ip6_output.c revision 1.117.2.2
1 /* $NetBSD: ip6_output.c,v 1.117.2.2 2007/06/09 23:58:13 ad 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 1.117.2.2 2007/06/09 23:58:13 ad 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 if (ifp != NULL) 686 in6_ifstat_inc(ifp, ifs6_out_discard); 687 goto bad; 688 } 689 if (rt == NULL) { 690 /* 691 * If in6_selectroute() does not return a route entry, 692 * dst may not have been updated. 693 */ 694 rtcache_setdst(ro, sin6tosa(&dst_sa)); 695 } 696 697 /* 698 * then rt (for unicast) and ifp must be non-NULL valid values. 699 */ 700 if ((flags & IPV6_FORWARDING) == 0) { 701 /* XXX: the FORWARDING flag can be set for mrouting. */ 702 in6_ifstat_inc(ifp, ifs6_out_request); 703 } 704 if (rt != NULL) { 705 ia = (struct in6_ifaddr *)(rt->rt_ifa); 706 rt->rt_use++; 707 } 708 709 /* 710 * The outgoing interface must be in the zone of source and 711 * destination addresses. We should use ia_ifp to support the 712 * case of sending packets to an address of our own. 713 */ 714 if (ia != NULL && ia->ia_ifp) 715 origifp = ia->ia_ifp; 716 else 717 origifp = ifp; 718 719 src0 = ip6->ip6_src; 720 if (in6_setscope(&src0, origifp, &zone)) 721 goto badscope; 722 bzero(&src_sa, sizeof(src_sa)); 723 src_sa.sin6_family = AF_INET6; 724 src_sa.sin6_len = sizeof(src_sa); 725 src_sa.sin6_addr = ip6->ip6_src; 726 if (sa6_recoverscope(&src_sa) || zone != src_sa.sin6_scope_id) 727 goto badscope; 728 729 dst0 = ip6->ip6_dst; 730 if (in6_setscope(&dst0, origifp, &zone)) 731 goto badscope; 732 /* re-initialize to be sure */ 733 bzero(&dst_sa, sizeof(dst_sa)); 734 dst_sa.sin6_family = AF_INET6; 735 dst_sa.sin6_len = sizeof(dst_sa); 736 dst_sa.sin6_addr = ip6->ip6_dst; 737 if (sa6_recoverscope(&dst_sa) || zone != dst_sa.sin6_scope_id) 738 goto badscope; 739 740 /* scope check is done. */ 741 742 if (rt == NULL || IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) { 743 if (dst == NULL) 744 dst = satocsin6(rtcache_getdst(ro)); 745 KASSERT(dst != NULL); 746 } else if (opt && opt->ip6po_nextroute.ro_rt != NULL) { 747 /* 748 * The nexthop is explicitly specified by the 749 * application. We assume the next hop is an IPv6 750 * address. 751 */ 752 dst = (struct sockaddr_in6 *)opt->ip6po_nexthop; 753 } else if ((rt->rt_flags & RTF_GATEWAY)) 754 dst = (struct sockaddr_in6 *)rt->rt_gateway; 755 else if (dst == NULL) 756 dst = satocsin6(rtcache_getdst(ro)); 757 758 /* 759 * XXXXXX: original code follows: 760 */ 761 if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) 762 m->m_flags &= ~(M_BCAST | M_MCAST); /* just in case */ 763 else { 764 struct in6_multi *in6m; 765 766 m->m_flags = (m->m_flags & ~M_BCAST) | M_MCAST; 767 768 in6_ifstat_inc(ifp, ifs6_out_mcast); 769 770 /* 771 * Confirm that the outgoing interface supports multicast. 772 */ 773 if (!(ifp->if_flags & IFF_MULTICAST)) { 774 ip6stat.ip6s_noroute++; 775 in6_ifstat_inc(ifp, ifs6_out_discard); 776 error = ENETUNREACH; 777 goto bad; 778 } 779 780 IN6_LOOKUP_MULTI(ip6->ip6_dst, ifp, in6m); 781 if (in6m != NULL && 782 (im6o == NULL || im6o->im6o_multicast_loop)) { 783 /* 784 * If we belong to the destination multicast group 785 * on the outgoing interface, and the caller did not 786 * forbid loopback, loop back a copy. 787 */ 788 KASSERT(dst != NULL); 789 ip6_mloopback(ifp, m, dst); 790 } else { 791 /* 792 * If we are acting as a multicast router, perform 793 * multicast forwarding as if the packet had just 794 * arrived on the interface to which we are about 795 * to send. The multicast forwarding function 796 * recursively calls this function, using the 797 * IPV6_FORWARDING flag to prevent infinite recursion. 798 * 799 * Multicasts that are looped back by ip6_mloopback(), 800 * above, will be forwarded by the ip6_input() routine, 801 * if necessary. 802 */ 803 if (ip6_mrouter && (flags & IPV6_FORWARDING) == 0) { 804 if (ip6_mforward(ip6, ifp, m) != 0) { 805 m_freem(m); 806 goto done; 807 } 808 } 809 } 810 /* 811 * Multicasts with a hoplimit of zero may be looped back, 812 * above, but must not be transmitted on a network. 813 * Also, multicasts addressed to the loopback interface 814 * are not sent -- the above call to ip6_mloopback() will 815 * loop back a copy if this host actually belongs to the 816 * destination group on the loopback interface. 817 */ 818 if (ip6->ip6_hlim == 0 || (ifp->if_flags & IFF_LOOPBACK) || 819 IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst)) { 820 m_freem(m); 821 goto done; 822 } 823 } 824 825 /* 826 * Fill the outgoing inteface to tell the upper layer 827 * to increment per-interface statistics. 828 */ 829 if (ifpp) 830 *ifpp = ifp; 831 832 /* Determine path MTU. */ 833 if ((error = ip6_getpmtu(ro_pmtu, ro, ifp, &finaldst, &mtu, 834 &alwaysfrag)) != 0) 835 goto bad; 836 #ifdef IPSEC 837 if (needipsectun) 838 mtu = IPV6_MMTU; 839 #endif 840 841 /* 842 * The caller of this function may specify to use the minimum MTU 843 * in some cases. 844 * An advanced API option (IPV6_USE_MIN_MTU) can also override MTU 845 * setting. The logic is a bit complicated; by default, unicast 846 * packets will follow path MTU while multicast packets will be sent at 847 * the minimum MTU. If IP6PO_MINMTU_ALL is specified, all packets 848 * including unicast ones will be sent at the minimum MTU. Multicast 849 * packets will always be sent at the minimum MTU unless 850 * IP6PO_MINMTU_DISABLE is explicitly specified. 851 * See RFC 3542 for more details. 852 */ 853 if (mtu > IPV6_MMTU) { 854 if ((flags & IPV6_MINMTU)) 855 mtu = IPV6_MMTU; 856 else if (opt && opt->ip6po_minmtu == IP6PO_MINMTU_ALL) 857 mtu = IPV6_MMTU; 858 else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) && 859 (opt == NULL || 860 opt->ip6po_minmtu != IP6PO_MINMTU_DISABLE)) { 861 mtu = IPV6_MMTU; 862 } 863 } 864 865 /* 866 * clear embedded scope identifiers if necessary. 867 * in6_clearscope will touch the addresses only when necessary. 868 */ 869 in6_clearscope(&ip6->ip6_src); 870 in6_clearscope(&ip6->ip6_dst); 871 872 /* 873 * If the outgoing packet contains a hop-by-hop options header, 874 * it must be examined and processed even by the source node. 875 * (RFC 2460, section 4.) 876 */ 877 if (exthdrs.ip6e_hbh) { 878 struct ip6_hbh *hbh = mtod(exthdrs.ip6e_hbh, struct ip6_hbh *); 879 u_int32_t dummy1; /* XXX unused */ 880 u_int32_t dummy2; /* XXX unused */ 881 882 /* 883 * XXX: if we have to send an ICMPv6 error to the sender, 884 * we need the M_LOOP flag since icmp6_error() expects 885 * the IPv6 and the hop-by-hop options header are 886 * continuous unless the flag is set. 887 */ 888 m->m_flags |= M_LOOP; 889 m->m_pkthdr.rcvif = ifp; 890 if (ip6_process_hopopts(m, (u_int8_t *)(hbh + 1), 891 ((hbh->ip6h_len + 1) << 3) - sizeof(struct ip6_hbh), 892 &dummy1, &dummy2) < 0) { 893 /* m was already freed at this point */ 894 error = EINVAL;/* better error? */ 895 goto done; 896 } 897 m->m_flags &= ~M_LOOP; /* XXX */ 898 m->m_pkthdr.rcvif = NULL; 899 } 900 901 #ifdef PFIL_HOOKS 902 /* 903 * Run through list of hooks for output packets. 904 */ 905 if ((error = pfil_run_hooks(&inet6_pfil_hook, &m, ifp, PFIL_OUT)) != 0) 906 goto done; 907 if (m == NULL) 908 goto done; 909 ip6 = mtod(m, struct ip6_hdr *); 910 #endif /* PFIL_HOOKS */ 911 /* 912 * Send the packet to the outgoing interface. 913 * If necessary, do IPv6 fragmentation before sending. 914 * 915 * the logic here is rather complex: 916 * 1: normal case (dontfrag == 0, alwaysfrag == 0) 917 * 1-a: send as is if tlen <= path mtu 918 * 1-b: fragment if tlen > path mtu 919 * 920 * 2: if user asks us not to fragment (dontfrag == 1) 921 * 2-a: send as is if tlen <= interface mtu 922 * 2-b: error if tlen > interface mtu 923 * 924 * 3: if we always need to attach fragment header (alwaysfrag == 1) 925 * always fragment 926 * 927 * 4: if dontfrag == 1 && alwaysfrag == 1 928 * error, as we cannot handle this conflicting request 929 */ 930 tlen = m->m_pkthdr.len; 931 tso = (m->m_pkthdr.csum_flags & M_CSUM_TSOv6) != 0; 932 if (opt && (opt->ip6po_flags & IP6PO_DONTFRAG)) 933 dontfrag = 1; 934 else 935 dontfrag = 0; 936 937 if (dontfrag && alwaysfrag) { /* case 4 */ 938 /* conflicting request - can't transmit */ 939 error = EMSGSIZE; 940 goto bad; 941 } 942 if (dontfrag && (!tso && tlen > IN6_LINKMTU(ifp))) { /* case 2-b */ 943 /* 944 * Even if the DONTFRAG option is specified, we cannot send the 945 * packet when the data length is larger than the MTU of the 946 * outgoing interface. 