ip6_output.c revision 1.109
1 /* $NetBSD: ip6_output.c,v 1.109 2006/12/15 21:18:55 joerg 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 1.109 2006/12/15 21:18:55 joerg 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 #include <net/net_osdep.h> 107 108 #ifdef PFIL_HOOKS 109 extern struct pfil_head inet6_pfil_hook; /* XXX */ 110 #endif 111 112 struct ip6_exthdrs { 113 struct mbuf *ip6e_ip6; 114 struct mbuf *ip6e_hbh; 115 struct mbuf *ip6e_dest1; 116 struct mbuf *ip6e_rthdr; 117 struct mbuf *ip6e_dest2; 118 }; 119 120 static int ip6_pcbopt __P((int, u_char *, int, struct ip6_pktopts **, 121 int, int)); 122 static int ip6_getpcbopt __P((struct ip6_pktopts *, int, struct mbuf **)); 123 static int ip6_setpktopt __P((int, u_char *, int, struct ip6_pktopts *, int, 124 int, int, int)); 125 static int ip6_setmoptions __P((int, struct ip6_moptions **, struct mbuf *)); 126 static int ip6_getmoptions __P((int, struct ip6_moptions *, struct mbuf **)); 127 static int ip6_copyexthdr __P((struct mbuf **, caddr_t, int)); 128 static int ip6_insertfraghdr __P((struct mbuf *, struct mbuf *, int, 129 struct ip6_frag **)); 130 static int ip6_insert_jumboopt __P((struct ip6_exthdrs *, u_int32_t)); 131 static int ip6_splithdr __P((struct mbuf *, struct ip6_exthdrs *)); 132 static int ip6_getpmtu __P((struct route_in6 *, struct route_in6 *, 133 struct ifnet *, struct in6_addr *, u_long *, int *)); 134 static int copypktopts __P((struct ip6_pktopts *, struct ip6_pktopts *, int)); 135 136 #ifdef RFC2292 137 static int ip6_pcbopts __P((struct ip6_pktopts **, struct mbuf *, 138 struct socket *)); 139 #endif 140 141 #define IN6_NEED_CHECKSUM(ifp, csum_flags) \ 142 (__predict_true(((ifp)->if_flags & IFF_LOOPBACK) == 0 || \ 143 (((csum_flags) & M_CSUM_UDPv6) != 0 && udp_do_loopback_cksum) || \ 144 (((csum_flags) & M_CSUM_TCPv6) != 0 && tcp_do_loopback_cksum))) 145 146 /* 147 * IP6 output. The packet in mbuf chain m contains a skeletal IP6 148 * header (with pri, len, nxt, hlim, src, dst). 149 * This function may modify ver and hlim only. 150 * The mbuf chain containing the packet will be freed. 151 * The mbuf opt, if present, will not be freed. 152 * 153 * type of "mtu": rt_rmx.rmx_mtu is u_long, ifnet.ifr_mtu is int, and 154 * nd_ifinfo.linkmtu is u_int32_t. so we use u_long to hold largest one, 155 * which is rt_rmx.rmx_mtu. 156 */ 157 int 158 ip6_output( 159 struct mbuf *m0, 160 struct ip6_pktopts *opt, 161 struct route_in6 *ro, 162 int flags, 163 struct ip6_moptions *im6o, 164 struct socket *so, 165 struct ifnet **ifpp /* XXX: just for statistics */ 166 ) 167 { 168 struct ip6_hdr *ip6, *mhip6; 169 struct ifnet *ifp, *origifp; 170 struct mbuf *m = m0; 171 int hlen, tlen, len, off; 172 boolean_t tso; 173 struct route_in6 ip6route; 174 struct rtentry *rt = NULL; 175 struct sockaddr_in6 *dst, src_sa, dst_sa; 176 int error = 0; 177 struct in6_ifaddr *ia = NULL; 178 u_long mtu; 179 int alwaysfrag, dontfrag; 180 u_int32_t optlen = 0, plen = 0, unfragpartlen = 0; 181 struct ip6_exthdrs exthdrs; 182 struct in6_addr finaldst, src0, dst0; 183 u_int32_t zone; 184 struct route_in6 *ro_pmtu = NULL; 185 int hdrsplit = 0; 186 int needipsec = 0; 187 #ifdef IPSEC 188 int needipsectun = 0; 189 struct secpolicy *sp = NULL; 190 191 ip6 = mtod(m, struct ip6_hdr *); 192 #endif /* IPSEC */ 193 194 #ifdef DIAGNOSTIC 195 if ((m->m_flags & M_PKTHDR) == 0) 196 panic("ip6_output: no HDR"); 197 198 if ((m->m_pkthdr.csum_flags & 199 (M_CSUM_TCPv4|M_CSUM_UDPv4|M_CSUM_TSOv4)) != 0) { 200 panic("ip6_output: IPv4 checksum offload flags: %d", 201 m->m_pkthdr.csum_flags); 202 } 203 204 if ((m->m_pkthdr.csum_flags & (M_CSUM_TCPv6|M_CSUM_UDPv6)) == 205 (M_CSUM_TCPv6|M_CSUM_UDPv6)) { 206 panic("ip6_output: conflicting checksum offload flags: %d", 207 m->m_pkthdr.csum_flags); 208 } 209 #endif 210 211 M_CSUM_DATA_IPv6_HL_SET(m->m_pkthdr.csum_data, sizeof(struct ip6_hdr)); 212 213 #define MAKE_EXTHDR(hp, mp) \ 214 do { \ 215 if (hp) { \ 216 struct ip6_ext *eh = (struct ip6_ext *)(hp); \ 217 error = ip6_copyexthdr((mp), (caddr_t)(hp), \ 218 ((eh)->ip6e_len + 1) << 3); \ 219 if (error) \ 220 goto freehdrs; \ 221 } \ 222 } while (/*CONSTCOND*/ 0) 223 224 bzero(&exthdrs, sizeof(exthdrs)); 225 if (opt) { 226 /* Hop-by-Hop options header */ 227 MAKE_EXTHDR(opt->ip6po_hbh, &exthdrs.ip6e_hbh); 228 /* Destination options header(1st part) */ 229 MAKE_EXTHDR(opt->ip6po_dest1, &exthdrs.ip6e_dest1); 230 /* Routing header */ 231 MAKE_EXTHDR(opt->ip6po_rthdr, &exthdrs.ip6e_rthdr); 232 /* Destination options header(2nd part) */ 233 MAKE_EXTHDR(opt->ip6po_dest2, &exthdrs.ip6e_dest2); 234 } 235 236 #ifdef IPSEC 237 if ((flags & IPV6_FORWARDING) != 0) { 238 needipsec = 0; 239 goto skippolicycheck; 240 } 241 242 /* get a security policy for this packet */ 243 if (so == NULL) 244 sp = ipsec6_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND, 0, &error); 245 else { 246 if (IPSEC_PCB_SKIP_IPSEC(sotoinpcb_hdr(so)->inph_sp, 247 IPSEC_DIR_OUTBOUND)) { 248 needipsec = 0; 249 goto skippolicycheck; 250 } 251 sp = ipsec6_getpolicybysock(m, IPSEC_DIR_OUTBOUND, so, &error); 252 } 253 254 if (sp == NULL) { 255 ipsec6stat.out_inval++; 256 goto freehdrs; 257 } 258 259 error = 0; 260 261 /* check policy */ 262 switch (sp->policy) { 263 case IPSEC_POLICY_DISCARD: 264 /* 265 * This packet is just discarded. 266 */ 267 ipsec6stat.out_polvio++; 268 goto freehdrs; 269 270 case IPSEC_POLICY_BYPASS: 271 case IPSEC_POLICY_NONE: 272 /* no need to do IPsec. */ 273 needipsec = 0; 274 break; 275 276 case IPSEC_POLICY_IPSEC: 277 if (sp->req == NULL) { 278 /* XXX should be panic ? */ 279 printf("ip6_output: No IPsec request specified.\n"); 280 error = EINVAL; 281 goto freehdrs; 282 } 283 needipsec = 1; 284 break; 285 286 case IPSEC_POLICY_ENTRUST: 287 default: 288 printf("ip6_output: Invalid policy found. %d\n", sp->policy); 289 } 290 291 skippolicycheck:; 292 #endif /* IPSEC */ 293 294 if (needipsec && 295 (m->m_pkthdr.csum_flags & (M_CSUM_UDPv6|M_CSUM_TCPv6)) != 0) { 296 in6_delayed_cksum(m); 297 m->m_pkthdr.csum_flags &= ~(M_CSUM_UDPv6|M_CSUM_TCPv6); 298 } 299 300 /* 301 * Calculate the total length of the extension header chain. 302 * Keep the length of the unfragmentable part for fragmentation. 303 */ 304 optlen = 0; 305 if (exthdrs.ip6e_hbh) optlen += exthdrs.ip6e_hbh->m_len; 306 if (exthdrs.ip6e_dest1) optlen += exthdrs.ip6e_dest1->m_len; 307 if (exthdrs.ip6e_rthdr) optlen += exthdrs.ip6e_rthdr->m_len; 308 unfragpartlen = optlen + sizeof(struct ip6_hdr); 309 /* NOTE: we don't add AH/ESP length here. do that later. */ 310 if (exthdrs.ip6e_dest2) optlen += exthdrs.ip6e_dest2->m_len; 311 312 /* 313 * If we need IPsec, or there is at least one extension header, 314 * separate IP6 header from the payload. 315 */ 316 if ((needipsec || optlen) && !hdrsplit) { 317 if ((error = ip6_splithdr(m, &exthdrs)) != 0) { 318 m = NULL; 319 goto freehdrs; 320 } 321 m = exthdrs.ip6e_ip6; 322 hdrsplit++; 323 } 324 325 /* adjust pointer */ 326 ip6 = mtod(m, struct ip6_hdr *); 327 328 /* adjust mbuf packet header length */ 329 m->m_pkthdr.len += optlen; 330 plen = m->m_pkthdr.len - sizeof(*ip6); 331 332 /* If this is a jumbo payload, insert a jumbo payload option. */ 333 if (plen > IPV6_MAXPACKET) { 334 if (!hdrsplit) { 335 if ((error = ip6_splithdr(m, &exthdrs)) != 0) { 336 m = NULL; 337 goto freehdrs; 338 } 339 m = exthdrs.ip6e_ip6; 340 hdrsplit++; 341 } 342 /* adjust pointer */ 343 ip6 = mtod(m, struct ip6_hdr *); 344 if ((error = ip6_insert_jumboopt(&exthdrs, plen)) != 0) 345 goto freehdrs; 346 optlen += 8; /* XXX JUMBOOPTLEN */ 347 ip6->ip6_plen = 0; 348 } else 349 ip6->ip6_plen = htons(plen); 350 351 /* 352 * Concatenate headers and fill in next header fields. 353 * Here we have, on "m" 354 * IPv6 payload 355 * and we insert headers accordingly. Finally, we should be getting: 356 * IPv6 hbh dest1 rthdr ah* [esp* dest2 payload] 357 * 358 * during the header composing process, "m" points to IPv6 header. 359 * "mprev" points to an extension header prior to esp. 360 */ 361 { 362 u_char *nexthdrp = &ip6->ip6_nxt; 363 struct mbuf *mprev = m; 364 365 /* 366 * we treat dest2 specially. this makes IPsec processing 367 * much easier. the goal here is to make mprev point the 368 * mbuf prior to dest2. 369 * 370 * result: IPv6 dest2 payload 371 * m and mprev will point to IPv6 header. 372 */ 373 if (exthdrs.ip6e_dest2) { 374 if (!hdrsplit) 375 panic("assumption failed: hdr not split"); 376 exthdrs.ip6e_dest2->m_next = m->m_next; 377 m->m_next = exthdrs.ip6e_dest2; 378 *mtod(exthdrs.ip6e_dest2, u_char *) = ip6->ip6_nxt; 379 ip6->ip6_nxt = IPPROTO_DSTOPTS; 380 } 381 382 #define MAKE_CHAIN(m, mp, p, i)\ 383 do {\ 384 if (m) {\ 385 if (!hdrsplit) \ 386 panic("assumption failed: hdr not split"); \ 387 *mtod((m), u_char *) = *(p);\ 388 *(p) = (i);\ 389 p = mtod((m), u_char *);\ 390 (m)->m_next = (mp)->m_next;\ 391 (mp)->m_next = (m);\ 392 (mp) = (m);\ 393 }\ 394 } while (/*CONSTCOND*/ 0) 395 /* 396 * result: IPv6 hbh dest1 rthdr dest2 payload 397 * m will point to IPv6 header. mprev will point to the 398 * extension header prior to dest2 (rthdr in the above case). 399 */ 400 MAKE_CHAIN(exthdrs.ip6e_hbh, mprev, nexthdrp, IPPROTO_HOPOPTS); 401 MAKE_CHAIN(exthdrs.ip6e_dest1, mprev, nexthdrp, 402 IPPROTO_DSTOPTS); 403 MAKE_CHAIN(exthdrs.ip6e_rthdr, mprev, nexthdrp, 404 IPPROTO_ROUTING); 405 406 M_CSUM_DATA_IPv6_HL_SET(m->m_pkthdr.csum_data, 407 sizeof(struct ip6_hdr) + optlen); 408 409 #ifdef IPSEC 410 if (!needipsec) 411 goto skip_ipsec2; 412 413 /* 414 * pointers after IPsec headers are not valid any more. 415 * other pointers need a great care too. 416 * (IPsec routines should not mangle mbufs prior to AH/ESP) 417 */ 418 exthdrs.ip6e_dest2 = NULL; 419 420 { 421 struct ip6_rthdr *rh = NULL; 422 int segleft_org = 0; 423 struct ipsec_output_state state; 424 425 if (exthdrs.ip6e_rthdr) { 426 rh = mtod(exthdrs.ip6e_rthdr, struct ip6_rthdr *); 427 segleft_org = rh->ip6r_segleft; 428 rh->ip6r_segleft = 0; 429 } 430 431 bzero(&state, sizeof(state)); 432 state.m = m; 433 error = ipsec6_output_trans(&state, nexthdrp, mprev, sp, flags, 434 &needipsectun); 435 m = state.m; 436 if (error) { 437 /* mbuf is already reclaimed in ipsec6_output_trans. */ 438 m = NULL; 439 switch (error) { 440 case EHOSTUNREACH: 441 case ENETUNREACH: 442 case EMSGSIZE: 443 case ENOBUFS: 444 case ENOMEM: 445 break; 446 default: 447 printf("ip6_output (ipsec): error code %d\n", error); 448 /* FALLTHROUGH */ 449 case ENOENT: 450 /* don't show these error codes to the user */ 451 error = 0; 452 break; 453 } 454 goto bad; 455 } 456 if (exthdrs.ip6e_rthdr) { 457 /* ah6_output doesn't modify mbuf chain */ 458 rh->ip6r_segleft = segleft_org; 459 } 460 } 461 skip_ipsec2:; 462 #endif 463 } 464 465 /* 466 * If there is a routing header, replace destination address field 467 * with the first hop of the routing header. 