ip6_output.c revision 1.204
1 /* $NetBSD: ip6_output.c,v 1.204 2018/04/18 07:17:49 maxv Exp $ */ 2 /* $KAME: ip6_output.c,v 1.172 2001/03/25 09:55:56 itojun Exp $ */ 3 4 /* 5 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project. 6 * All rights reserved. 7 * 8 * Redistribution and use in source and binary forms, with or without 9 * modification, are permitted provided that the following conditions 10 * are met: 11 * 1. Redistributions of source code must retain the above copyright 12 * notice, this list of conditions and the following disclaimer. 13 * 2. Redistributions in binary form must reproduce the above copyright 14 * notice, this list of conditions and the following disclaimer in the 15 * documentation and/or other materials provided with the distribution. 16 * 3. Neither the name of the project nor the names of its contributors 17 * may be used to endorse or promote products derived from this software 18 * without specific prior written permission. 19 * 20 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND 21 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 22 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 23 * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE 24 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 25 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 26 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 27 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 28 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 29 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 30 * SUCH DAMAGE. 31 */ 32 33 /* 34 * Copyright (c) 1982, 1986, 1988, 1990, 1993 35 * The Regents of the University of California. All rights reserved. 36 * 37 * Redistribution and use in source and binary forms, with or without 38 * modification, are permitted provided that the following conditions 39 * are met: 40 * 1. Redistributions of source code must retain the above copyright 41 * notice, this list of conditions and the following disclaimer. 42 * 2. Redistributions in binary form must reproduce the above copyright 43 * notice, this list of conditions and the following disclaimer in the 44 * documentation and/or other materials provided with the distribution. 45 * 3. Neither the name of the University nor the names of its contributors 46 * may be used to endorse or promote products derived from this software 47 * without specific prior written permission. 48 * 49 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 50 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 51 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 52 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 53 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 54 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 55 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 56 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 57 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 58 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 59 * SUCH DAMAGE. 60 * 61 * @(#)ip_output.c 8.3 (Berkeley) 1/21/94 62 */ 63 64 #include <sys/cdefs.h> 65 __KERNEL_RCSID(0, "$NetBSD: ip6_output.c,v 1.204 2018/04/18 07:17:49 maxv Exp $"); 66 67 #ifdef _KERNEL_OPT 68 #include "opt_inet.h" 69 #include "opt_inet6.h" 70 #include "opt_ipsec.h" 71 #endif 72 73 #include <sys/param.h> 74 #include <sys/malloc.h> 75 #include <sys/mbuf.h> 76 #include <sys/errno.h> 77 #include <sys/socket.h> 78 #include <sys/socketvar.h> 79 #include <sys/syslog.h> 80 #include <sys/systm.h> 81 #include <sys/proc.h> 82 #include <sys/kauth.h> 83 84 #include <net/if.h> 85 #include <net/route.h> 86 #include <net/pfil.h> 87 88 #include <netinet/in.h> 89 #include <netinet/in_var.h> 90 #include <netinet/ip6.h> 91 #include <netinet/ip_var.h> 92 #include <netinet/icmp6.h> 93 #include <netinet/in_offload.h> 94 #include <netinet/portalgo.h> 95 #include <netinet6/in6_offload.h> 96 #include <netinet6/ip6_var.h> 97 #include <netinet6/ip6_private.h> 98 #include <netinet6/in6_pcb.h> 99 #include <netinet6/nd6.h> 100 #include <netinet6/ip6protosw.h> 101 #include <netinet6/scope6_var.h> 102 103 #ifdef IPSEC 104 #include <netipsec/ipsec.h> 105 #include <netipsec/ipsec6.h> 106 #include <netipsec/key.h> 107 #endif 108 109 110 #include <net/net_osdep.h> 111 112 extern pfil_head_t *inet6_pfil_hook; /* XXX */ 113 114 struct ip6_exthdrs { 115 struct mbuf *ip6e_ip6; 116 struct mbuf *ip6e_hbh; 117 struct mbuf *ip6e_dest1; 118 struct mbuf *ip6e_rthdr; 119 struct mbuf *ip6e_dest2; 120 }; 121 122 static int ip6_pcbopt(int, u_char *, int, struct ip6_pktopts **, 123 kauth_cred_t, int); 124 static int ip6_getpcbopt(struct ip6_pktopts *, int, struct sockopt *); 125 static int ip6_setpktopt(int, u_char *, int, struct ip6_pktopts *, kauth_cred_t, 126 int, int, int); 127 static int ip6_setmoptions(const struct sockopt *, struct in6pcb *); 128 static int ip6_getmoptions(struct sockopt *, struct in6pcb *); 129 static int ip6_copyexthdr(struct mbuf **, void *, int); 130 static int ip6_insertfraghdr(struct mbuf *, struct mbuf *, int, 131 struct ip6_frag **); 132 static int ip6_insert_jumboopt(struct ip6_exthdrs *, u_int32_t); 133 static int ip6_splithdr(struct mbuf *, struct ip6_exthdrs *); 134 static int ip6_getpmtu(struct rtentry *, struct ifnet *, u_long *, int *); 135 static int copypktopts(struct ip6_pktopts *, struct ip6_pktopts *, int); 136 static int ip6_ifaddrvalid(const struct in6_addr *, const struct in6_addr *); 137 static int ip6_handle_rthdr(struct ip6_rthdr *, struct ip6_hdr *); 138 139 #ifdef RFC2292 140 static int ip6_pcbopts(struct ip6_pktopts **, struct socket *, struct sockopt *); 141 #endif 142 143 static int 144 ip6_handle_rthdr(struct ip6_rthdr *rh, struct ip6_hdr *ip6) 145 { 146 struct ip6_rthdr0 *rh0; 147 struct in6_addr *addr; 148 struct sockaddr_in6 sa; 149 int error = 0; 150 151 switch (rh->ip6r_type) { 152 case IPV6_RTHDR_TYPE_0: 153 rh0 = (struct ip6_rthdr0 *)rh; 154 addr = (struct in6_addr *)(rh0 + 1); 155 156 /* 157 * construct a sockaddr_in6 form of the first hop. 158 * 159 * XXX we may not have enough information about its scope zone; 160 * there is no standard API to pass the information from the 161 * application. 162 */ 163 sockaddr_in6_init(&sa, addr, 0, 0, 0); 164 error = sa6_embedscope(&sa, ip6_use_defzone); 165 if (error != 0) 166 break; 167 memmove(&addr[0], &addr[1], 168 sizeof(struct in6_addr) * (rh0->ip6r0_segleft - 1)); 169 addr[rh0->ip6r0_segleft - 1] = ip6->ip6_dst; 170 ip6->ip6_dst = sa.sin6_addr; 171 /* XXX */ 172 in6_clearscope(addr + rh0->ip6r0_segleft - 1); 173 break; 174 default: /* is it possible? */ 175 error = EINVAL; 176 } 177 178 return error; 179 } 180 181 /* 182 * Send an IP packet to a host. 183 */ 184 int 185 ip6_if_output(struct ifnet * const ifp, struct ifnet * const origifp, 186 struct mbuf * const m, const struct sockaddr_in6 * const dst, 187 const struct rtentry *rt) 188 { 189 int error = 0; 190 191 if (rt != NULL) { 192 error = rt_check_reject_route(rt, ifp); 193 if (error != 0) { 194 m_freem(m); 195 return error; 196 } 197 } 198 199 if ((ifp->if_flags & IFF_LOOPBACK) != 0) 200 error = if_output_lock(ifp, origifp, m, sin6tocsa(dst), rt); 201 else 202 error = if_output_lock(ifp, ifp, m, sin6tocsa(dst), rt); 203 return error; 204 } 205 206 /* 207 * IP6 output. The packet in mbuf chain m contains a skeletal IP6 208 * header (with pri, len, nxt, hlim, src, dst). 209 * 210 * This function may modify ver and hlim only. The mbuf chain containing the 211 * packet will be freed. The mbuf opt, if present, will not be freed. 212 * 213 * Type of "mtu": rt_rmx.rmx_mtu is u_long, ifnet.ifr_mtu is int, and 214 * nd_ifinfo.linkmtu is u_int32_t. So we use u_long to hold largest one, 215 * which is rt_rmx.rmx_mtu. 216 */ 217 int 218 ip6_output( 219 struct mbuf *m0, 220 struct ip6_pktopts *opt, 221 struct route *ro, 222 int flags, 223 struct ip6_moptions *im6o, 224 struct in6pcb *in6p, 225 struct ifnet **ifpp /* XXX: just for statistics */ 226 ) 227 { 228 struct ip6_hdr *ip6, *mhip6; 229 struct ifnet *ifp = NULL, *origifp = NULL; 230 struct mbuf *m = m0; 231 int tlen, len, off; 232 bool tso; 233 struct route ip6route; 234 struct rtentry *rt = NULL, *rt_pmtu; 235 const struct sockaddr_in6 *dst; 236 struct sockaddr_in6 src_sa, dst_sa; 237 int error = 0; 238 struct in6_ifaddr *ia = NULL; 239 u_long mtu; 240 int alwaysfrag, dontfrag; 241 u_int32_t optlen = 0, plen = 0, unfragpartlen = 0; 242 struct ip6_exthdrs exthdrs; 243 struct in6_addr finaldst, src0, dst0; 244 u_int32_t zone; 245 struct route *ro_pmtu = NULL; 246 int hdrsplit = 0; 247 int needipsec = 0; 248 #ifdef IPSEC 249 struct secpolicy *sp = NULL; 250 #endif 251 struct psref psref, psref_ia; 252 int bound = curlwp_bind(); 253 bool release_psref_ia = false; 254 255 #ifdef DIAGNOSTIC 256 if ((m->m_flags & M_PKTHDR) == 0) 257 panic("ip6_output: no HDR"); 258 if ((m->m_pkthdr.csum_flags & 259 (M_CSUM_TCPv4|M_CSUM_UDPv4|M_CSUM_TSOv4)) != 0) { 260 panic("ip6_output: IPv4 checksum offload flags: %d", 261 m->m_pkthdr.csum_flags); 262 } 263 if ((m->m_pkthdr.csum_flags & (M_CSUM_TCPv6|M_CSUM_UDPv6)) == 264 (M_CSUM_TCPv6|M_CSUM_UDPv6)) { 265 panic("ip6_output: conflicting checksum offload flags: %d", 266 m->m_pkthdr.csum_flags); 267 } 268 #endif 269 270 M_CSUM_DATA_IPv6_HL_SET(m->m_pkthdr.csum_data, sizeof(struct ip6_hdr)); 271 272 #define MAKE_EXTHDR(hp, mp) \ 273 do { \ 274 if (hp) { \ 275 struct ip6_ext *eh = (struct ip6_ext *)(hp); \ 276 error = ip6_copyexthdr((mp), (void *)(hp), \ 277 ((eh)->ip6e_len + 1) << 3); \ 278 if (error) \ 279 goto freehdrs; \ 280 } \ 281 } while (/*CONSTCOND*/ 0) 282 283 memset(&exthdrs, 0, sizeof(exthdrs)); 284 if (opt) { 285 /* Hop-by-Hop options header */ 286 MAKE_EXTHDR(opt->ip6po_hbh, &exthdrs.ip6e_hbh); 287 /* Destination options header (1st part) */ 288 MAKE_EXTHDR(opt->ip6po_dest1, &exthdrs.ip6e_dest1); 289 /* Routing header */ 290 MAKE_EXTHDR(opt->ip6po_rthdr, &exthdrs.ip6e_rthdr); 291 /* Destination options header (2nd part) */ 292 MAKE_EXTHDR(opt->ip6po_dest2, &exthdrs.ip6e_dest2); 293 } 294 295 /* 296 * Calculate the total length of the extension header chain. 297 * Keep the length of the unfragmentable part for fragmentation. 298 */ 299 optlen = 0; 300 if (exthdrs.ip6e_hbh) 301 optlen += exthdrs.ip6e_hbh->m_len; 302 if (exthdrs.ip6e_dest1) 303 optlen += exthdrs.ip6e_dest1->m_len; 304 if (exthdrs.ip6e_rthdr) 305 optlen += exthdrs.ip6e_rthdr->m_len; 306 unfragpartlen = optlen + sizeof(struct ip6_hdr); 307 /* NOTE: we don't add AH/ESP length here. do that later. */ 308 if (exthdrs.ip6e_dest2) 309 optlen += exthdrs.ip6e_dest2->m_len; 310 311 #ifdef IPSEC 312 if (ipsec_used) { 313 /* Check the security policy (SP) for the packet */ 314 sp = ipsec6_check_policy(m, in6p, flags, &needipsec, &error); 315 if (error != 0) { 316 /* 317 * Hack: -EINVAL is used to signal that a packet 318 * should be silently discarded. This is typically 319 * because we asked key management for an SA and 320 * it was delayed (e.g. kicked up to IKE). 321 */ 322 if (error == -EINVAL) 323 error = 0; 324 goto freehdrs; 325 } 326 } 327 #endif 328 329 if (needipsec && 330 (m->m_pkthdr.csum_flags & (M_CSUM_UDPv6|M_CSUM_TCPv6)) != 0) { 331 in6_delayed_cksum(m); 332 m->m_pkthdr.csum_flags &= ~(M_CSUM_UDPv6|M_CSUM_TCPv6); 333 } 334 335 /* 336 * If we need IPsec, or there is at least one extension header, 337 * separate IP6 header from the payload. 