIT, MT_HEADER); 814 if (m == NULL) { 815 error = ENOBUFS; 816 goto bad; 817 } 818 ml_enqueue(ml, m); 819 if ((error = m_dup_pkthdr(m, m0, M_DONTWAIT)) != 0) 820 goto bad; 821 m->m_data += max_linkhdr; 822 mhip6 = mtod(m, struct ip6_hdr *); 823 *mhip6 = *ip6; 824 m->m_len = sizeof(struct ip6_hdr); 825 826 if ((error = ip6_insertfraghdr(m0, m, hlen, &ip6f)) != 0) 827 goto bad; 828 ip6f->ip6f_offlg = htons((off - hlen) & ~7); 829 if (off + len >= tlen) 830 len = tlen - off; 831 else 832 ip6f->ip6f_offlg |= IP6F_MORE_FRAG; 833 834 m->m_pkthdr.len = hlen + sizeof(struct ip6_frag) + len; 835 mhip6->ip6_plen = htons(m->m_pkthdr.len - 836 sizeof(struct ip6_hdr)); 837 for (mlast = m; mlast->m_next; mlast = mlast->m_next) 838 ; 839 mlast->m_next = m_copym(m0, off, len, M_DONTWAIT); 840 if (mlast->m_next == NULL) { 841 error = ENOBUFS; 842 goto bad; 843 } 844 845 ip6f->ip6f_reserved = 0; 846 ip6f->ip6f_ident = id; 847 ip6f->ip6f_nxt = nextproto; 848 } 849 850 ip6stat_add(ip6s_ofragments, ml_len(ml)); 851 m_freem(m0); 852 return (0); 853 854 bad: 855 ip6stat_inc(ip6s_odropped); 856 ml_purge(ml); 857 m_freem(m0); 858 return (error); 859 } 860 861 int 862 ip6_copyexthdr(struct mbuf **mp, caddr_t hdr, int hlen) 863 { 864 struct mbuf *m; 865 866 if (hlen > MCLBYTES) 867 return (ENOBUFS); /* XXX */ 868 869 MGET(m, M_DONTWAIT, MT_DATA); 870 if (!m) 871 return (ENOBUFS); 872 873 if (hlen > MLEN) { 874 MCLGET(m, M_DONTWAIT); 875 if ((m->m_flags & M_EXT) == 0) { 876 m_free(m); 877 return (ENOBUFS); 878 } 879 } 880 m->m_len = hlen; 881 if (hdr) 882 memcpy(mtod(m, caddr_t), hdr, hlen); 883 884 *mp = m; 885 return (0); 886 } 887 888 /* 889 * Insert jumbo payload option. 890 */ 891 int 892 ip6_insert_jumboopt(struct ip6_exthdrs *exthdrs, u_int32_t plen) 893 { 894 struct mbuf *mopt; 895 u_int8_t *optbuf; 896 u_int32_t v; 897 898 #define JUMBOOPTLEN 8 /* length of jumbo payload option and padding */ 899 900 /* 901 * If there is no hop-by-hop options header, allocate new one. 902 * If there is one but it doesn't have enough space to store the 903 * jumbo payload option, allocate a cluster to store the whole options. 904 * Otherwise, use it to store the options. 905 */ 906 if (exthdrs->ip6e_hbh == 0) { 907 MGET(mopt, M_DONTWAIT, MT_DATA); 908 if (mopt == NULL) 909 return (ENOBUFS); 910 mopt->m_len = JUMBOOPTLEN; 911 optbuf = mtod(mopt, u_int8_t *); 912 optbuf[1] = 0; /* = ((JUMBOOPTLEN) >> 3) - 1 */ 913 exthdrs->ip6e_hbh = mopt; 914 } else { 915 struct ip6_hbh *hbh; 916 917 mopt = exthdrs->ip6e_hbh; 918 if (m_trailingspace(mopt) < JUMBOOPTLEN) { 919 /* 920 * XXX assumption: 921 * - exthdrs->ip6e_hbh is not referenced from places 922 * other than exthdrs. 923 * - exthdrs->ip6e_hbh is not an mbuf chain. 924 */ 925 int oldoptlen = mopt->m_len; 926 struct mbuf *n; 927 928 /* 929 * XXX: give up if the whole (new) hbh header does 930 * not fit even in an mbuf cluster. 931 */ 932 if (oldoptlen + JUMBOOPTLEN > MCLBYTES) 933 return (ENOBUFS); 934 935 /* 936 * As a consequence, we must always prepare a cluster 937 * at this point. 938 */ 939 MGET(n, M_DONTWAIT, MT_DATA); 940 if (n) { 941 MCLGET(n, M_DONTWAIT); 942 if ((n->m_flags & M_EXT) == 0) { 943 m_freem(n); 944 n = NULL; 945 } 946 } 947 if (!n) 948 return (ENOBUFS); 949 n->m_len = oldoptlen + JUMBOOPTLEN; 950 memcpy(mtod(n, caddr_t), mtod(mopt, caddr_t), 951 oldoptlen); 952 optbuf = mtod(n, u_int8_t *) + oldoptlen; 953 m_freem(mopt); 954 mopt = exthdrs->ip6e_hbh = n; 955 } else { 956 optbuf = mtod(mopt, u_int8_t *) + mopt->m_len; 957 mopt->m_len += JUMBOOPTLEN; 958 } 959 optbuf[0] = IP6OPT_PADN; 960 optbuf[1] = 0; 961 962 /* 963 * Adjust the header length according to the pad and 964 * the jumbo payload option. 965 */ 966 hbh = mtod(mopt, struct ip6_hbh *); 967 hbh->ip6h_len += (JUMBOOPTLEN >> 3); 968 } 969 970 /* fill in the option. */ 971 optbuf[2] = IP6OPT_JUMBO; 972 optbuf[3] = 4; 973 v = (u_int32_t)htonl(plen + JUMBOOPTLEN); 974 memcpy(&optbuf[4], &v, sizeof(u_int32_t)); 975 976 /* finally, adjust the packet header length */ 977 exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN; 978 979 return (0); 980 #undef JUMBOOPTLEN 981 } 982 983 /* 984 * Insert fragment header and copy unfragmentable header portions. 985 */ 986 int 987 ip6_insertfraghdr(struct mbuf *m0, struct mbuf *m, int hlen, 988 struct ip6_frag **frghdrp) 989 { 990 struct mbuf *n, *mlast; 991 992 if (hlen > sizeof(struct ip6_hdr)) { 993 n = m_copym(m0, sizeof(struct ip6_hdr), 994 hlen - sizeof(struct ip6_hdr), M_DONTWAIT); 995 if (n == NULL) 996 return (ENOBUFS); 997 m->m_next = n; 998 } else 999 n = m; 1000 1001 /* Search for the last mbuf of unfragmentable part. */ 1002 for (mlast = n; mlast->m_next; mlast = mlast->m_next) 1003 ; 1004 1005 if ((mlast->m_flags & M_EXT) == 0 && 1006 m_trailingspace(mlast) >= sizeof(struct ip6_frag)) { 1007 /* use the trailing space of the last mbuf for fragment hdr */ 1008 *frghdrp = (struct ip6_frag *)(mtod(mlast, caddr_t) + 1009 mlast->m_len); 1010 mlast->m_len += sizeof(struct ip6_frag); 1011 m->m_pkthdr.len += sizeof(struct ip6_frag); 1012 } else { 1013 /* allocate a new mbuf for the fragment header */ 1014 struct mbuf *mfrg; 1015 1016 MGET(mfrg, M_DONTWAIT, MT_DATA); 1017 if (mfrg == NULL) 1018 return (ENOBUFS); 1019 mfrg->m_len = sizeof(struct ip6_frag); 1020 *frghdrp = mtod(mfrg, struct ip6_frag *); 1021 mlast->m_next = mfrg; 1022 } 1023 1024 return (0); 1025 } 1026 1027 int 1028 ip6_getpmtu(struct rtentry *rt, struct ifnet *ifp, u_long *mtup) 1029 { 1030 u_int mtu, rtmtu; 1031 int error = 0; 1032 1033 if (rt != NULL) { 1034 mtu = rtmtu = atomic_load_int(&rt->rt_mtu); 1035 if (mtu == 0) 1036 mtu = ifp->if_mtu; 1037 else if (mtu < IPV6_MMTU) { 1038 /* RFC8021 IPv6 Atomic Fragments Considered Harmful */ 1039 mtu = IPV6_MMTU; 1040 } else if (mtu > ifp->if_mtu) { 1041 /* 1042 * The MTU on the route is larger than the MTU on 1043 * the interface! This shouldn't happen, unless the 1044 * MTU of the interface has been changed after the 1045 * interface was brought up. Change the MTU in the 1046 * route to match the interface MTU (as long as the 1047 * field isn't locked). 1048 */ 1049 mtu = ifp->if_mtu; 1050 if (!(rt->rt_locks & RTV_MTU)) 1051 atomic_cas_uint(&rt->rt_mtu, rtmtu, mtu); 1052 } 1053 } else { 1054 mtu = ifp->if_mtu; 1055 } 1056 1057 *mtup = mtu; 1058 return (error); 1059 } 1060 1061 /* 1062 * IP6 socket option processing. 