468 */ 469 if (exthdrs.ip6e_rthdr) { 470 struct ip6_rthdr *rh; 471 struct ip6_rthdr0 *rh0; 472 struct in6_addr *addr; 473 struct sockaddr_in6 sa; 474 475 rh = (struct ip6_rthdr *)(mtod(exthdrs.ip6e_rthdr, 476 struct ip6_rthdr *)); 477 finaldst = ip6->ip6_dst; 478 switch (rh->ip6r_type) { 479 case IPV6_RTHDR_TYPE_0: 480 rh0 = (struct ip6_rthdr0 *)rh; 481 addr = (struct in6_addr *)(rh0 + 1); 482 483 /* 484 * construct a sockaddr_in6 form of 485 * the first hop. 486 * 487 * XXX: we may not have enough 488 * information about its scope zone; 489 * there is no standard API to pass 490 * the information from the 491 * application. 492 */ 493 bzero(&sa, sizeof(sa)); 494 sa.sin6_family = AF_INET6; 495 sa.sin6_len = sizeof(sa); 496 sa.sin6_addr = addr[0]; 497 if ((error = sa6_embedscope(&sa, 498 ip6_use_defzone)) != 0) { 499 goto bad; 500 } 501 ip6->ip6_dst = sa.sin6_addr; 502 (void)memmove(&addr[0], &addr[1], 503 sizeof(struct in6_addr) * 504 (rh0->ip6r0_segleft - 1)); 505 addr[rh0->ip6r0_segleft - 1] = finaldst; 506 /* XXX */ 507 in6_clearscope(addr + rh0->ip6r0_segleft - 1); 508 break; 509 default: /* is it possible? */ 510 error = EINVAL; 511 goto bad; 512 } 513 } 514 515 /* Source address validation */ 516 if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) && 517 (flags & IPV6_UNSPECSRC) == 0) { 518 error = EOPNOTSUPP; 519 ip6stat.ip6s_badscope++; 520 goto bad; 521 } 522 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) { 523 error = EOPNOTSUPP; 524 ip6stat.ip6s_badscope++; 525 goto bad; 526 } 527 528 ip6stat.ip6s_localout++; 529 530 /* 531 * Route packet. 532 */ 533 /* initialize cached route */ 534 if (ro == 0) { 535 ro = &ip6route; 536 bzero((caddr_t)ro, sizeof(*ro)); 537 } 538 ro_pmtu = ro; 539 if (opt && opt->ip6po_rthdr) 540 ro = &opt->ip6po_route; 541 dst = (struct sockaddr_in6 *)&ro->ro_dst; 542 543 /* 544 * if specified, try to fill in the traffic class field. 545 * do not override if a non-zero value is already set. 546 * we check the diffserv field and the ecn field separately. 547 */ 548 if (opt && opt->ip6po_tclass >= 0) { 549 int mask = 0; 550 551 if ((ip6->ip6_flow & htonl(0xfc << 20)) == 0) 552 mask |= 0xfc; 553 if ((ip6->ip6_flow & htonl(0x03 << 20)) == 0) 554 mask |= 0x03; 555 if (mask != 0) 556 ip6->ip6_flow |= htonl((opt->ip6po_tclass & mask) << 20); 557 } 558 559 /* fill in or override the hop limit field, if necessary. */ 560 if (opt && opt->ip6po_hlim != -1) 561 ip6->ip6_hlim = opt->ip6po_hlim & 0xff; 562 else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) { 563 if (im6o != NULL) 564 ip6->ip6_hlim = im6o->im6o_multicast_hlim; 565 else 566 ip6->ip6_hlim = ip6_defmcasthlim; 567 } 568 569 #ifdef IPSEC 570 if (needipsec && needipsectun) { 571 struct ipsec_output_state state; 572 573 /* 574 * All the extension headers will become inaccessible 575 * (since they can be encrypted). 576 * Don't panic, we need no more updates to extension headers 577 * on inner IPv6 packet (since they are now encapsulated). 578 * 579 * IPv6 [ESP|AH] IPv6 [extension headers] payload 580 */ 581 bzero(&exthdrs, sizeof(exthdrs)); 582 exthdrs.ip6e_ip6 = m; 583 584 bzero(&state, sizeof(state)); 585 state.m = m; 586 state.ro = (struct route *)ro; 587 state.dst = (struct sockaddr *)dst; 588 589 error = ipsec6_output_tunnel(&state, sp, flags); 590 591 m = state.m; 592 ro_pmtu = ro = (struct route_in6 *)state.ro; 593 dst = (struct sockaddr_in6 *)state.dst; 594 if (error) { 595 /* mbuf is already reclaimed in ipsec6_output_tunnel. */ 596 m0 = m = NULL; 597 m = NULL; 598 switch (error) { 599 case EHOSTUNREACH: 600 case ENETUNREACH: 601 case EMSGSIZE: 602 case ENOBUFS: 603 case ENOMEM: 604 break; 605 default: 606 printf("ip6_output (ipsec): error code %d\n", error); 607 /* FALLTHROUGH */ 608 case ENOENT: 609 /* don't show these error codes to the user */ 610 error = 0; 611 break; 612 } 613 goto bad; 614 } 615 616 exthdrs.ip6e_ip6 = m; 617 } 618 #endif /* IPSEC */ 619 620 /* adjust pointer */ 621 ip6 = mtod(m, struct ip6_hdr *); 622 623 bzero(&dst_sa, sizeof(dst_sa)); 624 dst_sa.sin6_family = AF_INET6; 625 dst_sa.sin6_len = sizeof(dst_sa); 626 dst_sa.sin6_addr = ip6->ip6_dst; 627 if ((error = in6_selectroute(&dst_sa, opt, im6o, ro, &ifp, &rt, 0)) 628 != 0) { 629 switch (error) { 630 case EHOSTUNREACH: 631 ip6stat.ip6s_noroute++; 632 break; 633 case EADDRNOTAVAIL: 634 default: 635 break; /* XXX statistics? */ 636 } 637 if (ifp != NULL) 638 in6_ifstat_inc(ifp, ifs6_out_discard); 639 goto bad; 640 } 641 if (rt == NULL) { 642 /* 643 * If in6_selectroute() does not return a route entry, 644 * dst may not have been updated. 645 */ 646 *dst = dst_sa; /* XXX */ 647 } 648 649 /* 650 * then rt (for unicast) and ifp must be non-NULL valid values. 651 */ 652 if ((flags & IPV6_FORWARDING) == 0) { 653 /* XXX: the FORWARDING flag can be set for mrouting. */ 654 in6_ifstat_inc(ifp, ifs6_out_request); 655 } 656 if (rt != NULL) { 657 ia = (struct in6_ifaddr *)(rt->rt_ifa); 658 rt->rt_use++; 659 } 660 661 /* 662 * The outgoing interface must be in the zone of source and 663 * destination addresses. We should use ia_ifp to support the 664 * case of sending packets to an address of our own. 665 */ 666 if (ia != NULL && ia->ia_ifp) 667 origifp = ia->ia_ifp; 668 else 669 origifp = ifp; 670 671 src0 = ip6->ip6_src; 672 if (in6_setscope(&src0, origifp, &zone)) 673 goto badscope; 674 bzero(&src_sa, sizeof(src_sa)); 675 src_sa.sin6_family = AF_INET6; 676 src_sa.sin6_len = sizeof(src_sa); 677 src_sa.sin6_addr = ip6->ip6_src; 678 if (sa6_recoverscope(&src_sa) || zone != src_sa.sin6_scope_id) 679 goto badscope; 680 681 dst0 = ip6->ip6_dst; 682 if (in6_setscope(&dst0, origifp, &zone)) 683 goto badscope; 684 /* re-initialize to be sure */ 685 bzero(&dst_sa, sizeof(dst_sa)); 686 dst_sa.sin6_family = AF_INET6; 687 dst_sa.sin6_len = sizeof(dst_sa); 688 dst_sa.sin6_addr = ip6->ip6_dst; 689 if (sa6_recoverscope(&dst_sa) || zone != dst_sa.sin6_scope_id) 690 goto badscope; 691 692 /* scope check is done. */ 693 goto routefound; 694 695 badscope: 696 ip6stat.ip6s_badscope++; 697 in6_ifstat_inc(origifp, ifs6_out_discard); 698 if (error == 0) 699 error = EHOSTUNREACH; /* XXX */ 700 goto bad; 701 702 routefound: 703 if (rt && !IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) { 704 if (opt && opt->ip6po_nextroute.ro_rt != NULL) { 705 /* 706 * The nexthop is explicitly specified by the 707 * application. We assume the next hop is an IPv6 708 * address. 709 */ 710 dst = (struct sockaddr_in6 *)opt->ip6po_nexthop; 711 } else if ((rt->rt_flags & RTF_GATEWAY)) 712 dst = (struct sockaddr_in6 *)rt->rt_gateway; 713 } 714 715 /* 716 * XXXXXX: original code follows: 717 */ 718 if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) 719 m->m_flags &= ~(M_BCAST | M_MCAST); /* just in case */ 720 else { 721 struct in6_multi *in6m; 722 723 m->m_flags = (m->m_flags & ~M_BCAST) | M_MCAST; 724 725 in6_ifstat_inc(ifp, ifs6_out_mcast); 726 727 /* 728 * Confirm that the outgoing interface supports multicast. 729 */ 730 if (!(ifp->if_flags & IFF_MULTICAST)) { 731 ip6stat.ip6s_noroute++; 732 in6_ifstat_inc(ifp, ifs6_out_discard); 733 error = ENETUNREACH; 734 goto bad; 735 } 736 737 IN6_LOOKUP_MULTI(ip6->ip6_dst, ifp, in6m); 738 if (in6m != NULL && 739 (im6o == NULL || im6o->im6o_multicast_loop)) { 740 /* 741 * If we belong to the destination multicast group 742 * on the outgoing interface, and the caller did not 743 * forbid loopback, loop back a copy. 744 */ 745 ip6_mloopback(ifp, m, dst); 746 } else { 747 /* 748 * If we are acting as a multicast router, perform 749 * multicast forwarding as if the packet had just 750 * arrived on the interface to which we are about 751 * to send. The multicast forwarding function 752 * recursively calls this function, using the 753 * IPV6_FORWARDING flag to prevent infinite recursion. 754 * 755 * Multicasts that are looped back by ip6_mloopback(), 756 * above, will be forwarded by the ip6_input() routine, 757 * if necessary. 758 */ 759 if (ip6_mrouter && (flags & IPV6_FORWARDING) == 0) { 760 if (ip6_mforward(ip6, ifp, m) != 0) { 761 m_freem(m); 762 goto done; 763 } 764 } 765 } 766 /* 767 * Multicasts with a hoplimit of zero may be looped back, 768 * above, but must not be transmitted on a network. 769 * Also, multicasts addressed to the loopback interface 770 * are not sent -- the above call to ip6_mloopback() will 771 * loop back a copy if this host actually belongs to the 772 * destination group on the loopback interface. 773 */ 774 if (ip6->ip6_hlim == 0 || (ifp->if_flags & IFF_LOOPBACK) || 775 IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst)) { 776 m_freem(m); 777 goto done; 778 } 779 } 780 781 /* 782 * Fill the outgoing inteface to tell the upper layer 783 * to increment per-interface statistics. 784 */ 785 if (ifpp) 786 *ifpp = ifp; 787 788 /* Determine path MTU. */ 789 if ((error = ip6_getpmtu(ro_pmtu, ro, ifp, &finaldst, &mtu, 790 &alwaysfrag)) != 0) 791 goto bad; 792 #ifdef IPSEC 793 if (needipsectun) 794 mtu = IPV6_MMTU; 795 #endif 796 797 /* 798 * The caller of this function may specify to use the minimum MTU 799 * in some cases. 800 * An advanced API option (IPV6_USE_MIN_MTU) can also override MTU 801 * setting. The logic is a bit complicated; by default, unicast 802 * packets will follow path MTU while multicast packets will be sent at 803 * the minimum MTU. If IP6PO_MINMTU_ALL is specified, all packets 804 * including unicast ones will be sent at the minimum MTU. Multicast 805 * packets will always be sent at the minimum MTU unless 806 * IP6PO_MINMTU_DISABLE is explicitly specified. 807 * See RFC 3542 for more details. 808 */ 809 if (mtu > IPV6_MMTU) { 810 if ((flags & IPV6_MINMTU)) 811 mtu = IPV6_MMTU; 812 else if (opt && opt->ip6po_minmtu == IP6PO_MINMTU_ALL) 813 mtu = IPV6_MMTU; 814 else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) && 815 (opt == NULL || 816 opt->ip6po_minmtu != IP6PO_MINMTU_DISABLE)) { 817 mtu = IPV6_MMTU; 818 } 819 } 820 821 /* 822 * clear embedded scope identifiers if necessary. 823 * in6_clearscope will touch the addresses only when necessary. 824 */ 825 in6_clearscope(&ip6->ip6_src); 826 in6_clearscope(&ip6->ip6_dst); 827 828 /* 829 * If the outgoing packet contains a hop-by-hop options header, 830 * it must be examined and processed even by the source node. 831 * (RFC 2460, section 4.) 832 */ 833 if (exthdrs.ip6e_hbh) { 834 struct ip6_hbh *hbh = mtod(exthdrs.ip6e_hbh, struct ip6_hbh *); 835 u_int32_t dummy1; /* XXX unused */ 836 u_int32_t dummy2; /* XXX unused */ 837 838 /* 839 * XXX: if we have to send an ICMPv6 error to the sender, 840 * we need the M_LOOP flag since icmp6_error() expects 841 * the IPv6 and the hop-by-hop options header are 842 * continuous unless the flag is set. 843 */ 844 m->m_flags |= M_LOOP; 845 m->m_pkthdr.rcvif = ifp; 846 if (ip6_process_hopopts(m, (u_int8_t *)(hbh + 1), 847 ((hbh->ip6h_len + 1) << 3) - sizeof(struct ip6_hbh), 848 &dummy1, &dummy2) < 0) { 849 /* m was already freed at this point */ 850 error = EINVAL;/* better error? */ 851 goto done; 852 } 853 m->m_flags &= ~M_LOOP; /* XXX */ 854 m->m_pkthdr.rcvif = NULL; 855 } 856 857 #ifdef PFIL_HOOKS 858 /* 859 * Run through list of hooks for output packets. 