ip6_output.c revision 1.99
1 /* $NetBSD: ip6_output.c,v 1.99 2006/07/08 19:58:40 rpaulo 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 1.99 2006/07/08 19:58:40 rpaulo 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/ip6_var.h> 95 #include <netinet6/in6_pcb.h> 96 #include <netinet6/nd6.h> 97 #include <netinet6/ip6protosw.h> 98 #include <netinet6/scope6_var.h> 99 100 #ifdef IPSEC 101 #include <netinet6/ipsec.h> 102 #include <netkey/key.h> 103 #endif /* IPSEC */ 104 105 #include <net/net_osdep.h> 106 107 #ifdef PFIL_HOOKS 108 extern struct pfil_head inet6_pfil_hook; /* XXX */ 109 #endif 110 111 struct ip6_exthdrs { 112 struct mbuf *ip6e_ip6; 113 struct mbuf *ip6e_hbh; 114 struct mbuf *ip6e_dest1; 115 struct mbuf *ip6e_rthdr; 116 struct mbuf *ip6e_dest2; 117 }; 118 119 static int ip6_pcbopt __P((int, u_char *, int, struct ip6_pktopts **, 120 int, int)); 121 static int ip6_getpcbopt __P((struct ip6_pktopts *, int, struct mbuf **)); 122 static int ip6_setpktopt __P((int, u_char *, int, struct ip6_pktopts *, int, 123 int, int, int)); 124 static int ip6_setmoptions __P((int, struct ip6_moptions **, struct mbuf *)); 125 static int ip6_getmoptions __P((int, struct ip6_moptions *, struct mbuf **)); 126 static int ip6_copyexthdr __P((struct mbuf **, caddr_t, int)); 127 static int ip6_insertfraghdr __P((struct mbuf *, struct mbuf *, int, 128 struct ip6_frag **)); 129 static int ip6_insert_jumboopt __P((struct ip6_exthdrs *, u_int32_t)); 130 static int ip6_splithdr __P((struct mbuf *, struct ip6_exthdrs *)); 131 static int ip6_getpmtu __P((struct route_in6 *, struct route_in6 *, 132 struct ifnet *, struct in6_addr *, u_long *, int *)); 133 static int copypktopts __P((struct ip6_pktopts *, struct ip6_pktopts *, int)); 134 135 #ifdef RFC2292 136 static int ip6_pcbopts __P((struct ip6_pktopts **, struct mbuf *, 137 struct socket *)); 138 #endif 139 140 #define IN6_NEED_CHECKSUM(ifp, csum_flags) \ 141 (__predict_true(((ifp)->if_flags & IFF_LOOPBACK) == 0 || \ 142 (((csum_flags) & M_CSUM_UDPv6) != 0 && udp_do_loopback_cksum) || \ 143 (((csum_flags) & M_CSUM_TCPv6) != 0 && tcp_do_loopback_cksum))) 144 145 /* 146 * IP6 output. The packet in mbuf chain m contains a skeletal IP6 147 * header (with pri, len, nxt, hlim, src, dst). 148 * This function may modify ver and hlim only. 149 * The mbuf chain containing the packet will be freed. 150 * The mbuf opt, if present, will not be freed. 151 * 152 * type of "mtu": rt_rmx.rmx_mtu is u_long, ifnet.ifr_mtu is int, and 153 * nd_ifinfo.linkmtu is u_int32_t. so we use u_long to hold largest one, 154 * which is rt_rmx.rmx_mtu. 155 */ 156 int 157 ip6_output(m0, opt, ro, flags, im6o, so, ifpp) 158 struct mbuf *m0; 159 struct ip6_pktopts *opt; 160 struct route_in6 *ro; 161 int flags; 162 struct ip6_moptions *im6o; 163 struct socket *so; 164 struct ifnet **ifpp; /* XXX: just for statistics */ 165 { 166 struct ip6_hdr *ip6, *mhip6; 167 struct ifnet *ifp, *origifp; 168 struct mbuf *m = m0; 169 int hlen, tlen, len, off; 170 struct route_in6 ip6route; 171 struct rtentry *rt = NULL; 172 struct sockaddr_in6 *dst, src_sa, dst_sa; 173 int error = 0; 174 struct in6_ifaddr *ia = NULL; 175 u_long mtu; 176 int alwaysfrag, dontfrag; 177 u_int32_t optlen = 0, plen = 0, unfragpartlen = 0; 178 struct ip6_exthdrs exthdrs; 179 struct in6_addr finaldst, src0, dst0; 180 u_int32_t zone; 181 struct route_in6 *ro_pmtu = NULL; 182 int hdrsplit = 0; 183 int needipsec = 0; 184 #ifdef IPSEC 185 int needipsectun = 0; 186 struct secpolicy *sp = NULL; 187 188 ip6 = mtod(m, struct ip6_hdr *); 189 #endif /* IPSEC */ 190 191 M_CSUM_DATA_IPv6_HL_SET(m->m_pkthdr.csum_data, sizeof(struct ip6_hdr)); 192 193 #define MAKE_EXTHDR(hp, mp) \ 194 do { \ 195 if (hp) { \ 196 struct ip6_ext *eh = (struct ip6_ext *)(hp); \ 197 error = ip6_copyexthdr((mp), (caddr_t)(hp), \ 198 ((eh)->ip6e_len + 1) << 3); \ 199 if (error) \ 200 goto freehdrs; \ 201 } \ 202 } while (/*CONSTCOND*/ 0) 203 204 bzero(&exthdrs, sizeof(exthdrs)); 205 if (opt) { 206 /* Hop-by-Hop options header */ 207 MAKE_EXTHDR(opt->ip6po_hbh, &exthdrs.ip6e_hbh); 208 /* Destination options header(1st part) */ 209 MAKE_EXTHDR(opt->ip6po_dest1, &exthdrs.ip6e_dest1); 210 /* Routing header */ 211 MAKE_EXTHDR(opt->ip6po_rthdr, &exthdrs.ip6e_rthdr); 212 /* Destination options header(2nd part) */ 213 MAKE_EXTHDR(opt->ip6po_dest2, &exthdrs.ip6e_dest2); 214 } 215 216 #ifdef IPSEC 217 if ((flags & IPV6_FORWARDING) != 0) { 218 needipsec = 0; 219 goto skippolicycheck; 220 } 221 222 /* get a security policy for this packet */ 223 if (so == NULL) 224 sp = ipsec6_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND, 0, &error); 225 else { 226 if (IPSEC_PCB_SKIP_IPSEC(sotoinpcb_hdr(so)->inph_sp, 227 IPSEC_DIR_OUTBOUND)) { 228 needipsec = 0; 229 goto skippolicycheck; 230 } 231 sp = ipsec6_getpolicybysock(m, IPSEC_DIR_OUTBOUND, so, &error); 232 } 233 234 if (sp == NULL) { 235 ipsec6stat.out_inval++; 236 goto freehdrs; 237 } 238 239 error = 0; 240 241 /* check policy */ 242 switch (sp->policy) { 243 case IPSEC_POLICY_DISCARD: 244 /* 245 * This packet is just discarded. 246 */ 247 ipsec6stat.out_polvio++; 248 goto freehdrs; 249 250 case IPSEC_POLICY_BYPASS: 251 case IPSEC_POLICY_NONE: 252 /* no need to do IPsec. */ 253 needipsec = 0; 254 break; 255 256 case IPSEC_POLICY_IPSEC: 257 if (sp->req == NULL) { 258 /* XXX should be panic ? */ 259 printf("ip6_output: No IPsec request specified.\n"); 260 error = EINVAL; 261 goto freehdrs; 262 } 263 needipsec = 1; 264 break; 265 266 case IPSEC_POLICY_ENTRUST: 267 default: 268 printf("ip6_output: Invalid policy found. %d\n", sp->policy); 269 } 270 271 skippolicycheck:; 272 #endif /* IPSEC */ 273 274 if (needipsec && 275 (m->m_pkthdr.csum_flags & (M_CSUM_UDPv6|M_CSUM_TCPv6)) != 0) { 276 in6_delayed_cksum(m); 277 m->m_pkthdr.csum_flags &= ~(M_CSUM_UDPv6|M_CSUM_TCPv6); 278 } 279 280 /* 281 * Calculate the total length of the extension header chain. 282 * Keep the length of the unfragmentable part for fragmentation. 283 */ 284 optlen = 0; 285 if (exthdrs.ip6e_hbh) optlen += exthdrs.ip6e_hbh->m_len; 286 if (exthdrs.ip6e_dest1) optlen += exthdrs.ip6e_dest1->m_len; 287 if (exthdrs.ip6e_rthdr) optlen += exthdrs.ip6e_rthdr->m_len; 288 unfragpartlen = optlen + sizeof(struct ip6_hdr); 289 /* NOTE: we don't add AH/ESP length here. do that later. */ 290 if (exthdrs.ip6e_dest2) optlen += exthdrs.ip6e_dest2->m_len; 291 292 /* 293 * If we need IPsec, or there is at least one extension header, 294 * separate IP6 header from the payload. 295 */ 296 if ((needipsec || optlen) && !hdrsplit) { 297 if ((error = ip6_splithdr(m, &exthdrs)) != 0) { 298 m = NULL; 299 goto freehdrs; 300 } 301 m = exthdrs.ip6e_ip6; 302 hdrsplit++; 303 } 304 305 /* adjust pointer */ 306 ip6 = mtod(m, struct ip6_hdr *); 307 308 /* adjust mbuf packet header length */ 309 m->m_pkthdr.len += optlen; 310 plen = m->m_pkthdr.len - sizeof(*ip6); 311 312 /* If this is a jumbo payload, insert a jumbo payload option. */ 313 if (plen > IPV6_MAXPACKET) { 314 if (!hdrsplit) { 315 if ((error = ip6_splithdr(m, &exthdrs)) != 0) { 316 m = NULL; 317 goto freehdrs; 318 } 319 m = exthdrs.ip6e_ip6; 320 hdrsplit++; 321 } 322 /* adjust pointer */ 323 ip6 = mtod(m, struct ip6_hdr *); 324 if ((error = ip6_insert_jumboopt(&exthdrs, plen)) != 0) 325 goto freehdrs; 326 optlen += 8; /* XXX JUMBOOPTLEN */ 327 ip6->ip6_plen = 0; 328 } else 329 ip6->ip6_plen = htons(plen); 330 331 /* 332 * Concatenate headers and fill in next header fields. 333 * Here we have, on "m" 334 * IPv6 payload 335 * and we insert headers accordingly. Finally, we should be getting: 336 * IPv6 hbh dest1 rthdr ah* [esp* dest2 payload] 337 * 338 * during the header composing process, "m" points to IPv6 header. 339 * "mprev" points to an extension header prior to esp. 340 */ 341 { 342 u_char *nexthdrp = &ip6->ip6_nxt; 343 struct mbuf *mprev = m; 344 345 /* 346 * we treat dest2 specially. this makes IPsec processing 347 * much easier. the goal here is to make mprev point the 348 * mbuf prior to dest2. 349 * 350 * result: IPv6 dest2 payload 351 * m and mprev will point to IPv6 header. 352 */ 353 if (exthdrs.ip6e_dest2) { 354 if (!hdrsplit) 355 panic("assumption failed: hdr not split"); 356 exthdrs.ip6e_dest2->m_next = m->m_next; 357 m->m_next = exthdrs.ip6e_dest2; 358 *mtod(exthdrs.ip6e_dest2, u_char *) = ip6->ip6_nxt; 359 ip6->ip6_nxt = IPPROTO_DSTOPTS; 360 } 361 362 #define MAKE_CHAIN(m, mp, p, i)\ 363 do {\ 364 if (m) {\ 365 if (!hdrsplit) \ 366 panic("assumption failed: hdr not split"); \ 367 *mtod((m), u_char *) = *(p);\ 368 *(p) = (i);\ 369 p = mtod((m), u_char *);\ 370 (m)->m_next = (mp)->m_next;\ 371 (mp)->m_next = (m);\ 372 (mp) = (m);\ 373 }\ 374 } while (/*CONSTCOND*/ 0) 375 /* 376 * result: IPv6 hbh dest1 rthdr dest2 payload 377 * m will point to IPv6 header. mprev will point to the 378 * extension header prior to dest2 (rthdr in the above case). 379 */ 380 MAKE_CHAIN(exthdrs.ip6e_hbh, mprev, nexthdrp, IPPROTO_HOPOPTS); 381 MAKE_CHAIN(exthdrs.ip6e_dest1, mprev, nexthdrp, 382 IPPROTO_DSTOPTS); 383 MAKE_CHAIN(exthdrs.ip6e_rthdr, mprev, nexthdrp, 384 IPPROTO_ROUTING); 385 386 M_CSUM_DATA_IPv6_HL_SET(m->m_pkthdr.csum_data, 387 sizeof(struct ip6_hdr) + optlen); 388 389 #ifdef IPSEC 390 if (!needipsec) 391 goto skip_ipsec2; 392 393 /* 394 * pointers after IPsec headers are not valid any more. 395 * other pointers need a great care too. 396 * (IPsec routines should not mangle mbufs prior to AH/ESP) 397 */ 398 exthdrs.ip6e_dest2 = NULL; 399 400 { 401 struct ip6_rthdr *rh = NULL; 402 int segleft_org = 0; 403 struct ipsec_output_state state; 404 405 if (exthdrs.ip6e_rthdr) { 406 rh = mtod(exthdrs.ip6e_rthdr, struct ip6_rthdr *); 407 segleft_org = rh->ip6r_segleft; 408 rh->ip6r_segleft = 0; 409 } 410 411 bzero(&state, sizeof(state)); 412 state.m = m; 413 error = ipsec6_output_trans(&state, nexthdrp, mprev, sp, flags, 414 &needipsectun); 415 m = state.m; 416 if (error) { 417 /* mbuf is already reclaimed in ipsec6_output_trans. */ 418 m = NULL; 419 switch (error) { 420 case EHOSTUNREACH: 421 case ENETUNREACH: 422 case EMSGSIZE: 423 case ENOBUFS: 424 case ENOMEM: 425 break; 426 default: 427 printf("ip6_output (ipsec): error code %d\n", error); 428 /* FALLTHROUGH */ 429 case ENOENT: 430 /* don't show these error codes to the user */ 431 error = 0; 432 break; 433 } 434 goto bad; 435 } 436 if (exthdrs.ip6e_rthdr) { 437 /* ah6_output doesn't modify mbuf chain */ 438 rh->ip6r_segleft = segleft_org; 439 } 440 } 441 skip_ipsec2:; 442 #endif 443 } 444 445 /* 446 * If there is a routing header, replace destination address field 447 * with the first hop of the routing header. 