338 */ 339 if ((needipsec || optlen) && !hdrsplit) { 340 if ((error = ip6_splithdr(m, &exthdrs)) != 0) { 341 m = NULL; 342 goto freehdrs; 343 } 344 m = exthdrs.ip6e_ip6; 345 hdrsplit++; 346 } 347 348 /* adjust pointer */ 349 ip6 = mtod(m, struct ip6_hdr *); 350 351 /* adjust mbuf packet header length */ 352 m->m_pkthdr.len += optlen; 353 plen = m->m_pkthdr.len - sizeof(*ip6); 354 355 /* If this is a jumbo payload, insert a jumbo payload option. */ 356 if (plen > IPV6_MAXPACKET) { 357 if (!hdrsplit) { 358 if ((error = ip6_splithdr(m, &exthdrs)) != 0) { 359 m = NULL; 360 goto freehdrs; 361 } 362 m = exthdrs.ip6e_ip6; 363 hdrsplit++; 364 } 365 /* adjust pointer */ 366 ip6 = mtod(m, struct ip6_hdr *); 367 if ((error = ip6_insert_jumboopt(&exthdrs, plen)) != 0) 368 goto freehdrs; 369 optlen += 8; /* XXX JUMBOOPTLEN */ 370 ip6->ip6_plen = 0; 371 } else 372 ip6->ip6_plen = htons(plen); 373 374 /* 375 * Concatenate headers and fill in next header fields. 376 * Here we have, on "m" 377 * IPv6 payload 378 * and we insert headers accordingly. Finally, we should be getting: 379 * IPv6 hbh dest1 rthdr ah* [esp* dest2 payload] 380 * 381 * during the header composing process, "m" points to IPv6 header. 382 * "mprev" points to an extension header prior to esp. 383 */ 384 { 385 u_char *nexthdrp = &ip6->ip6_nxt; 386 struct mbuf *mprev = m; 387 388 /* 389 * we treat dest2 specially. this makes IPsec processing 390 * much easier. the goal here is to make mprev point the 391 * mbuf prior to dest2. 392 * 393 * result: IPv6 dest2 payload 394 * m and mprev will point to IPv6 header. 395 */ 396 if (exthdrs.ip6e_dest2) { 397 if (!hdrsplit) 398 panic("assumption failed: hdr not split"); 399 exthdrs.ip6e_dest2->m_next = m->m_next; 400 m->m_next = exthdrs.ip6e_dest2; 401 *mtod(exthdrs.ip6e_dest2, u_char *) = ip6->ip6_nxt; 402 ip6->ip6_nxt = IPPROTO_DSTOPTS; 403 } 404 405 #define MAKE_CHAIN(m, mp, p, i)\ 406 do {\ 407 if (m) {\ 408 if (!hdrsplit) \ 409 panic("assumption failed: hdr not split"); \ 410 *mtod((m), u_char *) = *(p);\ 411 *(p) = (i);\ 412 p = mtod((m), u_char *);\ 413 (m)->m_next = (mp)->m_next;\ 414 (mp)->m_next = (m);\ 415 (mp) = (m);\ 416 }\ 417 } while (/*CONSTCOND*/ 0) 418 /* 419 * result: IPv6 hbh dest1 rthdr dest2 payload 420 * m will point to IPv6 header. mprev will point to the 421 * extension header prior to dest2 (rthdr in the above case). 422 */ 423 MAKE_CHAIN(exthdrs.ip6e_hbh, mprev, nexthdrp, IPPROTO_HOPOPTS); 424 MAKE_CHAIN(exthdrs.ip6e_dest1, mprev, nexthdrp, 425 IPPROTO_DSTOPTS); 426 MAKE_CHAIN(exthdrs.ip6e_rthdr, mprev, nexthdrp, 427 IPPROTO_ROUTING); 428 429 M_CSUM_DATA_IPv6_HL_SET(m->m_pkthdr.csum_data, 430 sizeof(struct ip6_hdr) + optlen); 431 } 432 433 /* Need to save for pmtu */ 434 finaldst = ip6->ip6_dst; 435 436 /* 437 * If there is a routing header, replace destination address field 438 * with the first hop of the routing header. 439 */ 440 if (exthdrs.ip6e_rthdr) { 441 struct ip6_rthdr *rh; 442 443 rh = mtod(exthdrs.ip6e_rthdr, struct ip6_rthdr *); 444 445 error = ip6_handle_rthdr(rh, ip6); 446 if (error != 0) 447 goto bad; 448 } 449 450 /* Source address validation */ 451 if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) && 452 (flags & IPV6_UNSPECSRC) == 0) { 453 error = EOPNOTSUPP; 454 IP6_STATINC(IP6_STAT_BADSCOPE); 455 goto bad; 456 } 457 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) { 458 error = EOPNOTSUPP; 459 IP6_STATINC(IP6_STAT_BADSCOPE); 460 goto bad; 461 } 462 463 IP6_STATINC(IP6_STAT_LOCALOUT); 464 465 /* 466 * Route packet. 467 */ 468 /* initialize cached route */ 469 if (ro == NULL) { 470 memset(&ip6route, 0, sizeof(ip6route)); 471 ro = &ip6route; 472 } 473 ro_pmtu = ro; 474 if (opt && opt->ip6po_rthdr) 475 ro = &opt->ip6po_route; 476 477 /* 478 * if specified, try to fill in the traffic class field. 479 * do not override if a non-zero value is already set. 480 * we check the diffserv field and the ecn field separately. 481 */ 482 if (opt && opt->ip6po_tclass >= 0) { 483 int mask = 0; 484 485 if ((ip6->ip6_flow & htonl(0xfc << 20)) == 0) 486 mask |= 0xfc; 487 if ((ip6->ip6_flow & htonl(0x03 << 20)) == 0) 488 mask |= 0x03; 489 if (mask != 0) 490 ip6->ip6_flow |= htonl((opt->ip6po_tclass & mask) << 20); 491 } 492 493 /* fill in or override the hop limit field, if necessary. */ 494 if (opt && opt->ip6po_hlim != -1) 495 ip6->ip6_hlim = opt->ip6po_hlim & 0xff; 496 else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) { 497 if (im6o != NULL) 498 ip6->ip6_hlim = im6o->im6o_multicast_hlim; 499 else 500 ip6->ip6_hlim = ip6_defmcasthlim; 501 } 502 503 #ifdef IPSEC 504 if (needipsec) { 505 int s = splsoftnet(); 506 error = ipsec6_process_packet(m, sp->req); 507 splx(s); 508 509 /* 510 * Preserve KAME behaviour: ENOENT can be returned 511 * when an SA acquire is in progress. Don't propagate 512 * this to user-level; it confuses applications. 513 * XXX this will go away when the SADB is redone. 514 */ 515 if (error == ENOENT) 516 error = 0; 517 518 goto done; 519 } 520 #endif 521 522 /* adjust pointer */ 523 ip6 = mtod(m, struct ip6_hdr *); 524 525 sockaddr_in6_init(&dst_sa, &ip6->ip6_dst, 0, 0, 0); 526 527 /* We do not need a route for multicast */ 528 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) { 529 struct in6_pktinfo *pi = NULL; 530 531 /* 532 * If the outgoing interface for the address is specified by 533 * the caller, use it. 534 */ 535 if (opt && (pi = opt->ip6po_pktinfo) != NULL) { 536 /* XXX boundary check is assumed to be already done. */ 537 ifp = if_get_byindex(pi->ipi6_ifindex, &psref); 538 } else if (im6o != NULL) { 539 ifp = if_get_byindex(im6o->im6o_multicast_if_index, 540 &psref); 541 } 542 } 543 544 if (ifp == NULL) { 545 error = in6_selectroute(&dst_sa, opt, &ro, &rt, true); 546 if (error != 0) 547 goto bad; 548 ifp = if_get_byindex(rt->rt_ifp->if_index, &psref); 549 } 550 551 if (rt == NULL) { 552 /* 553 * If in6_selectroute() does not return a route entry, 554 * dst may not have been updated. 555 */ 556 error = rtcache_setdst(ro, sin6tosa(&dst_sa)); 557 if (error) { 558 goto bad; 559 } 560 } 561 562 /* 563 * then rt (for unicast) and ifp must be non-NULL valid values. 564 */ 565 if ((flags & IPV6_FORWARDING) == 0) { 566 /* XXX: the FORWARDING flag can be set for mrouting. */ 567 in6_ifstat_inc(ifp, ifs6_out_request); 568 } 569 if (rt != NULL) { 570 ia = (struct in6_ifaddr *)(rt->rt_ifa); 571 rt->rt_use++; 572 } 573 574 /* 575 * The outgoing interface must be in the zone of source and 576 * destination addresses. We should use ia_ifp to support the 577 * case of sending packets to an address of our own. 578 */ 579 if (ia != NULL && ia->ia_ifp) { 580 origifp = ia->ia_ifp; 581 if (if_is_deactivated(origifp)) 582 goto bad; 583 if_acquire(origifp, &psref_ia); 584 release_psref_ia = true; 585 } else 586 origifp = ifp; 587 588 src0 = ip6->ip6_src; 589 if (in6_setscope(&src0, origifp, &zone)) 590 goto badscope; 591 sockaddr_in6_init(&src_sa, &ip6->ip6_src, 0, 0, 0); 592 if (sa6_recoverscope(&src_sa) || zone != src_sa.sin6_scope_id) 593 goto badscope; 594 595 dst0 = ip6->ip6_dst; 596 if (in6_setscope(&dst0, origifp, &zone)) 597 goto badscope; 598 /* re-initialize to be sure */ 599 sockaddr_in6_init(&dst_sa, &ip6->ip6_dst, 0, 0, 0); 600 if (sa6_recoverscope(&dst_sa) || zone != dst_sa.sin6_scope_id) 601 goto badscope; 602 603 /* scope check is done. */ 604 605 /* Ensure we only send from a valid address. */ 606 if ((ifp->if_flags & IFF_LOOPBACK) == 0 && 607 (error = ip6_ifaddrvalid(&src0, &dst0)) != 0) 608 { 609 char ip6buf[INET6_ADDRSTRLEN]; 610 nd6log(LOG_ERR, 611 "refusing to send from invalid address %s (pid %d)\n", 612 IN6_PRINT(ip6buf, &src0), curproc->p_pid); 613 IP6_STATINC(IP6_STAT_ODROPPED); 614 in6_ifstat_inc(origifp, ifs6_out_discard); 615 if (error == 1) 616 /* 617 * Address exists, but is tentative or detached. 618 * We can't send from it because it's invalid, 619 * so we drop the packet. 620 */ 621 error = 0; 622 else 623 error = EADDRNOTAVAIL; 624 goto bad; 625 } 626 627 if (rt != NULL && (rt->rt_flags & RTF_GATEWAY) && 628 !IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) 629 dst = satocsin6(rt->rt_gateway); 630 else 631 dst = satocsin6(rtcache_getdst(ro)); 632 633 /* 634 * XXXXXX: original code follows: 635 */ 636 if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) 637 m->m_flags &= ~(M_BCAST | M_MCAST); /* just in case */ 638 else { 639 bool ingroup; 640 641 m->m_flags = (m->m_flags & ~M_BCAST) | M_MCAST; 642 643 in6_ifstat_inc(ifp, ifs6_out_mcast); 644 645 /* 646 * Confirm that the outgoing interface supports multicast. 647 */ 648 if (!(ifp->if_flags & IFF_MULTICAST)) { 649 IP6_STATINC(IP6_STAT_NOROUTE); 650 in6_ifstat_inc(ifp, ifs6_out_discard); 651 error = ENETUNREACH; 652 goto bad; 653 } 654 655 ingroup = in6_multi_group(&ip6->ip6_dst, ifp); 656 if (ingroup && (im6o == NULL || im6o->im6o_multicast_loop)) { 657 /* 658 * If we belong to the destination multicast group 659 * on the outgoing interface, and the caller did not 660 * forbid loopback, loop back a copy. 661 */ 662 KASSERT(dst != NULL); 663 ip6_mloopback(ifp, m, dst); 664 } else { 665 /* 666 * If we are acting as a multicast router, perform 667 * multicast forwarding as if the packet had just 668 * arrived on the interface to which we are about 669 * to send. The multicast forwarding function 670 * recursively calls this function, using the 671 * IPV6_FORWARDING flag to prevent infinite recursion. 672 * 673 * Multicasts that are looped back by ip6_mloopback(), 674 * above, will be forwarded by the ip6_input() routine, 675 * if necessary. 676 */ 677 if (ip6_mrouter && (flags & IPV6_FORWARDING) == 0) { 678 if (ip6_mforward(ip6, ifp, m) != 0) { 679 m_freem(m); 680 goto done; 681 } 682 } 683 } 684 /* 685 * Multicasts with a hoplimit of zero may be looped back, 686 * above, but must not be transmitted on a network. 687 * Also, multicasts addressed to the loopback interface 688 * are not sent -- the above call to ip6_mloopback() will 689 * loop back a copy if this host actually belongs to the 690 * destination group on the loopback interface. 691 */ 692 if (ip6->ip6_hlim == 0 || (ifp->if_flags & IFF_LOOPBACK) || 693 IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst)) { 694 m_freem(m); 695 goto done; 696 } 697 } 698 699 /* 700 * Fill the outgoing inteface to tell the upper layer 701 * to increment per-interface statistics. 702 */ 703 if (ifpp) 704 *ifpp = ifp; 705 706 /* Determine path MTU. */ 707 /* 708 * ro_pmtu represent final destination while 709 * ro might represent immediate destination. 710 * Use ro_pmtu destination since MTU might differ. 711 */ 712 if (ro_pmtu != ro) { 713 union { 714 struct sockaddr dst; 715 struct sockaddr_in6 dst6; 716 } u; 717 718 /* ro_pmtu may not have a cache */ 719 sockaddr_in6_init(&u.dst6, &finaldst, 0, 0, 0); 720 rt_pmtu = rtcache_lookup(ro_pmtu, &u.dst); 721 } else 722 rt_pmtu = rt; 723 error = ip6_getpmtu(rt_pmtu, ifp, &mtu, &alwaysfrag); 724 if (rt_pmtu != NULL && rt_pmtu != rt) 725 rtcache_unref(rt_pmtu, ro_pmtu); 726 if (error != 0) 727 goto bad; 728 729 /* 730 * The caller of this function may specify to use the minimum MTU 731 * in some cases. 732 * An advanced API option (IPV6_USE_MIN_MTU) can also override MTU 733 * setting. The logic is a bit complicated; by default, unicast 734 * packets will follow path MTU while multicast packets will be sent at 735 * the minimum MTU. If IP6PO_MINMTU_ALL is specified, all packets 736 * including unicast ones will be sent at the minimum MTU. Multicast 737 * packets will always be sent at the minimum MTU unless 738 * IP6PO_MINMTU_DISABLE is explicitly specified. 739 * See RFC 3542 for more details. 740 */ 741 if (mtu > IPV6_MMTU) { 742 if ((flags & IPV6_MINMTU)) 743 mtu = IPV6_MMTU; 744 else if (opt && opt->ip6po_minmtu == IP6PO_MINMTU_ALL) 745 mtu = IPV6_MMTU; 746 else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) && 747 (opt == NULL || 748 opt->ip6po_minmtu != IP6PO_MINMTU_DISABLE)) { 749 mtu = IPV6_MMTU; 750 } 751 } 752 753 /* 754 * clear embedded scope identifiers if necessary. 