947 * Notify the error by sending IPV6_PATHMTU ancillary data as 948 * well as returning an error code (the latter is not described 949 * in the API spec.) 950 */ 951 u_int32_t mtu32; 952 struct ip6ctlparam ip6cp; 953 954 mtu32 = (u_int32_t)mtu; 955 bzero(&ip6cp, sizeof(ip6cp)); 956 ip6cp.ip6c_cmdarg = (void *)&mtu32; 957 pfctlinput2(PRC_MSGSIZE, 958 rtcache_getdst(ro_pmtu), &ip6cp); 959 960 error = EMSGSIZE; 961 goto bad; 962 } 963 964 /* 965 * transmit packet without fragmentation 966 */ 967 if (dontfrag || (!alwaysfrag && (tlen <= mtu || tso))) { 968 /* case 1-a and 2-a */ 969 struct in6_ifaddr *ia6; 970 int sw_csum; 971 972 ip6 = mtod(m, struct ip6_hdr *); 973 ia6 = in6_ifawithifp(ifp, &ip6->ip6_src); 974 if (ia6) { 975 /* Record statistics for this interface address. */ 976 ia6->ia_ifa.ifa_data.ifad_outbytes += m->m_pkthdr.len; 977 } 978 #ifdef IPSEC 979 /* clean ipsec history once it goes out of the node */ 980 ipsec_delaux(m); 981 #endif 982 983 sw_csum = m->m_pkthdr.csum_flags & ~ifp->if_csum_flags_tx; 984 if ((sw_csum & (M_CSUM_UDPv6|M_CSUM_TCPv6)) != 0) { 985 if (IN6_NEED_CHECKSUM(ifp, 986 sw_csum & (M_CSUM_UDPv6|M_CSUM_TCPv6))) { 987 in6_delayed_cksum(m); 988 } 989 m->m_pkthdr.csum_flags &= ~(M_CSUM_UDPv6|M_CSUM_TCPv6); 990 } 991 992 KASSERT(dst != NULL); 993 if (__predict_true(!tso || 994 (ifp->if_capenable & IFCAP_TSOv6) != 0)) { 995 error = nd6_output(ifp, origifp, m, dst, rt); 996 } else { 997 error = ip6_tso_output(ifp, origifp, m, dst, rt); 998 } 999 goto done; 1000 } 1001 1002 if (tso) { 1003 error = EINVAL; /* XXX */ 1004 goto bad; 1005 } 1006 1007 /* 1008 * try to fragment the packet. case 1-b and 3 1009 */ 1010 if (mtu < IPV6_MMTU) { 1011 /* path MTU cannot be less than IPV6_MMTU */ 1012 error = EMSGSIZE; 1013 in6_ifstat_inc(ifp, ifs6_out_fragfail); 1014 goto bad; 1015 } else if (ip6->ip6_plen == 0) { 1016 /* jumbo payload cannot be fragmented */ 1017 error = EMSGSIZE; 1018 in6_ifstat_inc(ifp, ifs6_out_fragfail); 1019 goto bad; 1020 } else { 1021 struct mbuf **mnext, *m_frgpart; 1022 struct ip6_frag *ip6f; 1023 u_int32_t id = htonl(ip6_randomid()); 1024 u_char nextproto; 1025 #if 0 /* see below */ 1026 struct ip6ctlparam ip6cp; 1027 u_int32_t mtu32; 1028 #endif 1029 1030 /* 1031 * Too large for the destination or interface; 1032 * fragment if possible. 1033 * Must be able to put at least 8 bytes per fragment. 1034 */ 1035 hlen = unfragpartlen; 1036 if (mtu > IPV6_MAXPACKET) 1037 mtu = IPV6_MAXPACKET; 1038 1039 #if 0 1040 /* 1041 * It is believed this code is a leftover from the 1042 * development of the IPV6_RECVPATHMTU sockopt and 1043 * associated work to implement RFC3542. 1044 * It's not entirely clear what the intent of the API 1045 * is at this point, so disable this code for now. 1046 * The IPV6_RECVPATHMTU sockopt and/or IPV6_DONTFRAG 1047 * will send notifications if the application requests. 1048 */ 1049 1050 /* Notify a proper path MTU to applications. */ 1051 mtu32 = (u_int32_t)mtu; 1052 bzero(&ip6cp, sizeof(ip6cp)); 1053 ip6cp.ip6c_cmdarg = (void *)&mtu32; 1054 pfctlinput2(PRC_MSGSIZE, 1055 rtcache_getdst(ro_pmtu), &ip6cp); 1056 #endif 1057 1058 len = (mtu - hlen - sizeof(struct ip6_frag)) & ~7; 1059 if (len < 8) { 1060 error = EMSGSIZE; 1061 in6_ifstat_inc(ifp, ifs6_out_fragfail); 1062 goto bad; 1063 } 1064 1065 mnext = &m->m_nextpkt; 1066 1067 /* 1068 * Change the next header field of the last header in the 1069 * unfragmentable part. 1070 */ 1071 if (exthdrs.ip6e_rthdr) { 1072 nextproto = *mtod(exthdrs.ip6e_rthdr, u_char *); 1073 *mtod(exthdrs.ip6e_rthdr, u_char *) = IPPROTO_FRAGMENT; 1074 } else if (exthdrs.ip6e_dest1) { 1075 nextproto = *mtod(exthdrs.ip6e_dest1, u_char *); 1076 *mtod(exthdrs.ip6e_dest1, u_char *) = IPPROTO_FRAGMENT; 1077 } else if (exthdrs.ip6e_hbh) { 1078 nextproto = *mtod(exthdrs.ip6e_hbh, u_char *); 1079 *mtod(exthdrs.ip6e_hbh, u_char *) = IPPROTO_FRAGMENT; 1080 } else { 1081 nextproto = ip6->ip6_nxt; 1082 ip6->ip6_nxt = IPPROTO_FRAGMENT; 1083 } 1084 1085 if ((m->m_pkthdr.csum_flags & (M_CSUM_UDPv6|M_CSUM_TCPv6)) 1086 != 0) { 1087 if (IN6_NEED_CHECKSUM(ifp, 1088 m->m_pkthdr.csum_flags & 1089 (M_CSUM_UDPv6|M_CSUM_TCPv6))) { 1090 in6_delayed_cksum(m); 1091 } 1092 m->m_pkthdr.csum_flags &= ~(M_CSUM_UDPv6|M_CSUM_TCPv6); 1093 } 1094 1095 /* 1096 * Loop through length of segment after first fragment, 1097 * make new header and copy data of each part and link onto 1098 * chain. 1099 */ 1100 m0 = m; 1101 for (off = hlen; off < tlen; off += len) { 1102 struct mbuf *mlast; 1103 1104 MGETHDR(m, M_DONTWAIT, MT_HEADER); 1105 if (!m) { 1106 error = ENOBUFS; 1107 ip6stat.ip6s_odropped++; 1108 goto sendorfree; 1109 } 1110 m->m_pkthdr.rcvif = NULL; 1111 m->m_flags = m0->m_flags & M_COPYFLAGS; 1112 *mnext = m; 1113 mnext = &m->m_nextpkt; 1114 m->m_data += max_linkhdr; 1115 mhip6 = mtod(m, struct ip6_hdr *); 1116 *mhip6 = *ip6; 1117 m->m_len = sizeof(*mhip6); 1118 error = ip6_insertfraghdr(m0, m, hlen, &ip6f); 1119 if (error) { 1120 ip6stat.ip6s_odropped++; 1121 goto sendorfree; 1122 } 1123 ip6f->ip6f_offlg = htons((u_int16_t)((off - hlen) & ~7)); 1124 if (off + len >= tlen) 1125 len = tlen - off; 1126 else 1127 ip6f->ip6f_offlg |= IP6F_MORE_FRAG; 1128 mhip6->ip6_plen = htons((u_int16_t)(len + hlen + 1129 sizeof(*ip6f) - sizeof(struct ip6_hdr))); 1130 if ((m_frgpart = m_copy(m0, off, len)) == 0) { 1131 error = ENOBUFS; 1132 ip6stat.ip6s_odropped++; 1133 goto sendorfree; 1134 } 1135 for (mlast = m; mlast->m_next; mlast = mlast->m_next) 1136 ; 1137 mlast->m_next = m_frgpart; 1138 m->m_pkthdr.len = len + hlen + sizeof(*ip6f); 1139 m->m_pkthdr.rcvif = (struct ifnet *)0; 1140 ip6f->ip6f_reserved = 0; 1141 ip6f->ip6f_ident = id; 1142 ip6f->ip6f_nxt = nextproto; 1143 ip6stat.ip6s_ofragments++; 1144 in6_ifstat_inc(ifp, ifs6_out_fragcreat); 1145 } 1146 1147 in6_ifstat_inc(ifp, ifs6_out_fragok); 1148 } 1149 1150 /* 1151 * Remove leading garbages. 1152 */ 1153 sendorfree: 1154 m = m0->m_nextpkt; 1155 m0->m_nextpkt = 0; 1156 m_freem(m0); 1157 for (m0 = m; m; m = m0) { 1158 m0 = m->m_nextpkt; 1159 m->m_nextpkt = 0; 1160 if (error == 0) { 1161 struct in6_ifaddr *ia6; 1162 ip6 = mtod(m, struct ip6_hdr *); 1163 ia6 = in6_ifawithifp(ifp, &ip6->ip6_src); 1164 if (ia6) { 1165 /* 1166 * Record statistics for this interface 1167 * address. 1168 */ 1169 ia6->ia_ifa.ifa_data.ifad_outbytes += 1170 m->m_pkthdr.len; 1171 } 1172 #ifdef IPSEC 1173 /* clean ipsec history once it goes out of the node */ 1174 ipsec_delaux(m); 1175 #endif 1176 KASSERT(dst != NULL); 1177 error = nd6_output(ifp, origifp, m, dst, rt); 1178 } else 1179 m_freem(m); 1180 } 1181 1182 if (error == 0) 1183 ip6stat.ip6s_fragmented++; 1184 1185 done: 1186 /* XXX Second if is invariant? */ 1187 if (ro == &ip6route) 1188 rtcache_free(ro); 1189 else if (ro_pmtu == &ip6route) 1190 rtcache_free(ro_pmtu); 1191 1192 #ifdef IPSEC 1193 if (sp != NULL) 1194 key_freesp(sp); 1195 #endif /* IPSEC */ 1196 #ifdef FAST_IPSEC 1197 if (sp != NULL) 1198 KEY_FREESP(&sp); 1199 #endif /* FAST_IPSEC */ 1200 1201 1202 return (error); 1203 1204 freehdrs: 1205 m_freem(exthdrs.ip6e_hbh); /* m_freem will check if mbuf is 0 */ 1206 m_freem(exthdrs.ip6e_dest1); 1207 m_freem(exthdrs.ip6e_rthdr); 1208 m_freem(exthdrs.ip6e_dest2); 1209 /* FALLTHROUGH */ 1210 bad: 1211 m_freem(m); 1212 goto done; 1213 badscope: 1214 ip6stat.ip6s_badscope++; 1215 in6_ifstat_inc(origifp, ifs6_out_discard); 1216 if (error == 0) 1217 error = EHOSTUNREACH; /* XXX */ 1218 goto bad; 1219 } 1220 1221 static int 1222 ip6_copyexthdr(struct mbuf **mp, void *hdr, int hlen) 1223 { 1224 struct mbuf *m; 1225 1226 if (hlen > MCLBYTES) 1227 return (ENOBUFS); /* XXX */ 1228 1229 MGET(m, M_DONTWAIT, MT_DATA); 1230 if (!m) 1231 return (ENOBUFS); 1232 1233 if (hlen > MLEN) { 1234 MCLGET(m, M_DONTWAIT); 1235 if ((m->m_flags & M_EXT) == 0) { 1236 m_free(m); 1237 return (ENOBUFS); 1238 } 1239 } 1240 m->m_len = hlen; 1241 if (hdr) 1242 bcopy(hdr, mtod(m, void *), hlen); 1243 1244 *mp = m; 1245 return (0); 1246 } 1247 1248 /* 1249 * Process a delayed payload checksum calculation. 1250 */ 1251 void 1252 in6_delayed_cksum(struct mbuf *m) 1253 { 1254 uint16_t csum, offset; 1255 1256 KASSERT((m->m_pkthdr.csum_flags & (M_CSUM_UDPv6|M_CSUM_TCPv6)) != 0); 1257 KASSERT((~m->m_pkthdr.csum_flags & (M_CSUM_UDPv6|M_CSUM_TCPv6)) != 0); 1258 KASSERT((m->m_pkthdr.csum_flags 1259 & (M_CSUM_UDPv4|M_CSUM_TCPv4|M_CSUM_TSOv4)) == 0); 1260 1261 offset = M_CSUM_DATA_IPv6_HL(m->m_pkthdr.csum_data); 1262 csum = in6_cksum(m, 0, offset, m->m_pkthdr.len - offset); 1263 if (csum == 0 && (m->m_pkthdr.csum_flags & M_CSUM_UDPv6) != 0) { 1264 csum = 0xffff; 1265 } 1266 1267 offset += M_CSUM_DATA_IPv6_OFFSET(m->m_pkthdr.csum_data); 1268 if ((offset + sizeof(csum)) > m->m_len) { 1269 m_copyback(m, offset, sizeof(csum), &csum); 1270 } else { 1271 *(uint16_t *)(mtod(m, char *) + offset) = csum; 1272 } 1273 } 1274 1275 /* 1276 * Insert jumbo payload option. 1277 */ 1278 static int 1279 ip6_insert_jumboopt(struct ip6_exthdrs *exthdrs, u_int32_t plen) 1280 { 1281 struct mbuf *mopt; 1282 u_int8_t *optbuf; 1283 u_int32_t v; 1284 1285 #define JUMBOOPTLEN 8 /* length of jumbo payload option and padding */ 1286 1287 /* 1288 * If there is no hop-by-hop options header, allocate new one. 1289 * If there is one but it doesn't have enough space to store the 1290 * jumbo payload option, allocate a cluster to store the whole options. 1291 * Otherwise, use it to store the options. 