448 */ 449 if (exthdrs.ip6e_rthdr) { 450 struct ip6_rthdr *rh; 451 struct ip6_rthdr0 *rh0; 452 struct in6_addr *addr; 453 struct sockaddr_in6 sa; 454 455 rh = (struct ip6_rthdr *)(mtod(exthdrs.ip6e_rthdr, 456 struct ip6_rthdr *)); 457 finaldst = ip6->ip6_dst; 458 switch (rh->ip6r_type) { 459 case IPV6_RTHDR_TYPE_0: 460 rh0 = (struct ip6_rthdr0 *)rh; 461 addr = (struct in6_addr *)(rh0 + 1); 462 463 /* 464 * construct a sockaddr_in6 form of 465 * the first hop. 466 * 467 * XXX: we may not have enough 468 * information about its scope zone; 469 * there is no standard API to pass 470 * the information from the 471 * application. 472 */ 473 bzero(&sa, sizeof(sa)); 474 sa.sin6_family = AF_INET6; 475 sa.sin6_len = sizeof(sa); 476 sa.sin6_addr = addr[0]; 477 if ((error = sa6_embedscope(&sa, 478 ip6_use_defzone)) != 0) { 479 goto bad; 480 } 481 ip6->ip6_dst = sa.sin6_addr; 482 (void)memmove(&addr[0], &addr[1], 483 sizeof(struct in6_addr) * 484 (rh0->ip6r0_segleft - 1)); 485 addr[rh0->ip6r0_segleft - 1] = finaldst; 486 /* XXX */ 487 in6_clearscope(addr + rh0->ip6r0_segleft - 1); 488 break; 489 default: /* is it possible? */ 490 error = EINVAL; 491 goto bad; 492 } 493 } 494 495 /* Source address validation */ 496 if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) && 497 (flags & IPV6_UNSPECSRC) == 0) { 498 error = EOPNOTSUPP; 499 ip6stat.ip6s_badscope++; 500 goto bad; 501 } 502 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) { 503 error = EOPNOTSUPP; 504 ip6stat.ip6s_badscope++; 505 goto bad; 506 } 507 508 ip6stat.ip6s_localout++; 509 510 /* 511 * Route packet. 512 */ 513 /* initialize cached route */ 514 if (ro == 0) { 515 ro = &ip6route; 516 bzero((caddr_t)ro, sizeof(*ro)); 517 } 518 ro_pmtu = ro; 519 if (opt && opt->ip6po_rthdr) 520 ro = &opt->ip6po_route; 521 dst = (struct sockaddr_in6 *)&ro->ro_dst; 522 523 /* 524 * if specified, try to fill in the traffic class field. 525 * do not override if a non-zero value is already set. 526 * we check the diffserv field and the ecn field separately. 527 */ 528 if (opt && opt->ip6po_tclass >= 0) { 529 int mask = 0; 530 531 if ((ip6->ip6_flow & htonl(0xfc << 20)) == 0) 532 mask |= 0xfc; 533 if ((ip6->ip6_flow & htonl(0x03 << 20)) == 0) 534 mask |= 0x03; 535 if (mask != 0) 536 ip6->ip6_flow |= htonl((opt->ip6po_tclass & mask) << 20); 537 } 538 539 /* fill in or override the hop limit field, if necessary. */ 540 if (opt && opt->ip6po_hlim != -1) 541 ip6->ip6_hlim = opt->ip6po_hlim & 0xff; 542 else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) { 543 if (im6o != NULL) 544 ip6->ip6_hlim = im6o->im6o_multicast_hlim; 545 else 546 ip6->ip6_hlim = ip6_defmcasthlim; 547 } 548 549 #ifdef IPSEC 550 if (needipsec && needipsectun) { 551 struct ipsec_output_state state; 552 553 /* 554 * All the extension headers will become inaccessible 555 * (since they can be encrypted). 556 * Don't panic, we need no more updates to extension headers 557 * on inner IPv6 packet (since they are now encapsulated). 558 * 559 * IPv6 [ESP|AH] IPv6 [extension headers] payload 560 */ 561 bzero(&exthdrs, sizeof(exthdrs)); 562 exthdrs.ip6e_ip6 = m; 563 564 bzero(&state, sizeof(state)); 565 state.m = m; 566 state.ro = (struct route *)ro; 567 state.dst = (struct sockaddr *)dst; 568 569 error = ipsec6_output_tunnel(&state, sp, flags); 570 571 m = state.m; 572 ro_pmtu = ro = (struct route_in6 *)state.ro; 573 dst = (struct sockaddr_in6 *)state.dst; 574 if (error) { 575 /* mbuf is already reclaimed in ipsec6_output_tunnel. */ 576 m0 = m = NULL; 577 m = NULL; 578 switch (error) { 579 case EHOSTUNREACH: 580 case ENETUNREACH: 581 case EMSGSIZE: 582 case ENOBUFS: 583 case ENOMEM: 584 break; 585 default: 586 printf("ip6_output (ipsec): error code %d\n", error); 587 /* FALLTHROUGH */ 588 case ENOENT: 589 /* don't show these error codes to the user */ 590 error = 0; 591 break; 592 } 593 goto bad; 594 } 595 596 exthdrs.ip6e_ip6 = m; 597 } 598 #endif /* IPSEC */ 599 600 /* adjust pointer */ 601 ip6 = mtod(m, struct ip6_hdr *); 602 603 bzero(&dst_sa, sizeof(dst_sa)); 604 dst_sa.sin6_family = AF_INET6; 605 dst_sa.sin6_len = sizeof(dst_sa); 606 dst_sa.sin6_addr = ip6->ip6_dst; 607 if ((error = in6_selectroute(&dst_sa, opt, im6o, ro, &ifp, &rt, 0)) 608 != 0) { 609 switch (error) { 610 case EHOSTUNREACH: 611 ip6stat.ip6s_noroute++; 612 break; 613 case EADDRNOTAVAIL: 614 default: 615 break; /* XXX statistics? */ 616 } 617 if (ifp != NULL) 618 in6_ifstat_inc(ifp, ifs6_out_discard); 619 goto bad; 620 } 621 if (rt == NULL) { 622 /* 623 * If in6_selectroute() does not return a route entry, 624 * dst may not have been updated. 625 */ 626 *dst = dst_sa; /* XXX */ 627 } 628 629 /* 630 * then rt (for unicast) and ifp must be non-NULL valid values. 631 */ 632 if ((flags & IPV6_FORWARDING) == 0) { 633 /* XXX: the FORWARDING flag can be set for mrouting. */ 634 in6_ifstat_inc(ifp, ifs6_out_request); 635 } 636 if (rt != NULL) { 637 ia = (struct in6_ifaddr *)(rt->rt_ifa); 638 rt->rt_use++; 639 } 640 641 /* 642 * The outgoing interface must be in the zone of source and 643 * destination addresses. We should use ia_ifp to support the 644 * case of sending packets to an address of our own. 645 */ 646 if (ia != NULL && ia->ia_ifp) 647 origifp = ia->ia_ifp; 648 else 649 origifp = ifp; 650 651 src0 = ip6->ip6_src; 652 if (in6_setscope(&src0, origifp, &zone)) 653 goto badscope; 654 bzero(&src_sa, sizeof(src_sa)); 655 src_sa.sin6_family = AF_INET6; 656 src_sa.sin6_len = sizeof(src_sa); 657 src_sa.sin6_addr = ip6->ip6_src; 658 if (sa6_recoverscope(&src_sa) || zone != src_sa.sin6_scope_id) 659 goto badscope; 660 661 dst0 = ip6->ip6_dst; 662 if (in6_setscope(&dst0, origifp, &zone)) 663 goto badscope; 664 /* re-initialize to be sure */ 665 bzero(&dst_sa, sizeof(dst_sa)); 666 dst_sa.sin6_family = AF_INET6; 667 dst_sa.sin6_len = sizeof(dst_sa); 668 dst_sa.sin6_addr = ip6->ip6_dst; 669 if (sa6_recoverscope(&dst_sa) || zone != dst_sa.sin6_scope_id) 670 goto badscope; 671 672 /* scope check is done. */ 673 goto routefound; 674 675 badscope: 676 ip6stat.ip6s_badscope++; 677 in6_ifstat_inc(origifp, ifs6_out_discard); 678 if (error == 0) 679 error = EHOSTUNREACH; /* XXX */ 680 goto bad; 681 682 routefound: 683 if (rt && !IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) { 684 if (opt && opt->ip6po_nextroute.ro_rt) { 685 /* 686 * The nexthop is explicitly specified by the 687 * application. We assume the next hop is an IPv6 688 * address. 689 */ 690 dst = (struct sockaddr_in6 *)opt->ip6po_nexthop; 691 } else if ((rt->rt_flags & RTF_GATEWAY)) 692 dst = (struct sockaddr_in6 *)rt->rt_gateway; 693 } 694 695 /* 696 * XXXXXX: original code follows: 697 */ 698 if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) 699 m->m_flags &= ~(M_BCAST | M_MCAST); /* just in case */ 700 else { 701 struct in6_multi *in6m; 702 703 m->m_flags = (m->m_flags & ~M_BCAST) | M_MCAST; 704 705 in6_ifstat_inc(ifp, ifs6_out_mcast); 706 707 /* 708 * Confirm that the outgoing interface supports multicast. 709 */ 710 if (!(ifp->if_flags & IFF_MULTICAST)) { 711 ip6stat.ip6s_noroute++; 712 in6_ifstat_inc(ifp, ifs6_out_discard); 713 error = ENETUNREACH; 714 goto bad; 715 } 716 717 IN6_LOOKUP_MULTI(ip6->ip6_dst, ifp, in6m); 718 if (in6m != NULL && 719 (im6o == NULL || im6o->im6o_multicast_loop)) { 720 /* 721 * If we belong to the destination multicast group 722 * on the outgoing interface, and the caller did not 723 * forbid loopback, loop back a copy. 724 */ 725 ip6_mloopback(ifp, m, dst); 726 } else { 727 /* 728 * If we are acting as a multicast router, perform 729 * multicast forwarding as if the packet had just 730 * arrived on the interface to which we are about 731 * to send. The multicast forwarding function 732 * recursively calls this function, using the 733 * IPV6_FORWARDING flag to prevent infinite recursion. 734 * 735 * Multicasts that are looped back by ip6_mloopback(), 736 * above, will be forwarded by the ip6_input() routine, 737 * if necessary. 738 */ 739 if (ip6_mrouter && (flags & IPV6_FORWARDING) == 0) { 740 if (ip6_mforward(ip6, ifp, m) != 0) { 741 m_freem(m); 742 goto done; 743 } 744 } 745 } 746 /* 747 * Multicasts with a hoplimit of zero may be looped back, 748 * above, but must not be transmitted on a network. 749 * Also, multicasts addressed to the loopback interface 750 * are not sent -- the above call to ip6_mloopback() will 751 * loop back a copy if this host actually belongs to the 752 * destination group on the loopback interface. 753 */ 754 if (ip6->ip6_hlim == 0 || (ifp->if_flags & IFF_LOOPBACK) || 755 IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst)) { 756 m_freem(m); 757 goto done; 758 } 759 } 760 761 /* 762 * Fill the outgoing inteface to tell the upper layer 763 * to increment per-interface statistics. 764 */ 765 if (ifpp) 766 *ifpp = ifp; 767 768 /* Determine path MTU. */ 769 if ((error = ip6_getpmtu(ro_pmtu, ro, ifp, &finaldst, &mtu, 770 &alwaysfrag)) != 0) 771 goto bad; 772 #ifdef IPSEC 773 if (needipsectun) 774 mtu = IPV6_MMTU; 775 #endif 776 777 /* 778 * The caller of this function may specify to use the minimum MTU 779 * in some cases. 780 * An advanced API option (IPV6_USE_MIN_MTU) can also override MTU 781 * setting. The logic is a bit complicated; by default, unicast 782 * packets will follow path MTU while multicast packets will be sent at 783 * the minimum MTU. If IP6PO_MINMTU_ALL is specified, all packets 784 * including unicast ones will be sent at the minimum MTU. Multicast 785 * packets will always be sent at the minimum MTU unless 786 * IP6PO_MINMTU_DISABLE is explicitly specified. 787 * See RFC 3542 for more details. 788 */ 789 if (mtu > IPV6_MMTU) { 790 if ((flags & IPV6_MINMTU)) 791 mtu = IPV6_MMTU; 792 else if (opt && opt->ip6po_minmtu == IP6PO_MINMTU_ALL) 793 mtu = IPV6_MMTU; 794 else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) && 795 (opt == NULL || 796 opt->ip6po_minmtu != IP6PO_MINMTU_DISABLE)) { 797 mtu = IPV6_MMTU; 798 } 799 } 800 801 /* 802 * clear embedded scope identifiers if necessary. 803 * in6_clearscope will touch the addresses only when necessary. 804 */ 805 in6_clearscope(&ip6->ip6_src); 806 in6_clearscope(&ip6->ip6_dst); 807 808 /* 809 * If the outgoing packet contains a hop-by-hop options header, 810 * it must be examined and processed even by the source node. 811 * (RFC 2460, section 4.) 812 */ 813 if (exthdrs.ip6e_hbh) { 814 struct ip6_hbh *hbh = mtod(exthdrs.ip6e_hbh, struct ip6_hbh *); 815 u_int32_t dummy1; /* XXX unused */ 816 u_int32_t dummy2; /* XXX unused */ 817 818 /* 819 * XXX: if we have to send an ICMPv6 error to the sender, 820 * we need the M_LOOP flag since icmp6_error() expects 821 * the IPv6 and the hop-by-hop options header are 822 * continuous unless the flag is set. 823 */ 824 m->m_flags |= M_LOOP; 825 m->m_pkthdr.rcvif = ifp; 826 if (ip6_process_hopopts(m, (u_int8_t *)(hbh + 1), 827 ((hbh->ip6h_len + 1) << 3) - sizeof(struct ip6_hbh), 828 &dummy1, &dummy2) < 0) { 829 /* m was already freed at this point */ 830 error = EINVAL;/* better error? */ 831 goto done; 832 } 833 m->m_flags &= ~M_LOOP; /* XXX */ 834 m->m_pkthdr.rcvif = NULL; 835 } 836 837 #ifdef PFIL_HOOKS 838 /* 839 * Run through list of hooks for output packets. 