860 */ 861 if ((error = pfil_run_hooks(&inet6_pfil_hook, &m, ifp, PFIL_OUT)) != 0) 862 goto done; 863 if (m == NULL) 864 goto done; 865 ip6 = mtod(m, struct ip6_hdr *); 866 #endif /* PFIL_HOOKS */ 867 /* 868 * Send the packet to the outgoing interface. 869 * If necessary, do IPv6 fragmentation before sending. 870 * 871 * the logic here is rather complex: 872 * 1: normal case (dontfrag == 0, alwaysfrag == 0) 873 * 1-a: send as is if tlen <= path mtu 874 * 1-b: fragment if tlen > path mtu 875 * 876 * 2: if user asks us not to fragment (dontfrag == 1) 877 * 2-a: send as is if tlen <= interface mtu 878 * 2-b: error if tlen > interface mtu 879 * 880 * 3: if we always need to attach fragment header (alwaysfrag == 1) 881 * always fragment 882 * 883 * 4: if dontfrag == 1 && alwaysfrag == 1 884 * error, as we cannot handle this conflicting request 885 */ 886 tlen = m->m_pkthdr.len; 887 tso = (m->m_pkthdr.csum_flags & M_CSUM_TSOv6) != 0; 888 if (opt && (opt->ip6po_flags & IP6PO_DONTFRAG)) 889 dontfrag = 1; 890 else 891 dontfrag = 0; 892 893 if (dontfrag && alwaysfrag) { /* case 4 */ 894 /* conflicting request - can't transmit */ 895 error = EMSGSIZE; 896 goto bad; 897 } 898 if (dontfrag && (!tso && tlen > IN6_LINKMTU(ifp))) { /* case 2-b */ 899 /* 900 * Even if the DONTFRAG option is specified, we cannot send the 901 * packet when the data length is larger than the MTU of the 902 * outgoing interface. 903 * Notify the error by sending IPV6_PATHMTU ancillary data as 904 * well as returning an error code (the latter is not described 905 * in the API spec.) 906 */ 907 u_int32_t mtu32; 908 struct ip6ctlparam ip6cp; 909 910 mtu32 = (u_int32_t)mtu; 911 bzero(&ip6cp, sizeof(ip6cp)); 912 ip6cp.ip6c_cmdarg = (void *)&mtu32; 913 pfctlinput2(PRC_MSGSIZE, (struct sockaddr *)&ro_pmtu->ro_dst, 914 (void *)&ip6cp); 915 916 error = EMSGSIZE; 917 goto bad; 918 } 919 920 /* 921 * transmit packet without fragmentation 922 */ 923 if (dontfrag || (!alwaysfrag && (tlen <= mtu || tso))) { 924 /* case 1-a and 2-a */ 925 struct in6_ifaddr *ia6; 926 int sw_csum; 927 928 ip6 = mtod(m, struct ip6_hdr *); 929 ia6 = in6_ifawithifp(ifp, &ip6->ip6_src); 930 if (ia6) { 931 /* Record statistics for this interface address. */ 932 ia6->ia_ifa.ifa_data.ifad_outbytes += m->m_pkthdr.len; 933 } 934 #ifdef IPSEC 935 /* clean ipsec history once it goes out of the node */ 936 ipsec_delaux(m); 937 #endif 938 939 sw_csum = m->m_pkthdr.csum_flags & ~ifp->if_csum_flags_tx; 940 if ((sw_csum & (M_CSUM_UDPv6|M_CSUM_TCPv6)) != 0) { 941 if (IN6_NEED_CHECKSUM(ifp, 942 sw_csum & (M_CSUM_UDPv6|M_CSUM_TCPv6))) { 943 in6_delayed_cksum(m); 944 } 945 m->m_pkthdr.csum_flags &= ~(M_CSUM_UDPv6|M_CSUM_TCPv6); 946 } 947 948 if (__predict_true(!tso || 949 (ifp->if_capenable & IFCAP_TSOv6) != 0)) { 950 error = nd6_output(ifp, origifp, m, dst, rt); 951 } else { 952 error = ip6_tso_output(ifp, origifp, m, dst, rt); 953 } 954 goto done; 955 } 956 957 if (tso) { 958 error = EINVAL; /* XXX */ 959 goto bad; 960 } 961 962 /* 963 * try to fragment the packet. case 1-b and 3 964 */ 965 if (mtu < IPV6_MMTU) { 966 /* path MTU cannot be less than IPV6_MMTU */ 967 error = EMSGSIZE; 968 in6_ifstat_inc(ifp, ifs6_out_fragfail); 969 goto bad; 970 } else if (ip6->ip6_plen == 0) { 971 /* jumbo payload cannot be fragmented */ 972 error = EMSGSIZE; 973 in6_ifstat_inc(ifp, ifs6_out_fragfail); 974 goto bad; 975 } else { 976 struct mbuf **mnext, *m_frgpart; 977 struct ip6_frag *ip6f; 978 u_int32_t id = htonl(ip6_randomid()); 979 u_char nextproto; 980 #if 0 /* see below */ 981 struct ip6ctlparam ip6cp; 982 u_int32_t mtu32; 983 #endif 984 985 /* 986 * Too large for the destination or interface; 987 * fragment if possible. 988 * Must be able to put at least 8 bytes per fragment. 989 */ 990 hlen = unfragpartlen; 991 if (mtu > IPV6_MAXPACKET) 992 mtu = IPV6_MAXPACKET; 993 994 #if 0 995 /* 996 * It is believed this code is a leftover from the 997 * development of the IPV6_RECVPATHMTU sockopt and 998 * associated work to implement RFC3542. 999 * It's not entirely clear what the intent of the API 1000 * is at this point, so disable this code for now. 1001 * The IPV6_RECVPATHMTU sockopt and/or IPV6_DONTFRAG 1002 * will send notifications if the application requests. 1003 */ 1004 1005 /* Notify a proper path MTU to applications. */ 1006 mtu32 = (u_int32_t)mtu; 1007 bzero(&ip6cp, sizeof(ip6cp)); 1008 ip6cp.ip6c_cmdarg = (void *)&mtu32; 1009 pfctlinput2(PRC_MSGSIZE, (struct sockaddr *)&ro_pmtu->ro_dst, 1010 (void *)&ip6cp); 1011 #endif 1012 1013 len = (mtu - hlen - sizeof(struct ip6_frag)) & ~7; 1014 if (len < 8) { 1015 error = EMSGSIZE; 1016 in6_ifstat_inc(ifp, ifs6_out_fragfail); 1017 goto bad; 1018 } 1019 1020 mnext = &m->m_nextpkt; 1021 1022 /* 1023 * Change the next header field of the last header in the 1024 * unfragmentable part. 1025 */ 1026 if (exthdrs.ip6e_rthdr) { 1027 nextproto = *mtod(exthdrs.ip6e_rthdr, u_char *); 1028 *mtod(exthdrs.ip6e_rthdr, u_char *) = IPPROTO_FRAGMENT; 1029 } else if (exthdrs.ip6e_dest1) { 1030 nextproto = *mtod(exthdrs.ip6e_dest1, u_char *); 1031 *mtod(exthdrs.ip6e_dest1, u_char *) = IPPROTO_FRAGMENT; 1032 } else if (exthdrs.ip6e_hbh) { 1033 nextproto = *mtod(exthdrs.ip6e_hbh, u_char *); 1034 *mtod(exthdrs.ip6e_hbh, u_char *) = IPPROTO_FRAGMENT; 1035 } else { 1036 nextproto = ip6->ip6_nxt; 1037 ip6->ip6_nxt = IPPROTO_FRAGMENT; 1038 } 1039 1040 if ((m->m_pkthdr.csum_flags & (M_CSUM_UDPv6|M_CSUM_TCPv6)) 1041 != 0) { 1042 if (IN6_NEED_CHECKSUM(ifp, 1043 m->m_pkthdr.csum_flags & 1044 (M_CSUM_UDPv6|M_CSUM_TCPv6))) { 1045 in6_delayed_cksum(m); 1046 } 1047 m->m_pkthdr.csum_flags &= ~(M_CSUM_UDPv6|M_CSUM_TCPv6); 1048 } 1049 1050 /* 1051 * Loop through length of segment after first fragment, 1052 * make new header and copy data of each part and link onto 1053 * chain. 1054 */ 1055 m0 = m; 1056 for (off = hlen; off < tlen; off += len) { 1057 struct mbuf *mlast; 1058 1059 MGETHDR(m, M_DONTWAIT, MT_HEADER); 1060 if (!m) { 1061 error = ENOBUFS; 1062 ip6stat.ip6s_odropped++; 1063 goto sendorfree; 1064 } 1065 m->m_pkthdr.rcvif = NULL; 1066 m->m_flags = m0->m_flags & M_COPYFLAGS; 1067 *mnext = m; 1068 mnext = &m->m_nextpkt; 1069 m->m_data += max_linkhdr; 1070 mhip6 = mtod(m, struct ip6_hdr *); 1071 *mhip6 = *ip6; 1072 m->m_len = sizeof(*mhip6); 1073 error = ip6_insertfraghdr(m0, m, hlen, &ip6f); 1074 if (error) { 1075 ip6stat.ip6s_odropped++; 1076 goto sendorfree; 1077 } 1078 ip6f->ip6f_offlg = htons((u_int16_t)((off - hlen) & ~7)); 1079 if (off + len >= tlen) 1080 len = tlen - off; 1081 else 1082 ip6f->ip6f_offlg |= IP6F_MORE_FRAG; 1083 mhip6->ip6_plen = htons((u_int16_t)(len + hlen + 1084 sizeof(*ip6f) - sizeof(struct ip6_hdr))); 1085 if ((m_frgpart = m_copy(m0, off, len)) == 0) { 1086 error = ENOBUFS; 1087 ip6stat.ip6s_odropped++; 1088 goto sendorfree; 1089 } 1090 for (mlast = m; mlast->m_next; mlast = mlast->m_next) 1091 ; 1092 mlast->m_next = m_frgpart; 1093 m->m_pkthdr.len = len + hlen + sizeof(*ip6f); 1094 m->m_pkthdr.rcvif = (struct ifnet *)0; 1095 ip6f->ip6f_reserved = 0; 1096 ip6f->ip6f_ident = id; 1097 ip6f->ip6f_nxt = nextproto; 1098 ip6stat.ip6s_ofragments++; 1099 in6_ifstat_inc(ifp, ifs6_out_fragcreat); 1100 } 1101 1102 in6_ifstat_inc(ifp, ifs6_out_fragok); 1103 } 1104 1105 /* 1106 * Remove leading garbages. 1107 */ 1108 sendorfree: 1109 m = m0->m_nextpkt; 1110 m0->m_nextpkt = 0; 1111 m_freem(m0); 1112 for (m0 = m; m; m = m0) { 1113 m0 = m->m_nextpkt; 1114 m->m_nextpkt = 0; 1115 if (error == 0) { 1116 struct in6_ifaddr *ia6; 1117 ip6 = mtod(m, struct ip6_hdr *); 1118 ia6 = in6_ifawithifp(ifp, &ip6->ip6_src); 1119 if (ia6) { 1120 /* 1121 * Record statistics for this interface 1122 * address. 1123 */ 1124 ia6->ia_ifa.ifa_data.ifad_outbytes += 1125 m->m_pkthdr.len; 1126 } 1127 #ifdef IPSEC 1128 /* clean ipsec history once it goes out of the node */ 1129 ipsec_delaux(m); 1130 #endif 1131 error = nd6_output(ifp, origifp, m, dst, rt); 1132 } else 1133 m_freem(m); 1134 } 1135 1136 if (error == 0) 1137 ip6stat.ip6s_fragmented++; 1138 1139 done: 1140 /* XXX Second if is invariant? */ 1141 if (ro == &ip6route) 1142 rtcache_free((struct route *)ro); 1143 else if (ro_pmtu == &ip6route) 1144 rtcache_free((struct route *)ro_pmtu); 1145 1146 #ifdef IPSEC 1147 if (sp != NULL) 1148 key_freesp(sp); 1149 #endif /* IPSEC */ 1150 1151 return (error); 1152 1153 freehdrs: 1154 m_freem(exthdrs.ip6e_hbh); /* m_freem will check if mbuf is 0 */ 1155 m_freem(exthdrs.ip6e_dest1); 1156 m_freem(exthdrs.ip6e_rthdr); 1157 m_freem(exthdrs.ip6e_dest2); 1158 /* FALLTHROUGH */ 1159 bad: 1160 m_freem(m); 1161 goto done; 1162 } 1163 1164 static int 1165 ip6_copyexthdr(mp, hdr, hlen) 1166 struct mbuf **mp; 1167 caddr_t hdr; 1168 int hlen; 1169 { 1170 struct mbuf *m; 1171 1172 if (hlen > MCLBYTES) 1173 return (ENOBUFS); /* XXX */ 1174 1175 MGET(m, M_DONTWAIT, MT_DATA); 1176 if (!m) 1177 return (ENOBUFS); 1178 1179 if (hlen > MLEN) { 1180 MCLGET(m, M_DONTWAIT); 1181 if ((m->m_flags & M_EXT) == 0) { 1182 m_free(m); 1183 return (ENOBUFS); 1184 } 1185 } 1186 m->m_len = hlen; 1187 if (hdr) 1188 bcopy(hdr, mtod(m, caddr_t), hlen); 1189 1190 *mp = m; 1191 return (0); 1192 } 1193 1194 /* 1195 * Process a delayed payload checksum calculation. 1196 */ 1197 void 1198 in6_delayed_cksum(struct mbuf *m) 1199 { 1200 uint16_t csum, offset; 1201 1202 KASSERT((m->m_pkthdr.csum_flags & (M_CSUM_UDPv6|M_CSUM_TCPv6)) != 0); 1203 KASSERT((~m->m_pkthdr.csum_flags & (M_CSUM_UDPv6|M_CSUM_TCPv6)) != 0); 1204 KASSERT((m->m_pkthdr.csum_flags 1205 & (M_CSUM_UDPv4|M_CSUM_TCPv4|M_CSUM_TSOv4)) == 0); 1206 1207 offset = M_CSUM_DATA_IPv6_HL(m->m_pkthdr.csum_data); 1208 csum = in6_cksum(m, 0, offset, m->m_pkthdr.len - offset); 1209 if (csum == 0 && (m->m_pkthdr.csum_flags & M_CSUM_UDPv6) != 0) { 1210 csum = 0xffff; 1211 } 1212 1213 offset += M_CSUM_DATA_IPv6_OFFSET(m->m_pkthdr.csum_data); 1214 if ((offset + sizeof(csum)) > m->m_len) { 1215 m_copyback(m, offset, sizeof(csum), &csum); 1216 } else { 1217 *(uint16_t *)(mtod(m, caddr_t) + offset) = csum; 1218 } 1219 } 1220 1221 /* 1222 * Insert jumbo payload option. 1223 */ 1224 static int 1225 ip6_insert_jumboopt(exthdrs, plen) 1226 struct ip6_exthdrs *exthdrs; 1227 u_int32_t plen; 1228 { 1229 struct mbuf *mopt; 1230 u_int8_t *optbuf; 1231 u_int32_t v; 1232 1233 #define JUMBOOPTLEN 8 /* length of jumbo payload option and padding */ 1234 1235 /* 1236 * If there is no hop-by-hop options header, allocate new one. 1237 * If there is one but it doesn't have enough space to store the 1238 * jumbo payload option, allocate a cluster to store the whole options. 1239 * Otherwise, use it to store the options. 