1063 */ 1064 int 1065 ip6_ctloutput(int op, struct socket *so, int level, int optname, 1066 struct mbuf *m) 1067 { 1068 int privileged, optdatalen, uproto; 1069 void *optdata; 1070 struct inpcb *inp = sotoinpcb(so); 1071 int error, optval; 1072 struct proc *p = curproc; /* For IPsec and rdomain */ 1073 u_int rtableid, rtid = 0; 1074 1075 error = optval = 0; 1076 1077 privileged = (inp->inp_socket->so_state & SS_PRIV); 1078 uproto = (int)so->so_proto->pr_protocol; 1079 1080 if (level != IPPROTO_IPV6) 1081 return (EINVAL); 1082 1083 rtableid = p->p_p->ps_rtableid; 1084 1085 switch (op) { 1086 case PRCO_SETOPT: 1087 switch (optname) { 1088 /* 1089 * Use of some Hop-by-Hop options or some 1090 * Destination options, might require special 1091 * privilege. That is, normal applications 1092 * (without special privilege) might be forbidden 1093 * from setting certain options in outgoing packets, 1094 * and might never see certain options in received 1095 * packets. [RFC 2292 Section 6] 1096 * KAME specific note: 1097 * KAME prevents non-privileged users from sending or 1098 * receiving ANY hbh/dst options in order to avoid 1099 * overhead of parsing options in the kernel. 1100 */ 1101 case IPV6_RECVHOPOPTS: 1102 case IPV6_RECVDSTOPTS: 1103 if (!privileged) { 1104 error = EPERM; 1105 break; 1106 } 1107 /* FALLTHROUGH */ 1108 case IPV6_UNICAST_HOPS: 1109 case IPV6_MINHOPCOUNT: 1110 case IPV6_HOPLIMIT: 1111 1112 case IPV6_RECVPKTINFO: 1113 case IPV6_RECVHOPLIMIT: 1114 case IPV6_RECVRTHDR: 1115 case IPV6_RECVPATHMTU: 1116 case IPV6_RECVTCLASS: 1117 case IPV6_V6ONLY: 1118 case IPV6_AUTOFLOWLABEL: 1119 case IPV6_RECVDSTPORT: 1120 if (m == NULL || m->m_len != sizeof(int)) { 1121 error = EINVAL; 1122 break; 1123 } 1124 optval = *mtod(m, int *); 1125 switch (optname) { 1126 1127 case IPV6_UNICAST_HOPS: 1128 if (optval < -1 || optval >= 256) 1129 error = EINVAL; 1130 else { 1131 /* -1 = kernel default */ 1132 inp->inp_hops = optval; 1133 } 1134 break; 1135 1136 case IPV6_MINHOPCOUNT: 1137 if (optval < 0 || optval > 255) 1138 error = EINVAL; 1139 else 1140 inp->inp_ip6_minhlim = optval; 1141 break; 1142 1143 #define OPTSET(bit) \ 1144 do { \ 1145 if (optval) \ 1146 inp->inp_flags |= (bit); \ 1147 else \ 1148 inp->inp_flags &= ~(bit); \ 1149 } while (/*CONSTCOND*/ 0) 1150 #define OPTBIT(bit) (inp->inp_flags & (bit) ? 1 : 0) 1151 1152 case IPV6_RECVPKTINFO: 1153 OPTSET(IN6P_PKTINFO); 1154 break; 1155 1156 case IPV6_HOPLIMIT: 1157 { 1158 struct ip6_pktopts **optp; 1159 1160 optp = &inp->inp_outputopts6; 1161 error = ip6_pcbopt(IPV6_HOPLIMIT, 1162 (u_char *)&optval, sizeof(optval), optp, 1163 privileged, uproto); 1164 break; 1165 } 1166 1167 case IPV6_RECVHOPLIMIT: 1168 OPTSET(IN6P_HOPLIMIT); 1169 break; 1170 1171 case IPV6_RECVHOPOPTS: 1172 OPTSET(IN6P_HOPOPTS); 1173 break; 1174 1175 case IPV6_RECVDSTOPTS: 1176 OPTSET(IN6P_DSTOPTS); 1177 break; 1178 1179 case IPV6_RECVRTHDR: 1180 OPTSET(IN6P_RTHDR); 1181 break; 1182 1183 case IPV6_RECVPATHMTU: 1184 /* 1185 * We ignore this option for TCP 1186 * sockets. 1187 * (RFC3542 leaves this case 1188 * unspecified.) 1189 */ 1190 if (uproto != IPPROTO_TCP) 1191 OPTSET(IN6P_MTU); 1192 break; 1193 1194 case IPV6_V6ONLY: 1195 /* 1196 * make setsockopt(IPV6_V6ONLY) 1197 * available only prior to bind(2). 1198 * see ipng mailing list, Jun 22 2001. 1199 */ 1200 if (inp->inp_lport || !IN6_IS_ADDR_UNSPECIFIED( 1201 &inp->inp_laddr6)) { 1202 error = EINVAL; 1203 break; 1204 } 1205 /* No support for IPv4-mapped addresses. */ 1206 if (!optval) 1207 error = EINVAL; 1208 else 1209 error = 0; 1210 break; 1211 case IPV6_RECVTCLASS: 1212 OPTSET(IN6P_TCLASS); 1213 break; 1214 case IPV6_AUTOFLOWLABEL: 1215 OPTSET(IN6P_AUTOFLOWLABEL); 1216 break; 1217 1218 case IPV6_RECVDSTPORT: 1219 OPTSET(IN6P_RECVDSTPORT); 1220 break; 1221 } 1222 break; 1223 1224 case IPV6_TCLASS: 1225 case IPV6_DONTFRAG: 1226 case IPV6_USE_MIN_MTU: 1227 if (m == NULL || m->m_len != sizeof(optval)) { 1228 error = EINVAL; 1229 break; 1230 } 1231 optval = *mtod(m, int *); 1232 { 1233 struct ip6_pktopts **optp; 1234 optp = &inp->inp_outputopts6; 1235 error = ip6_pcbopt(optname, (u_char *)&optval, 1236 sizeof(optval), optp, privileged, uproto); 1237 break; 1238 } 1239 1240 case IPV6_PKTINFO: 1241 case IPV6_HOPOPTS: 1242 case IPV6_RTHDR: 1243 case IPV6_DSTOPTS: 1244 case IPV6_RTHDRDSTOPTS: 1245 { 1246 /* new advanced API (RFC3542) */ 1247 u_char *optbuf; 1248 int optbuflen; 1249 struct ip6_pktopts **optp; 1250 1251 if (m && m->m_next) { 1252 error = EINVAL; /* XXX */ 1253 break; 1254 } 1255 if (m) { 1256 optbuf = mtod(m, u_char *); 1257 optbuflen = m->m_len; 1258 } else { 1259 optbuf = NULL; 1260 optbuflen = 0; 1261 } 1262 optp = &inp->inp_outputopts6; 1263 error = ip6_pcbopt(optname, optbuf, optbuflen, optp, 1264 privileged, uproto); 1265 break; 1266 } 1267 #undef OPTSET 1268 1269 case IPV6_MULTICAST_IF: 1270 case IPV6_MULTICAST_HOPS: 1271 case IPV6_MULTICAST_LOOP: 1272 case IPV6_JOIN_GROUP: 1273 case IPV6_LEAVE_GROUP: 1274 error = ip6_setmoptions(optname, 1275 &inp->inp_moptions6, 1276 m, inp->inp_rtableid); 1277 break; 1278 1279 case IPV6_PORTRANGE: 1280 if (m == NULL || m->m_len != sizeof(int)) { 1281 error = EINVAL; 1282 break; 1283 } 1284 optval = *mtod(m, int *); 1285 1286 switch (optval) { 1287 case IPV6_PORTRANGE_DEFAULT: 1288 inp->inp_flags &= ~(IN6P_LOWPORT); 1289 inp->inp_flags &= ~(IN6P_HIGHPORT); 1290 break; 1291 1292 case IPV6_PORTRANGE_HIGH: 1293 inp->inp_flags &= ~(IN6P_LOWPORT); 1294 inp->inp_flags |= IN6P_HIGHPORT; 1295 break; 1296 1297 case IPV6_PORTRANGE_LOW: 1298 inp->inp_flags &= ~(IN6P_HIGHPORT); 1299 inp->inp_flags |= IN6P_LOWPORT; 1300 break; 1301 1302 default: 1303 error = EINVAL; 1304 break; 1305 } 1306 break; 1307 1308 case IPSEC6_OUTSA: 1309 error = EINVAL; 1310 break; 1311 1312 case IPV6_AUTH_LEVEL: 1313 case IPV6_ESP_TRANS_LEVEL: 1314 case IPV6_ESP_NETWORK_LEVEL: 1315 case IPV6_IPCOMP_LEVEL: 1316 #ifndef IPSEC 1317 error = EINVAL; 1318 #else 1319 if (m == NULL || m->m_len != sizeof(int)) { 1320 error = EINVAL; 1321 break; 1322 } 1323 optval = *mtod(m, int *); 1324 1325 if (optval < IPSEC_LEVEL_BYPASS || 1326 optval > IPSEC_LEVEL_UNIQUE) { 1327 error = EINVAL; 1328 break; 1329 } 1330 1331 switch (optname) { 1332 case IPV6_AUTH_LEVEL: 1333 if (optval < IPSEC_AUTH_LEVEL_DEFAULT && 1334 suser(p)) { 1335 error = EACCES; 1336 break; 1337 } 1338 inp->inp_seclevel.sl_auth = optval; 1339 break; 1340 1341 case IPV6_ESP_TRANS_LEVEL: 1342 if (optval < IPSEC_ESP_TRANS_LEVEL_DEFAULT && 1343 suser(p)) { 1344 error = EACCES; 1345 break; 1346 } 1347 inp->inp_seclevel.sl_esp_trans = optval; 1348 break; 1349 1350 case IPV6_ESP_NETWORK_LEVEL: 1351 if (optval < IPSEC_ESP_NETWORK_LEVEL_DEFAULT && 1352 suser(p)) { 1353 error = EACCES; 1354 break; 1355 } 1356 inp->inp_seclevel.sl_esp_network = optval; 1357 break; 1358 1359 case IPV6_IPCOMP_LEVEL: 1360 if (optval < IPSEC_IPCOMP_LEVEL_DEFAULT && 1361 suser(p)) { 1362 error = EACCES; 1363 break; 1364 } 1365 inp->inp_seclevel.sl_ipcomp = optval; 1366 break; 1367 } 1368 #endif 1369 break; 1370 case SO_RTABLE: 1371 if (m == NULL || m->m_len < sizeof(u_int)) { 1372 error = EINVAL; 1373 break; 1374 } 1375 rtid = *mtod(m, u_int *); 1376 if (inp->inp_rtableid == rtid) 1377 break; 1378 /* needs privileges to switch when already set */ 1379 if (rtableid != rtid && rtableid != 0 && 1380 (error = suser(p)) != 0) 1381 break; 1382 error = in_pcbset_rtableid(inp, rtid); 1383 break; 1384 case IPV6_PIPEX: 1385 if (m != NULL && m->m_len == sizeof(int)) 1386 inp->inp_pipex = *mtod(m, int *); 1387 else 1388 error = EINVAL; 1389 break; 1390 1391 default: 1392 error = ENOPROTOOPT; 1393 break; 1394 } 1395 break; 1396 1397 case PRCO_GETOPT: 1398 switch (optname) { 1399 1400 case IPV6_RECVHOPOPTS: 1401 case IPV6_RECVDSTOPTS: 1402 case IPV6_UNICAST_HOPS: 1403 case IPV6_MINHOPCOUNT: 1404 case IPV6_RECVPKTINFO: 1405 case IPV6_RECVHOPLIMIT: 1406 case IPV6_RECVRTHDR: 1407 case IPV6_RECVPATHMTU: 1408 1409 case IPV6_V6ONLY: 1410 case IPV6_PORTRANGE: 1411 case IPV6_RECVTCLASS: 1412 case IPV6_AUTOFLOWLABEL: 1413 case IPV6_RECVDSTPORT: 1414 switch (optname) { 1415 1416 case IPV6_RECVHOPOPTS: 1417 optval = OPTBIT(IN6P_HOPOPTS); 1418 break; 1419 1420 case IPV6_RECVDSTOPTS: 1421 optval = OPTBIT(IN6P_DSTOPTS); 1422 break; 1423 1424 case IPV6_UNICAST_HOPS: 1425 optval = inp->inp_hops; 1426 break; 1427 1428 case IPV6_MINHOPCOUNT: 1429 optval = inp->inp_ip6_minhlim; 1430 break; 1431 1432 case IPV6_RECVPKTINFO: 1433 optval = OPTBIT(IN6P_PKTINFO); 1434 break; 1435 1436 case IPV6_RECVHOPLIMIT: 1437 optval = OPTBIT(IN6P_HOPLIMIT); 1438 break; 1439 1440 case IPV6_RECVRTHDR: 1441 optval = OPTBIT(IN6P_RTHDR); 1442 break; 1443 1444 case IPV6_RECVPATHMTU: 1445 optval = OPTBIT(IN6P_MTU); 1446 break; 1447 1448 case IPV6_V6ONLY: 1449 optval = 1; 1450 break; 1451 1452 case IPV6_PORTRANGE: 1453 { 1454 int flags; 1455 flags = inp->inp_flags; 1456 if (flags & IN6P_HIGHPORT) 1457 optval = IPV6_PORTRANGE_HIGH; 1458 else if (flags & IN6P_LOWPORT) 1459 optval = IPV6_PORTRANGE_LOW; 1460 else 1461 optval = 0; 1462 break; 1463 } 1464 case IPV6_RECVTCLASS: 1465 optval = OPTBIT(IN6P_TCLASS); 1466 break; 1467 1468 case IPV6_AUTOFLOWLABEL: 1469 optval = OPTBIT(IN6P_AUTOFLOWLABEL); 1470 break; 1471 1472 case IPV6_RECVDSTPORT: 1473 optval = OPTBIT(IN6P_RECVDSTPORT); 1474 break; 1475 } 1476 if (error) 1477 break; 1478 m->m_len = sizeof(int); 1479 *mtod(m, int *) = optval; 1480 break; 1481 1482 case IPV6_PATHMTU: 1483 { 1484 u_long pmtu = 0; 1485 struct ip6_mtuinfo mtuinfo; 1486 struct ifnet *ifp; 1487 struct rtentry *rt; 1488 1489 if (!(so->so_state & SS_ISCONNECTED)) 1490 return (ENOTCONN); 1491 1492 rt = in6_pcbrtentry(inp); 1493 if (!rtisvalid(rt)) 1494 return (EHOSTUNREACH); 1495 1496 ifp = if_get(rt->rt_ifidx); 1497 if (ifp == NULL) 1498 return (EHOSTUNREACH); 1499 /* 1500 * XXX: we dot not consider the case of source 1501 * routing, or optional information to specify 1502 * the outgoing interface. 1503 */ 1504 error = ip6_getpmtu(rt, ifp, &pmtu); 1505 if_put(ifp); 1506 if (error) 1507 break; 1508 if (pmtu > IPV6_MAXPACKET) 1509 pmtu = IPV6_MAXPACKET; 1510 1511 bzero(&mtuinfo, sizeof(mtuinfo)); 1512 mtuinfo.ip6m_mtu = (u_int32_t)pmtu; 1513 optdata = (void *)&mtuinfo; 1514 optdatalen = sizeof(mtuinfo); 1515 if (optdatalen > MCLBYTES) 1516 return (EMSGSIZE); /* XXX */ 1517 if (optdatalen > MLEN) 1518 MCLGET(m, M_WAIT); 1519 m->m_len = optdatalen; 1520 bcopy(optdata, mtod(m, void *), optdatalen); 1521 break; 1522 } 1523 1524 case IPV6_PKTINFO: 1525 case IPV6_HOPOPTS: 1526 case IPV6_RTHDR: 1527 case IPV6_DSTOPTS: 1528 case IPV6_RTHDRDSTOPTS: 1529 case IPV6_TCLASS: 1530 case IPV6_DONTFRAG: 1531 case IPV6_USE_MIN_MTU: 1532 error = ip6_getpcbopt(inp->inp_outputopts6, 1533 optname, m); 1534 break; 1535 1536 case IPV6_MULTICAST_IF: 1537 case IPV6_MULTICAST_HOPS: 1538 case IPV6_MULTICAST_LOOP: 1539 case IPV6_JOIN_GROUP: 1540 case IPV6_LEAVE_GROUP: 1541 error = ip6_getmoptions(optname, 1542 inp->inp_moptions6, m); 1543 break; 1544 1545 case IPSEC6_OUTSA: 1546 error = EINVAL; 1547 break; 1548 1549 case IPV6_AUTH_LEVEL: 1550 case IPV6_ESP_TRANS_LEVEL: 1551 case IPV6_ESP_NETWORK_LEVEL: 1552 case IPV6_IPCOMP_LEVEL: 1553 #ifndef IPSEC 1554 m->m_len = sizeof(int); 1555 *mtod(m, int *) = IPSEC_LEVEL_NONE; 1556 #else 1557 m->m_len = sizeof(int); 1558 switch (optname) { 1559 case IPV6_AUTH_LEVEL: 1560 optval = inp->inp_seclevel.sl_auth; 1561 break; 1562 1563 case IPV6_ESP_TRANS_LEVEL: 1564 optval = 1565 inp->inp_seclevel.sl_esp_trans; 1566 break; 1567 1568 case IPV6_ESP_NETWORK_LEVEL: 1569 optval = 1570 inp->inp_seclevel.sl_esp_network; 1571 break; 1572 1573 case IPV6_IPCOMP_LEVEL: 1574 optval = inp->inp_seclevel.sl_ipcomp; 1575 break; 1576 } 1577 *mtod(m, int *) = optval; 1578 #endif 1579 break; 1580 case SO_RTABLE: 1581 m->m_len = sizeof(u_int); 1582 *mtod(m, u_int *) = inp->inp_rtableid; 1583 break; 1584 case IPV6_PIPEX: 1585 m->m_len = sizeof(int); 1586 *mtod(m, int *) = inp->inp_pipex; 1587 break; 1588 1589 default: 1590 error = ENOPROTOOPT; 1591 break; 1592 } 1593 break; 1594 } 1595 return (error); 1596 } 1597 1598 int 1599 ip6_raw_ctloutput(int op, struct socket *so, int level, int optname, 1600 struct mbuf *m) 1601 { 1602 int error = 0, optval; 1603 const int icmp6off = offsetof(struct icmp6_hdr, icmp6_cksum); 1604 struct inpcb *inp = sotoinpcb(so); 1605 1606 if (level != IPPROTO_IPV6) 1607 return (EINVAL); 1608 1609 switch (optname) { 1610 case IPV6_CHECKSUM: 1611 /* 1612 * For ICMPv6 sockets, no modification allowed for checksum 1613 * offset, permit "no change" values to help existing apps. 1614 * 1615 * RFC3542 says: "An attempt to set IPV6_CHECKSUM 1616 * for an ICMPv6 socket will fail." 1617 * The current behavior does not meet RFC3542. 1618 */ 1619 switch (op) { 1620 case PRCO_SETOPT: 1621 if (m == NULL || m->m_len != sizeof(int)) { 1622 error = EINVAL; 1623 break; 1624 } 1625 optval = *mtod(m, int *); 1626 if (optval < -1 || 1627 (optval > 0 && (optval % 2) != 0)) { 1628 /* 1629 * The API assumes non-negative even offset 1630 * values or -1 as a special value. 1631 */ 1632 error = EINVAL; 1633 } else if (so->so_proto->pr_protocol == 1634 IPPROTO_ICMPV6) { 1635 if (optval != icmp6off) 1636 error = EINVAL; 1637 } else 1638 inp->inp_cksum6 = optval; 1639 break; 1640 1641 case PRCO_GETOPT: 1642 if (so->so_proto->pr_protocol == IPPROTO_ICMPV6) 1643 optval = icmp6off; 1644 else 1645 optval = inp->inp_cksum6; 1646 1647 m->m_len = sizeof(int); 1648 *mtod(m, int *) = optval; 1649 break; 1650 1651 default: 1652 error = EINVAL; 1653 break; 1654 } 1655 break; 1656 1657 default: 1658 error = ENOPROTOOPT; 1659 break; 1660 } 1661 1662 return (error); 1663 } 1664 1665 /* 1666 * initialize ip6_pktopts. beware that there are non-zero default values in 1667 * the struct. 