840 */ 841 if ((error = pfil_run_hooks(&inet6_pfil_hook, &m, ifp, PFIL_OUT)) != 0) 842 goto done; 843 if (m == NULL) 844 goto done; 845 ip6 = mtod(m, struct ip6_hdr *); 846 #endif /* PFIL_HOOKS */ 847 /* 848 * Send the packet to the outgoing interface. 849 * If necessary, do IPv6 fragmentation before sending. 850 * 851 * the logic here is rather complex: 852 * 1: normal case (dontfrag == 0, alwaysfrag == 0) 853 * 1-a: send as is if tlen <= path mtu 854 * 1-b: fragment if tlen > path mtu 855 * 856 * 2: if user asks us not to fragment (dontfrag == 1) 857 * 2-a: send as is if tlen <= interface mtu 858 * 2-b: error if tlen > interface mtu 859 * 860 * 3: if we always need to attach fragment header (alwaysfrag == 1) 861 * always fragment 862 * 863 * 4: if dontfrag == 1 && alwaysfrag == 1 864 * error, as we cannot handle this conflicting request 865 */ 866 tlen = m->m_pkthdr.len; 867 868 if (opt && (opt->ip6po_flags & IP6PO_DONTFRAG)) 869 dontfrag = 1; 870 else 871 dontfrag = 0; 872 873 if (dontfrag && alwaysfrag) { /* case 4 */ 874 /* conflicting request - can't transmit */ 875 error = EMSGSIZE; 876 goto bad; 877 } 878 if (dontfrag && tlen > IN6_LINKMTU(ifp)) { /* case 2-b */ 879 /* 880 * Even if the DONTFRAG option is specified, we cannot send the 881 * packet when the data length is larger than the MTU of the 882 * outgoing interface. 883 * Notify the error by sending IPV6_PATHMTU ancillary data as 884 * well as returning an error code (the latter is not described 885 * in the API spec.) 886 */ 887 u_int32_t mtu32; 888 struct ip6ctlparam ip6cp; 889 890 mtu32 = (u_int32_t)mtu; 891 bzero(&ip6cp, sizeof(ip6cp)); 892 ip6cp.ip6c_cmdarg = (void *)&mtu32; 893 pfctlinput2(PRC_MSGSIZE, (struct sockaddr *)&ro_pmtu->ro_dst, 894 (void *)&ip6cp); 895 896 error = EMSGSIZE; 897 goto bad; 898 } 899 900 /* 901 * transmit packet without fragmentation 902 */ 903 if (dontfrag || (!alwaysfrag && tlen <= mtu)) { /* case 1-a and 2-a */ 904 struct in6_ifaddr *ia6; 905 int sw_csum; 906 907 ip6 = mtod(m, struct ip6_hdr *); 908 ia6 = in6_ifawithifp(ifp, &ip6->ip6_src); 909 if (ia6) { 910 /* Record statistics for this interface address. */ 911 ia6->ia_ifa.ifa_data.ifad_outbytes += m->m_pkthdr.len; 912 } 913 #ifdef IPSEC 914 /* clean ipsec history once it goes out of the node */ 915 ipsec_delaux(m); 916 #endif 917 918 sw_csum = m->m_pkthdr.csum_flags & ~ifp->if_csum_flags_tx; 919 if ((sw_csum & (M_CSUM_UDPv6|M_CSUM_TCPv6)) != 0) { 920 if (IN6_NEED_CHECKSUM(ifp, 921 sw_csum & (M_CSUM_UDPv6|M_CSUM_TCPv6))) { 922 in6_delayed_cksum(m); 923 } 924 m->m_pkthdr.csum_flags &= ~(M_CSUM_UDPv6|M_CSUM_TCPv6); 925 } 926 927 error = nd6_output(ifp, origifp, m, dst, rt); 928 goto done; 929 } 930 931 /* 932 * try to fragment the packet. case 1-b and 3 933 */ 934 if (mtu < IPV6_MMTU) { 935 /* path MTU cannot be less than IPV6_MMTU */ 936 error = EMSGSIZE; 937 in6_ifstat_inc(ifp, ifs6_out_fragfail); 938 goto bad; 939 } else if (ip6->ip6_plen == 0) { 940 /* jumbo payload cannot be fragmented */ 941 error = EMSGSIZE; 942 in6_ifstat_inc(ifp, ifs6_out_fragfail); 943 goto bad; 944 } else { 945 struct mbuf **mnext, *m_frgpart; 946 struct ip6_frag *ip6f; 947 u_int32_t id = htonl(ip6_randomid()); 948 u_char nextproto; 949 #if 0 /* see below */ 950 struct ip6ctlparam ip6cp; 951 u_int32_t mtu32; 952 #endif 953 954 /* 955 * Too large for the destination or interface; 956 * fragment if possible. 957 * Must be able to put at least 8 bytes per fragment. 958 */ 959 hlen = unfragpartlen; 960 if (mtu > IPV6_MAXPACKET) 961 mtu = IPV6_MAXPACKET; 962 963 #if 0 964 /* 965 * It is believed this code is a leftover from the 966 * development of the IPV6_RECVPATHMTU sockopt and 967 * associated work to implement RFC3542. 968 * It's not entirely clear what the intent of the API 969 * is at this point, so disable this code for now. 970 * The IPV6_RECVPATHMTU sockopt and/or IPV6_DONTFRAG 971 * will send notifications if the application requests. 972 */ 973 974 /* Notify a proper path MTU to applications. */ 975 mtu32 = (u_int32_t)mtu; 976 bzero(&ip6cp, sizeof(ip6cp)); 977 ip6cp.ip6c_cmdarg = (void *)&mtu32; 978 pfctlinput2(PRC_MSGSIZE, (struct sockaddr *)&ro_pmtu->ro_dst, 979 (void *)&ip6cp); 980 #endif 981 982 len = (mtu - hlen - sizeof(struct ip6_frag)) & ~7; 983 if (len < 8) { 984 error = EMSGSIZE; 985 in6_ifstat_inc(ifp, ifs6_out_fragfail); 986 goto bad; 987 } 988 989 mnext = &m->m_nextpkt; 990 991 /* 992 * Change the next header field of the last header in the 993 * unfragmentable part. 994 */ 995 if (exthdrs.ip6e_rthdr) { 996 nextproto = *mtod(exthdrs.ip6e_rthdr, u_char *); 997 *mtod(exthdrs.ip6e_rthdr, u_char *) = IPPROTO_FRAGMENT; 998 } else if (exthdrs.ip6e_dest1) { 999 nextproto = *mtod(exthdrs.ip6e_dest1, u_char *); 1000 *mtod(exthdrs.ip6e_dest1, u_char *) = IPPROTO_FRAGMENT; 1001 } else if (exthdrs.ip6e_hbh) { 1002 nextproto = *mtod(exthdrs.ip6e_hbh, u_char *); 1003 *mtod(exthdrs.ip6e_hbh, u_char *) = IPPROTO_FRAGMENT; 1004 } else { 1005 nextproto = ip6->ip6_nxt; 1006 ip6->ip6_nxt = IPPROTO_FRAGMENT; 1007 } 1008 1009 if ((m->m_pkthdr.csum_flags & (M_CSUM_UDPv6|M_CSUM_TCPv6)) 1010 != 0) { 1011 if (IN6_NEED_CHECKSUM(ifp, 1012 m->m_pkthdr.csum_flags & 1013 (M_CSUM_UDPv6|M_CSUM_TCPv6))) { 1014 in6_delayed_cksum(m); 1015 } 1016 m->m_pkthdr.csum_flags &= ~(M_CSUM_UDPv6|M_CSUM_TCPv6); 1017 } 1018 1019 /* 1020 * Loop through length of segment after first fragment, 1021 * make new header and copy data of each part and link onto 1022 * chain. 1023 */ 1024 m0 = m; 1025 for (off = hlen; off < tlen; off += len) { 1026 struct mbuf *mlast; 1027 1028 MGETHDR(m, M_DONTWAIT, MT_HEADER); 1029 if (!m) { 1030 error = ENOBUFS; 1031 ip6stat.ip6s_odropped++; 1032 goto sendorfree; 1033 } 1034 m->m_pkthdr.rcvif = NULL; 1035 m->m_flags = m0->m_flags & M_COPYFLAGS; 1036 *mnext = m; 1037 mnext = &m->m_nextpkt; 1038 m->m_data += max_linkhdr; 1039 mhip6 = mtod(m, struct ip6_hdr *); 1040 *mhip6 = *ip6; 1041 m->m_len = sizeof(*mhip6); 1042 error = ip6_insertfraghdr(m0, m, hlen, &ip6f); 1043 if (error) { 1044 ip6stat.ip6s_odropped++; 1045 goto sendorfree; 1046 } 1047 ip6f->ip6f_offlg = htons((u_int16_t)((off - hlen) & ~7)); 1048 if (off + len >= tlen) 1049 len = tlen - off; 1050 else 1051 ip6f->ip6f_offlg |= IP6F_MORE_FRAG; 1052 mhip6->ip6_plen = htons((u_int16_t)(len + hlen + 1053 sizeof(*ip6f) - sizeof(struct ip6_hdr))); 1054 if ((m_frgpart = m_copy(m0, off, len)) == 0) { 1055 error = ENOBUFS; 1056 ip6stat.ip6s_odropped++; 1057 goto sendorfree; 1058 } 1059 for (mlast = m; mlast->m_next; mlast = mlast->m_next) 1060 ; 1061 mlast->m_next = m_frgpart; 1062 m->m_pkthdr.len = len + hlen + sizeof(*ip6f); 1063 m->m_pkthdr.rcvif = (struct ifnet *)0; 1064 ip6f->ip6f_reserved = 0; 1065 ip6f->ip6f_ident = id; 1066 ip6f->ip6f_nxt = nextproto; 1067 ip6stat.ip6s_ofragments++; 1068 in6_ifstat_inc(ifp, ifs6_out_fragcreat); 1069 } 1070 1071 in6_ifstat_inc(ifp, ifs6_out_fragok); 1072 } 1073 1074 /* 1075 * Remove leading garbages. 1076 */ 1077 sendorfree: 1078 m = m0->m_nextpkt; 1079 m0->m_nextpkt = 0; 1080 m_freem(m0); 1081 for (m0 = m; m; m = m0) { 1082 m0 = m->m_nextpkt; 1083 m->m_nextpkt = 0; 1084 if (error == 0) { 1085 struct in6_ifaddr *ia6; 1086 ip6 = mtod(m, struct ip6_hdr *); 1087 ia6 = in6_ifawithifp(ifp, &ip6->ip6_src); 1088 if (ia6) { 1089 /* 1090 * Record statistics for this interface 1091 * address. 1092 */ 1093 ia6->ia_ifa.ifa_data.ifad_outbytes += 1094 m->m_pkthdr.len; 1095 } 1096 #ifdef IPSEC 1097 /* clean ipsec history once it goes out of the node */ 1098 ipsec_delaux(m); 1099 #endif 1100 error = nd6_output(ifp, origifp, m, dst, rt); 1101 } else 1102 m_freem(m); 1103 } 1104 1105 if (error == 0) 1106 ip6stat.ip6s_fragmented++; 1107 1108 done: 1109 if (ro == &ip6route && ro->ro_rt) { /* brace necessary for RTFREE */ 1110 RTFREE(ro->ro_rt); 1111 } else if (ro_pmtu == &ip6route && ro_pmtu->ro_rt) { 1112 RTFREE(ro_pmtu->ro_rt); 1113 } 1114 1115 #ifdef IPSEC 1116 if (sp != NULL) 1117 key_freesp(sp); 1118 #endif /* IPSEC */ 1119 1120 return (error); 1121 1122 freehdrs: 1123 m_freem(exthdrs.ip6e_hbh); /* m_freem will check if mbuf is 0 */ 1124 m_freem(exthdrs.ip6e_dest1); 1125 m_freem(exthdrs.ip6e_rthdr); 1126 m_freem(exthdrs.ip6e_dest2); 1127 /* FALLTHROUGH */ 1128 bad: 1129 m_freem(m); 1130 goto done; 1131 } 1132 1133 static int 1134 ip6_copyexthdr(mp, hdr, hlen) 1135 struct mbuf **mp; 1136 caddr_t hdr; 1137 int hlen; 1138 { 1139 struct mbuf *m; 1140 1141 if (hlen > MCLBYTES) 1142 return (ENOBUFS); /* XXX */ 1143 1144 MGET(m, M_DONTWAIT, MT_DATA); 1145 if (!m) 1146 return (ENOBUFS); 1147 1148 if (hlen > MLEN) { 1149 MCLGET(m, M_DONTWAIT); 1150 if ((m->m_flags & M_EXT) == 0) { 1151 m_free(m); 1152 return (ENOBUFS); 1153 } 1154 } 1155 m->m_len = hlen; 1156 if (hdr) 1157 bcopy(hdr, mtod(m, caddr_t), hlen); 1158 1159 *mp = m; 1160 return (0); 1161 } 1162 1163 /* 1164 * Process a delayed payload checksum calculation. 1165 */ 1166 void 1167 in6_delayed_cksum(struct mbuf *m) 1168 { 1169 uint16_t csum, offset; 1170 1171 KASSERT((m->m_pkthdr.csum_flags & (M_CSUM_UDPv6|M_CSUM_TCPv6)) != 0); 1172 KASSERT((~m->m_pkthdr.csum_flags & (M_CSUM_UDPv6|M_CSUM_TCPv6)) != 0); 1173 KASSERT((m->m_pkthdr.csum_flags 1174 & (M_CSUM_UDPv4|M_CSUM_TCPv4|M_CSUM_TSOv4)) == 0); 1175 1176 offset = M_CSUM_DATA_IPv6_HL(m->m_pkthdr.csum_data); 1177 csum = in6_cksum(m, 0, offset, m->m_pkthdr.len - offset); 1178 if (csum == 0 && (m->m_pkthdr.csum_flags & M_CSUM_UDPv6) != 0) { 1179 csum = 0xffff; 1180 } 1181 1182 offset += M_CSUM_DATA_IPv6_OFFSET(m->m_pkthdr.csum_data); 1183 if ((offset + sizeof(csum)) > m->m_len) { 1184 m_copyback(m, offset, sizeof(csum), &csum); 1185 } else { 1186 *(uint16_t *)(mtod(m, caddr_t) + offset) = csum; 1187 } 1188 } 1189 1190 /* 1191 * Insert jumbo payload option. 1192 */ 1193 static int 1194 ip6_insert_jumboopt(exthdrs, plen) 1195 struct ip6_exthdrs *exthdrs; 1196 u_int32_t plen; 1197 { 1198 struct mbuf *mopt; 1199 u_int8_t *optbuf; 1200 u_int32_t v; 1201 1202 #define JUMBOOPTLEN 8 /* length of jumbo payload option and padding */ 1203 1204 /* 1205 * If there is no hop-by-hop options header, allocate new one. 1206 * If there is one but it doesn't have enough space to store the 1207 * jumbo payload option, allocate a cluster to store the whole options. 1208 * Otherwise, use it to store the options. 1209 */ 1210 if (exthdrs->ip6e_hbh == 0) { 1211 MGET(mopt, M_DONTWAIT, MT_DATA); 1212 if (mopt == 0) 1213 return (ENOBUFS); 1214 mopt->m_len = JUMBOOPTLEN; 1215 optbuf = mtod(mopt, u_int8_t *); 1216 optbuf[1] = 0; /* = ((JUMBOOPTLEN) >> 3) - 1 */ 1217 exthdrs->ip6e_hbh = mopt; 1218 } else { 1219 struct ip6_hbh *hbh; 1220 1221 mopt = exthdrs->ip6e_hbh; 1222 if (M_TRAILINGSPACE(mopt) < JUMBOOPTLEN) { 1223 /* 1224 * XXX assumption: 1225 * - exthdrs->ip6e_hbh is not referenced from places 1226 * other than exthdrs. 1227 * - exthdrs->ip6e_hbh is not an mbuf chain. 1228 */ 1229 int oldoptlen = mopt->m_len; 1230 struct mbuf *n; 1231 1232 /* 1233 * XXX: give up if the whole (new) hbh header does 1234 * not fit even in an mbuf cluster. 