1292 */ 1293 if (exthdrs->ip6e_hbh == 0) { 1294 MGET(mopt, M_DONTWAIT, MT_DATA); 1295 if (mopt == 0) 1296 return (ENOBUFS); 1297 mopt->m_len = JUMBOOPTLEN; 1298 optbuf = mtod(mopt, u_int8_t *); 1299 optbuf[1] = 0; /* = ((JUMBOOPTLEN) >> 3) - 1 */ 1300 exthdrs->ip6e_hbh = mopt; 1301 } else { 1302 struct ip6_hbh *hbh; 1303 1304 mopt = exthdrs->ip6e_hbh; 1305 if (M_TRAILINGSPACE(mopt) < JUMBOOPTLEN) { 1306 /* 1307 * XXX assumption: 1308 * - exthdrs->ip6e_hbh is not referenced from places 1309 * other than exthdrs. 1310 * - exthdrs->ip6e_hbh is not an mbuf chain. 1311 */ 1312 int oldoptlen = mopt->m_len; 1313 struct mbuf *n; 1314 1315 /* 1316 * XXX: give up if the whole (new) hbh header does 1317 * not fit even in an mbuf cluster. 1318 */ 1319 if (oldoptlen + JUMBOOPTLEN > MCLBYTES) 1320 return (ENOBUFS); 1321 1322 /* 1323 * As a consequence, we must always prepare a cluster 1324 * at this point. 1325 */ 1326 MGET(n, M_DONTWAIT, MT_DATA); 1327 if (n) { 1328 MCLGET(n, M_DONTWAIT); 1329 if ((n->m_flags & M_EXT) == 0) { 1330 m_freem(n); 1331 n = NULL; 1332 } 1333 } 1334 if (!n) 1335 return (ENOBUFS); 1336 n->m_len = oldoptlen + JUMBOOPTLEN; 1337 bcopy(mtod(mopt, void *), mtod(n, void *), 1338 oldoptlen); 1339 optbuf = mtod(n, u_int8_t *) + oldoptlen; 1340 m_freem(mopt); 1341 mopt = exthdrs->ip6e_hbh = n; 1342 } else { 1343 optbuf = mtod(mopt, u_int8_t *) + mopt->m_len; 1344 mopt->m_len += JUMBOOPTLEN; 1345 } 1346 optbuf[0] = IP6OPT_PADN; 1347 optbuf[1] = 0; 1348 1349 /* 1350 * Adjust the header length according to the pad and 1351 * the jumbo payload option. 1352 */ 1353 hbh = mtod(mopt, struct ip6_hbh *); 1354 hbh->ip6h_len += (JUMBOOPTLEN >> 3); 1355 } 1356 1357 /* fill in the option. */ 1358 optbuf[2] = IP6OPT_JUMBO; 1359 optbuf[3] = 4; 1360 v = (u_int32_t)htonl(plen + JUMBOOPTLEN); 1361 bcopy(&v, &optbuf[4], sizeof(u_int32_t)); 1362 1363 /* finally, adjust the packet header length */ 1364 exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN; 1365 1366 return (0); 1367 #undef JUMBOOPTLEN 1368 } 1369 1370 /* 1371 * Insert fragment header and copy unfragmentable header portions. 1372 */ 1373 static int 1374 ip6_insertfraghdr(struct mbuf *m0, struct mbuf *m, int hlen, 1375 struct ip6_frag **frghdrp) 1376 { 1377 struct mbuf *n, *mlast; 1378 1379 if (hlen > sizeof(struct ip6_hdr)) { 1380 n = m_copym(m0, sizeof(struct ip6_hdr), 1381 hlen - sizeof(struct ip6_hdr), M_DONTWAIT); 1382 if (n == 0) 1383 return (ENOBUFS); 1384 m->m_next = n; 1385 } else 1386 n = m; 1387 1388 /* Search for the last mbuf of unfragmentable part. */ 1389 for (mlast = n; mlast->m_next; mlast = mlast->m_next) 1390 ; 1391 1392 if ((mlast->m_flags & M_EXT) == 0 && 1393 M_TRAILINGSPACE(mlast) >= sizeof(struct ip6_frag)) { 1394 /* use the trailing space of the last mbuf for the fragment hdr */ 1395 *frghdrp = (struct ip6_frag *)(mtod(mlast, char *) + 1396 mlast->m_len); 1397 mlast->m_len += sizeof(struct ip6_frag); 1398 m->m_pkthdr.len += sizeof(struct ip6_frag); 1399 } else { 1400 /* allocate a new mbuf for the fragment header */ 1401 struct mbuf *mfrg; 1402 1403 MGET(mfrg, M_DONTWAIT, MT_DATA); 1404 if (mfrg == 0) 1405 return (ENOBUFS); 1406 mfrg->m_len = sizeof(struct ip6_frag); 1407 *frghdrp = mtod(mfrg, struct ip6_frag *); 1408 mlast->m_next = mfrg; 1409 } 1410 1411 return (0); 1412 } 1413 1414 static int 1415 ip6_getpmtu(struct route *ro_pmtu, struct route *ro, struct ifnet *ifp, 1416 const struct in6_addr *dst, u_long *mtup, int *alwaysfragp) 1417 { 1418 u_int32_t mtu = 0; 1419 int alwaysfrag = 0; 1420 int error = 0; 1421 1422 if (ro_pmtu != ro) { 1423 union { 1424 struct sockaddr dst; 1425 struct sockaddr_in6 dst6; 1426 } u; 1427 1428 /* The first hop and the final destination may differ. */ 1429 sockaddr_in6_init(&u.dst6, dst, 0, 0, 0); 1430 rtcache_lookup(ro_pmtu, &u.dst); 1431 } 1432 if (ro_pmtu->ro_rt != NULL) { 1433 u_int32_t ifmtu; 1434 1435 if (ifp == NULL) 1436 ifp = ro_pmtu->ro_rt->rt_ifp; 1437 ifmtu = IN6_LINKMTU(ifp); 1438 mtu = ro_pmtu->ro_rt->rt_rmx.rmx_mtu; 1439 if (mtu == 0) 1440 mtu = ifmtu; 1441 else if (mtu < IPV6_MMTU) { 1442 /* 1443 * RFC2460 section 5, last paragraph: 1444 * if we record ICMPv6 too big message with 1445 * mtu < IPV6_MMTU, transmit packets sized IPV6_MMTU 1446 * or smaller, with fragment header attached. 1447 * (fragment header is needed regardless from the 1448 * packet size, for translators to identify packets) 1449 */ 1450 alwaysfrag = 1; 1451 mtu = IPV6_MMTU; 1452 } else if (mtu > ifmtu) { 1453 /* 1454 * The MTU on the route is larger than the MTU on 1455 * the interface! This shouldn't happen, unless the 1456 * MTU of the interface has been changed after the 1457 * interface was brought up. Change the MTU in the 1458 * route to match the interface MTU (as long as the 1459 * field isn't locked). 1460 */ 1461 mtu = ifmtu; 1462 if (!(ro_pmtu->ro_rt->rt_rmx.rmx_locks & RTV_MTU)) 1463 ro_pmtu->ro_rt->rt_rmx.rmx_mtu = mtu; 1464 } 1465 } else if (ifp) { 1466 mtu = IN6_LINKMTU(ifp); 1467 } else 1468 error = EHOSTUNREACH; /* XXX */ 1469 1470 *mtup = mtu; 1471 if (alwaysfragp) 1472 *alwaysfragp = alwaysfrag; 1473 return (error); 1474 } 1475 1476 /* 1477 * IP6 socket option processing. 1478 */ 1479 int 1480 ip6_ctloutput(int op, struct socket *so, int level, int optname, 1481 struct mbuf **mp) 1482 { 1483 int privileged, optdatalen, uproto; 1484 void *optdata; 1485 struct in6pcb *in6p = sotoin6pcb(so); 1486 struct mbuf *m = *mp; 1487 int error, optval; 1488 int optlen; 1489 struct lwp *l = curlwp; /* XXX */ 1490 1491 optlen = m ? m->m_len : 0; 1492 error = optval = 0; 1493 privileged = (l == 0 || kauth_authorize_generic(l->l_cred, 1494 KAUTH_GENERIC_ISSUSER, NULL)) ? 0 : 1; 1495 uproto = (int)so->so_proto->pr_protocol; 1496 1497 if (level == IPPROTO_IPV6) { 1498 switch (op) { 1499 case PRCO_SETOPT: 1500 switch (optname) { 1501 #ifdef RFC2292 1502 case IPV6_2292PKTOPTIONS: 1503 /* m is freed in ip6_pcbopts */ 1504 error = ip6_pcbopts(&in6p->in6p_outputopts, 1505 m, so); 1506 break; 1507 #endif 1508 1509 /* 1510 * Use of some Hop-by-Hop options or some 1511 * Destination options, might require special 1512 * privilege. That is, normal applications 1513 * (without special privilege) might be forbidden 1514 * from setting certain options in outgoing packets, 1515 * and might never see certain options in received 1516 * packets. [RFC 2292 Section 6] 1517 * KAME specific note: 1518 * KAME prevents non-privileged users from sending or 1519 * receiving ANY hbh/dst options in order to avoid 1520 * overhead of parsing options in the kernel. 1521 */ 1522 case IPV6_RECVHOPOPTS: 1523 case IPV6_RECVDSTOPTS: 1524 case IPV6_RECVRTHDRDSTOPTS: 1525 if (!privileged) { 1526 error = EPERM; 1527 break; 1528 } 1529 /* FALLTHROUGH */ 1530 case IPV6_UNICAST_HOPS: 1531 case IPV6_HOPLIMIT: 1532 case IPV6_FAITH: 1533 1534 case IPV6_RECVPKTINFO: 1535 case IPV6_RECVHOPLIMIT: 1536 case IPV6_RECVRTHDR: 1537 case IPV6_RECVPATHMTU: 1538 case IPV6_RECVTCLASS: 1539 case IPV6_V6ONLY: 1540 if (optlen != sizeof(int)) { 1541 error = EINVAL; 1542 break; 1543 } 1544 optval = *mtod(m, int *); 1545 switch (optname) { 1546 1547 case IPV6_UNICAST_HOPS: 1548 if (optval < -1 || optval >= 256) 1549 error = EINVAL; 1550 else { 1551 /* -1 = kernel default */ 1552 in6p->in6p_hops = optval; 1553 } 1554 break; 1555 #define OPTSET(bit) \ 1556 do { \ 1557 if (optval) \ 1558 in6p->in6p_flags |= (bit); \ 1559 else \ 1560 in6p->in6p_flags &= ~(bit); \ 1561 } while (/*CONSTCOND*/ 0) 1562 1563 #ifdef RFC2292 1564 #define OPTSET2292(bit) \ 1565 do { \ 1566 in6p->in6p_flags |= IN6P_RFC2292; \ 1567 if (optval) \ 1568 in6p->in6p_flags |= (bit); \ 1569 else \ 1570 in6p->in6p_flags &= ~(bit); \ 1571 } while (/*CONSTCOND*/ 0) 1572 #endif 1573 1574 #define OPTBIT(bit) (in6p->in6p_flags & (bit) ? 1 : 0) 1575 1576 case IPV6_RECVPKTINFO: 1577 #ifdef RFC2292 1578 /* cannot mix with RFC2292 */ 1579 if (OPTBIT(IN6P_RFC2292)) { 1580 error = EINVAL; 1581 break; 1582 } 1583 #endif 1584 OPTSET(IN6P_PKTINFO); 1585 break; 1586 1587 case IPV6_HOPLIMIT: 1588 { 1589 struct ip6_pktopts **optp; 1590 1591 #ifdef RFC2292 1592 /* cannot mix with RFC2292 */ 1593 if (OPTBIT(IN6P_RFC2292)) { 1594 error = EINVAL; 1595 break; 1596 } 1597 #endif 1598 optp = &in6p->in6p_outputopts; 1599 error = ip6_pcbopt(IPV6_HOPLIMIT, 1600 (u_char *)&optval, 1601 sizeof(optval), 1602 optp, 1603 privileged, uproto); 1604 break; 1605 } 1606 1607 case IPV6_RECVHOPLIMIT: 1608 #ifdef RFC2292 1609 /* cannot mix with RFC2292 */ 1610 if (OPTBIT(IN6P_RFC2292)) { 1611 error = EINVAL; 1612 break; 1613 } 1614 #endif 1615 OPTSET(IN6P_HOPLIMIT); 1616 break; 1617 1618 case IPV6_RECVHOPOPTS: 1619 #ifdef RFC2292 1620 /* cannot mix with RFC2292 */ 1621 if (OPTBIT(IN6P_RFC2292)) { 1622 error = EINVAL; 1623 break; 1624 } 1625 #endif 1626 OPTSET(IN6P_HOPOPTS); 1627 break; 1628 1629 case IPV6_RECVDSTOPTS: 1630 #ifdef RFC2292 1631 /* cannot mix with RFC2292 */ 1632 if (OPTBIT(IN6P_RFC2292)) { 1633 error = EINVAL; 1634 break; 1635 } 1636 #endif 1637 OPTSET(IN6P_DSTOPTS); 1638 break; 1639 1640 case IPV6_RECVRTHDRDSTOPTS: 1641 #ifdef RFC2292 1642 /* cannot mix with RFC2292 */ 1643 if (OPTBIT(IN6P_RFC2292)) { 1644 error = EINVAL; 1645 break; 1646 } 1647 #endif 1648 OPTSET(IN6P_RTHDRDSTOPTS); 1649 break; 1650 1651 case IPV6_RECVRTHDR: 1652 #ifdef RFC2292 1653 /* cannot mix with RFC2292 */ 1654 if (OPTBIT(IN6P_RFC2292)) { 1655 error = EINVAL; 1656 break; 1657 } 1658 #endif 1659 OPTSET(IN6P_RTHDR); 1660 break; 1661 1662 case IPV6_FAITH: 1663 OPTSET(IN6P_FAITH); 1664 break; 1665 1666 case IPV6_RECVPATHMTU: 1667 /* 1668 * We ignore this option for TCP 1669 * sockets. 