755 * in6_clearscope will touch the addresses only when necessary. 756 */ 757 in6_clearscope(&ip6->ip6_src); 758 in6_clearscope(&ip6->ip6_dst); 759 760 /* 761 * If the outgoing packet contains a hop-by-hop options header, 762 * it must be examined and processed even by the source node. 763 * (RFC 2460, section 4.) 764 * 765 * XXX Is this really necessary? 766 */ 767 if (ip6->ip6_nxt == IPPROTO_HOPOPTS) { 768 u_int32_t dummy1; /* XXX unused */ 769 u_int32_t dummy2; /* XXX unused */ 770 int hoff = sizeof(struct ip6_hdr); 771 772 if (ip6_hopopts_input(&dummy1, &dummy2, &m, &hoff)) { 773 /* m was already freed at this point */ 774 error = EINVAL; 775 goto done; 776 } 777 778 ip6 = mtod(m, struct ip6_hdr *); 779 } 780 781 /* 782 * Run through list of hooks for output packets. 783 */ 784 if ((error = pfil_run_hooks(inet6_pfil_hook, &m, ifp, PFIL_OUT)) != 0) 785 goto done; 786 if (m == NULL) 787 goto done; 788 ip6 = mtod(m, struct ip6_hdr *); 789 790 /* 791 * Send the packet to the outgoing interface. 792 * If necessary, do IPv6 fragmentation before sending. 793 * 794 * the logic here is rather complex: 795 * 1: normal case (dontfrag == 0, alwaysfrag == 0) 796 * 1-a: send as is if tlen <= path mtu 797 * 1-b: fragment if tlen > path mtu 798 * 799 * 2: if user asks us not to fragment (dontfrag == 1) 800 * 2-a: send as is if tlen <= interface mtu 801 * 2-b: error if tlen > interface mtu 802 * 803 * 3: if we always need to attach fragment header (alwaysfrag == 1) 804 * always fragment 805 * 806 * 4: if dontfrag == 1 && alwaysfrag == 1 807 * error, as we cannot handle this conflicting request 808 */ 809 tlen = m->m_pkthdr.len; 810 tso = (m->m_pkthdr.csum_flags & M_CSUM_TSOv6) != 0; 811 if (opt && (opt->ip6po_flags & IP6PO_DONTFRAG)) 812 dontfrag = 1; 813 else 814 dontfrag = 0; 815 816 if (dontfrag && alwaysfrag) { /* case 4 */ 817 /* conflicting request - can't transmit */ 818 error = EMSGSIZE; 819 goto bad; 820 } 821 if (dontfrag && (!tso && tlen > IN6_LINKMTU(ifp))) { /* case 2-b */ 822 /* 823 * Even if the DONTFRAG option is specified, we cannot send the 824 * packet when the data length is larger than the MTU of the 825 * outgoing interface. 826 * Notify the error by sending IPV6_PATHMTU ancillary data as 827 * well as returning an error code (the latter is not described 828 * in the API spec.) 829 */ 830 u_int32_t mtu32; 831 struct ip6ctlparam ip6cp; 832 833 mtu32 = (u_int32_t)mtu; 834 memset(&ip6cp, 0, sizeof(ip6cp)); 835 ip6cp.ip6c_cmdarg = (void *)&mtu32; 836 pfctlinput2(PRC_MSGSIZE, 837 rtcache_getdst(ro_pmtu), &ip6cp); 838 839 error = EMSGSIZE; 840 goto bad; 841 } 842 843 /* 844 * transmit packet without fragmentation 845 */ 846 if (dontfrag || (!alwaysfrag && (tlen <= mtu || tso))) { 847 /* case 1-a and 2-a */ 848 struct in6_ifaddr *ia6; 849 int sw_csum; 850 int s; 851 852 ip6 = mtod(m, struct ip6_hdr *); 853 s = pserialize_read_enter(); 854 ia6 = in6_ifawithifp(ifp, &ip6->ip6_src); 855 if (ia6) { 856 /* Record statistics for this interface address. */ 857 ia6->ia_ifa.ifa_data.ifad_outbytes += m->m_pkthdr.len; 858 } 859 pserialize_read_exit(s); 860 861 sw_csum = m->m_pkthdr.csum_flags & ~ifp->if_csum_flags_tx; 862 if ((sw_csum & (M_CSUM_UDPv6|M_CSUM_TCPv6)) != 0) { 863 if (IN6_NEED_CHECKSUM(ifp, 864 sw_csum & (M_CSUM_UDPv6|M_CSUM_TCPv6))) { 865 in6_delayed_cksum(m); 866 } 867 m->m_pkthdr.csum_flags &= ~(M_CSUM_UDPv6|M_CSUM_TCPv6); 868 } 869 870 KASSERT(dst != NULL); 871 if (__predict_true(!tso || 872 (ifp->if_capenable & IFCAP_TSOv6) != 0)) { 873 error = ip6_if_output(ifp, origifp, m, dst, rt); 874 } else { 875 error = ip6_tso_output(ifp, origifp, m, dst, rt); 876 } 877 goto done; 878 } 879 880 if (tso) { 881 error = EINVAL; /* XXX */ 882 goto bad; 883 } 884 885 /* 886 * try to fragment the packet. case 1-b and 3 887 */ 888 if (mtu < IPV6_MMTU) { 889 /* path MTU cannot be less than IPV6_MMTU */ 890 error = EMSGSIZE; 891 in6_ifstat_inc(ifp, ifs6_out_fragfail); 892 goto bad; 893 } else if (ip6->ip6_plen == 0) { 894 /* jumbo payload cannot be fragmented */ 895 error = EMSGSIZE; 896 in6_ifstat_inc(ifp, ifs6_out_fragfail); 897 goto bad; 898 } else { 899 const u_int32_t id = htonl(ip6_randomid()); 900 struct mbuf **mnext, *m_frgpart; 901 const int hlen = unfragpartlen; 902 struct ip6_frag *ip6f; 903 u_char nextproto; 904 #if 0 /* see below */ 905 struct ip6ctlparam ip6cp; 906 u_int32_t mtu32; 907 #endif 908 909 if (mtu > IPV6_MAXPACKET) 910 mtu = IPV6_MAXPACKET; 911 912 #if 0 913 /* 914 * It is believed this code is a leftover from the 915 * development of the IPV6_RECVPATHMTU sockopt and 916 * associated work to implement RFC3542. 917 * It's not entirely clear what the intent of the API 918 * is at this point, so disable this code for now. 919 * The IPV6_RECVPATHMTU sockopt and/or IPV6_DONTFRAG 920 * will send notifications if the application requests. 921 */ 922 923 /* Notify a proper path MTU to applications. */ 924 mtu32 = (u_int32_t)mtu; 925 memset(&ip6cp, 0, sizeof(ip6cp)); 926 ip6cp.ip6c_cmdarg = (void *)&mtu32; 927 pfctlinput2(PRC_MSGSIZE, 928 rtcache_getdst(ro_pmtu), &ip6cp); 929 #endif 930 931 /* 932 * Must be able to put at least 8 bytes per fragment. 933 */ 934 len = (mtu - hlen - sizeof(struct ip6_frag)) & ~7; 935 if (len < 8) { 936 error = EMSGSIZE; 937 in6_ifstat_inc(ifp, ifs6_out_fragfail); 938 goto bad; 939 } 940 941 mnext = &m->m_nextpkt; 942 943 /* 944 * Change the next header field of the last header in the 945 * unfragmentable part. 946 */ 947 if (exthdrs.ip6e_rthdr) { 948 nextproto = *mtod(exthdrs.ip6e_rthdr, u_char *); 949 *mtod(exthdrs.ip6e_rthdr, u_char *) = IPPROTO_FRAGMENT; 950 } else if (exthdrs.ip6e_dest1) { 951 nextproto = *mtod(exthdrs.ip6e_dest1, u_char *); 952 *mtod(exthdrs.ip6e_dest1, u_char *) = IPPROTO_FRAGMENT; 953 } else if (exthdrs.ip6e_hbh) { 954 nextproto = *mtod(exthdrs.ip6e_hbh, u_char *); 955 *mtod(exthdrs.ip6e_hbh, u_char *) = IPPROTO_FRAGMENT; 956 } else { 957 nextproto = ip6->ip6_nxt; 958 ip6->ip6_nxt = IPPROTO_FRAGMENT; 959 } 960 961 if ((m->m_pkthdr.csum_flags & (M_CSUM_UDPv6|M_CSUM_TCPv6)) 962 != 0) { 963 if (IN6_NEED_CHECKSUM(ifp, 964 m->m_pkthdr.csum_flags & 965 (M_CSUM_UDPv6|M_CSUM_TCPv6))) { 966 in6_delayed_cksum(m); 967 } 968 m->m_pkthdr.csum_flags &= ~(M_CSUM_UDPv6|M_CSUM_TCPv6); 969 } 970 971 /* 972 * Loop through length of segment after first fragment, 973 * make new header and copy data of each part and link onto 974 * chain. 975 */ 976 m0 = m; 977 for (off = hlen; off < tlen; off += len) { 978 struct mbuf *mlast; 979 980 MGETHDR(m, M_DONTWAIT, MT_HEADER); 981 if (!m) { 982 error = ENOBUFS; 983 IP6_STATINC(IP6_STAT_ODROPPED); 984 goto sendorfree; 985 } 986 m_reset_rcvif(m); 987 m->m_flags = m0->m_flags & M_COPYFLAGS; 988 *mnext = m; 989 mnext = &m->m_nextpkt; 990 m->m_data += max_linkhdr; 991 mhip6 = mtod(m, struct ip6_hdr *); 992 *mhip6 = *ip6; 993 m->m_len = sizeof(*mhip6); 994 995 ip6f = NULL; 996 error = ip6_insertfraghdr(m0, m, hlen, &ip6f); 997 if (error) { 998 IP6_STATINC(IP6_STAT_ODROPPED); 999 goto sendorfree; 1000 } 1001 1002 /* Fill in the Frag6 Header */ 1003 ip6f->ip6f_offlg = htons((u_int16_t)((off - hlen) & ~7)); 1004 if (off + len >= tlen) 1005 len = tlen - off; 1006 else 1007 ip6f->ip6f_offlg |= IP6F_MORE_FRAG; 1008 ip6f->ip6f_reserved = 0; 1009 ip6f->ip6f_ident = id; 1010 ip6f->ip6f_nxt = nextproto; 1011 1012 mhip6->ip6_plen = htons((u_int16_t)(len + hlen + 1013 sizeof(*ip6f) - sizeof(struct ip6_hdr))); 1014 if ((m_frgpart = m_copy(m0, off, len)) == NULL) { 1015 error = ENOBUFS; 1016 IP6_STATINC(IP6_STAT_ODROPPED); 1017 goto sendorfree; 1018 } 1019 for (mlast = m; mlast->m_next; mlast = mlast->m_next) 1020 ; 1021 mlast->m_next = m_frgpart; 1022 1023 m->m_pkthdr.len = len + hlen + sizeof(*ip6f); 1024 m_reset_rcvif(m); 1025 IP6_STATINC(IP6_STAT_OFRAGMENTS); 1026 in6_ifstat_inc(ifp, ifs6_out_fragcreat); 1027 } 1028 1029 in6_ifstat_inc(ifp, ifs6_out_fragok); 1030 } 1031 1032 sendorfree: 1033 m = m0->m_nextpkt; 1034 m0->m_nextpkt = 0; 1035 m_freem(m0); 1036 for (m0 = m; m; m = m0) { 1037 m0 = m->m_nextpkt; 1038 m->m_nextpkt = 0; 1039 if (error == 0) { 1040 struct in6_ifaddr *ia6; 1041 int s; 1042 ip6 = mtod(m, struct ip6_hdr *); 1043 s = pserialize_read_enter(); 1044 ia6 = in6_ifawithifp(ifp, &ip6->ip6_src); 1045 if (ia6) { 1046 /* 1047 * Record statistics for this interface 1048 * address. 1049 */ 1050 ia6->ia_ifa.ifa_data.ifad_outbytes += 1051 m->m_pkthdr.len; 1052 } 1053 pserialize_read_exit(s); 1054 KASSERT(dst != NULL); 1055 error = ip6_if_output(ifp, origifp, m, dst, rt); 1056 } else 1057 m_freem(m); 1058 } 1059 1060 if (error == 0) 1061 IP6_STATINC(IP6_STAT_FRAGMENTED); 1062 1063 done: 1064 rtcache_unref(rt, ro); 1065 if (ro == &ip6route) 1066 rtcache_free(&ip6route); 1067 #ifdef IPSEC 1068 if (sp != NULL) 1069 KEY_SP_UNREF(&sp); 1070 #endif 1071 if_put(ifp, &psref); 1072 if (release_psref_ia) 1073 if_put(origifp, &psref_ia); 1074 curlwp_bindx(bound); 1075 1076 return error; 1077 1078 freehdrs: 1079 m_freem(exthdrs.ip6e_hbh); 1080 m_freem(exthdrs.ip6e_dest1); 1081 m_freem(exthdrs.ip6e_rthdr); 1082 m_freem(exthdrs.ip6e_dest2); 1083 /* FALLTHROUGH */ 1084 bad: 1085 m_freem(m); 1086 goto done; 1087 1088 badscope: 1089 IP6_STATINC(IP6_STAT_BADSCOPE); 1090 in6_ifstat_inc(origifp, ifs6_out_discard); 1091 if (error == 0) 1092 error = EHOSTUNREACH; /* XXX */ 1093 goto bad; 1094 } 1095 1096 static int 1097 ip6_copyexthdr(struct mbuf **mp, void *hdr, int hlen) 1098 { 1099 struct mbuf *m; 1100 1101 if (hlen > MCLBYTES) 1102 return ENOBUFS; /* XXX */ 1103 1104 MGET(m, M_DONTWAIT, MT_DATA); 1105 if (!m) 1106 return ENOBUFS; 1107 1108 if (hlen > MLEN) { 1109 MCLGET(m, M_DONTWAIT); 1110 if ((m->m_flags & M_EXT) == 0) { 1111 m_free(m); 1112 return ENOBUFS; 1113 } 1114 } 1115 m->m_len = hlen; 1116 if (hdr) 1117 memcpy(mtod(m, void *), hdr, hlen); 1118 1119 *mp = m; 1120 return 0; 1121 } 1122 1123 /* 1124 * Process a delayed payload checksum calculation. 1125 */ 1126 void 1127 in6_delayed_cksum(struct mbuf *m) 1128 { 1129 uint16_t csum, offset; 1130 1131 KASSERT((m->m_pkthdr.csum_flags & (M_CSUM_UDPv6|M_CSUM_TCPv6)) != 0); 1132 KASSERT((~m->m_pkthdr.csum_flags & (M_CSUM_UDPv6|M_CSUM_TCPv6)) != 0); 1133 KASSERT((m->m_pkthdr.csum_flags 1134 & (M_CSUM_UDPv4|M_CSUM_TCPv4|M_CSUM_TSOv4)) == 0); 1135 1136 offset = M_CSUM_DATA_IPv6_HL(m->m_pkthdr.csum_data); 1137 csum = in6_cksum(m, 0, offset, m->m_pkthdr.len - offset); 1138 if (csum == 0 && (m->m_pkthdr.csum_flags & M_CSUM_UDPv6) != 0) { 1139 csum = 0xffff; 1140 } 1141 1142 offset += M_CSUM_DATA_IPv6_OFFSET(m->m_pkthdr.csum_data); 1143 if ((offset + sizeof(csum)) > m->m_len) { 1144 m_copyback(m, offset, sizeof(csum), &csum); 1145 } else { 1146 *(uint16_t *)(mtod(m, char *) + offset) = csum; 1147 } 1148 } 1149 1150 /* 1151 * Insert jumbo payload option. 1152 */ 1153 static int 1154 ip6_insert_jumboopt(struct ip6_exthdrs *exthdrs, u_int32_t plen) 1155 { 1156 struct mbuf *mopt; 1157 u_int8_t *optbuf; 1158 u_int32_t v; 1159 1160 #define JUMBOOPTLEN 8 /* length of jumbo payload option and padding */ 1161 1162 /* 1163 * If there is no hop-by-hop options header, allocate new one. 1164 * If there is one but it doesn't have enough space to store the 1165 * jumbo payload option, allocate a cluster to store the whole options. 1166 * Otherwise, use it to store the options. 1167 */ 1168 if (exthdrs->ip6e_hbh == NULL) { 1169 MGET(mopt, M_DONTWAIT, MT_DATA); 1170 if (mopt == 0) 1171 return (ENOBUFS); 1172 mopt->m_len = JUMBOOPTLEN; 1173 optbuf = mtod(mopt, u_int8_t *); 1174 optbuf[1] = 0; /* = ((JUMBOOPTLEN) >> 3) - 1 */ 1175 exthdrs->ip6e_hbh = mopt; 1176 } else { 1177 struct ip6_hbh *hbh; 1178 1179 mopt = exthdrs->ip6e_hbh; 1180 if (M_TRAILINGSPACE(mopt) < JUMBOOPTLEN) { 1181 const int oldoptlen = mopt->m_len; 1182 struct mbuf *n; 1183 1184 /* 1185 * Assumptions: 1186 * - exthdrs->ip6e_hbh is not referenced from places 1187 * other than exthdrs. 