1235 */ 1236 if (oldoptlen + JUMBOOPTLEN > MCLBYTES) 1237 return (ENOBUFS); 1238 1239 /* 1240 * As a consequence, we must always prepare a cluster 1241 * at this point. 1242 */ 1243 MGET(n, M_DONTWAIT, MT_DATA); 1244 if (n) { 1245 MCLGET(n, M_DONTWAIT); 1246 if ((n->m_flags & M_EXT) == 0) { 1247 m_freem(n); 1248 n = NULL; 1249 } 1250 } 1251 if (!n) 1252 return (ENOBUFS); 1253 n->m_len = oldoptlen + JUMBOOPTLEN; 1254 bcopy(mtod(mopt, caddr_t), mtod(n, caddr_t), 1255 oldoptlen); 1256 optbuf = mtod(n, u_int8_t *) + oldoptlen; 1257 m_freem(mopt); 1258 mopt = exthdrs->ip6e_hbh = n; 1259 } else { 1260 optbuf = mtod(mopt, u_int8_t *) + mopt->m_len; 1261 mopt->m_len += JUMBOOPTLEN; 1262 } 1263 optbuf[0] = IP6OPT_PADN; 1264 optbuf[1] = 0; 1265 1266 /* 1267 * Adjust the header length according to the pad and 1268 * the jumbo payload option. 1269 */ 1270 hbh = mtod(mopt, struct ip6_hbh *); 1271 hbh->ip6h_len += (JUMBOOPTLEN >> 3); 1272 } 1273 1274 /* fill in the option. */ 1275 optbuf[2] = IP6OPT_JUMBO; 1276 optbuf[3] = 4; 1277 v = (u_int32_t)htonl(plen + JUMBOOPTLEN); 1278 bcopy(&v, &optbuf[4], sizeof(u_int32_t)); 1279 1280 /* finally, adjust the packet header length */ 1281 exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN; 1282 1283 return (0); 1284 #undef JUMBOOPTLEN 1285 } 1286 1287 /* 1288 * Insert fragment header and copy unfragmentable header portions. 1289 */ 1290 static int 1291 ip6_insertfraghdr(m0, m, hlen, frghdrp) 1292 struct mbuf *m0, *m; 1293 int hlen; 1294 struct ip6_frag **frghdrp; 1295 { 1296 struct mbuf *n, *mlast; 1297 1298 if (hlen > sizeof(struct ip6_hdr)) { 1299 n = m_copym(m0, sizeof(struct ip6_hdr), 1300 hlen - sizeof(struct ip6_hdr), M_DONTWAIT); 1301 if (n == 0) 1302 return (ENOBUFS); 1303 m->m_next = n; 1304 } else 1305 n = m; 1306 1307 /* Search for the last mbuf of unfragmentable part. */ 1308 for (mlast = n; mlast->m_next; mlast = mlast->m_next) 1309 ; 1310 1311 if ((mlast->m_flags & M_EXT) == 0 && 1312 M_TRAILINGSPACE(mlast) >= sizeof(struct ip6_frag)) { 1313 /* use the trailing space of the last mbuf for the fragment hdr */ 1314 *frghdrp = (struct ip6_frag *)(mtod(mlast, caddr_t) + 1315 mlast->m_len); 1316 mlast->m_len += sizeof(struct ip6_frag); 1317 m->m_pkthdr.len += sizeof(struct ip6_frag); 1318 } else { 1319 /* allocate a new mbuf for the fragment header */ 1320 struct mbuf *mfrg; 1321 1322 MGET(mfrg, M_DONTWAIT, MT_DATA); 1323 if (mfrg == 0) 1324 return (ENOBUFS); 1325 mfrg->m_len = sizeof(struct ip6_frag); 1326 *frghdrp = mtod(mfrg, struct ip6_frag *); 1327 mlast->m_next = mfrg; 1328 } 1329 1330 return (0); 1331 } 1332 1333 static int 1334 ip6_getpmtu(ro_pmtu, ro, ifp, dst, mtup, alwaysfragp) 1335 struct route_in6 *ro_pmtu, *ro; 1336 struct ifnet *ifp; 1337 struct in6_addr *dst; 1338 u_long *mtup; 1339 int *alwaysfragp; 1340 { 1341 u_int32_t mtu = 0; 1342 int alwaysfrag = 0; 1343 int error = 0; 1344 1345 if (ro_pmtu != ro) { 1346 /* The first hop and the final destination may differ. */ 1347 struct sockaddr_in6 *sa6_dst = 1348 (struct sockaddr_in6 *)&ro_pmtu->ro_dst; 1349 if (ro_pmtu->ro_rt && 1350 ((ro_pmtu->ro_rt->rt_flags & RTF_UP) == 0 || 1351 !IN6_ARE_ADDR_EQUAL(&sa6_dst->sin6_addr, dst))) { 1352 RTFREE(ro_pmtu->ro_rt); 1353 ro_pmtu->ro_rt = (struct rtentry *)NULL; 1354 } 1355 if (ro_pmtu->ro_rt == NULL) { 1356 bzero(sa6_dst, sizeof(*sa6_dst)); /* for safety */ 1357 sa6_dst->sin6_family = AF_INET6; 1358 sa6_dst->sin6_len = sizeof(struct sockaddr_in6); 1359 sa6_dst->sin6_addr = *dst; 1360 1361 rtalloc((struct route *)ro_pmtu); 1362 } 1363 } 1364 if (ro_pmtu->ro_rt) { 1365 u_int32_t ifmtu; 1366 1367 if (ifp == NULL) 1368 ifp = ro_pmtu->ro_rt->rt_ifp; 1369 ifmtu = IN6_LINKMTU(ifp); 1370 mtu = ro_pmtu->ro_rt->rt_rmx.rmx_mtu; 1371 if (mtu == 0) 1372 mtu = ifmtu; 1373 else if (mtu < IPV6_MMTU) { 1374 /* 1375 * RFC2460 section 5, last paragraph: 1376 * if we record ICMPv6 too big message with 1377 * mtu < IPV6_MMTU, transmit packets sized IPV6_MMTU 1378 * or smaller, with fragment header attached. 1379 * (fragment header is needed regardless from the 1380 * packet size, for translators to identify packets) 1381 */ 1382 alwaysfrag = 1; 1383 mtu = IPV6_MMTU; 1384 } else if (mtu > ifmtu) { 1385 /* 1386 * The MTU on the route is larger than the MTU on 1387 * the interface! This shouldn't happen, unless the 1388 * MTU of the interface has been changed after the 1389 * interface was brought up. Change the MTU in the 1390 * route to match the interface MTU (as long as the 1391 * field isn't locked). 1392 */ 1393 mtu = ifmtu; 1394 if (!(ro_pmtu->ro_rt->rt_rmx.rmx_locks & RTV_MTU)) 1395 ro_pmtu->ro_rt->rt_rmx.rmx_mtu = mtu; 1396 } 1397 } else if (ifp) { 1398 mtu = IN6_LINKMTU(ifp); 1399 } else 1400 error = EHOSTUNREACH; /* XXX */ 1401 1402 *mtup = mtu; 1403 if (alwaysfragp) 1404 *alwaysfragp = alwaysfrag; 1405 return (error); 1406 } 1407 1408 /* 1409 * IP6 socket option processing. 1410 */ 1411 int 1412 ip6_ctloutput(op, so, level, optname, mp) 1413 int op; 1414 struct socket *so; 1415 int level, optname; 1416 struct mbuf **mp; 1417 { 1418 int privileged, optdatalen, uproto; 1419 void *optdata; 1420 struct in6pcb *in6p = sotoin6pcb(so); 1421 struct mbuf *m = *mp; 1422 int error, optval; 1423 int optlen; 1424 struct proc *p = curproc; /* XXX */ 1425 1426 optlen = m ? m->m_len : 0; 1427 error = optval = 0; 1428 privileged = (p == 0 || kauth_authorize_generic(p->p_cred, KAUTH_GENERIC_ISSUSER, &p->p_acflag)) ? 0 : 1; 1429 uproto = (int)so->so_proto->pr_protocol; 1430 1431 if (level == IPPROTO_IPV6) { 1432 switch (op) { 1433 case PRCO_SETOPT: 1434 switch (optname) { 1435 #ifdef RFC2292 1436 case IPV6_2292PKTOPTIONS: 1437 /* m is freed in ip6_pcbopts */ 1438 error = ip6_pcbopts(&in6p->in6p_outputopts, 1439 m, so); 1440 break; 1441 #endif 1442 1443 /* 1444 * Use of some Hop-by-Hop options or some 1445 * Destination options, might require special 1446 * privilege. That is, normal applications 1447 * (without special privilege) might be forbidden 1448 * from setting certain options in outgoing packets, 1449 * and might never see certain options in received 1450 * packets. [RFC 2292 Section 6] 1451 * KAME specific note: 1452 * KAME prevents non-privileged users from sending or 1453 * receiving ANY hbh/dst options in order to avoid 1454 * overhead of parsing options in the kernel. 1455 */ 1456 case IPV6_RECVHOPOPTS: 1457 case IPV6_RECVDSTOPTS: 1458 case IPV6_RECVRTHDRDSTOPTS: 1459 if (!privileged) { 1460 error = EPERM; 1461 break; 1462 } 1463 /* FALLTHROUGH */ 1464 case IPV6_UNICAST_HOPS: 1465 case IPV6_HOPLIMIT: 1466 case IPV6_FAITH: 1467 1468 case IPV6_RECVPKTINFO: 1469 case IPV6_RECVHOPLIMIT: 1470 case IPV6_RECVRTHDR: 1471 case IPV6_RECVPATHMTU: 1472 case IPV6_RECVTCLASS: 1473 case IPV6_V6ONLY: 1474 if (optlen != sizeof(int)) { 1475 error = EINVAL; 1476 break; 1477 } 1478 optval = *mtod(m, int *); 1479 switch (optname) { 1480 1481 case IPV6_UNICAST_HOPS: 1482 if (optval < -1 || optval >= 256) 1483 error = EINVAL; 1484 else { 1485 /* -1 = kernel default */ 1486 in6p->in6p_hops = optval; 1487 } 1488 break; 1489 #define OPTSET(bit) \ 1490 do { \ 1491 if (optval) \ 1492 in6p->in6p_flags |= (bit); \ 1493 else \ 1494 in6p->in6p_flags &= ~(bit); \ 1495 } while (/*CONSTCOND*/ 0) 1496 1497 #ifdef RFC2292 1498 #define OPTSET2292(bit) \ 1499 do { \ 1500 in6p->in6p_flags |= IN6P_RFC2292; \ 1501 if (optval) \ 1502 in6p->in6p_flags |= (bit); \ 1503 else \ 1504 in6p->in6p_flags &= ~(bit); \ 1505 } while (/*CONSTCOND*/ 0) 1506 #endif 1507 1508 #define OPTBIT(bit) (in6p->in6p_flags & (bit) ? 1 : 0) 1509 1510 case IPV6_RECVPKTINFO: 1511 #ifdef RFC2292 1512 /* cannot mix with RFC2292 */ 1513 if (OPTBIT(IN6P_RFC2292)) { 1514 error = EINVAL; 1515 break; 1516 } 1517 #endif 1518 OPTSET(IN6P_PKTINFO); 1519 break; 1520 1521 case IPV6_HOPLIMIT: 1522 { 1523 struct ip6_pktopts **optp; 1524 1525 #ifdef RFC2292 1526 /* cannot mix with RFC2292 */ 1527 if (OPTBIT(IN6P_RFC2292)) { 1528 error = EINVAL; 1529 break; 1530 } 1531 #endif 1532 optp = &in6p->in6p_outputopts; 1533 error = ip6_pcbopt(IPV6_HOPLIMIT, 1534 (u_char *)&optval, 1535 sizeof(optval), 1536 optp, 1537 privileged, uproto); 1538 break; 1539 } 1540 1541 case IPV6_RECVHOPLIMIT: 1542 #ifdef RFC2292 1543 /* cannot mix with RFC2292 */ 1544 if (OPTBIT(IN6P_RFC2292)) { 1545 error = EINVAL; 1546 break; 1547 } 1548 #endif 1549 OPTSET(IN6P_HOPLIMIT); 1550 break; 1551 1552 case IPV6_RECVHOPOPTS: 1553 #ifdef RFC2292 1554 /* cannot mix with RFC2292 */ 1555 if (OPTBIT(IN6P_RFC2292)) { 1556 error = EINVAL; 1557 break; 1558 } 1559 #endif 1560 OPTSET(IN6P_HOPOPTS); 1561 break; 1562 1563 case IPV6_RECVDSTOPTS: 1564 #ifdef RFC2292 1565 /* cannot mix with RFC2292 */ 1566 if (OPTBIT(IN6P_RFC2292)) { 1567 error = EINVAL; 1568 break; 1569 } 1570 #endif 1571 OPTSET(IN6P_DSTOPTS); 1572 break; 1573 1574 case IPV6_RECVRTHDRDSTOPTS: 1575 #ifdef RFC2292 1576 /* cannot mix with RFC2292 */ 1577 if (OPTBIT(IN6P_RFC2292)) { 1578 error = EINVAL; 1579 break; 1580 } 1581 #endif 1582 OPTSET(IN6P_RTHDRDSTOPTS); 1583 break; 1584 1585 case IPV6_RECVRTHDR: 1586 #ifdef RFC2292 1587 /* cannot mix with RFC2292 */ 1588 if (OPTBIT(IN6P_RFC2292)) { 1589 error = EINVAL; 1590 break; 1591 } 1592 #endif 1593 OPTSET(IN6P_RTHDR); 1594 break; 1595 1596 case IPV6_FAITH: 1597 OPTSET(IN6P_FAITH); 1598 break; 1599 1600 case IPV6_RECVPATHMTU: 1601 /* 1602 * We ignore this option for TCP 1603 * sockets. 1604 * (RFC3542 leaves this case 1605 * unspecified.) 1606 */ 1607 if (uproto != IPPROTO_TCP) 1608 OPTSET(IN6P_MTU); 1609 break; 1610 1611 case IPV6_V6ONLY: 1612 /* 1613 * make setsockopt(IPV6_V6ONLY) 1614 * available only prior to bind(2). 1615 * see ipng mailing list, Jun 22 2001. 1616 */ 1617 if (in6p->in6p_lport || 1618 !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr)) { 1619 error = EINVAL; 1620 break; 1621 } 1622 #ifdef INET6_BINDV6ONLY 1623 if (!optval) 1624 error = EINVAL; 1625 #else 1626 OPTSET(IN6P_IPV6_V6ONLY); 1627 #endif 1628 break; 1629 case IPV6_RECVTCLASS: 1630 #ifdef RFC2292 1631 /* cannot mix with RFC2292 XXX */ 1632 if (OPTBIT(IN6P_RFC2292)) { 1633 error = EINVAL; 1634 break; 1635 } 1636 #endif 1637 OPTSET(IN6P_TCLASS); 1638 break; 1639 1640 } 1641 break; 1642 1643 case IPV6_OTCLASS: 1644 { 1645 struct ip6_pktopts **optp; 1646 u_int8_t tclass; 1647 1648 if (optlen != sizeof(tclass)) { 1649 error = EINVAL; 1650 break; 1651 } 1652 tclass = *mtod(m, u_int8_t *); 1653 optp = &in6p->in6p_outputopts; 1654 error = ip6_pcbopt(optname, 1655 (u_char *)&tclass, 1656 sizeof(tclass), 1657 optp, 1658 privileged, uproto); 1659 break; 1660 } 1661 1662 case IPV6_TCLASS: 1663 case IPV6_DONTFRAG: 1664 case IPV6_USE_MIN_MTU: 1665 if (optlen != sizeof(optval)) { 1666 error = EINVAL; 1667 break; 1668 } 1669 optval = *mtod(m, int *); 1670 { 1671 struct ip6_pktopts **optp; 1672 optp = &in6p->in6p_outputopts; 1673 error = ip6_pcbopt(optname, 1674 (u_char *)&optval, 1675 sizeof(optval), 1676 optp, 1677 privileged, uproto); 1678 break; 1679 } 1680 1681 #ifdef RFC2292 1682 case IPV6_2292PKTINFO: 1683 case IPV6_2292HOPLIMIT: 1684 case IPV6_2292HOPOPTS: 1685 case IPV6_2292DSTOPTS: 1686 case IPV6_2292RTHDR: 1687 /* RFC 2292 */ 1688 if (optlen != sizeof(int)) { 1689 error = EINVAL; 1690 break; 1691 } 1692 optval = *mtod(m, int *); 1693 switch (optname) { 1694 case IPV6_2292PKTINFO: 1695 OPTSET2292(IN6P_PKTINFO); 1696 break; 1697 case IPV6_2292HOPLIMIT: 1698 OPTSET2292(IN6P_HOPLIMIT); 1699 break; 1700 case IPV6_2292HOPOPTS: 1701 /* 1702 * Check super-user privilege. 