1240 */ 1241 if (exthdrs->ip6e_hbh == 0) { 1242 MGET(mopt, M_DONTWAIT, MT_DATA); 1243 if (mopt == 0) 1244 return (ENOBUFS); 1245 mopt->m_len = JUMBOOPTLEN; 1246 optbuf = mtod(mopt, u_int8_t *); 1247 optbuf[1] = 0; /* = ((JUMBOOPTLEN) >> 3) - 1 */ 1248 exthdrs->ip6e_hbh = mopt; 1249 } else { 1250 struct ip6_hbh *hbh; 1251 1252 mopt = exthdrs->ip6e_hbh; 1253 if (M_TRAILINGSPACE(mopt) < JUMBOOPTLEN) { 1254 /* 1255 * XXX assumption: 1256 * - exthdrs->ip6e_hbh is not referenced from places 1257 * other than exthdrs. 1258 * - exthdrs->ip6e_hbh is not an mbuf chain. 1259 */ 1260 int oldoptlen = mopt->m_len; 1261 struct mbuf *n; 1262 1263 /* 1264 * XXX: give up if the whole (new) hbh header does 1265 * not fit even in an mbuf cluster. 1266 */ 1267 if (oldoptlen + JUMBOOPTLEN > MCLBYTES) 1268 return (ENOBUFS); 1269 1270 /* 1271 * As a consequence, we must always prepare a cluster 1272 * at this point. 1273 */ 1274 MGET(n, M_DONTWAIT, MT_DATA); 1275 if (n) { 1276 MCLGET(n, M_DONTWAIT); 1277 if ((n->m_flags & M_EXT) == 0) { 1278 m_freem(n); 1279 n = NULL; 1280 } 1281 } 1282 if (!n) 1283 return (ENOBUFS); 1284 n->m_len = oldoptlen + JUMBOOPTLEN; 1285 bcopy(mtod(mopt, caddr_t), mtod(n, caddr_t), 1286 oldoptlen); 1287 optbuf = mtod(n, u_int8_t *) + oldoptlen; 1288 m_freem(mopt); 1289 mopt = exthdrs->ip6e_hbh = n; 1290 } else { 1291 optbuf = mtod(mopt, u_int8_t *) + mopt->m_len; 1292 mopt->m_len += JUMBOOPTLEN; 1293 } 1294 optbuf[0] = IP6OPT_PADN; 1295 optbuf[1] = 0; 1296 1297 /* 1298 * Adjust the header length according to the pad and 1299 * the jumbo payload option. 1300 */ 1301 hbh = mtod(mopt, struct ip6_hbh *); 1302 hbh->ip6h_len += (JUMBOOPTLEN >> 3); 1303 } 1304 1305 /* fill in the option. */ 1306 optbuf[2] = IP6OPT_JUMBO; 1307 optbuf[3] = 4; 1308 v = (u_int32_t)htonl(plen + JUMBOOPTLEN); 1309 bcopy(&v, &optbuf[4], sizeof(u_int32_t)); 1310 1311 /* finally, adjust the packet header length */ 1312 exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN; 1313 1314 return (0); 1315 #undef JUMBOOPTLEN 1316 } 1317 1318 /* 1319 * Insert fragment header and copy unfragmentable header portions. 1320 */ 1321 static int 1322 ip6_insertfraghdr(m0, m, hlen, frghdrp) 1323 struct mbuf *m0, *m; 1324 int hlen; 1325 struct ip6_frag **frghdrp; 1326 { 1327 struct mbuf *n, *mlast; 1328 1329 if (hlen > sizeof(struct ip6_hdr)) { 1330 n = m_copym(m0, sizeof(struct ip6_hdr), 1331 hlen - sizeof(struct ip6_hdr), M_DONTWAIT); 1332 if (n == 0) 1333 return (ENOBUFS); 1334 m->m_next = n; 1335 } else 1336 n = m; 1337 1338 /* Search for the last mbuf of unfragmentable part. */ 1339 for (mlast = n; mlast->m_next; mlast = mlast->m_next) 1340 ; 1341 1342 if ((mlast->m_flags & M_EXT) == 0 && 1343 M_TRAILINGSPACE(mlast) >= sizeof(struct ip6_frag)) { 1344 /* use the trailing space of the last mbuf for the fragment hdr */ 1345 *frghdrp = (struct ip6_frag *)(mtod(mlast, caddr_t) + 1346 mlast->m_len); 1347 mlast->m_len += sizeof(struct ip6_frag); 1348 m->m_pkthdr.len += sizeof(struct ip6_frag); 1349 } else { 1350 /* allocate a new mbuf for the fragment header */ 1351 struct mbuf *mfrg; 1352 1353 MGET(mfrg, M_DONTWAIT, MT_DATA); 1354 if (mfrg == 0) 1355 return (ENOBUFS); 1356 mfrg->m_len = sizeof(struct ip6_frag); 1357 *frghdrp = mtod(mfrg, struct ip6_frag *); 1358 mlast->m_next = mfrg; 1359 } 1360 1361 return (0); 1362 } 1363 1364 static int 1365 ip6_getpmtu(ro_pmtu, ro, ifp, dst, mtup, alwaysfragp) 1366 struct route_in6 *ro_pmtu, *ro; 1367 struct ifnet *ifp; 1368 struct in6_addr *dst; 1369 u_long *mtup; 1370 int *alwaysfragp; 1371 { 1372 u_int32_t mtu = 0; 1373 int alwaysfrag = 0; 1374 int error = 0; 1375 1376 if (ro_pmtu != ro) { 1377 /* The first hop and the final destination may differ. */ 1378 struct sockaddr_in6 *sa6_dst = 1379 (struct sockaddr_in6 *)&ro_pmtu->ro_dst; 1380 if (!IN6_ARE_ADDR_EQUAL(&sa6_dst->sin6_addr, dst)) 1381 rtcache_free((struct route *)ro_pmtu); 1382 else 1383 rtcache_check((struct route *)ro_pmtu); 1384 if (ro_pmtu->ro_rt == NULL) { 1385 bzero(sa6_dst, sizeof(*sa6_dst)); /* for safety */ 1386 sa6_dst->sin6_family = AF_INET6; 1387 sa6_dst->sin6_len = sizeof(struct sockaddr_in6); 1388 sa6_dst->sin6_addr = *dst; 1389 rtcache_init((struct route *)ro_pmtu); 1390 } 1391 } 1392 if (ro_pmtu->ro_rt != NULL) { 1393 u_int32_t ifmtu; 1394 1395 if (ifp == NULL) 1396 ifp = ro_pmtu->ro_rt->rt_ifp; 1397 ifmtu = IN6_LINKMTU(ifp); 1398 mtu = ro_pmtu->ro_rt->rt_rmx.rmx_mtu; 1399 if (mtu == 0) 1400 mtu = ifmtu; 1401 else if (mtu < IPV6_MMTU) { 1402 /* 1403 * RFC2460 section 5, last paragraph: 1404 * if we record ICMPv6 too big message with 1405 * mtu < IPV6_MMTU, transmit packets sized IPV6_MMTU 1406 * or smaller, with fragment header attached. 1407 * (fragment header is needed regardless from the 1408 * packet size, for translators to identify packets) 1409 */ 1410 alwaysfrag = 1; 1411 mtu = IPV6_MMTU; 1412 } else if (mtu > ifmtu) { 1413 /* 1414 * The MTU on the route is larger than the MTU on 1415 * the interface! This shouldn't happen, unless the 1416 * MTU of the interface has been changed after the 1417 * interface was brought up. Change the MTU in the 1418 * route to match the interface MTU (as long as the 1419 * field isn't locked). 1420 */ 1421 mtu = ifmtu; 1422 if (!(ro_pmtu->ro_rt->rt_rmx.rmx_locks & RTV_MTU)) 1423 ro_pmtu->ro_rt->rt_rmx.rmx_mtu = mtu; 1424 } 1425 } else if (ifp) { 1426 mtu = IN6_LINKMTU(ifp); 1427 } else 1428 error = EHOSTUNREACH; /* XXX */ 1429 1430 *mtup = mtu; 1431 if (alwaysfragp) 1432 *alwaysfragp = alwaysfrag; 1433 return (error); 1434 } 1435 1436 /* 1437 * IP6 socket option processing. 1438 */ 1439 int 1440 ip6_ctloutput(op, so, level, optname, mp) 1441 int op; 1442 struct socket *so; 1443 int level, optname; 1444 struct mbuf **mp; 1445 { 1446 int privileged, optdatalen, uproto; 1447 void *optdata; 1448 struct in6pcb *in6p = sotoin6pcb(so); 1449 struct mbuf *m = *mp; 1450 int error, optval; 1451 int optlen; 1452 struct lwp *l = curlwp; /* XXX */ 1453 1454 optlen = m ? m->m_len : 0; 1455 error = optval = 0; 1456 privileged = (l == 0 || kauth_authorize_generic(l->l_cred, 1457 KAUTH_GENERIC_ISSUSER, &l->l_acflag)) ? 0 : 1; 1458 uproto = (int)so->so_proto->pr_protocol; 1459 1460 if (level == IPPROTO_IPV6) { 1461 switch (op) { 1462 case PRCO_SETOPT: 1463 switch (optname) { 1464 #ifdef RFC2292 1465 case IPV6_2292PKTOPTIONS: 1466 /* m is freed in ip6_pcbopts */ 1467 error = ip6_pcbopts(&in6p->in6p_outputopts, 1468 m, so); 1469 break; 1470 #endif 1471 1472 /* 1473 * Use of some Hop-by-Hop options or some 1474 * Destination options, might require special 1475 * privilege. That is, normal applications 1476 * (without special privilege) might be forbidden 1477 * from setting certain options in outgoing packets, 1478 * and might never see certain options in received 1479 * packets. [RFC 2292 Section 6] 1480 * KAME specific note: 1481 * KAME prevents non-privileged users from sending or 1482 * receiving ANY hbh/dst options in order to avoid 1483 * overhead of parsing options in the kernel. 1484 */ 1485 case IPV6_RECVHOPOPTS: 1486 case IPV6_RECVDSTOPTS: 1487 case IPV6_RECVRTHDRDSTOPTS: 1488 if (!privileged) { 1489 error = EPERM; 1490 break; 1491 } 1492 /* FALLTHROUGH */ 1493 case IPV6_UNICAST_HOPS: 1494 case IPV6_HOPLIMIT: 1495 case IPV6_FAITH: 1496 1497 case IPV6_RECVPKTINFO: 1498 case IPV6_RECVHOPLIMIT: 1499 case IPV6_RECVRTHDR: 1500 case IPV6_RECVPATHMTU: 1501 case IPV6_RECVTCLASS: 1502 case IPV6_V6ONLY: 1503 if (optlen != sizeof(int)) { 1504 error = EINVAL; 1505 break; 1506 } 1507 optval = *mtod(m, int *); 1508 switch (optname) { 1509 1510 case IPV6_UNICAST_HOPS: 1511 if (optval < -1 || optval >= 256) 1512 error = EINVAL; 1513 else { 1514 /* -1 = kernel default */ 1515 in6p->in6p_hops = optval; 1516 } 1517 break; 1518 #define OPTSET(bit) \ 1519 do { \ 1520 if (optval) \ 1521 in6p->in6p_flags |= (bit); \ 1522 else \ 1523 in6p->in6p_flags &= ~(bit); \ 1524 } while (/*CONSTCOND*/ 0) 1525 1526 #ifdef RFC2292 1527 #define OPTSET2292(bit) \ 1528 do { \ 1529 in6p->in6p_flags |= IN6P_RFC2292; \ 1530 if (optval) \ 1531 in6p->in6p_flags |= (bit); \ 1532 else \ 1533 in6p->in6p_flags &= ~(bit); \ 1534 } while (/*CONSTCOND*/ 0) 1535 #endif 1536 1537 #define OPTBIT(bit) (in6p->in6p_flags & (bit) ? 1 : 0) 1538 1539 case IPV6_RECVPKTINFO: 1540 #ifdef RFC2292 1541 /* cannot mix with RFC2292 */ 1542 if (OPTBIT(IN6P_RFC2292)) { 1543 error = EINVAL; 1544 break; 1545 } 1546 #endif 1547 OPTSET(IN6P_PKTINFO); 1548 break; 1549 1550 case IPV6_HOPLIMIT: 1551 { 1552 struct ip6_pktopts **optp; 1553 1554 #ifdef RFC2292 1555 /* cannot mix with RFC2292 */ 1556 if (OPTBIT(IN6P_RFC2292)) { 1557 error = EINVAL; 1558 break; 1559 } 1560 #endif 1561 optp = &in6p->in6p_outputopts; 1562 error = ip6_pcbopt(IPV6_HOPLIMIT, 1563 (u_char *)&optval, 1564 sizeof(optval), 1565 optp, 1566 privileged, uproto); 1567 break; 1568 } 1569 1570 case IPV6_RECVHOPLIMIT: 1571 #ifdef RFC2292 1572 /* cannot mix with RFC2292 */ 1573 if (OPTBIT(IN6P_RFC2292)) { 1574 error = EINVAL; 1575 break; 1576 } 1577 #endif 1578 OPTSET(IN6P_HOPLIMIT); 1579 break; 1580 1581 case IPV6_RECVHOPOPTS: 1582 #ifdef RFC2292 1583 /* cannot mix with RFC2292 */ 1584 if (OPTBIT(IN6P_RFC2292)) { 1585 error = EINVAL; 1586 break; 1587 } 1588 #endif 1589 OPTSET(IN6P_HOPOPTS); 1590 break; 1591 1592 case IPV6_RECVDSTOPTS: 1593 #ifdef RFC2292 1594 /* cannot mix with RFC2292 */ 1595 if (OPTBIT(IN6P_RFC2292)) { 1596 error = EINVAL; 1597 break; 1598 } 1599 #endif 1600 OPTSET(IN6P_DSTOPTS); 1601 break; 1602 1603 case IPV6_RECVRTHDRDSTOPTS: 1604 #ifdef RFC2292 1605 /* cannot mix with RFC2292 */ 1606 if (OPTBIT(IN6P_RFC2292)) { 1607 error = EINVAL; 1608 break; 1609 } 1610 #endif 1611 OPTSET(IN6P_RTHDRDSTOPTS); 1612 break; 1613 1614 case IPV6_RECVRTHDR: 1615 #ifdef RFC2292 1616 /* cannot mix with RFC2292 */ 1617 if (OPTBIT(IN6P_RFC2292)) { 1618 error = EINVAL; 1619 break; 1620 } 1621 #endif 1622 OPTSET(IN6P_RTHDR); 1623 break; 1624 1625 case IPV6_FAITH: 1626 OPTSET(IN6P_FAITH); 1627 break; 1628 1629 case IPV6_RECVPATHMTU: 1630 /* 1631 * We ignore this option for TCP 1632 * sockets. 1633 * (RFC3542 leaves this case 1634 * unspecified.) 1635 */ 1636 if (uproto != IPPROTO_TCP) 1637 OPTSET(IN6P_MTU); 1638 break; 1639 1640 case IPV6_V6ONLY: 1641 /* 1642 * make setsockopt(IPV6_V6ONLY) 1643 * available only prior to bind(2). 1644 * see ipng mailing list, Jun 22 2001. 1645 */ 1646 if (in6p->in6p_lport || 1647 !