ip6_output.c revision 1.107
1 /* $NetBSD: ip6_output.c,v 1.107 2006/12/02 18:59:17 dyoung 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 1.107 2006/12/02 18:59:17 dyoung 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 if (ro == &ip6route && ro->ro_rt) { /* brace necessary for RTFREE */ 1141 RTFREE(ro->ro_rt); 1142 } else if (ro_pmtu == &ip6route && ro_pmtu->ro_rt) { 1143 RTFREE(ro_pmtu->ro_rt); 1144 } 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 (ro_pmtu->ro_rt != NULL && 1381 ((ro_pmtu->ro_rt->rt_flags & RTF_UP) == 0 || 1382 !IN6_ARE_ADDR_EQUAL(&sa6_dst->sin6_addr, dst))) { 1383 RTFREE(ro_pmtu->ro_rt); 1384 ro_pmtu->ro_rt = (struct rtentry *)NULL; 1385 } 1386 if (ro_pmtu->ro_rt == NULL) { 1387 bzero(sa6_dst, sizeof(*sa6_dst)); /* for safety */ 1388 sa6_dst->sin6_family = AF_INET6; 1389 sa6_dst->sin6_len = sizeof(struct sockaddr_in6); 1390 sa6_dst->sin6_addr = *dst; 1391 1392 rtalloc((struct route *)ro_pmtu); 1393 } 1394 } 1395 if (ro_pmtu->ro_rt != NULL) { 1396 u_int32_t ifmtu; 1397 1398 if (ifp == NULL) 1399 ifp = ro_pmtu->ro_rt->rt_ifp; 1400 ifmtu = IN6_LINKMTU(ifp); 1401 mtu = ro_pmtu->ro_rt->rt_rmx.rmx_mtu; 1402 if (mtu == 0) 1403 mtu = ifmtu; 1404 else if (mtu < IPV6_MMTU) { 1405 /* 1406 * RFC2460 section 5, last paragraph: 1407 * if we record ICMPv6 too big message with 1408 * mtu < IPV6_MMTU, transmit packets sized IPV6_MMTU 1409 * or smaller, with fragment header attached. 1410 * (fragment header is needed regardless from the 1411 * packet size, for translators to identify packets) 1412 */ 1413 alwaysfrag = 1; 1414 mtu = IPV6_MMTU; 1415 } else if (mtu > ifmtu) { 1416 /* 1417 * The MTU on the route is larger than the MTU on 1418 * the interface! This shouldn't happen, unless the 1419 * MTU of the interface has been changed after the 1420 * interface was brought up. Change the MTU in the 1421 * route to match the interface MTU (as long as the 1422 * field isn't locked). 1423 */ 1424 mtu = ifmtu; 1425 if (!(ro_pmtu->ro_rt->rt_rmx.rmx_locks & RTV_MTU)) 1426 ro_pmtu->ro_rt->rt_rmx.rmx_mtu = mtu; 1427 } 1428 } else if (ifp) { 1429 mtu = IN6_LINKMTU(ifp); 1430 } else 1431 error = EHOSTUNREACH; /* XXX */ 1432 1433 *mtup = mtu; 1434 if (alwaysfragp) 1435 *alwaysfragp = alwaysfrag; 1436 return (error); 1437 } 1438 1439 /* 1440 * IP6 socket option processing. 1441 */ 1442 int 1443 ip6_ctloutput(op, so, level, optname, mp) 1444 int op; 1445 struct socket *so; 1446 int level, optname; 1447 struct mbuf **mp; 1448 { 1449 int privileged, optdatalen, uproto; 1450 void *optdata; 1451 struct in6pcb *in6p = sotoin6pcb(so); 1452 struct mbuf *m = *mp; 1453 int error, optval; 1454 int optlen; 1455 struct lwp *l = curlwp; /* XXX */ 1456 1457 optlen = m ? m->m_len : 0; 1458 error = optval = 0; 1459 privileged = (l == 0 || kauth_authorize_generic(l->l_cred, 1460 KAUTH_GENERIC_ISSUSER, &l->l_acflag)) ? 0 : 1; 1461 uproto = (int)so->so_proto->pr_protocol; 1462 1463 if (level == IPPROTO_IPV6) { 1464 switch (op) { 1465 case PRCO_SETOPT: 1466 switch (optname) { 1467 #ifdef RFC2292 1468 case IPV6_2292PKTOPTIONS: 1469 /* m is freed in ip6_pcbopts */ 1470 error = ip6_pcbopts(&in6p->in6p_outputopts, 1471 m, so); 1472 break; 1473 #endif 1474 1475 /* 1476 * Use of some Hop-by-Hop options or some 1477 * Destination options, might require special 1478 * privilege. That is, normal applications 1479 * (without special privilege) might be forbidden 1480 * from setting certain options in outgoing packets, 1481 * and might never see certain options in received 1482 * packets. [RFC 2292 Section 6] 1483 * KAME specific note: 1484 * KAME prevents non-privileged users from sending or 1485 * receiving ANY hbh/dst options in order to avoid 1486 * overhead of parsing options in the kernel. 1487 */ 1488 case IPV6_RECVHOPOPTS: 1489 case IPV6_RECVDSTOPTS: 1490 case IPV6_RECVRTHDRDSTOPTS: 1491 if (!privileged) { 1492 error = EPERM; 1493 break; 1494 } 1495 /* FALLTHROUGH */ 1496 case IPV6_UNICAST_HOPS: 1497 case IPV6_HOPLIMIT: 1498 case IPV6_FAITH: 1499 1500 case IPV6_RECVPKTINFO: 1501 case IPV6_RECVHOPLIMIT: 1502 case IPV6_RECVRTHDR: 1503 case IPV6_RECVPATHMTU: 1504 case IPV6_RECVTCLASS: 1505 case IPV6_V6ONLY: 1506 if (optlen != sizeof(int)) { 1507 error = EINVAL; 1508 break; 1509 } 1510 optval = *mtod(m, int *); 1511 switch (optname) { 1512 1513 case IPV6_UNICAST_HOPS: 1514 if (optval < -1 || optval >= 256) 1515 error = EINVAL; 1516 else { 1517 /* -1 = kernel default */ 1518 in6p->in6p_hops = optval; 1519 } 1520 break; 1521 #define OPTSET(bit) \ 1522 do { \ 1523 if (optval) \ 1524 in6p->in6p_flags |= (bit); \ 1525 else \ 1526 in6p->in6p_flags &= ~(bit); \ 1527 } while (/*CONSTCOND*/ 0) 1528 1529 #ifdef RFC2292 1530 #define OPTSET2292(bit) \ 1531 do { \ 1532 in6p->in6p_flags |= IN6P_RFC2292; \ 1533 if (optval) \ 1534 in6p->in6p_flags |= (bit); \ 1535 else \ 1536 in6p->in6p_flags &= ~(bit); \ 1537 } while (/*CONSTCOND*/ 0) 1538 #endif 1539 1540 #define OPTBIT(bit) (in6p->in6p_flags & (bit) ? 1 : 0) 1541 1542 case IPV6_RECVPKTINFO: 1543 #ifdef RFC2292 1544 /* cannot mix with RFC2292 */ 1545 if (OPTBIT(IN6P_RFC2292)) { 1546 error = EINVAL; 1547 break; 1548 } 1549 #endif 1550 OPTSET(IN6P_PKTINFO); 1551 break; 1552 1553 case IPV6_HOPLIMIT: 1554 { 1555 struct ip6_pktopts **optp; 1556 1557 #ifdef RFC2292 1558 /* cannot mix with RFC2292 */ 1559 if (OPTBIT(IN6P_RFC2292)) { 1560 error = EINVAL; 1561 break; 1562 } 1563 #endif 1564 optp = &in6p->in6p_outputopts; 1565 error = ip6_pcbopt(IPV6_HOPLIMIT, 1566 (u_char *)&optval, 1567 sizeof(optval), 1568 optp, 1569 privileged, uproto); 1570 break; 1571 } 1572 1573 case IPV6_RECVHOPLIMIT: 1574 #ifdef RFC2292 1575 /* cannot mix with RFC2292 */ 1576 if (OPTBIT(IN6P_RFC2292)) { 1577 error = EINVAL; 1578 break; 1579 } 1580 #endif 1581 OPTSET(IN6P_HOPLIMIT); 1582 break; 1583 1584 case IPV6_RECVHOPOPTS: 1585 #ifdef RFC2292 1586 /* cannot mix with RFC2292 */ 1587 if (OPTBIT(IN6P_RFC2292)) { 1588 error = EINVAL; 1589 break; 1590 } 1591 #endif 1592 OPTSET(IN6P_HOPOPTS); 1593 break; 1594 1595 case IPV6_RECVDSTOPTS: 1596 #ifdef RFC2292 1597 /* cannot mix with RFC2292 */ 1598 if (OPTBIT(IN6P_RFC2292)) { 1599 error = EINVAL; 1600 break; 1601 } 1602 #endif 1603 OPTSET(IN6P_DSTOPTS); 1604 break; 1605 1606 case IPV6_RECVRTHDRDSTOPTS: 1607 #ifdef RFC2292 1608 /* cannot mix with RFC2292 */ 1609 if (OPTBIT(IN6P_RFC2292)) { 1610 error = EINVAL; 1611 break; 1612 } 1613 #endif 1614 OPTSET(IN6P_RTHDRDSTOPTS); 1615 break; 1616 1617 case IPV6_RECVRTHDR: 1618 #ifdef RFC2292 1619 /* cannot mix with RFC2292 */ 1620 if (OPTBIT(IN6P_RFC2292)) { 1621 error = EINVAL; 1622 break; 1623 } 1624 #endif 1625 OPTSET(IN6P_RTHDR); 1626 break; 1627 1628 case IPV6_FAITH: 1629 OPTSET(IN6P_FAITH); 1630 break; 1631 1632 case IPV6_RECVPATHMTU: 1633 /* 1634 * We ignore this option for TCP 1635 * sockets. 