1703 * See comments for IPV6_RECVHOPOPTS. 1704 */ 1705 if (!privileged) 1706 return (EPERM); 1707 OPTSET2292(IN6P_HOPOPTS); 1708 break; 1709 case IPV6_2292DSTOPTS: 1710 if (!privileged) 1711 return (EPERM); 1712 OPTSET2292(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); /* XXX */ 1713 break; 1714 case IPV6_2292RTHDR: 1715 OPTSET2292(IN6P_RTHDR); 1716 break; 1717 } 1718 break; 1719 #endif 1720 case IPV6_PKTINFO: 1721 case IPV6_HOPOPTS: 1722 case IPV6_RTHDR: 1723 case IPV6_DSTOPTS: 1724 case IPV6_RTHDRDSTOPTS: 1725 case IPV6_NEXTHOP: 1726 { 1727 /* new advanced API (RFC3542) */ 1728 u_char *optbuf; 1729 int optbuflen; 1730 struct ip6_pktopts **optp; 1731 1732 #ifdef RFC2292 1733 /* cannot mix with RFC2292 */ 1734 if (OPTBIT(IN6P_RFC2292)) { 1735 error = EINVAL; 1736 break; 1737 } 1738 #endif 1739 1740 if (m && m->m_next) { 1741 error = EINVAL; /* XXX */ 1742 break; 1743 } 1744 if (m) { 1745 optbuf = mtod(m, u_char *); 1746 optbuflen = m->m_len; 1747 } else { 1748 optbuf = NULL; 1749 optbuflen = 0; 1750 } 1751 optp = &in6p->in6p_outputopts; 1752 error = ip6_pcbopt(optname, 1753 optbuf, optbuflen, 1754 optp, privileged, uproto); 1755 break; 1756 } 1757 #undef OPTSET 1758 1759 case IPV6_MULTICAST_IF: 1760 case IPV6_MULTICAST_HOPS: 1761 case IPV6_MULTICAST_LOOP: 1762 case IPV6_JOIN_GROUP: 1763 case IPV6_LEAVE_GROUP: 1764 error = ip6_setmoptions(optname, 1765 &in6p->in6p_moptions, m); 1766 break; 1767 1768 case IPV6_PORTRANGE: 1769 optval = *mtod(m, int *); 1770 1771 switch (optval) { 1772 case IPV6_PORTRANGE_DEFAULT: 1773 in6p->in6p_flags &= ~(IN6P_LOWPORT); 1774 in6p->in6p_flags &= ~(IN6P_HIGHPORT); 1775 break; 1776 1777 case IPV6_PORTRANGE_HIGH: 1778 in6p->in6p_flags &= ~(IN6P_LOWPORT); 1779 in6p->in6p_flags |= IN6P_HIGHPORT; 1780 break; 1781 1782 case IPV6_PORTRANGE_LOW: 1783 in6p->in6p_flags &= ~(IN6P_HIGHPORT); 1784 in6p->in6p_flags |= IN6P_LOWPORT; 1785 break; 1786 1787 default: 1788 error = EINVAL; 1789 break; 1790 } 1791 break; 1792 1793 #ifdef IPSEC 1794 case IPV6_IPSEC_POLICY: 1795 { 1796 caddr_t req = NULL; 1797 size_t len = 0; 1798 if (m) { 1799 req = mtod(m, caddr_t); 1800 len = m->m_len; 1801 } 1802 error = ipsec6_set_policy(in6p, optname, req, 1803 len, privileged); 1804 } 1805 break; 1806 #endif /* IPSEC */ 1807 1808 default: 1809 error = ENOPROTOOPT; 1810 break; 1811 } 1812 if (m) 1813 (void)m_free(m); 1814 break; 1815 1816 case PRCO_GETOPT: 1817 switch (optname) { 1818 #ifdef RFC2292 1819 case IPV6_2292PKTOPTIONS: 1820 /* 1821 * RFC3542 (effectively) deprecated the 1822 * semantics of the 2292-style pktoptions. 1823 * Since it was not reliable in nature (i.e., 1824 * applications had to expect the lack of some 1825 * information after all), it would make sense 1826 * to simplify this part by always returning 1827 * empty data. 1828 */ 1829 *mp = m_get(M_WAIT, MT_SOOPTS); 1830 (*mp)->m_len = 0; 1831 break; 1832 #endif 1833 1834 case IPV6_RECVHOPOPTS: 1835 case IPV6_RECVDSTOPTS: 1836 case IPV6_RECVRTHDRDSTOPTS: 1837 case IPV6_UNICAST_HOPS: 1838 case IPV6_RECVPKTINFO: 1839 case IPV6_RECVHOPLIMIT: 1840 case IPV6_RECVRTHDR: 1841 case IPV6_RECVPATHMTU: 1842 1843 case IPV6_FAITH: 1844 case IPV6_V6ONLY: 1845 case IPV6_PORTRANGE: 1846 case IPV6_RECVTCLASS: 1847 switch (optname) { 1848 1849 case IPV6_RECVHOPOPTS: 1850 optval = OPTBIT(IN6P_HOPOPTS); 1851 break; 1852 1853 case IPV6_RECVDSTOPTS: 1854 optval = OPTBIT(IN6P_DSTOPTS); 1855 break; 1856 1857 case IPV6_RECVRTHDRDSTOPTS: 1858 optval = OPTBIT(IN6P_RTHDRDSTOPTS); 1859 break; 1860 1861 case IPV6_UNICAST_HOPS: 1862 optval = in6p->in6p_hops; 1863 break; 1864 1865 case IPV6_RECVPKTINFO: 1866 optval = OPTBIT(IN6P_PKTINFO); 1867 break; 1868 1869 case IPV6_RECVHOPLIMIT: 1870 optval = OPTBIT(IN6P_HOPLIMIT); 1871 break; 1872 1873 case IPV6_RECVRTHDR: 1874 optval = OPTBIT(IN6P_RTHDR); 1875 break; 1876 1877 case IPV6_RECVPATHMTU: 1878 optval = OPTBIT(IN6P_MTU); 1879 break; 1880 1881 case IPV6_FAITH: 1882 optval = OPTBIT(IN6P_FAITH); 1883 break; 1884 1885 case IPV6_V6ONLY: 1886 optval = OPTBIT(IN6P_IPV6_V6ONLY); 1887 break; 1888 1889 case IPV6_PORTRANGE: 1890 { 1891 int flags; 1892 flags = in6p->in6p_flags; 1893 if (flags & IN6P_HIGHPORT) 1894 optval = IPV6_PORTRANGE_HIGH; 1895 else if (flags & IN6P_LOWPORT) 1896 optval = IPV6_PORTRANGE_LOW; 1897 else 1898 optval = 0; 1899 break; 1900 } 1901 case IPV6_RECVTCLASS: 1902 optval = OPTBIT(IN6P_TCLASS); 1903 break; 1904 1905 } 1906 if (error) 1907 break; 1908 *mp = m = m_get(M_WAIT, MT_SOOPTS); 1909 m->m_len = sizeof(int); 1910 *mtod(m, int *) = optval; 1911 break; 1912 1913 case IPV6_PATHMTU: 1914 { 1915 u_long pmtu = 0; 1916 struct ip6_mtuinfo mtuinfo; 1917 struct route_in6 *ro = (struct route_in6 *)&in6p 1918 ->in6p_route; 1919 1920 if (!(so->so_state & SS_ISCONNECTED)) 1921 return (ENOTCONN); 1922 /* 1923 * XXX: we dot not consider the case of source 1924 * routing, or optional information to specify 1925 * the outgoing interface. 1926 */ 1927 error = ip6_getpmtu(ro, NULL, NULL, 1928 &in6p->in6p_faddr, &pmtu, NULL); 1929 if (error) 1930 break; 1931 if (pmtu > IPV6_MAXPACKET) 1932 pmtu = IPV6_MAXPACKET; 1933 1934 memset(&mtuinfo, 0, sizeof(mtuinfo)); 1935 mtuinfo.ip6m_mtu = (u_int32_t)pmtu; 1936 optdata = (void *)&mtuinfo; 1937 optdatalen = sizeof(mtuinfo); 1938 if (optdatalen > MCLBYTES) 1939 return (EMSGSIZE); /* XXX */ 1940 *mp = m = m_get(M_WAIT, MT_SOOPTS); 1941 if (optdatalen > MLEN) 1942 MCLGET(m, M_WAIT); 1943 m->m_len = optdatalen; 1944 memcpy(mtod(m, void *), optdata, optdatalen); 1945 break; 1946 } 1947 1948 #ifdef RFC2292 1949 case IPV6_2292PKTINFO: 1950 case IPV6_2292HOPLIMIT: 1951 case IPV6_2292HOPOPTS: 1952 case IPV6_2292RTHDR: 1953 case IPV6_2292DSTOPTS: 1954 switch (optname) { 1955 case IPV6_2292PKTINFO: 1956 optval = OPTBIT(IN6P_PKTINFO); 1957 break; 1958 case IPV6_2292HOPLIMIT: 1959 optval = OPTBIT(IN6P_HOPLIMIT); 1960 break; 1961 case IPV6_2292HOPOPTS: 1962 optval = OPTBIT(IN6P_HOPOPTS); 1963 break; 1964 case IPV6_2292RTHDR: 1965 optval = OPTBIT(IN6P_RTHDR); 1966 break; 1967 case IPV6_2292DSTOPTS: 1968 optval = OPTBIT(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); 1969 break; 1970 } 1971 *mp = m = m_get(M_WAIT, MT_SOOPTS); 1972 m->m_len = sizeof(int); 1973 *mtod(m, int *) = optval; 1974 break; 1975 #endif 1976 case IPV6_PKTINFO: 1977 case IPV6_HOPOPTS: 1978 case IPV6_RTHDR: 1979 case IPV6_DSTOPTS: 1980 case IPV6_RTHDRDSTOPTS: 1981 case IPV6_NEXTHOP: 1982 case IPV6_OTCLASS: 1983 case IPV6_TCLASS: 1984 case IPV6_DONTFRAG: 1985 case IPV6_USE_MIN_MTU: 1986 error = ip6_getpcbopt(in6p->in6p_outputopts, 1987 optname, mp); 1988 break; 1989 1990 case IPV6_MULTICAST_IF: 1991 case IPV6_MULTICAST_HOPS: 1992 case IPV6_MULTICAST_LOOP: 1993 case IPV6_JOIN_GROUP: 1994 case IPV6_LEAVE_GROUP: 1995 error = ip6_getmoptions(optname, 1996 in6p->in6p_moptions, mp); 1997 break; 1998 1999 #ifdef IPSEC 2000 case IPV6_IPSEC_POLICY: 2001 { 2002 caddr_t req = NULL; 2003 size_t len = 0; 2004 if (m) { 2005 req = mtod(m, caddr_t); 2006 len = m->m_len; 2007 } 2008 error = ipsec6_get_policy(in6p, req, len, mp); 2009 break; 2010 } 2011 #endif /* IPSEC */ 2012 2013 2014 2015 2016 default: 2017 error = ENOPROTOOPT; 2018 break; 2019 } 2020 break; 2021 } 2022 } else { 2023 error = EINVAL; 2024 if (op == PRCO_SETOPT && *mp) 2025 (void)m_free(*mp); 2026 } 2027 return (error); 2028 } 2029 2030 int 2031 ip6_raw_ctloutput(op, so, level, optname, mp) 2032 int op; 2033 struct socket *so; 2034 int level, optname; 2035 struct mbuf **mp; 2036 { 2037 int error = 0, optval, optlen; 2038 const int icmp6off = offsetof(struct icmp6_hdr, icmp6_cksum); 2039 struct in6pcb *in6p = sotoin6pcb(so); 2040 struct mbuf *m = *mp; 2041 2042 optlen = m ? m->m_len : 0; 2043 2044 if (level != IPPROTO_IPV6) { 2045 if (op == PRCO_SETOPT && *mp) 2046 (void)m_free(*mp); 2047 return (EINVAL); 2048 } 2049 2050 switch (optname) { 2051 case IPV6_CHECKSUM: 2052 /* 2053 * For ICMPv6 sockets, no modification allowed for checksum 2054 * offset, permit "no change" values to help existing apps. 2055 * 2056 * XXX RFC3542 says: "An attempt to set IPV6_CHECKSUM 2057 * for an ICMPv6 socket will fail." The current 2058 * behavior does not meet RFC3542. 2059 */ 2060 switch (op) { 2061 case PRCO_SETOPT: 2062 if (optlen != sizeof(int)) { 2063 error = EINVAL; 2064 break; 2065 } 2066 optval = *mtod(m, int *); 2067 if ((optval % 2) != 0) { 2068 /* the API assumes even offset values */ 2069 error = EINVAL; 2070 } else if (so->so_proto->pr_protocol == 2071 IPPROTO_ICMPV6) { 2072 if (optval != icmp6off) 2073 error = EINVAL; 2074 } else 2075 in6p->in6p_cksum = optval; 2076 break; 2077 2078 case PRCO_GETOPT: 2079 if (so->so_proto->pr_protocol == IPPROTO_ICMPV6) 2080 optval = icmp6off; 2081 else 2082 optval = in6p->in6p_cksum; 2083 2084 *mp = m = m_get(M_WAIT, MT_SOOPTS); 2085 m->m_len = sizeof(int); 2086 *mtod(m, int *) = optval; 2087 break; 2088 2089 default: 2090 error = EINVAL; 2091 break; 2092 } 2093 break; 2094 2095 default: 2096 error = ENOPROTOOPT; 2097 break; 2098 } 2099 2100 if (op == PRCO_SETOPT && m) 2101 (void)m_free(m); 2102 2103 return (error); 2104 } 2105 2106 #ifdef RFC2292 2107 /* 2108 * Set up IP6 options in pcb for insertion in output packets or 2109 * specifying behavior of outgoing packets. 2110 */ 2111 static int 2112 ip6_pcbopts(pktopt, m, so) 2113 struct ip6_pktopts **pktopt; 2114 struct mbuf *m; 2115 struct socket *so; 2116 { 2117 struct ip6_pktopts *opt = *pktopt; 2118 int error = 0; 2119 struct proc *p = curproc; /* XXX */ 2120 int priv = 0; 2121 2122 /* turn off any old options. */ 2123 if (opt) { 2124 #ifdef DIAGNOSTIC 2125 if (opt->ip6po_pktinfo || opt->ip6po_nexthop || 2126 opt->ip6po_hbh || opt->ip6po_dest1 || opt->ip6po_dest2 || 2127 opt->ip6po_rhinfo.ip6po_rhi_rthdr) 2128 printf("ip6_pcbopts: all specified options are cleared.\n"); 2129 #endif 2130 ip6_clearpktopts(opt, -1); 2131 } else 2132 opt = malloc(sizeof(*opt), M_IP6OPT, M_WAITOK); 2133 *pktopt = NULL; 2134 2135 if (!m || m->m_len == 0) { 2136 /* 2137 * Only turning off any previous options, regardless of 2138 * whether the opt is just created or given. 