1188 * - exthdrs->ip6e_hbh is not an mbuf chain. 1189 */ 1190 KASSERT(mopt->m_next == NULL); 1191 1192 /* 1193 * Give up if the whole (new) hbh header does not fit 1194 * even in an mbuf cluster. 1195 */ 1196 if (oldoptlen + JUMBOOPTLEN > MCLBYTES) 1197 return ENOBUFS; 1198 1199 /* 1200 * At this point, we must always prepare a cluster. 1201 */ 1202 MGET(n, M_DONTWAIT, MT_DATA); 1203 if (n) { 1204 MCLGET(n, M_DONTWAIT); 1205 if ((n->m_flags & M_EXT) == 0) { 1206 m_freem(n); 1207 n = NULL; 1208 } 1209 } 1210 if (!n) 1211 return ENOBUFS; 1212 1213 n->m_len = oldoptlen + JUMBOOPTLEN; 1214 bcopy(mtod(mopt, void *), mtod(n, void *), 1215 oldoptlen); 1216 optbuf = mtod(n, u_int8_t *) + oldoptlen; 1217 m_freem(mopt); 1218 mopt = exthdrs->ip6e_hbh = n; 1219 } else { 1220 optbuf = mtod(mopt, u_int8_t *) + mopt->m_len; 1221 mopt->m_len += JUMBOOPTLEN; 1222 } 1223 optbuf[0] = IP6OPT_PADN; 1224 optbuf[1] = 0; 1225 1226 /* 1227 * Adjust the header length according to the pad and 1228 * the jumbo payload option. 1229 */ 1230 hbh = mtod(mopt, struct ip6_hbh *); 1231 hbh->ip6h_len += (JUMBOOPTLEN >> 3); 1232 } 1233 1234 /* fill in the option. */ 1235 optbuf[2] = IP6OPT_JUMBO; 1236 optbuf[3] = 4; 1237 v = (u_int32_t)htonl(plen + JUMBOOPTLEN); 1238 memcpy(&optbuf[4], &v, sizeof(u_int32_t)); 1239 1240 /* finally, adjust the packet header length */ 1241 exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN; 1242 1243 return 0; 1244 #undef JUMBOOPTLEN 1245 } 1246 1247 /* 1248 * Insert fragment header and copy unfragmentable header portions. 1249 * 1250 * *frghdrp will not be read, and it is guaranteed that either an 1251 * error is returned or that *frghdrp will point to space allocated 1252 * for the fragment header. 1253 * 1254 * On entry, m contains: 1255 * IPv6 Header 1256 * On exit, it contains: 1257 * IPv6 Header -> Unfragmentable Part -> Frag6 Header 1258 */ 1259 static int 1260 ip6_insertfraghdr(struct mbuf *m0, struct mbuf *m, int hlen, 1261 struct ip6_frag **frghdrp) 1262 { 1263 struct mbuf *n, *mlast; 1264 1265 if (hlen > sizeof(struct ip6_hdr)) { 1266 n = m_copym(m0, sizeof(struct ip6_hdr), 1267 hlen - sizeof(struct ip6_hdr), M_DONTWAIT); 1268 if (n == NULL) 1269 return ENOBUFS; 1270 m->m_next = n; 1271 } else 1272 n = m; 1273 1274 /* Search for the last mbuf of unfragmentable part. */ 1275 for (mlast = n; mlast->m_next; mlast = mlast->m_next) 1276 ; 1277 1278 if ((mlast->m_flags & M_EXT) == 0 && 1279 M_TRAILINGSPACE(mlast) >= sizeof(struct ip6_frag)) { 1280 /* use the trailing space of the last mbuf for the fragment hdr */ 1281 *frghdrp = (struct ip6_frag *)(mtod(mlast, char *) + 1282 mlast->m_len); 1283 mlast->m_len += sizeof(struct ip6_frag); 1284 } else { 1285 /* allocate a new mbuf for the fragment header */ 1286 struct mbuf *mfrg; 1287 1288 MGET(mfrg, M_DONTWAIT, MT_DATA); 1289 if (mfrg == NULL) 1290 return ENOBUFS; 1291 mfrg->m_len = sizeof(struct ip6_frag); 1292 *frghdrp = mtod(mfrg, struct ip6_frag *); 1293 mlast->m_next = mfrg; 1294 } 1295 1296 return 0; 1297 } 1298 1299 static int 1300 ip6_getpmtu(struct rtentry *rt, struct ifnet *ifp, u_long *mtup, 1301 int *alwaysfragp) 1302 { 1303 u_int32_t mtu = 0; 1304 int alwaysfrag = 0; 1305 int error = 0; 1306 1307 if (rt != NULL) { 1308 u_int32_t ifmtu; 1309 1310 if (ifp == NULL) 1311 ifp = rt->rt_ifp; 1312 ifmtu = IN6_LINKMTU(ifp); 1313 mtu = rt->rt_rmx.rmx_mtu; 1314 if (mtu == 0) 1315 mtu = ifmtu; 1316 else if (mtu < IPV6_MMTU) { 1317 /* 1318 * RFC2460 section 5, last paragraph: 1319 * if we record ICMPv6 too big message with 1320 * mtu < IPV6_MMTU, transmit packets sized IPV6_MMTU 1321 * or smaller, with fragment header attached. 1322 * (fragment header is needed regardless from the 1323 * packet size, for translators to identify packets) 1324 */ 1325 alwaysfrag = 1; 1326 mtu = IPV6_MMTU; 1327 } else if (mtu > ifmtu) { 1328 /* 1329 * The MTU on the route is larger than the MTU on 1330 * the interface! This shouldn't happen, unless the 1331 * MTU of the interface has been changed after the 1332 * interface was brought up. Change the MTU in the 1333 * route to match the interface MTU (as long as the 1334 * field isn't locked). 1335 */ 1336 mtu = ifmtu; 1337 if (!(rt->rt_rmx.rmx_locks & RTV_MTU)) 1338 rt->rt_rmx.rmx_mtu = mtu; 1339 } 1340 } else if (ifp) { 1341 mtu = IN6_LINKMTU(ifp); 1342 } else 1343 error = EHOSTUNREACH; /* XXX */ 1344 1345 *mtup = mtu; 1346 if (alwaysfragp) 1347 *alwaysfragp = alwaysfrag; 1348 return (error); 1349 } 1350 1351 /* 1352 * IP6 socket option processing. 1353 */ 1354 int 1355 ip6_ctloutput(int op, struct socket *so, struct sockopt *sopt) 1356 { 1357 int optdatalen, uproto; 1358 void *optdata; 1359 struct in6pcb *in6p = sotoin6pcb(so); 1360 struct ip_moptions **mopts; 1361 int error, optval; 1362 int level, optname; 1363 1364 KASSERT(solocked(so)); 1365 KASSERT(sopt != NULL); 1366 1367 level = sopt->sopt_level; 1368 optname = sopt->sopt_name; 1369 1370 error = optval = 0; 1371 uproto = (int)so->so_proto->pr_protocol; 1372 1373 switch (level) { 1374 case IPPROTO_IP: 1375 switch (optname) { 1376 case IP_ADD_MEMBERSHIP: 1377 case IP_DROP_MEMBERSHIP: 1378 case IP_MULTICAST_IF: 1379 case IP_MULTICAST_LOOP: 1380 case IP_MULTICAST_TTL: 1381 mopts = &in6p->in6p_v4moptions; 1382 switch (op) { 1383 case PRCO_GETOPT: 1384 return ip_getmoptions(*mopts, sopt); 1385 case PRCO_SETOPT: 1386 return ip_setmoptions(mopts, sopt); 1387 default: 1388 return EINVAL; 1389 } 1390 default: 1391 return ENOPROTOOPT; 1392 } 1393 case IPPROTO_IPV6: 1394 break; 1395 default: 1396 return ENOPROTOOPT; 1397 } 1398 switch (op) { 1399 case PRCO_SETOPT: 1400 switch (optname) { 1401 #ifdef RFC2292 1402 case IPV6_2292PKTOPTIONS: 1403 error = ip6_pcbopts(&in6p->in6p_outputopts, so, sopt); 1404 break; 1405 #endif 1406 1407 /* 1408 * Use of some Hop-by-Hop options or some 1409 * Destination options, might require special 1410 * privilege. That is, normal applications 1411 * (without special privilege) might be forbidden 1412 * from setting certain options in outgoing packets, 1413 * and might never see certain options in received 1414 * packets. [RFC 2292 Section 6] 1415 * KAME specific note: 1416 * KAME prevents non-privileged users from sending or 1417 * receiving ANY hbh/dst options in order to avoid 1418 * overhead of parsing options in the kernel. 1419 */ 1420 case IPV6_RECVHOPOPTS: 1421 case IPV6_RECVDSTOPTS: 1422 case IPV6_RECVRTHDRDSTOPTS: 1423 error = kauth_authorize_network(kauth_cred_get(), 1424 KAUTH_NETWORK_IPV6, KAUTH_REQ_NETWORK_IPV6_HOPBYHOP, 1425 NULL, NULL, NULL); 1426 if (error) 1427 break; 1428 /* FALLTHROUGH */ 1429 case IPV6_UNICAST_HOPS: 1430 case IPV6_HOPLIMIT: 1431 case IPV6_FAITH: 1432 1433 case IPV6_RECVPKTINFO: 1434 case IPV6_RECVHOPLIMIT: 1435 case IPV6_RECVRTHDR: 1436 case IPV6_RECVPATHMTU: 1437 case IPV6_RECVTCLASS: 1438 case IPV6_V6ONLY: 1439 error = sockopt_getint(sopt, &optval); 1440 if (error) 1441 break; 1442 switch (optname) { 1443 case IPV6_UNICAST_HOPS: 1444 if (optval < -1 || optval >= 256) 1445 error = EINVAL; 1446 else { 1447 /* -1 = kernel default */ 1448 in6p->in6p_hops = optval; 1449 } 1450 break; 1451 #define OPTSET(bit) \ 1452 do { \ 1453 if (optval) \ 1454 in6p->in6p_flags |= (bit); \ 1455 else \ 1456 in6p->in6p_flags &= ~(bit); \ 1457 } while (/*CONSTCOND*/ 0) 1458 1459 #ifdef RFC2292 1460 #define OPTSET2292(bit) \ 1461 do { \ 1462 in6p->in6p_flags |= IN6P_RFC2292; \ 1463 if (optval) \ 1464 in6p->in6p_flags |= (bit); \ 1465 else \ 1466 in6p->in6p_flags &= ~(bit); \ 1467 } while (/*CONSTCOND*/ 0) 1468 #endif 1469 1470 #define OPTBIT(bit) (in6p->in6p_flags & (bit) ? 1 : 0) 1471 1472 case IPV6_RECVPKTINFO: 1473 #ifdef RFC2292 1474 /* cannot mix with RFC2292 */ 1475 if (OPTBIT(IN6P_RFC2292)) { 1476 error = EINVAL; 1477 break; 1478 } 1479 #endif 1480 OPTSET(IN6P_PKTINFO); 1481 break; 1482 1483 case IPV6_HOPLIMIT: 1484 { 1485 struct ip6_pktopts **optp; 1486 1487 #ifdef RFC2292 1488 /* cannot mix with RFC2292 */ 1489 if (OPTBIT(IN6P_RFC2292)) { 1490 error = EINVAL; 1491 break; 1492 } 1493 #endif 1494 optp = &in6p->in6p_outputopts; 1495 error = ip6_pcbopt(IPV6_HOPLIMIT, 1496 (u_char *)&optval, 1497 sizeof(optval), 1498 optp, 1499 kauth_cred_get(), uproto); 1500 break; 1501 } 1502 1503 case IPV6_RECVHOPLIMIT: 1504 #ifdef RFC2292 1505 /* cannot mix with RFC2292 */ 1506 if (OPTBIT(IN6P_RFC2292)) { 1507 error = EINVAL; 1508 break; 1509 } 1510 #endif 1511 OPTSET(IN6P_HOPLIMIT); 1512 break; 1513 1514 case IPV6_RECVHOPOPTS: 1515 #ifdef RFC2292 1516 /* cannot mix with RFC2292 */ 1517 if (OPTBIT(IN6P_RFC2292)) { 1518 error = EINVAL; 1519 break; 1520 } 1521 #endif 1522 OPTSET(IN6P_HOPOPTS); 1523 break; 1524 1525 case IPV6_RECVDSTOPTS: 1526 #ifdef RFC2292 1527 /* cannot mix with RFC2292 */ 1528 if (OPTBIT(IN6P_RFC2292)) { 1529 error = EINVAL; 1530 break; 1531 } 1532 #endif 1533 OPTSET(IN6P_DSTOPTS); 1534 break; 1535 1536 case IPV6_RECVRTHDRDSTOPTS: 1537 #ifdef RFC2292 1538 /* cannot mix with RFC2292 */ 1539 if (OPTBIT(IN6P_RFC2292)) { 1540 error = EINVAL; 1541 break; 1542 } 1543 #endif 1544 OPTSET(IN6P_RTHDRDSTOPTS); 1545 break; 1546 1547 case IPV6_RECVRTHDR: 1548 #ifdef RFC2292 1549 /* cannot mix with RFC2292 */ 1550 if (OPTBIT(IN6P_RFC2292)) { 1551 error = EINVAL; 1552 break; 1553 } 1554 #endif 1555 OPTSET(IN6P_RTHDR); 1556 break; 1557 1558 case IPV6_FAITH: 1559 OPTSET(IN6P_FAITH); 1560 break; 1561 1562 case IPV6_RECVPATHMTU: 1563 /* 1564 * We ignore this option for TCP 1565 * sockets. 1566 * (RFC3542 leaves this case 1567 * unspecified.) 1568 */ 1569 if (uproto != IPPROTO_TCP) 1570 OPTSET(IN6P_MTU); 1571 break; 1572 1573 case IPV6_V6ONLY: 1574 /* 1575 * make setsockopt(IPV6_V6ONLY) 1576 * available only prior to bind(2). 1577 * see ipng mailing list, Jun 22 2001. 1578 */ 1579 if (in6p->in6p_lport || 1580 !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr)) { 1581 error = EINVAL; 1582 break; 1583 } 1584 #ifdef INET6_BINDV6ONLY 1585 if (!optval) 1586 error = EINVAL; 1587 #else 1588 OPTSET(IN6P_IPV6_V6ONLY); 1589 #endif 1590 break; 1591 case IPV6_RECVTCLASS: 1592 #ifdef RFC2292 1593 /* cannot mix with RFC2292 XXX */ 1594 if (OPTBIT(IN6P_RFC2292)) { 1595 error = EINVAL; 1596 break; 1597 } 1598 #endif 1599 OPTSET(IN6P_TCLASS); 1600 break; 1601 1602 } 1603 break; 1604 1605 case IPV6_OTCLASS: 1606 { 1607 struct ip6_pktopts **optp; 1608 u_int8_t tclass; 1609 1610 error = sockopt_get(sopt, &tclass, sizeof(tclass)); 1611 if (error) 1612 break; 1613 optp = &in6p->in6p_outputopts; 1614 error = ip6_pcbopt(optname, 1615 (u_char *)&tclass, 1616 sizeof(tclass), 1617 optp, 1618 kauth_cred_get(), uproto); 1619 break; 1620 } 1621 1622 case IPV6_TCLASS: 1623 case IPV6_DONTFRAG: 1624 case IPV6_USE_MIN_MTU: 1625 case IPV6_PREFER_TEMPADDR: 1626 error = sockopt_getint(sopt, &optval); 1627 if (error) 1628 break; 1629 { 1630 struct ip6_pktopts **optp; 1631 optp = &in6p->in6p_outputopts; 1632 error = ip6_pcbopt(optname, 1633 (u_char *)&optval, 1634 sizeof(optval), 1635 optp, 1636 kauth_cred_get(), uproto); 1637 break; 1638 } 1639 1640 #ifdef RFC2292 1641 case IPV6_2292PKTINFO: 1642 case IPV6_2292HOPLIMIT: 1643 case IPV6_2292HOPOPTS: 1644 case IPV6_2292DSTOPTS: 1645 case IPV6_2292RTHDR: 1646 /* RFC 2292 */ 1647 error = sockopt_getint(sopt, &optval); 1648 if (error) 1649 break; 1650 1651 switch (optname) { 1652 case IPV6_2292PKTINFO: 1653 OPTSET2292(IN6P_PKTINFO); 1654 break; 1655 case IPV6_2292HOPLIMIT: 1656 OPTSET2292(IN6P_HOPLIMIT); 1657 break; 1658 case IPV6_2292HOPOPTS: 1659 /* 1660 * Check super-user privilege. 