1636 * (RFC3542 leaves this case 1637 * unspecified.) 1638 */ 1639 if (uproto != IPPROTO_TCP) 1640 OPTSET(IN6P_MTU); 1641 break; 1642 1643 case IPV6_V6ONLY: 1644 /* 1645 * make setsockopt(IPV6_V6ONLY) 1646 * available only prior to bind(2). 1647 * see ipng mailing list, Jun 22 2001. 1648 */ 1649 if (in6p->in6p_lport || 1650 !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr)) { 1651 error = EINVAL; 1652 break; 1653 } 1654 #ifdef INET6_BINDV6ONLY 1655 if (!optval) 1656 error = EINVAL; 1657 #else 1658 OPTSET(IN6P_IPV6_V6ONLY); 1659 #endif 1660 break; 1661 case IPV6_RECVTCLASS: 1662 #ifdef RFC2292 1663 /* cannot mix with RFC2292 XXX */ 1664 if (OPTBIT(IN6P_RFC2292)) { 1665 error = EINVAL; 1666 break; 1667 } 1668 #endif 1669 OPTSET(IN6P_TCLASS); 1670 break; 1671 1672 } 1673 break; 1674 1675 case IPV6_OTCLASS: 1676 { 1677 struct ip6_pktopts **optp; 1678 u_int8_t tclass; 1679 1680 if (optlen != sizeof(tclass)) { 1681 error = EINVAL; 1682 break; 1683 } 1684 tclass = *mtod(m, u_int8_t *); 1685 optp = &in6p->in6p_outputopts; 1686 error = ip6_pcbopt(optname, 1687 (u_char *)&tclass, 1688 sizeof(tclass), 1689 optp, 1690 privileged, uproto); 1691 break; 1692 } 1693 1694 case IPV6_TCLASS: 1695 case IPV6_DONTFRAG: 1696 case IPV6_USE_MIN_MTU: 1697 if (optlen != sizeof(optval)) { 1698 error = EINVAL; 1699 break; 1700 } 1701 optval = *mtod(m, int *); 1702 { 1703 struct ip6_pktopts **optp; 1704 optp = &in6p->in6p_outputopts; 1705 error = ip6_pcbopt(optname, 1706 (u_char *)&optval, 1707 sizeof(optval), 1708 optp, 1709 privileged, uproto); 1710 break; 1711 } 1712 1713 #ifdef RFC2292 1714 case IPV6_2292PKTINFO: 1715 case IPV6_2292HOPLIMIT: 1716 case IPV6_2292HOPOPTS: 1717 case IPV6_2292DSTOPTS: 1718 case IPV6_2292RTHDR: 1719 /* RFC 2292 */ 1720 if (optlen != sizeof(int)) { 1721 error = EINVAL; 1722 break; 1723 } 1724 optval = *mtod(m, int *); 1725 switch (optname) { 1726 case IPV6_2292PKTINFO: 1727 OPTSET2292(IN6P_PKTINFO); 1728 break; 1729 case IPV6_2292HOPLIMIT: 1730 OPTSET2292(IN6P_HOPLIMIT); 1731 break; 1732 case IPV6_2292HOPOPTS: 1733 /* 1734 * Check super-user privilege. 1735 * See comments for IPV6_RECVHOPOPTS. 1736 */ 1737 if (!privileged) 1738 return (EPERM); 1739 OPTSET2292(IN6P_HOPOPTS); 1740 break; 1741 case IPV6_2292DSTOPTS: 1742 if (!privileged) 1743 return (EPERM); 1744 OPTSET2292(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); /* XXX */ 1745 break; 1746 case IPV6_2292RTHDR: 1747 OPTSET2292(IN6P_RTHDR); 1748 break; 1749 } 1750 break; 1751 #endif 1752 case IPV6_PKTINFO: 1753 case IPV6_HOPOPTS: 1754 case IPV6_RTHDR: 1755 case IPV6_DSTOPTS: 1756 case IPV6_RTHDRDSTOPTS: 1757 case IPV6_NEXTHOP: 1758 { 1759 /* new advanced API (RFC3542) */ 1760 u_char *optbuf; 1761 int optbuflen; 1762 struct ip6_pktopts **optp; 1763 1764 #ifdef RFC2292 1765 /* cannot mix with RFC2292 */ 1766 if (OPTBIT(IN6P_RFC2292)) { 1767 error = EINVAL; 1768 break; 1769 } 1770 #endif 1771 1772 if (m && m->m_next) { 1773 error = EINVAL; /* XXX */ 1774 break; 1775 } 1776 if (m) { 1777 optbuf = mtod(m, u_char *); 1778 optbuflen = m->m_len; 1779 } else { 1780 optbuf = NULL; 1781 optbuflen = 0; 1782 } 1783 optp = &in6p->in6p_outputopts; 1784 error = ip6_pcbopt(optname, 1785 optbuf, optbuflen, 1786 optp, privileged, uproto); 1787 break; 1788 } 1789 #undef OPTSET 1790 1791 case IPV6_MULTICAST_IF: 1792 case IPV6_MULTICAST_HOPS: 1793 case IPV6_MULTICAST_LOOP: 1794 case IPV6_JOIN_GROUP: 1795 case IPV6_LEAVE_GROUP: 1796 error = ip6_setmoptions(optname, 1797 &in6p->in6p_moptions, m); 1798 break; 1799 1800 case IPV6_PORTRANGE: 1801 optval = *mtod(m, int *); 1802 1803 switch (optval) { 1804 case IPV6_PORTRANGE_DEFAULT: 1805 in6p->in6p_flags &= ~(IN6P_LOWPORT); 1806 in6p->in6p_flags &= ~(IN6P_HIGHPORT); 1807 break; 1808 1809 case IPV6_PORTRANGE_HIGH: 1810 in6p->in6p_flags &= ~(IN6P_LOWPORT); 1811 in6p->in6p_flags |= IN6P_HIGHPORT; 1812 break; 1813 1814 case IPV6_PORTRANGE_LOW: 1815 in6p->in6p_flags &= ~(IN6P_HIGHPORT); 1816 in6p->in6p_flags |= IN6P_LOWPORT; 1817 break; 1818 1819 default: 1820 error = EINVAL; 1821 break; 1822 } 1823 break; 1824 1825 #ifdef IPSEC 1826 case IPV6_IPSEC_POLICY: 1827 { 1828 caddr_t req = NULL; 1829 size_t len = 0; 1830 if (m) { 1831 req = mtod(m, caddr_t); 1832 len = m->m_len; 1833 } 1834 error = ipsec6_set_policy(in6p, optname, req, 1835 len, privileged); 1836 } 1837 break; 1838 #endif /* IPSEC */ 1839 1840 default: 1841 error = ENOPROTOOPT; 1842 break; 1843 } 1844 if (m) 1845 (void)m_free(m); 1846 break; 1847 1848 case PRCO_GETOPT: 1849 switch (optname) { 1850 #ifdef RFC2292 1851 case IPV6_2292PKTOPTIONS: 1852 /* 1853 * RFC3542 (effectively) deprecated the 1854 * semantics of the 2292-style pktoptions. 1855 * Since it was not reliable in nature (i.e., 1856 * applications had to expect the lack of some 1857 * information after all), it would make sense 1858 * to simplify this part by always returning 1859 * empty data. 1860 */ 1861 *mp = m_get(M_WAIT, MT_SOOPTS); 1862 (*mp)->m_len = 0; 1863 break; 1864 #endif 1865 1866 case IPV6_RECVHOPOPTS: 1867 case IPV6_RECVDSTOPTS: 1868 case IPV6_RECVRTHDRDSTOPTS: 1869 case IPV6_UNICAST_HOPS: 1870 case IPV6_RECVPKTINFO: 1871 case IPV6_RECVHOPLIMIT: 1872 case IPV6_RECVRTHDR: 1873 case IPV6_RECVPATHMTU: 1874 1875 case IPV6_FAITH: 1876 case IPV6_V6ONLY: 1877 case IPV6_PORTRANGE: 1878 case IPV6_RECVTCLASS: 1879 switch (optname) { 1880 1881 case IPV6_RECVHOPOPTS: 1882 optval = OPTBIT(IN6P_HOPOPTS); 1883 break; 1884 1885 case IPV6_RECVDSTOPTS: 1886 optval = OPTBIT(IN6P_DSTOPTS); 1887 break; 1888 1889 case IPV6_RECVRTHDRDSTOPTS: 1890 optval = OPTBIT(IN6P_RTHDRDSTOPTS); 1891 break; 1892 1893 case IPV6_UNICAST_HOPS: 1894 optval = in6p->in6p_hops; 1895 break; 1896 1897 case IPV6_RECVPKTINFO: 1898 optval = OPTBIT(IN6P_PKTINFO); 1899 break; 1900 1901 case IPV6_RECVHOPLIMIT: 1902 optval = OPTBIT(IN6P_HOPLIMIT); 1903 break; 1904 1905 case IPV6_RECVRTHDR: 1906 optval = OPTBIT(IN6P_RTHDR); 1907 break; 1908 1909 case IPV6_RECVPATHMTU: 1910 optval = OPTBIT(IN6P_MTU); 1911 break; 1912 1913 case IPV6_FAITH: 1914 optval = OPTBIT(IN6P_FAITH); 1915 break; 1916 1917 case IPV6_V6ONLY: 1918 optval = OPTBIT(IN6P_IPV6_V6ONLY); 1919 break; 1920 1921 case IPV6_PORTRANGE: 1922 { 1923 int flags; 1924 flags = in6p->in6p_flags; 1925 if (flags & IN6P_HIGHPORT) 1926 optval = IPV6_PORTRANGE_HIGH; 1927 else if (flags & IN6P_LOWPORT) 1928 optval = IPV6_PORTRANGE_LOW; 1929 else 1930 optval = 0; 1931 break; 1932 } 1933 case IPV6_RECVTCLASS: 1934 optval = OPTBIT(IN6P_TCLASS); 1935 break; 1936 1937 } 1938 if (error) 1939 break; 1940 *mp = m = m_get(M_WAIT, MT_SOOPTS); 1941 m->m_len = sizeof(int); 1942 *mtod(m, int *) = optval; 1943 break; 1944 1945 case IPV6_PATHMTU: 1946 { 1947 u_long pmtu = 0; 1948 struct ip6_mtuinfo mtuinfo; 1949 struct route_in6 *ro = (struct route_in6 *)&in6p 1950 ->in6p_route; 1951 1952 if (!