2139 */ 2140 free(opt, M_IP6OPT); 2141 return (0); 2142 } 2143 2144 /* set options specified by user. */ 2145 if (p && !kauth_authorize_generic(p->p_cred, KAUTH_GENERIC_ISSUSER, &p->p_acflag)) 2146 priv = 1; 2147 if ((error = ip6_setpktopts(m, opt, NULL, priv, 2148 so->so_proto->pr_protocol)) != 0) { 2149 ip6_clearpktopts(opt, -1); /* XXX: discard all options */ 2150 free(opt, M_IP6OPT); 2151 return (error); 2152 } 2153 *pktopt = opt; 2154 return (0); 2155 } 2156 #endif 2157 2158 /* 2159 * initialize ip6_pktopts. beware that there are non-zero default values in 2160 * the struct. 2161 */ 2162 void 2163 ip6_initpktopts(struct ip6_pktopts *opt) 2164 { 2165 2166 memset(opt, 0, sizeof(*opt)); 2167 opt->ip6po_hlim = -1; /* -1 means default hop limit */ 2168 opt->ip6po_tclass = -1; /* -1 means default traffic class */ 2169 opt->ip6po_minmtu = IP6PO_MINMTU_MCASTONLY; 2170 } 2171 2172 #define sin6tosa(sin6) ((struct sockaddr *)(sin6)) /* XXX */ 2173 static int 2174 ip6_pcbopt(int optname, u_char *buf, int len, struct ip6_pktopts **pktopt, 2175 int priv, int uproto) 2176 { 2177 struct ip6_pktopts *opt; 2178 2179 if (*pktopt == NULL) { 2180 *pktopt = malloc(sizeof(struct ip6_pktopts), M_IP6OPT, 2181 M_WAITOK); 2182 ip6_initpktopts(*pktopt); 2183 } 2184 opt = *pktopt; 2185 2186 return (ip6_setpktopt(optname, buf, len, opt, priv, 1, 0, uproto)); 2187 } 2188 2189 static int 2190 ip6_getpcbopt(struct ip6_pktopts *pktopt, int optname, struct mbuf **mp) 2191 { 2192 void *optdata = NULL; 2193 int optdatalen = 0; 2194 struct ip6_ext *ip6e; 2195 int error = 0; 2196 struct in6_pktinfo null_pktinfo; 2197 int deftclass = 0, on; 2198 int defminmtu = IP6PO_MINMTU_MCASTONLY; 2199 struct mbuf *m; 2200 2201 switch (optname) { 2202 case IPV6_PKTINFO: 2203 if (pktopt && pktopt->ip6po_pktinfo) 2204 optdata = (void *)pktopt->ip6po_pktinfo; 2205 else { 2206 /* XXX: we don't have to do this every time... */ 2207 memset(&null_pktinfo, 0, sizeof(null_pktinfo)); 2208 optdata = (void *)&null_pktinfo; 2209 } 2210 optdatalen = sizeof(struct in6_pktinfo); 2211 break; 2212 case IPV6_OTCLASS: 2213 /* XXX */ 2214 return (EINVAL); 2215 case IPV6_TCLASS: 2216 if (pktopt && pktopt->ip6po_tclass >= 0) 2217 optdata = (void *)&pktopt->ip6po_tclass; 2218 else 2219 optdata = (void *)&deftclass; 2220 optdatalen = sizeof(int); 2221 break; 2222 case IPV6_HOPOPTS: 2223 if (pktopt && pktopt->ip6po_hbh) { 2224 optdata = (void *)pktopt->ip6po_hbh; 2225 ip6e = (struct ip6_ext *)pktopt->ip6po_hbh; 2226 optdatalen = (ip6e->ip6e_len + 1) << 3; 2227 } 2228 break; 2229 case IPV6_RTHDR: 2230 if (pktopt && pktopt->ip6po_rthdr) { 2231 optdata = (void *)pktopt->ip6po_rthdr; 2232 ip6e = (struct ip6_ext *)pktopt->ip6po_rthdr; 2233 optdatalen = (ip6e->ip6e_len + 1) << 3; 2234 } 2235 break; 2236 case IPV6_RTHDRDSTOPTS: 2237 if (pktopt && pktopt->ip6po_dest1) { 2238 optdata = (void *)pktopt->ip6po_dest1; 2239 ip6e = (struct ip6_ext *)pktopt->ip6po_dest1; 2240 optdatalen = (ip6e->ip6e_len + 1) << 3; 2241 } 2242 break; 2243 case IPV6_DSTOPTS: 2244 if (pktopt && pktopt->ip6po_dest2) { 2245 optdata = (void *)pktopt->ip6po_dest2; 2246 ip6e = (struct ip6_ext *)pktopt->ip6po_dest2; 2247 optdatalen = (ip6e->ip6e_len + 1) << 3; 2248 } 2249 break; 2250 case IPV6_NEXTHOP: 2251 if (pktopt && pktopt->ip6po_nexthop) { 2252 optdata = (void *)pktopt->ip6po_nexthop; 2253 optdatalen = pktopt->ip6po_nexthop->sa_len; 2254 } 2255 break; 2256 case IPV6_USE_MIN_MTU: 2257 if (pktopt) 2258 optdata = (void *)&pktopt->ip6po_minmtu; 2259 else 2260 optdata = (void *)&defminmtu; 2261 optdatalen = sizeof(int); 2262 break; 2263 case IPV6_DONTFRAG: 2264 if (pktopt && ((pktopt->ip6po_flags) & IP6PO_DONTFRAG)) 2265 on = 1; 2266 else 2267 on = 0; 2268 optdata = (void *)&on; 2269 optdatalen = sizeof(on); 2270 break; 2271 default: /* should not happen */ 2272 #ifdef DIAGNOSTIC 2273 panic("ip6_getpcbopt: unexpected option\n"); 2274 #endif 2275 return (ENOPROTOOPT); 2276 } 2277 2278 if (optdatalen > MCLBYTES) 2279 return (EMSGSIZE); /* XXX */ 2280 *mp = m = m_get(M_WAIT, MT_SOOPTS); 2281 if (optdatalen > MLEN) 2282 MCLGET(m, M_WAIT); 2283 m->m_len = optdatalen; 2284 if (optdatalen) 2285 memcpy(mtod(m, void *), optdata, optdatalen); 2286 2287 return (error); 2288 } 2289 2290 void 2291 ip6_clearpktopts(struct ip6_pktopts *pktopt, int optname) 2292 { 2293 if (optname == -1 || optname == IPV6_PKTINFO) { 2294 if (pktopt->ip6po_pktinfo) 2295 free(pktopt->ip6po_pktinfo, M_IP6OPT); 2296 pktopt->ip6po_pktinfo = NULL; 2297 } 2298 if (optname == -1 || optname == IPV6_HOPLIMIT) 2299 pktopt->ip6po_hlim = -1; 2300 if (optname == -1 || optname == IPV6_TCLASS) 2301 pktopt->ip6po_tclass = -1; 2302 if (optname == -1 || optname == IPV6_NEXTHOP) { 2303 if (pktopt->ip6po_nextroute.ro_rt) { 2304 RTFREE(pktopt->ip6po_nextroute.ro_rt); 2305 pktopt->ip6po_nextroute.ro_rt = NULL; 2306 } 2307 if (pktopt->ip6po_nexthop) 2308 free(pktopt->ip6po_nexthop, M_IP6OPT); 2309 pktopt->ip6po_nexthop = NULL; 2310 } 2311 if (optname == -1 || optname == IPV6_HOPOPTS) { 2312 if (pktopt->ip6po_hbh) 2313 free(pktopt->ip6po_hbh, M_IP6OPT); 2314 pktopt->ip6po_hbh = NULL; 2315 } 2316 if (optname == -1 || optname == IPV6_RTHDRDSTOPTS) { 2317 if (pktopt->ip6po_dest1) 2318 free(pktopt->ip6po_dest1, M_IP6OPT); 2319 pktopt->ip6po_dest1 = NULL; 2320 } 2321 if (optname == -1 || optname == IPV6_RTHDR) { 2322 if (pktopt->ip6po_rhinfo.ip6po_rhi_rthdr) 2323 free(pktopt->ip6po_rhinfo.ip6po_rhi_rthdr, M_IP6OPT); 2324 pktopt->ip6po_rhinfo.ip6po_rhi_rthdr = NULL; 2325 if (pktopt->ip6po_route.ro_rt) { 2326 RTFREE(pktopt->ip6po_route.ro_rt); 2327 pktopt->ip6po_route.ro_rt = NULL; 2328 } 2329 } 2330 if (optname == -1 || optname == IPV6_DSTOPTS) { 2331 if (pktopt->ip6po_dest2) 2332 free(pktopt->ip6po_dest2, M_IP6OPT); 2333 pktopt->ip6po_dest2 = NULL; 2334 } 2335 } 2336 2337 #define PKTOPT_EXTHDRCPY(type) \ 2338 do { \ 2339 if (src->type) { \ 2340 int hlen = (((struct ip6_ext *)src->type)->ip6e_len + 1) << 3;\ 2341 dst->type = malloc(hlen, M_IP6OPT, canwait); \ 2342 if (dst->type == NULL && canwait == M_NOWAIT) \ 2343 goto bad; \ 2344 memcpy(dst->type, src->type, hlen); \ 2345 } \ 2346 } while (/*CONSTCOND*/ 0) 2347 2348 static int 2349 copypktopts(struct ip6_pktopts *dst, struct ip6_pktopts *src, int canwait) 2350 { 2351 dst->ip6po_hlim = src->ip6po_hlim; 2352 dst->ip6po_tclass = src->ip6po_tclass; 2353 dst->ip6po_flags = src->ip6po_flags; 2354 if (src->ip6po_pktinfo) { 2355 dst->ip6po_pktinfo = malloc(sizeof(*dst->ip6po_pktinfo), 2356 M_IP6OPT, canwait); 2357 if (dst->ip6po_pktinfo == NULL && canwait == M_NOWAIT) 2358 goto bad; 2359 *dst->ip6po_pktinfo = *src->ip6po_pktinfo; 2360 } 2361 if (src->ip6po_nexthop) { 2362 dst->ip6po_nexthop = malloc(src->ip6po_nexthop->sa_len, 2363 M_IP6OPT, canwait); 2364 if (dst->ip6po_nexthop == NULL && canwait == M_NOWAIT) 2365 goto bad; 2366 memcpy(dst->ip6po_nexthop, src->ip6po_nexthop, 2367 src->ip6po_nexthop->sa_len); 2368 } 2369 PKTOPT_EXTHDRCPY(ip6po_hbh); 2370 PKTOPT_EXTHDRCPY(ip6po_dest1); 2371 PKTOPT_EXTHDRCPY(ip6po_dest2); 2372 PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */ 2373 return (0); 2374 2375 bad: 2376 if (dst->ip6po_pktinfo) free(dst->ip6po_pktinfo, M_IP6OPT); 2377 if (dst->ip6po_nexthop) free(dst->ip6po_nexthop, M_IP6OPT); 2378 if (dst->ip6po_hbh) free(dst->ip6po_hbh, M_IP6OPT); 2379 if (dst->ip6po_dest1) free(dst->ip6po_dest1, M_IP6OPT); 2380 if (dst->ip6po_dest2) free(dst->ip6po_dest2, M_IP6OPT); 2381 if (dst->ip6po_rthdr) free(dst->ip6po_rthdr, M_IP6OPT); 2382 2383 return (ENOBUFS); 2384 } 2385 #undef PKTOPT_EXTHDRCPY 2386 2387 struct ip6_pktopts * 2388 ip6_copypktopts(struct ip6_pktopts *src, int canwait) 2389 { 2390 int error; 2391 struct ip6_pktopts *dst; 2392 2393 dst = malloc(sizeof(*dst), M_IP6OPT, canwait); 2394 if (dst == NULL && canwait == M_NOWAIT) 2395 return (NULL); 2396 ip6_initpktopts(dst); 2397 2398 if ((error = copypktopts(dst, src, canwait)) != 0) { 2399 free(dst, M_IP6OPT); 2400 return (NULL); 2401 } 2402 2403 return (dst); 2404 } 2405 2406 void 2407 ip6_freepcbopts(struct ip6_pktopts *pktopt) 2408 { 2409 if (pktopt == NULL) 2410 return; 2411 2412 ip6_clearpktopts(pktopt, -1); 2413 2414 free(pktopt, M_IP6OPT); 2415 } 2416 2417 /* 2418 * Set the IP6 multicast options in response to user setsockopt(). 2419 */ 2420 static int 2421 ip6_setmoptions(optname, im6op, m) 2422 int optname; 2423 struct ip6_moptions **im6op; 2424 struct mbuf *m; 2425 { 2426 int error = 0; 2427 u_int loop, ifindex; 2428 struct ipv6_mreq *mreq; 2429 struct ifnet *ifp; 2430 struct ip6_moptions *im6o = *im6op; 2431 struct route_in6 ro; 2432 struct in6_multi_mship *imm; 2433 struct proc *p = curproc; /* XXX */ 2434 2435 if (im6o == NULL) { 2436 /* 2437 * No multicast option buffer attached to the pcb; 2438 * allocate one and initialize to default values. 2439 */ 2440 im6o = (struct ip6_moptions *) 2441 malloc(sizeof(*im6o), M_IPMOPTS, M_WAITOK); 2442 2443 if (im6o == NULL) 2444 return (ENOBUFS); 2445 *im6op = im6o; 2446 im6o->im6o_multicast_ifp = NULL; 2447 im6o->im6o_multicast_hlim = ip6_defmcasthlim; 2448 im6o->im6o_multicast_loop = IPV6_DEFAULT_MULTICAST_LOOP; 2449 LIST_INIT(&im6o->im6o_memberships); 2450 } 2451 2452 switch (optname) { 2453 2454 case IPV6_MULTICAST_IF: 2455 /* 2456 * Select the interface for outgoing multicast packets. 2457 */ 2458 if (m == NULL || m->m_len != sizeof(u_int)) { 2459 error = EINVAL; 2460 break; 2461 } 2462 bcopy(mtod(m, u_int *), &ifindex, sizeof(ifindex)); 2463 if (ifindex != 0) { 2464 if (ifindex < 0 || if_indexlim <= ifindex || 2465 !ifindex2ifnet[ifindex]) { 2466 error = ENXIO; /* XXX EINVAL? */ 2467 break; 2468 } 2469 ifp = ifindex2ifnet[ifindex]; 2470 if ((ifp->if_flags & IFF_MULTICAST) == 0) { 2471 error = EADDRNOTAVAIL; 2472 break; 2473 } 2474 } else 2475 ifp = NULL; 2476 im6o->im6o_multicast_ifp = ifp; 2477 break; 2478 2479 case IPV6_MULTICAST_HOPS: 2480 { 2481 /* 2482 * Set the IP6 hoplimit for outgoing multicast packets. 2483 */ 2484 int optval; 2485 if (m == NULL || m->m_len != sizeof(int)) { 2486 error = EINVAL; 2487 break; 2488 } 2489 bcopy(mtod(m, u_int *), &optval, sizeof(optval)); 2490 if (optval < -1 || optval >= 256) 2491 error = EINVAL; 2492 else if (optval == -1) 2493 im6o->im6o_multicast_hlim = ip6_defmcasthlim; 2494 else 2495 im6o->im6o_multicast_hlim = optval; 2496 break; 2497 } 2498 2499 case IPV6_MULTICAST_LOOP: 2500 /* 2501 * Set the loopback flag for outgoing multicast packets. 2502 * Must be zero or one. 2503 */ 2504 if (m == NULL || m->m_len != sizeof(u_int)) { 2505 error = EINVAL; 2506 break; 2507 } 2508 bcopy(mtod(m, u_int *), &loop, sizeof(loop)); 2509 if (loop > 1) { 2510 error = EINVAL; 2511 break; 2512 } 2513 im6o->im6o_multicast_loop = loop; 2514 break; 2515 2516 case IPV6_JOIN_GROUP: 2517 /* 2518 * Add a multicast group membership. 2519 * Group must be a valid IP6 multicast address. 2520 */ 2521 if (m == NULL || m->m_len != sizeof(struct ipv6_mreq)) { 2522 error = EINVAL; 2523 break; 2524 } 2525 mreq = mtod(m, struct ipv6_mreq *); 2526 if (IN6_IS_ADDR_UNSPECIFIED(&mreq->ipv6mr_multiaddr)) { 2527 /* 2528 * We use the unspecified address to specify to accept 2529 * all multicast addresses. Only super user is allowed 2530 * to do this. 2531 */ 2532 if (kauth_authorize_generic(p->p_cred, KAUTH_GENERIC_ISSUSER, &p->p_acflag)) 2533 { 2534 error = EACCES; 2535 break; 2536 } 2537 } else if (!IN6_IS_ADDR_MULTICAST(&mreq->ipv6mr_multiaddr)) { 2538 error = EINVAL; 2539 break; 2540 } 2541 2542 /* 2543 * If no interface was explicitly specified, choose an 2544 * appropriate one according to the given multicast address. 