ip6_output.c revision 1.114.2.3
1 /* $NetBSD: ip6_output.c,v 1.114.2.3 2007/05/07 10:56:04 yamt 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 1.114.2.3 2007/05/07 10:56:04 yamt Exp $"); 66 67 #include "opt_inet.h" 68 #include "opt_inet6.h" 69 #include "opt_ipsec.h" 70 #include "opt_pfil_hooks.h" 71 72 #include <sys/param.h> 73 #include <sys/malloc.h> 74 #include <sys/mbuf.h> 75 #include <sys/errno.h> 76 #include <sys/protosw.h> 77 #include <sys/socket.h> 78 #include <sys/socketvar.h> 79 #include <sys/systm.h> 80 #include <sys/proc.h> 81 #include <sys/kauth.h> 82 83 #include <net/if.h> 84 #include <net/route.h> 85 #ifdef PFIL_HOOKS 86 #include <net/pfil.h> 87 #endif 88 89 #include <netinet/in.h> 90 #include <netinet/in_var.h> 91 #include <netinet/ip6.h> 92 #include <netinet/icmp6.h> 93 #include <netinet/in_offload.h> 94 #include <netinet6/in6_offload.h> 95 #include <netinet6/ip6_var.h> 96 #include <netinet6/in6_pcb.h> 97 #include <netinet6/nd6.h> 98 #include <netinet6/ip6protosw.h> 99 #include <netinet6/scope6_var.h> 100 101 #ifdef IPSEC 102 #include <netinet6/ipsec.h> 103 #include <netkey/key.h> 104 #endif /* IPSEC */ 105 106 #ifdef FAST_IPSEC 107 #include <netipsec/ipsec.h> 108 #include <netipsec/ipsec6.h> 109 #include <netipsec/key.h> 110 #include <netipsec/xform.h> 111 #endif 112 113 114 #include <net/net_osdep.h> 115 116 #ifdef PFIL_HOOKS 117 extern struct pfil_head inet6_pfil_hook; /* XXX */ 118 #endif 119 120 struct ip6_exthdrs { 121 struct mbuf *ip6e_ip6; 122 struct mbuf *ip6e_hbh; 123 struct mbuf *ip6e_dest1; 124 struct mbuf *ip6e_rthdr; 125 struct mbuf *ip6e_dest2; 126 }; 127 128 static int ip6_pcbopt __P((int, u_char *, int, struct ip6_pktopts **, 129 int, int)); 130 static int ip6_getpcbopt __P((struct ip6_pktopts *, int, struct mbuf **)); 131 static int ip6_setpktopt __P((int, u_char *, int, struct ip6_pktopts *, int, 132 int, int, int)); 133 static int ip6_setmoptions __P((int, struct ip6_moptions **, struct mbuf *)); 134 static int ip6_getmoptions __P((int, struct ip6_moptions *, struct mbuf **)); 135 static int ip6_copyexthdr __P((struct mbuf **, void *, int)); 136 static int ip6_insertfraghdr __P((struct mbuf *, struct mbuf *, int, 137 struct ip6_frag **)); 138 static int ip6_insert_jumboopt __P((struct ip6_exthdrs *, u_int32_t)); 139 static int ip6_splithdr __P((struct mbuf *, struct ip6_exthdrs *)); 140 static int ip6_getpmtu(struct route *, struct route *, struct ifnet *, 141 const struct in6_addr *, u_long *, int *); 142 static int copypktopts __P((struct ip6_pktopts *, struct ip6_pktopts *, int)); 143 144 #ifdef RFC2292 145 static int ip6_pcbopts __P((struct ip6_pktopts **, struct mbuf *, 146 struct socket *)); 147 #endif 148 149 #define IN6_NEED_CHECKSUM(ifp, csum_flags) \ 150 (__predict_true(((ifp)->if_flags & IFF_LOOPBACK) == 0 || \ 151 (((csum_flags) & M_CSUM_UDPv6) != 0 && udp_do_loopback_cksum) || \ 152 (((csum_flags) & M_CSUM_TCPv6) != 0 && tcp_do_loopback_cksum))) 153 154 /* 155 * IP6 output. The packet in mbuf chain m contains a skeletal IP6 156 * header (with pri, len, nxt, hlim, src, dst). 157 * This function may modify ver and hlim only. 158 * The mbuf chain containing the packet will be freed. 159 * The mbuf opt, if present, will not be freed. 160 * 161 * type of "mtu": rt_rmx.rmx_mtu is u_long, ifnet.ifr_mtu is int, and 162 * nd_ifinfo.linkmtu is u_int32_t. so we use u_long to hold largest one, 163 * which is rt_rmx.rmx_mtu. 164 */ 165 int 166 ip6_output( 167 struct mbuf *m0, 168 struct ip6_pktopts *opt, 169 struct route *ro, 170 int flags, 171 struct ip6_moptions *im6o, 172 struct socket *so, 173 struct ifnet **ifpp /* XXX: just for statistics */ 174 ) 175 { 176 struct ip6_hdr *ip6, *mhip6; 177 struct ifnet *ifp, *origifp; 178 struct mbuf *m = m0; 179 int hlen, tlen, len, off; 180 bool tso; 181 struct route ip6route; 182 struct rtentry *rt = NULL; 183 const struct sockaddr_in6 *dst = NULL; 184 struct sockaddr_in6 src_sa, dst_sa; 185 int error = 0; 186 struct in6_ifaddr *ia = NULL; 187 u_long mtu; 188 int alwaysfrag, dontfrag; 189 u_int32_t optlen = 0, plen = 0, unfragpartlen = 0; 190 struct ip6_exthdrs exthdrs; 191 struct in6_addr finaldst, src0, dst0; 192 u_int32_t zone; 193 struct route *ro_pmtu = NULL; 194 int hdrsplit = 0; 195 int needipsec = 0; 196 #ifdef IPSEC 197 int needipsectun = 0; 198 struct secpolicy *sp = NULL; 199 200 ip6 = mtod(m, struct ip6_hdr *); 201 #endif /* IPSEC */ 202 #ifdef FAST_IPSEC 203 struct secpolicy *sp = NULL; 204 int s; 205 #endif 206 207 208 #ifdef DIAGNOSTIC 209 if ((m->m_flags & M_PKTHDR) == 0) 210 panic("ip6_output: no HDR"); 211 212 if ((m->m_pkthdr.csum_flags & 213 (M_CSUM_TCPv4|M_CSUM_UDPv4|M_CSUM_TSOv4)) != 0) { 214 panic("ip6_output: IPv4 checksum offload flags: %d", 215 m->m_pkthdr.csum_flags); 216 } 217 218 if ((m->m_pkthdr.csum_flags & (M_CSUM_TCPv6|M_CSUM_UDPv6)) == 219 (M_CSUM_TCPv6|M_CSUM_UDPv6)) { 220 panic("ip6_output: conflicting checksum offload flags: %d", 221 m->m_pkthdr.csum_flags); 222 } 223 #endif 224 225 M_CSUM_DATA_IPv6_HL_SET(m->m_pkthdr.csum_data, sizeof(struct ip6_hdr)); 226 227 #define MAKE_EXTHDR(hp, mp) \ 228 do { \ 229 if (hp) { \ 230 struct ip6_ext *eh = (struct ip6_ext *)(hp); \ 231 error = ip6_copyexthdr((mp), (void *)(hp), \ 232 ((eh)->ip6e_len + 1) << 3); \ 233 if (error) \ 234 goto freehdrs; \ 235 } \ 236 } while (/*CONSTCOND*/ 0) 237 238 bzero(&exthdrs, sizeof(exthdrs)); 239 if (opt) { 240 /* Hop-by-Hop options header */ 241 MAKE_EXTHDR(opt->ip6po_hbh, &exthdrs.ip6e_hbh); 242 /* Destination options header(1st part) */ 243 MAKE_EXTHDR(opt->ip6po_dest1, &exthdrs.ip6e_dest1); 244 /* Routing header */ 245 MAKE_EXTHDR(opt->ip6po_rthdr, &exthdrs.ip6e_rthdr); 246 /* Destination options header(2nd part) */ 247 MAKE_EXTHDR(opt->ip6po_dest2, &exthdrs.ip6e_dest2); 248 } 249 250 #ifdef IPSEC 251 if ((flags & IPV6_FORWARDING) != 0) { 252 needipsec = 0; 253 goto skippolicycheck; 254 } 255 256 /* get a security policy for this packet */ 257 if (so == NULL) 258 sp = ipsec6_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND, 0, &error); 259 else { 260 if (IPSEC_PCB_SKIP_IPSEC(sotoinpcb_hdr(so)->inph_sp, 261 IPSEC_DIR_OUTBOUND)) { 262 needipsec = 0; 263 goto skippolicycheck; 264 } 265 sp = ipsec6_getpolicybysock(m, IPSEC_DIR_OUTBOUND, so, &error); 266 } 267 268 if (sp == NULL) { 269 ipsec6stat.out_inval++; 270 goto freehdrs; 271 } 272 273 error = 0; 274 275 /* check policy */ 276 switch (sp->policy) { 277 case IPSEC_POLICY_DISCARD: 278 /* 279 * This packet is just discarded. 280 */ 281 ipsec6stat.out_polvio++; 282 goto freehdrs; 283 284 case IPSEC_POLICY_BYPASS: 285 case IPSEC_POLICY_NONE: 286 /* no need to do IPsec. */ 287 needipsec = 0; 288 break; 289 290 case IPSEC_POLICY_IPSEC: 291 if (sp->req == NULL) { 292 /* XXX should be panic ? */ 293 printf("ip6_output: No IPsec request specified.\n"); 294 error = EINVAL; 295 goto freehdrs; 296 } 297 needipsec = 1; 298 break; 299 300 case IPSEC_POLICY_ENTRUST: 301 default: 302 printf("ip6_output: Invalid policy found. %d\n", sp->policy); 303 } 304 305 skippolicycheck:; 306 #endif /* IPSEC */ 307 308 /* 309 * Calculate the total length of the extension header chain. 310 * Keep the length of the unfragmentable part for fragmentation. 311 */ 312 optlen = 0; 313 if (exthdrs.ip6e_hbh) optlen += exthdrs.ip6e_hbh->m_len; 314 if (exthdrs.ip6e_dest1) optlen += exthdrs.ip6e_dest1->m_len; 315 if (exthdrs.ip6e_rthdr) optlen += exthdrs.ip6e_rthdr->m_len; 316 unfragpartlen = optlen + sizeof(struct ip6_hdr); 317 /* NOTE: we don't add AH/ESP length here. do that later. */ 318 if (exthdrs.ip6e_dest2) optlen += exthdrs.ip6e_dest2->m_len; 319 320 #ifdef FAST_IPSEC 321 /* Check the security policy (SP) for the packet */ 322 323 /* XXX For moment, we doesn't support packet with extented action */ 324 if (optlen !=0) 325 goto freehdrs; 326 327 sp = ipsec6_check_policy(m,so,flags,&needipsec,&error); 328 if (error != 0) { 329 /* 330 * Hack: -EINVAL is used to signal that a packet 331 * should be silently discarded. This is typically 332 * because we asked key management for an SA and 333 * it was delayed (e.g. kicked up to IKE). 334 */ 335 if (error == -EINVAL) 336 error = 0; 337 goto freehdrs; 338 } 339 #endif /* FAST_IPSEC */ 340 341 342 if (needipsec && 343 (m->m_pkthdr.csum_flags & (M_CSUM_UDPv6|M_CSUM_TCPv6)) != 0) { 344 in6_delayed_cksum(m); 345 m->m_pkthdr.csum_flags &= ~(M_CSUM_UDPv6|M_CSUM_TCPv6); 346 } 347 348 349 /* 350 * If we need IPsec, or there is at least one extension header, 351 * separate IP6 header from the payload. 352 */ 353 if ((needipsec || optlen) && !hdrsplit) { 354 if ((error = ip6_splithdr(m, &exthdrs)) != 0) { 355 m = NULL; 356 goto freehdrs; 357 } 358 m = exthdrs.ip6e_ip6; 359 hdrsplit++; 360 } 361 362 /* adjust pointer */ 363 ip6 = mtod(m, struct ip6_hdr *); 364 365 /* adjust mbuf packet header length */ 366 m->m_pkthdr.len += optlen; 367 plen = m->m_pkthdr.len - sizeof(*ip6); 368 369 /* If this is a jumbo payload, insert a jumbo payload option. */ 370 if (plen > IPV6_MAXPACKET) { 371 if (!hdrsplit) { 372 if ((error = ip6_splithdr(m, &exthdrs)) != 0) { 373 m = NULL; 374 goto freehdrs; 375 } 376 m = exthdrs.ip6e_ip6; 377 hdrsplit++; 378 } 379 /* adjust pointer */ 380 ip6 = mtod(m, struct ip6_hdr *); 381 if ((error = ip6_insert_jumboopt(&exthdrs, plen)) != 0) 382 goto freehdrs; 383 optlen += 8; /* XXX JUMBOOPTLEN */ 384 ip6->ip6_plen = 0; 385 } else 386 ip6->ip6_plen = htons(plen); 387 388 /* 389 * Concatenate headers and fill in next header fields. 390 * Here we have, on "m" 391 * IPv6 payload 392 * and we insert headers accordingly. Finally, we should be getting: 393 * IPv6 hbh dest1 rthdr ah* [esp* dest2 payload] 394 * 395 * during the header composing process, "m" points to IPv6 header. 396 * "mprev" points to an extension header prior to esp. 397 */ 398 { 399 u_char *nexthdrp = &ip6->ip6_nxt; 400 struct mbuf *mprev = m; 401 402 /* 403 * we treat dest2 specially. this makes IPsec processing 404 * much easier. the goal here is to make mprev point the 405 * mbuf prior to dest2. 406 * 407 * result: IPv6 dest2 payload 408 * m and mprev will point to IPv6 header. 409 */ 410 if (exthdrs.ip6e_dest2) { 411 if (!hdrsplit) 412 panic("assumption failed: hdr not split"); 413 exthdrs.ip6e_dest2->m_next = m->m_next; 414 m->m_next = exthdrs.ip6e_dest2; 415 *mtod(exthdrs.ip6e_dest2, u_char *) = ip6->ip6_nxt; 416 ip6->ip6_nxt = IPPROTO_DSTOPTS; 417 } 418 419 #define MAKE_CHAIN(m, mp, p, i)\ 420 do {\ 421 if (m) {\ 422 if (!hdrsplit) \ 423 panic("assumption failed: hdr not split"); \ 424 *mtod((m), u_char *) = *(p);\ 425 *(p) = (i);\ 426 p = mtod((m), u_char *);\ 427 (m)->m_next = (mp)->m_next;\ 428 (mp)->m_next = (m);\ 429 (mp) = (m);\ 430 }\ 431 } while (/*CONSTCOND*/ 0) 432 /* 433 * result: IPv6 hbh dest1 rthdr dest2 payload 434 * m will point to IPv6 header. mprev will point to the 435 * extension header prior to dest2 (rthdr in the above case). 