(so->so_state & SS_ISCONNECTED)) 1953 return (ENOTCONN); 1954 /* 1955 * XXX: we dot not consider the case of source 1956 * routing, or optional information to specify 1957 * the outgoing interface. 1958 */ 1959 error = ip6_getpmtu(ro, NULL, NULL, 1960 &in6p->in6p_faddr, &pmtu, NULL); 1961 if (error) 1962 break; 1963 if (pmtu > IPV6_MAXPACKET) 1964 pmtu = IPV6_MAXPACKET; 1965 1966 memset(&mtuinfo, 0, sizeof(mtuinfo)); 1967 mtuinfo.ip6m_mtu = (u_int32_t)pmtu; 1968 optdata = (void *)&mtuinfo; 1969 optdatalen = sizeof(mtuinfo); 1970 if (optdatalen > MCLBYTES) 1971 return (EMSGSIZE); /* XXX */ 1972 *mp = m = m_get(M_WAIT, MT_SOOPTS); 1973 if (optdatalen > MLEN) 1974 MCLGET(m, M_WAIT); 1975 m->m_len = optdatalen; 1976 memcpy(mtod(m, void *), optdata, optdatalen); 1977 break; 1978 } 1979 1980 #ifdef RFC2292 1981 case IPV6_2292PKTINFO: 1982 case IPV6_2292HOPLIMIT: 1983 case IPV6_2292HOPOPTS: 1984 case IPV6_2292RTHDR: 1985 case IPV6_2292DSTOPTS: 1986 switch (optname) { 1987 case IPV6_2292PKTINFO: 1988 optval = OPTBIT(IN6P_PKTINFO); 1989 break; 1990 case IPV6_2292HOPLIMIT: 1991 optval = OPTBIT(IN6P_HOPLIMIT); 1992 break; 1993 case IPV6_2292HOPOPTS: 1994 optval = OPTBIT(IN6P_HOPOPTS); 1995 break; 1996 case IPV6_2292RTHDR: 1997 optval = OPTBIT(IN6P_RTHDR); 1998 break; 1999 case IPV6_2292DSTOPTS: 2000 optval = OPTBIT(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); 2001 break; 2002 } 2003 *mp = m = m_get(M_WAIT, MT_SOOPTS); 2004 m->m_len = sizeof(int); 2005 *mtod(m, int *) = optval; 2006 break; 2007 #endif 2008 case IPV6_PKTINFO: 2009 case IPV6_HOPOPTS: 2010 case IPV6_RTHDR: 2011 case IPV6_DSTOPTS: 2012 case IPV6_RTHDRDSTOPTS: 2013 case IPV6_NEXTHOP: 2014 case IPV6_OTCLASS: 2015 case IPV6_TCLASS: 2016 case IPV6_DONTFRAG: 2017 case IPV6_USE_MIN_MTU: 2018 error = ip6_getpcbopt(in6p->in6p_outputopts, 2019 optname, mp); 2020 break; 2021 2022 case IPV6_MULTICAST_IF: 2023 case IPV6_MULTICAST_HOPS: 2024 case IPV6_MULTICAST_LOOP: 2025 case IPV6_JOIN_GROUP: 2026 case IPV6_LEAVE_GROUP: 2027 error = ip6_getmoptions(optname, 2028 in6p->in6p_moptions, mp); 2029 break; 2030 2031 #ifdef IPSEC 2032 case IPV6_IPSEC_POLICY: 2033 { 2034 caddr_t req = NULL; 2035 size_t len = 0; 2036 if (m) { 2037 req = mtod(m, caddr_t); 2038 len = m->m_len; 2039 } 2040 error = ipsec6_get_policy(in6p, req, len, mp); 2041 break; 2042 } 2043 #endif /* IPSEC */ 2044 2045 2046 2047 2048 default: 2049 error = ENOPROTOOPT; 2050 break; 2051 } 2052 break; 2053 } 2054 } else { 2055 error = EINVAL; 2056 if (op == PRCO_SETOPT && *mp) 2057 (void)m_free(*mp); 2058 } 2059 return (error); 2060 } 2061 2062 int 2063 ip6_raw_ctloutput(op, so, level, optname, mp) 2064 int op; 2065 struct socket *so; 2066 int level, optname; 2067 struct mbuf **mp; 2068 { 2069 int error = 0, optval, optlen; 2070 const int icmp6off = offsetof(struct icmp6_hdr, icmp6_cksum); 2071 struct in6pcb *in6p = sotoin6pcb(so); 2072 struct mbuf *m = *mp; 2073 2074 optlen = m ? m->m_len : 0; 2075 2076 if (level != IPPROTO_IPV6) { 2077 if (op == PRCO_SETOPT && *mp) 2078 (void)m_free(*mp); 2079 return (EINVAL); 2080 } 2081 2082 switch (optname) { 2083 case IPV6_CHECKSUM: 2084 /* 2085 * For ICMPv6 sockets, no modification allowed for checksum 2086 * offset, permit "no change" values to help existing apps. 2087 * 2088 * XXX RFC3542 says: "An attempt to set IPV6_CHECKSUM 2089 * for an ICMPv6 socket will fail." The current 2090 * behavior does not meet RFC3542. 2091 */ 2092 switch (op) { 2093 case PRCO_SETOPT: 2094 if (optlen != sizeof(int)) { 2095 error = EINVAL; 2096 break; 2097 } 2098 optval = *mtod(m, int *); 2099 if ((optval % 2) != 0) { 2100 /* the API assumes even offset values */ 2101 error = EINVAL; 2102 } else if (so->so_proto->pr_protocol == 2103 IPPROTO_ICMPV6) { 2104 if (optval != icmp6off) 2105 error = EINVAL; 2106 } else 2107 in6p->in6p_cksum = optval; 2108 break; 2109 2110 case PRCO_GETOPT: 2111 if (so->so_proto->pr_protocol == IPPROTO_ICMPV6) 2112 optval = icmp6off; 2113 else 2114 optval = in6p->in6p_cksum; 2115 2116 *mp = m = m_get(M_WAIT, MT_SOOPTS); 2117 m->m_len = sizeof(int); 2118 *mtod(m, int *) = optval; 2119 break; 2120 2121 default: 2122 error = EINVAL; 2123 break; 2124 } 2125 break; 2126 2127 default: 2128 error = ENOPROTOOPT; 2129 break; 2130 } 2131 2132 if (op == PRCO_SETOPT && m) 2133 (void)m_free(m); 2134 2135 return (error); 2136 } 2137 2138 #ifdef RFC2292 2139 /* 2140 * Set up IP6 options in pcb for insertion in output packets or 2141 * specifying behavior of outgoing packets. 2142 */ 2143 static int 2144 ip6_pcbopts(pktopt, m, so) 2145 struct ip6_pktopts **pktopt; 2146 struct mbuf *m; 2147 struct socket *so; 2148 { 2149 struct ip6_pktopts *opt = *pktopt; 2150 int error = 0; 2151 struct lwp *l = curlwp; /* XXX */ 2152 int priv = 0; 2153 2154 /* turn off any old options. */ 2155 if (opt) { 2156 #ifdef DIAGNOSTIC 2157 if (opt->ip6po_pktinfo || opt->ip6po_nexthop || 2158 opt->ip6po_hbh || opt->ip6po_dest1 || opt->ip6po_dest2 || 2159 opt->ip6po_rhinfo.ip6po_rhi_rthdr) 2160 printf("ip6_pcbopts: all specified options are cleared.\n"); 2161 #endif 2162 ip6_clearpktopts(opt, -1); 2163 } else 2164 opt = malloc(sizeof(*opt), M_IP6OPT, M_WAITOK); 2165 *pktopt = NULL; 2166 2167 if (!m || m->m_len == 0) { 2168 /* 2169 * Only turning off any previous options, regardless of 2170 * whether the opt is just created or given. 2171 */ 2172 free(opt, M_IP6OPT); 2173 return (0); 2174 } 2175 2176 /* set options specified by user. */ 2177 if (l && !kauth_authorize_generic(l->l_cred, KAUTH_GENERIC_ISSUSER, 2178 &l->l_acflag)) 2179 priv = 1; 2180 if ((error = ip6_setpktopts(m, opt, NULL, priv, 2181 so->so_proto->pr_protocol)) != 0) { 2182 ip6_clearpktopts(opt, -1); /* XXX: discard all options */ 2183 free(opt, M_IP6OPT); 2184 return (error); 2185 } 2186 *pktopt = opt; 2187 return (0); 2188 } 2189 #endif 2190 2191 /* 2192 * initialize ip6_pktopts. beware that there are non-zero default values in 2193 * the struct. 