1670 * (RFC3542 leaves this case 1671 * unspecified.) 1672 */ 1673 if (uproto != IPPROTO_TCP) 1674 OPTSET(IN6P_MTU); 1675 break; 1676 1677 case IPV6_V6ONLY: 1678 /* 1679 * make setsockopt(IPV6_V6ONLY) 1680 * available only prior to bind(2). 1681 * see ipng mailing list, Jun 22 2001. 1682 */ 1683 if (in6p->in6p_lport || 1684 !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr)) { 1685 error = EINVAL; 1686 break; 1687 } 1688 #ifdef INET6_BINDV6ONLY 1689 if (!optval) 1690 error = EINVAL; 1691 #else 1692 OPTSET(IN6P_IPV6_V6ONLY); 1693 #endif 1694 break; 1695 case IPV6_RECVTCLASS: 1696 #ifdef RFC2292 1697 /* cannot mix with RFC2292 XXX */ 1698 if (OPTBIT(IN6P_RFC2292)) { 1699 error = EINVAL; 1700 break; 1701 } 1702 #endif 1703 OPTSET(IN6P_TCLASS); 1704 break; 1705 1706 } 1707 break; 1708 1709 case IPV6_OTCLASS: 1710 { 1711 struct ip6_pktopts **optp; 1712 u_int8_t tclass; 1713 1714 if (optlen != sizeof(tclass)) { 1715 error = EINVAL; 1716 break; 1717 } 1718 tclass = *mtod(m, u_int8_t *); 1719 optp = &in6p->in6p_outputopts; 1720 error = ip6_pcbopt(optname, 1721 (u_char *)&tclass, 1722 sizeof(tclass), 1723 optp, 1724 privileged, uproto); 1725 break; 1726 } 1727 1728 case IPV6_TCLASS: 1729 case IPV6_DONTFRAG: 1730 case IPV6_USE_MIN_MTU: 1731 if (optlen != sizeof(optval)) { 1732 error = EINVAL; 1733 break; 1734 } 1735 optval = *mtod(m, int *); 1736 { 1737 struct ip6_pktopts **optp; 1738 optp = &in6p->in6p_outputopts; 1739 error = ip6_pcbopt(optname, 1740 (u_char *)&optval, 1741 sizeof(optval), 1742 optp, 1743 privileged, uproto); 1744 break; 1745 } 1746 1747 #ifdef RFC2292 1748 case IPV6_2292PKTINFO: 1749 case IPV6_2292HOPLIMIT: 1750 case IPV6_2292HOPOPTS: 1751 case IPV6_2292DSTOPTS: 1752 case IPV6_2292RTHDR: 1753 /* RFC 2292 */ 1754 if (optlen != sizeof(int)) { 1755 error = EINVAL; 1756 break; 1757 } 1758 optval = *mtod(m, int *); 1759 switch (optname) { 1760 case IPV6_2292PKTINFO: 1761 OPTSET2292(IN6P_PKTINFO); 1762 break; 1763 case IPV6_2292HOPLIMIT: 1764 OPTSET2292(IN6P_HOPLIMIT); 1765 break; 1766 case IPV6_2292HOPOPTS: 1767 /* 1768 * Check super-user privilege. 1769 * See comments for IPV6_RECVHOPOPTS. 1770 */ 1771 if (!privileged) 1772 return (EPERM); 1773 OPTSET2292(IN6P_HOPOPTS); 1774 break; 1775 case IPV6_2292DSTOPTS: 1776 if (!privileged) 1777 return (EPERM); 1778 OPTSET2292(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); /* XXX */ 1779 break; 1780 case IPV6_2292RTHDR: 1781 OPTSET2292(IN6P_RTHDR); 1782 break; 1783 } 1784 break; 1785 #endif 1786 case IPV6_PKTINFO: 1787 case IPV6_HOPOPTS: 1788 case IPV6_RTHDR: 1789 case IPV6_DSTOPTS: 1790 case IPV6_RTHDRDSTOPTS: 1791 case IPV6_NEXTHOP: 1792 { 1793 /* new advanced API (RFC3542) */ 1794 u_char *optbuf; 1795 int optbuflen; 1796 struct ip6_pktopts **optp; 1797 if (!m) { 1798 error = EINVAL; 1799 break; 1800 } 1801 1802 #ifdef RFC2292 1803 /* cannot mix with RFC2292 */ 1804 if (OPTBIT(IN6P_RFC2292)) { 1805 error = EINVAL; 1806 break; 1807 } 1808 #endif 1809 1810 if (m && m->m_next) { 1811 error = EINVAL; /* XXX */ 1812 break; 1813 } 1814 1815 optbuf = mtod(m, u_char *); 1816 optbuflen = m->m_len; 1817 optp = &in6p->in6p_outputopts; 1818 error = ip6_pcbopt(optname, optbuf, optbuflen, 1819 optp, privileged, uproto); 1820 break; 1821 } 1822 #undef OPTSET 1823 1824 case IPV6_MULTICAST_IF: 1825 case IPV6_MULTICAST_HOPS: 1826 case IPV6_MULTICAST_LOOP: 1827 case IPV6_JOIN_GROUP: 1828 case IPV6_LEAVE_GROUP: 1829 error = ip6_setmoptions(optname, 1830 &in6p->in6p_moptions, m); 1831 break; 1832 1833 case IPV6_PORTRANGE: 1834 if (!m) { 1835 error = EINVAL; 1836 break; 1837 } 1838 optval = *mtod(m, int *); 1839 1840 switch (optval) { 1841 case IPV6_PORTRANGE_DEFAULT: 1842 in6p->in6p_flags &= ~(IN6P_LOWPORT); 1843 in6p->in6p_flags &= ~(IN6P_HIGHPORT); 1844 break; 1845 1846 case IPV6_PORTRANGE_HIGH: 1847 in6p->in6p_flags &= ~(IN6P_LOWPORT); 1848 in6p->in6p_flags |= IN6P_HIGHPORT; 1849 break; 1850 1851 case IPV6_PORTRANGE_LOW: 1852 in6p->in6p_flags &= ~(IN6P_HIGHPORT); 1853 in6p->in6p_flags |= IN6P_LOWPORT; 1854 break; 1855 1856 default: 1857 error = EINVAL; 1858 break; 1859 } 1860 break; 1861 1862 1863 #if defined(IPSEC) || defined(FAST_IPSEC) 1864 case IPV6_IPSEC_POLICY: 1865 { 1866 void *req = NULL; 1867 size_t len = 0; 1868 if (m) { 1869 req = mtod(m, void *); 1870 len = m->m_len; 1871 } 1872 error = ipsec6_set_policy(in6p, optname, req, 1873 len, privileged); 1874 } 1875 break; 1876 #endif /* IPSEC */ 1877 1878 default: 1879 error = ENOPROTOOPT; 1880 break; 1881 } 1882 if (m) 1883 (void)m_free(m); 1884 break; 1885 1886 case PRCO_GETOPT: 1887 switch (optname) { 1888 #ifdef RFC2292 1889 case IPV6_2292PKTOPTIONS: 1890 /* 1891 * RFC3542 (effectively) deprecated the 1892 * semantics of the 2292-style pktoptions. 1893 * Since it was not reliable in nature (i.e., 1894 * applications had to expect the lack of some 1895 * information after all), it would make sense 1896 * to simplify this part by always returning 1897 * empty data. 1898 */ 1899 *mp = m_get(M_WAIT, MT_SOOPTS); 1900 (*mp)->m_len = 0; 1901 break; 1902 #endif 1903 1904 case IPV6_RECVHOPOPTS: 1905 case IPV6_RECVDSTOPTS: 1906 case IPV6_RECVRTHDRDSTOPTS: 1907 case IPV6_UNICAST_HOPS: 1908 case IPV6_RECVPKTINFO: 1909 case IPV6_RECVHOPLIMIT: 1910 case IPV6_RECVRTHDR: 1911 case IPV6_RECVPATHMTU: 1912 1913 case IPV6_FAITH: 1914 case IPV6_V6ONLY: 1915 case IPV6_PORTRANGE: 1916 case IPV6_RECVTCLASS: 1917 switch (optname) { 1918 1919 case IPV6_RECVHOPOPTS: 1920 optval = OPTBIT(IN6P_HOPOPTS); 1921 break; 1922 1923 case IPV6_RECVDSTOPTS: 1924 optval = OPTBIT(IN6P_DSTOPTS); 1925 break; 1926 1927 case IPV6_RECVRTHDRDSTOPTS: 1928 optval = OPTBIT(IN6P_RTHDRDSTOPTS); 1929 break; 1930 1931 case IPV6_UNICAST_HOPS: 1932 optval = in6p->in6p_hops; 1933 break; 1934 1935 case IPV6_RECVPKTINFO: 1936 optval = OPTBIT(IN6P_PKTINFO); 1937 break; 1938 1939 case IPV6_RECVHOPLIMIT: 1940 optval = OPTBIT(IN6P_HOPLIMIT); 1941 break; 1942 1943 case IPV6_RECVRTHDR: 1944 optval = OPTBIT(IN6P_RTHDR); 1945 break; 1946 1947 case IPV6_RECVPATHMTU: 1948 optval = OPTBIT(IN6P_MTU); 1949 break; 1950 1951 case IPV6_FAITH: 1952 optval = OPTBIT(IN6P_FAITH); 1953 break; 1954 1955 case IPV6_V6ONLY: 1956 optval = OPTBIT(IN6P_IPV6_V6ONLY); 1957 break; 1958 1959 case IPV6_PORTRANGE: 1960 { 1961 int flags; 1962 flags = in6p->in6p_flags; 1963 if (flags & IN6P_HIGHPORT) 1964 optval = IPV6_PORTRANGE_HIGH; 1965 else if (flags & IN6P_LOWPORT) 1966 optval = IPV6_PORTRANGE_LOW; 1967 else 1968 optval = 0; 1969 break; 1970 } 1971 case IPV6_RECVTCLASS: 1972 optval = OPTBIT(IN6P_TCLASS); 1973 break; 1974 1975 } 1976 if (error) 1977 break; 1978 *mp = m = m_get(M_WAIT, MT_SOOPTS); 1979 m->m_len = sizeof(int); 1980 *mtod(m, int *) = optval; 1981 break; 1982 1983 case IPV6_PATHMTU: 1984 { 1985 u_long pmtu = 0; 1986 struct ip6_mtuinfo mtuinfo; 1987 struct route *ro = &in6p->in6p_route; 1988 1989 if (!(so->so_state & SS_ISCONNECTED)) 1990 return (ENOTCONN); 1991 /* 1992 * XXX: we dot not consider the case of source 1993 * routing, or optional information to specify 1994 * the outgoing interface. 1995 */ 1996 error = ip6_getpmtu(ro, NULL, NULL, 1997 &in6p->in6p_faddr, &pmtu, NULL); 1998 if (error) 1999 break; 2000 if (pmtu > IPV6_MAXPACKET) 2001 pmtu = IPV6_MAXPACKET; 2002 2003 memset(&mtuinfo, 0, sizeof(mtuinfo)); 2004 mtuinfo.ip6m_mtu = (u_int32_t)pmtu; 2005 optdata = (void *)&mtuinfo; 2006 optdatalen = sizeof(mtuinfo); 2007 if (optdatalen > MCLBYTES) 2008 return (EMSGSIZE); /* XXX */ 2009 *mp = m = m_get(M_WAIT, MT_SOOPTS); 2010 if (optdatalen > MLEN) 2011 MCLGET(m, M_WAIT); 2012 m->m_len = optdatalen; 2013 memcpy(mtod(m, void *), optdata, optdatalen); 2014 break; 2015 } 2016 2017 #ifdef RFC2292 2018 case IPV6_2292PKTINFO: 2019 case IPV6_2292HOPLIMIT: 2020 case IPV6_2292HOPOPTS: 2021 case IPV6_2292RTHDR: 2022 case IPV6_2292DSTOPTS: 2023 switch (optname) { 2024 case IPV6_2292PKTINFO: 2025 optval = OPTBIT(IN6P_PKTINFO); 2026 break; 2027 case IPV6_2292HOPLIMIT: 2028 optval = OPTBIT(IN6P_HOPLIMIT); 2029 break; 2030 case IPV6_2292HOPOPTS: 2031 optval = OPTBIT(IN6P_HOPOPTS); 2032 break; 2033 case IPV6_2292RTHDR: 2034 optval = OPTBIT(IN6P_RTHDR); 2035 break; 2036 case IPV6_2292DSTOPTS: 2037 optval = OPTBIT(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); 2038 break; 2039 } 2040 *mp = m = m_get(M_WAIT, MT_SOOPTS); 2041 m->m_len = sizeof(int); 2042 *mtod(m, int *) = optval; 2043 break; 2044 #endif 2045 case IPV6_PKTINFO: 2046 case IPV6_HOPOPTS: 2047 case IPV6_RTHDR: 2048 case IPV6_DSTOPTS: 2049 case IPV6_RTHDRDSTOPTS: 2050 case IPV6_NEXTHOP: 2051 case IPV6_OTCLASS: 2052 case IPV6_TCLASS: 2053 case IPV6_DONTFRAG: 2054 