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr)) { 1648 error = EINVAL; 1649 break; 1650 } 1651 #ifdef INET6_BINDV6ONLY 1652 if (!optval) 1653 error = EINVAL; 1654 #else 1655 OPTSET(IN6P_IPV6_V6ONLY); 1656 #endif 1657 break; 1658 case IPV6_RECVTCLASS: 1659 #ifdef RFC2292 1660 /* cannot mix with RFC2292 XXX */ 1661 if (OPTBIT(IN6P_RFC2292)) { 1662 error = EINVAL; 1663 break; 1664 } 1665 #endif 1666 OPTSET(IN6P_TCLASS); 1667 break; 1668 1669 } 1670 break; 1671 1672 case IPV6_OTCLASS: 1673 { 1674 struct ip6_pktopts **optp; 1675 u_int8_t tclass; 1676 1677 if (optlen != sizeof(tclass)) { 1678 error = EINVAL; 1679 break; 1680 } 1681 tclass = *mtod(m, u_int8_t *); 1682 optp = &in6p->in6p_outputopts; 1683 error = ip6_pcbopt(optname, 1684 (u_char *)&tclass, 1685 sizeof(tclass), 1686 optp, 1687 privileged, uproto); 1688 break; 1689 } 1690 1691 case IPV6_TCLASS: 1692 case IPV6_DONTFRAG: 1693 case IPV6_USE_MIN_MTU: 1694 if (optlen != sizeof(optval)) { 1695 error = EINVAL; 1696 break; 1697 } 1698 optval = *mtod(m, int *); 1699 { 1700 struct ip6_pktopts **optp; 1701 optp = &in6p->in6p_outputopts; 1702 error = ip6_pcbopt(optname, 1703 (u_char *)&optval, 1704 sizeof(optval), 1705 optp, 1706 privileged, uproto); 1707 break; 1708 } 1709 1710 #ifdef RFC2292 1711 case IPV6_2292PKTINFO: 1712 case IPV6_2292HOPLIMIT: 1713 case IPV6_2292HOPOPTS: 1714 case IPV6_2292DSTOPTS: 1715 case IPV6_2292RTHDR: 1716 /* RFC 2292 */ 1717 if (optlen != sizeof(int)) { 1718 error = EINVAL; 1719 break; 1720 } 1721 optval = *mtod(m, int *); 1722 switch (optname) { 1723 case IPV6_2292PKTINFO: 1724 OPTSET2292(IN6P_PKTINFO); 1725 break; 1726 case IPV6_2292HOPLIMIT: 1727 OPTSET2292(IN6P_HOPLIMIT); 1728 break; 1729 case IPV6_2292HOPOPTS: 1730 /* 1731 * Check super-user privilege. 1732 * See comments for IPV6_RECVHOPOPTS. 1733 */ 1734 if (!privileged) 1735 return (EPERM); 1736 OPTSET2292(IN6P_HOPOPTS); 1737 break; 1738 case IPV6_2292DSTOPTS: 1739 if (!privileged) 1740 return (EPERM); 1741 OPTSET2292(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); /* XXX */ 1742 break; 1743 case IPV6_2292RTHDR: 1744 OPTSET2292(IN6P_RTHDR); 1745 break; 1746 } 1747 break; 1748 #endif 1749 case IPV6_PKTINFO: 1750 case IPV6_HOPOPTS: 1751 case IPV6_RTHDR: 1752 case IPV6_DSTOPTS: 1753 case IPV6_RTHDRDSTOPTS: 1754 case IPV6_NEXTHOP: 1755 { 1756 /* new advanced API (RFC3542) */ 1757 u_char *optbuf; 1758 int optbuflen; 1759 struct ip6_pktopts **optp; 1760 1761 #ifdef RFC2292 1762 /* cannot mix with RFC2292 */ 1763 if (OPTBIT(IN6P_RFC2292)) { 1764 error = EINVAL; 1765 break; 1766 } 1767 #endif 1768 1769 if (m && m->m_next) { 1770 error = EINVAL; /* XXX */ 1771 break; 1772 } 1773 if (m) { 1774 optbuf = mtod(m, u_char *); 1775 optbuflen = m->m_len; 1776 } else { 1777 optbuf = NULL; 1778 optbuflen = 0; 1779 } 1780 optp = &in6p->in6p_outputopts; 1781 error = ip6_pcbopt(optname, 1782 optbuf, optbuflen, 1783 optp, privileged, uproto); 1784 break; 1785 } 1786 #undef OPTSET 1787 1788 case IPV6_MULTICAST_IF: 1789 case IPV6_MULTICAST_HOPS: 1790 case IPV6_MULTICAST_LOOP: 1791 case IPV6_JOIN_GROUP: 1792 case IPV6_LEAVE_GROUP: 1793 error = ip6_setmoptions(optname, 1794 &in6p->in6p_moptions, m); 1795 break; 1796 1797 case IPV6_PORTRANGE: 1798 optval = *mtod(m, int *); 1799 1800 switch (optval) { 1801 case IPV6_PORTRANGE_DEFAULT: 1802 in6p->in6p_flags &= ~(IN6P_LOWPORT); 1803 in6p->in6p_flags &= ~(IN6P_HIGHPORT); 1804 break; 1805 1806 case IPV6_PORTRANGE_HIGH: 1807 in6p->in6p_flags &= ~(IN6P_LOWPORT); 1808 in6p->in6p_flags |= IN6P_HIGHPORT; 1809 break; 1810 1811 case IPV6_PORTRANGE_LOW: 1812 in6p->in6p_flags &= ~(IN6P_HIGHPORT); 1813 in6p->in6p_flags |= IN6P_LOWPORT; 1814 break; 1815 1816 default: 1817 error = EINVAL; 1818 break; 1819 } 1820 break; 1821 1822 #ifdef IPSEC 1823 case IPV6_IPSEC_POLICY: 1824 { 1825 caddr_t req = NULL; 1826 size_t len = 0; 1827 if (m) { 1828 req = mtod(m, caddr_t); 1829 len = m->m_len; 1830 } 1831 error = ipsec6_set_policy(in6p, optname, req, 1832 len, privileged); 1833 } 1834 break; 1835 #endif /* IPSEC */ 1836 1837 default: 1838 error = ENOPROTOOPT; 1839 break; 1840 } 1841 if (m) 1842 (void)m_free(m); 1843 break; 1844 1845 case PRCO_GETOPT: 1846 switch (optname) { 1847 #ifdef RFC2292 1848 case IPV6_2292PKTOPTIONS: 1849 /* 1850 * RFC3542 (effectively) deprecated the 1851 * semantics of the 2292-style pktoptions. 1852 * Since it was not reliable in nature (i.e., 1853 * applications had to expect the lack of some 1854 * information after all), it would make sense 1855 * to simplify this part by always returning 1856 * empty data. 1857 */ 1858 *mp = m_get(M_WAIT, MT_SOOPTS); 1859 (*mp)->m_len = 0; 1860 break; 1861 #endif 1862 1863 case IPV6_RECVHOPOPTS: 1864 case IPV6_RECVDSTOPTS: 1865 case IPV6_RECVRTHDRDSTOPTS: 1866 case IPV6_UNICAST_HOPS: 1867 case IPV6_RECVPKTINFO: 1868 case IPV6_RECVHOPLIMIT: 1869 case IPV6_RECVRTHDR: 1870 case IPV6_RECVPATHMTU: 1871 1872 case IPV6_FAITH: 1873 case IPV6_V6ONLY: 1874 case IPV6_PORTRANGE: 1875 case IPV6_RECVTCLASS: 1876 switch (optname) { 1877 1878 case IPV6_RECVHOPOPTS: 1879 optval = OPTBIT(IN6P_HOPOPTS); 1880 break; 1881 1882 case IPV6_RECVDSTOPTS: 1883 optval = OPTBIT(IN6P_DSTOPTS); 1884 break; 1885 1886 case IPV6_RECVRTHDRDSTOPTS: 1887 optval = OPTBIT(IN6P_RTHDRDSTOPTS); 1888 break; 1889 1890 case IPV6_UNICAST_HOPS: 1891 optval = in6p->in6p_hops; 1892 break; 1893 1894 case IPV6_RECVPKTINFO: 1895 optval = OPTBIT(IN6P_PKTINFO); 1896 break; 1897 1898 case IPV6_RECVHOPLIMIT: 1899 optval = OPTBIT(IN6P_HOPLIMIT); 1900 break; 1901 1902 case IPV6_RECVRTHDR: 1903 optval = OPTBIT(IN6P_RTHDR); 1904 break; 1905 1906 case IPV6_RECVPATHMTU: 1907 optval = OPTBIT(IN6P_MTU); 1908 break; 1909 1910 case IPV6_FAITH: 1911 optval = OPTBIT(IN6P_FAITH); 1912 break; 1913 1914 case IPV6_V6ONLY: 1915 optval = OPTBIT(IN6P_IPV6_V6ONLY); 1916 break; 1917 1918 case IPV6_PORTRANGE: 1919 { 1920 int flags; 1921 flags = in6p->in6p_flags; 1922 if (flags & IN6P_HIGHPORT) 1923 optval = IPV6_PORTRANGE_HIGH; 1924 else if (flags & IN6P_LOWPORT) 1925 optval = IPV6_PORTRANGE_LOW; 1926 else 1927 optval = 0; 1928 break; 1929 } 1930 case IPV6_RECVTCLASS: 1931 optval = OPTBIT(IN6P_TCLASS); 1932 break; 1933 1934 } 1935 if (error) 1936 break; 1937 *mp = m = m_get(M_WAIT, MT_SOOPTS); 1938 m->m_len = sizeof(int); 1939 *mtod(m, int *) = optval; 1940 break; 1941 1942 case IPV6_PATHMTU: 1943 { 1944 u_long pmtu = 0; 1945 struct ip6_mtuinfo mtuinfo; 1946 struct route_in6 *ro = (struct route_in6 *)&in6p 1947 ->in6p_route; 1948 1949 if (!(so->so_state & SS_ISCONNECTED)) 1950 return (ENOTCONN); 1951 /* 1952 * XXX: we dot not consider the case of source 1953 * routing, or optional information to specify 1954 * the outgoing interface. 1955 */ 1956 error = ip6_getpmtu(ro, NULL, NULL, 1957 &in6p->in6p_faddr, &pmtu, NULL); 1958 if (error) 1959 break; 1960 if (pmtu > IPV6_MAXPACKET) 1961 pmtu = IPV6_MAXPACKET; 1962 1963 memset(&mtuinfo, 0, sizeof(mtuinfo)); 1964 mtuinfo.ip6m_mtu = (u_int32_t)pmtu; 1965 optdata = (void *)&mtuinfo; 1966 optdatalen = sizeof(mtuinfo); 1967 if (optdatalen > MCLBYTES) 1968 return (EMSGSIZE); /* XXX */ 1969 *mp = m = m_get(M_WAIT, MT_SOOPTS); 1970 if (optdatalen > MLEN) 1971 MCLGET(m, M_WAIT); 1972 m->m_len = optdatalen; 1973 memcpy(mtod(m, void *), optdata, optdatalen); 1974 break; 1975 } 1976 1977 #ifdef RFC2292 1978 case IPV6_2292PKTINFO: 1979 case IPV6_2292HOPLIMIT: 1980 case IPV6_2292HOPOPTS: 1981 case IPV6_2292RTHDR: 1982 case IPV6_2292DSTOPTS: 1983 switch (optname) { 1984 case IPV6_2292PKTINFO: 1985 optval = OPTBIT(IN6P_PKTINFO); 1986 break; 1987 case IPV6_2292HOPLIMIT: 1988 optval = OPTBIT(IN6P_HOPLIMIT); 1989 break; 1990 case IPV6_2292HOPOPTS: 1991 optval = OPTBIT(IN6P_HOPOPTS); 1992 break; 1993 case IPV6_2292RTHDR: 1994 optval = OPTBIT(IN6P_RTHDR); 1995 break; 1996 case IPV6_2292DSTOPTS: 1997 optval = OPTBIT(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); 1998 break; 1999 } 2000 *mp = m = m_get(M_WAIT, MT_SOOPTS); 2001 m->m_len = sizeof(int); 2002 *mtod(m, int *) = optval; 2003 break; 2004 #endif 2005 case IPV6_PKTINFO: 2006 case IPV6_HOPOPTS: 2007 case IPV6_RTHDR: 2008 case IPV6_DSTOPTS: 2009 case IPV6_RTHDRDSTOPTS: 2010 case IPV6_NEXTHOP: 2011 case IPV6_OTCLASS: 2012 case IPV6_TCLASS: 2013 case IPV6_DONTFRAG: 2014 case IPV6_USE_MIN_MTU: 2015 error = ip6_getpcbopt(in6p->in6p_outputopts, 2016 optname, mp); 2017 break; 2018 2019 case IPV6_MULTICAST_IF: 2020 case IPV6_MULTICAST_HOPS: 2021 case IPV6_MULTICAST_LOOP: 2022 case IPV6_JOIN_GROUP: 2023 case IPV6_LEAVE_GROUP: 2024 error = ip6_getmoptions(optname, 2025 in6p->in6p_moptions, mp); 2026 break; 2027 2028 #ifdef IPSEC 2029 case IPV6_IPSEC_POLICY: 2030 { 2031 caddr_t req = NULL; 2032 size_t len = 0; 2033 if (m) { 2034 req = mtod(m, caddr_t); 2035 len = m->m_len; 2036 } 2037 error = ipsec6_get_policy(in6p, req, len, mp); 2038 break; 2039 } 2040 #endif /* IPSEC */ 2041 2042 2043 2044 2045 default: 2046 error = ENOPROTOOPT; 2047 break; 2048 } 2049 break; 2050 } 2051 } else { 2052 error = EINVAL; 2053 if (op == PRCO_SETOPT && *mp) 2054 (void)m_free(*mp); 2055 } 2056 return (error); 2057 } 2058 2059 int 2060 ip6_raw_ctloutput(op, so, level, optname, mp) 2061 int op; 2062 struct socket *so; 2063 int level, optname; 2064 struct mbuf **mp; 2065 { 2066 int error = 0, optval, optlen; 2067 const int icmp6off = offsetof(struct icmp6_hdr, icmp6_cksum); 2068 struct in6pcb *in6p = sotoin6pcb(so); 2069 struct mbuf *m = *mp; 2070 2071 optlen = m ? m->m_len : 0; 2072 2073 if (level != IPPROTO_IPV6) { 2074 if (op == PRCO_SETOPT && *mp) 2075 (void)m_free(*mp); 2076 return (EINVAL); 2077 } 2078 2079 switch (optname) { 2080 case IPV6_CHECKSUM: 2081 /* 2082 * For ICMPv6 sockets, no modification allowed for checksum 2083 * offset, permit "no change" values to help existing apps. 2084 * 2085 * XXX RFC3542 says: "An attempt to set IPV6_CHECKSUM 2086 * for an ICMPv6 socket will fail." The current 2087 * behavior does not meet RFC3542. 2088 */ 2089 switch (op) { 2090 case PRCO_SETOPT: 2091 if (optlen != sizeof(int)) { 2092 error = EINVAL; 2093 break; 2094 } 2095 optval = *mtod(m, int *); 2096 if ((optval % 2) != 0) { 2097 /* the API assumes even offset values */ 2098 error = EINVAL; 2099 } else if (so->so_proto->pr_protocol == 2100 IPPROTO_ICMPV6) { 2101 if (optval != icmp6off) 2102 error = EINVAL; 2103 } else 2104 in6p->in6p_cksum = optval; 2105 break; 2106 2107 case PRCO_GETOPT: 2108 if (so->so_proto->pr_protocol == IPPROTO_ICMPV6) 2109 optval = icmp6off; 2110 else 2111 optval = in6p->in6p_cksum; 2112 2113 *mp = m = m_get(M_WAIT, MT_SOOPTS); 2114 m->m_len = sizeof(int); 2115 *mtod(m, int *) = optval; 2116 break; 2117 2118 default: 2119 error = EINVAL; 2120 break; 2121 } 2122 break; 2123 2124 default: 2125 error = ENOPROTOOPT; 2126 break; 2127 } 2128 2129 if (op == PRCO_SETOPT && m) 2130 (void)m_free(m); 2131 2132 return (error); 2133 } 2134 2135 #ifdef RFC2292 2136 /* 2137 * Set up IP6 options in pcb for insertion in output packets or 2138 * specifying behavior of outgoing packets. 2139 */ 2140 static int 2141 ip6_pcbopts(pktopt, m, so) 2142 struct ip6_pktopts **pktopt; 2143 struct mbuf *m; 2144 struct socket *so; 2145 { 2146 struct ip6_pktopts *opt = *pktopt; 2147 int error = 0; 2148 struct lwp *l = curlwp; /* XXX */ 2149 int priv = 0; 2150 2151 /* turn off any old options. */ 2152 if (opt) { 2153 #ifdef DIAGNOSTIC 2154 if (opt->ip6po_pktinfo || opt->ip6po_nexthop || 2155 opt->ip6po_hbh || opt->ip6po_dest1 || opt->ip6po_dest2 || 2156 opt->ip6po_rhinfo.ip6po_rhi_rthdr) 2157 printf("ip6_pcbopts: all specified options are cleared.\n"); 2158 #endif 2159 ip6_clearpktopts(opt, -1); 2160 } else 2161 opt = malloc(sizeof(*opt), M_IP6OPT, M_WAITOK); 2162 *pktopt = NULL; 2163 2164 if (!m || m->m_len == 0) { 2165 /* 2166 * Only turning off any previous options, regardless of 2167 * whether the opt is just created or given. 