436 */ 437 MAKE_CHAIN(exthdrs.ip6e_hbh, mprev, nexthdrp, IPPROTO_HOPOPTS); 438 MAKE_CHAIN(exthdrs.ip6e_dest1, mprev, nexthdrp, 439 IPPROTO_DSTOPTS); 440 MAKE_CHAIN(exthdrs.ip6e_rthdr, mprev, nexthdrp, 441 IPPROTO_ROUTING); 442 443 M_CSUM_DATA_IPv6_HL_SET(m->m_pkthdr.csum_data, 444 sizeof(struct ip6_hdr) + optlen); 445 446 #ifdef IPSEC 447 if (!needipsec) 448 goto skip_ipsec2; 449 450 /* 451 * pointers after IPsec headers are not valid any more. 452 * other pointers need a great care too. 453 * (IPsec routines should not mangle mbufs prior to AH/ESP) 454 */ 455 exthdrs.ip6e_dest2 = NULL; 456 457 { 458 struct ip6_rthdr *rh = NULL; 459 int segleft_org = 0; 460 struct ipsec_output_state state; 461 462 if (exthdrs.ip6e_rthdr) { 463 rh = mtod(exthdrs.ip6e_rthdr, struct ip6_rthdr *); 464 segleft_org = rh->ip6r_segleft; 465 rh->ip6r_segleft = 0; 466 } 467 468 bzero(&state, sizeof(state)); 469 state.m = m; 470 error = ipsec6_output_trans(&state, nexthdrp, mprev, sp, flags, 471 &needipsectun); 472 m = state.m; 473 if (error) { 474 rh = mtod(exthdrs.ip6e_rthdr, struct ip6_rthdr *); 475 /* mbuf is already reclaimed in ipsec6_output_trans. */ 476 m = NULL; 477 switch (error) { 478 case EHOSTUNREACH: 479 case ENETUNREACH: 480 case EMSGSIZE: 481 case ENOBUFS: 482 case ENOMEM: 483 break; 484 default: 485 printf("ip6_output (ipsec): error code %d\n", error); 486 /* FALLTHROUGH */ 487 case ENOENT: 488 /* don't show these error codes to the user */ 489 error = 0; 490 break; 491 } 492 goto bad; 493 } 494 if (exthdrs.ip6e_rthdr) { 495 /* ah6_output doesn't modify mbuf chain */ 496 rh->ip6r_segleft = segleft_org; 497 } 498 } 499 skip_ipsec2:; 500 #endif 501 } 502 503 /* 504 * If there is a routing header, replace destination address field 505 * with the first hop of the routing header. 506 */ 507 if (exthdrs.ip6e_rthdr) { 508 struct ip6_rthdr *rh; 509 struct ip6_rthdr0 *rh0; 510 struct in6_addr *addr; 511 struct sockaddr_in6 sa; 512 513 rh = (struct ip6_rthdr *)(mtod(exthdrs.ip6e_rthdr, 514 struct ip6_rthdr *)); 515 finaldst = ip6->ip6_dst; 516 switch (rh->ip6r_type) { 517 case IPV6_RTHDR_TYPE_0: 518 rh0 = (struct ip6_rthdr0 *)rh; 519 addr = (struct in6_addr *)(rh0 + 1); 520 521 /* 522 * construct a sockaddr_in6 form of 523 * the first hop. 524 * 525 * XXX: we may not have enough 526 * information about its scope zone; 527 * there is no standard API to pass 528 * the information from the 529 * application. 530 */ 531 bzero(&sa, sizeof(sa)); 532 sa.sin6_family = AF_INET6; 533 sa.sin6_len = sizeof(sa); 534 sa.sin6_addr = addr[0]; 535 if ((error = sa6_embedscope(&sa, 536 ip6_use_defzone)) != 0) { 537 goto bad; 538 } 539 ip6->ip6_dst = sa.sin6_addr; 540 (void)memmove(&addr[0], &addr[1], 541 sizeof(struct in6_addr) * 542 (rh0->ip6r0_segleft - 1)); 543 addr[rh0->ip6r0_segleft - 1] = finaldst; 544 /* XXX */ 545 in6_clearscope(addr + rh0->ip6r0_segleft - 1); 546 break; 547 default: /* is it possible? */ 548 error = EINVAL; 549 goto bad; 550 } 551 } 552 553 /* Source address validation */ 554 if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) && 555 (flags & IPV6_UNSPECSRC) == 0) { 556 error = EOPNOTSUPP; 557 ip6stat.ip6s_badscope++; 558 goto bad; 559 } 560 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) { 561 error = EOPNOTSUPP; 562 ip6stat.ip6s_badscope++; 563 goto bad; 564 } 565 566 ip6stat.ip6s_localout++; 567 568 /* 569 * Route packet. 570 */ 571 /* initialize cached route */ 572 if (ro == NULL) { 573 memset(&ip6route, 0, sizeof(ip6route)); 574 ro = &ip6route; 575 } 576 ro_pmtu = ro; 577 if (opt && opt->ip6po_rthdr) 578 ro = &opt->ip6po_route; 579 580 /* 581 * if specified, try to fill in the traffic class field. 582 * do not override if a non-zero value is already set. 583 * we check the diffserv field and the ecn field separately. 584 */ 585 if (opt && opt->ip6po_tclass >= 0) { 586 int mask = 0; 587 588 if ((ip6->ip6_flow & htonl(0xfc << 20)) == 0) 589 mask |= 0xfc; 590 if ((ip6->ip6_flow & htonl(0x03 << 20)) == 0) 591 mask |= 0x03; 592 if (mask != 0) 593 ip6->ip6_flow |= htonl((opt->ip6po_tclass & mask) << 20); 594 } 595 596 /* fill in or override the hop limit field, if necessary. */ 597 if (opt && opt->ip6po_hlim != -1) 598 ip6->ip6_hlim = opt->ip6po_hlim & 0xff; 599 else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) { 600 if (im6o != NULL) 601 ip6->ip6_hlim = im6o->im6o_multicast_hlim; 602 else 603 ip6->ip6_hlim = ip6_defmcasthlim; 604 } 605 606 #ifdef IPSEC 607 if (needipsec && needipsectun) { 608 struct ipsec_output_state state; 609 610 /* 611 * All the extension headers will become inaccessible 612 * (since they can be encrypted). 613 * Don't panic, we need no more updates to extension headers 614 * on inner IPv6 packet (since they are now encapsulated). 615 * 616 * IPv6 [ESP|AH] IPv6 [extension headers] payload 617 */ 618 bzero(&exthdrs, sizeof(exthdrs)); 619 exthdrs.ip6e_ip6 = m; 620 621 bzero(&state, sizeof(state)); 622 state.m = m; 623 state.ro = ro; 624 state.dst = rtcache_getdst(ro); 625 626 error = ipsec6_output_tunnel(&state, sp, flags); 627 628 m = state.m; 629 ro_pmtu = ro = state.ro; 630 dst = satocsin6(state.dst); 631 if (error) { 632 /* mbuf is already reclaimed in ipsec6_output_tunnel. */ 633 m0 = m = NULL; 634 m = NULL; 635 switch (error) { 636 case EHOSTUNREACH: 637 case ENETUNREACH: 638 case EMSGSIZE: 639 case ENOBUFS: 640 case ENOMEM: 641 break; 642 default: 643 printf("ip6_output (ipsec): error code %d\n", error); 644 /* FALLTHROUGH */ 645 case ENOENT: 646 /* don't show these error codes to the user */ 647 error = 0; 648 break; 649 } 650 goto bad; 651 } 652 653 exthdrs.ip6e_ip6 = m; 654 } 655 #endif /* IPSEC */ 656 #ifdef FAST_IPSEC 657 if (needipsec) { 658 s = splsoftnet(); 659 error = ipsec6_process_packet(m,sp->req); 660 661 /* 662 * Preserve KAME behaviour: ENOENT can be returned 663 * when an SA acquire is in progress. Don't propagate 664 * this to user-level; it confuses applications. 665 * XXX this will go away when the SADB is redone. 666 */ 667 if (error == ENOENT) 668 error = 0; 669 splx(s); 670 goto done; 671 } 672 #endif /* FAST_IPSEC */ 673 674 675 676 /* adjust pointer */ 677 ip6 = mtod(m, struct ip6_hdr *); 678 679 bzero(&dst_sa, sizeof(dst_sa)); 680 dst_sa.sin6_family = AF_INET6; 681 dst_sa.sin6_len = sizeof(dst_sa); 682 dst_sa.sin6_addr = ip6->ip6_dst; 683 if ((error = in6_selectroute(&dst_sa, opt, im6o, ro, 684 &ifp, &rt, 0)) != 0) { 685 switch (error) { 686 case EHOSTUNREACH: 687 ip6stat.ip6s_noroute++; 688 break; 689 case EADDRNOTAVAIL: 690 default: 691 break; /* XXX statistics? */ 692 } 693 if (ifp != NULL) 694 in6_ifstat_inc(ifp, ifs6_out_discard); 695 goto bad; 696 } 697 if (rt == NULL) { 698 /* 699 * If in6_selectroute() does not return a route entry, 700 * dst may not have been updated. 701 */ 702 rtcache_setdst(ro, sin6tosa(&dst_sa)); 703 } 704 705 /* 706 * then rt (for unicast) and ifp must be non-NULL valid values. 707 */ 708 if ((flags & IPV6_FORWARDING) == 0) { 709 /* XXX: the FORWARDING flag can be set for mrouting. */ 710 in6_ifstat_inc(ifp, ifs6_out_request); 711 } 712 if (rt != NULL) { 713 ia = (struct in6_ifaddr *)(rt->rt_ifa); 714 rt->rt_use++; 715 } 716 717 /* 718 * The outgoing interface must be in the zone of source and 719 * destination addresses. We should use ia_ifp to support the 720 * case of sending packets to an address of our own. 721 */ 722 if (ia != NULL && ia->ia_ifp) 723 origifp = ia->ia_ifp; 724 else 725 origifp = ifp; 726 727 src0 = ip6->ip6_src; 728 if (in6_setscope(&src0, origifp, &zone)) 729 goto badscope; 730 bzero(&src_sa, sizeof(src_sa)); 731 src_sa.sin6_family = AF_INET6; 732 src_sa.sin6_len = sizeof(src_sa); 733 src_sa.sin6_addr = ip6->ip6_src; 734 if (sa6_recoverscope(&src_sa) || zone != src_sa.sin6_scope_id) 735 goto badscope; 736 737 dst0 = ip6->ip6_dst; 738 if (in6_setscope(&dst0, origifp, &zone)) 739 goto badscope; 740 /* re-initialize to be sure */ 741 bzero(&dst_sa, sizeof(dst_sa)); 742 dst_sa.sin6_family = AF_INET6; 743 dst_sa.sin6_len = sizeof(dst_sa); 744 dst_sa.sin6_addr = ip6->ip6_dst; 745 if (sa6_recoverscope(&dst_sa) || zone != dst_sa.sin6_scope_id) 746 goto badscope; 747 748 /* scope check is done. */ 749 750 if (rt == NULL || IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) { 751 if (dst == NULL) 752 dst = satocsin6(rtcache_getdst(ro)); 753 KASSERT(dst != NULL); 754 } else if (opt && opt->ip6po_nextroute.ro_rt != NULL) { 755 /* 756 * The nexthop is explicitly specified by the 757 * application. We assume the next hop is an IPv6 758 * address. 759 */ 760 dst = (struct sockaddr_in6 *)opt->ip6po_nexthop; 761 } else if ((rt->rt_flags & RTF_GATEWAY)) 762 dst = (struct sockaddr_in6 *)rt->rt_gateway; 763 else if (dst == NULL) 764 dst = satocsin6(rtcache_getdst(ro)); 765 766 /* 767 * XXXXXX: original code follows: 768 */ 769 if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) 770 m->m_flags &= ~(M_BCAST | M_MCAST); /* just in case */ 771 else { 772 struct in6_multi *in6m; 773 774 m->m_flags = (m->m_flags & ~M_BCAST) | M_MCAST; 775 776 in6_ifstat_inc(ifp, ifs6_out_mcast); 777 778 /* 779 * Confirm that the outgoing interface supports multicast. 780 */ 781 if (!(ifp->if_flags & IFF_MULTICAST)) { 782 ip6stat.ip6s_noroute++; 783 in6_ifstat_inc(ifp, ifs6_out_discard); 784 error = ENETUNREACH; 785 goto bad; 786 } 787 788 IN6_LOOKUP_MULTI(ip6->ip6_dst, ifp, in6m); 789 if (in6m != NULL && 790 (im6o == NULL || im6o->im6o_multicast_loop)) { 791 /* 792 * If we belong to the destination multicast group 793 * on the outgoing interface, and the caller did not 794 * forbid loopback, loop back a copy. 795 */ 796 KASSERT(dst != NULL); 797 ip6_mloopback(ifp, m, dst); 798 } else { 799 /* 800 * If we are acting as a multicast router, perform 801 * multicast forwarding as if the packet had just 802 * arrived on the interface to which we are about 803 * to send. The multicast forwarding function 804 * recursively calls this function, using the 805 * IPV6_FORWARDING flag to prevent infinite recursion. 806 * 807 * Multicasts that are looped back by ip6_mloopback(), 808 * above, will be forwarded by the ip6_input() routine, 809 * if necessary. 810 */ 811 if (ip6_mrouter && (flags & IPV6_FORWARDING) == 0) { 812 if (ip6_mforward(ip6, ifp, m) != 0) { 813 m_freem(m); 814 goto done; 815 } 816 } 817 } 818 /* 819 * Multicasts with a hoplimit of zero may be looped back, 820 * above, but must not be transmitted on a network. 821 * Also, multicasts addressed to the loopback interface 822 * are not sent -- the above call to ip6_mloopback() will 823 * loop back a copy if this host actually belongs to the 824 * destination group on the loopback interface. 825 */ 826 if (ip6->ip6_hlim == 0 || (ifp->if_flags & IFF_LOOPBACK) || 827 IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst)) { 828 m_freem(m); 829 goto done; 830 } 831 } 832 833 /* 834 * Fill the outgoing inteface to tell the upper layer 835 * to increment per-interface statistics. 836 */ 837 if (ifpp) 838 *ifpp = ifp; 839 840 /* Determine path MTU. */ 841 if ((error = ip6_getpmtu(ro_pmtu, ro, ifp, &finaldst, &mtu, 842 &alwaysfrag)) != 0) 843 goto bad; 844 #ifdef IPSEC 845 if (needipsectun) 846 mtu = IPV6_MMTU; 847 #endif 848 849 /* 850 * The caller of this function may specify to use the minimum MTU 851 * in some cases. 852 * An advanced API option (IPV6_USE_MIN_MTU) can also override MTU 853 * setting. The logic is a bit complicated; by default, unicast 854 * packets will follow path MTU while multicast packets will be sent at 855 * the minimum MTU. If IP6PO_MINMTU_ALL is specified, all packets 856 * including unicast ones will be sent at the minimum MTU. Multicast 857 * packets will always be sent at the minimum MTU unless 858 * IP6PO_MINMTU_DISABLE is explicitly specified. 