2194 */ 2195 void 2196 ip6_initpktopts(struct ip6_pktopts *opt) 2197 { 2198 2199 memset(opt, 0, sizeof(*opt)); 2200 opt->ip6po_hlim = -1; /* -1 means default hop limit */ 2201 opt->ip6po_tclass = -1; /* -1 means default traffic class */ 2202 opt->ip6po_minmtu = IP6PO_MINMTU_MCASTONLY; 2203 } 2204 2205 #define sin6tosa(sin6) ((struct sockaddr *)(sin6)) /* XXX */ 2206 static int 2207 ip6_pcbopt(int optname, u_char *buf, int len, struct ip6_pktopts **pktopt, 2208 int priv, int uproto) 2209 { 2210 struct ip6_pktopts *opt; 2211 2212 if (*pktopt == NULL) { 2213 *pktopt = malloc(sizeof(struct ip6_pktopts), M_IP6OPT, 2214 M_WAITOK); 2215 ip6_initpktopts(*pktopt); 2216 } 2217 opt = *pktopt; 2218 2219 return (ip6_setpktopt(optname, buf, len, opt, priv, 1, 0, uproto)); 2220 } 2221 2222 static int 2223 ip6_getpcbopt(struct ip6_pktopts *pktopt, int optname, struct mbuf **mp) 2224 { 2225 void *optdata = NULL; 2226 int optdatalen = 0; 2227 struct ip6_ext *ip6e; 2228 int error = 0; 2229 struct in6_pktinfo null_pktinfo; 2230 int deftclass = 0, on; 2231 int defminmtu = IP6PO_MINMTU_MCASTONLY; 2232 struct mbuf *m; 2233 2234 switch (optname) { 2235 case IPV6_PKTINFO: 2236 if (pktopt && pktopt->ip6po_pktinfo) 2237 optdata = (void *)pktopt->ip6po_pktinfo; 2238 else { 2239 /* XXX: we don't have to do this every time... */ 2240 memset(&null_pktinfo, 0, sizeof(null_pktinfo)); 2241 optdata = (void *)&null_pktinfo; 2242 } 2243 optdatalen = sizeof(struct in6_pktinfo); 2244 break; 2245 case IPV6_OTCLASS: 2246 /* XXX */ 2247 return (EINVAL); 2248 case IPV6_TCLASS: 2249 if (pktopt && pktopt->ip6po_tclass >= 0) 2250 optdata = (void *)&pktopt->ip6po_tclass; 2251 else 2252 optdata = (void *)&deftclass; 2253 optdatalen = sizeof(int); 2254 break; 2255 case IPV6_HOPOPTS: 2256 if (pktopt && pktopt->ip6po_hbh) { 2257 optdata = (void *)pktopt->ip6po_hbh; 2258 ip6e = (struct ip6_ext *)pktopt->ip6po_hbh; 2259 optdatalen = (ip6e->ip6e_len + 1) << 3; 2260 } 2261 break; 2262 case IPV6_RTHDR: 2263 if (pktopt && pktopt->ip6po_rthdr) { 2264 optdata = (void *)pktopt->ip6po_rthdr; 2265 ip6e = (struct ip6_ext *)pktopt->ip6po_rthdr; 2266 optdatalen = (ip6e->ip6e_len + 1) << 3; 2267 } 2268 break; 2269 case IPV6_RTHDRDSTOPTS: 2270 if (pktopt && pktopt->ip6po_dest1) { 2271 optdata = (void *)pktopt->ip6po_dest1; 2272 ip6e = (struct ip6_ext *)pktopt->ip6po_dest1; 2273 optdatalen = (ip6e->ip6e_len + 1) << 3; 2274 } 2275 break; 2276 case IPV6_DSTOPTS: 2277 if (pktopt && pktopt->ip6po_dest2) { 2278 optdata = (void *)pktopt->ip6po_dest2; 2279 ip6e = (struct ip6_ext *)pktopt->ip6po_dest2; 2280 optdatalen = (ip6e->ip6e_len + 1) << 3; 2281 } 2282 break; 2283 case IPV6_NEXTHOP: 2284 if (pktopt && pktopt->ip6po_nexthop) { 2285 optdata = (void *)pktopt->ip6po_nexthop; 2286 optdatalen = pktopt->ip6po_nexthop->sa_len; 2287 } 2288 break; 2289 case IPV6_USE_MIN_MTU: 2290 if (pktopt) 2291 optdata = (void *)&pktopt->ip6po_minmtu; 2292 else 2293 optdata = (void *)&defminmtu; 2294 optdatalen = sizeof(int); 2295 break; 2296 case IPV6_DONTFRAG: 2297 if (pktopt && ((pktopt->ip6po_flags) & IP6PO_DONTFRAG)) 2298 on = 1; 2299 else 2300 on = 0; 2301 optdata = (void *)&on; 2302 optdatalen = sizeof(on); 2303 break; 2304 default: /* should not happen */ 2305 #ifdef DIAGNOSTIC 2306 panic("ip6_getpcbopt: unexpected option\n"); 2307 #endif 2308 return (ENOPROTOOPT); 2309 } 2310 2311 if (optdatalen > MCLBYTES) 2312 return (EMSGSIZE); /* XXX */ 2313 *mp = m = m_get(M_WAIT, MT_SOOPTS); 2314 if (optdatalen > MLEN) 2315 MCLGET(m, M_WAIT); 2316 m->m_len = optdatalen; 2317 if (optdatalen) 2318 memcpy(mtod(m, void *), optdata, optdatalen); 2319 2320 return (error); 2321 } 2322 2323 void 2324 ip6_clearpktopts(struct ip6_pktopts *pktopt, int optname) 2325 { 2326 if (optname == -1 || optname == IPV6_PKTINFO) { 2327 if (pktopt->ip6po_pktinfo) 2328 free(pktopt->ip6po_pktinfo, M_IP6OPT); 2329 pktopt->ip6po_pktinfo = NULL; 2330 } 2331 if (optname == -1 || optname == IPV6_HOPLIMIT) 2332 pktopt->ip6po_hlim = -1; 2333 if (optname == -1 || optname == IPV6_TCLASS) 2334 pktopt->ip6po_tclass = -1; 2335 if (optname == -1 || optname == IPV6_NEXTHOP) { 2336 if (pktopt->ip6po_nextroute.ro_rt) { 2337 RTFREE(pktopt->ip6po_nextroute.ro_rt); 2338 pktopt->ip6po_nextroute.ro_rt = NULL; 2339 } 2340 if (pktopt->ip6po_nexthop) 2341 free(pktopt->ip6po_nexthop, M_IP6OPT); 2342 pktopt->ip6po_nexthop = NULL; 2343 } 2344 if (optname == -1 || optname == IPV6_HOPOPTS) { 2345 if (pktopt->ip6po_hbh) 2346 free(pktopt->ip6po_hbh, M_IP6OPT); 2347 pktopt->ip6po_hbh = NULL; 2348 } 2349 if (optname == -1 || optname == IPV6_RTHDRDSTOPTS) { 2350 if (pktopt->ip6po_dest1) 2351 free(pktopt->ip6po_dest1, M_IP6OPT); 2352 pktopt->ip6po_dest1 = NULL; 2353 } 2354 if (optname == -1 || optname == IPV6_RTHDR) { 2355 if (pktopt->ip6po_rhinfo.ip6po_rhi_rthdr) 2356 free(pktopt->ip6po_rhinfo.ip6po_rhi_rthdr, M_IP6OPT); 2357 pktopt->ip6po_rhinfo.ip6po_rhi_rthdr = NULL; 2358 if (pktopt->ip6po_route.ro_rt) { 2359 RTFREE(pktopt->ip6po_route.ro_rt); 2360 pktopt->ip6po_route.ro_rt = NULL; 2361 } 2362 } 2363 if (optname == -1 || optname == IPV6_DSTOPTS) { 2364 if (pktopt->ip6po_dest2) 2365 free(pktopt->ip6po_dest2, M_IP6OPT); 2366 pktopt->ip6po_dest2 = NULL; 2367 } 2368 } 2369 2370 #define PKTOPT_EXTHDRCPY(type) \ 2371 do { \ 2372 if (src->type) { \ 2373 int hlen = (((struct ip6_ext *)src->type)->ip6e_len + 1) << 3;\ 2374 dst->type = malloc(hlen, M_IP6OPT, canwait); \ 2375 if (dst->type == NULL && canwait == M_NOWAIT) \ 2376 goto bad; \ 2377 memcpy(dst->type, src->type, hlen); \ 2378 } \ 2379 } while (/*CONSTCOND*/ 0) 2380 2381 static int 2382 copypktopts(struct ip6_pktopts *dst, struct ip6_pktopts *src, int canwait) 2383 { 2384 dst->ip6po_hlim = src->ip6po_hlim; 2385 dst->ip6po_tclass = src->ip6po_tclass; 2386 dst->ip6po_flags = src->ip6po_flags; 2387 if (src->ip6po_pktinfo) { 2388 dst->ip6po_pktinfo = malloc(sizeof(*dst->ip6po_pktinfo), 2389 M_IP6OPT, canwait); 2390 if (dst->ip6po_pktinfo == NULL && canwait == M_NOWAIT) 2391 goto bad; 2392 *dst->ip6po_pktinfo = *src->ip6po_pktinfo; 2393 } 2394 if (src->ip6po_nexthop) { 2395 dst->ip6po_nexthop = malloc(src->ip6po_nexthop->sa_len, 2396 M_IP6OPT, canwait); 2397 if (dst->ip6po_nexthop == NULL && canwait == M_NOWAIT) 2398 goto bad; 2399 memcpy(dst->ip6po_nexthop, src->ip6po_nexthop, 2400 src->ip6po_nexthop->sa_len); 2401 } 2402 PKTOPT_EXTHDRCPY(ip6po_hbh); 2403 PKTOPT_EXTHDRCPY(ip6po_dest1); 