2168 */ 2169 free(opt, M_IP6OPT); 2170 return (0); 2171 } 2172 2173 /* set options specified by user. */ 2174 if (l && !kauth_authorize_generic(l->l_cred, KAUTH_GENERIC_ISSUSER, 2175 &l->l_acflag)) 2176 priv = 1; 2177 if ((error = ip6_setpktopts(m, opt, NULL, priv, 2178 so->so_proto->pr_protocol)) != 0) { 2179 ip6_clearpktopts(opt, -1); /* XXX: discard all options */ 2180 free(opt, M_IP6OPT); 2181 return (error); 2182 } 2183 *pktopt = opt; 2184 return (0); 2185 } 2186 #endif 2187 2188 /* 2189 * initialize ip6_pktopts. beware that there are non-zero default values in 2190 * the struct. 2191 */ 2192 void 2193 ip6_initpktopts(struct ip6_pktopts *opt) 2194 { 2195 2196 memset(opt, 0, sizeof(*opt)); 2197 opt->ip6po_hlim = -1; /* -1 means default hop limit */ 2198 opt->ip6po_tclass = -1; /* -1 means default traffic class */ 2199 opt->ip6po_minmtu = IP6PO_MINMTU_MCASTONLY; 2200 } 2201 2202 #define sin6tosa(sin6) ((struct sockaddr *)(sin6)) /* XXX */ 2203 static int 2204 ip6_pcbopt(int optname, u_char *buf, int len, struct ip6_pktopts **pktopt, 2205 int priv, int uproto) 2206 { 2207 struct ip6_pktopts *opt; 2208 2209 if (*pktopt == NULL) { 2210 *pktopt = malloc(sizeof(struct ip6_pktopts), M_IP6OPT, 2211 M_WAITOK); 2212 ip6_initpktopts(*pktopt); 2213 } 2214 opt = *pktopt; 2215 2216 return (ip6_setpktopt(optname, buf, len, opt, priv, 1, 0, uproto)); 2217 } 2218 2219 static int 2220 ip6_getpcbopt(struct ip6_pktopts *pktopt, int optname, struct mbuf **mp) 2221 { 2222 void *optdata = NULL; 2223 int optdatalen = 0; 2224 struct ip6_ext *ip6e; 2225 int error = 0; 2226 struct in6_pktinfo null_pktinfo; 2227 int deftclass = 0, on; 2228 int defminmtu = IP6PO_MINMTU_MCASTONLY; 2229 struct mbuf *m; 2230 2231 switch (optname) { 2232 case IPV6_PKTINFO: 2233 if (pktopt && pktopt->ip6po_pktinfo) 2234 optdata = (void *)pktopt->ip6po_pktinfo; 2235 else { 2236 /* XXX: we don't have to do this every time... */ 2237 memset(&null_pktinfo, 0, sizeof(null_pktinfo)); 2238 optdata = (void *)&null_pktinfo; 2239 } 2240 optdatalen = sizeof(struct in6_pktinfo); 2241 break; 2242 case IPV6_OTCLASS: 2243 /* XXX */ 2244 return (EINVAL); 2245 case IPV6_TCLASS: 2246 if (pktopt && pktopt->ip6po_tclass >= 0) 2247 optdata = (void *)&pktopt->ip6po_tclass; 2248 else 2249 optdata = (void *)&deftclass; 2250 optdatalen = sizeof(int); 2251 break; 2252 case IPV6_HOPOPTS: 2253 if (pktopt && pktopt->ip6po_hbh) { 2254 optdata = (void *)pktopt->ip6po_hbh; 2255 ip6e = (struct ip6_ext *)pktopt->ip6po_hbh; 2256 optdatalen = (ip6e->ip6e_len + 1) << 3; 2257 } 2258 break; 2259 case IPV6_RTHDR: 2260 if (pktopt && pktopt->ip6po_rthdr) { 2261 optdata = (void *)pktopt->ip6po_rthdr; 2262 ip6e = (struct ip6_ext *)pktopt->ip6po_rthdr; 2263 optdatalen = (ip6e->ip6e_len + 1) << 3; 2264 } 2265 break; 2266 case IPV6_RTHDRDSTOPTS: 2267 if (pktopt && pktopt->ip6po_dest1) { 2268 optdata = (void *)pktopt->ip6po_dest1; 2269 ip6e = (struct ip6_ext *)pktopt->ip6po_dest1; 2270 optdatalen = (ip6e->ip6e_len + 1) << 3; 2271 } 2272 break; 2273 case IPV6_DSTOPTS: 2274 if (pktopt && pktopt->ip6po_dest2) { 2275 optdata = (void *)pktopt->ip6po_dest2; 2276 ip6e = (struct ip6_ext *)pktopt->ip6po_dest2; 2277 optdatalen = (ip6e->ip6e_len + 1) << 3; 2278 } 2279 break; 2280 case IPV6_NEXTHOP: 2281 if (pktopt && pktopt->ip6po_nexthop) { 2282 optdata = (void *)pktopt->ip6po_nexthop; 2283 optdatalen = pktopt->ip6po_nexthop->sa_len; 2284 } 2285 break; 2286 case IPV6_USE_MIN_MTU: 2287 if (pktopt) 2288 optdata = (void *)&pktopt->ip6po_minmtu; 2289 else 2290 optdata = (void *)&defminmtu; 2291 optdatalen = sizeof(int); 2292 break; 2293 case IPV6_DONTFRAG: 2294 if (pktopt && ((pktopt->ip6po_flags) & IP6PO_DONTFRAG)) 2295 on = 1; 2296 else 2297 on = 0; 2298 optdata = (void *)&on; 2299 optdatalen = sizeof(on); 2300 break; 2301 default: /* should not happen */ 2302 #ifdef DIAGNOSTIC 2303 panic("ip6_getpcbopt: unexpected option\n"); 2304 #endif 2305 return (ENOPROTOOPT); 2306 } 2307 2308 if (optdatalen > MCLBYTES) 2309 return (EMSGSIZE); /* XXX */ 2310 *mp = m = m_get(M_WAIT, MT_SOOPTS); 2311 if (optdatalen > MLEN) 2312 MCLGET(m, M_WAIT); 2313 m->m_len = optdatalen; 2314 if (optdatalen) 2315 memcpy(mtod(m, void *), optdata, optdatalen); 2316 2317 return (error); 2318 } 2319 2320 void 2321 ip6_clearpktopts(struct ip6_pktopts *pktopt, int optname) 2322 { 2323 if (optname == -1 || optname == IPV6_PKTINFO) { 2324 if (pktopt->ip6po_pktinfo) 2325 free(pktopt->ip6po_pktinfo, M_IP6OPT); 2326 pktopt->ip6po_pktinfo = NULL; 2327 } 2328 if (optname == -1 || optname == IPV6_HOPLIMIT) 2329 pktopt->ip6po_hlim = -1; 2330 if (optname == -1 || optname == IPV6_TCLASS) 2331 pktopt->ip6po_tclass = -1; 2332 if (optname == -1 || optname == IPV6_NEXTHOP) { 2333 rtcache_free((struct route *)&pktopt->ip6po_nextroute); 2334 if (pktopt->ip6po_nexthop) 2335 free(pktopt->ip6po_nexthop, M_IP6OPT); 2336 pktopt->ip6po_nexthop = NULL; 2337 } 2338 if (optname == -1 || optname == IPV6_HOPOPTS) { 2339 if (pktopt->ip6po_hbh) 2340 free(pktopt->ip6po_hbh, M_IP6OPT); 2341 pktopt->ip6po_hbh = NULL; 2342 } 2343 if (optname == -1 || optname == IPV6_RTHDRDSTOPTS) { 2344 if (pktopt->ip6po_dest1) 2345 free(pktopt->ip6po_dest1, M_IP6OPT); 2346 pktopt->ip6po_dest1 = NULL; 2347 } 2348 if (optname == -1 || optname == IPV6_RTHDR) { 2349 if (pktopt->ip6po_rhinfo.ip6po_rhi_rthdr) 2350 free(pktopt->ip6po_rhinfo.ip6po_rhi_rthdr, M_IP6OPT); 2351 pktopt->ip6po_rhinfo.ip6po_rhi_rthdr = NULL; 2352 rtcache_free((struct route *)&pktopt->ip6po_route); 2353 } 2354 if (optname == -1 || optname == IPV6_DSTOPTS) { 2355 if (pktopt->ip6po_dest2) 2356 free(pktopt->ip6po_dest2, M_IP6OPT); 2357 pktopt->ip6po_dest2 = NULL; 2358 } 2359 } 2360 2361 #define PKTOPT_EXTHDRCPY(type) \ 2362 do { \ 2363 if (src->type) { \ 2364 int hlen = (((struct ip6_ext *)src->type)->ip6e_len + 1) << 3;\ 2365 dst->type = malloc(hlen, M_IP6OPT, canwait); \ 2366 if (dst->type == NULL && canwait == M_NOWAIT) \ 2367 goto bad; \ 2368 memcpy(dst->type, src->type, hlen); \ 2369 } \ 2370 } while (/*CONSTCOND*/ 0) 2371 2372 static int 2373 copypktopts(struct ip6_pktopts *dst, struct ip6_pktopts *src, int canwait) 2374 { 2375 dst->ip6po_hlim = src->ip6po_hlim; 2376 dst->ip6po_tclass = src->ip6po_tclass; 2377 dst->ip6po_flags = src->ip6po_flags; 2378 if (src->ip6po_pktinfo) { 2379 dst->ip6po_pktinfo = malloc(sizeof(*dst->ip6po_pktinfo), 2380 M_IP6OPT, canwait); 2381 if (dst->ip6po_pktinfo == NULL && canwait == M_NOWAIT) 2382 goto bad; 2383 *dst->ip6po_pktinfo = *src->ip6po_pktinfo; 2384 } 2385 if (src->ip6po_nexthop) { 2386 dst->ip6po_nexthop = malloc(src->ip6po_nexthop->sa_len, 2387 M_IP6OPT, canwait); 2388 if (dst->ip6po_nexthop == NULL && canwait == M_NOWAIT) 2389 goto bad; 2390 memcpy(dst->ip6po_nexthop, src->ip6po_nexthop, 2391 src->ip6po_nexthop->sa_len); 2392 } 2393 PKTOPT_EXTHDRCPY(ip6po_hbh); 2394 PKTOPT_EXTHDRCPY(ip6po_dest1); 2395 PKTOPT_EXTHDRCPY(ip6po_dest2); 2396 PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */ 2397 return (0); 2398 2399 bad: 2400 if (dst->ip6po_pktinfo) free(dst->ip6po_pktinfo, M_IP6OPT); 2401 if (dst->ip6po_nexthop) free(dst->ip6po_nexthop, M_IP6OPT); 2402 if (dst->ip6po_hbh) free(dst->ip6po_hbh, M_IP6OPT); 2403 if (dst->ip6po_dest1) free(dst->ip6po_dest1, M_IP6OPT); 2404 if (dst->ip6po_dest2) free(dst->ip6po_dest2, M_IP6OPT); 2405 if (dst->ip6po_rthdr) free(dst->ip6po_rthdr, M_IP6OPT); 2406 2407 return (ENOBUFS); 2408 } 2409 #undef PKTOPT_EXTHDRCPY 2410 2411 struct ip6_pktopts * 2412 ip6_copypktopts(struct ip6_pktopts *src, int canwait) 2413 { 2414 int error; 2415 struct ip6_pktopts *dst; 2416 2417 dst = malloc(sizeof(*dst), M_IP6OPT, canwait); 2418 if (dst == NULL && canwait == M_NOWAIT) 2419 return (NULL); 2420 ip6_initpktopts(dst); 2421 2422 if ((error = copypktopts(dst, src, canwait)) != 0) { 2423 free(dst, M_IP6OPT); 2424 return (NULL); 2425 } 2426 2427 return (dst); 2428 } 2429 2430 void 2431 ip6_freepcbopts(struct ip6_pktopts *pktopt) 2432 { 2433 if (pktopt == NULL) 2434 return; 2435 2436 ip6_clearpktopts(pktopt, -1); 2437 2438 free(pktopt, M_IP6OPT); 2439 } 2440 2441 /* 2442 * Set the IP6 multicast options in response to user setsockopt(). 2443 */ 2444 static int 2445 ip6_setmoptions(optname, im6op, m) 2446 int optname; 2447 struct ip6_moptions **im6op; 2448 struct mbuf *m; 2449 { 2450 int error = 0; 2451 u_int loop, ifindex; 2452 struct ipv6_mreq *mreq; 2453 struct ifnet *ifp; 2454 struct ip6_moptions *im6o = *im6op; 2455 struct route_in6 ro; 2456 struct in6_multi_mship *imm; 2457 struct lwp *l = curlwp; /* XXX */ 2458 2459 if (im6o == NULL) { 2460 /* 2461 * No multicast option buffer attached to the pcb; 2462 * allocate one and initialize to default values. 2463 */ 2464 im6o = (struct ip6_moptions *) 2465 malloc(sizeof(*im6o), M_IPMOPTS, M_WAITOK); 2466 2467 if (im6o == NULL) 2468 return (ENOBUFS); 2469 *im6op = im6o; 2470 im6o->im6o_multicast_ifp = NULL; 2471 im6o->im6o_multicast_hlim = ip6_defmcasthlim; 2472 im6o->im6o_multicast_loop = IPV6_DEFAULT_MULTICAST_LOOP; 2473 LIST_INIT(&im6o->im6o_memberships); 2474 } 2475 2476 switch (optname) { 2477 2478 case IPV6_MULTICAST_IF: 2479 /* 2480 * Select the interface for outgoing multicast packets. 2481 */ 2482 if (m == NULL || m->m_len != sizeof(u_int)) { 2483 error = EINVAL; 2484 break; 2485 } 2486 bcopy(mtod(m, u_int *), &ifindex, sizeof(ifindex)); 2487 if (ifindex != 0) { 2488 if (if_indexlim <= ifindex || !ifindex2ifnet[ifindex]) { 2489 error = ENXIO; /* XXX EINVAL? */ 2490 break; 2491 } 2492 ifp = ifindex2ifnet[ifindex]; 2493 if ((ifp->if_flags & IFF_MULTICAST) == 0) { 2494 error = EADDRNOTAVAIL; 2495 break; 2496 } 2497 } else 2498 ifp = NULL; 2499 im6o->im6o_multicast_ifp = ifp; 2500 break; 2501 2502 case IPV6_MULTICAST_HOPS: 2503 { 2504 /* 2505 * Set the IP6 hoplimit for outgoing multicast packets. 2506 */ 2507 int optval; 2508 if (m == NULL || m->m_len != sizeof(int)) { 2509 error = EINVAL; 2510 break; 2511 } 2512 bcopy(mtod(m, u_int *), &optval, sizeof(optval)); 2513 if (optval < -1 || optval >= 256) 2514 error = EINVAL; 2515 else if (optval == -1) 2516 im6o->im6o_multicast_hlim = ip6_defmcasthlim; 2517 else 2518 im6o->im6o_multicast_hlim = optval; 2519 break; 2520 } 2521 2522 case IPV6_MULTICAST_LOOP: 2523 /* 2524 * Set the loopback flag for outgoing multicast packets. 2525 * Must be zero or one. 2526 */ 2527 if (m == NULL || m->m_len != sizeof(u_int)) { 2528 error = EINVAL; 2529 break; 2530 } 2531 bcopy(mtod(m, u_int *), &loop, sizeof(loop)); 2532 if (loop > 1) { 2533 error = EINVAL; 2534 break; 2535 } 2536 im6o->im6o_multicast_loop = loop; 2537 break; 2538 2539 case IPV6_JOIN_GROUP: 2540 /* 2541 * Add a multicast group membership. 2542 * Group must be a valid IP6 multicast address. 