case IPV6_USE_MIN_MTU: 2055 error = ip6_getpcbopt(in6p->in6p_outputopts, 2056 optname, mp); 2057 break; 2058 2059 case IPV6_MULTICAST_IF: 2060 case IPV6_MULTICAST_HOPS: 2061 case IPV6_MULTICAST_LOOP: 2062 case IPV6_JOIN_GROUP: 2063 case IPV6_LEAVE_GROUP: 2064 error = ip6_getmoptions(optname, 2065 in6p->in6p_moptions, mp); 2066 break; 2067 2068 #if defined(IPSEC) || defined(FAST_IPSEC) 2069 case IPV6_IPSEC_POLICY: 2070 { 2071 void *req = NULL; 2072 size_t len = 0; 2073 if (m) { 2074 req = mtod(m, void *); 2075 len = m->m_len; 2076 } 2077 error = ipsec6_get_policy(in6p, req, len, mp); 2078 break; 2079 } 2080 #endif /* IPSEC */ 2081 2082 2083 2084 2085 default: 2086 error = ENOPROTOOPT; 2087 break; 2088 } 2089 break; 2090 } 2091 } else { 2092 error = EINVAL; 2093 if (op == PRCO_SETOPT && *mp) 2094 (void)m_free(*mp); 2095 } 2096 return (error); 2097 } 2098 2099 int 2100 ip6_raw_ctloutput(int op, struct socket *so, int level, int optname, 2101 struct mbuf **mp) 2102 { 2103 int error = 0, optval, optlen; 2104 const int icmp6off = offsetof(struct icmp6_hdr, icmp6_cksum); 2105 struct in6pcb *in6p = sotoin6pcb(so); 2106 struct mbuf *m = *mp; 2107 2108 optlen = m ? m->m_len : 0; 2109 2110 if (level != IPPROTO_IPV6) { 2111 if (op == PRCO_SETOPT && *mp) 2112 (void)m_free(*mp); 2113 return (EINVAL); 2114 } 2115 2116 switch (optname) { 2117 case IPV6_CHECKSUM: 2118 /* 2119 * For ICMPv6 sockets, no modification allowed for checksum 2120 * offset, permit "no change" values to help existing apps. 2121 * 2122 * XXX RFC3542 says: "An attempt to set IPV6_CHECKSUM 2123 * for an ICMPv6 socket will fail." The current 2124 * behavior does not meet RFC3542. 2125 */ 2126 switch (op) { 2127 case PRCO_SETOPT: 2128 if (optlen != sizeof(int)) { 2129 error = EINVAL; 2130 break; 2131 } 2132 optval = *mtod(m, int *); 2133 if ((optval % 2) != 0) { 2134 /* the API assumes even offset values */ 2135 error = EINVAL; 2136 } else if (so->so_proto->pr_protocol == 2137 IPPROTO_ICMPV6) { 2138 if (optval != icmp6off) 2139 error = EINVAL; 2140 } else 2141 in6p->in6p_cksum = optval; 2142 break; 2143 2144 case PRCO_GETOPT: 2145 if (so->so_proto->pr_protocol == IPPROTO_ICMPV6) 2146 optval = icmp6off; 2147 else 2148 optval = in6p->in6p_cksum; 2149 2150 *mp = m = m_get(M_WAIT, MT_SOOPTS); 2151 m->m_len = sizeof(int); 2152 *mtod(m, int *) = optval; 2153 break; 2154 2155 default: 2156 error = EINVAL; 2157 break; 2158 } 2159 break; 2160 2161 default: 2162 error = ENOPROTOOPT; 2163 break; 2164 } 2165 2166 if (op == PRCO_SETOPT && m) 2167 (void)m_free(m); 2168 2169 return (error); 2170 } 2171 2172 #ifdef RFC2292 2173 /* 2174 * Set up IP6 options in pcb for insertion in output packets or 2175 * specifying behavior of outgoing packets. 2176 */ 2177 static int 2178 ip6_pcbopts(struct ip6_pktopts **pktopt, struct mbuf *m, struct socket *so) 2179 { 2180 struct ip6_pktopts *opt = *pktopt; 2181 int error = 0; 2182 struct lwp *l = curlwp; /* XXX */ 2183 int priv = 0; 2184 2185 /* turn off any old options. */ 2186 if (opt) { 2187 #ifdef DIAGNOSTIC 2188 if (opt->ip6po_pktinfo || opt->ip6po_nexthop || 2189 opt->ip6po_hbh || opt->ip6po_dest1 || opt->ip6po_dest2 || 2190 opt->ip6po_rhinfo.ip6po_rhi_rthdr) 2191 printf("ip6_pcbopts: all specified options are cleared.\n"); 2192 #endif 2193 ip6_clearpktopts(opt, -1); 2194 } else 2195 opt = malloc(sizeof(*opt), M_IP6OPT, M_WAITOK); 2196 *pktopt = NULL; 2197 2198 if (!m || m->m_len == 0) { 2199 /* 2200 * Only turning off any previous options, regardless of 2201 * whether the opt is just created or given. 2202 */ 2203 free(opt, M_IP6OPT); 2204 return (0); 2205 } 2206 2207 /* set options specified by user. */ 2208 if (l && !kauth_authorize_generic(l->l_cred, KAUTH_GENERIC_ISSUSER, 2209 NULL)) 2210 priv = 1; 2211 if ((error = ip6_setpktopts(m, opt, NULL, priv, 2212 so->so_proto->pr_protocol)) != 0) { 2213 ip6_clearpktopts(opt, -1); /* XXX: discard all options */ 2214 free(opt, M_IP6OPT); 2215 return (error); 2216 } 2217 *pktopt = opt; 2218 return (0); 2219 } 2220 #endif 2221 2222 /* 2223 * initialize ip6_pktopts. beware that there are non-zero default values in 2224 * the struct. 2225 */ 2226 void 2227 ip6_initpktopts(struct ip6_pktopts *opt) 2228 { 2229 2230 memset(opt, 0, sizeof(*opt)); 2231 opt->ip6po_hlim = -1; /* -1 means default hop limit */ 2232 opt->ip6po_tclass = -1; /* -1 means default traffic class */ 2233 opt->ip6po_minmtu = IP6PO_MINMTU_MCASTONLY; 2234 } 2235 2236 #define sin6tosa(sin6) ((struct sockaddr *)(sin6)) /* XXX */ 2237 static int 2238 ip6_pcbopt(int optname, u_char *buf, int len, struct ip6_pktopts **pktopt, 2239 int priv, int uproto) 2240 { 2241 struct ip6_pktopts *opt; 2242 2243 if (*pktopt == NULL) { 2244 *pktopt = malloc(sizeof(struct ip6_pktopts), M_IP6OPT, 2245 M_WAITOK); 2246 ip6_initpktopts(*pktopt); 2247 } 2248 opt = *pktopt; 2249 2250 return (ip6_setpktopt(optname, buf, len, opt, priv, 1, 0, uproto)); 2251 } 2252 2253 static int 2254 ip6_getpcbopt(struct ip6_pktopts *pktopt, int optname, struct mbuf **mp) 2255 { 2256 void *optdata = NULL; 2257 int optdatalen = 0; 2258 struct ip6_ext *ip6e; 2259 int error = 0; 2260 struct in6_pktinfo null_pktinfo; 2261 int deftclass = 0, on; 2262 int defminmtu = IP6PO_MINMTU_MCASTONLY; 2263 struct mbuf *m; 2264 2265 switch (optname) { 2266 case IPV6_PKTINFO: 2267 if (pktopt && pktopt->ip6po_pktinfo) 2268 optdata = (void *)pktopt->ip6po_pktinfo; 2269 else { 2270 /* XXX: we don't have to do this every time... */ 2271 memset(&null_pktinfo, 0, sizeof(null_pktinfo)); 2272 optdata = (void *)&null_pktinfo; 2273 } 2274 optdatalen = sizeof(struct in6_pktinfo); 2275 break; 2276 case IPV6_OTCLASS: 2277 /* XXX */ 2278 return (EINVAL); 2279 case IPV6_TCLASS: 2280 if (pktopt && pktopt->ip6po_tclass >= 0) 2281 optdata = (void *)&pktopt->ip6po_tclass; 2282 else 2283 optdata = (void *)&deftclass; 2284 optdatalen = sizeof(int); 2285 break; 2286 case IPV6_HOPOPTS: 2287 if (pktopt && pktopt->ip6po_hbh) { 2288 optdata = (void *)pktopt->ip6po_hbh; 2289 ip6e = (struct ip6_ext *)pktopt->ip6po_hbh; 2290 optdatalen = (ip6e->ip6e_len + 1) << 3; 2291 } 2292 break; 2293 case IPV6_RTHDR: 2294 if (pktopt && pktopt->ip6po_rthdr) { 2295 optdata = (void *)pktopt->ip6po_rthdr; 2296 ip6e = (struct ip6_ext *)pktopt->ip6po_rthdr; 2297 optdatalen = (ip6e->ip6e_len + 1) << 3; 2298 } 2299 break; 2300 case IPV6_RTHDRDSTOPTS: 2301 if (pktopt && pktopt->ip6po_dest1) { 2302 optdata = (void *)pktopt->ip6po_dest1; 2303 ip6e = (struct ip6_ext *)pktopt->ip6po_dest1; 2304 optdatalen = (ip6e->ip6e_len + 1) << 3; 2305 } 2306 break; 2307 case IPV6_DSTOPTS: 2308 if (pktopt && pktopt->ip6po_dest2) { 2309 optdata = (void *)pktopt->ip6po_dest2; 2310 ip6e = (struct ip6_ext *)pktopt->ip6po_dest2; 2311 optdatalen = (ip6e->ip6e_len + 1) << 3; 2312 } 2313 break; 2314 case IPV6_NEXTHOP: 2315 if (pktopt && pktopt->ip6po_nexthop) { 2316 optdata = (void *)pktopt->ip6po_nexthop; 2317 optdatalen = pktopt->ip6po_nexthop->sa_len; 2318 } 2319 break; 2320 case IPV6_USE_MIN_MTU: 2321 if (pktopt) 2322 optdata = (void *)&pktopt->ip6po_minmtu; 2323 else 2324 optdata = (void *)&defminmtu; 2325 optdatalen = sizeof(int); 2326 break; 2327 case IPV6_DONTFRAG: 2328 if (pktopt && ((pktopt->ip6po_flags) & IP6PO_DONTFRAG)) 2329 on = 1; 2330 else 2331 on = 0; 2332 optdata = (void *)&on; 2333 optdatalen = sizeof(on); 2334 break; 2335 default: /* should not happen */ 2336 #ifdef DIAGNOSTIC 2337 panic("ip6_getpcbopt: unexpected option\n"); 2338 #endif 2339 return (ENOPROTOOPT); 2340 } 2341 2342 if (optdatalen > MCLBYTES) 2343 return (EMSGSIZE); /* XXX */ 2344 *mp = m = m_get(M_WAIT, MT_SOOPTS); 2345 if (optdatalen > MLEN) 2346 MCLGET(m, M_WAIT); 2347 m->m_len = optdatalen; 2348 if (optdatalen) 2349 memcpy(mtod(m, void *), optdata, optdatalen); 2350 2351 return (error); 2352 } 2353 2354 void 2355 ip6_clearpktopts(struct ip6_pktopts *pktopt, int optname) 2356 { 2357 if (optname == -1 || optname == IPV6_PKTINFO) { 2358 if (pktopt->ip6po_pktinfo) 2359 free(pktopt->ip6po_pktinfo, M_IP6OPT); 2360 pktopt->ip6po_pktinfo = NULL; 2361 } 2362 if (optname == -1 || optname == IPV6_HOPLIMIT) 2363 pktopt->ip6po_hlim = -1; 2364 if (optname == -1 || optname == IPV6_TCLASS) 2365 pktopt->ip6po_tclass = -1; 2366 if (optname == -1 || optname == IPV6_NEXTHOP) { 2367 rtcache_free(&pktopt->ip6po_nextroute); 2368 if (pktopt->ip6po_nexthop) 2369 free(pktopt->ip6po_nexthop, M_IP6OPT); 2370 pktopt->ip6po_nexthop = NULL; 2371 } 2372 if (optname == -1 || optname == IPV6_HOPOPTS) { 2373 if (pktopt->ip6po_hbh) 2374 free(pktopt->ip6po_hbh, M_IP6OPT); 2375 pktopt->ip6po_hbh = NULL; 2376 } 2377 if (optname == -1 || optname == IPV6_RTHDRDSTOPTS) { 2378 if (pktopt->ip6po_dest1) 2379 free(pktopt->ip6po_dest1, M_IP6OPT); 2380 pktopt->ip6po_dest1 = NULL; 2381 } 2382 if (optname == -1 || optname == IPV6_RTHDR) { 2383 if (pktopt->ip6po_rhinfo.ip6po_rhi_rthdr) 2384 free(pktopt->ip6po_rhinfo.ip6po_rhi_rthdr, M_IP6OPT); 2385 pktopt->ip6po_rhinfo.ip6po_rhi_rthdr = NULL; 2386 rtcache_free(&pktopt->ip6po_route); 2387 } 2388 if (optname == -1 || optname == IPV6_DSTOPTS) { 2389 if (pktopt->ip6po_dest2) 2390 free(pktopt->ip6po_dest2, M_IP6OPT); 2391 pktopt->ip6po_dest2 = NULL; 2392 } 2393 } 2394 2395 #define PKTOPT_EXTHDRCPY(type) \ 2396 do { \ 2397 if (src->type) { \ 2398 int hlen = (((struct ip6_ext *)src->type)->ip6e_len + 1) << 3;\ 2399 dst->type = malloc(hlen, M_IP6OPT, canwait); \ 2400 if (dst->type == NULL && canwait == M_NOWAIT) \ 2401 goto bad; \ 2402 memcpy(dst->type, src->type, hlen); \ 2403 } \ 2404 } while (/*CONSTCOND*/ 0) 2405 2406 static int 2407 copypktopts(struct ip6_pktopts *dst, struct ip6_pktopts *src, int canwait) 2408 { 2409 dst->ip6po_hlim = src->ip6po_hlim; 2410 dst->ip6po_tclass = src->ip6po_tclass; 2411 dst->ip6po_flags = src->ip6po_flags; 2412 if (src->ip6po_pktinfo) { 2413 dst->ip6po_pktinfo = malloc(sizeof(*dst->ip6po_pktinfo), 2414 M_IP6OPT, canwait); 2415 if (dst->ip6po_pktinfo == NULL && canwait == M_NOWAIT) 2416 goto bad; 2417 *dst->ip6po_pktinfo = *src->ip6po_pktinfo; 2418 } 2419 if (src->ip6po_nexthop) { 2420 dst->ip6po_nexthop = malloc(src->ip6po_nexthop->sa_len, 2421 M_IP6OPT, canwait); 2422 if (dst->ip6po_nexthop == NULL && canwait == M_NOWAIT) 2423 goto bad; 2424 memcpy(dst->ip6po_nexthop, src->ip6po_nexthop, 2425 src->ip6po_nexthop->sa_len); 2426 } 2427 PKTOPT_EXTHDRCPY(ip6po_hbh); 2428 PKTOPT_EXTHDRCPY(ip6po_dest1); 2429 PKTOPT_EXTHDRCPY(ip6po_dest2); 2430 PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */ 2431 return (0); 2432 2433 bad: 2434 if (dst->ip6po_pktinfo) free(dst->ip6po_pktinfo, M_IP6OPT); 2435 if (dst->ip6po_nexthop) free(dst->ip6po_nexthop, M_IP6OPT); 2436 if (dst->ip6po_hbh) free(dst->ip6po_hbh, M_IP6OPT); 2437 if (dst->ip6po_dest1) free(dst->ip6po_dest1, M_IP6OPT); 2438 if (dst->ip6po_dest2) free(dst->ip6po_dest2, M_IP6OPT); 2439 if (dst->ip6po_rthdr) free(dst->ip6po_rthdr, M_IP6OPT); 2440 2441 return (ENOBUFS); 2442 } 2443 #undef PKTOPT_EXTHDRCPY 2444 2445 struct ip6_pktopts * 2446 ip6_copypktopts(struct ip6_pktopts *src, int canwait) 2447 { 2448 int error; 2449 struct ip6_pktopts *dst; 2450 2451 dst = malloc(sizeof(*dst), M_IP6OPT, canwait); 2452 if (dst == NULL && canwait == M_NOWAIT) 2453 return (NULL); 2454 ip6_initpktopts(dst); 2455 2456 if ((error = copypktopts(dst, src, canwait)) != 0) { 2457 free(dst, M_IP6OPT); 2458 return (NULL); 2459 } 2460 2461 return (dst); 2462 } 2463 2464 void 2465 ip6_freepcbopts(struct ip6_pktopts *pktopt) 2466 { 2467 if (pktopt == NULL) 2468 return; 2469 2470 ip6_clearpktopts(pktopt, -1); 2471 2472 free(pktopt, M_IP6OPT); 2473 } 2474 2475 /* 2476 * Set the IP6 multicast options in response to user setsockopt(). 2477 */ 2478 static int 2479 ip6_setmoptions(int optname, struct ip6_moptions **im6op, struct mbuf *m) 2480 { 2481 int error = 0; 2482 u_int loop, ifindex; 2483 struct ipv6_mreq *mreq; 2484 struct ifnet *ifp; 2485 struct ip6_moptions *im6o = *im6op; 2486 struct route ro; 2487 struct in6_multi_mship *imm; 2488 struct lwp *l = curlwp; /* XXX */ 2489 2490 if (im6o == NULL) { 2491 /* 2492 * No multicast option buffer attached to the pcb; 2493 * allocate one and initialize to default values. 2494 */ 2495 im6o = (struct ip6_moptions *) 2496 malloc(sizeof(*im6o), M_IPMOPTS, M_WAITOK); 2497 2498 if (im6o == NULL) 2499 return (ENOBUFS); 2500 *im6op = im6o; 2501 im6o->im6o_multicast_ifp = NULL; 2502 im6o->im6o_multicast_hlim = ip6_defmcasthlim; 2503 im6o->im6o_multicast_loop = IPV6_DEFAULT_MULTICAST_LOOP; 2504 LIST_INIT(&im6o->im6o_memberships); 2505 } 2506 2507 switch (optname) { 2508 2509 case IPV6_MULTICAST_IF: 2510 /* 2511 * Select the interface for outgoing multicast packets. 2512 */ 2513 if (m == NULL || m->m_len != sizeof(u_int)) { 2514 error = EINVAL; 2515 break; 2516 } 2517 bcopy(mtod(m, u_int *), &ifindex, sizeof(ifindex)); 2518 if (ifindex != 0) { 2519 if (if_indexlim <= ifindex || !ifindex2ifnet[ifindex]) { 2520 error = ENXIO; /* XXX EINVAL? */ 2521 break; 2522 } 2523 ifp = ifindex2ifnet[ifindex]; 2524 if ((ifp->if_flags & IFF_MULTICAST) == 0) { 2525 error = EADDRNOTAVAIL; 2526 break; 2527 } 2528 } else 2529 ifp = NULL; 2530 im6o->im6o_multicast_ifp = ifp; 2531 break; 2532 2533 case IPV6_MULTICAST_HOPS: 2534 { 2535 /* 2536 * Set the IP6 hoplimit for outgoing multicast packets. 2537 */ 2538 int optval; 2539 if (m == NULL || m->m_len != sizeof(int)) { 2540 error = EINVAL; 2541 break; 2542 } 2543 bcopy(mtod(m, u_int *), &optval, sizeof(optval)); 2544 if (optval < -1 || optval >= 256) 2545 error = EINVAL; 2546 else if (optval == -1) 2547 im6o->im6o_multicast_hlim = ip6_defmcasthlim; 2548 else 2549 im6o->im6o_multicast_hlim = optval; 2550 break; 2551 } 2552 2553 case IPV6_MULTICAST_LOOP: 2554 /* 2555 * Set the loopback flag for outgoing multicast packets. 2556 * Must be zero or one. 2557 */ 2558 if (m == NULL || m->m_len != sizeof(u_int)) { 2559 error = EINVAL; 2560 break; 2561 } 2562 bcopy(mtod(m, u_int *), &loop, sizeof(loop)); 2563 if (loop > 1) { 2564 error = EINVAL; 2565 break; 2566 } 2567 im6o->im6o_multicast_loop = loop; 2568 break; 2569 2570 case IPV6_JOIN_GROUP: 2571 /* 2572 * Add a multicast group membership. 2573 * Group must be a valid IP6 multicast address. 2574 */ 2575 if (m == NULL || m->m_len != sizeof(struct ipv6_mreq)) { 2576 error = EINVAL; 2577 break; 2578 } 2579 mreq = mtod(m, struct ipv6_mreq *); 2580 if (IN6_IS_ADDR_UNSPECIFIED(&mreq->ipv6mr_multiaddr)) { 2581 /* 2582 * We use the unspecified address to specify to accept 2583 * all multicast addresses. Only super user is allowed 2584 * to do this. 2585 */ 2586 if (kauth_authorize_generic(l->l_cred, 2587 KAUTH_GENERIC_ISSUSER, NULL)) 2588 { 2589 error = EACCES; 2590 break; 2591 } 2592 } else if (!IN6_IS_ADDR_MULTICAST(&mreq->ipv6mr_multiaddr)) { 2593 error = EINVAL; 2594 break; 2595 } 2596 2597 /* 2598 * If no interface was explicitly specified, choose an 2599 * appropriate one according to the given multicast address. 2600 */ 2601 if (mreq->ipv6mr_interface == 0) { 2602 union { 2603 struct sockaddr dst; 2604 struct sockaddr_in6 dst6; 2605 } u; 2606 2607 /* 2608 * Look up the routing table for the 2609 * address, and choose the outgoing interface. 2610 * XXX: is it a good approach? 2611 */ 2612 memset(&ro, 0, sizeof(ro)); 2613 sockaddr_in6_init(&u.dst6, &mreq->ipv6mr_multiaddr, 0, 2614 0, 0); 2615 rtcache_setdst(&ro, &u.dst); 2616 rtcache_init(&ro); 2617 ifp = (ro.ro_rt != NULL) ? ro.ro_rt->rt_ifp : NULL; 2618 rtcache_free(&ro); 2619 } else { 2620 /* 2621 * If the interface is specified, validate it. 2622 */ 2623 if (if_indexlim <= mreq->ipv6mr_interface || 2624 !ifindex2ifnet[mreq->ipv6mr_interface]) { 2625 error = ENXIO; /* XXX EINVAL? */ 2626 break; 2627 } 2628 ifp = ifindex2ifnet[mreq->ipv6mr_interface]; 2629 } 2630 2631 /* 2632 * See if we found an interface, and confirm that it 2633 * supports multicast 2634 */ 2635 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) { 2636 error = EADDRNOTAVAIL; 2637 break; 2638 } 2639 2640 if (in6_setscope(&mreq->ipv6mr_multiaddr, ifp, NULL)) { 2641 error = EADDRNOTAVAIL; /* XXX: should not happen */ 2642 break; 2643 } 2644 2645 /* 2646 * See if the membership already exists. 2647 */ 2648 for (imm = im6o->im6o_memberships.lh_first; 2649 imm != NULL; imm = imm->i6mm_chain.le_next) 2650 if (imm->i6mm_maddr->in6m_ifp == ifp && 2651 IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr, 2652 &mreq->ipv6mr_multiaddr)) 2653 break; 2654 if (imm != NULL) { 2655 error = EADDRINUSE; 2656 break; 2657 } 2658 /* 2659 * Everything looks good; add a new record to the multicast 2660 * address list for the given interface. 2661 */ 2662 imm = in6_joingroup(ifp, &mreq->ipv6mr_multiaddr, &error, 0); 2663 if (imm == NULL) 2664 break; 2665 LIST_INSERT_HEAD(&im6o->im6o_memberships, imm, i6mm_chain); 2666 break; 2667 2668 case IPV6_LEAVE_GROUP: 2669 /* 2670 * Drop a multicast group membership. 2671 * Group must be a valid IP6 multicast address. 2672 */ 2673 if (m == NULL || m->m_len != sizeof(struct ipv6_mreq)) { 2674 error = EINVAL; 2675 break; 2676 } 2677 mreq = mtod(m, struct ipv6_mreq *); 2678 2679 /* 2680 * If an interface address was specified, get a pointer 2681 * to its ifnet structure. 2682 */ 2683 if (mreq->ipv6mr_interface != 0) { 2684 if (if_indexlim <= mreq->ipv6mr_interface || 2685 !ifindex2ifnet[mreq->ipv6mr_interface]) { 2686 error = ENXIO; /* XXX EINVAL? */ 2687 break; 2688 } 2689 ifp = ifindex2ifnet[mreq->ipv6mr_interface]; 2690 } else 2691 ifp = NULL; 2692 2693 /* Fill in the scope zone ID */ 2694 if (ifp) { 2695 if (in6_setscope(&mreq->ipv6mr_multiaddr, ifp, NULL)) { 2696 /* XXX: should not happen */ 2697 error = EADDRNOTAVAIL; 2698 break; 2699 } 2700 } else if (mreq->ipv6mr_interface != 0) { 2701 /* 2702 * XXX: This case would happens when the (positive) 2703 * index is in the valid range, but the corresponding 2704 * interface has been detached dynamically. The above 2705 * check probably avoids such case to happen here, but 2706 * we check it explicitly for safety. 