2545 */ 2546 if (mreq->ipv6mr_interface == 0) { 2547 struct sockaddr_in6 *dst; 2548 2549 /* 2550 * Look up the routing table for the 2551 * address, and choose the outgoing interface. 2552 * XXX: is it a good approach? 2553 */ 2554 ro.ro_rt = NULL; 2555 dst = (struct sockaddr_in6 *)&ro.ro_dst; 2556 bzero(dst, sizeof(*dst)); 2557 dst->sin6_family = AF_INET6; 2558 dst->sin6_len = sizeof(*dst); 2559 dst->sin6_addr = mreq->ipv6mr_multiaddr; 2560 rtalloc((struct route *)&ro); 2561 if (ro.ro_rt == NULL) { 2562 error = EADDRNOTAVAIL; 2563 break; 2564 } 2565 ifp = ro.ro_rt->rt_ifp; 2566 rtfree(ro.ro_rt); 2567 } else { 2568 /* 2569 * If the interface is specified, validate it. 2570 */ 2571 if (mreq->ipv6mr_interface < 0 || 2572 if_indexlim <= mreq->ipv6mr_interface || 2573 !ifindex2ifnet[mreq->ipv6mr_interface]) { 2574 error = ENXIO; /* XXX EINVAL? */ 2575 break; 2576 } 2577 ifp = ifindex2ifnet[mreq->ipv6mr_interface]; 2578 } 2579 2580 /* 2581 * See if we found an interface, and confirm that it 2582 * supports multicast 2583 */ 2584 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) { 2585 error = EADDRNOTAVAIL; 2586 break; 2587 } 2588 2589 if (in6_setscope(&mreq->ipv6mr_multiaddr, ifp, NULL)) { 2590 error = EADDRNOTAVAIL; /* XXX: should not happen */ 2591 break; 2592 } 2593 2594 /* 2595 * See if the membership already exists. 2596 */ 2597 for (imm = im6o->im6o_memberships.lh_first; 2598 imm != NULL; imm = imm->i6mm_chain.le_next) 2599 if (imm->i6mm_maddr->in6m_ifp == ifp && 2600 IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr, 2601 &mreq->ipv6mr_multiaddr)) 2602 break; 2603 if (imm != NULL) { 2604 error = EADDRINUSE; 2605 break; 2606 } 2607 /* 2608 * Everything looks good; add a new record to the multicast 2609 * address list for the given interface. 2610 */ 2611 imm = in6_joingroup(ifp, &mreq->ipv6mr_multiaddr, &error, 0); 2612 if (imm == NULL) 2613 break; 2614 LIST_INSERT_HEAD(&im6o->im6o_memberships, imm, i6mm_chain); 2615 break; 2616 2617 case IPV6_LEAVE_GROUP: 2618 /* 2619 * Drop a multicast group membership. 2620 * Group must be a valid IP6 multicast address. 2621 */ 2622 if (m == NULL || m->m_len != sizeof(struct ipv6_mreq)) { 2623 error = EINVAL; 2624 break; 2625 } 2626 mreq = mtod(m, struct ipv6_mreq *); 2627 2628 /* 2629 * If an interface address was specified, get a pointer 2630 * to its ifnet structure. 2631 */ 2632 if (mreq->ipv6mr_interface != 0) { 2633 if (mreq->ipv6mr_interface < 0 || 2634 if_indexlim <= mreq->ipv6mr_interface || 2635 !ifindex2ifnet[mreq->ipv6mr_interface]) { 2636 error = ENXIO; /* XXX EINVAL? */ 2637 break; 2638 } 2639 ifp = ifindex2ifnet[mreq->ipv6mr_interface]; 2640 } else 2641 ifp = NULL; 2642 2643 /* Fill in the scope zone ID */ 2644 if (ifp) { 2645 if (in6_setscope(&mreq->ipv6mr_multiaddr, ifp, NULL)) { 2646 /* XXX: should not happen */ 2647 error = EADDRNOTAVAIL; 2648 break; 2649 } 2650 } else if (mreq->ipv6mr_interface != 0) { 2651 /* 2652 * XXX: This case would happens when the (positive) 2653 * index is in the valid range, but the corresponding 2654 * interface has been detached dynamically. The above 2655 * check probably avoids such case to happen here, but 2656 * we check it explicitly for safety. 2657 */ 2658 error = EADDRNOTAVAIL; 2659 break; 2660 } else { /* ipv6mr_interface == 0 */ 2661 struct sockaddr_in6 sa6_mc; 2662 2663 /* 2664 * The API spec says as follows: 2665 * If the interface index is specified as 0, the 2666 * system may choose a multicast group membership to 2667 * drop by matching the multicast address only. 2668 * On the other hand, we cannot disambiguate the scope 2669 * zone unless an interface is provided. Thus, we 2670 * check if there's ambiguity with the default scope 2671 * zone as the last resort. 2672 */ 2673 bzero(&sa6_mc, sizeof(sa6_mc)); 2674 sa6_mc.sin6_family = AF_INET6; 2675 sa6_mc.sin6_len = sizeof(sa6_mc); 2676 sa6_mc.sin6_addr = mreq->ipv6mr_multiaddr; 2677 error = sa6_embedscope(&sa6_mc, ip6_use_defzone); 2678 if (error != 0) 2679 break; 2680 mreq->ipv6mr_multiaddr = sa6_mc.sin6_addr; 2681 } 2682 2683 /* 2684 * Find the membership in the membership list. 2685 */ 2686 for (imm = im6o->im6o_memberships.lh_first; 2687 imm != NULL; imm = imm->i6mm_chain.le_next) { 2688 if ((ifp == NULL || imm->i6mm_maddr->in6m_ifp == ifp) && 2689 IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr, 2690 &mreq->ipv6mr_multiaddr)) 2691 break; 2692 } 2693 if (imm == NULL) { 2694 /* Unable to resolve interface */ 2695 error = EADDRNOTAVAIL; 2696 break; 2697 } 2698 /* 2699 * Give up the multicast address record to which the 2700 * membership points. 2701 */ 2702 LIST_REMOVE(imm, i6mm_chain); 2703 in6_leavegroup(imm); 2704 break; 2705 2706 default: 2707 error = EOPNOTSUPP; 2708 break; 2709 } 2710 2711 /* 2712 * If all options have default values, no need to keep the mbuf. 2713 */ 2714 if (im6o->im6o_multicast_ifp == NULL && 2715 im6o->im6o_multicast_hlim == ip6_defmcasthlim && 2716 im6o->im6o_multicast_loop == IPV6_DEFAULT_MULTICAST_LOOP && 2717 im6o->im6o_memberships.lh_first == NULL) { 2718 free(*im6op, M_IPMOPTS); 2719 *im6op = NULL; 2720 } 2721 2722 return (error); 2723 } 2724 2725 /* 2726 * Return the IP6 multicast options in response to user getsockopt(). 2727 */ 2728 static int 2729 ip6_getmoptions(optname, im6o, mp) 2730 int optname; 2731 struct ip6_moptions *im6o; 2732 struct mbuf **mp; 2733 { 2734 u_int *hlim, *loop, *ifindex; 2735 2736 *mp = m_get(M_WAIT, MT_SOOPTS); 2737 2738 switch (optname) { 2739 2740 case IPV6_MULTICAST_IF: 2741 ifindex = mtod(*mp, u_int *); 2742 (*mp)->m_len = sizeof(u_int); 2743 if (im6o == NULL || im6o->im6o_multicast_ifp == NULL) 2744 *ifindex = 0; 2745 else 2746 *ifindex = im6o->im6o_multicast_ifp->if_index; 2747 return (0); 2748 2749 case IPV6_MULTICAST_HOPS: 2750 hlim = mtod(*mp, u_int *); 2751 (*mp)->m_len = sizeof(u_int); 2752 if (im6o == NULL) 2753 *hlim = ip6_defmcasthlim; 2754 else 2755 *hlim = im6o->im6o_multicast_hlim; 2756 return (0); 2757 2758 case IPV6_MULTICAST_LOOP: 2759 loop = mtod(*mp, u_int *); 2760 (*mp)->m_len = sizeof(u_int); 2761 if (im6o == NULL) 2762 *loop = ip6_defmcasthlim; 2763 else 2764 *loop = im6o->im6o_multicast_loop; 2765 return (0); 2766 2767 default: 2768 return (EOPNOTSUPP); 2769 } 2770 } 2771 2772 /* 2773 * Discard the IP6 multicast options. 2774 */ 2775 void 2776 ip6_freemoptions(im6o) 2777 struct ip6_moptions *im6o; 2778 { 2779 struct in6_multi_mship *imm; 2780 2781 if (im6o == NULL) 2782 return; 2783 2784 while ((imm = im6o->im6o_memberships.lh_first) != NULL) { 2785 LIST_REMOVE(imm, i6mm_chain); 2786 in6_leavegroup(imm); 2787 } 2788 free(im6o, M_IPMOPTS); 2789 } 2790 2791 /* 2792 * Set IPv6 outgoing packet options based on advanced API. 2793 */ 2794 int 2795 ip6_setpktopts(control, opt, stickyopt, priv, uproto) 2796 struct mbuf *control; 2797 struct ip6_pktopts *opt, *stickyopt; 2798 int priv, uproto; 2799 { 2800 struct cmsghdr *cm = 0; 2801 2802 if (control == NULL || opt == NULL) 2803 return (EINVAL); 2804 2805 ip6_initpktopts(opt); 2806 if (stickyopt) { 2807 int error; 2808 2809 /* 2810 * If stickyopt is provided, make a local copy of the options 2811 * for this particular packet, then override them by ancillary 2812 * objects. 2813 * XXX: copypktopts() does not copy the cached route to a next 2814 * hop (if any). This is not very good in terms of efficiency, 2815 * but we can allow this since this option should be rarely 2816 * used. 2817 */ 2818 if ((error = copypktopts(opt, stickyopt, M_NOWAIT)) != 0) 2819 return (error); 2820 } 2821 2822 /* 2823 * XXX: Currently, we assume all the optional information is stored 2824 * in a single mbuf. 2825 */ 2826 if (control->m_next) 2827 return (EINVAL); 2828 2829 for (; control->m_len; control->m_data += CMSG_ALIGN(cm->cmsg_len), 2830 control->m_len -= CMSG_ALIGN(cm->cmsg_len)) { 2831 int error; 2832 2833 if (control->m_len < CMSG_LEN(0)) 2834 return (EINVAL); 2835 2836 cm = mtod(control, struct cmsghdr *); 2837 if (cm->cmsg_len == 0 || cm->cmsg_len > control->m_len) 2838 return (EINVAL); 2839 if (cm->cmsg_level != IPPROTO_IPV6) 2840 continue; 2841 2842 error = ip6_setpktopt(cm->cmsg_type, CMSG_DATA(cm), 2843 cm->cmsg_len - CMSG_LEN(0), opt, priv, 0, 1, uproto); 2844 if (error) 2845 return (error); 2846 } 2847 2848 return (0); 2849 } 2850 2851 /* 2852 * Set a particular packet option, as a sticky option or an ancillary data 2853 * item. "len" can be 0 only when it's a sticky option. 2854 * We have 4 cases of combination of "sticky" and "cmsg": 2855 * "sticky=0, cmsg=0": impossible 2856 * "sticky=0, cmsg=1": RFC2292 or RFC3542 ancillary data 2857 * "sticky=1, cmsg=0": RFC3542 socket option 2858 * "sticky=1, cmsg=1": RFC2292 socket option 2859 */ 2860 static int 2861 ip6_setpktopt(int optname, u_char *buf, int len, struct ip6_pktopts *opt, 2862 int priv, int sticky, int cmsg, int uproto) 2863 { 2864 int minmtupolicy; 2865 2866 if (!sticky && !cmsg) { 2867 #ifdef DIAGNOSTIC 2868 printf("ip6_setpktopt: impossible case\n"); 2869 #endif 2870 return (EINVAL); 2871 } 2872 2873 /* 2874 * IPV6_2292xxx is for backward compatibility to RFC2292, and should 2875 * not be specified in the context of RFC3542. Conversely, 2876 * RFC3542 types should not be specified in the context of RFC2292. 2877 */ 2878 if (!cmsg) { 2879 switch (optname) { 2880 case IPV6_2292PKTINFO: 2881 case IPV6_2292HOPLIMIT: 2882 case IPV6_2292NEXTHOP: 2883 case IPV6_2292HOPOPTS: 2884 case IPV6_2292DSTOPTS: 2885 case IPV6_2292RTHDR: 2886 case IPV6_2292PKTOPTIONS: 2887 return (ENOPROTOOPT); 2888 } 2889 } 2890 if (sticky && cmsg) { 2891 switch (optname) { 2892 case IPV6_PKTINFO: 2893 case IPV6_HOPLIMIT: 2894 case IPV6_NEXTHOP: 2895 case IPV6_HOPOPTS: 2896 case IPV6_DSTOPTS: 2897 case IPV6_RTHDRDSTOPTS: 2898 case IPV6_RTHDR: 2899 case IPV6_USE_MIN_MTU: 2900 case IPV6_DONTFRAG: 2901 case IPV6_OTCLASS: 2902 case IPV6_TCLASS: 2903 return (ENOPROTOOPT); 2904 } 2905 } 2906 2907 switch (optname) { 2908 #ifdef RFC2292 2909 case IPV6_2292PKTINFO: 2910 #endif 2911 case IPV6_PKTINFO: 2912 { 2913 struct ifnet *ifp = NULL; 2914 struct in6_pktinfo *pktinfo; 2915 2916 if (len != sizeof(struct in6_pktinfo)) 2917 return (EINVAL); 2918 2919 pktinfo = (struct in6_pktinfo *)buf; 2920 2921 /* 2922 * An application can clear any sticky IPV6_PKTINFO option by 2923 * doing a "regular" setsockopt with ipi6_addr being 2924 * in6addr_any and ipi6_ifindex being zero. 2925 * [RFC 3542, Section 6] 2926 */ 2927 if (optname == IPV6_PKTINFO && opt->ip6po_pktinfo && 2928 pktinfo->ipi6_ifindex == 0 && 2929 IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) { 2930 ip6_clearpktopts(opt, optname); 2931 break; 2932 } 2933 2934 if (uproto == IPPROTO_TCP && optname == IPV6_PKTINFO && 2935 sticky && !