1661 * See comments for IPV6_RECVHOPOPTS. 1662 */ 1663 error = 1664 kauth_authorize_network(kauth_cred_get(), 1665 KAUTH_NETWORK_IPV6, 1666 KAUTH_REQ_NETWORK_IPV6_HOPBYHOP, NULL, 1667 NULL, NULL); 1668 if (error) 1669 return (error); 1670 OPTSET2292(IN6P_HOPOPTS); 1671 break; 1672 case IPV6_2292DSTOPTS: 1673 error = 1674 kauth_authorize_network(kauth_cred_get(), 1675 KAUTH_NETWORK_IPV6, 1676 KAUTH_REQ_NETWORK_IPV6_HOPBYHOP, NULL, 1677 NULL, NULL); 1678 if (error) 1679 return (error); 1680 OPTSET2292(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); /* XXX */ 1681 break; 1682 case IPV6_2292RTHDR: 1683 OPTSET2292(IN6P_RTHDR); 1684 break; 1685 } 1686 break; 1687 #endif 1688 case IPV6_PKTINFO: 1689 case IPV6_HOPOPTS: 1690 case IPV6_RTHDR: 1691 case IPV6_DSTOPTS: 1692 case IPV6_RTHDRDSTOPTS: 1693 case IPV6_NEXTHOP: { 1694 /* new advanced API (RFC3542) */ 1695 void *optbuf; 1696 int optbuflen; 1697 struct ip6_pktopts **optp; 1698 1699 #ifdef RFC2292 1700 /* cannot mix with RFC2292 */ 1701 if (OPTBIT(IN6P_RFC2292)) { 1702 error = EINVAL; 1703 break; 1704 } 1705 #endif 1706 1707 optbuflen = sopt->sopt_size; 1708 optbuf = malloc(optbuflen, M_IP6OPT, M_NOWAIT); 1709 if (optbuf == NULL) { 1710 error = ENOBUFS; 1711 break; 1712 } 1713 1714 error = sockopt_get(sopt, optbuf, optbuflen); 1715 if (error) { 1716 free(optbuf, M_IP6OPT); 1717 break; 1718 } 1719 optp = &in6p->in6p_outputopts; 1720 error = ip6_pcbopt(optname, optbuf, optbuflen, 1721 optp, kauth_cred_get(), uproto); 1722 1723 free(optbuf, M_IP6OPT); 1724 break; 1725 } 1726 #undef OPTSET 1727 1728 case IPV6_MULTICAST_IF: 1729 case IPV6_MULTICAST_HOPS: 1730 case IPV6_MULTICAST_LOOP: 1731 case IPV6_JOIN_GROUP: 1732 case IPV6_LEAVE_GROUP: 1733 error = ip6_setmoptions(sopt, in6p); 1734 break; 1735 1736 case IPV6_PORTRANGE: 1737 error = sockopt_getint(sopt, &optval); 1738 if (error) 1739 break; 1740 1741 switch (optval) { 1742 case IPV6_PORTRANGE_DEFAULT: 1743 in6p->in6p_flags &= ~(IN6P_LOWPORT); 1744 in6p->in6p_flags &= ~(IN6P_HIGHPORT); 1745 break; 1746 1747 case IPV6_PORTRANGE_HIGH: 1748 in6p->in6p_flags &= ~(IN6P_LOWPORT); 1749 in6p->in6p_flags |= IN6P_HIGHPORT; 1750 break; 1751 1752 case IPV6_PORTRANGE_LOW: 1753 in6p->in6p_flags &= ~(IN6P_HIGHPORT); 1754 in6p->in6p_flags |= IN6P_LOWPORT; 1755 break; 1756 1757 default: 1758 error = EINVAL; 1759 break; 1760 } 1761 break; 1762 1763 case IPV6_PORTALGO: 1764 error = sockopt_getint(sopt, &optval); 1765 if (error) 1766 break; 1767 1768 error = portalgo_algo_index_select( 1769 (struct inpcb_hdr *)in6p, optval); 1770 break; 1771 1772 #if defined(IPSEC) 1773 case IPV6_IPSEC_POLICY: 1774 if (ipsec_enabled) { 1775 error = ipsec_set_policy(in6p, optname, 1776 sopt->sopt_data, sopt->sopt_size, 1777 kauth_cred_get()); 1778 break; 1779 } 1780 /*FALLTHROUGH*/ 1781 #endif /* IPSEC */ 1782 1783 default: 1784 error = ENOPROTOOPT; 1785 break; 1786 } 1787 break; 1788 1789 case PRCO_GETOPT: 1790 switch (optname) { 1791 #ifdef RFC2292 1792 case IPV6_2292PKTOPTIONS: 1793 /* 1794 * RFC3542 (effectively) deprecated the 1795 * semantics of the 2292-style pktoptions. 1796 * Since it was not reliable in nature (i.e., 1797 * applications had to expect the lack of some 1798 * information after all), it would make sense 1799 * to simplify this part by always returning 1800 * empty data. 1801 */ 1802 break; 1803 #endif 1804 1805 case IPV6_RECVHOPOPTS: 1806 case IPV6_RECVDSTOPTS: 1807 case IPV6_RECVRTHDRDSTOPTS: 1808 case IPV6_UNICAST_HOPS: 1809 case IPV6_RECVPKTINFO: 1810 case IPV6_RECVHOPLIMIT: 1811 case IPV6_RECVRTHDR: 1812 case IPV6_RECVPATHMTU: 1813 1814 case IPV6_FAITH: 1815 case IPV6_V6ONLY: 1816 case IPV6_PORTRANGE: 1817 case IPV6_RECVTCLASS: 1818 switch (optname) { 1819 1820 case IPV6_RECVHOPOPTS: 1821 optval = OPTBIT(IN6P_HOPOPTS); 1822 break; 1823 1824 case IPV6_RECVDSTOPTS: 1825 optval = OPTBIT(IN6P_DSTOPTS); 1826 break; 1827 1828 case IPV6_RECVRTHDRDSTOPTS: 1829 optval = OPTBIT(IN6P_RTHDRDSTOPTS); 1830 break; 1831 1832 case IPV6_UNICAST_HOPS: 1833 optval = in6p->in6p_hops; 1834 break; 1835 1836 case IPV6_RECVPKTINFO: 1837 optval = OPTBIT(IN6P_PKTINFO); 1838 break; 1839 1840 case IPV6_RECVHOPLIMIT: 1841 optval = OPTBIT(IN6P_HOPLIMIT); 1842 break; 1843 1844 case IPV6_RECVRTHDR: 1845 optval = OPTBIT(IN6P_RTHDR); 1846 break; 1847 1848 case IPV6_RECVPATHMTU: 1849 optval = OPTBIT(IN6P_MTU); 1850 break; 1851 1852 case IPV6_FAITH: 1853 optval = OPTBIT(IN6P_FAITH); 1854 break; 1855 1856 case IPV6_V6ONLY: 1857 optval = OPTBIT(IN6P_IPV6_V6ONLY); 1858 break; 1859 1860 case IPV6_PORTRANGE: 1861 { 1862 int flags; 1863 flags = in6p->in6p_flags; 1864 if (flags & IN6P_HIGHPORT) 1865 optval = IPV6_PORTRANGE_HIGH; 1866 else if (flags & IN6P_LOWPORT) 1867 optval = IPV6_PORTRANGE_LOW; 1868 else 1869 optval = 0; 1870 break; 1871 } 1872 case IPV6_RECVTCLASS: 1873 optval = OPTBIT(IN6P_TCLASS); 1874 break; 1875 1876 } 1877 if (error) 1878 break; 1879 error = sockopt_setint(sopt, optval); 1880 break; 1881 1882 case IPV6_PATHMTU: 1883 { 1884 u_long pmtu = 0; 1885 struct ip6_mtuinfo mtuinfo; 1886 struct route *ro = &in6p->in6p_route; 1887 struct rtentry *rt; 1888 union { 1889 struct sockaddr dst; 1890 struct sockaddr_in6 dst6; 1891 } u; 1892 1893 if (!(so->so_state & SS_ISCONNECTED)) 1894 return (ENOTCONN); 1895 /* 1896 * XXX: we dot not consider the case of source 1897 * routing, or optional information to specify 1898 * the outgoing interface. 1899 */ 1900 sockaddr_in6_init(&u.dst6, &in6p->in6p_faddr, 0, 0, 0); 1901 rt = rtcache_lookup(ro, &u.dst); 1902 error = ip6_getpmtu(rt, NULL, &pmtu, NULL); 1903 rtcache_unref(rt, ro); 1904 if (error) 1905 break; 1906 if (pmtu > IPV6_MAXPACKET) 1907 pmtu = IPV6_MAXPACKET; 1908 1909 memset(&mtuinfo, 0, sizeof(mtuinfo)); 1910 mtuinfo.ip6m_mtu = (u_int32_t)pmtu; 1911 optdata = (void *)&mtuinfo; 1912 optdatalen = sizeof(mtuinfo); 1913 if (optdatalen > MCLBYTES) 1914 return (EMSGSIZE); /* XXX */ 1915 error = sockopt_set(sopt, optdata, optdatalen); 1916 break; 1917 } 1918 1919 #ifdef RFC2292 1920 case IPV6_2292PKTINFO: 1921 case IPV6_2292HOPLIMIT: 1922 case IPV6_2292HOPOPTS: 1923 case IPV6_2292RTHDR: 1924 case IPV6_2292DSTOPTS: 1925 switch (optname) { 1926 case IPV6_2292PKTINFO: 1927 optval = OPTBIT(IN6P_PKTINFO); 1928 break; 1929 case IPV6_2292HOPLIMIT: 1930 optval = OPTBIT(IN6P_HOPLIMIT); 1931 break; 1932 case IPV6_2292HOPOPTS: 1933 optval = OPTBIT(IN6P_HOPOPTS); 1934 break; 1935 case IPV6_2292RTHDR: 1936 optval = OPTBIT(IN6P_RTHDR); 1937 break; 1938 case IPV6_2292DSTOPTS: 1939 optval = OPTBIT(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); 1940 break; 1941 } 1942 error = sockopt_setint(sopt, optval); 1943 break; 1944 #endif 1945 case IPV6_PKTINFO: 1946 case IPV6_HOPOPTS: 1947 case IPV6_RTHDR: 1948 case IPV6_DSTOPTS: 1949 case IPV6_RTHDRDSTOPTS: 1950 case IPV6_NEXTHOP: 1951 case IPV6_OTCLASS: 1952 case IPV6_TCLASS: 1953 case IPV6_DONTFRAG: 1954 case IPV6_USE_MIN_MTU: 1955 case IPV6_PREFER_TEMPADDR: 1956 error = ip6_getpcbopt(in6p->in6p_outputopts, 1957 optname, sopt); 1958 break; 1959 1960 case IPV6_MULTICAST_IF: 1961 case IPV6_MULTICAST_HOPS: 1962 case IPV6_MULTICAST_LOOP: 1963 case IPV6_JOIN_GROUP: 1964 case IPV6_LEAVE_GROUP: 1965 error = ip6_getmoptions(sopt, in6p); 1966 break; 1967 1968 case IPV6_PORTALGO: 1969 optval = ((struct inpcb_hdr *)in6p)->inph_portalgo; 1970 error = sockopt_setint(sopt, optval); 1971 break; 1972 1973 #if defined(IPSEC) 1974 case IPV6_IPSEC_POLICY: 1975 if (ipsec_used) { 1976 struct mbuf *m = NULL; 1977 1978 /* 1979 * XXX: this will return EINVAL as sopt is 1980 * empty 1981 */ 1982 error = ipsec_get_policy(in6p, sopt->sopt_data, 1983 sopt->sopt_size, &m); 1984 if (!error) 1985 error = sockopt_setmbuf(sopt, m); 1986 break; 1987 } 1988 /*FALLTHROUGH*/ 1989 #endif /* IPSEC */ 1990 1991 default: 1992 error = ENOPROTOOPT; 1993 break; 1994 } 1995 break; 1996 } 1997 return (error); 1998 } 1999 2000 int 2001 ip6_raw_ctloutput(int op, struct socket *so, struct sockopt *sopt) 2002 { 2003 int error = 0, optval; 2004 const int icmp6off = offsetof(struct icmp6_hdr, icmp6_cksum); 2005 struct in6pcb *in6p = sotoin6pcb(so); 2006 int level, optname; 2007 2008 KASSERT(sopt != NULL); 2009 2010 level = sopt->sopt_level; 2011 optname = sopt->sopt_name; 2012 2013 if (level != IPPROTO_IPV6) { 2014 return ENOPROTOOPT; 2015 } 2016 2017 switch (optname) { 2018 case IPV6_CHECKSUM: 2019 /* 2020 * For ICMPv6 sockets, no modification allowed for checksum 2021 * offset, permit "no change" values to help existing apps. 2022 * 2023 * XXX RFC3542 says: "An attempt to set IPV6_CHECKSUM 2024 * for an ICMPv6 socket will fail." The current 2025 * behavior does not meet RFC3542. 2026 */ 2027 switch (op) { 2028 case PRCO_SETOPT: 2029 error = sockopt_getint(sopt, &optval); 2030 if (error) 2031 break; 2032 if ((optval % 2) != 0) { 2033 /* the API assumes even offset values */ 2034 error = EINVAL; 2035 } else if (so->so_proto->pr_protocol == 2036 IPPROTO_ICMPV6) { 2037 if (optval != icmp6off) 2038 error = EINVAL; 2039 } else 2040 in6p->in6p_cksum = optval; 2041 break; 2042 2043 case PRCO_GETOPT: 2044 if (so->so_proto->pr_protocol == IPPROTO_ICMPV6) 2045 optval = icmp6off; 2046 else 2047 optval = in6p->in6p_cksum; 2048 2049 error = sockopt_setint(sopt, optval); 2050 break; 2051 2052 default: 2053 error = EINVAL; 2054 break; 2055 } 2056 break; 2057 2058 default: 2059 error = ENOPROTOOPT; 2060 break; 2061 } 2062 2063 return (error); 2064 } 2065 2066 #ifdef RFC2292 2067 /* 2068 * Set up IP6 options in pcb for insertion in output packets or 2069 * specifying behavior of outgoing packets. 2070 */ 2071 static int 2072 ip6_pcbopts(struct ip6_pktopts **pktopt, struct socket *so, 2073 struct sockopt *sopt) 2074 { 2075 struct ip6_pktopts *opt = *pktopt; 2076 struct mbuf *m; 2077 int error = 0; 2078 2079 KASSERT(solocked(so)); 2080 2081 /* turn off any old options. */ 2082 if (opt) { 2083 #ifdef DIAGNOSTIC 2084 if (opt->ip6po_pktinfo || opt->ip6po_nexthop || 2085 opt->ip6po_hbh || opt->ip6po_dest1 || opt->ip6po_dest2 || 2086 opt->ip6po_rhinfo.ip6po_rhi_rthdr) 2087 printf("ip6_pcbopts: all specified options are cleared.\n"); 2088 #endif 2089 ip6_clearpktopts(opt, -1); 2090 } else { 2091 opt = malloc(sizeof(*opt), M_IP6OPT, M_NOWAIT); 2092 if (opt == NULL) 2093 return (ENOBUFS); 2094 } 2095 *pktopt = NULL; 2096 2097 if (sopt == NULL || sopt->sopt_size == 0) { 2098 /* 2099 * Only turning off any previous options, regardless of 2100 * whether the opt is just created or given. 2101 */ 2102 free(opt, M_IP6OPT); 2103 return (0); 2104 } 2105 2106 /* set options specified by user. */ 2107 m = sockopt_getmbuf(sopt); 2108 if (m == NULL) { 2109 free(opt, M_IP6OPT); 2110 return (ENOBUFS); 2111 } 2112 2113 error = ip6_setpktopts(m, opt, NULL, kauth_cred_get(), 2114 so->so_proto->pr_protocol); 2115 m_freem(m); 2116 if (error != 0) { 2117 ip6_clearpktopts(opt, -1); /* XXX: discard all options */ 2118 free(opt, M_IP6OPT); 2119 return (error); 2120 } 2121 *pktopt = opt; 2122 return (0); 2123 } 2124 #endif 2125 2126 /* 2127 * initialize ip6_pktopts. beware that there are non-zero default values in 2128 * the struct. 