2707 */ 2708 error = EADDRNOTAVAIL; 2709 break; 2710 } else { /* ipv6mr_interface == 0 */ 2711 struct sockaddr_in6 sa6_mc; 2712 2713 /* 2714 * The API spec says as follows: 2715 * If the interface index is specified as 0, the 2716 * system may choose a multicast group membership to 2717 * drop by matching the multicast address only. 2718 * On the other hand, we cannot disambiguate the scope 2719 * zone unless an interface is provided. Thus, we 2720 * check if there's ambiguity with the default scope 2721 * zone as the last resort. 2722 */ 2723 bzero(&sa6_mc, sizeof(sa6_mc)); 2724 sa6_mc.sin6_family = AF_INET6; 2725 sa6_mc.sin6_len = sizeof(sa6_mc); 2726 sa6_mc.sin6_addr = mreq->ipv6mr_multiaddr; 2727 error = sa6_embedscope(&sa6_mc, ip6_use_defzone); 2728 if (error != 0) 2729 break; 2730 mreq->ipv6mr_multiaddr = sa6_mc.sin6_addr; 2731 } 2732 2733 /* 2734 * Find the membership in the membership list. 2735 */ 2736 for (imm = im6o->im6o_memberships.lh_first; 2737 imm != NULL; imm = imm->i6mm_chain.le_next) { 2738 if ((ifp == NULL || imm->i6mm_maddr->in6m_ifp == ifp) && 2739 IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr, 2740 &mreq->ipv6mr_multiaddr)) 2741 break; 2742 } 2743 if (imm == NULL) { 2744 /* Unable to resolve interface */ 2745 error = EADDRNOTAVAIL; 2746 break; 2747 } 2748 /* 2749 * Give up the multicast address record to which the 2750 * membership points. 2751 */ 2752 LIST_REMOVE(imm, i6mm_chain); 2753 in6_leavegroup(imm); 2754 break; 2755 2756 default: 2757 error = EOPNOTSUPP; 2758 break; 2759 } 2760 2761 /* 2762 * If all options have default values, no need to keep the mbuf. 2763 */ 2764 if (im6o->im6o_multicast_ifp == NULL && 2765 im6o->im6o_multicast_hlim == ip6_defmcasthlim && 2766 im6o->im6o_multicast_loop == IPV6_DEFAULT_MULTICAST_LOOP && 2767 im6o->im6o_memberships.lh_first == NULL) { 2768 free(*im6op, M_IPMOPTS); 2769 *im6op = NULL; 2770 } 2771 2772 return (error); 2773 } 2774 2775 /* 2776 * Return the IP6 multicast options in response to user getsockopt(). 2777 */ 2778 static int 2779 ip6_getmoptions(int optname, struct ip6_moptions *im6o, struct mbuf **mp) 2780 { 2781 u_int *hlim, *loop, *ifindex; 2782 2783 *mp = m_get(M_WAIT, MT_SOOPTS); 2784 2785 switch (optname) { 2786 2787 case IPV6_MULTICAST_IF: 2788 ifindex = mtod(*mp, u_int *); 2789 (*mp)->m_len = sizeof(u_int); 2790 if (im6o == NULL || im6o->im6o_multicast_ifp == NULL) 2791 *ifindex = 0; 2792 else 2793 *ifindex = im6o->im6o_multicast_ifp->if_index; 2794 return (0); 2795 2796 case IPV6_MULTICAST_HOPS: 2797 hlim = mtod(*mp, u_int *); 2798 (*mp)->m_len = sizeof(u_int); 2799 if (im6o == NULL) 2800 *hlim = ip6_defmcasthlim; 2801 else 2802 *hlim = im6o->im6o_multicast_hlim; 2803 return (0); 2804 2805 case IPV6_MULTICAST_LOOP: 2806 loop = mtod(*mp, u_int *); 2807 (*mp)->m_len = sizeof(u_int); 2808 if (im6o == NULL) 2809 *loop = ip6_defmcasthlim; 2810 else 2811 *loop = im6o->im6o_multicast_loop; 2812 return (0); 2813 2814 default: 2815 return (EOPNOTSUPP); 2816 } 2817 } 2818 2819 /* 2820 * Discard the IP6 multicast options. 2821 */ 2822 void 2823 ip6_freemoptions(struct ip6_moptions *im6o) 2824 { 2825 struct in6_multi_mship *imm; 2826 2827 if (im6o == NULL) 2828 return; 2829 2830 while ((imm = im6o->im6o_memberships.lh_first) != NULL) { 2831 LIST_REMOVE(imm, i6mm_chain); 2832 in6_leavegroup(imm); 2833 } 2834 free(im6o, M_IPMOPTS); 2835 } 2836 2837 /* 2838 * Set IPv6 outgoing packet options based on advanced API. 2839 */ 2840 int 2841 ip6_setpktopts(struct mbuf *control, struct ip6_pktopts *opt, 2842 struct ip6_pktopts *stickyopt, int priv, int uproto) 2843 { 2844 struct cmsghdr *cm = 0; 2845 2846 if (control == NULL || opt == NULL) 2847 return (EINVAL); 2848 2849 ip6_initpktopts(opt); 2850 if (stickyopt) { 2851 int error; 2852 2853 /* 2854 * If stickyopt is provided, make a local copy of the options 2855 * for this particular packet, then override them by ancillary 2856 * objects. 2857 * XXX: copypktopts() does not copy the cached route to a next 2858 * hop (if any). This is not very good in terms of efficiency, 2859 * but we can allow this since this option should be rarely 2860 * used. 2861 */ 2862 if ((error = copypktopts(opt, stickyopt, M_NOWAIT)) != 0) 2863 return (error); 2864 } 2865 2866 /* 2867 * XXX: Currently, we assume all the optional information is stored 2868 * in a single mbuf. 2869 */ 2870 if (control->m_next) 2871 return (EINVAL); 2872 2873 for (; control->m_len; control->m_data += CMSG_ALIGN(cm->cmsg_len), 2874 control->m_len -= CMSG_ALIGN(cm->cmsg_len)) { 2875 int error; 2876 2877 if (control->m_len < CMSG_LEN(0)) 2878 return (EINVAL); 2879 2880 cm = mtod(control, struct cmsghdr *); 2881 if (cm->cmsg_len == 0 || cm->cmsg_len > control->m_len) 2882 return (EINVAL); 2883 if (cm->cmsg_level != IPPROTO_IPV6) 2884 continue; 2885 2886 error = ip6_setpktopt(cm->cmsg_type, CMSG_DATA(cm), 2887 cm->cmsg_len - CMSG_LEN(0), opt, priv, 0, 1, uproto); 2888 if (error) 2889 return (error); 2890 } 2891 2892 return (0); 2893 } 2894 2895 /* 2896 * Set a particular packet option, as a sticky option or an ancillary data 2897 * item. "len" can be 0 only when it's a sticky option. 2898 * We have 4 cases of combination of "sticky" and "cmsg": 2899 * "sticky=0, cmsg=0": impossible 2900 * "sticky=0, cmsg=1": RFC2292 or RFC3542 ancillary data 2901 * "sticky=1, cmsg=0": RFC3542 socket option 2902 * "sticky=1, cmsg=1": RFC2292 socket option 2903 */ 2904 static int 2905 ip6_setpktopt(int optname, u_char *buf, int len, struct ip6_pktopts *opt, 2906 int priv, int sticky, int cmsg, int uproto) 2907 { 2908 int minmtupolicy; 2909 2910 if (!sticky && !cmsg) { 2911 #ifdef DIAGNOSTIC 2912 printf("ip6_setpktopt: impossible case\n"); 2913 #endif 2914 return (EINVAL); 2915 } 2916 2917 /* 2918 * IPV6_2292xxx is for backward compatibility to RFC2292, and should 2919 * not be specified in the context of RFC3542. Conversely, 2920 * RFC3542 types should not be specified in the context of RFC2292. 2921 */ 2922 if (!cmsg) { 2923 switch (optname) { 2924 case IPV6_2292PKTINFO: 2925 case IPV6_2292HOPLIMIT: 2926 case IPV6_2292NEXTHOP: 2927 case IPV6_2292HOPOPTS: 2928 case IPV6_2292DSTOPTS: 2929 case IPV6_2292RTHDR: 2930 case IPV6_2292PKTOPTIONS: 2931 return (ENOPROTOOPT); 2932 } 2933 } 2934 if (sticky && cmsg) { 2935 switch (optname) { 2936 case IPV6_PKTINFO: 2937 case IPV6_HOPLIMIT: 2938 case IPV6_NEXTHOP: 2939 case IPV6_HOPOPTS: 2940 case IPV6_DSTOPTS: 2941 case IPV6_RTHDRDSTOPTS: 2942 case IPV6_RTHDR: 2943 case IPV6_USE_MIN_MTU: 2944 case IPV6_DONTFRAG: 2945 case IPV6_OTCLASS: 2946 case IPV6_TCLASS: 2947 return (ENOPROTOOPT); 2948 } 2949 } 2950 2951 switch (optname) { 2952 #ifdef RFC2292 2953 case IPV6_2292PKTINFO: 2954 #endif 2955 case IPV6_PKTINFO: 2956 { 2957 struct ifnet *ifp = NULL; 2958 struct in6_pktinfo *pktinfo; 2959 2960 if (len != sizeof(struct in6_pktinfo)) 2961 return (EINVAL); 2962 2963 pktinfo = (struct in6_pktinfo *)buf; 2964 2965 /* 2966 * An application can clear any sticky IPV6_PKTINFO option by 2967 * doing a "regular" setsockopt with ipi6_addr being 2968 * in6addr_any and ipi6_ifindex being zero. 2969 * [RFC 3542, Section 6] 2970 */ 2971 if (optname == IPV6_PKTINFO && opt->ip6po_pktinfo && 2972 pktinfo->ipi6_ifindex == 0 && 2973 IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) { 2974 ip6_clearpktopts(opt, optname); 2975 break; 2976 } 2977 2978 if (uproto == IPPROTO_TCP && optname == IPV6_PKTINFO && 2979 sticky && !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) { 2980 return (EINVAL); 2981 } 2982 2983 /* validate the interface index if specified. */ 2984 if (pktinfo->ipi6_ifindex >= if_indexlim) { 2985 return (ENXIO); 2986 } 2987 if (pktinfo->ipi6_ifindex) { 2988 ifp = ifindex2ifnet[pktinfo->ipi6_ifindex]; 2989 if (ifp == NULL) 2990 return (ENXIO); 2991 } 2992 2993 /* 2994 * We store the address anyway, and let in6_selectsrc() 2995 * validate the specified address. This is because ipi6_addr 2996 * may not have enough information about its scope zone, and 2997 * we may need additional information (such as outgoing 2998 * interface or the scope zone of a destination address) to 2999 * disambiguate the scope. 3000 * XXX: the delay of the validation may confuse the 3001 * application when it is used as a sticky option. 3002 */ 3003 if (opt->ip6po_pktinfo == NULL) { 3004 opt->ip6po_pktinfo = malloc(sizeof(*pktinfo), 3005 M_IP6OPT, M_NOWAIT); 3006 if (opt->ip6po_pktinfo == NULL) 3007 return (ENOBUFS); 3008 } 3009 memcpy(opt->ip6po_pktinfo, pktinfo, sizeof(*pktinfo)); 3010 break; 3011 } 3012 3013 #ifdef RFC2292 3014 case IPV6_2292HOPLIMIT: 3015 #endif 3016 case IPV6_HOPLIMIT: 3017 { 3018 int *hlimp; 3019 3020 /* 3021 * RFC 3542 deprecated the usage of sticky IPV6_HOPLIMIT 3022 * to simplify the ordering among hoplimit options. 