2543 */ 2544 if (m == NULL || m->m_len != sizeof(struct ipv6_mreq)) { 2545 error = EINVAL; 2546 break; 2547 } 2548 mreq = mtod(m, struct ipv6_mreq *); 2549 if (IN6_IS_ADDR_UNSPECIFIED(&mreq->ipv6mr_multiaddr)) { 2550 /* 2551 * We use the unspecified address to specify to accept 2552 * all multicast addresses. Only super user is allowed 2553 * to do this. 2554 */ 2555 if (kauth_authorize_generic(l->l_cred, 2556 KAUTH_GENERIC_ISSUSER, &l->l_acflag)) 2557 { 2558 error = EACCES; 2559 break; 2560 } 2561 } else if (!IN6_IS_ADDR_MULTICAST(&mreq->ipv6mr_multiaddr)) { 2562 error = EINVAL; 2563 break; 2564 } 2565 2566 /* 2567 * If no interface was explicitly specified, choose an 2568 * appropriate one according to the given multicast address. 2569 */ 2570 if (mreq->ipv6mr_interface == 0) { 2571 struct sockaddr_in6 *dst; 2572 2573 /* 2574 * Look up the routing table for the 2575 * address, and choose the outgoing interface. 2576 * XXX: is it a good approach? 2577 */ 2578 bzero(&ro, sizeof(ro)); 2579 dst = (struct sockaddr_in6 *)&ro.ro_dst; 2580 dst->sin6_family = AF_INET6; 2581 dst->sin6_len = sizeof(*dst); 2582 dst->sin6_addr = mreq->ipv6mr_multiaddr; 2583 rtcache_init((struct route *)&ro); 2584 if (ro.ro_rt == NULL) { 2585 error = EADDRNOTAVAIL; 2586 break; 2587 } 2588 ifp = ro.ro_rt->rt_ifp; 2589 rtcache_free((struct route *)&ro); 2590 } else { 2591 /* 2592 * If the interface is specified, validate it. 2593 */ 2594 if (if_indexlim <= mreq->ipv6mr_interface || 2595 !ifindex2ifnet[mreq->ipv6mr_interface]) { 2596 error = ENXIO; /* XXX EINVAL? */ 2597 break; 2598 } 2599 ifp = ifindex2ifnet[mreq->ipv6mr_interface]; 2600 } 2601 2602 /* 2603 * See if we found an interface, and confirm that it 2604 * supports multicast 2605 */ 2606 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) { 2607 error = EADDRNOTAVAIL; 2608 break; 2609 } 2610 2611 if (in6_setscope(&mreq->ipv6mr_multiaddr, ifp, NULL)) { 2612 error = EADDRNOTAVAIL; /* XXX: should not happen */ 2613 break; 2614 } 2615 2616 /* 2617 * See if the membership already exists. 2618 */ 2619 for (imm = im6o->im6o_memberships.lh_first; 2620 imm != NULL; imm = imm->i6mm_chain.le_next) 2621 if (imm->i6mm_maddr->in6m_ifp == ifp && 2622 IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr, 2623 &mreq->ipv6mr_multiaddr)) 2624 break; 2625 if (imm != NULL) { 2626 error = EADDRINUSE; 2627 break; 2628 } 2629 /* 2630 * Everything looks good; add a new record to the multicast 2631 * address list for the given interface. 2632 */ 2633 imm = in6_joingroup(ifp, &mreq->ipv6mr_multiaddr, &error, 0); 2634 if (imm == NULL) 2635 break; 2636 LIST_INSERT_HEAD(&im6o->im6o_memberships, imm, i6mm_chain); 2637 break; 2638 2639 case IPV6_LEAVE_GROUP: 2640 /* 2641 * Drop a multicast group membership. 2642 * Group must be a valid IP6 multicast address. 2643 */ 2644 if (m == NULL || m->m_len != sizeof(struct ipv6_mreq)) { 2645 error = EINVAL; 2646 break; 2647 } 2648 mreq = mtod(m, struct ipv6_mreq *); 2649 2650 /* 2651 * If an interface address was specified, get a pointer 2652 * to its ifnet structure. 2653 */ 2654 if (mreq->ipv6mr_interface != 0) { 2655 if (if_indexlim <= mreq->ipv6mr_interface || 2656 !ifindex2ifnet[mreq->ipv6mr_interface]) { 2657 error = ENXIO; /* XXX EINVAL? */ 2658 break; 2659 } 2660 ifp = ifindex2ifnet[mreq->ipv6mr_interface]; 2661 } else 2662 ifp = NULL; 2663 2664 /* Fill in the scope zone ID */ 2665 if (ifp) { 2666 if (in6_setscope(&mreq->ipv6mr_multiaddr, ifp, NULL)) { 2667 /* XXX: should not happen */ 2668 error = EADDRNOTAVAIL; 2669 break; 2670 } 2671 } else if (mreq->ipv6mr_interface != 0) { 2672 /* 2673 * XXX: This case would happens when the (positive) 2674 * index is in the valid range, but the corresponding 2675 * interface has been detached dynamically. The above 2676 * check probably avoids such case to happen here, but 2677 * we check it explicitly for safety. 2678 */ 2679 error = EADDRNOTAVAIL; 2680 break; 2681 } else { /* ipv6mr_interface == 0 */ 2682 struct sockaddr_in6 sa6_mc; 2683 2684 /* 2685 * The API spec says as follows: 2686 * If the interface index is specified as 0, the 2687 * system may choose a multicast group membership to 2688 * drop by matching the multicast address only. 2689 * On the other hand, we cannot disambiguate the scope 2690 * zone unless an interface is provided. Thus, we 2691 * check if there's ambiguity with the default scope 2692 * zone as the last resort. 2693 */ 2694 bzero(&sa6_mc, sizeof(sa6_mc)); 2695 sa6_mc.sin6_family = AF_INET6; 2696 sa6_mc.sin6_len = sizeof(sa6_mc); 2697 sa6_mc.sin6_addr = mreq->ipv6mr_multiaddr; 2698 error = sa6_embedscope(&sa6_mc, ip6_use_defzone); 2699 if (error != 0) 2700 break; 2701 mreq->ipv6mr_multiaddr = sa6_mc.sin6_addr; 2702 } 2703 2704 /* 2705 * Find the membership in the membership list. 2706 */ 2707 for (imm = im6o->im6o_memberships.lh_first; 2708 imm != NULL; imm = imm->i6mm_chain.le_next) { 2709 if ((ifp == NULL || imm->i6mm_maddr->in6m_ifp == ifp) && 2710 IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr, 2711 &mreq->ipv6mr_multiaddr)) 2712 break; 2713 } 2714 if (imm == NULL) { 2715 /* Unable to resolve interface */ 2716 error = EADDRNOTAVAIL; 2717 break; 2718 } 2719 /* 2720 * Give up the multicast address record to which the 2721 * membership points. 2722 */ 2723 LIST_REMOVE(imm, i6mm_chain); 2724 in6_leavegroup(imm); 2725 break; 2726 2727 default: 2728 error = EOPNOTSUPP; 2729 break; 2730 } 2731 2732 /* 2733 * If all options have default values, no need to keep the mbuf. 2734 */ 2735 if (im6o->im6o_multicast_ifp == NULL && 2736 im6o->im6o_multicast_hlim == ip6_defmcasthlim && 2737 im6o->im6o_multicast_loop == IPV6_DEFAULT_MULTICAST_LOOP && 2738 im6o->im6o_memberships.lh_first == NULL) { 2739 free(*im6op, M_IPMOPTS); 2740 *im6op = NULL; 2741 } 2742 2743 return (error); 2744 } 2745 2746 /* 2747 * Return the IP6 multicast options in response to user getsockopt(). 2748 */ 2749 static int 2750 ip6_getmoptions(optname, im6o, mp) 2751 int optname; 2752 struct ip6_moptions *im6o; 2753 struct mbuf **mp; 2754 { 2755 u_int *hlim, *loop, *ifindex; 2756 2757 *mp = m_get(M_WAIT, MT_SOOPTS); 2758 2759 switch (optname) { 2760 2761 case IPV6_MULTICAST_IF: 2762 ifindex = mtod(*mp, u_int *); 2763 (*mp)->m_len = sizeof(u_int); 2764 if (im6o == NULL || im6o->im6o_multicast_ifp == NULL) 2765 *ifindex = 0; 2766 else 2767 *ifindex = im6o->im6o_multicast_ifp->if_index; 2768 return (0); 2769 2770 case IPV6_MULTICAST_HOPS: 2771 hlim = mtod(*mp, u_int *); 2772 (*mp)->m_len = sizeof(u_int); 2773 if (im6o == NULL) 2774 *hlim = ip6_defmcasthlim; 2775 else 2776 *hlim = im6o->im6o_multicast_hlim; 2777 return (0); 2778 2779 case IPV6_MULTICAST_LOOP: 2780 loop = mtod(*mp, u_int *); 2781 (*mp)->m_len = sizeof(u_int); 2782 if (im6o == NULL) 2783 *loop = ip6_defmcasthlim; 2784 else 2785 *loop = im6o->im6o_multicast_loop; 2786 return (0); 2787 2788 default: 2789 return (EOPNOTSUPP); 2790 } 2791 } 2792 2793 /* 2794 * Discard the IP6 multicast options. 2795 */ 2796 void 2797 ip6_freemoptions(im6o) 2798 struct ip6_moptions *im6o; 2799 { 2800 struct in6_multi_mship *imm; 2801 2802 if (im6o == NULL) 2803 return; 2804 2805 while ((imm = im6o->im6o_memberships.lh_first) != NULL) { 2806 LIST_REMOVE(imm, i6mm_chain); 2807 in6_leavegroup(imm); 2808 } 2809 free(im6o, M_IPMOPTS); 2810 } 2811 2812 /* 2813 * Set IPv6 outgoing packet options based on advanced API. 2814 */ 2815 int 2816 ip6_setpktopts(control, opt, stickyopt, priv, uproto) 2817 struct mbuf *control; 2818 struct ip6_pktopts *opt, *stickyopt; 2819 int priv, uproto; 2820 { 2821 struct cmsghdr *cm = 0; 2822 2823 if (control == NULL || opt == NULL) 2824 return (EINVAL); 2825 2826 ip6_initpktopts(opt); 2827 if (stickyopt) { 2828 int error; 2829 2830 /* 2831 * If stickyopt is provided, make a local copy of the options 2832 * for this particular packet, then override them by ancillary 2833 * objects. 2834 * XXX: copypktopts() does not copy the cached route to a next 2835 * hop (if any). This is not very good in terms of efficiency, 2836 * but we can allow this since this option should be rarely 2837 * used. 2838 */ 2839 if ((error = copypktopts(opt, stickyopt, M_NOWAIT)) != 0) 2840 return (error); 2841 } 2842 2843 /* 2844 * XXX: Currently, we assume all the optional information is stored 2845 * in a single mbuf. 2846 */ 2847 if (control->m_next) 2848 return (EINVAL); 2849 2850 for (; control->m_len; control->m_data += CMSG_ALIGN(cm->cmsg_len), 2851 control->m_len -= CMSG_ALIGN(cm->cmsg_len)) { 2852 int error; 2853 2854 if (control->m_len < CMSG_LEN(0)) 2855 return (EINVAL); 2856 2857 cm = mtod(control, struct cmsghdr *); 2858 if (cm->cmsg_len == 0 || cm->cmsg_len > control->m_len) 2859 return (EINVAL); 2860 if (cm->cmsg_level != IPPROTO_IPV6) 2861 continue; 2862 2863 error = ip6_setpktopt(cm->cmsg_type, CMSG_DATA(cm), 2864 cm->cmsg_len - CMSG_LEN(0), opt, priv, 0, 1, uproto); 2865 if (error) 2866 return (error); 2867 } 2868 2869 return (0); 2870 } 2871 2872 /* 2873 * Set a particular packet option, as a sticky option or an ancillary data 2874 * item. "len" can be 0 only when it's a sticky option. 2875 * We have 4 cases of combination of "sticky" and "cmsg": 2876 * "sticky=0, cmsg=0": impossible 2877 * "sticky=0, cmsg=1": RFC2292 or RFC3542 ancillary data 2878 * "sticky=1, cmsg=0": RFC3542 socket option 2879 * "sticky=1, cmsg=1": RFC2292 socket option 2880 */ 2881 static int 2882 ip6_setpktopt(int optname, u_char *buf, int len, struct ip6_pktopts *opt, 2883 int priv, int sticky, int cmsg, int uproto) 2884 { 2885 int minmtupolicy; 2886 2887 if (!sticky && !cmsg) { 2888 #ifdef DIAGNOSTIC 2889 printf("ip6_setpktopt: impossible case\n"); 2890 #endif 2891 return (EINVAL); 2892 } 2893 2894 /* 2895 * IPV6_2292xxx is for backward compatibility to RFC2292, and should 2896 * not be specified in the context of RFC3542. Conversely, 2897 * RFC3542 types should not be specified in the context of RFC2292. 2898 */ 2899 if (!cmsg) { 2900 switch (optname) { 2901 case IPV6_2292PKTINFO: 2902 case IPV6_2292HOPLIMIT: 2903 case IPV6_2292NEXTHOP: 2904 case IPV6_2292HOPOPTS: 2905 case IPV6_2292DSTOPTS: 2906 case IPV6_2292RTHDR: 2907 case IPV6_2292PKTOPTIONS: 2908 return (ENOPROTOOPT); 2909 } 2910 } 2911 if (sticky && cmsg) { 2912 switch (optname) { 2913 case IPV6_PKTINFO: 2914 case IPV6_HOPLIMIT: 2915 case IPV6_NEXTHOP: 2916 case IPV6_HOPOPTS: 2917 case IPV6_DSTOPTS: 2918 case IPV6_RTHDRDSTOPTS: 2919 case IPV6_RTHDR: 2920 case IPV6_USE_MIN_MTU: 2921 case IPV6_DONTFRAG: 2922 case IPV6_OTCLASS: 2923 case IPV6_TCLASS: 2924 return (ENOPROTOOPT); 2925 } 2926 } 2927 2928 switch (optname) { 2929 #ifdef RFC2292 2930 case IPV6_2292PKTINFO: 2931 #endif 2932 case IPV6_PKTINFO: 2933 { 2934 struct ifnet *ifp = NULL; 2935 struct in6_pktinfo *pktinfo; 2936 2937 if (len != sizeof(struct in6_pktinfo)) 2938 return (EINVAL); 2939 2940 pktinfo = (struct in6_pktinfo *)buf; 2941 2942 /* 2943 * An application can clear any sticky IPV6_PKTINFO option by 2944 * doing a "regular" setsockopt with ipi6_addr being 2945 * in6addr_any and ipi6_ifindex being zero. 2946 * [RFC 3542, Section 6] 2947 */ 2948 if (optname == IPV6_PKTINFO && opt->ip6po_pktinfo && 2949 pktinfo->ipi6_ifindex == 0 && 2950 IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) { 2951 ip6_clearpktopts(opt, optname); 2952 break; 2953 } 2954 2955 if (uproto == IPPROTO_TCP && optname == IPV6_PKTINFO && 2956 sticky && !