2404 PKTOPT_EXTHDRCPY(ip6po_dest2); 2405 PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */ 2406 return (0); 2407 2408 bad: 2409 if (dst->ip6po_pktinfo) free(dst->ip6po_pktinfo, M_IP6OPT); 2410 if (dst->ip6po_nexthop) free(dst->ip6po_nexthop, M_IP6OPT); 2411 if (dst->ip6po_hbh) free(dst->ip6po_hbh, M_IP6OPT); 2412 if (dst->ip6po_dest1) free(dst->ip6po_dest1, M_IP6OPT); 2413 if (dst->ip6po_dest2) free(dst->ip6po_dest2, M_IP6OPT); 2414 if (dst->ip6po_rthdr) free(dst->ip6po_rthdr, M_IP6OPT); 2415 2416 return (ENOBUFS); 2417 } 2418 #undef PKTOPT_EXTHDRCPY 2419 2420 struct ip6_pktopts * 2421 ip6_copypktopts(struct ip6_pktopts *src, int canwait) 2422 { 2423 int error; 2424 struct ip6_pktopts *dst; 2425 2426 dst = malloc(sizeof(*dst), M_IP6OPT, canwait); 2427 if (dst == NULL && canwait == M_NOWAIT) 2428 return (NULL); 2429 ip6_initpktopts(dst); 2430 2431 if ((error = copypktopts(dst, src, canwait)) != 0) { 2432 free(dst, M_IP6OPT); 2433 return (NULL); 2434 } 2435 2436 return (dst); 2437 } 2438 2439 void 2440 ip6_freepcbopts(struct ip6_pktopts *pktopt) 2441 { 2442 if (pktopt == NULL) 2443 return; 2444 2445 ip6_clearpktopts(pktopt, -1); 2446 2447 free(pktopt, M_IP6OPT); 2448 } 2449 2450 /* 2451 * Set the IP6 multicast options in response to user setsockopt(). 2452 */ 2453 static int 2454 ip6_setmoptions(optname, im6op, m) 2455 int optname; 2456 struct ip6_moptions **im6op; 2457 struct mbuf *m; 2458 { 2459 int error = 0; 2460 u_int loop, ifindex; 2461 struct ipv6_mreq *mreq; 2462 struct ifnet *ifp; 2463 struct ip6_moptions *im6o = *im6op; 2464 struct route_in6 ro; 2465 struct in6_multi_mship *imm; 2466 struct lwp *l = curlwp; /* XXX */ 2467 2468 if (im6o == NULL) { 2469 /* 2470 * No multicast option buffer attached to the pcb; 2471 * allocate one and initialize to default values. 2472 */ 2473 im6o = (struct ip6_moptions *) 2474 malloc(sizeof(*im6o), M_IPMOPTS, M_WAITOK); 2475 2476 if (im6o == NULL) 2477 return (ENOBUFS); 2478 *im6op = im6o; 2479 im6o->im6o_multicast_ifp = NULL; 2480 im6o->im6o_multicast_hlim = ip6_defmcasthlim; 2481 im6o->im6o_multicast_loop = IPV6_DEFAULT_MULTICAST_LOOP; 2482 LIST_INIT(&im6o->im6o_memberships); 2483 } 2484 2485 switch (optname) { 2486 2487 case IPV6_MULTICAST_IF: 2488 /* 2489 * Select the interface for outgoing multicast packets. 2490 */ 2491 if (m == NULL || m->m_len != sizeof(u_int)) { 2492 error = EINVAL; 2493 break; 2494 } 2495 bcopy(mtod(m, u_int *), &ifindex, sizeof(ifindex)); 2496 if (ifindex != 0) { 2497 if (if_indexlim <= ifindex || !ifindex2ifnet[ifindex]) { 2498 error = ENXIO; /* XXX EINVAL? */ 2499 break; 2500 } 2501 ifp = ifindex2ifnet[ifindex]; 2502 if ((ifp->if_flags & IFF_MULTICAST) == 0) { 2503 error = EADDRNOTAVAIL; 2504 break; 2505 } 2506 } else 2507 ifp = NULL; 2508 im6o->im6o_multicast_ifp = ifp; 2509 break; 2510 2511 case IPV6_MULTICAST_HOPS: 2512 { 2513 /* 2514 * Set the IP6 hoplimit for outgoing multicast packets. 2515 */ 2516 int optval; 2517 if (m == NULL || m->m_len != sizeof(int)) { 2518 error = EINVAL; 2519 break; 2520 } 2521 bcopy(mtod(m, u_int *), &optval, sizeof(optval)); 2522 if (optval < -1 || optval >= 256) 2523 error = EINVAL; 2524 else if (optval == -1) 2525 im6o->im6o_multicast_hlim = ip6_defmcasthlim; 2526 else 2527 im6o->im6o_multicast_hlim = optval; 2528 break; 2529 } 2530 2531 case IPV6_MULTICAST_LOOP: 2532 /* 2533 * Set the loopback flag for outgoing multicast packets. 2534 * Must be zero or one. 2535 */ 2536 if (m == NULL || m->m_len != sizeof(u_int)) { 2537 error = EINVAL; 2538 break; 2539 } 2540 bcopy(mtod(m, u_int *), &loop, sizeof(loop)); 2541 if (loop > 1) { 2542 error = EINVAL; 2543 break; 2544 } 2545 im6o->im6o_multicast_loop = loop; 2546 break; 2547 2548 case IPV6_JOIN_GROUP: 2549 /* 2550 * Add a multicast group membership. 2551 * Group must be a valid IP6 multicast address. 2552 */ 2553 if (m == NULL || m->m_len != sizeof(struct ipv6_mreq)) { 2554 error = EINVAL; 2555 break; 2556 } 2557 mreq = mtod(m, struct ipv6_mreq *); 2558 if (IN6_IS_ADDR_UNSPECIFIED(&mreq->ipv6mr_multiaddr)) { 2559 /* 2560 * We use the unspecified address to specify to accept 2561 * all multicast addresses. Only super user is allowed 2562 * to do this. 2563 */ 2564 if (kauth_authorize_generic(l->l_cred, 2565 KAUTH_GENERIC_ISSUSER, &l->l_acflag)) 2566 { 2567 error = EACCES; 2568 break; 2569 } 2570 } else if (!IN6_IS_ADDR_MULTICAST(&mreq->ipv6mr_multiaddr)) { 2571 error = EINVAL; 2572 break; 2573 } 2574 2575 /* 2576 * If no interface was explicitly specified, choose an 2577 * appropriate one according to the given multicast address. 2578 */ 2579 if (mreq->ipv6mr_interface == 0) { 2580 struct sockaddr_in6 *dst; 2581 2582 /* 2583 * Look up the routing table for the 2584 * address, and choose the outgoing interface. 2585 * XXX: is it a good approach? 2586 */ 2587 bzero(&ro, sizeof(ro)); 2588 dst = (struct sockaddr_in6 *)&ro.ro_dst; 2589 dst->sin6_family = AF_INET6; 2590 dst->sin6_len = sizeof(*dst); 2591 dst->sin6_addr = mreq->ipv6mr_multiaddr; 2592 rtalloc((struct route *)&ro); 2593 if (ro.ro_rt == NULL) { 2594 error = EADDRNOTAVAIL; 2595 break; 2596 } 2597 ifp = ro.ro_rt->rt_ifp; 2598 rtfree(ro.ro_rt); 2599 } else { 2600 /* 2601 * If the interface is specified, validate it. 2602 */ 2603 if (if_indexlim <= mreq->ipv6mr_interface || 2604 !ifindex2ifnet[mreq->ipv6mr_interface]) { 2605 error = ENXIO; /* XXX EINVAL? */ 2606 break; 2607 } 2608 ifp = ifindex2ifnet[mreq->ipv6mr_interface]; 2609 } 2610 2611 /* 2612 * See if we found an interface, and confirm that it 2613 * supports multicast 2614 */ 2615 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) { 2616 error = EADDRNOTAVAIL; 2617 break; 2618 } 2619 2620 if (in6_setscope(&mreq->ipv6mr_multiaddr, ifp, NULL)) { 2621 error = EADDRNOTAVAIL; /* XXX: should not happen */ 2622 break; 2623 } 2624 2625 /* 2626 * See if the membership already exists. 2627 */ 2628 for (imm = im6o->im6o_memberships.lh_first; 2629 imm != NULL; imm = imm->i6mm_chain.le_next) 2630 if (imm->i6mm_maddr->in6m_ifp == ifp && 2631 IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr, 2632 &mreq->ipv6mr_multiaddr)) 2633 break; 2634 if (imm != NULL) { 2635 error = EADDRINUSE; 2636 break; 2637 } 2638 /* 2639 * Everything looks good; add a new record to the multicast 2640 * address list for the given interface. 2641 */ 2642 imm = in6_joingroup(ifp, &mreq->ipv6mr_multiaddr, &error, 0); 2643 if (imm == NULL) 2644 break; 2645 LIST_INSERT_HEAD(&im6o->im6o_memberships, imm, i6mm_chain); 2646 break; 2647 2648 case IPV6_LEAVE_GROUP: 2649 /* 2650 * Drop a multicast group membership. 