1668 */ 1669 void 1670 ip6_initpktopts(struct ip6_pktopts *opt) 1671 { 1672 bzero(opt, sizeof(*opt)); 1673 opt->ip6po_hlim = -1; /* -1 means default hop limit */ 1674 opt->ip6po_tclass = -1; /* -1 means default traffic class */ 1675 opt->ip6po_minmtu = IP6PO_MINMTU_MCASTONLY; 1676 } 1677 1678 int 1679 ip6_pcbopt(int optname, u_char *buf, int len, struct ip6_pktopts **pktopt, 1680 int priv, int uproto) 1681 { 1682 struct ip6_pktopts *opt; 1683 1684 if (*pktopt == NULL) { 1685 *pktopt = malloc(sizeof(struct ip6_pktopts), M_IP6OPT, 1686 M_WAITOK); 1687 ip6_initpktopts(*pktopt); 1688 } 1689 opt = *pktopt; 1690 1691 return (ip6_setpktopt(optname, buf, len, opt, priv, 1, uproto)); 1692 } 1693 1694 int 1695 ip6_getpcbopt(struct ip6_pktopts *pktopt, int optname, struct mbuf *m) 1696 { 1697 void *optdata = NULL; 1698 int optdatalen = 0; 1699 struct ip6_ext *ip6e; 1700 int error = 0; 1701 struct in6_pktinfo null_pktinfo; 1702 int deftclass = 0, on; 1703 int defminmtu = IP6PO_MINMTU_MCASTONLY; 1704 1705 switch (optname) { 1706 case IPV6_PKTINFO: 1707 if (pktopt && pktopt->ip6po_pktinfo) 1708 optdata = (void *)pktopt->ip6po_pktinfo; 1709 else { 1710 /* XXX: we don't have to do this every time... */ 1711 bzero(&null_pktinfo, sizeof(null_pktinfo)); 1712 optdata = (void *)&null_pktinfo; 1713 } 1714 optdatalen = sizeof(struct in6_pktinfo); 1715 break; 1716 case IPV6_TCLASS: 1717 if (pktopt && pktopt->ip6po_tclass >= 0) 1718 optdata = (void *)&pktopt->ip6po_tclass; 1719 else 1720 optdata = (void *)&deftclass; 1721 optdatalen = sizeof(int); 1722 break; 1723 case IPV6_HOPOPTS: 1724 if (pktopt && pktopt->ip6po_hbh) { 1725 optdata = (void *)pktopt->ip6po_hbh; 1726 ip6e = (struct ip6_ext *)pktopt->ip6po_hbh; 1727 optdatalen = (ip6e->ip6e_len + 1) << 3; 1728 } 1729 break; 1730 case IPV6_RTHDR: 1731 if (pktopt && pktopt->ip6po_rthdr) { 1732 optdata = (void *)pktopt->ip6po_rthdr; 1733 ip6e = (struct ip6_ext *)pktopt->ip6po_rthdr; 1734 optdatalen = (ip6e->ip6e_len + 1) << 3; 1735 } 1736 break; 1737 case IPV6_RTHDRDSTOPTS: 1738 if (pktopt && pktopt->ip6po_dest1) { 1739 optdata = (void *)pktopt->ip6po_dest1; 1740 ip6e = (struct ip6_ext *)pktopt->ip6po_dest1; 1741 optdatalen = (ip6e->ip6e_len + 1) << 3; 1742 } 1743 break; 1744 case IPV6_DSTOPTS: 1745 if (pktopt && pktopt->ip6po_dest2) { 1746 optdata = (void *)pktopt->ip6po_dest2; 1747 ip6e = (struct ip6_ext *)pktopt->ip6po_dest2; 1748 optdatalen = (ip6e->ip6e_len + 1) << 3; 1749 } 1750 break; 1751 case IPV6_USE_MIN_MTU: 1752 if (pktopt) 1753 optdata = (void *)&pktopt->ip6po_minmtu; 1754 else 1755 optdata = (void *)&defminmtu; 1756 optdatalen = sizeof(int); 1757 break; 1758 case IPV6_DONTFRAG: 1759 if (pktopt && ((pktopt->ip6po_flags) & IP6PO_DONTFRAG)) 1760 on = 1; 1761 else 1762 on = 0; 1763 optdata = (void *)&on; 1764 optdatalen = sizeof(on); 1765 break; 1766 default: /* should not happen */ 1767 #ifdef DIAGNOSTIC 1768 panic("%s: unexpected option", __func__); 1769 #endif 1770 return (ENOPROTOOPT); 1771 } 1772 1773 if (optdatalen > MCLBYTES) 1774 return (EMSGSIZE); /* XXX */ 1775 if (optdatalen > MLEN) 1776 MCLGET(m, M_WAIT); 1777 m->m_len = optdatalen; 1778 if (optdatalen) 1779 bcopy(optdata, mtod(m, void *), optdatalen); 1780 1781 return (error); 1782 } 1783 1784 void 1785 ip6_clearpktopts(struct ip6_pktopts *pktopt, int optname) 1786 { 1787 if (optname == -1 || optname == IPV6_PKTINFO) { 1788 if (pktopt->ip6po_pktinfo) 1789 free(pktopt->ip6po_pktinfo, M_IP6OPT, 0); 1790 pktopt->ip6po_pktinfo = NULL; 1791 } 1792 if (optname == -1 || optname == IPV6_HOPLIMIT) 1793 pktopt->ip6po_hlim = -1; 1794 if (optname == -1 || optname == IPV6_TCLASS) 1795 pktopt->ip6po_tclass = -1; 1796 if (optname == -1 || optname == IPV6_HOPOPTS) { 1797 if (pktopt->ip6po_hbh) 1798 free(pktopt->ip6po_hbh, M_IP6OPT, 0); 1799 pktopt->ip6po_hbh = NULL; 1800 } 1801 if (optname == -1 || optname == IPV6_RTHDRDSTOPTS) { 1802 if (pktopt->ip6po_dest1) 1803 free(pktopt->ip6po_dest1, M_IP6OPT, 0); 1804 pktopt->ip6po_dest1 = NULL; 1805 } 1806 if (optname == -1 || optname == IPV6_RTHDR) { 1807 if (pktopt->ip6po_rhinfo.ip6po_rhi_rthdr) 1808 free(pktopt->ip6po_rhinfo.ip6po_rhi_rthdr, M_IP6OPT, 0); 1809 pktopt->ip6po_rhinfo.ip6po_rhi_rthdr = NULL; 1810 if (pktopt->ip6po_route.ro_rt) { 1811 rtfree(pktopt->ip6po_route.ro_rt); 1812 pktopt->ip6po_route.ro_rt = NULL; 1813 } 1814 } 1815 if (optname == -1 || optname == IPV6_DSTOPTS) { 1816 if (pktopt->ip6po_dest2) 1817 free(pktopt->ip6po_dest2, M_IP6OPT, 0); 1818 pktopt->ip6po_dest2 = NULL; 1819 } 1820 } 1821 1822 #define PKTOPT_EXTHDRCPY(type) \ 1823 do {\ 1824 if (src->type) {\ 1825 size_t hlen;\ 1826 hlen = (((struct ip6_ext *)src->type)->ip6e_len + 1) << 3;\ 1827 dst->type = malloc(hlen, M_IP6OPT, M_NOWAIT);\ 1828 if (dst->type == NULL)\ 1829 goto bad;\ 1830 memcpy(dst->type, src->type, hlen);\ 1831 }\ 1832 } while (/*CONSTCOND*/ 0) 1833 1834 int 1835 copypktopts(struct ip6_pktopts *dst, struct ip6_pktopts *src) 1836 { 1837 dst->ip6po_hlim = src->ip6po_hlim; 1838 dst->ip6po_tclass = src->ip6po_tclass; 1839 dst->ip6po_flags = src->ip6po_flags; 1840 if (src->ip6po_pktinfo) { 1841 dst->ip6po_pktinfo = malloc(sizeof(*dst->ip6po_pktinfo), 1842 M_IP6OPT, M_NOWAIT); 1843 if (dst->ip6po_pktinfo == NULL) 1844 goto bad; 1845 *dst->ip6po_pktinfo = *src->ip6po_pktinfo; 1846 } 1847 PKTOPT_EXTHDRCPY(ip6po_hbh); 1848 PKTOPT_EXTHDRCPY(ip6po_dest1); 1849 PKTOPT_EXTHDRCPY(ip6po_dest2); 1850 PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */ 1851 return (0); 1852 1853 bad: 1854 ip6_clearpktopts(dst, -1); 1855 return (ENOBUFS); 1856 } 1857 #undef PKTOPT_EXTHDRCPY 1858 1859 void 1860 ip6_freepcbopts(struct ip6_pktopts *pktopt) 1861 { 1862 if (pktopt == NULL) 1863 return; 1864 1865 ip6_clearpktopts(pktopt, -1); 1866 1867 free(pktopt, M_IP6OPT, 0); 1868 } 1869 1870 /* 1871 * Set the IP6 multicast options in response to user setsockopt(). 1872 */ 1873 int 1874 ip6_setmoptions(int optname, struct ip6_moptions **im6op, struct mbuf *m, 1875 unsigned int rtableid) 1876 { 1877 int error = 0; 1878 u_int loop, ifindex; 1879 struct ipv6_mreq *mreq; 1880 struct ifnet *ifp; 1881 struct ip6_moptions *im6o = *im6op; 1882 struct in6_multi_mship *imm; 1883 struct proc *p = curproc; /* XXX */ 1884 1885 if (im6o == NULL) { 1886 /* 1887 * No multicast option buffer attached to the pcb; 1888 * allocate one and initialize to default values. 1889 */ 1890 im6o = malloc(sizeof(*im6o), M_IPMOPTS, M_WAITOK); 1891 if (im6o == NULL) 1892 return (ENOBUFS); 1893 *im6op = im6o; 1894 im6o->im6o_ifidx = 0; 1895 im6o->im6o_hlim = ip6_defmcasthlim; 1896 im6o->im6o_loop = IPV6_DEFAULT_MULTICAST_LOOP; 1897 LIST_INIT(&im6o->im6o_memberships); 1898 } 1899 1900 switch (optname) { 1901 1902 case IPV6_MULTICAST_IF: 1903 /* 1904 * Select the interface for outgoing multicast packets. 