859 * See RFC 3542 for more details. 860 */ 861 if (mtu > IPV6_MMTU) { 862 if ((flags & IPV6_MINMTU)) 863 mtu = IPV6_MMTU; 864 else if (opt && opt->ip6po_minmtu == IP6PO_MINMTU_ALL) 865 mtu = IPV6_MMTU; 866 else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) && 867 (opt == NULL || 868 opt->ip6po_minmtu != IP6PO_MINMTU_DISABLE)) { 869 mtu = IPV6_MMTU; 870 } 871 } 872 873 /* 874 * clear embedded scope identifiers if necessary. 875 * in6_clearscope will touch the addresses only when necessary. 876 */ 877 in6_clearscope(&ip6->ip6_src); 878 in6_clearscope(&ip6->ip6_dst); 879 880 /* 881 * If the outgoing packet contains a hop-by-hop options header, 882 * it must be examined and processed even by the source node. 883 * (RFC 2460, section 4.) 884 */ 885 if (exthdrs.ip6e_hbh) { 886 struct ip6_hbh *hbh = mtod(exthdrs.ip6e_hbh, struct ip6_hbh *); 887 u_int32_t dummy1; /* XXX unused */ 888 u_int32_t dummy2; /* XXX unused */ 889 890 /* 891 * XXX: if we have to send an ICMPv6 error to the sender, 892 * we need the M_LOOP flag since icmp6_error() expects 893 * the IPv6 and the hop-by-hop options header are 894 * continuous unless the flag is set. 895 */ 896 m->m_flags |= M_LOOP; 897 m->m_pkthdr.rcvif = ifp; 898 if (ip6_process_hopopts(m, (u_int8_t *)(hbh + 1), 899 ((hbh->ip6h_len + 1) << 3) - sizeof(struct ip6_hbh), 900 &dummy1, &dummy2) < 0) { 901 /* m was already freed at this point */ 902 error = EINVAL;/* better error? */ 903 goto done; 904 } 905 m->m_flags &= ~M_LOOP; /* XXX */ 906 m->m_pkthdr.rcvif = NULL; 907 } 908 909 #ifdef PFIL_HOOKS 910 /* 911 * Run through list of hooks for output packets. 912 */ 913 if ((error = pfil_run_hooks(&inet6_pfil_hook, &m, ifp, PFIL_OUT)) != 0) 914 goto done; 915 if (m == NULL) 916 goto done; 917 ip6 = mtod(m, struct ip6_hdr *); 918 #endif /* PFIL_HOOKS */ 919 /* 920 * Send the packet to the outgoing interface. 921 * If necessary, do IPv6 fragmentation before sending. 922 * 923 * the logic here is rather complex: 924 * 1: normal case (dontfrag == 0, alwaysfrag == 0) 925 * 1-a: send as is if tlen <= path mtu 926 * 1-b: fragment if tlen > path mtu 927 * 928 * 2: if user asks us not to fragment (dontfrag == 1) 929 * 2-a: send as is if tlen <= interface mtu 930 * 2-b: error if tlen > interface mtu 931 * 932 * 3: if we always need to attach fragment header (alwaysfrag == 1) 933 * always fragment 934 * 935 * 4: if dontfrag == 1 && alwaysfrag == 1 936 * error, as we cannot handle this conflicting request 937 */ 938 tlen = m->m_pkthdr.len; 939 tso = (m->m_pkthdr.csum_flags & M_CSUM_TSOv6) != 0; 940 if (opt && (opt->ip6po_flags & IP6PO_DONTFRAG)) 941 dontfrag = 1; 942 else 943 dontfrag = 0; 944 945 if (dontfrag && alwaysfrag) { /* case 4 */ 946 /* conflicting request - can't transmit */ 947 error = EMSGSIZE; 948 goto bad; 949 } 950 if (dontfrag && (!tso && tlen > IN6_LINKMTU(ifp))) { /* case 2-b */ 951 /* 952 * Even if the DONTFRAG option is specified, we cannot send the 953 * packet when the data length is larger than the MTU of the 954 * outgoing interface. 955 * Notify the error by sending IPV6_PATHMTU ancillary data as 956 * well as returning an error code (the latter is not described 957 * in the API spec.) 958 */ 959 u_int32_t mtu32; 960 struct ip6ctlparam ip6cp; 961 962 mtu32 = (u_int32_t)mtu; 963 bzero(&ip6cp, sizeof(ip6cp)); 964 ip6cp.ip6c_cmdarg = (void *)&mtu32; 965 pfctlinput2(PRC_MSGSIZE, 966 rtcache_getdst(ro_pmtu), &ip6cp); 967 968 error = EMSGSIZE; 969 goto bad; 970 } 971 972 /* 973 * transmit packet without fragmentation 974 */ 975 if (dontfrag || (!alwaysfrag && (tlen <= mtu || tso))) { 976 /* case 1-a and 2-a */ 977 struct in6_ifaddr *ia6; 978 int sw_csum; 979 980 ip6 = mtod(m, struct ip6_hdr *); 981 ia6 = in6_ifawithifp(ifp, &ip6->ip6_src); 982 if (ia6) { 983 /* Record statistics for this interface address. */ 984 ia6->ia_ifa.ifa_data.ifad_outbytes += m->m_pkthdr.len; 985 } 986 #ifdef IPSEC 987 /* clean ipsec history once it goes out of the node */ 988 ipsec_delaux(m); 989 #endif 990 991 sw_csum = m->m_pkthdr.csum_flags & ~ifp->if_csum_flags_tx; 992 if ((sw_csum & (M_CSUM_UDPv6|M_CSUM_TCPv6)) != 0) { 993 if (IN6_NEED_CHECKSUM(ifp, 994 sw_csum & (M_CSUM_UDPv6|M_CSUM_TCPv6))) { 995 in6_delayed_cksum(m); 996 } 997 m->m_pkthdr.csum_flags &= ~(M_CSUM_UDPv6|M_CSUM_TCPv6); 998 } 999 1000 KASSERT(dst != NULL); 1001 if (__predict_true(!tso || 1002 (ifp->if_capenable & IFCAP_TSOv6) != 0)) { 1003 error = nd6_output(ifp, origifp, m, dst, rt); 1004 } else { 1005 error = ip6_tso_output(ifp, origifp, m, dst, rt); 1006 } 1007 goto done; 1008 } 1009 1010 if (tso) { 1011 error = EINVAL; /* XXX */ 1012 goto bad; 1013 } 1014 1015 /* 1016 * try to fragment the packet. case 1-b and 3 1017 */ 1018 if (mtu < IPV6_MMTU) { 1019 /* path MTU cannot be less than IPV6_MMTU */ 1020 error = EMSGSIZE; 1021 in6_ifstat_inc(ifp, ifs6_out_fragfail); 1022 goto bad; 1023 } else if (ip6->ip6_plen == 0) { 1024 /* jumbo payload cannot be fragmented */ 1025 error = EMSGSIZE; 1026 in6_ifstat_inc(ifp, ifs6_out_fragfail); 1027 goto bad; 1028 } else { 1029 struct mbuf **mnext, *m_frgpart; 1030 struct ip6_frag *ip6f; 1031 u_int32_t id = htonl(ip6_randomid()); 1032 u_char nextproto; 1033 #if 0 /* see below */ 1034 struct ip6ctlparam ip6cp; 1035 u_int32_t mtu32; 1036 #endif 1037 1038 /* 1039 * Too large for the destination or interface; 1040 * fragment if possible. 1041 * Must be able to put at least 8 bytes per fragment. 1042 */ 1043 hlen = unfragpartlen; 1044 if (mtu > IPV6_MAXPACKET) 1045 mtu = IPV6_MAXPACKET; 1046 1047 #if 0 1048 /* 1049 * It is believed this code is a leftover from the 1050 * development of the IPV6_RECVPATHMTU sockopt and 1051 * associated work to implement RFC3542. 1052 * It's not entirely clear what the intent of the API 1053 * is at this point, so disable this code for now. 1054 * The IPV6_RECVPATHMTU sockopt and/or IPV6_DONTFRAG 1055 * will send notifications if the application requests. 1056 */ 1057 1058 /* Notify a proper path MTU to applications. */ 1059 mtu32 = (u_int32_t)mtu; 1060 bzero(&ip6cp, sizeof(ip6cp)); 1061 ip6cp.ip6c_cmdarg = (void *)&mtu32; 1062 pfctlinput2(PRC_MSGSIZE, 1063 rtcache_getdst(ro_pmtu), &ip6cp); 1064 #endif 1065 1066 len = (mtu - hlen - sizeof(struct ip6_frag)) & ~7; 1067 if (len < 8) { 1068 error = EMSGSIZE; 1069 in6_ifstat_inc(ifp, ifs6_out_fragfail); 1070 goto bad; 1071 } 1072 1073 mnext = &m->m_nextpkt; 1074 1075 /* 1076 * Change the next header field of the last header in the 1077 * unfragmentable part. 1078 */ 1079 if (exthdrs.ip6e_rthdr) { 1080 nextproto = *mtod(exthdrs.ip6e_rthdr, u_char *); 1081 *mtod(exthdrs.ip6e_rthdr, u_char *) = IPPROTO_FRAGMENT; 1082 } else if (exthdrs.ip6e_dest1) { 1083 nextproto = *mtod(exthdrs.ip6e_dest1, u_char *); 1084 *mtod(exthdrs.ip6e_dest1, u_char *) = IPPROTO_FRAGMENT; 1085 } else if (exthdrs.ip6e_hbh) { 1086 nextproto = *mtod(exthdrs.ip6e_hbh, u_char *); 1087 *mtod(exthdrs.ip6e_hbh, u_char *) = IPPROTO_FRAGMENT; 1088 } else { 1089 nextproto = ip6->ip6_nxt; 1090 ip6->ip6_nxt = IPPROTO_FRAGMENT; 1091 } 1092 1093 if ((m->m_pkthdr.csum_flags & (M_CSUM_UDPv6|M_CSUM_TCPv6)) 1094 != 0) { 1095 if (IN6_NEED_CHECKSUM(ifp, 1096 m->m_pkthdr.csum_flags & 1097 (M_CSUM_UDPv6|M_CSUM_TCPv6))) { 1098 in6_delayed_cksum(m); 1099 } 1100 m->m_pkthdr.csum_flags &= ~(M_CSUM_UDPv6|M_CSUM_TCPv6); 1101 } 1102 1103 /* 1104 * Loop through length of segment after first fragment, 1105 * make new header and copy data of each part and link onto 1106 * chain. 1107 */ 1108 m0 = m; 1109 for (off = hlen; off < tlen; off += len) { 1110 struct mbuf *mlast; 1111 1112 MGETHDR(m, M_DONTWAIT, MT_HEADER); 1113 if (!m) { 1114 error = ENOBUFS; 1115 ip6stat.ip6s_odropped++; 1116 goto sendorfree; 1117 } 1118 m->m_pkthdr.rcvif = NULL; 1119 m->m_flags = m0->m_flags & M_COPYFLAGS; 1120 *mnext = m; 1121 mnext = &m->m_nextpkt; 1122 m->m_data += max_linkhdr; 1123 mhip6 = mtod(m, struct ip6_hdr *); 1124 *mhip6 = *ip6; 1125 m->m_len = sizeof(*mhip6); 1126 error = ip6_insertfraghdr(m0, m, hlen, &ip6f); 1127 if (error) { 1128 ip6stat.ip6s_odropped++; 1129 goto sendorfree; 1130 } 1131 ip6f->ip6f_offlg = htons((u_int16_t)((off - hlen) & ~7)); 1132 if (off + len >= tlen) 1133 len = tlen - off; 1134 else 1135 ip6f->ip6f_offlg |= IP6F_MORE_FRAG; 1136 mhip6->ip6_plen = htons((u_int16_t)(len + hlen + 1137 sizeof(*ip6f) - sizeof(struct ip6_hdr))); 1138 if ((m_frgpart = m_copy(m0, off, len)) == 0) { 1139 error = ENOBUFS; 1140 ip6stat.ip6s_odropped++; 1141 goto sendorfree; 1142 } 1143 for (mlast = m; mlast->m_next; mlast = mlast->m_next) 1144 ; 1145 mlast->m_next = m_frgpart; 1146 m->m_pkthdr.len = len + hlen + sizeof(*ip6f); 1147 m->m_pkthdr.rcvif = (struct ifnet *)0; 1148 ip6f->ip6f_reserved = 0; 1149 ip6f->ip6f_ident = id; 1150 ip6f->ip6f_nxt = nextproto; 1151 ip6stat.ip6s_ofragments++; 1152 in6_ifstat_inc(ifp, ifs6_out_fragcreat); 1153 } 1154 1155 in6_ifstat_inc(ifp, ifs6_out_fragok); 1156 } 1157 1158 /* 1159 * Remove leading garbages. 1160 */ 1161 sendorfree: 1162 m = m0->m_nextpkt; 1163 m0->m_nextpkt = 0; 1164 m_freem(m0); 1165 for (m0 = m; m; m = m0) { 1166 m0 = m->m_nextpkt; 1167 m->m_nextpkt = 0; 1168 if (error == 0) { 1169 struct in6_ifaddr *ia6; 1170 ip6 = mtod(m, struct ip6_hdr *); 1171 ia6 = in6_ifawithifp(ifp, &ip6->ip6_src); 1172 if (ia6) { 1173 /* 1174 * Record statistics for this interface 1175 * address. 1176 */ 1177 ia6->ia_ifa.ifa_data.ifad_outbytes += 1178 m->m_pkthdr.len; 1179 } 1180 #ifdef IPSEC 1181 /* clean ipsec history once it goes out of the node */ 1182 ipsec_delaux(m); 1183 #endif 1184 KASSERT(dst != NULL); 1185 error = nd6_output(ifp, origifp, m, dst, rt); 1186 } else 1187 m_freem(m); 1188 } 1189 1190 if (error == 0) 1191 ip6stat.ip6s_fragmented++; 1192 1193 done: 1194 /* XXX Second if is invariant? */ 1195 if (ro == &ip6route) 1196 rtcache_free(ro); 1197 else if (ro_pmtu == &ip6route) 1198 rtcache_free(ro_pmtu); 1199 1200 #ifdef IPSEC 1201 if (sp != NULL) 1202 key_freesp(sp); 1203 #endif /* IPSEC */ 1204 #ifdef FAST_IPSEC 1205 if (sp != NULL) 1206 KEY_FREESP(&sp); 1207 #endif /* FAST_IPSEC */ 1208 1209 1210 return (error); 1211 1212 freehdrs: 1213 m_freem(exthdrs.ip6e_hbh); /* m_freem will check if mbuf is 0 */ 1214 m_freem(exthdrs.ip6e_dest1); 1215 m_freem(exthdrs.ip6e_rthdr); 1216 m_freem(exthdrs.ip6e_dest2); 1217 /* FALLTHROUGH */ 1218 bad: 1219 m_freem(m); 1220 goto done; 1221 badscope: 1222 ip6stat.ip6s_badscope++; 1223 in6_ifstat_inc(origifp, ifs6_out_discard); 1224 if (error == 0) 1225 error = EHOSTUNREACH; /* XXX */ 1226 goto bad; 1227 } 1228 1229 static int 1230 ip6_copyexthdr(mp, hdr, hlen) 1231 struct mbuf **mp; 1232 void *hdr; 1233 int hlen; 1234 { 1235 struct mbuf *m; 1236 1237 if (hlen > MCLBYTES) 1238 return (ENOBUFS); /* XXX */ 1239 1240 MGET(m, M_DONTWAIT, MT_DATA); 1241 if (!m) 1242 return (ENOBUFS); 1243 1244 if (hlen > MLEN) { 1245 MCLGET(m, M_DONTWAIT); 1246 if ((m->m_flags & M_EXT) == 0) { 1247 m_free(m); 1248 return (ENOBUFS); 1249 } 1250 } 1251 m->m_len = hlen; 1252 if (hdr) 1253 bcopy(hdr, mtod(m, void *), hlen); 1254 1255 *mp = m; 1256 return (0); 1257 } 1258 1259 /* 1260 * Process a delayed payload checksum calculation. 