2651 * Group must be a valid IP6 multicast address. 2652 */ 2653 if (m == NULL || m->m_len != sizeof(struct ipv6_mreq)) { 2654 error = EINVAL; 2655 break; 2656 } 2657 mreq = mtod(m, struct ipv6_mreq *); 2658 2659 /* 2660 * If an interface address was specified, get a pointer 2661 * to its ifnet structure. 2662 */ 2663 if (mreq->ipv6mr_interface != 0) { 2664 if (if_indexlim <= mreq->ipv6mr_interface || 2665 !ifindex2ifnet[mreq->ipv6mr_interface]) { 2666 error = ENXIO; /* XXX EINVAL? */ 2667 break; 2668 } 2669 ifp = ifindex2ifnet[mreq->ipv6mr_interface]; 2670 } else 2671 ifp = NULL; 2672 2673 /* Fill in the scope zone ID */ 2674 if (ifp) { 2675 if (in6_setscope(&mreq->ipv6mr_multiaddr, ifp, NULL)) { 2676 /* XXX: should not happen */ 2677 error = EADDRNOTAVAIL; 2678 break; 2679 } 2680 } else if (mreq->ipv6mr_interface != 0) { 2681 /* 2682 * XXX: This case would happens when the (positive) 2683 * index is in the valid range, but the corresponding 2684 * interface has been detached dynamically. The above 2685 * check probably avoids such case to happen here, but 2686 * we check it explicitly for safety. 2687 */ 2688 error = EADDRNOTAVAIL; 2689 break; 2690 } else { /* ipv6mr_interface == 0 */ 2691 struct sockaddr_in6 sa6_mc; 2692 2693 /* 2694 * The API spec says as follows: 2695 * If the interface index is specified as 0, the 2696 * system may choose a multicast group membership to 2697 * drop by matching the multicast address only. 2698 * On the other hand, we cannot disambiguate the scope 2699 * zone unless an interface is provided. Thus, we 2700 * check if there's ambiguity with the default scope 2701 * zone as the last resort. 2702 */ 2703 bzero(&sa6_mc, sizeof(sa6_mc)); 2704 sa6_mc.sin6_family = AF_INET6; 2705 sa6_mc.sin6_len = sizeof(sa6_mc); 2706 sa6_mc.sin6_addr = mreq->ipv6mr_multiaddr; 2707 error = sa6_embedscope(&sa6_mc, ip6_use_defzone); 2708 if (error != 0) 2709 break; 2710 mreq->ipv6mr_multiaddr = sa6_mc.sin6_addr; 2711 } 2712 2713 /* 2714 * Find the membership in the membership list. 2715 */ 2716 for (imm = im6o->im6o_memberships.lh_first; 2717 imm != NULL; imm = imm->i6mm_chain.le_next) { 2718 if ((ifp == NULL || imm->i6mm_maddr->in6m_ifp == ifp) && 2719 IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr, 2720 &mreq->ipv6mr_multiaddr)) 2721 break; 2722 } 2723 if (imm == NULL) { 2724 /* Unable to resolve interface */ 2725 error = EADDRNOTAVAIL; 2726 break; 2727 } 2728 /* 2729 * Give up the multicast address record to which the 2730 * membership points. 2731 */ 2732 LIST_REMOVE(imm, i6mm_chain); 2733 in6_leavegroup(imm); 2734 break; 2735 2736 default: 2737 error = EOPNOTSUPP; 2738 break; 2739 } 2740 2741 /* 2742 * If all options have default values, no need to keep the mbuf. 2743 */ 2744 if (im6o->im6o_multicast_ifp == NULL && 2745 im6o->im6o_multicast_hlim == ip6_defmcasthlim && 2746 im6o->im6o_multicast_loop == IPV6_DEFAULT_MULTICAST_LOOP && 2747 im6o->im6o_memberships.lh_first == NULL) { 2748 free(*im6op, M_IPMOPTS); 2749 *im6op = NULL; 2750 } 2751 2752 return (error); 2753 } 2754 2755 /* 2756 * Return the IP6 multicast options in response to user getsockopt(). 2757 */ 2758 static int 2759 ip6_getmoptions(optname, im6o, mp) 2760 int optname; 2761 struct ip6_moptions *im6o; 2762 struct mbuf **mp; 2763 { 2764 u_int *hlim, *loop, *ifindex; 2765 2766 *mp = m_get(M_WAIT, MT_SOOPTS); 2767 2768 switch (optname) { 2769 2770 case IPV6_MULTICAST_IF: 2771 ifindex = mtod(*mp, u_int *); 2772 (*mp)->m_len = sizeof(u_int); 2773 if (im6o == NULL || im6o->im6o_multicast_ifp == NULL) 2774 *ifindex = 0; 2775 else 2776 *ifindex = im6o->im6o_multicast_ifp->if_index; 2777 return (0); 2778 2779 case IPV6_MULTICAST_HOPS: 2780 hlim = mtod(*mp, u_int *); 2781 (*mp)->m_len = sizeof(u_int); 2782 if (im6o == NULL) 2783 *hlim = ip6_defmcasthlim; 2784 else 2785 *hlim = im6o->im6o_multicast_hlim; 2786 return (0); 2787 2788 case IPV6_MULTICAST_LOOP: 2789 loop = mtod(*mp, u_int *); 2790 (*mp)->m_len = sizeof(u_int); 2791 if (im6o == NULL) 2792 *loop = ip6_defmcasthlim; 2793 else 2794 *loop = im6o->im6o_multicast_loop; 2795 return (0); 2796 2797 default: 2798 return (EOPNOTSUPP); 2799 } 2800 } 2801 2802 /* 2803 * Discard the IP6 multicast options. 2804 */ 2805 void 2806 ip6_freemoptions(im6o) 2807 struct ip6_moptions *im6o; 2808 { 2809 struct in6_multi_mship *imm; 2810 2811 if (im6o == NULL) 2812 return; 2813 2814 while ((imm = im6o->im6o_memberships.lh_first) != NULL) { 2815 LIST_REMOVE(imm, i6mm_chain); 2816 in6_leavegroup(imm); 2817 } 2818 free(im6o, M_IPMOPTS); 2819 } 2820 2821 /* 2822 * Set IPv6 outgoing packet options based on advanced API. 2823 */ 2824 int 2825 ip6_setpktopts(control, opt, stickyopt, priv, uproto) 2826 struct mbuf *control; 2827 struct ip6_pktopts *opt, *stickyopt; 2828 int priv, uproto; 2829 { 2830 struct cmsghdr *cm = 0; 2831 2832 if (control == NULL || opt == NULL) 2833 return (EINVAL); 2834 2835 ip6_initpktopts(opt); 2836 if (stickyopt) { 2837 int error; 2838 2839 /* 2840 * If stickyopt is provided, make a local copy of the options 2841 * for this particular packet, then override them by ancillary 2842 * objects. 2843 * XXX: copypktopts() does not copy the cached route to a next 2844 * hop (if any). This is not very good in terms of efficiency, 2845 * but we can allow this since this option should be rarely 2846 * used. 2847 */ 2848 if ((error = copypktopts(opt, stickyopt, M_NOWAIT)) != 0) 2849 return (error); 2850 } 2851 2852 /* 2853 * XXX: Currently, we assume all the optional information is stored 2854 * in a single mbuf. 2855 */ 2856 if (control->m_next) 2857 return (EINVAL); 2858 2859 for (; control->m_len; control->m_data += CMSG_ALIGN(cm->cmsg_len), 2860 control->m_len -= CMSG_ALIGN(cm->cmsg_len)) { 2861 int error; 2862 2863 if (control->m_len < CMSG_LEN(0)) 2864 return (EINVAL); 2865 2866 cm = mtod(control, struct cmsghdr *); 2867 if (cm->cmsg_len == 0 || cm->cmsg_len > control->m_len) 2868 return (EINVAL); 2869 if (cm->cmsg_level != IPPROTO_IPV6) 2870 continue; 2871 2872 error = ip6_setpktopt(cm->cmsg_type, CMSG_DATA(cm), 2873 cm->cmsg_len - CMSG_LEN(0), opt, priv, 0, 1, uproto); 2874 if (error) 2875 return (error); 2876 } 2877 2878 return (0); 2879 } 2880 2881 /* 2882 * Set a particular packet option, as a sticky option or an ancillary data 2883 * item. "len" can be 0 only when it's a sticky option. 