1905 */ 1906 if (m == NULL || m->m_len != sizeof(u_int)) { 1907 error = EINVAL; 1908 break; 1909 } 1910 memcpy(&ifindex, mtod(m, u_int *), sizeof(ifindex)); 1911 if (ifindex != 0) { 1912 ifp = if_get(ifindex); 1913 if (ifp == NULL) { 1914 error = ENXIO; /* XXX EINVAL? */ 1915 break; 1916 } 1917 if (ifp->if_rdomain != rtable_l2(rtableid) || 1918 (ifp->if_flags & IFF_MULTICAST) == 0) { 1919 error = EADDRNOTAVAIL; 1920 if_put(ifp); 1921 break; 1922 } 1923 if_put(ifp); 1924 } 1925 im6o->im6o_ifidx = ifindex; 1926 break; 1927 1928 case IPV6_MULTICAST_HOPS: 1929 { 1930 /* 1931 * Set the IP6 hoplimit for outgoing multicast packets. 1932 */ 1933 int optval; 1934 if (m == NULL || m->m_len != sizeof(int)) { 1935 error = EINVAL; 1936 break; 1937 } 1938 memcpy(&optval, mtod(m, u_int *), sizeof(optval)); 1939 if (optval < -1 || optval >= 256) 1940 error = EINVAL; 1941 else if (optval == -1) 1942 im6o->im6o_hlim = ip6_defmcasthlim; 1943 else 1944 im6o->im6o_hlim = optval; 1945 break; 1946 } 1947 1948 case IPV6_MULTICAST_LOOP: 1949 /* 1950 * Set the loopback flag for outgoing multicast packets. 1951 * Must be zero or one. 1952 */ 1953 if (m == NULL || m->m_len != sizeof(u_int)) { 1954 error = EINVAL; 1955 break; 1956 } 1957 memcpy(&loop, mtod(m, u_int *), sizeof(loop)); 1958 if (loop > 1) { 1959 error = EINVAL; 1960 break; 1961 } 1962 im6o->im6o_loop = loop; 1963 break; 1964 1965 case IPV6_JOIN_GROUP: 1966 /* 1967 * Add a multicast group membership. 1968 * Group must be a valid IP6 multicast address. 1969 */ 1970 if (m == NULL || m->m_len != sizeof(struct ipv6_mreq)) { 1971 error = EINVAL; 1972 break; 1973 } 1974 mreq = mtod(m, struct ipv6_mreq *); 1975 if (IN6_IS_ADDR_UNSPECIFIED(&mreq->ipv6mr_multiaddr)) { 1976 /* 1977 * We use the unspecified address to specify to accept 1978 * all multicast addresses. Only super user is allowed 1979 * to do this. 1980 */ 1981 if (suser(p)) 1982 { 1983 error = EACCES; 1984 break; 1985 } 1986 } else if (!IN6_IS_ADDR_MULTICAST(&mreq->ipv6mr_multiaddr)) { 1987 error = EINVAL; 1988 break; 1989 } 1990 1991 /* 1992 * If no interface was explicitly specified, choose an 1993 * appropriate one according to the given multicast address. 1994 */ 1995 if (mreq->ipv6mr_interface == 0) { 1996 struct rtentry *rt; 1997 struct sockaddr_in6 dst; 1998 1999 memset(&dst, 0, sizeof(dst)); 2000 dst.sin6_len = sizeof(dst); 2001 dst.sin6_family = AF_INET6; 2002 dst.sin6_addr = mreq->ipv6mr_multiaddr; 2003 rt = rtalloc(sin6tosa(&dst), RT_RESOLVE, rtableid); 2004 if (rt == NULL) { 2005 error = EADDRNOTAVAIL; 2006 break; 2007 } 2008 ifp = if_get(rt->rt_ifidx); 2009 rtfree(rt); 2010 } else { 2011 /* 2012 * If the interface is specified, validate it. 2013 */ 2014 ifp = if_get(mreq->ipv6mr_interface); 2015 if (ifp == NULL) { 2016 error = ENXIO; /* XXX EINVAL? */ 2017 break; 2018 } 2019 } 2020 2021 /* 2022 * See if we found an interface, and confirm that it 2023 * supports multicast 2024 */ 2025 if (ifp == NULL || ifp->if_rdomain != rtable_l2(rtableid) || 2026 (ifp->if_flags & IFF_MULTICAST) == 0) { 2027 if_put(ifp); 2028 error = EADDRNOTAVAIL; 2029 break; 2030 } 2031 /* 2032 * Put interface index into the multicast address, 2033 * if the address has link/interface-local scope. 2034 */ 2035 if (IN6_IS_SCOPE_EMBED(&mreq->ipv6mr_multiaddr)) { 2036 mreq->ipv6mr_multiaddr.s6_addr16[1] = 2037 htons(ifp->if_index); 2038 } 2039 /* 2040 * See if the membership already exists. 2041 */ 2042 LIST_FOREACH(imm, &im6o->im6o_memberships, i6mm_chain) 2043 if (imm->i6mm_maddr->in6m_ifidx == ifp->if_index && 2044 IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr, 2045 &mreq->ipv6mr_multiaddr)) 2046 break; 2047 if (imm != NULL) { 2048 if_put(ifp); 2049 error = EADDRINUSE; 2050 break; 2051 } 2052 /* 2053 * Everything looks good; add a new record to the multicast 2054 * address list for the given interface. 2055 */ 2056 imm = in6_joingroup(ifp, &mreq->ipv6mr_multiaddr, &error); 2057 if_put(ifp); 2058 if (!imm) 2059 break; 2060 LIST_INSERT_HEAD(&im6o->im6o_memberships, imm, i6mm_chain); 2061 break; 2062 2063 case IPV6_LEAVE_GROUP: 2064 /* 2065 * Drop a multicast group membership. 2066 * Group must be a valid IP6 multicast address. 2067 */ 2068 if (m == NULL || m->m_len != sizeof(struct ipv6_mreq)) { 2069 error = EINVAL; 2070 break; 2071 } 2072 mreq = mtod(m, struct ipv6_mreq *); 2073 if (IN6_IS_ADDR_UNSPECIFIED(&mreq->ipv6mr_multiaddr)) { 2074 if (suser(p)) { 2075 error = EACCES; 2076 break; 2077 } 2078 } else if (!IN6_IS_ADDR_MULTICAST(&mreq->ipv6mr_multiaddr)) { 2079 error = EINVAL; 2080 break; 2081 } 2082 2083 /* 2084 * Put interface index into the multicast address, 2085 * if the address has link-local scope. 2086 */ 2087 if (IN6_IS_ADDR_MC_LINKLOCAL(&mreq->ipv6mr_multiaddr)) { 2088 mreq->ipv6mr_multiaddr.s6_addr16[1] = 2089 htons(mreq->ipv6mr_interface); 2090 } 2091 2092 /* 2093 * If an interface address was specified, get a pointer 2094 * to its ifnet structure. 2095 */ 2096 if (mreq->ipv6mr_interface == 0) 2097 ifp = NULL; 2098 else { 2099 ifp = if_get(mreq->ipv6mr_interface); 2100 if (ifp == NULL) { 2101 error = ENXIO; /* XXX EINVAL? */ 2102 break; 2103 } 2104 } 2105 2106 /* 2107 * Find the membership in the membership list. 2108 */ 2109 LIST_FOREACH(imm, &im6o->im6o_memberships, i6mm_chain) { 2110 if ((ifp == NULL || 2111 imm->i6mm_maddr->in6m_ifidx == ifp->if_index) && 2112 IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr, 2113 &mreq->ipv6mr_multiaddr)) 2114 break; 2115 } 2116 2117 if_put(ifp); 2118 2119 if (imm == NULL) { 2120 /* Unable to resolve interface */ 2121 error = EADDRNOTAVAIL; 2122 break; 2123 } 2124 /* 2125 * Give up the multicast address record to which the 2126 * membership points. 2127 */ 2128 LIST_REMOVE(imm, i6mm_chain); 2129 in6_leavegroup(imm); 2130 break; 2131 2132 default: 2133 error = EOPNOTSUPP; 2134 break; 2135 } 2136 2137 /* 2138 * If all options have default values, no need to keep the option 2139 * structure. 2140 */ 2141 if (im6o->im6o_ifidx == 0 && 2142 im6o->im6o_hlim == ip6_defmcasthlim && 2143 im6o->im6o_loop == IPV6_DEFAULT_MULTICAST_LOOP && 2144 LIST_EMPTY(&im6o->im6o_memberships)) { 2145 free(*im6op, M_IPMOPTS, sizeof(**im6op)); 2146 *im6op = NULL; 2147 } 2148 2149 return (error); 2150 } 2151 2152 /* 2153 * Return the IP6 multicast options in response to user getsockopt(). 2154 */ 2155 int 2156 ip6_getmoptions(int optname, struct ip6_moptions *im6o, struct mbuf *m) 2157 { 2158 u_int *hlim, *loop, *ifindex; 2159 2160 switch (optname) { 2161 case IPV6_MULTICAST_IF: 2162 ifindex = mtod(m, u_int *); 2163 m->m_len = sizeof(u_int); 2164 if (im6o == NULL || im6o->im6o_ifidx == 0) 2165 *ifindex = 0; 2166 else 2167 *ifindex = im6o->im6o_ifidx; 2168 return (0); 2169 2170 case IPV6_MULTICAST_HOPS: 2171 hlim = mtod(m, u_int *); 2172 m->m_len = sizeof(u_int); 2173 if (im6o == NULL) 2174 *hlim = ip6_defmcasthlim; 2175 else 2176 *hlim = im6o->im6o_hlim; 2177 return (0); 2178 2179 case IPV6_MULTICAST_LOOP: 2180 loop = mtod(m, u_int *); 2181 m->m_len = sizeof(u_int); 2182 if (im6o == NULL) 2183 *loop = ip6_defmcasthlim; 2184 else 2185 *loop = im6o->im6o_loop; 2186 return (0); 2187 2188 default: 2189 return (EOPNOTSUPP); 2190 } 2191 } 2192 2193 /* 2194 * Discard the IP6 multicast options. 