2129 */ 2130 void 2131 ip6_initpktopts(struct ip6_pktopts *opt) 2132 { 2133 2134 memset(opt, 0, sizeof(*opt)); 2135 opt->ip6po_hlim = -1; /* -1 means default hop limit */ 2136 opt->ip6po_tclass = -1; /* -1 means default traffic class */ 2137 opt->ip6po_minmtu = IP6PO_MINMTU_MCASTONLY; 2138 opt->ip6po_prefer_tempaddr = IP6PO_TEMPADDR_SYSTEM; 2139 } 2140 2141 #define sin6tosa(sin6) ((struct sockaddr *)(sin6)) /* XXX */ 2142 static int 2143 ip6_pcbopt(int optname, u_char *buf, int len, struct ip6_pktopts **pktopt, 2144 kauth_cred_t cred, int uproto) 2145 { 2146 struct ip6_pktopts *opt; 2147 2148 if (*pktopt == NULL) { 2149 *pktopt = malloc(sizeof(struct ip6_pktopts), M_IP6OPT, 2150 M_NOWAIT); 2151 if (*pktopt == NULL) 2152 return (ENOBUFS); 2153 2154 ip6_initpktopts(*pktopt); 2155 } 2156 opt = *pktopt; 2157 2158 return (ip6_setpktopt(optname, buf, len, opt, cred, 1, 0, uproto)); 2159 } 2160 2161 static int 2162 ip6_getpcbopt(struct ip6_pktopts *pktopt, int optname, struct sockopt *sopt) 2163 { 2164 void *optdata = NULL; 2165 int optdatalen = 0; 2166 struct ip6_ext *ip6e; 2167 int error = 0; 2168 struct in6_pktinfo null_pktinfo; 2169 int deftclass = 0, on; 2170 int defminmtu = IP6PO_MINMTU_MCASTONLY; 2171 int defpreftemp = IP6PO_TEMPADDR_SYSTEM; 2172 2173 switch (optname) { 2174 case IPV6_PKTINFO: 2175 if (pktopt && pktopt->ip6po_pktinfo) 2176 optdata = (void *)pktopt->ip6po_pktinfo; 2177 else { 2178 /* XXX: we don't have to do this every time... */ 2179 memset(&null_pktinfo, 0, sizeof(null_pktinfo)); 2180 optdata = (void *)&null_pktinfo; 2181 } 2182 optdatalen = sizeof(struct in6_pktinfo); 2183 break; 2184 case IPV6_OTCLASS: 2185 /* XXX */ 2186 return (EINVAL); 2187 case IPV6_TCLASS: 2188 if (pktopt && pktopt->ip6po_tclass >= 0) 2189 optdata = (void *)&pktopt->ip6po_tclass; 2190 else 2191 optdata = (void *)&deftclass; 2192 optdatalen = sizeof(int); 2193 break; 2194 case IPV6_HOPOPTS: 2195 if (pktopt && pktopt->ip6po_hbh) { 2196 optdata = (void *)pktopt->ip6po_hbh; 2197 ip6e = (struct ip6_ext *)pktopt->ip6po_hbh; 2198 optdatalen = (ip6e->ip6e_len + 1) << 3; 2199 } 2200 break; 2201 case IPV6_RTHDR: 2202 if (pktopt && pktopt->ip6po_rthdr) { 2203 optdata = (void *)pktopt->ip6po_rthdr; 2204 ip6e = (struct ip6_ext *)pktopt->ip6po_rthdr; 2205 optdatalen = (ip6e->ip6e_len + 1) << 3; 2206 } 2207 break; 2208 case IPV6_RTHDRDSTOPTS: 2209 if (pktopt && pktopt->ip6po_dest1) { 2210 optdata = (void *)pktopt->ip6po_dest1; 2211 ip6e = (struct ip6_ext *)pktopt->ip6po_dest1; 2212 optdatalen = (ip6e->ip6e_len + 1) << 3; 2213 } 2214 break; 2215 case IPV6_DSTOPTS: 2216 if (pktopt && pktopt->ip6po_dest2) { 2217 optdata = (void *)pktopt->ip6po_dest2; 2218 ip6e = (struct ip6_ext *)pktopt->ip6po_dest2; 2219 optdatalen = (ip6e->ip6e_len + 1) << 3; 2220 } 2221 break; 2222 case IPV6_NEXTHOP: 2223 if (pktopt && pktopt->ip6po_nexthop) { 2224 optdata = (void *)pktopt->ip6po_nexthop; 2225 optdatalen = pktopt->ip6po_nexthop->sa_len; 2226 } 2227 break; 2228 case IPV6_USE_MIN_MTU: 2229 if (pktopt) 2230 optdata = (void *)&pktopt->ip6po_minmtu; 2231 else 2232 optdata = (void *)&defminmtu; 2233 optdatalen = sizeof(int); 2234 break; 2235 case IPV6_DONTFRAG: 2236 if (pktopt && ((pktopt->ip6po_flags) & IP6PO_DONTFRAG)) 2237 on = 1; 2238 else 2239 on = 0; 2240 optdata = (void *)&on; 2241 optdatalen = sizeof(on); 2242 break; 2243 case IPV6_PREFER_TEMPADDR: 2244 if (pktopt) 2245 optdata = (void *)&pktopt->ip6po_prefer_tempaddr; 2246 else 2247 optdata = (void *)&defpreftemp; 2248 optdatalen = sizeof(int); 2249 break; 2250 default: /* should not happen */ 2251 #ifdef DIAGNOSTIC 2252 panic("ip6_getpcbopt: unexpected option\n"); 2253 #endif 2254 return (ENOPROTOOPT); 2255 } 2256 2257 error = sockopt_set(sopt, optdata, optdatalen); 2258 2259 return (error); 2260 } 2261 2262 void 2263 ip6_clearpktopts(struct ip6_pktopts *pktopt, int optname) 2264 { 2265 if (optname == -1 || optname == IPV6_PKTINFO) { 2266 if (pktopt->ip6po_pktinfo) 2267 free(pktopt->ip6po_pktinfo, M_IP6OPT); 2268 pktopt->ip6po_pktinfo = NULL; 2269 } 2270 if (optname == -1 || optname == IPV6_HOPLIMIT) 2271 pktopt->ip6po_hlim = -1; 2272 if (optname == -1 || optname == IPV6_TCLASS) 2273 pktopt->ip6po_tclass = -1; 2274 if (optname == -1 || optname == IPV6_NEXTHOP) { 2275 rtcache_free(&pktopt->ip6po_nextroute); 2276 if (pktopt->ip6po_nexthop) 2277 free(pktopt->ip6po_nexthop, M_IP6OPT); 2278 pktopt->ip6po_nexthop = NULL; 2279 } 2280 if (optname == -1 || optname == IPV6_HOPOPTS) { 2281 if (pktopt->ip6po_hbh) 2282 free(pktopt->ip6po_hbh, M_IP6OPT); 2283 pktopt->ip6po_hbh = NULL; 2284 } 2285 if (optname == -1 || optname == IPV6_RTHDRDSTOPTS) { 2286 if (pktopt->ip6po_dest1) 2287 free(pktopt->ip6po_dest1, M_IP6OPT); 2288 pktopt->ip6po_dest1 = NULL; 2289 } 2290 if (optname == -1 || optname == IPV6_RTHDR) { 2291 if (pktopt->ip6po_rhinfo.ip6po_rhi_rthdr) 2292 free(pktopt->ip6po_rhinfo.ip6po_rhi_rthdr, M_IP6OPT); 2293 pktopt->ip6po_rhinfo.ip6po_rhi_rthdr = NULL; 2294 rtcache_free(&pktopt->ip6po_route); 2295 } 2296 if (optname == -1 || optname == IPV6_DSTOPTS) { 2297 if (pktopt->ip6po_dest2) 2298 free(pktopt->ip6po_dest2, M_IP6OPT); 2299 pktopt->ip6po_dest2 = NULL; 2300 } 2301 } 2302 2303 #define PKTOPT_EXTHDRCPY(type) \ 2304 do { \ 2305 if (src->type) { \ 2306 int hlen = (((struct ip6_ext *)src->type)->ip6e_len + 1) << 3;\ 2307 dst->type = malloc(hlen, M_IP6OPT, canwait); \ 2308 if (dst->type == NULL) \ 2309 goto bad; \ 2310 memcpy(dst->type, src->type, hlen); \ 2311 } \ 2312 } while (/*CONSTCOND*/ 0) 2313 2314 static int 2315 copypktopts(struct ip6_pktopts *dst, struct ip6_pktopts *src, int canwait) 2316 { 2317 dst->ip6po_hlim = src->ip6po_hlim; 2318 dst->ip6po_tclass = src->ip6po_tclass; 2319 dst->ip6po_flags = src->ip6po_flags; 2320 dst->ip6po_minmtu = src->ip6po_minmtu; 2321 dst->ip6po_prefer_tempaddr = src->ip6po_prefer_tempaddr; 2322 if (src->ip6po_pktinfo) { 2323 dst->ip6po_pktinfo = malloc(sizeof(*dst->ip6po_pktinfo), 2324 M_IP6OPT, canwait); 2325 if (dst->ip6po_pktinfo == NULL) 2326 goto bad; 2327 *dst->ip6po_pktinfo = *src->ip6po_pktinfo; 2328 } 2329 if (src->ip6po_nexthop) { 2330 dst->ip6po_nexthop = malloc(src->ip6po_nexthop->sa_len, 2331 M_IP6OPT, canwait); 2332 if (dst->ip6po_nexthop == NULL) 2333 goto bad; 2334 memcpy(dst->ip6po_nexthop, src->ip6po_nexthop, 2335 src->ip6po_nexthop->sa_len); 2336 } 2337 PKTOPT_EXTHDRCPY(ip6po_hbh); 2338 PKTOPT_EXTHDRCPY(ip6po_dest1); 2339 PKTOPT_EXTHDRCPY(ip6po_dest2); 2340 PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */ 2341 return (0); 2342 2343 bad: 2344 if (dst->ip6po_pktinfo) free(dst->ip6po_pktinfo, M_IP6OPT); 2345 if (dst->ip6po_nexthop) free(dst->ip6po_nexthop, M_IP6OPT); 2346 if (dst->ip6po_hbh) free(dst->ip6po_hbh, M_IP6OPT); 2347 if (dst->ip6po_dest1) free(dst->ip6po_dest1, M_IP6OPT); 2348 if (dst->ip6po_dest2) free(dst->ip6po_dest2, M_IP6OPT); 2349 if (dst->ip6po_rthdr) free(dst->ip6po_rthdr, M_IP6OPT); 2350 2351 return (ENOBUFS); 2352 } 2353 #undef PKTOPT_EXTHDRCPY 2354 2355 struct ip6_pktopts * 2356 ip6_copypktopts(struct ip6_pktopts *src, int canwait) 2357 { 2358 int error; 2359 struct ip6_pktopts *dst; 2360 2361 dst = malloc(sizeof(*dst), M_IP6OPT, canwait); 2362 if (dst == NULL) 2363 return (NULL); 2364 ip6_initpktopts(dst); 2365 2366 if ((error = copypktopts(dst, src, canwait)) != 0) { 2367 free(dst, M_IP6OPT); 2368 return (NULL); 2369 } 2370 2371 return (dst); 2372 } 2373 2374 void 2375 ip6_freepcbopts(struct ip6_pktopts *pktopt) 2376 { 2377 if (pktopt == NULL) 2378 return; 2379 2380 ip6_clearpktopts(pktopt, -1); 2381 2382 free(pktopt, M_IP6OPT); 2383 } 2384 2385 int 2386 ip6_get_membership(const struct sockopt *sopt, struct ifnet **ifp, 2387 struct psref *psref, void *v, size_t l) 2388 { 2389 struct ipv6_mreq mreq; 2390 int error; 2391 struct in6_addr *ia = &mreq.ipv6mr_multiaddr; 2392 struct in_addr *ia4 = (void *)&ia->s6_addr32[3]; 2393 2394 error = sockopt_get(sopt, &mreq, sizeof(mreq)); 2395 if (error != 0) 2396 return error; 2397 2398 if (IN6_IS_ADDR_UNSPECIFIED(ia)) { 2399 /* 2400 * We use the unspecified address to specify to accept 2401 * all multicast addresses. Only super user is allowed 2402 * to do this. 2403 */ 2404 if (kauth_authorize_network(curlwp->l_cred, KAUTH_NETWORK_IPV6, 2405 KAUTH_REQ_NETWORK_IPV6_JOIN_MULTICAST, NULL, NULL, NULL)) 2406 return EACCES; 2407 } else if (IN6_IS_ADDR_V4MAPPED(ia)) { 2408 // Don't bother if we are not going to use ifp. 2409 if (l == sizeof(*ia)) { 2410 memcpy(v, ia, l); 2411 return 0; 2412 } 2413 } else if (!IN6_IS_ADDR_MULTICAST(ia)) { 2414 return EINVAL; 2415 } 2416 2417 /* 2418 * If no interface was explicitly specified, choose an 2419 * appropriate one according to the given multicast address. 2420 */ 2421 if (mreq.ipv6mr_interface == 0) { 2422 struct rtentry *rt; 2423 union { 2424 struct sockaddr dst; 2425 struct sockaddr_in dst4; 2426 struct sockaddr_in6 dst6; 2427 } u; 2428 struct route ro; 2429 2430 /* 2431 * Look up the routing table for the 2432 * address, and choose the outgoing interface. 2433 * XXX: is it a good approach? 2434 */ 2435 memset(&ro, 0, sizeof(ro)); 2436 if (IN6_IS_ADDR_V4MAPPED(ia)) 2437 sockaddr_in_init(&u.dst4, ia4, 0); 2438 else 2439 sockaddr_in6_init(&u.dst6, ia, 0, 0, 0); 2440 error = rtcache_setdst(&ro, &u.dst); 2441 if (error != 0) 2442 return error; 2443 rt = rtcache_init(&ro); 2444 *ifp = rt != NULL ? 2445 if_get_byindex(rt->rt_ifp->if_index, psref) : NULL; 2446 rtcache_unref(rt, &ro); 2447 rtcache_free(&ro); 2448 } else { 2449 /* 2450 * If the interface is specified, validate it. 2451 */ 2452 *ifp = if_get_byindex(mreq.ipv6mr_interface, psref); 2453 if (*ifp == NULL) 2454 return ENXIO; /* XXX EINVAL? */ 2455 } 2456 if (sizeof(*ia) == l) 2457 memcpy(v, ia, l); 2458 else 2459 memcpy(v, ia4, l); 2460 return 0; 2461 } 2462 2463 /* 2464 * Set the IP6 multicast options in response to user setsockopt(). 2465 */ 2466 static int 2467 ip6_setmoptions(const struct sockopt *sopt, struct in6pcb *in6p) 2468 { 2469 int error = 0; 2470 u_int loop, ifindex; 2471 struct ipv6_mreq mreq; 2472 struct in6_addr ia; 2473 struct ifnet *ifp; 2474 struct ip6_moptions *im6o = in6p->in6p_moptions; 2475 struct in6_multi_mship *imm; 2476 2477 KASSERT(in6p_locked(in6p)); 2478 2479 if (im6o == NULL) { 2480 /* 2481 * No multicast option buffer attached to the pcb; 2482 * allocate one and initialize to default values. 2483 */ 2484 im6o = malloc(sizeof(*im6o), M_IPMOPTS, M_NOWAIT); 2485 if (im6o == NULL) 2486 return (ENOBUFS); 2487 in6p->in6p_moptions = im6o; 2488 im6o->im6o_multicast_if_index = 0; 2489 im6o->im6o_multicast_hlim = ip6_defmcasthlim; 2490 im6o->im6o_multicast_loop = IPV6_DEFAULT_MULTICAST_LOOP; 2491 LIST_INIT(&im6o->im6o_memberships); 2492 } 2493 2494 switch (sopt->sopt_name) { 2495 2496 case IPV6_MULTICAST_IF: { 2497 int s; 2498 /* 2499 * Select the interface for outgoing multicast packets. 2500 */ 2501 error = sockopt_get(sopt, &ifindex, sizeof(ifindex)); 2502 if (error != 0) 2503 break; 2504 2505 s = pserialize_read_enter(); 2506 if (ifindex != 0) { 2507 if ((ifp = if_byindex(ifindex)) == NULL) { 2508 pserialize_read_exit(s); 2509 error = ENXIO; /* XXX EINVAL? */ 2510 break; 2511 } 2512 if ((ifp->if_flags & IFF_MULTICAST) == 0) { 2513 pserialize_read_exit(s); 2514 error = EADDRNOTAVAIL; 2515 break; 2516 } 2517 } else 2518 ifp = NULL; 2519 im6o->im6o_multicast_if_index = if_get_index(ifp); 2520 pserialize_read_exit(s); 2521 break; 2522 } 2523 2524 case IPV6_MULTICAST_HOPS: 2525 { 2526 /* 2527 * Set the IP6 hoplimit for outgoing multicast packets. 2528 */ 2529 int optval; 2530 2531 error = sockopt_getint(sopt, &optval); 2532 if (error != 0) 2533 break; 2534 2535 if (optval < -1 || optval >= 256) 2536 error = EINVAL; 2537 else if (optval == -1) 2538 im6o->im6o_multicast_hlim = ip6_defmcasthlim; 2539 else 2540 im6o->im6o_multicast_hlim = optval; 2541 break; 2542 } 2543 2544 case IPV6_MULTICAST_LOOP: 2545 /* 2546 * Set the loopback flag for outgoing multicast packets. 2547 * Must be zero or one. 2548 */ 2549 error = sockopt_get(sopt, &loop, sizeof(loop)); 2550 if (error != 0) 2551 break; 2552 if (loop > 1) { 2553 error = EINVAL; 2554 break; 2555 } 2556 im6o->im6o_multicast_loop = loop; 2557 break; 2558 2559 case IPV6_JOIN_GROUP: { 2560 int bound; 2561 struct psref psref; 2562 /* 2563 * Add a multicast group membership. 