2884 * We have 4 cases of combination of "sticky" and "cmsg": 2885 * "sticky=0, cmsg=0": impossible 2886 * "sticky=0, cmsg=1": RFC2292 or RFC3542 ancillary data 2887 * "sticky=1, cmsg=0": RFC3542 socket option 2888 * "sticky=1, cmsg=1": RFC2292 socket option 2889 */ 2890 static int 2891 ip6_setpktopt(int optname, u_char *buf, int len, struct ip6_pktopts *opt, 2892 int priv, int sticky, int cmsg, int uproto) 2893 { 2894 int minmtupolicy; 2895 2896 if (!sticky && !cmsg) { 2897 #ifdef DIAGNOSTIC 2898 printf("ip6_setpktopt: impossible case\n"); 2899 #endif 2900 return (EINVAL); 2901 } 2902 2903 /* 2904 * IPV6_2292xxx is for backward compatibility to RFC2292, and should 2905 * not be specified in the context of RFC3542. Conversely, 2906 * RFC3542 types should not be specified in the context of RFC2292. 2907 */ 2908 if (!cmsg) { 2909 switch (optname) { 2910 case IPV6_2292PKTINFO: 2911 case IPV6_2292HOPLIMIT: 2912 case IPV6_2292NEXTHOP: 2913 case IPV6_2292HOPOPTS: 2914 case IPV6_2292DSTOPTS: 2915 case IPV6_2292RTHDR: 2916 case IPV6_2292PKTOPTIONS: 2917 return (ENOPROTOOPT); 2918 } 2919 } 2920 if (sticky && cmsg) { 2921 switch (optname) { 2922 case IPV6_PKTINFO: 2923 case IPV6_HOPLIMIT: 2924 case IPV6_NEXTHOP: 2925 case IPV6_HOPOPTS: 2926 case IPV6_DSTOPTS: 2927 case IPV6_RTHDRDSTOPTS: 2928 case IPV6_RTHDR: 2929 case IPV6_USE_MIN_MTU: 2930 case IPV6_DONTFRAG: 2931 case IPV6_OTCLASS: 2932 case IPV6_TCLASS: 2933 return (ENOPROTOOPT); 2934 } 2935 } 2936 2937 switch (optname) { 2938 #ifdef RFC2292 2939 case IPV6_2292PKTINFO: 2940 #endif 2941 case IPV6_PKTINFO: 2942 { 2943 struct ifnet *ifp = NULL; 2944 struct in6_pktinfo *pktinfo; 2945 2946 if (len != sizeof(struct in6_pktinfo)) 2947 return (EINVAL); 2948 2949 pktinfo = (struct in6_pktinfo *)buf; 2950 2951 /* 2952 * An application can clear any sticky IPV6_PKTINFO option by 2953 * doing a "regular" setsockopt with ipi6_addr being 2954 * in6addr_any and ipi6_ifindex being zero. 2955 * [RFC 3542, Section 6] 2956 */ 2957 if (optname == IPV6_PKTINFO && opt->ip6po_pktinfo && 2958 pktinfo->ipi6_ifindex == 0 && 2959 IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) { 2960 ip6_clearpktopts(opt, optname); 2961 break; 2962 } 2963 2964 if (uproto == IPPROTO_TCP && optname == IPV6_PKTINFO && 2965 sticky && !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) { 2966 return (EINVAL); 2967 } 2968 2969 /* validate the interface index if specified. */ 2970 if (pktinfo->ipi6_ifindex >= if_indexlim) { 2971 return (ENXIO); 2972 } 2973 if (pktinfo->ipi6_ifindex) { 2974 ifp = ifindex2ifnet[pktinfo->ipi6_ifindex]; 2975 if (ifp == NULL) 2976 return (ENXIO); 2977 } 2978 2979 /* 2980 * We store the address anyway, and let in6_selectsrc() 2981 * validate the specified address. This is because ipi6_addr 2982 * may not have enough information about its scope zone, and 2983 * we may need additional information (such as outgoing 2984 * interface or the scope zone of a destination address) to 2985 * disambiguate the scope. 2986 * XXX: the delay of the validation may confuse the 2987 * application when it is used as a sticky option. 2988 */ 2989 if (opt->ip6po_pktinfo == NULL) { 2990 opt->ip6po_pktinfo = malloc(sizeof(*pktinfo), 2991 M_IP6OPT, M_NOWAIT); 2992 if (opt->ip6po_pktinfo == NULL) 2993 return (ENOBUFS); 2994 } 2995 memcpy(opt->ip6po_pktinfo, pktinfo, sizeof(*pktinfo)); 2996 break; 2997 } 2998 2999 #ifdef RFC2292 3000 case IPV6_2292HOPLIMIT: 3001 #endif 3002 case IPV6_HOPLIMIT: 3003 { 3004 int *hlimp; 3005 3006 /* 3007 * RFC 3542 deprecated the usage of sticky IPV6_HOPLIMIT 3008 * to simplify the ordering among hoplimit options. 3009 */ 3010 if (optname == IPV6_HOPLIMIT && sticky) 3011 return (ENOPROTOOPT); 3012 3013 if (len != sizeof(int)) 3014 return (EINVAL); 3015 hlimp = (int *)buf; 3016 if (*hlimp < -1 || *hlimp > 255) 3017 return (EINVAL); 3018 3019 opt->ip6po_hlim = *hlimp; 3020 break; 3021 } 3022 3023 case IPV6_OTCLASS: 3024 if (len != sizeof(u_int8_t)) 3025 return (EINVAL); 3026 3027 opt->ip6po_tclass = *(u_int8_t *)buf; 3028 break; 3029 3030 case IPV6_TCLASS: 3031 { 3032 int tclass; 3033 3034 if (len != sizeof(int)) 3035 return (EINVAL); 3036 tclass = *(int *)buf; 3037 if (tclass < -1 || tclass > 255) 3038 return (EINVAL); 3039 3040 opt->ip6po_tclass = tclass; 3041 break; 3042 } 3043 3044 #ifdef RFC2292 3045 case IPV6_2292NEXTHOP: 3046 #endif 3047 case IPV6_NEXTHOP: 3048 if (!priv) 3049 return (EPERM); 3050 3051 if (len == 0) { /* just remove the option */ 3052 ip6_clearpktopts(opt, IPV6_NEXTHOP); 3053 break; 3054 } 3055 3056 /* check if cmsg_len is large enough for sa_len */ 3057 if (len < sizeof(struct sockaddr) || len < *buf) 3058 return (EINVAL); 3059 3060 switch (((struct sockaddr *)buf)->sa_family) { 3061 case AF_INET6: 3062 { 3063 struct sockaddr_in6 *sa6 = (struct sockaddr_in6 *)buf; 3064 int error; 3065 3066 if (sa6->sin6_len != sizeof(struct sockaddr_in6)) 3067 return (EINVAL); 3068 3069 if (IN6_IS_ADDR_UNSPECIFIED(&sa6->sin6_addr) || 3070 IN6_IS_ADDR_MULTICAST(&sa6->sin6_addr)) { 3071 return (EINVAL); 3072 } 3073 if ((error = sa6_embedscope(sa6, ip6_use_defzone)) 3074 != 0) { 3075 return (error); 3076 } 3077 break; 3078 } 3079 case AF_LINK: /* eventually be supported? */ 3080 default: 3081 return (EAFNOSUPPORT); 3082 } 3083 3084 /* turn off the previous option, then set the new option. */ 3085 ip6_clearpktopts(opt, IPV6_NEXTHOP); 3086 opt->ip6po_nexthop = malloc(*buf, M_IP6OPT, M_NOWAIT); 3087 if (opt->ip6po_nexthop == NULL) 3088 return (ENOBUFS); 3089 memcpy(opt->ip6po_nexthop, buf, *buf); 3090 break; 3091 3092 #ifdef RFC2292 3093 case IPV6_2292HOPOPTS: 3094 #endif 3095 case IPV6_HOPOPTS: 3096 { 3097 struct ip6_hbh *hbh; 3098 int hbhlen; 3099 3100 /* 3101 * XXX: We don't allow a non-privileged user to set ANY HbH 3102 * options, since per-option restriction has too much 3103 * overhead. 