2195 */ 2196 void 2197 ip6_freemoptions(struct ip6_moptions *im6o) 2198 { 2199 struct in6_multi_mship *imm; 2200 2201 if (im6o == NULL) 2202 return; 2203 2204 while (!LIST_EMPTY(&im6o->im6o_memberships)) { 2205 imm = LIST_FIRST(&im6o->im6o_memberships); 2206 LIST_REMOVE(imm, i6mm_chain); 2207 in6_leavegroup(imm); 2208 } 2209 free(im6o, M_IPMOPTS, sizeof(*im6o)); 2210 } 2211 2212 /* 2213 * Set IPv6 outgoing packet options based on advanced API. 2214 */ 2215 int 2216 ip6_setpktopts(struct mbuf *control, struct ip6_pktopts *opt, 2217 struct ip6_pktopts *stickyopt, int priv, int uproto) 2218 { 2219 u_int clen; 2220 struct cmsghdr *cm = 0; 2221 caddr_t cmsgs; 2222 int error; 2223 2224 if (control == NULL || opt == NULL) 2225 return (EINVAL); 2226 2227 ip6_initpktopts(opt); 2228 if (stickyopt) { 2229 int error; 2230 2231 /* 2232 * If stickyopt is provided, make a local copy of the options 2233 * for this particular packet, then override them by ancillary 2234 * objects. 2235 * XXX: copypktopts() does not copy the cached route to a next 2236 * hop (if any). This is not very good in terms of efficiency, 2237 * but we can allow this since this option should be rarely 2238 * used. 2239 */ 2240 if ((error = copypktopts(opt, stickyopt)) != 0) 2241 return (error); 2242 } 2243 2244 /* 2245 * XXX: Currently, we assume all the optional information is stored 2246 * in a single mbuf. 2247 */ 2248 if (control->m_next) 2249 return (EINVAL); 2250 2251 clen = control->m_len; 2252 cmsgs = mtod(control, caddr_t); 2253 do { 2254 if (clen < CMSG_LEN(0)) 2255 return (EINVAL); 2256 cm = (struct cmsghdr *)cmsgs; 2257 if (cm->cmsg_len < CMSG_LEN(0) || cm->cmsg_len > clen || 2258 CMSG_ALIGN(cm->cmsg_len) > clen) 2259 return (EINVAL); 2260 if (cm->cmsg_level == IPPROTO_IPV6) { 2261 error = ip6_setpktopt(cm->cmsg_type, CMSG_DATA(cm), 2262 cm->cmsg_len - CMSG_LEN(0), opt, priv, 0, uproto); 2263 if (error) 2264 return (error); 2265 } 2266 2267 clen -= CMSG_ALIGN(cm->cmsg_len); 2268 cmsgs += CMSG_ALIGN(cm->cmsg_len); 2269 } while (clen); 2270 2271 return (0); 2272 } 2273 2274 /* 2275 * Set a particular packet option, as a sticky option or an ancillary data 2276 * item. "len" can be 0 only when it's a sticky option. 2277 */ 2278 int 2279 ip6_setpktopt(int optname, u_char *buf, int len, struct ip6_pktopts *opt, 2280 int priv, int sticky, int uproto) 2281 { 2282 int minmtupolicy; 2283 2284 switch (optname) { 2285 case IPV6_PKTINFO: 2286 { 2287 struct ifnet *ifp = NULL; 2288 struct in6_pktinfo *pktinfo; 2289 2290 if (len != sizeof(struct in6_pktinfo)) 2291 return (EINVAL); 2292 2293 pktinfo = (struct in6_pktinfo *)buf; 2294 2295 /* 2296 * An application can clear any sticky IPV6_PKTINFO option by 2297 * doing a "regular" setsockopt with ipi6_addr being 2298 * in6addr_any and ipi6_ifindex being zero. 2299 * [RFC 3542, Section 6] 2300 */ 2301 if (opt->ip6po_pktinfo && 2302 pktinfo->ipi6_ifindex == 0 && 2303 IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) { 2304 ip6_clearpktopts(opt, optname); 2305 break; 2306 } 2307 2308 if (uproto == IPPROTO_TCP && 2309 sticky && !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) { 2310 return (EINVAL); 2311 } 2312 2313 if (pktinfo->ipi6_ifindex) { 2314 ifp = if_get(pktinfo->ipi6_ifindex); 2315 if (ifp == NULL) 2316 return (ENXIO); 2317 if_put(ifp); 2318 } 2319 2320 /* 2321 * We store the address anyway, and let in6_selectsrc() 2322 * validate the specified address. This is because ipi6_addr 2323 * may not have enough information about its scope zone, and 2324 * we may need additional information (such as outgoing 2325 * interface or the scope zone of a destination address) to 2326 * disambiguate the scope. 2327 * XXX: the delay of the validation may confuse the 2328 * application when it is used as a sticky option. 2329 */ 2330 if (opt->ip6po_pktinfo == NULL) { 2331 opt->ip6po_pktinfo = malloc(sizeof(*pktinfo), 2332 M_IP6OPT, M_NOWAIT); 2333 if (opt->ip6po_pktinfo == NULL) 2334 return (ENOBUFS); 2335 } 2336 bcopy(pktinfo, opt->ip6po_pktinfo, sizeof(*pktinfo)); 2337 break; 2338 } 2339 2340 case IPV6_HOPLIMIT: 2341 { 2342 int *hlimp; 2343 2344 /* 2345 * RFC 3542 deprecated the usage of sticky IPV6_HOPLIMIT 2346 * to simplify the ordering among hoplimit options. 2347 */ 2348 if (sticky) 2349 return (ENOPROTOOPT); 2350 2351 if (len != sizeof(int)) 2352 return (EINVAL); 2353 hlimp = (int *)buf; 2354 if (*hlimp < -1 || *hlimp > 255) 2355 return (EINVAL); 2356 2357 opt->ip6po_hlim = *hlimp; 2358 break; 2359 } 2360 2361 case IPV6_TCLASS: 2362 { 2363 int tclass; 2364 2365 if (len != sizeof(int)) 2366 return (EINVAL); 2367 tclass = *(int *)buf; 2368 if (tclass < -1 || tclass > 255) 2369 return (EINVAL); 2370 2371 opt->ip6po_tclass = tclass; 2372 break; 2373 } 2374 case IPV6_HOPOPTS: 2375 { 2376 struct ip6_hbh *hbh; 2377 int hbhlen; 2378 2379 /* 2380 * XXX: We don't allow a non-privileged user to set ANY HbH 2381 * options, since per-option restriction has too much 2382 * overhead. 2383 */ 2384 if (!priv) 2385 return (EPERM); 2386 2387 if (len == 0) { 2388 ip6_clearpktopts(opt, IPV6_HOPOPTS); 2389 break; /* just remove the option */ 2390 } 2391 2392 /* message length validation */ 2393 if (len < sizeof(struct ip6_hbh)) 2394 return (EINVAL); 2395 hbh = (struct ip6_hbh *)buf; 2396 hbhlen = (hbh->ip6h_len + 1) << 3; 2397 if (len != hbhlen) 2398 return (EINVAL); 2399 2400 /* turn off the previous option, then set the new option. */ 2401 ip6_clearpktopts(opt, IPV6_HOPOPTS); 2402 opt->ip6po_hbh = malloc(hbhlen, M_IP6OPT, M_NOWAIT); 2403 if (opt->ip6po_hbh == NULL) 2404 return (ENOBUFS); 2405 memcpy(opt->ip6po_hbh, hbh, hbhlen); 2406 2407 break; 2408 } 2409 2410 case IPV6_DSTOPTS: 2411 case IPV6_RTHDRDSTOPTS: 2412 { 2413 struct ip6_dest *dest, **newdest = NULL; 2414 int destlen; 2415 2416 if (!priv) /* XXX: see the comment for IPV6_HOPOPTS */ 2417 return (EPERM); 2418 2419 if (len == 0) { 2420 ip6_clearpktopts(opt, optname); 2421 break; /* just remove the option */ 2422 } 2423 2424 /* message length validation */ 2425 if (len < sizeof(struct ip6_dest)) 2426 return (EINVAL); 2427 dest = (struct ip6_dest *)buf; 2428 destlen = (dest->ip6d_len + 1) << 3; 2429 if (len != destlen) 2430 return (EINVAL); 2431 /* 2432 * Determine the position that the destination options header 2433 * should be inserted; before or after the routing header. 