1261 */ 1262 void 1263 in6_delayed_cksum(struct mbuf *m) 1264 { 1265 uint16_t csum, offset; 1266 1267 KASSERT((m->m_pkthdr.csum_flags & (M_CSUM_UDPv6|M_CSUM_TCPv6)) != 0); 1268 KASSERT((~m->m_pkthdr.csum_flags & (M_CSUM_UDPv6|M_CSUM_TCPv6)) != 0); 1269 KASSERT((m->m_pkthdr.csum_flags 1270 & (M_CSUM_UDPv4|M_CSUM_TCPv4|M_CSUM_TSOv4)) == 0); 1271 1272 offset = M_CSUM_DATA_IPv6_HL(m->m_pkthdr.csum_data); 1273 csum = in6_cksum(m, 0, offset, m->m_pkthdr.len - offset); 1274 if (csum == 0 && (m->m_pkthdr.csum_flags & M_CSUM_UDPv6) != 0) { 1275 csum = 0xffff; 1276 } 1277 1278 offset += M_CSUM_DATA_IPv6_OFFSET(m->m_pkthdr.csum_data); 1279 if ((offset + sizeof(csum)) > m->m_len) { 1280 m_copyback(m, offset, sizeof(csum), &csum); 1281 } else { 1282 *(uint16_t *)(mtod(m, char *) + offset) = csum; 1283 } 1284 } 1285 1286 /* 1287 * Insert jumbo payload option. 1288 */ 1289 static int 1290 ip6_insert_jumboopt(exthdrs, plen) 1291 struct ip6_exthdrs *exthdrs; 1292 u_int32_t plen; 1293 { 1294 struct mbuf *mopt; 1295 u_int8_t *optbuf; 1296 u_int32_t v; 1297 1298 #define JUMBOOPTLEN 8 /* length of jumbo payload option and padding */ 1299 1300 /* 1301 * If there is no hop-by-hop options header, allocate new one. 1302 * If there is one but it doesn't have enough space to store the 1303 * jumbo payload option, allocate a cluster to store the whole options. 1304 * Otherwise, use it to store the options. 1305 */ 1306 if (exthdrs->ip6e_hbh == 0) { 1307 MGET(mopt, M_DONTWAIT, MT_DATA); 1308 if (mopt == 0) 1309 return (ENOBUFS); 1310 mopt->m_len = JUMBOOPTLEN; 1311 optbuf = mtod(mopt, u_int8_t *); 1312 optbuf[1] = 0; /* = ((JUMBOOPTLEN) >> 3) - 1 */ 1313 exthdrs->ip6e_hbh = mopt; 1314 } else { 1315 struct ip6_hbh *hbh; 1316 1317 mopt = exthdrs->ip6e_hbh; 1318 if (M_TRAILINGSPACE(mopt) < JUMBOOPTLEN) { 1319 /* 1320 * XXX assumption: 1321 * - exthdrs->ip6e_hbh is not referenced from places 1322 * other than exthdrs. 1323 * - exthdrs->ip6e_hbh is not an mbuf chain. 1324 */ 1325 int oldoptlen = mopt->m_len; 1326 struct mbuf *n; 1327 1328 /* 1329 * XXX: give up if the whole (new) hbh header does 1330 * not fit even in an mbuf cluster. 1331 */ 1332 if (oldoptlen + JUMBOOPTLEN > MCLBYTES) 1333 return (ENOBUFS); 1334 1335 /* 1336 * As a consequence, we must always prepare a cluster 1337 * at this point. 1338 */ 1339 MGET(n, M_DONTWAIT, MT_DATA); 1340 if (n) { 1341 MCLGET(n, M_DONTWAIT); 1342 if ((n->m_flags & M_EXT) == 0) { 1343 m_freem(n); 1344 n = NULL; 1345 } 1346 } 1347 if (!n) 1348 return (ENOBUFS); 1349 n->m_len = oldoptlen + JUMBOOPTLEN; 1350 bcopy(mtod(mopt, void *), mtod(n, void *), 1351 oldoptlen); 1352 optbuf = mtod(n, u_int8_t *) + oldoptlen; 1353 m_freem(mopt); 1354 mopt = exthdrs->ip6e_hbh = n; 1355 } else { 1356 optbuf = mtod(mopt, u_int8_t *) + mopt->m_len; 1357 mopt->m_len += JUMBOOPTLEN; 1358 } 1359 optbuf[0] = IP6OPT_PADN; 1360 optbuf[1] = 0; 1361 1362 /* 1363 * Adjust the header length according to the pad and 1364 * the jumbo payload option. 1365 */ 1366 hbh = mtod(mopt, struct ip6_hbh *); 1367 hbh->ip6h_len += (JUMBOOPTLEN >> 3); 1368 } 1369 1370 /* fill in the option. */ 1371 optbuf[2] = IP6OPT_JUMBO; 1372 optbuf[3] = 4; 1373 v = (u_int32_t)htonl(plen + JUMBOOPTLEN); 1374 bcopy(&v, &optbuf[4], sizeof(u_int32_t)); 1375 1376 /* finally, adjust the packet header length */ 1377 exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN; 1378 1379 return (0); 1380 #undef JUMBOOPTLEN 1381 } 1382 1383 /* 1384 * Insert fragment header and copy unfragmentable header portions. 1385 */ 1386 static int 1387 ip6_insertfraghdr(m0, m, hlen, frghdrp) 1388 struct mbuf *m0, *m; 1389 int hlen; 1390 struct ip6_frag **frghdrp; 1391 { 1392 struct mbuf *n, *mlast; 1393 1394 if (hlen > sizeof(struct ip6_hdr)) { 1395 n = m_copym(m0, sizeof(struct ip6_hdr), 1396 hlen - sizeof(struct ip6_hdr), M_DONTWAIT); 1397 if (n == 0) 1398 return (ENOBUFS); 1399 m->m_next = n; 1400 } else 1401 n = m; 1402 1403 /* Search for the last mbuf of unfragmentable part. */ 1404 for (mlast = n; mlast->m_next; mlast = mlast->m_next) 1405 ; 1406 1407 if ((mlast->m_flags & M_EXT) == 0 && 1408 M_TRAILINGSPACE(mlast) >= sizeof(struct ip6_frag)) { 1409 /* use the trailing space of the last mbuf for the fragment hdr */ 1410 *frghdrp = (struct ip6_frag *)(mtod(mlast, char *) + 1411 mlast->m_len); 1412 mlast->m_len += sizeof(struct ip6_frag); 1413 m->m_pkthdr.len += sizeof(struct ip6_frag); 1414 } else { 1415 /* allocate a new mbuf for the fragment header */ 1416 struct mbuf *mfrg; 1417 1418 MGET(mfrg, M_DONTWAIT, MT_DATA); 1419 if (mfrg == 0) 1420 return (ENOBUFS); 1421 mfrg->m_len = sizeof(struct ip6_frag); 1422 *frghdrp = mtod(mfrg, struct ip6_frag *); 1423 mlast->m_next = mfrg; 1424 } 1425 1426 return (0); 1427 } 1428 1429 static int 1430 ip6_getpmtu(struct route *ro_pmtu, struct route *ro, struct ifnet *ifp, 1431 const struct in6_addr *dst, u_long *mtup, int *alwaysfragp) 1432 { 1433 u_int32_t mtu = 0; 1434 int alwaysfrag = 0; 1435 int error = 0; 1436 1437 if (ro_pmtu != ro) { 1438 union { 1439 struct sockaddr dst; 1440 struct sockaddr_in6 dst6; 1441 } u; 1442 1443 /* The first hop and the final destination may differ. */ 1444 sockaddr_in6_init(&u.dst6, dst, 0, 0, 0); 1445 rtcache_lookup(ro_pmtu, &u.dst); 1446 } 1447 if (ro_pmtu->ro_rt != NULL) { 1448 u_int32_t ifmtu; 1449 1450 if (ifp == NULL) 1451 ifp = ro_pmtu->ro_rt->rt_ifp; 1452 ifmtu = IN6_LINKMTU(ifp); 1453 mtu = ro_pmtu->ro_rt->rt_rmx.rmx_mtu; 1454 if (mtu == 0) 1455 mtu = ifmtu; 1456 else if (mtu < IPV6_MMTU) { 1457 /* 1458 * RFC2460 section 5, last paragraph: 1459 * if we record ICMPv6 too big message with 1460 * mtu < IPV6_MMTU, transmit packets sized IPV6_MMTU 1461 * or smaller, with fragment header attached. 1462 * (fragment header is needed regardless from the 1463 * packet size, for translators to identify packets) 1464 */ 1465 alwaysfrag = 1; 1466 mtu = IPV6_MMTU; 1467 } else if (mtu > ifmtu) { 1468 /* 1469 * The MTU on the route is larger than the MTU on 1470 * the interface! This shouldn't happen, unless the 1471 * MTU of the interface has been changed after the 1472 * interface was brought up. Change the MTU in the 1473 * route to match the interface MTU (as long as the 1474 * field isn't locked). 1475 */ 1476 mtu = ifmtu; 1477 if (!(ro_pmtu->ro_rt->rt_rmx.rmx_locks & RTV_MTU)) 1478 ro_pmtu->ro_rt->rt_rmx.rmx_mtu = mtu; 1479 } 1480 } else if (ifp) { 1481 mtu = IN6_LINKMTU(ifp); 1482 } else 1483 error = EHOSTUNREACH; /* XXX */ 1484 1485 *mtup = mtu; 1486 if (alwaysfragp) 1487 *alwaysfragp = alwaysfrag; 1488 return (error); 1489 } 1490 1491 /* 1492 * IP6 socket option processing. 1493 */ 1494 int 1495 ip6_ctloutput(int op, struct socket *so, int level, int optname, 1496 struct mbuf **mp) 1497 { 1498 int privileged, optdatalen, uproto; 1499 void *optdata; 1500 struct in6pcb *in6p = sotoin6pcb(so); 1501 struct mbuf *m = *mp; 1502 int error, optval; 1503 int optlen; 1504 struct lwp *l = curlwp; /* XXX */ 1505 1506 optlen = m ? m->m_len : 0; 1507 error = optval = 0; 1508 privileged = (l == 0 || kauth_authorize_generic(l->l_cred, 1509 KAUTH_GENERIC_ISSUSER, NULL)) ? 0 : 1; 1510 uproto = (int)so->so_proto->pr_protocol; 1511 1512 if (level == IPPROTO_IPV6) { 1513 switch (op) { 1514 case PRCO_SETOPT: 1515 switch (optname) { 1516 #ifdef RFC2292 1517 case IPV6_2292PKTOPTIONS: 1518 /* m is freed in ip6_pcbopts */ 1519 error = ip6_pcbopts(&in6p->in6p_outputopts, 1520 m, so); 1521 break; 1522 #endif 1523 1524 /* 1525 * Use of some Hop-by-Hop options or some 1526 * Destination options, might require special 1527 * privilege. That is, normal applications 1528 * (without special privilege) might be forbidden 1529 * from setting certain options in outgoing packets, 1530 * and might never see certain options in received 1531 * packets. [RFC 2292 Section 6] 1532 * KAME specific note: 1533 * KAME prevents non-privileged users from sending or 1534 * receiving ANY hbh/dst options in order to avoid 1535 * overhead of parsing options in the kernel. 1536 */ 1537 case IPV6_RECVHOPOPTS: 1538 case IPV6_RECVDSTOPTS: 1539 case IPV6_RECVRTHDRDSTOPTS: 1540 if (!privileged) { 1541 error = EPERM; 1542 break; 1543 } 1544 /* FALLTHROUGH */ 1545 case IPV6_UNICAST_HOPS: 1546 case IPV6_HOPLIMIT: 1547 case IPV6_FAITH: 1548 1549 case IPV6_RECVPKTINFO: 1550 case IPV6_RECVHOPLIMIT: 1551 case IPV6_RECVRTHDR: 1552 case IPV6_RECVPATHMTU: 1553 case IPV6_RECVTCLASS: 1554 case IPV6_V6ONLY: 1555 if (optlen != sizeof(int)) { 1556 error = EINVAL; 1557 break; 1558 } 1559 optval = *mtod(m, int *); 1560 switch (optname) { 1561 1562 case IPV6_UNICAST_HOPS: 1563 if (optval < -1 || optval >= 256) 1564 error = EINVAL; 1565 else { 1566 /* -1 = kernel default */ 1567 in6p->in6p_hops = optval; 1568 } 1569 break; 1570 #define OPTSET(bit) \ 1571 do { \ 1572 if (optval) \ 1573 in6p->in6p_flags |= (bit); \ 1574 else \ 1575 in6p->in6p_flags &= ~(bit); \ 1576 } while (/*CONSTCOND*/ 0) 1577 1578 #ifdef RFC2292 1579 #define OPTSET2292(bit) \ 1580 do { \ 1581 in6p->in6p_flags |= IN6P_RFC2292; \ 1582 if (optval) \ 1583 in6p->in6p_flags |= (bit); \ 1584 else \ 1585 in6p->in6p_flags &= ~(bit); \ 1586 } while (/*CONSTCOND*/ 0) 1587 #endif 1588 1589 #define OPTBIT(bit) (in6p->in6p_flags & (bit) ? 1 : 0) 1590 1591 case IPV6_RECVPKTINFO: 1592 #ifdef RFC2292 1593 /* cannot mix with RFC2292 */ 1594 if (OPTBIT(IN6P_RFC2292)) { 1595 error = EINVAL; 1596 break; 1597 } 1598 #endif 1599 OPTSET(IN6P_PKTINFO); 1600 break; 1601 1602 case IPV6_HOPLIMIT: 1603 { 1604 struct ip6_pktopts **optp; 1605 1606 #ifdef RFC2292 1607 /* cannot mix with RFC2292 */ 1608 if (OPTBIT(IN6P_RFC2292)) { 1609 error = EINVAL; 1610 break; 1611 } 1612 #endif 1613 optp = &in6p->in6p_outputopts; 1614 error = ip6_pcbopt(IPV6_HOPLIMIT, 1615 (u_char *)&optval, 1616 sizeof(optval), 1617 optp, 1618 privileged, uproto); 1619 break; 1620 } 1621 1622 case IPV6_RECVHOPLIMIT: 1623 #ifdef RFC2292 1624 /* cannot mix with RFC2292 */ 1625 if (OPTBIT(IN6P_RFC2292)) { 1626 error = EINVAL; 1627 break; 1628 } 1629 #endif 1630 OPTSET(IN6P_HOPLIMIT); 1631 break; 1632 1633 case IPV6_RECVHOPOPTS: 1634 #ifdef RFC2292 1635 /* cannot mix with RFC2292 */ 1636 if (OPTBIT(IN6P_RFC2292)) { 1637 error = EINVAL; 1638 break; 1639 } 1640 #endif 1641 OPTSET(IN6P_HOPOPTS); 1642 break; 1643 1644 case IPV6_RECVDSTOPTS: 1645 #ifdef RFC2292 1646 /* cannot mix with RFC2292 */ 1647 if (OPTBIT(IN6P_RFC2292)) { 1648 error = EINVAL; 1649 break; 1650 } 1651 #endif 1652 OPTSET(IN6P_DSTOPTS); 1653 break; 1654 1655 case IPV6_RECVRTHDRDSTOPTS: 1656 #ifdef RFC2292 1657 /* cannot mix with RFC2292 */ 1658 if (OPTBIT(IN6P_RFC2292)) { 1659 error = EINVAL; 1660 break; 1661 } 1662 #endif 1663 OPTSET(IN6P_RTHDRDSTOPTS); 1664 break; 1665 1666 case IPV6_RECVRTHDR: 1667 #ifdef RFC2292 1668 /* cannot mix with RFC2292 */ 1669 if (OPTBIT(IN6P_RFC2292)) { 1670 error = EINVAL; 1671 break; 1672 } 1673 #endif 1674 OPTSET(IN6P_RTHDR); 1675 break; 1676 1677 case IPV6_FAITH: 1678 OPTSET(IN6P_FAITH); 1679 break; 1680 1681 case IPV6_RECVPATHMTU: 1682 /* 1683 * We ignore this option for TCP 1684 * sockets. 1685 * (RFC3542 leaves this case 1686 * unspecified.) 