2564 * Group must be a valid IP6 multicast address. 2565 */ 2566 bound = curlwp_bind(); 2567 ifp = NULL; 2568 error = ip6_get_membership(sopt, &ifp, &psref, &ia, sizeof(ia)); 2569 if (error != 0) { 2570 KASSERT(ifp == NULL); 2571 curlwp_bindx(bound); 2572 return error; 2573 } 2574 2575 if (IN6_IS_ADDR_V4MAPPED(&ia)) { 2576 error = ip_setmoptions(&in6p->in6p_v4moptions, sopt); 2577 goto put_break; 2578 } 2579 /* 2580 * See if we found an interface, and confirm that it 2581 * supports multicast 2582 */ 2583 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) { 2584 error = EADDRNOTAVAIL; 2585 goto put_break; 2586 } 2587 2588 if (in6_setscope(&ia, ifp, NULL)) { 2589 error = EADDRNOTAVAIL; /* XXX: should not happen */ 2590 goto put_break; 2591 } 2592 2593 /* 2594 * See if the membership already exists. 2595 */ 2596 LIST_FOREACH(imm, &im6o->im6o_memberships, i6mm_chain) { 2597 if (imm->i6mm_maddr->in6m_ifp == ifp && 2598 IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr, 2599 &ia)) 2600 goto put_break; 2601 } 2602 if (imm != NULL) { 2603 error = EADDRINUSE; 2604 goto put_break; 2605 } 2606 /* 2607 * Everything looks good; add a new record to the multicast 2608 * address list for the given interface. 2609 */ 2610 IFNET_LOCK(ifp); 2611 imm = in6_joingroup(ifp, &ia, &error, 0); 2612 IFNET_UNLOCK(ifp); 2613 if (imm == NULL) 2614 goto put_break; 2615 LIST_INSERT_HEAD(&im6o->im6o_memberships, imm, i6mm_chain); 2616 put_break: 2617 if_put(ifp, &psref); 2618 curlwp_bindx(bound); 2619 break; 2620 } 2621 2622 case IPV6_LEAVE_GROUP: { 2623 struct ifnet *in6m_ifp; 2624 /* 2625 * Drop a multicast group membership. 2626 * Group must be a valid IP6 multicast address. 2627 */ 2628 error = sockopt_get(sopt, &mreq, sizeof(mreq)); 2629 if (error != 0) 2630 break; 2631 2632 if (IN6_IS_ADDR_V4MAPPED(&mreq.ipv6mr_multiaddr)) { 2633 error = ip_setmoptions(&in6p->in6p_v4moptions, sopt); 2634 break; 2635 } 2636 /* 2637 * If an interface address was specified, get a pointer 2638 * to its ifnet structure. 2639 */ 2640 if (mreq.ipv6mr_interface != 0) { 2641 if ((ifp = if_byindex(mreq.ipv6mr_interface)) == NULL) { 2642 error = ENXIO; /* XXX EINVAL? */ 2643 break; 2644 } 2645 } else 2646 ifp = NULL; 2647 2648 /* Fill in the scope zone ID */ 2649 if (ifp) { 2650 if (in6_setscope(&mreq.ipv6mr_multiaddr, ifp, NULL)) { 2651 /* XXX: should not happen */ 2652 error = EADDRNOTAVAIL; 2653 break; 2654 } 2655 } else if (mreq.ipv6mr_interface != 0) { 2656 /* 2657 * XXX: This case would happens when the (positive) 2658 * index is in the valid range, but the corresponding 2659 * interface has been detached dynamically. The above 2660 * check probably avoids such case to happen here, but 2661 * we check it explicitly for safety. 2662 */ 2663 error = EADDRNOTAVAIL; 2664 break; 2665 } else { /* ipv6mr_interface == 0 */ 2666 struct sockaddr_in6 sa6_mc; 2667 2668 /* 2669 * The API spec says as follows: 2670 * If the interface index is specified as 0, the 2671 * system may choose a multicast group membership to 2672 * drop by matching the multicast address only. 2673 * On the other hand, we cannot disambiguate the scope 2674 * zone unless an interface is provided. Thus, we 2675 * check if there's ambiguity with the default scope 2676 * zone as the last resort. 2677 */ 2678 sockaddr_in6_init(&sa6_mc, &mreq.ipv6mr_multiaddr, 2679 0, 0, 0); 2680 error = sa6_embedscope(&sa6_mc, ip6_use_defzone); 2681 if (error != 0) 2682 break; 2683 mreq.ipv6mr_multiaddr = sa6_mc.sin6_addr; 2684 } 2685 2686 /* 2687 * Find the membership in the membership list. 2688 */ 2689 LIST_FOREACH(imm, &im6o->im6o_memberships, i6mm_chain) { 2690 if ((ifp == NULL || imm->i6mm_maddr->in6m_ifp == ifp) && 2691 IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr, 2692 &mreq.ipv6mr_multiaddr)) 2693 break; 2694 } 2695 if (imm == NULL) { 2696 /* Unable to resolve interface */ 2697 error = EADDRNOTAVAIL; 2698 break; 2699 } 2700 /* 2701 * Give up the multicast address record to which the 2702 * membership points. 2703 */ 2704 LIST_REMOVE(imm, i6mm_chain); 2705 in6m_ifp = imm->i6mm_maddr->in6m_ifp; 2706 IFNET_LOCK(in6m_ifp); 2707 in6_leavegroup(imm); 2708 /* in6m_ifp should not leave thanks to in6p_lock */ 2709 IFNET_UNLOCK(in6m_ifp); 2710 break; 2711 } 2712 2713 default: 2714 error = EOPNOTSUPP; 2715 break; 2716 } 2717 2718 /* 2719 * If all options have default values, no need to keep the mbuf. 2720 */ 2721 if (im6o->im6o_multicast_if_index == 0 && 2722 im6o->im6o_multicast_hlim == ip6_defmcasthlim && 2723 im6o->im6o_multicast_loop == IPV6_DEFAULT_MULTICAST_LOOP && 2724 LIST_EMPTY(&im6o->im6o_memberships)) { 2725 free(in6p->in6p_moptions, M_IPMOPTS); 2726 in6p->in6p_moptions = NULL; 2727 } 2728 2729 return (error); 2730 } 2731 2732 /* 2733 * Return the IP6 multicast options in response to user getsockopt(). 2734 */ 2735 static int 2736 ip6_getmoptions(struct sockopt *sopt, struct in6pcb *in6p) 2737 { 2738 u_int optval; 2739 int error; 2740 struct ip6_moptions *im6o = in6p->in6p_moptions; 2741 2742 switch (sopt->sopt_name) { 2743 case IPV6_MULTICAST_IF: 2744 if (im6o == NULL || im6o->im6o_multicast_if_index == 0) 2745 optval = 0; 2746 else 2747 optval = im6o->im6o_multicast_if_index; 2748 2749 error = sockopt_set(sopt, &optval, sizeof(optval)); 2750 break; 2751 2752 case IPV6_MULTICAST_HOPS: 2753 if (im6o == NULL) 2754 optval = ip6_defmcasthlim; 2755 else 2756 optval = im6o->im6o_multicast_hlim; 2757 2758 error = sockopt_set(sopt, &optval, sizeof(optval)); 2759 break; 2760 2761 case IPV6_MULTICAST_LOOP: 2762 if (im6o == NULL) 2763 optval = IPV6_DEFAULT_MULTICAST_LOOP; 2764 else 2765 optval = im6o->im6o_multicast_loop; 2766 2767 error = sockopt_set(sopt, &optval, sizeof(optval)); 2768 break; 2769 2770 default: 2771 error = EOPNOTSUPP; 2772 } 2773 2774 return (error); 2775 } 2776 2777 /* 2778 * Discard the IP6 multicast options. 2779 */ 2780 void 2781 ip6_freemoptions(struct ip6_moptions *im6o) 2782 { 2783 struct in6_multi_mship *imm, *nimm; 2784 2785 if (im6o == NULL) 2786 return; 2787 2788 /* The owner of im6o (in6p) should be protected by solock */ 2789 LIST_FOREACH_SAFE(imm, &im6o->im6o_memberships, i6mm_chain, nimm) { 2790 struct ifnet *ifp; 2791 2792 LIST_REMOVE(imm, i6mm_chain); 2793 2794 ifp = imm->i6mm_maddr->in6m_ifp; 2795 IFNET_LOCK(ifp); 2796 in6_leavegroup(imm); 2797 /* ifp should not leave thanks to solock */ 2798 IFNET_UNLOCK(ifp); 2799 } 2800 free(im6o, M_IPMOPTS); 2801 } 2802 2803 /* 2804 * Set IPv6 outgoing packet options based on advanced API. 2805 */ 2806 int 2807 ip6_setpktopts(struct mbuf *control, struct ip6_pktopts *opt, 2808 struct ip6_pktopts *stickyopt, kauth_cred_t cred, int uproto) 2809 { 2810 struct cmsghdr *cm = 0; 2811 2812 if (control == NULL || opt == NULL) 2813 return (EINVAL); 2814 2815 ip6_initpktopts(opt); 2816 if (stickyopt) { 2817 int error; 2818 2819 /* 2820 * If stickyopt is provided, make a local copy of the options 2821 * for this particular packet, then override them by ancillary 2822 * objects. 2823 * XXX: copypktopts() does not copy the cached route to a next 2824 * hop (if any). This is not very good in terms of efficiency, 2825 * but we can allow this since this option should be rarely 2826 * used. 2827 */ 2828 if ((error = copypktopts(opt, stickyopt, M_NOWAIT)) != 0) 2829 return (error); 2830 } 2831 2832 /* 2833 * XXX: Currently, we assume all the optional information is stored 2834 * in a single mbuf. 2835 */ 2836 if (control->m_next) 2837 return (EINVAL); 2838 2839 /* XXX if cm->cmsg_len is not aligned, control->m_len can become <0 */ 2840 for (; control->m_len > 0; control->m_data += CMSG_ALIGN(cm->cmsg_len), 2841 control->m_len -= CMSG_ALIGN(cm->cmsg_len)) { 2842 int error; 2843 2844 if (control->m_len < CMSG_LEN(0)) 2845 return (EINVAL); 2846 2847 cm = mtod(control, struct cmsghdr *); 2848 if (cm->cmsg_len == 0 || cm->cmsg_len > control->m_len) 2849 return (EINVAL); 2850 if (cm->cmsg_level != IPPROTO_IPV6) 2851 continue; 2852 2853 error = ip6_setpktopt(cm->cmsg_type, CMSG_DATA(cm), 2854 cm->cmsg_len - CMSG_LEN(0), opt, cred, 0, 1, uproto); 2855 if (error) 2856 return (error); 2857 } 2858 2859 return (0); 2860 } 2861 2862 /* 2863 * Set a particular packet option, as a sticky option or an ancillary data 2864 * item. "len" can be 0 only when it's a sticky option. 2865 * We have 4 cases of combination of "sticky" and "cmsg": 2866 * "sticky=0, cmsg=0": impossible 2867 * "sticky=0, cmsg=1": RFC2292 or RFC3542 ancillary data 2868 * "sticky=1, cmsg=0": RFC3542 socket option 2869 * "sticky=1, cmsg=1": RFC2292 socket option 2870 */ 2871 static int 2872 ip6_setpktopt(int optname, u_char *buf, int len, struct ip6_pktopts *opt, 2873 kauth_cred_t cred, int sticky, int cmsg, int uproto) 2874 { 2875 int minmtupolicy; 2876 int error; 2877 2878 if (!sticky && !cmsg) { 2879 #ifdef DIAGNOSTIC 2880 printf("ip6_setpktopt: impossible case\n"); 2881 #endif 2882 return (EINVAL); 2883 } 2884 2885 /* 2886 * IPV6_2292xxx is for backward compatibility to RFC2292, and should 2887 * not be specified in the context of RFC3542. Conversely, 2888 * RFC3542 types should not be specified in the context of RFC2292. 2889 */ 2890 if (!cmsg) { 2891 switch (optname) { 2892 case IPV6_2292PKTINFO: 2893 case IPV6_2292HOPLIMIT: 2894 case IPV6_2292NEXTHOP: 2895 case IPV6_2292HOPOPTS: 2896 case IPV6_2292DSTOPTS: 2897 case IPV6_2292RTHDR: 2898 case IPV6_2292PKTOPTIONS: 2899 return (ENOPROTOOPT); 2900 } 2901 } 2902 if (sticky && cmsg) { 2903 switch (optname) { 2904 case IPV6_PKTINFO: 2905 case IPV6_HOPLIMIT: 2906 case IPV6_NEXTHOP: 2907 case IPV6_HOPOPTS: 2908 case IPV6_DSTOPTS: 2909 case IPV6_RTHDRDSTOPTS: 2910 case IPV6_RTHDR: 2911 case IPV6_USE_MIN_MTU: 2912 case IPV6_DONTFRAG: 2913 case IPV6_OTCLASS: 2914 case IPV6_TCLASS: 2915 case IPV6_PREFER_TEMPADDR: /* XXX not an RFC3542 option */ 2916 return (ENOPROTOOPT); 2917 } 2918 } 2919 2920 switch (optname) { 2921 #ifdef RFC2292 2922 case IPV6_2292PKTINFO: 2923 #endif 2924 case IPV6_PKTINFO: 2925 { 2926 struct in6_pktinfo *pktinfo; 2927 2928 if (len != sizeof(struct in6_pktinfo)) 2929 return (EINVAL); 2930 2931 pktinfo = (struct in6_pktinfo *)buf; 2932 2933 /* 2934 * An application can clear any sticky IPV6_PKTINFO option by 2935 * doing a "regular" setsockopt with ipi6_addr being 2936 * in6addr_any and ipi6_ifindex being zero. 2937 * [RFC 3542, Section 6] 2938 */ 2939 if (optname == IPV6_PKTINFO && opt->ip6po_pktinfo && 2940 pktinfo->ipi6_ifindex == 0 && 2941 IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) { 2942 ip6_clearpktopts(opt, optname); 2943 break; 2944 } 2945 2946 if (uproto == IPPROTO_TCP && optname == IPV6_PKTINFO && 2947 sticky && !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) { 2948 return (EINVAL); 2949 } 2950 2951 /* Validate the interface index if specified. */ 2952 if (pktinfo->ipi6_ifindex) { 2953 struct ifnet *ifp; 2954 int s = pserialize_read_enter(); 2955 ifp = if_byindex(pktinfo->ipi6_ifindex); 2956 if (ifp == NULL) { 2957 pserialize_read_exit(s); 2958 return ENXIO; 2959 } 2960 pserialize_read_exit(s); 2961 } 2962 2963 /* 2964 * We store the address anyway, and let in6_selectsrc() 2965 * validate the specified address. This is because ipi6_addr 2966 * may not have enough information about its scope zone, and 2967 * we may need additional information (such as outgoing 2968 * interface or the scope zone of a destination address) to 2969 * disambiguate the scope. 2970 * XXX: the delay of the validation may confuse the 2971 * application when it is used as a sticky option. 