2434 */ 2435 switch (optname) { 2436 case IPV6_RTHDRDSTOPTS: 2437 newdest = &opt->ip6po_dest1; 2438 break; 2439 case IPV6_DSTOPTS: 2440 newdest = &opt->ip6po_dest2; 2441 break; 2442 } 2443 2444 /* turn off the previous option, then set the new option. */ 2445 ip6_clearpktopts(opt, optname); 2446 *newdest = malloc(destlen, M_IP6OPT, M_NOWAIT); 2447 if (*newdest == NULL) 2448 return (ENOBUFS); 2449 memcpy(*newdest, dest, destlen); 2450 2451 break; 2452 } 2453 2454 case IPV6_RTHDR: 2455 { 2456 struct ip6_rthdr *rth; 2457 int rthlen; 2458 2459 if (len == 0) { 2460 ip6_clearpktopts(opt, IPV6_RTHDR); 2461 break; /* just remove the option */ 2462 } 2463 2464 /* message length validation */ 2465 if (len < sizeof(struct ip6_rthdr)) 2466 return (EINVAL); 2467 rth = (struct ip6_rthdr *)buf; 2468 rthlen = (rth->ip6r_len + 1) << 3; 2469 if (len != rthlen) 2470 return (EINVAL); 2471 2472 switch (rth->ip6r_type) { 2473 case IPV6_RTHDR_TYPE_0: 2474 if (rth->ip6r_len == 0) /* must contain one addr */ 2475 return (EINVAL); 2476 if (rth->ip6r_len % 2) /* length must be even */ 2477 return (EINVAL); 2478 if (rth->ip6r_len / 2 != rth->ip6r_segleft) 2479 return (EINVAL); 2480 break; 2481 default: 2482 return (EINVAL); /* not supported */ 2483 } 2484 /* turn off the previous option */ 2485 ip6_clearpktopts(opt, IPV6_RTHDR); 2486 opt->ip6po_rthdr = malloc(rthlen, M_IP6OPT, M_NOWAIT); 2487 if (opt->ip6po_rthdr == NULL) 2488 return (ENOBUFS); 2489 memcpy(opt->ip6po_rthdr, rth, rthlen); 2490 break; 2491 } 2492 2493 case IPV6_USE_MIN_MTU: 2494 if (len != sizeof(int)) 2495 return (EINVAL); 2496 minmtupolicy = *(int *)buf; 2497 if (minmtupolicy != IP6PO_MINMTU_MCASTONLY && 2498 minmtupolicy != IP6PO_MINMTU_DISABLE && 2499 minmtupolicy != IP6PO_MINMTU_ALL) { 2500 return (EINVAL); 2501 } 2502 opt->ip6po_minmtu = minmtupolicy; 2503 break; 2504 2505 case IPV6_DONTFRAG: 2506 if (len != sizeof(int)) 2507 return (EINVAL); 2508 2509 if (uproto == IPPROTO_TCP || *(int *)buf == 0) { 2510 /* 2511 * we ignore this option for TCP sockets. 2512 * (RFC3542 leaves this case unspecified.) 2513 */ 2514 opt->ip6po_flags &= ~IP6PO_DONTFRAG; 2515 } else 2516 opt->ip6po_flags |= IP6PO_DONTFRAG; 2517 break; 2518 2519 default: 2520 return (ENOPROTOOPT); 2521 } /* end of switch */ 2522 2523 return (0); 2524 } 2525 2526 /* 2527 * Routine called from ip6_output() to loop back a copy of an IP6 multicast 2528 * packet to the input queue of a specified interface. 2529 */ 2530 void 2531 ip6_mloopback(struct ifnet *ifp, struct mbuf *m, struct sockaddr_in6 *dst) 2532 { 2533 struct mbuf *copym; 2534 struct ip6_hdr *ip6; 2535 2536 /* 2537 * Duplicate the packet. 2538 */ 2539 copym = m_copym(m, 0, M_COPYALL, M_NOWAIT); 2540 if (copym == NULL) 2541 return; 2542 2543 /* 2544 * Make sure to deep-copy IPv6 header portion in case the data 2545 * is in an mbuf cluster, so that we can safely override the IPv6 2546 * header portion later. 2547 */ 2548 if ((copym->m_flags & M_EXT) != 0 || 2549 copym->m_len < sizeof(struct ip6_hdr)) { 2550 copym = m_pullup(copym, sizeof(struct ip6_hdr)); 2551 if (copym == NULL) 2552 return; 2553 } 2554 2555 #ifdef DIAGNOSTIC 2556 if (copym->m_len < sizeof(*ip6)) { 2557 m_freem(copym); 2558 return; 2559 } 2560 #endif 2561 2562 ip6 = mtod(copym, struct ip6_hdr *); 2563 if (IN6_IS_SCOPE_EMBED(&ip6->ip6_src)) 2564 ip6->ip6_src.s6_addr16[1] = 0; 2565 if (IN6_IS_SCOPE_EMBED(&ip6->ip6_dst)) 2566 ip6->ip6_dst.s6_addr16[1] = 0; 2567 2568 if_input_local(ifp, copym, dst->sin6_family); 2569 } 2570 2571 /* 2572 * Chop IPv6 header off from the payload. 2573 */ 2574 int 2575 ip6_splithdr(struct mbuf *m, struct ip6_exthdrs *exthdrs) 2576 { 2577 struct mbuf *mh; 2578 struct ip6_hdr *ip6; 2579 2580 ip6 = mtod(m, struct ip6_hdr *); 2581 if (m->m_len > sizeof(*ip6)) { 2582 MGET(mh, M_DONTWAIT, MT_HEADER); 2583 if (mh == NULL) { 2584 m_freem(m); 2585 return ENOBUFS; 2586 } 2587 M_MOVE_PKTHDR(mh, m); 2588 m_align(mh, sizeof(*ip6)); 2589 m->m_len -= sizeof(*ip6); 2590 m->m_data += sizeof(*ip6); 2591 mh->m_next = m; 2592 m = mh; 2593 m->m_len = sizeof(*ip6); 2594 bcopy((caddr_t)ip6, mtod(m, caddr_t), sizeof(*ip6)); 2595 } 2596 exthdrs->ip6e_ip6 = m; 2597 return 0; 2598 } 2599 2600 u_int32_t 2601 ip6_randomid(void) 2602 { 2603 return idgen32(&ip6_id_ctx); 2604 } 2605 2606 void 2607 ip6_randomid_init(void) 2608 { 2609 idgen32_init(&ip6_id_ctx); 2610 } 2611 2612 /* 2613 * Compute significant parts of the IPv6 checksum pseudo-header 2614 * for use in a delayed TCP/UDP checksum calculation. 2615 */ 2616 static __inline u_int16_t __attribute__((__unused__)) 2617 in6_cksum_phdr(const struct in6_addr *src, const struct in6_addr *dst, 2618 u_int32_t len, u_int32_t nxt) 2619 { 2620 u_int32_t sum = 0; 2621 const u_int16_t *w; 2622 2623 w = (const u_int16_t *) src; 2624 sum += w[0]; 2625 if (!IN6_IS_SCOPE_EMBED(src)) 2626 sum += w[1]; 2627 sum += w[2]; sum += w[3]; sum += w[4]; sum += w[5]; 2628 sum += w[6]; sum += w[7]; 2629 2630 w = (const u_int16_t *) dst; 2631 sum += w[0]; 2632 if (!IN6_IS_SCOPE_EMBED(dst)) 2633 sum += w[1]; 2634 sum += w[2]; sum += w[3]; sum += w[4]; sum += w[5]; 2635 sum += w[6]; sum += w[7]; 2636 2637 sum += (u_int16_t)(len >> 16) + (u_int16_t)(len /*& 0xffff*/); 2638 2639 sum += (u_int16_t)(nxt >> 16) + (u_int16_t)(nxt /*& 0xffff*/); 2640 2641 sum = (u_int16_t)(sum >> 16) + (u_int16_t)(sum /*& 0xffff*/); 2642 2643 if (sum > 0xffff) 2644 sum -= 0xffff; 2645 2646 return (sum); 2647 } 2648 2649 /* 2650 * Process a delayed payload checksum calculation. 2651 */ 2652 void 2653 in6_delayed_cksum(struct mbuf *m, u_int8_t nxt) 2654 { 2655 int nxtp, offset; 2656 u_int16_t csum; 2657 2658 offset = ip6_lasthdr(m, 0, IPPROTO_IPV6, &nxtp); 2659 if (offset <= 0 || nxtp != nxt) 2660 /* If the desired next protocol isn't found, punt. */ 2661 return; 2662 csum = (u_int16_t)(in6_cksum(m, 0, offset, m->m_pkthdr.len - offset)); 2663 2664 switch (nxt) { 2665 case IPPROTO_TCP: 2666 offset += offsetof(struct tcphdr, th_sum); 2667 break; 2668 2669 case IPPROTO_UDP: 2670 offset += offsetof(struct udphdr, uh_sum); 2671 if (csum == 0) 2672 csum = 0xffff; 2673 break; 2674 2675 case IPPROTO_ICMPV6: 2676 offset += offsetof(struct icmp6_hdr, icmp6_cksum); 2677 break; 2678 } 2679 2680 if ((offset + sizeof(u_int16_t)) > m->m_len) 2681 m_copyback(m, offset, sizeof(csum), &csum, M_NOWAIT); 2682 else 2683 *(u_int16_t *)(mtod(m, caddr_t) + offset) = csum; 2684 } 2685 2686 void 2687 in6_proto_cksum_out(struct mbuf *m, struct ifnet *ifp) 2688 { 2689 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *); 2690 2691 /* some hw and in6_delayed_cksum need the pseudo header cksum */ 2692 if (m->m_pkthdr.csum_flags & 2693 (M_TCP_CSUM_OUT|M_UDP_CSUM_OUT|M_ICMP_CSUM_OUT)) { 2694 int nxt, offset; 2695 u_int16_t csum; 2696 2697 offset = ip6_lasthdr(m, 0, IPPROTO_IPV6, &nxt); 2698 if (ISSET(m->m_pkthdr.csum_flags, M_TCP_TSO) && 2699 in_ifcap_cksum(m, ifp, IFCAP_TSOv6)) { 2700 csum = in6_cksum_phdr(&ip6->ip6_src, &ip6->ip6_dst, 2701 htonl(0), htonl(nxt)); 2702 } else { 2703 csum = in6_cksum_phdr(&ip6->ip6_src, &ip6->ip6_dst, 2704 htonl(m->m_pkthdr.len - offset), htonl(nxt)); 2705 } 2706 if (nxt == IPPROTO_TCP) 2707 offset += offsetof(struct tcphdr, th_sum); 2708 else if (nxt == IPPROTO_UDP) 2709 offset += offsetof(struct udphdr, uh_sum); 2710 else if (nxt == IPPROTO_ICMPV6) 2711 offset += offsetof(struct icmp6_hdr, icmp6_cksum); 2712 if ((offset + sizeof(u_int16_t)) > m->m_len) 2713 m_copyback(m, offset, sizeof(csum), &csum, M_NOWAIT); 2714 else 2715 *(u_int16_t *)(mtod(m, caddr_t) + offset) = csum; 2716 } 2717 2718 if (m->m_pkthdr.csum_flags & M_TCP_CSUM_OUT) { 2719 if (!ifp || !