1687 */ 1688 if (uproto != IPPROTO_TCP) 1689 OPTSET(IN6P_MTU); 1690 break; 1691 1692 case IPV6_V6ONLY: 1693 /* 1694 * make setsockopt(IPV6_V6ONLY) 1695 * available only prior to bind(2). 1696 * see ipng mailing list, Jun 22 2001. 1697 */ 1698 if (in6p->in6p_lport || 1699 !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr)) { 1700 error = EINVAL; 1701 break; 1702 } 1703 #ifdef INET6_BINDV6ONLY 1704 if (!optval) 1705 error = EINVAL; 1706 #else 1707 OPTSET(IN6P_IPV6_V6ONLY); 1708 #endif 1709 break; 1710 case IPV6_RECVTCLASS: 1711 #ifdef RFC2292 1712 /* cannot mix with RFC2292 XXX */ 1713 if (OPTBIT(IN6P_RFC2292)) { 1714 error = EINVAL; 1715 break; 1716 } 1717 #endif 1718 OPTSET(IN6P_TCLASS); 1719 break; 1720 1721 } 1722 break; 1723 1724 case IPV6_OTCLASS: 1725 { 1726 struct ip6_pktopts **optp; 1727 u_int8_t tclass; 1728 1729 if (optlen != sizeof(tclass)) { 1730 error = EINVAL; 1731 break; 1732 } 1733 tclass = *mtod(m, u_int8_t *); 1734 optp = &in6p->in6p_outputopts; 1735 error = ip6_pcbopt(optname, 1736 (u_char *)&tclass, 1737 sizeof(tclass), 1738 optp, 1739 privileged, uproto); 1740 break; 1741 } 1742 1743 case IPV6_TCLASS: 1744 case IPV6_DONTFRAG: 1745 case IPV6_USE_MIN_MTU: 1746 if (optlen != sizeof(optval)) { 1747 error = EINVAL; 1748 break; 1749 } 1750 optval = *mtod(m, int *); 1751 { 1752 struct ip6_pktopts **optp; 1753 optp = &in6p->in6p_outputopts; 1754 error = ip6_pcbopt(optname, 1755 (u_char *)&optval, 1756 sizeof(optval), 1757 optp, 1758 privileged, uproto); 1759 break; 1760 } 1761 1762 #ifdef RFC2292 1763 case IPV6_2292PKTINFO: 1764 case IPV6_2292HOPLIMIT: 1765 case IPV6_2292HOPOPTS: 1766 case IPV6_2292DSTOPTS: 1767 case IPV6_2292RTHDR: 1768 /* RFC 2292 */ 1769 if (optlen != sizeof(int)) { 1770 error = EINVAL; 1771 break; 1772 } 1773 optval = *mtod(m, int *); 1774 switch (optname) { 1775 case IPV6_2292PKTINFO: 1776 OPTSET2292(IN6P_PKTINFO); 1777 break; 1778 case IPV6_2292HOPLIMIT: 1779 OPTSET2292(IN6P_HOPLIMIT); 1780 break; 1781 case IPV6_2292HOPOPTS: 1782 /* 1783 * Check super-user privilege. 1784 * See comments for IPV6_RECVHOPOPTS. 1785 */ 1786 if (!privileged) 1787 return (EPERM); 1788 OPTSET2292(IN6P_HOPOPTS); 1789 break; 1790 case IPV6_2292DSTOPTS: 1791 if (!privileged) 1792 return (EPERM); 1793 OPTSET2292(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); /* XXX */ 1794 break; 1795 case IPV6_2292RTHDR: 1796 OPTSET2292(IN6P_RTHDR); 1797 break; 1798 } 1799 break; 1800 #endif 1801 case IPV6_PKTINFO: 1802 case IPV6_HOPOPTS: 1803 case IPV6_RTHDR: 1804 case IPV6_DSTOPTS: 1805 case IPV6_RTHDRDSTOPTS: 1806 case IPV6_NEXTHOP: 1807 { 1808 /* new advanced API (RFC3542) */ 1809 u_char *optbuf; 1810 int optbuflen; 1811 struct ip6_pktopts **optp; 1812 if (!m) { 1813 error = EINVAL; 1814 break; 1815 } 1816 1817 #ifdef RFC2292 1818 /* cannot mix with RFC2292 */ 1819 if (OPTBIT(IN6P_RFC2292)) { 1820 error = EINVAL; 1821 break; 1822 } 1823 #endif 1824 1825 if (m && m->m_next) { 1826 error = EINVAL; /* XXX */ 1827 break; 1828 } 1829 1830 optbuf = mtod(m, u_char *); 1831 optbuflen = m->m_len; 1832 optp = &in6p->in6p_outputopts; 1833 error = ip6_pcbopt(optname, optbuf, optbuflen, 1834 optp, privileged, uproto); 1835 break; 1836 } 1837 #undef OPTSET 1838 1839 case IPV6_MULTICAST_IF: 1840 case IPV6_MULTICAST_HOPS: 1841 case IPV6_MULTICAST_LOOP: 1842 case IPV6_JOIN_GROUP: 1843 case IPV6_LEAVE_GROUP: 1844 error = ip6_setmoptions(optname, 1845 &in6p->in6p_moptions, m); 1846 break; 1847 1848 case IPV6_PORTRANGE: 1849 if (!m) { 1850 error = EINVAL; 1851 break; 1852 } 1853 optval = *mtod(m, int *); 1854 1855 switch (optval) { 1856 case IPV6_PORTRANGE_DEFAULT: 1857 in6p->in6p_flags &= ~(IN6P_LOWPORT); 1858 in6p->in6p_flags &= ~(IN6P_HIGHPORT); 1859 break; 1860 1861 case IPV6_PORTRANGE_HIGH: 1862 in6p->in6p_flags &= ~(IN6P_LOWPORT); 1863 in6p->in6p_flags |= IN6P_HIGHPORT; 1864 break; 1865 1866 case IPV6_PORTRANGE_LOW: 1867 in6p->in6p_flags &= ~(IN6P_HIGHPORT); 1868 in6p->in6p_flags |= IN6P_LOWPORT; 1869 break; 1870 1871 default: 1872 error = EINVAL; 1873 break; 1874 } 1875 break; 1876 1877 1878 #if defined(IPSEC) || defined(FAST_IPSEC) 1879 case IPV6_IPSEC_POLICY: 1880 { 1881 void *req = NULL; 1882 size_t len = 0; 1883 if (m) { 1884 req = mtod(m, void *); 1885 len = m->m_len; 1886 } 1887 error = ipsec6_set_policy(in6p, optname, req, 1888 len, privileged); 1889 } 1890 break; 1891 #endif /* IPSEC */ 1892 1893 default: 1894 error = ENOPROTOOPT; 1895 break; 1896 } 1897 if (m) 1898 (void)m_free(m); 1899 break; 1900 1901 case PRCO_GETOPT: 1902 switch (optname) { 1903 #ifdef RFC2292 1904 case IPV6_2292PKTOPTIONS: 1905 /* 1906 * RFC3542 (effectively) deprecated the 1907 * semantics of the 2292-style pktoptions. 1908 * Since it was not reliable in nature (i.e., 1909 * applications had to expect the lack of some 1910 * information after all), it would make sense 1911 * to simplify this part by always returning 1912 * empty data. 1913 */ 1914 *mp = m_get(M_WAIT, MT_SOOPTS); 1915 (*mp)->m_len = 0; 1916 break; 1917 #endif 1918 1919 case IPV6_RECVHOPOPTS: 1920 case IPV6_RECVDSTOPTS: 1921 case IPV6_RECVRTHDRDSTOPTS: 1922 case IPV6_UNICAST_HOPS: 1923 case IPV6_RECVPKTINFO: 1924 case IPV6_RECVHOPLIMIT: 1925 case IPV6_RECVRTHDR: 1926 case IPV6_RECVPATHMTU: 1927 1928 case IPV6_FAITH: 1929 case IPV6_V6ONLY: 1930 case IPV6_PORTRANGE: 1931 case IPV6_RECVTCLASS: 1932 switch (optname) { 1933 1934 case IPV6_RECVHOPOPTS: 1935 optval = OPTBIT(IN6P_HOPOPTS); 1936 break; 1937 1938 case IPV6_RECVDSTOPTS: 1939 optval = OPTBIT(IN6P_DSTOPTS); 1940 break; 1941 1942 case IPV6_RECVRTHDRDSTOPTS: 1943 optval = OPTBIT(IN6P_RTHDRDSTOPTS); 1944 break; 1945 1946 case IPV6_UNICAST_HOPS: 1947 optval = in6p->in6p_hops; 1948 break; 1949 1950 case IPV6_RECVPKTINFO: 1951 optval = OPTBIT(IN6P_PKTINFO); 1952 break; 1953 1954 case IPV6_RECVHOPLIMIT: 1955 optval = OPTBIT(IN6P_HOPLIMIT); 1956 break; 1957 1958 case IPV6_RECVRTHDR: 1959 optval = OPTBIT(IN6P_RTHDR); 1960 break; 1961 1962 case IPV6_RECVPATHMTU: 1963 optval = OPTBIT(IN6P_MTU); 1964 break; 1965 1966 case IPV6_FAITH: 1967 optval = OPTBIT(IN6P_FAITH); 1968 break; 1969 1970 case IPV6_V6ONLY: 1971 optval = OPTBIT(IN6P_IPV6_V6ONLY); 1972 break; 1973 1974 case IPV6_PORTRANGE: 1975 { 1976 int flags; 1977 flags = in6p->in6p_flags; 1978 if (flags & IN6P_HIGHPORT) 1979 optval = IPV6_PORTRANGE_HIGH; 1980 else if (flags & IN6P_LOWPORT) 1981 optval = IPV6_PORTRANGE_LOW; 1982 else 1983 optval = 0; 1984 break; 1985 } 1986 case IPV6_RECVTCLASS: 1987 optval = OPTBIT(IN6P_TCLASS); 1988 break; 1989 1990 } 1991 if (error) 1992 break; 1993 *mp = m = m_get(M_WAIT, MT_SOOPTS); 1994 m->m_len = sizeof(int); 1995 *mtod(m, int *) = optval; 1996 break; 1997 1998 case IPV6_PATHMTU: 1999 { 2000 u_long pmtu = 0; 2001 struct ip6_mtuinfo mtuinfo; 2002 struct route *ro = &in6p->in6p_route; 2003 2004 if (!(so->so_state & SS_ISCONNECTED)) 2005 return (ENOTCONN); 2006 /* 2007 * XXX: we dot not consider the case of source 2008 * routing, or optional information to specify 2009 * the outgoing interface. 2010 */ 2011 error = ip6_getpmtu(ro, NULL, NULL, 2012 &in6p->in6p_faddr, &pmtu, NULL); 2013 if (error) 2014 break; 2015 if (pmtu > IPV6_MAXPACKET) 2016 pmtu = IPV6_MAXPACKET; 2017 2018 memset(&mtuinfo, 0, sizeof(mtuinfo)); 2019 mtuinfo.ip6m_mtu = (u_int32_t)pmtu; 2020 optdata = (void *)&mtuinfo; 2021 optdatalen = sizeof(mtuinfo); 2022 if (optdatalen > MCLBYTES) 2023 return (EMSGSIZE); /* XXX */ 2024 *mp = m = m_get(M_WAIT, MT_SOOPTS); 2025 if (optdatalen > MLEN) 2026 MCLGET(m, M_WAIT); 2027 m->m_len = optdatalen; 2028 memcpy(mtod(m, void *), optdata, optdatalen); 2029 break; 2030 } 2031 2032 #ifdef RFC2292 2033 case IPV6_2292PKTINFO: 2034 case IPV6_2292HOPLIMIT: 2035 case IPV6_2292HOPOPTS: 2036 case IPV6_2292RTHDR: 2037 case IPV6_2292DSTOPTS: 2038 switch (optname) { 2039 case IPV6_2292PKTINFO: 2040 optval = OPTBIT(IN6P_PKTINFO); 2041 break; 2042 case IPV6_2292HOPLIMIT: 2043 optval = OPTBIT(IN6P_HOPLIMIT); 2044 break; 2045 case IPV6_2292HOPOPTS: 2046 optval = OPTBIT(IN6P_HOPOPTS); 2047 break; 2048 case IPV6_2292RTHDR: 2049 optval = OPTBIT(IN6P_RTHDR); 2050 break; 2051 case IPV6_2292DSTOPTS: 2052 optval = OPTBIT(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); 2053 break; 2054 } 2055 *mp = m = m_get(M_WAIT, MT_SOOPTS); 2056 m->m_len = sizeof(int); 2057 *mtod(m, int *) = optval; 2058 break; 2059 #endif 2060 case IPV6_PKTINFO: 2061 case IPV6_HOPOPTS: 2062 case IPV6_RTHDR: 2063 case IPV6_DSTOPTS: 2064 case IPV6_RTHDRDSTOPTS: 2065 case IPV6_NEXTHOP: 2066 case IPV6_OTCLASS: 2067 case IPV6_TCLASS: 2068 case IPV6_DONTFRAG: 2069 case IPV6_USE_MIN_MTU: 2070 error = ip6_getpcbopt(in6p->in6p_outputopts, 2071 optname, mp); 2072 break; 2073 2074 case IPV6_MULTICAST_IF: 2075 case IPV6_MULTICAST_HOPS: 2076 case IPV6_MULTICAST_LOOP: 2077 case IPV6_JOIN_GROUP: 2078 case IPV6_LEAVE_GROUP: 2079 error = ip6_getmoptions(optname, 2080 in6p->in6p_moptions, mp); 2081 break; 2082 2083 #if defined(IPSEC) || defined(FAST_IPSEC) 2084 case IPV6_IPSEC_POLICY: 2085 { 2086 void *req = NULL; 2087 size_t len = 0; 2088 if (m) { 2089 req = mtod(m, void *); 2090 len = m->m_len; 2091 } 2092 error = ipsec6_get_policy(in6p, req, len, mp); 2093 break; 2094 } 2095 #endif /* IPSEC */ 2096 2097 2098 2099 2100 default: 2101 error = ENOPROTOOPT; 2102 break; 2103 } 2104 break; 2105 } 2106 } else { 2107 error = EINVAL; 2108 if (op == PRCO_SETOPT && *mp) 2109 (void)m_free(*mp); 2110 } 2111 return (error); 2112 } 2113 2114 int 2115 ip6_raw_ctloutput(op, so, level, optname, mp) 2116 int op; 2117 struct socket *so; 2118 int level, optname; 2119 struct mbuf **mp; 2120 { 2121 int error = 0, optval, optlen; 2122 const int icmp6off = offsetof(struct icmp6_hdr, icmp6_cksum); 2123 struct in6pcb *in6p = sotoin6pcb(so); 2124 struct mbuf *m = *mp; 2125 2126 optlen = m ? m->m_len : 0; 2127 2128 if (level != IPPROTO_IPV6) { 2129 if (op == PRCO_SETOPT && *mp) 2130 (void)m_free(*mp); 2131 return (EINVAL); 2132 } 2133 2134 switch (optname) { 2135 case IPV6_CHECKSUM: 2136 /* 2137 * For ICMPv6 sockets, no modification allowed for checksum 2138 * offset, permit "no change" values to help existing apps. 2139 * 2140 * XXX RFC3542 says: "An attempt to set IPV6_CHECKSUM 2141 * for an ICMPv6 socket will fail." The current 2142 * behavior does not meet RFC3542. 2143 */ 2144 switch (op) { 2145 case PRCO_SETOPT: 2146 if (optlen != sizeof(int)) { 2147 error = EINVAL; 2148 break; 2149 } 2150 optval = *mtod(m, int *); 2151 if ((optval % 2) != 0) { 2152 /* the API assumes even offset values */ 2153 error = EINVAL; 2154 } else if (so->so_proto->pr_protocol == 2155 IPPROTO_ICMPV6) { 2156 if (optval != icmp6off) 2157 error = EINVAL; 2158 } else 2159 in6p->in6p_cksum = optval; 2160 break; 2161 2162 case PRCO_GETOPT: 2163 if (so->so_proto->pr_protocol == IPPROTO_ICMPV6) 2164 optval = icmp6off; 2165 else 2166 optval = in6p->in6p_cksum; 2167 2168 *mp = m = m_get(M_WAIT, MT_SOOPTS); 2169 m->m_len = sizeof(int); 2170 *mtod(m, int *) = optval; 2171 break; 2172 2173 default: 2174 error = EINVAL; 2175 break; 2176 } 2177 break; 2178 2179 default: 2180 error = ENOPROTOOPT; 2181 break; 2182 } 2183 2184 if (op == PRCO_SETOPT && m) 2185 (void)m_free(m); 2186 2187 return (error); 2188 } 2189 2190 #ifdef RFC2292 2191 /* 2192 * Set up IP6 options in pcb for insertion in output packets or 2193 * specifying behavior of outgoing packets. 