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) { 2957 return (EINVAL); 2958 } 2959 2960 /* validate the interface index if specified. */ 2961 if (pktinfo->ipi6_ifindex >= if_indexlim) { 2962 return (ENXIO); 2963 } 2964 if (pktinfo->ipi6_ifindex) { 2965 ifp = ifindex2ifnet[pktinfo->ipi6_ifindex]; 2966 if (ifp == NULL) 2967 return (ENXIO); 2968 } 2969 2970 /* 2971 * We store the address anyway, and let in6_selectsrc() 2972 * validate the specified address. This is because ipi6_addr 2973 * may not have enough information about its scope zone, and 2974 * we may need additional information (such as outgoing 2975 * interface or the scope zone of a destination address) to 2976 * disambiguate the scope. 2977 * XXX: the delay of the validation may confuse the 2978 * application when it is used as a sticky option. 2979 */ 2980 if (opt->ip6po_pktinfo == NULL) { 2981 opt->ip6po_pktinfo = malloc(sizeof(*pktinfo), 2982 M_IP6OPT, M_NOWAIT); 2983 if (opt->ip6po_pktinfo == NULL) 2984 return (ENOBUFS); 2985 } 2986 memcpy(opt->ip6po_pktinfo, pktinfo, sizeof(*pktinfo)); 2987 break; 2988 } 2989 2990 #ifdef RFC2292 2991 case IPV6_2292HOPLIMIT: 2992 #endif 2993 case IPV6_HOPLIMIT: 2994 { 2995 int *hlimp; 2996 2997 /* 2998 * RFC 3542 deprecated the usage of sticky IPV6_HOPLIMIT 2999 * to simplify the ordering among hoplimit options. 3000 */ 3001 if (optname == IPV6_HOPLIMIT && sticky) 3002 return (ENOPROTOOPT); 3003 3004 if (len != sizeof(int)) 3005 return (EINVAL); 3006 hlimp = (int *)buf; 3007 if (*hlimp < -1 || *hlimp > 255) 3008 return (EINVAL); 3009 3010 opt->ip6po_hlim = *hlimp; 3011 break; 3012 } 3013 3014 case IPV6_OTCLASS: 3015 if (len != sizeof(u_int8_t)) 3016 return (EINVAL); 3017 3018 opt->ip6po_tclass = *(u_int8_t *)buf; 3019 break; 3020 3021 case IPV6_TCLASS: 3022 { 3023 int tclass; 3024 3025 if (len != sizeof(int)) 3026 return (EINVAL); 3027 tclass = *(int *)buf; 3028 if (tclass < -1 || tclass > 255) 3029 return (EINVAL); 3030 3031 opt->ip6po_tclass = tclass; 3032 break; 3033 } 3034 3035 #ifdef RFC2292 3036 case IPV6_2292NEXTHOP: 3037 #endif 3038 case IPV6_NEXTHOP: 3039 if (!priv) 3040 return (EPERM); 3041 3042 if (len == 0) { /* just remove the option */ 3043 ip6_clearpktopts(opt, IPV6_NEXTHOP); 3044 break; 3045 } 3046 3047 /* check if cmsg_len is large enough for sa_len */ 3048 if (len < sizeof(struct sockaddr) || len < *buf) 3049 return (EINVAL); 3050 3051 switch (((struct sockaddr *)buf)->sa_family) { 3052 case AF_INET6: 3053 { 3054 struct sockaddr_in6 *sa6 = (struct sockaddr_in6 *)buf; 3055 int error; 3056 3057 if (sa6->sin6_len != sizeof(struct sockaddr_in6)) 3058 return (EINVAL); 3059 3060 if (IN6_IS_ADDR_UNSPECIFIED(&sa6->sin6_addr) || 3061 IN6_IS_ADDR_MULTICAST(&sa6->sin6_addr)) { 3062 return (EINVAL); 3063 } 3064 if ((error = sa6_embedscope(sa6, ip6_use_defzone)) 3065 != 0) { 3066 return (error); 3067 } 3068 break; 3069 } 3070 case AF_LINK: /* eventually be supported? */ 3071 default: 3072 return (EAFNOSUPPORT); 3073 } 3074 3075 /* turn off the previous option, then set the new option. */ 3076 ip6_clearpktopts(opt, IPV6_NEXTHOP); 3077 opt->ip6po_nexthop = malloc(*buf, M_IP6OPT, M_NOWAIT); 3078 if (opt->ip6po_nexthop == NULL) 3079 return (ENOBUFS); 3080 memcpy(opt->ip6po_nexthop, buf, *buf); 3081 break; 3082 3083 #ifdef RFC2292 3084 case IPV6_2292HOPOPTS: 3085 #endif 3086 case IPV6_HOPOPTS: 3087 { 3088 struct ip6_hbh *hbh; 3089 int hbhlen; 3090 3091 /* 3092 * XXX: We don't allow a non-privileged user to set ANY HbH 3093 * options, since per-option restriction has too much 3094 * overhead. 3095 */ 3096 if (!priv) 3097 return (EPERM); 3098 3099 if (len == 0) { 3100 ip6_clearpktopts(opt, IPV6_HOPOPTS); 3101 break; /* just remove the option */ 3102 } 3103 3104 /* message length validation */ 3105 if (len < sizeof(struct ip6_hbh)) 3106 return (EINVAL); 3107 hbh = (struct ip6_hbh *)buf; 3108 hbhlen = (hbh->ip6h_len + 1) << 3; 3109 if (len != hbhlen) 3110 return (EINVAL); 3111 3112 /* turn off the previous option, then set the new option. */ 3113 ip6_clearpktopts(opt, IPV6_HOPOPTS); 3114 opt->ip6po_hbh = malloc(hbhlen, M_IP6OPT, M_NOWAIT); 3115 if (opt->ip6po_hbh == NULL) 3116 return (ENOBUFS); 3117 memcpy(opt->ip6po_hbh, hbh, hbhlen); 3118 3119 break; 3120 } 3121 3122 #ifdef RFC2292 3123 case IPV6_2292DSTOPTS: 3124 #endif 3125 case IPV6_DSTOPTS: 3126 case IPV6_RTHDRDSTOPTS: 3127 { 3128 struct ip6_dest *dest, **newdest = NULL; 3129 int destlen; 3130 3131 if (!priv) /* XXX: see the comment for IPV6_HOPOPTS */ 3132 return (EPERM); 3133 3134 if (len == 0) { 3135 ip6_clearpktopts(opt, optname); 3136 break; /* just remove the option */ 3137 } 3138 3139 /* message length validation */ 3140 if (len < sizeof(struct ip6_dest)) 3141 return (EINVAL); 3142 dest = (struct ip6_dest *)buf; 3143 destlen = (dest->ip6d_len + 1) << 3; 3144 if (len != destlen) 3145 return (EINVAL); 3146 /* 3147 * Determine the position that the destination options header 3148 * should be inserted; before or after the routing header. 3149 */ 3150 switch (optname) { 3151 case IPV6_2292DSTOPTS: 3152 /* 3153 * The old advanced API is ambiguous on this point. 3154 * Our approach is to determine the position based 3155 * according to the existence of a routing header. 3156 * Note, however, that this depends on the order of the 3157 * extension headers in the ancillary data; the 1st 3158 * part of the destination options header must appear 3159 * before the routing header in the ancillary data, 3160 * too. 3161 * RFC3542 solved the ambiguity by introducing 3162 * separate ancillary data or option types. 3163 */ 3164 if (opt->ip6po_rthdr == NULL) 3165 newdest = &opt->ip6po_dest1; 3166 else 3167 newdest = &opt->ip6po_dest2; 3168 break; 3169 case IPV6_RTHDRDSTOPTS: 3170 newdest = &opt->ip6po_dest1; 3171 break; 3172 case IPV6_DSTOPTS: 3173 newdest = &opt->ip6po_dest2; 3174 break; 3175 } 3176 3177 /* turn off the previous option, then set the new option. */ 3178 ip6_clearpktopts(opt, optname); 3179 *newdest = malloc(destlen, M_IP6OPT, M_NOWAIT); 3180 if (*newdest == NULL) 3181 return (ENOBUFS); 3182 memcpy(*newdest, dest, destlen); 3183 3184 break; 3185 } 3186 3187 #ifdef RFC2292 3188 case IPV6_2292RTHDR: 3189 #endif 3190 case IPV6_RTHDR: 3191 { 3192 struct ip6_rthdr *rth; 3193 int rthlen; 3194 3195 if (len == 0) { 3196 ip6_clearpktopts(opt, IPV6_RTHDR); 3197 break; /* just remove the option */ 3198 } 3199 3200 /* message length validation */ 3201 if (len < sizeof(struct ip6_rthdr)) 3202 return (EINVAL); 3203 rth = (struct ip6_rthdr *)buf; 3204 rthlen = (rth->ip6r_len + 1) << 3; 3205 if (len != rthlen) 3206 return (EINVAL); 3207 switch (rth->ip6r_type) { 3208 case IPV6_RTHDR_TYPE_0: 3209 if (rth->ip6r_len == 0) /* must contain one addr */ 3210 return (EINVAL); 3211 if (rth->ip6r_len % 2) /* length must be even */ 3212 return (EINVAL); 3213 if (rth->ip6r_len / 2 != rth->ip6r_segleft) 3214 return (EINVAL); 3215 break; 3216 default: 3217 return (EINVAL); /* not supported */ 3218 } 3219 /* turn off the previous option */ 3220 ip6_clearpktopts(opt, IPV6_RTHDR); 3221 opt->ip6po_rthdr = malloc(rthlen, M_IP6OPT, M_NOWAIT); 3222 if (opt->ip6po_rthdr == NULL) 3223 return (ENOBUFS); 3224 memcpy(opt->ip6po_rthdr, rth, rthlen); 3225 break; 3226 } 3227 3228 case IPV6_USE_MIN_MTU: 3229 if (len != sizeof(int)) 3230 return (EINVAL); 3231 minmtupolicy = *(int *)buf; 3232 if (minmtupolicy != IP6PO_MINMTU_MCASTONLY && 3233 minmtupolicy != IP6PO_MINMTU_DISABLE && 3234 minmtupolicy != IP6PO_MINMTU_ALL) { 3235 return (EINVAL); 3236 } 3237 opt->ip6po_minmtu = minmtupolicy; 3238 break; 3239 3240 case IPV6_DONTFRAG: 3241 if (len != sizeof(int)) 3242 return (EINVAL); 3243 3244 if (uproto == IPPROTO_TCP || *(int *)buf == 0) { 3245 /* 3246 * we ignore this option for TCP sockets. 3247 * (RFC3542 leaves this case unspecified.) 3248 */ 3249 opt->ip6po_flags &= ~IP6PO_DONTFRAG; 3250 } else 3251 opt->ip6po_flags |= IP6PO_DONTFRAG; 3252 break; 3253 3254 default: 3255 return (ENOPROTOOPT); 3256 } /* end of switch */ 3257 3258 return (0); 3259 } 3260 3261 /* 3262 * Routine called from ip6_output() to loop back a copy of an IP6 multicast 3263 * packet to the input queue of a specified interface. Note that this 3264 * calls the output routine of the loopback "driver", but with an interface 3265 * pointer that might NOT be lo0ifp -- easier than replicating that code here. 3266 */ 3267 void 3268 ip6_mloopback(ifp, m, dst) 3269 struct ifnet *ifp; 3270 struct mbuf *m; 3271 struct sockaddr_in6 *dst; 3272 { 3273 struct mbuf *copym; 3274 struct ip6_hdr *ip6; 3275 3276 copym = m_copy(m, 0, M_COPYALL); 3277 if (copym == NULL) 3278 return; 3279 3280 /* 3281 * Make sure to deep-copy IPv6 header portion in case the data 3282 * is in an mbuf cluster, so that we can safely override the IPv6 3283 * header portion later. 3284 */ 3285 if ((copym->m_flags & M_EXT) != 0 || 3286 copym->m_len < sizeof(struct ip6_hdr)) { 3287 copym = m_pullup(copym, sizeof(struct ip6_hdr)); 3288 if (copym == NULL) 3289 return; 3290 } 3291 3292 #ifdef DIAGNOSTIC 3293 if (copym->m_len < sizeof(*ip6)) { 3294 m_freem(copym); 3295 return; 3296 } 3297 #endif 3298 3299 ip6 = mtod(copym, struct ip6_hdr *); 3300 /* 3301 * clear embedded scope identifiers if necessary. 3302 * in6_clearscope will touch the addresses only when necessary. 3303 */ 3304 in6_clearscope(&ip6->ip6_src); 3305 in6_clearscope(&ip6->ip6_dst); 3306 3307 (void)looutput(ifp, copym, (struct sockaddr *)dst, NULL); 3308 } 3309 3310 /* 3311 * Chop IPv6 header off from the payload. 3312 */ 3313 static int 3314 ip6_splithdr(m, exthdrs) 3315 struct mbuf *m; 3316 struct ip6_exthdrs *exthdrs; 3317 { 3318 struct mbuf *mh; 3319 struct ip6_hdr *ip6; 3320 3321 ip6 = mtod(m, struct ip6_hdr *); 3322 if (m->m_len > sizeof(*ip6)) { 3323 MGETHDR(mh, M_DONTWAIT, MT_HEADER); 3324 if (mh == 0) { 3325 m_freem(m); 3326 return ENOBUFS; 3327 } 3328 M_MOVE_PKTHDR(mh, m); 3329 MH_ALIGN(mh, sizeof(*ip6)); 3330 m->m_len -= sizeof(*ip6); 3331 m->m_data += sizeof(*ip6); 3332 mh->m_next = m; 3333 m = mh; 3334 m->m_len = sizeof(*ip6); 3335 bcopy((caddr_t)ip6, mtod(m, caddr_t), sizeof(*ip6)); 3336 } 3337 exthdrs->ip6e_ip6 = m; 3338 return 0; 3339 } 3340 3341 /* 3342 * Compute IPv6 extension header length. 3343 */ 3344 int 3345 ip6_optlen(in6p) 3346 struct in6pcb *in6p; 3347 { 3348 int len; 3349 3350 if (!in6p->in6p_outputopts) 3351 return 0; 3352 3353 len = 0; 3354 #define elen(x) \ 3355 (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0) 3356 3357 len += elen(in6p->in6p_outputopts->ip6po_hbh); 3358 len += elen(in6p->in6p_outputopts->ip6po_dest1); 3359 len += elen(in6p->in6p_outputopts->ip6po_rthdr); 3360 len += elen(in6p->in6p_outputopts->ip6po_dest2); 3361 return len; 3362 #undef elen 3363 } 3364
Indexes created Tue Sep 23 07:09:52 GMT 2025