3104 */ 3105 if (!priv) 3106 return (EPERM); 3107 3108 if (len == 0) { 3109 ip6_clearpktopts(opt, IPV6_HOPOPTS); 3110 break; /* just remove the option */ 3111 } 3112 3113 /* message length validation */ 3114 if (len < sizeof(struct ip6_hbh)) 3115 return (EINVAL); 3116 hbh = (struct ip6_hbh *)buf; 3117 hbhlen = (hbh->ip6h_len + 1) << 3; 3118 if (len != hbhlen) 3119 return (EINVAL); 3120 3121 /* turn off the previous option, then set the new option. */ 3122 ip6_clearpktopts(opt, IPV6_HOPOPTS); 3123 opt->ip6po_hbh = malloc(hbhlen, M_IP6OPT, M_NOWAIT); 3124 if (opt->ip6po_hbh == NULL) 3125 return (ENOBUFS); 3126 memcpy(opt->ip6po_hbh, hbh, hbhlen); 3127 3128 break; 3129 } 3130 3131 #ifdef RFC2292 3132 case IPV6_2292DSTOPTS: 3133 #endif 3134 case IPV6_DSTOPTS: 3135 case IPV6_RTHDRDSTOPTS: 3136 { 3137 struct ip6_dest *dest, **newdest = NULL; 3138 int destlen; 3139 3140 if (!priv) /* XXX: see the comment for IPV6_HOPOPTS */ 3141 return (EPERM); 3142 3143 if (len == 0) { 3144 ip6_clearpktopts(opt, optname); 3145 break; /* just remove the option */ 3146 } 3147 3148 /* message length validation */ 3149 if (len < sizeof(struct ip6_dest)) 3150 return (EINVAL); 3151 dest = (struct ip6_dest *)buf; 3152 destlen = (dest->ip6d_len + 1) << 3; 3153 if (len != destlen) 3154 return (EINVAL); 3155 /* 3156 * Determine the position that the destination options header 3157 * should be inserted; before or after the routing header. 3158 */ 3159 switch (optname) { 3160 case IPV6_2292DSTOPTS: 3161 /* 3162 * The old advanced API is ambiguous on this point. 3163 * Our approach is to determine the position based 3164 * according to the existence of a routing header. 3165 * Note, however, that this depends on the order of the 3166 * extension headers in the ancillary data; the 1st 3167 * part of the destination options header must appear 3168 * before the routing header in the ancillary data, 3169 * too. 3170 * RFC3542 solved the ambiguity by introducing 3171 * separate ancillary data or option types. 3172 */ 3173 if (opt->ip6po_rthdr == NULL) 3174 newdest = &opt->ip6po_dest1; 3175 else 3176 newdest = &opt->ip6po_dest2; 3177 break; 3178 case IPV6_RTHDRDSTOPTS: 3179 newdest = &opt->ip6po_dest1; 3180 break; 3181 case IPV6_DSTOPTS: 3182 newdest = &opt->ip6po_dest2; 3183 break; 3184 } 3185 3186 /* turn off the previous option, then set the new option. */ 3187 ip6_clearpktopts(opt, optname); 3188 *newdest = malloc(destlen, M_IP6OPT, M_NOWAIT); 3189 if (*newdest == NULL) 3190 return (ENOBUFS); 3191 memcpy(*newdest, dest, destlen); 3192 3193 break; 3194 } 3195 3196 #ifdef RFC2292 3197 case IPV6_2292RTHDR: 3198 #endif 3199 case IPV6_RTHDR: 3200 { 3201 struct ip6_rthdr *rth; 3202 int rthlen; 3203 3204 if (len == 0) { 3205 ip6_clearpktopts(opt, IPV6_RTHDR); 3206 break; /* just remove the option */ 3207 } 3208 3209 /* message length validation */ 3210 if (len < sizeof(struct ip6_rthdr)) 3211 return (EINVAL); 3212 rth = (struct ip6_rthdr *)buf; 3213 rthlen = (rth->ip6r_len + 1) << 3; 3214 if (len != rthlen) 3215 return (EINVAL); 3216 switch (rth->ip6r_type) { 3217 case IPV6_RTHDR_TYPE_0: 3218 if (rth->ip6r_len == 0) /* must contain one addr */ 3219 return (EINVAL); 3220 if (rth->ip6r_len % 2) /* length must be even */ 3221 return (EINVAL); 3222 if (rth->ip6r_len / 2 != rth->ip6r_segleft) 3223 return (EINVAL); 3224 break; 3225 default: 3226 return (EINVAL); /* not supported */ 3227 } 3228 /* turn off the previous option */ 3229 ip6_clearpktopts(opt, IPV6_RTHDR); 3230 opt->ip6po_rthdr = malloc(rthlen, M_IP6OPT, M_NOWAIT); 3231 if (opt->ip6po_rthdr == NULL) 3232 return (ENOBUFS); 3233 memcpy(opt->ip6po_rthdr, rth, rthlen); 3234 break; 3235 } 3236 3237 case IPV6_USE_MIN_MTU: 3238 if (len != sizeof(int)) 3239 return (EINVAL); 3240 minmtupolicy = *(int *)buf; 3241 if (minmtupolicy != IP6PO_MINMTU_MCASTONLY && 3242 minmtupolicy != IP6PO_MINMTU_DISABLE && 3243 minmtupolicy != IP6PO_MINMTU_ALL) { 3244 return (EINVAL); 3245 } 3246 opt->ip6po_minmtu = minmtupolicy; 3247 break; 3248 3249 case IPV6_DONTFRAG: 3250 if (len != sizeof(int)) 3251 return (EINVAL); 3252 3253 if (uproto == IPPROTO_TCP || *(int *)buf == 0) { 3254 /* 3255 * we ignore this option for TCP sockets. 3256 * (RFC3542 leaves this case unspecified.) 3257 */ 3258 opt->ip6po_flags &= ~IP6PO_DONTFRAG; 3259 } else 3260 opt->ip6po_flags |= IP6PO_DONTFRAG; 3261 break; 3262 3263 default: 3264 return (ENOPROTOOPT); 3265 } /* end of switch */ 3266 3267 return (0); 3268 } 3269 3270 /* 3271 * Routine called from ip6_output() to loop back a copy of an IP6 multicast 3272 * packet to the input queue of a specified interface. Note that this 3273 * calls the output routine of the loopback "driver", but with an interface 3274 * pointer that might NOT be lo0ifp -- easier than replicating that code here. 3275 */ 3276 void 3277 ip6_mloopback(ifp, m, dst) 3278 struct ifnet *ifp; 3279 struct mbuf *m; 3280 struct sockaddr_in6 *dst; 3281 { 3282 struct mbuf *copym; 3283 struct ip6_hdr *ip6; 3284 3285 copym = m_copy(m, 0, M_COPYALL); 3286 if (copym == NULL) 3287 return; 3288 3289 /* 3290 * Make sure to deep-copy IPv6 header portion in case the data 3291 * is in an mbuf cluster, so that we can safely override the IPv6 3292 * header portion later. 3293 */ 3294 if ((copym->m_flags & M_EXT) != 0 || 3295 copym->m_len < sizeof(struct ip6_hdr)) { 3296 copym = m_pullup(copym, sizeof(struct ip6_hdr)); 3297 if (copym == NULL) 3298 return; 3299 } 3300 3301 #ifdef DIAGNOSTIC 3302 if (copym->m_len < sizeof(*ip6)) { 3303 m_freem(copym); 3304 return; 3305 } 3306 #endif 3307 3308 ip6 = mtod(copym, struct ip6_hdr *); 3309 /* 3310 * clear embedded scope identifiers if necessary. 3311 * in6_clearscope will touch the addresses only when necessary. 3312 */ 3313 in6_clearscope(&ip6->ip6_src); 3314 in6_clearscope(&ip6->ip6_dst); 3315 3316 (void)looutput(ifp, copym, (struct sockaddr *)dst, NULL); 3317 } 3318 3319 /* 3320 * Chop IPv6 header off from the payload. 3321 */ 3322 static int 3323 ip6_splithdr(m, exthdrs) 3324 struct mbuf *m; 3325 struct ip6_exthdrs *exthdrs; 3326 { 3327 struct mbuf *mh; 3328 struct ip6_hdr *ip6; 3329 3330 ip6 = mtod(m, struct ip6_hdr *); 3331 if (m->m_len > sizeof(*ip6)) { 3332 MGETHDR(mh, M_DONTWAIT, MT_HEADER); 3333 if (mh == 0) { 3334 m_freem(m); 3335 return ENOBUFS; 3336 } 3337 M_MOVE_PKTHDR(mh, m); 3338 MH_ALIGN(mh, sizeof(*ip6)); 3339 m->m_len -= sizeof(*ip6); 3340 m->m_data += sizeof(*ip6); 3341 mh->m_next = m; 3342 m = mh; 3343 m->m_len = sizeof(*ip6); 3344 bcopy((caddr_t)ip6, mtod(m, caddr_t), sizeof(*ip6)); 3345 } 3346 exthdrs->ip6e_ip6 = m; 3347 return 0; 3348 } 3349 3350 /* 3351 * Compute IPv6 extension header length. 3352 */ 3353 int 3354 ip6_optlen(in6p) 3355 struct in6pcb *in6p; 3356 { 3357 int len; 3358 3359 if (!in6p->in6p_outputopts) 3360 return 0; 3361 3362 len = 0; 3363 #define elen(x) \ 3364 (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0) 3365 3366 len += elen(in6p->in6p_outputopts->ip6po_hbh); 3367 len += elen(in6p->in6p_outputopts->ip6po_dest1); 3368 len += elen(in6p->in6p_outputopts->ip6po_rthdr); 3369 len += elen(in6p->in6p_outputopts->ip6po_dest2); 3370 return len; 3371 #undef elen 3372 } 3373
Indexes created Tue Sep 23 02:09:52 GMT 2025