2194 */ 2195 static int 2196 ip6_pcbopts(pktopt, m, so) 2197 struct ip6_pktopts **pktopt; 2198 struct mbuf *m; 2199 struct socket *so; 2200 { 2201 struct ip6_pktopts *opt = *pktopt; 2202 int error = 0; 2203 struct lwp *l = curlwp; /* XXX */ 2204 int priv = 0; 2205 2206 /* turn off any old options. */ 2207 if (opt) { 2208 #ifdef DIAGNOSTIC 2209 if (opt->ip6po_pktinfo || opt->ip6po_nexthop || 2210 opt->ip6po_hbh || opt->ip6po_dest1 || opt->ip6po_dest2 || 2211 opt->ip6po_rhinfo.ip6po_rhi_rthdr) 2212 printf("ip6_pcbopts: all specified options are cleared.\n"); 2213 #endif 2214 ip6_clearpktopts(opt, -1); 2215 } else 2216 opt = malloc(sizeof(*opt), M_IP6OPT, M_WAITOK); 2217 *pktopt = NULL; 2218 2219 if (!m || m->m_len == 0) { 2220 /* 2221 * Only turning off any previous options, regardless of 2222 * whether the opt is just created or given. 2223 */ 2224 free(opt, M_IP6OPT); 2225 return (0); 2226 } 2227 2228 /* set options specified by user. */ 2229 if (l && !kauth_authorize_generic(l->l_cred, KAUTH_GENERIC_ISSUSER, 2230 NULL)) 2231 priv = 1; 2232 if ((error = ip6_setpktopts(m, opt, NULL, priv, 2233 so->so_proto->pr_protocol)) != 0) { 2234 ip6_clearpktopts(opt, -1); /* XXX: discard all options */ 2235 free(opt, M_IP6OPT); 2236 return (error); 2237 } 2238 *pktopt = opt; 2239 return (0); 2240 } 2241 #endif 2242 2243 /* 2244 * initialize ip6_pktopts. beware that there are non-zero default values in 2245 * the struct. 2246 */ 2247 void 2248 ip6_initpktopts(struct ip6_pktopts *opt) 2249 { 2250 2251 memset(opt, 0, sizeof(*opt)); 2252 opt->ip6po_hlim = -1; /* -1 means default hop limit */ 2253 opt->ip6po_tclass = -1; /* -1 means default traffic class */ 2254 opt->ip6po_minmtu = IP6PO_MINMTU_MCASTONLY; 2255 } 2256 2257 #define sin6tosa(sin6) ((struct sockaddr *)(sin6)) /* XXX */ 2258 static int 2259 ip6_pcbopt(int optname, u_char *buf, int len, struct ip6_pktopts **pktopt, 2260 int priv, int uproto) 2261 { 2262 struct ip6_pktopts *opt; 2263 2264 if (*pktopt == NULL) { 2265 *pktopt = malloc(sizeof(struct ip6_pktopts), M_IP6OPT, 2266 M_WAITOK); 2267 ip6_initpktopts(*pktopt); 2268 } 2269 opt = *pktopt; 2270 2271 return (ip6_setpktopt(optname, buf, len, opt, priv, 1, 0, uproto)); 2272 } 2273 2274 static int 2275 ip6_getpcbopt(struct ip6_pktopts *pktopt, int optname, struct mbuf **mp) 2276 { 2277 void *optdata = NULL; 2278 int optdatalen = 0; 2279 struct ip6_ext *ip6e; 2280 int error = 0; 2281 struct in6_pktinfo null_pktinfo; 2282 int deftclass = 0, on; 2283 int defminmtu = IP6PO_MINMTU_MCASTONLY; 2284 struct mbuf *m; 2285 2286 switch (optname) { 2287 case IPV6_PKTINFO: 2288 if (pktopt && pktopt->ip6po_pktinfo) 2289 optdata = (void *)pktopt->ip6po_pktinfo; 2290 else { 2291 /* XXX: we don't have to do this every time... */ 2292 memset(&null_pktinfo, 0, sizeof(null_pktinfo)); 2293 optdata = (void *)&null_pktinfo; 2294 } 2295 optdatalen = sizeof(struct in6_pktinfo); 2296 break; 2297 case IPV6_OTCLASS: 2298 /* XXX */ 2299 return (EINVAL); 2300 case IPV6_TCLASS: 2301 if (pktopt && pktopt->ip6po_tclass >= 0) 2302 optdata = (void *)&pktopt->ip6po_tclass; 2303 else 2304 optdata = (void *)&deftclass; 2305 optdatalen = sizeof(int); 2306 break; 2307 case IPV6_HOPOPTS: 2308 if (pktopt && pktopt->ip6po_hbh) { 2309 optdata = (void *)pktopt->ip6po_hbh; 2310 ip6e = (struct ip6_ext *)pktopt->ip6po_hbh; 2311 optdatalen = (ip6e->ip6e_len + 1) << 3; 2312 } 2313 break; 2314 case IPV6_RTHDR: 2315 if (pktopt && pktopt->ip6po_rthdr) { 2316 optdata = (void *)pktopt->ip6po_rthdr; 2317 ip6e = (struct ip6_ext *)pktopt->ip6po_rthdr; 2318 optdatalen = (ip6e->ip6e_len + 1) << 3; 2319 } 2320 break; 2321 case IPV6_RTHDRDSTOPTS: 2322 if (pktopt && pktopt->ip6po_dest1) { 2323 optdata = (void *)pktopt->ip6po_dest1; 2324 ip6e = (struct ip6_ext *)pktopt->ip6po_dest1; 2325 optdatalen = (ip6e->ip6e_len + 1) << 3; 2326 } 2327 break; 2328 case IPV6_DSTOPTS: 2329 if (pktopt && pktopt->ip6po_dest2) { 2330 optdata = (void *)pktopt->ip6po_dest2; 2331 ip6e = (struct ip6_ext *)pktopt->ip6po_dest2; 2332 optdatalen = (ip6e->ip6e_len + 1) << 3; 2333 } 2334 break; 2335 case IPV6_NEXTHOP: 2336 if (pktopt && pktopt->ip6po_nexthop) { 2337 optdata = (void *)pktopt->ip6po_nexthop; 2338 optdatalen = pktopt->ip6po_nexthop->sa_len; 2339 } 2340 break; 2341 case IPV6_USE_MIN_MTU: 2342 if (pktopt) 2343 optdata = (void *)&pktopt->ip6po_minmtu; 2344 else 2345 optdata = (void *)&defminmtu; 2346 optdatalen = sizeof(int); 2347 break; 2348 case IPV6_DONTFRAG: 2349 if (pktopt && ((pktopt->ip6po_flags) & IP6PO_DONTFRAG)) 2350 on = 1; 2351 else 2352 on = 0; 2353 optdata = (void *)&on; 2354 optdatalen = sizeof(on); 2355 break; 2356 default: /* should not happen */ 2357 #ifdef DIAGNOSTIC 2358 panic("ip6_getpcbopt: unexpected option\n"); 2359 #endif 2360 return (ENOPROTOOPT); 2361 } 2362 2363 if (optdatalen > MCLBYTES) 2364 return (EMSGSIZE); /* XXX */ 2365 *mp = m = m_get(M_WAIT, MT_SOOPTS); 2366 if (optdatalen > MLEN) 2367 MCLGET(m, M_WAIT); 2368 m->m_len = optdatalen; 2369 if (optdatalen) 2370 memcpy(mtod(m, void *), optdata, optdatalen); 2371 2372 return (error); 2373 } 2374 2375 void 2376 ip6_clearpktopts(struct ip6_pktopts *pktopt, int optname) 2377 { 2378 if (optname == -1 || optname == IPV6_PKTINFO) { 2379 if (pktopt->ip6po_pktinfo) 2380 free(pktopt->ip6po_pktinfo, M_IP6OPT); 2381 pktopt->ip6po_pktinfo = NULL; 2382 } 2383 if (optname == -1 || optname == IPV6_HOPLIMIT) 2384 pktopt->ip6po_hlim = -1; 2385 if (optname == -1 || optname == IPV6_TCLASS) 2386 pktopt->ip6po_tclass = -1; 2387 if (optname == -1 || optname == IPV6_NEXTHOP) { 2388 rtcache_free(&pktopt->ip6po_nextroute); 2389 if (pktopt->ip6po_nexthop) 2390 free(pktopt->ip6po_nexthop, M_IP6OPT); 2391 pktopt->ip6po_nexthop = NULL; 2392 } 2393 if (optname == -1 || optname == IPV6_HOPOPTS) { 2394 if (pktopt->ip6po_hbh) 2395 free(pktopt->ip6po_hbh, M_IP6OPT); 2396 pktopt->ip6po_hbh = NULL; 2397 } 2398 if (optname == -1 || optname == IPV6_RTHDRDSTOPTS) { 2399 if (pktopt->ip6po_dest1) 2400 free(pktopt->ip6po_dest1, M_IP6OPT); 2401 pktopt->ip6po_dest1 = NULL; 2402 } 2403 if (optname == -1 || optname == IPV6_RTHDR) { 2404 if (pktopt->ip6po_rhinfo.ip6po_rhi_rthdr) 2405 free(pktopt->ip6po_rhinfo.ip6po_rhi_rthdr, M_IP6OPT); 2406 pktopt->ip6po_rhinfo.ip6po_rhi_rthdr = NULL; 2407 rtcache_free(&pktopt->ip6po_route); 2408 } 2409 if (optname == -1 || optname == IPV6_DSTOPTS) { 2410 if (pktopt->ip6po_dest2) 2411 free(pktopt->ip6po_dest2, M_IP6OPT); 2412 pktopt->ip6po_dest2 = NULL; 2413 } 2414 } 2415 2416 #define PKTOPT_EXTHDRCPY(type) \ 2417 do { \ 2418 if (src->type) { \ 2419 int hlen = (((struct ip6_ext *)src->type)->ip6e_len + 1) << 3;\ 2420 dst->type = malloc(hlen, M_IP6OPT, canwait); \ 2421 if (dst->type == NULL && canwait == M_NOWAIT) \ 2422 goto bad; \ 2423 memcpy(dst->type, src->type, hlen); \ 2424 } \ 2425 } while (/*CONSTCOND*/ 0) 2426 2427 static int 2428 copypktopts(struct ip6_pktopts *dst, struct ip6_pktopts *src, int canwait) 2429 { 2430 dst->ip6po_hlim = src->ip6po_hlim; 2431 dst->ip6po_tclass = src->ip6po_tclass; 2432 dst->ip6po_flags = src->ip6po_flags; 2433 if (src->ip6po_pktinfo) { 2434 dst->ip6po_pktinfo = malloc(sizeof(*dst->ip6po_pktinfo), 2435 M_IP6OPT, canwait); 2436 if (dst->ip6po_pktinfo == NULL && canwait == M_NOWAIT) 2437 goto bad; 2438 *dst->ip6po_pktinfo = *src->ip6po_pktinfo; 2439 } 2440 if (src->ip6po_nexthop) { 2441 dst->ip6po_nexthop = malloc(src->ip6po_nexthop->sa_len, 2442 M_IP6OPT, canwait); 2443 if (dst->ip6po_nexthop == NULL && canwait == M_NOWAIT) 2444 goto bad; 2445 memcpy(dst->ip6po_nexthop, src->ip6po_nexthop, 2446 src->ip6po_nexthop->sa_len); 2447 } 2448 PKTOPT_EXTHDRCPY(ip6po_hbh); 2449 PKTOPT_EXTHDRCPY(ip6po_dest1); 2450 PKTOPT_EXTHDRCPY(ip6po_dest2); 2451 PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */ 2452 return (0); 2453 2454 bad: 2455 if (dst->ip6po_pktinfo) free(dst->ip6po_pktinfo, M_IP6OPT); 2456 if (dst->ip6po_nexthop) free(dst->ip6po_nexthop, M_IP6OPT); 2457 if (dst->ip6po_hbh) free(dst->ip6po_hbh, M_IP6OPT); 2458 if (dst->ip6po_dest1) free(dst->ip6po_dest1, M_IP6OPT); 2459 if (dst->ip6po_dest2) free(dst->ip6po_dest2, M_IP6OPT); 2460 if (dst->ip6po_rthdr) free(dst->ip6po_rthdr, M_IP6OPT); 2461 2462 return (ENOBUFS); 2463 } 2464 #undef PKTOPT_EXTHDRCPY 2465 2466 struct ip6_pktopts * 2467 ip6_copypktopts(struct ip6_pktopts *src, int canwait) 2468 { 2469 int error; 2470 struct ip6_pktopts *dst; 2471 2472 dst = malloc(sizeof(*dst), M_IP6OPT, canwait); 2473 if (dst == NULL && canwait == M_NOWAIT) 2474 return (NULL); 2475 ip6_initpktopts(dst); 2476 2477 if ((error = copypktopts(dst, src, canwait)) != 0) { 2478 free(dst, M_IP6OPT); 2479 return (NULL); 2480 } 2481 2482 return (dst); 2483 } 2484 2485 void 2486 ip6_freepcbopts(struct ip6_pktopts *pktopt) 2487 { 2488 if (pktopt == NULL) 2489 return; 2490 2491 ip6_clearpktopts(pktopt, -1); 2492 2493 free(pktopt, M_IP6OPT); 2494 } 2495 2496 /* 2497 * Set the IP6 multicast options in response to user setsockopt(). 2498 */ 2499 static int 2500 ip6_setmoptions(optname, im6op, m) 2501 int optname; 2502 struct ip6_moptions **im6op; 2503 struct mbuf *m; 2504 { 2505 int error = 0; 2506 u_int loop, ifindex; 2507 struct ipv6_mreq *mreq; 2508 struct ifnet *ifp; 2509 struct ip6_moptions *im6o = *im6op; 2510 struct route ro; 2511 struct in6_multi_mship *imm; 2512 struct lwp *l = curlwp; /* XXX */ 2513 2514 if (im6o == NULL) { 2515 /* 2516 * No multicast option buffer attached to the pcb; 2517 * allocate one and initialize to default values. 2518 */ 2519 im6o = (struct ip6_moptions *) 2520 malloc(sizeof(*im6o), M_IPMOPTS, M_WAITOK); 2521 2522 if (im6o == NULL) 2523 return (ENOBUFS); 2524 *im6op = im6o; 2525 im6o->im6o_multicast_ifp = NULL; 2526 im6o->im6o_multicast_hlim = ip6_defmcasthlim; 2527 im6o->im6o_multicast_loop = IPV6_DEFAULT_MULTICAST_LOOP; 2528 LIST_INIT(&im6o->im6o_memberships); 2529 } 2530 2531 switch (optname) { 2532 2533 case IPV6_MULTICAST_IF: 2534 /* 2535 * Select the interface for outgoing multicast packets. 2536 */ 2537 if (m == NULL || m->m_len != sizeof(u_int)) { 2538 error = EINVAL; 2539 break; 2540 } 2541 bcopy(mtod(m, u_int *), &ifindex, sizeof(ifindex)); 2542 if (ifindex != 0) { 2543 if (if_indexlim <= ifindex || !ifindex2ifnet[ifindex]) { 2544 error = ENXIO; /* XXX EINVAL? */ 2545 break; 2546 } 2547 ifp = ifindex2ifnet[ifindex]; 2548 if ((ifp->if_flags & IFF_MULTICAST) == 0) { 2549 error = EADDRNOTAVAIL; 2550 break; 2551 } 2552 } else 2553 ifp = NULL; 2554 im6o->im6o_multicast_ifp = ifp; 2555 break; 2556 2557 case IPV6_MULTICAST_HOPS: 2558 { 2559 /* 2560 * Set the IP6 hoplimit for outgoing multicast packets. 2561 */ 2562 int optval; 2563 if (m == NULL || m->m_len != sizeof(int)) { 2564 error = EINVAL; 2565 break; 2566 } 2567 bcopy(mtod(m, u_int *), &optval, sizeof(optval)); 2568 if (optval < -1 || optval >= 256) 2569 error = EINVAL; 2570 else if (optval == -1) 2571 im6o->im6o_multicast_hlim = ip6_defmcasthlim; 2572 else 2573 im6o->im6o_multicast_hlim = optval; 2574 break; 2575 } 2576 2577 case IPV6_MULTICAST_LOOP: 2578 /* 2579 * Set the loopback flag for outgoing multicast packets. 