3023 */ 3024 if (optname == IPV6_HOPLIMIT && sticky) 3025 return (ENOPROTOOPT); 3026 3027 if (len != sizeof(int)) 3028 return (EINVAL); 3029 hlimp = (int *)buf; 3030 if (*hlimp < -1 || *hlimp > 255) 3031 return (EINVAL); 3032 3033 opt->ip6po_hlim = *hlimp; 3034 break; 3035 } 3036 3037 case IPV6_OTCLASS: 3038 if (len != sizeof(u_int8_t)) 3039 return (EINVAL); 3040 3041 opt->ip6po_tclass = *(u_int8_t *)buf; 3042 break; 3043 3044 case IPV6_TCLASS: 3045 { 3046 int tclass; 3047 3048 if (len != sizeof(int)) 3049 return (EINVAL); 3050 tclass = *(int *)buf; 3051 if (tclass < -1 || tclass > 255) 3052 return (EINVAL); 3053 3054 opt->ip6po_tclass = tclass; 3055 break; 3056 } 3057 3058 #ifdef RFC2292 3059 case IPV6_2292NEXTHOP: 3060 #endif 3061 case IPV6_NEXTHOP: 3062 if (!priv) 3063 return (EPERM); 3064 3065 if (len == 0) { /* just remove the option */ 3066 ip6_clearpktopts(opt, IPV6_NEXTHOP); 3067 break; 3068 } 3069 3070 /* check if cmsg_len is large enough for sa_len */ 3071 if (len < sizeof(struct sockaddr) || len < *buf) 3072 return (EINVAL); 3073 3074 switch (((struct sockaddr *)buf)->sa_family) { 3075 case AF_INET6: 3076 { 3077 struct sockaddr_in6 *sa6 = (struct sockaddr_in6 *)buf; 3078 int error; 3079 3080 if (sa6->sin6_len != sizeof(struct sockaddr_in6)) 3081 return (EINVAL); 3082 3083 if (IN6_IS_ADDR_UNSPECIFIED(&sa6->sin6_addr) || 3084 IN6_IS_ADDR_MULTICAST(&sa6->sin6_addr)) { 3085 return (EINVAL); 3086 } 3087 if ((error = sa6_embedscope(sa6, ip6_use_defzone)) 3088 != 0) { 3089 return (error); 3090 } 3091 break; 3092 } 3093 case AF_LINK: /* eventually be supported? */ 3094 default: 3095 return (EAFNOSUPPORT); 3096 } 3097 3098 /* turn off the previous option, then set the new option. */ 3099 ip6_clearpktopts(opt, IPV6_NEXTHOP); 3100 opt->ip6po_nexthop = malloc(*buf, M_IP6OPT, M_NOWAIT); 3101 if (opt->ip6po_nexthop == NULL) 3102 return (ENOBUFS); 3103 memcpy(opt->ip6po_nexthop, buf, *buf); 3104 break; 3105 3106 #ifdef RFC2292 3107 case IPV6_2292HOPOPTS: 3108 #endif 3109 case IPV6_HOPOPTS: 3110 { 3111 struct ip6_hbh *hbh; 3112 int hbhlen; 3113 3114 /* 3115 * XXX: We don't allow a non-privileged user to set ANY HbH 3116 * options, since per-option restriction has too much 3117 * overhead. 3118 */ 3119 if (!priv) 3120 return (EPERM); 3121 3122 if (len == 0) { 3123 ip6_clearpktopts(opt, IPV6_HOPOPTS); 3124 break; /* just remove the option */ 3125 } 3126 3127 /* message length validation */ 3128 if (len < sizeof(struct ip6_hbh)) 3129 return (EINVAL); 3130 hbh = (struct ip6_hbh *)buf; 3131 hbhlen = (hbh->ip6h_len + 1) << 3; 3132 if (len != hbhlen) 3133 return (EINVAL); 3134 3135 /* turn off the previous option, then set the new option. */ 3136 ip6_clearpktopts(opt, IPV6_HOPOPTS); 3137 opt->ip6po_hbh = malloc(hbhlen, M_IP6OPT, M_NOWAIT); 3138 if (opt->ip6po_hbh == NULL) 3139 return (ENOBUFS); 3140 memcpy(opt->ip6po_hbh, hbh, hbhlen); 3141 3142 break; 3143 } 3144 3145 #ifdef RFC2292 3146 case IPV6_2292DSTOPTS: 3147 #endif 3148 case IPV6_DSTOPTS: 3149 case IPV6_RTHDRDSTOPTS: 3150 { 3151 struct ip6_dest *dest, **newdest = NULL; 3152 int destlen; 3153 3154 if (!priv) /* XXX: see the comment for IPV6_HOPOPTS */ 3155 return (EPERM); 3156 3157 if (len == 0) { 3158 ip6_clearpktopts(opt, optname); 3159 break; /* just remove the option */ 3160 } 3161 3162 /* message length validation */ 3163 if (len < sizeof(struct ip6_dest)) 3164 return (EINVAL); 3165 dest = (struct ip6_dest *)buf; 3166 destlen = (dest->ip6d_len + 1) << 3; 3167 if (len != destlen) 3168 return (EINVAL); 3169 /* 3170 * Determine the position that the destination options header 3171 * should be inserted; before or after the routing header. 3172 */ 3173 switch (optname) { 3174 case IPV6_2292DSTOPTS: 3175 /* 3176 * The old advanced API is ambiguous on this point. 3177 * Our approach is to determine the position based 3178 * according to the existence of a routing header. 3179 * Note, however, that this depends on the order of the 3180 * extension headers in the ancillary data; the 1st 3181 * part of the destination options header must appear 3182 * before the routing header in the ancillary data, 3183 * too. 3184 * RFC3542 solved the ambiguity by introducing 3185 * separate ancillary data or option types. 3186 */ 3187 if (opt->ip6po_rthdr == NULL) 3188 newdest = &opt->ip6po_dest1; 3189 else 3190 newdest = &opt->ip6po_dest2; 3191 break; 3192 case IPV6_RTHDRDSTOPTS: 3193 newdest = &opt->ip6po_dest1; 3194 break; 3195 case IPV6_DSTOPTS: 3196 newdest = &opt->ip6po_dest2; 3197 break; 3198 } 3199 3200 /* turn off the previous option, then set the new option. */ 3201 ip6_clearpktopts(opt, optname); 3202 *newdest = malloc(destlen, M_IP6OPT, M_NOWAIT); 3203 if (*newdest == NULL) 3204 return (ENOBUFS); 3205 memcpy(*newdest, dest, destlen); 3206 3207 break; 3208 } 3209 3210 #ifdef RFC2292 3211 case IPV6_2292RTHDR: 3212 #endif 3213 case IPV6_RTHDR: 3214 { 3215 struct ip6_rthdr *rth; 3216 int rthlen; 3217 3218 if (len == 0) { 3219 ip6_clearpktopts(opt, IPV6_RTHDR); 3220 break; /* just remove the option */ 3221 } 3222 3223 /* message length validation */ 3224 if (len < sizeof(struct ip6_rthdr)) 3225 return (EINVAL); 3226 rth = (struct ip6_rthdr *)buf; 3227 rthlen = (rth->ip6r_len + 1) << 3; 3228 if (len != rthlen) 3229 return (EINVAL); 3230 switch (rth->ip6r_type) { 3231 case IPV6_RTHDR_TYPE_0: 3232 if (rth->ip6r_len == 0) /* must contain one addr */ 3233 return (EINVAL); 3234 if (rth->ip6r_len % 2) /* length must be even */ 3235 return (EINVAL); 3236 if (rth->ip6r_len / 2 != rth->ip6r_segleft) 3237 return (EINVAL); 3238 break; 3239 default: 3240 return (EINVAL); /* not supported */ 3241 } 3242 /* turn off the previous option */ 3243 ip6_clearpktopts(opt, IPV6_RTHDR); 3244 opt->ip6po_rthdr = malloc(rthlen, M_IP6OPT, M_NOWAIT); 3245 if (opt->ip6po_rthdr == NULL) 3246 return (ENOBUFS); 3247 memcpy(opt->ip6po_rthdr, rth, rthlen); 3248 break; 3249 } 3250 3251 case IPV6_USE_MIN_MTU: 3252 if (len != sizeof(int)) 3253 return (EINVAL); 3254 minmtupolicy = *(int *)buf; 3255 if (minmtupolicy != IP6PO_MINMTU_MCASTONLY && 3256 minmtupolicy != IP6PO_MINMTU_DISABLE && 3257 minmtupolicy != IP6PO_MINMTU_ALL) { 3258 return (EINVAL); 3259 } 3260 opt->ip6po_minmtu = minmtupolicy; 3261 break; 3262 3263 case IPV6_DONTFRAG: 3264 if (len != sizeof(int)) 3265 return (EINVAL); 3266 3267 if (uproto == IPPROTO_TCP || *(int *)buf == 0) { 3268 /* 3269 * we ignore this option for TCP sockets. 3270 * (RFC3542 leaves this case unspecified.) 3271 */ 3272 opt->ip6po_flags &= ~IP6PO_DONTFRAG; 3273 } else 3274 opt->ip6po_flags |= IP6PO_DONTFRAG; 3275 break; 3276 3277 default: 3278 return (ENOPROTOOPT); 3279 } /* end of switch */ 3280 3281 return (0); 3282 } 3283 3284 /* 3285 * Routine called from ip6_output() to loop back a copy of an IP6 multicast 3286 * packet to the input queue of a specified interface. Note that this 3287 * calls the output routine of the loopback "driver", but with an interface 3288 * pointer that might NOT be lo0ifp -- easier than replicating that code here. 3289 */ 3290 void 3291 ip6_mloopback(struct ifnet *ifp, struct mbuf *m, 3292 const struct sockaddr_in6 *dst) 3293 { 3294 struct mbuf *copym; 3295 struct ip6_hdr *ip6; 3296 3297 copym = m_copy(m, 0, M_COPYALL); 3298 if (copym == NULL) 3299 return; 3300 3301 /* 3302 * Make sure to deep-copy IPv6 header portion in case the data 3303 * is in an mbuf cluster, so that we can safely override the IPv6 3304 * header portion later. 3305 */ 3306 if ((copym->m_flags & M_EXT) != 0 || 3307 copym->m_len < sizeof(struct ip6_hdr)) { 3308 copym = m_pullup(copym, sizeof(struct ip6_hdr)); 3309 if (copym == NULL) 3310 return; 3311 } 3312 3313 #ifdef DIAGNOSTIC 3314 if (copym->m_len < sizeof(*ip6)) { 3315 m_freem(copym); 3316 return; 3317 } 3318 #endif 3319 3320 ip6 = mtod(copym, struct ip6_hdr *); 3321 /* 3322 * clear embedded scope identifiers if necessary. 3323 * in6_clearscope will touch the addresses only when necessary. 3324 */ 3325 in6_clearscope(&ip6->ip6_src); 3326 in6_clearscope(&ip6->ip6_dst); 3327 3328 (void)looutput(ifp, copym, (const struct sockaddr *)dst, NULL); 3329 } 3330 3331 /* 3332 * Chop IPv6 header off from the payload. 3333 */ 3334 static int 3335 ip6_splithdr(struct mbuf *m, struct ip6_exthdrs *exthdrs) 3336 { 3337 struct mbuf *mh; 3338 struct ip6_hdr *ip6; 3339 3340 ip6 = mtod(m, struct ip6_hdr *); 3341 if (m->m_len > sizeof(*ip6)) { 3342 MGETHDR(mh, M_DONTWAIT, MT_HEADER); 3343 if (mh == 0) { 3344 m_freem(m); 3345 return ENOBUFS; 3346 } 3347 M_MOVE_PKTHDR(mh, m); 3348 MH_ALIGN(mh, sizeof(*ip6)); 3349 m->m_len -= sizeof(*ip6); 3350 m->m_data += sizeof(*ip6); 3351 mh->m_next = m; 3352 m = mh; 3353 m->m_len = sizeof(*ip6); 3354 bcopy((void *)ip6, mtod(m, void *), sizeof(*ip6)); 3355 } 3356 exthdrs->ip6e_ip6 = m; 3357 return 0; 3358 } 3359 3360 /* 3361 * Compute IPv6 extension header length. 3362 */ 3363 int 3364 ip6_optlen(struct in6pcb *in6p) 3365 { 3366 int len; 3367 3368 if (!in6p->in6p_outputopts) 3369 return 0; 3370 3371 len = 0; 3372 #define elen(x) \ 3373 (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0) 3374 3375 len += elen(in6p->in6p_outputopts->ip6po_hbh); 3376 len += elen(in6p->in6p_outputopts->ip6po_dest1); 3377 len += elen(in6p->in6p_outputopts->ip6po_rthdr); 3378 len += elen(in6p->in6p_outputopts->ip6po_dest2); 3379 return len; 3380 #undef elen 3381 } 3382
Indexes created Wed Oct 01 07:09:59 GMT 2025