(ifp->if_capabilities & IFCAP_CSUM_TCPv6) || 2720 ip6->ip6_nxt != IPPROTO_TCP || 2721 ifp->if_bridgeidx != 0) { 2722 tcpstat_inc(tcps_outswcsum); 2723 in6_delayed_cksum(m, IPPROTO_TCP); 2724 m->m_pkthdr.csum_flags &= ~M_TCP_CSUM_OUT; /* Clear */ 2725 } 2726 } else if (m->m_pkthdr.csum_flags & M_UDP_CSUM_OUT) { 2727 if (!ifp || !(ifp->if_capabilities & IFCAP_CSUM_UDPv6) || 2728 ip6->ip6_nxt != IPPROTO_UDP || 2729 ifp->if_bridgeidx != 0) { 2730 udpstat_inc(udps_outswcsum); 2731 in6_delayed_cksum(m, IPPROTO_UDP); 2732 m->m_pkthdr.csum_flags &= ~M_UDP_CSUM_OUT; /* Clear */ 2733 } 2734 } else if (m->m_pkthdr.csum_flags & M_ICMP_CSUM_OUT) { 2735 in6_delayed_cksum(m, IPPROTO_ICMPV6); 2736 m->m_pkthdr.csum_flags &= ~M_ICMP_CSUM_OUT; /* Clear */ 2737 } 2738 } 2739 2740 #ifdef IPSEC 2741 int 2742 ip6_output_ipsec_lookup(struct mbuf *m, const struct ipsec_level *seclevel, 2743 struct tdb **tdbout) 2744 { 2745 struct tdb *tdb; 2746 struct m_tag *mtag; 2747 struct tdb_ident *tdbi; 2748 int error; 2749 2750 /* 2751 * Check if there was an outgoing SA bound to the flow 2752 * from a transport protocol. 2753 */ 2754 2755 /* Do we have any pending SAs to apply ? */ 2756 error = ipsp_spd_lookup(m, AF_INET6, sizeof(struct ip6_hdr), 2757 IPSP_DIRECTION_OUT, NULL, seclevel, &tdb, NULL); 2758 if (error || tdb == NULL) { 2759 *tdbout = NULL; 2760 return error; 2761 } 2762 /* Loop detection */ 2763 for (mtag = m_tag_first(m); mtag != NULL; mtag = m_tag_next(m, mtag)) { 2764 if (mtag->m_tag_id != PACKET_TAG_IPSEC_OUT_DONE) 2765 continue; 2766 tdbi = (struct tdb_ident *)(mtag + 1); 2767 if (tdbi->spi == tdb->tdb_spi && 2768 tdbi->proto == tdb->tdb_sproto && 2769 tdbi->rdomain == tdb->tdb_rdomain && 2770 !memcmp(&tdbi->dst, &tdb->tdb_dst, 2771 sizeof(union sockaddr_union))) { 2772 /* no IPsec needed */ 2773 tdb_unref(tdb); 2774 *tdbout = NULL; 2775 return 0; 2776 } 2777 } 2778 *tdbout = tdb; 2779 return 0; 2780 } 2781 2782 int 2783 ip6_output_ipsec_pmtu_update(struct tdb *tdb, struct route *ro, 2784 struct in6_addr *dst, int ifidx, int rtableid, int transportmode) 2785 { 2786 struct rtentry *rt = NULL; 2787 int rt_mtucloned = 0; 2788 2789 /* Find a host route to store the mtu in */ 2790 if (ro != NULL) 2791 rt = ro->ro_rt; 2792 /* but don't add a PMTU route for transport mode SAs */ 2793 if (transportmode) 2794 rt = NULL; 2795 else if (rt == NULL || (rt->rt_flags & RTF_HOST) == 0) { 2796 struct sockaddr_in6 sin6; 2797 int error; 2798 2799 memset(&sin6, 0, sizeof(sin6)); 2800 sin6.sin6_family = AF_INET6; 2801 sin6.sin6_len = sizeof(sin6); 2802 sin6.sin6_addr = *dst; 2803 sin6.sin6_scope_id = in6_addr2scopeid(ifidx, dst); 2804 error = in6_embedscope(dst, &sin6, NULL, NULL); 2805 if (error) { 2806 /* should be impossible */ 2807 return error; 2808 } 2809 rt = icmp6_mtudisc_clone(&sin6, rtableid, 1); 2810 rt_mtucloned = 1; 2811 } 2812 DPRINTF("spi %08x mtu %d rt %p cloned %d", 2813 ntohl(tdb->tdb_spi), tdb->tdb_mtu, rt, rt_mtucloned); 2814 if (rt != NULL) { 2815 atomic_store_int(&rt->rt_mtu, tdb->tdb_mtu); 2816 if (ro != NULL && ro->ro_rt != NULL) { 2817 rtfree(ro->ro_rt); 2818 ro->ro_rt = rtalloc(&ro->ro_dstsa, RT_RESOLVE, 2819 rtableid); 2820 } 2821 if (rt_mtucloned) 2822 rtfree(rt); 2823 } 2824 return 0; 2825 } 2826 2827 int 2828 ip6_output_ipsec_send(struct tdb *tdb, struct mbuf *m, struct route *ro, 2829 int tunalready, int fwd) 2830 { 2831 struct mbuf_list ml; 2832 struct ifnet *encif = NULL; 2833 struct ip6_hdr *ip6; 2834 struct in6_addr dst; 2835 u_int len; 2836 int error, ifidx, rtableid, tso = 0; 2837 2838 #if NPF > 0 2839 /* 2840 * Packet filter 2841 */ 2842 if ((encif = enc_getif(tdb->tdb_rdomain, tdb->tdb_tap)) == NULL || 2843 pf_test(AF_INET6, fwd ? PF_FWD : PF_OUT, encif, &m) != PF_PASS) { 2844 m_freem(m); 2845 return EACCES; 2846 } 2847 if (m == NULL) 2848 return 0; 2849 /* 2850 * PF_TAG_REROUTE handling or not... 2851 * Packet is entering IPsec so the routing is 2852 * already overruled by the IPsec policy. 2853 * Until now the change was not reconsidered. 2854 * What's the behaviour? 2855 */ 2856 #endif 2857 2858 /* Check if we can chop the TCP packet */ 2859 ip6 = mtod(m, struct ip6_hdr *); 2860 if (ISSET(m->m_pkthdr.csum_flags, M_TCP_TSO) && 2861 m->m_pkthdr.ph_mss <= tdb->tdb_mtu) { 2862 tso = 1; 2863 len = m->m_pkthdr.ph_mss; 2864 } else 2865 len = sizeof(struct ip6_hdr) + ntohs(ip6->ip6_plen); 2866 2867 /* Check if we are allowed to fragment */ 2868 dst = ip6->ip6_dst; 2869 ifidx = m->m_pkthdr.ph_ifidx; 2870 rtableid = m->m_pkthdr.ph_rtableid; 2871 if (ip_mtudisc && tdb->tdb_mtu && 2872 len > tdb->tdb_mtu && tdb->tdb_mtutimeout > gettime()) { 2873 int transportmode; 2874 2875 transportmode = (tdb->tdb_dst.sa.sa_family == AF_INET6) && 2876 (IN6_ARE_ADDR_EQUAL(&tdb->tdb_dst.sin6.sin6_addr, &dst)); 2877 error = ip6_output_ipsec_pmtu_update(tdb, ro, &dst, ifidx, 2878 rtableid, transportmode); 2879 if (error) { 2880 ipsecstat_inc(ipsec_odrops); 2881 tdbstat_inc(tdb, tdb_odrops); 2882 m_freem(m); 2883 return error; 2884 } 2885 ipsec_adjust_mtu(m, tdb->tdb_mtu); 2886 m_freem(m); 2887 return EMSGSIZE; 2888 } 2889 /* propagate don't fragment for v6-over-v6 */ 2890 if (ip_mtudisc) 2891 SET(m->m_pkthdr.csum_flags, M_IPV6_DF_OUT); 2892 2893 /* 2894 * Clear these -- they'll be set in the recursive invocation 2895 * as needed. 2896 */ 2897 m->m_flags &= ~(M_BCAST | M_MCAST); 2898 2899 if (tso) { 2900 error = tcp_chopper(m, &ml, encif, len); 2901 if (error) 2902 goto done; 2903 } else { 2904 CLR(m->m_pkthdr.csum_flags, M_TCP_TSO); 2905 in6_proto_cksum_out(m, encif); 2906 ml_init(&ml); 2907 ml_enqueue(&ml, m); 2908 } 2909 2910 KERNEL_LOCK(); 2911 while ((m = ml_dequeue(&ml)) != NULL) { 2912 /* Callee frees mbuf */ 2913 error = ipsp_process_packet(m, tdb, AF_INET6, tunalready); 2914 if (error) 2915 break; 2916 } 2917 KERNEL_UNLOCK(); 2918 done: 2919 if (error) { 2920 ml_purge(&ml); 2921 ipsecstat_inc(ipsec_odrops); 2922 tdbstat_inc(tdb, tdb_odrops); 2923 } 2924 if (!error && tso) 2925 tcpstat_inc(tcps_outswtso); 2926 if (ip_mtudisc && error == EMSGSIZE) 2927 ip6_output_ipsec_pmtu_update(tdb, ro, &dst, ifidx, rtableid, 0); 2928 return error; 2929 } 2930 #endif /* IPSEC */ 2931