2580 * Must be zero or one. 2581 */ 2582 if (m == NULL || m->m_len != sizeof(u_int)) { 2583 error = EINVAL; 2584 break; 2585 } 2586 bcopy(mtod(m, u_int *), &loop, sizeof(loop)); 2587 if (loop > 1) { 2588 error = EINVAL; 2589 break; 2590 } 2591 im6o->im6o_multicast_loop = loop; 2592 break; 2593 2594 case IPV6_JOIN_GROUP: 2595 /* 2596 * Add a multicast group membership. 2597 * Group must be a valid IP6 multicast address. 2598 */ 2599 if (m == NULL || m->m_len != sizeof(struct ipv6_mreq)) { 2600 error = EINVAL; 2601 break; 2602 } 2603 mreq = mtod(m, struct ipv6_mreq *); 2604 if (IN6_IS_ADDR_UNSPECIFIED(&mreq->ipv6mr_multiaddr)) { 2605 /* 2606 * We use the unspecified address to specify to accept 2607 * all multicast addresses. Only super user is allowed 2608 * to do this. 2609 */ 2610 if (kauth_authorize_generic(l->l_cred, 2611 KAUTH_GENERIC_ISSUSER, NULL)) 2612 { 2613 error = EACCES; 2614 break; 2615 } 2616 } else if (!IN6_IS_ADDR_MULTICAST(&mreq->ipv6mr_multiaddr)) { 2617 error = EINVAL; 2618 break; 2619 } 2620 2621 /* 2622 * If no interface was explicitly specified, choose an 2623 * appropriate one according to the given multicast address. 2624 */ 2625 if (mreq->ipv6mr_interface == 0) { 2626 union { 2627 struct sockaddr dst; 2628 struct sockaddr_in6 dst6; 2629 } u; 2630 2631 /* 2632 * Look up the routing table for the 2633 * address, and choose the outgoing interface. 2634 * XXX: is it a good approach? 2635 */ 2636 memset(&ro, 0, sizeof(ro)); 2637 sockaddr_in6_init(&u.dst6, &mreq->ipv6mr_multiaddr, 0, 2638 0, 0); 2639 rtcache_setdst(&ro, &u.dst); 2640 rtcache_init(&ro); 2641 ifp = (ro.ro_rt != NULL) ? ro.ro_rt->rt_ifp : NULL; 2642 rtcache_free(&ro); 2643 } else { 2644 /* 2645 * If the interface is specified, validate it. 2646 */ 2647 if (if_indexlim <= mreq->ipv6mr_interface || 2648 !ifindex2ifnet[mreq->ipv6mr_interface]) { 2649 error = ENXIO; /* XXX EINVAL? */ 2650 break; 2651 } 2652 ifp = ifindex2ifnet[mreq->ipv6mr_interface]; 2653 } 2654 2655 /* 2656 * See if we found an interface, and confirm that it 2657 * supports multicast 2658 */ 2659 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) { 2660 error = EADDRNOTAVAIL; 2661 break; 2662 } 2663 2664 if (in6_setscope(&mreq->ipv6mr_multiaddr, ifp, NULL)) { 2665 error = EADDRNOTAVAIL; /* XXX: should not happen */ 2666 break; 2667 } 2668 2669 /* 2670 * See if the membership already exists. 2671 */ 2672 for (imm = im6o->im6o_memberships.lh_first; 2673 imm != NULL; imm = imm->i6mm_chain.le_next) 2674 if (imm->i6mm_maddr->in6m_ifp == ifp && 2675 IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr, 2676 &mreq->ipv6mr_multiaddr)) 2677 break; 2678 if (imm != NULL) { 2679 error = EADDRINUSE; 2680 break; 2681 } 2682 /* 2683 * Everything looks good; add a new record to the multicast 2684 * address list for the given interface. 2685 */ 2686 imm = in6_joingroup(ifp, &mreq->ipv6mr_multiaddr, &error, 0); 2687 if (imm == NULL) 2688 break; 2689 LIST_INSERT_HEAD(&im6o->im6o_memberships, imm, i6mm_chain); 2690 break; 2691 2692 case IPV6_LEAVE_GROUP: 2693 /* 2694 * Drop a multicast group membership. 2695 * Group must be a valid IP6 multicast address. 2696 */ 2697 if (m == NULL || m->m_len != sizeof(struct ipv6_mreq)) { 2698 error = EINVAL; 2699 break; 2700 } 2701 mreq = mtod(m, struct ipv6_mreq *); 2702 2703 /* 2704 * If an interface address was specified, get a pointer 2705 * to its ifnet structure. 2706 */ 2707 if (mreq->ipv6mr_interface != 0) { 2708 if (if_indexlim <= mreq->ipv6mr_interface || 2709 !ifindex2ifnet[mreq->ipv6mr_interface]) { 2710 error = ENXIO; /* XXX EINVAL? */ 2711 break; 2712 } 2713 ifp = ifindex2ifnet[mreq->ipv6mr_interface]; 2714 } else 2715 ifp = NULL; 2716 2717 /* Fill in the scope zone ID */ 2718 if (ifp) { 2719 if (in6_setscope(&mreq->ipv6mr_multiaddr, ifp, NULL)) { 2720 /* XXX: should not happen */ 2721 error = EADDRNOTAVAIL; 2722 break; 2723 } 2724 } else if (mreq->ipv6mr_interface != 0) { 2725 /* 2726 * XXX: This case would happens when the (positive) 2727 * index is in the valid range, but the corresponding 2728 * interface has been detached dynamically. The above 2729 * check probably avoids such case to happen here, but 2730 * we check it explicitly for safety. 2731 */ 2732 error = EADDRNOTAVAIL; 2733 break; 2734 } else { /* ipv6mr_interface == 0 */ 2735 struct sockaddr_in6 sa6_mc; 2736 2737 /* 2738 * The API spec says as follows: 2739 * If the interface index is specified as 0, the 2740 * system may choose a multicast group membership to 2741 * drop by matching the multicast address only. 2742 * On the other hand, we cannot disambiguate the scope 2743 * zone unless an interface is provided. Thus, we 2744 * check if there's ambiguity with the default scope 2745 * zone as the last resort. 2746 */ 2747 bzero(&sa6_mc, sizeof(sa6_mc)); 2748 sa6_mc.sin6_family = AF_INET6; 2749 sa6_mc.sin6_len = sizeof(sa6_mc); 2750 sa6_mc.sin6_addr = mreq->ipv6mr_multiaddr; 2751 error = sa6_embedscope(&sa6_mc, ip6_use_defzone); 2752 if (error != 0) 2753 break; 2754 mreq->ipv6mr_multiaddr = sa6_mc.sin6_addr; 2755 } 2756 2757 /* 2758 * Find the membership in the membership list. 2759 */ 2760 for (imm = im6o->im6o_memberships.lh_first; 2761 imm != NULL; imm = imm->i6mm_chain.le_next) { 2762 if ((ifp == NULL || imm->i6mm_maddr->in6m_ifp == ifp) && 2763 IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr, 2764 &mreq->ipv6mr_multiaddr)) 2765 break; 2766 } 2767 if (imm == NULL) { 2768 /* Unable to resolve interface */ 2769 error = EADDRNOTAVAIL; 2770 break; 2771 } 2772 /* 2773 * Give up the multicast address record to which the 2774 * membership points. 2775 */ 2776 LIST_REMOVE(imm, i6mm_chain); 2777 in6_leavegroup(imm); 2778 break; 2779 2780 default: 2781 error = EOPNOTSUPP; 2782 break; 2783 } 2784 2785 /* 2786 * If all options have default values, no need to keep the mbuf. 2787 */ 2788 if (im6o->im6o_multicast_ifp == NULL && 2789 im6o->im6o_multicast_hlim == ip6_defmcasthlim && 2790 im6o->im6o_multicast_loop == IPV6_DEFAULT_MULTICAST_LOOP && 2791 im6o->im6o_memberships.lh_first == NULL) { 2792 free(*im6op, M_IPMOPTS); 2793 *im6op = NULL; 2794 } 2795 2796 return (error); 2797 } 2798 2799 /* 2800 * Return the IP6 multicast options in response to user getsockopt(). 2801 */ 2802 static int 2803 ip6_getmoptions(optname, im6o, mp) 2804 int optname; 2805 struct ip6_moptions *im6o; 2806 struct mbuf **mp; 2807 { 2808 u_int *hlim, *loop, *ifindex; 2809 2810 *mp = m_get(M_WAIT, MT_SOOPTS); 2811 2812 switch (optname) { 2813 2814 case IPV6_MULTICAST_IF: 2815 ifindex = mtod(*mp, u_int *); 2816 (*mp)->m_len = sizeof(u_int); 2817 if (im6o == NULL || im6o->im6o_multicast_ifp == NULL) 2818 *ifindex = 0; 2819 else 2820 *ifindex = im6o->im6o_multicast_ifp->if_index; 2821 return (0); 2822 2823 case IPV6_MULTICAST_HOPS: 2824 hlim = mtod(*mp, u_int *); 2825 (*mp)->m_len = sizeof(u_int); 2826 if (im6o == NULL) 2827 *hlim = ip6_defmcasthlim; 2828 else 2829 *hlim = im6o->im6o_multicast_hlim; 2830 return (0); 2831 2832 case IPV6_MULTICAST_LOOP: 2833 loop = mtod(*mp, u_int *); 2834 (*mp)->m_len = sizeof(u_int); 2835 if (im6o == NULL) 2836 *loop = ip6_defmcasthlim; 2837 else 2838 *loop = im6o->im6o_multicast_loop; 2839 return (0); 2840 2841 default: 2842 return (EOPNOTSUPP); 2843 } 2844 } 2845 2846 /* 2847 * Discard the IP6 multicast options. 2848 */ 2849 void 2850 ip6_freemoptions(im6o) 2851 struct ip6_moptions *im6o; 2852 { 2853 struct in6_multi_mship *imm; 2854 2855 if (im6o == NULL) 2856 return; 2857 2858 while ((imm = im6o->im6o_memberships.lh_first) != NULL) { 2859 LIST_REMOVE(imm, i6mm_chain); 2860 in6_leavegroup(imm); 2861 } 2862 free(im6o, M_IPMOPTS); 2863 } 2864 2865 /* 2866 * Set IPv6 outgoing packet options based on advanced API. 2867 */ 2868 int 2869 ip6_setpktopts(control, opt, stickyopt, priv, uproto) 2870 struct mbuf *control; 2871 struct ip6_pktopts *opt, *stickyopt; 2872 int priv, uproto; 2873 { 2874 struct cmsghdr *cm = 0; 2875 2876 if (control == NULL || opt == NULL) 2877 return (EINVAL); 2878 2879 ip6_initpktopts(opt); 2880 if (stickyopt) { 2881 int error; 2882 2883 /* 2884 * If stickyopt is provided, make a local copy of the options 2885 * for this particular packet, then override them by ancillary 2886 * objects. 2887 * XXX: copypktopts() does not copy the cached route to a next 2888 * hop (if any). This is not very good in terms of efficiency, 2889 * but we can allow this since this option should be rarely 2890 * used. 2891 */ 2892 if ((error = copypktopts(opt, stickyopt, M_NOWAIT)) != 0) 2893 return (error); 2894 } 2895 2896 /* 2897 * XXX: Currently, we assume all the optional information is stored 2898 * in a single mbuf. 2899 */ 2900 if (control->m_next) 2901 return (EINVAL); 2902 2903 for (; control->m_len; control->m_data += CMSG_ALIGN(cm->cmsg_len), 2904 control->m_len -= CMSG_ALIGN(cm->cmsg_len)) { 2905 int error; 2906 2907 if (control->m_len < CMSG_LEN(0)) 2908 return (EINVAL); 2909 2910 cm = mtod(control, struct cmsghdr *); 2911 if (cm->cmsg_len == 0 || cm->cmsg_len > control->m_len) 2912 return (EINVAL); 2913 if (cm->cmsg_level != IPPROTO_IPV6) 2914 continue; 2915 2916 error = ip6_setpktopt(cm->cmsg_type, CMSG_DATA(cm), 2917 cm->cmsg_len - CMSG_LEN(0), opt, priv, 0, 1, uproto); 2918 if (error) 2919 return (error); 2920 } 2921 2922 return (0); 2923 } 2924 2925 /* 2926 * Set a particular packet option, as a sticky option or an ancillary data 2927 * item. "len" can be 0 only when it's a sticky option. 2928 * We have 4 cases of combination of "sticky" and "cmsg": 2929 * "sticky=0, cmsg=0": impossible 2930 * "sticky=0, cmsg=1": RFC2292 or RFC3542 ancillary data 2931 * "sticky=1, cmsg=0": RFC3542 socket option 2932 * "sticky=1, cmsg=1": RFC2292 socket option 2933 */ 2934 static int 2935 ip6_setpktopt(int optname, u_char *buf, int len, struct ip6_pktopts *opt, 2936 int priv, int sticky, int cmsg, int uproto) 2937 { 2938 int minmtupolicy; 2939 2940 if (!sticky && !cmsg) { 2941 #ifdef DIAGNOSTIC 2942 printf("ip6_setpktopt: impossible case\n"); 2943 #endif 2944 return (EINVAL); 2945 } 2946 2947 /* 2948 * IPV6_2292xxx is for backward compatibility to RFC2292, and should 2949 * not be specified in the context of RFC3542. Conversely, 2950 * RFC3542 types should not be specified in the context of RFC2292. 2951 */ 2952 if (!cmsg) { 2953 switch (optname) { 2954 case IPV6_2292PKTINFO: 2955 case IPV6_2292HOPLIMIT: 2956 case IPV6_2292NEXTHOP: 2957 case IPV6_2292HOPOPTS: 2958 case IPV6_2292DSTOPTS: 2959 case IPV6_2292RTHDR: 2960 case IPV6_2292PKTOPTIONS: 2961 return (ENOPROTOOPT); 2962 } 2963 } 2964 if (sticky && cmsg) { 2965 switch (optname) { 2966 case IPV6_PKTINFO: 2967 case IPV6_HOPLIMIT: 2968 case IPV6_NEXTHOP: 2969 case IPV6_HOPOPTS: 2970 case IPV6_DSTOPTS: 2971 case IPV6_RTHDRDSTOPTS: 2972 case IPV6_RTHDR: 2973 case IPV6_USE_MIN_MTU: 2974 case IPV6_DONTFRAG: 2975 case IPV6_OTCLASS: 2976 case IPV6_TCLASS: 2977 return (ENOPROTOOPT); 2978 } 2979 } 2980 2981 switch (optname) { 2982 #ifdef RFC2292 2983 case IPV6_2292PKTINFO: 2984 #endif 2985 case IPV6_PKTINFO: 2986 { 2987 struct ifnet *ifp = NULL; 2988 struct in6_pktinfo *pktinfo; 2989 2990 if (len != sizeof(struct in6_pktinfo)) 2991 return (EINVAL); 2992 2993 pktinfo = (struct in6_pktinfo *)buf; 2994 2995 /* 2996 * An application can clear any sticky IPV6_PKTINFO option by 2997 * doing a "regular" setsockopt with ipi6_addr being 2998 * in6addr_any and ipi6_ifindex being zero. 2999 * [RFC 3542, Section 6] 3000 */ 3001 if (optname == IPV6_PKTINFO && opt->ip6po_pktinfo && 3002 pktinfo->ipi6_ifindex == 0 && 3003 IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) { 3004 ip6_clearpktopts(opt, optname); 3005 break; 3006 } 3007 3008 if (uproto == IPPROTO_TCP && optname == IPV6_PKTINFO && 3009 sticky && !