2972 */ 2973 if (opt->ip6po_pktinfo == NULL) { 2974 opt->ip6po_pktinfo = malloc(sizeof(*pktinfo), 2975 M_IP6OPT, M_NOWAIT); 2976 if (opt->ip6po_pktinfo == NULL) 2977 return (ENOBUFS); 2978 } 2979 memcpy(opt->ip6po_pktinfo, pktinfo, sizeof(*pktinfo)); 2980 break; 2981 } 2982 2983 #ifdef RFC2292 2984 case IPV6_2292HOPLIMIT: 2985 #endif 2986 case IPV6_HOPLIMIT: 2987 { 2988 int *hlimp; 2989 2990 /* 2991 * RFC 3542 deprecated the usage of sticky IPV6_HOPLIMIT 2992 * to simplify the ordering among hoplimit options. 2993 */ 2994 if (optname == IPV6_HOPLIMIT && sticky) 2995 return (ENOPROTOOPT); 2996 2997 if (len != sizeof(int)) 2998 return (EINVAL); 2999 hlimp = (int *)buf; 3000 if (*hlimp < -1 || *hlimp > 255) 3001 return (EINVAL); 3002 3003 opt->ip6po_hlim = *hlimp; 3004 break; 3005 } 3006 3007 case IPV6_OTCLASS: 3008 if (len != sizeof(u_int8_t)) 3009 return (EINVAL); 3010 3011 opt->ip6po_tclass = *(u_int8_t *)buf; 3012 break; 3013 3014 case IPV6_TCLASS: 3015 { 3016 int tclass; 3017 3018 if (len != sizeof(int)) 3019 return (EINVAL); 3020 tclass = *(int *)buf; 3021 if (tclass < -1 || tclass > 255) 3022 return (EINVAL); 3023 3024 opt->ip6po_tclass = tclass; 3025 break; 3026 } 3027 3028 #ifdef RFC2292 3029 case IPV6_2292NEXTHOP: 3030 #endif 3031 case IPV6_NEXTHOP: 3032 error = kauth_authorize_network(cred, KAUTH_NETWORK_IPV6, 3033 KAUTH_REQ_NETWORK_IPV6_HOPBYHOP, NULL, NULL, NULL); 3034 if (error) 3035 return (error); 3036 3037 if (len == 0) { /* just remove the option */ 3038 ip6_clearpktopts(opt, IPV6_NEXTHOP); 3039 break; 3040 } 3041 3042 /* check if cmsg_len is large enough for sa_len */ 3043 if (len < sizeof(struct sockaddr) || len < *buf) 3044 return (EINVAL); 3045 3046 switch (((struct sockaddr *)buf)->sa_family) { 3047 case AF_INET6: 3048 { 3049 struct sockaddr_in6 *sa6 = (struct sockaddr_in6 *)buf; 3050 3051 if (sa6->sin6_len != sizeof(struct sockaddr_in6)) 3052 return (EINVAL); 3053 3054 if (IN6_IS_ADDR_UNSPECIFIED(&sa6->sin6_addr) || 3055 IN6_IS_ADDR_MULTICAST(&sa6->sin6_addr)) { 3056 return (EINVAL); 3057 } 3058 if ((error = sa6_embedscope(sa6, ip6_use_defzone)) 3059 != 0) { 3060 return (error); 3061 } 3062 break; 3063 } 3064 case AF_LINK: /* eventually be supported? */ 3065 default: 3066 return (EAFNOSUPPORT); 3067 } 3068 3069 /* turn off the previous option, then set the new option. */ 3070 ip6_clearpktopts(opt, IPV6_NEXTHOP); 3071 opt->ip6po_nexthop = malloc(*buf, M_IP6OPT, M_NOWAIT); 3072 if (opt->ip6po_nexthop == NULL) 3073 return (ENOBUFS); 3074 memcpy(opt->ip6po_nexthop, buf, *buf); 3075 break; 3076 3077 #ifdef RFC2292 3078 case IPV6_2292HOPOPTS: 3079 #endif 3080 case IPV6_HOPOPTS: 3081 { 3082 struct ip6_hbh *hbh; 3083 int hbhlen; 3084 3085 /* 3086 * XXX: We don't allow a non-privileged user to set ANY HbH 3087 * options, since per-option restriction has too much 3088 * overhead. 3089 */ 3090 error = kauth_authorize_network(cred, KAUTH_NETWORK_IPV6, 3091 KAUTH_REQ_NETWORK_IPV6_HOPBYHOP, NULL, NULL, NULL); 3092 if (error) 3093 return (error); 3094 3095 if (len == 0) { 3096 ip6_clearpktopts(opt, IPV6_HOPOPTS); 3097 break; /* just remove the option */ 3098 } 3099 3100 /* message length validation */ 3101 if (len < sizeof(struct ip6_hbh)) 3102 return (EINVAL); 3103 hbh = (struct ip6_hbh *)buf; 3104 hbhlen = (hbh->ip6h_len + 1) << 3; 3105 if (len != hbhlen) 3106 return (EINVAL); 3107 3108 /* turn off the previous option, then set the new option. */ 3109 ip6_clearpktopts(opt, IPV6_HOPOPTS); 3110 opt->ip6po_hbh = malloc(hbhlen, M_IP6OPT, M_NOWAIT); 3111 if (opt->ip6po_hbh == NULL) 3112 return (ENOBUFS); 3113 memcpy(opt->ip6po_hbh, hbh, hbhlen); 3114 3115 break; 3116 } 3117 3118 #ifdef RFC2292 3119 case IPV6_2292DSTOPTS: 3120 #endif 3121 case IPV6_DSTOPTS: 3122 case IPV6_RTHDRDSTOPTS: 3123 { 3124 struct ip6_dest *dest, **newdest = NULL; 3125 int destlen; 3126 3127 /* XXX: see the comment for IPV6_HOPOPTS */ 3128 error = kauth_authorize_network(cred, KAUTH_NETWORK_IPV6, 3129 KAUTH_REQ_NETWORK_IPV6_HOPBYHOP, NULL, NULL, NULL); 3130 if (error) 3131 return (error); 3132 3133 if (len == 0) { 3134 ip6_clearpktopts(opt, optname); 3135 break; /* just remove the option */ 3136 } 3137 3138 /* message length validation */ 3139 if (len < sizeof(struct ip6_dest)) 3140 return (EINVAL); 3141 dest = (struct ip6_dest *)buf; 3142 destlen = (dest->ip6d_len + 1) << 3; 3143 if (len != destlen) 3144 return (EINVAL); 3145 /* 3146 * Determine the position that the destination options header 3147 * should be inserted; before or after the routing header. 3148 */ 3149 switch (optname) { 3150 case IPV6_2292DSTOPTS: 3151 /* 3152 * The old advanced API is ambiguous on this point. 3153 * Our approach is to determine the position based 3154 * according to the existence of a routing header. 3155 * Note, however, that this depends on the order of the 3156 * extension headers in the ancillary data; the 1st 3157 * part of the destination options header must appear 3158 * before the routing header in the ancillary data, 3159 * too. 3160 * RFC3542 solved the ambiguity by introducing 3161 * separate ancillary data or option types. 3162 */ 3163 if (opt->ip6po_rthdr == NULL) 3164 newdest = &opt->ip6po_dest1; 3165 else 3166 newdest = &opt->ip6po_dest2; 3167 break; 3168 case IPV6_RTHDRDSTOPTS: 3169 newdest = &opt->ip6po_dest1; 3170 break; 3171 case IPV6_DSTOPTS: 3172 newdest = &opt->ip6po_dest2; 3173 break; 3174 } 3175 3176 /* turn off the previous option, then set the new option. */ 3177 ip6_clearpktopts(opt, optname); 3178 *newdest = malloc(destlen, M_IP6OPT, M_NOWAIT); 3179 if (*newdest == NULL) 3180 return (ENOBUFS); 3181 memcpy(*newdest, dest, destlen); 3182 3183 break; 3184 } 3185 3186 #ifdef RFC2292 3187 case IPV6_2292RTHDR: 3188 #endif 3189 case IPV6_RTHDR: 3190 { 3191 struct ip6_rthdr *rth; 3192 int rthlen; 3193 3194 if (len == 0) { 3195 ip6_clearpktopts(opt, IPV6_RTHDR); 3196 break; /* just remove the option */ 3197 } 3198 3199 /* message length validation */ 3200 if (len < sizeof(struct ip6_rthdr)) 3201 return (EINVAL); 3202 rth = (struct ip6_rthdr *)buf; 3203 rthlen = (rth->ip6r_len + 1) << 3; 3204 if (len != rthlen) 3205 return (EINVAL); 3206 switch (rth->ip6r_type) { 3207 case IPV6_RTHDR_TYPE_0: 3208 if (rth->ip6r_len == 0) /* must contain one addr */ 3209 return (EINVAL); 3210 if (rth->ip6r_len % 2) /* length must be even */ 3211 return (EINVAL); 3212 if (rth->ip6r_len / 2 != rth->ip6r_segleft) 3213 return (EINVAL); 3214 break; 3215 default: 3216 return (EINVAL); /* not supported */ 3217 } 3218 /* turn off the previous option */ 3219 ip6_clearpktopts(opt, IPV6_RTHDR); 3220 opt->ip6po_rthdr = malloc(rthlen, M_IP6OPT, M_NOWAIT); 3221 if (opt->ip6po_rthdr == NULL) 3222 return (ENOBUFS); 3223 memcpy(opt->ip6po_rthdr, rth, rthlen); 3224 break; 3225 } 3226 3227 case IPV6_USE_MIN_MTU: 3228 if (len != sizeof(int)) 3229 return (EINVAL); 3230 minmtupolicy = *(int *)buf; 3231 if (minmtupolicy != IP6PO_MINMTU_MCASTONLY && 3232 minmtupolicy != IP6PO_MINMTU_DISABLE && 3233 minmtupolicy != IP6PO_MINMTU_ALL) { 3234 return (EINVAL); 3235 } 3236 opt->ip6po_minmtu = minmtupolicy; 3237 break; 3238 3239 case IPV6_DONTFRAG: 3240 if (len != sizeof(int)) 3241 return (EINVAL); 3242 3243 if (uproto == IPPROTO_TCP || *(int *)buf == 0) { 3244 /* 3245 * we ignore this option for TCP sockets. 3246 * (RFC3542 leaves this case unspecified.) 3247 */ 3248 opt->ip6po_flags &= ~IP6PO_DONTFRAG; 3249 } else 3250 opt->ip6po_flags |= IP6PO_DONTFRAG; 3251 break; 3252 3253 case IPV6_PREFER_TEMPADDR: 3254 { 3255 int preftemp; 3256 3257 if (len != sizeof(int)) 3258 return (EINVAL); 3259 preftemp = *(int *)buf; 3260 switch (preftemp) { 3261 case IP6PO_TEMPADDR_SYSTEM: 3262 case IP6PO_TEMPADDR_NOTPREFER: 3263 case IP6PO_TEMPADDR_PREFER: 3264 break; 3265 default: 3266 return (EINVAL); 3267 } 3268 opt->ip6po_prefer_tempaddr = preftemp; 3269 break; 3270 } 3271 3272 default: 3273 return (ENOPROTOOPT); 3274 } /* end of switch */ 3275 3276 return (0); 3277 } 3278 3279 /* 3280 * Routine called from ip6_output() to loop back a copy of an IP6 multicast 3281 * packet to the input queue of a specified interface. Note that this 3282 * calls the output routine of the loopback "driver", but with an interface 3283 * pointer that might NOT be lo0ifp -- easier than replicating that code here. 3284 */ 3285 void 3286 ip6_mloopback(struct ifnet *ifp, struct mbuf *m, 3287 const struct sockaddr_in6 *dst) 3288 { 3289 struct mbuf *copym; 3290 struct ip6_hdr *ip6; 3291 3292 copym = m_copy(m, 0, M_COPYALL); 3293 if (copym == NULL) 3294 return; 3295 3296 /* 3297 * Make sure to deep-copy IPv6 header portion in case the data 3298 * is in an mbuf cluster, so that we can safely override the IPv6 3299 * header portion later. 3300 */ 3301 if ((copym->m_flags & M_EXT) != 0 || 3302 copym->m_len < sizeof(struct ip6_hdr)) { 3303 copym = m_pullup(copym, sizeof(struct ip6_hdr)); 3304 if (copym == NULL) 3305 return; 3306 } 3307 3308 #ifdef DIAGNOSTIC 3309 if (copym->m_len < sizeof(*ip6)) { 3310 m_freem(copym); 3311 return; 3312 } 3313 #endif 3314 3315 ip6 = mtod(copym, struct ip6_hdr *); 3316 /* 3317 * clear embedded scope identifiers if necessary. 3318 * in6_clearscope will touch the addresses only when necessary. 3319 */ 3320 in6_clearscope(&ip6->ip6_src); 3321 in6_clearscope(&ip6->ip6_dst); 3322 3323 (void)looutput(ifp, copym, (const struct sockaddr *)dst, NULL); 3324 } 3325 3326 /* 3327 * Chop IPv6 header off from the payload. 3328 */ 3329 static int 3330 ip6_splithdr(struct mbuf *m, struct ip6_exthdrs *exthdrs) 3331 { 3332 struct mbuf *mh; 3333 struct ip6_hdr *ip6; 3334 3335 ip6 = mtod(m, struct ip6_hdr *); 3336 if (m->m_len > sizeof(*ip6)) { 3337 MGETHDR(mh, M_DONTWAIT, MT_HEADER); 3338 if (mh == NULL) { 3339 m_freem(m); 3340 return ENOBUFS; 3341 } 3342 M_MOVE_PKTHDR(mh, m); 3343 MH_ALIGN(mh, sizeof(*ip6)); 3344 m->m_len -= sizeof(*ip6); 3345 m->m_data += sizeof(*ip6); 3346 mh->m_next = m; 3347 mh->m_len = sizeof(*ip6); 3348 memcpy(mtod(mh, void *), (void *)ip6, sizeof(*ip6)); 3349 m = mh; 3350 } 3351 exthdrs->ip6e_ip6 = m; 3352 return 0; 3353 } 3354 3355 /* 3356 * Compute IPv6 extension header length. 3357 */ 3358 int 3359 ip6_optlen(struct in6pcb *in6p) 3360 { 3361 int len; 3362 3363 if (!in6p->in6p_outputopts) 3364 return 0; 3365 3366 len = 0; 3367 #define elen(x) \ 3368 (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0) 3369 3370 len += elen(in6p->in6p_outputopts->ip6po_hbh); 3371 len += elen(in6p->in6p_outputopts->ip6po_dest1); 3372 len += elen(in6p->in6p_outputopts->ip6po_rthdr); 3373 len += elen(in6p->in6p_outputopts->ip6po_dest2); 3374 return len; 3375 #undef elen 3376 } 3377 3378 /* 3379 * Ensure sending address is valid. 3380 * Returns 0 on success, -1 if an error should be sent back or 1 3381 * if the packet could be dropped without error (protocol dependent). 3382 */ 3383 static int 3384 ip6_ifaddrvalid(const struct in6_addr *src, const struct in6_addr *dst) 3385 { 3386 struct sockaddr_in6 sin6; 3387 int s, error; 3388 struct ifaddr *ifa; 3389 struct in6_ifaddr *ia6; 3390 3391 if (IN6_IS_ADDR_UNSPECIFIED(src)) 3392 return 0; 3393 3394 memset(&sin6, 0, sizeof(sin6)); 3395 sin6.sin6_family = AF_INET6; 3396 sin6.sin6_len = sizeof(sin6); 3397 sin6.sin6_addr = *src; 3398 3399 s = pserialize_read_enter(); 3400 ifa = ifa_ifwithaddr(sin6tosa(&sin6)); 3401 if ((ia6 = ifatoia6(ifa)) == NULL || 3402 ia6->ia6_flags & (IN6_IFF_ANYCAST | IN6_IFF_DUPLICATED)) 3403 error = -1; 3404 else if (ia6->ia6_flags & IN6_IFF_TENTATIVE) 3405 error = 1; 3406 else if (ia6->ia6_flags & IN6_IFF_DETACHED && 3407 (sin6.sin6_addr = *dst, ifa_ifwithaddr(sin6tosa(&sin6)) == NULL)) 3408 /* Allow internal traffic to DETACHED addresses */ 3409 error = 1; 3410 else 3411 error = 0; 3412 pserialize_read_exit(s); 3413 3414 return error; 3415 } 3416
Indexes created Tue Sep 30 20:09:53 GMT 2025