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) { 2936 return (EINVAL); 2937 } 2938 2939 /* validate the interface index if specified. */ 2940 if (pktinfo->ipi6_ifindex >= if_indexlim || 2941 pktinfo->ipi6_ifindex < 0) { 2942 return (ENXIO); 2943 } 2944 if (pktinfo->ipi6_ifindex) { 2945 ifp = ifindex2ifnet[pktinfo->ipi6_ifindex]; 2946 if (ifp == NULL) 2947 return (ENXIO); 2948 } 2949 2950 /* 2951 * We store the address anyway, and let in6_selectsrc() 2952 * validate the specified address. This is because ipi6_addr 2953 * may not have enough information about its scope zone, and 2954 * we may need additional information (such as outgoing 2955 * interface or the scope zone of a destination address) to 2956 * disambiguate the scope. 2957 * XXX: the delay of the validation may confuse the 2958 * application when it is used as a sticky option. 2959 */ 2960 if (opt->ip6po_pktinfo == NULL) { 2961 opt->ip6po_pktinfo = malloc(sizeof(*pktinfo), 2962 M_IP6OPT, M_NOWAIT); 2963 if (opt->ip6po_pktinfo == NULL) 2964 return (ENOBUFS); 2965 } 2966 memcpy(opt->ip6po_pktinfo, pktinfo, sizeof(*pktinfo)); 2967 break; 2968 } 2969 2970 #ifdef RFC2292 2971 case IPV6_2292HOPLIMIT: 2972 #endif 2973 case IPV6_HOPLIMIT: 2974 { 2975 int *hlimp; 2976 2977 /* 2978 * RFC 3542 deprecated the usage of sticky IPV6_HOPLIMIT 2979 * to simplify the ordering among hoplimit options. 2980 */ 2981 if (optname == IPV6_HOPLIMIT && sticky) 2982 return (ENOPROTOOPT); 2983 2984 if (len != sizeof(int)) 2985 return (EINVAL); 2986 hlimp = (int *)buf; 2987 if (*hlimp < -1 || *hlimp > 255) 2988 return (EINVAL); 2989 2990 opt->ip6po_hlim = *hlimp; 2991 break; 2992 } 2993 2994 case IPV6_OTCLASS: 2995 if (len != sizeof(u_int8_t)) 2996 return (EINVAL); 2997 2998 opt->ip6po_tclass = *(u_int8_t *)buf; 2999 break; 3000 3001 case IPV6_TCLASS: 3002 { 3003 int tclass; 3004 3005 if (len != sizeof(int)) 3006 return (EINVAL); 3007 tclass = *(int *)buf; 3008 if (tclass < -1 || tclass > 255) 3009 return (EINVAL); 3010 3011 opt->ip6po_tclass = tclass; 3012 break; 3013 } 3014 3015 #ifdef RFC2292 3016 case IPV6_2292NEXTHOP: 3017 #endif 3018 case IPV6_NEXTHOP: 3019 if (!priv) 3020 return (EPERM); 3021 3022 if (len == 0) { /* just remove the option */ 3023 ip6_clearpktopts(opt, IPV6_NEXTHOP); 3024 break; 3025 } 3026 3027 /* check if cmsg_len is large enough for sa_len */ 3028 if (len < sizeof(struct sockaddr) || len < *buf) 3029 return (EINVAL); 3030 3031 switch (((struct sockaddr *)buf)->sa_family) { 3032 case AF_INET6: 3033 { 3034 struct sockaddr_in6 *sa6 = (struct sockaddr_in6 *)buf; 3035 int error; 3036 3037 if (sa6->sin6_len != sizeof(struct sockaddr_in6)) 3038 return (EINVAL); 3039 3040 if (IN6_IS_ADDR_UNSPECIFIED(&sa6->sin6_addr) || 3041 IN6_IS_ADDR_MULTICAST(&sa6->sin6_addr)) { 3042 return (EINVAL); 3043 } 3044 if ((error = sa6_embedscope(sa6, ip6_use_defzone)) 3045 != 0) { 3046 return (error); 3047 } 3048 break; 3049 } 3050 case AF_LINK: /* eventually be supported? */ 3051 default: 3052 return (EAFNOSUPPORT); 3053 } 3054 3055 /* turn off the previous option, then set the new option. */ 3056 ip6_clearpktopts(opt, IPV6_NEXTHOP); 3057 opt->ip6po_nexthop = malloc(*buf, M_IP6OPT, M_NOWAIT); 3058 if (opt->ip6po_nexthop == NULL) 3059 return (ENOBUFS); 3060 memcpy(opt->ip6po_nexthop, buf, *buf); 3061 break; 3062 3063 #ifdef RFC2292 3064 case IPV6_2292HOPOPTS: 3065 #endif 3066 case IPV6_HOPOPTS: 3067 { 3068 struct ip6_hbh *hbh; 3069 int hbhlen; 3070 3071 /* 3072 * XXX: We don't allow a non-privileged user to set ANY HbH 3073 * options, since per-option restriction has too much 3074 * overhead. 3075 */ 3076 if (!priv) 3077 return (EPERM); 3078 3079 if (len == 0) { 3080 ip6_clearpktopts(opt, IPV6_HOPOPTS); 3081 break; /* just remove the option */ 3082 } 3083 3084 /* message length validation */ 3085 if (len < sizeof(struct ip6_hbh)) 3086 return (EINVAL); 3087 hbh = (struct ip6_hbh *)buf; 3088 hbhlen = (hbh->ip6h_len + 1) << 3; 3089 if (len != hbhlen) 3090 return (EINVAL); 3091 3092 /* turn off the previous option, then set the new option. */ 3093 ip6_clearpktopts(opt, IPV6_HOPOPTS); 3094 opt->ip6po_hbh = malloc(hbhlen, M_IP6OPT, M_NOWAIT); 3095 if (opt->ip6po_hbh == NULL) 3096 return (ENOBUFS); 3097 memcpy(opt->ip6po_hbh, hbh, hbhlen); 3098 3099 break; 3100 } 3101 3102 #ifdef RFC2292 3103 case IPV6_2292DSTOPTS: 3104 #endif 3105 case IPV6_DSTOPTS: 3106 case IPV6_RTHDRDSTOPTS: 3107 { 3108 struct ip6_dest *dest, **newdest = NULL; 3109 int destlen; 3110 3111 if (!priv) /* XXX: see the comment for IPV6_HOPOPTS */ 3112 return (EPERM); 3113 3114 if (len == 0) { 3115 ip6_clearpktopts(opt, optname); 3116 break; /* just remove the option */ 3117 } 3118 3119 /* message length validation */ 3120 if (len < sizeof(struct ip6_dest)) 3121 return (EINVAL); 3122 dest = (struct ip6_dest *)buf; 3123 destlen = (dest->ip6d_len + 1) << 3; 3124 if (len != destlen) 3125 return (EINVAL); 3126 /* 3127 * Determine the position that the destination options header 3128 * should be inserted; before or after the routing header. 3129 */ 3130 switch (optname) { 3131 case IPV6_2292DSTOPTS: 3132 /* 3133 * The old advanced API is ambiguous on this point. 3134 * Our approach is to determine the position based 3135 * according to the existence of a routing header. 3136 * Note, however, that this depends on the order of the 3137 * extension headers in the ancillary data; the 1st 3138 * part of the destination options header must appear 3139 * before the routing header in the ancillary data, 3140 * too. 3141 * RFC3542 solved the ambiguity by introducing 3142 * separate ancillary data or option types. 3143 */ 3144 if (opt->ip6po_rthdr == NULL) 3145 newdest = &opt->ip6po_dest1; 3146 else 3147 newdest = &opt->ip6po_dest2; 3148 break; 3149 case IPV6_RTHDRDSTOPTS: 3150 newdest = &opt->ip6po_dest1; 3151 break; 3152 case IPV6_DSTOPTS: 3153 newdest = &opt->ip6po_dest2; 3154 break; 3155 } 3156 3157 /* turn off the previous option, then set the new option. */ 3158 ip6_clearpktopts(opt, optname); 3159 *newdest = malloc(destlen, M_IP6OPT, M_NOWAIT); 3160 if (*newdest == NULL) 3161 return (ENOBUFS); 3162 memcpy(*newdest, dest, destlen); 3163 3164 break; 3165 } 3166 3167 #ifdef RFC2292 3168 case IPV6_2292RTHDR: 3169 #endif 3170 case IPV6_RTHDR: 3171 { 3172 struct ip6_rthdr *rth; 3173 int rthlen; 3174 3175 if (len == 0) { 3176 ip6_clearpktopts(opt, IPV6_RTHDR); 3177 break; /* just remove the option */ 3178 } 3179 3180 /* message length validation */ 3181 if (len < sizeof(struct ip6_rthdr)) 3182 return (EINVAL); 3183 rth = (struct ip6_rthdr *)buf; 3184 rthlen = (rth->ip6r_len + 1) << 3; 3185 if (len != rthlen) 3186 return (EINVAL); 3187 switch (rth->ip6r_type) { 3188 case IPV6_RTHDR_TYPE_0: 3189 if (rth->ip6r_len == 0) /* must contain one addr */ 3190 return (EINVAL); 3191 if (rth->ip6r_len % 2) /* length must be even */ 3192 return (EINVAL); 3193 if (rth->ip6r_len / 2 != rth->ip6r_segleft) 3194 return (EINVAL); 3195 break; 3196 default: 3197 return (EINVAL); /* not supported */ 3198 } 3199 /* turn off the previous option */ 3200 ip6_clearpktopts(opt, IPV6_RTHDR); 3201 opt->ip6po_rthdr = malloc(rthlen, M_IP6OPT, M_NOWAIT); 3202 if (opt->ip6po_rthdr == NULL) 3203 return (ENOBUFS); 3204 memcpy(opt->ip6po_rthdr, rth, rthlen); 3205 break; 3206 } 3207 3208 case IPV6_USE_MIN_MTU: 3209 if (len != sizeof(int)) 3210 return (EINVAL); 3211 minmtupolicy = *(int *)buf; 3212 if (minmtupolicy != IP6PO_MINMTU_MCASTONLY && 3213 minmtupolicy != IP6PO_MINMTU_DISABLE && 3214 minmtupolicy != IP6PO_MINMTU_ALL) { 3215 return (EINVAL); 3216 } 3217 opt->ip6po_minmtu = minmtupolicy; 3218 break; 3219 3220 case IPV6_DONTFRAG: 3221 if (len != sizeof(int)) 3222 return (EINVAL); 3223 3224 if (uproto == IPPROTO_TCP || *(int *)buf == 0) { 3225 /* 3226 * we ignore this option for TCP sockets. 3227 * (RFC3542 leaves this case unspecified.) 3228 */ 3229 opt->ip6po_flags &= ~IP6PO_DONTFRAG; 3230 } else 3231 opt->ip6po_flags |= IP6PO_DONTFRAG; 3232 break; 3233 3234 default: 3235 return (ENOPROTOOPT); 3236 } /* end of switch */ 3237 3238 return (0); 3239 } 3240 3241 /* 3242 * Routine called from ip6_output() to loop back a copy of an IP6 multicast 3243 * packet to the input queue of a specified interface. Note that this 3244 * calls the output routine of the loopback "driver", but with an interface 3245 * pointer that might NOT be lo0ifp -- easier than replicating that code here. 3246 */ 3247 void 3248 ip6_mloopback(ifp, m, dst) 3249 struct ifnet *ifp; 3250 struct mbuf *m; 3251 struct sockaddr_in6 *dst; 3252 { 3253 struct mbuf *copym; 3254 struct ip6_hdr *ip6; 3255 3256 copym = m_copy(m, 0, M_COPYALL); 3257 if (copym == NULL) 3258 return; 3259 3260 /* 3261 * Make sure to deep-copy IPv6 header portion in case the data 3262 * is in an mbuf cluster, so that we can safely override the IPv6 3263 * header portion later. 3264 */ 3265 if ((copym->m_flags & M_EXT) != 0 || 3266 copym->m_len < sizeof(struct ip6_hdr)) { 3267 copym = m_pullup(copym, sizeof(struct ip6_hdr)); 3268 if (copym == NULL) 3269 return; 3270 } 3271 3272 #ifdef DIAGNOSTIC 3273 if (copym->m_len < sizeof(*ip6)) { 3274 m_freem(copym); 3275 return; 3276 } 3277 #endif 3278 3279 ip6 = mtod(copym, struct ip6_hdr *); 3280 /* 3281 * clear embedded scope identifiers if necessary. 3282 * in6_clearscope will touch the addresses only when necessary. 3283 */ 3284 in6_clearscope(&ip6->ip6_src); 3285 in6_clearscope(&ip6->ip6_dst); 3286 3287 (void)looutput(ifp, copym, (struct sockaddr *)dst, NULL); 3288 } 3289 3290 /* 3291 * Chop IPv6 header off from the payload. 3292 */ 3293 static int 3294 ip6_splithdr(m, exthdrs) 3295 struct mbuf *m; 3296 struct ip6_exthdrs *exthdrs; 3297 { 3298 struct mbuf *mh; 3299 struct ip6_hdr *ip6; 3300 3301 ip6 = mtod(m, struct ip6_hdr *); 3302 if (m->m_len > sizeof(*ip6)) { 3303 MGETHDR(mh, M_DONTWAIT, MT_HEADER); 3304 if (mh == 0) { 3305 m_freem(m); 3306 return ENOBUFS; 3307 } 3308 M_MOVE_PKTHDR(mh, m); 3309 MH_ALIGN(mh, sizeof(*ip6)); 3310 m->m_len -= sizeof(*ip6); 3311 m->m_data += sizeof(*ip6); 3312 mh->m_next = m; 3313 m = mh; 3314 m->m_len = sizeof(*ip6); 3315 bcopy((caddr_t)ip6, mtod(m, caddr_t), sizeof(*ip6)); 3316 } 3317 exthdrs->ip6e_ip6 = m; 3318 return 0; 3319 } 3320 3321 /* 3322 * Compute IPv6 extension header length. 3323 */ 3324 int 3325 ip6_optlen(in6p) 3326 struct in6pcb *in6p; 3327 { 3328 int len; 3329 3330 if (!in6p->in6p_outputopts) 3331 return 0; 3332 3333 len = 0; 3334 #define elen(x) \ 3335 (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0) 3336 3337 len += elen(in6p->in6p_outputopts->ip6po_hbh); 3338 len += elen(in6p->in6p_outputopts->ip6po_dest1); 3339 len += elen(in6p->in6p_outputopts->ip6po_rthdr); 3340 len += elen(in6p->in6p_outputopts->ip6po_dest2); 3341 return len; 3342 #undef elen 3343 } 3344
Indexes created Sun Sep 21 20:09:37 GMT 2025