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) { 3010 return (EINVAL); 3011 } 3012 3013 /* validate the interface index if specified. */ 3014 if (pktinfo->ipi6_ifindex >= if_indexlim) { 3015 return (ENXIO); 3016 } 3017 if (pktinfo->ipi6_ifindex) { 3018 ifp = ifindex2ifnet[pktinfo->ipi6_ifindex]; 3019 if (ifp == NULL) 3020 return (ENXIO); 3021 } 3022 3023 /* 3024 * We store the address anyway, and let in6_selectsrc() 3025 * validate the specified address. This is because ipi6_addr 3026 * may not have enough information about its scope zone, and 3027 * we may need additional information (such as outgoing 3028 * interface or the scope zone of a destination address) to 3029 * disambiguate the scope. 3030 * XXX: the delay of the validation may confuse the 3031 * application when it is used as a sticky option. 3032 */ 3033 if (opt->ip6po_pktinfo == NULL) { 3034 opt->ip6po_pktinfo = malloc(sizeof(*pktinfo), 3035 M_IP6OPT, M_NOWAIT); 3036 if (opt->ip6po_pktinfo == NULL) 3037 return (ENOBUFS); 3038 } 3039 memcpy(opt->ip6po_pktinfo, pktinfo, sizeof(*pktinfo)); 3040 break; 3041 } 3042 3043 #ifdef RFC2292 3044 case IPV6_2292HOPLIMIT: 3045 #endif 3046 case IPV6_HOPLIMIT: 3047 { 3048 int *hlimp; 3049 3050 /* 3051 * RFC 3542 deprecated the usage of sticky IPV6_HOPLIMIT 3052 * to simplify the ordering among hoplimit options. 3053 */ 3054 if (optname == IPV6_HOPLIMIT && sticky) 3055 return (ENOPROTOOPT); 3056 3057 if (len != sizeof(int)) 3058 return (EINVAL); 3059 hlimp = (int *)buf; 3060 if (*hlimp < -1 || *hlimp > 255) 3061 return (EINVAL); 3062 3063 opt->ip6po_hlim = *hlimp; 3064 break; 3065 } 3066 3067 case IPV6_OTCLASS: 3068 if (len != sizeof(u_int8_t)) 3069 return (EINVAL); 3070 3071 opt->ip6po_tclass = *(u_int8_t *)buf; 3072 break; 3073 3074 case IPV6_TCLASS: 3075 { 3076 int tclass; 3077 3078 if (len != sizeof(int)) 3079 return (EINVAL); 3080 tclass = *(int *)buf; 3081 if (tclass < -1 || tclass > 255) 3082 return (EINVAL); 3083 3084 opt->ip6po_tclass = tclass; 3085 break; 3086 } 3087 3088 #ifdef RFC2292 3089 case IPV6_2292NEXTHOP: 3090 #endif 3091 case IPV6_NEXTHOP: 3092 if (!priv) 3093 return (EPERM); 3094 3095 if (len == 0) { /* just remove the option */ 3096 ip6_clearpktopts(opt, IPV6_NEXTHOP); 3097 break; 3098 } 3099 3100 /* check if cmsg_len is large enough for sa_len */ 3101 if (len < sizeof(struct sockaddr) || len < *buf) 3102 return (EINVAL); 3103 3104 switch (((struct sockaddr *)buf)->sa_family) { 3105 case AF_INET6: 3106 { 3107 struct sockaddr_in6 *sa6 = (struct sockaddr_in6 *)buf; 3108 int error; 3109 3110 if (sa6->sin6_len != sizeof(struct sockaddr_in6)) 3111 return (EINVAL); 3112 3113 if (IN6_IS_ADDR_UNSPECIFIED(&sa6->sin6_addr) || 3114 IN6_IS_ADDR_MULTICAST(&sa6->sin6_addr)) { 3115 return (EINVAL); 3116 } 3117 if ((error = sa6_embedscope(sa6, ip6_use_defzone)) 3118 != 0) { 3119 return (error); 3120 } 3121 break; 3122 } 3123 case AF_LINK: /* eventually be supported? */ 3124 default: 3125 return (EAFNOSUPPORT); 3126 } 3127 3128 /* turn off the previous option, then set the new option. */ 3129 ip6_clearpktopts(opt, IPV6_NEXTHOP); 3130 opt->ip6po_nexthop = malloc(*buf, M_IP6OPT, M_NOWAIT); 3131 if (opt->ip6po_nexthop == NULL) 3132 return (ENOBUFS); 3133 memcpy(opt->ip6po_nexthop, buf, *buf); 3134 break; 3135 3136 #ifdef RFC2292 3137 case IPV6_2292HOPOPTS: 3138 #endif 3139 case IPV6_HOPOPTS: 3140 { 3141 struct ip6_hbh *hbh; 3142 int hbhlen; 3143 3144 /* 3145 * XXX: We don't allow a non-privileged user to set ANY HbH 3146 * options, since per-option restriction has too much 3147 * overhead. 3148 */ 3149 if (!priv) 3150 return (EPERM); 3151 3152 if (len == 0) { 3153 ip6_clearpktopts(opt, IPV6_HOPOPTS); 3154 break; /* just remove the option */ 3155 } 3156 3157 /* message length validation */ 3158 if (len < sizeof(struct ip6_hbh)) 3159 return (EINVAL); 3160 hbh = (struct ip6_hbh *)buf; 3161 hbhlen = (hbh->ip6h_len + 1) << 3; 3162 if (len != hbhlen) 3163 return (EINVAL); 3164 3165 /* turn off the previous option, then set the new option. */ 3166 ip6_clearpktopts(opt, IPV6_HOPOPTS); 3167 opt->ip6po_hbh = malloc(hbhlen, M_IP6OPT, M_NOWAIT); 3168 if (opt->ip6po_hbh == NULL) 3169 return (ENOBUFS); 3170 memcpy(opt->ip6po_hbh, hbh, hbhlen); 3171 3172 break; 3173 } 3174 3175 #ifdef RFC2292 3176 case IPV6_2292DSTOPTS: 3177 #endif 3178 case IPV6_DSTOPTS: 3179 case IPV6_RTHDRDSTOPTS: 3180 { 3181 struct ip6_dest *dest, **newdest = NULL; 3182 int destlen; 3183 3184 if (!priv) /* XXX: see the comment for IPV6_HOPOPTS */ 3185 return (EPERM); 3186 3187 if (len == 0) { 3188 ip6_clearpktopts(opt, optname); 3189 break; /* just remove the option */ 3190 } 3191 3192 /* message length validation */ 3193 if (len < sizeof(struct ip6_dest)) 3194 return (EINVAL); 3195 dest = (struct ip6_dest *)buf; 3196 destlen = (dest->ip6d_len + 1) << 3; 3197 if (len != destlen) 3198 return (EINVAL); 3199 /* 3200 * Determine the position that the destination options header 3201 * should be inserted; before or after the routing header. 3202 */ 3203 switch (optname) { 3204 case IPV6_2292DSTOPTS: 3205 /* 3206 * The old advanced API is ambiguous on this point. 3207 * Our approach is to determine the position based 3208 * according to the existence of a routing header. 3209 * Note, however, that this depends on the order of the 3210 * extension headers in the ancillary data; the 1st 3211 * part of the destination options header must appear 3212 * before the routing header in the ancillary data, 3213 * too. 3214 * RFC3542 solved the ambiguity by introducing 3215 * separate ancillary data or option types. 3216 */ 3217 if (opt->ip6po_rthdr == NULL) 3218 newdest = &opt->ip6po_dest1; 3219 else 3220 newdest = &opt->ip6po_dest2; 3221 break; 3222 case IPV6_RTHDRDSTOPTS: 3223 newdest = &opt->ip6po_dest1; 3224 break; 3225 case IPV6_DSTOPTS: 3226 newdest = &opt->ip6po_dest2; 3227 break; 3228 } 3229 3230 /* turn off the previous option, then set the new option. */ 3231 ip6_clearpktopts(opt, optname); 3232 *newdest = malloc(destlen, M_IP6OPT, M_NOWAIT); 3233 if (*newdest == NULL) 3234 return (ENOBUFS); 3235 memcpy(*newdest, dest, destlen); 3236 3237 break; 3238 } 3239 3240 #ifdef RFC2292 3241 case IPV6_2292RTHDR: 3242 #endif 3243 case IPV6_RTHDR: 3244 { 3245 struct ip6_rthdr *rth; 3246 int rthlen; 3247 3248 if (len == 0) { 3249 ip6_clearpktopts(opt, IPV6_RTHDR); 3250 break; /* just remove the option */ 3251 } 3252 3253 /* message length validation */ 3254 if (len < sizeof(struct ip6_rthdr)) 3255 return (EINVAL); 3256 rth = (struct ip6_rthdr *)buf; 3257 rthlen = (rth->ip6r_len + 1) << 3; 3258 if (len != rthlen) 3259 return (EINVAL); 3260 switch (rth->ip6r_type) { 3261 case IPV6_RTHDR_TYPE_0: 3262 if (rth->ip6r_len == 0) /* must contain one addr */ 3263 return (EINVAL); 3264 if (rth->ip6r_len % 2) /* length must be even */ 3265 return (EINVAL); 3266 if (rth->ip6r_len / 2 != rth->ip6r_segleft) 3267 return (EINVAL); 3268 break; 3269 default: 3270 return (EINVAL); /* not supported */ 3271 } 3272 /* turn off the previous option */ 3273 ip6_clearpktopts(opt, IPV6_RTHDR); 3274 opt->ip6po_rthdr = malloc(rthlen, M_IP6OPT, M_NOWAIT); 3275 if (opt->ip6po_rthdr == NULL) 3276 return (ENOBUFS); 3277 memcpy(opt->ip6po_rthdr, rth, rthlen); 3278 break; 3279 } 3280 3281 case IPV6_USE_MIN_MTU: 3282 if (len != sizeof(int)) 3283 return (EINVAL); 3284 minmtupolicy = *(int *)buf; 3285 if (minmtupolicy != IP6PO_MINMTU_MCASTONLY && 3286 minmtupolicy != IP6PO_MINMTU_DISABLE && 3287 minmtupolicy != IP6PO_MINMTU_ALL) { 3288 return (EINVAL); 3289 } 3290 opt->ip6po_minmtu = minmtupolicy; 3291 break; 3292 3293 case IPV6_DONTFRAG: 3294 if (len != sizeof(int)) 3295 return (EINVAL); 3296 3297 if (uproto == IPPROTO_TCP || *(int *)buf == 0) { 3298 /* 3299 * we ignore this option for TCP sockets. 3300 * (RFC3542 leaves this case unspecified.) 3301 */ 3302 opt->ip6po_flags &= ~IP6PO_DONTFRAG; 3303 } else 3304 opt->ip6po_flags |= IP6PO_DONTFRAG; 3305 break; 3306 3307 default: 3308 return (ENOPROTOOPT); 3309 } /* end of switch */ 3310 3311 return (0); 3312 } 3313 3314 /* 3315 * Routine called from ip6_output() to loop back a copy of an IP6 multicast 3316 * packet to the input queue of a specified interface. Note that this 3317 * calls the output routine of the loopback "driver", but with an interface 3318 * pointer that might NOT be lo0ifp -- easier than replicating that code here. 3319 */ 3320 void 3321 ip6_mloopback(ifp, m, dst) 3322 struct ifnet *ifp; 3323 struct mbuf *m; 3324 const struct sockaddr_in6 *dst; 3325 { 3326 struct mbuf *copym; 3327 struct ip6_hdr *ip6; 3328 3329 copym = m_copy(m, 0, M_COPYALL); 3330 if (copym == NULL) 3331 return; 3332 3333 /* 3334 * Make sure to deep-copy IPv6 header portion in case the data 3335 * is in an mbuf cluster, so that we can safely override the IPv6 3336 * header portion later. 3337 */ 3338 if ((copym->m_flags & M_EXT) != 0 || 3339 copym->m_len < sizeof(struct ip6_hdr)) { 3340 copym = m_pullup(copym, sizeof(struct ip6_hdr)); 3341 if (copym == NULL) 3342 return; 3343 } 3344 3345 #ifdef DIAGNOSTIC 3346 if (copym->m_len < sizeof(*ip6)) { 3347 m_freem(copym); 3348 return; 3349 } 3350 #endif 3351 3352 ip6 = mtod(copym, struct ip6_hdr *); 3353 /* 3354 * clear embedded scope identifiers if necessary. 3355 * in6_clearscope will touch the addresses only when necessary. 3356 */ 3357 in6_clearscope(&ip6->ip6_src); 3358 in6_clearscope(&ip6->ip6_dst); 3359 3360 (void)looutput(ifp, copym, (const struct sockaddr *)dst, NULL); 3361 } 3362 3363 /* 3364 * Chop IPv6 header off from the payload. 3365 */ 3366 static int 3367 ip6_splithdr(m, exthdrs) 3368 struct mbuf *m; 3369 struct ip6_exthdrs *exthdrs; 3370 { 3371 struct mbuf *mh; 3372 struct ip6_hdr *ip6; 3373 3374 ip6 = mtod(m, struct ip6_hdr *); 3375 if (m->m_len > sizeof(*ip6)) { 3376 MGETHDR(mh, M_DONTWAIT, MT_HEADER); 3377 if (mh == 0) { 3378 m_freem(m); 3379 return ENOBUFS; 3380 } 3381 M_MOVE_PKTHDR(mh, m); 3382 MH_ALIGN(mh, sizeof(*ip6)); 3383 m->m_len -= sizeof(*ip6); 3384 m->m_data += sizeof(*ip6); 3385 mh->m_next = m; 3386 m = mh; 3387 m->m_len = sizeof(*ip6); 3388 bcopy((void *)ip6, mtod(m, void *), sizeof(*ip6)); 3389 } 3390 exthdrs->ip6e_ip6 = m; 3391 return 0; 3392 } 3393 3394 /* 3395 * Compute IPv6 extension header length. 3396 */ 3397 int 3398 ip6_optlen(in6p) 3399 struct in6pcb *in6p; 3400 { 3401 int len; 3402 3403 if (!in6p->in6p_outputopts) 3404 return 0; 3405 3406 len = 0; 3407 #define elen(x) \ 3408 (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0) 3409 3410 len += elen(in6p->in6p_outputopts->ip6po_hbh); 3411 len += elen(in6p->in6p_outputopts->ip6po_dest1); 3412 len += elen(in6p->in6p_outputopts->ip6po_rthdr); 3413 len += elen(in6p->in6p_outputopts->ip6po_dest2); 3414 return len; 3415 #undef elen 3416 } 3417
Indexes created Tue Sep 30 20:09:53 GMT 2025